Você está na página 1de 66

TOPOLOGY

A READING GUIDE
for
MATH 513
at
Binghamton University
by
Matthew G. Brin
1-st Edition
Department of Mathematical Sciences
Binghamton University
State University of New York
c 2012 Matthew G. Brin
All rights reserved.
ii
On the text
The text for this course is Topology, second edition, by James R. Munkres,
published 2000 by Prentice Hall, NJ.
These notes are a guide to the reading. They will be added to as the semester
goes on.
Most of the mathematics in this course will be learned by reading the book.
That is a very deliberate choice on the part of the instructor (me). Much of
my work in these notes and in class will be helping you learn how to learn
mathematics from a book.
My intention as I write this preface is to write chapters and sections of these
notes that correspond to the chapters and sections of the book. The notes will
contain what to look for, and perhaps will indentify parts to emphasize, parts
that might be more dicult, possibly parts to ignore, and parts that might
be of particular importance. There may be additional material introduced in
these notes of varying amounts. There may also be alternative points of view
than those given in the book, and there will sometimes be either alternative or
additional notation.
There will be material in these notes that may not at rst appear to be
mathematical, and perhaps never appear to be mathematical.
After writing a few pages, I nd that I refer to the Munkres book either with
the word book or the word text. This guide will never be referred to with either
of those words and their meaning will always be the Munkres book. If I wish
to refer to this guide it will be with the word guide or the word notes.
iii
iv ON THE TEXT
A note to the reader
The books A note to the reader is on Pages xv and xvi in the text. Read
it (the note) carefully. Do not worry if the four examples on Page xvi are
incomprehensible. You are not expected to be familiar with either the words or
notation used in any of the examples. The four examples will have been dened
and discussed to some extent by Page 118 in the text.
v
vi A NOTE TO THE READER
Part I
General Topology
1
Chapter 1
Set theory and logic
On the material before Section 1. The important sentence starts four lines
from the bottom of Page 3, starting with In this book, . . . Since this is
not a course in set theory, you will have to learn how to manipulate sets by
watching rather than having all the rules laid out for you at the beginning.
Since the manipulations will be few, this is not too much of a burden, but it is
a burden nonetheless. You will need to be attentative.
1.1 Fundamental concepts
As you read this section, you will need to focus on two concepts that are not
strictly mathematical content. They are language and interactions. We
start with language. Interactions will be discussed later in the section.
1.1.1 Language
Part of learning mathematics is learning how to communicate it. Saying you
know how to solve a problem but dont know how to explain the solution means
that you have not learned enough. Missing from your learning would be math-
ematical language adequate to explain the solution.
The language consists of words and symbols. Symbols are usually reserved
for very elementary and frequently used concepts, so I will treat them dierently.
Words, or terms, are usually dened. Denitions have three parts.
1. A general idea. Example: A derivative tells you how fast the value of
a function changes in comparison to the change of the argument of the
function.
2. The technical denition. Example:
f

(x
0
) = lim
xx0
f(x) f(x
0
)
x x
0
.
3
4 CHAPTER 1. SET THEORY AND LOGIC
3. The explanation of how the technical denition expresses the general idea.
Example: From the denition of limit (not given), for any given > 0 there
is a > 0 so that if |x x
0
| is less than , then

f(x) f(x
0
)
x x
0
f

(x
0
)

<
which leads to
f

(x
0
) <
f(x) f(x
0
)
x x
0
< f

(x
0
) + ,
and we see that the ratio given by change of f over change of x can be
kept as close to f

