Algebra 1

Introduction to Commutative Algebra

Lecture Notes, Summer 2019

 Contents

Introduction 4

Chapter 1. Rings 5
1.1. Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2. The Spectrum of a Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3. Radicals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4. Local Rings and Rings of Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Chapter 2. Modules and Integral Extensions 22

2.1. Modules - Basics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2. Free and Finitely Generated Modules . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3. Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4. Localization of Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5. Integral Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.6. Going Up and Going Down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.7. Noether Normalization Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

Chapter 3. Hilbert’s Nullstellensatz and some Algebraic Geometry 47

3.1. Jacobson Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2. Hilbert’s Nullstellensatz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3. The Dimension of a Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.4. Zero Sets and Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.5. The Zariski-Topology on Ank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.6. Morphisms of Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.7. Some examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

Chapter 4. Noetherian Rings and Modules 62

4.1. Dimension Theory of Noetherian Rings . . . . . . . . . . . . . . . . . . . . . . . 66
4.2. Primary Decomposition in Noetherian Rings . . . . . . . . . . . . . . . . . . 70

Chapter 5. Regular Rings 75

5.1. Valuation Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2. Discrete Valuation Rings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.3. Dedekind Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4. The Class Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.5. Modules over PIDs and Projective Modules . . . . . . . . . . . . . . . . . . . 85

Appendix A. Prerequisites - Rings 89

A.1. Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

Appendix B. Categories 90
B.1. General Categories and Functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2
 CONTENTS 3

B.2. Additive and Abelian Categories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

B.3. Some Homological Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Appendix C. Further Remarks - Modules 93
C.1. Projective Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
C.2. Tensor Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
C.3. Localization of Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
C.4. Local-Global . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
C.5. Structure Theorems for Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Appendix. Bibliography 100
Appendix. Index 101
 Introduction

These are my lecture notes for the course Algebra 1, held by Dr. Thorsten
Heidersdorf in the summer term 20191. You can find the current version
on my website (https://pankratius.gitlab.io/notes). There is also
a version on the course hompage, but it might be outdated. If you find
mistakes (there are still a lot) or have suggestions, please send me an e-mail
to [email protected]. I want to thank everyone who already pointed
some of them out to me and apologize for the long time it took me to fix
them.
The recomended literature for this course is [AM94], [EE95] and
[MR89]. I also like to use [Alu09].
Dr. Heidersdorf introduced categories and (exact) functors in lecture 6.
I decided to put this (and a bit more) in a seperate appendix, which will be
added during the semester. For now, the reader is e.g. refered to [Ste19].

Future Aaron here: Me introducing categories and stuff in the appendix

did not happen during the semester. Overall, the appendix is a huge mess. I
hope I will be able to fix this during the summer, before the new semester
starts. I also plan to add some more stuff which I found interesting (and not
too far away from the lecture) as well as the missing proof of the last few
lectures. I also want to include the results of the exercise sheets, but still
have to figure out the right places. The last lecture (lecture 23) was a big
black box, and I am still not sure up to what detail I will be able to fix this.
And one more thing – I hope I will be able to do the same next semester
for Algebraic Geometry 1, so stay tuned... .

1Last change: 2019-07-10 10:31:42 +0200; Current commit: 05ab381

 CHAPTER 1

Rings

Convention. In this lecture rings are assumed to be

i) commutative: for all ab = ba holds for all a, b ∈ R,
ii) unital: there is an element 1 = 1R ∈ R such that 1a = a for all
a ∈ R.
Ring homomorphisms f : R → S always respect the unit, i.e. f (1R ) = 1S
holds.

1.1. Ideals
Definitionb 1.A. Let I ⊆ R be an ideal. Then I is a proper ideal or proper
if I 6= R.
Definition 1.1.
i) A proper ideal p ( R is a prime ideal if for all x, y ∈ R with xy ∈ p,
already x ∈ p or y ∈ p holds.
ii) A proper ideal m ( R is a maximal ideal if there is no ideal I with
m ( I ( R.

Notationb 1.B. I try to follow Dr. Heidersdorf’s way of naming ideals,

with one typographical addition: ordinary ideals are denoted as I, J,..., prime
ideals as p,... and maximal ideals as m,... .
Lemma 1.2. Let I ⊆ R be an ideal.
i) The following are equivalent:
a) I is a prime ideal.
b) R/I is an integral domain.

ii) The following are equivalent:

a) I is a maximal ideal.
b) R/I is a field.

Proof. We denote the coset of an element a ∈ R in R/I by a.

i) Let I be a prime ideal, and let x, y ∈ R/I such that 0 = x · y = xy.
This means that xy is in I. As I is prime, x ∈ I or y ∈ I follows,
and hence x = 0 or y = 0.
Now assume that R/I is an integral domain. Let x, y ∈ R with
xy ∈ I. Hence 0 = xy = x · y, and as R/I is an integral domain,
x = 0 or y = 0 follows. So x ∈ I or y ∈ I.
5
6 1. RINGS

ii) Let I be maximal, and x ∈/ I. Consider the ideal generated by x and

I, hI, xi. As I is maximal and I ⊆ hI, xi, we have hI, xi = R = h1i.
So there are z ∈ I, y ∈ R such that 1 = xy + z. So in R/I, we have
1 = xy + z = x · y + z = x · y,
which shows that x is a unit in R/I.
Now assume that R/I is a field and let J be an ideal with
I ⊆ J ⊆ R. If there is an x ∈ J such that x ∈
/ I, then x is invertible
in R/I. So there are z ∈ I, y ∈ R such that 1 = xy + z. As z ∈ J
and x ∈ J this implies 1 ∈ J, and hence J = R.

Corollary 1.3. Let I be an ideal. If I is maximal then I is prime.
The following is a consequence of Zorn’s Lemma and the ideal correspon-
dence:
Lemma 1.4. Let R 6= 0 be a ring.
i) R contains a maximal ideal.
ii) Every ideal of R is contained in some maximal ideal.

Corollary 1.5.
/ R× is contained in some maximal ideal of R.
i) Every x ∈
ii) The units of R are given by the complement of the union over all
maximal ideals m:
[
R× = R \ m.
m maximal ideal
iii) Let m ⊆ R be a maximal ideal in a local ring R and x ∈ m. Then
1 + x is a unit in R.

Proof.
i) Consider the ideal generated by x. As x is not a unit hxi ( h1i
holds. So by Corollary 1.5, there is a maximal ideal containing hxi,
and in particular x.
ii) Let x ∈/ m for all maximal ideals m. Then hxi ( m for all maximal
ideals m, and hence hxi = h1i.

Example 1.6. Consider the case R = Z. Then the prime ideals are h0i
and hpi, for every prime number p.
Lemmab 1.C. Let R be a ring, and consider the polynomial ring R[X1 , . . . , Xn ]
in n variables. Then for every 0 ≤ m ≤ n there is an isomorphism
R[X1 , . . . , Xn ]/hX1 , . . . , Xm i ∼
= R[Xm+1 , . . . , Xn ].
Example 1.7. Let R be an integral domain. Consider the polynomial ring
in n variables, R[X1 , . . . , Xn ]. Let m ≤ n and consider the ideal hX1 , . . . , Xm i.
Then by Lemmab 1.C, we have R[X1 , . . . , Xn ]/hX1 , . . . Xm i ∼
= R[Xm+1 , . . . , Xn ].
As R is an integral domain, R[Xm+1 , . . . , Xn ] is too. So by Lemma 1.2,
 1.2. THE SPECTRUM OF A RING 7

hX1 , . . . , Xm i is a prime ideal. However, hX1 , . . . , Xm i is not necessarily

maximal:
Consider the case m = 2, n = 2: the quotient R[X1 , X2 ]/hX1 , X2 i is
isomorphic to R. So, again by Lemma 1.2, hX1 , X2 i is maximal if and only
if R is a field.

1.2. The Spectrum of a Ring

Definition 1.8. Let R be a ring, M ⊆ R a set. We define
 
p is a prime ideal of R,
Z(M ) := p ⊂ R .
M ⊆ p.
Lemma 1.9. For every set M ⊆ R
Z (M ) = Z (hM i)
holds.
Example 1.10.
i) Z (h1i) = ∅. Z (h0i) is the set of all prime ideals.
ii) Let m be a maximal ideal. Then Z(m) = {m}. For prime ideals, the
converse is also true: Let p be a prime ideal with Z(p) = {p}.
By Lemma 1.4, there is a maximal ideal m containing p. So
{p, m} ⊆ Z (p), which implies p = m. Hence p is maximal.
iii) Consider Z, and let n ∈ Z. Then Z (hni) = {p | p prime, p divides n}.

Definition 1.11. Let X be a set, and V a system of subsets of X. We say

V defines a topology on X if the following holds:
i) arbitrary intersections of elements of V are again in V ;
ii) finite unions of elements of V are again in V ;
iii) X and ∅ are in V .

In this case, the elements of V are called closed sets.

Proposition 1.12. Let I, J ⊆ R be ideals and {Iλ }λ∈Λ a collection of
ideals of R.
i) If I ⊆ J then Z(I) ⊇ Z(J);
ii) Z(IJ)
P = Z(I)
∪TZ(J);
iii) Z λ∈Λ Iλ = λ∈Λ Z (Iλ ).

Proof.
i) Every prime ideal that contains J also contains I, hence Z(J) ⊆ Z(I).
ii) IJ is a subset of both I and J. So by i), Z(IJ) ⊇ Z(I) and
Z(IJ) ⊇ Z(J); hence Z(IJ) ⊇ Z(I) ∪ Z(J).
Let p be a prime ideal with IJ ⊆ p. Assume I * p, and let
x ∈ I be an element with x ∈ / p. For all y ∈ J the product xy is an
element of IJ ⊆ p. As p is a prime ideal, y ∈ p follows, and hence
J ⊆ p. This implies Z(IJ) ⊆ Z(I) ∪ Z(J).
8 1. RINGS

iii) Let p be a prime ideal that contains

P all of theTIλ . As ideals are Pclosed

under addition, p also contains λ∈Λ Iλ . SoP λ∈Λ Z(Iλ ) ⊆ Z λ∈Λ Iλ .
On the other hand, every
P Iλ is a subset of λ∈Λ Iλ , so every prime
ideal that contains
T λ∈Λ Iλ in particular contains each Iλ , and
P
hence Z λ∈Λ Iλ ⊆ λ∈Λ Z(Iλ ).


Corollary 1.13. The collection of the Z(I) for all ideals I of R define a
topology on the set of all prime ideals of R.
Definition 1.14. The spectrum Spec R of R is the set of all prime ideals of
R with the topology from Corollary 1.13. This topology is called the Zariski
topology.
Definitionb 1.D. The set of all maximal ideals of R is denoted by MaxSpec R.
Definition 1.15. A set in Spec R is open if it is of the form Spec R \ Z(I),
for an ideal I.
Recall the following fact about quotient rings:
Proposition 1.16. Let I ⊆ R be an ideal and ϕ : R → R0 a ring homo-
morphism such that I ⊆ ker ϕ. Then there is a unique ring homomorphism
ϕ0 : R/I → R0 such that the following diagram commutes:
ϕ
R R0

ϕ0
R/I
In the case I = ker ϕ, ϕ0 is a ring isomorphism
∼
R/ ker ϕ im ϕ.

Lemma 1.17. Let ϕ : R → R0 be a ring homomorphism and p ⊂ R0 a

prime ideal. Then the preimage ϕ−1 (p) ⊂ R is again a prime ideal.
Proof. Consider the composition
ϕ
ϕ̃ : R R0 R0 /p.
Then ker ϕ̃ = ϕ−1 (p). By Proposition 1.16, there exists a unique, injective
ring homomorphism ϕ0 : R/ϕ−1 (p) → R0 /p such that the diagram
ϕ
R R0 R0 /p

R/ϕ−1 (p) ϕ0

commutes.
This identifies R/ϕ−1 (p) with a subring of R0 /p. By applying Lemma 1.2
twice, we get that ϕ−1 (p) is indeed a prime ideal. 
Remarkb 1.E. This statement is in general not true for maximal ideals:
Consider the embedding Z ,→ Q. Then the preimage of the maximal ideal
h0i ⊆ Q is h0i, but h0i is not maximal in Z.
 1.2. THE SPECTRUM OF A RING 9

Proposition 1.18. Every ring homomorphism ϕ : R → R0 induces a

continous map
ϕ# : Spec R0 −→ Spec R
p 7−→ ϕ−1 (p) .
Proof. By Lemma 1.17, ϕ# is well-defined.
Let I ⊆ R be an ideal. Then
 −1 n o
ϕ# (Z(I)) = p ∈ Spec R0 ϕ# (p) ∈ Z(I)
= p ∈ Spec R0 I ⊆ ϕ−1 (p)


= p ∈ Spec R0 ϕ (I) ⊆ p


= Z (ϕ(I)) ,

so ϕ# is continous, as preimages of closed sets are closed. 

Notationb 1.F. The map ϕ# is also denoted as Spec ϕ.

Corollary 1.19. The assignment of the spectrum to a ring can be in-
tepreted as a functor
Spec : CRingop −→ Top.
In particular: Isomorphic rings have homeomorphic spectra.
Remark 1.20.
i) Let R be an integral domain. This is equivalent to h0i being a prime
ideal. Now let p be any prime ideal. Then h0i is in every open
subset containing p. So Spec R is not Hausdorff.
ii) For any prime ideal p ∈ Spec R,
\
{p} = Z(I) = Z(p).
I⊆p
I ideal

So p is maximal if and only if {p} is closed in Spec R (by Exam-

ple 1.10).
iii) Let R be an integral domain. Then h0i ∈ Spec R is a point with

{h0i} = Z (h0i) = Spec R.

Remarkb 1.G. Let I ⊆ R be an ideal. Then the projection π : R  R/I

induces a homeomorphism
π # : Z(I) −→
∼
Spec R/I
p 7−→ π(p).
Lemma 1.21. Spec R is quasi-compact: for every covering
[
Spec R = Uλ
λ∈Λ
10 1. RINGS

where each of the Uλ is an open subset of Spec R and Λ an arbitrary index

set, there are finitely many Uλ1 , . . . , Uλn such that
n
[
Spec R = Uλi .
i=1
Proof. This will be on the first exercise sheet. 

End of Lecture 1

1.3. Radicals
We now want to find an equivalent characterisation of Z(I) = Z(J) for
two ideals I, J of R.
Definition 1.22. Let I ⊆ R be an ideal. The radical of I is
√
I := {x ∈ R | there exists an n > 0 such that xn ∈ I} .
Definitionb 1.H. An element x ∈ R is called nilpotent if there is an n > 0
such that xn = 0.
Exampleb 1.I.
i) The zero element is always nilpotent.
ii) In an integral domain, there are no non-zero nilpotent elements.

Lemma √ 1.23.
i) I is an ideal
p√ of R.
√
ii) I√⊆ I = I.
iii) I√= R if and only if I = R.
iv) R/ I has no non-zero nilpotent elements.

Proof. √
i) Let x, y ∈ I and m, n > 0 such that xm , y n ∈ I. Then
m+n−1
X m + n − 1
m+n−1
(x + y) = xi y m+n−1−i .
i
i=0

Now by assumption ∈ I for i ≥ m and y m+n−1−i

xi √ ∈ I for i < m.
So the whole sum is in I, and hence x + y ∈ I. Furthermore, we
have for any r ∈ R
(rx)m = rm xm
which is in I. √ √ p√
ii) By setting n = 1, we get p I ⊆ I, and hence I ⊆ I. For the
√ n
reverse inclusion, let x ∈ I and n > 0 such that x ∈ I. Then
there is a m > 0 such that xnm = (xn )m ∈ I. √
iii) As 1n = 1 for all n > 0, 1 ∈ I if and only if 1 ∈ I. p
√ √ √ √
iv) Let z ∈ R/ I with z n = 0. Then z n ∈ I, so z ∈ I = I,
which is equivalent to z = 0.

 1.3. RADICALS 11

Definition 1.24. √
i) An ideal p
I is a radical ideal if I = I holds.
ii) Nil R := h0i is the nilradical .
iii) If R has no non-zero nilpotent elements then R is reduced .

Exampleb 1.J. Let I ⊆ R be an ideal and π : R → I the canonical

projection. Then the nilradical of R/I is given by
Nil R/I = {x ∈ R/I | xn = 0 for a n > 0}
= {π(x) | x ∈ R, xn ∈ I for a n > 0}
√ 
=π I .

Lemma 1.25. An ideal I is a radical ideal if and only if R/I is reduced.

Proof. We have the following equivalences:
I is a radical ideal ⇐⇒ for all x ∈ R, n > 0 with xn it holds that x ∈ I
⇐⇒ in R/I : xn = 0 implies x = 0
⇐⇒ R/I is reduced.


Definition 1.26. We call Rred := R/ Nil R the reduced ring associated to

R.
p
Example 1.27. Let R = Z and I = hai for an 0 6= a ∈ Z. How does hai
look like? Consider the decomposition into prime factors
ml
a = pm
1 · . . . · pl .
1

p
Then hai p= hp1 . . . pl i:
If x ∈ hai, then there is a n > 0 such that xn ∈ hai. So a divides xn , and
hence p1 , . . . pl divide x, which implies x ∈ hp1 . . . pl i. Let now x ∈ hp1 . . . pl i.
n n n m1 ml
Choose n ≥ max {m p1 , . . . , ml }. Then x ∈ hp1 . . . pl i ⊆ hp1 . . . pl i = hai
which implies x ∈ hai.
Proposition 1.28. For any ideal I
√ \
I= p
p prime
I⊆p

holds.
√
Proof. Let x ∈ I and n > 0 such that xn ∈ I. Let p be a prime ideal
with I ⊆ p. Then xn ∈ p, and as p is a prime ideal, this already implies
x ∈ p.
For the converse, assume √that x is in the intersection of all prime ideals
that contain I and that x ∈/ I. We now want to use Zorn’s Lemma in a
non-obvious way to arrive at a contradiction. For that, define
 
J is an ideal of R,
Σ := J ⊂R .
for all n > 0: xn ∈/ J.
12 1. RINGS
√
First note that I ∈ Σ, as x ∈ / I; so Σ 6= ∅. Furthermore, Σ is partially
ordered by inclusion. Let (Jt )t∈T be a non-empty chain in Σ and consider
[
J˜ := Jt .
t∈T

Then J˜ contains I, is an ideal1 and does not contain xn for any n > 0. So J˜
is an upper bound of (Jt )t∈T . By Zorn’s Lemma, this implies that Σ contains
a maximal element p̃. We now show that p̃ is a prime ideal:
Let a, b ∈ R \ p̃. Then hai + p̃, hbi + p̃ strictly contain p̃ and hence cannot
be in Σ. By definition of Σ, there are now m, n > 0 such that xm ∈ hai + p̃
and y n ∈ hbi + p̃. So there are c, d ∈ R and r, s ∈ p̃ such that xm = ac + r
and xn = bd + s. Now
xm+n (ac + r) · (bd + s)
= abcd
|{z} + rbd + sac + rs,
| {z }
∈habi ∈p̃

so xn+m ∈ habi + p̃. If ab would be an element of p̃, then habi ⊆ p̃, which
would imply xn+m ∈ p̃. Therefore ab cannot be an element of p̃, which is
equivalent to p̃ being prime.
T
But by the original assumption, x ∈ p p, and hence x ∈ p̃. This is a
I⊆p
contradiction as p̃ is an element of Σ. 
Corollary 1.29.
i) The nilradical of R is given the intersection of all prime ideals of R:
\
Nil R = p.
p prime

ii)b The canonical projection π : R → R/ Nil R induces a homeomor-

phism
∼
π # : Spec R Spec (R/ Nil R) .

Proof. The spectrum of the quotient by Nil R is given by

Spec R/ Nil R = {prime ideals of R/ Nil R}
= {π (p) | p prime, Nil R ⊆ p}
= {π (p) | p prime} .

Exampleb 1.K. Consider R := Z/aZ. Then the nilradical is given by
\
Nil R = hpi = hp1 . . . pm i,
p prime
p|n

1Note that in general, unions of ideals are not ideals.

 1.3. RADICALS 13

where a = pm ml b
1 · . . . · pl . Note that this recovers the results of Example 1.J
1

and Example 1.27.

In particular, this shows that the nilradical of a ring is not necessarily a
prime ideal.
√ 
Corollary 1.30. For any ideal I of R, Z(I) = Z I holds.
√ √ 
Proof. As I ⊆ I, Z I ⊆ Z(I) follows (c.f. Proposition 1.12). On
the other hand, as
√ \
I= p,
p prime
I⊆p
√ √ 
every prime ideal that contains I also contains I. So Z(I) = Z I . 

Corollary 1.31. For all ideals I, J√the following

√ holds:
i) Z(J) ⊆ Z(I) if and only if √J ⊇ √ I.
ii) Z(I) = Z(J) if and only if I = J.

√ √
Proof. By symmetry, √  to prove i). If I ⊆ J, then by
√  it suffices
Proposition 1.12 Z I ⊇ Z J holds. Corollary 1.30 now implies
Z(I) ⊇ Z(J).
For the other direction, assume Z(J) ⊆ Z(I). Then every prime ideal
that contains J also contains I. Now
√ \
I= p
p prime
I⊆p
\ √
⊆ p= J.
p prime
J⊆p

We now introduce a notion similar to the nilradical, but for maximal

ideals:
Definition 1.32. The Jacobson radical Jac R is defined as the intersection
of all maximal ideals of R:
\
Jac R := m.
m maximal
Proposition 1.33. The Jacobson radical of R is given by
Jac R = x ∈ R 1 − ax ∈ R× for all a ∈ R .


Proof.
⊆ “: Let x ∈ Jac R and a ∈ R. Assume 1 − ax is not a unit in R.
”
Then by Corollary 1.5 there is a maximal ideal m containing 1 − ax.
But then
1 = (1 − ax) + ax
14 1. RINGS

As both summands are in m, 1 ∈ m would follow, which is a contra-

diction.

⊇ “: Let x ∈ R be an element such that 1 − ax is a unit for every

”
a ∈ R, and let m be a maximal ideal that does not contain x. Then
hxi + m = h1i, so there is a a ∈ R and y ∈ m such that 1 = ax + y.
But then y = 1 − ax would be a unit, which is not possible (as
y ∈ m).


1.4. Local Rings and Rings of Fractions

Definition 1.34. A ring R is local if it contains exactly one maximal ideal.
Exampleb 1.L.
i) Every field is a local ring, with maximal ideal h0i.
ii) Let p ⊆ R be a prime ideal, S := R \ p and the localization
Rp := S −1 R. Then Rp is a local ring:
Consider the ideal
na o
I := a ∈ p, s ∈ S .
s
Let b/t ∈
/ I. Hence b ∈ / p, which implies b ∈ S. So b/t is a unit in
Rp . This shows that every ideal J ⊆ Rp with J ( I contains a unit.

Lemma 1.35. Let I ( R be an ideal. Then the following are equivalent:

i) R is a local ring with maximal ideal I;
ii) R \ I ⊆ R× ;
iii) R \ I = R× .

Proof. This follows from Corollary 1.5. 

End of Lecture 2
Lemma 1.36. Let m ⊂ R be a maximal ideal. If 1 + x is a unit in R for
every x ∈ m, then R is a local ring with maximal ideal m.
Proof. Let b ∈ R \ m. As m is maximal, hbi + m = R. So there are
a ∈ R, x ∈ m such that ab + x = 1. Hence ab = 1 − x ∈ R× , by assumption.
But then hbi = 1, and hence b is a unit in R. The claim follows now from
Lemma 1.35. 
Lemmab 1.M. Let R be a local ring, with maximal ideal m. Then 1 + x
is a unit for every x ∈ m.
Proof. This follows direclty from Proposition 1.33. 
Lemmab 1.N. Let ϕ : R → k be a surjective ring homomorphism, where
k is a field. Then ker ϕ is a maximal ideal of R.
Proof. By Proposition 1.16, there is a unique isomorphism R/ ker ϕ ∼
= k.
The claim now follows from Lemma 1.2. 
 1.4. LOCAL RINGS AND RINGS OF FRACTIONS 15

Example 1.37.
i) For every prime number p and every n ≥ 1, R := Z/pn Z is a local
ring: The prime ideals in R are in one-to-one, order preserving
correspondence with the prime ideals in Z that contain hpn i. But
the only prime ideal that contains hpn i is hpi. So R has only one
prime ideal, given by hpi, which has to be maximal.
Note that for all n > 1, Z/pn Z is a finite local ring, which is
not an integral domain, so in particular not a field.
ii) Let R be a local ring with maximal ideal m, and consider the ring
of formal power series
(∞ )
X
i
R[[t]] := ai t ai ∈ R
i=0

(for more on this ring, see A.1.1).

There is a well-defined evaluation map

ev : R[[t]] −→ R/m
X∞
ai ti 7−→ a0
i=0

with kernel ker (ev) = m + hti. As ev is surjective, Lemmab 1.N

implies that ker (ev) is a maximal ideal of R[[t]].
It is also the only maximal ideal: Let f ∈ ker ev, so
∞
X
f= ai ti
i=0

with a0 ∈ m. Consider now 1 + f . We construct an inverse g for

1 + f , so that the claim follows from Lemma 1.36: Let g ∈ R[[t]] be
a polynomial of the form
∞
X
g= bi ti .
i=0

The condition (1 + f )g = 1 is equivalent to requiring (1 + a0 )b0 = 1

and

(1 + a0 ) bn + a1 bn−1 + . . . + an b0 = 0 for all n > 0.

