M3P8 Algebra III

Lectured by Dr David Helm

Typed by David Kurniadi Angdinata
Autumn 2018

Modules

Rings

Integral domains

Integrally closed domains

Unique factorisation domains

Principal ideal domains

Euclidean domains

Fields

Dedekind domains

Noetherian domains

Noetherian rings

Noetherian modules

Syllabus
Rings. Homomorphisms, ideals, and quotients. Factorisation. The Chinese remainder theorem. Fields
and field extensions. Finite fields. R-modules. Noetherian rings and modules. Polynomial rings in
several variables. Integral extensions and algebraic integers. Dedekind domains. Integers in number
fields. Introduction to algebraic geometry.

1
M3P8 Algebra III Contents

Contents
0 Introduction 4

1 Basic definitions and examples 5

1.1 Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Polynomial rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Subrings and extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Integral domains and rings of fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Homomorphisms, ideals, and quotients 10

2.1 Homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Evaluation homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Images, kernels, and ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Ideals: examples and basic operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Prime and maximal ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Factorisation 14
3.1 Divisibility, units, associates, and irreducibles . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Unique factorisation domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Principal ideal domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 Euclidean domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4 The Chinese remainder theorem 18

4.1 Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 The Chinese remainder theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

5 Fields and field extensions 21

5.1 Prime fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.2 Field extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.3 Extensions generated by one element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.4 Algebraic extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

6 Finite fields 25
6.1 Finite fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.2 The Frobenius automorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.3 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.4 The multiplicative group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.5 Uniqueness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

7 R-modules 29
7.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.2 Submodules, quotients, and direct sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.3 Module homomorphisms, kernels, and images . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.4 Free modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.5 Generators and relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

8 Noetherian rings and modules 35

8.1 Definitions and basic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
8.2 Finitely generated modules over Noetherian rings . . . . . . . . . . . . . . . . . . . . . . . . . 36

2
M3P8 Algebra III Contents

9 Polynomial rings in several variables 38

9.1 The Hilbert basis theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
9.2 Polynomial rings over UFDs are UFDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
9.3 Irreducible polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

10 Integral extensions and algebraic integers 46

10.1 Integral extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

11 Dedekind domains 49
11.1 Dedekind domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
11.2 Ideal class groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

12 Integers in number fields 55

12.1 Integer rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
12.2 Trace and norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
12.3 The main result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

13 Introduction to algebraic geometry 60

13.1 Algebraically closed fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
13.2 Affine algebraic sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
13.3 Proof of the Nullstellensatz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3
M3P8 Algebra III 0 Introduction

0 Introduction
Lecture 1
This course is an introduction to ring theory. The topics covered will include ideals, factorisation, the theory Friday
of field extensions, finite fields, polynomial rings in several variables, and the theory of modules. In addition 05/10/18
to the lecture notes, the following will cover much of the material we will be studying.
• M Artin, Algebra, 1991

Rings are contexts in which it makes sense to add and multiply. For example, Z, Q, R, C, polynomials,
functions {0, 1} → R, and Z/nZ are rings. The goals of this course include
• to unify arguments that apply in all of the above contexts, and
• to study relationships between different rings.

The applications of rings include

• number theory, by studying extensions of Z in which particular Diophantine equations have solutions,
for example n = x2 + y 2 = (x + iy) (x − iy), to study solutions in Z [i] and pass to result about Z,
• algebraic geometry, by the study of zero sets of polynomials in several variables via rings of functions,
and

• topology, by cohomology classes of topological spaces.

The official notes are integrated in these unofficial notes.

4
M3P8 Algebra III 1 Basic definitions and examples

1 Basic definitions and examples

1.1 Rings
Recall the definition of a commmutative ring.

Definition 1.1.1. A commutative ring with identity R is a set together with two binary operations

+R : R × R → R, ·R : R × R → R,

addition and multiplication, and two distinguished elements

0R , 1R ,

such that the following holds.

• The operation +R makes R into an abelian group with identity 0R , that is

– for all r ∈ R,
0R +R r = r +R 0R = 0R ,
– for all r, s, t ∈ R,
(r +R s) +R t = r +R (s +R t) ,
– for all r, s ∈ R,
r +R s = s +R r,
– for all r ∈ R, there exists −r ∈ R such that

r +R (−r) = (−r) +R r = 0R .

• The operation ·R is associative and commutative with identity 1R . That is,

– for all r ∈ R,
1R ·R r = r ·R 1R = 1R ,
– for all r, s, t ∈ R,
(r ·R s) ·R t = r ·R (s ·R t) ,
– for all r, s ∈ R,
r ·R s = s ·R r.

• Multiplication distributes over addition. That is,

– for all r, s, t ∈ R,
r ·R (s +R t) = r ·R s +R r ·R t,
– for all r, s, t ∈ R,
(s +R t) ·R r = s ·R r +R t ·R r.
There is some redundancy here, of course. I have written things this way so that one obtains the definition
of a noncommutative ring simply by removing the condition that multiplication is commutative. In this
course, however, all rings will be commutative. When it is clear from the context what ring we are working
with, we will write 0R and 1R as 0 and 1, a +R b as a + b, and a ·R b as ab.

5
M3P8 Algebra III 1 Basic definitions and examples

Proposition 1.1.2. Let R be a ring. Then for all r ∈ R, r ·R 0R = 0R .

Proof.
r ·R 0R = r ·R (0R +R 0R ) = r ·R 0R +R r ·R 0R .
Thus
0R = − (r ·R 0R ) +R (r ·R 0R ) = − (r ·R 0R ) +R (r ·R 0R +R r ·R 0R ) = r ·R 0R .

Note. Some definitions of rings require 1R 6= 0R in R.

We will not do this. However, it is not hard to see the following.
Proposition 1.1.3. If 0R = 1R , then R is the one-element ring {0R }.

Proof. We certainly have

r = 1R ·R r = 0R ·R r.
On the other hand
0R ·R r = (0R +R 0R ) ·R r = 0R ·R r +R 0R ·R r,
and subtracting 0R ·R r from both sides we find that 0R ·R r = 0R .
Definition 1.1.4. A ring R is a field if R 6= {0R } and every nonzero element of R has a multiplicative
inverse. That is, for every r ∈ R \ {0R } there exists r−1 ∈ R such that

rr−1 = r−1 r = 1R .

We do not consider the zero ring {0R } to be a field. We have seen many examples of rings at this point.
Example.
• The sets Z, Q, R, C are all rings with their usual notion of addition and multiplication. All of them but
Z are in fact fields.

• We have the ring Z/nZ of integers mod n. Let n ∈ Z>0 , and recall that a and b are said to be
congruent mod n if a − b is divisible by n. It is easy to check that this is an equivalence relation on
Z. Moreover, since any a ∈ Z can uniquely be written as qn + r with q, r ∈ Z and 0 ≤ r < n, the set

{[0]n , . . . , [n − 1]n }

is a complete list of the equivalence classes under this relation, where [a]n denotes the set of all integers
congruent to a mod n. We denote this n-element set by Z/nZ, and we can define addition and
multiplication in Z/nZ by setting

[a]n + [b]n = [a + b]n ,

[a]n [b]n = [ab]n .

This defines a ring structure on Z/nZ, once one checks that it is well-defined. This is the first example
of a general construction we will see more of later, the quotient of a ring by an ideal.

6
M3P8 Algebra III 1 Basic definitions and examples

1.2 Polynomial rings

Lecture 2
A very important class of rings that we will study are the polynomial rings. Let R be any ring. Then we Monday
can form a new ring R [X], called the ring of polynomials in X with coefficients in R. Informally, a 08/10/18
polynomial in R [X] is a finite sum of the form

r0 + · · · + rn X n , n ∈ Z≥0 , r0 , . . . , rn ∈ R.

If n > m, we consider r0 + · · · + rn X n to represent the same polynomial of R [X] as s0 + · · · + sm X m if

ri = si for i ≤ m and ri = 0R for i > m. That is, you can pad out a polynomial with terms of the form
0R X i without changing it. From a formal standpoint, it is better to define a polynomial to be an infinite
sum
X∞
ri X i = r0 + r1 X + . . . , ri ∈ R,
n=0

in which all but finitely many ri are zero. This makes it easier to define addition and multiplication. The
degree of such an expression is the largest i such that ri is nonzero. We add and multiply in R [X] just as
we would any other polynomials.
∞ ∞ ∞
! !
X X X
i i
ri X +R[X] si X = (ri +R si ) X i ,
i=0 i=0 i=0
 
∞ ∞ ∞
! ! i
X X X X
ri X i ·R[X] si X i =  (rj ·R si−j ) X i .
i=0 i=0 i=0 j=0

What about polynomial rings in more than one variable? Since the construction of polynomial rings takes
an arbitrary ring as input, one can iterate it. Start with a ring R, and consider first the ring R [X] and then
the ring (R [X]) [Y ]. A polynomial of this has the form
 
X∞ X∞
 rij X j  Y i , rij ∈ R.
i=0 j=0

On the other hand, we can consider the ring (R [Y ]) [X], whose polynomials have the form
 
X∞ ∞
X
 rij Y j  X i , rij ∈ R.
i=0 j=0

Alternatively, we could consider the ring R [X, Y ] whose polynomials are formal expressions of the form
∞
X
rij X i Y j , rij ∈ R,
i,j=0

with only finitely many nonzero coefficients rij and define addition and multiplication in the usual way. It
is not hard to see that all three approaches yield the same ring. If we identify the elements
   
X∞ ∞
X X∞ X∞ X∞
 rij X j  Y i ∈ (R [X]) [Y ] ,  rij Y j  X i ∈ (R [Y ]) [X] , rij X i Y j ∈ R [X, Y ] ,
i=0 j=0 i=0 j=0 i,j=0

we see that addition and multiplication in any of these three rings gives the same answer. We will therefore
primarily use notation like R [X, Y ] for polynomial rings in multiple variables, but we will occasionally need
to know that this is the same as (R [X]) [Y ] or (R [Y ]) [X]. The identifications we have made here are an
example of isomorphisms of rings, a notion we will make precise later.

7
M3P8 Algebra III 1 Basic definitions and examples

1.3 Subrings and extensions

Definition 1.3.1. Let R be a ring. A subset S of R is a subring of R if
• 0R , 1R , −1R ∈ S, and
• S is closed under +R and ·R , so if r, s ∈ S, then so are r +R s and r ·R s.
Example. Z is a subring of R, which is itself a subring of C.
Subrings inherit the additive and multiplicative structures from the ring that contains them, and are thus
themselves rings. It is easy to see that the intersection of two subrings of R, or even an arbitrary collection
of subrings of R, is also a subring of R.
Definition 1.3.2. Now let S ⊆ R be a subring of a ring R, and let α be an element of R. We can then
form a subring S [α] of R, called the subring of R generated by α over S, as follows. An element of R
lies in S [α] if and only if it can be expressed in the form
r0 + · · · + rn αn , n ∈ Z∗ , r0 , . . . , rn ∈ S.
This operation is known as adjoining the element α to the ring S.
Example. Let i denote a square root of −1 in C, and consider the subring Z [i] of C formed by Z ⊆ C and
i. This consists of all complex numbers that can be expressed as polynomials in i with integer coefficients.
Note that such an expression need not be unique. For instance the element 1 + i of Z [i] can also be written
as 2 + i + i2 , and −1 = i2 = i6 = i + i3 + i10 .
Indeed, since i2 = −1, the following holds.
Proposition 1.3.3. We can uniquely express any element a0 + · · · + an in of Z [i] as a + bi for a, b ∈ Z.
P∞
Proof. Given n=0 an in with only finitely many an nonzero, set a = a0 −a2 +· · · ∈ Z and b = a1 −a3 +· · · ∈ Z.
Then
∞
X
an in = a + bi.
n=0
This expression is clearly unique, as if a + bi = c + di in C for a, b, c, d ∈ Z, then a = c and b = d.
If α is more complicated then the elements of R [α] may well be harder to describe, and indeed, a nice
description might not exist at all.
Example.
• If α is the real cube root of 2, then every element of Z [α] can be uniquely expressed as a + bα + cα2 ,
where a, b, c ∈ Z.
P∞
• In Z [π], any element
P∞ hasna uniqueP∞ expression in the form n=0 an π n for all but finitely many an are
n
zero. Suppose n=0 an π = n=0 bn π , then
∞
X
0= (an − bn ) π n .
n=0

Since π is transcendental, this polynomial must be zero. Thus each an = bn .

• The elements of Z [1/2] can be expressed uniquely as a/b, where b is a power of 2 and a is odd unless
b = 1. If α is a root of the polynomial x2 − x/2 + 1 then α2 ∈ Z [α] and α2 = α/2 − 1. Can show that
every element of Z [α] can be uniquely expressed as a + bα, where a, b lies in Z [1/2], but there are pairs
a, b such that a + bα does not lie in Z [α]. For which pairs a, b of elements of Z [1/2] does a + bα lie in
Z [α]?
An alternative way of defining the ring S [α] is to note that it is the smallest subring of R containing S and
α. In one direction, any such subring contains every expression of the form r0 + · · · + rn αn , with ri ∈ S,
so any subring of R containing S and α contains S [α]. One can thus construct S [α] as the intersection of
every subring of R containing S and α. Since the intersection of any collection of subrings of R is a subring
of R it is clear that this intersection is equal to S [α] as defined above.

8
M3P8 Algebra III 1 Basic definitions and examples

1.4 Integral domains and rings of fractions

Lecture 3
Definition 1.4.1. A zero divisor in a ring R is a nonzero element r of R such that there exists a nonzero Wednesday
s ∈ R with rs = 0. A ring R in which there are no zero divisors is called an integral domain. 10/10/18
Example. Z is an integral domain and any subring of a field is an integral domain, but Z/6Z is not an
integral domain, as [2] [3] is zero mod 6 even though neither [2] nor [3] is zero mod 6.
If R is an integral domain, then we can form the field of fractions of R in analogy to the way we build Q
from Z.
Definition 1.4.2. Let R be an integral domain. The field of fractions K (R) is the set of equivalence
classes of expressions of the form a/b, where a, b are elements of R with b nonzero, and a/b is equivalent to
a0 /b0 if and only if ab0 = a0 b. We add and multiply elements of K (R) just as we do for fractions.

a a0 ab0 + ba0
+ 0
= ,
b b bb0
a a0 aa0
· = .
b b0 bb0
Then K (R) is a field, and it contains R in a natural way as a subring if we identify r with r/1R ∈ K (R).
0R 1R
0K(R) = , 1K(R) = .
1R 1R
If a 6= 0 in R, then b/a ∈ K (R), so
a b ab 1
· = ∼ .
b a ba 1
The field K (R) is in some sense the smallest field containing R as a subring. When we talk about homomorph-
isms and isomorphisms, we will be able to state this more precisely. More generally, let the multiplicative
system S be a subset of R that contains 1R , does not contain 0R and is closed under multiplication. That
is, if a, b are in S then so is ab. For any integral domain R and any multiplicative system S, we can define
S −1 R to be the subring of K (R) consisting of all fractions of the form a/b with b ∈ S. It is easy to see that
this is closed under addition and multiplication, and defines a ring in between R and K (R).

Example. If R = Z and S is the set of powers of 2, then S −1 R = Z [1/2]. On the other hand, if S is the
set of odd integers, then S −1 R is the set of all rational numbers of the form a/b with b odd.
In general S −1 R is the smallest subring of K (R) containing R in which every element of S has a multiplicative
inverse, that is 1/b ∈ S for all b ∈ S. The process of obtaining S −1 R from R is called localisation and is
an extremely powerful tool. One can even make sense of it when R is not an integral domain, but one has
to be more careful. The equivalence relation on fractions is trickier, for example. We will not discuss this in
this course but it will be quite useful in future courses.

9
M3P8 Algebra III 2 Homomorphisms, ideals, and quotients

2 Homomorphisms, ideals, and quotients

2.1 Homomorphisms
Let R and S be rings. A ring homomorphism from R to S is, roughly, a way of interpreting elements of R
as elements of S, in a way that is compatible with the addition and multiplication laws on R and S. More
precisely is the following.
Definition 2.1.1. A function f : R → S is a ring homomorphism if
1. f (1R ) = 1S ,
2. for all r, r0 ∈ R,
f (r +R r0 ) = f (r) +S f (r0 ) ,

3. for all r, r0 ∈ R,
f (r ·R r0 ) = f (r) ·S f (r0 ) .

Note. If f is a homomorphism then f (0R ) = 0S . This is because

f (0R ) = f (0R + 0R ) = f (0R ) +S f (0R ) .

Adding the additive inverse, in S, of f (0R ) to both sides gives 0S = f (0R ). Thus we do not need to require
this as an axiom. On the other hand we do need to require f (1R ) = 1S . For certain R, S one can construct
examples of maps f : R → S that satisfy properties 2 and 3 of the definition without satisfying property 1.
Definition 2.1.2. A bijective homomorphism f : R → S is called an isomorphism. Write S ∼ = R for S
is isomorphic to R. In this case one verifies easily that the inverse map f −1 : S → R is also a bijective
homomorphism.
Example.
• If R is a subring of S, then the inclusion of R into S is a homomorphism. This is just a fancy way of
saying that the addition and multiplication on R are induced from the corresponding operations on S.
In particular the inclusions Z ⊂ Q ⊂ R ⊂ C are all homomorphisms.
• The composition of two homomorphisms is a homomorphism, as is easily checked from the definitions.
• The map Z → Z/nZ that takes m ∈ Z into its congruence class mod n is a ring homomorphism.
In fact, this is a special case of the following construction.
Proposition 2.1.3. Let R be any ring. Then there is a unique ring homomorphism f : Z → R such that

1R + · · · + 1R
 n>0
f (n) = 0R n=0.

− (1R + · · · + 1R ) n<0


Proof. Let f : Z → R be a homomorphism. Then, directly from the definition, we have f (0) = 0R and
f (1) = 1R . In particular for all n > 0,

f (n) = f (1 + · · · + 1) = 1R + · · · + 1R ,

where there are n copies of 1R in the sum. Moreover, since

0R = f (n + (−n)) = f (n) + f (−n) ,

we find that f (−n) is the additive inverse of − (1R + · · · + 1R ). Thus f (n) is determined, for all n, completely
by the fact that f is a homomorphism. In the converse direction, it is not hard to check that the map defined
above is in fact a homomorphism.
Thus, for any ring R, we can regard an integer as an element of R via this homomorphism.

10
M3P8 Algebra III 2 Homomorphisms, ideals, and quotients

2.2 Evaluation homomorphisms

Let R be a ring, and consider the ring R [X] of polynomials in X with coefficients in R. If s is an element of
R, then we can define a homomorphism R [X] → R by evaluation at s. More precisely, given an element
of R [X] of the form
P (X) = r0 + · · · + rn X n , n ∈ Z≥0 , ri ∈ R.
Then P (s) for s ∈ R is defined to be

P (s) = r0 + · · · + rn sn ∈ R.

Consider the map

φs : R [X] −→ R
.
P (X) 7−→ P (s)
In effect, it substitutes s for X. It is easy to check that this is in fact a ring homomorphism. More generally,
if R and S are rings and f : R → S is a homomorphism, and s is an element of S, then we can define a map

φs,f : R [X] −→ S
.
r0 + · · · + rn X n 7−→ f (r0 ) + · · · + f (rn ) sn

That is, by applying f to the coefficients and substituting s for X. Again, this is clearly a homomorphism.
The evaluation homomorphisms φs,f are a fundamental property of polynomial rings. In some sense, they
are the reason polynomial rings are worth studying. In fact, the ring R [X] is uniquely characterised by
the fact that homomorphisms from R [X] to S are in bijection with pairs (s, f ), where f : R → S is a
homomorphism and s is an element of S.

2.3 Images, kernels, and ideals

Definition 2.3.1. Let f : R → S be a homomorphism. The image of f is

Im (f ) = {f (r) | r ∈ R} ⊆ S.

The kernel of f is
Ker (f ) = {r ∈ R | f (r) = 0} ⊆ R.

The image of a homomorphism f : R → S is easily seen to be a subring of S.

Example. If R is a subring of S, f : R → S is the inclusion and s lies in S, then the image of the map
φs,f : R [X] → S is precisely the subring R [s] of S.
By contrast, the kernel of a homomorphism f is almost never a subring of R. For instance, subrings contain
the identity. However, we have the following. Lecture 4
Friday
Definition 2.3.2. A nonempty subset I of R is an ideal of R if I is closed under addition, that is for all 12/10/18
elements i, j of I, i + j is an element of I, and for all elements i of I and r of R, ri is an element of I.
Then one can verify, directly from the definition, that the kernel of any homomorphism f : R → S is an ideal
of R. Any ideal of R contains 0R , and conversely the subset {0R } of R is an ideal, called the zero ideal.
Note. A homomorphism f : R → S is injective if and only if its kernel is the zero ideal. Forward direction
is easy. Conversely, if f (x) = f (y), f (x − y) = 0, so x − y ∈ Ker (f ). If Ker (f ) = {0}, x = y.
The kernel of the homomorphism Z → R is either the zero ideal, or the ideal of multiples of n in Z for
some n > 0. We say that R has characteristic zero or characteristic n, respectively. If not zero, the
characteristic of R is the smallest n such that the sum of n copies of 1R is equal to zero.