(x
0
) as we like by keeping x suciently close to x
0
.
When possible all three parts of a denition should be learned thoroughly.
When you get to the denition of compact spaces in Chapter 3, the general
idea will elude you. It eludes everyone when they rst learn it and continues to
elude for some time after. The notion takes much getting used to. So when the
notion is introduced, the only part available is the technical denition. Note
that this gives you permission to proceed with your learning when you cannot
nd a general idea behind a concept. Stated more strongly, you do not have
permission to halt your progress just because you come across a term whose
general idea eludes you.
In this course, elusiveness of general ideas will be rare. The rule about de-
nitions will be that the technical denitions must always be learned completely,
immediately and precisely. Further, unless absolutely impossible, the other two
parts should be learned as quickly and as thoroughly as possible. This will be
almost all the time. The bottom line is that at all times you must know the
technical denitions of the words that are being used in the course.
1.1.2 On reading Section 1
Omit nothing from this section.
In reading Section 1, identify all terms that are dened. The use of boldface
type is a broad hint, but there may be terms that the author forgot to identify
this way.
The notion of equality of sets needs comment. If a and b are integers, then
a = b means they are the same integer and no more needs to be said. But if
A and B are sets, then the book claims that A = B means that they are the
same set. But the book does not claim that no more needs to be said, because
it goes on to say more. Does this give equality of sets a denition?
The bottom of Page 4 and top of Page 5 give two ways of specifying sets. It
leaves out one. For example the set of natural numbers (according to me) is
{0, 1, 2, 3, 4, . . .}.
The dot, dot, dot notation is common and you should get used to it. Perhaps
the book regards this as part of the notation discussed at the bottom of Page 4.
1.1. FUNDAMENTAL CONCEPTS 5
The book mentions at the top of Page 5 that the notation there denes B
as a subset of a given set A. This is often built into the notation as
B = {x A | x satises some property}
which would be read out loud as B is the set of all x in A that satisfy the given
property.
In Page 5, a formal denition of A B is given as
A B = {x | x A or x B}.
Why is the word or in that denition and not the word and? After all,
A B contains all the elements of A and all the elements of B. Answering this
question requires full understanding of the technique for specifying sets given
at the top of Page 5.
In the same vein, why is the word and in the denition
A B = {x | x A and x B}?
The statement about 8 lines from the bottom of Page 6 saying for each
object x, the relation x does not hold is often taken as a denition of the
empty set. A formal approach to set theory would have as an axiom that such
a set exists.
The long discussion on Page 7 leading to vacuously true is important. It
includes a careful discussion of when an If . . . , then . . . statement is to be
considered true.
The discussion on Page 8 leaves out the inverse. If P Q is a statement,
then the inverse of the statement is (not P) (not Q). What is the relation
between the inverse and the converse?
On Page 9, the relation P Q is often read as P is equivalent to Q, or
more loquatiously as P is logically equivalent to Q. The phrase if and only if
is often shortened to i in print and in writing. There is no way to read i
out loud except to say if and only if.
The logical quantiers in the second half of Page 9 are what make mathe-
matics hard and what make it interesting. They are important enough to have
their own symbols. I will use them often on the board for two reasons. One is
that they save writing. The other is that they help organize. That will become
apparent when we return later to discuss the rule on the bottom of Page 9 more
thoroughly.
The symbol for for all is , and the symbol for at least one (which is
more usually said as there exists) is .
In the bottom of Page 9 and the top of Page 10, the book discusses statements
P and Q that may or may or not hold depending on the value of some x. This is
usually revealed symbolically by writing P(x) and Q(x) to show the dependence
on x. An example of such a P(x) would be x > 7 which is either true or false
depending on the value of x. Thus for every x A, statement P(x) holds
becomes
x A, P(x)
6 CHAPTER 1. SET THEORY AND LOGIC
and for at least one x A, statement Q(x) holds becomes
x A, Q(x).
1.1.3 Relations and operations
All of , = and are relations. Whatever they are given to work with, the
result takes on one of two values: true or false. On the other hand, and
are operations. Whatever they are given, the result is a mathematical object.
With these particular operations the result is a set. Operations such as + and
generally give a result that is a number, although a is dened on sets in
Page 10 whose result is a set. You should look back through the material of
Section 1 up to the start of Rules of Set Theory on Page 10 and identify all
the relations and operations that you can nd.
This leads us to the topic of interactions.
1.1.4 Interactions
Mathematics takes place where concepts meet.
We have barely enough symbols to see how they interact. To shorten state-
ments even more, we invent for not as in (not Q) is written as (Q). The
rule at the bottom of Page 9 becaomes
(x A, P(x)) x A, (P(x)).
This is an example of an interaction. It shows how interacts with . Whenever
two operators (each can be either a relation or an operation) are introduced
in the same subject, it is often useful to see how they interact. The material on
Pages 10 and 11 on the Rules of Set Theory discuss the interactions of ,
and as operations on sets.
As the operations and relations accumulate, you should take as a worthwhile
exercise to see how they interact. This is an open ended assignment. You will
never nish. But it is good to keep pecking away at it. Many of the exercises
at the end of the section on Pages 14 and 15 explore such interactions.
More on language
The formula
(x A, P(x)) x A, (P(x)) (1.1)
illustrates the power of the symbols to shorten statements. What formula similar
to (1.1) expresses the content of the top few lines on Page 10?
If P(x, y) is a statement that depends on two values (consider x > y as
an example), then we can look at
x, (y, P(x, y)) (1.2)
and
y, (x, P(x, y)). (1.3)
1.1. FUNDAMENTAL CONCEPTS 7
The rst says that for each x you can nd a y that cooperates with the
particular x to make P(x, y) true. Thus one can choose a dierent y for each
x. The second says that there is one y that works for every single x to make
P(x, y) true. In the second statement, the burden on the one y is much greater
than on any single y used in the rst statement.
This distinction is crucial. Which statement is true if P(x, y) is take to be
x > y?
As an exercise, negate each of (1.2) and (1.3). This exercise is supposed to
convince you of the value of the use of symbols and formulas such as (1.1) and
its counterpars. If you are not convinced, try to nd an ecient algorithm to
negate combinations of and using the formulas.
How would you negate
x, (y, (z, R(x, y, z)))?
1.1.5 Back to Section 1
The topic Collections of Sets starting near the bottom of Page 11 contains
the notion of a set of sets. This concept and the union and intersection of a set
of sets in Arbitrary Unions and Intersections go together. In Collections of
Sets is the denition of the power set of a set. This will enter many discussions
shortly.
Cartesian products come next and are based on ordered pairs. Grammat-
ically, an ordered pair is a construct (a, b) using the ve symbols ( a ,
b ) with behavior dened by saying that two such pairs, (a, b) and (c, d) are
equal if and only if a = c and b = d. The small type inset on Page 13 shows
that ordered pairs can be dened using only the notion of set. To make good
on this claim one has to prove that the statement
{{a}, {a, b}} = {{c}, {c, d}}
is true if and only if a = c and b = d. This looks obvious until it is pointed out
that a = b is not an assumption. The proof of the if and only if can be taken
as an exercise, but you should be warned that it is annoying and when nished,
only mildly satisfying.
The discussion near the bottom of Page 13 leaves out one more use of (a, b).
If a and b are integers, then (a, b) is often used to denote the greatest common
divisor (GCD) of a and b. It is very common in mathematics to have one
notation used for several concepts.
1.1.6 The exercises on Page 1415
You should do them all, but you might nto regard that as practical. There is
also the problem of what is meant by check in Problem 1, or determine in
Problems 2, 5 and 10, and prove in problem 9. Problems 3, 4, 6, 7 and 8 are
more straightforward and should be done rst. The word show in Problem 8
should give no diculty.
8 CHAPTER 1. SET THEORY AND LOGIC
Proving statements will be dealt with piecemeal over the next few sections.
For Problems 1, 2, 5, 9 and 10, do the best you can. Do as many parts of
Problem 2 as you can stomach. The more the better.
More work
See how many interactions you can analyze. You have learned that distributes
over and vice versa. Does distribute over and? Does distribute over
or? How about ? Not all answers are yes.
Note: There are symbols for and (it is ) and or (it is ). I rarely use
them (the words are already short enough) and the book does not use them at
all. They are used heavily in books on logic.
1.2 Functions
The book comes close to messing up the denition of a function, but saves itself
in the end. It is denitely guilty of unnecessary generalization.
By Page 16, you know that a function f : A B must have a value f(x) for
every x A. This is so important a property for a function that it should not
have been delayed full page after the denition started. The rule of assignment
dened on Page 15 will have no use in the rest of the course.
The other part of the denition of a function appears as the top line of
Page 16. The two parts of the denition of a function are often combined into
statements of existence and uniqueness. If f : A B is a function, then for
every x A, the exists a value for f(x) and this value is unique.
As an exercise see if you can dig out the existence part from the denition
of a function in the book.
The words range and image are treated dierently in dierent books. You
will have to learn how this book uses them and get used to it. The books
interpretation of these two words agrees with many other books. I am not bold
enough to say most.
The technical denitions of injective and surjective could use quantiers.
What should the be and where should they go?
The only other point to make about this section is that it could be clearer
on the double burden that the notation f
1
carries. When f is a bijection, then
f
1
is a function. However, if f is not a bijection, then f
1
still has meaning.
Technically if f : A B is a function, then f
1
is a function from P(B), the
power set of B, to P(A), the power set of A. Less technically, if Y is a subset
of B, then f
1
(Y ) is a subset of A as dened in the middle of Page 19.
The rest of the section should be read carefully, omitting nothing. An im-
portant concept is introduced in Problem 5 at the end of the section and is
discussed below.
1.2. FUNCTIONS 9
1.2.1 The exercises on Pages 2021
There are now more things to interact. There are f and f
1
available to in-
teract with each other and with the various constructs in Chapter 1. Not all
interactions are as nice as some others and the exercises work through them. It
is extremely important to know which work out nicely and which do not.
The word show starts to be more important here, making this a good time
to start discussions of proofs.
Some not very hard and fast rules are as follows.
1. Be guided by the conclusion. It is often a mistake to start with the hy-
pothesis and try to nd your way to the conclusion.
2. To prove P Q, assume P and then prove Q.
3. To prove that P and Q is true, rst prove P and then prove Q.
4. To prove that P or Q is true, assume one (say P, for example) is false
and use that assumption to prove the other (Q in the example).
5. To prove that P and Q imply R just assume both P and Q and then
prove R.
6. To prove that (P or Q) implies R you have to do twice the work. You
have to prove P implies R and you have to prove Q implies R. This
is known as proof by cases.
7. To prove that x, P(x) is true, let x be arbitrary and prove P(x) for that
arbitrary x. It is against the rules to make x less than arbitrary. Thus it
is not valid to say Let x be arbitrary. Now assume x = 7. Now prove
P(x). It is valid to say Let x be arbitrary. Now x > 7 or x 7 is true.
We will prove that x > 7 implies P(x) and we will prove x 7 implies
P(x). Explain how this example combines two of the points made above.
8. To use x, P(x) in a hypothesis, just assume P(x) is true every time an x
comes up in the proof.
9. To prove x, P(x) nd an x that works.
10. To use x, P(x) in a hypothesis, let x be arbitrary in every way except
that you assume that P(x) is true about that x. It is against the rules
to make x any less arbitrary. Thus it is not valid to say Let x be such
that P(x) is true. Now assume x = 7. Now prove the desired conclusion.
This comes up in using an assumed twice. Example: Assume f and g
are both continuous and an > 0 and an x are given. The denition of
continuity guarantees a value of > 0 that has some nice properties that
we need not state. It is wrong to say Let > 0 be such that something
nice happens around f(x), and let > 0 be such that something nice
happens around g(x). The second use of the assumed is wrong because
the same letter is used for g that was used for f. That makes the
10 CHAPTER 1. SET THEORY AND LOGIC
picked for g not arbitrary. It has been assumed to be the same that
was used for f. It is correct to say Let
1
> 0 be such that something
nice happens around f(x), and let
2
> 0 be such that something nice
happens around g(x).
If these rules are not familiar to you, then you will have to refer back to
them from time to time.
One diculty in applying these rules is deciding at each state of a proof what
the general shape is of what you are trying to prove. This will be discussed in
class for a while.
Some of the above rules can be validated by an exercise with truth tables.
For 4, show that (P or Q) (P Q). Find a similar statement that
validates 6.
I assume no one will have trouble with applying (A B and B C)
implies (A C) or the simpler (A and (A B)) implies B. The rst is
referred to as the transitive property of , and the second is called modus
ponens.
Now that you have been given way too much information about proofs, try
all the problems on Pages 20 and 21.
Several concepts are introduced in Problem 5 on Page 21. This makes the
problem an important one. The two parts of the denition of a function and the
two concepts one-toone and onto have nice symmetries. This problem is
an exploration of those symmetries. What part of the denition of a function is
like the denition of one-to-one and how does it dier? Answer the question
with one-to-one replaced by onto.
1.3 Relations
Every function is a relation but not every relation is a function. Thus the
denitions of a function and a relation can use much of the same words and
symbols. The book does this and a function in the book has been dened as
the set of all ordered pairs (a, f(a)) in A B as a runs over A. In standard
notation, this is the set
{(a, b) AB | b = f(a)}.
As just mentioned, this is just the graph of f.
In words, this is just the set of all (a, b) in AB that are related by having
b be the image of a.
The same approach is taken with relations. The denition makes a relation
the set of all related pairs. The main dierence between function and relation
is that a function has restrictions and a relation has none. Any set of ordered
pairs makes a relation. This is not at all true of functions.
The book only denes relations on one set, so the pairs live in A A. One
can have a relation between two dierent sets. The relation was born in is a
relation from the set of people to the set of countries. One can have a relation
1.3. RELATIONS 11
between elements of more sets. The relation was born in during is a relation
between the set of people, the set of countries and the set of years. You can all
think of diculties with these examples (people born at sea?) which is why we
tend to stick to mathematics most of the time.
Example 1 in the book is another collection of non-mathematical examples.
Hopefully equivalence relations are well known to you. If not, commit all
denitions to deep memory. Lemma 3.1 should be gone over until its key points
have become obvious.
The connection between equivalence relations and partitions is complete.
The bottom half of Page 23 is really dealing with a one-to-one correspondence
and its inverse.
In the paragraph above the denition there is a process that takes an equiv-
alence relation on A and derives from it a partition of A. Thus if E is the set of
equivalence relations on A and P is the set of partitions on A, we have created a
function f : E P. In the paragraphs after the denition a function is dened
(that we may as well call g) that goes in the other direction.
The paragraph at the bottom of the page shows something about this con-
struction. What is it showing? Is it about g? Or is it about f? And what
exactly is being shown? Once this is done, you should realize that all this could
have been made a lot simpler by showing that f and g are inverses of each other
and using Problem 5 of the previous section. In fact, you should take as an
exercise that f and g are inverses of each other.
The next point is that functions give partitions and thus equivalence rela-
tions. If f : A B is a function, then
{f
1
(b) | b B}
is a collection of subsets of A which (if you ignore the empty sets that show
up if f is not onto) is a partition of A. If f is onto, then the empty sets dont
show up. The corresponding equivalence relation on A is dened by x y i
f(x) = f(y).
The book does not bring in the map from a set with an equivalence relation
to the set of equivalence classes. This relates to the partition that comes from a
function in the previous paragraph, but this can wait until it arises. It is related
to the observations in Problem 4 in this section.
The subsection on order relations is loaded with denitions. Unfortunately
for you, they are all in common use and need to be memorized. The dictionary
order relation will need some thought to get used to and you should devote some
energy to that. It comes with several alternate names. My favorite is lexical
order since it uses few letters. Lazier people call it lex. People that like lots
of syllables call it lexicographical order.
To show the care needed in learning the denitions, note that not all sets
that are bounded above have a largest element. Find an example.
The main point of the least upper bound property is that the real numbers
have this property. That will just be assumed in this course. In another course,
it can be proven if a given construct of the reals is assumed.
12 CHAPTER 1. SET THEORY AND LOGIC
The problems are starting to get a little harder. Do as many as you can.
Problems in this section break into two categories. One category is of general
sounding problems. These are 2, 4, 7, 9, 11, 13, 14. The other category (the
ones not mentioned) are less general and are about specic examples. Both
types are valuable and you should not stick only to one type if you dont end up
doing all the problems. Note that some problems (13 is a good example) cannot
just be done by relying on the formalities to get you through. Some thought
needs to be involved. This requires complete familiarity with the denitions,
and often an understanding of the correspondence between the denition and
the idea behind the concept.
1.3.1 Uniqueness
If something s is claimed to be unique with respect to a property, that means
that any candidate c with that property must have c = s. This looks like pieces
of the denition of one-to-one, and it should. The denition of one-to-one says
that an element that is mapped to a particular value is unique.
One shows uniqueness by assuming two items with a given property and
proving that they must be the same.
Show that a largest element of a set is unique.
This section does not cover partial orders. They show up in the last section
of the chapter in an exercise.
1.4 The integers and the real numbers
This section is a rather mixed assemblage of facts.
1.4.1 The reals
The assumptions on the reals (1)(8) from the bottom of Page 30 to the top
third of Page 31 seem to leave some things out. I dont know how to prove that
every x R has a x R so that x + (x) = 0. I dont know how to prove
that every x = 0 in R has an x
1
so that x x
1
= 1.
After the positive integers are dened it is possible to give one more property
of the reals that says:
(9) For every real number x there is a positive integer n so that x < n.
This turns out to be provable from (1)(8) and this is done on Page 33, so
it doesnt really need its own number.
But, I think you can prove (8) from (1)(7). You might need to assume
1 > 0 which is part of Exercise 2. Try to prove (8). I can give a hint if you need
one.
The problem with the listed properties is that no one really refers back to
them every time they use the real numbers. However, I claim that every fact
about the real numbers follows from properties (1)(8), even though it would
1.4. THE INTEGERS AND THE REAL NUMBERS 13
be a pain to prove all these facts from these basic properties. The exercises give
enough samples of things that follow from these assumptions.
To see the importance of (9), try to prove that there is no positive real
number that is a lower bound for the set
_
1
n
| n Z
+
_
rst by assuming (9) and then by not assuming (9). Of course, (9) is provable
from (1)(8), but by not assuming (9) I mean by not proving (9) rst.
1.4.2 Induction
The other major topic in the section is induction. You should be familiar with
induction, but are probably not familiar with it in the form given in the book.
Item (2) in the middle of Page 32 is the key. If you have a statement P(n)
about elements of n Z
+
and you wish to prove that
n Z
+
, P(n)
then the usual technique is to prove P(1) and then prove that
n Z
+
,
_
P(n) P(n + 1)
_
.
The step above is called the inductive step.
The connection of this usual form of induction with (2) on Page 32 of the
book is to let T be dened by
T = {n Z
+
| P(n)}.
In words T is the set of all n Z
+
that make P(n) true.
The usual induction proof is essentially a proof that T is inductive. That
is, 1 T and for every x T, we have x = 1 T. Now by (2) on Page 32 of
the book, we get that T is all of Z
+
and we have that P(n) is true for every
n Z
+
.
Note that this connection is used in the proof of Theorem 4.1. Read the
proof of 4.1 carefully to nd the use of this connection.
If we call the procedure of proof by induction that you are familiar with
ordinary induction, then Theorem 4.2 (after translation) gives strong induc-
tion. The translation from 4.2 to a method of induction is the same as the
translation from (2) on Page 32 to ordinary induction.
To make the translation smoother, let us describe ordinary induction as
true for 1, together with true for n implies true for n + 1 implies true for all
n Z
+
.
The translation of 4.2 gives true for everything less than n implies true for
n implies true for all n Z
+
. Of course this is too brief. Expanding it says
the following.
14 CHAPTER 1. SET THEORY AND LOGIC
If for each n Z
+
, you can prove P(n) by assuming P(j) for each j Z
+
with j < n, then you have proven P(n) for every n Z
+
. The justication
from 4.2 is the same as the justication of ordinary induction from (2) on Page
32. Let T be the set of all n Z
+
that make P(n) true and show that true
for everything less than n implies true for n means that T satises all the
hypotheses stated about A in Theorem 4.2, and so T = Z
+
.
The main point is that strong induction is justied by Theorem 4.1 which is
proven using ordinary induction. So strong induction is as justied a technique
of proof as ordinary induction.
So why is strong induction called strong. It is because it is easier to use. To
prove P(n), you are allowed to assume all of P(j) for j Z
+
with j < n. In
ordinary induction you are only allowed to assume P(n 1).
To appreciate the dierence try to prove the fundamental theorem of arith-
metic with strong induction and without it. The fundamental theorem of arith-
metic says that every n Z
+
is either 1, a prime or a product of a nite number
of primes.
One last comment on strong induction is how to prove P(n) under the as-
sumption that P(j) is true for all j Z
+
with j < n when n = 1. You will note
that there are no j Z
+
with j < 1. So you are supposed to prove P(1) under
no assumptions at all. That is, you are just supposed to prove P(1), period.
This is the same as in ordinary induction except that in ordinary induction it
has to be stated separately. In strong induction, the fact that you have to prove
P(1) directly is there, but it is hidden in the main part of the statement of
strong induction.
1.4.3 The exercises
The theme of the exercises is that you really can prove everything you know
about the reals from (1)(8). Once that point is made, it is not all that necessary
to go on.
In Problem 1, you should go on until bored. Much of what is in Problem 1
belongs in an algebra course.
Problem 2 deals with order. It might be good to go at least as far as (g).
Problem 3 justies everything on Page 32 and is important. The theme of
(a) will be repeated in many settings in many courses and is crucial. The theme
will be identied later. A dierent part of the theme continues into (b) and is
just as important.
Problem 4(a) is a bore, but 4(b) is important.
With a few exceptions 5 through 11 continue the theme of everything follows
from (1)(8) and get less important as you plow through them. They dont
teach you anything new about the reals, and the techniques used have little to
do with a topology course. They are more appropriate for an analysis course
for some and for an algebra course for others.
Note that 8(a) is a repeat of 13 from the previous section.
8(b) was already mentioned in the discussion in these notes about the rst
part of this section. 8(c) is more of an analysis problem.
1.5. CARTESIAN PRODUCTS 15
Problem 9 might be of more interest, but once again, there are no surprising
statements. The surprise might be that the parts might require a lot of thought.
Lastly, unless I missed something, I do not nd an exercise that says: for
each n Z
+
, there is no j Z
+
with n < j < n + 1. (Although, Exercise 5(c)
comes close.) Try proving the statement that I gave for just n = 1. If you get
n = 1, then the other values of n follow quickly.
1.4.4 A pedagogical point
The exercises build on each other. If you try to prove every exercise by always
going back to (1)(8), you will do way too much work. One hard part of learning
a mathematical subject is that the things that are proven true one day may be
crucial a day later. This means that as time goes on, you have a larger and
larger set of facts that you need to remember so that you can function as you
go deeper into the subject. This is often the aspect of mathematics that gives
students the greatest problem. To avoid this problem, work to remember the
facts as they accumulate.
1.5 Cartesian products
This is a major section.
First feature: a second way to refer to lots of sets. In Section 1.1 on Pages 11
and 12, the concept of a set of sets was introduced for the purpose of discussing
arbitrary unions and arbitrary intersections. Now we have the concept of an
indexed family of sets. This is just a function whose domain is a set and whose
range is a set of sets. The notation is a bit more like the notation for a sequence.
This requires a digression which is not really a digression since it is really a
special case.
A sequence is a function whose domain is the positive integers Z
+
(if the
sequence is an innite sequence) or a set like {1, l2, . . . , n} if the sequence is a
nite sequence. See the discussion on Page 37. If s is the sequence, we write s
i
instead of s(i).
In an indexed family of sets, the domain is the indexed set J and if A is the
family, then we write A

for a given J instead of A().


