Download as pdf or txt
Download as pdf or txt
You are on page 1of 49

A Study in Commutative Algebra

Hasitha Nipun

May 1, 2023
Contents

Preface 1

1 A Review of Ring Theory 2


1.1 Rings and ring homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Ideals and quotient rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Zero divisors, nilpotent elements and units . . . . . . . . . . . . . . . . . . . . . 3
1.4 Prime ideals and maximal ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Nilradical and Jacobson radical . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Operations on ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.7 Extension and contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Some solved problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 Euclidean Domains 15
2.1 Basic definitions and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Properties of Euclidean domains . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Some solved problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3 Principal Ideal Domains and Unique Factorization Domains 22


3.1 Principal Ideal Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Unique Factorization Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Some solved problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4 Modules 27
4.1 Modules and module homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Submodules and quotient modules . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Operations on submodules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4 Direct sum and product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.5 Finitely generated modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.6 Exact sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.7 Tensor product of modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.8 Some solved problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

i
5 Localization 43
5.1 Rings of Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

ii
Preface

Commutative algebra is essentially the study of commutative rings and their modules. It is
a rich and diverse area of mathematics with applications in various fields, including algebraic
geometry, algebraic number theory, and algebraic topology.

Chapter 1 serves as a review of ring theory, covering essential concepts on rings and ideals.
Chapter 2 is a study of Euclidean domains, examining their basic definitions and properties.
Chapter 3 focuses on principal ideal domains and unique factorization domains, exploring their
properties and structures. Chapter 4 explores the fundamental concepts of modules, including
submodules, operations on submodules, direct sums, exact sequences and mainly, tensor prod-
ucts. Chapter 5 is devoted to localization, which introduces ”denominators” for specific ring
elements and is crucial for studying algebraic objects at prime ideals.

Throughout this text, the following conventions and terminologies are employed: by a “ring”
we shall always mean a commutative ring, with an identity element; We do not exclude the
possibility in that 1 might be equal to 0. (i.e. The only ring satisfying this property is the zero
ring.); Every non-zero ring homomorphism preserves the identity; A subring of a ring contains
the identity.

Chapter 2 and 3 are based on [6], while the other chapters are based on [2]. The last section
of each chapter is reserved for some solved problems, and almost all of these problems were
taken from [2] and [6].

Commutative algebra provides fundamental tools and techniques for studying algebraic
geometry and algebraic number theory, two prominent areas of research. By mastering the
core ideas and techniques of commutative algebra, one can develop a solid understanding of
topics such as varieties, the Nullstellensatz, and homological algebra, which have applications
in algebraic geometry.

1
1. A Review of Ring Theory

In this chapter We shall provide some basic definitions, and some elementary properties of
rings which are mostly well-known. Subsequently, we will discuss ideals of a ring and the basic
operations that can be performed on them.

1.1 Rings and ring homomorphisms


Definition 1.1.1. A ring A is a set together with two binary operations (called addition and
multiplication) satisfying the following axioms:
1. (A, +) is an abelian group.

2. Multiplication is associative: (ab)c = a(bc) for all a, b, c ∈ A.

3. Multiplication is distributive over addition: a(b + c) = ab + ac, (a + b)c = ac + bc for all


a, b, c ∈ A.
The ring A is commutative if ab = ba for all a, b ∈ A; and A is said to have an identity element
if there exists 1 ∈ A with a1 = 1a = a for all a ∈ A.
We shall consider only commutative rings with an identity. Moreover, identity element in a
ring is unique. By a subring of a ring A we shall mean a subset S of A which is closed under
addition and multiplication and contains the identity element of A.
Definition 1.1.2. Let A and B be rings. A mapping f : A → B is a ring homomorphism if it
respects addition, multiplication and the identity element. That is:
1. f (a + b) = f (a) + f (b) for all a, b ∈ A (So, f is a homomorphism of abelian groups).

2. f (ab) = f (a)f (b) for all a, b ∈ A.

3. f(1)=1.
The kernel of the ring homomorphism f , denoted ker f , is the set of elements of A that map to
0 in B.
If f : A → B and g : B → C are ring homomorphisms, so is their composition g ◦f : A → C.
A bijective ring homomorphism is called an isomorphism. A ∼ = B means that there is a ring
isomorphism f : A → B.

2
1.2 Ideals and quotient rings
An ideal a of a ring A is an additive subgroup of A which is closed under the multiplication by
elements from A, i.e., Aa = {xa | x ∈ A, a ∈ a} ⊆ a. If a ̸= A, a is a proper ideal. The quotient
group A/a inherits a uniquely defined multiplication from A which makes it into a ring, called
the quotient ring (or residue-class ring) A/a. The elements of A/a are the cosets x + a of a in
A, where x ∈ A; and the natural projection map π : A → A/a given by x 7→ x + a is a surjective
ring homomorphism. We shall use the notation x ≡ y( mod a) sometimes to say that x−y ∈ a.
If f : A → B is a ring homomorphism then ker f is an ideal of A and Im f (= f (A)) is a subring
of B. Furthermore, if b is an ideal of B, then f −1 (b) = {x ∈ A : f (x) ∈ b} is an ideal of A. The
following so-called fundamental theorem of ring homomorphisms is particularly important.
Theorem 1.2.1. Let φ : A → B be a ring homomorphism with a ⊆ ker φ where a is an ideal of
A. Then φ induces a unique ring homomorphism φ : A/a → B such that φ ◦ π = φ. Moreover,
if φ is surjective then A/ ker φ ∼
= B is an isomorphism of rings.
We shall frequently use the following result, which is sometimes known as the correspondence
theorem [3].
Proposition 1.2.2. Let a be an ideal of A. The correspondence A ←→ A/a is an inclusion pre-
serving bijection between the ideals b of A that contain a and the ideals b of A/a. Furthermore,
b ⊇ a is an ideal of A if and only if b/a is an ideal of A/a.

1.3 Zero divisors, nilpotent elements and units


Let A be a ring. An element x ∈ A is called a zero divisor if xy = 0 for some non-zero y ∈ A.
An integral domain is a ring with identity 1 ̸= 0 where there are no non-zero zero divisors.
The ring of integers Z, polynomial ring R[x1 , ..., xn ] over an integral domain R are examples of
integral domains.
An element x ∈ A is called nilpotent if xn = 0 for some integer n > 0. A nilpotent element
is a zero divisor unless A = 0, but not conversely in general. So, an integral domain does not
contain non-zero nilpotent elements.
An element x ∈ A is called a unit if xy = 1 for some y ∈ A. The element y is called
the (multiplicative) inverse of x. In a ring, the inverse of an element (if exists) is unique,
and denoted by x−1 . The set of all units of A is sometimes denoted as A× ; the set A× is a
multiplicative abelian group. A field is a ring where each nonzero element is a unit. Any field
is an integral domain; but there exist integral domains (for example, Z) that are not fields.
For a fixed x ∈ A, the multiples ax(a ∈ A) of x forms an ideal which is known as the
principal ideal of A; this ideal is denoted by (x) or Ax. The zero ideal (0) is denoted by just 0
when there is no confusion. x ∈ A is a unit if and only if (x) = A = (1).
Proposition 1.3.1. Let A ̸= 0 be a ring. Then the following are equivalent.
(i) A is a field;
(ii) the only ideals in A are 0 and (1);
(iii) every non-zero homomorphism f : A → B (B ̸= 0 is a ring) is injective.

3
Proof. (i) =⇒ (ii) Clearly, 0 and (1) are ideals of A. If a ⊆ A is any non-zero ideal, then a
contains a unit x. Now, a ⊇ (x) = (1). So, a = (1).
(ii) =⇒ (iii) Since f is non-zero, ker f is proper ideal of A. Hence, ker f = 0. So, f is injective.
(iii) =⇒ (i) Let z ∈ A be a non-unit element. Then (x) ̸= (1). Thus, B = A/(x) is not the zero
ring. Now, the natural projection f : A → B is a non-zero ring homomorphism onto B with
ker f = (x). Since f is injective, (x) = 0. So, x = 0.

1.4 Prime ideals and maximal ideals


Definition 1.4.1. Let A be a ring.

• An ideal p ̸= A of A is prime if ab ∈ p implies a ∈ p or b ∈ p.

• An ideal m ̸= A of A is maximal if the only ideals containing m are m itself and A.

We have the following equivalent characterizations for prime and maximal ideals.

Proposition 1.4.1. Let p and m be ideals of the ring A.

i p is prime if and only if A/p is an integral domain.

ii m is maximal if and only if A/m is a field.

Proof. i. p is prime ⇔ p ̸= A and ab ∈ p implies a ∈ p or b ∈ p in A ⇔ A/p ̸= 0 and ab = ab = 0


implies a = 0 or b = 0 in A/p ⇔ A/p is an integral domain.
ii. m is maximal ⇔ ideals containing m are only m and A ⇔ ideals of A/m are only 0 and A/m
(by proposition 1.2.2) ⇔ A/m is a field (by proposition 1.3.1).

Corollary 1.4.2. In a ring A, every maximal ideal is prime.


Proof. By the proposition 1.4.1 m is maximal ⇔ A/m is a field ⇒ A/m is an integral domain
⇒ m is prime.
Converse of corollary 1.4.2 is not true; for instance, 0 is a prime ideal in Z but not maximal.
It is easy to see that 0 is a prime ideal in A if and only if A is an integral domain.

Proposition 1.4.3. Let f : A → B be a ring homomorphism. Then the following hold.

(i) If q ⊆ B is a prime ideal then f −1 (q) is a prime ideal of A.

(ii) If f is surjective and p ⊆ A is a prime ideal then f (p) is a prime ideal of B.

Proof. (i) It is easy to check that f −1 (q) is an ideal of A. The map φ : A → B/q given by
a 7→ f (a) + q is a homomorphism with ker φ = f −1 (q). Hence, A/f −1 (q) is isomorphic to Im φ,
which is a subring of B/q and also an integral domain since q is prime. So, f −1 (q) is also prime.
(ii) This verification is straightforward and can be accomplished by relying on definitions.

4
However, if m is a maximal ideal of B, it does not guarantee that the preimage f −1 (m) ,
is necessarily a maximal in A. For instance, consider the inclusion homomorphism ι : Z → Q;
zero ideal is maximal in Q but ι−1 (0) = 0 is not maximal in Z. Before moving to the next
theorem let us recall the Zorn’s lemma [5, p. 53].
1
Zorn’s lemma: If S is a nonempty partially ordered set in which every chain has an upper
bound then S has a maximal element.
Prime ideals play a crucial role in entire commutative algebra. The following proposition
guarantees the existence of prime ideals in a ring; proof relies on Zorn’s Lemma.
Theorem 1.4.4 (Krull’s theorem). Any ring A ̸= 0 has at least one maximal ideal.
Proof. Let S be the set of all proper ideals in A. Order S by inclusion so that S is a partially
S set. Also, S is non-empty as 0 ∈ S. Let C = (aα ) be any chain of ideals in S. Take
ordered
a = α aα .
Claim: a is an ideal of A.
Let a, b ∈ a. Then a ∈ aα and b ∈ aβ for some aα , aβ ∈ C. So, either aα ⊆ aβ or aβ ⊆ aα ; in
either case a − b ∈ C. Since each aα ∈ C is closed under multiplication by elements of A, so is
a. This proves the claim.
Note that 1 ∈ / a as 1 ∈
/ aα for each aα ∈ C. Therefore, a ∈ S and a is an upper bound for C.
By Zorn’s lemma S possesses a maximal element.
Corollary 1.4.5. If a is a proper ideal of the ring A, then a is contained in some maximal
ideal of A.
First proof. By 1.4.4 A/a has a maximal ideal m. Let π : A → A/a be the natural projection.
Then π −1 (m) is and ideal of A; and m := π −1 (m) contains a by proposition 1.2.2. If m is not
maximal then there exists an ideal a with a ⊊ m ⊊ A. Hence, again by proposition 1.2.2 π(b)
is a proper ideal of A/a containing m which contradicts the maximality of m.
Second proof. Let S be the set of all proper ideals of A which contain a. Then use a similar
argument as in the proof of proposition 1.4.4.
Corollary 1.4.6. An element x ∈ A is non-unit if and only if x is contained in some maximal
ideal.
Proof. If x ∈ A is non-unit then (x) is a proper ideal of A and hence contained in a maximal
ideal by corollary 1.4.5. Conversely, suppose x is in some ideal m and x is a unit. Then
(1) = (x) ⊆ m ⊊ (1) is a contradiction.
A ring A is called local if it has precisely one maximal ideal m. The field A/m is called the
residue field of A. Fields are local rings. A semi-local ring is a ring with only a finite number
if maximal ideals.
Proposition 1.4.7. Let A be a ring.
1
A partially ordered set is a set S with a relation “≤” which is reflexive (x ≤ x for all x ∈ S), anti-
symmetric (x ≤ y and y ≤ x implies x = y for all x, y ∈ S) and transitive (x ≤ y and y ≤ z implies x ≤ z for
all x, y, z ∈ S)

5
(i) If m ≠ (1) is an ideal of A such that every x ∈ A \ m is a unit in A, then Then A is a
local ring and m its maximal ideal.
(ii) If m is a maximal ideal in A such that every element of the form 1 + x where x ∈ m is a
unit in A. Then A is a local ring.
Proof. (i) Let a ̸= (1) be an ideal of A. Then x ∈ a ⇒ x ∈ / A× ⇒ x ∈/ A \ m ⇒ x ∈ m. So,
a ⊆ m. Thus, m is the only maximal ideal of A.
(ii) Let a ∈ A \ m. Since m is maximal, the ideal generated by a and m is (1); then there exist
b ∈ A and t ∈ m such that ab + t = 1. Now, ab = 1 − t is a unit in A by assumption. Thus, a
is a unit and hence by part (i), A is a local ring.

1.5 Nilradical and Jacobson radical


Definition 1.5.1. The set of all nilpotent elements of a ring A is called the nilradical of A and
it denoted by R(A).
Proposition 1.5.1. R(A) is an ideal of A; and A/R(A) has no non-zero nilpotent elements.
Proof. Let a, b ∈ R(A); say am = 0 and bn = 0 for some positive integers m, n. From the
binomial theorem (which is valid in any commutative ring), we have (a + b)n+m is a sum of
integer multiples of products of the form ar bs where r + s = m + n; it is impossible to have
both r < m and s < n; hence each of these products vanishes and consequently, (a + b)m+n = 0.
Whence a + b ∈ R(A). Also for any x ∈ A, we have (xa)n = xn an = 0. So, xa ∈ R(A). Thus,
R(A) is an ideal of A.
Let x + R(A) ∈ A/R(A). Then (x + R(A))n = 0 in A/R(A) ⇒ xn ∈ R(A) ⇒ (xn )k = 0 for
some k ∈ N ⇒ (xn )k = 0 ⇒ x ∈ R(A) ⇒ x + R(A) = 0 in A/R(A).

The following proposition gives an alternating definition for the nilradical of a ring.

