v

Table of Contents

Abstract iii

Acknowledgements iv

1 Introduction 1
1.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 R-modules 8
2.1 R-module homomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Submodule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Simple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Regular, Quasi-regular, and The Jacobson Radical 18

3.1 Regular Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Jacobson Radical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Quasi-regular . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4 Nilpotency, Idempotency, and Artinian Rings 31

4.1 Nilpotence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Artinian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 Idempotency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

5 Wedderburn-Artin Theorems 42
5.1 Semi-simple Artinian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.2 Simple Artinian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6 Implications 53

Bibliography 61
 1

Chapter 1

Introduction

We begin our path by introducing the notion of an R-module. An R-module

“M ” is an abelian group under addition that is susceptible to the influence of a ring R.
We will go into full detail about R-modules in the next chapter. To give an idea, the
most popular types of examples for R-modules are vector spaces over fields. If V is a
vector space over a field F , then V is an F -module. In fact, one could argue that the idea
of R-modules is an attempt to generalize the idea of vector spaces. Another approach to
R-modules is that they are the generalization of abelian groups, seeing how every abelian
group under addition is a Z-module.
Naturally, if the elements of R can influence the elements of M , it begs the ques-
tion: Are all rings efficient scalars for their modules? Since such a question is immensely
broad, we put our attention into a ring R and its “simple” R-modules. In short, a simple
R-module is an R-module that does not contain any other non-zero R-modules. With the
constraint of simple R-modules, the answer to our previous question comes in the form
of the Jacobson radical. We will see that the Jacobson radical is an ideal of R, and in the
case for which the Jacobson radical happens to be the trivial ideal, we say the ring R is
semi-simple. Thereafter, we eventually come upon a nice way to describe the Jacobson
radical of a ring.
Once the Jacobson radical is demystified, we make our move toward the Wedderburn-
Artin theorems. First, we explore the consequences of our findings with the added as-
sumption that our rings are Artinian, which may be thought of as an abstraction of
finite-dimensionality for vector spaces. Inevitably, we come across the idea of nilpotence
 2

and idempotence, which prove to be of vast help. Once we have the necessary theorems,
we discover the structure of semi-simple Artinian rings. In a turn of events, it becomes
apparent that the structure of semi-simple Artinian rings depends on the structure of
simple Artinian rings. Accordingly, we embark on our last mission: to find the structure
of simple Artinian rings.
After everything is said and done, we will explore concrete examples on how
some semi-simple Artinian rings break down. Maschke’s Theorem reveals how one can
easily create semi-simple Artinian rings. Once we get our hands on a semi-simple Artinian
ring, we will decompose it into a direct sum of simple Artinian rings.
This work is inspired primarily by Herstein’s Noncommutative Rings, [Her]. For-
mal definitions are taken mostly from Hungerford’s Algebra, [Hun], as a strong foundation
for Ring Theory. Nevertheless, this paper is written so that no extra knowledge is nec-
essary to understand this work other than the bare bones of abstract algebra. Although
we shall consider every theorem in the noncommutative, nonunital way, this paper has
been written with examples that are welcoming to those of us who are used to having
elements commute and multiplicative identities!

1.1 Preliminaries

Before we begin, we should become accustomed with a few results from Ring
Theory. We assume the reader has some previous knowledge on both Ring Theory and
Group Theory. Still, we will quickly review a few theorems and lemmas that will fre-
quently be called upon throughout this paper. Refer to [Hun] if needed for a more in
depth review of Group Theory and Ring Theory.
A common theme throughout many of the proofs in this paper will be dealing
with the sets A + B and AB, which we define as follows:

Definition 1.1.1. For any nonempty subsets A, B of a ring R, A + B = {a + b | a ∈

P
A, b ∈ B} and AB = { i,j ai bj | ai ∈ A, bj ∈ B}.

We define AB as sums of products in order to guarantee AB will be closed under

addition. On that note, we can state our first lemma.

Lemma 1.1.1. If A, B are (left, right, two-sided) ideals of R, then both A + B and AB
are (left, right, two-sided) ideals of R.
 3

Proof. We prove the case for left ideals A, B of R. First, since 0 ∈ A and 0 ∈ B, then
0 + 0 = 0 ∈ A + B and 0 · 0 = 0 ∈ AB. Second, Let x ∈ A + B with x = a + b for
P
some a ∈ A and b ∈ B, and let y ∈ AB with y = i,j ai bj for some ai ∈ A and bj ∈ B.
Since a, ai ∈ A for all i and b ∈ B, where A and B are left ideals, then we know −a,
−ai ∈ A and −b ∈ B for all i. Hence, (−a) + (−b) = −(a + b) = −x ∈ A + B and
0
P P P
i,j (−ai )bj = i,j −(ai bj ) = −( i,j ai bj ) = −y ∈ AB. Third, suppose x, x ∈ A + B
with x = a + b and x0 = a0 + b0 for some a, a0 ∈ A and b, b0 ∈ B, and suppose y,
y 0 ∈ AB with y = i,j ai bj and y 0 = i,j a0i bj for some ai , a0i ∈ A and all bj ∈ B. Then,
P P

x + x0 = (a + b) + (a0 + b0 ) = (a + a0 ) + (b + b0 ) ∈ A + B and y + y 0 = i,j ai bj + i,j a0i bj =

P P

0
P
i,j (ai + ai )bj ∈ AB. Finally, let r ∈ R, x ∈ A + B with x = a + b for some a ∈ A
P
and b ∈ B, and y ∈ AB with y = i,j ai bj for some ai ∈ A and some bj ∈ B. Then,
P P
rx = r(a + b) = ra + rb ∈ A + B and ry = r( i,j ai bj ) = i,j (rai )bj ∈ AB on account
of A, B being ideals of R. Hence, both A + B and AB are left ideals of R.

Instinctively, we can define sums and products of ideals with more than two sum-
mands using the appropriate analogues and still have the previous result hold. Specifically,
P
A1 +· · ·+An = {a1 +· · ·+an | ai ∈ Ai , i = 1, . . . , n} and A1 · · · An = { j a1j · · · anj | aij ∈
Ai , i = 1, . . . , n} will be ideals of R provided A1 , . . . , An are ideals of R. Consequently, if
A is an ideal of R, then the set An = { a1 · · · an | ai ∈ A} (= A · · · A n-fold) is an ideal
P

of R.
Since we have established some facts about ideals, we now draw our focus on
some facts about ring homomorphisms.

Canonical Projection Theorem. If f : R → S is a homomorphism of rings, then the

kernel of f is an ideal in R. Conversely, if I is an ideal in R, then the map π : R → R/I
given by r 7→ r + I is a surjective homomorphism of rings with kernel I.

Proof. Suppose R, S are rings and f : R → S is a homomorphism of rings. Consider

Ker f = {r ∈ R | f (r) = 0S }. Of course, f being a homomorphism of rings implies
f (0R ) = 0S , so 0R ∈ Ker f . Next, if a ∈ Ker f , then 0S = f (0R ) = f (a + −(a)) =
f (a) + f (−a) = 0S + f (−a) = f (−a), so −a ∈ Ker f . Now, suppose a, b ∈ Ker f , then
f (a + b) = f (a) + f (b) = 0S + 0S = 0S and f (ab) = f (a)f (b) = (0S )2 = 0S , so both
ab and a + b belong to Ker f . Lastly, let r ∈ R and i ∈ Ker f , then f (ri) = f (r)f (i) =
f (r)0S = 0S = 0S f (r) = f (i)f (r) = f (ir), so both ri and ir belong to Ker f for any r in
R, so Ker f is an ideal of R.
 4

Now, suppose I is an ideal of R and let the map π : R → R/I be given by

r 7→ r + I. Since I is an ideal of R, π is a mapping between two rings. First, let a, b ∈ R;
then, f (a + b) = (a + b) + I = (a + I) + (b + I) = f (a) + f (b) and f (ab) = ab + I =
(a + I)(b + I) = f (a)f (b), so f is indeed a ring homomorphism. Next, if r + I is an
arbitrary element of R/I, then surely r ∈ R and f (r) = r + I, thus f is surjective. Lastly,
note that Ker f = {r ∈ R | f (r) = 0 + I} = {r ∈ R | r + I = I} = {r ∈ R | r ∈ I} = I.

Any mapping defined in the same way as π is called a canonical projection.

There will be many instances throughout this paper in which canonical projections are
used. On the other hand, noting that the kernel of a ring homomorphism is an ideal of
the domain will prove to be very valuable in the proofs to come.
The next two theorems are must-haves when it comes to ring homomorphisms.
Although it is a very well-known and essential fact of Ring Theory, we introduce the First
Isomorphism Theorem of rings in this paper solely for its extensive reappearance in many
of the proofs. As for the Second Isomorphism Theorem of rings, a few crucial theorems
rely on it.

First Isomorphism Theorem. If f : R → S is a homomorphism of rings, then f

∼ f (R).
induces an isomorphism of rings R/Ker f =

Proof. Let R, S be rings and let f : R → S be a homomorphism of rings. As we have

seen, Ker f is an ideal of R and R/Ker f is a ring. Moreover, f (R) is a subring of S.
Naturally, f (0R ) = 0S . Therefore, since f (0R ) ∈ f (R), then 0S ∈ f (R). Next, suppose
x ∈ f (R), then there exists r ∈ R so that x = f (r). Since r ∈ R, then −r ∈ R, meaning
f (−r) = −f (r) = −x ∈ f (R). So, f (R) has additive inverses. Now, suppose x, y ∈ f (R)
and let r, s ∈ R so that x = f (r) and y = f (s). Then, x+y = f (r)+f (s) = f (r+s) ∈ f (R)
and xy = f (r)f (s) = f (rs) ∈ f (R) since R is a ring. Therefore, f (R) is a subring of S.
Lastly, consider the function φ : R/Ker f → f (R) given by φ(r + I) = f (r).
We will show φ is well-defined. Let r + Ker f and s + Ker f be elements of R/Ker f
with r + Ker f = s + Ker f . Then r − s ∈ Ker f , meaning f (r − s) = 0. However,
f (r − s) = f (r) − f (s) = 0 implies f (r) = f (s), so φ(r + Ker f ) = φ(s + Ker f ), φ is
well-defined. Hence, the mapping φ : R/Ker f → f (R) given by φ(r + I) = f (r) is a
mapping of rings.
Now, we move on to show that φ is an isomorphism. First, we will show φ is a
bijective function. Of course, if f (r) = f (s), then f (r − s) = 0, meaning r − s ∈ Ker f .
 5

Hence r + Ker f = s + Ker f , so f is injective. Moreover, let y ∈ f (R) with y = f (r)

for some r ∈ R. Then x = r + Ker f ∈ R/Ker f and φ(x) = φ(r + Ker f ) = f (r) = y,
so φ is surjective. Next, we will show φ is a homomorphism of rings. Let r + Ker f ,
s + Ker f ∈ R/Ker f , then φ( (r + Ker f ) + (s + Ker f ) ) = φ( (r + s) + Ker f ) = f (r + s) =
f (r) + f (s) = φ(r + Ker f ) + φ(s + Ker f ) and φ( (r + Ker f )(s + Ker f ) ) = φ(rs + Ker f ) =
f (rs) = f (r)f (s) = φ(r + Ker f )φ(s + Ker f ). Hence, φ is an isomorphism, meaning
R/Ker f ∼= f (R).

Second Isomorphism Theorem. If I, J are ideals of a ring R, then there is an iso-

morphism of rings (I + J)/J ∼
= I/(I ∩ J).

Proof. Let I, J be ideals of R. Recall that I + J is therefore an ideal of R. Moreover,

J ⊂ I + J. If j ∈ J, then j = 0 + j ∈ I + J. Consequently, J ⊂ (I + J) and both J, I + J
being ideals of R imply that J is an ideal of I + J. Hence, (I + J)/J is a ring. Now,
consider the mapping φ : I → (I + J)/J with φ(i) = i + J for all i ∈ I. Naturally, we
regard i = i + 0 as an element of I + J, so i + J = (i + 0) + J ∈ (I + J)/J. Now, if a, b ∈ I
with a = b, then a − b = 0 ∈ J, so a + J = b + j. Hence, φ is well-defined. Of course, if
a, b ∈ I are arbitrary, then φ(a + b) = (a + b) + J = (a + J) + (b + J) = φ(a) + φ(b) and
φ(ab) = ab + J = (a + J)(b + J) = φ(a)φ(b), so φ is a ring homomorphism. Moreover,
φ is surjective: (I + J)/J = {(i + j) + J | i ∈ I, j ∈ J} = {i + J | i ∈ I} = φ(I). Lastly,
note that Ker φ = {i ∈ I | i + J = 0 + J} = {i ∈ I | i ∈ J} = I ∩ J. Hence, by the First
Isomorphism Theorem, I/(I ∩ J) ∼ = (I + J)/J as desired.

The following lemma is a useful fact to have. Not only will it help prove the
next theorem, but it will also come in handy for future theorems.

Lemma 1.1.2. If f : R → S is a ring homomorphism and J is an ideal of S, then f −1 (J)

is an ideal of R that contains Ker f .

Proof. Let f : R → S be a ring homomorphism and J be any ideal of S. Consider the set
f −1 (J) = {r ∈ R | f (r) ∈ J}. Since f (0R ) = 0S and 0S ∈ J, then 0R ∈ f −1 (J). Moreover,
if a ∈ f −1 (J), then f (−a) = −f (a) ∈ J, so −a ∈ f −1 (J). Also, if a, b ∈ f −1 (J), then
f (a + b) = f (a) + f (b) ∈ J on since f (a) and f (b) belong to J, an additive group. Thus
a + b ∈ f −1 (J). Lastly, let r ∈ R and i ∈ f −1 (J), then note that f (ri) = f (r)f (i)
and f (ir) = f (i)f (r). Since f (r) ∈ S and f (i) ∈ J, where J is a two-sided ideal, then
 6

both f (ri) and f (ir) belong to J. Thus, for any r ∈ R and any i ∈ f −1 (J), both ri
and ir belong to f −1 (J). Hence, f −1 (J) is an ideal of R. Lastly, let x ∈ Ker f , then
f (x) = 0S ∈ J. In other words, x ∈ f −1 (J), meaning Ker f ⊂ f −1 (J).

Note that if I is an ideal of R, and if we have f : R → S as a ring homomorphism,

it need not imply that f (I) is an ideal of S. However, if f happened to be surjective,
then in fact f (I) would be an ideal of S.
Since we will deal with many quotient rings, the following theorem will prove to
be useful. While the following theorem is worded for rings and ideals, there is a similar
statement for Group Theory that we might make note of in the future. We will see the
correspondence between the ideals of a quotient ring and the original ring it came from.

Correspondence Theorem. If I is an ideal in a ring R, then every ideal in R/I is

uniquely of the form J/I, where J is an ideal of R which contains I.

Proof. Let I be an ideal of R and let J/I be any ideal of R/I. Recall that the map
π : R → R/I given by r 7→ r + I is a surjective homomorphism of rings with kernel I.
By the previous lemma, we have that both φ−1 (J/I) is an ideal of R and I ⊂ π −1 (J/I).
If x ∈ J, then π(x) ∈ J/I, meaning that x ∈ π −1 (J/I). Hence, J ⊂ π −1 (J/I). Now, if
x ∈ π −1 (J/I), then we know that π(x) ∈ J/I. Yet, if x + I ∈ J/I, then x ∈ J. Hence, we
have π −1 (J/I) ⊂ J. Therefore, π −1 (J/I) = J, meaning J is an ideal of R that contains
I.
To show uniqueness, we will show that π(π −1 (J/I)) = J/I and π −1 (π(J)) = J.
Let x ∈ π(π −1 (J/I)), then x = π(r), where r ∈ π −1 (J/I). Since r ∈ π −1 (J/I), then
π(r) ∈ J/I. Since x = π(r), then x ∈ J/I. Hence, π(π −1 (J/I)) ⊂ J/I. Now, suppose x ∈
J/I. Then, x = j + I, for some j ∈ J. Since π(j) = x ∈ J/I, then j ∈ π −1 (J/I). Lastly,
since x = π(j) and j ∈ π −1 (J/I), then x ∈ π(π −1 (J/I)). Thus, J/I ⊂ π(π −1 (J/I)),
meaning π(π −1 (J/I)) = J/I. Next, suppose x ∈ J. Then, π(x) ∈ π(J). However, if
π(x) ∈ π(J), then x ∈ π −1 (π(J)). So, J ⊂ π −1 (π(J)). On the other hand, suppose
x ∈ π −1 (π(J)). If so, then π(x) ∈ π(J). Therefore, for some j ∈ J, π(x) = π(j).
Consequently, since π is a ring homomorphism, π(x) − π(j) = π(x − j) = 0, meaning
x − j ∈ Ker π = I. Since I ⊂ J, then we have x − j ∈ J. If x − j = j 0 , for some
j 0 ∈ J, then we have x = j + j 0 ∈ J, since J is an ideal of R. Hence, x ∈ J, meaning
π −1 (π(J)) ⊂ J and thus π −1 (π(J)) = J.
 7

Now, while we assume the reader is familiar with a field, we will define what is
meant by a division ring. Since we will not assume commutativity in any proof, division
rings will be abundant in this paper. One way to visualize a division ring is simply by
thinking of a field without commutativity.

Definition 1.1.2. A ring D with identity 1D 6= 0 in which every nonzero element is a

unit is called a division ring.

The next part our preliminaries will consider is the ring of n × n matrices over a
division ring. In particular, we will see that the ring of endomorphisms over vector spaces
will be of great importance. As one can guess, we will see the relation between the ring
of endomorphisms of a vector space V over a division ring D with dimension n and the
ring of n × n matrices over D. Nevertheless, for the purposes of brevity, we will define
what we mean by the ring of n × n matrices over a division ring.

Definition 1.1.3. We will use M atn (D) to denote the ring of n×n matrices over a
division ring D.

Finally, we will use Zorn’s lemma and the axiom of choice for some key theorems.
Zorn’s lemma will aid us in proving a familiar theorem about maximal ideals in a ring
with a new twist. As for the axiom of choice, having it at our disposal will grant us an
alternative definition for an Artinian ring, which will be used throughout the second half
of the paper. So, it is highly recommended that the reader become familiar with these
two statements.

Axiom of Choice. Let C be a collection of nonempty sets. Then we can choose a member
from each set in that collection.

Zorn’s Lemma. Every non-empty partially ordered set in which every chain has an
upper bound contains at least one maximal element.
 8

Chapter 2

R-modules

In order to identify the structure of simple and semi-simple rings, the idea of
R-modules is fundamental. Regardless, R-modules on their own are a rich subject to
study. We shall cover the basics of modules before going deeper into Ring Theory topics.
Without further ado, let us observe the most important definition of this paper:

Definition 2.0.1. Let R be a ring. A left R-module is an additive abelian group A

together with a function R × A → A ( the image of (r, a) being denoted by ra) such that
for all r, s ∈ R and a, b ∈ A:

(i) r(a + b) = ra + rb

(ii) (r + s)a = ra + sa

(iii) r(sa) = (rs)a

If R has an identity element 1R and

(iv) 1R a = a, ∀a ∈ A

then A is said to be a unitary R-module If R is a division ring, then a unitary R-module

is called a (left) vector space.

