# 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
November 30, 2000 [http://www.math.miami.edu/∼ec/book/]
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 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. 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 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.
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 55
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 94
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
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.
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.
In this book, the only applications of the HMP are 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 set with n elements, Y is a set with m
elements, and f : X →Y is a function.
1) If n = m, then f is injective iﬀ f is surjective iﬀ f is bijective.
2) If n > m, then f is not injective.
3) If n < m, 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
n = m = 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 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.
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
. (Thus if T = Z
+
, ¦x
t
¦ = ¦x
1
, x
2
, ...¦.) For each s ∈ 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 n and m elements, then
Y
T
has m
n
elements. In particular, when T = ¦1, 2, 3¦, Y
T
= Y Y Y has
m
3
elements. Show that if m ≥ 3, the subset of Y
{1,2,3}
of all injective functions has
m(m− 1)(m− 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 n elements, {(T) has 2
n
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”.
6) A set Y is said to be countable if it is ﬁnite or if there is a bijection from N to
Chapter 1 Background 13
Y. Consider the following three collections.
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. Show there is a bijection from ¦0, 1¦
N
to the real numbers R. (This is
not so easy.)
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.
We are now ready for our ﬁrst theorem with any guts.
Theorem If a and b are relatively prime and a [ bc, then 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 n is an integer which is not 0,1,
or -1. Then n 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
k
and
positive integers s
1
, s
2
, ..., s
k
such that n = ±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 or 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.
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
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, the least common multiple of a and b is c = ab/(a, b). In
particular, if 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)

2 is irrational. (Proof: Suppose

2 = m/n where (m, n) = 1. Then
2n
2
= m
2
and if n > 1, n and m have a common prime factor.
Since this is impossible, n = 1, and so

2 is an integer. This is a

2 is irrational.)
6) Suppose c is an integer greater than 1. Then

c is rational iﬀ

c is an
integer.
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
. 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 = 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 all the concepts are new.
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) 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
.
(vi) 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.
(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. 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. 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 is a cyclic group of order n. Then ∃ a ∈ G with G =
¦a
o
, a
1
, . . . , a
n−1
¦. Suppose H is a subgroup of G with more than one element. Let
m be the smallest integer with o < m < n and a
m
∈ H. Then m[n and a
m
generates
H. The case where G is an inﬁnite cyclic group is left as an exercise. Note that Z
is an additive cyclic group and it was shown in the previous chapter that subgroups
of Z are 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.
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, we used H 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. This is really much 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 bijection is given by
F(Ha) = a
−1
H. 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 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.
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.
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 deﬁned
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 abelian group.
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.
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 when we study 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
.
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 integers
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 seven basic properties of permutations. They are all easy except
4), which is rather delicate. Properties 8), 9), and 10) are listed solely for reference.
Theorem
1) Disjoint cycles commute. (This is obvious.)
2) Every 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 transpostions.
(Proof: (a
1
, ..., a
n
) = (a
1
, a
2
)(a
1
, a
3
) (a
1
, a
n
). )
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
.
The following parts are not included in this course. They are presented here merely
for reference.
8) For any n = 4, A
n
is simple, i.e., has no proper normal subgroups.
9) For any n ≥ 3, A
n
is generated by its 3-cycles.
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) The proof presented here uses polynomials in n variables with real
coeﬃcients. Since polynomials will not be introduced until Chapter 3, the student
may skip the proof until after that chapter. Suppose S = ¦1, ..., n¦. If σ is a
permutation on S and p = p(x
1
, ..., x
n
) is a polynomial in n variables, deﬁne σ(p)
to be the polynomial p(x
(1)σ
, ..., x
(n)σ
). Thus if p = x
1
x
2
2
+ x
1
x
3
, and σ is the trans-
position (1, 2), then σ(p) = x
2
x
2
1
+ x
2
x
3
. Note that if σ
1
and σ
2
are permutations,
σ
2

1
(p)) = (σ
1
σ
2
)(p). Now let p be the product of all (x
i
−x
j
) where 1 ≤ i < j ≤ n.
(For example, if n = 3, p = (x
1
−x
2
)(x
1
−x
3
)(x
2
−x
3
).) If σ is a permutation on S,
then for each 1 ≤ i, j ≤ n with i = j, σ(p) has (x
i
−x
j
) or (x
j
−x
i
) as a factor. Thus
σ(p) = ±p. A careful examination shows that if σ
i
is a transposition, σ
i
(p) = −p.
Any permutation σ is the product of transpositions, σ = σ
1
σ
2
σ
t
. Thus if σ(p) = p,
t must be even, and if σ(p) = −p, t must be odd.
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 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 G
1
G
2
and G
2
G
1
are isomor-
phic.
Exercise If o(a
1
) = n and o(a
2
) = m, 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 that Z
mn
is isomorphic to Z
n
Z
m
iﬀ
(n, m) = 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
.
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.
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. (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., 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
¯
. Therefore 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. 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. A non-zero element a ∈ R is called
a zero divisor provided ∃ 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 Chapter 2 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 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.
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 or right (left, 2-sided) ideals of R, then
¸
t∈T
I
t
is a
right (left, 2-sided) ideal of R.
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 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
f
−1
(0
¯
).
Here is a list of the basic properties of ring homomorphisms. Much of this
work has already been done in 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
¯
to 1
¯
).
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 an 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 R is a ring, T is a non-void set, 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 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].
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
Suppose n and m are relatively prime integers with n, m > 1. There is an exercise
in Chapter 2 to show that Z
nm
and Z
n
Z
m
are isomorphic as groups. It will now be
shown that they are also isomorphic as rings. (For a useful and elegant generalization
of this theorem, see the Appendix.)
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 rings.
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.)
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 homomorphism
f(m) = m1
¯
= m
¯
. The non-negative integer n with ker(f) = nZ is called the charac-
teristic 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.
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.
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 for R.
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.
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.
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.
Exercise Suppose R is a ﬁnite Boolean ring. Show that R ≈ Z
2
Z
2
Z
2
.
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 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, V A, 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. 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
.
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
There are 3 types of elementary row and column operations on a 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 following 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 is the coset D + ker(f) in R
n
.
(See part 7 of the section on Homomorphisms in Chapter 2.)
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 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 say that d is skew-symmetric.
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 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 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 much 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.
Convention Unless otherwise stated, the word “R-module” (or sometimes just
“module”) will mean “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 a module because the axioms
will 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.
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 If R is a subring of a ring T, then T, with scalar multiplication deﬁned
by ring multiplication, is an R-module. In particular, R is a Q-module. If f : Q →R
is a Z-module homomorphism, must f be a Q-module homomorphism?
Homomorphisms on R
n
R
n
as an R-module In Chapter 4 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
. 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 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. 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. 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
= MMM 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 m
1
, . . . , m
n
∈ R
m
,
deﬁne A ∈ R
m,n
to be the matrix with column i = m
i
. Then f deﬁned by f(B) = AB
is the unique homomorphism from R
n
to R
m
with f(e
i
) = m
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. 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.
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.
Exercise Suppose M and N are R-modules, and A ⊂ M, B ⊂ N are submodules.
Show (M ⊕N)/(A ⊕B) is isomorphic to (M/A) ⊕(N/B).
Exercise Suppose R is a commutative ring, M is an R-module, and n ≥ 1. Deﬁne
a function ∝: 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
.
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
1) Is 2Z a summand of Z?
2) Is 4Z
8
a summand of Z
8
?
3) Is 3Z
12
a summand of Z
12
?
4) Suppose n, m > 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 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. If ∃ a basis for M, M is said to
be a free R-module. 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
.
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
. 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 M and N are modules, f : M →N is a homomorphism, and
S = ¦s
t
¦ is a basis for M. 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
It will now be shown that any free R-module is isomorphic to one of the
canonical free R-modules.
Theorem An R-module N is free iﬀ ∃ an index set T such that N ≈

t∈T
R
t
. In
particular, N has a ﬁnite free basis of n elements iﬀ N ≈ R
n
.
Proof If N is isomorphic to a free module, N is certainly free. Now suppose N
has a free basis ¦s
t
¦. Then the homomorphism f : ⊕R
t
→ N with f(e
t
) = s
t
sends
the canonical basis for ⊕R
t
to the basis for N. 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).
The next exercise is to 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 λ =

