5 Lecture Notes in General Topology

Lecture Notes in General Topology
Lectures by Dr Sheng-Chi Liu
Throughout these notes, signifies end proof, N signifies end of exam-

ple, and marks the end of exercise.
Much of the theory herein is made easier to understand and more
sensible by judicious use of pictures and graphs. The author of these notes
is lazy and has not included such pictures. The reader is encouraged to
try to draw them for themselves.
Table of Contents
Table of Contents i
Lecture 1 Laying the groundwork 1

1.1 Topological space . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Lecture 2 Basis 3
2.1 Basis for a topology . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Subbasis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Lecture 3 Many basic notions 5

3.1 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Interior and closure . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 Hausdorff space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.4 Finite product spaces . . . . . . . . . . . . . . . . . . . . . . . . . 7
Lecture 4 Continuity 8
4.1 Continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2 Homeomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.3 Constructing continuous functions . . . . . . . . . . . . . . . . . 10
Lecture 5 Infinite product spaces 12

5.1 Box topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.2 Product topology . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.3 Metric topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4 Normed linear spaces . . . . . . . . . . . . . . . . . . . . . . . . . 14
Lecture 6 Connectedness and compactness 16

6.1 Norms on finite dimensional vector spaces . . . . . . . . . . . . . 16
6.2 Uniform convergence . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.3 Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Notes by Jakob Streipel. Last updated December 4, 2020.
i
ii TABLE OF CONTENTS
Lecture 7 Different kinds of connectedness 20

7.1 More connectedness . . . . . . . . . . . . . . . . . . . . . . . . . 20
7.2 Path connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7.3 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Lecture 8 Tychonoff’s theorem 24

8.1 Open maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
8.2 The Tychonoff theorem, finite case . . . . . . . . . . . . . . . . . 24
8.3 Axiom of choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
8.4 Zorn’s lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Lecture 9 Tychonoff’s theorem 27

9.1 The infinite case . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
9.2 Other notions of compactness . . . . . . . . . . . . . . . . . . . . 28
Lecture 10 Compactnesses 31
10.1 Proof continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
10.2 Uniform continuity . . . . . . . . . . . . . . . . . . . . . . . . . . 32
10.3 Local compactness . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.4 One-point compactification . . . . . . . . . . . . . . . . . . . . . 33
Lecture 11 Countability and separation axioms 34

11.1 Countability axioms . . . . . . . . . . . . . . . . . . . . . . . . . 34
11.2 Separation axiom . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
11.3 Urysohn’s lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Lecture 12 Urysohn’s lemma 38

12.1 Proof of Urysohn’s lemma . . . . . . . . . . . . . . . . . . . . . . 38
12.2 The Urysohn metrisation theorem . . . . . . . . . . . . . . . . . 41
Lecture 13 Urysohn’s metrisation theorem 41

13.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Lecture 14 Completely regular space and Stone–Čech compact-

ification 44
14.1 Completely regular space . . . . . . . . . . . . . . . . . . . . . . 44
14.2 Stone–Čech compactification . . . . . . . . . . . . . . . . . . . . 46
Lecture 15 Analysis 47
15.1 Stone–Čech compactification . . . . . . . . . . . . . . . . . . . . 47
15.2 Selected topological results in analysis . . . . . . . . . . . . . . . 49
15.3 Baire category theorem . . . . . . . . . . . . . . . . . . . . . . . 50
Lecture 16 Baire category theorem 50

16.1 Baire space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
16.2 Applications of the Baire category theorem . . . . . . . . . . . . 52
Lecture 17 Quotient spaces and topological groups 55

17.1 Quotient spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
17.2 Topological group . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
TABLE OF CONTENTS iii
Lecture 18 Topological groups 58

18.1 Topological groups . . . . . . . . . . . . . . . . . . . . . . . . . . 58
18.2 Orbit spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Lecture 19 Homotopy theory 61

19.1 Homotopy of paths . . . . . . . . . . . . . . . . . . . . . . . . . . 61
19.2 Fundamental group . . . . . . . . . . . . . . . . . . . . . . . . . . 62
19.3 Changing base point . . . . . . . . . . . . . . . . . . . . . . . . . 63
Lecture 20 Covering spaces 63

20.1 Simply connected space . . . . . . . . . . . . . . . . . . . . . . . 63
20.2 Induced homomorphism . . . . . . . . . . . . . . . . . . . . . . . 64
20.3 Covering spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
20.4 Lifting and universal covering spaces . . . . . . . . . . . . . . . . 64
20.5 Existence of path lifting . . . . . . . . . . . . . . . . . . . . . . . 65
Lecture 21 Universal covering space 66

21.1 Lifting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Lecture 22 Homotopy type 68

22.1 Fundamental group of the unit circle . . . . . . . . . . . . . . . . 68
22.2 Homotopy type . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Lecture 23 Fundamental group calculations 71

23.1 Fundamental group of the punctured plane . . . . . . . . . . . . 71
23.2 Fundamental group of S n . . . . . . . . . . . . . . . . . . . . . . 71
23.3 nth homotopy group πn (X) . . . . . . . . . . . . . . . . . . . . . 73
23.4 Fundamental group of the torus . . . . . . . . . . . . . . . . . . . 73
Lecture 24 Retraction 74
24.1 Fundamental group of projective space . . . . . . . . . . . . . . . 74
24.2 Brouwer fixed-point theorem . . . . . . . . . . . . . . . . . . . . 74
Lecture 25 Brouwer fixed-point theorem 76

25.1 Brouwer fixed-point theorem . . . . . . . . . . . . . . . . . . . . 76
25.2 Application of the Brouwer fixed-point theorem . . . . . . . . . . 77
25.3 The Borsuk–Ulam theorem . . . . . . . . . . . . . . . . . . . . . 78
Lecture 26 The Borsuk–Ulam theorem 79

26.1 The Borsuk–Ulam theorem . . . . . . . . . . . . . . . . . . . . . 79
26.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
26.3 Invariance of domain . . . . . . . . . . . . . . . . . . . . . . . . . 80
Lecture 27 Invariance of domain 81

27.1 Invariance of domain . . . . . . . . . . . . . . . . . . . . . . . . . 81
Index 84
iv TABLE OF CONTENTS
LAYING THE GROUNDWORK 1
Lecture 1 Laying the groundwork

There are two main purposes of topology:
First: Classify geometric objects. For instance, are [0, 1], (0, 1), and the real
line the same? Are they different?
Or in higher dimension, compare the closed square [0, 1] × [0, 1], the open
square (0, 1) × (0, 1) and R2 .
Or in three dimensions, compare a sphere and a cube; a doughnut (torus)
and a double doughnut (double torus).
Of course to answer this we must first determine what we mean when we
say two objects are the same or different.
Second: To do analysis. Some recap: we talk about functions f : R → R or
f : I → R, I ⊂ R, being continuous. Likewise f : Rn → Rn , and so on.
In many fields we have occasion to talk about functions on more abstract
spaces. Say, f : X → Y : what does it mean for f to be continuous in this
setting?
For example, we might walk about X = Q, or the Galois group of Q/Q, or
an algebraic variety, and so on.
Experience with analysis says that, in order to talk about abstract continuity
in a sensible way, we need to consider a certain collection of subsets of X.
1.1 Topological space

Definition 1.1.1 (Topology). Let X be a set. A collection T of subsets of X
is a topology for X if T has the following properties:
(i) ∅ and X are in T ,
(ii) any union of sets in T is in T , and
(iii) any finite intersection of sets in T is in T .
Definition 1.1.2 (Topological space, open set, (open) neighbourhood). A topo-

logical space is a pair (X, T ) where X is a set and T is a topology for X.
The sets in T are called open sets.
An open set U ∈ T is called an (open) neighbourhood of x ∈ X if x ∈ U .
Some trivial examples.
Example 1.1.3 (Trivial topology). If T = ∅, X , then T is a topology on

X called the trivial topology or indiscrete topology.

This is useless since there is no structure at all—there are no open sets
distinguishing any points. N
Example 1.1.4. The set T composed of all subsets of X is called the discrete
topology on X.
This is useless for a different reason: everything is an open set, so every set
in it looks geometrically the same. N
Our first nontrivial example:

Date: August 24th, 2020.
2 LAYING THE GROUNDWORK

Example 1.1.5. Let X = a, b, c . There are many different topologies for
X. We have as before T1 = ∅, X , and
T2 = ∅, X, a , b , b , a, b , a, c , b, c ,

but there are more.

Clearly(?) (X, T1 ) 6= (X, T2 ), even though the underlying space X is the
same in both.
Next, let’s consider T3 = ∅, X, a , a, b . Clearly the first axiom of

a topology is satisfied—trying the last two axioms we find that they, too, are
satisfied.
This is not immediately obvious, because:
Remark 1.1.6. Not every of subsets of X is a topology
collection for X. For in-
stance, T4 = ∅, X, a , b is not a topology, since a ∪ b = a, b 6∈
T4 , so the second axiom isviolated.
T

Similarly, 5 = ∅, X, a, b , b, c is not a topology either, since a, b ∩
b, c = b 6∈ T5 .

N
Definition 1.1.7 (Closed set). A subset S of X is defined to be closed if X \ S
is open (which of course implies there’s an underlying topology we’re working
with).
By De Morgan’s law, we therefore have the following properties of closed
sets:
(i) X \ ∅ = X and X \ X = ∅ are closed,
(ii) any intersection of closed sets is closed, and
(iii) any finite union of closed sets is closed.
The following exercises review some basic set theory which will soon come
in handy.
Exercise 1.1. State De Morgan’s laws and verify them.
Exercise 1.2. (a) What is an equivalence relation on a set?
(b) Let “∼” be an equivalence relation on a set A. Let C1 and C2 be two
equivalence classes. Show that either C1 = C2 or C1 ∩ C2 = ∅.
Exercise 1.3. Let X be a set and let T be a family of subsets U of X such that
X \ U is finite, together with the empty set ∅. Show that T is a topology. (T
is called the cofinite topology of X.)
Exercise 1.4. (a) Show that Q is countable.
(b) Is the polynomial ring Q[x] countable? Explain why?
∞
Q
Exercise 1.5. (a) Let A = 0, 1 . Define AN := Ai , where Ai = A for
i=1
i = 1, 2, . . . . Show that AN is uncountable.
(b) Show that R is uncountable.
Exercise 1.6. Let X be a topological space. Show that if U is open in X and A
is closed in X, then U \ A is open and A \ U is closed in X.
BASIS 3
Definition 1.1.8 ((Strictly) finer topology). Suppose T and U are two topolo-
gies for X. If T ⊃ U (respectively T ) U ), then we say that T is finer
(respectively strictly finer) than U .
In other words, an open set in (X, U ) is also open in (X, T ), but not
necessarily the other way around.
Example 1.1.9. Consider in R the sets I1 = (0, 1), I2 = ( 21 , 2), and I3 = (2, 3)
are open. But they are not the only open sets: we are allowed to take arbitrary
unions and finite intersections. So, for instance, I1 ∪ I3 = (0, 1) ∪ (2, 3) is also
open. N
However the open sets themselves are still the building blocks of open sets;
the open sets themselves can be complicated, like the last example, but we can
generate all of them by only looking at open intervals.
Lecture 2 Basis
2.1 Basis for a topology

In R2 ,
the basic building block of an open set is an open ball, B(a, r) = x |x −
1| < r . But this is not the only open set. The key, however, is that for any
point in an arbitrary open set, we can find an open ball around the point that
is contained in the set.
Definition 2.1.1 (Basis). Let X be a topological space. A collection B of

open subsets of X is a basis for the topology of X if every open subset of X is
a union of sets in B.
In other words, the basis generalises the set of open balls in R2 :
Theorem 2.1.2. A collection B of open subsets of a topological space X is a

basis for the topology on X if and only if for each x ∈ X and each neighbourhood
U of x, there exists V ∈ B such that x ∈ V and V ⊂ U .
Proof. For the forward direction, suppose B is a basis. Then for any open
neighbourhood U of x, [
U= Vα ,
α
where Vα ∈ B by the definition of basis. Hence, since x ∈ U , we must have

x ∈ Vα0 ∈ B for some α0 , and Vα0 ⊂ U .
For the converse, let U be an open sets of X. For each x ∈ U , by our
assumption there exists some Vx ∈ B such that x ∈ Vx ⊂ U . Hence
[
U= Vx .
x∈U
Theorem 2.1.3. A collection B of subsets of X is a basis for the topology on

X if and only if B has the following properties:
(i) Each x ∈ X lies in at least one set in B.

Date: August 27th, 2020.
4 BASIS
(ii) If U, V ∈ B, and x ∈ U ∩ V , then there exists W ∈ B such that x ∈ W

and W ⊂ U ∩ V .
Proof. In the forward direction, (i) follows from X being open. For (ii), since
U ∩ V is open, we must have some W ∈ B between x and U ∩ V by the previous
theorem.
For the opposite direction, Let T be the collection of all subsets of X that
are unions of sets in B.
By convention, we also take ∅ ∈ T , as the empty union.
If T is a topology, then by definition B is a basis for it. Hence we claim
that T is a topology for X.
First, if x ∈ X, by (i), x ∈ Ux ∈ B. Hence
[
X= Ux ∈ T ,
x∈X
so the first two conditions for a topology are satisfied.

Arbitrary union of elements in T is in T by definition of T in this case.
We need to show that a finite intersection of sets in T is also in T . Since
the intersection is finite, it suffices to show that the intersection of two sets in
T is in T .
Suppose U, V ∈ T . Since U and V are unions of sets in B, there exist
U0 , V0 ∈ B such that x ∈ U0 ⊂ U and x ∈ V0 ⊂ V . Hence x ∈ U0 ∩ V0 . Hence
by (ii), there exists some Wx ∈ B such that x ∈ Wx ⊂ U0 ∩ V0 ⊂ U ∩ V .
So [
U ∩V = Wx ,
x∈U ∩V
so by definition U ∩ V ∈ T , being the union of sets in B.
Exercise 2.1. Let B and B 0 be bases for the topologies T and T 0 respectively,
on X. Then the following are equivalent:
(i) T 0 is finer than T (i.e., T ⊂ T 0 ),
(ii) For each x ∈ X and U ∈ B containing x, there exists U 0 ∈ B 0 such that

x ∈ U0 ⊂ U.
Example 2.1.4. Let X = R. Let B be the collection of all open intervals

(a, b), a < b. The topology T with basis B is called the standard topology
on R. N
Example 2.1.5. Let X = R and B 0 be the collection of all half-open intervals

[a, b). The topology T 0 with basis B 0 is called the lower limit topology on
R. N
Note that T 0 is strictly finer than T . I.e., T 0 ) T .

To see this, notice how for each x ∈ R and x ∈ (a, b), we need to find [c, d)
such that x ∈ [c, d) ⊂ (a, b). This is clearly easy: pick and a < c ≤ x and
x < d ≤ b. That it is strictly finer is clear: you can’t always do the opposite: if
x = c, there exists no (a, b) such that c ∈ (a, b) ⊂ [c, d).
2.2 Subbasis 5
2.2 Subbasis
Exercise 2.2. Let X be a set and let C be a collection of subsets of X.
(a) There exists a unique smallest topology T on X such that C ⊂ T .
(b) Let B be the collection of subsets of X consisting of X, ∅, and all finite
intersections of sets in C . Show that B is a basis for T generated by
C.
Definition 2.2.1 (Generated topology, subbasis). The topology T is called
the topology generated by C and C is called a subbasis for T .
Lecture 3 Many basic notions

3.1 Subspaces
Definition 3.1.1 (Subspace, subspace topology). Let (X, T ) be a topological
space. Let Y be a subset of X. Then the collection
TY = Y ∩ U U ∈ T

is a topology on Y called the subspace topology or relative topology.

We call (Y, TY ) a subspace of (X, T ).
Remark 3.1.2. A (relative) open subset of Y need not be open in X.
Example 3.1.3. Take X = R, and Y = [0, ∞). The set (−1, 1) is open in X,
but the relative open set in Y is [0, 1), which is not open in X. N
Lemma 3.1.4. Let Y be a subspace of X. If U is (relatively) open in Y and Y
is open in X, then U is open in X.
Proof. By definition, U being open in Y means U = Y ∩ V for some V open in
X. Moreover Y is open in X, so by definition of topology, Y ∩ V = U is open
in X, being the finite intersection of open sets in X.
Exercise 3.1. Let Y be a subspace of X. If B is a basis for the topology on X,
then
BY = U ∩ Y U ∈ B

is a basis for the subspace topology on Y .

