Notes on Topology

Notes on Topology
These are links to PostScript files containing notes for various topics in topology.

I learned topology at M.I.T. from Topology: A First Course by James Munkres. At the time, the first
edition was just coming out; I still have the photocopies we were given before the printed version was
ready!

Hence, I'm a bit biased: I still think Munkres' book is the best book to learn from. The writing is clear
and lively, the choice of topics is still pretty good, and the exercises are wonderful. Munkres also has a
gift for naming things in useful ways (The Pasting Lemma, the Sequence Lemma, the Tube Lemma).

(I like Paul Halmos's suggestion that things be named in descriptive ways. On the other hand, some
names in topology are terrible --- "first countable" and "second countable" come to mind. They are
almost as bad as "regions of type I" and "regions of type II" which you still sometimes encounter in
books on multivariable calculus.)

I've used Munkres both of the times I've taught topology, the most recent occasion being this summer
(1999). The lecture notes below follow the order of the topics in the book, with a few minor variations.

Here are some areas in which I decided to do things differently from Munkres:

● I prefer to motivate continuity by recalling the epsilon-delta definition which students see in
analysis (or calculus); therefore, I took the pointwise definition as a starting point, and derived
the inverse image version later.
● I stated the results on quotient maps so as to emphasize the universal property.
● I prefer Zorn's lemma to the Maximal Principle; for example, in constructing connected
components, I use Zorn's lemma to construct maximal connected sets.
● I did Tychonoff's theorem at the same time as the other stuff on compactness. The proof for a
finite product is long enough that I decided to save some time by omitting it and just doing the
general case.
● It seems simpler to me to do Urysohn's lemma by indexing the level sets with dyadic rationals in
[0,1] rather than rationals.

● How do I view/print the notes?

● How may I use/distribute the notes? How did you write them?

file:///F|/Downloads/BOOKS/General%20Topology/notes/Files/Notes%20on%20Topology.htm (1 of 2)2004/07/26 03:08:14 •.•

 Notes on Topology

● funct.ps (86,912 bytes; 5 pages): Review of topics in set theory.

● top.ps (168,720 bytes; 6 pages): Topological spaces, bases.
● closed.ps (55,509 bytes; 4 pages): Closed sets and limit points.
● haus.ps (132,025 bytes; 2 pages): Hausdorff spaces.
● contin.ps (100,021 bytes; 3 pages): Continuous functions.
● homeo.ps (83,891 bytes; 3 pages): Homeomorphisms.
● prod.ps (51,962 bytes; 3 pages): The product topology.
● metric.ps (140,993 bytes; 8 pages): Metric spaces.
● quot.ps (116,395 bytes; 5 pages): Quotient spaces.
● conn.ps (82,400 bytes; 4 pages): Connectedness.
● realconn.ps (55,934 bytes; 3 pages): Connected subsets of the reals.
● pathco.ps (45,136 bytes; 3 pages): Path connectedness, local connectedness.
● compact.ps (240,878 bytes; 8 pages): Compactness (including Tychonoff's theorem).
● compreal.ps (69,020 bytes; 3 pages): Compact subsets of the reals.
● limcomp.ps (107,691 bytes; 4 pages): Limit point compactness and sequential compactness.
● loccpt.ps (70,399 bytes; 3 pages): Local compactness; the one-point compactification.
● countab.ps (40,734 bytes; 2 pages): The countability axioms.
● sep.ps (139,569 bytes; 5 pages): The separation axioms: regularity and normality.
● urysohn.ps (69,043 bytes; 2 pages): Urysohn's lemma.
● tietze.ps (45,448 bytes; 3 pages): The Tietze Extension Theorem.

Send comments about this page to: mailto:[email protected]

Last updated: August 16, 2000

Bruce Ikenaga's Home Page

Math Department Home Page

Millersville University Home Page

Copyright 1999 by Bruce Ikenaga

file:///F|/Downloads/BOOKS/General%20Topology/notes/Files/Notes%20on%20Topology.htm (2 of 2)2004/07/26 03:08:14 •.•

 6-14-1999
Review: Set Theory, Functions, and Relations

This is a brief review of material on set theory, functions, and relations. I'll concentrate on things which
are often not covered (or given cursory treatment) in courses on mathematical proof.
De nition. Let A and B be sets, and let f : A ! B be a function from A to B .
1. f is injective if for all a1 a2 2 A, f (a1 ) = f (a2 ) implies a1 = a2.
2. f is surjective if for all b 2 B , there is an element a 2 A such that f (a) = b.
3. f is bijective if f is injective and surjective.
De nition. Let f : A ! B be a function. Let B0  B .
1. The image of f is the set
f (A) = ff (a) j a 2 Ag :
2. The inverse image of B0 is the set
f ;1 (B0 ) = fa 2 A j f (a) 2 B0 g :
Remark. f is surjective if and only if f (A) = B .

Example. (a) Let f : A ! B and g h : B ! C be functions. Suppose that f is surjective and g  f = h  f .

Prove that g = h.
Let b 2 B . I must show that g(b) = h(b).
Since f is surjective, I may nd an element a 2 A such that f (a) = b. Then g  f = h  f implies that
g(f (a)) = h(f (a)) so g(b) = h(b):
Therefore, g = h.
(b) Let f : A ! B . Suppose that for all g h : B ! C , if g  f = h  f , then g = h. Prove that f is surjective.
First, I'll take care of the cases where A =  or A contains a single element.
If A =  and B = , then f is vacuously surjective: Every element of B (there are none) is an image of
an element of A.
If A =  and B 6= , then B contains at least one element b. Dene g : B ! f0 1g by g(b) = 0 and
h : B ! f0 1g by h(b) = 1. It is vacuously true that g(f (a)) = h(f (a)) for all a 2 A (since there are no a's
in A), but g 6= h. Therefore, the hypothesis isn't satised, and this case is ruled out.
Suppose A consists of a single element a. If B = , then f is vacuously surjective. If B consists of a
single element, then f is obviously surjective.
Finally, suppose that B contains more than one element. f (a) is an element of B  suppose that b0 2 B
and b0 6= f (a). Consider the functions g h : B ! B given by
g(b) = b for all b 2 B
h(b) = fb (a) ifif bb 6= b0
= b0 :
Then g(f (a)) = f (a) = h(f (a)), so g  f = h  f . However, g 6= h. Therefore, the hypothesis isn't
satised, and this case is ruled out.
1
 This concludes the proof for the trivial cases.
Assume that A contains more than one element.
Suppose, on the contrary, that f (A) 6= B . I'll show that the hypothesis is violated.
If b 2 f (A), choose an element ab 2 A such that f (ab ) = b. Since A contains more than one element, I
may nd elements a1 a2 2 A such that a1 6= a2 . Dene g h : B ! A by

g(b) = aab ifif bb 22= ff ((AA)) and g(b) = aab ifif bb 22= ff ((A )
A) :
1 2

If a 2 A, then f (a) 2 f (A), so

g(f (a)) = af (a) = h(f (a)):
Thus, g  f = h  f .
However, since f (A) 6= B , there is an element b0 2 B ; f (A). For this element, g(b0 ) = a1 , but
h(b0) = a2 . Therefore, g 6= h.
This proves that the hypothesis is violated. This contradiction in turn establishes that f (A) = B , so f
is surjective.
Remark. This result is often expressed by saying that surjective functions are right-cancellable. The dual
result is also true: Injective functions are left-cancellable.

Example. Let f : A ! B , and suppose that B1 B2  B . Prove that f ;1 (B1  B2 ) = f ;1 (B1 )  f ;1 (B2 ).
Suppose a 2 f ;1 (B1  B2 ). Then f (a) 2 B1  B2 , so f (a) 2 B1 or f (a) 2 B2 .
If f (a) 2 B1 , then a 2 f ;1 (B1 ), so a 2 f ;1 (B1 )  f ;1 (B2 ). If f (a) 2 B2 , then a 2 f ;1 (B2 ), so
a 2 f ;1 (B1 )  f ;1 (B2 ).
Therefore, a 2 f ;1 (B1 )  f ;1 (B2 ), so f ;1 (B1  B2 )  f ;1 (B1 )  f ;1 (B2 ).
Conversely, suppose that a 2 f ;1 (B1 )  f ;1 (B2 ), so a 2 f ;1 (B1 ) or a 2 f ;1 (B2 ).
If a 2 f ;1 (B1 ), then f (a) 2 B1 , so f (a) 2 B1  B2 , and hence a 2 f ;1 (B1  B2 ). If a 2 f ;1 (B2 ), then
f (a) 2 B2 , so f (a) 2 B1  B2 , and hence a 2 f ;1 (B1  B2 ).
Therefore, a 2 f ;1 (B1  B2 ), so f ;1 (B1 )  f ;1 (B2 )  f ;1 (B1  B2 ).
This proves that f ;1 (B1  B2 ) = f ;1 (B1 )  f ;1 (B2 ).

De nition. Let A be a set. A relation < on A is a simple order (or linear order) if:
1. (Comparability) For all a b 2 A, exactly one of the following is true: a < b, b < a, or a = b.
2. (Nonre
exivity) There is no element a 2 A such that a < a.
3. (Transitivity) For all a b c 2 A, if a < b and b < c,then a < c.

Example. Let < be a simple order on a set A. The dictionary order (or lexicographic order) on A  A
is given by: (a b) < (a0 b0) if and only if
1. a < a0  or
2. a = a0 , and b < b0 .
The words in a dictionary are ordered by their rst letter. Among words with the same rst letter, they
are ordered by their second letter, and so on. This explains the name of this order relation.
To give a specic example, in R2 under the dictionary order, (1 4) < (2 4) and (1 6) < (1 8).

2
De nition. If A is an ordered set, a b 2 A, and a < b, then the open interval from a to b is the set
(a b) = fx j a < x < bg:
It may happen that this set is empty. In that case, since there are no elements \between" a and b, it's
natural to say that a is the predecessor of b and b is the successor of a.
Unfortunately, \(a b)" can mean a point in a Cartesian product or an open interval in a simple order.
To avoid confusion, I'll sometimes write a  b for the point when both concepts occur in the same discussion.

Example. Consider the dictionary order on R2 . The open interval from the point 1  2 to the point 2  4 is
the union of:
1. The open ray f1  y j y > 2g.
2. The lines fx  y j 1 < x < 2g.
3. The open ray f2  y j y < 4g.

2x4

1x2

De nition. Suppose A and B are sets with simple orders <A and<B , respectively. Let f : A ! B be a
function. f is order-preserving if for all a1 a2 2 A,
a1 <A a2 implies f (a1 ) <B f (a2 ):

Example. With the usual order relation on R, dene f : R ! R by f (x) = x3 . Then f is order-preserving.
In fact, a function R ! R is order-preserving if it is an increasing function.

De nition. Let A and B be ordered sets. A and B have the same order type if there's an order-preserving
bijection from A to B .
De nition. A set is well-ordered if every nonempty subset has a smallest element.
The Well-Ordering Axiom says that every set can be well-ordered. This is an axiom of set theory
it's equivalent to other results which can be taken as axioms, such as the Axiom of Choice and Zorn's
lemma.
Well-ordering means that for a given set, there is some order relation on the set relative to which the
set is well-ordered. That well-ordering will usually have little to do with familiar order relations.
3
De nition. Let A be an ordered set, and let B  A.
1. c 2 B is the largest element of B if b < c for all b 2 B , b 6= c.
Likewise, c 2 B is the smallest element of B if c < b for all b 2 B , b 6= c.
2. An element a 2 A is an upper bound for B if b < a for all b 2 B , b 6= a. If B has an upper bound, B
is bounded above.
Likewise, an element a 2 A is a lower bound for B if a < b for all b 2 B , b 6= a. If B has a lower
bound, B is bounded below.
3. If the set of upper bounds of B is nonempty, then the smallest element of that set is the least upper
bound for B . It is denoted lub B or sup B.
Likewise, if the set of lower bounds of B is nonempty, the largest element of that set is the greatest
lower bound for B . It is denoted glb B or inf B.
4. A set has the least upper bound property if every nonempty subset that is bounded above has a
least upper bound.
A set has the greatest lower bound property if every nonempty subset that is bounded below has
a greatest lower bound.

Example. The real numbers with the usual ordering satisfy the least upper bound property. In fact, it is
one of the axioms for the real numbers.

Mathematics is based on a collection of axioms for set theory. They are called the Zermelo-Fraenkel
Axioms with the Axiom of Choice (often abbreviated to ZFC).
While the Axiom of Choice is accepted and used by most mathematicians, the situation was dierent
earlier in this century. Some mathematicians felt that the Choice Axiom was problematic and refused to use
it others took care to note when they used the Choice Axiom in a proof. Many of the objections stemmed
from the often startling results which can be proved using the axiom.
Axiom of Choice. Let fXa g be a collection of disjoint nonempty sets. There exists a set X containing
exactly one element from each Xa .
The Axiom of Choice is equivalent to the Well-Ordering Axiom, and also to an important result known
as Zorn's lemma. The statement of Zorn's lemma requires some preliminaries.
De nition. A relation on a set X is a partial order if:
1. (Re
exivity) x x for all x 2 X .
2. (Antisymmetry) If x y and y x, then x = y.
3. (Transitivity) If x y and y z , then x z .

Example. The set of subsets of a set is partially ordered by inclusion.

Example. The usual \less than or equal to" relation on R is a partial order.
4
De nition. Let X be a partially ordered set.
1. A subset Y  X is totally ordered (or linearly ordered) if for all x y 2 Y , either x y or y x.
A totally ordered subset is also said to be a chain.
2. If Y  X , then Y has an upper bound if there is an element x 2 X such that y x for all y 2 Y .
3. An element x 2 X is maximal if y 2 X and x y implies x = y.
Zorn's Lemma. If X is a nonempty partially ordered set in which every chain has an upper bound, then
X has maximal elements.
Zorn's lemma is an immensely useful form of the Axiom of Choice. For example, you may have seen a
proof that every nite-dimensional vector space has a basis. Zorn's lemma can be used to show that every
vector space has a basis.

Example. Let G be a group and let g 2 G, g 6= 1. There is a subgroup H < G such that g 2= H and H is
maximal among subgroups with this property.
Let S be the collection of subgroups of G which do not contain g. S 6= , since f1g 2 S . S is partially
ordered by inclusion.
Let T be a chain in S . Let

H0 = H:
H 2T
Thus, H0 is just the union of all the elements of the chain.
I claim that H0 is a subgroup of G which does not contain g.
First, 1 2 H for all H 2 T , so certainly 1 2 H0.
If h h0 2 H0, then h 2 H and h0 2 H 0 for some H H 0 2 T . But T is a chain, so either H  H 0 or
H  H.
0
Suppose H  H 0. Then h h0 2 H 0 , so (since H 0 is a subgroup) hh0;1 2 H 0 . Therefore, hh0;1 2 H0, and
hence H0 is a subgroup.
If g 2 H0, then g 2 H for some H 2 T . But H 2 S , so g 2= H by denition of S . Therefore, g 2= H0 .
I've shown that H0 is a subgroup of G which does not contain g, so H0 2 S . Since H  H0 for all
H 2 T , it follows that H0 is an upper bound for T in S .
Since every chain has an upper bound, Zorn's lemma implies that S has a maximal element H . Then
H is a subgroup of G which is maximal among subgroups which do not contain g.

c 1999 by Bruce Ikenaga 5

 6-15-1999
Topological Spaces
A function f : R ! R is continuous at x = a if and only if
For every > 0, there is a  > 0 such that if  > jx ; aj > 0, then > jf (x) ; f (a)j.
In words,
You can make f (x) arbitrarily close to f (a) by making x su ciently close to a.
Continuity is important in analysis and calculus. The idea of \closeness" is an important part of the
denition of continuity. What would \closeness" mean in an abstract setting | one in which there's no way
to measure the distance between two points?
A topology for a space is, roughly speaking, a specication of all "closeness relations" in the space.
De nition. Let X be a set. A topology on X is a collection T of subsets of X satisfying the following
conditions:
1.  2 T and X 2 T .
2. The union of elements of T is an element of T .
3. The nite intersection of elements of T is an element of T .
The elements of T are called open sets. The set X together with the topology T is a topological
space.

Example. If X is a set, take D to be P (X ), the power set of X . (Recall that P (X ) is the set of all subsets
of X .)
D is clearly a topology on X  it is called the discrete topology. In the discrete topology, all subsets
are open.
The other extreme is the indiscrete topology on X . In this case, I = f X g. Again, it is clear that
this is a topology on X .

De nition. If T1 and T2 are topologies on a set X , then T1 is ner than T2 (or T2 is coarser than T1) if
T1  T2. In other words, ner topologies contain more open sets.
Thus, if D is the discrete topology on X , then D is ner than any other topology on X . If I is the
indiscrete topology on X , then I is coarser than any other topology on X .
It is, of course, possible for two topologies on X to be incomparable, i.e. neither may be ner than the
other.
Analysts often refer to \stronger" and \weaker" topologies, but their usage is often at variance with the
usage of topologists.

Example. Let X = f1 2 3g, and consider the following topologies on X :

T1 = f f1g f2g f1 2gX g
T2 = f f1g f1 2g X g
T3 = f f1 2g X g:
T4 = f f2 3g X g:
1
 1
1 2 3

2 1 2 3

3 1 2 3

4 1 2 3

Then T1 is ner than T2 , and T2 is ner than T3. On the other hand, T4 is not comparable with the
other three.
As this example shows, topology does not necessarily concern itself with the concepts of measurement
or distance. They might not be applicable to the space under consideration.

Example. The standard topology on R is the topology whose open sets consist of arbitrary unions of open
intervals.
The empty set is considered to be a degenerate open interval, and R can be written as a union of open
intervals.
A union of unions of open intervals is a union of open intervals.
It remains to show that this topology is closed under nite intersections. It's fairly obvious that two
open intervals either do not intersect, or intersect in an open interval I won't write out the details. The
general case for nite intersections reduces to this case by distributing intersections over unions.

Example. The standard topology on Rn is obtained by starting with open balls

B (x r) = fy 2 Rn j jx ; yj < rg:
(B (x r) is called the open r-ball centered at x.) The open sets are arbitrary unions of open balls.
I'll defer the verication that this gives a topology, since it will follow easily once I develop the notion
of a basis.

The examples above give a typical way of dening a topology: Start with a collection of subsets and
then extend the collection to a topology by taking arbitrary unions. The next denition makes precise the
conditions I need to be able to do this.
De nition. Let X be a set. A collection B of subsets of X is a basis if:
1. Every element of X is contained in some element of B.
2. If B1  B2 2 B and x 2 B1 \ B2 , there is an element B3 2 B such that
x 2 B3  B1 \ B2 :
B2
B1
x
B3

2
 The elements of B are called basic subsets, or basis elements.
Remark. By induction, it follows that if B is a basis, B1  : : : Bn 2 B, and x 2 B1 \ \ Bn , then there is
an element B 2 B such that
x 2 B  B1 \ \ Bn :

Example. The open intervals in R form a basis.

If x 2 R, then x 2 (x ; 1 x + 1) (for instance).
If x is contained in the intersection of two open intervals, the intersection is itself an open interval which
contains x and is contained in the intersection.
The open balls B (x r) in Rn form a basis. Obviously, every point in Rn is contained in some open ball.
Take open balls B (x1  r1) and B (x2  r2). Assume that they intersect, and let x 2 B (x1  r1) \ B (x2  r2).
I have to nd an open ball centered at x which is contained in the intersection.
Let s1 = r1 ; jx ; x1j. Then B (x s1 )  B (x1  r1).
r1 - | x - x1 |

| x - x1 |

x1

B(x1 ; r1 )

Likewise, let s2 = r2 ; jx ; x2j. Then B (x s2 )  B (x2  r2).

Therefore, if I let r = min(s1  s2 ), then
B (x r)  B (x1  r1) \ B (x2  r2):
This shows that the open balls in Rn form a basis.

Proposition. Let X be a set, and let B be a basis. Dene T to be the collection of subsets U  X satisfying
the following property:
U 2 T if and only if for all x 2 U , there is an element B 2 B such that x 2 B  U .
Then T is a topology on X .
T is the topology generated by the basis B.
Proof. The empty set is in T  it satises the dening condition vacuously, since it has no elements.
X 2 T : If x 2 X , the rst axiom for a basis implies that there is an element B 2SB such that x 2 B  X .
Let fUa g be a collection of elements of T . I need to show that the union U = a Ua is in T .
Let x 2 U . Then x 2 Ua for some a. Since Ua 2 T , there is an element B 2 B such that x 2 B  Ua .
Then
x 2 B  Ua  U:
Hence, U 2 T .
3
 T
Let V1  : : : Vn 2 T . I must show that V = ni=1 Vi is in T . Let x 2 V . Then x 2 Vi for all i. Since
Vi 2 T , for each i I may nd an element
T Bi 2 B such that x 2 Bi  Vi. By the remark above, I may nd an
element B 2 B such that x 2 B  ni=1 Bi . Then

x2B
\n Bi  \n Vi = V:
i=1 i=1
Hence, V 2 T .
Therefore, T is a topology.
Notice that elements of a basis are automatically elements of the topology they generate: Basic sets are
open.
De nition. If T is a topology and B is a basis, then B is a basis for T if the topology generated by B is
the same as T .
Lemma. Let X be a set, B a basis, and T the topology generated by B. Then T is equal to the collection
of all unions of elements of B.
Proof. Since basis elements are open, unions of basis elements are open. Therefore, the collection of all
unions of basis elements is contained in T .
Conversely, let U 2 T . I must express U Sas a union of basis elements. For each x 2 U , nd a basis
element Bx such that x 2 Bx  U . Then U = x2U Bx .
Here are two more facts about bases. The proofs are straightforward.
1. Let T1 and T2 be topologies on X generated by bases B1 and B2 , respectively. Then T1 is ner than T2
if and only if for every U 2 B2 and x 2 U , there is an element V 2 B1 such that x 2 V  U .
2. Given a topology T on X , a subcollection B of open sets is a basis for T if for all U 2 T and x 2 U ,
there is an element B 2 B such that x 2 B  U .
It's possible to go to a \lower" level and generate a topology by taking unions and nite intersections.
De nition. If X is a set, a subbasis is a collection of subsets of X whose union is X .
Proposition. If S is a subbasis in X , the collection of all unions of nite intersections of elements of S is a
topology on X .
Proof. Let T denote the collection of all unions of nite intersections of elements of S .
The union of the elements of the subbasis is X , so X 2 T . Taking the empty union gives the empty set,
so  2 T .
The union of unions of nite intersections of elements of S are unions of nite intersections of elements
of S , so such a union is an element of T .
The verication that T is closed under nite intersections is routine, using the fact that intersections
distribute over unions. I'll omit the details.