¸
¸
λ
1
.
λ
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) f(λ) is the linear combination of the columns of A, f(λ) = f(e
1
λ
1
+ +
e
n
λ
n
) = v
1
λ
1
+ +v
n
λ
n
.
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.
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.
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.
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. When we
originally deﬁned basis, we could have called it “indexed free basis” or even “ordered
free basis”.
Two sequences cannot begin to be equal unless they have the same index set.
We will follow the classical convention that an index set with n elements must 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.
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 chap-
ter 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 that 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
Chapter 5 Linear Algebra 85
is 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) : 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) = (1, −1). Find the matrix A ∈ R
2
which represents f with respect to the
basis ¦(1, 2), (−1, 1)¦. 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 or trace. We now focus on the case where R is a ﬁeld, and
show that in this case, every R-module is free. Thus any ﬁnitely generated R-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
¯
or r = 0
¯
.
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.
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 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, W is an F-module, and f : M → W
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
.
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.) Parts 1) and 1
t
)
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 if B ∈ F
n
, f(B) is a linear combination of those vectors.
90 Linear Algebra Chapter 5
Thus the image of f is the submodule of F
m
generated by the columns of A, and
its dimension is the rank of f. 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 Suppose M is a ﬁnite dimensional vector space over a ﬁeld F, and
f : M → M is an endomorphism. The rank of f is deﬁned to be the dimension of
the image of f. It follows from the work above that this is the same as the rank of
any matrix representing f.
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
¯
.
Extend v
1
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. Do not use 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.)
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 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 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 polyno-
mial 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
¯
.
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, 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
) (s
1
, .., s
n
) = r
1
s
1
+ +r
n
s
n
. S = ¦e
1
, .., e
n
¦ will be called the canonical or
standard orthonormal basis. The next theorem shows that this inner product has an
amazing geometry.
Theorem If u, u ∈ 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.
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) and w
2
= (0, 1, 0). W is
generated by the sequence ¦w
1
, w
2
, v¦ where v = (1, 2, 3). 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 a future exercise 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 S =
¦f(u
1
), .., f(u
n
)¦ is an orthonormal sequence in V . Then S is independent and thus
S is a basis and thus f is an isomorphism. 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 .
Now 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 Suppose y, z ∈ R
n
. Then the dot product y z is the matrix product y
t
z.
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
(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 2 2 matrix. Let λ be an eigenvalue for A and ¦v
1
, v
2
¦ be an orthonor-
mal basis for R
2
with Av
1
= λv
1
. Then w.r.t this basis, the transformation 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 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.
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 integers. Here it 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 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 on
group theory at the end of this section.
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 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
a = b
1
b
2
b
n
= c
1
c
2
c
m
with each b
i
, 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.
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.
R is a unique factorization domain (UFD) means R is a FD and 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 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, 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) If ¦I
t
¦
t∈T
is a collection of ideals, ∃ t
0
∈ T such that if t is any element in T
with I
t
⊃ I
t
0
, then I
t
= I
t
0
. (The ideal I
t
0
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
to
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 1) is false and I ⊂ R is an ideal not ﬁnitely generated. Then ∃ a
sequence a
1
, a
2
, . . . of elements in I with a
1
R+a
2
R+ +a
n
R properly contained
in a
1
R + a
2
R + + a
n+1
R for each n ≤ 1. Thus 3) is false and 1) ⇔ 3). The
proof that 2) ⇒3) is immediate. If 2) is false, ∃ a sequence of ideals I
1
⊂ I
2
⊂ . . .
with each inclusion proper. Thus 3) is false and so 2) ⇔ 3).
Deﬁnition If R satisﬁes these properties, R is said to be Noetherian, or it is said
to satisfy the ascending chain condition. This is a useful property satisﬁed by many
of the classical rings in mathematics. Having three deﬁnitions makes this property
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. Then ∃
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 of a and b factors as
Chapter 6 Appendix 113
a ﬁnite product of irreducibles. This gives a ﬁnite factorization of c into irreducibles
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 over C 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.)
Note The combination of the last two theorems shows that Noetherian is a ubiq-
uitous property which is satisﬁed by many of the basic rings in commutative algebra.
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
114 Appendix Chapter 6
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.
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.
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.
A B C
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 commutative ring, 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, it 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 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.
Euclidean Domains
The ring Z possesses the Euclidean algorithm and the polynomial ring F[x] has
the division algorithm. The concept of Euclidean domain is an abstraction of these
properties. The axioms are so miniscule that it is surprising you get this much juice
out of them. However they are exactly what you need, and 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(i) ∀ i ∈ 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.
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 ¦h
1
, h
2
, ..., h
t
¦. Let ¦g
1
, g
2
, ..., g
n
¦ be a free basis for B, where n ≥ t. The
composition
R
t

−→A

−→B

−→R
n
e
i
−→h
i
g
i
−→e
i
is represented by a matrix (a
i,j
) ∈ R
n,t
where h
i
= a
1,i
g
1
+ a
2,i
g
2
+ + a
n,i
g
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 ¦h
1
, h
2
, ..., h
t
¦ and ¦g
1
, g
2
, ..., g
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 be 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.
Proof There exists and 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.
Let d = p
s
1
1
p
s
2
2
p
st
t
be the prime factorization of d. 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. 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.
To give perspective to this section, here is a brief discussion of torsion submodules.
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 g in the additive group has the property that gd
t
= 0
¯
. So
every u ∈ 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 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 matrices.
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 division algorithm in the chapter on rings.) 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].
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, see the section in Chapter 5.)
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.
Exercise 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
+ λ
i
2
+ +λ
i
n
= 0
¯
for each 1 ≤ i ≤ n. Then λ
i
= 0
¯
for 1 ≤ i ≤ n.
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
n
¸

. 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 is from the section on determinants in Chapter 4.
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 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 g : M → N is an isomorphism, then H(g) : H(N) → H(M) is an
isomorphism, and 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, 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.
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 has a basis ¦v
1
, . . . , v
m
¦ and N has 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 g

(w

i
) is a homomorphism fromM 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 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 has a ﬁnite free 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.
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 [http://www.math.miami.edu/∼ec/book/]

November 30, 2000

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.

miami. Unfortunately mathematics is a diﬃcult and heavy subject. Teaching abstract algebra and linear algebra as distinct courses results in a loss of synergy and a loss of momentum. but it requires a lot of organization. I hope the students and professors who try it. Basic algebra is a subject of incredible elegance and utility. I would have made the book shorter. Because after you extract it. because mathematics is learned in hindsight. This book works best when viewed lightly and read as a story. This book is my attempt at that organization.edu .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. The style and approach of this book is to make it a little lighter. enjoy it. Also I am convinced it is easier to build a course from a base than to extract it from a big book. FL 33124 ec@math. H. The goal is to stay focused and go forward. E. you still have to build it. but I did not have any more time. 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. Connell Department of Mathematics University of Miami Coral Gables.

solutions of equations. strips.vi Outline Chapter 1 Background and Fundamentals of Mathematics 1 3 5 13 14 Sets. projections Notation for the logic of mathematics Integers. equivalence relations Functions. right and left inverses. and scalar matrices Elementary operations and elementary matrices Systems of equations 53 55 56 57 59 . order. domains. Cartesian products Relations. invertible matrices Transpose Triangular. partial orderings. 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. unique factorization Chapter 2 Groups 19 21 25 27 31 34 Groups. the integers mod n Homomorphisms Permutations. Hausdorﬀ maximality principle. bijections. quotient groups. cosets Normal subgroups. subgroups. scalar multiplication for additive groups Subgroups. the symmetric groups Product of groups Chapter 3 Rings Rings Units. diagonal.

Gram-Schmidt orthonormalization Orthogonal matrices. the classical adjoint Similarity. submodules Homomorphisms Homomorphisms on Rn Cosets and quotient modules Products and coproducts Summands Independence. and free basis Characterization of free modules Uniqueness of dimension Change of basis Vector spaces.vii Determinants. and characteristic polynomial Chapter 5 Linear Algebra 68 69 71 74 75 77 78 79 82 83 85 90 91 92 93 94 96 98 102 103 60 64 Modules. 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 . square matrices over ﬁelds. generating sets. rank of a matrix Geometric interpretation of determinant Linear functions approximate diﬀerentiable functions locally The transpose principle Nilpotent homomorphisms Eigenvalues. trace. characteristic roots Jordan canonical form Inner product spaces.

viii .

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

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

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

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

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

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

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

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

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

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

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

1} then α ◦ β(f ) = f . P(T ) has 2n elements. Show that if T is a ﬁnite set with n elements. For any Y and Yt . and be 0 when t ∈ S. then Y T is the inﬁnite product Y × Y × · · · . 3}. 4) If S is any subset of T . Y T = Y × Y × Y has m3 elements. when T = {1. Then Y T = t∈T 2) Suppose each of Y1 and Y2 is a non-void set. then Y T has mn elements. Deﬁne f : T → {0. denote γ(t) by γ(t) = ft : T → {0. (This is the fundamental property of Cartesian products presented in the two previous theorems. Show that if T and Y are ﬁnite sets with n and m elements. Show that if S ⊂ T then β ◦ α(S) = S. 1}T is a function and show that it cannot be surjective. . Thus α is a bijection and β = α−1 . Show that f is not in the image of γ and thus γ cannot be surjective. the subset of Y {1. 1}. 1}T 5) Suppose γ : T → {0. That is.) 3) Deﬁne P(T ). y2 . Show that if m ≥ 3. 1] → R which have an inﬁnite number of derivatives. Use the fundamental theorem of calculus to show that D is surjective.3} of all injective functions has m(m − 1)(m − 2) elements. It is included here for students who wish to do a little more set theory. deﬁne its characteristic function χS : T → {0. Exercise This exercise is not used elsewhere in this text and may be omitted. Deﬁne D : A0 → A by D(f ) = df /dx. let Yt be a copy of Y for each t ∈ T. 2. If T = N. 1} by f (t) = 0 if ft (t) = 1.12 Background Chapter 1 A Calculus Exercise Let A be the collection of all functions f : [0. then the set {0. Suppose T is a non-void set. 1}T by α(S) = χS . 1}T → P(T ) by β(f ) = f −1 (1). Deﬁne α : P(T ) → {0. Y N is the set of all inﬁnite sequences (y1 . and f (t) = 1 if ft (t) = 0. 6) A set Y is said to be countable if it is ﬁnite or if there is a bijection from N to . .2. 1) If Y is a non-void set. This shows that if T is an inﬁnite set. and if f : T → {0. deﬁne Y T to be the collection of all functions with domain T and range Y . 1} by | letting χS (t) be 1 when t ∈ S. If t ∈ T . . In particular. These injective functions are called permutations on Y taken 3 at a time. Deﬁne β : {0. P(T ) ←→ {0. the power set of T . T . to be the collection of all subsets of T (including the null set). Show there is a natural bijection from (Y1 × Y2 )T to Y1T × Y2T . Use the mean value theorem to show that D is injective.) where each yi ∈ Y . Let A0 ⊂ A be the subcollection of those functions f with f (0) = 0. 1}T represents a “higher order of inﬁnity”.