By Lemmab 1.M 1 + a0 is a unit in R, so such a b0 exists. Assuming

that b0 , . . . bn−1 are already constructed, the equation

(1 + a0 ) bn + a1 bn−1 + . . . + an b0 = 0

can be re-written as

bn := − (1 + a0 )−1 · (an b0 + . . . + a1 bn−1 ) ,

which is well-defined, as (1 + a0 ) is a unit. By induction, we obtain

an inverse g for 1 + f .
16 1. RINGS

iii) Consider Rn with the standard topology, and let X ⊆ Rn be an

open subset with 0 ∈ X. We define an equivalence relation on the
set of tuples
 
U ⊆ X an open subset with 0 ∈ U ,
(U, f ) ,
f : U → R continuous.
by setting (U1 , f ) ∼ (U2 , g) if there is an open subset W ⊆ X such
that 0 ∈ W , W ⊆ U1 ∩ U2 and f |W = g|W . The equivalence class
of (U, f ) is denoted by [U, f ] and is called a germ at 0:

U1 U2
W

Consider now the set of all germs at 0

 
U ⊆ X an open subset with 0 ∈ U ,
F0 := [U, f ] ,
f : U → R continuous.
which is called the stalk at 0. We can define a ring structure on F0
by setting
[U1 , f1 ] + [U2 , f2 ] := [U1 ∩ U2 , f1 + f2 ]
and
[U1 , f1 ] · [U2 , f2 ] := [U1 ∩ U2 , f1 · f2 ],
which is inherited from the pointwise ring structure on functions
to R. Note that this is well-defined: If [U1 , f1 ] = [U10 , f10 ] such that
f1 |W1 = f10 |W1 for an open subset W1 ⊆ U1 ∩U10 and [U2 , f2 ] = [U20 , f20 ]
such that f2 |W2 = f20 |W2 , then f1 + g1 = f2 + g2 and f1 · g1 = f2 · g2
on W1 ∩ W2 ⊆ U1 ∩ U2 .
The stalk F0 is a local ring: Consider the the ring homomor-
phism
ϕ : F0 −→ R
[U, f ] 7−→ f (0).
This is well-defined, as all functions in the germ [U, f ] agree on 0.
As ϕ ([X, f = c]) = c for all c ∈ R, we see that ϕ is surjective and
hence ker ϕ is a maximal ideal (Lemmab 1.N).
Let now [U, f ] ∈ ker ϕ. As f is continuous, there is an open
neighbourhood W of 0 such that 1 + f (x) 6= 0 for all x ∈ W . Hence
[W, 1/(1 + f )] is the unit element of F0 . The claim now follows from
Lemma 1.36.

Remarkb 1.O. Germs and stalks can be defined in the more general context
of (pre-)sheaves. The example we considered is for the sheaf of continous
functions Rn → R. For more, the reader is refered to [Vak18, Chapter 2].
Remarkb 1.P. Here are some more facts on R[[t]] for a general ring R:
 1.4. LOCAL RINGS AND RINGS OF FRACTIONS 17

i) As R-modules the map

Y ∞
X
R[[t]] → R, ai ti 7→ (ai )i∈N
N i=0

is a well-defined isomorphism.
Q It is a classical (non-trivial) result
that the infinite product N Z is not a free.
ii) The units in R[[t]] are of the form a0 + . . ., where a0 is a unit in R.
This similar to the case where R is local.
iii) So we can still describe the maximal ideals MaxSpec R[[t]]: by ii),
every maximal ideal necessarily contains t and hence corresponds
to a maximal ideal of R[[t]]/hti ∼= R. So we have
MaxSpec R[[t]] = {m + hti | m ∈ MaxSpec R} .
Since for any ideal I ⊆ R the maximal ideals over I + hti are
precisley of the form m + hti for the maximal ideals m over I ⊆ m
the assignment
MaxSpec R[[t]] ←→ MaxSpec R
m 7−→ m ∩ R
m + hti ←−[ m
is a homeomorphism (where we equip MaxSpec R[[t]] and MaxSpec R
with the subspace topology).

The following is a recollection of basic facts about rings of fractions.

More details can be found in [Sch19, 5.17].
Definition 1.38. A subset S ⊆ R is called multiplicative if
i) 1 ∈ S;
ii) for all a, b ∈ S it holds that ab ∈ S.

Remark/Definition 1.39. Let S ⊆ R be a multiplicative set. We can

define an equivalence relation on S × R by
(s, a) ∼ (t, b) if there is a u ∈ S such that u (ta − sb) = 0.
Denote by S −1 R the set of equivalence classes of ∼, and by a/s or a
s the
equivalence class [(s, a)].
We can define a ring structure on S −1 R by setting
a b at + bs
+ :=
s t st
and
a b ab
· := .
s t st
To show that this is well-defined requires S to be multiplicative and some
work, so we will not do this here.
The ring S −1 R is called the ring of fractions with respect to S. Mostly,
S will be left implicit, so that we refer to S −1 R only as the ring of fractions.
Lemma 1.40.
18 1. RINGS

i) The map
η : R −→ S −1 R
a
a 7−→
1
is a ring homomorphism. The elements of S are invertible in S −1 R:
×
η (S) ⊆ S −1 R .

ii) If R has no zero-divisors, then S −1 R has no zero-divisors too.

iii) If S has no zero-divisors, then η is injective.

Proof. Ommited. 
Proposition 1.41. Let S ⊆ R be a multiplicative subset and g : R → R0
a ring homomorphism such that g (S) ⊆ (R0 )× . Then g factors over the ring
of fractions: there is a unique ring homomorphism g 0 : S −1 R → R0 such that
g
R R0
η
g0
S −1 R
commutes.
Proof. Ommited. 
Example 1.42.
i) Let S = {1}. Then S −1 R ∼
= R.
ii) Let 0 6= a ∈ R be an element and
S := {an | n > 0} .
Then Ra := S −1 R is called the localization at a.
iii) Let p be a prime ideal and S := R \ p. Then S is a multiplicative
subset. Then Rp := S −1 R is the localization at p.

Example 1.43. Consider the case R = Z. Then for any prime number p,
the localization at p is given by
 
a
Zp = a ∈ Z, n ≥ 0 .
pn
The localization at a prime ideal p = hpi however is given by
na o
Zp = a, b ∈ Z, p does not divide b .
b
Note that Zp and Zp are different from each other; both are extremly different
from Z/pZ!
Lemmab 1.Q. For an element x ∈ R, the following are equivalent:
i) x = 0,
ii) η(x) = 0 for all prime ideals p ∈ Spec R,
iii) η(x) = 0 for all maximal ideals m ∈ MaxSpec R.

Proof. Ommited. 
 1.4. LOCAL RINGS AND RINGS OF FRACTIONS 19

Definition 1.44. Let ϕ : R → R0 be a ring homomorphism.

i) Let J ⊆ R0 be an ideal. We denote by
J ∩ R := ϕ−1 (J) ⊆ R
the contraction of J by ϕ.
ii) Let I ⊆ R be an ideal. We denote by
IR0 := hϕ(I)i ⊆ R0
the extension of I by ϕ.
In both cases, the map ϕ is often left implicit.
Lemma 1.45. Let ϕ : R → R0 be a ring homomorphism.
i) Extensions and contractions of ideals by ϕ are again ideals.
ii) For all ideals I ⊆ R and J ⊆ R0 , I ⊆ (IR0 ) ∩ R = I and
J ⊇ (J ∩ R) R0 = J holds.

Theorem 1.46. Let S ⊆ R be a multiplicative subset.

i) If I ⊆ R is an ideal, then

na o
I S −1 R = a ∈ I, s ∈ S .
s
ii) If I ⊆ R is an ideal, then
I S −1 R ∩ R = {a ∈ R | there is a n ∈ S such that na ∈ I} .

iii) If J ⊆ S −1 R is an ideal, then

(J ∩ R) S −1 R = J.
iv) If p is a prime ideal in R with p ∩ S = ∅, then p S −1 R is a prime

ideal in S −1 R.
v) The maps
Spec S −1 R ←→ {p ∈ Spec R | p ∩ S = ∅}
q 7−→ q ∩ R
−1


p S R ←−[ p
are mutually inverse and preseve inclusions.

Remarkb 1.R. The statement in v) is actually stronger: If we consider

the set
{p ∈ Spec R | p ∩ S = ∅}
with the subspace topology in Spec R, then the map q 7→ q ∩ R is actually a
homeomorphism.
Proof. Ommited. 

Remarkb 1.S. In Theorem 1.46, all statements about contractions and

extensions of ideals are with respect to the canonical inclusion η : R → S −1 R.
Proof of Theorem 1.46.
20 1. RINGS

i) Let a/s ∈ S −1 R with a ∈ I and s ∈ S. Then

a a 1
= ·
s 1 s
1
= η(a) · ,
s
−1


which is in I S R , as η(a)

is.
−1
Let now x ∈ I S R , so there are ai ∈ I, bi ∈ R and si ∈ S
such that
X bi ai
x= · .
si 1
i
Set  
Y X Y
s := si and a := bi  sj aj  .
i i i6=j
Then s ∈ S, as S is a multiplicative set, and a ∈ I, as I is an ideal.
As x = a/s, the claim follows.
ii) Let a ∈ IS −1 R ∩ R. Then a/1 ∈ I S −1
R . We now have the

following chain of equivalences a ∈ I S −1 R if and only if

a b
there are b ∈ I, t ∈ S, such that=
1 t
⇐⇒ there are b ∈ I and s, t ∈ S with s(ta − b) = 0
⇐⇒ there is a n ∈ S such that na ∈ I,
where the first equivalence follows from i), the second from the defi-
nition of the localization and the third from the following argument:
If there are such b, t and s, then (st)a = sb. But n := st is in S
(as s and t are) and sb is in I, as b is. On the other hand, if there
is a n ∈ S such that b := na ∈ I, then for t := n and s := 1 we have
1 · (na − na) = 0.
iii) (J ∩ R) S −1 R ⊆ J is always true (Lemma 1.45). For the other
inclusion, let x = a/s ∈ J. Then
a s a
= ·
1 1 s
and hence a ∈ J ∩ R. So a/s ∈ (J ∩ R) S −1 R, by i).
iv) Let p ∈ Spec R be a prime ideal
with p∩S = ∅, and let a/s, b/t ∈ S −1 R\p S −1 R .

If (ab)/(st) ∈ S −1 R \ p S −1 R . Then there are c ∈ p, u ∈ S such

that (ab)/(st) = c/u (by i). By the definition of the localization,
there is now a v ∈ S such that
v (uab − stc) = 0.
So (vu) ab = (vst)c ∈ p, as p is an ideal. But since ab ∈
/ p this
implies vu ∈ p ∩ S, contradicting p ∩ S = ∅.

Corollary 1.47. Let p ⊆ R be a prime ideal.
i) There is a bijection
Spec Rp {p0 ∈ Spec R | p0 ∩ (R \ p) = ∅} = {p0 ∈ Spec R | p0 ⊆ p} .
 1.4. LOCAL RINGS AND RINGS OF FRACTIONS 21

ii) Rp is a local ring with maximal ideal pRp .

Corollary 1.48. The map

Spec η : Spec S −1 R Spec R
is injective, with image
{p ∈ Spec R | p ∩ S = ∅} .
Definition 1.49. In the special case S = {an | n ≥ 0} the image of Spec η
is D(a) := Spec R \ Z(hai), which is called a principal open subset.
Remarkb 1.T. The principal open subsets form a basis for the Zariski
topology: for every open subset U ⊆ Spec R and a point x ∈ U there is a
principal open subset D(a) such that x ∈ D(a) ⊆ U .

End of Lecture 3
 CHAPTER 2

Modules and Integral Extensions

2.1. Modules - Basics

Definition 2.1. An R-module M (M, +, ·) is an abelian group (M, +)
together with a map
· : R × M −→ M
a, x 7−→ ax
such that
i) (a + b)x = ax + bx,
ii) a(x + y) = ax + ay,
iii) a(bx) = (ab)x,
iv) 1R x = x
for all x, y ∈ M and a, b ∈ R.
Example 2.2.
i) Let k be a field. Then k-modules are precisely k-vector spaces.
ii) Let I ⊆ R be an ideal. Then I can be regarded as an R-module,
since it is closed under addition and multiplication by elements in
R.
iii) Consider R = Z. Let G be an abelian group. Then G is a Z-module,
by setting
nx := x + . . . + x
| {z }
n times
and (−1)x := −x.
iv) Let ϕ : R → R0 be a ring homomorphism and let M be an R0 -module.
Then M can be regarded as an R-module, by setting
ay := ϕ(a)y
for all a ∈ R. This is called restriction of scalars.

Definition 2.3. Let M, M 0 be R-modules and f : M → M 0 a map. We

say f is an R-linear map if
i) f (x + x0 ) = f (x) + f (x0 ),
ii) f (ax) = af (x)
for all a ∈ R and x, x0 ∈ M .
Remark 2.4.
i) Composition of R-linear maps are again R-linear: If f : M → N and
g : N → O are R-linear maps, then their composition g ◦ f : M → O
is R-linear too.
22
 2.1. MODULES - BASICS 23

ii) For all R-modules M , the identity map id : M → M, x 7→ x is

R-linear.
iii) Let f : M → N be a bijective R-linear map. Then the inverse
f −1 : N → M is R-linear too. In this case, we say f is an isomor-
phism of R-modules.
So we can construct a category R-Mod, where objects are R-modules and
morphisms are R-linear maps.
iv) For two R-modules M, N , the set of R-linear maps
homR (M, N ) := f : M → M 0 f is R-linear


is an R-module, by setting f +g : x 7→ f (x)+g(x) and af : x 7→ f (ax)

for all f, g ∈ homR (M, N ) and a ∈ R. In the notation, the ring R is
sometimes ommited and we just write hom(M, N ) for homR (M, N ).
So R-Mod is an abelian and a pre-R-linear category.
v) For an R-module M , the set of R-linear maps EndR (M ) := homR (M, M )
has also non-commutative ring structure, by setting f g : x 7→ (f ◦g)(x)
for all f, g ∈ EndR (M ). We call EndR (M ) the set of R-linear endo-
morphism, and f an (R-linear) endomorphism.

Remarkb 2.A. Using the restriction of scalars from Example 2.2, we

obtain a functor Fϕ : R0 -Mod → R-Mod for all rings R, R0 and ring
homomorphisms ϕ : R → R0 . This functor is faithful : for all R0 -modules
M, N , the induced map homR0 (M, N ) → homR (M, N ) is injective.
Example 2.5. Let M be a R-module. The map
M −→ homR (R, M )
x 7−→ [a 7→ ax]
is an isomorphism of R-modules, with inverse
homR (R, M ) −→ M
f 7−→ f (1).
Example 2.6. Let R be a ring. Then an R[t]-module is the same“ as an
”
R-module M , together with an endomorphism f : M → M .
If M is an R-module, we can define the an R[t]-structure on M by setting
tm := f (m) for all m ∈ M and extending linearly:
n n
!
X X
ai ti (m) := ai f (i) (m).
i=0 i=0
If, on the other hand, M is an R[X]-module, we can regard M as an
R-module, by restriction of scalars for the embedding R ,→ R[t]. We also
get an endomorphism f : M → M , defined by m 7→ tm.
Definition 2.7. Let M be an R-module. A subset M 0 ⊆ M is an R-
submodule if
i) M 0 is a subgroup of (M, +) and
ii) for all a ∈ R and x ∈ M 0 , it holds that ax ∈ M 0 .
If the ring R is clear, we will often refer to M 0 just as a submodule.
24 2. MODULES AND INTEGRAL EXTENSIONS

Exampleb 2.B.
i) Let k be a field. Then k-submodules of k-modules are precisely
k-subspaces.
ii) Let I ⊆ R be an ideal. Then I is an R-submodule of R.
iii) Let H ⊆ G be a subgroup of an abelian group G. Then H is a
Z-submodule.

Proposition 2.8. Let M 0 ⊆ M be a submodule. Then the quotient group

M/M 0 becomes an R-module, by setting

· : R × R/M 0 −→ M/M 0
r, x + M 0 −→ (rx) + M 0 .

The quotient map M → M/M 0 is R-linear.

Proof. Ommited. 

Definition 2.9. Let f : M → N be an R-linear map.

i) The kernel of f is defined as

ker f := {x ∈ M | f (x) = 0} .

ii) The image of f is defined as

im f := {f (x) | x ∈ M } .

Proposition 2.10. Let f : M → N be an R-linear map. The kernel of f

is a submodule of M , the image of f is a submodule of N .
Definition 2.11. Let f : M → N be an R-linear map. The cokernel of f
is defined as coker f := N/ im f .
Lemmab 2.C. Let f : M → N be an R-linear map.
i) The kernel of f is trivial if and only if f is injective.
ii) The cokernel of f is trivial if and only if f is surjective.

Proposition 2.12. Let f : M → N be an R-linear map.

i) Let M 0 ⊆ M be a submodule such that M 0 ⊆ ker f . Then there
is a unique R-linear map f : M/M 0 → N such that the following
diagram commutes:

f
M N

f
M/M 0
 2.2. FREE AND FINITELY GENERATED MODULES 25

∼
ii) There is a unique isomorphism f˜ : M/ ker f −
→ im f such that the
following diagram commutes:
f
M N

∼
M/ ker f im f
f˜

Proof. Ommited. 

End of Lecture 4

2.2. Free and Finitely Generated Modules

Lemma 2.13. Let MT be an R-module, (Mi ) a family of submodules of M .
Then the intersection i Mi and the sum
( )
X X mi ∈ Mi ,
Mi := mi
mi 6= 0 for only finitely many i.
i i
are submodules of M .
Proof. Ommited. 
Definition 2.14.
i) Let M1 , M2 ⊆ M be submodules of an R-module M . If M1 ∩M2 = {0}
and M1 + M2 = M , we write M1 ⊕int M2 = M . This construction
is called the (internal) direct sum of M1 and M2 .
ii) Let (Mi ) be a family of R-modules. ThenQ the product of the {Mi }
is defined as the cartesian product i Mi , with component-wise
addition and scalar multiplication.
iii) Let (Mi ) be a family of R-modules. Then the direct sum (or
coproduct) {Mi } is defined as the submodule
 Q  Y
M (mi ) ∈ i Mi ,
Mi := (mi ) ⊆ Mi .
mi 6= 0 for only finitely many i.
i i

Remarkb 2.D.
i) If we regard two submodules M1 , M2 ⊆ M of an R-module M with
M1 ⊕int M2 = M as R-modules, then M1 ⊕ M2 ∼ = M = M1 ⊕int M2 .
So we will not distinguish further between the two notions.
ii) If (Mi ) is a finite family of R-modules, then the product and direct
sum of the Mi are equal.
iii) The product of a family (Mi ) is indeed a product in the category
R-Mod. The coproduct of a family (Mi ) is indeed a coprodcut in
the category R-Mod.

Definition 2.15. Let {xi } be a family of elements in an R-module M .

26 2. MODULES AND INTEGRAL EXTENSIONS

i) We say {xi } is a generating system of M , if every x ∈ M can be

written as a finite linear combination of xi .
ii) We define the subspace Lin {xi } generated by {xi } as

X 
ai ∈ I,
Lin {xi } := ai xi .
ai 6= 0 for only finitely many i.

iii) We say {xi } is linearly independent if for all tuples (x1 , . . . , xn ) of

elements from {xi } there are no ai ∈ R such that

a1 x1 + . . . + an xn = 0.

iv) We say {xi } is a basis if it is a linearly independent generating

system.
v) M is a free R-module if there is an index set I such that M ∼
L
= i∈I R.
vi) M is aLfinite-free R-module if there is a finite index set I such that
M∼ = i∈I R.
vii) M is a finitely generated R-module if M has a finite generating
system.

Definitionb 2.E. Let M be an R-module. We say M is finitely presented

or that M is an R-module of finite presentation if there are integers n, m
and an exact sequence of the form

Rm Rn M 0
Exampleb 2.F.
i) If M is finitely presented then M is already finitely generated.
ii) The converse is in general not true: The sequence Rm → Rn → M
being exact is equivalent to saying that Rm → Rn is a kernel for
the projection Rn → M . Consider now any non-noetherian ring
and I ⊆ R an ideal which is not finitely generated. Then R/I is not
finitely presented. As a concrete example, there is no presentation
for k[t1 , . . .]/ht1 , . . .i.

Lemmab 2.G. Let M be an R-module.

i) M is free if and only if M has a basis.
ii) M is finite-free if and only if it is finitely generated and free.

Remarkb 2.H. The alternative characterizations of a basis, as known from

linear algebra for vector spaces, does not hold for general modules: Consider
R = Z and M = Z. Then Z has a basis given by {1}. However, the subset
{2, 3} is also linear independet, and the subset {2} does not generate all of
Z.
 2.2. FREE AND FINITELY GENERATED MODULES 27

Remark 2.16. Let M be a free R-module with basis {xi }i∈I . Let M 0
be another R-module and {yi }i∈I a subset of M 0 . Then there is a unique
R-linear map f : M → M 0 such that the diagram
∃!f
M M0

f
{xi }i∈I

commutes. Here, the map of sets f : {xi }i∈I is given by f (xi ) := yi for all
i ∈ I. This is called the universal property of a free module.
If M 0 is a finitely generated R-module, then in particular there is a
surjective map Rn → M 0 .
Example 2.17.
i) For R = Z and M = Z/pZ, M is not a free Z-module, as every
element has torsion.
ii) For R = Z and M = Q, M is not finitely generated as Z-module.

Definition 2.18. Let M be an R-module and I ⊆ R an ideal. We define

the submodule IM as
IM := Lin {am | a ∈ I, m ∈ M } .
Lemmab 2.I. Let M be an R-module and I ⊆ R an ideal. Then R/IM
has the structure of an R/I-module.
Lemma 2.19. Let M be an R-module such that
Rn ∼
=M ∼ = Rm
for n, m ∈ N. Then m = n follows.
Proof. Let m ⊆ R be a maximal ideal, so R/m is a field. Then
dimR/mR M/mM = dimR/mR Rn /mRn
= dimR/mR (R/mR)n
= n,
so n is uniquely determined. 
Remark 2.20 (Linear Algebra for Modules). For n, m ∈ N, we can identify
homR (Rn , Rm ) with the set Mat(n × m, R) of n × m-matrices with entries
in R. For an endomorphism f : Rn → Rn , we have that f is an isomorphism
if and only if its associated matrix is invertible.
Using the Leibniz formula, we define the determinant of a n × n matrix
A = (ai,j ) ∈ Mat (n × n, R) as
X
det A := sgn(σ)a1,σ(1) . . . an,σ(n) .
σ∈Sn

Using the determinant, we associate to every matrix A = (ai,j ) ∈ Mat(n×n, R)

the adjugate matrix adju M , which is defined as
 
(adju A)i,j := (−1)i+j det Aj,i
28 2. MODULES AND INTEGRAL EXTENSIONS

where the matrix Aj,i ∈ Mat ((n − 1) × (n − 1), R) is obtained from A by

removing the j-th row and the i-th column. It then holds that
A adju A = (adju A) A = det A · En
where En denotes the n × n unit matrix. As the determinant stays multi-
plicative for matrices with entries in an arbitrary ring, we have that a matrix
with entries in R is invertible if and only if its determinant is a unit in R.
Proposition 2.21 (Cayley-Hamilton for Modules). Let M be a finitely
generated R-module, I ⊆ R an ideal and f : M → M an endomorphism, such
that f (M ) ⊆ IM . Then there is a monic polynomial p = ni=0 ai ti ∈ R[t]
P
such that ai ∈ I for i 6= n and p(f ) = 0 holds.
Proof. Let {m1 , . . . , mn } be a generating set for M . Then f (mi ) ∈ IM
for all 1 ≤ i ≤ n, and so there are aij ∈ I for 1 ≤ i, j ≤ n such that
n
X
(∗) f (mj ) = aij mi .
i=1
By Example 2.6, we can regard M as an R[t]-module, with the action
tx := f (x) for all x ∈ M . The condition (∗) then reads as
Xn
(tδij − aij ) mj = 0 for all 1 ≤ j ≤ n.
i=0
 
Consider now the matrices A := (tδij − aij )i,j and B := adju A. Then
BA = det A · En , and hence
 
det (tδij − aij )i,j (mj ) = 0.
 
The claim follows for the polynomial p := det (tδij − aij )i,j . (That p is
monic and the non-leading coefficients are in I follows from the Leibniz-
formula.) 
Lemma 2.22. Let M be a finitely generated R-module and f : M → M a
surjective R-linear map. Then f is already an isomorphism.
Proof. As in Example 2.6, we consider M as an R[t1 ] module. Consider
the ideal I := ht1 i ⊆ R[t1 ]. As f is surjective, IM = M follows. By
Proposition 2.21 (for the endomorphism id : M → M ), there is a polynomial
p = tn2 + an−1 tn−1
2 + . . . + a0 ∈ R[t1 ][t2 ] such that p(id) = 0. So there is a
polynomial q ∈ R[t1 ] with
id = t1 q(t1 ).
Evaluating at f gives the invese. 
We will now state and prove two version of Nakayama’s Lemma:
Lemma 2.23 (Nakayama - the general one“). Let M be a finitely gener-
”
ated R-module and I ⊆ R an ideal with IM = M .
i) There is an a ∈ I such that am = m for all m ∈ M .
ii) There is a x ∈ R such that 1 − x ∈ I and xM = 0.

Proof.
 2.2. FREE AND FINITELY GENERATED MODULES 29

i) Apply Proposition 2.21 to id : M → M . Then there are a0 , . . . , an−1 ∈ I

such that
id = (−(an−1 + . . . + a0 )) id .
As I is an ideal, −(an−1 + . . . + a0 ) ∈ I.
ii) For x := 1 + an−1 + . . . + a0 , where the a0 , . . . , an−1 are choosen as
in the proof of i), the claim follows.

Lemma 2.24 (Nakayama - the classical one“). Let M be a finitely gen-
”
erated R-module, and I ⊆ Jac R an ideal. Then IM = M if and only if
M = 0.
Proof. If M = 0, then IM = 0. If IM = M , then by Lemma 2.23
there is an x ∈ R such that 1 − x ∈ Jac R. Now by Proposition 1.33,
x = 1 − (1 − x) ∈ R× . But this implies 1m = 0 for all m ∈ M . So
M = 0. 
Lemmab 2.J. Let M be a finitely generated R-module, and M 0 a submod-
ule. Then M/M 0 is again finitely generated.
Proof. By Remark 2.16, there is a surjective map Rn → M . Extend
this to a surjective map Rn → M → M/M 0 . 
Remarkb 2.K. It is in general not true that submodules of finitely gener-
ated modules are again finitely generated: Let
I1 ( I2 ( . . . ( R
be a strictly increasing chain of ideals in a ring R. Then the ideal
∞
[
I := Ii
i=1
is not finitely generated as an R-module.
Proof. I is an ideal, since it is the union over a chain of ideals. Further-
more, I 6= R, since there is no ideal Ii with 1 ∈ Ii . If there were f1 , . . . , fn
such that I = hf1 , . . . , fn i then there would be a m such that f1 , . . . , fn ∈ Im .
But then I ⊆ Im ( Im+1 ( I, which is not possible. So I is not a finitely
generated R-module. 
So consider now the polynomial ring in infintely many variables over a
ring R 6= 0, R0 := R[t1 , t2 , . . .]. Then the chain of ideals
ht1 i ( ht1 , t2 i ( . . . ( R0
is strictly increasing, so the submodule ht1 , t2 , . . .i is not finitely generated
as an R-module.
This observation leads to the following definition: A ring R is noetherian
if it satisfies one of the following equivalent conditions:
i) All ideals I ⊆ R are finitely generated as R-modules.
ii) There is no strictly increasing chain of ideals.
iii) Every non-empty set M of ideals of R has an inclusion-maximal
element.
30 2. MODULES AND INTEGRAL EXTENSIONS

We have already encountered noetherian rings: A field is noetherian, since

there are no proper ideals other than h0i. We showed (without using the
name) last semester that PIDs are noetherian (in the proof that PIDs are
factorial, [Sch19, Satz 5.30]).
In the same vein, we call an R-module M noetherian if it satisfies one of
the following equivalent conditions:
i) Every submodule N ⊆ M is finitely generated.
ii) There is no strictly increasing chain of submodules.
iii) Every non-empty set M of submodules of M has an inclusion-
maximal element.
It can be shown that an R-module M is noetherian if and only if it is finitely
generated and R/ AnnR (M ) is a noetherian ring (a proof can be found in
[Fra17, 1.3, Prop. 5].)
Corollary 2.25. Let M be a finitely generated R-module, N ⊆ M a
submodule and I ⊆ Jac R such that

M = IM + N.

