Documente Academic
Documente Profesional
Documente Cultură
Editors
F. W. Gehring
P.R. Halmos
C.C. Moore
Robert B. Reisel
Elementary Theory of
Metric Spaces
A Course in Constructing
Mathematical Proofs
Springer-Verlag
New York Heidelberg
Berlin
Robert B. Reisel
Loyola University of Chicago
Department of Mathematical Sciences
Chicago, Illinois 60626
U.S.A.
Editorial Board
F. W. Gehring
University of Michigan
Department of Mathematics
Ann Arbor, Michigan 48104
U.S.A.
P.R. Halmos
Indiana University
Department of Mathematics
Bloomington, Indiana 47401
U.S.A.
C.C. Moore
University of California at Berkeley
Department of Mathematics
Berkeley, California 94720
9 8 7 6 5 432 1
ISBN-13: 978-0-387-90706-2
e-ISBN-13: 978-1-4613-8188-4
DOT: 10.1007/978-1-4613-8188-4
Preface
In science
In other
mathematicians.
They must,
of course, start with easy proofs and build to more complicated ones.
This is usually done in courses, like abstract algebra or real
analysis, in the junior and senior years of college, but this is
almost too late in the curriculum.
a text for an earlier course that would emphasize proofs while teaching useful and important mathematics.
As I said, students learn to construct proofs by actually working
out proofs.
Instead
the definitions and some indication of what the ideas mean so that
the book is self-contained.
of selected results.
~ight
cause
the appendix until they have written their own proofs, or at least
tried to do so.
essentials of the theory of metric spaces and also do all the proofs,
the course could be accelerated by having the students read the
proofs in the appendix and only try to construct proofs for the other
results.
I think that the best way to use this book is in a seminar; I
give some suggestions for this below.
Neverthe-
less, the book develops all the ideas of metric spaces that are
needed in a course in real analysis of functions of a single variable.
The book also gives the students experience with mathematics in an
axiomatic setting.
The
first three chapters cover the essential material that everyone should
know in order to understand modern analysis, namely, the elements of
set theory, particularly the concept of a mapping, the basic geometrical ideas of metric spaces, and the properties of continuous mappings.
The remaining three chapters take up sequences and completeness,
connectedness, and compactness.
involve general metric spaces, they are slanted toward the space of
real numbers, which is the most important elementary application of
these ideas and the one that is most familiar to the students.
I have
not included any discussion of product spaces, which would allow these
results to be extended to higher dimensional spaces, because I think
that product spaces are studied more effectively in the context of a
topology course.
more difficult proofs and would extend the book beyond what could be
reasonably covered in a single course.
Mathematical
the seminar and the course meets two hours a week for a semester.
Assignments are made so that each student's report takes about a half
hour.
The students are told to write out their IIlork completely, not
At the start
should be made to correct them in the class with the other students
making suggestions.
It
might not be possible to cover the entire book, so I have made some
suggestions in the various chapters as to what might be omitted.
I want to mention a few technical points about the book.
Results
Within a chapter
A reference
definitions are expressed'by using the word "if" rather than the
phrase "if and only if," although the latter is actually meant.
have followed this convention.
Finally, the
I would like
to list the names of all my colleagues who have used forms of this
book over the years, but I am sure that I would forget to include some
of those who left Loyola University years ago.
CONTENTS
Preface.
14
34
52
59
Chapter V. Connectedness.
69
75
Afterword.
85
89
Appendix S. Solutions.
91
Index.
119
Chapter 0:
Some Ideas of Logic
We use
There are a
I will give you an opportunity to use what you have learned by asking
you to construct some
si~ple
proofs.
combine sentences.
In this
To avoid misunderstandings, I
Therefore, in
important, only its "truth value" counts, that is, we speak only of
its truth or falsity.
One way to modify a statement is to negate it.
If "pl1 is a
This can be
not-P
In such a table
each line shows the connection between the truth values of "p" and
"not-P".
For example, the first line asserts that when "p" is true,
"not-P" is false.
P and Q
P or Q
Note that the statement "P or Q" is defined to be true if one or the
other or both of the statements "P" or "Q" are true.
This is not
Sometimes
"or" is used like this, but at other times "or" is used in the sense
that "p .or Q" is true when precisely one of the "p" or "Q" is true,
but not when they are both true.
In mathematics "or" is
P -.. Q
In ordinary
speech we do not say "If P, then Q" when we kn.ow that "P" is false,
so there is really no conflict between this strict logical usage and
everyday speech.
is a true statement.
Experience has shown that the above truth table is the most useful way
of defining a statement of the form "If P, then Q".
-+
This can
only happen in the first line of the truth table, so you can conclude
that "Q" is also true.
if "P
-+
Q" and "Po are both true, then "Q" is also true.
+-+
+-+
+-+
Q" is true when "po and "Q" have the same truth value
+-+
-+
Q) and (Q
It is easy to see
P}".
-+
To show this
(P
Q) and (Q
-+
-+
-+
-+
P)
+-+
Q".
Since we are
only concerned with the truth values of a statement, we can say that
"(P
-+
Q) and (Q
-+
+-+
(Two
are identical.)
"P
++
->-
->-
-+
Q".
->-
Therefore, in the proof you assume that "P" is true and then
form:
Given:
Prove:
Proof:
The "Proof" will be a sequence of steps that lead from "P"
(and
->-
which is "Q
->
positive of "P
P".)
->-
->-
->-
Q",
->-
Q".
not-Q
not-p
not-Q -+ not-P
(Compare the last column with the truth table for "P
->
Q".)
Therefore,
if you have to prove a result of the form "P -+ Qn and you want to do
so indirectly, you would write:
Given:
not-Q
Prove:
not-P
Then you will try to find a sequence of steps that lead from "not-Q"
(and previously known results) to "not-P".
A second kind of indirect proof is "proof by contradiction".
Here to prove "p .,. Q" you assume that this statement is false and try
to derive a contradiction.
Prove:
a contradiction
Then you will try to find a sequence of steps that lead from "p" and
"not-Q" to a contradiction.
what the contradiction will be, so you are working a little blindly.
However, you are starting with two statements in the "given", so you
have more to work with.
is the following.
statement, this means that you must have the fourth line of the truth
table for the statement "R".
("P ... Q" is false) is actually false, so "p .,. Q" is true, as you wanted
to prove.
One technical point should be observed here.
In the indirect
Many
times you will have a universal statement, that is, one that says that
all the objects of a class have a certain property, for example, "All
differentiable functions are continuous."
statement says that there is an object of the class that does not have
the property.
Similarly, the
negation of a particular statement, that is, one that says that some
object of the class has a property, is a universal statement, saying
that no object of the class has that property.
Sometimes it is diffi-
cult to decide exactly how to write such negations, but usually common
sense will suffice.
You should show your work to your instructor so that he can correct
any errors you might have made and comment on your style.
WOP~ED
1.
Prove:
B iff A u B
B.
First prove
that an element of the first set is in the second and vice versa, so
this part of the proof will itself consist of two parts.
In the first
of these two parts you have to prove "x s A u B .,. x s B" and in the
second, "x s B .,. x s A u B".
(i)
Given:
x s A u B
Prove:
Proof:
1.
A u B
2.
A or x
3.
B
(Given)
E
(Definition of union)
(Given)
4.
5.
x E B
(Definition of
~)
(Steps 2 and 4)
A c
Prove:
A u B
Proof:
l.
(Given)
2.
A u B
(Definition of union)
Given:
B".
To prove "A
B".
A u B
Prove:
Proof:
l.
(Given)
2.
A u B
(Definition of union)
3.
A u B
4.
(Given)
B
[]
(Definition of equality)
co~~on
error is to
merely repeat the statement of the problem in the "Given" and the
"Prove".
For example, in the last part of the above proof, you might
write:
Given:
A u B
Prove:
B.
2.
Prove:
(B u C)
(A 0 B)
(A 0 C)
The proof will consist of two parts because you have to prove the
equality of sets.
(i)
A n (B u C)
Given:
Prove:
Proof:
l.
2.
A and x
3.
A and (x
(A
B) u (A
C)
(B u C)
(Given)
(Definition of
B u C
B or x
0)
C)
(Definition of u)
(x
4.