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

for each n ≥ 1. . We say that G is closed under negation and closed under addition. not both zero. then m = −4 and r = 3. For example. b with b = 0. The Integers In this section. Here we will establish the following three basic properties of the integers.. b. then n factors uniquely as the product of primes. Exercise Use induction to show that... then G is a subgroup of Z. 1) 2) If G is a subgroup of Z. 3) All of this will follow from long division. −2n. Note that b | a ⇔ the rational number a/b is an integer ⇔ ∃! m such that a = bm ⇔ a ∈ bZ. −17 = 5(−4) + 3. b divides a “m times with a remainder of r”. This fact is written as b | a. ± 1 divides anything . 1 + 2 + · · · + n = n(n + 1)/2.14 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. If n ≥ 2.e. Also n divides a and b with the same remainder iﬀ n divides (a − b). In other words. elements of Z. ∃! m and r with 0 ≤ r <|b| and a = bm + r. .}. i. n. this is impossible. 0 does not divide anything. P (n) is true. which we now state formally. 0. lower case letters a... Deﬁnition If r = 0.. c. and its positive generator is the greatest common divisor of a and b. Euclidean Algorithm Given a. will represent integers. Deﬁnition A non-void subset G ⊂ Z is a subgroup provided (g ∈ G ⇒ −g ∈ G) and (g1 .. if a = −17 and b = 5. . Suppose P (1) is true and for each n ≥ 1. Note Anything (except 0) divides 0. If n = 0. Since P (m − 1) is true.. g2 ∈ G ⇒ (g1 + g2 ) ∈ G).. then ∃ n ≥ 0 such that G = nZ. 2. Then for each n ≥ 1. . . then ∃ a smallest positive integer m such that P (m) is false. Proof If the theorem is false. the set of integers which n divides is nZ = {nm : m ∈ Z} = {. −n. 2n. If a and b are integers. P (n) ⇒ P (n + 1). we say that b divides a or a is a multiple of b..

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

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

Suppose | a |> 1 and | b |> 1.. Thus if there is no common prime in their factorizations.. and show that it is equal to (180 · 28)/(180.) √ c is rational iﬀ √ c is an 2) 3) 3) 4) 5) 6) Suppose c is an integer greater than 1. Then 1 k (a. Note that c is a multiple of a and b. n =√ and so 2 is an integer. the picture becomes transparent. . In particular. pk . 22 · 54 · 7) = 22 · 5. i. ∃ m.28). n) = 1. then n is a multiple of c. For example (23 · 5 · 11. Exercise Find (180.e. . Then the only 1 k divisors or a are of the form ±pt1 · · · ptk where 0 ≤ ti ≤ si for i = 1. ﬁnd the greatest common divisor of 180 and 28. 28).. Theorem (Summary) 1) Suppose | a |> 1 has prime factorization a = ±ps1 · · · psk . n and m have a common prime factor. Then 2n2 = m2 and if n > 1. b) = 1 iﬀ there is no common prime in their factorizations. Let {p1 .e. Let ui be the minimum of si and ti . the least common multiple of a and b is c = ab/(a. Finally. n with ma + nb = 1. 1 k If | a |> 1 and | b |> 1. Then no prime would divide (p1 p2 · · · pk + 1).. Then c = pv1 · · · pvk is the least 1 k common multiple of a and b. .. .. 28). if a and b are relatively prime. b) = pu1 · · · puk . Find integers m and n such that 180m + 28n = (180. . . b). i. There is an inﬁnite number of primes. √ Since this is impossible. (Proof: Suppose there were only a ﬁnite number of primes p1 . pk } be the union of the distinct primes of their factorizations. Find the least common multiple of 180 and 28. contradiction and therefore 2 is irrational. (Proof: Suppose 2 = m/n where (m. ﬁnd the positive generator of the subgroup generated by {180. and if n is a multiple of a and b. p2 . Here are some of the basic properties of the integers in this light.. then their least common multiple is just their product. k. 1 k Let vi be the maximum of si and ti . then (a. Thus a = ±ps1 · · · psk where 0 ≤ si and 1 k b = ±pt1 · · · ptk where 0 ≤ ti . This is a 1. . Then integer..Chapter 1 Background 17 Now that we have unique factorization and part 3) above.) √ √ 2 is irrational.28}.

You move the top card to the bottom. Now let b = an + an−1 + · · · + a0 . and its positive generator is the gcd of the elements of S. its nth root is an irrational number. it says that [a] = [b] in Z3 . a0 = an 10n + an−1 10n−1 + · · · + a0 where 0 ≤ ai ≤ 9. Show that the set of all linear combinations of the elements of S is a subgroup of Z. 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. Exercise Show that if the nth root of an integer is a rational number. Gm is a subgroup of Z. then the second card to the bottom. and this will be the selected card. More generally. Exercise Show that a positive integer is divisible by 3 iﬀ the sum of its digits is divisible by 3. and show that 3 divides a and b with the same remainder. Take the six cards face down in your left hand and the selected card in your right hand. . Thus if p is a prime. an } is a ﬁnite collection of integers with |ai | > 1 for 1 ≤ i ≤ n. Express the gcd and the lcm in terms of the prime factorizations of the ai . Exercise Show that the gcd of S = {90. let a = an an−1 . There is a unique positive real number x with xn = c.. . n2 . where it is shown that any non-zero element of Z7 has order 7. That is. Ask your friends to give you a number between one and seven (not allowing one). it will be more transparent later on.. and ﬁnd integers n1 . suppose c and n are integers greater than 1. 42} is 2.18 Background Chapter 1 Exercise We have deﬁned the greatest common divisor (gcd) and the least common multiple (lcm) of a pair of integers. Put your hands behind your back and place the selected card on top. and bring the seven cards in front in your left hand. a2 . Show that if x is rational. and announce you will place the selected card in with the other six. Although this is a straightforward exercise in long division. 70. moving the top two cards to the bottom and turning the third card face up on top. Now let G = (90Z) ∩ (70Z) ∩ (42Z) and ﬁnd the positive integer n with G = nZ. Now suppose n ≥ 2 and S = {a1 . In the language of the next chapter. but they are not to know where. Deﬁne the gcd and the lcm of the elements of S and develop their properties.. . n3 such that 90n1 + 70n2 + 42n3 = 2. Then repeat the process.. . then it itself is an integer. Also ﬁnd the lcm of the elements of S. G2 . then it is an integer. leaving it face up on top. Magic? Stay tuned for Chapter 2. and then you turn over the third card. Exercise Show that if each of G1 . Suppose they say three. .. then G = G1 ∩ G2 ∩ · · · ∩ Gm is also a subgroup of Z. Continue until there is only one card face down.

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

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

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

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

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

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 −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, we used H 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. This is really much 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) 2) 3) 4) 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.

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 bijection is given by F (Ha) = a−1 H. 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 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. 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. 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.

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

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

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

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

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

2. ak ).. bk } is the same collection in some diﬀerent order.) Every permutation can be written (non-uniquely) as the product of transpostions. and is denoted by (a1 . then g ◦ f is 3) 4) 5) 6) .. Theorem 1) 2) Disjoint cycles commute.ak−1 ak is called a k-cycle.. For these we have a special notation. (This is easy. They are all easy except 4).. . ak } is a collection of distinct integers Deﬁnition with 1 ≤ ai ≤ n.ak a1 A 2-cycle is called a transposition. Properties 8)... 3) is an even permutation.....) Every permutation can be written uniquely (except for order) as the product of disjoint cycles. In the other case.. b2 . 2)(1. g ∈ Sn .. ) The parity of the number of these transpositions is unique. 9).. Deﬁnition Listed here are seven basic properties of permutations. which is rather delicate. a1 a2 . ak the matrix represents f ∈ Sn deﬁned by (ai )f = bi for 1 ≤ i ≤ k. .. f is said to be an even permutation. A k-cycle is even (odd) iﬀ k is odd (even). Suppose f.. then p is even iﬀ q is even. . and {b1 . an ) = (a1 . b1 b2 . (Proof: (a1 .. Then a1 a2 . and 10) are listed solely for reference. bk and (a)f = a for all other a. a2 )(a1 .. f is an odd permutation. For example (1. The cycles (a1 .32 be represented by a matrix. .. c ) are disjoint provided ai = cj for all 1 ≤ i ≤ k and 1 ≤ j ≤ . . The composition of two permutations is computed by applying the matrix on the left ﬁrst and the matrix on the right second. a2 . a2 a3 .. a2 .. 3) = (1.. In this case. There is a special type of permutation called a cycle. 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. {a1 .. Groups Chapter 2 Suppose 1 < k ≤ n.. ak ) and (c1 .. . an ). (This is obvious. a3 ) · · · (a1 ..

If σ is a permutation on S and p = p(x1 . σ(p) has (xi − xj ) or (xj − xi ) as a factor.. and Sn /An ≈ Z2 . i. then σ(p) = x2 x2 + x2 x3 .7. n)} generates Sn . p = (x1 − x2 )(x1 − x3 )(x2 − x3 ). Its kernel (the subgroup of even permutations) is denoted by An and is called the alternating group. (Of course there are subgroups of Sn which cannot be generated by two elements).3.. . Suppose S = {1.. 2. if n = 3. the student may skip the proof until after that chapter. 10) Sn can be generated by two elements. Since polynomials will not be introduced until Chapter 3. {(1.. then for each 1 ≤ i. 6 5 4 3 1 7 2 Write (1.3. Note that if σ1 and σ2 are permutations. Now let p be the product of all (xi −xj ) where 1 ≤ i < j ≤ n.. t must be even.. deﬁne σ(p) to be the polynomial p(x(1)σ .1) as the product of disjoint cycles.7.1)(1. . For any n ≥ 3. They are presented here merely for reference. Proof of 4) The proof presented here uses polynomials in n variables with real coeﬃcients.7)(2.. j ≤ n with i = j. Thus if σ(p) = p.. An is simple. An is generated by its 3-cycles.7)(2..5.6. Thus An is a normal subgroup of index 2. A careful examination shows that if σi is a transposition. Exercise 1) 1 2 3 4 5 6 7 as the product of disjoint cycles.5. 2).Chapter 2 Groups 33 odd. (Obvious. σ = σ1 ·σ2 ···σt . (For example. x(n)σ ). Thus σ(p) = ±p. σi (p) = −p. then g ◦ f is even.4) as the product of disjoint cycles. and if σ(p) = −p. Any permutation σ is the product of transpositions... Write (3. has no proper normal subgroups. t must be odd. Which of these permutations are odd and which are even? Write .4)(3.e.. 1 σ2 (σ1 (p)) = (σ1 ·σ2 )(p). n}. (1.) If σ is a permutation on S. xn ) is a polynomial in n variables. . 2). Thus if p = x1 x2 + x1 x3 . In fact. If f and g are both even or both odd. ..6.) 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. The following parts are not included in this course. 8) 9) For any n = 4. and σ is the trans2 position (1.