Then M = N holds.
Proof. We have

I(M/N ) = (IM + N )/N

= M/N.

Then by Lemmab 2.J and Lemma 2.24, M/N = 0 follows, which implies
M = N. 

End of Lecture 5
Lemma 2.26. Let M be a finitely-generated R-module, m1 , . . . , mn ∈ M
and I ⊆ Jac R an ideal. Then the following are equivalent:
i) The set {mi }i generates M as an R-module.
ii) The set {mi }i generates M/IM as an R/I-module.

Proof. If the set {mi }i generates M/IM as an R/I-module, then it

also generates R/I as an R-module, since the canonical map R → R/I is
surjective. So
hm1 , . . . , mn i + IM = M

and by Corollary 2.25, hm1 , . . . , mn i = M follows. 

2.3. Algebras
Definition 2.27. Let R be a ring.
i) A R-algebra consists of a tuple (R0 , ϕ), where R0 is a ring and
ϕ : R → R0 is a ring homomorphism.
 2.4. LOCALIZATION OF MODULES 31

ii) Let (R0 , ϕ0 ), (R00 , φ00 ) be R-algebras. A ring homomorphism f : R0 → R00

is an R-algebra homomorphism if the following diagram commutes:
f
R0 R00

ϕ0 ϕ00
R
iii) LetR0 be an R-algebra. We say R0 is finitely generated as R-algebra
if there is a n ≥ 0 and b1 , . . . , bn ∈ R0 such that the evaluation map
evb1 ,...,bn : R[t1 , . . . , tn ] −→ R0
ti 7−→ bi
is surjective. In this case, we will also sometimes say that R0 is an
R-algebra of finite type.

Remarkb 2.L. The morphism ϕ in the definition of an R-algebra is often

left implicit.
Warning 2.28. Being finitely generated as R-module implies being finitely
generated as an R-algebra. The converse is in general not true: For example,
the polynomial ring R[t1 , . . . , tn ] is not a finitely generated R-module.

2.4. Localization of Modules

Remark/Definition 2.29. Let M be an R-module and S ⊆ R be a
multiplicative set. We can define an equivalence relation on S × M by
(s, x) ∼ (t, y) if there is a u ∈ S such that u (tx − sy) = 0.
Denote by S −1 M the set of equivalence classes of ∼, and by x/s or xs the
equivalence class [(s, x)].
We can define a S −1 R-module structure on S −1 M by setting
x y tx + sy
+ :=
s t st
and
a y ay
· := .
s t st
To show that this is well-defined requires S to be multiplicative and is
analogous to the localization of rings.
The module S −1 M is called the localization of M by S.
Further remkarks on the localization of modules can be found in C.3.
Proposition 2.30. Localization is functorial: Given a ring R and a multi-
plicative subset S, we can define a functor
S −1 (−) : R-Mod −→ S −1 R-Mod
M 7−→ S −1 M
which sends an R-linear map f : M → N to the induced S −1 R-linear map
S −1 f : S −1 M −→ S −1 N
x f (x)
7−→ .
s s
32 2. MODULES AND INTEGRAL EXTENSIONS

Proof. We need to show that S −1 f is always a well-defined S −1 R-linear

map. We only show well-definedness: Let x/s = y/t in S −1 M . So there
exists a u ∈ S such that u(tx − sy) = 0. As f is R-linear, this implies
u (tf (x) − sf (y)) = 0 and hence f (x)/s = f (y)/t. 
Definition 2.31.
i) A sequence of the form
... Mi−1 Mi Mi+1 ... ,
where the Mi are R-modules and the fi are R-linear maps is called
exact at i if im fi = ker fi+1 . We say this sequence is exact if it is
exact at every i.
ii) An exact sequence of the form
0 M0 M N 0
is a short-exact sequence.

Lemma 2.32.
f
i) The sequence 0 M0 M is exact if and only if f is injective.
g
ii) The sequence M N 0 is exact if and only if g is surjective.
iii) The sequence 0 M0 h M 0 is exact if and only if h is an
isomorphism.
iv) Let M 0 ⊆ M be a submodule. Then the sequence
0 M0 M M/M 0 0
is short-exact.

Lemmab 2.M. Let

0 M0 M M 00 0
be a short-exact sequence of R-modules. Then: M is finitely generated if
and only if both M 0 and M 00 are.
Proof. This is on Exercise Sheet 3. 
Lemma 2.33. Let S ⊆ R be a multiplicative set. Then the localization-
functor F : R-Mod → S −1 R-Mod is exact.
Proof. F is additive: Let f, f 0 : M → N be two R-linear maps. Then
(f + f 0 ) (x)
F f + f 0 (x/s) =


s
f (x) + f 0 (x)
=
s
f (x) f 0 (x)
= +
s s
= F (f ) (x/s) + F f 0 (x/s) .

Let now y/t ∈ ker F (g), so g(y)/t = 0. Now by the definition of the
localization there is a u ∈ S such that ug(y) = 0, and hence uy ∈ ker g. As
 2.5. INTEGRAL EXTENSIONS 33

the original sequence is short exact, there is a x ∈ M such that f (x) = uy.
Then F (f ) (x/(ut)) = (uy)/(ut) = y/t, so y/t is in the image of F (f ). 
Corollary 2.34. Let
N ⊆ M
be a submodule of an R-module M . Then
S −1 (M/N ) ∼
= S −1 M / S −1 N .
Proof. Consider the short-exact sequence
0 N M M/N 0 .
Since localization is an exact functor, the sequence

0 S −1 N S −1 M S −1 (M/N ) 0

is exact too. Hence S −1 M/S −1 N ∼

= S −1 (M/N ). 

End of Lecture 6

2.5. Integral Extensions

Convention. In this and the remaining sections of this chapter, we will
consider R-algebras R0 , induced by ring homomorphisms ϕ : R → R0 . In
the notation, ϕ will be ommited, i.e. for a ∈ R and b ∈ R0 , ab := ϕ(a)b.
→. In this
If ϕ is injective, we will show this pictorial by a hooked arrow ,−
case we further identify R with a subset of R0 . So for an element a ∈ R and
→ R0 , a ∈ R0 means ϕ(a) ∈ R0 .
ϕ : R ,−
Definition 2.35. Let R0 be an R-algebra.
i) An element a ∈ R0 is integral over R if there is a monic polynomial
p ∈ R[t] such that p(a) = 0, i.e. there are cn−1 , . . . , c0 ∈ R such
that
an + cn−1 an−1 + . . . + c0 = 0.
ii) The ring R0 is integral over R if all a ∈ R0 are integral over R.
iii) The set
R := a ∈ R0 a integral over R


is the integral closure of R in R0 .

iv) The ring R is integraly closed in R0 if R = im (ϕ : R → R0 ).
v) The ring R0 is finite over R if R0 is finitely generated as R-module.

Example 2.36. Consider the inclusion Z ,− → Q: Let a ∈ Q be integral

over Z. Then there is a monic polynomial p ∈ Z[t] such that p(a) = 0, i.e.
there are cn−1 , . . . , c0 ∈ Z such that
0 = an + cn−1 an−1 + . . . + c0 .
Let now a = r/s where r, s ∈ Z are coprime. Then
rn = −s cn−1 rn−1 + . . . + c0 sn

and hence s divides r. So s ∈× Z = {±1}, which implies a ∈ Z and thus

Z = Z.
This argument still holds if we replace Z by a general factorial ring R
and Q by Quot R.
34 2. MODULES AND INTEGRAL EXTENSIONS

Warning 2.37. Unlike in the special case of fields, the condition that a
polynomial p is monic is necessary, as it is in general not possible to invert
the leading coefficient. For example, a = 1/2 ∈ Q is a root of 2t − 1 ∈ Z[t],
but still not integral over Z.
Remark/Definition 2.38. Let R0 be an R-algebra and b1 , . . . , bn ∈ R0 .
We denote by
 
 X 
R[b1 , . . . , bn ] := ai1 ,...,in bi1 . . . bin ai1 ,...,in ∈ R ⊆ R0
 
i1 ,...,in

the smallest R-subalgebra of R0 which contains all of the b1 , . . . , bn .

Lemma 2.39. Let R0 be an R-algebra, b ∈ R0 . Then the following are
equivalent:
i) b is integral over R.
ii) The R-subalgebra R[b] is a finitely generated R-module.
iii) There is a R-subalgebra R̃ ⊆ R0 such that R[b] ⊆ R̃ ⊆ R0 , R̃ is a
finitely generated R-module and b ∈ R̃.

Proof. i) =⇒ ii): Let p = tn + cn−1 t

n−1 + . . . + c ∈ R[t] be a
0
polynomial such that p(b) = 0. Then the set 1, b, . . . , bn−1 generates R[b]
as R-module.
ii) =⇒ iii): We can simply choose R̃ := R[b].
iii) =⇒ i): If we regard R̃ as an R-module, left-multiplication with b
(i.e. the map fb : R0 → R0 , m 7→ b · m) is an R-linear endomorphism of R0 .
As R̃ is finitely generated as an R-module, Cayley-Hamilton (Propo-
sition 2.21) implies that there is a monic polynomial p ∈ R[t] such that
p(fb ) = 0. So in particular we have 0 = p(fb )(1) = p(b). 

Lemma 2.40. Let ϕ : R → R0 and ϕ0 : R0 → R00 be ring homomorphisms,

and consider R0 as an R-algebra and R00 as an R0 -algebra.
i) If R0 is finite over R and R00 is finite over R0 , then R00 is finite over
R.
ii) If R0 is integral over R and R00 is integral over R0 , then R00 is integral
over R.

Proof. Ommited. 

Corollary 2.41. Let R0 be an R-algebra. Then the following are equivalent:

i) R0 is finite over R.
ii) There are elements b1 , . . . , bn ∈ R0 which are integral over R such
that R0 = R[b1 , . . . , bn ].
iii) R0 is an R-algebra of finite type and integral over R.

Corollary 2.42. Let R0 be an R-algebra. Then the integral closure R ⊆ R0

is an R-subalgebra.
 2.6. GOING UP AND GOING DOWN 35

Remarkb 2.N. Let ϕ : R → R0 be a ring homomorphism and let S ⊆ R

be a multiplicative set. Then the localization of R0 by S as an R-module
and the localization of R0 by f (S) as a ring are isomorphic as R-algebras.
Lemmab 2.O. Let R0 be an R-algebra via the ring homomorphism ϕ : R → R0
and S ⊆ R a multiplicative set. Then the localization S −1 R0 is still a ring
and the induced map of S −1 R-modules
S −1 (ϕ) : S −1 R −→ S −1 R0
a ϕ(a)
7−→
s s
is also a ring homomorphism.
Lemma 2.43. Let R0 be an R-algebra and R0 integral over R.
i) Let I ⊆ R0 be an ideal and J := R ∩ I. Then R0 /I is integral over
R/J.
ii) Let S ⊆ R be a multiplicative set. We can regard the localized
S −1 R-module S −1 R0 as an S −1 R module via the induced map from
Lemmab 2.O. Then S −1 R0 is integral over S −1 R.

Proof. Ommited. 

2.6. Going Up and Going Down

Definitionb 2.P. Let R0 be an R-algebra via the ring homomorphism
ϕ : R → R0 . If ϕ is injective and R0 integral over R, we say that R0 is an
integral extension of R.
Lemma 2.44. Let R and R0 be integral domains and R0 an integral exten-
sion of R. Then R0 is a field if and only if R is a field.
Proof. Assume that R0 is a field and let a ∈ R \ {0}. We want to show
that the inverse b of a, which exists in R0 , is an element
P ofi R. As b is integral
over R, there are c0 , . . . , cn−1 ∈ R such that bn = ci b . Since b = an−1 bn ,
we have
n−1
X
b= ci an−1−i
i=0

so b ∈ R. Note that we did not need that R or R0 is an integral domain for

this direction.
Let now a ∈ R0 \ {0}. As a is integral over R, R[a] is a finite dimensional
R-vector space (R is assumed to be a field). Consider now the map
fa : R[a] −→ R[a]
m 7−→ am.
This is R-linear and injective, as R0 is an integral domain, and hence bijective.
So there is a b ∈ R0 such that ab = 1. 

Lemma 2.45. Let R0 be an R-algebra which is integral over R and q ⊆ R0

a prime ideal. Set p := q ∩ R. Then R/p → R0 /q is an integral extension.
36 2. MODULES AND INTEGRAL EXTENSIONS

Proof. By Lemma 2.43 we have that R0 /q is integral over R/p. Let

ϕ : R → R0 be the ring homomorphism that induces the R-algebra structure
on R0 . Then for the composition
ϕ
ϕ:R R0 R0 /q
we have p = R ∩ q = ker ϕ. So we get a factorisation of the form
ϕ
R R0 R0 /q

R/p ϕ0

and ϕ0 is injective. So R0 /q is an integral extension of R/p. 

Remarkb 2.Q. In the lecture, the claim of Lemma 2.45 was only made for
→ R0 . But this is not necessary, since
injective ring homomorphisms ϕ : R ,−
0
the induced map ϕ is injective, even if ϕ was not (this is also the version
stated in [Fra18b, Prop. 6.8]).
Corollary 2.46. Let R0 be an R-algebra which is interal over R and q ⊆ R0
a prime ideal. Set p := q ∩ R. Then q is maximal if and only if p is maximal.
Proof. By Lemma 2.45, R/p ,− → R0 /q is an integral extension. The
claim now follwos from Lemma 2.44 and Lemma 1.2 
Lemma 2.47 (3am-Lemma). Let R0 be an integral extension of R and
q1 ⊆ q2 ∈ Spec R0 prime ideals with q1 ∩ R = q2 ∩ R. Then already q1 = q2
holds.
Proof. Let p := q1 ∩ R = q2 ∩ R and consider Rp0 as the localized
R-module or equivalently the localization by ϕ (p). (We have Rp0 6= 0 as ϕ is
injective.) We then have the following commutative diagram
ϕ, integral
R R0
η η0

Rp Rp0
ϕ0 , integral

with maps
a
η 0 : R0 → Rp0 , a 7→
1
and
ϕ0 : Rp −→ Rp0
a ϕ(a) ϕ(a)
7−→ =
s ϕ(s) s
where we identify s ∈ R and ϕ(s) ∈ R0 .
We now have that q0i := qi Rp0 is a prime ideal in Rp0 for i = 1, 2, as
qi ∩ (ϕ (R \ p)) = ∅.
By the commutativity of the above diagram we get
q0i ∩ Rp ∩ R = q0i ∩ R0 ∩ R

= qi Rp0 ∩ R0 ∩ R.



 2.6. GOING UP AND GOING DOWN 37

As p is prime in R, we can use Theorem 1.46, v) to get qi Rp0 ∩ R0 = qi and

hence
q0i ∩ Rp ∩ R = qi ∩ R

= p.

So, by Theorem 1.46, iii), we have that

q0i ∩ Rp ∩ R Rp



pRp =
= q0i ∩ Rp .

Now by Corollary 1.47, ii) we have that Rp is a local ring with maximal ideal
pRp , so by Corollary 2.46 we have that both q01 and q2 are maximal in Rp0 .
But since we assumed q1 ⊆ q2 this implies q01 = q02 .
Using Theorem 1.46 one last time, we get
q1 = q1 · Rp0 ∩ R0

= q01 ∩ R0
= q02 ∩ R0
= q2 ,
which finishes the proof. 

Remarkb 2.R. The nickname “3am-Lemma“ comes from a characteri-

zation of its proof Dr. Heidersdorf gave in the lecture. A more suitable
description is that for the induced map ϕ# : Spec R0 → Spec R there is no
proper inclusion in thefibres.
Remarkb 2.S. The 3am-lemma should also be true if ϕ : R → R0 is not
assumed to be injective, c.f. [Sta19, 00GT]. One way of seeing this should
be the following :
Let q1 ⊆ q2 ⊆ R0 be the prime ideals in question, and assume only that
R is integral over R. By Lemma 2.45, we get that ϕ : R/p → R0 /q1 is an
0

integral extension. Denote by π 0 : R0 → R0 /q1 the canonical projection. We

now can apply our version of the 3am-lemma (Lemma 2.47) to ϕ to get that
π 0 (q1 ) = π 0 (q2 ). So by Remarkb 1.G, we get q1 = q2 .

End of Lecture 7
Lemma 2.48 (Lying Over). Let ϕ : R ,− → R0 be an integral extension of
R. Then for all prime ideals p ∈ Spec R there is a prime ideal q ∈ Spec R0
such that q ∩ R = p, i.e. the induced morphism Spec ϕ : Spec R0 → Spec R
is surjective.
Proof. Let p ∈ Spec R be a prime ideal. Consider the commutative
diagram
ϕ, integral
R R0
η η0
ϕ0 , integral
Rp Rp0
38 2. MODULES AND INTEGRAL EXTENSIONS

We have that ϕ0 is injective, as localization is exact (Lemma 2.33). So in

particular, Rp0 6= 0 and there is a maximal ideal n ⊆ Rp0 . Since ϕ0 is an
integral extension n ∩ Rp is again maximal (Corollary 2.46).
By Corollary 1.47, Rp is a local ring with maximal ideal pRp and thus
n ∩ Rp = pRp . Set q := n ∩ R0 ⊆ R0 . Then:
q ∩ R = n ∩ R0 ∩ R

= (n ∩ Rp ) ∩ R
= (pRp ) ∩ R
= p.

Definition 2.49. Let ϕ : R → R0 be an algebra.
i) We say that ϕ satisfies going up if given a chain of prime ideals
p1 ⊆ p2 ⊆ . . . ⊆ pn
in R and a chain of prime ideals
q1 ⊆ q2 ⊆ . . . qm
in R0 with m ≤ n and pi = qi ∩ R for all 1 ≤ i ≤ m there are prime
ideals qm+1 , . . . , qn such that the following holds:
• pi = qi ∩ R for all m + 1 ≤ i ≤ n; and
• the ideals qm+1 , . . . , qn fit into the chain
q1 ⊆ . . . ⊆ qm+1 ⊆ . . . ⊆ qn .
ii) We say that ϕ satisfies going down if given a chain of prime ideals
p1 ⊇ p2 ⊇ . . . ⊇ pn
in R and a chain of prime ideals
q1 ⊇ q2 ⊇ . . . qm
with m ≤ n and pi = qi ∩ R for all 1 ≤ i ≤ m there are prime ideals
qm+1 , . . . , qn such that the following holds:
• pi = qi ∩ R for all m + 1 ≤ i ≤ n; and
• the ideals qm+1 , . . . , qn fit into the chain
q1 ⊇ . . . ⊇ qm+1 ⊇ . . . ⊇ qn .

Lemma 2.50. Let ϕ : R → R0 be a ring homomorphism.

i) The following are equivalent:
a) ϕ satisfies going down.
b) For all prime ideals q ∈ Spec R0 and p := q ∩ R, the induced
map Spec Rq0 → Spec Rp is surjective.

ii) The following are equivalent:

a) ϕ satisfies going up.
b) For all prime ideals q ∈ Spec R0 and p := q ∩ R, the induced
map Spec(R/p) → Spec(R0 /q) is surjective.
 2.6. GOING UP AND GOING DOWN 39

c) The induced map ϕ# : Spec R0 → Spec R is closed: images of

closed sets in Spec R0 under ϕ# are closed in Spec R.

Proof. This will be on the 5th exercise sheet. 

Theorem 2.51 (Going Up for Integral Extensions). → R0 be an
Let ϕ : R ,−
integral extension. Then ϕ satisfies going up.
Proof. Let p1 , p2 ∈ Spec R and q1 ∈ Spec R0 be prime ideals such that
p1 ⊆ p2 and q1 ∩ R = p1 . We now have the following commutative diagram
ϕ, integral
R R0
π π0

R/p1 R0 /q1
ϕ0 , integral

where ϕ0 is an integral extension by Lemma 2.43 and Lemma 2.45. The ideal
p2 /p1 is prime in R/p1 (Remarkb 1.G) and hence Lying Over (Lemma 2.48)
implies that there is a prime ideal q02 ∈ Spec R0 /q1 such that q02 ∩R/p1 = p2 /p1 .
Consider now the prime ideal q2 := q02 ∩ R0 . Then q1 ⊆ q2 and
q2 ∩ R = q02 ∩ R0 ∩ R

= (p2 /p1 ) ∩ R
= p2 .

Remarkb 2.T. There are also extension that are not integral but still
satisfy going up: A trivial example is the embedding Q ,− → R or more
generaly any field extension that is not algebraic.
As another example, consider the embedding Z ,− → Z[t]. This is not
integral (because by Corollary 2.41 this would imply that Z[t] is a finite
Z-module). However, for every prime ideal q ∈ Z the image pZ[t] is prime
too.
Going Down for Integral over Normal.
Definition 2.52. Let ϕ : R → R0 be an algebra and I ⊆ R an ideal.
i) An element b ∈ R0 is integral over I if there is a monic polynomial
p ∈ R[t] such that p(b) = 0 and the non-leading coefficients of p are
in I.
ii) The set
I := b ∈ R0 b is integral over I .


is the integral closure of I in R.

Lemmab 2.U. Let ϕ : R → R0 be an algebra and I ⊆ R an ideal. Then

an element b ∈ R0 is integral over I if and only if there is a n > 0 such that
bn is integral over I.
Lemma 2.53. Let ϕ : R →√ R0 be an algebra and I ⊆ R an ideal. Consider
the ideal IR ⊆ R. Then I = IR ⊆ R.
40 2. MODULES AND INTEGRAL EXTENSIONS

Proof. If b ∈ I ⊆ R then there are a0 , . . . , an−1 ∈ I such that

bn = an−1 bn−1 + . . . + a0 ,
√
so bn ∈ IR and hence b ∈ IR.
Let now b ∈ IR (this suffices, by Lemmab 2.U), so there are ai , . . . , an ∈ I
and c1 , . . . , cn ∈ R such that b = a1 c1 + . . . + an cn . Since the ci are integral
over R, the module M := R[c1 , . . . , cn ] is finitely generated (by Lemma 2.39).
Consider now the R-linear map
fb : M −→ M
n
x 7−→ xb .
Then im fb ⊆ IM , and hence by Cayley-Hamilton (Proposition 2.21) there
is a monic polynomial p ∈ R[t] with non-leading coefficients in I such that
p(fb ) = 0. So in particular, p(bn ) = 0, hence b is integral over I (again by
Lemmab 2.U). 

Definition 2.54. Let R be an integral domain. We say R is normal if R

is integrally closed in its quotient field.
Example 2.55. Every factorial ring is normal.
Lemmab 2.V. Being normal is a local property: For an integral domain,
the following are equivalent:
i) R is normal.
ii) Rp is normal for all p ∈ Spec R.
iii) Rm is normal for all m ∈ MaxSpec R.

Lemmab 2.W. Let ϕ : R ,− → R0 be an injective map, with R, R0 integral

domains.
i) This induces a field extension Quot R ,−→ Quot R0 .
0
→ R is integral, then Quot R ,−
ii) If R ,− → Quot R0 is too,
→ R0 is finite, then Quot R ,−
iii) If R ,− → Quot R0 is too.

Proof.
i) Since ϕ is injective, every non-zero element in R gets mapped to a
unit in Quot R0 . So by the univesal property of the localization, we
get an induced map
ϕ
R R0

Quot R Quot R0

ii) Localizing R, R0 (as modules) at S := R \ {0} gives an intgegral ex-

tension (Lemma 2.43) Quot R = S −1 R ,−→ S −1 R0 . By Lemma 2.44,
we get that S R is a field, and hence S −1 R0 = Quot R0 . So
−1 0

→ Quot R0 is algebraic.
Quot R ,−
iii) Localizing once again at S := R \ {0} gives a finite (localizing is
exact) extension Quot R = S −1 R ,− → S −1 R0 . Arguing as in ii),
 2.6. GOING UP AND GOING DOWN 41

Quot R0 = S −1 R0 follows, so S −1 R0 is a finite Quot R-vector space.


Lemma 2.56. Let ϕ : R ,− → R0
be an integral extension, with R, R0
integral
0
domains and R normal. Let b ∈ R be integral over some ideal I ⊆ R. Then
b/1 is algebraic over Quot √
R and the non-leading coefficients of its minimal
polynomial are already in I.
Proof. By Lemmab 2.W, there is indeed a minimal polynomial for b.
Set K := Quot R, and consider the intermediate field K ⊆ K[b] ⊆ Quot R0 .
Let L be a field extension of K[b] such that the minimal polynomial p of b is
a product of linear factors in L (e.g. the algebraic closure of K[b]). So p has
the form
p = (t − x1 ) . . . (t − xn ) ,
where the x1 , . . . , xn are elements of L.
As b is integral over R, there is a monic polynomial f ∈ R[t] ⊆ K[t]
with f (b) = 0. By the minimal property of the minimal polynomial, p | f
follows and hence all roots of p are also roots of f which shows that the
x1 , . . . , xn are integral over I. Since the non-leading coefficients of p are in
K[x1 , . . . , xn ] we get that they are integral over I too (Lemmab 2.U).
Since we assumed that R is normal, we get that √ the ai are√already in R.
By applying Lemmab 2.U to R0 = K, we get I = I, so ai ∈ I. 
Lemma 2.57. Let ϕ : R → R0 be a ring homomorphism and p ∈ Spec R a
prime ideal. Then the following are equivalent:
i) There is a prime ideal q ∈ Spec R0 such that p = ϕ−1 (q).
ii) It holds that ϕ−1 (ϕ(p)R0 ) = p.

Proof. This will be on the 5th exercise sheet. 

Theorem 2.58 (Going Down). Let ϕ : R → R0
be an integral extension.
0
Assume that R, R are integral domains and that R is normal. Then ϕ
satisfies going down.

End of Lecture 8
Proof. Let p1 ⊆ p2 ⊆ R and q2 ⊆ R0 be prime ideals such that
p2 = q2 ∩ R. We want to find a prime ideal q1 ∈ R0 such that p1 = q1 ∩ R
and q1 ⊆ q2 :
R0 ∃q1 ⊆ q2
ϕ

R p1 ⊆ p2
Consider now the map
ϕ η
ϕ0 : R R0 Rq0 2 .