Proposition 1.5.2. The nilradical of the ring A is equal to the intersection of all prime ideals
of A.
Proof. Let P be the intersection of all prime ideals of A. Let p be any prime ideal. For any
f ∈ R(A), we have f n = 0 ∈ p for some n ∈ N. But since p is prime, we get f ∈ p. So, f ∈ P
and hence R(A) ⊆ P.
For the reverse containment, suppose f ∈/ R(A). We show that f ∈ / P. Let

S = {a ⊆ A is an ideal : f n ∈
/ a for all n ∈ N}.

Then S is Snon-empty as 0 ∈ S. Order S by inclusion. Let C = (aα ) be any chain of ideals in S.


Take p = α aα . Then p is an ideal of A. Also, for each n ∈ N, f n ∈ / p as as f n ∈
/ aα for each
aα ∈ C. Therefore, p ∈ S and p is an upper bound for C. By Zorn’s lemma S has a maximal
element. Let q be a maximal element of S. We show that q is a prime ideal. Let x, y ∈ / q. Then
the ideals q + (x), q + (y) are strictly containing q and hence do not belong to S. Therefore,
f n ∈ q + (x) and f m ∈ q + (y) for some m, n ∈ N. This implies f m+n ∈ q + (xy); thus the ideal
q + (xy) does not belong to S and hence xy ∈ / q showing that q is a prime ideal. Since q is an
element of S, f ∈/ q. But q is a prime ideal; hence f ∈/ P.

6
Definition 1.5.2. The intersection of all maximal ideals of a ring A is called the Jacobson
radical of A and it denoted by J(A).
In any ring, since all maximal ideals are prime, the nilradical is contained in the Jacobson
radical. Proposition 1.5.3 gives the characterization of Jacobson radical.

Proposition 1.5.3. Let A be a ring. Then x ∈ J(A) if and only if 1 − xy is a unit in A for
all y ∈ A.
Proof. ⇒: Suppose x ∈ J(A) and 1 − xy is not a unit in A for some y ∈ A. By corollary 1.4.6
1 − xy is contained in some maximal ideal m. But x ∈ J(A) ⊆ m. So, xy ∈ m implying 1 ∈ m
which is impossible.
⇐: Suppose x ∈ / J(A). Then x ∈ / m for some maximal ideal m. Thus x and m generates the
ideal (1). Then there exists y ∈ A and u ∈ m such that xy + m = 1. Now, 1 − xy = m ∈ m
and hence cannot be a unit by corollary 1.4.6.

1.6 Operations on ideals


Let A be a ring and X ⊆ A. Then the ideal generated by X is denoted by (X) or ⟨X⟩. That
is, the intersection of all ideals in A containing X. Note that the empty set generates the zero
ideal. Let L(A) be the set of all ideals of A. Then L(A) is a partially ordered set with respect
to inclusion.
If a, b ⊆ A are ideals then we define their sum a + b as a + b := {x + y : x ∈ a and y ∈ b}.
It is easy to check that
P a + b is the smallest ideal containing both a and a. More generally, we
can define the sum i∈I ai of any family (ai )i∈I of ideals ai of A as follows:
( )
X X
ai := xi | xi ∈ ai ∀i ∈ I and only finitely many xi are non-zero
i∈I i∈I
P
It is straightforward to verify that i∈I ai the smallest ideal of A which contains all the ideals
ai . The intersection of any family (ai )i∈I of ideals ai is an ideal of A. Moreover, it is the largest
ideal contained in each ai . This discussion readily leads to the following proposition.

Proposition 1.6.1. L(A) is a complete lattice. For any S ⊆ L(A) we have


\
glb S = a
a∈S
X
lub S = ai
ai ∈S

where glb S and lub S denote greatest lower bound and least upper bound of S, respectively.
P S
Proposition 1.6.2. Let (ai )i∈I be any family of ideals of A. Then i∈I ai = i∈I ai .

Proof. We will give a more general proof in chapter 5. See proposition 4.3.2

7
If a and b are two ideals in the ring A, their product ab is defined as
( n )
X
ab := xi yi : xi ∈ a, yi ∈ b, n ∈ N
i=1

In a similar way, we can define the product of finitely many ideals of A. Thus, the powers
an where n ∈ N of an ideal a is defined; it is the ideal generated by the products of the form
x1 x2 · · · xn where each xi ∈ a. By convention, a0 = (1).
Examples 1.6.3.
• Let A = Z, a = (m), b = (n) where m, n ∈ Z. Then we have a ∩ b = ⟨lcm(m, n)⟩,
a + b = ⟨gcd(m, n)⟩ and ab = ⟨mn⟩.
• Let A = k[x1 , ..., xn ] where k is a field. Let a = ⟨x1 , ..., xn ⟩. So a consists of the polyno-
mials with zero constant term. Then for m ∈ N, am is generated by monomials of degree
at least m.
The operations sum, intersection and product of ideals in a ring are all commutative and
associative. Also, the ideal multiplication is distributive over addition (i.e., a(b + c)). In the
ring of integers the sum and intersection of ideals distributes over each other. But in general,
ideal intersection does not distribute over ideal addition. That is,
a ∩ (b + c) ̸= [a ∩ b] + [a ∩ c] (1.1)
The following example illustrates the inequality 1.1.
Examples 1.6.4. Let A = Z[x]. Take a = ⟨x⟩ , b = ⟨x + 1⟩ and c = ⟨x − 1⟩.
Clearly, 2x ∈ a and 2x ∈ b + c. Hence, 2x ∈ a ∩ (b + c). We show that 2x ∈
/ [a ∩ b] + [a ∩ c].
Claim: a ∩ b = ⟨x(x + 1)⟩ and a ∩ c = ⟨x(x − 1)⟩
Clearly, ⟨x(x + 1)⟩ ⊆ a ∩ b.
For the reverse containment let f (x) ∈ a ∩ b.
Then f (x) = xp(x) = (x + 1)q(x) and for some p(x), q(x) ∈ Z[x]. Note that,
f (x) = f (x)(x + 1) − xf (x) = x(x + 1)p(x) − x(x + 1)q(x) = x(x + 1)[p(x) − q(x)]
Thus, f (x) ∈ ⟨x(x + 1)⟩. So, a ∩ b ⊆ ⟨x(x + 1)⟩.
Similarly we can show that a ∩ c = ⟨x(x − 1)⟩. This proves the claim.
If 2x ∈ [a ∩ b] + [a ∩ c] then 2x = x(x + 1)g1 (x) + x(x − 1)g2 (x) for some g1 (x), g2 (x) ∈ Z[x].
But this is impossible as the degree of the polynomial in the left hand side is one while that of
the polynomial in the right hand side is at least two. So, 2x ∈
/ [a ∩ b] + [a ∩ c].
Equality can be achieved in the expression 1.1 under certain conditions which are given in
the modular law.
Proposition 1.6.5 (Modular law). If b ⊆ a or c ⊆ a then a ∩ (b + c) = [a ∩ b] + [a ∩ c]
Proof. Suppose b ⊆ a. Any x ∈ a ∩ (b + c) can be written as x = a = b + c for some a ∈ a, b ∈ b
and c ∈ c. since b ⊆ a we have c = a−b ∈ a∩c. Therefore, x = b+c ∈ b+[a∩c] = [a∩b]+[a∩c].
So, a ∩ (b + c) ⊆ [a ∩ b] + [a ∩ c].
On the other hand, [a ∩ b] + [a ∩ c] ⊆ a and [a ∩ b] + [a ∩ c] ⊆ b + c. Hence, [a ∩ b] + [a ∩ c] ⊆
a ∩ (b + c).

8
For any two ideals a and b we have, (a + b)(a ∩ b) ⊆ ab because (a + b)(a ∩ b) = a(a ∩ b) +
b(a ∩ b) ⊆ ab. Since the generating set of ab is contained in a ∩ b we have ab ⊆ a ∩ b.
Definition 1.6.1. Two ideals a and b are said to be coprime (or comaximal) if a + b = (1).
If a and b are coprime then a∩b = (1)(a∩b) = (a+b)(a∩b) ⊆ ab. It follows that a∩b = ab
whenever a and b are coprime. Clearly two ideals a and b are coprime if and only if there exist
a ∈ a and b ∈ b such that a + b = 1.
Let A1 , ..., An be rings. Their direct product, denoted by ni=1 Ai is defined as
Q

n
Y
A= Ai := {(xi , ..., xn )| xi ∈ Ai } .
i=1

Note that, A = ni=1 Ai is a ring under the coordinate-wise addition and multiplication. The
Q
projection mapping pi : A −→ Ai defined by (x1 , ..., xn ) 7−→ xi for each i = 1, ..., n is an onto
ring homomorphism.
Theorem 1.6.6 (Chinese remainder theorem). Let A be a ring and a1 , ..., an ideals of A. Define
the homomorphism
Yn
ϕ : A −→ (A/ai )
i=1

by ϕ(x) = (x + a1 , ..., x + an ). Then,


Q T
(i) If ai , aj are coprime whenever i ̸= j, then ai = ai .

(ii) ϕ is surjective ⇔ ai , aj are coprime whenever i ̸= j.


T
(iii) ϕ is injective ⇔ ai = (0).

Proof. (i) We use the induction on n. Q We have already


Tn−1 proved for n = 2. Suppose n > 2 and
n−1
the result is true for a1 , ..., an−1 . Thus, i=1 ai = i=1 ai . Since ai +an = (1) for i = 1, ..., n−1,
we get equations x1 + yi = 1 where xi ∈ ai and yi ∈ an and hence
n−1
Y n−1
Y
xi = (1 − yi ) ≡ 1( mod an )
i=1 i=1
Qn−1
Thus, an + i=1 ai = (1) and therefore,
n n−1
! n−1
! n
Y Y \ \
ai = ai an = ai ∩ an = ai
i=1 i=1 i=1 i=1

(ii) Suppose ϕ is surjective. Let ei = (0, ..., 1, ..., 0) where unity is in the i th coordinate and
zero elsewhere. Let ϕi be the projection ϕi : A → A/ai . Now for each i = 1, ..., n there exists
xi ∈ A such that ϕ(xi ) = ei . Note that ϕj (xi ) = 0 for i ̸= j and ϕj (xi ) = 1 for i = j. Observe
that, ϕi (1 − xi ) = ϕi (1) − ϕi (xi ) = 1 − 1 = 0. So, 1 − xi ∈ ker ϕi = ai . Now let i ̸= j. Then
1 = (1 − xi ) + xi ∈ ai + aj . Hence, ai + aj = (1).
For the converse, suppose that ai , aj are coprime whenever i ̸= j. Since ei (1 ≤ i ≤ n) generate

9
Qn
i=1 (A/ai ) it is enough to find ai ∈ A such that ϕ(ai ) = ei . Let i ̸= Q j. Since ai + aj = (1)
there exists pij ∈ ai and qij ∈ aj such that pij + qij = 1. Choose ai = k̸=i qik . Then ai ∈ ak
and
!
Y Y Y Y
ϕi (ai ) = ϕi qik = ϕi (qik ) = ϕi (1 − pik ) = ϕi (1) − ϕi (pik ) = 1 − 0 = 1.
k̸=i k̸=i k̸=i k̸=i

Therefore, ai − 1 ∈ ai . It follows that, ϕi (ai ) = ei .


T
(iii) ϕ is injective ⇔ ker ϕ = (0). But ker ϕ = ai .
The union a ∪ b of ideals is not necessarily an ideal. For instance, 2Z and 3Z are ideals in
Z, but their union is not; as 5 = 2 + 3 ∈
/ 2Z ∪ 3Z.

Proposition 1.6.7. Let A be a ring.

(i) If p1 , ..., pn are primes ideals of A and a ⊆ A is an ideal such that a ⊆ ni=1 pi then a ⊆ pi
S
for some i.
Tn
T p ⊆ A is a prime ideal such that i=1 ai ⊆ p then ai ⊆ p
(ii) If a1 , ..., an are ideals of A and
for some i. Moreover, if p = ai then p = ai .

Proof. (i) We use use induction on n to prove


n
[
a ⊈ pi for i ∈ {1, ..., n} =⇒ a ⊈ pi .
i=1

This is clearly true for n = 1. Suppose n > 1 and the statement holds for n − 1. Suppose a ⊈ pi
for each i = 1, ..., n. Then for each i we have
[
a ⊈ pj for i ∈ {1, ..., n} \ {i} =⇒ a ⊈ pj .
j̸=i
S
each i, there exists xi ∈ a such that xi ∈
Hence for S / j̸=i pj . If xi ∈
/ pi for some i = 1, ..., n
/ nj=1 pj . So we are done. If xi ∈ pi for each i = 1, ..., n then consider the element
then xi ∈
/ pi since each monomial is not in any pi . So, a ⊈ ni=1 pi .
y = ni=1 nj̸=i xi . Then y ∈ a but y ∈
P Q S

(ii) To the contrary


Qn supposeTn ai ⊈ p for all i. Then there exists xi ∈ ai such that xi ∈ / p.
Thus, x1 · · · xn ∈ i=1 ai ⊆ i=1 ai ⊆ p. So, x1 · · ·Txn ∈ p. But this
T contradicts with xi ∈ p as
p is a prime ideal. Thus, ai ⊆ p for some i. If p = ai then p = ai ⊆ ai and hence p = ai for
some i.

Definition 1.6.2. If a and b are ideals of a ring A their ideal quotient (a : b) is defined as

(a : b) = {x ∈ A| xb ⊆ a}

It is easy to check that (a : b) is an ideal of A, which is sometimes referred to as a colon


ideal. If b is a principal ideal (say b = (x)), we write (a : x) instead of ((a : (x)). When a is

10
the zero ideal (0 : b) is known as the annihilator of b; it is precisely the set of all x ∈ A such
that xb = 0 for all b ∈ b. Annihilator of b is also denoted by Ann(b): with this notation the
set D of all zero divisors in A is denoted by
[
D= Ann(x).
x̸=0

Examples 1.6.8.
• Let A = Z, a = (m) and b = (n). Write m = pα1 1 · · · pαk k , m = pβ1 1 · · · pβkk where p1 , ..., pk
are primes and α1 , . . . , αk , β1 , . . . , βk are non-negative integers. Then

(a : b) = {x ∈ Z| xn ∈ mZ} = qZ

where q = m/ gcd(m, n).


Proposition 1.6.9. The following properties hold for ideal quotient.
(i) a ⊆ (a : b)

(ii) (a : b)b ⊆ a

(iii) ((a : b) : c) = (a : bc) = ((a : c) : b)


T T
(iv) ( i ai : b) = i (ai : b)
P T
(v) (a : i bi ) = i (a : bi )
Proof. (i) x ∈ a. For any b ∈ b we have, xb ∈ (x) ⊆ a. So, x ∈ (a : b).

(ii) Let xb ∈ (a : b)b where x ∈ (a : b) and b ∈ b. By definition, xb ⊆ a. Thus, xb ∈ a.