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

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

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

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

. Rn is a ring. ¯ ¯ ¯ ¯ ¯ (−a) · b = a · (−b) = −(a · b). a polynomical ring in n variables. b ∈ R and n. This scalar multiplication can be written on the right or left. In the next chapter. Under these operations. Theorem 1) 2) Suppose R is a ring and a. 2 = 1 + 1. Also if R is any commutative ring. (This follows from the distributive law and the previous theorem. the real numbers R. the integers mod n. ¯ Theorem 0 can never be a unit. Rn is never commutative. More 2) . since R is an additive abelian group. Theorem 1) Suppose a. Therefore 1 = 0. 1 is always a unit. it has a scalar multiplication over Z. (na) · (mb) = (nm)(a · b). i. that is. ¯ Of course.e. m ∈ Z. and the complex numbers C.38 Rings Chapter 3 rational numbers Q. . Now suppose R is any ring. xn ]. Recall that. we will deﬁne R[x1 . operations of addition and multiplication of matrices will be deﬁned. and the next theorem shows it relates nicely to the ring multiplication. This is a basic example of a non-commutative ring.) Let n = n1. It will be shown later that Zn . x2 . n ≥ 1. even if R is commutative. n may be 0 even though n = 0. scalar ¯ ¯ ¯ ¯ ¯ ¯ multiplication by n is the same as ring multiplication by n. They should be worked as exercises. . b ∈ R. and Rn is the collection of all n×n matrices over R. a · 0 = 0 · a = 0. For example. a−1 is also a ¯ −1 ¯ unit with (a )−1 = a. If a is a unit. na = an. The next two theorems show that ring multiplication behaves as you would wish it to. ¯ ¯ 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.. has a natural multiplication under which it is a commutative ring. If n > 1. . The product of units is a unit with (a · b)−1 = b−1 · a−1 . Then na = n · a.

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

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

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

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

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

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

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

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

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

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

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

all the non-zero elements of R have the same order.. there is one and only one ring homomorphism f : Z → R. and thus f (1) is a group generator. .. nt are integers. (For a useful and elegant generalization of this theorem. Thus f is injective iﬀ R has characteristic 0 iﬀ 1 has inﬁnite order. Since their least common multiple is n. Deﬁnition Suppose R is a ring and f : Z → R is the natural ring homomorphism f (m) = m1 = m. the order of f (1) is n.. (See the fourth exercise on page 36. where n = n1 n2 · · nt . 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. Proof We wish to show that the order of f (1) is n. Theorem If R is a ring. ¯ It is an interesting fact that.. It will now be shown that they are also isomorphic as rings. the kernel of f is nZ. There is an exercise in Chapter 2 to show that Znm and Zn × Zm are isomorphic as groups. see the Appendix..) Exercise Show that if a is an integer and p is a prime.50 The Chinese Remainder Theorem Rings Chapter 3 Suppose n and m are relatively prime integers with n. if R is a domain. . but it shows that in ring theory. .. The element f (1)m = ([1]. . Thus Zn and Zn1 × · · ×Znt are isomorphic rings. the ring of integers is a “cornerstone”. Characteristic The following theorem is just an observation. ft ) : Z → Zn1 × · · ×Znt is surjective. . Furthermore. It is given by f (m) = m1 = m. nj ) = 1 for all i = j. The non-negative integer n with ker(f ) = nZ is called the charac¯ ¯ teristic of R.) Then the ring homomorphism f = (f1 . (Note that the bracket symbol is used ambiguously. ¯ If f is not injective.) Theorem Suppose n1 . nt . m > 1. and (ni .... [1])m = ([m]. each ni > 1. it has the same last digit as b5 ... then [a] = [ap ] in Zp (Fermat’s Little Theorem). Let fi : Z → Zni be deﬁned by fi (a) = [a]. the characteristic of R is the order of 1. [m]) is zero iﬀ m is a multiple of each of n1 . and thus f is surjective. Use this and the Chinese Remainder Theorem to show that if b is a positive integer. .

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

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

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

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

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

At ∈ Rn.n to Rn. Triangular Matrices If A ∈ Rn . ai.56 Transpose Matrices Chapter 4 Notation For the remainder of this chapter on matrices. j) term is the (j. Theorem If A ∈ Rn is strictly upper (or lower) triangular.j = 0 for all i = j. Of course. then A is invertible iﬀ At is invertible. Theorem 1) 2) 3) 4) 5) (At )t = A (A + B)t = At + B t If c ∈ R. Deﬁnition If T is any ring. I − B is invertible.j = 0 for all i ≥ j (all j ≥ i). an element t ∈ T is said to be nilpotent provided ∃n n such that t = 0. Transpose is a function from Rm. suppose R is a commutative ring. (Ac)t = At c (AB)t = B t At If A ∈ Rn . A is diagonal if it is upper and lower triangular. At is also square. Exercise Characterize those invertible matrices A ∈ R2 which have A−1 = At . i.e. Rn is non-commutative. So row i (column i) of A becomes column i (row i) of At . In this case (A−1 )t = (At )−1 . (1 − t) is a unit with inverse 1 + t + t2 + · · · + tn−1 . Note that if A is upper (lower) triangular. In this case. The geometry of this theorem will become transparent later in Chapter 5 when the matrix A deﬁnes an R-module endomorphism on Rn . then A is upper (lower) triangular provided ai.j = 0 for all i > j (all j > i). If A ∈ Rm.n . If A is a square matrix. Proof The way to understand this is just multiply it out for n = 2 and n = 3. A is strictly upper (lower) triangular provided ai. Show that they form a subgroup of Gl2 (R). I = I t = (AA−1 )t = (A−1 )t At . then At is lower (upper) triangular. . Proof of 5) Suppose A is invertible. If A is an n-dimensional row vector. for n > 1. Thus if T = Rn and B is a nilpotent matrix.m is the matrix whose (i. then An = 0.. i) term of A. then At is an n-dimensional column vector.m .

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

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

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

For each n ≥ 1 and each commutative ring R.) Suppose B ∈ Rm is invertible. c d Deﬁnition Let A = (ai. For n = 1. Then f is a group homomorphism and also f (Dc) = f (D)c for any c ∈ R. and sign(σ) = −1 if σ is an odd permutation. 2. (Note that here we are writing the permutation functions as σ(i) and not as (i)σ. | (a) | = a.) all σ . this says that f is an R-module homomorphism.σ(n) . In this section we simply present the basic properties. The determinant is deﬁned by | A |= sign(σ) a1..σ(1) · a2. AX = C has a solution iﬀ C ∈ image(f ). let sign(σ) = 1 if σ is an even permutation. (See part 7 of the section on Homomorphisms in Chapter 2. Its solution set is ker(f ). the solution set is the coset D + ker(f ) in Rn . this agrees with the deﬁnition above. Then AX = C and (BA)X = BC have the same set of solutions. In the language of the next chapter. n). The proper development of this concept requires a study of multilinear forms. Determinants The concept of determinant is one of the most amazing in all of mathematics. 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.j ) ∈ Rn . Thus we may perform any row operation on both sides of the equation and not change the solution set. For n = 2.60 Matrices Chapter 4 Deﬁne f : Rn → Rm by f (D) = AD..σ(2) · · · an. Check that for n = 2. = ad − bc. . If A ∈ Rm is invertible.. The following theorem summarizes what we already know about solutions of linear equations in this setting. then AX = C has the unique solution X = A−1 C. If σ is a permutation on (1. determinant is a function from Rn a b to R. If D ∈ Rn is one solution. Theorem 1) 2) AX = 0 is called the homogeneous equation. which is given in Chapter 6.

Since R is commutative.σ(1) · a2. ¯ Proof For simplicity.1 · aτ (2). . τ is the inverse of σ and thus there are two ways to deﬁne determinant.2 · · · aτ (n).1 · aτ γ(2). . You may view an n × n matrix A as a sequence of n column vectors or as a sequence of n row vectors. . It follows that the determinant of a matrix is equal to the determinant of its transpose.n = aτ γ(1). 2. . Ar−1 . etc. Summary Determinant is a function d : Rn → R. A2 . . An )| Proof This is immediate from the deﬁnition of determinant and the distributive law of multiplication in the ring R.n and this summation has n! terms and Theorem n! is an even number. If two columns of A are equal. then |A| = 0.n . c ∈ R. ¯ Theorem Suppose 1 ≤ r ≤ n. Cr ∈ Rn.2 · · · aτ (n). A1 = A2 .1 · · · aτ (n). . .n . Therefore |A| = 0. . . .σ(n) contains exactly one factor from each row and one factor from each column. . An )| + c|(A1 .σ(2) · · · an.2 · · · aτ (n). Then for any τ .5 . Corollary |A| = |At |.σ(1) · a2. . the second from the second column. . i.e. We wish to show that τ = σ −1 and thus sign(σ) = sign(τ ). . . and since sign(τ )=−sign(τ γ). Anyway. . aAr + cCr . suppose σ(2) = 5.5 . An ) where each Ai ∈ Rn. . . Let γ be the transposition which interchanges one and two. . To reduce the abstraction. and a. Then the ﬁrst expression will contain the factor a2. n) such that a1. This means we write the matrix A as A = (A1 . .σ(2) · · · an. This means that there is a permutation τ on (1.σ(n) = aτ (1). all τ .2 · · · aτ γ(n). it will appear as aτ (5). In the second expression. .. Ar−1 . Then |(A1 . . these pairs cancel in the summation.1 = Rn . we may rearrange the factors so that the ﬁrst comes from the ﬁrst column. In the language used in the Appendix. Now |A| = sign(τ )aτ (1). and so τ (5) = 2. .1 · aτ (2). aτ (1). Here we will use column vectors. . a1. assume the ﬁrst two columns are equal. Ar+1 . An )| = a|(A1 .σ(1) · · · an. This pairs up the n! terms of the summation.Chapter 4 Matrices 61 For each σ.n . The next two theorems say that d is skew-symmetric.1 · aτ (2). .1 .σ(n) = all τ sign(τ )aτ (1). . . Cr . Ar+1 . . the two previous theorems say that d is an alternating multilinear form. Theorem |A| = all σ sign(σ)a1.