A (unitary) right R-module is defined similarly via a function A × R → A

denoted (a, r) 7→ ar and satisfying the obvious analogues for (i)-(iv). Still, from now
on we will regard “R-module” as “left R-module” unless otherwise specified. Note that
vector spaces over fields are actually a special case of R-modules, meaning the analogies
 9

we have been making thus far are well-founded. On the other hand, with the assumption
that R is merely a ring and A some arbitrary additive abelian group, many of the linear
algebra theorems are put on hold.
From the definition, we can contemplate the fact than the product of a ring
element with a module element is another module element. That is, once a ring ele-
ment makes contact with a module element, we are talking about element in the module
thereafter. Let us establish some basic properties of these module elements.
Now, if A is an R-module, 0A is the additive identity of A, and 0R is the additive
identity of R, then ∀r ∈ R, ∀a ∈ A:

r0A = 0A and 0R a = 0A

Proof. Since A is an R-module, then we know that if r ∈ R, then r0A ∈ A. let b = r0A ,
then b = r0A = r(0A + 0A ) = r0A + r0A = b + b. Since b ∈ A, −b ∈ A. Hence,
0A = b + (−b) = b + b + (−b) = b, so b = 0A , or r0A = 0A as desired. Similarly, if a ∈ A,
then 0R a ∈ A. Let c = 0R a, then c = 0R a = (0R + 0R )a = 0R a + 0R a = c + c, then
0A = c + (−c) = c + c + (−c) = c = 0R a.

Another notion that we will need to establish before we move on is the handling
of additive inverses with respect to R-modules. That is, establishing the fact that ∀r ∈ R,
∀a ∈ A:

(−r)a = −(ra) = r(−a)

Proof. All that is necessary is to show that both (−r)a and r(−a) are the additive inverses
of ra. Simple, since ra+(−r)a = (r +(−r))a = 0R a = 0A and ra+r(−a) = r(a+(−a)) =
r0A = 0A . A is abelian, thus adding the elements to the left will be the same as adding
them from the right. Since additive inverses are unique, then (−r)a = −(ra) = r(−a).

There are many examples that show the idea of R-modules is not too foreign to
the rest of ring theory. We will observe examples that show how natural R-modules can
exists. Here are some examples of R-modules.

Example 2.0.1. Every additive abelian group G is a Z-module, with na = a + · · · + a

n-fold and (−n)a = −na for n > 0 and a ∈ G.
 10

Example 2.0.2. If R is a ring and S is a subring of R, then R is an S-module with the

multiplication of R.

Example 2.0.3. If R is a ring, then R[x] (the ring of polynomials in x with coefficients
in R) is an R-module with the scalar multiplication for functions.

Example 2.0.4. If I is a left ideal of R, then I is a left R-module. Similarly, a right

ideal of R is a right R-module.

Example 2.0.5. Let I be a left ideal of R. The additive quotient group R/I is not quite
a ring. Still, with the action of R on R/I as r(s + I) = rs + I, R/I is an R-module. This
action is well-defined, for if a + I, b + I ∈ R/I and a + I = b + I, then for any r ∈ R we
have that a − b ∈ I, meaning r(a − b) = ra − rb ∈ I, and thus ra + I = rb + I.

Example 2.0.6. If R is a ring, then M atn (R) is an R-module with the scalar multipli-
cation r(aij ) = (raij ) for any (aij ) ∈ M atn (R).

2.1 R-module homomorphism

In the comparison of R-modules to vector spaces over fields, R-module homo-

morphisms are the answer to linear transformations. In fact, when R is a division ring,
an R-module homomorphism is indeed called a linear transformation. The full impact
of the propositions following the next definition will be seen extensively once we focus
on tackling the second Wedderburn-Artin theorem. For now, we will establish key facts
about R-module homomorphisms.

Definition 2.1.1. A left R-module homomorphism is a homomorphism f : A → B of

abelian groups where both A and B are R-modules, and also the following condition is
met:

• ∀r ∈ R, ∀a ∈ A, f (ra) = r(f (a))

Example 2.1.1. Every homomorphism of abelian groups f : A → B is a Z-module

homomorphism since f (na) = nf (a), for all a ∈ A and n ∈ Z.

Example 2.1.2. If R is a ring, then the mapping φ : R[x] → R[x] with φ(f (x)) = x(f (x))
(left multiplication by x) for all f ∈ R[x] is an R-module homomorphism, but not a ring
homomorphism.
 11

Proposition 2.1.1. If A and B are R-modules, then HomR (A, B) (the set of all R-
module homomorphisms A → B) is an abelian group with f + g given on a ∈ A by
(f + g)(a) = f (a) + g(a). Additive identity is the zero map.

Proof. HomR (A, B) 6= ∅ since B being an R-module implies B is an abelian group that
has the additive identity 0. Consider the function f : A → B with f (a) = 0, ∀a ∈ A. The
function f is a homomorphism since f (a + b) = 0 = 0 + 0 = f (a) + f (b), ∀a, b ∈ A. Also,
f (ra) = 0 = r0 = rf (a), hence f belongs in HomR (A, B). Closure is given by having
defined addition of functions as pointwise addition. Meaning, if f, g ∈ HomR (A, B),
then ∀a ∈ A, (f + g)(a) ∈ HomR (A, B) since (f + g)(a) = f (a) + g(a) ∈ B. Moreover,
(f + g)(a + b) = f (a + b) + g(a + b) = f (a) + g(a) + f (b) + g(b) = (f + g)(a) + (f + g)(b)
and (f + g)(ra) = f (ra) + g(ra) = rf (a) + rg(a) = r(f (a) + g(a)) = r((f + g)(a)).
Next, if f ∈ HomR (A, B), consider g : A → B with g(a) = −f (a), ∀a ∈ A. Since
(f + g)(a) = f (a) + g(a) = f (a) + (−f (a)) = 0, ∀a ∈ A, then f + g is the zero map,
meaning g is the additive inverse of f . Still, we must show g ∈ HomR (A, B). Clearly,
g(a + b) = −(f (a + b)) = −(f (a) + f (b)) = (−f (a)) + (−f (b)) = g(a) + g(b), so g is
a homomorphism. Similarly, g(ra) = −(f (ra)) = −(rf (a)) = r(−f (a)) = rg(a), so
g ∈ HomR (A, B). Since we are seeing HomR (A, B) as a group under addition, it is
understood that the operation is associative.

Proposition 2.1.2. HomR (A, A) is a ring with identity, where multiplication is compo-
sition of functions.

Proof. If A is an R-module, them we have already established why HomR (A, A) is a group
under addition. To make sense of HomR (A, A) as a ring, we will define multiplication of
two functions of HomR (A, A) to be their composition of functions. Hence, to prove that
HomR (A, A) is closed under multiplication, we will show that if f, g ∈ HomR (A, A), then
f ◦ g ∈ HomR (A, A). Let f, g ∈ HomR (A, A) and let a, b ∈ A. Then, (f ◦ g)(a + b) =
f (g(a + b)) = f (g(a) + g(b)) = f (g(a)) + f (g(b)) = (f ◦ g)(a) + (f ◦ g)(b). Hence,
f ◦ g is a homomorphism. Now, let f, g ∈ HomR (A, A), a ∈ A, and r ∈ R. To show
f ◦ g is linear over R, and thus and R-module homomorphism, consider (f ◦ g)(ra):
(f ◦ g)(ra) = f (g(ra)) = f (rg(a)) = rf (g(a)) = r(f ◦ g)(a). Hence, HomR (A, A) is
closed under multiplication, and thus a ring. HomR (A, A) has the map i : A → A, with
i(a) = a, ∀a ∈ A as the multiplicative identity: f (i(a)) = f (a) = i(f (a)), ∀a ∈ A, so
f ◦ i = f = i ◦ f.
 12

Recall, HomR (A, A) is denoted as EndR (A). As we have shown, EndR (A) is
a ring with composition of functions as multiplication. One thing that has been hiding
in the background is that the elements of HomR (A, B) act on the elements of A. Even
more interesting than that is how EndR (A) has the necessary power over the elements
of A to satisfy the R-module axioms. In particular, EndR (A) is a unitary ring, where
as HomR (A, B) is just an additive group. Moreover, the “product” of an element of
EndR (A) with an element of A necessarily belongs to A, unlike HomR (A, B). Additively,
while HomR (A, B) would satisfy parts (i) and (ii) from the definition of R-modules,
EndR (A) undoubtedly satisfies (i)-(iv). In light of this entire paragraph, we can claim
the following proposition.

Proposition 2.1.3. A is a left EndR (A)-module with f a = f (a), for any f ∈ EndR (A)
and a ∈ A.

As we have seen, V a vector space over a division ring D is a special case in the
topic of modules. Nevertheless, EndD (V ) is certainly a ring as we have just shown. Of
course, EndD (V ) is still the set of linear transformations from V onto itself. That being
said, we will recall the connecting between the set of D-linear transformations from V
onto itself and the ring of n × n matrices over D. While we know these are isomorphic
when D is a field, we will observe the slight change if we restrict ourselves to division
rings.

Definition 2.1.2. Let R be a ring. Then, the opposite ring of R, denoted Rop , is a ring
with the same elements as R, the same addition as R, and multiplication “ ·” given by

a · b = ba

where ba is the product in R.

Theorem 2.1.1. EndD (V ) ∼

= M atn (Dop ) , for V a finite dimensional vector space over
a division ring D with dimension n.

Proof. Let V be an n-dimensional vector space over the division ring D with basis B =
{v1 , . . . , vn }. Define Φ : EndD (V ) → M atn (Dop ) with Φ(f ) = (aij ), where f (vd ) =
P
c acd vc . To prove Φ is injective, suppose Φ(f ) = Φ(g). If so, then since every entry in
both matrices agree, we have that f (vi ) = g(vi ) for i ∈ {1, . . . , n}. Since both f and g
 13

agree in outputs for the basis vectors, then it must be that f = g. Now, for surjective,
suppose A ∈ M atn (Dop ). Let u1 , . . . , un be the column vectors of A in V with respect
to B, and define h : V → V to be a D-linear mapping that sends h(vi ) = ui for all
i. Since h is a D-linear transformation from V to itself, we have that h ∈ EndD (V ).
Since h ∈ EndD (V ) and Φ(h) = A, Φ is surjective. Now, suppose f , g ∈ Endn (D) and
P P
consider Φ(f +g). Since we have that (f +g)(vd ) = f (vd )+g(vd ) = c acd vc + c bcd vc =
P
c (acd + bcd )vc , then we can see that the i-jth entry for Φ(f + g) is precisely aij + bij .
Hence, Φ(f +g) = (aij +bij ) = (aij )+(bij ) = Φ(f )+Φ(g). Now, for multiplication, we need
to show that Φ(f ◦ g) = Φ(f )Φ(g), where ◦ is composition of functions (the multiplication
P P
of EndD (V )). Relative to the basis B, say f (vk ) = i aik vi and g(vj ) = k bkj vk , for
i, j, k ∈ {1, . . . , n}. If Φ(f ) = A and Φ(g) = B, we can say that the aik are the
entries of A and bkj are the entries of B. Moreover, since A, B ∈ M atn (Dop ), then
the i-jth entry of AB is l (ail · blj ), where · is the operation of Dop as described in
P
P
Definition 2.1.2. On that note, we can see that (f ◦ g)(vj ) = f (g(vj )) = f ( k bkj vk ) =
P P P P P P P
k bkj f (vk ) = k bkj ( i aik vi ) = i ( k bkj aik )vi = i ( k (aik · bkj ))vi . In other
P
words, ( k (aik · bkj )) is the i-jth entry of the matrix corresponding to Φ(f ◦ g). As we
P
stated before, ( k (aik · bkj )) is the i-jth entry of the matrix AB. Hence, we have that
Φ(f ◦ g) = AB = Φ(f )Φ(g). Thus, Φ is an isomorphism of rings between EndD (V ) and
M atn (Dop ).

2.2 Submodule

In abstract algebra, we have subgroups in group theory and we have subrings

in ring theory. Clearly, module theory cannot lag behind on this aspect! Sadly, starting
from the next section and thereafter, we will focus on R-modules that do not have proper,
nontrivial submodules. In any case, we will explore briefly the idea of submodules.

Definition 2.2.1. Let R be a ring, A an R-module and B a nonempty subset of A. We

say that B is a submodule of A provided that B is an additive subgroup of A and rb ∈ B
for all r ∈ R, b ∈ B.

Observe that merely being a subgroup of an R-module is not enough to claim

the subgroup is a submodule. Mainly, the extra condition needed is that the product of
a ring element and a subgroup element lands inside the subgroup. Also note that if A
 14

is an R-module, it is not guaranteed that for any subset B of A, RB ⊂ B; only, it is

guaranteed that RA ⊂ A.

Example 2.2.1. As stated before, if I is an ideal of R, then I is an R-module. Of

course, R itself is an R-module and I is an additive subgroup of R. Hence, all (left, right,
two-sided) ideals of R are submodules of the R-module R.

Example 2.2.2. Recall, Q is an additive subgroup of R. Since both are abelian groups,
they are Z-modules. Hence, Q is a submodule of R over Z.

Example 2.2.3. Consider the ring M at2 (Q). Of course, M at2 (R) is a M at2 (Q)-module.
Consider the subgroup S = {(aij ) ∈ M at2 (R) | aij = 0 if i 6= j} of M at2 (R). Even though
S is a subgroup of M at2 (R), S is not a submodule of M at2 (R) under M at2 (Q).

Example 2.2.4. let R be a ring and f : A → B be an R-module homomorphism. As we

will see in a future proof, Ker f is a submodule of A and Im f = f (A) is a submodule of
B.

2.3 Simple

The word “simple” will be the most recurring word in this paper. In order to
be in the same page, we will define “simple” in the context of both modules and rings.
As we will see, these are not necessarily the same idea.

Definition 2.3.1. A left module A over a ring R is simple, or irreducible, if RA 6= (0)

and A has no proper submodules. A ring R is simple if R2 6= (0) and R has no proper
(two-sided) ideals.

Suppose you have a ring R that is simple. Since R2 6= (0) and R is its own an
R-module, the definitions for simple modules and simple rings agree thus far. However,
a ring with no ideals could still have nontrivial left (or right) ideals, thus R could have
nontrivial proper submodules.

Example 2.3.1. If D is a division ring, then D has no nontrivial, proper left, right, or
two-sided ideals. Recall that if I is a nonzero ideal of D, and if a 6= 0 is an element of I,
then there exists b ∈ D so that ba = 1D ∈ I, meaning D = D1D ⊂ I. Hence, we can see
that D is both a simple ring and a simple D-module.
 15

Proposition 2.3.1. If D is a division ring, then M atn (D) is a simple ring

Proof. Let D be a division ring and suppose I is an ideal of M atn (D) such that I 6= (0);
we will show I = M atn (D). Let A = (aij ) be a nonzero matrix of I. Since A is nonzero,
without loss of generality suppose the entry amk 6= 0. Let the matrix Exy ∈ M atn (D)
be one where the entry exy = 1 and all other entries are zero. If so, then it is easy
to check that Elm AEkl = amk Ell for l = 1, 2, . . . , n. Moreover, since A ∈ I, an ideal,
then amk Ell ∈ I for l = 1, 2, . . . , n. Furthermore, if we let E ∈ M atn (D) be the
identity matrix, then we can see that nl=1 amk Ell = amk E belongs to I. Of course,
P

((amk )−1 E)(amk E) = E, so amk E is a unit of M atn (D) that belongs in I. Hence,
I = M atn (D).

Example 2.3.2. Let D be a division ring and let R = M atn (D) with n > 1. For each k
so that 1 ≤ k ≤ n, Ik = {(aij ) ∈ R | aij = 0 for j 6= k} is a simple left R-module (a quick
proof of this is given in Example 4.3.4). Note that while R is a simple ring, the existence
of left, right ideals shows R is not a simple R-module.

Definition 2.3.2. A left ideal I of a ring R is said to be a minimal left ideal if I 6= (0)
and for every left ideal J such that (0) ⊂ J ⊂ I, either J = (0) or J = I

Example 2.3.3. A left ideal I of R such that RI 6= (0) is a simple left R-module if and
only if I is a minimal left ideal.

Proposition 2.3.2. Every simple left R-module is cyclic; in fact, if A is simple, then
A = Ra, ∀a ∈ A, (a 6= 0).

Proof. Let A be a simple left R-module and let a ∈ A, a 6= 0. First, we will show that Ra
is a subgroup, and then a submodule, of A. Ra = {ra | r ∈ R} is a subset of A since a ∈ A
and A is an R-module implies that ∀r ∈ R, ra ∈ A. First, 0R ∈ R, so 0R a = 0A ∈ Ra,
so Ra is non-empty. Second, if ra, sa ∈ Ra, then ra + sa = (r + s)a ∈ Ra, meaning
there is closure in Ra. Third, if ra ∈ Ra, then r ∈ R, so −r ∈ R, so (−r)a ∈ Ra,
meaning −ra ∈ Ra. Since Ra has identity, closure, and inverses, then Ra is a subgroup
of A. For submodule, we must know show that ∀r ∈ R, x ∈ Ra, rx ∈ Ra. Of course, if
x ∈ Ra, then x = sa, for some s ∈ R. So, rx = r(sa) = (rs)a ∈ Ra, meaning Ra is a
submodule. Since Ra is a submodule of the simple R-module A, then by Definition 2.3.1
either Ra = A or Ra = (0). Now, consider B = {b ∈ A | Rb = (0)}. Clearly 0A ∈ B since
 16

r0A = 0A for all r ∈ R (and 0A ∈ A), So B is non-empty. If x, y ∈ B, then ∀r ∈ R,

r(x + y) = rx + ry = 0A + 0A = 0A , meaning R(x + y) = (0), or that x + y ∈ B. Plus, if
x ∈ B, then ∀r ∈ R, rx = 0A . Hence, r(−x) = −rx = −0A = 0A for all r ∈ R, meaning
−x ∈ B. In short, B is a subgroup of A. To show B is a submodule, let r ∈ R and b ∈ B.
To show rb ∈ B, consider R(rb). If s ∈ R, then s(rb) = s(0) = 0; meaning, s(rb) = 0
∀s ∈ R, or R(rb) = (0), and thus rb ∈ B. Since A is simple and B its submodule, again
either B = A or B = (0). However, if B = A, then ∀x ∈ A, Rx = (0), meaning RA = (0),
contradicting A as a simple R-module. Thus, B = (0). Hence, if a 6= 0, then a ∈
/ B,
meaning Ra 6= (0), thus Ra = A.

The following theorem is the most powerful result from R-module homomor-
phism and simple R-modules: Schur’s lemma. While it has many forms throughout
abstract algebra, we shall focus on the module theory version.

Schur’s Lemma. Let A be a simple R-module and B any R-module:

(i) Every nonzero R-module homomorphism f : A → B is injective.

(ii) Every nonzero R-module homomorphism g : B → A is surjective.

(iii) EndR (A) is a division ring.

