# Elements of

Abstract and Linear Algebra
E. H. Connell
ii
E.H. Connell
Department of Mathematics
University of Miami
P.O. Box 249085
Coral Gables, Florida 33124 USA
ec@math.miami.edu
Mathematical Subject Classiﬁcations (1991): 12-01, 13-01, 15-01, 16-01, 20-01
c (1999 E.H. Connell
March 20, 2004
iii
Introduction
In 1965 I ﬁrst taught an undergraduate course in abstract algebra. It was fun to
teach because the material was interesting and the class was outstanding. Five of
those students later earned a Ph.D. in mathematics. Since then I have taught the
course about a dozen times from various texts. Over the years I developed a set of
lecture notes and in 1985 I had them typed so they could be used as a text. They
now appear (in modiﬁed form) as the ﬁrst ﬁve chapters of this book. Here were some
of my motives at the time.
1) To have something as short and inexpensive as possible. In my experience,
students like short books.
2) To avoid all innovation. To organize the material in the most simple-minded
straightforward manner.
3) To order the material linearly. To the extent possible, each section should use
the previous sections and be used in the following sections.
4) To omit as many topics as possible. This is a foundational course, not a topics
course. If a topic is not used later, it should not be included. There are three
good reasons for this. First, linear algebra has top priority. It is better to go
forward and do more linear algebra than to stop and do more group and ring
theory. Second, it is more important that students learn to organize and write
proofs themselves than to cover more subject matter. Algebra is a perfect place
to get started because there are many “easy” theorems to prove. There are
many routine theorems stated here without proofs, and they may be considered
as exercises for the students. Third, the material should be so fundamental
that it be appropriate for students in the physical sciences and in computer
science. Zillions of students take calculus and cookbook linear algebra, but few
take abstract algebra courses. Something is wrong here, and one thing wrong
is that the courses try to do too much group and ring theory and not enough
matrix theory and linear algebra.
5) To oﬀer an alternative for computer science majors to the standard discrete
mathematics courses. Most of the material in the ﬁrst four chapters of this text
is covered in various discrete mathematics courses. Computer science majors
might beneﬁt by seeing this material organized from a purely mathematical
viewpoint.
iv
Over the years I used the ﬁve chapters that were typed as a base for my algebra
courses, supplementing them as I saw ﬁt. In 1996 I wrote a sixth chapter, giving
enough material for a full ﬁrst year graduate course. This chapter was written in the
same “style” as the previous chapters, i.e., everything was right down to the nub. It
hung together pretty well except for the last two sections on determinants and dual
spaces. These were independent topics stuck on at the end. In the academic year
1997-98 I revised all six chapters and had them typed in LaTeX. This is the personal
background of how this book came about.
It is diﬃcult to do anything in life without help from friends, and many of my
friends have contributed to this text. My sincere gratitude goes especially to Marilyn
Gonzalez, Lourdes Robles, Marta Alpar, John Zweibel, Dmitry Gokhman, Brian
Coomes, Huseyin Kocak, and Shulim Kaliman. To these and all who contributed,
this book is fondly dedicated.
This book is a survey of abstract algebra with emphasis on linear algebra. It is
intended for students in mathematics, computer science, and the physical sciences.
The ﬁrst three or four chapters can stand alone as a one semester course in abstract
algebra. However they are structured to provide the background for the chapter on
linear algebra. Chapter 2 is the most diﬃcult part of the book because groups are
written in additive and multiplicative notation, and the concept of coset is confusing
at ﬁrst. After Chapter 2 the book gets easier as you go along. Indeed, after the
ﬁrst four chapters, the linear algebra follows easily. Finishing the chapter on linear
algebra gives a basic one year undergraduate course in abstract algebra. Chapter 6
continues the material to complete a ﬁrst year graduate course. Classes with little
background can do the ﬁrst three chapters in the ﬁrst semester, and chapters 4 and 5
in the second semester. More advanced classes can do four chapters the ﬁrst semester
and chapters 5 and 6 the second semester. As bare as the ﬁrst four chapters are, you
still have to truck right along to ﬁnish them in one semester.
The presentation is compact and tightly organized, but still somewhat informal.
The proofs of many of the elementary theorems are omitted. These proofs are to
be provided by the professor in class or assigned as homework exercises. There is a
non-trivial theorem stated without proof in Chapter 4, namely the determinant of the
product is the product of the determinants. For the proper ﬂow of the course, this
theorem should be assumed there without proof. The proof is contained in Chapter 6.
The Jordan form should not be considered part of Chapter 5. It is stated there only
as a reference for undergraduate courses. Finally, Chapter 6 is not written primarily
v
This text is written with the conviction that it is more eﬀective to teach abstract
and linear algebra as one coherent discipline rather than as two separate ones. Teach-
ing abstract algebra and linear algebra as distinct courses results in a loss of synergy
and a loss of momentum. Also with this text the professor does not extract the course
from the text, but rather builds the course upon it. I am convinced it is easier to
build a course from a base than to extract it from a big book. Because after you
extract it, you still have to build it. The bare bones nature of this book adds to its
ﬂexibility, because you can build whatever course you want around it. Basic algebra
is a subject of incredible elegance and utility, but it requires a lot of organization.
This book is my attempt at that organization. Every eﬀort has been extended to
make the subject move rapidly and to make the ﬂow from one topic to the next as
seamless as possible. The student has limited time during the semester for serious
study, and this time should be allocated with care. The professor picks which topics
to assign for serious study and which ones to “wave arms at”. The goal is to stay
focused and go forward, because mathematics is learned in hindsight. I would have
made the book shorter, but I did not have any more time.
When using this text, the student already has the outline of the next lecture, and
each assignment should include the study of the next few pages. Study forward, not
just back. A few minutes of preparation does wonders to leverage classroom learning,
and this book is intended to be used in that manner. The purpose of class is to
learn, not to do transcription work. When students come to class cold and spend
the period taking notes, they participate little and learn little. This leads to a dead
class and also to the bad psychology of “OK, I am here, so teach me the subject.”
Mathematics is not taught, it is learned, and many students never learn how to learn.
Professors should give more direction in that regard.
Unfortunately mathematics is a diﬃcult and heavy subject. The style and
approach of this book is to make it a little lighter. This book works best when
viewed lightly and read as a story. I hope the students and professors who try it,
enjoy it.
E. H. Connell
Department of Mathematics
University of Miami
Coral Gables, FL 33124
ec@math.miami.edu
vi
Outline
Chapter 1 Background and Fundamentals of Mathematics
Sets, Cartesian products 1
Relations, partial orderings, Hausdorﬀ maximality principle, 3
equivalence relations
Functions, bijections, strips, solutions of equations, 5
right and left inverses, projections
Notation for the logic of mathematics 13
Integers, subgroups, unique factorization 14
Chapter 2 Groups
Groups, scalar multiplication for additive groups 19
Subgroups, order, cosets 21
Normal subgroups, quotient groups, the integers mod n 25
Homomorphisms 27
Permutations, the symmetric groups 31
Product of groups 34
Chapter 3 Rings
Rings 37
Units, domains, ﬁelds 38
The integers mod n 40
Ideals and quotient rings 41
Homomorphisms 42
Polynomial rings 45
Product of rings 49
The Chinese remainder theorem 50
Characteristic 50
Boolean rings 51
Chapter 4 Matrices and Matrix Rings
Addition and multiplication of matrices, invertible matrices 53
Transpose 56
Triangular, diagonal, and scalar matrices 56
Elementary operations and elementary matrices 57
Systems of equations 59
vii
Similarity, trace, and characteristic polynomial 64
Chapter 5 Linear Algebra
Modules, submodules 68
Homomorphisms 69
Homomorphisms on R
n
71
Cosets and quotient modules 74
Products and coproducts 75
Summands 77
Independence, generating sets, and free basis 78
Characterization of free modules 79
Uniqueness of dimension 82
Change of basis 83
Vector spaces, square matrices over ﬁelds, rank of a matrix 85
Geometric interpretation of determinant 90
Linear functions approximate diﬀerentiable functions locally 91
The transpose principle 92
Nilpotent homomorphisms 93
Eigenvalues, characteristic roots 95
Jordan canonical form 96
Inner product spaces, Gram-Schmidt orthonormalization 98
Orthogonal matrices, the orthogonal group 102
Diagonalization of symmetric matrices 103
Chapter 6 Appendix
The Chinese remainder theorem 108
Prime and maximal ideals and UFD
s
109
Splitting short exact sequences 114
Euclidean domains 116
Jordan blocks 122
Jordan canonical form 123
Determinants 128
Dual spaces 130
viii
1 2 3 4
5 6 7 8
9 11 10
Abstract algebra is not only a major subject of science, but it is also
magic and fun. Abstract algebra is not all work and no play, and it is
certainly not a dull boy. See, for example, the neat card trick on page
18. This trick is based, not on sleight of hand, but rather on a theorem
in abstract algebra. Anyone can do it, but to understand it you need
some group theory. And before beginning the course, you might ﬁrst try
your skills on the famous (some would say infamous) tile puzzle. In this
puzzle, a frame has 12 spaces, the ﬁrst 11 with numbered tiles and the
last vacant. The last two tiles are out of order. Is it possible to slide the
tiles around to get them all in order, and end again with the last space
vacant? After giving up on this, you can study permutation groups and
Chapter 1
Background and Fundamentals of
Mathematics
This chapter is fundamental, not just for algebra, but for all ﬁelds related to mathe-
matics. The basic concepts are products of sets, partial orderings, equivalence rela-
tions, functions, and the integers. An equivalence relation on a set A is shown to be
simply a partition of A into disjoint subsets. There is an emphasis on the concept
of function, and the properties of surjective, injective, and bijective. The notion of a
solution of an equation is central in mathematics, and most properties of functions
can be stated in terms of solutions of equations. In elementary courses the section
on the Hausdorﬀ Maximality Principle should be ignored. The ﬁnal section gives a
proof of the unique factorization theorem for the integers.
Notation Mathematics has its own universally accepted shorthand. The symbol
∃ means “there exists” and ∃! means “there exists a unique”. The symbol ∀ means
“for each” and ⇒ means “implies”. Some sets (or collections) are so basic they have
their own proprietary symbols. Five of these are listed below.
N = Z
+
= the set of positive integers = ¦1, 2, 3, ...¦
Z = the ring of integers = ¦..., −2, −1, 0, 1, 2, ...¦
Q = the ﬁeld of rational numbers = ¦a/b : a, b ∈ Z, b = 0¦
R = the ﬁeld of real numbers
C = the ﬁeld of complex numbers = ¦a +bi : a, b ∈ R¦ (i
2
= −1)
Sets Suppose A, B, C,... are sets. We use the standard notation for intersection
and union.
A ∩ B = ¦x : x ∈ A and x ∈ B¦ = the set of all x which are elements
1
2 Background Chapter 1
of A and B.
A ∪ B = ¦x : x ∈ A or x ∈ B¦ = the set of all x which are elements of
A or B.
Any set called an index set is assumed to be non-void. Suppose T is an index set and
for each t ∈ T, A
t
is a set.
¸
t∈T
A
t
= ¦x : ∃ t ∈ T with x ∈ A
t
¦
¸
t∈T
A
t
= ¦x : if t ∈ T, x ∈ A
t
¦ = ¦x : ∀t ∈ T, x ∈ A
t
¦
Let ∅ be the null set. If A∩ B = ∅, then A and B are said to be disjoint.
Deﬁnition Suppose each of A and B is a set. The statement that A is a subset
of B (A ⊂ B) means that if a is an element of A, then a is an element of B. That
is, a ∈ A ⇒a ∈ B. If A ⊂ B we may say A is contained in B, or B contains A.
Exercise Suppose each of A and B is a set. The statement that A is not a subset
of B means .
Theorem (De Morgan’s laws) Suppose S is a set. If C ⊂ S (i.e., if C is a subset
of S), let C

, the complement of C in S, be deﬁned by C

= S−C = ¦x ∈ S : x ∈ C¦.
Then for any A, B ⊂ S,
(A ∩ B)

= A

∪ B

and
(A ∪ B)

= A

∩ B

Cartesian Products If X and Y are sets, X Y = ¦(x, y) : x ∈ X and y ∈ Y ¦.
In other words, the Cartesian product of X and Y is deﬁned to be the set of all
ordered pairs whose ﬁrst term is in X and whose second term is in Y .
Example RR = R
2
= the plane.
Chapter 1 Background 3
Deﬁnition If each of X
1
, ..., X
n
is a set, X
1
X
n
= ¦(x
1
, ..., x
n
) : x
i
∈ X
i
for 1 ≤ i ≤ n¦ = the set of all ordered n-tuples whose i-th term is in X
i
.
Example R R = R
n
= real n-space.
Question Is (RR
2
) = (R
2
R) = R
3
?
Relations
If A is a non-void set, a non-void subset R ⊂ A A is called a relation on A. If
(a, b) ∈ R we say that a is related to b, and we write this fact by the expression a ∼ b.
Here are several properties which a relation may possess.
1) If a ∈ A, then a ∼ a. (reﬂexive)
2) If a ∼ b, then b ∼ a. (symmetric)
2

) If a ∼ b and b ∼ a, then a = b. (anti-symmetric)
3) If a ∼ b and b ∼ c, then a ∼ c. (transitive)
Deﬁnition A relation which satisﬁes 1), 2

), and 3) is called a partial ordering.
In this case we write a ∼ b as a ≤ b. Then
1) If a ∈ A, then a ≤ a.
2

) If a ≤ b and b ≤ a, then a = b.
3) If a ≤ b and b ≤ c, then a ≤ c.
Deﬁnition A linear ordering is a partial ordering with the additional property
that, if a, b ∈ A, then a ≤ b or b ≤ a.
Example A = R with the ordinary ordering, is a linear ordering.
Example A = all subsets of R
2
, with a ≤ b deﬁned by a ⊂ b, is a partial ordering.
Hausdorﬀ Maximality Principle (HMP) Suppose S is a non-void subset of A
and ∼ is a relation on A. This deﬁnes a relation on S. If the relation satisﬁes any
of the properties 1), 2), 2

), or 3) on A, the relation also satisﬁes these properties
when restricted to S. In particular, a partial ordering on A deﬁnes a partial ordering
4 Background Chapter 1
on S. However the ordering may be linear on S but not linear on A. The HMP is
that any linearly ordered subset of a partially ordered set is contained in a maximal
linearly ordered subset.
Exercise Deﬁne a relation on A = R
2
by (a, b) ∼ (c, d) provided a ≤ c and
b ≤ d. Show this is a partial ordering which is linear on S = ¦(a, a) : a < 0¦. Find
at least two maximal linearly ordered subsets of R
2
which contain S.
One of the most useful applications of the HMP is to obtain maximal monotonic
collections of subsets.
Deﬁnition A collection of sets is said to be monotonic if, given any two sets of
the collection, one is contained in the other.
Corollary to HMP Suppose X is a non-void set and A is some non-void
collection of subsets of X, and S is a subcollection of A which is monotonic. Then ∃
a maximal monotonic subcollection of A which contains S.
Proof Deﬁne a partial ordering on A by V ≤ W iﬀ V ⊂ W, and apply HMP.
The HMP is used twice in this book. First, to show that inﬁnitely generated
vector spaces have free bases, and second, in the Appendix, to show that rings have
maximal ideals (see pages 87 and 109). In each of these applications, the maximal
monotonic subcollection will have a maximal element. In elementary courses, these
results may be assumed, and thus the HMP may be ignored.
Equivalence Relations A relation satisfying properties 1), 2), and 3) is called
an equivalence relation.
Exercise Deﬁne a relation on A = Z by n ∼ m iﬀ n − m is a multiple of 3.
Show this is an equivalence relation.
Deﬁnition If ∼ is an equivalence relation on A and a ∈ A, we deﬁne the equiva-
lence class containing a by cl(a) = ¦x ∈ A : a ∼ x¦.
Chapter 1 Background 5
Theorem
1) If b ∈ cl(a) then cl(b) = cl(a). Thus we may speak of a subset of A
being an equivalence class with no mention of any element contained
in it.
2) If each of U, V ⊂ A is an equivalence class and U ∩ V = ∅, then
U = V .
3) Each element of A is an element of one and only one equivalence class.
Deﬁnition A partition of A is a collection of disjoint non-void subsets whose union
is A. In other words, a collection of non-void subsets of A is a partition of A provided
any a ∈ A is an element of one and only one subset of the collection. Note that if A
has an equivalence relation, the equivalence classes form a partition of A.
Theorem Suppose A is a non-void set with a partition. Deﬁne a relation on A by
a ∼ b iﬀ a and b belong to the same subset of the partition. Then ∼ is an equivalence
relation, and the equivalence classes are just the subsets of the partition.
Summary There are two ways of viewing an equivalence relation — one is as a
relation on A satisfying 1), 2), and 3), and the other is as a partition of A into
disjoint subsets.
Exercise Deﬁne an equivalence relation on Z by n ∼ m iﬀ n −m is a multiple
of 3. What are the equivalence classes?
Exercise Is there a relation on R satisfying 1), 2), 2

) and 3) ? That is, is there
an equivalence relation on R which is also a partial ordering?
Exercise Let H ⊂ R
2
be the line H = ¦(a, 2a) : a ∈ R¦. Consider the collection
of all translates of H, i.e., all lines in the plane with slope 2. Find the equivalence
relation on R
2
deﬁned by this partition of R
2
.
Functions
Just as there are two ways of viewing an equivalence relation, there are two ways
of deﬁning a function. One is the “intuitive” deﬁnition, and the other is the “graph”
or “ordered pairs” deﬁnition. In either case, domain and range are inherent parts of
the deﬁnition. We use the “intuitive” deﬁnition because everyone thinks that way.
6 Background Chapter 1
Deﬁnition If X and Y are (non-void) sets, a function or mapping or map with
domain X and range Y , is an ordered triple (X, Y, f) where f assigns to each x ∈ X
a well deﬁned element f(x) ∈ Y . The statement that (X, Y, f) is a function is written
as f : X →Y or X
f
→Y .
Deﬁnition The graph of a function (X, Y, f) is the subset Γ ⊂ X Y deﬁned
by Γ = ¦(x, f(x)) : x ∈ X¦. The connection between the “intuitive” and “graph”
viewpoints is given in the next theorem.
Theorem If f : X → Y , then the graph Γ ⊂ X Y has the property that each
x ∈ X is the ﬁrst term of one and only one ordered pair in Γ. Conversely, if Γ is a
subset of X Y with the property that each x ∈ X is the ﬁrst term of one and only
ordered pair in Γ, then ∃! f : X → Y whose graph is Γ. The function is deﬁned by
“f(x) is the second term of the ordered pair in Γ whose ﬁrst term is x.”
Example Identity functions Here X = Y and f : X → X is deﬁned by
f(x) = x for all x ∈ X. The identity on X is denoted by I
X
or just I : X →X.
Example Constant functions Suppose y
0
∈ Y . Deﬁne f : X → Y by f(x) =
y
0
for all x ∈ X.
Restriction Given f : X →Y and a non-void subset S of X, deﬁne f [ S : S →Y
by (f [ S)(s) = f(s) for all s ∈ S.
Inclusion If S is a non-void subset of X, deﬁne the inclusion i : S → X by
i(s) = s for all s ∈ S. Note that inclusion is a restriction of the identity.
Composition Given W
f
→X
g
→Y deﬁne g ◦ f : W →Y by
(g ◦ f)(x) = g(f(x)).
Theorem (The associative law of composition) If V
f
→ W
g
→ X
h
→ Y , then
h ◦ (g ◦ f) = (h ◦ g) ◦ f. This may be written as h ◦ g ◦ f.
Chapter 1 Background 7
Deﬁnitions Suppose f : X →Y .
1) If T ⊂ Y , the inverse image of T is a subset of X, f
−1
(T) = ¦x ∈ X :
f(x) ∈ T¦.
2) If S ⊂ X, the image of S is a subset of Y , f(S) = ¦f(s) : s ∈ S¦ =
¦y ∈ Y : ∃s ∈ S with f(s) = y¦.
3) The image of f is the image of X , i.e., image (f) = f(X) =
¦f(x) : x ∈ X¦ = ¦y ∈ Y : ∃x ∈ X with f(x) = y¦.
4) f : X →Y is surjective or onto provided image (f) = Y i.e., the image
is the range, i.e., if y ∈ Y , f
−1
(y) is a non-void subset of X.
5) f : X →Y is injective or 1-1 provided (x
1
= x
2
) ⇒f(x
1
) = f(x
2
), i.e.,
if x
1
and x
2
are distinct elements of X, then f(x
1
) and f(x
2
) are
distinct elements of Y .
6) f : X →Y is bijective or is a 1-1 correspondence provided f is surjective
and injective. In this case, there is function f
−1
: Y →X with f
−1
◦ f =
I
X
: X →X and f ◦ f
−1
= I
Y
: Y →Y . Note that f
−1
: Y →X is
also bijective and (f
−1
)
−1
= f.
Examples
1) f : R →R deﬁned by f(x) = sin(x) is neither surjective nor injective.
2) f : R →[−1, 1] deﬁned by f(x) = sin(x) is surjective but not injective.
3) f : [0, π/2] →R deﬁned by f(x) = sin(x) is injective but not surjective.
4) f : [0, π/2] →[0, 1] deﬁned by f(x) = sin(x) is bijective. (f
−1
(x) is
written as arcsin(x) or sin
−1
(x).)
5) f : R →(0, ∞) deﬁned by f(x) = e
x
is bijective. (f
−1
(x) is written as
ln(x).)
Note There is no such thing as “the function sin(x).” A function is not deﬁned
unless the domain and range are speciﬁed.
8 Background Chapter 1
Exercise Show there are natural bijections from (R R
2
) to (R
2
R) and
from (R
2
R) to R R R. These three sets are disjoint, but the bijections
between them are so natural that we sometimes identify them.
Exercise Suppose X is a set with 6 elements and Y is a ﬁnite set with n elements.
1) There exists an injective f : X →Y iﬀ n .
2) There exists a surjective f : X →Y iﬀ n .
3) There exists a bijective f : X →Y iﬀ n .
Pigeonhole Principle Suppose X is a ﬁnite set with m elements, Y is a ﬁnite
set with n elements, and f : X →Y is a function.
1) If m = n, then f is injective iﬀ f is surjective iﬀ f is bijective.
2) If m > n, then f is not injective.
3) If m < n, then f is not surjective.
If you are placing 6 pigeons in 6 holes, and you run out of pigeons before you ﬁll
the holes, then you have placed 2 pigeons in one hole. In other words, in part 1) for
m = n = 6, if f is not surjective then f is not injective. Of course, the pigeonhole
principle does not hold for inﬁnite sets, as can be seen by the following exercise.
Exercise Show there is a function f : Z
+
→ Z
+
which is injective but not
surjective. Also show there is one which is surjective but not injective.
Exercise Suppose f : [−2, 2] → R is deﬁned by f(x) = x
2
. Find f
−1
(f([1, 2])).
Also ﬁnd f(f
−1
([3, 5])).
Exercise Suppose f : X → Y is a function, S ⊂ X and T ⊂ Y . Find the
relationship between S and f
−1
(f(S)). Show that if f is injective, S = f
−1
(f(S)).
Also ﬁnd the relationship between T and f(f
−1
(T)). Show that if f is surjective,
T = f(f
−1
(T)).
Strips We now deﬁne the vertical and horizontal strips of X Y .
If x
0
∈ X, ¦(x
0
, y) : y ∈ Y ¦ = (x
0
Y ) is called a vertical strip.
If y
0
∈ Y, ¦(x, y
0
) : x ∈ X¦ = (X y
0
) is called a horizontal strip.
Chapter 1 Background 9
Theorem Suppose S ⊂ X Y . The subset S is the graph of a function with
domain X and range Y iﬀ each vertical strip intersects S in exactly one point.
This is just a restatement of the property of a graph of a function. The purpose
of the next theorem is to restate properties of functions in terms of horizontal strips.
Theorem Suppose f : X →Y has graph Γ. Then
1) Each horizontal strip intersects Γ in at least one point iﬀ f is .
2) Each horizontal strip intersects Γ in at most one point iﬀ f is .
3) Each horizontal strip intersects Γ in exactly one point iﬀ f is .
Solutions of Equations Now we restate these properties in terms of solutions of
equations. Suppose f : X → Y and y
0
∈ Y . Consider the equation f(x) = y
0
. Here
y
0
is given and x is considered to be a “variable”. A solution to this equation is any
x
0
∈ X with f(x
0
) = y
0
. Note that the set of all solutions to f(x) = y
0
is f
−1
(y
0
).
Also f(x) = y
0
has a solution iﬀ y
0
∈ image(f) iﬀ f
−1
(y
0
) is non-void.
Theorem Suppose f : X →Y .
1) The equation f(x) = y
0
has at least one solution for each y
0
∈ Y iﬀ
f is .
2) The equation f(x) = y
0
has at most one solution for each y
0
∈ Y iﬀ
f is .
3) The equation f(x) = y
0
has a unique solution for each y
0
∈ Y iﬀ
f is .
Right and Left Inverses One way to understand functions is to study right and
left inverses, which are deﬁned after the next theorem.
Theorem Suppose X
f
→Y
g
→W are functions.
1) If g ◦ f is injective, then f is injective.
10 Background Chapter 1
2) If g ◦ f is surjective, then g is surjective.
3) If g ◦ f is bijective, then f is injective and g is surjective.
Example X = W = ¦p¦, Y = ¦p, q¦, f(p) = p, and g(p) = g(q) = p. Here
g ◦ f is the identity, but f is not surjective and g is not injective.
Deﬁnition Suppose f : X → Y is a function. A left inverse of f is a function
g : Y → X such that g ◦ f = I
X
: X → X. A right inverse of f is a function
h : Y →X such that f ◦ h = I
Y
: Y →Y .
Theorem Suppose f : X →Y is a function.
1) f has a right inverse iﬀ f is surjective. Any such right inverse must be
injective.
2) f has a left inverse iﬀ f is injective. Any such left inverse must be
surjective.
Corollary Suppose each of X and Y is a non-void set. Then ∃ an injective
f : X → Y iﬀ ∃ a surjective g : Y → X. Also a function from X to Y is bijective
iﬀ it has a left inverse and a right inverse iﬀ it has a left and right inverse.
Note The Axiom of Choice is not discussed in this book. However, if you worked
1) of the theorem above, you unknowingly used one version of it. For completeness,
we state this part of 1) again.
The Axiom of Choice If f : X → Y is surjective, then f has a right inverse
h. That is, for each y ∈ Y , it is possible to choose an x ∈ f
−1
(y) and thus to deﬁne
h(y) = x.
Note It is a classical theorem in set theory that the Axiom of Choice and the
Hausdorﬀ Maximality Principle are equivalent. However in this text we do not go
that deeply into set theory. For our purposes it is assumed that the Axiom of Choice
and the HMP are true.
Exercise Suppose f : X → Y is a function. Deﬁne a relation on X by a ∼ b if
f(a) = f(b). Show this is an equivalence relation. If y belongs to the image of f,
then f
−1
(y) is an equivalence class and every equivalence class is of this form. In the
next chapter where f is a group homomorphism, these equivalence classes will be
called cosets.
Chapter 1 Background 11
Projections If X
1
and X
2
are non-void sets, we deﬁne the projection maps
π
1
: X
1
X
2
→X
1
and π
2
: X
1
X
2
→X
2
by π
i
(x
1
, x
2
) = x
i
.
Theorem If Y, X
1
, and X
2
are non-void sets, there is a 1-1 correspondence
between ¦functions f: Y →X
1
X
2
¦ and ¦ordered pairs of functions (f
1
, f
2
) where
f
1
: Y →X
1
and f
2
: Y →X
2
¦.
Proof Given f, deﬁne f
1
= π
1
◦ f and f
2
= π
2
◦ f. Given f
1
and f
2
deﬁne
f : Y → X
1
X
2
by f(y) = (f
1
(y), f
2
(y)). Thus a function from Y to X
1
X
2
is
merely a pair of functions from Y to X
1
and Y to X
2
. This concept is displayed in
the diagram below. It is summarized by the equation f = (f
1
, f
2
).
X
1
X
2
X
1
X
2
Y
·
,

»

f
1
f
2
f
π
1
π
2
products.
Deﬁnition Suppose T is an index set and for each t ∈ T, X
t
is a non-void set.
Then the product
¸
t∈T
X
t
=
¸
X
t
is the collection of all sequences ¦x
t
¦
t∈T
= ¦x
t
¦
where x
t
∈ X
t
. Formally these sequences are functions α from T to
¸
X
t
with each
α(t) in X
t
and written as α(t) = x
t
. If T = ¦1, 2, . . . , n¦ then ¦x
t
¦ is the ordered
n-tuple (x
1
, x
2
, . . . , x
n
). If T = Z
+
then ¦x
t
¦ is the sequence (x
1
, x
2
, . . .). For any T
and any s in T, the projection map π
s
:
¸
X
t
→X
s
is deﬁned by π
s
(¦x
t
¦) = x
s
.
Theorem If Y is any non-void set, there is a 1-1 correspondence between
¦functions f : Y →
¸
X
t
¦ and ¦sequences of functions ¦f
t
¦
t∈T
where f
t
: Y → X
t
¦.
Given f, the sequence ¦f
t
¦ is deﬁned by f
t
= π
t
◦ f. Given ¦f
t
¦, f is deﬁned by
f(y) = ¦f
t
(y)¦.
12 Background Chapter 1
A Calculus Exercise Let A be the collection of all functions f : [0, 1] → R
which have an inﬁnite number of derivatives. Let A
0
⊂ A be the subcollection of
those functions f with f(0) = 0. Deﬁne D : A
0
→A by D(f) = df/dx. Use the mean
value theorem to show that D is injective. Use the fundamental theorem of calculus
to show that D is surjective.
Exercise This exercise is not used elsewhere in this text and may be omitted. It
is included here for students who wish to do a little more set theory. Suppose T is a
non-void set.
1) If Y is a non-void set, deﬁne Y
T
to be the collection of all functions with domain
T and range Y . Show that if T and Y are ﬁnite sets with m and n elements, then
Y
T
has n
m
elements. In particular, when T = ¦1, 2, 3¦, Y
T
= Y Y Y has
n
3
elements. Show that if n ≥ 3, the subset of Y
{1,2,3}
of all injective functions has
n(n − 1)(n − 2) elements. These injective functions are called permutations on Y
taken 3 at a time. If T = N, then Y
T
is the inﬁnite product Y Y . That is,
Y
N
is the set of all inﬁnite sequences (y
1
, y
2
, . . .) where each y
i
∈ Y . For any Y and
T, let Y
t
be a copy of Y for each t ∈ T. Then Y
T
=
¸
t∈T
Y
t
.
2) Suppose each of Y
1
and Y
2
is a non-void set. Show there is a natural bijection
from (Y
1
Y
2
)
T
to Y
T
1
Y
T
2
. (This is the fundamental property of Cartesian products
presented in the two previous theorems.)
3) Deﬁne {(T), the power set of T, to be the collection of all subsets of T (including
the null set). Show that if T is a ﬁnite set with m elements, {(T) has 2
m
elements.
4) If S is any subset of T, deﬁne its characteristic function χ
S
: T → ¦0, 1¦ by
letting χ
S
(t) be 1 when t ∈ S, and be 0 when t ∈[ S. Deﬁne α : {(T) → ¦0, 1¦
T
by
α(S) = χ
S
. Deﬁne β : ¦0, 1¦
T
→ {(T) by β(f) = f
−1
(1). Show that if S ⊂ T then
β ◦ α(S) = S, and if f : T → ¦0, 1¦ then α ◦ β(f) = f. Thus α is a bijection and
β = α
−1
.
{(T) ←→¦0, 1¦
T
5) Suppose γ : T →¦0, 1¦
T
is a function and show that it cannot be surjective. If
t ∈ T, denote γ(t) by γ(t) = f
t
: T → ¦0, 1¦. Deﬁne f : T → ¦0, 1¦ by f(t) = 0 if
f
t
(t) = 1, and f(t) = 1 if f
t
(t) = 0. Show that f is not in the image of γ and thus
γ cannot be surjective. This shows that if T is an inﬁnite set, then the set ¦0, 1¦
T
represents a “higher order of inﬁnity than T”.
6) An inﬁnite set Y is said to be countable if there is a bijection from the positive
Chapter 1 Background 13
integers N to Y. Show Q is countable but the following three collections are not.
i) {(N), the collection of all subsets of N.
ii) ¦0, 1¦
N
, the collection of all functions f : N →¦0, 1¦.
iii) The collection of all sequences (y
1
, y
2
, . . .) where each y
i
is 0 or 1.
We know that ii) and iii) are equal and there is a natural bijection between i)
and ii). We also know there is no surjective map from N to ¦0, 1¦
N
, i.e., ¦0, 1¦
N
is
uncountable. Finally, show there is a bijection from ¦0, 1¦
N
to the real numbers R.
(This is not so easy. To start with, you have to decide what the real numbers are.)
Notation for the Logic of Mathematics
Each of the words “Lemma”, “Theorem”, and “Corollary” means “true state-
ment”. Suppose A and B are statements. A theorem may be stated in any of the
following ways:
Theorem Hypothesis Statement A.
Conclusion Statement B.
Theorem Suppose A is true. Then B is true.
Theorem If A is true, then B is true.
Theorem A ⇒B (A implies B ).
There are two ways to prove the theorem — to suppose A is true and show B is
true, or to suppose B is false and show A is false. The expressions “A ⇔ B”, “A is
equivalent to B”, and “A is true iﬀ B is true ” have the same meaning (namely, that
A ⇒B and B ⇒A).
The important thing to remember is that thoughts and expressions ﬂow through
the language. Mathematical symbols are shorthand for phrases and sentences in the
English language. For example, “x ∈ B ” means “x is an element of the set B.” If A
is the statement “x ∈ Z
+
” and B is the statement “x
2
∈ Z
+
”, then “A ⇒ B”means
“If x is a positive integer, then x
2
is a positive integer”.
Mathematical Induction is based upon the fact that if S ⊂ Z
+
is a non-void
subset, then S contains a smallest element.
14 Background Chapter 1
Theorem Suppose P(n) is a statement for each n = 1, 2, ... . Suppose P(1) is
true and for each n ≥ 1, P(n) ⇒P(n + 1). Then for each n ≥ 1, P(n) is true.
Proof If the theorem is false, then ∃ a smallest positive integer m such that
P(m) is false. Since P(m−1) is true, this is impossible.
Exercise Use induction to show that, for each n ≥ 1, 1+2+ +n = n(n+1)/2.
The Integers
In this section, lower case letters a, b, c, ... will represent integers, i.e., elements
of Z. Here we will establish the following three basic properties of the integers.
1) If G is a subgroup of Z, then ∃ n ≥ 0 such that G = nZ.
2) If a and b are integers, not both zero, and G is the collection of all linear
combinations of a and b, then G is a subgroup of Z, and its
positive generator is the greatest common divisor of a and b.
3) If n ≥ 2, then n factors uniquely as the product of primes.
All of this will follow from long division, which we now state formally.
Euclidean Algorithm Given a, b with b = 0, ∃! m and r with 0 ≤ r <[b[ and
a = bm + r. In other words, b divides a “m times with a remainder of r”. For
example, if a = −17 and b = 5, then m = −4 and r = 3, −17 = 5(−4) + 3.
Deﬁnition If r = 0, we say that b divides a or a is a multiple of b. This fact is
written as b [ a. Note that b [ a ⇔ the rational number a/b is an integer ⇔ ∃! m
such that a = bm ⇔ a ∈ bZ.
Note Anything (except 0) divides 0. 0 does not divide anything.
± 1 divides anything . If n = 0, the set of integers which n divides
is nZ = ¦nm : m ∈ Z¦ = ¦..., −2n, −n, 0, n, 2n, ...¦. Also n divides
a and b with the same remainder iﬀ n divides (a −b).
Deﬁnition A non-void subset G ⊂ Z is a subgroup provided (g ∈ G ⇒ −g ∈ G)
and (g
1
, g
2
∈ G ⇒(g
1
+g
2
) ∈ G). We say that G is closed under negation and closed
Chapter 1 Background 15
Theorem If n ∈ Z then nZ is a subgroup. Thus if n = 0, the set of integers
which n divides is a subgroup of Z.
The next theorem states that every subgroup of Z is of this form.
Theorem Suppose G ⊂ Z is a subgroup. Then
1) 0 ∈ G.
2) If g
1
and g
2
∈ G, then (m
1
g
1
+m
2
g
2
) ∈ G for all integers m
1
, m
2
.
3) ∃! non-negative integer n such that G = nZ. In fact, if G = ¦0¦
and n is the smallest positive integer in G, then G = nZ.
Proof Since G is non-void, ∃ g ∈ G. Now (−g) ∈ G and thus 0 = g + (−g)
belongs to G, and so 1) is true. Part 2) is straightforward, so consider 3). If G = 0,
it must contain a positive element. Let n be the smallest positive integer in G. If
g ∈ G, g = nm +r where 0 ≤ r < n. Since r ∈ G, it must be 0, and g ∈ nZ.
Now suppose a, b ∈ Z and at least one of a and b is non-zero.
Theorem Let G be the set of all linear combinations of a and b, i.e., G =
¦ma +nb : m, n ∈ Z¦. Then
1) G contains a and b.
2) G is a subgroup. In fact, it is the smallest subgroup containing a and b.
It is called the subgroup generated by a and b.
3) Denote by (a, b) the smallest positive integer in G. By the previous
theorem, G = (a, b)Z, and thus (a, b) [ a and (a, b) [ b. Also note that
∃ m, n such that ma +nb = (a, b). The integer (a, b) is called
the greatest common divisor of a and b.
4) If n is an integer which divides a and b, then n also divides (a, b).
Proof of 4) Suppose n [ a and n [ b i.e., suppose a, b ∈ nZ. Since G is the
smallest subgroup containing a and b, nZ ⊃ (a, b)Z, and thus n [ (a, b).
Corollary The following are equivalent.
1) a and b have no common divisors, i.e., (n [ a and n [ b) ⇒n = ±1.
16 Background Chapter 1
2) (a, b) = 1, i.e., the subgroup generated by a and b is all of Z.
3) ∃ m, n ∈Z with ma +nb = 1.
Deﬁnition If any one of these three conditions is satisﬁed, we say that a and b
are relatively prime.
This next theorem is the basis for unique factorization.
Theorem If a and b are relatively prime with a not zero, then a[bc ⇒a[c.
Proof Suppose a and b are relatively prime, c ∈ Z and a [ bc. Then there exist
m, n with ma + nb = 1, and thus mac + nbc = c. Now a [ mac and a [ nbc. Thus
a [ (mac +nbc) and so a [ c.
Deﬁnition A prime is an integer p > 1 which does not factor, i.e., if p = ab then
a = ±1 or a = ±p. The ﬁrst few primes are 2, 3, 5, 7, 11, 13, 17,... .
Theorem Suppose p is a prime.
1) If a is an integer which is not a multiple of p, then (p, a) = 1. In other
words, if a is any integer, (p, a) = p or (p, a) = 1.
2) If p [ ab then p [ a or p [ b.
3) If p [ a
1
a
2
a
n
then p divides some a
i
. Thus if each a
i
is a prime,
then p is equal to some a
i
.
Proof Part 1) follows immediately from the deﬁnition of prime. Now suppose
p [ ab. If p does not divide a, then by 1), (p, a) = 1 and by the previous theorem, p
must divide b. Thus 2) is true. Part 3) follows from 2) and induction on n.
The Unique Factorization Theorem Suppose a is an integer which is not 0,1,
or -1. Then a may be factored into the product of primes and, except for order, this
factorization is unique. That is, ∃ a unique collection of distinct primes p
1
, p
2
, ..., p
k
and positive integers s
1
, s
2
, ..., s
k
such that a = ±p
s
1
1
p
s
2
2
p
s
k
k
.
Proof Factorization into primes is obvious, and uniqueness follows from 3) in the
theorem above. The power of this theorem is uniqueness, not existence.
Chapter 1 Background 17
Now that we have unique factorization and part 3) above, the picture becomes
transparent. Here are some of the basic properties of the integers in this light.
Theorem (Summary)
1) Suppose [ a[> 1 has prime factorization a = ±p
s
1
1
p
s
k
k
. Then the only
divisors of a are of the form ±p
t
1
1
p
t
k
k
where 0 ≤ t
i
≤ s
i
for i = 1, ..., k.
2) If [ a [> 1 and [ b [> 1, then (a, b) = 1 iﬀ there is no common prime in
their factorizations. Thus if there is no common prime in their
factorizations, ∃ m, n with ma +nb = 1, and also (a
2
, b
2
) = 1.
3) Suppose [ a[> 1 and [ b[> 1. Let ¦p
1
, . . . , p
k
¦ be the union of the distinct
primes of their factorizations. Thus a = ±p
s
1
1
p
s
k
k
where 0 ≤ s
i
and
b = ±p
t
1
1
p
t
k
k
where 0 ≤ t
i
. Let u
i
be the minimum of s
i
and t
i
. Then
(a, b) = p
u
1
1
p
u
k
k
. For example (2
3
5 11, 2
2
5
4
7) = 2
2
5.
3

) Let v
i
be the maximum of s
i
and t
i
. Then c = p
v
1
1
p
v
k
k
is the least
(positive) common multiple of a and b. Note that c is a multiple of
a and b, and if n is a multiple of a and b, then n is a multiple of c.
Finally, if a and b are positive, their least common multiple is
c = ab/(a, b), and if in addition a and b are relatively prime,
then their least common multiple is just their product.
4) There is an inﬁnite number of primes. (Proof: Suppose there were only
a ﬁnite number of primes p
1
, p
2
, ..., p
k
. Then no prime would divide
(p
1
p
2
p
k
+ 1).)
5) Suppose c is an integer greater than 1. Then

c is rational iﬀ

c is an
integer. In particular,

2 and

3 are irrational. (Proof: If

c is
rational, ∃ positive integers a and b with

c = a/b and (a, b) = 1.
If b > 1, then it is divisible by some prime, and since cb
2
= a
2
, this
prime will also appear in the prime factorization of a. This is a
contradiction and thus b = 1 and

c is an integer.) (See the ﬁfth
exercise below.)
Exercise Find (180,28), i.e., ﬁnd the greatest common divisor of 180 and 28,
i.e., ﬁnd the positive generator of the subgroup generated by ¦180,28¦. Find integers
m and n such that 180m + 28n = (180, 28). Find the least common multiple of 180
and 28, and show that it is equal to (180 28)/(180, 28).
18 Background Chapter 1
Exercise We have deﬁned the greatest common divisor (gcd) and the least com-
mon multiple (lcm) of a pair of integers. Now suppose n ≥ 2 and S = ¦a
1
, a
2
, .., a
n
¦
is a ﬁnite collection of integers with [a
i
[ > 1 for 1 ≤ i ≤ n. Deﬁne the gcd and the
lcm of the elements of S and develop their properties. Express the gcd and the lcm
in terms of the prime factorizations of the a
i
. When is the lcm of S equal to the
product a
1
a
2
a
n
? Show that the set of all linear combinations of the elements of
S is a subgroup of Z, and its positive generator is the gcd of the elements of S.
Exercise Show that the gcd of S = ¦90, 70, 42¦ is 2, and ﬁnd integers n
1
, n
2
, n
3
such that 90n
1
+ 70n
2
+ 42n
3
= 2. Also ﬁnd the lcm of the elements of S.
Exercise Show that if each of G
1
, G
2
, ..., G
m
is a subgroup of Z, then
G
1
∩ G
2
∩ ∩ G
m
is also a subgroup of Z. Now let G = (90Z) ∩ (70Z) ∩ (42Z)
and ﬁnd the positive integer n with G = nZ.
Exercise Show that if the nth root of an integer is a rational number, then it
itself is an integer. That is, suppose c and n are integers greater than 1. There is a
unique positive real number x with x
n
= c. Show that if x is rational, then it is an
integer. Thus if p is a prime, its nth root is an irrational number.
Exercise Show that a positive integer is divisible by 3 iﬀ the sum of its digits is
divisible by 3. More generally, let a = a
n
a
n−1
. . . a
0
= a
n
10
n
+ a
n−1
10
n−1
+ + a
0
where 0 ≤ a
i
≤ 9. Now let b = a
n
+a
n−1
+ +a
0
, and show that 3 divides a and b
with the same remainder. Although this is a straightforward exercise in long division,
it will be more transparent later on. In the language of the next chapter, it says that
[a] = [b] in Z
3
.
Card Trick Ask friends to pick out seven cards from a deck and then to select one
to look at without showing it to you. Take the six cards face down in your left hand
and the selected card in your right hand, and announce you will place the selected
card in with the other six, but they are not to know where. Put your hands behind
your back and place the selected card on top, and bring the seven cards in front in
allowing one). Suppose they say three. You move the top card to the bottom, then
the second card to the bottom, and then you turn over the third card, leaving it face
up on top. Then repeat the process, moving the top two cards to the bottom and
turning the third card face up on top. Continue until there is only one card face
down, and this will be the selected card. Magic? Stay tuned for Chapter 2, where it
is shown that any non-zero element of Z
7
has order 7.
Chapter 2
Groups
Groups are the central objects of algebra. In later chapters we will deﬁne rings and
modules and see that they are special cases of groups. Also ring homomorphisms and
module homomorphisms are special cases of group homomorphisms. Even though
the deﬁnition of group is simple, it leads to a rich and amazing theory. Everything
presented here is standard, except that the product of groups is given in the additive
notation. This is the notation used in later chapters for the products of rings and
modules. This chapter and the next two chapters are restricted to the most basic
topics. The approach is to do quickly the fundamentals of groups, rings, and matrices,
and to push forward to the chapter on linear algebra. This chapter is, by far and
above, the most diﬃcult chapter in the book, because group operations may be written
as addition or multiplication, and also the concept of coset is confusing at ﬁrst.
Deﬁnition Suppose G is a non-void set and φ : G G → G is a function. φ is
called a binary operation, and we will write φ(a, b) = a b or φ(a, b) = a+b. Consider
the following properties.
1) If a, b, c ∈ G then a (b c) = (a b) c. If a, b, c ∈ G then a + (b +c) = (a +b) +c.
2) ∃ e = e
G
∈ G such that if a ∈ G ∃ 0
¯
=0
¯
G
∈ G such that if a ∈ G
e a = a e = a. 0
¯
+a = a+0
¯
= a.
3) If a ∈ G, ∃b ∈ G with a b = b a = e If a ∈ G, ∃b ∈ G with a +b = b +a = 0
¯
(b is written as b = a
−1
). (b is written as b = −a).
4) If a, b ∈ G, then a b = b a. If a, b ∈ G, then a +b = b +a.
Deﬁnition If properties 1), 2), and 3) hold, (G, φ) is said to be a group. If we
write φ(a, b) = a b, we say it is a multiplicative group. If we write φ(a, b) = a + b,
19
20 Groups Chapter 2
we say it is an additive group. If in addition, property 4) holds, we say the group is
abelian or commutative.
Theorem Let (G, φ) be a multiplicative group.
(i) Suppose a, c, ¯ c ∈ G. Then a c = a ¯ c ⇒ c = ¯ c.
Also c a = ¯ c a ⇒ c = ¯ c.
In other words, if f : G →G is deﬁned by f(c) = a c, then f is injective.
Also f is bijective with f
−1
given by f
−1
(c) = a
−1
c.
(ii) e is unique, i.e., if ¯ e ∈ G satisﬁes 2), then e = ¯ e. In fact,
if a, b ∈ G then (a b = a) ⇒(b = e) and (a b = b) ⇒ (a = e).
Recall that b is an identity in G provided it is a right and left
identity for any a in G. However, group structure is so rigid that if
∃ a ∈ G such that b is a right identity for a, then b = e.
Of course, this is just a special case of the cancellation law in (i).
(iii) Every right inverse is an inverse, i.e., if a b = e then b = a
−1
.
Also if b a = e then b = a
−1
. Thus inverses are unique.
(iv) If a ∈ G, then (a
−1
)
−1
= a.
(v) The multiplication a
1
a
2
a
3
= a
1
(a
2
a
3
) = (a
1
a
2
) a
3
is well-deﬁned.
In general, a
1
a
2
a
n
is well deﬁned.
(vi) If a, b ∈ G, (a b)
−1
= b
−1
a
−1
. Also (a
1
a
2
a
n
)
−1
=
a
−1
n
a
−1
n−1
a
−1
1
.
(vii) Suppose a ∈ G. Let a
0
= e and if n > 0, a
n
= a a (n times)
and a
−n
= a
−1
a
−1
(n times). If n
1
, n
2
, ..., n
t
∈ Z then
a
n
1
a
n
2
a
nt
= a
n
1
+···+nt
. Also (a
n
)
m
= a
nm
.
Finally, if G is abelian and a, b ∈ G, then (a b)
n
= a
n
b
n
.
Exercise. Write out the above theorem where G is an additive group. Note that
part (vii) states that G has a scalar multiplication over Z. This means that if a is in
G and n is an integer, there is deﬁned an element an in G. This is so basic, that we
state it explicitly.
Theorem. Suppose G is an additive group. If a ∈ G, let a0 =0
¯
and if n > 0,
let an = (a + +a) where the sum is n times, and a(−n) = (−a) + (−a) +(−a),
Chapter 2 Groups 21
which we write as (−a − a − a). Then the following properties hold in general,
except the ﬁrst requires that G be abelian.
(a +b)n = an +bn
a(n +m) = an +am
a(nm) = (an)m
a1 = a
Note that the plus sign is used ambiguously — sometimes for addition in G
and sometimes for addition in Z. In the language used in Chapter 5, this theorem
states that any additive abelian group is a Z-module. (See page 71.)
Exercise Suppose G is a non-void set with a binary operation φ(a, b) = a b which
satisﬁes 1), 2) and [ 3

) If a ∈ G, ∃b ∈ G with a b = e]. Show (G, φ) is a group,
i.e., show b a = e. In other words, the group axioms are stronger than necessary.
If every element has a right inverse, then every element has a two sided inverse.
Exercise Suppose G is the set of all functions from Z to Z with multiplication
deﬁned by composition, i.e., f g = f ◦ g. Note that G satisﬁes 1) and 2) but not 3),
and thus G is not a group. Show that f has a right inverse in G iﬀ f is surjective,
and f has a left inverse in G iﬀ f is injective (see page 10). Also show that the set
of all bijections from Z to Z is a group under composition.
Examples G = R, G = Q, or G = Z with φ(a, b) = a +b is an additive
abelian group.
Examples G = R−0 or G = Q−0 with φ(a, b) = ab is a multiplicative
abelian group.
G = Z −0 with φ(a, b) = ab is not a group.
G = R
+
= ¦r ∈ R : r > 0¦ with φ(a, b) = ab is a multiplicative
abelian group.
Subgroups
Theorem Suppose G is a multiplicative group and H ⊂ G is a non-void subset
satisfying
1) if a, b ∈ H then a b ∈ H
and 2) if a ∈ H then a
−1
∈ H.
22 Groups Chapter 2
Then e ∈ H and H is a group under multiplication. H is called a subgroup of G.
Proof Since H is non-void, ∃a ∈ H. By 2), a
−1
∈ H and so by 1), e ∈ H. The
associative law is immediate and so H is a group.
Example G is a subgroup of G and e is a subgroup of G. These are called the
improper subgroups of G.
Example If G = Z under addition, and n ∈ Z, then H = nZ is a subgroup of
Z. By a theorem in the section on the integers in Chapter 1, every subgroup of Z
is of this form (see page 15). This is a key property of the integers.
Exercises Suppose G is a multiplicative group.
1) Let H be the center of G, i.e., H = ¦h ∈ G : g h = h g for all g ∈ G¦. Show
H is a subgroup of G.
2) Suppose H
1
and H
2
are subgroups of G. Show H
1
∩ H
2
is a subgroup of G.
3) Suppose H
1
and H
2
are subgroups of G, with neither H
1
nor H
2
contained in
the other. Show H
1
∪ H
2
is not a subgroup of G.
4) Suppose T is an index set and for each t ∈ T, H
t
is a subgroup of G.
Show
¸
t∈T
H
t
is a subgroup of G.
5) Furthermore, if ¦H
t
¦ is a monotonic collection, then
¸
t∈T
H
t
is a subgroup of G.
6) Suppose G= ¦all functions f : [0, 1] →R¦. Deﬁne an addition on G by
(f +g)(t) = f(t) +g(t) for all t ∈ [0, 1]. This makes G into an abelian group.
Let K be the subset of G composed of all diﬀerentiable functions. Let H
be the subset of G composed of all continuous functions. What theorems
in calculus show that H and K are subgroups of G? What theorem shows
that K is a subset (and thus subgroup) of H?
Order Suppose G is a multiplicative group. If G has an inﬁnite number of
Chapter 2 Groups 23
elements, we say that o(G), the order of G, is inﬁnite. If G has n elements, then
o(G) = n. Suppose a ∈ G and H = ¦a
i
: i ∈ Z¦. H is an abelian subgroup of G
called the subgroup generated by a. We deﬁne the order of the element a to be the
order of H, i.e., the order of the subgroup generated by a. Let f : Z → H be the
surjective function deﬁned by f(m) = a
m
. Note that f(k + l) = f(k) f(l) where
the addition is in Z and the multiplication is in the group H. We come now to the
ﬁrst real theorem in group theory. It says that the element a has ﬁnite order iﬀ f
is not injective, and in this case, the order of a is the smallest positive integer n
with a
n
= e.
Theorem Suppose a is an element of a multiplicative group G, and
H = ¦a
i
: i ∈ Z¦. If ∃ distinct integers i and j with a
i
= a
j
, then a has some ﬁnite
order n. In this case H has n distinct elements, H = ¦a
0
, a
1
, . . . , a
n−1
¦, and a
m
= e
iﬀ n[m. In particular, the order of a is the smallest positive integer n with a
n
= e,
and f
−1
(e) = nZ.
Proof Suppose j < i and a
i
= a
j
. Then a
i−j
= e and thus ∃ a smallest positive
integer n with a
n
= e. This implies that the elements of ¦a
0
, a
1
, ..., a
n−1
¦ are distinct,
and we must show they are all of H. If m ∈ Z, the Euclidean algorithm states that
∃ integers q and r with 0 ≤ r < n and m = nq + r. Thus a
m
= a
nq
a
r
= a
r
, and
so H = ¦a
0
, a
1
, ..., a
n−1
¦, and a
m
= e iﬀ n[m. Later in this chapter we will see that
f is a homomorphism from an additive group to a multiplicative group and that,
in additive notation, H is isomorphic to Z or Z
n
.
Exercise Write out this theorem for G an additive group. To begin, suppose a is
an element of an additive group G, and H = ¦ai : i ∈ Z¦.
Exercise Show that if G is a ﬁnite group of even order, then G has an odd number
of elements of order 2. Note that e is the only element of order 1.
Deﬁnition A group G is cyclic if ∃ an element of G which generates G.
Theorem If G is cyclic and H is a subgroup of G, then H is cyclic.
Proof Suppose G = ¦a
i
: i ∈ Z¦ is a cyclic group and H is a subgroup
of G. If H = e, then H is cyclic, so suppose H = e. Now there is a small-
est positive integer m with a
m
∈ H. If t is an integer with a
t
∈ H, then by
the Euclidean algorithm, m divides t, and thus a
m
generates H. Note that in
the case G has ﬁnite order n, i.e., G = ¦a
0
, a
1
, . . . , a
n−1
¦, then a
n
= e ∈ H,
and thus the positive integer m divides n. In either case, we have a clear picture
of the subgroups of G. Also note that this theorem was proved on page 15 for the
24 Groups Chapter 2
Cosets Suppose H is a subgroup of a group G. It will be shown below that H
partitions G into right cosets. It also partitions G into left cosets, and in general
these partitions are distinct.
Theorem If H is a subgroup of a multiplicative group G, then a ∼ b deﬁned by
a ∼ b iﬀ a b
−1
∈ H is an equivalence relation. If a ∈ G, cl(a) = ¦b ∈ G : a ∼ b¦ =
¦h a : h ∈ H¦ = Ha. Note that a b
−1
∈ H iﬀ b a
−1
∈ H.
If H is a subgroup of an additive group G, then a ∼ b deﬁned by a ∼ b iﬀ
(a − b) ∈ H is an equivalence relation. If a ∈ G, cl(a) = ¦b ∈ G : a ∼ b¦ = ¦h + a :
h ∈ H¦ = H +a. Note that (a −b) ∈ H iﬀ (b −a) ∈ H.
Deﬁnition These equivalence classes are called right cosets. If the relation is
deﬁned by a ∼ b iﬀ b
−1
a ∈ H, then the equivalence classes are cl(a) = aH and
they are called left cosets. H is a left and right coset. If G is abelian, there is no
distinction between right and left cosets. Note that b
−1
a ∈ H iﬀ a
−1
b ∈ H.
In the theorem above, H is used to deﬁne an equivalence relation on G, and thus
a partition of G. We now do the same thing a diﬀerent way. We deﬁne the right
cosets directly and show they form a partition of G. You might ﬁnd this easier.
Theorem Suppose H is a subgroup of a multiplicative group G. If a ∈ G, deﬁne
the right coset containing a to be Ha = ¦h a : h ∈ H¦. Then the following hold.
1) Ha = H iﬀ a ∈ H.
2) If b ∈ Ha, then Hb = Ha, i.e., if h ∈ H, then H(h a) = (Hh)a = Ha.
3) If Hc ∩ Ha = ∅, then Hc = Ha.
4) The right cosets form a partition of G, i.e., each a in G belongs to one and
only one right coset.
5) Elements a and b belong to the same right coset iﬀ a b
−1
∈ H iﬀ b a
−1
∈ H.
Proof There is no better way to develop facility with cosets than to prove this
theorem. Also write this theorem for G an additive group.
Theorem Suppose H is a subgroup of a multiplicative group G.
Chapter 2 Groups 25
1) Any two right cosets have the same number of elements. That is, if a, b ∈ G,
f : Ha →Hb deﬁned by f(h a) = h b is a bijection. Also any two left cosets
have the same number of elements. Since H is a right and left coset, any
two cosets have the same number of elements.
2) G has the same number of right cosets as left cosets. The function F deﬁned
by F(Ha) = a
−1
H is a bijection from the collection of right cosets to the left
cosets. The number of right (or left) cosets is called the index of H in G.
3) If G is ﬁnite, o(H) (index of H) = o(G) and so o(H) [ o(G). In other words,
o(G)/o(H) = the number of right cosets = the number of left cosets.
4) If G is ﬁnite, and a ∈ G, then o(a) [ o(G). (Proof: The order of a is the order
of the subgroup generated by a, and by 3) this divides the order of G.)
5) If G has prime order, then G is cyclic, and any element (except e) is a generator.
(Proof: Suppose o(G) = p and a ∈ G, a = e. Then o(a) [ p and thus o(a) = p.)
6) If o(G) = n and a ∈ G, then a
n
= e. (Proof: a
o(a)
= e and n = o(a) (o(G)/o(a)) .)
Exercises
i) Suppose G is a cyclic group of order 4, G = ¦e, a, a
2
, a
3
¦ with a
4
= e. Find the
order of each element of G. Find all the subgroups of G.
ii) Suppose G is the additive group Z and H = 3Z. Find the cosets of H.
iii) Think of a circle as the interval [0, 1] with end points identiﬁed. Suppose G = R
under addition and H = Z. Show that the collection of all the cosets of H
can be thought of as a circle.
iv) Let G = R
2
under addition, and H be the subgroup deﬁned by
H = ¦(a, 2a) : a ∈ R¦. Find the cosets of H. (See the last exercise on p 5.)
Normal Subgroups
We would like to make a group out of the collection of cosets of a subgroup H. In
26 Groups Chapter 2
general, there is no natural way to do that. However, it is easy to do in case H is a
normal subgroup, which is described below.
Theorem If H is a subgroup of a group G, then the following are equivalent.
1) If a ∈ G, then aHa
−1
= H
2) If a ∈ G, then aHa
−1
⊂ H
3) If a ∈ G, then aH = Ha
4) Every right coset is a left coset, i.e., if a ∈ G, ∃ b ∈ G with Ha = bH.
Proof 1) ⇒ 2) is obvious. Suppose 2) is true and show 3). We have (aHa
−1
)a ⊂
Ha so aH ⊂ Ha. Also a(a
−1
Ha) ⊂ aH so Ha ⊂ aH. Thus aH = Ha.
3) ⇒ 4) is obvious. Suppose 4) is true and show 3). Ha = bH contains a, so
bH = aH because a coset is an equivalence class. Thus aH = Ha.
Finally, suppose 3) is true and show 1). Multiply aH = Ha on the right by a
−1
.
Deﬁnition If H satisﬁes any of the four conditions above, then H is said to be a
normal subgroup of G. (This concept goes back to Evariste Galois in 1831.)
Note For any group G, G and e are normal subgroups. If G is an abelian group,
then every subgroup of G is normal.
Exercise Show that if H is a subgroup of G with index 2, then H is normal.
Exercise Show the intersection of a collection of normal subgroups of G is a
normal subgroup of G. Show the union of a monotonic collection of normal subgroups
of G is a normal subgroup of G.
Exercise Let A ⊂ R
2
be the square with vertices (−1, 1), (1, 1), (1, −1), and
(−1, −1), and G be the collection of all “isometries” of A onto itself. These are
bijections of A onto itself which preserve distance and angles, i.e., which preserve dot
product. Show that with multiplication deﬁned as composition, G is a multiplicative
group. Show that G has four rotations, two reﬂections about the axes, and two
reﬂections about the diagonals, for a total of eight elements. Show the collection of
rotations is a cyclic subgroup of order four which is a normal subgroup of G. Show
that the reﬂection about the x-axis together with the identity form a cyclic subgroup
of order two which is not a normal subgroup of G. Find the four right cosets of this
subgroup. Finally, ﬁnd the four left cosets of this subgroup.
Chapter 2 Groups 27
Quotient Groups Suppose N is a normal subgroup of G, and C and D are
cosets. We wish to deﬁne a coset E which is the product of C and D. If c ∈ C and
d ∈ D, deﬁne E to be the coset containing c d, i.e., E = N(c d). The coset E does
not depend upon the choice of c and d. This is made precise in the next theorem,
which is quite easy.
Theorem Suppose G is a multiplicative group, N is a normal subgroup, and
G/N is the collection of all cosets. Then (Na) (Nb) = N(a b) is a well de-
ﬁned multiplication (binary operation) on G/N, and with this multiplication, G/N
is a group. Its identity is N and (Na)
−1
= (Na
−1
). Furthermore, if G is ﬁnite,
o(G/N) = o(G)/o(N).
Proof Multiplication of elements in G/N is multiplication of subsets in G.
(Na) (Nb) = N(aN)b = N(Na)b = N(a b). Once multiplication is well deﬁned,
the group axioms are immediate.
Exercise Write out the above theorem for G an additive group. In the additive
abelian group R/Z, determine those elements of ﬁnite order.
Example Suppose G = Z under +, n > 1, and N = nZ. Z
n
, the group of
integers mod n is deﬁned by Z
n
= Z/nZ. If a is an integer, the coset a + nZ is
denoted by [a]. Note that [a] + [b] = [a + b], −[a] = [−a], and [a] = [a + nl] for any
integer l. Any additive abelian group has a scalar multiplication over Z, and in this
case it is just [a]m = [am]. Note that [a] = [r] where r is the remainder of a divided
by n, and thus the distinct elements of Z
n
are [0], [1], ..., [n − 1]. Also Z
n
is cyclic
because each of [1] and [−1] = [n −1] is a generator. We already know that if p is a
prime, any non-zero element of Z
p
is a generator, because Z
p
has p elements.
Theorem If n > 1 and a is any integer, then [a] is a generator of Z
n
iﬀ (a, n) = 1.
Proof The element [a] is a generator iﬀ the subgroup generated by [a] contains
[1] iﬀ ∃ an integer k such that [a]k = [1] iﬀ ∃ integers k and l such that ak +nl = 1.
Exercise Show that a positive integer is divisible by 3 iﬀ the sum of its digits is
divisible by 3. Note that [10] = [1] in Z
3
. (See the ﬁfth exercise on page 18.)
Homomorphisms
Homomorphisms are functions between groups that commute with the group op-
erations. It follows that they honor identities and inverses. In this section we list
28 Groups Chapter 2
the basic properties. Properties 11), 12), and 13) show the connections between coset
groups and homomorphisms, and should be considered as the cornerstones of abstract
algebra. As always, the student should rewrite the material in additive notation.
Deﬁnition If G and
¯
G are multiplicative groups, a function f : G →
¯
G is a
homomorphism if, for all a, b ∈ G, f(a b) = f(a) f(b). On the left side, the group
operation is in G, while on the right side it is in
¯
G. The kernel of f is deﬁned by
ker(f) = f
−1
(¯ e) = ¦a ∈ G : f(a) = ¯ e¦. In other words, the kernel is the set of
solutions to the equation f(x) = ¯ e. (If
¯
G is an additive group, ker(f) = f
−1
(0
¯
).)
Examples The constant map f : G →
¯
G deﬁned by f(a) = ¯ e is a homomorphism.
If H is a subgroup of G, the inclusion i : H → G is a homomorphism. The function
f : Z → Z deﬁned by f(t) = 2t is a homomorphism of additive groups, while the
function deﬁned by f(t) = t +2 is not a homomorphism. The function h : Z →R−0
deﬁned by h(t) = 2
t
is a homomorphism from an additive group to a multiplicative
group.
We now catalog the basic properties of homomorphisms. These will be helpful
later on in the study of ring homomorphisms and module homomorphisms.
Theorem Suppose G and
¯
G are groups and f : G →
¯
G is a homomorphism.
1) f(e) = ¯ e.
2) f(a
−1
) = f(a)
−1
. The ﬁrst inverse is in G, and the second is in
¯
G.
3) f is injective ⇔ ker(f) = e.
4) If H is a subgroup of G, f(H) is a subgroup of
¯
G. In particular, image(f) is
a subgroup of
¯
G.
5) If
¯
H is a subgroup of
¯
G, f
−1
(
¯
H) is a subgroup of G. Furthermore, if
¯
H is
normal in
¯
G, then f
−1
(
¯
H) is normal in G.
6) The kernel of f is a normal subgroup of G.
7) If ¯ g ∈
¯
G, f
−1
(¯ g) is void or is a coset of ker(f), i.e., if f(g) = ¯ g then
f
−1
(¯ g) = Ng where N= ker(f). In other words, if the equation f(x) = ¯ g has a
Chapter 2 Groups 29
solution, then the set of all solutions is a coset of N= ker(f). This is a key fact
which is used routinely in topics such as systems of equations and linear
diﬀerential equations.
8) The composition of homomorphisms is a homomorphism, i.e., if h :
¯
G →
=
G
is
a homomorphism, then h ◦ f : G →
=
G
is a homomorphism.
9) If f : G →
¯
G is a bijection, then the function f
−1
:
¯
G →G is a homomorphism.
In this case, f is called an isomorphism, and we write G ≈
¯
G. In the case
G =
¯
G, f is also called an automorphism.
10) Isomorphisms preserve all algebraic properties. For example, if f is an
isomorphism and H ⊂ G is a subset, then H is a subgroup of G
iﬀ f(H) is a subgroup of
¯
G, H is normal in G iﬀ f(H) is normal in
¯
G, G is
cyclic iﬀ
¯
G is cyclic, etc. Of course, this is somewhat of a cop-out, because an
algebraic property is one that, by deﬁnition, is preserved under isomorphisms.
11) Suppose H is a normal subgroup of G. Then π : G →G/H deﬁned by
π(a) = Ha is a surjective homomorphism with kernel H. Furthermore, if
f : G →
¯
G is a surjective homomorphism with kernel H, then G/H ≈
¯
G
(see below).
12) Suppose H is a normal subgroup of G. If H ⊂ ker(f), then
¯
f : G/H →
¯
G
deﬁned by
¯
f(Ha) = f(a) is a well-deﬁned homomorphism making
the following diagram commute.
G
¯
G
G/H
f
·

π
¯
f
Thus deﬁning a homomorphism on a quotient group is the same as deﬁning a
homomorphism on the numerator which sends the denominator to ¯ e. The
image of
¯
f is the image of f and the kernel of
¯
f is ker(f)/H. Thus if H = ker(f),
¯
f is injective, and thus G/H ≈ image(f).
13) Given any group homomorphism f, domain(f)/ker(f) ≈ image(f). This is
the fundamental connection between quotient groups and homomorphisms.
30 Groups Chapter 2
14) Suppose K is a group. Then K is an inﬁnite cycle group iﬀ K is isomorphic to
the integers under addition, i.e., K ≈ Z. K is a cyclic group of order n iﬀ
K ≈ Z
n
.
Proof of 14) Suppose
¯
G = K is generated by some element a. Then f : Z →K
deﬁned by f(m) = a
m
is a homomorphism from an additive group to a multiplicative
group. If o(a) is inﬁnite, f is an isomorphism. If o(a) = n, ker(f) = nZ and
¯
f : Z
n
→K is an isomorphism.
Exercise If a is an element of a group G, there is always a homomorphism from Z
to G which sends 1 to a. When is there a homomorphism from Z
n
to G which sends [1]
to a? What are the homomorphisms from Z
2
to Z
6
? What are the homomorphisms
from Z
4
to Z
8
?
Exercise Suppose G is a group and g is an element of G, g = e.
1) Under what conditions on g is there a homomorphism f : Z
7
→G with
f([1]) = g ?
2) Under what conditions on g is there a homomorphism f : Z
15
→G with
f([1]) = g ?
3) Under what conditions on G is there an injective homomorphism f : Z
15
→G ?
4) Under what conditions on G is there a surjective homomorphism f : Z
15
→G ?
Exercise We know every ﬁnite group of prime order is cyclic and thus abelian.
Show that every group of order four is abelian.
Exercise Let G = ¦h : [0, 1] → R : h has an inﬁnite number of derivatives¦.
Then G is a group under addition. Deﬁne f : G → G by f(h) =
dh
dt
= h

. Show f
is a homomorphism and ﬁnd its kernel and image. Let g : [0, 1] → R be deﬁned by
g(t) = t
3
−3t + 4. Find f
−1
(g) and show it is a coset of ker(f).
Exercise Let G be as above and g ∈ G. Deﬁne f : G →G by f(h) = h

+ 5h

+
6t
2
h. Then f is a group homomorphism and the diﬀerential equation h

+5h

+6t
2
h =
g has a solution iﬀ g lies in the image of f. Now suppose this equation has a solution
and S ⊂ G is the set of all solutions. For which subgroup H of G is S an H-coset?
Chapter 2 Groups 31
Exercise Suppose G is a multiplicative group and a ∈ G. Deﬁne f : G → G to
be conjugation by a, i.e., f(g) = a
−1
g a. Show that f is a homomorphism. Also
show f is an automorphism and ﬁnd its inverse.
Permutations
Suppose X is a (non-void) set. A bijection f : X → X is called a permutation
on X, and the collection of all these permutations is denoted by S = S(X). In this
setting, variables are written on the left, i.e., f = (x)f. Therefore the composition
f ◦g means “f followed by g”. S(X) forms a multiplicative group under composition.
Exercise Show that if there is a bijection between X and Y , there is an iso-
morphism between S(X) and S(Y ). Thus if each of X and Y has n elements,
S(X) ≈ S(Y ), and these groups are called the symmetric groups on n elements.
They are all denoted by the one symbol S
n
.
Exercise Show that o(S
n
) = n!. Let X = ¦1, 2, ..., n¦, S
n
= S(X), and H =
¦f ∈ S
n
: (n)f = n¦. Show H is a subgroup of S
n
which is isomorphic to S
n−1
. Let
g be any permutation on X with (n)g = 1. Find g
−1
Hg.
The next theorem shows that the symmetric groups are incredibly rich and com-
plex.
Theorem (Cayley’s Theorem) Suppose G is a multiplicative group with n
elements and S
n
is the group of all permutations on the set G. Then G is isomorphic
to a subgroup of S
n
.
Proof Let h : G →S
n
be the function which sends a to the bijection h
a
: G →G
deﬁned by (g)h
a
= g a. The proof follows from the following observations.
1) For each given a, h
a
is a bijection from G to G.
2) h is a homomorphism, i.e., h
a·b
= h
a
◦ h
b
.
3) h is injective and thus G is isomorphic to image(h) ⊂ S
n
.
The Symmetric Groups Now let n ≥ 2 and let S
n
be the group of all permu-
tations on ¦1, 2, ..., n¦. The following deﬁnition shows that each element of S
n
may
32 Groups Chapter 2
be represented by a matrix.
Deﬁnition Suppose 1 < k ≤ n, ¦a
1
, a
2
, ..., a
k
¦ is a collection of distinct inte-
gers with 1 ≤ a
i
≤ n, and ¦b
1
, b
2
, ..., b
k
¦ is the same collection in some diﬀerent order.
Then the matrix

a
1
a
2
... a
k
b
1
b
2
... b
k

represents f ∈ S
n
deﬁned by (a
i
)f = b
i
for 1 ≤ i ≤ k,
and (a)f = a for all other a. The composition of two permutations is computed by
applying the matrix on the left ﬁrst and the matrix on the right second.
There is a special type of permutation called a cycle. For these we have a special
notation.
Deﬁnition

a
1
a
2
...a
k−1
a
k
a
2
a
3
...a
k
a
1

is called a k-cycle, and is denoted by (a
1
, a
2
, ..., a
k
).
A 2-cycle is called a transposition. The cycles (a
1
, ..., a
k
) and (c
1
, ..., c