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

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

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

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

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

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

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

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

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

then T . then f is also a Z-module homomorphism. Proof The deﬁnitions are the same except expressed in diﬀerent language. This makes R a right R-module denoted by RR (or just R). In ¯ ¯ other words. Exercise If R is a subring of a ring T .n an R-module. R as a right R-module Let M = R and deﬁne scalar multiplication on the right by ar = a · r. is an R-module. Then f = g. ∃! homomorphism h : R → M with h(1) = m.Chapter 5 Linear Algebra 71 Abelian groups are Z-modules On page 21. etc. This is the same as the deﬁnition before for Rn when n = 1. While group theory in general is quite separate from linear algebra. must f be a Q-module homomorphism? Homomorphisms on Rn Rn as an R-module In Chapter 4 it was shown that the additive abelian group Rm. g : R → M are homomorphisms with f (1) = ¯ g(1). Proof Suppose f (1) = g(1). Note that this is the only way M can be a Zmodule.1 . If f : Q → R is a Z-module homomorphism. Summary Additive abelian groups are “the same things” as Z-modules. and if M is abelian. Theorem Suppose N is a subset of R. the properties are satisﬁed to make M a Z-module. Given m ∈ M . Thus . In particular. h : R → M deﬁned by h(r) = mr is a homomorphism. because a1 = a. Then f (r) = f (1 · r) = f (1)r = g(1)r = g(1 · r) = ¯ ¯ ¯ ¯ ¯ ¯ g(r). R is a Q-module. The properties listed there were exactly those needed to make Rm. We begin with the case n = 1. If m ∈ M . with scalar multiplication deﬁned by ring multiplication. scalar multiplication is just ring multiplication. HomR (R. M ) ≈ M . That is. the study of additive abelian groups is a special case of the study of R-modules. Then N is a submodule of RR (R R) iﬀ N is a right (left) ideal of R. it is shown that any additive group M admits a scalar multiplication by integers. Furthermore. Theorem Suppose M = MR and f. Of particular importance is the case Rn = R ⊕ · · ⊕R = Rn. a2 = a + a. if f : M → N is a group homomorphism of abelian groups.n admits a scalar multiplication by elements in R.

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

then f + g is given by the matrix A + C. That is. Rn ) and Rn are isomorphic as rings. This corollary shows one way non-commutative rings arise. These properties are made explicit in the next two theorems. In the case where the domain and range are the same. The automorphisms correspond to the invertible matrices. We now return to the general theory of modules (over some given ring R).n .n are isomorphic as additive groups. Matrices over R give R-module homomorphisms! Furthermore. if m1 . Conversely. deﬁne A ∈ Rm. Theorem If f : Rn → Rm is the homomorphism given by A ∈ Rm. Then f : Rn → Rm deﬁned by f (B) = AB is a homomorphism with f (ei ) = column i of A. Thus HomR (Rn . It is the matrix which represents the composition of the functions.n and g : Rm → Rp is the homomorphism given by C ∈ Rp. Rm ) and Rm. g : Rn → Rm are given by matrices A. they are isomorphic as R-modules. Theorem Suppose A = (ai. . then g ◦ f : Rn → Rp is given by CA ∈ Rp.j ) ∈ Rm. The previous theorem reveals where matrix multiplication comes from. Even if R is commutative. it is one of the great classical facts of linear algebra.n . If R is commutative. Then f deﬁned by f (B) = AB is the unique homomorphism from Rn to Rm with f (ei ) = mi . C ∈ Rm. Rn is never commutative unless n = 1. . . Rm ) ≈ Rm. and multiplication of matrices corresponds to composition of homomorphisms.n . . composition of homomorphisms corresponds to multiplication of matrices. mn ∈ Rm .n . Theorem If f.n to be the matrix with column i = mi . Proof This is just the associative law of matrix multiplication. C(AB) = (CA)B. we have the following elegant corollary. Even though this follows easily from the previous theorem and properties of matrices.Chapter 5 Linear Algebra 73 Now let’s examine the special case M = Rm and show HomR (Rn .m . . namely as endomorphism rings. Corollary HomR (Rn . addition of matrices corresponds to addition of homomorphisms.

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

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

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

l) = k + l. because L ≈ M/K. Theorem Consider M1 = M1 ⊕0 as a submodule of M1 ⊕M2 . Exercise R is a module over Q. Thus f is an isomorphism iﬀ K + L = M and K ∩ L = 0. . f (k. This abuse of notation allows us to ¯ avoid talking about “internal” and “external” direct sums. Show (M ⊕ N )/(A ⊕ B) is isomorphic to (M/A) ⊕ (N/B). it need not be unique. it will be easy. If such an L exists. and Q ⊂ R is a submodule. This is exactly what you would expect. and the next theorem is almost as intuitive. ¯ Suppose M is a module and K = {(m. and A ⊂ M . m) : m ∈ M } ⊂ M ⊕ M . In this case we write K ⊕ L = M . this is the same as there exists a submodule L with K + L = M and K ∩ L = 0. Is Q a summand of R? With the material at hand. Then the projection ¯ map π2 : M1 ⊕ M2 → M2 is a surjective homomorphism with kernel M1 . −a) : a ∈ K ∩ L}. Suppose R is a commutative ring. Deﬁnition Suppose K is a submodule of M . but it will be unique up to ¯ isomorphism. Linear Algebra 77 Suppose M and N are R-modules. Then the image of f is K + L and the kernel of f is {(a. B ⊂ N are submodules. The statement that K is a summand of M means ∃ a submodule L of M with K ⊕ L = M . Show that M ⊕ N is isomorphic to Exercise Suppose M and N are R-modules. and n ≥ 1. Theorem Suppose K and L are submodules of M and f : K ⊕ L → M is the natural homomorphism. Before deﬁning summand.Chapter 5 Exercise N ⊕ M. Deﬁne Exercise a function ∝: HomR (Rn . M and 0 are always summands of M . this is not an easy question. here are two theorems for background. According to the previous theorem. Of course. M is an R-module. M ) → M n which is a R-module isomorphism. Show Exercise K is a submodule of M ⊕ M which is a summand. Thus (M1 ⊕ M2 )/M1 is isomorphic to M2 . Later on. Summands One basic question in algebra is “When does a module split as the sum of two modules?”.

Let R be a commutative ring and T = Rn . Let S be the sequence {st }t∈T = {st }.. Let SR be the set of all linear combinations st1 r1 + · · +stn rn . Rn ). then S is said to be a basis or free basis for M . Deﬁnition Suppose M is an R-module. such that the linear combination st1 r1 + · · +stn rn = 0. 2.. to avoid dizziness.. not all zero. st ∈ M . then S is dependent. rn in R. . In this case any v ∈ M can be written uniquely as a linear combination of elements in S. n} and ⊕Rt = Rn . When is nZmn a summand of Zmn ? Chapter 5 Exercise If T is a ring. Theorem For each t ∈ T . Then {ec }c∈T is a basis for ⊕Rt called ¯ ¯ the canonical basis or standard basis. ¯ S is independent.. . Also if ∃ distinct ¯ elements t1 and t2 in T with st1 = st2 . let Rt = RR and for each c ∈ T . Generating Sets. then S is dependent.. M is said to be a free R-module. Show Rn is a left T -module and ﬁnd HomT (Rn . . Independence. let ec ∈ ⊕Rt = t∈T Rt be ec = {rt } where rc = l and rt = 0 if t = c. m > 1. . These concepts work ﬁne for an inﬁnite index set T because linear combination means ﬁnite linear combination. Otherwise. There is a exercise on page 57 to show that the center of T is the subring of scalar matrices. and elements r1 . SR is a submodule of M called the submodule generated by S. deﬁne the center of T to be the subring {t : ts = st for all s ∈ T }. and Free Basis This section is a generalization and abstraction of the brief section Homomorphisms on Rn . tn in T .. The statement that S is dependent means ∃ a ﬁnite number of distinct elements t1 . However. You might try ﬁrst the case T = {1.. The next two theorems are obvious. Note that if some st = 0.78 Exercise 1) 2) 3) 4) Linear Algebra Answer the following questions about abelian groups. the student should consider the case where T is ﬁnite. T is an index set. Is 2Z a summand of Z? Is 4Z8 a summand of Z8 ? Is 3Z12 a summand of Z12 ? Suppose n. and for each t ∈ T .. If ∃ a basis for M . If S is independent and generates M . except for the confusing notation.