Example. (Order topologies) Let X be a set linearly ordered by a relation <. Consider the following
collection of subsets of X :
1. All intervals s b), where b > s and s is the smallest element of X (if there is one).
2. All intervals (a l], where a < l and l is the largest element of X (if there is one).
3. All intervals (a b), where a < b.
I claim that this collection is a basis.
4
 First, if x 2 X , I must nd an element of the basis containing x. If x is the largest element, take an
interval of the form (a x] if x is the smallest element, take an interval of the form x b). If x is neither
largest nor smallest, I may nd a, b such that a < x < b. Then x 2 (a b).
To verify the second axiom for a basis, just observe that the intersection of two intervals is either empty,
or an interval.
The topology generated by this basis is called the order topology. On R, for instance, this topology
is the standard topology.

Example. (Partition topologies) Let X be a set, and let P be a partition of X . Then P is a basis: the
union of the sets in the collection is X , and since the sets do not intersect, the second basis axiom is satised
vacuously.
The topology generated by P is called the partition topology on X . The open sets are unions of
partition elements.
Note that if the partition consists of X itself, the partition topology is the indiscrete topology on X .
And if the partition elements are the points of X , the partition topology is the discrete topology.

Example. (Product Topology) Suppose (X S ) and (Y T ) are topological spaces. (That is, S is the
topology on X and T is the topology on Y .)
Dene a basis on the product X  Y by
B = fU  V j U 2 S  V 2 T g :
I need to verify that this is a basis.
First, if (x y) 2 X  Y , nd open sets U  X , V  Y such that x 2 U  X and y 2 V  Y . Then
(x y) 2 U  V 2 B.
Now let U1 and U2 be open in X , and let V1 and V2 be open in Y . Suppose that
(x y) 2 (U1  V1 ) \ (U2  V2):
Note that
(U1  V1 ) \ (U2  V2 ) = (U1 \ U2 )  (V1 \ V2 ):
Find open sets U  X and V  Y such that x 2 U  U1 \ U2 and y 2 V  V1 \ V2. Then U  V 2 B,
and
(x y) 2 U  V  (U1 \ U2)  (V1 \ V2):
Therefore, B is a basis.
The topology generated by B is the product topology on X  Y .
In fact, if C is a basis for the topology on X and D is a basis for the topology on Y , then
E = fU  V j U 2 C V 2 Dg
is a basis for the product topology on X  Y .
For instance, give R the standard topology generated by the open intervals. Then the product topology
on R2 is generated by products of open intervals, i.e. open rectangles. In fact, this topology is the same
as the topology generated by open balls. This will follow from the next lemma, which gives an easy way of
telling when two topologies are the same.

Lemma. Let T1 and T2 be topologies on X . Then T1  T2 if and only if for every U 2 T1 and x 2 U , there
is an element V 2 T2 such that x 2 V  U .
5
Proof. Suppose T1  T2. If U 2 T1 and x 2 U , then U 2 T2, so x 2 U  U , and the condition holds.
Conversely, suppose that for every U 2 T1 and x 2 U , there is an element V 2 T2 such that x 2 V  U .
Let U be open in T1. I want to show that U 2 T2. For each x 2 U , nd an element Vx 2 T2 such that
x 2 Vx  U . Then Sx2U Vx 2 T2, but clearly Sx2U = U . Therefore, U 2 T2. Hence, T1  T2.
Corollary. The product topology on R2 is the same as the topology generated by the open balls.
Proof. The picture below indicates the idea:

x
x

Example. (Subspace topology) Let X be a topological space, and let A  X . Dene a topology on A by
stipulating that the open sets be the intersections of open sets of X with A.
It is easy to check that this is a topology. It is called the subspace topology on A. When a subset A
of X is given the subspace topology, A is said to be a subspace of X .
It is also easy to check that if B is a basis for the topology on X , then
fU \ A j U 2 Bg
is a basis for the subspace topology on A.

Remark. If A  X , a subset of A that is open in the subspace topology need not be open in X .
For example, consider the subset Z R, where R has the standard topology. Then f0g is open in the
subspace topology on Z, since (;1 1) is open in R and f0g = Z\ (;1 1).
However, f0g is not open in R. Take 0 2 f0g. There is no open interval (a b) in R such that 0 2 (a b) 
f0g.
In fact, the subspace topology on Zis just the discrete topology.
Theorem. Let X and Y be topological spaces. Let A be a subspace of X , and let B be a subspace of Y .
The product topology on A  B agrees with topology it inherits as a subset of X  Y .
Proof. Let U be open in A and let V be open in B . Thus,
U = A \ U 0 and V = B \ V 0 
where U 0 is open in X and V 0 is open in Y . U  V is a basic open set in the product topology.
Now
U  V = (A \ U 0 )  (B \ V 0) = (A  B ) \ (U 0  V 0 ):
Therefore, U  V is open in the subspace topology on A  B . Thus, the product topology is contained
in the subspace topology.
Going the other way, consider a basic open set (A  B ) \ (U 0  V 0 ) in the subspace topology on A  B .
Then
( A  B ) \ (U 0  V 0 ) = (A \ U 0 )  (B \ V 0 ) :
This is a product of an open set in A and an open set in B , so it's open in the product topology on
A  B.
Since an arbitrary open set in the subspace topology on A  B is a union of basic open sets, it follows
that the subspace topology is contained in the product topology.
c 1999 by Bruce Ikenaga 6
 6-16-1999
Closed Sets and Limit Points
De nition. If X is a topological space, a subset A X is closed if X ; A is open.

Example. In the standard topology on R, closed intervals are closed and points are closed.
Note, however, than points need not be closed in an arbitrary topological space.
In Rn, the closed r-ball centered at x
B (x r) = fy 2 Rn j jx ; yj  rg
is closed.

Closed sets satisfy complementary properties relative to open sets.

Proposition. Let X be a topological space.
1. X and  are closed.
2. Finite unions of closed sets are closed.
3. Arbitrary intersections of closed sets are closed.
Proof. 1. X is closed, since X ; X =  is open.  is closed, since X ;  = X is open.
2. Let C1 : : : Cn be closed. For each i, write Ci = X ; Ui , where Ui is open. Then

X;
n
Ci = X ;
n
(X ; Ui ) =
\n (X ; (X ; Ui )) = \n Ui:
i=1 i=1 i=1 i=1
Tni S
Ui is open, because it's a nite intersection of open sets. Therefore, ni=1 Ci is closed.
=1

3. Let fCa ga2A be a family of closed sets. For each a 2 A, write Ca = X ; Ua , where Ua is open. Then
\ \  !
X ; Ca = X ; (X ; Ua ) = X ; X ; Ua = Ua :
a2A a 2A a2A a2A
Sa2A Ua is open, because it's a union of open sets. Therefore, Ta2A Ca is closed.
It's possible to construct a topology by specifying the closed sets rather than the open sets. Then you
can de ne the open sets to be the complements of closed sets.

 
Example. Give R the standard topology. The intervals ; n1 n1 for n = 1 2 : : : are open sets. However,
their intersection is
\1 ; 1 1  = f0g
n n
n=1
which is a closed subset of R.
1
  1
Likewise, the intervals n1 n ;
n are closed sets. However, their union is
1 1n ; 1  = (0 1)
n=1 n n
which is open in R.

Open sets in the subspace topology are intersections of the subspace with open sets in the big space.
The next result says that the same is true of closed sets.
Proposition. Let X be a topological space, let Y be a subspace of X , and let C Y . C is closed in Y if
and only if C = Y \ D, where D is closed in X .
Note that \closed in Y " means \closed in the subspace topology on Y ".
Proof. Let D be closed in X . Write D = X ; U , where U is open in X . Then
D \ Y = (X ; U ) \ Y = Y ; (Y \ U ):
(If x 2 (X ; U ) \ Y , then x 2 Y and x 2= U , so x 2= Y \ U . This means x 2 Y ; (Y \ U ). Conversely, if
x 2 Y ; (Y \ U ), then x 2 Y , and x 2= U since x 2= Y \ U . Therefore, x 2 X ; U , so x 2 (X ; U ) \ Y .)
Y \ U is open in Y , so Y ; (Y \ U ) is closed in Y .
Conversely, suppose C is closed in Y . Then Y ; C is open in Y , so Y ; C = Y \ U , where U is open in
X . Hence,
C = Y ; (Y ; C ) = Y ; (Y \ U ) = (X ; U ) \ Y:
X ; U is closed in X , so C is the intersection of a closed set in X with Y .

Proposition. Let X be a topological space, let Y be a subspace of X and let Z Y .

1. If Z is open in Y and Y is open in X , then Z is open in X .
2. If Z is closed in Y and Y is closed in X , then Z is closed in X .
Proof. I'll prove the rst assertion the proof of the second is similar.
Suppose Z is open in Y and Y is open in X . Z = Y \ U , where U is open in X . But Y is open in X ,
so Z is an intersection of two open sets of X | hence, it is open in X .
De nition. Let X be a topological space, and let Y X .
1. The closure of Y (denoted Y or cl(Y )) is the intersection of all closed sets containing Y .
2. The interior of Y (denoted Y_ or int(Y )) is the union of all open sets contained in Y .
Notice that the closed set X contains every subset Y , so the intersection formed in taking the closure
is not taken over an empty collection.
Remarks.
1. The closure of a set is closed, and the interior of a set is open.
2. The closure of a set is the smallest closed set containing the set likewise, the interior of a set is the
largest open set contained in the set.
3. A set Y is closed if and only if Y = Y  a set Y is open if and only if Y = Y_ .
2
Example. In the standard topology on R, the closure of (a b) is a b] and the interior of a b] is (a b).
More interestingly, Q = R. (A set whose closure is the whole space is said to be dense thus, the
rationals are dense in the reals.)

Example. The closure of even a single point can be quite large. For example, give Zthe topology whose
open sets are the open intervals (;n n) for n 2 Z+ , together with  and Z.
What is the closure of f1g? The only closed sets which contain 1 are Zand Z; (;1 1). Now (;1 1) =
f0 g, so the intersection of these two sets is Z; f0g. Thus, the closure of a single point turns out to be
everything in Zbut 0!

Lemma. Let X be a topological space, let Y be a subspace, and let Z Y . Then

clY Z = Y \ Z:
Here clY Z is the closure of Z in Y (with the subspace topology), whereas Z is the closure of Z in X .
Proof. clY Z is closed in Y , so clY Z = C \ Y , where C is closed in X . Now
Z clY Z = C \ Y C
so C is a closed set in X containing Z .
Therefore, Z C , so
Z \ Y C \ Y = clY Z:
Conversely, Z is closed in X and contains Z , so Z \ Y is closed set in Y containing Z . Therefore,
clY Z Z \ Y .
Hence, clY Z = Y \ Z .
Terminology.
1. If X is a topological space and x 2 X , a neighborhood of x is any open set containing x.
Some authors use \neighborhood" to mean a set containing x in its interior.
2. If X and Y are sets, I'll say that X intersects Y or X meets Y if X \ Y 6= .
Lemma. Let X be a topological space, and let Y X . x 2 Y if and only if every neighborhood of x
intersects Y .
Proof. Suppose x 2 Y . Let U be a neighborhood of x. Suppose on the contrary that U \ Y = . Then
U X ; Y , so X ; U  X ; (X ; Y ) = Y .
Now X ; U is a closed set containing Y , so X ; U  Y . But then x 2 Y X ; U implies x 2= U ,
contrary to the assumption that U was a neighborhood of x.
Conversely, suppose every neighborhood of x intersects y. Let C be a closed set containing Y . I must
show that x 2 C  since C is arbitrary, this will prove that x 2 Y .
Suppose on the contrary that x 2= C . Y C implies X ; Y  X ; C | that is, X ; C does not intersect
Y . Further, x 2 X ; C , and X ; C is open. Thus, X ; C is a neighborhood of x which does not meet Y ,
contrary to assumption. Therefore, x 2 C , which is what I wanted to show.
Remark. If B is a basis for the topology on X , then x 2 Y if and only if every B 2 B intersects Y .
In a sense, the closure of a set consists of the original set together with other points that are \close
to" the original set | you might picture these additional points as \boundary points". I want to make the
notion of a \boundary point" precise.
3
De nition. Let X be a topological space, Y X . A point x 2 X is a limit point of Y is every neighborhood
of x intersects Y in some point other than x.
Example. In the standard topology on R, every point of a b] is a limit point of (a b).
On the other hand, 0 is not a limit point of (1 2). For example, the open set (;0:5 0:5) contains 0, but
 1 (1 2).+
does not intersect
The set n n 2 Z has 0 as its only limit point.
Zhas no limit points in R. About every point in R; Z, you can construct an open interval which does
not contain any integers. And if n 2 Z, the open interval (n ; 0:5 n + 0:5) intersects Zin the point n.
Every real number is a limit point of Q, since every open interval about a real number contains rational
numbers.

De nition. If Y is a subset of a topological space X , the derived set Y 0 is the set of limit points of Y .
Proposition. If Y is a subset of a topological space X , then
Y = A  A0 :
Proof. Suppose x 2 Y  Y 0.
If x 2 Y , then x 2 Y .
If x 2= Y , then x 2 Y 0 , so every neighborhood of x meets Y | in a point other than x, but this fact
isn't relevant. Since every neighborhood of x meets Y , an earlier result implies that x 2 Y .
Conversely, suppose that x 2 Y . If x 2 Y , then x 2 Y  Y 0, and I'm done.
Suppose then that x 2= Y . Since x 2 Y , every neighborhood of x intersects Y . But the intersection can't
consist of x, since x 2= Y . Therefore, every neighborhood of x intersects Y in a point other than x. Hence,
x 2 Y 0, and certainly x 2 Y  Y 0.
Corollary. Let X be a topological space, and let Y X . Then Y is closed if and only if Y 0 Y .
Proof. If Y is closed, then Y = Y = Y  Y 0, so y0 Y .
Conversely, if Y 0 Y , then Y  Y 0 = Y , so Y = Y , and Y is closed.
The last result is often expressed by saying that a closed set contains all its limit points. Thus, you
often show a set is closed by taking a limit point and showing that it lies in the set.

c 1999 by Bruce Ikenaga 4

 6-17-1999
Hausdor Spaces
If you try to study topological spaces without any conditions, the variety becomes overwhelming |
there is little you can say about topological spaces in general. In order to say more, you have to assume
more. The Hausdor condition is a common assumption mathematicians make about topological spaces.
De nition. A topological space is Hausdor (or 2 )if for all 2 , 6= , there are neighborhoods
X T x y X x y

U of and of such that \ = .

x V y U V

y
V

U
x

Remark. An easy induction shows that if is Hausdor and 1X x ::: x n 2 X , then there are neighborhoods
U1 of 1,
x , n of n, such that the 's are pairwise disjoint.
::: U x U

Example. R is Hausdor in the standard topology.

Let and be distinct points in R. Let = 12 j ; j. Then ( ;
x y  x y x  x + ) and ( ;
 y  y + ) are disjoint


open sets containing and , respectively.

x y

ε ε ε ε

x y

Example. Consider the topology on X = f1 2 3g in which the open sets are , f1 2g, and f1 2 3g.

1 2 3

X is not Hausdor , since it isn't possible to nd disjoint neighborhoods containing 1 and 2.

Proposition. The product of Hausdor spaces is Hausdor .

Proof. Let and be Hausdor spaces. Let ( 1 1) ( 2 2 ) 2  , and assume that ( 1 1 ) 6= ( 2 2).
X Y x y x y X Y x y x y

I want to nd disjoint neighborhoods in  containing the points.

X Y

Since the points aren't the same, they must di er in at least one coordinate. Assume 1 =
6 2 (the x x

argument is similar if instead 1 6= 2.y y

1
 Since is Hausdor , I can nd disjoint neighborhoods 1 of 1 and 2 of 2. Then
X U x U x U1  Y and U2  Y
are disjoint neighborhoods in  , ( 1 1 ) 2 1  , and ( 2 2 ) 2 2  .
X Y x y U Y x y U Y

(x1 , y1 )
y1

(x2 , y2 )
y2

U1 x1 U2 x2 X

Therefore,  is Hausdor .
X Y

Proposition. A subspace of a Hausdor space is Hausdor .

Proof. Let be Hausdor , and let be a subspace of . Let 1 2 2 , where 1 6= 2 .
X Y X y y Y y y

Since is Hausdor , there are disjoint neighborhoods 1 of 1 and 2 of 2 in . Then 1 \ is a

X U y U y X U Y

neighborhood of 1 in and 2 \ is a neighborhood of 2 in , and 1 \ and 2 \ are disjoint.

y Y U Y y Y U Y U Y

Therefore, is Hausdor . Y

Proposition. If is Hausdor and 1

X n 2 , then f 1 ng is closed.
x : : :x X x ::: x

Proof. First, I'll show that a single point is closed. Let 2 . I claim that ; f g is open. x X X x

Take 2 ; f g. Since 6= , I may nd disjoint neighborhoods of and of . Since doesn't

y X x x y U x V y V

intersect and 2 , it follows that  ; f g. Thus, is a neighborhood of contained in ; f g.

U x U V X x V y X x

Since was arbitrary, ; f g is open.

y X x

Since a nite union of closed sets is closed, if 1 n 2 , then f 1 ng is closed. x : : :x X x ::: x

If is a limit point of a set , then every neighborhood of meets in a point other than . In a
y Y y Y y

Hausdor space, every neighborhood of a limit point meets the set in in nitely many points.
Proposition. Let be Hausdor , and let  . is a limit point of if and only if every neighborhood
X Y X x Y

of meets in innitely many points.

x Y

Proof. If every neighborhood of meets in innitely many points, then surely every neighborhood meets
x Y

Y in a point other than . Therefore, is a limit point of .

Y x Y

Conversely, suppose is a limit point of . Let be a neighborhood of . I must show that meets
x Y U x U

Y in innitely many points.

Suppose meets in only nitely many points. Then meets ; f g in only nitely many points
U Y U Y x

x 1 ::: n. (Note that must meet ; f g in at least one point, since is a limit point of .)
x U Y x x Y

Now ; f 1 X ng is open and contains , so \ ( ; f 1

x ::: x ng) is open and contains . But
x U X x ::: x x

in forming this set I've thrown out the points 1 n where meets , so it follows that \ ( ;
x ::: x U Y U X

f 1
x :::ng) does not meet at all! This contradicts the fact that every neighborhood of must intersect
x Y x

Y .
Hence, intersects in innitely many points.
U Y


c 1999 by Bruce Ikenaga 2
 6-21-1999
Continuous Functions
For a function f : R ! R, f is continuous at x = a if for every > 0, there is a  > 0 such that if
> x j ; aj > 0 then j ( ) ; f (a)j:
> f x

Interpreted in terms of intervals | i.e. basic open sets in R | this says that for any interval I containing
( ), there is an interval J containing a, such that f maps J into I .
f a
This motivates the following de nition.
De nition. Let f : X ! Y be a function, where X and Y are topological spaces. f is continuous at x 2 X
if for every neighborhood V of f (x), there is a neighborhood U of x such that f (U )  V .
V
U
x f
f(x) f(U)

f is continuous (on X ) if it is continuous at every point of X .

The de nition I've given is often called the pointwise de nition of continuity. I'll give the equivalent
global form shortly.

Example. Any function f : Rn ! Rm that is continuous in the - sense is continuous according to the
de nition above.

Example. Let X be a topological space. The identity function id : X ! X is continuous.

To prove this, let x 2 X and let V be a neighborhood of id(x) = x. Then V is a neighborhood of x
which satis es id(V )  V .

Example. Let X and Y be topological spaces, and let y 2 Y . The constant function cy : X ! Y de ned by
cy (x) = y for all x 2 X is continuous.
To see this, suppose x 2 X and V is a neighborhood of cy (x) = y. X is a neighborhood of x, and
cy (X ) = fyg  V . Therefore, cy is continuous at x.

Example. Let Y be a subspace of the topological space X . The inclusion map i : Y ,! X given by i(y) = y
for all y 2 Y is continuous.
To prove this, let y 2 Y and let V be a neighborhood of i(y) = y in X . Since V is open in X , V \ Y is
open in Y , and y 2 V \ Y . Clearly, i(V \ Y )  V , so V \ Y is a neighborhood of y in Y which i maps into
V . Therefore,  is continuous at y.

Example. The composite of continuous functions is continuous.

To be speci c, suppose that X , Y , and Z are topological spaces, f : X ! Y is continuous at x, and
g : Y ! Z is continuous at f (x). Then g  f : X ! Z is continuous at x.
Thus, if f is continuous on X and g is continuous on Y , then g  f is continuous on X .
1
 To prove the result at a point, assume that f : X ! Y is continuous at x and g : Y ! Z is continuous
at f (x). Let W be a neighborhood of g(f (x)) in Z . By continuity of g at f (x), there is a neighborhood V
of f (x) such that g(V )  W .
By continuity of f at x, there is a neighborhood U of x such that f (U )  V .
U
x
f
V

f(x)

f(U)
g
W

f(V)
g(f(U))

g(f(x))

Then g(f (U ))  g(V )  W . so g  f is continuous at x.

Example. The restriction of a continuous function is continuous.

Speci cally, let f : X ! Y be a continuous function between topological spaces. Let U be a subspace
of X . Then f jU : U ! Y de ned by f jU (x) = f (x) for all x 2 U is continuous.
The result is immediate from the last two examples, since f jU = f  i, where i : U ,! X is the inclusion
map, and the composite of continuous functions is continuous.