Theorem 3.1.5. Let Y be a subspace of X. Then a subset E of Y is (relatively)
closed in Y if and only if E is the intersection of Y and a closed subset of X.
Proof. Assume E is relatively closed in Y . In other words, Y \ E is open in Y .
Hence Y \ E = Y ∩ U for some open subset U of X.
Take complements (in X) and we get
E = Y ∩ (X \ U )
Since U is open in X, X \ U is open.
For the converse direction, assume E = Y ∩ C where C is a closed subset of
X. Hence Y \ E = Y ∩ (X \ C), where X \ C is open in X. Hence Y \ E is
relatively open in Y , making E closed in Y .
Date: September 1st, 2020.
6 MANY BASIC NOTIONS
3.2 Interior and closure

Definition 3.2.1 (Interior, closure). Let X be a topological space. Let S ⊂ X
be a subset.
(i) The interior S̊ or int(S) is defined as the union of all open sets contained
in S.
(ii) The closure S of S is defined as the intersection of all closed sets con-
taining S.
Remark 3.2.2. Since arbitrary unions of open sets is open, int(S) is open. Sim-
ilarly, arbitrary intersections of closed sets is closed, so S is closed.
Exercise 3.2. Let X be a topological space and let S ⊂ X. Then x ∈ S if and
only if every open neighbourhood U of x intersects S.
Theorem 3.2.3. Let Y be a subspace of X and let E be a subset of Y . Then
the relative closure of E in Y is E ∩ Y , where E is the closure of E in X.
Proof. Let B denote the closure of E in Y . In other words, we want to show
that B = E ∩ Y .
Since E is closed in X, the set E ∩ Y is closed in Y , and this set contains
E. Hence by definition of closure, B ⊂ (E ∩ U ).
On the other hand, B is closed in Y , meaning that B = Y ∩ C where C is
closed in X. Also, E ⊂ B. Hence E ⊂ C. Hence E ⊂ C. Take intersection,
and we have
E ∩ Y ⊂ C ∩ Y = B.
Hence B = E ∩ Y .
We can generalise the notion of a convergent sequence to topological spaces,
even if we don’t have a metric:
Definition 3.2.4 (Limit point, convergence, isolated point). Let X be a topo-
logical space.
(i) Let S ⊂ X be a subset. A point x ∈ X is called a limit point of S if
every neighbourhood of x intersects with S in some point other than x.

(ii) A sequence an ⊂ X converges to x ∈ X if for every neighbourhood
U of x, there exists some N ∈ N such that an ∈ U for all n > N .
(iii) A point s ∈ S is called an isolated
point of S if there exists a neighbour-
hood U of s such that U ∩ S = s .
Exercise 3.3. Let T be the cofinite topology on Z.
∞
(a) Show that the sequence n n=1 converges in (Z, T ) to each point in Z.
(b) Describe the convergent sequences in (Z, T ).
Proposition 3.2.5. Let X be a topological space and let S ⊂ X be a subset.
Then S is the union of the set of limit points of S and the isolated points of S.
Exercise 3.4. Prove Proposition 3.2.5.
An immediate consequence of this is:
Corollary 3.2.6. A subset S ⊂ X is closed if and only if S contains all of its
limit points.
3.3 Hausdorff space 7
3.3 Hausdorff space

Definition 3.3.1 (Hausdorff space). A topological space X is called a Haus-
dorff space or T2 -space if for each pair x1 , x2 ∈ X, x1 6= x2 , there exists
neighbourhoods U1 and U2 of x1 and x2 respectively, such that U1 ∩ U2 = ∅.
Theorem 3.3.2. Let X be a Hausdorff space. Then every single-point set is
closed.

Proof. Let E = x0 . We want to show that X \ E is open.
If x ∈ X \ E, then since X is Hausdorff there exist open sets Ux and Vx such
that x0 ∈ Ux and x ∈ Vx with Ux ∩ Vx = ∅. Hence Vx ⊂ X \ E.
Taking the union of all such x,
\
X \E = Vx ,
x∈X\E
an arbitrary union of open sets, so X \ E is open, whence E is closed.

Exercise 3.5. Show
that X is Hausdorff if and only if the diagonal set 4 =
(x, x) x ∈ X is a closed unset of X × X.
Exercise 3.6. Let A be a subset of a topological space X. Let Y be a Hausdorff
space. Suppose f : A → Y be a continuous function that has a continuous
extension to g : A → Y (i.e., g(a) = f (a) for all a ∈ A). Show that g is uniquely
determined by f .
One way to think of Hausdorff spaces is that it guarantees the existence of
small open neighbourhoods.
Theorem 3.3.3. Let X be a Hausdorff space and let S ⊂ X be a subset. Then
x ∈ X is a limit point of S if and only if every neighbourhood of x contains
infinitely many points of S.
Proof. The converse direction follows from the definition of limit point.
For the forward direction, suppose there exists a neighbourhood U of x such
that S ∩ U contains only finitely many points. Thus

U ∩ (S \ x ) = a1 , a2 , . . . , an

is some finite set. By
Theorem
3.3.2, a1 , a2 , .
. . , an , being the
finite intersec-
tion of closed sets ai , is closed. Hence X \ a1 , a2 , . . . , an is open.
Take
V := U ∩ (X \ a1 , a2 , . . . , an ).

This is open, and x ∈ V . Hence V ∩ (S \ x ) = ∅.
But this contradicts x being a limit point of S, since we found a neighbour-
hood of V that doesn’t meet S in a point different from x.
3.4 Finite product spaces

Definition 3.4.1 (Product topology). Let X1 , X2 , . . . , Xn be topological spaces.
The product topology on X1 × X2 × · · · × Xn is the topology for which a basis
B of open sets is given by
B = U1 × U2 × · · · × Un Ui is open in Xi for i = 1, 2, . . . .

8 CONTINUITY
We should check that this satisfies the criterion for being a basis in Theo-
rem 2.1.3. The first one, (x1 , x2 , . . . , xn ) ∈ X1 × X2 × · · · × Xn ∈ B, is trivial.
Secondly, if
(x1 , x2 , . . . , xn ) ∈ (U1 × U2 × · · · × Un ) ∩ (V1 × V2 × · · · × Vn ),
then the point is also in (U1 ∩ V1 ) × · · · × (Un ∩ Vn ). Each of the constituent

factors is open, and the product is open, so this is in B. Clearly, this product
of intersections is contained in the intersection of products.
Proposition 3.4.2. Let Bi be a basis for the topology on Xi , i = 1, 2, . . . , n.

Then
C = V1 × V2 × · · · × Vn Vi ∈ Bi , i = 1, 2, . . . , n

is a basis for the product topology on X1 × X2 × · · · × Xn .
Proof. We apply Theorem 2.1.2. Let x = (x1 , x2 , . . . , xn ) ∈ X1 × X2 × · · · × Xn .

Let W be an open neighbourhood of x.
Then there exists a basis element U1 × U2 × · · · × Un such that Ui is open in
Xi and xi ∈ Ui .
Since Bi is a basis for Xi , there exists Vi ∈ Bi so that x ∈ Vi ⊂ Ui . Thus
x ∈ V1 × V2 × · · · × Vn ⊂ U1 × U2 × · · · × Un ⊂ W
where V1 × V2 × · · · × Vn ∈ C . Hence by Theorem 2.1.2, C is a basis for the

product topology on X1 × X2 × · · · × Xn .
Exercise 3.7. Let X and Y be topological spaces. Define the projections π1 : X ×

Y → X by π1 (x, y) = x and π2 : X × Y → Y by π2 (x, y) = y. Then
B = π1−1 (U ) U open in X ∪ π2−1 (V ) V open in Y

is a subbasis for the topology of X × Y . So the projection maps pull back open
sets to open sets in the product topology.
In other words, the product topology makes the projection map continuous.
In a sense, this is a better description of the product topology, because it works
for infinite products too—more on that in future.
Lecture 4 Continuity
4.1 Continuous functions
Recall from analysis:
As a local condition, a function f : R → R is continuous at x0 if for any
ε > 0 there exists δ > 0 such that |x − x0 | < δ implies |f (x) − f (x0 )| < ε.
As a global condition, f : R → R is continuous on R if f is continuous at
all points in R. An important result from analysis is that this is equivalent to
f −1 (V ) is open in R for any open V ⊂ R.
This second option is the correct/productive way to generalise continuity to
arbitrary topological spaces, since all we have is a notion of openness:
Date: September 3rd, 2020.
4.1 Continuous functions 9
Definition 4.1.1 (Continuous function). Let X and Y be topological spaces.

(i) A function f : X → Y is continuous if f −1 (V ) is open in X for any open
V in Y .
(ii) A function f : X → Y is continuous at the point x0 ∈ X if for any
open subset V ⊂ Y such that f (x0 ) ∈ V , there exists an open U ⊂ X
such that x0 ∈ U and f (U ) ⊂ V .
The correspondence in the analysis case still works:
Theorem 4.1.2. Let X and Y be topological spaces. A function f : X → Y is
continuous if and only if f is continuous at every point of X.
Proof. For the forward direction, let x0 ∈ X and let V be an open neighbour-
hood of f (x0 ). Since f is continuous, f −1 (V ) is open in X. Taking U = f −1 (V ),
we see that x0 ∈ U and f (U ) ⊂ V .
For the converse, let V ⊂ Y be open. Let x ∈ f −1 (V ). Since f is continuous
at x, there exists an open neighbourhood Ux of x such that f (Ux ) ⊂ V . Let
[
U= Ux ,
x∈f −1 (V )
a union of open sets, hence open. In addition, x ∈ f −1 (V ), x ∈ Ux ⊂ U ,

so f −1 (V ) ⊂ U . Moreover, f (Ux ) ⊂ V for each x implies f (U ) ⊂ V , so
f −1 (V ) ⊃ V . Thus f −1 (V ) = U , open in X.
Theorem 4.1.3. Let X and Y be topological spaces. Let f : X → Y . Then the
following are equivalent:
(i) f is continuous.
(ii) For any subset E ⊂ X, f (E) ⊂ f (E).
(iii) For any closed C ⊂ Y , f −1 (C) is closed in X.
Proof. First, (i) implies (ii). We need to show that for x ∈ E, f (x) ∈ f (E). Let
V be an open neighbourhood of f (x)—we want to show V meets f (E). Now
∈ f −1 (V ) is an open neighbourhood of x ∈ X. Since x ∈ E, f −1 (V ) ∩ E 6= ∅.
Hence there exists some x0 ∈ f −1 (V ) ∩ E, and therefore f (x0 ) ∈ V ∩ f (E).
Thus V ∩ f (E) 6= ∅, and f (x) ∈ f (E).
Next, (ii) implies (iii). Let C ⊂ Y be closed. Set E = f −1 (C). To show that
E is closed, it suffices to show E ⊂ E. For x ∈ E, we have f (x) ∈ f (E) ⊂ f (E)
by (ii). This is contained in C = C, so f (x) ∈ C and hence x ∈ f −1 (C) = E.
Finally, (iii) implies (i). Let U ⊂ Y be open. Then C = Y \ U is closed. By
assumption f −1 (C) is closed in X. Hence
f −1 (C) = f −1 (Y \ U ) = f −1 (Y ) \ f −1 (U ) = X \ f −1 (U ),
so f −1 (U ) is open.
Proposition 4.1.4. Let X1 , X2 , X3 be topological spaces. Let f : X1 → X2 and
g : X2 → X3 be continuous. Then g ◦ f : X1 → X3 is continuous.
Proof. This is obvious: pull an open set U ⊂ X3 back to X1 via X2 .
10 CONTINUITY

Exercise 4.1. Let X be a topological space and let Eα α∈Λ be a collection of
S
subsets such that X = Eα . Let f : X → Y . Suppose f E is continuous for
α
α∈Λ
all α ∈ Λ.
(a) Show that if Λ is finite and Eα is closed for all α ∈ Λ, then f is continuous.
(b) Give an example where Λ is countable and Eα is closed for all α ∈ Λ, but
f is not continuous.

(c) Suppose Eα α∈Λ is locally finite, that is each x ∈ X has a neighbour-
hood that intersects only finitely many Eα , and Eα is closed for all α ∈ Λ.
Show that f is continuous.
4.2 Homeomorphism
Definition 4.2.1 (Homeomorphism). Let X and Y be topological spaces. Let
f : X → Y be one-to-one and onto. The map f is called a homeomorphism
if both f and f −1 are continuous.
In other words, f is a homeomorphism if is has the property that U is open
in X if and only if f (U ) is open in Y .
Definition 4.2.2 (Homeomorphic spaces). The topological spaces X and Y

are homeomorphic if there exists a homeomorphism f : X → Y .
Classifying objects and/or spaces as homeomorphic or not is an important

question in topology.
For example, the cube and the sphere are homeomorphic, but the sphere and
the torus are not. This is very much the realm of algebraic topology, assigning
an algebraic quantity/number to a topological space that is invariant under
homeomorphism.
Definition 4.2.3 (Topological property). A property of a topological space is

called a topological property if is is preserved under homeomorphism.
Example 4.2.4. Metrisability (being able to assign a metric) is a topological

property. N
Example 4.2.5. Let f : (−1, 1) → R be defined by

x
f (x) = .
1 − x2
This map is a homeomorphism, showing that (−1, 1) is homeomorphic to R. N
4.3 Constructing continuous functions

Proposition 4.3.1. Let X and Y be topological spaces.

(i) If f : X → Y is continuous and Z ⊂ X is a subspace, then f Z : X → Y
is continuous.1

1 By f we mean f restricted to Z.
Z
4.3 Constructing continuous functions 11
S
(ii) If X = Vα , Vα open and f Vα is continuous, then f : X → Y is contin-
α
uous.
Proof. (i) Let U ⊂ Y be open. We have
(f Z )−1 (U ) = f −1 (U ) ∩ Z

is open in Z (using the subspace topology) since f 1 (U ) is open in X. Hence

f Z is continuous.
(ii) Let U ⊂ Y be open. Then

[ [
Vα ⊃ f −1 (U ) = (Vα ∩ f −1 (U )) = (f Vα )−1 (U ).

X=
α α
The right-hand side is open in X since Vα is, so f −1 (U ) is open in X.
Theorem 4.3.2 (The pasting lemma). Let X = A ∪ B, where A and B are

closed in X. Let f : A → Y and g : B → Y be continuous. Suppose f (x) = g(x)
for all x ∈ A ∩ B. Then
h: X → Y
defined by h(x) = f (x) if x ∈ A and h(x) = g(x) if x ∈ B is a continuous map.
Proof. Let C ⊂ Y be closed. Since f is continuous, f −1 (C) is closed in A. Also,

A is closed in X. Hence f −1 (C) is closed in X. Similarly for g: g −1 (C) is closed
in X.
Now
h−1 (C) = f −1 (C) ∪ g −1 (C)
is closed in X since it is the union of two closed sets. Hence h is continuous.
Theorem 4.3.3. Let f : Z → X × Y be given by f (z) = (f1 (z), f2 (z)), where

f1 : Z → X and f2 : Z → Y . (These maps f1 , f2 are called coordinate func-
tions of f .) Then f is continuous if and only if f1 , f2 are continuous.
Proof. Note how the projection maps
π1 : X → Y → X, π1 (x, y) = x
π2 : X → Y → Y, π2 (x, y) = y
are continuous, since π1−1 (U ) = U × Y is open in X × Y if U is open in X.

Now this becomes easy: Assume f is continuous. Then f1 (z) = π1 (f (z)) is
continuous by Proposition 4.1.4. Similarly for f2 .
For the converse, take basis elements U × V for the topology on X × Y . We
need to show f −1 (U × V ) is open. We have
f −1 (U × V ) = f1−1 (Y ) ∩ f2−1 (V ).
Each piece is open in Z since f1 and f2 are continuous. Hence f −1 (Y × V ) is

open, so f is continuous.
12 INFINITE PRODUCT SPACES
Lecture 5 Infinite product spaces

Let xα α∈I
be an indexed family of topological spaces. Define
Y
X := Xα = x = (xα )α∈I xα ∈ Xα .
α∈I
This index set I could be finite, infinite, countable, uncountable, etc.

We want to define a topology on X.
5.1 Box topology

This is the topology generated by the basis
nY o
B= Uα Uα is open in Xα .

α∈I
This is indeed a basis for a topology on X, called the box topology.

This seems very natural, and follows the definition we gave for the product
topology of finite products. However, there’s a problem:
Remark 5.1.1. This topology contains too many open sets, and as a result it is
not very useful.
By not useful we mean that many expected things break down:
Example 5.1.2. Assume Xα are compact for all α ∈ I. The product space
Y
Xα
α∈I
need not be compact, under the box topology. N
Example 5.1.3. Assume Xα are connected for all α ∈ I. The product might
be totally disconnected, meaning the connected component contains only one
point. N
All by way of saying: we want a better topology.
5.2 Product topology

This time, we take the basis B to be all sets of the form
Y
Uα ,
α∈I
with Uα open in Xα for all α ∈ I, like with the box topology, only this time
Uα = Xα for all but finitely many α.
Hence the topology on X generated by B, called the product topology,
has much fewer open sets than the box topology (if I is infinite), i.e.:
Remark 5.2.1. If the index set I is finite, then the box topology and the product
topology coincide.
Date: September 10th, 2020.
5.3 Metric topology 13
Remark 5.2.2. Since this is much more useful than the box topology, we shall
assume that Y
X= Xα
α
is endowed with the product topology unless otherwise specified.

Define the projection map
πα : X → Xα
by πα (x) = xα , where x = (xα )α∈I .