(iii) To show that ((a : b) : c) = (a : bc) let x ∈ ((a : b) : c). Then xc ⊆ (a : b). Thus,
xcb ⊆ a. It follows that x ∈ (a : cb) = (a : bc). Conversely let y ∈ (a : bc). Now for
any c ∈ c we have ycb ⊆ ycb = ybc ⊆ a. Hence yc ⊆ (a : b) implying z ∈ ((a : b) : c).
So, ((a : b) : c) = (a : bc). Also, since (yb)c ⊆ a we have yb ∈ (a : c) and hence
y ∈ ((a : c) : b). Thus, (a : bc) ⊆ ((a : c) : b). For the reverse inclusion let z ∈ ((a : c) : b). So,
zb ⊆ (a : c) ⇒ (zb)c = z(bc) ⊆ a ⇒ z ∈ (a : bc).
T T T
(iv) x ∈ ( i ai : b) ⇔ xb ⊆ i ai ⇔ xb ⊆ ai for each i ⇔ x ∈ i (ai : b)
P P
T Let x ∈ (a : i bi ). Then xbTi ⊆ x i bi ⊆ a. Now, xbi ⊆ a for each i implies
(v)
x ∈ i (a : bi ). On the other
Phand x ∈ i (a : bi ) implies x ∈ (a : bi ) P
for each i. That is xbi ⊆ a
for each i which implies x i bi ⊆ a which is equivalent to x ∈ (a : i bi ).

Definition 1.6.3. The radical of an ideal a in a ring A, denoted by r(a) or a, is defined as

r(a) = {x ∈ A : xn ∈ a for some n ∈ N}

11
Proposition 1.6.10. r(a) is an ideal of A.
Proof. Consider the natural projection ϕ : A −→ A/a. Observe that,

x ∈ r(a) ⇔ xn ∈ a ⇔ (x + a)n ∈ a ⇔ ϕ(x)n ∈ a ⇔ ϕ(x) ∈ R(A/a) ⇔ x ∈ ϕ−1 (R(A/a))

By proposition 1.5.1, R(A/a) is an ideal and hence, so is ϕ−1 (R(A/a)).

Proposition 1.6.11. The following properties hold for radical of an ideal.

(i) a ⊆ r(a)

(ii) r(r(a)) = r(a)

(iii) r(ab) = r(a ∩ b) = r(a) ∩ r(b)

(iv) r(a) = (1) ⇔ a = (1)

(v) r(a + b) = r(r(a) + r(b))

(vi) If p is prime then r(pn ) = p for all n > 0

Proof. (i) For any x ∈ a, x = x1 ∈ a.

(ii) Let x ∈ r(r(a)). Then (xn )m ∈ a for some m, n ∈ N. That is, xnm ∈ a. So, x ∈ r(a).
Thus, r(r(a)) ⊆ r(a). Also, r(a) ⊆ r(r(a)) by part (i).

(iii) We show that r(ab) ⊆ r(a ∩ b) ⊆ r(a) ∩ r(b) ⊆ r(ab). Since ab ⊆ a ∩ b we get r(ab) ⊆
r(a ∩ b). Now, if x ∈ r(a ∩ b) then xn ∈ a ∩ b for some n ∈ N. That is, xn ∈ a and
xn ∈ b ⇒ x ∈ r(a) ∩ (b) ⇒ xn ∈ a, xm ∈ b for some m, n ∈ N. ⇒ xn+m = xn xm ∈ ab ⇒
x ∈ r(ab).

(iv) r(a) = (1) ⇔ (1) ⊆ r(a) ⇔ for any x ∈ A, xn ∈ A for some n ∈ N ⇔ 1 ∈ a ⇔ a = (1).

(v) By part (i), a ⊆ r(a) and b ⊆ r(b). This implies a + b ⊆ r(a) + r(b) ⇒ r(a + b) ⊆
r(r(a) + r(b)). Conversely, if x ∈ r(r(a) + r(b)) then xn ∈ r(a) + r(b) for some n ∈ N.
So, xn = pa + qb for some p, q ∈ A, a ∈ r(a), b ∈ r(b). Thus, an1 ∈ a, bn2 ∈ b for some
n1 , n2 ∈ N. Choose m ∈ N such that m > n1 , n2 . Then (xn )m ∈ r(a + b) by considering
the powers of the terms in the binomial expansion of (pa + qb)m .

(vi) Let x ∈ r(pn ). Then xk ∈ p ⊆ p for some k. So, x ∈ p or xk−1 ∈ p. Repeating this
argument at most k − 1 times we get x ∈ p. On the other hand, y ∈ p ⇒ y ∈ pn is clear.

A radical ideal is defined as an ideal whose radical is equal to itself. By part (vi) of the
proposition 1.6.11 every prime ideal is radical. The converse is not true in general; for example,
in Z the ideal 6Z is radical but not prime. More generally, for a subset E of the ring A we may
define r(E) = {x ∈ A : xn ∈ E for some n ∈ N}. It is not necessarily an ideal; for example,
r({x}) is not an ideal in Z[x] as 0 ∈
/ r({x}).
S S
Proposition 1.6.12. For any family of subsets Eα , r ( α Eα ) = α r(Eα ).

12
n
Eα for some n ∈ N ⇒ xn ∈ Eα0 for some α0 . ⇒ x ∈ r(Eα0 ) ⇒
S S
S x ∈ r ( α Eα ) ⇒ x ∈ α S
Proof.
x ∈ α r(Eα ). SConversely, y ∈ S α r(Eα ) ⇒ y ∈ r(Eα0 ) for some α0 ⇒ y m ∈ Eα0 for some
m
m ∈ N ⇒ y ∈ α Eα ⇒ y ∈ r ( α Eα ).
S
Proposition 1.6.13. Let D be the set of all zero divisors of A. Then D = x̸=0 r(Ann(x)).
S  S
Proof. D = r(D) = r x̸=0 Ann(x) = x̸=0 r(Ann(x)).

Examples 1.6.14.
• If A = Z, a = (m), let piT (1 ≤ i ≤ r) be the distinct prime divisors of m. Then we claim
= ri=1 (pi ).
that r(a) = (p1 · · · pr ) T
We shall show r(a) ⊆ ri=1 (pi ) ⊆ (pi · · · pr ) ⊆ r(a). Note that, r(a) ⇒ xn ∈ (m) for some
n > 0 m|xn ⇒ p1 · · · pr |xn ⇒ pi |xn for each n
Tr i ⇒ x ∈ (pi ) for each i ⇒ x ∈ (pi ) for each i
(since each (pi ) is a prime ideal) ⇒ x ∈ i=1 (pi ) ⇒ x ∈ (pi ) · · · (pr ) (since (pi )+(pj ) = (1)
for i ̸= j) ⇒ x ∈ (pi · · · pr ) ⇒ x = x1 ∈ (m) x ∈ r((m)).

Proposition 1.6.15. a and a are coprime if and only if r(a) and r(a) are coprime.
Proof. r(a+b) = r(r(a)+r(b)) = r((1)) = (1) by part (v) of the proposition 1.6.11. Conversely,
a + b = (1) ⇒ r(a) + r(b) because a ⊆ r(a) and b ⊆ r(b).

1.7 Extension and contraction


Definition 1.7.1. Let f : A → B be a ring homomorphism, and a ⊆ A, b ⊆ B be ideals.

• The extension of a denoted by ae is the ideal generated by f (a) in B.

• The contraction of b denoted by bc is f −1 (b) in A.

1.8 Some solved problems


1.1. Let x be a nilpotent element of a ring A. Show that 1 + x is a unit of A. Deduce that the
sum of a nilpotent element and a unit is a unit.
Solution: x ∈ R(A) ⇒ x ∈ J(A). So, 1 − xy ∈ A× for all y ∈ A. In particular, 1 + x ∈ A× . If
u ∈ A× then 1 + xu−1 ∈ A× . Now, x + u = u(xu−1 + 1) ∈ A× .

1.2. A ring A is such that every ideal not contained in the nilradical contains a non-zero
idempotent (that is, an element e such that e2 = e ̸= 0). Prove that the nilradical and
Jacobson radical of A are equal.
Solution: We have R(A) ⊆ J(A) for any ring A. For the reverse containment suppose x ∈ /
R(A). Then (x) ⊈ R(A). Now, there exists 0 ̸= e ∈ (x) such that e2 = e. Write e = xt for
some t ∈ A. But 1 − e is a zero divisor of A since (1 − e)e = 0 and e ̸= 0. So, 1 − e = 1 − xt is
not a unit. Thus, x ∈ J(A) by proposition 1.5.3.

13
1.3. Let A be a ring in which every element x satisfies xn = x for some n > 1 (depending on
x). Show that every prime ideal in A is maximal.
Solution: Let p ⊆ A be a prime ideal and let x ∈ A\p. Then A/p is an integral domain
and x = x + p is a non-zero element in A/p. Now, xn = (x + p)n = xn + p = x + p = x ⇒
x(xn−1 − 1) = 0 for some n > 1. Since A/p is an integral domain and x ̸= 0, we have xn−1 = 1.
Thus, each non-zero element in A/p is a unit; so A/p is a field. Hence p is maximal.

1.4. Let A be a ring ̸= 0. Show that the set of prime ideals of A has minimal elements with
respect to inclusion.
Solution: Let X be the set of all prime ideals of A. Order TX by inclusion. Let T := p1 ⊇
p2 ⊇ ... be any descending chain of ideals in X. Choose p = ∞ i=1 pi . Then p is an ideal of A.
Suppose xy ∈ p and x ∈ / p. Then xy ∈ pi for each i and x ∈ / pj for some j. It follows that
x∈/ pk for each k ≥ j. Since pk is prime, y ∈ pk for each k ≥ j. Since pr ⊆ pj for each r < j,
we have y ∈ pr for each r < j. Thus, y ∈ pi for all i. Hence y ∈ p. So, p ∈ X. (i.e., p is a lower
bound for T in X). By Zorn’s lemma, X has at least one minimal element.

1.5.
Solution:

14
2. Euclidean Domains

There are a number of classes of integral domains with more algebraic structure than the
generic integral domains. Euclidean domains is one such class, which are characterized by a
division algorithm that generalizes the familiar algorithm in the ring of integers. Several other
such classes will be discussed in the next chapter. This chapter begins with a comprehensive
overview of definitions and examples of Euclidean domains; followed by a detailed discussion
of their properties.

2.1 Basic definitions and examples


We begin by introducing the notion of a norm on an integral domain A, which serves as a way
of measuring the “size” of elements in A.
Definition 2.1.1. Any function N : A → Z+ ∪ {0} with N (0) = 0 is called a norm on the
integral domain A. If N (a) > 0 for a ̸= 0 define N to be a positive norm.
Note that it is possible for an integral domain to have different norms.
Definition 2.1.2. The integral domain A is said to be a Euclidean Domain (ED) if there is a
norm N on A such that for any a, b ∈ A with b ̸= 0 there exist q, r ∈ A with a = qb + r with
r = 0 or N (r) < N (b).
Thus, ED is an integral domain that possess a division algorithm. Having a division algo-
rithm on an integral domain A is significant because it enables the use of Euclidean algorithm
to compute the greatest common divisor of two elements in A. We shall discuss this in the
section 2.2.
Examples 2.1.1.
• Any field k is a ED with respect to any norm as for any a, b ∈ k with b ̸= 0, a = qb + 0
where q = ab−1 .
• Z is a ED with the norm given by N (a) = |a|.
• k[x] is a ED when k is a field with respect to the norm N (p(x)) = degree of p(x).
• The Gaussian integers Z[i] is a ED with the norm given by N (a + bi) = a2 + b2 .
For a square-free integer D, the quadratic integer ring OQ(√D) is defined by

15
OQ(√D) = Z[ω] = {a + bω | a, b ∈ Z}
where
(√
if D ≡ 2, 3( mod 4)
D
ω= √
1+ D
.
if D ≡ 1( mod 4)
2
√ √
OQ(√D) is a subring of the quadratic field Q( D) = {a + b D | a, b ∈ Q} and is an integral
domain. Note that when D = −1 we have the ring Z[i] if Gaussian integers.
1

Definition 2.1.3. Let D be a square-free
√ integer. The field norm √of the quadratic field Q( D)
is defined to be the function ϕD : Q( D) → Q given by ϕD (a + b D) = a2 − Db2 .
The field norm
√ defined in the definition 2.1.3 is multiplicative, i.e. ϕD (αβ) = ϕD (α)ϕD (β)
for all α, β ∈ Q( D). The determination of the values of D for which the ring OQ(√D) is a ED
with respect to the restricted field norm ϕD is a central problem in algebraic number theory.
For negative square-free integers we have the following theorem.

Theorem 2.1.2. Let D be a negative square-free integer. Then the following are true.

1. D ≡ 1( mod 4) and OQ(√D) is a ED with respect to the norm given in the definition
2.1.3 if and only if D = −3, −7, −11.

2. D ≡ 2, 3( mod 4) and OQ(√D) is a ED with respect to the norm given in the definition
2.1.3 if and only if D = −1, −2.

Proof. See the proofs given for Theorem 2.2.3 and Theorem 2.2.5 in [1].
The determination of the positive square-free integers D for which OQ(√D) is a ED with
respect to the field norm ϕD in definition 2.1.3 is much more difficult. However, Chatland and
Davenport [4] established the following result in 1950.

Theorem 2.1.3. Let D be a positive square-free integer. Then the following are true.

1. D ≡ 1( mod 4) and OQ(√D) is a ED with respect to the field norm ϕD if and only if
D = 5, 13, 17, 21, 29, 33, 37, 41, 73.

2. D ≡ 2, 3( mod 4) and OQ(√D) is a ED with respect to the field norm ϕD if and only if
D = 2, 3, 6, 7, 11, 19, 57.

2.2 Properties of Euclidean domains


An important implication of division algorithm for an integral domain A is that it forces every
ideal of A to be principal.

Proposition 2.2.1. If a is any non-zero ideal in the Euclidean domain A then a = (a) for
some non-zero element a ∈ a of minimum norm.
1
The field norm of a quadratic field may not be a norm in the sense of definition 2.1.1.

16
Proof. Let a ∈ a be a non-zero element with minimum norm (such an a exists since the set
{N (x) | x ∈ a} ⊂ Z has a minimum element by the Well-ordering of Z). Clearly, (a) ⊆ a as
a ∈ a. For the reverse containment pick any x ∈ a and apply the division algorithm to write
x = qa + r where r = 0 or N (r) < N (a). But then r = x − qa ∈ a. By the minimality of the
norm of a, we see that r = 0. Thus, x = qa ∈ (a) showing a ⊆ (a).
With the aid of the proposition 2.2.1 one can prove that certain integral domains A are not
Euclidean with respect to any norm by showing the existence of an ideal in A which is not
principal. For instance, the polynomial ring Z[x] is not an ED as the ideal (2, x) in Z[x] is not
principal.
One of the prominent properties of a ED is the greatest common divisor of any two non-zero
elements always exists. The notion of greatest common divisor that we have seen in Z can be
made precise to a general ring as follows.