11
M3P8 Algebra III 2 Homomorphisms, ideals, and quotients

2.4 Ideals: examples and basic operations

If r is an element of R, then any ideal containing R contains any multiple sr of R, for any r in S. Conversely,
one checks easily that the set {sr | s ∈ R} is an ideal of R. It is known as the ideal of R generated by r, and
denoted hri. An ideal generated by one element in this way is called a principal ideal.
Note. The ideal generated by 1R , or more generally by any element of R with a multiplicative inverse, is all
of R. This ideal is called the unit ideal of R.
Proposition 2.4.1. R is a field if and only if the only ideals of R are the zero ideal {0} and unit ideal R.
Proof. If R is a field, let I ⊆ R be a nonzero ideal. There exists r ∈ I 6= 0. Then for all s ∈ R, sr−1 (r) ∈ I,

so s ∈ I for all s ∈ R. Conversely, if R has only the zero ideal and the unit ideal, let r ∈ R 6= 0, and let
I = {sr | s ∈ R}. This is an ideal that is not the zero ideal, so it is all of R. In particular, 1 ∈ I, so there
exists s ∈ R such that sr = 1.
More generally is the following.
Definition 2.4.2. If S is a subset of elements of R, then any ideal containing S consists of all elements of
R of the form
r0 s0 + · · · + rn sn , n ∈ Z≥0 , ri ∈ R, si ∈ S.
The set of all elements of this form is an ideal of R, known as the ideal of R generated by S, and denoted
hSi. It is the intersection of all the ideals of R containing S. It is also the smallest ideal of R containing S.
If S has one element, hSi is a principal ideal. We will show soon that any ideal of Z is a principal ideal,
as is any ideal of the ring K [X] for any field K. You may well have seen this in last year’s algebra course.
On the other hand, there are rings in which not every ideal is principal. For example, the ideal hX, Y i of
K [X, Y ] is not a principal ideal. Given ideals I and J there are several ways to create new ideals.
Example.
• If I, J are ideals, then the intersection I ∩ J is an ideal. Note that if I and J are given by generators,
it might be hard to find generators for the intersection. Certainly it is not enough to intersect the
generating sets.
• The union of ideals is not usually an ideal. Taking R = Z, h3i ∪ h5i contains 3, 5 but not 3 + 5.
• If I, J are ideals, then the sum I + J is an ideal, which are all expressions of the form i + j for i ∈ I and
j ∈ J. It is the smallest ideal containing both I and J, or equivalently the ideal generated by I ∪ J.
• If I, J are ideals, the product I · J or IJ is the ideal generated by elements of the form ij with i ∈ I
and j ∈ J. This may be strictly larger than the set of such products. For example, consider the
product of the ideals I = hX, Y i and J = hZ, W i in R = K [X, Y, Z, W ] for K a field. The product
IJ = hXZ, XW, Y Z, Y W i contains XZ + Y W , but the latter is not a product of an element in I with
an element in J.
• If I, J are general ideals, the product of ideals I and J is always contained in the intersection of I and
J, but the two need not be equal, even in simple rings like Z. h3i · h3i = h9i ⊆ Z and h3i ∩ h3i = h3i.

2.5 Quotients
Let R be a ring and let I be an ideal of R. If x, y are elements of R, we say that x is congruent to y mod I
if x − y is in I. This is an equivalence relation on R. We denote the equivalence class of r by r + I, or as the
alternative notations [r]I , r. It is the set {r + s | s ∈ I}. Let R/I denote the set of equivalence classes on R
mod I. This set has the natural structure of a ring. The additive and multiplicative identities are 0R + I
and 1R + I, respectively, and addition and multiplication are defined by
(r + I) + (s + I) = (r + s) + I,
(r + I) · (s + I) = (rs) + I.
respectively. One has to check that these are well-defined, but this is not difficult. The ring R/I is called
the quotient of R by the ideal I.

12
M3P8 Algebra III 2 Homomorphisms, ideals, and quotients

Example. If R = Z and I is the ideal generated by n, then R/I is the ring Z/nZ that we have already seen.
There is a natural quotient homomorphism, reduction mod I,
R
R −→
I .
r 7−→ r+I

This homomorphism is surjective with kernel I. We then have the following.

Proposition 2.5.1 (Universal property of the quotient). Let I ⊆ R be an ideal and let f : R → S be a
homomorphism, and suppose that the kernel of f contains I. Then there is a unique homomorphism
R
f: → S,
I
such that for all r ∈ R, f (r) = f (r + I).
Proof. Note that f is necessarily unique, as every element of R/I has the form r + I for some r. We must
thus show that it is well-defined and gives a homomorphism. If r + I = r0 + I, then r and r0 differ by an
element of I, so f (r − r0 ) = 0, so f (r) = f (r0 ) since I is contained in the kernel of f . Thus f is well-defined.
Checking that it is a homomorphism follows from f is a homomorphism.
Note. The kernel of f in Proposition 2.5.1 above is just the image of the kernel of f in R/I. If the kernel of
f is equal to I, this image is the zero ideal and f is injective. In particular, any homomorphism of R to S
can be thought of as an isomorphism of some quotient of R with a subring of S.
Example. Let R ⊆ S be a subring, α ∈ S, and ι : R → S be the inclusion map. Recall that we have an
evaluation at α by φι,α : R [X] → S. Image of this is R [α]. Let I = Ker (φι,α ). Then φι,α descends to a
map φι,α : R [α] /I → S that is injective with image R [α]. So R [α] is isomorphic to a quotient of R [X].
Lecture 5
Monday
2.6 Prime and maximal ideals 15/10/18
Definition 2.6.1. An ideal I of R is prime if the quotient R/I is an integral domain. It is maximal if
R/I is a field.
Note. As fields are integral domains, every maximal ideal is prime. The converse need not hold, of course.
The zero ideal in Z is prime but not maximal.
Lemma 2.6.2. An ideal I is prime if and only if for every pair of elements s, r in R such that rs is in I,
either r is in I or s is in I.
Proof. This is just a restatement of Definition 2.6.1. R/I is an integral domain if and only if for all whenever
two elements r + I and s + I in R/I satisfy (r + I) (s + I) = 0 + I in R/I, either r + I = 0 + I or s + I = 0 + I
in R/I. This is the same as saying rs lies in I if and only if either r or s lies in I.
Lemma 2.6.3. An ideal I is maximal if and only if the only ideals of R containing I are I and the unit
ideal R.
This justifies the name maximal for such ideals.
Proof. First suppose that R/I is a field. Recall that R/I is a field if and only if only ideals of R/I are {0}
and R/I. Given an ideal J ⊆ R/I, let Je be the preimage of J under R → R/I. Je is an ideal containing I
and contained in R. Then J is either the zero ideal of R/I, in which case Je is contained in, and thus equal
to, I, or J is all of R/I, in which case Je contains I and an element of 1R + I, so Je contains 1R and is thus
the unit ideal of R. Conversely, if the only ideals of R containing I are I and the unit ideal, then for any r
in R \ I, the ideal of R generated by I and r contains 1R . We can thus write 1R = rs + i, where i ∈ I and
s ∈ R. This means that s + I and r + I are multiplicative inverses of each other in R/I, so R/I is a field.

13
M3P8 Algebra III 3 Factorisation

3 Factorisation
In these notes R always denotes an integral domain.

3.1 Divisibility, units, associates, and irreducibles

Definition 3.1.1. Let r, s be elements of R. We say r divides s, denoted r | s, if there exists r0 ∈ R
with rr0 = s, or, equivalently, s lies in the principal ideal hri generated by r. An element r that divides 1R
is called a unit of R, or, equivalently, hri = R. The set of units in R forms a group under multiplication
denoted R∗ .
For any element r ∈ R and any unit u of R, both u and ur divide r.
Definition 3.1.2. The set of elements of R of the form ur, with r ∈ R∗ are called associates of R.
That is, r, r0 are associates if r = ur0 for a unit u ∈ R∗ . This implies r | r0 , that is there exists u0 with
u0 u = 1 and u0 r = r0 .
Note. The principal ideals hri and hr0 i are equal if and only if r and r0 are associates.
Definition 3.1.3. A nonzero element r of R is called irreducible if r is not a unit and the only elements
of R that divide r are the units and the associates of r.

3.2 Unique factorisation domains

An interesting question is when elements of rings admit unique factorisations into irreducibles? To that end
we define the following.
Definition 3.2.1. A unique factorisation domain (UFD) is a ring R in which
1. every nonzero, nonunit element r of R admits a factorisation as a finite product of irreducibles in R,
and
2. if
r = p1 . . . pn = q1 . . . qm ∈ R
are two factorisations of r as products of irreducibles pi , qi , then n = m and, after permuting the qi ,
each qi is an associate of pi .
Both conditions can fail.
Example.
• There are certainly domains in which 1 can fail, although they are somewhat exotic. One example is
to take the rational polynomial ring R = C X Q with coefficients in C, whose entries are finite formal
sums
XN
ai X ni , ai ∈ C, ni ∈ Q≥0 .
i=0

Any such expression of R is a polynomial in X 1/n for some

 n. The element X

nof this ring is not a unit,
and also not a finite product of irreducibles. In C X Q , X factors as X 1/n . X has no factorisation
into irreducibles in R.
√ 
• Even if 1 holds, 2 often fails. The classic example of this is R = Z −5 , in which
√ √
2, 3, 1 + −5, 1 − −5

are all irreducibles, none are associates of each other, yet

√
√

(2) (3) = 1 + −5 1 − −5 .

14
M3P8 Algebra III 3 Factorisation

We will show later that a very mild finiteness condition on a domain R, the condition that R is Noetherian,
actually guarantees that 1 holds. Another way to interpret condition 2 is as follows.
Definition 3.2.2. We say an element r of R is prime if the principal ideal hri of R is a prime ideal. In
other words, for any s, s0 in R, if r divides ss0 , then r | s or r | s0 .

Lemma 3.2.3. Prime elements are irreducible.

Proof. If r is prime and s divides r, we can write r = ss0 . Then since r divides ss0 we have that either r
divides s, in which case rs00 = s, then ss0 s00 = s and s0 s00 = 1, so r is an associate of s, or r divides s0 , in
which case s0 = rs00 , then r = srs00 and ss00 = 1, so r is an associate of s0 and s is a unit.
The converse is not necessarily true, but we have the following observation as criteria for R to be a UFD.

Proposition 3.2.4. Let R be a domain in which condition 1 holds. Then condition 2 above holds for R if
and only if every irreducible element of R is prime.
Proof. First suppose condition 2 holds, and let r be an irreducible element of R. If r divides ab, we can
write rs = ab for some s ∈ R. Expanding out s, a, b as products of irreducibles we see that r is an associate
of some irreducible dividing a or b, so r is prime. Conversely, if every irreducible element of R is prime, and
we have products of irreducibles
p1 . . . pn = q1 . . . qm ,
then, since p1 is prime, it divides the product q1 . . . qm and is thus an associate of some qi . We can thus
cancel p1 from the left and qi from the right, after introducing a unit on one side. This is possible because
R is an integral domain. Repeating the process we find that, up to reordering the terms and multiplying by
units, the two expressions coincide.

3.3 Principal ideal domains

Definition 3.3.1. An integral domain R is a principal ideal domain (PID) if every ideal of R is a
principal ideal.
Theorem 3.3.2. Every PID is a UFD.

We first show 1. It is true for units trivially.

Lemma 3.3.3. Let R be a PID. Then every nonzero nonunit r ∈ R has a irreducible divisor.
Proof. Fix r = r0 ∈ R. We first show r has an irreducible factor. If r0 is irreducible we are done. Otherwise
if r0 is not irreducible, we can choose an r1 , not a unit nor an associate of r0 , such that r1 divides r0 , so
r0 = r1 s1 with r1 , s1 not units. If r1 is not irreducible we choose r2 similarly, and repeat. If this process
ever terminates we have found an irreducible divisor of r. Suffices to show this terminates. Suppose it does
not terminate. We obtain an increasing tower of ideals

hr0 i ( hr1 i ( . . . .

Let I be the union of all these ideals generated by r0 , r1 , . . . . Then I is an ideal, so it is generated by some
element s ∈ I. Thus s divides ri for all i. On the other hand, s lives in some hrj i, so rj divides s. Thus
s is an associate of rj , and therefore an associate of ri for all i > j, that is I ⊆ hrj i. This contradicts our
construction, because hrj+1 i ⊆ I and hrj+1 i =6 hrj i.
Thus r has an irreducible divisor s0 .

15
M3P8 Algebra III 3 Factorisation

Lemma 3.3.4. Let R be a PID. Every nonzero nonunit r ∈ R is a finite product of irreducibles.
Proof. Consider rs−1 −1
0 . If this is a unit we are done. If not let s1 be an irreducible divisor of rs0 . If
−1
r (s0 s1 ) is a unit we are done. Otherwise repeat. We obtain a sequence of irreducibles s0 , s1 , . . . such that
s0 . . . si divides r for all i, so
r = r0 s0 = r0 r1 s1 = . . . ,
with r0 , r1 , . . . irreducible. If this process ever terminates we are done. Suppose it does not. Then we have
a strictly increasing tower of ideals
hri ( hs0 i ( hs1 i ( . . . .
This cannot continue forever. Arguing as above we arrive at a contradiction.

Now we show 2.
Proof of Theorem 3.3.2. It suffices to show that in a PID every irreducible is prime. Let r ∈ R be irreducible,
and suppose that r divides st. Want r | s or r | t. Let q be a generator of the ideal hr, si of R, so hr, si = hqi.
Then q divides r, so either q is a unit or q is an associate of r. If q is an associate of r, then since q divides
s, r divides s. On the other hand, if q is a unit, then the ideal generated by r and s is the unit ideal and
1 ∈ hr, si, so we can write 1 = xr + ys for x, y elements of R. We then have t = xrt + yst, and since r divides
both yst and xrt, r divides t.
Lecture 6
Wednesday
3.4 Euclidean domains 16/10/18

One technique for proving that rings are PIDs is Euclid’s algorithm. We formalise this in an abstract setting
as follows.
Definition 3.4.1. Let R be an integral domain.
• A Euclidean norm on R is a function N : R → Z≥0 such that for all a, b ∈ R, with b 6= 0, there exists
q, r ∈ R such that a = qb + r, and either r = 0 or N (r) < N (b).

• An integral domain R is called a Euclidean domain if there is a Euclidean norm on R.

Theorem 3.4.2. Any Euclidean domain is a PID.
Proof. Let R be a Euclidean domain, N be a Euclidean norm on R, and I ⊆ R be a nonzero ideal of R. Let
n be the smallest integer such that there exists a nonzero element a ∈ I with N (a) = n minimal, that is
if b ∈ I and b 6= 0, then N (b) < N (a). Claim that I = hai. Then for any b ∈ I, we can write b = qa + r
with N (r) < N (a) unless r = 0. But since N (a) is the smallest possible norm in I, we must have r = 0, so
b = qa. Thus I is generated by a and we are done.

16
M3P8 Algebra III 3 Factorisation

3.5 Examples
Example.
• The classic example of a Euclidean domain is Z, with

N (x) = |x| , qx ∈ Z.
2
• The ring Z [i] is a Euclidean domain, with N (z) = zz = |z| , so
2
N (x + yi) = |x + yi| = x2 + y 2 .

To see this, note that given a and b in Z [i] for b 6= 0, set q 0 = a/b ∈ Q [i]. Write

q 0 = x0 + iy 0 , x0 , y 0 ∈ Q.

Let x and y be the closest integers to x0 and y 0 , such that

1
|x − x0 | ,|y − y 0 | ≤ ,
2
and set q = x + iy in Z [i] and r = a − bq. Then
a  2 2 2 N (b)
2 2 2
N (r) = |r| = |a − bq| = a − b + (q − q 0 ) = |b (q − q 0 )| = |b| |q − q 0 | ≤ .
b 2

• Similar arguments can be used to prove that Z [α] is a Euclidean domain for
√ √ √
α = −2, α = −1+2 −3 , α = −1+2 −7 .

Beyond this one needs other tricks, and for most α unique factorisation fails.
• A critical example is the polynomial ring K [X] for K a field. Here we can take N (P (X)) to be the
degree of P (X). Then, given polynomials P (X) , T (X) ∈ K [X] and T (X) 6= 0, we can use polynomial
long division to write
P (X) = Q (X) T (X) + R (X) ,
for some Q (X) with the degree of R (X) strictly less than that of T (X), unless T (X) is constant, in
which case we can make R (X) = 0. To prove this, fix T (X). If deg (T (X)) = 0, T (X) is constant, so
T (X) = c 6= 0 ∈ K. Take
1
Q (X) = P (X) , R (X) = 0.
c
Otherwise induct on deg (P (X)). If deg (P (X)) < deg (T (X)), set

R (X) = P (X) , Q (X) = 0.

Suppose the claim is true for polynomials of degree n and P (X) has degree n + 1, so
n+1
X d
X
P (X) = ai X i , T (X) = bi X i , d < n + 1.
i=0 i=0

Then
an+1 n+1−d
S (X) = P (X) − X T (X)
bd
has degree n. By inductive hypothesis there exist Q (X) , R (X) with deg (R (X)) < deg (T (X)) such
that S (X) = Q (X) T (X) + R (X), so
 
an+1 n+1−d
P (X) = X + Q (X) T (X) + R (X) .
bd

Later, will show if R is a UFD, then R [X] is also a UFD.

17
M3P8 Algebra III 4 The Chinese remainder theorem

4 The Chinese remainder theorem

In elementary number theory, let m1 , m2 ∈ Z be relatively prime and a1 , a2 ∈ Z. Then there exists a ∈ Z
such that
a ≡ a1 mod m1 , a ≡ a2 mod m2 .
Moreover, a is unique up to congruence mod m1 m2 . A question is given ideals I1 , . . . , Ir and a1 , . . . , ar ∈ R,
when can we find a a ∈ R with

a ∈ a1 + I1 , ..., a ∈ ar + Ir ?

4.1 Products
Definition 4.1.1. Let R1 , . . . , Rn be rings. The direct product R1 × · · · × Rn is a ring whose elements
are n-tuples (r1 , . . . , rn ) with ri ∈ Ri for all i. The addition and multiplication are given componentwise.

(r1 , . . . , rn ) + (r10 , . . . , rn0 ) = (r1 + r10 , . . . , rn + rn0 ) ,

(r1 , . . . , rn ) (r10 , . . . , rn0 ) = (r1 r10 , . . . , rn rn0 ) .

Note. The product comes with natural homomorphisms πi for all i, the projection onto the i-th factor,
defined by
πi : R1 × · · · × Rn −→ Ri
.
(r1 , . . . , rn ) 7−→ ri
The product also comes with the following universal property.
Theorem 4.1.2 (Universal property of the product). Let S, R1 , . . . , Rn be any rings. For any homomorph-
isms
f1 : S → R1 , ..., f n : S → Rn ,
there exists a unique homomorphism
f : S → R1 × · · · × Rn ,
such that πi ◦ f = f for all i.
Proof. Given fi , the homomorphism f is defined by

f (s) = (f1 (t) , . . . , fn (t)) .

Then (πi ◦ f ) (s) = fi (s). For uniqueness, if (π ◦ g) (s) = fi (s) for all i, then

g (s) = (f1 (s) , . . . , fn (s)) = f (s) .

Q
More generally, if I is any index set, and for each i ∈ I we have a ring Ri , we can define the product i Ri .
An element r of this product is a choice, for each i ∈ I, of an element of Ri . We write such an element as
(ri )i∈I . For each j ∈ I we have a map
Q
πj : i Ri −→ Rj
.
(ri )i∈I 7−→ rj

Such a product satisfies a very similar universal property. For any collection

fi : S → Ri

of maps for each i ∈ I, we get a unique map

Y
f :S→ Ri ,
i

such that πj ◦ f = fj .

18
M3P8 Algebra III 4 The Chinese remainder theorem

4.2 The Chinese remainder theorem

Let R be a ring, and let I1 , . . . , Ir be a finite collection of ideals of R. We have the natural maps
R R
R→ , ..., R→ ,
I1 Ir
which are surjective with kernel Ij . Consider the product map
R R
R→ × ··· × .
I1 Ir
It is easy to see that the kernel of this map is the set of r ∈ R such that r maps to zero in R/Ij for all j.
That is, the kernel is the intersection
I1 ∩ · · · ∩ Ir .
Call this ideal J. We thus have an injective embedding
R R R
,→ × ··· × .
J I1 Ir
A natural question to ask is, what can we say about the image? In other words, given congruence classes
mod I1 , . . . , Ir , when is there a single element of R that lives in all those congruence classes simultaneously?
Note. Because the above map is injective, if one such element exists, then there is a unique congruence class
mod J that satisfies all of the required congruences.
Of course, without further hypotheses we cannot expect this map to be surjective. Think about what happens
when I1 = I2 , for instance. Nonetheless, we have the following.
Definition 4.2.1. We will say I1 , . . . , Ir are pairwise relatively prime if for each i 6= j, the sum Ii + Ij
is the unit ideal in R.
If R = Z, then Ii = hni i, and {Ii } is pairwise relatively prime if and only if for all i 6= j, ni and nj are
relatively prime.
Theorem 4.2.2. Let R be a ring and I1 , . . . , Ir be pairwise relatively prime ideals. Then the natural map
R R R
,→ × ··· ×
J I1 Ir
is an isomorphism.
Proof. We have to prove it is surjective. Fix any tuple (c1 , . . . , cr ) of elements of R. We need to find c ∈ R
such that c ∈ ci + Ii for all i. It suffices to construct, for each i, an element ei of R that is congruent to 1
mod Ii and zero mod Ij for j 6= i. Suppose we have such an element. Then the element
c = c1 e1 + · · · + cr er
is congruent to cj mod Ij for all j. Given i, j with i 6= j, we know that Ii + Ij is the unit ideal. That is, we
can write
aij + bij = 1, aij ∈ Ii , bij ∈ Ij .
Then aij is congruent to 1 mod Ij and zero mod Ii as an element of
R R
× ··· × ,
I1 Ir
so aij has zero in the i-th place and one in the j-th place. Then for any j we can take
Y
ej = aij ,
i6=j

and ej will be congruent to 1 mod Ij and zero mod Ii for all j 6= i, so ej has one only in the j-th place. So
R R
R→ × ··· ×
I1 Ir
is surjective. The result follows.

19
M3P8 Algebra III 4 The Chinese remainder theorem

4.3 Examples
When R = Z, then every ideal is principal, so we can write Ij = hnj i for all j. The condition that Ii + Ij
is the unit ideal becomes the condition that ni ∈ Z are pairwise relatively prime. In this case the ideal J is
generated by the product n of the ni . Specialising, we find the version of the Chinese remainder theorem
from elementary number theory.
Theorem 4.3.1. If {nj ∈ Z} is a finite collection of pairwise relatively prime integers, and n is their product,
then for any c1 , . . . , cr ∈ Z, there exists c ∈ Z unique up to congruence mod n such that c is congruent to ci
mod ni for all i.
Now let K be a field and take R = K [X]. If c1 , . . . , cr ∈ K are distinct elements of K, the ideals

Ii = hX − ci i ⊆ R

are such that

Ii + Ij = hX − ci i + hX − cj i 3 ci − cj ∈ K ∗ ,
so contains 1. That is, Ii + Ij is the unit ideal in R and the ideals Ii are pairwise relatively prime. Moreover,
for each i, Ii is the kernel of the evaluation map

fi : R −→ K
.
P (X) 7−→ P (ci )

Let
f : R −→ K × ··· × K
.
P (X) 7−→ (P (c1 ) , . . . , P (cr ))
Then the following diagram commutes.
f
R K × ··· × K
∼
.
R R R
∼ × ··· ×
J I1 Ir

Chinese remainder theorem implies that f is surjective. We thus have an isomorphism

R
−→ K
Ii ,
P (X) 7−→ P (ci )

for all polynomials P . We thus obtain the following.

