Algebra I:

Commutative Algebra

Alexander Schmitt

Berlin, Winter 2012/2013

1
i
Table of Contents

Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

I Basic Theory of Rings and their Ideals . . . . . . . . . . . . . . . . . . . . . 1

I.1 Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
I.2 Ideals and Quotient Rings . . . . . . . . . . . . . . . . . . . . . . . . . . 8
I.3 Zero Divisors, Nilpotent Elements, and Units . . . . . . . . . . . . . . . 11
I.4 Prime Ideals and Maximal Ideals . . . . . . . . . . . . . . . . . . . . . . 12
I.5 Irreducible Elements and Prime Elements . . . . . . . . . . . . . . . . . 19
I.6 Factorial Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
I.7 The Nilradical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
I.8 Operations on Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
I.9 Algebraic Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

II Noetherian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
II.1 Chain Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
II.2 Artinian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
II.3 Localization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
II.4 Primary Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

III The Nullstellensatz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

III.1 Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
III.2 Finite Ring Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
III.3 The Nullstellensatz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
III.4 Noether Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
III.5 Normal Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

IV Dimension Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

IV.1 Krull Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
IV.2 The Going-Up Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
IV.3 The Transcendence Degree of a Field . . . . . . . . . . . . . . . . . . . . 122
IV.4 The Dimension of an Algebraic Variety . . . . . . . . . . . . . . . . . . 123
IV.5 Krull’s Principal Ideal Theorem . . . . . . . . . . . . . . . . . . . . . . . 130

i
Preface

IV.6 Embedding Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

IV.7 Singular Points of Algebraic Varieties . . . . . . . . . . . . . . . . . . . 135
IV.8 Regularity and Normality . . . . . . . . . . . . . . . . . . . . . . . . . . 148

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

ii
Preface

The present text contains notes on my course “Algebra I” at Freie Universität Berlin dur-
ing the winter term 2012/2013. The course “Algebra I” is part of a cycle of three courses
providing an introduction to algebraic geometry. It is also meant as a continuation of my
course “Algebra und Zahlentheorie” (see [30]).
The basic objects we will be studying are commutative rings and their ideals. There
are many different motivations for looking at these objects. A ring can be, for example,
viewed as a domain of numbers with which we would like to compute. In certain situa-
tions, ideals are valuable generalizations of numbers.1 A ring can also consist of regular
functions on an affine algebraic variety. If we consider algebraic varieties over an alge-
braically closed field such as the field of complex numbers, then this algebra completely
determines the variety. Ideals in the algebra correspond to subvarieties, e.g., points of the
variety. It is one of the merits of commutative algebra that it provides a unified framework
for, among other things, arithmetic and algebro-geometric investigations.
The first chapter presents the language of rings and their ideals. Many operations on
rings and ideals which will be used throughout the text are presented. I would like to
highlight two topics: The first one is the section on factorial rings. It shows how we may
generalize the main theorem of elementary number theory, i.e., the unique factorization of
natural numbers into powers of prime numbers, to other settings and problems which will
usually occur. The second one is the spectrum of a ring. It attaches to a commutative ring
a geometric object. This is a fundamental construction of modern algebraic geometry.
Noetherian rings are rings which satisfy a crucial finiteness condition. This condition
is fulfilled by important rings occuring in number theory and algebraic geometry, such
as orders in number fields and coordinate algebras of algebraic varieties. A central result
is the decomposition of ideals in noetherian rings into primary ideals and its uniqueness
properties. This is a vast generalization of the main theorem of elementary number theory
and has also an important geometric interpretation.
As the main topic of the courses “Algebra I-III” is algebraic geometry, the remaining
two chapters deal with subjects of a more geometric nature. The third chapter focusses
on Hilbert’s Nullstellensatz. The Nullstellensatz provides the dictionary between finitely
generated algebras over an algebraically closed field k and their (radical) ideals on the
one hand and algebraic varieties defined over k and their subvarieties on the other hand. It
explains the fundamental role of commutative algebra in algebraic geometry. We present
an elementary proof due to Munshi. Another central topic is Noether’s normalization
1
This is an idea of Kummer.

iii
Preface

theorem which supplies important information on the structure of algebraic varieties and
will be used over and over again in the fourth chapter. The third chapter also develops the
notion of modules over a ring.
The fourth chapter begins the study of the geometry of algebraic varieties. The Krull
dimension of a ring is introduced and investigated. If the ring in question is the coordi-
nate algebra of an affine algebraic variety, this provides a basic geometric invariant. We
will check various properties which we intuitively expect from the notion of dimension.
Finally, we will study singularities of algebraic varieties. In this context, we will also
discuss the relation between the intricate notion of normality of rings and affine algebraic
varieties and singularities.
Dr. Juan Pons Llopis and Anna Wißdorf proofread the manuscript and suggested var-
ious corrections and improvements. The biographical data of mathematicians were taken
from Wikipedia. These notes are heavily based on the books [1] and [14]. Other important
sources are [4], [8], and [11].

Alexander Schmitt
Berlin, March 2013

iv
I
Basic Theory of Rings and their Ideals

In an introductory course on linear algebra, one usually works over fields. In commutative
algebra, fields are replaced by more general objects, namely commutative rings with iden-
tity element. In linear algebra, you rarely talk about the fields themselves. Certainly, you
can do some explicit computations over the field of rational numbers or finite fields, and,
for the theory of the Jordan normal form, you need the ground field to be algebraically
closed. Apart from that, you don’t worry much about the “internal” structure of the field.
In various respects, rings have a much richer structure than fields. To begin with, you
should think of the ring of integers Z or the polynomial ring k[x] over a field k. In these
rings, you may investigate when a number or a polynomial divides another one. This
leads you to certain “indivisible” objects which you call prime numbers or irreducible
polynomials. In fields, there are no counterparts to these concepts, because any non-zero
element is a unit. During the development of algebra, it turned out that it is more useful
to work with ideals than with the ring elements themselves.1 In this section, we will first
define rings and look at some basic examples. Then, we will develop the notion of ideals
and explain how to compute with them. As a motivation in Kummer’s2 spirit, we will also
look at prime factorization. In that context, we will see its failure in some rings and the
class of factorial rings in which prime factorizations do exist.

I.1 Rings
A ring is a tuple (R, 0, +, ·) which consists of an abelian group (R, 0, +) (see [30], Defini-
tion II.1.1 and II.1.4) and a map

· : R × R −→ R
(a, b) 7−→ a · b
1
The term ideal goes back to Kummer’s notion of “ideale Zahl”, an extension of the concept of number
or ring
√ element that permits to generalize the prime factorization in the ring of integers to some rings such
as Z[ −5] (see Section I.5).
2
Ernst Eduard Kummer (1810 - 1893) was a German mathematician.

1
I. Basic Theory of Rings and their Ideals

which is associative, i.e.,

∀a, b, c ∈ R : a · (b · c) = (a · b) · c,

such that the distributive laws hold, i.e.,

∀a, b, c ∈ R : a · (b + c) = a · b + a · c,
(a + b) · c = a · c + b · c.

We will refer to “+” as the addition and to “·” as the multiplication. In the sequel, we
will write R rather than (R, 0, +, ·) for the datum of a ring.
A ring R is called commutative, if

∀a, b ∈ R : a · b = b · a.

An element 1 in a ring R is an identity element, if

∀a ∈ R : 1 · a = a = a · 1.

I.1.1 Note. A ring R can have at most one identity element. For, if 1, 1′ ∈ R are identity
elements, we have
1 = 1 · 1′ = 1′ .
Let R be a ring. For a ∈ R, we let −a be the additive inverse of a, i.e., the element for
which a + (−a) = 0 holds.

I.1.2 Properties. Let R be a ring.

i) For a ∈ R, we have a · 0 = 0 · a = 0.
ii) For a, b ∈ R, we have (−a) · b = −(a · b) = a · (−b). In particular, (−a) · (−b) = a · b.

Proof. i) We use 0 = 0 + 0, so that

0 · a = (0 + 0) · a = 0 · a + 0 · a.

Add −(0 · a) to both sides to get

0 = 0 + 0 · a = 0 · a.

Similarly, one shows a · 0 = 0.

ii) Using i), we see

a · (−b) + a · b = a · (−b + b) = a · 0 = 0.

Thus, a · (−b) = −(a · b). In the same vein, one shows (−a) · b = −(a · b). 
I.1.3 Examples. i) R = {0} with the only possible addition and multiplication is a ring with
identity element 1 = 0. (It is not a field (compare [30], Definition III.1.1.)!) Note that R
is the only ring with identity element in which 1 = 0 holds. Indeed, let R , {0} be a ring
with identity element 1 and pick a ∈ R \ {0}. Then,

1 · a = a , 0 = 0 · a.

2
 I.1. Rings

Hence, 1 , 0.
ii) Let (R, 0, +) be an abelian group and define

· : R × R −→ R
(a, b) 7−→ 0

Then, (R, 0, +, ·) is a ring. If R , {0}, then R has no identity element.

iii) The integers form the ring Z.
iv) Fields are rings, e.g., Q, R, C, Fq , q = pn , n ≥ 1 and p a prime number.
v) Suppose R is a ring and X is a set. We introduce

Map(X, R) := f : X −→ R | f is a set theoretic map .

For f, g ∈ Map(X, R), we form

f + g : X −→ R
x 7−→ f (x) + g(x)

and

f · g : X −→ R
x 7−→ f (x) · g(x).

Moreover, we set

0 : X −→ R
x 7−→ 0.

The tuple (Map(X, R), 0, +, ·) is a ring. Observe that Map(X, R) is commutative if R is so

and that Map(X, R) possesses an identity element if R does. In fact, if 1 ∈ R is the identity
element of R, then

1 : X −→ R
x 7−→ 1

is the identity element in Map(X, R). Note, however, that Map(X, R) will, in general, not
be a field, even if R is one (see Example I.3.1, iv).
vi) Suppose X is a topological space. Then,

C (X, R) := f : X −→ R | f is continuous

together with 0, 1, +, · as in v) is a commutative ring with identity element.

vii) Suppose X ⊂ C is an open subset (compare [31], Definition III.2.1, ii), and
III.3.10, i). Then,

O (X) := f : X −→ C | f is holomorphic
together with 0, 1, +, · as in v) is a commuative ring with identity element.
viii) Let (G, 0, +) be an abelian group. We look at

End(G) := f : G −→ G | f is a group homomorphism

3
I. Basic Theory of Rings and their Ideals

and define
0 : G −→ G
g 7−→ 0
and, for f, h ∈ End(G),
f + h : G −→ G
g 7−→ f (g) + h(g).
Composition of maps provides us with the multiplication: For f, h ∈ End(G), we set
f ◦ h : G −→ G


g 7−→ f h(g) .
We leave it to the reader to verify that (End(G), 0, +, ·) is a ring with identity element idG .
It is, in general, not a commutative ring (see Exercise I.1.4).
ix) Let k be a field. The vector space Mn (k) of (n×n)-matrices with entries in k forms a
ring with respect to componentwise addition and matrix multiplication (see [33], Kapitel
III). The unit matrix En is the identity element. Note that Mn (k) is non-commutative if
and only if n ≥ 2.
x) Suppose R1 , ..., Rn are rings. The cartesian product R1 × · · · × Rn equipped with
componentwise addition and multiplication is again a ring. It is called the direct product
of the rings R1 , ..., Rn .
I.1.4 Exercise. Describe End(Z × Z) in terms of (2 × 2)-matrices and give two elements
of that ring which do not commute with each other.
Let R be a ring with identity element. A subset S ⊂ R is a subring, if S is a subgroup
of (R, +, 0) (see [30], Definition II.4.1), 1 ∈ S and a · b ∈ S , if a, b ∈ S .

From now on, all rings are supposed to be commutative and to possess an
identity element.

Let R, S be two rings. A homomorphism from R to S is a map f : R −→ S , such that

⋆ ∀a, b ∈ R : f (a + b) = f (a) + f (b), i.e., f is a homomorphism of the underlying
abelian groups,
⋆ ∀a, b ∈ R : f (a · b) = f (a) · f (b),
⋆ f (1) = 1.
I.1.5 Remark. One may be tempted to believe that the third condition is a consequence
of the second. However, the proof for the analogous statement in group theory ([30],
Lemma II.3.4, i) requires the existence of inverse elements. This cannot be assumed for
multiplication. Indeed, choosing R = {0} and S , {0}, we see that it may be false. A more
sophisticated example, using the construction in Example I.1.3, x), is the following:
ϕ: Z −→ Z × Z
k 7−→ (k, 0).

4
 I.1. Rings

I.1.6 Exercise. Let f : R −→ S be a homomorphism of rings. Check that


Im( f ) := b ∈ S | ∃a ∈ R : b = f (a)
is a subring of S .
Is

Ker( f ) := a ∈ R | f (a) = 0
a subring of R?
Suppose R is a ring and a ∈ R. We set
⋆ a0 := 1,
⋆ an+1 := a · an , n ∈ N.
As usual, the exponential rule
∀a ∈ R∀m, n ∈ N : am+n = am · an
holds true.

Polynomial Rings
Let R be a ring. A polynomial over R in the indeterminate x should be an expression of
the form
p = a0 + a1 · x + · · · + an · xn with a0 , ..., an ∈ R.
We have natural rules for adding and multiplying two polynomials. They are based on the
multiplication in R, the distributive laws and
∀m, n ∈ N : xm · xn = xm+n .
In this way, we obtain the polynomial ring R[x]. This description is not satisfactory,
because it uses the mathematically undefined terms “indeterminate” and “formal expres-
sion”. In the following definition, we will characterize the polynomial ring by its (univer-
sal) property rather than by a construction. We advise the reader to pay special attention
to this procedure as this kind of approach will become more and more important during
the course.
A polynomial ring over R in the indeterminate x is a triple (T, x, ι) which consists of a
ring T , an element x ∈ T , and a homomorphism ι : R −→ T , such that the following uni-
versal property holds: For every ring S , every element s ∈ S , and every homomorphism
ϕ : R −→ S , there exists a unique homomorphism Φ : T −→ S , such that
⋆ Φ ◦ ι = ϕ,
⋆ Φ(x) = s.
The stated property may be best remembered by the diagram
R❅
ι ⑧⑧⑧ ❅❅❅ ϕ
⑧ ❅❅
⑧⑧ ❅
⑧
⑧
T ❴ ❴ ❴∃!Φ❴ ❴ ❴ ❴/ S .
x7−→s

5
I. Basic Theory of Rings and their Ideals

Notation. We set R[x] := T .

I.1.7 Remark. Given two polynomial rings (T, x, ι) and (T ′ , x′ , ι′ ), there are, by definition,
uniquely determined homomorphisms
⋆ Φ : T −→ T ′ with a) Φ ◦ ι = ι′ and b) Φ(x) = x′ ;
⋆ Φ′ : T ′ −→ T with a) Φ′ ◦ ι′ = ι and b) Φ′ (x′ ) = x.
Observe that, for Φ′ ◦ Φ, we have
⋆ (Φ′ ◦ Φ) ◦ ι = ι,
⋆ (Φ′ ◦ Φ)(x) = x.
The uniqueness statement in the definition of a polynomial ring, applied to S = T , ϕ = ι
and s = x shows
Φ′ ◦ Φ = idT .
Similarly, we verify
Φ ◦ Φ′ = idT ′ .
We do not only see that T and T ′ are isomorphic, there is also a distinguished isomorphism
between T and T ′ that respects the extra data x, ι and x′ , ι′ , respectively. One says
The tuples (T, x, ι) and (T ′ , x′ , ι′ ) are canonically isomorphic.
This means that, in dealing with a polynomial ring, we need just to remember its uni-
versal property as it is completely determined by it. This aspect will be very useful in
other situations, e.g., the tensor product ([1], p. 24f) or the fibered product of schemes
([11], Chapter II.3) are usually remembered by their universal properties and not by their
(involved) constructions.
I.1.8 Theorem. Let R be a ring. Then, there exists a polynomial ring (T, x, ι) over R.
Furthermore, the homomorphism ι is injective.
Proof. We define

T := f : N −→ R | f (n) = 0 for all but finitely many n ∈ N ,

x: N −→ R(
0, if n , 1
n 7−→ ,
1, if n = 1
and
ι : R −→ T ( !
0, if n , 0
a 7−→ n 7−→ .
a, if n = 0
We define addition as before, i.e., for f, g ∈ T , we set
f + g: N −→ R
n 7−→ f (n) + g(n).

6
 I.1. Rings

The product of f, g ∈ T is defined in the following way:

f · g: N −→ R
X
n
n 7−→ f (k) · g(n − k).
k=0

The reader should verify that f + g and f · g do belong to T , i.e., vanish in all but finitely
many points, f, g ∈ T . Furthermore, the following properties are readily verified:
⋆ T is a commutative ring with identity element ι(1).
⋆ ι is an injective ring homomorphism.
Let us write a map f : N −→ R as a sequence (a0, a1, a2, ...) with ak = f (k), k ∈ N. Then,
⋆ x = (0, 1, 0, ...),
⋆ ι(a) = (a, 0, 0, ...), a ∈ R,
⋆ (a0 , a1, a2 , ...) + (b0 , b1 , b2 , ...) = (a0 + b0 , a1 + b1 , a2 + b2 , ...),
⋆ (a0 , a1, a2 , ...) · (b0 , b1, b2 , ...) = (a0 · b0 , a0 · b1 + a1 · b0 , ...),
⋆ xk = (0, ..., 0, 1, 0, ...) with 1 at the (k + 1)-st place, k ∈ N, i.e.,

xk : N −→ (R
0, if n , k
n 7−→ .
1, if n = k

Let f = (a0 , a1 , a2 , ...) ∈ T . If ak = 0 for all k ∈ N, we simply write f = 0. Otherwise, let


n := max k ∈ N | ak , 0 .

Then, using the above formulae, we have the identity

f = a0 + a1 · x + a2 · x2 + · · · + an · xn .

This representation is unique.

Using the above notation, we find, for a given homomorphism ϕ : R −→ S , s ∈ S , and
a homomorphism Φ : T −→ S , satisfying Φ ◦ ι = ϕ and Φ(x) = s, that

Φ(a0 +a1 · x+· · ·+an · xn ) = ϕ(a0 )+ϕ(a1 )· s +· · ·+ϕ(an )· sn , n ∈ N, a0 , ..., an ∈ R. (I.1)

This shows that Φ is uniquely determined, if it exists. On the other hand, for given
ϕ : R −→ S and s ∈ S , Equation (I.1) defines a set theoretic map Φ : T −→ S , and it is
readily checked that it is a ring homomorphism with Φ(x) = s and Φ ◦ ι = ϕ. 
We can now recursively define the polynomial ring in the indeterminates x1 , ..., xn+1 as

R[x1 , ..., xn+1] = R[x1 , ..., xn][xn+1 ]. (I.2)

I.1.9 Exercise. Characterize R[x1 , ..., xn+1 ] by a universal property which is not recursive.

7
I. Basic Theory of Rings and their Ideals

Let S be a ring, R ⊂ S a subring, and y1 , ..., yn ∈ S . By the universal property of the

polynomial ring, there is a unique homomorphism

R[x1 , ..., xn] −→ S

xi 7−→ yi , i = 1, ..., n.

We say that y1 , ..., yn are algebraically independent over R, if this homomorphism is in-
jective.

I.2 Ideals and Quotient Rings

Let R be a ring. A subset I ⊂ R is called an ideal, if
⋆ I is a subgroup of (R, 0, +),
⋆ ∀a ∈ I∀r ∈ R : r · a ∈ I, or, for short, R · I ⊂ I.
The role of ideals in ring theory is somewhat similar to the role of normal subgroups in
group theory (see [30], Abschnitt II.9). We will come back to this below.
I.2.1 Examples and properties. i) For any ring R, the subsets {0} and R are ideals.
ii) Let R be a ring and I ⊂ R an ideal. Then,

I=R ⇐⇒ 1 ∈ I.

The implication “=⇒” is clear. For “⇐=”, let a ∈ R. Since a = a · 1, it follows that a
belows to I.
iii) Suppose k is a field. Then, the only ideals are {0} and k. For, if I , {0} is an ideal
of k, there exists an element a ∈ I \ {0} ⊂ k \ {0}. Since 1 = a−1 · a, we see that 1 ∈ I, and
ii) implies I = k.
iv) If R is a ring and a ∈ R an element, then

hai := r · a | r ∈ R

is an ideal of R. It is called the principal ideal generated by a and is the smallest ideal of
R which contains a.
v) Suppose R = Z. We first recall that, for integers k, m ∈ Z, the relation m|k means
that there is an integer l ∈ Z with k = l · m which is equivalent to k ∈ hmi. Now, suppose
I ⊂ Z is an ideal. The zero ideal is the principal ideal h0i. If I is a nonzero ideal, it
contains some integer k , 0. By the definition of an ideal, it also contains −k = (−1) · k.
It follows that I contains a positive integer. By the least element principle ([27], Satz
1.3.22), we can define

m := min n ∈ N | n > 0 ∧ n ∈ I .
We claim I = hmi. Suppose k ∈ I is a nonzero element. There are integers x, y ∈ Z (see
[30], Satz I.4.4, ii), such that

x · k + y · m = gcd(k, m).

This shows gcd(k, m) ∈ I. By definition of m, we have m ≤ gcd(k, m). This means

m = gcd(k, m) and is equivalent to m|k, i.e., k ∈ hmi.

8
 I.2. Ideals and Quotient Rings

vi) Let R be a ring and X a set. In Example I.1.3, v), we introduced the ring Map(X, R).
Let Y ⊂ X be a subset. Then,

I := f : X −→ R | ∀y ∈ Y : f (y) = 0

is an ideal of Map(X, R).

vii) Let R, S be rings and ϕ : R −→ S a homomorphism. Then, the kernel

Ker(ϕ) := x ∈ R | ϕ(x) = 0

is an ideal of R. More generally, for every ideal J ⊂ S , the preimage ϕ−1 (J) ⊂ R is an
ideal. (Note Ker(ϕ) = ϕ−1 ({0}).)
viii) The inclusion Z ⊂ Q is a ring homomorphism. Its image is not an ideal, i.e., the
image of an ideal is, in general, not an ideal.
ix) Let R, S be rings and ϕ : R −→ S a surjective homomorphism. Then, the image
of an ideal in R is an ideal in S . In fact, let I ⊂ R be an ideal. Then, ϕ(I) is a subgroup
of (S , 0, +) ([30], Lemma II.4.4). Now, let s ∈ S and b ∈ ϕ(I). Then, there exist elements
r ∈ R and a ∈ I with ϕ(r) = s and ϕ(a) = b. We see

s · b = ϕ(r) · ϕ(a) = ϕ(r · a) ∈ ϕ(I).

x) If I, J ⊂ R are ideals, then I ∩ J and


I + J = a+b|a ∈ I ∧ b ∈ J

are ideals, too. In particular, we have, for elements a1 , ..., a s ∈ R the ideal

h a1 , ..., a s i = ha1 i + · · · + ha s i.

It is the smallest ideal of R which contains a1 , ..., a s .

Part iii) and v) illustrate how ideals reflect the algebraic structure of the respective
ring. We will see many more examples of this kind.
Let us spend a few words why ideals are important. For this, let R be a ring and I ⊂ R
an ideal. Then, the set R/I of residue classes inherits the structure of an abelian group
(see [30], Satz II.9.4).

Notation. Write [a] for the class


a + I = a + r | r ∈ I } ∈ R/I, a ∈ R.

Now, we try the following multiplication on R/I:

· : R/I × R/I −→ R/I

[a], [b] 7−→ [a · b].

We need to verify that this is well-defined. Given a, a′, b, b′ ∈ R, such that [a] = [a′ ] and
[b] = [b′ ], there exist elements x, y ∈ I with

a′ = a + x and b′ = b + y.

9
I. Basic Theory of Rings and their Ideals

We get
a′ · b′ = a · b + a · y + b · x + x · y.

By definition of an ideal, we have

a·y+b·x+x·y∈ I and [a′ · b′ ] = [a · b].

Note that

π : R −→ R/I
a 7−→ [a]

is a group homomorphism which satisfies

∀a, b ∈ R : π(a · b) = [a · b] = [a] · [b] = π(a) · π(b).

This shows that the multiplication in R/I satisfies associativity and that the distributive
laws hold. So, R/I inherits a ring structure, such that π is a surjective ring homomorphism.

I.2.2 Lemma. The assignment J 7−→ π−1 (J) induces an inclusion preserving bijection
between the set of ideals of R/I and the set of ideals of R that contain I.

I.2.3 Exercise. Prove this lemma. In particular, give the map from the set of ideals in R
that contain I to the set of ideals of S that is inverse to the map described in the lemma.
I.2.4 Exercises. i) Let R be a ring and I ⊂ R an ideal. Show that the pair (R/I, π), consist-
ing of the quotient ring R/I and the surjection π : R −→ R/I, a 7−→ [a], has the following
universal property (compare [30], Satz II.9.7): For every ring S and every homomorphism
ϕ : R −→ S , such that
I ⊂ Ker(ϕ),

there is a unique homomorphism ϕ : R/I −→ S with

ϕ = ϕ ◦ π;

R
⑤⑤ ❄❄❄
π ⑤⑤ ❄❄ϕ
⑤⑤ ❄❄
~⑤⑤ ❄
∃!ϕ
R/I ❴ ❴ ❴ ❴ ❴ ❴ ❴/ S .

ii) Let f : R −→ S be a homomorphism of rings. Prove the first isomorphism theo-

rem (compare [30], II.10.1): We have Im( f ) = Im( f ) (compare Exercise I.1.6), f being
the induced homomorphism from Part i), and

f : R/Ker( f ) −→ Im( f )

is an isomorphism.

10
 I.3. Zero Divisors, Nilpotent Elements, and Units

I.3 Zero Divisors, Nilpotent Elements, and Units

Let R be a ring. An element a ∈ R is called a zero divisor, if there exists an element b , 0,
such that a · b = 0. The ring R is called an integral domain, if R , {0} and 0 is its only
zero divisor.
I.3.1 Examples. i) The ring Z of integers is an integral domain.
ii) Fields are integral domains.
iii) If R is an integral domain, then the polynomial ring R[x] is also an integral domain.
In particular, if R is a field, then the polynomial ring k[x1 , ..., xn] in n variables is an integral
domain.
iv) Assume that X contains at least two distinct elements x1 , x2 and R , {0} is a ring.
Then, Map(X, R) contains non-trivial zero-divisors. For example, we look at

fi : X −→ R
(
0, if x , xi
x 7−→ , i = 1, 2.
1, if x = xi

Then, fi , 0, i = 1, 2, but f1 · f2 = 0.
Let R be a ring. An element a ∈ R is nilpotent, if there exists a natural number n with
n
a = 0.
I.3.2 Remark. Obviously, a nilpotent element is a zero divisor. The converse does not
hold. E.g., f1 and f2 in Example I.3.1, iv), are zero divisors but not nilpotent.
I.3.3 Example. Let k be a field and n ≥ 2 a natural number. We look at the principal ideal
hxn i ⊂ k[x] and the quotient ring R := k[x]/hxn i. Then, [x]k = [xk ] , 0, for 1 ≤ k < n, but
[x]n = [xn ] = 0, i.e., [x] is a nilpotent element of R.
I.3.4 Exercise. Let n ≥ 2 be a natural number. Describe the zero divisors and nilpotent
elements in the residue ring Z/hni in terms of the prime factorization (see [30], Kapitel I)
of n.
Let R be a ring. An element a ∈ R is called a unit, if there exists an element b ∈ R,
such that a · b = 1. Observe that the element b is uniquely determined. (Indeed, for
b, b′ ∈ R with a · b = 1 = a · b′ , we find b = b · 1 = b · (a · b′ ) = (b · a) · b′ = 1 · b′ = b′ .)

Notation. We write a−1 := b.

I.3.5 Remarks. i) The set


R⋆ := a ∈ R | a is a unit of R

of units in the ring R is an abelian group with respect to multiplication in R with identity
element 1.
ii) An element a ∈ R is a unit if and only if hai = R.
I.3.6 Examples. i) The units of the ring Z of integers form the group Z⋆ = {±1}.
ii) Let R be an integral domain, then (R[x])⋆ = R⋆ (compare Exercise I.3.10).
I.3.7 Exercise. Describe the units of Z/hni for n ≥ 1 (compare [30], Abschnitt III.2).

11
I. Basic Theory of Rings and their Ideals

I.3.8 Exercises (Units and nilpotent elements). i) Let R be a ring and n ∈ R a nilpotent
element. Show that 1 + n is a unit.
ii) Deduce that the sum u + n of a unit u ∈ R and a nilpotent element n ∈ R is a unit.

I.3.9 Lemma. Let R , {0} be a ring. The following conditions are equivalent:
i) R is a field.
ii) The subsets {0} and R are the only ideals of R.
iii) Every homomorphism ϕ : R −→ S to a nonzero ring S , {0} is injective.

Proof. For the implication “i)=⇒ii)”, see Example I.2.1, iii).

“ii)=⇒iii)”: If S , {0}, then 1 < Ker(ϕ) (see Example I.2.1). So, Ker(ϕ) is a proper
ideal of R. The assumption yields Ker(ϕ) = {0}. As for group homomorphisms (see [30],
Lemma II.4.5), this implies that ϕ is injective.
“iii)=⇒i)”: For an element a ∈ R \ {0}, we define S := R/hai. Then, π : R −→ S is a
surjective ring homomorphism with π(a) = π(0) = 0. So, π is not injective. This implies
S = {0} or, equivalently, hai = R. In view of Remark I.3.5, ii), this shows that a is a
unit. 

I.3.10 Exercise (Units in polynomial rings). Let R be a ring and f = a0 + a1 x + · · · + an xn ∈

R[x] a polynomial. Show that f is a unit if and only if a0 is a unit and a1 , . . . , an are
nilpotent elements.
Instructions.

• For “=⇒”, use Exercise I.3.8.

• For “⇐=”, let g = b0 + b1 x + · · · + bm xm ∈ R[x] be a polynomial with f · g = 1.

Prove by induction on r that

ar+1
n · bm−r = 0, r = 0, ..., m. (I.3)

To this end write f · g = c0 + c1 · x + · · · + cm+n · xm+n and look at ar+1

n · cm+n−r−1 .
Finally, deduce from (I.3) that an is nilpotent and conclude by induction on n.

I.4 Prime Ideals and Maximal Ideals

Let R be a ring. An ideal I ⊂ R is called a prime ideal, if I , R and

∀a, b ∈ R : a·b∈ I =⇒ (a ∈ I ∨ b ∈ I),

and a maximal ideal, if I , R and there is no ideal J of R with I ( J ( R.

Notation. It is customary to use gothic letters for prime and maximal ideals, e.g., p, m.

I.4.1 Proposition. Let R be a ring and I ⊂ R an ideal, then

i) I is a prime ideal if and only if R/I is an integral domain.
ii) I is a maximal ideal if and only if R/I is a field.

12
 I.4. Prime Ideals and Maximal Ideals

Proof. i) This is immediate from the definitions.

ii) “=⇒”: Let I , R be a maximal ideal. According to Lemma I.2.2, the ideals of R/I
correspond to the ideals in R which contain I. These are I and R. So, {0} and R/I are the
only ideals of R/I. By Lemma I.3.9, R/I is a field. The converse implication “⇐=” is
obtained by a similar reasoning. 

I.4.2 Corollary. A maximal ideal is a prime ideal.

I.4.3 Remarks. i) Let R be a ring. The zero ideal {0} is a prime ideal if and only if R is an
integral domain.
ii) Let ϕ : R −→ S be a homomorphism of rings. If q ⊂ S is a prime ideal, then
p := ϕ−1 (q) is a prime ideal of R. Indeed, we have

1<q =⇒ 1 < p,

so that p is a proper ideal. By Exercise I.2.4, ii), there is the induced injective homomor-
phism ϕ : R/p −→ S /q, [r] 7−→ [ϕ(r)]. Since S /q is an integral domain, the same holds
for R/p. Now, apply Proposition I.4.1, i).
iii) Let ϕ : R −→ S be, as before, a ring homomorphism. If m ⊂ S is a maximal ideal,
then q := f −1 (m) needs not be maximal. Look for example at the inclusion ϕ : Z ⊂ Q.
Then, {0} ⊂ Q is a maximal ideal, because Q is a field, but {0} = ϕ−1 ({0}) is not a maximal
ideal of Z.

I.4.4 Theorem. In every ring R , {0}, there exist maximal ideals.

I.4.5 Remark (The axiom of choice). i) The proof of this theorem requires the axiom of
choice. Like the theorem that every vector space has a basis, this theorem is actually
equivalent to the axiom of choice (see, e.g., [3]).
ii) Since, by Corollary I.4.2, maximal ideals are prime ideals, Theorem I.4.4 shows
that every non-zero ring contains a prime ideal. It might be interesting to know that the
statement “Every ring R , {0} possesses a prime ideal.” is actually weaker than the
axiom of choice. It is equivalent to the axiom (BPI) that every non-zero boolean ring (see
Exercise I.4.16) contains a prime ideal. We refer the reader to the paper [25] for more
details and references.
The axiom of choice, in turn, is equivalent to Zorn’s3 lemma that we shall now for-
mulate. Let (S , ≤) be a partially ordered set. This means that S is a set and “≤” a relation
on S which satisfies the following properties:

⋆ Reflexivity: ∀s ∈ S : s ≤ s.

⋆ Antisymmetry: ∀s1 , s2 ∈ S : (s1 ≤ s2 ∧ s2 ≤ s1 ) ⇐⇒ s1 = s2 .

⋆ Transitivity: ∀s1 , s2 , s3 ∈ S : s1 ≤ s2 ∧ s2 ≤ s3 =⇒ s1 ≤ s3 .

I.4.6 Note. The relation “≤” corresponds to a subset U ⊂ S × S , such that

∀s1 , s2 ∈ S : s1 ≤ s2 ⇐⇒ (s1 , s2 ) ∈ U.
3
Max August Zorn (1906 - 1993) was a German mathematician who emigrated to the USA because of
the Nazi policies.

13
I. Basic Theory of Rings and their Ideals

A chain in S is a subset T ⊂ S , such that, for t1 , t2 ∈ T , one has t1 ≤ t2 or t2 ≤ t1 , i.e., “≤”

induces a total ordering on T .
An upper bound for a chain T is an element u ∈ S , such that t ≤ u holds for all t ∈ T .
Observe that it is not required that u belongs to T .
A maximal element of S is an element m ∈ S , such that, for s ∈ S , m ≤ s implies
m = s.

I.4.7 Zorn’s lemma. Let (S , ≤) be a nonempty partially ordered set, such that every chain
T in S has an upper bound. Then, S contains at least one maximal element.

Proof. See [35], Satz 5.13. 

Proof of Theorem I.4.4. We look at the set

S := I ⊂ R | I , R and I is an ideal .

This set contains {0} and is therefore nonempty, and it is partially ordered by inclusion
“⊂”. Let T be a chain in S and define
[
J := I.
I∈T

Claim. The set J is an upper bound for T , i.e., J ∈ S .

We first verify that J , R. This follows, because 1 < I, I ∈ T . Next, we show that J is
an ideal. To see that J is a subgroup of R, note first that 0 ∈ J, because 0 ∈ I for all I ∈ T .
Next, assume a1 , a2 ∈ J. Then, there are ideals I1 , I2 ∈ T with a1 ∈ I1 and a2 ∈ I2 . Since
T is a chain, we have I2 ⊂ I1 or I1 ⊂ I2 . We assume the latter. Then, a1 + a2 ∈ I2 ⊂ J.
Finally, let a ∈ J and r ∈ R. Then, there is an ideal I ∈ T with a ∈ I. Since I is an ideal,
r · a ∈ I and, thus, r · a ∈ J. For r = −1, this gives −a ∈ J (see Property I.1.2, ii) and
completes the proof. X
By Zorn’s lemma I.4.7, S contains a maximal element m. By definition, m is a maxi-
mal ideal. 

I.4.8 Corollary. i) Every ideal I ⊂ R is contained in a maximal ideal.

ii) Every element a ∈ R which is not a unit is contained in a maximal ideal.

Proof. i) We may apply the theorem to the ring R/I and use Lemma I.2.2 or modify the
above proof.
ii) If a is not a unit, then hai ( R. Hence, we may conclude by i). 
A ring R with exactly one maximal ideal is called a local ring. In this case, the
field R/m is called the residue field. A ring with only finitely many maximal ideals is a
semilocal ring.
I.4.9 Example. A field is a local ring with maximal ideal {0}.

I.4.10 Proposition. i) Let R , {0} be a ring and m ( R an ideal, such that every element
a ∈ R \ m is a unit. Then, R is a local ring with maximal ideal m.
ii) Let R , {0} be a ring and m ⊂ R a maximal ideal, such that 1 + m ⊂ R⋆ . Then, R
is a local ring.

14
 I.4. Prime Ideals and Maximal Ideals

Proof. i) Let I ( R be an ideal. Then, no element of a ∈ I is a unit. By assumption,

I ⊂ m.
ii) Let a ∈ R \ m. Then, hai + m = R, because m is a maximal ideal. So, there exist
r ∈ R and b ∈ m, such that
r · a + b = 1.
We see that r · a ∈ 1 + m, so that r · a is a unit. It follows that a is a unit. (To see this, note
((r · a)−1 · r) · a = 1.) We now conclude by i). 

I.4.11 Examples. i) In Z, every ideal is principal, i.e., of the form hmi for some integer
m ∈ Z. Here, hmi is a prime ideal if and only if m = 0 or m is a prime number. For a
prime number p, F p = Z/hpi is the field with p elements. In particular, every non-zero
prime ideal in Z is a maximal ideal.
ii) An integral domain R in which every ideal is principal is called a principal ideal
domain. Examples for principal ideal domains include the ring of integers Z and the
polynomial ring k[x] over a field k (see Exercise I.4.15, i). In a principal ideal domain,
every nonzero prime ideal is maximal. Indeed, let hai be a nonzero prime ideal, i.e.,
a , 0, and hbi be an ideal with
hai ( hbi.
Thus, there exists an element r ∈ R with r · b = a. If b < hai, we must have r ∈ hai.
Choose s ∈ R with r = s · a. So, b · s · a = a, i.e., (b · s − 1) · a = 0. Since a , 0 and R is
an integral domain, we infer b · s = 1. So, b is a unit and hbi = R.
iii) Power series rings. Let R be a ring. We look again at

Map(N, R) = Sequences (a0 , a1 , a2 , ...) | ak ∈ R, k ∈ N .

The arithmetic operations are as follows:

⋆ As addition, we use again componentwise addition:

(a0 , a1, a2 , ...) + (b0 , b1 , b2 , ...) = (a0 + b0 , a1 + b1 , a2 + b2 , ...).

⋆ As multiplication, we use, as for polynomial rings, the Cauchy4 product:

(a0 , a1 , a2, ...) · (b0 , b1 , b2 , ...) = (c0 , c1 , c2 , ...)

with
X
n
cn := ak · bn−k , n ∈ N. (I.4)
k=0

P
∞
Notation. We write a sequence (a0 , a1 , a2, ...) in Map(N, R) as ak · xk . Such an expres-
k=0
sion is called a formal power series over R. The ring of all formal power series over R is
denoted by R[[x]].

Remark. Note that the polynomial ring R[x] is a subring (see Page 4) of R[[x]].
4
Augustin-Louis Cauchy (1789 - 1857), French mathematician.

15
I. Basic Theory of Rings and their Ideals

P
∞
Proposition. A formal power series ak · xk is a unit in R[[x]] if and only if a0 is a unit
k=0
in R.
We leave it to the reader to verify this easy consequence of (I.4).
Corollary. If k is a field, then k[[x]] is a local ring with maximal ideal
 ∞ 

 X  

 i 
hxi =  a i · x  a i ∈ k, i = 1, 2, 3, ... 
 .
 
i=1

P
∞
Assume k = C. A formal power series ai · xi is convergent, if its radius of conver-
i=0
gence (see [31], Definition II.3.8) is positive. We set
 

 X∞  

C{x} :=  p = ai · x ∈ C[[x]]  p is convergent  .
 i

i=0

Proposition. i) C{x} ⊂ C[[x]] is a subring.

ii) C[x] ⊂ C{x}.
iii) C{x} is a local ring with maximal ideal hxi.
iv) Let k be a field. For a ∈ k, the principal ideal hx − ai is a maximal ideal of k[x].
If a1 , a2 , then hx − a1 i , hx − a2 i. This means that k[x] is not a local ring. If k is
algebraically closed (see [31], Satz IV.5.11), then all maximal ideals are of the form
hx − ai, a ∈ k.
Note. If k is algebraically closed, a non-constant polynomial p ∈ k[x] is irreducible5 if
and only if its degree is 1.
I.4.12 Exercises6 . The ring C{x} may be interpreted in terms of complex functions. Let

S := (U, f ) | U ⊂ C open, 0 ∈ U, f : U −→ C holomorphic .

We introduce the following relation on S:

(U, f ) ∼ (V, g) :⇐⇒ ∃0 ∈ W ⊂ U ∩ V open : f|W ≡ g|W .

i) Prove that “∼” is an equivalence relation on S.

The equivalence classes are germs of holomorphic functions at 0. We write the equiv-
alence class of (U, f ) ∈ S as [U, f ].
ii) Show that addition and multiplication of complex functions endow the set S of
germs of holomorphic functions at 0 with the structure of a ring and that

ev : S −→ C
[U, f ] −
7 → f (0)

is a ring homomorphism.7
5
See Page 20 for the definition of an irreducible element in an integral domain.
6
The necessary prerequisites for these exercises are contained [31], especially Kapitel IV
7
Note that 0 is the only point at which it makes sense to evaluate a germ.

16
 I.4. Prime Ideals and Maximal Ideals

iii) For a germ [U, f ] ∈ S, let T f,0 be its Taylor series with expansion point 0. Show
that
T : S −→ C{x}
[U, f ] −
7 → T f,0
is an isomorphism of rings, such that


T −1 hxi = [U, f ] ∈ S | ev( f ) = 0 .
This exercise illustrates the name “local”: It comes exactly from the context of such
rings of germs of functions at a point, here the origin. Germs are, roughly speaking, the
local functions at the given point.
I.4.13 Exercises (Prime ideals). Determine all prime and maximal ideals of the following
rings: i) R, ii) Z, iii) C[x], and iv) R[x].
I.4.14 Exercise. Let R , {0} be a ring. Show that the set Σ of prime ideals of R has a
minimal element with respect to inclusion.
I.4.15 Exercises. Let k be a field.
i) Prove that the polynomial ring k[x] over k is a principal ideal domain.
ii) Prove that k is algebraically closed if and only if, for every maximal ideal m ⊂ k[x],
there exists an element a ∈ k with m = hx − ai.
I.4.16 Exercises (Boolean rings). i) Let R be a ring such that every element x ∈ R satisfies
xn = x for some n > 1. Show that every prime ideal p of R is a maximal ideal.
ii) A ring R is called boolean8, if every element x ∈ R verifies x2 = x. Show that
2x = x + x = 0 holds true for every element x in a boolean ring R.
iii) Let R , {0} be a boolean ring and p ⊂ R a prime ideal. Show that p is a maximal
ideal and that R/p is a field of two elements.

The Spectrum of a Ring

The following set of exercises contains the first steps of associating with a ring a geometric
object which contains all the information about the ring.
I.4.17 Exercises (The spectrum of a ring). Let R , {0} be a ring. We define

Spec(R) := p ⊂ R | p is a prime ideal .
For an ideal I ⊂ R, we set

V(I) := p ∈ Spec(R) | I ⊂ p .
Establish the following properties:
i) V(0) = Spec(R), V(R) = ∅.
P
ii) Let Ik , k ∈ K, be a family of ideals in R. Their sum Ik is the ideal of all linear
P k∈K
combinations ak with ak ∈ Ik , k ∈ K, almost all zero (see also Page 33). Then,
k∈K
X  \
V Ik = V(Ik ).
k∈K k∈K
8
George Boole (1815 - 1864), English mathematician, philosopher and logician.

17
I. Basic Theory of Rings and their Ideals

iii) For two ideals I and J of R,

V(I ∩ J) = V(I) ∪ V(J).

Remark. Call a subset Z ⊂ Spec(R) Zariski9 closed, if there is an ideal I ⊂ R with Z = V(I)
and a subset U ⊂ Spec(R) Zariski open, if the complement Z = Spec(R) \ U is Zariski
closed. The above properties say:
i’) The empty set and Spec(R) are Zariski open.
ii’) The union of an arbitrary family of Zariski open subsets is Zariski open.
iii’) The intersection of two Zariski open subsets is Zariski open.
So,

T := U ⊂ Spec(R) | U is Zariski open
is a topology (see [18], Section 1.2) on Spec(R), the Zariski topology.
iv) Let f : R −→ S be a homomorphism of rings. Define

f # : Spec(S ) −→ Spec(R)
p 7−→ f −1 (p).

Show that f # is continuous in the Zariski topology.