Example. If f : X ! Y is a continuous function between topological spaces and Y is a subspace of Z , then

the function f~ : X ! Z de ned by f~(x) = f (x) for all x 2 X is continuous. (f~ is said to be obtained by
expanding the range.)
This result is also easy, since f~ = i  f , where i : Y ,! Z is the inclusion map and the composite of
continuous functions is continuous.

Next, I'll show that the pointwise de nition of continuity is equivalent to the following global de nition.
Proposition. Let f : X ! Y be a function between topological spaces. f is continuous if and only if for
every open set V in Y , f ;1 (V ) is open in X .
This is expressed more concisely by saying that the inverse image of an open set is open.
Proof. Suppose f is continuous, and let V be open in Y . I want to show that f ;1 (V ) is open in X .
Let x 2 f ;1 (V ). I want to nd a neighborhood of x contained in f ;1 (V ). Now f (x) 2 V , so by
continuity, I can nd a neighborhood U of x such that f (U )  V . Then U  f ;1 (V ), and U is the desired
neighborhood of x. This proves that f ;1 (V ) is open.
Conversely, suppose that the inverse image of an open set is open. I want to show that f is continuous.
Let x 2 X , and let V be a neighborhood of f (x). I want to nd a neighborhood U of x such that f (U )  V .
By assumption, f ;1 (V ) is open. Since f (x) 2 V , it follows that x 2 f ;1 (V ). Moreover, f (f ;1 (V ))  V .
Thus, f ;1 (V ) is a neighborhood of x that f maps into V . Hence, f is continuous.
I believe that the name of the next result is due to James R. Munkres it's an enormously useful tool
for constructing continuous functions.
Lemma. (The Pasting Lemma) Let X and Y be topological spaces, and let U and V be open sets in X
2
such that X = U  V . Suppose that f : U ! Y and g : V ! Y are continuous, and
f (x) = g(x) for all x 2 U \ V:

Then the function h : X ! Y de ned by

f (x) if x 2 U
h(x) =
g(x) if x 2 V

is continuous.
Proof. Let W be open in Y . Then
h (W ) = f (W )  g (W ):
;1 ;1 ;1
By continuity, f ;1 (W ) is open in U  since U is open in X , f ;1 (W ) is open in X .
By continuity, g;1 (W ) is open in V  since V is open in X , g;1 (W ) is open in X .
Therefore, f ;1 (W )  g;1 (W ) is open in X , so h;1(W ) is open in X . By the preceding result, h is
continuous.
An analogous result holds if U and V are closed the proof follows from the next result, which says that
you can also express continuity in terms of closed sets.
Proposition. Let f : X ! Y be a function between topological spaces. The following statements are
equivalent:
1. f is continuous.
2. For all U  X , f (U )  f (U ).
3. The inverse image of a closed set is closed.
Proof. (1 ) 2) Suppose f is continuous, and let U  X . Since U = U  U 0 , I have
f (U ) = f (U  U 0 ) = f (U )  f (U 0 ):

Clearly, f (U )  f (U ) I need to show that f (U 0 )  f (U ).

Let x be a limit point of U , and let V be a neighborhood of f (x). I must show that V intersects f (U ).
Since f ;1 (V ) is a neighborhood of x and x is a limit point of U , there is a y 2 f ;1 (V ) \ U such that y 6= x.
Then ; ;1 
f (y) 2 f f (V ) \ U  f (f (V )) \ f (U )  V \ f (U ):
;1
This proves that V intersects f (U ), so f (x) 2 f (U ). Thus, f (U 0 )  f (U ), and so f (U )  f (U 0 )  f (U ).
Therefore, f (U )  f (U ).
(2 ) 3) Suppose that for all U  X , f (U )  f (U ). I want to show that the inverse image of a closed set is
closed.
Let C be closed in Y . I want to show that f ;1 (C ) is closed in X . Now
 
f f ;1 (C )  f (f ;1 (C ))  C = C
; 
where the rst inclusion follows from statement 2, the second inclusion follows from f f ;1 (C )  C , and
the last equality follows from the fact that C is closed.
Thus, f ;1 (C )  f ;1 (C ). But clearly, f ;1 (C )  f ;1 (C ), so f ;1 (C ) = f ;1 (C ), and f ;1 (C ) is closed.
(3 ) 1) Suppose the inverse image of a closed set is closed. I want to show that f is continuous.
Let V be an open subset of Y . Then Y ; V is closed, so by assumption f ;1 (Y ; V ) is closed. But
f (Y ; V ) = f (Y ) ; f (V ) = X ; f (V ):
;1 ;1 ;1 ;1
Therefore, X ; f ;1 (V ) is closed, so f ;1 (V ) is open. Therefore, f is continuous.

c 1999 by Bruce Ikenaga 3

 6-22-1999
Homeomorphisms
De nition. Let X and Y be topological spaces, and let f : X ! Y . f is a homeomorphism if f is bijective,
;
and both f and f 1 are continuous.
If there is a homeomorphism between topological spaces X and Y , then X and Y are homeomorphic.
Homeomorphic spaces are \the same" as topological spaces, in the same way that isomorphic groups
are \the same" as groups. The big problem in topology | which is much too di cult to deal with in this
generality | is to classify all spaces up to homeomorphism.
The following properties are obvious:
1. The identity map of a topological space is a homeomorphism.
2. The composite of homeomorphisms is a homeomorphism.
3. The inverse of a homeomorphism is a homeomorphism.

Example. De ne f : R ! R by
f(x) = x + k
where k 2 R. Then f is a homeomorphism.
It's clear that f is continuous, since the inverse image of an open interval is an open interval | it's just
a translate of the original interval. Then the inverse function f ;1 : R ! R given by f ;1 (x) = x ; k is also
continuous, since it can be written as f ;1 (x) = x + (;k).
De ne g : R ! R by
g(x) = rx
where r 2 R and r 6= 0. Then g is a homeomorphism.
Again, it's clear that g is continuous, since the inverse image of an open interval is an open interval.
Replacing r with r1 shows that the inverse g;1 : R ! R is continuous.
From these facts, it follows that any two nonempty open intervals in R (with the subspace topologies)
are homeomorphic. For if (a b) is a nonempty open interval, then:
1. f(x) = x ; a maps (a b) to (0 b ; a).
2. g(x) = b ;x a maps (0 b ; a) to (0 1).
f and g are homeomorphisms, by observations above. This proves that every open interval is homeo-
morphic to (0 1). But homeomorphism is transitive (since the composite of homeomorphisms is a homeo-
morphism), so any two nonempty
open intervals are homeomorphic.

The function tan : ;
2
2 ! R is continuous, and its inverse arctan : R ! ;
2
2 is continuous.



Therefore, ;
2
2 is homeomorphic to R. But every nonempty open interval is homeomorphic to ;
2
2 ,
so it follows that every nonempty open interval is homeomorphic to R.

De nition. Let f : X ! Y be a function between topological spaces. f is open if f takes open sets to open
sets. Likewise, f is closed if f takes closed sets to closed sets.
Lemma. Homeomorphisms are open maps and closed maps.
Proof. I'll do the open map case the closed map case is similar.
1
 Let f : X ! Y be a homeomorphism, and let U be open in X. I want to show that f(U) is open in Y .
Since f : X ! Y is a homeomorphism, it has a continuous inverse f ;1 . By continuity, (f ;1 );1 (U) is
open in Y . But (f ;1 );1 (U) = f(U), so f(U) is open in Y .

Example. You can tell that two spaces are not homeomorphic if you can nd a topological property that one
has that the other does not. In a sense, this is a circular observation: A topological property is a property
preserved by homeomorphisms. So this observation becomes useful as you develop a repertoire of properties
that are preserved by homeomorphisms.
For example, the circle n  o
S 1 = (x y) 2 R2  x2 + y2 = 1
with the subspace topology is not homeomorphic to the interval (0 1) in R with the subspace topology. One
reason: Removing a point from (0 1) leaves a disjoint union of two sets open in (0 1), but removing a point
from S 1 does not leave a disjoint union of two sets open in S 1 .
Take for granted that the assertion about S 1 is true. To show that this dierence implies that the
spaces aren't homeomorphic, suppose f : (0 1) ! S 1 is a homeomorphism. Take x 2 (0 1), so (0 1) ;fxg =
(0 x)  (x 1), a disjoint union of open subsets of (0 1).
Since f is a homemorphism, f(0 x)] and f(x 1)] are disjoint open subsets of S 1 , and S 1 ; ff(x)g =
f(0 x)]  f(x 1)]. This contradicts my claim about S 1 , so S 1 and (0 1) aren't homeomorphic.
The assertion about S 1 will follow from results I'll prove later about connected set.

Example. Removing a point from R leaves a disjoint union of two open sets. The same is not true when a
point is removed from Rn for n  2. Therefore, R is not homeomorphic to Rn for n  2.
The fact that Rm and Rn aren't homeomorphic for m 6= n is a fact called invariance of domain the
best proofs use methods from algebraic topology.

The standard informal example of homeomorphic spaces is that a coee cup is homeomorphic to a
doughnut. Popular accounts often say that two spaces are homeomorphic if you can imagine deforming one
into the other with cutting, tearing, or breaking.

One of these things is not e

likthe others.

In the picture above, the three solid objects are homeomorphic the object shaped like a ring is not
homeomorphic to the others. This kind of assertion (like the one about the coee cup and the doughnut)
must be taken in the right spirit. No one would set out to prove the assertion by writing down speci c
homeomorphisms!
The notion of homeomorphism does not include the idea of \deforming" one thing into another. In the
rst place, a deformation requires a \big space" inside of which the deformation takes place. In the second
place, a deformation is something which takes places in time. Neither of these notions is part of the de nition
2
of a homeomorphism. The notion of isotopy comes closer to the idea of a continuous deformation of one
space into another.
Homeomorphism is also independent of the embedding of a space in a larger space. For example, the
trefoil knot on the left and the circle on the right are homeomorphic as topological spaces:

In order to get a hold of the obvious dierence that exists between the two, you need to consider the
complements of these sets in R3. In this case, the ambient space R3 is responsible for the \knottedness" of
the trefoil.

c 1999 by Bruce Ikenaga 3

 6-23-1999
The Product Topology
If X and Y are topological spaces, the product topology on X Y is the topology generated by the
basis consisting of all sets U V , where U is open in X and V is open in Y . This may be extended to
a product of nitely many spaces in the obvious fashion. How do you dene a topology on a product of
innitely many spaces?
Q
De nition. Let fXa ga2A be a family of topological
Q spaces. The box topology on a2A Xa is the topology
generated by the basis consisting of sets a2A Ua , where Ua is open in Xa .
Q
Let B be the collection
Q
of sets a2A Ua , where UaQis open in Xa . I want to show that B is a basis.
Let (xa )a2A 2 a2A Xa . Since Xa is open in Xa , a2A Xa is an element of B, and it obviously contains
(xa )a2A . ;Q Q

Let (xa )a2AQ2 a2A Ua \ a2A Va , where Ua and Va are open in Xa for each a. Then Ua Va is
open in Xa , so a2A (Ua \ Va ) is an element of B. Moreover,
 !
Y Y Y
(xa )a2A 2 (Ua \ Va )  Ua \ Va :
a2A a2A a2A
This proves that B is a basis.
There is another way to generate a topology on a product. I'll need some preliminary results.
T
Lemma. If fTaga2A is a family of topologies on a set X , then T = a2A Ta is a topology on X .
Proof. Since 
X 2 Ta for all a, it follows that 
X 2 T .
If fUiSgi2I is a collection of open sets S in T , then for each a, I have Ui 2 Ta for all i. Since Ta is a
topology, i2I Ui 2 Ta for all a. Hence, i2I Ui 2 T . Thus, T is closed under arbitrary unions.
If fU1
: : :
Ung is a collection of open sets in T , then for each a, I have Ui 2 Ta for all i = 1
: : :
n. Since
Ta is a topology, Tni=1 Ui 2 Ta for all a. Hence, Tni=1 Ui 2 T . Thus, T is closed under nite intersections.
Therefore, T is a topology.
De nition. Let fXa ga2A be a family of topological spaces, let X be a set, and let ffa : X ! Xa j a 2 Ag
be a family of functions. The topology induced on X by the family ffa ga2A is the smallest topology
on X which makes all the fa 's continuous.
There is at least one topology on X which makes all the fa 's continuous, namely the discrete topology.
If I intersect all the topologies on X which make the fa 's continuous, I get a topology (by the last lemma)
which makes all the fa 's continuous. It is clearly the smallest such topology, in the sense that it's contained
in every topology which makes all the fa 's continuous.
While this construction accomplishes the goal of making the fa 's continuous, it would be nice to have
a more concrete description of the open sets.
Lemma. Let fXa ga2A be a family of topological spaces, let X be a set, and let ffa : X ! Xa j a 2 Ag be
a family of functions. The topology induced on X by the fa 's is the same as the topology generated by the
subbasis n open o
fa;1 (U ) j U  Xa
a 2 A :
Proof. First, this collection is a subbasis: If x 2 X , x a 2 A and consider fa (x) 2 Xa . Then Xa is open in
Xa and fa;1 (Xa ) is an element of the collection which contains x.
Now let S be the topology induced on X by the fa 's and let T be the topology generated by the subbasis.
If U is open in Xa and fa;1 (U ) is an element of the subbasis, then fa;1 (U ) is open in T . Since U and
fa were arbitrary, this shows that T makes all the fa 's continuous. But S is the smallest topology which
makes the fa 's continuous. Hence, S  T .
1
 Conversely, let U be open in Xa and let fa;1 (U ) be an element of the subbasis. I claim that fa;1 (U ) is
an open set in S . For if not, then fa is not continuous, since the inverse image under fa of the open set U
isn't open in X .
Now I know that all the elements of the subbasis are contained in S . But S is a topology, so arbitrary
unions of nite intersections of elements of the subbasis are also contained in S . Therefore, T  S .
Hence, the two topologies are the same.
Here's an important property of this topology.
Proposition. Let fXa ga2A be a family of topological spaces, let X be a set, and let ffa : X ! Xa j a 2 Ag
be a family of functions. Give X the topology induced on X by the fa 's, and let W be a topological space.
A function g : W ! X is continuous if and only if fa  g : W ! Xa is continuous for all a.
Proof. If g is continuous, then fa  g : W ! Xa is continuous for all a, since fa  g is a composite of continuous
maps.
Conversely, suppose fa  g : W ! Xa is continuous for all a. I want to show that g is continuous. Let
U be open in X  I must show that g;1 (U ) is open in W .
To begin with, take U to be a subbasic open set. Thus, suppose U = fa;1 (V ), where V is open in Xa .
Then
g;1(U ) = g;1 (fa;1 (V )) = (fa  g);1 (V ):
Since fa  g is continuous, (fa  g);1 (V ) is open in W .
For a general open set U , simply write U as a union of a nite intersections of subbasic open sets, then
use the fact that g;1 (;) commutes with unions and intersections.
Q
Now specialize to the case where X = a2A Xa , and give X the topology induced by the family of
projection maps ( )
Y
b : Xa ! Xb j b 2 A :
a2A
The topology generated in this way is called the product topology it is the smallest topology which
makes all the projection maps continuous.
Notice that I've already used the name \product topology" in the case of a product of nitely many
spaces to mean the topology generated by the basis consisting of products of open sets from the factors. In
fact, if there are nitely many factors, these two topologies are the same.
To see this, it's helpful to get an explicit description for the basic open sets in the product topology. A
typical basis element has the form
a;11 (U1 ) \ a;21 (U2 ) \ : : : \ a;1 (Un )

n

where Ui is open in Xa . Note that a;1 (U ) \ a;1 (V ) = a;1 (U \ V ), so I can combine terms with the same
i
index. Thus, I may as well assume that a1 , a2, : : : , an are distinct.
In this case, the set above is

Y
Va
where Va = X a if a 6= a1
a2
: : :
an
a2A
Ui if a = ai , i = 1
2
: : :
n :
Thus, a typical basis element for the product topology is a product of sets open in the factors, where all
but
nitely many of these sets are the whole spaces. In the case where there are only nitely many factors,
this is simply products of sets open in the factors { the same as the product topology I dened earlier.
The next result says that a function into a product is continuous if and only if its factors are continuous.
Q
Corollary. Let fXa ga2A be a family of topological
Q spaces, and give a2A Xa the product topology. Let W
be a topological space, and let g : W ! a2A Xa be a function. g is continuous if and only if a  g is
continuous for all a.
2
Example. Dene f : R ! R3 by
f (t) = (cos t
sin t
t):
In this case, the projections are
(1  f )(t) = cos t
(2  f )(t) = sin t
(3  f )(t) = t:
Since these are continuous as functions R ! R, it follows from the corollary that f is continuous as well.

A nal note: A basis element in the box topology is a product of sets open in the factors. Since this is
more general than the condition for basis elements of the product topology, it follows that the box topology
is
ner than the product topology.

c 1999 by Bruce Ikenaga 3

 7-17-1999
Metric Spaces
De nition. If X is a set, a metric on X is a function d : X X ! R such that:
1. d(x y)  0 for all x y 2 X d(x y) = 0 if and only if x = y.
2. d(x y) = d(y x) for all x y 2 X .
3. (Triangle Inequality) For all x y z 2 X ,
d(x y) + d(y z )  d(x z ):
Lemma. If X is a set with a metric, the collection of open balls
B (x ) = fy 2 X j d(x y) < g
is a basis.
Proof. If x 2 X , then x 2 B (x 1).
Suppose B (x1 1) and B (x2 2) are open balls. Let x 2 B (x1 1) \ B (x2 2).
Let
 = min(1 ; d(x x1) 2 ; d(x x2)):
Then x 2 B (x )  B (x1 1) \ B (x2 2). Therefore, the collection of open balls forms a basis.
De nition. If X is a set with a metric, the metric topology on X is the topology generated by the basis
consisting of open balls B (x ), where x 2 X and  > 0. A metric space consists of a set X together with
a metric d, where X is given the metric topology induced by d.
Remark. In generating a metric topology, it suces to consider balls of rational radius.

Example. The usual metric on Rn is dened by

X
n
!1=2
d(x y) = (xi ; yi ) 2 :
i=1

where x = (x1 : : : xn) and y = (y1 : : : yn).

It is clear that d(x y)  0 and that d(x x) = 0 for all x y 2 Rn . If x = (x1 : : : xn) and y = (y1 : : : yn )
and d(x y) = 0, then !
X
n 1=2
X
n
(xi ; yi )2 = 0 so (xi ; yi )2 = 0:
i=1 i=1
This is only possible if (xi ; yi ) = 0 for all i, and this in turn implies that xi = yi for all i. Therefore,
2
x = y.
It is obvious that d(x y) = d(y x) for all x y 2 Rn.
Note that d(x y) = jx ; yj, where j  j is the standard norm which gives the length of a vector. Now
juj2 = u  u, where  denotes the dot product in Rn. By standard properties of the dot product,
ju + vj2 = (u + v)  (u + v) = u  u + 2u  v + v  v = juj2 + 2u  v + jvj2  juj2 + 2jujjvj + jvj2 = (juj + jvj)2:
(The inequality follows from the Schwarz inequality ju  vj  jujjvj.) Then
ju + vj  juj + jvj:
1
 Now let u = x ; y and v = y ; z . Then
jx ; z j  jx ; yj + jy ; z j or d(x z )  d(x y) + d(y z ):
Here is a proof of the Schwarz inequality in case you ahven't seen it. Given u = (u1 : : : un) v =
(v1 : : : vn) 2 Rn , I want to show that ju  vj  jujjvj I'll show that ju  vj2  juj2jvj2 , and the result will
follow by taking square roots.
Set A = juj2, C = jvj2, and B = ju  vj. I want to show that B 2  AC .
If A = 0, then u = 0, and the result is obvious. Assume then that A > 0. For all x 2 R,
X
n
(uix + vi )2  0
i=1
X
n X
n X
n
x2
u2i + 2x ui vi + vi2  0
i=1 i=1 i=1
x A + 2xB + C  0
2

Take x = ; BA . The last inequality yields

B2  A ; 2 B  B + C  0
A2 A
; BA + C  0
2

AC  C
This completes the proof of the Schwarz inequality.
Thus, the standard metric on Rn satises the axioms for a metric. Obviously, the metric topology is
just the standard topology.

Lemma. (Comparison Lemma for Metric Topologies) Let d and d be metrics on X inducing topologies
0

T and T . T is ner than T if and only if for all x 2 X and  > 0, there is a  > 0 such that Bd (x  ) 
0 0

Bd (x ).
0

Proof. Suppose rst that T  T . Let x 2 X , and let  > 0. Bd (x ) is open in T , so it's open in T . Since
0
0
0

the open d-balls form a basis for T , there is an open ball Bd (x  ) such that
x 2 Bd (x  )  Bd (x ): 0

Conversely, suppose that for all x 2 X and  > 0, there is a  > 0 such that Bd (x  )  Bd (x ). I want
0

to show that T  T .
0

Let U be open in T . I want to show that it's open in T . Let x 2 U . Since the d -balls form a basis for
0 0

T , there is an  > 0 such that

0

x 2 Bd (x )  U:0

By assumption, there is a  > 0 such that

x 2 Bd (x  )  Bd (x ): 0

Therefore, x 2 Bd (x  )  U .
Now Bd (x  ) is a T -open set containing x and contained in U . Since x 2 U was arbitrary, U is open in
T . Therefore, T  T . 0

2
 The standard metric on Rn is unbounded, in the sense that you can nd pairs of points which are
arbitrarily far apart. However, you can always replace a metric with a bounded metric which gives the same
topology.
De nition. If (X d) is a metric space and Y  X , then Y is bounded if there is an M 2 R such that
d(x y)  M for all x y 2 Y:
Lemma. Let X be a metric space with metric d. Dene
d(x y) = min(d(x y) 1):
1. d is a metric.
2. d and d induce the same topology on X .
Proof. 1. Let x y 2 X . Since d(x y)  0, d(x y) = min(d(x y) 1)  0, and
d(x x) = min(d(x x) 1) = min(0 1) = 0:
If d(x y) = min(d(x y) 1) = 0, then d(x y) = 0, so x = y. This shows that the rst metric axiom holds.
Since d(x y) = min(d(x y) 1) = min(d(y x) 1) = d(y x), the second metric axiom holds.
To verify the third axiom, take x y z 2 X . Begin by noting that if either d(x y)  1 or d(y z )  1,
then d(x y) = 1 or d(y z ) = 1. Therefore,
d(x y) + d(y z )  1  d(x z ):
Assume that d(x y) < 1 and d(y z ) < 1. Then
d(x y) + d(y z ) = d(x y) + d(y z )  d(x z )  d(x z ):
This veries the third axiom, so d is a metric.
2. Observe that for 0 <  < 1, Bd (x ) = Bd (x ). The idea is to apply the Comparison Lemma, shrinking
balls if necessary to make their radii less than 1.
Let x 2 X and let  > 0.
If  < 1, then x 2 Bd (x ) = Bd (x ).
If   1, then
x 2 Bd (x 0:5) = Bd (x 0:5)  Bd (x ):
0

Therefore, the d-topology is ner than the d -topology. The other inclusion follows by simply swapping
0

the d's and d 's.