Theorem 4.3.2. For any c1 , . . . , cn ∈ K, there is a polynomial P (X) in K [X], unique up to congruence
mod
(X − a1 ) . . . (X − an ) ,
such that P (ai ) = ci for all i.

20
M3P8 Algebra III 5 Fields and field extensions

5 Fields and field extensions

Lecture 7
Next we will use K [X] is a PID for K a field to study fields systematically. Friday
19/10/18
5.1 Prime fields
Let K be a field. We have a unique ring homomorphism

ι : Z −→ K
, n ≥ 0.
n 7−→ nK = 1K + · · · + 1K

Let I be the kernel. Then Z/I ,→ K so Z/I is an integral domain, so I is a prime ideal. Thus I is either
the zero ideal {0}, if K has characteristic zero, or the ideal hpi for some prime p of Z. In the former case
I = {0}, the injection Z ,→ K extends to an inclusion

Q −→ K
a
aK .
7−→ (ιa) ιb−1 =
b bK
In the latter case I = hpi, we get an injection of the field Fp ,

Z/pZ ,→ K,

which we often denote Fp when we think of it as a field. Upshot is that every field K contains exactly one
of Q or Fp for p prime, in exactly one way depending on its characteristic. This field is called the prime
field of K, and it is contained in K in a unique way.

5.2 Field extensions

The prime fields are in some sense the smallest possible fields. Once we know they exist, it makes sense to
study fields by studying pairs K, L of fields such that K ⊆ L of fields, trying to relate L to K.
Definition 5.2.1. A field extension is such a pair of fields K, L with K ⊆ L, and is often denoted L/K.

Note. Such an inclusion of fields L/K makes L into a K-vector space, that is a vector space over K.
Definition 5.2.2. We say that a field extension L/K is finite if L is finite dimensional as a K-vector space.
If this is the case, the degree of such an extension is the dimension of L as a K-vector space dimK (L), and
is denoted [L : K].

Proposition 5.2.3. Let K ⊆ L ⊆ M be fields. Then M/K is finite if and only if M/L and L/K are both
finite. If this is the case then
[M : K] = [M : L] [L : K] .
Proof. First suppose that M/K is finite. Then L is a K-subspace of M , so finite dimensional as a K-vector
space. Moreover, there exists a K-basis m1 , . . . , mr , and this basis spans M over K and thus also over L.
Thus M is finite dimensional as an L-vector space, so M/L is finite. Conversely, suppose L/K and M/L are
finite. Let e1 , . . . , en be a K-basis for L, and let f1 , . . . , fn be an L-basis for M . Then claim that

e1 f1 , . . . , e1 fm , . . . , en f1 , . . . , en fm

is a K-basis for M . Every element x of M can be expressed uniquely as

c1 f1 + · · · + cm fm , ci ∈ L.

Each ci in turn can be expressed as

d1,i e1 + · · · + dn,i en , dj,i ∈ K.

21
M3P8 Algebra III 5 Fields and field extensions

Thus we can express x as

d1,1 e1 f1 + · · · + dn,1 en f1 + · · · + d1,m e1 fm + · · · + dn,m en fm .

In particular the set

{ei fj | 1 ≤ i ≤ n, 1 ≤ j ≤ m}
spans M over K. In this case the degree of L over K is n and the degree of M over L is m, so it remains to
show that {ei fj } is linearly independent over K. Suppose we have elements di,j of K such that
X
di,j ei fj = 0.
i,j

Then, regrouping, we find that !

X X
di,j ei fj = 0
j i

is an L-linear combination of the fj that is zero. Since the fj are linearly independent over L we must have
X
di,j ei = 0,
i

for all j. Since the ei are linearly independent over K we must have di,j = 0 for all i, j.
Lecture 8
Monday
5.3 Extensions generated by one element 22/10/18
Let L/K be a field extension, and let α be an element of L.
Definition 5.3.1. We let K (α) denote the subfield of L consisting of all elements of L that can be expressed
in the form
P (α)
,
Q (α)
where P and Q are polynomials with coefficients in K and Q (α) is not zero. This is the smallest subfield of
L containing K and α.
Recall that if R, S are rings, f : R → S is a homomorphism, and α ∈ S, then have

φf,a : Pn R [X]i −→ S
n .
f (ri ) αi
P
i=1 ri X 7−→ i=1

We also have a natural map

K [X] −→ K (α) ⊆ L
.
P (X) 7−→ P (α)
the inclusion on K. It is a ring homomorphism. Let I be the kernel of this homomorphism. We then get an
injection of K [X] /I into the field K (α). Thus K [X] /I is an integral domain, so I is a prime ideal of K [X].
Since K [X] is a PID, every nonzero prime ideal is maximal. There are thus two cases. In the first I is the
zero ideal that is not maximal. That is, there is no nonzero polynomial Q in K [X] such that Q (α) is zero
in L. We say that α is transcendental over K in this case. In the second I is an ideal hQi for Q ∈ K [X] a
nonzero irreducible polynomial that is a maximal ideal of K [X]. In this case we say α is algebraic over K.
Definition 5.3.2. K (X) is the field of rational functions on X,
( )
P (X)
P, Q ∈ K [X] , Q 6= 0
Q (X)
K (X) = .
∼

22
M3P8 Algebra III 5 Fields and field extensions

Assume first that α is transcendental over K, that is I = {0}. Recall

I = {P (X) ∈ K [X] | P (α) = 0} .

So in this case there is no nonzero polynomial P ∈ K [X] with P (α) = 0. In this case the map taking P (X)
to P (α) is an injection of K [X] into K (α) ⊆ L. In particular every nonzero element of K [X] gets sent to
a nonzero, hence invertible, element of L. Thus the map from K [X] to L extends to an injective map from
the field of fractions of K [X],
K (X) −→ L
P (X) P (α) .
7−→
Q (X) Q (α)
By definition of K (α), this map is surjective so the image of this map is K (α). In particular K (X) and
K (α) are isomorphic.
Note. In this case K (α) is infinite dimensional as a K-vector space. It contains a subspace isomorphic to
K [X], for instance.
If α is algebraic over K, then I is a nonzero maximal ideal of the PID K [X], so it is generated by a
single irreducible polynomial Q (X) in K [X]. As a consequence, since the units in K [X] are just the
constant polynomials, the polynomial Q (X) is well-defined up to a constant factor. It is called the minimal
polynomial of α. By definition, it divides every polynomial P (X) such that P (α) = 0. Since hQ (X)i is
maximal, the ring K [X] / hQ (X)i is a field. Recall that for any P ∈ K [X], can write P (X) uniquely as

A (X) Q (X) + R (X) , deg (R) < deg (Q) .

So
1, . . . , X deg(Q)−1
are a K-basis of K [X] / hQ (X)i. So its dimension as a K-vector space is equal to the degree of Q (X). The
map K [X] → K (α) ⊆ L descends to an injection of K [X] / hQ (X)i into L. Since its image is a subfield of
K (α) containing K and α, this map is an isomorphism
K [X]
K (α) ∼
= .
hQ (X)i
Thus in this case the extension K (α) /K is a finite extension, of degree equal to the degree of Q (X). To
summarise, extend K by a single element by
• building K [X], and
• either passing to field of fractions K (X) to form a transcendental extension, or choosing an irreducible
polynomial Q to form an algebraic extension K [X] / hQ (X)i.
Slightly informally, instead of K [X] / hQ (X)i, we sometimes write K (α), where α is a root of Q (X).

5.4 Algebraic extensions

Definition 5.4.1. An extension L/K is algebraic if every element of L is algebraic over K.
Proposition 5.4.2. If L/K is finite, then L/K is algebraic.
Proof. Suppose not. Let L/K be finite, and suppose α ∈ L is transcendental over K. Then we have an
injection
K [X] −→ K (α) ⊆ L
.
X 7−→ α
Since K [X] is a polynomial ring in K (α), it is an infinite dimensional K-vector space, so L cannot be finite
over K. More explicitly, there is also the following argument. Let d be the dimension of L over K. Then for
any α, the set 1, . . . , αd must be linearly dependent over K. This gives a nonzero polynomial P such that
P (α) = 0.

23
M3P8 Algebra III 5 Fields and field extensions

Corollary 5.4.3. Let L/K be a field extension, and suppose α, β are elements of L algebraic over K. Then
α + β and αβ are algebraic over K. Moreover, if α is nonzero then α−1 is algebraic over K.
Proof. Consider the chain of extensions
K ⊆ K (α) ⊆ K (α, β) ,
where we write K (α, β) for (K (α)) (β). Since α is algebraic over K, K (α) is finite over K, of degree deg (α).
Since β is algebraic over K, it is also algebraic over K (α), so K (α, β) is finite over K (α), of degree at most
deg (β). Thus K (α, β) is algebraic over K, of degree at most (deg (α)) (deg (β)). On the other hand, we also
have a chain of extensions
K ⊆ K (α + β) ⊆ K (α, β) ,
so K (α + β) is finite over K, of degree at most (deg (α)) (deg (β)). Hence α + β is algebraic over K. The
proofs for αβ and α−1 are similar.
Corollary 5.4.4. For any extension L/K, let Lalg be the subset of L consisting of all elements of L that
are algebraic over K. Then Lalg is a field.
Proof. We have seen that Lalg is closed

n under addition, multiplication, and taking inverses. For example, if
a0 + · · · + an αn = 0, then a0 α−1 + · · · + an = 0.

Example. In particular, the subfield Q ⊆ C of complex numbers that are algebraic over Q is a field, called
the field of algebraic numbers.

5.5 Example
Example. Consider the polynomial X 2 + X + 1 in F2 [X]. It has no roots in F2 , so it is irreducible, as a
polynomial of degree two any nontrivial factor would be linear. The other polynomials of degree two are
2
X 2, X 2 + X = X (X + 1) , X 2 + 1 = (X + 1) ,
so X 2 + X + 1 is the unique irreducible polynomial of degree two. Thus the quotient F2 [X] / X 2 + X + 1
is a field extension of degree two of F2 , which is denoted F4 . Its four elements are 0, 1, X, X + 1, or more
precisely, their classes mod X 2 + X + 1 .

· 0 1 X X +1
0 0 0 0 0
1 0 1 X X +1 .
X 0 X X +1 1
X +1 0 X +1 1 X
Note that
2
X 2 = −X − 1 = X + 1, X 2 + X + 1 = 0, (X + 1) = X, X 3 = X (X + 1) = 1
in F4 . In particular the multiplicative group of F4 is cyclic of order three. This is not particularly surprising,
as all groups of order three are cyclic. We will see later, though, that the multiplicative group of any finite
field is cyclic.
Proposition 5.5.1. Let K be a field with four elements. Then K ∼
= F4 .
Proof. Let α ∈ K with α 6= 0 and α 6= 1. Consider 1, α, α2 . Since K has dimension two over F2 , there is a
linear dependence. So there exists a polynomial P in F2 [X] of degree at most two such that P (α) = 0. In
fact P must be irreducible of degree two. If it is divisible by something of degree one, then a polynomial of
degree one vanishes on α, so α = 0 or α = 1. So α2 + α + 1 = 0. The map
F2 [X] −→ K
.
X 7−→ α

descends to F2 [X] / X 2 + X + 1 → K. So F4 embeds in K. Thus K ∼

= F4 .

24
M3P8 Algebra III 6 Finite fields

6 Finite fields
Lecture 9
6.1 Finite fields Wednesday
24/10/18
Let K be a finite field. That is, a field with only finitely many elements. Then K has characteristic p for
some prime p, and is in particular a finite dimensional Fp -vector space. Thus its order is a power pr of p
for r ∈ Z>0 . If we fix a particular prime power pr , then two questions naturally arise. Does there exist a
field of order pr ? If so, can we classify fields of order pr up to isomorphism? We will see that in fact, up to
isomorphism, there is a unique field Fpr of order pr .

6.2 The Frobenius automorphism

Let p be a prime. For any ring R, the map x 7→ xp on R certainly satisfies
p
(xy) = xp y p , x, y ∈ R.

On the other hand,    

p p p p−1 p
(x + y) = x + xy + ··· + xp−1 y + y p .
1 p−1
Now the binomial coefficients satisfy
 
p p!
p = , 1 ≤ i ≤ p − 1,
i i! (p − i)!

so if R has characteristic p, we have

p
(x + y) = xp + y p .
Thus, when R has characteristic p, the map x 7→ xp is a ring homomorphism from R to R, called the
Frobenius endomorphism of R. If R is a field of characteristic p, then the Frobenius endomorphism
is injective. If in addition R is finite, then any injective map from R to R is surjective. In particular
the Frobenius endomorphism is a bijective and an isomorphism from R to R when R is a finite field of
characteristic p. In this case we call the map x 7→ xp the Frobenius automorphism. Composing the
r
Frobenius endomorphism with itself, we find that for any r, x 7→ xp is also an endomorphism of any ring R
of characteristic p.
Example. Let R = F4 . y 7→ y 2 gives

0 7→ 0, 1 7→ 1, X 7→ X + 1, X + 1 7→ X.
r
Let K be a field of pr elements. Then αp = α for all α ∈ K. If α = 0, clear. Otherwise α ∈ K ∗ , so K ∗
r r
is an abelian group of order pr − 1. Lagrange’s theorem implies that αp −1 = 1, so αp = α. We have the
following.
r
Proposition 6.2.1. Let K be a field of characteristic p, such that αp = α for all α ∈ K. Let P (X) ∈ K [X]
r r
be an irreducible factor of X p − X over K [X]. Then every element β of K [X] / hP (X)i satisfies β p = β.
Proof. Let d = deg (P ). Can write
β = c0 + · · · + cd−1 X d−1 .
r
Moreover, since P (X) = 0 in K [X] / hP (X)i and P (X) divides X p − X, we have
r
Xp = X

in K [X] / hP (X)i. Thus

r r r
 r d−1  r d−1
β p = cp0 + · · · + cpd−1 X p = c0 + · · · + cd−1 X p = c0 + · · · + cd−1 X d−1 = β.

25
M3P8 Algebra III 6 Finite fields

Corollary 6.2.2. There exists a field K of characteristic p such that

r
1. αp = α for all α ∈ K, and
r
2. the polynomial X p − X of K [X] factors into linear factors over K [X].
Proof. Let K0 = Fp . K0 satisfies 1. We construct a tower of fields

K0 = Fp ( K1 ( K2 ( . . .
r
all satisfying 1 as follows. Suppose we have constructed Ki satisfying 1. If X p − X factors into linear factors
r
over Ki [X], we are done. Otherwise, choose a nonlinear irreducible factor Pi (X) of X p − X in Ki [X] of
degree at least two, and set
Ki [X]
Ki+1 = .
hPi (X)i
Then Ki+1 is strictly larger than Ki and still satisfies 1. On the other hand, in any field Ki satisfying 1,
r
every element is a root of X p − X, so #Ki ≤ pr for all i. Since this polynomial can have at most pr roots,
this process must eventually terminate.
r
Since X p − X has degree pr , we expect the field K constructed above to have pr elements. So it suffices to
r
show that over any field K of characteristic p, X p − X has no repeated roots. To prove this we need an
additional tool.

6.3 Derivatives
Definition 6.3.1. Let R be a ring, and let

P (X) = r0 + · · · + rd X d

be an element of R [X]. The derivative P 0 (X) of P (X) is the polynomial

r1 + · · · + drd X d−1 .

Note. Just as for differentiation in calculus, we have a Leibniz rule. For P, Q ∈ R [X],
0
(P Q) (X) = P (X) Q0 (X) + P 0 (X) Q (X) ,

by reducing to P, Q monomials.
From this we deduce the following.
Lemma 6.3.2. Let K be a field, and let P (X) be a polynomial in K [X] with a multiple root in K. Then
P (X) and P 0 (X) have a common factor of degree greater than zero.
Proof. Let α ∈ K be the multiple root. Then we can write
2
P (X) = (X − α) Q (X) .

Applying the Leibniz rule we get

2
P 0 (X) = 2 (X − α) Q (X) + (X − α) Q0 (X) ,

and it is clear that X − α divides both P (X) and P 0 (X).

r
Corollary 6.3.3. Let K be a field of characteristic p. Then X p − X has no repeated roots in K.
r
Proof. Let P (X) = X p − X. Then P 0 (X) = −1, so P (X) and P 0 (X) have no common factor.
Corollary 6.3.4. There exists a finite field of pr elements.
Proof. The field K we constructed has pr elements.

26
M3P8 Algebra III 6 Finite fields

6.4 The multiplicative group

Lecture 10
Rather than show immediately that there is a unique finite field of pr elements, we make a detour to study Friday
the multiplicative group of a finite field. This is not strictly necessary to prove uniqueness, but will simplify 26/10/18
the proof, and is of interest in its own right. Let K denote a field of pr elements. The goal of this section is
to show that K ∗ is cyclic.
Note. As a multiplicative group, K ∗ is an abelian group of order pr − 1, so by Lagrange’s theorem, we have
r
αp −1 = 1 for all α ∈ K ∗ .
Recall for an abelian group A, operation written additively, that the order of an element a of A is the smallest
d ∈ Z>0 such that da = 0. The order of an element a of A divides the order of A. If d0 a = 0 for some d0 ∈ Z
then the order of a divides d0 . The order of an element a of K ∗ is the smallest d ∈ Z>0 such that ad = 1.
r
Since ap −1 = 1, the order of a is a divisor of pr − 1. On the other hand, if d is a divisor of pr − 1, then any
element of order dividing d is a root of the polynomial X d − 1. Since K is a field, this polynomial has at
most d roots, so we find that there are at most d elements of K ∗ of order dividing d. Order of any element
r r
divides pr − 1. Know X p −1 − 1 has pr − 1 distinct roots in K. For d | pr − 1, X d − 1 | X p −1 − 1, so X d − 1
has exactly d roots in K. That is, for all d | pr − 1, K ∗ has exactly d elements of order dividing d. In fact,
we have the following.
Proposition 6.4.1. Let A be a finite abelian group of order n, and suppose that A has exactly d elements
of order dividing d, for all d dividing n. Then A is cyclic.
In particular K ∗ is cyclic. The remainder of this section will be devoted to proving Proposition 6.4.1. As
a corollary, we deduce that the multiplicative group of any finite field is cyclic. Consider the cyclic group
Z/nZ. The order of any element in this group is a divisor of n.
Definition 6.4.2. For n ∈ Z, we let Φ (n) denote the number of elements in (Z/nZ, +) of exact order n.
This equals to the number of elements t ∈ Z for 1 ≤ t ≤ n such that (t, n) = 1.
Note. Since [1] in Z/nZ has order n, Φ (n) is nonzero for all n.
Lemma 6.4.3. For any d dividing n, the cyclic group Z/nZ contains a unique subgroup of order d, and any
element of Z/nZ of order dividing d is contained in this subgroup.
Proof. The cyclic h nsubgroup
i C of Z/nZ generated by n/d is clearly a subgroup of order d. This has d elements
[0] , . . . , (d − 1) . Conversely, if x is an element of a subgroup of Z/nZ of order d, then the order of x
d
divides d, so dx is divisible by n, and hence, by unique factorisation, x is divisible by n/d. Thus x is in C
and the claim follows.
As a consequence, we deduce the following.
Corollary 6.4.4. For any d dividing n, Φ (d) is the number of elements of Z/nZ of order d.
Corollary 6.4.5. For any n ∈ Z, we have
X
Φ (d) = n.
d|n

Proof. Since every element of Z/nZ has order d for some d dividing n, the sum over all possible d dividing
n of the number of elements of order d is just the number of elements of Z/nZ, which is n.
Proof of Proposition 6.4.1. Let A be as in Proposition 6.4.1. We must show that A contains an element of
order n. In fact, we will show, by induction on d, that A contains exactly Φ (d) elements of order d for all
d | n. In particular, A has Φ (n) > 0 elements of order n, so it is cyclic. If d = 1, the only element of order
one is the identity of A. Since Φ (1) = 1 the base case holds. Assume the claim is true for all d0 < d. A has d
elements of order dividing d, and Φ (d) elements of order d0 for d0 | d and d0 < d, so the number of elements
of exact order d is X
d− Φ (d0 ) .
d0 |d, d0 <d

By Corollary 6.4.5, this is precisely Φ (d).

27
M3P8 Algebra III 6 Finite fields

6.5 Uniqueness
We now turn to the question of showing that any two fields of pr elements are isomorphic. Let K be such a
field. The cyclicity of K ∗ immediately shows the following.
Proposition 6.5.1. Any finite field K of characteristic p is generated over Fp by a single element α ∈ K.

Proof. Let α be an element of K, that generates K ∗ as an abelian group. Then Fp (α) is contained in K,
but contains αn for all n, so contains K ∗ , hence K = Fp (α).
As a corollary, we deduce the following.
Proposition 6.5.2. For any prime p and any r ∈ Z>0 , there exists an irreducible polynomial P (X) ∈ Fp [X]
of degree r in Fp [X].
Proof. Let K be a finite field of pr elements, let α be an element of K that generates K over Fp , and let P
the minimal polynomial of α over Fp . We then have a surjective map

Fp [X] −→ K
.
X 7−→ α

Its kernel is generated by the irreducible polynomial P (X) of degree deg (P ) = [Fp (α) : Fp ] = r.
We also have the following trick.
r
−1
Lemma 6.5.3. Every irreducible polynomial P (X) of degree r in Fp [X] is a divisor of X p − 1.
r r
Proof. Let K = Fp (α) where α is a root of P . #K = pr so αp − α is zero in K. So P (X) | X p − X.