) are disjoint
provided a
i
= c
j
for all 1 ≤ i ≤ k and 1 ≤ j ≤ .
Listed here are eight basic properties of permutations. They are all easy except
4), which takes a little work. Properties 9) and 10) are listed solely for reference.
Theorem
1) Disjoint cycles commute. (This is obvious.)
2) Every nonidentity permutation can be written uniquely (except for order) as
the product of disjoint cycles. (This is easy.)
3) Every permutation can be written (non-uniquely) as the product of transposi-
tions. (Proof: I = (1, 2)(1, 2) and (a
1
, ..., a
k
) = (a
1
, a
2
)(a
1
, a
3
) (a
1
, a
k
). )
4) The parity of the number of these transpositions is unique. This means that if
f is the product of p transpositions and also of q transpositions, then p is
even iﬀ q is even. In this case, f is said to be an even permutation. In the other
case, f is an odd permutation.
5) A k-cycle is even (odd) iﬀ k is odd (even). For example (1, 2, 3) = (1, 2)(1, 3) is
an even permutation.
6) Suppose f, g ∈ S
n
. If one of f and g is even and the other is odd, then g ◦ f is
Chapter 2 Groups 33
odd. If f and g are both even or both odd, then g ◦ f is even. (Obvious.)
7) The map h : S
n
→Z
2
deﬁned by h(even)= [0] and h(odd)= [1] is a
homomorphism from a multiplicative group to an additive group. Its kernel (the
subgroup of even permutations) is denoted by A
n
and is called the alternating
group. Thus A
n
is a normal subgroup of index 2, and S
n
/A
n
≈ Z
2
.
8) If a, b, c and d are distinct integers in ¦1, 2, . . . , n¦, then (a, b)(b, c) = (a, c, b)
and (a, b)(c, d) = (a, c, d)(a, c, b). Since I = (1, 2, 3)
3
, it follows that for
n ≥ 3, every even permutation is the product of 3-cycles.
The following parts are not included in this course. They are presented here merely
for reference.
9) For any n = 4, A
n
is simple, i.e., has no proper normal subgroups.
10) S
n
can be generated by two elements. In fact, ¦(1, 2), (1, 2, ..., n)¦ generates S
n
.
(Of course there are subgroups of S
n
which cannot be generated by two
elements).
Proof of 4) It suﬃces to prove if the product of t transpositions is the identity I
on ¦1, 2, . . . , n¦, then t is even. Suppose this is false and I is written as t transposi-
tions, where t is the smallest odd integer this is possible. Since t is odd, it is at least 3.
Suppose for convenience the ﬁrst transposition is (a, n). We will rewrite I as a prod-
uct of transpositions σ
1
σ
2
σ
t
where (n)σ
i
= (n) for 1 ≤ i < t and (n)σ
t
= n, which
will be a contradiction. This can be done by inductively “pushing n to the right”
using the equations below. If a, b, and c are distinct integers in ¦1, 2, . . . , n − 1¦,
then (a, n)(a, n) = I, (a, n)(b, n) = (a, b)(a, n), (a, n)(a, c) = (a, c)(c, n), and
(a, n)(b, c) = (b, c)(a, n). Note that (a, n)(a, n) cannot occur here because it would
result in a shorter odd product. (Now you may solve the tile puzzle on page viii.)
Exercise
1) Write

1 2 3 4 5 6 7
6 5 4 3 1 7 2

as the product of disjoint cycles.
Write (1,5,6,7)(2,3,4)(3,7,1) as the product of disjoint cycles.
Write (3,7,1)(1,5,6,7)(2,3,4) as the product of disjoint cycles.
Which of these permutations are odd and which are even?
34 Groups Chapter 2
2) Suppose (a
1
, . . . , a
k
) and (c
1
, . . . , c

) are disjoint cycles. What is the order of
their product?
3) Suppose σ ∈ S
n
. Show that σ
−1
(1, 2, 3)σ = ((1)σ, (2)σ, (3)σ). This shows
that conjugation by σ is just a type of relabeling. Also let τ = (4, 5, 6) and
ﬁnd τ
−1
(1, 2, 3, 4, 5)τ.
4) Show that H = ¦σ ∈ S
6
: (6)σ = 6¦ is a subgroup of S
6
and ﬁnd its right
cosets and its left cosets.
5) Let A ⊂ R
2
be the square with vertices (−1, 1), (1, 1), (1, −1), and (−1, −1),
and G be the collection of all isometries of A onto itself. We know from a
previous exercise that G is a group with eight elements. It follows from Cayley’s
theorem that G is isomorphic to a subgroup of S
8
. Show that G is isomorphic
to a subgroup of S
4
.
6) If G is a multiplicative group, deﬁne a new multiplication on the set G by
a ◦ b = b a. In other words, the new multiplication is the old multiplication
in the opposite order. This deﬁnes a new group denoted by G
op
, the opposite
group. Show that it has the same identity and the same inverses as G, and
that f : G →G
op
deﬁned by f(a) = a
−1
is a group isomorphism. Now consider
the special case G = S
n
. The convention used in this section is that an element
of S
n
is a permutation on ¦1, 2, . . . , n¦ with the variable written on the left.
Show that an element of S
op
n
is a permutation on ¦1, 2, . . . , n¦ with the variable
written on the right. (Of course, either S
n
or S
op
n
may be called the symmetric
group, depending on personal preference or context.)
Product of Groups
The product of groups is usually presented for multiplicative groups. It is pre-
sented here for additive groups because this is the form that occurs in later chapters.
As an exercise, this section should be rewritten using multiplicative notation. The
two theorems below are transparent and easy, but quite useful. For simplicity we
ﬁrst consider the product of two groups, although the case of inﬁnite products is only
slightly more diﬃcult. For background, read ﬁrst the two theorems on page 11.
Theorem Suppose G
1
and G
2
1
G
2
by (a
1
, a
2
) +(b
1
, b
2
) = (a
1
+b
1
, a
2
+b
2
). This operation makes G
1
G
2
into a group.
Its “zero” is (0
¯
1
, 0
¯
2
) and −(a
1
, a
2
) = (−a
1
, −a
2
). The projections π
1
: G
1
G
2
→G
1
Chapter 2 Groups 35
and π
2
: G
1
G
2
→G
2
are group homomorphisms. Suppose G is an additive group.
We know there is a bijection from ¦functions f : G →G
1
G
2
¦ to ¦ordered pairs of
functions (f
1
, f
2
) where f
1
: G → G
1
and f
2
: G → G
2
¦. Under this bijection, f is a
group homomorphism iﬀ each of f
1
and f
2
is a group homomorphism.
Proof It is transparent that the product of groups is a group, so let’s prove
the last part. Suppose G, G
1
, and G
2
are groups and f = (f
1
, f
2
) is a function
from G to G
1
G
2
. Now f(a + b) = (f
1
(a + b), f
2
(a + b)) and f(a) + f(b) =
(f
1
(a), f
2
(a)) +(f
1
(b), f
2
(b)) = (f
1
(a) +f
1
(b), f
2
(a) +f
2
(b)). An examination of these
two equations shows that f is a group homomorphism iﬀ each of f
1
and f
2
is a group
homomorphism.
Exercise Suppose G
1
and G
2
are groups. Show that G
1
G
2
and G
2
G
1
are
isomorphic.
Exercise If o(a
1
) = m and o(a
2
) = n, ﬁnd the order of (a
1
, a
2
) in G
1
G
2
.
Exercise Show that if G is any group of order 4, G is isomorphic to Z
4
or Z
2
Z
2
.
Show Z
4
is not isomorphic to Z
2
Z
2
. Show Z
12
is isomorphic to Z
4
Z
3
. Finally,
show that Z
mn
is isomorphic to Z
m
Z
n
iﬀ (m, n) = 1.
Exercise Suppose G
1
and G
2
are groups and i
1
: G
1
→ G
1
G
2
is deﬁned by
i
1
(g
1
) = (g
1
, 0
¯
2
). Show i
1
is an injective group homomorphism and its image is a
normal subgroup of G
1
G
2
. Usually G
1
is identiﬁed with its image under i
1
, so G
1
may be considered to be a normal subgroup of G
1
G
2
. Let π
2
: G
1
G
2
→ G
2
be the projection map deﬁned in the Background chapter. Show π
2
is a surjective
homomorphism with kernel G
1
. Therefore (G
1
G
2
)/G
1
≈ G
2
as you would expect.
Exercise Let R be the reals under addition. Show that the addition in the
product RR is just the usual addition in analytic geometry.
Exercise Suppose n > 2. Is S
n
isomorphic to A
n
G where G is a multiplicative
group of order 2 ?
One nice thing about the product of groups is that it works ﬁne for any ﬁnite
number, or even any inﬁnite number. The next theorem is stated in full generality.
36 Groups Chapter 2
Theorem Suppose T is an index set, and for any t ∈ T, G
t
¸
t∈T
G
t
=
¸
G
t
by ¦a
t
¦ + ¦b
t
¦ = ¦a
t
+ b
t
¦. This op-
eration makes the product into a group. Its “zero” is ¦0
¯
t
¦ and −¦a
t
¦ = ¦−a
t
¦.
Each projection π
s
:
¸
G
t
→ G
s
is a group homomorphism. Suppose G is an ad-
ditive group. Under the natural bijection from ¦functions f : G →
¸
G
t
¦ to
¦sequences of functions ¦f
t
¦
t∈T
where f
t
: G → G
t
¦, f is a group homomorphism
iﬀ each f
t
is a group homomorphism. Finally, the scalar multiplication on
¸
G
t
by integers is given coordinatewise, i.e., ¦a
t
¦n = ¦a
t
n¦.
¸
G
t
is coordinatewise.
Exercise Suppose s is an element of T and π
s
:
¸
G
t
→G
s
is the projection map
deﬁned in the Background chapter. Show π
s
is a surjective homomorphism and ﬁnd
its kernel.
Exercise Suppose s is an element of T and i
s
: G
s

¸
G
t
is deﬁned by i
s
(a) =
¦a
t
¦ where a
t
= 0
¯
if t = s and a
s
= a. Show i
s
is an injective homomorphism
and its image is a normal subgroup of
¸
G
t
. Thus each G
s
may be considered to be
a normal subgroup of
¸
G
t
.
Exercise Let f : Z → Z
30
Z
100
be the homomorphism deﬁned by f(m) =
([4m], [3m]). Find the kernel of f. Find the order of ([4], [3]) in Z
30
Z
100
.
Exercise Let f : Z → Z
90
Z
70
Z
42
be the group homomorphism deﬁned by
f(m) = ([m], [m], [m]). Find the kernel of f and show that f is not surjective. Let
g : Z → Z
45
Z
35
Z
21
be deﬁned by g(m) = ([m], [m], [m]). Find the kernel of
g and determine if g is surjective. Note that the gcd of ¦45, 35, 21¦ is 1. Now let
h : Z → Z
8
Z
9
Z
35
be deﬁned by h(m) = ([m], [m], [m]). Find the kernel of h
and show that h is surjective. Finally suppose each of b, c, and d is greater than 1
and f : Z → Z
b
Z
c
Z
d
is deﬁned by f(m) = ([m], [m], [m]). Find necessary and
suﬃcient conditions for f to be surjective (see the ﬁrst exercise on page 18).
Exercise Suppose T is a non-void set, G is an additive group, and G
T
is the
collection of all functions f : T →G with addition deﬁned by (f +g)(t) = f(t) +g(t).
Show G
T
is a group. For each t ∈ T, let G
t
= G. Note that G
T
is just another way
of writing
¸
t∈T
G
t
. Also note that if T = [0, 1] and G = R, the addition deﬁned on
G
T
is just the usual addition of functions used in calculus. (For the ring and module
versions, see exercises on pages 44 and 69.)
Chapter 3
Rings
Rings are additive abelian groups with a second operation called multiplication. The
connection between the two operations is provided by the distributive law. Assuming
the results of Chapter 2, this chapter ﬂows smoothly. This is because ideals are also
normal subgroups and ring homomorphisms are also group homomorphisms. We do
not show that the polynomial ring F[x] is a unique factorization domain, although
with the material at hand, it would be easy to do. Also there is no mention of prime
or maximal ideals, because these concepts are unnecessary for our development of
linear algebra. These concepts are developed in the Appendix. A section on Boolean
rings is included because of their importance in logic and computer science.
Suppose R is an additive abelian group, R = 0
¯
, and R has a second binary
operation (i.e., map from R R to R) which is denoted by multiplication. Consider
the following properties.
1) If a, b, c ∈ R, (a b) c = a (b c). (The associative property
of multiplication.)
2) If a, b, c ∈ R, a (b +c) = (a b) + (a c) and (b +c) a = (b a) + (c a).
(The distributive law, which connects addition and
multiplication.)
3) R has a multiplicative identity, i.e., there is an element
1
¯
= 1
¯
R
∈ R such that if a ∈ R, a 1
¯
= 1
¯
a = a.
4) If a, b ∈ R, a b = b a. (The commutative property for
multiplication.)
Deﬁnition If 1), 2), and 3) are satisﬁed, R is said to be a ring. If in addition 4)
is satisﬁed, R is said to be a commutative ring.
Examples The basic commutative rings in mathematics are the integers Z, the
37
38 Rings Chapter 3
rational numbers Q, the real numbers R, and the complex numbers C. It will be shown
later that Z
n
, the integers mod n, has a natural multiplication under which it is a
commutative ring. Also if R is any commutative ring, we will deﬁne R[x
1
, x
2
, . . . , x
n
],
a polynomical ring in n variables. Now suppose R is any ring, n ≥ 1, and R
n
is the
collection of all nn matrices over R. In the next chapter, operations of addition and
multiplication of matrices will be deﬁned. Under these operations, R
n
is a ring. This
is a basic example of a non-commutative ring. If n > 1, R
n
is never commutative,
even if R is commutative.
The next two theorems show that ring multiplication behaves as you would wish
it to. They should be worked as exercises.
Theorem Suppose R is a ring and a, b ∈ R.
1) a 0
¯
= 0
¯
a = 0
¯
. Since R = 0
¯
, it follows that 1
¯
= 0
¯
.
2) (−a) b = a (−b) = −(a b).
Recall that, since R is an additive abelian group, it has a scalar multiplication
over Z (page 20). This scalar multiplication can be written on the right or left, i.e.,
na = an, and the next theorem shows it relates nicely to the ring multiplication.
Theorem Suppose a, b ∈ R and n, m ∈ Z.
1) (na) (mb) = (nm)(a b). (This follows from the distributive
law and the previous theorem.)
2) Let n
¯
= n1
¯
. For example, 2
¯
= 1
¯
+ 1
¯
. Then na = n
¯
a, that is, scalar
multiplication by n is the same as ring multiplication by n
¯
.
Of course, n
¯
may be 0
¯
even though n = 0.
Units
Deﬁnition An element a of a ring R is a unit provided ∃ an element a
−1
∈ R
with a a
−1
= a
−1
a = 1
¯
.
Theorem 0
¯
can never be a unit. 1
¯
is always a unit. If a is a unit, a
−1
is also a
unit with (a
−1
)
−1
= a. The product of units is a unit with (a b)
−1
= b
−1
a
−1
. More
Chapter 3 Rings 39
generally, if a
1
, a
2
, ..., a
n
are units, then their product is a unit with (a
1
a
2
a
n
)
−1
=
a
−1
n
a
−1
n−1
a
−1
1
. The set of all units of R forms a multiplicative group denoted by
R

. Finally if a is a unit, (−a) is a unit and (−a)
−1
= −(a
−1
).
In order for a to be a unit, it must have a two-sided inverse. It suﬃces to require
a left inverse and a right inverse, as shown in the next theorem.
Theorem Suppose a ∈ R and ∃ elements b and c with b a = a c = 1
¯
. Then
b = c and so a is a unit with a
−1
= b = c.
Proof b = b 1
¯
= b (a c) = (b a) c = 1
¯
c = c.
Corollary Inverses are unique.
Domains and Fields In order to deﬁne these two types of rings, we ﬁrst consider
the concept of zero divisor.
Deﬁnition Suppose R is a commutative ring. An element a ∈ R is called a zero
divisor provided it is non-zero and ∃ a non-zero element b with a b = 0
¯
. Note that
if a is a unit, it cannot be a zero divisor.
Theorem Suppose R is a commutative ring and a ∈ (R−0
¯
) is not a zero divisor.
Then (a b = a c) ⇒b = c. In other words, multiplication by a is an injective map
from R to R. It is surjective iﬀ a is a unit.
Deﬁnition A domain (or integral domain) is a commutative ring such that, if
a = 0
¯
, a is not a zero divisor. A ﬁeld is a commutative ring such that, if a = 0
¯
, a is
a unit. In other words, R is a ﬁeld if it is commutative and its non-zero elements
form a group under multiplication.
Theorem A ﬁeld is a domain. A ﬁnite domain is a ﬁeld.
Proof A ﬁeld is a domain because a unit cannot be a zero divisor. Suppose R is
a ﬁnite domain and a = 0
¯
. Then f : R → R deﬁned by f(b) = a b is injective and,
by the pigeonhole principle, f is surjective. Thus a is a unit and so R is a ﬁeld.
40 Rings Chapter 3
Exercise Let C be the additive abelian group R
2
. Deﬁne multiplication by
(a, b) (c, d) = (ac − bd, ad + bc). Show C is a commutative ring which is a ﬁeld.
Note that 1
¯
= (1, 0) and if i = (0, 1), then i
2
= −1
¯
.
Examples Z is a domain. Q, R, and C are ﬁelds.
The Integers Mod n
The concept of integers mod n is fundamental in mathematics. It leads to a neat
little theory, as seen by the theorems below. However, the basic theory cannot be
completed until the product of rings is deﬁned. (See the Chinese Remainder Theorem
on page 50.) We know from page 27 that Z
n
Theorem Suppose n > 1. Deﬁne a multiplication on Z
n
by [a] [b] = [ab]. This
is a well deﬁned binary operation which makes Z
n
into a commutative ring.
Proof Since [a +kn] [b +l n] = [ab +n(al +bk +kl n)] = [ab], the multiplication
is well-deﬁned. The ring axioms are easily veriﬁed.
Theorem Suppose n > 1 and a ∈ Z. Then the following are equivalent.
1) [a] is a generator of the additive group Z
n
.
2) (a, n) = 1.
3) [a] is a unit of the ring Z
n
.
Proof We already know from page 27 that 1) and 2) are equivalent. Recall that
if b is an integer, [a]b = [a] [b] = [ab]. Thus 1) and 3) are equivalent, because each
says ∃ an integer b with [a]b = [1].
Corollary If n > 1, the following are equivalent.
1) Z
n
is a domain.
2) Z
n
is a ﬁeld.
3) n is a prime.
Proof We already know 1) and 2) are equivalent, because Z
n
is ﬁnite. Suppose
3) is true. Then by the previous theorem, each of [1], [2],...,[n − 1] is a unit, and
thus 2) is true. Now suppose 3) is false. Then n = ab where 1 < a < n, 1 < b < n,
Chapter 3 Rings 41
[a][b] = [0], and thus [a] is a zero divisor and 1) is false.
Exercise List the units and their inverses for Z
7
and Z
12
. Show that (Z
7
)

is
a cyclic group but (Z
12
)

is not. Show that in Z
12
the equation x
2
= 1
¯
has four
solutions. Finally show that if R is a domain, x
2
= 1
¯
can have at most two solutions
in R (see the ﬁrst theorem on page 46).
Subrings Suppose S is a subset of a ring R. The statement that S is a subring
of R means that S is a subgroup of the group R, 1
¯
∈ S , and (a, b ∈ S ⇒a b ∈ S).
Then clearly S is a ring and has the same multiplicative identity as R. Note that Z
is a subring of Q, Q is a subring of R, and R is a subring of C. Subrings do not play
a role analogous to subgroups. That role is played by ideals, and an ideal is never a
subring (unless it is the entire ring). Note that if S is a subring of R and s ∈ S, then
s may be a unit in R but not in S. Note also that Z and Z
n
have no proper subrings,
and thus occupy a special place in ring theory, as well as in group theory.
Ideals and Quotient Rings
Ideals in ring theory play a role analagous to normal subgroups in group theory.
Deﬁnition A subset I of a ring R is a

left
right
2−sided

ideal provided it is a subgroup
of the additive group R and if a ∈ R and b ∈ I, then

a b ∈ I
b a ∈ I
a b and b a ∈ I

. The
word “ideal ” means “2-sided ideal”. Of course, if R is commutative, every right or
left ideal is an ideal.
Theorem Suppose R is a ring.
1) R and 0
¯
are ideals of R. These are called the improper ideals.
2) If ¦I
t
¦
t∈T
is a collection of right (left, 2-sided) ideals of R, then
¸
t∈T
I
t
is a
right (left, 2-sided) ideal of R. (See page 22.)
42 Rings Chapter 3
3) Furthermore, if the collection is monotonic, then
¸
t∈T
I
t
is a right (left, 2-sided)
ideal of R.
4) If a ∈ R, I = aR is a right ideal. Thus if R is commutative, aR is an ideal,
called a principal ideal. Thus every subgroup of Z is a principal ideal,
because it is of the form nZ.
5) If R is a commutative ring and I ⊂ R is an ideal, then the following are
equivalent.
i) I = R.
ii) I contains some unit u.
iii) I contains 1
¯
.
Exercise Suppose R is a commutative ring. Show that R is a ﬁeld iﬀ R contains
no proper ideals.
The following theorem is just an observation, but it is in some sense the beginning
of ring theory.
Theorem Suppose R is a ring and I ⊂ R is an ideal, I = R. Since I is a normal
subgroup of the additive group R, R/I is an additive abelian group. Multiplication
of cosets deﬁned by (a +I) (b +I) = (ab +I) is well-deﬁned and makes R/I a ring.
Proof (a + I) (b + I) = a b + aI + Ib + II ⊂ a b + I. Thus multiplication
is well deﬁned, and the ring axioms are easily veriﬁed. The multiplicative identity is
(1
¯
+I).
Observation If R = Z, n > 1, and I = nZ, the ring structure on Z
n
= Z/nZ
is the same as the one previously deﬁned.
Homomorphisms
Deﬁnition Suppose R and
¯
R are rings. A function f : R →
¯
R is a ring homo-
morphism provided
1) f is a group homomorphism
2) f(1
¯
R
) = 1
¯
¯
R
and
3) if a, b ∈ R then f(a b) = f(a) f(b). (On the left, multiplication
Chapter 3 Rings 43
is in R, while on the right multiplication is in
¯
R.)
The kernel of f is the kernel of f considered as a group homomorphism, namely
ker(f) = f
−1
(0
¯
).
Here is a list of the basic properties of ring homomorphisms. Much of this
work has already been done by the theorem in group theory on page 28.
Theorem Suppose each of R and
¯
R is a ring.
1) The identity map I
R
: R →R is a ring homomorphism.
2) The zero map from R to
¯
R is not a ring homomorphism
(because it does not send 1
¯
R
to 1
¯
¯
R
).
3) The composition of ring homomorphisms is a ring homomorphism.
4) If f : R →
¯
R is a bijection which is a ring homomorphism,
then f
−1
:
¯
R →R is a ring homomorphism. Such an f is called
a ring isomorphism. In the case R =
¯
R, f is also called a
ring automorphism.
5) The image of a ring homomorphism is a subring of the range.
6) The kernel of a ring homomorphism is an ideal of the domain.
In fact, if f : R →
¯
R is a homomorphism and I ⊂
¯
R is an ideal,
then f
−1
(I) is an ideal of R.
7) Suppose I is an ideal of R, I = R, and π : R →R/I is the
natural projection, π(a) = (a +I). Then π is a surjective ring
homomorphism with kernel I. Furthermore, if f : R →
¯
R is a surjective
ring homomorphism with kernel I, then R/I ≈
¯
R (see below).
8) From now on the word “homomorphism” means “ring homomorphism”.
Suppose f : R →
¯
R is a homomorphism and I is an ideal of R, I = R.
If I ⊂ ker(f), then
¯
f : R/I →
¯
R deﬁned by
¯
f(a +I) = f(a)
44 Rings Chapter 3
is a well-deﬁned homomorphism making the following diagram commute.
R
¯
R
R/I
f
·

π
¯
f
Thus deﬁning a homomorphism on a quotient ring is the same as
deﬁning a homomorphism on the numerator which sends the
denominator to zero. The image of
¯
f is the image of f, and
the kernel of
¯
f is ker(f)/I. Thus if I = ker(f),
¯
f is
injective, and so R/I ≈ image (f).
Proof We know all this on the group level, and it is only necessary
to check that
¯
f is a ring homomorphism, which is obvious.
9) Given any ring homomorphism f, domain(f)/ker(f) ≈ image(f).
Exercise Find a ring R with a proper ideal I and an element b such that b is not
a unit in R but (b +I) is a unit in R/I.
Exercise Show that if u is a unit in a ring R, then conjugation by u is an
automorphism on R. That is, show that f : R →R deﬁned by f(a) = u
−1
a u is
a ring homomorphism which is an isomorphism.
Exercise Suppose T is a non-void set, R is a ring, and R
T
is the collection of
all functions f : T → R. Deﬁne addition and multiplication on R
T
point-wise. This
means if f and g are functions from T to R, then (f + g)(t) = f(t) + g(t) and
(f g)(t) = f(t)g(t). Show that under these operations R
T
is a ring. Suppose S is a
non-void set and α : S →T is a function. If f : T →R is a function, deﬁne a function
α

(f) : S →R by α

(f) = f ◦ α. Show α

: R
T
→R
S
is a ring homomorphism.
Exercise Now consider the case T = [0, 1] and R = R. Let A ⊂ R
[0,1]
be the
collection of all C

functions, i.e., A =¦f : [0, 1] →R : f has an inﬁnite number of
derivatives¦. Show A is a ring. Notice that much of the work has been done in the
previous exercise. It is only necessary to show that A is a subring of the ring R
[0,1]
.
Chapter 3 Rings 45
Polynomial Rings
In calculus, we consider real functions f which are polynomials, f(x) = a
0
+a
1
x +
+a
n
x
n
. The sum and product of polynomials are again polynomials, and it is easy
to see that the collection of polynomial functions forms a commutative ring. We can
do the same thing formally in a purely algebraic setting.
Deﬁnition Suppose R is a commutative ring and x is a “variable” or “symbol”.
The polynomial ring R[x] is the collection of all polynomials f = a
0
+a
1
x + +a
n
x
n
where a
i
∈ R. Under the obvious addition and multiplication, R[x] is a commutative
ring. The degree of a non-zero polynomial f is the largest integer n such that a
n
= 0
¯
,
and is denoted by n = deg(f). If the top term a
n
= 1
¯
, then f is said to be monic.
To be more formal, think of a polynomial a
0
+ a
1
x + as an inﬁnite sequence
(a
0
, a
1
, ...) such that each a
i
∈ R and only a ﬁnite number are non-zero. Then
(a
0
, a
1
, ...) + (b
0
, b
1
, ...) = (a
0
+b
0
, a
1
+b
1
, ...) and
(a
0
, a
1
, ...) (b
0
, b
1
, ...) = (a
0
b
0
, a
0
b
1
+a
1
b
0
, a
0
b
2
+a
1
b
1
+a
2
b
0
, ...).
Note that on the right, the ring multiplication a b is written simply as ab, as is
often done for convenience.
Theorem If R is a domain, R[x] is also a domain.
Proof Suppose f and g are non-zero polynomials. Then deg(f)+deg(g) = deg(fg)
and thus fg is not 0
¯
. Another way to prove this theorem is to look at the bottom
terms instead of the top terms. Let a
i
x
i
and b
j
x
j
be the ﬁrst non-zero terms of f
and g. Then a
i
b
j
x
i+j
is the ﬁrst non-zero term of fg.
Theorem (The Division Algorithm) Suppose R is a commutative ring, f ∈
R[x] has degree ≥ 1 and its top coeﬃcient is a unit in R. (If R is a ﬁeld, the
top coeﬃcient of f will always be a unit.) Then for any g ∈ R[x], ∃! h, r ∈ R[x]
such that g = fh +r with r = 0
¯
or deg(r) < deg(f).
Proof This theorem states the existence and uniqueness of polynomials h and
r. We outline the proof of existence and leave uniqueness as an exercise. Suppose
f = a
0
+ a
1
x + +a
m
x
m
where m ≥ 1 and a
m
is a unit in R. For any g with
deg(g) < m, set h = 0
¯
and r = g. For the general case, the idea is to divide f into g
until the remainder has degree less than m. The proof is by induction on the degree
of g. Suppose n ≥ m and the result holds for any polynomial of degree less than
46 Rings Chapter 3
n. Suppose g is a polynomial of degree n. Now ∃ a monomial bx
t
with t = n − m
and deg(g − fbx
t
) < n. By induction, ∃ h
1
and r with fh
1
+ r = (g − fbx
t
) and
deg(r) < m. The result follows from the equation f(h
1
+bx
t
) +r = g.
Note If r = 0
¯
we say that f divides g. Note that f = x − c divides g iﬀ c is
a root of g, i.e., g(c) = 0
¯
. More generally, x −c divides g with remainder g(c).
Theorem Suppose R is a domain, n > 0, and g(x) = a
0
+ a
1
x + + a
n
x
n
is a
polynomial of degree n with at least one root in R. Then g has at most n roots. Let
c
1
, c
2
, .., c
k
be the distinct roots of g in the ring R. Then ∃ a unique sequence of
positive integers n
1
, n
2
, .., n
k
and a unique polynomial h with no root in R so that
g(x) = (x − c
1
)
n
1
(x − c
k
)
n
k
h(x). (If h has degree 0, i.e., if h = a
n
, then we say
“all the roots of g belong to R”. If g = a
n
x
n
, we say “all the roots of g are 0
¯
”.)
Proof Uniqueness is easy so let’s prove existence. The theorem is clearly true
for n = 1. Suppose n > 1 and the theorem is true for any polynomial of degree less
than n. Now suppose g is a polynomial of degree n and c
1
is a root of g. Then ∃
a polynomial h
1
with g(x) = (x − c
1
)h
1
. Since h
1
has degree less than n, the result
follows by induction.
Note If g is any non-constant polynomial in C[x], all the roots of g belong to C,
i.e., C is an algebraically closed ﬁeld. This is called The Fundamental Theorem of
Algebra, and it is assumed without proof for this textbook.
Exercise Suppose g is a non-constant polynomial in R[x]. Show that if g has
odd degree then it has a real root. Also show that if g(x) = x
2
+ bx + c, then it has
a real root iﬀ b
2
≥ 4c, and in that case both roots belong to R.
Deﬁnition A domain T is a principal ideal domain (PID) if, given any ideal I,
∃ t ∈ T such that I = tT. Note that Z is a PID and any ﬁeld is PID.
Theorem Suppose F is a ﬁeld, I is a proper ideal of F[x], and n is the smallest
positive integer such that I contains a polynomial of degree n. Then I contains a
unique polynomial of the form f = a
0
+ a
1
x + +a
n−1
x
n−1
+ x
n
and it has the
property that I = fF[x]. Thus F[x] is a PID. Furthermore, each coset of I can be
written uniquely in the form (c
0
+c
1
x + +c
n−1
x
n−1
+I).
Proof. This is a good exercise in the use of the division algorithm. Note this is
similar to showing that a subgroup of Z is generated by one element (see page 15).
Chapter 3 Rings 47
Theorem. Suppose R is a subring of a commutative ring C and c ∈ C. Then
∃! homomorphism h : R[x] → C with h(x) = c and h(r) = r for all r ∈ R. It is
deﬁned by h(a
0
+ a
1
x + +a
n
x
n
) = a
0
+ a
1
c + +a
n
c
n
, i.e., h sends f(x) to f(c).
The image of h is the smallest subring of C containing R and c.
This map h is called an evaluation map. The theorem says that adding two
polynomials in R[x] and evaluating is the same as evaluating and then adding in C.
Also multiplying two polynomials in R[x] and evaluating is the same as evaluating
and then multiplying in C. In street language the theorem says you are free to send
x wherever you wish and extend to a ring homomorphism on R[x].
Exercise Let C = ¦a + bi : a, b ∈ R¦. Since R is a subring of C, there exists a
homomorphism h : R[x] → C which sends x to i, and this h is surjective. Show
ker(h) = (x
2
+ 1)R[x] and thus R[x]/(x
2
+ 1) ≈ C. This is a good way to look
at the complex numbers, i.e., to obtain C, adjoin x to R and set x
2
= −1.
Exercise Z
2
[x]/(x
2
+ x + 1) has 4 elements. Write out the multiplication table
for this ring and show that it is a ﬁeld.
Exercise Show that, if R is a domain, the units of R[x] are just the units of R.
Thus if F is a ﬁeld, the units of F[x] are the non-zero constants. Show that [1] +[2]x
is a unit in Z
4
[x].
In this chapter we do not prove F[x] is a unique factorization domain, nor do
we even deﬁne unique factorization domain. The next deﬁnition and theorem are
included merely for reference, and should not be studied at this stage.
Deﬁnition Suppose F is a ﬁeld and f ∈ F[x] has degree ≥ 1. The statement
that g is an associate of f means ∃ a unit u ∈ F[x] such that g = uf. The statement
that f is irreducible means that if h is a non-constant polynomial which divides f,
then h is an associate of f.
We do not develop the theory of F[x] here. However, the development is easy
because it corresponds to the development of Z in Chapter 1. The Division Algo-
rithm corresponds to the Euclidean Algorithm. Irreducible polynomials correspond
to prime integers. The degree function corresponds to the absolute value function.
One diﬀerence is that the units of F[x] are non-zero constants, while the units of Z
48 Rings Chapter 3
are just ±1. Thus the associates of f are all cf with c = 0
¯
while the associates of an
integer n are just ±n. Here is the basic theorem. (This theory is developed in full in
the Appendix under the topic of Euclidean domains.)
Theorem Suppose F is a ﬁeld and f ∈ F[x] has degree ≥ 1. Then f factors as the
product of irreducibles, and this factorization is unique up to order and associates.
Also the following are equivalent.
1) F[x]/(f) is a domain.
2) F[x]/(f) is a ﬁeld.
3) f is irreducible.
Deﬁnition Now suppose x and y are “variables”. If a ∈ R and n, m ≥ 0, then
ax
n
y
m
= ay
m
x
n
is called a monomial. Deﬁne an element of R[x, y] to be any ﬁnite
sum of monomials.
Theorem R[x, y] is a commutative ring and (R[x])[y] ≈ R[x, y] ≈ (R[y])[x]. In
other words, any polynomial in x and y with coeﬃcients in R may be written as a
polynomial in y with coeﬃcients in R[x], or as a polynomial in x with coeﬃcients in
R[y].
Side Comment It is true that if F is a ﬁeld, each f ∈ F[x, y] factors as the
product of irreducibles. However F[x, y] is not a PID. For example, the ideal
I = xF[x, y] +yF[x, y] = ¦f ∈ F[x, y] : f(0
¯
, 0
¯
) = 0
¯
¦ is not principal.
If R is a commutative ring and n ≥ 2, the concept of a polynomial ring in
n variables works ﬁne without a hitch. If a ∈ R and v
1
, v
2
, ..., v
n
are non-negative
integers, then ax
v
1
1
x
v
2
2
x
vn
n
is called a monomial. Order does not matter here.
Deﬁne an element of R[x
1
, x
2
, ..., x
n
] to be any ﬁnite sum of monomials. This
gives a commutative ring and there is canonical isomorphism R[x
1
, x
2
, ..., x
n
] ≈
(R[x
1
, x
2
, ..., x
n−1
])[x
n
]. Using this and induction on n, it is easy to prove the fol-
lowing theorem.
Theorem If R is a domain, R[x
1
, x
2
, ..., x
n
] is a domain and its units are just the
units of R.
Chapter 3 Rings 49
Exercise Suppose R is a commutative ring and f : R[x, y] → R[x] is the eval-
uation map which sends y to 0
¯
. This means f(p(x, y)) = p(x, 0
¯
). Show f is a ring
homomorphism whose kernel is the ideal (y) = yR[x, y]. Use the fact that “the do-
main mod the kernel is isomorphic to the image” to show R[x, y]/(y) is isomorphic
to R[x]. That is, if you adjoin y to R[x] and then factor it out, you get R[x] back.
Product of Rings
The product of rings works ﬁne, just as does the product of groups.
Theorem Suppose T is an index set and for each t ∈ T, R
t
is a ring. On the
¸
t∈T
R
t
=
¸
R
t
, deﬁne multiplication by ¦r
t
¦ ¦s
t
¦ = ¦r
t
s
t
¦.
Then
¸
R
t
is a ring and each projection π
s
:
¸
R
t
→ R
s
is a ring homomorphism.
Suppose R is a ring. Under the natural bijection from ¦functions f : R →
¸
R
t
¦
to ¦sequences of functions ¦f
t
¦
t∈T
where f
t
: R → R
t
¦, f is a ring homomorphism
iﬀ each f
t
is a ring homomorphism.
Proof We already know f is a group homomorphism iﬀ each f
t
is a group homo-
morphism (see page 36). Note that ¦1
¯
t
¦ is the multiplicative identity of
¸
R
t
, and
f(1
¯
R
) = ¦1
¯
t
¦ iﬀ f
t
(1
¯
R
) = 1
¯
t
for each t ∈ T. Finally, since multiplication is deﬁned
coordinatewise, f is a ring homomorphism iﬀ each f
t
is a ring homomorphism.
Exercise Suppose R and S are rings. Note that R 0 is not a subring of R S
because it does not contain (1
¯
R
, 1
¯
S
). Show R0
¯
is an ideal and (RS/R0
¯
) ≈ S.
Suppose I ⊂ R and J ⊂ S are ideals. Show I J is an ideal of RS and every
ideal of R S is of this form.
Exercise Suppose R and S are commutative rings. Show T = R S is not a
domain. Let e = (1, 0) ∈ RS and show e
2
= e, (1 −e)
2
= (1 −e), R0 = eT,
and 0 S = (1 −e)T.
Exercise If T is any ring, an element e of T is called an idempotent provided
e
2
= e. The elements 0 and 1 are idempotents called the trivial idempotents. Suppose
T is a commutative ring and e ∈ T is an idempotent with 0 = e = 1. Let R = eT
and S = (1 − e)T. Show each of the ideals R and S is a ring with identity, and
f : T →RS deﬁned by f(t) = (et, (1−e)t) is a ring isomorphism. This shows that
a commutative ring T splits as the product of two rings iﬀ it contains a non-trivial
idempotent.
50 Rings Chapter 3
The Chinese Remainder Theorem
The natural map from Z to Z
m
Z
n
is a group homomorphism and also a ring
homomorphism. If m and n are relatively prime, this map is surjective with kernel
mnZ, and thus Z
mn
and Z
m
Z
n
are isomorphic as groups and as rings. The next
theorem is a classical generalization of this. (See exercise three on page 35.)
Theorem Suppose n
1
, ..., n
t
are integers, each n
i
> 1, and (n
i
, n
j
) = 1 for all
i = j. Let f
i
: Z → Z
n
i
be deﬁned by f
i
(a) = [a]. (Note that the bracket symbol is
used ambiguously.) Then the ring homomorphism f = (f
1
, .., f
t
) : Z →Z
n
1
Z
nt
is surjective. Furthermore, the kernel of f is nZ, where n = n
1
n
2
n
t
. Thus Z
n
and Z
n
1
Z
nt
are isomorphic as rings, and thus also as groups.
Proof We wish to show that the order of f(1) is n, and thus f(1) is a group
generator, and thus f is surjective. The element f(1)m = ([1], .., [1])m = ([m], .., [m])
is zero iﬀ m is a multiple of each of n
1
, .., n
t
. Since their least common multiple is n,
the order of f(1) is n. (See the fourth exercise on page 36 for the case t = 3.)
Exercise Show that if a is an integer and p is a prime, then [a] = [a
p
] in Z
p
(Fermat’s Little Theorem). Use this and the Chinese Remainder Theorem to show
that if b is a positive integer, it has the same last digit as b
5
.
Characteristic
The following theorem is just an observation, but it shows that in ring theory, the
ring of integers is a “cornerstone”.
Theorem If R is a ring, there is one and only one ring homomorphism f : Z →R.
It is given by f(m) = m1
¯
= m
¯
. Thus the subgroup of R generated by 1
¯
is a subring
of R isomorphic to Z or isomorphic to Z
n
for some positive integer n.
Deﬁnition Suppose R is a ring and f : Z → R is the natural ring homomor-
phism f(m) = m1
¯
= m
¯
. The non-negative integer n with ker(f) = nZ is called the
characteristic of R. Thus f is injective iﬀ R has characteristic 0 iﬀ 1
¯
has inﬁnite
order. If f is not injective, the characteristic of R is the order of 1
¯
.
It is an interesting fact that, if R is a domain, all the non-zero elements of R
have the same order. (See page 23 for the deﬁnition of order.)
Chapter 3 Rings 51
Theorem Suppose R is a domain. If R has characteristic 0, then each non-zero
a ∈ R has inﬁnite order. If R has ﬁnite characteristic n, then n is a prime and each
non-zero a ∈ R has order n.
Proof Suppose R has characteristic 0, a is a non-zero element of R, and m is a
positive integer. Then ma = m
¯
a cannot be 0
¯
because m
¯
, a = 0
¯
and R is a domain.
Thus o(a) = ∞. Now suppose R has characteristic n. Then R contains Z
n
as a
subring, and thus Z
n
is a domain and n is a prime. If a is a non-zero element of R,
na = n
¯
a = 0
¯
a = 0
¯
and thus o(a)[n and thus o(a) = n.
Exercise Show that if F is a ﬁeld of characteristic 0, F contains Q as a subring.
That is, show that the injective homomorphism f : Z → F extends to an injective
homomorphism
¯
f : Q →F.
Boolean Rings
This section is not used elsewhere in this book. However it ﬁts easily here, and is
included for reference.
Deﬁnition A ring R is a Boolean ring if for each a ∈ R, a
2
= a, i.e., each
element of R is an idempotent.
Theorem Suppose R is a Boolean ring.
1) R has characteristic 2. If a ∈ R, 2a = a +a = 0
¯
, and so a = −a.
Proof (a +a) = (a +a)
2
= a
2
+ 2a
2
+a
2
= 4a. Thus 2a = 0
¯
.
2) R is commutative.
Proof (a +b) = (a +b)
2
= a
2
+ (a b) + (b a) +b
2
= a + (a b) −(b a) +b. Thus a b = b a.
3) If R is a domain, R ≈ Z
2
.
Proof Suppose a = 0
¯
. Then a (1
¯
−a) = 0
¯
and so a = 1
¯
.
4) The image of a Boolean ring is a Boolean ring. That is, if I is an ideal
of R with I = R, then every element of R/I is idempotent and thus
R/I is a Boolean ring. It follows from 3) that R/I is a domain iﬀ R/I
is a ﬁeld iﬀ R/I ≈ Z
2
. (In the language of Chapter 6, I is a prime
ideal iﬀ I is a maximal ideal iﬀ R/I ≈ Z
2
).
52 Rings Chapter 3
Suppose X is a non-void set. If a is a subset of X, let a

= (X−a) be a complement
of a in X. Now suppose R is a non-void collection of subsets of X. Consider the
following properties which the collection R may possess.
1) a ∈ R ⇒ a

∈ R.
2) a, b ∈ R ⇒ (a ∩ b) ∈ R.
3) a, b ∈ R ⇒ (a ∪ b) ∈ R.
4) ∅ ∈ R and X ∈ R.
Theorem If 1) and 2) are satisﬁed, then 3) and 4) are satisﬁed. In this case, R
is called a Boolean algebra of sets.
Proof Suppose 1) and 2) are true, and a, b ∈ R. Then a ∪b = (a

∩b

)

belongs to
R and so 3) is true. Since R is non-void, it contains some element a. Then ∅ = a ∩a

and X = a ∪ a

belong to R, and so 4) is true.
Theorem Suppose R is a Boolean algebra of sets. Deﬁne an addition on R by
a +b = (a ∪ b) −(a ∩ b). Under this addition, R is an abelian group with 0
¯
= ∅ and
a = −a. Deﬁne a multiplication on R by a b = a ∩ b. Under this multiplication R
becomes a Boolean ring with 1
¯
= X.
Exercise Let X = ¦1, 2, ..., n¦ and let R be the Boolean ring of all subsets of
X. Note that o(R) = 2
n
. Deﬁne f
i
: R → Z
2
by f
i
(a) = [1] iﬀ i ∈ a. Show each
f
i
is a homomorphism and thus f = (f
1
, ..., f
n
) : R → Z
2
Z
2
Z
2
is a ring
homomorphism. Show f is an isomorphism. (See exercises 1) and 4) on page 12.)
Exercise Use the last exercise on page 49 to show that any ﬁnite Boolean ring is
isomorphic to Z
2
Z
2
Z
2
, and thus also to the Boolean ring of subsets above.
Note Suppose R is a Boolean ring. It is a classical theorem that ∃ a Boolean
algebra of sets whose Boolean ring is isomorphic to R. So let’s just suppose R is
a Boolean algebra of sets which is a Boolean ring with addition and multiplication
deﬁned as above. Now deﬁne a ∨ b = a ∪ b and a ∧ b = a ∩ b. These operations cup
and cap are associative, commutative, have identity elements, and each distributes
over the other. With these two operations (along with complement), R is called a
Boolean algebra. R is not a group under cup or cap. Anyway, it is a classical fact
that, if you have a Boolean ring (algebra), you have a Boolean algebra (ring). The
advantage of the algebra is that it is symmetric in cup and cap. The advantage of
the ring viewpoint is that you can draw from the rich theory of commutative rings.
Chapter 4
Matrices and Matrix Rings
We ﬁrst consider matrices in full generality, i.e., over an arbitrary ring R. However,
after the ﬁrst few pages, it will be assumed that R is commutative. The topics,
such as invertible matrices, transpose, elementary matrices, systems of equations,
and determinant, are all classical. The highlight of the chapter is the theorem that a
square matrix is a unit in the matrix ring iﬀ its determinant is a unit in the ring.
This chapter concludes with the theorem that similar matrices have the same deter-
minant, trace, and characteristic polynomial. This will be used in the next chapter
to show that an endomorphism on a ﬁnitely generated vector space has a well-deﬁned
determinant, trace, and characteristic polynomial.
Deﬁnition Suppose R is a ring and m and n are positive integers. Let R
m,n
be
the collection of all mn matrices
A = (a
i,j
) =

¸
¸
¸
a
1,1
. . . a
1,n
.
.
.
.
.
.
a
m,1
. . . a
m,n
¸

where each entry a
i,j
∈ R.
A matrix may be viewed as m n-dimensional row vectors or as n m-dimensional
column vectors. A matrix is said to be square if it has the same number of rows
as columns. Square matrices are so important that they have a special notation,
R
n
= R
n,n
. R
n
is deﬁned to be the additive abelian group R R R.
To emphasize that R
n
does not have a ring structure, we use the “sum” notation,
R
n
= R⊕R⊕ ⊕R. Our convention is to write elements of R
n
as column vectors,
i.e., to identify R
n
with R
n,1
. If the elements of R
n
are written as row vectors, R
n
is
identiﬁed with R
1,n
.
53
54 Matrices Chapter 4
Addition of matrices To “add” two matrices, they must have the same number
of rows and the same number of columns, i.e., addition is a binary operation R
m,n

R
m,n
→R
m,n
. The addition is deﬁned by (a
i,j
) +(b
i,j
) = (a
i,j
+b
i,j
), i.e., the i, j term
of the sum is the sum of the i, j terms. The following theorem is just an observation.
Theorem R
m,n
is an additive abelian group. Its “zero” is the matrix 0 = 0
m,n
all of whose terms are zero. Also −(a
i,j
) = (−a
i,j
R
m,n
≈ R
mn
.
Scalar multiplication An element of R is called a scalar. A matrix may be
“multiplied” on the right or left by a scalar. Right scalar multiplication is deﬁned
by (a
i,j
)c = (a
i,j
c). It is a function R
m,n
R → R
m,n
. Note in particular that
scalar multiplication is deﬁned on R
n
. Of course, if R is commutative, there is no
distinction between right and left scalar multiplication.
Theorem Suppose A, B ∈ R
m,n
and c, d ∈ R. Then
(A+B)c = Ac +Bc
A(cd) = (Ac)d
and A1 = A
This theorem is entirely transparent. In the language of the next chapter, it merely
states that R
m,n
is a right module over the ring R.
Multiplication of Matrices The matrix product AB is deﬁned iﬀ the number
of columns of A is equal to the number of rows of B. The matrix AB will have the
same number of rows as A and the same number of columns as B, i.e., multiplication
is a function R
m,n
R
n,p
→R
m,p
. The product (a
i,j
)(b
i,j
) is deﬁned to be the matrix
whose (s, t) term is a
s,1
b
1,t
+ + a
s,n
b
n,t
, i.e., the dot product of row s of A
with column t of B.
Exercise Consider real matrices A =

a b
c d

, U =

2 0
0 1

, V =

0 1
1 0

,
and W =

1 2
0 1

. Find the matrices AU, UA, AV, VA, AW, and WA.
Chapter 4 Matrices 55
Deﬁnition The identity matrix I
n
∈ R
n
is the square matrix whose diagonal terms
are 1 and whose oﬀ-diagonal terms are 0.
Theorem Suppose A ∈ R
m,n
.
1) 0
p,m
A = 0
p,n
A0
n,p
= 0
m,p
2) I
m
A = A = AI
n
Theorem (The distributive laws) (A+B)C = AC +BC and
C(A+B) = CA +CB whenever the
operations are deﬁned.
Theorem (The associative law for matrix multiplication) Suppose A ∈ R
m,n
,
B ∈ R
n,p
, and C ∈ R
p,q
. Then (AB)C = A(BC). Note that ABC ∈ R
m,q
.
Proof We must show that the (s, t) terms are equal. The proof involves writing
it out and changing the order of summation. Let (x
i,j
) = AB and (y
i,j
) = BC.
Then the (s, t) term of (AB)C is
¸
i
x
s,i
c
i,t
=
¸
i

¸
j
a
s,j
b
j,i

c
i,t
=
¸
i,j
a
s,j
b
j,i
c
i,t
=
¸
j
a
s,j

¸
i
b
j,i
c
i,t

=
¸
j
a
s,j
y
j,t
which is the (s, t) term of A(BC).
Theorem For each ring R and integer n ≥ 1, R
n
is a ring.
Proof This elegant little theorem is immediate from the theorems above. The
units of R
n
are called invertible or non-singular matrices. They form a group under
multiplication called the general linear group and denoted by GL
n
(R) = (R
n
)

.
Exercise Recall that if A is a ring and a ∈ A, then aA is right ideal of A. Let
A = R
2
and a = (a
i,j
) where a
1,1
= 1 and the other entries are 0. Find aR
2
and R
2
a.
Show that the only ideal of R
2
containing a is R
2
itself.
Multiplication by blocks Suppose A, E ∈ R
n
, B, F ∈ R
n,m
, C, G ∈ R
m,n
, and
D, H ∈ R
m
. Then multiplication in R
n+m
is given by

A B
C D

E F
G H

=

AE +BG AF +BH
CE +DG CF +DH

.
56 Matrices Chapter 4
Transpose
Notation For the remainder of this chapter on matrices, suppose R is a commu-
tative ring. Of course, for n > 1, R
n
is non-commutative.
Transpose is a function from R
m,n
to R
n,m
. If A ∈ R
m,n
, A
t
∈ R
n,m
is the matrix
whose (i, j) term is the (j, i) term of A. So row i (column i) of A becomes column
i (row i) of A
t
. If A is an n-dimensional row vector, then A
t
is an n-dimensional
column vector. If A is a square matrix, A
t
is also square.
Theorem 1) (A
t
)
t
= A
2) (A+B)
t
= A
t
+B
t
3) If c ∈ R, (Ac)
t
= A
t
c
4) (AB)
t
= B
t
A
t
5) If A ∈ R
n
, then A is invertible iﬀ A
t
is invertible.
In this case (A
−1
)
t
= (A
t
)
−1
.
Proof of 5) Suppose A is invertible. Then I = I
t
= (AA
−1
)
t
= (A
−1
)
t
A
t
.
Exercise Characterize those invertible matrices A ∈ R
2
which have A
−1
= A
t
.
Show that they form a subgroup of GL
2
(R).
Triangular Matrices
If A ∈ R
n
, then A is upper (lower) triangular provided a
i,j
= 0 for all i > j (all
j > i). A is strictly upper (lower) triangular provided a
i,j
= 0 for all i ≥ j (all j ≥ i).
A is diagonal if it is upper and lower triangular, i.e., a
i,j
= 0 for all i = j. Note
that if A is upper (lower) triangular, then A
t
is lower (upper) triangular.
Theorem If A ∈ R
n
is strictly upper (or lower) triangular, then A
n
= 0.
Proof The way to understand this is just multiply it out for n = 2 and n = 3.
The geometry of this theorem will become transparent later in Chapter 5 when the
matrix A deﬁnes an R-module endomorphism on R
n
(see page 93).
Deﬁnition If T is any ring, an element t ∈ T is said to be nilpotent provided ∃n
such that t
n
= 0. In this case, (1 − t) is a unit with inverse 1 + t + t
2
+ + t
n−1
.
Thus if T = R
n
and B is a nilpotent matrix, I −B is invertible.
Chapter 4 Matrices 57
Exercise Let R = Z. Find the inverse of

¸
¸
1 2 −3
0 1 4
0 0 1
¸

.
Exercise Suppose A =

¸
¸
¸
¸
¸
¸
¸
a
1
a
2
0

0
a
n
¸

is a diagonal matrix, B ∈ R
m,n
,
and C ∈ R
n,p
. Show that BA is obtained from B by multiplying column i of B
by a
i
. Show AC is obtained from C by multiplying row i of C by a
i
. Show A is a
unit in R
n
iﬀ each a
i
is a unit in R.
Scalar matrices A scalar matrix is a diagonal matrix for which all the diagonal
terms are equal, i.e., a matrix of the form cI
n
. The map R → R
n
which sends c to
cI
n
is an injective ring homomorphism, and thus we may consider R to be a subring
of R
n
. Multiplying by a scalar is the same as multiplying by a scalar matrix, and
thus scalar matrices commute with everything, i.e., if B ∈ R
n
, (cI
n
)B = cB = Bc =
B(cI
n
). Recall we are assuming R is a commutative ring.
Exercise Suppose A ∈ R
n
and for each B ∈ R
n
, AB = BA. Show A is a scalar
matrix. For n > 1, this shows how non-commutative R
n
is.
Elementary Operations and Elementary Matrices
Suppose R is a commutative ring and A is a matrix over R. There are 3 types of
elementary row and column operations on the matrix A. A need not be square.
Type 1 Multiply row i by some Multiply column i by some
unit a ∈ R. unit a ∈ R.
Type 2 Interchange row i and row j. Interchange column i and column j.
Type 3 Add a times row j Add a times column i
to row i where i = j and a to column j where i = j and a
is any element of R. is any element of R.
58 Matrices Chapter 4
Elementary Matrices Elementary matrices are square and invertible. There
are three types. They are obtained by performing row or column operations on the
identity matrix.
Type 1 B =

¸
¸
¸
¸
¸
¸
¸
¸
¸
1
1 0
a
1
0 1
1
¸

where a is a unit in R.
Type 2 B =

¸
¸
¸
¸
¸
¸
¸
¸
¸
1
0 1
1
1
1 0
1
¸

Type 3 B =

¸
¸
¸
¸
¸
¸
¸
¸
¸
1
1 a
i,j
1
1
0 1
1
¸

where i = j and a
i,j
is
any element of R.
In type 1, all the oﬀ-diagonal elements are zero. In type 2, there are two non-zero
oﬀ-diagonal elements. In type 3, there is at most one non-zero oﬀ-diagonal element,
and it may be above or below the diagonal.
Exercise Show that if B is an elementary matrix of type 1,2, or 3, then B is
invertible and B
−1
is an elementary matrix of the same type.
The following theorem is handy when working with matrices.
Theorem Suppose A is a matrix. It need not be square. To perform an elemen-
tary row (column) operation on A, perform the operation on an identity matrix to
obtain an elementary matrix B, and multiply on the left (right). That is, BA = row
operation on A and AB = column operation on A. (See the exercise on page 54.)
Chapter 4 Matrices 59
Exercise Suppose F is a ﬁeld and A ∈ F
m,n
.
1) Show ∃ invertible matrices B ∈ F
m
and C ∈ F
n
such that BAC = (d
i,j
)
where d
1,1
= = d
t,t
= 1 and all other entries are 0. The integer t is
called the rank of A. (See page 89 of Chapter 5.)
2) Suppose A ∈ F
n
is invertible. Show A is the product of elementary
matrices.
3) A matrix T is said to be in row echelon form if, for each 1 ≤ i < m, the
ﬁrst non-zero term of row (i + 1) is to the right of the ﬁrst non-zero
term of row i. Show ∃ an invertible matrix B ∈ F
m
such that BA is in
row echelon form.
4) Let A =

3 11
0 4

and D =

3 11
1 4

. Write A and D as products
of elementary matrices over Q. Is it possible to write them as products
of elementary matrices over Z?
For 1), perform row and column operations on A to reach the desired form. This
shows the matrices B and C may be selected as products of elementary matrices.
Part 2) also follows from this procedure. For part 3), use only row operations. Notice
that if T is in row-echelon form, the number of non-zero rows is the rank of T.
Systems of Equations
Suppose A = (a
i,j
) ∈ R
m,n
and C =

¸
¸
¸
¸
c
1

c
m
¸

∈ R
m
= R
m,1
. The system
a
1,1
x
1
+ + a
1,n
x
n
= c
1
.
.
.
.
.
.
.
.
.
a
m,1
x
1
+ +a
m,n
x
n
= c
m
of m equations in n unknowns, can be written as one
matrix equation in one unknown, namely as (a
i,j
)

¸
¸
¸
¸
x
1

x
n
¸

=

¸
¸
¸
¸
c
1

c
m
¸

or AX = C.
60 Matrices Chapter 4
Deﬁne f : R
n
→R
m
by f(D) = AD. Then f is a group homomorphism and also
f(Dc) = f(D)c for any c ∈ R. In the language of the next chapter, this says that
f is an R-module homomorphism. The next theorem summarizes what we already
know about solutions of linear equations in this setting.
Theorem
1) AX = 0 is called the homogeneous equation. Its solution set is ker(f).
2) AX = C has a solution iﬀ C ∈ image(f). If D ∈ R
n
is one
solution, the solution set f
−1
(C) is the coset D + ker(f) in R
n
.
(See part 7 of the theorem on homomorphisms in Chapter 2, page 28.)
3) Suppose B ∈ R
m
is invertible. Then AX = C and (BA)X = BC have
the same set of solutions. Thus we may perform any row operation
on both sides of the equation and not change the solution set.
4) If m = n and A ∈ R
m
is invertible, then AX = C has the unique
solution X = A
−1
C.
The geometry of systems of equations over a ﬁeld will not become really trans-
parent until the development of linear algebra in Chapter 5.
Determinants
The concept of determinant is one of the most amazing in all of mathematics.
The proper development of this concept requires a study of multilinear forms, which
is given in Chapter 6. In this section we simply present the basic properties.
For each n ≥ 1 and each commutative ring R, determinant is a function from R
n
to R. For n = 1, [ (a) [ = a. For n = 2,

a b
c d

Deﬁnition Let A = (a
i,j
) ∈ R
n
. If σ is a permutation on ¦1, 2, ..., n¦, let sign(σ) =
1 if σ is an even permutation, and sign(σ) = −1 if σ is an odd permutation. The
determinant is deﬁned by [ A [=
¸
all σ
sign(σ) a
1,σ(1)
a
2,σ(2)
a
n,σ(n)
. Check that for
n = 2, this agrees with the deﬁnition above. (Note that here we are writing the
permutation functions as σ(i) and not as (i)σ.)
Chapter 4 Matrices 61
For each σ, a
1,σ(1)
a
2,σ(2)
a
n,σ(n)
contains exactly one factor from each row and
one factor from each column. Since R is commutative, we may rearrange the factors
so that the ﬁrst comes from the ﬁrst column, the second from the second column, etc.
This means that there is a permutation τ on ¦1, 2, . . . , n¦ such that a
1,σ(1)
a
n,σ(n)
=
a
τ(1),1
a
τ(n),n
. We wish to show that τ = σ
−1
and thus sign(σ) = sign(τ). To
reduce the abstraction, suppose σ(2) = 5. Then the ﬁrst expression will contain
the factor a
2,5
. In the second expression, it will appear as a
τ(5),5
, and so τ(5) = 2.
Anyway, τ is the inverse of σ and thus there are two ways to deﬁne determinant. It
follows that the determinant of a matrix is equal to the determinant of its transpose.
Theorem [A[ =
¸
all σ
sign(σ)a
1,σ(1)
a
2,σ(2)
a
n,σ(n)
=
¸
all τ
sign(τ)a
τ(1),1
a
τ(2),2
a
τ(n),n
.
Corollary [A[ = [A
t
[.
You may view an n n matrix A as a sequence of n column vectors or as a
sequence of n row vectors. Here we will use column vectors. This means we write the
matrix A as A = (A
1
, A
2
, . . . , A
n
) where each A
i
∈ R
n,1
= R
n
.
Theorem If two columns of A are equal, then [A[ = 0
¯
.
Proof For simplicity, assume the ﬁrst two columns are equal, i.e., A
1
= A
2
.
Now [A[ =
¸
all τ
sign(τ)a
τ(1),1
a
τ(2),2
a
τ(n),n
and this summation has n! terms and
n! is an even number. Let γ be the transposition which interchanges one and two.
Then for any τ, a
τ(1),1
a
τ(2),2
a
τ(n),n
= a
τγ(1),1
a
τγ(2),2
a
τγ(n),n
. This pairs up
the n! terms of the summation, and since sign(τ)=−sign(τγ), these pairs cancel
in the summation. Therefore [A[ = 0
¯
.
Theorem Suppose 1 ≤ r ≤ n, C
r
∈ R
n,1
, and a, c ∈ R. Then [(A
1
, . . . , A
r−1
,
aA
r
+cC
r
, A
r+1
, . . . , A
n
)[ = a[(A
1
, . . . , A
n
)[ + c[(A
1
, . . . , A
r−1
, C
r
, A
r+1
, . . . , A
n
)[
Proof This is immediate from the deﬁnition of determinant and the distributive
law of multiplication in the ring R.
Summary Determinant is a function d : R
n
→ R. In the language used in the
Appendix, the two previous theorems say that d is an alternating multilinear form.
The next two theorems show that alternating implies skew-symmetric (see page 129).
62 Matrices Chapter 4
Theorem Interchanging two columns of A multiplies the determinant by minus
one.
Proof For simplicity, show that [(A
2
, A
1
, A
3
, . . . , A
n
)[ = −[A[. We know 0
¯
=
[(A
1
+ A
2
, A
1
+ A
2
, A
3
, . . . , A
n
)[ = [(A
1
, A
1
, A
3
, . . . , A
n
)[ + [(A
1
, A
2
, A
3
, . . . , A
n
)[ +
[(A
2
, A
1
, A
3
, . . . , A
n
)[ + [(A
2
, A
2
, A
3
, . . . , A
n
)[. Since the ﬁrst and last of these four
terms are zero, the result follows.
Theorem If τ is a permutation of (1, 2, . . . , n), then
[A[ = sign(τ)[(A
τ(1)
, A
τ(2)
, . . . , A
τ(n)
)[.
Proof The permutation τ is the ﬁnite product of transpositions.
Exercise Rewrite the four preceding theorems using rows instead of columns.
The following theorem is just a summary of some of the work done so far.
Theorem Multiplying any row or column of matrix by a scalar c ∈ R, multiplies
the determinant by c. Interchanging two rows or two columns multiplies the determi-
nant by −1. Adding c times one row to another row, or adding c times one column
to another column, does not change the determinant. If a matrix has two rows equal
or two columns equal, its determinant is zero. More generally, if one row is c times
another row, or one column is c times another column, then the determinant is zero.
There are 2n ways to compute [ A[; expansion by any row or expansion by any
column. Let M
i,j
be the determinant of the (n − 1) (n − 1) matrix obtained by
removing row i and column j from A. Let C
i,j
= (−1)
i+j
M
i,j
. M
i,j
and C
i,j
are
called the (i, j) minor and cofactor of A. The following theorem is useful but the
proof is a little tedious and should not be done as an exercise.
Theorem For any 1 ≤ i ≤ n, [ A[= a
i,1
C
i,1
+ a
i,2
C
i,2
+ + a
i,n
C
i,n
. For any
1 ≤ j ≤ n, [ A[= a
1,j
C
1,j
+a
2,j
C
2,j
+ +a
n,j
C
n,j
. Thus if any row or any column is
zero, the determinant is zero.
Exercise Let A =

¸
¸
a
1
a
2
a
3
b
1
b
2
b
3
c
1
c
2
c
3
¸

. The determinant of A is the sum of six terms.
Chapter 4 Matrices 63
Write out the determinant of A expanding by the ﬁrst column and also expanding by
the second row.
Theorem If A is an upper or lower triangular matrix, [ A[ is the product of the
diagonal elements. If A is an elementary matrix of type 2, [ A [= −1. If A is an
elementary matrix of type 3, [ A[= 1.
Proof We will prove the ﬁrst statement for upper triangular matrices. If A ∈ R
2
is an upper triangular matrix, then its determinant is the product of the diagonal
elements. Suppose n > 2 and the theorem is true for matrices in R
n−1
. Suppose
A ∈ R
n
is upper triangular. The result follows by expanding by the ﬁrst column.
An elementary matrix of type 3 is a special type of upper or lower triangular
matrix, so its determinant is 1. An elementary matrix of type 2 is obtained from the
identity matrix by interchanging two rows or columns, and thus has determinant −1.
Theorem (Determinant by blocks) Suppose A ∈ R
n
, B ∈ R
n,m
, and D ∈ R
m
.
Then the determinant of