We are going to show that under ϕ0 ,

p1 Rq0 2 ∩ R = p1


(∗)
42 2. MODULES AND INTEGRAL EXTENSIONS

holds. Assume for now that this is the case. Then Lemma 2.57 implies that
there is a prime ideal q01 ⊆ Rq0 2 such that q01 ∩ R = p1 . For q1 := q01 ∩ R0 we
have that q1 ∩ R = p1 , and q1 ⊆ q2 (by Corollary 1.47, i)), which shows that
ϕ satisfies going down.
We now prove (∗): Note first that p1 ⊆ p1 Rq0 2 ∩ R is always true. So it

suffices to show p1 Rq0 2 ∩ R ⊆ p1 . We will do this by contradiction:

Let x ∈ p1 Rq0 2 ∩R be any element and consider ϕ0 (x) ∈ p1 Rq0 2 , which we

will identify with x. As p1 Rq0 2 = (p1 R0 ) Rq0 2 there are y ∈ p1 R0 and s ∈ R0 \ q2

such that x = y/s (Theorem 1.46, i)). p Now by Lemmab 2.U, y is integral
√
over p1 (we have y ∈ p1 R0 ⊆ p1 R0 = p1 R, since R0 is integral over R.).
Let py ∈ (Quot (R)) [t] be the minimal polynomial of y over K := Quot(R)
(c.f. Lemmab 2.W). By Lemma 2.56, the non-leading coefficients of py , say
√
a0 , . . . , an−1 , are in p1 = p1 (since p1 is a prime ideal).
Assume now that x ∈ / p1 , so in particular x 6= 0 and hence s has the form
y/x ∈ Quot R0 , and 1/x ∈ K. This implies that the minimal polynomial ps
of s over k has the form
an−1 n−1 a0
tn + t + ... + n.
t x
Set ãi := ai /x n−i for (i = 0, . . . , n − 1). Since s ∈ R0 is integral over R (by
assumption on ϕ), we have by Lemma 2.56 that the ãi are already in R.
Now in R, we have ãi xn−1 = ai ∈ p1 , and since we assumed x ∈ / p1 , this
implies ãi ∈ p1 for all i = 0, . . . , n − 1. But then
sn = − ãn−1 sn−1 + . . . + ã0 ∈ p1 R ⊆ p2 R0 q2 ,

contradicting s ∈
/ q2 . 
More Examples of Going Down. The proofs for the following propo-
sitions will (hopefully) be added in the future.
Theoremb 2.X. Let ϕ : R → R0 be flat, i.e. R0 is flat as an R-module.
Then ϕ satisfies going down.
Theoremb 2.Y. Let ϕ : R → R0 be a ring homomorphism such that
Spec ϕ : Spec R0 → Spec R is open (i.e. maps open sets to open sets.). Then
ϕ satisfies going down.

2.7. Noether Normalization Lemma

Theorem 2.59 (Noether Normalization Lemma (NNL)). Let k be a field,
and A a finitely generated k-algebra. Let
I1 ⊆ I2 ⊆ . . . ⊆ Im ( A
be a chain of ideals in A. Then there is a n ≥ 0, a1 , . . . , an ∈ A and
0 ≤ h1 ≤ . . . ≤ hm ≤ n such that:
i) the elements a1 , . . . , an are algebraically independent over k;
ii) k[a1 , . . . , an ] ⊆ A is a finite ring extension; and
iii) Il ∩ k[a1 , . . . , an ] = (a1 , . . . , ahl ).

We will only prove i) and ii). The proof of iii) can be found in [Fra18b].
 2.7. NOETHER NORMALIZATION LEMMA 43

Lemma 2.60. Let k be a field, and 0 6= f ∈ k[t1 , . . . , tn ] a non-zero

polynomial. Then there are r1 , . . . , rn−1 ∈ N such that after the substitution
ti := Yi + trni (1 ≤ i ≤ n − 1), the polynomial f has the form
f = ctm m−1
n + h1 tn + . . . + hm ∈ k[Y1 , . . . , Yn−1 , tn ],
for a m > 0, c ∈ k × and h1 , . . . , hm ∈ k[Y1 , . . . , Yn−1 ].
Proof. Assume f has the form
X
f= bσ tσ1 1 . . . tσnn .
σ∈Nn
After substituting ti := Yi +trni
for (yet to be determined) ri ≥ 0, this becomes
X
f= bσ (Y1 + trn1 )σ1 . . . (Yn−1 + trnn−1 )σn−1 tσnn .
σ∈Nn
Let now τ ∈ Nn be a specific multi-index. Define
e (τ ) := τ1 r1 + . . . + τn−1 rn−1 + τn
With this notation, we obtain a factorisation of the τ -summand of f which
reads as follows:
n−1
Y
τn
bτ tn (Yi + trni )τi = bτ te(τ
n +
)

i=1
(terms where tn has stricly lower degrees) .
Claim 1. The ri can be choosen in such a way that for each different pair of
multi-indices σ, τ ∈ Nn , the associated exponents e(σ), e(τ ) are different too:
By definition, there is a M > 0 such that bσ = 0 for all multi-indices
σ ∈ Nn \{0, . . . , M − 1}. Set now r1 := M, r2 := M 2 , . . . , rn−1 := M n−1 ,
and let σ ∈ Nn be a multi-index with bσ 6= 0. Then the value of
n−1
X
e(σ) = σn + σi ri = σn + σ1 M + . . . + σn−1 M n−1
i=1
is uniquely determined by the values of the σ1 , . . . , σn (since the M -adic
expansion of a natural number is unique, and M was chosen in such a
way that σi < M for all 1 ≤ i ≤ n).

Now for such a choice of ri , there is a unique multi-index σ ∈ Nn such that

the corresponding exponent e(σ) is maximal and bσ 6= 0. After re-grouping
the expansion of f in decreasing order of powers of tn , the claim follows with
m := e(σ) and c := bσ . 
Proof of NNL (Theorem 2.59). Denote by x1 , . . . , xm ∈ A a set of
generators of A. We want to show that there are algebraically independent
a1 , . . . , an such that the ring homomorphism k[a1 , . . . , an ] → A is injec-
tive and A is a finitely-generated k[a1 , . . . , an ]-module. We will do this by
induction on the number m of generators:
The case m = 0 is trivial. So assume NNL holds for m − 1 gener-
atos. If the x1 , . . . , xm are algebraically independent then the canonical
map k[x1 , . . . , xm ] → A is indeed injective, and A is a finitely generated
k[x1 , . . . , xm ]-module.
44 2. MODULES AND INTEGRAL EXTENSIONS

If, however, the xi are not algebraically independent, then there is a poly-
nomial 0 6= f ∈ k[t1 , . . . , tm ] such that f (x1 , . . . , xm ) = 0. Set yi := xi − xrmi
for 1 ≤ i ≤ 0 and (yet to be determined) ri . We then have
rm−1

0 = f y1 + xrm1 , . . . , ym−1 + xm−1 .
But by Lemma 2.60, there is a set of exponents ri such that
r
0 = f (y1 + xrm1 , . . . , ym−1 + xm−1
m−1
) = cxdm + h1 xd−1
m + . . . + hd
with h1 , . . . , hd ∈ k[y1 , . . . , ym−1 ] and c ∈ k × . So xm is integral over
k[y1 , . . . , ym−1 ] and k[y1 , . . . , ym−1 ][xm ] is a finite k[y1 , . . . , ym−1 ]-module
(Corollary 2.41). By induction hypothesis, there are algebraically indepen-
→ k[y1 , . . . , ym−1 ] is finite. Hence
dent a1 , . . . , an such that k[a1 , . . . , an ] ,−
k[a1 , . . . , an ] k[y1 , . . . , ym−1 ] k[y1 , . . . , ym−1 ][xm ] = A
is finite (Lemma 2.40), which proves the claim. 

End of Lecture 9
Notation. Let s ∈ R and consider the multiplicative set S := 1, s, s2 , . . . .

Then the localization of an R-module M at S is denoted by
M [s−1 ] := S −1 M.
Lemmab 2.Z. Let R0 be an R-algebra of finite type, and S ⊆ R a mul-
tiplicative subset. Then the localization S −1 R0 of R0 (as R-algebra) is an
S −1 R-algebra of finite type.
Proof. Since R0 is of finite type, there is surjective ring homomorphism
R[t1 , . . . , tn ]  R0 . Since localization is exact, we get an induced epimorphism
S −1 R [t1 , . . . , tn ] ∼ S −1 (R[t1 , . . . , tn ]) S −1 R0 .


The following is a generalization of NNL to integral domains:
Proposition 2.61. Let R be an integral domain and R0 an R-algebra of
finite type.
i) There exists an element s ∈ R \ {0} and elements b1 , . . . , bn ∈ R0
such that
• the elements b1 , . . . , bn are algebraically independent over the
fraction field Quot R; and
• the ring extension
R[s−1 ][b1 , . . . , bn ] R0 [s−1 ]
is finite.
ii) For all prime ideals p ⊆ R[s−1 ] there is a prime ideal q ⊆ R0 [s−1 ]
such that q ∩ R[s−1 ] = p.
iii) For all prime ideals p ∈ Spec R[s−1 ] and the prime ideal q which
was constructed in ii), it holds that
Quot (R/(p ∩ R)) = Quot R[s−1 ]/p ⊆ Quot R0 [s−1 ]/q

= Quot R0 /(q ∩ R0 ) ,


 2.7. NOETHER NORMALIZATION LEMMA 45

and the extension is a finite field extension.

Proof.
i) Set S := R \ {0}. Then the induced extension

k := Quot R = S −1 R S −1 R0

is of finite type (Lemmab 2.Z) and we can apply Noethers Normal-

ization Lemma (Theorem 2.59) to get algebraically independent
elements b01 , . . . , b0n ∈ S −1 R0 such that k[b01 , . . . , b0n ] ,−
→ S −1 R0 is a
finite ring extension. Choose now representatives bi ∈ R0 and si ∈ S
such that b0i = bi /si . Then the bi are already algebraically indepen-
dent over k and k[b1 , . . . , bn ] = k[b01 , . . . , b0n ] (since 1/si is in k for
all i).
As R0 is an R-algebra of finite type, there are c1 , . . . , cm ∈ R0
(not necessarily algebraically independent) such that R[c1 , . . . , cm ] = R0 .
As S −1 R0 is finite over k[b1 , . . . , bn ] it is in particular integral over
k[b1 , . . . , bn ] (Corollary 2.41) and hence there are monic polynomials
fi ∈ k[b1 , . . . , bn ][t] such that fi (ci /1) = 0 in S −1 R0 (for 1 ≤ i ≤ m).
We can now choose a u ∈ S such that the coefficients of all of the
fi are already in the image of the morphism
R[u−1 ][b1 , . . . , bn ] −→ k[b1 , . . . , bn ]
which is induced by the embedding R[u−1 ] ,− → k. Then there
are monic polynomials gi ∈ R[u−1 ][b1 , . . . , bn ][t] such that gi gets
mapped to fi for all 1 ≤ i ≤ m. Since 0
0 −1 0 −1 −1 0

bi , ci ∈ R we have
gi ∈ R [u ]. So gi ∈ ker R [u ] → S R , as g(ci ) gets mapped
to f (ci /1) = 0 (by construction of the fi ). Hence there are
vi ∈ R\{0} such that vi gi (ci ) = 0 in R0 [u−1 ]. Define now v := v1 . . . vm
and s := vu.
By the universal property of the localization at u and the fact
that R[u−1 ][b1 , . . . , bn ] we get a morphism
ψ : R[u−1 ][b1 , . . . , bn ] −→ R[s−1 ][b1 , . . . , bn ],
which also induces a morphism between the corresponding polyno-
mial rings. Set hi := ψ(gi ) ∈ R[s−1][b1 , . . . , bn ][t]. Then the hi are
monic, since ψ is a ring homomorphism, and hi (ci )0. So the ci are in-
tegral over R[s−1 ][b1 , . . . , bn ], and thus R[s−1 ][b1 , . . . , bn ] ,−
→ R0 [s−1 ]
0 −1 −1
is finite (note R [s ] = R[s ][c1 , . . . , cm ] and then apply Corol-
lary 2.41).
ii) Let p ∈ Spec R[s−1 ] be a prime ideal. Define
p0 := pR[s−1 ][b1 , . . . , bn ] + hb1 , . . . , bn i.
Then p0 ∩ R[s−1 ] = p the map
R[s−1 ][b1 , . . . , bn ] −→ R[s−1 ]/p
1 7−→ 1
bi 7−→ 0
46 2. MODULES AND INTEGRAL EXTENSIONS

induces an isomorphism
R[s−1 ][b1 , . . . , bn ]/p0 −→
∼
R[s−1 ]/p.
(Note that this is well-defined, since the bi are algebraically inde-
pendent). As p is a prime ideal, we get that p0 is too.
Using Lying Over (Lemma 2.48) for the finite integral extension
(by i))
R[s−1 ][b1 , . . . , bn ] R0 [s−1 ]
we get that there is a prime ideal q sin Spec R0 [s−1 ] such that
q ∩ R[s−1 ][b1 , . . . , bn ] = p0 . Thus
q ∩ R[s−1 ][b1 , . . . , bn ] ∩ R[s−1 ] = p.

iii) Consider the following commutative diagram

finite
R[s−1 ][b1 , . . . , bn ] integral
R0 [s−1 ]

R[s−1 ][b1 , . . . , bn ]/p0 R0 [s−1 ]/q

Then
R[s−1 ]/p R0 [s−1 ]/q
is an integral extension too (Lemma 2.45) and finite, since
R[s−1 ][b1 , . . . , bn ] R0 [s−1 ]
is.
We also have that
Quot R[s−1 ]/p ,−
→ Quot R0 [s−1 ]/q
is finite, by Lemmab 2.W.
As s ∈/ p ∩ R, we have

.

R[s−1 ] p p · R[s−1 ] p ∼= Rp∩R /(Rp∩R ) ,
which implies
Quot R[s−1 ]/p ∼
= Quot R/p ∩ R,
as for any ring A and b ∈ Spec A it holds that Quot A/b ∼
= Ab / (b · Ab ).


 CHAPTER 3

Hilbert’s Nullstellensatz and some Algebraic

Geometry

3.1. Jacobson Rings

Definition 3.1. A ring R is a Jacobson ring if for all prime ideals p ⊆ R
it holds that
\
p= m.
p⊆m
m maximal
Lemma 3.2. For a ring R, the following are equivalent:
i) R is a Jacobson ring.
ii) For all prime ideals p ⊆ R and a ∈ R \ p, there is a maximal ideal
m such that p ⊆ m and a ∈ / m.
iii) For all ideals I ⊆ R it holds that
√ \
I= m.
I⊆m
m maximal

Proof. i) ⇐⇒ iii) follows from Proposition 1.28. 

Example 3.3.
i) Fields are Jacobson rings.
ii) If R is a local ring, which has only one prime ideal, the R is a
Jacobson ring.

Exampleb 3.A.
i) If R be a noetherian domain such that every non-zero prime ideal
is maximal and R has infinitely many maximal ideals, then R is
Jacobson: Since every non-zero ideal is maximal, it suffices to show
h0i = Jac R. For that, it suffices that every non-zero x ∈ R is only
contained in finitely many prime ideals, i.e. that Z(x) is finite. But
Z(x) is isomorphic to Spec R/hxi. Now R/hxi is noetherian and
every prime ideal is minimal. It is a general fact that noetherian
rings have only finitely many minimal prime ideals so the claim
follows.
ii) Claim i) implies that all factorial rings with infinitly many prime
ideals are Jacobson (like Z). Note that for a factorial ring the
condition Jac R = 0 suffices for being jacobson.
47
48 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

iii) Let on the other hand R be a domain which has only finitely many
prime ideals. Then R cannot be Jacobson: We have
0 6= m1 · . . . · mn ⊆ m1 ∩ . . . ∩ mn ,
for the maximal ideals m1 , . . . , mn .

Lemma 3.4. Let ϕ : R ,− → R0 be an integral extension. Assume R is a

0
Jacobson ring. Then R is too.
Proof. Let q ∈ Spec R0 be a prime ideal and set
\
J := m.
q⊆m
m∈MaxSpec R

We first show that J ∩ R = q ∩ R := p:

Since R is a Jacobson ring p is the intersection of all maximal ideals
containing it. Now for any maximal ideal m ∈ MaxSpec R with p ⊆ m, going
up (Theorem 2.51) implies that there is a prime ideal n ∈ Spec R0 with
n ∩ R = m and q ⊆ n. By Lemma 2.44 we get that n is maximal too, and
J ∩ R = p follows.
We are now in the following situation:
integral
R R0 p=R∩q q

Rp integral
R0 qRp0 .

Since q∩R = p, we have ϕ (R \ p)∩q = ∅, so qRq0 is prime in Rp0 . Furthermore,

we have qRp0 ∩ Rp = pRp . Since pRp is a maximal ideal (Corollary 1.47),
Lemma 2.44 implies that qRp0 is a maximal ideal.
As q ⊆ J we have qRp0 ⊆ JRp0 . Since (R ∩ J) ∩ (R \ p) = ∅ we have that
JRp0 is a proper ideal, so qRp0 = JRq0 follows. But now
J ⊆ JRp0 ∩ R0 = qRq0 ∩ R0

= q,
so J = q follows. 

3.2. Hilbert’s Nullstellensatz

Theorem 3.5 (Generalized Hilbert’s Nullstellensatz, GHNS). Let R be a
Jacobson ring and R0 an R-algebra of finite type. Then
i) R0 is a Jacobson ring too.
ii) For all maximal ideals n ⊆ R0 it holds that:
• The ideal m := n ∩ R is a maximal ideal.
• For this m, R/m ,−→ R0 /n is a finite field extension.

Proof.
i) Let q ∈ Spec R0 be a prime ideal and b ∈ R0 \ q. By Lemma 3.2, we
want to show that there is a maximal ideal n ∈ MaxSpec R0 such
that q ⊆ n and b ∈
/ n.
 3.2. HILBERT’S NULLSTELLENSATZ 49

Consider the ring extension:

R̃ := R/ (q ∩ R) R0 /q (R0 /q) [b−1 ] =: R̃0 .

Now R̃ is an integral domain and R̃ ,− → R̃0 is an R̃0 -algebra of finite

type. So by Proposition 2.61, there is a s̃ ∈ R̃ \ {0} such that:

For all p̃ ∈ Spec R̃ with s̃ ∈
 / p there is a
0
q̃ ∈ Spec R̃ such that
(∗)  q̃ ∩R̃ = p̃ and 
the extension Quot R̃/p̃ ,− → Quot R̃0 /q̃ is finite


Let s ∈ R be a preimage of s̃ under R  R̃, so s ∈

/ q ∩ R. Since R
is jacobson, there is a maximal ideal m ∈ MaxSpec R such that
q ∩ R ⊆ m and s ∈
/ m.
Hence s̃ is not contained in the maximal ideal mR̃ ∈ MaxSpec R̃.
By applying (∗) to m̃ := mR̃ we get a prime ideal ñ ∈ Spec R̃0 such
that ñ ∩ R̃ = m̃.
Consider now the finite field extension Quot R̃ ,− → Quot R̃0 .
We now have Quot R̃/m̃ = R̃/m as m̃ is maximal. We also have
R̃/m̃ = R/m and R̃0 /ñ = R0 /n (where n := ñ ∩ R0 ). So now we
arrive at the following commutative diagram:
finite
R R0

R/m R0 /n

finite
Quot R̃/m̃ Quot R̃0 /ñ

So R/m ,− → R0 /n is integral, and hence n is a maximal ideal

(Lemma 2.44).
Since n is the preimage of a prime ideal in quotient R0 /q, it
folows that q ⊆ n.
By the description of prime ideals in the localization (Theo-
rem 1.46) we get a bijection:
Spec R0 /q [b−1 ] −→
 0
∼
q ∈ Spec R0 q ⊆ q0 , b ∈
/ q0

q̃ 7−→ q̃ ∩ R0
So b ∈
/ n, which concludes the proof.
ii) Let q ∈ MaxSpec R0 be a maximal ideal. By applying the con-
struction for b = 1, we get a maximal ideal m ∈ MaxSpec R with
q ∩ R ⊆ m and a maximal ideal n ∈ MaxSpec R0 with n ∩ R = m
and q ⊆ n. Since q is a maximal ideal, q = n follows. So
n ∩ R = q ∩ R = m, which is maximal.
We also have that Quot R̃/m̃ ,− → Quot R̃0 /ñ is finite. Since
Quot R̃/m̃ = R/m̃ and Quot B̃/ñ = B/q, the claim follows.

50 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

End of Lecture 10
Corollary 3.6. Let k be a field and A a k-algebra of finite type. Then
i) A is a Jacobson ring.
ii) For all maximal ideals m ∈ MaxSpec A the map k → A → A/m is
a finite field extension.
iii) The maximal ideals of A are given by
MaxSpec A = {p ∈ Spec A | k → Quot A/p is finite} .
iv) Let f : A → B be a homomorphism of k-algebras of finite type and
m ∈ MaxSpec B a maximal ideal. Then m ∩ A is maximal too.

Proof.
i) This is just i) of GHNS (Theorem 3.5).
ii) By ii) of GHNS, m ∩ k is a maximal ideal in k. But since k is a
field, this implies that m ∩ k = h0i. So by the second part of ii),
→ A/m is a finite field extension.
k ,−
iii) The inclusion ⊆“ is just ii). For the other direction let p ⊆ A be a
”
prime ideal and assume that in
k A/p Quot A/p
the composition k ,− → Quot A/p is a finite field extension. Then
→ Quot A/p is necessarily finite, and so in particular integral.
A/p ,−
Now by Lemma 2.44, this implies that A/p is a field, and hence p
is a maximal ideal.
iv) Since B is of finite type, it is in particular an A-algebra of finite
type. By applying ii) of GHNS to A → B, the claim follows.

Corollary 3.7. Denote by k-Algf.t. the category whose objects are k-
algebras of finite type and whose morphisms are k-algebra homomorphisms
maps. Then MaxSpec(−) induces a contravariant functor
MaxSpec(−) : k-Algf.t. −→ Top
A 7−→ MaxSpec A
which maps k-algebra homomorphisms ϕ : A → B to the restriction of ϕ#
to MaxSpec B.
Proof. We only show that the restriction of ϕ# is well-defined, i.e. that
we indeed get a map ϕ# : MaxSpec B → MaxSpec A. But this is just iv) of
Corollary 3.6. 
Theorem 3.8 (Weak Nullstellensatz). Let k be an algebraically closed field.
For a tupel x = (x1 , . . . , xn ) ∈ k n , denote by mx the ideal
mx := ht1 − x1 , t2 − x2 , . . . , tn − xn i ⊆ k[t1 , . . . , tn ].
Then mx is a maximal ideal and the assignment
k n −→ MaxSpec k[t1 , . . . , tn ]
x 7−→ mx
 3.3. THE DIMENSION OF A RING 51

is a bijection.
Proof. As k[t1 , . . . , tn ]/mx ∼
= k we find that mx is indeed a maximal
ideal. We also have mx 6= my for x, y ∈ k n with x 6= y.
It remains to show the surjectivity: Let m ∈ MaxSpec k[t1 , . . . , tn ] be a
maximal ideal. Then by Corollary 3.6, we have that k ,− → A/m is a finite
field extension. As k is algebraically closed, there are no non-trivial algebraic
field extensions of k, so in particular there are no finite extensions. So we
have k ∼ = a/m.
∼
Denote by π the map π : A  A/m −→ k and set xi := π(ti ). Then
ti − xi ∈ ker π, and so mx = ht1 − x1 , . . . , tn − xn i ⊆ ker π = m. But as mx
is maximal and m 6= A, it follows that mx = m. 
Remarkb 3.B. That a field satiesfies the Weak Nullstellensatz is equiva-
lent to k being algebraically closed, since it implies that every irreducible
polynomial in k[t] is of the form t − a for a a ∈ k.

3.3. The Dimension of a Ring

Definition 3.9. Let R be a ring.
i) The dimension of R is defined as
 
there is an ascending chain of prime ideals
dim R := sup l ∈ N
p0 ( p1 ( . . . ( pl
if it exists and dim R := ∞ otherwise.
ii) Let p ∈ Spec R be a prime ideal. Then the height is defined as
 
there is an ascending chain of prime ideals
ht p := sup l ∈ N
p0 ( p1 ( . . . ( pl ⊆ p
if it exists and ht p := ∞ otherwise.

Example 3.10.
i) Let k be a field. Then dim k = 0, as every prime ideal is maximal.
ii) Let R be a prinicipal ideal domain which is not a field. Then every
ascending chain of prime ideals in R is of the form h0i ( hpi, for a
prime element p ∈ R. So dim R = 1.
iii) Consider the polynomial ring in n variables over a field k. Then
0 ( ht1 i ( ht1 , t2 i ( . . . ( ht1 , . . . , tn i
is stricly ascending, so dim k[t1 , . . . , tn ] ≥ n.

Lemma 3.11. Let p ∈ Spec R be a prime ideal.

i) It holds that ht p = dim Rp .
ii) It holds that dim R ≥ dim R/p + ht p.
iii) The dimension of R is local in the following sense:
dim R = sup {dim Rm | m ∈ MaxSpec R}
= sup {ht m | m ∈ MaxSpec R} .
52 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

Proof.
i) By the classification of prime ideals in Rp (Corollary 1.47), there is
a order-preserving bijection
∼
Spec Rp {p0 ∈ Spec R | p0 ⊆ p} .

Now the statement is now just the definition of height and dimension.
ii) This follows from the classification of prime ideals of R/p (Remarkb 1.G)
and the definition of height and dimension.
iii) After localising at a maximal ideal, the image of an ascending chain
is still an ascending chain. Vice versa, every ascending chain in the
localization at a maximal ideal can be lifted to an ascending chain
in R. This shows the first equality. The second equality follows
from i).


Definitionb 3.C. Let p0 ( . . . ( pn be an ascending chain in R. We say

that this chain is maximal if there is no prime ideal p ∈ Spec R with p ( p0
or pi ( p ( pi+1 (for 0 ≤ i ≤ n − 1) or pn ( p.
Lemma 3.12. Let R be a ring with dim R < ∞. Assume all maximal
chains in R have the same length. Let p ∈ Spec R be a prime ideal.
i) All maximal chains in R/p have the same length.
ii) It holds that dim R = dim R/p + ht p.
iii) It holds that dim Rp = dim R if p is a maximal ideal.