Definition 2.2.1.
Let A be a ring and let a, b ∈ A with b ̸= 0.

(1) b is said to divide a if there exists an element c ∈ A such that a = bc, in this case we
write b|a.

(2) A greatest common divisor of a and b is a non-zero element d such that,

i. d|a and d|b, and


ii. if d′ |a and d′ |b then d′ |d.

A greatest common divisor of a and b will be denoted by gcd(a, b) or simply (abusing


the notation) by (a, b).

We have the following sufficient condition for the existence of greatest common divisor of
two elements of a ring.

Proposition 2.2.2. If a and b are non-zero elements in a ring A such that the ideal generated
by a and b is a principal ideal (d), i.e., (a, b) = (d), then d = gcd(a, b).
Proof. If (a, b) = (d) then (a) ⊆ (d) and (b) ⊆ (d), or equivalently, d|a and d|b. If d′ |a and d′ |b
then (a) ⊆ (d′ ) and (b) ⊆ (d′ ). Thus, (d) = (a, b) ⊆ (a) + (b) ⊆ (d′ ) implying d′ |d.
Proposition 2.2.2 justifies the usage of the notation (a, b) to denote both greatest common
divisor of a and b, and the ideal generated by a and b.
Let a be the ideal of A generated by a and b. The two conditions given in the definition of
greatest common divisor (see (2) of the definition 2.2.1) can be translated into the language of
ideals and therefore become respectively:

i. a is contained in the principal ideal (d), and

ii. if (d′ ) is any principal ideal containing a then (d) is contained in (d′ )

17
In other words, greatest common divisor (if it exists) is a generator for the smallest principal
ideal containing both a and b. There are integral domains with elements whose greatest common
divisors do not exist [7].
The converse of the proposition 2.2.2 does not hold. For, in the ring Z[x], 2 and x generate
a maximal, non-principal ideal; hence (1) is the smallest principal ideal containing both 2 and
x. So, gcd(2, x) = 1.
The next proposition states that in an integral domain, two principal ideals are coincide
precisely when their generators are associates2 .
Proposition 2.2.3. Let A be an integral domain. Then (a) = (b) for some non-zero a, b ∈ A
if and only if a and b are associates.
Proof. Suppose (a) = (b). Then a ∈ (a), and hence a = ub for some u ∈ A. Similarly,
b = va for some v ∈ A. Now, a = uva, and since A is an integral domain and a ̸= 0, we
must have uv = vu = 1. So, u is a unit. Conversely if a = ub for some unit u in A, then
(a) = aA = ubA = buA = bA = (b).
Now let us discuss the uniqueness of greatest common divisors.
Corollary 2.2.4. Let A be an integral domain. Suppose a, b ∈ A and gcd(a, b) exists. Then
gcd(a, b) is unique up to associates.
Proof. Suppose both d and d′ are greatest common divisors of a and b. Then d and d′ generates
the same principal ideal and hence they are associates by the proposition 2.2.3.
Coming back to Euclidean domains we see that greatest common divisors always exist; in
fact, if A is a ED and a, b ∈ A then (a, b) = (d) for some d ∈ A by proposition 2.2.1 also
d = gcd(a, b) by proposition 2.2.2. Furthermore, in a ED greatest common divisor of two
elements can be computed by using the Euclidean algorithm. We shall provide some such
computations at the end of this chapter.
Finally, we explore the existence of universal side divisors, a property possessed by Euclidean
domains that are not fields.
Definition 2.2.2. For any integral domain A let à = A× ∪ {0} denote the collection of units
of A together with 0. An element u ∈ A − Ã is called a universal side divisor if for every x ∈ A
there is some z ∈ Ã such that u divides x − z in A.
Notice that an if u is a universal side divisor in the integral domain A, then any element
x ∈ A can be written as x = qu + z where z is either zero or a unit, i.e., there is a type of
“division algorithm” for u. Thus the existence of universal side divisors is a “weakening” of the
Euclidean condition.
Proposition 2.2.5. Let A be an integral domain that is not a field. If A is a ED then A has
universal side divisors.
Proof. Suppose A is Euclidean with respect to some norm N and let u be an element of A − Ã
(this is non-empty since A is not a field) of minimal norm. For any x ∈ A, we can find q ∈ A
such that x = qu + z with z is either 0 or N (z) < N (u). In either case the minimality of u
implies r ∈ Ã. Hence u is a universal side divisor in A. This completes the proof.
2
In a ring A, two elements a and b are called associates if a = ub for some unit u in A

18
We can show that a certain integral domain A which is not a field, is not Euclidean (with
respect to any norm) by showing that A has no universal side divisors. These types of problems
will be encountered at the end of this chapter.

2.3 Some solved problems


2.1. Let A be a ED. Let m be the minimum integer in the set of norms of non-zero elements
of A. Prove that every non-zero element of A of norm m is a unit. Deduce that a nonzero
element of norm zero (if such an element exists) is a unit.
Solution: Let u ∈ A \ {0} be such that N (u) = m. Since A is a ED we can find q ∈ A such
that 1 = qu + r with r = 0 or N (r) < N (u) = m. Since u has minimal norm we must have
r = 0 and hence 1 = qu implying that u is a unit. Since any non-zero element with zero norm
has minimal norm, the assertion follows.

2.2. Let A be a ED.


(a) Prove that if gcd(a, b) = 1 and a divides bc, then a divides c. More generally, show that
a
if a divides bc with non-zero a, b then gcd(a,b) divides c.

(b) Consider the Diophantine Equation ax + by = N where a, b and N are integers and a, b
are non-zero. Suppose x0 , y0 is a solution: ax0 + by0 = N . Prove that the full set of
solutions to this equation is given by
b a
x = x0 + m , y = y0 − m
gcd(a, b) gcd(a, b)
as m ranges over the integers.

Solution:
(a) Suppose gcd(a, b) = 1 and a|bc in A. Then we can find x, y ∈ A such that 1 = ax + by.
Multiplying by c we get c = acx + bcy. Since a|bc, bc = at for some t ∈ A. Now,
c = acx + aty = a(cx + ty) which tells that a|c in A.
Now suppose a, b are non-zero and a|bc in A. Let d = gcd(a, b). Now we can find x, y ∈ A
such that d = ax + by. Multiplying by c we get cd = acx + bcy. Again, bc = at for some
t ∈ A. Now, cd = acx + aty = a(cx + ty). Since A is an integral domain we may write
a
c = ad (cx + ty) which means gcd(a,b) divides c in A.
 
b a b
(b) Conversely, if x = x0 + m gcd(a,b) and y = y0 − m gcd(a,b) then ax + by = a x0 + m gcd(a,b) +
 
a
b y0 − m gcd(a,b) = ax0 + by0 = N as x0 , y0 is a solution. So, x, y is a solution.


2.3. Consider the quadratic integer ring OQ(√D) which is a subring of the quadratic field Q( D)

with the field norm N (a + b D) = a2 − Db2 . Prove that OQ(√D) is Euclidean with respect to
N for D = −1, −2.

19
Solution: Let D ∈ {−1, −2}. First, note that N is indeed a norm on OQ(√D) . Let α =
√ √
a1 + b1 D, β = a2 + b2 D ∈ OQ(√D with b ̸= 0. Now,
√ √ √
α a1 + b 1 D (a1 + b1 D)(a2 − b2 D) a1 a2 − b1 b2 D2 (b1 a2 − a1 b2 ) √ √
= √ = = + D = r + s D
β a2 + b 2 D a22 − Db22 a22 − Db22 a22 − Db22
a1 a2 −b1 b2 D2 (b1 a2 −a1 b2 )
where r = a22 −Db22
∈ Q and s = ∈ Q. Choose p, q ∈ Z such that |r − p| < 1/2 and
a22 −Db22
√ √
|s − q|√< 1/2. Let θ = (r√
− p) + (s − q) D√and set γ = βθ. Then γ = (r − p)β + (s −√q)β D =
(r + s D)β − pβ − qβ D = a − (p + q D)β. Thus γ ∈ OQ(√D) and a = (p + q D)β + γ.
Note that,  2  2
2 2 1 1 1 + |D|
N (θ) = (r − p) − D(s − q) ≤ + |D| = .
2 2 4
Multiplicativity of the field norm yields,
 
1 + |D|
N (γ) = N (β)N (θ) ≤ N (b) < N (b)
4

as D ∈ {−1, −2}. Thus, a = (p + q D)β + γ where γ ∈ OQ(√D) with N (γ) < N (b).
Note: γ = 0 ⇔ θ = 0 ⇔ r = p and s = q ⇔ r, s ∈ Z.

2.4. Prove that the quadratic integer ring OQ(√D) is not a ED with respect to any norm for
D = −43.
√ ^ √
Solution: We show that Z[ D] = OQ(√D) has no universal side divisors. Note that Z[ D] =
√ × √ √ √
Z[ D] ∪ {0} = {0, ±1}. Let a + b D ∈ Z[ D]. Then N (a + b D) = a + ab + 11b2 = 2
√ 2 2
a2 + ab + 1−D b . For D = −43, N (a + b D) = a2 + ab + 11b2 = a + 2b + 45b
 2
4 4
≥ 12. So, the
smallest non-zero values for N are 1 (for units ±1), 4 (for ±2), 9 (for ±3), 11 (for b = 1, a = 0).
For x = 2, u must divide √ one of 2 − 0 or 2 ± 1. That is u is a non-unit divisor of 2 or 3. If
2 = αβ for some α, β ∈ Z[ D]. Then 4 = N (2) = N (α)N (β). But for this, one of α or β must
has norm 1. i.e., equals ±1. Hence the only divisors of 2 are {±1, ±2}. By a similar argument,
only divisors√ of 3 are {±1, ±3}. Now the set of all possible values for√u are only {±2, ±3}.
Take x = 1+ 2−43 . Then none of x, x ± 1 are divisible by ±2 or ±3 in Z[ D].

2.5. Prove that the quotient ring Z[i]/a is finite for any nonzero ideal a of Z[i].
Solution: Since Z[i] is a ED it is a PID. Let a be any non-zero ideal of Z[i]. Then a = (α)
for some α ∈ Z[i]. Now let β + a ∈ Z[i]/a. By division algorithm there exists γ, δ ∈ Z[i] such
that β = δα + γ with N (γ) < N (α). This implies that β − γ ∈ a, i.e., β + a = γ + a. Thus,
each element of Z[i]/a is represented by an element γ whose norm is not exceeding N (α). In
fact there are finitely many integers a, b such that a2 + b2 ≤ N (α). Therefore, Z[i]/a has only
finitely many elements.

2.6. Let A be a commutative ring and let a, b ∈ A be non-zero. A least common multiple of a
and b is an element e of A such that

(i) a|e and b|e, and

20
(ii) if a|e′ and b|e′ then e|e′ .

(a) Prove that a least common multiple of a and b (if such exists) is a generator for the unique
largest principal ideal contained in (a) ∩ (b).

(b) Deduce that any two nonzero elements in a Euclidean Domain have a least common
multiple which is unique up to multiplication by a unit.
ab
(c) Prove that in a ED the least common multiple of a and b is gcd(a,b)
.

Solution:

(a) Suppose lcm(a, b) exists, say e. Since a|e and b|e we get (e) ⊆ (a) and (e) ⊆ (b) respec-
tively. So, (e) ⊆ (a) ∩ (b). Now, if (e′ ) ⊆ (a) ∩ (b) then a|e′

21
3. Principal Ideal Domains and Unique
Factorization Domains

In the previous chapter, we discussed about EDs, a special class of integral domains. Building
upon that foundation, this chapter delves into the examination of two additional classes of
integral domains: Principal Ideal Domains (PIDs) and Unique Factorization Domains (UFDs).

3.1 Principal Ideal Domains


Definition 3.1.1. A Principal Ideal Domain (PID) is an integral domain where each ideal is
principal.
From the proposition 2.2.1, it follows that EDs are PIDs. Thus, every result for PIDs holds
for EDs. The ring of integers Z is a PID. However, the converse of the proposition 2.2.1 does
not hold. In other words, there are hPIDs √ i
that are not Euclidean. In the upcoming exercises,
1+ 19
we will demonstrate that the ring Z 2
, which was shown not to be an ED in the previous
chapter, is nevertheless a PID. Let A be a PID and let a, b ∈ A be non-zero. We can collect
the facts about greatest common divisors that discussed in the previous chapter.
Let d be a generator for the principal ideal generated by a and b. Then
1. d is a greatest common divisor of a and b
2. d can be written as an A-linear combination of a and b, i.e., there are elements x and y
in A such that d = ax + by.
3. d is unique up to multiplication by a unit of A
Recall that maximal ideals are always prime ideals but the converse is not true in general; but
it is true for non-zero prime ideals in a PID.
Proposition 3.1.1. Every non-zero prime ideal in a PID is maximal.
Proof. Suppose (p) be a non-zero prime ideal in the PID A and let a = (m) be any ideal
containing (p). As p ∈ (m), p = am for some a ∈ A. Since (p) is a prime ideal, we have a ∈ (p)
or m ∈ (p) . If a ∈ (p) then write a = pb. Now, p = am = pbm and thus, bm = 1. i.e. m is a
unit. So, a = A. On the other hand, m ∈ (p) we have (m) = (p). Therefore, (p) is a maximal
ideal in A.

22
Next proposition states that a polynomial ring over a field is a PID.
Proposition 3.1.2. If F is a field then F [x] is a ED and hence is a PID.
Proof. See [6, p. 299].
The converse for the proposition 3.1.2 is also true.
Proposition 3.1.3. If A is a ring such that the polynomial ring A[x] is a PID (or a ED), then
A is a field.
Proof. Suppose A[x] is a PID. Since A is a subring of A[x], A must be an integral domain.
Since A[x]/(x) is isomorphic to the integral domain A, The ideal (x) is a non-zero prime ideal
in A[x]. From the proposition 3.1.1, (x) is a maximal ideal, hence the quotient A[x]/(x) is a
field.
Definition 3.1.2. A positive norm N on an Integral domain A is defined to be a Dedekind-
Hasse norm if for each non-zero a, b ∈ A either a is an element of the ideal (b) or there is a
non-zero element in the ideal (a, b) of norm strictly smaller than the norm of b (i.e., either b
divides a in A or there exist s, t ∈ A with 0 < N (sa − tb) < N (b)).
Observe that the integral domain A is Euclidean with respect to N if it satisfy the condition
given in the definition of Dedekind-Hasse norm with s = 1; so this is a weakening of the
Euclidean condition.
Definition 3.1.3. Almost Euclidean domain is an integral domain that possesses a Dedekind-
Hasse norm.
Proposition 3.1.4. Any almost ED is a PID.
Proof. Suppose A be an almost ED. Let a be any non-zero ideal in A and let b be a non-zero
element of a with minimal norm N (b). Given a ∈ a, for any x, y ∈ A, we have ax + by ∈ a. By
the minimality of b, we cannot have 0 < N (ax + by) < N (b). Since A is almost Euclidean with
respect to N we must have a = bq for some q ∈ A. Thus, a = (b) is principal.