A and x
B) or (x
A and x
C)
It says that
You should
think about this long enough to see that this is also what Step 4
says.
This can also be done more formally by noting that Step 3 has
the logical form "P and (Q or R)" and Step 4 has the logical form
"(P and Q) or (P and R)".
5.
(x
B) or (x
C)
(Definition of 0)
6.
(ii)
Given:
Prove:
(A
A
(A 0
B)
B) u (A
(A 0 C)
(Definition of u)
C)
(B u C)
Proof:
The proof of this part just consists of the steps of the proof of
part (i)
3.
in reverse order.
Prove:
(i)
Given:
B
A
iff
~
X E
Prove:
B
cB
x E cA
cA
cB
[]
10
Proof:
l.
cB
(Given)
2.
t-
3.
4.
5.
x t
6.
x t A
7.
(Given)
-7
(Definition of =.)
-7
t A
(Detachment, Steps 2, 5)
(Definition of complement)
cA
(Contrapositive of 4)
This is not an
Comment.
Given:
cA
cB
::>
Prove:
Proof:
l.
(Given)
2.
t-
cA
(Definition of complement)
3.
cA
4.
cB
-7
cA
(Definition of
5.
t-
cA
-7
t-
cB
(Contrapositive of 4)
6.
i-
cB
7.
::>
(Given)
cB
~)
(Detachment, Steps 2, 5)
[J
(Definition of complement)
writing the second part of the proof by referring to the first part
in a certain way.
could be done.
::>
cB.
result about the double complement, you will get the statement that is
proved in the second part of the above proof.
11
4.
Prove:
c(A u B)
cA n cB
and
c(A n B)
cA u cB
way that logic tells us how to negate "or" and "and" statements.
The
If
you are not familiar with these rules, you can easily prove them by
writing the appropriate truth tables.
(i)
(ii)
Given:
<:
c(A u B)
Prove:
cA n cB
Proof:
l.
c(A u B)
(Given)
2.
A u B
(Definition of complement)
3.
not-(x
4.
A and x
5.
cA and x
6.
cA n cB
Given:
x E cA n cB
Prove:
x E c(A u B)
A or x
<:
B
E
cB
B)
(Definition of union)
(See above)
(Definition of complement)
(Def. of intersection)
The proof here just reverses all the steps of the preceding proof.
You should write it out yourself.
Another way of
and "B" by "cB" and then manipulate the result to get the second
equation.
[1
12
5.
you start with an element of the supposed subset and try to prove
that it is in the larger set, that is, if x
~,
then x
A.
I will
A,
~.
Given:
Prove:
Proof:
1.
(Definition of
[J
~)
Note that the proof is extremely short because of the use of the
contrapositive.
it to a triviality.
6.
is false, that is, that there is more than one empty set, and derive
a contradiction.
Given:
~'
is an empty set
is an empty set
Prove:
a contradiction
Proof:
l.
~'
(Exercise 5 with
~'
as the set A)
~'
(Definition of equality)
~'
(Given)
3.
4.
5.
contradiction
r!
as the set A)
~'
2.
(Steps 3 and 4)
[J
13
Good luck!
Chapter I:
Sets and Mappings
In this chapter you will learn some of the basic ideas of set
theory, particularly those associated with mappings or functions.
These will be used extensively in the following chapters and, indeed,
in all of mathematics.
theory.
and ordinal numbers, and foundational problems, that are not directly
needed in this book.
the use of sets, so I do not give many examples of the familiar terms.
Of course, you should not look in other books to find proofs of the
theorems and the exercises of this chapter, because that would defeat
the purpose of the book.
sion of the topics taken up here, you can look at books in set theory.
1.
set U and that all the sets that will be considered are composed of
elements of u.
is denoted by "a
A".
they have the same elements, that is, for every element x, x is an
element of A iff x is
a~
element of B.
15
as follows:
(Ilx) (x
("\i"
A ++
B).
is the logical symbol for "for every" and ":er" for "there exists".
B" or "B
A set A is a subset of
:J
true:
(\ix) (x E A ->- X E B),
Band B
It follows that A
A.
I p(x)}" will
denote the set of all the elements of U for which p(x) is true when
the element is substituted for x.
p(x)}" or "{x
If A is a set, "{x
x E A
and
The
{x
E A
A n B
{x
A and x
B},
A - B
{x
A and x
B},
cA
U - A
{x
or x
B},
x i A},
(In case it is necessary to
one empty set and this will justify the use of the definite article
"the" when referring to the empty set.
16
EXERCISES.
In the following problems you are asked to prove only a few of
the many properties of sets and their operations.
you enough experience so that you will be able to prove any other
properties that might come up later.
1.
Prove:
(Hint.
B.
A c B iff A u B
that A u B
B.
First assume A
B and prove
Prove:
(Hint.
A n (B u C) =
B.)
(A n B) u (A n C).
Prove:
4.
Prove:
c(A u B)
5.
(Hint.
cB.
(cA)
n (cB)
and c(A n B)
(cA)
(cB).
(Hint.
2.
B iff cA
statement.
6.
Assume that there are two empty sets and use Exercise 5.)
17
equal iff a
c and b
d.)
numbers, then RxR is just the set of points in the plane of analytic
geometry.
(I will always use the letter "R" as the symbol for the
is a system consist-
There is no
deep reason why I am using Greek letters for the indexing set and for
the index, but it helps in avoiding confusion with the sets of the
family. )
y s r)".
If f
{I, 2, .. , n} or if r
=~,
family need not be distinct, that is, it is possible that for two
different elements a and S in f, the sets Aa and AS might be equal.
(See the COIT@ent below.)
If (Ay
ysr
If
= {I,
2,
{x
{x
A2
An"
18
Comment.
YEf
iff
(:;ry) (x
A )
iff
Ny) (x
A ).
y
= {l, 2, 3}
lYE r),
lYE r} is the
By the defini-
It is always possible
example, the set of sets {A,B}, you could take r to be {1,2}, or any
two-element set, and let Al
A and A2
B.
this set of sets and the union of the family are the same, that is,
A u B
EXERCISES.
1.*
Prove:
(AxB) n (CxD)
2.
Prove:
To do this,
assume that [(P and Q) or (R and S)] is true and prove that the other
expression must be true.
19
Prove:
If A
4.*
Prove:
c( u A )
YEf Y
A' and B
(cA ) and c( n
Y
Prove:
5.
yEf
Comment.
YEf
A'xB'.
A)
Y
(cA ).
yEf
(B n A ).
y
To show
{I}, B
{2}, C
{3}, and D
Such an example
is, a set with only one element, and all the sets are different.
Then AxB
{(1,2),
{2,4}, so (A u C)x(B u D)
{(1,2),
(1,4),
(3,2),
{I,3}
(3,4)}
MAPPINGS.
Let X and Y be nonempty sets, possibly identical.
A mapping
what the sets X and Y are, then this mapping may be denoted just by
20
The
If f : X - Y is a
The symbol
Note
that if A f X, flA and f are not equal because their domains are
different.
Let f:X_Y be a mapping, A be a subset of X, and B be a subset
of Y.
{y E Y
(:;[x) (x
A and y
f (x))},
It is customary
21
Since these ideas are probably new to you, I will give some
Let f:R~R be the mapping given by f(x) = x 2 for every
examples.
x in R.
If A'
f(A)
{4, 9}.
If
B = {-5, 1, 4}, then f+(B) = {-I, 1, -2, 2}, that is, it is the set
of all real numbers whose square is -5, 1, or 4.
If B'
Comment.
with the direct image and working with the inverse image.
If you are
construct such an x.
Study
these proofs before you try to prove the remaining parts of these
theorems.
THEOREM 3.1.
f(Yl)
(ii)
f(X)
(iii)
(A
(iv)
f ( u Ay)
ysf
(v)
f ( n Ay)
ysf
(i)
Yl
c
ysf
n
ysf
f(A )
y
f (Ay ).
22
equality does not necessarily hold in Theorem 3.1 (v); in this example
you may take f = {l,2}.
THEOREM 3.2.
Let
f:X~Y
be a mapping.
f+On
fO
(ii)
f+(Y)
(iii)
..,.