Indexed families of sets are not needed in a discussion of unions or intersec-
tions since intersecting the sets X and X and Y is the same as intersecting the
sets X and Y . That is, it doesnt matter that X was mentioned twice the rst
time and not the second. The result is the same.
However, X X Y is not the same as X Y . It makes a dierence that
X is mentioned twice the rst time and not the second.
The idea that a product of sets is built out of tuples is not hard to get
used to. The idea that tuples can be be of arbitrary large size and are really
just functions on an indexing set takes work to get used to. You should allocate
enough time to to this.
16 CHAPTER 1. SET THEORY AND LOGIC
This book starts gently by rst making the jump from nite tuples to
tuples indexed over Z
+
. Later, tuples indexed over arbitrary indexing sets
will be introduced.
In the exercises 1 and 2 are boring but worth doing. In 3, (a) and (b) are
good; (c) is not really a good introduction to a large topic, but you should look
anyway; (d) needs careful thought. Problem 4 needs the notation defeined in
the middle of Page 38 and should be atempted. Problem 5 is reasonable. The
notation is that of Example 3 on Page 38.
1.6 Finite sets
There is little in this section worth looking at. Even the denition on Page 39
is of minor importance. The purpose of the chapter is to demonstrate that a
lot of intuitive statements can acutally be proven rigorously. Those that nd
this interesting should read the section. The rest shoulc read the statements of
Corollaries 6.3 through 6.8 carefully, but not the proofs. Do not do the exercises.
1.7 Countable and uncountable sets
The important thing to develop is a way to see reasonably quickly which innite
sets are countable and which innite sets are uncountable. This is not to imply
that such questions are always easy, but often they are.
First point: appreciate the denitions on Pages 44 and 45.
Next appreciate the techniques in Examples 1 and 2 on the same pages.
Example 1 is quite easy, but important. Example 2 is a bit more invovled, but
its arguments are incredibly standard.
The book leaves out the important notion of having the same cardinality.
Two sets have the same cardinality if there is a bijection between them. Of
course that means that this is not exactly a new notion. It is simply giving a
new word to the relation has a bijection between. However, the important
point is that the relation is an equivalence relation (check that it is) and in
particular is transitive. Thus to show that two sets have the same cardinality it
suces to show that they have the same cardinality of some third set. This turns
one problem into two problems, but sometimes each of the two new problems is
much easier than the original.
Theorems 7.1 and 7.2 are important in statement, and the proofs are almost
as important. The proofs are not as important as material that comes later
that can be used to replace some of the proofs given. The easy argument for
(1) (2) does not get replaced later and should be read carefully as it is.
The material on Pages 47 and the top of Page 48 is rather overblown and
should be read only once and not that critically.
The statements 7.3 through 7.7 are critical. They need to be committed to
memory and used when needed. THe proofs are nice too and you should learn
them.
1.8. THE PRINCIPLE OF RECURSIVE DEFINITION 17
The key to 7.7 is the parenthetical remark at the top of Page 50. You have
to decide how that justies the sentence that follows the remark.
The technique in the proof of 7.7 is almost always referred to as Cantors
diagonal argument. This is the same technique used in the proof of the equally
important 7.8. The important part of the proof of 7.8 is that g in the second
paragraph cannot be surjective. The formula for B in the middle of the proof
would have been more clearly stated as
B = {a A | a / g(a)}.
Every book likes to give the argument that B is not in the image of g that
is given in the last paragraph of the proof. I dont quite like it. I prefer a more
direct argument that shows
a A, B = g(a).
You should write out what this argument would consist of. It does not need
proof by contradiction.
You should also decide why the proof of 7.8 is also called the Cantor diagonal
argument.
In spite of the fact that the last paragraph in the section downplays the use
of the Cantor diagonal argument to prove that R is uncountable, it is good
practice to build such a proof. Prove that the set of all real numbers in (0, 1)
whose decimal expansion uses only the digits 5 and 6 is uncountable using a
Cantor diagonal argument.
1.7.1 Exercises
Do Problem 1 in steps. If you try to do it in one step, you will be missing some
of the points of the section.
Problem 2 is not that rewarding, but it pays to check that you can do it.
Problem 3 is worth thinking about.
The work in 4 is in (a). Again, prove it in steps, and be sure to use results
from the section.
Problem 5 is good practice.
Problem 6(b) is famous (it is named) and the statement should be memorized
even if you dont do the problem.
The rest of the problems are optional.
1.8 The Principle of Recursive Denition
This should be skipped.
18 CHAPTER 1. SET THEORY AND LOGIC
1.9 Innite sets and the axiom of choice
1.10 Well-ordered sets
1.11 The maximum principle
These sections put too much work into the material. I will talk about what is
needed in class.
Chapter 2
Topological spaces and
continuous functions
The title of this chapter is the subject of topology. Unfotunately, the book
defers the denition of continuous functions until the seventh section of this
chapter. This causes problems in motivation and understanding. Rather than
try to change the order of the book, I will just point out that the practicality
and motivation of several of the denitions will not become apparent until the
section on continuous functions.
The outline of the rst six sections is partly laid out in the second sentence
of the chapter. To add more detail, topological spaces are rst dened, then a
section tells how they are often specied, then a section gives one construction of
topological spaces, then two sections tell how new topological spaces can be built
from existing topological spaces, and then a section discusses some fundamental
concepts associated to topological spaces.
The other half of the subject of topology comes after the six sections just
mentioned. Continuous functions will be discussed together with whatever else
is introduced from that point on in the book.
2.12 Topological spaces
Since topological spaces form half of the subject and since the other half of
the subject, continuous functions, have topological spaces as their domains and
ranges, it is clear that the denition of a topological space must be learned
thoroughly and learned immediately. Unfortunately, the denition is hard to
motivate. We will not do so here, but will give some comments.
1. There are no interesting topologies on nite sets. Thus putting energy
into building funny topologies on a set with, say, seven elements is a waste of
time. If the rst sentence of this comment is too strong and sweeping for some,
I will modify it and say that there are no interesting topologies on nite sets in
this course.
19
20 CHAPTER 2. TOPOLOGICAL SPACES
2. There are two useful topologies on nite sets. If you nd them interesting,
then you are allowed to disagree with the point I make above. The two useful
topologies are introduced in Example 2 on Page 77. One of them is extremely
important. Even though it is important, I still say that it is not interesting. The
two topologies will be referred to often. Example 3 is also somewhat useful, but
mostly as an example.
3. Any denition by properties, such as the denition of a topology, must
be immediately followed by examples. Otherwise the denition remains too
abstract. This means that you should learn all the examples as thoroughly as
possible. However, given my rst point above, you are allowed to forget the
material in Example 1 as soon as you read it.
4. The denition of a topology allows for many strange cooked up exam-
ples. The power of the denition is that it covers such a wide sweep of natural
examples that are useful to mathematical investigation. The fact that the de-
nition is exible enough to build funny structures that no one would look at is
not part of its power, nor any evidence of any weakness.
5. The denitions of coarser and ner are given a bit too early. They carry
no intuition until continuous functions are dened. In spite of the fact that the
denitions carry no intuition, they can be useful. Since they talk about one set
(the topology) being contained in another, the concepts can be used to show
that two topologies are equal. This is done in the next section.
6. If you have been introduced to open subsets of the real line in a previous
course but not in a course on topology, then the word open hear can lead to
confusion. In a topology, the open sets are those that are declared to be open
when the topology is specied. In the section on bases, there will be a topology
on the real line where the open sets are not the open sets that you are used to.
This example is important for several reasons, and its strange open sets is one
of the reasons.
7. The book does not adopt the usual convention that a topological space
always refers to a pair (X, T ) where X is a set and T is the topology on X.
There is good reason for not insisting on the pair point of view. Very few people
adhere to the pair view in the way that they write and the way that they talk.
However, the pair point of view lies behind the way most people refer to a
topological space. The book gives one sentence to this view immediately after
the denition of a topological space. However, it is much more important than
this. The wording of Problem 5 on Page 92 suers tremendously because the
pair point of view is so thoroughly suppressed in the book.
You will note that there are no exercises in this section.
2.13 Basis for a topology
Topologies should be specied by saying what is in the topology. This is often
hard. So what is often done is to specify enough of what is in the topology so
that the rest of the topology will follow naturally from the part that is specied.
This is similar to giving a basis for a vector space or a generating set for a group
2.13. BASIS FOR A TOPOLOGY 21
or subgroup. One technique is referred to as giving a basis for a topology. The
use of the same word as is used in linear algebra is deliberate but there are
no parallels in the details. There is no notion of linear combination, linear
independence, and so on, and you should not look for them. The use of bases is
not the only technique to specify a topology and a more fundamental technique
is labeled by the word subbasis. In spite of the fact that it is more fundamental,
it is used less often. However, it is used, and must be learned.
There are two key set theory points in the section.
One is that if A and B are collections of sets, then
_
_
AA
A
_

_
_
BB
B
_
=
_
(A,B)AB
(A B).
This goes back to the older way of denoting a collection of sets. You
should prove the statement above. It is quite easy.
The book abandons the non-indexed collection notation in spite of the
fact that it uses it before the indexed collection notation. Thus you should
also state (and prove) the parallel statement in which A

for I and B

for
J are given and we wish to consider
_
_
I
A

_
_
_
J
B

_
_
.
You might not notice that this is essentially proven on Page 79.
The reason for introducing this is that the section does not emphasize unions
of sets. It gets around this by the use of (2) in the denition of a basis. The
connection between (2) and unions of sets is the following.
Lemma 2.13.1 Let A be a collection of sets. Then S equals a union of sets
from A if and only if for every x S there is an element A of A so that
x A S.
You should do several things with this lemma. You should prove it. However,
rst you must understand it. The statement that S equals a union of sets from
A does not mean that it is the union of all the sets in A. To give notation to
the notion, it means that there is a subset A

of A and
S =
_
AA

A.
You should then prove that the denition of basis in which (2) is replaced
by (2a): Every intersection of two basis elements is a union of basis elements
is equvalent to the denition of basis given in the book.
You should then use this equivalent denition and the facts given above
about intersections of unions to give an argument dierent than the one given
on Page 79 that the collection T dened there is a topology.
22 CHAPTER 2. TOPOLOGICAL SPACES
You should then use the lemma to shorten part of the proof of Lemma 13.2.
Lemma 13.3 together with Example 4 illustrates how the notions of coarser
and ner can be used.
The topologies introduced on the bottom of Page 81 and the top of Page 82
are important. One, the standard topology, is important because it is standard.
The others are important because they are not standard.
The notion of subbasis is given near the bottom of Page 82. The notion
should be learned. Do not worry if the motivating question given before the
denition never occurred to you. I doubt it occurred to very many people just
after they learned the denition of basis.
You should have the impression at this point that the subject is dominated
by manipulation of sets. Any weakness that you have with the manipulation of
sets will hold you back from this point on.
2.13.1 Exercises
Exercise 1 is a test to see if you have absorbed the lemma that is given above.
If you have, then the exercise can be done in one line.
I see no use for Exercise 2.
Exercise 3 is worth doing.
Exercise 4(a) and (b) are useful. The smallest topology part of (b) is related
to the notion of basis and subbasis. See the next problem. If you nd (c) useful,
then you should do it.
The relevance of the part of 4(b) that uses with the word smallest is shown
in Exercise 5. It explains the use of the phrase generated by right after the
denition of basis.
You should do as much of the various parts of Exercises 6, 7, and 8 as needed
to check your understanding. Parts involve tools about the real line from the
previous chapter.
2.14 The order topology
This section and the two that follow it give ways of building topologies. That
there are no exercises given until after the three sections indicates the common
theme that runs through the three sections. These three sections do not give
every technique for building topologies. There are more given later in the chap-
ter, more given later in the book, and more that are not even covered in the
book.
The order topology is an important technique, but not a dominant one. It
shows up with some regularity, but not with great frequency. However, it is a
great source of examples. Questions like Why dont properties P, Q and R
imply property X? can often be answered with a topology built from some
order.
The material should be learned. Pay particular attention near the end of
the section to the use of basis and subbasis.
2.15. THE PRODUCT TOPOLOGY ON X Y 23
2.15 The product topology on X Y
The product topology on X Y seems natural. It is. But it is not natural
for the obvious reason that the denition is natural. The obvious reason is
that its basis is the usual (wrong) rst guess as to what would be a reasonable
topology to put on X Y if X and Y have topologies. This leads to some
points.
The usual rst guess is given as the basis in the denition of the product
topology on Page 86. My apologies if this would not have been your rst guess.
You should understand the books example as to why it is wrong to use the
basis as the full topology.
The product of two topological spaces is generalized later to an arbitrary
product of topological spaces. The sophistication gained after the study of the
product of two spaces leads to a usual rst guess for a topology on an arbitrary
product. This usual rst guess is not wrong in that it does not form a topology.
The usual rst guess is wrong because it is not the one given by the usual
denition. The reason for this is based on an omission in the current section.
The current section completely omits motivation. It cannot give any motivation
because there has been no mention of continuous functions. So you will have to
learn the material in this section without motivation. If you want motivation,
you will have to peek ahead to Theorem 18.4 on Page 110 and Theorem 19.6 on
Page 117. The motivation from these theorems is not complete enough for me.
When we get there, I will give more motivation.
Theorem 15.2 gives a better idea of what to expect from the topology that is
put on an arbitrary product, but it will not be clear why it should give a better
idea.
All the material in the section should be learned.
2.16 The subspace topology
Subspace topologies are everywhere. The previous two sections can build quite a
few topologies, but with the subspace topology, the supply of examples explodes.
More topologies are introduced by the subspace topology than by any other
technique. I have no way to justify this statement, but I dont particularly care.
Perhaps there can be doubts about this statement, but no one can claim that
it is obviously false.
Subspace topologies introduce the word in that cannot be omitted from
this point on. Since any subset of a topological space can be a topological space,
and since the notion of open changes as you move from one space to another,
it is now forbidden to use the word open without putting the word in behind
it.
If R is given the order topology then [0, 1) is not open in R, it is not open
in (1, 1), but it is open in [0, 1), it is open in [0, 2), it is open in [0, ) if all of
(1, 1), [0, 1), [0, 2) and [0, ) are given the subspace topology.
24 CHAPTER 2. TOPOLOGICAL SPACES
All of the material in the section should be learned. This includes the ex-
amples.
Lemma 16.2 can be shortened to say open in open is open which is cute
and worth remembering, but it does voilate the rule of always following the
word open by the word in.
2.16.1 exercises
Exercise 1 is not hard, but it is a study in being careful with denitions. The
same applies to Exercise 2.
Exercise 3 is a study in the proper use of the word in.
Exercise 4 is sadly misplaced without the introduction of continous functions,
but it is important anyway and should be done. It is the rst exercise involving
a function in this chapter and it would be good if you can gure out which
properties of functions and their interactions with sets are used in the arguments.
The wording of Exercise 5 is confusing. The book says let X and X