3.2 Unique Factorization Domains


Let’s begin by introducing some terminology.
Definition 3.2.1. Let A be an integral domain.
1. Suppose x ∈ A is non-zero and is not a unit. Then x is called irreducible in A if whenever
x = ab with a, b ∈ A, at least one of a or b must be a unit in A. Otherwise x is said to
be reducible.
2. The nonzero element p ∈ A is called prime in A if the ideal (p) generated by p is a prime
ideal. In other words, a nonzero element p is a prime if it is not a unit and whenever
p|ab for any a, b ∈ A, then either p|a or p|b.
3. Two elements a and b of A differing by a unit are said to be associate in A (i.e., a = ub
for some unit u in A).

23
Next proposition states that primes elements are irreducible in an integral domain.

Proposition 3.2.1. Let A be an integral domain and suppose p ∈ A is prime. Then p is


irreducible in A.
Proof. Suppose (p) is a non-zero prime ideal and p = ab. Then ab = p ∈ (p), thus one of a or
b, say a, is in (p). Thus a = px for some x. This implies p = ab = pxb so xb = 1 and hence b is
a unit. This shows that p is irreducible.
The converse of proposition 3.2.1 is not true in general as we shown in the example below.
√ √
Examples 3.2.2. Consider the ring √ Z[ −5]. First, we show√that 3 is irreducible in Z[ −5].
Suppose 3 = αβ for some α, β ∈ Z[ −5]. Write β = a + b −5 for some a, b ∈ Z. Taking
the norm of the first equation we have 9 = N (α)(a2 + 5b2 ). Since both N (α) and a2 + 5b2 are
integers this forces that a2 + 5b2 = 1, 3 or 9. But, a2 + 5b2 = 3 has no integer solutions. If
a2 + 5b2 = 1 then a = ±1 and hence, β is a unit. If a2 + 5b2 = 9 then N√ √ so, α = 2±1
(α) = 1 and
is a unit. Therefore, 3 is irreducible.
√ is not a prime since (2 + √−5)(2 − −5) = 3 is
But 3 √
divisible by 3, but neither 2 + −5 nor 2 − −5 is divisible by 3 in Z[ −5].
In a PID, notions of prime and irreducible coincide as stated in the following proposition.

Proposition 3.2.3. Let A be a PID. An element p ∈ A is prime if and only if p is irreducible.


Proof. By proposition 3.2.1, any prime element is irreducible. For the converse, suppose p ∈ A
is a prime. Let m be any ideal containing (p); since A is a PID, m = (m) is a principal ideal.
As p ∈ (m), p = am for some a. But p is irreducible so by definition either a or m is a unit.
This means either (p) = (m) or (m) = (1), respectively. Thus the only ideals containing (p) are
(p) or (1), i.e., (p) is a maximal ideal. Since maximal ideals are prime ideals, this completes
the proof.
By the proposition 3.2.3, primes and irreducibles√are same in rings Z and k[x] where k is a
√ irreducible but not a prime in Z[ −2], proposition 3.2.3 gives another way
field. Since 3 is an
to prove that Z[ −2] is not a PID.
Next we give the definition of a UFD.

Definition 3.2.2. A UFD is an integral domain A in which every non-zero non-unit element
a ∈ A has the following two properties:

(1) The element a can be written as a finite product of irreducibles pi of A (not necessarily
distinct): a = p1 p2 · · · pn and

(2) the decomposition in (1) is unique up to associates; i.e., if a = q1 q2 · · · qm is another


factorization of a into irreducibles, then there exists a bijective map ϕ : {1, 2, ..., n} →
{1, 2, ..., m} such that pi is associated to qϕ(i) for each i = 1, 2, ..., n.

Any field k is trivially a UFD since every nonzero element is a unit, so there are no elements
satisfying properties (1) and (2) in the definition 3.2.2. The following are some examples of
integral domains that are not UFDs.

Examples 3.2.4.

24
• Consider the subring S of Gaussian integers given by S = Z[2i] = {a + 2ib : a, b ∈ Z} is
an integral domain but not a UFD: the elements 2 and 2i are irreducibles and 4 ∈ S has
two factorizations 4 = 2 · 2 = −2i · 2i but 2 and 2i are not associates since there is no
unit u in S such that 2 = 2iu.

• The quadratic integer ring Z[ √−5] is not√ a UFD since the element 9 has two different
factorizations 9 = 3 · 3 = (2 − 5)(2 + 5)
Here is the version of proposition 3.2.3 for UFDs.

Proposition 3.2.5. Let A be a UFD. An element p ∈ A is prime if and only if p is irreducible.


Proof. If p ∈ A is a prime then p is irreducible by the proposition 3.2.1. Conversely, let p ∈ A
be irreducible. Suppose p|ab for some a, b ∈ A; i.e., ab = pc for some c ∈ A. Writing a and
b as a product of irreducibles, note that from the last equation and from the uniqueness of
the decomposition into irreducibles of ab that the irreducible element p must be associate to
one of the irreducibles occurring either in the factorization of a or in the factorization of b.
Without loss of generality we may assume that p is associate to one of the irreducibles in the
factorization of a. That is a can be written as a product a = (up)p1 · · · pn for u a unit and some
(possibly empty set of) irreducibles p1 , · · · , pn · But then p divides a, completing the proof.
The following proposition states that the greatest common divisor of two non-zero elements
in a UFD always exists.

Proposition 3.2.6. Let A be a UFD and let a, b ∈ A be non-zero and suppose

a = upe11 pe22 · · · penn and b = vpf11 pf22 · · · pfnn

are prime factorizations for a and b, where u and v are units, the primes p1 , p2 , ..., pn are distinct
and the exponents e1 , fi ≥ 0. Then the element
min(e1 ,f1 ) min(e2 ,f2 )
d = p1 p2 · · · pnmin(en ,fn )

is a greatest common divisor for a and b.


Proof. Exponents of each of the primes occurring in d are no larger than the exponents occurring
in the factorizations of both a and b. Thus, d divides both a and b. Let c be any common
divisor of a and b and let c = q1g1 q2g2 · · · qm gm
be the prime factorization of c. Since each qi divides
c, hence divides a and b. Since qi is a prime, it divides one of the primes pj . In particular, up
to associates the primes occurring in c must be a subset of the primes occurring in a and b; i.e.,
{q1 , q2 , ..., qm } ⊆ {p1 , p2 , ..., pn }. Similarly, the exponents for the primes occurring in c must be
no larger than those occurring in d. This implies that c divides d. Therefore, d = gcd(a, b).
Now, we reach a crucial result that establishes a connection among the rings we have
discussed so far.

Proposition 3.2.7. Every PID is a UFD. In particular, every ED is a UFD.


Proof. See [6, p. 287] for the first part. Second statement follows from the first as every ED is
a PID.

25
Corollary 3.2.8 (Fundamental Theorem of Arithmetic). The ring of integers Z is a UFD.
Proof. Since Z is a ED, the result follows by proposition 3.2.7.
We conclude this section by giving the proof for the converse of the proposition 3.1.4 using
the notion of a UFD.

Proposition 3.2.9. Any PID is an almost ED.


Proof. Let A be a PID. By proposition 3.2.7, A is a UFD. Define the function ϕ : A → N ∪ {0}
by 
 0 if a = 0
ϕ(a) = 1 if a ∈ A×
 n
2 if a ∈ A − Ã and i1 i2 · · · in where ik ’s are irreducibles
where à = A× ∪ {0}. Clearly, ϕ is a positive norm. For any non-zero elements a, b ∈ A, the
ideal generated by a and b is principal by assumption. Say a = (a, b) = (r) for some non-zero
r ∈ A. If b divides a then a = (b). Otherwise, a ̸= (b) and since b ∈ a, b = xr for some x ∈ A.
So, ϕ(r) ≤ ϕ(b). Note that, x is not a unit as a ̸= (b). So, ϕ(x) > 1 and hence ϕ(r) < ϕ(b).
Now we can choose x0 , y0 ∈ R such that r = ax0 + by0 . Therefore, 0 < ϕ(ax0 + by0 ) < ϕ(b) and
hence ϕ is a Dedekind-Hasse norm.

3.3 Some solved problems

26
4. Modules

The notion of a module, obtained by a modest generalization of that of a vector space, is a


fundamental and indispensable tool in commutative algebra. The theory of modules, which
provides a natural extension of the ideas and techniques of linear algebra to more general
structures such as rings and fields, plays an important role in the theory of linear algebra.
In this chapter, we will explore the theory of modules and their various properties, including
operations, finitely generated modules, exact sequences, and the tensor product of modules.

4.1 Modules and module homomorphisms


Definition 4.1.1. Let A be a ring. By an A-module, or a module over A, we shall mean an
additive abelian group M together with a linear action A × M → M , described by (a, x) → ax,
such that the following axioms are satisfied:

1. a(x + y) = ax + ay

2. (a + b)x = ax + bx

3. (ab)x = a(bx)

4. 1x = x

where a, b ∈ A and x, y ∈ M .
Equivalently, M is an abelian group together with a ring homomorphism A → End(M ),
where End(M ) is the ring of endomorphisms of the abelian group M .
When A is a field k, the axioms for an A -module are precisely the same as those for a vector
space V over k, so that vector spaces are just special types of modules which arise when the
underlying ring is a field.
Here are some familiar examples for modules.

Examples 4.1.1.
• A ring A itself is an A-module, where the action of a ring element on a module element
is just the usual multiplication in the ring A. More generally, an ideal a of A is an
A-module.

27
• Every abelian group M can be considered as a Z-module here the action Z × M → M is
given by (m, x) → mx where

a + a + · · · + a (n times)
 if n > 0
na = 0 if n = 0

−a − a − · · · − a (−n times) if n < 0

Conversely, if M is a module over the ring Z, then it is automatically an abelian group,


since any module over Z is a generalization of an abelian group. So, Z-modules are
precisely the abelian groups.
• If A is a unitary ring and n is a positive integer consider the abelian group An of all
n-tuples of elements of A under the component-wise addition
(x1 , ..., xn ) + (y1 , ..., yn ) = (x1 + y1 , ..., xn + yn )
Define a left action A × An → An in the obvious way, namely by
a(x1 , ..., xn ) = (ax1 , ..., axn )
Then An becomes an A-module. Similarly, if k is a field then k n is an k-vector space
which is known as the affine n-space over k.
• Let A = k[x] where k is a field; an A-module is a k-vector space with a linear transfor-
mation [6, p. 340].
Let M and N be A-modules. A map f : M → N is called an A-module homomorphism (or
is A-linear) if it respects the A-module structures of M and N , i.e.,
1. f (x + y) = f (x) + f (y)
2. f (ax) = af (x)
for all x, y ∈ M and a ∈ A. Thus f is a homomorphism of abelian groups which commutes
with the action of each a ∈ A. If A is a field, an A-module homomorphism is known as a linear
transformation of vector spaces. If f : M → N and g : N → L are A-module homomorphisms
so is their composition g ◦ f : M → N . A bijective A-module homomorphism is called an
isomorphism. M ∼ = N means that M and N are isomorphic as A-modules.
The set of all A-module homomorphisms from M to N can be turned into an A-module by
defining addition and scalar multiplication by the rules
1. (f + g)(x) = f (x) + g(x)
2. (af )(x) = af (x)
for all x ∈ M . This A-module is denoted by HomA (M, N ) or just by Hom(M, N ) when the
underlying ring A is clear from the context. It is easy to verify that HomA (M, N ) is indeed an
A-module. Note that when M = N , Hom(M, M ) is the ring of endomorphisms End(M ) which
is non-commutative in general. Suppose M, M ′ , N, N ′ are A-modules and u : M ′ → M, v :
N → N ′ are A-module homomorphisms. Define
u : Hom(M, N ) → Hom(M ′ , N ) by u(f ) = f ◦ u and
v : Hom(M, N ) → Hom(M, N ′ ) by v(f ) = v ◦ f
It is easy to check that u and v are A-module homomorphisms and we say that u and
v are induced from u and v respectively. Moreover, for any A-module M , HomA (A, M ) is
isomorphic to M as A-modules. In fact, the map given by f 7→ f (1), ∀f ∈ HomA (A, M ) is an
isomorphism.

28
4.2 Submodules and quotient modules
Let M be an A-module. By a submodule of M we mean a subgroup N of M which is closed
under the action of the elements in A. Thus, submodules of M are just subsets of M which
are themselves modules under the restricted operations. It is clear that a non-empty subset N
of M is a submodule of M if and only if x − y ∈ M and ax ∈ M for all x, y ∈ M , a ∈ A;
or equivalently, ax + y ∈ M . If A is a field, then submodules are same as subspaces. The
intersection of any family of submodules of M is a submodule of M . Below are a few instances
of submodules:

Examples 4.2.1.
• Considering the ring A as an A-module, the submodules of A are precisely the ideals of
A.

• If M is an abelian group then the submodules of the Z-module M are precisely the sub-
groups of M .

• Let k be a field, V be a k-vector space and let ϕ ∈ End(V ). In the example 4.1.1 We men-
tioned that V is an k[x]-module. Submodules of V are exactly the ϕ-invariant subspaces
[6, p. 341].
Let N be a submodule of the A-module M . The additive, abelian quotient group M/N can
be made into an A-module by defining the action of A by a(x + N ) = (ax) + N for all a ∈ A,
x + N ∈ M/N . It is straightforward to see that this action is well-defined and the axioms for
an A-module are satisfied by M/N . As we saw in ideals, there is a one-to-one order-preserving
correspondence between submodules of M which contain N , and submodules of M/N (the
statement for ideals is a special case).
If f : M → N is an A-module homomorphism, then kernel of f given by ker f = {x ∈
M : f (x) = 0} is a submodule of M ; the image of f given by Imf = {f (x) : x ∈ M } is
a submodule of N ; and the cokernel of f given by Cokerf = N/Imf is a quotient module
of N . If M ′ is a submodule of M such that M ′ ⊆ ker f , then f gives rise to an A-module
homomorphism f : M/M ′ → N , defined by f (x + M ′ ) := f (x). This map f is well-defined
because if x + M ′ = y + M ′ then x − y ∈ M ′ ⊆ ker f , hence f (x − y) = 0, that is f (x) = f (y)
which in turn implies that f (x + M ′ ) = f (y + M ′ ). Note that the kernel of f is ker f /M ′ and f
is onto Imf . The homomorphism f is said to be induced by f . In particular, taking M ′ = ker f ,
we have the following isomorphism of A-modules

M/ ker f ∼
= Imf (4.1)

The equation 4.1 is sometimes known as the first isomorphism theorem for modules.

4.3 Operations on submodules


In this section, we discuss some operations on submodules; sum, intersection and multiplication
by an ideal. It is noteworthy that most of these operations are analogous to operations on ideals
examined in chapter 1.