Proof. First, suppose A and B are as assumed, and suppose f : A → B is a nonzero

R-module homomorphism. Consider Ker f = {a ∈ A | f (a) = 0}. We know Ker f is
a subgroup of A. We will show Ker f is a submodule. Let r ∈ R and k ∈ Ker f .
Since f is an R-module homomorphism and k ∈ A (Ker f ⊂ A), then f (rk) = rf (k).
Since k ∈ Ker f , then f (k) = 0, so f (rk) = rf (k) = r0 = 0, so rk ∈ Ker f . Thus,
rk ∈ Ker f for all r ∈ R, k ∈ Ker f , then Ker f is a submodule of A. Since A is simple,
then either Ker f = A or Ker f = (0). However, if Ker f = A, then f is the zero
map, which we are assuming it is not. Hence, Ker f = (0). Now, if f (a) = f (b), then
f (a) − f (b) = f (a − b) = 0. So, a − b ∈ Ker f = (0), so a − b = 0, or a = b. Therefore, f
is injective.
Second, suppose A and B are as assumed, and suppose g : B → A is a nonzero
R-module homomorphism. Consider the image of B under g, g(B) = {g(b) | b ∈ B}.
Again, g(B) as the image of a homomorphism is a subgroup of the codomain A. Now,
we will show g(B) is a submodule of A. Let r ∈ R and let y ∈ g(B). Consider ry. If
 17

y ∈ g(B), then ∃x ∈ B so that y = g(x). Since g is an R-module homomorphism, then

ry = rg(x) = g(rx). Since r ∈ R, x ∈ B and B an R-module, then rx ∈ B. Therefore,
ry = g(rx) and rx ∈ B, then ry ∈ g(B). Now that we know g(B) is a submodule of A,
by simplicity either g(B) = (0) or g(B) = A. If g(B) = (0), then it would mean that g is
the zero mapping, which is not. Therefore g(B) = A, which means g is surjective.
Third, Suppose A is a simple R-module. As we have seen before, EndR (A) is a
ring with unity. Now, let h ∈ EndR (A), h 6= 0. By (i) and (ii), we know h is injective
and surjective. Since h is bijective, then h has an inverse h−1 : A → A. Lastly, to say
h−1 ∈ EndR (A), we need to show h−1 is an R-module homomorphism. Let a, b ∈ A.
Since h is bijective, then ∃x, y ∈ A so that a = h(x) and b = h(y) (Note this implies that
h−1 (a) = x and h−1 (b) = y). Then h−1 (a + b) = h−1 (h(x) + h(y)) = h−1 (h(x + y)) =
x + y = h−1 (a) + h−1 (b), so h−1 is a homomorphism. Now, let a ∈ A and r ∈ R. Since
h is bijective, say a = h(x) for some x ∈ A. Then h−1 (ra) = h−1 (rh(x)) = h−1 (h(rx)) =
rx = rh−1 (a), so h−1 is an R-module homomorphism.

Mainly, part (iii) will be of great importance. By Proposition 2.1.3, A being an

R-module implies A is an EndR (A)-module. As we can see from Schur’s lemma, whenever
A is a simple R-module, then A is a vector space over the division ring EndR (A). Just
to let the weirdness sink in, notice this says that a simple module is a vector space with
functions as scalars. Then again, the act of multiplying a vector by a scalar to get another
vector makes it sound like scalars induce endomorphisms on a vector space. So, perhaps
it is not so different after all. Also, let us not dismiss the fact that every R-module
endomorphism of a simple module is automatically bijective.
 18

Chapter 3

Regular, Quasi-regular, and The

Jacobson Radical

Now that we have familiarized ourselves with the basics of R-modules, we will
finally lean more towards ring theory. While we covered the definition of a simple ring, we
have yet to mention the definition of a semi-simple ring. The reason for withholding the
definition so far is that the definition for a semi-simple ring depends on what is known as
the Jacobson radical of a ring. Consequently, before understanding the Jacobson radical
we needed to dig into R-modules beforehand. Regardless, there are still a few notions we
must cover before jumping into the Jacobson radical.

3.1 Regular Ideals

As we mentioned before, if I is a left ideal of R, then R/I is a left R-module.

Upon close examination, we can see that there is a specific type of ideal that would
guarantee R/I to have no submodules. However, in order to call it simple, having no
nontrivial submodules is half of it. To guarantee the second condition, R(R/I) 6= (0), we
will need the idea of regular ideals.

Definition 3.1.1. A left ideal I of a ring R is left regular if there exists a ∈ R such that
r − ra ∈ I for every r ∈ R. Similarly, a right ideal J of R is right regular if there exists
b ∈ R such that r − br ∈ J for every r ∈ R.

While the definition for regular ideals seems wild, it is not as uncommon as one
 19

might think. We shall contemplate some examples of familiar ideals that are regular.

Example 3.1.1. If R is a ring with unity, then every (left, right, two-sided) ideal of R
is regular. That is, for any ideal I of R, r − r(1R ) = 0 ∈ I and r − (1R )r = 0 ∈ I for any
r in R.

Example 3.1.2. Consider the non unital ring 2Z and its ideal 10Z. Since 2Z is commuta-
tive, right regular and left regular are the same notion. Now, consider −4 ∈ 2Z: if n ∈ 2Z,
then n = 2k for some integer k. Thus, n − n(−4) = n + 4n = 5n = 5(2k) = 10k ∈ 10Z
for any n in 2Z. So, 10Z is a regular ideal of 2Z.

There is an application of Zorn’s lemma that declares the existence of maximal

ideals on rings with unity. However, asking for a multiplicative identity is much more
than necessary. As we shall prove, a ring need only have a proper regular ideal in order
to contain a maximal ideal.

Lemma 3.1.1. If ρ is a proper regular left ideal of R, then there exists a maximal regular
left ideal ρ0 of R so that ρ ⊂ ρ0 .

Proof. Let ρ be a proper, regular left ideal of R and let a ∈ R so that r − ra ∈ ρ for all
r ∈ R. Note that a ∈
/ ρ, or else ra ∈ ρ meaning r ∈ ρ, ∀r ∈ R, or ρ = R. Let M be the
set of all proper left ideals of R which contain ρ. Note once more that if ρ0 ∈ M, then
/ ρ0 . Otherwise, since r − ra ∈ ρ and ρ ⊂ ρ0 means both r − ra and ra are in ρ0 for all
a∈
r ∈ R, then r ∈ ρ, ∀r ∈ R contradicting ρ0 as being a proper left ideal of R. Now, we will
apply Zorn’s lemma. Since ρ ⊂ ρ, then ρ ∈ M, meaning M is nonempty. Moreover, M
is a partially ordered set under set containment. Lastly, Suppose ρ ⊂ ρ1 ⊂ ρ2 ⊂ ρ3 ⊂ ...
is a chain of ideals from M. Let U be the union of all the ideals from the chain of ideals.
Clearly, U is an upperbound to the chain of ideals as for set containment. Hence, all that
is needed to show is that the set U belongs in M. Hence, we have to show U is a left ideal
of R that contains ρ. In fact, we know ρ ⊂ U since U is the union of sets that contain ρ.
Now, since ρ ⊂ U , then ρ being a left ideal of R implies 0 ∈ U , so U is nonempty. Next,
let u ∈ U . Since u ∈ U , ∃ρu in the chain of ideals so that u ∈ ρu . Since ρu is a left ideal,
it is certainly an additive subgroup of R, thus, −u ∈ ρu . Since ρu ⊂ U , −u ∈ U , meaning
elements of U have additive inverses. Now, let x, y ∈ U . As explained before, x ∈ ρx and
y ∈ ρy for some ρx , ρy in the chain of ideals. Since both ρx and ρy are both in the chain,
 20

either ρx ⊂ ρy or ρy ⊂ ρx . Without loss of generality, suppose ρx ⊂ ρy . Then, x ∈ ρx

implies x ∈ ρy and since both x and y are in ρy , ρy is a subring of R, then x + y ∈ ρy and
xy ∈ ρy . In other words, x + y and xy belong in U , making U closed under addition and
multiplication. Now that we have shown U is a subring of R, we need to establish U as a
left ideal. Let r ∈ R and i ∈ U . Say i ∈ ρi , where ρi is a left ideal from the chain. Then,
by definition, ri ∈ ρi . So, ri ∈ U , making U a left ideal of R. Therefore, since U is a an
upperbound on the chain of ideals and U is in M, Zorn’s lemma indicates that there is
a maximal element of M. Call the maximal element ρ0 . Note that if a proper left ideal
J of R happened to contain ρ0 , it would imply that J ∈ M, which in turn means J is
contained in ρ0 . Therefore, ρ0 is a maximal left ideal of R. Lastly, as hinted before, ρ0 is
regular since r − ra ∈ ρ and ρ ⊂ ρ0 .

Now that we have made ourselves familiar with regular ideals, let us digress to
our previous discussion of creating simple modules. If R is a ring and I is a left ideal of
R, then R/I will be without nontrivial submodules provided I is a maximal left ideal of
R. However, to make sure R will not annihilate R/I, having I as a regular ideal will do
the trick. Surprisingly, we will discover that any simple left R-module is isomorphic to a
quotient group R/I where I is as maximal left ideal which is also regular.

Theorem 3.1.1. A left R-module A is simple ⇐⇒ A ∼

= R/ρ for some maximal left
ideal ρ, which is regular.

Proof. Let A be a simple R-module and let a ∈ A, a 6= 0. Recall that R is also an

R-module and that A = Ra. Consider the mapping φ : R → Ra with φ(r) = ra. Suppose
r, s ∈ R. Then, φ(r + s) = (r + s)a = ra + sa = φ(r) + φ(s), meaning φ is a group
homomorphism. Now, to show that φ is an left R-module homomorphism, let c ∈ R and
r ∈ R, then φ(cr) = (cr)a = c(ra) = cφ(r). Moreover, it is obvious that φ is surjective.
Let ρ = Kerφ = {r ∈ R | ra = 0}, then R/ρ ∼= A by the . Next, we will show that ρ is
maximal left ideal. Since ρ = Kerφ, then ρ is an additive subgroup of R. Now, if r, s
∈ ρ, then (rs)a = r(sa) = r(0) = 0, so rs ∈ ρ; thus, ρ is a subring of R. To show it is
a left ideal, let r ∈ R and i ∈ ρ. Since (ri)a = r(ia) = r(0) = 0, then ri ∈ ρ, meaning ρ
is a left ideal. To show that ρ is left maximal, suppose that instead there is an left ideal
J of R such that ρ ⊂ J. Then J/ρ is a submodule of R/ρ. However, since A is simple
∼ R/ρ, then either J/ρ = (0) or J/ρ = R/ρ. In other words, either J = ρ
module, and A =
 21

or J = R. Therefore, ρ is maximal left ideal of R. Finally, to show ρ is regular, recall

that a ∈ A and A = Ra. Since a ∈ A, then a ∈ Ra. Therefore, ∃s ∈ R so that a = sa.
Then, for all r ∈ R, ra = rsa, which means ra − rsa = (r − rs)a = 0. In other words,
∀r ∈ R, r − rs ∈ ρ where s ∈ R, then ρ is regular.
For the converse, suppose A ∼
= R/ρ for some maximal regular left ideal ρ. Since
ρ is maximal, R/ρ will have no submodules. By the Correspondence Theorem, any
submodule of R/ρ is a submodule of the form J/ρ, where J is an ideal of R and ρ ⊂ J.
Thus, either J = ρ or J = R, meaning the submodule J/ρ is either the trivial subgroup
or R/ρ itself. Now, to call R/ρ simple, we must show that R(R/ρ) 6= (0). Suppose that
instead, R(R/ρ) = (0). We know that ρ is a maximal regular left ideal, and suppose that
s ∈ R so that r − rs ∈ ρ for all r ∈ R. Since R(R/ρ) = (0), then surely r(s + ρ) = ρ,
∀r ∈ R. This implies that rs ∈ ρ, ∀r ∈ R. However, if both r −rs ∈ ρ and rs ∈ ρ, ∀r ∈ R,
then r ∈ ρ, ∀r ∈ R. Hence, R ⊂ ρ, which contradicts ρ being a maximal ideal.

3.2 Jacobson Radical

When we talk about A being an R-module, we can think of the elements of A

as objects that are being influenced by elements from R. That being said, we want to be
aware of elements of R which “eliminate” the elements of A. On that note, we turn to
the annihilator of an R-module M .

Definition 3.2.1. If M is an R-module, then A(M ) = {x ∈ R | xM = (0)} is the left

annihilator of M .

Example 3.2.1. Consider Z6 with the standard Z-module structure for abelian groups.
What is A(Z6 )? Note that even though 3 · 2 = 0 and 3 · 4 = 0, 3 does not annihilate all
of Z6 . Hence, 3 ∈
/ A(Z6 ). Upon close examination, we can see A(Z6 ) = 6Z. In general,
A(Zn ) = nZ.

Note that in the previous example, the annihilator of the module was an ideal of
the ring. This is not an accident, as we shall see in the next lemma. In fact, the definition
of an annihilator is strong enough to make ideals from mere subset of a module.

Lemma 3.2.1. If A is a left R-module and B ⊂ A, then A(B) = {r ∈ R | rb = 0,

∀b ∈ B} is a left ideal of R. If B is a submodule of A, then A(B) is a two-sided ideal of
R.
 22

Proof. First, we will show that if B ⊂ A, then A(B) is a left ideal. First, 0R ∈ A(B),
so A(B) is nonempty. Second, suppose r, s ∈ A(B). Let b ∈ B and consider (r + s)b =
rb + sb = 0 + 0 = 0. Since (r + s)b = 0 for all b ∈ B, then r + s ∈ A(B). Third, if
r ∈ A(B), then (−r)b = −rb = −0 = 0, meaning −r ∈ A(B). Thus, A(B) is a subgroup
of R. Now, for subring, let r, s ∈ A(B) and let b ∈ B. Since (rs)b = r(sb) = r(0) = 0 for
all b ∈ B, then rs ∈ A(B). Now that we know A(B) is a subring of R, we will now show
that A(B) is a left ideal. Let r ∈ R and i ∈ A(B). To show ri ∈ A(B), let b ∈ B. Then,
(ri)b = r(ib) = r(0) = 0, ∀b ∈ B. Hence, ri ∈ A(B), making it a left ideal.
Now, suppose that this time B is a submodule of A. Since submodules are
subsets, when know that A(B) is a left ideal. To show it is an ideal, we will show it is
also a right ideal. Let r ∈ R, i ∈ A(B) and b ∈ B. To show ir ∈ A(B), we must show
that (ir)b = 0 ∀b ∈ B. We know (ir)b = i(rb), however, we do not know if rb = 0 since
r is just a random element of R and b a random element of B. The difference from last
time is that we know that B is an submodule of A, which means that we definitely know
rb ∈ B. Since i ∈ A(B) and rb ∈ B, then we know i(rb) = 0, which means (ir)b = 0,
meaning A(B) is a two-sided ideal.

Clearly, the zero element is not the only element of a ring that can annihilate
its module elements. However, if zero is the only element that ubiquitously annihilates
every module element, we call the module faithful. In the case where the module also
happens to be simple, we say the ring is primitive.

Definition 3.2.2. A (left) module M is faithful if its (left) annihilator A(M ) is (0). A
ring R is (left) primitive if there exists a simple faithful (left) R-module.

Example 3.2.2. Recall that Q as a subring of R implies R is a Q-module. Now, since

A(R) is an ideal of Q and clearly not all rational numbers annihilate R, then R is a
faithful Q-module.

Example 3.2.3. Continuing with the above example, since R is also an R-module where
only zero annihilates all of R, then R is a faithful R-module. Moreover, since submodules
of R translates to ideals of R in this case, we can see that R is a simple module. Therefore,
R is a primitive ring.

The reason why fields were used as examples foreshadows an interesting fact
about commutative primitive rings. We will take a closer look at primitive rings right
 23

before tackling the last Wedderburn-Artin theorem. For the moment, let us observe one
more fact concerning annihilators.

Lemma 3.2.2. If M is a left R-module, then M is a faithful left R/A(M )-module.

Proof. Let M be a left R-module. Recall from Lemma 3.2.1 that, since M is a submodule
of itself, A(M ) is an ideal of R. So, R/A(M ) is a ring. Define for m ∈ M and r +A(M ) ∈
R/A(M ), the action (r+A(M ))m = rm. Let us show this action is well-defined: Suppose
a + A(M ) = b + A(M ). Then, a − b ∈ A(M ). Therefore, (a − b)m = 0 for all m ∈ M . So,
am − bm = 0 implies am = bm for all m ∈ M . In short, (a + A(M ))m = (b + A(M ))m for
all m ∈ M , so the action of R/A(M ) on M is well-defined. Now, recall that M is an R-
module, so verifying the module axioms is almost trivial. First, let r + A(M ) ∈ R/A(M )
and let m, n ∈ M . Then, (r + A(M ))(n + m) = r(n + m) = rn + rm = (r + A(M ))n +
(r + A(M ))m. Second, let r + A(M ), s + A(M ) ∈ R/A(M ) and let m ∈ M , then
[(r + A(M )) + (s + A(M ))]m = (r + s + A(M ))m = (r + s)m = rm + sm = (r +
A(M ))m+(s+A(M ))m. Third, let r+A(M ), s+A(M ) ∈ R/A(M ) and let m ∈ M , then
[(r+A(M ))(s+A(M ))]m = (rs+A(M ))m = (rs)m = r(sm) = (r+A(M ))[(s+A(M ))m].
Thus, M is an R/A(M )-module. Now, to show M is faithful, suppose (r + A(M ))m = 0
for all m ∈ M . Then, rm = 0 for all m ∈ M . This implies that r ∈ A(M ); hence, if
(r + A(M ))m = 0, then r + A(M ) = 0 + A(M ), the zero element.

Example 3.2.4. Recall that Z6 is a Z-module. Notice that the set C = {0, 2, 4} is a
simple submodule of Z6 with A(C) = 3Z. By Lemma 3.2.2, C is a faithful Z/3Z-module.
Hence, the ring Z/3Z is primitive.

Given how a ring R could in fact have elements that turn all the module elements
into the zero element, it would be nice to have a way to categorize a ring by how volatile
it might be to its modules. While the annihilator of an R-module M can tell us which
elements annihilate the elements of M , this set is only specific to the relevant R-module.
Moreover, considering all of the modules of a ring sounds pretty daunting, so we would
like to focus on the simple modules of a ring first. On that note, we introduce The
Jacobson Radical.

Definition 3.2.3. The Jacobson radical of R, written as J(R), is the set of all elements
of R which annihilate all the simple R-modules. If R has no simple R-modules, then
 24

J(R) = R and we call R a radical ring. On the other hand, If J(R) = (0), then we say R
is semi-simple.

We have finally revealed the meaning of a semi-simple ring! In abstract algebra,

the relationship between simple objects and semi-simple objects is, well, simple. We
usually refer to semi-simple objects as those which are some type of product of other
simple objects. While this definition for a semi-simple ring sounds out of place, it is
simply a foreshadowing for things to come.
Notice that the Jacobson radical is the intersection of all simple module annihi-
T
lators. That is, J(R) = A(M ), where each M is a simple R-module of R. Respectively,
since each A(M ) is a two-sided ideal of R, then J(R) is also a two-sided ideal of R. Unfor-
tunately, a ring could have many random modules, so exploring the annihilators of every
single module to a identify ring’s Jacobson radical is not a task we want to undertake.
We now look for a better structure for the Jacobson radical of a ring.

Definition 3.2.4. If I is a left ideal of R, then (I : R) = {x ∈ R | xR ⊂ I}.