denote
a single set in . . . , but it really means that X and X

denote the same set in


. . . . Or, if that is too wordy, then it simply means X = X

. Then the book


says in the topologies . . . which is hard to interpret. Are they sets in a larger
topological space? I dont see any reason why they should be. The problem
makes perfect sense if instead of the wording given in the problem it said that
T is a topology on X and T

is a topology on X

with X = X

. Similar
treatment would apply to Y and Y

. Of course, if the pair point of view had


been emphasized, then (X, T ) and (X, T

) could have been given as topological


spaces and the commonality of the two sets would have been clear.
Exercises 6 through 10 test understanding of the denitions. The word
compare in 9 and 10 is vague. I suppose that it might mean that you should
see if the words ner or coarser apply to the pair of topologies mentioned.
Other forms of comparison get into the realm of comparison of properties, and
properties of topological spaces and their subsets have not yet been covered.
2.17 Closed sets and limit points
This is a large section with a lot of material. It is all important.
2.17.1 Closed sets
Some comments.
1. Closed sets are complements of open sets, but closed is not the opposite
of open. The book makes this point emphatically and you should internalize
all the discussion that goes around it.
2. Just as open must be followed by in, so must closed be followed by
in for the same reasons.
3. Closed has a connection to limit and convergence that is easier to absorb
than the connection of these notions to open. If something that is capable
2.17. CLOSED SETS AND LIMIT POINTS 25
to converging (sequence, for a familiar example, or net or lter for unfamiliar
examples) is contained in a closed set and it manages to converge to a limit,
then the limit will be in the closed set. Thus elements outside a closed set are
far away from the closed set.
4. Statements about closed sets come with proofs. At this point, it is just as
important to know the proofs as it is to know the statements. All the basic facts
are triumphs of basic facts from set theory. For example, Theorem 17.1 is built
on the DeMorgan laws. Theorem 17.2 is built on the way that set dierence
copperates with intersection.
5. Theorem 17.5 is important since it motivates the denition of limit point
on the next page. It also motivates the other denition of closed which is
given as the second half of the statement of Corollary 17.7.
The phrase U is a neighborhood of x is discussed on Page 96. On Page
97, the book points out that dierent authors mean dierent things by the
word neighborhood. The book adopts one convention, but I would like to not
contradict the book but at least not adopt the convention. Thus I would like
to always use the phrase u is an open neighborhood of x to add emphasis
(and perhaps clarity in the presence of authors of other books) to the openness
assumption of the book.
More on that phrase. Many people (including me) say U is open about x
with the word about implying that x is in U. Of course this means that U is
an (open) neighborhood of x. This sort of wording often gets used in proofs with
words like Let U be open about x to mean that U is an open neighborhood
of x.
In the denition of limit point is the provision other than x itself. This is
annoying. Get used to it. Study the example. The point (pun unavoidable) is
that a set A can contain an element that really does not deserve to be a limit
point of A.
The key to the alternate denition of closed given by Corollary 17.7 is
Theorem 17.6.
By the end of Corollary 17.7, you will have a train of ideas that leads to
the corollary starting with the dention of a closed set. You should familiarize
yourself with the entire train of ideas. That is, know the proofs.
2.17.2 Hausdor spaces
We encounter our rst restriction. The property Hausdor is a property that
a topological space can either have or not have. The book denes Hausdor
space without dening Hausdor. The word is used without being followed
by space so often (see Problem 9 on Page 101) that it should be introduced
that way. So we say that a topological space is Hausdor if it is a Hausdor
space.
The word Hausdor becomes a restriction if one only looks at Hausdor
spaces. Or it is a restriction if one has a theorm that needs Hausdor as a
hypothesis. Restrictions are common. The denition of topological space is so
exible that it is hard to nd theorems that apply to all topological spaces. The
26 CHAPTER 2. TOPOLOGICAL SPACES
interesting (but restrictive) hypotheses are the ones that show up as hypotheses
of the largest number of theorems. The hypothesis is Hausdor is one of the
most interesting from this point of view.
As often happens (and as mentioned in the book) it is possible to modify a
hypothesis or dention to get denitions with similar but not identical eect.
The book points out that T
1
is close to Hausdor. Problem 15 on Page 101
shows that there is a denition of T
1
equivalent to the one given on Page 99
that makes T
1
look even more like Hausdor.
One of the motivations the book uses for the notion of Hausdor is the
behavior of the limit of a sequence. The denition of limit of a sequence is a
combination of the denition of converges to on Page 98 and the denition
at the top of Page 100. You could try to compare this to the dention of limit
of sequence of real numbers that you are used to, but that can wait until a
bit later. Theorem 17.10 about limits of sequences is a good illustration of the
power of the assumption that a space is Hausdor.
Theorem 17.11 is good evidence that Hausdor topologies are not all that
rare.
The role of the open sets in the denition of convergence of a sequence will
be discussed in class.
2.17.3 Exercises
Exercise 1 is a converse to Theorem 17.1. Exercises 2 and 3 have corresponding
facts about open sets to compare to. Exercises 4 and 5 are basic.
Exercises 6 and 7 are important in that it is easy to get parts wrong and
it is important to know why. Part (b) of 6 and Problem 7 need be considered
carefully.
Exercises 8 and 9 are a good reviews of the denitions.
Exercises 10 through 12 review the basic denitions. Exercise 13 is a good
test of how thoroughly you have learned the material and how exible you are
in approach. There is more than one way to approach this problem and some
are easier than others. This makes the problem sound hard. It is not, but it
can appear hard if you are not yet used to the terms.
Exercises 14, 16, 17 and 18 are tests of how well you can work with the
concepts of the section.
Exercise 15 was mentioned above as giving another way to dene T
1
.
This book sometimes introduces important notions in exercises. Exercises
19 and 20 do that with the notion of boundary. They are important exercises.
Exercise 21 has a star. It is somewhat harder than the others and less
fundamental. It is more of a puzzle than an exercise. However, the no more
than 14 aspect can be handled neatly with a few key observations. In a certain
sense, one must develop an understanding of the algebraic behavior of the
two operations closure and complement. To do this, give each a symbol
(obviously it will not do to use c for both) and see what happens when they
are combined in various ways.
2.18. CONTINUOUS FUNCTIONS 27
2.18 Continuous functions
2.18.1 Immediate concerns
Certain exercises are imperative after reading the denition of a continuous
function.
1. In the exercises, Exercise 1 is an implication. Both directions are true and
should be written out.
2. Another is the content of the rst paragraph on Page 103.
3. Another is the content of the second paragraph of Page 103.
4. Another is the equivalence of (1) and (3) of Theorem 18.1.
5. Another is the equivalence of (1) and (4) of Theorem 18.1.
Of course most of the above have been done for you already. You should
learn them well and perhaps try to prove them without looking at the book.
Item (2) of Theorem 18.1 is nice, but not used often. However, it is a good
exercise in certain techniques.
Other aspects of the section are as follows.
2.18.2 Restrictions and inclusions
Theorem 18.2 is full of trivialities, but they are important in that they review
the denitions of several concepts, most importantly the concept of the subspace
topology. Item (b) in that theorem lies behind a couple of the other items.
2.18.3 Continuity in pieces
Item (f) of Theorem 18.2, Theorem 18.3 and Exercise 9 go together. They
combine into one result something that can be split into two steps.
Step 1. There is a notion of generating a topology that is not the same
as generating a subgroup or the notion summarized by the words basis or
subbasis. Let X be a set with topology T. Let A be a family of subsets of
X. Each A A gets the subspace topology from T. We say that A generates
the topology T if for every B X, we have that B is in T if and only if for
every A A, it is true that B A is open in A under the subspace topology
on A. The only if direction will always be true and is useless. It is the if
direction that is important. It basically says that testing for openness can be
done on small sets (the sets in A).
1a. You can show that if A is a collection of open sets in X whose union is
all of X, then A generates the topology T. This is the content of Item (f) of
Theorem 18.2 and the proof can be derived from the proof of this item.
1b. You can show that if A is a nite collection of closed sets in X whose
union is all of X, then A generates the topology T. This is the content of
Exercise 9(a) and if you can do one, you can do the other.
28 CHAPTER 2. TOPOLOGICAL SPACES
1c. You can show that if A is a locally nite collection of closed sets in X
whose union is all of X, then A generates the topology T. This is the content
of Exercise 9(c) and if you can do one, you can do the other.
Step 2. You can show that in the setting of Step 1, if Agenerates the topology
T, then a function f : X Y is continuous if and only if its restriction to every
A A is continuous using the subspace topology on A. Of course only the if
direction is interesting. This is the easier of the two steps. It basically says that
testing for continuity can be done on sets in a collection if testing for openness
can also be done on the sets in the collection.
The summary is that Item (f) of Theorem 18.2, Theorem 18.3 and Exercise
9 are really statements about collections that generate a topology.
Theorem 18.3 and its mild generalization Ecercise 9(a) are used all the time.
The claim that
f(x) =
_
2x, x 1,
x + 1, x 1
is continuous is usually argued in calculus by saying that the two parts of the
denition agree at x = 1. This check only argues that f is a function. What
is really going on is the pasting lemma (18.3). It is clear from the formulas
(both are dierentiable) that the restrictions of f to the closed sets (, 1] and
[1, ) are both continuous, and the union of these two sets is all of R. Now the
pasting lemma guarantees the continuity of f on all of R.
2.18.4 Homeomorphisms
Homeomorphisms are extremely important. They are to topological spaces what
isomorphisms are to algebraic structures (groups, vector spaces, etc.). A home-
omorphism f from X with topology T to Y with topology S carries open sets
in T as a one-to-one correspondence to the open sets in S. (This sentence is
even worth proving as a simple exercise.) Thus homeomorphisms carry the
topological structure of one space over to the topological structure of the other.
By and large, anything dened in a topology course using only open sets
(we will see denitions using other means later that behave badly under home-
omorphisms) that apply to one space will apply to a space homeomorphic to it
if properly carried over by a homeomorphism. So for example, if f, X and Y
are as in the last paragraph, and A X and x X are given, then x is a limit
point of A if and only if f(x) is a limit point of f(A). Find an example where
some aspect of this last sentence fails if f is only known to be a continuous
function.
The book denes topological property rather obliquely. It does so because
this is the right denition. The denition is in terms of an eect rather than a
cause. If you cannot change something by a homeomorphism, then whatever it
is that is not changing must be a topological property. Thus the denition of
topological property is dened in terms of the behavior of the property and
not on the manner in which the property was dened. This takes some getting
used to.
2.19. THE PRODUCT TOPOLOGY 29
Denitions that discuss only open sets are obviously topological and the fact
that they are topological surprises no one. We will meet quite a few obvious
ones. There are various other properties that can be dened on a topological
space where the denition does not look all that topological. We will not meet
the less obvious ones in this course, but you will in algebraic topology. For
example, it was not known for a while that the homology groups (wait until
algebraic topology) as dened before (say) 1940 were topological properties.
They are and the proof took some doing.
2.18.5 Products
Theorem 18.4 partly justies the denition of the product topology on X Y .
It does not completely justify it. It leaves open the question of whether there
is another topology that could be dened on X Y that would also have the
property guaranteed by Theorem 18.4. We will discuss this question in due
time.
2.18.6 Exercises
Exercise 1 has been mentioned and the content of Exercise 2 has been mentioned
when f is a homeomorphism. Exercise 2 only assumes that f is continuous.
Some properties are preserved by continuous functions. It is not necessary to
have homeomorphisms all the time.
Exercise 3 combines larger and smaller topologies with continuous functions.
It is easy and should be done.
Exercise 4 is used later and should be done and also remembered.
Exercise 5 is easy if you think about the denitions carefully.
Exercise 6 is easy, tricky if you have not seen things like it before and not
crucial.
Exercise 8 is a good review of facts, denitions and previous results.
Exercise 9 has been discussed.
Exercise 10 is useful.
Exercises 11 and 12 go together, are nice but not crucial.
Exercise 13 is a nice argument that uses a lot of techniques that are useful.
It will be gone over. You should attempt it.
2.19 The product topology
When the number of factors of a product is innite, then the obvious topology
is no longer the topology with the desired properties. This is explored to some
extent in the book. The right denition is given minor motivation before the
denition, and given much stronger motivation after the dention. But it is
never explained why the right denition is the only right denition.
The discussion through the top half of Page 113 is ne as far as it goes.
The denition of the Cartesian product of an indexed family of sets has been
30 CHAPTER 2. TOPOLOGICAL SPACES
mentioned in class before and is necessary for the discussion. The break of the
denition into two parts (the two denitions on Page 113) is good.
The comments in the second paragraph of Page 114 need to be emphasized
more. The comments say that there are two ways of looking at the set of
functions from J to X. One way is just what the words say: it is the set of
functions from J to X. The other way is to think of it as a product of copies
of X indexed over J. The reason that this is important is that this section will
put a topology on that product if X is a topological space. That means the set
of functions from J to X will be given a topology. This lets us talk about a
space of fuctions rather than just a set of functions. This will come up again
in important ways.
The dentions on Page 114 are needed. The box topology must be un-
derstood only as far as knowing that it is the topology with the undesirable
properties.
The basis for the product topology discussed at the top of Page 115 is more
important than the subbasis that is used in the denition at the bottom of Page
114. The subbasis does have some uses in arguments, but the basis is much
more likely to be used when trying to prove something.
Theorem 19.1 is important only in that its statement eciently discusses the
dierences between the box and product topologies.
Theorem 19.2 is a typical theorem of the type it suces to consider a basis
instead of all open sets in a denition.
The next three theorems are straightforward.
Theorem 19.6 and the example that follows makes clear the important dif-
ferences between the box and product topologies. The importance of Theorem
19.6 is hidden in the fact that it has an if and only if statement in it, and
not just an if statement. Exercise 10 goes a little way towards justifying the
product topology as the right topology to put on a Cartesian product. To com-
plete the justication, we need some preliminary exercises. We dont need them
all, but once we start, we may as well give them all.
Let f : A B be a function. If A and B have topologies, then we can
discuss the continuity of f. We can discuss what makes it easier for f to be
continuous and what makes it harder.
Which makes it easier for f to be continuous, a large topology on B or a
small topology? If a topolgy is given on A what is the smallest topology on B
that makes f continuous? What is the largest? Is there a topology on B that
guarantees the continuity of f no matter what topology A has?
Which makes it easier for f to be continous, a large topology on A or a
small topology? If a topolgy is given on B what is the smallest topology on A
that makes f continuous? What is the largest? Is there a topology on A that
guarantees the continuity of f no matter what topology B has?
The last two paragraphs asked for examples of topologies that guarantee
continuity no matter what the other topology is. Give an example of a func-
tion that is guaranteed to be continuous no matter what topologies are used on
both domain and range? Show that there is such a function between any two
sets?
2.19. THE PRODUCT TOPOLOGY 31
In the setting of Theorem 19.6 a test is made which determines the continuity
of a function f. The result of the test has to do with the topology on the
product, but the details of the test (checking the continuity of the f