I.4.18 Exercises (Principal open subsets). Let R be a ring and X := Spec(R). For f ∈ R,
set X f := X \ V(h f i).
i) Show that the X f , f ∈ R, form a basis for the Zariski topology, i.e., for every Zariski
open subset U ⊂ X, there is a subset F ⊂ R, such that
[
U= Xf .
f ∈F
P
Hint: For an ideal I ⊂ R, one has I = h f i.
f ∈I
ii) Prove that X f ∩ Xg = X f ·g , f, g ∈ R.
iii) Check that X f = X holds if and only if f is a unit.
iv) Show that X is quasi-compact, i.e., every open covering of X possesses a finite
subcovering.
I.4.19 Exercises. Let R be a ring and X := Spec(R). √
i) Show that, for an ideal I ∈ R, one has V(I) = V( I).
ii) For a subset Z ⊂ X, define the ideal
\
I(Z) := p.
p∈Z

Show that

√
I V(I) = I
holds for every ideal I ⊂ R.
iii) Let Z ⊂ X be a closed subset. Prove that


V I(Z) = Z.
9
Oscar Zariski (1899 - 1986), was an American mathematician of Russian origin.

18
 I.5. Irreducible Elements and Prime Elements

I.4.20 Exercises (Boolean rings). Let R be a boolean ring (see Exercise I.4.16). Set X :=
Spec(R).
i) Show that, for f ∈ R, the set X f is both open and closed (in the Zariski topology).
ii) Let f1 ,..., fn ∈ R and

I := h f1 , ..., fn i := h f1 i + · · · + h fn i.

Prove that I is a principal ideal.

iii) Suppose f1 ,..., fn ∈ R. Demonstrate that there is an element f ∈ R, such that

X f := X f1 ∪ · · · ∪ X fn .

I.4.21 Exercises (The spectrum of Z[x]). The aim of this exercise is to determine all prime
and maximal ideals of Z[x].
i) Show that a prime ideal p which is not principal contains two irreducible polyno-
mials f1 and f2 with f1 ∤ f2 and f2 ∤ f1 .
ii) Explain why the greatest common divisor of f1 and f2 in Q[x] is 1, so that there are
polynomials g1 , g2 ∈ Q[x] with f1 · g1 + f2 · g2 = 1.
iii) Deduce from ii) that the intersection Z ∩ p is non-zero and therefore of the form
hpi for some prime number p ∈ Z.
iv) Infer that a non-principal prime ideal p ⊂ Z[x] is of the form h p, f i where p ∈ Z is
a prime number and f ∈ Z[x] is a primitive polynomial (see Page 28) of positive degree,
such that its class f ∈ F p [x] is irreducible. Is such an ideal maximal?
v) Now, describe all prime and all maximal ideals of Z[x].
Remark. A picture of Spec(Z[x]) may be found in the books [21], Example H, page 74f,
and [6], Section II.4.3.
I.4.22 Exercise (The spectrum of a product). Let R1 , R2 be non-zero rings. Describe the
spectrum of R1 × R2 in terms of the spectra of R1 and R2 . (Don’t forget to think about the
topology of the respective spaces.)

I.5 Irreducible Elements and Prime Elements

In this section, R is assumed to be an integral domain. The prime factorization in the
ring of integers ([30], Kapitel I) is the most important tool of elementary number theory.
To state it, we just need the relation of divisibility among two integers. This can equally
well be defined and studied in any integral domain.

I.5.1 Question. Is there a prime factorization in R?

We will see that the answer is no, in general. This motivates two developments. First,
we may single out the class of rings for which the answer is yes, so-called factorial rings,
and study some examples and properties of these rings. Second, we can generalize the
concept of prime factorization by allowing also ideals in the factorization. This will lead
to the primary decomposition of ideals (see Section II.4).
Let a, b ∈ R. We say that b divides a, if there exists an element c ∈ R, such that

a = b · c.

19
I. Basic Theory of Rings and their Ideals

I.5.2 Notation. b|a.

The reader may check the following properties of the divisibility relation (see [30],
Eigenschaften I.2.2, for the corresponding statements in the ring Z.)
I.5.3 Properties. i) Let a ∈ R. Then, 1|a and a|a.
ii) Let a, b, c ∈ R. If c|b and b|a, then also c|a.
iii) Let a1 , ..., an, b ∈ R be elements with b|ai , i = 1, ..., n. For all r1 , ..., rn ∈ R, we have

b(r1 · a1 + · · · + rn · an ).

iv) Let b ∈ R be an element with b|1. Then, b is a unit of R.

v) Let a ∈ R and u ∈ R⋆ be a unit. For every b ∈ R, the relation b|a implies (b · u)|a
and b|(a · u).
vi) Let a, b ∈ R. Then,
b|a ⇐⇒ hai ⊂ hbi.
Two elements a, b ∈ R are associated, if there exists a unit u ∈ R⋆ , such that

a = b · u.

I.5.4 Notation. a ∼ b.
I.5.5 Lemma. i) The relation “∼” is an equivalence relation.
ii) Let a, b ∈ R. Then, the following conditions are equivalent: ⋆) a ∼ b, ⋆⋆) a|b∧b|a,
and ⋆ ⋆ ⋆) hai = hbi.

Proof. i) This is very easy to check directly. It is also an immediate consequence of ii).
ii) Condition ⋆) clearly implies ⋆⋆). Condition ⋆⋆) and ⋆ ⋆ ⋆) are equivalent by
Property I.5.3, vi). So, assume that a|b and b|a and let r, s ∈ R be such that a = b · r and
b = a · s. Then,
a = (r · s) · a.
This is equivalent to
(1 − r · s) · a = 0.
Since R is an integral domain,10 we have a = 0 or 1 = r · s. In the first case, b = a · s = 0.
In the second case, r and s are units of R. In both cases, a and b are associated. 
We have now two options to generalize the notion of a prime number in the ring of
integers (compare [30], Definition I.3.1 and Satz I.4.5): An element p ∈ R is called a
prime element, if p , 0, p < R⋆ , and

∀a, b ∈ R: p|(a · b) =⇒ p|a or p|b.

An element q ∈ R is irreducible, ifp , 0, p < R⋆ , and

∀a, b ∈ R: q = a · b =⇒ a ∈ R⋆ or b ∈ R⋆ .

In other words, the only divisors of q are units or associated elements.

10
Here, our general assumption becomes important.

20
 I.5. Irreducible Elements and Prime Elements

I.5.6 Examples. i) Let k be a field. Then, there are no prime or irreducible elements in k.
ii) An integer m ∈ Z is an irreducible element if and only if it is a prime number. A
prime number is a prime element, by [30], Satz I.4.5. Since a prime element is irreducible
(Proposition I.5.7, i), every prime element in Z is a prime number.

I.5.7 Proposition. Let p ∈ R be an element.

i) If p is a prime element, then p is irreducible.
ii) The element p is a prime element if and only if hpi is a prime ideal
iii) The element p is irreducible if and only if there is no element a ∈ R with

hpi ( hai ( R,

i.e., if hpi is maximal among the proper principal ideals of R.

Proof. i) Let a, b ∈ R be such that p = a · b. Since p|p, we have p|(a · b) and p|a or p|b.
Let us assume p|a and let c ∈ R be such that a = p · c. We find p = (b · c) · p. Since p , 0
and R is an integral domain, it follows that b · c = 1 and that b is a unit.
ii) “=⇒”: For a, b with a · b ∈ hpi, we have p|(a · b). Since p is a prime element, this
implies p|a or p|b, i.e., a ∈ hpi or b ∈ hpi. The converse is similar.
iii) “=⇒”: Suppose a ∈ R is an element with hpi ⊂ hai, i.e., a|p. Let b ∈ R be such
that a · b = p. Then, a ∈ R⋆ or b ∈ R⋆ , that is hai = R or hai = hpi.
“⇐=”: For a, b ∈ R with p = a · b, we have hpi ⊂ hai. By assumption, hai = hpi or
hai = R. In the first case b ∈ R⋆ , and, in the second case, a ∈ R⋆ . 

I.5.8 Corollary. Suppose R is a principal ideal domain. Then, any irreducible element
p ∈ R is a prime element.

I.5.9 Important example. We look at the ring

h√ i n √  o
R := Z −5 := k + l · −5 ∈ C  k, l ∈ Z .

It is a subring of the field C of complex numbers and therefore an integral domain. In

order to study divisibility among elements in R, we introduce the norm map

N : R −→ Z
√
k + l · −5 7−→ k2 + 5 · l2 .

It satisfies
∀a, b ∈ R : N(a · b) = N(a) · N(b). (I.5)
We can list elements of small norm:

⋆ An element of a ∈ R has norm 1 if and only if a = ±1.

⋆ There are no elements a ∈ R of norm 2 or 3.

√
⋆ The elements of norm 9 are ±3 and ±(2 ± −5).

This has already several interesting consequences:

Claim. The units of R are ±1.

21
I. Basic Theory of Rings and their Ideals

The elements ±1 are clearly units of R. Conversely, (I.5) shows that a unit u ∈ R
satisfies N(u) = 1. Our previous observation says that this is equivalent to u = ±1. X
√
Claim. The elements ±3 and ±(2 ± −5) are irreducible.
We explain the argument for 3. Equation (I.5) shows that a divisor a of 3 has norm 1
or 9, because there is√no element of norm 3. If N(a)
√ = ±1, then a = ±1. If N(a) = 9, then
a = ±3 or a = ±(2 ± −5). The elements ±(2 ± −5) do not divide 3. X
√
Claim. The elements ±3 and ±(2 ± −5) are not prime.
Again, we show the assertion for 3. We use the equation
√ √
9 = 3 · 3 = (2 − −5) · (2 + −5). (I.6)
√
It shows 3|9. By our previous discussion, 3 ∤ (2 ± −5). X
I.5.10 Exercise. Prove that every element a ∈ R which is neither zero nor a unit can be
written as a product of irreducible elements.
I.5.11 Remark. i) There are several related observations resulting from our discussion, in
particular, Equation (I.6). Every element in R may be written as a product of irreducible
elements, but this factorization is, in general, not unique up to associated element. For
example, 9 has two essentially distinct factorizations. Not every irreducible element is
a prime element. There are elements such as 9 which cannot be written as products of
prime elements. We will clarify these matters in the following section.
ii) The fact that there are numbers which cannot be written as products of prime el-
ements led to the idea of “ideal numbers” and eventually of ideals which may be used
to obtain such a factorization nevertheless. An example looks as follows (see Exercise
I.8.9): The ideals
√ √
p1 := h 3, 2 + −5 i and p2 := h 3, 2 − −5 i

are prime ideals in R with

h9i = p21 · p22 = p21 ∩ p22 .
√
iii) The ring R is the ring of integers in the number field Q( −5). Every number
field K has such a ring of integers OK . The question whether there is a prime factorization
or not in OK is of great importance for algebraic number theory. It is, for example, related
to the famous equation x p + y p = z p , p a prime number, of Fermat.11 Chapter I of [23]
contains detailed information on these topics.

I.6 Factorial Rings

In this section, R will be an integral domain. We will study the following properties:
(F1) For every element a ∈ R\({0}∪R⋆ ), there are a natural number r ≥ 1 and irreducible
elements q1 , ..., qr ∈ R with
a = q1 · · · · · qr .
11
Pierre de Fermat (1601 or 1607/8 - 1665) was a French lawyer and mathematician.

22
 I.6. Factorial Rings

(F2) For every element a ∈ R \ ({0} ∪ R⋆ ), there are a natural number r ≥ 1 and prime
elements p1 , ..., pr ∈ R with
a = p1 · · · · · pr .

(F3) If we are given natural numbers r, t ≥ 1 and irreducible elements q1 , ..., qr , s1, ..., st ∈
R with
q1 · · · · · qr = s 1 · · · · · s t ,
then r = t and there is a permutation σ : { 1, ..., r } −→ { 1, ..., r }, such that

∀i ∈ { 1, ..., r } : qi ∼ sσ(i) .

(F4) Every irreducible element of R is a prime element.

I.6.1 Theorem. The following conditions on the integral domain R are equivalent:
i) The properties (F1) and (F3) hold in R.
ii) The properties (F1) and (F4) hold in R.
iii) Property (F2) holds in R.

Proof. “i)=⇒ii)”. Let q be an irreducible element and a, b ∈ R ring elements with q|a · b.
If a = 0 or b = 0, there is nothing to show. If a ∈ R⋆ , it follows that q|b, and, if b ∈ R⋆ ,
we have q|a. So, we may assume a, b ∈ R \ ({0} ∪ R⋆ ). Let c ∈ R be such that

a · b = q · c.

Since q is irreducible, we must have c < R⋆ and, obviously, c , 0. By (F1), there are
natural numbers r, t, v ≥ 1 and irreducible elements q1 , ..., qr , s1 , ..., st , u1 , ..., uv ∈ R with

a = q1 · · · · · qr , b = s1 · · · · · st , and c = u1 · · · · · uv .

The identity
q1 · · · · · qr · s 1 · · · · · s t = q · u1 · · · · · uv
and (F3) show that there is an index i0 ∈ { 1, ..., r } or an index j0 ∈ { 1, ..., t } with

q ∼ qi0 or q ∼ s j0 ,

so that
q|a or q|b.
“ii)=⇒iii)”. This is trivial.
“iii)=⇒i)”. By Proposition I.5.7, i), every prime element of R is irreducible. This
implies that (F1) holds true in R.
Claim. Property (F4) is verified by R.
In fact, let q ∈ R be an irreducible element. There are a natural number r ≥ 1 and
prime elements p1 , ..., pr with
q = p1 · · · · · pr .
Since prime elements aren’t units, the irreducibility of q implies r = 1 and q = p1 . X

23
I. Basic Theory of Rings and their Ideals

Now, let q1 , ..., qr , s1, ..., st ∈ R be irreducible elements with

q1 · · · · · qr = s 1 · · · · · s t .

We proceed by induction on r. If r = 1, then t = 1, because q1 is irreducible. In general,

there is an index i ∈ { 1, ..., r } with s1 |qi , because, according to (F4), s1 is a prime element.
Since qi is irreducible, this implies s1 ∼ qi . We may clearly assume i = 1. There is a unit
u ∈ R with s1 = u · q1 . We infer

q1 · q2 · · · · · qr = q1 · (u · s2 ) · s3 · · · · · st .

Using the fact that R is an integral domain, this equation implies

q2 · · · · · qr = (u · s2 ) · s3 · · · · · st ,

and we may conclude by induction. 

A factorial ring is an integral domain which satisfies Conditions (F1) - (F4).

I.6.2 Exercise (Chains of ideals in principal ideal domains). Let R be a principal ideal
domain.
i) Let
hr1 i ⊂ hr2 i ⊂ · · · ⊂ hrk i ⊂ hrk+1 i ⊂ · · ·
be an ascending chain of (principal) ideals. Show that this sequence becomes stationary,12
i.e., there is an index k0 ∈ N, such that

hrk i = hrk0 i for all k ≥ k0 .

ii) Use Part i) to show that (F1) (see Page 22) holds in a principal domain (compare
[8], Chapter II, Lemma 4.3.4) and conclude that a principal ideal domain is factorial.
I.6.3 Example. Let k be a field. By Exercise I.4.15, i), the polynomial ring k[x] is a
principal ideal domain and, therefore, by the previous exercise a factorial ring.
The next aim is to prove the existence of more factorial rings.

I.6.4 Theorem (Gauß). Let R be a factorial ring. Then, the polynomial ring R[x] is
factorial, too.

By Example I.6.3, the theorem is true, if R is a field. We would like to use this result.
This is possible, because we may associate with any integral domain in a canonical way
a field.

Quotient Fields
In order to define what a quotient field is, we will recur again to a universal property.
Let R be an integral domain. A quotient field of R is a pair (Q(R), ι) which consists of
a field Q(R) and an injective homomorphism ι : R −→ Q(R) and satisfies the following
12
This property will be studied in detail in Chapter II.1.

24
 I.6. Factorial Rings

property: For every field K and every injective homomorphism ϕ : R −→ K, there is a

unique homomorphism Φ : Q(R) −→ K with

Φ ◦ ι = ϕ.

The corresponding diagram looks as follows:

R
③ ❄❄❄
ι ③③③ ❄❄ϕ
③③③ ❄❄
❄
}③ ∃!Φ
Q(R) ❴ ❴ ❴ ❴ ❴ ❴ ❴ / K.

The universal property expresses that Q(R) is the smallest field that contains R. In order to
construct it, we obviously have to invert the elements of R \ {0}. The formal construction
proceeds along the lines of the construction of the field Q of rational numbers from the
ring Z of integers ([27], Abschnitt 1.5).
For (a, b), (c, d) ∈ R × (R \ {0}), we write

(a, b) ∼ (c, d) :⇐⇒ a · d = b · c.

I.6.5 Proposition. The relation “∼” is an equivalence relation on R × (R \ {0}).

Proof. We leave this as an exercise. 
In the following, we write
a
b
for the equivalence class [a, b] of (a, b) ∈ R × (R \ {0}). We will also abusively write a for
the class a/1, a ∈ R. In this notation, we declare the addition

+ : Q(R) × Q(R) −→ Q(R)

 
a c a·d+b·c
, 7−→
b d b·d
and the multiplication

· : Q(R) × Q(R) −→ Q(R)

a c  a·c
, 7−→ .
b d b·d
I.6.6 Theorem. i) The tuple (Q, 0, +, ·, 1) is a field.
ii) The map

ι : R −→ Q(R)
a
a 7−→ a =
1
is an injective ring homomorphism.13
iii) The pair (Q(R), ι) is a quotient field of R.
Proof. Everything works as for Z and Q. So, we leave the proof as an exercise. 
13
This justifies our abusive notation.

25
I. Basic Theory of Rings and their Ideals

Greatest Common Divisors

We need some more concepts in factorial rings which generalize their counterparts in the
ring Z of integers in order to compare factorizations in the rings R[x] and Q(R)[x].
For ring elements a1 , ..., an ∈ R, a common divisor of a1 , ..., an is an element d ∈ R
with
d|ai , i = 1, ..., n,
and

cd(a1 , ..., an) := d ∈ R | d is a common divisor of a1 , ..., an
is the set of common divisors of a1 , ..., an.

I.6.7 Properties. Let n ≥ 1 be a positive natural number, a1 , ..., an ∈ R ring elements and
u ∈ R⋆ a unit.
i) For every ring element d ∈ R we have

d ∈ cd(a1 , ..., an) ⇐⇒ hdi ⊃ ha1 i + · · · + han i.

ii) It is always true that R⋆ ⊂ cd(a1 , ..., an).

iii) We have cd(a1 , ..., an, u) = R⋆ .
iv) We have cd(a1 , ..., an, 0) = cd(a1 , ..., an).
v) The property 0 ∈ cd(a1 , ..., an) holds if and only if a1 = · · · = an = 0.

Proof. For i), observe that, for a ∈ R, d|a holds if and only if hdi ⊃ hai (Property I.5.3,
vi). So,


d ∈ cd(a1 , ..., an) ⇐⇒ hdi ⊃ ha1 i ∪ · · · ∪ han i .
The fact that hdi is an ideal implies


hdi ⊃ ha1 i ∪ · · · ∪ han i ⇐⇒ hdi ⊃ ha1 i + · · · + han i.

The rest of the asserted properties is straightforward to verify, and we leave the proofs
to the reader. 

The elements a1 , ..., an ∈ R are coprime, if

cd(a1 , ..., an) ⊂ R⋆ .

By Property I.6.7, ii), this condition is equivalent to cd(a1 , ..., an) = R⋆ .

Given a1 , ..., an ∈ R, a greatest common divisor of a1 , ..., an is an element d ∈ cd(a1 , ...,
an ) with the property
∀d ′ ∈ cd(a1 , ..., an) : d ′ |d.
We let

gcd(a1 , ..., an) = d ∈ R | d is a greatest common divisor of a1 , ..., an

be the set of greatest common divisors of a1 , ..., an.

26
 I.6. Factorial Rings

I.6.8 Properties. Let a1 , ..., an ∈ R be ring elements.

i) Let d, d ′ ∈ R be elements with d ∈ gcd(a1 , ..., an) and d ∼ d ′. Then, d ′ ∈ gcd(a1 , ...,
an ).
ii) If d, d ′ ∈ gcd(a1 , ..., an) are greatest common divisors of a1 , ..., an, then d ∼ d ′ , i.e.,
a greatest common divisor is determined up to units.
iii) Suppose there is an index i0 ∈ { 1, ..., n } with ai0 , 0 and d ∈ gcd(a1 , ..., an). Then,
the elements a′1 , ..., a′n ∈ R with ai = d · a′i are coprime.

Proof. i) This is obvious. ii) This is a direct consequence of Property I.5.3, v). iii) Let
t ∈ cd(a′1 , ..., a′n) and write a′i = a′′i · t for a suitable ring element a′′i ∈ R, i = 1, ..., n. It
follows that d · t ∈ cd(a1 , ..., an). By definition of a greatest common divisor, (d · t)|d. As
usual, we infer that t ∈ R⋆ is a unit. So, we have shown cd(a′1 , ..., a′n) ⊂ R⋆ as required. 

The concept of a greatest common divisor has been defined in any integral domain. In
general, it need not exist.
√
I.6.9 Example.
√ We look again at the ring Z [ −5].√ The element a = 9 has the divisors√ ±1,
±3, ±(2 ± −5),√ ±9, and the element b := 3 · (2 + −5)
√ has the divisors ±1, ±3, ±(2 + −5)
and ±3 · (2 + −5). Since the elements 3 and (2 + −5) are not associated, it follows that
the elements a and b do not have a greatest common divisor.

I.6.10 Proposition. Assume that R is a factorial ring. Then, for n ≥ 1, ring elements
a1 , ..., an which are not all zero do have a greatest common divisor.

Proof. By Property I.6.7, iv) and v), we may suppose ai , 0, i = 1, ...., n, and by Property
I.6.7, iii), we may assume ai < R⋆ , i = 1, ..., n. Since R is factorial, we may find prime
elements p1 , ..., pr with pi / p j for 1 ≤ i < j ≤ r and natural numbers k j (ai ), j = 1, ..., r,
i = 1, ..., n, with
ai ∼ pk11 (ai ) · · · · · prkr (ai ) , i = 1, ..., n.

Now, set

m j := min k j (ai ) | i = 1, ..., n , j = 1, ..., r.

It is readily verified that

d := pm1 1 · · · · · pmr r

is a greatest common divisor of a1 , ..., an. 

I.6.11 Lemma. Let R be a factorial ring. Then, every element x ∈ Q(R) can be written
as x = a/b with a ∈ R and b ∈ R \ {0} coprime.

Proof. We pick α ∈ R and β ∈ R \ {0}. By Proposition I.6.10, α and β have a greatest

common divisor. Let d be one, a ∈ R and b ∈ R \ {0} be elements with α = a · d and
β = b · d. According to Property I.6.8, iii), a and b are coprime. It is also clear that
x = α/β = a/b. 

27
I. Basic Theory of Rings and their Ideals

Primitive Polynomials
For the rest of this section, we assume that R is a factorial ring. We need to compare
irreducible elements in the rings R[x] and Q(R)[x]. The key concept for doing so is the
one of a primitive polynomial: A polynomial f ∈ R[x] is primitive, if its coefficients are
coprime.
I.6.12 Example. If k is a field, then every polynomial f ∈ k[x] \ {0} is primitive.

I.6.13 Lemma. i) Let f ∈ R[x] be a polynomial of positive degree. If f is irreducible,

then f is primitive.
ii) If f ∈ R[x] is a primitive polynomial and f is irreducible in Q(R)[x], then f is
irreducible in R[x].

In i), we need to assume that f is non-constant, because the element 2 ∈ Z[x] is

irreducible but not primitive. Part ii) does not work, if we don’t assume that f is primitive.
In fact, the polynomial 2x+2 ∈ Q[x] is irreducible. In Z[x], the equation 2·(x+1) = 2x+2
expresses 2x + 2 as a product of two non-units, so that 2x + 2 is not an irreducible element
of Z[x]. (This is the only subtlety of the lemma.)

Proof of Lemma I.6.13. i) Let d be a greatest common divisor of the coefficients of f .

(This exists by Proposition I.6.10.) By Property I.6.8, iii), there is a primitive polynomial
g ∈ R[x] with f = d · g. Since f is irreducible and g ∈ R[x] is not a unit, because
deg(g) = deg( f ) > 0, d must be a unit.
ii) Let g, h ∈ R[x] polynomials with f = g · h. This is also an equation in Q(R)[x]. By
assumption, g ∈ Q(R)[x]⋆ = Q(R) \ {0} or h ∈ Q(R)⋆ . It suffices to treat the first case. We
have g ∈ R \ {0}, and g clearly is a common divisor of the coefficients of f . Since f is
primitive, we have g ∈ R⋆ = R[x]⋆ as required. 

I.6.14 Lemma (Gauß). Let f, g ∈ R[x] be primitive polynomials. Then, f · g ∈ R[x] is

primitive, too.

Proof. For every prime element p ∈ R, we have the surjection

̺ p : R[x] −→ (R/hpi)[x]
a0 + a1 · x + · · · + an · xn 7−→ [a0 ] + [a1 ] · x + · · · + [an ] · xn .

Formally, it is associated with the homomorphism π p : R −→ R/hpi and the element

x ∈ (R/hpi)[x] (see Page 5). Using prime factorization in R, we have the following:

∀ f ∈ R[x] : f is primitive ⇐⇒ ∀ prime elements p ∈ R : ̺ p ( f ) , 0. (I.7)

By assumption, we have ̺ p ( f ) , 0 and ̺ p (g) , 0, p ∈ R a prime element. Now, hpi is

a prime ideal (Proposition I.5.7, ii). So, R/hpi and (R/hpi)[x] are integral domains. The
inequality
∀ prime elements p ∈ R : ̺ p ( f · g) = ̺ p ( f ) · ̺ p (g) , 0
gives the result. 

28
 I.6. Factorial Rings

I.6.15 Remark. We can rephrase the above proof also in terms of ideals. The kernel p of
̺ p is a prime ideal (see Proposition I.4.1, i), p ∈ R a prime element. In the proof, we have
exploited the property
f < p ∧ g < p =⇒ f · g < p
of the prime ideal p.

I.6.16 Lemma. i) For every polynomial g ∈ Q(R)[x] \ {0}, there is a number a ∈ Q(R),
such that the polynomial a · g belongs to R[x] and is primitive.
ii) If f, g ∈ R[x] are polynomials, g is primitive, and a ∈ Q(R) is a number with
f = a · g, then a ∈ R.

Proof. i) We write g = α0 + α1 · x + · · · + αn · xn with α0 , ..., αn ∈ Q(R) and pick elements

ri ∈ R, si ∈ R \ {0} with αi = ri /si , i = 0, ..., n. Set s := s0 · · · · · sn . Then, we obviously
have s · g ∈ R[x] \ {0}. Since R is factorial, we may find a greatest common divisor d of
the coefficients of s · g. There exists a polynomial f ∈ R[x] with s · g = d · f . By Property
I.6.8, iii), f is primitive. Altogether, we may choose a = s/d.
ii) The case f = 0 is trivial. Otherwise, we may write a = r/s with r, s ∈ R coprime
(see Lemma I.6.11). The equation f = a · g may be rewritten as

s · f = r · g.

This shows that s divides all the coefficients of r · g. Since s is coprime to r, prime
factorization in R implies that s divides all the coefficients of g. The primitivity of g
implies that s is a unit. So, a = r/s ∈ R as asserted. 

The following result finally relates the irreducible elements of R[x] and Q(R)[x].

I.6.17 Proposition. i) Let f ∈ R[x] be a non-constant primitive polynomial and g ∈ R[x].

If the relation f |g holds in Q(R)[x], then it holds in R[x], too.
ii) If f ∈ R[x] is a non-constant irreducible polynomial, then f is irreducible as an
element of Q(R)[x].

Proof. i) If g = 0, there is nothing to show. Otherwise, there is a polynomial h ∈ Q(R)[x]\

{0} with g = f · h. By Lemma I.6.16, i), there is a number a ∈ Q(R), such that a · h is a
primitive polynomial in R[x]. Now, we look at the equation
1
g= · f · (a · h).
a
By Lemma I.6.14, f · (a · h) is primitive. According to Lemma I.6.16, ii), 1/a ∈ R, so that
h = (1/a) · (a · h) ∈ R[x].
ii) Suppose we could write f = g · h with g, h ∈ Q(R)[x]. We pick a ∈ Q(R), such that
a · h ∈ R[x] is a primitive polynomial, and look at the equation
!
1
f = · g · (a · h).
a

By Part i), we have (1/a) · g ∈ R[x]. Since f is irreducible as an element of R[x], we

conclude that (1/a) · g ∈ R⋆ or a · h ∈ R⋆ . This implies g ∈ Q(R)⋆ or h ∈ Q(R)⋆ . 

29
I. Basic Theory of Rings and their Ideals

Proof of Theorem I.6.4. We will verify Conditions (F1) and (F3) (see Page 22f).
Step 1. We show that (F1) holds, i.e., that every element of R[x] \ ({0} ∪ R⋆ ) may be
written as a product of irreducible elements. We do this by induction on the degree n.
n = 0. Let f ∈ R[x] \ ({0} ∪ R⋆ ) be a non-zero constant. Since R is factorial, f can be
written as a product of irreducible elements in R. Finally, every irreducible element of R
is an irreducible element of R[x].
n −→ n + 1. Let f ∈ R[x] be a polynomial of degree n + 1. There are a ring element
a ∈ R and a primitive polynomial g ∈ R[x] with f = a · g. Since the assertion holds
for a, it suffices to factorize g. If g is irreducible, there is nothing to show. If g is not
irreducible, there exist elements g1 , g2 ∈ R[x] \ ({0} ∪ R⋆ ) with g = g1 · g2 . Since g is
primitive and g1 , g2 aren’t units, g1 and g2 cannot be constant. It follows deg(g1 ) < deg(g)
and deg(g2 ) < deg(g). By induction hypothesis, g1 and g2 may be written as products of
irreducible polynomials. The same is true for g = g1 · g2 .
Step 2. Here, we check (F3). Let c1 , ..., cm , d1, ..., dn ∈ R be irreducible elements and
p1 , ..., p s , q1 , ..., qt ∈ R[x] \ R be irreducible polynomials of positive degree, such that

c1 · · · · · cm · p1 · · · · · p s = d1 · · · · · dn · q1 · · · · · qt . (I.8)

By Lemma I.6.13, i), the polynomials p1 , ..., p s and q1 , ..., qt are primitive. By Lemma
I.6.14, the polynomials p1 · · · · · p s and q1 · · · · · qt are primitive, too. It is easy to infer from
Equation (I.8) that
c1 · · · · · cm ∼ d1 · · · · · dn .
Since R is a factorial ring, m = n and there is a permutation σ : { 1, ..., n } −→ { 1, ..., n }
with
ci ∼ dσ(i) , i = 1, ..., n.
We conclude
p1 · · · · · p s ∼ q1 · · · · · qt . (I.9)
Now, we look at this relation in the ring Q(R)[x]. By Proposition I.6.17, ii), the poly-
nomials p1 , ..., p s , q1, ..., qt are irreducible in Q(R)[x]. We already know that Q(R)[x] is
factorial (see Example I.6.3). From (I.9), we now infer that s = t and that there is a
permutation τ : { 1, ..., t } −→ { 1, ..., t } with

pi ∼ qτ(i) in Q(R)[x], i = 1, ..., t.

With Lemma I.5.5, ii), we rewrite this as

pi |qτ(i) and qτ(i) |pi in Q(R)[x], i = 1, ..., t.

Next, Proposition I.6.17 shows

pi |qτ(i) and qτ(i) |pi in R[x],

and, thus, by Lemma I.5.5, ii),

pi ∼ qτ(i) in R[x], i = 1, ..., t.

This gives the assertion and concludes the proof. 

30
 I.7. The Nilradical

I.7 The Nilradical

I.7.1 Proposition. Let R be a ring and

N := a ∈ R | a is nilpotent .
i) The subset N of R is an ideal.
ii) The quotient ring R/N has no nilpotent element other than 0.
Proof. i) Clearly, 0 ∈ N. Suppose a, b ∈ N and choose exponents m ≥ 1 and n ≥ 1 with
am = 0 and bn = 0. For 0 ≤ i < m, we have m + n − 1 − i ≥ n. This shows that every
summand of the right hand sum in
X m + n − 1!
m+n−1
m+n−1
(a + b) = · ai · bm+n−1−i
i=0
i

is zero, so that (a + b)m+n−1 = 0 and a + b ∈ N. Assume r ∈ R and a ∈ N and let n ≥ 1

be an exponent with an = 0. Then, (r · a)n = rn · an = 0 and r · a ∈ N. In the special case
r = −1, we get −a ∈ N.
ii) Let a ∈ R be such that [a] ∈ R/N is nilpotent. Let n ≥ 1 be such that [an ] = [a]n = 0.
This means an ∈ N. Hence, we may find an exponent m ≥ 1 with
am·n = (an )m = 0.
This shows a ∈ N and [a] = 0. 
The ideal N is called the nilradical of R.
I.7.2 Proposition. Let R , {0} be a ring. Then, the nilradical is the intersection of all
prime ideals in R: \
N= p.
p⊂R
prime ideal

Proof. We first show that the nilradical is contained in every prime ideal. To this end, let
a ∈ N be a nilpotent element, n ≥ 1 an exponent with an = 0, and p ⊂ R a prime ideal.
We obviously have an ∈ p. So, we may define

l := min m ≥ 1 | am ∈ p .
Assume l > 1. Then, al = a · al−1 ∈ p. By definition, a ∈ p or al−1 ∈ p. Both conclusions
contradict the choice of l. The only way out is l = 1 and a ∈ p.
Now, let a ∈ R be an element which is not nilpotent. We will prove the existence of
a prime ideal which does not contain a. Denote by Σ the set of ideals I ⊂ R with the
property that an < I for all n ≥ 1. This set contains the zero ideal {0} and is, therefore,
non-empty. By Zorn’s lemma I.4.7 (compare the proof of Theorem I.4.4), it contains a
maximal element. Let p ∈ Σ be a maximal element. If we can show that p is a prime
ideal, we are clearly done. Suppose b, c ∈ R \ p. Then, p ( p + hbi and p ( p + hci. By
definition of p, there are exponents m ≥ 1 and n ≥ 1 with am ∈ p + hbi and an ∈ p + hci.
We infer
am+n ∈ p + hb · ci.
So, p + hb · ci < Σ and p ( p + hb · ci. This shows b · c < p as required. 

31
I. Basic Theory of Rings and their Ideals

I.7.3 Remark. In the above proof, we have used Zorn’s lemma, i.e., the axiom of choice
in its full strength. In Section II.3, we will see that the statement that Proposition I.7.2
holds for every non-zero ring is equivalent to the fact that every non-zero ring has a prime
ideal (compare Remark I.4.5). Note the fact that, in a non-zero ring, 1 < N, so that R must
contain a prime ideal, if Proposition I.7.2 holds.14

I.8 Operations on Ideals

Here, we will discuss various ways to construct new ideals from given ones. These con-
structions and their properties will be used throughout the following text.

Intersections, Sums, and Products

Let R be a ring, S an index set and (Is )s∈S a family of ideals in R indexed by the set S .
Then, it is a straightforward task to verify that the intersection
\
Is
s∈S

is also an ideal. This basic observation makes possible the following important construc-
tion:

I.8.1 Lemma. Let R be a ring and X ⊂ R. The ring R possesses one and only one ideal
I(X) which contains X and is contained in any ideal containing X.

Proof. Obviously, the ideal \

I(X) := I
I⊂R ideal
X⊂I

does the trick. 

The reader should compare this to the corresponding construction in group theory
([30], Satz II.4.10). The ideal I(X) is the ideal generated by X.

Definition. We will often write hXi for I(X).

I.8.2 Remark. The ideal I(X) is identical to the ideal of all finite R-linear combinations of
elements in X:
X  
I(X) = ra · a  ra ∈ R, a ∈ X, all but finitely many are zero .
a∈X

I.8.3 Examples. i) The ideal generated by the empty set is the zero ideal, h∅i = {0}.
ii) For a ∈ R, the ideal I({a}) agrees with the principal ideal hai generated by a (see
Example I.2.1, iv).
iii) More generally, for finitely many elements a1 , ..., an ∈ R, we have (compare Ex-
ample I.2.1, x)


I { a1, ..., an } = h a1 , ..., an i = ha1 i + · · · + han i.
14
The intersection over an empty index set is, by definition, the whole ring R.

32
 I.8. Operations on Ideals

We say that an ideal I ⊂ R is finitely generated, if there are a natural number n and
elements a1 , ..., an ∈ R with I = h a1 , ..., an i.
The above construction has an important special case: For S an index set and a family
of ideals (Is )s∈S in R indexed by the set S , the sum is the ideal
X [  X 
Is := Is = a s | a s ∈ Is , s ∈ S , all but finitely many zero .
s∈S s∈S s∈S
P
I.8.4 Examples. i) For a subset X ⊂ R, we find I(X) = hai.
a∈X
ii) The union of ideals is, in general, not an ideal. This means that we get, in general,
[ X
Is ( Is .
s∈S s∈S

For example, h2i ∪ h3i ⊂ Z is not an ideal. We have 2 ∈ (h2i ∪ h3i) and 3 ∈ (h2i ∪ h3i),
but 5 < (h2i ∪ h3i). Note also that

h2i + h3i = Z,

because 1 = −2 + 3 ∈ h2i + h3i.

For ideals I, J ⊂ R, the product is the ideal

X 
 n
I · J = I { a · b | a ∈ I, b ∈ J } = ai · bi  n ∈ N, ai ∈ I, bi ∈ J, i = 1, ..., n .
i=1

The powers of an ideal I ⊂ R are defined recursively via

⋆ I 0 := R,

⋆ I n+1 := I · I n , n ∈ N.

I.8.5 Examples. i) Let R = Z and a, b ∈ Z integers. We find the following description of

the above constructions:

⋆ hai + hbi = hci with c the greatest common divisor of a and b.

⋆ hai ∩ hbi = hci with c the least common multiple of a and b (see [30], Definition
I.4.12).

⋆ hai · hbi = ha · bi.

ii) Let k be a field, k[x1 , ..., xn] the polynomial ring in n variables (see Equation (I.2)
and the following exercise) and I := h x1 , ..., xn i. Then, for m ≥ 1,

Im = polynomials which contain only monomials of degree at least m
( X )

= ai1 ,...,in · x1 · · · · · xn  ai1 ,...,in ∈ R, (i1 , ..., in) ∈ N : i1 + · · · + in ≥ m .
i1 in  ×n

(i1 ,...,in )∈N×n :

i1 +···+in ≥m

Note that R/I m contains nilpotent elements different from zero, if m ≥ 2.

33
I. Basic Theory of Rings and their Ideals

I.8.6 Example. Let R be a ring and I, J ⊂ R ideals. We always have

I · J ⊂ I ∩ J.

In general, this inclusion is strict. For example, the discussion in Example I.8.5, i), shows
that, for integers a, b ∈ Z, the equation

hai · hbi = hai ∩ hbi

holds if and only if a and b are coprime.

We say that I and J are coprime, if I + J = R, and claim that, for coprime ideals
I, J ⊂ R, we have
I · J = I ∩ J.
We have to verify the inclusion “⊃”. Pick r ∈ I and s ∈ J with r + s = 1 and let a ∈ I ∩ J.
Then, the equations
a = a · 1 = a · (r + s) = |{z}
a · r + |{z}
a·s
∈I·J ∈I·J

shows that a is an element of I · J.

I.8.7 Remarks. i) The operations of taking intersections, sums, and products of ideals are
commutative and associative, and the following distributive law holds:

For ideals I, J, K ⊂ R : I · (J + K) = I · J + I · K.

The reader may verify this as an exercise.

ii) Let I, J, K be ideals of the ring R and assume J ⊂ I or K ⊂ I. Then, we also have

I ∩ (J + K) = (I ∩ J) + (I ∩ K).

Here, the inclusion “⊃” is obvious. For the converse inclusion, we assume J ⊂ I. Let
b ∈ J and c ∈ K be elements with b + c ∈ I. Together with b ∈ I, this implies c ∈ I.
I.8.8 Examples. i) Let k be a field and R = k[x, y, z] the polynomial ring in three variables
over k. We assert
h x · y, x · z, y · z i = h x, y i ∩ h x, z i ∩ h y, z i.
The inclusion “⊂” is immediate. For the converse, we apply the rules from Remark I.8.7.
We have


h x, y i ∩ h x, z i = hxi + hyi ∩ h x, z i



= hxi ∩ h x, z i + hyi ∩ h x, z i .

Since hyi and h x, z i are not coprime, the rules do not give the intersection hyi ∩ h x, z i.
Here, we apply prime factorization. Let f ∈ h x, z i. Write f as a product of irreducible
factors. Since y is irreducible, it must be associated with one of these irreducible factors,
provided f ∈ hyi. We see f ∈ h x · y, y · z i and, thus, hyi ∩ h x, z i ⊂ h x · y, y · z i. The
converse inclusion is obvious. We continue as follows:



hxi ∩ h x, z i + hyi ∩ h x, z i = hxi + h x · y, y · z i = h x, y · z i.

34
 I.8. Operations on Ideals

Next


h x, y · z i ∩ h y, z i = hxi + hy · zi ∩ h y, z i



= hxi ∩ h y, z i + hy · zi ∩ h y, z i
= hx · y, x · zi + h y · z i
= h x · y, x · z, y · z i.

ii) If we have two finitely generated ideals I = h f1 , ..., f s i and J = h g1 , ..., gt i in a ring
R, then one easily sees

I · J = h f1 · g1 , ..., f1 · gt , ..., f s · g1 , ..., f s · gt i.

A similar result holds, if the ideals are specified by possibly infinite sets of generators.
We may apply this to the setting of the previous example:

h x, y i · h x, z i · h y, z i = h x2 · y, x2 · z, x · y2 , x · y · z, x · z2 , y2 · z, y · z2 i.

We infer
h x, y i · h x, z i · h y, z i ( h x, y i ∩ h x, z i ∩ h y, z i.
√
I.8.9 Exercise (A primary decomposition in Z[ −5]). Define
√ √
p1 := h 3, 2 + −5 i and p2 := h 3, 2 − −5 i

i) Show that p1 and p2 are prime ideals.

ii) Verify that the ideals p21 and p22 are coprime.
iii) Demonstrate that h9i = p21 · p22 = p21 ∩ p22 .

The Chinese Remainder Theorem

We generalize Example I.8.6.
I.8.10 Lemma. Let R be a ring and I1 , ..., In ⊂ R ideals, such that Ik and Il are coprime
for 1 ≤ k < l ≤ n. Then,15
Y
n \
n
Ik := I1 · · · · · In = Ik .
k=1 k=1

Proof. We proceed by induction on n. The case n = 2 has already been dealt with in
Example I.8.6.
n − 1 −→ n. We set
Y
n−1
J := Ik .
k=1

By induction hypothesis, we have

\
n−1
J= Ik .
k=1
15
The associativity of the product shows that the definition is independent of the brackets one implicitly
has to insert into the middle term, or, in other words, grants that the middle term is well-defined.

35
I. Basic Theory of Rings and their Ideals

It suffices to show that J and In are coprime. By assumption, there are elements rk ∈ Ik
and sk ∈ In with
rk + sk = 1, k = 1, ..., n − 1.
There is an element s ∈ In with
Y
n−1 Y
n−1
rk = (1 − sk ) = 1 − s.
k=1 k=1

Since the element on the left hand side belongs to J, we see that 1 ∈ J + In . 
Now, we would like to generalize the Chinese remainder theorem from elementary
number theory in the formulation [30], III.1.4. We place ourselves in the situation of the
lemma. To make the notation more precise, we denote the class of a ∈ R in the quotient
ring R/Ik by [a]k , k = 1, ..., n. For k = 1, ..., n, there is the canonical surjection

πk : R −→ R/Ik
a 7−→ [a]k .

Using the cartesian product of rings (Example I.1.3, x), we define

n

ϕ : R −→ R/Ik
k=1


a 7−→ [a]1 , ..., [a]n .

Note that
\
n
ker(ϕ) = Ik .
k=1

By the isomorphism theorem, we get an induced injective homomorphism

\
n n

ϕ : R/ Ik −→ R/Ik .
k=1 k=1

Next, we would like to study when ϕ or, equivalently, ϕ is surjective. Let

n

ei := (0, ..., 0, 1, 0, ..., 0) ∈ R/Ik
k=1

be the element which has the entry 1 at the i-th place, i = 1, ..., n. Then, it is readily
checked that
ϕ is surjective ⇐⇒ ∀i ∈ { 1, ..., n } : ei ∈ im(ϕ).
Let us look at the condition e1 ∈ im(ϕ) in more detail. If it is satisfied, then there is an
element a ∈ R with [a]1 = 1, i.e., 1 − a ∈ I1 , and [a]k = 0, that is a ∈ Ik , for k = 2, ..., n. In
particular, we have
1 = (1 − a) + a ∈ I1 + Ik , k = 2, ..., n.
Conversely, assume that I1 and Ik are coprime, for k = 2, ..., n. Then, there are elements
rk ∈ I1 and sk ∈ Ik with
1 = rk + sk , k = 2, ..., n.

36
 I.8. Operations on Ideals

Set
Y
n Y
n
a := sk = (1 − rk ).
k=2 k=2

The second description of a shows [a]1 = 1 and the first [a]k = 0, k = 2, ..., n. It follows
ϕ(a) = e1 . Our discussion shows:

I.8.11 Lemma. The homomorphism ϕ is surjective if and only if the ideals Ik and Il are
coprime, for 1 ≤ k < l ≤ n.