While an left ideal I of a ring R absorbs ring elements from the left, this new
set consists of elements of R that absorb ring elements into I from the right. Naturally,
if I is two-sided, then I ⊂ (I : R). In any case, we will see this set is of great importance
in moving our focus away from the modules themselves.

Proposition 3.2.1. If M = R/ρ for some maximal regular left ideal ρ, then A(M ) =
(ρ : R), and (ρ : R) is the largest two-sided ideal of R inside ρ.

Proof. Suppose M = R/ρ for some maximal regular left ideal ρ. Let x ∈ A(M ), Then
xM = (0); meaning, x(r + ρ) = ρ for all r ∈ R. Therefore, xr ∈ ρ, ∀r ∈ R. Hence,
xR ⊂ ρ and thus x ∈ (ρ : R). Now, let x ∈ (ρ : R). Then, since xR ⊂ ρ, xr ∈ ρ, ∀r ∈ R.
Then, ρ = xr + ρ = x(r + ρ), ∀r ∈ R. So x(R/ρ) = (0), implying x ∈ A(M ). Therefore,
A(M ) = (ρ : R). Moreover, Since ρ is regular, ∃a ∈ R so that r − ra ∈ ρ for all r ∈ R.
Let x ∈ (ρ : R). Consequently, x ∈ R, then x − xa ∈ ρ. However, since x ∈ (ρ : R)
and a ∈ R, then xa ∈ ρ. If xa ∈ ρ and x − xa ∈ ρ, then x ∈ ρ. Hence, (ρ : R) ⊂ ρ.
To show that (ρ : R) is the largest ideal of ρ, suppose there is an ideal J of ρ so that
(ρ : R) ⊂ J ⊂ ρ. If j ∈ J, then j ∈ R. Now, j ∈ J, a two-sided ideal, and J ⊂ ρ implies
jr ∈ ρ for all r ∈ R; hence by definition, j ∈ (ρ : R). Therefore, J ⊂ (ρ : J).
 25

T
Hence, for any ring R, J(R) = (ρ : R), where ρ runs over all regular maximal
left ideal of R. Now, all there is left to do is to keep track of (ρ : R) for each maximal
regular left ideals ρ of R. While it sounds better than looking at every simple module
of a ring, it still feels unsatisfactory. We can still do better, as we will see in the next
section.

3.3 Quasi-regular

If the idea of regular ideals seemed bizarre, then we are in for a treat with the
introduction of quasi-regular ideals. While the definition of a regular ideal states nothing
about the element of the ideal, a quasi-regular ideal is an ideal where all of its elements
are quasi-regular.

Definition 3.3.1. An element a ∈ R is said to be left-quasi-regular if there is an element

b ∈ R such that b + a + ba = 0. We call b a left-quasi-inverse of a. Similarly, c ∈ R
is said to be right-quasi-regular if there is an element d ∈ R such that c + d + cd = 0
and d is the right-quasi-inverse of c. We say that a (left, right, two-sided) ideal of R
is left-quasi-regular if all of its elements are left-quasi-regular. Similarly, a (left, right,
two-sided) ideal of R is right-quasi-regular if all of its elements are right-quasi-regular.

Example 3.3.1. In R, every element is quasi-regular except −1. If (−1) + y + (−1)y = 0,

x
then −1 = 0. If neither x nor y is −1 and y is the quasi-inverse of x, then y = − x+1 .

Example 3.3.2. In Z, the only quasi-regular elements are 0 and −2, where they are
x
both their own quasi-inverse. (Again, observe the solutions to y = − x+1 for x, y ∈ Z.)

Apparently quasi-regular elements are not as strange as we originally thought.

Certainly, a question one might have is “why do we care for quasi-regular elements in a
ring?” In a weird turn of events, quasi-regular elements are the key for a better structure
of the Jacobson radical. Foremost, let us state an important fact about quasi-regular
ideals.

Lemma 3.3.1. If I is a left-quasi-regular left ideal of R, then I ⊂ J(R).

Proof. Let I be a left-quasi-regular left ideal of R. Suppose then that I 6⊂ J(R). If so,
then for some simple R-module M , we must have IM 6= (0). In turn, it must mean that
 26

there is at least one element m ∈ M so that Im 6= (0). Consequently, it turns out that
Im is a submodule of M . Since 0 ∈ I, 0m = 0 ∈ Im, so Im is nonempty. Now, if a ∈ Im,
then a = im for some i ∈ I. Since i ∈ I, then −i ∈ I, meaning (−i)m = −im = −a ∈ Im.
Now, if a, b ∈ Im, say a = im and b = jm, then a + b = im + jm = (i + j)m ∈ Im.
Now that we know Im is a subgroup of M , let us confirm the module part. Let r ∈ R
and let b ∈ Im with b = jm, j ∈ I. Consider rb = r(jm) = (rj)m: since r ∈ R
and j ∈ I, I a left ideal of R, then rj ∈ I. Hence, rb ∈ Im for all r ∈ R, b ∈ Im,
so Im is a submodule of M . However, since Im is a nonzero submodule of M , then
Im = M . Well, −m ∈ M implies −m ∈ Im, so ∃t ∈ I so that −m = tm. Moreover, t is
a member of a left-quasi-regular ideal, so ∃s ∈ R so that s + t + st = 0. It follows that
0 = 0m = (s + t + st)m = sm + tm + stm = sm + (−m) + s(−m) = sm − sm − m = −m.
Since −m = 0, then m = 0, which would make Im = (0), a contradiction. Therefore, it
must be that I ⊂ J(R).

As Lemma 3.3.1 shows, from now on we can instantly declare that any quasi-
regular ideal is contained in the Jacobson radical. This will prove be immensely useful.
One application we can observe is that if a ring is semi-simple, then the only quasi-regular
ideal is the trivial ideal. Still, the most crutial implication of Lemma 3.3.1 is the following
theorem.
T
Theorem 3.3.1. J(R) = ρ, where ρ runs over all maximal regular left ideals of R.
T T
Proof. We know that J(R) = (ρ : R) and that (ρ : R) ⊂ ρ, therefore J(R) ⊂
ρ, where
T
ρ runs over all the maximal regular left ideals. Now, for the converse, suppose x ∈ ρ.
Consider the set S = {y +yx | y ∈ R}. We claim R = S. Otherwise, suppose S is a proper
subset of R. As it happens, the set S is a left ideal of R. Since 0 ∈ R, then 0+0x = 0 ∈ S.
Next, assume a ∈ S, then ∃c ∈ R so that a = c + cx. Since c ∈ R, then −c ∈ R, meaning
(−c) + (−c)x ∈ S. Of course, (−c) + (−c)x = −c − cx = −(c + cx) = −a, so −a ∈ S.
Moreover, if a and b belong in S, say a = c + cx and b = d + dx for some c, d ∈ R, then
a + b = (c + cx) + (d + dx) = (c + d) + (c + d)x ∈ S and ab = a(d + dx) = (ad) + (ad)x ∈ S
(Note this last part shows both closure under multiplication and the general ideal of how
S is a left ideal). Therefore, S is a left ideal of R. Furthermore, since r − r(−x) ∈ S for
all r ∈ R, then S is regular. By Lemma 3.1.1, since S is a proper regular left ideal of R,
T
S is contained in a maximal regular left ideal ρ0 . Now, since ρ is the intersection of all
 27

the maximal regular left ideal of R and ρ0 is a maximal regular left ideal of R, then we
see that x ∈ ρ0 . However, since y + yx ∈ ρ0 (S ⊂ ρ0 ) and yx ∈ ρ0 (x is in the left ideal
ρ0 ) for all y ∈ R, then R ⊂ ρ0 . This contradicts ρ0 being a maximal ideal. Therefore, S
is not a proper subset but instead all of R. Now, since R = {y + yx | y ∈ R} and −x ∈ R,
T
∃w ∈ R so that −x = w + wx, or in other words w + x + wx = 0. In conclusion, if x ∈ ρ,
T
then x is left-quasi-regular. This makes ρ a left-quasi-regular left ideal of R, and thus
T
ρ ⊂ J(R) by Lemma 3.3.1.

Finally, we have a pleasant way to describe the Jacobson radical of a ring. In

order to examine the radical of a ring, we must first take note of the ring’s maximal
regular left ideals. For example, note that since in a division ring the trivial ideal is
maximal and regular (D has unity), then every division ring is semi-simple. On the other
hand, the next example show why the ring of integers is semi-simple

Example 3.3.3. Since Z has unity, every ideal of Z is regular. Now, recall that every
maximal ideal of Z is of the form pZ, where p is a prime number. Moreover, if n and m
are integers, then nZ ∩ mZ = dZ, where d is the least common multiple of n and m. The
point being, the intersection of all maximal ideals of Z is principal and generated by the
only common multiple across all prime numbers: 0. Hence, J(Z) = (0)

Let us backtrack a bit. Notice that in the proof, since we showed that both
T T
J(R) = ρ and that ρ is a left-quasi-regular ideal of R, then J(R) is a left-quasi-
regular ideal of R. Moreover, by Lemma 3.3.1, J(R) is the biggest (contains every other)
left-quasi-regular ideal of R. Hence, we have the next theorem

Theorem 3.3.2. J(R) is a left-quasi-regular ideal of R and contains all the left-quasi-
regular left ideals of R. Moreover, J(R) is the unique maximal left-quasi-regular left ideal
of R.

Now, suppose there are elements a, b and c in a ring R where b is the left-
quasi-inverse of a and c is the right-quasi-inverse of a. So, we know b + a + ba = 0 and
a + c + ac = 0. Then, we have (b + a + ba)c = bc + ac + bac = 0 and b(a + c + ac) =
ba + bc + bac = 0. In other words, we get bc + ac + bac = ba + bc + bac. Subtract bc and
bac from both sides, we get ba = ac. Again, since b + a + ba = 0 = a + c + ac, subtracting
the elements that are equal we get b = c. Therefore, if an element has a left and a right
quasi-inverse, the two quasi-inverses are the same.
 28

Let us look specifically at the elements of J(R). If a ∈ J(R), and thus a is

left quasi-regular, then there exists a0 ∈ J(R) so that that a0 + a + a0 a = 0. Hence,
a0 = −a − a0 a ∈ J(R) since a ∈ J(R). Now, since a0 ∈ J(R), then a0 is also left-quasi-
regular element and there exists a00 ∈ R so that a00 + a0 + a00 a0 = 0. Since a is the left-
quasi-inverse of a0 and a00 is the right-quasi-inverse of a0 , we have by the above paragraph,
a = a00 . Therefore, we can say that a is right-quasi-regular (0 = a00 +a0 +a00 a0 = a+a0 +aa0 ).
As a result, J(R) is a right-quasi-regular ideal of R.
The fact that J(R) is a left-quasi-regular ideal and a right-quasi-regular ideal
of R implies that there is only one Jacobson radical. Indeed, if everything we have done
thus far is in regards to left modules and left ideals, then we should regard J(R) as the
left Jacobson radical. Similarly, if we do the analogues for right modules and right ideals,
we would end up with a right Jacobson radical. Denote J(R)l as the left Jacobson radical
and J(R)r as the right Jacobson radical. Note that the analogue for Lemma 3.3.1 would
state that any right-quasi-regular ideal of R is contained in J(R)r . Since we have shown
J(R)l is right-quasi-regular, then J(R)l ⊂ J(R)r . Similarly, we could prove that J(R)r
is left-quasi-regular, hence J(R)r ⊂ J(R)l . As we all know, this implies J(R)l = J(R)r ,
meaning there is only one Jacobson radical! As a last note, we can therefore say that the
Jacobson radical is the intersection of all maximal regular right ideals of R.
Another note of interest is an analogy to Lemma 3.2.2 for the Jacobson radical.
Lemma 3.2.2 depicts how a module can turn into a faithful module over a different ring.
The following theorem shows how one can go from a ring to a semi-simple ring.

Theorem 3.3.3. J(R/J(R)) = (0); that is, R/J(R) is semi-simple.

Proof. Consider R/J(R) and let ρ be a maximal regular left ideal of R. We know that
J(R) ⊂ ρ by Theorem 3.3.1. Now, we will prove that ρ/J(R) is maximal in R/J(R).
Suppose ρ/J(R) ⊂ I/J(R) ⊂ R/J(R) and ρ/J(R) 6= I/J(R). Then, consequently ρ ⊂
I ⊂ R, where ρ 6= I. However, since ρ is maximal in R, it implies I = R. Hence,
I/J(R) = R/J(R), meaning ρ/J(R) is maximal in R/J(R). Moreover, since ρ is left
regular in R, ρ/J(R) will also be regular in R/J(R). Say a ∈ R so that r − ra ∈ ρ,
∀r ∈ R, then (r + J(R)) − (r + J(R))(a + J(R)) = (r + J(R)) − (ra + J(R)) = (r −
ra) + J(R) ∈ ρ/J(R) ∀r ∈ R. In short, the image of all of the maximal regular left
ideals of R will remain maximal regular left ideals in R/J(R). However, this does not
imply that all of the maximal regular left ideals of R/J(R) come from R. On that
 29

T T
note, if J(R) = ρ, then ( ρ)/J(R) = J(R)/J(R) = (0) in R/J(R). One thing
T T T
that we can say is that ( ρ)/J(R) = (ρ/J(R)). Let x + J(R) ∈ ( ρ)/J(R), then
T T
x ∈ ρ, so x + J(R) ∈ ρ/J(R) for all ρ, and thus x + J(R) ∈ (ρ/J(R)). On the
T
other hand, let x + J(R) ∈ (ρ/J(R)), then x + J(R) ∈ ρ/J(R) for all ρ. Moreover,
if x + J(R) ∈ ρ/J(R) for each ρ, then x ∈ ρ for each ρ: Say x + J(R) = r + J(R)
for some r ∈ ρ, then x − r ∈ J(R) ⊂ ρ; since both x − r and r belong to ρ, then
T
x ∈ ρ. Therefore, x + J(R) ∈ ρ/J(R) for all ρ implies x is in all ρ and thus x ∈ ρ,
T
meaning x + J(R) ∈ ( ρ)/J(R). Recalling that J(R/J(R)) is the intersection of all
maximal regular left ideals of R/J(R), we can show why R/J(R) is a semi-simple ring:
T T
J(R/J(R)) ⊂ (ρ/J(R)) = ( ρ)/J(R) = (0).

Let us talk about the relation between the two-sided ideals of a ring and the
ring’s Jacobson radical. Let A be an ideal of a ring R. If A is an ideal of R, then A is a
subring of R. So, if A is an ideal of a ring, then A itself is a ring. If that is the case, can
we relate the Jacobson radical of A with the Jacobson radical of R? The answer turns
out to be quite satisfying, as we see in the following theorem and corollary.

Theorem 3.3.4. If A is an ideal of R, then J(A) = A ∩ J(R).

Proof. Let A be an ideal of R. Now, let a ∈ A ∩ J(R). So, a ∈ J(R). Since J(R) is
left-quasi-regular, then ∃b ∈ R so that b + a + ba = 0. Solving for b, we get b = −a − ba.
Since also a ∈ A, an ideal, and b ∈ R, then both ba and −a belong in A, meaning
b = −a − ba ∈ A. Explicitly, this shows that A ∩ J(R) is a left-quasi-regular ideal of A,
therefore A ∩ J(R) ⊂ J(A) by Lemma 3.3.1.
Now, suppose ρ is a maximal regular left ideal of R. Let ρA = ρ ∩ A. If A ⊂ ρ,
then ρA = A, meaning J(A) ⊂ ρA . In other words, J(A) is a subset of ρA for all maximal
regular left ideal ρ of R that contain A. Then, let us examine all of the maximal regular
left ideals of R that do not contain A. If A 6⊂ ρ, then the maximality of ρ forces A+ρ = R.
That is, since both A and ρ are left ideals of R, then A + ρ is a left ideal of R. Moreover,
since A 6⊂ ρ, then ∃a ∈ A so that a ∈
/ ρ. Since a ∈ A and 0 ∈ ρ, then a = a + 0 ∈ A + ρ.
Therefore, ρ is a proper subset of A + ρ. As a result, since ρ is a left maximal ideal of R
and A + ρ a bigger left ideal than ρ in R, then A + ρ = R. Now, we can apply the Second
Isomorphism Theorem:

R/ρ = A+ρ ∼
= A
= A/ρA
ρ A∩ρ
 30

Now, since R/ρ is simple, then A/ρA is simple, making ρA a maximal left ideal
of A. Moreover, since ρ is regular, ∃b ∈ R so that x − xb ∈ ρ for all x in R. Yet, since
b ∈ R = A+ρ, b = a+r for some a ∈ A, r ∈ ρ. Then, x−xb = x−x(a+r) = x−xa−xr ∈ ρ.
Of course, since r ∈ ρ, xr ∈ ρ, which implies that really x−xa ∈ ρ for all x ∈ R. Therefore,
if we let instead x ∈ A, then we know x − xa ∈ ρ and also x − xa ∈ A (A is an ideal,
so both x and xa belong in A), meaning ∃a ∈ A so that x − xa ∈ ρA for all x ∈ A; so
ρA is a regular left ideal in A specifically. In fact, ρA is a maximal regular left ideal of
(specifically) A. Therefore, whether A ⊂ ρ or A 6⊂ ρ, J(A) ⊂ ρA for all maximal regular
T T
left ideals ρ of R. Hence, J(A) ⊂ ρA = ( ρ) ∩ A = J(R) ∩ A. So, A ∩ J(R) ⊂ J(A)
and J(A) ⊂ A ∩ J(R), so A ∩ J(R) = J(A).

Corollary 3.3.1. If R is semi-simple, then so is every ideal of R.

Proof. If J(R) = 0, then J(A) = J(R)∩A = (0)∩A = (0), meaning A is semi-simple.

 31

Chapter 4

Nilpotency, Idempotency, and

Artinian Rings

Now that we have covered semi-simple rings, let us move on to Artinian rings.
We will see that an Artinian ring is a ring in which every descending chain of ideals
becomes stationary. Meaning, there is an ending point to every descending chain of
proper ideals. Notice there are two key components for a desending chain of ideals in an
Artinian ring: finiteness and stability. Finiteness in the respect that, in a chain of ideals,
there is eventually an end to the difference in size of the ideals and a final destination
is reached. Stability in the sense that once a certain point is reached, the ideals remain
the same thereafter. As we shall see, nilpotency and idempotency respectively exemplify
these characteristics and are closely related to Artinian rings.
In this chapter, we will see what nilpotency, idempotency, and Artinian rings
have to say about the Jacobson radical. First, we will explore what nilpotency has to do
with the Jacobson radical. Next, we will start assuming our ring to be Artinian and see
what implications arise. Finally, idempotency will prove to be ever present in Artinian
rings. Once we explore these three facets of Artinian rings, we will make our way to the
Wedderburn-Artin theorems.
 32

4.1 Nilpotence

Nilpotence, in essence, is to be extinguished by oneself. That is, a nilpotent

entity is one that becomes nothing after multiple applications of itself. What this means
for ring theory, we observe in the following definition.

Definition 4.1.1. Let R be a ring, then:

(i) a ∈ R is nilpotent if there exists n ∈ N such that an = 0.

(ii) (right, left, two-sided) ideal I is nil if all elements of I are nilpotent.

(iii) An ideal J is nilpotent if there exists m ∈ N such that J m = (0). It is equivalent to

say that there exists m ∈ N such that a1 a2 · · · am = 0, ∀ai ∈ J.