Corollary 6.5.4. Any two finite fields K, K 0 of cardinality pr are isomorphic.

Proof. Choose α ∈ K such that α generates K over Fp . We can then write

Fp [X]
K = Fp (α) ∼
= ,
hP (X)i

where P (X) is the minimal polynomial of α over Fp . In particular P (X) is irreducible of degree r. Since
r r
P (X) divides X p −1 − 1 in Fp [X], it also divides X p −1 − 1 in K 0 [X]. Since in K 0 [X], the latter factors
into linear factors, P (X) also factors into linear factors over K 0 . In particular there exists a root α0 ∈ K 0 of
P (X) in K 0 [X] such that P (α0 ) = 0. Then the map

Fp [X] −→ K0
X 7−→ α0

has kernel hP (X)i and induces a map

Fp [X]
K → → K0
hP (X)i .
Q (α) 7→ Q (X) 7→ Q (α0 )

Since this is map of fields from K to K 0 that takes α to α0 it is injective. Since both fields K, K 0 have the
same cardinality pr , it is also surjective and an isomorphism.
If K = Q, Q [X] / X 2 − p are pairwise nonisomorphic extensions of degree α for every prime p. Lecture 11
Lecture 11 is a problem class. Monday
29/10/18

28
M3P8 Algebra III 7 R-modules

7 R-modules
Lecture 12
7.1 Definitions Wednesday
31/10/18
Definition 7.1.1. An R-module M is a set, together with two operations

+M : M × M → M, ·M : R × M → M,

such that
1. (M, +) makes M into an abelian group with identity 0M ,

2. for all r ∈ R and m, m0 ∈ M ,

r (m + m0 ) = rm + rm0 ,

3. for all r, r0 ∈ R and m ∈ M ,

(r + r0 ) = rm + r0 m,

4. for all r, r0 ∈ R and m ∈ M ,

(rr0 ) m = r (r0 m) ,

5. for all m ∈ M ,
1R · m = m.

Note. For an abelian group M , let End (M ) denote the set of homomorphisms M → M of abelian groups.
End (M ) is a noncommutative ring. 2 if and only if for all r ∈ R, ·r : M → M lives in End (M ). 3, 4, 5 if
and only if the map R → End (M ) given by 2 is a homomorphism of rings.
Example.

• The usual addition and multiplication on R naturally makes R into an R-module. More generally, any
ideal I of R is an R-module with the usual addition and multiplication.
• If f : R → S, then f makes S into an R-module, where the addition + is the usual addition in S, and
the multiplication law is defined by

r · s = f (r) ·S s, r ∈ S, s ∈ S.

In particular any quotient R/I is an R-module.

• More generally, if f : R → S is a homomorphism, and M is any S-module, then M is also an R-module
via
r · m = f (r) · m.
In particular, R → R/I lets us treat any R/I-module M as an R-module. Note that if M is an
R/I-module, then for all r ∈ I and m ∈ M , r · m = 0. We say that I annihilates M in this situation.
Conversely, if M is an R-module and r · m = 0 for all r ∈ I and m ∈ M , then M naturally has the
structure of an R/I-module. Given r + I ∈ R/I and m ∈ M , we define

(r + I) · m = rm.

If r + I = r0 + I, then r − r0 ∈ I, so

rm − r0 m = (r − r0 ) m = 0,

by assumption.

29
M3P8 Algebra III 7 R-modules

• Let R = Z, and let M be an abelian group. Then M has the unique natural structure of Z-module, as
follows. Property 3 from the module axioms shows that

m + · · · + m
 n>0
n·m= 0 n=0.

(−m) + · · · + (−m) n < 0


Thus the multiplication law Z × M → M is forced on us, and one checks that it does satisfy properties
2 to 5 above. Informally, we say that abelian groups are Z-modules.
• If R is a field, then R-modules are just R-vector spaces.
• Let S be a set, and let MS be the set of R-valued functions f : S → R. We add and multiply pointwise.
For f, g ∈ MS , we can define
f + g = s 7→ f (s) + g (s) ,
rf = s 7→ r · f (s) .
MS is clearly an R-module.
• Also of interest is the R-submodule FS of MS that consists of functions f : S → R such that f (s) = 0R
for all but finitely many s. The R-module FS is called the free R-module on the set S and will be very
important for us.

7.2 Submodules, quotients, and direct sums

Definition 7.2.1. Let M be an R-module. A subset N of M is an R-submodule of M if N is closed under
addition and multiplication by elements of R. That is, N is an additive subgroup of M , and for all r ∈ R,
we have rN ⊆ N . In particular, the ideals of R are just the R-submodules of R.
Definition 7.2.2. If S is any subset of M , we define the R-submodule of M generated by S to be the set
of all elements of M of the form
r1 s1 + · · · + rn sn ,
where the ri are elements of R and the si are elements of S. It is the smallest R-submodule of M containing
S.
Definition 7.2.3. An R-module M is a finitely generated R-module if M admits a finite subset S of M
such that the R-submodule of M generated by S is all of M . We say S is a generating set for M .
Definition 7.2.4. Let M be an R-module and N be an R-submodule of M . We say two elements m, m0
of M are congruent mod N if their difference m − m0 lies in N . This is easily seen to be an equivalence
relation, and the equivalence classes are the cosets of the form m + N , for m ∈ M . The set of equivalence
classes is denoted M/N . It has the natural structure of an R-module, where
(m + N ) + (m0 + N ) = (m + m0 ) + N,
r · (m + N ) = (rm) + N.
This R-module is called the quotient of M by N . If m + N = m0 + N , then m − m0 ∈ N , so
rm − rm0 = r (m − m0 ) ∈ N.
So well-defined. Have a natural map M → M/N taking m to m + N .
Definition 7.2.5. Given two R-modules M1 and M2 , the direct sum M1 ⊕ M2 is the set of ordered pairs
(m1 , m2 ) with
(m1 , m2 ) + (m01 , m02 ) = (m1 + m01 , m2 + m02 ) ,
r (m1 , m2 ) = (rm1 , rm2 ) ,
for m1 , m01 ∈ M1 , m2 , m02 ∈ M2 , and r ∈ R.

30
M3P8 Algebra III 7 R-modules

Example. Let M be an R-module and I an ideal of R. Then we can form the R-submodule IM of M
consisting of all elements of M of the form
i1 m1 + · · · + ir mr ,
where the ij are in I and the mj are in M . This is an R-submodule of M , so we can form the quotient
M/IM . Then M/IM is certainly an R-module, but it is also an R/I-module. One can define multiplication
R M M
× −→
I IM IM .
(r + I, m + IM ) 7−→ rm + IM
As always one has to check that this is well-defined, but this is straightforward. We need that if r − r0 lies
in I, and m − m0 lies in IM , then rm − r0 m0 lies in IM . But
rm − r0 m0 = (r − r0 ) m + r0 (m − m0 ) ,
which is clearly in IM .

7.3 Module homomorphisms, kernels, and images

Definition 7.3.1. A map f : M → N of R-modules is called a homomorphism of R-modules if
• f is a homomorphism of the underlying abelian groups, and
• for all r ∈ R and m ∈ M ,
f (rm) = rf (m) .
Warning that a ring homomorphism R → R satisfies
f (rr0 ) = f (r) f (r0 ) ,
but an R-module homomorphism R → R satisfies
f (rr0 ) = rf (r0 ) .
Definition 7.3.2. The kernel of f : M → N is the set
{m ∈ M | f (m) = 0} ,
and the image of f : M → N is the set
{n ∈ N | ∃m ∈ M, f (m) = n} .
It is easy to see that the kernel and image of a homomorphism of R-modules f : M → N are R-submodules
of M and N , respectively.
Note. In particular there is a natural homomorphism
M
M −→
N .
m 7−→ m+N
This homomorphism has the following universal property, exactly analogous to the universal property of the
quotient construction for rings.
Proposition 7.3.3 (Universal property of the quotient). Let N be an R-submodule of M , and let f : M →
M 0 be an R-module homomorphism whose kernel contains N . Then there is unique homomorphism
M
f: → M 0,
N
such that f (m + N ) = f (m) for all m ∈ M . In particular the kernel of f is the image of Ker (f ) in M/N .
Proof. The proof is identical to that for quotient rings, and will be omitted.

31
M3P8 Algebra III 7 R-modules

7.4 Free modules

Definition 7.4.1. Let M be an R-module. A subset S of M is a basis for M if the following two conditions
hold.
• S spans M over R. For all m ∈ M , there exist s1 , . . . , sn ∈ S finite and r1 , . . . , rn ∈ R such that
m = r1 s1 + · · · + rn sn , that is the R-submodule of M generated by S is all of M .
• S is R-linearly independent. For any collection s1 , . . . , sn of distinct elements of S, and any
r1 , . . . , rn ∈ R, r1 s1 + · · · + rn sn is nonzero in M unless all ri are zero.
Definition 7.4.2. An R-module M that has a basis S is called a free R-module. The cardinality n of the
basis S is called the rank of the free R-module M over R.
Lecture 13
Remark 7.4.3. If R is a field, then the notion of a basis for an R-module coincides with the usual notion for Friday
vector spaces. In this case, at least if one assumes the axiom of choice, every R-module has a basis. When 02/11/18
R is not a field only very special R-modules have bases. For instance any quotient R/I of R, for I a nonzero
ideal, has no basis.
Example. The ring R is a free R-module of rank one over R, with basis {1R }. More generally any unit
u ∈ R∗ gives a basis of R as an R-module.
Recall that the free R-module FS on a set S was defined to be the set of functions f : S → R such that
f (s) = 0 for all but finitely many s ∈ S. For each s ∈ S, we have an element es of FS defined by es (t) = 0
for all t ∈ S with t 6= s, es (s) = 1. Claim that the es form a basis for FS . In particular, given f : R → S
with f (s) = 0 for all but finitely many s, let s1 , . . . , sn be the set of elements in S on which f (si ) is nonzero.
Set ri = f (si ). Claim that
f = r1 es1 + · · · + rn esn .
If f (s) = 0, then s ∈
/ {s1 , . . . , sn } so esi (s) = 0 for all i. For any i, esi (sj ) = 0 if i 6= j and esi (si ) = 1, so
n
!
X
ri esi (sj ) = rj = f (sj ) .
i=1

Then f can be written as r1 es1 + · · · + rn esn , so the es span FS . On the other hand, for all s1 , . . . , sn ∈ S
distinct with
X n
ri esi = 0,
i=1
Pn
i=1 ri esi takes the value ri by evaluating at si for all i, and thus is only the zero function when all ri are
zero for all i, so we do have R-linear independence. Thus FS is free, justifying its name.
Proposition 7.4.4. Let F1 , F2 be free R-modules with basis S1 , S2 . Then F1 ⊕ F2 is free with basis
{(s, 0) | s ∈ S1 } ∪ {(0, s0 ) | s0 ∈ S2 } .
Moreover, if F1 and F2 are free of finite ranks n1 and n2 respectively, then F1 ⊕ F2 is free of rank n1 + n2 .
Proof. For linear independence, let s1 , . . . , sm ∈ S1 and s01 , . . . , s0l ∈ S2 be distinct. Suppose we have
r1 , . . . , rm , r10 , . . . , rl0 ∈ R such that
r1 (s1 , 0) + · · · + rm (sm , 0) + r10 (0, s01 ) + · · · + rl0 (0, s0l ) = 0.
Then
r1 s1 + · · · + rm sm = 0 ∈ M1 , r10 s01 + · · · + rl0 s0l = 0 ∈ M2 ,
so ri , ri0 = 0. For spanning set, let (m, m0 ) ∈ M1 ⊕ M2 . Write
m = r1 s1 + · · · + rm sm , si ∈ S1 , m0 = r10 s01 + · · · + rl0 s0l , s0i ∈ S2 ,
then
(m, m0 ) = r1 (s1 , 0) + · · · + rm (sm , 0) + r10 (0, s01 ) + · · · + rl0 (0, s0l ) .
Thus S1 ∪ S2 is a basis for F1 ⊕ F2 , which immediately proves the claim.

32
M3P8 Algebra III 7 R-modules

Free modules have the following universal property.

Proposition 7.4.5 (Universal property of free modules). Let FS be a free R-module on a set S. Then for
any R-module M , and any map of sets f : S → M , there is a unique homomorphism of R-modules

φf : FS → M,

such that φf (es ) = f (s) for all s ∈ S.

Proof. Define φf by X
φf (g) = g (s) f (s) .
s∈S, g(s)6=0

Note that this is a finite sum since all but finitely many s have g (s) = 0. Then it is clear that this is a
homomorphism of R-modules. On the other hand suppose φ is any other map FS → N with φ (es ) = f (s)
for all s. Then we can write X
g= g (s) es ,
s∈S, g(s)6=0

again a finite sum, so X X

φ (g) = g (s) φ (es ) = g (s) f (s) ,
s∈S, g(s)6=0 s∈S, g(s)6=0

so uniqueness is clear.
The image of φf is the R-submodule of N generated by the elements f (s), for s ∈ S.
Corollary 7.4.6. Let M be a free R-module with a basis T for M . Let S be any set of the same cardinality
as T , and let g : T → S be any bijection. Then the map φf : FS → M is an isomorphism. In particular, any
two free R-modules of the same rank are isomorphic.
Proof. The map φf : FS → M is such that φf (es ) = f (s). Since elements of T are linearly independent,
this map is injective. Suppose φf (g) = 0. Can write
X
g= ri esi ,
i
P
for si distinct, then φf (g) = i ri f (si ). Since si are distinct, f (si ) are distinct elements of T , so
X
ri f (si ) = 0.
i

So ri = 0, so g = 0. Since elements of T span M , this map is surjective. Given m ∈ M , write

X
m= ri ti .
i

For all i, find si , with f (si ) = ti . Then

!
X X
φf ri esi = ri ti = m.
i i

Thus M is isomorphic to FS . Since M was arbitrary, any R-module of rank equal to the cardinality of S is
isomorphic to FS and the result follows.

Note. It is also true, but harder to prove, that if M, N are free of different ranks, then M  N .

33
M3P8 Algebra III 7 R-modules

7.5 Generators and relations

Lecture 14
Now let M be any R-module, and let S = {m1 , . . . , mt } be a finite subset of M generating M . Then we Monday
have a natural map 05/11/18
FS −→ M
,
ei 7−→ mi ∈ S
and this map is surjective. In particular, let K be the kernel of this map, then M = ∼ FS /K. Elements of the
kernel K are called relations among P S. Explicitly, an element of K is a map f : S → R such that f (s) = 0
for all but finitely many s, and s∈S f (s) s = 0. In other words, each element of K encodes a linear relation
among the elements of S. It is a measure of how far the elements of S are from being linearly independent.
Let T = {k1 , . . . , ks } be a subset of K that generates K. Then in the same way as above, we get a surjection
FT −→ K
,
ei 7−→ ki
with FT a free module of rank s. Composing with the inclusion of K in FS gives us a map φ : FT → FS
whose image is K. The map φ determines M up to isomorphism with the quotient FS /K, and hence with
FS /φ (FT ). A description of a module as a quotient of a free module by the image of a map of free modules
is called a presentation of M . If both modules have finite rank the presentation is called finite. A module
that has a finite presentation is called finitely presented. Put another way, a presentation is a description
of a module M in terms of
• a generating set S for M , and
• a generating set T for the linear relations satisfied by S.
When S and T are finite we can encode a presentation in a matrix, called the presentation matrix. Write
S = {e1 , . . . , et } and T = {f1 , . . . , fs }. Then φ is determined by φ (f1 ) , . . . , φ (fs ). For each i we can write
φ (fi ) as a sum
X t
rij esj ,
j=1

and let A be the s by t matrix whose i, j entry is rij . Then A gives a map from Rt to Rs , and the quotient
of Rs by the R-submodule ARt of Rs is isomorphic to M .
Example.
• Let R = Z and M = Z/nZ generated by [1]n . The map Z → M is the quotient map with kernel hni.
So presentation matrix is just (n).
√  √
• Let R = Z −5 and I = 2, 1 + −5 . Then
R2 −→ I √
.
(e1 , e2 ) 7−→ 2, 1 + −5
√
√

Since (2) (3) = 1 + −5 1 − −5 ,
√
√

2e2 − 1 + −5 e1 7→ 0, 3e1 − 1 − −5 e2 7→ 0.
√
√

Claim that the two relations 1 + −5 e1 − 2e2 and 3e1 − 1 − −5 e2 generate K. Let ae1 + be2
√
√


be a relation, so a, b ∈ R and 2a + 1 − −5 b = 0, that is a = 1 + −5 /2 b. A question is for
√


which b does 1 + −5 /2 b lie in R? Claim that the set of such b is an ideal J of R. 2 ∈ J and
√ √ √ √
1 − −5 ∈ J so J contains 2, 1 − −5 . 1 ∈ / J, since 2, 1 − −5 is maximal, J = 2, 1 − −5 .
So we have a presentation matrix
 √ 
1 + −5 3√
: R2 → R2 .
−2 −1 + −5

General idea is if we have a presentation matrix A : Rt → Rs for M , with s rows and t columns, then BAC
is also a presentation matrix for M , where B is s × s and C is t × t, and B and C are invertible matrices
with inverse matrix entries in R.

34
M3P8 Algebra III 8 Noetherian rings and modules

8 Noetherian rings and modules

8.1 Definitions and basic properties
Definition 8.1.1. Let R be a ring and let M be an R-module. We say M is Noetherian if every increasing
infinite chain
M 1 ⊆ M2 ⊆ . . .
of R-submodules Mi of M is eventually constant. That is, for any such chain, there exists N such that
we have Mi = MN for all i ≥ N . A ring R is Noetherian if R is Noetherian as an R-module over itself. Since
the R-submodules of R are just the ideals of R, a ring R is Noetherian if every increasing infinite chain

I1 ⊆ I2 ⊆ . . .

of ideals Ij of R is eventually constant.

The following result about Noetherian R-modules is fundamental.
Theorem 8.1.2. An R-module M is Noetherian if and only if every R-submodule N of M is finitely
generated.

Proof. Suppose first that M is Noetherian, and let N be an R-submodule of M . Choose an element n0 of
N , and let N0 be the R-submodule of N generated by n0 . If N0 is all of N , then N is finitely generated.
Otherwise, choose n1 in N \ N0 , and let N1 be the R-submodule of N generated by n0 and n1 . If N is not
finitely generated, we may continue this process indefinitely, choosing for each i an ni in N \ Ni−1 , which is
nonempty since N is not finitely generated, and letting Ni be generated by n0 , . . . , ni . In this way we obtain
a strictly increasing infinite chain
N0 ( N1 ( . . .
of R-submodules of M , contradicting the fact that M is Noetherian. Conversely, suppose that every R-
submodule of M is finitely generated, and let

M0 ⊆ M1 ⊆ . . .

be an increasing chain. We must show that this chain is eventually constant. Let N be the union of the
R-submodules Mi . Note that N is an R-submodule of M . Thus N is finitely generated, say by n1 , . . . , ns .
If n1 , n2 ∈ N , then there exist i, j with n1 ∈ Mi and n2 ∈ Mj . If d ≥ i, j, n1 , n2 ∈ Md , so n1 + n2 ∈ Md , so
n1 + n2 ∈ N . Since N is the union of the Mj , there exist i1 , . . . , is such that nj is in Mij for all j. Let d be
the largest of the ij . Then Md contains n1 , . . . , ns so it contains N . In particular for any d0 ≥ d we have

N ⊆ Md ⊆ Md0 ⊆ N,

so N = Md = Md0 for all such d0 and the chain is constant after Md .

Lecture 15
Wednesday
Corollary 8.1.3. Let R be a PID. Then R is Noetherian.
07/11/18
Proof. Every ideal of R is principal, hence finitely generated.
Example.

• Any field is Noetherian.

 
• The ring C X Q≥0 is not Noetherian. The ideal consisting of all elements with no constant term is
not finitely generated.

35
M3P8 Algebra III 8 Noetherian rings and modules

8.2 Finitely generated modules over Noetherian rings

The plan is
• to show that Noetherianness has strong consequences, and

• to use these properties to show R is Noetherian implies that R [X] is Noetherian and other consequences.
The goal of this section is to prove the following theorem.
Theorem 8.2.1. Any finitely generated R-module M over a Noetherian ring R is Noetherian.
We proceed in several steps. First note the following.

Proposition 8.2.2. Let M be a Noetherian R-module. Then for any R-submodule N of M ,

1. N is Noetherian, and
2. M/N is Noetherian.

Proof.
1. Since M is Noetherian, any R-submodule of M is finitely generated, and thus any R-submodule of N
is finitely generated.
2. Given a R-submodule N 0 of M/N , let N f0 be its preimage in N under the canonical quotient map
f : M → M/N . We have a surjection from N f0 to N 0 induced by f . N
f0 ⊆ M , so Nf0 is finitely
generated, say by n1 , . . . , ns . Claim that

f (n1 ) , . . . , f (ns )

generate N 0 . Given n ∈ N 0 , there exists n f0 such that f (e

e∈N n) = n. Write

e = r1 n1 + · · · + rs ns ,
n ri ∈ R.