29
Definition 4.3.1. Let M be an A-module and X be a subset of M . Define ⟨X⟩ to be the
submodule of M generated by X, that is, the intersection of all submodules of M containing X.
Definition 4.3.2. Let M be an A-module and X be a non-empty subset of M . Then x ∈ M
linear combination of elements of X if there exist ai ∈ A, xi ∈ M and n ∈ N such that
is a P
x = ni=1 ai xi . The set of all linear combinations of elements of X will be denoted by LC(X).
Proposition 4.3.1. Let M be an A-module and X be a subset of M . Then
(
{0} if X = ∅
⟨X⟩ =
LC(X) if X ̸= ∅

Proof. Clearly, {0} is the smallest submodule that contains X if X = ∅. Suppose X ̸= ∅. Note
that LC(X) is a submodule of M . Also, X ⊆ LC(X) as x = 1x ∈ LC(X) for any x ∈ X.
Since ⟨X⟩ is the smallest submodule containing X, we have ⟨X⟩ ⊆ LC(X). Conversely, every
linear combination of elements of X belongs to every submodule that contains X and thus we
have LC(X) ∈ ⟨X⟩.
Definition 4.3.3.PLet M be an A-module and let (Mi )i∈I Pbe a family of submodules of M .
Define their sum i∈I Mi to be the set of all finite sums i∈I xi where xi ∈ Mi for all i ∈ I
and almost all the xi (i.e, all but a finite number) are zero.
Proposition
P 4.3.2. Let M be an A-module and let (Mi )i∈I be a family of submodules of M .
Then
P i∈I M
Si is the smallest submodule of M which contains all the Mi . In other words,
i∈I Mi = i∈I Mi .
S P
Proof. A linear combination of elements of i∈I Mi is precisely a sum of the form i∈I xi where
xi ∈ Mi for all i ∈ I and almost all the xi are zero.
If M1 and M2 are two submodules of an A-module M their sum M1 + M2 is the smallest
submodule containing both M1 and M2 ; and their intersection M1 ∩M2 is the largest submodule
of M that contained in both M1 and M2 with respect to inclusion. Thus the submodules of
M form a complete lattice with respect to inclusion in which greatest lower bound is the
intersection and least upper bound is the sum. Parts i. and ii. of the proposition 4.3.3 are
known as the second and third isomorphism theorems respectively.
Proposition 4.3.3.
i. Let M be an A-module. If M1 and M2 are submodules of M then (M1 + M2 )/M1 ∼
=
M2 /(M1 ∩ M2 ).
ii. If L ⊇ M ⊇ N are A-modules then (L/N )/(M/N ) ∼
= (L/M ).
Proof. i. Consider the inclusion homomorphism ι : M2 → M1 + M2 and the projection map
π : M1 + M2 → (M1 + M2 )/M1 . Their composition π ◦ ι : M2 → (M1 + M2 )/M1 is a surjective
homomorphism with ker π ◦ ι = M1 ∩ M2 . Whence the result follows by the first isomorphism
theorem (equation 4.1).
ii. Define θ : L/N → L/M by x + N 7→ x + M then θ is well-defined homomorphism of
L/N onto L/M with ker θ = M/N . Now the result follows by the first isomorphism theorem
(equation 4.1).

30
Although we cannot in general define the product of two submodules, we can define the
product of a submodule by an ideal.
Definition 4.3.4. Let M be an PA-module and let a be an ideal of A. Define their product aM
to be the set of all finite sums ai xi where ai ∈ a and xi ∈ M .
Notice that aM defined in the definition 4.3.4 is a submodule of M . If N and P are
submodules of an A-module M , define (N : P ) to be the set of all a ∈ A such that aP ⊆ N . It
is easy to check that (N : P ) is an ideal of A. If N = 0 is the zero submodule, we have (0 : P );
which is the set of all a ∈ A such that aP = 0; this ideal is called the annihilator of P and
denoted by AnnA P or just by AnnP when the underlying ring A is clear. An A-module M is
said to be faithful if AnnM = 0. If the ideal a is a subset of AnnM we can make M into an
A/a-module by defining an action of A/a on M as follows: for each x ∈ M and a + a ∈ A/a
let (a + a)x = ax. This action is well-defined as if a + a = b + a, then a − b ∈ a ⊆ AnnM , thus
for any x ∈ M , (a − b)x = 0 that is, ax = bx; also one easily checks that this action makes M
into an A/a-module. Moreover, if AnnM = a then M is faithful as A/a-module; in particular
if a ⊆ A is a maximal ideal then M is a A/a-vector space.
Proposition 4.3.4. Let N and P be submodules of an A-module M . Then the following are
true.
i. Ann(N + P ) = AnnN ∩ AnnP .

ii. (N : P ) = Ann((N + P )/N ).


Proof. i. x ∈ Ann(N + P ) ⇔ x(N + P ) = 0 ⇔ xN + xP = 0 ⇔ xN = 0 and xP = 0
⇔ x ∈ AnnN and x ∈ AnnP ⇔ x ∈ AnnN ∩ AnnP
ii. x ∈ (N : P ) ⇔ xP ⊆ N ⇔ xN + xP ⊆ N ⇔ x((N + P )/N ) = 0 in (N + P )/N
⇔ x ∈ Ann((N + P )/N )
If x ∈ M then the set {ax : a ∈ A} is a submodule of M , denoted by Ax or (x). If
M = LC(X) for some X ⊆ M then X is said to be a set of generators of M . M is said to be
finitely generated if there exists some finite X = {x1 , ..., xn } ⊆ M such that M = LC(X); in
this case we write M = (x1 , ..., xn ).

4.4 Direct sum and product


Definition
L 4.4.1. Let (Mi )i∈I be a family of A-modules. Their external direct sum, denoted
by i∈I Mi , is the set of all families (xi )i∈I such that xi ∈ Mi for each i ∈ I and almost all xi
are zero. L
Note that the external direct sum i∈I Mi defined in the definition 4.4.1 can be made into
an A-module by defining addition and scalar multiplication as follows:

(xi ) + (yi ) = (xi + yi ) (4.2)


a(xi ) = (axi ) (4.3)
L
If (Mi )i∈I be a family of submodules of an A-module M , then the direct sum i∈I Mi defined
in the definition 4.4.1 is referred to as internal direct sum of Mi ’s.

31
Definition
Q 4.4.2. Let (Mi )i∈I be a family of A-modules. Their direct product, denoted by
Qof all families (xi )i∈I such that xi ∈ Mi for each i ∈ I.
i∈I Mi , is the set
We can make i∈I Mi into an A-module Q by defining addition and scalar multiplication as
in equations 4.2 and 4.3 for all (xi ), (yi ) ∈ i∈I Mi , a ∈ A. Observe that the difference between
the definitions 4.4.1 and 4.4.2 is that in the definition 4.4.2, we have dropped the constraint
that “almost all xi are zero”. Therefore, direct sum and direct product are the same if the
index set I is finite.
Proposition 4.4.1. The ring A is isomorphic to a direct sum of finitely many ideals of A if
and only if A is isomorphic to a direct product of finitely many rings.
Proof.LSuppose A ∼ = a1 ⊕ · · · ⊕ an where ai is an ideal of A for each i = 1, ..., n. Define
bi = j̸=i aj for each i. then the map ϕi : A → ai given by (a1 , ..., an ) 7−→ ai is a surjective
A-module homomorphism with ker ϕi = bi . Hence A/bi ∼ = ai as A-modules. Qn But each ideal ai

is a ring (isomorphic to A/bi ),Qit follows that A = A/b1 ⊕ · · · ⊕ A/bn = i=1 A/bi .
Conversely, suppose that A ∼ n
= i=1 Ai where Ai ’s are rings. Then the set of all elements of A of
the form (0, Q..., 0, ai .0, ..., 0) with ai ∈ Ai is an ideal ai of A. Since the ring A and each Q ideal ai are
A-modules, i=1 Ai and the direct sum a1 ⊕···⊕an are A-modules. Now, the map ϕ : ni=1 Ai →
n

a1 ⊕ · · · ⊕ an given by ϕ(a1 , ..., an ) = ((a1 , 0, ..., 0), ..., (0, ..., 0, ai , 0, ..., 0), ..., (0, ..., 0, an )) is an
isomorphism of A-modules. Thus the ring A is isomorphic to a direct sum of finitely many
ideals of A.

4.5 Finitely generated modules


In this section, we will introduce free modules and delve into the structure of finitely generated
modules. We will provide proofs for various relevant results, including the Nakayama’s lemma.
Definition 4.5.1. A subset X of an A-module M is called linearly independent if a1 x1 + · · · +
an xn = 0 implies a1 = · · · = an = 0 for all a1 , ..., an ∈ A and x1 , ..., xn ∈ X. If M is generated
by a linearly independent subset X, then X is called a basis of M . A free module is a module
with non-empty basis.
Proposition
L 4.5.1. An A-module M is free if and only if M is isomorphic to an A-module of
the form x∈X Mx where each Mx ∼
= A (as an A-module).
Proof. Suppose M is free and let X ̸= ∅ be a basis for M . For each x ∈ X, define ϕx : Mx → A
by ax 7→ a where Mx is the submodule of M L
generated by x. Then ϕx is an isomorphism of
A-modules. Hence we have, M ∼
L
= x∈X Ax = x∈X Mx .
ψ L φ
Conversely, suppose that M ∼= x∈X Mx with Mx ∼ = A (as an A-module) for each x. Since A is
finitely generated as an A-module (generated by 1), so is Mx (generated by φ−1 (1)). It follows
that the family (ex )x∈X where
(
φ−1 (1) if x = y
(ex )y =
0 if x ̸= y

Mx . Thus, the family (ψ −1 (ex ))x∈X is a basis for M .


L
is a basis for x∈X

32
A free module M with basis X ⊆ M is denoted by A(X) ; the cardinality of X is called the
rank of M . For a fixed non-negative integer n, a free A-module of rank n is an A-module that
is isomorphic to A ⊕ · · · ⊕ A (n summands), which is denoted by An ; by convention A0 is the
zero module. The next proposition gives universal property of free modules.
Proposition 4.5.2. For any set S there is a free A-module F (S) on S and F (S) satisfies the
following universal property: if M is any Amodule and φ : S → M is any map of sets, then
there exists a unique A-module homomorphism h : F (S) → M such that the following diagram
commutes.
ι
S F (S)

φ ∃! h
M

Proof. If S = ∅ then take F (S) to be the zero module. Suppose S ̸= ∅. Let F (S) = {f :
S → A : f (s) = 0 for almost all s ∈ S}. Make F (S) into an A-module by defining point-wise
addition and point-wise multiplication by the elements of A as
(f + g)(s) = f (s) + g(s)
(af )(s) = af (s)

for all a ∈ A, s ∈ S and f, g ∈ F (S). If s ∈ S define fs ∈ F (S) as


(
1 if x = s
fs (x) :=
0 otherwise

If f ∈ F (S) then there are s1 , ..., sn ∈ S such that f = f (s1 )fs1 + · · · + f (sn )fsn . Note that
f (si ) ∈ A and fsi ∈ F (S) for each i. Since this representation is unique, {fs : s ∈ S} is a basis
for F (S). Thus, F (S) is a free module. For the universal property, suppose φ : S → M . Define
h : F (S) → M so that !
Xn n
X
h ai f s i = ai φ(si )
i=1 i=1
.
Then h is a well-defined homomorphism. By definition, the restriction of h to S equals φ.
Finally, since F (S) is generated by S, once we know the values of an A-module homomorphism
on S, its value on every element of F (S) are uniquely determined. So, h is the unique extension
of φ to all of F (S).

Proposition 4.5.3. Let M be an A-module. Then M is finitely generated if and only if M is


isomorphic to a quotient of An for some integer n > 0.
Proof. Suppose M = (x1 , ..., xn ). Define ϕ : An −→ M by (a1 , ..., an ) 7−→ a1 x1 + · · · + an xn .
Then ϕ is an A-module homomorphism onto M . Hence, An / ker ϕ ∼ = M.
∼ n
For the other direction, suppose M = A /K for some n > 0 and for some submodule K of

33
An . Let ei = (0, ..., 0, 1, 0, ..., 0) (1 being at the i th place). Then ei (1 ≤ i ≤ n) generate An .
Consider the natural projection π : An −→ An /K which is onto An /K. Then π(ei ) (1 ≤ i ≤ n)
generate An /K. Thus, M is finitely generated.
Proposition 4.5.4. Let M be a finitely generated A-module, let a be an ideal of A and let
ϕ ∈ End(M ) be such that ϕ(M ) ⊆ aM . Then ϕ satisfies an equation of the form
ϕn + a1 ϕn−1 + · · · + an = 0
where ai ∈ a for each i = 1, ..., n.
Proof. Let M =P (x1 , ..., xn ). Since ϕ(M ) ⊆ aM , we have ϕ(xi ) ∈ aM for each i. Thus for each
i, write ϕ(xi ) = nj=1 aij xj where aij ∈ a; this can also be written as
n
X
(δij ϕ − aij )xj = 0 (4.4)
j=1

where δij is the Kronecker delta. Let D be the n × n matrix whose ij th entry is δij ϕ − aij .
Then the system of equations in 4.4 can be written as
Dx = 0 (4.5)
   
x1 0
 x2  0
where x =  ..  and 0 =  .. . Multiplying left of the equation 4.5 by adjoint of D, we get
   
. .
xn 0
det(D)Ix = 0. This implies that det(D)I ∈ End(M ) annihilates each xi . But xi (1 ≤ i ≤ n)
generate M , hence det(D)I should be the zero endomorphism. Expanding out the determinant
of D, we have an equation of the required form.
Corollary 4.5.5. Let M be a finitely generated A-module, let a be an ideal of A such that
aM = M . Then there exists x ≡ 1( mod a) such that xM = 0.
Proof. Choose ϕ to be the identity map. Then ϕ ∈ End(M ) and ϕ(M ) = M = aM . By
proposition 4.5.4 we have (ϕn +a1 ϕn−1 +···+an )(y) = 0 for all y ∈ M ; that is, y+a1 y+···+an y =
0 for all y ∈ M ; or equivalently, xy = 0 for all y ∈ M where x = 1 + a1 + · · · + an . Thus, x ≡ 1(
mod a) and xM = 0.
Theorem 4.5.6 (Nakayama’s lemma). Let M be a finitely generated A-module and a be an
ideal of A contained in the Jacobson radical J(A) of A. Then aM = M implies M = 0.
First proof. By corollary 4.5.5 there exists x ∈ A such that x ≡ 1( mod a) and xM = 0. Now,
1 − x ∈ a ⊆ J(A). By proposition 1.5.3, x is a unit in A. So, M = x−1 xM = x−1 0 = 0.
Second proof. To the contrary, assume M ̸= 0. Let {x1 , ..., xn } be a minimal set of generators
of M . Since xn ∈ M = aM , write xn = a1 x1 + · · · + an xn for some ai ∈ a(1 ≤ i ≤ n). Now,
(1 − an )xn = a1 x1 + · · · + an−1 xn−1 . Because an ∈ J(A), by proposition 1.5.3 1 − an is a unit in
A. Let (1 − an )−1 = b. Thus, xn = b(1 − an )xn = ba1 x1 + · · · + ban−1 xn−1 . But bai ∈ a for each
i = 1, ..., n − 1. So, xn belongs to a submodule of M generated by x1 , ..., xn−1 contradicting the
fact that {x1 , ..., xn } is a minimal set of generators of M .