Proof. Let q0 ( . . . ( qr be a maximal chain in R/p. We can lift this

to a chain in R and complete it to a maximal chain of the form
p0 ( . . . ( pm = p = q0 ( . . . ( qr .
Now ht p ≥ m and dim R/p ≥ r. By the assumption that every maximal
chain in R has the same length, it follows that m+r = dim R. By Lemma 3.11
ii), we have
dim R ≥ dim R/p + ht p = m + r = dim R.
So ht p = m and dim R/p = r follows. This shows ii). For i), note that the
completion of the lift of the ideal chain in R is independend from r, so r has
to be necessarily the same for all prime chains in R/p. Claim iii) now follows
from ii), since for a maximal ideal, we have dim R/p = 0. 

Proposition 3.13. Let ϕ : R ,− → R0 be an integral ring extension.

i) It holds that dim R = dim R0 .
ii) For all q ∈ Spec R0 it holds that dim Rq∩R ≥ dim Rq0 .
iii) If ϕ satisfies going down then dim Rq∩R = dim Rq0 .

Proof.
i) Let q0 ( . . . ( ql be an ascending chain in R0 . Then q0 ∩R ( . . . ( ql ∩R
is an ascending chain in R (That the inclusions are strict follows
from Lemma 2.47). This shows dim R ≥ dim R0 .
 3.3. THE DIMENSION OF A RING 53

Let p0 ( . . . ( pd be an ascending chain in R. Then by Ly-

ing Over (Lemma 2.48) there is a prime ideal q0 ∈ Spec R0 with
q0 ∩ R = p. Now by Theorem 2.51 there is a chain q0 ( . . . ( qd in
R0 such that the following diagram is commutative:

q0 ( q1 ( ... ( qd

p0 ( p1 ( ... ( pd

That the inclusions are strict follows from Lemma 2.47.

ii) Let q0 ( . . . ql = q be a chain in R0 . This gives a chain of the form
q0 ∩ R ( . . . ( ql ∩ R in R. By applying Lemma 3.11 twice, the
claim follows.
iii) Let p0 ( . . . pl = q ∩ R be a chain in R. By applying going up, we
can lift this to a chain of the form

q0 ) ql−1 ) ... ) q0

q∩R ) pl−1 ) ... ) p0

So dim Rq0 ≥ dim Rq∩R . The claim now follows from ii).


Lemmab 3.D (Going Between). Let k be a field and R a k-algebra of finite

type. Let R ,−→ R0 be an integral ring extension. Let p1 ( p2 ( p3 be prime
ideals in k and q1 ( q3 prime ideals in R0 . Assume that q1 ∩ R = p1 and
q3 ∩ R = p3 . Then there is a prime ideal q2 ∈ Spec R0 satisfying q1 ( q2 ( q3 .
Proof. This is on the sixth exercise sheet. 

Remarkb 3.E. Note that in the setting of going between, it does not
necessarily hold that q2 ∩ R = p2 . A counterexample can also be found in
the solutions to sheet 6.
Theorem 3.14. Let k be a field.
i) It holds that dim k[t1 , . . . , tn ] = n.
ii) All maximal chains in k[t1 , . . . , tn ] = n have the same length.

Proof. We will do this by induction on n. In the case n = 1, k[t] is an

integral domain, and hence has dimension 1. In particular, maximal chains
have necessarily the same length.
In the general case, consider a chain of prime ideals p0 ( . . . ( pm in
k[t1 , . . . , tn ]. Without loss of generality, we can assume that this chain is
maximal. This implies that p = h0i (since k[t1 , . . . , tn ] is an integral domain),
pm is a maximal ideal and p1 = hf i, where f is an irreducible polynomial
(since k[t1 , . . . , tn ] is factorial). Using Lemma 2.60, we can assume that f
is a monic polynomial in tn , with non-leading coefficients in k[t1 , . . . , tn−1 ].
54 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

→ k[t1 , . . . , tn ]/hf i. Consider

So we get an integral extension k[t1 , . . . , tn−1 ] ,−
now the images of the original chain under the following maps:
π # : Spec k[t1 , . . . , tn ] −→ Spec k[t1 , . . . , tn ]/p1
pi 7−→ pi /p1
ϕ# : Spec k[t1 , . . . , tn ]/p1 −→ Spec ϕk [t1 , . . . , tn−1 ]
q 7−→ q ∩ k[t1 , . . . , tn−1 ],
for all prime ideals pi in the chain we started with. Then both maps preserve
prime ideals (the image of a prime ideal under π # is again prime, since π
maps into a quotient), strict inclusions of prime ideals (ϕ# by Lemma 2.47).
Furthermore, maximal chains are mapped to maximal chains:
In the chain
0 ( p2 /p1 ∩ k[t1 , . . . , tn−1 ] ( . . . ( pm /p1 ∩ k[t1 , . . . , tn−1 ]
the ideal p2 /p1 ∩ k[t1 , . . . , tn−1 ] is again maximal, by Corollary 3.6 iii). Now
by Going Between (Lemmab 3.D), the chain is indeed maximal. But by
the induction hypothesis, the length of a maximal chain in k[t1 , . . . , tn−1 ] is
exactly n − 1, for all maximal chains. Using the prime ideal correspondence
for the quotient, the claim follows. 
Let k be a field and A a k-algebra of finite type. Then by the Noether
Normalization Lemma (Theorem 2.59), there are algebraically independet
elements e1 , . . . , en such that A is integral over k[e1 , . . . , en ].
Corollary 3.15. In this case, n = dim A holds.
Proof. By Proposition 3.13, we have dim A = dim k[e1 , . . . , en ] and
dim k[e1 , . . . , en ] = n by Theorem 3.14. 
Corollary 3.16. Let A be a k-algebra of finite type and assume that A is
an integral domain. Then all maximal chains in A have the same length.
Proof. Let ϕ : k[t1 , . . . , tn ]  A, then A ∼ = k[t1 , . . . , tn ]/ ker ϕ. Since
A is an integral domain it follows that ker ϕ is a prime ideal. Since by
Theorem 3.14 all chains in k[t1 , . . . , tn ] have the same length and ker ϕ is
prime, the same is true for k[t1 , . . . , tn ]/ ker ϕ (by Lemma 3.12). 
Corollaryb 3.F. Let A be a k-algebra of finite type and integral domain.
Then for every maximal ideal m ∈ MaxSpec A, we have dim Am = dim A.
Proof. This now follows from the more general result Lemma 3.12. 

End of Lecture 11

3.4. Zero Sets and Varieties

Definition 3.17.
i) Let k be a field and n ≥ 1 an integer. We denote by Ank := k n the
affine space.
ii) Let S ⊆ k[t1 , . . . , tn ] be a subset. We denote by
V (S) := {a ∈ Ank | f (a) = 0 for all f ∈ S} ,
 3.4. ZERO SETS AND VARIETIES 55

the vanishing set of S. Here, f (a) = f ((a1 , . . . , an )) is a short-hand

notation for the image of f under the evaluation morphism Xi 7→ ai .
iii) Subsets of Ank , which are of the form V (T ) for a subset T ⊆ k[t1 , . . . , tn ],
are called varieties or algebraic subsets.

Example 3.18. These are  some examples of varieties:

i) V (x2 + y 2 − 1) = (a, b) ∈ k 2 a2 + b2 = 1 .

ii) V (x · y) = (a, b) ∈ k 2 a = 0 or b = 0 .


iii) V ({x − a, y − b}) = {(a, b)}.

Lemma 3.19.
i) Let S1 ⊆ S2 ⊆ k[t1 , . . . , tn ]. Then V (S1 ) ⊇ V (S2 ).
ii) Let S ⊆ k[t1 , . . . , tn ]. Then V (S) = V (hSi).

Proof. Ommited. 
Example 3.20. We want to describe the varieties in A1k for k algebraically
closed. Let I ⊆ k[t] be an ideal. Since k[t] is a prinicipal ideal domain, there
is a f ∈ k[t] such that I = hf i. Then, using Lemma 3.19 we get
V (I) = V ({f })
= {x ∈ k | f (x) = 0}
= {x ∈ k | (t − a1 ) · . . . · (t − an ) = 0}
= {a1 , . . . , an } .
Note that it was crucial that k is algebraically closed, since otherwise we
would have not obtained a factorisation of f .
Definition 3.21.
i) Let X ⊆ Ank be a subset. Define
I(X) := {f ∈ k[t1 , . . . , tn ] | f (x) = 0 for all x ∈ X} .
ii) Let X ⊆ Ank be a subset. Define A(X) := k[t1 , . . . , tn ]/ I(X) as the
ring of polynomial functions on X or the coordinate ring at X.
Note that A(X) is a finitely-generated k-algebra.
Proof. I(X) is indeed an ideal, so this is well-defined. 
Remarkb 3.G. Note that A (Ank ) = k[t1 , . . . , tn ].
Example 3.22. Let a := (a1 , . . . , an ) ∈ Ank . Then
I(a) = ma = (t1 − a1 , . . . , tn − an ) .
Definition 3.23. Let Y ⊆ X ⊆ Ank and S ⊆ A(X).
i) The set VX (S) := {x ∈ X | f (x) = 0 for all f ∈ S} is a subvariety
of X.
ii) We define IX (Y ) := {f ∈ A(X) | f (y) = 0 for all y ∈ Y } .

Lemma 3.24. Let X ⊆ Ank be a variety, Y, Y 0 ⊆ X and S, S 0 ⊆ A(X).

56 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

i) If Y ⊆ Y 0 , then IX (Y ) ⊇ IX (Y 0 ). If S ⊂ S 0 , then VX (S) ⊇ VX (S 0 ).

ii) It holds that Y ⊆ VX (IX (Y )) and S ⊆ IX (VX (S)).
iii) If Y is a subvariety of X, then Y = VX (IX (Y )).
iv) If Y is a subvariety of X, then A(X)/ IX (Y ) = A(Y ).
v) It holds that IX (Y1 ∪ Y2 ) = IX (Y1 ) ∩ I(Y2 ).

Proof. i), ii) and v) are clear from the definition. For iii), note that we
only need to show VX (IX (Y )) ⊆ Y , since the other inclusion follows from ii).
But since Y is a subvariety of X, there is a S ⊆ A(X) such that Y = VX (S).
Then, using i) and ii), we get
ii) i)
S ⊆ IX (VX (S)) = I(Y ) =⇒ VX (IX (Y )) ⊆ VX (S) = Y.
For iv), note that the restricion map
A(X) −→ A(Y )
f 7−→ f |Y
is well-defined and surjective, with kernel IX (Y ). 
Notation 3.25. In the following, if the subset X is clear, we will ommit it
in the notation for I and V.
Remark 3.26. In light of Lemma 3.24 it seems reasonable to ask if
I(V (J)) ⊆ J holds for a general ideal. But in general, this is far from
being true:
i) For the ideal
J := h(t − a1 )k1 · . . . · (t − an )kn i ⊆ C[t]
with ki ≥ 0 and ai ∈ C, we have V (J) = {a1 , . . . , an }. So I(V (J))
consists of all polynomials which have each of the factors t − ai at
least once. So if one of the ki is greater than 1 J is a proper subset
of I(V (J)).
ii) Consider
J := ht2 + 1i ⊆ R[t].
Then I(J) = ∅, so I(V (J)) = I(∅) = R[t].

End of Lecture 12

Lemma 3.27. Let Y ⊆ Ank be a subset. Then V (I (Y )) = Y .

Proof. now that Y ⊆ V (I(Y )), by Lemma 3.24. Since V (I(Y )) is
closed, we get Y ⊆ V (I(Y )).
On the other hand, Y is closed. So by definition, there is an ideal
J ⊆ k[t1 , . . . , tn ] such that Y = V (J). Since J ⊆ I(Y ), we have
Y = V (J) ⊆ V (I(Y )) .

We now prove yet another Hilbert’s Nullstellensatz:
 3.5. THE ZARISKI-TOPOLOGY ON An
k 57

Theorem 3.28 (Hilbert’s

√ Nullstellensatz). Let J ⊆ k[t1 , . . . , tn ] be an ideal.
Then I (V (J)) = J holds.
Proof. We have
V (J) = {a ∈ Ank | f (a) = 0 for all f ∈ J}
= {a ∈ Ank | J ⊆ ma := ker (eva : k[t1 , . . . , tn ] → k)}
and
I(V (J)) = {f ∈ k[t1 , . . . , tn ] | f (a) = 0 for all a ∈ V (J)} .
Hence f ∈ I(V (J)) if and only if for all a ∈ Ank with J ⊆ ma it holds that
J ∈ ma , so
\
I(V (J)) = ma .
a∈An
k
J⊆ma

By the Weak Nullstellensatz (Theorem 3.8) we have

MaxSpec k[t1 , . . . , tn ] = {ma | a ∈ Ank } ,
and hence
\
I(V (J)) = m
m∈MaxSpec k[t1 ,...,tn ]
J⊆m
√
= J,
as k[t1 , . . . , tn ] is a Jacobson ring (Corollary 3.6, and then Lemma 3.2). 

3.5. The Zariski-Topology on Ank

Convention. In the following, k always denote an algebraically closed
field.
Lemma 3.29. Let I, J be ideals in k[t1 , . . . , tn ] and {Il }l∈L a family of
ideals in k[t1 , . . . , tn ]. P T
i) It holds that V ( l Il ) = l V (Il ).
ii) It holds that V (IJ) = V (I ∩ J) = V (I) ∪ V (J).

Proof. P
i) WeP have V T( l Il ) ⊆ V (Il ) for all l ∈ L, by Lemma 3.19. So
V ( l Il ) ⊆ l V (IT l ). P
Let now x ∈ P l V (Il ). Then for any polynomial f ∈ l Il ,
f (x) = 0, so x ∈ V ( l Il ).
ii) We have V (J) ⊇ V (I ∩ J) ⊇ V (I) ∪ V (J). Let x ∈ V (IJ), with
x∈ / V (I). Then there is a f ∈ I such that f (x) 6= 0. Now for any
g ∈ J, we have f g ∈ IJ and hence (f g)(x) = 0. So g(x) = 0 and
thus x ∈ V (J).

Proposition 3.30. By the above lemma, the subsets of the form V (I) ⊆ Ank
satisfy the axioms of closed sets of a topology on Ank . This topology is the
Zariski-Topology on Ank
58 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

Remark 3.31. The  Zariski-Topology on Ank is quite weird:

i) Let X := x ∈ A1C |x| ≤ 1 be the unit disk in C. As closed
subsets of A1C are finite sets (Example 3.20), we get X = A1C .
ii) Let ϕ : A1C → A1C be bijective. Then preimages of finite sets are
finite. So ϕ is already continous.

Lemmab 3.H (Prime Avoidance). Let p, p1 , . . . , pm ∈ Spec R be prime

. . . , In ⊆ R ideals.
ideals and I, I1 ,T
i) If IT⊆ m i=1 pi , then there exists an 1 ≤ i ≤ m such that I ⊆ pi .
n
ii) If Ti=1 Ii ⊆ p, then there is a 1 ≤ i ≤ n such that Ii ⊆ p.
iii) If ni=1 Ii = p, then there is a 1 ≤ i ≤ n such that Ii = p.
iv) Parts ii) also holds if one of the ideals is not prime.

Proof. This is on the seventh exercise sheet. (Part iv) actually not). 
Definition b 3.I. Let I ⊆ R be an ideal. We say I is a radical ideal if
√
I = I holds.
Definitionb 3.J. A topological space X is called irreducible if every de-
composition X = X1 ∪ X2 in closed subsets X1 , X2 implies that X1 or X2
equals X.
Theorem 3.32.
i) The maps
{varieties in Ank } ←→ {radical ideals in k[t1 , . . . , tn ]}
Y 7−→ I(Y )
V (J) ←−[ J
are mutually inverse bijections.
ii) The bijection from i) restricts to a bijection
{closed, irreducible subsets of Ank } ←→ Spec k[t1 , . . . , tn ].
iii) The bijection from i) restricts to a bijection
{{x} ⊆ Ank } ←→ MaxSpec k[t1 , . . . , tn ].

Proof. p
i) Since k is an integral domain, we have I(X) = I(X) for all
X ⊆ Ank . If X ⊆ Ank is a variety then by Lemma 3.27, we have
V (I(X)) = X. Since the closed sets in Ank are precisley the varieties,
we get V (I(X)) = X.
Let J ⊆ k[t1 , . . . , tn ] be a radical√ideal. By the Nullstellensatz
(Theorem 3.28), we have I(V (J)) = J = J.
ii) Let Y ⊆ Ank be closed and irreducible. We want to show that I(Y )
is a prime ideal. We do this by contradiction:
Assume there are polynomials f, g ∈ k[t1 , . . . , tn ] \ I(Y ) such
that f g ∈ I(Y ). Now V (I(Y ) + hf i) , V (I(Y ) + hgi) ⊆ Y are proper
 3.5. THE ZARISKI-TOPOLOGY ON An
k 59

subsets of Y , because if V (I(Y ) + hf i) = Y , then f (y) = 0 would

hold for every y ∈ Y which would mean f ∈ I(Y ).
As Y is irreducible it holds that
V (I(Y ) + hf i) ∪ V (I(Y ) + hgi) ( Y
By Lemma 3.24, we have
V (I(Y ) + hf i) ∪ V (I(Y ) + hgi) = V ((I(Y ) + hf i) · (I(Y ) + hgi))
We also have
(I(Y ) + hf i) · (I(Y ) + hgi) ⊆ I(Y ) + hf gi
= I(Y ),
as we assume f g ∈ I(Y ). Putting all of this together, we get
Y = V (I(Y ))
= V (I(Y ) + hf gi)
⊆ V (I(Y ) + hf i) ∩ V (I(Y ) + hgi)
( Y,
which is not possible. This shows that the I(Y ) is indeed a prime
ideal.
For the other map, let p ∈ Spec k[t1 , . . . , tn ] be a prime ideal and
assume there is a decomposition V (p) = V (I1 ) ∩ V (I2 ) = V (I1 I2 )
for some ideals I1 , I2 ⊆ k[t1 , . . . , tn ]. Using part i), we get
√ p
p = p = I1 I2 ⊇ I1 ∩ I2 ,
which implies I1 ⊆ p or I2 ⊆ p (by Prime Avoidance, Lemmab 3.H,
ii)). Thus V (p) ⊆ V (I1 ) or V (p) ⊆ V (I2 ) and so V (p) is indeed
irreducible.
iii) Let m ∈ MaxSpec k[t1 , . . . , tn ] be a maximal ideal. Then by the
Weak Nullstellensatz (Theorem 3.8) it is of the form m = ma for an
a ∈ Ank . We now have V (ma ) = {a}.
For the other direction, let a ∈ Ank be a point. We then have
I ({a}) = {f ∈ k[t1 , . . . , tn ] | f (a) = 0} = ma .


Corollary 3.33. Let X ⊆ Ank be a variety.
i) The maps
{closed subsets of X} ←→ {radical ideals in A(X)}
Y 7−→ IX (Y )
VX (J) ←−[ J
are mutually inverse bijections.
ii) The bijection from i) restricts to a bijection
{closed, irreducible subsets of X} ←→ Spec A(X).
60 3. HILBERT’S NULLSTELLENSATZ AND SOME ALGEBRAIC GEOMETRY

iii) The bijection from i) restricts to a bijection

{{x} ⊆ X} ←→ MaxSpec A(X).

Definition 3.34. Let X be a topological space. A closed, irreducible subset

C ⊆ X is an irreducible component if for all closed subsets Z ⊆ X with
C ⊆ Z already C = Z follows.
Corollary 3.35. Let X ⊆ Ank be a variety. Then the maps

{irreducible components of X} ←→ {minimal prime ideals in A(X)}

Y 7−→ IX (Y )
VX (J) ←−[ J

are mutually inverse bijections.

End of Lecture 13

3.6. Morphisms of Varieties

Definition 3.36. Let X ⊆ Ank be a variety and A(X) the coordinate ring.
i) We say a function ϕ : X → k is regular if there is a polynomial
f ∈ k[t1 , . . . , tn ] such that ϕ(a) = f (a) for all a ∈ X. We denote
the set of regular functions X → k is denoted by O(X).
ii) Let X, X 0 be varieties, with X ⊆ Ank and X 0 ⊆ Am k . We say a
function (ϕ1 , . . . , ϕm ) = ϕ : X → X 0 is regular or a morphism of
affine varieties if all components are regular. We denote the set of
regular functions X → X 0 by Hom(X, X 0 ).

Remarkb 3.K. The set O(X) has a natural ring structure, which is given
by (f + g)(x) := f (x) + g(x) and (f · g)(x) := f (x)g(x). That’s why it is also
called the ring of regular functions on X.
Lemma 3.37. The ring O(X) of regular functions and the coordinate ring
A(X) are isomorphic.
Proof. Consider the ring homomorphism

k[t1 , . . . , tn ] −→ O(X)
f 7−→ (x 7→ f (x)) .

Then by definition of O(X) this map is surjective. The kernel is given

precisley by I(X), so A(X) = k[t1 , . . . , tn ]/ I(X) ∼
= O(X). 

Theorem 3.38. Let X ⊆ Ank , X 0 ⊆ Am

k be varieties. Then there is a
bijection

Hom(X, X 0 ) −→
∼
hom(O(X 0 ), O(X)).
 3.7. SOME EXAMPLES 61

Proof. Let ϕ : X → X 0 be a morphism of affine varieties, then ϕ

induces a ring homomorphism
ϕ∗ : O(X 0 ) −→ O(X)
 ψ
  ϕ ψ

X0 −
→k 7−→ → X0 −
X− →k .
We show that the assignment
(−)∗ : Hom(X, X 0 ) −→ hom(O(X 0 ), O(X))
ϕ 7−→ ϕ∗
is a bijection:
For that, note that ϕ is uniquely determined by ϕ∗ : Consider the coordi-
nate function
yi : X 0 −→ k
(y1 , . . . , ym ) 7−→ yi .
Then ϕ∗ (yi ) = ϕi for all 1 ≤ i ≤ m. So in particular, ϕ∗ is injective.
For the surjectivity of (−)∗ , let β : O(X 0 ) → O(X) be a ring homomor-
phism. Define a map
ϕ = (ϕ1 , . . . , ϕm ) : X → Am
k
by ϕi := β(yi ). Then ϕ∗ = β. It remains to show that im ϕ ⊆ X 0 . For
that, let a ∈ X. Since X 0 is a variety, it suffices to show ϕ(a) ∈ V (I(X 0 )).
Using the isomorphism from Lemma 3.37, we can associate to g an element
g ∈ A(X 0 ) ∼
= O(X 0 ), and as g ∈ I(X 0 ) we have g = 0. Hence ϕ∗ (g) = 0 and
thus g(ϕ(a)) = 0 for all a ∈ X. 

3.7. Some examples

This part of the lecture will be added at some point in the future.
 CHAPTER 4

Noetherian Rings and Modules

Definition 4.1. Let R be a ring and M be an R-module.

i) We say M is noetherian if every ascending chain M0 ⊆ M1 ⊆ . . . of
submodules of M terminates after finitely many steps, i.e. there is
a n ∈ N such that Mk = Mn for all k ≥ n.
ii) We say M is artinian if every descending chain M0 ⊇ M1 ⊇ . . . of
submodules of M terminates after finitely many steps, i.e. there is
a n ∈ N such that Mk = Mn for all k ≥ n.
iii) We say R is noetherian/artinian if it is noetherian/artinian as an
R-module.

Example 4.2.
i) Every field is noetherian and artinian.
ii) Let V be a vector space. Then V is noetherian if and only if it is
artinian if and only if it is finite dimensional.
iii) The ring of integers Z is noetherian: For every chain of ideals
I0 ⊆ I1 ⊆ . . ., the ideal I1 is generated by a single element a ∈ Z.
Since Z/aZ is finite, there are only finitely many ideals lying over
I1 .
However, Z is not artinian: The chain Z ) 2Z ) 4Z ) . . . is
stricly descending but does not terminate.
iv) The polynomial in infinitely many variables over a field k[t1 , . . .] is
neither noetherian nor artinian: The chain ht1 i ( ht1 , t2 i ( . . . is
stricly increasing but does not terminate; the chain ht1 i ) ht21 i ) . . .
is stricly decreasing but does not terminate.

Remarkb 4.A. There are noetherian rings with dim R = ∞. An example

can be found in [EE95, Exercise 9.6].

End of Lecture 14
Lemma 4.3. Let M be an R-module.
i) The following are equivalent:
a) M is noetherian.
b) Every non-empty family of submodules of M has an inclusion-
maximal element.
c) Every submodule of M is finitely generated.

ii) M is artinian if and only if every non-empty family of M has an

inclusion-minimal element.

62
 4. NOETHERIAN RINGS AND MODULES 63

Proof.
i) For a) =⇒ b), let (Ni )i be a family of submodules of M . Then
(Ni ) is partially ordered by inclusion, and since M is noetherian,
every chain of elements from (Ni ) has an upper bound (namely the
subspace that terminates the chain). So by Zorn’s Lemma, there is
an inclusion-maximal subspace of (Ni ).
For b) =⇒ c), let N be a submodule of M and let (Ni ) be
the family of finitely-generated submodules of N . Then (Ni ) is
non-empty, since {0} ⊆ N is finitely-generated. So by b), there
is an inclusion-maximal subspace P ∈ (Ni ). Assume that P is a
proper submodule of N , and let P be generated by the elements
p1 , . . . , pk . Then there is an element pk+1 ∈ N \ P . But now the
subspace hp1 , . . . , pk+1 i is a finitely-generated submodule of N that
has P as a proper subset. This contradicts the maximality of P .
Finally, for c) =⇒ a), let M0 ( M1 ( . . . be an ascending
chain of submodules of M . Consider the subspace M̃ := ∪Mi . Then
by assumption, M̃ is finitely generated. So there is a i ≥ 0 such
that the set of generators is in Mk for all k ≥ i. So Mk = Mi for
all k ≥ i follows, and hence the chain terminates.
ii) That M being artinian implies that every non-empty family of
subspaces has an inclusion-minimal element can be shown analogous
to a) =⇒ b) in i). For the other direction, note that every
descending chain of submodules of M has an inclusion-minimal
element which necessarily terminates the chain.


Corollary 4.4. Every principal ideal domain is a noetherian ring.

Remark 4.5. Let I ⊆ R be an ideal and M an R-module. Then the quo-
tient M/IM is both an R/I-module and an R-module. Using the definition
of a noetherian module, we get that M/IM is noetherian as an R-module if
and only if it is noetherian as an R/I-module (since statements about chains
of submodules are independent from the ground ring, and R-submodules are
precisley R/I-submodules).
Lemma 4.6. Let M be an R-module and
0 N M N0 0
be a short-exact sequence of R-modules.
i) M is noetherian if and only if both N and N 0 are.
ii) M is artinian if and only if both N and N 0 are.