Then
n) = r1 f (n1 ) + · · · + rs f (ns ) .
n = f (e

36
M3P8 Algebra III 8 Noetherian rings and modules

Proposition 8.2.3. Let M be an R-module, let N be a Noetherian R-submodule of M , and suppose that
M/N is Noetherian. Then M is Noetherian.
Proof. Let M 0 be a R-submodule of M . Then M 0 ∩ N is a R-submodule of M , hence finitely generated. Let
a1 , . . . , as ∈ M 0 ∩ N generate M 0 ∩ N . Let M 0 denote the image of M 0 in M/N . This is a R-submodule of
M/N and thus finitely generated. Let b1 , . . . , bt ∈ M 0 ⊆ M/N generate M 0 , and choose elements b1 , . . . , bt
of M 0 mapping to b1 , . . . , bt in M/N , respectively. We now show that
a1 , . . . , as , b1 , . . . , bt
is a generating set for M , proving the claim. Given any m ∈ M 0 , let m be its image in M/N under
0

f : M 0 → M 0 . Then we can write m as a sum

r1 b1 + · · · + rt bt , r1 , . . . , rt ∈ R.
Let
m0 = m − r1 b1 − · · · − rt bt .
Then the image of m0 in M/N is
f (m0 ) = m − r1 b1 − · · · − rt bt = 0.
So m0 lies in N .
m ∈ M 0, r1 b1 ∈ M 0 , ..., rt bt ∈ M 0 ,
so m0 also lies in M 0 . So it lies in M 0 ∩ N . We can thus write m0 as
q1 a1 + · · · + qs as , q1 , . . . , qs ∈ R.
We then have
m = q1 a1 + · · · + qs as + r1 b1 + · · · + rt bt ,
proving the claim.
Corollary 8.2.4. Let M and N are Noetherian R-modules, then so is M ⊕ N .
Proof. We have a surjection
M ⊕ N −→ M
.
(m, n) 7−→ m
Its kernel K is the set of pairs of M ⊕ N of the form (0, n), which is isomorphic to N by the map
N −→ K
,
n 7−→ (0, n)
and hence Noetherian. The surjection M ⊕ N → M descends to an isomorphism
M ⊕N ∼
= M,
K
so that (M ⊕ N ) /K is Noetherian. Thus M ⊕ N is Noetherian.
Now assume R is Noetherian. Then R, R ⊕ R, . . . are all Noetherian R-modules, that is the following.
Corollary 8.2.5. If R is Noetherian, then any free R-module of finite rank is Noetherian.
Proof. A free R-module of rank s is the direct sum of s copies of R, each of which is Noetherian as an
R-module when R is Noetherian.
Proof of Theorem 8.2.1. Let M be a finitely generated R-module, and let m1 , . . . , ms be a set of generators
for M . Then if Rs is a free R-module of rank s, with generators e1 , . . . , es , we have a surjection
Rs −→ M
.
ei 7−→ mi
Let K be the kernel. Then M is isomorphic to Rs /K, and Rs is a Noetherian R-module, so M is Noetherian
as well.

37
M3P8 Algebra III 9 Polynomial rings in several variables

9 Polynomial rings in several variables

9.1 The Hilbert basis theorem
In this section, we will use the ideas of the previous section to establish the following key result about
polynomial rings, known as the Hilbert basis theorem.
Theorem 9.1.1 (Hilbert basis theorem). Let R be a Noetherian ring. Then R [X] is Noetherian.
Lecture 16
Over a field, if Friday
09/11/18
Q (X) = an X n + . . . , P (X) = bm X m + . . . , m ≥ n, an , bm 6= 0,

then  
bm m−n
deg P (X) − X Q (X) < deg (P (X)) .
an
Over a ring, this only goes so far. Over Z, cannot use a multiple of 3X + 4 to reduce the degree of an X n + . . .
unless 3 | an . Let
P (X) = b0 + · · · + bn X n , bn ∈ R∗ .
We say that bn is the leading coefficient of P (X). In general, if I have Q1 (X) , . . . , Qr (X) with degrees
d1 , . . . , dr and leading coefficients a1 , . . . , ar and P (X) of degree d ≥ d1 , . . . , dr then there exist n1 , . . . , nr ∈
R such that
deg P (X) − n1 X d−d1 Q1 (X) − · · · − nr X d−dr Qr (X) < d,

if and only if leading coefficient of P (X) is in the ideal generated by a1 , . . . , ar .

Lemma 9.1.2. Let R be Noetherian and I ⊆ R [X] be an ideal. Let J ⊆ R be the set of leading coefficients
of polynomials in I. That is, the set of a ∈ R such that there exists a polynomial P (X) in I with leading
coefficient a. Then J is an ideal of R.
Proof. Certainly if a ∈ J is the leading coefficient of P (X) ∈ I such that

P (X) = aX n + . . . ,

then for any r ∈ R, ra is the leading coefficient of

rP (X) = raX n + . . . ,

so ra ∈ J, so J is closed under multiplication. On the other hand, if a, b ∈ J are the leading coefficients of
P (X) and Q (X) in I, then let n, m be the degrees of

P (X) = aX n + . . . , Q (X) = bX m + . . .

respectively. Without loss of generality we may assume n ≥ m. Then a + b is the leading coefficient of

P (X) + X n−m Q (X) = (a + b) X n+m + . . . ,

and the latter polynomial is in I so a + b ∈ J. Thus J is closed under addition, and is therefore an ideal.

38
M3P8 Algebra III 9 Polynomial rings in several variables

Now since R is Noetherian, J is finitely generated, say by a1 , . . . , as ∈ R. By definition of J, there are thus
polynomials P1 , . . . , Ps in I, of degrees d1 , . . . , dn , such that
Pi = ai X di + . . .
has leading coefficient ai for all i. Let N be the largest of the di .
Lemma 9.1.3. Given Q (X) ∈ I of degree d ≥ N . Then there exist R1 (X) , . . . , Rs (X) ∈ R [X] such that
Q (X) − R1 (X) P1 (X) − · · · − Rs (X) Ps (X)
has degree less than N .
Proof. The proof is by induction on d and the base case d < N is clear by setting Ri = 0 for all i. Suppose
the claim is true for polynomials of degree less than or equal to d − 1, with d ≥ N . Let a ∈ J be the leading
coefficient of
Q (X) = aX d + . . . ,
so that Q (X) − aX d has degree at most d − 1. Since a lies in J we can write
a = r1 a1 + · · · + rs as .
Then the leading term of the polynomial
r1 X d−d1 P1 (X) + · · · + rs X d−ds Ps (X)
is aX d , so the difference
Q (X) − r1 X d−d1 P1 (X) − · · · − rs X d−ds Ps (X)
has degree at most d − 1 and lies in I. By the inductive hypothesis this difference is an R [X]-linear
combination of the Pi (X),
R1 (X) P1 (X) + · · · + Rs (X) Ps (X) .
So
Q (X) = R1 (X) + r1 X d−d1 P1 (X) + · · · + Rs (X) + rs X d−ds Ps (X)

is as well.
The following proof is due to Emmy Noether, and is a vast simplification of Hilbert’s original proof.
Proof of Theorem 9.1.1. Let I be an ideal of R [X]. We want to show that I is finitely generated. Let
I≤N = I ∩ R [X]≤N
be the subset of I consisting of all polynomials of degree at most N . Then I≤N is an R-submodule of the
R-module R [X]≤N of all polynomials of degree at most N . The latter is free of rank N + 1 and generated by
1, . . . , X N as an R-module, so it is finitely generated, hence Noetherian. In particular since R is Noetherian
I≤N is also a finitely generated R-module. Let T1 (X) , . . . , Tk (X) generate I≤N as an R-module. We will
show that
P1 (X) , . . . , Ps (X) , T1 (X) , . . . , Tk (X)
generate I as an R [X]-module. More precisely, we will show that Q (X) is an R [X]-linear combination of
the Pi (X) and Tj (X). Given Q (X) ∈ I, there exist R1 (X) , . . . , Rs (X) ∈ R [X] such that
Q (X) = R1 (X) P1 (X) + · · · + Rs (X) Ps (X) + T (X) ,
with T (X) ∈ I≤N . There exist r1 , . . . , rk ∈ R such that
T (X) = r1 T1 (X) + · · · + rk Tk (X) ,
so
Q (X) = R1 (X) P1 (X) + · · · + Rs (X) Ps (X) + r1 T1 (X) + · · · + rk Tk (X) .

39
M3P8 Algebra III 9 Polynomial rings in several variables

As a corollary, we deduce the following.

Corollary 9.1.4. Let R be any field or PID, or indeed any Noetherian ring. Then for any n, the ring
R [X1 , . . . , Xn ] is Noetherian.
An observation is that if R is Noetherian and I ⊆ R is an ideal, then R/I is Noetherian. Let J be an ideal
of R/I and Je be preimage of J in R. There exist je1 , . . . , jen generating Je over R. Let ji = jei + I ∈ R/I.
These lie in J and generate J over R/I. In particular, any quotient of polynomial ring over a field or PID
is Noetherian. Indeed, since any quotient of a Noetherian ring is Noetherian, we can say more.
Definition 9.1.5. Let R be a ring. An R-algebra is a ring S together with a homomorphism f : R → S.
If S is an R-algebra, we say that S is finitely generated as an R-algebra over R if there exists a finite set
of elements s1 , . . . , sn ∈ S such that every element of S can be expressed as a polynomial in the si with
coefficients in R. Equivalently, S is generated over R by s1 , . . . , sn if the homomorphism

R [X1 , . . . , Xn ] −→ S
,
Xi 7−→ si

where f : R → S, is surjective.
Note. Any finitely generated R-algebra S is isomorphic to a quotient R [X1 , . . . , Xn ] /I for some n and some
ideal I. Thus we can rephrase the Hilbert basis theorem as saying that if R is Noetherian, then any finitely
generated R-algebra is Noetherian.
Lecture 17
Lecture 17 is a problem class. Monday
12/11/18
9.2 Polynomial rings over UFDs are UFDs Lecture 18
Wednesday
Our next goal is to study factorisation in polynomial rings of the form R [X]. Z [X] is not a PID nor a UFD.
14/11/18
The idea is to relate factorisations in Z [X] to factorisations in Q [X]. Warning that irreducibility in Q [X]
does not imply irreducibility in Z [X].
Example. 3x + 15 is irreducible in Q [X]. In Z [X]

3x + 15 = 3 (x + 15) .

Certainly if R is not a UFD then we cannot expect to have unique factorisation in R [X], since we do not even
have it in R. Assume R is a UFD. Then the ring R [X] might be quite complicated, but R [X] is contained in
a much simpler ring where we do understand factorisation, the ring K [X], where K is the field of fractions
of R. Our goal will thus be to compare factorisations in K [X] with factorisations in R [X]. Fundamental
question is can we turn factorisations in K [X] of P (X) ∈ R [X] into factorisations in R [X]? The key to
doing this is the following result, often called Gauss’ lemma.
Theorem 9.2.1 (Gauss’ lemma). Let R be a UFD and let K be its field of fractions. Let P (X) ∈ R [X],
and let Q (X) be a polynomial in K [X] that divides P (X) in K [X]. Then there is an element α ∈ K ∗ such
that αQ [X] lies in R [X], and divides P (X) in R [X]. In particular, if P (X) is reducible in K [X], then
P (X) is also reducible in R [X].
Proof. Write
P (X) = Q (X) T (X) ∈ K [X] ,
and choose nonzero elements e1 , e2 ∈ R such that e1 Q (X) and e2 T (X) have coefficients in R, and so that
the greatest common divisor of the coefficients of Q (X) is one, as is the greatest common divisor of the
coefficients of T (X). Letting d = e1 e2 , we have

dP (X) = Q0 (X) T 0 (X) , Q0 (X) = e1 Q (X) , T 0 (X) = e2 T (X) .

Suppose d is not a unit in R. Then d is divisible by an irreducible element q of R. Since R is a UFD,

irreducibles are prime, so the ideal of R generated by q is a prime ideal. Thus R/ hqi is an integral domain,

40
M3P8 Algebra III 9 Polynomial rings in several variables

so R/ hqi [X] is as well. Moreover, if Q0 (X) and T 0 (X) are the images of Q0 (X) and T 0 (X) mod hqi in
R/ hqi [X], then we have
dP (X) = Q0 (X) T 0 (X) ,
so 0 = Q0 (X) T 0 (X) in R/ hqi [X]. Since R/ hqi [X] is an integral domain we must have either Q0 (X) = 0 or
T 0 (X) = 0 in R/ hqi [X]. Without loss of generality assume Q0 (X) = 0. Then all the coefficients of Q0 (X)
are divisible by q. Thus

d1 P (X) = Q1 (X) T1 (X) , Q1 (X) , T1 (X) ∈ R [X] ,

and Q1 (X) is a multiple of Q (X) in K [X]. If d1 is a unit, done. Otherwise write d1 = d2 q1 for q1 irreducible.
Same trick implies that

d2 P (X) = Q2 (X) T2 (X) , Q2 (X) , T2 (X) ∈ R [X] ,

and Q2 (X) is a multiple of Q1 (X) in K [X]. Repeat, contradicting our construction of Q0 (X). Thus α = d
is a unit in P (X), and we have
1 1
P (X) = e1 Q (X) e2 T (X) , e1 Q (X) , e2 T (X) ∈ R [X] .
d d

Note. The converse to the last claim of Theorem 9.2.1 is not true. If P (X) is reducible in R [X], it might
be irreducible in K [X].

Example. The polynomial 7x factors into irreducibles as 7 · x in Z [X], but since 7 is a unit in Q [X], 7x is
irreducible in Q [X].
The following lemma shows that this kind of thing is all that can happen, however.
Proposition 9.2.2. Let P (X) in R [X] be a polynomial and suppose that the greatest common divisor of
all of its coefficients is one. Then P (X) is irreducible in K [X] if and only if it is also irreducible in R [X].
Proof. Suppose P (X) is irreducible in R [X], and write

P (X) = Q (X) T (X) , Q (X) , T (X) ∈ R [X] ,

where Q (X) and T (X) are nonunits. If Q (X) or T (X) were constant with degree zero then it would divide
every coefficient of P (X) and thus divide the GCD of those coefficients, making it a unit. Thus Q (X) and
T (X) are nonconstant with positive degree and the factorisation

P (X) = Q (X) T (X)

is also a nontrivial factorisation in K [X], so P (X) is reducible in K [X]. Conversely suppose P is reducible
in K [X]. Then there exist Q (X) ∈ K [X] with

0 < deg (Q) < deg (P ) , Q (X) | P (X) ∈ K [X] .

Gauss’ lemma shows that there exist α ∈ K ∗ such that

αQ (X) ∈ R [X] , αQ (X) | P (X) ∈ R [X] .

41
M3P8 Algebra III 9 Polynomial rings in several variables

We are now in a position to prove the following.

Theorem 9.2.3. If R is a UFD, then R [X] is a UFD.
Proof. For existence of factorisations, let P (X) be an element of R [X]. We must show that P (X) factors
into irreducibles. Let d be the greatest common divisor of the coefficients of P (X), and write

P (X) = dQ (X) ,

where the greatest common divisor of the coefficients of Q (X) is one. Since R is a UFD, d factors into
irreducibles q1 , . . . , qs in R, and these remain irreducible in R [X], so it suffices to show that Q (X) factors
into irreducibles. Factor Q (X) into irreducibles in K [X],

Q (X) = Q1 (X) . . . Qr (X) .

By Gauss’ lemma, there exist scalars α1 , . . . , αr ∈ K ∗ such that

α1 . . . αr = 1, αi Qi (X) ∈ R [X] .

Let Q0i (X) = αi Qi (X). GCD of coefficients of Q01 (X) , . . . , Q0r (X) is one, so Q01 (X) , . . . , Q0r (X) are irredu-
cible in R [X] since they are irreducible in K [X]. For uniqueness of factorisations, it remains to show that
if P (X) ∈ R [X] is irreducible in R [X] and divides A (X) B (X) in R [X] for A (X) , B (X) ∈ R [X], then
P (X) divides either A (X) or B (X) in R [X].
• If P (X) is constant, then P (X) = c is irreducible in R. In R/ hci [X] a domain,

0 = A (X) B (X) ,

so A (X) = 0 or B (X) = 0, so c | A (X) or c | B (X).

• If P (X) is nonconstant, since P (X) is irreducible in R [X] it is irreducible in K [X] by Gauss’ lemma,
and hence divides either A (X) or B (X) in K [X]. Suppose P (X) divides A (X) in K [X]. Then

A (X) = P (X) Q (X)

in K [X]. Then there is an element α = r/s ∈ K ∗ for r, s ∈ R and r 6= 0 such that

αP (X) ∈ R [X] , αP (X) | A (X) ∈ R [X] , A (X) = αP (X) α−1 Q (X) ∈ R [X] ,

by Gauss’ lemma. On the other hand, since P (X) is irreducible in R [X] the GCD of its coefficients
is one, so the only way αP (X) lies in R [X] is if s is a unit and α lies in R. Thus α−1 Q (X) ∈ R [X],
α ∈ R, and P (X) ∈ R [X], so P (X) also divides A (X).

Corollary 9.2.4. If K is a UFD, a field, or a PID, then K [X1 , . . . , Xn ] is a UFD for any n.
Warning that quotients of UFDs are only rarely UFDs themselves.
√ 
Example. Z [X] is a UFD, but Z [X] / X 2 + 5 = Z −5 is not a UFD.

42
M3P8 Algebra III 9 Polynomial rings in several variables

9.3 Irreducible polynomials

Lecture 19
A question is how can we test if P (X) ∈ K [X] is irreducible? We will now use the results of the previous Friday
section to obtain criteria for proving polynomials are irreducible. We begin with some trivial observations. 16/11/18
Lemma 9.3.1. Let K be any field, and P (X) ∈ K [X] of degree two or three. Then P (X) is irreducible if
and only if P (X) has no root in K.
Proof. Any nontrivial factor of P (X) would have to have degree one or two. Either way, if P (X) is reducible
it must have a linear factor.
Slightly less trivially, if K is finite there is a necessary and sufficient criterion for irreducibility. Let K = Fq
be a finite field with q = ps elements.
r
Lemma 9.3.2. X q − X is the product of P (X) ∈ Fq [X] irreducible, monic of degree dividing r.
Proof. Let P (X) be irreducible monic of degree d | r. Consider K (α), where α is a root of P (X). Thus
d r r
K (α) has order q d . So αq = α. Since d | r, αq = α. So α is a root of X q − X. But P (X) is the minimal
polynomial of α, so r
P (X) | X q − X.
2 r
Suppose P (X) | X q − X. Write
r 2
X q − X = P (X) Q (X) .
Take derivatives,
2
−1 = 2P (X) P 0 (X) Q (X) + P (X) Q0 (X) .
r
Since P (X) - −1, this is impossible. Finally, let P (X) ∈ K [X] irreducible be a divisor of X q − X. Let
r
K 0 = Fqr contain K. X q − X factors into linear factors over K 0 . So there exists α ∈ K 0 such that P (α) = 0.
P (X) is the minimal polynomial of α over K, so have an injection
K [X]
−→ K0
hP (X)i .
X 7−→ α

n
Order of K [X] / hP (X)i is q deg(P ) and order of K 0 is q r , so q r = q deg(P ) , so deg (P ) | r.
Corollary 9.3.3. Let P (X) in Fq [X] have degree d. Then P (X) is irreducible if and only if the greatest
r
common divisor of P (X) and X q − X is one for all 1 ≤ r < d.
r
Proof. If the polynomial P (X) is irreducible, it does not divide X q − X for r < d. Conversely, if P (X) is
reducible, there exists an irreducible polynomial Q (X) of degree 0 < r < d such that Q (X) | P (X), and
r
then Q (X) | X q − X in Fq [X].
Having obtained a satisfactory criterion for finite fields, the next simplest case to look at this that of Q [X].
This is already much more complicated. We will take advantage of the fact that Z [X] lives inside Q [X]. In
fact, all of our tricks will work in the following more general situation. R is a UFD with field of fractions
K, and we consider polynomials over K [X]. As we have seen, irreducibility over K is closely related to
irreducibility in R [X]. Let P (X) be a polynomial in K [X] and d = deg (P ). We can multiply P (X) by
scalars without substantially changing its factorisation, so we can assume that P (X) is monic. In general
there might be denominators in the coefficients of P (X), but note that for any r ∈ R, if
P (X) = c0 + · · · + cd−1 X d−1 + X d ,
then define a polynomial Qr (X) by
 
d X
Qr (X) = r P = c0 rd + · · · + cd−1 rX d−1 + X d .
r
It is easy to see that Qr (X) is irreducible in K [X] if and only if P (X) is. Moreover, we can choose r so that
Qr (X) has coefficients in R. We are thus reduced to the problem of deciding whether a monic polynomial
with coefficients in R is irreducible in K [X]. Moreover, we have shown that such a polynomial Qr (X) is
irreducible in K [X] if and only if it is irreducible in R [X]. Therefore a question is given Q (X) monic in
R [X], how can we prove or test irreducibility? We therefore get the following nice criterion for irreducibility.

43
M3P8 Algebra III 9 Polynomial rings in several variables

Proposition 9.3.4. Let Q (X) be a monic polynomial in R [X], and let p be a prime ideal of R. Suppose
that the mod p reduction Q (X) is irreducible in R/p [X]. Then Q (X) is irreducible in R [X].
Proof. Suppose Q (X) were reducible in R [X]. Since Q (X) is monic, Q (X) must factor as

A (X) B (X) ,

where both A (X) and B (X) are not units. Can assume A (X) , B (X) are monic of degree deg (A) , deg (B) >
0, since leading coefficients of A, B multiply to one. Then Q (X) factors in R/p [X] as

A (X) B (X) ,


where both are monic of positive degree between 1 and deg Q (X) − 1, so Q (X) is also reducible.
This means, for instance, that we can show that a monic polynomial in Z [X] is irreducible if we can find
even one prime p for which it is irreducible mod p.
Example.
X 2 + aX + b ∈ Z [X] ,
with a, b odd is irreducible in Q [X]. Its reduction mod 2 is

X 2 + X + 1,

which is irreducible in F2 [X].

Unfortunately, even when the polynomial is irreducible we will not always be able to do this.
Example. The polynomial X 4 + 1 is irreducible in Z [X] and Q [X], but reducible mod p for every p. You
can prove this with some elementary number theory.
4
X 4 + 1 = (X + 1)
2
in F2 [X]. If p is odd, X 4 + 1 has a common factor, and in fact divides X p − X. In fact
2
−1
X4 + 1 | Xp − 1.

If p = 2k + 1,
p2 = 4k 2 + 4k + 1 = 4 k 2 + k + 1 = 8m + 1,

since k 2 + k is even. 2
−1
Xp − 1 = X 8m − 1 = X 4 + 1 X 8m−4 − · · · − 1 .

There is another sufficient criterion for irreducibility by reducing mod p, known as Eisenstein’s criterion.
Proposition 9.3.5 (Eisenstein’s criterion). Let

Q (X) = a0 + · · · + an−1 X n−1 + X n

be a monic polynomial in R [X], and let p be a prime ideal of R. Suppose that

• for 0 ≤ i ≤ n − 1, ai ∈ p, and
/ p2 .
• a0 ∈

Then Q (X) is irreducible in R [X].