Out of all the topics outside basic ring theory, nilpotency is the least strange.
While studying rings such as Zn and M atn (D), nilpotent elements have been present.

Example 4.1.1. In Zpk , clearly p is nilpotent. Moreover, any multiple of p is also

nilpotent.

Example 4.1.2. For any ring R and any positive integer n, any matrix in M atn (R) that
is either upper triangular or lower triangular is nilpotent.

Lemma 4.1.1. If a is nilpotent, then a is quasi-regular (left and right).

Proof. Say am = 0, construct b = −a + a2 − a3 + ... + (−1)m−1 am−1 . Then, we get that:

b + a + ba = −a + a2 − a3 + . . . + (−1)m−1 am−1 + a

+ −a2 + a3 + · · · + (−1)m−2 am−1 + (−1)m−1 am

= (−1)m−1 am
=0

Similarly:

a + b + ab = a + −a + a2 − a3 + . . . + (−1)m−1 am−1

+ −a2 + a3 + · · · + (−1)m−2 am−1 + (−1)m−1 am

= (−1)m−1 am
=0

Therefore, a is both left and right quasi-regular.

 33

Lemma 4.1.1 illustrates the importance of nilpotency. The lemma implies that
every nil right or left ideal of a ring R is left or right quasi-regular, thus contained in J(R).
On the other hand, if R is semi-simple, then there are no nonzero nil ideals. Similarly, if
R is a nil ideal of R, then it is a radical ring.

Example 4.1.3. Recall that in Zpk , the set P = {pm | m ∈ Z} contains only nilpotent
elements. Additionally, P is an ideal of Zpk , so P is a nil ideal. Therefore, P itself as a
ring is a radical ring (that is, J(P ) = P ).

Lemma 4.1.2. All nilpotent ideals are nil ideals.

Proof. Let I be a nilpotent ideal of a ring R and let m ∈ N so that I m = (0). So, if a ∈ I,
then am ∈ I m . Hence, am ∈ (0), meaning am = 0. Thus, every element of I is nilpotent,
making I a nil ideal.

Since all nil ideals are contained in the Jacobson radical of a ring, then all
nilpotent ideals are contained in the radical. Thus, if a ring R is semi-simple, then we
know there are no nonzero nilpotent (and no nil) ideals. Again, if R is a nilpotent ideal
of R, then R is a radical ring.

Example 4.1.4. Let n be a positive integer and consider Zn . Suppose n = pk11 · · · pkr r ,
where pi are prime numbers and ki ≥ 1 for all i. Let m = p1 · · · pr , then m is nilpotent
in Zn . Particularly, if k = max{ki | 1 ≤ i ≤ r}, then mk = pk1 · · · pkr = nq, for some q ∈ Z.
On that note, the ideal M = {ms | s ∈ Z} is nilpotent since M k = (0). Again, as a ring,
M is a radical ring.

4.2 Artinian Rings

Finally, we can discuss Artinian rings. Artinian rings generalize finite rings as
well as rings that are finite-dimensional vector spaces over division rings. In addition,
Artinian rings are seen as the duals to Noetherian rings in respect to chains of ideals.
Regardless, Artinian rings are those in which any non-empty set of ideals has a minimal
element under set containment. We shall show that an Artinian ring is equivalently a ring
in which any descending chain of ideals terminates. Afterwards, we shall briefly explore
facts about Artinian rings.
 34

Definition 4.2.1. A ring is said to be (left, right) Artinian if any non-empty set of (left,
right) ideals has a minimal element.

From now on, we shall say Artinian for left Artinian, as we still have many left
ideals to deal with. There is a famous, equivalent definition for Artinian rings that we
will use frequently. Theorem 4.2.1 illustrates what an Artinian ring means in terms of
chains of ideals. In various cases, we will deal with descending chains of ideals. Therefore,
Theorem 4.2.1 will be of great use. Also, note that the Axiom of Choice is necessary for
the converse of the theorem.

Theorem 4.2.1. A ring R is Artinian if and only if any descending chain of left ideals
of R becomes stationary.

Proof. Let R be an Artinian ring and let I1 ⊃ I2 ⊃ I3 ⊃ · · · be a descending chain of left

ideals of R. Consider the set M = {Ii } composed of these ideals. Since R is Artinian,
then M has a minimal element Im . However, if k > m, then Ik ⊂ Im . Hence, Ii = Im for
i ≥ m.
Now, let R be a ring in which any descending chain of left ideals of R becomes
stationary. Let M be a non-empty set of left ideals of R. Using the Axiom of Choice,
pick I1 ∈ M. If I1 is minimal, we are done. So, suppose I1 is not minimal. Then, we
pick I2 ∈ M such that I1 ⊃ I2 . Again, either I2 is minimal, or we continue to pick an
I3 ∈ M such that I2 ⊃ I3 . In this manner, suppose we get a chain I1 ⊃ I2 ⊃ I3 ⊃ · · · .
Since this chain of left ideals becomes stationary, then there is an m so that Im = Ii for
i ≥ m. Then, Im is a minimal element of M.

Artinian rings posses multiple helpful properties. For this paper, we shall recall
a handful of facts that will prove to be invaluable.

Lemma 4.2.1. If R is a Artinian ring and I 6= (0) is a left ideal of R, then I contains
a minimal left ideal of R.

Proof. Let R be an Artinian ring and let I 6= (0) be a left ideal of R. Let M be the set
of all nontrivial left ideals of R that are contained in I. Of course, I ⊂ I, so M is not
empty. Since M is a set of left ideals of R and R is Artinian, then there exists a left ideal
µ in M that is minimal in M. To prove that µ is a minimal ideal of R, suppose there
exists a left ideal J 6= (0) of R such that (0) ⊂ J ⊂ µ. If so, then J ⊂ µ ⊂ I and J being
 35

nontrivial left ideal of R implies J ∈ M. However, if J is a proper subset of µ, then we

have a contradiction of µ being a minimal element of M. Hence, J = µ, meaining µ is a
minimal left ideal of R.

Example 4.2.1. Any division ring is Artinian. Recall that every (left, right, two-sided)
ideal is either the trivial ideal or the entire ring.

Theorem 4.2.2. If D is a division ring, then M atn (D) is Artinian.

Proof. Let D be a division ring and let I1 ⊃ I2 ⊃ I3 ⊃ · · · be a descending chain of left

ideals of M atn (D). Note that M atn (D) is a D-module, meaning M atn (D) is a vector
space over D. Moreover, if I is a left ideal of M atn (D), then I is a left vector subspace of
M atn (D). By definition of left ideal, we know I ⊂ M atn (D), 0 ∈ I and I is closed under
addition. To show I is closed under scalar multiplication, suppose a ∈ D and v ∈ I. Let e
be the n × n identity matrix, then av = aev = (ae)v. Since ae is an element of M atn (D)
and v belongs to a left ideal of M atn (D), then (ae)v ∈ I. Hence, av ∈ I, we have that I
is closed under scalar multiplication and thus a left subspace of M atn (D). Therefore, the
descending chain of left ideals I1 ⊃ I2 ⊃ I3 ⊃ · · · is also a descending chain of left vector
subspaces of M atn (D). Of course, every subspace Ii has a basis Bi , and since M atn (D)
is n2 -dimensional, then we know that both |B1 | ≥ |B2 | ≥ |B3 | ≥ · · · and |B1 | ≤ n2 . Since
{|B1 |, |B2 |, |B3 |, · · · } is a set on non-negative integers, it must have a minimal element,
say |Bm |. Since, for any k ≥ m, |Bm | = |Bk | (|Bm | ≤ |Bk | and |Bm | ≥ |Bk |) and Im ⊃ Ik ,
Ik as a subspace of equal dimension to Im implies Im = Ik for any k ≥ m. Hence,
the descending chain of left ideals I1 ⊃ I2 ⊃ I3 ⊃ · · · becomes stationary. Therefore,
M atn (D) is an Artinian ring.

Lemma 4.2.2. A homomorphic image of an Artinian ring is Artinian

Proof. Let R be an Artinian ring and let f : R → S be a ring homomorphism. We

want to show that f (R) is an Artinian ring. Certainly, f (R) is a ring since f is a ring
homomorphism. Now, suppose J1 ⊃ J2 ⊃ J3 ⊃ · · · is a descending chain of left ideals
of f (R). Then, we know that f −1 (Ji ) is a left ideal of R for all i. Moreover, f −1 (Ji ) ⊃
f −1 (Ji+1 ) for all i: if x ∈ f −1 (Ji+1 ), then f (x) ∈ Ji+1 ⊂ Ji , meaning x ∈ f −1 (Ji ).
Therefore, f −1 (J1 ) ⊃ f −1 (J2 ) ⊃ f −1 (J3 ) ⊃ · · · is a chain of descending ideals of R. Since
R is Artinian, there exists an ideal I that is minimal in the chain. In other words, I is a
 36

left ideal of R and I ⊂ f −1 (Ji ) for all i. On that note, we can see that f (I) ⊂ Ji for all i.
Let x ∈ f (I) with x = f (a), for some a ∈ I. Since I ⊂ f −1 (Ji ) for all i, then a ∈ f −1 (Ji )
for all i. Therefore, f (a) ∈ Ji , meaning x ∈ Ji . Lastly, in order to claim f (I) is the
minimal element in the descending chain, we must be sure that f (I) is an ideal of f (R).
Of course, I being a subring of R makes f (I) a subring of f (R). Now, let f (r) ∈ f (R)
and f (i) ∈ f (I), then f (r)f (i) = f (ri) ∈ f (I) since ri ∈ I. Hence, f (I) is a left ideal of
f (R), making f (I) the minimal element in the chain J1 ⊃ J2 ⊃ J3 ⊃ · · · .

Recall that in the previous section, we discovered that all nilpotent ideals of a
ring R are contained in the Jacobson radical J(R). Mainly, we noted that all nilpotent
ideals are nil ideals and all nil ideals are quasi-regular ideals. In turn, not only did J(R)
contain all of the quasi-regular ideals of R, but J(R) itself was quasi-regular. Similarly,
since we know that J(R) contains all nil and nilpotent ideals, can we say that J(R) is
either of them?

Theorem 4.2.3. If R is Artinian, then J(R) is nilpotent.

Proof. Let J = J(R); consider the descending chain of left ideals J ⊃ J 2 ⊃ . . . ⊃ J n ⊃ . . .

. Since R is Artinian, there is an integer n such that J n = J n+1 = . . . = J 2n = . . . . We
want to prove that J n = (0); so, assume that J n 6= (0). Let W = {x ∈ J | J n x = (0)};
we will show W is an ideal of R. First, 0 ∈ J (since J is an ideal) and j0 = 0, ∀j ∈ J n ,
so 0 ∈ W . Next, suppose that a ∈ J and ja = 0, ∀j ∈ J n . Then, −a ∈ J and
j(−a) = −ja = −0 = 0, ∀j ∈ J n , meaning −a ∈ W . Next, let a, b ∈ W ; then,
j(a + b) = ja + jb = 0 + 0 = 0 and j(ab) = (ja)b = 0b = 0, ∀j ∈ J n . Of course, J an ideal
implies both a + b and ab belong to J, so W is closed under addition and multiplication.
Now, suppose r ∈ R and w ∈ W and consider both rw and wr. Since w ∈ J, both rw
and wr belong to J. Now, let j ∈ J n ; then, jrw = (jr)w = 0 and jwr = (jw)r = 0r = 0
(j ∈ J n , an ideal, implies jr ∈ J n , thus w annihilates it). Therefore, W is an ideal.
Now, if J n ⊂ W , then J n J n = (0); meaning, J n = J 2n = J n J n = (0), proving our
statement. So naturally, consider the case in which J n 6⊂ W . If so, then R/W is a ring
and J n /W 6= (0) is an ideal.
Now, consider the elements x+W such that (J n /W )(x+W ) = (0). If (J n /W )(x+
W ) = (0), then (j +W )(x+W ) = W for all j ∈ J n . In other words, jx ∈ W for all j ∈ J n ,
therefore, we can say J n x ⊂ W . Consequently, we can deduce (0) = J n (J n x) = J n J n x =
 37

J 2n x = J n x, placing x in W . Therefore, if (J n /W )(x + W ) = (0), then x + W = 0 + W ,

the zero element of R/W .
Finally, since J n /W 6= (0) and R/W the homomorphic image of an Artinian
ring, Lemma 4.2.1 implies that there exists a minimal left ideal µ of R/W contained in
J n /W . However, µ being a minimal left ideal of R/W implies that µ is a simple R/W -
module and thus annihilated by J(R/W ). Additionally, J n /W ⊂ J(R/W ): If a ∈ J n ,
then a ∈ J(R), so there exists b ∈ R so that b + a + ba = 0. Hence, (b + W ) + (a +
W ) + (b + W )(a + W ) = (b + a + ba) + W = 0 + W ∈ R/W , meaning that J n /W is
a left-quasi-regular ideal of R/W and thus a subset of J(R/W ). That being said, since
J n /W ⊂ J(R/W ) and µ is a simple R/W -module, then (J n /W )µ = (0). From the last
line of the paragraph above, this implies that every element of µ is the zero element,
contradicting µ being a minimal left ideal (minimal ideals are nonzero ideals). Therefore,
J n = (0) and thus J(R) is nilpotent.

While we have seen that every nilpotent ideal is a nil ideal, the converse is not
always true. However, if a nil ideal is an ideal of an Artinian ring, then Theorem 4.2.3
shows why such an ideal must also be nilpotent.

Corollary 4.2.1. If R is Artinian, then any nil ideal of R is nilpotent.

Proof. Let I be a nil one-sided ideal. Since every element of I is nilpotent, then by
Lemma 4.1.1 every element of I is quasi-regular. Hence, I ⊂ J(R). Now, since R is
Artinian, J(R) is nilpotent; let m ∈ N such that (J(R))m = (0). Hence, all there is
left to do is prove I m ⊂ (J(R))m . Let x ∈ I m with x = a1 a2 . . . am , where ai ∈ I for
i = 1, 2, . . . , m. Since I ⊂ J(R), then we can say x = a1 a2 . . . am , with ai ∈ J(R) for
i = 1, 2, . . . , m. Hence, x ∈ (J(R))m . So, I m ⊂ (J(R))m ⊂ (0), so I m is nilpotent.

4.3 Idempotency

Idempotent elements are those which remain unchanged when raised to a higher
power. Mainly, as we may notice, the square of an element need only be itself in order
to have the element to be idempotent. In any case, since idempotent elements are ever
present in Artinian rings, we shall prove some valuable theorems involving idempotent
elements. But first, let us look at the definition, and some examples, of idempotent
elements.
 38

Definition 4.3.1. An element e 6= 0 in R is an idempotent if e2 = e.

Example 4.3.1. In Z10 , both 5 and 6 are idempotent.

Example 4.3.2. Let D be a division ring and consider M atn (D) for n > 1. Let Exy be
the n × n matrix with exy = 1 and all other entries equal to zero (1 ≤ x, y ≤ n). Then,
not only are all Eii idempotent, but also E11 + Enn . For a concrete example, in M at2 (Z),
any matrix such that a11 = a12 = k and a21 = a22 = 1 − k for k ∈ Z is idempotent.

Lemma 4.3.1. Let R be a ring having no nonzero nilpotent ideals. Suppose that µ 6= (0)
is a minimal left ideal of R. Then µ = Re for some idempotent e ∈ R.

Proof. Suppose R is a ring having no non-zero nilpotent ideals and suppose that µ 6= (0)
is a minimal left ideal of R. Since R has no non-zero nilpotent ideal and µ 6= (0), then we
know µ2 6= (0). So, it is safe to say that there exists x ∈ µ so that µx 6= (0). Moreover,
µx is a left ideal of R. First, since µ is a left ideal, 0 ∈ µ, therefore 0x = 0 ∈ µx, so µx is
non-empty. Second, suppose a ∈ µx with a = mx, where m ∈ µ, then −m ∈ µ and thus
(−m)x = −mx = −a ∈ µx, so µx has inverses. Third, suppose a, b ∈ µx with a = mx,
b = nx (m, n ∈ µ), then a + b = mx + nx = (m + n)x ∈ µx (m + n ∈ µ). Moreover,
ab = (mx)(nx) = ((mx)n)x ∈ µx since n ∈ µ (a left ideal) implies (mx)n ∈ µ (this last
part shows also why µx is a left ideal of R). Additionally, since x ∈ µ, then mx ∈ µ for
all m ∈ µ, meaning µx ⊂ µ. Since µx ⊂ µ, µx 6= (0), then µ being a minimal ideal of R
implies µx = µ. Therefore, since x ∈ µ, there exists an element e ∈ µ so that x = ex. On
that note, by multiplying e on the left, we get ex = e2 x; in other words, (e2 − e)x = 0.
Consider the set C = {c ∈ µ | cx = 0}; C is a left ideal of R. First, 0 ∈ µ and 0x = 0,
so 0 ∈ C. Second, let c ∈ C. Then, cx = 0 implies (−c)x = −cx = −0 = 0. Of course,
c ∈ C implies c ∈ µ, so −c ∈ µ; Hence, C has inverses. Third, let c, d ∈ C, then c + d ∈ µ
and (c + d)x = cx + dx = 0 + 0 = 0, so c + d ∈ C. Lastly, let r ∈ R and c ∈ C, then
c ∈ µ implies rc ∈ µ and (rc)x = r(cx) = r0 = 0, so C is a left ideal of R. By definition,
C ⊂ µ. Since µ is a minimal left ideal, either C = µ or C = (0). However, if C = µ,
then it would imply µx = (0). Of course, µx 6= (0), so C 6= µ. Thus, C = (0). Since
e2 − e ∈ C, then e2 − e = 0, so e2 = e. Also, since µx 6= (0), we can be sure that x 6= 0.
Consequently, since ex = x 6= 0, we can be sure that e 6= 0. We conclude that e is an
idempotent of R. Now, consider the left principal ideal Re. If re ∈ Re, since e ∈ µ, then
re ∈ µ; so, Re ⊂ µ. Again, µ being a minimal ideal of R implies Re = (0) or Re = µ.
 39

Still, since e ∈ µ and µ ⊂ R, we can say ee = e2 = e ∈ Re; since e 6= 0, then Re 6= (0).

Thus, Re = µ for some idempotent e in R.

There are a few things to note about the previous lemma. First, recall that
a semi-simple ring is a ring that has no nonzero nilpotent ideals. Hence, we can apply
Lemma 4.3.1 to any semi-simple ring. Furthermore, since any nonzero left ideal of an
Artinian ring contains a minimal left ideal, we shall take full advantage of this lemma
for the structure of semi-simple Artinian rings. As for rings that are Artinian but not
semi-simple, there is a theorem that guarantees the existence of idempotent elements on
nonzero left ideals. First, however, we must prove the following peculiar lemma.

Lemma 4.3.2. Let R be a ring and suppose that for some a ∈ R, a2 − a is nilpotent.
Then either a is nilpotent or, for some polynomial q(x) with integer coefficients, e = aq(a)
is a non-zero idempotent.