Proof. By duality, it suffices to show i). Without loss of generality, we

can assume that N is a submodule of M and that N 0 = M/N . Assume first
that M is noetherian. Let N0 ⊆ N1 ⊆ N2 ⊆ . . . be a chain of submodules
of M . We can regard this as a chain in N and since M is noetherian, it
terminates in M at a module Nn . So the original chain in N terminates in
Nn too. Analogously, let P0 ⊆ P1 ⊆ . . . be a chain in M/N and denote by
q : M  M/N the canonical projection to the quotient. Set Mk := q −1 (Pk )
64 4. NOETHERIAN RINGS AND MODULES

for all k. Then M0 ⊆ M1 ⊆ . . . is an ascending chain in M , which terminates

in a submodule Mn since M is noetherian. Then the original chain terminates
in Pn = q(Mn ) (the equality holds since q is surjective.)
For the other direction, let M0 ⊆ M1 ⊆ . . . be an ascending chain in
M . Set Nk := Mk ∩ N and Pk := (Mk + N ) /N . So we get ascending chains
N0 ⊆ N1 ⊆ . . . and P0 ⊆ P1 ⊆ . . . in N and M/N respectivley. Since N and
M/N are noetherian, there is a n > 0 such that Pk = Pn and Nk = Nn for
all k ≥ n. Then the original chain in M terminates in n too:
Denote by i : N ,−→ M the inclusion of N into M and let x ∈ Mk for a
k ≥ n. Then there is a x0 ∈ Mn such that q(x) = q(x0 ) (as the chain termi-
nates in the quotient and q is surjective). So x − x0 ∈ ker q = im i and hence
there is a y ∈ N such that i(y) = x−x0 . This implies y ∈ i−1 (Mk ) = i−1 (Mn ).
Hence x = i(y) + x0 , which implies x ∈ Mn , as i(y) and x0 are. 

Corollary 4.7. Let R be a noetherian ring.

i) Let M, N be R-modules. Then M ⊕ N is noetherian if and only if
M, N are.
ii) Let M be a finitely-generated R-module. Then M is noetherian.

Proof.
i) This follows from the previous lemma, using the short-exact sequence
0 → M → M ⊕ N → N → 0.
ii) By i), Rn is noetherian for all n > 0. Since M is finitely-generated
as R-module, it is isomorphic to a quotient of Rn for a n > 0. So
by the previous lemma, M is noetherian.


Theorem 4.8 (Hilbert’s Basissatz, HBS). Let R be a noetherian ring.

Then the polynomial ring R[t] is also a noetherian ring.
Proof. We will do this by contradiction - assume R[t] is not noetherian.
So by Lemma 4.3 there is an ideal I ⊆ R[t] which is not finitely generated.
We can now inductivley choose elements f0 , f1 , . . . ∈ I which have the
following properties: The polynomial f0 ∈ I has minimal degree among all
polynomials in I. We then choose fn+1 as a polynomial of minimal degree
in I \ hf0 , . . . , fn i for all n > 0.
In this way, we get an infinite sequence

f0 , f1 , . . . ∈ I such that fn+1 ∈

/ hf0 , . . . , fn i.

Set dn := deg fn . Then, by construction, we have dn+1 ≥ dn . Let now ak be

the leading coefficient of fk = ak X dk + (lower order terms in X). This yields
the ascending chain ha0 i ⊆ ha0 , a1 i ⊆ . . . in R. Now since R is noetherian,
this terminates for a n and hence the leading coefficient of fn+1 , i.e. an+1 , is
of the form
an+1 = c0 a0 + . . . + cn an ,

for some c0 , . . . , cn ∈ R.
 4. NOETHERIAN RINGS AND MODULES 65

Consider now the polynomial

n
X
0
fn+1 := fn+1 − ck tdn+1 −dk fk .
k=0

Then the coefficient of tdn+1 is

n
X
an+1 − ck ak = 0,
k=0

by the above observation. So fn+1 0 0

is a polynomial with deg fn+1 < deg fn+1
0 0
P n d −d
and fn+1 ∈/ hf0 , . . . , fn i (as otherwise fn = fn+1 + k=0 ck t n+1 k f would
k
also be in hf0 , . . . , fn i). But this is a contradiction to the minimality of
fn+1 . 

Corollary 4.9. Let R be a noetherian ring and A an R-algebra of finite

type. Then A is a noetherian ring.
Proof. This is on exercise sheet 8. 

Remark 4.10.
i) Let X ⊆ Ank be a variety. Then A(X) is noetherian and hence every
ideal I ⊆ A(X) is finitely generated. So every subvariety of X is
already determined by finitely many polynomial equations.
ii) By the ascending chain condition for A(X), we get that every chain
of subvarities X0 ⊇ X1 ⊇ . . . terminates.
iii) Assume X has infinitely many points a1 , a2 , . . . ∈ X ⊆ A n
Sk . This
gives an ascending chain of closed subsets, by setting Xn := k≤n {ak }.
Then this chain corresponds to a descending chain of ideals in A(X)
which does not become stationary. So A(X) cannot be artinian.

Proposition 4.11.
i) If R is an artinian ring, then R has only finitely many maximal
ideals and all prime ideals are maximal.
ii) For a ring R the following are equivalent
a) R is artinian.
b) R is noetherian and every prime ideal is maximal.

Proof. This is on exercise sheet 8. 

Proposition 4.12. Let R be a ring and A an R-algebra of finite type

which is integral over R. Then for p ∈ Spec R, there are only finitely many
prime ideals in A which lie over p. This means that the induced map
Spec A → Spec R has finite fibre.
Proof. This in on exercise sheet 8. 

End of Lecture 15
66 4. NOETHERIAN RINGS AND MODULES

4.1. Dimension Theory of Noetherian Rings

Definition 4.13. Let p ∈ Spec R be a prime ideal and n ∈ N. We define
p(n) := (pn Rp ) ∩ R which is called the n-th symbolic power of p.
Lemma 4.14. Let p ∈ Spec R be a prime ideal.
i) It holds that
p(n) = {a ∈ R | there is a s ∈ R \ p such that sa ∈ pn } .
ii) The chain p(0) ⊇ p(1) ⊇ . . . is descending and p(0) = R, p(1) = p.
iii) It holds that pn ⊆ p(n) for all n ∈ N.
iv) It holds that
p(n) Rp = pn Rp = (pRp )n .

Proof.
i) For the inclusion “⊇“, let a ∈ R and assume that there is a s ∈ R \ p
such that sa ∈ pn . Then in Rp we have that a/1 = sa/s ∈ pn Rp
and hence a ∈ (pn Rp ) ∩ R.
For the reverse inclusion, let a ∈ (pn Rp ) ∩ R. Then a/1 ∈ pn Rp
and so there is a b ∈ pn and a t ∈ R \ p such that a/1 = b/t. Hence
there exists a u ∈ R \ p such that uta = ub ∈ pn . So s := ut ∈ /p
(since p is prime) and sa = ub ∈ pn .
ii) This is clear.
iii) This follows from i).
iv) The first equality is just Lemma 1.45 applied to p. For the second
n
P direction “⊆“, let b ∈ p . Then there are bij ∈ p
equality and the
such that b = i bi,1 . . . bi,n . For s ∈ R \ p we have
b X bi,1 bi,2 bi,n
= · ... ∈ (pRp )n
s s 1 1
i

For the inclusion ⊇“, let c ∈ (pRp )n . Then there are bi,j ∈ p
”
and si,j ∈ R \ p such that
X bi,1 bi,n
c= ... .
si,1 si,n
i
Set
  
 
Y X  Y n
 Y 
s := si,j and b := 
 sij 
  bi,j 
.
i,j i 1≤j≤n j=1
j6=i

Then b ∈ R \ p and b ∈ p(n) and thus c = b/s ∈ pn Rp .


Definitionb 4.B. Let a ∈ R and p ∈ Spec R be a prime ideal. We say that
p is minimal over a if a ∈ p holds and there is no prime ideal q ∈ Spec R
such that hai ⊆ q ( p holds.
 4.1. DIMENSION THEORY OF NOETHERIAN RINGS 67

Theorem 4.15 (Krull’s Principal Ideal Theorem). Let R be a noetherian

ring and p ∈ Spec R be a prime ideal. If there is an element a ∈ R such that
p is minimal over a then dim Rp ≤ 1 holds.
Proof. Let q0 ⊆ q ( p be a chain of prime ideals. We want to show
q = q0 . For that, consider the following simplifications
R/q0 R/q0 p .


R

Then (R/q0 )p is a local noetherian integral domain and by various prime

ideal coefficients, it suffices to show q (R/q0 )p = 0. So we show the following:
Let (R, p) be a local noetherian integral domain such that the unique maximal
ideal p is minimal over an element a ∈ R. Then all other prime ideals
q ∈ Spec R \ MaxSpec R are zero.
Claim 1. Let q ∈ Spec R be a prime ideal. Then q(n) ⊆ q(n+1) + hai.
The quotient R/hai is a noetherian ring and since there are no prime
ideals between hai and m, we have dim R/hai = 0. So R/hai is artinian
(Proposition 4.11, ii)). Consider now the prime ideal chain
   
q(0) + hai /hai ⊇ q(1) + hai /hai ⊇ . . .

in R/hai. Then this terminates (as R/hai is artinian) and so there

is a n ≥ 0 such that q(n) + hai ⊆ q(n+1) + hai, and in particular
q(n) ⊆ q(n+1) + hai.

Claim 2. In the above situation, it holds that q(n) = q(n+1) + pq(n) .

The inclusion ⊇ is clear since q(n+1) ⊆ q(n) (Lemma 4.14). For the other
inclusion let b ∈ q(n) . Then by the above claim we have b = c + ar with
c ∈ q(n+1) and r ∈ R. So ar = b − c ∈ q(n) . By Lemma 4.14 i), there
is a s ∈ R \ q such that s · ar ∈ qn . As p is minimal over a, we have
a∈ / q and thus sa · r ∈ qn implies r ∈ q(n) (again by Lemma 4.14,i)).
Now this gives b = c + ar, with c ∈ q(n+1) , r ∈ q(n) and a ∈ p and thus
b ∈ q(n+1) + pq(n) .

We now apply the Nakayama Corollary 2.25 to M := q(n) , N := q(n+1)

and I = p (Note that I = Jac R, since R is a local ring). Then as M = N +IM
we get q(n) = q(n+1) .
Consider now the localization Rq . By applying Lemma 4.14 iv) twice we
have
(qRq )n = q(n) Rq
= q(n+1) Rq
= (qRq )n+1 .

NoWe now apply the classical Nakayama Lemma (Lemma 2.24) to M = (qRq )n ,
I = qRq = Jac Rq , and since

M = (qRq )n = (qRq )n+1 = IM

68 4. NOETHERIAN RINGS AND MODULES

6 0, this implies qn = 0 and as R is a domain, this

this yields M = 0. As Rq =
ultimately shows q = 0. 
Lemma 4.16. Let R be an artinian ring. Then the Jacobson ideal Jac R
is nilpotent, i.e. there is a k ≥ 1 such that (Jac R)k 0.
Proof. As R is an artinian ring, Proposition 4.11 ii) implies that
N := Nil R = Jac R. The chain N ⊇ N 2 ⊇ . . . is decreasing and hence
terminates (as R is artinian). So there is k ≥ 1 such that N k = N k+1 =: a.
Assume a 6= 0. Then the set
 
b is an ideal
Σ := b⊆R
b · a 6= 0
is not empty as a ∈ Σ. Now Σ is partially ordered by inclusion and R being
artinian implies that Σ has an inclusion-minimal element c (by Lemma 4.3).
Now by the minimality condition c = hxi for an element x ∈ R (as there
is an x ∈ c with xa 6= 0). We also have xa ⊆ hxi, as (xa) a = xa2 = xa 6= 0
and hence xa ⊆ hxi.
So there is a y ∈ a wit x = xy. Hence x = xy k = 0, contradicting
x 6= 0. 
Lemma 4.17. Let (R, m) be a local noetherian ring with unique maximal
ideal m and I ( R a proper ideal. Then the following are equivalent:
i) There is a n ≥ 1 such that mn ⊆ I.
ii) For all prime ideals q ∈ Spec R with I ⊆ p it already holds that
p = m.
iii) It holds that dim R/I = 0.
iv) The ring R/I is artinian.

An ideal satisfying any of the above conditions is called an ideal of

definition.
Proof of Lemma 4.17.
Claim 1. Let (R, m) be a local ring such that there is a k ≥ 1 with mk = 0.
Then Spec R = {m}.
For all p ∈ Spec R prime we have p ⊆ m. Now let b ∈ m. Then bk = 0 ∈ p
and so b ∈ p.

Applying this to R/I proves i) =⇒ ii). Now ii) =⇒ iii) is just the definition
of dim R/I.
iii) =⇒ iv): R/I is still a noetherian ring and dim R/I = 0 is equivalent
to all prime ideals of R/I being already maximal. The claim now follows
from Proposition 4.11.
iv) =⇒ i): By Lemma 4.16 we have that Jac R/I is nilpotent. But since
Jac R/I = m, this just means that there is a n ≥ 1 with mn ⊆ I. 
Theorem 4.18. Let (R, m) be a local noetherian ring.
i) If I = ha1 , . . . , al i is an ideal of definition, then dim R ≤ l.
ii) Assume dim R = d. Then there is an ideal of definition which is
generated by d elements.
 4.1. DIMENSION THEORY OF NOETHERIAN RINGS 69

Corollary 4.19. Let R be a noetherian ring, a1 , . . . , al ∈ R and p ∈ Spec R

be a minimal prime ideal of ha1 , . . . , al i. Then dim Rp ≤ l.
Proof. After localising at p the ring Rp is a noetherian local ring. Now
the ideal ha1 , . . . , al i is an ideal of definition, by Lemma 4.17 ii). 
Corollary 4.20. Let A be a k-algebra of finite type and a domain. For
an element 0 6= x ∈ A, let p be a minimal prime ideal of x. Then
dim A/p = dim A − 1.
Proof. By Corollary 4.19 we have dim Ap ≤ 1. Now in the case
dim Ap = 0 this means that p is a minimal prime ideal of A and hence
p = h0i (as A is an integral domain). But then x = 0, contradicting the
assumption x 6= 0.
So dim Ap = 1. Now by Lemma 3.12 and Corollary 3.16 we have
dim A = dim A/p + dim Ap and hence
dim A/p = dim A − dim Ap = dim A − 1.


End of Lecture 16
Lemma 4.21. Let R be a noetherian ring and I ⊆ R an ideal. Then there
are only finitely many prime ideals q ∈ Spec R which are minimal over I.
Proof. This will be on Exercise Sheet 10. 
Proof of Theorem 4.18. We prove both claims by induction, on l
and d respectivley.
i) The case l = 0 is trivial, and the case l = 1 is precisley Theorem 4.15
(Note that m is minimal over hai and that dim R = ht m holds for
a local ring, Lemma 3.11). So assume the claim holds for ideals of
definition in local rings which are generated by l − 1 elements.
Let q ∈ Spec R be a prime ideal such that q ( m and that there
is no prime ideal between q and m. So I 6⊆ q (since m is minimal
over I) and hence we can assume a1 ∈ / q.
Consider now the ideal ha1 i+q. Then this is an ideal of definition,
by the maximality of q. By Lemma 4.17 there is a n ≥ 1 such that
mn ⊆ ha1 i + q and in particular g2 , . . . , gl ∈ q, c2 , . . . , cl ∈ R such
that ani = ci a1 + gi for all i ≥ 2.
Claim 1. The ideal hg2 , . . . , gl , a1 i is an ideal of definition.
r r
P I ⊆ ha1 , g2 , . . . , gl i: An element x ∈ I is of
Set r := ln. Then
the form x = ν cν ai1 ,ν . . . air ,ν . Now in each of the summands,
each of the ai appears with a power ≥ n, and so the claim
follows from the above observation.
Now since I is an ideal of definition, there is an s > 0 such
that ms ⊆ I and so in total mrs ⊆ I r ⊆ ha1 , g2 , . . . , gl i follows,
which implies that ha1 , g2 , . . . , gl i is an ideal of definition.

Consider now the quotient R/hg2 , . . . , gl i, in which q := q/hg2 , . . . , gl i

and m := m/hg2 , . . . , gl i are prime ideals. By Claim 1 we ge get that
70 4. NOETHERIAN RINGS AND MODULES

a1 ∈
/ q and hence that m is minimal over a1 . Theorem 4.15 now
implies that dim (R/hg2 , . . . , gl i)m ≤ 1.
We now have
dim (R/hg2 , . . . , gl i)m = dim Rm /hg2 , . . . , gl i and
dim (R/hg2 , . . . , gl i)q = dim Rq /hg2 , . . . , gl i,
and since q ( m this implies dim Rq /hg2 , . . . , gl i = 0. So by
Lemma 4.17, hg2 , . . . , gl i is an ideal of definition in Rq . By the
induction hypothesis, we get dim Rq ≤ l − 1.
Let now p0 ( . . . ( pd be a chain in R such that m = pd and
that there is no prime ideal between pd−1 and m. We then have
dim Rpd−1 ≥ d − 1 and by the above reasoning dim Rpd−1 ≤ l − 1.
So d ≤ l, which proves i).
ii) Let q ∈ Spec R be prime with ht q = d − 1. Then by the induction
hypothesis, there are b1 , . . . , bd−1 ∈ Rq such that hb1 , . . . , bd−1 i is
an ideal of definition in Rq . Now bi = ai /si for some ai ∈ R and
si ∈ R \ q and so hb1 , . . . , bd−1 i = ha1 /1, . . . , ad−1 /1i. Hence in R,
q is minimal over I := ha1 , . . . , ad−1 i. By Lemma 4.21, there are
only finitely many prime ideals q1 , . . . , qr which are minimal over
ha1 , . . . , ad−1 i.
Claim 2. We have m 6⊆ q1 ∪ . . . ∪ qr .
If m ⊆ q1 ∪ . . . ∪ qr then by prime avoidance (Lemmab 3.H)
there is an i such that m ⊆ qi and hence m = qi . So m is a
minimal prime ideal of ha1 , . . . , ad−1 i and by i), this would imply
dim R ≤ d − 1.

Let now ad ∈ m \ (q1 ∪ . . . ∪ qr ). Then ha1 , . . . , ad−1 , ad i is an ideal

of definition of R, since otherwise, there would be a prime ideal
p ∈ Spec R with ha1 , . . . , ad i ⊆ p ( m. But this would give a chain
q1 ( pm for an 1 ≤ i ≤ r, contradicting ht qi ≤ d − 1.


4.2. Primary Decomposition in Noetherian Rings

Definition 4.22. Let R be a ring and I ( R a proper ideal. We say I is a
primary ideal if for all a, b ∈ R with ab ∈ I it already holds that a ∈ I or
bn ∈ I, for a n > 0.
Example 4.23.
i) Prime ideals are in particular primary ideals.
ii) Let R be a principal ideal domain. Then an ideal I is primary if
and only if I = pn holds, for a n > 0 and a prime ideal p ∈ Spec R.

√
Lemma 4.24. Let I be a primary ideal. Then I is a prime ideal.
√ n ∈ I. So
Proof. Let ab ∈ I. Then there is a n > 0 such that (ab)
√ √
n nm
√ I is primary, a ∈ I or b ∈ I follows, and hence a ∈ I or b ∈ I.
since
So I is indeed a prime ideal. 
 4.2. PRIMARY DECOMPOSITION IN NOETHERIAN RINGS 71

Remark 4.25. Since √ \

I= p,
p∈Spec R
I⊆p
√
we have that p := I is the smallest prime ideal which contains I. For this
p, we say that I is p-primary.
Lemma 4.26. I ( R be a proper ideal. Then I is a primary ideal if and
only if every zero-divisor of R/I is nilpotent.
Proof. Let ab = 0. Then ab ∈ I and hence a ∈ I or bn ∈ I, which
n
implies a = 0 or b = 0. This works in the other direction as well. 
Lemma 4.27. Let J ⊆ I ( R be ideals. Then I is primary if and only if
I/J is primary in R/J.
Lemma 4.28. Let R be a ring. Then the following are equivalent:
i) R has only one prime ideal.
ii) p
Every element in R is either a unit or nilpotent.
iii) h0i is a maximal ideal.

Proof. This is on Exercise Sheet 10. 

Lemma √4.29. Let m be a maximal ideal and I ( R an ideal such that
i) I = m; or
ii) mn ⊆ I ⊆ m holds for a n > 0.
Then I is m-primary.
Proof. p
i) In R/I, the ideal h0i is maximal. So the claim follow from
Lemma 4.28. √ √
ii) The√square root is monotonical increasing, hence p = pn ⊆ p = p,
so I = p follows. Now this is just i).

Definition 4.30. Let I ( R be an ideal. A primary decomposition of I is
a finite set of primary ideal I1 , . . . , Ir such that I = I1 ∩ . . . ∩ Ir holds.
Theorem 4.31. Let R be a noetherian ring. Then every proper ideal I ( R
has a primary decomposition.
Proof. Assume there are some ideals that do not have a primary de-
composition. Since R is noetherian, there is an ideal I which is maximal
with this property. Then in the ring R0 := R/I, the zero ideal h0i is the only
ideal which does not have a primary decomposition (Lemma 4.26).
Assume that h0i is not primary in R0 . Then there are a, b ∈ R0 such that
ab = 0, but a 6= 0 and bn 6= 0 for all n ≥ 1. Consider now the ascending
chain
Ann b ⊆ Ann b2 ⊆ Ann b3 ⊆ . . . ,
which terminates, since R/I is noetherian. So there is a n ≥ 1 such that
Ann bn = Ann bn+1 .
72 4. NOETHERIAN RINGS AND MODULES

Claim 1. For this n, hai ∩ hbn i = 0 holds.

Let x ∈ hai ∩ hbn i. Then x = ca = dbn for some c, d ∈ R0 holds.
Now by assumption on a, b, we have 0 = cab = dbn+1 , and hence
d ∈ Ann bn+1 = Ann bn . So x = dbn = 0 follows.

Since hai, hbn i =

6 0, both of them have a primary decomposition, and hence
n
hai ∩ hb i has too. But this contradicts Claim 1.
So h0i is primary. But this contradicts the original assumption that h0i
does not have a primary decomposition. 

Remark 4.32.
i) Let A(X) be the coordinate ring of an affine variety X. Let
I ( A(X) be an ideal and I = I1 ∩ . . . ∩ Ir a primary decom-
position. Then
V (I) = V (I1 ) ∪ . . . ∪ V (Ir )
p  p 
=V I1 ∪ . . . ∪ V Ir .

So the primary decomposition of I induces a decomposition of V (I)

into irreducible components.
ii) Let I = h(x − a1 )k1 . . . (x − an )kn i. Then V (I) = {a1 , . . . , an } and
a primary decomposition of I is given by
(x − a1 )k1 ∩ . . . ∩ (x − an )kn .

End of Lecture 17
Lemma 4.33. Let p ∈ Spec R be a prime ideal and I1 , I2 ⊆ R two p-
primary ideals. Then the intersection I1 ∩ I2 is p-primary too.
√ √ √
Proof. It holds that I1 ∩ I2 = I1 ∩ I2 = p.
Furthermore, I1 ∩ I2 is a primary ideal: Let ab ∈ I1 ∩ I2 . Then a ∈ I1
or b ∈ p and a ∈ I2 or b ∈ p. So a ∈ I1 ∩ I2 or b ∈ p, showing that I1 ∩ I2 is
indeed primary. 

Definition
√ 4.34. Let I = I1 ∩ . . . ∩ Ir be a primary decomposition, with
pi := Ii . We say this decomposition is minimal if the following two
conditions are satisfied:
i) none of the Ii is redundant: for all 1 ≤ i ≤ r it holds that
\
Ij 6⊆ Ii .
i6=j

ii) The pi are pairwise different: pi 6= pj holds for i 6= j.

Proposition 4.35. If an ideal I ⊆ R in an (arbitrary) ring R has a primary

decomposition, then I has also a minmal primary decomposition.
Proof. Part i) is clear, and part ii) follows from Lemma 4.33. 
 4.2. PRIMARY DECOMPOSITION IN NOETHERIAN RINGS 73

Notation 4.36. Let N ⊆ M be a submodule and m ∈ M . We write

N : m := {a ∈ R | am ∈ N } .
Lemma 4.37. In the above case, N : m is an ideal of R.
Proof. This is immediate. 
Lemma 4.38. Let I be√a p-primary ideal. Then for all a ∈ R it holds that
√
I : a = R if a ∈ R and I : a = p if a ∈ / I.
√ √
Proof. If a ∈ I, then I : a = R and hence I : a = R = R.
In the other case, let b ∈ I : a. Since I is primary, b ∈ p follows. Now
I ⊆ I : a ⊆ p and hence
√ √ √
p = I ⊆ I : a ⊆ p,
since the square root is monotonical. 
Definition 4.39. Let I ⊆ R be an ideal.
√ p ∈ Spec R is associated to I if there is an a ∈ R such
i) A prime ideal
that p = I : a. The set of prime ideals associated to I is denoted
by Ass(I).
ii) The inclusion-minimal prime ideals in Ass(I) are called isolated
prime ideals of I, all others embedded prime ideals of I.

Proposition 4.40.
√ Let I = I1 ∩. . .∩Ir be a minimal primary decomposition
of I, with pi := Ii . Then {p1 , . . . , pl } = Ass(I). So in particular, the number
of primary ideals in the decomposition does not depend on the decomposition.
Proof. We first show pi ∈ Ass(I): Since T the primary decomposition is
minmal, there is an a ∈ R such that a ∈ i6=j Ij and a ∈ / Ii for an 1 ≤ i ≤ r.
Now I : a = (I1 : a) ∩ . . . ∩ (Ir : a) and hence
√ p
I : a = (I1 : a) ∩ . . . ∩ (Ir : a)
p p
= I1 : a ∩ . . . Ir : a
4.38 p
= R ∩ . . . ∩ R ∩ Ii : a ∩ R ∩ . . . ∩ R
4.38 p
= Ii : a = pi ,
where we used Lemma 4.38 in the marked equalities. So the inclusion ⊆ “
”
follows. √
For the other direction, let p ∈ Ass(I), so there is an a ∈ R with I : a.
Now
√
p= I:a
p p
= I1 : a ∩ . . . ∩ Ir : a.
√
Now by prime avoidance (Lemmab 3.H) we get p = Ii : a for an 1 ≤ i ≤ r,
and hence (again by Lemma 4.33) p = pi (since p = R is not possible.) 
Proposition 4.41. Let R be a noetherian ring and I ⊆ R an ideal.
i) The isolated prime ideal of I are precisley the minimal prime ideals
over I.
74 4. NOETHERIAN RINGS AND MODULES

ii) There are only finitely many minimal prime ideals over I.

Proof. This is on Exercise Sheet 10. 