We combine this observation with Lemma I.8.10:

I.8.12 Chinese remainder theorem. Let R be a ring and I1 , ..., In ⊂ R ideals, such that Ik
and Il are coprime for 1 ≤ k < l ≤ n. Then,

\
n Y
n n

R/ Ik = R/ Ik  R/Ik .
k=1 k=1 k=1

I.8.13 Example. Let a ∈ Z be an integer and

a = pν11 · · · · · pνnn

its prime factorization. Then,

Z/hai  Z/hpν1 i × · · · × Z/hpνn i.

1 n

I.8.14 Exercise (The Chinese remainder theorem). i) What is the smallest (positive) mul-
tiple of 10 which has remainder 2 when divided by 3, and remainder 3 when divided by
7?
ii) Let k be a field. Describe the ring k[x]/hx2 − 1i.

Ideal Quotients
Let R be a ring and I, J ⊂ R ideals. Then, the ideal quotient of I by J is set to be

(I : J) := a ∈ R | a · J ⊂ I .

It is straightforward to check that (I : J) is an ideal. For an ideal I ⊂ R, we call (h0i : I)

the annihilator of I.

Notation. ⋆ Ann(I) := (h0i : I).

⋆ For a ∈ R, we set Ann(a) := Ann(hai).

I.8.15 Remark. The union [

D := Ann(a)
a∈R\{0}

is the set of zero divisors of the ring R.

37
I. Basic Theory of Rings and their Ideals

I.8.16 Example. Let R = Z and a, b ∈ Z be integers. We look at the prime factorizations

Y Y
a= pµ p and b = pν p .
p prime number p prime number

There exists a positive integer c with

hai : hbi = hci.
Its prime factorization Y
c= pγ p
p prime number

is determined by
γ p = max{ µ p − ν p , 0 } = µ p − min{ µ p , ν p }, p a prime number.
We see
a
c= .
gcd(a, b)
I.8.17 Properties. Let R be a ring, S and T index sets, and I, J, K, Is , s ∈ S , and Jt ,
t ∈ T , ideals in R. Then, the following properties are verified:
i) I ⊂ (I : J).
ii) (I : J) · J ⊂ I.




iii)  (I : J) : K = I : (J · K) = (I : K) : J .
\ \
iv) (Is : J) = (Is : J).
s∈S
 X  \ s∈S

v) I : Jt = (I : Jt ).
t∈T t∈T

The verifications are left as an exercise to the reader.

Radicals of Ideals
Let R be a ring and I ⊂ R an ideal. The radical of I is
√ 
I := a ∈ R | ∃k ≥ 1 : ak ∈ I .
√
Note that I is the preimage of the nilradical N of the ring R/I under the canonical pro-
jection π : R −→ R/I. Proposition I.7.2, therefore, gives the following
√
I.8.18 Corollary. The radical I is the intersection of all prime ideals which contain I.
I.8.19 Properties. Let R be a ring, I, J ⊂ R ideals, and p ⊂ R a prime ideal of R. Then,
we have the √ following properties:
i) Iq⊂ I.
√ √
ii) I = I.
√
iii) √ I = R if√and only√if I =√R.
iv) I · J = qI ∩ J = I ∩ J.
√ √ √
v) I + J = I + J.
√
vi) For every n > 0, one has pn = p.

38
 I.8. Operations on Ideals

The proof is again left to the reader.

I.8.20 Examples. i) Let R = Z and a ∈ Z an integer. As usual, we look at the prime
factorization
a = pν11 · · · · · pνnn .
This means, in particular, that νi > 0, i = 1, ..., n. Then, it is readily checked that
p
hai = hp1 · · · · · pn i.

ii) We return to Example I.8.8, ii). More precisely, we look at the ideal

I = h x2 · y, x2 · z, x · y2 , x · y · z, y2 · z, y · z2 i.

Note x · y < I, but (x · y)2 = x · (x · y2 ) ∈ I.

√
An ideal I ⊂ R which is its own radical, I = I, is a radical ideal. Radical ideals play
an important role in algebraic geometry (see Section I.9).

Extension and Contraction of Ideals

Let R, S be rings and ϕ : R −→ S a homomorphism. It is important to compare the ideals
of R and S .16 Recall from Example I.2.1, viii), that the image ϕ(I) of an ideal I ⊂ R need
not be an ideal. For this reason, we have to resort to the construction in Lemma I.8.1: Let
I ⊂ R. Then, the ideal


I e := I ϕ(I)
of S generated by the image ϕ(I) of I is the extension of I via ϕ. We have
X
n  
e
I = si · f (ai )  n ∈ N, si ∈ S , ai ∈ I, i = 1, ..., n .
i=1

I.8.21 Remarks. If p ⊂ R is a prime ideal, then pe ⊂ S need not be a prime ideal.17

For example, for the inclusion ι : Z −→ Q, the extension of hpi, p a prime number,
is always Q which is not a prime ideal. Another example is the canonical projection
π : k[x, y] −→ k[x, y]/hx · yi, k a field. Since k[x, y] is an integral domain, h0i is a prime
ideal of k[x, y]. Obviously, h0ie = h0i. But the quotient ring k[x, y]/hx·yi has zero divisors,
e.g., [x] and [y]. Therefore, h0i ⊂ k[x, y]/hx · yi is not a prime ideal.
In the above setting, let J ⊂ S be an ideal. Then,

J c := ϕ−1 (J)

is always an ideal of S . It is called the contraction of J via ϕ. Recall that qc is a prime

ideal of R, if q is a prime ideal of S .
16
In Exercise I.4.17, iv), you have already encountered the induced continuous map ϕ# : Spec(S ) −→
Spec(R).
17
So, there is, in general, no induced map ϕ⋆ : Spec(R) −→ Spec(S ).

39
I. Basic Theory of Rings and their Ideals

I.8.22 Remark. By Exercise I.2.4, we can factorize ϕ : R −→ S as

π ι
R −→ T := R/ker(ϕ) ֒→ S .
Here, π is the canonical projection and ι := ϕ is injective. Regarding the surjection
π : R −→ T,
we have
⋆ If I ⊂ R is an ideal, then π(I) ⊂ T is an ideal (Example I.2.1).
⋆ The map K ⊂ R/ker(ϕ) 7−→ π−1 (K) ⊂ R induces an inclusion preserving bijection
between the set of ideals of R/ker(ϕ) and the set of ideals of R containing ker(ϕ).18
The above discussion reveals that most of the complications encountered in the exten-
sion and contraction of ideals show up for injective ring homomorphisms. Here is a nice
classical example from number theory.
I.8.23 Example. We look at the ring of Gaußian integers:
Z[i] =  k + l · i ∈ C | k, l ∈ Z .
This ring is euclidean (see [8], Chapter II, Definition 2.2.4, for the definition of a euclidean
ring and [8], Beispiel 2.2.5, 3), or [23], I.(1.2), for the proof of the above-mentioned fact).
In particular, it is a principal ideal domain ([8], Satz 2.2.6). The prime elements of Z[i]
are described in [23], I.(1.4).
Let ϕ : Z −→ Z[i] be the natural inclusion. Using the classification of prime elements
in Z[i], one finds the following:
⋆ h2ie = h(1 + i)2 i.
⋆ If p ≡ 1 mod 4, then hpie is the product of two prime ideals. For example, h5ie =
h2 + ii · h2 − ii.
⋆ If p ≡ 3 mod 4, then hpie is a prime ideal.
This is just the ideal theoretic description of the classical result of number theory that an
odd prime number can be written as the sum of two squares if and only if it is congruent
to 1 modulo 4 (see [23], Chapter I, §1).
We list further rules for the extension and contraction of ideals which the reader should
prove on her or his own.
I.8.24 Properties. Let R, S be rings, I, I1 , I2 be ideals in R, J, J1 , J2 ideals in S , and
f : R −→ S a ring homomorphism. Then:
i) I ⊂ I ec , J ⊃ J ce .
ii) I e = I ece , J c = J cec .
iii) (I1 + I2 )e = I1e + I2e , (J1 + J2 )c ⊃ J1c + J2c .
iv) (I1 ∩ I2 )e ⊂ I1e ∩ I2e , (J1 ∩ J2 )c = J1c ∩ J2c .
v) (I1 · I2 )e = I1e · I2e , (J1 · J2 )c ⊃ J1c · J2c .
vi) (I1 : I2 )e ⊂ (I1e : I2e ), (J1 : J2 )c ⊂ (J1c : J2c ).
√ √ √ √
vii) ( I)e ⊂ I e , ( J)c = J c .
18
Be aware that, for distinct ideals I and I ′ which do not both contain ker(ϕ), it may very well happen
that ϕ(I) = ϕ(I ′ ). Can you think of an easy example?

40
 I.9. Algebraic Sets

I.9 Algebraic Sets

So far, we have developed the calculus of ideals. In addition, we studied the problem
of prime factorizations in integral domains. The failure of it led to Kummer’s idea that
ideals are generalizations of numbers which should help to save the prime factorization.
This is a strong motivation for introducing ideals. The main motivation to study ideals
in this course comes, however, from algebraic geometry. We will now illustrate how
basic notions concerning systems of polynomial equations may be elegantly treated in the
language of ideals. To do this, we will need some of the operations on ideals that we have
discussed. This is why we treat this material at such a late stage.
Roughly speaking, algebraic geometry is the study of solutions of systems of polyno-
mial equations over a field or even a ring. Let us formalize this. Let k be a field. The
n-dimensional affine space over k is19

Ank :=  (a1 , ..., an) | ai ∈ k, i = 1, ..., n .

Let F ⊂ k[x1 , ..., xn] be a subset. The set

V(F) := p = (a1 , ..., an) ∈ Ank | ∀ f ∈ F : f (p) = f (a1, ..., an) = 0

is the vanishing locus of F. Its points are those which simultaneously solve all the poly-
nomial equations f = 0, f ∈ F. Note that F may be an infinite set.
A subset Z ⊂ Ank is algebraic, if there is a subset F ⊂ k[x1 , ..., xn] with

Z = V(F).

For a subset S ⊂ Ank , we introduce


I(S ) := f ∈ k[x1 , ..., xn ] | ∀p ∈ S : f (p) = 0 .

Note that I(S ) is an ideal (compare Example I.2.1, vi). It is the ideal of regular functions
vanishing on S or simply the ideal of S .

I.9.1 Properties. Let S , Y, Z ⊂ Ank and F, G, H ⊂ k[x1 , ..., xn ] be subsets. Then, one has:
i) The ideal I(S ) of S is a radical ideal.
ii) I(Y ∪ Z) = I(Y) ∩ I(Z).
iii) If Y ⊂ Z, then I(Y) ⊃ I(Z).
iv) If G ⊂ H, then V(G) ⊃ V(H).
v) Let hFi be the ideal generated by F (see Lemma I.8.1). Then,


V hFi = V(F).

Proof. i) Let f ∈ k[x1 , ..., xn] and l ≥ 1 be such that f l ∈ I(S ). This means that f l (p) =
( f (p))l = 0 for all p ∈ S . Since the field k has no nilpotent element besides 0, this
condition is equivalent to f (p) = 0 for all p ∈ S , i.e., to f ∈ I(S ).
ii), iii), and iv) are trivial.
19
One uses this notation rather than kn to indicate that we are not interested in the k-vector space structure
of that set.

41
I. Basic Theory of Rings and their Ideals

v) Since F ⊂ hFi, we have V(F) ⊃ V(hFi). Now, let p ∈ V(F) and h ∈ hFi. We
have to show that h(p) = 0. There are a natural number s, elements f1 , ..., f s ∈ F and
r1 , ..., r s ∈ k[x1 , ..., xn ] with
X
s
h= ri · fi .
i=1
It follows that
X
s
h(p) = ri (p) · fi (p) = 0,
i=1
because fi (p) = 0, i = 1, ..., s. 
Property I.9.1, v), tells us that, when dealing with algebraic sets, we may restrict to
vanishing loci of ideals. The set theoretic operations on algebraic sets reflect very nicely
in the operations on ideals.
I.9.2 Properties. i) The empty set ∅ and the affine space Ank are algebraic sets.
ii) Let I, J ⊂ k[x1 , ..., xn] be ideals. Then,
V(I) ∪ V(J) = V(I ∩ J) = V(I · J).
iii) Let K be an index set and (Ik )k∈K be a family of ideals in k[x1 , ..., xn]. Then,
\ X 
V(Ik ) = V Ik .
k∈K k∈K

Proof. i) We have ∅ = V(h1i) and A = V(h0i). n

k
ii) Since I ∩ J ⊂ I and I ∩ J ⊂ J, we have V(I ∩ J) ⊃ V(I) ∪ V(J), by Property
I.9.1, iv). We also have I · J ⊂ I ∩ J, so that V(I · J) ⊃ V(I ∩ J). Let us show that
V(I · J) ⊂ V(I) ∪ V(J). To this end, let p ∈ V(I · J) and assume p < V(I). Then, we find a
polynomial f ∈ I with f (p) , 0. For every g ∈ J, we obviously have f · g ∈ I · J, so that
∀g ∈ J : f (p) · g(p) = ( f · g)(p) = 0.
Since f (p) , 0 and k is a field, this means
∀g ∈ J : g(p) = 0
and shows that p ∈ V(J).
iii) It is readily verified that
\ [ 
V(Ik ) = V Ik .
k∈K k∈K
P S
By definition (see Page I.8), Ik is the ideal generated by Ik . So, we conclude by
k∈K k∈K
Property I.9.1, v). 
Property I.9.2 tells us that the algebraic subsets of Ank fullfil the axioms of the closed
subsets of a topological space (see [18], Abschnitt 1.3). Therefore, we say that a subset
Z ⊂ Ank is Zariski closed, if it is algebraic. A subset U ⊂ Ank is Zariski open, if its
complement Ank \ U is Zariski closed.
By Property I.9.2

T := U ⊂ Ank | U is Zariski open
is a topology on Ank , the Zariski topology.

42
 I.9. Algebraic Sets

I.9.3 Remark. i) We already see some of the nice features of algebraic geometry emerge:
the subtle interplay of topological, geometric and algebraic tools. The reader should be
careful about the topology. It is very distinct from, e.g., the usual topology on Rn or Cn
(see the following exercise).
ii) In Property I.9.2, ii), we give two ideals which yield the union V(I) ∪ V(J), I, J ⊂
k[x1 , ..., xn] ideals, namely, I ∩ J and I · J. The first description has the advantage that
I ∩ J will be a radical ideal, if I and J are radical ideals (Property I.8.19, iv). The second
description allows to determine easily equations for V(I) ∪ V(J) from known generators
for I and J. In fact, if ( fk )k∈K generate I and (gl )l∈L generate J, then ( fk · gl )(k,l)∈K×L generate
I · J (compare Example I.8.8, ii).
I.9.4 Exercises. i) Describe the Zariski open subsets of A1k . (Recall that k[x] is a principal
ideal domain.)
ii) Check that the Zariski topology on A2k is not the product topology (see [18], Section
2.1) on A1k × A1k , the factors being endowed with the Zariski topology.
Let (X, T ) be a topological space and S ⊂ X a subset. Recall ([18], Definition, p. 11)
that the closure S of S is the smallest closed subset of X that contains S , i.e.,
\
S = Z.
Z⊂X closed:
S ⊂Z

I.9.5 Lemma. i) For every ideal I ⊂ k[x1 , ..., xn ], one has

√

I ⊂ I V(I) .
ii) Let S ⊂ Ank be a subset. Then,


S = V I(S ) .
Proof. i) Clearly, I ⊂ I(V(I)). Hence,
√ q



I ⊂ I V(I) = I V(I) .

For the last equation, we used Property I.9.1, i).

ii) We clearly have S ⊂ V(I(S )). Since V(I(S )) is closed, we also find S ⊂ V(I(S )).
Let S ⊂ Z ⊂ Ank be a closed subset. There is an ideal I ⊂ k[x1 , ..., xn] with Z = V(I). We
see



Z = V(I) ⊃ V I(Z) ⊃ V I(S ) .
The first inclusion is a consequence of I ⊂ I(Z) and the second one of I(Z) ⊂ I(S ) which,
in turn, follows from Z ⊃ S . 
We have maps
n o n o
Φ : Algebraic sets in Ank −→ Radical ideals in k[x1 , ..., xn]
Z 7−→ I(Z)
and
n o n o
Ψ : Radical ideals in k[x1 , ..., xn ] −→ Algebraic sets in Ank
I 7−→ V(I).

43
I. Basic Theory of Rings and their Ideals

Lemma I.9.5, ii), shows

Ψ ◦ Φ = id .

In particular, Φ is injective and Ψ is surjective.

Hilbert’s20 Nullstellensatz III.3.4 asserts that, if k is algebraically closed, then also

Φ ◦ Ψ = id .

Then, we have translated the theory of algebraic sets into the theory of ideals in rings.
If k is not algebraically closed, we cannot expect such a result. Look, for example,
at the prime ideal21 I := hx2 + y2 + 1i ⊂ k[x, y]. Obviously, V(I) = ∅ and I(V(I)) = h1i.
I.9.6 Examples. i) We look at the ideals Ix = h y, z i, Iy = h x, z i, and Iz = h x, y i inside
k[x, y, z]. Then, V(Ix ) is the x-axis, V(Iy ) the y-axis, and V(Iz ) the z-axis. The ideals Ix ·Iy ·Iz
and Ix ∩ Iy ∩ Iz were computed in Example I.8.8. By Property I.9.2, ii), both V(Ix · Iy · Iz )
and V(Ix ∩ Iy ∩ Iz ) consist of the union of the coordinate axes.

∪ ∪ =

Note that



Ix · Iy · Iz ( Ix ∩ Iy ∩ Iz ⊂ I V(Ix ∩ Iy ∩ Iz ) = I V(Ix · Iy · Iz ) .

We remark that Ix · Iy · Iz , I(V(Ix · Iy · Iz )) is already implied by the fact that Ix · Iy · Iz is

not a radical ideal (see Example I.8.20, ii).
ii) We look at the ideals I1 := h y2 − x3 , z i and I2 := h x, y i in the ring k[x, y, z]. The
vanishing locus V(I1 ) is Neil’s parabola ([28], Beispiel 4.1.2, iv) inside the (x, y)-plane
and V(I2 ) is, as before, the z-axis. The union of these two objects is V(I1 · I2 ). By Example
I.8.8, ii),
I1 · I2 = h x · z, y · z, x · y2 − x4 , y3 − x3 · y i.

Again, I1 · I2 is not a radical ideal. For example, y2 − x3 < I1 · I2 but

(y2 − x3 )2 = y · y · (y2 − x3 ) − x2 · x · (y2 − x3 ) = y · (y3 − x3 · y) − x2 · (x · y2 − x4 ) ∈ I1 · I2 .

We see I1 · I2 ( I(V(I1 · I2 )).

20
David Hilbert (1862 - 1943), German mathematician.
21 2
x + y2 + 1 is an irreducible polynomial, because, if it could be written as the product of two linear
polynomials, it would have zeroes.

44
 I.9. Algebraic Sets

∪ =

I.9.7 Remark. Let Z ⊂ Ank be an algebraic set and I(Z) ⊂ k[x1 , ..., xn] its ideal. The ring

k[Z] := k[x1 , ..., xn]/I(Z)

is called the coordinate algebra of Z. It describes Z as an “abstract object”, i.e., without

reference to the inclusion ι : Z ֒→ Ank . We will elaborate on this later (see, e.g., the
following exercise).
I.9.8 Exercises (Maps between algebraic sets). Let k be a field and Z ⊂ Ank an algebraic
set. Its algebra of regular functions is

k[Z] := k[x1 , ..., xn ]/I(Z).

Note that an element f ∈ k[Z] defines indeed a function f : Z −→ k. Let W ⊂ Amk and
Z ⊂ Ank be algebraic sets and F : W −→ Z a map. Write the induced map F : W −→ Z ⊂
Ank as w 7−→ ( f1 (w), ..., fn(w)). We say that F is regular, if fi is a regular function on W,
i = 1, ..., n.
i) Show that a regular map F : W −→ Z induces a homomorphism

F ⋆ : k[Z] −→ k[W]

of k-algebras.
ii) Suppose ϕ : k[Z] −→ k[W] is a homomorphism of k-algebras. Show that there is a
unique regular map F : W −→ Z with F ⋆ = ϕ.
Let I ⊂ k[x1 , ..., xn ] be an ideal. Then, we have its vanishing locus V(I). As a subset
of Ank , it is endowed with an induced Zariski topology. On the other hand, we introduced
Spec(k[x1 , ..., xn]/I). It carries a topology that was also called the Zariski topology. It is
important not to confuse these two objects.
I.9.9 Remark. By Lemma I.2.2 and Exercise I.2.3, the surjection

π : k[x1 , ..., xn] −→ k[x1 , ..., xn]/I

induces an injective map

π# : Spec(k[x1 , ..., xn]/I) ֒→ Spec(k[x1 , ..., xn]),

and the Zariski topology on Spec(k[x1 , ..., xn]/I) agrees with the subspace topology (see
[18], Section 2.1).

45
I. Basic Theory of Rings and their Ideals

In order to get an idea about the difference of these two objects, let us first relate them
to each other. First, note that, for a point p = (a1 , ..., an) ∈ Ank ,

m p := h x1 − a1 , ..., xn − an i

is a maximal ideal.22 This provides us with the map

Φ: Ank−→ Spec k[x1 , ..., xn]
p 7−→ m p .

I.9.10 Lemma. For every point p = (a1 , ..., an) and every ideal I ⊂ k[x1 , ..., xn], we have

p ∈ V(I) ⇐⇒ I ⊂ h x1 − a1 , ..., xn − an i.

Proof. Let p ∈ V(I). Then, h x1 −a1 , ..., xn −an i ⊂ I({p}) ( h1i. Since h x1 −a1 , ..., xn −an i
is a maximal ideal, we must have equality. Thus, we see


h x1 − a1 , ..., xn − an i = I({p}) ⊃ I V(I) ⊃ I.

Conversely, I ⊂ h x1 − a1 , ..., xn − an i yields

{p} = V h x1 − a1 , ..., xn − an i ⊂ V(I).

This completes the proof. 

For any ideal I ⊂ k[x1 , ..., xn], the sets V(I) ⊂ Ank and V(I) ⊂ Spec(k[x1 , ..., xn]) are,
thus, related by


V(I) = Φ−1 V(I) . (I.10)
In particular, Φ is continuous in the Zariski topology.
I.9.11 Exercise. Let R be a ring. A point p ∈ Spec(R) is closed, if {p} is a Zariski closed
subset of Spec(R). Show that p ∈ Spec(R) is a closed point if and only if p is a maximal
ideal.
In order to get a precise understanding of Φ and, therefore, of the relation between Ank
and Spec(k[x1 , ..., xn ]), we need to understand the maximal ideals of k[x1 , ..., xn].
If k is algebraically closed, Hilbert’s Nullstellensatz III.3.4 shows that all maximal
ideals of k[x1 , ..., xn] are of the form h x1 − a1 , ..., xn − an i, (a1 , ..., an) ∈ Ank . So, Φ identifies
Ank with the set of closed points of Spec(k[x1 , ..., xn]). Note that Spec(k[x1 , ..., xn]) contains
non-closed points, e.g., h0i, and, if n ≥ 2, hxi i, i = 1, ..., n and, more generally, h f i,
f ∈ k[x1 , ..., xn] an irreducible polynomial.
If k is not algebraically closed, then Spec(k[x1 , ..., xn]) contains in some way the
points of AnK , K/k a finite extension field. Let us look at the extension C/R. The set
{ (±i, 1) ∈ A2C } corresponds to the maximal ideal h x2 + 1, y − 1 i ⊂ R[x, y]. Indeed,

R[x, y]/h x2 + 1, y − 1 i  R[x]/hx2 + 1i  C.

22
The canonical map (see Page 5 and Exercise I.1.9) k[x1 , ..., xn ] −→ k that sends xi to ai , i = 1, ..., n,
factorizes over an isomorphism k[x1 , ..., xn ]/h x1 − a1 , ..., xn − an i −→ k.

46
 I.9. Algebraic Sets

The fact that h x2 + 1, y − 1 i gives two points is reflected by the fact that we have two
R-linear isomorphisms between R[x]/hx2 + 1i and C, namely, we may map x either to i or
to −i.
The above discussion carries over to V(I) and Spec(k[x1 , ..., xn ]/I), I ⊂ k[x1 , ..., xn ] an
ideal. If k is not algebraically closed, it may happen that V(I) is empty (see Page 44).
For a non-zero ring R, the set Spec(R) is not empty, because R contains a prime ideal
(Theorem I.4.4).

47
II
Noetherian Rings

In this chapter, we present the concept of a noetherian ring.1 Noetherian rings are char-
acterized by a finiteness condition which may be phrased in several ways. Each of the
characterizations is given in terms of ideals of the ring. These rings are the most impor-
tant ones appearing in algebraic geometry. For example, all coordinate algebras of alge-
braic varieties are noetherian rings. We also introduce a major tool for constructing new
rings, namely localization. The main theorem of this chapter is the existence of primary
decompositions of ideals in noetherian rings which has already been alluded to several
times. It vastly generalizes prime factorizations and has also a geometric flavour: It is
closely related to the decomposition of an algebraic set into its irreducible components.
Primary decompositions are not unique, in general, but many important ingredients in a
primary decomposition are determined by the respective ideal. In order to prove these
uniqueness statements, we carefully investigate the extension and contraction of ideals
under localization.

II.1 Chain Conditions

A ring R is noetherian, if every ideal I ⊂ R is finitely generated, i.e., there are a natural
number n ≥ 1 and elements a1 , ..., an ∈ I, such that

I = h a1 , ..., an i.

Let R be a ring. An ascending chain in R is a sequence (Ik )k∈N of ideals with the
property
∀k ∈ N : Ik ⊂ Ik+1 .
Likewise, a descending chain in R is a sequence (Ik )k∈N of ideals, such that

∀k ∈ N : Ik ⊃ Ik+1 .
1
Emmy Noether (1882 - 1935), German mathematician.

49
II. Noetherian Rings

Let (Ik )k∈N be an ascending or descending chain in R. It is stationary, if there is an index

k0 with
∀k ≥ k0 : Ik = Ik0 .
Here are two alternative characterizations of noetherian rings.

II.1.1 Theorem. Let R be a ring. Then, the following conditions on R are equivalent:
i) The ring R is noetherian.
ii) The ring R satisfies the ascending chain condition (ACC), that is, every ascending
chain in R is stationary.
iii) Every non-empty subset Σ of ideals in R contains an element which is maximal
with respect to inclusion.

Let R , {0} be a notherian ring. Then, we can define Σ as the set of all proper ideals
of R. Thus, the theorem asserts, in particular, that every non-zero noetherian ring has
a maximal ideal. This statement is weaker than Theorem I.4.4 which asserts that every
non-zero ring has a maximal ideal. In the proof of Theorem II.1.1, we use the following:

II.1.2 Axiom of dependent choice. Let X be a non-empty set and R ⊂ X × X a relation,

satisfying the following property:

∀x ∈ X∃y ∈ X : (x, y) ∈ R.

Then, there is a sequence (xk )k∈N , such that

∀k ∈ N : (xk , xk+1 ) ∈ R.

If we do not assume the axiom of choice, this is really an axiom: One might try
to define the sequence (xk )k∈N by recursion. Having already constructed x0 , ..., xk , there
is an element y ∈ X with (xk , y) ∈ R. But y is, in general, not uniquely defined. We
have to choose one. The choice, of course, depends on the previous element. So, the
construction of (xk )k∈N requires countably many choices and these are not possible without
some axiom. For a deeper discussion of this axiom, we refer the reader to [12], Chapter
2.

Proof of Theorem II.1.1. “i)=⇒ii)”. Let (Ik )k∈N be an ascending chain of ideals. Then,
[
I := Ik
k∈N

is also an ideal. By assumption, it is finitely generated. Pick n ≥ 1 and a1 , ..., an ∈ I with

h a1 , ..., an i = I. There are natural numbers k1 , ..., kn with ai ∈ Iki , i = 1, ..., n. Set

k0 := max{ k1 , ..., kn }.

Then, ai ∈ Ik0 , i = 1, ..., n. We see:

∀k ≥ k0 : Ik0 ⊂ Ik ⊂ I ⊂ Ik0 .

This shows that Ik0 = Ik , k ≥ k0 .

50
 II.1. Chain Conditions

“ii)=⇒iii)”. Suppose that iii) does not hold, and let Σ be a non-empty set of ideals in
R without maximal element. Then, there exists an ascending chain (Ik )k∈N with

∀k ∈ N : Ik ∈ Σ ∧ Ik ( Ik+1 .

We found an ascending chain in R which is not stationary, a contradiction.

“iii)=⇒i)”. Let I ⊂ R be an ideal. We define Σ as the set of all finitely generated ideals
of R which are contained in I. Note that h0i ∈ Σ, so that Σ , ∅. Let J ∈ Σ be a maximal
element. We have J ⊂ I, and J is finitely generated. We have to show that J = I. If this
were wrong, we could pick an element a ∈ I \ J. Then,

J ( J + hai.

This implies that J + hai is not finitely generated. On the other hand, there exist n ≥ 1
and a1 , ..., an ∈ J with h a1 , ..., an i = J. But then

J + hai = h a1 , ..., an, a i,

a contradiction. 
II.1.3 Exercise. Where was the axiom of dependent choice used in the above proof? Spec-
ify the set X and the relation R to which it was applied.
II.1.4 Corollary. Let R, S be rings and ϕ : R −→ S a surjective ring homomorphism. If
R is noetherian, then so is S .
Proof. Let J ⊂ S be an ideal of S . Then,

J = ϕ(I), I := ϕ−1 (J).

Since I is an ideal of R and R is noetherian, there exists a natural number n ≥ 1 and

elements a1 , ..., an ∈ I with I = h a1 , ..., an i. Obviously, J is generated by the images
ϕ(a1 ), ..., ϕ(an).
We can also argue by using ACC: By Exercise I.2.4, ii),

S  R/ker(ϕ).

The ideals of S are in inclusion preserving bijection to the ideals of R containing ker(ϕ)
(Lemma I.2.2). So, the ascending chain condition in R clearly implies the ascending chain
condition in S . 
II.1.5 Examples. i) Fields are noetherian rings.
ii) Principal ideal domains are noetherian rings. Recall that Z and the polynomial ring
k[x] over a field k are examples for principal ideal domains and, thus, for noetherian rings.
iii) We let R := C 0 ([0, 1]) be the ring of continuous functions on the interval [0, 1]
(compare Example I.1.3, vi). Set
n  o
Ik := f ∈ R  f|[0,1/k] ≡ 0 , k ≥ 1.

Then, (Ik )k≥1 is an ascending chain which is not stationary. In particular, R is not a noethe-
rian ring.

51
II. Noetherian Rings

II.1.6 Hilbert’s basis theorem. If R is a noetherian ring, then the polynomial ring R[x]
is also a noetherian ring.
Proof. Assume that R[x] is not noetherian. Let I ⊂ R[x] be an ideal which is not finitely
generated and X the set of all finite subsets of I. It is obviously non-empty. For x, y ∈ X,
we write x < y, if x ⊂ y, #y = #x + 1, and the unique element f ∈ y \ x satisfies

f < hxi ∧ ∀g ∈ I \ hxi : deg(g) ≥ deg( f ).

Since I is not finitely generated, it is clear that we find for each x ∈ X an element y ∈ X
with x < y. By the axiom of dependent choice II.1.2, there is a sequence (xk )k∈N with
xk ∈ X and xk < xk+1 , k ∈ N. This defines the sequence ( fk )k≥1 with

{ fk } = xk \ xk−1 , k ≥ 1.

Finally, we obtain the sequence (ak )k≥1 in R in which ak is the leading coefficient of fk ,
i.e., the coefficient of xdeg( fk ) in fk , k ≥ 1.
Claim. ∀k ≥ 1 : h a1 , ..., ak i ( h a1, ..., ak , ak+1 i.
If we had h a1 , ..., ak i = h a1 , ..., ak , ak+1 i, then there were elements r1 , ..., rk ∈ R with
X
k
ak+1 = r i · ai .
i=1

The definition of the relation “<” implies

∀k : deg( fk ) ≤ deg( fk+1 ).

Therefore, we can define

X
k
g := fk+1 − ri · xdeg( fk+1 )−deg( fi ) · fi . (II.1)
i=1

But then,
deg(g) < deg( fk+1 ).
By construction of “<”, fk+1 has the least degree among all elements of I that are not
contained in hx0 i + h f1, ..., fk i. So, we must have g ∈ hx0 i + h f1, ..., fk i. Now, (II.1) shows
fk+1 ∈ hx0 i + h f1 , ..., fk i, and this is a contradiction. X
This claim shows that (h a1 , ..., ak i)k≥1 is a non-stationary ascending chain in R. This
contradicts the assumption that R is noetherian. 

II.1.7 Corollary. Let R be a noetherian ring, e.g., a field. Then, the polynomial ring
R[x1 , ..., xn] is noetherian, n ≥ 1.
II.1.8 Remark (An application to algebraic sets). Let k be a field, n ≥ 1 a natural number,
k[x1 , ..., xn] the polynomial ring in n variables, and F ⊂ k[x1 , ..., xn] a possibly infinite
subset. We claim that there are a natural number m ≥ 1 and elements f1 , ..., fm ∈ F with


V(F) = V { f1 , ..., fm } .

52
 II.1. Chain Conditions

In particular, every algebraic set may be defined by finitely many equations. According
to Property I.9.1, v),


V(F) = V hFi .
Now, k[x1 , ..., xn] is a noetherian ring, so that hFi is finitely generated. Pick a natural
number s ≥ 1 and elements g1 , ..., g s with
hFi = h g1 , ..., g s i.
Then,


V(F) = V { g1, ..., g s } ,
but the elements g1 , ..., g s need not belong to F. However, there are elements
ai j ∈ R and fi j ∈ F, j = 1, ..., mi, i = 1, ..., s,
such that
X
mi
gi = ai j · f i j , i = 1, ..., s.
j=1
We obviously have
hFi = h f11 , ..., f1m1 , ..., f s1, ..., f sms i
and, consequently,




V(F) = V hFi = V h f11 , ..., f1m1 , ..., f s1, ..., f sms i = V { f11 , ..., f1m1 , ..., f s1, ..., f sms } .
Let R be a ring. An R-algebra is a ring S together with a ring homomorphism ϕ : R −→
S . An R-algebra S is finitely generated, if there are elements f1 , ..., fn ∈ S , such that the
homomorphism
R[x1 , ..., xn ] −→ S
that is associated with ϕ and the assignment xi 7−→ fi , i = 1, ..., n, is surjective (compare
Exercise I.1.9). One writes abusively
S = R[ f1 , ..., fn].
II.1.9 Proposition. If R is a noetherian ring and S is a finitely generated R-algebra, then
S is also a noetherian ring.
II.1.10 Example. Let k be a field, n ≥ 1 a natural number, and Z ⊂ Ank an algebraic set.
Then, the coordinate algebra
k[x1 , ..., xn ]/I(Z)
is a finitely generated k-algebra and, hence, a noetherian ring.
II.1.11 Exercise (Noetherian rings and spaces). i) Let k be a field. Then, one may define
the polynomial ring R := k[x1 , x2 , x3 , ...] in the infinitely many variables xi , i ≥ 1. Is R
noetherian?
ii) A topological space X is called noetherian, if it satisfies the descending chain
condition for closed subsets, i.e., for any sequence
Z1 ⊃ Z2 ⊃ · · ·
of closed subsets of X, there is an index k0 , such that Zk = Zk0 , for every k ≥ k0 . Let R be
a noetherian ring. Show that Spec(R) is a noetherian topological space.
iii) Give an example of a non-noetherian ring R, such that Spec(R) consists of just one
point (and is, therefore, noetherian).

53
II. Noetherian Rings

II.2 Artinian Rings

A ring satisfies the descending chain condition (DCC), if every descending chain of ideals
in R is stationary. A ring which satisfies the descending chain condition is called an
artinian2 ring. It might be surprising that the descending chain condition is much more
restrictive than the ascending chain condition. In fact, it turns out that artinian rings are
very special noetherian rings (see [1], Theorem 8.5). We will not discuss this in full detail.
Instead we give some indications.

II.2.1 Proposition. In an artinian ring, every prime ideal is maximal.

Proof. Let R be an artinian ring and p ⊂ R a prime ideal. Then, R/p is an integral domain.
Lemma I.2.2 shows that R/p satisfies the descending chain condition (compare the proof
of Corollary II.1.4). Let x ∈ R/p be a non-zero element. Consider the descending chain


hxk i k∈N .

For some k0 ≥ 0,
hxk0 i = hxk0 +1 i.
Thus, there exists an element a ∈ R/p with xk0 = a · xk0 +1 . In other words,

xk0 · (1 − a · x) = 0.

Since R/p is an integral domain, xk0 , 0 and, so, 1 − a · x = 0 and x is a unit.

Every element x ∈ (R/p) \ {0} is a unit. This shows that R/p is a field and p is a
maximal ideal. 
II.2.2 Examples. Let k be a field.
i) Let R be a k-algebra which is finite dimensional as a k-vector space. Then, R satisfies
the ascending and descending chain condition for dimension reasons. (Ideals in R are, in
particular, sub vector spaces.) So, R is both noetherian and artinian.
ii) For n ≥ 1, k[x]/hxn i is a k-algebra which has dimension n as k-vector space. By i),
it is an artinian ring.

II.3 Localization
Let R be a ring. A subset S ⊂ R is multiplicatively closed, if

⋆ 1 ∈ S,

⋆ ∀s, t ∈ S : s · t ∈ S .
II.3.1 Example. Let R be a ring, f ∈ R, and p ⊂ R a prime ideal. Then, the following
subsets of R are multiplicatively closed:

⋆ S := a ∈ R | a is not a zero divisor .
 
⋆ S := f k | k ∈ N = 1, f, f 2, ... .
2
Emil Artin (1898 - 1962), Austrian mathematician.

54
 II.3. Localization

⋆ S := R \ p.

Let R be a ring and S ⊂ R a multiplicatively closed subset. We define a relation “∼” on

R × S by:

∀(a, s), (b, t) ∈ R × S : (a, s) ∼ (b, t) :⇐⇒ ∃u ∈ S : u · (a · t − b · s) = 0.

II.3.2 Lemma. The relation “∼” is an equivalence relation on R × S .

Proof. Reflexivity and symmetry follow easily from 1 ∈ S . For transitivity, let (a, s),
(b, t), (c, u) ∈ R × S with (a, s) ∼ (b, t) and (b, t) ∼ (c, u). This means that there exist
elements v, w ∈ S with

v · (a · t − b · s) = 0 and w · (b · u − c · t) = 0.

We multiply the first equation by u · w, the second by v · s and find

u·v·w·a·t−u·v·w·b·s = 0
u · v · w · b · s − v · w · c · t · s = 0.

Adding these two yields

t · v · w · (a · u − c · s) = u · v · w · a · t − v · w · c · t · s = 0.

Since S is multiplicatively closed, t · v · w ∈ S , and we see (a, s) ∼ (c, u). 

II.3.3 Remarks. Assume that R is an integral domain and 0 < S . Then,3

∀(a, s), (b, t) ∈ R × S : (a, s) ∼ (b, t) ⇐⇒ a · t − b · s = 0.

If R does contain zero divisors, it is necessary to formulate the relation “∼” as above in
order to get an equivalence relation.
In the following, we write
a
s
for the equivalence class of (a, s), a ∈ R, s ∈ S , and set
 a  
RS :=  a ∈ R, s ∈ S .
s
II.3.4 Notation. i) Other common symbols for RS are S −1 R and R[S −1 ].
ii) Let f ∈ R and S = { f k | k ∈ N }. Then, we write R f for RS .
iii) Let p ⊂ R be a prime ideal and S := R \ p. We will write Rp instead of RS .

Next, we equip RS with the structure of a ring. The addition is defined via

+ : RS × RS −→ RS
!
a b a·t+b·s
, 7−→ .
s t s·t
3
Compare Page 25.

55
II. Noetherian Rings

We first check that this is well-defined. Let (a, s), (a′, s′), (b, t) ∈ R × S with

(a, s) ∼ (a′ , s′).

We need to show
(a · t + b · s, s · t) ∼ (a′ · t + b · s′ , s′ · t). (II.2)
For this, we observe

∆ := (a · t + b · s) · s′ · t − (a′ · t + b · s′ ) · s · t = a · s′ · t2 − a′ · s · t2 = (a · s′ − a′ · s) · t2 .

If u ∈ S annihilates a · s′ − a′ · s, then u also annihilates ∆. This establishes (II.2). X

The addition is clearly commutative, and
0
0 :=
1
is the neutral element. Finally, we check associativity: Let a/s, b/t, and c/u ∈ RS . We
compute
!
a b c a·t+b·s c
+ + = +
s t u s·t u
a·t·u+b·s·u+c·s·t
=
s·t·u
a b·u+c·t
= +
s t · u!
a b c
= + + .
s t u
The multiplication is defined as

· : RS × RS −→ RS
!
a b a·b
, 7−→ .
s t s·t
Again, we verify that this is well-defined. For (a, s), (a′ , s′), (b, t) ∈ R × S with

(a, s) ∼ (a′ , s′),

we have to establish
(a · b, s · t) ∼ (a′ · b, s′ · t). (II.3)
We form
∆ := a · b · s′ · t − a′ · b · s · t = (a · s′ − a′ · s) · b · t.
If u ∈ S annihilates a · s′ − a′ · s, then u also annihilates ∆. We infer (II.3). X
It is immediate that multiplication is commutative and associative and that
1
1 :=
1
is the neutral element.

56
 II.3. Localization

The last thing we verify is the distributive law. For a/s, b/t, and c/u ∈ RS , we find
!
a b c a b·u+c·t
· + = ·
s t u s t·u
a·b·u+a·c·t
=
s·t·u
a·b·s·u+a·c·s·t
=
s2 · t · u
a·b a·c
= +
s·t s·u
a b a c
= · + · .
s t s u
So, (RS , +, ·, 0, 1) is a ring.
II.3.5 Example. i) If R is an integral domain and S := R \ {0}, then RS is the quotient field
of R that was already considered in Section I.6.
ii) In an arbitrary ring R,

S := s ∈ R | s is not a zero divisor

is a multiplicatively closed subset. In this case, Q(R) := RS is the total ring of fractions
of R.
iii) The ring RS is the zero ring if and only if 0 ∈ S . In fact, 0/1 = 1/1 is equivalent
to the existence of an element s ∈ S with s · 1 = 0. Multiplicatively closed subsets of R
which contain zero are, e.g., { 0, 1 }, R.
We look at the homomorphism

ϕ : R −→ RS
a
a 7−→ .
1
Note that
a
= ϕ(a) = 0 ⇐⇒ ∃s ∈ S : s · a = 0.
1
This shows [
ker(ϕ) = Ann(s).
s∈S

In particular, if R is an integral domain, ϕ is injective.

II.3.6 Proposition. i) For every ideal I ⊂ R, the extension of I via ϕ is

 a  
e
I = J := 
 a ∈ I, s ∈ S .
s
ii) For an ideal I ⊂ R,

I e = RS ⇐⇒ I ∩ S , ∅.

iii) Every ideal J ⊂ RS is an extended ideal.

57
II. Noetherian Rings

iv) For an ideal I ⊂ R, we have

[

I ec = I : hsi
s∈S

v) An ideal I ⊂ R is a contracted ideal if and only if, for all s ∈ S , [s] is not a zero
divisor in R/I.
Proof. i) It is clear that ϕ(I) ⊂ J ⊂ I e . Since I e is the smallest ideal of RS which contains
ϕ(I), it suffices to show that J is an ideal. But this is immediate.
ii) “⇐=”. If s ∈ I ∩ S , then
s
1 = ∈ Ie.
s
“=⇒”. For a ∈ I and s ∈ S , the equation
a
=1
s
implies that there is element t ∈ S with
t · a − s · t = t · (a − s) = 0.
So, s · t ∈ I ∩ S .
iii) By Property I.8.24, i), J ce ⊂ J. So, we have to show J ⊂ J ce . Let a/s ∈ J. Then,
a s a
= · ∈ J.
1 1 s
This shows a ∈ J c and a/s ∈ J ce .
iv) “⊂”. Let a ∈ I ec . This means that there exist b ∈ I and s ∈ S with
a b
= .
1 s
So, there exists an element t ∈ S with
t · (a · s − b) = 0.
For u := s · t ∈ S , we obtain


u · a ∈ I, i.e., a ∈ I : hui .
“⊃”. Let s ∈ S and a ∈ (I : hsi). Then, b := a · s ∈ I, so that
a b
= ∈ Ie.
1 s
This shows a ∈ I ec .
v) Since I ⊂ I ec (Property I.8.24, i), we have to investigate the condition I ec ⊂ I. By
iv), it is equivalent to


∀s ∈ S : I : hsi ⊂ I.
This condition is, in turn, equivalent to
∀s ∈ S ∀a ∈ R : a·s∈ I =⇒ a ∈ I.
This is equivalent to the condition that [s] ∈ R/I is not a zero divisor, s ∈ S . 