0

It follows that boundedness is not a topological notion, since every subset is bounded in the standard
bounded metric.

Example. The square metric on Rn is given by

(x y) = max(jx1 ; y1 j : : : jxn ; yn j):
Relative to this metric, B (x ) is an n-cube centered at x with sides of length 2.
First, I'll show that  is a metric. Let x = (x1 : : : xn) y = (y1 : : : yn ) 2 Rn.
Clearly, (x y)  0 and (x x) = 0. If (x y) = 0, then jxi ; yi j = 0 for all i, so x = y.
It's also obvious that (x y) = (y x).
3
 If x y z 2 Rn, then for each j ,
(x y) + (y z ) = max
i
fjxi ; yi jg + max
i
fjyi ; zi jg  jxj ; yj j + jyj ; zj j  jxj ; zj j:
Therefore,
(x y) + (y z )  max
j
fjxj ; zj jg = (x z ):
Thus,  is a metric.

Lemma.  induces the same topology on Rn as the standard metric.

Proof. The idea of the proof is depicted below.

Note that
X !1=2
;  n
(x y) = max
i
fjxi ; yi jg = jxj ; yj j = (xj ; yj )2 1=2
(xj ; yj )2 = d(x y)
i=1

X !1=2  1=2 p ;
n
 p
d(x y) = (xj ; yj )
2
 n  max
i
f( x i ; yi ) g
2
= n (x y)2 1=2 = n(x y):
i=1
These inequalities may be used to get -balls contained in d-balls and d-balls contained in -balls by
the Comparison Lemma, this shows that the topologies are the same.
Lemma. The square metric induces the product topology on Rn.
Proof. If x = (x1 : : : xn) 2 Rn , then
B (x ) = (x1 ;  x1 + ) (x2 ;  x1 + )    (xn ;  xn + ):
The set on the right is open in the product topology. Since the square metric-basic sets are open in the
product topology, any square metric-open set is open in the product topology.
Conversely, let
U = (a1 b1) (a2 b2) : : : (an bn):
It is easy to check that sets of this form comprise a basis for the product topology.
Let x = (x1 : : : xn) 2 U , so ai < xi < bi for all i. Dene
 = min
i
fxi ; ai bi ; xig:
Then
(xi ;  xi + )  (ai bi) for all i:
So Y Y
B (x ) = (xi ;  xi + )  (ai bi) = U:
i i

4
 It follows that U is open in the square metric topology. Since the product topology basic sets are open
in the square metric topology, any product topology open set is open in the square metric topology.
Lemma. Metric topologies are Hausdor.
 
Proof. Let (X d) be a metric space, and let x and y be distinct points of X . Let  = d(x y). Then B x 2
 
and B y 2 are disjoint open sets in the metric topology which contain x and y, respectively.
Lemma. If (X d), (Y d ) are metric spaces, the - denition of continuity is valid. That is, a map f : X ! Y
0

is continuous at x 2 X if and only if for every  > 0, there is a  > 0 such that if  > d(x y) implies that
 > d (f (x) f (y)).
0

Proof. First, suppose that f : X ! Y is continuous at x 2 X . Let  > 0, and consider the ball B (f (x) ).
Since this is an open set containing f (x), continuity implies that there is a  > 0 such that
f (B (x  ))  B (f (x) ):
Now consider the conclusion to be established. Suppose y 2 X satises  > d(x y). Then y 2 B (x  ),
so f (y) 2 f (B (x  )). Therefore, f (y) 2 B (f (x) ), so  > d (f (x) f (y)).
0

Conversely, suppose that for every  > 0, there is a  > 0 such that if  > d(x y) implies that
 > d (f (x) f (y)). I want to show that f is continuous.
0

Let x 2 X , and let V be an open set in Y containing f (x). I want to nd a neighborhood U of x such
that f (U )  V .
Since the -balls form a basis for the metric topology, I may nd an  > 0 such that f (x) 2 B (f (x)   V .
By assumption, there is a  > 0 such that if  > d(x y) implies that  > d (f (x) f (y)).
0

Now consider the ball B (x  ). This is an open set containing x. If y 2 B (x  ), then  > d(x y) > 0.
Therefore,  > d (f (x) f (y)), so f (y) 2 B (f (x) ). This shows that f (B (x  ))  B (f (x) )  V , so f is
0

continuous.
De nition. If X is a set, a sequence in X is a function x : Z+ ! X .
It's customary to write xn for x(n) in this situation, and to abuse terminology by referring to the
collection fxn g as \the sequence".
De nition. Let X be a topological space. A sequence fxng converges to a point x 2 X if for every
neighborhood U of x, there is an integer N such that xn 2 U for all n  N .
xn ! x means that fxn g converges to x.
Lemma. Let X be a Hausdor space. Convergent sequences converge to unique points.
Proof. Let xn ! x and xn ! y. I want to show that x = y.
Suppose x 6= y. Since X is Hausdor, I can nd disjoint neighborhoods U of x and V of y. Since
xn ! x, I can nd an integer M such that n  M implies xn 2 U . Since xn ! y, I can nd an integer N
such that n  N implies xn 2 V . Therefore, for n  max(M N ), I have xn 2 U \ V = . This is nonsense,
so x = y.
In particular, limits of sequences are unique in metric spaces.
Lemma. (The Sequence Lemma) Let X be a topological space, let Y  X , and let x 2 X . If there is a
sequence fxng with xn 2 Y for all n and xn ! x, then x 2 Y . The converse is true if X is a metric space.
Proof. Suppose that there is a sequence fxng with xn 2 Y for all n and xn ! x. Let U be a neighborhood
of x. Find an integer N such that xn 2 U for all n  N . Obviously, U meets Y . This proves  that x2Y.
1
Conversely, suppose that X is a metric space and x 2 Y . For each n 2 Z , the ball B x n meets Y ,
+
 
so I may choose xn 2 B x n1 \ Y . I claim that xn ! x.

5
 Let U be a neighborhood of x. Since the open balls form a basis for the  metric
 topology, I may nd
1 1
 > 0 such that B (x )  U then I may nd N 2 Z such that n < , so B x N  B (x ).
+

1 1
 1  1
For all n  N , I have n < N , so xn 2 B x N . Since B x N  U , I have xn 2 U for all n  N .
This proves that xn ! x.
Theorem. Let X be a metric space, let Y be a topological space, and let f : X ! Y . f is continuous if and
only if xn ! x in X implies that f (xn ) ! f (x) in Y .
More succinctly, continuous functions carry convergent sequences to convergent sequences.
Proof. Suppose f is continuous, and suppose xn ! x in X . Let V be a neighborhood of f (x) in Y . By
continuity, there is a neighborhood U of x such that f (U )  V .
Since xn ! x, there is an integer N such that xn 2 U for all n  N . Then f (xn ) 2 f (U )  V for all
n  N . This proves that f (xn ) ! f (x).
Conversely, suppose that xn ! x in X implies that f (xn ) ! f (x) in Y . To show f is continuous, it will
suce to show that for all A  X , I have f (A)  f (A).
Thus, take x 2 clA. I want to show that f (x) 2 f (A).
Now X is a metric space and x 2 clA, so by the Sequence Lemma, there is a sequence of points fxng  A
with xn ! x. By hypothesis, this implies that f (xn ) ! f (x). Since ff (xn )g is a sequence in f (A), the
Sequence Lemma implies that f (x) 2 f (A). Therefore, f is continuous.
De nition. Let ffn : X ! Y g be a sequence of functions from X to Y , where Y is a metric space. ffng
converges uniformly to a function f : X ! Y if for every  > 0, there is an integer N such that
d(fn(x) f (x)) <  for n  N and x 2 X:
Theorem. Let ffn : X ! Y g be a sequence of continuous functions from X to Y , where Y is a metric space.
If ffng converges uniformly to f : X ! Y , then f is continuous.
This is often expressed by saying that a uniform limit of continuous functions is continuous.
Proof. Let f (a) 2 Y and let B (f (a) ) be a neighborhood of f (a). I want to nd a neighborhood U of a
such that f (U )  B (f (a) ).
First, uniform continuity implies that there is an integer N such that
d(fn (x) f (x)) < 3 for n  N x 2 X: ()
In particular,
d(fN (a) f (a)) < 3 :
 
fN is continuous, so there is a neighborhood U of a such that fN (U )  B fN (a) 3 . Thus,

d (fN (x) fN (a)) < 3 for all x 2 U:

Moreover, restricting (*) to n = N and x 2 U , I have
d(fN (x) f (x)) < 3 for all x 2 U:
Therefore, the triangle inequality implies that
d(f (x) f (a))  d(fN (x) f (x)) + d (fN (x) fN (a)) + d(fN (a) f (a)) < 3 + 3 + 3 = 
6
for all x 2 U .
This proves that f is continuous.

Example. For n 2 Z+ , let fn : R ! R be dened by

fn (x) = nxx+ 1 :

For xed x, nlim nxx+ 1 = 0. Thus, this sequence of functions converges pointwise to the constant
!1
function 0.
The picture below shows the graphs of fn for n = 1 2 3 4 5 on the interval 0  x  1.
0.5

0.4

0.3

0.2

0.1

0.2 0.4 0.6 0.8 1

I will show that the convergence is uniform on the interval 0  x  1. Thus, choose  > 0 I must nd
an integer N such that if n  N , then
jfn (x)j < :
1 , f is an increasing function f (1) = 1 , so it follows that
Since fn (x) = (nx +
0

1)2 n n
n+1

jfn (x)j < n +1 1 for 0  x  1:

Now choose N such that N 1+ 1 < . Then if n  N ,

jfn(x)j < n +1 1  N 1+ 1 < :

This proves that fn converges uniformly to 0 on 0  x  1.

Example. For n 2 Z+ , let fn : R ! R be dened by

fn (x) = nx1+ 1 :

For xed x 6= 0, nlim nx1+ 1 = 0. Thus, this sequence of functions converges pointwise to the constant
function 0 for x 6= 0. It converges pointwise to 1 for x = 0.
!1

7
 The picture below shows the graphs of fn for n = 1 2 3 4 5 on the interval 0  x  1.
0.2 0.4 0.6 0.8 1

0.8

0.6

0.4

0.2

I will show that the convergence is not uniform on the interval 0 < x < 1. In fact, I will show that there
is no integer N such that if n  N , then
jfn (x)j < 41 :
1 1
To see this, it suces to note that fn n = 2 , so there will always be a point in 0 < x < 1 where the
function exceeds 14 .
Therefore, ffng converges pointwise, but not uniformly, to the zero function.

c 1999 by Bruce Ikenaga 8

 7-5-1999
Quotient Spaces
You may have encountered quotient groups or quotient rings in an algebra course. In those situa-
tions, the idea was to turn a set of cosets of a subgroup or an ideal into a group or a ring.
A set of cosets is a partition of a group or a ring. In the topological setting, a quotient space arises as
a topology on a set of partition elements. You can use this to construct new topological spaces by \gluing"
and \pasting".
If X and Y are topological spaces, a function f : X ! Y is continuous if and only if when V is open in
Y , f ;1 is open in X .
De nition. Let f : X ! Y , where X and Y are topological spaces. f is a quotient map if:
1. f is surjective.
2. V is open in Y if and only if f ;1 (V ) is open in X .
Remarks.
1. Quotient maps are continuous.
2. If f : X ! Y is a homeomorphism, then f is a quotient map.
Since f is a homeomorphism, it's continuous hence, if V is open in Y , then f ;1 (V ) is open in X .
Conversely, suppose V
Y and f ;1 (V ) is open in X . Since f ;1 is continuous,
V = f f ;1 (V ) = f ;1 ;1 f ;1 (V ) is open:
;
;
;

3. The composite of two quotient maps is a quotient map.

In the denition of a quotient map, there are topologies on both X and Y . More often, you start with
a topological space and a partition of the space, then try to topologize the set of partition elements so that
the associated function (carrying an element of the space to the partition element containing it) is a quotient
map.
Lemma. Let X be a topological space, let Y be a set, and let f : X ! Y be a surjection. Dene a collection
of subsets of Y by requiring that V 2 T if and only if f ;1 (V ) is open in X . Then T is a topology on Y .
T is called the quotient topology on Y .
Proof. f ;1 () =  is open in X , so  is open in Y . f ;1 (Y ) = X is open in X , so Y is open in Y .
Suppose fVi gi2I is a collection of subsets of Y such that f ;1 (Vi ) is open in X for all i 2 I . Now
 !
 
f ;1 Vi = f ;1 (Vi )

i2I i2I
S S
and i2I f ;1 (Vi ) is open in X . Therefore, i2I Vi is open in Y .
Finally, suppose V1, : : :Vn is a collection of subsets of Y such that f ;1 (Vi ) is open in X for i = 1
: : :
n.
 !
n
\ n
\
f ;1 Vi = f ;1 (Vi )

i=1 i=1
and ni=1 f ;1 (Vi ) is open in X . Therefore, ni=1 Vi is open in Y .
T T

Therefore, T is a topology.
1
Example. (Partition Topologies) Let X be a topological space, and let X  = fXa ga2A be a partition of
X . Dene f : X ! X  by
f (x) = Xa if x 2 Xa :
This makes sense, because every element of X lies in exactly one Xa .
f is clearly surjective the quotient topology on X  induced by f is called the partition topology.
In fact, all quotient topologies are partition topologies. For if X is a topological space and f : X ! Y is
surjective, the collection of inverse images ff ;1 (y) j y 2 Y g forms a partition of X . The partition topology
induced by this partition is the same as the quotient topology induced by f .
In applications, you often want to \glue" parts of a space together to make a new space. In this case,
the partition elements are as follow. A point that isn't being glued to anything else is in a partition element
by itself. Otherwise, a group of points that are being glued together form a partition element.
For example, take the interval 0
1]. Partition 0
1] by dening f0
1g to be one partition element and
dening fxg to be a partition element for 0 < x < 1. This has the eect of gluing the endpoints together
the quotient space is homeomorphic to a circle.

glue the
endpoints

glue the
sides

Similarly, if you take a rectangle and glue two opposite sides together, you get a cylinder. In this case,
the partition elements consist of individual points not on the glued sides and pairs of corresponding points
on the sides being glued.
If you twist the rectangle before gluing the opposite sides, the resulting quotient space is a Mobius strip.
If you glue both pairs of opposite sides together with no twisting, you get a torus.

Lemma. Let X and Y be topological spaces. If f : X ! Y is continuous, open, and surjective, then f is a
quotient map.
Proof. If V is open in Y , then f ;1 (V ) is open in X by continuity.
Suppose V
Y and f ;1 (V ) is open in X . Since f is open, f (f ;1 (V )) is open since f is surjective,
V = f (f ;1 (V )). Therefore, V is open.
Hence, f is a quotient map.
In general, quotient maps do not need to be open.
Covering spaces provide important examples of quotient maps they are fundamental objects in
algebraic topology.
De nition. Let p : E ! B be a continuous surjection. ` p is a covering map if every point x 2 B has
a neighborhood U such that p;1(U ) is a disjoint union a2A Ua of open sets and pjUa : Ua ! U is a
homeomorphism for all a 2 A.
A neighborhood U which satises this condition is evenly covered.
2
 Thus, p;1(U ) looks like a stack of pancakes, where each pancake is carried homeomorphically onto U
by p.

Ua

U
x

Lemma. Covering maps are open.

Proof. Let p : E ! B be a covering map, and let U
E be open. I want to show that p(U ) is open in B .
Take x 2 U , so p(x) 2 p(U ). I need to nd a neighborhood V of p`(x) contained in p(U ). Let W be an
evenly covered neighborhood of p(x). Since p;1 (W ) is a disjoint union a2A Wa of open sets homeomorphic
to W and since x 2 p;1 (W ), x must lie in one of the Wa 's.
Thus, let Wa be a neighborhood of x mapped homeomorphically onto W by p. Then Wa \ U is open,
p maps it homeomorphically onto p(Wa \ U ), and p(x) 2 p(Wa \ U ). I claim that p(Wa \ U ) is open in B .
To see this, observe that pjWa : Wa ! p(Wa ) = W is a homeomorphism, so the inverse pj;W1a : W ! Wa
is continuous. Now Wa \ U is open in E , so it's open in Wa . Hence, (pj;W1a );1 (Wa \ U ) = p(Wa \ U ) is open
in p(Wa ). But p(Wa ) = W is open in B , so p(Wa \ U ) is open in B .
I've found a neighborhood p(Wa \ U ) of x which is contained in p(U ). Therefore, p(U ) is open, and p
is an open map.
Corollary. Covering maps are quotient maps.

Example. The map e : R ! S 1 given by

e(t) = (cos(2t)
sin(2t))

is a covering map it wraps the real line around the circle. Each open interval of length 1 in R is mapped
homeomorphically to the circle minus a point.

Example. If X  Y is a product of topological spaces, the projection maps 1 : X  Y ! X and 2 :

X  Y ! Y are open, continuous, and surjective. Therefore, they are quotient maps.

The next result tells you how to construct a map out of a quotient space it is analogous to the results
in algebra which tell you how to construct group maps out of quotient groups, or ring maps out of quotient
rings.
De nition. Let p : X ! Y be a quotient map. A ber of p is a set p;1 (y) for y 2 Y .
Theorem. (Universal Property for Quotients) Let p : X ! Y be a quotient map. Then any continuous
map f : X ! Z which is constant on each ber of p factors uniquely through p that is, there is a unique
3
continuous map f~ : Y ! Z such that fp~ = f .
X
@
p @@f
? f~ @@R
Y -Z
Proof. Let y 2 Y . Since f is constant on p;1(y), I may take any x 2 p;1 (y) and dene
f~(y) = f (x):
~ = f . I have to show that f~ is continuous.
By construction, fp
Let U be open in Z . I needto showf~;1 (U ) is open in Y . By denition of the quotient topology, f~;1 (U )
is open in Y if and only if p;1 f~;1 (U ) is open in X . But since fp ~ = f,
 
p;1 f~;1 (U ) = f ;1 (U )

which is open in X by continuity of f .
This proves that f~ is continuous.
To prove uniqueness, suppose that f~1 and f~2 are continuous maps Y ! Z such that f~1 p = f and
f~2 p = f . Then f~1 p = f = f~2 p since p is surjective, it is right-cancellable, so f~1 = f~2 .
If you think of the quotient topology as \gluing" or identifying the elements of each ber, the universal
property says that the quotient topology is the \smallest" way of accomplishing this \gluing", in the sense
that any other map that accomplishes this \gluing" receives a map from the quotient space.
Corollary. Let f : X ! Z be a continuous surjective map and suppose p : X ! X  is the quotient map
obtained from the partition of X by the bers ff ;1 (z ) j z 2 Z g.
1. There is a continuous bijection f~ : X  ! Z such that fp ~ = f.
2. f~ is a homeomorphism if and only if f is a quotient map.
Proof. 1. p is the function which sends x 2 X to the partition element f ;1 (z ) 2 X  , where z = f (x). f is
trivially constant on each ber, so the Universal Property yields a continuous map f~ : X  ! Z such that
~ = f.
fp
To be explicit, consider a ber f ;1 (z ) 2 X  . Let x 2 f ;1 (z ). Then p(x) = f ;1 (z ), so
;

f~ f ;1 (z ) = f~ (p(x)) = f (x) = z:
The function g : Z ! X  given by g(z ) = f ;1 (z ) is obviously the inverse of f~ as a set map, so f~ is
bijective. Thus, f~ is a continuous bijection.
2. If f~ is a homeomorphism, then f = fp ~ is a composite of quotient maps, so it's a quotient map.
Conversely, suppose f is a quotient map. p is constant on each ber of f , so the Universal Property
applied to the quotient map f yields a continuous map h : Z ! X  such that hf = p.
Now fhf ~ = f . That is, fh
~ makes the outer triangle below commute:
X
;; @@
f;
;
p @f@
Z
; h - X? f~ -@R Z
4
 ~ with idZ . By the uniqueness part of the Universal
But the outer triangle clearly commutes if I replace fh
~ = idZ .
Property applied to the quotient map f , it follows that fh
~ ~
Likewise, hfp = p. In other words, hf makes the outer triangle below commute:
X
; @@
p;; f
; @p

;
 ? h @@R
X f
~
-Z - X
The outer triangle will commute if I replace hf~ with idX . By the uniqueness part of the Universal
Property applied to the quotient map p, it follows that hf~ = idX .
Therefore, h = f~;1 . (In fact, h is the map g dened in the rst part of the proof.)
Since f~ is a continuous bijection with a continuous inverse, it's a homeomorphism.
Corollary. Let f : X ! Z be a continuous surjective map and suppose p : X ! X  is the quotient map
obtained from the partition of X by the bers ff ;1 (z ) j z 2 Z g. If Z is Hausdor, then so is X  .
Proof. Suppose f ;1 (z1 ) and f ;1 (z2 ) are distinct elements of X  . Then z1 6= z2 , so I may nd disjoint
neighborhoods U1 of z1 and U2 of z2 in Z .
Look at f~;1 (U1 ) and f~;1 (U2 ). These are open in X  , since f~ is continuous. Since U1 and U2 are
disjoint, f~;1 (U;1 ) and f~
;1 (U2 ) are disjoint.
Finally, f~ f ;1 (z1 ) = z1 2 U1 shows that f ;1 (z1 ) 2 f~;1 (U1 ), and likewise f ;1 (z2 ) 2 f~;1 (U2 ).
Thus, I have disjoint neighborhoods of f ;1 (z1 ) and f ;1 (z2 ) in X  , so X  is Hausdor.

c 1999 by Bruce Ikenaga 5

 7-26-1999
Connected Spaces
De nition. A space is connected if it cannot be written as the disjoint union of two nonempty open sets.
A separation of a topological space X consists of two disjoint nonempty open sets U and V such that
X = U V.
Thus, a connected space is one that cannot be separated.