A B
O D

is [ A[[ D[.
Proof Expand by the ﬁrst column and use induction on n.
The following remarkable theorem takes some work to prove. We assume it here
without proof. (For the proof, see page 130 of the Appendix.)
Theorem The determinant of the product is the product of the determinants,
i.e., if A, B ∈ R
n
, [ AB[ = [ A[[ B[. Thus [ AB[ = [ BA[ and if C is invertible,
[ C
−1
AC [ = [ACC
−1
[ = [ A[.
Corollary If A is a unit in R
n
, then [ A[ is a unit in R and [ A
−1
[ = [ A[
−1
.
Proof 1 = [ I [ = [ AA
−1
[ = [ A[[ A
−1
[ .
One of the major goals of this chapter is to prove the converse of the preceding
corollary.
Classical adjoint Suppose R is a commutative ring and A ∈ R
n
. The classical
i,j
)
t
, i.e., the matrix whose (j, i) term is the (i, j) cofactor. Before
64 Matrices Chapter 4
we consider the general case, let’s examine 2 2 matrices.
If A =

a b
c d

then (C
i,j
) =

d −c
−b a

and so (C
i,j
)
t
=

d −b
−c a

. Then
A(C
i,j
)
t
= (C
i,j
)
t
A =

[ A[ 0
0 [ A[

= [ A[ I. Thus if [ A[ is a unit in R, A is
invertible and A
−1
= [ A[
−1
(C
i,j
)
t
. In particular, if [ A[ = 1, A
−1
=

d −b
−c a

.
Here is the general case.
Theorem If R is commutative and A ∈ R
n
, then A(C
i,j
)
t
= (C
i,j
)
t
A = [ A[ I.
Proof We must show that the diagonal elements of the product A(C
i,j
)
t
are all
[ A[ and the other elements are 0. The (s, s) term is the dot product of row s of A
with row s of (C
i,j
) and is thus [ A[ (computed by expansion by row s). For s = t,
the (s, t) term is the dot product of row s of A with row t of (C
i,j
). Since this is the
determinant of a matrix with row s = row t, the (s, t) term is 0. The proof that
(C
i,j
)
t
A = [A[I is similar and is left as an exercise.
We are now ready for one of the most beautiful and useful theorems in all of
mathematics.
Theorem Suppose R is a commutative ring and A ∈ R
n
. Then A is a unit in
R
n
iﬀ [ A[ is a unit in R. (Thus if R is a ﬁeld, A is invertible iﬀ [ A[ = 0.) If A is
invertible, then A
−1
= [ A[
−1
(C
i,j
)
t
. Thus if [ A[ = 1, A
−1
= (C
i,j
)
t
, the classical
Proof This follows immediately from the preceding theorem.
Exercise Show that any right inverse of A is also a left inverse. That is, suppose
A, B ∈ R
n
and AB = I. Show A is invertible with A
−1
= B, and thus BA = I.
Similarity
Suppose A, B ∈ R
n
. B is said to be similar to A if ∃ an invertible C ∈ R
n
such
that B = C
−1
AC, i.e., B is similar to A iﬀ B is a conjugate of A.
Theorem B is similar to B.
Chapter 4 Matrices 65
B is similar to A iﬀ A is similar to B.
If D is similar to B and B is similar to A, then D is similar to A.
“Similarity” is an equivalence relation on R
n
.
Proof This is a good exercise using the deﬁnition.
Theorem Suppose A and B are similar. Then [ A[ = [ B[ and thus A is invertible
iﬀ B is invertible.
Proof Suppose B = C
−1
AC. Then [ B[ = [ C
−1
AC [ = [ACC
−1
[ = [ A[.
Trace Suppose A = (a
i,j
) ∈ R
n
. Then the trace is deﬁned by trace(A) = a
1,1
+
a
2,2
+ +a
n,n
. That is, the trace of A is the sum of its diagonal terms.
One of the most useful properties of trace is trace(AB) = trace(BA) whenever AB
and BA are deﬁned. For example, suppose A = (a
1
, a
2
, ..., a
n
) and B = (b
1
, b
2
, ..., b
n
)
t
.
Then AB is the scalar a
1
b
1
+ + a
n
b
n
while BA is the n n matrix (b
i
a
j
). Note
that trace(AB) = trace(BA). Here is the theorem in full generality.
Theorem Suppose A ∈ R
m,n
and B ∈ R
n,m
. Then AB and BA are square
matrices with trace(AB) = trace(BA).
Proof This proof involves a change in the order of summation. By deﬁnition,
trace(AB) =
¸
1≤i≤m
a
i,1
b
1,i
+ +a
i,n
b
n,i
=
¸
1≤i≤m
1≤j≤n
a
i,j
b
j,i
=
¸
1≤j≤n
b
j,1
a
1,j
+ +b
j,m
a
m,j
=
trace(BA).
Theorem If A, B ∈ R
n
, trace(A+B) = trace(A) + trace(B) and
trace(AB) = trace(BA).
Proof The ﬁrst part of the theorem is immediate, and the second part is a special
case of the previous theorem.
Theorem If A and B are similar, then trace(A) = trace(B).
Proof trace(B) = trace(C
−1
AC) = trace(ACC
−1
) = trace(A).
66 Matrices Chapter 4
Summary Determinant and trace are functions from R
n
to R. Determinant is a
multiplicative homomorphism and trace is an additive homomorphism. Furthermore
[ AB[ = [ BA[ and trace(AB) = trace(BA). If A and B are similar, [ A[ = [ B[ and
trace(A) = trace(B).
Exercise Suppose A ∈ R
n
and a ∈ R. Find [aA[ and trace(aA).
Characteristic polynomials If A ∈ R
n
, the characteristic polynomial CP
A
(x) ∈
R[x] is deﬁned by CP
A
(x) = [ (xI − A) [. Any λ ∈ R which is a root of CP
A
(x) is
called a characteristic root of A.
Theorem CP
A
(x) = a
0
+ a
1
x + + a
n−1
x
n−1
+ x
n
where trace(A) = −a
n−1
and [ A[ = (−1)
n
a
0
.
Proof This follows from a direct computation of the determinant.
Theorem If A and B are similar, then they have the same characteristic polyno-
mials.
Proof Suppose B = C
−1
AC. CP
B
(x) = [ (xI −C
−1
AC) [ = [ C
−1
(xI −A)C[ =
[ (xI −A)[ = CP
A
(x).
Exercise Suppose R is a commutative ring, A =

a b
c d

is a matrix in R
2
, and
CP
A
(x) = a
0
+ a
1
x + x
2
. Find a
0
and a
1
and show that a
0
I + a
1
A + A
2
= 0, i.e.,
show A satisﬁes its characteristic polynomial. In other words, CP
A
(A) = 0.
Exercise Suppose F is a ﬁeld and A ∈ F
2
. Show the following are equivalent.
1) A
2
= 0.
2) [ A [= trace(A) = 0.
3) CP
A
(x) = x
2
.
4) ∃ an elementary matrix C such that C
−1
AC is strictly upper triangular.
Note This exercise is a special case of a more general theorem. A square matrix
over a ﬁeld is nilpotent iﬀ all its characteristic roots are 0
¯
iﬀ it is similar to a strictly
upper triangular matrix. This remarkable result cannot be proved by matrix theory
alone, but depends on linear algebra (see pages 93, 94, and 98).
Chapter 5
Linear Algebra
The exalted position held by linear algebra is based upon the subject’s ubiquitous
utility and ease of application. The basic theory is developed here in full generality,
i.e., modules are deﬁned over an arbitrary ring R and not just over a ﬁeld. The
tern as in the chapters on groups and rings. We give a simple proof that if R is a
commutative ring and f : R
n
→ R
n
is a surjective R-module homomorphism, then
f is an isomorphism. This shows that ﬁnitely generated free R-modules have a well
deﬁned dimension, and simpliﬁes some of the development of linear algebra. It is in
this chapter that the concepts about functions, solutions of equations, matrices, and
generating sets come together in one uniﬁed theory.
After the general theory, we restrict our attention to vector spaces, i.e., modules
over a ﬁeld. The key theorem is that any vector space V has a free basis, and thus
if V is ﬁnitely generated, it has a well deﬁned dimension, and incredible as it may
seem, this single integer determines V up to isomorphism. Also any endomorphism
f : V →V may be represented by a matrix, and any change of basis corresponds to
conjugation of that matrix. One of the goals in linear algebra is to select a basis so
that the matrix representing f has a simple form. For example, if f is not injective,
then f may be represented by a matrix whose ﬁrst column is zero. As another
example, if f is nilpotent, then f may be represented by a strictly upper triangular
matrix. The theorem on Jordan canonical form is not proved in this chapter, and
should not be considered part of this chapter. It is stated here in full generality only
for reference and completeness. The proof is given in the Appendix. This chapter
concludes with the study of real inner product spaces, and with the beautiful theory
relating orthogonal matrices and symmetric matrices.
67
68 Linear Algebra Chapter 5
Deﬁnition Suppose R is a ring and M is an additive abelian group. The state-
ment that M is a right R-module means there is a scalar multiplication
M R → M satisfying (a
1
+a
2
)r = a
1
r +a
2
r
(m, r) → mr a(r
1
+r
2
) = ar
1
+ar
2
a(r
1
r
2
) = (ar
1
)r
2
a1
¯
= a
for all a, a
1
, a
2
∈ M and r, r
1
, r
2
∈ R.
The statement that M is a left R-module means there is a scalar multiplication
R M → M satisfying r(a
1
+a
2
) = ra
1
+ra
2
(r, m) → rm (r
1
+r
2
)a = r
1
a +r
2
a
(r
1
r
2
)a = r
1
(r
2
a)
1
¯
a = a
Note that the plus sign is used ambiguously, as addition in M and as addition in R.
Notation The fact that M is a right (left) R-module will be denoted by M = M
R
(M =
R
M). If R is commutative and M = M
R
then left scalar multiplication deﬁned
by ra = ar makes M into a left R-module. Thus for commutative rings, we may write
the scalars on either side. In this text we stick to right R-modules.
Convention Unless otherwise stated, it is assumed that R is a ring and the word
“R-module” (or sometimes just “module”) means “right R-module”.
Theorem Suppose M is an R-module.
1) If r ∈ R, then f : M →M deﬁned by f(a) = ar is a homomorphism of
¯
M
)r = 0
¯
M
.
2) If a ∈ M, a0
¯
R
= 0
¯
M
.
3) If a ∈ M and r ∈ R, then (−a)r = −(ar) = a(−r).
Proof This is a good exercise in using the axioms for an R-module.
Chapter 5 Linear Algebra 69
Submodules If M is an R-module, the statement that a subset N ⊂ M is a
submodule means it is a subgroup which is closed under scalar multiplication, i.e., if
a ∈ N and r ∈ R, then ar ∈ N. In this case N will be an R-module because the
axioms will automatically be satisﬁed. Note that 0
¯
and M are submodules, called the
improper submodules of M.
Theorem Suppose M is an R-module, T is an index set, and for each t ∈ T,
N
t
is a submodule of M.
1)
¸
t∈T
N
t
is a submodule of M.
2) If ¦N
t
¦ is a monotonic collection,
¸
t∈T
N
t
is a submodule.
3) +
t∈T
N
t
= ¦all ﬁnite sums a
1
+ +a
m
: each a
i
belongs
to some N
t
¦ is a submodule. If T = ¦1, 2, .., n¦,
then this submodule may be written as
N
1
+N
2
+ +N
n
= ¦a
1
+a
2
+ +a
n
: each a
i
∈ N
i
¦.
Proof We know from page 22 that versions of 1) and 2) hold for subgroups, and
in particular for subgroups of additive abelian groups. To ﬁnish the proofs it is only
necessary to check scalar multiplication, which is immediate. Also the proof of 3) is
immediate. Note that if N
1
and N
2
are submodules of M, N
1
+ N
2
is the smallest
submodule of M containing N
1
∪ N
2
.
Exercise Suppose T is a non-void set, N is an R-module, and N
T
is the collection
of all functions f : T →N with addition deﬁned by (f +g)(t) = f(t)+g(t), and scalar
multiplication deﬁned by (fr)(t) = f(t)r. Show N
T
is an R-module. (We know from
the last exercise in Chapter 2 that N
T
is a group, and so it is only necessary to check
scalar multiplication.) This simple fact is quite useful in linear algebra. For example,
in 5) of the theorem below, it is stated that Hom
R
(M, N) forms an abelian group.
So it is only necessary to show that Hom
R
(M, N) is a subgroup of N
M
. Also in 8) it
is only necessary to show that Hom
R
(M, N) is a submodule of N
M
.
Homomorphisms
Suppose M and N are R-modules. A function f : M → N is a homomorphism
(i.e., an R-module homomorphism) provided it is a group homomorphism and if
a ∈ M and r ∈ R, f(ar) = f(a)r. On the left, scalar multiplication is in M and on
the right it is in N. The basic facts about homomorphisms are listed below. Much
70 Linear Algebra Chapter 5
of this work has already been done in the chapter on groups (see page 28).
Theorem
1) The zero map M →N is a homomorphism.
2) The identity map I : M →M is a homomorphism.
3) The composition of homomorphisms is a homomorphism.
4) The sum of homomorphisms is a homomorphism. If f, g : M →N are
homomorphisms, deﬁne (f +g) : M →N by (f +g)(a) = f(a) +g(a).
Then f +g is a homomorphism. Also (−f) deﬁned by (−f)(a) = −f(a)
is a homomorphism. If h : N →P is a homomorphism,
h ◦ (f +g) = (h ◦ f) + (h ◦ g). If k : P →M is a homomorphism,
(f +g ) ◦ k = (f ◦ k) + (g ◦ k).
5) Hom
R
(M, N) = Hom(M
R
, N
R
), the set of all homomorphisms from M
to N, forms an abelian group under addition. Hom
R
(M, M), with
multiplication deﬁned to be composition, is a ring.
6) If a bijection f : M →N is a homomorphism, then f
−1
: N →M is also
a homomorphism. In this case f and f
−1
are called isomorphisms. A
homomorphism f : M →M is called an endomorphism. An isomorphism
f : M →M is called an automorphism. The units of the endomorphism
ring Hom
R
(M, M) are the automorphisms. Thus the automorphisms on
M form a group under composition. We will see later that if M = R
n
,
Hom
R
(R
n
, R
n
) is just the matrix ring R
n
and the automorphisms
are merely the invertible matrices.
7) If R is commutative and r ∈ R, then g : M →M deﬁned by g(a) = ar
is a homomorphism. Furthermore, if f : M →N is a homomorphism,
fr deﬁned by (fr)(a) = f(ar) = f(a)r is a homomorphism.
8) If R is commutative, Hom
R
(M, N) is an R-module.
9) Suppose f : M →N is a homomorphism, G ⊂ M is a submodule,
and H ⊂ N is a submodule. Then f(G) is a submodule of N
and f
−1
(H) is a submodule of M. In particular, image(f) is a
submodule of N and ker(f) = f
−1
(0
¯
) is a submodule of M.
Proof This is just a series of observations.
Chapter 5 Linear Algebra 71
Abelian groups are Z-modules On page 21, it is shown that any additive
group M admits a scalar multiplication by integers, and if M is abelian, the properties
are satisﬁed to make M a Z-module. Note that this is the only way M can be a Z-
module, because a1 = a, a2 = a + a, etc. Furthermore, if f : M → N is a group
homomorphism of abelian groups, then f is also a Z-module homomorphism.
Summary Additive abelian groups are “the same things” as Z-modules. While
group theory in general is quite separate from linear algebra, the study of additive
abelian groups is a special case of the study of R-modules.
Exercise R-modules are also Z-modules and R-module homomorphisms are also
Z-module homomorphisms. If M and N are Q-modules and f : M → N is a
Z-module homomorphism, must it also be a Q-module homomorphism?
Homomorphisms on R
n
R
n
as an R-module On page 54 it was shown that the additive abelian
group R
m,n
admits a scalar multiplication by elements in R. The properties listed
there were exactly those needed to make R
m,n
an R-module. Of particular importance
is the case R
n
= R ⊕ ⊕R = R
n,1
(see page 53). We begin with the case n = 1.
R as a right R-module Let M = R and deﬁne scalar multiplication on the right
by ar = a r. That is, scalar multiplication is just ring multiplication. This makes
R a right R-module denoted by R
R
(or just R). This is the same as the deﬁnition
before for R
n
when n = 1.
Theorem Suppose R is a ring and N is a subset of R. Then N is a submodule
of R
R
(
R
R) iﬀ N is a right (left) ideal of R.
Proof The deﬁnitions are the same except expressed in diﬀerent language.
Theorem Suppose M = M
R
and f, g : R →M are homomorphisms with f(1
¯
) =
g(1
¯
). Then f = g. Furthermore, if m ∈ M, ∃! homomorphism h : R → M with
h(1
¯
) = m. In other words, Hom
R
(R, M) ≈ M.
Proof Suppose f(1
¯
) = g(1
¯
). Then f(r) = f(1
¯
r) = f(1
¯
)r = g(1
¯
)r = g(1
¯
r) =
g(r). Given m ∈ M, h : R → M deﬁned by h(r) = mr is a homomorphism. Thus
72 Linear Algebra Chapter 5
evaluation at 1
¯
gives a bijection from Hom
R
(R, M) to M, and this bijection is clearly
a group isomorphism. If R is commutative, it is an isomorphism of R-modules.
In the case M = R, the above theorem states that multiplication on left by some
m ∈ R deﬁnes a right R-module homomorphism from R to R, and every module
homomorphism is of this form. The element m should be thought of as a 1 1
matrix. We now consider the case where the domain is R
n
.
Homomorphisms on R
n
Deﬁne e
i
∈ R
n
by e
i
=

¸
¸
¸
¸
¸
¸
¸
0
¯

1
¯
i

0
¯
¸

. Note that any

¸
¸
¸
¸
¸
¸
¸
r
1

r
n
¸

can be written uniquely as e
1
r
1
+ +e
n
r
n
. The sequence ¦e
1
, .., e
n
¦ is called the
canonical free basis or standard basis for R
n
.
Theorem Suppose M = M
R
and f, g : R
n
→ M are homomorphisms with
f(e
i
) = g(e
i
) for 1 ≤ i ≤ n. Then f = g. Furthermore, if m
1
, m
2
, ..., m
n
∈ M, ∃!
homomorphism h : R
n
→ M with h(e
i
) = m
i
for 1 ≤ i ≤ m. The homomorphism
h is deﬁned by h(e
1
r
1
+ +e
n
r
n
) = m
1
r
1
+ +m
n
r
n
.
Proof The proof is straightforward. Note this theorem gives a bijection from
Hom
R
(R
n
, M) to M
n
= M M M and this bijection is a group isomorphism.
We will see later that the product M
n
is an R-module with scalar multiplication
deﬁned by (m
1
, m
2
, .., m
n
)r = (m
1
r, m
2
r, .., m
n
r). If R is commutative so that
Hom
R
(R
n
, M) is an R-module, this theorem gives an R-module isomorphism from
Hom
R
(R
n
, M) to M
n
.
This theorem reveals some of the great simplicity of linear algebra. It does not
matter how complicated the ring R is, or which R-module M is selected. Any
R-module homomorphism from R
n
to M is determined by its values on the basis,
and any function from that basis to M extends uniquely to a homomorphism from
R
n
to M.
Exercise Suppose R is a ﬁeld and f : R
R
→ M is a non-zero homomorphism.
Show f is injective.
Chapter 5 Linear Algebra 73
Now let’s examine the special case M = R
m
and show Hom
R
(R
n
, R
m
) ≈ R
m,n
.
Theorem Suppose A = (a
i,j
) ∈ R
m,n
. Then f : R
n
→R
m
deﬁned by f(B) = AB
is a homomorphism with f(e
i
) = column i of A. Conversely, if v
1
, . . . , v
n
∈ R
m
, deﬁne
A ∈ R
m,n
to be the matrix with column i = v
i
. Then f deﬁned by f(B) = AB is
the unique homomorphism from R
n
to R
m
with f(e
i
) = v
i
.
Even though this follows easily from the previous theorem and properties of ma-
trices, it is one of the great classical facts of linear algebra. Matrices over R give
tion of homomorphisms, and multiplication of matrices corresponds to composition
of homomorphisms. These properties are made explicit in the next two theorems.
Theorem If f, g : R
n
→ R
m
are given by matrices A, C ∈ R
m,n
, then f + g is
given by the matrix A+C. Thus Hom
R
(R
n
, R
m
) and R
m,n
groups. If R is commutative, they are isomorphic as R-modules.
Theorem If f : R
n
→ R
m
is the homomorphism given by A ∈ R
m,n
and g :
R
m
→R
p
is the homomorphism given by C ∈ R
p,m
, then g ◦ f : R
n
→R
p
is given by
CA ∈ R
p,n
. That is, composition of homomorphisms corresponds to multiplication
of matrices.
Proof This is just the associative law of matrix multiplication, C(AB) = (CA)B.
The previous theorem reveals where matrix multiplication comes from. It is the
matrix which represents the composition of the functions. In the case where the
domain and range are the same, we have the following elegant corollary.
Corollary Hom
R
(R
n
, R
n
) and R
n
are isomorphic as rings. The automorphisms
correspond to the invertible matrices.
This corollary shows one way non-commutative rings arise, namely as endomor-
phism rings. Even if R is commutative, R
n
is never commutative unless n = 1.
We now return to the general theory of modules (over some given ring R).
74 Linear Algebra Chapter 5
Cosets and Quotient Modules
After seeing quotient groups and quotient rings, quotient modules go through
without a hitch. As before, R is a ring and module means R-module.
Theorem Suppose M is a module and N ⊂ M is a submodule. Since N is a
normal subgroup of M, the additive abelian quotient group M/N is deﬁned. Scalar
multiplication deﬁned by (a + N)r = (ar + N) is well-deﬁned and gives M/N the
structure of an R-module. The natural projection π : M → M/N is a surjective
homomorphism with kernel N. Furthermore, if f : M →
¯
M is a surjective homomor-
phism with ker(f) = N, then M/N ≈
¯
M (see below).
Proof On the group level, this is all known from Chapter 2 (see pages 27 and 29).
It is only necessary to check the scalar multiplication, which is obvious.
The relationship between quotients and homomorphisms for modules is the same
as for groups and rings, as shown by the next theorem.
Theorem Suppose f : M →
¯
M is a homomorphism and N is a submodule of M.
If N ⊂ ker(f), then
¯
f : (M/N) →
¯
M deﬁned by
¯
f(a + N) = f(a) is a well-deﬁned
homomorphism making the following diagram commute.
M
¯
M
M/N
f
·

π
¯
f
Thus deﬁning a homomorphism on a quotient module is the same as deﬁning a homo-
morphism on the numerator that sends the denominator to 0
¯
. The image of
¯
f is the
image of f, and the kernel of
¯
f is ker(f)/N. Thus if N = ker(f),
¯
f is injective, and
thus (M/N) ≈image(f). Therefore for any homomorphism f, (domain(f)/ker(f)) ≈
image(f).
Proof On the group level this is all known from Chapter 2 (see page 29). It is
only necessary to check that
¯
f is a module homomorphism, and this is immediate.
Chapter 5 Linear Algebra 75
Theorem Suppose M is an R-module and K and L are submodules of M.
i) The natural homomorphism K →(K + L)/L is surjective with kernel
K ∩ L. Thus (K/K ∩ L)

→(K +L)/L is an isomorphism.
ii) Suppose K ⊂ L. The natural homomorphism M/K →M/L is surjective
with kernel L/K. Thus (M/K)/(L/K)

→M/L is an isomorphism.
Examples These two examples are for the case R = Z, i.e., for abelian groups.
1) M = Z K = 3Z L = 5Z K ∩ L = 15Z K +L = Z
K/K ∩ L = 3Z/15Z ≈ Z/5Z = (K +L)/L
2) M = Z K = 6Z L = 3Z (K ⊂ L)
(M/K)/(L/K) = (Z/6Z)/(3Z/6Z) ≈ Z/3Z = M/L
Products and Coproducts
Inﬁnite products work ﬁne for modules, just as they do for groups and rings.
This is stated below in full generality, although the student should think of the ﬁnite
case. In the ﬁnite case something important holds for modules that does not hold
for non-abelian groups or rings – namely, the ﬁnite product is also a coproduct. This
makes the structure of module homomorphisms much more simple. For the ﬁnite
case we may use either the product or sum notation, i.e., M
1
M
2
M
n
=
M
1
⊕M
2
⊕ ⊕M
n
.
Theorem Suppose T is an index set and for each t ∈ T, M
t
is an R-module. On
¸
t∈T
M
t
=
¸
M
t
deﬁne scalar multiplication by ¦m
t
¦r =
¦m
t
r¦. Then
¸
M
t
is an R-module and, for each s ∈ T, the natural projection
π
s
:
¸
M
t
→M
s
is a homomorphism. Suppose M is a module. Under the natural 1-1
correspondence from ¦functions f : M →
¸
M
t
¦ to ¦sequence of functions ¦f
t
¦
t∈T
where f
t
: M →M
t
¦, f is a homomorphism iﬀ each f
t
is a homomorphism.
Proof We already know from Chapter 2 that f is a group homomorphism iﬀ each
f
t
is a group homomorphism. Since scalar multiplication is deﬁned coordinatewise,
f is a module homomorphism iﬀ each f
t
is a module homomorphism.
76 Linear Algebra Chapter 5
Deﬁnition If T is ﬁnite, the coproduct and product are the same module. If T
is inﬁnite, the coproduct or sum
¸
t∈T
M
t
=

t∈T
M
t
= ⊕M
t
is the submodule of
¸
M
t
consisting of all sequences ¦m
t
¦ with only a ﬁnite number of non-zero terms. For
each s ∈ T, the inclusion homomorphisms i
s
: M
s
→⊕M
t
is deﬁned by i
s
(a) = ¦a
t
¦
where a
t
= 0
¯
if t = s and a
s
= a. Thus each M
s
may be considered to be a submodule
of ⊕M
t
.
Theorem Suppose M is an R-module. There is a 1-1 correspondence from
¦homomorphisms g : ⊕M
t
→ M¦ and ¦sequences of homomorphisms ¦g
t
¦
t∈T
where
g
t
: M
t
→ M¦ . Given g, g
t
is deﬁned by g
t
= g ◦ i
t
. Given ¦g
t
¦, g is deﬁned by
g(¦m
t
¦) =
¸
t
g
t
(m
t
). Since there are only a ﬁnite number of non-zero terms, this
sum is well deﬁned.
For T = ¦1, 2¦ the product and sum properties are displayed in the following
commutative diagrams.
M
1
⊕M
2
M
1
M
2
M
1
M
2
M
1
⊕M
2
M M
π
1 i
1
π
2 i
2
f
g
f
1
f
2
g
1
g
2
, ,
·
`

»

´
Theorem For ﬁnite T, the 1-1 correspondences in the above theorems actually
produce group isomorphisms. If R is commutative, they give isomorphisms of R-
modules.
Hom
R
(M, M
1
⊕ ⊕M
n
) ≈ Hom
R
(M, M
1
) ⊕ ⊕Hom
R
(M, M
n
) and
Hom
R
(M
1
⊕ ⊕M
n
, M) ≈ Hom
R
(M
1
, M) ⊕ ⊕Hom
R
(M
n
, M)
Proof Let’s look at this theorem for products with n = 2. All it says is that if f =
(f
1
, f
2
) and h = (h
1
, h
2
), then f +h = (f
1
+h
1
, f
2
+h
2
). If R is commutative, so that
the objects are R-modules and not merely additive groups, then the isomorphisms
are module isomorphisms. This says merely that fr = (f
1
, f
2
)r = (f
1
r, f
2
r).
Chapter 5 Linear Algebra 77
Exercise Suppose M and N are R-modules. Show that M ⊕N is isomorphic to
N ⊕ M. Now suppose A ⊂ M, B ⊂ N are submodules and show (M ⊕ N)/(A ⊕B)
is isomorphic to (M/A) ⊕ (N/B). In particular, if a ∈ R and b ∈ R, then
(R ⊕ R)/(aR ⊕ bR) is isomorphic to (R/aR) ⊕ (R/bR). For example, the abelian
group (Z ⊕ Z)/(2Z ⊕ 3Z) is isomorphic to Z
2
⊕ Z
3
. These isomorphisms are trans-
parent and are used routinely in algebra without comment (see Th 4, page 118).
Exercise Suppose R is a commutative ring, M is an R-module, and n ≥ 1. Deﬁne
a function α : Hom
R
(R
n
, M) →M
n
which is a R-module isomorphism.
Summands
One basic question in algebra is “When does a module split as the sum of two
modules?”. Before deﬁning summand, here are two theorems for background.
Theorem Consider M
1
= M
1
⊕0
¯
as a submodule of M
1
⊕M
2
. Then the projection
map π
2
: M
1
⊕ M
2
→ M
2
is a surjective homomorphism with kernel M
1
. Thus
(M
1
⊕M
2
)/M
1
is isomorphic to M
2
. (See page 35 for the group version.)
This is exactly what you would expect, and the next theorem is almost as intuitive.
Theorem Suppose K and L are submodules of M and f : K ⊕ L → M is the
natural homomorphism, f(k, l) = k + l. Then the image of f is K + L and the
kernel of f is ¦(a, −a) : a ∈ K ∩ L¦. Thus f is an isomorphism iﬀ K + L = M and
K ∩ L = 0
¯
. In this case we write K ⊕ L = M. This abuse of notation allows us to
avoid talking about “internal” and “external” direct sums.
Deﬁnition Suppose K is a submodule of M. The statement that K is a summand
of M means ∃ a submodule L of M with K ⊕ L = M. According to the previous
theorem, this is the same as there exists a submodule L with K + L = M and
K ∩ L = 0
¯
. If such an L exists, it need not be unique, but it will be unique up to
isomorphism, because L ≈ M/K. Of course, M and 0
¯
are always summands of M.
Exercise Suppose M is a module and K = ¦(m, m) : m ∈ M¦ ⊂ M ⊕M. Show
K is a submodule of M ⊕M which is a summand.
Exercise R is a module over Q, and Q ⊂ R is a submodule. Is Q a summand of
R? With the material at hand, this is not an easy question. Later on, it will be easy.
78 Linear Algebra Chapter 5
(See the third exercise on page 35.)
1) Is 2Z a summand of Z?
2) Is 2Z
4
a summand of Z
4
?
3) Is 3Z
12
a summand of Z
12
?
4) Suppose m, n > 1. When is nZ
mn
a summand of Z
mn
?
Exercise If T is a ring, deﬁne the center of T to be the subring ¦t : ts =
st for all s ∈ T¦. Let R be a commutative ring and T = R
n
. There is a exercise
on page 57 to show that the center of T is the subring of scalar matrices. Show R
n
is a left T-module and ﬁnd Hom
T
(R
n
, R
n
).
Independence, Generating Sets, and Free Basis
This section is a generalization and abstraction of the brief section Homomor-
phisms on R
n
. These concepts work ﬁne for an inﬁnite index set T because linear
combination means ﬁnite linear combination. However, to avoid dizziness, the student
should ﬁrst consider the case where T is ﬁnite.
Deﬁnition Suppose M is an R-module, T is an index set, and for each t ∈ T,
s
t
∈ M. Let S be the sequence ¦s
t
¦
t∈T
= ¦s
t
¦. The statement that S is dependent
means ∃ a ﬁnite number of distinct elements t
1
, ..., t
n
in T, and elements r
1
, .., r
n
in
R, not all zero, such that the linear combination s
t
1
r
1
+ +s
tn
r
n
= 0
¯
. Otherwise,
S is independent. Note that if some s
t
= 0
¯
, then S is dependent. Also if ∃ distinct
elements t
1
and t
2
in T with s
t
1
= s
t
2
, then S is dependent.
Let SR be the set of all linear combinations s
t
1
r
1
+ +s
tn
r
n
. SR is a submodule
of M called the submodule generated by S. If S is independent and generates M,
then S is said to be a basis or free basis for M. In this case any v ∈ M can be written
uniquely as a linear combination of elements in S. An R-module M is said to be a
free R-module if it is zero or if it has a basis. The next two theorems are obvious,
except for the confusing notation. You might try ﬁrst the case T = ¦1, 2, ..., n¦ and
⊕R
t
= R
n
(see p 72).
Theorem For each t ∈ T, let R
t
= R
R
and for each c ∈ T, let e
c
∈ ⊕R
t
=

t∈T
R
t
be e
c
= ¦r
t
¦ where r
c
= l
¯
and r
t
= 0
¯
if t = c. Then ¦e
c
¦
c∈T
is a basis for ⊕R
t
called
the canonical basis or standard basis.
Chapter 5 Linear Algebra 79
Theorem Suppose N is an R-module and M is a free R-module with a basis
¦s
t
¦. Then ∃ a 1-1 correspondence from the set of all functions g : ¦s
t
¦ →N and the
set of all homomorphisms f : M → N. Given g, deﬁne f by f(s
t
1
r
1
+ +s
tn
r
n
) =
g(s
t
1
)r
1
+ +g(s
tn
)r
n
. Given f, deﬁne g by g(s
t
) = f(s
t
). In other words, f is
completely determined by what it does on the basis S, and you are “free” to send the
basis any place you wish and extend to a homomorphism.
Recall that we have already had the preceding theorem in the case S is the canon-
ical basis for M = R
n
(p 72). The next theorem is so basic in linear algebra that it
is used without comment. Although the proof is easy, it should be worked carefully.
Theorem Suppose N is a module, M is a free module with basis S = ¦s
t
¦, and
f : M →N is a homomorphism. Let f(S) be the sequence ¦f(s
t
)¦ in N.
1) f(S) generates N iﬀ f is surjective.
2) f(S) is independent in N iﬀ f is injective.
3) f(S) is a basis for N iﬀ f is an isomorphism.
4) If h : M →N is a homomorphism, then f = h iﬀ f [ S = h [ S.
Exercise Let (A
1
, .., A
n
) be a sequence of n vectors with each A
i
∈ Z
n
.
Show this sequence is linearly independent over Z iﬀ it is linearly independent over Q.
Is it true the sequence is linearly independent over Z iﬀ it is linearly independent
over R? This question is diﬃcult until we learn more linear algebra.
Characterization of Free Modules
Any free R-module is isomorphic to one of the canonical free R-modules ⊕R
t
.
This is just an observation, but it is a central fact in linear algebra.
Theorem A non-zero R-module M is free iﬀ ∃ an index set T such that
M ≈

t∈T
R
t
. In particular, M has a ﬁnite free basis of n elements iﬀ M ≈ R
n
.
Proof If M is isomorphic to ⊕R
t
then M is certainly free. So now suppose M
has a free basis ¦s
t
¦. Then the homomorphism f : M → ⊕R
t
with f(s
t
) = e
t
sends
the basis for M to the canonical basis for ⊕R
t
. By 3) in the preceding theorem, f is
an isomorphism.
80 Linear Algebra Chapter 5
Exercise Suppose R is a commutative ring, A ∈ R
n
, and the homomorphism
f : R
n
→ R
n
deﬁned by f(B) = AB is surjective. Show f is an isomorphism, i.e.,
show A is invertible. This is a key theorem in linear algebra, although it is usually
stated only for the case where R is a ﬁeld. Use the fact that ¦e
1
, .., e
n
¦ is a free basis
for R
n
.
The next exercise is routine, but still informative.
Exercise Let R = Z, A =

2 1 0
3 2 −5

and f: Z
3
→ Z
2
be the group homo-
morphism deﬁned by A. Find a non-trivial linear combination of the columns of A
which is 0
¯
. Also ﬁnd a non-zero element of kernel(f).
If R is a commutative ring, you can relate properties of R as an R-module to
properties of R as a ring.
Exercise Suppose R is a commutative ring and v ∈ R, v = 0
¯
.
1) v is independent iﬀ v is .
2) v is a basis for R iﬀ v generates R iﬀ v is .
Note that 2) here is essentially the ﬁrst exercise for the case n = 1. That is, if
f : R →R is a surjective R-module homomorphism, then f is an isomorphism.
Relating these concepts to matrices
The theorem stated below gives a summary of results we have already had. It
shows that certain concepts about matrices, linear independence, injective homo-
morphisms, and solutions of equations, are all the same — they are merely stated in
diﬀerent language. Suppose A ∈ R
m,n
and f : R
n
→R
m
is the homomorphism associ-
ated with A, i.e., f(B) = AB. Let v
1
, .., v
n
∈ R
m
be the columns of A, i.e., f(e
i
) = v
i
= column i of A. Let B =

¸
¸
b
1
.
b
n
¸

represent an element of R
n
and C =

¸
¸
c
1
.
c
m
¸

Chapter 5 Linear Algebra 81
represent an element of R
m
.
Theorem
1) The element f(B) is a linear combination of the columns of A, that is
f(B) = f(e
1
b
1
+ +e
n
b
n
) = v
1
b
1
+ +v
n
b
n
. Thus the image of f
is generated by the columns of A. (See bottom of page 89.)
2) ¦v
1
, .., v
n
¦ generates R
m
iﬀ f is surjective iﬀ (for any C ∈ R
m
, AX = C
has a solution).
3) ¦v
1
, .., v
n
¦ is independent iﬀ f is injective iﬀ AX = 0
¯
has a unique
solution iﬀ (∃ C ∈ R
m
such that AX = C has a unique solution).
4) ¦v
1
, .., v
n
¦ is a basis for R
m
iﬀ f is an isomorphism iﬀ (for any C ∈ R
m
,
AX = C has a unique solution).
Relating these concepts to square matrices
We now look at the preceding theorem in the special case where n = m and R
is a commutative ring. So far in this chapter we have just been cataloging. Now we
prove something more substantial, namely that if f : R
n
→ R
n
is surjective, then f
is injective. Later on we will prove that if R is a ﬁeld, injective implies surjective.
Theorem Suppose R is a commutative ring, A ∈ R
n
, and f : R
n
→R
n
is deﬁned
by f(B) = AB. Let v
1
, .., v
n
∈ R
n
be the columns of A, and w
1
, .., w
n
∈ R
n
= R
1,n
be the rows of A. Then the following are equivalent.
1) f is an automorphism.
2) A is invertible, i.e., [ A [ is a unit in R.
3) ¦v
1
, .., v
n
¦ is a basis for R
n
.
4) ¦v
1
, .., v
n
¦ generates R
n
.
5) f is surjective.
2
t
) A
t
is invertible, i.e., [ A
t
[ is a unit in R.
3
t
) ¦w
1
, .., w
n
¦ is a basis for R
n
.
82 Linear Algebra Chapter 5
4
t
) ¦w
1
, .., w
n
¦ generates R
n
.
Proof Suppose 5) is true and show 2). Since f is onto, ∃ u
1
, ..., u
n
∈ R
n
with
f(u
i
) = e
i
. Let g : R
n
→R
n
be the homomorphism satisfying g(e
i
) = u
i
. Then f ◦ g
is the identity. Now g comes from some matrix D and thus AD = I. This shows that
A has a right inverse and is thus invertible. Recall that the proof of this fact uses
determinant, which requires that R be commutative (see the exercise on page 64).
We already know the ﬁrst three properties are equivalent, 4) and 5) are equivalent,
and 3) implies 4). Thus the ﬁrst ﬁve are equivalent. Furthermore, applying this
result to A
t
shows that the last three properties are equivalent to each other. Since
[ A [=[ A
t
[, 2) and 2
t
) are equivalent.
Uniqueness of Dimension
There exists a ring R with R
2
≈ R
3
as R-modules, but this is of little interest. If
R is commutative, this is impossible, as shown below. First we make a convention.
Convention For the remainder of this chapter, R will be a commutative ring.
Theorem If f : R
m
→R
n
is a surjective R-module homomorphism, then m ≥ n.
Proof Suppose k = n − m is positive. Deﬁne h : (R
m
⊕ R
k
= R
n
) → R
n
by
h(u, v) = f(u). Then h is a surjective homomorphism, and by the previous section,
also injective. This is a contradiction and thus m ≥ n.
Corollary If f : R
m
→R
n
is an isomorphism, then m = n.
Proof Each of f and f
−1
is surjective, so m = n by the previous theorem.
Corollary If ¦v
1
, .., v
m
¦ generates R
n
, then m ≥ n.
Proof The hypothesis implies there is a surjective homomorphism R
m
→R
n
. So
this follows from the ﬁrst theorem.
Lemma Suppose M is a f.g. module (i.e., a ﬁnite generated R-module). Then
if M has a basis, that basis is ﬁnite.
Chapter 5 Linear Algebra 83
Proof Suppose U ⊂ M is a ﬁnite generating set and S is a basis. Then any
element of U is a ﬁnite linear combination of elements of S, and thus S is ﬁnite.
Theorem Suppose M is a f.g. module. If M has a basis, that basis is ﬁnite
and any other basis has the same number of elements. This number is denoted by
dim(M), the dimension of M. (By convention, 0
¯
is a free module of dimension 0.)
Proof By the previous lemma, any basis for M must be ﬁnite. M has a basis of
n elements iﬀ M ≈ R
n
. The result follows because R
n
≈ R
m
iﬀ n = m.
Change of Basis
Before changing basis, we recall what a basis is. Previously we deﬁned generat-
ing, independence, and basis for sequences, not for collections. For the concept of
generating it matters not whether you use sequences or collections, but for indepen-
dence and basis, you must use sequences. Consider the columns of the real matrix
A =

2 3 2
1 4 1

. If we consider the column vectors of A as a collection, there are
only two of them, yet we certainly don’t wish to say the columns of A form a basis for
R
2
. In a set or collection, there is no concept of repetition. In order to make sense,
we must consider the columns of A as an ordered triple of vectors, and this sequence
is dependent. In the deﬁnition of basis on page 78, basis is deﬁned for sequences, not
for sets or collections.
Two sequences cannot begin to be equal unless they have the same index set.
Here we follow the classical convention that an index set with n elements will be
¦1, 2, .., n¦, and thus a basis for M with n elements is a sequence S = ¦u
1
, .., u
n
¦
or if you wish, S = (u
1
, .., u
n
) ∈ M
n
. Suppose M is an R-module with a basis of
n elements. Recall there is a bijection α : Hom
R
(R
n
, M) → M
n
deﬁned by α(h) =
(h(e
1
), .., h(e
n
)). Now h : R
n
→M is an isomorphism iﬀ α(h) is a basis for M.
Summary The point of all this is that selecting a basis of n elements for M
is the same as selecting an isomorphism from R
n
to M, and from this viewpoint,
change of basis can be displayed by the diagram below.
Endomorphisms on R
n
are represented by square matrices, and thus have a de-
terminant and trace. Now suppose M is a f.g. free module and f : M → M is a
homomorphism. In order to represent f by a matrix, we must select a basis for M
(i.e., an isomorphism with R
n
). We will show that this matrix is well deﬁned up to
similarity, and thus the determinant, trace, and characteristic polynomial of f are
well-deﬁned.
84 Linear Algebra Chapter 5
Deﬁnition Suppose M is a free module, S = ¦u
1
, .., u
n
¦ is a basis for M, and
f : M → M is a homomorphism. The matrix A = (a
i,j
) ∈ R
n
of f w.r.t. the basis
S is deﬁned by f(u
i
) = u
1
a
1,i
+ +u
n
a
n,i
. (Note that if M = R
n
and u
i
= e
i
, A is
the usual matrix associated with f).
Theorem Suppose T = ¦v
1
, .., v
n
¦ is another basis for M and B ∈ R
n
is the
matrix of f w.r.t. T. Deﬁne C = (c
i,j
) ∈ R
n
by v
i
= u
1
c
1,i
+ +u
n
c
n,i
. Then C is
invertible and B = C
−1
AC, i.e., A and B are similar. Therefore [A[ = [B[,
trace(A)=trace(B), and A and B have the same characteristic polynomial (see page
66 of chapter 4).
Conversely, suppose C = (c
i,j
) ∈ R
n
is invertible. Deﬁne T = ¦v
1
, .., v
n
¦ by
v
i
= u
1
c
1,i
+ +u
n
c
n,i
. Then T is a basis for M and the matrix of f w.r.t. T is
B = C
−1
AC. In other words, conjugation of matrices corresponds to change of basis.
Proof The proof follows by seeing that the following diagram is commutative.
R
n
R
n
R
n
R
n
M M C C
A
B
≈ ≈
≈ ≈
≈ ≈
e
i
v
i
e
i
u
i
v
i
e
i
u
i
e
i
f

· ·

´

»

»
´

»

´

´

»

The diagram also explains what it means for A to be the matrix of f w.r.t. the
basis S. Let h : R
n
→ M be the isomorphism with h(e
i
) = u
i
for 1 ≤ i ≤ n. Then
the matrix A ∈ R
n
is the one determined by the endomorphism h
−1
◦f ◦h : R
n
→R
n
.
In other words, column i of A is h
−1
(f(h(e
i
))).
An important special case is where M = R
n
and f : R
n
→ R
n
is given by some
matrix W. Then h is given by the matrix U whose i
th
column is u
i
and A =
U
−1
WU. In other words, W represents f w.r.t. the standard basis, and U
−1
WU
represents f w.r.t. the basis ¦u
1
, .., u
n
¦.
Deﬁnition Suppose M is a f.g. free module and f : M →M is a homomorphism.
Deﬁne [f[ to be [A[, trace(f) to be trace(A), and CP
f
(x) to be CP
A
(x), where A is
Chapter 5 Linear Algebra 85
the matrix of f w.r.t. some basis. By the previous theorem, all three are well-deﬁned,
i.e., do not depend upon the choice of basis.
Exercise Let R = Z and f : Z
2
→ Z
2
be deﬁned by f(D) =

3 3
0 −1

D.
Find the matrix of f w.r.t. the basis

2
1

,

3
1
¸
.
Exercise Let L ⊂ R
2
be the line L = ¦(r, 2r)
t
: r ∈ R¦. Show there is one
and only one homomorphism f : R
2
→ R
2
which is the identity on L and has
f((−1, 1)
t
) = (1, −1)
t
. Find the matrix A ∈ R
2
which represents f with respect
to the basis ¦(1, 2)
t
, (−1, 1)
t
¦. Find the determinant, trace, and characteristic
polynomial of f. Also ﬁnd the matrix B ∈ R
2
which represents f with respect to
the standard basis. Finally, ﬁnd an invertible matrix C ∈ R
2
with B = C
−1
AC.
Vector Spaces
So far in this chapter we have been developing the theory of linear algebra in
general. The previous theorem, for example, holds for any commutative ring R, but
it must be assumed that the module M is free. Endomorphisms in general will not
have a determinant, trace, or characteristic polynomial. We now focus on the case
where R is a ﬁeld F, and show that in this case, every F-module is free. Thus any
ﬁnitely generated F-module will have a well-deﬁned dimension, and endomorphisms
on it will have well-deﬁned determinant, trace, and characteristic polynomial.
In this section, F is a ﬁeld. F-modules may also be called vector spaces and
F-module homomorphisms may also be called linear transformations.
Theorem Suppose M is an F-module and v ∈ M. Then v = 0
¯
iﬀ v is independent.
That is, if v ∈ V and r ∈ F, vr = 0
¯
implies v = 0
¯
in M or r = 0
¯
in F.
Proof Suppose vr = 0
¯
and r = 0
¯
. Then 0
¯
= (vr)r
−1
= v1
¯
= v.
Theorem Suppose M = 0
¯
is an F-module and v ∈ M. Then v generates M iﬀ v
is a basis for M. Furthermore, if these conditions hold, then M ≈ F
F
, any non-zero
element of M is a basis, and any two elements of M are dependent.
86 Linear Algebra Chapter 5
Proof Suppose v generates M. Then v = 0
¯
and is thus independent by the
previous theorem. In this case M ≈ F, and any non-zero element of F is a basis, and
any two elements of F are dependent.
Theorem Suppose M = 0
¯
is a ﬁnitely generated F-module. If S = ¦v
1
, .., v
m
¦
generates M, then any maximal independent subsequence of S is a basis for M. Thus
any ﬁnite independent sequence can be extended to a basis. In particular, M has a
ﬁnite free basis, and thus is a free F-module.
Proof Suppose, for notational convenience, that ¦v
1
, .., v
n
¦ is a maximal inde-
pendent subsequence of S, and n < i ≤ m. It must be shown that v
i
is a linear
combination of ¦v
1
, .., v
n
¦. Since ¦v
1
, .., v
n
, v
i
¦ is dependent, ∃ r
1
, ..., r
n
, r
i
not all
zero, such that v
1
r
1
+ +v
n
r
n
+v
i
r
i
= 0
¯
. Then r
i
= 0
¯
and v
i
= −(v
1
r
1
+ +v
n
r
n
)r
−1
i
.
Thus ¦v
1
, .., v
n
¦ generates S and thus all of M. Now suppose T is a ﬁnite indepen-
dent sequence. T may be extended to a ﬁnite generating sequence, and inside that
sequence it may be extended to a maximal independent sequence. Thus T extends
to a basis.
After so many routine theorems, it is nice to have one with real power. It not
only says any ﬁnite independent sequence can be extended to a basis, but it can be
extended to a basis inside any ﬁnite generating set containing it. This is one of the
theorems that makes linear algebra tick. The key hypothesis here is that the ring
is a ﬁeld. If R = Z, then Z is a free module over itself, and the element 2 of Z is
independent. However it certainly cannot be extended to a basis. Also the ﬁniteness
hypothesis in this theorem is only for convenience, as will be seen momentarily.
Since F is a commutative ring, any two bases of M must have the same number
of elements, and thus the dimension of M is well deﬁned (see theorem on page 83).
Theorem Suppose M is an F-module of dimension n, and ¦v
1
, ..., v
m
¦ is an
independent sequence in M. Then m ≤ n and if m = n, ¦v
1
, .., v
m
¦ is a basis.
Proof ¦v
1
, .., v
m
¦ extends to a basis with n elements.
The next theorem is just a collection of observations.
Theorem Suppose M and N are ﬁnitely generated F-modules.
Chapter 5 Linear Algebra 87
1) M ≈ F
n
iﬀ dim(M) = n.
2) M ≈ N iﬀ dim(M) = dim(N).
3) F
m
≈ F
n
iﬀ n = m.
4) dim(M ⊕N) = dim(M) + dim(N).
Here is the basic theorem for vector spaces in full generality.
Theorem Suppose M = 0
¯
is an F-module and S = ¦v
t
¦
t∈T
generates M.
1) Any maximal independent subsequence of S is a basis for M.
2) Any independent subsequence of S may be extended to a maximal
independent subsequence of S, and thus to a basis for M.
3) Any independent subsequence of M can be extended to a basis for M.
In particular, M has a free basis, and thus is a free F-module.
Proof The proof of 1) is the same as in the case where S is ﬁnite. Part 2) will
follow from the Hausdorﬀ Maximality Principle. An independent subsequence of S is
contained in a maximal monotonic tower of independent subsequences. The union of
these independent subsequences is still independent, and so the result follows. Part
3) follows from 2) because an independent sequence can always be extended to a
generating sequence.
Theorem Suppose M is an F-module and K ⊂ M is a submodule.
1) K is a summand of M, i.e., ∃ a submodule L of M with K ⊕L = M.
2) If M is f.g., then dim(K) ≤ dim(M) and K = M iﬀ dim(K) = dim(M).
Proof Let T be a basis for K. Extend T to a basis S for M. Then S−T generates
a submodule L with K ⊕L = M. Part 2) follows from 1).
Corollary Q is a summand of R. In other words, ∃ a Q-submodule V ⊂ R
with Q⊕V = R as Q-modules. (See exercise on page 77.)
Proof Q is a ﬁeld, R is a Q-module, and Q is a submodule of R.
Corollary Suppose M is a f.g. F-module, N is an F-module, and f : M → N
is a homomorphism. Then dim(M) = dim(ker(f)) + dim(image(f)).
88 Linear Algebra Chapter 5
Proof Let K = ker(f) and L ⊂ M be a submodule with K ⊕ L = M. Then
f [ L : L →image(f) is an isomorphism.
Exercise Suppose R is a domain with the property that, for R-modules, every
submodule is a summand. Show R is a ﬁeld.
Exercise Find a free Z-module which has a generating set containing no basis.
Exercise The real vector space R
2
is generated by the sequence S =
¦(π, 0), (2, 1), (3, 2)¦. Show there are three maximal independent subsequences of
S, and each is a basis for R
2
. (Row vectors are used here just for convenience.)
The real vector space R
3
is generated by S = ¦(1, 1, 2), (1, 2, 1), (3, 4, 5), (1, 2, 0)¦.
Show there are three maximal independent subsequences of S and each is a basis
for R
3
. You may use determinant.
Square matrices over ﬁelds
This theorem is just a summary of what we have for square matrices over ﬁelds.
Theorem Suppose A ∈ F
n
and f : F
n
→ F
n
is deﬁned by f(B) = AB. Let
v
1
, .., v
n
∈ F
n
be the columns of A, and w
1
, .., w
n
∈ F
n
= F
1,n
be the rows of A.
Then the following are equivalent.
1) ¦v
1
, .., v
n
¦ is independent, i.e., f is injective.
2) ¦v
1
, .., v
n
¦ is a basis for F
n
, i.e., f is an automorphism, i.e., A is
invertible, i.e., [ A [ = 0
¯
.
3) ¦v
1
, .., v
n
¦ generates F
n
, i.e., f is surjective.
1
t
) ¦w
1
, .., w
n
¦ is independent.
2
t
) ¦w
1
, .., w
n
¦ is a basis for F
n
, i.e., A
t
is invertible, i.e., [ A
t
[ = 0
¯
.
3
t
) ¦w
1
, .., w
n
¦ generates F
n
.
Chapter 5 Linear Algebra 89
Proof Except for 1) and 1
t
), this theorem holds for any commutative ring R.
(See the section Relating these concepts to square matrices, pages 81 and 82.)
Parts 1) and 1
t
) follow from the preceding section.
Exercise Add to this theorem more equivalent statements in terms of solutions
of n equations in n unknowns.
Overview Suppose each of X and Y is a set with n elements and f : X →Y is a
function. By the pigeonhole principle, f is injective iﬀ f is bijective iﬀ f is surjective.
Now suppose each of U and V is a vector space of dimension n and f : U → V is
a linear transformation. It follows from the work done so far that f is injective iﬀ
f is bijective iﬀ f is surjective. This shows some of the simple and deﬁnitive nature
of linear algebra.
Exercise Let A = (A
1
, .., A
n
) be an nn matrix over Z with column i = A
i

Z
n
. Let f : Z
n
→ Z
n
be deﬁned by f(B) = AB and
¯
f : R
n
→ R
n
be deﬁned by
¯
f(C) = AC. Show the following are equivalent. (See the exercise on page 79.)
1) f : Z
n
→Z
n
is injective.
2) The sequence (A
1
, .., A
n
) is linearly independent over Z.
3) [A[ = 0.
4)
¯
f : R
n
→R
n
is injective.
5) The sequence (A
1
, .., A
n
) is linearly independent over R.
Rank of a matrix Suppose A ∈ F
m,n
. The row (column) rank of A is deﬁned
to be the dimension of the submodule of F
n
(F
m
) generated by the rows (columns)
of A.
Theorem If C ∈ F
m
and D ∈ F
n
are invertible, then the row (column) rank of
A is the same as the row (column) rank of CAD.
Proof Suppose f : F
n
→ F
m
is deﬁned by f(B) = AB. Each column of A
is a vector in the range F
m
, and we know from page 81 that each f(B) is a linear
90 Linear Algebra Chapter 5
combination of those vectors. Thus the image of f is the submodule of F
m
generated
by the columns of A, and its dimension is the column rank of A. This dimension
is the same as the dimension of the image of g ◦ f ◦ h : F
n
→ F
m
, where h is any
automorphism on F
n
and g is any automorphism on F
m
. This proves the theorem
for column rank. The theorem for row rank follows using transpose.
Theorem If A ∈ F
m,n
, the row rank and the column rank of A are equal. This
number is called the rank of A and is ≤ min¦m, n¦.
Proof By the theorem above, elementary row and column operations change
neither the row rank nor the column rank. By row and column operations, A may be
changed to a matrix H where h
1,1
= = h
t,t
= 1
¯
and all other entries are 0
¯
(see the
ﬁrst exercise on page 59). Thus row rank = t = column rank.
Exercise Suppose A has rank t. Show that it is possible to select t rows and t
columns of A such that the determined t t matrix is invertible. Show that the rank
of A is the largest integer t such that this is possible.
Exercise Suppose A ∈ F
m,n
has rank t. What is the dimension of the solution
set of AX = 0
¯
?
Deﬁnition If N and M are ﬁnite dimensional vector spaces and f : N →M is a
linear transformation, the rank of f is the dimension of the image of f. If f : F
n
→F
m
is given by a matrix A, then the rank of f is the same as the rank of the matrix A.
Geometric Interpretation of Determinant
Suppose V ⊂ R
n
is some nice subset. For example, if n = 2, V might be the
interior of a square or circle. There is a concept of the n-dimensional volume of V .
For n = 1, it is length. For n = 2, it is area, and for n = 3 it is “ordinary volume”.
Suppose A ∈ R
n
and f : R
n
→R
n
is the homomorphism given by A. The volume of
V does not change under translation, i.e., V and V +p have the same volume. Thus
f(V ) and f(V +p) = f(V ) +f(p) have the same volume. In street language, the next
theorem says that “f multiplies volume by the absolute value of its determinant”.
Theorem The n-dimensional volume of f(V ) is ±[A[(the n-dimensional volume
of V ). Thus if [A[ = ±1, f preserves volume.
Chapter 5 Linear Algebra 91
Proof If [A[ = 0, image(f) has dimension < n and thus f(V ) has n-dimensional
volume 0. If [A[ = 0 then A is the product of elementary matrices (see page 59)
and for elementary matrices, the theorem is obvious. The result follows because the
determinant of the composition is the product of the determinants.
Corollary If P is the n-dimensional parallelepiped determined by the columns
v
1
, .. , v
n
of A, then the n-dimensional volume of P is ±[A[.
Proof Let V = [0, 1] [0, 1] = ¦e
1
t
1
+ +e
n
t
n
: 0 ≤ t
i
≤ 1¦. Then
P = f(V ) = ¦v
1
t
1
+ +v
n
t
n
: 0 ≤ t
i
≤ 1¦.
Linear functions approximate diﬀerentiable functions locally
We continue with the special case F = R. Linear functions arise naturally in
business, science, and mathematics. However this is not the only reason that linear
algebra is so useful. It is a central fact that smooth phenomena may be approx-
imated locally by linear phenomena. Without this great simpliﬁcation, the world
of technology as we know it today would not exist. Of course, linear transforma-
tions send the origin to the origin, so they must be adjusted by a translation. As
a simple example, suppose h : R → R is diﬀerentiable and p is a real number. Let
f : R →R be the linear transformation f(x) = h

(p)x. Then h is approximated near
p by g(x) = h(p) +f(x −p) = h(p) +h

(p)(x −p).
Now suppose V ⊂ R
2
is some nice subset and h = (h
1
, h
2
) : V → R
2
is injective
and diﬀerentiable. Deﬁne the Jacobian by J(h)(x, y) =

∂h
1
∂x
∂h
1
∂y
∂h
2
∂x
∂h
2
∂y

and for each
(x, y) ∈ V , let f(x, y) : R
2
→ R
2
be the homomorphism deﬁned by J(h)(x, y).
Then for any (p
1
, p
2
) ∈ V , h is approximated near (p
1
, p
2
) (after translation) by
f(p
1
, p
2
). The area of V is

V
1dxdy. From the previous section we know that
any homomorphism f multiplies area by [ f [. The student may now understand
the following theorem from calculus. (Note that if h is the restriction of a linear
transformation from R
2
to R
2
, this theorem is immediate from the previous section.)
Theorem Suppose the determinant of J(h)(x, y) is non-negative for each
(x, y) ∈ V . Then the area of h(V ) is

V
[ J(h) [ dxdy.
92 Linear Algebra Chapter 5
The Transpose Principle
We now return to the case where F is a ﬁeld (of arbitrary characteristic). F-
modules may also be called vector spaces and submodules may be called subspaces.
The study of R-modules in general is important and complex. However the study of
F-modules is short and simple – every vector space is free and every subspace is a
summand. The core of classical linear algebra is not the study of vector spaces, but
the study of homomorphisms, and in particular, of endomorphisms. One goal is to
show that if f : V → V is a homomorphism with some given property, there exists
a basis of V so that the matrix representing f displays that property in a prominent
manner. The next theorem is an illustration of this.
Theorem Let F be a ﬁeld and n be a positive integer.
1) Suppose V is an n-dimensional vector space and f : V →V is a
homomorphism with [f[ = 0
¯
. Then ∃ a basis of V such that the matrix
representing f has its ﬁrst row zero.
2) Suppose A ∈ F
n
has [A[ = 0
¯
. Then ∃ an invertible matrix C such that
C
−1
AC has its ﬁrst row zero.
3) Suppose V is an n-dimensional vector space and f : V →V is a
homomorphism with [f[ = 0. Then ∃ a basis of V such that the matrix
representing f has its ﬁrst column zero.
4) Suppose A ∈ F
n
has [A[ = 0
¯
. Then ∃ an invertible matrix D such that
D
−1
AD has its ﬁrst column zero.
We ﬁrst wish to show that these 4 statements are equivalent. We know that
1) and 2) are equivalent and also that 3) and 4) are equivalent because change of
basis corresponds to conjugation of the matrix. Now suppose 2) is true and show
4) is true. Suppose [A[ = 0
¯
. Then [A
t
[ = 0
¯
and by 2) ∃ C such that C
−1
A
t
C has
ﬁrst row zero. Thus (C
−1
A
t
C)
t
= C
t
A(C
t
)
−1
has ﬁrst row column zero. The result
follows by deﬁning D = (C
t
)
−1
. Also 4) implies 2).
This is an example of the transpose principle. Loosely stated, it is that theorems
about change of basis correspond to theorems about conjugation of matrices and
theorems about the rows of a matrix correspond to theorems about the columns of a
matrix, using transpose. In the remainder of this chapter, this will be used without
further comment.
Chapter 5 Linear Algebra 93
Proof of the theorem We are free to select any of the 4 parts, and we select
part 3). Since [ f [= 0, f is not injective and ∃ a non-zero v
1
∈ V with f(v
1
) = 0
¯
.
Now v
1
is independent and extends to a basis ¦v
1
, .., v
n
¦. Then the matrix of f w.r.t
this basis has ﬁrst column zero.
Exercise Let A =

3π 6
2π 4

. Find an invertible matrix C ∈ R
2
so that C
−1
AC
has ﬁrst row zero. Also let A =

¸
¸
0 0 0
1 3 4
2 1 4
¸

and ﬁnd an invertible matrix D ∈ R
3
so that D
−1
Exercise Suppose M is an n-dimensional vector space over a ﬁeld F, k is an
integer with 0 < k < n, and f : M → M is an endomorphism of rank k. Show
there is a basis for M so that the matrix representing f has its ﬁrst n −k rows zero.
Also show there is a basis for M so that the matrix representing f has its ﬁrst n −k
columns zero. Work these out directly without using the transpose principle.
Nilpotent Homomorphisms
In this section it is shown that an endomorphism f is nilpotent iﬀ all of its char-
acteristic roots are 0
¯
iﬀ it may be represented by a strictly upper triangular matrix.
Deﬁnition An endomorphism f : V →V is nilpotent if ∃ m with f
m
= 0
¯
. Any
f represented by a strictly upper triangular matrix is nilpotent (see page 56).
Theorem Suppose V is an n-dimensional vector space and f : V → V is a
nilpotent homomorphism. Then f
n
= 0
¯
and ∃ a basis of V such that the matrix
representing f w.r.t. this basis is strictly upper triangular. Thus the characteristic
polynomial of f is CP
f
(x) = x
n
.
Proof Suppose f = 0
¯
is nilpotent. Let t be the largest positive integer with
f
t
= 0
¯
. Then f
t
(V ) ⊂ f
t−1
(V ) ⊂ ⊂ f(V ) ⊂ V . Since f is nilpotent, all of these
inclusions are proper. Therefore t < n and f
n
= 0
¯
. Construct a basis for V by
starting with a basis for f
t
(V ), extending it to a basis for f
t−1
(V ), etc. Then the
matrix of f w.r.t. this basis is strictly upper triangular.
Note To obtain a matrix which is strictly lower triangular, reverse the order of
the basis.
94 Linear Algebra Chapter 5
Exercise Use the transpose principle to write 3 other versions of this theorem.
Theorem Suppose V is an n-dimensional vector space and f : V → V is a
homomorphism. Then f is nilpotent iﬀ CP
f
(x) = x
n
. (See the exercise at the end
of Chapter 4 for the case n = 2.)
Proof Suppose CP
f
(x) = x
n
. For n = 1 this implies f = 0
¯
, so suppose n > 1.
Since the constant term of CP
f
(x) is 0
¯
, the determinant of f is 0
¯
. Thus ∃ a basis
of V such that the matrix A representing f has its ﬁrst column zero. Let B ∈ F
n−1
be the matrix obtained from A by removing its ﬁrst row and ﬁrst column. Now
CP
A
(x) = x
n
= xCP
B
(x). Thus CP
B
(x) = x
n−1
and by induction on n, B is
nilpotent and so ∃ C such that C
−1
BC is strictly upper triangular. Then

¸
¸
¸
¸
¸
¸
¸
1 0 0
0
C
−1

0
¸

¸
¸
¸
¸
¸
¸
¸
0 ∗ ∗

B

0
¸

¸
¸
¸
¸
¸
¸
¸
1 0 0
0
C

0
¸

=

¸
¸
¸
¸
¸
¸
¸
0 ∗ ∗
0
C
−1
BC

0
¸

is strictly upper triangular.
Exercise Suppose F is a ﬁeld, A ∈ F
3
is a strictly lower triangular matrix of
rank 2, and B =

¸
¸
0 0 0
1 0 0
0 1 0
¸

. Using conjugation by elementary matrices, show there
is an invertible matrix C so that C
−1
AC = B. Now suppose V is a 3-dimensional
vector space and f : V →V is a nilpotent endomorphism of rank 2. We know f can
be represented by a strictly lower triangular matrix. Show there is a basis ¦v
1
, v
2
, v
3
¦
for V so that B is the matrix representing f. Also show that f(v
1
) = v
2
, f(v
2
) = v
3
,
and f(v
3
) = 0
¯
. In other words, there is a basis for V of the form ¦v, f(v), f
2
(v)¦
with f
3
(v) = 0
¯
.
Exercise Suppose V is a 3-dimensional vector space and f : V →V is a nilpotent
endomorphism of rank 1. Show there is a basis for V so that the matrix representing
f is

¸
¸
0 0 0
1 0 0
0 0 0
¸

.
Chapter 5 Linear Algebra 95
Eigenvalues
Our standing hypothesis is that V is an n-dimensional vector space over a ﬁeld F
and f : V →V is a homomorphism.
Deﬁnition An element λ ∈ F is an eigenvalue of f if ∃ a non-zero v ∈ V with
f(v) = λv. Any such v is called an eigenvector. E
λ
⊂ V is deﬁned to be the set of
all eigenvectors for λ (plus 0
¯
). Note that E
λ
= ker(λI − f) is a subspace of V . The
next theorem shows the eigenvalues of f are just the characteristic roots of f.
Theorem If λ ∈ F then the following are equivalent.
1) λ is an eigenvalue of f, i.e., (λI −f) : V →V is not injective.
2) [ (λI −f) [= 0
¯
.
3) λ is a characteristic root of f, i.e., a root of the characteristic
polynomial CP
f
(x) =[ (xI −A) [, where A is any matrix representing f.
Proof It is immediate that 1) and 2) are equivalent, so let’s show 2) and 3)
are equivalent. The evaluation map F[x] → F which sends h(x) to h(λ) is a ring
homomorphism (see theorem on page 47). So evaluating (xI − A) at x = λ and
taking determinant gives the same result as taking the determinant of (xI − A) and
evaluating at x = λ. Thus 2) and 3) are equivalent.
The nicest thing you can say about a matrix is that it is similar to a diagonal
matrix. Here is one case where that happens.
Theorem Suppose λ
1
, .., λ
k
are distinct eigenvalues of f, and v
i
is an eigenvector
of λ
i
for 1 ≤ i ≤ k. Then the following hold.
1) ¦v
1
, .., v
k
¦ is independent.
2) If k = n, i.e., if CP
f
(x) = (x −λ
1
) (x −λ
n
), then ¦v
1
, .., v
n
¦ is a
basis for V . The matrix of f w.r.t. this basis is the diagonal matrix whose
(i, i) term is λ
i
.
Proof Suppose ¦v
1
, .., v
k
¦ is dependent. Suppose t is the smallest positive integer
such that ¦v
1
, .., v
t
¦ is dependent, and v
1
r
1
+ +v
t
r
t
= 0
¯
is a non-trivial linear
combination. Note that at least two of the coeﬃcients must be non-zero. Now
(f −λ
t
)(v
1
r
1
+ +v
t
r
t
) = v
1

1
−λ
t
)r
1
+ +v
t−1

t−1
−λ
t
)r
t−1
+0
¯
= 0
¯
is a shorter
96 Linear Algebra Chapter 5
non-trivial linear combination. This is a contradiction and proves 1). Part 2) follows
from 1) because dim(V ) = n.
Exercise Let A =

0 1
−1 0

∈ R
2
. Find an invertible C ∈ C
2
such that
C
−1
AC is diagonal. Show that C cannot be selected in R
2
. Find the characteristic
polynomial of A.
Exercise Suppose V is a 3-dimensional vector space and f : V →V is an endo-
morphism with CP
f
(x) = (x−λ)
3
. Show that (f −λI) has characteristic polynomial
x
3
and is thus a nilpotent endomorphism. Show there is a basis for V so that the
matrix representing f is

¸
¸
λ 0 0
1 λ 0
0 1 λ
¸

,

¸
¸
λ 0 0
1 λ 0
0 0 λ
¸

or

¸
¸
λ 0 0
0 λ 0
0 0 λ
¸

.
We could continue and ﬁnally give an ad hoc proof of the Jordan canonical form,
but in this chapter we prefer to press on to inner product spaces. The Jordan form
will be developed in Chapter 6 as part of the general theory of ﬁnitely generated
modules over Euclidean domains. The next section is included only as a convenient
reference.
Jordan Canonical Form
This section should be just skimmed or omitted entirely. It is unnecessary for the
rest of this chapter, and is not properly part of the ﬂow of the chapter. The basic
facts of Jordan form are summarized here simply for reference.
The statement that a square matrix B over a ﬁeld F is a Jordan block means that
∃ λ ∈ F such that B is a lower triangular matrix of the form
B =

¸
¸
¸
¸
¸
¸
¸
λ 0
1 λ

0 1 λ
¸

. B gives a homomorphism g : F
m
→F
m
with g(e
m
) = λe
m
and g(e
i
) = e
i+1
+λe
i
for 1 ≤ i < m. Note that CP
B
(x) = (x −λ)
m
and so λ is the
only eigenvalue of B, and B satisﬁes its characteristic polynomial, i.e., CP
B
(B) = 0
¯
.
Chapter 5 Linear Algebra 97
Deﬁnition A matrix D ∈ F
n
is in Jordan form if ∃ Jordan blocks B
1
, .. , B
t
such
that D =

¸
¸
¸
¸
¸
¸
¸
¸
¸
B
1
B
2
0

0
B
t
¸

. Suppose D is of this form and B
i
∈ F
n
i
has
eigenvalue λ
i
. Then n
1
+ +n
t
= n and CP
D
(x) = (x−λ
1
)
n
1
(x −λ
t
)
nt
. Note that
a diagonal matrix is a special case of Jordan form. D is a diagonal matrix iﬀ each
n
i
= 1, i.e., iﬀ each Jordan block is a 1 1 matrix.
Theorem If A ∈ F
n
, the following are equivalent.
1) ∃ an invertible C ∈ F
n
such that C
−1
AC is in Jordan form.
2) ∃ λ
1
, .., λ
n
∈ F (not necessarily distinct) such that CP
A
(x) = (x −λ
1
)
(x −λ
n
). (In this case we say that all the eigenvalues of A belong to F.)
Theorem Jordan form (when it exists) is unique. This means that if A and D are
similar matrices in Jordan form, they have the same Jordan blocks, except possibly
in diﬀerent order.
The reader should use the transpose principle to write three other versions of the
ﬁrst theorem. Also note that we know one special case of this theorem, namely that
if A has n distinct eigenvalues in F, then A is similar to a diagonal matrix. Later on
it will be shown that if A is a symmetric real matrix, then A is similar to a diagonal
matrix.
Let’s look at the classical case A ∈ R
n
. The complex numbers are algebraically
closed. This means that CP
A
(x) will factor completely in C[x], and thus ∃ C ∈ C
n
with C
−1
AC in Jordan form. C may be selected to be in R
n
iﬀ all the eigenvalues
of A are real.
Exercise Find all real matrices in Jordan form that have the following charac-
teristic polynomials: x(x − 2), (x − 2)
2
, (x − 2)(x − 3)(x − 4), (x − 2)(x − 3)
2
,
(x −2)
2
(x −3)
2
, (x −2)(x −3)
3
.
Exercise Suppose D ∈ F
n
is in Jordan form and has characteristic polynomial
a
0
+a
1
x + +x
n
. Show a
0
I +a
1
D + +D
n
= 0
¯
, i.e., show CP
D
(D) = 0
¯
.
98 Linear Algebra Chapter 5
Exercise (Cayley-Hamilton Theorem) Suppose E is a ﬁeld and A ∈ E
n
.
Assume the theorem that there is a ﬁeld F containing E such that CP
A
(x) factors
completely in F[x]. Thus ∃ an invertible C ∈ F
n
such that D = C
−1
AC is in Jordan
form. Use this to show CP
A
(A) = 0
¯
. (See the second exercise on page 66.)
Exercise Suppose A ∈ F
n
is in Jordan form. Show A is nilpotent iﬀ A
n
= 0
¯
iﬀ CP
A
(x) = x
n
. (Note how easy this is in Jordan form.)
Inner Product Spaces
The two most important ﬁelds for mathematics and science in general are the
real numbers and the complex numbers. Finitely generated vector spaces over R or
C support inner products and are thus geometric as well as algebraic objects. The
theories for the real and complex cases are quite similar, and both could have been
treated here. However, for simplicity, attention is restricted to the case F = R.
In the remainder of this chapter, the power and elegance of linear algebra become
transparent for all to see.
Deﬁnition Suppose V is a real vector space. An inner product (or dot product)
on V is a function V V →R which sends (u, v) to u v and satisﬁes
1) (u
1
r
1
+u
2
r
2
) v = (u
1
v)r
1
+ (u
2
v)r
2
for all u
1
, u
2
, v ∈ V
v (u
1
r
1
+u
2
r
2
) = (v u
1
)r
1
+ (v u
2
)r
2
and r
1
, r
2
∈ R.
2) u v = v u for all u, v ∈ V .
3) u u ≥ 0 and u u = 0 iﬀ u = 0
¯
for all u ∈ V .
Theorem Suppose V has an inner product.
1) If v ∈ V , f : V →R deﬁned by f(u) = u v is a homomorphism.
Thus 0
¯
v = 0.
2) Schwarz’ inequality. If u, v ∈ V, (u v)
2
≤ (u u)(v v).
Proof of 2) Let a =

v v and b =

u u. If a or b is 0, the result is obvious.
Suppose neither a nor b is 0. Now 0 ≤ (ua ±vb) (ua ±vb) = (u u)a
2
±2ab(u v)+
(v v)b
2
= b
2
a
2
±2ab(u v)+a
2
b
2
. Dividing by 2ab yields 0 ≤ ab±(u v) or [ u v [≤ ab.
Chapter 5 Linear Algebra 99
Theorem Suppose V has an inner product. Deﬁne the norm or length of a vector
v by |v| =

v v. The following properties hold.
1) |v| = 0 iﬀ v = 0
¯
.
2) |vr| = |v| [ r [.
3) [ u v [ ≤ |u||v|. (Schwarz’ inequality)
4) |u +v| ≤ |u| +|v|. (The triangle inequality)
Proof of 4) |u + v|
2
= (u + v) (u + v) = |u|
2
+ 2(u v) + |v|
2
≤ |u|
2
+
2|u||v| +|v|
2
= (|u| +|v|)
2
.
Deﬁnition An Inner Product Space (IPS) is a real vector space with an
inner product. Suppose V is an IPS. A sequence ¦v
1
, .., v
n
¦ is orthogonal provided
v
i
v
j
= 0 when i = j. The sequence is orthonormal if it is orthogonal and each
vector has length 1, i.e., v
i
v
j
= δ
i,j
for 1 ≤ i, j ≤ n.
Theorem If S = ¦v
1
, .., v
n
¦ is an orthogonal sequence of non-zero vectors in an
IPS V, then S is independent. Furthermore

v
1
|v
1
|
, ,
v
n
|v
n
|
¸
is orthonormal.
Proof Suppose v
1
r
1
+ +v
n
r
n
= 0
¯
. Then 0 = (v
1
r
1
+ +v
n
r
n
) v
i
= r
i
(v
i
v
i
)
and thus r
i
= 0. Thus S is independent. The second statement is transparent.
It is easy to deﬁne an inner product, as is shown by the following theorem.
Theorem Suppose V is a real vector space with a basis S = ¦v
1
, .., v
n
¦. Then
there is a unique inner product on V which makes S an orthornormal basis. It is
given by the formula (v
1
r
1
+ +v
n
r
n
) (v
1
s
1
+ +v
n
s
n
) = r
1
s
1
+ +r
n
s
n
.
Convention R
n
will be assumed to have the standard inner product deﬁned by
(r
1
, .., r
n
)
t
(s
1
, .., s
n
)
t
= r
1
s
1
+ +r
n
s
n
. S = ¦e
1
, .., e
n
¦ will be called the canonical
or standard orthonormal basis (see page 72). The next theorem shows that this
inner product has an amazing geometry.
Theorem If u, v ∈ R
n
, u v = |u||v| cos Θ where Θ is the angle between u
100 Linear Algebra Chapter 5
and v.
Proof Let u = (r
1
, .., r
n
) and v = (s
1
, .., s
n
). By the law of cosines |u − v|
2
=
|u|
2
+|v|
2
−2|u||v| cos Θ. So (r
1
−s
1
)
2
+ +(r
n
−s
n
)
2
= r
2
1
+ +r
2
n
+s
2
1
+
+s
2
n
−2|u||v| cos Θ. Thus r
1
s
1
+ +r
n
s
n
= |u||v| cos Θ.
Exercise This is a simple exercise to observe that hyperplanes in R
n
are cosets.
Suppose f : R
n
→R is a non-zero homomorphism given by a matrix A = (a
1
, .., a
n
) ∈
R
1,n
. Then L = ker(f) is the set of all solutions to a
1
x
1
+ +a
n
x
n
= 0, i.e., the
set of all vectors perpendicular to A. Now suppose b ∈ R and C =