58
 II.3. Localization

II.3.7 Corollary. Let R, S , and ϕ : R −→ RS be as in the proposition.

i) The assignment p 7−→ pe gives an inclusion preserving bijection between the prime
ideals of R with p ∩ S = ∅ and the prime ideals of RS .
ii) Let p ⊂ R be a prime ideal. Then, Rp is a local ring with maximal ideal pe .

Proof. i) Let p ⊂ R be a prime ideal with p ∩ S = ∅. By Part ii) of the proposition, pe is

a proper ideal of R. We show that it is a prime ideal. Let a, b ∈ R and s, t ∈ S with

a·b a b
= · ∈ pe .
s·t s t
So, there exist c ∈ p and u ∈ S with

a·b c
= ,
s·t u
i.e.,
∃v ∈ S : v · (a · b · u − c · s · t) = 0.
We see
(a · b) · (u · v) = (s · t · v) · c ∈ p.
Since u, v < p, we have u · v < p and a · b ∈ p. This means a ∈ p or b ∈ p and

a b
∈ pe or ∈ pe .
s t
Finally, R/p is an integral domain (Proposition I.4.1, i). Since p ∩ S = ∅, we have
0 , [s] ∈ R/p, i.e., [s] is not a zero divisor, s ∈ S . Proposition II.3.6, v), shows pec = p.
ii) This follows immediately from i): For a prime ideal q ⊂ R, q ∩ (R \ p) = ∅ is
equivalent to q ⊂ p. 

II.3.8 Exercise (The universal property of localization). Let R be a ring, S ⊂ R a mul-

tiplicatively closed subset, and ϕ : R −→ RS , x 7−→ x/1, the canonical homomorphism
to the localization. Show that the pair (RS , ϕ) has the following universal property: For
any ring T and any homomorphism ψ : R −→ T , such that ψ(S ) ⊂ T ⋆ , there is a unique
homomorphism ψS : RS −→ T with ψ = ψS ◦ ϕ.
To conclude this section, we will give another proof of Proposition I.7.2 which only
uses the axiom that every non-zero ring possesses a prime ideal. This axiom is weaker
than the axiom of choice (see Remark I.4.5, ii).

Alternative proof of Proposition I.7.2. The inclusion “⊂” is obtained as before. For the
converse inclusion, let f ∈ R\N. Then, the multiplicatively closed subset S = { f k | k ∈ N }
does not contain 0, so that R f , {0}, by Example II.3.5, iii). Let q ⊂ R f be a prime ideal.
Then, p := qc is a prime ideal of R with

p ∩ f k | k ∈ N = ∅,

i.e., f < p. 

59
II. Noetherian Rings

II.4 Primary Decomposition

In the ring of integers, we need powers of prime numbers in order to factorize all positive
integers. We now need the ideal theoretic analogs for powers of prime numbers. It turns
out that these are quite subtle.
Let R be a ring. An ideal q ⊂ R is primary, if the following property holds:

∀a, b ∈ R : a·b∈q =⇒ a∈q ∨ ∃k ≥ 1 : bk ∈ q.

II.4.1 Lemma. An ideal q ⊂ R is a primary ideal if and only if every zero divisor in the
ring R/q is nilpotent.

Proof. Assume that q is a primary ideal and that b ∈ R is an element, such that [b] ∈ R/q
is a zero divisor. Then, there is an element a ∈ R with [a] , 0 and [a · b] = [a] · [b] = 0.
This means that a · b ∈ q but a < q. By assumption, there is an exponent k ≥ 1 with bk ∈ q,
i.e., [b]k = 0.
Now, assume that every zero divisor in R/q is nilpotent. Let a, b ∈ R with a · b ∈ q.
This means [a] · [b] = [a · b] = 0 in R/q. So, [a] = 0 or [b] is a zero divisor and there
exists a natural number k ≥ 1 with [bk ] = [b]k = 0. This shows a ∈ q or there is a natural
number k ≥ 1 with bk ∈ q. 

II.4.2 Examples. i) A prime ideal is a primary ideal.

ii) If ϕ : R −→ S is a ring homomorphism and q ⊂ S is a primary ideal, then qc ⊂ R is
a primary ideal. In fact, by Exercise I.2.4, ii), we have an injective homomorphism

ϕ : R/qc −→ S /q,

so that we may apply Lemma II.4.1.

√
II.4.3 Lemma. Let R be a ring and q ⊂ R a primary ideal, then the radical q is a prime
ideal.
√
Proof. Let a, b ∈ R be elements with a · b ∈ q. There is an exponent k ≥ 1 with

ak · bk = (a · b)k ∈ q.

Then, ak ∈ q or there is an exponent l ≥ 1 with

bk·l = (bk )l ∈ q.
√ √
This shows a ∈ q or b ∈ q as asserted. 

Let q ⊂ R be an ideal and p a prime ideal. We say that q is a p-primary ideal, if it is a

√
primary ideal with q = p.
II.4.4 Examples. i) The primary ideals of Z are exactly the ideals of the form hpk i, p a
prime number, k ≥ 1 a natural number. First, let q ⊂ Z be a primary ideal. According
√
to Lemma II.4.3, q is a prime ideal, i.e., of the form hpi, for some prime number p. By
Example I.8.20, i), q = hpk i, for some natural number k ≥ 1. Second, let k ≥ 1 be a
natural number and p a prime number. For integers a, b ∈ Z, a · b ∈ hpk i means pk |(a · b).

60
 II.4. Primary Decomposition

Then, pk divides a or p divides b and, then, pk divides bk . It follows that a ∈ hpk i or

bk ∈ hpk i.
ii) Let k be a field, R = k[x, y], and q := h x, y2 i. Then,

R/q  k[y]/hy2 i.

The zero divisors in that ring form the ideal hyi. The elements of that ideal are all nilpo-
tent. So, q is a primary ideal. Its radical is
√
q = h x, y i =: p.

Observe
p2 = h x2 , x · y, y2 i ( q = h x, y2 i ( p = h x, y i.
This shows:
Not every primary ideal is the power of a prime ideal.
iii) Let k be a field and

R := k[x, y, z]/h x · y − z2 i.

We look at


p := [x], [z] .
Since
R/p = k[y],
the ideal p is a prime ideal. We claim that the ideal p2 is not primary. We have

[x] · [y] = [z]2 ∈ p2 .

Here, [x] < p2 and, for all k ≥ 1, [y]k < p2 . We note:

Not every power of a prime ideal is a primary ideal.
The phenomenon in Example II.4.4, iii), cannot be observed in rings in which every
non-zero prime ideal is a maximal ideal, such as principal ideal domains. More generally,
the following holds true:
√
II.4.5 Lemma. Let R be a ring and q ⊂ R an ideal, such that m = q is a maximal ideal.
Then, q is an m-primary ideal.
√
Proof. By Corollary I.8.18, q is the intersection of all prime ideals which contain q. The
assumption implies that m is the only prime ideal which contains q, i.e., R/q is a local
ring. Let m be the image of m in R/q. The elements of m are nilpotent, and the elements
of (R/q) \ m are units. In particular, every zero divisor in R/q is nilpotent. By Lemma
II.4.1, q is an m-primary ideal. 
II.4.6 Primary decomposition in noetherian rings. Let R be a noetherian ring and
I ( R a proper ideal. Then, there exist a natural number m ≥ 1 and primary ideals
q1 , ..., qm , such that
I = q1 ∩ · · · ∩ qm .

61
II. Noetherian Rings

Proof. Step 1. An ideal q ⊂ R is irreducible, if, for ideals I, J ⊂ R, the equality

q=I∩J
implies q = I or q = J.
Claim. For every proper ideal I ( R, there exist a natural number m ≥ 1 and irreducible
ideals q1 , ..., qm , such that
I = q1 ∩ · · · ∩ qm .
Let Σ be the set of proper ideals in R which cannot be written as the intersection of
finitely many irreducible ideals. Assume that Σ is non-empty. By Theorem II.1.1, this set
contains a maximal element a. The ideal a is not irreducible. So, there exist ideals a ( I
and a ( J with
a = I ∩ J.
Note that this implies I ( R and J ( R, By definition of a, the ideals I and J are finite
intersections of irreducible ideals. But then, a is also an intersection of finitely many
irreducible ideals, a contradiction. X
Step 2. By Step 1, it remains to be shown that every irreducible ideal is primary. Let
q ⊂ R be an irreducible ideal. In view of Lemma I.2.2, we may replace R by R/q. So,
without loss of generality, we may assume q = h0i. Let a, b ∈ R with a · b = 0 and a , 0.
We look at the ascending chain


Ann(bk ) k∈N .
Since R is noetherian, there exists a natural number k0 ∈ N with
Ann(bk0 ) = Ann(bk0 +1 ).
Claim. hai ∩ hbk0 i = h0i.
Let c ∈ hai ∩ hbk0 i. We write c = r · bk0 for a suitable element r ∈ R. Since c ∈ hai, we
have
0 = c · b = r · bk0 +1 .
This means
r ∈ Ann(bk0 +1 ) = Ann(bk0 ),
so that c = r · bk0 = 0. X
Since h0i is irreducible and h0i , hai, we have h0i = hbk0 i, i.e., bk0 = 0. 
The first step of the above proof bears strong resemblance to the proof of the existence
of a prime factorization in the ring of integers ([30], Satz I.3.2).
II.4.7 Corollary. Let R be a noetherian ring and I ⊂ R a radical ideal. Then, there exist
prime ideals p1 , ..., pm with
I = p1 ∩ · · · ∩ pm .
Proof. Let
I = q1 ∩ · · · ∩ qm
be a primary decomposition. According to our assumption and Property I.8.19, iv), we
have √ √ √
I = I = q1 ∩ · · · ∩ qm .
√
By Lemma II.4.3, the radical ideal pi := qi is a prime ideal, i = 1, ..., m. 

62
 II.4. Primary Decomposition

II.4.8 Remarks. i) The theorem is also known as the Lasker4 –Noether theorem.
ii) A ring in which every ideal can be written as a finite intersection of primary ideals
is called a laskerian ring. The above theorem states that every noetherian ring is laskerian.
However, there exist laskerian rings which are not noetherian. An example is contained
in [9].
iii) The zero ideal in the ring C 0 ([0, 1]) of continuous functions on the unit interval
does not admit a primary decomposition (Exercise II.4.11).
II.4.9 Caution. In the proof of the Lasker–Noether theorem II.4.6, we showed that every
irreducible ideal in a noetherian ring is a primary ideal. The converse does not hold: Let
k be a field and R := k[x, y].5 We have

h x2 , x · y, y2 i = h x, y2 i ∩ h y, x2 i.

All occurring ideals have the maximal ideal m = h x, y i as radical, so they are all m-
primary, by Lemma II.4.5.
The reader should also look at the proof of Lemma II.4.12. There, several primary
ideals are intersected to obtain a new, non-irreducible, primary ideal. In order to obtain
uniqueness statements, we have to include primary ideals which are not irreducible. This
explains, in particular, why we work with primary ideals rather than with irreducible ones.
II.4.10 Exercise (Maximal ideals in rings of continuous functions). In this exercise, we
work in the ring R := C 0 ([0, 1]) of continuous functions on the interval [0, 1] ⊂ R.
i) Show that, for a point x ∈ [0, 1],

m x := f ∈ R | f (x) = 0

is a maximal ideal in R.
ii) Let m ⊂ R be a maximal ideal of R. Show that there exists a point x ∈ [0, 1] with
m = m x . (Hint. Use the compactness of [0, 1].)
II.4.11 Exercise (An ideal without primary decomposition). Let R be as in the previous
exercise.
i) Let q ⊂ R be a primary ideal. Show that there is a unique point x ∈ [0, 1] with
q ⊂ mx .
ii) Conclude that the zero ideal h0i ⊂ R cannot be written as the intersection of finitely
many primary ideals.

The First Uniqueness Theorem

For the considerations in this part, we do not need to assume that the ring R is noetherian.
Unfortunately, primary decompositions are, in general, not unique. First, there are
some stupid reasons for non-uniqueness which we will eliminate first. A primary decom-
position
I = q1 ∩ · · · ∩ qm
is minimal or irredundant, if the following two properties hold:
4
Emanuel Lasker (1868 - 1941), German chess player, mathematician, and philosopher.
5
Example taken from http://math.stackexchange.com/questions/28620/primary-ideals-
of-noetherian-rings-which-are-not-irreducible

63
II. Noetherian Rings

\
i) ∀i ∈ { 1, ..., m }: q j 1 qi .
j∈{ 1,...,m }\{i}

√ √
ii) For 1 ≤ i < j ≤ m: qi , q j .

II.4.12 Lemma. Let R be a ring, I ⊂ R an ideal, and

I = q1 ∩ · · · ∩ qm

a primary decomposition. Then, the ideal possesses also a minimal primary decomposi-
tion
I = s1 ∩ · · · ∩ sn .

Proof. Step 1. Given any primary decomposition, one gets a primary decomposition
satisfying Condition i), by just removing some of the primary components.
Step 2. The idea is to collect the primary components with the same radical in one
primary component. For this, we need the following result:

Claim. Let p ⊂ R a prime ideal and r1 , r2 ⊂ R two p-primary ideals. Then, the intersection

r := r1 ∩ r2

is also a p-primary ideal.

We first compute the radical with the help of Property I.8.19, iv):
√ √ √
r1 ∩ r2 = r1 ∩ r2 = p ∩ p = p.

Now, we show that r1 ∩ r2 is a primary ideal. Let a, b ∈ R with a · b ∈ r1 ∩ r2 and

a < r1 ∩ r2 . We may assume without loss of generality that a < r1 . Then, there exists an
exponent k ≥ 1 with bk ∈ r1 . So,
√ √
b ∈ r1 = r1 ∩ r2 .

This means that there is also an exponent l ≥ 1 with bl ∈ r1 ∩ r2 . X

Let p1 , ..., pn be the distinct prime ideals with
 √ √
p1 , ..., pn = q1 , ..., qm .

Set \
 √
I j := i ∈ { 1, ..., m } | qi = p j and rj = qi , j = 1, ..., n.
i∈I j

By the claim,
I = r1 ∩ · · · ∩ rn (II.4)
is a primary decomposition with satisfies Condition ii). If the primary decomposition we
started with satisfied Condition i), (II.4) will also satisfy Condition i). Otherwise we apply
Step 1 to (II.4). 

64
 II.4. Primary Decomposition

II.4.13 Example. Let k be a field and R = k[x, y]. Here, we have

h x2 , x · y i = hxi ∩ h x, y i2 = hxi ∩ h x2 , y i. (II.5)
The ideal hxi is a prime ideal, and
p p
h x, y i = h x2 , y i = h x, y i2 .
This is a maximal ideal, so h x2 , y i and h x, y i2 are primary ideals. Therefore, (II.5) con-
tains two distinct minimal primary decompositions of the same ideal.
We can understand the underlying geometry at a heuristic level. The algebraic set
V(x2 ) is the y-axis in A2k . We should count it twice. The set V(x · y) is the union of the
coordinate axes. The intersection V(x2 ) ∩ V(x · y) consists of the y-axis. But the origin
occurs with multiplicity 2, because it occurs both in V(x2 ) and in V(x · y) with multiplicity
2. This fact is reflected by the above primary decompositions.

The origin occurs with This gives an “embedded”

multiplicity two in point of multiplicity two
both “varieties”. in the intersection.

∩ b = b

II.4.14 First uniqueness theorem. Let R be a ring, I ⊂ R an ideal,

I = q1 ∩ · · · ∩ qm
√
a minimal primary decomposition, and pi := qi , i = 1, ..., m. Then,
q
 n
o
p1 , ..., pm = p ⊂ R | p is a prime ideal : ∃r ∈ R with p = I : hri .
In particular, the set { p1 , ..., pm } depends only on I and not on the primary decomposition.
II.4.15 Corollary. Let R be a noetherian ring. For every radical ideal I ⊂ R, there exist
uniquely determined distinct prime ideals p1 , ..., pm with
I = p1 ∩ · · · ∩ pm .
Let I ( R be a proper ideal. A prime ideal p ⊂ R is associated with I, if there is an
element r ∈ R with q


p = I : hri .
The set of prime ideals associated with I is denoted by
Ass(I).
The prime ideals in Ass(I) which are minimal with respect to inclusion among the ideals
in Ass(I) are called isolated. The remaining ones are called embedded.

65
II. Noetherian Rings

II.4.16 Example. We return to Example II.4.13 and the primary decomposition

h x2 , x · y i = hxi ∩ h x, y i2 .

The associated prime ideals are hxi and h x, y i, and we have

hxi ⊂ h x, y i.

So, hxi is an isolated associated prime ideal and h x, y i an embedded one. The corre-
sponding geometric objects are the y-axis V(hxi) and the origin V(h x, y i). The origin is
embedded into the y-axis. This explains the terminology.
We need some preparations for the proof of Theorem II.4.14.

II.4.17 Prime avoidance. Let R be a ring.

i) Suppose p1 , ..., pm ⊂ R are prime ideals and I ⊂ R an ideal with
[
m
I⊂ pi .
i=1

Then, there exists an index i0 ∈ { 1, ..., m } with

I ⊂ pi0 .

ii) Let I1 , ..., Im ⊂ R be ideals and p ⊂ R a prime ideal. If

\
m
p⊃ I j,
j=1

then, there exists an index j0 ∈ { 1, ..., m } with

p ⊃ I j0 .

Moreover, if
\
m
p= I j,
j=1

then, there exists an index j0 ∈ { 1, ..., m } with

p = I j0 ,

i.e., prime ideals are irreducible.

Proof. i) We prove the following statement6 by induction on m:

[
m
∀i ∈ { 1, ..., m } : I 1 pi =⇒ I1 pi .
i=1
6
This explains the name “prime avoidance”: If I avoids the prime ideals p1 , ..., pm , then it also avoids
their union.

66
 II.4. Primary Decomposition

The case m = 1 is trivial. For the induction step “m −→ m + 1”, we may choose elements
a1 , ..., am+1 ∈ I with

ai < p j , j ∈ { 1, ..., m + 1 } \ {i}, i = 1, ..., m + 1.

If there is an index i0 ∈ { 1, ..., m + 1 } with ai0 < pi0 , we are done. Thus, we need to
consider the case ai ∈ pi , i = 1, ..., m + 1. We then form

X
m+1
b := a1 · · · · · ai−1 · ai+1 · · · · · am+1 ∈ I.
i=1

Let i ∈ { 1, ..., m + 1 }. The i-th summand of b is not contained in pi , but all the other
summands are. It follows b < pi , i = 1, ..., m + 1.
ii) Suppose that p 2 I j , j = 1, ..., m. Choose elements a j ∈ I j \ p, j = 1, ..., m. Then,

Y
m \
m
a1 · · · · · am ∈ Ij ⊂ I j,
j=1 j=1

but
a1 · · · · · am < p.
This is a contradiction.
Assume
\
m
p= I j.
j=1

By what has already been proved, there is an index j0 ∈ { 1, ..., m } with p ⊃ I j0 . We clearly
have p ⊂ I j0 and, thus, p = I j0 . 

II.4.18 Lemma. Let R be a ring, p ⊂ R a prime ideal, q a p-primary ideal, and a ∈ R.

Then,


 R, if a ∈ q


q : hai = 
 a p-primary ideal, if a < q .

 q, if a < p

Proof. The assertion for a ∈ q is clear. Next, let us assume a < p. If r ∈ (q : hai), we have
√
r · a ∈ q. Since a < p = q, we have r ∈ q. This shows (q : hai) ⊂ q. The other inclusion
is trivial.
Next, we just assume a < q. Let us first determine the radical of (q : hai). For
√
r ∈ (q : hai), we have r · a ∈ q. Since a < q, we have r ∈ q = p. This implies


q ⊂ q : hai ⊂ p.
√
Taking radicals yields (q : hai) = p. √
Finally, assume that r, s ∈ R are elements with r · s ∈ (q : hai) and s < p = (q : hai).
We have to check that r ∈ (q : hai). With a · r · s ∈ q and s < p, we find a · r ∈ q, i.e.,
r ∈ (q : hai). 

67
II. Noetherian Rings

Proof of Theorem II.4.14. Let I ⊂ R be an ideal and

I = q1 ∩ · · · ∩ qm

a minimal primary decomposition. By Property I.8.17, iv),

\
m ! \m



∀a ∈ R : I : hai = qi : hai = qi : hai .
i=1 i=1

With Lemma II.4.18, this gives

q q

\ \
m


∀a ∈ R : I : hai = qi : hai = pi . (II.6)
i=1 i∈{ 1,...,m | a<qi }

√
a) Let a ∈ R be an element, such that (I : hai) is a prime ideal.√By (II.6) and Propo-
sition II.4.17, ii), there is an index i0 ∈ { 1, ..., m | a < qi } with pi0 = (I : hai).
b) Let i0 ∈ { 1, ..., m }. By the minimality of the primary decomposition, there is an
element !/
\
ai0 ∈ q j qi0 .
j∈{ 1,...,m }\{i0 }

Lemma II.4.18 and (II.6) show

q


I : hai0 i = pi0 .

This concludes the proof. 

II.4.19 Exercise (Irreducible sets). A topological space X is called irreducible, if it is non-

empty, and, if X1 and X2 are closed subsets, such that X = X1 ∪ X2 , then X1 = X or X2 = X.
Let X be a topological space and Y a subset of X. Then, Y inherits a topology as follows:
A subset U ⊂ Y is said to be open, if there is an open subset U e ⊂ X with U = Y ∩ U. e We
call a subset Y ⊂ X irreducible, if it is irreducible with respect to the induced topology.
i) Let X be a noetherian topological space and Z a closed subset. Show that there are
irreducible closed subsets Z1 ,...,Zr , such that

⋆ Z = Z1 ∪ · · · ∪ Zr ,

⋆ Zi 1 Z j , for i , j.

Show also that these closed subsets are uniquely determined. The sets Zi , i = 1, ..., r, are
called the irreducible components of Z.
ii) Let R be a noetherian ring and I ⊂ R an ideal. What is the relation between the
primary decomposition of I and the above decomposition of the closed subset V(I) ⊂
Spec(R) into irreducible components?
II.4.20 Exercise (Primary ideals). Show the following: In the polynomial ring Z[t], a) the
ideal m = h2, ti is maximal and b) the ideal q = h4, ti is m-primary, but c) q is not a power
of m.

68
 II.4. Primary Decomposition

II.4.21 Exercise (A primary decomposition). Let k be a field and R := k[x, y, z]. Set
p1 := hx, yi, p2 := hx, zi, and m := hx, y, zi.
i) Show that p1 and p2 are prime ideals, while m is maximal.
ii) Let I := p1 · p2 . Show that

I = p1 ∩ p2 ∩ m2

and that this a minimal primary decomposition of I.

iii) Which components are isolated and which are embedded?
II.4.22 Exercise (A primary decomposition). We work in the ring R = k[x1 , x2 , x3 , x4 ], k a
field. Show that

h x1 x2 − x4 , x1 x3 − x4 , x2 x3 − x4 i =
= h x1 , x2 , x4 i ∩ h x1 , x3 , x4 i ∩ h x2 , x3 , x4 i ∩ h x1 − x2 , x2 − x3 , x21 − x4 i

is a minimal primary decomposition.

The Second Uniqueness Theorem

II.4.23 Proposition. Let R be a ring, S ⊂ R a multiplicatively closed subset, p ⊂ R a
prime ideal, q ⊂ R a p-primary ideal, and

ϕ : R −→ RS

the natural localization homomorphism. Then:

i) The assertion S ∩ p , ∅ is equivalent to the assertion qe = RS .
ii) If S ∩ p = ∅, then qe is a pe -primary ideal and qec = q.
Proof. i) By Proposition II.3.6, ii), the extended ideal qe equals RS if and only if S ∩q , ∅.
Thus, we have to prove:
Claim. S ∩ p , ∅ ⇐⇒ S ∩ q , ∅.
“⇐=”. This is clear, because q ⊂ p.
“=⇒”. Let s ∈ S ∩ p. Since p is the radical of q, there is an exponent k ≥ 1 with sk ∈ q.
Furthermore, S is a multiplicatively closed subset, so that sk ∈ S . X
ii) We use the description [

qec = q : hsi
s∈S

from Proposition II.3.6, iv). Suppose s ∈ S and r ∈ (q : hsi). Then, r · s ∈ q. Since

sk < q, k ≥ 1, we must have r ∈ q. This shows qec ⊂ q. The other inclusion is contained in
Property I.8.24, i), so that
qec = q.
By Proposition II.3.6, iii), J ce = J holds for every ideal J ⊂ RS . So, in order to verify
√
qe = pe , it suffices to check √ e
c
q = pec .
With Property I.8.19, iv), we verify this as follows:
√ e
c √ ec √
q = q = q = p = pec .

69
II. Noetherian Rings

Finally, we check that qe is a primary ideal. Suppose we are given a, b ∈ R and s, t ∈ R

with
a·b a b
= · ∈ qe .
s·t s t
Then, a · b ∈ qec = q, so that a ∈ q or bk ∈ q, for some k ≥ 1. This implies a/s ∈ qe or
(b/t)k = bk /tk ∈ qe , for some k ≥ 1. 

II.4.24 Proposition. Let R be a ring, S ⊂ R a multiplicatively closed subset, I ⊂ R an

ideal, and
\m
I= qi
i=1
√
a minimal primary decomposition. Set pi := qi , i = 1, ..., m, and suppose that the
numbering is such that

∃m0 ∈ { 0, ..., m }∀i ∈ { 1, ..., m } : S ∩ pi , ∅ ⇐⇒ i > m0 .

Then,
\
m0 \
m0
e
I = qei and ec
I = qi , (II.7)
i=1 i=1

and these are both minimal primary decompositions.

Proof. Using Proposition II.3.6, iii), and Property I.8.24, iv), it is enough to check the
second equality in (II.7). We first apply Proposition II.3.6, iv), and Property I.8.17, iv), to
see m 
[
[ \  [ \ m


ec
I = I : hsi = 
 qi : hsi = qi : hsi .
s∈S s∈S i=1 s∈S i=1

Our assertion amounts to

[\
m

\0 m
qi : hsi = qi .
s∈S i=1 i=1

“⊂”. For s ∈ S and 1 ≤ i ≤ m0 , we have s < pi . This inclusion, therefore, follows

from Lemma II.4.18.
“⊃”. For m0 + 1 ≤ i ≤ m, S ∩ pi , ∅. By Proposition II.4.23, i), we may pick an
element si ∈ S ∩ qi . Then,

∀i ∈ { m0 + 1, ..., m } : s := sm0 +1 · · · · · sm ∈ qi .

Lemma II.4.18 shows

∀i ∈ { m0 + 1, ..., m } : qi : hsi = R.

T
m
The second equation in (II.7) is a minimal primary decomposition, because I = qi
i=1
is one. By Proposition II.4.23, ii), qei is a pei -primary ideal, i = 1, ..., m0. This shows that
the first equation in (II.7) is also a primary decomposition. To check that it is minimal,
we may contract it to R. This is the second equation in (II.7) which is minimal. 

70
 II.4. Primary Decomposition

II.4.25 Second uniqueness theorem. Let R be a ring, I ⊂ R an ideal,

\
m
I= qi (II.8)
i=1

a minimal primary decomposition, and


Ass(I) = p1 , ..., pm .

i) Let { i1 , ..., in } ⊂ { 1, ..., m } be a subset, such that pi j is an isolated associated prime

ideal of I, j = 1, ..., n. Then, the intersection
\
n
qi j
j=1

does not depend on (II.8).

ii) Let p ∈ Ass(I) be an associated prime ideal and set

M(I, p) = r ∈ Ass(I) | r ⊂ p .

Let
 √
{ i1, ..., in } = i ∈ { 1, ..., m } | qi ∈ M(I, p) .
Then, the intersection
\
n
qi j
j=1

does not depend on (II.8).

Proof. For i), we define

S := R \ (pi1 ∪ · · · ∪ pin )
and, for ii),
S := R \ p,
and look at the localization map ϕ : R −→ RS . In both cases,
\
n
I ec = qi j ,
j=1

by Proposition II.4.24. 
II.4.26 Remark. In general, we may associate with any subset { pi1 , ..., pim } ⊂ Ass(I) an
ideal which is the intersection of some of the primary ideals in (II.8) and which does not
depend on I.
II.4.27 Corollary. The primary ideals in a minimal primary decomposition of I which
correspond to the isolated associated prime ideals are uniquely determined by I.

Proof. Let p ⊂ R be an isolated associated prime ideal. We apply i) to {p} ⊂ Ass(I) or ii),
noting M(I, p) = {p}. 

71
II. Noetherian Rings

To further familiarize us with primary decompositions, we study the primary decom-

position of the zero ideal.
II.4.28 Theorem. Let R be a noetherian ring. Write


{ pi1 , ..., pk } = Ass h0i
and let


{ pi1 , ..., pim } ⊂ Ass h0i
be the subset of isolated prime ideals associated with h0i. Then:
i) The nilradical of R is
p
N = h0i = p1 ∩ · · · ∩ pk = pi1 ∩ · · · ∩ pim .
ii) Let p ⊂ R be a prime ideal. Then, there exists an index j ∈ { 1, ..., m } with
pi j ⊂ p,
i.e., the isolated prime ideals associated with h0i are the minimal prime ideals of R.
iii) The zero divisors of R are exactly the elements of the set
[
k
pi .
i=1

Proof. i) Let
h0i = q1 ∩ · · · ∩ qk
√
be a minimal primary decomposition with pi = qi , i = 1, ..., k. Then,
p √ √ √
h0i = q1 ∩ · · · ∩ qk = q1 ∩ · · · ∩ qk = p1 ∩ · · · ∩ pk = pi1 ∩ · · · ∩ pim .
√
ii) By Proposition I.7.2, h0i ⊂ p, so that
pi1 ∩ · · · ∩ pim ⊂ p.
By prime avoidance (Proposition II.4.17), there is an index j ∈ { 1, ..., m } with pi j ⊂ p.
iii) Assume first that a ∈ R is a zero divisor. There is an element b ∈ R \ {0} with
a · b = 0. Since h0i = q1 ∩ · · · ∩ qk , there is an index i0 ∈ { 1, ..., k } with b < qi0 . Since qi0
√
is a primary ideal, it follows that a ∈ qi0 = pi0 .
Next, let
a ∈ p1 ∪ · · · ∪ pk .
If a is nilpotent, then a is a zero divisor and we are done. Otherwise, we may choose
an index i0 ∈ { 1, ..., k } and an exponent l ≥ 1 with al ∈ qi0 . Note that the existence of
non-nilpotent zero divisors implies k ≥ 2. We have


qi0 · q1 ∩ · · · ∩ qi0 −1 ∩ qi0 +1 ∩ · · · ∩ qk ⊂ q1 ∩ · · · ∩ qk = h0i.
Since the primary decomposition is minimal, we infer
J := q1 ∩ · · · ∩ qi0 −1 ∩ qi0 +1 ∩ · · · ∩ qk , h0i.
Choose b ∈ J \ {0}. Then, al · b = 0. Let n ≥ 0 be the largest natural number with an · b , 0.
We see that a · (an · b) = an+1 · b = 0 and that a is a zero divisor. 

72
III
The Nullstellensatz

We prove here Hilbert’s famous Nullstellensatz. It completes our discussion from Section
I.9: Over an algebraically closed ground field k, there is a correspondence between alge-
braic sets in Ank and radical ideals in k[x1 , ..., xn]. This result, therefore, translates algebraic
geometry in affine spaces into commutative algebra and establishes the close ties between
these two areas of mathematics. There are many proofs of the Nullstellensatz. We will
present a very elementary one which is a variant of an argument due to Munshi. The proof
uses some basic facts about integral ring extensions. In order to speak about integral ring
extensions, we need the language of modules. For these reasons, we take the opportunity
to develop the language of modules and the formalism of finite ring extensions in some
detail. We will also discuss normal rings and (Noether) normalization.

III.1 Modules
Let R be a ring. An R-module is an abelian group (M, +, 0) together with a scalar multi-
plication
· : R × M 7−→ M,
such that the following conditions are satisfied:
⋆ ∀a ∈ R∀x, y ∈ M: a · (x + y) = a · x + a · y.
⋆ ∀a, b ∈ R∀x ∈ M: (a + b) · x = a · x + b · x.
⋆ ∀a, b ∈ R∀x ∈ M: (a · b) · x = a · (b · x).
⋆ ∀x ∈ M: 1 · x = x.
III.1.1 Remark. Let (M, +, 0) be an abelian group. The datum of a scalar multiplication
· : R × M −→ M is equivalent to the datum of a ring homomorphism1
ϕ : R −→ End(M).
1
The ring End(M), which is, in general, non-commutative, was described in Example I.1.3, viii).

73
III. The Nullstellensatz

In fact, let a scalar multiplication · : R × M −→ M be given. The first property states

that

µa : M −→ M
x 7−→ a · x

is an endomorphism of the abelian group M. So, we can define

ϕ : R −→ End(M)
a 7−→ µa .

The fourth property says that ϕ(1) = idM , the second and third property express that ϕ is
compatible with addition and multiplication.
If we are give a ring homomorphism2

ϕ : R −→ End(M),

then it is easy to check that

· : R × M −→ M
(a, x) 7−→ ϕ(a)(x)

is a scalar multiplication.
III.1.2 Examples. i) If k is a field, a k-module is the same as a k-vector space (see [33],
§7).
ii) A Z-module is the same as an abelian group. In fact, the endomorphisms µn : M −→
M (see Remark III.1.1), n ∈ N, satisfy the recursion formula

∀n ∈ N∀x ∈ M : µn+1 (x) = (n + 1) · x = n · x + 1 · x = n · x + x = µn (x) + x. (III.1)

Furthermore, it is easy to see that µ−1 (x) = −x (compare Property I.1.2, ii). Thus, µ0 = 0
and


∀n ∈ N∀x ∈ M : µ−n (x) = (−n)· x = (−1)·n · x = n· (−1)· x) = n·(−x) = µn (−x). (III.2)

So, the µk , k ∈ Z, or, equivalently, the scalar multiplication “·” are completely determined
by the condition µ1 = id M . In particular, on every abelian group, there exists at most one
scalar multiplication · : Z × M −→ M.
On the other hand, given an abelian group M, we can start with the constant map
µ0 : M −→ M, x 7−→ 0, and define µn by recursion (see [27], Satz 1.3.8) via (III.1) for
all natural numbers and, using (III.2), for all integers. It is then checked with various
inductions (compare [27], Satz 1.3.12) that

·: Z×M −→ M
(k, x) 7−→ µk (x)

is a scalar multiplication.
2
This implies that ϕ(1) = id M .

74
 III.1. Modules

iii) Let k be a field and R = k[t] the polynomial ring in one variable over k. Any
R-module M is, in particular, a k-vector space, and

f := µt : M −→ M
x 7−→ t · x

is a k-linear map. So, an R-module determines a pair (M, f ) which consists of a k-vector
space M and a k-linear map f : M −→ M.
Suppose, conversely, that M is a k-vector space and f : M −→ M is a k-linear map.
Set
X
n  
i 
k[ f ] := λi · f  n ∈ N, λi ∈ k, i = 0, ..., n ⊂ End(M).
i=0

This is a commutative subring of End(M). By the universal property of the polynomial

ring (Page 5), idk and f ∈ k[ f ] define a (surjective) homomorphism k[t] −→ k[ f ] and,
thus, a homomorphism
ϕ : k[t] −→ End(M).
In this way, we get an R-module structure on M. It satisfies µt = f .
Altogether, we can say that R-modules identify with pairs (M, f ) composed of a k-
vector space M and a k-linear endomorphism f of M. Such objects are intensely studied in
any introduction to linear algebra, especially in the situation when M is finite dimensional.
iv) Let R, S be rings and ϕ : R −→ S a ring homomorphism. Then,

· : R × S −→ S
(a, b) 7−→ ϕ(a) · b

equips S with the structure of an R-module. In particular, we can use idR : R −→ R to

view R as an R-module.
Let R be a ring and M, N R-modules. An R-module homomorphism is a map ϕ : M −→ N,
such that
⋆ ∀x, y ∈ M: ϕ(x + y) = ϕ(x) + ϕ(y).
⋆ ∀a ∈ R∀x ∈ M: ϕ(a · x) = a · ϕ(x).
Module homomorphisms are the maps that are compatible with the given group struc-
tures and scalar multiplications on the modules and, therefore, allow to compare different
modules. As usual, there are different modules which are in a certain sense “indistin-
guishable”. These are modules which are related by an isomorphism. The “categorical”
definition of an isomorphism reads as follows: Let R be a ring and M, N R-modules. A
map ϕ : M −→ N is an isomorphism of R-modules, if
⋆ ϕ is an R-module homomorphism,
⋆ there exists an R-module homomorphism ψ : N −→ M with ϕ ◦ ψ = id M and ψ ◦ ϕ =
idN .
III.1.3 Exercise. Let R be a ring, M, N R-Modules, and ϕ : M −→ N an R-module homo-
morphism. Show that ϕ is an isomorphism if and only if ϕ is bijective.

75
III. The Nullstellensatz

III.1.4 Example and Exercise. Let k be a field, R = k[x], and (M, f ) and (N, g) k-vector
spaces endowed with k-linear endomorphisms. According to Example III.1.2, iii), they
define R-modules. A map ϕ : M −→ N is a homomorphism of R-modules if and only if it
is k-linear and verifies
ϕ ◦ f = g ◦ ϕ,
i.e., the diagram
ϕ
M / N
f g
 ϕ 
M / N
commutes.
The classification of finite dimensional C[x]-modules is provided by the theory of the
Jordan3 normal form (see [33], §54; compare Page 87ff).

Constructions
Let R be a ring and M an R-module. A submodule is a subset N ⊂ M, such that

⋆ N , ∅.

⋆ ∀x, y ∈ N: x + y ∈ N.

⋆ ∀a ∈ R∀x ∈ N: a · x ∈ N.

Note that, for x ∈ N, we have −x = (−1) · x ∈ N. Let x0 ∈ N (N is non-empty). Then,

0 = x0 − x0 ∈ N. This proves that N is a subgroup of M. By the third property, N is also
equipped with a scalar multiplication. So, N inherits the structure of an R-module.
III.1.5 Example. Let R be a ring. According to Example III.1.2, iv), we may view R as an
R-module. The submodules of R are the ideals.
Suppose that M is an R-module and that N ⊂ M is a submodule. Then, we can form
the group M/N (see [30], Satz II.9.4). We set

· : R × M/N −→ M/N


a, [x] −→ [a · x].

The reader should verify that this is well-defined and equips M/N with the structure of an
R-module. It is the quotient module of M by N.
III.1.6 Example and Exercise. Let R be a ring, M, N R-modules, and ϕ : M −→ N a
homomorphism.
i) Then,

⋆ im(ϕ) is a submodule of N.

⋆ ker(ϕ) is a submodule of M.

⋆ The module N/im(ϕ) is called the cokernel of ϕ.

3
Marie Ennemond Camille Jordan (1838 - 1922), french mathematician.

76
 III.1. Modules

ii) The first isomorphism theorem holds, i.e.,

M/ker(ϕ)  im(ϕ).

Let I , ∅ be a possibly infinite index set and (Mi )i∈I a family of R-modules indexed
by I. The cartesian product 
Mi
i∈I

together with componentwise addition and scalar multiplication is an R-module. It is

called the direct product of (Mi )i∈I .
The set
M    
Mi := (xi )i∈I ∈ Mi  xi = 0 for all but finitely many i ∈ I
i∈I i∈I

is a submodule of Mi . It is referred to as the direct sum of (Mi )i∈I . Note that the direct
L i∈I 
sum Mi equals the direct product Mi if and only if I is finite.
i∈I i∈I
If M is an R-module and I , ∅ an index set, we may define Mi := M, i ∈ I. In this
case, we set M M
M := Mi .
i∈I i∈I

For convenience, we also define M

M := {0}.
i∈∅

Let M, N be R-modules, then


HomR (M, N) := ϕ : M −→ N | ϕ is a homomorphism of R-modules

together with the addition

∀ϕ, ψ ∈ HomR (M, N) : ϕ + ψ : M −→ N, x 7−→ ϕ(x) + ψ(x),

and the scalar multiplication

∀a ∈ R∀ϕ ∈ HomR (M, N) : a · ϕ : M −→ N, x 7−→ a · ϕ(x),

is an R-module, too. The neutral element for the addition is the zero homomorphism
0 : M −→ N, x 7−→ 0.
III.1.7 Exercises (The universal properties of theL
direct sum and the direct product). i) Let
R be a ring, (Mi )i∈I a family of R-modules, and Mi its direct sum. Define, for k ∈ I,
i∈I
M
jk : Mk −→ Mi
i∈I
(
m, if i = k
m 7−→ (mi )i∈I with mi = .
0, if i , k

77
III. The Nullstellensatz

L
Prove that Mi has the following universal property: Given an R-module N and a
i∈I
collection
L of homomorphisms fk : Mk −→ N, k ∈ I, there is a unique homomorphism
f: Mi −→ N with f ◦ jk = fk , k ∈ I. In other words,
i∈I
M  
Hom R Mi , N  Hom R (Mi , N) .
i∈I i∈I

ii) Let R be a ring, (Mi )i∈I a family of R-modules, and Mi its direct product. Define,
i∈I
for k ∈ I,

pk : Mi −→ Mk
i∈I
(mi )i∈I 7−→ mk .

Show that Mi has the following universal property: Given an R-module N and a col-
i∈I
lection of homomorphisms fk : N −→ Mk , k ∈ I, there is a unique homomorphism

f : N −→ Mi with pk ◦ f = fk , k ∈ I. In other words,
i∈I
   
Hom R N, Mi  Hom R (N, Mi ) .
i∈I i∈I

Let R be a ring, M an R-module, and S ⊂ M a subset. The submodule generated by

S is (X )
\ n 
hS i := N= ai · xi  n ≥ 1, ai ∈ R, xi ∈ S , i = 1, ..., n .
N⊂M submodule i=1
S ⊂N

We say that the R-module M is finitely generated, if the there is a finite subset S ⊂ M
with
M = hS i.
An R-module M is free, if there are an index set I and an isomorphism
M
ϕ: R −→ M.
i∈I

If I = { 1, ..., n }, we write
M
n
⊕n
R := R.
i=1

III.1.8 Exercise (Finitely generated modules). Show that an R-module M is finitely gen-
erated if and only if there exist a natural number n ∈ N and a surjection

ϕ : R⊕n −→ M

of R-modules.
III.1.9 Exercise (The rank of a free module). Let R be a ring, s, t ∈ N natural numbers,
and ϕ : R⊕s −→ R⊕t a surjective map. Prove that s ≥ t. In particular, R⊕s  R⊕t if and
only if s = t.

78
 III.1. Modules

L
If M is an R-module and there are a finite index set I and an isomorphism ϕ : R −→
i∈I
M, we call the number #I the rank of M. By the last exercise, this is well-defined.
If k is a field, then every vector space has a basis4 (see [33], §28) and, so, every
k-module is free. Many concepts of linear algebra, such as the matrix formalism, may
be extended to free modules. However, the condition of freeness is rather restrictive, in
general. The following lemma gives a first illustration for this.
III.1.10 Lemma. Let R be an integral domain and I ⊂ R an ideal. Then, I is a free
R-module if and only if it is a principal ideal.
Proof. Let I be a principal ideal. If I = h0i, there is nothing to show. Otherwise, I = hai
with a , 0. Consider

ϕ : R −→ R
r 7−→ r · a

Since R is an integral domain, ϕ is injective. The image of ϕ is hai, so that hai  R as

R-module.
Assume I is a free R-module. If it is of rank 0 or 1, then I is a principal ideal. Other-
wise, there is an injective homomorphism ϕ : R⊕2 −→ I. Set f := ϕ(1, 0) and g := ϕ(0, 1).
The elements f and g are non-zero, because ϕ is injective. Observe

ϕ(g, − f ) = g · ϕ(1, 0) − f · ϕ(0, 1) = g · f − f · g = 0.

Since (g, − f ) , 0, this contradicts the injectivity of ϕ. 

Let M be an R-module. The torsion submodule of M is

Tors(M) := x ∈ M | ∃ non-zero divisor a ∈ R : a · x = 0 .

If Tors(M) = {0}, then M is said to be torsion free. If M = Tors(M), then M is said to be

a torsion module.
III.1.11 Remark. Let us briefly verify that Tors(M) is indeed a submodule. Clearly, 0 ∈
Tors(M). Let x ∈ Tors(M), a ∈ R, and b ∈ R a non-zero divisor with b · x = 0. Then,
b · (a · x) = a · (b · x) = a · 0 = 0, so that a · x ∈ Tors(M). If x, y ∈ Tors(M) and a, b ∈ R
are non-zero divisors with a · x = 0 and b · y = 0. Then, a · b is not a zero divisor and

(a · b) · (x + y) = (a · b) · x + (a · b) · y = b · (a · x) + a · (b · y) = 0.