Lemma 4.42. Let S ⊆ R be multiplicative closed and I a p-primary ideal.
Denote by ϕ : R → S −1 R the canonical map into the localization.
i) If S ∩ p 6= ∅, then IS −1 R = S −1 R.
ii) If S ∩ p = ∅, then (IS −1 R) ∩ R = I and IS −1 R is pS −1 R-primary.

√
Proof. If S ∩ p 6= ∅, then there is a s ∈ S with s ∈ p = I. So sn ∈ I
for a n ≥ 1. Now
1 sn
= n ∈ IS −1 R
1 s
and hence (IS −1 R) ∩ R = R.
Assume now S ∩ p = ∅ and let a ∈ (S −1 R) ∩ R. Then a/1 ∈ IS −1 R and
hence there are q ∈ I, s, n ∈ S such that n(q − as) = 0. Now ans = nq ∈ I.
Since I is p-primary, this implies a ∈ I or ns ∈ p, which in this case means
a ∈ I. 
Proposition 4.43.
√ Let I = I1 ∩. . .∩Ir be a minimal primary decomposition
of I, with pi := Ii . If pi is minimal over I, then (IRpi )∩R = Ii . In particular,
the corresponding Ii do not depend on the decomposition.
Proof. Let S ⊆ R be a multiplicative set, then
(I1 ∩ . . . ∩ Ir )S −1 R = I1 S −1 R ∩ . . . ∩ Ir S −1 R ,

and hence
r
\
−1
Ii S −1 R .



IS R =
i=1
Set now S := R \ pi , such that pi is minimal over I. Then S ∩ pi = ∅ and
S ∩ pj 6= ∅ for i 6= j, since pj 6⊆ pi . Hence by Lemma 4.42 we get
(IRpi ) ∩ R = (Ii Rpi ) ∩ R.

Remark 4.44. It is possible to defined primary decomposition in the more
general context of modules:
i) Let M be an R-module and p ∈ Spec R. We say that p is associated
to M if there is an m ∈ M such that p = Ann m (Note that this
does not coincide with the definition of an associated prime ideal
for an ideal I ⊆ R, regarded as an R-module).
ii) We say that a submodule N ⊆ M is primary if it has an associated
primary ideal. It can be shown that every proper submodule N has
a decomposition N = N1 ∩ . . . ∩ Nr into primary submodules (if R
is noetherian).
iii) The uniqueness results are similar to the ones for ideals. (ToDo: do
this in more detail).
 CHAPTER 5

Regular Rings

Remarkb 5.A. Let (R, m) be a regular noetherian ring. Then R has finite
Krull-dimension.
Proof. The maximal ideal m is an ideal of definition. Since R is noether-
ian, m is finitely generated, and hence dim R ≤ number of generators of m,
by Theorem 4.18. 

Notation. Let (R, m) be a local ring. If not otherwise mentioned, we

denote by k := R/m the fraction field of m.
Lemma 5.1. Let (R, m) be a local noetherian ring with d = dim R.
i) It holds that d ≤ dimk m/m.
ii) We have d = dimk m/m2 if and only if m is generated by d elements.

Proof. This is on exercise sheet 9 (Hint: Use Nakayama.). 

Definition 5.2.
i) Let (R, m) be a local noetherian ring. We say R is regular if
d = dimk m/m2 holds.
ii) We say a noetherian ring R is regular if all localization Rp with
p ∈ Spec R are regular in the sense of i).
iii) Let X be a variety. We say a point a ∈ X is regular if A(X)I(a) is
a regular local ring.

Remarkb 5.B.
i) Note that regular rings are by definition noetherian.
ii) It is not clear that the definitions of regular rings are consistent
(i.e. that for a regular local ring (in the sense of i)) the dimension
equality is satisfied for all localizations at prime ideals). But this
seems to be the case ([Sta19, 00NN]) or [Fra18a, Page 33, Cor. 1]

End of Lecture 18
Lemmab 5.C. Let (R, m) be a local noetherian ring. Then R is a field if
and only if dimk m/m2 = 0.
Proof. The one direction is clear. If, on the other hand, dimk m/m2 = 0,
then this is equivalent to m = m2 . By Nakayama (Lemma 2.24), m = 0
follows and hence R is a field. 

75
76 5. REGULAR RINGS

Lemmab 5.D. Let (R, m) be a local noetherian ring and f ∈ m. Then

dim R/hf i ≤ dim R − 1. If f is not contained in any of the minimal prime
ideals of R then equality holds.
Proof. Let x1 , . . . , xd be elements in R/hf i such that hx1 , . . . , xd i is an
ideal of definition and d = dim R/hf i (These exist, by Theorem 4.18, ii)).
Then hf, x1 . . . , xd i is an ideal of definition of R, and hence dim R ≤ dim R/hf i+1
(by Theorem 4.18, i)).
If f is not contained in any minimal prime ideal of R, then every chain
in R/hf i can be lifted to chain of prime ideals which is at least one prime
ideal away from being maximal, and hence equality follows. 
Lemmab 5.E. Let (R, m) be a regular local ring and f ∈ m \ h0i.
i) Set R := R/hf i, m := m/hf i and k := R/m. Then
(
2 dim R, if f ∈ m2
dimk m/m = .
dim R − 1, if f ∈ / m2
ii) Assume that dim R = dim R − 1. Then R is regular if and only if
f∈ / m2 .
/ m2 , then dim R = dim R − 1 and R is regular.
iii) If f ∈

Proof. The canonical short-exact sequence

0 hf i m m 0,
where m := m/hf i, now induces the following big commutative square:
0 0 0

0 hf i ∩ m2 m2 m2 0

0 hf i m m 0

hf i/ hf i ∩ m2 m/m2 m/m2


0 0

0 0 0
Now all three columns and the two upper rows are exact, and hence the
lower one is too (by the 9-lemma). So it is in particular exact as sequence of
k-vector spaces and hence
dimk m/m2 = dimk hf i/ hf i ∩ m2 + dimk m/m2

holds (Note that m/m2 is finite-dimensional, since R is noetherian. Further-

more, R/m ∼ = R/m, by the third isomorphism theorem.).
Since R is regular, we have dim R = dimk m/m2 and hence
dimk m/m2 = dimk m/m2 − dimk hf i/ hf i ∩ m2

= dim R − dimk hf i/ hf i ∩ m2 .

As dimk hf i/ hf i ∩ m2 ≤ 1 and dimk hf i/ hf i ∩ m2 = 0 if and only if

f ∈ m2 , then dim R = dim R − 1 implies that m/m2 is regular if and only if

f∈/ m2 . This shows i) and ii).
 5.1. VALUATION RINGS 77

For iii), note that dim R ≤ dimk m/m2 = dim R − 1, by i) and Lemma 5.1.
Furthermore, by Lemmab 5.D, we have dim R ≥ dim R − 1. Hence equality
follows, and R is regular by ii). 

Corollaryb 5.F. Let (R, m) be a local noetherian integral domain and

/ m2 .
f ∈ m \ h0i. Then (R/hf i, m) is a regular local ring if and only if f ∈
Proof. By Lemmab 5.D, dim R/hf i = dim R − 1 holds (since f =
6 0 and
MinSpec R = {h0i}). The claim now follows directly from Lemmab 5.D. 

Proposition 5.3. Every regular local ring is an integral domain.

Proof. We will do this by induction on n := dim R. The case n = 0 is
Lemmab 5.C, which applies since R is regular.
In the general case n > 0, we will show that h0i is prime: Denote by
p1 , . . . , pr the minmal prime ideals over h0i.
Claim 1. We have m 6⊆ m2 ∪ p1 ∪ . . . ∪ pr .
If m ⊆ m2 ∪ . . . ∪ pr , then by Prime Avoidance (Lemmab 3.H), we have
m ⊆ m2 or m ⊆ pi for an 1 ≤ i ≤ r. If m2 = m, then dim R = 0 follows,
since R is regular.
So assume m = pi . Then already m = pj for all 1 ≤ j ≤ r follows,
and hence MinSpec R = MaxSpec R = m follows and hence dim R = 0.

/ m2 , p1 , . . . , pr . By Lemmab 5.E iii), we have that

So there is a a ∈ m with a ∈
R := R/hai is regular with dim R = dim R − 1. The induction hypothesis
now implies that hai is prime and hence there is an 1 ≤ i ≤ r such that
pi ⊆ hai. As a ∈/ pi and a ∈ m, we have mpi = pi , which implies pi = h0i (by
Nakayama, Lemma 2.24). 

5.1. Valuation Rings

Lemma 5.4. Let (R, m) be a 1-dimensional regular local ring. Then:
i) The maximal ideal m is a primary ideal.
ii) For every non-zero a ∈ R there is a unique n ≥ 1 such that hai = mn .

Proof.
i) Since dimk m/m2 = 1, m is generated by one element (Lemma 5.1,ii)).
ii) Since R is an integral domain (Proposition 5.3), h0i is the only mini-
mal prime ideal and R being 1-dimensional implies Spec R = {h0i, m}.
So for every non-zero a ∈ R, hai is an ideal of definition, and hence
there is a minimal n ≥ 1 such that mn ⊆ hai (Lemma 4.17). By i),
m = hti for a t ∈ R and hence there is a b ∈ R such that tn = ba. If
b ∈ m, then there is a b0 such that b = b0 t and hence tn = b0 ta which
would imply tn−1 = b0 a, contradicting the minimality of n. So b is
a unit, and hence mn = hai.


Definition 5.5.
78 5. REGULAR RINGS

i) A totally ordered group is an abelian group (G, +) with a total order

≤ such that for all pairs m ≤ n and k ∈ G already m + k ≤ n + k
follows.
ii) Let G be a totally ordered group. We extend the ordering and group
strucutre on G to the set G∪{∞} by a ≤ ∞ and a+∞ := ∞+a := ∞
for all a ∈ G.
iii) Let K be a field and G a totally ordered group. A valution on K is a
group homomorphism ν : K × → G such that ν(a+b) ≥ min {ν(a), ν(b)}
if a + b 6= 0. We extend ν to K be setting ν(0) := ∞.
iv) Let ν : K × → G be a valuation on K. Then Rν := {a ∈ K | ν(a) ≥ 0}
is called the valuation ring of ν.
v) The subgroup ν (K × ) ⊆ G is called the valuation subgroup of ν.

Proof. The valuation ring Rν is indeed a ring: We have ν(0) = ∞ ≥ 0

(by definition), ν(1) = 0 (since ν is a group homomorphism) and for all
a, b ∈ K × it holds that ν(a + b) ≥ min {0, 0} = 0 and ν(ab) = 0 + 0 = 0. 
Notation. More generally, we say that a ring R is a valuation ring if there
is a valuation ν : K → G such that R = Rν .
Example 5.6.
i) On every field, there is the trivial valuation K × → {0}.
ii) Let R be a factorial ring, set K := Quot R and let p be a prime
element of R. Now for every element b ∈ R, there is a unique
maximal m ≥ 0 such that b = apm . So every element b0 ∈ k has a
unique decomposition of the form b0 = a0 pn with n ∈ Z, such that
a0 is quotient of two elements from R that are both not divisible by
p. Define a map ν : K × → Z, apn 7→ n. Then this is a valuation.
The valuation ring is given by
Rν = apn a ∈× K, n ≥ 0, p 6| a ∪ {0} = Rhpi ,


and the value group is given by Z.

iii) Let K be a field and consider the field
( )
X
n
L := an t an ∈ K, {n ∈ Z | an 6= 0} has a lower bound .
n∈Z

Then
L× −→ Z
X
an tn 7−→ min {n ∈ Z | an 6= 0}
is a valuation on L, with valuation ring K[[t]]. Note that this is a
special case of ii), with R = K[[t]].

End of Lecture 19
Lemma 5.7. Let R = Rν be a valuation ring.
i) The ring R is an integral domain.
ii) For all a ∈ K × it holds that a ∈ R or a−1 ∈ R.
 5.1. VALUATION RINGS 79

iii) For all a, b ∈ R it holds that ν(a) ≤ ν(b) if and only if b ∈ hai.
iv) The group of units of R is given by R× = ker ν.
v) It holds that R is a local ring, with unique maximal ideal
m = {a ∈ R | ν(a) > 0} .
vi) The ring R is normal.

Proof.
i) Since R is a subring of a field, it is an integral domain.
ii) Since ν is a group homomorphism, we have
ν(a) + ν(a−1 ) = ν(1) = 0.
So ν(a) ≥ 0 or ν(a−1 ) ≥ 0.
iii) Assume that a = 0. Then ν(b) ≥ ∞ if and only if ν(b) = ∞ if and
only if b = 0 if and only if b ∈ h0i, since R is an integral domain.
So assume a 6= 0. Then ν(a) ≤ ν(b) if and only if ν(b/a) ≥ 0 if
and only if b/a ∈ R if and only if b ∈ hai.
iv) Let a ∈ R \ {0}. Then a ∈ R× if and only if a−1 ∈ R if and only if
ν(a−1 ) = −ν(a) ≥ 0 if and only if ν(a) = 0.
v) Since ν is a group homomorphism, m is an ideal. Let now I ⊆ R be
an ideal such that I 6⊆ m. Then I contains an element a ∈ R such
that ν(a) = 0, and hence I = h1i, by iv).
vi) The fraction field of R is given by K. Let now a ∈ K × be integral
over R. So there are cn−1 , . . . , c0 such that
an + cn−1 an−1 + . . . + c0 = 0.
/ R. Then a−1 ∈ R by ii). Hence
Assume a ∈
a = − cn−1 + cn−2 a−1 + . . . + an−1 c0 ,

and all summands are in R. But this is a contradiction.


Proposition 5.8. For a ring R, the following are equivalent:
i) R is a valuation ring.
ii) R is an integral domain such that for all a ∈ (Quot R)× it holds
that a ∈ R or a−1 ∈ R.

Proof. The direction i) =⇒ ii) follows from Lemma 5.7. For the other
direction, set K := Quot R, G := K × /R× (as quotient of abelian groups) and
denote by a the image of a ∈ K × in G. We make G into a totally ordered
abelian group by setting:
a ≤ b if and only if b/a ∈ R.
Then this is well-defined, since for units c1 , c2 ∈ R× , we have a/b ∈ R if
and only if c1 /c2 · a/b ∈ R. It is antisymmetric, since a/b ∈ R and b/a ∈ R
implies that there is a unit c ∈ R× such that a/b = c, transitive and by
assumption a ≤ b or b ≤ a always holds. The relation is also compatible with
the group structure on G, since b/a ∈ R implies bc/ac ∈ R for all c ∈ K × .
80 5. REGULAR RINGS

Denote by ν : K × → G the quotient map. Then ν(a) ≤ ν(b) if and only

if b/a ∈ R. Then ν is indeed a valuation on K × : By construction, ν is a
group homomorphism, and if a/b ∈ R then (a + b)/b = a/b + 1 ∈ R, and
hence ν(a + b) ≥ ν(b) follows. 
Remark 5.9. Let Rν ⊆ K be a valuation ring for the valuation ν : K × → G.
Then this already determines ν in the following sense: It holds that K = Quot R,
ν (K × ) = K × /R× and ν(a) ≤ ν(b) if and only if ν(b/a) ≤ 0 if and only
b/a ∈ R.

5.2. Discrete Valuation Rings

Proposition 5.10. Let R be a valuation ring with maximal ideal m. Then
the following are equivalent:
i) R is noetherian but not a field.
ii) R is a principal ideal domain but not a field.
iii) The valuation group of R is given by Z.

Definition 5.11. In this case, we say that R is a discrete valuation ring

Proof. This will be added in the near future. 
Proposition 5.12. Let R be a local noetherian ring. Then the following
are equivalent:
i) R is a discrete valuation ring.
ii) R is a prinicipal ideal domain.
iii) R is a one-dimensional factorial ring.
iv) R is a one-dimensional normal integral domain.
v) R is a one-dimensional regular local ring.

Proof. This will be added in the near future. 

End of Lecture 20

5.3. Dedekind Rings

Definition 5.13. Let R be a one-dimensional noetherian integral domain
such that all localizations at prime ideals Rp are discrete valuation rings.
Then we say that R is a Dedekind domain.
Remarkb 5.G. By Proposition 5.12, this is well-defined.
Remark 5.14. A one-dimensional noetherian integral domain is a Dedekind
domain if and onyl if it is normal.
Proof. By Lemmab 2.V, being normal is a local property. By Proposi-
tion 5.12, a localization Rp of R is a discrete valuation ring if and only if it
is normal. 
Example 5.15.
i) The ring of integers Z, and more generally, every principal ideal
domain is a Dedekind domain.
 5.3. DEDEKIND RINGS 81

ii) Let X be a smooth curve. Then the coordinate ring A(X) is a

Dedekind domain.

Definitionb 5.H. Let Q ,− → K be a finite field extension. Then the integral

closure of Z in K is the ring of integers and denoted by OK .
Theorem 5.16. Let Q ,− → K be a finite field extension. Then the ring of
integers OK is a Dedekind domain.
Proof. Since OK is a subring of Q, it is an integral domain. By
definition of OK , the extension Z ,−
→ OK is integral and since dim Z = 1 it
follows that dim OK = 1 (by Proposition 3.13). Furthermore, every element
a ∈ Quot OK is integral over Z and so by definition already in OK . The
difficult part is to show that OK is noetherian – we first need two more
claims:
Claim 1. Let m ∈ Z be an integer. Then OK /mOK has only finitely many
elements.
Consider first the case m = p for a prime number p. Now R/pR is
a Fp -vector space, since OK is a Z-module. So it suffices to show
dimFp OK /pOK ≤ dimQ K: Let b1 , . . . , bn ∈ OK /pOK be linearly inde-
pendent. If there are λ1 , . . . , λn ∈ Q such that

λ1 b1 + . . . + λn bn = 0,

then (by factoring out the common denominator) we can assume that
λ1 , . . . , λn ∈ Z and that not all λi are divisible by p. But then in OK ,
we get
λ1 b1 + . . . + λn bn = 0,
contradicting that b1 , . . . , bn are linearly independent. So each linearly
independet subsest of OK /mOK lifts to a linearly independet subset of
OK ⊆ K and hence dimFp OK /mOK ≤ dimQ OK ≤ dimQ K.
In the general case, note first that for any ring R a short-exact
sequence of R-modules

0 M0 M M 00 0

the module M is finite if and only if both M 0 and M 00 are finite. Fur-
thermore, for a Z-module M and all m1 , m2 ∈ Z the sequence

0 M/m1 M/(m1 m2 )M M/m2 M 0

is short-exact.
So if m = pk11 , . . . , pknn for prime number p1 , . . . , pn ∈ Z then the
above observations show that for each prime number, OK /pki i is finite
and hence OK /(pn1 1 . . . pknn )OK is finite too.

Claim 2. Let I ⊆ OK be a non-zero ideal. Then there is a non-zero m ∈ Z

such that m ∈ I.
82 5. REGULAR RINGS

Assume to the contrary that there is no such m, i.e. I ∩ Z = {0}. Now

the morphism Z/I ∩ Z ,−
→ OK /I is integral which implies
dim OK /I = dim Z/(I ∩ Z) = dim Z = 1.
But since OK is an integral domain and I 6= 0, dim OK /I < dim OK holds.
This is a contradiction (we noted earlier that dim OK = dim Z = 1).

We now show that OK is noetherian by showing that every ideal I ⊆ OK is

finitely generated: Let I ⊆ OK be an ideal. By Claim 2, there is a m ∈ I ∩ Z.
Now I/hmi is a submodule of OK /hmi, and by Claim 1 I/hmi is finite, so
in particular finitely generated. By Lemmab 2.M, this already implies that I
is finitely generated. 
Theorem 5.17. Let R be a Dedekind domain.
i) Let p ∈ MaxSpec R be a maximal ideal and I ⊆ R an ideal. Then I
is p-primary if and only if I = pk for a unique k ≥ 0.
ii) Every ideal I has a decomposition of the form
I = pk11 · . . . · pknn
where k1 , . . . , kn ≥ 1 and p1 , . . . , pn = Ass(I). This decomposition
is unique up to permutation.

Proof.
i) The direction “ ⇐= “ is true in any ring. For the other direction,
we note that IRp 6= 0 and that (by definition) Rp is a discrete
valuation ring. So by Proposition 5.12 and Lemma 5.4 there is a
unique k ≥ 0 such that IRp = (pRp )k = pk Rp . Now by Lemma 4.42
I = pk follows.
ii) Since R is noetherian, there is a minimal primary decomposition of
I such that I = I1 ∩ . . . In and Ass(I) = {p1 , . . . , pn }. By part i),
there are ki ≥ 1 such that Ii = pki 1 . Since the pi are maximal and
coprime, we get
pk11 ∩ . . . ∩ pknn = pk11 · . . . · pknn .


Lemma 5.18. Let R be a Dedekind domain.
i) For all collections p1 , . . . , pn of maximal ideals and natural numbers
k1 , . . . , kn , l1 , . . . , ln it holds that
pk11 · . . . · pknn ⊆ pl11 · . . . · plnn if and only if li ≤ ki for all i.
ii) Let a ∈ R be non-zero. Then there is a decomposition of the form
ν (a)
hai = p11 · . . . · pνnn (a)
where {p1 , . . . , pn } = Ass(hai) and νi : Rpi → Quot(Rpi ) is the
valuation on the localization.

Proof.
 5.4. THE CLASS GROUP 83

i) We first need the following basic facts about localizations

Fact 1. Let R be any ring (note necessarily noetherian, ...).
a) Let I, J ⊆ R be ideals in an arbitrary ring R. Then I ⊆ J if
and only if IRm ⊆ JRm for all m ∈ MaxSpec R.
b) Let R → R0 be a ring homomorphism. Then
(IJ)R0 = (IR0 )(JR0 ).

Now
pk11 · . . . · pknn ⊆ pl11 · . . . · plnn
if and only if (pi Rpi )ki ⊆ (pi Rpi )li if and only if ki ≤ li (since Rpi is
a discrete valuation ring). The result now follows from Fact 1 a)
and b) (and the fact that pi Rpk = Rp for i 6= j).
ii) Let
ν (a)
hai = p11 · . . . · pνnn (a)
with {p1 , . . . , pn } = Ass(hai). Then
haiRp1 = (pi Rpi )ki ,
which is exactly νi (a).


5.4. The Class Group

Definition 5.19. Let R be an integral domain and set K := Quot R.
i) A fractional ideal is an R-submodule of K such that there is a
non-zero a ∈ R with aI ⊆ R.
ii) Let a1 , . . . , an ∈ K. We denote by
ha1 , . . . , an i := Ra1 + . . . + Ran
the fractional ideal which is generated by a1 , . . . , an . We say a
fractional ideal I is finitely generated if there are a1 , . . . , an such
that I = ha1 , . . . , an i. We say I is a principal fractional ideal if
there is a a ∈ K such that I = hai.

Example 5.20.
i) Every ideal of R is a fractional ideal in K.
ii) The fractional ideal generated by some elements a1 , . . . , an is indeed
a fractional ideal.

Lemma 5.21. Let I be a fractional ideal such that aI ⊆ R is finitely

generated. Then I is finitely generated as fractional ideal.
Definition 5.22. Let I, J ⊆ R be submodules of K.
i) Set ( n )
X
I · J := ai bi n ≥ 0, a ∈ I, b ∈ J ,
i=1
84 5. REGULAR RINGS

and
I :K J := {a ∈ K | aJ ⊆ I} .
ii) We say an R-submodule I ⊆ K is invertible if there is an R-
submodule J ⊆ K such that I · J = K.

Lemma 5.23. Let I, J ⊆ R be submodules of K.

i) Assume I · J = R. Then J = R :K I.
ii) If I = hai with a 6= 0 is a principal fractional ideal then I is
invertible.
iii) If I is invertible then I is a fractional ideal.

Proof.
i) Assume R = IJ. We now have
J ⊆ R :K I = (R :K I) (IJ) ⊆ RJ = J
and hence R :K I = J.
ii) If I = hai then hai · h1/ai = R.
P I · (R :K I) = R, so there are aiP
iii) Let I be invertible, i.e. ∈ I and
bi ∈ R :K I such that ai bi = 1. Let now b ∈ I then b = ai (bi b)
with bi b ∈ R. Let a be the product of the denominators of the ai ,
then ab ∈ R and hence aI ⊆ R. So I is a fractional ideal.


End of Lecture 21
Technical Remark. The proofs of the statements of this lecture will be
added (hopefully) int the beginnig of September.
Proposition 5.24. Let R be a Dedekind domain. Then every fractional
ideal is invertible.
Corollary 5.25. Let R be a Dedekind ring. Then the set of fractional
ideals becomes an group with the multiplication of fractional ideals as
binary operation and the ring R as unit. We denote this group by Div(R).
Furthermore, Div(R) is abelian.
Lemma 5.26. The set Prin(R) principal fractional ideals are a subgroup
of Div(R).
Definition 5.27. The quotient Div(R)/ Prin(R) is denoted by Cl(R) and
called the class group of R.
Remark 5.28. Here are two fun facts about the class group, which we are
not able to prove in this lecture:
i) For every abelian group G there is a Dedekind ring R such that
G∼ = Cl(R).
ii) If OK is the ring of integers of a number field, then Cl(OK ) is finite.
The class number of K is defined as |Cl(OK )|.
 5.5. MODULES OVER PIDS AND PROJECTIVE MODULES 85

Proposition 5.29. Let I be an invertible ideal in a Dedekind ring R. Then

I = pk11 · . . . · pknn
for distinct prime ideals p1 , . . . , pn and unique integers k1 , . . . , kn ∈ Z. This
representation is unique up to permutation of maximal ideals.
Corollary 5.30. If R is a Dedekind domain, then the group of fractional
ideals is the free abelian group generated by the maximal ideals.
Theorem 5.31. Let R be a Dedekind domain. Then the following are
equivalent:
i) R is a prinicipal ideal domain.
ii) R is factorial.
iii) The class group Cl(R) is trivial.