Then
B = U U = πα−1 (Uα )∩πα−1 (Uα2 )∩· · ·∩πα−1

1 2 m
(Uαm ) for finitely many α1 , . . . , αm .
Hence πα−1 (Uα ) is open in X, meaning:

Q
Proposition 5.2.3. (i) The product topology on Xα is the smallest topol-
α∈I
ogy such that πα , α ∈ I, is continuous.
Q
(ii) If Xα is Hausdorff for all α ∈ I, then Xα is also Hausdorff in both the
α∈I
box topology and the product topology.
(iii) Let Y
f: Z → Xα
α∈I
be given by f (z) = (fα (z))α∈I where fα : Z → Xα . Then f is continuous

if and only if fα is continuous for all α ∈ I.
5.3 Metric topology

Definition 5.3.1 (Metric). A metric on a set X is a function d : X × X → R
that has the following properties:
(i) d(x, y) ≥ 0 for all x, y ∈ X and d(x, y) = 0 if and only if x = y;
(ii) d(x, y) = d(y, x), i.e., d is symmetric; and
(iii) d(x, z) ≤ d(x, y) + d(y, z) for all x, y, z ∈ X, i.e., d satisfies the triangle
inequality.
Given ε > 0, x ∈ X, define

B(x, ε) = y ∈ X d(x, y) < ε ,
the ε-ball centred at x.

Definition 5.3.2 (Metric topology). Let d be a metric on a set X. The collec-
tion of B(x, ε) for all x ∈ X and ε > 0 is a basis for a topology on X, called the
metric topology on X induced by d.
Definition 5.3.3 (Metric space). Let X be a topological space.
14 INFINITE PRODUCT SPACES
(i) X is called metrisable if there exists a metric d on X that induces the

topology on X.
(ii) A metric space X is a metrisable space X together with a a specific
metric d that induces the topology.
So the metrisability of a topological space, whether a metric exists, is a
topological problem. On the other hand, if you start with a metric space,
understanding that space is an analysis problem.
Exercise 5.2. Assume that X is metrisable and X is homeomorphic to Y . Show
that Y is also metrisable (i.e. metrisability is a topological property).
Definition 5.3.4 (Bounded). Let (X, d) be a metric space. A subset S ⊂ X is
bounded if there exists some M > 0 such that d(x, y) ≤ M for all x, y ∈ S.
Proposition 5.3.5. The limit of a convergent sequence in a metric space is
unique.
Proof. This is analysis: Suppose xn → x and xn → y. We have
d(x, y) ≤ d(x, xn ) + d(xn , y) → 0
by the triangle inequality, meaning that d(x, y) = 0, so x = y.
5.4 Normed linear spaces

Many useful metric spaces are vector spaces endowed with a metric that arises
from a norm, e.g., Rn .
Definition 5.4.1 (Norm). Let X be a vector space over F = R or C. A norm
is a function k·k : X → R such that
(i) kxk ≥ 0 for all x ∈ X and kxk = 0 if and only if x = 0;
(ii) kcxk = |c|kXk for all x ∈ X and c ∈ F ; and
(iii) we have the triangle inequality,
kx + yk ≤ kxk + kyk
for all x, y ∈ X.
Norms induce metrics:
Lemma 5.4.2. Let (X, k·k) be a normed linear space. Then d(x, y) := kx − yk
for x, y ∈ X is a metric on X.
Proof. Clearly d(x, y) ≥ 0 and d(x, y) = 0 if and only of kx − yk = 0 if and only
if x − y = 0 if and only if x = y.
The metric is symmetric since
d(x, y) = kx − yk = k(−1)(y − x)k = |−1|ky − xk = d(y, x).
Finally
d(x, z) = kx − zk = k(x − y) + (y − z)k

≤ kx − yk + ky − zk = d(x, y) + d(y, z).
5.4 Normed linear spaces 15
Definition 5.4.3 (Normed topology). Let (X, k·k) be a normed linear space.
The topology induced by the metric d(x, y) = kx − yk is called the normed
topology of x induced by k·k.
Example 5.4.4. The Euclidean space Rn is a dimension n vector space over
R. There are several common norms and metrics on this:
(i) The Euclidean norm and metric: For x = (x1 , x2 , . . . , xn ), we have the
usual Euclidean norm
kxk = (x21 + x22 + · · · + x2n )1/2 ,
which induces the usual Euclidean metric

1/2
d(x, y) = kx − yk = (x1 − y2 )2 + · · · + (xn − yn )2 .
(ii) The p-norm, k·kp , for 1 ≤ p < ∞, defined by
kxkp = (|x1 |p + · · · + |xn |p )1/p ,
so that for p = 2, k·k2 = k·k.

(iii) The supremum norm, k·k∞ :
kxk∞ = max|xi |. N
i
Here’s a natural and interesting question: we have one and the same set
X = Rn , with three different kinds of norms and metrics. How do their open
sets, in the induced metric topologies, relate?
The answer is that the topologies are the same, which we will discover below.
Definition 5.4.5 (Equivalent norms). Let k·k1 and k·k2 be two norms on
a linear space X. We say that k·k1 and k·k2 are equivalent if there exist
constants c and k such that
ckxk1 ≤ kxk2 ≤ kkxk1
for all x ∈ X. Note that the constants c and k do not depend on x.

Theorem 5.4.6. Let k·k1 and k·k2 be two norms on a linear space X. Then
k·k1 and k·k2 are equivalent if and only if they induce the same topology on X.
Proof. Assume the norms are equivalent, i.e.,
ckxk1 ≤ kxk2 ≤ kkxk1
for all x ∈ X. Then for any ε > 0,

ε n εo
B1 x, = y ∈ X kx − yk1 < ⊂ y ∈ X kx − yk2 < ε = B2 (x, ε).

k k
Similarly,
B2 (x, cε) ⊂ B1 (x, ε).
Given B2 (x0 , ε), open in (X, k·k2 ), we want to show that B2 (x0 , ε) is open in
(X, k·k1 ).
16 CONNECTEDNESS AND COMPACTNESS
For y ∈ B2 (x0 , ε), there exists εy > 0 such that B2 (y, εy ) ⊂ B2 (x0 , ε) since
the ball is open in k·k2 .
ε
We know B1 (y, ky ) ⊂ B2 (y, εy ), where notably the first ball is open in
(X, k·k1 ). Thus ε
y
[
B2 (x0 , ε) = B1 y,
k
y∈B2 (x0 ,ε)
is open in (X, k·k1 ).

Similarly, any open set in (X, k·k1 ) is also open in (X, k·k2 ). Hence the two
spaces have the same topology.
For the converse, assume the topologies are the same. Consider B1 (0, 1) =
x ∈ X kxk1 < 1 . This is open in both (X, k·k1 ) and (X, k·k2 ) since the
topologies are equal.
In particular, there exists r > 0 such that

B2 (0, r) = x ∈ X kxk2 < r ⊂ B1 (0, 1).
In other words, if kxk2 < r, then kxk1 < 1. Take C = 2r . Then if kxk2 ≤ C
implies kxk1 < 1.
cx
Then for any x 6= 0 in X, consider y = kxk 2
. Then kyk2 = c. Hence
kyk1 < 1, so cx
< 1,

kxk2

which rearranged means

ckxk1 < kxk2
for all x 6= 0 (the x = 0 case is trivial).
The same argument shows
kxk2 ≤ kkxk1
for some k > 0. Hence the norms k·k1 and k·k2 are equivalent.
Lecture 6 Connectedness and compactness

6.1 Norms on finite dimensional vector spaces
Theorem 6.1.1. If X is a finite dimensional vector space, then any two norms
on X are equivalent.

Proof. Let e1 , e2 , . . . , en be a basis for X over the field F . For
n
X
x= ci ei ,
i=1
ci ∈ F , define the sup-norm

kxk∞ := max |c1 |, |c2 |, . . . , |cn | .

6.1 Norms on finite dimensional vector spaces 17
Let k·k be any norm on X. We want to show that k·k is equivalent to k·k∞ .
If so, then any two norms on X induce the same topology on X, which in turn
implies any two norms are equivalent.
By the triangle inequality,
Xn X n
kxk = ci ei ≤ |ci |kei k ≤ Kkxk∞

i=1 i=1
for all x ∈ X by taking

n
X
K= kei k.
i=1
In other words,
kxk ≤ Kkxk∞
for all x ∈ X.
Next we need to show that there exists some c > 0 such that ckxk∞ ≤ kxk
for all x ∈ X. This requires two things from analysis which we will also discuss
later:
Definition 6.1.2 (Compact set). A set is compact if every open cover has a
finite subcover.
A continuous function f : X → R has a maximum and a minimum on a

compact set.
With this in mind, let B = xkxk∞ = 1 . Then B is compact in (X, k·k∞ ).
Now we claim that k·k : (X, k·k∞ ) → R is continuous under k·k∞ .
To see this, given ε > 0 we need
to show that there exists δ > 0 such that
kx − yk∞ < δ implies kxk − kyk < ε. By the triangle inequality,

kxk − kyk ≤ kx − yk < Kkx − yk∞ < ε
ε
by taking δ = K . Since k·k : X → R is continuous and B is compact with
respect to k·k∞ , the norm k·k must have a minimum on B, say c > 0.
In other words, if kxk∞ = 1, then kxk ≥ c. Then for any nonzero x ∈ X,
consider y = kxkx ∞ . Then kyk∞ = 1, so kyk ≥ c, whereby
x
≥ c,

kxk∞

or in other words
kxk ≥ ckxk∞ .
We leave it as an exercise to check the two details left out in the above
argument:
Exercise 6.1. Verify that k·k∞ is indeed a norm on X.

Exercise 6.2. Verify that B = x kxk∞ = 1 is compact in (X, k·k∞ ).
Corollary 6.1.3. All norms on Rn are equivalent. I.e., Rn has a unique norm
topology.
18 CONNECTEDNESS AND COMPACTNESS
6.2 Uniform convergence

Definition 6.2.1 (Uniform convergence). Let X be a topological space and
(Y, d) be a metric space. Let fn : X → Y , n = 1, 2, 3, . . . . We say fn
converges uniformly to a function f : X → Y if for any ε > 0 there exists
N ∈ N such that
d(fn (x), f (x)) < ε
for all n ≥ N and all x ∈ X.
Theorem 6.2.2. Let X be a topological space and (Y, d) be a metric space. Let
fn : X → Y be continuous, n = 1, 2, 3, . . . . Suppose fn converges uniformly
to f . Then f is continuous.
Proof. Let V be open in Y . We need to show f −1 (V ) is open in X. It suffices

to show that for any x0 ∈ f −1 (V ), there exists a neighbourhood U of x0 such
that U ⊂ f −1 (V ), which is equivalent to f (U ) ⊂ V .
Let y0 = f (x0 ) ∈ V in Y . Choose ε > 0 such that B(y0 , ε) ⊂ V , which is
doable since V is open. Since fn → f uniformly, there exists N ∈ N such that
d(fn (x), f (x)) < 4ε for all n ≥ N and all x ∈ X.
Since fN is continuous at x0 , there exists a neighbourhood U of x0 such that
fN (U ) ⊂ B(fN (x0 ), 2ε ) ⊂ V .
We claim that f (U ) ⊂ B(y0 , ε) ⊂ V . This is easy: it’s the triangle inequality!
For x ∈ U ,
d(f (x), f (x0 )) ≤ d(f (x), fN (x)) + f (dN (x), fN (x0 )) + d(fN (x0 ), f (x0 ))
ε ε ε
< + + = ε.
4 2 4
Hence f (x) ∈ B(y0 , ε).
We know from analysis that if the convergence is not uniform, the limiting
function needn’t be continuous,
but it might be continuous on some set.
So, question: Let fn be a sequence of continuous functions. Suppose
fn (x) → f (x) pointwise for all x ∈ X. How large is the set

x ∈ X f is continuous at x ?
The answer is this: if X is a compact Hausdorff space or a complete metric

space, then this set is dense in X. We will come back to this problem later.
(This is a consequence of the Baire category theorem.)
6.3 Connectedness
Definition 6.3.1 (Connected, disconnected). (i) A topological space X is
called disconnected if there exist open sets U and V such that X = U tV ,
i.e. U, V 6= ∅ and U ∩ V = ∅.
(ii) X is connected if it is not disconnected.
Remark 6.3.2. If X is disconnected, then the U and V in question are also closed
sets, since U = X \ V .
We immediately get the following characterisation:
6.3 Connectedness 19
Proposition 6.3.3. A topological space X is connected if and only if the only

subsets of X that are both open and closed are ∅ and X.
Proof. Assume X is connected, and suppose there exists a set U 6= ∅, X that
is both open and closed. Then V = X \ U is open and closed, and X = U ∪ V ,
and by construction U ∩ V = ∅. Hence X is disconnected, a contradiction.
Suppose X is disconnected. In other words, there exist open U, V 6= ∅ with
U ∩ V = ∅ and X = U ∪ V . Then U = X \ V is closed, being the complement
of an open set. Hence U is both open and closed, a contradiction.
Definition 6.3.4 (Connected subspace). A subset of a topological space is
connected if it is connected in the relative topology.
Example 6.3.5. Consider Q ⊂ R. Then Q is disconnected.
To see this, take any irrational number a ∈ R \ Q. Set U = Q ∩ (a, ∞) and
V = Q ∩ (−∞, a). Then U and V are open in Q by the definition of the relative
topology, they don’t meet, and Q = U ∪ V . N
Remark 6.3.6. The only connected subsets of Q are singletons, or one-point sets.
Thus, Q is totally disconnected.
Definition 6.3.7 (Totally disconnected space). A topological space is totally
disconnected if the only connected subsets are one-point sets.
Exercise 6.3. Show that any countable metric space is totally disconnected.
Theorem 6.3.8. Let f : X → Y be a continuous function. Suppose X is con-
nected. Then f (X) is connected.
Proof. Let E ⊂ f (X) be both open and closed. We need to show E = ∅ or
E = f (X).
Since f is continuous, f −1 (E) is open and closed in X. Since X is connected,
f −1 (E) = ∅ or f −1 (E) = X, and we are done.

Theorem 6.3.9. Let Eα α∈Λ be a family of connected subsets of X. Assume
Eα ∩ Eβ 6= ∅ for all α, β ∈ Λ. Then
[
Eα
α∈Λ
is connected.
Proof. Let [
E= Eα
α∈Λ
and let F 6= ∅ be a subset of E that is both open and closed. We need to show
that F = E, which is equivalent to Eα ⊂ F for all α.
Let x ∈ F . Then x ∈ Eα0 for some α0 ∈ Λ since F ⊂ E.
Now F ∩Eα0 6= ∅, since they meet in at least x, and this set is both open and
closed in Eα0 . But Eα0 is by assumption connected, meaning F ∩ Eα0 = Eα0 is
the only option. In other words, Eα0 ⊂ F .
For any β ∈ Λ, F ∩ Eβ is open and closed in Eβ . But this can’t be empty,
since Eα0 ⊂ F . I.e.,
(Eα0 ∩ Eβ ) ⊂ (F ∩ Eβ )
implies F ∩ Eβ 6= ∅, so F ∩ Eβ = Eβ . Hence Eβ ⊂ F for all β ∈ Λ. Thus
E = F.
20 DIFFERENT KINDS OF CONNECTEDNESS
Definition 6.3.10 (Connected component). Let X be a topological space and

x ∈ X. The connected component of x, denoted C(x), is the union of all
connected subsets that contain x.
From Theorem 6.3.9 we get immediately:
Corollary 6.3.11. Two connected components of X either coincide or are dis-
joint.
Exercise 6.4. A connected component of X is always closed.
Lecture 7 Different kinds of connectedness

7.1 More connectedness
Theorem 7.1.1. (i) A finite product of connected spaces is connected.

(ii) Let Xα α∈Λ be a family of connected spaces. Then
Y
X := Xα
α∈Λ
is connected. (By convention, this product has the product topology.)

Proof. (i) It suffices to consider the product of two connected spaces X and Y .
The result then follows from induction.
Fix y0 ∈ Y . For x ∈ X, let

Tx (X × y0 ) ∪ ( x × Y ).

Then X × y0 and x × Y are connected since X and Y are (they’re home-
omorphic, we’ve just added a single point), and

(X × y0 ) ∩ ( x × Y ) = (x, y0 ) 6= ∅.
By Theorem 6.3.9, this implies Tx is connected. Notice how

[
X ×Y = Tx
x∈X

and for any x1 , x2 ∈ X, Tx1 ∩ Tx2 6= ∅ (they meet in X × y0 ). Hence by
Theorem 6.3.9 X × Y is connected.
(ii) First fix a base point b = (bα )α∈Λ ∈ X.

For any finite subset α1 , α2 , . . . , αn ⊂ Λ, define

X ⊃ X(α1 , α2 , . . . , αn ) = (xα )α∈Λ xα = bα for α 6= α1 , α2 , . . . , αn .
So, up to order of multiplication, this looks like

Xα1 × Xα2 × · · · × Xαn × bα α6=α .
i

7.2 Path connectedness 21
Thus X(α1 , . . . , αn ) is connected.

Let [
Y :=X(α1 , α2 , . . . , αn ),

where the union runs over finite subsets α1 , α2 , . . . , αn ⊂ Λ.
Notice how b ∈ X(α1 , α2 , . . . , αn ), so any X(α1 , α2 , . . . , αn )∩X(β, β, . . . , β) 6=
∅. Thus by the previous part Y is connected.
Notice how up until this point we have not used the topology of X. We
claim Y = X. The result then follows from the fact that if A if connected, then
A is connected.
For any x = (xα )α∈Λ ∈ X and any neighbourhood U of x, we need to show
U ∩ Y 6= ∅.
Q
We can assume U = Uα is a basis element. I.e., Uα = Xα for α 6∈
α∈Λ
α1 , α2 , . . . , αn .
Take y = (yα ) by
(
xα for α ∈ α1 , α2 , . . . , αn
yα =
bα for α ∈
6 α1 , α2 , . . . , αn .
Then y = (yα ) ∈ X(α1 , . . . , αn ) ⊂ Y and also y ∈ U . Thus Y ∩ U 6= ∅, and we

are done.
Exercise 7.1. Show that if A is connected, then A is connected.