Example. If X is connected and Y is homeomorphic to X , then Y is connected.

Example. A point is connected (considered as a subspace of an arbitrary topological space).

Example. R is connected in the standard topology.

On the other hand, Zis not connected when considered as a subspace of R. (The subspace topology is
the discrete topology in this case.)
More generally, a set with more than one point is not connected in the discrete topology.

Example. If X = U V is a separation of X , then U and V are both open and closed.

Conversely, if A is a subset of a topological space X , A 6=  X , and A is both open and closed, then
X = A (X ; A) is a separation of X .
To put it another way, a space is connected if and only if there is no proper nontrivial subset which is
both open and closed.
For example, any (nontrivial) partition topology is not connected.

Proposition. Let X = U V be a separation of X . If Y  X is connected, then Y  U or Y  V .

Proof. Suppose on the contrary that Y \ U 6=  and Y \ V 6= . Then Y \ U and Y \ V are nonempty open
subsets of V whose union is V , contradicting the fact that Y is connected.
If a space X is not connected, what are the biggest connected subspaces of X ?
T
Lemma. Let fSUa ga2A be a collection of connected subsets of a topological space X . Suppose that a2A Ua 6=
. Then U = a2A Ua is connected.
T
Proof. Suppose on the contrary that U = V W is a separation of U . Let x 2 a2A Ua . Then x 2 V or
x 2 W without loss of generality, suppose x 2 V .
For each a, the connected set Ua is contained in either V or W , by the preceding result. But x 2 Ua for
all a, so all the Ua 's meet V . Therefore, Ua  V for all a, and hence U  V .
This means that W = , which contradicts the fact that V and W separated U .
Therefore, U is connected.
Theorem. Let X be a topological space.
1. Every point x 2 X is contained in a unique maximal connected subset Cx of X .
2. Cx = Cy or Cx \ Cy = .
1
 To say that Cx is maximal means that Cx is not properly contained in any other connected subset of
X.
Proof. 1. Let x 2 X . Consider the collection C of connected subsets of X which contain x. C is nonempty,
since fxg 2 C . Order C by inclusion. I'll use Zorn's Lemma to show S that C has maximal elements.
Let D be a chain in C . Then every U 2 D contains x, so V = U 2D U is a union of connected sets with
a point in common. By the preceding result, V is connected since it contains x and contains every U 2 D,
it is an upper bound for D in C .
By Zorn's Lemma, C has maximal elements.
Next, I'll show that C has a unique maximal element.
If U and V are maximal elements of C , then U V is a connected set, since U and V are connected
and have x in common. Thus, U V is a connected set containing x, so it's an element of C . Since U and
V are maximal and are contained in U V , this is only possible if U = V .
Note that if Cx is the unique maximal element in C , any connected set containing Cx would be a
connected set containing x | hence, an element of C containing the maximal element Cx . Therefore, there
is no connected set which properly contains Cx.
2. Suppose Cx \ Cy 6= . Let z 2 Cx \ Cy . Then Cx Cy is connected, since it's the union of connected
sets having a point in common. Moreover, Cx Cy contains the maximal sets Cx and Cy . By part 1, this is
impossible unless Cx = Cy .
The maximal connected sets described by the theorem form a partition of X . They are called the
connected components of X . Here is another sense in which they are maximal.
Lemma. Let X be a topological space. Every connected subset of X is contained in a component of X .
Proof. Let U be a connected subset of X . U must intersect some component of X suppose U \ C 6= , where
U is a component. Then U C is a union of connected sets with nonempty intersection, so it's connected.
Since C  U C , and since C is maximal, I must have C = U C . Therefore, U  C .
Theorem. The continuous image of a connected set is connected.
Proof. Let f : X ! Y be continuous and surjective, and suppose X is connected. I want to show that Y is
connected. Q
Suppose on the contrary that Y = U V is a separation of Y . Since U and V are open, f ;1 (U )
and f ;1Q(V ) are open since U and V are disjoint, f ;1 (U ) and f ;1 (V ) are disjoint. Therefore, X =
f ;1 (U ) f ;1 (V ) is a separation, contradicting the fact that X is connected.
The theorem will be useful once I have some connected spaces to play with.
I'm working toward showing that the product of connected spaces is connected the preliminaries are
interesting in their own right.
Q
Lemma. Let X be a topological space, let Y  X , and let Y = A B be a partition of Y . A and B separate
Y if and only if neither A nor B contains an X -limit point of the other.
Proof. Suppose A and B separate Y . Then A and B are closed in Y , so
A \ Y = clY A = A:
Suppose x is an X -limit point of A, so x 2 A. If x 2 B , then x 2 Y , so x 2 A \ Y = A. But then
x 2 A \ B , contradicting the fact that A and B are disjoint.
Therefore, B contains no X -limit points of A.
A similar argument shows that A contains to X -limit points of B .
Conversely, suppose neither A nor B contains an X -limit point of the other. Let A0 denote the X -limit
points of A. Then A0 \ Y is the set of X -limit points of A contained in Y by assumption, none of these are
in B , so any such points must be in A. That is, A0 \ Y  A. Then
A \ Y = (A A0) \ Y = (A \ Y ) (A0 \ Y )  A A = A  A \ Y:
2
 Therefore, A = A \ Y , so A is closed in Y .
A similar argument shows that B is closed in Y .
Since A and B partition Y , they form a separation of Y .
Lemma. If A is connected and
AB A
then B is connected.
Heuristically, if you add some limit points of a connected set to the set, you get a connected set.
Q
Proof. Suppose B = U V is a separation of B . Thus, U and V are disjoint nonempty open subsets of B .
Since A is connected, A lies entirely in U or V . Without loss of generality, say A  U . Taking closures in
X yields A  U .
U and V are disjoint, and by the preceding lemma V contains no limit points of U . Thus, A \ V = .
But B  A, so B doesn't intersect V , either. This means that V is empty, which is a contradiction.
Therefore, B is connected.
Corollary. The closure of a connected set is connected.
Corollary. Connected components of a space are closed.
Proof. If C is a connected component of X , then C is a connected set containing C , by the previous corollary.
By maximality, C = C , so C is closed.
Lemma. The product of nitely many connected spaces is connected.
Proof. It suces to prove this for two spaces, since the general result will follow by induction.
Thus, suppose X and Y are connected. I want to show that X  Y is connected. The idea is to use the
fact that a union of connected sets with a point in common is connected.
To do this, picture X  Y as a plane. I'll decompose it as the union of \crosses": each \cross" consists
of the union of a xed \horizontal line" with an arbitrary \vertical line".
Y

{x} x Y

X x {y0 }
y0

x
X

Fix y0 2 Y . For each x 2 X , consider the set

Cx = (X  fy0 g) (fxg  Y ):
X  fy0 g is connected, since it's a homeomorphic copy of X fxg  Y is connected, since it's a homeo-
morphic copy of Y . The sets have the point (x y0) in common. Therefore, Cx is connected.
Next, the union of the Cx 's is connected, because it's a union of connected sets having X  fy0g in
common. But the union of the Cx 's is X  Y , so X  Y is connected.
Theorem. An arbitrary product of connected spaces is connected.
Proof. First, I'll choose a point in the product. I'll construct a union of connected sets having this point in
common the earlier result on unions shows that the union is connected. Finally, I'll show that the closure
of the union is the whole product this will prove that the product is connected, by the earlier result on
closures.
3
 Q
Let X = a2A Xa be a product of connected spaces. I want to show that X is connected. Let
y = (ya )a2A 2 X .
For each nite collection fa1 : : : ang of indices from A, dene
X (a1 : : : an) = (xa )a2A j xa = ya if a 6= a1 : : : ang:
Thus, a point of X (a1 : : : an) agrees with y except possibly in the a1 , : : : , an positions.
(Note that this doesn't mean that such an x doesn't agree with y in the a1 , : : : , an positions it means
only that such an x does agree with y in the other positions.)
I claim that X (a1 : : : an) is connected. I'll do this by showing that it's homeomorphic to a nite
product of Xa 's, which is connected by the preceding lemma.
Dene  : Xa1   Xa ! X (a1 : : : an) by
n

(xa1 : : : xa )]a = xy a ifif aa =

i ai for some i
6= ai for any i :
n
a
Thus,  plugs xa1 , : : : , xa into the a1, : : : , an components and uses y for the other components.
n
Dene  : X (a1 : : : an) ! Xa1   Xa by n

(xa )] = (xa1 : : : xa ):
n

 and  are easily seen to be inverses, so  is bijective.

A basis element for the product topology on Xa1   Xa is a product of open sets from the factors.
n

Q on X (a1 : : : an) consists of the Q

A basis element for the subspace topology intersection of X (a1 : : : an) with
a basis element for the product X = a2A Xa . A basis element for X = a2A Xa consists of a product of
open sets from the factors, where at most nitely many are not the whole space.
With this description in mind, it is clear that  carries a basis element for Xa1   Xa onto a n
basis element for X (a1 : : : an), and  carries a basis element for X (a1 : : : an) onto a basis element for
Xa1   Xa . Therefore,  is a homeomorphism.
n
Since Xa1   Xa is a nite product of connected spaces, it's connected. Therefore, X (a1 : : : an)
n
is connected.
Since y 2 X (a1 : : : an) for any nite subset fa1 : : : ang  A, the union

n  o
X 0 = X (a1 : : : an)  fa1 : : : ang  A
is a union of connected sets with a point in common, henceQ connected.
Finally, I claim that X 0 = X . Let x 2 X , and let a2A Ua be a neighborhood of x, where Ua is open
in Xa and Ua = Xa for all but nitely many a's. Let a1 : : : an be the indices for which Ua 6= Xa . Dene
z = (za )a2A by
za = xyaa ifif aa =
i ai for some i
6= ai for any i :
Q
Then z 2 a2A Ua and z 2 X (a1 : : : an)  X 0 .
Thus, every (basic) neighborhood of x intersects X 0 . Since x 2 X was arbitrary, it follows that X 0 = X .
Since X is the closure of a connected set, it's connected.

c 1999 by Bruce Ikenaga 4

 8-10-1999
Connected Subsets of the Real Line
Proposition. 0 1] is connected in the standard topology.
Proof. Suppose 0 1] = is a separation, so and are nonempty, disjoint, open sets.
U V U V

The idea of the proof is to nd a point on the \boundary" between and , then show that such a U V

point can't be in either or . U V

Without loss of generality, suppose that 0 2 . Let U

S = f 2 0 1] j 0 ]  g
s s U :

h
Since 0i 2 , is nonempty. Moreover, since 0 2 and is open, there is an 0 such that 0 )  .
S

Then 0 2  , so 2 2 . Thus, 6= f0g.


U
S

S S
U U  >  U

is a nonempty set which is bounded above, so it has a least upper bound. Let
S

s0 = sup S:

Since 6= f0g, 0 0.S s >

0 is the sort of \boundary point" I alluded to above. Before using it to obtain a contradiction, I'll
s

prove two preliminary results.

Claim 1: If 2 and 0  s , then 2 .
S r < s r S

0 ]  0 ] 
r s U since s 2 S:
Therefore, 2 . r S

Claim 2: If 0  0 , then 2 and 2 .

s < s s S s U

Suppose 0  0 but 2 . If
s < s , then 2 , else 2 by Claim 1. Since no element of is
s = S t > s t = S s S S

greater than , it follows that is an upper bound for . But

s 0 , and 0 is the least upper bound for .
s S s < s s S

This contradiction implies that 2  thus, 0 ]  , so 2 . s S s U s U

Finally, I'll obtain a contradiction by showing that 0 can't be in or its complement. s S

Suppose 0 2 , so 0 0]  . Note that 0 1 | for if 0 = 1, then 0 1] = 0 0]  , contradicting

s S s U s < s s U

the fact that is nonempty. V

Now 0 0 1, 0 2 , and is open, so there is an 0 such that ( 0 ; 0 + )  . Hence,

< s < s U U  > s  s  U

h0 i
+ 2 = 0 0]
h i
s0

s s0 s0 +2  
U

so + 2 2 . This contradicts the fact that 0 is the least upper bound for .
s0

S s S

It follows that 0 2 . Therefore, 0 0] 6 . By Claim 2, 0 0)  , so 0 2 . Therefore, 0 2 .

s = S s U s U s = U s V

Now 0 0 and is open, so there is an

s > V 0 such that ( 0 ; 0 ]  . In particular, 0 ; 2 2 , and
 > s  s V s

V

this contradicts Claim 2.

It follows that there is no such separation, and hence 0 1] is connected.
Corollary. Any closed interval in R is connected.
Proof. Any closed interval in R is homeomorphic to 0 1].
Corollary. R is connected.
1
Proof.
1
R= ; n n ]
n=1
represents R as a union of connected sets having points in common (speci cally, the points in ;1 1] are
common to all of the intervals). By an earlier result, this proves that R is connected.
Corollary. Rn is connected.
Proof. Rn is a product of connected spaces, so it's connected.
Corollary. Any open interval in R is connected.
Proof. Any open interval in R is homeomorphic to R.
In fact, more is true. Let be an ordered set with at least two elements. is a linear continuum if
L L

L has the least upper bound property (every subset bounded above has a least upper bound) and if x < y

implies there exists such that

z . For example, R is a linear continuum. One may show that every
x < z < y

linear continuum is connected.

You can show that many sets in Rn are connected by using the results above and general stu about
connected sets.

Example. 1 is a connected subset of R2 .

S

To see this, note that the continuous function : R ! 1 given by ( ) = (cos sin ) maps the
f S f t t t

connected set R onto 1 . S

More generally, if : R ! R2 is continuous, then the image is connected. For example, the graph of a
f

function = ( ) de ned for all 2 R is a connected set.

y f x x

Example. The topologist's sine curve consists of the graph of = sin 1 for 0 y  1. The closed < x

topologist's sine curve consists of the union of the topologist's sine curve with the segment f(0 ) j ;1 
x
y

y  1g.

The topologist's sine curve is connected, because it's a continuous image of a connected set. The closed
topologist's sine curve is the closure of the topologist's sine curve in R2, so it's connected as well.

Here's a familiar result from calculus.

Theorem. (Intermediate Value Theorem) Let : ! be a continuous function, where is connected
f X Y X

and is an ordered set with the order topology. Let 2 , and let be an element of between ( )
Y a b X d Y f a

and ( ) (so ( )   ( ) or ( )   ( )). Then = ( ) for some 2 .

f b f a d f b f b d f a d f c c X

2
Proof. If = ( ) or = ( ), then I'm done. So assume 6= ( ) ( ). Then ( )
d f a d f b d f a f b f a ( ) or
< d < f b

()
f b ( ).
< d < f a

( ) is connected, since is connected. Consider the sets ( ) \ (;1 ) and ( ) \ ( +1). These
f X X f X d f X d

are disjoint open subsets of ( ) they are nonempty, since one contains ( ) and the other contains ( ).
f X f a f b

Therefore, their union isn't all of ( ). Hence, ( ) \ f g is nonempty.

f X f X d

An element ( ) \ f g is an element ( ) for 2 such that ( ) = .

f X d f c c X f c d

c 1999 by Bruce Ikenaga 3

 7-5-1999
Path-Connectedness and Local Connectedness

De nition. Let X be a topological space. A path in X is a continuous function : 0
1] ! X . (0) and
(1) are the endpoints of the path is a path from (0) to (1).
De nition. Let X be a topological space. X is path connected if for all x
y 2 X , there is a path in X
from x to y.

Example. Points are path connected.

Example. Rn is path connected. If x
y 2 Rn, then

(t) = (1 ; t)x + ty
is a path in Rn from x to y.

Lemma. A path connected space is connected.

Proof. Suppose X is path connected, and suppose that X = U  V is a separation of X . Let x 2 U and let
y 2 V . By assumption there is a path : 0
1] ! X such that (0) = x and (1) = y.
;1 (U ) and ;1(V ) are disjoint open sets in 0
1], and 0
1] = ;1 (U )  ;1 (V ). This contradicts the
fact that 0
1] is connected.
Therefore, X is connected.

Example. The topologist's sine curve S consists of the graph of y = sin x1 for 0 < x  1.
The topologist's sine curve is connected, since it's a continuous image of the connected set (0
1]. The
space T = S f(0
0)g is also connected, since it's the union of S with one of its limit points in R2 . However,
T is not path connected.
To prove this, I'll show that the point (0
0) cannot be connected by a path to the point (1
sin1).
Suppose on the contrary that : 0
1] ! S is a path from (0
0) to (1
sin1). Notice that (0
0) is closed
in T , so ;1((0
0)) is a closed subset of 0
1]. I will show that ;1((0
0)) is also open in 0
1] this will
prove that ;1 ((0
0)) = 0
1].
Let x 2 ;1 ((0
0)). Let V be the open rectangle (;0
5:0:5)  (;0:5
0:5). The intersection of V with S
is an open set in S containing (0
0) | that is, a neighborhood of (0
0) in S . By continuity, I may nd an
open interval (a
b) in R containing x such that
(0
1] \ (a
b))  V \ S:
The intersection 0
1] \ (a
b) is an open interval or a half-open interval, so it's connected. Therefore,
(0
1] \ (a
b)) is connected. V \ S is a disjoint union of open arcs in S with the point (0
0). Hence,
(0
1] \ (a
b)) must map entirely into one of these components.
However, I know the point x 2 0
1] \ (a
b) maps to (0
0). Therefore,
(0
1] \ (a
b)) = (0
0):
Thus, 0
1] \ (a
b) is a neighborhood of x contained in ;1((0
0)). This proves that ;1 ((0
0)) is open.
1
 Since ;1 ((0
0)) is a nonempty, open, and closed, and since it's a subset of the connected set 0
1], it
must be all of 0
1]. Thus, (0
1]) = (0
0), contradicting the fact that is a path from (0
0) to (1
sin 1).
It follows that there is no such path, and therefore T isn't path connected.

Connected components were dened as the maximalconnected subsets of a space their existence followed
from Zorn's Lemma. The path components of a space are dened in a similar fashion. I need a lemma
which is analogous to one I prove for connected sets.
Lemma. Let fUa ga2A beSa collection of path connected subsets of a topological space X . Suppose that
T
a2A Ua 6= . Then U = a2A Ua is path connected.
Proof. Let y
z 2 U . I must show that y and z can be joined by a path in U .
Suppose y 2 Ua and z 2 Ub . Since x 2 Ua and Ua is path connected, x and y can be joined by a path
in Ua . Since x 2 Ub and Ub is path connected, x and z can be joined by a path in Ub .
Concatenating the two paths produces a path in Ua  Ub  U joining y and z . Therefore, U is path
connected.
Theorem. Let X be a topological space.
1. Every point x 2 X is contained in a unique maximal path connected subset Cx of X .
2. Cx = Cy or Cx \ Cy = .
To say that Cx is maximal means that Cx is not properly contained in any other path connected subset
of X .
The proof will be omitted it is essentially the same as the result for connected components. You
can verify that the connected components proof required only the lemma on unions of connected sets with
nonempty intersection I just proved the analogous lemma for path connected sets above.
De nition. Let X be a topological space. The path components of X are the maximal path connected
subsets of X .
De nition. Let X be a topological space.
1. X is locally connected if for every x 2 X and every neighborhood U of x, there is a connected
neighborhood V of x such that V  U .
2. X is locally path connected if for every x 2 X and every neighborhood U of x, there is a path
connected neighborhood V of x such that V  U .

Example. Since a path connected neighborhood is a connected neighborhood, every locally path connected
space is locally connected.

Example. Rn is locally path connected, since a ball B (x ) is path connected.

Example. Let S denote the topologist's sine curve:

1
 
S= x
sin 
0<x1 :
x
Let T = S f(0
0)g. T is not locally connected a neighborhood of (0
0) consists of (0
0) together with
a disjoint union of open arcs, which is not a connected set.
2
Theorem. A space is locally connected if and only if the components of any open subset are open.
In particular, the components of a locally connected space are open.
Proof. Suppose X is connected and U  X is open. Let C be a component of U . I want to show C is open.
Let x 2 C . By local connectedness, there is a connected neighborhood V of x such that x 2 V  U .
Now V  C is a union of connected sets having the point x in common since C is maximal, V  C = C , so
V  C.
Since every point of C has a neighborhood contained in C , C is open.
Conversely, suppose that components of open subsets are open. Let x 2 X , and let U be a neighborhood
of x. I must nd a connected neighborhood V of x such that x 2 V  U . Take V to be the connected
component of U which contains x. By assumption, V is open, and x 2 V  U .
Hence, X is locally connected.
An essentially identical argument proves the following.
Theorem. A space is locally path connected if and only if the path components of any open set are open.