(B -=- B ')
(iv)
{.- ( u
YEf
By)
(v)
(vi)
EXERCISE 2.
*
Construct a counterexample of a
a property analogous to Theorem 3.2 (vi) does not hold for direct
images.
Proof of Theorem 3.1 (iv) .
(i)
Given:
y E f ( u Ay)
YEf
Prove:
y E
Proof:
l.
y E f ( u Ay)
Ysf
2.
(ax) (x E
3.
f (A )
YEf
'I
YEf
(Given)
A
and y
f (x))
(DeL of image)
(ax) (ay) (x E A
'I
and y
f (x))
(Def. of union)
23
4.
5.
(<IY) (y E flAy)
6.
YEr
(ii) Given:
Prove:
y E
YEr
(DeL of image)
f (A )
Y
(Def. of union)
flAy)
y E f( U
Ay)
yEf
Proof:
Comment.
[]
There is
a statement when values from some universal set are substituted for
x and y.
The statements "(<Ix) (Vy) P (x, y)" and "(VY) (<Ix) P (x, y)" do
"works" for all the y, but the second allows each y to have a different x.
It should then be clear that you can go from the first to the
necessarily true.
(i)
In symbols,
Given:
Prove:
x E
f+ (
YEr
U
yEf
By)
f+(B)
24
Proof:
l.
x E f+(
2.
f(x)
3.
(Def. of union)
4.
(ay) (x E f+(B y ) )
5.
x E
YEr
u
yEy
x E
YEf
Comment.
(Given)
B
y
f+(B
YEr
(ii) Given:
By)
(Def. of union)
f+(B y )
Proof:
x E f+(
Proof:
YEr
By)
[]
prove the remaining parts of Theorem 3.1 and Theorem 3.2 and do the
two exercises.
THEOREM 3.3.
(i)*
EXERCISE 3.
Let
f:X~Y
f(f+(B
be a mapping.
Then
f:X~Y
gf:X~Z
and
g:Y~Z
be
25
comment.
"g
mappings.
mappings, so I use the simpler notation "gf" for the composite mapping.
Note that in order to construct a composite mapping gf, the codemain
of f must equal the domain of g.
Let
and
f:X~Y
be mappings.
g:Y~Z
If C is a
f+(g+(C)).
APPENDIX
example.
EXAMPLE.
Let
(See Figure 1.
= x2 .
It is not necessary to
always have the zeroes lined up or to use the same scale on both
lines. )
-2
-2
-1
-1
Figure 1
At each point on the lower line, say at the point x, draw an arrow
whose origin or tail is at that point and whose head is on the upper
line at the point x 2 .
2 to 4, from -1 to 1, etc'.
26
Picture the
domain as some set, like a line, and the codomain as another set above
the first.
If, for
example, the mapping had the real plane as the domain and the real
line as the codomain, the picture would look like Figure 2.
Figure 2
Usually only a few typical arrows are drawn, but if the domain
and codomain are small finite sets, all the arrows can perhaps be
drawn.
EXAMPLE.
Let X
{a,b,c,d} and Y
{1,2,3}.
Suppose
ftc) = 2
f (d)
f:x~Y
is defined by
f(a) = I ,
f(b) = 2,
1.
27
/~.
abc
Figure 3
The various concepts of this section and later ones can be nicely
illustrated by this process.
Let
be a mapping and A
f:x~Y
f:R~R,
where
[0,4]
-2
-2
, [ t
-1
XI 1/.
-1
Figure 4
To picture the inverse image f~(B) start by taking B as a subset of Y, the upper set.
[1,4].
:::?11L
-2
-1
-2
-1
Figure 5
[-2,-1]
[1,2].
28
f:X~Y
and
g:Y~Z
are mappings.
The mapping f
and are obtained by combining the arrows of f and g, that is, start
with a point x in X, draw the arrow from x to f(x) in Y and then the
arrow from f(x) to g(f(x
in Z.
from x to g(f(x.
g(f(x
Figure 6
pictures when you are trying to construct a proof or when you are
presenting your results.
proof, they can suggest what should be done and they can make proofs
more understandable.
29
4.
f:X~Y
a surjection if f(X)
its codomain.
To prove that
f:X~Y
Let
f:~~~,
by f(x) = x + 1.
(y - 1) + 1
y.
Be
g:~~~
the range is the set of even integers and the codomain is the set of
all integers.
The solution is
Of course, to
prove that a mapping is not surjective, you just have to produce one
element of the codomain that is not in the range.
EXERCISE 1.
Let
f:X~Y
be a mapping.
Then
surjective.
(Hint.
30
If
f:X~Y,
and
g:Y~Z
h:Y~Z
hf, then g
are mappings
h.
The mappings g and h are given as having the same domain and
EXERCISE 3.
ix:X~X
If it is
clear what set is being considered, the symbol "i" may be used in
place of "i x ".
THEOREM 4.3.
If
f:X~Y
THEOREM 4.4.
If
f:X~Y
and
g:Y~X
f.
f:X~Y
(f(x) = f(x'}}
to the contrapositive:
(x
(x = x').
i x')
(f(x)
i f(x'.
31
Examples.
by g(x)
Let
2x.
that is 2x
g:~~~,
where
It follows that x
x', so g is injective.
g(x'),
The
To
prove that a mapping is not injective, it suffices to find two different elements of the domain which have the same image, like the 2
and -2 above.
EXERCISE 4.
pendent concepts, that is, a mapping can be one without being the
other.
(iii) both surjective and injective, and (iv) neither surjective nor
injective.
THEOREM 4.5.
(i)
Let
f:X~Y
f+(f(A
be a mapping.
Then
injective;
(ii)*
n f(A')
of X iff f is injective;
(iii)
= f(cA)
for
in the case where the subsets are singletons, that is, sets with
one element.)
THEOREM 4.6.
If
g:X~Y,
h:X~Y
and
f:Y~Z
EXERCISE 6.
If
f:X~Y
and
g:Y~X
g:Y~Z
be mappings.
Let
f:X~Y
and
Then
(i)
(ii)
eA:A~X
eA:A~X
is injective.
f:X~Y
If
If f:x~Y is a bijection, then a mapping f-l:y~X can be defined by the rule that f-l(y)
x iff f(x)
y.
f:X~Y.
mapping.
(Hint.
Point
The mapping
f:~~~,
seen to be a bijection.
defined by f(x)
x + 1 is easily
x + 1.
To find
the inverse mapping f-l:f~~, you have to solve this equation for
33
x, that is, x = Y - 1.
Then f-l(y) = Y - 1.
domain and the codomain are the same set, the inverse is often
written with the letter "x", that is, f-l(x) = x - 1.
Such usage
ordinarily does not cause confusion, but if it does for you, don't
use it.
If f:X~Y is a bijection, then f-lf
THEOREM 5.2.*
ff- l = i y .
THEORW15.3.
If f:X_Y and
gf = iX and fg = i y
(Hint.
COROLLARY 5.4.
Comment.
g:Y~X
is a bijection.
gf
iy
iX and fg
g:Y~X
for which
g should be.
THEOREM 5.5.
If
f:X~Y
and
g:Y~Z
Chapter II:
Metric Spaces
In this chapter you will begin the study of metric spaces, the
main topic of this book.
If you do so,
ping
d:XxX~R
(Ml)
(M2)
d (x,y)
iff x = y,
35
(M3)
d(x,y)
(M4)
d(x,y) + d(y,z)
Y and d
d'.
This is
called the usual metric on R and the phrase "the metric space R" will
always refer to this metric space.
EXERCISE 1.
Properties (Ml) ,
(M2) and
(M3) are obviously true and (M4) follows from the theorem that the
sum of the lengths of two sides of a triangle is greater than or equal
to the length of the third side.
It is
36
graphs.
x21,
are the same, the metrics are different, so (R 2 ,d) and (R 2 ,d') are
different metric spaces.
Prove that (R 2 ,d') is a metric space.
EXERCISE 2.*
>
(Hint.
Prove
max [a + c, b + dl.)
0, if x = y
d(x,y)
1, if x 'I y.
had much experience with least upper bounds of sets of real numbers.