44
M3P8 Algebra III 9 Polynomial rings in several variables

Proof. Suppose Q (X) is reducible. Then we can write

Q (X) = A (X) B (X) ∈ R [X] ,
with A (X) and B (X) monic of positive degree less than deg (Q (X)). Reducing mod p we find that
R
Q (X) = X n = A (X) B (X) ∈ [X] .
p
In particular, since R/p is an integral domain, one of A (0) or B (0) is zero, say A (0) = 0. Write
A (X) = X d S (X) , S (0) 6= 0.
Degree d term of A (X) B (X) is S (0) B (0). d < n, so
S (0) B (0) = 0.
So B (0) = 0, so both A (0) = B (0) = 0. But then the constant terms A (0) and B (0) of A (X) and B (X)
both lie in p, so the constant term
a0 = Q (0) = A (0) B (0)
of Q (X) = A (X) B (X) must lie in p2 , contradicting our assumptions.
Corollary 9.3.6. X 4 + 1 is irreducible.
4
Proof. X 4 + 1 is irreducible if and only if (X + 1) + 1 is irreducible.
4
(X + 1) + 1 = X 4 + 4X 3 + 6X 4 + 4X 2 + 2
satisfies Eisenstein’s criterion mod 2.
Lecture 20
Monday
Example. Let F [X, Y, Z] for F field be a polynomial ring, such as Z [X, Y, Z]. Can write
19/11/18
F [X, Y, Z] = F [X, Y ] [Z] , R = F [X, Y ] ,
or
F [X, Y, Z] = F [X, Z] [Y ] = F [Y, Z] [X] .
Can also think of
F (X, Y, Z) ⊆ F (X) [Y, Z] = F (X) [Y ] [Z] , R = F (X) [Y ] .
• Let
P (X, Y ) = X 4 + X 2 Y 2 + Y 2 + XY ∈ C [X, Y ] .
Take R = C [X] and K = C (X). P (X, Y ) is quadratic in Y with coefficients in R,
P (X, Y ) = X 2 + 1 Y 2 + X · Y + X 4 .

GCD of coefficients is one, so it is irreducible if and only if it is irreducible in K [Y ], if and only if it

has no root in K, if and only if its discriminant is not a square in K. Discriminant is
X 2 − 4X 4 X 2 + 1 = X 2 − 4X 6 − 4X 4 = X 2 1 − 4X 4 − 4X 2 ,

which is not a square, so P (X, Y ) is irreducible.

•
P (X, Y, Z) = Z 5 + X 3 Y 4 Z + 2X 2 Y Z 3 − XY Z + Y 3 ∈ C [X, Y ] [Z]
is irreducible if it is irreducible mod X.
P (Y, Z) = Z 5 + Y 3 ∈ C [Z] [Y ]
is irreducible if and only if it is irreducible in C (Z) [Y ], if and only if it has no root, if and only if −Z 5 is
not a cube in C (Z), if and only if −Z 5 is not a cube in C [Z], which is clear from unique factorisation.

45
M3P8 Algebra III 10 Integral extensions and algebraic integers

10 Integral extensions and algebraic integers

10.1 Integral extensions
Definition 10.1.1. α ∈ C is an algebraic integer if there exists a monic polynomial P (X) ∈ Z [X] such
that P (α) = 0.
Note. If Q (X) is the minimal polynomial of α over Q, monic, then Q (X) | P (X). By Gauss’ lemma, there
exists α ∈ Q∗ such that αQ (X) ∈ Z [X] and αQ (X) | P (X) in Z [X]. Since Q (X) is monic, α ∈ Z. Since
P (X) is monic, α | 1. So α = ±1 and Q (X) ∈ Z [X].

Definition 10.1.2. Let R be a subring of a ring S, and α an element of S. We say α is integral over R if
there exists a monic polynomial P (X) ∈ R [X] with coefficients in R such that P (α) = 0 in S.
We can characterise integral elements in the following way.
Proposition 10.1.3. An element α ∈ S is integral over R if and only if the subring R [α] of S is a finitely
generated R-module.

Proof. Suppose α is integral over R, so that there exists a monic polynomial P (X) in R [X] with P (α) = 0.
Then R [α] is a quotient of R [X] / hP (X)i so 1, . . . , αd−1 , where d is the degree of P (X), span R [α] over
R. Given x ∈ R [α], can write
x = Q (α) = rn αn + · · · + r0 .
Write
Q (X) = P (X) T (X) + A (X) ∈ R [X] , deg (A (X)) < deg (P (X)) ,
so Q (α) = P (α) T (α) + A (α).

A (X) = a0 + · · · + ad−1 X d−1 , d = deg (P (X)) , ai ∈ R,

so
x = Q (α) = A (α) = a0 + · · · + ad−1 αd−1 .
Conversely, if R [α] is finitely generated as an R-module, say by x1 , . . . , xr ∈ R [α], we can write

xi = Qi (α) , Qi (X) ∈ R [X] .

Let n be larger than the degree di of all the Qi (X). We can write αn ∈ R [α] as
r
X r
X
si xi = si Qi (α) , si ∈ R.
i=1 i=1

So let
r
X
P (X) = X n − si Qi (X) ∈ R [X] .
i=1

Then P (X) is a monic polynomial with coefficients in R such that P (α) = 0.

Definition 10.1.4. Let R be a subring of S. We say that S is integral over R if every element of S is
integral over R.
Proposition 10.1.5. Suppose R is a Noetherian ring, and S is a ring containing R that is finitely generated
as an R-module. Then S is a Noetherian ring and is integral over R.
Proof. Let α ∈ S. The ring R [α] is an R-submodule of S, so it is finitely generated as an R-module, so α is
integral over R. Every ideal of S is an R-submodule of S, thus finitely generated as an R-module since R is
Noetherian, and hence also finitely generated as an S-module, so S is a Noetherian ring.

46
M3P8 Algebra III 10 Integral extensions and algebraic integers

Lemma 10.1.6. Let R ⊆ S ⊆ T be rings, such that S is finitely generated as an R-module and T is finitely
generated as an S-module. Then T is finitely generated as an R-module.
Proof. Let t1 , . . . , tl generate T over S, and let s1 , . . . , sm generate S over R. Then for any element t of T ,
we can write
Xl
t= ai ti , ai ∈ S.
i=1

We can further write

m
X
ai = bji sj , bji ∈ R,
j=1

so that
l X
X m
t= bji sj ti ,
i=1 j=1

so that T is generated over R by the elements si tj .

Corollary 10.1.7. Let R ⊆ S ⊆ T , with R Noetherian. If T is integral over S and S is integral over R,
then T is integral over R.
Proof. Let t ∈ T . Then t satisfies a polynomial

P (X) = X n + sn−1 X n−1 + · · · + s0 ∈ S [X] , si ∈ S,

such that P (t) = 0. Consider the subring

S 0 = R [s0 , . . . , sn−1 ] ⊆ S.

Since each si is integral over R, s0 is in particular integral over R and si is integral over R [s0 , . . . , si−1 ].
Thus R [s0 ] is a finitely generated R-module and R [s0 , . . . , si ] is a finitely generated R [s0 , . . . , si−1 ]-module
for each i by induction. By Lemma 10.1.6 above, S 0 is a finitely generated R-module. Since t is integral
over S 0 , S 0 [t] is a finitely generated S 0 -module, and hence a finitely generated R-module by Lemma 10.1.6.
Since R [t] is contained in S 0 [t] and R is a Noetherian ring, R [t] is a finitely generated R-module and thus t
is integral over R.
Corollary 10.1.8. Let R be a Noetherian subring of S and suppose α, β ∈ S are integral over R. Then αβ
and α + β are integral over R.
Proof. The ring R [α] is a finitely generated R-module and thus integral over R. Since β is integral over R
it is integral over R [α]. Thus R [α, β] = R [α] [β] is integral over R [α] and hence over R by Lemma 10.1.6.
Since α + β and αβ lie in R [α, β] they are integral over R.
Lecture 21
Definition 10.1.9. Let R be a Noetherian subring of S. The integral closure of R in S is the subset Wednesday
of S consisting of all elements s ∈ S that are integral over R. This is a subring of R. We say that R is 21/11/18
integrally closed in S if every element in S that is integral over R is contained in R, so R is equal to its
integral closure in S. If R is an integral domain, we say that R is integrally closed if R is integrally closed
in its field of fractions K.
Lemma 10.1.10. Let R be a Noetherian subring of S, and let R0 be the integral closure of R in S. Then
R0 is integrally closed in S.
Proof. Let t be an element of R integral over R0 . Then R0 [t] is a finitely generated R0 -module, so is integral
over R0 , and R0 is integral over R. Thus R0 [t] is integral over R, so t is integral over R and thus lies in
R0 .
√  √  √
Example. Z −3 is integral over Z. As a Z-module, Z −3 is generated by 1, −3. It is not integrally
√
closed. 1+ 2 −3 is in the field of fractions of Z −3 and is a root of X 2 − X + 1.
√ 

47
M3P8 Algebra III 10 Integral extensions and algebraic integers

Theorem 10.1.11. Let R be a UFD. Then R is integrally closed.

Proof. Let K be the field of fractions of R, and suppose α ∈ K is integral over R. Want α ∈ R. Then
there exists a monic polynomial P (X) in R [X] with coefficients in R such that P (α) = 0. Then (X − α)
is an element of K [X] dividing P (X). By Gauss’ lemma there is a λ ∈ K ∗ such that λ (X − α) is in R [X]
and divides P (X) in R [X]. Clearly λ must lie in R, and on the other hand divide the leading coefficient of
P (X), which is one. Thus λ ∈ R∗ is a unit, so since (X − α) ∈ R [X] we must have α ∈ R.
This suggests to number theorists to take an extension K/Q finite, and let OK , the ring of integers of K, be
the integral closure of Z in K.
Note. For x, y ∈ Q,
√
p  √

Q x =Q y2 x = Q y x .
√ 
We now focus on a specific class of examples. Let d ∈ Z be squarefree and let K = Q d . This is precisely
the set of elements of K that are integral over Z. That is, that satisfy
√ a monic polynomial with integral
coefficients. What is OK ? Every element of K is of the form a√+ b d for a, b ∈ Q. A question is when is
this an algebraic integer? Need the minimal polynomial of a + b d to have integer coefficients. We have the
following lemma.
Lemma 10.1.12. Let α ∈ K and suppose α is integral over Z. Then the minimal polynomial of α, taken to
be monic, has integer coefficients.
Proof. Let Q (X) be the minimal polynomial of α, normalised so it is monic. Since α is integral over Z, there
is a monic polynomial P (X), with integer coefficients, such that P (α) = 0. Then Q (X) divides P (X) in
Q [X]. By Gauss’ lemma, there exists β ∈ Q∗ such that βQ (X) has integer coefficients and divides P (X)
in Z [X]. Since Q (X) is monic, β lies in Z. Since βQ (X) divides P (X) we see that β divides one, by
comparing leading coefficients, so β is a unit and Q (X) lies in Z [X].
√
Let α = a + b d with a, b ∈ Q. Then the minimal polynomial of α over Q is
  √    √ 
X − a − b d = X 2 − 2aX + a2 − b2 d .


X − a+b d

Thus α is an algebraic integer if and only if 2a, a2 − b2 d ∈ Z.

• Suppose this is the case, and that a ∈ Z. Then b2 d ∈ Z. Suppose b ∈ / Z. There exists a prime p
dividing denominator of b in lowest terms. Since b2 d ∈ Z must have p2 | d but we took d squarefree.
So b ∈ Z.
• On the other hand, suppose that a = m/2 where m is odd. Then if a2 −b2 d ∈ Z we have m2 /4−b2 d ∈ Z
and so m2 − 4b2 d is a multiple of four. So 4b2 d = x where m2 − x ∈ Z is a multiple of four. Since
2
(2k + 1) = 4k 2 + 4k + 1 ≡ 1 mod 4,

so m2 ≡ 1 mod 4, this can only happen if d is odd and b = n/2 with n odd. We then have m2 − n2 d
is a multiple of four. Since m2 , n2 ∈ Z are odd they are congruent to 1 mod 4, so this is only possible
if d is congruent to 1 mod 4.
√ √
Thus if α is an algebraic integer, either α = a + b d with a, b ∈ Z, or α = m+n 2
d
with m, n ∈ Z odd and
d congruent to 1 mod 4.
√  Conversely, it is easy to check that all such elements are algebraic integers. To
summarise, if K = Q d for d squarefree, we thus have
 h√ i n √ o
Z d = a + b d | a, b ∈ Z d ≡ 2, 3 mod 4
OK = h √ i n √ o .
Z 1+ d = m+n d | n, m ∈ Z, n ≡ m mod 2 d≡1 mod 4
2 2

Note. The results in this section are also true without any Noetherian hypotheses, but the proofs are more
difficult, and require machinery we have not covered.

48
M3P8 Algebra III 11 Dedekind domains

11 Dedekind domains
Lecture 22
11.1 Dedekind domains Friday
23/11/18
For number theorists, it is often convenient to work in a ring of the form Z [α], where α ∈ C is an algebraic
integer, or more generally in some subring O of C that is integral over Z. Unfortunately, unique factorisation
only rarely holds in such rings. If O is integrally closed, however, there is a substitute for unique factorisation
that is often good enough, unique factorisation of ideals. In this section we develop the ideas behind this
result, in the more general context of what are called Dedekind domains.
Definition 11.1.1. An integral domain R is called a Dedekind domain if
• R is Noetherian,
• R is integrally closed, and
• every nonzero prime ideal of R is maximal, so R has dimension one.
Example.
• In particular, any PID is a Dedekind domain. We have seen that every nonzero prime ideal is maximal
in a PID, and PIDs are certainly Noetherian. They are integrally closed because any UFD is integrally
closed.
• The rings OK , the integral closure of Z in K, with K a quadratic extension of Q are
 h√ i
Z d d ≡ 2, 3 mod 4
OK = h √ i .
1+ d
Z
2 d ≡ 1 mod 4

OK ⊆ K is an integral domain. OK is finitely generated by a single element as a Z-algebra. It is thus

Noetherian. OK is also integrally closed. We proved on example sheet 2 that Z [X] / hP (X)i for P (X)
monic and irreducible with coefficients in Z has dimension one, that is every nonzero prime of such a
ring is maximal.
More generally, we have the following.
Theorem 11.1.2. Let R be a PID with field of fractions K, and let L be a finite extension of K. Let S be
the integral closure of R in L. Then S is a Dedekind domain.
We will prove this later in the course, under a mild additional hypothesis on the extension L/K. In particular,
let R = Z and K = Q. The ring of integers OL in L/Q a finite extension is a Dedekind domain. Also in
particular let R = F [X] for F a field, K = F (X), and L/K finite. Integral closure of R in K is also
Dedekind.
Example.
K [X, Y ]
,
hY 2 − X 3 − aX − bi
where X 3 + aX + b is a squarefree polynomial in K [X, Y ].
The reason Dedekind domains are interesting to us is that the nonzero ideals in a Dedekind domain factor
uniquely as products of prime ideals. The idea to study factorisation of ideals into prime ideals comes from
the following observation.
Lemma 11.1.3. Let p be a prime ideal of any ring R, let I, J ⊆ R be ideals, and suppose that p contains
IJ. Then either p contains I or p contains J.
Proof. Suppose that p does not contains I, and fix an r ∈ I such that r is not in p. Then for all s ∈ J, the
product rs lies in IJ and hence in p. Since r does not lie in p, and p is prime, we must have s ∈ p.

49
M3P8 Algebra III 11 Dedekind domains

Note the resemblance of this to the property, p | ab implies p | a or p | b for p irreducible, which holds in UFDs
and implies unique factorisation. We might hope that the above result thus implies unique factorisation into
primes for arbitrary rings, but this is too much to ask for. The problem is that ideal multiplication is usually
badly behaved compared to multiplication of elements in integral domains. Recall that for I, J ⊆ R ideals,
IJ is the ideal generated by all elements of the form rs for r ∈ I and s ∈ J. In particular if r1 , . . . , rn
generate I and s1 , . . . , sm generate J then
r1 s1 , . . . , r1 sm , . . . , rn s1 , . . . , rn sm
generate IJ.
√ 
Example. R = Z −3 is not a Dedekind domain, since it fails to be integrally closed. Then the ideal
√
I = 2, 1 + −3 is prime, and we have
√ 2 √ √ √
2, 1 + −3 = 4, 2 + 2 −3, −2 + 2 −3 = 4, 2 + 2 −3 .
There is thus a chain of inclusions
√ √
2, 1 + −3 = I ) h2i ) I 2 = 4, 2 + 2 −3 ,
so the ideal h2i is not a product of prime ideals. Worse is I 2 = h2i I but I 6= h2i, so cannot cancel ideals.
Dedekind domains give precisely the context where this does not happen. In order to make this precise, we
first define the following.
Definition 11.1.4. Let R be a Noetherian integral domain. A fractional ideal of R is a finitely generated
nonzero R-submodule of the field of fractions K of R. A principal fractional ideal is an R-submodule
R · x ⊆ K of K finitely generated by a single nonzero element x for x ∈ K ∗ .
Example. The subgroup of Q generated by 3/5 is a principal fractional ideal of Z. Indeed, every fractional
ideal of Z, or any PID, is principal.
More generally, let R be a Noetherian integral domain, and let I be the R-submodule of K generated by
r1 , . . . , rn ∈ K. Then by definition I is a fractional ideal of R. On the other hand, we can clear denominators.
There exists an r ∈ R, nonzero, such that rri lies in R for all i. Then rI is generated by elements of R, so
is an ideal J of R, and I = 1r J. Thus the fractional ideals of R are precisely the subsets of K of the form
1
r J, where r is a nonzero element of R and J is an ideal of R. Let I and J be fractional ideals of R. The
product IJ is the R-submodule of K generated by all products of the form rs for r ∈ I and s ∈ J. It is a
fractional ideal of R. The multiplication I, J 7→ IJ is an associative and commutative operation.
Note. R is a fractional ideal of R, and RJ = J for any fractional ideal J, so R is an identity element for this
operation.
For a nonzero ideal I of R, let I −1 denote the set
{r ∈ K ∗ | rI ⊆ R} .
Then I −1 is clearly an R-submodule of K. If r ∈ I is nonzero, then rI −1 , by definition, is contained in R,
so I −1 is contained in 1r · R and is thus a fractional ideal. Warning that in a general ring, I 7→ I −1 is not
always a good inverse operation. II −1 ⊆ R but need not equal R. For a prime ideal p of R, and n ∈ Z>0 ,
−n −1 n


define p = p . We then have the following.
Theorem 11.1.5. Let R be a Dedekind domain. Then
• the set of fractional ideals of R form a group under multiplication I, J 7→ IJ, identity R, and inverse
I 7→ I −1 , and
• moreover, any fractional ideal I of R factors uniquely as
pn1 1 . . . pns s , ni ∈ Z,
where the pi are nonzero prime ideals.

50
M3P8 Algebra III 11 Dedekind domains

The proof of this statement will occur in several steps. We first show the following. Lecture 23
Monday
Proposition 11.1.6. Let I be a nonzero ideal of a Noetherian ring R. Then there exist nonzero primes 26/11/18
p1 , . . . , pr and n1 , . . . , nr ∈ Z>0 such that I contains

pn1 1 . . . pnr r .

Proof. Suppose the claim fails for some I. Then there exists an ideal I such that
• I does not contain a product of primes, but
• every ideal containing I does.
Suppose not. Then let I0 be an ideal satisfying 1. Since 2 does not hold for I0 there exists I1 ) I0 such that
1 holds for I1 . Then 2 cannot hold for I1 , so there exists I2 ) I1 such that 1 holds for I2 , etc, so get infinite
increasing chain
I0 ( I1 ( . . . ,
contradicting Noetherianness of R. Fix such an I. Certainly I cannot be prime. So there exist a, b ∈ R with
ab ∈ I but a and b not in I. Then the ideals I + hai and I + hbi both strictly contain I, so the claim holds
for both of these ideals by 2, so

I + hai ⊇ p1 . . . pr , I + hbi ⊇ q1 . . . qs ,

for pi and qj prime ideals. Then it also holds for their product

(I + hai) (I + hbi) = I 2 + hai I + hbi I + habi ,

but this product is contained in I. Thus

p1 . . . pr q1 . . . qs ⊆ (I + hai) (I + hbi) ⊆ I

as well and we have a contradiction to 1.

Note. If the claim holds for an ideal I then it holds for any ideal containing I, and that if the claim holds
for I and J then it holds for I ∩ J.
Next, we show that prime ideals have multiplicative inverses. To do so we use the following lemma.
Lemma 11.1.7. Let R be a Dedekind domain with field of fractions K, and let x be an element of K that
is not in R, and let I be any nonzero ideal of R. Then xI is not contained in I.
Proof. Suppose xI were contained in I. Let a ∈ I, and for each i let Mi be the ideal of I generated by
a, . . . , axi , so
Mi = a, . . . , axi ⊆ I.
This is an increasing tower of ideals of R. In particular, since R is Noetherian, it is eventually constant, that
is Mi+1 = Mi for some i. Then axi+1 can be expressed as an R-linear combination of the axj , that is there
exist r0 , . . . , ri ∈ R such that we have
Xi
axi+1 = rj axj .
j=0

Since a 6= 0 and R is an integral domain we can cancel the a, so

i
X
xi+1 = rj xj ,
j=0

so x satisfies a monic polynomial with coefficients in R. Thus x is integral over R. Since R is integrally
closed and x does not lie in R this is a contradiction.
When R is not integrally closed this is false.

51
M3P8 Algebra III 11 Dedekind domains

√
1+ −3
√ 
Example. If R = Z −3 , I ⊂ R, and x = 2 , then
√ √ √
xI = 1 + −3, −1 + −3 = 2, 1 + −3 = I.

Proposition 11.1.8. Let p be a nonzero prime ideal of a Dedekind domain R. Then p−1 · p = R.
Proof. We first show that there is an element x ∈ p−1 such that x ∈
/ R. Let a be an element of p, and a 6= 0,
so that we have hai ⊂ p. Choose a minimal set of nonzero primes p1 , . . . , pr such that

p1 . . . pr ⊆ hai .