) do not.
So if Theorem 19.6 is to be true, then this forces the topology on the product
to make a certain set of functions continuous. This is from the if part of
the conclusion. Does this force the product topology to be large or small? (Or
perhaps it should be asked whether it forces the product topology to be not
too large or not too small?)
Note that if Theorem 19.6 is to be true, then any function that does not
pass the test must not be continous. This is from the only if part of the
conclusion. Does this force the topology to be large or small? (Or perhaps it
should be asked whether it forces the product topology to be not too large or
not too small?)
We now try to make large and small more specic. To one of the
questions just asked, the answer is the product topology cannot be too small.
This can be made more specic by stating what the topology must contain. To
the other question, the answer is the product topology cannot be too large.
This can be made more specic by stating what the topology must be contained
in.
It can be shown that the same collection of sets can be used in both answers.
In an attempt to make things less confusing, let P be the collection of sets in the
product topology. Of course we know that P is what the book calls the product
topology, but let us pretend that we dont. It is a perfectly good topology and
we can use it no matter what we call it. Then we have the following.
Consider the identity function from a product to itself. Since the product
is both domain and range, we have to use the words domain and range to
distinguish the two roles that the product plays. Put the topology P on the
domain. This is to play the role of A in the statement of Theorem 19.6. Assume
that there is a topology on the range that satises the conclusions of Theorem
19.6. That is, any function into the range is continuous if (we just use the if
part here) all the f

are continuous. The important observation at this point


is that all the f

are in fact continuous because we are using P on the domain.


Thus if Theorem 19.6 is to hold, the identity must be continous. Argue that
this means that the product topology (i.e., the right topology on the product
that cooperates with 19.6) must be contained in P. This makes specic the
not too large aspect of the product topology.
Now we deal with the not too small aspect. Let Q be a topology on the
product that is contained in P but that is strictly smaller than P. Argue that
for some the topology Q does not have one of the subbasic sets dened at
the bottom of Page 114. Then argue that for this , the projection

is not
continuous.
Now we return to the identity function from the product to the product. Put
the topology Q on both domain and range and pretend that we think that Q is
the right topololgy to be called the product topology on the range. Argue that
this would violate Theorem 19.6 because it would make a fuction continuous
(the identity) that does not pass the test. Conclude that the right topology
32 CHAPTER 2. TOPOLOGICAL SPACES
to put on the product cannot be strictly smaller than P.
The the only topology on the product that cooperates with all aspects of
Theorem 19.6 must both contain P and be contained in P. This justies calling
P the product topology.
2.19.1 Exercises
Exercises 1, 2 and 3 are routine and are good checks of understanding. Exercise
2 requires care in interpreting denitions.
Exercise 4 is a check that denitions are understood. What is the homeo-
morphism?
Exercise 5 is another check of denitions, and such checks should be getting
tiresome by now.
Exercises 6, 7 and 8 are more advanced checks of understanding the topolo-
gies. They are also more fun.
Exercise 9 is too trivial to spend any time on. If you dont get it skip it.
Exercise 10 is not worth it at this point. If you went through the justication
of the product topology, then the concepts in the exercise wont add much. Some
of the ideas get repeated later.
An extra exercise is worth it.
Let {X

}
J
be an indexed collection of topological spaces and let
Y =

J
X

be given the product topology. Let c be an element of Y and let be in J.


We dene a function f : X

Y . Since f(x) for x X

will be an element
of a product, we must say what its coordinates are. That is, we must dene
f(x)() for every J. We set
f(x)() =
_
x, = ,
c(), = .
Sometimes the image of f is called the slice parallel to X

through c. Prove
that f is a homeomorphism onto its image (which inherits the subspace topology
from Y ).
2.20 The metric topology
The metric topology is important enough to get two sections in the book. Ev-
erything on Pages 119 and 120 are important.
The top two paragraphs of Page 121 are opinions and are somewhat ques-
tionable.
The two denitions on Page 121 are important as is the discussion of the
example of a property that is not topological. Another more compelling example
of a property that is not topological will come up later.
2.20. THE METRIC TOPOLOGY 33
The note at the top of Page 122 is important. It says that small sets deter-
mine the topology of a metric space.
The denitions on Page 122 are important. The comment about the Eu-
clidean metric is valid. (The book consistently uses names without capitalizing
them. Most books follow the tradition that all names are capitalized with one
exception. The word abelian, named for Niels Henrik Abel, is never capital-
ized.) The square root in the Euclidean metric makes it a pain to work with.
The square metric has other names and is much easier to work with.
The argument that the two metrics considered on Page 122 give the same
topology is important. Lemma 20.2 is a key lemma and it should be appreciated
for the fact that several containments are involved (one metric contained in
another, one ball contained in another) and that the containments might not go
in a direction that you might guess without thinking it through. It is important
that you dont remember which way the containments go. Instead you should
get familiar enough with the pictures to have in mind so that you can tell which
way the containments go in just a few seconds.
Theorem 20.3 does two things. It shows that two metrics give the same
topology, and it shows that they give the same topology as the product topology.
The techniques of the second part are much more important than the techniques
of the rst part.
The calculations behind the rst part really come from the pictures of the
balls in the two metrics. The pictures are not given in the book and you should
work out what they are.
The second part is really an exercise in working with the word max. Since
the word max is never used in the argument, you should see where it ts in.
The discussion on Page 124 is important and justes the need for Theorem
20.1.
Recall that X
J
can be thought of in two ways. One is that it is a set of
functions and the other is that it is a product of copies of X indexed over J.
The rest of the section discusses topologies to put on this set of fuctions (giving
dierent spaces of functions) when X = R. We already have one topology, the
product topology. The uniform metric gives another.
The uniform metric is more important than Theorem 20.4 would have you
believe. We got used to the idea that the product topology is the right
topology to put on a product and that the box topology is wrong. However,
that does not mean that they are both useless. The box topology has its uses,
rare though they may be. The uniform topology is used quite often in spite of
the fact that it diers from the product topology. The uniform metric is dened
here for functions into R. It can be dened for functions into any metric space
with no extra work, but R will do for now.
The proof of Theorem 20.5 is long. However, it is important in that it gives
more examples of how to work with sup and max.
34 CHAPTER 2. TOPOLOGICAL SPACES
2.20.1 Exercises
Exercise 1(b) introduces what is often known as the taxi metric. It is not crucial.
Part (b) is important for analysis. It is not crucial here either.
Exercise 2 is easy, but not crucial.
Exercise 3 is more important.
Exercise 4 is good if you want to understand the topologies. It is more
important to understand the examples with respect to the product and uniform
topologies, but if you are going to do that, you may as well include the box
topology.
Exercise 5 is a good test of understanding. The notation R