Remark 7.1.2. The result does not necessarily hold if X is endowed with the
box topology instead of the product topology.
Exercise 7.2. The only connected subsets of R are ∅, single-point sets, and
(finite or infinite) intervals (open, closed, or half-open).
This, together with Theorem 6.3.8 saying that a continuous function sends
connected sets to connected sets, gives us a slick proof of a famous Calculus
result:
Theorem 7.1.3 (Intermediate value theorem). Let X be connected and f : X →

R be continuous. If f (a) < f (b) for some a, b ∈ X, then for any r ∈ R with
f (a) < r < f (b) there exists c ∈ X such that f (c) = r.
7.2 Path connectedness

Definition 7.2.1 (Path, path connected). Let X be a topological space. Let
x, y ∈ X.
(i) A path in X from x to y is a continuous function r : [0, 1] → X such that

r(0) = x and r(1) = y.
(ii) X is path-connected if for any x, y ∈ X, there exists a path from x to y.
Theorem 7.2.2. A path-connected space is connected.

22 DIFFERENT KINDS OF CONNECTEDNESS
Proof. Fix x0 ∈ X. For each x ∈ X, let rx : [0, 1] → X be a path from x0 to

x. Since rx is continuous and [0, 1] is connected, the path traced by rx ([0, 1]) is
connected. Likewise, a path ry from x0 to y ∈ X gives a connected set ry ([0, 1]).
Then [
X= rx ([0, 1])
x∈X
is connected.
Remark 7.2.3. Connected does not imply path connected.

Counterexample 7.2.4. Let E = (0, y)−1 ≤ y ≤ 1 and F = (x, sin y1 )0 <

x ≤ 1 . Take X = E ∪ F = F .
Then X is connected (because it is the closure of a connected set) but it is
not path-connected. N
Exercise 7.3. Verify that this set (known as the topologist’s sine curve) is
not path-connected.
Remark 7.2.5. This gives an example showing that the closure of a path con-
nected set is not necessarily path connected.
Definition 7.2.6 (Locally path-connected). A space X is locally path-connected

if, for each open subset V of X and each x ∈ V , there is a neighbourhood U of
x such that x can be joined to any point of U by a path in V .
Exercise 7.4. Prove that the path components of a locally path-connected space
coincide with the connected components.
Exercise 7.5. Show that an open subset of Rn is connected if and only if it is
path-connected.
7.3 Compactness
Definition 7.3.1 (Compact). A topological space X is compact if every open
cover of X has a finite subcover.
In other words, for any family of open sets Uα α∈Λ with
[
X= Uα
α∈Λ
there exits α1 , α2 , . . . , αn ∈ Λ such that

n
[
X= Uαi .
i=1
Definition 7.3.2 (Compact subset). A subset E ⊂ X is compact if E is

compact in the relative topology.
Proposition 7.3.3. Any finite union of compact sets is compact.
Theorem 7.3.4. A closed subspace of a compact topological space is compact.

7.3 Compactness 23

Proof. Let X be compact and E ⊂ X closed. For any open cover Uα α∈Λ
of
E, [
(X \ E) ∪ Uα
α∈Λ
is an open cover of X (since X \ E is open because E is closed). Since X is

compact, there exists α1 , . . . , αn ∈ Λ such that
n
[
(X \ E) ∪ Uα
i=1
n
is an open cover of X. Hence Uαi i=1
is an open cover of E.
Lemma 7.3.5. Let X be a Hausdorff space. Let E ⊂ X be a compact subset.
Then for each x ∈ X \ E, there exists disjoint open sets U and V such that
x ∈ U and E ⊂ V .
Proof. Let x ∈ X \ E be fixed. For each y ∈ E, there exists disjoint open sets
x ∈ Uy and y ∈ Vy since X is Hausdorff.
Uy and Vy with
Then Vy y∈E is an open cover of E. Since E is compact there exists a
finite open subcover, say Vy1 , Vy2 , . . . , Vyn .
Then taking
V := Vy1 ∪ Vy2 ∪ · · · ∪ Vyn
we have E ⊂ V and taking
U := Uy1 ∩ Uy2 ∩ · · · ∩ Uyn .
This is open since it is a finite intersection of open sets, and by construction

V ∩ U = ∅.
Theorem 7.3.6. A compact subset in a Hausdorff space is closed.
Proof. Let X be Hausdorff and E ⊂ X compact. By the previous lemma, for
each x ∈ X \ E, there exists an open U with x ∈ U and U ∈ X \ E. Hence
X \ E is open, and E is closed.
Exercise 7.7. Let X be a Hausdorff space. Suppose A and B are disjoint compact
subsets of X. Show that there exist disjoint open sets U and V such that A ⊂ U
and B ⊂ V .
Theorem 7.3.7. Let X be a compact space and Y be any topological space.
Suppose f : X → Y is continuous. Then f (X) is a compact subset of Y .

Proof. Let Uα α∈Λ be an open cover of f (X). Then
f −1(Uα )

α∈Λ
is an open cover of X. Since X is compact, we can find a finite subcover, say
X = f −1 (Uα1 ) ∪ f −1 (Uα2 ) ∪ · · · ∪ f −1 (Uαn ).
Mapping back,
f (X) ⊂ Uα1 ∪ Uα2 ∪ · · · ∪ Uαn .
Hence f (X) is compact.
24 TYCHONOFF’S THEOREM
Like with the Intermediate value theorem above, this gives a slick proof of
the Extreme value theorem in analysis.
Recall another theorem from analysis:
Theorem 7.3.8 (Heine–Borel theorem). Let E ⊂ Rn . Then E is compact if
and only if E is closed and bounded.
Remark 7.3.9. Let X be a metric space. Then E ⊂ X implies E is closed and
bounded (closed because metric spaces are Hausdorff).
The converse is not true in general, though it is true for any finite-dimensional
vector space.
Lecture 8 Tychonoff’s theorem

8.1 Open maps
Theorem 8.1.1. Let X be a compact space and Y be a Hausdorff space. Sup-
pose f : X → Y is continuous. If f is one-to-one, then f is a homeomorphism
of X and f (X).
Proof. It suffices to show f is an open mapping, i.e., f sends open sets to open
sets, or f (U ) is open in f (X) for all open U ⊂ X.
Note that X \ U is closed, being the complement of an open set. Since X
is compact, X \ U is also compact. Thus f (X \ U ) = f (X) \ f (U ) because f is
one-to-one. This image is compact in f (X) ⊂ Y by Theorem 7.3.7. Since Y is
Hausdorff, f (X) must be closed by Theorem 7.3.6. Hence f (X) \ f (U ) is closed
in f (X), so f (U ) is open in f (X).
8.2 The Tychonoff theorem, finite case

We want to prove that an arbitrary product of compact spaces is compact.
There are two natural cases: the finite product case and the infinite product
case. We start with the finite product case.
Lemma 8.2.1. Let X be a topological space and let B be a basis for the topology
of X. If every open cover of X by sets in B has a finite subcover, then X is
compact.
Note that compact means any cover has a finite subcover. This lemma says
that we really only need to consider the basis elements.

Proof. Let Uα α∈Λ be an open cover of X. For each x ∈ X, x ∈ Uαx for
some αx ∈ Λ. Since Uαx we can choose Vx ∈ B such that x ∈ Vx ⊂ Uαx . Then
Vx x∈X ⊂ B is an open cover of X.

By assumption there exist x1 , x2 , . . . , xn ∈ X such that

X = Vx1 ∪ Vx2 ∪ · · · ∪ Vxn ,
but each Vxi ⊂ Uαi , so
X = Uαx1 ∪ Uαx2 ∪ · · · ∪ Uαxn
is a finite subcover from the original family.
Date: September 22nd, 2020.
8.3 Axiom of choice 25
Theorem 8.2.2 (Tychonoff’s theorem, finite case). If X1 , X2 , . . . , Xn are com-

pact spaces, then X1 × X2 × · · · × Xn is compact.
Proof. It suffices to consider the case n = 2. The theorem follows by induction.

Let X = X1 × X2 . Let
C = Uα × Vα α∈Λ

be an open cover of X with Uα open in X1 and Vα open in X2 .

By the above Lemma it suffices show that C has a finite subcover.
to
For any fixed y ∈ X2 , X1 × y is compact (since it is homeomorphic to
X1 ). Hence there exists a finite cover of X1 × y in C , say

Uy,1 × Vy,1 , Uy,2 × Vy,2 , ... Uy,n × Vy,n .
Then
Vy := Vy,1 ∩ Vy,2 ∩ · · · ∩ Vy,n
is an open neighbourhood of y in X2 (it’s a finite intersection of open sets).
Note that for π2 : X1 × X2 → X2 defined by π2 (x, y) = y, π2−1 (Vy ) is covered
by
Uy,1 × Vy,1 , Uy,2 × Vy,2 , . . . Uy,n × Vy,n .

Since Vy y∈X2 is an open cover of X2 and X2 is compact,
X2 = Vy1 ∪ Vy2 ∪ · · · ∪ Vyn
for some finite collection y1 , y2 , . . . , yn ∈ X2 . Now
X = π2−1 (Vy1 ) ∪ π2−1 (Vy2 ) ∪ · · · ∪ π2−1 (Vyn )
and each π2−1 (Vyi ) is covered by finitely many sets in C .

Hence X is covered by finitely many sets in C , finishing the proof.
8.3 Axiom of choice

In order to prove the infinite case of Tychonoff’s theorem we have to talk about:
Axiom 8.3.1 (Axiom of choice). Given a family of nonempty sets, it is possible

to select precisely oneelement
from each member of the family.
In other words, if Sα α∈Λ is a family of nonempty sets indexed by Λ, then
there exists a function f on Λ such that f (α) ∈ Sα for all α ∈ Λ.
This seems very natural, but it can’t be proved from basic set theory axioms.
It also has some seemingly unnatural consequences, like the Banach–Tarsky
theorem.
8.4 Zorn’s lemma

Definition 8.4.1 (Partially ordered set). A partially ordered set is a nonempty
set S with a relation “≤” satisfying
(i) x ≤ x for all x ∈ S (reflexivity);

(ii) if x ≤ y and y ≤ z, then x ≤ z (transitivity); and

(iii) if x ≤ y and y ≤ x, then x = y (antisymmetry).
Note that this does not say that any two elements can be compared, only
that if they can be compared, the above holds.
Definition 8.4.2 (Totally ordered set). Let S be a partially ordered set. Let
E ⊂ S be a subset.
(i) E is totally ordered if for any x, y ∈ E we have x ≤ y or y ≤ x.
(ii) An upper bound for E is an element x ∈ S such that y ≤ x for all y ∈ E.
(iii) An element x ∈ E is a maximal element of E if x ≤ y for some y ∈ E,
then y = x.
Remark 8.4.3. The distinction between upper bound and maximal element is
that the upper bound only needs to come from S; there could be two upper
bounds of E in S that are not mutually comparable.
Lemma 8.4.4 (Zorn’s lemma). Let S 6= ∅ be a partially ordered set. Suppose
each totally ordered subset of S has an upper bound. Then S has a maximal
element.
Remark 8.4.5. The Axiom of choice and Zorn’s lemma are equivalent.
Unfortunately, the infinite product case of Tychonoff’s theorem is also equiv-
alent to the Axiom of choice. So what we will do is assume Zorn’s lemma and
show that Tychonoff’s theorem follows.
But first:
Theorem 8.4.6 (Alexander subbasis theorem). Let X be a topological space
and let B be a subbasis for the topology of X.
If every open cover of X by sets in B has a finite subcover, then X is
compact.
Proof. Suppose X is not compact. In other words, there exists an open cover
C of X such that no finite subfamily of C covers X.
We will show that there exists a cover of X by sets in B that has no finite
subcover of X, which is a contradiction.
Consider families D of open subsets in X such that C ⊂ D and no finite
collection of sets in D covers X.
Set P = D ordered by inclusion. Then P 6= ∅ (since C ∈ P), and P

is partially
ordered.
Let Dα α∈Λ be a totally ordered subset of P. Set

[
D= Dα .
α∈Λ
Clearly D is an upper bound for Dα ; we claim that D ∈ P.

I.e., no finite collection of sets in D covers X.

To see this, suppose there exists a finite collection of sets in D that covers
X, say
Uα1 , Uα2 , . . . , Uαn ,
TYCHONOFF’S THEOREM 27
from Dα1 ⊂ Dα2 ⊂ · · · ⊂ Dαn . Then these sets must belong to Dα for some
α ∈ Λ, which contradicts Dα having no such finite subcover.
By Zorn’s lemma, P therefore has a maximal element, say E .
Finally we claim that E ∩ B covers X. If true, since E ⊂ P says this is a
cover of X by sets in B that has no finite subcover. This is a contradiction.
Let x ∈ X. It suffices to show that there exists V ∈ E ∩ B such that x ∈ V .
Choose U ∈ E such that x ∈ U . Since B is a subbasis there exists
V1 , V2 , . . . , Vm ∈ B such that
x ∈ V1 ∩ V2 ∩ · · · ∩ Vm ⊂ U.
If Vj ∈ E for some j, then we are done since x ∈ Vj ∈ E ∩ B.

If no Vj ∈ E for all j, we will derive a contradiction. We have E ∪ Vj 6∈ P

for j = 1, 2, . . . , m since E is maximal in P. This means for each j, there exists

finitely many open sets Wj,1 , . . . , Wj,nj ∈ E such that
nj
[
X = Vj ∪ Wj,i .
i=1
Take intersections over j, so

m nj
[ [
\
X= Vj ∪ Wj,i ⊂ (V1 ∩ V2 ∩ · · · ∩ Vm ) ∪ Wi,j .
j=1 i=1 i,j
But V1 ∩ V2 ∩ · · · ∩ Vm ⊂ U by choice of U , so
[
X=U∪ Wi,j ,
i,j
but all of these are in E , so E has a finite subcover of X. This contradicts

E ∈ P.
Lecture 9 Tychonoff’s theorem

9.1 The infinite case
We will now use the Alexander subbasis theorem in order to prove:

Theorem 9.1.1 (Tychonoff’s theorem, infinite case). Let Xα α∈Λ be a family
of compact spaces. Then
Y
X= Xα
α∈Λ
is compact (in the product topology).
Proof. By the Alexander subbasis theorem, X is compact if it satisfies the fol-

lowing property: Let B be a subbasis for the topology on X. If for any family
D of subsets in B, D has no finite subcover of X, then D does not cover X.
Now take B to be the subbasis for the product topology of X of the form
−1
α (U α ), α ∈ Λ, where Uα is open in Xα . These pullbacks look like Uα ×
π
Xβ β6=α .
Let D be a family of subsets such that D ⊂ B and D has no finite subcover
of X.
We want to show that D does not cover X.
For any β ∈ Λ, consider the open subsets V ⊂ Xβ such that πβ−1 (V ) ∈ D.
The family of such sets V cannot cover Xβ . Otherwise, since Xβ is compact,
we would have a finite subcover of Xβ , say V1 , V2 , . . . , Vn , for which πβ−1 (Vi ) ∈ D.
But those pullbacks would cover X, and D has no finite subcover of X.
Hence we can choose a point
xβ ∈ Xβ \ V πβ−1 (V ) ∈ D and V open in Xβ .

(If no such V exist, we are just subtracting the empty set.)

Then
x = (xβ )β∈Λ ∈ X
is not included in any sets in D by construction.

Hence D does not cover X since it misses at least x.
Remark 9.1.2. Note that the existence of this x is by Axiom of choice.

In fact, the infinite case of Tychonoff’s theorem is equivalent to the Axiom
of choice.
Exercise 9.1. Show that Tychonoff’s theorem implies the Axiom of choice.
9.2 Other notions of compactness

Recall that compact means every open cover has a finite subcover. There are
other types of compactness:
Definition 9.2.1 (Limit point compactness). A space X is limit point com-

pact if every infinite subset of X has a limit point.
Definition 9.2.2 (Sequential compactness). A space X is sequentially com-

pact if every sequence in X has a convergent subsequence.
Note that this last notion, in Rn , is usually discussed in terms of the Bolzano–
Weierstrass theorem, saying that any bounded sequence has a convergent
subsequence (since a bounded sequence is contained in a compact set).
Note that in Rn , all of these notions are equivalent. We will show that in
fact in any metrisable space, all three notions are equivalent.
Recall that x ∈ X is a limit point of a subset A ⊂ X if for every open

neighbourhood U of x, we have U ∩(A\ x ) 6= ∅. In other words, x ∈ A \ x .
Theorem 9.2.3. If X is compact, then X is limit point compact.
Note that we did not assume X is metrisable here—compactness is stronger

than limit point compactness.
9.2 Other notions of compactness 29
Proof. We need to show that if A ⊂ X is infinite, then A has a limit point. This
is equivalent to showing that if A has no limit point, then A is a finite set.
Assume A has no limit point.
We claim A is closed, i.e. X \ A is open.