Proof. Let R be a ring and suppose that for some a ∈ R, k ∈ N, (a2 −a)k = 0. Expanding
(a2 − a)k = a2k + . . . + (−a)k = 0 and solving for ak , we get ak = ak+1 p(a), where
p(x) has integer coefficients. Of course, ak+1 = ak a, then ak = ak+1 p(a) = ak ap(a) =
(ak+1 p(a))ap(a) = ak+2 p(a)2 . Therefore, repeating this process, eventually we get that
ak = a2k p(a)k . Now, either ak = 0 or ak 6= 0. However, if ak = 0, then by definition
a is nilpotent. So, suppose ak 6= 0, then e = ak p(a)k 6= 0 and e2 = a2k p(a)2k =
(a2k p(a)k )p(a)k = ak p(a)k = e. Say q(a) = ak−1 p(a)k , then e is the non-zero idempotent
we were looking for.

Theorem 4.3.1. If R is an Artinian ring and I 6= (0) is a non-nilpotent left ideal of R,

then I contains a non-zero idempotent element.

Proof. Let R be an Artinian ring and I 6= (0) be a non-nilpotent left ideal of R. By

Theorem 4.2.3, I being non-nilpotent implies I 6⊂ J(R). Consider the ring R/J(R). By
Theorem 3.3.3, we know R/J(R) is semi-simple; Thus, R/J(R) contains no non-zero
nilpotent ideals. Now, since I 6⊂ J(R), then I/J(R) 6= (0). Moreover, since R/J(R) is
the image of an Artinian ring and I/J(R) is a non-zero ideal, then there is a minimal
ideal µ/J(R) inside of I/J(R). Consequently, since R/J(R) has no non-zero nilpotent
ideals and µ/J(R) is minimal in R/J(R), then there is an idempotent element e0 +J(R) ∈
µ/J(R). Let a ∈ µ map onto e0 +J(R); then a2 −a maps onto (e0 +J(R))2 −(e0 +J(R)) =
 40

((e0 )2 − e0 ) + J(R) = J(R), meaning a2 − a ∈ J(R). Therefore, Since R being Artinian

makes J(R) nilpotent, then a2 − a is nilpotent. By Lemma 4.3.2, either a is nilpotent
or e = aq(a) is an idempotent element. Suppose that for some k ∈ N, ak = 0, then
ak +J(R) = 0+J(R). However, J(R) = ak +J(R) = (a+J(R))k = (e0 +J(R))k = e0 +J(R)
contradicts e0 + J(R) as being idempotent. Thus, a is not nilpotent. Since e = aq(a) is
an idempotent and a ∈ µ, then e ∈ µ. Finally, µ/J(R) ⊂ I/J(R) implies µ ⊂ I. Hence,
e ∈ I is an idempotent contained in I.

The last result of this chapter will be used to find minimal ideals. The more
we progress, the more is evident that minimal ideals are becoming integral to the study
of semi-simple Artinian rings. Still, the following theorem will be used mainly in our
conclusion chapter and will allow us to test and find minimal ideals by using idempotent
elements.

Theorem 4.3.2. Let R be a ring having no nonzero nilpotent ideals and suppose that e
is an idempotent element in R. Then, Re is a minimal left ideal of R if and only if eRe
is a division ring.

Proof. Let R be a ring having no nonzero nilpotent ideals and suppose Re is a minimal
left ideal of R with e being idempotent. First, we can see that eRe is a ring with unity
e. First, e0e = 0 ∈ eRe. Second, if eae ∈ eRe, then e(−a)e = −eae ∈ eRe. Third,
if eae, ebe ∈ eRe, then eae + ebe = e(ae + be) = e((a + b)e) = e(a + b)e ∈ eRe and
(eae)(ebe) = eae2 be = e(aeb)e ∈ eRe. To show e is the multiplicative identity of eRe,
surely eee = e ∈ eRe and for any eae ∈ eRe, we have e(eae) = e2 ae = eae = eae2 =
(eae)e. Lastly, to prove eRe is a division ring, consider a nonzero element eae ∈ eRe.
Obviously, Reae is a principal left ideal of R. Now, note that Reae ⊂ Re: if reae ∈ Reae,
then since rea ∈ R, we get (rea)e ∈ Re. Moreover, since e ∈ R, then e(eae) = eae ∈ Reae,
meaning Reae 6= (0). Since Reae is a nonzero left ideal inside the minimal left ideal Re,
then Re = Reae. Since e = ee ∈ Re, there exists b ∈ R so that e = beae. Upon
further calculation, we see that e2 = ebeae, meaning e = (ebe)(eae). Therefore, eae has
a multiplicative inverse, which means eRe is a division ring.
Now, for the converse, suppose R is a ring that has no nonzero nilpotent ideals
and e an idempotent of R with eRe being a division ring. To prove Re is a minimal
left ideal, suppose µ ⊂ Re and µ 6= (0). If eµ = (0), then we can see that µ2 = µµ ⊂
 41

(Re)µ = (0), meaning µ is a nonzero nilpotent ideal of R, contradicting the hypothesis.

So, eµ 6= (0). On that note, since eµ 6= (0), there must be some m ∈ µ such that em 6= 0.
Since µ ⊂ Re, then m ∈ Re. Consequently, em ∈ eRe. Since em 6= 0 and eRe is a
division ring, there must be a multiplicative inverse ese ∈ eRe so that (ese)(em) = e.
Upon close inspection, we can see that e = (esee)m ∈ Rm and also since m ∈ µ, a left
ideal, then Rm ⊂ µ. In short, e belongs to µ. Since e ∈ µ, then Re ⊂ µ. Therefore,
µ = Re, so Re is a minimal ideal of R.

Corollary 4.3.1. If R has no nonzero nilpotent ideals and if e is an idempotent in R,

then Re is a minimal left ideal of R if and only if eR is a minimal right ideal of R.

Proof. Since the right analogue of Theorem 4.3.2 holds, we can say that eR is a minimal
right ideal if and only if eRe is a division ring. Thus, Re being a minimal ideal of R is
logically equivalent to eR being a minimal ideal of R.

Example 4.3.3. Since Z10 is unitary, every ideal of Z10 is regular. By intersecting the
only two maximal ideals of Z10 , h2i and h5i, we see that Z10 is semi-simple. Therefore,
Z10 has no nonzero nilpotent ideals. Of course, since there are a finite number of ideals,
Z10 is Artinian. Recall that 6 ∈ Z10 is idempotent. We can see that 6Z10 = {0, 2, 4, 6, 8}
is a principal ideal of Z10 and is minimal since |6Z10 | = 5, a prime number. Therefore,
6Z10 6 is a division ring. In fact, since Z10 is commutative, 6Z10 6 = 6Z10 is a field with 6
as its unity.

Example 4.3.4. Let D be a division ring and let R = M atn (D) with n > 1. For each
k so that 1 ≤ k ≤ n, Ik = {(aij ) ∈ R | aij = 0 for j 6= k}. Recall that matrix Ekk is
idempotent in R. Moreover, Ik = R(Ekk ) for all k. Since Ik2 6= (0) for any k, we have that
Ik is a minimal ideal of R if and only if (Ekk )R(Ekk ) is a division ring. As it turns out,
(Ekk )R(Ekk ) = {cEkk | c ∈ D} is a division ring; Ekk is the unity of (Ekk )R(Ekk ), and if
cEkk 6= 0, then c 6= 0, so there exists d ∈ D such that cd = 1, then dEkk ∈ (Ekk )R(Ekk )
and (cEkk )(dEkk ) = Ekk . Hence, Ik is a minimal ideal for all k.
 42

Chapter 5

Wedderburn-Artin Theorems

Finally, we have arrived at the main purpose of this paper. The first Wedderburn-
Artin theorem illustrates the structure of semi-simple Artinian rings. As a result, all
theorems leading to it will be about semi-simple Artinian rings. The second Wedderburn-
Artin theorem demontrates the structure of simple Artinian rings. As for the theorems
leading to it, primitivity will play an important role.

5.1 Semi-simple Artinian Rings

First, we will establish the fact that every left ideal of a semi-simple Artinian
ring is principal and generated by an idempotent element. This is an improvement from
Theorem 4.3.1. In addition, the two corollaries that follow Theorem 5.1.1 will detail
important information about two-sided ideals of semi-simple Artinian rings.

Theorem 5.1.1. Let R be a semi-simple Artinian ring and I 6= (0) is a left ideal of R.
Then, I = Re for some idempotent e ∈ R.

Proof. Let R be a semi-simple Artinian ring and I 6= (0) be a left ideal of R. Since R is
semi-simple, I is not nilpotent. Moreover, since I is a non-zero non-nilpotent left ideal of
R, by Theorem 4.3.1, there exists an idempotent e in I. Let A(e) = {x ∈ I | xe = 0}, we
will show A(e) is a left ideal of R. Of course, 0 ∈ I and 0e = 0, so 0 ∈ A(e). Next, say
a ∈ A(e), then −a ∈ I and (−a)e = −ae = −0 = 0, so −a ∈ A(e). Also, if a, b ∈ A(e),
then a + b ∈ I and (a + b)e = ae + be = 0 + 0 = 0, so a + b ∈ A(e). Finally, suppose r ∈ R
and a ∈ A(e), then ra ∈ I and (ra)e = r(ae) = r0 = 0, so ra ∈ A(e). Now, consider the
 43

set M = {A(e) | e2 = e 6= 0 and e ∈ I}, the set of all A(e) for idempotent elements e from
I. Since we know there is at least one idempotent element in I, then M is a non-empty
set of left ideals of R. Since R is Artinian, M has a minimal element A(e0 ).
Suppose A(e0 ) 6= (0). As a nonzero, nonnilpotent left ideal of the semi-simple
Artinian ring R, A(e0 ) must have an idempotent e1 . By the definition of A(e0 ), e1 ∈ I
and e1 e0 = 0. Consider the element e∗ = e0 + e1 − e0 e1 . Not only is e∗ ∈ I, but e∗ is
idempotent:

(e∗ )2 = (e0 + e1 − e0 e1 )(e0 + e1 − e0 e1 )

= [(e0 )2 + e1 e0 − e0 e1 e0 ] + [e0 e1 + (e1 )2 − e0 (e1 )2 ] + [−(e0 )2 e1 − e0 e1 e0 + e0 e1 e0 e1 ]
= [e0 + (0) − e0 (0)] + [e0 e1 + e1 − e0 e1 ] + [−e0 e1 − e0 (0) + e0 (0)e1 ]
= e0 + e1 − e0 e1 .
= e∗

Moreover, e1 e∗ = e1 (e0 + e1 − e0 e1 ) = e1 6= 0, So se can safely claim e∗ 6= 0.

Since we have proved e∗ is an idempotent of I, then A(e∗ ) belongs to M. Upon close
examination, it turns out A(e∗ ) ⊂ A(e0 ). Let x ∈ A(e∗ ), then x(e0 + e1 − e0 e1 ) = 0.
If x(e0 + e1 − e0 e1 ) = 0, then x(e0 + e1 − e0 e1 )e0 = 0. However, x(e0 + e1 − e0 e1 )e0 =
x(e0 )2 + xe1 e0 − xe0 e1 e0 = xe0 , meaning xe0 = 0, or x ∈ A(e0 ). Moreover, e1 e∗ =
/ A(e∗ ), so A(e∗ ) is a proper
e1 (e0 + e1 − e0 e1 ) = e1 e0 + (e1 )2 − e1 e0 e1 = e1 6= 0 implies e1 ∈
subset of A(e0 ). We have arrived at a contradiction to A(e0 ) being a minimal element of
M. So, A(e0 ) 6= (0) cannot happen.
Now, suppose A(e0 ) = (0). Note that for any x ∈ I, (x−xe0 )e0 = xe0 −x(e0 )2 =
xe0 − xe0 = 0, meaning x − xe0 ∈ A(e0 ) = (0). This would imply x = xe0 for all x ∈ I.
Therefore, I = Ie0 ⊂ Re0 . However, e0 ∈ I, a left ideal of R, implies Re0 ⊂ I. Therefore,
I = Re0 for some idempotent e0 .

Corollary 5.1.1. If R is semi-simple Artinian ring and A 6= (0) is an ideal of R, then

A = Re = eR, for some idempotent e in the center of R.

Proof. Suppose R is a semisimple Artinian ring and A 6= (0) is an ideal of R. Since A

is an ideal, then certainly it is a left ideal. Hence, by the previous theorem, A = Re
for some idempotent element e in R. Note that if x ∈ A, then xe = x: if x ∈ A, then
 44

x = re for some r ∈ R, meaning xe = (re)e = re2 = re = x. Now, consider the right

ideal B = {x − ex | x ∈ A}. Clearly, e(x − ex) = ex − e2 x = ex − ex = 0 for all x ∈ A,
so eB = (0). Therefore, AB = ReB = R(0) = (0). Consequently, B ⊂ A implies
B 2 ⊂ AB = (0), so B 2 = (0). Since R is semi-simple and B is nilpotent, then B = (0).
Hence, x − ex = 0 implies x = ex for all x ∈ A. Therefore, if x ∈ A, then x = ex ∈ eR,
so A ⊂ eR. Of course, since e ∈ R, ee = e ∈ Re; yet e ∈ A, A an ideal, implies eR ⊂ A.
Thus, A = eR as desired.
Note that we have ex = x = xe for all x ∈ A, meaning e is a unity for A. To
prove that e is in the center of R, let y ∈ R. Since ye ∈ A and e is a unity for A, then
surely e(ye) = ye. Similarly, ey ∈ A and e a unity for A implies (ey)e = ey. Therefore,
for all y ∈ R, ey = eye = ye, so e is in the center of R.

Corollary 5.1.2. A semi-simple Artinian ring has a two-sided unit element.

Proof. A ring R is also an ideal of R. Therefore, as we have seen from the previous
corollary, there is a multiplicative identity element in the ideal R for all of the elements
of R. In other words, R has unity.

In order to tackle the First Wedderburn-Artin theorem, we will use the notion
of a Peirce decomposition. As we have shown, semi-simple Artinian rings always contain
a two-sided unit element. Suppose we have an idempotent element e not equal to 1. Then
for any element x of a ring R, we can say x = 1x = 1x + xe − xe = xe + x(1 − e); in other
words, R = Re + R(1 − e). As it will turn out, the sum between Re and R(1 − e) is a
T
direct sum; that is, Re R(1 − e) = (0). This is a Peirce decomposition. Let us observe
the statement and the proof before I give everything away in this paragraph.

Lemma 5.1.1. An ideal of a semi-simple Artinian ring is a semi-simple Artinian ring

Proof. Let R be a semi-simple Artinian ring and A 6= (0) be an ideal of R. Of course, A

is a ring in its own right. Moreover, by Corollary 3.3.1, R being semi-simple makes A a
semi-simple ring. Now, all there is to do is show A is Artinian.
Since R is a semi-simple Artinian ring and A 6= (0) is an ideal, we know by
Corollary 5.1.1 that A = Re for an idempotent e in the center of R. Moreover, we know
R has unity 1 ∈ R. Suppose x ∈ R, then x = 1x + xe − xe = xe + (1 − e)x; therefore,
R = Re + R(1 − e). Note that 1 − e is in the center of R : r(1 − e) = r1 − re = 1r − er =
 45

(1 − e)r, ∀r ∈ R. Of course, R(1 − e) is an ideal (principal) of R. Next, we will show

Re ∩ R(1 − e) = (0). Suppose x ∈ Re ∩ R(1 − e). Note that if x ∈ Re ∩ R(1 − e), then
x = re and x = s(1 − e) for some r, s ∈ R. Therefore, xe = (re)e = re2 = re = x and
xe = [s(1 − e)]e = s(e − e2 ) = s(0) = 0, then x = 0. So, if x ∈ Re ∩ R(1 − e), then x = 0
and so Re ∩ R(1 − e) = (0). Hence, R is the direct sum of ideals A and R(1 − e). Now,
we can apply the Second Isomorphism Theorem:

R/R(1 − e) = A+R(1−e) ∼
= A
= A/(0) ∼
=A
R(1−e) A∩R(1−e)

Since A is isomorphic to R/R(1 − e), as a homomorphic image of the Artinian

ring R, A is Artinian.

To tackle the first Wedderburn-Artin theorem, we shall use a Peirce decomposi-

tion over and over until we cannot longer do so. Note that minimal ideals always exist in
Artinian rings. For a ring R, since R is an ideal of R, either R is minimal or it contains
a minimal ideal. Suppose A is a proper minimal ideal of R, then we know there exists an
idempotent element e in R so that A = Re. Hence, we can see that R = A ⊕ R(1 − e).
Again, since R(1 − e) is an ideal of R, either it is simple or there is a minimal ideal inside
of it that is also generated by an idempotent element. Suppose we say R(1 − e) = R1 and
e1 generates a minimal ideal A1 in R1 , then R(1 − e) = A1 ⊕ R1 (1 − e1 ). In short, R is
starting to decompose as R = A ⊕ A1 ⊕ . . .. The ring being Artinian will guarantee that
this decomposition does not go on forever. Lastly, it should be noted that these minimal
ideals are simple Artinian rings, hence our interest in the structure of simple Artinian
rings.

Wedderburn-Artin 1. A semi-simple Artinian ring is the direct sum of a finite number

of simple Artinian rings.

Proof. Let A 6= (0) be a minimal ideal in a semi-simple ring R. We will show A is a

simple ring. Of course, R being semi-simple and A 6= (0) imply A2 6= (0). Now, suppose
B 6= (0) is an ideal of A, then ABA is an ideal of R and ABA ⊂ B. We can see that
ABA is a subset of B since B is an ideal of A, therefore elements of B should absorb
elements of A from both the left and the right. However, since B being an ideal of A does
not guarantee B to be an ideal of R, we will have to formally show ABA is an ideal of R.
Mainly, since ABA is a subgroup of A (Both A and B are ideals of A, so ABA is definitely
 46

an ideal of A), and thus a subgroup of R, we need only show that ABA absorbs elements
P
of R both from the left and from the right. Let r ∈ R and i,j,k ai bj ak ∈ ABA, then
P P P P
r( i,j,k ai bj ak ) = i,j,k (rai )bj ak ∈ ABA and ( i,j,k ai bj ak )r = i,j,k ai bj (ak r) ∈ ABA
since both rai and ak r belong to the ideal A for all i, k.
Now, since A is also a semi-simple Artinian ring by Corollary 5.1.2, there is
1A ∈ A guaranteeing ABA 6= (0). Hence, by the minimality of A in R, ABA = A. Of
course, A = ABA ⊂ B ⊂ A implies B = A, therefore A is a simple ring.
Recall from the Peirce decomposition, R = A ⊕ T0 , where T0 is an ideal of R. Of
course, T0 an ideal of R implies T0 is also a semi-simple Artinian ring by Lemma 5.1.1.
Therefore, since T0 contains a minimal ideal by Lemma 4.2.1, repeat the process and pick
a minimal ideal A1 of R lying in T0 . As we have seen, A1 is a simple Artinian ring and by
Peirce decomposition T0 = A1 ⊕ T1 . Hence, we get R = A ⊕ T0 = A ⊕ A1 ⊕ T1 . Continuing
we get A = A0 , A1 , · · · , Ak , . . ., distinct, simple ideals of R, each of which is Artinian,
and such that the sums A0 + · · · + Ak are all direct. The sum is direct since if i 6= j and
T T
Ai Aj 6= (0), then Ai Aj is an ideal smaller than both Ai and Aj , contradicting their
minimality. Lastly, we claim that for some n ∈ N, R = A0 ⊕ · · · ⊕ An . That is, eventually
we will reach a Tn−1 that is minimal, making Tn−1 = An and finishing the process. If
not, and this process goes on forever, define R0 = A0 ⊕ A1 ⊕ · · · , R1 = A1 ⊕ A2 ⊕ · · · ,
Rm = Am ⊕ Am+1 ⊕ · · · . Then, R0 ) R1 ) · · · ) Rm ) · · · is a descending chain of
proper ideals of R that will not stop, contradicting R as Artinian. Thus, it must be that
R = A0 ⊕ · · · ⊕ An for some n in N.