This illustrates the appearance of non-zero divisors in the definition of the torsion sub-
module.
III.1.12 Examples. i) A free R-module is torsion free.
ii) If k is a field, then every k-module is free and, in particular, torsion free.
iii) If A is an abelian group, i.e., a Z-module (Example III.1.2, ii), then x ∈ A lies in
Tors(A) if and only it is an element of finite order.
iv) Let k be a field and (M, f ) a k[x]-module (see Example III.1.2, iii). If M is a
finite dimensional k-vector space, then (M, f ) is a torsion module. In fact, since the
4
if and only if the axiom of choice is admitted ([12], Theorem 4.44)

79
III. The Nullstellensatz

vector space Endk (M) = Homk (M, M) is also finite dimensional, the powers f k , k ∈ N,
are linearly dependent. So, there is a polynomial p ∈ k[x] with p( f ) = 0,5 and, for every
x ∈ M,
p · x = p( f )(x) = 0.
Let M be an R-module and x ∈ M, then

Ann(x) := AnnR (x) := a ∈ R | a · x = 0

is an ideal in R. It is the annihilator of x.

If N ⊂ M is a submodule, we set
 \
Ann(N) := AnnR (N) := a ∈ R | ∀x ∈ N : a · x = 0 = Ann(x).
x∈N

This is the annihilator of N.

III.1.13 Remark. Let M be an R-module. Note that M is in a natural way a module over
the ring R/Ann(M).

Modules over Principal Ideal Domains

The main theorem on finite abelian groups (see [30], II.13.5) classifies all finitely gener-
ated Z-modules. It generalizes to modules over a principal ideal domain R. We start with
the following

III.1.14 Theorem. Let R be a principal ideal domain and M a finitely generated R-

module. Then,

⋆ M/Tors(M) is a free R-module.

⋆ The map

ϕ : M/Tors(M) ⊕ Tors(M) −→ M
(x, y) 7−→ x + y

is an isomorphism.

The rank of the free R-module M/Tors(M) is called the rank of M and is denoted by
rk(M).
III.1.15 Remark. Let M, N be finitely generated R-modules and ϕ : M −→ N a homomor-
phism. It induces a homomorphism

ϕ : M/Tors(M) −→ N/Tors(N).

If ϕ is surjective, then so is ϕ. In particular, rk(M) ≥ rk(N) in that case (see Exercise

III.1.9).
We need several preparations to prove this result.
5
The theorem of Cayley–Hamilton ([33], §36) asserts that we may take p to be the characteristic poly-
nomial of f .

80
 III.1. Modules

III.1.16 Proposition. Let R be a principal ideal domain, m ∈ N a natural number, and

M a free module of rank m. Then, every submodule N of M is free of rank n for some
n ∈ { 0, ..., m }.
Proof. Without loss of generality, we may assume M = R⊕m . We set

Mk := (a1 , ..., ak , 0, ..., 0) ∈ R⊕m | ai ∈ R, i = 1, ..., k
and
Nk := N ∩ Mk , k = 1, ..., m.
We will prove inductively that Nk is a free module, k = 1, ..., m. Obviously, M1 is isomor-
phic to R (as R-module). Thus, N1 is isomorphic to a submodule of R, i.e., to an ideal of
R (Example III.1.5). Lemma III.1.10 shows that N1 is free of rank 0 or 1.
For k ≥ 2, we set

I := b ∈ R | ∃b1 , ..., bk−1 ∈ R : (b1 , ..., bk−1 , b, 0, ..., 0) ∈ Nk .
This is an ideal in R. Pick an element a ∈ R with I = hai and elements a1 , ..., ak−1 ∈ R
with
x0 := (a1 , ..., ak−1 , a, 0, ..., 0) ∈ Nk .
If a = 0, then Nk = Nk−1 and Nk is free by induction hypothesis. For the rest of the
argument, we assume a , 0. For every element x ∈ Nk , there exists an element r ∈ R with
x − r · x0 ∈ Nk−1 . This shows that the homomorphism
ϕk : Nk−1 ⊕ R −→ Nk
(x, r) 7−→ x + r · x0
is surjective. It is also injective, because
Nk−1 ∩ hx0 i = {0}.
So, Nk is free of rank rk(Nk−1 ) + 1. Note that the rank of Nk is at most k. 
III.1.17 Remark. The proposition is false, if R is not a principal ideal domain. In fact, R
is a free module of rank 1. If R contains an ideal which is not principal, we may use the
argument given in the proof of Lemma III.1.10. If R is not an integral domain, then we
pick a non-trivial zero divisor b. Then,


Ann hbi = Ann(b) , {0}.
But, for a non-zero free module, the annihilator is clearly {0}.
III.1.18 Corollary. Let R be a principal ideal domain and M a finitely generated R-
module. Then, every submodule N of M is finitely generated.
Proof. The assumption means that there exist a natural number m ∈ N and a surjection
ϕ : R⊕m −→ M.
Now, ϕ−1 (N) is a submodule of M and, therefore, free. Let n ∈ { 1, ..., m } be its rank and
choose an isomorphism ψ : R⊕n −→ ϕ−1 (N). Then,
ψ ϕ|ϕ−1 (N)
R⊕n −→ ϕ−1 (N) −→ N
is a surjection, and N is finitely generated. 

81
III. The Nullstellensatz

III.1.19 Remark. The corollary is true for every noetherian ring (see Proposition III.1.30
and Lemma III.1.29).
III.1.20 Proposition. Let R be a principal ideal domain and M a finitely generated tor-
sion free R-module. Then, M is free.
Proof. We may clearly assume M , {0}. Let us start with a surjection (see Exercise
III.1.8)
ϕ : R⊕m −→ M.
Let ei := (0, ..., 0, 1, 0, ..., 0), 1 being the i-th entry, and xi := ϕ(ei ), i = 1, ..., m. A subset
X ⊂ M is linearly independent, if

λ1 · x1 + · · · + λn · xn = 0 =⇒ λ1 = · · · = λn = 0

holds for all n ≥ 1, for all pairwise distinct x1 , ..., xn ∈ X, and for all λ1 , ..., λn ∈ R. Note
that, in a non-zero torsion free module, any set of cardinality 1 is linearly independent. Let
n ∈ { 1, ..., m } be the maximal cardinality of a linearly independent subset of { x1 , ..., xm }
and fix a linearly independent subset { xi1 , ..., xin } of { x1 , ..., xm } with n elements. For
i = 1, ..., m, there exist ring elements ai , ai1 , ..., ain ∈ R with

ai · xi + ai1 · xi1 + · · · + ain · xin = 0.

Note that ai , 0, because { xi1 , ..., xin } is linearly independent, i = 1, ..., m. So,

a := a1 · · · · · am , 0.

For i = 1, ..., m, we have

a · xi ∈ N := h xi1 , ..., xin i
and, thus,
a · M ⊂ N.
Note that N is a free module of rank n. By Proposition III.1.16, a · M is also free. Finally,

µa : M −→ M
x 7−→ a · x

is injective, because a , 0 and M is torsion free. Since µa maps M onto a · M, the module
M is free. 
III.1.21 Proposition. Let R be an arbitrary ring, M an R-module, N a free R-module,
and ϕ : M −→ N a surjection. Then, there exists a submodule P ⊂ M, such that
⋆ ϕ|P : P −→ N is an isomorphism,
⋆ the homomorphism

ψ : P ⊕ ker(ϕ) −→ N
(x, y) 7−→ x + y

is an isomorphism.

82
 III.1. Modules

Proof. Let I be an index set, such that

M
N R.
i∈I

Let ei = (ei j ) j∈I be the tuple with eii = 1 and ei j = 0, j ∈ I \ {i}. We pick elements xi ∈ M
with
ϕ(xi ) = ei , i ∈ I.
By the universal property of the direct sum (Exercise III.1.7, i), there is a unique homo-
morphism ̺ : N −→ M that maps ei to xi , i ∈ I. We set

P := im(̺).

Clearly, ϕ ◦ ̺ = idN . In particular, ̺ is injective and, therefore, maps N isomorphically to

P. It also follows that ϕ|P is inverse to ̺.
For x ∈ M, we have



ϕ x − (̺ ◦ ϕ)(x) = ϕ(x) − (ϕ ◦ ̺) ϕ(x) = ϕ(x) − ϕ(x) = 0

and


x = (̺ ◦ ϕ)(x) + x − (̺ ◦ ϕ)(x) .
| {z } | {z }
∈P ∈ker(ϕ)

It follows that ψ is surjective. It is also injective, because P ∩ ker(ϕ) = {0}. 

Proof of Theorem III.1.14. We first show that M/Tors(M) is free. By Proposition III.1.14,
it is enough to verify that this module is torsion free. Let [x] ∈ M/Tors(M) be a torsion
element. There exists a non-zero divisor b with [b · x] = b · [x] = 0. This means b · x ∈
Tors(M). So, there is a non-zero divisor a ∈ R with

(a · b) · x = a · (b · x) = 0.

Since a · b is a non-zero divisor, x ∈ Tors(M) and [x] = 0. To conclude, we apply

Proposition III.1.21 to the surjection M −→ M/Tors(M). 
Let R be a principal ideal domain. In order to understand all finitely generated modules
over R, we need to understand the finitely generated torsion modules. This we will do now.
Let P ⊂ R be a subset, such that for every prime element q ∈ R, there exists one and only
one element p ∈ P with q ∼ p, i.e., P is a set of representatives for the equivalence classes
of prime elements in R with respect to the equivalence relation “being associated”.
Let M , {0} be a non-trivial torsion module over R. In this case, Ann(M) is a non-zero
proper ideal of R. Pick a generator a ∈ R \ (R⋆ ∪ {0}) for Ann(M). The idea is to use the
prime factorization of a to decompose M further.
For any ring element b ∈ R, let

µb : M −→ M
x 7−→ b · x

and
Mb := ker(µb ).

83
III. The Nullstellensatz

III.1.22 Proposition. Let c, d ∈ R be coprime elements and b = c · d. Then,

Mb  Mc ⊕ Md .

Proof. It is evident that Mc ⊂ Mb and Md ⊂ Mc . Note that hci + hdi = h1i. So, let r, s ∈ R
with
r · c + s · d = 1. (III.3)
With this equation, we see that Mc ∩ Md = {0}.
Next, let x ∈ Mb . By (III.3), we have

x = 1 · x = r · (c · x) + s · (d · x).

Now, d · x ∈ Mc and c · x ∈ Md . 
There are distinct elements p1 , ..., p s ∈ P and positive integers k1 , ..., ks with

hai = h pk11 · · · · · pkss i.

By Proposition III.1.22,
M  M pk1 ⊕ · · · ⊕ M pkss .
1

III.1.23 Proposition. Suppose there are an element p ∈ P and a natural number k ≥ 1

with
M = M pk .
Then, there exist positive integers l1 , ..., lt , such that

M  R/hpl1 i ⊕ · · · ⊕ R/hplt i.

We need some preparations for the proof. Let M be any R-mdoule. We call elements
x1 , ..., xu ∈ M independent, if

∀λ1 , ..., λu ∈ R : λ1 · x1 + · · · + λu · xu = 0 =⇒ λ1 · x1 = · · · = λu · xu = 0.

This condition is weaker than linear independence. In fact, independent elements can
exist in torsion modules whereas linearly independent elements can’t.
In the set-up of Proposition III.1.23, M is said to be a p-torsion module. The number

e := min{ k ∈ N | M = M pk }

is the exponent exp p (M) of M. In the following, M is assumed to be a p-torsion module.

Let x ∈ M \ {0}. We call

s := min{ t ∈ N | pt · x = 0 }

the order ord p (x) of x. Clearly,

s ≥ 1, ord p (x) ≤ exp p (M),

and there is an element x0 ∈ M for which equality is achieved.

84
 III.1. Modules

III.1.24 Lemma. In the above setting, let b ∈ R be an element with b · x = 0. Then,

pord p (x) |b.
Proof. We may write b = p s · m with p and m coprime. Assume s < o := ord p (x). As in
the proof of Proposition III.1.22, we find r, s ∈ R with
1 = r · po−s + s · m.
Now,
x = r · po−s · x + s · m · x
has order at most s. This is a contradiction to the definition of the order. 
III.1.25 Lemma. Let M be a p-torsion module of exponent e and x0 ∈ M an element of
order e. Assume y1 , ..., yn ∈ M/hx0 i are non-zero independent elements. Then, there exist
elements x1 , ..., xn ∈ M, such that
⋆ [xi ] = yi , i = 1, ..., n,
⋆ ord p (xi ) = ord p (yi ), i = 1, ..., n,
⋆ x0 , x1 , ..., xn are independent.
Proof. Let y ∈ M/hx0 i, o := ord p (y), and x′ ∈ M an element with [x′ ] = y. Obviously,
ord p (x′ ) ≥ ord p (y).
There exist 0 ≤ s ≤ e and an element m ∈ R which is coprime to m, such that
po · x′ = p s · m · x0 .
If s = e, then po · x′ = 0, and ord p (x′ ) ≤ ord p (y), so that finally ord p (x′ ) = ord p (y).
If s < e, then p s · m · x0 has order e − s, by Lemma III.1.24. This means that x′ has
order o + e − s. But, we also know
o + e − s ≤ e, i.e., o ≤ s.
In this case,
x := x′ − p s−o · m · x0
has order o and [x] = y.
These considerations show that we may find x1 , ..., xn ∈ M, satisfying the first two
properties. Finally, let a0 , a1 , ..., an ∈ R with
a0 · x0 + a1 · x1 + · · · + an · xn = 0.
Then,
a1 · y1 + · · · + an · yn = 0.
By assumption, ai · yi = 0, i = 1, ..., n. According to Lemma III.1.24,
pord p (yi ) |ai , i = 1, ..., n.
But then
ai · xi = 0, i = 1, ..., n,
and also a · x0 = 0. 

85
III. The Nullstellensatz

From now on, we assume that M is finitely generated. Let p ∈ P and suppose M is a
p-torsion module. By Corollary III.1.18, M p is finitely generated. According to Remark
III.1.13, M p is a module over R/hpi. The latter is a field. We shall denote it by K(p). So,
we may associate with M the number

dimK(p) (M p ) ∈ N.

Proof of Proposition III.1.23. Let x0 ∈ M be an element of order r and M = M/hx0 i. We

prove
dimK(p) (M p ) < dimK(p) (M p ).

Note that elements x1 , ..., xn ∈ M p or elements y1 , ..., yn ∈ M p are independent if and only
if they are K(p)-linearly independent (Lemma III.1.24).
Let y1 , ..., yn ∈ M p form a K(p)-basis. By Lemma III.1.25, we can lift these elements
to K(p)-linear independent elements x1 , ..., xn ∈ M p . The elements x0 , x1 , ..., xn ∈ M are
independent. Then, pr−1 · x0 , x1 , ..., xn are also independent. Observe that pr−1 · x0 ∈ M p .
We see
dimK(p) (M p ) ≥ n + 1 = dimK(p) (M p ) + 1.

We now prove the result by induction on d := dimK(p) (M). If d = 0, we claim that

M = {0}. Indeed, if M , {0} and x ∈ M \ {0}, then o := ord p (x) ≥ 1 and po−1 · x is an
element of order 1, i.e., a non-zero element in M p .
Now, suppose the result holds for all natural numbers < dimK(p) (M p ). Pick an element
x0 ∈ M of order e = exp0 (M). Then,


dimK(p) M/hx0 i p < dimK(p) (M p ).

By induction hypothesis, M/hx0 i is generated by independent elements y1 , ..., yn . We con-

struct x1 , ..., xn ∈ M as in Lemma III.1.25. Then, x0 , x1 , ..., xn generate M and are inde-
pendent. This means
M = hx0 i ⊕ hx1 i ⊕ · · · ⊕ hxn i.
Furthermore,
hxi i  R/hpord p (xi ) i, i = 0, 1, ..., n.
This finishes the proof. 

III.1.26 Theorem (Torsion modules over principal ideal domains). Let M be a finitely
generated torsion module. Then, there are positive integers s, t1, ..., t s , s distinct prime
elements p1 , ..., p s ∈ P, and positive integers 1 ≤ ki1 ≤ · · · ≤ kiti , i = 1, ..., s, such that
k1t k
M  R/hpk111 i ⊕ · · · ⊕ R/hp1 1 i ⊕ · · · ⊕ R/hpkss1 i ⊕ · · · ⊕ R/hp ssts i. (III.4)

The integers s, t1 , ..., t s, ki j , j = 1, ..., ti, i = 1, ..., s, and the prime elements6 p1 , ..., p s ∈
P are uniquely determined by M.
6
They depend, of course, on the choice of P.

86
 III.1. Modules

Proof. The existence follows from the previous discussion. The uniqueness follows from
a careful look at submodules of the form N p where N is constructed in some way from M
and p ∈ P is a prime element.
First, we observe

{ p1 , ..., p s } = p ∈ P | M p , {0} .
Hence, s and p1 , ..., p s are uniquely determined.
Next, we point out that


Z/hpk i p = hpk−1 i/hpk i, p ∈ P, k ≥ 1.

This shows
ti = dimK(pi ) (M pi ), i = 1, ..., s.
Likewise one sees
 

# j ∈ { 1, ..., t s } | ki j ≥ l = dimK(pi ) M/pl−1
i · M pi , l ≥ 1, i = 1, ..., s.

From these numbers, one may clearly determine k11 , ..., ksts . 
As an application of this result, let us derive the theorem on the Jordan normal form.
Let R = C[x] and let (M, f ) be a pair in which M is a finite dimensional complex vector
space and f : M −→ M is an endomorphism. By Example III.1.2, ii), and Exercise
III.1.12, iv), this defines a torsion module over C[x].
We decompose it according to Theorem III.1.26. Since C is algebraically closed, a
polynomial p ∈ C[x] is irreducible if and only if it is linear, i.e., of the form c · (x − λ), for
some c ∈ C⋆ , λ ∈ C. We need to understand the C[x]-modules


Mλ,k = C[x]/ (x − λ)k , λ ∈ C, k ≥ 1.

The elements
 
vi := (x − λ)k+1−i , i = 1, ..., k,
form a C-basis for Mλ,k . Set v0 := 0. We have, for i = 1, ..., k,
   
x· vi = λ · vi +(x−λ)· vi = λ · vi +(x−λ)· (x−λ)k+1−i = λ · vi + (x−λ)k+1−(i−1) = λ · vi +vi−1 .

Thus, with respect to the ordered C-basis (v1 , ..., vk ) of Mλ,k , multiplication by x is de-
scribed by the matrix  
 λ 1 0 · · · 0 
 
 0 λ . . . . . . ... 
 
0  , λ ∈ C.
 .. . . . . . . . . . 
 . 
 ..
. λ 1 
..
 .
 
0 ··· ··· 0 λ
III.1.27 Theorem (Finitely generated modules over principal ideal domains). Let R be a
principal ideal domain and M a finitely generated R-module. Then, there are a positive
integer t and elements a1 , ..., at ∈ R \ {0} with

ha1 i ⊃ · · · ⊃ hat i

87
III. The Nullstellensatz

and
M  R⊕rk(M) ⊕ R/ha1 i ⊕ · · · ⊕ R/hat i. (III.5)
Moreover, if a′1 , ..., a′u are other elements with these properties, then t = u and ai ∼ a′i ,
i = 1, ..., t.

Proof. By Theorem III.1.14, it suffices to look at the case rk(M) = 0, i.e., that M is a
torsion module. We use Theorem III.1.26. Let t := max{ ti | i = 1, ..., s }. Define

li1 := · · · := li(t−ti ) := 0 and li(t−ti + j) := ki j , j = 1, ..., ti, i = 1, ..., s,

and
aν = pl11ν · · · · · plssν , ν = 1, ..., t.
By the Chinese remainder theorem I.8.12, these elements clearly have the required prop-
erties.
Since we recover the decomposition (III.4) from (III.5), the asserted uniqueness fol-
lows from the corresponding statement in Theorem III.1.26. 
Note that this theorem includes the main theorem on finitely generated abelian groups
([30], Satz II.13.5).

Noetherian Modules
Let R be a ring and M an R-module. We say that M is noetherian, if every submodule N
of M is finitely generated.
III.1.28 Remark. i) We may view R as an R-module (Example III.1.2, iv). Then, the
submodules of R are the ideals of R (see Example III.1.5). So, R is noetherian as R-
module if and only if R is noetherian as ring.
ii) If R is not noetherian, then there exist finitely generated R-modules which are not
noetherian. In fact, R itself is such an example. It is free of rank 1 as R-module and
contains an ideal I which is not finitely generated.

III.1.29 Lemma. Let M be an R-module and N ⊂ M a submodule. Then, M is noetherian

if and only if N and M/N are noetherian.

Proof. We will use the canonical surjection

π : M −→ M/N.

Assume first that M is noetherian. Every submodule of N is also a submodule of M.

Therefore, N is also noetherian. The fact that any submodule of M/N is finitely generated
is shown with an argument similar to the one in the proof Corollary III.1.18.
Next, assume that N and M/N are noetherian and P is a submodule of M. Let
x1 , ..., xm ∈ P be generators of the module N ∩ P, y1 , ..., yn ∈ M/N generators for π(P) ⊂
M/N, and xm+i ∈ P with π(xm+i ) = yi , i = 1, ..., n. Then, one easily checks that x1 , ..., xm+n
generate P. 

III.1.30 Proposition. Let R be a noetherian ring, M a finitely generated R-module. Then,

M is a noetherian R-module.

88
 III.1. Modules

Proof. There exist a natural number m ≥ 0 and a surjection ϕ : R⊕m −→ M. By Lemma

III.1.29, it suffices to show that R⊕m is noetherian. This will be done by induction on m.
m = 1. This was explained in Remark III.1.28.
m −→ m + 1. Let

N := (0, ..., 0, r) ∈ R⊕(n+1) | r ∈ R .
This is a submodule of R⊕m . It is evidently isomorphic to R. The quotient R⊕(m+1) /N is
isomorphic to R⊕m . In fact,

ϕ : R⊕m −→ R⊕(m+1) −→ R⊕(m+1) /M

(a1 , ..., am ) 7−→ (a1 , ..., am, 0)
(a1 , ..., am, am+1 ) 7−→ [a1 , ..., am , am+1 ]

is an isomorphism. By the induction hypothesis, N and R⊕(m+1) /N are noetherian. By

Lemma III.1.29, R⊕(m+1) is noetherian, too. 
III.1.31 The Nakayama7 lemma. Let R be a local ring with maximal ideal m. Let M be
a finitely generated R-module, such that

m · M = M.

Then,
M = {0}.
Proof. Assume M , {0} and let m ≥ 1 be minimal, such that there is a surjection
ϕ : R⊕m −→ M. Define ei as in the proof of Proposition III.1.21 and set xi := ϕ(ei ),
i = 1, ..., m. Since m · M = M, there are elements a1 , ..., at ∈ m, such that

xm = a1 · x1 + · · · + am · xm ,

i.e.,
(1 − am ) · xm = a1 · x1 + · · · + am−1 · xm−1 .
Since am ∈ m and R is a local ring, (1 − am ) is a unit. This shows that xm is a linear
combination of x1 , ..., xm−1 , so that M is generated by x1 , ..., xm−1 . This contradicts the
choice of m. 
We will discuss two applications of this result.
III.1.32 Krull’s8 intersection theorem. Let R be a noetherian local ring with maximal
ideal m. Then, \
I := mk = {0}.
k∈N

Proof. The intersection I is an ideal. Set


Σ := J ⊆ R | J is an ideal with J ∩ I = m · I .

Note that m · I ∈ Σ, so that Σ , ∅. Let J0 be a maximal element of Σ (see Theorem II.1.1).

By definition, m · I ⊂ J0 .
7
Tadashi Nakayama (1912 - 1964), japanese mathematician.
8
Wolfgang Krull (1899 - 1971), german mathematician.

89
III. The Nullstellensatz

Claim. There is a natural number k ∈ N, such that mk ⊆ J0 .

If the claim is correct, we have
I ⊂ mk ⊂ J0 ,
so that
I ⊂ J0 ∩ I = m · I.
We see
I = m · I.
Hence, I = {0}, by the Nakayama lemma III.1.31. X
Let us prove the claim. Since m is finitely generated, it suffices to check:
∀ f ∈ m∃l ≥ 1 : f l ∈ J0 .
In fact, let f1 , ..., fn generate m and assume fili ∈ J0 , i = 1, ..., n. For k ≥ 1, mk is generated
by the monomials (compare Example I.8.8, ii)
f1k1 · · · · · fnkn with k1 + · · · + kn = k. (III.6)
If k ≥ l1 + · · · + ln , there is an index i0 ∈ { 1, ..., n } with ki0 ≥ li0 , and then the monomial in
(III.6) belongs to J0 .
Let f ∈ m and look at the ascending chain


J0 : h f k i k∈N .
There is an exponent k0 ≥ 1, such that



J0 : h f k0 i = J0 : h f k0 +1 i .
Choose x ∈ (J0 + h f k0 i) ∩ I. This means
∃y ∈ J0 ∃a ∈ R : x = y + a · f k0 .
We see
f k0 +1 · a = f · x − f · y ∈ m · I + J0 = J0
and conclude a ∈ (J0 : h f k0 +1 i) = (J0 : h f k0 i). Consequently, a · f k0 ∈ J0 and x ∈ J0 . This
implies



J0 + h f k0 i ∩ I = J0 ∩ I = m · I and J0 + h f k0 i ∈ Σ.
Since J0 is maximal in Σ and J0 ⊂ J0 + h f k0 i, we infer J0 = J0 + h f k0 i, that is f k0 ∈ J0 . 
Let M and N be finitely generated R-modules and ϕ : M −→ N a homomorphism of
R-modules. We also have the natural quotient map
π : N −→ C := coker(ϕ) = N/im(ϕ).
Note that we get induced homomorphisms
ϕ : M := M/(m · M) −→ N := M/(m · N)
and
π : N −→ C := C/(m · C).

90
 III.2. Finite Ring Extensions

III.1.33 Lemma. i) We have C  coker(ϕ).

ii) The homomorphism ϕ is surjective if and only if ϕ is surjective.
Proof. ii) If ϕ is surjective, then ϕ is obviously surjective, too. If ϕ is surjective, then
coker(ϕ) = {0}. By i), C = {0}. The Nakayama lemma III.1.31 implies C = {0}, i.e., the
surjectivity of ϕ.
i) We have to show that
ker(π) = im(ϕ).
The inclusion “⊃” is clear. For “⊂”, let y ∈ N be such that π([y]) = [π(y)] = 0. So, there
are an element m ∈ m and an element z ∈ N with π(y) = m · [z] in C. This, in turn, means
that there is an element u ∈ im(ϕ) with

y = u + m · z.

We see that, in N, we have

[y] = [u],
and this element belongs to im(ϕ). 
III.1.34 Remark. The proof actually shows that the kernel of the induced surjection

N −→ C

is
im(ϕ) + m · N.
III.1.35 Exercise. Let R be a local ring with maximal ideal m, M a finitely generated
R-module, and N ⊂ M a submodule. Prove that, if

M = m · M + N,

then
M = N.

III.2 Finite Ring Extensions

Let R, S be rings and ϕ : R −→ S a ring homomorphism. We say that ϕ is a finite ring
extension, if ϕ is injective and S is finitely generated as an R-module.
III.2.1 Example. If K and L are fields, then a finite field extension K ⊂ L, i.e., dimK (L) <
∞, is an example for a finite ring extension.
For the following definitions, we assume that ϕ is injective. An element s ∈ S is
integral over R, if there are a positive integer n > 0 and elements a1 , ..., an ∈ R with9

sn + a1 · sn−1 + · · · + an−1 · s + an = 0.

III.2.2 Example. Let K ⊂ L a field extension. Then, y ∈ L is integral over K if and only if
it is algebraic over K (see [8], Satz III.1.6.2, i). If R and S aren’t fields, it is important to
keep in mind that the coefficient of the highest occurring power of s is 1.
9
The symbol “·” refers to the R-module structure of S (Example III.1.2, iv).

91
III. The Nullstellensatz

The ring R is integrally closed in S , if

∀s ∈ S : s is integral over R ⇐⇒ s ∈ R.
We say that R is integrally closed, if R is integrally
√ closed in its total ring of fractions (see
Example II.3.5, ii). A ring R is reduced, if h0i = {0}, i.e., R does not contain non-trivial
nilpotent elements. It is normal, if it is integrally closed and reduced.
III.2.3 Examples. i) We define
R := C[x, y]/hx2 − y3 i.
One readily checks that the polynomial x2 − y3 is irreducible, so that R is an integral
domain.
We will show that R is not normal. For this, we look at the element
x
t := ∈ Q(R)
y
in the quotient field of R. We have

2 x2 y3
t = 2 = 2 =y and t3 = t · y = x.
y y
In particular, t is integral over R, but is is not contained in R. This shows that R is not
normal.
To conclude this example, let us compute the normalization of R, i.e., its integral
closure in Q(R). We look at
S := R[t] ⊂ Q(R).
We see that
S  C[x, y, t]/h x2 − y3 , t2 − y, t3 − x i.
Moreover, one checks that
ϕ : C[ϑ] −→ S
ϑ 7−→ t
and
ψ: S −→ C[ϑ]
t 7−→ ϑ
x 7−→ ϑ3
y 7−→ ϑ2
are homomorphisms which are inverse to each other. It is easy to verify that the poly-
nomial ring C[ϑ] is a normal ring. Hence, so is S . The integral extension ν : R ⊂ S is,
therefore, called the normalization. Using the isomorphism S  C[ϑ] just described, it is
given as
ν : S −→ C[ϑ]
x 7−→ t3
y 7−→ t2 .

92
 III.2. Finite Ring Extensions

Its geometric counterpart is the map

f: A1C −→ V(x2 − y3 ) ⊂ A2k

t 7−→ (t3 , t2 ).

f
−→
AC
1

ii) Set
R := C[x, y]/h x · y i.
We write abusively x, y for [x], [y] ∈ R. Note that R is not an integral domain but reduced
and that
h0i = hxi ∩ hyi
is a minimal primary decomposition. By Theorem II.4.28, iii), we have

{ zero divisors of R } = hxi ∪ hyi.

This shows that x + y is not a zero divisor, so that we may form the non-zero element
x
u := ∈ Q(R)
x+y

in the total ring of fractions of R. It is not contained in R, but satisfies the integrality
condition

2 x2 x x·y=0 x · (x + y) x x x
u −u = 2
− = 2
− = − = 0.
(x + y) x+y (x + y) x+y x+y x+y

This shows that R is not normal.

iii) Set R := C[x]/hx2 i. An element a + b · [x] is a unit if and only if a , 0, a, b ∈ C.
Otherwise, it is nilpotent. We see that Q(R) = R. The ring R is integrally closed, but not
normal, because it is not reduced.
The notion of normality is an important concept in commutative algebra and algebraic
geometry. The above examples already suggest that the notion of normality is related to
singularities. We will study normal rings and normalizations in more detail in Section
III.5 and IV.8.
Let K ⊂ L be a field extension. If it is finite, then every element of L is algebraic over
K. Conversely, if α ∈ L is algebraic over K, then the subfield K(α) ⊂ L it generates is a
finite extension of K. The sum and product of algebraic elements are algebraic, and so on.
The reader may consult, e.g., [8], Satz III.1.6.2, ii), for this. We will now prove similar
results in the realm of commutative rings.

93
III. The Nullstellensatz

III.2.4 Proposition. Let ϕ : R −→ S be an injective ring homomorphism and s ∈ S . The

following conditions are equivalent:
i) The element s is integral over R.
ii) The R-module R[s] ⊂ S is finitely generated.
iii) There is a finitely generated R-module T ⊂ S which contains R[s].
iv) There exists an R[s]-module M, such that
⋆ M is finitely generated as R-module,

⋆ AnnR[s] (M) = y ∈ R[s] | y · M = {0} } = {0}.
Proof. “i)=⇒ii)”. Let n ≥ 1 and a1 , ..., an ∈ R, such that

sn + a1 · sn−1 + · · · + an−1 · s + an = 0.

Then, R[s] is generated as an R-module by 1, s, ..., sn−1. In fact, set

p(x) := xn + a1 · xn−1 + · · · + an−1 · x + an ∈ R[x].

For u ∈ R[s], there is a polynomial g ∈ R[x] with u = g(s). Since the leading coefficient of
p is a unit in R, polynomial division is possible. This implies that there are polynomials
q, r ∈ R[x], such that
g=q· p+r
and deg(r) < n. Then,
u = g(s) = r(s) ∈ h 1, s, ..., sn−1 i.
“ii)=⇒iii)”. We may take T = R[s].
“iii)=⇒iv)”. Set M := T . For y ∈ AnnR[s] (M) ⊂ S , we have y · 1 = 0 and, thus, y = 0.
“iv)=⇒i)”. Let M be generated as an R-module by the elements x1 , ..., xm . There are
elements ai j ∈ R, i, j = 1, ..., m, such that

∀i ∈ { 1, ..., m } : s · xi = ai1 · x1 + · · · + aim · xm .

We form the (m × m)-matrix

B := s · Em − (ai j )i, j=1,...,m ∈ Matm R[s]

with entries in the ring R[s]. By definition

 
 x1 
 
B ·  ...  = 0. (III.7)
 
xm

We note that the theory of determinants works over every commutative ring. In particular,
we have Cramer’s rule10 ([33], §27): For i, j = { 1, ..., m }, let Bi j ∈ Matm−1 (R[s]) be the
matrix that is obtained from B by deleting the i-th column and the j-th row and bi j :=
(−1)i+ j · det(Bi j ). The matrix Bad := (bi j )i, j=1,...,m is the adjoint matrix of B and satisfies

Bad · B = det(B).
10
Gabriel Cramer (1704 - 1752), swiss mathematician.

94
 III.2. Finite Ring Extensions

Multiplying Equation (III.7) by Bad yields the conclusion

∀i ∈ { 1, ..., m } : det(B) · xi = 0, i.e., det(B) ∈ AnnR[s] (M).

By assumption, det(B) = 0. Expanding the determinant of B provides us with an integral-

ity equation for s. 

This proposition has several important consequences.

III.2.5 Corollary. Let R, S be rings and ϕ : R −→ S an injective ring homomorphism.

i) Suppose n ≥ 1 and s1 , ..., sn ∈ S are integral over R. Then, the R-subalgebra
R[s1 , ..., sn] ⊂ S is finitely generated as R-module.
ii) If s, t ∈ S are integral over R, then so are s + t and s · t. In particular,

T := s ∈ S | s is integral over R

is a subring of S .

The subring T in Part ii) of the corollary is called the integral closure of R in S . In the
proof of the above corollary, we use the following

III.2.6 Lemma. Suppose A, B, C are rings and f : A −→ B and g : B −→ C are homo-

morphisms. If B is finitely generated as A-module and C is finitely generated as B-module,
then C is also finitely generated as A-module.

Proof. Suppose x1 , ..., xm ∈ B generate B as A-module and that y1 , ..., yn ∈ C generate C

as B-module. Using the B-module structure of C, we introduce the elements

xi · y j , i = 1, ..., m, j = 1, ..., n.

It is readily checked that these elements generate C as A-module. 

Proof of Corollary III.2.5. i) We prove this result by induction on n. The case n = 1 is

Part ii) of Proposition III.2.4, ii).
n −→ n + 1. For the induction step, we write (compare (I.2))

R[s1 , ..., sn+1] = R[s1 , ..., sn][sn+1 ].

By induction hypothesis, R[s1 , ..., sn] is finitely generated as R-module. Since sn+1 is
integral over R, it is also integral over R[s1 , ..., sn]. By Part ii) of Proposition III.2.4,
R[s1 , ..., sn, sn+1 ] is a finitely generated R[s1 , ..., sn]-module. Lemma III.2.6 shows that
R[s1 , ..., sn, sn+1 ] is also finitely generated as R-module.
ii) By Part i), we know that the R-module R[s, t] ⊂ S is finitely generated. Note
R[s + t] ⊂ R[s, t] and R[s · t] ⊂ R[s, t]. Part iii) of Proposition III.2.4, iii), says that s + t
and s · t are integral over R. 

III.2.7 Corollary. Let R, S , T be rings and ϕ : R −→ S and ψ : S −→ T injective ring

homomorphisms. If S is integral over R and T is integral over S , then T is also integral
over R.

95
III. The Nullstellensatz

Proof. Let t ∈ T . There exist a natural number n ≥ 1 and elements b1 , ..., bn ∈ S with

tn + b1 · tn−1 + · · · + bn−1 · t + bn = 0.

Set S ′ := R[b1 , ..., bn]. The elements b1 , ..., bn are integral over R. By Corollary III.2.5, S ′
is a finitely generated R-module. The element t is integral over S ′ . So,

S ′ [t] = R[b1 , ..., bn, t]

is a finitely generated S ′ -module. Lemma III.2.6 proves that R[b1 , ..., bn, t] is a finitely
generated R-module. It contains R[t]. By Part iii) of Proposition III.2.4, t is integral over
R. 

III.2.8 Corollary. Let R, S be rings, ϕ : R −→ S an injective homomorphism, and T ⊂ S

the integral closure of R in S . Then, T is integrally closed in S .

III.2.9 Example. This corollary can be applied to the homomorphism ϕ : R −→ Q(R). If T

is the integral closure of R in Q(R), then Q(T ) = Q(R) (Exercise III.2.11, ii). This means
that T is an integrally closed ring.
III.2.10 Exercise (Integral ring extensions). Let R, S 1 ,...,S n be rings and fi : R −→ S i ,
i = 1, ..., n, integral ring extensions. Show that
n

f : R −→ Si
i=1


x 7−→ f1 (x), ..., fn(x)

is also an integral ring extension.

III.2.11 Exercises (Total rings of fractions and integral ring extensions). Let R be a ring
and Q(R) its total ring of fractions.
i) Show that the homomorphism

λR : R −→ Q(R)
a
a 7−→
1
is injective.
ii) Let S ⊂ Q(R) be a subring, containing λR (R). Prove that

ψ : Q(R) −→ Q(S )
a λR (a)
7−→
s λR (s)
is an isomorphism, so that, in particular, Q(Q(R)) = Q(R).
iii) Give an example of rings R, S , an injective ring homomorphism ϕ : R −→ S , and
an element a ∈ R, such that

⋆ a ∈ R is not a zero divisor,

⋆ ϕ(a) ∈ S is a zero divisor.

96
 III.3. The Nullstellensatz

Conclude that there is no ring homomorphism ψ : Q(R) −→ Q(S ), such that the diagram
ϕ
R / S
λR λS
 ψ 
Q(R) / Q(S )
commutes.
iv) Let R be a ring and S ⊂ Q(R) the integral closure of R in Q(R). Demonstrate that
S is an integrally closed ring with Q(R) = Q(S ).
III.2.12 Exercise (Normal rings). i) Let R be an integral domain and S ⊂ R a multiplica-
tively closed subset. Note that one may interpret the localization RS as a subring of the
quotient field Q(R). Using this interpretation, show that the following identity holds in
Q(R): \
R= Rm .
m⊂R
maximal ideal

Hint. Let s be an element of the right hand intersection. Look at { r ∈ R | r · s ∈ R }.

ii) Prove that an integral domain R is normal if and only if Rm is normal for every
maximal ideal m ⊂ R.

III.3 The Nullstellensatz

In this section, we will present a fairly elementary proof of Hilbert’s Nullstellensatz, fol-
lowing Swan11 [34]. It is due to Munshi [22]. The next section contains a more geometric
proof, based on Noether’s normalization theorem.
III.3.1 Nullstellensatz — Field theoretic version. Let k be a field and R a finitely gen-
erated k-algebra. If R is a field, then k ⊂ R is a finite field extension.
III.3.2 Exercise (Maximal ideals). Suppose that k is an algebraically closed field and that
m ⊂ k[x1 , ..., xn] is a maximal ideal. Prove that there exists a point (a1 , ..., an) ∈ Ank with

m = h x1 − a1 , ..., xn − an i.

III.3.3 Weak Nullstellensatz. Let k be an algebraically closed field, n ≥ 1 a natural

number, and I ( k[x1 , ..., xn] a proper ideal. Then,

V(I) , ∅.

Proof. There is a maximal ideal m with I ⊂ m (Corollary I.4.8, i). By Exercise III.3.2,
there is a point (a1 , ..., an) ∈ Ank , such that

m = h x1 − a1 , ..., xn − an i.

Then,
(a1 , ..., an) ∈ V(I).
This proves the claim. 
11
Richard Gordon Swan (∗ 1933), US mathematician.

97
III. The Nullstellensatz

III.3.4 Strong Nullstellensatz. Let k be an algebraically closed field, n ≥ 1 a natural

number, and I ⊂ k[x1 , ..., xn] an ideal. Then,

√
I V(I) = I.

Proof. We apply the trick of Rabinovich. Let I = h f1 , ..., fm i ⊂ k[x1 , ..., xn ] and f ∈
I(V(I)) \ {0}. This means


∀a = (a1 , ..., an) ∈ Ank : ∀i ∈ { 1, ..., m } : fi (a) = 0 =⇒ f (a) = 0.

We infer that the polynomials f1 , ..., fm , (1 − x0 · f ) ∈ k[x0 , x1 , ..., xn] do not have a common
zero in An+1 k . By the weak Nullstellensatz III.3.3, there exist polynomials b0 , ..., bm ∈
k[x0 , ..., xm ] with
b0 · (1 − x0 · f ) + b1 · f1 + · · · + bm · fm = 1. (III.8)
Define

ϕ : k[x0 , ..., xm ] −→ k[x1 , ..., xn] f

1
x0 7−→
f
xi 7−→ xi , i = 1, ..., n.

With ci := ϕ(bi ), i = 1, ..., m, Equation (III.8) yields

c1 · f1 + · · · + cm · fm = 1. (III.9)

By construction of the localization R[x1 , ..., xn ] f , there exists a natural number s ∈ N with

di := f s · ci ∈ R[x1 , ..., xn ], i = 1, ..., m.

Multiplying Equation (III.9) by f s gives

d1 · f 1 + · · · + dm · f m = f s
√
and shows f s ∈ I and f ∈ I. 

III.3.5 Corollary. Let k be an algebraically closed field, n ≥ 1 a natural number. Then,

the maps Φ and Ψ defined on Page 43 are bijections which are inverse to each other.

Next, we will prepare the proof of Theorem III.3.1.

III.3.6 Lemma. Let R be an integral domain and R[x] the polynomial ring in one variable
over R. Then, there does not exist an element f ∈ R[x], such that the localization R[x] f is
a field.

Proof. Assume to the contrary that f ∈ R[x] is an element, such that R[x] f is a field.
Obviously, we must have deg( f ) ≥ 1. In particular, 1 + f , 0. There are a polynomial
g ∈ R[x] and an exponent k ∈ N, such that the equation
1 g
= k
1+ f f

98
 III.3. The Nullstellensatz

holds in R(x) := Q(R[x]). It implies the equation

f k = (1 + f ) · g (III.10)

in R[x]. We pass to the ring S := R[x]/h1 + f i. Then, [ f ] = −1 in S and (III.10) gives

[ f ]k = 0. We find
(−1)k = 0
in S . This means S = {0} and h1 + f i = R[x]. So, 1 + f is a unit in R[x]. But this is
impossible, because deg(1 + f ) > 0 (Exercise I.3.10). 

III.3.7 Lemma. Let ϕ : R −→ S be an integral ring extension. Then,

R ∩ S ⋆ = R⋆ .

Proof. The inclusion “⊃” is obvious. For the converse inclusion, let a ∈ R ∩ S ⋆ . Then,
there is an element b ∈ S with a · b = 1. Since b is integral over R, there are a positive
integer n ≥ 1 and elements a1 , ..., an ∈ R with

bn + a1 · bn−1 + · · · + an−1 · b + an = 0.

We multiply this by an−1 and find

b = −a1 − a2 · a − · · · − an−1 · an−2 − an · an−1 .

This shows b ∈ R and a ∈ R⋆ . 

III.3.8 Exercise. Let R and S be integral domains, ϕ : R −→ S an integral ring extension,

and n ⊂ S be a maximal ideal and m := n ∩ R. Show that m is a maximal ideal in R.

III.3.9 Lemma. Let ϕ : R −→ S be an integral ring extension. If S is a field, then so is R.

Proof. This is a direct consequence of Lemma III.3.7: R \ {0} = R ∩ (S \ {0}) = R ∩ S ⋆ =

R⋆ . 

The central ingredient in the proof of the Nullstellensatz is

III.3.10 Proposition. Let R be an integral domain, n ≥ 1 a positive integer, and m ⊂

R[x1 , ..., xn] a maximal ideal with

m ∩ R = {0}.

Then, there exists an element a ∈ R, such that

⋆ Ra is a field,

⋆ Ra ⊂ R[x1 , ..., xn]/m is a finite field extension.

99
III. The Nullstellensatz

Proof. We prove the result by induction on n.

n = 1. Let f ∈ m \ {0} be a non-constant element and l ≥ 1 be its degree. Write
f = a0 · xl + a1 · xl−1 + · · · + al−1 · x + al .
Since R ∩ m = 0, by assumption, we have a0 < m. Thus, we obtain the injective homo-
morphism (compare Exercise II.3.8)
ϕ : Ra0 −→ R[x]/m = R[ξ], ξ := [x].
Observe that
1 ai
g := · f = xl + b1 · xl−1 + · · · + bl−1 · x + bl , bi := , i = 1, ..., l,
a0 a0
is a polynomial in Ra0 [x] with
g(ξ) = 0.
This shows that ϕ is an integral ring extension. By Lemma III.3.9, Ra0 is a field. Obvi-
ously, ϕ is a field extension of degree at most l.
n −→ n + 1. Set S i := R[xi ], i = 1, ..., n + 1. We apply the induction hypothesis to S i
and the polynomial ring
S i [x1 , ..., xi−1 , xi+1 , ..., xn+1 ], i = 1, ..., n + 1.
We infer from Proposition III.3.10 that
S i ∩ m , {0}, i = 1, ..., n + 1.
Pick non-zero elements
fi ∈ S i ∩ m
and write
fi = ai0 · xli + ai1 · xl−1
i + · · · + ail−1 · xi + ail , i = 1, ..., n + 1.
As before, we may assume that ai0 < m, i = 1, ..., n + 1, and
a := a10 · · · · · al+1
0 < m.