For any x ∈ X \ A, x is not a limit point of A. Hence x 6∈ A \ x = A.
This means x ∈ X \ A which is open, so there exists an open neighbourhood U
of x such that U ⊂ X \ A. Therefore x ∈ U ⊂ (X \ A) ⊂ X \ A. Hence X \ A
is open, so A is closed.
Our goal is to show that A is finite. To this end, notice how for each x ∈ A,
since x is not a limit point
of A, there exists an openneighbourhood
Ux of x
such that Ux ∩ (A \ x ) = ∅. This means Ux ∩ A = x , i.e., every point in
A is an isolated point.
Now consider
C = X \ A ∪ Ux x ∈ A .

This is an open cover of X, and X is compact so there exists a finite subcover,

say
X \ A, Ux1 , Ux2 , . . . , Uxn .
Thus
X = (X \ A) ∪ (Ux1 ∪ Ux2 ∪ · · · ∪ Uxn ).
In particular this means the union in the parenthesis covers A, so
A = A ∩ (Ux1 ∪ Ux2 ∪ · · · ∪ Uxn ),
and each of those intersections is just xi , so

A = x1 , x2 , . . . , xn
is a finite set.
Remark 9.2.4. In general, limit point compactness does not imply compactness.

Counterexample 9.2.5. Let Y = a, b be a two point set with the trivial
topology, i.e., T = ∅, Y . Let X = Z+ × Y where

Z+ = n > 0 n ∈ Z
with the discrete topology (so every single point is an open set).
Then X is limit point compact. In fact, for any nonempty A ⊂ X, A has a
limit point. If (n, a) ∈ A, then (n, b) isa limit point
of A.
But X is not compact. Take C = Un = n × T n ∈ Z+ . This is an

open cover of X, but it has no finite subcover since removing any set makes it
miss a point of Z+ . N
Theorem 9.2.6. Let X be a metrisable space. Then the following are equiva-
lent:
(i) X is compact.
(ii) X is limit point compact.
(iii) X is sequentially compact.

Proof. That (i) implies (ii) is true in general: this is Theorem 9.2.3.

To show (ii) implies
(iii), let xn be a sequence in X. Consider the set
A = xn n ∈ Z+ .
First, if A is a finite set, then xn must repeat at least one element infinitely
many times. I.e.,there exists some x ∈ X such
that x = xn for infinitely many
n. This means xn has a subsequence xnk = x n which is a constant
k
sequence, hence convergent (to x).
Second, if A is an infinite set, then since X is limit point
compact A must
have a limit point, say x. We select a subsequence of xn that converges to
x as follows.
For any δ > 0, B(x, δ) ∩ A has infinitely many points since x is a limit
point, because otherwise there exists some sufficiently small δ 0 > 0 such that
B(x, δ 0 ) ∩ A = ∅, contradicting x being a limit point of A.
Thus we can choose n1 < n2 < n3 < . . . such that xni ∈ B(x, 1i ) ∩ A. Then
∞
xni i=1 converges to x.
(Note that this is really the same argument one uses to prove the Bolzano–
Weierstrass theorem in Rn .)
Finally, (iii) implies (i). Assume X is sequentially compact with a metric d.
First, an important lemma:
Lemma 9.2.7 (Lebesgue number lemma). Let X be a metrisable space and

suppose X is (sequentially) compact.
For any open cover C of X there exists δ > 0 (depending on C ) such that
for each subset V of X with diameter diam(V ) < δ, there exists U ∈ C such
that V ⊂ U .
Note that this is uniform in V ; it only depends on C . Here
Definition 9.2.8 (Diameter). The diameter of a subset V in a metric space

is defined by

diam(V ) = sup d(x1 , x2 ) x1 , x2 ∈ V .
We prove this by contradiction: Assume there is no δ > 0 such that V ⊂ X

with diam(V ) < δ that has an element of C containing it.
Then for each n ∈ N there exists Vn ⊂ X such that diam(Vn ) < n1 and Vn is
not contained in any element of C .
Choose xn ∈ Vn for each n; this sequence must have a convergent subse-
quence, say xni , since X is sequentially compact.
Assume xni converges to x ∈ X. Since C covers X, x ∈ V for some V ∈ C .
Since V is open, there exists some ε > 0 such that B(x, ε) ⊂ V . Choose ni large
enough such that n1i < 2ε and d(xni , x) < 2ε .
Then for any y ∈ Vni ,
1 ε
d(y, x) ≤ d(y, xni ) + d(xni , x) < + < ε.
ni 2
Hence y ∈ B(x, ε) ⊂ V . I.e., Vni ⊂ V ∈ C contradicts no element of C

containing any Vn .
COMPACTNESSES 31
Lecture 10 Compactnesses
10.1 Proof continued
Proof of Theorem 9.2.6, continued. We are left with (iii) implying (i), for which
we showed the Lebesgue number lemma.
Next, if X is sequentially compact, then given any ε > 0, there exists a finite
cover of X by ε-balls (in other words, X is totally bounded).
We show this by contradiction. Assume there exists some ε > 0 such that
X cannot be covered by finitely many ε-balls.
Choose x1 ∈ X, then x2 ∈ X \ B(x1 , ε), x3 ∈ X \ (B(x1 , ε) ∪ B(x2 , ε)), and
so on. The sets we are removing are nonempty by the assumption of ε-balls not
covering X.
Hence we get a sequence xn in X that does not have a convergent sub-
sequence, since d(xn+1 , xi ) ≥ ε for i = 1, 2, . . . , n. This contradicts X being
sequentially compact.
Third: we are not equipped to show that if X is sequentially compact, then
X is compact.
Let C be an open cover of X. By Lebesgue number lemma, there exists
some δ > 0 such that for any V ⊂ X with diam(V ) < δ, we have V ⊂ U for
some U ∈ C .
Take ε = 3δ . By the second step, we know X can be covered by finitely many
ε-balls. In other words,
X = B(x1 , ε) ∪ B(x2 , ε) ∪ · · · ∪ B(xn , ε).
Now diam(B(xi , ε)) < δ, so B(xi , ε) ⊂ Ui for some Ui ∈ C . Hence from the
covering of balls we get a finite cover
X = U1 ∪ U2 ∪ · · · ∪ Un
of X.
Remark 10.1.1. There exist sequentially compact spaces that are not compact.
There also exist compact spaces that are not sequentially compact.
For posterity we write down the definition from inside the proof:
Definition 10.1.2 (Totally bounded). A metric space X is totally bounded
if for any ε > 0 there is a finite cover of X by ε-balls.
In other words, we showed in the second step of the Theorem 9.2.6 that
Corollary 10.1.3. If X is a compact metric space, then X is totally bounded.
The converse is not true: we will show later that if X is a metric space, X
is compact if and only if X is complete and totally bounded.
Exercise 10.1. Let X be a metric space.
(a) Show that if E is a compact subset of X, then E is closed and bounded.
(b) Give an example to show the converse is false. In other words, find a metric
space in which not every closed and bounded subset if compact.
32 COMPACTNESSES
10.2 Uniform continuity

Definition 10.2.1 (Uniform continuity). Let (X, dX ) and (Y, dY ) be two met-
ric spaces. A function f : X → Y is uniformly continuous if for any ε > 0
there exists δ > 0 such that for any x1 , x2 ∈ X, dX (x1 , x2 ) < δ implies
dY (f (x1 ), f (x2 )) < ε.
Theorem 10.2.2. Let (X, dX ) and (Y, dY ) be metric spaces. Let f : X → Y be

continuous. Assume X is compact. Then f is uniformly continuous.
Proof. Given ε > 0, for each y ∈ Y let Vy = B(y, 2ε ) ⊂ Y . Since f is continuous,
C = f −1 (Vy ) y ∈ Y

is an open cover of X. Since X is compact, by Lebesgue number lemma there

exists δ > 0 such that for any U ⊂ X with diam(U ) ≤ 2δ, U is contained in
some element of C .
Hence for any x1 , x2 ∈ X with dX (x1 , x2 ) < δ, we have x2 ∈ B(x1 , δ). Since
the diameter of this is bounded by 2δ, this is contained in some f −1 (Vy ).
Hence f (x1 ), f (x2 ) ∈ Vy = B(y, 2ε ), so dY (f (x1 ), f (x2 )) < ε.
Exercise 10.2. Let X and Y be two topological spaces. Assume

that Y is
compact. Let N be an open set of X × Y containing x0 × Y for some
x0 ∈ X. Show that there exists an open neighbourhood W of x0 in X such that
(W × Y ) ⊂ N .
Exercise 10.3. Let X and Y be two topological spaces.
(a) Assume that Y is compact. Show that the projection map π1 : X ×Y → X

is a closed map. That is, for any closed subset E of X ×Y , π1 (E) is closed.
(b) Is (a) still true if we do not assume Y is compact?
Exercise 10.4. Let X and Y be two topological spaces. Let f : X → Y be a

map. Let

Gf = (x, f (x)) ∈ X × Y x ∈ X
be the graph of f .
(a) Assume that Y is compact Hausdorff. Show that f is continuous if and

only if Gf is closed in X × Y .
(b) Is (a) true if Y is Hausdorff but not compact?
Exercise 10.5. Let (X, d) be a compact metric space. Let f : X → X be a

(continuous) map. Let 0 < α < 1 be a real number. Suppose that
d(f (x), f (y)) ≤ αd(x, y)
for all x, y ∈ X. Show that there exists a unique point x ∈ X such that f (x) = x.
This is known as the Banach fixed-point theorem.
10.3 Local compactness 33
10.3 Local compactness

Definition 10.3.1 (Local compactness). (i) A space X is locally compact
at x ∈ X if there is an open neighbourhood of x such that U is compact.
(ii) A space X is locally compact if X is locally compact at every point
x ∈ X.
Example 10.3.2. The space Rn is locally compact. (But it is not compact; it
is unbounded.) N
Exercise 10.6. Let X be a Hausdorff space. Prove that X is locally compact at
x ∈ X if and only if for every open neighbourhood U of x, there is a neighbour-
hood V such that V ⊂ U and V is compact.
10.4 One-point compactification

n
Let X be a locally compact Hausdorff space which is not compact
(e.g. R ).
Take a point, denoted by ∞, with ∞ 6∈ X. Set Y = X ∪ ∞ .
We define a topology (i.e. a collection of open sets) T on Y such that Y is
compact. In particular, we have the open sets
(i) U where U ⊂ X is open in X,
(ii) Y \ C where C ⊂ X is compact.
Example 10.4.1. The model for this is taking the number line R, identifying
both ‘ends’ of the number line as ∞, and gluing them together to form a circle.
N
Exercise 10.7. Show that T as defined above is a topology on Y .

Theorem 10.4.2. The space Y = X ∪ ∞ is a compact Hausdorff space, the
relative topology for X inherited from Y coincides with the original topology on
X, and X = Y in Y .
Proof. First let us show that the two topologies coincide. Let (X, C ) be the
original topology on X. Then we want to show that (X, C ) = (X, T ), and
X =Y.
For any open set U in Y , we have two options: either U ⊂ X, so U ∈ C , in
which case U ∩ X = U , where the left-hand side is open in T .
If U = Y \ C where C ⊂ X is compact, then U ∩ X = X \ C. But X is
Hausdorff and C is compact, so C is closed, meaning that X \ C ∈ C . is open.
Hence (X, T ) ⊂ (X, C ).
On the other hand, for U ∈ C , U ∈ T . Hence (X, C ) ⊂ (X, T ).
Next, we show X = Y . Since X is not compact, for any open set Y \ C
containing ∞, (Y \ C) ∩ X = X \ C 6= ∅. Hence ∞ ∈ X.
Now let us show Y is compact. Let A be an open cover of Y . Then Y \C ∈ A
for some C ⊂ X compact (since those are the neighbourhoods of ∞). Then
A \ Y \C

is an open cover of C. But C is compact, so there is a finite subcover A 0 ⊂ A

of C. Then
A0∪ Y \C

34 COUNTABILITY AND SEPARATION AXIOMS
is a finite subcover of Y .
That Y is Hausdorff is easy. Let x, y ∈ Y . If x, y ∈ X, then since X is
Hausdorff, there exist open U, V ⊂ X such that x ∈ U , y ∈ V , and U ∩ V = ∅.
If x ∈ X and y = ∞, since X is locally compact, there exists an open U
such that x ∈ U and U is compact. Then y = ∞ ∈ Y \ U , which is open in Y ,
and (Y \ U ) ∩ U = ∅.

Example 10.4.3. The
one-point compactification
R ∪ ∞ is homeomorphic
to S 1 = (x, y) ∈ R2 x2 + y 2 = 1 . N
Similarly,

Example 10.4.4. The one-point compactification Rn ∪ ∞ ' S n , where
S n = (x1 , x2 , . . . , xn ) ∈ Rn+1 x21 + x22 + · · · + x2n = 1 .

N

Example 10.4.5. The one-point compactification C ∪ ∞ is the Riemann
sphere. N
Example 10.4.6. Since (0, 1) and R are homeomorphic, so must their one-point
compactification be, i.e.
(0, 1) ∪ ∞ ' R ∪ ∞ = S 1 .

On the other hand, we can compactify (0, 1) by adding both endpoints, i.e.,
[0, 1]. But these two spaces are not homeomorphic. N
Remark 10.4.7. There can be many compactifications of a space; the one-point

compactification is just one method.
We will introduce another compactification: the Stone–Čech compactifica-
tion, which is useful for analysis.
Lecture 11 Countability and separation axioms

11.1 Countability axioms
Definition 11.1.1 (First-countable). A topological space is first-countable
∞ if
for each x ∈ X there exists a sequence of open neighbourhoods Un n=1 of x
such that each neighbourhood of x includes one of the Un .
Exercise 11.1. (a) Any metric space is first-countable.
(b) Let X be first-countable and let A ⊂ X. Then x ∈ A (so x is a limit point

of A) if and only if there is a sequence of points in A converging to x.
Definition 11.1.2 (Second-countable). A topological space X is second-countable

if X has a countable basis for its topology.
Exercise 11.2. A second-countable space is first-countable.

Date: October 1st, 2020.
11.2 Separation axiom 35
Exercise 11.3. Let X be a compact Hausdorff space. Show that X is metrisable

if and only if X is second-countable.
Remark 11.1.3. There exist metric spaces that are not second-countable. (Though
they are not Euclidean spaces; there we can make a countable basis for the
topology using rational intervals.)
Theorem 11.1.4 (Lindelöf’s theorem). Let X be a second-countable topological
space. Then every open cover of X has a countable subcover.
Proof. Let A be an open cover of X. Let B be a countable basis for the
topology on X.
Let C ⊂ B consisting of sets U ∈ B such that U ⊂ V for some V ∈ A .
We claim that C is a cover of X.
This is easy: for each x ∈ X, x ∈ V for some V ∈ A since A covers X.
Since V is open and B is a basis, there must exist some U ∈ B such that
x ∈ U ⊂ V . Hence x ∈ U C .
Now C ⊂∞B, and the latter is countable, so the former is countable. Write
C = Un n=1 , and for each Un choose Vn ∈ A such that Un ⊂ Vn . Then
∞
Vn n=1 ⊂ A is a countable subcover of X.
Definition 11.1.5 (Dense, separable). (i) A subset E ⊂ X is dense if E =

X.
(ii) A topological space X is separable if there is a countable subset of X
that is dense in X.
Example 11.1.6. The real numbers R is separable, because Q ⊂ R is dense;
Q = R. N
Theorem 11.1.7. If X is second-countable, then X is separable.
∞
Proof. Let Un n=1 be a basis for X. For each Un 6= ∅, choose xn ∈ Un . Then
∞
E = xn n=1 is a countable set which is dense in X.
To see this, pick an arbitrary x ∈ X. For any open neighbourhood U of
x, we need to show that U ∩ E 6= ∅, so that x is a limit point of E. But U
is open, so there exists some Un in the basis with Un ⊂ U , and xn ∈ Un , so
xn ∈ U ∩ E 6= ∅.
Example 11.1.8. As one might show in measure theory or functional analysis
classes, Lp (X) is separable for 1 ≤ p < ∞. N
However L∞ (X) is not separable; this is the main difference between Lp and
∞
L .
Exercise 11.4. Show the above; that Lp (X) is separable for 1 ≤ p < ∞, but not
for p = ∞.
11.2 Separation axiom

These are denoted with T because ‘separation axiom’ is ‘Trennungsaxiom’ in
German.
Definition 11.2.1 (T1 -space). A space X is a T1 -space if for disjoint x, y ∈ X,
there exists an open set U such that x 6∈ U and y ∈ U .
36 COUNTABILITY AND SEPARATION AXIOMS
Proposition 11.2.2. X is a T1 -space if and only if single-point sets are closed.

Proof. Forwards, fix x ∈ X. Foreach y ∈ X \ x , there exists an open Uy
such that x 6∈ Uy . I.e., Uy ⊂ X \ x , and hence
[
X\ x = Uy

y∈X\ x
is open.
Conversely, for distinct x, y ∈ X, X \ x is open since x is closed.
Hence there exists an open U such that y ∈ U and x 6∈ U .
Exercise 11.5. Let X be a topological space and let X0 be the topological space
made up of the set X with the cofinite topology. Show that the identity map
from X to X0 is continuous if and only if X is a T1 -space.
Definition 11.2.3 (T2 -space, Hausdorff). A space X is a T2 -space if for dis-

tinct points x, y ∈ X there exist disjoint open sets U and V such that x ∈ U
and y ∈ V .
In other words, T2 means Hausdorff.