In particular, the path components of a locally path connected space are open.
The relationship between components and path components is described by the following result.
Proposition. Let X be a topological space.
1. Every path component is a subset of a component.
2. If X is locally path connected, then the path components and components coincide.
Proof. 1. A path component is path connected, and path connected sets are connected. Since every
connected set is contained in a component (i.e. a maximal connected set), every path component is contained
in a component.
2. Suppose X is locally path connected. Let C be a component. I have to show that C is a path component.
Let P be any path component contained in C . (For instance, let x 2 C , and take P to be the path
component containing x.) Let P 0 be the union of the path components of C other than P .
Since X is locally path connected, P is open. X is locally connected, so C is open again, since X is
locally path connected, the path components of the open set C are open. Hence, P 0 is open, since it's a
union of open sets.
Now C = P  P 0 is a disjoint union of open sets. Since C is connected, P and P 0 can't be nonempty
since P 6= , I have P 0 = . Thus, C has no path components other than P , which means that C = P .

c 1999 by Bruce Ikenaga 3

 7-11-1999
Compact Spaces
In the real line, compactness is equivalent to being closed and bounded. The Heine-Borel Theorem says
that a closed and bounded subset B of R has the property that if fUi gi2I is a collection of open intervals
whose union contains B (i.e., the collection covers B ), then some nite subcollection of the U 's also covers
B . This is the appropriate way to generalize \closed and bounded" to arbitary spaces.
De nition.
S
1. A topological space X is covered by a collection fUi gi2I of subsets if X = i2I Ui .
2. If X is Sa topological space, Y X , and fUi gi2I is a collection of subsets of X , then fUi gi2I covers Y
if Y i2I Ui .
3. If fUigi2I covers X and each Ui is open, then fUi gi2I is an open cover of X .
4. If a subcollection of a covering also covers the space, the subcollection is a subcovering.

Example. The collection

1  
f(n n + 1) j n 2 Zg  n ; 2 n + 12  n 2 Z

is an open cover of R.
Every point
of R ; Zis contained
 in an interval of the form (n n + 1) for some n 2 Z. Each integer n
1 1
is contained in n ; 2 n + 2 . Thus, the collection covers R since each element of the collection is open,
it is an open cover.
The collection of open intervals with rational endpoints is also an open cover of R.
More generally, the collection of open balls with rational centers and rational radii forms an open cover
of Rn .

De nition. A space X is compact if for every open cover fUi gi2I of X , some nite subcollection fU1 : : : Ung
also covers X .
To say it another way, a space is compact if every covering has a nite subcovering.
It is clear that compactness is preserved by homeomorphisms: If X is compact and X is homeomorphic
to Y , then Y is compact.

Example. Any nite set of points (with any topology) is compact, since any open cover can contain at most
nitely many (distinct) open sets.

Example. R is not compact.

Consider the following open cover of R:

 
f(n n + 1) j n 2 Zg  n ; 21 n + 21  n 2 Z :
1


Each integer n is contained in exactly one element of the cover, namely n ; 12 n + 12 . Therefore,
any nite subcollection contains at most nitely many integers, so no nite subcollection can cover R.
Since every open interval (a b) is homeomorphic to R, open intervals aren't compact.
On the other hand, I'll show later that a closed interval a b] is compact.

The following condition is equivalent to compactness it is extremely important in analysis.

De nition. Let X be a topological space. A collection of subsets C of X satises the nite intersection
condition if the intersection of any nite subcollection
T of C is nonempty.
In other words, if C1 : : : Cn 2 C , then ni=1 Ci 6= .
Theorem. Let X be a topological space. X is compact if and only if every collection of closed sets satisfying
the nite intersection condition has nonempty intersection.
Proof. Suppose X is compact, and let fCigi2I be a collectionTof closed subsets of X such that every nite
T I want to show that i2I Ci 6= .
subcollection has nonempty intersection.
Suppose on the contrary that i2I Ci = . Then
\ 
X = X ; Ci = (X ; Ci ):
i2I i2I
Therefore, fX ; Ci gi2I is an open cover of X . By compactness, some nite subollection fX ; C1 : : : X ;
Cng covers X . Thus,
n \n
X = (X ; Ci ) = X ; Ci :
i=1 i=1
Tn
Hence, i=1TCi = , contrary to assumption.
Therefore, i2I Ci 6= .
Conversely, suppose that every collection of closed subsets of X satisfying the nite intersection condition
has nonempty intersection. Let fUi gi2I be an open cover of X . I want to show that fUi gi2I has a nite
subcover.
Suppose on the contrary that no nite subcollection of fUi gi2I covers X . Consider the complements
fX ; Ui gi2I . This is a collection of closed sets. If fX ; U1 : : : X ; Un g is a nite subcollection, then
\
n n
(X ; Ui ) = X ; Ui 6= :
i=1 i=1
S
For X 6= ni=1 Ui , since no nite subcollection of fUi gi2I covers X .
Thus, every nite subcollection of fX ; Ui gi2I has nonempty intersection. By assumption, fX ; Ui gi2I
has nonempty intersection. Therefore,
 \
X ; Ui = (X ; Ui ) 6= :
i2I i2I
This contradicts the fact that fUi gi2I covers X .
Hence, some nite subcollection of fUi gi2I covers X , and X is compact.
Thus, the intersection condition is really just a translation of the denition of compactness from open
to closed sets. In any situation, you can choose the version that is easier to apply.
The results that follow describe how compactness behaves in connection with other topological concepts.
Lemma. Let X be a topological space, and let Y X . Y is compact if and only if every open cover of Y
has a nite subcover.
2
 Reminder: say that fUi gi2I is an open cover of Y as a subset of X means that each Ui is open in X ,
S Uto. That
and Y i2I i is, the Ui 's are subsets of X , and not necessarily subsets of Y .
Proof. Suppose Y is compact. Let fUi gi2I be an open cover of Y by sets open in X . Then fUi \ Y gi2I
is an open cover of Y by sets open in Y . By compactness, some nite subcollection fU1 \ Y : : : Un \ Y g
covers Y . Then fU1 : : : Ung covers Y .
Conversely, suppose that every open cover of Y has a nite subcover. Let fUi gi2I be an open cover of
Y by sets open in Y . For each i, Ui = Y \ Vi , where Vi is open in X .
Now fVi gi2I is an open cover of Y by sets open in X , so by assumption a nite subcollection fV1 : : : Vng
covers Y . Then fY \ V1 : : : Y \ Vn g = fU1 : : : Ung covers Y . Therefore, Y is compact.
One way of putting this is: Compactness is intrinsic to the space. That is, to say Y is compact (in the
sense of the \open cover" property) has the same meaning whether Y is considered as a subspace of another
space or as a space in its own right. In contrast, whether a set is closed or not depends on the topology of
the ambient space.
Theorem. A closed subset of a compact space is compact.
Proof. Let X be a compact topological space, and let C be a closed subset of X . Let fUi gi2I be an open
cover of C , and consider the collection fUi gi2I  fX ; C g. This is an open cover of X , so by compactness
it has a nite subcover.
Ui

X-C

If X ; C occurs in this subcover, throw it out. The remaining sets fU1 : : : Ung cover C . Thus, I've
found a nite subcover of C , so C is compact.
Theorem. A compact subset of a Hausdor space is closed.
Proof. Let X be a Hausdor space, and let C be a compact subset of X . I'll show that the complement of
C is open.
Let x 2 X ; C . I want to nd a neighborhood of x that is contained in X ; C .
Since X is Hausdor, for each y 2 C , there are disjoint neighborhoods Uy of x and Vy of y. Now
fVy gy2Y is an open cover of C , so by compactness there is a nite subcover fVy1 : : : Vy g. Let n

n
\
U= Uy : i
i=1
Uy
1
Uy
2

Uy Vy
3 2

y2
y1

Vy y3
1

C Vy
3

3
 If z 2 U , then z 2 Uy for i = 1 : : : n. Hence, z 2= Vy for i = 1 : : : n. But fVy1 , : : : , Vy g cover C , so
z 2= C , i.e. z 2 X ; C . Thus, U is a neighborhood of x contained in X ; C . Hence, X ; C is open, so C is
i i n

closed.
Remark. In the course of the proof, I showed that in a Hausdor space, a compact set (C ) and a point in
the complement (x) may be separated by disjoint open sets.
Theorem. The continuous image of a compact space is compact.
Proof. Let X be a compact topological space, let Y be a topological space, and let f : X ! Y be a surjective
continuous function. I want to show that Y is compact.
Let fUi gi2I be an open cover of Y . Then ff ;1 (Ui )gi2I is an open cover of X , so by compactness I may
nd a nite subcover ff ;1 (U1 ) : : : f ;1 (Un )g.
Now if y 2 Y , then y = f (x) for some x 2 X . Find i 2 f1 : : : ng such that x 2 f ;1 (Ui ). Then
y = f (x) 2 Ui . This shows that fU1 : : : Un g covers Y .
Hence, Y is compact.
The proof of the following theorem makes nice use of the last three results.
Theorem. Let X and Y be topological spaces, where X is compact and Y is Hausdor. Let f : X ! Y be
a continuous bijection. Then f is a homeomorphism.
Proof. I need to show that f ;1 is continuous. I will show that the inverse image of a closed set is closed.
Let C be closed in X  I want to show that (f ;1 );1 (C ) = f (C ) is closed in Y . Since X is compact and
C is closed, C is compact. Since f is continuous and C is compact, f (C ) is compact. Since Y is Hausdor
and f (C ) is compact, f (C ) is closed.
Therefore, f ;1 is continuous, and f is a homeomorphism.

Example. Dene f : 0 1) ! S 1 by
f (t) = (cos(2t) sin(2t)):

S1

0 1

f is a continuous bijection, and S 1 is surely Hausdor. However, f is not a homeomorphism. Intuitively,

f ;1\unwraps" the circle onto the interval it fails to be continuous at the point (1 0), where the circle needs
to be \cut" in order to do the unwrapping. This is not continuous, because points on either side of the cut
point which start o close together wind up far apart, at opposite ends of 0 1).
Let X be the union of countably many copies of 0 1) and countably many copies of S 1 . Construct a
continuous bijection of X to itself by dening the map to be the \wrapping map" f from one of the 0 1)'s
4
to one of the S 1 's and the identity maps id : 0 1) ! 0 1) and id : S 1 ! S 1 on all the other pieces.

0 1

id

id

0 1 0 1

id

id

0 1 0 1

The inverse fails to be continuous for the same reason that f is not continuous.
This gives an example of a continuous bijection from a Hausdor space to itself which is not a homeo-
morphism.

The following result can be used to show that a nite product of compact spaces is compact it's of
interest in its own right.
Theorem. (The Tube Lemma) Let X and Y be topological spaces, and suppose Y is compact. Let x 2 X ,
and let W X  Y be an open set containing fxg  Y . Then there is a neighborhood U of x such that
U  Y W.
Y
w

{x} x Y

UxY

x X
U

Proof. For each y 2 Y , there is a neighborhood Uy  Vy of (x y) such that Uy is a neighborhood of x in X ,

Vy is a neighborhood of y in Y , and Uy  Vy W . It follows that fUy  Vy gy2Y is an open cover of fxg Y .
Now fxg  Y is compact, since it's homeomorphic to Y . Therefore, there is a nite subcollection
fUu1  Vy1 : : :Uy  Vy g which covers fxg  Y .
T n n

Let U = in=1 Uy . U is an open set containing x. If (z y) 2 U  Y , then y 2 Vy for some i 2 f1 : : : ng.

i i
So (z y) 2 U  Vy W , and hence U  Y W .
i

Example.
  
W = f(0 y) j y 2 Rg  (x y) j y <  x1  x 6= 0
5
is an open subset of R2 which contains the y-axis (i.e. f0g  R.
y

However, there is no open subset U of R such that

f(0 y) j y 2 Rg U  R W:
The Tube Lemma does not apply because R is not compact.

Lemma. Let X be a topological space. Let D be a collection of subsets of X which is maximal with respect
to satisfying the nite intersection condition. Then:
1. D is closed under nite intersections.
2. If C X and C meets every element of D, then C 2 D.
T
Proof. 1. Suppose D1 : : : Dn 2 D. I must show that D = ni=1 Di 2 D.
Consider the set D0 = D  fDg. I claim that D0 satises
T the nite intersection condition.
Take E1 : : : Em 2 D0 . If E1 : : : Em 2 D, then mi=1 Ei 6= , since D satises the nite intersection
condition.
Otherwise, one of the E 's is D without loss of generality, say E1 = D and E2 : : : Em 2 D. Then
\
m \
n
Ei = D \ (E2 \ \ Em ) = Di \ (E2 \ \ Em ) 6= 
i=1 i=1
since this is a nite intersection of elements of D.
Now D0 satises the nite intersection condition and it contains D. By maximality of D, D0 = D, which
means that D 2 D.
2. Suppose C X and C meets every element of D. I want to show that C 2 D.
Consider the set D0 = D  fC g. I claim that D0 satises
Tm the nite intersection condition.
Take E1 : : : Em 2 D . If E1 : : : Em 2 D, then i=1 Ei 6= , since D satises the nite intersection
0
condition.
Otherwise, one of the E 's is C  without loss of generality, say E1 = C and E2 : : : Em 2 D. Now
E2 \ \ Em 2 D by part 1, so
C \ (E2 \ \ Em ) 6= 
since C meets every element of D.
Now D0 satises the nite intersection condition and it contains D. By maximality of D, D0 = D, which
means that C 2 D.
All of the proofs of Tychono's theorem are sophisticated it's a dicult and important result. Lang
1] credits the following proof to Nicholas Bourbaki, the nom-de-plume of a group of mathematicians who
wrote an extremely inuential series of expository monographs.
6
 Q
Theorem. (Tychono
) Let fXa ga2A be a family of compact topological spaces. Then X = a2A Xa is
compact.
Proof. LetTC be a collection of closed subsets of X satisfying the nite intersection condition. I need to
show that C 2C C 6= .
Step 1. Consider the set ; of collections of subsets of X (closed or otherwise) which satisfy the nite
intersection condition and contain C . I claim that ; has maximal elements.
Order ; by inclusion.
Let ;0 be a chain in ;. Let 
F = G:
G2;0
Since each G 2 ;0 contains C , F also contains C . I claim that F satises the nite intersection condition.
Let F1 : : : Fn 2 F . Suppose F1 2 G1 , : : : , Fn 2 Gn , where G1 : : : Gn 2 ;0 . Since ;0 is a chain, there is
an index k 2 f1 : : : ng such that Fi 2 Gk for 1 i 6= n. Since Gk satises the nite intersection condition,
F1 \ \ Fn 6= . This proves that F satises the nite intersection condition.
Since F satises the nite intersection condition and contains C , it's an element of ;, and it's clearly an
upper bound for ;0 . By Zorn's lemma, ; has maximal elements.
Step 2. Let D be a maximal element of ;. I'll construct an element x 2 D2D D.
T
Let a : X ! Xa be the ath projection map. Fix a 2 A, and consider the family of sets fa DgD2D .
For any nite subcollection fa (D1 ) : : : a(Dn )g, the intersection D1 \ \ Dn is nonempty, because D
satises the nite intersection condition. Since
a (D1 \ \ Dn ) a (D1 ) \ \ a(Dn )
it follows that a(D1 ) \ \ a (Dn ) 6= .
Therefore, faDgD2D satises the nite intersection condition.
Hence, faD gD2D satises the nite intersection condition.
Now faD gD2D is a family of closed subsets of the compact space Xa . Therefore, there is an element
\
xa 2 a D:
D2D
Dene x = (xa )a2A .
Step 3. Let Ua be a neighborhood of xa in Xa . I claim that a;1(Ua ) 2 D.
Since Ua is a neighborhood of xa and xa 2 a D for all D 2 D, Ua meets a (D) for all D 2 D. But if
Ua \ a(D) 6= , then a;1 (Ua ) \ D 6= . To see this, note that if z 2 Ua \ a (D), then z = a (y) for some
y 2 D and so a(y) 2 Ua . Thus, y 2 a;1 (Ua ), and hence y 2 a;1 (Ua ) \ D.
Thus, a;1(Ua ) meets every set D 2 D. By the second part of the lemma, a;1 (Ua ) 2 D.
Step 4. Every basic open set containing x is contained in D.
The sets a;1 (Ua ) for a 2 A and Ua a neighborhood of xa in Xa form a subbasis for the open sets
containing x. A basic open set containing x is a nite intersection of a;1 (Ua )-sets such a nite intersection
is in D, by the rst part of the lemma.
T
Step 5. Finally, I'll show that x 2 C 2C C .
Let D 2 D, and let U be a basic open set containing x. Since D satises the nite intersection condition,
U \D = 6 . Since every basic open set containing x meets D, it follows that x 2 D .
T set in D, then x 2 D. But all the sets in C are closed, and C D, so x 2 C
In particular, if D is a closed
for all C 2 C . Therefore, x 2 C 2C C .
7
 This completes the proof that X is compact.

1] Serge Lang, Real analysis (2nd edition). Reading, Massachusetts: Addison-Wesley Publishing Company,
1983. ISBN 0-201-14179-5]

c 1999 by Bruce Ikenaga 8

 7-11-1999
Compact Sets and the Real Numbers
Theorem. If X is an ordered set satisfying the least upper bound property, then any closed interval a b] in
X is compact.
Proof. The outline of the proof is as follows. Let fUi gi2I be an open cover of a b]. Construct the set C
of points y 2 a b] such that a y] can be covered by a nite subcollection of fUi gi2I . Take the least upper
bound c of C and show that c 2 C. Finally, show that c = b.
Step 1. Suppose a
x < b. I claim that for some y > x, the interval x y] can be covered by at most two
elements of fUi gi2I .
If x has an immediate successor x + 1, then x x + 1] has only two elements, so it can be covered by at
most two U's.
If x does not have an immediate successor, nd Ui containing x. Pick y0 > x such that x y0 )  Ui this
is possible since Ui is open. Since x does not have an immediate successor, there is an element y such that
x < y < y0 . Then x y]  Ui , and x y] is covered by a single element of fUi gi2I .
Step 2. Now let
C = fy 2 (a b] j a y] can be covered by nitely many Ui g:
By Step 1, there is an element y > a such that a y] can be covered by at most two elements of fUi gi2I .
Therefore, C is nonempty. Let c be the least upper bound of C in a b].
Step 3. I claim that c 2 C.
Find Ui containing c. Ui is open and c > a, so I may nd an interval (d c]  Ui . Since d can't be an
upper bound for C, there is an element of C larger than d. Let c0 2 C, where d < c0 < c. Then a c0] can
be covered by nitely many U's, and c0 c]  Ui . Therefore, a c] = a c0]  c0 c] can be covered by nitely
many U's. Hence, c 2 C.
Step 4. I claim that c = b.
Suppose that c < b. By Step 1, there is a y > c such that c y] can be covered by at most two elements
of fUi gi2I . Since c 2 C, a c] can be covered by nitely many elements of fUi gi2I . So a y] = a c]  c y]
can be covered by nitely many elements of fUi gi2I , and therefore y 2 C. This contradicts the fact that c
was the least upper bound of C. Hence, c = b.
Since b 2 C, a b] can be covered by nitely many elements of fUi gi2I . Therefore, a b] is compact.
Corollary. A closed interval a b] in R is compact.
Theorem. (Heine-Borel) A subset C  Rn is compact if and only if it is closed and bounded relative to
the standard metric or the square metric.
Remark. Recall that the standard metric d and the square metric  are related by
p
(x y)
d(x y)
n  (x y):
Hence, a set bounded relative to one metric is bounded relative to the other.
Proof. Suppose that C is compact. Rn is Hausdor, so C is closed. The collection of balls fB(0 n) j n 2 Z+ g
is an open cover of Rn , so it is an open cover of C. By compactness, there is a nite subcover the element
of the subcover with the largest radius contains C, so C is bounded.
Conversely, suppose C is a closed and bounded subset of Rn. Suppose that (x y)
s for all x y 2 C.
Fix x 2 C and let t = (x 0). Then for all y 2 C,
d(y 0)
d(y x) + d(x 0) = s + t:
1
 It follows that C  ;(s + t) s + t]n. However, ;(s + t) s + t]n is a product of compact spaces, so it's
compact. Thus, C is a closed subset of the compact set ;(s + t) s + t]n, so it is compact.
Theorem. Let X be a compact topological space, let Y be an ordered set with the order topology, and let
f : X ! Y be continuous. There are points a b 2 X such that
f(a)
f(x)
f(b) for all x 2 X:
Proof. f(X) is a compact subset of Y . I'll show that f(X) has a largest element and a smallest element.
Suppose that f(X) does not have a largest element. Then the collection of open rays
f(;1 f(x)) j x 2 X g
cover f(X). By compactness, I may nd x1 : : : xn 2 X such that
f(1 f(x1)) : : : (;1 f(xn))g
cover f(X).
Find m 2 f1 : : : ng such that
f(xm ) = maxff(xi ) j 1
i
ng:
Then f(X)  (;1 f(xm )), but f(xm ) 2= (;1 f(xm )). This contradiction shows that f(X) has a
largest element a similar argument shows that f(X) has a smallest element.
In the case where X = a b] is a closed interval in R and Y = R, this is the familiar result from calculus
which says that a continuous function on a closed interval has a max and a min on the interval.
The next result and its corollary are more amusing than important, since you've probably seen the
Cantor diagonalization proof of the uncountability of 0 l]. But this topological proof does not make any
reference to representations of real numbers as innite decimal.
Theorem. Let X be a (nonempty) compact Hausdor space. Suppose every point of X is a limit point of
X. Then X is uncountable.
Proof.
Step 1. Let U be a nonempty open subset of X, and let x 2 X. I'll construct an open set V  U such that
x 2= V .
If U = fxg, then U is a neighborhood of x which contains only x. This contradicts the assumption that
every point of X is a limit point of X. Now x may not be in U in the rst place, but this shows that in any
event, U can't consist of x alone.
Thus, I may choose y 2 U such that y 6= x. Since X is Hausdor, there are disjoint neighborhoods A of
x and B of y. Let V = B \ U. Now
V =B \U B  X ;A
and X ; A is closed. Therefore, V  X ; A. In particular, since x 2 A, V does not contain x.
A

U
V=U B
U

2
Step 2. Next, I'll show that there is no surjection f : Z+ ! X.
Suppose that f : Z+ ! X. For the purposes of numbering, take V0 = X. By Step 1, I may nd a
nonempty open set V1 such that f(1) 2= V1.
Suppose inductively that n > 1, and I have found a nonempty open set Vn;1  Vn;2 such that
f(n ; 1) 2= Vn;1 . Use Step 1 to nd a nonempty open set Vn  Vn;1 such that f(n) 2= Vn .
Consider the collection fVn gn2Z+ . The collection has the nite intersection condition, since
V1 V2 V3 : : ::
T
By compactness, there is a point x 2 n2Z+ Vn . Then x 6= f(n) for any n 2 Z+ , so f is not surjective.
Since there is no surjective function from Z+ to X, X is uncountable.
Corollary. The interval 0 1] in R is uncountable.
Obviously, it follows that R and open intervals in R are uncountable.