¸
¸
¸
¸
c
1

c
n
¸

∈ R
n
has f(C) = b. Then f
−1
(b) is the set of all solutions to a
1
x
1
+ +a
n
x
n
= b which
is the coset L+C, and this the set of all solutions to a
1
(x
1
−c
1
) + +a
n
(x
n
−c
n
) = 0.
Gram-Schmidt orthonormalization
Theorem (Fourier series) Suppose W is an IPS with an orthonormal basis
¦w
1
, .., w
n
¦. Then if v ∈ W, v = w
1
(v w
1
) + +w
n
(v w
n
).
Proof v = w
1
r
1
+ +w
n
r
n
and v w
i
= (w
1
r
1
+ +w
n
r
n
) w
i
= r
i
Theorem Suppose W is an IPS, Y ⊂ W is a subspace with an orthonormal basis
¦w
1
, .., w
k
¦, and v ∈ W−Y . Deﬁne the projection of v onto Y by p(v) = w
1
(v w
1
)+
+w
k
(vw
k
), and let w = v−p(v). Then (ww
i
) = (v−w
1
(vw
1
)−w
k
(vw
k
))w
i
= 0.
Thus if w
k+1
=
w
|w|
, then ¦w
1
, .., w
k+1
¦ is an orthonormal basis for the subspace
generated by ¦w
1
, .., w
k
, v¦. If ¦w
1
, .., w
k
, v¦ is already orthonormal, w
k+1
= v.
Theorem (Gram-Schmidt) Suppose W is an IPS with a basis ¦v
1
, .., v
n
¦.
Then W has an orthonormal basis ¦w
1
, .., w
n
¦. Moreover, any orthonormal sequence
in W extends to an orthonormal basis of W.
Proof Let w
1
=
v
1
|v
1
|
. Suppose inductively that ¦w
1
, .., w
k
¦ is an orthonormal
basis for Y , the subspace generated by ¦v
1
, .., v
k
¦. Let w = v
k+1
− p(v
k+1
) and
Chapter 5 Linear Algebra 101
w
k+1
=
w
|w|
. Then by the previous theorem, ¦w
1
, .., w
k+1
¦ is an orthonormal basis
for the subspace generated by ¦w
1
, .., w
k
, v
k+1
¦. In this manner an orthonormal basis
for W is constructed. Notice that this construction deﬁnes a function h which sends
a basis for W to an orthonormal basis for W (see topology exercise on page 103).
Now suppose W has dimension n and ¦w
1
, .., w
k
¦ is an orthonormal sequence in
W. Since this sequence is independent, it extends to a basis ¦w
1
, .., w
k
, v
k+1
, .., v
n
¦.
The process above may be used to modify this to an orthonormal basis ¦w
1
, .., w
n
¦.
Exercise Let f : R
3
→ R be the homomorphism deﬁned by the matrix (2,1,3).
Find an orthonormal basis for the kernel of f. Find the projection of (e
1
+ e
2
) onto
ker(f). Find the angle between e
1
+e
2
and the plane ker(f).
Exercise Let W = R
3
have the standard inner product and Y ⊂ W be the
subspace generated by ¦w
1
, w
2
¦ where w
1
= (1, 0, 0)
t
and w
2
= (0, 1, 0)
t
. W is
generated by the sequence ¦w
1
, w
2
, v¦ where v = (1, 2, 3)
t
. As in the ﬁrst theorem
of this section, let w = v − p(v), where p(v) is the projection of v onto Y , and set
w
3
=
w
|w|
. Find w
3
and show that for any t with 0 ≤ t ≤ 1, ¦w
1
, w
2
, (1 −t)v +tw
3
¦
is a basis for W. This is a key observation for an exercise on page 103 showing O(n)
is a deformation retract of GL
n
(R).
Isometries Suppose each of U and V is an IPS. A homomorphism f : U → V
is said to be an isometry provided it is an isomorphism and for any u
1
, u
2
in U,
(u
1
u
2
)
U
= (f(u
1
) f(u
2
))
V
.
Theorem Suppose each of U and V is an n-dimensional IPS, ¦u
1
, .., u
n
¦ is an
orthonormal basis for U, and f : U →V is a homomorphism. Then f is an isometry
iﬀ ¦f(u
1
), .., f(u
n
)¦ is an orthonormal sequence in V .
Proof Isometries certainly preserve orthonormal sequences. So suppose T =
¦f(u
1
), .., f(u
n
)¦ is an orthonormal sequence in V . Then T is independent and thus
T is a basis for V and thus f is an isomorphism (see the second theorem on page 79).
It is easy to check that f preserves inner products.
We now come to one of the deﬁnitive theorems in linear algebra. It is that, up to
isometry, there is only one inner product space for each dimension.
102 Linear Algebra Chapter 5
Theorem Suppose each of U and V is an n-dimensional IPS. Then ∃ an isometry
f : U →V. In particular, U is isometric to R
n
with its standard inner product.
Proof There exist orthonormal bases ¦u
1
, .., u
n
¦ for U and ¦v
1
, .., v
n
¦ for V .
By the ﬁrst theorem on page 79, there exists a homomorphism f : U → V with
f(u
i
) = v
i
, and by the previous theorem, f is an isometry.
Exercise Let f : R
3
→ R be the homomorphism deﬁned by the matrix (2,1,3).
Find a linear transformation h : R
2
→R
3
which gives an isometry from R
2
to ker(f).
Orthogonal Matrices
As noted earlier, linear algebra is not so much the study of vector spaces as it is
the study of endomorphisms. We now wish to study isometries from R
n
to R
n
.
We know from a theorem on page 90 that an endomorphism preserves volume iﬀ
its determinant is ±1. Isometries preserve inner product, and thus preserve angle and
distance, and so certainly preserve volume.
Theorem Suppose A ∈ R
n
and f : R
n
→ R
n
is the homomorphism deﬁned by
f(B) = AB. Then the following are equivalent.
1) The columns of A form an orthonormal basis for R
n
, i.e., A
t
A = I.
2) The rows of A form an orthonormal basis for R
n
, i.e., AA
t
= I.
3) f is an isometry.
Proof A left inverse of a matrix is also a right inverse (see the exercise on
page 64). Thus 1) and 2) are equivalent because each of them says A is invert-
ible with A
−1
= A
t
. Now ¦e
1
, .., e
n
¦ is the canonical orthonormal basis for R
n
, and
f(e
i
) is column i of A. Thus by the previous section, 1) and 3) are equivalent.
Deﬁnition If A ∈ R
n
satisﬁes these three conditions, A is said to be orthogonal.
The set of all such A is denoted by O(n), and is called the orthogonal group.
Theorem
1) If A is orthogonal, [ A [= ±1.
2) If A is orthogonal, A
−1
is orthogonal. If A and C are orthogonal, AC is
orthogonal. Thus O(n) is a multiplicative subgroup of GL
n
(R).
Chapter 5 Linear Algebra 103
3) Suppose A is orthogonal and f is deﬁned by f(B) = AB. Then f preserves
distances and angles. This means that if u, v ∈ R
n
, |u −v| =
|f(u)−f(v)| and the angle between u and v is equal to the angle between
f(u) and f(v).
Proof Part 1) follows from [A[
2
= [A[ [A
t
[ = [I[ = 1. Part 2) is imme-
diate, because isometries clearly form a subgroup of the multiplicative group of
all automorphisms. For part 3) assume f : R
n
→ R
n
is an isometry. Then
|u − v|
2
= (u − v) (u − v) = f(u − v) f(u − v) = |f(u − v)|
2
= |f(u) − f(v)|
2
.
The proof that f preserves angles follows from u v = |u||v|cosΘ.
Exercise Show that if A ∈ O(2) has [A[ = 1, then A =

cosΘ −sinΘ
sinΘ cosΘ

for
some number Θ. (See the exercise on page 56.)
Exercise (topology) Let R
n
≈ R
n
2
have its usual metric topology. This means
a sequence of matrices ¦A
i
¦ converges to A iﬀ it converges coordinatewise. Show
GL
n
(R) is an open subset and O(n) is closed and compact. Let h : GL
n
(R) →
O(n) be deﬁned by Gram-Schmidt. Show H : GL
n
(R) [0, 1] → GL
n
(R) deﬁned
by H(A, t) = (1 −t)A+th(A) is a deformation retract of GL
n
(R) to O(n).
Diagonalization of Symmetric Matrices
We continue with the case F = R. Our goals are to prove that, if A is a symmetric
matrix, all of its eigenvalues are real and that ∃ an orthogonal matrix C such that
C
−1
AC is diagonal. As background, we ﬁrst note that symmetric is the same as
Theorem Suppose A ∈ R
n
and u, v ∈ R
n
. Then (A
t
u) v = u (Av).
Proof If y, z ∈ R
n
, then the dot product y z, is the matrix product y
t
z, and
matrix multiplication is associative. Thus (A
t
u) v = (u
t
A)v = u
t
(Av) = u (Av).
Deﬁnition Suppose A ∈ R
n
. A is said to be symmetric provided A
t
= A. Note
that any diagonal matrix is symmetric. A is said to be self-adjoint if (Au)v = u(Av)
for all u, v ∈ R
n
. The next theorem is just an exercise using the previous theorem.
Theorem A is symmetric iﬀ A is self-adjoint.
104 Linear Algebra Chapter 5
Theorem Suppose A ∈ R
n
is symmetric. Then ∃ real numbers λ
1
, .., λ
n
(not
necessarily distinct) such that CP
A
(x) = (x − λ
1
)(x − λ
2
) (x − λ
n
). That is, all
the eigenvalues of A are real.
Proof We know CP
A
(x) factors into linears over C. If µ = a + bi is a complex
number, its conjugate is deﬁned by ¯ µ = a −bi. If h : C →C is deﬁned by h(µ) = ¯ µ,
then h is a ring isomorphism which is the identity on R. If w = (a
i,j
) is a complex
matrix or vector, its conjugate is deﬁned by ¯ w = (¯ a
i,j
). Since A ∈ R
n
is a real
symmetric matrix, A = A
t
=
¯
A
t
. Now suppose λ is a complex eigenvalue of A
and v ∈ C
n
is an eigenvector with Av = λv. Then λ(v
t
¯ v) = (λv)
t
¯ v = (Av)
t
¯ v =
(v
t
A)¯ v = v
t
(A¯ v) = v
t
(Av) = v
t
(λv) =
¯
λ(v
t
¯ v). Thus λ =
¯
λ and λ ∈ R. Or
you can deﬁne a complex inner product on C
n
by (w v) = w
t
¯ v. The proof then
reads as λ(v v) = (λv v) = (Av v) = (v Av) = (v λv) =
¯
λ(v v). Either way,
λ is a real number.
We know that eigenvectors belonging to distinct eigenvalues are linearly indepen-
dent. For symmetric matrices, we show more, namely that they are perpendicular.
Theorem Suppose A is symmetric, λ
1
, λ
2
∈ R are distinct eigenvalues of A, and
Au = λ
1
u and Av = λ
2
v. Then u v = 0.
Proof λ
1
(u v) = (Au) v = u (Av) = λ
2
(u v).
Review Suppose A ∈ R
n
and f : R
n
→ R
n
is deﬁned by f(B) = AB. Then A
represents f w.r.t. the canonical orthonormal basis. Let S = ¦v
1
, .., v
n
¦ be another
basis and C ∈ R
n
be the matrix with v
i
as column i. Then C
−1
AC is the matrix
representing f w.r.t. S. Now S is an orthonormal basis iﬀ C is an orthogonal matrix.
Summary Representing f w.r.t. an orthonormal basis is the same as conjugating
A by an orthogonal matrix.
Theorem Suppose A ∈ R
n
and C ∈ O(n). Then A is symmetric iﬀ C
−1
AC
is symmetric.
Proof Suppose A is symmetric. Then (C
−1
AC)
t
= C
t
A(C
−1
)
t
= C
−1
AC.
The next theorem has geometric and physical implications, but for us, just the
incredibility of it all will suﬃce.
Chapter 5 Linear Algebra 105
Theorem If A ∈ R
n
, the following are equivalent.
1) A is symmetric.
2) ∃ C ∈ O(n) such that C
−1
AC is diagonal.
Proof By the previous theorem, 2) ⇒ 1). Show 1) ⇒ 2). Suppose A is a
symmetric 22 matrix. Let λ be an eigenvalue for A and ¦v
1
, v
2
¦ be an orthonormal
basis for R
2
with Av
1
= λv
1
. Then w.r.t this basis, the transformation determined
by A is represented by

λ b
0 d

. Since this matrix is symmetric, b = 0.
Now suppose by induction that the theorem is true for symmetric matrices in
R
t
for t < n, and suppose A is a symmetric n n matrix. Denote by λ
1
, .., λ
k
the
distinct eigenvalues of A, k ≤ n. If k = n, the proof is immediate, because then there
is a basis of eigenvectors of length 1, and they must form an orthonormal basis. So
suppose k < n. Let v
1
, .., v
k
be eigenvectors for λ
1
, .., λ
k
with each | v
i
|= 1. They
may be extended to an orthonormal basis v
1
, .., v
n
. With respect to this basis, the
transformation determined by A is represented by

¸
¸
¸
¸
¸
¸
¸
¸
¸

¸
¸
¸
¸
λ
1

λ
k
¸

(B)
(0) (D)
¸

.
Since this is a symmetric matrix, B = 0 and D is a symmetric matrix of smaller
size. By induction, ∃ an orthogonal C such that C
−1
DC is diagonal. Thus conjugating
by

I 0
0 C

makes the entire matrix diagonal.
This theorem is so basic we state it again in diﬀerent terminology. If V is an IPS, a
linear transformation f : V →V is said to be self-adjoint provided (uf(v)) = (f(u)v)
for all u, v ∈ V .
Theorem If V is an n-dimensional IPS and f : V →V is a linear transformation,
then the following are equivalent.
2) ∃ an orthonormal basis ¦v
1
, ..., v
n
¦ for V with each
v
i
an eigenvector of f.
106 Linear Algebra Chapter 5
Exercise Let A =

2 2
2 2

. Find an orthogonal C such that C
−1
AC is diagonal.
Do the same for A =

2 1
1 2

.
Exercise Suppose A, D ∈ R
n
are symmetric. Under what conditions are A and D
similar? Show that, if A and D are similar, ∃ an orthogonal C such that D = C
−1
AC.
Exercise Suppose V is an n-dimensional real vector space. We know that V is
isomorphic to R
n
. Suppose f and g are isomorphisms from V to R
n
and A is a subset
of V . Show that f(A) is an open subset of R
n
iﬀ g(A) is an open subset of R
n
. This
shows that V , an algebraic object, has a god-given topology. Of course, if V has
an inner product, it automatically has a metric, and this metric will determine that
same topology. Finally, suppose V and W are ﬁnite-dimensional real vector spaces
and h : V →W is a linear transformation. Show that h is continuous.
Exercise Deﬁne E : C
n
→C
n
by E(A) = e
A
= I +A+(1/2!)A
2
+. This series
converges and thus E is a well deﬁned function. If AB = BA, then E(A + B) =
E(A)E(B). Since A and −A commute, I = E(0
¯
) = E(A − A) = E(A)E(−A), and
thus E(A) is invertible with E(A)
−1
= E(−A). Furthermore E(A
t
) = E(A)
t
, and
if C is invertible, E(C
−1
AC) = C
−1
E(A)C. Now use the results of this section to
prove the statements below. (For part 1, assume the Jordan form, i.e., assume any
A ∈ C
n
is similar to a lower triangular matrix.)
1) If A ∈ C
n
, then [ e
A
[= e
trace(A)
. Thus if A ∈ R
n
, [ e
A
[= 1
iﬀ trace(A) = 0.
2) ∃ a non-zero matrix N ∈ R
2
with e
N
= I.
3) If N ∈ R
n
is symmetric, then e
N
= I iﬀ N = 0
¯
.
4) If A ∈ R
n
and A
t
= −A, then e
A
∈ O(n).
Chapter 6
Appendix
The ﬁve previous chapters were designed for a year undergraduate course in algebra.
In this appendix, enough material is added to form a basic ﬁrst year graduate course.
Two of the main goals are to characterize ﬁnitely generated abelian groups and to
prove the Jordan canonical form. The style is the same as before, i.e., everything is
right down to the nub. The organization is mostly a linearly ordered sequence except
for the last two sections on determinants and dual spaces. These are independent
sections added on at the end.
Suppose R is a commutative ring. An R-module M is said to be cyclic if it can
be generated by one element, i.e., M ≈ R/I where I is an ideal of R. The basic
theorem of this chapter is that if R is a Euclidean domain and M is a ﬁnitely generated
R-module, then M is the sum of cyclic modules. Thus if M is torsion free, it is a
free R-module. Since Z is a Euclidean domain, ﬁnitely generated abelian groups
are the sums of cyclic groups – one of the jewels of abstract algebra.
Now suppose F is a ﬁeld and V is a ﬁnitely generated F-module. If T : V →V is
a linear transformation, then V becomes an F[x]-module by deﬁning vx = T(v). Now
F[x] is a Euclidean domain and so V
F[x]
is the sum of cyclic modules. This classical
and very powerful technique allows an easy proof of the canonical forms. There is a
basis for V so that the matrix representing T is in Rational canonical form. If the
characteristic polynomial of T factors into the product of linear polynomials, then
there is a basis for V so that the matrix representing T is in Jordan canonical form.
This always holds if F = C. A matrix in Jordan form is a lower triangular matrix
with the eigenvalues of T displayed on the diagonal, so this is a powerful concept.
In the chapter on matrices, it is stated without proof that the determinant of the
product is the product of the determinants. A proof of this, which depends upon the
classiﬁcation of certain types of alternating multilinear forms, is given in this chapter.
The ﬁnal section gives the fundamentals of dual spaces.
107
108 Appendix Chapter 6
The Chinese Remainder Theorem
On page 50 in the chapter on rings, the Chinese Remainder Theorem was proved
for the ring of integers. In this section this classical topic is presented in full generality.
Surprisingly, the theorem holds even for non-commutative rings.
Deﬁnition Suppose R is a ring and A
1
, A
2
, ..., A
m
are ideals of R. Then the sum
A
1
+ A
2
+ + A
m
is the set of all a
1
+ a
2
+ + a
m
with a
i
∈ A
i
. The product
A
1
A
2
A
m
is the set of all ﬁnite sums of elements a
1
a
2
a
m
with a
i
∈ A
i
. Note
that the sum and product of ideals are ideals and A
1
A
2
A
m
⊂ (A
1
∩A
2
∩ ∩A
m
).
Deﬁnition Ideals A and B of R are said to be comaximal if A+B = R.
Theorem If A and B are ideals of a ring R, then the following are equivalent.
1) A and B are comaximal.
2) ∃ a ∈ A and b ∈ B with a +b = 1
¯
.
3) π(A) = R/B where π : R →R/B is the projection.
Theorem If A
1
, A
2
, ..., A
m
and B are ideals of R with A
i
and B comaximal for
each i, then A
1
A
2
A
m
and B are comaximal. Thus A
1
∩ A
2
∩ ∩ A
m
and B
are comaximal.
Proof Consider π : R →R/B. Then π(A
1
A
2
A
m
) = π(A
1
)π(A
2
) π(A
m
) =
(R/B)(R/B) (R/B) = R/B.
Chinese Remainder Theorem Suppose A
1
, A
2
, ..., A
n
are pairwise comaximal
ideals of R, with each A
i
= R. Then the natural map π : R →R/A
1
R/A
2

R/A
n
is a surjective ring homomorphism with kernel A
1
∩ A
2
∩ ∩ A
n
.
Proof There exists a
i
∈ A
i
and b
i
∈ A
1
A
2
A
i−1
A
i+1
A
n
with a
i
+b
i
= 1
¯
. Note
that π(b
i
) = (0, .., 0, 1
¯
i
, 0, .., 0). If (r
1
+ A
1
, r
2
+ A
2
, ..., r
n
+ A
n
) is an element of the
range, it is the image of r
1
b
1
+r
2
b
2
++r
n
b
n
= r
1
(1
¯
−a
1
)+r
2
(1
¯
−a
2
)++r
n
(1
¯
−a
n
).
Theorem If R is commutative and A
1
, A
2
, ..., A
n
are pairwise comaximal ideals
of R, then A
1
A
2
A
n
= A
1
∩ A
2
∩ ∩ A
n
.
Proof for n = 2. Show A
1
∩A
2
⊂ A
1
A
2
. ∃ a
1
∈ A
1
and a
2
∈ A
2
with a
1
+a
2
= 1
¯
.
If c ∈ A
1
∩ A
2
, then c = c(a
1
+a
2
) ∈ A
1
A
2
.
Chapter 6 Appendix 109
Prime and Maximal Ideals and UFD
s
In the ﬁrst chapter on background material, it was shown that Z is a unique
factorization domain. Here it will be shown that this property holds for any principle
ideal domain. Later on it will be shown that every Euclidean domain is a principle
ideal domain. Thus every Euclidean domain is a unique factorization domain.
Deﬁnition Suppose R is a commutative ring and I ⊂ R is an ideal.
I is prime means I = R and if a, b ∈ R have ab ∈ I, then a or b ∈ I.
I is maximal means I = R and there are no ideals properly between I and R.
Theorem 0
¯
is a prime ideal of R iﬀ R is
0
¯
is a maximal ideal of R iﬀ R is
Theorem Suppose J ⊂ R is an ideal, J = R.
J is a prime ideal iﬀ R/J is
J is a maximal ideal iﬀ R/J is
Corollary Maximal ideals are prime.
Proof Every ﬁeld is a domain.
Theorem If a ∈ R is not a unit, then ∃ a maximal ideal I of R with a ∈ I.
Proof This is a classical application of the Hausdorﬀ Maximality Principle. Con-
sider ¦J : J is an ideal of R containing a with J = R¦. This collection contains a
maximal monotonic collection ¦V
t
¦
t∈T
. The ideal V =
¸
t∈T
V
t
does not contain 1
¯
and
thus is not equal to R. Therefore V is equal to some V
t
and is a maximal ideal
containing a.
Note To properly appreciate this proof, the student should work the exercise in
group theory at the end of this section (see page 114).
Deﬁnition Suppose R is a domain and a, b ∈ R. Then we say a ∼ b iﬀ there
exists a unit u with au = b. Note that ∼ is an equivalence relation. If a ∼ b, then a
110 Appendix Chapter 6
and b are said to be associates.
Examples If R is a domain, the associates of 1
¯
are the units of R, while the only
associate of 0
¯
is 0
¯
itself. If n ∈ Z is not zero, then its associates are n and −n.
If F is a ﬁeld and g ∈ F[x] is a non-zero polynomial, then the associates of g are
all cg where c is a non-zero constant.
The following theorem is elementary, but it shows how associates ﬁt into the
scheme of things. An element a divides b (a[b) if ∃! c ∈ R with ac = b.
Theorem Suppose R is a domain and a, b ∈ (R − 0
¯
). Then the following are
equivalent.
1) a ∼ b.
2) a[b and b[a.
3) aR = bR.
Parts 1) and 3) above show there is a bijection from the associate classes of R to
the principal ideals of R. Thus if R is a PID, there is a bijection from the associate
classes of R to the ideals of R. If an element of a domain generates a non-zero prime
ideal, it is called a prime element.
Deﬁnition Suppose R is a domain and a ∈ R is a non-zero non-unit.
1) a is irreducible if it does not factor, i.e., a = bc ⇒ b or c is a unit.
2) a is prime if it generates a prime ideal, i.e., a[bc ⇒ a[b or a[c.
Note If a is a prime and a[c
1
c
2
c
n
, then a[c
i
for some i. This follows from the
deﬁnition and induction on n. If each c
j
is irreducible, then a ∼ c
i
for some i.
Note If a ∼ b, then a is irreducible (prime) iﬀ b is irreducible (prime). In other
words, if a is irreducible (prime) and u is a unit, then au is irreducible (prime).
Note a is prime ⇒a is irreducible. This is immediate from the deﬁnitions.
Theorem Factorization into primes is unique up to order and associates, i.e., if
d = b
1
b
2
b
n
= c
1
c
2
c
m
with each b
i
and each c
i
prime, then n = m and for some
permutation σ of the indices, b
i
and c
σ(i)
are associates for every i. Note also ∃ a unit
u and primes p
1
, p
2
, . . . , p
t
where no two are associates and du = p
s
1
1
p
s
2
2
p
st
t
.
Chapter 6 Appendix 111
Proof This follows from the notes above.
Deﬁnition R is a factorization domain (FD) means that R is a domain and if a is
a non-zero non-unit element of R, then a factors into a ﬁnite product of irreducibles.
Deﬁnition R is a unique factorization domain (UFD) means R is a FD in which
factorization is unique (up to order and associates).
Theorem If R is a UFD and a is a non-zero non-unit of R, then a is irreducible
⇔ a is prime. Thus in a UFD, elements factor as the product of primes.
Proof Suppose R is a UFD, a is an irreducible element of R, and a[bc. If either
b or c is a unit or is zero, then a divides one of them, so suppose each of b and c is
a non-zero non-unit element of R. There exists an element d with ad = bc. Each of
b and c factors as the product of irreducibles and the product of these products is
the factorization of bc. It follows from the uniqueness of the factorization of ad = bc,
that one of these irreducibles is an associate of a, and thus a[b or a[c. Therefore
the element a is a prime.
Theorem Suppose R is a FD. Then the following are equivalent.
1) R is a UFD.
2) Every irreducible element of R is prime, i.e., a irreducible ⇔ a is prime.
Proof We already know 1) ⇒ 2). Part 2) ⇒ 1) because factorization into primes
is always unique.
This is a revealing and useful theorem. If R is a FD, then R is a UFD iﬀ each
irreducible element generates a prime ideal. Fortunately, principal ideal domains
have this property, as seen in the next theorem.
Theorem Suppose R is a PID and a ∈ R is non-zero non-unit. Then the following
are equivalent.
1) aR is a maximal ideal.
2) aR is a prime ideal, i.e., a is a prime element.
3) a is irreducible.
Proof Every maximal ideal is a prime ideal, so 1) ⇒ 2). Every prime element is
an irreducible element, so 2) ⇒ 3). Now suppose a is irreducible and show aR is a
maximal ideal. If I is an ideal containing aR, ∃ b ∈ R with I = bR. Since b divides
a, the element b is a unit or an associate of a. This means I = R or I = aR.
112 Appendix Chapter 6
Our goal is to prove that a PID is a UFD. Using the two theorems above, it
only remains to show that a PID is a FD. The proof will not require that ideals be
principally generated, but only that they be ﬁnitely generated. This turns out to
be equivalent to the property that any collection of ideals has a “maximal” element.
We shall see below that this is a useful concept which ﬁts naturally into the study of
unique factorization domains.
Theorem Suppose R is a commutative ring. Then the following are equivalent.
1) If I ⊂ R is an ideal, ∃ a ﬁnite set ¦a
1
, a
2
, ..., a
n
¦ ⊂ R such that I =
a
1
R +a
2
R + +a
n
R, i.e., each ideal of R is ﬁnitely generated.
2) Any non-void collection of ideals of R contains an ideal I which is maximal in
the collection. This means if J is an ideal in the collection with J ⊃ I, then
J = I. (The ideal I is maximal only in the sense described. It need not contain
all the ideals of the collection, nor need it be a maximal ideal of the ring R.)
3) If I
1
⊂ I
2
⊂ I
3
⊂ ... is a monotonic sequence of ideals, ∃ t
0
≥ 1 such that I
t
= I
t
0
for all t ≥ t
0
.
Proof Suppose 1) is true and show 3). The ideal I = I
1
∪ I
2
∪ . . . is ﬁnitely
generated and ∃ t
0
≥ 1 such that I
t
0
contains those generators. Thus 3) is true. Now
suppose 2) is true and show 1). Let I be an ideal of R, and consider the collection
of all ﬁnitely generated ideals contained in I. By 2) there is a maximal one, and it
must be I itself, and thus 1) is true. We now have 2)⇒1)⇒3), so suppose 2) is false
and show 3) is false. So there is a collection of ideals of R such that any ideal in the
collection is properly contained in another ideal of the collection. Thus it is possible
to construct a sequence of ideals I
1
⊂ I
2
⊂ I
3
. . . with each properly contained in
the next, and therefore 3) is false. (Actually this construction requires the Hausdorﬀ
Maximality Principle or some form of the Axiom of Choice, but we slide over that.)
Deﬁnition If R satisﬁes these properties, R is said to be Noetherian, or it is said
to satisfy the ascending chain condition. This property is satisﬁed by many of the
classical rings in mathematics. Having three deﬁnitions makes this property useful
and easy to use. For example, see the next theorem.
Theorem A Noetherian domain is a FD. In particular, a PID is a FD.
Proof Suppose there is a non-zero non-unit element that does not factor as the
ﬁnite product of irreducibles. Consider all ideals dR where d does not factor. Since
R is Noetherian, ∃ a maximal one cR. The element c must be reducible, i.e., c = ab
where neither a nor b is a unit. Each of aR and bR properly contains cR, and so each
Chapter 6 Appendix 113
of a and b factors as a ﬁnite product of irreducibles. This gives a ﬁnite factorization
of c into irreducibles, which is a contradiction.
Corollary A PID is a UFD. So Z is a UFD and if F is a ﬁeld, F[x] is a UFD.
You see the basic structure of UFD
s
is quite easy. It takes more work to prove
the following theorems, which are stated here only for reference.
Theorem If R is a UFD then R[x
1
, ..., x
n
] is a UFD. Thus if F is a ﬁeld,
F[x
1
, ..., x
n
] is a UFD. (This theorem goes all the way back to Gauss.)
If R is a PID, then the formal power series R[[x
1
, ..., x
n
]] is a UFD. Thus if F
is a ﬁeld, F[[x
1
, ..., x
n
]] is a UFD. (There is a UFD R where R[[x]] is not a UFD.
See page 566 of Commutative Algebra by N. Bourbaki.)
Theorem Germs of analytic functions on C
n
form a UFD.
Proof See Theorem 6.6.2 of An Introduction to Complex Analysis in Several Vari-
ables by L. H¨ ormander.
Theorem Suppose R is a commutative ring. Then R is Noetherian ⇒R[x
1
, ..., x
n
]
and R[[x
1
, ..., x
n
]] are Noetherian. (This is the famous Hilbert Basis Theorem.)
Theorem If R is Noetherian and I ⊂ R is a proper ideal, then R/I is Noetherian.
(This follows immediately from the deﬁnition. This and the previous theorem show
that Noetherian is a ubiquitous property in ring theory.)
Domains With Non-unique Factorizations Next are presented two of the
standard examples of Noetherian domains that are not unique factorization domains.
Exercise Let R = Z(

5) = ¦n +m

5 : n, m ∈ Z¦. Show that R is a subring of
R which is not a UFD. In particular 2 2 = (1 −

5) (−1 −

5) are two distinct
irreducible factorizations of 4. Show R is isomorphic to Z[x]/(x
2
−5), where (x
2
−5)
represents the ideal (x
2
− 5)Z[x], and R/(2) is isomorphic to Z
2
[x]/(x
2
− [5]) =
Z
2
[x]/(x
2
+ [1]), which is not a domain.
114 Appendix Chapter 6
Exercise Let R = R[x, y, z]/(x
2
− yz). Show x
2
− yz is irreducible and thus
prime in R[x, y, z]. If u ∈ R[x, y, z], let ¯ u ∈ R be the coset containing u. Show R
is not a UFD. In particular ¯ x ¯ x = ¯ y ¯ z are two distinct irreducible factorizations
of ¯ x
2
. Show R/(¯ x) is isomorphic to R[y, z]/(yz), which is not a domain. An easier
approach is to let f : R[x, y, z] → R[x, y] be the ring homomorphism deﬁned by
f(x) = xy, f(y) = x
2
, and f(z) = y
2
. Then S = R[xy, x
2
, y
2
] is the image of
f and S is isomorphic to R. Note that xy, x
2
, and y
2
are irreducible in S and
(xy)(xy) = (x
2
)(y
2
) are two distinct irreducible factorizations of (xy)
2
in S.
Exercise In Group Theory If G is an additive abelian group, a subgroup H
of G is said to be maximal if H = G and there are no subgroups properly between
H and G. Show that H is maximal iﬀ G/H ≈ Z
p
for some prime p. For simplicity,
consider the case G = Q. Which one of the following is true?
1) If a ∈ Q, then there is a maximal subgroup H of Q which contains a.
2) Q contains no maximal subgroups.
Splitting Short Exact Sequences
Suppose B is an R-module and K is a submodule of B. As deﬁned in the chapter
on linear algebra, K is a summand of B provided ∃ a submodule L of B with
K+L = B and K∩L = 0
¯
. In this case we write K⊕L = B. When is K a summand
of B? It turns out that K is a summand of B iﬀ there is a splitting map from
B/K to B. In particular, if B/K is free, K must be a summand of B. This is used
below to show that if R is a PID, then every submodule of R
n
is free.
Theorem 1 Suppose R is a ring, B and C are R-modules, and g : B → C is a
surjective homomorphism with kernel K. Then the following are equivalent.
1) K is a summand of B.
2) g has a right inverse, i.e., ∃ a homomorphism h : C →B with g ◦h = I : C →C.
(h is called a splitting map.)
Proof Suppose 1) is true, i.e., suppose ∃ a submodule L of B with K ⊕ L = B.
Then (g[L) : L → C is an isomorphism. If i : L → B is inclusion, then h deﬁned
by h = i ◦ (g[L)
−1
is a right inverse of g. Now suppose 2) is true and h : C → B
is a right inverse of g. Then h is injective, K + h(C) = B and K ∩ h(C) = 0
¯
.
Thus K ⊕h(C) = B.
Chapter 6 Appendix 115
Deﬁnition Suppose f : A → B and g : B → C are R-module homomorphisms.
The statement that 0 →A
f
→B
g
→C →0 is a short exact sequence (s.e.s) means
f is injective, g is surjective and f(A) = ker(g). The canonical split s.e.s. is A →
A ⊕ C → C where f = i
1
and g = π
2
. A short exact sequence is said to split if ∃
an isomorphism B

→A⊕C such that the following diagram commutes.
0 → A B C →0
A⊕C

f
g
i
1
π
2

`
`
`
`
`
`

·
We now restate the previous theorem in this terminology.
Theorem 1.1 A short exact sequence 0 → A → B → C → 0 splits iﬀ f(A) is
a summand of B, iﬀ B → C has a splitting map. If C is a free R-module, there is
a splitting map and thus the sequence splits.
Proof We know from the previous theorem f(A) is a summand of B iﬀ B → C
has a splitting map. Showing these properties are equivalent to the splitting of the
sequence is a good exercise in the art of diagram chasing. Now suppose C has a free
basis T ⊂ C, and g : B → C is surjective. There exists a function h : T → B such
that g ◦ h(c) = c for each c ∈ T. The function h extends to a homomorphism from
C to B which is a right inverse of g.
Theorem 2 If R is a domain, then the following are equivalent.
1) R is a PID.
2) Every submodule of R
R
is a free R-module of dimension ≤ 1.
This theorem restates the ring property of PID as a module property. Although
this theorem is transparent, 1)⇒2) is a precursor to the following classical result.
Theorem 3 If R is a PID and A ⊂ R
n
is a submodule, then A is a free R-module
of dimension ≤ n. Thus subgroups of Z
n
are free Z-modules of dimension ≤ n.
Proof From the previous theorem we know this is true for n = 1. Suppose n > 1
and the theorem is true for submodules of R
n−1
. Suppose A ⊂ R
n
is a submodule.
116 Appendix Chapter 6
Consider the following short exact sequences, where f : R
n−1
→R
n−1
⊕R is inclusion
and g = π : R
n−1
⊕R →R is the projection.
0 −→R
n−1
f
−→R
n−1
⊕R
π
−→R −→0
0 −→A∩ R
n−1
−→A −→π(A) −→0
By induction, A∩ R
n−1
is free of dimension ≤ n −1. If π(A) = 0
¯
, then A ⊂ R
n−1
.
If π(A) = 0
¯
, it is free of dimension 1 and thus the sequence splits by Theorem 1.1.
In either case, A is a free submodule of dimension ≤ n.
Exercise Let A ⊂ Z
2
be the subgroup generated by ¦(6, 24), (16, 64)¦. Show A
is a free Z-module of dimension 1. Also show the s.e.s. Z
4
×3
−→ Z
12
−→ Z
3
splits
but Z
×2
−→Z −→Z
2
and Z
2
×2
−→Z
4
−→Z
2
do not (see top of page 78).
Euclidean Domains
The ring Z possesses the Euclidean algorithm and the polynomial ring F[x] has
the division algorithm (pages 14 and 45). The concept of Euclidean domain is an
abstraction of these properties, and the eﬃciency of this abstraction is displayed in
this section. Furthermore the ﬁrst axiom, φ(a) ≤ φ(ab), is used only in Theorem
2, and is sometimes omitted from the deﬁnition. Anyway it is possible to just play
around with matrices and get some deep results. If R is a Euclidean domain and M
is a ﬁnitely generated R-module, then M is the sum of cyclic modules. This is one of
the great classical theorems of abstract algebra, and you don’t have to worry about
it becoming obsolete. Here N will denote the set of all non-negative integers, not
just the set of positive integers.
Deﬁnition A domain R is a Euclidean domain provided ∃ φ : (R−0
¯
) −→N such
that if a, b ∈ (R −0
¯
), then
1) φ(a) ≤ φ(ab).
2) ∃ q, r ∈ R such that a = bq +r with r = 0
¯
or φ(r) < φ(b).
Examples of Euclidean Domains
Z with φ(n) = [n[.
A ﬁeld F with φ(a) = 1 ∀ a = 0
¯
or with φ(a) = 0 ∀ a = 0
¯
.
F[x] where F is a ﬁeld with φ(f = a
0
+a
1
x + +a
n
x
n
) = deg(f).
Z[i] = ¦a +bi : a, b ∈ Z¦ = Gaussian integers with φ(a +bi) = a
2
+b
2
.
Chapter 6 Appendix 117
Theorem 1 If R is a Euclidean domain, then R is a PID and thus a UFD.
Proof If I is a non-zero ideal, then ∃ b ∈ I −0
¯
satisfying φ(b) ≤ φ(a) ∀ a ∈ I −0
¯
.
Then b generates I because if a ∈ I − 0
¯
, ∃ q, r with a = bq + r. Now r ∈ I and
r = 0
¯
⇒φ(r) < φ(b) which is impossible. Thus r = 0
¯
and a ∈ bR so I = bR.
Theorem 2 If R is a Euclidean domain and a, b ∈ R −0
¯
, then
φ(1
¯
) is the smallest integer in the image of φ.
a is a unit in R iﬀ φ(a) = φ(1
¯
).
a and b are associates ⇒ φ(a) = φ(b).
Proof This is a good exercise. However it is unnecessary for Theorem 3 below.
The following remarkable theorem is the foundation for the results of this section.
Theorem 3 If R is a Euclidean domain and (a
i,j
) ∈ R
n,t
is a non-zero matrix,
then by elementary row and column operations (a
i,j
) can be transformed to

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
d
1
0 0
0 d
2
.
.
.
.
.
.
d
m
0
0 0
¸

where each d
i
= 0
¯
, and d
i
[d
i+1
for 1 ≤ i < m. Also d
1
generates the ideal of R
generated by the entries of (a
i,j
).
Proof Let I ⊂ R be the ideal generated by the elements of the matrix A = (a
i,j
).
If E ∈ R
n
, then the ideal J generated by the elements of EA has J ⊂ I. If E is
invertible, then J = I. In the same manner, if E ∈ R
t
is invertible and J is the ideal
generated by the elements of AE, then J = I. This means that row and column
operations on A do not change the ideal I. Since R is a PID, there is an element
d
1
with I = d
1
R, and this will turn out to be the d
1
displayed in the theorem.
The matrix (a
i,j
) has at least one non-zero element d with φ(d) a miminum.
However, row and column operations on (a
i,j
) may produce elements with smaller
118 Appendix Chapter 6
φ values. To consolidate this approach, consider matrices obtained from (a
i,j
) by a
ﬁnite number of row and column operations. Among these, let (b
i,j
) be one which
has an entry d
1
= 0 with φ(d
1
) a minimum. By elementary operations of type 2, the
entry d
1
may be moved to the (1, 1) place in the matrix. Then d
1
will divide the other
entries in the ﬁrst row, else we could obtain an entry with a smaller φ value. Thus
by column operations of type 3, the other entries of the ﬁrst row may be made zero.
In a similar manner, by row operations of type 3, the matrix may be changed to the
following form.

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
d
1
0 0
0
.
.
. c
ij
0
¸

Note that d
1
divides each c
i,j
, and thus I = d
1
R. The proof now follows by induction
on the size of the matrix.
This is an example of a theorem that is easy to prove playing around at the
blackboard. Yet it must be a deep theorem because the next two theorems are easy
consequences.
Theorem 4 Suppose R is a Euclidean domain, B is a ﬁnitely generated free R-
module and A ⊂ B is a non-zero submodule. Then ∃ free bases ¦a
1
, a
2
, ..., a
t
¦ for A
and ¦b
1
, b
2
, ..., b
n
¦ for B, with t ≤ n, and such that each a
i
= d
i
b
i
, where each d
i
= 0
¯
,
and d
i
[d
i+1
for 1 ≤ i < t. Thus B/A ≈ R/d
1
⊕R/d
2
⊕ ⊕R/d
t
⊕R
n−t
.
Proof By Theorem 3 in the section Splitting Short Exact Sequences, A has a
free basis ¦v
1
, v
2
, ..., v
t
¦. Let ¦w
1
, w
2
, ..., w
n
¦ be a free basis for B, where n ≥ t. The
composition
R
t

−→A

−→B

−→R
n
e
i
−→v
i
w
i
−→e
i
is represented by a matrix (a
i,j
) ∈ R
n,t
where v
i
= a
1,i
w
1
+a
2,i
w
2
+ +a
n,i
w
n
. By
the previous theorem, ∃ invertible matrixes U ∈ R
n
and V ∈ R
t
such that
Chapter 6 Appendix 119
U(a
i,j
)V =

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
d
1
0 0
0 d
2
0
.
.
. 0
.
.
.
d
t
0 0
¸

with d
i
[d
i+1
. Since changing the isomorphisms R
t

−→A and B

−→R
n
corresponds
to changing the bases ¦v
1
, v
2
, ..., v
t
¦ and ¦w
1
, w
2
, ..., w
n
¦, the theorem follows.
Theorem 5 If R is a Euclidean domain and M is a ﬁnitely generated R-module,
then M ≈ R/d
1
⊕R/d
2
⊕ ⊕R/d
t
⊕R
m
where each d
i
= 0
¯
, and d
i
[d
i+1
for 1 ≤ i < t.
Proof By hypothesis ∃ a ﬁnitely generated free module B and a surjective homo-
morphism B −→M −→0. Let A be the kernel, so 0 −→A

−→B −→M −→0 is
a s.e.s. and B/A ≈ M. The result now follows from the previous theorem.
The way Theorem 5 is stated, some or all of the elements d
i
may be units, and for
such d
i
, R/d
i
= 0
¯
. If we assume that no d
i
is a unit, then the elements d
1
, d
2
, ..., d
t
are
called invariant factors. They are unique up to associates, but we do not bother with
that here. If R = Z and we select the d
i
to be positive, they are unique. If R = F[x]
and we select the d
i
to be monic, then they are unique. The splitting in Theorem 5
is not the ultimate because the modules R/d
i
may split into the sum of other cyclic
modules. To prove this we need the following Lemma.
Lemma Suppose R is a PID and b and c are non-zero non-unit elements of R.
Suppose b and c are relatively prime, i.e., there is no prime common to their prime
factorizations. Then bR and cR are comaximal ideals. (See p 108 for comaximal.)
Proof There exists an a ∈ R with aR = bR + cR. Since a[b and a[c, a is a
unit, so R = bR +cR.
Theorem 6 Suppose R is a PID and d is a non-zero non-unit element of R.
Assume d = p
s
1
1
p
s
2
2
p
st
t
is the prime factorization of d (see bottom of p 110). Then
the natural map R/d

−→R/p
s
1
1
⊕ ⊕ R/p
st
t
is an isomorphism of R-modules.
(The elements p
s
i
i
are called elementary divisors of R/d.)
Proof If i = j, p
s
i
i
and p
s
j
j
are relatively prime. By the Lemma above, they are
120 Appendix Chapter 6
comaximal and thus by the Chinese Remainder Theorem, the natural map is a ring
isomorphism (page 108). Since the natural map is also an R-module homomorphism,
it is an R-module isomorphism.
This theorem carries the splitting as far as it can go, as seen by the next exercise.
Exercise Suppose R is a PID, p ∈ R is a prime element, and s ≥ 1. Then the
R-module R/p
s
has no proper submodule which is a summand.
Torsion Submodules This will give a little more perspective to this section.
Deﬁnition Suppose M is a module over a domain R. An element m ∈ M is said
to be a torsion element if ∃ r ∈ R with r = 0
¯
and mr = 0
¯
. This is the same as
saying m is dependent. If R = Z, it is the same as saying m has ﬁnite order. Denote
by T(M) the set of all torsion elements of M. If T(M) = 0
¯
, we say that M is torsion
free.
Theorem 7 Suppose M is a module over a domain R. Then T(M) is a submodule
of M and M/T(M) is torsion free.
Proof This is a simple exercise.
Theorem 8 Suppose R is a Euclidean domain and M is a ﬁnitely generated
R-module which is torsion free. Then M is a free R-module, i.e., M ≈ R
m
.
Proof This follows immediately from Theorem 5.
Theorem 9 Suppose R is a Euclidean domain and M is a ﬁnitely generated
R-module. Then the following s.e.s. splits.
0 −→T(M) −→M −→M/T(M) −→0
Proof By Theorem 7, M/T(M) is torsion free. By Theorem 8, M/T(M) is a free
R-module, and thus there is a splitting map. Of course this theorem is transparent
anyway, because Theorem 5 gives a splitting of M into a torsion part and a free part.
Chapter 6 Appendix 121
Note It follows from Theorem 9 that ∃ a free submodule V of M such that T(M)⊕
V = M. The ﬁrst summand T(M) is unique, but the complementary summand V is
not unique. V depends upon the splitting map and is unique only up to isomorphism.
To complete this section, here are two more theorems that follow from the work
we have done.
Theorem 10 Suppose T is a domain and T

is the multiplicative group of units
of T. If G is a ﬁnite subgroup of T

, then G is a cyclic group. Thus if F is a ﬁnite
ﬁeld, the multiplicative group F

is cyclic. Thus if p is a prime, (Z
p
)

is cyclic.
Proof This is a corollary to Theorem 5 with R = Z. The multiplicative group G
is isomorphic to an additive group Z/d
1
⊕Z/d
2
⊕ ⊕ Z/d
t
where each d
i
> 1 and
d
i
[d
i+1
for 1 ≤ i < t. Every u in the additive group has the property that ud
t
= 0
¯
.
So every g ∈ G is a solution to x
dt
− 1
¯
= 0
¯
. If t > 1, the equation will have degree
less than the number of roots, which is impossible. Thus t = 1 and so G is cyclic.
Exercise For which primes p and q is the group of units (Z
p
Z
q
)

a cyclic group?
We know from Exercise 2) on page 59 that an invertible matrix over a ﬁeld is the
product of elementary matrices. This result also holds for any invertible matrix over
a Euclidean domain.
Theorem 11 Suppose R is a Euclidean domain and A ∈ R
n
is a matrix with
non-zero determinant. Then by elementary row and column operations, A may be
transformed to a diagonal matrix

¸
¸
¸
¸
¸
d
1
0
d
2
.
.
.
0 d
n
¸

where each d
i
= 0
¯
and d
i
[d
i+1
for 1 ≤ i < n. Also d
1
generates the ideal generated
by the entries of A. Furthermore A is invertible iﬀ each d
i
is a unit. Thus if A is
invertible, A is the product of elementary matrices.
122 Appendix Chapter 6
Proof It follows from Theorem 3 that A may be transformed to a diagonal matrix
with d
i
[d
i+1
. Since the determinant of A is not zero, it follows that each d
i
= 0
¯
.
Furthermore, the matrix A is invertible iﬀ the diagonal matrix is invertible, which is
true iﬀ each d
i
is a unit. If each d
i
is a unit, then the diagonal matrix is the product
of elementary matrices of type 1. Therefore if A is invertible, it is the product of
elementary matrices.
Exercise Let R = Z, A =

3 11
0 4

and D =

3 11
1 4

. Perform elementary
operations on A and D to obtain diagonal matrices where the ﬁrst diagonal element
divides the second diagonal element. Write D as the product of elementary matri-
ces. Find the characteristic polynomials of A and D. Find an elementary matrix B
over Z such that B
−1
AB is diagonal. Find an invertible matrix C in R
2
such that
C
−1
DC is diagonal. Show C cannot be selected in Q
2
.
Jordan Blocks
In this section, we deﬁne the two special types of square matrices used in the
Rational and Jordan canonical forms. Note that the Jordan block B(q) is the sum
of a scalar matrix and a nilpotent matrix. A Jordan block displays its eigenvalue
on the diagonal, and is more interesting than the companion matrix C(q). But as
we shall see later, the Rational canonical form will always exist, while the Jordan
canonical form will exist iﬀ the characteristic polynomial factors as the product of
linear polynomials.
Suppose R is a commutative ring, q = a
0
+ a
1
x + + a
n−1
x
n−1
+ x
n
∈ R[x]
is a monic polynomial of degree n ≥ 1, and V is the R[x]-module V = R[x]/q.
V is a torsion module over the ring R[x], but as an R-module, V has a free basis
¦1, x, x
2
, . . . , x
n−1
¦. (See the last part of the last theorem on page 46.) Multipli-
cation by x deﬁnes an R-module endomorphism on V , and C(q) will be the ma-
trix of this endomorphism with respect to this basis. Let T : V → V be deﬁned
by T(v) = vx. If h(x) ∈ R[x], h(T) is the R-module homomorphism given by
multiplication by h(x). The homomorphism from R[x]/q to R[x]/q given by
multiplication by h(x), is zero iﬀ h(x) ∈ qR[x]. That is to say q(T) = a
0
I + a
1
T+
+ T
n
is the zero homomorphism, and h(T) is the zero homomorphism iﬀ
h(x) ∈ qR[x]. All of this is supposed to make the next theorem transparent.
Theorem Let V have the free basis ¦1, x, x
2
, ..., x
n−1
¦. The companion matrix
Chapter 6 Appendix 123
representing T is
C(q) =

¸
¸
¸
¸
¸
¸
¸
¸
0 . . . . . . 0 −a
0
1 0 . . . 0 −a
1
0 1 0 −a
2
.
.
.
.
.
.
.
.
.
.
.
.
0 . . . . . . 1 −a
n−1
¸

The characteristic polynomial of C(q) is q, and [C(q)[ = (−1)
n
a
0
. Finally, if h(x) ∈
R[x], h(C(q)) is zero iﬀ h(x) ∈ qR[x].
Theorem Suppose λ ∈ R and q(x) = (x − λ)
n
. Let V have the free basis
¦1, (x −λ), (x −λ)
2
, . . . , (x −λ)
n−1
¦. Then the matrix representing T is
B(q) =

¸
¸
¸
¸
¸
¸
¸
¸
λ 0 . . . . . . 0
1 λ 0 . . . 0
0 1 λ
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 . . . . . . 1 λ
¸

The characteristic polynomial of B(q) is q, and [B(q)[ = λ
n
= (−1)
n
a
0
. Finally, if
h(x) ∈ R[x], h(B(q)) is zero iﬀ h(x) ∈ qR[x].
Note For n = 1, C(a
0
+ x) = B(a
0
+x) = (−a
0
). This is the only case where a
block matrix may be the zero matrix.
Note In B(q), if you wish to have the 1
s
above the diagonal, reverse the order of
the basis for V .
Jordan Canonical Form
We are ﬁnally ready to prove the Rational and Jordan forms. Using the previous
sections, all that’s left to do is to put the pieces together. (For an overview of Jordan
form, read ﬁrst the section in Chapter 5, page 96.)
124 Appendix Chapter 6
Suppose R is a commutative ring, V is an R-module, and T : V → V is an
R-module homomorphism. Deﬁne a scalar multiplication V R[x] → V by
v(a
0
+a
1
x + +a
r
x
r
) = va
0
+T(v)a
1
+ +T
r
(v)a
r
.
Theorem 1 Under this scalar multiplication, V is an R[x]-module.
This is just an observation, but it is one of the great tricks in mathematics.
Questions about the transformation T are transferred to questions about the module
V over the ring R[x]. And in the case R is a ﬁeld, R[x] is a Euclidean domain and so
we know almost everything about V as an R[x]-module.
Now in this section, we suppose R is a ﬁeld F, V is a ﬁnitely generated F-module,
T : V →V is a linear transformation and V is an F[x]-module with vx = T(v). Our
goal is to select a basis for V such that the matrix representing T is in some simple
form. A submodule of V
F[x]
is a submodule of V
F
which is invariant under T. We
know V
F[x]
is the sum of cyclic modules from Theorems 5 and 6 in the section on
Euclidean Domains. Since V is ﬁnitely generated as an F-module, the free part of
this decomposition will be zero. In the section on Jordan Blocks, a basis is selected
for these cyclic modules and the matrix representing T is described. This gives the
Rational Canonical Form and that is all there is to it. If all the eigenvalues for T are
in F, we pick another basis for each of the cyclic modules (see the second theorem in
the section on Jordan Blocks). Then the matrix representing T is called the Jordan
Canonical Form. Now we say all this again with a little more detail.
From Theorem 5 in the section on Euclidean Domains, it follows that
V
F[x]
≈ F[x]/d
1
⊕F[x]/d
2
⊕ ⊕F[x]/d
t
where each d
i
is a monic polynomial of degree ≥ 1, and d
i
[d
i+1
. Pick ¦1, x, x
2
, . . . , x
m−1
¦
as the F-basis for F[x]/d
i
where m is the degree of the polynomial d
i
.
Theorem 2 With respect to this basis, the matrix representing T is

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
C(d
1
)
C(d
2
)
.
.
.
C(d
t
)
¸

Chapter 6 Appendix 125
The characteristic polynomial of T is p = d
1
d
2
d
t
and p(T) = 0
¯
. This is a type
of canonical form but it does not seem to have a name.
Now we apply Theorem 6 to each F[x]/d
i
. This gives V
F[x]
≈ F[x]/p
s
1
1
⊕ ⊕
F[x]/p
sr
r
where the p
i
are irreducible monic polynomials of degree at least 1. The p
i
need not be distinct. Pick an F-basis for each F[x]/p
s
i
i
as before.
Theorem 3 With respect to this basis, the matrix representing T is

¸
¸
¸
¸
¸
¸
¸
¸
C(p
s
1
1
)
C(p
s
2
2
) 0
0
.
.
.
C(p
sr
r
)
¸

The characteristic polynomial of T is p = p
s
1
1
p
sr
r
and p(T) = 0
¯
. This is called
the Rational canonical form for T.
Now suppose the characteristic polynomial of T factors in F[x] as the product of
linear polynomials. Thus in the Theorem above, p
i
= x −λ
i
and
V
F[x]
≈ F[x]/(x −λ
1
)
s
1
⊕ ⊕F[x]/(x −λ
r
)
sr
is an isomorphism of F[x]-modules. Pick ¦1, (x − λ
i
), (x − λ
i
)
2
, . . . , (x − λ
i
)
m−1
¦ as
the F-basis for F[x]/(x −λ
i
)
s
i
where m is s
i
.
Theorem 4 With respect to this basis, the matrix representing T is

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
B((x −λ
1
)
s
1
)
0
B((x −λ
2
)
s
2
)
0
.
.
.
B((x −λ
r
)
sr
)
¸

126 Appendix Chapter 6
The characteristic polynomial of T is p = (x−λ
1
)
s
1
(x−λ
r
)
sr
and p(T) = 0
¯
. This
is called the Jordan canonical form for T. Note that the λ
i
need not be distinct.
Note A diagonal matrix is in Rational canonical form and in Jordan canonical
form. This is the case where each block is one by one. Of course a diagonal matrix
is about as canonical as you can get. Note also that if a matrix is in Jordan form,
its trace is the sum of the eigenvalues and its determinant is the product of the
eigenvalues. Finally, this section is loosely written, so it is important to use the
transpose principle to write three other versions of the last two theorems.
Exercise Suppose F is a ﬁeld of characteristic 0 and T ∈ F
n
has trace(T
i
) = 0
¯
for 0 < i ≤ n. Show T is nilpotent. Let p ∈ F[x] be the characteristic polynomial of
T. The polynomial p may not factor into linears in F[x], and thus T may have no
conjugate in F
n
which is in Jordan form. However this exercise can still be worked
using Jordan form. This is based on the fact that there exists a ﬁeld
¯
F containing F
as a subﬁeld, such that p factors into linears in
¯
F[x]. This fact is not proved in this
book, but it is assumed for this exercise. So ∃ an invertible matrix U ∈
¯
F
n
so that
U
−1
TU is in Jordan form, and of course, T is nilpotent iﬀ U
−1
TU is nilpotent. The
point is that it suﬃcies to consider the case where T is in Jordan form, and to show
the diagonal elements are all zero.
So suppose T is in Jordan form and trace (T
i
) = 0
¯
for 1 ≤ i ≤ n. Thus trace
(p(T)) = a
0
n where a
0
is the constant term of p(x). We know p(T) = 0
¯
and thus
trace (p(T)) = 0
¯
, and thus a
0
n = 0
¯
. Since the ﬁeld has characteristic 0, a
0
= 0
¯
and so 0
¯
is an eigenvalue of T. This means that one block of T is a strictly lower
triangular matrix. Removing this block leaves a smaller matrix which still satisﬁes
the hypothesis, and the result follows by induction on the size of T. This exercise
illustrates the power and facility of Jordan form. It also has a cute corollary.
Corollary Suppose F is a ﬁeld of characteristic 0, n ≥ 1, and (λ
1
, λ
2
, .., λ
n
) ∈ F
n
satisﬁes λ
i
1

i
2
+ +λ
i
n
= 0
¯
for each 1 ≤ i ≤ n. Then λ
i
= 0
¯
for 1 ≤ i ≤ n.
Minimal polynomials To conclude this section here are a few comments on the
minimal polynomial of a linear transformation. This part should be studied only if
you need it. Suppose V is an n-dimensional vector space over a ﬁeld F and T : V →V
is a linear transformation. As before we make V a module over F[x] with T(v) = vx.
Chapter 6 Appendix 127
Deﬁnition Ann(V
F[x]
) is the set of all h ∈ F[x] which annihilate V , i.e., which
satisfy V h = 0
¯
. This is a non-zero ideal of F[x] and is thus generated by a unique
monic polynomial u(x) ∈ F(x), Ann(V
F[x]
) = uF[x]. The polynomial u is called the
minimal polynomial of T. Note that u(T) = 0
¯
and if h(x) ∈ F[x], h(T) = 0
¯
iﬀ
h is a multiple of u in F[x]. If p(x) ∈ F[x] is the characteristic polynomial of T,
p(T) = 0
¯
and thus p is a multiple of u.
Now we state this again in terms of matrices. Suppose A ∈ F
n
is a matrix
representing T. Then u(A) = 0
¯
and if h(x) ∈ F[x], h(A) = 0
¯
iﬀ h is a multiple of
u in F[x]. If p(x) ∈ F[x] is the characteristic polynomial of A, then p(A) = 0
¯
and
thus p is a multiple of u. The polynomial u is also called the minimal polynomial of
A. Note that these properties hold for any matrix representing T, and thus similar
matrices have the same minimal polynomial. If A is given to start with, use the linear
transformation T : F
n
→F
n
determined by A to deﬁne the polynomial u.
Now suppose q ∈ F[x] is a monic polynomial and C(q) ∈ F
n
is the compan-
ion matrix deﬁned in the section Jordan Blocks. Whenever q(x) = (x − λ)
n
, let
B(q) ∈ F
n
be the Jordan block matrix also deﬁned in that section. Recall that q is
the characteristic polynomial and the minimal polynomial of each of these matrices.
This together with the rational form and the Jordan form will allow us to understand
the relation of the minimal polynomial to the characteristic polynomial.
Exercise Suppose A
i
∈ F
n
i
has q
i
as its characteristic polynomial and its minimal
polynomial, and A =

¸
¸
¸
¸
¸
A
1
0
A
2
.
.
.
0 A
r
¸

. Find the characteristic polynomial
and the minimal polynomial of A.
Exercise Suppose A ∈ F
n
.
1) Suppose A is the matrix displayed in Theorem 2 above. Find the characteristic
and minimal polynomials of A.
2) Suppose A is the matrix displayed in Theorem 3 above. Find the characteristic
and minimal polynomials of A.
3) Suppose A is the matrix displayed in Theorem 4 above. Find the characteristic
and minimal polynomials of A.
128 Appendix Chapter 6
4) Suppose λ ∈ F. Show λ is a root of the characteristic polynomial of A iﬀ λ
is a root of the minimal polynomial of A. Show that if λ is a root, its order
in the characteristic polynomial is at least as large as its order in the minimal
polynomial.
5) Suppose
¯
F is a ﬁeld containing F as a subﬁeld. Show that the minimal poly-
nomial of A ∈ F
n
is the same as the minimal polynomial of A considered as a
matrix in
¯
F
n
. (This funny looking exercise is a little delicate.)
6) Let F = R and A =

¸
¸
5 −1 3
0 2 0
−3 1 −1
¸

. Find the characteristic and minimal
polynomials of A.
Determinants
In the chapter on matrices, it is stated without proof that the determinant of the
product is the product of the determinants (see page 63). The purpose of this section
is to give a proof of this. We suppose R is a commutative ring, C is an R-module,
n ≥ 2, and B
1
, B
2
, . . . , B
n
is a sequence of R-modules.
Deﬁnition A map f : B
1
⊕ B
2
⊕ ⊕ B
n
→ C is R-multilinear means that if
1 ≤ i ≤ n, and b
j
∈ B
j
for j = i, then f[(b
1
, b
2
, . . . , B
i
, . . . , b
n
) deﬁnes an R-linear
map from B
i
to C.
Theorem The set of all R-multilinear maps is an R-module.
Proof From the ﬁrst exercise in Chapter 5, the set of all functions from B
1
⊕B
2

⊕ B
n
to C is an R-module (see page 69). It must be seen that the R-multilinear
maps form a submodule. It is easy to see that if f
1
and f
2
are R-multilinear, so is
f
1
+f
2
. Also if f is R-multilinear and r ∈ R, then (fr) is R-multilinear.
From here on, suppose B
1
= B
2
= = B
n
= B.
Deﬁnition
1) f is symmetric means f(b
1
, . . . , b
n
) = f(b
τ(1)
, . . . , b
τ(n)
) for all
permutations τ on ¦1, 2, . . . , n¦.
2) f is skew-symmetric if f(b
1
, . . . , b
n
) = sign(τ)f(b
τ(1)
, . . . , b
τ(n)
) for all τ.
Chapter 6 Appendix 129
3) f is alternating if f(b
1
, . . . , b
n
) = 0
¯
whenever some b
i
= b
j
for i = j.
Theorem
i) Each of these three types deﬁnes a submodule of the set of all
R-multilinear maps.
ii) Alternating ⇒ skew-symmetric.
iii) If no element of C has order 2, then alternating ⇐⇒ skew-symmetric.
Proof Part i) is immediate. To prove ii), assume f is alternating. It suﬃcies to
show that f(b
1
, ..., b
n
) = −f(b
τ(1)
, ..., b
τ(n)
) where τ is a transposition. For simplicity,
assume τ = (1, 2). Then 0
¯
= f(b
1
+ b
2
, b
1
+ b
2
, b
3
, ..., b
n
) = f(b
1
, b
2
, b
3
, ..., b
n
) +
f(b
2
, b
1
, b
3
, ..., b
n
) and the result follows. To prove iii), suppose f is skew symmetric
and no element of C has order 2, and show f is alternating. Suppose for convenience
that b
1
= b
2
and show f(b
1
, b
1
, b
3
, . . . , b
n
) = 0
¯
. If we let τ be the transposition (1, 2),
we get f(b
1
, b
1
, b
3
, . . . , b
n
) = −f(b
1
, b
1
, b
3
, . . . , b
n
), and so 2f(b
1
, b
1
, b
3
, . . . , b
n
) = 0
¯
,
and the result follows.
Now we are ready for determinant. Suppose C = R. In this case multilinear
maps are usually called multilinear forms. Suppose B is R
n
with the canonical basis
¦e
1
, e
2
, . . . , e
n
¦. (We think of a matrix A ∈ R
n
as n column vectors, i.e., as an element
of B ⊕B ⊕ ⊕B.) First we recall the deﬁnition of determinant.
Suppose A = (a
i,j
) ∈ R
n
. Deﬁne d : B⊕B⊕ ⊕B →R by d(a
1,1
e
1
+a
2,1
e
2
+ +
a
n,1
e
n
, ....., a
1,n
e
1
+a
2,n
e
2
+ +a
n,n
e
n
) =
¸
all τ
sign(τ)(a
τ(1),1
a
τ(2),2
a
τ(n),n
) = [A[.
The next theorem follows from the section on determinants on page 61.
Theorem d is an alternating multilinear form with d(e
1
, e
2
, . . . , e
n
) = 1
¯
.
If c ∈ R, dc is an alternating multilinear form, because the set of alternating forms
is an R-module. It turns out that this is all of them, as seen by the following theorem.
Theorem Suppose f : B ⊕ B ⊕. . . ⊕B → R is an alternating multilinear form.
Then f = df(e
1
, e
2
, . . . , e
n
). This means f is the multilinear form d times the scalar
f(e
1
, e
2
, ..., e
n
). In other words, if A = (a
i,j
) ∈ R
n
, then f(a
1,1
e
1
+ a
2,1
e
2
+ +
a
n,1
e
n
, ....., a
1,n
e
2
+ a
2,n
e
2
+ + a
n,n
e
n
) = [A[f(e
1
, e
2
, ..., e
n
). Thus the set of alter-
nating forms is a free R-module of dimension 1, and the determinant is a generator.
130 Appendix Chapter 6
Proof For n = 2, you can simply write it out. f(a
1,1
e
1
+a
2,1
e
2
, a
1,2
e
1
+a
2,2
e
2
) =
a
1,1
a
1,2
f(e
1
, e
1
) +a
1,1
a
2,2
f(e
1
, e
2
) +a
2,1
a
1,2
f(e
2
, e
1
) +a
2,1
a
2,2
f(e
2
, e
2
) = (a
1,1
a
2,2

a
1,2
a
2,1
)f(e
1
, e
2
) = [A[f(e
1
, e
2
). For the general case, f(a
1,1
e
1
+a
2,1
e
2
+ +
a
n,1
e
n
, ....., a
1,n
e
1
+a
2,n
e
2
+ +a
n,n
e
n
) =
¸
a
i
1
,1
a
i
2
,2
a
in,n
f(e
i
1
, e
i
2
, ..., e
in
) where
the sum is over all 1 ≤ i
1
≤ n, 1 ≤ i
2
≤ n, ..., 1 ≤ i
n
≤ n. However, if any i
s
= i
t
for s = t, that term is 0 because f is alternating. Therefore the sum is
just
¸
all τ
a
τ(1),1
a
τ(2),2
a
τ(n),n
f(e
τ(1)
, e
τ(2)
, . . . , e
τ(n)
) =
¸
all τ
sign(τ)a
τ(1),1
a
τ(2),2
a
τ(n),n
f(e
1
, e
2
, . . . , e
n
) = [A[f(e
1
, e
2
, ..., e
n
).
This incredible classiﬁcation of these alternating forms makes the proof of the
following theorem easy. (See the third theorem on page 63.)
Theorem If C, A ∈ R
n
, then [CA[ = [C[[A[.
Proof Suppose C ∈ R
n
. Deﬁne f : R
n
→R by f(A) = [CA[. In the notation of
the previous theorem, B = R
n
and R
n
= R
n
⊕ R
n
⊕ ⊕ R
n
. If A ∈ R
n
, A =
(A
1
, A
2
, ..., A
n
) where A
i
∈ R
n
is column i of A, and f : R
n
⊕ ⊕ R
n
→ R
has f(A
1
, A
2
, ..., A
n
) = [CA[. Use the fact that CA = (CA
1
, CA
2
, ..., CA
n
) to
show that f is an alternating multilinear form. By the previous theorem, f(A) =
[A[f(e
1
, e
2
, ..., e
n
). Since f(e
1
, e
2
, ..., e
n
) = [CI[ = [C[, it follows that [CA[ = f(A) =
[A[[C[.
Dual Spaces
The concept of dual module is basic, not only in algebra, but also in other areas
such as diﬀerential geometry and topology. If V is a ﬁnitely generated vector space
over a ﬁeld F, its dual V

is deﬁned as V

= Hom
F
(V, F). V

is isomorphic to V , but
in general there is no natural isomorphism from V to V

. However there is a natural
isomorphism from V to V
∗∗
, and so V

is the dual of V and V may be considered
to be the dual of V

. This remarkable fact has many expressions in mathematics.
For example, a tangent plane to a diﬀerentiable manifold is a real vector space. The
union of these spaces is the tangent bundle, while the union of the dual spaces is the
cotangent bundle. Thus the tangent (cotangent) bundle may be considered to be the
dual of the cotangent (tangent) bundle. The sections of the tangent bundle are called
vector ﬁelds while the sections of the cotangent bundle are called 1-forms.
In algebraic topology, homology groups are derived from chain complexes, while
cohomology groups are derived from the dual chain complexes. The sum of the
cohomology groups forms a ring, while the sum of the homology groups does not.
Chapter 6 Appendix 131
Thus the concept of dual module has considerable power. We develop here the basic
theory of dual modules.
Suppose R is a commutative ring and W is an R-module.
Deﬁnition If M is an R-module, let H(M) be the R-module H(M)=Hom
R
(M, W).
If M and N are R-modules and g : M → N is an R-module homomorphism, let
H(g) : H(N) → H(M) be deﬁned by H(g)(f) = f ◦ g. Note that H(g) is an
R-module homomorphism.
M N
W
f
g
H(g)(f) = f ◦ g

·
`
`
`
`
`
`
``
Theorem
i) If M
1
and M
2
are R-modules, H(M
1
⊕M
2
) ≈ H(M
1
) ⊕H(M
2
).
ii) If I : M →M is the identity, then H(I) : H(M) →H(M) is the
identity.
iii) If M
1
g
−→M
2
h
−→M
3
are R-module homomorphisms, then H(g)◦H(h) =
H(h ◦ g). If f : M
3
→W is a homomorphism, then
(H(g) ◦ H(h))(f) = H(h ◦ g)(f) = f ◦ h ◦ g.
M
1
M
2
M
3
g
h
f
f ◦ h
W
f ◦ h ◦ g

·

`
`
`
`
`
`
Note In the language of the category theory, H is a contravariant functor from
the category of R-modules to itself.
132 Appendix Chapter 6
Theorem If M and N are R-modules and g : M → N is an isomorphism, then
H(g) : H(N) →H(M) is an isomorphism with H(g
−1
) = H(g)
−1
.
Proof
I
H(N)
= H(I
N
) = H(g ◦ g
−1
) = H(g
−1
) ◦ H(g)
I
H(M)
= H(I
M
) = H(g
−1
◦ g) = H(g) ◦ H(g
−1
)
Theorem
i) If g : M →N is a surjective homomorphism, then H(g) : H(N) →H(M)
is injective.
ii) If g : M →N is an injective homomorphism and g(M) is a summand
of N, then H(g) : H(N) →H(M) is surjective.
iii) If R is a ﬁeld and g : M →N is a homomorphism, then g is surjective
(injective) iﬀ H(g) is injective (surjective).
Proof This is a good exercise.
For the remainder of this section, suppose W = R
R
. In this case H(M) =
Hom
R
(M, R) is denoted by H(M) = M

and H(g) is denoted by H(g) = g

.
Theorem Suppose M has a ﬁnite free basis ¦v
1
, ..., v
n
¦. Deﬁne v

i
∈ M

by
v

i
(v
1
r
1
+ + v
n
r
n
) = r
i
. Thus v

i
(v
j
) = δ
i,j
. Then v

1
, . . . , v

n
is a free basis for
M

, called the dual basis. Therefore M

is free and is isomorphic to M.
Proof First consider the case of R
n
= R
n,1
, with basis ¦e
1
, . . . , e
n
¦ where e
i
=

¸
¸
¸
¸
¸
¸
¸
0

1
i

0
¸

.
We know (R
n
)

≈ R
1,n
, i.e., any homomorphism from R
n
to R is given by a 1 n
matrix. Now R
1,n
is free with dual basis ¦e