5.5. Modules over PIDs and Projective Modules

Lemma 5.32. Let R be a principal ideal domain and N a finitely generated
free R-module. Then every submodule M ⊆ N is free too and rg M ≤ rg N .
Remarkb 5.I. The statement remains true even if N is not finitely gener-
ated, although a different proof is needed.
Lemma 5.33. Let R be a principal ideal domain and M a finitely-generated
R-module. Then M is free if and only if it is torsion-free.
Lemma 5.34. Let f : M → N be a surjective R-linear map into a free
R-module N . Then M ∼
= ker f ⊕ N .
Corollary 5.35. Let M be a finitely generated R-module. Then there is
a decomposition M ∼
= M 0 ⊕ T (M ) with M 0 free.
Definition 5.36. An R-module P is projective if for all R-linear maps
P → M 00 and surjective R-linear maps M  M 00 there is an R-linear map
P → M such that the diagram
P
∃

M M 00
commutes.
Example 5.37. If P is a free R-module, then P is projective.
Definition 5.38. Let
f g
0 M0 M M 00 0
be a short-exact sequence of R-modules. A split of g is an R-linear map
σ : M 00 → M such that gσ = idM 00 . If there is a split of g, we say that g
splits or that the sequence is split-exact.
Lemma 5.39. Let
f g
0 M0 M M 00 0
86 5. REGULAR RINGS

∼
→ M 0 ⊕ M 00 such
be split-exact. Then there is a unique isomorphism h : M −
that the diagram
f g
0 M0 M M 00 0
∼

0 M0 M M 00 0
commutes.

End of Lecture 22
Technical Remark. Dr. Heidersdorf said in the lecture that the content of
the following is not relevant for the first exam. Moreover, he uses some pretty
advanced facts about modules. I am not sure if I will be able to add their
proofs in the future. The reader can find more on this topic in [Sta19, 05E3].
Lemma 5.40. Let P be an R-module. Then the following are equivalent:
i) P is projective.
ii) For every surjective map π : M  M 00 the induced map
π∗ : hom(P, M ) → hom(P, M 00 )
is surjective.
iii) For every surjective map π : F  M 00 with F free the induced map
π∗ : hom(P, M ) → hom(P, M 00 ) is surjective.
iv) P is a direct summand of a free R-module, i.e. there is an R-module
Q and a free R-module F such that F ∼ = P ⊕ Q.
v) For every surjective map g : M → P the induced sequence
g
0 ker g M P 0
splits.

Definition 5.41. Let M be an R-module.

i) We say M is locally finitely-generated if for all prime ideals p ∈ Spec R
there is an element f ∈ R \ p such that Mf is a finitely-generated
Rf -module.
ii) We say M is locally finitely-presented if for all prime ideals p ∈ Spec R
there is an element f ∈ R \ p such that Mf is a finitely-presented
Rf -module. We say M is locally free of finite rank if for all prime
ideals p ∈ Spec R there is an element f ∈ R \ p such that Mf is a
finitely-generated free Rf -module.
iii) We say M is locally free of rank n if for all prime ideals p ∈ Spec R
there is an element f ∈ R \ p such that Mf is isomorphic to Rfn .

Theorem 5.42. Let M be an R-module. Then the following are equivalent:

i) M is finitely-generated and projective.
ii) M is finitely-presented and Mm is a free Rm -module for all maximal
ideals m ∈ MaxSpec R.
iii) M is locally-free of finite rank.
 5.5. MODULES OVER PIDS AND PROJECTIVE MODULES 87

Definition 5.43. Let M be an R-module. We say M is invertible if there

is an R-module N such that M ⊗R N ∼
= R.
Remark 5.44. If R is an integral domain, then the notions of invertible
modules in the above sense and invertible R-modules in Quot R do not
necessarily coincide.
Lemma 5.45. Let M be an R-module.
i) If M is invertible, then M is already finitely-generated.
ii) If R is a local ring and M is invertible, then M is free of rank 1.

Lemma 5.46. Let M be an R-module. The following are equivalent:

i) M is invertible.
ii) M is finitely generated and Mm ∼
= Rm for all m ∈ MaxSpec R.
iii) R is locally free of rank 1.
In this case, M is already finitely-presented and for every R-module N with
M ⊗R N ∼ = R it already holds that N ∼ = hom(M, R).
Definitionb 5.J. The R-module hom(M, R) is the dual of M and is also
denoted by M ∨ .
Lemmab 5.K. If M is a finitely-presented R-module, N any R-module
and S ⊆ R a multiplicative subset then there is an isomorphism
∼
S −1 (homR (M, N )) −→ homS −1 R S −1 M, S −1 N

. So in particular, if M is a finitely-presented R-module, then

(homR (M, R)) ∼
m = homR (Mm , Rm )
m

for all maximal ideals m ∈ MaxSpec R.

Proof. This is [Wei94, 3.3.7]. 
Definition 5.47. Let R be an integral domain and M ⊆ Quot R a frac-
tional ideal. Then M is a local principal ideal if Mm is fractional principal
ideal in Quot Rm for all m ∈ MaxSpec R.
Lemma 5.48. Let R be an integral domain and M, N ⊆ Quot R fractional
ideals.
i) There is a surjective linear map π : M ⊗R N  M · N .
ii) If M is a local prinicipal ideal, then π is already an isomorphism.

Lemma 5.49. Let R be an integral domain and M an invertible fractional

ideal. Then M is finitely generated.
Lemma 5.50. Let R be an integral domain and M a fractional ideal. Then
M is invertible if and only if M is a non-zero fractional principal ideal.
Lemma 5.51. Let R be an integral domain and M a fractional ideal. Then
the following are equivalent:
i) M is an invertible fractional ideal.
ii) M is finitely generated a locally principal ideal.
88 5. REGULAR RINGS

Theorem 5.52. Let R be an intgral domain. Then the following are

equivalent:
i) M is an invertible fractional ideal.
ii) M is an invertible module.
iii) M is projective.

Definition 5.53. Let R be any ring. We define the Picard group Pic(R)
of R as the set of isomorphism classes of invertible R-modules, where the
multiplication is given by the tensor product.
Lemma 5.54. Let R be an integral domain. Then every invertible module
is isomorphic to a fractional ideal.
Theorem 5.55. If R is a Dedekind Domain, then Pic(R) is isomorphic to
the class group Cl(R).

End of Lecture 23

End of Algebra 1
 APPENDIX A

Prerequisites - Rings

A.1. Basics
We recall some basic facts and definitions about rings which can be found
in [Alu09, Chapter 3],[Str18, Sch19].
Proposition A.1 (Chinese Remainder Theorem). Let R 6= 0 be a ring,
and I1 , . . . , Ir ⊆ R ideals such that Ii + Ij = R for all i 6= j. Then there is a
surjective ring homomorphism
ϕ : R −→ R/I1 × . . . × R/Ir
r 7−→ (r, . . . , r) ,
and ker ϕ = I1 ∩ . . . ∩ Ir . In particular there is a ring isomorphism
∼
R/ (I1 ∩ . . . ∩ Ir ) R/I1 × . . . × R/Ir .
A.1.1. Formal Power Series. Under construction.

89
 APPENDIX B

Categories

This chapter is currently under construction. The reader can find the
necessary (and much more!) material e.g. in [Ste19].

B.1. General Categories and Functors

B.1.1. Yoneda-Lemma.
Lemma B.1 (Yoneda-Lemma).
i) Let F : C → Set be a functor and A ∈ C an object. Then for every
element u ∈ F (A), there is a natural transformation
ηu : C (A, −) −→ F
 f 
ηu (B) : A − →B 7−→ F (f )(u) (for all B ∈ C) .

The maps
{natural transformations C (A, −) → F } −→ F (A)
(η : C (A, −) → F ) 7−→ ηA (idA )
(ηu : C (A, −) → F ) ←−[ u
are mutually inverse and natural in F and A.
ii) Let F : C op → Set be a functor and A ∈ C an object. Then for
every element u ∈ F (A), there is a natural transformation
ηu : C (−, A) −→ F
 f 
ηu (B) : A − →B 7−→ F (f )(u) (for all B ∈ C) .

The maps
{natural transformations C (−, A) → F } −→ F (A)
(η : C (A, −) → F ) 7−→ ηA (idA )
(ηu : C (A, −) → F ) ←−[ u
are mutually inverse and natural in F and A.

Lemma B.2 (Yoneda-Embedding). Let C be a category.

i) The functor
C op −→ Fun (C, Set)
X 7−→ C(X, −)
is fully faithful.
90
 B.3. SOME HOMOLOGICAL ALGEBRA 91

ii) The functor

C −→ Fun (C op , Set)
Y 7−→ C(−, Y )

is fully faithful.

Definition B.3. Let F : C → Set be a functor. An object A ∈ C is a

representing object of F if there is a natural isomorphism

∼
C(A, −) −→ F.

B.2. Additive and Abelian Categories

B.3. Some Homological Algebra
B.3.1. Some Diagram Lemmas.
Lemma B.4 (5-Lemma). Let A be an abelian category. Let

0 X1 X2 X3 X4 X5 0
f1 f2 f3 f4 f5

0 Y1 Y2 Y3 Y4 Y5 0

be a commutative diagram in A with exact rows. Then

i) If f2 , f4 are epimorphisms and f5 is a monomorphism then f3 is an
epimorphisms too.
ii) If f2 , f4 are monomorphism and f1 is an epimorphisms then f3 is a
monomorphism.
iii) If f2 , f4 are isomorphisms, f1 an epimorphisms and f5 a monomor-
phism then f3 is an isomorphism.

Proof. Ommited. 

Lemma B.5 (Snake Lemma). Let A be an abelian category. Let

i p
X0 X X 00 0
f0 f f 00

0 Y0 j
Y q Y 00
92 B. CATEGORIES

be a commutative diaram in A with exact rows. Then there are induced

morphisms such that the following diagram commutes
ĩ p̃
ker f 0 ker f ker f 00

i p
X0 X X 00 0
f0 f f 00

0 Y0 j
Y q Y 00

j̃ q̃
coker f 0 coker f coker f 00
i) There is a morphism δ : ker f 00 → coker f 0 such that the sequence
ĩ p̃
ker f 0 ker f ker f 00

δ
j̃ q̃
coker f 0 coker f coker f 00
is exact.
ii) If i is a monomorphism, then ĩ is a monomorphism too.
iii) If q is an epimorphisms, then q̃ is an epimorphisms too.

Proof. Ommited. 
 APPENDIX C

Further Remarks - Modules

C.1. Projective Modules

Definition C.1. Let M be an R-module. We say M is a projective R-
module if for all epimorphisms f : X → Y and morphisms p : M → Y there
is a morphism f 0 : M → X such that the diagram
M
f0
p

X Y 0
commutes.
Example C.2.
i) Let M be a free R-module. Then M is projective: Let (ei ) be a
basis for M . For each epimorphisms f : X → Y and morphism
M → Y we can choose for a basis element ei a preimage under f ,
say xi . Then the assignment ei 7→ xi can be extended to an R-linear
map M → X, by the universal property of the free module.
ii) The Z-module Z/2Z is not projective: Consider the projection
Z  Z/2Z. Now any lift of the identity

Z/2Z
?

Z Z/2Z

would necessarily be the zero-morphism (a morphism f : Z/2Z → Z

has to satisfy 2f (1) = f (0) = 0) but then the diagram is far from
being commutative.

C.2. Tensor Products

Definition C.3. Let M, N, P be R-modules and f : M × M 0 → P a map.
We say h is R-bilinear if for all x ∈ M , y ∈ N the maps
f (x, −) : N → P, y 0 7→ f (x, y 0 )
f (−, y) : M → P, x0 7→ f (x0 , y)
are R-linear.
Proposition C.4 (Existence of Tensor Products 1). Let M, N be R-
modules.
93
94 C. FURTHER REMARKS - MODULES

i) Then there exists an R-module M ⊗R N and an R-bilinear map

g : M × N → M ⊗R N such that for all other R-bilinear maps
f : M × N → P there is a unique R-linear map f 0 : M ⊗R N → P
such that the diagram
f
M ×N P
d
∃!f 0
M ⊗R N

commutes.
ii) The tensor product is unique in the following sense: If there is
another R-module T with an R-bilinear map h : M × N → T
and the property that every R-bilinear map f : M × N → P
factors uniquely over T , then there exists a unique isomorphism of
∼
R-modules λ : M ⊗R N −→ T such that the diagram
g
M ×N M ⊗R N
∃!λ
h
T

commutes.

Proof. This is completely analogous to the proof for vector spaces,

which can be found in [Sch18, 16.2]. A proof where vector space“ is
”
replaced with module“ can be found in [Fra18b, Prop. 2.23]. 
”
There is another way of defining tensor products, c.f. [Bra16]:
Proposition C.5 (Existence of Tensor Products 2). Let M, N be R-
modules. Consider the bilinear-functor

bilinR (M, N ; −) : R-Mod → Set

which sends an R-module P to the set of R-bilinear maps M × N → P and

an R-linear map f : P → Q to the induced map

bilinR (M, N ; P ) −→ bilinR (M, N ; Q)

   f

h h
M ×N − →P 7−→ M ×N − →P − →Q .

Then there exists a representing object M ⊗R N for bilinR (M, N ; −), i.e.
a natural isomorphism bilinR (M, N ; −) → homR (M × N, −).
The notions of a tensor product from Proposition C.4 and Proposition C.5
are the same. We call the R-module M ⊗R N the tensor product of M and
N over R.
We will now use the Yoneda-Embedding (Lemma B.2) to show several
properties of the tensor product:
 C.2. TENSOR PRODUCTS 95

Lemma C.6. Let M, N, M 0 , N 0 be R-modules and let f : M → M 0 ,

g : N → N 0 be R-linear maps. Then there is a unique R-linear map
M ⊗R N −→ M 0 ⊗R N 0
a ⊗ b 7−→ f (a) ⊗ g(b).
Proof. The maps f, g induce a natural transformation
η : bilinR (M 0 , N 0 ; −) −→ bilinR (M, N ; −)
 
(f,g)
 
0 0 h 0 0 h
ηP : M × N − →P 7−→ M × N −−−→ M × N − →P ,

where the map (f, g) : M × N → M 0 × N 0 is given by (a, b) 7→ (f (a), g(b)).

This is indeed natural as the diagram
ηP
bilinR (M 0 , N 0 ; P ) bilinR (M, N ; P )
g◦− g◦−

bilinR (M 0 , N 0 ; P 0 ) ηP 0 bilinR (M, N ; P 0 )

g
− P 0.
commutes for all R-linear maps P →
By Proposition C.5 this corresponds to a natural transformation
homR (M 0 ⊗R N 0 , −) → homR (M ⊗R N, −).
Using the Yoneda embedding (Lemma B.2, i)), this corresponds to a unique
R-linear map M ⊗R N → M 0 ⊗R N 0 . 

Lemma C.7. Let M, N be R-modules. Then there is a R-linear isomor-

∼
phism M ⊗R N −→ N ⊗R M .
Proof. There is a natural isomorphism
η : bilinR (M × N, −) −→ bilinR (N × M, −)
   
(f,g) (g,f )
ηP : M × N −−−→ P 7−→ N × M −−−→ P .

So by the Yoneda embedding, this corresponds to a R-linear isomorphism

∼
M ⊗R N −→ N ⊗R M . 

Proposition C.8 (Tensor-Hom-Adjunction). Let N be an R-module.

i) Consider the assignment
F : R-Mod → R-Mod, M 7→ M ⊗R N
which sends an R-linear map f : M → M 0 to the induced R-linear
map F (f ) := f ⊗ idN : M ⊗R N → M 0 ⊗R N from Lemma C.6.
Then this defines an additive functor R-Mod → R-Mod.
ii) Denote by G the covariant hom-functor
G : R-Mod −→ R-Mod
P 7−→ homR (N, P )
   
h h∗
→ P0
P − 7−→ homR (N, P ) −→ homR (N, P 0 ) .
96 C. FURTHER REMARKS - MODULES

Then for all M, P ∈ R-Mod, there is well-defined map

φ : homR (M ⊗R N, P ) −→ homR (M, homR (N, P ))
f 7−→ (x 7→ (y 7→ f (x ⊗ y))) .
iii) The pair (F, G, φ) is an adjunction:
For all M, P ∈ R-Mod, the map φ from ii) is an isomorphism
of R-modules, natural in M and P in the sense that the diagrams

homR (M 0 ⊗R N, P ) homR (M 0 , homR (N, P ))

(f ⊗ id)∗ f∗

homR (M ⊗R N, P ) homR (M, homR (N, P ))

and
homR (M ⊗R N, P ) homR (M, homR (N, P ))
g∗ (g∗ )∗

homR (M ⊗R N, P 0 ) homR (M, homR (N, P 0 ))

commute for all R-linear maps f : M → M 0 , g : N → N 0 .

Proof. Consider the maps

bilinR (M, N ; P ) ←→ homR (M, homR (N, P ))
 f

M ×N − →P 7−→ (b 7→ f (−, b))
  ψ

(a × b 7→ ψ (b) (a)) ←−[ b 7→ M − →P .

Then these are well-defined natural bijections. So we get for all M, N, P ∈ R-Mod
a natural isomorphism
bilinR (M × N, P ) ∼
= homR (M, homR (N, P )) .
But since
bilinR (M × N, P ) ∼
= homR (M ⊗R N, P )
by the universal property of the tensor-product, the claim follows. 
Proposition C.9. Let N be an R-module and let
f g
0 M0 M M 00 0
be a short-exact sequence of R-modules. Then the sequence
f ⊗ id g ⊗ id
M0 M M 00 0
is again exact. So the tensor-product functor is right-exact.
Proof. This follows from the fact that the tensor product functor is
left-adjoint. 
Corollary C.10. Let M be an R-module and I ⊆ R an ideal. Then
R/I ⊗R M ∼
= M/IM as R-modules.
 C.2. TENSOR PRODUCTS 97

Proof. Consider the exact sequence

0 I R R/IR .
Tensoring with M gives the exact sequence
I ⊗R M R ⊗R M R/IR ⊗R M .
So R/IR ⊗ M ∼
= coker (I ⊗ M → M, i ⊗ 7→ im), which is precisley M/IM .

Example C.11. It is in general not true that the tensor product functor
is left-exact: For that, consider the injective map Z → Z, x 7→ 2x. Then
tensoring with Z/2Z gives an induced morphism
Z ⊗Z Z/2Z −→ Z ⊗Z Z
x ⊗ 1 7−→ 2x ⊗ 1.
But now we have 2x ⊗ 1 = x ⊗ 2 = 0, so the induced homomorphism is
the zero-morphism, and in particular not injective. (Note that we have
Z ⊗Z Z/2Z ∼
= Z/2Z).
C.2.1. Flat Modules. We saw in Example C.11 that the tensor prod-
uct functor is in general not exact. This leads to the following definition:

Definition C.12. An R-module N is flat if the tensor-product functor

− ⊗R N is exact.
Example C.13.
i) The ring R as a module is always flat, since M ⊗R R ∼ = N for all
R-modules N .
ii) More generally, any projective R-module P is flat. If P is finitely
presented, then the converse is also true.

C.2.2. Extension of Scalars. Let ϕ : R → R0 be a ring homomor-

phism and M an R-module. Then ϕ induces the structure of an R-module
on R0 and hence we can form the tensor product R0 ⊗R M . Now this induces
the structure of an R0 -module on R0 ⊗R M given by
r10 r20 ⊗ m := r10 r20 ⊗ m,

which can be extended linearly. This is called extensions of scalars or base

change.
Proposition C.14.
i) Extension of scalars is functorial: For every R-module homomor-
phism f : M → M 0 there is an induced R0 -linear morphism
R0 ⊗R M → R0 ⊗R M 0 . Denote this functor by F : R-Mod → R0 -Mod.
ii) Recall that by Example 2.2, we can regard every R0 -module as an
R-module. This defines a functor G : R0 -Mod → R-Mod.
iii) The pair F : R-Mod R0 -Mod : G defines an adjunction.

Proof.
98 C. FURTHER REMARKS - MODULES

i) For every R-linear map f : M → M 0 , we get an R-bilinear map

R0 × M −→ R0 × M 0
r0 , m 7−→ r0 ⊗ f (m).
This induces an R-linear map R0 ⊗R M → R0 ⊗ M 0 , given on ele-
mentary tensors by r0 ⊗ m 7→ r0 ⊗ f (m). Cleary, this map is also
R0 -linear.
ii) An R0 -linear map M → M 0 is in particular R-linear, since the action
of R on M is given by r.m := ϕ(r)m.
iii) The assignment
homR0 (M ⊗R R0 , M 0 ) −→ homR (M, M 0 )
f f0
M ⊗R R0 −
→ M 0 7−→ (M −→ M 0 , m 7→ f (1 ⊗ m))
is a natural bijection.


C.3. Localization of Modules

Proposition C.15. Let M be an R-module, S ⊆ R a multiplicative set.
Then the map
S −1 M −→∼
S −1 R ⊗R M
x 1
7−→ ⊗x
r r
is an isomorphism of S −1 R-modules.

C.4. Local-Global
For more, the reader is refered to [Sta19, 00EN].
Proposition C.16 (being zero is local). Let M be an R-module. Then
the following are equivalent:
i) M = 0.
ii) Mp = 0 for all prime ideals p ⊆ R.
iii) Mm = 0 for all maximal ideal m ⊆ R.

Proposition C.17 (being injective/surjective is local). Let f : M → N

be an R-linear map.
i) The following are equivalent:
a) f is injective.
b) fp : Mp → Np is injective for all prime ideals p ⊆ R.
c) fm : Mm → Nm is injective for all maximal ideals m ⊆ R.

ii) The following are equivalent:

a) f is surjective.
b) fp : Mp → Np is surjective for all prime ideals p ⊆ R.
c) fm : Mm → Nm is surjective for all maximal ideals m ⊆ R.
 C.5. STRUCTURE THEOREMS FOR MODULES 99

C.5. Structure Theorems for Modules

C.5.1. Modules over PIDs.
Definition C.18. Let R be a ring. We say R is a principal ideal domain
if R is an integral domain and every ideal I ⊆ R is of the form hai for an
a ∈ R.
The following facts about modules over PIDs are taken from [Fra18b,
Chapter 4]
Proposition C.19. Let R be a PID and F a free R-module and M ⊆ F
a submodule. Then M is free.
Theorem C.20. Let R be a PID and M a finitely generated R-module.
Then there is a r ≥ 0, an r ≥ 0 and prime elements p1 , . . . , pn ∈ R such that
for all 1 ≤ i ≤ n there are 1 ≤ si1 ≤ . . . ≤ siti with
 
n M li  si 
M∼
M M
= Rr  A/ pi j 
i=1 j=1
 Bibliography

[Alu09] Paolo Aluffi, Algebra: chapter 0, Vol. 104, American Mathematical Soc., 2009.
[AM94] M.F. Atiyah and I.G. MacDonald, Introduction to commutative algebra, Addison-
Wesley series in mathematics, Avalon Publishing, 1994.
[Bra16] M. Brandenburg, Einführung in die kategorientheorie: Mit ausführlichen erklärun-
gen und zahlreichen beispielen, Springer Berlin Heidelberg, 2016.
[EE95] D. Eisenbud and P.D. Eisenbud, Commutative algebra: With a view toward
algebraic geometry, Graduate Texts in Mathematics, Springer, 1995.
[Fra17] J. Franke, Algebra 1 (lecture notes) (2017), available at https://github.com/
Nicholas42/AlgebraFranke/blob/master/AlgebraI/Alg1.pdf.
[Fra18a] , Homological methods in commutative algebra (lecture notes) (2018), avail-
able at https://github.com/Nicholas42/AlgebraFranke/raw/master/HomAlg/
HomAlg.pdf.
[Fra18b] H. Franzen, Algebra 1 (lecture notes) (2018), available at https://github.com/
lkempf/AlgebraStroppel.
[MR89] H. Matsumura and M. Reid, Commutative ring theory, Cambridge Studies in
Advanced Mathematics, Cambridge University Press, 1989.
[Sch18] J. Schröer, Lineare algebra (lecture notes) (2018).
[Sch19] , Einführung in die algebra (lecture notes) (2019).
[Sta19] The Stacks project authors, The stacks project, 2019.
[Ste19] J. Stelzner, Foundations of representation theory (lec-
ture notes) (2019), available at https://github.com/cionx/
foundations-in-representation-theory-notes-ws-18-19.
[Str18] C. Stroppel, Einführung in die algebra (lecture notes) (2018), available at https:
//github.com/lkempf/AlgebraStroppel.
[Vak18] R. Vakil, MATH 216: Foundations of Algebraic Geometry (2018), available at
http://math.stanford.edu/~vakil/216blog/FOAGnov1817public.pdf.
[Wei94] C. Weibel, An introduction to homological algebra, Graduate Texts in Mathematics,
Cambridge University Press, 1994.

100
 Index

MaxSpec, 8 of a module, 26
germ, 16
affine going down, 38
space, 54 going up, 38
algebra, 30 group
of finite type, 31 class, 84
algebraic subset, 55 totally ordered, 78
artinian, 62
height
basis of a prime ideal, 51
of a module, 26 Hilbert
of a topology, 21 Nullstellensatz, 48
bilinear
map of R-modules, 93 ideal
contraction of, 19
class group, 84 fractional, 83
closure maximal, 5
integral, 33 prime, 5
cokernel, 24 primary, 70
contraction, 19 proper, 5
coordinate ring, 55 radical, 11, 58
coproduct image, 24
of R-modules, 25 integral
over an ideal, 39
decomposition closure, 33
primary, 71 element, 33
Dedekind domain, 80 extension, 35
determinant, 27 over another ring, 33
dimension integral closure
of a ring, 51 of an ideal, 39
direct sum, 25 irreducible
dual component, 60
of a module, 87 isomorphism
of R-modules, 23
exact
sequence, 32 jacobson
ring, 47
finite type, 31 Jacobson radical, 13
finitely generated
as algebra, 31 kernel, 24
as module, 26
formal power series, 15 Lemma
free Nakayama, 28
module, 26 linear map, 22
linearly independent subset, 26
generating system local

101
102 INDEX

ring, 14 sequence
localization exact, 32
at a prime ideal, 18 short exact, 32
at an element, 18 space
lying over, 37 affine, 54
spectrum, 8
module, 22 split, 85
finitely generated, 26 stalk, 16
free, 26 submodule, 23
invertible, 87 subset
localization, 31 multiplicative, 17
projective, 93 subvariety, 55
morphism symbolic power, 66
of modules, 22
Theorem
Nakayama Lemma, 28 Cayley-Hamilton, 28
nilpotent element, 10 theorem
nilradical, 11 Chinese Remainer, 89
noetherian, 62 Hilbert’s Nullstellensatz, 48
normal topology, 7
ring, 40 basis of, 21
object valuation, 78
representing, 91 valuation ring
of scalars discrete, 80
restriction, 22 valution
ring, 78
picard group, 88
subgroup, 78
primary
variety, 55
decomposition, 71
ideal, 70 Zariski Topology
primary decomposition on Ank ., 57
minimal, 72 Zariski topology
prime ideal on Spec R, 8
height, 51
associated, 73
embedded, 73
isolated, 73
principal open subset, 21
product
of R-modules, 25

radical, 10
jacobson, 13
nil-, 11
regular, 60
ring, 75
representing object, 91
restriction
of scalars, 22
ring
local, 14
of formal power series, 15
of fractions, 17
of polynomial functions, 55
PID, 99
jacobson, 47
reduced, 11
regular, 75