Remark 11.2.4. Every T2 -space is a T1 space.
Definition 11.2.5 (T3 -space, regular space). A space X is T3 or regular if X

is T1 and for any closed subset E ⊂ X and x ∈ X \ E there exist disjoint open
sets U and V such that x ∈ U and E ⊂ V .
Exercise 11.6. Show that a subspace of a regular space is regular.

Remark 11.2.6. Since in a T1 -space, single-point sets are closed, T3 means we
can separate single points, so T3 implies T2 (or Hausdorff).
Exercise 11.7. Let B be the collection of subsets of R of the form (a, b), and
(a, b) ∩ Q, where −∞ < a < b < ∞. Prove the following:
(a) B is a basis of open sets for a topology T on R.
(b) (R, T ) is a Hausdorff space.
(c) R \ Q is T -closed.
(d) Let f : (R, T ) → R be a continuous map. If f = 0 on R \ Q, then f = 0

on R.
(e) (R, T ) is not regular.
This provides an example of a Hausdorff space (T2 ) that is not regular (T2 ).
Definition 11.2.7 (T4 -space, normal space). A space X is T4 or normal if X

is T1 and for any disjoint closed subsets E and F there exist disjoint open sets
U and V such that E ⊂ U and F ⊂ V .
Remark 11.2.8. A subspace of a normal space need not be normal.

Remark 11.2.9. Again since a T4 has to be T1 , single-point sets are closed, so
T4 implies T3 .
11.3 Urysohn’s lemma 37
In other words,
T4 =⇒ T3 =⇒ T2 =⇒ T1 .
Theorem 11.2.10. Every metric space is normal (i.e. T4 ).
Proof. Let (X, d) be a metric space. Obviously X is T2 (Hausdorff); just take a

ball of radius a half the distance between two points.
Let E and F be two disjoint closed sets. For each x ∈ E, there exists a
radius r(x) > 0 such that B(x, r(x)) ∩ F = ∅ (since F is closed, so X \ F is
open).
Similarly, for each y ∈ F there exists a radius r(y) > 0 such that B(y, r(y))∩
E = ∅.
Then any two B(x, r(x)) and B(y, r(y)) might intersect; but worst case sce-
nario we only need to shrink by a factor of two (inspired by the Hausdorff
argument above), so let
[ r(x)
U= B x,
2
x∈E
and r(y)
[
V = B y,
2
y∈F
are open sets. Clearly E ⊂ U and F ⊂ V .

We claim U ∩ V = ∅. If not, suppose U ∩ V 6= ∅, so there exists z ∈ U ∩ V .
Then r(x) r(y)
z ∈ B x, ∩ B y,
2 2
for some x ∈ E and y ∈ F . This implies
r(x) r(y)
d(x, y) ≤ d(x, z) + d(z, y) < + ≤ max r(x), r(y) .
2 2
But this means x ∈ B(y, r(y)) or y ∈ B(x, r(y)), which is a contradiction, since
r(x) and r(y) were radii specifically chosen to separate x and y.
11.3 Urysohn’s lemma

This asks the question: given two disjoint subsets, can they be separated by a
continuous function? The answer is yes, if the space is normal.
To prove this we first need the following:
Lemma 11.3.1. A topological space X is normal if and only if X is T1 and

given a closed subset E ⊂ X and an open set W ⊃ E, there exist an open set U
such that
E ⊂ U ⊂ U ⊂ W.
Proof. The forward direction is fairly straightforward. Assume X is normal (so

automatically T1 ) and suppose E ⊂ W where E is closed and W is open. Then
E and X \ W are disjoint closed sets.
Hence, since X is normal, we can separate E and X \ W by disjoint open
sets U and V such that E ⊂ U and (X \ W ) ⊂ V . Thus U ⊂ (X \ V ) ⊂ W
(since U ⊂ X \ V , but X \ V is closed).
38 URYSOHN’S LEMMA
For the converse, let E and F be disjoint sets. Then W = X \ F ⊃ E, and

W is open. So by assumption there exists an open set U such that E ⊂ U ⊂
U ⊂ W = X \ F.
Then U and X \ U are disjoint open sets and E ⊂ U and F ⊂ (X \ U ).
Before we state and prove Urysohn’s lemma, recall that a dyadic rational
number is a rational number of the form 2mn where m, n ∈ Z.
The dyadic rational numbers clearly are dense in R (e.g. in [0, 1], keep taking
middle points; they’re all dyadic).
Lecture 12 Urysohn’s lemma

12.1 Proof of Urysohn’s lemma
Theorem 12.1.1 (Urysohn’s lemma). Let E and F be two disjoint closed sub-
sets of a normal space X. Then there exists a continuous function f : X → [0, 1]
such that f = 0 on E and f = 1 on F .
Remark 12.1.2. The interval [0, 1] can be replaced by any closed interval [a, b]
so that f = a on E and f = b on F .
Proof. Let V = X \ F . Since F is closed, V is open, and E ⊂ V . By
Lemma 11.3.1, there exist an open set U 12 such that
E ⊂ U 21 ⊂ U 21 ⊂ V.
Since E is closed and U 12 is open, we can insert, using the lemma, another pair
U 14 ⊂ U 14 between those, and similarly between U 12 and V we insert U 34 ⊂ U 34 .
Hence
E ⊂ U 14 ⊂ U 14 ⊂ U 21 ⊂ U 21 ⊂ U 34 ⊂ U 34 ⊂ V.
Now repeat this process ad nauseam between each closed-open pair, at each
finite step adding sets for more and more dyadic rational numbers between 0
and 1. Hence we construct open sets Ur for every dyadic rational 0 < r < 1
such that
Ur ⊂ Us
for 0 < r < s < 1,
E ⊂ Ur
for all 0 < r < 1, and
Ur ⊂ V
for all 0 < r < 1. Now the idea is to use the boundaries of the Ur as level
curves f = r for a function f : X → [0,
1]. I.e., define f (x) = 0 if x ∈ Ur for all
0 < r < 1 and f (x) = sup r x 6∈ Ur .
Thus 0 ≤ f (x) ≤ 1, f = 0 on E, and f = 1 on F .
It remains to verify that f is continuous. Let x ∈ X, and let us show that f
is continuous at x.
First, suppose f (x) = 0. For any ε > 0, take a dyadic rational r such that
0 < r < ε. Then x ∈ Ur (since otherwise f (x) ≥ r > 0). Hence Ur is an open
Date: October 6th, 2020.
12.1 Proof of Urysohn’s lemma 39
neighbourhood of x. For any y ∈ Ur , |f (y) − f (x)| = |f (y)| ≤ r < ε. Hence

f (Ur ) ⊂ B(f (x), ε), so f is continuous at x.
Second, suppose f (x) = 1, meaning that x 6∈ Ur for any 0 < r < 1 (since
otherwise f (x) ≤ r < 1). Thus x 6∈ Ur (because for r1 < r2 we have Ur1 ⊂
Ur1 ⊂ Ur2 ⊂ Ur2 by construction, so we can always perturb r slightly). For any
ε > 0 there is a dyadic rational s such that 1 − ε < s < 1 (so 0 < 1 − s < ε).
Since x 6∈ Us , take W = X \ Us , which is open. Hence W is an open
neighbourhood of x.
For any y ∈ W , we have y 6∈ Us , hence s ≤ f (y) ≤ 1. Hence 0 ≤ 1 − f (y) ≤
1 − s < ε, so |f (x) − f (y)| < ε. Thus f (W ) ⊂ B(f (x), ε), so f is continuous at
x.
Third, suppose 0 < f (x) < 1. For any ε > 0, take dyadic rationals r and s
such that
f (x) − ε < r < f (x) < s < f (x) + ε.
Then x 6∈ Ut for all r < t < f (x). Since Ur ⊂ Ut for r < t, this means x 6∈ Ur .
On the other hand x ∈ Us , thus W = Us \ Ur is an open neighbourhood of
x. For y ∈ W , we have r ≤ f (y) ≤ s, implying that |f (y) − f (x)| < ε. This
means f (W ) ⊂ B(f (x), ε), so f is continuous at x.
A consequence of this theorem is:
Theorem 12.1.3 (Tietze extension theorem). Let X be a normal space and let
Y ⊂ X be a closed subset. Let f be a bounded continuous real-valued function
on Y . Then there exists a bounded continuous real-valued function h on X such
that h = f on Y .
Proof. We should assume f is not constant on Y , since otherwise we just take

h to be the same constant on X.
Step 1: Let c0 = sup |f (y)| > 0 (since f is not constant). Let
y∈Y
n c0 o
E0 = y ∈ Y f (y) ≤ −

3
and n c0 o
F0 = y ∈ Y f (y) ≥ .

3
Then E0 and F0 are closed (since they’re continuous pullbacks of closed sets) and
disjoint subsets of X. By Urysohn’s lemma there exists a continuous real-valued
function g0 on X such that g0 = − c30 on E0 and g0 = c30 on F0 , and
c0 c0
− ≤ g0 (x) ≤
3 3
c0
for all x ∈ X. In particular, |g0 (x)| ≤ 3 for all x ∈ X, and
2
|f (x) − g0 (x)| ≤ c0
3
for all x ∈ Y (since f is only defined on Y ). In other words we have constructed
a continuous function g0 which approximates f with a maximum error of 23 c0 .
40 URYSOHN’S LEMMA
∞
Step 2: Construct a sequence gn (x) n=0
inductively such that |gn (x)| ≤
2n
3n+1 c0 for all x ∈ X and
∞
X 2n+1
f (x) − gk (x) ≤ n+1 c0

3
k=0
for all x ∈ Y . We do this by considering f1 (x) = f (x) − g0 (x) on Y , let

2
c1 = sup |f1 (x)| ≤ c0 ,
y∈Y 3
n c1 o
E1 = y ∈ Y f )1(y) ≤ − ,

3
and n c1 o
F1 = y ∈ Y f1 ≥ .

3
Use step 1 to show there exists a g1 (x) with
1 2
|g1 (x)| ≤ c1 ≤ 2 c0
3 3
for all x ∈ X and
2 22
|f (x) − g0 (x) −g1 (x)| ≤ c1 ≤ 2 c0
| {z } 3 3
=f1 (x)
for all x ∈ Y . Then rinse and repeat with f2 = f (x) − g0 (x) − g1 (x), and so on.
Step 3: Let
Xn
hn (x) = gk (x).
k=0

Then hn (x) converges to a bounded continuous function h(x) on X and
h = f on Y .
We show this by showing hn (x) is uniformly Cauchy in X, which means
hn → h for some continuous function h on X. For n ≥ m,
|hn (x) − hm (x)| = |gm+1 (x) + · · · + gn (x)| ≤ |gm+1 (x)| + · · · + |gn (x)|
2 m+1 2 n c
0
≤ + ··· +
3 3 3
2 m+1 2 2 2 c
0
≤ 1+ + + ...
3 3 3 3
2 m+1 1 c 2 m+1
0
≤ 2 = c0 → 0
3 1− 3 3 3

uniformly (for all x ∈ X) as m → ∞. Hence hn (x) is uniformly Cauchy.
Thus hn (x) → h(x) is continuous.
Moreover
∞ ∞ n
X X 2 c0
|h(x)| ≤ |gn (x)| ≤ = c0 .
n=0 n=0
3 3
Hence h is bounded on X (by the same bound as f on Y ).

12.2 The Urysohn metrisation theorem 41
Finally, for x ∈ Y ,
2 n+1
|f (x) − h(x)| ≤ |f (x) − hn (x)| + |hn (x) − h(x)| ≤ c0 + |hn (x) − h(x)|.
3
The last term converges to 0 uniformly for all x ∈ X, and the first part also
goes to 0 uniformly as n goes to 0. Hence f (x) = h(x) for all x ∈ Y .
We should always keep in our mind that metric spaces are particular exam-
ples of normal spaces.
12.2 The Urysohn metrisation theorem

In general, spaces need not be metric (nor normal), but it is an interesting
question to ask whether we can assign a metric fo a space—to metrise it.
There are many results in this direction, we will talk about one of them,
namely that a regular space (T3 ) with a countable basis (second countable) is
metrisable.
Lecture 13 Urysohn’s metrisation theorem

13.1 Preliminaries
Lemma 13.1.1. Every regular space X with a countable basis is normal.
Proof. Let B be a countable basis for X. Let E and F be two disjoint closed
sets in X. For each x ∈ E, by regularity there exist open sets U and V such
that x ∈ U , F ⊂ V , and U ∩ V = ∅.
Since x ∈ U ⊂ (X \V ), where X \V is closed, U ⊂ (X \V ). Thus U ∩F = ∅.
Now since U is open, take Ux ∈ B such that x ∈ Ux ⊂ U . Repeat
∞ this for
every x ∈ E. Since B is countable, we have an open cover Ui i=1 of E with
Ui ∩ F = ∅, with Ui ∈ B. ∞
Similarly F has an open cover Vi i=1 with Vi ∩ E = ∅ and Vi ∈ B.
Let
n
[
Un0 := Un \ Vi
i=1
and
n
[
Vn0 := Vn \ Ui ,
i=1
both open. We claim

∞
[
E ⊂ U 0 := Un0
n=1
and
∞
[
0
F ⊂ V := Vn0 ,
n=1
and importantly U 0 ∩ V 0 = ∅.
42 URYSOHN’S METRISATION THEOREM
For x ∈ E, x ∈ Un for some n. Since Vi ∩ E = ∅ for all i we have x ∈ Un0 .

Hence x ∈ U 0 , so E ⊂ U 0 . Similarly F ⊂ V 0 .
Now suppose U 0 ∩ V 0 6= ∅. In other words, there exists x ∈ U 0 ∩ V 0 meaning
that x ∈ Un0 and x ∈ Vm0 for some n and m.
If n ≥ m, then
[n
x ∈ Un0 = Un \ Vi ,
i=1
so x 6∈ Vm , but then x 6∈ Vm0 , a contradiction.

Similarly if n ≤ m, then x ∈ Vm0 , meaning x 6∈ Un , so x 6∈ Un0 .
Lemma 13.1.2. Let X be a regular space with a countable basis. Then there
exist a countable collection of continuous functions fn : X → [0, 1] such that for
any x0 ∈ X and any neighbourhood U of x0 there exist an index n such that
fn (x0 ) > 0 and fn = 0 on X \ U .
∞
Proof. Let Bn n=1 be a countable basis for X. For each pair Bn , Bm with
Bn ⊂ Bm we have that Bn and X \ Bm are disjoint closed sets.
By Lemma 13.1.1, X is normal, and so we can apply Theorem 12.1.1 to
construct a continuous function gn,m : X → [0, 1] with gn,m (Bn ) = 1 and
gn,m (X \ Bm) = 0.
Reindex gn,m to fn . Now for x0 ∈ X and x0 ∈ U ⊂ X open, we can
choose Bn and Bm such that x0 ∈ Bn ⊂ Bm ⊂ U (by regularity).
Then gn,m (x0 ) = 1 and gn,m (X \ Bm ) = 0. Hence g(X \ U ) = 0 since
Bm ⊂ U .
Now the idea for Urysoh’s metrisation theorem is to show that X, using this
∞
Q
previous lemma, embeds homeomorphically to the infinite product RN = R.
i=1
Let
¯ b) := min |a − b|, 1

d(a,
be the standard bounded metric on R.
Exercise 13.1. Verify that d¯ is a metric on R.
Let Y = RN . For x = (xi ) and y = (yi ) in Y , define
¯ i , yi ) o
n d(x
D(x, y) = sup .
1≤i<∞ i
Exercise 13.2. Verify that D is a metric on Y .
Proposition 13.1.3. D is a metric on Y that induces the product topology on

Y = RN . In other words, RN (with the product topology) is metrisable.
Proof. Let U be an open set in the metric topology and let x = (xi ) ∈ U .
We want to show that there exists an open set V in the product topology
such that x ∈ V ⊂ U . (Thus U is open in the product topology.)
Since U is open in the metric topology, choose ε > 0 such that BD (x, ε) ⊂ U .
Take N ∈ N large enough such that N1 < ε. Let
V = (x1 − ε, x1 + ε) × · · · × (xN − ε, xN + ε) × R × R × · · · × R.
13.1 Preliminaries 43
This is open in the product topology (it is a basis element). For y = (yi ) ∈ V ,
if i ≥ N then
¯ i , yi )
d(x 1
≤ < ε.
i N
If i < N , then
¯ i , yi )
d(x ε
≤ < ε.
i i
Thus, taking supremum, D(x, y) < ε, so y ∈ BD (x, ε), meaning that V ⊂
Bd (x, ε) < U , as desired.
Q∞
Conversely, let V = Vi be a basis element for the product topology. In
i=1
other words, Vi = R for i > N for some N ∈ N and Vi is open in R for i ≤ N .
For x = (xi ) ∈ V there exists an open U in the metric topology such that
x ∈ U ⊂ V . To see this, for each i ≤ N , choose 0 < εi < 1 such that
(xi − εi , xi + εi ) ⊂ Vi . Define
nε o
i
ε = min .
1≤i≤N i
Let U = BD (x, ε). Let y = (yi ) ∈ U . If i < N , then yi ∈ Vi = R. If i ≤ N , then

¯ i , yi )
d(x εi
≤ D(x, y) < ε <
i i
¯ i , yi ) < εi < 1. Hence
for all i ≤ N , so d(x
|xi − yi | < εi ,
so yi ∈ (xi − εi , xi + εi ) = Vi . Therefore U ⊂ V .
We are now equipped to prove:
Theorem 13.1.4 (Urysohn metrisation theorem). Every regular space X with

a countable basis is metrisable.