1
, . . . , e

n
¦ where e

i
= (0, . . . , 0, 1
i
, 0, . . . , 0).
For the general case, let g : R
n

→ M be given by g(e
i
) = v
i
. Then g

: M

→ (R
n
)

sends v

i
to e

i
. Since g

is an isomorphism, ¦v

1
, . . . , v

n
¦ is a basis for M

.
Theorem Suppose M is a free module with a basis ¦v
1
, . . . , v
m
¦ and N is a free
module with a basis ¦w
1
, . . . , w
n
¦ and g : M → N is the homomorphism given by
A = (a
i,j
) ∈ R
n,m
. This means g(v
j
) = a
1,j
w
1
+ + a
n,j
w
n
. Then the matrix of
g

: N

→M

with respect to the dual bases, is given by A
t
.
Chapter 6 Appendix 133
Proof Note that g

(w

i
) is a homomorphism from M to R. Evaluation on v
j
gives
g

(w

i
)(v
j
) = (w

i
◦ g)(v
j
) = w

i
(g(v
j
)) = w

i
(a
1,j
w
1
+ +a
n,j
w
n
) = a
i,j
. Thus g

(w

i
)
= a
i,1
v

1
+ +a
i,m
v

m
, and thus g

is represented by A
t
.
Exercise If U is an R-module, deﬁne φ
U
: U

⊕ U → R by φ
U
(f, u) = f(u).
Show that φ
U
is R-bilinear. Suppose g : M → N is an R-module homomorphism,
f ∈ N

and v ∈ M. Show that φ
N
(f, g(v)) = φ
M
(g

(f), v). Now suppose M =
N = R
n
and g : R
n
→ R
n
is represented by a matrix A ∈ R
n
. Suppose f ∈ (R
n
)

and v ∈ R
n
. Use the theorem above to show that φ : (R
n
)

⊕ R
n
→ R has the
property φ(f, Av) = φ(A
t
f, v). This is with the elements of R
n
and (R
n
)

written as
column vectors. If the elements of R
n
are written as column vectors and the elements
of (R
n
)

are written as row vectors, the formula is φ(f, Av) = φ(fA, v). Of course
this is just the matrix product fAv. Dual spaces are confusing, and this exercise
should be worked out completely.
Deﬁnition “Double dual” is a “covariant” functor, i.e., if g : M → N is
a homomorphism, then g
∗∗
: M
∗∗
→ N
∗∗
. For any module M, deﬁne α : M → M
∗∗
by α(m) : M

→ R is the homomorphism which sends f ∈ M

to f(m) ∈ R, i.e.,
α(m) is given by evaluation at m. Note that α is a homomorphism.
Theorem If M and N are R-modules and g : M →N is a homomorphism, then
the following diagram is commutative.
M M
∗∗
N N
∗∗
α
α
g
g
∗∗

· ·
Proof On M, α is given by α(v) = φ
M
(−, v). On N, α(u) = φ
N
(−, u).
The proof follows from the equation φ
N
(f, g(v)) = φ
M
(g

(f), v).
Theorem If M is a free R-module with a ﬁnite basis ¦v
1
, . . . , v
n
¦, then
α : M →M
∗∗
is an isomorphism.
Proof ¦α(v
1
), . . . , α(v
n
)¦ is the dual basis of ¦v

1
, . . . , v

n
¦, i.e., α(v
i
) = (v

i
)

.
134 Appendix Chapter 6
Note Suppose R is a ﬁeld and C is the category of ﬁnitely generated vector spaces
over R. In the language of category theory, α is a natural equivalence between the
identity functor and the double dual functor.
Note For ﬁnitely generated vector spaces, α is used to identify V and V
∗∗
. Under
this identiﬁcation V

is the dual of V and V is the dual of V

. Also, if ¦v
1
, . . . , v
n
¦
is a basis for V and ¦v

i
, . . . , v

n
¦ its dual basis, then ¦v
1
, . . . , v
n
¦ is the dual basis
for ¦v

1
, . . . , v

n
¦.
In general there is no natural way to identify V and V

. However for real inner
product spaces there is.
Theorem Let R = R and V be an n-dimensional real inner product space.
Then β : V →V

given by β(v) = (v, −) is an isomorphism.
Proof β is injective and V and V

have the same dimension.
Note If β is used to identify V with V

, then φ
V
: V

⊕ V → R is just the dot
product V ⊕V →R.
Note If ¦v
1
, . . . , v
n
¦ is any orthonormal basis for V, ¦β(v
1
), . . . , β(v
n
)¦ is the dual
basis of ¦v
1
, . . . , v
n
¦, that is β(v
i
) = v

i
. The isomorphism β : V → V

deﬁnes an
inner product on V

, and under this structure, β is an isometry. If ¦v
1
, . . . , v
n
¦ is
an orthonormal basis for V, ¦v

1
, . . . , v

n
¦ is an orthonormal basis for V

. Also, if U
is another n-dimensional IPS and f : V → U is an isometry, then f

: U

→ V

is an isometry and the following diagram commutes.
V V

U U

β
β
f f

·
`
Exercise Suppose R is a commutative ring, T is an inﬁnite index set, and
for each t ∈ T, R
t
= R. Show (

t∈T
R
t
)

is isomorphic to R
T
=
¸
t∈T
R
t
. Now let
T = Z
+
, R = R, and M =

t∈T
R
t
. Show M

is not isomorphic to M.
Index
Abelian group, 20, 71
Algebraically closed ﬁeld, 46, 97
Alternating group, 32
Ascending chain condition, 112
Associate elements in a domain, 47, 109
Automorphism
of groups, 29
of modules, 70
of rings, 43
Axiom of choice, 10
Basis or free basis
canonical or standard for R
n
, 72, 79
of a module, 78, 83
Bijective or one-to-one correspondence,7
Binary operation, 19
Boolean algebras, 52
Boolean rings, 51
Cancellation law
in a group, 20
in a ring, 39
Cartesian product, 2, 11
Cayley’s theorem, 31
Cayley-Hamilton theorem, 66, 98, 125
Center of group, 22
Change of basis, 83
Characteristic of a ring, 50
Characteristic polynomial
of a homomorphism, 85, 95
of a matrix, 66
Chinese remainder theorem, 50, 108
Classical adjoint of a matrix, 63
Cofactor of a matrix, 62
Comaximal ideals, 108, 120
Commutative ring, 37
Complex numbers, 1, 40, 46, 47, 97, 104
Conjugate, 64
Conjugation by a unit, 44
Contravariant functor, 131
Coproduct or sum of modules, 76
Coset, 24, 42, 74
Cycle, 32
Cyclic
group, 23
module, 107
Determinant
of a homomorphism, 85
of a matrix, 60, 128
Diagonal matrix, 56
Dimension of a free module, 83
Division algorithm, 45
Domain
euclidean, 116
integral domain, 39
of a function, 5
principal ideal, 46
unique factorization, 111
Dual basis, 132
Dual spaces, 130
Eigenvalues, 95
Eigenvectors, 95
Elementary divisors, 119, 120
Elementary matrices, 58
135
136 Index
Elementary operations, 57, 122
Endomorphism of a module, 70
Equivalence class, 4
Equivalence relation, 4
Euclidean algorithm, 14
Euclidean domain, 116
Evaluation map, 47, 49
Even permutation, 32
Exponential of a matrix, 106
Factorization domain (FD), 111
Fermat’s little theorem, 50
Field, 39
Formal power series, 113
Fourier series, 100
Free basis, 72, 78, 79, 83
Free R-module, 78
Function or map, 6
bijective, 7
injective, 7
surjective, 7
Function space Y
T
as a group, 22, 36
as a module, 69
as a ring, 44
as a set, 12
Fundamental theorem of algebra, 46
Gauss, 113
General linear group GL
n
(R), 55
Generating sequence in a module, 78
Generators of Z
n
, 40
Geometry of determinant, 90
Gram-Schmidt orthonormalization, 100
Graph of a function, 6
Greatest common divisor, 15
Group, 19
abelian, 20
cyclic, 23
multiplicative, 19
symmetric, 31
Hausdorﬀ maximality principle, 3, 87,
109
Hilbert, 113
Homogeneous equation, 60
Homormophism
of groups, 23
of rings, 42
of modules, 69
Homomorphism of quotient
group, 29
module, 74
ring, 44
Ideal
left, 41
maximal, 109
of a ring, 41
prime, 109
principal, 42, 46
right, 41
Idempotent element in a ring, 49, 51
Image of a function, 7
Independent sequence in a module, 78
Index of a subgroup, 25
Index set, 2
Induction, 13
Injective or one-to-one, 7, 79
Inner product spaces, 98
Integers mod n, 27, 40
Integers, 1, 14
Invariant factors, 119
Inverse image, 7
Invertible or non-singular matrix, 55
Irreducible element, 47, 110
Isometries of a square, 26, 34
Isometry, 101
Isomorphism
Index 137
of groups, 29
of modules, 70
of rings, 43
Jacobian matrix, 91
Jordan block, 96, 123
Jordan canonical form, 96, 123, 125
Kernel, 28, 43, 70
Least common multiple, 17, 18
Linear combination, 78
Linear ordering, 3
Linear transformation, 85
Matrix
elementary, 58
invertible, 55
representing a linear transformation,
84
triangular, 56
Maximal
ideal, 109
independent sequence, 86, 87
monotonic subcollection, 4
subgroup, 114
Minimal polynomial, 127
Minor of a matrix, 62
Module over a ring, 68
Monomial, 48
Monotonic collection of sets, 4
Multilinear forms, 129
Multiplicative group of a ﬁnite ﬁeld, 121
Nilpotent
element, 56
homomorphism, 93
Noetherian ring, 112
Normal subgroup, 26
Odd permutation, 32
Onto or surjective, 7, 79
Order of an element or group, 23
Orthogonal group O(n), 102
Orthogonal vectors, 99
Orthonormal sequence, 99
Partial ordering, 3
Partition of a set, 5
Permutation, 31
Pigeonhole principle, 8, 39
Polynomial ring, 45
Power set, 12
Prime
element, 110
ideal, 109
integer, 16
Principal ideal domain (PID), 46
Principal ideal, 42
Product
of groups, 34, 35
of modules, 75
of rings, 49
of sets, 2, 11
Projection maps, 11
Quotient group, 27
Quotient module, 74
Quotient ring, 42
Range of a function, 6
Rank of a matrix, 59, 89
Rational canonical form, 107, 125
Relation, 3
Relatively prime
integers, 16
elements in a PID, 119
Right and left inverses of functions, 10
Ring, 38
Root of a polynomial, 46
Row echelon form, 59
Scalar matrix, 57
138 Index
Scalar multiplication, 21, 38, 54, 71
Short exact sequence, 115
Sign of a permutation, 60
Similar matrices, 64
Solutions of equations, 9, 59, 81
Splitting map, 114
Standard basis for R
n
, 72, 79
Strips (horizontal and vertical), 8
Subgroup, 14, 21
Submodule, 69
Subring, 41
Summand of a module, 77, 115
Surjective or onto, 7, 79
Symmetric groups, 31
Symmetric matrix, 103
Torsion element of a module, 121
Trace
of a homormophism, 85
of a matrix, 65
Transpose of a matrix, 56, 103, 132
Transposition, 32
Unique factorization,
in principal ideal domains, 113
of integers, 16
Unique factorization domain (UFD), 111
Unit in a ring, 38
Vector space, 67, 85
Volume preserving homomorphism, 90
Zero divisor in a ring, 39

ii

E.H. Connell Department of Mathematics University of Miami P.O. Box 249085 Coral Gables, Florida 33124 USA ec@math.miami.edu

Mathematical Subject Classiﬁcations (1991): 12-01, 13-01, 15-01, 16-01, 20-01

c 1999

E.H. Connell

March 20, 2004

iii

Introduction
In 1965 I ﬁrst taught an undergraduate course in abstract algebra. It was fun to teach because the material was interesting and the class was outstanding. Five of those students later earned a Ph.D. in mathematics. Since then I have taught the course about a dozen times from various texts. Over the years I developed a set of lecture notes and in 1985 I had them typed so they could be used as a text. They now appear (in modiﬁed form) as the ﬁrst ﬁve chapters of this book. Here were some of my motives at the time. 1) To have something as short and inexpensive as possible. In my experience, students like short books. 2) To avoid all innovation. To organize the material in the most simple-minded straightforward manner. 3) To order the material linearly. To the extent possible, each section should use the previous sections and be used in the following sections. 4) To omit as many topics as possible. This is a foundational course, not a topics course. If a topic is not used later, it should not be included. There are three good reasons for this. First, linear algebra has top priority. It is better to go forward and do more linear algebra than to stop and do more group and ring theory. Second, it is more important that students learn to organize and write proofs themselves than to cover more subject matter. Algebra is a perfect place to get started because there are many “easy” theorems to prove. There are many routine theorems stated here without proofs, and they may be considered as exercises for the students. Third, the material should be so fundamental that it be appropriate for students in the physical sciences and in computer science. Zillions of students take calculus and cookbook linear algebra, but few take abstract algebra courses. Something is wrong here, and one thing wrong is that the courses try to do too much group and ring theory and not enough matrix theory and linear algebra. 5) To oﬀer an alternative for computer science majors to the standard discrete mathematics courses. Most of the material in the ﬁrst four chapters of this text is covered in various discrete mathematics courses. Computer science majors might beneﬁt by seeing this material organized from a purely mathematical viewpoint.

it is learned. but it requires a lot of organization. you still have to build it.edu .” Mathematics is not taught. Basic algebra is a subject of incredible elegance and utility. Teaching abstract algebra and linear algebra as distinct courses results in a loss of synergy and a loss of momentum. I would have made the book shorter. Because after you extract it. E. The purpose of class is to learn. not just back. FL 33124 ec@math. I am convinced it is easier to build a course from a base than to extract it from a big book. The bare bones nature of this book adds to its ﬂexibility. they participate little and learn little. Also with this text the professor does not extract the course from the text. This book is my attempt at that organization. enjoy it. the student already has the outline of the next lecture. Professors should give more direction in that regard. Every eﬀort has been extended to make the subject move rapidly and to make the ﬂow from one topic to the next as seamless as possible. The professor picks which topics to assign for serious study and which ones to “wave arms at”. This book works best when viewed lightly and read as a story. The style and approach of this book is to make it a little lighter. Unfortunately mathematics is a diﬃcult and heavy subject. I am here. The student has limited time during the semester for serious study. The goal is to stay focused and go forward. A few minutes of preparation does wonders to leverage classroom learning. Connell Department of Mathematics University of Miami Coral Gables. and this book is intended to be used in that manner. and this time should be allocated with care. but I did not have any more time. I hope the students and professors who try it. but rather builds the course upon it. because you can build whatever course you want around it. H. Study forward. When students come to class cold and spend the period taking notes. When using this text.v This text is written with the conviction that it is more eﬀective to teach abstract and linear algebra as one coherent discipline rather than as two separate ones. This leads to a dead class and also to the bad psychology of “O K. and each assignment should include the study of the next few pages.miami. so teach me the subject. because mathematics is learned in hindsight. and many students never learn how to learn. not to do transcription work.

and scalar matrices Elementary operations and elementary matrices Systems of equations 53 56 56 57 59 . strips. quotient groups. invertible matrices Transpose Triangular. ﬁelds The integers mod n Ideals and quotient rings Homomorphisms Polynomial rings Product of rings The Chinese remainder theorem Characteristic Boolean rings Chapter 4 Matrices and Matrix Rings 37 38 40 41 42 45 49 50 50 51 Addition and multiplication of matrices. equivalence relations Functions.vi Outline Chapter 1 Background and Fundamentals of Mathematics 1 3 5 13 14 Sets. cosets Normal subgroups. projections Notation for the logic of mathematics Integers. bijections. right and left inverses. subgroups. Cartesian products Relations. the integers mod n Homomorphisms Permutations. solutions of equations. partial orderings. Hausdorﬀ maximality principle. scalar multiplication for additive groups Subgroups. order. the symmetric groups Product of groups Chapter 3 Rings Rings Units. diagonal. unique factorization Chapter 2 Groups 19 21 25 27 31 34 Groups. domains.

trace. and characteristic polynomial Chapter 5 Linear Algebra 68 69 71 74 75 77 78 79 82 83 85 90 91 92 93 95 96 98 102 103 60 64 Modules. Gram-Schmidt orthonormalization Orthogonal matrices.vii Determinants. submodules Homomorphisms Homomorphisms on Rn Cosets and quotient modules Products and coproducts Summands Independence. rank of a matrix Geometric interpretation of determinant Linear functions approximate diﬀerentiable functions locally The transpose principle Nilpotent homomorphisms Eigenvalues. characteristic roots Jordan canonical form Inner product spaces. and free basis Characterization of free modules Uniqueness of dimension Change of basis Vector spaces. the classical adjoint Similarity. square matrices over ﬁelds. generating sets. the orthogonal group Diagonalization of symmetric matrices Chapter 6 Appendix The Chinese remainder theorem Prime and maximal ideals and UFDs Splitting short exact sequences Euclidean domains Jordan blocks Jordan canonical form Determinants Dual spaces 108 109 114 116 122 123 128 130 .

And before beginning the course. not on sleight of hand. This trick is based. In this puzzle. Is it possible to slide the tiles around to get them all in order. a frame has 12 spaces.viii 1 5 9 2 6 11 3 7 10 4 8 Abstract algebra is not only a major subject of science. The last two tiles are out of order. but it is also magic and fun. and end again with the last space vacant? After giving up on this. for example. but rather on a theorem in abstract algebra. the ﬁrst 11 with numbered tiles and the last vacant. and it is certainly not a dull boy. the neat card trick on page 18. but to understand it you need some group theory. Anyone can do it. you can study permutation groups and learn the answer! . Abstract algebra is not all work and no play. See. you might ﬁrst try your skills on the famous (some would say infamous) tile puzzle.

.} Q = the ﬁeld of rational numbers = {a/b : a. injective. b = 0} R = the ﬁeld of real numbers C = the ﬁeld of complex numbers = {a + bi : a. . and most properties of functions can be stated in terms of solutions of equations.. An equivalence relation on a set A is shown to be simply a partition of A into disjoint subsets. 2. and bijective.Chapter 1 Background and Fundamentals of Mathematics This chapter is fundamental. The symbol ∀ means “for each” and ⇒ means “implies”. and the properties of surjective. C. 3. but for all ﬁelds related to mathematics. The basic concepts are products of sets. We use the standard notation for intersection and union.. not just for algebra.. are sets. 1. and the integers. Notation Mathematics has its own universally accepted shorthand. −1... b ∈ R} (i2 = −1) Sets Suppose A.. 2. 0. −2. . The symbol ∃ means “there exists” and ∃! means “there exists a unique”. A ∩ B = {x : x ∈ A and x ∈ B} = the set of all x which are elements 1 . In elementary courses the section on the Hausdorﬀ Maximality Principle should be ignored.. N = Z+ = the set of positive integers = {1. equivalence relations. Five of these are listed below. B.. partial orderings.} Z = the ring of integers = {. The notion of a solution of an equation is central in mathematics.. Some sets (or collections) are so basic they have their own proprietary symbols. The ﬁnal section gives a proof of the unique factorization theorem for the integers. functions. There is an emphasis on the concept of function. b ∈ Z.

x ∈ At } t∈T Let ∅ be the null set. At is a set. x ∈ At } = {x : ∀t ∈ T. the complement of C in S. or B contains A. Deﬁnition Suppose each of A and B is a set.e. a ∈ A ⇒ a ∈ B. then a is an element of B. . be deﬁned by C = S − C = {x ∈ S : x ∈ C}. If A ⊂ B we may say A is contained in B. The statement that A is a subset of B (A ⊂ B) means that if a is an element of A. Suppose T is an index set and for each t ∈ T . let C . That is. Exercise Suppose each of A and B is a set.. y) : x ∈ X and y ∈ Y }. if C is a subset of S). Any set called an index set is assumed to be non-void.2 of A and B. Theorem (De Morgan’s laws) Suppose S is a set. t∈T At = {x : ∃ t ∈ T with x ∈ At } At = {x : if t ∈ T. The statement that A is not a subset of B means . B ⊂ S. If C ⊂ S (i. In other words. then A and B are said to be disjoint. X × Y = {(x. (A ∩ B) = A ∪ B (A ∪ B) = A ∩ B and Cartesian Products If X and Y are sets. Then for any A. Background Chapter 1 A ∪ B = {x : x ∈ A or x ∈ B} = the set of all x which are elements of A or B. the Cartesian product of X and Y is deﬁned to be the set of all ordered pairs whose ﬁrst term is in X and whose second term is in Y . Example R × R = R2 = the plane. If A ∩ B = ∅.

b ∈ A. .. and we write this fact by the expression a ∼ b..Chapter 1 Background 3 Deﬁnition If each of X1 . 2). If a ∼ b. then (reﬂexive) (symmetric) a = b. Example Example A = R with the ordinary ordering. X1 × · · · × Xn = {(x1 . (anti-symmetric) a ∼ c. and 3) is called a partial ordering. then a ≤ a. Example Question R × · · · × R = Rn = real n-space. is a partial ordering.. Here are several properties which a relation may possess. xn ) : xi ∈ Xi for 1 ≤ i ≤ n} = the set of all ordered n-tuples whose i-th term is in Xi . the relation also satisﬁes these properties when restricted to S. 2 ) If a ≤ b and b ≤ a. Then 1) If a ∈ A. Deﬁnition A linear ordering is a partial ordering with the additional property that. a partial ordering on A deﬁnes a partial ordering . then b ∼ a. or 3) on A. Is (R × R2 ) = (R2 × R) = R3 ? Relations If A is a non-void set. then a ≤ c. In particular. a non-void subset R ⊂ A × A is called a relation on A. 2 ). If a ∼ b and b ∼ a. then a ≤ b or b ≤ a. In this case we write a ∼ b as a ≤ b. then a = b. then a ∼ a. 1) 2) 2) 3) If a ∈ A. If (a.. If the relation satisﬁes any of the properties 1). Hausdorﬀ Maximality Principle (HMP) Suppose S is a non-void subset of A and ∼ is a relation on A. if a. . 2 ). This deﬁnes a relation on S.. (transitive) Deﬁnition A relation which satisﬁes 1). b) ∈ R we say that a is related to b. 3) If a ≤ b and b ≤ c. with a ≤ b deﬁned by a ⊂ b. A = all subsets of R2 . then If a ∼ b and b ∼ c. Xn is a set.. is a linear ordering.

one is contained in the other. and thus the HMP may be ignored. Show this is a partial ordering which is linear on S = {(a. Deﬁnition A collection of sets is said to be monotonic if. we deﬁne the equivalence class containing a by cl(a) = {x ∈ A : a ∼ x}. One of the most useful applications of the HMP is to obtain maximal monotonic collections of subsets. Corollary to HMP Suppose X is a non-void set and A is some non-void collection of subsets of X. to show that inﬁnitely generated vector spaces have free bases. given any two sets of the collection. First. In elementary courses. Proof Deﬁne a partial ordering on A by V ≤ W iﬀ V ⊂ W. the maximal monotonic subcollection will have a maximal element. Exercise Deﬁne a relation on A = R2 by (a. a) : a < 0}. and 3) is called Exercise Deﬁne a relation on A = Z by n ∼ m iﬀ n − m is a multiple of 3. A relation satisfying properties 1). these results may be assumed. in the Appendix. and second. d) provided a ≤ c and b ≤ d. Deﬁnition If ∼ is an equivalence relation on A and a ∈ A. In each of these applications. Equivalence Relations an equivalence relation.4 Background Chapter 1 on S. and apply HMP. Show this is an equivalence relation. b) ∼ (c. The HMP is that any linearly ordered subset of a partially ordered set is contained in a maximal linearly ordered subset. The HMP is used twice in this book. and S is a subcollection of A which is monotonic. However the ordering may be linear on S but not linear on A. Then ∃ a maximal monotonic subcollection of A which contains S. . 2). to show that rings have maximal ideals (see pages 87 and 109). Find at least two maximal linearly ordered subsets of R2 which contain S.

If each of U. then U = V. In other words.Chapter 1 Theorem 1) Background 5 2) 3) If b ∈ cl(a) then cl(b) = cl(a). Deﬁnition A partition of A is a collection of disjoint non-void subsets whose union is A.e. Then ∼ is an equivalence relation. and the other is the “graph” or “ordered pairs” deﬁnition. Consider the collection of all translates of H. 2).. In either case. Deﬁne a relation on A by a ∼ b iﬀ a and b belong to the same subset of the partition. and the other is as a partition of A into disjoint subsets. Find the equivalence relation on R2 deﬁned by this partition of R2 . there are two ways of deﬁning a function. Summary There are two ways of viewing an equivalence relation — one is as a relation on A satisfying 1). What are the equivalence classes? Exercise Is there a relation on R satisfying 1). Thus we may speak of a subset of A being an equivalence class with no mention of any element contained in it. and the equivalence classes are just the subsets of the partition. Note that if A has an equivalence relation. a collection of non-void subsets of A is a partition of A provided any a ∈ A is an element of one and only one subset of the collection. Functions Just as there are two ways of viewing an equivalence relation. 2a) : a ∈ R}. domain and range are inherent parts of the deﬁnition. We use the “intuitive” deﬁnition because everyone thinks that way. V ⊂ A is an equivalence class and U ∩ V = ∅. i. Each element of A is an element of one and only one equivalence class. is there Exercise Let H ⊂ R2 be the line H = {(a. 2 ) and 3) ? an equivalence relation on R which is also a partial ordering? That is. and 3). the equivalence classes form a partition of A. Exercise Deﬁne an equivalence relation on Z by n ∼ m iﬀ n − m is a multiple of 3. Theorem Suppose A is a non-void set with a partition. . all lines in the plane with slope 2. One is the “intuitive” deﬁnition. 2).

The function is deﬁned by “f (x) is the second term of the ordered pair in Γ whose ﬁrst term is x. if Γ is a subset of X × Y with the property that each x ∈ X is the ﬁrst term of one and only ordered pair in Γ. f ) where f assigns to each x ∈ X a well deﬁned element f (x) ∈ Y . Y. . Y.6 Background Chapter 1 Deﬁnition If X and Y are (non-void) sets. then the graph Γ ⊂ X × Y has the property that each x ∈ X is the ﬁrst term of one and only one ordered pair in Γ. a function or mapping or map with domain X and range Y . f g deﬁne g ◦ f : W → Y by f g h Theorem (The associative law of composition) If V → W → X → Y . deﬁne f | S : S → Y by (f | S)(s) = f (s) for all s ∈ S. Deﬁnition The graph of a function (X. Composition Given W → X → Y (g ◦ f )(x) = g(f (x)). deﬁne the inclusion i : S → X by i(s) = s for all s ∈ S. Suppose y0 ∈ Y . f ) is a function is written f as f : X → Y or X → Y .” Example Identity functions Here X = Y and f : X → X is deﬁned by f (x) = x for all x ∈ X. then ∃! f : X → Y whose graph is Γ. The statement that (X. Example Constant functions y0 for all x ∈ X. f (x)) : x ∈ X}. Inclusion If S is a non-void subset of X. This may be written as h ◦ g ◦ f . Conversely. Y. The identity on X is denoted by IX or just I : X → X. The connection between the “intuitive” and “graph” viewpoints is given in the next theorem. then h ◦ (g ◦ f ) = (h ◦ g) ◦ f. Deﬁne f : X → Y by f (x) = Restriction Given f : X → Y and a non-void subset S of X. Theorem If f : X → Y . f ) is the subset Γ ⊂ X × Y deﬁned by Γ = {(x. is an ordered triple (X. Note that inclusion is a restriction of the identity.

Chapter 1 Deﬁnitions 1) Background Suppose f : X → Y .) Note There is no such thing as “the function sin(x). 7 If T ⊂ Y . Note that f −1 : Y → X is also bijective and (f −1 )−1 = f .” A function is not deﬁned unless the domain and range are speciﬁed. then f (x1 ) and f (x2 ) are distinct elements of Y . (f −1 (x) is written as ln(x). i. i. . 2) 3) 4) 5) 6) Examples 1) 2) 3) 4) f : R → R deﬁned by f (x) = sin(x) is neither surjective nor injective. If S ⊂ X..) (f −1 (x) is 5) f : R → (0.e. if y ∈ Y ... written as arcsin(x) or sin−1 (x). if x1 and x2 are distinct elements of X. the inverse image of T is a subset of X. f : [0. The image of f is the image of X . the image is the range. f : [0. f −1 (T ) = {x ∈ X : f (x) ∈ T }. f : X → Y is injective or 1-1 provided (x1 = x2 ) ⇒ f (x1 ) = f (x2 ). ∞) deﬁned by f (x) = ex is bijective.. π/2] → [0. image (f ) = f (X) = {f (x) : x ∈ X} = {y ∈ Y : ∃x ∈ X with f (x) = y}. there is function f −1 : Y → X with f −1 ◦ f = IX : X → X and f ◦ f −1 = IY : Y → Y . f −1 (y) is a non-void subset of X. f : X → Y is surjective or onto provided image (f ) = Y i.e. i. f (S) = {f (s) : s ∈ S} = {y ∈ Y : ∃s ∈ S with f (s) = y}.e. π/2] → R deﬁned by f (x) = sin(x) is injective but not surjective. 1] deﬁned by f (x) = sin(x) is surjective but not injective. In this case.e. f : R → [−1. f : X → Y is bijective or is a 1-1 correspondence provided f is surjective and injective. 1] deﬁned by f (x) = sin(x) is bijective. the image of S is a subset of Y .

. Find f −1 (f ([1. Exercise Suppose f : [−2. y0 ) : x ∈ X} = (X × y0 ) is called a horizontal strip. then f is not injective. Of course. 2] → R is deﬁned by f (x) = x2 . If m > n. . Pigeonhole Principle Suppose X is a ﬁnite set with m elements. then you have placed 2 pigeons in one hole. 5])). Also ﬁnd the relationship between T and f (f −1 (T )). If x0 ∈ X. as can be seen by the following exercise.8 Background Chapter 1 Exercise Show there are natural bijections from (R × R2 ) to (R2 × R) and from (R2 × R) to R × R × R. S ⊂ X and T ⊂ Y . T = f (f −1 (T )). Exercise Show there is a function f : Z+ → Z+ which is injective but not surjective. Exercise 1) 2) 3) Suppose X is a set with 6 elements and Y is a ﬁnite set with n elements. 1) 2) 3) If m = n. Find the relationship between S and f −1 (f (S)). If you are placing 6 pigeons in 6 holes. If m < n. if f is not surjective then f is not injective. 2])). then f is injective iﬀ f is surjective iﬀ f is bijective. Strips We now deﬁne the vertical and horizontal strips of X × Y . and f : X → Y is a function. y) : y ∈ Y } = (x0 × Y ) is called a vertical strip. in part 1) for m = n = 6. There exists an injective f : X → Y iﬀ n There exists a surjective f : X → Y iﬀ n There exists a bijective f : X → Y iﬀ n . {(x. {(x0 . Also show there is one which is surjective but not injective. Y is a ﬁnite set with n elements. . the pigeonhole principle does not hold for inﬁnite sets. These three sets are disjoint. Show that if f is injective. then f is not surjective. Exercise Suppose f : X → Y is a function. In other words. but the bijections between them are so natural that we sometimes identify them. and you run out of pigeons before you ﬁll the holes. If y0 ∈ Y. S = f −1 (f (S)). Show that if f is surjective. Also ﬁnd f (f −1 ([3.

which are deﬁned after the next theorem. The equation f (x) = y0 has at least one solution for each y0 ∈ Y iﬀ f is . The purpose of the next theorem is to restate properties of functions in terms of horizontal strips. Consider the equation f (x) = y0 . The subset S is the graph of a function with domain X and range Y iﬀ each vertical strip intersects S in exactly one point. Right and Left Inverses One way to understand functions is to study right and left inverses. . Note that the set of all solutions to f (x) = y0 is f −1 (y0 ). A solution to this equation is any x0 ∈ X with f (x0 ) = y0 . Also f (x) = y0 has a solution iﬀ y0 ∈ image(f ) iﬀ f −1 (y0 ) is non-void. Solutions of Equations Now we restate these properties in terms of solutions of equations. Here y0 is given and x is considered to be a “variable”. Theorem 1) Suppose X → Y → W are functions. Theorem 1) 2) 3) Suppose f : X → Y has graph Γ.Chapter 1 Background 9 Theorem Suppose S ⊂ X × Y . . Theorem 1) 2) 3) Suppose f : X → Y . This is just a restatement of the property of a graph of a function. then f is injective. The equation f (x) = y0 has at most one solution for each y0 ∈ Y iﬀ . f g . Suppose f : X → Y and y0 ∈ Y . f is The equation f (x) = y0 has a unique solution for each y0 ∈ Y iﬀ f is . Then Each horizontal strip intersects Γ in at least one point iﬀ f is Each horizontal strip intersects Γ in at most one point iﬀ f is Each horizontal strip intersects Γ in exactly one point iﬀ f is . If g ◦ f is injective.

f (p) = p. Here g ◦ f is the identity. However in this text we do not go that deeply into set theory. Suppose f : X → Y is a function. then f has a right inverse h. The Axiom of Choice If f : X → Y is surjective. q}. Also a function from X to Y is bijective iﬀ it has a left inverse and a right inverse iﬀ it has a left and right inverse. Deﬁnition Suppose f : X → Y is a function. it is possible to choose an x ∈ f −1 (y) and thus to deﬁne h(y) = x. Note It is a classical theorem in set theory that the Axiom of Choice and the Hausdorﬀ Maximality Principle are equivalent. Corollary Suppose each of X and Y is a non-void set. If y belongs to the image of f . . Theorem 1) 2) Suppose f : X → Y is a function. In the next chapter where f is a group homomorphism. Show this is an equivalence relation. A left inverse of f is a function g : Y → X such that g ◦ f = IX : X → X. f has a right inverse iﬀ f is surjective. Then ∃ an injective f : X → Y iﬀ ∃ a surjective g : Y → X. but f is not surjective and g is not injective. Y = {p.10 2) 3) Background If g ◦ f is surjective. Any such right inverse must be injective. then f −1 (y) is an equivalence class and every equivalence class is of this form. However. f has a left inverse iﬀ f is injective. Note The Axiom of Choice is not discussed in this book. If g ◦ f is bijective. for each y ∈ Y . Chapter 1 Example X = W = {p}. For our purposes it is assumed that the Axiom of Choice and the HMP are true. Any such left inverse must be surjective. these equivalence classes will be called cosets. then f is injective and g is surjective. For completeness. then g is surjective. and g(p) = g(q) = p. That is. A right inverse of f is a function h : Y → X such that f ◦ h = IY : Y → Y . you unknowingly used one version of it. if you worked 1) of the theorem above. we state this part of 1) again. Deﬁne a relation on X by a ∼ b if Exercise f (a) = f (b).

Proof Given f . there is a 1-1 correspondence between {functions f: Y → X1 × X2 } and {ordered pairs of functions (f1 . f2 (y)). Formally these sequences are functions α from T to Xt with each α(t) in Xt and written as α(t) = xt . . . Thus a function from Y to X1 × X2 is merely a pair of functions from Y to X1 and Y to X2 . the projection map πs : Xt → Xs is deﬁned by πs ({xt }) = xs . there is a 1-1 correspondence between {functions f : Y → Xt } and {sequences of functions {ft }t∈T where ft : Y → Xt }. Xt = Xt is the collection of all sequences {xt }t∈T = {xt } Then the product where xt ∈ Xt . f2 ) where f1: Y → X1 and f2 : Y → X2 }. xn ). For any T and any s in T . we deﬁne the projection maps π1 : X1 × X2 → X1 and π2 : X1 × X2 → X2 by πi (x1 . and X2 are non-void sets. Given f1 and f2 deﬁne f : Y → X1 × X2 by f (y) = (f1 (y).). . . x2 ) = xi . . t∈T . x2 . 2. f is deﬁned by f (y) = {ft (y)}. Theorem If Y is any non-void set. . . . . It is summarized by the equation f = (f1 . . Given f . Y f1   X1   ©   ' π1     d d f c d f2 d  d X1 × X 2 π2 E X2 One nice thing about this concept is that it works ﬁne for inﬁnite Cartesian products. This concept is displayed in the diagram below. Deﬁnition Suppose T is an index set and for each t ∈ T . If T = Z+ then {xt } is the sequence (x1 . x2 . .Chapter 1 Background 11 Projections If X1 and X2 are non-void sets. Given {ft }. Theorem If Y. the sequence {ft } is deﬁned by ft = πt ◦ f. f2 ). X1 . deﬁne f1 = π1 ◦ f and f2 = π2 ◦ f . Xt is a non-void set. n} then {xt } is the ordered n-tuple (x1 . If T = {1.

4) If S is any subset of T . let Yt be a copy of Y for each t ∈ T. to be the collection of all subsets of T (including the null set). . then Y T has nm elements. (This is the fundamental property of Cartesian products presented in the two previous theorems. Y T = Y × Y × Y has n3 elements. 1}. when T = {1. 1}T → P(T ) by β(f ) = f −1 (1). and be 0 when t ∈ S. 1}T 5) Suppose γ : T → {0.2. and if f : T → {0. It is included here for students who wish to do a little more set theory. P(T ) ←→ {0. Deﬁne f : T → {0. Use the mean value theorem to show that D is injective. 1) If Y is a non-void set. . Show that f is not in the image of γ and thus γ cannot be surjective. Thus α is a bijection and β = α−1 . Let A0 ⊂ A be the subcollection of those functions f with f (0) = 0.) 3) Deﬁne P(T ). 1} then α ◦ β(f ) = f . Show that if S ⊂ T then β ◦ α(S) = S. 1} by f (t) = 0 if ft (t) = 1. then the set {0. deﬁne its characteristic function χS : T → {0. This shows that if T is an inﬁnite set. Show that if n ≥ 3. 1}T is a function and show that it cannot be surjective. Y N is the set of all inﬁnite sequences (y1 . Show that if T and Y are ﬁnite sets with m and n elements.12 Background Chapter 1 A Calculus Exercise Let A be the collection of all functions f : [0. 1] → R which have an inﬁnite number of derivatives. In particular. denote γ(t) by γ(t) = ft : T → {0. P(T ) has 2m elements. T . Suppose T is a non-void set. Then Y T = t∈T 2) Suppose each of Y1 and Y2 is a non-void set. . deﬁne Y T to be the collection of all functions with domain T and range Y .3} of all injective functions has n(n − 1)(n − 2) elements. 1}T represents a “higher order of inﬁnity than T ”. If t ∈ T . 6) An inﬁnite set Y is said to be countable if there is a bijection from the positive .) where each yi ∈ Y . the subset of Y {1. Exercise This exercise is not used elsewhere in this text and may be omitted. Deﬁne D : A0 → A by D(f ) = df /dx. That is. Use the fundamental theorem of calculus to show that D is surjective. Deﬁne β : {0. y2 . Deﬁne α : P(T ) → {0. then Y T is the inﬁnite product Y × Y × · · · . 1} by letting χS (t) be 1 when t ∈ S. the power set of T . Show there is a natural bijection from (Y1 × Y2 )T to Y1T × Y2T . For any Y and Yt . and f (t) = 1 if ft (t) = 0. Show that if T is a ﬁnite set with m elements. 2. If T = N. 1}T by | α(S) = χS . These injective functions are called permutations on Y taken 3 at a time. 3}.

We know that ii) and iii) are equal and there is a natural bijection between i) and ii). iii) The collection of all sequences (y1 . “x ∈ B ” means “x is an element of the set B. ii) {0. (This is not so easy. Mathematical symbols are shorthand for phrases and sentences in the English language. To start with. i) P(N).) where each yi is 0 or 1. then B is true. Conclusion Statement B. Mathematical Induction is based upon the fact that if S ⊂ Z+ is a non-void subset.. 1}N is uncountable. If A is true. 1}N to the real numbers R. the collection of all subsets of N. show there is a bijection from {0.) Notation for the Logic of Mathematics Each of the words “Lemma”. 1}N . Show Q is countable but the following three collections are not. . and “Corollary” means “true statement”. The important thing to remember is that thoughts and expressions ﬂow through the language. or to suppose B is false and show A is false. then “A ⇒ B”means “If x is a positive integer. {0. y2 .Chapter 1 Background 13 integers N to Y. A ⇒ B (A implies B ). . the collection of all functions f : N → {0.e. “Theorem”.” If A is the statement “x ∈ Z+ ” and B is the statement “x2 ∈ Z+ ”. “A is equivalent to B”. then x2 is a positive integer”. i. you have to decide what the real numbers are. Suppose A and B are statements. then S contains a smallest element. Suppose A is true. that A ⇒ B and B ⇒ A). 1}. We also know there is no surjective map from N to {0. 1}N . Finally. . and “A is true iﬀ B is true ” have the same meaning (namely. For example. The expressions “A ⇔ B”. There are two ways to prove the theorem — to suppose A is true and show B is true. . Then B is true. A theorem may be stated in any of the following ways: Theorem Theorem Theorem Theorem Hypothesis Statement A.

± 1 divides anything . we say that b divides a or a is a multiple of b. 3) All of this will follow from long division. Since P (m − 1) is true. . then ∃ a smallest positive integer m such that P (m) is false. Then for each n ≥ 1.e. 2. .. If n ≥ 2. 1) 2) If G is a subgroup of Z. . Note that b | a ⇔ the rational number a/b is an integer ⇔ ∃! m such that a = bm ⇔ a ∈ bZ. Euclidean Algorithm Given a. Proof If the theorem is false. for each n ≥ 1. Exercise Use induction to show that. P (n) is true.. For example. b. g2 ∈ G ⇒ (g1 + g2 ) ∈ G). Here we will establish the following three basic properties of the integers. Deﬁnition A non-void subset G ⊂ Z is a subgroup provided (g ∈ G ⇒ −g ∈ G) and (g1 ... . i. In other words. 0 does not divide anything. and its positive generator is the greatest common divisor of a and b... If a and b are integers.. . then n factors uniquely as the product of primes. −n. The Integers In this section. Also n divides a and b with the same remainder iﬀ n divides (a − b). b divides a “m times with a remainder of r”.14 Background Chapter 1 Theorem Suppose P (n) is a statement for each n = 1. and G is the collection of all linear combinations of a and b. 2n. P (n) ⇒ P (n + 1). will represent integers. 0.}. 1 + 2 + · · · + n = n(n + 1)/2. which we now state formally. then ∃ n ≥ 0 such that G = nZ. Deﬁnition If r = 0. then m = −4 and r = 3. n. the set of integers which n divides is nZ = {nm : m ∈ Z} = {. b with b = 0. then G is a subgroup of Z. this is impossible. Suppose P (1) is true and for each n ≥ 1. ∃! m and r with 0 ≤ r <|b| and a = bm + r. −2n. If n = 0. −17 = 5(−4) + 3.. not both zero. if a = −17 and b = 5. We say that G is closed under negation and closed under addition. Note Anything (except 0) divides 0. This fact is written as b | a. lower case letters a.. elements of Z.. c.

Chapter 1 Background 15 Theorem If n ∈ Z then nZ is a subgroup. Since r ∈ G. In fact. Let n be the smallest positive integer in G. it is the smallest subgroup containing a and b.. b). n ∈ Z}. By the previous theorem. and so 1) is true.. Since G is the smallest subgroup containing a and b. if G = {0} and n is the smallest positive integer in G. b) is called the greatest common divisor of a and b. and g ∈ nZ. 4) Proof of 4) Suppose n | a and n | b i. b) | b. Now (−g) ∈ G and thus 0 = g + (−g) belongs to G. Part 2) is straightforward. . G is a subgroup. the set of integers which n divides is a subgroup of Z. It is called the subgroup generated by a and b. Thus if n = 0. Then 1) 2) 3) G contains a and b. so consider 3). it must be 0. then n also divides (a.e. Denote by (a. b)Z. If G = 0. m2 . suppose a. G = {ma + nb : m.e. b). Also note that ∃ m. and thus n | (a.e. If n is an integer which divides a and b. If g1 and g2 ∈ G. it must contain a positive element. ∃ g ∈ G. b). Now suppose a. The integer (a. b ∈ nZ. b)Z. i. then G = nZ. b) | a and (a. ∃! non-negative integer n such that G = nZ. The next theorem states that every subgroup of Z is of this form. Then 0 ∈ G. b ∈ Z and at least one of a and b is non-zero. Corollary 1) The following are equivalent. g = nm + r where 0 ≤ r < n. (n | a and n | b) ⇒ n = ±1. If g ∈ G. Proof Since G is non-void. b) the smallest positive integer in G. n such that ma + nb = (a. i. nZ ⊃ (a. then (m1 g1 + m2 g2 ) ∈ G for all integers m1 . G = (a. Theorem Let G be the set of all linear combinations of a and b.. a and b have no common divisors. and thus (a. Theorem 1) 2) 3) Suppose G ⊂ Z is a subgroup. In fact.

then a|bc ⇒ a|c. . ∃ m. Part 3) follows from 2) and induction on n.1. This next theorem is the basis for unique factorization.. the subgroup generated by a and b is all of Z. pk and positive integers s1 . Theorem 1) 2) 3) Suppose p is a prime.e. if a is any integer.. (p.. Proof Part 1) follows immediately from the deﬁnition of prime. If p does not divide a.. Thus a | (mac + nbc) and so a | c. sk such that a = ±ps1 ps2 · · · psk . this factorization is unique. 1 2 k Proof Factorization into primes is obvious. Thus 2) is true. then by 1). and uniqueness follows from 3) in the theorem above. or -1. In other words. a) = 1.. except for order.. s2 . Then there exist m.. then p is equal to some ai . a) = p or (p. . If p | ab then p | a or p | b. we say that a and b are relatively prime. then (p. The ﬁrst few primes are 2. a) = 1 and by the previous theorem. Then a may be factored into the product of primes and. If p | a1 a2 · · · an then p divides some ai . if p = ab then a = ±1 or a = ±p. n ∈Z with ma + nb = 1. 5. 13. . p2 . 7. n with ma + nb = 1. c ∈ Z and a | bc. p must divide b. The power of this theorem is uniqueness. Proof Suppose a and b are relatively prime.. Now a | mac and a | nbc. 11. The Unique Factorization Theorem Suppose a is an integer which is not 0. (p. 3. i. That is. not existence. If a is an integer which is not a multiple of p.. b) = 1. Thus if each ai is a prime. a) = 1. Now suppose p | ab.. . Theorem If a and b are relatively prime with a not zero.. Chapter 1 Deﬁnition If any one of these three conditions is satisﬁed. 17. and thus mac + nbc = c.e. Deﬁnition A prime is an integer p > 1 which does not factor.16 2) 3) Background (a. ∃ a unique collection of distinct primes p1 . i.

b).) 2) 3) 3) 4) 5) Exercise Find (180. (Proof: If c is √ rational..) (See the ﬁfth exercise below. then their least common multiple is just their product.28}. ﬁnd the greatest common divisor of 180 and 28. 28). . Then c = pv1 · · · pvk is the least 1 k (positive) common multiple of a and b. 1 k Let vi be the maximum of si and ti . n with ma + nb = 1.. Then 1 k (a. Finally. k. Thus if there is no common prime in their factorizations. this prime will also appear in the prime factorization of a.. Thus a = ±ps1 · · · psk where 0 ≤ si and 1 k b = ±pt1 · · · ptk where 0 ≤ ti . i. . then (a. Find integers m and n such that 180m + 28n = (180. if a and b are positive.. and if in addition a and b are relatively prime. then it is divisible by some prime. For example (23 · 5 · 11.) √ √ Suppose c is an integer √ greater than 1.Chapter 1 Background 17 Now that we have unique factorization and part 3) above. Let ui be the minimum of si and ti . pk .e. . and also (a2 . Then the only 1 k t divisors of a are of the form ±pt1 · · · pkk where 0 ≤ ti ≤ si for i = 1. . and show that it is equal to (180 · 28)/(180. 22 · 54 · 7) = 22 · 5. ∃ positive integers a and b with c = a/b and (a. ∃ m.. There is an inﬁnite number of primes. 28). (Proof: Suppose there were only a ﬁnite number of primes p1 . b) = 1. If b > 1. In particular.. Suppose | a |> 1 and | b |> 1. This is a √ contradiction and thus b = 1 and c is an integer. and if n is a multiple of a and b. p2 . pk } be the union of the distinct primes of their factorizations. . 1 If | a |> 1 and | b |> 1. their least common multiple is c = ab/(a. the picture becomes transparent. i. and since cb2 = a2 . Here are some of the basic properties of the integers in this light.28). b) = 1 iﬀ there is no common prime in their factorizations. b) = pu1 · · · puk .. . b2 ) = 1. Then no prime would divide (p1 p2 · · · pk + 1).e. Note that c is a multiple of a and b. . Then c is rational iﬀ c is an √ √ integer. Theorem (Summary) 1) Suppose | a |> 1 has prime factorization a = ±ps1 · · · psk . then n is a multiple of c. Let {p1 . Find the least common multiple of 180 and 28.. 2 and 3 are irrational. ﬁnd the positive generator of the subgroup generated by {180.

. let a = an an−1 . but they are not to know where. 42} is 2. its nth root is an irrational number. leaving it face up on top. Exercise Show that if each of G1 .. Exercise Show that if the nth root of an integer is a rational number. Also ﬁnd the lcm of the elements of S. Now let b = an + an−1 + · · · + a0 . Now suppose n ≥ 2 and S = {a1 . Thus if p is a prime. Suppose they say three. a2 . then G1 ∩ G2 ∩ · · · ∩ Gm is also a subgroup of Z. That is. When is the lcm of S equal to the product a1 a2 · · · an ? Show that the set of all linear combinations of the elements of S is a subgroup of Z. . Exercise Show that the gcd of S = {90. and ﬁnd integers n1 . suppose c and n are integers greater than 1. There is a unique positive real number x with xn = c.. n3 such that 90n1 + 70n2 + 42n3 = 2. Then repeat the process. . Exercise Show that a positive integer is divisible by 3 iﬀ the sum of its digits is divisible by 3. Deﬁne the gcd and the lcm of the elements of S and develop their properties. where it is shown that any non-zero element of Z7 has order 7. Magic? Stay tuned for Chapter 2. . and bring the seven cards in front in your left hand. and show that 3 divides a and b with the same remainder.. and then you turn over the third card. Now let G = (90Z) ∩ (70Z) ∩ (42Z) and ﬁnd the positive integer n with G = nZ.18 Background Chapter 1 Exercise We have deﬁned the greatest common divisor (gcd) and the least common multiple (lcm) of a pair of integers. it will be more transparent later on. Card Trick Ask friends to pick out seven cards from a deck and then to select one to look at without showing it to you. moving the top two cards to the bottom and turning the third card face up on top. Take the six cards face down in your left hand and the selected card in your right hand. n2 . Continue until there is only one card face down. 70. an } is a ﬁnite collection of integers with |ai | > 1 for 1 ≤ i ≤ n. Show that if x is rational. Gm is a subgroup of Z. . Ask your friends to give you a number between one and seven (not allowing one). a0 = an 10n + an−1 10n−1 + · · · + a0 where 0 ≤ ai ≤ 9. and this will be the selected card. it says that [a] = [b] in Z3 . then the second card to the bottom. then it is an integer. G2 . and announce you will place the selected card in with the other six. In the language of the next chapter. You move the top card to the bottom. More generally. Although this is a straightforward exercise in long division.. . and its positive generator is the gcd of the elements of S. Express the gcd and the lcm in terms of the prime factorizations of the ai . then it itself is an integer. Put your hands behind your back and place the selected card on top.

b) = a + b. φ) is said to be a group. c ∈ G then a · (b · c) = (a · b) · c. it leads to a rich and amazing theory. Consider the following properties. we say it is a multiplicative group. by far and above. rings. If a. c ∈ G then a + (b + c) = (a + b) + c. φ is called a binary operation. 19 . and also the concept of coset is confusing at ﬁrst. 2). b) = a+b. b ∈ G. The approach is to do quickly the fundamentals of groups. b ∈ G. ∃b ∈ G with a · b = b · a = e If a ∈ G. If we write φ(a. 2) ∃ e = eG ∈ G such that if a ∈ G e · a = a · e = a. Everything presented here is standard. 1) If a. the most diﬃcult chapter in the book. b. then a · b = b · a. If a. If we write φ(a. Deﬁnition Suppose G is a non-void set and φ : G × G → G is a function. b) = a·b or φ(a. ∃ 0=0G ∈ G such that if a ∈ G ¯ ¯ 0+a = a+0= a. b. and we will write φ(a. because group operations may be written as addition or multiplication. This chapter is. Also ring homomorphisms and module homomorphisms are special cases of group homomorphisms. Deﬁnition If properties 1). and matrices.Chapter 2 Groups Groups are the central objects of algebra. then a + b = b + a. ¯ ¯ 3) If a ∈ G. and 3) hold. This is the notation used in later chapters for the products of rings and modules. and to push forward to the chapter on linear algebra. 4) If a. This chapter and the next two chapters are restricted to the most basic topics. (b is written as b = −a). b) = a · b. ∃b ∈ G with a + b = b + a = 0 ¯ (b is written as b = a−1 ). except that the product of groups is given in the additive notation. (G. Even though the deﬁnition of group is simple. In later chapters we will deﬁne rings and modules and see that they are special cases of groups.

Theorem (i) Let (G.. If in addition. Recall that b is an identity in G provided it is a right and left identity for any a in G. . ¯ let an = (a + · · +a) where the sum is n times. However. ¯ ¯ ¯ Also c · a = c · a ⇒ c = c. b ∈ G. i. ¯ ¯ if a. if f : G → G is deﬁned by f (c) = a · c. then e = e. Then a · c = a · c ⇒ c = c. φ) be a multiplicative group. Thus inverses are unique. and a(−n) = (−a) + (−a) · · + (−a). If a ∈ G. This means that if a is in G and n is an integer. group structure is so rigid that if ∃ a ∈ G such that b is a right identity for a. Write out the above theorem where G is an additive group. i. b ∈ G. if G is abelian and a. c ∈ G. c. e is unique. In general. then (a · b)n = an · bn .. In fact.e. there is deﬁned an element an in G.. if a · b = e then b = a−1 . If n1 .20 Groups Chapter 2 we say it is an additive group.e. Also (an )m = anm . This is so basic. Also f is bijective with f −1 given by f −1 (c) = a−1 · c. this is just a special case of the cancellation law in (i). then f is injective. . let a0 =0 and if n > 0. Let a0 = e and if n > 0. 1 n−1 n Suppose a ∈ G. if e ∈ G satisﬁes 2). Finally. ¯ ¯ In other words. Also if b · a = e then b = a−1 . an = a · · · a (n times) and a−n = a−1 · · · a−1 (n times).. Suppose a. Every right inverse is an inverse. Also (a1 · a2 · · · an )−1 = a−1 · a−1 · · · a−1 . Suppose G is an additive group. then b = e. (ii) (iii) (iv) (v) (vi) (vii) Exercise. The multiplication a1 · a2 · a3 = a1 · (a2 · a3 ) = (a1 · a2 ) · a3 is well-deﬁned. Note that part (vii) states that G has a scalar multiplication over Z. If a. we say the group is abelian or commutative. property 4) holds. a1 · a2 · · · an is well deﬁned. If a ∈ G. n2 . Theorem. b ∈ G then (a · b = a) ⇒ (b = e) and (a · b = b) ⇒ (a = e). (a · b)−1 = b−1 · a−1 . then (a−1 )−1 = a. Of course.. nt ∈ Z then an1 · an2 · · · ant = an1 +···+nt . that we state it explicitly.

and f has a left inverse in G iﬀ f is injective (see page 10). Then the following properties hold in general. and thus G is not a group. b) = ab is a multiplicative abelian group. ∃b ∈ G with a · b = e]. Show (G. b) = a·b which satisﬁes 1). or G = Z with φ(a. In the language used in Chapter 5. 2) and [ 3 ) If a ∈ G. b ∈ H then a · b ∈ H 2) if a ∈ H then a−1 ∈ H. Exercise Suppose G is the set of all functions from Z to Z with multiplication deﬁned by composition. except the ﬁrst requires that G be abelian. i. If every element has a right inverse. G = Z − 0 with φ(a. In other words.) Exercise Suppose G is a non-void set with a binary operation φ(a. . Show that f has a right inverse in G iﬀ f is surjective. b) = a + b is an additive abelian group. G = R − 0 or G = Q − 0 with φ(a.Chapter 2 Groups 21 which we write as (−a − a · · − a). Note that G satisﬁes 1) and 2) but not 3).e. Examples G = R. (a + b)n a(n + m) a(nm) a1 = = = = an + bn an + am (an)m a Note that the plus sign is used ambiguously — sometimes for addition in G and sometimes for addition in Z. φ) is a group. G = Q. (See page 71. Subgroups Theorem satisfying and Suppose G is a multiplicative group and H ⊂ G is a non-void subset Examples 1) if a.. f · g = f ◦ g. G = R+ = {r ∈ R : r > 0} with φ(a.e. b) = ab is a multiplicative abelian group. i. Also show that the set of all bijections from Z to Z is a group under composition. this theorem states that any additive abelian group is a Z-module. show b · a = e.. b) = ab is not a group. the group axioms are stronger than necessary. then every element has a two sided inverse.

Example G is a subgroup of G and e is a subgroup of G. The associative law is immediate and so H is a group.e. a−1 ∈ H and so by 1). 1] → R}. Let H be the subset of G composed of all continuous functions. Let K be the subset of G composed of all diﬀerentiable functions. By 2). These are called the improper subgroups of G. Suppose H1 and H2 are subgroups of G. then Ht is a subgroup of G. with neither H1 nor H2 contained in the other. Show H1 ∪ H2 is not a subgroup of G. H is called a subgroup of G. What theorems in calculus show that H and K are subgroups of G? What theorem shows that K is a subset (and thus subgroup) of H? Order Suppose G is a multiplicative group. Deﬁne an addition on G by (f + g)(t) = f (t) + g(t) for all t ∈ [0. Suppose T is an index set and for each t ∈ T . e ∈ H. i. This is a key property of the integers. Show H1 ∩ H2 is a subgroup of G. Show Ht is a subgroup of G. Proof Since H is non-void. ∃a ∈ H. then H = nZ is a subgroup of Z. This makes G into an abelian group. 1]. every subgroup of Z is of this form (see page 15). Let H be the center of G. Suppose H1 and H2 are subgroups of G. H = {h ∈ G : g · h = h · g for all g ∈ G}. Example If G = Z under addition. t∈T Suppose G= {all functions f : [0. Show H is a subgroup of G. Exercises 1) Suppose G is a multiplicative group.. By a theorem in the section on the integers in Chapter 1. if {Ht } is a monotonic collection.22 Groups Chapter 2 Then e ∈ H and H is a group under multiplication. Ht is a subgroup of G. t∈T 2) 3) 4) 5) 6) Furthermore. and n ∈ Z. If G has an inﬁnite number of .

Note that in the case G has ﬁnite order n. then by the Euclidean algorithm. Exercise Show that if G is a ﬁnite group of even order. and H = {ai : i ∈ Z}.. and i H = {a : i ∈ Z}. If H = e. Proof Suppose G = {ai : i ∈ Z} is a cyclic group and H is a subgroup of G. Suppose a ∈ G and H = {ai : i ∈ Z}. We deﬁne the order of the element a to be the order of H. Also note that this theorem was proved on page 15 for the additive group Z. Then ai−j = e and thus ∃ a smallest positive integer n with an = e. m divides t. then G has an odd number of elements of order 2. and we must show they are all of H. the order of a is the smallest positive integer n with an = e. then a has some ﬁnite order n. i. and thus the positive integer m divides n. . a1 . a1 . Proof Suppose j < i and ai = aj . then o(G) = n. . an−1 }. We come now to the ﬁrst real theorem in group theory. .Chapter 2 Groups 23 elements. Let f : Z → H be the surjective function deﬁned by f (m) = am .. Note that f (k + l) = f (k) · f (l) where the addition is in Z and the multiplication is in the group H. Note that e is the only element of order 1. . . then H is cyclic. If m ∈ Z. so suppose H = e.. we say that o(G). In either case. and so H = {a0 . and in this case. suppose a is an element of an additive group G. Deﬁnition Theorem A group G is cyclic if ∃ an element of G which generates G...e. To begin. H is isomorphic to Z or Zn . H is an abelian subgroup of G called the subgroup generated by a.. It says that the element a has ﬁnite order iﬀ f is not injective. i. H = {a0 . in additive notation.. Later in this chapter we will see that f is a homomorphism from an additive group to a multiplicative group and that. . and am = e iﬀ n|m. Thus am = anq · ar = ar . If G is cyclic and H is a subgroup of G. an−1 }. then an = e ∈ H. . In particular. Now there is a smallest positive integer m with am ∈ H. is inﬁnite. This implies that the elements of {a0 . and f −1 (e) = nZ. and am = e iﬀ n|m. If G has n elements. G = {a0 . .e. the Euclidean algorithm states that ∃ integers q and r with 0 ≤ r < n and m = nq + r. an−1 }. and thus am generates H. an−1 } are distinct.. the order of a is the smallest positive integer n with an = e. we have a clear picture of the subgroups of G. Theorem Suppose a is an element of a multiplicative group G. a1 . In this case H has n distinct elements. . If t is an integer with at ∈ H. the order of G. If ∃ distinct integers i and j with ai = aj . the order of the subgroup generated by a. Exercise Write out this theorem for G an additive group. a1 . then H is cyclic. . .

24

Groups

Chapter 2

Cosets Suppose H is a subgroup of a group G. It will be shown below that H partitions G into right cosets. It also partitions G into left cosets, and in general these partitions are distinct. If H is a subgroup of a multiplicative group G, then a ∼ b deﬁned by Theorem −1 a ∼ b iﬀ a · b ∈ H is an equivalence relation. If a ∈ G, cl(a) = {b ∈ G : a ∼ b} = {h · a : h ∈ H} = Ha. Note that a · b−1 ∈ H iﬀ b · a−1 ∈ H. If H is a subgroup of an additive group G, then a ∼ b deﬁned by a ∼ b iﬀ (a − b) ∈ H is an equivalence relation. If a ∈ G, cl(a) = {b ∈ G : a ∼ b} = {h + a : h ∈ H} = H + a. Note that (a − b) ∈ H iﬀ (b − a) ∈ H. Deﬁnition These equivalence classes are called right cosets. If the relation is deﬁned by a ∼ b iﬀ b−1 · a ∈ H, then the equivalence classes are cl(a) = aH and they are called left cosets. H is a left and right coset. If G is abelian, there is no distinction between right and left cosets. Note that b−1 · a ∈ H iﬀ a−1 · b ∈ H. In the theorem above, H is used to deﬁne an equivalence relation on G, and thus a partition of G. We now do the same thing a diﬀerent way. We deﬁne the right cosets directly and show they form a partition of G. You might ﬁnd this easier. Theorem Suppose H is a subgroup of a multiplicative group G. If a ∈ G, deﬁne the right coset containing a to be Ha = {h · a : h ∈ H}. Then the following hold. Ha = H iﬀ a ∈ H. If b ∈ Ha, then Hb = Ha, i.e., if h ∈ H, then H(h · a) = (Hh)a = Ha. If Hc ∩ Ha = ∅, then Hc = Ha. The right cosets form a partition of G, i.e., each a in G belongs to one and only one right coset. 5) Elements a and b belong to the same right coset iﬀ a · b−1 ∈ H iﬀ b · a−1 ∈ H. 1) 2) 3) 4) Proof There is no better way to develop facility with cosets than to prove this theorem. Also write this theorem for G an additive group.

Theorem

Suppose H is a subgroup of a multiplicative group G.

Chapter 2 1)

Groups

25

Any two right cosets have the same number of elements. That is, if a, b ∈ G, f : Ha → Hb deﬁned by f (h · a) = h · b is a bijection. Also any two left cosets have the same number of elements. Since H is a right and left coset, any two cosets have the same number of elements. G has the same number of right cosets as left cosets. The function F deﬁned by F (Ha) = a−1 H is a bijection from the collection of right cosets to the left cosets. The number of right (or left) cosets is called the index of H in G. If G is ﬁnite, o(H) (index of H) = o(G) and so o(H) | o(G). In other words, o(G)/o(H) = the number of right cosets = the number of left cosets. If G is ﬁnite, and a ∈ G, then o(a) | o(G). (Proof: The order of a is the order of the subgroup generated by a, and by 3) this divides the order of G.) If G has prime order, then G is cyclic, and any element (except e) is a generator. (Proof: Suppose o(G) = p and a ∈ G, a = e. Then o(a) | p and thus o(a) = p.) If o(G) = n and a ∈ G, then an = e. (Proof: ao(a) = e and n = o(a) (o(G)/o(a)) .)

2)

3)

4)

5)

6)

Exercises i) Suppose G is a cyclic group of order 4, G = {e, a, a2 , a3 } with a4 = e. Find the order of each element of G. Find all the subgroups of G. Suppose G is the additive group Z and H = 3Z. Find the cosets of H. Think of a circle as the interval [0, 1] with end points identiﬁed. Suppose G = R under addition and H = Z. Show that the collection of all the cosets of H can be thought of as a circle. Let G = R2 under addition, and H be the subgroup deﬁned by H = {(a, 2a) : a ∈ R}. Find the cosets of H. (See the last exercise on p 5.) Normal Subgroups We would like to make a group out of the collection of cosets of a subgroup H. In

ii) iii)

iv)

26

Groups

Chapter 2

general, there is no natural way to do that. However, it is easy to do in case H is a normal subgroup, which is described below. Theorem 1) 2) 3) 4) If H is a subgroup of a group G, then the following are equivalent. If a ∈ G, then aHa−1 = H If a ∈ G, then aHa−1 ⊂ H If a ∈ G, then aH = Ha Every right coset is a left coset, i.e., if a ∈ G, ∃ b ∈ G with Ha = bH.

Proof 1) ⇒ 2) is obvious. Suppose 2) is true and show 3). We have (aHa−1 )a ⊂ Ha so aH ⊂ Ha. Also a(a−1 Ha) ⊂ aH so Ha ⊂ aH. Thus aH = Ha. 3) ⇒ 4) is obvious. Suppose 4) is true and show 3). Ha = bH contains a, so bH = aH because a coset is an equivalence class. Thus aH = Ha. Finally, suppose 3) is true and show 1). Multiply aH = Ha on the right by a−1 . Deﬁnition If H satisﬁes any of the four conditions above, then H is said to be a normal subgroup of G. (This concept goes back to Evariste Galois in 1831.) Note For any group G, G and e are normal subgroups. If G is an abelian group, then every subgroup of G is normal. Exercise Show that if H is a subgroup of G with index 2, then H is normal.

Exercise Show the intersection of a collection of normal subgroups of G is a normal subgroup of G. Show the union of a monotonic collection of normal subgroups of G is a normal subgroup of G. Exercise Let A ⊂ R2 be the square with vertices (−1, 1), (1, 1), (1, −1), and (−1, −1), and G be the collection of all “isometries” of A onto itself. These are bijections of A onto itself which preserve distance and angles, i.e., which preserve dot product. Show that with multiplication deﬁned as composition, G is a multiplicative group. Show that G has four rotations, two reﬂections about the axes, and two reﬂections about the diagonals, for a total of eight elements. Show the collection of rotations is a cyclic subgroup of order four which is a normal subgroup of G. Show that the reﬂection about the x-axis together with the identity form a cyclic subgroup of order two which is not a normal subgroup of G. Find the four right cosets of this subgroup. Finally, ﬁnd the four left cosets of this subgroup.

[n − 1]. Proof Multiplication of elements in G/N is multiplication of subsets in G. Also Zn is cyclic because each of [1] and [−1] = [n − 1] is a generator. Proof The element [a] is a generator iﬀ the subgroup generated by [a] contains [1] iﬀ ∃ an integer k such that [a]k = [1] iﬀ ∃ integers k and l such that ak + nl = 1. . It follows that they honor identities and inverses. and G/N is the collection of all cosets. The coset E does not depend upon the choice of c and d. We already know that if p is a prime. Theorem Suppose G is a multiplicative group. i. n > 1. Furthermore. Then (N a) · (N b) = N (a · b) is a well deﬁned multiplication (binary operation) on G/N . n) = 1. if G is ﬁnite.. In this section we list .Chapter 2 Groups 27 Quotient Groups Suppose N is a normal subgroup of G. Zn . This is made precise in the next theorem. and C and D are cosets.) Homomorphisms Homomorphisms are functions between groups that commute with the group operations.. Note that [a] = [r] where r is the remainder of a divided by n. [1]. (N a) · (N b) = N (aN )b = N (N a)b = N (a · b). N is a normal subgroup. determine those elements of ﬁnite order. If a is an integer. Note that [a] + [b] = [a + b]. Exercise Show that a positive integer is divisible by 3 iﬀ the sum of its digits is divisible by 3. Once multiplication is well deﬁned. and thus the distinct elements of Zn are [0]. In the additive abelian group R/Z. and [a] = [a + nl] for any integer l. Its identity is N and (N a)−1 = (N a−1 ). Exercise Write out the above theorem for G an additive group. then [a] is a generator of Zn iﬀ (a. and N = nZ. the group axioms are immediate.. Example Suppose G = Z under +. (See the ﬁfth exercise on page 18. the group of integers mod n is deﬁned by Zn = Z/nZ. Note that [10] = [1] in Z3 .e. G/N is a group. Theorem If n > 1 and a is any integer. −[a] = [−a]. We wish to deﬁne a coset E which is the product of C and D. and with this multiplication. o(G/N ) = o(G)/o(N ).. E = N (c · d). deﬁne E to be the coset containing c · d. and in this case it is just [a]m = [am]. the coset a + nZ is denoted by [a]. because Zp has p elements. which is quite easy. any non-zero element of Zp is a generator. Any additive abelian group has a scalar multiplication over Z. If c ∈ C and d ∈ D.

if H is ¯ ¯ normal in G. if the equation f (x) = g has a ¯ ¯ Suppose G and G are groups and f : G → G is a homomorphism. and the second is in G. the kernel is the set of e ¯ ¯ solutions to the equation f (x) = e. if f (g) = g then ¯ ¯ g ¯ −1 g ¯ f (¯) = N g where N = ker(f ). f (a · b) = f (a) · f (b). the student should rewrite the material in additive notation. a function f : G → G is a homomorphism if. 12).e. for all a. Furthermore. (If G is an additive group. and should be considered as the cornerstones of abstract algebra. ¯ If H is a subgroup of G. In other words. ¯ ¯ ¯ ¯ If H is a subgroup of G. f is injective ⇔ ker(f ) = e. In particular. f −1 (H) is a subgroup of G. In other words. the inclusion i : H → G is a homomorphism. while on the right side it is in G. f −1 (¯) is void or is a coset of ker(f ). ker(f ) = f −1 (0). the group ¯ operation is in G. ¯ ¯ f (a−1 ) = f (a)−1 . ¯ ¯ Deﬁnition If G and G are multiplicative groups. The function f : Z → Z deﬁned by f (t) = 2t is a homomorphism of additive groups. 5) 6) 7) . We now catalog the basic properties of homomorphisms.. then f −1 (H) is normal in G. The function h : Z → R − 0 deﬁned by h(t) = 2t is a homomorphism from an additive group to a multiplicative group. b ∈ G. If g ∈ G. As always.28 Groups Chapter 2 the basic properties. while the function deﬁned by f (t) = t + 2 is not a homomorphism. f (H) is a subgroup of G. image(f ) is ¯ a subgroup of G. The ﬁrst inverse is in G. Properties 11). The kernel of f is a normal subgroup of G. These will be helpful later on in the study of ring homomorphisms and module homomorphisms. and 13) show the connections between coset groups and homomorphisms.) ¯ ¯ ¯ Examples The constant map f : G → G deﬁned by f (a) = e is a homomorphism. The kernel of f is deﬁned by −1 ker(f ) = f (¯) = {a ∈ G : f (a) = e}. ¯ If H is a subgroup of G. Theorem 1) 2) 3) 4) f (e) = e. i. On the left side.

= 9) 10) Isomorphisms preserve all algebraic properties. f is called an isomorphism. then H is a subgroup of G ¯ ¯ iﬀ f (H) is a subgroup of G. ¯ f is injective. ¯ In this case. by deﬁnition. The ¯ ¯ ¯ image of f is the image of f and the kernel of f is ker(f )/H. and we write G ≈ G. then G/H ≈ G (see below). Then π : G → G/H deﬁned by π(a) = Ha is a surjective homomorphism with kernel H. ¯ ¯ If f : G → G is a bijection. i. For example. then f : G/H → G ¯(Ha) = f (a) is a well-deﬁned homomorphism making deﬁned by f the following diagram commute. This is a key fact which is used routinely in topics such as systems of equations and linear diﬀerential equations. if h : G →G is = a homomorphism. then the set of all solutions is a coset of N = ker(f ). then the function f −1 : G → G is a homomorphism. G π G/H c& & f & & ¯ E b & & ¯ G f Thus deﬁning a homomorphism on a quotient group is the same as deﬁning a homomorphism on the numerator which sends the denominator to e. domain(f )/ker(f ) ≈ image(f ). G is ¯ is cyclic. and thus G/H ≈ image(f ).. . If H ⊂ ker(f ). Of course. because an cyclic iﬀ G algebraic property is one that. H is normal in G iﬀ f (H) is normal in G. 13) Given any group homomorphism f . f is also called an automorphism. 11) Suppose H is a normal subgroup of G.e. This is the fundamental connection between quotient groups and homomorphisms. this is somewhat of a cop-out. Furthermore.Chapter 2 Groups 29 solution. ¯ ¯ 12) Suppose H is a normal subgroup of G. if f is an isomorphism and H ⊂ G is a subset. is preserved under isomorphisms. 8) ¯ The composition of homomorphisms is a homomorphism. Thus if H = ker(f ). if ¯ ¯ f : G → G is a surjective homomorphism with kernel H. In the case ¯ G = G. then h ◦ f : G →G is a homomorphism. etc.