c 1999 by Bruce Ikenaga 3

 7-12-1999
Limit Point Compactness and Sequential Compactness
De nition. Let X be a topological space.
1. X is limit point compact if every innite subset has a limit point.
2. X is sequentially compact if every sequence in X has a convergent subsequence.
If X is sequentially compact, then X is limit point compact. For every innite subset of X contains
an innite sequence the limit of a convergent subsequence is a limit point of the original subset. A partial
converse is contained in the following theorem.
Theorem. Let X be a topological space.
1. If X is compact, then X is limit point compact.
2. If X is a metric space and X is limit point compact, then X is sequentially compact.
Proof. 1. Let X be compact, and let S X be an innite subset. I must show that S has a limit point.
Suppose that S does not have a limit point. It is vacuously true that S contains all its limit points, so
S is closed. Since X is compact, S is compact.
Let x 2 S . Since x is not a limit point of S , there is a neighborhood Ux of x which does not intersect
S in a point other than x. fUx gx2S is an open cover of S , so by compactness there is a nite subcover
fUx1 : : : Ux g.
n
Now S fUx1  : : :  Ux , but each Ux meets S only in xi . Therefore, S = fx1 : : : xng, which
n i
contradicts the assumption that S is innite.
Therefore, S has a limit point.
2. Suppose X is a metric space and X is limit point compact. Let fxng be a sequence in X . I want to show
that fxng has a convergent subsequence.
First, I have to eliminate any repetitions in the sequence, so dene
S = fxn j n 2 Z+ g:
Thus, S is the set of points in fxng.
Suppose S is nite. Then for some x 2 X , xn = x for innitely many n 2 Z+ . Suppose this is true for
n1 n2 : : :. Then fxn gi2Z+ is a convergent subsequence of fxng, because all the terms are equal to x.
i
The only other possibility is that S is innite. Then

1by limit
 point compactness, S has a limit point x.
Consider the nested sequence of neighborhoods B x k  k 2 Z . Since metric spaces are Hausdor
+

1
and x is a limit point of S , each B x k contains innitely many points of S .
Choose n1 2 Z+ such that
xn1 2 B (x 1). Assuming that k > 1 and that nk1 has been chosen, choose
nk > nk;1 such that xn 2 B x k1 .
k

I now have xn ! x. For if  > 0, choose N so that n1 <  for k  N . Then xn 2 B (x ) for k  N .
k k
k

Example. The subsets f2n ; 1 2n] j n 2 Z+ g form a partition of Z+ . The corresponding partition topology
is called the odd-even topology. Let X denote Z+ with this topology.
X is limit point compact. To see this, let A be an arbitrary nonempty subset of X , and let a 2 A. For
some n 2 Z+ , a 2 2n ; 1 2n]. Notice that 2n ; 1 2n] is the smallest open set containing 2n ; 1 or 2n.
1
 If a = 2n ; 1, then 2n is a limit point of A, since every neighborhood of 2n contains a.
If a = 2n, then 2n ; 1 is a limit point of A, since every neighborhood of 2n ; 1 contains a.
In either case, A has a limit point. Since A was an arbitrary nonempty subset of X , surely every innite
subset of X has a limit point. Thus, X is limit point compact.
On the other hand, X is not sequentially compact. For example, the sequence f1 2 3 : ::g has no
convergent subsequence.
This example shows that limit point compactness does not in general imply sequential compactness.
In fact, X is not compact: The open cover f2n ; 1 2n] j n 2 Z+ g does not have a nite subcover. Thus,
limit point compactness does not in general imply compactness.

Recall that if Y is a subset of a metric space (X d), the diameter of Y is

diam(Y ) = supfd(x y) j x y 2 Y g:
Theorem. (Lebesgue Number Lemma) Let X be a sequentially compact metric space, and let fUi gi2I
be an open cover of X . There is an  > 0 such that every subset of X of diameter less than  is contained
in an element of fUi gi2I .
 is called a Lebesgue number for the cover.
Proof. Suppose on the contrary that no such  exists. Then for every n 2 Z+ , there is a set An of
diameter less than n1 which is not contained in any Ui . Let xn 2 An. I'll show that fxng has no convergent
subsequence, which will contradict sequential compactness.
Suppose on the contrary that fxn g is a convergent subsequence converging to x. Suppose that x 2 Ui .
k
Find  > 0 such that B (x ) Ui . Then choose nk suciently large so that
d (x xn ) < 2 and n1 < 2 :
k
k

Ui

1/n
k

B(x;ε ) xnk
x B(xnk ,1/nk )
x

ε
xn
An B(x;ε )

1

1

1

Since diam(An ) < n and xn 2 An , it follows that An B xn  n . Suppose y 2 B xn  n .
k k k k k k
k k k
Then
d(x y)  d(x xn ) + d(xn y)  2 + n1 < 2 + 2 = :
k k
k

1 
Therefore, B xn  n
k B (x ). Hence,
k


An B xn  n 1 B (x ) Ui :
k k
k

This contradicts the fact that An is not contained in any Ui . Therefore, there is a number  satisfying
k
the conclusion of the theorem.
Theorem. Let X be a metric space. The following are equivalent:
2
 1. X is compact.
2. X is limit point compact.
3. X is sequentially compact.
Proof. I've already proven (1 ) 2) and (2 ) 3). I need to prove (3 ) 1). Suppose then that X is
sequentially compact. I need to show that X is compact.
Step 1. For every  > 0, there is a nite covering of X by -balls.
Let x1 2 X and construct B (x1  ). If X = B (x1  ), then I'm done. Otherwise, choose x2 2 X ; B (x1  )
and construct B (x2  ). Again, if X = B (x1  )  B (x2  ), I'm done otherwise, choose x3 2 X ; (B (x1  ) 
B (x2  )). Keep going. Notice that each B (xn  ) contains only one xk , namely xn.
Suppose that the process does not terminate. Consider the sequence fxng. Iclaim that it has no
convergent subsequence. Indeed, if fxn g is a subsequence converging to x, then B x 2 can contain at
k

most one xn : If it contains two xn 's, then they are less than  apart. Therefore, there is an -ball about
k k
an xn containing another xn
k k
It follows that the process must terminate, so X is covered by a nite number of -balls.
Step 2. Every open cover of X contains a nite subcover.
Let fUigi2I be an open cover of X . Let  be a Lebesgue number for the cover. Cover X with a nite
number of 2 balls fB1 : : : Bn g. Each Bj has diameter less than , so each Bj is contained in some Ui . j

Then fUi j j = 1 : : : ng is a nite subcollection of fUigi2I which covers X . Therefore, X is compact.

j

De nition. Let X and Y be metric spaces, and let f : X ! Y . f is uniformly continuous if for every
 > 0, there is a  > 0 such that dX (a b) <  implies dY (f (a) f (b)) <  for all a b 2 X .
Theorem. Let X and Y be metric spaces, let f : X ! Y be a continuous function, and suppose X is
compact. Then f is uniformly continuous.
n   o n     o
Proof. Given  > 0, the open balls B y 2  y 2 Y cover Y . Therefore, f ;1 B y 2  y 2 Y is
an open cover of X .
By the Lebesgue  Lemma, there is a  > 0 such that every set of diameter less than  is
 Number

contained in a f B y 2 . In particular, if dX (a b) <  , then fa bg is a set of diameter less than  , so
; 1
    
a b 2 f ;1 B y 2 for some y. This means that f (a) f (b) 2 B y 2 , so dY (f (a) f (b)) < .

Example. Consider the function f : R ! R given by f (x) = x2. f is continuous, but f is not uniformly
continuous.
To show that f is not uniformly continuous, I'll show that there is no  > 0 such that if  > jx ; yj,
then 1 > jf (x) ; f (y)j. Suppose on the contrary that such a  exists. Choose x > 2 , and consider the points
x + 2 and x.
First,

x + 2 ; x = 2 < :


The points are less than  units apart.

However,

x + 2 + x = 2x + 2 > 2  2 + 2 = 4 + 2 > 4 :
3
 So


2



f x + 2 ; f (x) = x + 2 ; x2 = x + 2 ; x x + 2 + x = 2  x + 2 + x > 2  4 = 2:
The images are more than 1 unit apart.
This shows that f is not uniformly continuous.

c 1999 by Bruce Ikenaga 4

 7-16-1999
Local Compactness
De nition. Let X be a topological space. X is locally compact if for all x 2 X , there is a compact set C
and a neighborhood U of x such that x 2 U
C .
If X is compact, then X is locally compact: For any x 2 X , X is a compact set containing the
neighborhood X of x.

Example. If X is a space with the discrete topology, then X is locally compact. If x 2 X , then fxg is a
compact set containing the neighborhood fxg of x.

Example. Rn is locally compact: If x 2 Rn , the closed ball B (x 1) is compact, and it contains the open ball
B (x 1).

Theorem. Let X be a Hausdor space. X is locally compact if and only if for every x 2 X and every
neighborhood U of x, there is a neighborhood V of x such that V is compact and V
U .
Proof. Suppose for every x 2 X and every neighborhood U of x, there is a neighborhood V of x such that
V is compact and V
U . Choose any neighborhood U of x | for example, X | and nd a neighborhood
V of x with compact closure such that V
U . Then V is a compact set containing a neighborhood V of x,
so X is locally compact.
Conversely, suppose X is locally compact. Let x 2 X , and let U be a neighborhood of x. Let C be a
compact set which contains a neighborhood V of x.
C is compact and X is Hausdor , therefore C is closed. It follows that C ; U = C \ (X ; U ) is closed
since it's a subset of the compact set C , C ; U is compact. Note that x 2 U , so x 2= C ; U .
I showed earlier that in a Hausdor space, a compact set and a point not contained in it may be
separated by disjoint open sets. Thus, I may nd disjoint open sets W1 and W2 such that C ; U
W1 and
x 2 W2 .
U

C x

W2

W1

Next, I'll show that W2 \ V is compact. C is a closed set containing V , so V
C . But W2 \ V
V ,

so W2 \ V
V
C . Now W2 \ V is a closed subset of the compact set C , so W2 \ V is compact.
Moreover, W2 \ V
W2
X ; W1 . X ; W1 is closed, so W2 \ V
X ; W1. Since W2 \ V does not
meet W1, and since C ; U
W1, it follows that W2 \ V does not intersect C ; U . But W2 \ V is a subset
of C , so W2 \ V
U .
Thus, W2 \ V is a neighborhood of x with compact closure whose closure is contained in U .
Corollary. Let X be a locally compact Hausdor space, and let Y be a subspace of X . If Y is open or closed
in X , then Y is locally compact.
1
Proof. Suppose Y is open in X . Let x 2 Y . By the preceding result, I may nd a neighborhood U of x such
that U is compact and U
Y . Note that U is automatically open in Y . Thus, x is contained in a compact
set which contains a Y -neighborhood of x. Therefore, Y is locally compact.
Suppose Y is closed. Let y 2 Y , and let C be a compact set containing a neighborhood U of y. C is
closed, since X is Hausdor therefore, C \ Y is a closed subset of C . Since C is compact, C \ Y is compact.
Moreover, C \ Y contains U \ Y , which is a neighborhood of y. Therefore, Y is locally compact.

Example. Let S denote the topologist's sine curve

1
 
S= x
sin

0<x1 :
x

Let X = S  f(0
0)g. X is not locally compact.

Take a ball B ((0
0) ) about the origin. Consider the intersection of X with the line y = 2 . The
intersection is an in nite
 sequence of points with all but nitely many terms lying inside B ((0
0) ). The

sequence converges to 0
2 , which is not in X .
If there is a compact set C in X containing a neighborhood of (0
0), then that neighborhood contains
B ((0
0) ) \ X for  suciently small. By the argument above, such a neighborhood contains an in nite
subset of X with no limit point. The same is true for C , which contradicts the fact that compact sets are
limit point compact.

It is often useful to embed a space in a compact Hausdor space, because compact Hausdor spaces
are very nicely behaved. For example, I'll show later that compact Hausdor spaces are normal: Any two
disjoint closed sets can be separated by disjoint open sets. If a space is locally compact Hausdor , it can be
embedded in a compact Hausdor space called the one-point compacti
cation.
Lemma. Let X be a locally compact Hausdor space. De ne Y = X  f1g, where 1 is a point not in X .
Let T be the collection of subsets of Y consisting of:
1. Any open subset of X .
2. The complement in Y of a compact subset of X .
Then T is a topology on Y .
(Y
T ) is called the one-point compacti
cation of X .
Proof.  is open in X , so it's open in Y .  is compact, so Y = Y ;  is open in Y .
To verify the axioms for unions and intersections, it's necessary to take cases. I'll show the work for
unions the proof for intersections is similar.
The union of open sets in X is open in X , so it's open in Y .
If fCig is a family of compact subsets of X , then
 \
(Y ; Ci) = Y ; Ci:
i2I i 2I
T C is closed. Since it's contained
Now X is Hausdor , so compact subsets are closed. Therefore, S i2I i
in any one of the compact sets Ci , it's also compact. Therefore, i2I (Y ; Ci) is the complement in Y of a
compact set in X , so it's open in Y .
The remaining possibility is a union of sets open in X with complements of compact subsets of X . Using
the rst two cases, this reduces to showing that if U is open in X and C is a compact subset of X , then
U  (Y ; C ) is open in Y . But
U  (Y ; C ) = Y ; (C ; U ):

2
 C is closed in X , U is open in X , so C ; U is closed in X . But C ; U
C and C is compact, therefore
C ;U is compact. Hence, U  (Y ; C ) is a complement in Y of a compact set in X , so it's open in Y .
Here are some properties of the one-point compacti cation.
Lemma. Let X be a locally compact Hausdor space, and let Y be its one-point compacti cation.
1. The subspace topology on X is the same as the original topology on X . (That is, the inclusion of X
into Y is a homeomorphism onto its image.)
2. If X is not compact, then Y = X .
3. Y is compact.
4. Y is Hausdor .
Proof. 1. If U is open in X , then it's open in Y by de nition. Conversely, an open set in Y is either
an open set U in X (in which case U \ X = U is open in X ) or Y ; C , where C is compact in X . But
(Y ; C ) \ X = X ; C , which is open inX since C is closed in X .
2. Since the only point of Y not in X is 1, this amounts to showing that 1 is a limit point of X .
Neighborhoods of 1 are sets Y ; C , where C is compact, and since X is not compact, C 6= X . Thus, Y ; C
must intersect X , and so 1 2 X .
3. Let fUi gi2I be an open cover of Y . One of the U 's, say U0 , must contain 1, so U0 must be a set of
the form Y ; C , where C is compact in X . The sets Ui \ X for Ui 6= U0 form an open cover of C let
fUi1 \ X
: : :
Uin \ X g be a nite subcover. Then fUi1
: : :
Uin
U0 g covers Y .
4. Let x and y be distinct points of Y . If they're both in X , they may be separated by disjoint neighborhoods
in X (since X is Hausdor ), and these neighborhoods are also open in Y .
Otherwise, I'm trying to separate x 2 X from 1. Since X is locally compact, I may nd a compact set
C containing a neighborhood U of x. Then Y ; C and U are disjoint open sets in Y separating x and 1.
Corollary. A space X embeds as an open subset of a compact Hausdor space if and only if X is locally
compact Hausdor .

c 1999 by Bruce Ikenaga 3

 7-16-1999
The Countability Axioms
De nition. Let X be a topological space, and let x 2 X .
1. X has a countable basis at x if there is a countable collection of neighborhoods fBi gi2I of x such
that if U is a neighborhood of x, then Bi
U for some i.
2. X is rst countable if X has a countable basis at each point.
3. X is second countable if X has a countable basis for its topology.
Obviously, a second countable space is rst countable.

 
Example. Every metric space is rst countable. If X is a metric space and x 2 X , then B x n1  n 2 Z+
is a countable basis at x.

Example. Rn is second countable, since the balls with rational centers and radii form a countable basis for
the usual topology.
On the other hand, R with the discrete topology is not second countable.

Example. Consider R with the nite complement topology. Thus, the open sets are  and any set whose
complement is nite.
I claim that R does not have a countableTbasis at any point. Suppose, for example, that there is a
countable basis fB1 B2 : : :g at 0. I claim that 1i=1 Bi = f0g. Clearly, 0 is in the intersection I must show
that no nonzero point is in the intersection.
Suppose that x 6= 0. The complement ofTR ; fxg is nite, so it's an open set containing 0. Hence,
Bi
R ; fxg for some i. Then x 2= Bi , so x 2= 1i=1 Bi .
Thus,
\
1 1
R ; f0g = R ; Bi = (R ; Bi):
i=1 i=1
S (R ; B ) is countable. But R is uncountable, so R ; f0g is uncountable.
Each set Bi is nite, so 1 i=1 i
This contradiction shows that there is no countable basis at 0 obviously, the argument works for any x 2 R.
Thus, R is not rst countable (or second countable) in the nite complement topology.

The following results were proved for metric spaces. An examination of the proofs shows that they only
depended on the fact that metric spaces are rst countable.
Theorem. Let X and Y be topological spaces, and suppose X is rst countable.
1. Let A
X . x 2 A if and only if there is a sequence of points of A converging to x.
2. Let f : X ! Y . f is continuous if and only if whenever fxng is a convergent sequence in X , ff (xn )g is
a convergent sequence in Y .
Proposition.
1. A subspace of a rst countable space is rst countable.
1
 2. A countable product of rst countable spaces is rst countable.
3. A subspace of a second countable space is second countable.
4. A countable product of second countable spaces is second countable.
Proof. 1. Suppose X is rst countable, and suppose Y is a subspace of X . Let y 2 Y . Let fB1 B2 : : :g be
a countable basis at y in X . Then fB1 \ Y B2 \ Y : : :g is a countable basis at y in Y .
2. Suppose X1 X2 : : : are rst countable spaces. Let (xn ) 2 Q1
Q X . Let B be a countable basis for x
n=1 n n n
in Xn . Consider the collection U of product neighborhoods 1 n=1 Un , where U n =
Q Xn for all but
Q1 nitely
many n, and if Un 6= Xn , then Un 2 Bn . Then U is a countable basis at (xn ) in 1 n=1 Xn , so n=1 Xn is
rst countable.
3. Suppose X is second countable, and suppose Y is a subspace of X . Let fB1 B2 : : :g be a countable basis
for the topology on X . Then fB1 \ Y B2 \ Y : : :g is a countable basis for the subspace topology on Y .
4. Suppose X1 X2 : : : are second countable spaces. Let Q Bn be a countable basis for the topology on
Xn .Consider the collection U of product neighborhoods 1 n=1 Un , where Un = Xn for all but nitely many
nQ, 1and if Un 6= Xn , then Un 2 Bn . Then U is a countable basis for the product topology on Q1 n=1 Xn , so
n=1 X n is second countable.
De nition. Let X be a topological space.
1. X is Lindel
of if every open cover of X has a countable subcover.
2. X is separable if X has a countable dense subset.

Example. Any compact space is Lindelof, since every open cover has a nite subcover.

Example. R is separable, since Q is a countable dense subset of R.

Proposition. Let X be a second countable topological space.

1. X is Lindelof.
2. X is separable.
Proof. 1. Let U be an open cover of X . Let fB1 B2 : : :g be a countable basis for X . For each n 2 Z+ , let
Un = fU 2 U j Bn
U g:
Next, for each nonempty Un, choose Vn 2 Un.
For each x 2 X , there is a U 2 U such that x 2 U . Moreover, there is a basis element Bn such that
x 2 Bn
U . Then Un is nonempty, so Vn is de ned, and x 2 Bn
Vn.
Now x was an arbitrary point of X , so the collection fVn g covers X . fVn g is a countable subcover of U .
2. Let fB1 B2 : : :g be a countable basis for the topology of X . Choose xn 2 Bn for each n 2 Z+ . Then
fxng is a countable dense subset of X .


c 1999 by Bruce Ikenaga 2
 7-19-1999
The Separation Axioms
De nition. Let X be a space in which singletons are closed.
1. X is regular if a closed set and a point outside it can be separated by disjoint open sets.
2. X is normal if disjoint closed sets can be separated by disjoint open sets.

regular

normal

Obviously, a normal space is regular, and a regular space is Hausdor.

Terminology. Recall that a Hausdor space is also known as a T2 space.
X is a T0 space if given distinct points x y, there is a neighborhood U such that either x 2 U and y 2= U
or y 2 U and x 2= U .
X is a T1 space if given distinct points x y, there are neighborhoods U of x and V of y such that y 2= U
and x 2= V .
X is a T2 21 space if distinct points have neighborhoods whose closures are disjoint.
If you don't require that points be closed, then a space in which disjoint points and closed sets can be
separated is T3 , and a space in which disjoint closed sets can be separated is T4.
Lemma. Let X be a space in which singletons are closed.
1. X is regular if and only if for all x 2 X and every neighborhood U of x, there is a neighborhood V of
x such that V
U .
2. X is normal if and only if for every closed subset A of X and every open set U containing A, there is
an open set V containing A such that V
U .
Proof. 1. Suppose X is regular, x 2 X , and U is a neighborhood of x. Then X ; U is a closed set disjoint
from x, so by regularity I may nd disjoint open sets V containing x and W containing X ; U .
In particular, x 2 V
X ; U since X ; U is closed, V
X ; U .
Conversely, suppose that for all x 2 X and every neighborhood U of x, there is a neighborhood V of x
such that V
U . Let x 2 X and let C be a closed set which does not contain x.
X ; C is an open set containing x, so there is a neighborhood V of x such that V
X ; C . Then X ; V
is an open set containing C , and it is disjoint from the open set V containing x. Therefore, X is regular.
2. The proof for normality is the same as the proof for regularity, with the point x replaced with the closed
set A.
Corollary. If X is a regular space, then X is a T2 12 space.
Proof. Suppose X is regular. Let x and y be distinct points in X . Points in X are closed, so X ; fyg is a
neighborhood of x. By the preceding result, there is a neighborhood U of x such that U
X ; fyg. Again
by the preceding result, there is a neighborhood V of x such that V
U .
X ; U is a neighborhood of y, and X ; U
X ; U . Since X ; U is closed, X ; U
X ; U . Since
V
U and X ; U
X ; U , V and X ; U are disjoint.
1
 Thus, V is a neighborhood of x, X ; U is a neighborhood of y, and their closures are disjoint. Therefore,
X is T2 12 .
Corollary. If X is a locally compact Hausdor space, then X is regular.