Then we have in particular

p1 . . . pr ⊆ p,
so by Lemma 11.1.3 above we must have p = pi for some i. Without loss of generality we can take i = 1.
Then by our minimality assumption p2 . . . pr is not contained in hai. Take b to be an element of p2 . . . pr
that is not in hai. Then b/a ∈ K but not in R. Take x = b/a. On the other hand for any y ∈ p,

by
xy = , by ∈ p1 . . . pr ⊆ hai .
a
Thus xy lies in R. By definition, this means x lies in p−1 but not in R. Now consider p−1 · p. By definition
this is contained in R. Since p ⊆ R, 1 ∈ p−1 so p−1 · p contains p. Since p is a nonzero prime ideal it is
maximal, so we must have either p−1 · p = R or p−1 · p = p. Suppose the latter holds. Then in particular
multiplication by x sends p to p. This contradicts Lemma 11.1.7 above, so p−1 · p ) p.

Proposition 11.1.9. Let I be a nonzero ideal of a Dedekind domain R. Then there exists a fractional ideal
J of R such that IJ = R.
Proof. Suppose otherwise. Then there is a maximal nonzero ideal I of R for which no such J exists.
Proposition 11.1.8 shows that I is not a maximal ideal, so I is properly contained in some maximal ideal p
of R. Then p−1 is contained in I −1 . We thus have inclusions

I ⊆ Ip−1 ⊆ II −1 ⊆ R.

Suppose that Ip−1 = I. By Proposition 11.1.8 there exists x ∈ p−1 not in R, so we would have xI ⊂ I
contradicting Lemma 11.1.7 above. Thus Ip−1 strictly contains I and thus has an inverse

−1
J = Ip−1 .

Then Ip−1 ·J = R. But then I·p−1 J = R, so p−1 J ⊆ I −1 . Then II −1 ⊇ R. So II −1 = R, a contradiction.

Theorem 11.1.10. Let R be a Dedekind domain. Then the fractional ideals of R form a group under
multiplication.
Proof. We must show that every fractional ideal of R is invertible. Let I be such a fractional ideal of R.
Then there is r ∈ R such that rI is an ideal of R. Proposition 11.1.9 shows that rI has a multiplicative
inverse J, so I = 1r J −1 . Then I −1 = rJ is a multiplicative inverse for I, since

II −1 = 1r J −1 · rJ = JJ −1 = R.

Lecture 24
Lecture 24 is a problem class. Wednesday
28/11/18

52
M3P8 Algebra III 11 Dedekind domains

It remains to show that every fractional ideal of R factors uniquely as a product of prime powers. The hard Lecture 25
part is showing such factorisations exist, and we make heavy use of the fact that the fractional ideals are a Friday
group. Uniqueness is then almost an afterthought. 30/11/18

Proposition 11.1.11. Every fractional ideal in a Dedekind domain R is uniquely expressible as a product
of, possibly negative, prime powers
pn1 1 . . . pns s , ni ∈ Z,
where the pi are primes.
Proof. We first show that every nonzero ideal I in R is a product of nonnegative prime powers.
• Suppose otherwise, that for some ideal I of R, I cannot be expressed as a product of primes. Claim
that there is a largest ideal I 0 such that I 0 cannot be expressed as a product of primes but all J ) I
can. Take I0 = I if there exists I1 ) I that cannot be expressed as a product of primes, either all ideals
properly containing I1 can be or there exists I2 ) I1 that cannot be expressed as a product of primes.
Since R is Noetherian, the process terminates. Now let I 0 be as in the claim. Certainly I 0 6= R, and I 0
is not prime, since every maximal ideal of R is certainly such a product I 0 cannot be a maximal ideal.
Thus I 0 is properly contained in a maximal ideal p. Then

J = p−1 · I 0

is an ideal of R. Since p ⊇ I 0 , p−1 · p ⊇ J, so J ⊆ R. Since the nonzero fractional ideals of R form a

group this ideal J strictly contains I 0 and thus factors as a product of prime powers

p−1 · I 0 = J = pn1 1 . . . pns s .

But then
p · J = I 0 = ppn1 1 . . . pns s
is also a product of prime powers, contradicting our assumption.
• Now suppose that I is a fractional ideal. Then I = 1r J for some nonzero ideal J of R and some nonzero
element r of R. Since
hri = qm mt
1 . . . qt ,
1
J = pn1 1 . . . pns s
factor as products of prime powers, so does
−1
I = 1r J = hri J = pn1 1 . . . pns s q−m
1
1
. . . qt−mt .

It remains to show that such factorisations are unique. Suppose otherwise for a fractional ideal I. Then we
have a finite collection of distinct primes p1 , . . . , ps and q1 , . . . , qt , and two sequences

n1 , . . . , ns , m1 , . . . , mt ∈ Z,

such that
I = pn1 1 . . . pns s = qm mt
1 . . . qt ,
1

and we must show that mi = ni for all i. Suppose this is not the case. We can make all prime powers ni , mj
m
involved positive by cancelling pni i and qj j from both sides of the equation. We then get an expression of
the form
pn1 1 . . . pns s = qm mt
1 . . . qt ,
1

where the primes {pi } , {qj } are all distinct and all powers ai and bj are positive. Claim that both products
must be empty under these assumptions. Recall that if R is Noetherian, p prime, I, J ideals, then p contains
IJ implies that p contains I or p contains J. If one product, say the pi ’s, is nonempty, then since p1 divides
the left hand side it also divides the right hand side, and thus contains one of the qi ’s for some i. Since pi
and qj are maximal in a Dedekind domain, p1 = qi , which contradicts disjointness and is impossible.

53
M3P8 Algebra III 11 Dedekind domains

11.2 Ideal class groups

Let R be a Dedekind domain. Then the fractional ideals of R form a group, which we will denote I (R).
The principal fractional ideals are a subset of I (R) that is easily seen to be closed under multiplication and
inverses. If r, s ∈ K ∗ , then

−1 1
(rR) R
=
r
(rR) (sR) = rsR.

Denote this subgroup by P (R). We can then form a quotient group called the ideal class group of R.
Definition 11.2.1. The ideal class group of R, denoted A (R), is

I (R)
A= .
P (R)

Example. If R is a PID, A (R) = {e}.

In general A (R) is a measure of the failure of fractional ideals of R to be principal. That is, it measures the
failure of R to be a PID. We will show that if K is a finite extension of Q then the integral closure OK of
Z in K is a Dedekind domain. A fundamental result of algebraic number theory, which we will not prove, is
the following.
Theorem 11.2.2. If R is a Dedekind domain with field of fractions a finite extension of K, that is R = OK
for K/Q finite, then the ideal class group A (R) is a finite group.
The order of the ideal class group of OK is called the class number of K. In particular, if A (R) is finite,
say of order n, then pn is principal for every prime p. Warning that it is not true that A (R) is finite for
R Dedekind. The study of class groups and class numbers is a central part of modern number theory and
there are many, many open questions.
Example. If
C [x, y]
R= ,
hy 2 − x (x − 1) (x + 1)i
A (R) is uncountable.

54
M3P8 Algebra III 12 Integers in number fields

12 Integers in number fields

12.1 Integer rings
At one point I claimed that if R is a PID, K its field of fractions, L/K finite, then the integral closure S of R
in L is Dedekind. In fact, only need R is Dedekind. We will prove under a simplifying assumption, involving
the trace map. Let K be a finite extension of Q. Such an extension is called a number field. The integral
closure OK of Z in K is called the ring of integers of K. A fundamental result of number theory is that
OK is a Dedekind domain. The goal of this section is to prove this fact. Indeed, we will prove something
more general, but in order to do that we need to introduce some new concepts.

12.2 Trace and norm

Let L/K be a finite extension of fields, and let α be an element of L. Then we can regard L as a finite
dimensional K-vector space. Multiplication by α is then a K-linear map from L to L. If we choose a K-basis
β1 , . . . , βd for L, such a map
L −→ L
x 7−→ xα
is given by a d by d matrix Mα ∈ Md (K), with entries in K where d is the degree of L over K. That is,
(Mα )i,j is defined by
Xd
αβi = (Mα )i,j βj .
j=1

The matrix of course depends on the basis β1 , . . . , βd chosen, but its trace and determinant are elements of
K that depend only on α. We denote the trace of Mα by T rL/K (α) ∈ K and call it the trace of α with
respect to L/K. Similarly, the determinant of Mα is denoted NL/K (α) ∈ K and called the norm of α.

Lemma 12.2.1. The map

T rL/K : L −→ K
α 7−→ T rL/K (α)
is K-linear. If α, α0 ∈ L, and λ ∈ K, then

T rL/K (λα + α0 ) = λT rL/K (α) + T rL/K (α0 ) .

The map
NL/K : L −→ K
α 7−→ NL/K (α)
is multiplicative, so
NL/K (αα0 ) = NL/K (α) NL/K (α0 ) .

Proof. Since λ ∈ K, distributivity of multiplication over addition shows that, with respect to a fixed basis
of L over K, Mλα+α0 = λMα + Mα0 , so

T rL/K (λα + α0 ) = T r (Mλα+α0 ) = T r (λMα ) + T r (Mα0 ) = λT rL/K (α) + T rL/K (α0 ) .

Similarly Mαα0 = Mα Mα0 , by associativity of multiplication, so

NL/K (αα0 ) = det (Mαα0 ) = det (Mα ) det (Mα0 ) = NL/K (α) NL/K (α0 ) .

We will prove that if R is a PID or a Dedekind domain, such as Z, K is its field of fractions, L/K is finite
such that T rL/K is not the zero map, then the integral closure of R in L is Dedekind.

55
M3P8 Algebra III 12 Integers in number fields

Lecture 26
Proposition 12.2.2. Let L/K be a finite extension, and α an element of L. Let
Monday
Q (X) = X n + an−1 X n−1 + · · · + a0 03/12/18

be the minimal polynomial of α over K. Then

• T rL/K (α) = −dan−1 , and
n d
• NL/K (α) = ((−1) a0 ) ,
where d is the degree [L : K (α)] of L over K (α).
Proof.
• We first prove this when d = 1, so L = K (α) and n = [L : K]. Then we have seen 1, . . . , αn−1 is a
K-basis for L over K. With respect to this basis,
basis element ·α in terms of basis
1 α α
.. .. .. .
. . .
αn−2 αn−1 αn−1
αn−1 αn −an−1 αn−1 − · · · − a0
Mα has the matrix  
0 ... 0 −a0
1 ... 0 −a1 
Mα =  . .
 
.. .. ..
 .. . . . 
0 ... 1 −an−1
This matrix is called the companion matrix Cα of the polynomial Q (X).
n
T r (M ) = −an−1 , det (M ) = (−1) a0 ,
from which both claims can be easily deduced.
• In general,
K ⊆ K (α) ⊆ L.
Choose a basis β1 , . . . , βd for L over K (α). Then
β1 , . . . , β1 αn−1 , . . . , βd , . . . , βd αn−1
is a K-basis for L/K. With respect to this basis,
basis vector ·α in terms of basis
βi βi α βi α
.. .. .. .
. . .
n−2 n−1 n−1
βi α βi α βi α

βi αn−1 βi α n βi −an−1 αn−1 − · · · − a0
Mα is block diagonal, consisting of d blocks along the diagonal, each of which is the n by n matrix
above,  
Cα . . . 0
Mα =  ... .. ..  .

. . 
0 . . . Cα
d n d
T r (Mα ) = (T r (Cα )) · d = −dan−1 , det (Mα ) = (det (Cα )) = ((−1) a0 ) ,
so the claim follows.

56
M3P8 Algebra III 12 Integers in number fields

Remark 12.2.3. If L/K is finite, T rL/K (1) = [L : K], considered as an element of K. The map T rL/K :
L → K is sometimes the zero map. However, this does not happen if K has characteristic char (K) = 0 or
if the degree d = [L : K] is relatively prime to the characteristic of K, since the above Proposition 12.2.2
shows that T rL/K (1) = d 6= 0.
Example. Let K = F2 (t), the rational functions with coefficients in F2 .
 
K [X] 1
L= 2
= F 2 t2 .
hX − ti
T rL/K is K-linear. T rL/K (1) = 0 and T rL/K (X) = 0, so T rL/K (aX + b) = 0 for all a, b ∈ K.
Proposition 12.2.4. If L/K are finite fields, then T rL/K is not zero map.

12.3 The main result

We can now state our main result.
Theorem 12.3.1. Let R be a PID, or even a Dedekind domain, with field of fractions K, and let L/K be
a finite extension such that T rL/K is not the zero map. Let S be the integral closure of R in L. Then S is
a Dedekind domain.
To prove this, we must show three things about S, that S is Noetherian, that S is integrally closed, and that
S has dimension one, that is every nonzero prime ideal of S is maximal. We first show the following.
Lemma 12.3.2. The field of fractions of S is L.
Proof. In fact, we will show that every element of L can be expressed as s/r for s ∈ S and r ∈ R. Let x ∈ L,
and let
Q (X) = X n + an−1 X n−1 + · · · + a0 , ai ∈ K
be the minimal polynomial of x over K, where K is the field of fractions of R. We will show that there exists
r ∈ R∗ such that rx ∈ S. Then x = rx/r, so is in field of fractions of S. Let n be the degree of Q (X). For
each r ∈ R, let  
n X
Qr (X) = r Q = X n + ran−1 X n−1 + · · · + rn a0 .
r
We can find an r ∈ R such that Qr (X) has coefficients in R. But Q (x) = 0, so Qr (rx) = 0, so Qr (X) is
the minimal polynomial of rx, so it follows that for such r, rx is integral over R and thus lies in S.
Corollary 12.3.3. The ring S is integrally closed.
Proof. We have shown that the integral closure S of R in L is integrally closed in L. Since L is the field of
fractions of S, we have that S is integrally closed.
Next we show that S is Noetherian. In fact, we will show that S is a finitely generated R-module. Since R
is Noetherian it will then follow that S is Noetherian as an R-module, and hence also as an S-module. Then
every R-submodule of S is finitely generated over R, so every ideal of S is finitely generated as an R-module
and as an S-module. Thus S is Noetherian.

R K
.
S L
Lecture 27
To do this, choose a K-basis β1 , . . . , βd for L over K. We have seen that for each i there exists ri ∈ R∗ such Wednesday
that ri βi ∈ S, so, replacing βi by ri βi , we may assume that the βi all lie in S. Let M ⊆ S be the R-module 05/12/18
in S spanned by the βi , and let M ∗ denote

M ∗ = x ∈ L | ∀m ∈ M, T rL/K (xm) ∈ R ⊆ L.


57
M3P8 Algebra III 12 Integers in number fields

Claim that S ⊆ M ∗ . Since M ⊆ S, if x ∈ S, then xm ∈ S. So it suffices to show that for all s ∈ S,

T rL/K (s) ∈ R. Fix s, let Q (X) be the minimal polynomial of s over K. Since S is integral over R, s is
integral over R, so Q (s) has coefficients in R. If
Q (X) = X r + ar−1 X r−1 + · · · + a0 ,
last time showed that
T rL/K (s) = −nar−1 , n = [L : K (s)] .
This lies in R. We now have
M ⊆ S ⊆ M ∗.
Suffices to show M ∗ is finitely generated as an R-module.
Proposition 12.3.4. There exist β1∗ , . . . , βd∗ ∈ L such that
(
∗

1 i=j
T rL/K βi βj = .
0 i 6= j

Proof. Let A be the matrix where

Aij = T rL/K (βi βj ) ,
a d × d matrix with entries in K. If x = r1 β1 + · · · + rd βd , then
       
r1 r1 T r (β1 β1 ) + · · · + rd T r (β1 βd ) T r (β1 ) (r1 β1 + · · · + rd βd ) T r (β1 x)
A  ...  =  ..
=
..
=
..
.
       
. . .
rd r1 T r (βd β1 ) + · · · + rd T r (βd βd ) T r (βd ) (r1 β1 + · · · + rd βd ) T r (βd x)
If I have y1 , . . . , yd ∈ K, finding an element x = r1 β1 + · · · + rd βd ∈ L such that
T r (β1 x) = y1 , ..., T r (βd x) = yd ,
is equivalent to solving

T
T
A r1 ... rd = y1 ... yd ,
so βj∗ , if it exists, is
βj∗ = r1 β1 + · · · + rd βd ,
for r1 , . . . , rd a solution to

T
T
A r1 ... rd = 0 ... 1 ... 0 .
So suffices to show A is invertible. Suppose otherwise. Then there exists r1 , . . . , rd not all zero.

T
A r1 . . . rd = 0.
Then x = r1 β1 + · · · + rd βd ∈ L∗ is such that T rL/K (βi x) = 0 for all i. If this is true, can write y ∈ L as
b1 β 1 + · · · + bd β d .
T rL/K (xy) = b1 T rL/K (β1 x) + · · · + bd T rL/K (βd x) = 0.
So T rL/K (xy) = 0 for all y ∈ L. But x 6= 0. Setting y = x−1 z, we find T rL/K (z) = 0 for all z ∈ L. But we
assumed T rL/K 6= 0.
Corollary 12.3.5. M ∗ is a free R-module of rank d.
Proof. Claim that βj∗ ∈ M ∗ for all j, and the βj∗ generate M ∗ as an R-module. Every element of M can be
written as
r1 β1 + · · · + rd βd , ri ∈ R,
∗ ∗


so T rL/K βj m = rj ∈ R. Let x ∈ M . We can write
x = r1 β1∗ + · · · + rd βd∗ , ri ∈ K.
T rL/K (βi x) = ri , so ri ∈ R for all i.

58
M3P8 Algebra III 12 Integers in number fields

Corollary 12.3.6. S is a finitely generated R-module.

Proof. S is an R-submodule of the finitely generated R-module M ∗ , and, since R is Noetherian, S is therefore
finitely generated as an R-module.
Thus S is Noetherian. Now it remains to prove that every nonzero prime ideal p of S is maximal. Let p be
a nonzero prime ideal of S, and let s be an element of p. Consider the intersection q = p ∩ R. Then q is a
prime ideal of R. Claim that q = p ∩ R is a nonzero prime ideal of R. Let Q (X) be the minimal polynomial
of s over K, then s is integral, so Q (X) has coefficients in R. If

Q (X) = X r + ar−1 X r−1 + · · · + a0 , a0 6= 0,

where Q (X) is irreducible, we then have

0 = Q (s) = a0 + · · · + ar−1 sr−1 + sr ,

and thus lie in R. Rewriting, we get

−a0 = s a1 + · · · + ar−1 sr−2 + sr−1 .

In particular −a0 lies in the ideal generated by s, and hence in p. Moreover, since Q (X) is irreducible, a0
is a nonzero element of R, so q is nonzero since a0 ∈ q. Since R is Dedekind, thus q is a maximal ideal of R.
Have R ⊆ S → S/p, where kernel is R/q, a field. So get R/q → S/p. The ring S/p is an integral domain
containing the field R/q. Moreover, since S is a finitely generated R-module, if s1 , . . . , sr generate S as an
R-module, then they generate S/p as an R/q-module, so S/p is a finitely generated R/q-module. That is,
S/p is a finite dimensional R/q-vector space. We now show the following.
Lemma 12.3.7. Let K be an integral domain and let R be an integral domain containing K that is finite
dimensional as an K-vector space. Then R is a field.
Proof. Let r be a nonzero element of R. Have map

K [X] −→ R
.
x 7−→ r

Let I be the kernel. Since R is finite dimensional I = 0. Since R is an integral domain I is prime. So

K [X]
−→ R
I
x 7−→ r

is an injection, with K [X] /I a field. Then x has an inverse in K [X] /I, and this inverse maps to a
multiplicative inverse for r in R.
Lemma 12.3.7 shows that S/I is a field, so I is maximal. We have thus shown that S is Noetherian, integrally
closed, and that every nonzero prime ideal in S is maximal, so S is indeed a Dedekind domain. In particular,
for any finite extension K/Q, the integral closure OK of Z in K is a Dedekind domain, and thus has unique
factorisation of ideals. Another class of examples comes by taking K a field, letting L be a finite extension
of K (t) such that T rL/K(t) is nonzero, and letting R be the integral closure of K [t] in L. The field L is
called a function field, and the ring R is the ring of regular functions on a smooth affine algebraic
curve. Such rings R are also Dedekind domains, and they are of considerable interest in algebraic geometry.
They of course also have the unique factorisation property for ideals, and just like in ring of integers one can
consider the ideal class group. In this context, the ideal class group is also known as the Picard group.
It has a geometric interpretation in terms of line bundles on algebraic curves. Unlike in the number field
setting, the Picard group is often not a finite group.

59
M3P8 Algebra III 13 Introduction to algebraic geometry

13 Introduction to algebraic geometry

Lecture 28
The idea is Friday
• to study ideals in polynomial rings by studying geometry of the common zeros in ideal, and 07/12/18

• to study geometry of solutions to polynomial equations via ring theory.

13.1 Algebraically closed fields

Work best over algebraically closed fields.
Definition 13.1.1. Let K be a field. We say K is algebraically closed if every nonconstant polynomial
P (X) ∈ K [X] factors into linear factors.
In particular, the only irreducible polynomials are linear polynomials.
Theorem 13.1.2 (Fundamental theorem of algebra). The field C is algebraically closed.
We will not prove this in this course. Ultimately it requires some analysis. This is unsurprising, since the
construction of C is fundamentally an analytic one. We have the following characterisation of algebraically
closed fields.
Example.
• By contrast,
Q, R, Fpr , Fp (t)
are all not algebraically closed.
• Q is algebraically closed.
Lemma 13.1.3. A field K is algebraically closed if and only if every field extension L/K is either trivial,
that is L = K, or transcendental.
Proof. Suppose K is algebraically closed. Let L/K be an algebraic extension, and α ∈ L. Let P (X) be
the minimal polynomial of α over K. Then P (X) is irreducible, hence linear. But then α ∈ K, so L = K.
Conversely, if every field extension L/K is trivial or transcendental, then if P (X) is an irreducible polynomial
in K [X] we must have K [X] / hP (X)i = K, so P (X) must have degree one. Since every polynomial in
K [X] factors into irreducibles, K must be algebraically closed.

13.2 Affine algebraic sets

Fix an algebraically closed field K. In fact, it is harmless to take K = C throughout.
Definition 13.2.1. Let AnK , the affine n-space over K, denote the set K n of n-tuples of elements of K.
For S an arbitrary collection of elements of K [X1 , . . . , Xn ], we let Z (S) denote the subset

Z (S) = {(x1 , . . . , xn ) ∈ AnK | ∀P ∈ S, P (x1 , . . . , xn ) = 0} ⊆ AnK .