34
Corollary 4.5.7. Let M be a finitely generated A-module, N ⊆ M a submodule, a ⊆ J(A) an
ideal. If M = aM + N then M = N .
Proof. Note that, a(M/N ) = a((aM +N )/N ) = (aM +N )/N = M/N . Hence, by theorem 4.5.6 M =
N.
Let A be a local ring, m be its maximal ideal, and k = A/m be its residue field. Let M be
a finitely generated A-module. Then m ⊆ Ann(M/mM ) as m(x + mM ) = mx + mM = 0 in
M/mM for all x + mM ∈ M/mM . Hence, M/mM is an A/m-module. That is, M/mM is an
k-vector space V . Moreover, V is finite dimensional and we have the following result on V .

Proposition 4.5.8. Let xi (1 ≤ i ≤ n) be the elements of M whose images in M/mM form a


basis for V . Then xi generate M .
ι π
Proof. Let N = (x1 , ..., xn ) ⊆ M . We show that N = M . Consider the mapping N −→ M −→
M/mM . Then π ◦ ι maps N onto M/mM . Thus, N + M/mM = M . Hence, N = M by
proposition 4.5.7.

4.6 Exact sequences


Definition 4.6.1. A sequence of A-modules and A- homomorphisms is a diagram of the form
fi fi+1
· · · −→ Mi−1 −→ Mi −−→ Mi+1 −→ · · · (4.6)

Such a sequence is said to be exact at Mi if Imfi = ker fi+1 , and to be exact if it is exact at each
Mi ; and semi exact at Mi if Imfi ⊆ ker fi+1 (or equivalently, fi+1 ◦ fi = 0), and semi exact if
it is semi exact at each Mi .

It is immediate from the definition that


f
• 0 −→ M ′ −→ M is exact if and only if f is injective, and
g
• M −→ M ′′ −→ 0 is exact if and only if g is surjective.

Exact sequences of the form


f g
0 −→ M ′ −→ M −→ M ′′ −→ 0 (4.7)

are called short exact sequences. Observe that a sequence of the form 4.7 is short exact if and
only if f is injective, g is surjective, and Imf = ker g (or equivalently g induces an isomorphism
of Cokerf onto M ′ ). Any exact sequence 4.6 can be split into short exact sequences 0 −→
Ni −→ Mi −→ Ni+1 −→ 0 by taking Ni = Imfi = ker fi+1 for each i.

Proposition 4.6.1 (The four lemma). Suppose the following diagram of A-modules and A-
module homomorphisms

35
f g h
K L M N
α β γ δ
f′ g′ h′
K′ L′ M′ N′
commutes with rows exact. Then the following hold.
i. if α, γ are surjective and δ is injective then β is surjective.

ii. if β, δ are injective and α is surjective then γ is injective.


Proof. i. Let l′ ∈ L′ . Since γ is surjective, there exists m ∈ M such that γ(m) = g ′ (l′ ). By the
commutativity of the right-hand square, δ[h(m)] = h′ [γ(m)] = h′ [g ′ (l′ )] = 0. Thus h(m) ∈ ker δ.
But since δ is injective, h(m) = 0 implying m ∈ ker h = Img. Thus, m = g(l) for some l ∈ L.
Now, g ′ (l′ ) = γ(m) = γ[g(l)] = g ′ [β(l)] by the commutativity of the middle square. This gives
g ′ [l′ − β(l)] = 0 which means l′ − β(l) ∈ ker g ′ = Imf ′ . Then l′ − β(l) = f ′ (k ′ ) for some k ′ ∈ K ′ .
Since α is surjective choose k ∈ K such that α(k) = k ′ and by the commutativity of the left-
hand square, l′ − β(l) = f ′ (k ′ ) = f ′ [α(k)] = β[f (k)]. Hence, l′ = β[f (k)] + β(l) + β[f (k) + l].

ii. Let m ∈ ker γ. Then γ[h(m)] = h′ [γ(m)] = h′ (0) = 0 implying h(m) ∈ ker δ. Now,
h(m) = 0 as δ is injective. So, m ∈ ker h = Img. Thus, m = g(l) for some l ∈ L. From
the middle square, 0 = γ(m) = γ[g(l)] = g ′ [β(l)]. This gives, β(l) ker g ′ = Imf ′ and hence
β(l) = f ′ (k ′ ) for some k ′ ∈ K ′ . Since α is surjective, k ′ = α(k) for some k ∈ K. Now,
β(l) = f ′ [α(k)] = β[f (k)]. Because β is injective we must have l = f (k). Finally, m = g(l) =
g[f (k)] = 0.
Proposition 4.6.2 (The five lemma). Suppose the following diagram of A-modules and A-
module homomorphisms

M1 M2 M3 M4 M5
α1 α2 α3 α4 α5

N1 N2 N3 N4 N5

commutes with rows exact. If α1 , α2 , α4 , α5 are isomorphisms then so is α3 .


Proof. Considering the three squares in the left-hand side, from the part i. of the proposition
4.6.1 we get α3 is surjective. Considering the three squares in the right-hand side, from the
part ii. of the proposition 4.6.1 we get α3 is injective.
Corollary 4.6.3. Suppose the following diagram of A-modules and A-module homomorphisms

0 M2 M3 M4 0
α β γ

0 N2 N3 N4 0

commutes with rows exact. If α, γ are isomorphisms then so is β.


Proof. Take M1 = N1 = M5 = N5 = 0 in the proposition 4.6.2.

36
Proposition 4.6.4.
i. Let
f g
M ′ −→ M −→ M ′′ −→ 0 (4.8)
be a sequence of A-modules and homomorphisms. Then the sequence 4.8 is exact if and
only if for all A-modules N , the sequence
g f
0 −→ Hom(M ′′ , N ) −→ Hom(M, N ) −→ Hom(M ′ , N ) (4.9)

is exact.

ii. Let
f g
0 −→ N ′ −→ N −→ N ′′ (4.10)
be a sequence of A-modules and homomorphisms. Then the sequence 4.10 is exact if and
only if for all A-modules M , the sequence
g f
0 −→ Hom(M, N ′ ) −→ Hom(M, N ) −→ Hom(M, N ′′ ) (4.11)

is exact.
Proof. Suppose that the sequence 4.8 is exact.

4.7 Tensor product of modules


In this section, we study the tensor product of modules over a ring (which is commutative
and contains 1, as always). Loosely speaking, taking the tensor product of two modules is a
construction that forms another module, in which one can take “products” of two elements.
We will prove the universal property of tensor products and also discuss various properties of
tensor products.

Definition 4.7.1. Let M, N and P be A-modules. A mapping f : M × N −→ P is called


A-bilinear if for each x ∈ M the map y 7−→ f (x, y) from N into P is A-linear and for each
y ∈ N the map x 7−→ f (x, y) from M into P is also A-linear, i.e.,

f (x, a1 y1 + a2 y2 ) = a1 f (x, y1 ) + a2 f (x, y2 ) for all y1 , y2 ∈ N and a1 , a2 ∈ A


and
f (a1 x1 + a2 x2 , y) = a1 f (x1 , y) + a2 f (x2 , y) for all x1 , x2 ∈ M and a1 , a2 ∈ A.

The motivation for tensor product of two A-modules M and N is to construct an A-module
T with the property that the A-bilinear mappings from M × N into P are in a natural one
to one correspondence with the A-linear mappings from T into P , for any A-module P , more
precisely we have the following theorem:

37
Theorem 4.7.1. Let M and N be A-modules. Then there exist a pair (T, g) consisting of
an A-module T and an A-bilinear map g : M × N −→ T such that for given any A-module
P and any A-bilinear map f : M × N −→ P , there exist a unique A-module homomorphism
f ′ : T −→ P such that the following diagram commutes
f
M ×N P
g ⟲
∃! f ′
T
Moreover if (T ′ , g ′ ) is another pair with this property, then there exist a unique isomorphism
h : T −→ T ′ such that such that h ◦ g = g ′ .
Proof. Existence: Let F be the free A-module with basis the set M × N (i.e., F = A(M ×N ) ).
The elements of F are formal A-linear
Pn combinations of finitely many elements of M × N . i.e.,
an element in F is of the form i=1 ai (xi , yi ) where ai ∈ A, xi ∈ M and yi ∈ N . Let D be the
submodule of F generated by all elements of F of the following forms:
(x + x′ , y) − (x, y) − (x′ , y)
(x, y + y ′ ) − (x, y) − (x, y ′ )
(ax, y) − a(x, y)
(x, ay) − a(x, y)
where x, x′ ∈ M and y, y ′ ∈ N , and a ∈ A. Let T denote the quotient module F/D and
let x ⊗ y denotes the equivalence class to which the basis element (x, y) ∈ F belongs to (i.e.,
x ⊗ y = (x, y) + D. Now, T is generated by such elements and hence an arbitrary element of
T is of the form
Xn n
X n
X
ai (xi ⊗ yi ) = ai ((xi , yi ) + D) = ai (xi , yi ) + D.
i=1 i=1 i=1

For any x, x′ ∈ M and y, y ′ ∈ N we have (x + x′ , y) − (x′ , y) − (x, y) ∈ D =⇒ (x + x′ , y) −


(x′ , y) − (x, y) + D = D =⇒ (x + x′ , y) + D = (x, y) + D + (x′ , y) + D, or equivalently
(x + x′ ) ⊗ y = x ⊗ y + x′ ⊗ y. (4.12)
Moreover, if a ∈ A we get (ax, y) − a(x, y) ∈ D =⇒ (ax, y) − a(x, y) + D = D =⇒ (ax, y) + D =
a(x, y) + D. In other words,
(ax) ⊗ y = a(x ⊗ y). (4.13)
A similar argument shows that
x ⊗ (y + y ′ ) = x ⊗ y + x ⊗ y ′ (4.14)
and
x ⊗ (ay) = a(x ⊗ y) (4.15)
Define the map g : M × N −→ T by (x, y) 7−→ x ⊗ y. Now, equations 4.12, 4.13, 4.14 and 4.15
immediately imply that g is A-bilinear. Let P be any A-module and f : M × N −→ P be any
A-bilinear map. Since F is a free module with basis M × N , by Proposition 4.5.2 there exists
a unique A-module homomorphism f˜ : F −→ P such that the following diagram commutes

38
ι
M ×N F
∃! f˜
f
P

Now define f ′ : T −→ P by f ′ (x ⊗ y) = f (x, y) = f˜(x, y). It requires to prove that f ′ is


well-defined. For that we first show that f vanishes on all generators of D. Since f˜ is A-linear
and f is A-bilinear it follows that, f˜ ((x + x′ , y) − (x, y) − (x′ , y)) = f˜(x + x′ , y) − f˜(x, y) −
f˜(x′ , y) = f (x + x′ , y) − f (x, y) − f (x′ , y) = f (x, y) + f (x′ , y) − f (x, y) − f (x′ , y) = 0. Also note
that, f˜ ((ax, y) − a(x, y)) = f˜(ax, y) − af˜(x, y) = f (ax, y) − af (x, y) = af (x, y) − af (x, y) =
0 for each a ∈ A. Similarly it can be shown that f˜ ((x, y + y ′ ) − (x, y) − (x, y ′ )) = 0 and
f˜ ((x, ay) − a(x, y)) = 0. Thus, f˜ vanishes on the whole of D. Now if x ⊗ y = x′ ⊗ y ′ then
(x, y) + D = (x′ , y ′ ) + D which implies (x, y) − (x′ , y ′ ) ∈ D and hence f˜(x, y) = f˜(x′ , y ′ ). That
is, f ′ (x ⊗ y) = f ′ (x′ ⊗ y ′ ) proving that f ′ is well-defined.
Next we verify that f ′ is a homomorphism. For that we show f ′ is A-linear on the generators
of T . Let x ⊗ y and x′ ⊗ y ′ be any two generators of T . Then, f ′ ((x ⊗ y) + (x′ ⊗ y ′ )) =
f ′ (((x, y) + D) + ((x′ , y ′ ) + D)) = f ′ ((x + x′ , y + y ′ ) + D) = f˜(x + x′ , y + y ′ ) + f˜(D) =
f˜((x, y) + (x′ , y ′ )) + f˜(D) = f˜(x, y) + f˜(D) + f˜(x′ , y ′ ) + f˜(D) = f˜((x, y) + D) + f˜((x′ , y ′ ) + D) =
f ′ ((x, y)+D)+f ′ ((x′ , y ′ )+D) = f ′ (x⊗y)+f ′ (x′ ⊗y ′ ) and f ′ (a(x⊗y)) = f ′ (ax⊗y) = f (ax, y) =
af (x, y) = af ′ (x ⊗ y) for each a ∈ A. This implies that f ′ is a homomorphism. Next, for any
(x, y) ∈ M × N , f (x, y) = f ′ (x ⊗ y) = f ′ (g(x, y)). That is, f = f ′ ◦ g. Furthermore, the
homomorphism f ′ is uniquely defined by the generators x ⊗ y of F .
Uniqueness: Suppose that (T ′ , g ′ ) is another pair with this property. Then by replacing (P, f )
by (T ′ , g ′ ) we must have a unique A-module homomorphism h : T −→ T ′ such that g ′ = h ◦ g.
Interchanging the roles of T and T ′ we get an A-module homomorphism h′ : T ′ −→ T such
that g = h′ ◦ g ′ . Now, g = h′ ◦ h ◦ g and g ′ = h ◦ h′ ◦ g ′ . i.e., h and h′ make the following
diagrams commutative.

g g′
M ×N T M ×N T′
g′ g
h h′
idT idT ′
g T′ T
g′
h′ h

T T′

Since the identity maps idT and idT ′ also make the above diagrams commute, we must have
h′ ◦ h = idT and h ◦ h′ = idT ′ . That is h and h′ are inverses of each other and hence h is an
isomorphism.

39
The module T constructed in the Theorem 4.7.1 is called the tensor product of M and N , and
is denoted by M ⊗A N , or just by M ⊗ N when the underlying ring is clear. This A-module T
is generated by the “products” x ⊗ y where x ∈ M and y ∈ N . If (xi )i∈I , (yj )j∈J are families of
generators of M and N respectively, then the elements xi ⊗ yj generate M ⊗ N . In particular,
if M and N are finitely generated, so is M ⊗ N .

From the proof for the Theorem 4.7.1, for any x, x′ ∈ M, y, y ′ ∈ N and a ∈ A we have the
following:

(i) (x + x′ ) ⊗ y = x ⊗ y + x′ ⊗ y

(ii) x ⊗ (y + y ′ ) = x ⊗ y + x ⊗ y ′

(iii) (ax) ⊗ y = a(x ⊗ y) + x ⊗ (ay)

In particular, setting a = 0 in part (iii), we get 0 ⊗ y = 0 = x ⊗ 0.