Lemma 5.1.2. If R is a semi-simple Artinian ring and R = A1 ⊕ · · · ⊕ Ak where the Ai

are simple, then the Ai account for all the minimal ideals of R.

Proof. Recall that if A is both an ideal of R and simple, then A is a minimal ideal of
R. That is, if µ 6= (0) is an ideal of R and µ ⊂ A, then by simplicity, µ = A. Now, let
R be a semi-simple Artinian ring and R = A1 ⊕ · · · ⊕ Ak , where the Ai are simple. Let
B 6= (0) be a minimal ideal of R. Since R is a semi-simple Artinian ring, then we know
there exists 1 ∈ R. Hence, we know RB 6= (0). Note that RB = A1 B ⊕ · · · ⊕ Ak B 6= (0),
so it must be that for some i between 1 and k, Ai B 6= (0). Of course, both Ai , B being
ideals of R makes Ai B an ideal of R. Note that since Ai B ⊂ Ai , by the minimality of
Ai and Ai B 6= (0), then Ai B = Ai . Similarly, Ai B ⊂ B implies, by the minimality of B,
that Ai B = B. Thus, B = Ai for some i.
 47

5.2 Simple Artinian Rings

With everything we have covered up to this point, finding the structure of simple
Artinian rings will be swift. For simple Artinian rings, primitivity will make a comeback.
Consequently, we shall revisit R-modules; simple, faithful R-modules to be precise. The
one topic that will be new in this chapter is density. While there is a connection between
this use of density and density in the topological sense, density here refers to the ring of
endomorphisms over a vector space. Once we reach Jacobson’s density theorem, we will
be ready to prove the second Wedderburn-Artin theorem.
We shall now start using an alternative definition for primitive rings. Theo-
rem 5.2.1 will give us a definition that only deals in ring theory. Since we have managed
to put the Jacobson radical in terms of maximal regular ideals, this new definition con-
nects semi-simplicity with primitivity.

Theorem 5.2.1. A ring R is primitive ⇐⇒ ∃ maximal regular left ideal ρ ⊂ R such

that (ρ : R) = (0).

Proof. Suppose R is a primitive ring. Then, by definition, there exists a simple, faithful
R-module M . By Theorem 3.1.1, M ∼ = R/ρ for some maximal regular left ideal ρ. Since
A(M ) = (ρ : R) by Proposition 3.2.1 and M is faithful, then (ρ : R) = (0). Hence, there
exists a maximal regular left ideal ρ ⊂ R such that (ρ : R) = (0).
Now, suppose there exists a maximal regular left ideal ρ ⊂ R such that (ρ : R) =
(0). By Theorem 3.1.1, R/ρ is a simple R-module. Of course, since A(M ) = (ρ : R),
then R/ρ is faithful. Thus, R is primitive.

Recall that Proposition 3.2.1 implies J(R) to be the intersection of all (ρ : R)

over all maximal regular left ideals ρ. If just one (ρ : R) happened to be the trivial ideal,
R would then be semi-simple. In light of the alternative definition for primitivity, we can
see that any primitive ring R is therefore semi-simple.
Now, we take a moment to observe some facts about primitive rings.

Proposition 5.2.1. A simple ring which is semi-simple is primitive

Proof. Suppose a ring R is both simple and semi-simple. Consider the ideals (ρ : R) over
all of the maximal regular left ideals ρ of R. By the simplicity of R, either (ρ : R) = (0)
T
or (ρ : R) = R for each ρ. If (ρ : R) = R for every ρ, then ρ (ρ : R) = J(R) = R,
 48

contradicting R being semi-simple. Thus, for some maximal regular left ideal ρ0 , (ρ0 :
R) = (0). Hence, R is primitive.

Proposition 5.2.2. A simple ring R with unity is primitive

Proof. Let R be a simple ring and say 1 ∈ R. Therefore, the trivial ideal (0) is a regular.
By Lemma 3.1.1, there exists a maximal regular left ideal ρ in R that contains (0).
Therefore, R/ρ is a simple R-module. Moreover, on account of R/ρ being simple, by
definition R(R/ρ) 6= (0); thus, A(R/ρ) 6= R. Since R is simple and A(R/ρ) is an ideal,
then A(R/ρ) = (0). Hence, R/ρ is a simple, faithful R-module. Hence, R is primitive.

Proposition 5.2.3. A commutative ring R is primitive if and only if R is a field.

Proof. Suppose R is a primitive ring and M is a simple, faithful R-module. Then, A ∼

=
R/ρ for some maximal regular left ideal ρ of R. Since R is commutative, ρ is in fact a
two-sided ideal and ρ ⊂ (ρ : R) = A(M ) = (0). Moreover, since ρ = (0) is regular, there
exists e ∈ R such that r − re ∈ (0) for all r ∈ R. Meaning, r = re for all r ∈ R. Since R
is commutative, e is the multiplicative identity of R. Since R is a commutative ring with
unity and ρ = (0) is maximal, then R/(0) ∼ = R is a field.
Conversely, if R is a field, then R is simple and has unity. By the previous
proposition, R is primitive. Of course, R being a field means R is commutative.

Definition 5.2.1. Let M be a (left) vector space over a division ring ∆. A ring R is said
to be dense on M if for ever n ∈ N and v1 , · · · , vn in M which are linearly independent
over ∆ and arbitrary elements w1 , · · · , wn in M , there is an element r ∈ R such that
rvi = wi for i = 1, 2, · · · , n.

Why are we talking about rings and vector spaces? Recall that whenever we
have a simple R-module M , then by Schur’s Lemma the set of R-module endomorphisms,
denoted as EndR (M ), is a division ring. Consequently, M is thus not just an mere
EndR (M )-module, but a vector space over EndR (M ). Hence, we can see an example
where the elements of a ring R act on a vector space over a division ring. It is this
specific relation between ring, module, and division ring that Jacobson’s density theorem
explores.
Jacobson’s density theorem connects density and primitive rings. Additionally,
the second Wedderburn-Artin theorem depends heavily on Jacobson’s density theorem.
 49

As a sneak peek, know that it is possible to prove that a simple Artinian ring is semi-
simple, and thus by Proposition 5.2.1 it is primitive.

Jacobson Density Theorem. Let R be a primitive ring and let M be a simple, faithful
R-module. If ∆ = EndR (M ), then R is a dense ring of linear transformations of M over
∆.

Proof. Let R be a primitve ring and let M be a simple, faithful R-module. As we have
seen, M is a left vector space over ∆ = EndR (M ). Let V be a finite-dimensional subspace
of M over ∆. We will prove by induction on the dimension of V over ∆ that if V ⊂ M
is of finite dimension over ∆ and m ∈ M , m ∈
/ V , then there exists an r ∈ R such that
rV = (0) but rm 6= 0. Suppose that dim∆ (V ) = 0, then V = (0) meaning m 6= 0. Hence,
since M being simple implies R(M ) 6= (0), there exists r ∈ R such that rm 6= 0. Of
course, V = (0) implies rV = (0). Therefore, the statement is true for the base case
dim∆ (V ) = 0.
Now, suppose V = V0 +∆w, where dim∆ (V0 ) = dim∆ (V )−1 and where w ∈
/ V0 .
By the induction hypothesis, if A(V0 ) = {x ∈ R | xV0 = (0)}, then for y ∈ M , y ∈
/ V0 ,
there is an r ∈ A(V0 ) such that ry 6= 0; Notice this last fact shows that if y ∈
/ V0 , then
A(V0 )y 6= (0) since r ∈ A(V0 ) and ry 6= 0. In particular, since w ∈
/ V0 , then A(V0 )w 6= (0).
Recall that by the mere fact that V0 is a subset of M that A(V0 ) is a left ideal of R.
Therefore, we can show A(V0 )w is a submodule of M . Of course, for any a ∈ A(V0 ) ⊂ R,
aw ∈ M , so A(V0 )w ⊂ M . First, 0 ∈ A(V0 ) implies 0w = 0 ∈ A(V0 )w. Second, suppose
aw ∈ A(V0 )w with a ∈ A(V0 ), then −a ∈ A(V0 ) and (−a)w = −(aw) ∈ A(V0 )w. Third,
let aw, bw ∈ A(V0 )w with a, b ∈ A(V0 ), then aw + bw = (a + b)w ∈ A(V0 )w since
a + b ∈ A(V0 ). Lastly, let r ∈ R and aw ∈ A(V0 )w with a ∈ A(V0 ), then A(V0 ) being
a left ideal of R implies ra ∈ A(V0 ), meaning r(aw) = (ra)w ∈ A(V0 )w, thus A(V0 )w is
an R-module. Since we have confirmed A(V0 )w is a submodule of M , and M is simple,
A(V0 )w 6= (0), we have A(V0 )w = M .
Suppose there is an m ∈ M , m ∈
/ V so that whenever rV = (0) then rm = 0;
We will show this cannot happen. Let m ∈ M , m ∈
/ V , have the property that whenever
rV = (0) then rm = 0. Since M = A(V0 )w, consider the mapping τ : M → M by
τ (aw) = am, for all a in A(V0 ). First, we will show τ is well-defined. Say a1 w = a2 w
for some a1 , a2 ∈ A(V0 ), then (a1 − a2 )w = 0. Note that if v ∈ V with v = v0 + cw,
for some v0 ∈ V0 and c ∈ ∆, then a1 − a2 ∈ A(V0 ) ⊂ R and c ∈ EndR (M ) imply
 50

(a1 − a2 )v = (a1 − a2 )v0 + (a1 − a2 )(cu) = 0 + c[(a1 − a2 )w] = c0 = 0. In other

words, (a1 − a2 )V = (0). By m’s peculiar property, we have that (a1 − a2 )m = 0.
Therefore, a1 m = a2 m, meaning τ (a1 w) = τ (a2 w). Hence, τ is a well-defined function.
As it turns out, τ ∈ EndR (M ). Say aw, bw ∈ M , then τ (aw + bw) = τ ( (a + b)w ) =
(a+b)m = am+bm = τ (aw)+τ (bw). Moreover, let r ∈ R and aw ∈ M , then τ (r(aw)) =
τ ((ra)w) = (ra)m = r(am) = rτ (aw). Thus, τ is an R-module homomorphism from M
to M . Moreover, Since A(V0 ) ⊂ R, it must be that for any x ∈ M and any a ∈ A(V0 ),
τ (ax) = aτ (x). Therefore, since am = τ (aw) = aτ (w), then a(m − τ (w)) = 0 for
any a ∈ A(V0 ), meaning A(V0 )(m − τ (w)) = (0). Recall that one way to state our
induction hypothesis is that if y ∈
/ V0 , then A(V0 )y 6= (0). Therefore, it is also true to
say that if A(V0 )y = (0), then y ∈ V0 . The point being, if A(V0 )(m − τ (w)) = (0), then
(m − τ (w)) ∈ V0 . Hence, (m − τ (w)) + τ w = m ∈ V0 + ∆w = V , contradicting m ∈
/ V.
In conclusion, whenever there is an r ∈ R so that rV = (0), then it must be that rm 6= 0
for all m ∈ M , m ∈
/ V.
Now that we know we can find an r ∈ R for any m ∈ M , m ∈
/ V , so that
rV = (0) but rm 6= 0 for any finite dimensional subspace V of M , we can finally finish
this proof. Suppose you have arbitrary linearly independent vectors v1 , . . . , vn ∈ M over
∆ and arbitrary vectors w1 , . . . , wn in M . Let Vi = span{vj | j 6= i}, then vi ∈
/ Vi for all
i. Hence, we can find ri ∈ R so that ri vi 6= 0 but ri Vi = (0) for all i. Now, since ri vi 6= 0,
as we have seen before R(ri vi ) is a nonzero submodule of M , thus R(ri vi ) = M for all i.
In particular, since wi ∈ M , we can find si ∈ R so that si (ri vi ) = wi for all i. Let ti = si ri
for all i, then we see that both ti vi = wi and ti V = (si ri )V = si (ri V ) = si (0) = (0) for
all i. Hence, let t = t1 + · · · + tn , then tvk = (t1 + · · · + tn )vk = tk vk = wk for all i.
Therefore, R is dense on M .

Finally, we have arrived to the second Wedderburn-Artin theorem. Once we

discover the structure of simple Artinian rings, the true nature of semi-simple Artinian
rings will be revealed. Without further delay, let us dig into the last Wedderburn-Artin
theorem.

Wedderburn-Artin 2. Wedderburn: Let R be a simple Artinian ring. Then R is iso-

morphic to M atn D, for some division ring D and n ∈ N. Conversely, for any division
ring D, M atn D is a simple Artinian ring.
 51

Proof. Let R be a simple Artinian ring. Recall that R being simple implies either J(R) =
(0) or J(R) = R. Since R is Artinian, J(R) is nilpotent. However, by definition, R a
simple ring implies R2 6= (0), meaning R2 = R (R2 is an ideal of R). Therefore, R
is not a nilpotent ideal. Consequently, J(R) cannot be R (if J(R) = R, then R is
nilpotent), so J(R) = (0). Since R has been shown to be both simple and semi-simple,
by Proposition 5.2.1 R is primitive. Let M be a simple, faithful R-module, then M is a
left vector space over the division ring ∆ = EndR (M ).
We will show M is finite-dimensional over ∆ by contradiction. So, suppose that
instead {v1 , . . . , vn , . . .} is an infinite set of linearly independent vectors over ∆. Let
Bm = {v1 , . . . , vm }, then since Bm ⊂ M for all m ∈ N, A(Bm ) are all left ideals of R.
Moreover, by Jacobson’s density theorem, for any i ∈ N, A(Bi+1 ) ( A(Bi ). Of course,
if x ∈ A(Bi+1 ), then xvk = 0 for k = 1, 2, . . . , i + 1, meaning x ∈ A(Bi ). Moreover,
since vi+1 ∈
/ span(Bi ), by density there exists r ∈ R so that r(span(Bi )) = (0), but
rvi 6= 0; in other words, r ∈ A(Bi ) but r ∈
/ A(Bi+1 ). Therefore, A(B1 ) ) A(B2 ) ) · · · )
A(Bm ) ) · · · is a descending chain of proper left ideals of R that has never stops. This
is a contradiction to R being Artinian. Hence, M is finite-dimensional over ∆.
Now that we know M is finite-dimensional over ∆, say dim∆ (M ) = n, we can
prove R is isomorphic to End∆ (M ) and thus isomorphic to M atn (∆op ). Let {v1 , . . . , vn }
be a basis for M and define Φ : R → End∆ (M ) with Φ(r) = Φr and define Φr with
Φr (x) = rx for all x ∈ M . First, we will show Φ is injective. Suppose Φa = Φb , then
ax = bx for all x ∈ M . Hence, (a−b)x = 0 for all x ∈ M , meaning a−b ∈ A(M ). However,
M is a faithful R-module, so a − b ∈ (0), hence a = b. Next, since R acts densely on M ,
we can show Φ is surjective. Let f ∈ End∆ (M ); recall that a function f ∈ End∆ (M ) is
defined by the output of the basis elements under f . Since f (vi ) ∈ M for all i and R is
dense in M , there exists r ∈ R so that rvi = f (vi ) for all i. Since Φr (vi ) = f (vi ) for all
i, Φr and f agree with the outputs of the basis elements. Thus Φr = f . Hence, for any
f ∈ End∆ (M ), there exists r ∈ R so that Φ(r) = f , so Φ is surjective. Now, suppose
a, b ∈ R, then Φ(a + b) = Φa+b = (a + b)x = ax + bx = Φa + Φb = Φ(a) + Φ(b) and
Φ(ab) = Φab = (ab)x = a(bx) = Φa (bx) = Φa (Φb ) = Φa Φb (composition of functions is the
multiplication of End∆ (M )). Hence, Φ is an isomorphism between R and End∆ (M ), so
R∼
= End∆ (M ). Since M atn (∆op ) ∼
= End∆ (M ) by Theorem 2.1.1, then M atn (∆op ) ∼
= R.
Of course, since ∆ is a division ring, so is ∆op . Let D = ∆op , then R ∼
= M atn (D).
 52

Now, for the converse, we will show that for any division ring D, M atn (D) is a
simple Artinian ring. Of course, by Theorem 4.2.2 we saw that for any division ring D,
M atn (D) is an Artinian ring. Moreover, by Proposition 2.3.1, M atn (D) is a simple ring.
Hence, M atn (D) is a simple Artinian ring.
 53

Chapter 6

Implications

It is time we put our theorems to the test. Since the Wedderburn-Artin theorems
talk about semi-simple Artinian rings, we shall look at some of those. Mainly, we will
observe the impact of these theorems on some finite rings and finite-dimensional group
algebras over fields. These two types of rings exemplify the notion of Artinian rings.
First, we will take a look at Zn . Since Zn is finite for all n ∈ N, it is certainly
Artinian. That is, Zn has a finite number of elements, thus has a finite number of possible
subsets. If the number of subsets is finite, so must the number of ideals. Hence, any chain
of ideals will always stabilize. Now, while all Zn are all Artinian, not all are semi-simple.
Before we dive into theorems about Zn , let us fist observe one Zn that is semi-simple.
After we observe this example, we can then see what drives the theorems we shall explore
soon after.

Example 6.0.1. Consider the Artinian ring Z30 . Since Z30 is commutative and unitary,
all left ideals are two-sided and all ideals are regular. Hence, the Jacobson radical of
Z30 , J(Z30 ), is the intersection of all the maximal ideals of Z30 . The maximal ideals of
Z30 are h2i, h3i, and h5i, hence J(Z30 ) = h2i ∩ h3i ∩ h5i = (0), so Z30 is semi-simple.
Now, to find minimal ideals, we first find an idempotent element of Z30 and construct a
principal ideal with it. To check that is a minimal ideal, we test it using Theorem 4.3.2.
Luckily, finding an idempotent element in Z30 proves to not be as difficult as 6 fits the
bill. Now, let us observe 6Z30 = h6i = {0, 6, 12, 18, 24}. As we can see, h6i is not only
minimal, but a field! Now, as we did in First Wedderburn-Artin theorem, we have Z30 =
6Z30 ⊕ (1 − 6)Z30 . Note that 1 − 6 = 25, so we have Z30 = 6Z30 ⊕ 25Z30 . As it becomes
 54

evident, 25Z30 = h25i = {0, 5, 10, 15, 20, 25} is not a minimal ideal. Regardless, what
we must do is now find a minimal ideal of Z30 inside h25i. Note that 10 is also idempotent
in Z30 , and h10i is a minimal ideal of Z30 . Hence, if R1 = h25i, and since 25 is the unity
of R1 , then we have that R1 = 10R1 ⊕ (25 − 10)R1 . Note that if I is an ideal or a ring R
and e ∈ I is an idempotent element of R, then eI = eR: Since I ⊂ R, then eI ⊂ eR. On
the other hand, since e ∈ I, eR ⊂ I. Hence, e(eR) ⊂ e(I); of course, e(eR) = e2 R = eR,
we have eR ⊂ eI. Hence, we can simply say that h25i = h10i ⊕ h(25 − 10)i with respect
to Z30 , not R1 . Finally, (25 − 10)R1 = h15i is also minimal ideal, so we stop the Peirce
decomposition on Z30 . Therefore, we have Z30 = h6i ⊕ h10i ⊕ h15i, the direct sum of
simple (the minimal ideals are fields) Artinian rings.