Proof. By Lemma 13.1.2, we have a collection of continuous functions fn
satisfying the properties in the lemma (chiefly, they separate points in X).
Define F : X → Y = RN by
F (x) = (f1 (x), f2 (x), . . . ).
Since fn is continuous, F is continuous.

For x 6= y, there exists fn such that fn (x) > 0 = fn (y) (since the two points
can be separated by an open set). Hence F (x) 6= F (y), so F is one-to-one. (I.e.,
F is an embedding of X into RN ).
We claim that X is homeomorphic to Z = F (X) ⊂ Y . Since Y is metrisable
(that’s Proposition 13.1.3), so if F (X).
It suffices to show that F −1 is continuous (since F is continuous, one-to-one,
and onto already). In other words, show that F is an open mapping, i.e., F (U )
is open in Z for each open U ⊂ X.
Let x0 ∈ U and F (x0 ) = z0 ∈ F (U ). We want to find an open neighbourhood
between z0 and F (U ).
44COMPLETELY REGULAR SPACE AND STONE–ČECH COMPACTIFICATION
Choose an index N such that fN (x0 ) > 0 and fN (X \ U ) = 0 (by construc-

−1
tion of the family fn ). Let V = πN ((0, ∞)), where πN : Y = RN → R with
πN ((xi )) = xN is the projection on the N th coordinate, which is a continuous
pullback of an open set, so open in Y .
Let W = V ∩Z, open in Z. Then z0 ∈ W . Now it suffices to show W ⊂ F (U ).
For z ∈ W , z = F (x) for some x ∈ X, πN (z) = fN (x) > 0. Since Fn (X \
U ) = 0, x ∈ U . Thus z = F (x) ∈ F (U ).
Thus finally X is homeomorphic to a subspace of RN , which is metrisable,
so X is metrisable as well.
This idea, contained in the proof, that we can find a family of functions
that separate the points, is generally speaking more useful than the Urysohn
metrisation theorem itself.
N
Remark 13.1.5. In fact, we embed X into [0, 1] since fn : X → [0, 1].
In fact what we proved in effect is this:

Theorem 13.1.6 (Embedding theorem). Let X is a T1 -space. Suppose fα α∈J
is an indexed family of continuous functions fα : X → R such that for each
x ∈ X and neighbourhood U of x, there exists α ∈ J such that fα (x) > 0 and
fα (X \ U ) = 0.
Then F : X ,→ RJ defined by
F (x) = (fα (x))α∈J
is an embedding of X into RJ .
Lecture 14 Completely regular space and Stone–

Čech compactification
14.1 Completely regular space
Definition 14.1.1 (Completely regular). A space X is completely regular if
X is T1 and for each x0 ∈ X and each closed subset E ⊂ X with x0 6∈ E, there
exists a continuous function f : X → [0, 1] such that f (x0 ) = 1 and f (E) = 0.
Remark 14.1.2. From Lemma 13.1.2 we know that a regular space with count-
able basis is completely regular.
Remark 14.1.3. From Urysohn metrisation theorem we know that T4 (normal)
implies completely regular.
We also know completely regular implies T3 (regular), since e.g. if x0 is a
point and E is a closed set, then x0 ∈ f −1 (( 21 , 1]) and E ⊂ f −1 ([0, 13 )) are
disjoint open sets.
Hence completely regular is between T3 and T4 ; for this reason (and because
T3 and T4 got their names before completely regular did), completely regular is
referred to as T3.5 .
Definition 14.1.4 (Compactification). (i) A compactification of a space
X is a compact Hausdorff space Y such that X ⊂ Y and X = Y .
14.1 Completely regular space 45
(ii) Two compactifications Y1 and Y2 of X are equivalent if there exists a

homeomorphism h : Y1 → Y2 such that h(x) = x for all x ∈ X.
Example 14.1.5. As we have previously discussed, (0, 1) has a compactification
to [0, 1] and one to S 1 by adding a point at infinity—these are not equivalent. N
Proposition 14.1.6. (i) A subspace of a completely regular space is com-
pletely regular.
(ii) Every locally compact Hausdorff space is completely regular.
Lemma 14.1.7. Let X be a topological space and let Z be a compact Hausdorff
space. Let h : X → Z be a continuous embedding (i.e., continuous and one-to-
one). Then there exists a corresponding compactification Y of X satisfying the
following property:
There exists an embedding H : Y → Z such that H = h on X. (I.e., we can
extend h to an embedding of Y into Z.)
The compactification Y is unique up to equivalence.
Remark 14.1.8. This compactification Y is called the compactification in-
duced by h.
Proof. The idea is this: since h is an embedding, we identify X with h(X) = X0 .
Now Y0 = h(X) is closed in Z, but Z is a compact Hausdorff space so Y0 is
also compact Hausdorff. Hence we identify the compactification we want by Y0 ,
because Y0 is therefore a compactification of X0 . Now we need to carry this
back to X.
Choose a set E disjoint from X that is in one-to-one correspondence with
the set Y0 . I.e., k : E → Y0 is one-to-one and onto.
Define Y = X t E, which we want to identify with Y0 . Define H : Y → Y0
by H(x) = h(x) if x ∈ X and H(y) = k(y) if y ∈ E. Now Y0 is a topological
space, but we don’t yet have a topology on Y .
Define a topology on Y by: U is open in Y if and only if H(U ) is open in
Y0 . Then H is a homeomorphism (it’s an open mapping and bijective) and X
is a subspace of Y . It is clear that X = Y . Hence Y is a compactification of X.
Finally, uniqueness: Suppose Y1 and Y2 are two compactifications of X with
H1 : Y1 → Z and H2 : Y2 → Z embeddings, such that Hi (x) = h(x) for all
x ∈ X.
Hence Hi (X) = h(X) = X0 for both i, and Hi (Yi ) = h(X) = X0 = Y0 .
Hence
H1 H2−1
Y1 X0 Y2
is a composition of homeomorphisms, so H2−1 ◦ H1 is a homeomorphism of Y1

and Y2 and H2−1 ◦ H1 = IdX on X.
Corollary 14.1.9. A topological space X has a compactification Y if and only
if X is completely regular.
Proof. The forward direction is Proposition 14.1.6.
J
By the Embedding theorem, we can embed h : X ,→ [0, 1] . The right-hand
side is compact Hausdorff (it’s a product of compact Hausdorff spaces, so by
46COMPLETELY REGULAR SPACE AND STONE–ČECH COMPACTIFICATION
Tychonoff’s theorem, infinite case and Proposition 5.2.3 the product is too). By
Lemma 14.1.7, X has a compactification induced by h.
In general, there are many compactifications for X.
Example 14.1.10. Let X = (0, 1) ⊂ R.
(i) The one-point compactification comes from h : X → S 1 by h(t) = cos(2πt), sin(2πt)).

Then Y = X.
(ii) Consider Id : X → R, Id(t) = t. Then Y = X = [0, 1]. N
This raises a question: if Y is a compactification of X, can a continuous

real-valued function f on X be extended continuously to Y ?
We should assume that f is a bounded function—if it is not, then there is no
hope of extending it to Y since Y , being compact, cannot have an unbounded
image.
That’s not quite enough though:
Example 14.1.11. Consider again X = (0, 1) ⊂ R.
(i) Let Y ' S 1 be the one-point compactification of X. A bounded continuous

function f : X → R is extendable to Y if and only if
lim f (x) and lim f (x)

x→0+ x→1−
both exist and are equal (since in the compactification those endpoints
are the same).
(ii) On the other hand, let Y = [0, 1]. Then a bounded continuous function
f : X → R is extendable to Y if and only if
lim f (x) and lim f (x)

x→0+ x→1−
exist. Here they don’t need to be equal, since the endpoints are not
identified. N
The crux is that the definition of f : X → R is independent of the compact-

ification Y . But given a compactification Y , we can figure out if it’s extendable
or not.
14.2 Stone–Čech compactification

So, question: can we find a compactification Y so that given any bounded
continuous function, we can extend it continuously to Y ? The answer is yes!
This compactification Y is called the Stone–Čech compactification of X.
The downside is that it is not easy to describe Y precisely.
Theorem 14.2.1 (Stone–Čech compactification, existence). Let X be a com-

pletely regular space. There exists a compactification Y of X having the property
that every bounded continuous function f : X → R extends uniquely to a con-
tinuous function on Y .
ANALYSIS 47

Proof. Let fα α∈J be the collection of all bounded continuous real-valued
functions on X. For each α ∈ J, define Iα = [inf fα (X), sup fα (X), ⊂]R (since
fα is bounded).
Q
Then define h : X → Iα by h(x) = (fα (x))α∈J .
α∈J

Since X is completely regular, fα satisfies the condition inQthe Embedding
theorem. Note also that by Tychonoff’s theorem, infinite case, Iα is compact
α∈J
Hausdorff.
Let Y be the compactification of X induced by h. Then we have an embed-
ding
Y
H: Y → Iα
α∈J
with H = h on X.
The construction of Y is independent of X since we built it using the set of
all bounded continuous real-valued functions on X.
Q a bounded continuous f : X → R, then f = fβ for some β ∈ J.
Now given
Let πβ : Iα → Iβ be the projection map. Then
α∈J
H Q πβ
πβ ◦ H : Y Iα Iβ
α∈J
is continuous. For x ∈ X,
πβ ◦ H(x)πβ ◦ h(x) = πβ (fα (x)α∈J ) = fβ (x) = f (x),
so πβ ◦ H extends f = fβ to Y .
Lecture 15 Analysis
15.1 Stone–Čech compactification
The uniqueness of the extension of f to Y is a consequence of the following
lemma:
Lemma 15.1.1. Let Z be a Hausdorff space. Let E ⊂ X and let f : E → Z be

continuous. Then there is at most one extension of f to a continuous function
F : E → Z.
Proof. Suppose F1 , F2 : E → Z are two different extensions of f . Thus there

exists some x ∈ E \ E such that F1 (x) 6= F2 (x). These are in Z, and Z is
Hausdorff, so we can separate them: there exist open neighbourhoods U1 and
U2 of F1 (x) and F2 (x) respectively, and U1 ∩ U2 = ∅.
Since F1 and F2 are continuous at x, there exist an open neighbourhood V
of x such that F1 (V ) ⊂ U1 and F2 (V ) ⊂ U2 .
Since x ∈ E, this means V ∩ E 6= ∅. Thus there exist y ∈ V ∩ E. But
U1 3 F1 (y) = f (y) = F2 (y) ∈ U2 , contradicting U1 ∩ U2 = ∅.
48 ANALYSIS
Theorem 15.1.2 (Extension property). Let X be a completely regular space.

Let Y be a compactification of X as in Theorem 14.2.1. Let C be a compact
Hausdorff space.
Then for any continuous map f : X → C, f extends uniquely to a continuous
function F : Y → C.
Proof. By Proposition 14.1.6, C is completely regular. By the Embedding the-
J J
orem, C ,→ [0, 1] for some index set J. Hence we may assume C ⊂ [0, 1]
and f : X → C is given by f (x) = (fα (x))α∈J . Hence each fα : X → [0, 1]
is a bounded continuous function. By Theorem 14.2.1, each fα extends to a
continuous function Fα : Y → R.
Define F : Y → RJ by F (y) = (Fα (y))α∈J . This is an extension of f , and
F (y) is continuous since each coordinate Fα (y) is continuous. Now since F is
continuous,
F (Y ) = F (X) ⊂ F (X) = f (X) ⊂ C = C
since C is compact in a Hausdorff space, hence closed.
So if we have a Stone–Čech compactification we can always extend continu-
ous maps with image in a compact Hausdorff space.
Theorem 15.1.3 (Stone–Čech compactification, uniqueness). Let X be a com-
pletely regular space. Suppose Y1 and Y2 are two compactifications of X satis-
fying the extension property Theorem 15.1.2.
Then Y1 and Y2 are equivalent.
Proof. Consider the two inclusion maps ι1 : X ,→ Y1 and ι2 : X ,→ Y2 .
Since Y1 has the extension property, ι2 extends to F2 : Y1 → Y2 , i.e.,
Y1
ι1
F2
X ι2 =f2
Y2
Similarly ι1 = f1 extends to F1 : Y2 → Y1 .
Consider
F2 F1
F1 ◦ F2 : Y1 Y2 Y1 .
For x ∈ X,
F1 ◦ F2 (x) = F1 (ι2 (x)) = F1 (x) = ι1 (x) = x,
so F1 ◦ F2 = Id on X. Hence F1 ◦ F2 = Id on Y1 .
Similarly, F1 ◦ F2 = Id in Y2 . Hence Y1 and Y2 are homeomorphic, because
F1 and F2 are continuous inverses between the two.
Definition 15.1.4 (Stone–Čech compactification). The compactification of X
in Theorem 14.2.1 is called Stone–Čech compactification of X, denoted by
β(X).
It is characterised by the property that any continuous map f : X → C,
where C is compact Hausdorff, extends uniquely to a continuous map F : β(X) →
C. In a picture,
β(X)
F
X f
C
15.2 Selected topological results in analysis 49
Exercise 15.1. Let X be a completely regular space and let Y be an arbitrary

compactification of X. Let β(X) be the Stone–Čech compactification of X.
Show that there exists a continuous surjective closed map f : β(X) → Y that
equals the identity map on X.
Remark 15.1.5. This shows that every compactification of X is equivalent to a
quotient space of β(X).
15.2 Selected topological results in analysis

Definition 15.2.1. Let (X, d) be a metric space.
∞
(i) A sequence xn n=1 in X is Cauchy if for any ε > 0 there exist N ∈ N
such that d(xn , xm ) < ε for all n, m ≥ N .
(ii) (X, d) is complete if every Cauchy sequence in X converges.
Example 15.2.2. For instance, (Rn , |·|) is a complete metric space. N
Exercise 15.2. The space (RN , D) is complete, where
¯ i , yi ) o
n d(x
D(x, y) = sup
i i
and
¯ i , yi ) = min |xi − yi |, 1 .

d(x
Note that, as we showed in Proposition 13.1.3, the above induces the product
topology on RN , so this space is complete.
Recall that a metric space (X, d) is totally bounded if for every ε > 0 there
exists a finite cover of X by ε-balls.
It is clear that a compact metric space is totally bounded; just cover with
ε-balls around every point and use compactness to pick a finite subcover. The
converse is not true:
Theorem 15.2.3. A metric space (X, d) is compact if and only if it is complete

and totally bounded.

The forward direction, as mentioned, is trivial. For any ε > 0, B(x, ε) x ∈
Proof.
X is an open cover of X. Since X is compact, this cover has a finite subcover,
which is the cover we need for totally bounded.
∞
We also need to show it is complete, so let xn n=1 be a Cauchy sequence
in X. Since X is compact,
X is also sequentially compact (since
∞ it is a metric
space). Hence xn has a convergent subsequence, say xnk k=1 , converging
to say x ∈ X.
Then xn → x as well: for any ε > 0, since xnk → x as k → ∞, there exists
N1 ∈ N such that d(xnk , x) < 2ε for all nk ≥ N1 . On the other hand, since xn

is Cauchy, there exists N2 ∈ N such that d(xn , xm ) < 2ε for all n, m ≥ N2 .

Now let N = max N1 , N2 . Then
ε ε
d(xn , x) ≤ d(xn , xnk ) + d(xnk , x) < + = ε,
2 2
so xn → x.
50 BAIRE CATEGORY THEOREM
For the converse, assume (X, d) is complete and totally bounded. We claim
X is sequentially compact (hence X is compact, since in a metric space they
are equivalent).
In other words,
we need to show that every sequence has a convergent sub-
∞
sequence. Let xn n=1 ⊂ X. This is a standard partition argument; partition
the space into parts containing infinitely many terms of the sequence, keep
subdividing to identify a convergent subsequence.
So first cover X by finitely many balls of radius 1 (doable since X is totally
bounded). Then there exists a ball, say B1 , containing
infinitely many xn (since
finitely many balls but infinite sequence). Say xn n∈J ⊂ B1 for some infinite
1
set J1 ⊂ N.
Now repeat: consider xn n∈J1 and cover X by finitely many radius 21 balls.