We get the injective homomorphism

ϕ : Ra −→ R[x1 , ..., xn+1 ]/m = R[ξ1 , ..., ξn+1], ξi := [xi ], i = 1, ..., n + 1.
Note that
1 a10 · · · · · ai−1 i+1 n+1
0 · a0 · · · · · ai
= ∈ Ra , i = 1, ..., n + 1,
ai0 a
so that we can form the polynomial
1 bij
gi := i · f = xli + bi1 · xl−1
i + · · · + bil−1 · xi + bil , bij := , j = 1, ..., l,
a0 ai0
in Ra [xi ], i = 1, ..., n. We have
gi (ξi ) = 0, i = 1, ..., n.
We see that ϕ is an integral ring extension. As before, we conclude that Ra is a field and
ϕ is a finite field extension. 

100
 III.4. Noether Normalization

Proof of Theorem III.3.1. We find a positive integer n ≥ 1 and a surjection

ϕ : k[x1 , ..., xn] −→ R.

Its kernel is a maximal ideal m ⊂ k[x1 , ..., xn ]. Since the elements of k⋆ are units in
k[x1 , ..., xn], we have
k ∩ m = {0}.
In Proposition III.3.10, we must have Ra = k, so that this proposition immediately yields
the claim. 
III.3.11 Exercise (Study’s lemma12 ). Deduce the following result from the Nullstellensatz:
Let k be an algebraically closed field and f, g ∈ k[x1 , ..., xn ] polynomials. Assume that f
is irreducible and V( f ) ⊂ V(g). Show that f divides g in k[x1 , ..., xn].

III.4 Noether Normalization

Let k be an infinite field, e.g., an algebraically closed field. We first collect some elemen-
tary facts on the polynomial ring k[x1 , ..., xn ].
III.4.1 Lemma. Let f ∈ k[x1 , ..., xn] \ {0} be a non-zero polynomial. Then, there exists a
point (a1 , ..., an) ∈ Ank with
f (a1 , ..., an) , 0.
Proof. We prove this result by induction on n. For n = 1, observe that a polynomial
f ∈ k[x1 ] has at most deg( f ) zeroes and k is infinite.
n −→ n + 1. For the induction step, set S := k[x1 , ..., xn] and let f ∈ S [xn+1 ] be a
non-trivial polynomial and d ∈ N its degree. There are polynomials g0 , ..., gd ∈ S with
gd , 0, such that
f = gd · xdn+1 + · · · + g1 · xn+1 + g0 .
By the induction hypothesis, there is a point (a1 , ..., an) ∈ Ank with gd (a1 , ..., an) , 0. Then,

f (a1 , ..., an, xn+1 ) = gd (a1 , ..., an) · xdn+1 + · · · + g1 (a1 , ..., an) · xn+1 + g0 (a1 , ..., an)

is a non-trivial polynomial in k[xn+1 ]. There exists an element a ∈ k with

f (a1, ..., an, a) , 0,

and this finishes the proof. 

III.4.2 Exercises (Dominant regular maps). Let k be an algebraically closed field. An
affine algebraic variety is an irreducible algebraic set X ⊂ Ank . Recall that an algebraic
set Z ⊂ Ank is irreducible if and only if its coordinate algebra

k[Z] := k[x1 , ..., xn]/I(Z)

is an integral domain. A regular map (see Exercise I.9.8) f : X −→ Y between algebraic

varieties is dominant, if f (X) is dense in Y.
12
Christian Hugo Eduard Study (1862 - 1930), German mathematician.

101
III. The Nullstellensatz

i) Let F : X −→ Y be a regular map between algebraic varieties and F ⋆ : k[Y] −→ k[X]

the corresponding homomorphism of algebras. Show that F is dominant if and only if F ⋆
is injective.
ii) Let X be an algebraic variety. The function field of X is the quotient field


k(X) := Q k[X]

of the coordinate algebra k[X] of X. Show that a dominant morphism F : X −→ Y induces

a field extension F # : k(Y) −→ k(X), such that the diagram

F⋆
k[Y]
_
/ k[X]
_

 
F#
k(Y) / k(X)

commutes.
An element m ∈ k[x1 , ..., xn ] is a monomial, if there are natural numbers k1 , ..., kn ∈ N
with
m = xk11 · · · · · xknn .
The number
deg(m) = k1 + · · · + kn
is the degree of m. The set of monomials is a k-basis for k[x1 , ..., xn]. Let d ≥ 1 be a
natural number. A polynomial f ∈ k[x1 , ..., xn] is homogeneous of degree d, if it is a linear
combination of monomials of degree d.
III.4.3 Remark. Lemma III.4.1 shows that a polynomial f ∈ k[x1 , ..., xn] is homogeneous
of degree d if and only if

∀(a1 , ..., an) ∈ Ank ∀λ ∈ k : f (λ · a1 , ..., λ · an ) = λd · f (a1, ..., an).

The degree of a not necessarily homogeneous non-zero polynomial f ∈ k[x1 , ..., xn] \
{0} is the maximal degree of a monomial occuring with non-zero coefficient in f .
III.4.4 Noether normalization for hypersurfaces. Let d ≥ 1 be a natural number and
f ∈ k[x1 , ..., xn] a polynomial of degree d. Then, there are linear polynomials y1 , ..., yn ∈
k[x1 , ..., xn], polynomials g1 , ..., gd ∈ k[x1 , ..., xn−1 ], and a non-zero constant c ∈ k⋆ , such
that

ϕ : k[x1 , ..., xn] −→ k[x1 , ..., xn]

xi 7−→ yi , i = 1, ..., n,

is an isomorphism and

ϕ( f ) = f (y1 , ..., yn) = c · (xdn + g1 · xd−1

n + · · · + gd−1 · xn + gd ). (III.11)

Proof. There are uniquely determined homogeneous polynomials f0 , ..., fd ∈ k[x1 , ..., xn]
with
⋆ deg( fi ) = i, i = 0, ..., d,

102
 III.4. Noether Normalization

⋆ f = fd + fd−1 + · · · + f1 + f0 ,

⋆ fd , 0.

By Lemma III.4.1, there is an element (b1 , ..., bn) ∈ Ank with c := fd (b1 , ..., bn) , 0.
Note that (b1 , ..., bn) , 0, because d ≥ 1 and fd is homogeneous of degree d, so that
fd (0, ..., 0) = 0. After renumbering, we may assume bn , 0. Set

yn := bn · xn ,
yi := bi · xn + xi , i = 1, ..., n − 1.

We leave it to the reader to check that the homomorphism ϕ : k[x1 , .., xn] −→ k[x1 , ..., xn ],
xi 7−→ yi , i = 1, ..., n, is an isomorphism. Let h := ϕ( f ). We compute (compare Remark
III.4.3)

h(0, ..., 0, xn) = fd (b1 , ..., bn) · xdn + fd−1 (b1 , .., bn) · xd−1
n + · · · + f1 (b1 , ..., bn ) · xn + f0 (b1 , ..., bn ).

This implies that there are polynomials g1 , ..., gd ∈ k[x1 , ..., xn−1 ], such that Equation
(III.11) holds. 

III.4.5 Remarks. i) Let f be a polynomial as in (III.11). Then,

k[x1 , ..., xn ]/h f i

is a finitely generated k[x1 , ..., xn−1 ]-module. Indeed, it is generated by 1, [xn ], ..., [xd−1
n ]
(see the proof of Proposition III.2.4).
ii) Assume that k is algebraically closed. There is a geometric interpretation of the
lemma. We let f be a polynomial as in (III.11) and look at the projection

π : Ank −→ An−1k
(a1 , ..., an) 7−→ (a1 , ..., an−1 )

and its restriction

π := π|V( f ) : V( f ) −→ An−1
e k

to the zero set of f . Then,

⋆ e
π is surjective,

⋆ any fiber of eπ consists of at most d points, or exactly d points when counted with
multiplicity.

This means that the hypersurface V( f ) ⊂ Ank may be presented as a ramified covering of
degree d of affine (n − 1)-space An−1
k . This suggests also that the dimension of V( f ) is
n − 1. We will develop this in Chapter IV.

103
III. The Nullstellensatz

b
b

V( f )
b

b b b b

A1k
ramification/discriminant locus

III.4.6 The projection theorem. Let k be an algebraically closed field, I ⊂ k[x1 , .., xn]
an ideal, and f0 ∈ I an element for which there exist a natural number d ≥ 1 and elements
g1 , ..., gd ∈ k[x1 , ..., xn−1 ], such that
f0 = xdn + g1 · xd−1
n + · · · + gd−1 · xn + gd .
Set
⋆ X := V(I) ⊂ Ank ,
⋆ I ′ := I ∩ k[x1 , ..., xn−1 ],
⋆ X ′ := V(I ′ ) ⊂ An−1
k .

Then, the projection

π : Ank −→ An−1k
(a1 , ..., an) 7−→ (a1 , ..., an−1)
satisfies
π(X) = X ′ .
Proof. “⊂”. Let a = (a1 , ..., an−1) ∈ π(X). There, there exists an element b ∈ k with
(a, b) = (a1 , ..., an−1, b) ∈ X.
For f ∈ I ′ ⊂ I, we have f (a, b) = 0. Since f ∈ k[x1 , ..., xn−1], this means f (a) = 0. We
conclude a ∈ V(I ′ ).
“⊃”. Suppose a = (a1 , ..., an−1) < π(X). We will construct an element g ∈ I ′ with
g(a) , 0, so that a < X ′ .
Claim. Let f ∈ k[x1 , ..., xn]. Then, there exist an element h f ∈ I and polynomials p0 , ...,
pd−1 ∈ k[x1 , ..., xn−1 ] with pi (a) = 0, i = 0, ..., d − 1, such that
f = p0 + p1 · xn + · · · + pd−1 · xd−1
n + hf .

104
 III.4. Noether Normalization

We look at the homomorphism

ϕ : k[x1 , ..., xn ] −→ k[xn ]

xi 7−→ ai , i = 1, ..., n − 1,
xn 7−→ xn ,

i.e.,
ϕ( f ) = f (a, xn).
Since ϕ is surjective, the image ϕ(I) of I is an ideal. Note:


∃b ∈ k : b ∈ V ϕ(I) ⇐⇒ ∃b ∈ k∀ f ∈ I : f (a, b) = 0 ⇐⇒ a ∈ π(X).

Since we assumed a < π(X), we see V(ϕ(I)) = ∅. By the “Nullstellensatz in one variable”,
i.e., the definition of an algebraically closed field, ϕ(I) = k[xn ]. So, there exists an element
h′f ∈ I with
ϕ(h′f ) = ϕ( f ).
We set
g f := f − h′f .
We perform polynomial division by f0 (in k[x1 , ..., xn−1][xn ]). There are a polynomial
q ∈ k[x1 , ..., xn] and polynomials p0 , ..., pd−1 ∈ k[x1 , ..., xn−1 ], such that

X
d−1
g f = q · f0 + pi · xin .
i=0

We look at the equation

X
d−1
0 = g f (a, xn ) = q(a, xn) · f0 (a, xn ) + pi (a) · xin .
i=0

Now, deg( f0 (a, xn )) = d. This implies q(a, xn ) = 0 and pi (a) = 0, i = 0, ..., d − 1. With

h f := h′f + q · f0 ∈ I,

we find
X
d−1
f = hf + pi · xin
i=0

as asserted. X
Using this claim, we find polynomials pi j ∈ k[x1 , ..., xn−1 ] with pi j (a) = 0, i, j =
0, ..., d − 1, and hi ∈ I, i = 0, ..., d − 1, with

1 = p0,0 + p0,1 · xn + · · · + p0,d−1 · xd−1

n + h0
d−1
xn = p1,0 + p1,1 · xn + · · · + p1,d−1 · xn + h1
..
.
d−1
xn = pd−1,0 + pd−1,1 · xn + · · · + pd−1,d−1 · xd−1
n + hd−1 .

105
III. The Nullstellensatz

Define


A := Ed − (pi j )i, j=0,...,d−1 ∈ Matd k[x1 , ..., xn−1] .
The above system of equations can be rewritten as
   
 1   h0 
   
A ·  ...  =  ...  .
 d−1   
xn hd−1

Multiplying by the adjoint matrix (compare Page 94) Aad , we infer

   
 1   h′0 
   
det(A) ·  ...  =  ...  , (III.12)
 d−1   ′ 
xn hd−1

for appropriate polynomials h′0 , ..., h′n ∈ I. We would like to show that g := det(A) ∈
k[x1 , ..., xn−1] is the polynomial we are looking for. The first row in (III.12) shows g ∈ I
and, consequently, g ∈ I ′ . Finally, pi j (a) = 0, i, j = 0, ..., d − 1. This gives g(a) = 1. 
We pause a minute to give an alternative proof of the weak Nullstellensatz III.3.3.
Proof of Theorem III.3.3. We perform induction on n. For n = 0, 1, the theorem is true.
Since an algebraically closed field is infinite, we can apply Noether normalization for
hypersurfaces III.4.4 and assume without loss of generality that I contains an element f0
as in the projection theorem. If I ⊂ k[x1 , ..., xn ] is a proper ideal, so is I ′ ⊂ k[x1 , ..., xn−1 ].
By induction hypothesis, V(I ′ ) , ∅. Since


π V(I) = V(I ′ ),

we also have V(I) , ∅. 

III.4.7 Noether’s normalization theorem. Let k be an infinite field and I ⊂ k[x1 , ..., xn]
a proper ideal. Then, there are linear polynomials z1 , ..., zn ∈ k[x1 , ..., xn] and a natural
number r ≤ n, such that

ψ : k[x1 , ..., xn] −→ k[x1 , ..., xn]

xi 7−→ zi , i = 1, ..., n,

is an isomorphism,

k[x1 , ..., xr ] −→ k[x1 , ..., xn]/ψ(I)

xi 7−→ [xi ], i = 1, ..., r,

is a finite ring extension. If I , h0i, then r < n.

Proof. We perform induction on n. Note that the case I = h0i is trivial.
n = 1. Let f ∈ k[x] \ {0} be a polynomial with I = h f i. Then, k[x]/I is a finite
dimensional k-vector space of dimension at most deg( f ) (compare Remark III.4.5, i).
n −→ n + 1. Again, we may assume I , h0i. We let y1 , ..., yn and ϕ be as in Theorem
III.4.4, i.e., such that ϕ(I) contains an element f0 of the form given in (III.11). As was

106
 III.4. Noether Normalization

explained in Remark III.4.5, i), k[x1 , ..., xn ]/h f0 i is a finite module over k[x1 , ..., xn−1 ]. Set
J := ϕ(I) and
J ′ := J ∩ k[x1 , ..., xn−1 ].
Since f0 ∈ ϕ(I), we have the commutative diagram

k[x1 , ..., xn−1 ] / k[x1 , ..., xn ]/h f0 i

 
k[x1 , ..., xn−1]/J ′ / k[x1 , ..., xn ]/J

in which the vertical maps are surjective and the horizontal ones injective. So, the ring
k[x1 , ..., xn]/J is a finite module over k[x1 , ..., xn−1 ]/J ′ . Now, we apply the induction hy-
pothesis to J ′ ⊂ k[x1 , ..., xn−1 ]. (The reader should pay attention how to combine the
choice of y1 , ..., yn and the choice of elements in k[x1 , ..., xn−1] hidden in the application of
the induction hypotheses to J ′ to a single choice of elements z1 , ..., zn ∈ k[x1 , ..., xn].) 
III.4.8 Lemma. Assume that k is algebraically closed, and let I ⊂ k[x1 , ..., xn] be an ideal
and r ≤ n an integer, such that

k[x1 , ..., xr ] −→ k[x1 , ..., xn]/I

is injective and k[x1 , ..., xn]/I is a finite k[x1 , ..., xr ]-module. Then, the restriction

π := π|V(I) : V(I) −→ Ark

e

of the projection

π : Ank −→ Ark
(a1 , ..., an) 7−→ (a1 , ..., ar )

to the algebraic set V(I) is surjective.

Proof. Set si := [xi ] ∈ k[x1 , ..., xn]/I, i = r + 1, ...., n. Since si is integral over S :=
k[x1 , ..., xr ], there are a positive integer di ∈ Z and a polynomial

fi (xi ) = xdi i + g1 · xidi −1 + · · · + gdi −1 · xi + gdi ∈ S [xi ] (III.13)

with
fi (si ) = 0,
i.e.,
fi ∈ I ∩ k[x1 , ..., xi ], i = r + 1, ..., n.
Define, for i = r + 1, ..., n,

Ji := I ∩ k[x1 , ..., xi]

and

πi : Aik −→ Ai−1k
(a1 , ..., ai) 7−→ (a1 , ..., ai−1 ).

107
III. The Nullstellensatz

Due to (III.13), the projection theorem III.4.6 implies that πi maps V(Ji ) surjectively
onto V(Ji−1 ), i = r + 2, ..., n, and V(Jr+1 ) surjectively onto Ark . Since

π = πr+1 ◦ · · · ◦ πn ,

this gives the assertion. We will see in Remark IV.2.5, ii), a more conceptual proof of this
statement. 

III.4.9 Exercises (Noether normalization). We study the map

v: C −→ C3
t 7−→ (t3 , t4 , t5).

i) Show that the image of v is the set V(a) with

a := h z22 − z1 · z3 , z31 − z2 · z3 , z23 − z21 · z2 i.

ii) Prove that the projection π : C3 −→ C, (z1 , z2, z3 ) 7−→ z1 , gives a Noether normal-
ization of V(a), such that z2 is a primitive element for the corresponding field extension.
Determine the discriminant locus.
III.4.10 Exercise (Noether normalization). Let S be an integral domain and R ⊂ S a
subring, such that S is finitely generated as R-algebra. Prove that there are a non-zero
element f ∈ R \ {0}, a natural number n, and elements y1 , ..., yn ∈ S , such that

⋆ y1 , ..., yn are algebraically independent over R,13

⋆ the induced homomorphism ϕ : R f [y1 , ..., yn ] −→ S f is an integral ring extension.

III.4.11 Exercise (Universal property of normalization). An affine algebraic variety X is

said to be normal, if its coordinate algebra k[X] is normal (compare Exercise III.4.2). Let
k be an algebraically closed field and X an affine algebraic variety over k. Show that there
are a normal affine algebraic variety X e and a dominant regular map

e −→ X
ν: X

which are universal, i.e., for every normal algebraic variety Z and every dominant regular
map ϕ : Z −→ X, there is a unique regular map ϑ : Z −→ X e with ϕ = ν ◦ ϑ:

∃!ϑ e
Z ❃❴ ❴ ❴/ X
❃❃
❃❃
ν
ϕ ❃❃❃
 
X.

e ν) is the normalization of X.
The pair (X,
13
See Page 8 for the definition of algebraically independent elements.

108
 III.5. Normal Rings

III.5 Normal Rings

III.5.1 Lemma. Let R be a factorial ring.14 Then, R is a normal ring.

Proof. Recall that a factorial ring is an integral domain. Let Q(R) be the quotient field
of R and s ∈ Q(R) \ {0} a non-zero element which is integral over R. Pick d ≥ 1 and
a1 , ..., ad ∈ R with
sd + a1 · sd−1 + · · · + ad−1 · s + ad = 0. (III.14)
According to Lemma I.6.11, there are elements q, r ∈ R \ {0} which are coprime and
satisfy
q
s= .
r
By Equation (III.14),
rd−1 · sd ∈ R.
Therefore,
qd = rd · sd ∈ hri.
Since q and r are coprime, this is possible if and only if r is a unit of R. This implies
s ∈ R. 

III.5.2 Proposition. Let R be a normal noetherian local ring. Then, R is an integral

domain.
√
Proof (compare Example III.2.3, ii). By definition, a normal ring is reduced, i.e., h0i =
h0i. According to Corollary II.4.7, there are a natural number r ≥ 1 and prime ideals p1 ,
..., pr with
h0i = p1 ∩ · · · ∩ pr .
We have to prove that r = 1. So, let us assume r ≥ 2. We choose elements

⋆ f ∈ p1 \ (p2 ∪ · · · ∪ pr ) (compare Proposition II.4.17, i),

⋆ g ∈ (p2 ∩ · · · ∩ pr ) \ p1 .

Then,

⋆ f + g < p1 ∪ · · · ∪ pr , i.e., f + g is not a zero-divisor (Theorem II.4.28, iii),

⋆ f · g ∈ p1 · (p2 ∩ · · · ∩ pr ) ⊂ p1 ∩ · · · ∩ pr = h0i.

We look at the element (compare Example III.2.3)

f
u := ∈ Q(R).
f +g

It satisfies
f2 f ·g=0 f · ( f + g) f
u2 = = = = u, i.e., u · (u − 1) = 0.
( f + g)2 ( f + g)2 f +g
14
See Page 24.

109
III. The Nullstellensatz

This shows that u is integral over R. Since R is assumed to be integrally closed in Q(R),
this means that u ∈ R. Note u , 0, because f , 0, and u , 1, because g , 0. Hence, u
and (1 − u) are zero divisors in R. Thus, they are both contained in the maximal ideal m
of R. But then 1 = u + (1 − u) ∈ m, and this is impossible. 
III.5.3 Exercises15 . i) Let R1 and R2 be integral domains. Describe the total ring of frac-
tions Q(R1 × R2 ) in terms of the quotient fields Q(R1 ) and Q(R2 ).
ii) Use Part i) to construct a normal ring which is not an integral domain.
We look at integral domains R, S , and at an injective ring extension ϕ : R −→ S . By
the universal property of a quotient field (Page 25), it induces a field extension
e
ϕ : Q(R) −→ Q(S ).
III.5.4 Remark. If ϕ is a finite ring extension, then e
ϕ is a finite field extension.
We need to recall some results from the theory of fields. Let K be a field, n ≥ 1, and
f (x) = xn + a1 · xn−1 + · · · + an−1 · x + an
a monic polynomial in K[x], i.e., a polynomial with leading coefficient one. Let K ⊂ L
be a field extension, such that f splits in L, i.e., there are elements α1 , ..., αn ∈ L, such that
f (x) = (x − α1 ) · · · · · (x − αn ) in L[x].
The discriminant of f is Y
∆( f ) := (αi − α j ).
i, j∈{ 1,...,n }:
i, j

By the theory of symmetric functions ([17], Chapter IV, §6), ∆( f ) is a polynomial in

a1 , ..., an with integral coefficients and does not depend on L. In particular,
∆( f ) ∈ K.
III.5.5 Remark. We have ∆( f ) = 0 if and only if f has a multiple root in L.
An irreducible monic polynomial f ∈ K[x] is separable, if ∆( f ) , 0. An arbi-
trary non-constant monic polynomial f ∈ K[x] is separable, if its irreducible factors,
normalized to have leading coefficient one, are separable. A field k is perfect, if every
non-constant monic polynomial f ∈ K[x] is separable.
III.5.6 Examples. i) Every field of characteristic zero is perfect ([8], Korollar III.3.4.8;
Exercise III.5.7, ii).
ii) Every finite field is perfect ([8], Korollar III.4.1.9; Exercise III.5.8, ii).
iii) Every algebraically closed field is perfect. (This is trivial, because the only irre-
ducible polynomials are those of degree one.)
iv) Let k be a field of characteristic p > 0 and k(t) := Q(k[t]) the quotient field of the
polynomial ring in one variable over k. Then, k(t) is not perfect. In fact, the polynomial
x p − t is not separable, because
x p − t = (x − ϑ) p
for every field extension k(t) ⊂ L and every element ϑ ∈ L with ϑ p = t.
15
It might be good to recall Exercise I.4.22

110
 III.5. Normal Rings

III.5.7 Exercises (Separable polynomials). Let K be a field and f ∈ K[x] an irreducible

monic polynomial. Define the derivative f ′ of f by the usual rules of analysis ([27], Page
127).
i) Let K ⊂ L be a field extension, such that f and f ′ split over L. Show that f is
separable if and only if f and f ′ do not have a common zero in L.
ii) Prove that f is inseparable, i.e., not separable, if and only if f ′ = 0.
III.5.8 Exercises (Perfect fields). i) Let k be a perfect field of characteristic p > 0. Show
that, for every element a ∈ k and every s ≥ 1, there is an element b ∈ k with
s
b p = a.

ii) Prove that a field k with the property that, for every element a ∈ k, there is an
element b ∈ k with b p = a is perfect.
Hint. Use Exercise III.5.7, ii), to show that an inseparable polynomial f ∈ k[x] lies in the
subring k[x p ].
Let K ⊂ L be a finite field extension and α ∈ L. Then,

mα : L −→ L
v 7−→ α · v

is a K-linear automorphism. The minimal polynomial of α is the minimal polynomial

µα ∈ K[x] of the linear map mα (see [33], §36). It is, by definition, monic and irreducible.
We say that α ∈ L is separable over K, if µα ∈ K[x] is a separable polynomial. The field
extension K ⊂ L is separable, if every element α ∈ L is separable over K.
Let K ⊂ L be a field extension. An element α ∈ L is primitive for this field extension,
if L = K(α). Here, K(α) is the smallest subfield of L that contains K and α.
III.5.9 Remark. If the primitive element α ∈ L is algebraic over K, then K ⊂ L is a finite
extension and every element can be written as a polynomial in α with coefficients in K.
The degree of this polynomial can be chosen to be smaller than the degree of the minimal
polynomial µα of α. In fact, if r = deg(µα ), then 1, α, ..., αr−1 is a K-basis for L.

III.5.10 Theorem of the primitive element. Let K ⊂ L be a finite separable field exten-
sion. Then, it has a primitive element.

Proof. [8], III.4.2.3; [17], Theorem V.4.6. 

III.5.11 Lemma. Let R be an integral domain, K := Q(R) its quotient field, and K ⊂ L a
finite field extension.
i) If there exists a primitive element for the field extension, then there exists also a
primitive element which is integral over R.
ii) Assume that R is also normal, and let α ∈ L be an element which is integral over
R. Then, the minimal polynomial of α lies in R[x].

Proof. i) Let α ∈ L be a primitive element. Then, there are a natural number r ≥ 1 and
elements a1 , ..., ar ∈ K with

αr + a1 · αr−1 + · · · + ar−1 · α + ar = 0.

111
III. The Nullstellensatz

Since K is the quotient field of R, we may find an element s ∈ R \ {0} with s · ai ∈ R,

i = 1, ..., r. We multiply the above equation by sr and find

(α · s)r + (a1 · s) · (α · s)r−1 + · · · + (ar−1 · sr−1) · (α · s) + (ar · sr ) = 0.

This shows that the element α · s is integral over R. Finally, K(α · s) = K(α), because
s ∈ K⋆.
ii) Let L ⊂ L be a field extension, such that µα splits in L, and write

µα = (x − α1 ) · · · · · (x − αr ) in L[x],

for suitable elements α1 , ..., αr ∈ L. We may assume that the numbering is such that
α = α1 . Note that µα is also the minimal polynomial for αi , i = 2, ..., r. Thus, there are
K-linear isomorphisms
ψi : K[x]/hµα i −→ K(αi ) (III.15)
sending x to αi , i = 1, ..., r. We deduce that αi is integral over R, i = 1, ..., r. The
coefficients of µα are polynomials in α1 , ..., αr with integer coefficients. By Corollary
III.2.5, ii), they are integral over R. Since R is integrally closed in K, they must actually
belong to R, and this concludes the argument. 

III.5.12 Finiteness of integral closure I. Let R be a normal integral domain with quotient
field K := Q(R), K ⊂ L a finite separable field extension, and S the integral closure of R
in L.16 Pick a primitive element17 α ∈ S for the given field extension, and let ∆ ∈ R \ {0} be
the discriminant of the minimal polynomial µα ∈ R[x].18 Then, the following properties
hold true:
i) If r = deg(µα ), then S is contained in the R-submodule of L that is generated by the
elements
1 α αr−1
, , ..., .
∆ ∆ ∆
In particular, for f ∈ S , there exists a polynomial q ∈ R[x] with 0 ≤ deg(q) < deg(µα ),
such that
∆ · f = q(α).
ii) We have
S ∆ = R∆ [α].
In particular, R∆ [α] is normal.

III.5.13 Remark. Assume that R is noetherian. Since the R-module (1/∆) · R[α] is finitely
generated, the R-submodule S is also finitely generated (Proposition III.1.30), i.e., R ⊂ S
is a finite ring extension. The theorem does not necessarily hold, if K ⊂ L is not separable!
See [15], Theorem 100, for a counterexample. In the case of an inseparable extension,
we have to add additional assumptions on R for the theorem to remain true (see Theorem
III.5.14).
16
See Page 95.
17
See Theorem III.5.10 and Lemma III.5.11, i).
18
See Lemma III.5.11, ii)

112
 III.5. Normal Rings

Proof of Theorem III.5.12. i) We fix a finite field extension K ⊂ M, such that µα splits in
M. Let α1 , ..., αr ∈ M be such that

µα (x) = (x − α1 ) · · · · · (x − αr ) ∈ M[x].

We may assume α = α1 . Since µα belongs to R[x] and is monic, the equation µα (αi ) = 0
shows that αi is integral over R, i = 1, ..., r.
Next, let f ∈ S be an element of the integral closure of R in L. There are elements
q0 , ..., qr−1 ∈ K with
f = q0 + q1 · α + · · · + qr−1 · αr−1 .
We define
fi = q0 + q1 · αi + · · · + qr−1 · αr−1
i , i = 1, ..., r.
Note that f = f1 . By (III.15), there is a K-linear isomorphism K(α) −→ K(αi ) which
maps α to αi and, therefore, f to fi , so that fi is integral over R, i = 1, ..., r. We form the
matrix  
 1 α1 · · · αr−11 
 . . 
A :=  .. .. ..  ∈ Matr (M).
.
 
1 αr · · · αr−1
r

By definition,
   
 q0   f1 
   
A ·  ...  = 
..
. 
   
qr−1 fr
We multiply this equation by the adjoint matrix Aad of A (see Page 94) and find
   
 q0   p1 
   
det(A) ·  ...  = 
..
.  .
   
qr−1 pr

The entries of Aad are polynomials in the elements αij ∈ M, i = 1, ..., r, j = 0, ..., r − 1,
with integer coefficients. This implies that they are integral over R (see Corollary III.2.5,
ii). Since f1 , ..., fr are also integral over R, we infer that p1 , ..., pr are integral over R. Note
that det(A) is a Vandermonde19 determinant (see [19], Kapitel IV, §3, Beispiel 3). Hence,
Y
det(A) = (αi − α j ).
r≥i> j≥1

This implies that det(A) is integral over R. We finally see that

∆ · qi = ±det(A) · pi

is integral over R, i = 1, ..., r. Furthermore, ∆ · qi ∈ K and R is normal, so that ∆ · qi ∈ R,

i = 1, ..., r. We infer
∆ · f ∈ R[α].
19
Alexandre-Théophile Vandermonde (1735 - 1796), French musician, mathematician, and chemist.

113
III. The Nullstellensatz

ii) We clearly have

R[α] ⊂ S ,
and Part i) shows
S ⊂ R∆ [α].
This yields the assertion. 

III.5.14 Finiteness of integral closure II. Let k be an infinite perfect field,20 R a finitely
generated k-algebra and integral domain with quotient field K := Q(R), K ⊂ L a finite
field extension. Then, the integral closure S of R in L is a finitely generated R-module.

Proof. Let n ∈ N be a natural number, such that R may be generated over k by n elements.
By the Noether normalization theorem III.4.7, there are a natural number 0 ≤ r ≤ n and a
finite ring extension
ϕ : k[t1 , ..., tr ] −→ R.
This means that we may assume without loss of generality that R = k[t1 , ..., tr ]. In partic-
ular, the theorem of Gauß I.6.4 and Lemma III.5.1 show that R is normal.
By Theorem III.5.12, we only have to treat the case that K ⊂ L is an inseparable
extension. We explain how we may reduce to the separable case. We fix an algebraically
closed extension L ⊂ K ([8], Theorem III.2.1.8). Let p > 0 be the characteristic of the
field k. Let i = 1, ..., r, s ≥ 1, and ϑi,s ∈ K a root of the polynomial
s
x p − ti ∈ K[x].

Then,
s s
x p − ti = (x − ϑi,s ) p ∈ K[x].
In other words, ti has a unique p s -th root in K, i = 1, ..., r, s ≥ 1. Thus, we write
s
ti1/p := ϑi,s , i = 1, ..., r, s ≥ 1.

Define21
s s
Ks := k(t11/p , ..., tr1/p ), s ≥ 1.
Note
∀s ≥ 1 : Ks ⊂ Ks+1 .
Therefore, we may form [
K∞ := Ks .
s≥1

This is a subfield of K. For every element a ∈ K∞ , there is an index s0 ≥ 1 with a ∈ Ks0 ,

so that we may define

s(a) := min s ≥ 1 | a ∈ Ks . (III.16)
Claim. The field K∞ is perfect.
20
e.g., an algebraically closed field as in Example III.5.6, iii)
s s
21
By definition, this is the smallest subfield of K that contains k, t11/p , ..., tr1/p .

114
 III.5. Normal Rings

We apply Exercise III.5.8, ii). The fact that k is perfect (Exercise III.5.8) gives
 s
Ks = b ∈ K | b p ∈ K , s ≥ 1, (III.17)

i.e., Ks consists of the p s -th roots of elements of K, s ≥ 1. Now, let a ∈ K∞ . Since K

is algebraically closed, there is an element b ∈ K with b p = a. By Equation (III.17),
b ∈ Ks(a)+1 ⊂ K∞ . X
The composite L · K∞ of L and K∞ in K is the smallest subfield of K that contains both
L and K∞ . The extension K∞ ⊂ L · K∞ is finite and, by the claim, separable. Let β ∈ L · K∞
be a primitive element (Theorem III.5.10) and µβ ∈ K∞ [x] its minimal polynomial.
Claim. There is an index s0 ≥ 1 with L ⊂ Ks0 [β] and µβ ∈ Ks0 [x]. In particular, Ks0 [β] =
Ks0 (β).
We have L ⊂ K∞ [β]. Let b ∈ L, m ∈ N, and λ0 , ..., λm ∈ K∞ with

b = λ0 + λ1 · β + · · · + λm · β m .

Then, in the notation of (III.16),


b ∈ Ks [β], s := max s(λ0 ), ..., s(λm ) .

Now, let (b1 , ..., bn) be a K-basis for L. The previous discussion shows that there is an
index s0 ≥ 1 with bi ∈ Ks0 [β], i = 1, ..., n. This clearly implies L ⊂ Ks0 [β]. Likewise,
the fact that µβ ∈ K∞ [x] has only finitely many coefficients implies that we may suppose
µβ ∈ Ks0 [x]. X
We choose s0 so large that Ks0 [x] contains the minimal polynomial µβ of β. The
field extension Ks0 ⊂ Ks0 [β] is separable, because the µβ is a separable polynomial ([8],
Korollar III.3.4.13).
Since L ⊂ Ks0 [β] is a finite field extension, we may replace L by Ks0 [β]. In fact, the
integral closure S of R in L is an R-submodule of the integral closure T of R in Ks0 [β]. If
T is finitely generated as R-module, then so is S (Proposition III.1.30).
For the field extension Ks0 ⊂ Ks0 [β], we may apply Theorem III.5.12. It remains to
investigate the extension K ⊂ Ks0 . Recall that R = k[t1 , ..., tr ].
s0 s0
Claim. The elements t11/p , ..., tr1/p are algebraically independent over k.
Indeed, if q ∈ k[x1 , ..., xr ] is a polynomial with
s0 s
q(t11/p , ..., tr1/p 0 ) = 0,
s0 s0 s0
then u := q p 0 is a polynomial in the subring k[x1p , ..., xrp ] ⊂ k[x1 , ..., xr ]. So, u(t11/p , ...,
s

s0
tr1/p ) is actually a polynomial in t1 , ..., tr , i.e., there exists a polynomial u′ ∈ k[x1 , ..., xr ]
with
s0 s
u′ (t1 , ..., tr ) = u(t11/p , ..., tr1/p 0 ).
Since t1 , ..., tr are algebraically independent over k, u′ = 0. Using the injective map

N×r −→ N×r
(ν1 , ..., νr ) 7−→ (p s0 · ν1 , ..., p s0 · νr ),

115
III. The Nullstellensatz

we find a bijection between the monomials in u and the monomials in u′ . We infer u = 0

and q = 0. X
By this claim,

ι : k[x1 , ..., xr ] −→ Ks0

s0
xi 7−→ ti1/p , i = 1, ..., r,

is an injective homomorphism, and Ks0 is isomorphic to the quotient field of k[x1 , ..., xr ].
The polynomial
s0 s0
ring k[x1 , ..., xr ] is normal (Theorem I.6.4 and Lemma III.5.1), so that
k[t11/p , ..., tr1/p ] is integrally closed in Ks0 . It remains to show that
s0 s
k[t1 , ..., tr ] −→ k[t11/p , ..., tr1/p 0 ]
s0
is a finite ring extension. By Corollary III.2.5, i), it is enough to show that τi := ti1/p is
s0
integral over k[t1 , ..., tr ], i = 1, ..., r. Finally, τi satisfies the integrality equation τip −ti = 0,
i = 1, ..., r. 

III.5.15 Corollary (Finiteness of normalization). Let k be an infinite perfect field, R a

finitely generated k-algebra and integral domain, and R ⊂ S the normalization of R, i.e.,
the integral closure of the ring R in its quotient field Q(R).22 Then, the integral closure S
is a finitely generated R-module and, in particular, a finitely generated k-algebra.

Proof. By Theorem III.4.7, there are r ∈ N and a finite ring extension

ϕ : k[x1 , ..., xr ] −→ R.

By Corollary III.2.7, the integral closure of k[x1 , ..., xr ] in Q(R) equals the integral closure
of R in Q(R). So, we may apply Theorem III.5.14 to k[x1 , ..., xr ] and L := Q(R). 
III.5.16 Remark. Not every noetherian ring has a finite normalization. We refer to [24]
for a survey on counterexamples.

22
See Example III.2.9.

116
IV
Dimension Theory

The concept of dimension is very important in many areas of mathematics. That might be
partially due to the fact that we have a good intuition for dimension. It is our impression
that we live in a three-dimensional world, so that, adding time, we may visualize up to
four real dimensions. Now, it becomes an important task to introduce in a mathematically
rigid way a concept of dimension that matches our intuitive expectations. In commutative
algebra, this concept is the Krull dimension of a ring. It will be defined and discussed in
this chapter. For integral domains which are finitely generated over a field, it is possible to
identify the Krull dimension with the transcendence degree of the quotient field. Another
important concept is the embedding dimension of a local ring. In the geometric context,
the embedding dimension is the dimension of the tangent space at a point.1 With the
embedding dimension, we may define regular and singular points of algebraic varieties.
Working over the complex numbers, a point of a variety is regular if and only if the variety
looks locally2 around that point like an open subset of Ck , k the dimension of the variety,
i.e., is a complex manifold around that point. So, near regular points, the local geometry
is very easy, whereas, at singular points, it can become very intricate. We will see some
basic examples for this. We will also highlight some algebraic consequences of regularity.

IV.1 Krull Dimension

Let R be a ring. Its Krull dimension is

dim(R) := sup k ∈ N | ∃ prime ideals p0 ( · · · ( pk ∈ N ∪ {∞}.

IV.1.1 Examples. i) A field has Krull dimension zero. Geometrically, a field is the coor-
dinate algebra of a point. So, this matches our expectation that a point should be zero-
dimensional.
1
We recall that, for varieties over algebraically closed ground fields, we have a correspondence between
points of the variety and maximal ideals in the coordinate algebra. The local ring attached to a point of the
variety is the localization of the coordinate algebra of the variety at the corresponding maximal ideal.
2
in the euclidean topology of AnC

117
IV. Dimension Theory

ii) A principal ideal domain is either a field or has Krull dimension one. For example,
let k be a field. Then, k[x] is one-dimensional as ring. It is the coordinate algebra of
the affine line A1k . Again, our feeling that the affine line should be one-dimensional is
confirmed. It is worthwhile noting that Z also has Krull dimension one. For this reason,
one draws Spec(Z) in a fashion which supports this fact (see [21], Chapter II.1, Example
C).3 It is, indeed, one of the big achievements of modern algebraic geometry that rings
such as Z and k[x] may be described on an equal footing.
IV.1.2 Remarks. i) Let k be a field, n ∈ N, I ⊂ k[x1 , ..., xn] an ideal, and

R := k[x1 , ..., xn]/I.

Then, the Krull dimension of R is the supremum of all natural numbers s for which there
exist irreducible algebraic sets

V(I) ⊃ V0 ) · · · ) V s

in the affine n-space Ank . This illustrates the idea behind Krull dimension: Given two
irreducible subsets Y, Z ⊂ Ank , an inclusion Y ⊂ Z is only possible, if Y = Z or the
dimension of Z is larger than the one of Y. For example, an irreducible zero-dimensional
algebraic set is just a point. Any irreducible algebraic set properly containing it has to be
at least one-dimensional, and so on.
ii) Let k be a field, n ∈ N, and R := k[x1 , ..., xn]. We have the chain of prime ideals

h0i ( hx1 i ( · · · ( h x1 , ..., xn i.

This shows


dim k[x1 , ..., xn] ≥ n.

IV.1.3 Theorem. Assume that the field k is infinite. For any natural number n ∈ N, we
have dim(k[x1 , ..., xn]) = n.

The proof will be given on Page 121f.

IV.2 The Going-Up Theorem

We now investigate how dimension behaves under integral ring extensions.

IV.2.1 Lemma. Let R and S be integral domains and ϕ : R −→ S an integral ring

extension. Then, R is a field if and only if S is one.

Proof. We have already seen that R is a field, if S is one (Lemma III.3.9). Now, assume
that R is a field. Let y ∈ S \ {0} be a non-zero element. Choose n ≥ 1 minimal, such that
there exist elements a1 , ..., an ∈ R with

yn + a1 · yn−1 + · · · + an−1 · y + an = 0. (IV.1)

Accordingly, Z[x] should be two-dimensional. Here, you should consult your picture of the spectrum
3

of that ring (Exercise I.4.21). You will also find it in [21], Chapter II.1, Example H.

118
 IV.2. The Going-Up Theorem

Since we chose n to be minimal and S is an integral domain, we have an , 0. Equation

(IV.1) can be read as

an = −y · (yn−1 + a1 · yn−2 + · · · + an−1 ).

This shows that

1
y−1 = − · (yn−1 + a1 · yn−2 + · · · + an−1 )
an
is an element of S . 
IV.2.2 Lemma. Let R, S be rings, ϕ : R −→ S an integral ring extension, q ⊂ S a prime
ideal, and
p := q ∩ R.
Then, q is a maximal ideal in S if and only if p is a maximal ideal in R.4
Proof. The homomorphism ϕ induces the homomorphism

ϕ : R/p −→ S /q.

Here, R/p and S /q are integral domains, and ϕ is also an integral ring extension.
By Proposition I.4.1, ii), q is a maximal ideal in S if and only if S /q is a field. The
previous lemma says that S /q is a field if and only if R/p is a field. Using Proposition
I.4.1, ii), again, R/p is a field if and only if p is a maximal ideal in R. This completes the
proof. 
IV.2.3 Lemma. Let R, S be rings, ϕ : R −→ S an integral ring extension, and q ⊂ q′ ⊂ S
prime ideals. If
p := R ∩ q = R ∩ q′ ,
then
q = q′ .
Proof. The set T := R \ p is a multiplicatively closed subset of both R and S . By Exercise
II.3.8, the homomorphism ϕ induces a homomorphism

ϕT : RT −→ S T .

It is readily verified that ϕT is an integral ring extension, too. We use the commutative
diagram
ϕ
R /S (IV.2)

 ϕT 
RT / ST.
Define
n := qe ⊂ S T and n′ := q′e ⊂ S T .
Note that5
q ∩ T = ∅ = q′ ∩ T.
4
Compare Exercise III.3.8.
5
We use ϕ to view R as a subring of S and write intersections instead of preimages.