The notation x ⊗ y is ambiguous unless we specify the tensor product to which it belongs to.
For instance, if x ∈ M ′ and y ∈ N ′ where M ′ and N ′ are submodules of M and N respectively,
then it can be happen that x⊗y is zero in M ⊗N but z ⊗y is a non-zero element in M ′ ⊗N ′ . To
illustrate, set A = Z, M = Z, N = Z2 , M ′ = 2Z and N ′ = N . Take x = 2 and y = 1. Viewing
x ⊗ y as an element of M ⊗ N we get x ⊗ y = 2 ⊗ 1 = 2 · 1 ⊗ 1 = 2(1 ⊗ 1) = 1 ⊗ 2 · 1 = 1 ⊗ 0 = 0.
But if x ⊗ y = 0 in M ′ ⊗ N ′ then as 2 ⊗ 1̄ generates M ′ ⊗ N ′ , it must be the zero module
contradicting the Theorem 4.7.1.

However, we have the following result.


P
Proposition 4.7.2. If xi ∈ M, yi ∈ N such that xi ⊗ yi =P0 in M ⊗ N then there exist
finitely generated submodules M ′ of M and N ′ of N such that xi ⊗ yi = 0 in M ′ ⊗ N ′ .
P
Proof. We use P the notations used in thePproof of Theorem 4.7.1. Suppose xi ⊗ yi = 0 in
M ⊗ N . Then xi ⊗ yi ∈ D and hence xi ⊗ yi can be written as a finite sum of generators
of D. Let M ′ be the submodule of M generated by xi ’s and all the elements of M which occur
as the first coordinates of these generators of D, and define N ′ similarly (i.e., by including ”all
the elements of M which occur as first coordinates in these generators of D”, it ensures that
M ′ contains enough elements to be closed under the module operations, and similarly for N ′ .).
xi ⊗ yi = 0 in M ′ ⊗ N ′ .
P
Then

We can generalize the Theorem 4.7.1 into multilinear mappings f : M1 × · · · × Mr which yields
to “multi-tensor product” T = M1 ⊗ · · · ⊗ Mr that generated by all products x1 ⊗ · · · ⊗ xr where
xi ∈ Mi for i = 1, ..., r. More precisely we have:

40
4.8 Some solved problems
4.1. If M is a finite abelian group then M is naturally a Z-module. Can this action be extended
to make M into a Q-module?
Solution: No, not always. Consider the Z-module Z/2Z. If this were naturally a Q-module,
then it would have some element 21 · 1. This element would satisfy
 
1 1 1 1
·1+ ·1= + ·1=1·1=1
2 2 2 2

and in particular, it would have an order of at least three as an element of the group Z/2Z.
This is not possible. More generally, for any finite abelian group G, one can consider the action
1
of |G| to derive a contradiction. Thus, a finite abelian group never has a Q action compatible
with the natural Z action.
However, if an abelian group is divisible, then we can extend its natural Z action to a Q
action. Of course, nonzero divisible abelian groups are necessarily infinite, so this falls outside
the scope of the problem.

4.2. An element m of the A-module M is called a torsion element if am = 0 for some nonzero
element a ∈ A. The set of torsion elements is denoted

Tor(M ) = {m ∈ M | rm = 0 for some nonzero a ∈ A}

(a) Prove that if A is an integral domain then Tor(M ) is a submodule of M (called the torsion
submodule of M ).

(b) Give an example of a ring A and an A-module M such that Tor(M ) is not a submodule.
(Consider the torsion elements in the A-module A.)

(c) If A has zero divisors show that every nonzero A-module has nonzero torsion elements.
Solution: (a) Let A be an integral domain and observe that Tor(M ) is nonempty since it
contains zero. Then let x, y ∈ Tor(M ) and let a1 , a2 ∈ A be nonzero so that a1 x = 0 and
a2 y = 0. For an arbitrary a ∈ A we can notice that

a1 a2 (x + ay) = a1 a2 x + a1 a2 ay = a2 ra x + a1 aa2 y = a2 · 0 + a1 r · 0 = 0 + 0 = 0

Furthermore, observe that a1 a2 is nonzero since A is an integral domain, and so x+ay ∈ Tor(M ).
This proves that Tor(M ) is a submodule.

(b) Consider Z/6Z. The torsion elements of this ring as a module over itself are {0, 2, 3, 4}
which do not even form an additive subgroup, much less a submodule.

(c) Suppose A has zero divisors and let x, y ∈ A be nonzero so that xy = 0. Then for some
nonzero m ∈ M consider ym. If ym = 0 then m is a nonzero torsion element. Otherwise ym is
a nonzero torsion element since x(ym) = (xy)m = 0m = 0.

41
4.3. Show that the intersection of any nonempty collection of submodules of an A-module is a
submodule.
Solution: Let M be an A-module and let {Nα } be an arbitrary collection of submodules of M .
Let N = ∩α Nα . Notice that N is nonempty since each Nα must contain zero. Let x, y ∈ N .
Since each Nα is a submodule we have ax + y ∈ Nα for all a ∈ A and all α. We conclude that
x + ry ∈ N and hence N is a submodule of M .

4.4. Let N1 ⊆ N2 ⊆ · · · be an ascending chain of submodules of M . Prove that ∪∞


i=1 Ni is a
submodule of M .
Solution: Let N = ∪∞ i=1 Ni . Note that 0 ∈ N so N is nonempty. Then let x, y ∈ N . There
must exist Ni so that x, y ∈ Ni and by virtue of Ni being a submodule we will have ax + y ∈ Ni
for all a ∈ A and hence ax + y ∈ N . This proves that N is a submodule.

4.5. Let n ∈ Z+ , n > 1 and let A be the ring of n × n matrices with entries from a field F .
Let M be the set of n × n matrices with arbitrary elements of F in the first column and zeros
elsewhere. Show that M is a submodule of A when A is considered as a left module over itself,
but M is not a submodule of A when A is considered as a right A-module.
Solution: It is clear that M is an additive subgroup of the module A. When A acts on M from
the left M is invariant since the i-th column of am for a ∈ A and m ∈ M is just the product
of a with the i-th column in m. For i > 1 this column is zero and so must be a’s product with
it. Hence am ∈ M .
On the other hand when A acts from the right the columns in ma beyond the first may
nonzero, as illustrated by the small example below.
    
1 0 1 1 1 1
= ∈
/ M.
1 0 1 1 1 1

42
5. Localization

Localization introduces ”denominators” to rings or modules. It constructs a new ring/module


from an existing one. Localization is fundamental in algebraic geometry, linking to sheaf theory,
and originates from studying geometric objects locally near points.

5.1 Rings of Fractions


A well-known theorem states that each integral domain can be embedded in a field (See Theorem
11.7.2 in [2]). The proof of this theorem is analogous to the procedure by which one construct
the rational field Q from the ring of integers Z. In this section, we generalize this notion into
an arbitrary ring.

Definition 5.1.1. Let A be a ring. A multiplicatively closed subset S of A is a monoid under


multiplication with 1 ̸= 0.

Examples 5.1.1.
• For any f ∈ A, S = {f n }n≥0 = {1, f, f 2 , f 3 , ...} ⊆ A is multiplicatively closed.

• If p is a prime ideal of A then A/p is multiplicatively closed. In particular, A/{0} is


multiplicatively closed in an integral domain.

Proposition 5.1.2. Let A be a ring, S ⊂ A be multiplicatively closed, and a ⊂ A an ideal. If


S ∩ a = ∅, then there exists a prime ideal p containing a with S ∩ p = ∅.
Proof. Let
Σ := {b ⊂ A is ideal : S ∩ b = ∅ and a ⊆ b}
Order
S Σ by inclusion. Then Σ ̸= ∅ as a ∈ Σ. For any chain (bα ) of ideals in Σ, we have the ideal
α bα ∈ Σ and b is is an upper bound of the chain. By Zorn’s lemma, Σ possesses a maximal
element p. It remains to show that p is a prime ideal in A. To this end, let xy ∈ p and suppose
x∈ / p and y ∈ / p. Then [(x) + p] ∩ S ̸= ∅ and [(y) + p] ∩ S ̸= ∅. Since S is multiplicatively
closed, [(x) + p][(y) + p] ∩ S ̸= ∅. But, [(x) + p][(y) + p] = (xy) + (x)p + (y)p + p2 ⊆ p implying
S ∩ p ̸= ∅ which is a contradiction. Hence, x ∈ p or y ∈ p.

43
Let A be a ring, S ⊆ A be multiplicatively closed. Define a relation ∼ on A × S by

(a, s) ∼ (b, t) ⇔ (at − bs)x = 0 for some x ∈ S

This relation is clearly reflexive and symmetric. If (a, s) ∼ (b, t) and (b, t) ∼ (c, u) then
(at − bs)x = 0 and (bu − ct)y = 0 for some x, y ∈ S. Multiplying the first equation by uy and
the second by xs and adding gives (au − cs)txy = 0. As S is multiplicatively closed, txy ∈ S
and hence, (a, s) ∼ (c, u). So, ∼ is transitive.

Let a/s denote the equivalence class of (a, s) under ∼ and let S −1 A be the set of these equiva-
lence classes. Define addition and multiplication on S −1 A by

(a/s) + (b/t) = (at + bs) /st and (a/s) (b/t) = ab/st

It can be easily verified that these operations are well-defined and make S −1 A into a ring with
1 = 1/1. For each s ∈ S, s/1 is a unit in S −1 A. The ring S −1 A is called the ring of fractions
of A with respect to S.

We can also define f : A −→ S −1 A by f (x) = x/1 for each x ∈ A. It is evident that f is a


ring homomorphism. This is not in general injective. In fact, consider S = {1, 2, 4} in A = Z6 .
Then S −1 A = {0/1, 1/1, 2/1} ∼= Z3 ; and hence f is not injective. Furthermore, in this ring of
fractions S −1 A, (0, 1) ∼ (0, 2) ∼ (0, 4) ∼ (3, 1) ∼ (3, 2) ∼ (3, 4), and (1, 1) ∼ (1, 4) ∼ (2, 2) ∼
(4, 1) ∼ (4, 4) ∼ (5, 2), and (2, 1) ∼ (1, 2) ∼ (2, 4) ∼ (4, 2) ∼ (5, 1) ∼ (5, 4).

Proposition 5.1.3. If A is an integral domain and S is any multiplicatively closed subset not
containing 0 then f is injective.
Proof. If x1 /1 = x2 /1 with x1 , x2 ∈ A then (x1 , 1) ∼ (x2 , 1) which means (x1 − x2 )u = 0 for
some u ∈ S. Since A is an integral domain and u ̸= 0 we get (x1 − x2 ) = 0. That is, x1 = x2
and hence f is injective.
If A is an integral domain and S = A − {0} then for each a/s ∈ S −1 A with a ̸= 0, we have
a ∈ S and thus, s/a ∈ S −1 A. So, we get (a/s) (s/a) = 1/1 implying a/s is a unit in S −1 A.
Therefore, S −1 A is a field. In this case we call S −1 A as the field of fractions of A.

The ring S −1 A has the following universal property:

Proposition 5.1.4. Let g : A −→ B be a ring homomorphism such that g(s) is a unit in B


for all s ∈ S. Then there exists a unique ring homomorphism h : S −1 A −→ B such that the
following diagram commutes.
g
A B
f ⟲
∃! h
−1
S A

44
Proof. Define h : S −1 A −→ B by h(a/s) = g(a)g(s)−1 . To show that h is well-defined, suppose
a/s = a′ /s′ . Then (a, s) ∼ (a′ , s′ ) and hence (as′ − a′ s)x = 0 for some x ∈ S. This gives
0 = g(0) = g((as′ − a′ s)x) = [g(a)g(s′ ) − g(a′ )g(s)]g(x) = 0. Since g(x) is a unit in B, this
yields g(a)g(s′ ) − g(a′ )g(s) = 0, or equivalently, g(a)g(s′ ) = g(a′ )g(s). Again, as g(s) and
g(s′ ) are units in B, we have g(a)g(s)−1 = g(a′ )g(s′ )−1 . It is routine to check that h is a ring
homomorphism. Moreover, if a ∈ A, then (h ◦ f )(a) = h(a/1) = g(a)g(1)−1 = g(a). Finally, if
φ : S −1 A −→ B is any ring homomorphism such that φ ◦ f = g, then φ(a/s) = φ(a/1 · 1/s) =
φ(a/1)φ(1/s) = φ(a/1)φ(s/1)−1 = φ(f (a))φ(f (s))−1 = g(a)g(s)−1 = h(a/s) since 1/s is a unit
in B with inverse s/1. Thus h is uniquely determined on every element of S −1 A, completing
the proof.

Proposition 5.1.5. The ring S −1 A and the homomorphism f : A −→ S −1 A have the following
properties:

(i) f (s) is a unit in S −1 A for each s ∈ S;

(ii) f (a) = 0 ⇒ as = 0 for some s ∈ S;

(iii) Every element of S −1 A is of the form f (a)f (s)−1 for some a ∈ A and for some s ∈ S.

Proof. (i) f (s) = 1/s in S −1 A has the inverse s/1 for each s ∈ S.
(ii) f (a) = 0 ⇒ a/1 = 0/1 ⇒ as = 0 for some s ∈ S. (iii) Any a/s ∈ S −1 A can be written as
a/s = a/1 · 1/s = f (a)f (s)−1 .

On the other hand, above conditions (i), (ii), and (iii) characterize S −1 A up to isomorphism.

Proposition 5.1.6. If g : A −→ B is a ring homomorphism such that

(i) s ∈ S ⇒ g(s) is a unit in B;

(ii) g(a) = 0 ⇒ as = 0 for some s ∈ S;

(iii) Every element of B is of the form g(a)g(s)−1 for some a ∈ A and for some s ∈ S;

then there is a unique isomorphism h : S −1 A −→ B such that g = h ◦ f .

45
Bibliography

[1] Saban Alaca and Kenneth S Williams. Introductory algebraic number theory. 2003.

[2] Michael Artin. Algebra. Birkhäuser, 1998.

[3] Phani Bhushan Bhattacharya, Surender Kumar Jain, and SR Nagpaul. Basic abstract
algebra. Cambridge University Press, 1994.

[4] H Chatland and H Davenport. Euclid’s algorithm in real quadratic fields. Canadian Journal
of Mathematics, 2:289–296, 1950.

[5] Krzysztof Ciesielski. Set theory for the working mathematician. Number 39. Cambridge
University Press, 1997.

[6] David Steven Dummit and Richard M Foote. Abstract algebra, volume 3. Wiley Hoboken,
2004.

[7] Dinesh Khurana. On gcd and lcm in domains—a conjecture of gauss. Resonance, 8(6):72–79,
2003.

46

You might also like