Example. (Irrational slope topology 1, Example 75]) Let

X = Q Q 0
the points with rational coordinates in the closed upper half-plane. Fix an irrational number .
If x  2 R and  > 0, de ne a subset of the x-axis B (x ) by
B (x ) = fy 2 Q j jx ; yj < g:
Then for (x y) 2 X and  > 0, set



N ((x y) ) = f(x y)g  B x ; y   B x + y  :
N ((x y) ) consists of the point (x y) together with two open -intervals of rationals on the x-axis. The
intervals are centered at points on the x-axis which determine lines of slopes  and ; with the point (x y).
(x,y)

B(x - y/θ ,ε ) B(x + y/θ ,ε )

Note that if y = 0, N ((x y) ) reduces to an open interval of rationals about x in the x-axis. Moreover,
if x 2 Q, then N ((x 0) ) is a basic open set.
De ne a topology by taking as a basis all the sets N ((x y) ) for all real  > 0 and all (x y) 2 X . It's
clear that the sets cover X .
If (a b) 2 N ((x y) 1) \ N ((x y) 2 ), then either (a b) = (x y) or (a b) lies in the intersection of the
intervals in the x-axis.
In the rst case, if  = min(1 2), then N ((x y) ) is a basis element containing (a b) and contained
in N ((x y) 1 ) \ N ((x y) 2).


In the second case, (a b) = (a 0), and (a 0) 2 B x  y 1 \ B x  y 2 , and this is the intersection
of open intervals in the x-axis. Choosing a suciently small  > 0, I may nd an interval N ((a 0) ) contained
in this intersection.
If (a b) 2 N ((x1 y1) 1) \ N ((x2 y2 ) 2) where (x1 y1 ) 6= (x2 y2 ), then again (a b) must lie in the
intersection of intervals in the x-axis. As before, I may nd a basic interval containing x contained in the
intersection of the original intervals.
Therefore, the collections of sets N ((x y) ) for all real  > 0 and all (x y) 2 X forms a basis.
Next, I'll show that the closures of any two basic neighborhoods must intersect.
What is
N ((x y) )? It consists

of two diagonal strips of slopes  and ; emanating from the intervals
B x ; y  and B x + y  .

(x,y)

B(x - y/θ ,ε ) B(x + y/θ ,ε )

2
 To see this, observe that a basic

neighborhood
of a point in these strips will contain intervals in the
y
x-axis which intersect B x ;   or B x +   .y
It is clear that any two such pairs of strips must intersect. Therefore, the closures of any two basic
neighborhoods must intersect. By the result above, X is not regular.

I proved a special case of the rst part of the following proposition I'm repeating the proof of the second
part for the sake of completeness.
Proposition.
1. A product of Hausdor spaces is Hausdor.
2. A subspace of a Hausdor space is Hausdor.
Q
Proof. 1. Let fXa ga2A be a family of Hausdor spaces. Let (xa ) (ya ) be distinct points in a2A Xa . For
some index b, xb 6= yb . Choose disjoint open sets U V
Xb such that xb 2 U and yb 2 V .
De ne n n
Ua = X a if a 6= b and Va = Xa if a 6= b :
U ifQa = b V if a = b Q
Q Q
a2A Ua is a neighborhood of (xa ), a2A Va is a neighborhood of (ya ), and a2A Ua and a2A Va are
disjoint. Therefore, fXa ga2A is Hausdor.
2. Let X be a Hausdor space, and let Y
X . Let x y 2 Y . Find disjoint neighborhoods U of x and V of
y in X . Then U \ Y is a Y -neighborhood of x, V \ Y is a Y -neighborhood of y, and U \ Y and V \ y are
disjoint. Therefore, Y is Hausdor.
Proposition.
1. A product of regular spaces is regular.
2. A subspace of a regular space is regular.
Proof.
Q 1. Let fXa ga2A be a familyQof regular topological spaces. For each a, Xa is Hausdor. Therefore,
X Q in a2A Xa are closed.
a2A a is Hausdor, so points Q
Let (xa ) be a point in a2A Xa . Take a neighborhood of (xa) inQ a2A Xa replacing the neighborhood
with a smaller one if necessary, I may assume that it has the form a2A Ua , where Ua is open in Xa and
Ua = Xa for all but nitely many a's.
Since Xa is regular for each a, I may nd a neighborhood Q Va of xa in Xa such that Va
Ua .QIf a is an
index for which Ua = Xa , I will take Va = Xa as well. Then a2A Va is a neighborhood of (xa ) in a2A Xa ,
and Y Y Y
Va = Va
Ua :
a2A a2A a2A
Q
It follows by an earlier result that a2A Xa is regular.
2. Let X be a regular topological space, and let Y
X . A point of Y is closed in X , so it is closed in Y .
Let x 2 Y , and let C be a closed subset of Y which does not contain x. Write C = Y \ D, where D is
closed in X . Find disjoint open sets U and V in X such that x 2 U and D
V . Then Y \ U and Y \ V are
disjoint open sets in Y , x 2 Y \ U , and C
Y \ V . Therefore, Y is regular.
Remark. The preceding results are false for normal spaces: A subspace of a normal space need not be
normal, and a product of normal spaces need not be normal.
The following results say that the class of normal spaces is large enough to contain many interesting
spaces: speci cally, metric spaces, compact Hausdor spaces, and regular spaces having countable bases.
Theorem. Every metric space is normal.
3
Proof. Let (X d) be a metric space. Since metric spaces are Hausdor, points in X are closed.
Let C and D be disjoint closed sets in X . Since C
X ; D and X ; D is open, for every x 2 C I may
nd an open ball B (x x) contained in X ; D.
By a similar argument, for every y 2 D, there is a ball B (y y ) contained in X ; C .
Let


U = B x 2x and V = B y 2y :
 
x2C y2D
U and V are open sets, C
U , and D
V . I claim that U and
V are disjoint.

Suppose on the contrary that z 2 U \ V | say z 2 B x 2x for x 2 C and z 2 B y 2y for y 2 D.
Then
d(x y)  d(x z ) + d(y z ) < 2x + 2y :
If x  y , then
d(x y) < 2x + 2y  2y + 2y = y :
This means that x 2 B (y y ), so x 2= C , contradicting the fact that x 2 C .
Likewise, y  x implies that d(x y) < x , which in turns implies that y 2 B (x x ). This means that
y 2= D, contradicting the fact that y 2 D.
This proves my claim that U and V are disjoint. Since I've separated the closed sets C and D with
disjoint open sets U and V , it follows that X is normal.
Theorem. A compact Hausdor space is normal.
Proof. In a Hausdor space, points are closed.
I need to show that disjoint closed sets can be separated by disjoint open sets. First, recall that in a
Hausdor space, a point and a compact set that doesn't contain it can be separated by disjoint open sets.
Let C and D be disjoint closed subsets of the compact Hausdor space X . Since D is closed and X is
compact, D is compact. By the observation above, for each x 2 C I may nd disjoint open sets Ux and Vx
such that x 2 Ux and D
Vx .
fUx gx2C is an open cover of C C is compact, since it's a closed subset of a compact space. Let
fUx1 : : : Ux g be a nite subcover of C . De ne
n

n
 n
\
U= Ux and V =
i Vx :
i
i=1 i=1
U and V are open sets, C
U , and D
V . If y 2 U \ V , then y 2 Ux for some i and y 2 V
Vx , so
i i
y 2 Ux \ Vx = . This contradiction shows that U and V are disjoint.
i i
Therefore, X is normal.
Theorem. A regular space with a countable basis is normal.
Proof. Let X be a regular space with a coutable basis fBi gi2Z+ . Since X is regular, points are closed.
Let C and D be disjoint closed subsets of X . Let x 2 C . Since X ; D is an open set containing x,
I may nd a neighborhood U of x such that x 2 U
X ; D. Next, regularity implies that there is a
neighborhood V of x such that V
U . Finally, there is a basis element Bi containing x such that Bi
V .
Now Bi
V
U
X ; D, so Bi misses D.
Repeat this procedure for each x 2 C . I wind up up with a countable subcollection fCj g of fBi g which
covers C , and which satis es Cj \ D = for all j .
Likewise, I may nd a countable subcollection fDk g of fBi g which covers D, and which satis es Dk \ C =
for all k.
What I'd like to do is to take the unions of the two subcollections as my neighborhoods of C and D, but
these unions may not be disjoint. However, since the subcollections are countable, I can inductively adjust
the subcollections to produce new subcollections whose unions will be disjoint.
4
 Thus, for each j and k, de ne
j
 k

Cj0 = Cj ; Dk and Dk0 = Dk ; Cj :
k=1 j =1
Then let 1 1
 
A= Cj0 and B= Dk0 :
j =1 k=1
I claim that A and B are disjoint open sets, C
A, and D
B .
Each Cj0 and each Dk0 is open, since each is an open set minus a closed set. Therefore, A and B are
unions of open sets, so they are open.
The union of the Cj 's Scontains C . This is not changed by subtracting Dk 's, because these sets missed
C anyway. Therefore, A = 1 0
j =1 Cj contains C . Similarly, B contains D.
Finally, I'll show that A and B are disjoint. Suppose y 2 A \ B . Then y 2 Cj0 \ Dk0 for some j and k.
Without loss of generality, suppose that j  k. Now y 2 Cj0 implies y 2 Cj , but y 2 Dk0 implies y 2= Cj for
j  k. This contradiction proves that A \ B = .
Therefore, A and B are disjoint neighborhoods of C and D, and X is normal.

1] Lynn A. Steen and J. Arthur Seebach, Counterexamples in Topology. New York: Holt, Rinehart, and
Winston, Inc., 1970. ISBN 0-03-079485-4]

c 1999 by Bruce Ikenaga 5

 7-22-1999
Urysohn's Lemma
Urysohn's lemma is often expressed by saying that disjoint closed sets in a normal space can be
separated by a continuous function | that is, there is a continuous, real-valued function which is 0 on one
of the closed sets and 1 on the other.
Note that if A and B are disjoint closed sets in a topological space X , then X ; B is an open set
containing A. Conversely, if A is closed and U is an open set containing A, then A and X ; U are disjoint
closed sets.
Thus, Urysohn's lemma can be expressed in another form: In a normal space, given a closed set and an
open set containing it, there is a continuous, real-valued function which is 0 on the closed set and 1 outside
the open set. It is this version that I'll prove the discussion above shows that the other version is an easy
corollary.
The idea is that, for such a function to exist, there should be \level sets" | a set where the function is
equal to 41 , a set where the function is equal to 21 , and so on from 0 to 1. Going the other way, maybe I can
dene such a function by constructing the level sets.
U

A
1/4
1/2
3/4

There are various ways of constructing the level sets they dier in the way they index the level sets. I
will take the approach of 1] and 2], which index the level sets using the dyadic rationals in 0 1] | the
rationals which can be written in the form 2kn . The rst step is to show that, given an appropriate collection
of such level sets, one may dene a continuous function by using the level set indices in the obvious way.
Lemma. Let X and Y be topological spaces, let f : X ! Y , and let S be a subbasis for the topology on Y .
f is continuous if and only if f ;1 (U ) is open in X for every U 2 S .
Proof. Since subbasic sets are open, if f is continuous, then f ;1 (U ) is open in X for every U 2 S .
Conversely, suppose that f ;1 (U ) is open in X for every U 2 S . An arbitrary open set in Y is a union
of nite intersections of elements of S . Since f ;1 preserves arbitrary unions and arbitrary intersections, the
inverse image of an arbitrary open set in Y is open in X . Therefore, f is continuous.
Lemma. Let X be a topological space, and let
k


S = 2n
n  0 0  k  2 : n

Suppose that for each s 2 S , there is an open set Us in X such that if r s 2 S and r < s, then Ur  Us .
The function f : X ! R dened by

f (x) = inf
1
Rfr 2 S j x 2 Ur g ifif xx 22= UU1
1

is continuous.
1
 Note that some elements of S are represented by more than one fraction of the form 2kn | but repetitions
are eliminated by implication, since S is a set.
Proof. The open rays (;1 a), (b +1) for a b 2 R form a subbasis for the standard topology on R. It
therefore suces to show that f ;1 ((;1a)) and f ;1 ((b +1)) are open for all a b 2 R.
Consider the set f ;1 ((;1 a)). The range of f lies in 0 1]. Thus, if a > 1, then f ;1 ((;1 a)) = X ,
and if a  0, then f ;1 ((;1 a)) = . In either case, f ;1 ((;1 a)) isSopen.
Suppose then that 0 < a  1. I will show that f ;1 ((;1 a)) = r<a Ur .
Let x 2 f ;1 ((;1 a)), so f (x) < a  1. Since f (x) is the greatest lower bound of S the indices r such
that x 2 Ur , and since f (x) < aS, there is an index r < a such that x 2 Ur . Hence, x 2 r<a Ur .
Conversely, suppose x 2 r<a Ur . Suppose x 2 Ur , where r < a. Then f (x)  r < a, so x 2
f ;1 ((;1 a)). S
Thus, f ;1 ((;1 a)) = r<a Ur , so f ;1 ((;1 a)) is open.
Now consider the set f ;1 ((b +1)). The range of f lies in (;1 1]. Thus, if b  1, then f ;1 ((b +1)) =
, and if b < 0, then f ;1 ((b +1)) = X . In either case, f ;1 ((b +1))S is open.
Suppose then that 0  b < 1. I will show that f ;1 ((b +1)) = r>b (X ; Ur ).
Let x 2 f ;1 ((b +1)), so f (x) > b. Since in addition b < 1, I may nd r s 2 S such that b < r < s <
f (x) by construction, Ur  Us . Since f (x) is the greatest lower bound of the indices S t such that x 2 Ut,
and since f (x) > s, it follows that
S x 2
= U s . Hence, x 2
= U r , so x 2 X ; U r , and x 2 r>b (X ; Ur ).
Conversely, suppose x 2 r>b (X ; Ur ). Suppose x 2 X ; Ur , where r > b. Then f (x) 2= Ur , so the
fact that the U 's are nested implies that f (x) 2= Us for all s 2 S with s  r. Therefore, f (x)  r > b, so
x 2 f ;1 ((b +1)). S
Thus, f ;1 ((b +1)) = r>b (X ; Ur ), so f ;1 ((b +1)) is open.
Therefore, f is continuous.
Theorem. (Urysohn's Lemma) Let X be a normal space, let C be a closed subset of X , and let U be an
open set containing C . There is a continuous function f : X ! 0 1] such that f (C ) = 0 and f (X ; U ) = 1.
Proof. I'll dene a sequence of sets Ur indexed by


S = 2kn

n  0 0  k  2n
and satisfying Ur  Us for r < s.
Let U1 = U . By normality, I may nd an open set U0 such that C  U0  U0  U1 .
By normality, I may nd an open set U1=2 such that
U0  U1=2  U1=2  U1 :
By normality, I may nd open sets U1=4 and U3=4 such that
U0  U1=4  U1=4  U1=2
and
U1=2  U3=4  U3=4  U1 :
Keep going. By construction, the U 's satisfy the hypotheses of the lemma. The function f constructed
by the lemma satises the conclusion of the theorem.
Corollary. Let X be a normal space, and let A and B be disjoint closed subset of X . There is a continuous
function f : X ! 0 1] such that f (A) = 0 and f (B ) = 1.

1] Glen Bredon, Topology and Geometry.New York: Springer-Verlag New York, Inc., 1993. ISBN 0-387-
97926-3].
2] Serge Lang, Real Analysis (2nd edition). Reading, Massachusetts: Addison-Wesley Publishing Company,
Inc., 1983. ISBN 0-201-14179-5].
c 1999 by Bruce Ikenaga 2
 7-26-1999
The Tietze Extension Theorem
The Tietze Extension Theorem says that a continuous real-valued function on a closed subset of
a normal space may be extended to the entire space. I'll show that this extension condition is essentially
equivalent to normality.
The construction of the extension involves building a sequence of functions whose sum agrees with the
given function on the subspace.
Lemma. Let X be a normal space, let A be a closed subset of X , and let f : A ! R be a continuous function
satisfying jf (x)j
c for all x 2 A and some c 2 R. Then there is a continuous function g : X ! R such that
1. jg(x)j
31 c for all x 2 X .

2. jf (x) ; g(x)j
32 c for all x 2 A.

Proof. The sets
 
f ;1 13 c 1 and f ;1 ;1 ; 13 c
are disjoint closed sets in A since A
is closed, these sets are also closed in X . By Urysohn's lemma, I may
nd a continuous function g : X ! ; 13 c 31 c such that

1  1
1  1
g f ;1 ;1 ; 3 c = ; 3 c and g f ;1 3c 1 = 3 c:

By construction, jg(x)j
13 c for all x 2 X .

Now consider cases. If ;c
f (x)
; 31 c, then g(x) = ; 13 c, so jf (x) ; g(x)j
23 c.
If 31 c
f (x)
c, then g(x) = 31 c, so jf (x) ; g(x)j
32 c.
Finally, suppose ; 31 c < f (x) < 13 c. Since ; 13 c
g(x)
31 c, I again have jf (x) ; g(x)j
23 c.
Theorem. Let X be Hausdor. The following are equivalent:
1. X is normal.
2. If A is a closed subset of X , any continuous function f : A ! R extends to a continuous function
F : X ! R.
To say that F extends f means that F jA = f .
Remark. R may be replaced by a closed interval a b].
Proof. (2 ) 1) Suppose that if A is a closed subset of X , any continuous function f : A ! R extends to a
continuous function F : X ! R. Let C and D be disjoint closed subsets of X . Dene f : C  D ! R by
n
f (x) = 01 ifif xx 22 CD :
C  D is closed, so by assumption, I may extend f to a function F : X ! R. Pick disjoint open sets U
and V in R such that 0 2 U and 1 2 V . Then F ;1(U ) and F ;1(V ) are disjoint open sets in X , C  F ;1(U ),
and D  F ;1(V ). Therefore, X is normal.
1
(1 ) 2) Suppose rst that jf (x)j
c for all x 2 A and some c 2 R. I'll extend f to F : X ! R such that
jF (x)j
c.
By the Lemma, I may nd a continuous function g0 : X ! R such that
jg0(x)j
31 c for all x 2 X
jf (x) ; g0(x)j
23 c for all x 2 A:
Next, apply the Lemma to f ; g0 : A ! R to nd a function g1 : X ! R such that
jg1(x)j
13  23 c for all x 2 X
jf (x) ; g0(x) ; g1(x)j
23  23 c for all x 2 A:
Suppose n > 1 and gn;1 : X ! R has been dened and satises
n;1 
jgn;1(x)j
13  23 c for all x 2 X
2 n
jf (x) ; g0(x) ; g1(x) ; ; gn;1(x)j
3 c for all x 2 A:
Apply the Lemma to nd gn : X ! R satisfying
1 2 n
jgn(x)j
3  3 c for all x 2 X
n+1
jf (x) ; g0(x) ; g1(x) ; ; gn(x)j
32 c for all x 2 A:
X
1
I want to dene F (x) = gn (x). I must show that this series converges. To see this, note that
 n =0
2 n c, and that X 1 1 2 n
jgn(x)j
31  3 
3 3 c is a convergent geometric series. Therefore, P1n=0 gn(x)
n=0
converges absolutely by direct comparison.
Next, I want to show that F (x) is continuous. Let x 2 X , and let
X
k
sk (x) = gn (x)
n=0
be the kth partial sum.
If k > j , then
X X   X  k
X Xk 1 2 n
 k j
  k 
jsk(x) ; sj (x)j =  gn(x) ; gn(x) =  gn(x)
jgn(x)j

   c=
n=0 n=0 n=j +1 n=j +1 n=j +1 3 3
0 k;j 1
2  !

1  2 j +1 c  k;X (j +1)   BB 1 ; 3 CCC = 2 j+1 ; 2 k+1 c:
2 n = 1  2 j +1 c  B
3 3 n=0 3 3 3 @ 1; 2 A 3 3
3
2
 Fix j and let k ! 1: j+1
jF (x) ; sj (x)j
3 c: 2

Since the right side goes to 0 as j ! 1 independent of x, it follows that the partial sums converge
uniformly to F (x). Since the partial sums are continuous, it follows that F (x) is continuous.
For x 2 A, I have  
f (x) ; X k

2 k+1 c:
 g (
n x )  3
n=0
As k ! 1, the right side goes to 0, and the sum goes to F (x). Thus, F agrees with f on A.
Before I consider the case where f is unbounded, I need to observe something about the preceding
construction. I claim that if jf (x)j < c, I can construct an extension F 0(x) satisfying jF 0(x)j < c.
To see this, rst construct F (x) as above. Note that
X  X n
jF (x)j =  gn(x)
jgn(x)j
X 31  32 c = c:
1 1 1

n=0 n=0 n=0
Let B = F ;1 (f;cg  fcg). This is a closed set disjoint from A, since if x 2 A, jF (x)j = jf (x)j < c. Let
 be a continuous function such that (A) = 1 and (B ) = 0. Dene F 0(x) = (x)  F (x). F 0 still extends
f , but now jF 0(x)j < c.
Now I'll consider the unbounded case where f : A ! R.
Let  : R ! ; 2 2 be the arctangent function (x) = arctan x. Then   f : A ! ; 2 2 satises
j  f (x)j < 2 . Construct an extension F : X ! ; 2 2 . Then ;1  F : X ! R is a continuous function
which extends f .

c 1999 by Bruce Ikenaga 3