In other words, Z (S) is the set of common zeros of all the polynomials in S.
Example.
• S = {y − x (x − 1) (x + 1)} is y = x (x − 1) (x + 1).
• S = {y (y − x (x − 1) (x + 1))} is y = 0 or y = x (x − 1) (x + 1).
• S = {y, y − x (x − 1) (x + 1)} is y = 0 and y = x (x − 1) (x + 1).

• S = y 2 − x (x − 1) (x + 1) is connected.

60
M3P8 Algebra III 13 Introduction to algebraic geometry

Note that we have the following.

Lemma 13.2.2. Let S be a subset of K [X1 , . . . , Xn ] and let I be the ideal of K [X1 , . . . , Xn ] generated by
S. Then Z (I) = Z (S).
Proof. Since S ⊆ I, we have Z (I) ⊆ Z (S). On the other hand, let p = (x1 , . . . , xn ) ∈ Z (S). Then for all
P (X1 , . . . , Xn ) ∈ S, we have P (p) = 0. Since any polynomial P (X1 , . . . , Xn ) ∈ I can be expressed as

P = Q1 S1 + · · · + Qr Sr , Qi ∈ K [X1 , . . . , Xn ] , Si ∈ S,

But then we have

P (p) = Q1 (p) S1 (p) + · · · + Qr (p) Sr (p) = 0,
since Si (p) = 0 for all i. Thus we have that p ∈ Z (I).
From this and the Hilbert basis theorem we deduce, K [X1 , . . . , Xn ] is Noetherian and every ideal is finitely
generated, so for any subset S of polynomials in K [X1 , . . . , Xn ] there is a finite collection S 0 of polynomials
such that Z (S) = Z (S 0 ). Just let S 0 be a generating set for the ideal generated by S.
Definition 13.2.3. A subset T of AnK is called an affine algebraic set if T is for the form Z (I) for some
ideal I of K [X1 , . . . , Xn ].
Conversely, subsets of AnK define ideals of K [X1 , . . . , Xn ]. For T ⊆ AnK , let I (T ) denote the ideal

I (T ) = {P (X1 , . . . , Xn ) ∈ K [X1 , . . . , Xn ] | ∀t ∈ T, P (t) = 0} ⊆ K.

Claim that I (T ) is an ideal. If P (t) = 0 and Q (t) = 0 for all t ∈ T ,

P (t) + Q (t) = 0, R (t) P (t) = 0, t ∈ T, qR (X1 , . . . , Xn ) ∈ K [X1 , . . . , Xn ] .

Note. The operations T 7→ I (T ) and I 7→ Z (I) are both inclusion-reversing.

• If T ⊆ T 0 ⊆ AnK then I (T ) ⊇ I (T 0 ).
• If J ⊆ J 0 ⊆ K [X1 , . . . , Xn ] then Z (J) ⊇ Z (J 0 ).

We have the following.

Lemma 13.2.4. Let X be an affine algebraic set. Then Z (I (X)) = X.
Proof. First observe that certainly any element of I (X) vanishes at all points of X, so X ⊆ Z (I (X)). On
the other hand, since X is an affine algebraic set, X = Z (J) for some ideal J. Want Z (I (X)) ⊆ X, that is
Z (I (Z (J))) ⊆ Z (J). Since any P ∈ J vanishes on X we have J ⊆ I (Z (J)). Then Z (I (X)) ⊆ Z (J) = X,
so X = Z (I (X)).
This need not be equal for X arbitrary. If X is not an affine algebraic set, then Z (I (X)) is the smallest affine
algebraic set containing X. If Y = Z (J) contains X, then I (Y ) ⊆ I (X), so Z (I (X)) ⊆ Z (I (Y )) = Y .
We call Z (I (X)) the Zariski closure of X. This operation is the closure operation for a topology on AnK
called the Zariski topology which we will define later.

Example.

X = {(n, 0) | (n ∈ Z)} ⊆ A2C , I (X) = {P ∈ C [X, Y ] | ∀n ∈ Z, P (n, 0) = 0} .

P (X, 0) ∈ C [X] is a polynomial that vanishes at every n ∈ Z, so P (X, 0) = 0. So Y | P (X, Y ), so

I (X) = hY i. Thus Z (I (X)) is the Y -axis.

61
M3P8 Algebra III 13 Introduction to algebraic geometry

One might thus hope that similarly I (Z (J)) = J for any ideal J of K [X1 , . . . , Xn ]. This cannot be literally
true, for the following reason. Recall that
rad (I) = {r ∈ K [X1 , . . . , Xn ] | ∃m, rm ∈ I} .
An ideal I is radical if rad (I) = I.
Note. In fact, for any T , the ideal I (T ) is a radical ideal.
m
P m ∈ I (T ) ⇐⇒ ∀t ∈ T, P (t) =0 ⇐⇒ ∀t ∈ T, P (t) = 0 ⇐⇒ P ∈ I (T ) .
Thus if J is not a radical ideal we cannot have I (Z (J)) = J.
However, one does have the following.
Theorem 13.2.5 (Hilbert’s Nullstellensatz). Let K be an algebraically closed field. For any ideal J of
K [X1 , . . . , Xn ], we have I (Z (J)) = rad (J).
Note. This can only possibly hold over algebraically closed fields.
Example. For n = 1, let P (X) ∈ K [X].
I (Z (P (X))) = rad (P (X)) 6= R,
unless P (X) is constant, so Z (P (X)) 6= ∅.
Corollary 13.2.6. If J ∈ K [X1 , . . . , Xn ] an ideal is not the unit ideal, then Z (J) 6= ∅.
Proof. I (Z (J)) = rad (J) 6= h1i, so Z (J) 6= ∅.
Lecture 29
We will prove this theorem later on. For now, we note that since rad (rad (J)) = rad (J) for all ideals J, the Monday
maps X 7→ I (X) and J 7→ Z (J) define a bijection 10/12/18
affine algebraic subsets of AnK .
 
radical ideals J ⊆ K [X1 , . . . , Xn ] !
This bijection is inclusion-reversing, and it is interesting to ask what geometric properties of X are carried
to algebraic properties of I (X) via this bijection. For instance, the following holds.
Proposition 13.2.7. Let J ⊆ K [X1 , . . . , Xn ] be a radical ideal. Then J is maximal if and only if Z (J) is
a single point.
Proof. Suppose Z (J) = {p}, where p = (p1 , . . . , pn ). I (Z (J)) = I ({p}), so
X1 − p1 , . . . , Xn − pn ∈ I ({p}) .
Then
mp = hX1 − p1 , . . . , Xn − pn i ⊆ I ({p}) ( K [X1 , . . . , Xn ] .
Note that mp is maximal. Consider the map
K [X1 , . . . , Xn ] −→ K
,
Xi 7−→ pi
and K 7→ X. The kernel of this map is hXi − pi i = mp , so
K [X1 , . . . , Xn ] ∼
= K.
mp
So mp is maximal, must have I ({p}) = mp . Thus
J = rad (J) = I (Z (J)) = mp .
Conversely, suppose J is maximal. Z (J) 6= ∅, since I (Z (J)) = J but I (∅) is the unit ideal. So there exists
p ∈ Z (J) such that
J = I (Z (J)) ⊆ I ({p}) = mp ,
so J = mp .

62
M3P8 Algebra III 13 Introduction to algebraic geometry

Proposition 13.2.8. Let J1 , J2 ⊆ K [X1 , . . . , Xn ] be ideals. Then

Z (J1 + J2 ) = Z (J1 ) ∩ Z (J2 ) , Z (J1 ∩ J2 ) = Z (J1 ) ∪ Z (J2 ) .

Proof. Let p ∈ AnK be a point of Z (J1 + J2 ). Then every element of J1 +J2 vanishes at p, so since J1 ⊆ J1 +J2
we have that p ∈ Z (J1 ). Similarly p ∈ Z (J2 ), so

Z (J1 + J2 ) ⊆ Z (J1 ) ∩ Z (J2 ) .

Conversely, if p ∈ Z (J1 ) ∩ Z (J2 ), then for any element Q of J1 + J2 we can write

Q = R + S, R ∈ J1 , S ∈ J2 .

Then R (p) = S (p) = 0, so Q (p) = 0 and p ∈ Z (J1 + J2 ). The proof that

Z (J1 ∩ J2 ) = Z (J1 ) ∪ Z (J2 )

is similar, and will be omitted.

Corollary 13.2.9. Conversely, if X and Y are affine algebraic sets, then

I (X ∩ Y ) = rad (I (X) + I (Y )) , I (X ∪ Y ) = I (X) ∩ I (Y ) .

Proof. Follows from Z (J1 + J2 ) = Z (J1 ) ∩ Z (J1 ) and using the Nullstellensatz I (Z (J)) = rad (J).
Definition 13.2.10. An affine algebraic set X is irreducible if X cannot be written as the union Y ∪ Z of
two proper affine algebraic subsets Y and Z, that is Y, Z 6= X, ∅.
Example. Let
X = Z ({y (y − x (x − 1) (x + 1))}) .
Then
Z ({y (y − x (x − 1) (x + 1))}) = Z ({y}) ∪ Z ({y − x (x − 1) (x + 1)}) ,
so X is not irreducible.
Proposition 13.2.11. An affine algebraic set X is irreducible if and only if I (X) is prime.
Proof. Suppose X is irreducible, and let f, g be elements of K [X1 , . . . , Xn ] such that f g ∈ I (X). Then

X ⊆ Z (f g) = Z (f ) ∩ Z (g) .

In particular
X = (X ∩ Z (f )) ∪ (X ∩ Z (g)) .
Since X is irreducible we must have
X = X ∩ Z (f ) ⊆ Z (f ) ,
in which case f ∈ I (X), or
X = X ∩ Z (g) ⊆ Z (g) ,
in which case g ∈ I (X). So I (X) is prime. Conversely, suppose I (X) is prime, and that X = Y ∪ Z, where
Y and Z are affine algebraic subsets of X. Then

I (X) = I (Y ) ∩ I (Z) ,

so I (X) contains the product I (Y ) I (Z). Since I (X) is prime, either I (X) contains I (Y ), in which case
X = Y , or I (X) contains I (Z), in which case X = Z.

63
M3P8 Algebra III 13 Introduction to algebraic geometry

Definition 13.2.12. An affine algebraic set X has Krull dimension d if d is the length of the largest
increasing tower of irreducible affine algebraic subsets
X0 ( · · · ( Xd ⊆ X.
Example.
• Points have dimension zero.
• In A1 , affine algebraic sets are zeros of polynomials, so irreducibles in A1 are points on A1 , so A1 has
Krull dimension one.
• In fact, An has Krull dimension n for every n.
Definition 13.2.13. A ring R has Krull dimension d if d is the length of the largest increasing tower
p0 ( · · · ( pd
of prime ideals of R.
Example. If R is a domain,
• dimension zero implies that R is a field, and
• dimension one implies that every nonzero prime ideal is maximal.
Lecture 30
Lecture 30 is a problem class. Wednesday
The Hilbert basis theorem then gives us the following. 12/12/18
Proposition 13.2.14. Let X be an affine algebraic set. Then X can be written uniquely as a finite union
X1 ∪ · · · ∪ Xr such that each Xi is an irreducible affine algebraic set and no Xi is contained in Xj for i 6= j.
Proof. We first show that if X is not irreducible, then X can be written as Y ∪ Z with Y irreducible
and Z 6= X affine algebraic. Certainly we can write X = Y1 ∪ Z1 with Y1 , Z1 proper subsets of X. If
Y1 is irreducible we are done. Otherwise write Y1 = Y2 ∪ Z2 . Again, if Y2 is irreducible we can write
X = Y2 ∪ (Z1 ∪ Z2 ) and we are done. Otherwise, supposing this never terminates, we obtain
Y1 ) Y2 ) . . . , I (Y1 ) ( I (Y2 ) ( . . . ,
which is impossible since K [X1 , . . . , Xn ] is Noetherian. Now given X, if X is not irreducible we can write
X = Y1 ∪ Z1 with Y1 irreducible and Z1 6= X. If Z1 is not irreducible we write Z1 = Y2 ∪ Z2 with Y2
irreducible and Z2 6= Z1 . If this process ever terminates we have written X as a finite union of irreducibles.
Otherwise, we have
Z1 ) Z2 ) . . . ,
and as above this is impossible since K [X1 , . . . , Xn ] is Noetherian. For uniqueness, suppose we have
X = Y1 ∪ · · · ∪ Yr , X = Z1 ∪ · · · ∪ Zs ,
with the Yi and Zj irreducible, and with no Yi , or Zi , contained in Yj , or Zj , when i 6= j. Then I (Yi ) and
I (Zi ) are prime for all i. In particular I (Y1 ) is prime. Since Y1 ⊂ X, I (X) ⊂ I (Y1 ), so I (Y1 ) contains
I (Z1 ∪ · · · ∪ Zs ) = I (Z1 ) ∩ · · · ∩ I (Zs ) .
It follows that I (Y1 ) contains the product I (Z1 ) . . . I (Zs ). Thus I (Y1 ) contains I (Zj ) for some j. Similarly
I (Zj ) contains I (Yi ) for some i. Then I (Y1 ) ⊆ I (Yi ), so Yi ⊆ Y1 and we must have i = 1. Then Y1 = Zj .
Proceeding we show that each Yi is equal to some Zj and vice versa, proving uniqueness.
Translating this to a statement about ideals, we find the following.
Corollary 13.2.15. Every radical ideal in K [X1 , . . . , Xn ] is uniquely expressible as a finite intersection of
prime ideals, none of which contains any of the others.
This is a special case of a very general ring-theoretic phenomenon known as primary decomposition,
which was discovered via the sort of geometric considerations we see above.

64
M3P8 Algebra III 13 Introduction to algebraic geometry

13.3 Proof of the Nullstellensatz

The ideas above rely heavily on the correspondence between radical ideals and affine algebraic sets, and
thus ultimately on the Nullstellensatz. We now give a proof of the Nullstellensatz. The first step is to show
that to prove the Nullstellensatz it suffices to prove the following, seemingly much weaker, special case, the
so-called weak Nullstellensatz.
Theorem 13.3.1 (Weak Nullstellensatz). Let I be an ideal of K [X1 , . . . , Xn ] such that Z (I) is empty.
Then I is the unit ideal.
Proof of Theorem 13.2.5. Let J be an ideal of K [X1 , . . . , Xn ]. Clearly I (Z (J)) contains rad (J). We must
show the reverse containment. Let P be an element of I (Z (J)). We must show that P m lies in J for some
m. Consider the ring
K [X1 , . . . , Xn , T ] ,
and let Je be the ideal of K [X1 , . . . , Xn , T ] generated by the polynomials in I, together with the polynomial

1 − T P (X1 , . . . , Xn ) .

Consider the subset

  
Z Je = (x1 , . . . , xn , t) ∈ K n+1 | ∀Q ∈ J, Q (x1 , . . . , xn ) = 0, 1 − tP (x1 , . . . , xn ) = 0 ⊆ An+1
K .

 
In particular if (x1 , . . . , xn , t) lies in Z Je , then (x1 , . . . , xn ) lies in Z (J). Since P ∈ I (Z (J)) we have
P (x1 , . . . , xn ) = 0, so
1 − tP (x1 , . . . , xn ) = 1.
 
Thus Z Je is empty. By the weak Nullstellensatz, Je is the unit ideal, so there are polynomials Q0 , . . . , Qs
in K [X1 , . . . , Xn , T ], and R1 , . . . , Rs ∈ I, such that

1 = Q0 (1 − T P ) + Q1 R1 + · · · + Qs Rs .

Consider the map  

1
K [X1 , . . . , Xn , T ] −→ K X1 , . . . , Xn ,
P ,
1
T 7−→
P
and K [X1 , . . . , Xn ] 7→ K [X1 , . . . , Xn ]. Applying this map we find that
   
1 1
1 = Q1 X1 , . . . , Xn , R1 (X1 , . . . , Xn ) + · · · + Qs X1 , . . . , Xn , Rs (X1 , . . . , Xn ) ,
P P

in K [X1 , . . . , Xn , 1/P ]. Multiplying by a sufficiently large power of P , we get

   
m m 1 m 1
P = P Q1 X1 , . . . , Xn , R1 (X1 , . . . , Xn ) + · · · + P Qs X1 , . . . , Xn , Rs (X1 , . . . , Xn ) .
P P

Since for m sufficiently large  

m 1
P Qs X1 , . . . , Xn ,
P
is a polynomial in the Xi we find that P m ∈ I for m sufficiently large.

It remains to prove the weak Nullstellensatz. This requires some new ideas. The following approach is due
to Emmy Noether.

65
M3P8 Algebra III 13 Introduction to algebraic geometry

Definition 13.3.2. Let R be a K-algebra, that is a ring together with a map K → R. We say that elements
y1 , . . . , ys of R are algebraically independent over K if there is no nonzero polynomial P (X1 , . . . , Xs ) ∈
K [X1 , . . . , Xs ] such that P (y1 , . . . , ys ) = 0. Equivalently, y1 , . . . , ys are algebraically independent if and
only if the map
K [X1 , . . . , Xs ] −→ R
Xi 7−→ yi
is injective.
Proposition 13.3.3 (Noether’s normalisation lemma). Let K be a field, and let R be a finitely generated
K-algebra. Then there exists s ∈ Z≥0 , and algebraically independent elements y1 , . . . , ys of R such that R is
integral over K [y1 , . . . , ys ].
Proof. Write
K [X1 , . . . , Xm ]
R= .
I
We proceed by induction on m. The base case m = 0 is clear. Fix m and assume the claim is true for
m − 1. If I = 0 then the statement is also clear, with yi = Xi for all i. Otherwise let P (X1 , . . . , Xm ) ∈ I.
Renumbering the variables if necessary, we may assume P is a nonconstant polynomial in Xm with coefficients
in X1 , . . . , Xm−1 . Let d be the total degree of P . That is, the largest value of a1 + · · · + am for any monomial
i
cX1a1 . . . Xm am
appearing in P . Let ni = (1 + d) and Yi = Xi − Xm ni
for each i in 1, . . . , m − 1. Define
n1 nm−1
Q (X1 , . . . , Xm ) = P (X1 + Xm , . . . , Xm−1 + Xm , Xm ) .
Then Q (Y1 , . . . , Ym−1 , Xm ) is zero in R. We now claim that, up to a factor c ∈ K ∗ , Q (X1 , . . . , Xm ) is monic
when considered as a polynomial in Xm . Let cX1a1 . . . Xm am
be a monomial appearing in P (X1 , . . . , Xm ).
This monomial contributes the following terms to Q (X1 , . . . , Xm ).
n1 a1 nm−1 am−1 am
c (X1 − Xm ) . . . (Xm−1 − Xm ) Xm .
Moreover, each ni is greater than d and hence greater than the sum of the aj . It is thus clear that the term
N
of highest degree in the above expression is cXm , where
m−1
N = n1 a1 + · · · + nm−1 am−1 + am = a1 (1 + d) + · · · + am−1 (1 + d) + am .
Since 1 + d is greater than the sum of the exponents a appearing in any monomial of P (X1 , . . . , Xm ), the
N
terms cXm appearing in different monomials are all of different degree and thus cannot cancel. It follows
N
that the term of the form cXm of highest degree is the highest degree term in Q (X1 , . . . , Xm ), so that
(1/c) Q (X1 , . . . , Xm ) is monic in Xm . Write
N
1 X
n
Q (X1 , . . . , Xm ) = Hn (X1 , . . . , Xm−1 ) Xm .
c n=0

Since Q (Y1 , . . . , Ym−1 , Xm ) = 0, we have

N
X
Hn (Y1 , . . . , Ym−1 ) Xm = 0.
n=0

That is, Xm is integral over the subalgebra S = K [Y1 , . . . , Ym−1 ] of R. Since Xm generates R over S, it
follows that R is integral over S. On the other hand we have a map
K [Z1 , . . . , Zm−1 ] −→ S
.
Zi 7−→ Yi
Let J be the kernel. Then
K [Z1 , . . . , Zm−1 ]
S= .
J
Then by the inductive hypothesis there are algebraically independent elements y1 , . . . , ys ∈ S such that S is
integral over K [y1 , . . . , ys ]. Since R is integral over S, it follows that R is integral over K [y1 , . . . , ys ] and we
are done.

66
M3P8 Algebra III 13 Introduction to algebraic geometry

Corollary 13.3.4. Every maximal ideal of K [X1 , . . . , Xn ] is of the form

hX1 − p1 , . . . , Xn − pn i , p1 , . . . , pn ∈ K.

Proof. Let I be a maximal ideal of K [X1 , . . . , Xn ], and consider R = K [X1 , . . . , Xn ] /I. Then R is a field.
On the other hand, by Noether normalisation, there exist y1 , . . . , ys algebraically independent such that R
is integral over S = K [y1 , . . . , ys ]. Let x be a nonzero element of K [y1 , . . . , ys ]. Then x−1

lies in R. Since R
is integral over S there is a monic polynomial P with coefficients in S such that P x−1 = 0. We thus have

d−1

d X
x−1 = ai x−i ,
i=0

with ai ∈ S. Multiplying by xd−1 we find that

d−1
X
x−1 = ai xd−i−1 ,
i=0

so that x−1 is also in S. Thus S is a field. But since the yi are algebraically independent, S is also a
polynomial ring in s variables. Since no such ring is a field unless s = 0 we must have s = 0 and R is
integral over K. But then R is a finite dimensional K-vector space, hence a finite extension of K. Since K
is algebraically closed, the inclusion of K in R is an isomorphism. Thus for each i there is an element pi of
K such that Xi is equal to pi in R. Then Xi − pi is in I for all i, so I contains the ideal

hX1 − p1 , . . . , Xn − pn i .

Since the latter is clearly maximal it must be equal to I.

Proof of Theorem 13.3.1. Let I be an ideal of K [X1 , . . . , Xn ] such that I is not the unit ideal. Then I is
contained in some maximal ideal of K [X1 , . . . , Xn ], and thus in some ideal of the form

hX1 − p1 , . . . , Xn − pn i .

Then (p1 , . . . , pn ) lies in Z (I).

67