119
IV. Dimension Theory

Thus, Proposition II.3.6, v), shows

nc = q and n′c = q′ . (IV.3)
Next,
m := pe ⊂ RT .
By Corollary II.3.7, ii), it is the maximal ideal of RT . Using Proposition II.3.6, i), and the
maximality of m, our assumption gives
n ∩ RT = m = n′ ∩ RT .
We apply Lemma IV.2.2 to ϕT . It shows that n and n′ are maximal ideals in S T . Since
n ⊂ n′ , we infer n = n′ . So, Equation (IV.3) yields our claim. 
IV.2.4 Lying-over theorem. Let R, S be rings, ϕ : R −→ S an integral ring extension,
and p ⊂ R a prime ideal. Then, there is a prime ideal q ⊂ S with
p = q ∩ R.
IV.2.5 Remarks. i) The theorem says that the induced map
ϕ# : Spec(S ) −→ Spec(R)
is surjective.
ii) Note that, by Lemma IV.2.2, ϕ# maps maximal ideals to maximal ideals, i.e., ϕ# is
also surjective on the level of maximal ideals. Together with the correspondence between
the points of an algebraic set and the maximal ideals of its coordinate algebra provided by
the Nullstellensatz (Exercise III.3.2), we find a nice conceptual proof for Lemma III.4.8.
Proof of the lying-over theorem. As in the proof of Lemma IV.2.3, we localize at the mul-
tiplicatively closed subset T := R \ p and look at Diagram (IV.2). Let n ⊂ S T be a maxi-
mal ideal (Theorem I.4.4). Since ϕT is an integral ring extension, we conclude by Lemma
IV.2.2 that
m := n ∩ RT
is a maximal ideal of RT . Now, RT is a local ring (Corollary II.3.7, ii). So, m = pe and
mc = p. Finally,
q := nc ⊂ S
is a prime ideal. By construction, q ∩ R = p. 
IV.2.6 The going-up theorem. Let R, S be rings, ϕ : R −→ S an integral ring extension,
k > l natural numbers,
p0 ( · · · ( pk
prime ideals in R, and
q0 ( · · · ( ql
prime ideals in S , such that
∀i ∈ { 0, ..., l } : qi ∩ R = pi .
Then, one finds prime ideals
ql+1 ( · · · ( qk
in S with

120
 IV.2. The Going-Up Theorem

⋆ ql ( ql+1 ,

⋆ qi ∩ R = pi , i = l + 1, ..., k.

Proof. It clearly suffices to treat the case k = l + 1. This case follows from applying the
lying-over theorem to the integral ring extension

ϕl : R/pl −→ S /ql

and the prime ideal pk of R/pl (Lemma I.2.2). 

IV.2.7 Theorem. Let R, S be rings and ϕ : R −→ S an integral ring extension. Then, the
Krull dimensions of R and S are equal:

dim(R) = dim(S ).

Proof. The lying-over theorem and the going-up theorem clearly show that

dim(R) ≤ dim(S ).

Conversely, let
q0 ( · · · ( qk
be prime ideals in S and
pi := qi ∩ R.
By Lemma IV.2.3,
p0 ( · · · ( pk .
This implies
dim(R) ≥ dim(S )
and, therefore, the assertion of the theorem. 

Proof of Theorem IV.1.3. We proceed by induction on n. For n = 0, we are dealing with

the field k. It has Krull dimension 0 (Example IV.1.1).
Now, assume that the theorem is true for n, and let

p0 ( · · · ( pk

be prime ideals in the polynomial ring k[x1 , ..., xn+1 ]. Since k[x1 , ..., xn+1 ] is an integral
domain, we may assume p0 = h0i. Moreover, we may suppose k ≥ 1 (see Remark IV.1.2,
ii). Let f ∈ p1 \ {0}. Recall that k[x1 , ..., xn+1 ] is a factorial ring (Theorem I.6.4). Thus,
there are pairwise non-associated irreducible polynomials f1 , ..., f s ∈ k[x1 , ..., xn+1] and
positive integers k1 , ..., ks , such that

f ∼ f1k1 · · · · · f sks .

Since p1 is a prime ideal, there is an index i0 ∈ { 1, ..., s } with fi0 ∈ p1 . We get the prime
ideals
p0 ( h fi0 i ( p1 ( · · · ( pk .

121
IV. Dimension Theory

This shows that we may assume without loss of generality that p1 is a principal ideal, say,
p1 = hgi with g ∈ k[x1 , ..., xn+1] an irreducible polynomial. By the Noether normalization
theorem for hypersurfaces III.4.4, there is an integral ring extension

ϕ : k[x1 , ..., xn ] −→ k[x1 , ..., xn+1 ]/hgi

By Theorem IV.2.7 and the induction hypothesis,

dim k[x1 , ..., xn+1 ]/hgi = n.

This shows that k ≤ n + 1 and, consequently,

dim k[x1 , ..., xn+1 ] ≤ n + 1.

Together with Remark IV.1.2, ii), we obtain the desired equality. 

IV.3 The Transcendence Degree of a Field

Let k ⊂ K be a field extension. A subset S ⊂ K is said to be algebraically independent
over k, if every finite subset S ′ ⊂ S is algebraically independent over k (in the sense of
Page 8). The algebraically independent subsets of K are partially ordered by inclusion,
and ∅ is an algebraically independent subset. Zorn’s lemma I.4.7 gives:
IV.3.1 Lemma. The field K contains maximal algebraically independent subsets.
A maximal algebraically independent subset S ⊂ K is called a transcendence basis
for K over k. We say that K has finite transcendence degree over k, if there exists a finite
transcendence basis for K over k.
IV.3.2 Remark. The field extension k ⊂ K is algebraic if and only if ∅ is a transcendence
basis of K over k.
IV.3.3 Proposition. Assume that K has finite transcendence degree over k. Let S ⊂ K be
a transcendence basis and m := #S . Then, every transcendence basis of K over k has m
elements.
If K has finite transcendence degree over k and S ⊂ K is a transcendence basis, we
call
trdegk (K) := #S
the transcendence degree of K over k. By Proposition IV.3.3, this is well-defined. If K
does not have finite transcendence degree, we say that it has infinite transcendence degree
and write
trdegk (K) := ∞.
Proof of Proposition IV.3.3. We may assume that S has the minimal number of elements
among all transcendence bases for K. Write S = { s1, ..., sm } and let T = { t1, ..., tn } ⊂ K
be an algebraically independent subset with m ≤ n. We will then prove that there is a
subset T ′ ⊂ T with m elements, such that K is algebraic over k(T ′ ). This implies that
T \ T ′ is empty and, thus, m = n, and that T is a transcendence basis. The strategy is to
exchange elements of S by elements of T .

122
 IV.4. The Dimension of an Algebraic Variety

We set

r := max ̺ ∈ { 0, ..., m } | ∃i̺+1 , ..., im ∈ { 1, ..., m } :

K is algebraic over k(t1 , ..., t̺, si̺+1 , ..., sim ) .

We will lead the assumption r < m to a contradiction. After renumbering, we may as-
sume { ir+1, ..., im } = { r + 1, ..., m }. There exists a non-zero polynomial f ∈ k[x1 , ...., xm+1]
with
f (t1, ..., tr , sr+1 , ..., sm, tr+1) = 0 (in K).
Since t1 , ..., tr+1 are algebraically independent over k, there is an index i0 ∈ { r + 1, ..., m },
such that xi0 occurs in f , i.e., f < k[x1 , ..., xr , xr+1 , ..., xi0 −1 , xi0 +1 , ..., xm+1]. After renum-
bering, if necessary, we may assume that r + 1 is such an index. This means that sr+1
is algebraic over k(t1 , ..., tr+1, sr+2, ...., sm ) and, therefore, K is an algebraic extension of
k(t1 , ..., tr+1, sr+2, ...., sm ), contradicting the definition of r. 
IV.3.4 Example. Let k be a field. For n ∈ N, it follows readily that



trdegk k(x1 , ..., xn) = n, k(x1 , ..., xn) := Q k[x1 , ..., xn ] .

IV.4 The Dimension of an Algebraic Variety

IV.4.1 Theorem. Let k be an infinite field, n ≥ 1, and p ⊂ k[x1 , ..., xn] a prime ideal.
Then,


dim(k[x1 , ..., xn ]/p) = trdegk Q(k[x1 , ..., xn ]/p) .
Proof. By Noether normalization III.4.7, there are a natural number 0 ≤ m ≤ n and a
finite ring extension
ϕ : k[x1 , ..., xm ] −→ k[x1 , ..., xn ]/p.
The homomorphism ϕ induces a finite field extension (see Page 25)


e
ϕ : k(x1 , ..., xm ) = Q k[x1 , ..., xm ] −→ Q(k[x1 , ..., xn ]/p).

Since every finite field extension is algebraic ([8], Satz III.1.6.2, 1), we have



trdegk Q(k[x1 , ..., xn]/p) = trdegk k(x1 , ..., xm ) = m.

On the other hand, Theorem IV.2.7 gives

dim(k[x1 , ..., xn]/p) = dim k[x1 , ..., xm ] = m.

This settles the theorem. 

The Going-Down Theorem

Let R, T be rings, ψ : R −→ T an injective homomorphism, and I ⊂ R an ideal. We say
that α ∈ T is integral over I, if there exist a positive natural number n ≥ 1 and elements
a1 , ..., an ∈ I with
αn + a1 · αn−1 + · · · + an−1 · α + an = 0. (IV.4)

123
IV. Dimension Theory

IV.4.2 Lemma. Let R, S be integral domains, ϕ : R −→ S an integral ring extension,

e
ϕ : K −→ L the corresponding algebraic field extension, K := Q(R), L := Q(S ), and
I ⊂ R an ideal. √
i) An element α ∈ S is integral over I if and only if α ∈ I e .
ii) Assume that R is normal. Let α ∈ L be an element √ which is integral over I. Then,
the minimal polynomial µα ∈ K[x] of α has coefficients in I.

Proof. i) If √ α ∈ S is integral over I, then an integrality equation like (IV.4) shows αn ∈ I e ,

so that α ∈ I e . √
Conversely, suppose that α ∈ I e and pick s ≥ 1 with αs ∈ I e . There are m ≥ 1,
r1 , ..., rm ∈ I, and a1 , ..., am ∈ S with

α s = a1 · r 1 + · · · + am · r m .

Now, let M ⊂ S be the finitely generated R-submodule generated by a1 , ..., am , and set

ϕ : M −→ M
x 7−→ αs · x.

This is an endomorphism with ϕ(M) ⊂ I · M. As in the proof of “iv)=⇒i)” of Proposition

III.2.4, we construct a monic polynomial p ∈ R[x] with coefficients in I, such that p(αs )
annihilates M. If α , 0, then M , {0}. Since S is an integral domain, this implies that
p(αs ) = 0. Therefore, αs and α are integral over I.
ii) By Equation (IV.4) and the definition of a minimal polynomial, the minimal poly-
nomial µα of α divides p(x) := xn + a1 · xn−1 + · · · + an−1 · x + an . Therefore, the roots
α1 , ..., αr of µα are integral over I. The arguments in the proof of Lemma √ III.5.11, ii),
show that the coefficients of µα belong to R, and Part i) yields that they lie in I. 

IV.4.3 The going-down theorem. Let R, S be integral domains, ϕ : R −→ S an integral

ring extension, k > l natural numbers,

p0 ) · · · ) pk

prime ideals in R, and

q0 ) · · · ) ql
prime ideals in S , such that

∀i ∈ { 0, ..., l } : qi ∩ R = pi .

If R is normal, one finds prime ideals

ql+1 ) · · · ) qk

in S with

⋆ ql ) ql+1 ,

⋆ qi ∩ R = pi , i = l + 1, ..., k.

124
 IV.4. The Dimension of an Algebraic Variety

Proof. Again, it suffices to treat the case k = l + 1. We look at the localization6 S ql

and the homomorphism R −→ S ql . It is sufficient to show that, with respect to this
homomorphism,
pec
l+1 = pl+1 .

In fact, by Corollary II.3.7, pel+1 , the extension being taken with respect to ϕ, will do the
trick.
Let β ∈ pel+1 . Then, we may write β = α/s with α ∈ S · pl+1 and s ∈ S \ ql . Since S
is integral over R, it is clear that α is integral over pl+1 , and Lemma IV.4.2, ii), shows that
we may find n ≥ 1 and a1 , ..., an ∈ pl+1 with

αn + a1 · αn−1 + · · · + an−1 · α + an = 0. (IV.5)

In fact, we may assume that xn + a1 · xn−1 + · · · + an−1 · x + an is the minimal polynomial of

α. Now, assume that β ∈ pec n
l+1 \ {0}. Recall that s = α/β. We divide (IV.5) by β and find
the integrality equation
ai
sn + b1 · sn−1 + · · · + bn−1 · s + bn = 0, bi := , i = 1, ..., n, (IV.6)
βi

for s over K. We claim that q(x) := xn + b1 · xn−1 + · · · + bn−1 · x + bn is the minimal

polynomial of s. Indeed, if there were a polynomial r(x) ∈ K[X] of degree, say, l < n,
with r(s) = 0, then (βl · r)(α) = 0 and βl · r ∈ K[x]. This contradicts the choice of n. By
Lemma III.5.11, q(x) is a polynomial in R[x], i.e., bi ∈ R, and

βi · bi = ai ∈ pl+1 , i = 1, ..., n.

If β < pl+1 , then bi ∈ pl+1 , i = 1, ..., n. Equation (IV.6) proves that sn ∈ S · pl+1 ⊂ ql , and
this is impossible. So, after all, β ∈ pl+1 . 

A Refined Version of Noether Normalization

IV.4.4 Theorem. Let k be an infinite field, R a finitely generated k-algebra, n := dim(R),
and
I0 ( · · · ( Il
a chain of ideals in R. Then, there exist a finite ring extension

ϕ : k[t1 , ..., tn] −→ R

and natural numbers

0 ≤ k0 ≤ · · · ≤ kl ≤ n
with
ϕ−1 (I j ) = h t1 , ..., tk j i, j = 0, ..., l.

6
Corollary II.3.7 explains why this is natural.

125
IV. Dimension Theory

Proof. Step 1. We first reduce to the case when R is a polynomial ring. By definition,
there are a natural number m ≥ 1 and an ideal J ⊂ k[x1 , ..., xm ], such that

R = k[x1 , ..., xm ]/J.

By Theorem III.4.7, there is an integral ring extension

ψ : k[x1 , ..., xn] −→ R.

We get the chain

J0 ⊂ · · · ⊂ Jl , Jk := ψ−1 (Ik ), k = 0, ..., l.
Since the composition of finite ring extensions is a finite ring extension (Lemma III.2.6),
we may assume without loss of generality that R = k[x1 , ..., xn].
Step 2. We first consider the case l = 0. Here, we will use several inductions. To start
with, we look at the case I = h f i with f , 0.
Claim. The element f is transcendent over k.
According to the proof of Theorem III.4.4, we may assume that there are polynomials
g1 , ..., gr ∈ k[x2 , ..., xn], such that

f = xr1 + g1 · xr−1
1 + · · · + gr−1 · x1 + gr . (IV.7)

For a non-zero polynomial p(t) = tl + a1 · tl−1 + · · · + al−1 · t + al ∈ k[t], we have

p( f ) = xr·l
1 + terms involving x2 , ..., xn + lower order terms.

This is not zero. X

A similar argument shows that the elements f, x2, ..., xn are algebraically independent
over k (compare Section IV.3). The inclusion

k[ f, x2, ..., xn ] ⊂ k[x1 , x2 , ..., xn ]

is a finite ring extension,

xr1 + g1 · xr−1
1 + · · · + gr−1 · x1 + gr − f = 0

being an integrality equation for x1 . Together with the isomorphism

k[t1 , t2, ..., tn] −→ k[ f, x2 , ..., xn]

t1 7−→ f
ti 7−→ xi , i = 2, ..., n,

we obtain the integral ring extension ϑ with

ϑ−1 h f i = ht1 i.

Now, we are ready to prove the case l = 0 by induction on n. The case n = 1 is clear
by the foregoing discussion.

126
 IV.4. The Dimension of an Algebraic Variety

Next, assume that I ⊂ k[x1 , x2 , ..., xn ] is a non-trivial ideal. We let f ∈ I \ {0}. Suppose
that it has the form given in (IV.7) and construct an integral ring extension

ϑ : k[u1 , ..., un] −→ k[x1 , ..., xn]

with ϑ−1 (h f i) = hu1 i. So, we have an induced finite ring extension

ϑ : k[u2 , ..., un] = k[u1 , ..., un]/hu1 i −→ k[x1 , ..., xn]/h f i.

−1
Define I ′ as the image of I in the ring k[x1 , ..., xn ]/h f i and I ′′ := ϑ (I ′ ). By induction,
there are a natural number 2 ≤ k0 ≤ n and a finite ring extension

ψ : k[t2 , ..., tn] −→ k[u2 , ..., un]

with
ψ−1 (I ′′ ) = h t2 , ..., tk0 i.
Define ϕ as the composition of ϑ and the homomorphism

k[t1 , ..., tn] −→ k[u1 , ..., un]

t1 7−→ u1
ti 7−→ ψ(ti ), i = 2, ..., n.

It is an integral ring extension with

ϕ−1 (I) = h t1 , ..., tk0 i.

Step 3. Now, we can prove the general statement by induction on l. The case l = 0
has been settled in the previous step. Let I0 ⊂ · · · ⊂ Il+1 be a chain of ideals in k[x1 , ..., xn ].
By induction hypothesis, there are natural numbers 0 ≤ k0 ≤ · · · ≤ kl ≤ n and a finite ring
extension
ϑ : k[u1 , ..., un] −→ k[x1 , ..., xn]
with
ϑ−1 (I j ) = h u1 , ..., uk j i, j = 0, ..., l.
It induces a finite ring extension

ϑ : k[ukl +1 , ..., un] = k[u1 , ..., un]/h u1 , ..., ukl i −→ k[x1 , ..., xn]/Il .

We conclude as in Step 2. 
IV.4.5 Remark. Assume in the statement of Theorem IV.4.4 that

p0 ( · · · ( pl

is a chain of prime ideals. Then, by Lemma IV.2.3,

ϕ−1 (pi−1 ) ( ϕ−1 (pi ), i = 1, ..., l,

i.e.,
0 ≤ k0 < · · · < kl ≤ n,
and k0 = 0 holds if and only if p0 = h0i.

127
IV. Dimension Theory

IV.4.6 The chain theorem. Let k be an infinite field, n ≥ 1 a natural number, R a finitely
generated k-algebra and an integral domain, and n = dim(R). Every finite chain

p0 ( · · · ( pl (IV.8)

of prime ideals in R can be completed to a finite chain of length n, i.e., there are prime
ideals
p′0 ( · · · ( p′n
such that
{ p0 , ..., pl } ⊂ { p′0 , ..., p′n }.
Proof. We apply Theorem IV.4.4 to (IV.8). Let 0 ≤ k0 < · · · < kl ≤ n be natural numbers
and
ϕ : k[t1 , ..., tn] −→ R
be an integral ring extension with

ϕ−1 (pi ) = h t1 , ..., tki i, i = 0, ..., l.

Note that we may assume kl = n, because otherwise, by the going-up theorem IV.2.6,
there is an ideal pl ( pl+1 with ϕ−1 (pl+1 ) = h t1 , ..., tn i. This assumption and Remark IV.4.5
imply that, if l < n, there must be an index i0 ∈ { 1, ..., l } with

ki0 − ki0 −1 > 1.

We pass to the finite ring extension

ϕ : k[tκ+1 , ..., tn] := k[t1 , ..., tn]/h t1 , ..., tκ i −→ R/pi0 −1 κ := ki0 −1 ,

and look at the image p of pi0 in R/pi0 −1 . By construction,

ϕ−1 (p) = h tκ+1 , ..., tki0 i.

The assumption ki0 > κ + 1 implies

htκ+1 i ( h tκ+1 , ..., tki0 i

Since k[tκ+1 , ..., tn] is a normal ring (Theorem I.6.4 and Lemma III.5.1), we may apply the
going-down theorem IV.4.3. So, there exists an ideal

q(p

with
ϕ−1 (q) = htκ+1 i.
The preimage q of q in k[x1 , ..., xn] satisfies

pi0 −1 ( q ( pi0 .

So, we have increased the length of the given chain by one. Iterating the argument, if
necessary, we arrive at a chain of length n. 

128
 IV.4. The Dimension of an Algebraic Variety

Heights
Let R be a ring and p ⊂ R a prime ideal. The height of p is

ht(p) = sup k | ∃ prime ideals p0 ( · · · ( pk = p .

IV.4.7 Remark. By Corollary II.3.7, i), we have

ht(p) = dim(Rp ).

IV.4.8 Corollary to the chain theorem. Let k be an infinite field, n ≥ 1 a natural number,
q ⊂ k[x1 , ..., xn] a prime ideal, and

R := k[x1 , ..., xn]/q.

Then, for every prime ideal p ⊂ R, we have

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

In particular, if m ⊂ R is a maximal ideal, then

dim(R) = dim(Rm ).

We now prove a variant of Krull’s principal ideal theorem (see Theorem IV.5.6) for
k[x1 , ..., xn]. If Z ⊂ Ank is an algebraic set, we define


dim(Z) := dim k[Z]

to be the dimension of Z.

IV.4.9 Proposition. Let k be an algebraically closed field, n ≥ 1 a natural number, I ⊂

k[x1 , ..., xn] a radical ideal, and

V(I) = V1 ∪ · · · ∪ V s

the decomposition into irreducible components. Then, the following conditions are equiv-
alent:
i) The ideal I is a principal ideal.
ii) For i = 1, ..., s, one has dim(Vi ) = n − 1.

Proof. One reduces immediately to s = 1, i.e., to the case when I is a prime ideal.
“i)=⇒ii)”. This is again an application of Noether normalization for hypersurfaces
III.4.4.
“ii)=⇒i)”. Since I is a prime ideal, we may apply Corollary IV.4.8. The assumption is
equivalent to ht(I) = 1, i.e., h0i ( I is the maximal chain ending in I. As explained in the
proof of Theorem IV.1.3, we may always insert a non-zero principal ideal into that chain.
Therefore, I has to be a principal ideal. 

129
IV. Dimension Theory

IV.4.10 Exercise (Dimension). Let k be a field, R a finitely generated k-algebra, and

p1 , ..., ps the minimal prime ideals of R (see Theorem II.4.28, ii).
i) Assume that k is algebraically closed and that R is reduced. Write R = k[x1 , ..., xn]/I
for a suitable natural number n and a suitable radical ideal I ⊂ k[x1 , ..., xn ], and set
Z := V(I) ⊂ Ank . Define e pi ⊂ k[x1 , ..., xn ] as the preimage of pi under the projection
k[x1 , ..., xn] −→ k[x1 , ..., xn]/I, i = 1, ..., n. What is the geometric significance of the
varieties V(e pi ) ⊂ Ank , i = 1, ..., s?
ii) Set Ri := R/pi , i = 1, ..., s. Prove that
n
 o
dim(R) = max trdegk Q(Ri )  i = 1, ..., s .

iii) Let S ⊂ R be a subalgebra. Prove that dim(S ) ≤ dim(R).

Hint. If p1 , ..., p s ∈ R are algebraically independent over k, then k[p1 , ..., p s ] ⊂ R is an
integral domain.

IV.5 Krull’s Principal Ideal Theorem

Let R be a ring and M an R-module. The number

l(M) := sup l ∈ N | ∃ submodules M0 ( · · · ( Ml ∈ N ∪ {∞}

is called the length of M.

IV.5.1 Remarks. i) Let R be a ring. We may view R as an R-module. Then, l(R) is not the
same as dim(R). In fact, in the definition of l(R), all ideals are used, not just the prime
ideals.
ii) As Part i) already suggests, an R-module M, even a noetherian one, will usually
have infinite length. In other words, modules of finite length are quite special.
IV.5.2 Example. Let k be a field and V a k-vector space. Then, l(V) = dimk (V).

IV.5.3 Lemma. Let R be a ring, M an R-module, and M0 ( · · · ( Ml a maximal chain of

submodules. Then, l(M) = l.

Proof. Note that the maximality of the chain implies M0 = {0} and Ml = M. We proceed
by induction on l. The case l = 0 is trivial.
l − 1 −→ l. Let l′ ∈ N and M0′ ( · · · ( Ml′′ be a chain of submodules of M with
′
M0 = {0}. Set
n  o
r := max t ∈ { 0, ..., l′ }  Mt′ ⊂ Ml−1 .
Then,
Ml−1 + Mi′ = M, i = r + 1, ..., l′.
Observe that
′ ′ ′
Mi+1 = Mi+1 ∩ M = Mi+1 ∩ (Ml−1 + Mi′ ) = Mi′ + (Ml−1 ∩ Mi+1
′
), i = r + 1, ..., l′ − 1.

Since Mi′ ( Mi+1

′
, we have

(Ml−1 ∩ Mi′ ) ( (Ml−1 ∩ Mi+1

′
), i = r + 1, ..., l′ − 1.

130
 IV.5. Krull’s Principal Ideal Theorem

We form

M0′ ( · · · ( Mr′ ⊂ (Ml−1 ∩ Mr+1

′ ′
) ( (Ml−1 ∩ Mr+2 ) ( · · · ( (Ml−1 ∩ Ml′′ ).

This is a chain of length at least l′ − 1. Since M0 ( · · · ( Ml−1 is a maximal chain in Ml−1 ,

the induction hypothesis gives l′ − 1 ≤ l − 1, i.e., l′ ≤ l, and we are done. 
IV.5.4 Lemma. Let R be a ring, M an R-module, and N ⊂ M a submodule. Then,

l(M) = l(N) + l(M/N).

Proof. If N or M/N has infinite length, then it is easy to check that M has infinite length,
too. So, assume that N and M/N have finite length. Then, we have maximal sequences

{0} = N0 ( · · · ( Nl(N) = N and {0} = M 0 ( · · · ( M l(M/N) = M/N

of submodules of N and M/N, respectively. With the preimage Mi of M i under the pro-
jection M −→ M/N, i = 0, ..., l(M/N), we get the maximal chain

{0} = N0 ( · · · ( Nl(N) = N = M0 ( · · · ( Ml(M/N)

of submodules of M. Lemma IV.5.3 shows l(M) = l(N) + l(M/N). 

IV.5.5 Lemma. Let R be a noetherian local ring and M a finitely generated R-module.
Then, the following holds:


dim R/Ann(M) = 0 =⇒ l(M) < ∞.

Proof. We suppose M , 0.
Step 1. Let m be the maximal ideal of R and k := R/m. Since R and M are noetherian,
m is finitely generated, and, for every natural number s ∈ N, (m s · M)/(m s+1 · M) is a finite
dimensional k-vector space. Note, for s ∈ N,



l (m s · M)/(m s+1 · M) = dimk (m s · M)/(m s+1 · M)

and

X
s−1
s

l M/(m · M) = dimk (mi · M)/(mi+1 · M) .
i=0

The second formula results from Lemma IV.5.4.

Step 2. Let R := R/Ann(M) and m ⊂ R the image of m under the surjection R −→ R.
Then, R is a local ring with maximal ideal m. The assumption dim(R) = 0 implies that m
is also a minimal prime ideal of R. Therefore, it equals the radical of R (see Proposition
s
I.7.2). Since m is finitely generated, there is a power s ≥ 1 with m = h0i (compare
Example I.8.8, ii), i.e.,

m s ⊂ Ann(M) and m s · M = h0i.

Together with Step 1, we find our claim. 

Let R be a noetherian ring and x ∈ R \ R⋆ . A minimal prime ideal containing x is a
prime ideal x ∈ p ⊂ R whose image p ⊂ R/hxi is a minimal prime ideal.

131
IV. Dimension Theory

IV.5.6 Krull’s principal ideal theorem. Let R be a noetherian ring, x ∈ R \ R⋆ and p ⊂ R

a minimal prime ideal containing x. Then, ht(p) ≤ 1.

Proof. We look at the localization Rp . In view of Remark IV.4.7, we have to show that
dim(Rp ) ≤ 1. So, we have reduced to the case that R is a local ring and the minimal prime
ideal p containing x is the maximal ideal of R.
Suppose we have a chain r ( q ( p of prime ideals in R. We may pass to the quotient
R/r and, therefore, assume that r = h0i and that R is an integral domain.
Now, we work with the localization R −→ Rq . We form the descending chain

qnec + hxi ⊂ R, n ∈ N, (IV.9)

of ideals in R. We claim that this sequence becomes stationary. It gives rise to a de-
scending chain (pn )n∈N in R := R/hxi. Note that our construction implies that R is zero-
dimensional and, therefore, has finite length as R-module, by Lemma IV.5.5. For this
reason, the sequence (pn )n∈N becomes stationary. By Lemma I.2.2, this is also true for the
sequence (IV.9).
Pick a natural number n0 with
ec
qn0 ec + hxi = qn0 +1 + hxi.

In particular,
ec
qn0 ec ⊂ qn0 +1 + hxi.
ec
Let r ∈ qn0 ec . There are elements s ∈ qn0 +1 and a ∈ R with

r = s + a · x.

We infer
a · x = r − s.
Since x < q, we find, using Property I.8.24, ii),
r−s
a= ∈ qn0 ece ∩ R = qn0 e ∩ R = qn0 ec .
x
Our discussion shows
ec
qn0 ec = qn0 +1 + p · qn0 ec .
Using Exercise III.1.35, we find
ec
qn0 ec = qn0 +1 .
This yields
ece e
qn0 e = qn0 ece = qn0 +1 = qn0 +1 = qn0 e · qe
in Rq . Note that qe is the maximal ideal of Rq (Corollary II.4.7). We apply the Nakayama
lemma III.1.31 once more. It gives

qn0 e = h0i.

Since R is an integral domain, this implies that qn0 = h0i (see Page 57) and, thus, q = h0i.
This contradicts our assumption on q. 

132
 IV.6. Embedding Dimension

IV.5.7 Exercise (Krull’s Höhensatz). Let R be a noetherian ring and I ( R a proper ideal.
We say that I ⊂ p is a minimal prime ideal containing I, if the image p of p in R/I is a
minimal prime ideal of R/I.
Assume that there are s ∈ N and a1 , ..., a s ∈ R with h a1 , ..., a s i = I. Prove that

ht(p) ≤ s

for every minimal prime ideal p ⊂ R containing I. What is the geometric interpretation of
this statement?

IV.6 Embedding Dimension

Let R be a local notherian ring and m ⊂ R its maximal ideal. Then,

m/m2

is a vector space over the field R/m.

IV.6.1 Remark. The ideal m is finitely generated, because R is a noetherian ring. Let s ≥ 1
and a1 , ..., a s ∈ m be such that
h a1 , ..., a s i = m.
Then, the classes [a1 ], ..., [a s ] ∈ m/m2 generate m/m2 as a vector space over R/m, i.e.,

dimR/m (m/m2 ) ≤ s.

The natural number

edim(R) := dimR/m (m/m2 )
is called the embedding or tangential dimension of R.

IV.6.2 Proposition. Let a1 , ..., a s ∈ m be elements, such that the classes [a1 ], ..., [a s ] ∈
m/m2 generate m/m2 as a vector space over R/m. Then,

h a1 , ..., a s i = m.

In particular, the embedding dimension of R equals the minimal number of generators for
the maximal ideal m.

Proof. We look at the homomorphism

ϕ : R⊕s −→ m
(r1 , ..., r s ) 7−→ r1 · a1 + · · · + r s · a s

of R-modules. The assumption states that the induced homomorphism

ϕ : (R/m)⊕s −→ m/m2

is surjective. By Lemma III.1.33, ii), ϕ is surjective. 

133
IV. Dimension Theory

A noetherian local ring R is regular, if

dim(R) = edim(R).

IV.6.3 Remarks. i) Let R, S be not necessarily local rings and ϕ : R −→ S a surjective

homomorphism. If n ⊂ S is a maximal ideal in S , then m = ϕ−1 (n) is a maximal ideal in
R, and we have an induced surjection

ϕ : m/m2 −→ n/n2 .

ii) Let R be a not necessarily local ring, m ⊂ R a maximal ideal, and ϕ : R −→ Rm

the localization. Then, Rm is a local ring with maximal ideal n = me . Then, by Corollary
II.3.7, i),
nc = mec = m and (m2 )ec = m2 .
It is also evident (Proposition II.3.6) that

(m2 )e = n2 .

Thus, we have an induced injective homomorphism

ϕ : m/m2 −→ n/n2 .

It is also clear that ϕ is surjective and, therefore, an isomorphism.

IV.6.4 Examples. i) Let k be an algebraically closed field, n ≥ 1 a natural number, R :=
k[x1 , ..., xn] and m ⊂ R a maximal ideal. Then,

dimk (m/m2 ) = n.

This is obvious for m = h x1 , ..., xn i. In general, there is a point (a1 , ..., an) ∈ Ank with

m = h x1 − a1 , ..., xn − an i,

and

ϕ : k[x1 , ..., xn ] −→ k[x1 , ..., xn ]

xi 7−→ xi − ai , i = 1, ..., n,

is an automorphism which maps h x1 , ..., xn i to m. Corollary IV.4.8 and Remark IV.6.3,

ii), show that Rm is a regular local ring of dimension n.
ii) Let k be an algebraically closed field, n ≥ 1 a natural number, I ⊂ k[x1 , ..., xn] an
ideal, R := k[x1 , ..., xn ]/I, and m ⊂ R a maximal ideal. Then,

edim(Rm ) ≤ n.

In fact, let
π : k[x1 , ..., xn] −→ R
be the canonical surjection. There is a point (a1 , ..., an) ∈ Ank with


π h x1 − a1 , ..., xn − an i = m.

134
 IV.7. Singular Points of Algebraic Varieties

By Remark IV.6.1,
dim(m/m2 ) ≤ n.
The contention follows now from Remark IV.6.3, ii). This gives (a partial) explanation for
the name embedding dimension: Given an affine variety X with coordinate algebra R :=
k[X], k an algebraically closed field. Then, the maximum7 of the embedding dimensions
of the localizations of R at all maximal ideals is a lower bound for the minimal number n,
such that X can be embedded into Ank .

IV.7 Singular Points of Algebraic Varieties

We now work over an algebraically closed field k. Fix a natural number n ≥ 1, and a
radical ideal I ⊂ k[x1 , ..., xn]. Define

Z := V(I) ⊂ Ank

as the corresponding algebraic set and

R := k[x1 , ..., xn ]/I

as its coordinate algebra. According to Exercise III.3.2, the assignment


a 7−→ ma := f ∈ R | f (a) = 0

establishes a bijection between V(I) and the set of maximal ideals of the k-algebra R.
We say that a ∈ Z is a regular point of Z or that Z is non-singular at a, if Rma is a
regular local ring. Otherwise, we say that a is a singular point of Z or that Z is singular
at a. The k-vector space
T a Z := Homk (ma /m2a , k)
is the Zariski tangent space of Z at a. Observe that it is defined intrinsically in terms of
the coordinate algebra of Z, i.e., without reference to the embedding Z ⊂ Ank .
A derivation of R at a is a k-linear map

t : R −→ k

which satisfies the Leibniz8 rule

∀ f, g ∈ R : t( f · g) = t( f ) · g(a) + f (a) · t(g).

The space

Dera (R) := t ∈ Homk (R, k) | t is a derivation at a
of all derivations of R at a is a sub vector space of Homk (R, k).
Now, suppose t : R −→ k is a derivation of R at a. We find

t(1) = t(1 · 1) = 2 · t(1).

7
The above discussion shows that this maximum exists. In fact, it is bounded by m, if X ⊂ Am
k.
8
Gottfried Wilhelm von Leibniz (1646 - 1716), German mathematician and philosopher.

135
IV. Dimension Theory

This implies t(1) = 0 and, by k-linearity,

∀λ ∈ k : t(λ) = 0. (IV.10)

Note that

F : R −→ k ⊕ ma


f 7−→ f (a), f − f (a)

is an isomorphism of k-vector spaces. By (IV.10), t is determined by its restriction to ma .

Next, let f, g ∈ ma . Then,

t( f · g) = t( f ) · g(a) + f (a) · t(g) = 0.

Thus, t|ma factorizes over a k-linear map

t : ma /m2a −→ k.

IV.7.1 Proposition. The map

H : Dera (R) −→ T a Z = Homk (ma /m2a , k)

t 7−→ t

is an isomorphism of k-vector spaces.

Proof. The injectivity and k-linearity are clear from the definition and the above discus-
sion. Now, suppose we are given l ∈ Homk (ma /m2a , k). We set

t : R −→ k
 
f 7−→ l f − f (a) .

Now, let f, g ∈ R. Then,

f − f (a) · g − g(a) ∈ m2a .
We now compute



t( f · g) = t f · g − f − f (a) · g − g(a)


= t f · g(a) + f (a) · g − f (a) · g(a)
= t( f ) · g(a) + f (a) · t(g) − f (a) · g(a) · t(1)
= t( f ) · g(a) + f (a) · t(g).

So, t is a derivation of R at a with t = l, i.e., H is also surjective. 

IV.7.2 Example (Taylor9 expansion). Let a = (a1 , ..., an) ∈ Ank be a point and m := h x1 −
a1 , ..., xn − an i ⊂ k[x1 , ..., xn ] the corresponding maximal ideal. The homomorphism

τa : k[x1 , ..., xn ] −→ k[x1 , ..., xn]

xi 7−→ xi − ai , i = 1, ..., n,
9
Brook Taylor (1685 - 1731), English mathematician.

136
 IV.7. Singular Points of Algebraic Varieties

is a k-linear isomorphism. So, for a polynomial f ∈ k[x1 , ..., xn ], there exists a polynomial
T a f ∈ k[x1 , ..., xn] with

f (x1 , ..., xn ) = T a f (x1 − a1 , ..., xn − an ).

We call T a f the Taylor expansion of f at a. We have

X
n
T a f (x1 − a1 , ..., xn − an ) = f (a) + δ f,i (a) · (xi − ai ) + higher order terms.
i=1

The vector space ma /m2a has the basis [x1 − a1 ], ..., [xn − an ]. Let l1 , ..., ln be the dual basis
of Homk (ma /m2a , k). Then,

∀ f ∈ k[x1 , ..., xn] : li ( f ) = δ f,i (a).

On the other hand, we may define the partial derivatives

∂f
∈ k[x1 , ..., xn], i = 1, ..., n,
∂xi
by formally applying the rules we know from analysis ([28], Kapitel 5). Then, one finds
∂f
δ f,i (a) = (a), i = 1, ..., n. (IV.11)
∂xi
In fact, for i ∈ { 1, ..., n }, the operators δ·,i (a) and (∂/∂xi )(a) are both k-linear and satisfy
the Leibniz rule, i.e., for f, g ∈ k[x1 , ..., xn ],
∂( f · g) ∂f ∂g
δ f ·g,i (a) = δ f,i (a) · g(a) + f (a) · δg,i (a) and (a) = (a) · g(a) + f (a) · (a).
∂xi ∂xi ∂xi
For this reason, it is enough to check (IV.11) for constant polynomials and the polynomials
x1 , ..., xn. For all of these, (IV.11) is trivial.
IV.7.3 Remark. In principle, we could try to compute the Taylor expansion also up to
higher orders. Here, we have to be a little cautious: The usual Taylor formula ([28], Satz
7.2.2, [4], Kapitel I, Satz (2.9)) requires division by natural numbers of the form l1 !·· · ··ln!
with l1 + · · · + ln = l in the term of order l. So, the characteristic of our base field k must
be larger than l, if we would like to have the usual formula for the term of order l in the
Taylor expansion. If we allow, as we do, fields of characteristic 2, it makes only sense to
look at the linear terms of the Taylor expansion.
Now, let g1 , ..., g s ∈ k[x1 , ..., xn] be such that I = h g1 , ..., g s i. For every derivation
D : R −→ k at a, the composition
π D
e : k[x1 , ..., xn] −→
D R −→ k

is a derivation of k[x1 , ..., xn ] at a.

IV.7.4 Lemma. A derivation D e : k[x1 , ..., xn ] −→ k at a factorizes over a derivation
D : R −→ k of R at a if and only if

∀i ∈ { 1, ..., s } : e i ) = 0.
D(g

137
IV. Dimension Theory

Proof. The implication “=⇒” is trivial. For the converse, we have to show that D e vanishes
on all elements of I. An element of I has the form f1 ·g1 +· · ·+ f s ·g s for suitable polynomials
f1 , ..., f s ∈ k[x1 , ..., xn ]. Since a derivation is a k-linear, it suffices to look at elements of
the shape f · gi , f ∈ k[x1 , ..., xn], i ∈ { 1, ..., s }. For such an element we have
e f · gi ) = D(
D( e f ) · gi (a) + f (a) · D(g
e i ) = 0.

Indeed, the first summand vanishes, because a ∈ V(I) ⊂ V(gi ) and the second one by
assumption. 
Example IV.7.2 and Lemma IV.7.4 give the following description of the tangent space
of a variety at a point.
IV.7.5 Proposition. Let I = h g1 , ..., g s i ⊂ k[x1 , ..., xn ] be an ideal, R := k[x1 , ..., xn ]/I,
Z := V(I), and a = (a1 , ..., an) ∈ Z. Then,
( X n )
n  ∂gi
T a Z  (t1 , ..., tn) ∈ k  tj · (a) = 0, i = 1, ..., s .
j=1
∂x j

In the situation of Proposition IV.7.5, set G := (g1 , ..., g s ). The matrix

!
∂gi

JG := ∈ Mat s, n, k[x1 , ..., xn]
∂x j i=1,...,s
j=1,...,n

is the Jacobian matrix of the ordered tuple G of polynomials. Putting everything together,
we find the following result.
IV.7.6 Proposition (Jacobian criterion). Let I = h g1, ..., g s i ⊂ k[x1 , ..., xn] be an ideal,
R := k[x1 , ..., xn]/I, Z := V(I), and a = (a1 , ..., an) ∈ Z. Then,
!

∂gi
edim(Rma ) = n − rk JG (a) , JG (a) := (a) ∈ Mat(s, n, k).
∂x j i=1,...,s
j=1,...,n

Moreover, if all the irreducible components of Z have the same dimension, Rma is a regular
local ring if and only if


rk JG (a) = n − dim(Z).
The reader should compare this with the corresponding result in real analysis ([28],
Abschnitt 11.2). Given a set of elements g1 , ..., g s ∈ k[x1 , ..., xn], it is not obvious what the
dimension of V(g1 , ..., g s ) is. However, if s = 1, we know from Proposition IV.4.9 that it
is n − 1. In this case, the Jacobian criterion is easy to check. For g ∈ k[x1 , ..., xn ] \ {0},
the hypersurface V(g) is non-singular at a ∈ V(g) if and only if there is at least one index
j ∈ { 1, ..., n } with (∂g/∂x j )(a) , 0.
IV.7.7 Examples. i) We define the curve
Z := V(x3 − y2 − y3 ) ⊂ A2C
and compute the partial derivatives of f := x3 − y2 − y3 :
∂f ∂f
= 3x2 , = −y · (2 + 3y).
∂x ∂y
They both vanish at (0, 0) and (0, −2/3). The second point does not lie on Z. So, Z has
exactly one singular point at the origin.

138
 IV.7. Singular Points of Algebraic Varieties

ii) Set f := y · (3x2 − y2 ) − (x2 + y2 )2 and Z := V( f ) ⊂ A2C . This is a clover leaf.

The partial derivatives are

∂f ∂f
= 6xy − 4x · (x2 + y2 ), = 3x2 − 3y2 − 4y · (x2 + y2 ).
∂x ∂y

We look at points where both partial derivatives vanish. These points verify the equation

0 = 3x3 − 3xy2 − 6xy2 = 3x · (x2 − 3y2 ).

√
This gives x = 0 or x = ± 3 · y. In the first case, we must also have −y2 · (3 + 4y) = 0,
that is, y = 0 or y = −3/4. The origin (0, 0) is a singular point of Z, but (0, −3/4) does not
belong to the curve. In the second case, we plug the result into the second derivative:

0 = 9y2 − 3y2 − 4y · 4y2 = 2y2 · (3 − 8y).

√ √
So, y = 0 and x = 0, or y = 3/8 and x = 3 · 3/8. The point (3 · 3/8, 3/8) does not lie
on the curve.

139
IV. Dimension Theory

iii) We look at Cayley’s ruled surface

Z := V(x21 x3 + x32 + x1 x2 ) ⊂ A3C .

The partial derivatives of f := x21 x3 + x32 + x1 x2 are

∂f ∂f ∂f
= 2x1 x3 + x2 , = 3x22 + x1 , = x21 .
∂x1 ∂x2 ∂x3

They all vanish on the line

l := { x1 = 0 = x2 } ⊂ Z.
In other words, Z is singular along the line l.

Let us add a simple example which is not a hypersurface.

IV.7.8 Example (The twisted cubic10 ). We look at the regular map

ϕ: A1k −→ A3k
t 7−→ (t, t2, t3).

The points in the image clearly satisfy the equations11

x2 − y = 0, x3 − z = 0, y3 − z2 = 0.

Set g1 := x2 − y, g2 := x3 − z, g3 := y3 − z2 , I := h g1 , g2 , g3 i, and Z := V(I). The map

ψ : Z −→ A1k
(x, y, z) 7−→ x
10
In this example, we will slightly abuse notation, by not distinguishing between coordinate functions
and coordinates.
11
The third one being redundant.

140
 IV.7. Singular Points of Algebraic Varieties

is inverse to ϕ. Let us determine the corresponding homomorphisms of algebras (Exercise

I.9.8):

ϕ# : k[x, y, z]/I −→ k[t]

[x] 7−→ t
[y] 7−→ t2
[z] 7−→ t3

and

ψ# : k[t] −→ k[x, y, z]/I

t 7−→ [x].