I have given a short explanation of least upper bounds and greatest
lower bounds in an appendix to this section.
show that the concept of a metric space has more extensive application
than you might have suspected.)
37
Let X be the set of all bounded real valued functions defined on the
closed interval I = tx E R
10
x ~ l}.
EO
THEOREM 1.1.
for every x in X.
or an upper bound.
has no upper bound, but any real number less than or equal to 1 is a
lower bound.
The set X has no lower bound iff for every real number
The infinum of X is
38
X, then t
s.
Similarly, a least
More precisely,
v.
It is
Once you have proved that it has a lower bound, you can apply
for suprema.
EXERCISE 5.
State and
positive real number r, the open ball, closed ball, and sphere with
center p and radius r are, respectively, the sets
B(p;r) = {x
B*(p;r) = {x
E
E
d(p,x) < d,
39
S(p;r) = {x
d(p,x) = r}.
In situations where there are several metric spaces under consideration, the above symbols are modified by indicating the metric space
by a subscript, as in "Bx(p;r)".
EXAMPLE.
a ~
I a
x ~ b}.
< x
< b}
The paren-
thesis indicates that the number is not included in the interval and
the bracket indicates that it is included.
{x
R l a < x ~ b}.)
p - rand p + r.
EXERCISE 1.
Draw pictures of
B(p;r').
>
O.
there exist two open balls, one with center at p and the other at q,
that are disjoint, that is, their intersection is empty.
(Hint.
40
3.
OPEN SETS.
A set G in the metric space X is called an open set if for every
point p of G there exists a positive real number r such that the open
ball B(p;r) is contained in G.
The empty set and the whole space X are open sets.
THEOREM 3.2.*
(Hint.
If x
(Example 1.2).
B(p;r).
Of
open and that a singleton, that is, a one point set, is not open.
EXERCISE 2.
THEOREM 3.3.
YEr
(ii)
if
EXERCISE 3.
Then
is an open set,
is finite,
Gy is an open set.
y~r
B(p;r p ) contained in G.
Give examples of open sets which are not open balls for
and analysis.
ideas are defined in terms of open sets or that there are theorems
which give necessary and sufficient conditions for these new ideas in
terms of open sets.
4.
CLOSED SETS.
Let A be a subset of the metric space X.
A point p of X is called
(B(p;r) - {p})
nAt~.
numbers.
A.
To see this take any open ball with center at 0, say, B(O;r).
A
An open
42
(-3, -1).
is open.
THEOR&~
4.2.
The empty set and the whole space X are closed sets.
THEOREM 4.3.
a metric space.
(i)
YEf
(ii)
(Hint.
Then
Fy
is a closed set,
if r is finite,
YEf
F y is a closed set.
EXERCISE 2.
THEOREM 4.5.*
space X, then for any positive real number r, B(p;r) contains infinitely many distinct points of A.
(Hint.
43
COROLLARY 4.6.
In
CLOSURE OF A SET.
The closure of a subset A of a metric space X is defined to be
The closure
THEOREM 5.1.
In particular, ~
Comment.
~ and X = x.
Of course, A
A,
A point p is an element of
(i)
(ii)
(iii)
(iv)
A iff
A.
~.
A.
B, then
Ac
Then
B;
A is closed;
if F is a closed subset of X such that A
F, then A
F;
(v)
A u B = A u B and A n B
(Hint.
:::>
A n B.
To do this
(v).
44
Comment.
A is
A is
the smallest
A.
(iii) gives
A.
Ac
A.
Therefore, A' =
A.
upper bound.
An unbounded set is
bounded set, then for any positive real number K, however large, there
exist points p and q in A such that d(p,q) > K.
EXAMPLES.
(O,l)}.
I will
show that there exist points p and q in (0,1) for which d(p,q) > r.
This will prove that no number less than 1 is an upper bound for
{d(x,y)
45
(1 + 5)/21 = [-s[=
<
The
To prove this,
I (K
+ 1) - 11 = K.
THEOREM 6.1.
Then
(i)
(ii)
B implies alA)
(iii) 6(B(p;r))
(Hint.
a(B);
In (i) do not forget the case where the sets are unbounded.
6.2.
c(A) + a(B).
(Hint.
bounded.
6.3.
space is bounded.
46
(Hint.
The
For example, if p
Bx(p;r)
The latter
point of Y, By(p;r)
B~(p;r)
Y.
47
(See Exercise 1,
be taken for closed sets and for all other concepts of metric spaces.
THEOREM 7.2.*
EXERCISE 1.
7.3.
Then every
Y.
Y.
48
8.
INTERIOR OF A SET.
Let A be a subset of the metric space X.
A point p of A is
A.
Of course,
A.
EXERCISE 1.
Comment.
(i)
A=
(ii)
AO = c(Clx(cA)).
c(Intx(cA)),
the basis of the definition, Theorem 8.1 gives a way of proving these
properties by using the known results about closures.
I will give an
example of this process and I suggest you use it to prove the theorems
of this section.
EXAMPLE.
that AO is open.
(ii) below.)
I will prove
In particular,
THEOREM 8.3.
~o
and XO
X.
Then
(i)
if A
(ii)
AO is open;
(iii)
(iv)
B, then AO
A.
contained in A;
BO;
G, then AO
G;
49
(A n B)
AO n BO
(v)
EXERCISE 2.
and AO u BO c
(A u B) o.
(v).
BOUNDARY OF A SET.
Let A be a subset of the metric space X.
A point p of X is
~ ~
~ ~,
n.
EXERCISE 1.
set {a,b}.
What is the
Then Bd(A)
n (cA).
COROLLARY 9.2.
(i)
(ii)
Bd (A)
(iii)
Bd (A)
(iv)
Bd (A 0)
(Hint.
Then
Bd (cA) ;
Bd (A) ;
c
c
Bd (A) .
EXERCISE 2.
(iv).
50
THEOREM 9.3.
(i)*
(ii)
A.
THEOREM 9.4.
~;
(i)*
AO = A - Bd(A);
(ii)
A=
Comment.
Then
Then
A u Bd(A).
other results, but they should give you a "feeling" for open and
closed sets, interiors and closures.
Theorem 9.3
(ii) shows
Simi-
by discarding from the set all its boundary points, while (ii)
shows
DENSE SETS.
Let A be a subset of the metric space X.
in X if
A = x.
A is said to be dense
A~
X, since X is
A.
A.
numbers is dense in R.
51
THEOREM 10.2.
(cAlC =
~.
AFTERWORD.
In this chapter you have studied many of the elementary "geo-
chapter you will notice that I often express concepts, theorems and
proofs in terms of open sets.
If you
ever take a topology course, you will see that many of these ideas
and proofs carryover to topological spaces without change.
Chapter III:
Mappings of Metric Spaces
It is
mappings between these objects, but not just any mappings, rather
those that are closely related to the structure being examined.
the theory of metric spaces these are the continuous mappings.
In
In
this chapter you will learn precisely what such mappings are and what
are their elementary properties.
As in Chapter II not everything in this chapter must be covered.
I would suggest that you omit Section 2 if time is short or if you
did not study Sections 8 and 10 of Chapter II.
53
1.
CONTINUOUS MAPPINGS.
Let (X,d) and (Y,d') be metric spaces and f be a mapping from
X to Y.
< E.
A mapping
f:X~Y
which is
The mapping
f:X~Y
EXERCISE 1.
space.
EXERCISE 2.
f:X~Y
is continuous.
f:X~Y
be a constant
Prove that
f is continuous.
EXERCISE 3.
space X.
Let
ix:X~X
EXERCISE 4.
Let
f:R~R
f(x) = 1, if x > O.
<
0, and
54
EXERCISE 5.
metric space Y.
It gives a characteriza-
(See
1.3.
metric space Y.
THEOREM 1.4.
metric space Y.
f (A)
(Hint.
f (A).
>
O.
To
55
THEOREM 1.5.
Then
(i)
flA = fe A ;
(ii)
LE!1MA 2.3.
A)
By(f(xo);c).
56
EXERCISE.
f:R~R
to show that in
THEO~~
(Hint.
glA, then f
g.
Use
Theorem II 2.2.)
Comment;
For
ous function f(x) = 2 x for all real numbers x, that is, there can be
only one way to define 2 x for real numbers in such a way that the
function is continuous.