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

Exercise 2 1 0 and f: Z3 → Z2 be the group homo3 2 −5 morphism deﬁned by A. . Exercise 1) 2) Suppose R is a commutative ring and v ∈ R. Suppose A ∈ Rm... i... v = 0. injective homomorphisms. vn ∈ Rm be the columns of A. ¯ v is independent iﬀ v is v is a basis for R iﬀ v generates R iﬀ v is . linear independence. That is. then f is an isomorphism. en } is a free basis for Rn . It shows that certain concepts about matrices.  represent an element of Rn and C =  . Note that 2) here is essentially the ﬁrst exercise for the case n = 1. Use the fact that {e1 . Relating these concepts to matrices The theorem stated below gives a summary of results we have already had. A ∈ Rn . This is a key theorem in linear algebra. although it is usually stated only for the case where R is a ﬁeld. show A is invertible.  λn cm . f (ei ) = vi     λ1 c1     = column i of A. and the homomorphism f : Rn → Rn deﬁned by f (B) = AB is surjective.e. Also ﬁnd a non-zero element of kernel(f ). but still informative. i. if f : R → R is a surjective R-module homomorphism. ..80 Linear Algebra Chapter 5 Exercise Suppose R is a commutative ring. f (B) = AB. . A = The next exercise is to relate properties of R as an R-module to properties of R as a ring. The next exercise is routine...e. Let λ =  . Find a non-trivial linear combination of the columns of A which is 0.n and f : Rn → Rm is the homomorphism associated with A. Let v1 . Show f is an isomorphism. are all the same — they are merely stated in diﬀerent language. Let R = Z. i. and solutions of equations.e.

..Chapter 5 Linear Algebra 81 represent an element of Rm . . Theorem 1) f (λ) is the linear combination of the columns of A.. i.. 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.n be the rows of A. | At | is a unit in R. wn ∈ Rn = R1. . . vn } generates Rm iﬀ f is surjective iﬀ (for any C ∈ Rm .. {v1 . Let v1 . AX = C has a solution).e. and w1 . A ∈ Rn . . i. 1) 2) 3) 4) 5) 2t ) 3t ) f is an automorphism.. {v1 .. Theorem Suppose R is a commutative ring. f (λ) = f (e1 λ1 + · · + en λn ) = v1 λ1 + · · +vn λn .. Now we prove something more substantial. {v1 . Later on we will prove that if R is a ﬁeld... vn ∈ Rn be the columns of A. and f : Rn → Rn is deﬁned by f (B) = AB.. vn } is a basis for Rm iﬀ f is an isomorphism iﬀ (for any C ∈ Rm . f is surjective. namely that if f : Rn → Rn is surjective. wn } is a basis for Rn . . | A | is a unit in R. .... At is invertible. then f is injective... AX = C has a unique solution). vn } generates Rn . vn } is a basis for Rn . Then the following are equivalent. . {w1 . {v1 . . . 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. A is invertible. So far in this chapter we have just been cataloging. {v1 .. injective implies surjective..

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

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

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

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

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

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

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

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

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

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

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. ¯ Extend v1 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. Do not use 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.) 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

       

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

       

is strictly upper triangular.

Exercise 

Suppose F is a ﬁeld, A ∈ F3 is a lower triangular matrix of rank 2,  0 0 0 and B =  1 0 0 . Using conjugation by elementary matrices, show there is an   0 1 0 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 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 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. Suppose λ1 , .., λk are distinct eigenvalues of f , and vi is an eigenvector Theorem 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 C −1 AC −1 0 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 endomorphism. Show there is a basis for V so that the x3 and is thus a nilpotent      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. ¯

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

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

. u · v = u v cos Θ where Θ is the angle between u . It is given by the formula (v1 r1 + · · +vn rn ) · (v1 s1 + · · +vn sn ) = r1 s1 + · · +rn sn . u ∈ Rn . .... . vn } is orthogonal provided vi · vj = 0 when i = j. sn ) = r1 s1 + · · +rn sn . Then 0 = (v1 r1 + · · +vn rn ) · vi = ri (vi · vi ) ¯ and thus ri = 0. ¯ vr = v | r |. Convention Rn will be assumed to have the standard inner product deﬁned by (r1 . Furthermore v1 vn Theorem Proof Suppose v1 r1 + · · +vn rn = 0. | u · v |≤ u v . A sequence {v1 . vn }. Thus S is independent. . i.. en } will be called the canonical or standard orthonormal basis..e. u+v ≤ u + v . The following properties hold. . Deﬁne the norm or length of a vector v by v = v · v. then v1 vn is orthonormal.···. S = {e1 . Suppose V is an IPS. as is shown by the following theorem. + 2(u · v) + v 2 ≤ u 2 + Deﬁnition An Inner Product Space (IPS) is a real vector space with an inner product. rn ) · (s1 . . Theorem If u.. (Schwarz’ inequality) (The triangle inequality) 2 Proof of 4) 2 u v + v 2 u + v 2 = (u + v) · (u + v) = u = ( u + v )2 .Chapter 5 Linear Algebra 99 Theorem √ Suppose V has an inner product. The second statement is transparent. Theorem Suppose V is a real vector space with a basis S = {v1 .j for 1 ≤ i. If S = {v1 . vi · vj = δi. Then there is a unique inner product on V which makes S an orthornormal basis. The next theorem shows that this inner product has an amazing geometry. j ≤ n. The sequence is orthonormal if it is orthogonal and each vector has length 1. 1) 2) 3) 4) v = 0 iﬀ v = 0. ... ... It is easy to deﬁne an inner product. vn } is an orthogonal sequence of non-zero vectors.. S is independent..

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

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

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

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

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

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

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

This classical and very powerful technique allows an easy proof of the canonical forms. everything is right down to the nub. The basic theorem of this chapter is that if R is a Euclidean domain and M is a ﬁnitely generated R-module.e. Thus if M is torsion free. Suppose R is a commutative ring. then there is a basis for V so that the matrix representing T is in Jordan canonical form. There is a basis for V so that the matrix representing T is in Rational canonical form. A proof of this. An R-module M is said to be cyclic if it can be generated by one element. M ≈ R/I where I is an ideal of R. The style is the same as before. Now F [x] is a Euclidean domain and so VF [x] is the sum of cyclic modules. A matrix in Jordan form is a lower triangular matrix with the eigenvalues of T displayed on the diagonal.e. i. If T : V → V is a linear transformation.. In this appendix. In the chapter on matrices. If the characteristic polynomial of T factors into the product of linear polynomials. These are independent sections added on at the end. is given in this chapter. so this is a powerful concept. This always holds if F = C. ﬁnitely generated abelian groups are the sums of cyclic groups. it is stated without proof that the determinant of the product is the product of the determinants. Since Z is a Euclidean domain.. then V becomes an F [x]-module by deﬁning vx = T (v).Chapter 6 Appendix The ﬁve previous chapters were designed for a year undergraduate course in algebra. which depends upon the classiﬁcation of certain types of alternating multilinear forms. 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. i. Now suppose F is a ﬁeld and V is a ﬁnitely generated F -module. then M is the sum of cyclic modules. The ﬁnal section gives the fundamentals of dual spaces. 107 . 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 A1 A2 · · · An = A1 ∩ A2 ∩ · · · ∩ An . A2 . it is the image of r1 b1 +r2 b2 +···+rn bn = r1 (1 −a1 )+r2 (1 −a2 )+···+rn (1 −an ). . Show A1 ∩ A2 ⊂ A1 A2 . Chinese Remainder Theorem Suppose A1 . Proof for n = 2. Here it is presented in full generality. Then π : R → R/A1 ×R/A2 ×···×R/An is a surjective ring homomorphism with kernel A1 ∩ A2 ∩ · · · ∩ An . Proof There exists ai ∈ Ai and bi ∈ A1 A2 ···Ai−1 Ai+1 ···An with ai +bi = 1. 0. then the following are equivalent. An are pairwise comaximal ideals of R. 1i .. ∃ a1 ∈ A1 and a2 ∈ A2 with a1 + a2 = 1. ¯ π(A) = R/B where π : R → R/B is the projection.. If A and B are ideals of a ring R. rn + An ) is an element of the ¯ range. Am are ideals of R... ∃ a ∈ A and b ∈ B with a + b = 1. . . Deﬁnition Suppose R is a ring and A1 . 0).108 Appendix The Chinese Remainder Theorem Chapter 6 On page 50 in the chapter on rings.. Am and B are ideals of R with Ai and B comaximal for each i. . A and B are comaximal. The product A1 A2 · · · Am is the set of all ﬁnite sums of elements a1 a2 · · · am with ai ∈ Ai .. then c = c(a1 + a2 ) ∈ A1 A2 .. ¯ If c ∈ A1 ∩ A2 . Thus A1 ∩ A2 ∩ · · · ∩ Am and B are comaximal.. 0. Proof Consider π : R → R/B. . Then π(A1 A2 · · · Am ) = π(A1 )π(A2 ) · · · π(Am ) = (R/B)(R/B) · · · (R/B) = R/B..... then A1 A2 · · · Am and B are comaximal. . .. r2 + A2 . the theorem holds even for non-commutative rings. If (r1 + A1 . . A2 . Note ¯ that π(bi ) = (0.. A2 . the Chinese Remainder Theorem was proved for the integers. Deﬁnition Theorem 1) 2) 3) Ideals A and B of R are said to be comaximal if A + B = R. ¯ ¯ ¯ Theorem If R is commutative and A1 . A2 .... An are pairwise comaximal ideals of R. Note that the sum and product of ideals are ideals and A1 A2 · · · Am ⊂ (A1 ∩ A2 ∩ · · · ∩ Am ). Surprisingly.. with each Ai = R. Theorem If A1 . Then the sum A1 + A2 + · · · + Am is the set of all a1 + a2 + · · · + am with ai ∈ Ai ..