Under what conditions on g is there a homomorphism f : Z7 → G with f ([1]) = g ? Under what conditions on g is there a homomorphism f : Z15 → G with f ([1]) = g ? Under what conditions on G is there an injective homomorphism f : Z15 → G ? Under what conditions on G is there a surjective homomorphism f : Z15 → G ? 2) 3) 4) Exercise We know every ﬁnite group of prime order is cyclic and thus abelian. Then G is a group under addition. Show that every group of order four is abelian. Find f −1 (g) and show it is a coset of ker(f ). Then f is a group homomorphism and the diﬀerential equation h +5h +6t2 h = g has a solution iﬀ g lies in the image of f . Then f : Z → K deﬁned by f (m) = am is a homomorphism from an additive group to a multiplicative group. 1] → R be deﬁned by g(t) = t3 − 3t + 4. If o(a) = n. Exercise If a is an element of a group G. Deﬁne f : G → G by f (h) = dh = h . Then K is an inﬁnite cycle group iﬀ K is isomorphic to the integers under addition. When is there a homomorphism from Zn to G which sends [1] to a? What are the homomorphisms from Z2 to Z6 ? What are the homomorphisms from Z4 to Z8 ? Exercise 1) Suppose G is a group and g is an element of G. Exercise Let G be as above and g ∈ G. K is a cyclic group of order n iﬀ K ≈ Zn . Exercise Let G = {h : [0.. Now suppose this equation has a solution and S ⊂ G is the set of all solutions. If o(a) is inﬁnite.30 Groups Chapter 2 14) Suppose K is a group. there is always a homomorphism from Z to G which sends 1 to a. ¯ Proof of 14) Suppose G = K is generated by some element a. For which subgroup H of G is S an H-coset? . Let g : [0. Deﬁne f : G → G by f (h) = h + 5h + 6t2 h. ker(f ) = nZ and ¯ f : Zn → K is an isomorphism. 1] → R : h has an inﬁnite number of derivatives}. g = e.e. Show f dt is a homomorphism and ﬁnd its kernel and image. f is an isomorphism. K ≈ Z. i.

Deﬁne f : G → G to be conjugation by a. Also show f is an automorphism and ﬁnd its inverse. ha·b = ha ◦ hb .. Theorem (Cayley’s Theorem) Suppose G is a multiplicative group with n elements and Sn is the group of all permutations on the set G. They are all denoted by the one symbol Sn . i. 2. Exercise Show that o(Sn ) = n!. and H = {f ∈ Sn : (n)f = n}. i.e. Let X = {1. The proof follows from the following observations.. variables are written on the left. i. A bijection f : X → X is called a permutation on X.. there is an isomorphism between S(X) and S(Y ). n}. Let g be any permutation on X with (n)g = 1... 2. f = (x)f . Therefore the composition f ◦ g means “f followed by g”. S(X) forms a multiplicative group under composition. Permutations Suppose X is a (non-void) set. Then G is isomorphic to a subgroup of Sn . Thus if each of X and Y has n elements. and the collection of all these permutations is denoted by S = S(X). Proof Let h : G → Sn be the function which sends a to the bijection ha : G → G deﬁned by (g)ha = g · a.Chapter 2 Groups 31 Exercise Suppose G is a multiplicative group and a ∈ G.. n}. f (g) = a−1 · g · a. The Symmetric Groups Now let n ≥ 2 and let Sn be the group of all permutations on {1. h is a homomorphism. Show H is a subgroup of Sn which is isomorphic to Sn−1 .e. and these groups are called the symmetric groups on n elements.. Sn = S(X).e. . .. Show that f is a homomorphism. ha is a bijection from G to G. Exercise Show that if there is a bijection between X and Y . The following deﬁnition shows that each element of Sn may .. In this setting. The next theorem shows that the symmetric groups are incredibly rich and complex. h is injective and thus G is isomorphic to image(h) ⊂ Sn . S(X) ≈ S(Y ). 1) 2) 3) For each given a. Find g −1 Hg.

. .32 be represented by a matrix... b2 . (This is obvious. . a2 )(a1 . In this case. Listed here are eight basic properties of permutations. c ) are disjoint provided ai = cj for all 1 ≤ i ≤ k and 1 ≤ j ≤ . a1 a2 . bk } is the same collection in some diﬀerent order.. (This is easy. then g ◦ f is 3) 4) 5) 6) ...) Every nonidentity permutation can be written uniquely (except for order) as the product of disjoint cycles... which takes a little work.. bk and (a)f = a for all other a. . a3 ) · · · (a1 .. and is denoted by (a1 .ak−1 ak is called a k-cycle. 2) and (a1 . then p is even iﬀ q is even.... ak } is a collection of distinct integers with 1 ≤ ai ≤ n.. Suppose f. {a1 . .) Every permutation can be written (non-uniquely) as the product of transpositions. f is an odd permutation... f is said to be an even permutation. Deﬁnition a1 a2 . 2)(1. a2 a3 . 2)(1. The cycles (a1 . ak ). 3) = (1... 2. For example (1. ak ).. Properties 9) and 10) are listed solely for reference. a2 . ak ) and (c1 . . g ∈ Sn . Theorem 1) 2) Disjoint cycles commute. They are all easy except 4). b1 b2 .. For these we have a special notation. and {b1 .. If one of f and g is even and the other is odd. This means that if f is the product of p transpositions and also of q transpositions. ) The parity of the number of these transpositions is unique..... A k-cycle is even (odd) iﬀ k is odd (even). a2 . In the other case.ak a1 A 2-cycle is called a transposition. Groups Chapter 2 Deﬁnition Suppose 1 < k ≤ n. 3) is an even permutation. ak ) = (a1 ... There is a special type of permutation called a cycle. The composition of two permutations is computed by applying the matrix on the left ﬁrst and the matrix on the right second. ak Then the matrix represents f ∈ Sn deﬁned by (ai )f = bi for 1 ≤ i ≤ k. (Proof: I = (1.

b)(b. 2. . c. where t is the smallest odd integer this is possible. If f and g are both even or both odd. then (a. n). and (a.1) as the product of disjoint cycles. n)(b. (1.5. . b)(a. 6 5 4 3 1 7 2 Write (1. c) = (a. . {(1. b) and (a. 8) The following parts are not included in this course. then (a. n}. This can be done by inductively “pushing n to the right” using the equations below.4) as the product of disjoint cycles. b). 9) For any n = 4.. . c) = (b. n)(a.3. Since I = (1. Since t is odd. n) = I. If a.7. Which of these permutations are odd and which are even? Write . . b. (Of course there are subgroups of Sn which cannot be generated by two elements).. n). c)(c. n)} generates Sn . has no proper normal subgroups. it is at least 3. Proof of 4) It suﬃces to prove if the product of t transpositions is the identity I on {1. 2. b)(c. c)(a.4)(3. every even permutation is the product of 3-cycles. n). n)(a. . 3)3 . n)(b. c. . Thus An is a normal subgroup of index 2. 2. They are presented here merely for reference. (a. (a.. i.6.Chapter 2 Groups 33 odd. then g ◦ f is even. Its kernel (the subgroup of even permutations) is denoted by An and is called the alternating group.3. 2. An is simple. 2. (Now you may solve the tile puzzle on page viii. b. Note that (a. and Sn /An ≈ Z2 . n) cannot occur here because it would result in a shorter odd product. ..6.1)(1. . (Obvious.7)(2. 10) Sn can be generated by two elements. d)(a. c and d are distinct integers in {1. . then t is even. If a. . c.5.) 7) The map h : Sn → Z2 deﬁned by h(even)= [0] and h(odd)= [1] is a homomorphism from a multiplicative group to an additive group. . Suppose this is false and I is written as t transpositions. We will rewrite I as a product of transpositions σ1 σ2 · · · σt where (n)σi = (n) for 1 ≤ i < t and (n)σt = n. n − 1}. and c are distinct integers in {1. which will be a contradiction. In fact. Suppose for convenience the ﬁrst transposition is (a. d) = (a.7. n).) Exercise 1) 1 2 3 4 5 6 7 as the product of disjoint cycles. . Write (3. c) = (a.e. n) = (a. it follows that for n ≥ 3.7)(2. n)(a. n}. 2).

Deﬁne an addition on G1 × G2 by (a1 . op Show that an element of Sn is a permutation on {1. . . −1). a2 + b2 ). Let A ⊂ R2 be the square with vertices (−1. Show that it has the same identity and the same inverses as G. and G be the collection of all isometries of A onto itself. n} with the variable op written on the right. this section should be rewritten using multiplicative notation. −a2 ). Theorem Suppose G1 and G2 are additive groups. . 2. and that f : G → Gop deﬁned by f (a) = a−1 is a group isomorphism. either Sn or Sn may be called the symmetric group. a2 ) + (b1 . Its “zero” is (01 . Also let τ = (4. (3)σ). We know from a previous exercise that G is a group with eight elements. a2 ) = (−a1 . 1).) Product of Groups 3) 4) 5) 6) The product of groups is usually presented for multiplicative groups. 5. . Now consider the special case G = Sn . For simplicity we ﬁrst consider the product of two groups. deﬁne a new multiplication on the set G by a ◦ b = b · a. This operation makes G1 × G2 into a group. 6) and ﬁnd τ −1 (1. This deﬁnes a new group denoted by Gop . 2. depending on personal preference or context. (1. b2 ) = (a1 + b1 . The two theorems below are transparent and easy. although the case of inﬁnite products is only slightly more diﬃcult. n} with the variable written on the left. 3. . This shows that conjugation by σ is just a type of relabeling. the new multiplication is the old multiplication in the opposite order. . . . In other words. It is presented here for additive groups because this is the form that occurs in later chapters. . What is the order of their product? Suppose σ ∈ Sn . For background. It follows from Cayley’s theorem that G is isomorphic to a subgroup of S8 . 4. and (−1. (1. . Show that σ −1 (1. . 2. read ﬁrst the two theorems on page 11. 5)τ . but quite useful. . The convention used in this section is that an element of Sn is a permutation on {1. 2. Show that G is isomorphic to a subgroup of S4 . . c ) are disjoint cycles. . . . (Of course.34 2) Groups Chapter 2 Suppose (a1 . The projections π1 : G1 × G2 → G1 ¯ ¯ . −1). (2)σ. ak ) and (c1 . Show that H = {σ ∈ S6 : (6)σ = 6} is a subgroup of S6 and ﬁnd its right cosets and its left cosets. If G is a multiplicative group. 02 ) and −(a1 . As an exercise. 3)σ = ((1)σ. the opposite group. 1).

Suppose G. a2 ) in G1 × G2 . so let’s prove the last part. Exercise Suppose G1 and G2 are groups. Exercise Let R be the reals under addition. f2 ) where f1 : G → G1 and f2 : G → G2 }. Proof It is transparent that the product of groups is a group. Finally. Under this bijection. Exercise Suppose G1 and G2 are groups and i1 : G1 → G1 × G2 is deﬁned by i1 (g1 ) = (g1 . Exercise Suppose n > 2. f2 (a) + f2 (b)). f2 (a)) + (f1 (b). Show that the addition in the product R × R is just the usual addition in analytic geometry. Show that G1 × G2 and G2 × G1 are If o(a1 ) = m and o(a2 ) = n. f2 (a + b)) and f (a) + f (b) = (f1 (a). f2 (b)) = (f1 (a) + f1 (b). show that Zmn is isomorphic to Zm × Zn iﬀ (m. Is Sn isomorphic to An × G where G is a multiplicative group of order 2 ? One nice thing about the product of groups is that it works ﬁne for any ﬁnite number. n) = 1. . or even any inﬁnite number. An examination of these two equations shows that f is a group homomorphism iﬀ each of f1 and f2 is a group homomorphism. Show Z12 is isomorphic to Z4 × Z3 . and G2 are groups and f = (f1 . Suppose G is an additive group. Show Z4 is not isomorphic to Z2 × Z2 . f2 ) is a function from G to G1 × G2 . Exercise isomorphic. Now f (a + b) = (f1 (a + b). ﬁnd the order of (a1 . Exercise Show that if G is any group of order 4. G1 . f is a group homomorphism iﬀ each of f1 and f2 is a group homomorphism. Let π2 : G1 × G2 → G2 be the projection map deﬁned in the Background chapter. We know there is a bijection from {functions f : G → G1 × G2 } to {ordered pairs of functions (f1 . Usually G1 is identiﬁed with its image under i1 . Show i1 is an injective group homomorphism and its image is a ¯ normal subgroup of G1 × G2 . so G1 may be considered to be a normal subgroup of G1 × G2 . Show π2 is a surjective homomorphism with kernel G1 . The next theorem is stated in full generality. G is isomorphic to Z4 or Z2 ×Z2 .Chapter 2 Groups 35 and π2 : G1 × G2 → G2 are group homomorphisms. Therefore (G1 × G2 )/G1 ≈ G2 as you would expect. 02 ).

c. Its “zero” is {0t } and −{at } = {−at }. Show GT is a group. [m]. {at }n = {at n}. (For the ring and module versions. see exercises on pages 44 and 69. Note that the gcd of {45. [m]). Now let h : Z → Z8 × Z9 × Z35 be deﬁned by h(m) = ([m]. [m]). Exercise Let f : Z → Z30 × Z100 be the homomorphism deﬁned by f (m) = ([4m]. Note that GT is just another way of writing Gt .e.. [3]) in Z30 × Z100 . Exercise Suppose T is a non-void set.) . This opt∈T eration makes the product into a group. Find the kernel of f. the addition deﬁned on t∈T GT is just the usual addition of functions used in calculus. Thus each Gs may be considered to be a normal subgroup of Gt . Under the natural bijection from {functions f : G → Gt } to {sequences of functions {ft }t∈T where ft : G → Gt }. i. the scalar multiplication on Gt by integers is given coordinatewise. Exercise Let f : Z → Z90 × Z70 × Z42 be the group homomorphism deﬁned by f (m) = ([m]. Gt is an additive group. [m]. [m]. ¯ Each projection πs : Gt → Gs is a group homomorphism. let Gt = G. and d is greater than 1 and f : Z → Zb × Zc × Zd is deﬁned by f (m) = ([m]. Let g : Z → Z45 × Z35 × Z21 be deﬁned by g(m) = ([m]. Deﬁne an addition on Gt = Gt by {at } + {bt } = {at + bt }. 1] and G = R. Find the kernel of f and show that f is not surjective. Suppose G is an additive group. Finally suppose each of b. and GT is the collection of all functions f : T → G with addition deﬁned by (f + g)(t) = f (t) + g(t). [m]). 35. f is a group homomorphism iﬀ each ft is a group homomorphism. [3m]). G is an additive group. Find the kernel of g and determine if g is surjective. Proof The addition on Gt is coordinatewise. and for any t ∈ T .36 Groups Chapter 2 Theorem Suppose T is an index set. [m]). Also note that if T = [0. Exercise Suppose s is an element of T and πs : Gt → Gs is the projection map deﬁned in the Background chapter. 21} is 1. Exercise Suppose s is an element of T and is : Gs → Gt is deﬁned by is (a) = {at } where at = 0 if t = s and as = a. Find necessary and suﬃcient conditions for f to be surjective (see the ﬁrst exercise on page 18). [m]. Find the order of ([4]. Finally. Show πs is a surjective homomorphism and ﬁnd its kernel. Show is is an injective homomorphism ¯ and its image is a normal subgroup of Gt . Find the kernel of h and show that h is surjective. For each t ∈ T .

A section on Boolean rings is included because of their importance in logic and computer science.Chapter 3 Rings Rings are additive abelian groups with a second operation called multiplication. there is an element 1 = 1R ∈ R such that if a ∈ R. If in addition 4) is satisﬁed. R is said to be a ring. and 3) are satisﬁed. Assuming the results of Chapter 2. this chapter ﬂows smoothly. (The distributive law. (The associative property of multiplication. which connects addition and multiplication. 1) 2) If a. These concepts are developed in the Appendix. a · b = b · a. the 37 . c ∈ R.. it would be easy to do. ¯ ¯ ¯ ¯ If a. and R has a second binary ¯ operation (i. This is because ideals are also normal subgroups and ring homomorphisms are also group homomorphisms. b. i. map from R × R to R) which is denoted by multiplication.e. c ∈ R.e.. (a · b) · c = a · (b · c).) R has a multiplicative identity. b ∈ R. We do not show that the polynomial ring F [x] is a unique factorization domain. Also there is no mention of prime or maximal ideals. Consider the following properties. a · 1 = 1 · a = a.) 3) 4) Deﬁnition If 1). (The commutative property for multiplication. b. R = 0. 2). The connection between the two operations is provided by the distributive law. although with the material at hand. Suppose R is an additive abelian group. because these concepts are unnecessary for our development of linear algebra. Examples The basic commutative rings in mathematics are the integers Z. R is said to be a commutative ring.) If a. a · (b + c) = (a · b) + (a · c) and (b + c) · a = (b · a) + (c · a).

Rn is never commutative. na = an. Recall that. . we will deﬁne R[x1 . It will be shown later that Zn . ¯ Theorem 0 can never be a unit. b ∈ R. The product of units is a unit with (a · b)−1 = b−1 · a−1 . ¯ ¯ Units Deﬁnition An element a of a ring R is a unit provided ∃ an element a−1 ∈ R with a · a−1 = a−1 · a = 1. Rn is a ring. This is a basic example of a non-commutative ring. ¯ Of course. . scalar ¯ ¯ ¯ ¯ ¯ ¯ multiplication by n is the same as ring multiplication by n. This scalar multiplication can be written on the right or left. The next two theorems show that ring multiplication behaves as you would wish it to. a · 0 = 0 · a = 0. Since R = 0. Theorem 1) Suppose a. More 2) . Then na = n · a. even if R is commutative. They should be worked as exercises. Under these operations.. n may be 0 even though n = 0. x2 . a polynomical ring in n variables. m ∈ Z. xn ]. that is.e. and the next theorem shows it relates nicely to the ring multiplication. and Rn is the collection of all n×n matrices over R. i. . In the next chapter. For example. Theorem 1) 2) Suppose R is a ring and a. the integers mod n. has a natural multiplication under which it is a commutative ring. b ∈ R and n.) Let n = n1. operations of addition and multiplication of matrices will be deﬁned. 2 = 1 + 1. n ≥ 1. 1 is always a unit. the real numbers R. If n > 1. If a is a unit. . a−1 is also a ¯ −1 ¯ unit with (a )−1 = a. (This follows from the distributive law and the previous theorem. Now suppose R is any ring. and the complex numbers C. since R is an additive abelian group. it follows that 1 = 0.38 Rings Chapter 3 rational numbers Q. ¯ ¯ ¯ ¯ ¯ ¯ (−a) · b = a · (−b) = −(a · b). Also if R is any commutative ring. it has a scalar multiplication over Z (page 20). (na) · (mb) = (nm)(a · b).

an are units. as shown in the next theorem. (−a) is a unit and (−a)−1 = −(a−1 ). Suppose R is a ﬁnite domain and a = 0. we ﬁrst consider the concept of zero divisor. Then ¯ b = c and so a is a unit with a−1 = b = c. Proof b = b · 1 = b · (a · c) = (b · a) · c = 1 · c = c. R is a ﬁeld if it is commutative and its non-zero elements form a group under multiplication.. . a is not a zero divisor. It suﬃces to require a left inverse and a right inverse. if a1 . Finally if a is a unit. Thus a is a unit and so R is a ﬁeld. Deﬁnition A domain (or integral domain) is a commutative ring such that. a is ¯ ¯ a unit. Theorem Suppose a ∈ R and ∃ elements b and c with b · a = a · c = 1. f is surjective. it must have a two-sided inverse. An element a ∈ R is called a zero divisor provided it is non-zero and ∃ a non-zero element b with a · b = 0. a2 . then their product is a unit with (a1 · a2 · · · an )−1 = a−1 · a−1 · · · a−1 . A ﬁnite domain is a ﬁeld. ¯ Then (a · b = a · c) ⇒ b = c. Theorem A ﬁeld is a domain. ¯ ¯ Inverses are unique. Proof A ﬁeld is a domain because a unit cannot be a zero divisor. it cannot be a zero divisor. Theorem Suppose R is a commutative ring and a ∈ (R − 0) is not a zero divisor. if a = 0. ¯ by the pigeonhole principle. In other words. A ﬁeld is a commutative ring such that. if a = 0.Chapter 3 Rings 39 generally. Deﬁnition Suppose R is a commutative ring.. In other words. multiplication by a is an injective map from R to R. The set of all units of R forms a multiplicative group denoted by n−1 1 n R∗ .. Then f : R → R deﬁned by f (b) = a · b is injective and. Note that ¯ if a is a unit. Corollary Domains and Fields In order to deﬁne these two types of rings. In order for a to be a unit. It is surjective iﬀ a is a unit. .

[n − 1] is a unit. Suppose 3) is true. 1).. Now suppose 3) is false. Zn is a ﬁeld. and C are ﬁelds.40 Rings Chapter 3 Exercise Let C be the additive abelian group R2 .. [a]b = [a] · [b] = [ab]. Proof We already know 1) and 2) are equivalent. each of [1]. It leads to a neat little theory. Deﬁne multiplication by (a. the following are equivalent. because each says ∃ an integer b with [a]b = [1]. b) · (c. Show C is a commutative ring which is a ﬁeld. because Zn is ﬁnite. (See the Chinese Remainder Theorem on page 50. [2]. [a] is a unit of the ring Zn . The ring axioms are easily veriﬁed. However. d) = (ac − bd. [a] is a generator of the additive group Zn . Then the following are equivalent. Note that 1 = (1.) We know from page 27 that Zn is an additive abelian group. 0) and if i = (0. R. Theorem 1) 2) 3) Suppose n > 1 and a ∈ Z. Proof We already know from page 27 that 1) and 2) are equivalent. ad + bc). Thus 1) and 3) are equivalent. Corollary 1) 2) 3) If n > 1. Then by the previous theorem. as seen by the theorems below. Recall that if b is an integer. ¯ ¯ Examples Z is a domain. (a. . the basic theory cannot be completed until the product of rings is deﬁned. Then n = ab where 1 < a < n. The Integers Mod n The concept of integers mod n is fundamental in mathematics. Theorem Suppose n > 1. Since [a + kn] · [b + l n] = [ab + n(al + bk + kl n)] = [ab]. the multiplication Proof is well-deﬁned. then i2 = −1. Zn is a domain. n is a prime. Q. n) = 1. This is a well deﬁned binary operation which makes Zn into a commutative ring. Deﬁne a multiplication on Zn by [a] · [b] = [ab].. 1 < b < n.. and thus 2) is true.

R and 0 are ideals of R. Note that Z is a subring of Q. x2 = 1 can have at most two solutions ¯ in R (see the ﬁrst theorem on page 46). Exercise List the units and their inverses for Z7 and Z12 . and (a. then . Subrings Suppose S is a subset of a ring R. and thus occupy a special place in ring theory. Of course. That role is played by ideals. and R is a subring of C. then s may be a unit in R but not in S. every right or left ideal is an ideal. Show that (Z7 )∗ is a cyclic group but (Z12 )∗ is not. and thus [a] is a zero divisor and 1) is false. as well as in group theory. 2-sided) ideals of R. 1 ∈ S . and an ideal is never a subring (unless it is the entire ring). Note that if S is a subring of R and s ∈ S. ¯ If {It }t∈T is a collection of right (left. The     a · b and b · a ∈ I word “ideal ” means “2-sided ideal”. then right (left. ¯ Then clearly S is a ring and has the same multiplicative identity as R. Finally show that if R is a domain. Ideals and Quotient Rings Ideals in ring theory play a role analagous to normal subgroups in group theory. Show that in Z12 the equation x2 = 1 has four ¯ solutions.     Theorem 1) 2) Suppose R is a ring. Subrings do not play a role analogous to subgroups. These are called the improper ideals. Q is a subring of R. 2-sided) ideal of R. Note also that Z and Zn have no proper subrings. The statement that S is a subring of R means that S is a subgroup of the group R.Chapter 3 Rings 41 [a][b] = [0]. if R is commutative.) It is a t∈T .  left  Deﬁnition A subset I of a ring R is a  right ideal provided it is a subgroup    2−sided    a·b ∈ I    b·a∈I of the additive group R and if a ∈ R and b ∈ I. (See page 22. b ∈ S ⇒ a · b ∈ S).

I = R. called a principal ideal. Thus every subgroup of Z is a principal ideal. then the following are equivalent.42 3) Furthermore. and the ring axioms are easily veriﬁed. (On the left. iii) I contains 1. i) I = R. 2-sided) ideal of R. Show that R is a ﬁeld iﬀ R contains no proper ideals. R/I is an additive abelian group. Homomorphisms ¯ ¯ Deﬁnition Suppose R and R are rings. if the collection is monotonic. and I = nZ. The following theorem is just an observation. Multiplication of cosets deﬁned by (a + I) · (b + I) = (ab + I) is well-deﬁned and makes R/I a ring. 4) If a ∈ R. Thus if R is commutative. but it is in some sense the beginning of ring theory. b ∈ R then f (a · b) = f (a) · f (b). Thus multiplication is well deﬁned. ¯ Observation If R = Z. ii) I contains some unit u. because it is of the form nZ. If R is a commutative ring and I ⊂ R is an ideal. Proof (a + I) · (b + I) = a · b + aI + Ib + II ⊂ a · b + I. aR is an ideal. The multiplicative identity is (1 + I). ¯ 5) Exercise Suppose R is a commutative ring. Since I is a normal subgroup of the additive group R. Theorem Suppose R is a ring and I ⊂ R is an ideal. then t∈T Rings Chapter 3 It is a right (left. multiplication . I = aR is a right ideal. A function f : R → R is a ring homomorphism provided 1) 2) 3) f is a group homomorphism f (1R ) = 1R and ¯ ¯ ¯ if a. n > 1. the ring structure on Zn = Z/nZ is the same as the one previously deﬁned.

¯ Here is a list of the basic properties of ring homomorphisms.) 43 The kernel of f is the kernel of f considered as a group homomorphism. The kernel of a ring homomorphism is an ideal of the domain. The identity map IR : R → R is a ring homomorphism. while on the right multiplication is in R. f is also called a ring automorphism. Theorem 1) 2) ¯ Suppose each of R and R is a ring. Suppose I is an ideal of R. ¯ Suppose f : R → R is a homomorphism and I is an ideal of R. From now on the word “homomorphism” means “ring homomorphism”. π(a) = (a + I). if f : R → R is a surjective ¯ ring homomorphism with kernel I. ¯ ¯ ¯ The composition of ring homomorphisms is a ring homomorphism. Much of this work has already been done by the theorem in group theory on page 28. ¯ ¯ ¯ If I ⊂ ker(f ).Chapter 3 Rings ¯ is in R. Then π is a surjective ring ¯ homomorphism with kernel I. then f : R/I → R deﬁned by f (a + I) = f (a) 3) 4) 5) 6) 7) 8) . namely ker(f ) = f −1 (0). then f −1 (I) is an ideal of R. ¯ ¯ In fact. and π : R → R/I is the natural projection. I = R. The image of a ring homomorphism is a subring of the range. ¯ The zero map from R to R is not a ring homomorphism (because it does not send 1R to 1R ). ¯ If f : R → R is a bijection which is a ring homomorphism. −1 ¯ → R is a ring homomorphism. Furthermore. I = R. if f : R → R is a homomorphism and I ⊂ R is an ideal. Such an f is called then f : R ¯ a ring isomorphism. In the case R = R. then R/I ≈ R (see below).

Notice that much of the work has been done in the previous exercise. Suppose S is a non-void set and α : S → T is a function. domain(f )/ker(f ) ≈ image(f ). 1] and R = R. i. 9) Given any ring homomorphism f. It is only necessary to show that A is a subring of the ring R[0.. Exercise Find a ring R with a proper ideal I and an element b such that b is not a unit in R but (b + I) is a unit in R/I. Deﬁne addition and multiplication on RT point-wise. Exercise Now consider the case T = [0. Proof We know all this on the group level. A ={f : [0. .1] . Exercise Show that if u is a unit in a ring R. and it is only necessary ¯ to check that f is a ring homomorphism. Exercise Suppose T is a non-void set. which is obvious. The image of f is the image of f . f is injective. 1] → R : f has an inﬁnite number of derivatives}.e. This means if f and g are functions from T to R. R π c && f & E b & & ¯ R & & & ¯ f R/I Thus deﬁning a homomorphism on a quotient ring is the same as deﬁning a homomorphism on the numerator which sends the ¯ denominator to zero. show that f : R → R deﬁned by f (a) = u−1 · a · u is a ring homomorphism which is an isomorphism. and RT is the collection of all functions f : T → R.44 Rings Chapter 3 is a well-deﬁned homomorphism making the following diagram commute. If f : T → R is a function. Show that under these operations R T is a ring. and so R/I ≈ image (f ). deﬁne a function α∗ (f ) : S → R by α∗ (f ) = f ◦ α. R is a ring.1] be the collection of all C ∞ functions. Show A is a ring. and ¯ ¯ the kernel of f is ker(f )/I. then conjugation by u is an automorphism on R. That is. Let A ⊂ R[0. Show α∗ : RT → RS is a ring homomorphism. then (f + g)(t) = f (t) + g(t) and (f · g)(t) = f (t)g(t). Thus if I = ker(f ).

.. . Suppose n ≥ m and the result holds for any polynomial of degree less than .) + (b0 .). Note that on the right. R[x ] is a commutative ring.. We outline the proof of existence and leave uniqueness as an exercise. ∃! h.. b1 . the top coeﬃcient of f will always be a unit.. and it is easy to see that the collection of polynomial functions forms a commutative ring. a1 . ¯ To be more formal. . f (x ) = a0 + a1 x + · · +an x n . Under the obvious addition and multiplication. a0 b1 + a1 b0 . Deﬁnition Suppose R is a commutative ring and x is a “variable” or “symbol”.) such that each ai ∈ R and only a ﬁnite number are non-zero. Suppose f = a0 + a1 x + · · +am xm where m ≥ 1 and am is a unit in R. If the top term an = 1. Proof Suppose f and g are non-zero polynomials. as is often done for convenience.) · (b0 .. ¯ Proof This theorem states the existence and uniqueness of polynomials h and r.. ¯ and is denoted by n = deg(f ). a1 . Then ai bj xi+j is the ﬁrst non-zero term of f g. think of a polynomial a0 + a1 x + · · · as an inﬁnite sequence (a0 . the idea is to divide f into g ¯ until the remainder has degree less than m. set h = 0 and r = g. a1 + b1 .) = (a0 b0 ...) and (a0 . a1 . a0 b2 + a1 b1 + a2 b0 . . Another way to prove this theorem is to look at the bottom ¯ terms instead of the top terms. we consider real functions f which are polynomials. The polynomial ring R[x ] is the collection of all polynomials f = a0 + a1 x + · · +an x n where ai ∈ R. Then (a0 . We can do the same thing formally in a purely algebraic setting.. the ring multiplication a · b is written simply as ab. For the general case. f ∈ R[x ] has degree ≥ 1 and its top coeﬃcient is a unit in R.) Then for any g ∈ R[x ]. The proof is by induction on the degree of g. Theorem If R is a domain. For any g with deg(g) < m. r ∈ R[x ] such that g = f h + r with r = 0 or deg(r) < deg(f ). Theorem (The Division Algorithm) Suppose R is a commutative ring. . (If R is a ﬁeld. . The degree of a non-zero polynomial f is the largest integer n such that an = 0. R[x ] is also a domain. .) = (a0 + b0 ... b1 . then f is said to be monic.. The sum and product of polynomials are again polynomials..Chapter 3 Rings Polynomial Rings 45 In calculus. . Let ai xi and bj xj be the ﬁrst non-zero terms of f and g. Then deg(f )+deg(g) = deg(f g) and thus f g is not 0.

each coset of I can be written uniquely in the form (c0 + c1 x + · · +cn−1 x n−1 + I). (If h has degree 0. Also show that if g(x) = x2 + bx + c. n > 0. This is called The Fundamental Theorem of Algebra.. Theorem Suppose F is a ﬁeld. Let c1 . The result follows from the equation f (h1 + bxt ) + r = g.e. c2 . i. and g(x) = a0 + a1 x + · · · + an xn is a Theorem polynomial of degree n with at least one root in R. given any ideal I. Deﬁnition A domain T is a principal ideal domain (PID) if. Note that Z is a PID and any ﬁeld is PID.) ¯ Proof Uniqueness is easy so let’s prove existence. Note that f = x − c divides g iﬀ c is ¯ a root of g. C is an algebraically closed ﬁeld. Furthermore. This is a good exercise in the use of the division algorithm. . Note this is Proof.e. ck be the distinct roots of g in the ring R. we say “all the roots of g are 0”. ∃ h1 and r with f h1 + r = (g − f bxt ) and deg(r) < m. Note If g is any non-constant polynomial in C[x].46 Rings Chapter 3 n. and n is the smallest positive integer such that I contains a polynomial of degree n. x − c divides g with remainder g(c). I is a proper ideal of F [x ]. g(c) = 0. Show that if g has odd degree then it has a real root.. Exercise Suppose g is a non-constant polynomial in R[x]. Now suppose g is a polynomial of degree n and c1 is a root of g. Thus F [x ] is a PID. Now ∃ a monomial bxt with t = n − m and deg(g − f bxt ) < n.e. Suppose n > 1 and the theorem is true for any polynomial of degree less than n. n2 . Then ∃ a unique sequence of positive integers n1 . i. ∃ t ∈ T such that I = tT. similar to showing that a subgroup of Z is generated by one element (see page 15). Since h1 has degree less than n. if h = an . More generally... . and in that case both roots belong to R. then we say “all the roots of g belong to R”. By induction. . Then I contains a unique polynomial of the form f = a0 + a1 x + · · +an−1 x n−1 + x n and it has the property that I = f F [x ]. If g = an xn . Suppose g is a polynomial of degree n. nk and a unique polynomial h with no root in R so that g(x) = (x − c1 )n1 · · · (x − ck )nk h(x).. then it has a real root iﬀ b2 ≥ 4c.. and it is assumed without proof for this textbook. Then g has at most n roots. Note If r = 0 we say that f divides g. Then ∃ a polynomial h1 with g(x) = (x − c1 )h1 .. The theorem is clearly true for n = 1. ¯ Suppose R is a domain. the result follows by induction. all the roots of g belong to C. i.

i. Suppose R is a subring of a commutative ring C and c ∈ C. This is a good way to look at the complex numbers.. We do not develop the theory of F [x ] here. Write out the multiplication table for this ring and show that it is a ﬁeld. The next deﬁnition and theorem are included merely for reference. The image of h is the smallest subring of C containing R and c. Thus if F is a ﬁeld. then h is an associate of f . the units of R[x ] are just the units of R. Exercise Show that.e. One diﬀerence is that the units of F [x ] are non-zero constants. This map h is called an evaluation map. Also multiplying two polynomials in R[x ] and evaluating is the same as evaluating and then multiplying in C. adjoin x to R and set x2 = −1. and this h is surjective. The statement that f is irreducible means that if h is a non-constant polynomial which divides f . i. Then ∃! homomorphism h : R[x ] → C with h(x ) = c and h(r) = r for all r ∈ R. The Division Algorithm corresponds to the Euclidean Algorithm. Show ker(h) = (x2 + 1)R[x ] and thus R[x ]/(x 2 + 1) ≈ C. The statement that g is an associate of f means ∃ a unit u ∈ F [x] such that g = uf .Chapter 3 Rings 47 Theorem. In this chapter we do not prove F [x] is a unique factorization domain. Exercise Z2 [x ]/(x 2 + x + 1) has 4 elements. b ∈ R}. It is deﬁned by h(a0 + a1 x + · · +an x n ) = a0 + a1 c + · · +an cn . h sends f (x) to f (c). However. to obtain C. Deﬁnition Suppose F is a ﬁeld and f ∈ F [x] has degree ≥ 1. there exists a homomorphism h : R[x] → C which sends x to i. The theorem says that adding two polynomials in R[x ] and evaluating is the same as evaluating and then adding in C. The degree function corresponds to the absolute value function. Exercise Let C = {a + bi : a. Irreducible polynomials correspond to prime integers. while the units of Z .e. In street language the theorem says you are free to send x wherever you wish and extend to a ring homomorphism on R[x]. nor do we even deﬁne unique factorization domain.. the units of F [x ] are the non-zero constants. Show that [1] + [2]x is a unit in Z4 [x ]. Since R is a subring of C. the development is easy because it corresponds to the development of Z in Chapter 1. if R is a domain. and should not be studied at this stage.

y] is a commutative ring and (R[x ])[y] ≈ R[x . x2 . vn are non-negative v v v integers. xn ] is a domain and its units are just the .. y] to be any ﬁnite sum of monomials. Using this and induction on n. x2 .. and this factorization is unique up to order and associates. x2 . Theorem R[x .. the ideal I = xF [x.. If R is a domain. then ax n y m = ay m x n is called a monomial... it is easy to prove the following theorem.. the concept of a polynomial ring in n variables works ﬁne without a hitch. Deﬁne an element of R[x . xn ] to be any ﬁnite sum of monomials. y] : f (0.. y] = {f ∈ F [x. any polynomial in x and y with coeﬃcients in R may be written as a polynomial in y with coeﬃcients in R[x ]. However F [x . x2 .. Here is the basic theorem. y] is not a PID. This gives a commutative ring and there is canonical isomorphism R[x1 . 1) 2) 3) F [x ]/(f ) is a domain. Thus the associates of f are all cf with c = 0 while the associates of an ¯ integer n are just ±n. For example. or as a polynomial in x with coeﬃcients in R[y]. y] factors as the product of irreducibles. Deﬁne an element of R[x1 ..48 Rings Chapter 3 are just ±1. . 0) = 0} is not principal. m ≥ 0. Side Comment It is true that if F is a ﬁeld. In other words. xn ] ≈ (R[x1 .. then ax1 1 x2 2 · · · xn n is called a monomial. f is irreducible. R[x1 . xn−1 ])[xn ].) Theorem Suppose F is a ﬁeld and f ∈ F [x ] has degree ≥ 1. . . (This theory is developed in full in the Appendix under the topic of Euclidean domains.. Then f factors as the product of irreducibles. If a ∈ R and v1 . Theorem units of R. each f ∈ F [x .. . v2 . F [x ]/(f ) is a ﬁeld. Also the following are equivalent. If a ∈ R and n. . y] ≈ (R[y])[x ].. y] + yF [x.. Order does not matter here. Deﬁnition Now suppose x and y are “variables”. ¯ ¯ ¯ If R is a commutative ring and n ≥ 2.

Show f is a ring ¯ ¯ homomorphism whose kernel is the ideal (y) = yR[x. Exercise Suppose R and S are commutative rings. since multiplication is deﬁned ¯ ¯ ¯ ¯ coordinatewise. an element e of T is called an idempotent provided e2 = e. Suppose T is a commutative ring and e ∈ T is an idempotent with 0 = e = 1. Proof We already know f is a group homomorphism iﬀ each ft is a group homomorphism (see page 36). Theorem Suppose T is an index set and for each t ∈ T . 1S ). Under the natural bijection from {functions f : R → Rt } to {sequences of functions {ft }t∈T where ft : R → Rt }. Rt is a ring. Show T = R × S is not a domain. just as does the product of groups. and ¯ f (1R ) = {1t } iﬀ ft (1R ) = 1t for each t ∈ T. On the additive abelian group Rt = Rt . and 0 × S = (1 − e)T . Finally. That is. ¯ ¯ ¯ ¯ Suppose I ⊂ R and J ⊂ S are ideals. f is a ring homomorphism iﬀ each ft is a ring homomorphism. Note that {1t } is the multiplicative identity of Rt . Let e = (1. deﬁne multiplication by {rt } · {st } = {rt · st }. y]. y)) = p(x. t∈T . Exercise If T is any ring. if you adjoin y to R[x] and then factor it out. and f : T → R × S deﬁned by f (t) = (et. Show R × 0 is an ideal and (R × S/R × 0) ≈ S. (1 − e)t) is a ring isomorphism. Show I ×J is an ideal of R×S and every ideal of R × S is of this form. (1 − e)2 = (1 − e). Then Rt is a ring and each projection πs : Rt → Rs is a ring homomorphism. This means f (p(x. Let R = eT and S = (1 − e)T . This shows that a commutative ring T splits as the product of two rings iﬀ it contains a non-trivial idempotent. 0) ∈ R × S and show e2 = e. R × 0 = eT . you get R[x] back. y] → R[x] is the evaluation map which sends y to 0. Exercise Suppose R and S are rings. 0). Product of Rings The product of rings works ﬁne. Show each of the ideals R and S is a ring with identity. Use the fact that “the domain mod the kernel is isomorphic to the image” to show R[x. y]/(y) is isomorphic to R[x]. The elements 0 and 1 are idempotents called the trivial idempotents.Chapter 3 Rings 49 Exercise Suppose R is a commutative ring and f : R[x. Note that R × 0 is not a subring of R × S because it does not contain (1R . Suppose R is a ring. f is a ring homomorphism iﬀ each ft is a ring homomorphism.

the ring of integers is a “cornerstone”. Thus Zn and Zn1 × · · ×Znt are isomorphic as rings.. ft ) : Z → Zn1 × · · ×Znt is surjective. . The element f (1)m = ([1]. Proof We wish to show that the order of f (1) is n. Use this and the Chinese Remainder Theorem to show that if b is a positive integer. each ni > 1. Thus the subgroup of R generated by 1 is a subring ¯ ¯ ¯ of R isomorphic to Z or isomorphic to Zn for some positive integer n. nt are integers. and thus f (1) is a group generator. (See page 23 for the deﬁnition of order. The next theorem is a classical generalization of this. and thus f is surjective. but it shows that in ring theory. If m and n are relatively prime. Furthermore. if R is a domain. and thus Zmn and Zm × Zn are isomorphic as groups and as rings. [m]) is zero iﬀ m is a multiple of each of n1 ... .. then [a] = [ap ] in Zp (Fermat’s Little Theorem). nj ) = 1 for all i = j. Theorem If R is a ring. the kernel of f is nZ.. . this map is surjective with kernel mnZ. Characteristic The following theorem is just an observation. and (ni . there is one and only one ring homomorphism f : Z → R.) Exercise Show that if a is an integer and p is a prime.) Theorem Suppose n1 . where n = n1 n2 · · nt . the order of f (1) is n. . (See exercise three on page 35.. (Note that the bracket symbol is used ambiguously. (See the fourth exercise on page 36 for the case t = 3. all the non-zero elements of R have the same order..50 The Chinese Remainder Theorem Rings Chapter 3 The natural map from Z to Zm × Zn is a group homomorphism and also a ring homomorphism. nt . Let fi : Z → Zni be deﬁned by fi (a) = [a]. The non-negative integer n with ker(f ) = nZ is called the ¯ ¯ characteristic of R. Thus f is injective iﬀ R has characteristic 0 iﬀ 1 has inﬁnite ¯ order... If f is not injective. and thus also as groups. [1])m = ([m].. It is given by f (m) = m1 = m.) .) Then the ring homomorphism f = (f1 . Since their least common multiple is n.. ¯ It is an interesting fact that. . Deﬁnition Suppose R is a ring and f : Z → R is the natural ring homomorphism f (m) = m1 = m. the characteristic of R is the order of 1. it has the same last digit as b5 .

then every element of R/I is idempotent and thus R/I is a Boolean ring. If a ∈ R. Now suppose R has characteristic n. However it ﬁts easily here. 3) If R is a domain. then n is a prime and each non-zero a ∈ R has order n. ¯ 2 2 2 2 Proof (a + a) = (a + a) = a + 2a + a = 4a. and so a = −a. ¯ ¯ ¯ ¯ Thus o(a) = ∞.. Then ma = m · a cannot be 0 because m. (In the language of Chapter 6. a is a non-zero element of R. If R has characteristic 0. each element of R is an idempotent. and is included for reference. ¯ ¯ ¯ Exercise Show that if F is a ﬁeld of characteristic 0. . Thus 2a = 0. I is a prime ideal iﬀ I is a maximal ideal iﬀ R/I ≈ Z2 ). 2a = a + a = 0. ¯ ¯ ¯ ¯ 4) The image of a Boolean ring is a Boolean ring. a2 = a. If a is a non-zero element of R.Chapter 3 Rings 51 Theorem Suppose R is a domain. if I is an ideal of R with I = R. It follows from 3) that R/I is a domain iﬀ R/I is a ﬁeld iﬀ R/I ≈ Z2 . That is. Then a · (1 − a) = 0 and so a = 1. If R has ﬁnite characteristic n. Thus a · b = b · a. and thus Zn is a domain and n is a prime. and m is a positive integer. then each non-zero a ∈ R has inﬁnite order. Deﬁnition A ring R is a Boolean ring if for each a ∈ R. Theorem 1) Suppose R is a Boolean ring. ¯ R is commutative. R has characteristic 2. Boolean Rings This section is not used elsewhere in this book. That is. F contains Q as a subring. 2) Proof (a + b) = (a + b)2 = a2 + (a · b) + (b · a) + b2 = a + (a · b) − (b · a) + b. na = n · a = 0 · a = 0 and thus o(a)|n and thus o(a) = n. a = 0 and R is a domain. show that the injective homomorphism f : Z → F extends to an injective ¯ homomorphism f : Q → F . i.e. Proof Suppose R has characteristic 0. Proof Suppose a = 0. Then R contains Zn as a subring. R ≈ Z2 .

it contains some element a. It is a classical theorem that ∃ a Boolean algebra of sets whose Boolean ring is isomorphic to R. b ∈ R ⇒ (a ∩ b) ∈ R. Show each fi is a homomorphism and thus f = (f1 . . Under this addition. Since R is non-void. let a = (X −a) be a complement of a in X. Then a ∪ b = (a ∩ b ) belongs to R and so 3) is true. These operations cup and cap are associative. Theorem Suppose R is a Boolean algebra of sets. R is not a group under cup or cap. If a is a subset of X. fn ) : R → Z2 × Z2 × · · ×Z2 is a ring homomorphism. In this case. The advantage of the ring viewpoint is that you can draw from the rich theory of commutative rings. and thus also to the Boolean ring of subsets above. Under this multiplication R becomes a Boolean ring with 1 = X. Anyway.. Now deﬁne a ∨ b = a ∪ b and a ∧ b = a ∩ b. Then ∅ = a ∩ a and X = a ∪ a belong to R. So let’s just suppose R is a Boolean algebra of sets which is a Boolean ring with addition and multiplication deﬁned as above. Theorem If 1) and 2) are satisﬁed... n} and let R be the Boolean ring of all subsets of X. it is a classical fact that. R is called a Boolean algebra. (See exercises 1) and 4) on page 12. have identity elements. and each distributes over the other. 2. Deﬁne fi : R → Z2 by fi (a) = [1] iﬀ i ∈ a. Show f is an isomorphism. a. Now suppose R is a non-void collection of subsets of X.. then 3) and 4) are satisﬁed. ¯ Exercise Let X = {1.) Exercise Use the last exercise on page 49 to show that any ﬁnite Boolean ring is isomorphic to Z2 × Z2 × · · ×Z2 . 1) 2) 3) 4) a ∈ R ⇒ a ∈ R. The advantage of the algebra is that it is symmetric in cup and cap. b ∈ R ⇒ (a ∪ b) ∈ R. Note Suppose R is a Boolean ring.. Deﬁne a multiplication on R by a · b = a ∩ b. and a. you have a Boolean algebra (ring). Deﬁne an addition on R by a + b = (a ∪ b) − (a ∩ b). . Note that o(R) = 2n .. commutative. if you have a Boolean ring (algebra). With these two operations (along with complement). a. b ∈ R. and so 4) is true. Proof Suppose 1) and 2) are true.52 Rings Chapter 3 Suppose X is a non-void set. . ∅ ∈ R and X ∈ R. Consider the following properties which the collection R may possess. R is called a Boolean algebra of sets. R is an abelian group with 0 = ∅ and ¯ a = −a.

A matrix is said to be square if it has the same number of rows as columns. Deﬁnition Suppose R is a ring and m and n are positive integers.  A = (ai. we use the “sum” notation.Chapter 4 Matrices and Matrix Rings We ﬁrst consider matrices in full generality.1 . .1 . If the elements of Rn are written as row vectors. Rn is deﬁned to be the additive abelian group R × R × · · · × R. Square matrices are so important that they have a special notation. such as invertible matrices. The highlight of the chapter is the theorem that a square matrix is a unit in the matrix ring iﬀ its determinant is a unit in the ring. Let Rm.n be the collection of all m × n matrices a1. Rn is identiﬁed with R1. a1. .e. are all classical... .n . i. Rn = R ⊕ R ⊕ · · · ⊕ R.1 .j . and characteristic polynomial. The topics. Our convention is to write elements of R n as column vectors.  where each entry a ∈ R.  am. trace. and determinant. Rn = Rn. over an arbitrary ring R. . i. This chapter concludes with the theorem that similar matrices have the same determinant. i. elementary matrices.j ) =  . However.   53 . This will be used in the next chapter to show that an endomorphism on a ﬁnitely generated vector space has a well-deﬁned determinant.  . am.e.n A matrix may be viewed as m n-dimensional row vectors or as n m-dimensional column vectors.n  . it will be assumed that R is commutative. . To emphasize that Rn does not have a ring structure. to identify Rn with Rn. and characteristic polynomial.n . transpose. after the ﬁrst few pages. systems of equations. . trace.

Note in particular that scalar multiplication is deﬁned on Rn .n · bn.. The matrix AB will have the same number of rows as A and the same number of columns as B. Rm.p → Rm.t . In the language of the next chapter.n × R → Rm. VA. Exercise and W = Consider real matrices A = 1 2 0 1 a b c d . Scalar multiplication An element of R is called a scalar.n all of whose terms are zero. AV. j term of the sum is the sum of the i.e.n × Rn.n and c. Theorem Rm. the dot product of row s of A with column t of B.n . there is no distinction between right and left scalar multiplication.n . addition is a binary operation Rm. Find the matrices AU.p .j ) is deﬁned to be the matrix whose (s.j ). Then (A + B)c = Ac + Bc A(c + d) = Ac + Ad A(cd) = (Ac)d A1 = A and This theorem is entirely transparent.j ) + (bi. The addition is deﬁned by (ai.V = 0 1 1 0 .. Theorem Suppose A.j )c = (ai.j ) = (−ai. d ∈ R.n is an additive abelian group. i.j + bi. j terms. multiplication is a function Rm. i.n ≈ Rmn . Of course.n is a right module over the ring R. Multiplication of Matrices The matrix product AB is deﬁned iﬀ the number of columns of A is equal to the number of rows of B. it merely states that Rm. and W .t + · · · + as. Its “zero” is the matrix 0 = 0m. A matrix may be “multiplied” on the right or left by a scalar. The product (ai.. t) term is as. . The following theorem is just an observation.e.e.n × Rm.U = 2 0 0 1 .1 · b1. Furthermore. UA.j ).n → Rm. i.j · c). as additive groups.54 Matrices Chapter 4 Addition of matrices To “add” two matrices.j )(bi. they must have the same number of rows and the same number of columns. B ∈ Rm. It is a function Rm. i. Also −(ai. AW . A.j ) = (ai. if R is commutative. Right scalar multiplication is deﬁned by (ai. the i..e.

i ci.j ) where a1. Note that ABC ∈ Rm. t) term of (AB)C is as.j bj. They form a group under multiplication called the general linear group and denoted by GLn (R) = (Rn )∗ . Find aR2 and R2 a.p Im A = A = AIn (The distributive laws) (A + B)C = AC + BC C(A + B) = CA + CB and whenever the operations are deﬁned. Then (AB)C = A(BC).j ) = BC.n . Show that the only ideal of R2 containing a is R2 itself.t = as. B. Then multiplication in Rn+m is given by A C B D E G F H = AE + BG CE + DG AF + BH CF + DH .n . Exercise Recall that if A is a ring and a ∈ A. t) terms are equal. Proof We must show that the (s. Theorem For each ring R and integer n ≥ 1. Proof This elegant little theorem is immediate from the theorems above. C. H ∈ Rm .Chapter 4 Matrices 55 Deﬁnition The identity matrix In ∈ Rn is the square matrix whose diagonal terms are 1 and whose oﬀ-diagonal terms are 0.q .m . 0p.t = j as.j bj.j j i bj.j as.i ci. The units of Rn are called invertible or non-singular matrices. and C ∈ Rp. E ∈ Rn . and D.n A0n. Theorem 1) 2) Theorem Suppose A ∈ Rm.i ci. Rn is a ring. Multiplication by blocks Suppose A. .i ci. Let (xi. The proof involves writing it out and changing the order of summation. F ∈ Rn. t) term of A(BC). xs.p = 0m. G ∈ Rm. then aA is right ideal of A. Let A = R2 and a = (ai.t = Then the (s. B ∈ Rn.p .q .1 = 1 and the other entries are 0.j yj.m A = 0p.j ) = AB and (yi.t which is the (s.n . Theorem (The associative law for matrix multiplication) Suppose A ∈ Rm.t = i i j i.

Then I = I t = (AA−1 )t = (A−1 )t At . If A ∈ Rm. The geometry of this theorem will become transparent later in Chapter 5 when the matrix A deﬁnes an R-module endomorphism on Rn (see page 93). So row i (column i) of A becomes column i (row i) of At . Proof of 5) Exercise Characterize those invertible matrices A ∈ R2 which have A−1 = At .j = 0 for all i ≥ j (all j ≥ i).56 Transpose Notation tative ring. . In this case.m . then A is upper (lower) triangular provided ai.n . Matrices Chapter 4 For the remainder of this chapter on matrices.j = 0 for all i > j (all j > i). In this case (A−1 )t = (At )−1 .e. Triangular Matrices If A ∈ Rn . i) term of A.m is the matrix whose (i. At ∈ Rn.. Theorem 1) 2) 3) 4) 5) (At )t = A (A + B)t = At + B t If c ∈ R. Thus if T = Rn and B is a nilpotent matrix. an element t ∈ T is said to be nilpotent provided ∃n n such that t = 0. Proof The way to understand this is just multiply it out for n = 2 and n = 3.j = 0 for all i = j. (1 − t) is a unit with inverse 1 + t + t2 + · · · + tn−1 . Suppose A is invertible. Show that they form a subgroup of GL2 (R). then At is an n-dimensional column vector. Deﬁnition If T is any ring. (Ac)t = At c (AB)t = B t At If A ∈ Rn . j) term is the (j. If A is a square matrix. At is also square. A is diagonal if it is upper and lower triangular. Transpose is a function from Rm.n to Rn. i. A is strictly upper (lower) triangular provided ai. Note that if A is upper (lower) triangular. then An = 0. Theorem If A ∈ Rn is strictly upper (or lower) triangular. If A is an n-dimensional row vector. Rn is non-commutative. ai. then At is lower (upper) triangular. I − B is invertible. suppose R is a commuOf course. then A is invertible iﬀ At is invertible. for n > 1.

Multiplying by a scalar is the same as multiplying by a scalar matrix. a matrix of the form cIn . i.n .Chapter 4 Matrices 1 2 −3  Find the inverse of  0 1 4 . There are 3 types of elementary row and column operations on the matrix A.. For n > 1. Multiply column i by some unit a ∈ R. Type 1 Multiply row i by some unit a ∈ R. Scalar matrices A scalar matrix is a diagonal matrix for which all the diagonal terms are equal.  0 0 1    57 Exercise Let R = Z. Show AC is obtained from C by multiplying row i of C by ai . B ∈ Rm. Show A is a unit in Rn iﬀ each ai is a unit in R. Exercise Suppose A ∈ Rn and for each B ∈ Rn . Elementary Operations and Elementary Matrices Suppose R is a commutative ring and A is a matrix over R. i. this shows how non-commutative Rn is. AB = BA.p .                is a diagonal matrix. Type 2 Type 3 .e. and thus we may consider R to be a subring of Rn . and thus scalar matrices commute with everything. Add a times row j to row i where i = j and a is any element of R. Show A is a scalar matrix. The map R → Rn which sends c to cIn is an injective ring homomorphism. if B ∈ Rn . (cIn )B = cB = Bc = B(cIn ). a1 a2 0 · 0 · Exercise Suppose A = an and C ∈ Rn. Interchange row i and row j.. A need not be square. Show that BA is obtained from B by multiplying column i of B by ai . Recall we are assuming R is a commutative ring. Interchange column i and column j.e. Add a times column i to column j where i = j and a is any element of R.

In type 2. then B is invertible and B −1 is an elementary matrix of the same type.2. They are obtained by performing row or column operations on the identity matrix. (See the exercise on page 54. or 3. It need not be square. and multiply on the left (right). all the oﬀ-diagonal elements are zero. there are two non-zero oﬀ-diagonal elements. There are three types. In type 1.) . BA = row operation on A and AB = column operation on A. perform the operation on an identity matrix to obtain an elementary matrix B. Theorem Suppose A is a matrix.58 Matrices Chapter 4 Elementary Matrices Elementary matrices are square and invertible. In type 3. there is at most one non-zero oﬀ-diagonal element. Type 2 B=                               Type 3 B=          where i = j and ai. and it may be above or below the diagonal. The following theorem is handy when working with matrices.j 1 0 Type 1 B=                     where a is a unit in R. Exercise Show that if B is an elementary matrix of type 1.j is any element of R. To perform an elementary row (column) operation on A.  1 1 a 1 0 1 1 1 0 1 1 1 0 1 1 1 1 1 0 1 1 ai. That is.

59 where d1.1 x1 + · · · + a1.j ) A matrix T is said to be in row echelon form if. perform row and column operations on A to reach the desired form.n .j ) ∈ Rm. . . can be written as one . namely as (ai. Suppose A = (ai.1 .n xn = cm     c1 x1  ·   ·       or AX = C. Is it possible to write them as products matrices over Z? 4) For 1). the ﬁrst non-zero term of row (i + 1) is to the right of the ﬁrst non-zero term of row i. = matrix equation in one unknown. 3 0 of elementary of elementary Let A = 11 3 11 and D = .1 x1 + · · · + am. Write A and D as products 1 4 4 matrices over Q.j )   ·   ·  cm xn a1.) 2) 3) Suppose A ∈ Fn is invertible. . the number of non-zero rows is the rank of T . .n xn = . for each 1 ≤ i < m. Notice that if T is in row-echelon form. For part 3).n and C =  c1 · · cm      ∈ Rm = Rm. . The integer t is called the rank of A.Chapter 4 Exercise 1) Matrices Suppose F is a ﬁeld and A ∈ Fm. The system . am. Show ∃ invertible matrices B ∈ Fm and C ∈ Fn such that BAC = (di. use only row operations. Show A is the product of elementary matrices. Show ∃ an invertible matrix B ∈ Fm such that BA is in row echelon form. Systems of Equations     c1 .t = 1 and all other entries are 0. This shows the matrices B and C may be selected as products of elementary matrices. of m equations in n unknowns. Part 2) also follows from this procedure.1 = · · · = dt. . (See page 89 of Chapter 5.

the solution set f −1 (C) is the coset D + ker(f ) in Rn . which is given in Chapter 6.. For n = 2.) Suppose B ∈ Rm is invertible. If σ is a permutation on {1. Then AX = C and (BA)X = BC have the same set of solutions. If D ∈ Rn is one solution. (See part 7 of the theorem on homomorphisms in Chapter 2.j ) ∈ Rn . 2. Its solution set is ker(f ). this agrees with the deﬁnition above. 3) 4) The geometry of systems of equations over a ﬁeld will not become really transparent until the development of linear algebra in Chapter 5. The determinant is deﬁned by | A |= sign(σ) a1. Determinants The concept of determinant is one of the most amazing in all of mathematics. let sign(σ) = 1 if σ is an even permutation. For each n ≥ 1 and each commutative ring R. For n = 1.σ(1) · a2. The proper development of this concept requires a study of multilinear forms. and sign(σ) = −1 if σ is an odd permutation. In this section we simply present the basic properties. If m = n and A ∈ Rm is invertible. In the language of the next chapter. page 28. Thus we may perform any row operation on both sides of the equation and not change the solution set. | (a) | = a. The next theorem summarizes what we already know about solutions of linear equations in this setting.σ(2) · · · an.. = ad − bc. c d Deﬁnition Let A = (ai.60 Matrices Chapter 4 Deﬁne f : Rn → Rm by f (D) = AD.) . Check that for all σ n = 2. then AX = C has the unique solution X = A−1 C.. Then f is a group homomorphism and also f (Dc) = f (D)c for any c ∈ R. this says that f is an R-module homomorphism. . Theorem 1) 2) AX = 0 is called the homogeneous equation. n}.σ(n) . AX = C has a solution iﬀ C ∈ image(f ). (Note that here we are writing the permutation functions as σ(i) and not as (i)σ. determinant is a function from Rn a b to R.

Ar+1 . .n . We wish to show that τ = σ −1 and thus sign(σ) = sign(τ ).2 · · · aτ (n). . An )| + c|(A1 . .1 = Rn . ¯ Proof For simplicity. n} such that a1. .σ(1) · a2. It follows that the determinant of a matrix is equal to the determinant of its transpose. In the second expression. . Let γ be the transposition which interchanges one and two. Summary Determinant is a function d : Rn → R. An ) where each Ai ∈ Rn. and so τ (5) = 2.5 . Here we will use column vectors. . . it will appear as aτ (5). Then the ﬁrst expression will contain the factor a2..1 · aτ (2).σ(2) · · · an. . .2 · · · aτ (n). . .1 . ¯ Theorem Suppose 1 ≤ r ≤ n.σ(1) · · · an. Anyway. Theorem |A| = sign(σ)a1. . This means that there is a permutation τ on {1. c ∈ R. An )| = a|(A1 .1 · aτ (2). . these pairs cancel in the summation.e.n . . Ar−1 . all τ . .σ(1) · a2. Therefore |A| = 0.5 . Cr . i. 2. .n . Ar+1 . . You may view an n × n matrix A as a sequence of n column vectors or as a sequence of n row vectors. .σ(n) = sign(τ )aτ (1). . A1 = A2 . If two columns of A are equal. and since sign(τ )=−sign(τ γ). assume the ﬁrst two columns are equal.1 · aτ (2). we may rearrange the factors so that the ﬁrst comes from the ﬁrst column. suppose σ(2) = 5. In the language used in the Appendix. To reduce the abstraction. sign(τ )aτ (1).Chapter 4 Matrices 61 For each σ. a1. .1 · · · aτ (n). .n and this summation has n! terms and Now |A| = Theorem n! is an even number. . .2 · · · aτ (n).2 · · · aτ γ(n). . aAr + cCr . aτ (1). τ is the inverse of σ and thus there are two ways to deﬁne determinant. This means we write the matrix A as A = (A1 . . Then for any τ . the two previous theorems say that d is an alternating multilinear form. A2 . and a. etc. This pairs up the n! terms of the summation. all σ all τ Corollary |A| = |At |. The next two theorems show that alternating implies skew-symmetric (see page 129). Cr ∈ Rn. Then |(A1 .σ(n) = aτ (1). . then |A| = 0. the second from the second column.σ(n) contains exactly one factor from each row and one factor from each column. . . Ar−1 . Since R is commutative.n = aτ γ(1). An )| Proof This is immediate from the deﬁnition of determinant and the distributive law of multiplication in the ring R.1 · aτ γ(2).σ(2) · · · an.

j be the determinant of the (n − 1) × (n − 1) matrix obtained by removing row i and column j from A. More generally. Rewrite the four preceding theorems using rows instead of columns. . . An )| + |(A2 . .j Cn. If a matrix has two rows equal or two columns equal. An )| + |(A1 . . The permutation τ is the ﬁnite product of transpositions.j = (−1)i+j Mi. does not change the determinant. Interchanging two rows or two columns multiplies the determinant by −1.2 Ci. Since the ﬁrst and last of these four terms are zero. Aτ (2) . We know 0 = ¯ |(A1 + A2 . or adding c times one column to another column. Let Mi. Mi. . There are 2n ways to compute | A |. . multiplies the determinant by c.j C2. . . expansion by any row or expansion by any column. the determinant is zero. A3 . Theorem Multiplying any row or column of matrix by a scalar c ∈ R.j C1. . | A |= a1. . Thus if any row or any column is zero. 2.1 + ai. then |A| = sign(τ )|(Aτ (1) . A2 .j + a2.2 + · · · + ai. Let Ci.62 Theorem one. .n . An )| + |(A2 . j) minor and cofactor of A. | A |= ai. A1 + A2 . its determinant is zero.n Ci. A1 . . . . A3 . Aτ (n) )|. An )| = |(A1 . Theorem For any 1 ≤ i ≤ n. An )|.  c1 c2 c3   Exercise . then the determinant is zero. show that |(A2 . . The following theorem is just a summary of some of the work done so far. An )| = −|A|. n). . For any 1 ≤ j ≤ n. A1 . . The following theorem is useful but the proof is a little tedious and should not be done as an exercise. A3 . A2 . A3 . Theorem Proof Exercise If τ is a permutation of (1. A1 .j are called the (i. . Adding c times one row to another row. . . .j and Ci. A3 . a1 a2 a3  Let A =  b1 b2 b3 . A3 . The determinant of A is the sum of six terms. the result follows. .j . .1 Ci. .j + · · · + an. . . . Matrices Chapter 4 Interchanging two columns of A multiplies the determinant by minus Proof For simplicity. .j . . . . or one column is c times another column. . if one row is c times another row.

(For the proof. the matrix whose (j. An elementary matrix of type 3 is a special type of upper or lower triangular matrix. One of the major goals of this chapter is to prove the converse of the preceding corollary. B ∈ Rn . then its determinant is the product of the diagonal elements.. i. | A |= −1. We assume it here without proof. Before . Suppose A ∈ Rn is upper triangular. j) cofactor. so its determinant is 1. If A is an elementary matrix of type 2.Chapter 4 Matrices 63 Write out the determinant of A expanding by the ﬁrst column and also expanding by the second row. | C −1 AC | = |ACC −1 | = | A |. | A | is the product of the diagonal elements. We will prove the ﬁrst statement for upper triangular matrices. Classical adjoint Suppose R is a commutative ring and A ∈ Rn . (Determinant by blocks) Suppose A ∈ Rn . The following remarkable theorem takes some work to prove.e. An elementary matrix of type 2 is obtained from the identity matrix by interchanging two rows or columns. Suppose n > 2 and the theorem is true for matrices in Rn−1 . 1 = | I | = | AA−1 | = | A || A−1 | . i.) Theorem The determinant of the product is the product of the determinants. then | A | is a unit in R and | A−1 | = | A |−1 . and D ∈ Rm . If A ∈ R2 Proof is an upper triangular matrix..j )t . Thus | AB | = | BA | and if C is invertible. The result follows by expanding by the ﬁrst column. A B Then the determinant of is | A || D |. Corollary Proof If A is a unit in Rn . B ∈ Rn. Theorem If A is an upper or lower triangular matrix. if A.e.m . If A is an elementary matrix of type 3. i) term is the (i. | AB | = | A || B |. The classical adjoint of A is (Ci. and thus has determinant −1. O D Theorem Proof Expand by the ﬁrst column and use induction on n. see page 130 of the Appendix. | A |= 1.

i.64 we consider the general case.j )t are all | A | and the other elements are 0. We are now ready for one of the most beautiful and useful theorems in all of mathematics. Proof We must show that the diagonal elements of the product A(Ci.j )t = d −b −c a . .j ) and is thus | A | (computed by expansion by row s).) If A is invertible. B ∈ Rn and AB = I.j )t . That is. A is invertible iﬀ | A | = 0. Thus if | A | = 1. t) term is 0. (Thus if R is a ﬁeld.e. Theorem B is similar to B.j )t A = | A | I. if | A | = 1.j )t = (Ci.j ).j )t . and thus BA = I. suppose A. If A = a b c d then (Ci. The proof that (Ci. the (s. A is d −b −c a . In particular. t) term is the dot product of row s of A with row t of (Ci.j )t = (Ci. then A(Ci. then A−1 = | A |−1 (Ci. invertible and A−1 = | A |−1 (Ci.j )t .j ) = |A| 0 0 |A| d −c −b a Matrices Chapter 4 and so (Ci. Then A is a unit in Rn iﬀ | A | is a unit in R. Exercise Show that any right inverse of A is also a left inverse. let’s examine 2 × 2 matrices. Theorem If R is commutative and A ∈ Rn . Similarity Suppose A. B is similar to A iﬀ B is a conjugate of A. A−1 = (Ci. Thus if | A | is a unit in R. Since this is the determinant of a matrix with row s = row t. Show A is invertible with A−1 = B. For s = t. Then A(Ci. B is said to be similar to A if ∃ an invertible C ∈ Rn such that B = C −1 AC. A−1 = Here is the general case.j )t A = |A|I is similar and is left as an exercise. The (s. B ∈ Rn . s) term is the dot product of row s of A with row s of (Ci. Theorem Suppose R is a commutative ring and A ∈ Rn . the classical adjoint of A. the (s.j )t A = = | A | I.. Proof This follows immediately from the preceding theorem.

. Theorem Suppose A ∈ Rm. Theorem Suppose A and B are similar. then trace(A) = trace(B).m am. One of the most useful properties of trace is trace(AB) = trace(BA) whenever AB and BA are deﬁned.j ) ∈ Rn .j + · · ·+ bj.n . “Similarity” is an equivalence relation on Rn . bn )t .j = ai.. the trace of A is the sum of its diagonal terms. trace(A + B) = trace(A) + trace(B) and trace(AB) = trace(BA). Note that trace(AB) = trace(BA). .2 + · · · + an. bj. By deﬁnition. B ∈ Rn . Here is the theorem in full generality. then D is similar to A. Trace Suppose A = (ai. Theorem If A.1 b1. For example. Proof Suppose B = C −1 AC.i = trace(AB) = 1≤i≤m 1≤i≤m 1≤j≤n 1≤j≤n trace(BA). trace(B) = trace(C −1 AC) = trace(ACC −1 ) = trace(A). 65 Proof This is a good exercise using the deﬁnition. If D is similar to B and B is similar to A. and the second part is a special case of the previous theorem. .Chapter 4 Matrices B is similar to A iﬀ A is similar to B.1 + a2. Then the trace is deﬁned by trace(A) = a1. Proof This proof involves a change in the order of summation. b2 .. That is.m ...1 a1..n and B ∈ Rn.i = ai. an ) and B = (b1 . a2 .i + · · ·+ ai.j bj..n bn. Then AB and BA are square matrices with trace(AB) = trace(BA). Theorem Proof If A and B are similar. suppose A = (a1 . Then | A | = | B | and thus A is invertible iﬀ B is invertible. Then | B | = | C −1 AC | = |ACC −1 | = | A |. Then AB is the scalar a1 b1 + · · · + an bn while BA is the n × n matrix (bi aj ). Proof The ﬁrst part of the theorem is immediate.

then they have the same characteristic polynoCPB (x) = | (xI − C −1 AC) | = | C −1 (xI − A)C | = Proof Suppose B = C −1 AC. i. but depends on linear algebra (see pages 93. Any λ ∈ R which is a root of CPA (x) is called a characteristic root of A. If A and B are similar. Show the following are equivalent. | A |= trace(A) = 0. Exercise Suppose A ∈ Rn and a ∈ R.e. | A | = | B | and trace(A) = trace(B). Determinant is a multiplicative homomorphism and trace is an additive homomorphism.. . Theorem CPA (x) = a0 + a1 x + · · · + an−1 xn−1 + xn where trace(A) = −an−1 and | A | = (−1)n a0 . A = Exercise 1) 2) 3) 4) Suppose F is a ﬁeld and A ∈ F2 . In other words. Suppose R is a commutative ring. and c d CPA (x) = a0 + a1 x + x2 . Exercise a b is a matrix in R2 . 94. Note This exercise is a special case of a more general theorem. the characteristic polynomial CPA (x) ∈ R[x] is deﬁned by CPA (x) = | (xI − A) |. If A and B are similar.66 Matrices Chapter 4 Summary Determinant and trace are functions from Rn to R. and 98). CPA (A) = 0. Characteristic polynomials If A ∈ Rn . A square matrix over a ﬁeld is nilpotent iﬀ all its characteristic roots are 0 iﬀ it is similar to a strictly ¯ upper triangular matrix. ∃ an elementary matrix C such that C −1 AC is strictly upper triangular. Furthermore | AB | = | BA | and trace(AB) = trace(BA). | (xI − A) | = CPA (x). This remarkable result cannot be proved by matrix theory alone. Proof Theorem mials. show A satisﬁes its characteristic polynomial. This follows from a direct computation of the determinant. Find |aA| and trace(aA). Find a0 and a1 and show that a0 I + a1 A + A2 = 0. A2 = 0. CPA (x) = x2 .

this single integer determines V up to isomorphism. It is stated here in full generality only for reference and completeness. we restrict our attention to vector spaces. quotients. and any change of basis corresponds to conjugation of that matrix. and generating sets come together in one uniﬁed theory. Also any endomorphism f : V → V may be represented by a matrix. One of the goals in linear algebra is to select a basis so that the matrix representing f has a simple form. if f is nilpotent. It is in this chapter that the concepts about functions. This shows that ﬁnitely generated free R-modules have a well deﬁned dimension. We give a simple proof that if R is a commutative ring and f : Rn → Rn is a surjective R-module homomorphism. The elementary facts about cosets. then f may be represented by a strictly upper triangular matrix. The key theorem is that any vector space V has a free basis.. For example.e. it has a well deﬁned dimension. The basic theory is developed here in full generality. The proof is given in the Appendix. and with the beautiful theory relating orthogonal matrices and symmetric matrices. and homomorphisms follow the same pattern as in the chapters on groups and rings. 67 . and simpliﬁes some of the development of linear algebra.e. then f is an isomorphism. i. solutions of equations.. and thus if V is ﬁnitely generated. modules are deﬁned over an arbitrary ring R and not just over a ﬁeld. and incredible as it may seem. As another example. if f is not injective. matrices.Chapter 5 Linear Algebra The exalted position held by linear algebra is based upon the subject’s ubiquitous utility and ease of application. The theorem on Jordan canonical form is not proved in this chapter. and should not be considered part of this chapter. After the general theory. i. then f may be represented by a matrix whose ﬁrst column is zero. modules over a ﬁeld. This chapter concludes with the study of real inner product spaces.