Be careful!
such a continuous real function; I have only said that there can be
no more than one.
In the case
UNIFORM CONTINUITY.
Let (X,d) and (Y,d') be metric spaces and f be a mapping from
X to Y.
for every positive real number E there exists a positive real number
such that
h~en
<
E.
f:X~Y
is continuous
at a point x o ' the 6 that you must find usually depends on both
and
57
xo.
alone.
EXERCISE 1.*
of R.
f:X~R
f:X~Y
by f(x) = l/x.
is continuous.
Prove that f is
and
is uniformly continuous.
gf:X~Z
Comment.
f:X~Y
I mentioned at the
THEOREM 3.3.
continuous.
The last theorem leads to a generalization.
are metric spaces and
f:X~Y
58
THEOREJ>.l 3.4.
(i)
(ii)
f is uniformly continuous.
Comment.
An isometry
f:X~Y
Then
between the points of X and the points of Y in such a way that the
distance between a pair of points of X equals the distance between
the corresponding points of Y.
and Yare "essentially" the same, that is, as metric spaces they have
the same properties.
for example, have some algebraic property that the other does not
have, but any property that can be phrased in terms of metric concepts that holds for one space holds for the other also.
Usually we
regard two isometric spaces as being the same and do not distinguish
between them.
EXERCISE 2.
ix:X~Y
Prove that
Chapter IV:
Sequences in Metric Spaces
This
is followed by a look at how sequences are related to metric properties, like closure or continuity.
The material
60
1.
SEQUENCES.
A sequence in a set X is a mapping from the set N of positive
integers to X.
a:N~X
Thus, a sequence is
The sequence
"a:~~x".
bility for misunderstanding, this notation is used by most mathematicians, and it will probably cause you no trouble after a little
while.
Different sequences are denoted by different letters, for
example, by "(an)" or by "(x n )".
equal as mappings, that is,
(an) =
It is
For example,
r! (b n )
...
61
a sequence, if you presume that you know what occurs where the dots
are written.
A sequence (b n ) is a subsequence of the sequence (an) if there
exists an increasing mapping
:~~~
~,
Roughly
In fact, b l
a(l)
a2
2, b 2
a(2)
a4
2, etc.
(n)
)".
one the increasing mapping is denoted by "i" and the subscript notation
is used, while in the second the mapping notation is used.
mapping from
to
Since a
with the sequence notation, but I have found that it usually causes
confusion among students.
notation.
62
1.1.*
L~~
Let
:~~~
be an increasing mapping.
>
Then for
n.
In such a process
the first term, aI' is defined by some rule and then it is explained
how to define a n +l when the preceding terms, aI' a 2 , ... , an are
assumed to be known.
You
defined by recursion.
Assume that a l
If
Try to be as explicit as
(Hint.
integers has the property that every nonempty subset has a smallest
element.)
preceding proof, but you must show that, in principle, the recursion
63
A point p
or to be divergent.
EXERCISE 1.
p for every
P and b
THEOREM 2.1.
THEOREM 2.2.
(x n ), then
EXERCISE 3.
If
(x~(n
(x~(n
convergent subsequence.
THEOREM 2.3.*
Then a point
p.
n, xn E B(p;l/n).)
COROLLARY 2.4.
If
(x n ) is a
If
f:X~Y
be a mapping
(x n ) is a sequence of
CLUSTER POINTS OF
SEQUENCE.
every open ball with center at p there are infinitely many values of
the subscript n such that xn is in that ball.
65
THEOREM 3.1.
of the sequence.
EXERCISE 1.
(x~(n
of the sequence
that converges to p.
(Hint.
x~(n)
COROLLARY 3.3.
it is divergent.
EXERCISE 2.
If p is an accumulation point of A,
finit~
range,
f:X~Y
4.
CAUCHY SEQUENCES.
A sequence (x n ) of points of a metric space X is a Cauchy
the sequence after some index value are close to each other.
THEOREM 4.1.
sequence.
EXERCISE 1.
condition.
positive integer n.
section, this is typical of the metric spaces in which Cauchy sequences need not converge.
67
THEOREM 4.2.*
convergent.
THEOREM 4.5.*
(Hint.
Xn
B(xN+I;I).
Nand
find a suitable closed ball with center at x N+ I that contains all the
points of the sequence.)
THEOREH 4.6.
There is an important
book.)
This
Chapter V:
Connectedness
will look at the situation on the real line and prove an important
theorem - the Intermediate Value Theorem.
This chapter and the next are independent of each other.
If you
prefer, you can skip this chapter completely and go to the next, but
if you do study this chapter, I suggest you cover it completely.
One
1.
70
and A u B
X.
point is disconnected.
EXERCISE 2.
If X
A u B and A n B
~,
each other.)
You already know that if X is a metric space,
open and closed.
nected metric space these are the only subsets that are both open and
closed.
THEOREM 1.2.*
and X.
If
In
71
THEOREM 1.3.*
iff there exist open sets A and B in X with the following four prop(i) AnY of fi,
erties:
(iv) YeA u B.
(ii) B n Y of fi,
(iii)
(A n B) n Y = jil, and
word "closed".
EXERCISE 4.
connected subset of R by using Theorem 1.3 with open sets and again
with closed sets.
Comment.
you should assume that there exist open (or closed) subsets A and B
in X such that yeA u B and (A n B) n Y = fi and try to prove that
AnY or B n Y is empty.
THEOREM 1.4.
Y,
Y is
then ~ is connected.
In
connected.
THEOREM 1.5.*
Z ~
YEr
of
fi, then
CONNECTED SETS IN R.
It is convenient to extend the concept of interval on the real
line.
that for any two distinct points p and q of I, all the points between
p and q are also in I.
interval iff there exist two points p and q in I such that some point
72
x > a}
{x E R
point is an interval.
(Hint.
2.2.
Proof:
Let I = [a,b].
Since I is
E >
0, BR(p;E) n A f
so pEA
A.
A.
I and
Hence, p = b, so b
E
A.
=~.
[]
73
Use theorems
2 . 2 and 1. 5. )
Comment.
Actually, a
(It
~,
In particular, R is
connected.
3.
THEOREM 3.1.*
If
f:X~Y
f(X).
that A
G n f(X)
COROLLARY 3.2.
If
F n f(X).
f:X~Y
THEOREM 3.3.
Let
f:R~R
be a
f:R~R
that f(p) and f(q) have opposite signs, then by the Intermediate
Value Theorem you know that there is a root of the equation between
p and q.
Chapter VI:
Compactness
As you
will see, these two concepts are equivalent for metric spaces, but
since they are not equivalent for the more general case of topological
spaces, it is customary to study them separately.
perhaps they are, but after you have seen them used several times,
they cease to be tricks and become techniques.
76
cover of X if X
If
fl.
c
- r, then
(G y
r I ) is
E:
If a subfamily of a cover of X
If
is
A metric space X
interval (n - 1, n + 1).
Then (G n
subcover.
This is a contradiction.
metric space.
EXERCISE 2.
Show that
77
THEOREM 1.1.
YEr
(F y lYE r)
(G y
YEr
Gy ' there is
subset, adjoin one more open set to the family to get an open cover of
the whole space.)
LEMMA 1.5.
78
TllEOREM 1.6.
THEOREM 1.7.
(Hint.
(B(x;l)
x E A).)
EXERCISE 4.
THEOREM 2.1.
Let
f:X~Y
Let
f:X~Y
Let
f:X~Y
closed in Y.
EXERCISE.
f:R~R
and a
Let
f:X~Y
Prove that if F
THEOREM 2.5.
Let
X, then (f
f:X~Y
-1 <-
) (F)
f (F).)
By (f(p);E/2).
79
There
as required.
Comment.
[]
SEQUENTIAL COMPACTNESS.
A metric space X is said to be sequentially compact if every
80
(Hint.
points in X.
or {p}.)
n) covers X.
is bounded.
THEOREM 3.5.
>
m = 1, 2, ... , n, d(pm ,p.)
1
E,
1 < i
~
< m.