So, why did Z30 work so nicely? Well, we knew Z30 was Artinian, so that was
half the battle. If we also had that it was semi-simple, our theorems should do the rest
of the work. Why was Z30 semi-simple?

Lemma 6.0.1. Let n, m ∈ N. If m divides n, then Zn /hmi ∼

= Zm

Proof. Let n, m ∈ N and say m|n, where n = mq, q ∈ N. Consider the map φ : Zn → Zm
with φ(a) = a. First, we will show φ is well-defined. Suppose r, s ∈ Zn and r = s.
If so, then r ≡ s mod(n), meaning r − s = nk, where k ∈ Z. Since n = mq, then
r − s = (mq)k = m(qk). Thus, r ≡ s mod(m), so r = s in Zm . Now, suppose a, b ∈ Zn .
Recall that in both Zn and Zm , a + b = a + b and ab = ab. Thus, φ(a + b) = φ(a + b) =
a + b = a + b = φ(a) + φ(b) and φ(ab) = φ(ab) = ab = ab = φ(a)φ(b). Moreover, φ is
surjective. Let y ∈ Zm , then consider y ∈ Zn and note that φ(y) = y. Hence, φ is a ring
homomorphism from Zn to Zm with φ(Zn ) = Zm . Lastly, Kerφ = {a ∈ Zn | φ(a) = 0} =
{a ∈ Zn | m|a} = hmi. Therefore, by the First Isomorphism Theorem, Zn /hmi ∼
= Zm .

Theorem 6.0.1. If n = pk11 · · · pkr r , pi ’s all distinct primes, then J(Zn ) = hp1 · · · pr i.

Proof. Let n = pk11 · · · pkr r , where the pi ’s are all distinct primes. Of course, Zn is a unitary
commutative ring, so J(Zn ) is the intersection of all maximal ideal of Zn by Theorem 3.3.1.
Of course, with Lemma 6.0.1, we can quickly prove that hpi i is maximal in Zn for all i:
Zn /hpi i ∼
= Zpi , so Zn /hpi i being a field implies hpi i is a maximal ideal. Thus, J(Zn ) =
hp1 i ∩ · · · ∩ hpr i. Now, if x ∈ hp1 · · · pr i, then x = (p1 · · · pr )q, where q ∈ Zn . Therefore,
x ∈ hpi i for every i, thus x ∈ hp1 i∩· · ·∩hpr i. Likewise, if x ∈ hp1 i∩· · ·∩hpr i, then x ∈ hpi i
 55

for all i. If so, then pi divides x for all i. Hence, p1 · · · pr divides x; say, x = (p1 · · · pr )t,
for t ∈ Z. Hence, x = (p1 · · · pr )t ∈ hp1 · · · pr i, so hp1 · · · pr i = hp1 i ∩ · · · ∩ hpr i = J(Zn ),
as desired.

Corollary 6.0.1. if n = p1 · · · pr , pi all distinct, then Zn is semi-simple.

Proof. Since p1 · · · pr = 0 in Zp1 ···pr , then J(Zp1 ···pr ) = hp1 · · · pr i = (0).

Hence, we have pinpointed which Zn are semi-simple. Now that we know which
Zn to pick, let us step back and analyze our example on Z30 to generalize on how to
decompose a semi-simple Artinian Zn . First, we know from Lemma 5.1.2 that a semi-
simple Zn will split into direct sums of all its minimal ideals. Particularly, if n = p1 · · · pr ,
then the principal ideals h( pni )i will be all the minimal ideals of Zn (they are all of prime
order). With such knowledge beforehand, the Peirce decomposition is no longer necessary
for the structure of Zn .
Let us move on to group algebras. While Zn shows how to one might approach
finite semi-simple rings in general, group algebras will give us an insight on how to deal
with vector spaces over fields. This time around, looking for the order of ideals will be
rendered useless as a tactic to spot minimal ideals. Instead, ideas from linear algebra
will prove to have a better effect. The reason why we shall explore group algebras as
opposed to random vector spaces is the result of Maschke’s Theorem, which guarantees
semi-simplicity. Of course, let us define what a group algebra is and prove a lemma before
jumping right into the theorem.

Definition 6.0.1. Let F be a field and let G be a finite group of order n. We define
P
the group algebra of G over F , denoted by F (G), as the set { i αi gi | αi ∈ F, gi ∈ G} of
formal sums, where the group elements are considered to be linearly independent over F ,
addition is defined naturally, and multiplication is defined by the distributive laws and
with (αi gi )(βj gj ) = αi βj gi gj .

Lemma 6.0.2. Let N be a nilpotent linear transformation on a finite-dimensional vector

space V over a division ring D. Then, there exists basis W = {w1 , . . . , wn } of V such
that N (w1 ) = 0 and N (wi ) ∈ Span{w1 , . . . , wi−1 } for 2 ≤ i ≤ n.

Proof. Let N ∈ EndD (V ) be (non-zero) nilpotent. We will prove the lemma by induction
on the dimension of V over D.
 56

Say dimD (V ) = 1. Since N is nilpotent in EndD (V ), there exists k ∈ N so

that N k = 0 in EndD (V ). Let M = {s ∈ N | N s = 0}; since k ∈ M , as a non-empty
subset of the natural numbers, M must have a minimal element m. Note that m ≥ 2
(m 6= 1), or else it contradicts N as a non-zero map. Therefore, we can say N m−1 6= 0
in EndD (V ). Otherwise, m − 1 ∈ M , contradicting m as the smallest element in M . If
N m−1 6= 0 in EndD (V ), then there must be some w ∈ V so that N m−1 (w) 6= 0. Hence, let
w1 = N m−1 (w). Since w1 6= 0, W = {w1 } is a basis for V and N (w1 ) = N (N m−1 (w)) =
N m (w) = 0, as desired.
Now, assuming that we can find such a basis for any vector space with dimension
n−1 or less, we will show we can do so for a vector space of dimension n. Let V be a vector
space of dimension n and basis {v1 , . . . , vn }, and let N be a nilpotent linear transforma-
tion on V . Consider V 0 = Span{v1 , . . . , vn−1 }. By the induction hypothesis, V 0 has a
basis W = {w1 , . . . , wn−1 } of V 0 such that N (w1 ) = 0 and N (wi ) ∈ Span{w1 , . . . , wi−1 }
for 2 ≤ i ≤ n − 1. Now, let w ∈
/ Span(W ), w ∈ V . If N (w) ∈ Span(W ), then wn = w,
W = W ∪ {w} is the desired basis for V ; we are finished. So, suppose N (w) ∈
/ Span(W ).
Then, there exists l ∈ N so that N l (w) ∈
/ Span(W ) but N l+1 (w) ∈ Span(W ). Otherwise,
suppose that for all l ∈ N so that N l (w) ∈
/ Span(W ) then N l+1 (w) ∈
/ Span(W ), since
N (w) = N 1 (w) ∈
/ Span(W ), then really N d (w) ∈
/ Span(W ) for any natural number d.
However, since N is nilpotent, N k = 0 for some k ∈ N, surely N k (w) = 0 ∈ Span(W ). So,
since k ∈ N, N k (w) ∈
/ Span(W ), yet N k (w) ∈ Span(W ) since N k (w) = 0, contradiction.
Therefore, for some l ∈ N we have N l (w) ∈
/ Span(W ) but N l+1 (w) ∈ Span(W ). Hence,
let wn = N l (w), then W = W ∪ {N l (w)} is the basis for V that we are looking for.

Maschke’s Theorem. If G is a finite group of order n and F is a field of characteristic

0 or p, where p 6 |n, then F (G) is semi-simple.

Proof. Let G be a finite group of order n and let F be a field of characteristic 0 or p,

where p 6 |n. For a ∈ F (G) we define the mapping Ta : F (G) → F (G) by Ta (x) = ax
for all x ∈ F (G). This is called the left regular representation of F (G). As we can
see, for any a ∈ F (G), Ta is an F -module homomorphism from F (G) to itself and
thus an F -linear transformation on F (G). Surely, for any x, y ∈ F (G) and a ∈ F (G),
Ta (x + y) = a(x + y) = ax + ay = Ta (x) + Ta (y). Moreover, for any c ∈ F , a ∈ F (G) and
x ∈ F (G), Ta (cx) = a(cx) = a(c)x = (c)ax = c(ax) = cTa (x). Now, since we are seeing
F (G) as a vector space over F , we can say that, for any a ∈ F (G), Ta ∈ EndF (F (G)). Of
 57

course, by Theorem 2.1.1, EndF (F (G)) ∼

= M atn (F ) (since F is commutative, opposite
ring is not needed). Define, for any a ∈ F (G), Ma = [Ta (g1 ), . . . , Ta (gn )]; the matrix
composed of the coefficients of Ta (gi ) as the i-th column vector (just as in Theorem 2.1.1).
We will note two things about Mg , for g ∈ G. If e is the identity element of G,
then the trace of Me , tr(Me ), is n, and since eg = g = ge for any g ∈ G, then Me is the
identity matrix. On the other hand, if g 6= e, then tr(Mg ) = 0 since gh 6= h 6= hg for any
h ∈ G.
Let J be the Jacobson radical of F (G). Since F (G) is finite dimensional over F ,
thus Artinian, then J is nilpotent by Theorem 4.2.3. To prove J = (0), suppose instead
that J 6= (0). Let v ∈ J be a nonzero element. Say, v = a1 g1 + · · · + an gn , with at
least ak 6= 0. Since J is an ideal, then gk−1 v ∈ J and gk−1 v = ak e + a02 g20 + · · · + a0n gn0 ;
say gk−1 v = v 0 . Since v 0 ∈ J and J is nilpotent, then v 0 is nilpotent. However, since
v 0 is nilpotent, then Tv0 is a nilpotent linear transformation. Since Tv0 is a nilpotent
linear transformation, by Lemma 6.0.2 we have a basis W of F (G) so that Tv0 (w1 ) = 0
and Tv0 (wi ) ∈ Span{w1 , . . . , wi−1 } for 2 ≤ i ≤ n. Note that [Mv0 ]W is a matrix so
that the entries aij equal zero if i ≥ j, meaning tr([Mv0 ]W ) = 0. On that note, let P
be the change of basis matrix from {g1 , . . . , gn } to W, then [Mv0 ]W = P Mv0 P −1 and
tr([Mv0 ]W ) = tr((P Mv0 )P −1 ) = tr(P −1 (P Mv0 )) = tr(Mv0 ). Hence, tr(Mv0 ) = 0. On the
other hand, Tv0 = v 0 x = (ak e + a02 g20 + · · · + a0n gn0 )x = ak (ex) + a02 (g20 x) + · · · + a0n (gn0 x) =
ak Te + a02 Tg20 + · · · + a0n Tgn0 , so tr(Mv0 ) = ak tr(Me ) + a02 tr(Mg20 ) + · · · + a0n tr(Mgn0 ) = ak n,
so ak n = 0. Since ak 6= 0, then n = 0 in F . If F has characteristic 0 and n(ak ) = 0,
ak 6= 0, then n1F = nak (ak )−1 = 0, contradicting F as having characteristic 0. On the
other hand, if the characteristic of F is p, it must be that p|n, which contradicts the given
hypothesis. Hence, it must be that J = (0), so F (G) is semi-simple.

Let us explore an example of a group algebra decomposed as a direct sum of

simple rings. Note that we shall eventually use a strategy implemented in Jacobson, pg.
49 [J], for picking a particular basis.

Example 6.0.2. Let hαi be the cyclic group of order 3 generated by α. Consider the
group algebra R(hαi) = {c1 e + c2 α + c3 α2 | ci ∈ R}. By Maschke’s Theorem, we know
e+α+α2
R(hαi) is a semi-simple (and Artinian) ring. Note that the element 3 is a nonzero
e+α+α2
idempotent element of R(hαi). In fact, if we say ζ = 3 , then ζ(c1 e + c2 α + c3 ) =
c1 ζe + c2 ζα + c3 ζα2 = c1 ζ + c2 ζ + c3 ζ = (c1 + c2 + c3 )ζ. In other words, the product of any
 58

element of R(hαi) and ζ results in a scalar multiple of ζ. On that note, we can observe
how [R(hαi)]ζ is a field with unity ζ (ζ ∈ R(hαi) implies ζζ = ζ 2 = ζ ∈ [R(hαi)]ζ). By
Theorem 4.3.2, we have that [R(hαi)]ζ is a minimal ideal. Hence, R(hαi) = [R(hαi)]ζ ⊕
[R(hαi)](e − ζ). Sadly, proving [R(hαi)](e − ζ) is a minimal ideal (and a field) is no easy
task, so we will use a different strategy. Recall, R(hαi) is still a vector space over R, so
we shall pick basis B = {ζ, α − e, α2 − α} ; let u1 = ζ, u2 = α − e, and u3 = α2 − α.
Since B is a basis for R(hαi), we can say R(hαi) = SpanR {u1 } ⊕ SpanR {u2 , u3 }. From
our previous conversation on [R(hαi)]ζ, we can see that [R(hαi)]ζ = SpanR {u1 }. Hence,
it must be that SpanR {u2 , u3 } = [R(hαi)](e − ζ). Correspondingly, we will show that
SpanR {u2 , u3 } is a minimal ideal by showing that it does not contain a smaller ideal. Of
course, we will do so by contradiction. Suppose there exists an ideal I of R(hαi) such that
(0) ⊂ I ⊂ SpanR {u2 , u3 } and SpanR {u2 , u3 } 6= I. We can quickly show I is a vector
subspace of R(hαi). By definition, I being an ideal implies it is closed under addition
and contains zero (the zero vector). Now, to show I is closed under scalar multiplication,
let c ∈ R and v ∈ I, then cv = cev = (ce)v. Since ce ∈ R(hαi), v ∈ I, and I an ideal of
R(hαi), we have that (ce)v ∈ I; hence, cv ∈ I. Now, since I is a subspace of R(hαi) and
a proper subset of SpanR {u2 , u3 }, I must be either (0) or a 1-dimensional subspace of
SpanR {u2 , u3 }. If I is a 1-dimensional subspace of SpanR {u2 , u3 }, say the basis of I is
{v} (v 6= 0), it must be that for any x ∈ R(hαi), xv = λv for some λ ∈ R. Of course, if
x ∈ R(hαi), then x is a formal sum of powers of α. Hence, we need only check if there
exists a scalar λ ∈ R such that αv = λv, for some λ ∈ R. Since u2 , u3 form a basis for
SpanR {u2 , u3 }, αu2 = u3 and αu3 = −u2 − u3 , we can say the linear transformation on
SpanR {u2 , u3 } induced by α is
 
0 −1
Tα =  
1 −1

Hence, we must see if there exists scalar λ ∈ R such that Tα v = λv. Upon
a simple calculation of eigenvaules, we see that the characteristic polynomial of Tα is
λ2 + λ + 1, which is irreducible over R. Since no such λ exists in R, I cannot be 1-
dimensional. Hence, I = (0), which means SpanR {u2 , u3 } is a minimal ideal of R(hαi).
Therefore, R(hαi) = [R(hαi)]ζ ⊕ [R(hαi)](e − ζ) is the full Peirce decomposition of R(hαi).
Lastly, to show what these two simple rings are isomorphic to, consider the ring
of endomorphisms for both of the rings. Say, R1 = [R(hαi)]ζ and R2 = [R(hαi)](e − ζ).
 59

Recall that R1 , R2 being fields imply that they are primitive rings by Proposition 5.2.3.
As such, they must have simple, faithful modules which can be taken to be R1 and
R2 themselves respectively. On that note, D1 = EndR1 (R1 ) and D2 = EndR2 (R2 )
are division rings by Schur’s Lemma. Lastly, Wedderburn’s second theorem shows that
R1 ∼= EndD1 (R1 ) and R2 ∼= EndD2 (R2 ). Now, we need only consider what EndD1 (R1 )
2 = R1 ∼
and EndD (R2 ) are. For R1 , it is clear to see that R ∼ = EndD (R1 ). As for
1

EndD2 (R2 ), since we need D2 -module homomorphisms on R2 we need only check all of
the linear transformation of R2 that commute with α, or rather Tα . Hence consider the
2 × 2 matrices such that

     
a b 0 −1 0 −1 a b
  =  
c d 1 −1 1 −1 c d

Then we have that b = −c and d = a − c; a and c remain free. In fact, if E is the 2 × 2

identity matrix, then any matrix that commutes Tα is actually a linear combination of
E and Tα . Furthermore, if I = √1 E + √2 Tα , then I 2 = −E. Since we know R2 ∼
3 3
=
EndD2 (R2 ), and R2 a 2-dimensional field over R, and there is an element I such that
I 2 = −E, we can see that it is not hard to prove EndD (R2 ) ∼
2 = C, the complex numbers!
Hence, we have that R1 and R2 are 1 × 1 matrices over familiar fields. That is, R(hαi) =
M at1 (R) × M at1 (C).

In both examples, we had our rings be a direct sum of fields. Such is the
consequence of dealing with commutative rings. In general, if we assume our ring to be
commutative on top of being semi-simple Artinian, then the ring will decompose into a
direct sum of fields. That is, if a ring R is semi-simple Artinian, then the first Wedderburn-
Artin theorem states that it will be a finite direct sum of simple Artinian rings. Then,
Lemma 5.1.2 states that these simple rings are all of the minimal (two-sided) ideals of R.
Next, we observe Corollary 5.1.1 which states that every minimal ideal of R is of the form
Rei for some idempotent elements ei or R. Lastly, we note Theorem 4.3.2 and recall that
in the commutative case, both Re = eRe and eRe would be a commutative division ring.
Hence, the minimal ideals Rei would all be fields, meaning R is a direct sum of fields.
Our group algebra example also came out this way since F (G) is commutative if and
only if both F and G are commutative. Hence, one could try an example with a group G
 60

that is not commutative and have simple modules that are not fields. Sadly, this is the
end of the road. We have proved the Wedderburn-Artin theorems and have given quick
examples of the theorems in action. While we could explore the results of the theorems
in more dept, that sounds like a paper for another day. Quite honestly, this paper cold
have ended with chapter 5, but I figured a quick display of our theorems would be of
great use. Now, taking the examples we have covered thus far, the reader can explore
semi-simple rings on their own if they wish to. Good luck!
 61

Bibliography

[C] Curtis, C. W., Reiner, I. Representation Theory of Finite Groups and Associative
Algebras. Wiley, 1988

[Her] Herstein, I. N. Noncommutative Rings, Carus Mathematical Monographs, Vol. 15,

Mathematical Association of America, Washington, D.C., 1968.

[Hun] Hungerford, T., Algebra. Graduate Texts in Mathematics, Vol. 73, Springer-
Verlag, New York, 1974.

[J] Jacobson, N., Basic Algebra II, W. H. Freeman and Company, San Francisco,
1980.