is terrible.
Exercise 6 is a one idea exercise. What is the key idea?
I have no idea how important Exercise 7 is.
Exericse 8, is very important in analysis. Exercise 9 shows how much work
is needed to deal with the Euclidean metric. Exercise 10 is also more important
for analysis. None of these is crucial to this course, but they might be for others.
Exercise 11 can be skipped unless you think it is cute.
2.21 The metric topology (continued)
The rst section on the metric topology concentrated on the metric and a bit
on continuous functions. In a sense, the exercises on sequences were exercises
on continuity. (Give the set {0, 1, 2, . . . , } with larger than any n N the
order topology. Then an innite sequence x is a function on {0, 1, 2, . . .}. An
innite sequence x converges to L if and only if extending the function x to take
to L is continous.)
This section adds other topological notions to the discussion.
The rather quick statements in Page 129 are important. Exericse 1 forces
you to do one statement. The claim about the Hausdor property should be
done out in extreme detail just to see how it works. The bottom of the page
shows that metric spaces can handle continuity in much the same way that
continuity is handled in calculus.
Lemma 21.2 and Thoerem 21.3 have a common theme: in metric spaces
sequences determine properties. That this is false in non-metric spaces is crucial
to know. That this is true in the more general setting of rst countable spaces is
also important to know. The contruction of B
n
near the top of Page 131 shows
that in rst countable spaces, one can insist that the countable basis at each
point be a sequence of nested sets.
Lemmas 21.4 and 21.5 verify the obvious and should be done, but can be
skipped if you are feeling overwhelmed.
The topic of uniform convergence that starts near the bottom of Page 131
and continues through Theorem 21.6 is very important. Exercise 7 ties it to
topology. This point is made just below the middle of Page 132. (Recall that
the book only denes the uniform metric for functions into R. Exercise 7 holds
in much more generality with no more work.) The argument in the proof of
2.21. THE METRIC TOPOLOGY (CONTINUED) 35
Theorem 21.6 is often called the epsilon over three argument. The result of
Theorem 21.6 together with Exercise 7 and Lemma 21.2 says that in the uniform
metric on the set of functions from a topological space to a metric space, the
continuous functions form a closed set.
The book misses a point. It also is not careful with the way that it writes
out quantied sentences. The right way to say that the sequence of functions
(f
n
) converges uniformly to f is to say
> 0, N N, x X, n > N, (d(f
n
(x), f(x)) < ).
This allows us to compare this denition to the dention that the sequence
of functions (f
n
) converges pointwise to f. This reads
> 0, x X, N N, n > N, (d(f
n
(x), f(x)) < ).
The dierence is the reversal of a in uniform convergence to in point-
wise convergence. So in uniform convergence, one n works for every x, but in
pointwise convergence each x can have its own private N that works for it.
The relevance to the current topics is that there is a parallel to Exercise 7
for pointwise convergence.
Exercise 7
1
2
. Let X be a set, and let f
n
: X R be a sequence of functions.
Let R
X
have the product topology. Show that the sequence (f
n
) converges
pointwise to the function f : X R if and only if the sequence (f
n
) converges
to f as elements of the topological space R
X
with the product topology.
This begins to justify the existence of the uniform metric. The uniform
metric leads to a topology that is larger than the product topology. That
makes it harder to be continuous into the uniform metric than into the product
topology. For sequences, it makes it harder for sequences to converge. Thus in
the uniform metric it is not possible for a sequence of continous functions to
converge to a non-continous function. But in the product topology (the topology
of pointwise convergence) it is. There is an example in the exercises.
2.21.1 Exercises
Exercise 1 justies a quick remark at the beginning of the section.
Exercise 2 is trivial but it introduces an important concept.
Im not sure about Exercise 3. I suppose it is dierent from Theorem 20.5,
but I cant see it as an important dierence. The main change is from R to
arbitrary metric spaces. Perhaps it is just a good review of Theorem 20.5.
Exercise 4 is nice but not crucial.
Exercise 5 is hard to get excited about.
Exercise 6 is relevant if you have done Exercise 7
1
2
. What is the limit of the
sequence (f
n
) in the product topology (topology of pointwise convergence)? Is
it continuous?
Exericse 7 has already been mentioned.
It is hard to get excited about the rest of the exercises. Do any that look
interesting.
36 CHAPTER 2. TOPOLOGICAL SPACES
2.22 The quotient topology
We will cover this section in spite of the fact that the book labels it optional.
It is used too often to skip.
The quotient topology is easy to dene and a little harder to visualize. It is
also often set up in a form that seems dierent from its denition.
If f : X Y is a surjection from a topological space to a set, then there is
a largest topology on Y that makes f continous. This topology is the quotient
topology.
You should show that the denition in the book has this property.
The word saturated is useful and is dened on Page 137. The notions
of open map and closed map are moderately useful and their denitions are
easy to remember. Open maps are seen discussed more often than closed maps.
The projection from a product to a factor is an open map. This was an earlier
exercise. However, neither notion is crucial.
The verication that the quotient topology is a topology comes down to the
nice behavior of inverses of functions and is discussed on Page 138.
Page 139 is where the quotient topology is set up in a form that seems
dierent from the denition. We will argue that it is really not dierent.
Let f : X Y be a surjection. Let X be partitioned by the sets f
1
(t)
for t Y . As a partition, it might be suspected that these sets are equivalence
classes under some equivalence relation. They are. The equivalence relation on
X is the relation x is related to y if f(x) = f(y). The properties of = make
it trivial that this is an equivalence relation.
Now it is also trivial that there is a one-to-one correspondence between
the set of equivalence classes and Y . Further the one-to-one correspondence
is derived from f in that if [x] is the equivalence class containing x (it is the
class of all elements of X that map to the same place as x under f), then [x]
corrsponds to f(x) in Y .
So the denition at the top of Page 139 is not all that special. Every sur-
jection comes from a function that is like taking each element to an equivalence
class that contains it. On the other hand, every partition (set of equivalence
classes from an equivalence relation) leads to a surjection to the set of equiva-
lence classes by sending each element to the eqivalence class that contains it.
We continue the discussion without topologies for a minute.
In the situation above, Y is universal for the partition given. If g : X Z
is a function that preserves classes (if x and y are in the same class in that
they map to the same element of Y under f), then there is a unique function
h : Y Z so that hf = g. Just let h send f(x) to g(x). This is well dened
by the hypotheses, it is the only choice that works, and it denes h on all of Y
since f is onto.
If X is a topological space, then the quotient topology on Y lets us do the
discussion above all over again with topologies. We get that f is continous, and
if g is continous, then the unique h discussed above is continous by the denition
of the quotient topology.
2.22. THE QUOTIENT TOPOLOGY 37
If every surjection comes from a partition (eqivalence relation) and every
equivalence relation leads to a partition, why have both approaches? After all,
they lead to the same thing. The answer is convenience. In many situations
the description of the equivalence classes (partition elements) is quite easy. If I
want to glue a square to a disjoint triangle along an edge, then I can say the
following.
S A

















T
B
Let i be a one-to-one correspondence from side A of square S to side B of
triangle T. The obvious one will do. Now we put a partition on S T. Every
point not in A B gets its own private equivalence class consisting of just that
point. All other equivalence classes have two points each. These come from
points in A. For each x A we form the class {x, i(x)}. Now we take the
quotient topology. We say that we have glued T to S along A by the fuction
i.
This can be varied tremendously. For example, we can use functions that
are not one-to-one correspondences.
The book goes to great lengths to show how to visualize a quotient topology.
You must keep in mind that the cases considered are quite special. There is no
general way to visualize all quotient topologies. They can be quite messy.
(I just discovered that the book has been using xy since Page 85 for pairs
of elements in R R to avoid confusion with open intervals. I suppose this is
acceptable, but it will be found in no other book.)
Example 4 is less important than Example 5. This is especially true because
the book confesses that in Example 4 one can show that the quotient space is
the 2-sphere. Later when we are ready to show that this is true, we will come
back to the example. The picture of Example 5 is easier to deal with since the
equivalence classes are smaller. In Example 5, the largest equivalence class has
four points.
The points made at the bottom of Page 141 are worth knowing, but not
dwelling over. The main point is that very little is preserved by quotient maps.
The exact examples are not important for now.
Theorem 22.1 is probably not needed for a rst look at quotient spaces.
Theorem 22.2 was discussed above and gives a crucial property of quotient
topologies. Corollary 22.3 is an easy conequence.
The examples on Page 143 and Page 144 can be skipped for now. They show
how bad quotient maps can be.
A point made earlier about topological spaces is relevant. There are many
strange topologies. The ability of the denition of a topology to include such
strange spaces should not detract from the fact that the denition includes
many very useful topologies. Similarly, the ability of quotients to create strange
38 CHAPTER 2. TOPOLOGICAL SPACES
topologies should not detract from the fact that they create some very useful
topologies as well.
2.22.1 Exercises
The only exercise worth doing is 2 which introduces an important notion.
To give an exercise that is simple enough to do and real enough to be useful,
consider the following.
The topological space X is the closed interval [0, 1]. The topological space
Y is the closed interval [2, 3]. Give them each the standard topology that they
get from R. Consider two quotient topologies dened by equivalence classes.
For the rst (call it A) the only class with more than one point is the class
{1, 2}.
For the second (call it B) there are only two classes with more than one
point. One is the class {1, 2} and the other is the class {0, 3}.
Now give each of A and B the quotient topology from X Y .
Show that A is homeomorphic to [0, 1] with the standard topology and show
that B is homeomorphic to the unit circle {(x, y) | x
2
+ y
2
= 1} with the
topology that it inherits from the standard topology in the plane.
Recall that the standard topologies on R and R
2
are now known to be metric
topologies. This may help. Also remember that homeomorphisms are functions.
2.22.2 Topological groups
We will not cover topological groups. This is not an indication of any lack of
importance. The set of n n matrices with real entries can be regarded as
an organizational structure imposed on R
n
2
. Since R
n
2
is a topological space
(usual topology), the set of nn matrices gets a topology. Under addition, this
is a topological group. It is not that interesting. Much more interesting is the
set of nn matrices with non-zero determinant. Now we get a subspace of R
n
2
that is a topological group under multiplication. This is much more interesting.
Variations can be made. One can restrict the matrices and look only at
matrices of determinant one, or matrices of determinant 1, or matrices that
are orthogonal, etc. Many extremely important examples of topological spaces
are topological groups.
This nishes Chapter 2.
Chapter 3
Connectedness and
compactness
The intro to Chapter 3 (Pages 147148) is worth reading.
3.23 Connected spaces
It has been stated by some that most of topology concerns questions of connec-
tivity. I have no reason to dispute this.
Connectedness forms a theory. It consists of denitions and intermedi-
ate results that form a nice package that should be mastered before even
attempting any problems.
You will note that this section only presents the package. It gives no exam-
ples of connected spaces. This is unusual with denitions. Usually a dened
concept is immediately followed by examples. To ll in that gap, I give you
two exercises to be done at a point that you consider appropriate. (a) Prove
that any one-point topological space is connected. (b) Prove that any indiscrete
topological space is connected. That should hold you until the next section.
The denition of a separation has several details (1. non-empty, 2. disjoint,
3. open, 4. union is everything) that cannot be forgotten.
The alternative denition in the middle of Page 148 is useful, but not a
replacement.
Lemma 23.1 tries to combine two ideas. One is the passage to subspaces,
and the other is the view in terms of limit points. The combination is a massive
failure. It gives the impression that the two ideas are linked. They are not.
There is a statement that does not mention subspaces that deals only with
limit points. Then the statement that deals with subspaces alone is trivial. A
subspace is connected if and only if it is connected as a topological space under
the subspace topology.
The results from Lemma 23.2 through Theorem 23.5 should not only be
learned, but their proofs should be learned. This means being able to write
39
40 CHAPTER 3. CONNECTEDNESS AND COMPACTNESS
them all out with their proofs without looking at the book.
Theorem 23.6 depends on two facts. The rst is that being connected is
a topological property. That is, if two spaces are homeomorphic, then one is
connected if and only if the other is connected. Of course, this is easy given
Lemma 23.5.
The second fact I cannot nd anywhere in the book before it is used on Page
150. Since it concerns homeomorphisms, it has to come after Page 108, and
since it concerns products, it has to come in or after Section 19. I cannot nd
it. The statement is that if
P =

I
X

has the product topology, if is one index in I, and if f P is one element of


the product, then
f

= {g P | = , g() = f()}
is homeomorphic to X

. In fact the homoemorphism is just the projection

.
You should take as an exercise (which is just a review of denitions) that this
is true.
The set I label f