Since
t u B (Pi; E) ,
l~i~n
Then
For the
must exist a positive integer N such that for any pair of points of
the subsequence with subscripts larger than N, say p$(m) and p$(n)'
d(p$(m),p$(n
< E.
[]
is contained in at
Proof:
There
exists some Ga such that P EGa' and since Ga is open, there exists
a positive integer m such that B(p;2/m)
Ga.
Because p is a cluster
I claim
indirect proof.
If x
21m, so x
B(p;2/m).
Then,
82
[]
(Hint.
Since the
Use this to
COMPACT SUBSETS OF R.
THEOREM 4.1.
Let (G
c
YEr
Gy
A f ~ because a E A.
[a,x] is contained in a
[a,x l ].
If xl E A and
Furthermore, if
To see
Ga n [a,b].
Then,
Ix - r,
b.
E, b + E)
bound of A.
GS
E is not an upper
A.
E < X
b.
It follows,
83
~]
(b -
~,
b] is contained in a
[]
Comment.
continuous mapping.
f(p)
f:X~R
be a
EXERCISE.
Afterword
Mathematics is
been completed long ago, new results are still being discovered.
So
I hope
I have been successful, but before I stop, let me go over these two
goals with you.
Think back to your attitude toward mathematical proofs before
you started this book.
You
should know that when you are faced with a proof you have to recognize
what is given and what you have to prove.
86
some logic and have become familiar with enough mathematical techniques
so that in many cases you can actually construct the proof.
Of course,
However, you
What should you do when you come across a proof in your mathematical
reading?
However,
If
Thirdly, if a proof
It might, as
Don't stop
you now have, you should find most of the material understandable.
am sure that you will find it an interesting subject.
you, in any event, to begin to read mathematics.
I encourage
Goodbye!
Appendix M:
Mathematical Induction
The
Let (P n ) be a sequence of
PI is true;
(2)
(Iln) (P n is true).
Condition (2)
90
To
A common error
always true.
3.
positive integer n,
Appendix S:
Solutions
These
If your proof is
not the same as mine, do not assume that you are wrong but, at least,
view your proof with suspicion.
probably are right, but you should show your proof to someone else,
preferably your teacher.
Remember,
in your later work you will have to be the judge of your own work.
The second way this appendix can be used is to speed up the pace
of the work.
If you do
this, you will be able to cover the later chapters of this book and
so get a better foundation in the theory of metric spaces.
If you
92
Most proofs in
Many
times you will not find these explicitly stated in proofs that you will
be reading, but you should always supply them.
Chapter I.
1.
Prove:
(i)
Section 2.
(AxB) n (CxD)
(A n C) x (B n D).
Given:
Prove:
(x,y) s
Proof:
1.
(x,y) s (AxB)
2.
3.
(A n C)x(B n D)
n (CxD)
(Given)
(Def. of product)
4.
(x s A and x s C) and
(y s Band y s D)
(Rearrangement)
5.
(x sAn C) and (y s B n D)
(Def. of intersection)
6.
(x,y) s
(A n C)x(B n D)
(Def. of product)
93
Comment.
that is,
are obvious and will not interrupt the proofs to explain them.
(ii)
4.
Prove:
Given:
(x,y)
(A n C)x (B n D)
Prove:
(x,y)
(AxB)
Proof:
c( U Ay)
yef
(CxD)
[]
yf
(i)
(ii)
Given:
x e c (
Prove:
n
yef
Proof:
l.
x C(
2.
not-ex e
3.
yf
Ay)
(CAy)
U
yef
(Given)
Ay)
U
yf
Ay)
(Def. of complement)
(Def. of union)
4.
(Rule of negation)
5.
(Vy) (x CAy)
(Def. of complement)
6.
(Def. of intersection)
Given:
n
yef
Prove:
x c(
Proof:
yf
(CAy)
(CAy)
U
yf
Ay)
[]
94
Chapter I.
Section 3.
Theorem 3.1
(i).
f(~)
f(~)
is not empty
7'
(As sumpt io n )
l.
f(~)
2.
(ay) (y s f
3.
(ay) (ax) (x s
(~)
~
and y
f (x)
(Def. of image)
4.
Theorem 3.2
(i)
Contradiction since x s
(vi).
is false.
c(f+(B))
f+(CB)
Given:
x s f+ (cB)
Prove:
x s c (f+ (B) )
Proof:
l.
x s f+(CB)
(Given)
2.
f (x) s cB
3.
(Def. of complement)
4.
f(x) I- B
+
x t f (B)
5.
x s c (f+(B))
(Def. of complement)
There is now a second half to the proof, but this is done by reversing
the sequence of steps in the preceding proof.
Theorem 3.3
(i) .
f (f+ (B))
Given:
y s f(f
Prove:
y s B
Proof:
l.
2
[]
(B) )
y s f (f+ (B))
+
(ax) (x s f (B) and y
(Given)
f(x) )
(Def. of image)
3.
(ax) (f(x)
s Band y
f (x))
Y s B
[]
95
comment.
B.
Since y = f(x),
f (f + (B) )
surjective.
By Theorem 3.3(i), f(f+(B)) ~ B, so you must prove the reverse
inclusion under the assumption that f is surjective.
Given:
f is surjective
Bey
Prove:
f(f+(B))
Proof:
1.
2.
(ax) (x
X and y = f (x) )
3.
(ax) (x
(f is surjective)
(Def. of inverse image)
(Def. of image)
f(f+(Y)) = Y
Prove:
f is surjective
+
f (Y) = X
l.
+
2.
f(f (Y) )
Y
Proof:
Theorem 4.2.
3.
f(X) = Y
(Substitution)
4.
f is surjective []
(Def. of surjective)
96
Given:
f is surjective
gf
hf
=
E
Prove:
g(y) = hey)
Proof:
1.
2.
(:;Ix) (x
3.
(gf) (x)
(hf) (x)
4.
g(f(x
h(f(x
5.
g(y)
Theore~
4.5 (ii).
X and y = f (x) )
hey)
(f is surjective)
(gf = hf)
(Substitution)
[)
f(A n A')
f(A)
n f(A')
of X iff f is injective.
By Theorem 3.1 (v), f(A n A')
f(A)
f is injective
y
f (A)
Prove:
f(A n A')
Proof:
1.
f(A)
2.
f(A) and y
3.
(:;Ix) (x
n f (A' )
f(A')
E
f(A')
A and y = f(x
A' and
y = f (x'
4.
f(x) = f(x')
5.
x = x'
6.
(f is injective)
f (A n A')
f(x) = f(x')
f({x} n {x'})
Prove:
x = x'
f{{x})
n f({x'})
97
Proof:
fIx)
2.
fIx)
f (x')
3.
f ({x})
n f ({x' })
4.
f ({x} n -Lx' })
5.
{x} n {x'}
6.
x = x'
Theorem 4.7.
Given:
l.
I f gf
gf
(Theorem 3.1
(i) )
[]
f (x')
Prove:
Proof:
l.
fIx)
2.
g(f(x) )
g (f (x')
(gf) (x)
(gf) (x')
4.
ix (x)
5.
Chapter I.
f({x'})
ix
fIx)
x'
x'
fIx' )
(Uniqueness of images)
ix(x' )
[]
Section 5.
Theorem 5.1.
mapping.
(i)
Given:
y E Y
Prove:
(:;Ix) (x E X and f
Proof:
l.
y E Y
2.
3.
-1
(y)
x)
(f is surjective)
y)
x)
(Def. of f-l)
(ii) Next prove that X is the codomain of f-l, but this is clear
from the definition of the rule f-l.
98
(iii)
x.
To do this, assume that f-l(y) has two values and prove that they
are equal.
Proof:
Comment.
l.
2.
f(x l )
y and f(x 2 ) = y
3.
f(x l )
f(x 2 )
4.
xl
x2
[1
(f is injective)
this last part right after line 1 by saying that xl and x 2 are equal
because they are both equal to f-l(y).
On the other
hand, line 3 follows from line 2 because f(x l ) and f(x 2 ) do both equal
the single element y.
-1
99
Given:
Proof:
Let f(x)
(Notation)
f-l(y) = x
(DeL of f-l)
Section 1.
EXERCISE 2.
Prove:
ix(x)
[]
f-l(y)
(A
Proof:
max[a + c, b + d] .