So, the varieties A1k and Z are isomorphic and one dimensional. In particular, Z is non-
singular. Let us check this with the Jacobian criterion:
 
 2x −1 0 
 
JG =  3x2 0 −1  .
 
0 3y2 −2z

This matrix has everywhere rank at least two. We add (3y2 )× the first line and (−2z)× the
second line to the last line and find


6x · (y2 − xz) 0 0 .

This vanishes at every point of Z. So, the matrix has, in fact, rank two at every point of
Z.12
IV.7.9 Lemma. Let k be an algebraically closed field, n ≥ 1, f ∈ k[x1 , ..., xn] an irre-
ducible polynomial, and Z := V( f ) ⊂ Ank . Then,

Sing(Z) := a ∈ Z | Z is singular at a

is a proper Zariski closed subset of Z.

Proof. By the Jacobian criterion IV.7.6, we have
!
∂f ∂f
Sing(Z) = V f, , ..., .
∂x1 ∂xn
This shows that Sing(Z) is Zariski closed in Z. Assume Z = Sing(Z). This means
!
∂f
V( f ) ⊂ V ,
∂xi
i.e.,  ∂ f
f  , i = 1, ..., n,
∂xi
12
The fact that the rank of JG can never be three at a point of Z follows, indeed, from Theorem IV.7.16,
ii).

141
IV. Dimension Theory

by the lemma of Study III.3.11. Since deg(∂ f /∂xi ) < deg( f ), this is only possible if
∂f
= 0, i = 1, ..., n. (IV.12)
∂xi
In characteristic zero, it follows that f is constant. But, this is ruled out by the assumption
that f be irreducible.
If char(k) = p > 0, then (IV.12) implies that f is a polynomial in x1p , ..., xnp. The field
k is perfect, because it is algebraically closed (Example III.5.6, iii). So, we may find a
polynomial g ∈ k[x1 , ..., xn] with
f = g p.
Again, this contradicts the assumption that f be irreducible. 
We would like to extend this result to arbitrary varieties. For this, we need some
preparations.

Semicontinuity of the embedding dimension

Let R be a ring, r, s ∈ N, M ∈ Mat(r, s, R) an (r × s)-matrix with coefficients in R, and
1 ≤ t ≤ min{ r, s }. A t-minor of M is the determinant of a (t × t)-matrix which is obtained
from M by deleting r −t rows and s−t columns. It is an element of R and can be expressed
as a polynomial with integer coefficients in the entries of M. Recall the following basic
result from linear algebra ([7], 3.3.6, Satz):
IV.7.10 Lemma. Let K be a field, r, s ∈ N, M ∈ Mat(r, s, K), and 1 ≤ t ≤ min{ r, s }.
Then, the rank of the matrix M is at least t if and only if it has a non-vanishing t-minor.
Now, let k be an algebraically closed field, n ≥ 1, I ⊂ k[x1 , ..., xn] a radical ideal,
R := k[x1 , ..., xn]/I, Z := V(I) ⊂ Ank , and

M = ( fi j ) i=1,...,r ∈ Mat(r, s, R).

j=1,...,s

For every point a ∈ Z, we obtain the matrix

M(a) := ( fi j (a)) i=1,...,r ∈ Mat(r, s, k).

j=1,...,s

For t ∈ { 1, ..., min{ r, s } }, a t-minor of M is a regular function on Z, i.e., an element of R.

IV.7.11 Exercises. i) Let a ∈ Z be a point and t := rk(M(a)). Show that there is a Zariski
open subset a ∈ U ⊂ Z, such that


∀a ∈ U : rk M(a) ≥ t.

ii) Show that there are a non-empty Zariski open subset ∅ , U ⊂ Z and a natural
number t ∈ { 0, ..., min{ r, s } }, such that


∀a ∈ U : rk M(a) = t

and


∀a ∈ Z : rk M(a) ≤ t.

142
 IV.7. Singular Points of Algebraic Varieties

Suppose g1 , ..., gl ∈ k[x1 , ..., xn] generate I and set G := (g1 , ..., gl). Then, the above
discussion can be applied to the Jacobian matrix JG . An immediate consequence of Exer-
cise IV.7.11 is:
IV.7.12 Proposition. There are a non-empty Zariski open subset ∅ , U ⊂ Z and a
natural number t ∈ { n − min{ l, n }, ..., n }, such that

∀a ∈ U : edim(Rma ) = t, ma = f ∈ R | f (a) = 0 ,

and
∀a ∈ Z : edim(Rma ) ≥ t.
We will see later (Theorem IV.7.16) that t = dim(Z).

Principal open subsets

Let k be an algebraically closed field, n ≥ 1, and f ∈ k[x1 , ..., xn] \ {0}. Then, we call

D( f ) := Ank \ V( f )

the principal open subset associated with f (compare Exercise I.4.18).

IV.7.13 Remark. Let U ⊂ Ank be a non-empty Zariski open subset. Then, there is a non-
zero ideal I ⊂ k[x1 , ..., xn] with Ank \ U = V(I). Pick f ∈ I \ {0}. According to Property
I.9.1, iv), V(I) ⊂ V( f ), and, therefore,

D( f ) ⊂ U.

IV.7.14 Exercise. Let U ⊂ Ank be a non-empty Zariski open subset. Prove that there are
finitely many elements f1 , ..., f s ∈ k[x1 , ..., xn] \ {0} with

U = D( f1 ) ∪ · · · ∪ D( f s ).

The regular function f doesn’t vanish anywhere on D( f ). For this reason, we consider

k[x1 , ..., xn] f

as the algebra of regular functions on D( f ). There is an even better justification for

doing this: The homomorphisms

ϕ : k[x1 , ..., xn+1 ]/hxn+1 · f − 1i −→ k[x1 , ..., xn] f

xi 7−→ xi , i = 1, ..., n
1
xn+1 7−→
f
and

ψ : k[x1 , ..., xn ] f −→ k[x1 , ..., xn+1 ]/hxn+1 · f − 1i

xi 7−→ xi , i = 1, ..., n
1
7−→ xn+1
f

143
IV. Dimension Theory

are inverse to each other. Now, the algebra k[x1 , ..., xn+1 ]/hxn+1 · f − 1i is attached to the
hypersurface
V(xn+1 · f − 1) ⊂ An+1 k .

This hypersurface is the graph of the function

D( f ) −→ A1k
1
a 7−→ .
f (a)
The projection

π : An+1
k −→ Ank
(a1 , ..., an+1) 7−→ (a1 , ..., an)

induces a homeomorphism between V(xn+1 · f − 1) and D( f ). In this way, we may view

D( f ) as an affine algebraic variety. The inclusion

D( f ) ⊂ Ank

corresponds to the localization homomorphism

k[x1 , ..., xn] −→ k[x1 , ..., xn ] f .

IV.7.15 Example. Look at

D(x) = A1k \ {0} ⊂ A1k .
The above construction embeds A1k \ {0} as a hyperbola into A2k .

The same thing can be done for every algebraic variety. Indeed, let p ⊂ k[x1 , ..., xn ] be
a prime ideal, R := k[x1 , ..., xn]/p, and Z = V(p). For f ∈ R \ {0},

D( f ) := Z \ V( f )

is a non-empty open subset. By Property I.9.1, iv), and Exercise III.3.2, its points are in
one-to-one correspondence with the maximal ideals m ⊂ R, such that

f < m. (IV.13)

144
 IV.7. Singular Points of Algebraic Varieties

The set D( f ) inherits a topology from the Zariski topology of Z (compare [18], Section
2.1). On the other hand, R f is a finitely generated k-algebra, and thus defines an affine
algebraic variety Z f . By Corollary II.3.7, ii), the points of Z f correspond to the maximal
ideals of R, satisfying (IV.13). Again, Z f comes with a topology. The reader should check
that both topologies on the set of maximal ideals of R, satisfying (IV.13), thus obtained
do agree and that the localization R −→ R f gives (via Exercise I.9.8) rise to the inclusion
D( f ) ⊂ Z.

The singular locus of an algebraic variety

IV.7.16 Theorem. Let k be an algebraically closed field, n ≥ 1, p ⊂ k[x1 , ..., xn] a prime
ideal, S := k[x1 , ..., xn]/p, and Z := V(p).
i) The singular locus

Sing(Z) := a ∈ Z | Z is singular at a
is a proper Zariski closed subset of Z.
ii) For every point a ∈ Z, we have
edim(S ma ) ≥ dim(Z).
Proof. We will reduce to the case of a hypersurface (Lemma IV.7.9). In fact, we will
show that Z looks almost everywhere like a hypersurface. This is basically a consequence
of the theorem of the primitive element III.5.10.
Let L := Q(S ) be the quotient field of S . The proof of Theorem III.5.14 shows that
there is an injective homomorphism
ϕ : k[t1 , ..., t s ] −→ L,
such that the induced field extension
e
ϕ : K := k(t1 , ..., t s ) −→ L
is separable. Now, there are elements g1 , ..., g s , h1 , ..., h s ∈ S \ {0} with
gi
ϕ(ti ) = , i = 1, ..., s.
hi
By the theorem of the primitive element III.5.10, we may find g s+1 , h s+1 ∈ S \ {0}, such
that
g s+1
α :=
h s+1
is a primitive element for the field extension K ⊂ L. Set h := h1 · · · · · h s · h s+1 . Then, we
may view ϕ as a homomorphism:
ϕ : k[t1 , ..., t s ] −→ S h .
As a k-algebra, S h is generated by ξ1 := [x1 ], ..., ξn := [xn ], and ξn+1 := 1/h. Since
K −→ L is a finite ring extension, these elements are integral over K, i.e., we find natural
numbers si ≥ 1 and elements κi1 , ..., κisi ∈ K with

ξisi + κi1 · ξisi −1 + · · · + κisi −1 · ξi + κisi = 0, i = 1, ..., n + 1.

145
IV. Dimension Theory

Now, there are elements αij ∈ k[t1 , ..., t s ] and βij ∈ k[t1 , ..., t s ] \ {0} with

αij
κij = , j = 1, ..., si, i = 1, ..., n + 1.
βij

Setting
Y
n+1 Y
si
g := βij ,
i=1 j=1

we find a finite ring extension

b
ϕ : k[t1 , ..., t s ]g −→ (S h )g = S g·h

which induces K −→ L.13 Recall that, by our construction, there is a primitive element
α ∈ S g·h . Let ∆ ∈ k[t1 , ..., t s]g be as in Theorem III.5.12. We find a natural number t ∈ N
and an element ∆′ ∈ k[t1 , ..., t s], such that
∆′
∆= .
gt
We see that

S g·h·∆′ = (S g·h )∆′ = k[t1 , ..., t s ]g·∆ [α] = k[t1 , ..., t s ]g·∆ [t s+1 ]/hµα i.

The minimal polynomial µα of α has coefficients in k[t1 , ..., t s ]g (see Lemma III.5.11, ii).
We may find a polynomial γ ∈ k[t1 , ..., t s ] whose irreducible factors are among those of
g and ∆′ , such that f := γ · µα ∈ k[t1 , ..., t s, t s+1 ] is an irreducible polynomial (compare
Lemma I.6.16, i). Let
H := V( f ) ⊂ Aks+1 .
This is an irreducible hypersurface, and

H0 := H \ V(g · ∆′ )

is a dense open subset of H. By construction, H0 is isomorphic to the dense open subset

Z0 := Z \ V(g · h · ∆′ )

of Z. Lemma IV.7.9 yields that there is a dense open subset U ⊂ Z with

Sing(Z) ⊂ Z \ U.

By the previous discussion, the rank of the Jacobian is everywhere at most

t := n − dim(Z),

and the locus where it is strictly less than t is the vanishing locus of all (t × t)-minors of
the Jacobian. But this is also the singular locus of Z. This proves all the claims about
the singular locus. The statement on the embedding dimension follows from this and the
Jacobian criterion IV.7.6. 
13
By Exercise III.4.10, you should be familiar with this argument.

146
 IV.7. Singular Points of Algebraic Varieties

IV.7.17 Remark. In the above proof, we have shown that any algebraic variety has an
open subset which is isomorphic to an open subset of an irreducible hypersurface. You
will usually find this result in the form (see, e.g., [11], Proposition I.4.9)

Any algebraic variety is birationally equivalent to a hypersurface.

IV.7.18 Exercise (Singular and non-singular points I). i) Compute the Jacobian for the
following regular functions f : C2 −→ C. Which points are non-singular by the “Jacobian
criterion”? Find out which equation gives which curve in the following picture. Try to
explain why the remaining points really are singular.

a) f (x, y) = x2 − x4 − y4 , b) f (x, y) = xy − x6 − y6 , c) f (x, y) = x3 − y2 − x4 − y4 ,

d) f (x, y) = x2 y + xy2 − x4 − y4 .

Node Triple point Cusp Tacnode

ii) Do the same as in i) for the following regular functions f : C3 −→ C.

a) f (x, y, z) = xy2 − z2 , b) f (x, y, z) = x2 + y2 − z2 , c) f (x, y, z) = xy + x3 + y3 .

Double point Double line Pinch point

IV.7.19 Exercise (Singular and non-singular points II). Let k be an algebraically closed
field and R := k[x, y, z]/h xyz, z2 i.
i) Is the ring R reduced?
ii) Sketch the “variety” associated with R.
iii) Determine dim(R).
iv) For which maximal ideals m ⊂ R is Rm a regular local ring?

147
IV. Dimension Theory

IV.8 Regularity and Normality

IV.8.1 Lemma. Let R be a noetherian ring, l ∈ N a natural number, p0 ( · · · ( pl a chain
of prime ideals, q ⊃ pl a prime ideal, and x ∈ q. There exists a chain

p′0 ( · · · ( p′l−1 ⊂ q

of prime ideals with x ∈ p′0 .

Proof. l = 1. Here, we simply take p′0 := q.

l − 1 −→ l. If x ∈ pl−1 , we may apply the induction hypothesis with pl−1 instead of q
to get
p′0 ( · · · ( p′l−2 ⊂ pl−1 ( p′l−1 := pl ⊂ q.
If x < pl−1 , then we look at hxi + pl−2 1 pl−1 . Let q ⊂ R/(hxi + pl−2 ) be the image of q in
this ring. The ring (R/(hxi + pl−2 ))q contains a minimal prime ideal s (Exercise I.4.14 or
Theorem II.4.28). The ideal s corresponds to a prime ideal r ⊂ R with

pl−2 ( hxi + pl−2 ⊂ r ⊂ q.

We apply the induction hypothesis once more,14 and find a chain

p′0 ( · · · ( p′l−2 ⊂ r

with x ∈ p′0 . It suffices to show that r ( q.

For this, we look at the integral domain R := R/pl−2 . As usual, for an ideal I ⊂ R, we
denote by I its image in R. If r = q, then the chain

h0i ( pl−1 ( r

of prime ideal in R shows

ht(r) ≥ 2.
By contruction, r is a minimal prime ideal containing [x], the class of x in R. Krull’s
principal ideal theorem requires ht(r) ≤ 1. This contradiction shows that, indeed, r (
q. 

The following lemma provides an important tool for carrying out inductions on the
Krull dimension of a ring.

IV.8.2 Lemma. Let R be a noetherian local ring with maximal ideal m and x ∈ m not a
zero divisor. Then,


dim R/hxi = dim(R) − 1.

Proof. We may apply Lemma IV.8.1 to a maximal chain p0 ( · · · ( pn = m, n = dim(R),

of prime ideals in R, and q = m. It shows


dim R/hxi ≥ dim(R) − 1.
14
namely to the chain u0 ( · · · ( ul−1 with ui = pi , i = 1, ..., l − 1, ul−1 := r, and the prime ideal r

148
 IV.8. Regularity and Normality

On the other hand, in any maximal chain p0 ( · · · ( pn = m, p0 is a minimal prime

ideal. By assumption, x is not a zero divisor. Theorem II.4.28, iii), states that x is not
contained in any minimal prime ideal of R. For this reason


dim R/hxi ≤ dim(R) − 1,

and we infer dim(R/hxi) = dim(R) − 1 as asserted. 

IV.8.3 Proposition. Let R be a regular noetherian local ring. Then, R is an integral
domain.
We will need a slight generalization of prime avoidance (Proposition II.4.17, i):
IV.8.4 Lemma. Let R be a ring, I, J ⊂ R ideals, and p1 , ..., pr ⊂ R prime ideals, such that

J 1 I, J 1 pi , i = 1, ..., r.

Then,
J 1 I ∪ p1 ∪ · · · ∪ pr .
Proof. We abbreviate
P := p1 ∪ · · · ∪ pr .
If I ⊂ P, then Proposition II.4.17, i), applies directly. So, we may exclude this case. Note
also that, by Proposition II.4.17, i), J 1 P. Let x ∈ J \ I. If x < P, we are done. Otherwise,
we pick y ∈ J \ P and z ∈ I \ P. Then, it is readily verified that x + y · z ∈ J \ (I ∪ P). 
Proof of Proposition IV.8.3. Denote the maximal ideal of R by m. We proceed by induc-
tion on n := dim(R).
n = 0. In this case, the regularity of R implies m = h0i (compare Proposition IV.6.2).
n −→ n+ 1. By Theorem II.4.28, ii), a noetherian ring has only finitely many minimal
prime ideals, call them p1 , ..., pr . We assume dim(R) = n + 1 ≥ 1, so m is not a mimimal
prime ideal. The Nakayama lemma III.1.31 shows that m , m2 . By Lemma IV.8.4, we
may pick


x ∈ m \ m2 ∪ p1 ∪ · · · ∪ pr .
Note that this implies that x is not a zero divisor. Let x1 , ..., xn ∈ m be such that the
classes [x1 ], ..., [xn ], [x] in m/m2 form a basis for that (R/m)-vector space. Recall from
Proposition IV.6.2 that
m = h x1 , ..., xn , x i.
Now, we pass to the ring R := R/hxi. It is clear that

edim(R) = n.

By Lemma IV.8.2,
dim(R) = n.
Thus, R is also a regular ring. By induction hypothesis, it is an integral domain. This
means that hxi is a prime ideal.
Let q ⊂ Rhxi be a minimal prime ideal (Exercise I.4.14 or Theorem II.4.28). Then, by
Corollary II.3.7, i),
h0i ⊂ p := qc ⊂ hxi

149
IV. Dimension Theory

is a minimal prime ideal of R. We will show that it has to be zero, using the Nakayama
lemma III.1.31. In fact, let z ∈ p. Then, there exists an element y ∈ R with z = y · x. Since
x < p and p is a prime ideal, it follows that y ∈ p. We have shown

p ⊂ m · p ⊂ p.

As announced, the Nakayama lemma III.1.31 finishes the argument. 

IV.8.5 Exercise. Let R be a noetherian local ring. Show that

dim(R) ≤ edim(R).

IV.8.6 Theorem. Let R be a regular noetherian local ring. Then, R is a normal ring.
Proof. We set up an induction on the dimension n of R as in the proof of Proposition
IV.8.3. In the induction step n −→ n + 1, we find an element x ∈ m \ {0}, such that R/hxi
is a regular noetherian local ring of dimension n. By induction hypothesis, it is normal. It
remains to prove15
Claim. Let R be a noetherian local integral domain with maximal ideal m and x ∈ m\{0}.
If the ring R/hxi is normal, then so is the ring R.
We form the ideal \
I := hxl i.
l∈N

This ideal clearly has the property x · I = I and, thus, m · I = I. The Nakayama lemma
III.1.31 gives I = h0i. We conclude that, for every r ∈ R \ {0}, the number

l(r) := max l ∈ N | r ∈ hxl i

exists and that

r = u · xl(r)
for some element u ∈ R \ hxi. Since the ring R/hxi is normal, it is an integral domain, by
Proposition IV.8.3. This means that hxi is a prime ideal. Using this, we infer

∀r, s ∈ R \ {0} : l(r · s) = l(r) + l(s). (IV.14)

Let K := Q(R) be the quotient field of R, α ∈ R, and β ∈ R\{0}, such that s := α/β ∈ K
is integral over R. We have to show that R = S := R[s] ⊂ K. By the lying-over theorem
IV.2.4, there is a prime ideal q ⊂ S with q ∩ R = hxi. This implies

hxi = R ∩ (S · x). (IV.15)

We observe that l(α) ≥ l(β). To see this, let γ, δ ∈ R \ hxi be elements with α = γ · xl(α)
and β = δ · xl(β) . We have
γ = xl(β)−l(α) · δ · s.
So, l(β) > l(α) implies γ ∈ R ∩ (S · x) = hxi and contradicts the fact γ ∈ R \ hxi. We write
α′
s= with α′ = xl(α)−l(β) · γ.
δ
15
The proof of this claim was kindly supplied by Professor Markus Brodmann.

150
 IV.8. Regularity and Normality

We see that
Rδ = S δ .
Thus, the extension hxie of the prime ideal hxi via the localization R −→ Rδ = S δ is a
prime ideal of S δ , and
r := hxie ∩ S
is a prime ideal of S with r ∩ R = hxi. (We need this more specific prime ideal lying over
hxi for later purposes, namely (IV.16).)
We get the induced finite ring extension

R := R/hxi −→ S := S /r.

It induces a field extension

Q(R) −→ Q(S ).
The foregoing discussion shows

S ⊂ R[δ] ⊂ Q(R),

so that
Q(R) = Q(S ),
and the induction hypothesis gives R = S .
We deduce
S = R + r ⊂ R + hxie (IV.16)
s as R-module. Since s ∈ S and α = s · δ, we see

(IV.15)&hδi⊂R
α′ = s · δ ∈ R ∩ hδi + hxie = hδi + hxi.

This means that we may pick a, b ∈ R with

b·x
α′ = a · δ + b · x, so that s=a+ .
δ
We infer that S is generated as an R-algebra by s1 := b · x/δ. Pick N ∈ N, such that S is
generated as R-module by 1,..., s1N . Then, inside Q(R), we have

1
R⊂S ⊂ · R.
δN
Let us abbreviate
δN · bn · xn
N
tn := δ · sn1 = ∈ R, n ∈ N.
δn
Next, we introduce
s1 b
s2 :== .
x δ
We set S 1 := R[s1 ] = R[s] = S and S 2 := R[s2 ]. We claim that
1
R[s2 ] ⊂ · R.
δN

151
IV. Dimension Theory

For n ∈ N, we find
an
δN · sn2 = .
xn
By definition,
δn · an = (δN · bn ) · xn ∈ hxn i, n ∈ N.
Since δ ∈ R \ hxi, we have l(δ) = 0. By (IV.14), it follows that l(an ) ≥ n, n ∈ N. So, there
is an element un ∈ R with an = un · xn , n ∈ N. We conclude

δN · sn2 = un ∈ R, n ∈ N,

as asserted.
Obviously, the R-module (1/δN ) · R is finitely generated. According to Proposition
III.2.4, “iv)=⇒i)”, s2 is integral over R. We may replace s by s2 . If we apply the procedure
which led from s to s1 to s2 , we find an element s′2 . We next set s3 := s′2 /x and S 3 := R[s3 ].
This process can clearly be iterated and leads to a chain

1
R =: S 0 ⊂ S 1 ⊂ S 2 ⊂ · · · ⊂ S l ⊂ S l+1 ⊂ · · · ⊂ ·R
δN
of R-modules. Since R is a noetherian ring, this chain must become stationary. Let m ∈ N
be the first index for which S m = S m+1 holds true. Then,
sm
sm+1 = ∈ S m+1 = S m .
x
This shows
sm ∈ x · S m ⊂ m · S m . (IV.17)
Let t ∈ S m . Then, there exist a natural number M and elements r0 , ..., r M ∈ R, such that

t = r0 · 1 + r1 · sm + · · · + r M · smM .

By (IV.17), t ∈ R + m · S m . This proves

S m = R + m · S m.

By the Nakayama lemma III.1.31, applied to M = S m /R, we have S m = R. We see that

s ∈ S 1 ⊂ S m ⊂ R. This finishes the argument. 

IV.8.7 Remark. In fact, a stronger16 statement holds true, called the Auslander17 –Buchs-
baum18 theorem: A regular noetherian local ring R is factorial.
The proof requires some tools from homological algebra which we haven’t developed,
so far. References are the original paper [2], [5], Satz 10.10, or [20], Theorem 20.3.

IV.8.8 Proposition. Let R be a noetherian local ring with dim(R) = 1. Then, R is regular
if and only if it is normal.
16
see Lemma III.5.1
17
Maurice Auslander (1926 - 1994), American mathematician.
18
David Alvin Buchsbaum, born 1929, American mathematician.

152
 IV.8. Regularity and Normality

Proof. We have to prove that R is regular, if R is normal. According to Proposition IV.6.2,

we have to verify that the maximal ideal m of R is a principal ideal. Fix x ∈ m \ m2 ,19 and
assume
hxi ( m.
Since R is an integral domain (Proposition III.5.2) and dim(R) = 1, m is the only prime
ideal of R. Thus, by Corollary I.8.18,
p
m = hxi.

The ideal m is finitely generated, so that there is some natural number s with

m s ⊂ hxi,

and we set

s0 := min s ∈ N | m s ⊂ hxi .
Note that s0 > 1. Pick
y ∈ m s0 −1 \ hxi.
This gives
y · m ⊂ hxi,
and we assert
y · m ⊂ x · m.
Otherwise, there would be an element r ∈ R \ m, i.e., a unit of R, with r · x ∈ y · m. This
would give
x = r−1 · (r · x) ∈ y · m ⊂ m2 .
So, in the quotient field Q(R), we have
 y r y y 1
∀r ∈ N : · m ⊂ m, x·R ⊂ R, and R ⊂ · R.
x x x x
This shows that y/x is contained in a finitely generated R-submodule of Q(R). According
to Proposition III.2.4, iii), y/x is integral over R. Since R is normal, we have y/x ∈ R, so
that y ∈ hxi, a contradiction. 

In geometric language, the above proposition says that a normal affine algebraic curve
is smooth. More generally, the normalization of a possibly singular irreducible algebraic
curve is non-singular. So, in the realm of curves, normalization provides a canonical way
to attach to a singular irreducible curve a non-singular one.

IV.8.9 Proposition. Let k be an algebraically closed field, n ≥ 1, and Z ⊂ Ank an algebraic

variety. If Z is normal, then


dim Sing(Z) ≤ dim(Z) − 2.

In the proof, we will use the following result.

19
This set is non-empty by the Nakayama lemma III.1.31 and the assumption dim(R) = 1.

153
IV. Dimension Theory

IV.8.10 Lemma. Let R be a noetherian local integral domain, m ⊂ R its maximal ideal,
and x ∈ m \ {0}. If the ring R/hxi is regular, then R is regular, too.
Proof. By Lemma IV.8.2, dim(R/hxi) = dim(R) − 1 =: d. By assumption, there are
elements r1 , ..., rd , such that their images r1 , ..., rd generate the image m of m in R/hxi.
Then, m = h x, r1 , ..., rd i. Remark IV.6.1 and Proposition IV.6.2 show that R is a regular
local ring. 
Proof of Proposition IV.8.9. Let Reg(Z) := Z \ Sing(Z) be the open subset of nonsingular
points of Z. It suffices to show that

H ∩ Reg(Z) , ∅

holds true for every irreducible algebraic set H ⊂ Z with

dim(H) = dim(Z) − 1.

Let us fix such a subset H ⊂ Z. Then,


p := I(H) := f ∈ k[Z] | ∀a ∈ H : f (a) = 0

is a prime ideal in the coordinate algebra k[Z] of Z. We look at the localization k[Z]p of
the coordinate algebra at that prime ideal. Then:

⋆ k[Z]p is a normal local ring. (This is left as an exercise (compare Exercise III.2.12).
Note that k[Z]p has the same the quotient field as k[Z].)

⋆ dim(k[Z]p ) = 1 (Corollary IV.4.8).

By Proposition IV.8.8, the ring k[Z]p is regular. So, its maximal ideal pe is a principal
ideal. This means that there is an element r ∈ k[Z], such that

pe = r · k[Z]p .

Next, let r1 , ..., r s ∈ p be elements with p := h r1 , ..., r s i, and a1 , ..., a s ∈ k[Z], h1 , ..., h s ∈
k[Z] \ p elements with ri = ai /hi , i = 1, ..., s. Then, with

h := h1 · · · · · h s ,

we find the equality

p · k[Z]h = r · k[Z]h . (IV.18)
Note that k[Z]h is the coordinate algebra of the principal open subset D(h) ⊂ Z (see Page
143f). Since h < p, we have
D(h) ∩ H , ∅.
Let Reg(H) := H \ Sing(H) be the open subset of non-singular points of H. Since H is
irreducible, we have
D(h) ∩ Reg(H) , ∅.
We will show that
D(h) ∩ Reg(H) ⊂ Reg(Z).

154
 IV.8. Regularity and Normality

Let a ∈ D(h) ∩ Reg(H) and ma ⊂ k[Z] the maximal ideal of a. Then, h < ma . So, (IV.18)
gives
p · k[Z]ma = r · k[Z]ma . (IV.19)
Let ma be the image of ma in k[H] = k[Z]/p. One checks

k[H]ma  k[Z]ma /(p · k[Z]ma ).

By (IV.19),
k[H]ma = k[Z]ma /(r · k[Z]ma ).
Our choice of a implies that k[H]ma is a regular local ring. Therefore, Lemma IV.8.10
shows that k[Z]ma is a regular local ring, too. 
IV.8.11 Remark. The attentive reader will have noticed that, in the above proof, we used
only the fact that k[Z]p is normal for every prime ideal p ⊂ k[Z] of height one. In fact, for
any noetherian integral domain R, it is true that R is normal if and only if Rp is normal for
every prime ideal p ⊂ R of height one and
\
R= Rp .
p⊂R prime
ht(p)=1

We refer the reader to [4], Satz (13.25).

For more information on normal rings related to the above discussion, we refer the
reader to [4], p. 199ff, or [5], Section 11.

155
References

[1] M.F. Atiyah, I.G. Macdonald, Introduction to commutative algebra, Addison–

Wesley Publishing Co., Reading, Mass.–London–Don Mills, Ont. 1969, ix+128
pp.
[2] M. Auslander, D.A. Buchsbaum, Unique factorization in regular local rings, Proc.
Natl. Acad. Sci. USA 45 (1959), 733-734.
[3] B. Banaschewski, A new proof that “Krull implies Zorn”, Math. Log. Q. 40
(1994), 478-80.
[4] M. Brodmann, Algebraische Geometrie — Eine Einführung, Basler Lehrbücher,
Basel etc.: Birkhäuser Verlag, 1989, xv+470 S.
[5] M. Brodmann, Kommutative Algebra, lecture notes by R. Boldini and F. Rohrer,
http://www.math.uzh.ch/index.php?id=publikationen&key1=115.
[6] D. Eisenbud, J. Harris, The geometry of schemes, Graduate Texts in Mathematics,
197, Springer-Verlag, New York, 2000, x+294 pp.
[7] G. Fischer, Lineare Algebra. Eine Einführung für Studienanfänger, 17. akt. Aufl.,
Grundkurs Mathematik, Wiesbaden: Vieweg+Teubner, 2010, xxii+384 S.
[8] G. Fischer, R. Sacher, Einführung in die Algebra, 3., überarb. Aufl., Teubner Stu-
dienbücher, Mathematik, Stuttgart: B.G. Teubner, 1983, 240 S.
[9] R.W. Gilmer, Rings in which the unique primary decomposition theorem holds,
Proc. Amer. Math. Soc. 14 (1963), 777-781.
[10] Ph. Griffiths, J. Harris, Principles of algebraic geometry, 2. Aufl., Wiley Classics
Library, New York, NY: John Wiley & Sons Ltd., 1994, xii+813 S.
[11] R. Hartshorne, Algebraic Geometry, Graduate Texts in Mathematics, 52,
Springer-Verlag, New York-Heidelberg, 1977, xvi+496 pp.
[12] H. Herrlich, Axiom of choice, Lecture Notes in Mathematics, 1876, Springer-
Verlag, Berlin, 2006, xiv+194 pp.
[13] D. Huybrechts, Complex geometry. An introduction, Universitext, Berlin:
Springer, 2005, xii+309 S.

157
References

[14] Th. de Jong, G. Pfister, Local analytic geometry. Basic theory and applications,
Advanced Lectures in Mathematics, Braunschweig: Friedr. Vieweg & Sohn,
2000, xii+382 pp.

[15] I. Kaplansky, Commutative rings, 2nd revised ed., Chicago–London: The Univer-
sity of Chicago Press, 1974, viii+182 pp.

[16] S. Katok, p-adic analysis compared with real, Student Mathematical Library, 37,
American Mathematical Society, Providence, RI, 2007, xiv+152 S.

[17] S. Lang, Algebra, 3rd revised ed., Graduate Texts in Mathematics, vol. 211, New
York, NY: Springer, 2002, xv+914 pp.

[18] G. Laures, M. Szymik, Grundkurs Topologie, Spektrum Akademischer Verlag

GmbH, Heidelberg, 2009, x+242 pp.

[19] F. Lorenz, Lineare Algebra I, 3., überarb. Aufl., Mannheim: B.I. Wis-
senschaftsverlag, 1992, x+226 pp.

[20] H. Matsumura, Commutative ring theory, transl. from the Japanese by M. Reid,
Cambridge Studies in Advanced Mathematics, vol. 8, Cambridge etc.: Cambridge
University Press, 1989, 336 pp.

[21] D. Mumford, The red book of varieties and schemes, second, expanded edition, in-
cludes the Michigan lectures (1974) on curves and their Jacobians, with contribu-
tions by Enrico Arbarello, Lecture Notes in Mathematics, 1358, Springer-Verlag,
Berlin, 1999, x+306 pp.

[22] R. Munshi, Hilberts Nullstellensatz, Bull. Bombay Math. Colloq. 15 (1999), 20-
24.

[23] J. Neukirch, Algebraische Zahlentheorie, reprint of the 1992 original, Berlin:

Springer, 2007, xiv+595 pp.

[24] B. Olberding, Noetherian rings without finite normalization, in Progress in com-

mutative algebra 2. Closures, finiteness and factorization, edited by Ch. Fran-
cisco, L. Klingler, S. Sather-Wagstaff, and J.C. Vassilev, Walter de Gruyter GmbH
& Co. KG, Berlin, 2012, x+315 pp, 171-203.

[25] Y. Rav, Variants of Rado’s selection lemma and their applications, Math. Nachr.
79 (1977), 145-65.

[26] R. Remmert, G. Schumacher, Funktionentheorie 1, 5., neu bearb. Aufl., Berlin:

Springer, 2002, xx+402 S.

[27] A. Schmitt, Analysis I, lecture notes,

http://userpage.fu-berlin.de/∼aschmitt.

[28] A. Schmitt, Analysis II, lecture notes,

http://userpage.fu-berlin.de/∼aschmitt.

158
 References

[29] A. Schmitt, Analysis III, lecture notes,

http://userpage.fu-berlin.de/∼aschmitt.

[30] A. Schmitt, Algebra und Zahlentheorie, lecture notes,

http://userpage.fu-berlin.de/∼aschmitt.

[31] A. Schmitt, Funktionentheorie, lecture notes,

http://userpage.fu-berlin.de/∼aschmitt.

[32] R.-H. Schulz, Elementargeometrie, lecture notes,

http://page.mi.fu-berlin.de/rhschulz/Elgeo-Skript/elgeo.html

[33] V. Schulze, Lineare Algebra, lecture notes,

http://page.mi.fu-berlin.de/klarner/lina− skript.pdf

[34] R.G. Swan, On Munshi’s proof of the Nullstellensatz,

www.math.uchicago.edu/∼swan/nss.pdf

[35] H.-P. Tuschik, H. Wolter, Mathematische Logik — kurzgefaßt. Grundlagen,

Modelltheorie, Entscheidbarkeit, Mengenlehre, 2nd ed., Spektrum Hochschul-
taschenbuch, Heidelberg: Spektrum Akademischer Verlag, 2002, viii+214 pp.

159
Index

ACC, 48 basis for a topology, 18

addition, 2, 25, 53, 75 Boole, 17
adjoint matrix, 92, 104, 111 boolean ring, 17, 18
affine algebraic variety, 99 Buchsbaum, 150
normal —, 106
normalization of an —, 106 canonically isomorphic, 6
affine space, 40 Cauchy, 15
algebra Cauchy product, 15
coordinate —, 44 Cayley, 137
finitely generated —, 51 Cayley’s ruled surface, 137
algebra of regular functions, 44 chain, 13
algebra of regular functions on ascending —, 47
a principal open subset, 141 descending —, 47
algebra over a ring, 51 stationary —, 48
algebraic set, 40 chain of ideals, 24
algebraically closed, 16 chain theorem, 125, 127
algebraically independent, 8, 120 Chinese remainder theorem, 36
annihilator of a module, 78 closed point, 45
annihilator of an element, 78 closure, 42
annihilator of an ideal, 37 clover leaf, 136
antisymmetry, 13 cokernel, 74
Artin, 52 common divisor, 25
artinian ring, 52
greatest —, 26, 33
ascending chain, 47
commutative, 2, 4
ascending chain condition, 48
component
associated elements, 20
irreducible —, 66
associated prime ideal, 63
composite of field extensions, 113
associative, 2
contraction of an ideal, 39
Auslander, 150
convergent power series, 16
Auslander–Buchsbaum theorem, 150
coordinate algebra, 44
axiom of choice, 13
coprime elements, 26
axiom of dependent choice, 48
coprime ideals, 33
basis Cramer, 92
transcendence —, 120 Cramer’s rule, 92

161
Index

DCC, 52 finite transcendence degree, 120

degree of a monomial, 100 finitely generated algebra, 51
degree of a polynomial, 100 finitely generated ideal, 32
derivation, 133 finitely generated module, 76
descending chain, 47 finitely generated modules
descending chain condition, 52 over principal ideal domains, 85
dimension, 127 finiteness of integral closure, 110, 112
embedding —, 131 finiteness of normalization, 114
Krull —, 115 first isomorphism theorem, 10
tangential —, 131 first uniqueness theorem
dimension of an algebraic set, 127 for primary decompositions, 63
direct product of modules, 75 formal power series, 15
direct product of rings, 4 free module, 76
direct sum of modules, 75 function field of an algebraic variety, 100
discriminant locus, 102
discriminant of a polynomial, 108 Gauß, 24, 28
distributive law, 2, 33 gaußian integers, 39
divides, 19 germ, 16
divisor going-down theorem, 122
common —, 25 going-up theorem, 118
greatest common —, 26, 33 greatest common divisor, 26, 33
zero —, 11
dominant regular map, 99 height, 126
Hilbert, 43
element Hilbert’s basis theorem, 50
identity —, 2, 4 Hilbert’s Nullstellensatz, 43, 46
maximal —, 13 homogeneous polynomial, 100
prime —, 20 homomorphism, 4
embedded prime ideal, 63 module —, 73
embedding dimension, 131 ring —, 4
existence theorem for primary decompo- homomorphism of modules, 73
sitions in noetherian rings, 59 hyperbola, 142
exponent of a prime torsion module, 82
exponential rule, 5 ideal, 8
extension of an ideal, 38 finitely generated —, 32
irreducible —, 60
factorial ring, 24 maximal —, 12
Fermat, 22 primary —, 58
field prime —, 12
number —, 22 principal —, 8
perfect —, 108, 109 radical —, 38
quotient —, 24 ideal generated by a subset, 32
residue —, 14 ideal of an algebraic set, 41
field extension ideal of regular functions
composite of —s, 113 vanishing on an algebraic set, 41
separable —, 109 ideal quotient, 37
finite ring extension, 89 ideale Zahl, 1

162
 Index

identity element, 2, 4 minimal prime ideal

independent elements in a module, 82 containing an element, 129
infinite transcendence degree, 120 minimal prime ideal
integral closure, 93 containing an ideal, 130
integral domain, 11 minor of a matrix, 140
integral element, 89 module
integral element over an ideal, 121 annihilator of a —, 78
integrally closed inside another ring, 90 finitely generated —, 76
integrally closed ring, 90 free —, 76
irreducible algebraic set, 99 noetherian —, 86
irreducible component, 66 prime torsion —, 82
irreducible element, 20 quotient —, 74
irreducible ideal, 60 rank of a —
irreducible subset, 66 over a principal ideal domain, 78
irreducible topological space, 66 torsion —, 77
irredundant primary decomposition, 61 torsion free —, 77
isolated prime ideal, 63 module homomorphism, 73
isomorphism of modules, 73 module isomorphism, 73
module over a ring, 71
Jacobian criterion, 136 monic polynomial, 108
Jacobian matrix, 135 monomial, 100
Jordan, 74 degree of a —, 100
Jordan normal form, 74, 85 multiplication, 2, 25, 54
kernel, 9 multiplicatively closed subset, 52
Krull, 87 Munshi, 95
Krull dimension, 115 Nakayama, 87
Krull’s Höhensatz, 130 Nakayama lemma, 87
Krull’s intersection theorem, 87 nilpotent element, 11, 12
Krull’s principal ideal theorem, 127, 129 nilradical, 31
Kummer, 1, 40 Noether, 47
Lasker, 61 Noether normalization, 106
laskerian ring, 61 Noether normalization
least common multiple, 33 for hypersurfaces, 100
Leibniz, 133 Noether’s normalization theorem, 104
Leibniz rule, 133, 135 noetherian module, 86
length, 128 noetherian ring, 47
linearly independent elements noetherian topological space, 51
in a module, 80 non-singular point, 133
local ring, 14 norm map, 21
regular —, 131 normal affine algebraic variety, 106
lying-over theorem, 118 normal ring, 90
normal subgroup, 8
maximal element, 13 normalization of a ring, 90
maximal ideal, 12 normalization of an affine
minimal polynomial, 109 algebraic variety, 106
minimal primary decomposition, 61 Nullstellensatz, 43, 46, 95

163
Index

strong —, 96 quasi-compact, 18
weak —, 95 quotient field, 24
number field, 22 quotient module, 74

open subset, 66 Rabinovich, 96

order of an element radical, 38
in a prime torsion module, 82 radical ideal, 38
ramification locus, 102
partial derivatives, 134 ramified covering, 101
partially ordered set, 13 rank of a free module, 77
perfect field, 108, 109 rank of a module
polynomial, 5 over a principal ideal domain, 78
degree of a —, 100 reduced ring, 90
discriminant of a —, 108 reflexivity, 13
homogeneous —, 100 regular function, 44
monic —, 108 regular local ring, 131
primitive —, 27 regular map
separable —, 108, 109 dominant —, 99
polynomial ring, 5, 12 regular point, 133
power of an element, 5 residue field, 14
power of an ideal, 32 ring, 1
power series ring, 15 algebra over a —, 51
primary decomposition, 19, 34 artinian —, 52
first uniqueness theorem for —, 63 boolean —, 17, 18
second uniqueness theorem factorial —, 24
for —, 69 laskerian —, 61
primary decomposition local —, 14
in noetherian rings, 59 noetherian —, 47
primary ideal, 58 polynomial —, 5, 12
prime avoidance, 64, 146 power series —, 15
prime element, 20 semilocal —, 14
prime ideal, 12 ring homomorphism, 4
associated —, 63 ring of integers, 22
embedded —, 63
isolated —, 63 scalar multiplication, 71, 75
prime torsion module, 82 second uniqueness theorem
primitive element, 109 for primary decompositions, 69
theorem of the —, 109 semilocal ring, 14
primitive polynomial, 27 separable element, 109
principal ideal, 8, 77 separable field extension, 109
principal ideal domain, 15, 24 separable polynomial, 108, 109
principal open subset, 18, 141 set
product partially ordered —, 13
Cauchy —, 15 set of common divisors, 25
direct —, 4, 75 set of greatest common divisors, 26
product of ideals, 32 singular point, 133
projection theorem, 102 singularity, 91

164
 Index

spectrum of a product, 19 universal property of normalization, 106

spectrum of a ring, 17 universal property
spectrum of Z[x], 19 of the direct product, 75
stationary chain, 24, 48 universal property of the direct sum, 75
strong Nullstellensatz, 96 universal property of the polynomial
Study, 99 ring, 5
Study’s lemma, 99 universal property of the polynomial
submodule, 74 ring in several variables, 7
submodule generated by a subset, 76 universal property
subring, 4 of the quotient field, 24
subset universal property
irreducible —, 66 of the quotient ring, 10
multiplicatively closed —, 52 upper bound, 13
open —, 66
sum Vandermonde, 111
direct —, 75 Vandermonde determinant, 111
sum of ideals, 17, 32 vanishing locus, 40
Swan, 95 weak Nullstellensatz, 95
symmetric functions, 108
Zariski, 17
tangent space Zariski closed, 17, 42
Zariski —, 133 Zariski open, 17, 42
tangential dimension, 131 Zariski tangent space, 133
Taylor, 134 Zariski topology, 18, 42
Taylor expansion, 134 zero divisor, 11
theorem of Cayley–Hamilton, 78 zero homomorphism, 75
theorem of the primitive element, 109 Zorn, 13
topological space Zorn’s lemma, 14
irreducible —, 66
noetherian —, 51
torsion free module, 77
torsion module, 77
torsion modules
over principal ideal domains, 84
torsion submodule, 77
total ordering, 13
total ring of fractions, 55, 94
transcendence basis, 120
transcendence degree, 120
finite —, 120
infinite —, 120
transitivity, 13
trick of Rabinovich, 96
twisted cubic, 138

unit, 11, 12
universal property of localization, 57

165