Thus for commutative rings. a0R = 0M . a1 . then (−a)r = −(ar) = a(−r). it is assumed that R is a ring and the word “R-module” (or sometimes just “module”) means “right R-module”. ¯ ¯ If a ∈ M . a2 ∈ M and r. If R is commutative and M = MR then left scalar multiplication deﬁned by ra = ar makes M into a left R-module. r1 . Theorem 1) 2) 3) Proof Suppose M is an R-module. m) → rm satisfying r(a1 + a2 ) (r1 + r2 )a (r1 · r2 )a 1a ¯ = = = = ra1 + ra2 r1 a + r2 a r1 (r2 a) a Note that the plus sign is used ambiguously. we may write the scalars on either side. . as addition in M and as addition in R.68 Linear Algebra Chapter 5 Deﬁnition Suppose R is a ring and M is an additive abelian group. In this text we stick to right R-modules. Convention Unless otherwise stated. Notation The fact that M is a right (left) R-module will be denoted by M = MR (M = R M ). ¯ ¯ If a ∈ M and r ∈ R. This is a good exercise in using the axioms for an R-module. then f : M → M deﬁned by f (a) = ar is a homomorphism of additive groups. r2 ∈ R. The statement that M is a right R-module means there is a scalar multiplication M ×R → M (m. In particular (0M )r = 0M . r) → mr satisfying (a1 + a2 )r a(r1 + r2 ) a(r1 · r2 ) a1 ¯ = = = = a1 r + a2 r ar1 + ar2 (ar1 )r2 a for all a. If r ∈ R. The statement that M is a left R-module means there is a scalar multiplication R×M → M (r.

. then this submodule may be written as N1 + N2 + · · +Nn = {a1 + a2 + · · +an : each ai ∈ Ni }. (We know from the last exercise in Chapter 2 that N T is a group. On the left. T is an index set. N ) is a subgroup of N M . f (ar) = f (a)r. i.. t∈T 2) 3) +t∈T Nt = {all ﬁnite sums a1 + · · +am : each ai belongs to some Nt } is a submodule. In this case N will be an R-module because the axioms will automatically be satisﬁed. For example. Also in 8) it is only necessary to show that HomR (M. the statement that a subset N ⊂ M is a submodule means it is a subgroup which is closed under scalar multiplication. an R-module homomorphism) provided it is a group homomorphism and if a ∈ M and r ∈ R. and for each t ∈ T . Note that 0 and M are submodules. n}. 1) t∈T Nt is a submodule of M . Exercise Suppose T is a non-void set. Also the proof of 3) is immediate. if a ∈ N and r ∈ R. Nt is a submodule of M . 2.e. scalar multiplication is in M and on the right it is in N . which is immediate.. in 5) of the theorem below. N ) is a submodule of N M . and in particular for subgroups of additive abelian groups. N1 + N2 is the smallest submodule of M containing N1 ∪ N2 . N ) forms an abelian group. Proof We know from page 22 that versions of 1) and 2) hold for subgroups.. If T = {1.Chapter 5 Linear Algebra 69 Submodules If M is an R-module. Much . . then ar ∈ N. To ﬁnish the proofs it is only necessary to check scalar multiplication. it is stated that HomR (M. So it is only necessary to show that HomR (M. Homomorphisms Suppose M and N are R-modules. and scalar multiplication deﬁned by (f r)(t) = f (t)r. Show N T is an R-module.) This simple fact is quite useful in linear algebra. Note that if N1 and N2 are submodules of M . Theorem Suppose M is an R-module. called the ¯ improper submodules of M . and N T is the collection of all functions f : T → N with addition deﬁned by (f +g)(t) = f (t)+g(t). N is an R-module. A function f : M → N is a homomorphism (i. Nt is a submodule. If {Nt } is a monotonic collection. The basic facts about homomorphisms are listed below. and so it is only necessary to check scalar multiplication.e.

An isomorphism f : M → M is called an automorphism. g : M → N are homomorphisms. In particular. ¯ This is just a series of observations. G ⊂ M is a submodule. Also (−f ) deﬁned by (−f )(a) = −f (a) is a homomorphism. Theorem 1) 2) 3) 4) The zero map M → N is a homomorphism. HomR (M. NR ). The units of the endomorphism ring HomR (M. with multiplication deﬁned to be composition. if f : M → N is a homomorphism. N ) = Hom(MR . Rn ) is just the matrix ring Rn and the automorphisms are merely the invertible matrices. Suppose f : M → N is a homomorphism. is a ring. N ) is an R-module. M ). The identity map I : M → M is a homomorphism. We will see later that if M = R n . f r deﬁned by (f r)(a) = f (ar) = f (a)r is a homomorphism. HomR (M.70 Linear Algebra Chapter 5 of this work has already been done in the chapter on groups (see page 28). (f + g ) ◦ k = (f ◦ k) + (g ◦ k). If k : P → M is a homomorphism. If R is commutative and r ∈ R. 5) 6) 7) 8) 9) Proof . If h : N → P is a homomorphism. forms an abelian group under addition. h ◦ (f + g) = (h ◦ f ) + (h ◦ g). If a bijection f : M → N is a homomorphism. and H ⊂ N is a submodule. In this case f and f −1 are called isomorphisms. The composition of homomorphisms is a homomorphism. A homomorphism f : M → M is called an endomorphism. the set of all homomorphisms from M to N . HomR (Rn . Thus the automorphisms on M form a group under composition. If f. If R is commutative. then g : M → M deﬁned by g(a) = ar is a homomorphism. Then f (G) is a submodule of N and f −1 (H) is a submodule of M. deﬁne (f + g) : M → N by (f + g)(a) = f (a) + g(a). Furthermore. The sum of homomorphisms is a homomorphism. M ) are the automorphisms. image(f ) is a submodule of N and ker(f ) = f −1 (0) is a submodule of M . then f −1 : N → M is also a homomorphism. HomR (M. Then f + g is a homomorphism.

That is. ¯ Proof Suppose f (1) = g(1). Of particular importance is the case Rn = R ⊕ · · ⊕R = Rn. it is shown that any additive group M admits a scalar multiplication by integers. Given m ∈ M . This is the same as the deﬁnition before for Rn when n = 1. Proof The deﬁnitions are the same except expressed in diﬀerent language. and if M is abelian. Summary Additive abelian groups are “the same things” as Z-modules. Then f (r) = f (1 · r) = f (1)r = g(1)r = g(1 · r) = ¯ ¯ ¯ ¯ ¯ ¯ g(r). the properties are satisﬁed to make M a Z-module. Thus . We begin with the case n = 1. if m ∈ M . Furthermore. Furthermore. If M and N are Q-modules and f : M → N is a Z-module homomorphism. HomR (R.n an R-module. Theorem Suppose M = MR and f.n admits a scalar multiplication by elements in R. etc. the study of additive abelian groups is a special case of the study of R-modules.1 (see page 53). This makes R a right R-module denoted by RR (or just R). h : R → M deﬁned by h(r) = mr is a homomorphism. a2 = a + a. Then f = g. because a1 = a. ∃! homomorphism h : R → M with ¯ h(1) = m. R as a right R-module Let M = R and deﬁne scalar multiplication on the right by ar = a · r. Theorem Suppose R is a ring and N is a subset of R. scalar multiplication is just ring multiplication. Note that this is the only way M can be a Zmodule. While group theory in general is quite separate from linear algebra.Chapter 5 Linear Algebra 71 Abelian groups are Z-modules On page 21. Then N is a submodule of RR (R R) iﬀ N is a right (left) ideal of R. R-modules are also Z-modules and R-module homomorphisms are also Exercise Z-module homomorphisms. must it also be a Q-module homomorphism? Homomorphisms on Rn Rn as an R-module On page 54 it was shown that the additive abelian group Rm. g : R → M are homomorphisms with f (1) = ¯ g(1). then f is also a Z-module homomorphism. if f : M → N is a group homomorphism of abelian groups. M ) ≈ M . In other words. The properties listed there were exactly those needed to make Rm.

. Furthermore. 0 r1  ¯     ·   ·       Homomorphisms on Rn Deﬁne ei ∈ Rn by ei =  1i . Any R-module homomorphism from Rn to M is determined by its values on the basis. M ) is an R-module. Show f is injective.. . ∃! homomorphism h : Rn → M with h(ei ) = mi for 1 ≤ i ≤ m. M ) to M . Theorem Suppose M = MR and f. We will see later that the product M n is an R-module with scalar multiplication deﬁned by (m1 . If R is commutative so that HomR (Rn .72 Linear Algebra Chapter 5 evaluation at 1 gives a bijection from HomR (R. this theorem gives an R-module isomorphism from HomR (Rn . the above theorem states that multiplication on left by some m ∈ R deﬁnes a right R-module homomorphism from R to R. and every module homomorphism is of this form. en } is called the canonical free basis or standard basis for R n . mn r). or which R-module M is selected. M ) to M n = M × M × · · ×M and this bijection is a group isomorphism. In the case M = R. mn ∈ M . Note that any   ¯       ·  ·     0 rn ¯ can be written uniquely as e1 r1 + · · +en rn .. m2 . Exercise Suppose R is a ﬁeld and f : RR → M is a non-zero homomorphism. if m1 ... The homomorphism h is deﬁned by h(e1 r1 + · · +en rn ) = m1 r1 + · · +mn rn . Then f = g. it is an isomorphism of R-modules. M ) to M n ..     . . g : Rn → M are homomorphisms with f (ei ) = g(ei ) for 1 ≤ i ≤ n. The element m should be thought of as a 1 × 1 matrix. mn )r = (m1 r. and any function from that basis to M extends uniquely to a homomorphism from Rn to M . and this bijection is clearly ¯ a group isomorphism. m2 . m2 r. This theorem reveals some of the great simplicity of linear algebra. We now consider the case where the domain is Rn . ... Proof The proof is straightforward. Note this theorem gives a bijection from HomR (Rn . If R is commutative. It does not matter how complicated the ring R is... The sequence {e1 .

These properties are made explicit in the next two theorems. Corollary HomR (Rn .Chapter 5 Linear Algebra 73 Now let’s examine the special case M = Rm and show HomR (Rn . Thus HomR (Rn .m . then f + g is given by the matrix A + C. . Theorem Suppose A = (ai. Matrices over R give R-module homomorphisms! Furthermore. Rm ) ≈ Rm. g : Rn → Rm are given by matrices A. .n . composition of homomorphisms corresponds to multiplication of matrices. In the case where the domain and range are the same. we have the following elegant corollary. Conversely. C ∈ Rm. . Then f : Rn → Rm deﬁned by f (B) = AB is a homomorphism with f (ei ) = column i of A. Even though this follows easily from the previous theorem and properties of matrices. Rm ) and Rm. they are isomorphic as R-modules. addition of matrices corresponds to addition of homomorphisms. The previous theorem reveals where matrix multiplication comes from. . Proof This is just the associative law of matrix multiplication. Even if R is commutative.n and g : Rm → Rp is the homomorphism given by C ∈ Rp.n to be the matrix with column i = vi . If R is commutative. it is one of the great classical facts of linear algebra.j ) ∈ Rm.n are isomorphic as additive groups. This corollary shows one way non-commutative rings arise. Then f deﬁned by f (B) = AB is the unique homomorphism from Rn to Rm with f (ei ) = vi . We now return to the general theory of modules (over some given ring R). C(AB) = (CA)B. then g ◦ f : Rn → Rp is given by CA ∈ Rp. namely as endomorphism rings. Rn ) and Rn are isomorphic as rings. The automorphisms correspond to the invertible matrices. Rn is never commutative unless n = 1. Theorem If f : Rn → Rm is the homomorphism given by A ∈ Rm. deﬁne A ∈ Rm.n . That is. It is the matrix which represents the composition of the functions. . and multiplication of matrices corresponds to composition of homomorphisms.n . vn ∈ Rm . Theorem If f.n . if v1 .

as shown by the next theorem. the additive abelian quotient group M/N is deﬁned. The image of f is the ¯ ¯ ¯ image of f . Since N is a normal subgroup of M . quotient modules go through without a hitch. then M/N ≈ M (see below). Theorem Suppose M is a module and N ⊂ M is a submodule. Proof On the group level this is all known from Chapter 2 (see page 29). f is injective. Furthermore. Scalar multiplication deﬁned by (a + N )r = (ar + N ) is well-deﬁned and gives M/N the structure of an R-module. f M π c && & & & E b & & ¯ M & ¯ f M/N Thus deﬁning a homomorphism on a quotient module is the same as deﬁning a homo¯ morphism on the numerator that sends the denominator to 0. ¯ : (M/N ) → M deﬁned by f (a + N ) = f (a) is a well-deﬁned ¯ ¯ If N ⊂ ker(f ). Thus if N = ker(f ). The natural projection π : M → M/N is a surjective ¯ homomorphism with kernel N . (domain(f )/ker(f )) ≈ image(f ).74 Linear Algebra Cosets and Quotient Modules Chapter 5 After seeing quotient groups and quotient rings. this is all known from Chapter 2 (see pages 27 and 29). ¯ Theorem Suppose f : M → M is a homomorphism and N is a submodule of M . if f : M → M is a surjective homomor¯ phism with ker(f ) = N . It is ¯ only necessary to check that f is a module homomorphism. It is only necessary to check the scalar multiplication. and the kernel of f is ker(f )/N . As before. Therefore for any homomorphism f . The relationship between quotients and homomorphisms for modules is the same as for groups and rings. . then f homomorphism making the following diagram commute. R is a ring and module means R-module. and thus (M/N ) ≈ image(f ). which is obvious. and this is immediate. Proof On the group level.

.e.. ii) Suppose K ⊂ L. Under the natural 1-1 correspondence from {functions f : M → Mt } to {sequence of functions {ft }t∈T where ft : M → Mt }. This makes the structure of module homomorphisms much more simple. Suppose M is a module. These two examples are for the case R = Z. i. On the additive abelian group Mt = Mt deﬁne scalar multiplication by {mt }r = t∈T {mt r}. Since scalar multiplication is deﬁned coordinatewise. Theorem Suppose T is an index set and for each t ∈ T . In the ﬁnite case something important holds for modules that does not hold for non-abelian groups or rings – namely. Proof We already know from Chapter 2 that f is a group homomorphism iﬀ each ft is a group homomorphism. although the student should think of the ﬁnite case.. f is a homomorphism iﬀ each ft is a homomorphism. The natural homomorphism K → (K + L)/L is surjective with kernel ≈ K ∩ L. For the ﬁnite case we may use either the product or sum notation. The natural homomorphism M/K → M/L is surjective ≈ with kernel L/K. Then Mt is an R-module and. This is stated below in full generality. Thus (K/K ∩ L) → (K + L)/L is an isomorphism. M =Z K = 3Z L = 5Z K ∩ L = 15Z K/K ∩ L = 3Z/15Z ≈ Z/5Z = (K + L)/L K +L=Z Examples 1) 2) M =Z K = 6Z L = 3Z (K ⊂ L) (M/K)/(L/K) = (Z/6Z)/(3Z/6Z) ≈ Z/3Z = M/L Products and Coproducts Inﬁnite products work ﬁne for modules. the natural projection πs : Mt → Ms is a homomorphism. just as they do for groups and rings. for abelian groups.Chapter 5 Linear Algebra 75 Theorem i) Suppose M is an R-module and K and L are submodules of M . i. the ﬁnite product is also a coproduct. Mt is an R-module.e. Thus (M/K)/(L/K) → M/L is an isomorphism. for each s ∈ T . M1 × M2 × · · ×Mn = M1 ⊕ M2 ⊕ · · ⊕Mn . f is a module homomorphism iﬀ each ft is a module homomorphism.

f2 ) and h = (h1 . Since there are only a ﬁnite number of non-zero terms. f2 + h2 ). If T is inﬁnite. M1 ) ⊕ · · ⊕HomR (M. g is deﬁned by g({mt }) = gt (mt ).76 Linear Algebra Chapter 5 Deﬁnition If T is ﬁnite. . For T = {1. t M f 1      d d M f d 2 g1         E    d T s d M1   ©   ' f c π1 M1 ⊕ M 2 π2 d  d E g d g2 d d M2 M1 i1 M1 ⊕ M 2 ' i2 M2 Theorem For ﬁnite T . M ) ≈ HomR (M1 . gt is deﬁned by gt = g ◦ it . so that the objects are R-modules and not merely additive groups. M1 ⊕ · · ⊕Mn ) ≈ HomR (M. Thus each Ms may be considered to be a submodule ¯ of ⊕Mt . Theorem Suppose M is an R-module. HomR (M. this sum is well deﬁned. f2 )r = (f1 r. Given g. they give isomorphisms of Rmodules. If R is commutative. M ) and Proof Let’s look at this theorem for products with n = 2. the inclusion homomorphisms is : Ms → ⊕Mt is deﬁned by is (a) = {at } where at = 0 if t = s and as = a. then the isomorphisms are module isomorphisms. M ) ⊕ · · ⊕HomR (Mn . This says merely that f r = (f1 . h2 ). All it says is that if f = (f1 . Given {gt }. There is a 1-1 correspondence from {homomorphisms g : ⊕Mt → M } and {sequences of homomorphisms {gt }t∈T where gt : Mt → M } . Mn ) HomR (M1 ⊕ · · ⊕Mn . the coproduct or sum Mt = Mt = ⊕Mt is the submodule of Mt t∈T t∈T consisting of all sequences {mt } with only a ﬁnite number of non-zero terms. then f + h = (f1 + h1 . the coproduct and product are the same module. the 1-1 correspondences in the above theorems actually produce group isomorphisms. If R is commutative. 2} the product and sum properties are displayed in the following commutative diagrams. For each s ∈ T . f2 r).

The statement that K is a summand of M means ∃ a submodule L of M with K ⊕ L = M . Of course. −a) : a ∈ K ∩ L}. Then the projection ¯ map π2 : M1 ⊕ M2 → M2 is a surjective homomorphism with kernel M1 . M ) → M n which is a R-module isomorphism. In particular. Thus (M1 ⊕ M2 )/M1 is isomorphic to M2 . If such an L exists. . These isomorphisms are transparent and are used routinely in algebra without comment (see Th 4. f (k. Show K is a submodule of M ⊕ M which is a summand. Theorem Consider M1 = M1 ⊕0 as a submodule of M1 ⊕M2 . because L ≈ M/K. but it will be unique up to ¯ isomorphism. M and 0 are always summands of M . and n ≥ 1. Exercise Suppose R is a commutative ring. Thus f is an isomorphism iﬀ K + L = M and K ∩ L = 0. Is Q a summand of R? With the material at hand. if a ∈ R and b ∈ R. Show that M ⊕ N is isomorphic to N ⊕ M. Summands One basic question in algebra is “When does a module split as the sum of two modules?”. and Q ⊂ R is a submodule. (See page 35 for the group version. Before deﬁning summand. and the next theorem is almost as intuitive. M is an R-module. this is the same as there exists a submodule L with K + L = M and K ∩ L = 0. page 118).) This is exactly what you would expect. Deﬁne a function α : HomR (Rn . Exercise R is a module over Q. Later on. Now suppose A ⊂ M. In this case we write K ⊕ L = M . Suppose K and L are submodules of M and f : K ⊕ L → M is the Theorem natural homomorphism. For example. then (R ⊕ R)/(aR ⊕ bR) is isomorphic to (R/aR) ⊕ (R/bR). ¯ Exercise Suppose M is a module and K = {(m. According to the previous theorem. it will be easy. this is not an easy question. Then the image of f is K + L and the kernel of f is {(a. here are two theorems for background. the abelian group (Z ⊕ Z)/(2Z ⊕ 3Z) is isomorphic to Z2 ⊕ Z3 . m) : m ∈ M } ⊂ M ⊕ M . This abuse of notation allows us to ¯ avoid talking about “internal” and “external” direct sums.Chapter 5 Linear Algebra 77 Exercise Suppose M and N are R-modules. l) = k + l. Deﬁnition Suppose K is a submodule of M . B ⊂ N are submodules and show (M ⊕ N )/(A ⊕ B) is isomorphic to (M/A) ⊕ (N/B). it need not be unique.

and Free Basis This section is a generalization and abstraction of the brief section Homomorphisms on Rn ..78 Linear Algebra Chapter 5 Exercise Answer the following questions about abelian groups.. to avoid dizziness. Deﬁnition Suppose M is an R-module. Theorem For each t ∈ T . then S is said to be a basis or free basis for M . (See the third exercise on page 35. then S is dependent. rn in R. . st ∈ M . The statement that S is dependent means ∃ a ﬁnite number of distinct elements t1 . tn in T . . If S is independent and generates M . the student should ﬁrst consider the case where T is ﬁnite. deﬁne the center of T to be the subring {t : ts = st for all s ∈ T }. .. then S is dependent.. These concepts work ﬁne for an inﬁnite index set T because linear combination means ﬁnite linear combination. When is nZmn a summand of Zmn ? Exercise If T is a ring. Let S be the sequence {st }t∈T = {st }.. SR is a submodule of M called the submodule generated by S. T is an index set. and elements r1 . such that the linear combination st1 r1 + · · +stn rn = 0. let Rt = RR and for each c ∈ T ... In this case any v ∈ M can be written uniquely as a linear combination of elements in S.) 1) 2) 3) 4) Is 2Z a summand of Z? Is 2Z4 a summand of Z4 ? Is 3Z12 a summand of Z12 ? Suppose m. n > 1. However. Note that if some st = 0. 2. Show R n is a left T -module and ﬁnd HomT (Rn .. let ec ∈ ⊕Rt = Rt t∈T be ec = {rt } where rc = l and rt = 0 if t = c. Independence. ¯ S is independent. n} and ⊕Rt = Rn (see p 72). Generating Sets. You might try ﬁrst the case T = {1. . Otherwise. An R-module M is said to be a free R-module if it is zero or if it has a basis. except for the confusing notation. not all zero. Let R be a commutative ring and T = Rn . The next two theorems are obvious.e. There is a exercise on page 57 to show that the center of T is the subring of scalar matrices. and for each t ∈ T . i. Also if ∃ distinct ¯ elements t1 and t2 in T with st1 = st2 . Rn ).. Z-modules. Let SR be the set of all linear combinations st1 r1 + · · +stn rn . Then {ec }c∈T is a basis for ⊕Rt called ¯ ¯ the canonical basis or standard basis.

Chapter 5 Linear Algebra 79 Theorem Suppose N is an R-module and M is a free R-module with a basis {st }. M≈ t∈T Proof If M is isomorphic to ⊕Rt then M is certainly free. f (S) is independent in N iﬀ f is injective. Then ∃ a 1-1 correspondence from the set of all functions g : {st } → N and the set of all homomorphisms f : M → N . If h : M → N is a homomorphism.. Theorem Suppose N is a module. Given g. f is completely determined by what it does on the basis S. it should be worked carefully.. Let f (S) be the sequence {f (st )} in N . The next theorem is so basic in linear algebra that it is used without comment. M is a free module with basis S = {st }. This is just an observation. Theorem A non-zero R-module M is free iﬀ ∃ an index set T such that Rt . 1) 2) 3) 4) f (S) generates N iﬀ f is surjective. . . and you are “free” to send the basis any place you wish and extend to a homomorphism. deﬁne f by f (st1 r1 + · · +stn rn ) = g(st1 )r1 + · · +g(stn )rn . f (S) is a basis for N iﬀ f is an isomorphism. and f : M → N is a homomorphism. In other words. Although the proof is easy. Exercise Let (A1 . deﬁne g by g(st ) = f (st ). In particular. So now suppose M has a free basis {st }. Show this sequence is linearly independent over Z iﬀ it is linearly independent over Q. then f = h iﬀ f | S = h | S. An ) be a sequence of n vectors with each Ai ∈ Zn . f is an isomorphism. Is it true the sequence is linearly independent over Z iﬀ it is linearly independent over R? This question is diﬃcult until we learn more linear algebra. Recall that we have already had the preceding theorem in the case S is the canonical basis for M = Rn (p 72). Given f . but it is a central fact in linear algebra. By 3) in the preceding theorem. Characterization of Free Modules Any free R-module is isomorphic to one of the canonical free R-modules ⊕Rt . Then the homomorphism f : M → ⊕Rt with f (st ) = et sends the basis for M to the canonical basis for ⊕Rt . M has a ﬁnite free basis of n elements iﬀ M ≈ Rn .

en } is a free basis for Rn . ¯ v is independent iﬀ v is v is a basis for R iﬀ v generates R iﬀ v is . A = If R is a commutative ring. The next exercise is routine. and the homomorphism f : Rn → Rn deﬁned by f (B) = AB is surjective. A ∈ Rn . Let B =  ..  represent an element of Rn and C =  . i. Use the fact that {e1 .e. and solutions of equations. are all the same — they are merely stated in diﬀerent language.. i.e.80 Linear Algebra Chapter 5 Exercise Suppose R is a commutative ring. . Relating these concepts to matrices The theorem stated below gives a summary of results we have already had. show A is invertible.. injective homomorphisms. . you can relate properties of R as an R-module to properties of R as a ring.Let v1 .. . but still informative. f (B) = AB.. Suppose A ∈ Rm. Exercise 1) 2) Suppose R is a commutative ring and v ∈ R. linear independence.  bn cm . Find a non-trivial linear combination of the columns of A which is 0.e. It shows that certain concepts about matrices. This is a key theorem in linear algebra. Note that 2) here is essentially the ﬁrst exercise for the case n = 1. v = 0. then f is an isomorphism. f (ei ) = vi    b1 c1     = column i of A. Show f is an isomorphism. That is..n and f : Rn → Rm is the homomorphism associated with A. Also ﬁnd a non-zero element of kernel(f ). 2 1 0 and f: Z3 → Z2 be the group homo3 2 −5 morphism deﬁned by A. ¯ Exercise Let R = Z. vn ∈ Rm be the columns of A. although it is usually stated only for the case where R is a ﬁeld.. i. if f : R → R is a surjective R-module homomorphism.

. .. . {v1 . wn } is a basis for Rn .Chapter 5 Linear Algebra 81 represent an element of Rm . vn } is a basis for Rm iﬀ f is an isomorphism iﬀ (for any C ∈ Rm . {v1 . Then the following are equivalent. At is invertible. So far in this chapter we have just been cataloging... A ∈ Rn . | At | is a unit in R. AX = C has a solution).. Later on we will prove that if R is a ﬁeld. A is invertible. . | A | is a unit in R. (See bottom of page 89. Now we prove something more substantial.n be the rows of A.e. 1) 2) 3) 4) 5) 2t ) 3t ) f is an automorphism. i.. .. {v1 . {v1 . . vn ∈ Rn be the columns of A.. {w1 . Theorem 1) The element f (B) is a linear combination of the columns of A. and f : Rn → Rn is deﬁned by f (B) = AB.. . .. . that is f (B) = f (e1 b1 + · · +en bn ) = v1 b1 + · · +vn bn . Let v1 . Thus the image of f is generated by the columns of A. injective implies surjective... and w1 . namely that if f : Rn → Rn is surjective. vn } generates Rn . Theorem Suppose R is a commutative ring. . i. then f is injective... AX = C has a unique solution). wn ∈ Rn = R1. vn } is a basis for Rn . vn } generates Rm iﬀ f is surjective iﬀ (for any C ∈ Rm ... 2) 3) 4) Relating these concepts to square matrices We now look at the preceding theorem in the special case where n = m and R is a commutative ring.. f is surjective.) {v1 . vn } is independent iﬀ f is injective iﬀ AX = 0 has a unique ¯ solution iﬀ (∃ C ∈ Rm such that AX = C has a unique solution)..e.

g. Then h is a surjective homomorphism. Let g : Rn → Rn be the homomorphism satisfying g(ei ) = ui . as shown below. then m ≥ n. If R is commutative. R will be a commutative ring.82 4t ) {w1 . and by the previous section. a ﬁnite generated R-module).. Proof Suppose k = n − m is positive. This is a contradiction and thus m ≥ n. which requires that R be commutative (see the exercise on page 64). but this is of little interest. Since f is onto.. Then if M has a basis. Deﬁne h : (Rm ⊕ Rk = Rn ) → Rn by h(u. Furthermore. 2) and 2t ) are equivalent. So this follows from the ﬁrst theorem. then m ≥ n. Thus the ﬁrst ﬁve are equivalent. Now g comes from some matrix D and thus AD = I. 4) and 5) are equivalent. First we make a convention. Corollary Proof The hypothesis implies there is a surjective homomorphism R m → Rn .e... wn } generates Rn . vm } generates Rn . ∃ u1 . module (i. Then f ◦ g is the identity.. If {v1 . so m = n by the previous theorem. . This shows that A has a right inverse and is thus invertible.. also injective. applying this result to At shows that the last three properties are equivalent to each other. this is impossible. Linear Algebra Chapter 5 Proof Suppose 5) is true and show 2). un ∈ Rn with f (ui ) = ei . that basis is ﬁnite.. Corollary Proof If f : Rm → Rn is an isomorphism. Convention Theorem For the remainder of this chapter. Lemma Suppose M is a f.. . Since | A |=| At |. then m = n. . and 3) implies 4). We already know the ﬁrst three properties are equivalent. v) = f (u). If f : Rm → Rn is a surjective R-module homomorphism. Each of f and f −1 is surjective. . Recall that the proof of this fact uses determinant. Uniqueness of Dimension There exists a ring R with R2 ≈ R3 as R-modules.

h(en )). trace. change of basis can be displayed by the diagram below. Endomorphisms on Rn are represented by square matrices.. . The point of all this is that selecting a basis of n elements for M Summary is the same as selecting an isomorphism from Rn to M . . M has a basis of n elements iﬀ M ≈ Rn .. In order to represent f by a matrix. un } or if you wish. We will show that this matrix is well deﬁned up to similarity. an isomorphism with Rn ). independence. 2. (By convention.g. the dimension of M. you must use sequences.. basis is deﬁned for sequences. n}.Chapter 5 Linear Algebra 83 Proof Suppose U ⊂ M is a ﬁnite generating set and S is a basis.. but for independence and basis. . Now suppose M is a f. and thus the determinant. Here we follow the classical convention that an index set with n elements will be {1. and thus S is ﬁnite. Two sequences cannot begin to be equal unless they have the same index set. Change of Basis Before changing basis. and characteristic polynomial of f are well-deﬁned. If M has a basis. Now h : Rn → M is an isomorphism iﬀ α(h) is a basis for M ... there is no concept of repetition. and thus a basis for M with n elements is a sequence S = {u1 . Then any element of U is a ﬁnite linear combination of elements of S. In order to make sense.) ¯ Proof By the previous lemma. not for collections. Consider the columns of the real matrix 2 3 2 A= . S = (u1 . we must select a basis for M (i. yet we certainly don’t wish to say the columns of A form a basis for R2 . free module and f : M → M is a homomorphism. Theorem Suppose M is a f. Recall there is a bijection α : HomR (Rn .. not for sets or collections. M ) → M n deﬁned by α(h) = (h(e1 ). . . we must consider the columns of A as an ordered triple of vectors. The result follows because Rn ≈ Rm iﬀ n = m.. and from this viewpoint. Previously we deﬁned generating. that basis is ﬁnite and any other basis has the same number of elements. there are 1 4 1 only two of them. and thus have a determinant and trace. module. Suppose M is an R-module with a basis of n elements. If we consider the column vectors of A as a collection. and this sequence is dependent. un ) ∈ M n . This number is denoted by dim(M ).. any basis for M must be ﬁnite. In the deﬁnition of basis on page 78. For the concept of generating it matters not whether you use sequences or collections. and basis for sequences.g. In a set or collection.e. 0 is a free module of dimension 0. we recall what a basis is.

and f : M → M is a homomorphism. Then T is a basis for M and the matrix of f w. Deﬁnition Suppose M is a f. and U −1 W U represents f w.i + · · +un cn. An important special case is where M = Rn and f : Rn → Rn is given by some matrix W .r. trace(A)=trace(B).84 Linear Algebra Chapter 5 Deﬁnition Suppose M is a free module. Deﬁne T = {v1 . ..r. Deﬁne C = (ci.j ) ∈ Rn is invertible. column i of A is h−1 (f (h(ei ))). i. Then C is invertible and B = C −1 AC. Conversely. trace(f ) to be trace(A).i .t...t.j ) ∈ Rn of f w. free module and f : M → M is a homomorphism.r.i + · · +un an. the basis {u1 .r.g. vn } by vi = u1 c1. Theorem Suppose T = {v1 .t. In other words. Then h is given by the matrix U whose ith column is ui and A = U −1 W U.r.. . (Note that if M = Rn and ui = ei . Let h : Rn → M be the isomorphism with h(ei ) = ui for 1 ≤ i ≤ n. Proof The proof follows by seeing that the following diagram is commutative. T is B = C −1 AC. Then the matrix A ∈ Rn is the one determined by the endomorphism h−1 ◦ f ◦ h : Rn → Rn . suppose C = (ci. vn } is another basis for M and B ∈ Rn is the matrix of f w. where A is . the basis S. A is the usual matrix associated with f ).t..i + · · +un cn. . .i . un }. In other words. W represents f w. Deﬁne |f | to be |A|...i . The matrix A = (ai. In other words.e.. and A and B have the same characteristic polynomial (see page 66 of chapter 4).t. Rn B d s E ei s d d d d ≈d vi d  d ei        © ≈   vi   ©   Rn    C ≈ c M   ©   f E M d us ≈ id d E s d d d d  ≈  ui         ≈ C c Rn ©i   e  ed i A Rn The diagram also explains what it means for A to be the matrix of f w. un } is a basis for M . the basis S is deﬁned by f (ui ) = u1 a1. and CPf (x) to be CPA (x). the standard basis. S = {u1 . T . conjugation of matrices corresponds to change of basis.j ) ∈ Rn by vi = u1 c1. Therefore |A| = |B|.. A and B are similar.t.r.

Find the determinant. 1)t }. any non-zero element of M is a basis. The previous theorem. Then 0= (vr)r = v1 = v. and characteristic polynomial. ¯ That is. trace. and any two elements of M are dependent. In this section. Finally. Also ﬁnd the matrix B ∈ R2 which represents f with respect to the standard basis. By the previous theorem. some basis. −1)t . and show that in this case. ¯ ¯ ¯ ¯ Theorem Suppose M = 0 is an F -module and v ∈ M . Find the matrix A ∈ R2 which represents f with respect to the basis {(1. . or characteristic polynomial.Chapter 5 Linear Algebra 85 the matrix of f w. and characteristic polynomial of f . Theorem Suppose M is an F -module and v ∈ M . but it must be assumed that the module M is free. trace.r. Endomorphisms in general will not have a determinant. (−1. if these conditions hold. for example. 1)t ) = (1. every F -module is free. 3 3 0 −1 D. Vector Spaces So far in this chapter we have been developing the theory of linear algebra in general.. ﬁnd an invertible matrix C ∈ R2 with B = C −1 AC. We now focus on the case where R is a ﬁeld F . all three are well-deﬁned. Show there is one and only one homomorphism f : R2 → R2 which is the identity on L and has f ((−1.t. if v ∈ V and r ∈ F . the basis Exercise Let L ⊂ R2 be the line L = {(r. then M ≈ FF .r.e. Furthermore.t. vr = 0 implies v = 0 in M or r = 0 in F . F is a ﬁeld. holds for any commutative ring R. and endomorphisms on it will have well-deﬁned determinant. Then v = 0 iﬀ v is independent. 2)t . Find the matrix of f w. F -modules may also be called vector spaces and F -module homomorphisms may also be called linear transformations. Then v generates M iﬀ v ¯ is a basis for M . 2r)t : r ∈ R}. i. trace. do not depend upon the choice of basis. 3 1 . Thus any ﬁnitely generated F -module will have a well-deﬁned dimension. ¯ ¯ ¯ −1 Proof Suppose vr = 0 and r = 0. Exercise Let R = Z and f : Z2 → Z2 be deﬁned by f (D) = 2 1 .

Theorem Suppose M = 0 is a ﬁnitely generated F -module. vm } is an independent sequence in M.. Thus T extends to a basis. Then v = 0 and is thus independent by the ¯ previous theorem. The next theorem is just a collection of observations. .. and any two elements of F are dependent. Theorem Suppose M is an F -module of dimension n. Also the ﬁniteness hypothesis in this theorem is only for convenience. This is one of the theorems that makes linear algebra tick. such that v1 r1 + · · +vn rn + vi ri = 0. vm } extends to a basis with n elements. It not only says any ﬁnite independent sequence can be extended to a basis.. ∃ r1 . vn }. Thus any ﬁnite independent sequence can be extended to a basis. and thus the dimension of M is well deﬁned (see theorem on page 83). . and the element 2 of Z is independent. {v1 .. In this case M ≈ F . Proof Suppose. vm } ¯ generates M . Theorem Suppose M and N are ﬁnitely generated F -modules. vi } is dependent.. ri not all −1 zero... for notational convenience. Since F is a commutative ring.. . then any maximal independent subsequence of S is a basis for M . vn } generates S and thus all of M . M has a ﬁnite free basis. . After so many routine theorems. it is nice to have one with real power..86 Linear Algebra Chapter 5 Proof Suppose v generates M . If S = {v1 . . The key hypothesis here is that the ring is a ﬁeld. and {v1 . . It must be shown that vi is a linear combination of {v1 . vn } is a maximal independent subsequence of S. .. ¯ ¯ Thus {v1 . . Since {v1 ... then Z is a free module over itself. vn .. but it can be extended to a basis inside any ﬁnite generating set containing it. Then m ≤ n and if m = n. .. If R = Z. and n < i ≤ m. .. and thus is a free F -module.. and inside that sequence it may be extended to a maximal independent sequence. Then ri = 0 and vi = −(v1 r1 + · · +vn rn )ri .... rn . Proof {v1 . that {v1 . In particular. any two bases of M must have the same number of elements. T may be extended to a ﬁnite generating sequence. However it certainly cannot be extended to a basis. and any non-zero element of F is a basis. as will be seen momentarily.. Now suppose T is a ﬁnite independent sequence. vm } is a basis.

Then S −T generates a submodule L with K ⊕ L = M. K is a summand of M . Theorem 1) 2) Suppose M is an F -module and K ⊂ M is a submodule. Then dim(M ) = dim(ker(f )) + dim(image(f )). 87 Here is the basic theorem for vector spaces in full generality.e. Theorem 1) 2) 3) Suppose M = 0 is an F -module and S = {vt }t∈T generates M . Part 2) follows from 1). F -module. Any independent subsequence of S may be extended to a maximal independent subsequence of S. dim(M ⊕ N ) = dim(M ) + dim(N ). Proof The proof of 1) is the same as in the case where S is ﬁnite. Any independent subsequence of M can be extended to a basis for M . ¯ Any maximal independent subsequence of S is a basis for M . (See exercise on page 77. i. and Q is a submodule of R. Extend T to a basis S for M .) Proof Q is a ﬁeld. F m ≈ F n iﬀ n = m. Corollary Suppose M is a f. . An independent subsequence of S is contained in a maximal monotonic tower of independent subsequences. Corollary Q is a summand of R. If M is f. R is a Q-module.. ∃ a submodule L of M with K ⊕ L = M . M ≈ N iﬀ dim(M ) = dim(N ). In other words. ∃ a Q-submodule V ⊂ R with Q ⊕ V = R as Q-modules.g. Part 3) follows from 2) because an independent sequence can always be extended to a generating sequence.Chapter 5 1) 2) 3) 4) Linear Algebra M ≈ F n iﬀ dim(M ) = n. and thus to a basis for M . M has a free basis. The union of these independent subsequences is still independent.g. and thus is a free F -module. Part 2) will follow from the Hausdorﬀ Maximality Principle. N is an F -module. and f : M → N is a homomorphism. and so the result follows. then dim(K) ≤ dim(M ) and K = M iﬀ dim(K) = dim(M ). In particular. Proof Let T be a basis for K..

every submodule is a summand. f is injective.e.. 1. for R-modules. f is an automorphism. . | At |= 0. wn } is independent. You may use determinant.. (3. .. 1).. A is invertible. (2. ¯ {v1 . Theorem Suppose A ∈ Fn and f : F n → F n is deﬁned by f (B) = AB.e.. Show R is a ﬁeld. Exercise Find a free Z-module which has a generating set containing no basis. {w1 .. i. wn ∈ F n = F1. Square matrices over ﬁelds This theorem is just a summary of what we have for square matrices over ﬁelds.. wn } is a basis for F n . . 1). 0).e.e.. wn } generates F n . 1) 2) {v1 . vn ∈ F n be the columns of A. 2).... i. {w1 . vn } is independent. . and w1 .... vn } generates F n ... (3. (1.n be the rows of A. i.e.. Then f | L : L → image(f ) is an isomorphism. (1.. Let v1 . Show there are three maximal independent subsequences of S. vn } is a basis for F n . 0)}. . {v1 . (Row vectors are used here just for convenience. . | A |= 0. . i. Show there are three maximal independent subsequences of S and each is a basis for R3 . 3) 1t ) 2t ) 3t ) . 5). f is surjective.) The real vector space R3 is generated by S = {(1. i. i. Exercise The real vector space R2 is generated by the sequence S = {(π. ¯ {w1 . 2)}. At is invertible. i.. Exercise Suppose R is a domain with the property that. 2...e. and each is a basis for R2 . Then the following are equivalent.e..88 Linear Algebra Chapter 5 Proof Let K = ker(f ) and L ⊂ M be a submodule with K ⊕ L = M . 2.. . 4.

Show the following are equivalent. then the row (column) rank of A is the same as the row (column) rank of CAD. pages 81 and 82. The sequence (A1 . and we know from page 81 that each f (B) is a linear . f is injective iﬀ f is bijective iﬀ f is surjective. An ) is linearly independent over Z. Rank of a matrix Suppose A ∈ Fm. The sequence (A1 . This shows some of the simple and deﬁnitive nature of linear algebra. Exercise Add to this theorem more equivalent statements in terms of solutions of n equations in n unknowns.. . An ) be an n × n matrix over Z with column i = Ai ∈ n ¯ Z . By the pigeonhole principle. Now suppose each of U and V is a vector space of dimension n and f : U → V is a linear transformation. Let f : Zn → Zn be deﬁned by f (B) = AB and f : Rn → Rn be deﬁned by ¯(C) = AC. It follows from the work done so far that f is injective iﬀ f is bijective iﬀ f is surjective.) f 1) 2) 3) 4) 5) f : Zn → Zn is injective. Exercise Let A = (A1 . . Overview Suppose each of X and Y is a set with n elements and f : X → Y is a function.. . ¯ f : Rn → Rn is injective. this theorem holds for any commutative ring R. An ) is linearly independent over R.Chapter 5 Linear Algebra 89 Proof Except for 1) and 1t ).. The row (column) rank of A is deﬁned to be the dimension of the submodule of F n (F m ) generated by the rows (columns) of A. |A| = 0.) Parts 1) and 1t ) follow from the preceding section. (See the exercise on page 79.. Theorem If C ∈ Fm and D ∈ Fn are invertible.. Proof Suppose f : F n → F m is deﬁned by f (B) = AB. (See the section Relating these concepts to square matrices. Each column of A is a vector in the range F m .n ..

By the theorem above. For n = 2. the row rank and the column rank of A are equal. Thus the image of f is the submodule of F m generated by the columns of A. the rank of f is the dimension of the image of f . There is a concept of the n-dimensional volume of V . and for n = 3 it is “ordinary volume”. the next theorem says that “f multiplies volume by the absolute value of its determinant”. and its dimension is the column rank of A. This proves the theorem for column rank.1 = ·· = ht. By row and column operations. f preserves volume. A may be changed to a matrix H where h1. n}. it is length. V and V + p have the same volume. Thus f (V ) and f (V + p) = f (V ) + f (p) have the same volume.n has rank t. Show that it is possible to select t rows and t columns of A such that the determined t × t matrix is invertible. it is area.e. Suppose A ∈ Rn and f : Rn → Rn is the homomorphism given by A. The volume of V does not change under translation. What is the dimension of the solution set of AX = 0? ¯ Deﬁnition If N and M are ﬁnite dimensional vector spaces and f : N → M is a linear transformation.t = 1 and all other entries are 0 (see the ¯ ¯ ﬁrst exercise on page 59). Thus if |A| = ±1. i. If f : F n → F m is given by a matrix A. In street language. where h is any automorphism on F n and g is any automorphism on F m . For example. This number is called the rank of A and is ≤ min{m.90 Linear Algebra Chapter 5 combination of those vectors. V might be the interior of a square or circle. The theorem for row rank follows using transpose. Theorem The n-dimensional volume of f (V ) is ±|A|(the n-dimensional volume of V ). Geometric Interpretation of Determinant Suppose V ⊂ Rn is some nice subset.. Theorem If A ∈ Fm. Exercise Suppose A has rank t. Thus row rank = t = column rank.n . then the rank of f is the same as the rank of the matrix A. Show that the rank of A is the largest integer t such that this is possible. For n = 1. This dimension is the same as the dimension of the image of g ◦ f ◦ h : F n → F m . Exercise Suppose A ∈ Fm. . elementary row and column operations change Proof neither the row rank nor the column rank. if n = 2.

Without this great simpliﬁcation. suppose h : R → R is diﬀerentiable and p is a real number. The area of V is any homomorphism f multiplies area by | f |. Of course. . y) ∈ V . y) : R2 → R2 be the homomorphism deﬁned by J(h)(x. science. 1] × · · ×[0.. vn of A. Then h is approximated near p by g(x) = h(p) + f (x − p) = h(p) + h (p)(x − p). Deﬁne the Jacobian by J(h)(x. p2 ). y). However this is not the only reason that linear algebra is so useful. The result follows because the determinant of the composition is the product of the determinants. the world of technology as we know it today would not exist. Corollary If P is the n-dimensional parallelepiped determined by the columns v1 . Proof Let V = [0. let f (x. It is a central fact that smooth phenomena may be approximated locally by linear phenomena. so they must be adjusted by a translation. then the n-dimensional volume of P is ±|A|. Linear functions approximate diﬀerentiable functions locally We continue with the special case F = R. and mathematics. y) ∈ V . As a simple example. Let f : R → R be the linear transformation f (x) = h (p)x. . Now suppose V ⊂ R2 is some nice subset and h = (h1 .) Theorem Suppose the determinant of J(h)(x. From the previous section we know that (x. 1] = {e1 t1 + · · +en tn : 0 ≤ ti ≤ 1}.Chapter 5 Linear Algebra 91 Proof If |A| = 0. p2 ) (after translation) by f (p1 . this theorem is immediate from the previous section. p2 ) ∈ V . the theorem is obvious. The student may now understand the following theorem from calculus. image(f ) has dimension < n and thus f (V ) has n-dimensional volume 0. y) is non-negative for each V V ∂h1 ∂x ∂h2 ∂x ∂h1 ∂y ∂h2 ∂y and for each 1dxdy. h2 ) : V → R2 is injective and diﬀerentiable. . Then the area of h(V ) is | J(h) | dxdy. h is approximated near (p1 . Then for any (p1 . Linear functions arise naturally in business. linear transformations send the origin to the origin. Then P = f (V ) = {v1 t1 + · · +vn tn : 0 ≤ ti ≤ 1}. If |A| = 0 then A is the product of elementary matrices (see page 59) and for elementary matrices. (Note that if h is the restriction of a linear transformation from R2 to R2 . y) = (x.

92

Linear Algebra The Transpose Principle

Chapter 5

We now return to the case where F is a ﬁeld (of arbitrary characteristic). F modules may also be called vector spaces and submodules may be called subspaces. The study of R-modules in general is important and complex. However the study of F -modules is short and simple – every vector space is free and every subspace is a summand. The core of classical linear algebra is not the study of vector spaces, but the study of homomorphisms, and in particular, of endomorphisms. One goal is to show that if f : V → V is a homomorphism with some given property, there exists a basis of V so that the matrix representing f displays that property in a prominent manner. The next theorem is an illustration of this. Theorem 1) Let F be a ﬁeld and n be a positive integer. Suppose V is an n-dimensional vector space and f : V → V is a homomorphism with |f | = 0. Then ∃ a basis of V such that the matrix ¯ representing f has its ﬁrst row zero. Suppose A ∈ Fn has |A| = 0. Then ∃ an invertible matrix C such that ¯ C −1 AC has its ﬁrst row zero. Suppose V is an n-dimensional vector space and f : V → V is a homomorphism with |f | = 0. Then ∃ a basis of V such that the matrix representing f has its ﬁrst column zero. Suppose A ∈ Fn has |A| = 0. Then ∃ an invertible matrix D such that ¯ D −1 AD has its ﬁrst column zero.

2) 3)

4)

We ﬁrst wish to show that these 4 statements are equivalent. We know that 1) and 2) are equivalent and also that 3) and 4) are equivalent because change of basis corresponds to conjugation of the matrix. Now suppose 2) is true and show 4) is true. Suppose |A| = 0. Then |At | = 0 and by 2) ∃ C such that C −1 At C has ¯ ¯ ﬁrst row zero. Thus (C −1 At C)t = C t A(C t )−1 has ﬁrst row column zero. The result follows by deﬁning D = (C t )−1 . Also 4) implies 2). This is an example of the transpose principle. Loosely stated, it is that theorems about change of basis correspond to theorems about conjugation of matrices and theorems about the rows of a matrix correspond to theorems about the columns of a matrix, using transpose. In the remainder of this chapter, this will be used without further comment.

Chapter 5

Linear Algebra

93

Proof of the theorem We are free to select any of the 4 parts, and we select part 3). Since | f |= 0, f is not injective and ∃ a non-zero v1 ∈ V with f (v1 ) = 0. ¯ Now v1 is independent and extends to a basis {v1 , .., vn }. Then the matrix of f w.r.t this basis has ﬁrst column zero. Exercise Let A = 3π 6 2π 4 . Find an invertible matrix C ∈ R2 so that C −1 AC
 

0 0 0   has ﬁrst row zero. Also let A =  1 3 4  and ﬁnd an invertible matrix D ∈ R3 2 1 4 so that D −1 AD has ﬁrst column zero. Exercise Suppose M is an n-dimensional vector space over a ﬁeld F , k is an integer with 0 < k < n, and f : M → M is an endomorphism of rank k. Show there is a basis for M so that the matrix representing f has its ﬁrst n − k rows zero. Also show there is a basis for M so that the matrix representing f has its ﬁrst n − k columns zero. Work these out directly without using the transpose principle. Nilpotent Homomorphisms In this section it is shown that an endomorphism f is nilpotent iﬀ all of its characteristic roots are 0 iﬀ it may be represented by a strictly upper triangular matrix. ¯ Deﬁnition An endomorphism f : V → V is nilpotent if ∃ m with f m = 0. Any ¯ f represented by a strictly upper triangular matrix is nilpotent (see page 56). Theorem Suppose V is an n-dimensional vector space and f : V → V is a nilpotent homomorphism. Then f n = 0 and ∃ a basis of V such that the matrix ¯ representing f w.r.t. this basis is strictly upper triangular. Thus the characteristic polynomial of f is CPf (x) = xn . Proof Suppose f = 0 is nilpotent. Let t be the largest positive integer with ¯ t f = 0. Then f t (V ) ⊂ f t−1 (V ) ⊂ ·· ⊂ f (V ) ⊂ V . Since f is nilpotent, all of these ¯ inclusions are proper. Therefore t < n and f n = 0. Construct a basis for V by ¯ starting with a basis for f t (V ), extending it to a basis for f t−1 (V ), etc. Then the matrix of f w.r.t. this basis is strictly upper triangular. Note To obtain a matrix which is strictly lower triangular, reverse the order of the basis.

94 Exercise

Linear Algebra

Chapter 5

Use the transpose principle to write 3 other versions of this theorem.

Theorem Suppose V is an n-dimensional vector space and f : V → V is a homomorphism. Then f is nilpotent iﬀ CPf (x) = xn . (See the exercise at the end of Chapter 4 for the case n = 2.) Proof Suppose CPf (x) = xn . For n = 1 this implies f = 0, so suppose n > 1. ¯ Since the constant term of CPf (x) is 0, the determinant of f is 0. Thus ∃ a basis ¯ ¯ of V such that the matrix A representing f has its ﬁrst column zero. Let B ∈ Fn−1 be the matrix obtained from A by removing its ﬁrst row and ﬁrst column. Now CPA (x) = xn = xCPB (x). Thus CPB (x) = xn−1 and by induction on n, B is nilpotent and so ∃ C such that C −1 BC is strictly upper triangular. Then
       

1 0 · ·0 0 · C −1 · 0

       

0 ∗ · ·∗     ·     B       · 0

       

0 ∗ · ·∗ 1 0 · ·0    0  0    =  · C −1 BC · C     ·  ·  0 0

       

is strictly upper triangular.

Suppose F is ﬁeld, A ∈ F3 is a strictly lower triangular matrix of a  0 0 0   rank 2, and B =  1 0 0 . Using conjugation by elementary matrices, show there 0 1 0 is an invertible matrix C so that C −1 AC = B. Now suppose V is a 3-dimensional vector space and f : V → V is a nilpotent endomorphism of rank 2. We know f can be represented by a strictly lower triangular matrix. Show there is a basis {v1 , v2 , v3 } for V so that B is the matrix representing f . Also show that f (v1 ) = v2 , f (v2 ) = v3 , and f (v3 ) = 0. In other words, there is a basis for V of the form {v, f (v), f 2 (v)} ¯ with f 3 (v) = 0. ¯ Exercise Exercise Suppose V is a 3-dimensional vector space and f : V → V is a nilpotent endomorphism of rank 1. Show there is a basis for V so that the matrix representing   0 0 0 f is  1 0 0 .   0 0 0

Chapter 5

Linear Algebra Eigenvalues

95

Our standing hypothesis is that V is an n-dimensional vector space over a ﬁeld F and f : V → V is a homomorphism. Deﬁnition An element λ ∈ F is an eigenvalue of f if ∃ a non-zero v ∈ V with f (v) = λv. Any such v is called an eigenvector. Eλ ⊂ V is deﬁned to be the set of all eigenvectors for λ (plus 0). Note that Eλ = ker(λI − f ) is a subspace of V . The ¯ next theorem shows the eigenvalues of f are just the characteristic roots of f . Theorem 1) 2) 3) If λ ∈ F then the following are equivalent. λ is an eigenvalue of f , i.e., (λI − f ) : V → V is not injective. | (λI − f ) |= 0. ¯ λ is a characteristic root of f , i.e., a root of the characteristic polynomial CPf (x) = | (xI − A) |, where A is any matrix representing f .

Proof It is immediate that 1) and 2) are equivalent, so let’s show 2) and 3) are equivalent. The evaluation map F [x] → F which sends h(x) to h(λ) is a ring homomorphism (see theorem on page 47). So evaluating (xI − A) at x = λ and taking determinant gives the same result as taking the determinant of (xI − A) and evaluating at x = λ. Thus 2) and 3) are equivalent. The nicest thing you can say about a matrix is that it is similar to a diagonal matrix. Here is one case where that happens. Theorem Suppose λ1 , .., λk are distinct eigenvalues of f , and vi is an eigenvector of λi for 1 ≤ i ≤ k. Then the following hold. 1) 2) {v1 , .., vk } is independent. If k = n, i.e., if CPf (x) = (x − λ1 ) · · · (x − λn ), then {v1 , .., vn } is a basis for V . The matrix of f w.r.t. this basis is the diagonal matrix whose (i, i) term is λi .

Proof Suppose {v1 , .., vk } is dependent. Suppose t is the smallest positive integer such that {v1 , .., vt } is dependent, and v1 r1 + · · +vt rt = 0 is a non-trivial linear ¯ combination. Note that at least two of the coeﬃcients must be non-zero. Now (f − λt )(v1 r1 + · · +vt rt ) = v1 (λ1 − λt )r1 + · · +vt−1 (λt−1 − λt )rt−1 + 0 = 0 is a shorter ¯ ¯

96

Linear Algebra

Chapter 5

non-trivial linear combination. This is a contradiction and proves 1). Part 2) follows from 1) because dim(V ) = n. Exercise 0 1 ∈ R2 . Find an invertible C ∈ C2 such that −1 0 C −1 AC is diagonal. Show that C cannot be selected in R2 . Find the characteristic polynomial of A. Let A =

Exercise Suppose V is a 3-dimensional vector space and f : V → V is an endomorphism with CPf (x) = (x − λ)3 . Show that (f − λI) has characteristic polynomial x3 and is thus a nilpotent endomorphism. Show there is a basis for V so that the      λ 0 0 λ 0 0 λ 0 0       matrix representing f is  1 λ 0 ,  1 λ 0  or  0 λ 0 . 0 1 λ 0 0 λ 0 0 λ We could continue and ﬁnally give an ad hoc proof of the Jordan canonical form, but in this chapter we prefer to press on to inner product spaces. The Jordan form will be developed in Chapter 6 as part of the general theory of ﬁnitely generated modules over Euclidean domains. The next section is included only as a convenient reference. Jordan Canonical Form This section should be just skimmed or omitted entirely. It is unnecessary for the rest of this chapter, and is not properly part of the ﬂow of the chapter. The basic facts of Jordan form are summarized here simply for reference. The statement that a square matrix B over a ﬁeld F is a Jordan block means that ∃ λ ∈ F such that B is a lower triangular matrix of the form

B= 
 

λ   1 λ  0

0 ·

· 1 λ

   .   

B gives a homomorphism g : F m → F m with g(em ) = λem

and g(ei ) = ei+1 + λei for 1 ≤ i < m. Note that CPB (x) = (x − λ)m and so λ is the only eigenvalue of B, and B satisﬁes its characteristic polynomial, i.e., CPB (B) = 0. ¯

e.) Theorem Jordan form (when it exists) is unique. This means that if A and D are similar matrices in Jordan form.e. iﬀ each Jordan block is a 1 × 1 matrix. . . Theorem 1) 2) If A ∈ Fn . namely that if A has n distinct eigenvalues in F .. (x − 2)(x − 3)(x − 4). they have the same Jordan blocks. C may be selected to be in Rn iﬀ all the eigenvalues of A are real. Then n1 + · · +nt = n and CPD (x) = (x − λ1 )n1 · ·(x − λt )nt .Chapter 5 Deﬁnition  Linear Algebra 97 A matrix D ∈ Fn is in Jordan form if ∃ Jordan blocks B1 . (x − 2)2 (x − 3)2 . The complex numbers are algebraically closed. D is a diagonal matrix iﬀ each ni = 1. Also note that we know one special case of this theorem. then A is similar to a diagonal matrix. Note that a diagonal matrix is a special case of Jordan form. i. ¯ ¯ . This means that CPA (x) will factor completely in C[x].. Exercise Find all real matrices in Jordan form that have the following characteristic polynomials: x(x − 2). (x − 2)(x − 3)2 . ∃ λ1 .. ∃ an invertible C ∈ Fn such that C −1 AC is in Jordan form. The reader should use the transpose principle to write three other versions of the ﬁrst theorem.. i. Show a0 I + a1 D + · · +D n = 0. the following are equivalent.. . except possibly in diﬀerent order. (In this case we say that all the eigenvalues of A belong to F . λn ∈ F (not necessarily distinct) such that CPA (x) = (x − λ1 ) · · (x − λn ).     that D =          Suppose D is of this form and Bi ∈ Fni has eigenvalue λi . then A is similar to a diagonal matrix. show CPD (D) = 0. Let’s look at the classical case A ∈ Rn . (x − 2)(x − 3)3 . and thus ∃ C ∈ Cn with C −1 AC in Jordan form. Exercise Suppose D ∈ Fn is in Jordan form and has characteristic polynomial a0 + a1 x + · · +xn . Bt such B1 B2 · 0 · Bt 0      . Later on it will be shown that if A is a symmetric real matrix. (x − 2)2 .

v ∈ V. However. the power and elegance of linear algebra become transparent for all to see. u2 . Thus ∃ an invertible C ∈ Fn such that D = C −1 AC is in Jordan form.) Inner Product Spaces The two most important ﬁelds for mathematics and science in general are the real numbers and the complex numbers. Finitely generated vector spaces over R or C support inner products and are thus geometric as well as algebraic objects. (Note how easy this is in Jordan form. v ∈ V and r1 . attention is restricted to the case F = R. ¯ Schwarz’ inequality. In the remainder of this chapter. the result is obvious. r2 ∈ R. for all u1 . Show A is nilpotent iﬀ An = 0 ¯ iﬀ CPA (x) = xn . for all u. An inner product (or dot product) on V is a function V × V → R which sends (u. If a or b is 0. Dividing by 2ab yields 0 ≤ ab ± (u · v) or | u · v |≤ ab. Now 0 ≤ (ua ± vb) · (ua ± vb) = (u · u)a2 ± 2ab(u · v)+ (v · v)b2 = b2 a2 ± 2ab(u · v) + a2 b2 .) ¯ Exercise Suppose A ∈ Fn is in Jordan form. (See the second exercise on page 66. (u · v)2 ≤ (u · u)(v · v). and both could have been treated here. If v ∈ V . 2) 3) Theorem 1) 2) √ √ Proof of 2) Let a = v · v and b = u · u. v) to u · v and satisﬁes 1) (u1 r1 + u2 r2 ) · v = (u1 · v)r1 + (u2 · v)r2 v · (u1 r1 + u2 r2 ) = (v · u1 )r1 + (v · u2 )r2 u·v = v·u u · u ≥ 0 and u · u = 0 iﬀ u = 0 ¯ Suppose V has an inner product. Suppose neither a nor b is 0. Assume the theorem that there is a ﬁeld F containing E such that CPA (x) factors completely in F [x]. f : V → R deﬁned by f (u) = u · v is a homomorphism. If u. Use this to show CPA (A) = 0. for all u ∈ V . for simplicity.98 Linear Algebra Chapter 5 Exercise (Cayley-Hamilton Theorem) Suppose E is a ﬁeld and A ∈ En . Deﬁnition Suppose V is a real vector space. v ∈ V . . The theories for the real and complex cases are quite similar. Thus 0 · v = 0.

Furthermore v1 vn Theorem Proof Suppose v1 r1 + · · +vn rn = 0. S = {e1 . vn }.j for 1 ≤ i.. en } will be called the canonical or standard orthonormal basis (see page 72). u · v = u v cos Θ where Θ is the angle between u . (Schwarz’ inequality) (The triangle inequality) 2 u+v ≤ u + v ..···. The second statement is transparent. Thus S is independent. IPS V. then S is independent. Convention Rn will be assumed to have the standard inner product deﬁned by (r1 . sn )t = r1 s1 + · · +rn sn . Then 0 = (v1 r1 + · · +vn rn ) · vi = ri (vi · vi ) ¯ and thus ri = 0. as is shown by the following theorem. It is easy to deﬁne an inner product. . 2 Proof of 4) 2 u v + v u + v 2 = (u + v) · (u + v) = u = ( u + v )2 . vn } is an orthogonal sequence of non-zero vectors in an vn v1 . vi · vj = δi.e. ... Theorem If u... . . j ≤ n. Theorem Suppose V is a real vector space with a basis S = {v1 . i. A sequence {v1 . ¯ vr = v | r |.. rn )t · (s1 .. It is given by the formula (v1 r1 + · · +vn rn ) · (v1 s1 + · · +vn sn ) = r1 s1 + · · +rn sn . vn } is orthogonal provided vi · vj = 0 when i = j.Chapter 5 Linear Algebra 99 Theorem √Suppose V has an inner product. | u·v |≤ u v . + 2(u · v) + v 2 ≤ u 2 + Deﬁnition An Inner Product Space (IPS) is a real vector space with an inner product. .. v ∈ Rn . is orthonormal. . Deﬁne the norm or length of a vector v by v = v · v. 1) 2) 3) 4) v = 0 iﬀ v = 0. If S = {v1 ... Then there is a unique inner product on V which makes S an orthornormal basis. The following properties hold.. Suppose V is an IPS.. The next theorem shows that this inner product has an amazing geometry. The sequence is orthonormal if it is orthogonal and each vector has length 1.

w . Linear Algebra Chapter 5 Proof Let u = (r1 . and v ∈ W −Y ... . . an ) ∈ R1.. v}. vk }. wk+1 } is an orthonormal basis for the subspace Thus if wk+1 = w generated by {w1 . If {w1 . wn }.e. i. Thus r1 s1 + · · +rn sn = u v cos Θ. ... rn ) and v = (s1 . v} is already orthonormal. Then (w ·wi ) = (v −w1 (v ·w1 )··−wk (v ·wk ))·wi = 0. v = w1 (v · w1 ) + · · +wn (v · wn ). . Now suppose b ∈ R and C =   ∈ Rn  ·  cn −1 has f (C) = b. .. So (r1 − s1 )2 + · · +(rn − sn )2 = r1 + · · +rn + s2 + · · 1 +s2 − 2 u v cos Θ. wk . Moreover. the subspace generated by {v1 . Suppose inductively that {w1 . v1 Proof Let w1 = . then {w1 . Then W has an orthonormal basis {w1 . . and this the set of all solutions to a1 (x1 − c1 ) + · · +an(xn − cn ) = 0.. wk }... sn ).. n Exercise This is a simple exercise to observe that hyperplanes in Rn are cosets. vn }.. Proof v = w1 r1 + · · +wn rn and v · wi = (w1 r1 + · · +wn rn ) · wi = ri Theorem Suppose W is an IPS.. By the law of cosines u − v 2 = 2 2 2 u + v 2 − 2 u v cos Θ... . wk . any orthonormal sequence in W extends to an orthonormal basis of W . Suppose f : Rn → R is a non-zero homomorphism given by a matrix A = (a1 . . . . wk } is an orthonormal v1 basis for Y . Theorem (Gram-Schmidt) Suppose W is an IPS with a basis {v1 . and let w = v −p(v).. wk+1 = v.... wn }.. Y ⊂ W is a subspace with an orthonormal basis {w1 .n .. the   c1  ·    set of all vectors perpendicular to A.. Then if v ∈ W .100 and v. Let w = vk+1 − p(vk+1 ) and .. Gram-Schmidt orthonormalization Theorem (Fourier series) Suppose W is an IPS with an orthonormal basis {w1 . Then f (b) is the set of all solutions to a1 x1 + · · +an xn = b which is the coset L + C. Then L = ker(f ) is the set of all solutions to a1 x1 + · · +an xn = 0. Deﬁne the projection of v onto Y by p(v) = w1 (v ·w1 )+·· +wk (v ·wk ).. . ...

As in the ﬁrst theorem of this section. So suppose T = {f (u1 ). .. vk+1 }. Then T is independent and thus T is a basis for V and thus f is an isomorphism (see the second theorem on page 79). w2 . Isometries Suppose each of U and V is an IPS.. It is that. Find the projection of (e1 + e2 ) onto ker(f ). and f : U → V is a homomorphism. {w1 .. 2. . Then by the previous theorem. wn }. 1.. f (un )} is an orthonormal sequence in V . un } is an orthonormal basis for U .. .. Theorem Suppose each of U and V is an n-dimensional IPS. wk+1 } is an orthonormal basis w for the subspace generated by {w1 . It is easy to check that f preserves inner products. (u1 · u2 )U = (f (u1 ) · f (u2 ))V .3). We now come to one of the deﬁnitive theorems in linear algebra.... u2 in U . wk } is an orthonormal sequence in W . This is a key observation for an exercise on page 103 showing O(n) is a deformation retract of GLn (R). 0. Proof Isometries certainly preserve orthonormal sequences. where p(v) is the projection of v onto Y . (1 − t)v + tw3 } w is a basis for W . 0)t . Exercise Find an orthonormal basis for the kernel of f . wk . {u1 . Now suppose W has dimension n and {w1 ..1. there is only one inner product space for each dimension.. . In this manner an orthonormal basis for W is constructed.. Then f is an isometry iﬀ {f (u1 ). and set w w3 = .. vk+1 . Find w3 and show that for any t with 0 ≤ t ≤ 1. The process above may be used to modify this to an orthonormal basis {w1 . . 3)t . .. v} where v = (1.. W is generated by the sequence {w1 .. {w1 .. 0)t and w2 = (0. Find the angle between e1 + e2 and the plane ker(f ). w2 .Chapter 5 wk+1 = Linear Algebra 101 w . . . Since this sequence is independent. Notice that this construction deﬁnes a function h which sends a basis for W to an orthonormal basis for W (see topology exercise on page 103). Exercise Let W = R3 have the standard inner product and Y ⊂ W be the subspace generated by {w1 .. Let f : R3 → R be the homomorphism deﬁned by the matrix (2. w2 } where w1 = (1. wk . up to isometry. f (un )} is an orthonormal sequence in V . A homomorphism f : U → V is said to be an isometry provided it is an isomorphism and for any u1 . it extends to a basis {w1 . . vn }. let w = v − p(v). .

f is an isometry. Thus O(n) is a multiplicative subgroup of GLn (R). Proof There exist orthonormal bases {u1 . In particular. Orthogonal Matrices As noted earlier. We now wish to study isometries from Rn to Rn .1. and so certainly preserve volume.e. vn } for V . Proof A left inverse of a matrix is also a right inverse (see the exercise on page 64). and is called the orthogonal group. At A = I. Exercise Let f : R3 → R be the homomorphism deﬁned by the matrix (2..... Theorem 1) 2) If A is orthogonal. and thus preserve angle and distance. AAt = I. AC is orthogonal. i.102 Linear Algebra Chapter 5 Theorem Suppose each of U and V is an n-dimensional IPS. i. Thus 1) and 2) are equivalent because each of them says A is invertible with A−1 = At . If A is orthogonal. 1) and 3) are equivalent. The set of all such A is denoted by O(n). Theorem Suppose A ∈ Rn and f : Rn → Rn is the homomorphism deﬁned by f (B) = AB. and f (ei ) is column i of A. Now {e1 . Isometries preserve inner product. and by the previous theorem.. . linear algebra is not so much the study of vector spaces as it is the study of endomorphisms. Thus by the previous section. Then the following are equivalent. U is isometric to Rn with its standard inner product. .e.3). . en } is the canonical orthonormal basis for Rn . Deﬁnition If A ∈ Rn satisﬁes these three conditions.. un } for U and {v1 . 1) 2) 3) The columns of A form an orthonormal basis for Rn . there exists a homomorphism f : U → V with f (ui ) = vi .. A−1 is orthogonal. A is said to be orthogonal.. Find a linear transformation h : R2 → R3 which gives an isometry from R2 to ker(f ). f is an isometry. . By the ﬁrst theorem on page 79. | A |= ±1. If A and C are orthogonal. We know from a theorem on page 90 that an endomorphism preserves volume iﬀ its determinant is ±1. Then ∃ an isometry f : U → V. The rows of A form an orthonormal basis for Rn .

v ∈ Rn . (See the exercise on page 56. u − v = f (u)−f (v) and the angle between u and v is equal to the angle between f (u) and f (v). The next theorem is just an exercise using the previous theorem. This means that if u. z ∈ Rn . we ﬁrst note that symmetric is the same as self-adjoint. Proof Part 1) follows from |A|2 = |A| |At | = |I| = 1. Our goals are to prove that. and matrix multiplication is associative. t) = (1 − t)A + th(A) is a deformation retract of GLn (R) to O(n). Theorem A is symmetric iﬀ A is self-adjoint. then A = 2 cosΘ −sinΘ sinΘ cosΘ for some number Θ. Let h : GLn (R) → O(n) be deﬁned by Gram-Schmidt. This means a sequence of matrices {Ai } converges to A iﬀ it converges coordinatewise. As background. . Then u − v 2 = (u − v) · (u − v) = f (u − v) · f (u − v) = f (u − v) 2 = f (u) − f (v) 2 . Diagonalization of Symmetric Matrices We continue with the case F = R. Then f preserves distances and angles. For part 3) assume f : Rn → Rn is an isometry. if A is a symmetric matrix. Show GLn (R) is an open subset and O(n) is closed and compact. is the matrix product y t z. A is said to be self-adjoint if (Au)·v = u·(Av) for all u. 1] → GLn (R) deﬁned by H(A. Then (At u) · v = u · (Av). Show H : GLn (R) × [0. A is said to be symmetric provided At = A. Proof If y. Deﬁnition Suppose A ∈ Rn . all of its eigenvalues are real and that ∃ an orthogonal matrix C such that C −1 AC is diagonal.Chapter 5 3) Linear Algebra 103 Suppose A is orthogonal and f is deﬁned by f (B) = AB. The proof that f preserves angles follows from u · v = u v cosΘ. because isometries clearly form a subgroup of the multiplicative group of all automorphisms. Thus (At u) · v = (ut A)v = ut (Av) = u · (Av). Exercise Show that if A ∈ O(2) has |A| = 1. Part 2) is immediate. v ∈ Rn . v ∈ Rn . then the dot product y · z. Note that any diagonal matrix is symmetric.) Exercise (topology) Let Rn ≈ Rn have its usual metric topology. Theorem Suppose A ∈ Rn and u.