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

1) 2) 3) Suppose R is a domain and a. i. If n ∈ Z is not zero. a = bc ⇒ b or c is a unit. bi and cσ(i) are associates for every i.. it is called a prime element. while the only ¯ associate of 0 is 0 itself. Note If a is a prime and a|c1 c2 · · · cn .. the associates of 1 are the units of R. i. there is a bijection from the associate classes of R to the ideals of R. a is prime if it generates a prime ideal. The following theorem is elementary.e. Appendix Chapter 6 Examples If R is a domain. Note a is prime ⇒ a is irreducible. but it shows how associates ﬁt into the scheme of things. Note If a ∼ b. Deﬁnition 1) 2) Suppose R is a domain and a ∈ R is a non-zero non-unit. then a is irreducible (prime) iﬀ b is irreducible (prime). In other words. aR = bR. if a = b1 b2 ···bn = c1 c2 ···cm with each bi .110 and b are said to be associates.. Then the following are ¯ a ∼ b. a is irreducible if it does not factor. Theorem equivalent. If each cj is irreducible.e. ci prime. i.e. This is immediate from the deﬁnitions. . Thus if R is a PID. ¯ ¯ If F is a ﬁeld and g ∈ F [x] is a non-zero polynomial. b ∈ (R − 0). then a ∼ ci for some i. This follows from the deﬁnition and induction on n. then the associates of g are all cg where c is a non-zero constant. then n = m and for some permutation σ of the indices. If an element generates a non-zero prime ideal. Parts 1) and 3) above show there is a bijection from the associate classes of R to the principal ideals of R. An element a divides b (a|b) if ∃! c ∈ R with ac = b. Theorem Factorization into primes is unique up to order and associates. a|bc ⇒ a|b or a|c. a|b and b|a. then a|ci for some i. then au is irreducible (prime). then its associates are n and −n. if a is irreducible (prime) and u is a unit.

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

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

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

This is used below to show that if R is a PID.. i. Show R/(¯) is isomorphic to R[y. ¯ Exercise In Group Theory If G is an additive abelian group. i. Show that H is maximal iﬀ G/H ≈ Zp for some prime p. then every submodule of Rn is free. and g : B → C is a surjective homomorphism with kernel K. Theorem 1 Suppose R is a ring. if B/K is free. 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. y. g has a right inverse. y. then there is a maximal subgroup H of Q which contains a. Show R is isomorphic to Z[x]/(x2 − 5). K must be a summand of B. which is not a domain.) Proof Suppose 1) is true. B and C are R-modules. and R/(2) is isomorphic to Z2 [x]/(x2 − [5]) = Z2 [x]/(x2 + [1]). (h is called a splitting map. z]. z]/(x2 − yz). 1) 2) K is a summand of B.. If i : L → B is inclusion. In particular x · x = y · z are two distinct irreducible factorizations ¯ ¯ ¯ ¯ x of x2 . Show x2 − yz is irreducible and thus prime in R[x. z]. let u ∈ R be the coset containing u. Q contains no maximal subgroups. K + h(C) = B and K ∩ h(C) = 0. . K is a summand of B provided ∃ a submodule L of B with K + L = B and K ∩ L = 0. As deﬁned in the chapter on linear algebra. For simplicity. which is not a domain. a subgroup H of G is said to be maximal if H = G and there are no subgroups properly between H and G. z]/(yz). consider the case G = Q.114 Appendix Chapter 6 irreducible factorizations of 4. Then the following are equivalent. In particular. y. Exercise Let R = R[x. Splitting Short Exact Sequences Suppose B is an R-module and K is a submodule of B.e. If u ∈ R[x. Show R ¯ is not a UFD. Then h is injective. Then (g|L) : L → C is an isomorphism.e. suppose ∃ a submodule L of B with K ⊕ L = B. ¯ Thus K ⊕ h(C) = B. Now suppose 2) is true and h : C → B is a right inverse of g. ∃ a homomorphism h : C → B with g ◦ h = I : C → C. then h deﬁned by h = i ◦ (g|L)−1 is a right inverse of g. where (x2 − 5) represents the ideal (x2 − 5)Z[x]. Which one of the following is true? 1) 2) If a ∈ Q.

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

In either ¯ case. If π(A) = 0. and you don’t have to worry about it becoming obsolete. 0 −→ Rn−1 −→ Rn−1 ⊕ R −→ R −→ 0 0 −→ A ∩ Rn−1 −→ A −→ π(A) −→ 0 By induction.1. Z[i] = {a + bi : a. it is free of dimension 1 and the sequence splits by Theorem 1. This is one of the great classical theorems of abstract algebra. Show A is a free Z-module of dimension 1. However they are exactly what you need. (16. A is a free submodule of dimension ≤ n. Deﬁnition A domain R is a Euclidean domain provided ∃ φ : (R − 0) −→ N such ¯ that if a. b ∈ (R − 0). Euclidean Domains The ring Z possesses the Euclidean algorithm and the polynomial ring F [x] has the division algorithm. A ∩ Rn−1 is free of dimension ≤ n − 1.116 Appendix Chapter 6 Consider the following short exact sequences. where f : Rn−1 → Rn−1 ⊕ R is inclusion and g = π : Rn−1 ⊕ R → R is the projection. ¯ Examples of Euclidean Domains Z with φ(n) = |n|. then M is the sum of cyclic modules. The concept of Euclidean domain is an abstraction of these properties. If ¯ π(A) = 0. r ∈ R such that a = bq + r with r = 0 or φ(r) < φ(b). A ﬁeld F with φ(a) = 1 ∀ a = 0 or with φ(a) = 0 ∀ a = 0. b ∈ Z} = Gaussian integers with φ(a + bi) = a2 + b2 . Exercise Let A ⊂ Z2 be the subgroup generated by {(6. Here N will denote the set of all non-negative integers. If R is a Euclidean domain and M is a ﬁnitely generated R-module. then ¯ 1) φ(a) ≤ φ(ab). 2) ∃ q. The axioms are so miniscule that it is surprising you get this much juice out of them. not just the set of positive integers. 24). f π . then A ⊂ Rn−1 . ¯ ¯ F [x] where F is a ﬁeld with φ(f = a0 + a1 x + · · · + an xn ) = deg(f ). 64)}. and it is possible to just play around with matrices and get some deep results.

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

.i g2 + · · · + an.. consider matrices obtained from (ai. Proof By Theorem 3 in the section Splitting Short Exact Sequences..118 Appendix Chapter 6 φ values. .t where hi = a1. . Theorem 4 Suppose R is a Euclidean domain. Then d1 will divide the other entries in the ﬁrst row. The composition Rt −→ A −→ B −→ Rn ei −→ hi gi −→ ei ≈ ⊂ ≈ is represented by a matrix (ai. ¯ and di |di+1 for 1 ≤ i < t. at } for A and {b1 . ∃ invertible matrixes U ∈ Rn and V ∈ Rt such that . To consolidate this approach. bn } for B. else we could obtain an entry with a smaller φ value. Yet it must be a deep theorem because the next two theorems are easy consequences. .j ) ∈ Rn. let (bi.. This is an example of a theorem that is easy to prove playing around at the blackboard.. by row operations of type 3. Thus by column operations of type 3. ht }.. the entry d1 may be moved to the (1.j ) be one which has an entry d1 = 0 with φ(d1 ) a minimum. Thus B/A ≈ R/d1 ⊕ R/d2 ⊕ · · · ⊕ R/dt ⊕ Rn−t . with t ≤ n.. gn } be a free basis for B. where n ≥ t. Then ∃ free bases {a1 ..            d1 0 · · · 0 0 .. Let {g1 . Among these. The proof now follows by induction on the size of the matrix. the other entries of the ﬁrst row may be made zero.i g1 + a2.. In a similar manner. b2 . g2 .. h2 . and such that each ai = di bi . .i gn .. By the previous theorem. where each di = 0. and thus I = d1 R.j . 1) place in the matrix. a2 . .j ) by a ﬁnite number of row and column operations. 0            cij Note that d1 divides each ci. B is a ﬁnitely generated free Rmodule and A ⊂ B is a non-zero submodule. the matrix may be changed to the following form. A has a free basis {h1 . By elementary operations of type 2. .

gn }. then the elements d1 . Since a|b and a|c. The way Theorem 5 is stated.. there is no prime common to their prime factorizations. Theorem 5 If R is a Euclidean domain and M is a ﬁnitely generated R-module.. s1 s2 Let d = p1 p2 · · · pst be the prime factorization of d. Theorem 6 Suppose R is a PID and d is a non-zero non-unit element of R.) Proof If i = j. Proof There exists and a ∈ R with aR = bR + cR. If R = F [x] and we select the di to be monic.. so R = bR + cR. ht } and {g1 . R/di = 0. d2 .e. some or all of the elements di may be units.  dt       0 0 with di |di+1 .. . Then the natural map t ≈ R/d −→R/ps1 ⊕ · · · ⊕ R/pst is an isomorphism of R-modules. so 0 −→ A −→ B −→ M −→ 0 is a s. g2 . and for such di . and di |di+1 for 1 ≤ i < t. then M ≈ R/d1 ⊕R/d2 ⊕· · · ⊕R/dt ⊕Rm where each di = 0. Lemma Suppose R is a PID and b and c are non-zero non-unit elements of R... then they are unique. but we do not bother with that here. Since changing the isomorphisms Rt −→ A and B −→ Rn corresponds to changing the bases {h1 . If we assume that no di is a unit. psi and pj j are relatively prime.Chapter 6 Appendix      U (ai. The result now follows from the previous theorem...j )V =       119 d1 0 . a is a unit.s.. The splitting in Theorem 5 is not the ultimate because the modules R/di may be split into the sum of other cyclic modules. the theorem follows.. They are unique up to associates.. To prove this we need the following Lemma. they are i s ≈ . Then bR and cR are comaximal ideals. ¯ Proof By hypothesis ∃ a ﬁnitely generated free module B and a surjective homo⊂ morphism B −→ M −→ 0. By the Lemma above. .e. h2 . and B/A ≈ M. . 0 d2 0 ··· ≈ ··· 0  0    . (The elements psi t 1 i are called elementary divisors of R/d.. . they are unique. Suppose b and c are relatively prime. If R = Z and we select the di to be positive. dt ¯ are called invariant factors. . i. Let A be the kernel.