(Def. of max)
1.
2.
a + c
3.
4.
max[a + c, b + d]
max[a,b] and c
~
max[c,d]
max[a,b] + max[c,d]
<
max[a,b] + max[c,d]
(Def. of max)
(Lemma)
100
max [ IXI
d' (p,r)
EXERCISE 4.
I x
O<x~l}
I}.
upper bound.
Given:
Prove:
{ I f (x) - g(x) I I x
Proof:
(f is bounded)
(g is bounded)
X and g
X
r} has an upper bound
O.
g, then d(f,g)
Given:
Prove:
d(f,g)
Proof:
O.
g
~
g(x
sup{lf(x)-g(x)llxEI}>O
d(f,g)
if
101
{ t g (xl - h (xl t
r}
{ t f (xl - h (xl I
I}.
<
sup A + sup B.
For any x in I,
I(f(xl - g(xll + (g(xl - h(xl)t
t f (xl - h (xl I
<
sup A + sup B
<
[]
section 2.
Theorem 2.2.
there exist two disjoint open balls, one with center at p and the other
with center at q.
Let d(p,ql = r > O.
that these open balls are disjoint, that is, have an empty intersection.
Use a proof by contradiction.
Given:
Prove:
a contradiction
Proof:
d(p,xl
<
d(p,ql
<
d(p,xl + d(x,ql
(M4l
d(p,xl + d(q,xl
(M3l
B(p;r/2l
r
Comment.
B(q;r/2l
<
r/2
<
r/2 + r/2 = r
<
r , a contradiction
[]
You do not have to use the radius r/2 for the balls; any
102
Chapter II.
Section 3.
Theorem 3.2.
x E B(p;r)
Prove:
Proof:
Let s
r - d(p,x).
that y is in B(p;r).
d(y,x)
< s
< r.
Then
[]
B.
family equals G.
Since B(p;r p )
in G.
U
pEG
B(p;r )
P
so X E B(x;r x )
Comment.
pEG
B(p;r )
pEG
B(p;r p ).
[]
because for each p in G there are infinitely many real numbers that
103
requires some further study of sets of real numbers that I do not want
to go into.
Chapter II.
Section 4.
Theorem 4.1.
its complement
is open.
Proof:
F is closed
Theorem 4.5.
+-+
+-+
++
(Vp) (p
+-+
(Vp) (:!Ir) (p
cF and (B(p;r)
++
(Vp) (:!Ir) (p
+-+
(Vp) (:!Ir) (p
cF and B(p;r) c F)
++
cF is open
{p} ) n F =
~)
[]
accumulation point of A.
Comment.
~,
[]
If this
set were infinite, the set of all the distances d(p,a i ) might not have
a minimum.
104
Chapter II.
Section 5.
Theorem 5.2.
number r, B(p;r) n A f
Proof:
~.
Suppose pEA.
of
f~.
However, B(p;r) n A
:0
Therefore, assume
(B(p;r)
~.
~).
Therefore,
(8r) ((B(p;r) -
~).
this value of r.
Theorem 5.3 (ii).
Proof:
if ptA,
for
[]
A is closed.
of~.
Take p E Clx(A).
=
Then for
r - d(p,q),
Since q E A, B(q;s) n A f ~, so
Then
~,
where r is any
[]
of X which contain A.
Proof:
Let C = n{F
C.
F is closed and F
c
C.
:0
A}.
Then C is closed by
105
A=
C.
A~
Therefore, it
C.
[]
Section 6.
A nonempty subset of a metric space X is bounded iff
Suppose A is bounded.
Take any x in A.
so X
Then d(p,x)
< d(p,q)
I claim that A
B*(p;r).
B*(p;r), as required.
Conversely, suppose A
B*(p;r).
[]
Section 7.
Theorem 7.2.
Y.
G n Y.
A is open in Y
+--+
A ==
+->-A
++
pEA
u
pEA
( u
pEA
+->-A
+7
By(p;rp)
(BX (p;r p ) n y)
(Theorem 7.1)
BX (p;r p )) n y
G n Y, where G
(Exercise I 2.5)
u
pEA
Bx(p;r p ) is open in X.
[]
106
Theorem 7.4.
easily proved:
A is closed in Y
cy{A) is open in y
++
n Y
(C X (G)
u cx{y
[c X (G)
n y]
[cx{G) n Y]
So,
n Y
(Exercise I 2.4)
[c X (y) n y]
u ~
cx(G)
Chapter II.
(Exercise I 2.5)
n y.
cx{G)
F n y), where F
n Y)
cx{G).
Section 8.
Theorem 8.1.
(i)
n Y.
A = c{Intx{cA
Proof of (i):
and (iii
ptA
p E cA
++
Car) (B (p;r)
n A
++
Car) (B{p;r)
AO = c{cA).
~)
cAl
p E (cA)o.
Intx(cA).
Clx{CA)
c [Int x (c (cA]
c (Int x (A
(by (i
[]
[]
107
Chapter II.
Section 9.
Theorem 9.3.
(i)
Proof:
=~.
~.
~.
Theorem 9.4.
AO
=~.
(ii)
If P
Therefore, p
B(p;r) n A, so B(p;r)
nAt~.
A -
Hence, p
Bd(A)
A - Bd(A).
A and p t Bd(A).
A, so P
~.
This proves AO
A - Bd(A).
For every
nAt~.
Bd (A)
Bd (Al.
Bd(A) , so AO
r, p B(p;r), so B(p;r)
~.
[]
AO, then p
inclusion take p
A -
AO.
[]
Chapter II.
Section 10.
Theorem 10.1.
But A
F.
X, so X
F, which gives X = F.
A.
cA
(If A is empty,
Proof:
~.
Conversely, suppose
(ar) (B(p;r) n A
~).
Then
Since ptA,
~
cB(p;r).
108
theorem.
B(p;r)
~ ~,
Chapter III.
Section 1.
Theorem 1.1.
The mapping
f:X~y
[]
Proof:
f is continuous at x
-+
d' (fex),f(xo
< s)
[]
is an open set in x.
Proof:
Suppose f is continuous.
G, so (:3:s) (By(f(xo);S)
G).
open in y, f+(Byefexo);s
is open in x.
o.
Since Byefexo);s) is
But Xo s f+(By(f(xo);s,
Therefore,
109
Since x
arbitrary point of X, this proves that f is continuous.
is an
[]
Chapter III.
Section 2.
Theorem 2.4.
Let
AO.
Then
(~r)
(Bx(xo;r)
By(f(xO );)'
Theorem 2.5.
[]
then f = g.
Proof:
Take 0
=~.
= min
g(BX(Xo;o
[0 1 ,0 2 ],
~
Then f(BX(Xo;o
By (g(x o );2)'
Theorem II 10.2.
and 02
by
Since a
[]
A,
110
Chapter III.
Section 3.
EXERCISE 1.
by f(x) = l/x.
f:X~R
I~ - xlo I
But x > x
It follows that
Comment.
Ix
Then if Ix - xol
- xl
xXo
I~ x~1
X,
E(X )2
<
xx
(x )2
o
< 8, with x
< , so f
so
X <
is continuous at xo'
working backwards.
To prove that f is not uniformly continuous, I will use a proof
by contradiction.
corresponds to an arbitrary
as in the definition.
28
2
If(x ) - f(x) I
which is a contradiction.
Ix~
[]
8.
- x:1 =
Then x
However, for x
1- x~1
Take x
to be a
is in X and
111
Chapter IV.
Section 1.
EXERCISE 2.
$:~~~
= a($W)(n)
and
Then c n
bw(n)
Let
> $ (W (j
$ (W (i
Take i and j in
such
W:~~~
$:~~~
be an increasing mapping.
[]
positive integer n.
(i)
smallest element of N.
$(1)
(ii)
EXERCISE 4.
n for every
Assume $(k)
Therefore,
[]
Take X
as an infinite subset of any set and prove that there exists a sequence
of distinct points of X.
Proof:
Let X
N be infinite.
to be any element of X.
Take a l
Since X is infinite,
112
this set.
a.
n
The recursion
is then complete and this defines the sequence (an) of distinct points
of X.