λn (not necessarily distinct) such that CPA (x) = (x − λ1 )(x − λ2 ) · · · (x − λn ). The proof then ¯ ¯ · v). The next theorem has geometric and physical implications.t. namely that they are perpendicular.r. S. just the incredibility of it all will suﬃce. reads as λ(v · v) = (λv · v) = (Av · v) = (v · Av) = (v · λv) = λ(v λ is a real number.. Theorem Suppose A ∈ Rn and C ∈ O(n). Review Suppose A ∈ Rn and f : Rn → Rn is deﬁned by f (B) = AB. its conjugate is deﬁned by µ = a − bi. but for us. For symmetric matrices. Then (C −1 AC)t = C t A(C −1 )t = C −1 AC. vn } be another basis and C ∈ Rn be the matrix with vi as column i. Now S is an orthonormal basis iﬀ C is an orthogonal matrix. We know that eigenvectors belonging to distinct eigenvalues are linearly independent. Then A represents f w. Then u · v = 0. the canonical orthonormal basis. Now suppose λ is a complex eigenvalue of A symmetric matrix. Proof We know CPA (x) factors into linears over C. Proof λ1 (u · v) = (Au) · v = u · (Av) = λ2 (u · v). Then C −1 AC is the matrix representing f w. Then ∃ real numbers λ1 .j ) is a complex matrix or vector.. Then λ(v t v ) = (λv)t v = (Av)t v = ¯ ¯ ¯ t t t t ¯ t v ). If h : C → C is deﬁned by h(µ) = µ.. Let S = {v1 . ¯ ¯ then h is a ring isomorphism which is the identity on R. Summary Representing f w. Since A ∈ Rn is a real ¯ a t ¯t . Thus λ = λ and λ ∈ R. and Au = λ1 u and Av = λ2 v. That is.r. an orthonormal basis is the same as conjugating A by an orthogonal matrix.t. . Either way.j ). If w = (ai. Or ¯ (v A)¯ = v (A¯) = v (Av) = v (λv) = λ(v ¯ v v you can deﬁne a complex inner product on Cn by (w · v) = w t v .104 Linear Algebra Chapter 5 Theorem Suppose A ∈ Rn is symmetric. . Then A is symmetric iﬀ C −1 AC is symmetric.. If µ = a + bi is a complex number. A = A = A and v ∈ Cn is an eigenvector with Av = λv. Proof Suppose A is symmetric. its conjugate is deﬁned by w = (¯i. λ1 . λ2 ∈ R are distinct eigenvalues of A. Theorem Suppose A is symmetric. we show more. . all the eigenvalues of A are real.t.r.

and suppose A is a symmetric n × n matrix.. the   λ1 · · transformation determined by A is represented by Since this is a symmetric matrix.. 1) 2) f is self-adjoint.Chapter 5 Theorem 1) 2) Linear Algebra If A ∈ Rn .. Theorem If V is an n-dimensional IPS and f : V → V is a linear transformation. v2 } be an orthonormal basis for R2 with Av1 = λv1 . Let λ be an eigenvalue for A and {v1 . then the following are equivalent. . because then there is a basis of eigenvectors of length 1.               λk      (B)     . the transformation determined λ b by A is represented by . . Suppose A is a symmetric 2 × 2 matrix.t this basis. They may be extended to an orthonormal basis v1 . . b = 0. Show 1) ⇒ 2)... k ≤ n. vn . A is symmetric. and they must form an orthonormal basis. λk with each vi = 1. If k = n. . So suppose k < n. Thus conjugating I 0 makes the entire matrix diagonal. a linear transformation f : V → V is said to be self-adjoint provided (u·f (v)) = (f (u)·v) for all u.. 105 Proof By the previous theorem. vk be eigenvectors for λ1 .     (0) (D) . Let v1 .. λk the distinct eigenvalues of A... 2) ⇒ 1). ∃ C ∈ O(n) such that C −1 AC is diagonal. With respect to this basis..r. By induction.. Denote by λ1 . . ∃ an orthonormal basis {v1 . by 0 C This theorem is so basic we state it again in diﬀerent terminology. the proof is immediate. Then w. 0 d Now suppose by induction that the theorem is true for symmetric matrices in Rt for t < n. vn } for V with each vi an eigenvector of f . ∃ an orthogonal C such that C −1 DC is diagonal. Since this matrix is symmetric. If V is an IPS. B = 0 and D is a symmetric matrix of smaller size. v ∈ V . the following are equivalent.

i. 2 Exercise Suppose A. Thus if A ∈ Rn . Exercise Deﬁne E : Cn → Cn by E(A) = eA = I + A + (1/2!)A2 + ··. ∃ a non-zero matrix N ∈ R2 with eN = I. Furthermore E(At ) = E(A)t . an algebraic object. Find an orthogonal C such that C −1 AC is diagonal. This shows that V . and this metric will determine that same topology. | eA |= 1 iﬀ trace(A) = 0. and if C is invertible. if V has an inner product. ∃ an orthogonal C such that D = C −1 AC. then | eA |= etrace(A) . ¯ If A ∈ Rn and At = −A. then E(A + B) = E(A)E(B). We know that V is isomorphic to Rn . Show that h is continuous. Suppose f and g are isomorphisms from V to Rn and A is a subset of V . then eN = I iﬀ N = 0. D ∈ Rn are symmetric. E(C −1 AC) = C −1 E(A)C. assume the Jordan form. if A and D are similar. If N ∈ Rn is symmetric. assume any A ∈ Cn is similar to a lower triangular matrix. Now use the results of this section to prove the statements below. I = E(0) = E(A − A) = E(A)E(−A). Show that f (A) is an open subset of Rn iﬀ g(A) is an open subset of Rn . Exercise Suppose V is an n-dimensional real vector space. This series converges and thus E is a well deﬁned function. 2) 3) 4) . If AB = BA. suppose V and W are ﬁnite-dimensional real vector spaces and h : V → W is a linear transformation. then eA ∈ O(n). it automatically has a metric.106 Linear Algebra Chapter 5 Exercise Let A = Do the same for A = 2 2 2 1 2 . Of course. Since A and −A commute. 2 1 . (For part 1. and ¯ thus E(A) is invertible with E(A)−1 = E(−A). Under what conditions are A and D similar? Show that. Finally.) 1) If A ∈ Cn .. has a god-given topology.e.

This always holds if F = C. then M is the sum of cyclic modules. i. Thus if M is torsion free.. i. enough material is added to form a basic ﬁrst year graduate course. Suppose R is a commutative ring. 107 . The ﬁnal section gives the fundamentals of dual spaces. everything is right down to the nub.Chapter 6 Appendix The ﬁve previous chapters were designed for a year undergraduate course in algebra. Now F [x] is a Euclidean domain and so VF [x] is the sum of cyclic modules. An R-module M is said to be cyclic if it can be generated by one element. If the characteristic polynomial of T factors into the product of linear polynomials. so this is a powerful concept. then there is a basis for V so that the matrix representing T is in Jordan canonical form. The basic theorem of this chapter is that if R is a Euclidean domain and M is a ﬁnitely generated R-module. M ≈ R/I where I is an ideal of R. The style is the same as before.. In the chapter on matrices. ﬁnitely generated abelian groups are the sums of cyclic groups – one of the jewels of abstract algebra.e. In this appendix. There is a basis for V so that the matrix representing T is in Rational canonical form. A matrix in Jordan form is a lower triangular matrix with the eigenvalues of T displayed on the diagonal. Since Z is a Euclidean domain. Now suppose F is a ﬁeld and V is a ﬁnitely generated F -module.e. These are independent sections added on at the end. is given in this chapter. A proof of this. This classical and very powerful technique allows an easy proof of the canonical forms. it is stated without proof that the determinant of the product is the product of the determinants. it is a free R-module. The organization is mostly a linearly ordered sequence except for the last two sections on determinants and dual spaces. then V becomes an F [x]-module by deﬁning vx = T (v). If T : V → V is a linear transformation. which depends upon the classiﬁcation of certain types of alternating multilinear forms. Two of the main goals are to characterize ﬁnitely generated abelian groups and to prove the Jordan canonical form.

A2 . Deﬁnition Suppose R is a ring and A1 .. ∃ a1 ∈ A1 and a2 ∈ A2 with a1 + a2 = 1. . then c = c(a1 + a2 ) ∈ A1 A2 .. Note that the sum and product of ideals are ideals and A1 A2 · · · Am ⊂ (A1 ∩ A2 ∩ · · · ∩ Am ). it is the image of r1 b1 +r2 b2 +···+rn bn = r1 (1 −a1 )+r2 (1 −a2 )+···+rn (1 −an ). 0). Chinese Remainder Theorem Suppose A1 . ∃ a ∈ A and b ∈ B with a + b = 1. If A and B are ideals of a ring R. Show A1 ∩ A2 ⊂ A1 A2 .. 0. . ¯ ¯ ¯ Theorem If R is commutative and A1 . with each Ai = R..108 Appendix The Chinese Remainder Theorem Chapter 6 On page 50 in the chapter on rings. then the following are equivalent... Am and B are ideals of R with Ai and B comaximal for each i. Proof for n = 2.. ¯ If c ∈ A1 ∩ A2 . Surprisingly... Proof There exists ai ∈ Ai and bi ∈ A1 A2 · · · Ai−1Ai+1 · · · An with ai + bi = 1. A2 . . A2 .... r2 + A2 . Proof Consider π : R → R/B. If (r1 + A1 . In this section this classical topic is presented in full generality. 0. Am are ideals of R... Note ¯ that π(bi ) = (0. Theorem If A1 . then A1 A2 · · · Am and B are comaximal. Deﬁnition Theorem 1) 2) 3) Ideals A and B of R are said to be comaximal if A + B = R. . The product A1 A2 · · · Am is the set of all ﬁnite sums of elements a1 a2 · · · am with ai ∈ Ai . 1i . . . Thus A1 ∩ A2 ∩ · · · ∩ Am and B are comaximal. ¯ π(A) = R/B where π : R → R/B is the projection. An are pairwise comaximal ideals of R.. Then the natural map π : R → R/A1 × R/A2 × · · · × R/An is a surjective ring homomorphism with kernel A1 ∩ A2 ∩ · · · ∩ An .. . .. Then π(A1 A2 · · · Am ) = π(A1 )π(A2 ) · · · π(Am ) = (R/B)(R/B) · · · (R/B) = R/B. An are pairwise comaximal ideals of R. then A1 A2 · · · An = A1 ∩ A2 ∩ · · · ∩ An . the theorem holds even for non-commutative rings. Then the sum A1 + A2 + · · · + Am is the set of all a1 + a2 + · · · + am with ai ∈ Ai . rn + An ) is an element of the ¯ range. A and B are comaximal. A2 .. the Chinese Remainder Theorem was proved for the ring of integers..

then a or b ∈ I. Note that ∼ is an equivalence relation. Later on it will be shown that every Euclidean domain is a principle ideal domain. I is maximal means I = R and there are no ideals properly between I and R. then ∃ a maximal ideal I of R with a ∈ I. the student should work the exercise in group theory at the end of this section (see page 114). If a ∼ b. b ∈ R. Theorem 0 is a prime ideal of R iﬀ R is ¯ 0 is a maximal ideal of R iﬀ R is ¯ Suppose J ⊂ R is an ideal. The ideal V = Vt does not contain 1 and ¯ t∈T thus is not equal to R. This collection contains a maximal monotonic collection {Vt }t∈T . Note To properly appreciate this proof. Deﬁnition Suppose R is a domain and a. Thus every Euclidean domain is a unique factorization domain. then a . Then we say a ∼ b iﬀ there exists a unit u with au = b. J = R. Deﬁnition Suppose R is a commutative ring and I ⊂ R is an ideal. If a ∈ R is not a unit.Chapter 6 Appendix Prime and Maximal Ideals and UFDs 109 In the ﬁrst chapter on background material. I is prime means I = R and if a. Here it will be shown that this property holds for any principle ideal domain. J is a prime ideal iﬀ R/J is J is a maximal ideal iﬀ R/J is Maximal ideals are prime. it was shown that Z is a unique factorization domain. Consider {J : J is an ideal of R containing a with J = R}. Theorem Corollary Proof Theorem Every ﬁeld is a domain. Proof This is a classical application of the Hausdorﬀ Maximality Principle. b ∈ R have ab ∈ I. Therefore V is equal to some Vt and is a maximal ideal containing a.

Note also ∃ a unit u and primes p1 . Note a is prime ⇒ a is irreducible. while the only ¯ associate of 0 is 0 itself. a is prime if it generates a prime ideal. Theorem equivalent. it is called a prime element. ¯ ¯ If F is a ﬁeld and g ∈ F [x] is a non-zero polynomial.110 and b are said to be associates. . Deﬁnition 1) 2) Suppose R is a domain and a ∈ R is a non-zero non-unit. If an element of a domain generates a non-zero prime ideal. 1) 2) 3) Suppose R is a domain and a. then au is irreducible (prime). b ∈ (R − 0). Then the following are ¯ a ∼ b. An element a divides b (a|b) if ∃! c ∈ R with ac = b. . Note If a is a prime and a|c1 c2 · · · cn . then a|ci for some i. then a is irreducible (prime) iﬀ b is irreducible (prime). then its associates are n and −n. aR = bR. Note If a ∼ b. then a ∼ ci for some i. Parts 1) and 3) above show there is a bijection from the associate classes of R to the principal ideals of R. Thus if R is a PID.. if d = b1 b2 · · · bn = c1 c2 · · · cm with each bi and each ci prime. This is immediate from the deﬁnitions. then n = m and for some permutation σ of the indices. . a is irreducible if it does not factor. . This follows from the deﬁnition and induction on n. i. p2 .e. i. bi and cσ(i) are associates for every i. a = bc ⇒ b or c is a unit. Theorem Factorization into primes is unique up to order and associates. The following theorem is elementary. a|b and b|a. In other words. then the associates of g are all cg where c is a non-zero constant. the associates of 1 are the units of R. if a is irreducible (prime) and u is a unit.e. Appendix Chapter 6 Examples If R is a domain..e. pt where no two are associates and du = ps1 ps2 · · · pst . i.. If n ∈ Z is not zero. there is a bijection from the associate classes of R to the ideals of R. t 1 2 . but it shows how associates ﬁt into the scheme of things. If each cj is irreducible. a|bc ⇒ a|b or a|c.

then a factors into a ﬁnite product of irreducibles. Then the following are equivalent. Suppose R is a UFD. Proof We already know 1) ⇒ 2).e. Theorem Suppose R is a PID and a ∈ R is non-zero non-unit. R is a UFD. a is a prime element. Part 2) ⇒ 1) because factorization into primes is always unique. i. that one of these irreducibles is an associate of a. Since b divides a. If either Proof b or c is a unit or is zero. a is irreducible. Proof Every maximal ideal is a prime ideal.. aR is a prime ideal. 1) 2) 3) aR is a maximal ideal. It follows from the uniqueness of the factorization of ad = bc. Deﬁnition R is a factorization domain (FD) means that R is a domain and if a is a non-zero non-unit element of R. the element b is a unit or an associate of a. ∃ b ∈ R with I = bR. Fortunately. . Then the following are equivalent. This means I = R or I = aR.Chapter 6 Proof Appendix 111 This follows from the notes above. then R is a UFD iﬀ each irreducible element generates a prime ideal. If I is an ideal containing aR. principal ideal domains have this property. so 2) ⇒ 3). Now suppose a is irreducible and show aR is a maximal ideal.. Therefore the element a is a prime. Each of b and c factors as the product of irreducibles and the product of these products is the factorization of bc. elements factor as the product of primes. then a divides one of them. and thus a|b or a|c. then a is irreducible ⇔ a is prime. Theorem If R is a UFD and a is a non-zero non-unit of R. Every prime element is an irreducible element. Deﬁnition R is a unique factorization domain (UFD) means R is a FD in which factorization is unique (up to order and associates). as seen in the next theorem. so suppose each of b and c is a non-zero non-unit element of R. There exists an element d with ad = bc. i. Thus in a UFD. so 1) ⇒ 2).e. Every irreducible element of R is prime. Theorem 1) 2) Suppose R is a FD. and a|bc. a is an irreducible element of R. If R is a FD. This is a revealing and useful theorem. a irreducible ⇔ a is prime.

Since R is Noetherian.. i. ∃ t0 ≥ 1 such that It = It0 for all t ≥ t0 . The element c must be reducible. Each of aR and bR properly contains cR. Let I be an ideal of R.112 Appendix Chapter 6 Our goal is to prove that a PID is a UFD. nor need it be a maximal ideal of the ring R. Theorem A Noetherian domain is a FD. We shall see below that this is a useful concept which ﬁts naturally into the study of unique factorization domains. Using the two theorems above. The proof will not require that ideals be principally generated. Having three deﬁnitions makes this property useful and easy to use. an } ⊂ R such that I = a1 R + a2 R + · · · + an R. It need not contain all the ideals of the collection. then J = I. and thus 1) is true. Proof Suppose 1) is true and show 3). a2 . . each ideal of R is ﬁnitely generated.) Deﬁnition If R satisﬁes these properties. and it must be I itself. see the next theorem. and consider the collection of all ﬁnitely generated ideals contained in I.. Proof Suppose there is a non-zero non-unit element that does not factor as the ﬁnite product of irreducibles. but we slide over that.. ∃ a maximal one cR. is a monotonic sequence of ideals. a PID is a FD.. ... Any non-void collection of ideals of R contains an ideal I which is maximal in the collection.) If I1 ⊂ I2 ⊂ I3 ⊂ . with each properly contained in the next.. Thus 3) is true. i. so suppose 2) is false and show 3) is false. it only remains to show that a PID is a FD. . Theorem 1) 2) Suppose R is a commutative ring. We now have 2)⇒1)⇒3). and therefore 3) is false. By 2) there is a maximal one. (Actually this construction requires the Hausdorﬀ Maximality Principle or some form of the Axiom of Choice. (The ideal I is maximal only in the sense described. . and so each . Thus it is possible to construct a sequence of ideals I1 ⊂ I2 ⊂ I3 . . For example. This turns out to be equivalent to the property that any collection of ideals has a “maximal” element. Consider all ideals dR where d does not factor.e. is ﬁnitely generated and ∃ t0 ≥ 1 such that It0 contains those generators. Then the following are equivalent. 3) If I ⊂ R is an ideal. This property is satisﬁed by many of the classical rings in mathematics. In particular. Now suppose 2) is true and show 1). The ideal I = I1 ∪ I2 ∪ . R is said to be Noetherian. c = ab where neither a nor b is a unit. This means if J is an ideal in the collection with J ⊃ I. but only that they be ﬁnitely generated. ∃ a ﬁnite set {a1 .e. So there is a collection of ideals of R such that any ideal in the collection is properly contained in another ideal of the collection. or it is said to satisfy the ascending chain condition.

. Bourbaki. Corollary A PID is a UFD.2 of An Introduction to Complex Analysis in Several Variables by L. m ∈√ Z}. In particular 2 · 2 = (1 − 5) · (−1 − 5) are two distinct irreducible factorizations of 4. xn ]] is a UFD.6. which is not a domain. ...) Domains With Non-unique Factorizations Next are presented two of the standard examples of Noetherian domains that are not unique factorization domains.. (This follows immediately from the deﬁnition. Theorem If R is a UFD then R[x1 . Show R is isomorphic to Z[x]/(x2 − 5). which is a contradiction. o Theorem Suppose R is a commutative ring. Thus if F is a ﬁeld. This gives a ﬁnite factorization of c into irreducibles. xn ]] is a UFD.... xn ]] are Noetherian.. √ √ Exercise Let R = Z( 5) = {n + m 5 : n.. So Z is a UFD and if F is a ﬁeld. (There is a UFD R where R[[x]] is not a UFD.) Theorem Germs of analytic functions on Cn form a UFD. then the formal power series R[[x1 .. F [[x1 .. H¨rmander. See page 566 of Commutative Algebra by N. Thus if F is a ﬁeld. Show that R is a subring of √ R which is not a UFD. . xn ] and R[[x1 . . then R/I is Noetherian. Proof See Theorem 6. F [x] is a UFD.Chapter 6 Appendix 113 of a and b factors as a ﬁnite product of irreducibles.. where (x2 − 5) represents the ideal (x2 − 5)Z[x]. and R/(2) is isomorphic to Z2 [x]/(x2 − [5]) = Z2 [x]/(x2 + [1]).) If R is a PID. This and the previous theorem show that Noetherian is a ubiquitous property in ring theory. .) Theorem If R is Noetherian and I ⊂ R is a proper ideal.. (This is the famous Hilbert Basis Theorem. Then R is Noetherian ⇒ R[x1 . You see the basic structure of UFDs is quite easy... F [x1 .. .. xn ] is a UFD. which are stated here only for reference. . xn ] is a UFD. . (This theorem goes all the way back to Gauss.. It takes more work to prove the following theorems.

In particular. and g : B → C is a surjective homomorphism with kernel K. z] → R[x. and f (z) = y 2 . Show x2 − yz is irreducible and thus prime in R[x. An easier ¯ x approach is to let f : R[x. Show R ¯ is not a UFD. Show that H is maximal iﬀ G/H ≈ Zp for some prime p. K must be a summand of B. y. a subgroup H of G is said to be maximal if H = G and there are no subgroups properly between H and G. In particular x · x = y · z are two distinct irreducible factorizations ¯ ¯ ¯ ¯ of x2 . g has a right inverse. i. Splitting Short Exact Sequences Suppose B is an R-module and K is a submodule of B. Q contains no maximal subgroups. (h is called a splitting map. Then the following are equivalent. Now suppose 2) is true and h : C → B is a right inverse of g. f (y) = x2 .. then there is a maximal subgroup H of Q which contains a. If u ∈ R[x. Then h is injective.) Proof Suppose 1) is true. 1) 2) K is a summand of B. then h deﬁned by h = i ◦ (g|L)−1 is a right inverse of g. let u ∈ R be the coset containing u. y. z]/(yz). This is used below to show that if R is a PID. and y 2 are irreducible in S and (xy)(xy) = (x2 )(y 2 ) are two distinct irreducible factorizations of (xy)2 in S. y] be the ring homomorphism deﬁned by f (x) = xy. If i : L → B is inclusion. x2 .e. consider the case G = Q.e. i. . Note that xy. ∃ a homomorphism h : C → B with g ◦ h = I : C → C. When is K a summand ¯ of B? It turns out that K is a summand of B iﬀ there is a splitting map from B/K to B. z]. ¯ Thus K ⊕ h(C) = B. then every submodule of Rn is free. Which one of the following is true? 1) 2) If a ∈ Q. Then (g|L) : L → C is an isomorphism. Exercise In Group Theory If G is an additive abelian group. Theorem 1 Suppose R is a ring. y. x2 . In this case we write K ⊕ L = B. B and C are R-modules. For simplicity. y. As deﬁned in the chapter on linear algebra. Show R/(¯) is isomorphic to R[y.. suppose ∃ a submodule L of B with K ⊕ L = B.114 Appendix Chapter 6 Exercise Let R = R[x. Then S = R[xy. K is a summand of B provided ∃ a submodule L of B with K + L = B and K ∩ L = 0. K + h(C) = B and K ∩ h(C) = 0. z]. which is not a domain. if B/K is free. y 2 ] is the image of f and S is isomorphic to R. z]/(x2 − yz).

Chapter 6 Appendix 115 Deﬁnition Suppose f : A → B and g : B → C are R-module homomorphisms.e. If C is a free R-module. Although this theorem is transparent. iﬀ B → C has a splitting map.s) means f is injective. is A → A ⊕ C → C where f = i1 and g = π2 . A short exact sequence is said to split if ∃ ≈ an isomorphism B → A ⊕ C such that the following diagram commutes. Showing these properties are equivalent to the splitting of the sequence is a good exercise in the art of diagram chasing. f g The statement that 0 → A → B → C → 0 is a short exact sequence (s.1 A short exact sequence 0 → A → B → C → 0 splits iﬀ f (A) is a summand of B. Proof From the previous theorem we know this is true for n = 1. Suppose n > 1 and the theorem is true for submodules of Rn−1 . Proof We know from the previous theorem f (A) is a summand of B iﬀ B → C has a splitting map. g is surjective and f (A) = ker(g). . Thus subgroups of Zn are free Z-modules of dimension ≤ n. Now suppose C has a free basis T ⊂ C. 0→ A f   E B c g &2 π E & b & & C →0 i 1  ~  ≈ A⊕C & & We now restate the previous theorem in this terminology. Suppose A ⊂ Rn is a submodule. The canonical split s.e. then A is a free R-module of dimension ≤ n. and g : B → C is surjective. This theorem restates the ring property of PID as a module property. there is a splitting map and thus the sequence splits. There exists a function h : T → B such that g ◦ h(c) = c for each c ∈ T . Theorem 2 1) 2) If R is a domain.s. then the following are equivalent. Theorem 1. The function h extends to a homomorphism from C to B which is a right inverse of g. Theorem 3 If R is a PID and A ⊂ Rn is a submodule. R is a PID. 1)⇒2) is a precursor to the following classical result. Every submodule of RR is a free R-module of dimension ≤ 1.

and is sometimes omitted from the deﬁnition. it is free of dimension 1 and thus the sequence splits by Theorem 1. r ∈ R such that a = bq + r with r = 0 or φ(r) < φ(b). is used only in Theorem 2. ¯ ¯ F [x] where F is a ﬁeld with φ(f = a0 + a1 x + · · · + an xn ) = deg(f ). If π(A) = 0.116 Appendix Chapter 6 Consider the following short exact sequences. This is one of the great classical theorems of abstract algebra. ¯ In either case. ¯ Examples of Euclidean Domains Z with φ(n) = |n|. Furthermore the ﬁrst axiom. where f : R n−1 → Rn−1 ⊕ R is inclusion and g = π : Rn−1 ⊕ R → R is the projection. φ(a) ≤ φ(ab).s. b ∈ Z} = Gaussian integers with φ(a + bi) = a2 + b2 .1. Here N will denote the set of all non-negative integers. and the eﬃciency of this abstraction is displayed in this section. Show A ×3 is a free Z-module of dimension 1. ¯ If π(A) = 0. then ¯ 1) φ(a) ≤ φ(ab). and you don’t have to worry about it becoming obsolete. (16. 2) ∃ q. 24). Deﬁnition A domain R is a Euclidean domain provided ∃ φ : (R − 0) −→ N such ¯ that if a. Z[i] = {a + bi : a. Exercise Let A ⊂ Z2 be the subgroup generated by {(6. b ∈ (R − 0). Euclidean Domains The ring Z possesses the Euclidean algorithm and the polynomial ring F [x] has the division algorithm (pages 14 and 45). then M is the sum of cyclic modules. Also show the s. Anyway it is possible to just play around with matrices and get some deep results. A is a free submodule of dimension ≤ n. not just the set of positive integers. 0 −→ Rn−1 −→ Rn−1 ⊕ R −→ R −→ 0 0 −→ A ∩ Rn−1 −→ A −→ π(A) −→ 0 By induction. 64)}. Z4 −→ Z12 −→ Z3 splits ×2 ×2 but Z −→ Z −→ Z2 and Z2 −→ Z4 −→ Z2 do not (see top of page 78). A ∩ Rn−1 is free of dimension ≤ n − 1. then A ⊂ Rn−1 . The concept of Euclidean domain is an abstraction of these properties. f π . If R is a Euclidean domain and M is a ﬁnitely generated R-module. A ﬁeld F with φ(a) = 1 ∀ a = 0 or with φ(a) = 0 ∀ a = 0.e.

If E ∈ Rn . ¯ ¯ Theorem 2 If R is a Euclidean domain and a. If E is invertible. This means that row and column operations on A do not change the ideal I. This is a good exercise. and this will turn out to be the d1 displayed in the theorem.j ). . Since R is a PID. Proof The following remarkable theorem is the foundation for the results of this section. . However. if E ∈ Rt is invertible and J is the ideal generated by the elements of AE. . then ∃ b ∈ I − 0 satisfying φ(b) ≤ φ(a) ∀ a ∈ I − 0. then by elementary row and column operations (ai. Now r ∈ I and ¯ r = 0 ⇒ φ(r) < φ(b) which is impossible. 117 Proof If I is a non-zero ideal. r with a = bq + r. . row and column operations on (ai. dm 0 0 0              0 where each di = 0. However it is unnecessary for Theorem 3 below.j ) ∈ Rn.t is a non-zero matrix.j ). Theorem 3 If R is a Euclidean domain and (ai. b ∈ R − 0. then R is a PID and thus a UFD.Chapter 6 Theorem 1 Appendix If R is a Euclidean domain. ¯ ¯ Then b generates I because if a ∈ I − 0.j ) may produce elements with smaller . In the same manner. Proof Let I ⊂ R be the ideal generated by the elements of the matrix A = (ai.j ) can be transformed to              d1 0 · · · 0 d2 . then J = I.. and di |di+1 for 1 ≤ i < m. ¯ a and b are associates ⇒ φ(a) = φ(b). there is an element d1 with I = d1 R. ∃ q. then J = I. ¯ a is a unit in R iﬀ φ(a) = φ(1). The matrix (ai. then ¯ φ(1) is the smallest integer in the image of φ. then the ideal J generated by the elements of EA has J ⊂ I.j ) has at least one non-zero element d with φ(d) a miminum. Also d1 generates the ideal of R ¯ generated by the entries of (ai. Thus r = 0 and a ∈ bR so I = bR.

¯ and di |di+1 for 1 ≤ i < t. where each di = 0.i wn . Theorem 4 Suppose R is a Euclidean domain.. and thus I = d1 R. ∃ invertible matrixes U ∈ Rn and V ∈ Rt such that . a2 .... Then ∃ free bases {a1 .t where vi = a1. By the previous theorem. This is an example of a theorem that is easy to prove playing around at the blackboard. Yet it must be a deep theorem because the next two theorems are easy consequences.. vt }. and such that each ai = di bi . Thus B/A ≈ R/d1 ⊕ R/d2 ⊕ · · · ⊕ R/dt ⊕ Rn−t . the matrix may be changed to the following form.j ) by a ﬁnite number of row and column operations.. .j ) be one which has an entry d1 = 0 with φ(d1 ) a minimum. Among these. . Thus by column operations of type 3. Proof By Theorem 3 in the section Splitting Short Exact Sequences.j . By elementary operations of type 2.i w2 + · · · + an. the other entries of the ﬁrst row may be made zero.. w2 . The composition Rt −→ A −→ B −→ Rn ei −→ vi wi −→ ei ≈ ⊂ ≈ is represented by a matrix (ai. . Let {w1 . In a similar manner. . .118 Appendix Chapter 6 φ values. at } for A and {b1 . 0 cij            Note that d1 divides each ci..j ) ∈ Rn. where n ≥ t. 1) place in the matrix... else we could obtain an entry with a smaller φ value. Then d1 will divide the other entries in the ﬁrst row. with t ≤ n. wn } be a free basis for B.. the entry d1 may be moved to the (1. To consolidate this approach. v2 .            d1 0 · · · 0 0 .i w1 + a2. bn } for B. consider matrices obtained from (ai. b2 . A has a free basis {v1 . B is a ﬁnitely generated free Rmodule and A ⊂ B is a non-zero submodule. . by row operations of type 3.. The proof now follows by induction on the size of the matrix. let (bi.

.s. there is no prime common to their prime factorizations. d2 . dt are ¯ called invariant factors.) Proof There exists an a ∈ R with aR = bR + cR.) Proof s If i = j.j )V =           0 ··· ≈ dt   0 with di |di+1 . . Then t ≈ the natural map R/d −→R/ps1 ⊕ · · · ⊕ R/pst is an isomorphism of R-modules. . If R = Z and we select the di to be positive.. The result now follows from the previous theorem. then the elements d1 . If R = F [x] and we select the di to be monic. the theorem follows. then they are unique... R/di = 0. Theorem 6 Suppose R is a PID and d is a non-zero non-unit element of R. 0 d2 0 ··· 0  0    . and B/A ≈ M. t 1 s (The elements pi i are called elementary divisors of R/d. then M ≈ R/d1 ⊕ R/d2 ⊕ · · ·⊕ R/dt ⊕ Rm where each di = 0. . i.Chapter 6 Appendix  119 d1 0 . v2 .e. they are s . The way Theorem 5 is stated. They are unique up to associates. Since changing the isomorphisms Rt −→ A and B −→ Rn corresponds to changing the bases {v1 . they are unique. The splitting in Theorem 5 is not the ultimate because the modules R/di may split into the sum of other cyclic modules. Let A be the kernel.. s1 s2 Assume d = p1 p2 · · · pst is the prime factorization of d (see bottom of p 110). and for such di . Suppose b and c are relatively prime. some or all of the elements di may be units.. Then bR and cR are comaximal ideals.. so R = bR + cR. wn }. By the Lemma above. (See p 108 for comaximal.  .e..     ≈ U (ai. To prove this we need the following Lemma. a is a unit. w2 . but we do not bother with that here. and di |di+1 for 1 ≤ i < t... pi i and pj j are relatively prime. Lemma Suppose R is a PID and b and c are non-zero non-unit elements of R. Theorem 5 If R is a Euclidean domain and M is a ﬁnitely generated R-module. vt } and {w1 . . If we assume that no di is a unit. Since a|b and a|c. ¯ Proof By hypothesis ∃ a ﬁnitely generated free module B and a surjective homo⊂ morphism B −→ M −→ 0.. . so 0 −→ A −→ B −→ M −→ 0 is a s.

splits. Then the following s.. .e. M ≈ Rm . An element m ∈ M is said to be a torsion element if ∃ r ∈ R with r = 0 and mr = 0. Theorem 9 Suppose R is a Euclidean domain and M is a ﬁnitely generated R-module. Suppose R is a PID. Then M is a free R-module. the natural map is a ring isomorphism (page 108). Proof This is a simple exercise.e.120 Appendix Chapter 6 comaximal and thus by the Chinese Remainder Theorem. If R = Z. and thus there is a splitting map. Since the natural map is also an R-module homomorphism. Theorem 7 Suppose M is a module over a domain R. M/T (M ) is torsion free. Denote by T (M ) the set of all torsion elements of M . it is the same as saying m has ﬁnite order. as seen by the next exercise. i. Deﬁnition Suppose M is a module over a domain R. If T (M ) = 0. because Theorem 5 gives a splitting of M into a torsion part and a free part. it is an R-module isomorphism. M/T (M ) is a free R-module. Proof This follows immediately from Theorem 5. Then T (M ) is a submodule of M and M/T (M ) is torsion free. This theorem carries the splitting as far as it can go. and s ≥ 1. Theorem 8 Suppose R is a Euclidean domain and M is a ﬁnitely generated R-module which is torsion free. 0 −→ T (M ) −→ M −→ M/T (M ) −→ 0 Proof By Theorem 7. Then the Exercise R-module R/ps has no proper submodule which is a summand.s. we say that M is torsion ¯ free. By Theorem 8. p ∈ R is a prime element. This is the same as ¯ ¯ saying m is dependent. Of course this theorem is transparent anyway. Torsion Submodules This will give a little more perspective to this section.

Thus t = 1 and so G is cyclic. The multiplicative group G is isomorphic to an additive group Z/d1 ⊕ Z/d2 ⊕ · · · ⊕ Z/dt where each di > 1 and di |di+1 for 1 ≤ i < t. V depends upon the splitting map and is unique only up to isomorphism. To complete this section. but the complementary summand V is not unique.. 0       dn where each di = 0 and di |di+1 for 1 ≤ i < n. Every u in the additive group has the property that udt = 0. Theorem 11 Suppose R is a Euclidean domain and A ∈ Rn is a matrix with non-zero determinant. Thus if A is invertible. Thus if p is a prime. the equation will have degree ¯ ¯ less than the number of roots. Theorem 10 Suppose T is a domain and T ∗ is the multiplicative group of units of T . A may be transformed to a diagonal matrix       d1 d2 0 . If t > 1. Also d1 generates the ideal generated ¯ by the entries of A. Thus if F is a ﬁnite ﬁeld. (Zp )∗ is cyclic. A is the product of elementary matrices. Exercise For which primes p and q is the group of units (Zp ×Zq )∗ a cyclic group? We know from Exercise 2) on page 59 that an invertible matrix over a ﬁeld is the product of elementary matrices. which is impossible.Chapter 6 Appendix 121 Note It follows from Theorem 9 that ∃ a free submodule V of M such that T (M )⊕ V = M . If G is a ﬁnite subgroup of T ∗ . ¯ So every g ∈ G is a solution to xdt − 1 = 0. then G is a cyclic group. The ﬁrst summand T (M ) is unique. Proof This is a corollary to Theorem 5 with R = Z. Then by elementary row and column operations. the multiplicative group F ∗ is cyclic. This result also holds for any invertible matrix over a Euclidean domain. here are two more theorems that follow from the work we have done. Furthermore A is invertible iﬀ each di is a unit. . .

and C(q) will be the matrix of this endomorphism with respect to this basis. is zero iﬀ h(x) ∈ qR[x]. The companion matrix . V is a torsion module over the ring R[x]. 3 11 3 11 and D = . Write D as the product of elementary matrices. Find the characteristic polynomials of A and D. (See the last part of the last theorem on page 46. the matrix A is invertible iﬀ the diagonal matrix is invertible. Perform elementary 0 4 1 4 operations on A and D to obtain diagonal matrices where the ﬁrst diagonal element divides the second diagonal element. . xn−1 }. Let T : V → V be deﬁned by T (v) = vx. while the Jordan canonical form will exist iﬀ the characteristic polynomial factors as the product of linear polynomials. Exercise Let R = Z. If each di is a unit. it is the product of elementary matrices. q = a0 + a1 x + · · · + an−1 xn−1 + xn ∈ R[x] is a monic polynomial of degree n ≥ 1. . V has a free basis {1. and h(T ) is the zero homomorphism iﬀ h(x) ∈ qR[x]. Find an elementary matrix B over Z such that B −1 AB is diagonal.) Multiplication by x deﬁnes an R-module endomorphism on V . A Jordan block displays its eigenvalue on the diagonal. x2 . Theorem Let V have the free basis {1. If h(x) ∈ R[x]. Note that the Jordan block B(q) is the sum of a scalar matrix and a nilpotent matrix. The homomorphism from R[x]/q to R[x]/q given by multiplication by h(x). ¯ Furthermore. it follows that each di = 0. but as an R-module. That is to say q(T ) = a0 I + a1 T + · · · + T n is the zero homomorphism. All of this is supposed to make the next theorem transparent. and is more interesting than the companion matrix C(q). which is true iﬀ each di is a unit. and V is the R[x]-module V = R[x]/q. Therefore if A is invertible.122 Appendix Chapter 6 Proof It follows from Theorem 3 that A may be transformed to a diagonal matrix with di |di+1 . then the diagonal matrix is the product of elementary matrices of type 1. . Since the determinant of A is not zero.. But as we shall see later. x2 . Show C cannot be selected in Q2 .. h(T ) is the R-module homomorphism given by multiplication by h(x). Find an invertible matrix C in R2 such that C −1 DC is diagonal. . Suppose R is a commutative ring. we deﬁne the two special types of square matrices used in the Rational and Jordan canonical forms. the Rational canonical form will always exist.. x. . x. A = Jordan Blocks In this section. xn−1 }.

 . page 96. (For an overview of Jordan form. .  .. Note In B(q). (x − λ). 0 0 . . Note For n = 1. . if you wish to have the 1s above the diagonal.  1 λ  . h(C(q)) is zero iﬀ h(x) ∈ qR[x]. read ﬁrst the section in Chapter 5. if h(x) ∈ R[x]. 1 −an−1  The characteristic polynomial of C(q) is q. . 1 λ  0 λ The characteristic polynomial of B(q) is q..Chapter 6 Appendix 123 representing T is  C(q) =         0 . . . and |C(q)| = (−1)n a0 .  .. Let V have the free basis {1. . all that’s left to do is to put the pieces together.  . and |B(q)| = λn = (−1)n a0 . . 0   . Finally. if h(x) ∈ R[x]. Then the matrix representing T is   B(q) =  0   . . . This is the only case where a block matrix may be the zero matrix. . reverse the order of the basis for V . ... . Finally. . .. . . Theorem Suppose λ ∈ R and q(x) = (x − λ)n ... . .  0 .  .. . . .. . .  . (x − λ)n−1 }. . . 0 −a0  1 0 .. λ  1   0 . 0 −a1   0 1 0 −a2   . (x − λ)2 . C(a0 + x) = B(a0 + x) = (−a0 ). . . .. Jordan Canonical Form We are ﬁnally ready to prove the Rational and Jordan forms.. Using the previous sections.  . . h(B(q)) is zero iﬀ h(x) ∈ qR[x].) .. .

A submodule of VF [x] is a submodule of VF which is invariant under T . Now we say all this again with a little more detail. In the section on Jordan Blocks. x2 . and T : V → V is an R-module homomorphism. If all the eigenvalues for T are in F . R[x] is a Euclidean domain and so we know almost everything about V as an R[x]-module. . xm−1 } as the F -basis for F [x]/di where m is the degree of the polynomial di . the matrix representing T is          VF [x] ≈ F [x]/d1 ⊕ F [x]/d2 ⊕ · · · ⊕ F [x]/dt C(d1 )  C(d2 )   . Theorem 2 With respect to this basis. Questions about the transformation T are transferred to questions about the module V over the ring R[x]. Since V is ﬁnitely generated as an F -module. we suppose R is a ﬁeld F . V is an R[x]-module. x. From Theorem 5 in the section on Euclidean Domains. V is a ﬁnitely generated F -module. Theorem 1 Under this scalar multiplication. . Pick {1. T : V → V is a linear transformation and V is an F [x]-module with vx = T (v).. . C(dt )           . . it follows that where each di is a monic polynomial of degree ≥ 1. we pick another basis for each of the cyclic modules (see the second theorem in the section on Jordan Blocks). We know VF [x] is the sum of cyclic modules from Theorems 5 and 6 in the section on Euclidean Domains. a basis is selected for these cyclic modules and the matrix representing T is described. Now in this section. This is just an observation. And in the case R is a ﬁeld. and di |di+1 . V is an R-module.124 Appendix Chapter 6 Suppose R is a commutative ring. Then the matrix representing T is called the Jordan Canonical Form. the free part of this decomposition will be zero. but it is one of the great tricks in mathematics. Deﬁne a scalar multiplication V × R[x] → V by v(a0 + a1 x + · · · + ar xr ) = va0 + T (v)a1 + · · · + T r (v)ar . This gives the Rational Canonical Form and that is all there is to it. . Our goal is to select a basis for V such that the matrix representing T is in some simple form.

The pi r need not be distinct. This is called 1 r ¯ the Rational canonical form for T .. . . Thus in the Theorem above. the matrix representing T is              B((x − λ1 )s1 ) B((x − λ2 )s2 ) 0 . Now we apply Theorem 6 to each F [x]/di . This gives VF [x] ≈ F [x]/ps1 ⊕ · · · ⊕ 1 F [x]/psr where the pi are irreducible monic polynomials of degree at least 1. (x − λi )2 . (x − λi )m−1 } as the F -basis for F [x]/(x − λi )si where m is si . Pick an F -basis for each F [x]/psi as before. . Theorem 4 With respect to this basis.. Pick {1. . pi = x − λi and VF [x] ≈ F [x]/(x − λ1 )s1 ⊕ · · · ⊕ F [x]/(x − λr )sr is an isomorphism of F [x]-modules. . i Theorem 3 With respect to this basis. (x − λi ). C(psr ) r 0 The characteristic polynomial of T is p = ps1 · · · psr and p(T ) = 0.Chapter 6 Appendix 125 The characteristic polynomial of T is p = d1 d2 · · · dt and p(T ) = 0. the matrix representing T is C(ps1 ) 1   C(ps2 ) 2                 0 . This is a type ¯ of canonical form but it does not seem to have a name. 0              B((x − λr )sr ) . . Now suppose the characteristic polynomial of T factors in F [x] as the product of linear polynomials.

this section is loosely written. and to show the diagonal elements are all zero. but it is assumed for this exercise. Note that the λi need not be distinct. λ2 . Suppose V is an n-dimensional vector space over a ﬁeld F and T : V → V is a linear transformation. Then λi = 0 for 1 ≤ i ≤ n. its trace is the sum of the eigenvalues and its determinant is the product of the eigenvalues. So suppose T is in Jordan form and trace (T i ) = 0 for 1 ≤ i ≤ n. and the result follows by induction on the size of T . As before we make V a module over F [x] with T (v) = vx. Of course a diagonal matrix is about as canonical as you can get. Thus trace ¯ (p(T )) = a0 n where a0 is the constant term of p(x). Show T is nilpotent. n 2 ¯ ¯ Minimal polynomials To conclude this section here are a few comments on the minimal polynomial of a linear transformation. This is based on the fact that there exists a ﬁeld F containing F ¯ as a subﬁeld. a0 = 0 ¯ ¯ ¯ and so 0 is an eigenvalue of T . However this exercise can still be worked ¯ using Jordan form. Corollary Suppose F is a ﬁeld of characteristic 0. Note A diagonal matrix is in Rational canonical form and in Jordan canonical form.126 Appendix Chapter 6 The characteristic polynomial of T is p = (x − λ1 )s1 · · · (x − λr )sr and p(T ) = 0. λn ) ∈ F n i satisﬁes λ1 + λi + · · +λi = 0 for each 1 ≤ i ≤ n.. We know p(T ) = 0 and thus ¯ trace (p(T )) = 0. This fact is not proved in this ¯ book. n ≥ 1. T is nilpotent iﬀ U T U is nilpotent. and (λ1 . Since the ﬁeld has characteristic 0. and thus T may have no conjugate in Fn which is in Jordan form. . Finally. So ∃ an invertible matrix U ∈ Fn so that −1 −1 U T U is in Jordan form. This means that one block of T is a strictly lower ¯ triangular matrix. such that p factors into linears in F [x]. The polynomial p may not factor into linears in F [x]. This ¯ is called the Jordan canonical form for T. Removing this block leaves a smaller matrix which still satisﬁes the hypothesis. This part should be studied only if you need it. It also has a cute corollary. This exercise illustrates the power and facility of Jordan form. . This is the case where each block is one by one. so it is important to use the transpose principle to write three other versions of the last two theorems. The point is that it suﬃcies to consider the case where T is in Jordan form.. Exercise Suppose F is a ﬁeld of characteristic 0 and T ∈ Fn has trace(T i ) = 0 ¯ for 0 < i ≤ n. Note also that if a matrix is in Jordan form. Let p ∈ F [x] be the characteristic polynomial of T . and of course. and thus a0 n = 0.

If A is given to start with. If p(x) ∈ F [x] is the characteristic polynomial of T . The polynomial u is called the minimal polynomial of T. Recall that q is the characteristic polynomial and the minimal polynomial of each of these matrices. This is a non-zero ideal of F [x] and is thus generated by a unique ¯ monic polynomial u(x) ∈ F (x). The polynomial u is also called the minimal polynomial of A. ¯ Now we state this again in terms of matrices. h(A) = 0 iﬀ h is a multiple of ¯ ¯ u in F [x].   . . Find the characteristic and minimal polynomials of A. then p(A) = 0 and ¯ thus p is a multiple of u. Suppose A is the matrix displayed in Theorem 4 above. Find the characteristic and minimal polynomials of A. let B(q) ∈ Fn be the Jordan block matrix also deﬁned in that section. Suppose A ∈ Fn is a matrix representing T . Exercise 1) 2) 3) Suppose A ∈ Fn . Now suppose q ∈ F [x] is a monic polynomial and C(q) ∈ Fn is the companion matrix deﬁned in the section Jordan Blocks. Note that these properties hold for any matrix representing T ... . If p(x) ∈ F [x] is the characteristic polynomial of A. and thus similar matrices have the same minimal polynomial. Suppose A is the matrix displayed in Theorem 3 above. 0 polynomial. which satisfy V h = 0. use the linear transformation T : F n → F n determined by A to deﬁne the polynomial u.    Find the characteristic polynomial Suppose A is the matrix displayed in Theorem 2 above. Find the characteristic and minimal polynomials of A. This together with the rational form and the Jordan form will allow us to understand the relation of the minimal polynomial to the characteristic polynomial.e. Ann(VF [x]) = uF [x]. p(T ) = 0 and thus p is a multiple of u. h(T ) = 0 iﬀ ¯ ¯ h is a multiple of u in F [x]. Whenever q(x) = (x − λ)n . Note that u(T ) = 0 and if h(x) ∈ F [x]. i. Exercise Suppose Ai ∈ Fni has qi as its characteristic polynomial and its minimal       A1 A2 0 .Chapter 6 Appendix 127 Deﬁnition Ann(VF [x]) is the set of all h ∈ F [x] which annihilate V . and A = Ar and the minimal polynomial of A. Then u(A) = 0 and if h(x) ∈ F [x].

bn ) = f (bτ (1) . . .128 Appendix Chapter 6 4) Suppose λ ∈ F . 2. . B2 . . Find the characteristic and minimal 6) Let F = R and A =  0  −3 1 −1 polynomials of A. . so is f1 + f2 . n ≥ 2.   Determinants In the chapter on matrices. C is an R-module. . The purpose of this section is to give a proof of this. . it is stated without proof that the determinant of the product is the product of the determinants (see page 63). . . then (f r) is R-multilinear. Also if f is R-multilinear and r ∈ R. We suppose R is a commutative ring. b2 . (This funny looking exercise is a little delicate. . . .) 5 −1 3  2 0 . bn ) deﬁnes an R-linear map from Bi to C. . . . From here on. Show λ is a root of the characteristic polynomial of A iﬀ λ is a root of the minimal polynomial of A. . . . . . Deﬁnition A map f : B1 ⊕ B2 ⊕ · · · ⊕ Bn → C is R-multilinear means that if 1 ≤ i ≤ n. Deﬁnition 1) 2) f is symmetric means f (b1 . the set of all functions from B1 ⊕ B2 ⊕ · · · ⊕ Bn to C is an R-module (see page 69). Bi . . ¯ 5) Suppose F is a ﬁeld containing F as a subﬁeld. . bτ (n) ) for all τ . Proof From the ﬁrst exercise in Chapter 5. Bn is a sequence of R-modules. . Show that the minimal polynomial of A ∈ Fn is the same as the minimal polynomial of A considered as a ¯ matrix in Fn . It is easy to see that if f1 and f2 are R-multilinear. bτ (n) ) for all permutations τ on {1. Theorem The set of all R-multilinear maps is an R-module. bn ) = sign(τ )f (bτ (1) . Show that if λ is a root. and B1 . . its order in the characteristic polynomial is at least as large as its order in the minimal polynomial. . . It must be seen that the R-multilinear maps form a submodule. . . and bj ∈ Bj for j = i. then f |(b1 . suppose B1 = B2 = · · · = Bn = B. . . . f is skew-symmetric if f (b1 . . n}. .

b1 . In other words.. bn ) = 0.1 e2 + · · · + an. If we let τ be the transposition (1. . Now we are ready for determinant. then alternating ⇐⇒ skew-symmetric. and show f is alternating. Suppose B is Rn with the canonical basis {e1 . then f (a1. bn ) = −f (b1 . This means f is the multilinear form d times the scalar f (e1 . en }. bn )... a1.. en ). b1 . . 2). and so 2f (b1 . b3 . .n e2 + a2. a1. . . Theorem d is an alternating multilinear form with d(e1 . iii) If no element of C has order 2. .) First we recall the deﬁnition of determinant. For simplicity. bn ) + ¯ f (b2 . assume τ = (1. (We think of a matrix A ∈ Rn as n column vectors. en ). bn ) = −f (bτ (1) . b1 . b1 . .1 e2 +· · ·+ an. . ii) Alternating ⇒ skew-symmetric.. e2 .n ) = |A|.2 · · · aτ (n). b1 + b2 ... dc is an alternating multilinear form. .. ... .n e1 + a2. . as an element of B ⊕ B ⊕ · · · ⊕ B. . ¯ Each of these three types deﬁnes a submodule of the set of all R-multilinear maps... en ). b3 .. . . b3 .j ) ∈ Rn . . It suﬃcies to show that f (b1 . The next theorem follows from the section on determinants on page 61. To prove iii). bn ) = 0.. . e2 . ..n e2 + · · · + an.. b3 . b1 . bn ) = 0 whenever some bi = bj for i = j. Proof Part i) is immediate. Theorem Suppose f : B ⊕ B ⊕ . It turns out that this is all of them. e2 . bn ) = f (b1 ..e.Chapter 6 3) Theorem Appendix 129 f is alternating if f (b1 . if A = (ai. . as seen by the following theorem. .. . bn ) and the result follows. To prove ii). . . ¯ and the result follows. . . . ⊕ B → R is an alternating multilinear form.1 en . . . bτ (n) ) where τ is a transposition.1 en . e2 . Then f = df (e1 .1 e1 + a2.1 aτ (2). e2 . and the determinant is a generator.. b3 . . . Deﬁne d : B ⊕B ⊕· · ·⊕B → R by d(a1.n e2 + · · · + an.. . Thus the set of alternating forms is a free R-module of dimension 1. . . Suppose A = (ai.... .1 e1 +a2. Suppose for convenience that b1 = b2 and show f (b1 . . . en ) = 1... .n en ) = all τ sign(τ )(aτ (1)... Suppose C = R. b2 . b3 .. . In this case multilinear maps are usually called multilinear forms. . assume f is alternating. . i.. ¯ we get f (b1 .n en ) = |A|f (e1 . Then 0 = f (b1 + b2 . suppose f is skew symmetric and no element of C has order 2. . ¯ i) If c ∈ R.. because the set of alternating forms is an R-module. . .j ) ∈ Rn . 2). . b3 ...

1 )f (e1 . .2 a2.1 a2. en ) = |A|f (e1 . .. However there is a natural isomorphism from V to V ∗∗ .1 e1 + a2.2 e2 ) = a1.. .1 e1 + a2.1 a1. e1 ) + a2.2 − a1. while the sum of the homology groups does not.n en ) = ai1 ... it follows that |CA| = f (A) = |A||C|...n f (eτ (1) . A ∈ Rn . . Deﬁne f : Rn → R by f (A) = |CA|. . its dual V ∗ is deﬁned as V ∗ = HomF (V. If A ∈ Rn . . a1. The sum of the cohomology groups forms a ring.1 ai2 .130 Appendix Chapter 6 Proof For n = 2. . f (a1.2 f (e1 . V ∗ is isomorphic to V .n f (e1 . e2 . a tangent plane to a diﬀerentiable manifold is a real vector space.1 aτ (2). F ).. (See the third theorem on page 63.2 · · · aτ (n).. en ). .2 · · · ain . In algebraic topology.. en ) = |CI| = |C|. while cohomology groups are derived from the dual chain complexes. f (A) = |A|f (e1 .1 a1. a1.. . 1 ≤ i2 ≤ n. Use the fact that CA = (CA1 . Since f (e1 .2 f (e2 . Proof Suppose C ∈ Rn . not only in algebra.. you can simply write it out. .1 en . .. e2 . The sections of the tangent bundle are called vector ﬁelds while the sections of the cotangent bundle are called 1-forms. . but also in other areas such as diﬀerential geometry and topology. while the union of the dual spaces is the cotangent bundle. Dual Spaces The concept of dual module is basic.. and f : Rn ⊕ · · · ⊕ Rn → R has f (A1 . This incredible classiﬁcation of these alternating forms makes the proof of the following theorem easy. For the general case. The union of these spaces is the tangent bundle.. e2 ) + a2. .. B = Rn and Rn = Rn ⊕ Rn ⊕ · · · ⊕ Rn . if any is = it for s = t. CA2 . e2 ) = |A|f (e1 . ein ) where the sum is over all 1 ≤ i1 ≤ n.. then |CA| = |C||A|. An ) = |CA|. f (a1.1 e2 . e2 .2 · · · aτ (n).2 e1 + a2.n f (ei1 ... e1 ) + a1. CAn ) to show that f is an alternating multilinear form. An ) where Ai ∈ Rn is column i of A.. . .. . If V is a ﬁnitely generated vector space over a ﬁeld F .n e2 + · · · + an. that term is 0 because f is alternating. but in general there is no natural isomorphism from V to V ∗ . A = (A1 .1 a2.. ei2 ..1 aτ (2).. Therefore the sum is just all τ aτ (1). For example.1 e2 + · · · + an.. e2 ) = (a1. 1 ≤ in ≤ n. Thus the tangent (cotangent) bundle may be considered to be the dual of the cotangent (tangent) bundle. A2 ... e2 . By the previous theorem.2 f (e1 .n e1 + a2.1 a2. and so V ∗ is the dual of V and V may be considered to be the dual of V ∗ . This remarkable fact has many expressions in mathematics. eτ (n) ) = all τ sign(τ )aτ (1).2 f (e2 .. . homology groups are derived from chain complexes. A2 . e2 ). . In the notation of the previous theorem.) Theorem If C. However.. en ). eτ (2) ..

Chapter 6 Appendix 131 Thus the concept of dual module has considerable power. If M and N are R-modules and g : M → N is an R-module homomorphism. then H(g)◦H(h) = H(h ◦ g). g h iii) If M1 −→ M2 −→ M3 are R-module homomorphisms. We develop here the basic theory of dual modules. W ). let H(g) : H(N ) → H(M ) be deﬁned by H(g)(f ) = f ◦ g. then H(I) : H(M ) → H(M ) is the identity. If I : M → M is the identity. H(M1 ⊕ M2 ) ≈ H(M1 ) ⊕ H(M2 ). H is a contravariant functor from the category of R-modules to itself. Suppose R is a commutative ring and W is an R-module. g M     E N f H(g)(f ) = f ◦ g    ~  c W Theorem i) ii) If M1 and M2 are R-modules. If f : M3 → W is a homomorphism. Deﬁnition If M is an R-module. Note that H(g) is an R-module homomorphism. M1 g   E  M2 h  E M3 f c f ◦h◦g f   ◦ h      ~   q  W Note In the language of the category theory. . then (H(g) ◦ H(h))(f ) = H(h ◦ g)(f ) = f ◦ h ◦ g. let H(M ) be the R-module H(M )=HomR (M.

n i ≈ For the general case. . . . .j . vn } is a basis for M ∗ . suppose W = RR . . e∗ } where e∗ = (0. Thus vi (vj ) = δi. . .. i. {v1 .j ) ∈ Rn. let g : Rn → M be given by g(ei ) = vi . 1i . 0. iii) If R is a ﬁeld and g : M → N is a homomorphism. . vn is a free basis for M ∗ . . called the dual basis. vm } and N is a free module with a basis {w1 .. For the remainder of this section. . . 0. . . with basis {e1 . Then v1 . . . In this case H(M ) = HomR (M. Now R1. . then H(g) : H(N ) → H(M ) is surjective. . then H(g) : H(N ) → H(M ) is injective. ∗ Theorem Suppose M has a ﬁnite free basis {v1 . i     . This means g(vj ) = a1. 0). . . Then g ∗ : M ∗ → (Rn )∗ ∗ ∗ ∗ sends vi to e∗ . . Deﬁne vi ∈ M ∗ by ∗ ∗ ∗ ∗ vi (v1 r1 + · · · + vn rn ) = ri . is given by At . .n is free with dual basis {e1 . . en } where ei =  1i    · 0 We know (Rn )∗ ≈ R1.e.m . vn }. . .. Then the matrix of g ∗ : N ∗ → M ∗ with respect to the dual bases.  0   ·  n Proof First consider the case of R = Rn. . wn } and g : M → N is the homomorphism given by A = (ai. . .132 Appendix Chapter 6 Theorem If M and N are R-modules and g : M → N is an isomorphism.1 . Since g ∗ is an isomorphism.j w1 + · · · + an. If g : M → N is an injective homomorphism and g(M ) is a summand of N. Therefore M ∗ is free and is isomorphic to M .    Theorem Suppose M is a free module with a basis {v1 .n . . . . any homomorphism from Rn to R is given by a 1 × n ∗ matrix. Proof IH(N ) = H(IN ) = H(g ◦ g −1 ) = H(g −1 ) ◦ H(g) IH(M ) = H(IM ) = H(g −1 ◦ g) = H(g) ◦ H(g −1 ) Theorem i) ii) If g : M → N is a surjective homomorphism. then g is surjective (injective) iﬀ H(g) is injective (surjective). Proof This is a good exercise.. R) is denoted by H(M ) = M ∗ and H(g) is denoted by H(g) = g ∗ . .j wn . then H(g) : H(N ) → H(M ) is an isomorphism with H(g −1 ) = H(g)−1 . . . .

u) = f (u). Av) = φ(At f. v). then the following diagram is commutative.e. . Thus g ∗ (wi ) ∗ ∗ = ai. and this exercise should be worked out completely. This is with the elements of Rn and (Rn )∗ written as column vectors. Exercise If U is an R-module.e. α(vn )} is the dual basis of {v1 . v). i. Note that α is a homomorphism. Theorem If M is a free R-module with a ﬁnite basis {v1 .Chapter 6 Appendix 133 ∗ Proof Note that g ∗ (wi ) is a homomorphism from M to R.j wn ) = ai.j w1 + · · · + an. . f ∈ N ∗ and v ∈ M . u). vn }. g(v)) = φM (g ∗ (f ). Deﬁnition “Double dual” is a “covariant” functor.j . and thus g ∗ is represented by At .. α(u) = φN (−. Show that φU is R-bilinear. .1 v1 + · · · + ai. v). . . i. . α(vi ) = (vi )∗ . Theorem If M and N are R-modules and g : M → N is a homomorphism. then g ∗∗ : M ∗∗ → N ∗∗ . .. . Use the theorem above to show that φ : (Rn )∗ ⊕ Rn → R has the property φ(f. Now suppose M = N = Rn and g : Rn → Rn is represented by a matrix A ∈ Rn . If the elements of Rn are written as column vectors and the elements of (Rn )∗ are written as row vectors. Evaluation on vj gives ∗ ∗ ∗ ∗ ∗ g ∗ (wi )(vj ) = (wi ◦ g)(vj ) = wi (g(vj )) = wi (a1. v). Dual spaces are confusing. if g : M → N is a homomorphism.. i. Av) = φ(f A. . . α(m) is given by evaluation at m. For any module M . g(v)) = φM (g ∗ (f ). Proof ∗ ∗ ∗ {α(v1 ). v). Show that φN (f. the formula is φ(f. The proof follows from the equation φN (f. vn }. . M g c α E M ∗∗ g ∗∗ α E N ∗∗ c N Proof On M. deﬁne φU : U ∗ ⊕ U → R by φU (f. then α : M → M ∗∗ is an isomorphism. deﬁne α : M → M ∗∗ by α(m) : M ∗ → R is the homomorphism which sends f ∈ M ∗ to f (m) ∈ R.m vm . Of course this is just the matrix product f Av. α is given by α(v) = φM (−. Suppose g : M → N is an R-module homomorphism. Suppose f ∈ (Rn )∗ and v ∈ Rn . On N.e. . .

and M = t∈T Rt . If {v1 . and Rt . . vn } is any orthonormal basis for V. vn } is the dual basis ∗ ∗ for {v1 . that is β(vi ) = vi . However for real inner product spaces there is. Also. R = R. β(vn )} is the dual ∗ basis of {v1 . Proof β is injective and V and V ∗ have the same dimension. . . then f ∗ : U ∗ → V ∗ is an isometry and the following diagram commutes. . . . if U is another n-dimensional IPS and f : V → U is an isometry. Now let for each t ∈ T . . Rt = R. . vn }. . . α is a natural equivalence between the identity functor and the double dual functor. Note If {v1 . {v1 . . . vn } is an orthonormal basis for V ∗ . . . Note If β is used to identify V with V ∗ . . vn } ∗ ∗ is a basis for V and {vi . . . {β(v1 ). α is used to identify V and V ∗∗ . . if {v1 . . then {v1 . . . In general there is no natural way to identify V and V ∗ . and under this structure. . . Note For ﬁnitely generated vector spaces. . Under this identiﬁcation V ∗ is the dual of V and V is the dual of V ∗ . . −) is an isomorphism. . . T is an inﬁnite index set. β is an isometry. . . . Also. Show ( Rt )∗ is isomorphic to RT = t∈T t∈T T = Z+ . . V f c β E V∗ T f∗ U β E U∗ Exercise Suppose R is a commutative ring. The isomorphism β : V → V ∗ deﬁnes an inner product on V ∗ . . vn }. Then β : V → V ∗ given by β(v) = (v. then φV : V ∗ ⊕ V → R is just the dot product V ⊕ V → R. Show M ∗ is not isomorphic to M . . . vn } its dual basis. vn } is ∗ ∗ an orthonormal basis for V. . . . In the language of category theory.134 Appendix Chapter 6 Note Suppose R is a ﬁeld and C is the category of ﬁnitely generated vector spaces over R. Theorem Let R = R and V be an n-dimensional real inner product space.

51 Cancellation law in a group. 72. 109 Automorphism of groups. 97. 76 Coset. 43 Axiom of choice. 62 Comaximal ideals. 79 of a module. 120 Commutative ring. 116 integral domain. 50 Characteristic polynomial of a homomorphism. 74 Cycle. 24. 95 of a matrix. 66 Chinese remainder theorem. 29 of modules. 39 Cartesian product. 85. 120 Elementary matrices. 70 of rings. 78. 5 principal ideal. 71 Algebraically closed ﬁeld. 42.Index Abelian group. 46 unique factorization. 95 Elementary divisors. 32 Cyclic group. 60. 98. 20. 32 Ascending chain condition. 2.7 Binary operation. 1. 22 Change of basis. 128 Diagonal matrix. 39 of a function. 47. 119. 97 Alternating group. 19 Boolean algebras. 108 Classical adjoint of a matrix. 85 of a matrix. 45 Domain euclidean. 31 Cayley-Hamilton theorem. 56 Dimension of a free module. 46. 66. 37 Complex numbers. 64 Conjugation by a unit. 63 135 Cofactor of a matrix. 108. 95 Eigenvectors. 46. 83 Characteristic of a ring. 58 . 107 Determinant of a homomorphism. 132 Dual spaces. 111 Dual basis. 83 Bijective or one-to-one correspondence. 40. 47. 11 Cayley’s theorem. 130 Eigenvalues. 131 Coproduct or sum of modules. 10 Basis or free basis canonical or standard for Rn . 52 Boolean rings. 23 module. 112 Associate elements in a domain. 20 in a ring. 125 Center of group. 44 Contravariant functor. 83 Division algorithm. 50. 104 Conjugate.

34 Isometry. 113 Homogeneous equation. 72. 7. 23 of rings. 74 ring. 6 bijective. 41 Idempotent element in a ring. 55 Generating sequence in a module. 42 of modules. 1. 70 Equivalence class. 113 General linear group GLn (R). 69 as a ring. 47. 4 Equivalence relation. 36 as a module. 109 Hilbert. 109 of a ring. 3. 6 Greatest common divisor. 78 Index of a subgroup. 41 maximal. 100 Free basis. 122 Endomorphism of a module. 2 Induction. 116 Evaluation map. 78 Generators of Zn . 26. 50 Field. 41 prime. 106 Factorization domain (FD). 14 Euclidean domain. 47. 49. 46 right. 110 Isometries of a square. 49 Even permutation. 46 Gauss. 111 Fermat’s little theorem. 39 Formal power series. 119 Inverse image. 79 Inner product spaces. 42. 51 Image of a function. 7 Invertible or non-singular matrix. 22. 7 Independent sequence in a module. 25 Index set. 100 Graph of a function. 40 Geometry of determinant. 98 Integers mod n. 60 Homormophism of groups. 7 Function space Y T as a group. 78. 19 abelian. 87. 109 principal. 14 Invariant factors. 4 Euclidean algorithm. 69 Homomorphism of quotient group. 29 module. 44 Ideal left. 20 additive. 23 multiplicative. 83 Free R-module. 12 Fundamental theorem of algebra. 32 Exponential of a matrix. 27. 79. 55 Irreducible element. 20 cyclic. 57. 19 symmetric. 15 Group. 90 Gram-Schmidt orthonormalization. 101 Isomorphism . 44 as a set. 7 surjective. 31 Index Hausdorﬀ maximality principle.136 Elementary operations. 40 Integers. 78 Function or map. 7 injective. 113 Fourier series. 13 Injective or one-to-one.

11 Projection maps. 5 Permutation. 10 Ring. 99 Orthonormal sequence. 107. 123. 79 Order of an element or group. 6 Rank of a matrix. 75 of rings. 78 Linear ordering. 8. 109 integer. 125 Relation. 127 Minor of a matrix. 112 Normal subgroup. 56 Maximal ideal. 49 of sets. 121 Nilpotent element.Index of groups. 109 independent sequence. 39 Polynomial ring. 12 Prime element. 34. 17. 99 Partial ordering. 42 137 Range of a function. 46 Principal ideal. 16 elements in a PID. 110 ideal. 26 Odd permutation. 123 Jordan canonical form. 93 Noetherian ring. 70 Least common multiple. 129 Multiplicative group of a ﬁnite ﬁeld. 28. 23 Orthogonal group O(n). 62 Module over a ring. 96. 125 Kernel. 7. 87 monotonic subcollection. 56 homomorphism. 29 of modules. 35 of modules. 3 Linear transformation. 102 Orthogonal vectors. 55 representing a linear transformation. 31 Pigeonhole principle. 16 Principal ideal domain (PID). 32 Onto or surjective. 3 Relatively prime integers. 42 Product of groups. 4 Multilinear forms. 119 Right and left inverses of functions. 91 Jordan block. 2. 89 Rational canonical form. 86. 59. 27 Quotient module. 4 subgroup. 45 Power set. 59 Scalar matrix. 96. 74 Quotient ring. 38 Root of a polynomial. 70 of rings. 43 Jacobian matrix. 84 triangular. 11 Quotient group. 43. 85 Matrix elementary. 57 . 58 invertible. 48 Monotonic collection of sets. 3 Partition of a set. 18 Linear combination. 68 Monomial. 114 Minimal polynomial. 46 Row echelon form.

138 Scalar multiplication. 111 Unit in a ring. 67. 103. 31 Symmetric matrix. 132 Transposition. 121 Trace of a homormophism. 115 Surjective or onto. 103. 103 Torsion element of a module. 39 Index . 65 Transpose of a matrix. 69 Subring. 79 Strips (horizontal and vertical). 64 Solutions of equations. 9. 54. 85 of a matrix. in principal ideal domains. 21. 14. 90 Zero divisor in a ring. 21 Submodule. 16 Unique factorization domain (UFD). 114 Standard basis for Rn . 32 Unique factorization. 81 Splitting map. 8 Subgroup. 56. 60 Similar matrices. 72. 77. 38 Vector space. 7. 115 Sign of a permutation. 38. 79 Symmetric groups. 105 Short exact sequence. 59. 41 Summand of a module. 85 Volume preserving homomorphism. 113 of integers. 71 Self adjoint.