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

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

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

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

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

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

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

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

It is easy to see that if f1 and f2 are R-multilinear. . .   Determinants In the chapter on matrices. . We suppose R is a commutative ring. . . . . . . . bn ) deﬁnes an R-linear map from Bi to C. . Bi . f is skew-symmetric if f (b1 . Also if f is R-multilinear and r ∈ R. Proof From the ﬁrst exercise in Chapter 5. the set of all functions from B1 ⊕ B2 ⊕ · · · ⊕ Bn to C is an R-module (see page 69). . . 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). . . . . and B1 . and bj ∈ Bj for j = i. . (This funny looking exercise is a little delicate. . bn ) = sign(τ )f (bτ (1) . Theorem The set of all R-multilinear maps is an R-module. It must be seen that the R-multilinear maps form a submodule. . Find the characteristic and minimal 6) Let F = R and A =  0  −3 1 −1 polynomials of A. . . Show that if λ is a root. ¯ 5) Suppose F is a ﬁeld containing F as a subﬁeld. b2 . n}. . so is f1 + f2 .128 Appendix Chapter 6 4) Suppose λ ∈ F . bn ) = f (bτ (1) . then (f r) is R-multilinear. . . bτ (n) ) for all permutations τ on {1. . . Show λ is a root of the characteristic polynomial of A iﬀ λ is a root of the minimal polynomial of A.) 5 −1 3  2 0 . B2 . = Bn = B. . 2. Deﬁnition 1) 2) f is symmetric means f (b1 . . . From here on. then f |(b1 . . Show that the minimal polynomial of A ∈ Fn is the same as the minimal polynomial of A considered as a ¯ matrix in Fn . its order in the characteristic polynomial is at least as large as its order in the minimal polynomial. . Bn is a sequence of R-modules. . Deﬁnition A map f : B1 ⊕ B2 ⊕ · · · ⊕ Bn → C is R-multilinear means that if 1 ≤ i ≤ n. suppose B1 = B2 = . C is an R-module. bτ (n) ) for all τ . . . n ≥ 2.

a1. . . and show f is alternating. .. ii) Alternating ⇒ skew-symmetric. e2 . en }.. . To prove iii). b2 . b3 . ... 2).. as an element of B ⊕ B ⊕ · · · ⊕ B. . and so 2f (b1 . Then f = df (e1 . . b3 . bn ) = f (b1 .. en ). .. then f (a1. en ). . . bn ) = 0.. For simplicity. e2 .. iii) If no element of C has order 2.n e1 + a2. ..n e2 + a2. Now we are ready for determinant. bn ) = 0 whenever some bi = bj for i = j. Suppose A = (ai. bn ) + ¯ f (b2 . . Theorem Suppose f : B ⊕ B ⊕ .. Suppose B is Rn with the canonical basis {e1 . b3 . Theorem d is an alternating multilinear form with d(e1 . if A = (ai. . .. This means f is the multilinear form d times the scalar f (e1 ... 2). It turns out that this is all of them. b1 . dc is an alternating multilinear form. .n e2 + · · · + an. bn )..1 e2 + · · · + an.n en ) = |A|f (e1 . If we let τ be the transposition (1. . .2 · · · aτ (n). . Deﬁne d : B ⊕B ⊕· · ·⊕B → R by d(a1.. Suppose C = R. . because the set of alternating forms is an R-module. . . b1 . . . (We think of a matrix A ∈ Rn as n column vectors. b1 . . Then 0 = f (b1 + b2 . bn ) = −f (bτ (1) .n en ) = all τ sign(τ )(aτ (1). assume f is alternating. . . . Proof Part i) is immediate... bτ (n) ) where τ is a transposition.. .. Thus the set of alternating forms is a free R-module of dimension 1... b1 . b3 .1 en . b1 . . . . bn ) = 0. . . b3 . b1 + b2 .) First we recall the deﬁnition of determinant.. ¯ i) If c ∈ R. To prove ii). It suﬃcies to show that f (b1 . en ) = 1. e2 . bn ) and the result follows. In other words.. ..n ) = |A|..1 e1 +a2. ⊕ B → R is an alternating multilinear form. ..1 en . .. .. a1.1 e1 + a2. . and ¯ the result follows. . e2 . The next theorem is from the section on determinants in Chapter 4. en ). ¯ we get f (b1 . . Suppose for convenience that b1 = b2 and show f (b1 .e. b3 . In this case multilinear maps are usually called multilinear forms.Chapter 6 3) Theorem Appendix f is alternating if f (b1 . bn ) = −f (b1 .1 aτ (2).n e2 + · · · + an. then alternating ⇐⇒ skew-symmetric. . . e2 . .j ) ∈ Rn .j ) ∈ Rn . as seen by the following theorem. suppose f is skew symmetric and no element of C has order 2. i. . ¯ 129 Each of these three types deﬁnes a submodule of the set of all R-multilinear maps. assume τ = (1.. and the determinant is a generator.1 e2 +· · ·+ an... b3 .

then |CA| = |C||A|. and so V ∗ is the dual of V and V may be considered to be the dual of V ∗ .2 · · · ain .n e1 + a2. a1. ... . e2 ) = (a1.n en ) = ai1 . f (a1. e2 ) = |A|f (e1 . 1 ≤ in ≤ n. a tangent plane to a diﬀerentiable manifold is a real vector space. . Proof Suppose C ∈ Rn . eτ (n) ) = all τ sign(τ )aτ (1). if any is = it for s = t. e2 .1 e1 + a2.. . The sections of the tangent bundle are called vector ﬁelds while the sections of the cotangent bundle are called 1-forms.2 − a1. e2 ) + a2.n f (e1 .1 a2. However there is a natural isomorphism from V to V ∗∗ . . For the general case.. A2 .2 · · · aτ (n).1 )f (e1 . In the notation of the previous theorem. A = (A1 .2 e2 ) = a1. CA2 . e2 ). If V is a ﬁnitely generated vector space over a ﬁeld F . Thus the tangent (cotangent) bundle may be considered to be the dual of the cotangent (tangent) bundle... .2 f (e1 .. . Since f (e1 .1 a2. f (A) = |A|f (e1 . but also in other areas such as diﬀerential geometry and topology. e2 .1 aτ (2). However. eτ (2) . homology groups are derived from chain complexes. F ). Therefore the sum is just all τ aτ (1). .2 f (e2 . V ∗ is isomorphic to V . its dual V ∗ is deﬁned as V ∗ = HomF (V.1 a2.. 1 ≤ i2 ≤ n.n f (eτ (1) . .2 f (e2 . f (a1. not only in algebra... Use the fact that CA = (CA1 . . If A ∈ Rn .2 a2..1 e1 + a2. en ) = |A|f (e1 .n f (ei1 . while the union of the dual spaces is the cotangent bundle. .1 e2 + · · · + an.1 aτ (2). ei2 . Deﬁne f : Rn → R by f (A) = |CA|.. .2 e1 + a2.1 a1... Dual Spaces The concept of dual module is basic. ein ) where the sum is over all 1 ≤ i1 ≤ n. (See the third theorem on page 63. a1. . en ) = |CI| = |C|.n e2 + · · · + an.. e2 . This incredible classiﬁcation of these alternating forms makes the proof of the following theorem easy.2 · · · aτ (n). A ∈ Rn . The sum of the cohomology groups forms a ring.130 Appendix Chapter 6 Proof For n = 2.1 ai2 ....) Theorem If C...2 f (e1 .1 a1. it follows that |CA| = f (A) = |A||C|.. that term is 0 because f is alternating..1 en .. . B = Rn and Rn = Rn ⊕ Rn ⊕ · · · ⊕ Rn .. e1 ) + a1. e1 ) + a2. and f : Rn ⊕ · · · ⊕ Rn → R has f (A1 .. In algebraic topology. A2 . .1 e2 . but in general there is no natural isomorphism from V to V ∗ . The union of these spaces is the tangent bundle. CAn ) to show that f is an alternating multilinear form. e2 . An ) where Ai ∈ Rn is column i of A. By the previous theorem. while the sum of the homology groups does not. For example. while cohomology groups are derived from the dual chain complexes.. An ) = |CA|.. . . en ). This remarkable fact has many expressions in mathematics.. en ). you can simply write it out.. .

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

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

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

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

23 module. 60. 119. 83 Characteristic of a ring. 19 Boolean algebras. 97 Alternating group. 58 . 66 Chinese remainder theorem. 108 Classical adjoint of a matrix. 46 unique factorization. 39 Cartesian product. 76 Coset. 70 of rings. 50 Characteristic polynomial of a homomorphism. 66. 43 Axiom of choice. 20 in a ring. 29 of modules. 32 Cyclic group. 83 Division algorithm. 1. 97. 51 Cancellation law in a group. 22 Change of basis. 85 of a matrix.Index Abelian group. 79 of a module. 131 Coproduct or sum of modules. 116 integral domain. 71 Algebraically closed ﬁeld. 83 Bijective or one-to-one correspondence. 44 Contravariant functor. 98. 39 of a function. 64 Conjugation by a unit. 46. 95 Elementary divisors. 130 Eigenvalues. 125 Center of group.7 Binary operation. 50. 132 Dual spaces. 42. 120 Commutative ring. 47. 20. 24. 104 Conjugate. 31 Cayley-Hamilton theorem. 45 Domain euclidean. 128 Diagonal matrix. 112 Associate elements in a domain. 72. 107 Determinant of a homomorphism. 74 Cycle. 52 Boolean rings. 11 Cayley’s theorem. 2. 46. 120 Elementary matrices. 62 Comaximal ideals. 78. 109 Automorphism of groups. 40. 108. 37 Complex numbers. 56 Dimension of a free module. 32 Ascending chain condition. 47. 5 principal ideal. 95 Eigenvectors. 10 Basis or free basis canonical or standard for Rn . 111 Dual basis. 63 135 Cofactor of a matrix. 95 of a matrix. 85.

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

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

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