In the general case where X is an infinite subset of any set the
recursion process proceeds in the same way, except you cannot choose
a n + l to be the smallest element of X - {aI' a 2 , ... , an} because this
set need not even have any ordering defined on it.
However, since it
is not empty, it contains elements and you merely choose one of them
arbitrarily.
[]
Section 2.
Let A be a subset of a metric space X.
Then a point p
(x n ) by recursion as follows:
Define a sequence
1, 2,
(B(p;l/(n + 1))
n, xi
(B(p;l/i) - {p}) n A.
Since
sion; all the points of the sequence are in A and none equals p.
show that lim x
p, take any
To
< E,
so
113
Xn
B(p;l/n)
B(p;E), as required.
~,
>
p.
N, xn
But then
so p is an accumulation point of A.
[]
Section 3.
The'orem 3.2.
(x~(n))
converges to p.
Proof:
mapping
~:~~~
by recursion.
Let
~(l)
Construct an increasing
such that xm E B(p;l), that is, take the set of all the subscripts of
points of the sequence that lie in B(p;l) and choose the smallest one.
Assume that
~(l),
~(2),
<
<
<
~(l)
~(2)
~(n)
... ,
and
~(n)
x~(i)
(X n )
n.
~(n)
and
, there are
This defines
~,
(x~(n))
converges to p, suppose
E > 0 is given and take N to be a positive integer such that N > l/E.
Then if n > N,
x~(n)
E B(p;l/n)
Conversely, suppose
(x~(n))
B(p;l/N)
B(p;E), so lim
x~(n)
Since lim
= p.
x~(n)
114
B(P;).
Then
there exists an integer m such that (m) > max [M,N], for example
m = max [M,N] + 1 will work.
X(m)
Chapter IV.
[J
Section 4.
Theorem 4.2.
Theorem 4.5.
Proof:
d(xm,P) < /2 + /2
p.
d(xn,x m) +
[]
Let k
n, xn
+ 1)
B(~+l;k
d(xn'~+l)
max {d(xi'~+l)
~ B*(~+l;k
+ 1).
11
In particular,
~ i ~ N}.
(Theorem II
[]
Chapter IV.
Theorem 5.3.
Section 5.
If every bounded infinite subset of a metric space X has
115
Proof:
tion point in X.
If the range
If the
(xn ) converges.
v.
[]
Section 1.
Theorem 1.2.
Proof:
Assume X is disconnected.
and X.
and X = A u B.
Then A
and A
X.
and X.
=~
Let B
and X
cA.
and X,
A u B, so X is disconnected.
[]
iff there exist open sets A and B in X with the following four
properties:
(iv) YeA
(i) AnY
U
B.
t~,
(ii) B n Y
t~,
(iii)
(A n B) n Y
~,
and
word "closed".
Proof:
Y is disconnected
Bl
~,
such that A n Y
and Y
such that Al
(A n Y)
~,
n (B n Y)
~,
Al n Bl
(A n Y)
and Y
~,
Al u Bl
B n Y
(B n Y)
~,
++
there exist
116
But (A n Y)
iff yeA
B.
(B n Y)
(A n B)
n Y.
Also, Y
(A n Y)
(B n Y)
When
these results are substituted into the above conditions, you get the
statement of the theorem.
If (A
lYE f)
Proof:
Suppose
YEf
B are open in X.
Ay
n A
YEf Y
f~,
then
Let p E n Ay '
YEf
f~.
Chapter V.
Theorem 3.1.
n B
~.
f(X).
where A and
Since Ay c A u Band Ay n (A n B)
=~.
YEr
Ay is connected.
Section 3.
If
f:X~Y
~,
each y in r, so (u Ay)
YEf
Ay is connected.
YEf
A u B and (A n B) n (u Ay) =
YEr
[]
is connected in Y.
is f(X) itself.
It
is connected.
G n f(X)
F n f(X).
Then f+-(G)
in X since f is continuous.
is done as follows:
117
x g f+(G) ~ f(x)
f(x)
++
Then f+(G)
g G ++ f(x)
g
F n f(X)
++
f(x)
++
be that f+(G) = X.
Therefore, f(X)
G n f(X)
Chapter VI.
(G y
f(X), as required.
f(f+(G
~ ~,
so it must
G and so
[]
Section 1.
Theorem 1.4.
Proof:
f+(F)
connected space X.
g A
G n f(X) ++ f(x)
the family
(G y
YEf)
u cF.
If cF is in this
Take
Gy
[]
Section 3.
Theorem 3.3.
points in X.
(B(p;r p ) -
(B(p;r )
p
still a cover.
A
(I have written r i
in place of r
Pi
.)
Then
A n X
set in the union is either empty or a singleton, and since there are
118
[]
APPENDIX M
1.
Proof:
is the statement:
log (xn )
n log(x).
n log(x).
(1)
PI is obviously true.
(2)
(k + 1) log(x).
log (x k +l ) = log(x xk) = log (x) + log(xk ) = log (x) + k log (x)
(k + 1) log(x), as required.
[]
But,
Index
Accumulation point
41
32
Bijection
49
Boundary
Boundary point
49
16
Cartesian product
66
Cauchy sequence
Closure
50
44
Difference of sets
15
Disconnected space
69
64
20
Compact set
Embedding
32
Empty set
15
63
15
Complete space
17
Family
76
67
69
Image
Connected subset
70
Indirect proof
53
30
Identity mapping
Connected space
Continuous mapping
83
Heine-Borel Theorem
24
Composite mapping
Contrapositive
36
80
Epsilon net
76
Compact space
70
20
Domain
Cluster point
Complement
Dense
Divergent sequence
43
Codomain
76
Discrete metric
41
Closed set
Cover
Disconnected subset
38
Closed ball
83
Diameter
44
Bounded set
19
Counterexample
Bolzano-Weierstrass Theorem
63
Convergent sequence
20
Infinum
Injection
37
30
77
120
48
Interior
Range
48
Interior point
Intersection of a family
Intersection of sets
73
20
32
57
57
37
Lower bound
Mapping
34
Precompact
76
15
Subspace
46
Supremum
38
63
29
80
Totally bounded
44
Unbounded set
Union of a family
35
Metric space
Open set
38
56
Uniformly continuous
19.
Open ball
Sphere
Surjection
Limit of a sequence
79
Sequentially compact
Subset
Isometric spaces
Isometry
17
60
Sequence
Subcover
Inverse mapping
Metric
15
71
Inverse image
20
Restriction
Interval
20
38
40
15
Union of sets
upper bound
37
Usual metric on R
80
17
Usual metric on R2
35
35
Universitext
Editors: F.W. Gehring, P.R. Halmos, C.C. Moore
Chern: Complex Manifolds Without Potential Theory
Chorin/Marsden: A Mathematical Introduction to Fluid Mechanics
Cohn: A Classical Invitation to Algebraic Numbers and Class Fields
Curtis: Matrix Groups
van Dalen: Logic and Structure
Devlin: Fundamentals of Contemporary Set Theory
Edwards: A Formal Background to Mathematics 1: Logic, Sets and
Numbers
Edwards: . A Formal Background to Mathematics 2: A Critical Approach to
Elementary Analysis
Frauenthal: Mathematical Modeling in Epidemiology
Fuller: FORTRAN Programming: A Supplement for Calculus Courses
Gardiner: A First Course in Group Theory
Greub: Multilinear Algebra
Hajek/Havri:'mek: Mechanizing Hypothesis Formation: Mathematical
Foundations for a General Theory
Hermes: Introduction to Mathematical Logic
Kalbfleisch: Probability and Statistical Inference 1/11
Kelly/Matthews: The Non-Euclidean, Hyperbolic Plane: Its Structure and
Consistency
Kostrikin: Introduction to Algebra
Lu: Singularity Theory and an Introduction to Catastrophe Theory
Marcus: Number Fields
Meyer: Essential Mathematics for Applied Fields
Moise: Introductory Problem Courses in Analysis and Topology
Oden/Reddy: Variational Methods in Theoretical Mechanics
Reisel: Elementary Theory of Metric Spaces: A Course in Constructing
Mathematical Proofs
Rickart: Natural Function Algebras
Schreiber: Differential Forms: A Heuristic Introduction