above is the slice through f parallel to X

. The book
uses the words horizontal and vertical since it is looking at a product of
just two spaces.
Now Theorem 23.6 follows easily from this and the results that come above
it.
Example 7 gives the idea behind why an arbitrary product of connected
spaces is connected. Note that it uses Theorem 23.4 without giving the number.
This is an illustration of the fact that results from 23.2 through 23.5 are regarded
as so fundamental that everyone is expected to know them cold.
3.23.1 Exercises
Exercise 1 is easy. Exercise 2 is easy to do wrong. It is not an induction proof.
(I assume that n comes from Z
+
.) Induction only gives the union of nitely
many.
Exercise 3 is easy. Exercise 4 is not important.
Exercise 5 is important for the denition.
Exercise 6 is an important review of denitions.
Exercise 7 is a less important review of denitions.
Exercise 8 is interesting. It is covered more thorougly in an exercise from a
later section.
Exercise 9 is a good review of the techniques that prove Theorem 23.6. In
other words, learn the proof of Theorem 23.6 before looking at this problem.
Exercise 10 generalizes Example 7.
Exercise 11 gives a sort of converse to Theorem 23.5. It raises the question
of why there is not a more general converse to Theorem 23.5. So I will ask you
to nd a counterexample to a general converse to Theorem 23.5.
3.24. CONNECTED SUBSPACES OF THE REAL LINE 41
Exercise 12 should be attempted after you nd examples that show that
all the hypotheses are necessary. That is, nd a counterexample when X is
not connected (easy) and then nd a counterexample when Y is not connected
(almost as easy). Then do the problem.
3.24 Connected subspaces of the real line
The most famous connected topological spaces are the intervals in R.
Theorem 24.1 proves this. It does so in greater generality, but all that is
going on is that the proof for R was investigated to see what is really used in
the proof and then the result was written down only mentioning the hypotheses
that are needed. The proof that just assumes that L = R would be no simpler.
The proof of Theorem 24.1 should be learned. Note that this is probably
the most complicated proof seen so far in this book. Learning the proof is a
good exercise at this point. There will be more complicated proofs later and
the practice is necessary.
Theorem 24.3 justies the rst claim of the intro to Chapter 3.
The book likes the ordered square. I prefer well orders. It is a matter of
taste. That means I prefer Example 2 over Example 1. Both are good exercises
in understanding, but neither is critical.
Once some connected spaces are known, then more can be built using the
results of the previous section. Several of them get used in the topic dealing with
paths and path connectedness. Examples 3, 4, 5 are important. In Example
5, the book claims that a parallel to Theorem 23.5 works for path connected
spaces. Prove this parallel. Example 6 is not my favorite. Example 7 is part
of the general culture. The technique used in proving the claimed properties is
essential. It is an example of how to nd a rigorous proof of something obvious.
Note the use of the intermediate value theorem to make the existence of the
points t
n
rigorous.
3.24.1 Exercises
Exercise 1 uses the fact that connectedness is a topoological property. It also
uses the fact that if f : X Y is a homeomorphism and x X, then X {x}
is homeomorphic to Y {f(x)}. Why? This is used in (a) and (c).
Exercise 2 is cute, and illustrative of how certain facts can be used in odd
ways.
Exercise 3 is even cuter and more important. It comes down to a trick. A
hint is to ask what the graph of f has to look like for f not to have a xed
point.
Exercise 4 is easy but not essential.
Exercises 5 through 7 are pretty optional.
Exercise 8 is more important.
Exercise 9 is a nice trick.
Exercise 10 is rather important. Draw pictures.
42 CHAPTER 3. CONNECTEDNESS AND COMPACTNESS
Exercise 11 has its merits.
Exercise 12 is for those that like well orders. It gives an example that is part
of the culture.
3.25 Components and local connectedness
We will cover this section.
The section discusses components and path components. To distinguish or-
dinary components from path components, some books/people say connected
component to mean ordinary component. This bothers me no end. Compo-
nents of all kinds are connected. There are no non-connected components.
However, I have learned to live with it.
The notions of component and path component are developed in this section
is parallel. So there is ample opportunity to compare and contrast. There is
some contrast. The concept of locally path connected ties the two together in
Theorem 25.5.
It surprises me that I cannot nd anywhere that the book points out that
components are closed. They are. Prove it. It takes one line if you have learned
the machinery of connectedness.
Components are not open. It is the concept of local connectedness that
makes them open. This is the content of Theorem 25.3. The parallel for path
components is in Theorem 25.4.
The rest of the section covers the basics of the various notions. These are in
Theorems 25.1 and 25.2.
All denitions, statements and proofs should be learned.
3.25.1 Exercises
To emphasize the learning of proofs over the doing of exercises, I will mention
that there are very few exercises in the section that have any importance. They
are Exercises 4, 6, 8, 10, and none is really all that critical.
In Exercise 4, it is my impression that it is almost entirely covered by the
content of the section. Since the statement of Exercise 4 talks about subsets,
you should think about what is the minimum that has to be proven so that
material from the section can be applied.
Exercise 6 uses a notion that is nice but does not appear frequently. It does
give a good example of how properties are tweaked to discuss local properties
at or near a point.
Exercise 8 has some merit.
Exercise 10 deals with a concept that begs to be dened immediately after
the denition of a connected space. Items (a) and (b) are nice. I assume (c)
shows dierences between the various types of components.
3.26. COMPACT SPACES 43
3.26 Compact spaces
Compactness forms a theory. It consists of denitions and intermediate results
that form a nice package that should be mastered before even attempting any
problems. (Yes, these sentences were copied with one change from the beginning
of the section on connected spaces.)
There are many parallels between the books development of compact spaces
and the books development of connected spaces. They both start with deni-
tions and basic facts, but no examples. The basic facts include ways to get new
spaces with the desired property from old spaces, including continuous images
and products. The basic facts section is then followed by a section giving specic
examples of spaces with the desired property, mostly centered around the real
line. There is then a section about local properties. In compactness, there is
a section in between about dierent forms of compactness that would compare
only weakly to the relationship between connectedness and path connectedness.
The denition does not reveal what is truly going on. This is a situation in
which a general idea is not available to help with the details of the denition.
So the only thing to hold on to is the denition itself. The denition must be
learned correctly. A common error is to assume that to every open cover there is
another nite open cover so that every open set in the second cover is contained
in an open set from the rst cover. This is wrong. This does not make the
second cover a subcover of the rst, it makes the second cover a renement
of the rst. Renements do get used in dierent settings. See the bottom of
Page 245. It is important to remember that compactness uses subcovers and not
renements. Go over the denition until you understand the dierence between
a subcover and a renement.
Since there are no examples in the section, I will give a trivial exercise. Prove
that every topology on a nite set is compact.
I dislike Lemma 26.1. It should be an exercise. However, the denition that
comes just above it and the lemma itself do get referred to a lot and it pays to
have been through it. I will enforce a rule. If you ever use Lemma 26.1, then
you had better be able to prove it on the spot.
Theorems 26.2 and 26.3 invite confusion. Their statements should be remem-
bered without confusing the two of them, and their proofs should be learned.
The proof of Theorem 26.3 is a nice combination of the Hausdor property and
the forgotten lemma. The detailed argument at the end of the proof that shows
the obvious fact that U is disjoint from Y should be noted.
Lemma 26.4 is a practice lemma for a more important lemma of a similar
nature to come later.
Theorem 26.5 is a cousin of the theorem that says that the continuous image
of a connected space is connected. You should learn the proof backwards and
forwards. The book leaves some obvious facts to the reader. You should ll
them in. Why is the collection in the third line a covering of X? It is obvious
is not a proof. Why do the sets A
1
through A
n
cover Y ? Same comment.
Theorem 26.6 is too important to ignore. It is extremely useful. To prove
a homeomorphism, you have to show four things about a function: one-to-one,
44 CHAPTER 3. CONNECTEDNESS AND COMPACTNESS
onto, continuous, open. (Open in the sense of Problem 4 on Page 92.) Theorem
26.6 says that with the right hypotheses, you can do with the rst three.
Thoerem 26.7 is a key theorem. There are many ways to prove it incorrectly.
The correct proof uses a trick that must be learned. The trick is isolated as a
separate lemma (26.8) that the book calls the tube lemma. It is as good a
name as any. All should be learned.
The denition and companion lemma on the bottom of Page 169 are often
used. It simply translates the denition of compactness into closed sets instead
of open using de Morgans laws and the contrapositive. The specic application
mentioned at the end of the section on Page 170 is used often. If you forget this
topic, there will be at least one problem that you will nd almost impossible.
Keep in mind that the denitions are the only things you have to hold onto
in this topic.
3.26.1 Exercises
Exercise 1 is very important. It is trivial if you know the intermediate results
in the chapter. It is not if you dont.
Exercise 2 can be skipped. Exercise 3 should be done.
Exercise 4 is good. If you cannot do the rst sentence now, you will be able
to after seeing how things are done in a section or two. The second sentence is
a review of your examples of metric spaces. Not every metric space is R
n
.
Exercise 5 is a review of the techniques of proof from the section.
Exercise 6 is a review of the intermediate results.
Exericses 7 through 10 can be picked over for anything that sounds interest-
ing.
Exercise 11 does not have the hypotheses that every A in A is not empty. It
should. Then the conclusion would also be able to claim that Y is not empty.
This makes it more interesting. Why is the extra claim true under the extra
hypothesis? This is a warm up for the hint in the book that makes no sense if
not all the A in A are non-empty.
Exercises 12 and 13 are getting a bit technical.
3.27 Compact spaces of the real line
Just as you had to wait a section for examples of connected spaces, you had
to wait a section for examples of compact spaces. The reasons are the same.
The proofs are not trivial. The proof of Thoerem 27.1 is on the same order of
diculty of the proof that linear continua are connected. It is also presented
in similar generality. The proof for the closed unit interval is studied and the
material used is extracted to give a more general proof. However, the proof
for the closed unit interval would be just as complicated. The proof should be
learned. There are two styles of proof for this theorem. The book chooses the
one that is more topological. Its key is Step 2. The next two steps justify the
approach started in Step 2.
3.28. LIMIT POINT COMPACTNESS 45
Theorem 27.3 is great, but too many people remember it wrong. The theo-
rem does not say that a space is compact if and only if it is closed and bounded.
The term bounded only makes sense for metric spaces. The theorem does not
even say that a subspace of a metric space is compact if and only if it is closed
and bounded. It says exactly what it does say (see the book) and no more.
The books comment right after the proof of the theorem only goes part way
towards explaining the problems of remembering the wrong theorem. What can
you say that is stronger? For example, the set [0, ) is closed and not bounded
in the usual metric on R. Can you apply the comment in the book to show that
something goes wrong with the theorem if it is not quoted correctly?
Theorem 27.4 justies one of the comments in the intro to the chapter.
Starting on Page 175 several notions are introduced. Distance from a set
to a point, the Lebesgue number, uniform continuity, isolated points are all
important.
I dont like the proof in the book for Theorem 27.5. Its only merit is that it
does not start in the way that cannot work. The obvious way to start is to
take a nite subcover of A. Try to nish the following proof. For each x X,
let
x
be such that the
x
-ball around x lies in some element of A. The next
obvious thing to try is to look at the cover by all the
x
-balls for all the x X
and take a nite subcover of that. This does not work either. Next we take the
cover of X by the (
x
/2)-balls for all the x X. Now take a nite subcover of
that. Now nish the proof.
Another place that /2-balls show up is in the proof of Theorem 27.6. The-
orem 27.6 is also one of the main reasons for proving the existence of Lebesgue
numbers. There are other applications of the Lebesgue number, but they will
not show up anytime soon.
3.27.1 exercises
Exercise 1 is a curiousity.
Exercise 2 introduces important notions.
Exercise 3 is not essential.
Exercise 4 is a good review of facts.
Exercise 5 is an example of an important class of theorems. This is the
compact Hausdor version. Later we will see the complete metric space
version which tends to be applied more often.
Exercise 6 is critical culture. Everyone must know what the middle thirds
Cantor set is. No part of the problem is all that hard.
3.28 Limit point compactness
This section introduces two competing notions of compactness. The book admits
that one (limit point compactness) has at least one other name. The Balzano-
Weierstrass property is a very well known way to refer to it.
46 CHAPTER 3. CONNECTEDNESS AND COMPACTNESS
There are two points to the section. The rst is that it gives some intuition to
the notion of compactness. Compact spaces are not very roomy. It is impossible
to have an innite set with no limit points in a compact space. The subset Z
in R is innite and has no limit points. Thus R is not compact. The closed
unit interval is compact and there is no way to have a set with innitely many
points in [0, 1] without having at least one limit point in [0, 1]. Note that these
things have to be said carefully. The set A = {
1
n
| n Z
+
} is an innite set in
(0, 1). The statement that A has no limit point does not have enough words.
The point 0 is clearly a limit point. However, it is not in A. So the denition
on Page 178 could use the extra words in X at the end of the sentence for
clarity. However, they are implied since there is no other space being discussed.
When a subspace is discussed, the extra words become more important.
Sequential compactness adds to the intuition. It is also an extremely useful
concept. In a compact space, if you have a sequence that is trying to converge,
then it almost is succeeding. Just leave out some (perhaps innitely many)
terms and the rest will converge.
The second point of the section is the proof of Theorem 18.2. It is probably
the most complicated proof so far in the book (surpassing the connectedness and
compactness results about subsets of R). Do not just read it for understanding.
See how much you can learn. Learning the proofs of the properties of intervals
in R is a warm up for this one.
3.28.1 Exercises
There are few problems that are truly essential. The important notion of an
isometry is introduced in Exercise 6. The notion of countably compact is prob-
ably less important, but it is probably nice to know the denition. It is in
Exercise 4.
Exercise 7 is important. It drives a lot of machinery in several subjects. It
is sucient to do (a) to get the right avor. If desired (b) can be atempted. We
will come back to (d) after completeness is dened in Section 43.
3.29 Local compactness
We will not do the entire section. One main point of local compactness is
Theorem 29.1 and the denition that follows from it on the top of Page 185.
The proof of Theorem 291. is long but very elementary. It is long because there
are a lot of things to check. However, no check is all that complicated. It is
worthwhile to carefully go over such a proof to see how systematicall it proceeds
through all the necessary checks.
In Example 4, you are told that you may readily check that the one point
compactication of R is homeomorphic to the circle. How will you readily check
this? This is actually exercise 6.
The rest of the exercises can be picked through to see if you nd anything
interesting.
3.29. LOCAL COMPACTNESS 47
3.29.1 Nets
We will not cover nets. However, the rst two sentences on Page 187 are de-
nitely worth reading. Then even if you dont bother to read farther and learn
what nets are, at least you will know what their purpose is. It will also remind
you of an important aspect of metric spaces.
48 CHAPTER 3. CONNECTEDNESS AND COMPACTNESS
Chapter 4
Countability and separation
axioms
The real numbers R play a large role in topology. The usual topology gives a
good example of a topology. The real numbers are the values of a metric. The
real numbers motivate various properties. The last two points are combined to
ask what about a topological space makes it a metric space. If a topological
space is to turn out to be a metric space, then the real numbers have to show
up somehow. The properties of this chapter imitate enough of the real numbers
to make this happen.
4.30 The countability axioms
The three countability axioms most often encountered are rst countable,
second countable, and the badly named (the book points this out) separa-
ble. After these three, the next most often encountered is Lindelof. All of
these are dened in the rst three pages of the section.
The use of rst countable is immediate. It makes sequences as important
as they are in metric spaces. Second countable is harder to motivate. The best
motivation is that the real line is second countable and we are looking for useful
properties of the real line. In a sense, a second countable space cannont be too
big, whatever that means.
Separability needs the notion of dense which is dened on Page 191. Sepa-
rability also keeps a space from being too big.
One can ask what the relations are between the notions. The book gives
examples that show that some properties do not imply others. They are not
crucial to go through. The theorems are more important.
Note that the relations between these properties and standard constructions
such as subspace and product are considered important enought to explore.
Exercises 4, 5, 11, 13 are probably enough for now. Other problems can be
done if they look interesting.
49
50 CHAPTER 4. COUNTABILITY AND SEPARATION AXIOMS
4.31 The separation axioms
The book ends up dening T
1
(often called T
1
), Hausdor (often called T
2
), reg-
ular (often called T
3
), and normal (often called T
4
). (Sometimes the T versions
dier slightly from the non-T versions. Some books include T
1
as a part of T
3
,
for example, since T
1
makes a single point a closed set.) There exist axioms T
0
,
T
5
, T
6
, T
2
1
2
, T
3
1
2
and several more. There are 17 separation axioms in all listed
in Wikipedia. There are then about 17
2
questions about what implies what.
Few are interesting. The relation to subspaces and products is interesting.
Give a false proof that a subspace of a normal space is normal. Then nd
your mistake.
The reason that these are called separation axioms is that they are motivated
by the ability to separate sets by real functions. We say that a continuous
f : X R separates x X from y X if f(x) = f(y) and that if such
an f exists, then x and y are said to be separated by a function. If all pairs
of unequal points can be separated by a function, then Wikipedia calls such a
space completely Hausdor, and it is clear that this implies that the space is
T
2
.
One of the main results in this topic is that a strong version of being sepa-
rated by functions is obtainable without assuming anything about being sepa-
rated by functions. This is quite a remarkable achievement and is proven two
sections later.
The most important result in the section is Lemma 31.1(b). All it does is
give another view of normality. However, another view of something is often
worthwhile even if it is trivial to arrive at.
The consequence of 31.1(b) is that a certain situation divides itself into
two smaller identical situations. The given situation is that there is a closed set
A inside an open set U. The consequence is that we can nd an open V so that
A V V U
and the closed A in open U has been divided into a smaller (or left half) version
of closed A in open V and a larger (or right half) closed V in open U. This
is supposed to remind you of how things get divided up on the real line. For
example
(, 0] (, 1)
gets split into
(, 0] (,
1
2
) (,
1
2
] (, 1).
Exercise 2 is a variation on this. Exercise 3 is worthwhile, and Exercise 5 is
nice if you also explain why Hausdor is needed.
4.32 Normal spaces
Normal spaces are nice. Nice spaces are normal. The section supports this and
the proofs in the section are very important to learn.
4.33. URYSOHNS LEMMA 51
Exercuse 3 can be done if you appreciate Theorem 29.2 and its proof. Exer-
cise 4 is a classic.
4.33 Urysohns lemma
This is a big deal. I dont agree with everything the author says in the rst
two paragraphs of the section, but they are not worth arguing with. The main
fact about Urysohns lemma is that it makes real numbers appear in a situation
where the hypotheses give no hint that real numbers will appear.
I am not crazy of the way that the book writes down the proof. The key
is Lemma 31.1(b) and the comments that I make about it in the notes above.
The book does not make enough of a fuss about this, and never even mentions
that it is using the lemma. It just goes ahead and uses it. We will go over the
proof in detail in class.
The proof is a triumph of notation. Once the notation is agreed on, the
function is very natural to dene. All that remains to check is that it is con-
tinuous. This is the hard part (as the book mentions) and can be made a bit
easier by referring to subbases.
At the top of Page 211 is a discussion of the relationship between normality
and the ability to separate by a function. The important aspect of Urysohns
lemma is in the second paragraph on that page.
The discussion that follows about the ability to separate with functions is
nice enough, but we will skip it.
4.34 The Urysohn metrization theorem
We will skip this section. Urysohn invented his lemma to prove his metrization
theorem. It is a nice theorem and you should be aware that it exists. You
should look at its statement to appreciate how little it assumes. The idea of the
proof is to nd enough continuous functions from X to R so that given any two
points at least one of the fuctions separates them. Then using these functions
we can embed X into a product of copies of R. If there are only countably many
such functions, then we have only countably many copies of R. A product of
countably many copies of R can be given a metric and we are done. So the
diculty is nding a set of functions that is only countable that separates all
the pairs of points. This is where the countable basis comes in.
The Urysohn metrization theorem is not the last word in metrization the-
orems. It is not if and only if. There are such and we will not look at them
either. Metrization theorems used to be a much larger topic than they are now.
4.35 The Tietze extension theorem
This is another application of Urysohns lemma. It is used quite a bit. It is
a beautiful example of a countable process. One might say that the proof of
52 CHAPTER 4. COUNTABILITY AND SEPARATION AXIOMS
Urysohns lemma is a countable process as well, but it does not have quite the
same avor. We will go over the proof of the Tietze extension theorem in class.
Part of the point of the Tietze extension theorem is its connection to the
material discussed in Exercises 4, 5 and 6. This will also be discussed in class
to a certain extent.
4.36 Imbeddings of manifolds
We will not cover this section.
Chapter 5
The Tychono theorem
We will not cover this chapter. The Tychono theorem says that an arbitrary
product of compact spaces is compact. The proof uses the axiom of chioce. This
is not surprising since the Tychono theorem implies the axiom of choice. The
proof of this converse is rather cute. Google it to see if you can nd the proof or
try to nd the proof by yourself. The use of the nite intersection property is
what makes the proof go through. In fact, the proof of the Tychono theorem
from the axiom of choice that is given in the book also uses the nite intersection
property. Those that wish to read the proof are invited to do so. It is not all
that bad.
The section on the Stone-

Cech compactication makes more sophisticated


use of the axiom of choice and should be read by those in love with the axiom
of choice and its consequences.
53
54 CHAPTER 5. THE TYCHONOFF THEOREM
Chapter 6
Metrization theorems and
paracompactness
We will not cover this chapter either. As mentioned before, metrization the-
orems are easy to dismiss these days. Not so with paracompactness. It is a
rather important tool, but we will skip it for lack of time. The denition of
paracompactness (Page 253) is not at all revealing and the proof that all metric
spaces are paracompact is quite famous, quite long, and quite useful. Their use
is summarized in Lemma 41.6 and Theorem 41.7. The result is that certain
spaces admit partitions of unity. This phrase only makes sense after studying
the denition. The uses of partitions of unity will not be covered in this course,
but may show up in other courses.
55
56CHAPTER 6. METRIZATION THEOREMS AND PARACOMPACTNESS
Chapter 7
Complete metric spaces and
function spaces
This chapter covers several topics that are denitely related, but all of math-
ematics is related. The chapter could as easily have been divided into smaller
chapters or had parts separated out and given to other chapters.
7.43 Complete metric spaces
This topic could have been covered in any of several other chapters without
straining the outline. It is a fundamental topic in the subject of metric spaces
and relates as much to compactness as anything else. It is crucial for those
interested in analysis.
Many examples of non-compact spaces can be described by saying that cer-
tain points that should be there are not there. In (0, 1), the points 0 and 1
are missing. They help sequential compactness to fail. The sequence a
i
= 1/i
fails to have a convergent subsequence because the sequence itself deserves to
have a limit and the limit is not there.
Completeness studies sequences that deserve to converge. The metric space
setting makes the study work.
In metric spaces, Cauchy sequences (denition on Page 264) are the ones that
deserve to converge. The sentence after the denition is important. As an easy
exercise, prove that any convergent sequence is Cauchy and so the denition is
not silly.
The denition of completeness (same paragraph as denition of Cauchy)
assumes a metric space since the metric is used in the denition of Cauchy.
So the topic does not exist without a metric. However, two dierent metrics
might lead to the same topology and it is therefor not clear if the actual metric
makes a dierence in whether a space is complete. It does. On Page 264 is an
argument that certain changes of the metric make no dierence. But the book
gives no examples that point out that sometimes it does make a dierence. It
57
58CHAPTER 7. COMPLETE METRIC SPACES AND FUNCTION SPACES
almost does in the discussion of Example 2. This is similar to the example
(0, 1) discussed a couple of paragraphs up in these notes. Think of how you can
use Example 2 to make a space with 2 dierent metrics that induce the same
topology but that have one complete and the other not. This discussion is dealt
with in great detail in Exercise 6 on Page 270. We will discuss that shortly.
The point that closed subsets of complete spaces are complete should have
been made a lemma. The argument is shortly after the denitions on Page 264.
The next several lemmas and theorems develop the theory to culminate in
several important theorems.
Lemma 43.1 has as an immediate consequence that every compact metric
spaces is complete.
Theorem 43.2 points out that Euclidean spaces are complete. This could
have been done by starting with R, and then taking products. So the important
part of Theorem 43.2 is that R is complete. This rounds out a set of useful
observations about R. It is connected, complete and its closed intervals are
compact.
The results leading to Theorem 43.5 are a step by step development leading
to a grand statement about spaces of functions into complete spaces when the
function spaces are viewed as products of complete spaces and the product
is given the uniform metric. The discussion really started in Chapter 2 and
could have been made a lot shorter by going directly to arbitrary products of
arbitrary metric spaces, but perhaps the slow approach of the book helps. Even
the discussion in Chapter 7 could have been shortened.
Theorem 43.6 is an important combination of prior discussions. Not only
are the function spaces complete (which means that sequences of functions that
deserve to converge actually do converge) but if the functions in the sequence
are continuous, then so is the limit. This has practical applications outside of
topology.
The book revisits the uniform metric in the setting of bounded functions and
shows that two things that look the same are the same. This is the discussion
in the second half of Page 268.
Theorem 43.7 is a nice result that you should know, but the proof is not
crucial now. It has a avor of other proofs in the book that we have skipped.
In particular it has the avor of one of the metrization theorems. There is
another proof (Exercise 9 on Page 271) that has a avor of other proofs that we
have not done in the class either. If you have seen the rationals built from the
integers by looking at classes of pairs of numbers, then the outline of Exercise
9 will look familiar. Rationals solve the problem that not all division problems
in the integers have solutions in the integers. So you let the problems be their
own solutions. That is 3/4 is a problem that is coded by the pair (3, 4). We
say that (3, 4) solves the division problem 3/4. But certain problems deserve
the same solutions. For example 6/8 deserves the same solution as 3/4 because
3 8 = 6 4. So we put an equivalence relation on pairs by saying (a, b) (c, d)
if ad = bc. Then we do a lot of work with the equivalence classes and after the
work, we announce that we have invented the rationals. Along the way, addition,
subtraction, multiplication, division, and so forth have to be dened along with
7.44. A SPACE-FILLING CURVE 59
proofs that the operations are well dened. We also announce that we have made
the world better because now all division problems (except the usual problem
of dividing by zero) have solutions. Lastly, one shows that sending n Z to
(n, 1) in the new numbers is an algebraic embedding. In Exercise 9, the same
thing is done with Cauchy sequences. Every Cauchy sequence deserves a limit,
so you let the Cauchy sequence itself become its own limit. Sometimes dierent
sequences deserve the same limit. And the resulting theory resembles (dierent
details, same outline) the development of the rationals from the integers.
Theorem 43.7 and Exercise 10 are of a certain type. They improve something
and do so in such a minimal way that the result has a strong form of uniqueness.
What should be taken from this section is a knowledge of the denitions and
main results as a minimum. This is almost impossible to do correctly without
at least learning some of the easier lemmas. The proof of Thoerem 43.5 is so
easy and so important that it should be learned. The intricacies of the minor
variations of bounded metric, uniform metric, bounded functions are a pain and
are not terribly important ideas. Theorem 3.7 and Exercise 10 are not critical
in detail, but you should know the statements.
7.43.1 Exercises
Exercises 14 are nice technical reviews and can be skipped if you have decided
that you want to remain ignorant of much of this topic.
Exercise 5 will be important later. We will probably go over it in class, but
it is not to hard with the right trick. Start with the uniqueness. That is easy.
There is a lot more information in Exercise 7 of Section 28 that will help with
the existence.
Exercise 6 is important to know as a list of statements even if you never do
the exercises. Parts (a) and (b) are simply reviews of the denition. Parts (c)
and (d) are harder and the book acknowledges this by giving hints. However,
Parts (c) and (d) have consequences that are as important as the details. Part
(c) says that (0, 1) is topologically complete. That means that there is a metric
you can put on (0, 1) that gives the usual topology, but that makes every Cauchy
sequence converge. This is true, but it comes out giving the wrong impression.
The right way to say it is to say that it makes the non-convergent sequences
not be Cauchy any more. This has to be what is happening because if the
topology stays the same and the metric changes, then the convergent sequences
cannot change, but the Cauchy sequences can. This last point is important to
understand. As is pointed out at the end of (d), the irrationals are topologically
complete. This points out the power of changing the metric.
7.44 A space-lling curve
It is worth knowing that this exists. Study the statement of Theorem 44.1.
There is little to study. This theorem led to a deep study of dimension as a
topological topic. This theorem made it clear that dimension is not preserved
60CHAPTER 7. COMPLETE METRIC SPACES AND FUNCTION SPACES
by continuous surjections. Dimension theory is given a nice introduction in
Section 50 of the book. We will not get to it, but most of it is readable at
this time. The proof of the main result, Theorem 50.5, as given in the book
needs more material that comes just before Section 50, but the statement can
be appreciated now.
The most important sentence in the proof of Theorem 44.1 is the rst sen-
tence. Everything after that is detail. If you want to know how it is done, read
the book and look at the pictures. We will probably not have time to go over
this in class.
7.45 Compactness in metric spaces
You know that in R
n
with the usual metric the compact subsets are the ones
that are closed and bounded. Hopefully you also know that this does not apply
to all metric spaces. In this section, you get the right phrase that characterizes
compact subsets of arbitrary metric spaces. This comes in Thoerem 45.1 where
it is shown that complete and totally bounded is equivalent to compact in a
metric space.
The important aspect of this result is that it can be applied. Standard
examples of metric spaces where compactness can be elusive are function spaces.
We will try to get to applications of this section to function spaces if time allows.

Você também pode gostar