At least one ball, say B2 , contains infinitely many xn n∈J1 , say xn n∈J2 ⊂

B2 for some infinite set J2 ⊂ J1 . Rinse and repeat, so xn n∈J with Bk a ball
k
1
of radius 2k−1 and J1 ⊃ J2 ⊃ . . . .
∞
Take xnk k=1 with nk ∈ Jk . We shall show xnk is Cauchy. Thus
xnk → x for some x ∈ X since X is complete.
Given ε > 0, choose N ∈ N such that 2N1−2 < ε. For k, h > N , this means
nk , nj ∈ JN . Hence xnk , xnj ∈ BN , with radius 2N1−1 , so
1 1
d(xnk , xnh ) < 2 = N −2 < ε.
2N −1 2

Hence x nk is Cauchy.
15.3 Baire category theorem

∞ X is a Baire space if for any count-
Definition 15.3.1 (Baire space). A space
able collection of open dense sets Un n=1 in X,
∞
\
Un
n=1
is also dense in X.
The goal of this discussion is to show that a compact Hausdorff space is

Baire, and that a complete metric space is also Baire.
Lecture 16 Baire category theorem

16.1 Baire space
We ask what the definition of a Baire space is in terms of closet sets?
Let E ⊂ X be a subset. Then X \ E is dense in X (so X \ E = X) if and
only if every point in E is a limit point of X \ E.
This is true if and only if E contains no open subsets of X other than ∅, if
and only if E has no interior points, so has empty interior.
From this discussion it follows directly that
16.1 Baire space 51
Lemma 16.1.1. A space X is Baire if and only if for any countable collection
∞
of closed sets with empty interior En n=1 in X, the union
∞
[
En
n=1
has empty interior.
Theorem 16.1.2 (Baire category theorem). If X is a compact Hausdorff space

or X is a complete metric space, then X is a Baire space.
∞
Proof. Given a countable collection En n=1 of closed sets withe empty interior
in X, we want to show that
[∞
En
n=1
has empty interior. I.e., for any nonempty open set U ⊂ X, there exists x ∈ U
such that x 6∈ En for all n.
Since int(E1 ) = ∅ and U is open, we must have U 6⊂ E1 , so there exists
y1 ∈ U \ E1 .
Note that our assumption on X implies that X is regular. So there exists
an open neighbourhood U1 of y1 such that U1 ∩ E1 = ∅, and U1 ⊂ U .
Now consider U1 and E2 . Then by the same argument there exists y2 ∈
U1 \ E2 and open neighbourhood U2 of y2such that U2 ∩ E2 =∅ and
U2 ⊂ U1 .
Repeating this we obtain a sequence yn and open sets Un such that
yn ∈ Un−1 \ En , Un is an open neighbourhood of yn , Un ∩ En = ∅, and Un ⊂
Un−1 .
Moreover, if X is a metric space, we can also choose Un such that diam Un <
1
n.
Now we claim
∞
\
Un 6= ∅.
n=1
Then there exists some x in the intersection, so x 6∈ En for all n. But x ∈ U

since x ∈ Un for all n. Hence, if this claim is true, we are done.
We do this in two cases. First, X is compact Hausdorff. Assume
∞
\
Un = ∅.
n=1
Then given x ∈ X, there ∞ some n such that x 6∈ Un . Hence x ∈ X \Un ,

must exist
which is open. Hence X \ Un n=1 is an open cover of X. But X is compact,
m
so it has a finite subcover X \ Uni i=1 . I.e.,
m
[
X= (X \ Uni ).
i=1
Taking complement,
m
\
∅= Uni = Unm
i=1
since this is a sequence of shrinking sets. But ynm ∈ Unm , so it is nonempty,

which is a contradiction.
Secondly, suppose X is a complete metric space. This is a standard analysis
argument. Note that yn is a Cauchy sequence since for n > N , yn ∈ UN and
diam(UN ) < N1 . Now X is complete, so yn → x for some x ∈ X. For any fixed
n, yk ∈ Un for all k ≥ n. Then yk → x, so x ∈ Un fora any fixed n. Hence
∞
\
x∈ Un 6= ∅.
n=1
16.2 Applications of the Baire category theorem

This is clearly a topological result, but it has important applications.
Example 16.2.1. We give a topological proof that the real numbers R is un-
countable. ∞
Suppose R is countable. In other words, we can write R = xn n=1 . Each
one-point set is a closed set with empty interior.
But R is a complete metric space, so by the Baire category theorem it is a
Baire space. Hence
[∞

xn = R
n=1
has empty interior. But R obviously has interior, so this is a contradiction. N

Another question: Let fn : [0, 1] → R be continuous such that fn (x) → f (x)
pointwise.
We know f might not be continuous, since the continuity is not necessarily
uniform.
How large is the set x ∈ [0, 1] f is continuous at x ?
Lemma 16.2.2. Let X be a Baire space. Let Y ⊂ X be an open subspace.
Then Y is a Baire space.
∞
Proof. Let En n=1 be a countable collection of closed sets in Y with empty
interior. We need to show
∞
[
En
n=1
has empty interior.

Let En be the closure in X. Then En ∩ Y = En since En is closed in Y .
Notice that En has empty interior in X. Otherwise there exists nonempty
open U ⊂ X such that U ⊂ En , whence U ∩ Y ⊂ En and U ∩ Y is open in Y ,
so En has nonempty interior.
Since X is Baire,
[∞
En
n=1
has empty interior in X.

Now suppose
∞
[
En ⊂ Y
n=1
16.2 Applications of the Baire category theorem 53
has nonempty interior in Y . Then there exists some nonempty open W in Y

contained in this union. Thus W is also open in X. But then
∞
[ ∞
[
W ⊂ En ⊂ En ,
n=1 n=1
which is a contradiction to X being Baire.

Now we can answer our question.
Theorem 16.2.3. Let X be a Baire space and let (Y, d) be a metric space. Let
fn : X → Y be continuous for n = 1, 2, . . . . Assume fn (x) → f (x) pointwise for
some f : X → Y .
Then the set
E = x ∈ X f is continuous at x
is dense in X.
Proof. For any N ∈ N and ε > 0, define

EN (ε) = x ∈ X d(fn (x), fm (x)) ≤ ε for all n, m ≥ N .
In other words, the set of all x for which the fn are uniformly close.
For any fixed n and m, let

An,m (ε) = x ∈ X d(fn (x), fm (x)) ≤ ε .
Note that An,m (ε) is closed (it’s the pullback of a continuous map).
Thus we have \
En (ε) = An,m (ε)
n,m≥N
is an arbitrary intersection of closed sets, hence closed. Moreover,
E1 (ε) ⊂ E2 (ε) ⊂ . . . ,
and
∞
[
EN (ε) = X
N =1

since for any x ∈ X fixed, fn (x) converges in Y . Hence fn (x) is Cauchy,
so the tail is eventually close to each other, and so lies in some EN (ε).
Now let
∞
[
U (ε) := int(EN (ε)),
N =1
which is open in X.
We make two claims. First, U (ε) is dense in X for any ε > 0.
Second, f is continuous at each point in the set
∞
\ 1
C := U .
m=1
m
Since X is Baire, the first claim implies C is dense in X. Thus the theorem
follows by the second claim.
To prove the first claim it suffices to show for any nonempty open V in X,
there exists N such that V ∩ int(EN (ε)) 6= ∅.
By Lemma 16.2.2, V is Baire. Second, V ∩ EN (ε) is closed in V . Now
∞
[
X= EN (ε),
N =1
so
∞
[
V = (EN (ε) ∩ V )
N =1
where EN (ε) ∩ V are closed in V . Since V is Baire this means EN (ε) ∩ V
has nonempty interior in V for some N ∈ N. In other words, there exists a
nonempty open W in V such that
W ⊂ (EN (ε) ∩ V ).
But W is open in V and V is open in X, so W is also open in X. Thus
∅ 6= W ⊂ (int(EN (ε)) ∩ V ), and we are done.
Now to prove the second claim. Let x ∈ C. We want to show f is continuous
at x. I.e., for any ε > 0, there exists an open neighbourhood W of x such that
d(f (x), f (y)) < ε for all y ∈ W .
Choose k ∈ N such that k1 < 3ε . Since
∞
\ 1
x∈C= U ,
m=1
m
we must have x ∈ U ( k1 ). But

1 ∞
[ 1
U = int EN .
k k
N =1
So x ∈ int EM ( k1 ) for some M ∈ N. Since fM is continuous at x, there exists

an open neighbourhood W ∈ int EM ( k1 ) of x such that d(fM (x), fM (y)) 3ε for all
y ∈ W . Since W ⊂ int EM ( k1 ), for n > M and y ∈ W ,
1 ε
d(fn (y), fM (y)) ≤ < .
k 3
Now let n → ∞. Then
ε
d(f (y), fM (y)) <
3
for all y ∈ W . Thus for y ∈ W ,
d(f (y), f (x)) ≤ d(f (y), fM (y)) + d(fM (y), fM (x)) + d(fM (x), f (x)) < ε.
Hence f is continuous at x.
Exercise 16.1. Let X be a complete metric space. Let F be a subset of the
real-valued
functions on X. Suppose that for each x ∈ X, the set
continuous

Fx = f (x) f ∈ F is bounded. Show that there exists a non-empty open set
U of X such that F is uniformly bounded on U , i.e. there exists M ∈ N such
that |f (x)| ≤ M for all f ∈ F and for all x ∈ U .
0
Exercise 16.2. Let f : R → R be a differentiable function. Show that f : R → R
is continuous on a dense subset of R. Here f 0 is the usual derivative of f .
QUOTIENT SPACES AND TOPOLOGICAL GROUPS 55
Lecture 17 Quotient spaces and topological groups

17.1 Quotient spaces
The idea here is that we want to construct a new space from a known space.
Example 17.1.1. If we take a line segment from a point p to q, we can construct

a new space by gluing (identifying) p and q.
Given a rectangle we can identify the left-hand side L1 and the right-hand
side L2 with the same direction, in which case we get a cylinder.
If we now take this cylinder and glue the top circle C1 to the bottom circle
C2 , then we get a torus. N
We want to describe this process, this identifying or gluing, in an algebraic

way.
Let X be a topological space and let ∼ be an equivalence relation on X.
This gives us a partition of X into equivalence classes, denoted X/∼, called the
quotient space of X modulo ∼.
Now we want to give a topology on the quotient space, since X originally is
a topological space.
Note that we have a natural projection
π : X → X/∼
by π(x) = [x], the equivalence class of x. We would like how new topology to
make sure this projection map is continuous.
To this end we define the quotient topology on X/∼ by U ⊂ X/∼ is open
if and only if π −1 (U ) is open in X.
Exercise 17.1. Check that this in fact defined a topology on X/∼.
Remark 17.1.2. This is the smallest topology on X/∼ such that π is continuous.
Proposition 17.1.3. A function f : X/∼ → Y is continuous if and only if f ◦π

is continuous.
Theorem 17.1.4. Let X and Y be compact Hausdorff spaces. Let f : X → Y

be continuous and surjective. Define an equivalence relation on X by x0 ∼ x1
if and only if f (x0 ) = f (x1 ).
Then X/∼ is homeomorphic to Y .
Exercise 17.3. Verify that the relation ∼ in Theorem 17.1.4 is an equivalence

relation.
Proof. In a picture, we want to construct
f
X Y
π
g
X/∼
Date: October 22nd, 2020.

56 QUOTIENT SPACES AND TOPOLOGICAL GROUPS
so that g : X/∼ → Y is one-to-one and onto, and is also an open mapping.

To this end we define g : X/∼ → Y by g([x]) := f (x).
First, this definition depends on the representative x of the equivalence class
[x], so we need to check that g is well-defined. Fortunately, this is obvious: if
[x0 ] = [x1 ], i.e., x0 ∼ x1 , then by definition f (x0 ) = f (x1 ).
Second, since g ◦ π = f and f is continuous, Proposition 17.1.3 implies g is
continuous.
Third, g is one-to-one and onto. If g([x0 ]) = g([x1 ]), then f (x0 ) = f (x1 ), so
by definition x0 ∼ x1 . In other words, [x0 ] = [x1 ], so g is one-to-one. Onto is
trivial: since f is onto, so is g.
Finally, to show g is a homeomorphism it remains to check that g −1 is also
continuous. In other words, show that g is an open map (so sends open sets to
open sets). For U ⊂ X/∼, we need to show that g(U ) is open in Y . This is
equivalent to showing Y \ g(U ) is closed in Y . Since g is onto,
Y \ g(U ) = g(X/∼) \ g(U )
and since g is one-to-one
g(X/∼) \ g(U ) = g (X/∼) \ U = f (X \ π −1 (U )).

Now π −1 (U ) is open in X since U is open, hence X \π −1 (U ) is closed in X, which

is compact Hausdorff. Hence X \ π −1 (U ) is compact. Since f is continuous, this
means f (X \ π −1 (U )) is compact in Y , and since Y is also compact Hausdorff,
f (X \ π −1 (U )) must be closed in Y .
Thus g(U ) has closed complement, so g(U ) is open in Y .
Example 17.1.5. Consider f : [0, 1] → S 1 by f (t) = e2πit , 0 ≤ t ≤ 1.

Note that f (0) = f (1) = 1, and these are the only two points with the same
image.
Hence the quotient space [0, 1]/∼ ∼ = S 1 , which is the very act of gluing or
identifying the endpoints as discussed informally above. N
Example 17.1.6. Now consider X = [0, 1] × [0, 1]. Define f : X → S 1 × S 1 by

f (s, t) = (e2πis , e2πit ), 0 ≤ s, t ≤ 1.
Let us look at when f (s0 , t0 ) = f (s1 , t1 ). Clearly f (s0 , 0) = f (s0 , 1) since
the x-coordinates are the same and the y-coordinates differ by an integer, and
the period is 1. Similarly, f (0, t0 ) = f (1, t0 ).
Hence X/∼ is homeomorphic to the torus, and so S 1 × S 1 is in fact a way
to describe the torus. N
Exercise 17.4. Let X/∼ be the quotient space determined by an equivalence

relation “∼” on a topological space X. Prove the following:
(a) If X is compact, then X/∼ is compact.
(b) If X is connected, then X/∼ is connected.
(c) if X is path-connected, then X/∼ is path-connected.

17.2 Topological group 57
17.2 Topological group

Definition 17.2.1 (Topological group). A topological group G is both a T1
topological space and a group such that the two structures (topological and
group) are compatible.
In other words, the group operation G × G → G by (a, b) 7→ ab is continuous
(in the sense of the topological structure) and G → G by a 7→ a−1 is continuous.
Exercise 17.5. In fact, a topological group is always a regular space. Prove

this!
Example 17.2.2. The real numbers R with the usual topology is a topological
group, where the group operation is ordinary addition. N

Example 17.2.3. Consider the unit circle S 1 = eiθ 0 ≤ θ ≤ 2π ⊂ R2 with
the subspace topology. This is a topological group under multiplication. N
Example 17.2.4. Consider f : R → S 1 defined by f (r) = e2πir . Then f (r0 ) =

f (r1 ) if and only if r0 ≡ r1 (mod Z).
Clearly f is surjective, so the quotient space R/∼ = R/Z ∼ = S 1 are homeo-
morphic.
Now note that S 1 is a topological group under multiplication. On the other
hand, R is an abelian group, so every subgroup is normal. Hence R/Z is also a
group (under addition). In fact it is a topological group.
Now this means that the homeomorphism above says the topological struc-
tures of R/Z and S 1 are the equivalent, but are the group structures? The
answer is yes—the ∼ = above is also a group homomorphism (so in particular an
isomorphism). N
Proposition 17.2.5. Let G be a topological group. For any a ∈ G, the left

translation La : G → G defined by L1 (g) = ag and the right translation Ra : G →
G defined by Ra (g) = ga are homeomorphisms.
Proof. Obvious: La and Ra are both one-to-one and onto (their inverses are
La−1 and Ra−1 ).
It remains to check continuity, which in this case means it suffices to check
La and Ra are continuous, since they are their own inverses for some a.
But this is obvious:
embed multiplication
G G×G G
g (a, g) ag
La
is the composition of two continuous maps (embedding and the group multipli-
cation). Likewise for Ra .
Corollary 17.2.6. Let G be a topological group and let V be an open neigh-

bourhood of the identity e ∈ G. Then gV is an open neighbourhood of g ∈ G.
Proof. This is now obvious: the left translation Lg is a homeomorphism by

Proposition 17.2.5, so it sends open sets to open sets.
58 TOPOLOGICAL GROUPS
Similarly, given an open neighbourhood U of g, we can translate by g −1 and

get an open neighbourhood of the identity e.
This means that the topology of a topological group is completely determined
by neighbourhoods of the identity.
Theorem 17.2.7. Let G be a topological group. Let K denote the connected

component of G containing the identity e ∈ G. Then K is a closed normal
subgroup of G.
Proof. That K is closed is easy: a connected component is always closed in any

topological space (this is Exercise 6.4).
Next we need to show that K is a subgroup. In other words, we need to
check ba−1 ∈ K for any a, b ∈ K.
For a ∈ K, Ka−1 = Ra−1 (K) is connected (since Ra−1 is continuous, sending
connected sets to connected sets). Moreover e = aa−1 ∈ Ka−1 . In other words
e ∈ K (since K is the connected component of e). Thus for any b ∈ K, ba−1 ∈ K.
Next, K is normal. In other words, for any g ∈ G, gKg −1 ⊂ K. But this
can be rewritten as
gKg −1 = Lg (Rg−1 (K)),
which is again connected. It also contains e, since geg −1 = e. Hence gKg −1 ⊂
K. Thus K is indeed a closed normal subgroup of G.
Theorem 17.2.8. Let G be a connected topological group. Then for any open
neighbourhood V of the identity e ∈ G, the subgroup generated by V is G. In
other words, hV i = G.
Note that hV i is by definition the smallest subgroup of G containing V (it

is made up of arbitrary products of elements of V and their inverses).
Lecture 18 Topological groups

Proof. Let H = hV i. For any h ∈ H, hV is an open neighbourhood of h (since
left-translation is a homeomorphism, so sends open set to open set). Moreover
hV ⊂ H since H is a group. Hence every point in H is interior, so H is open in
G.
Now we claim G \ U is also open, meaning H is closed. But G is connected,
so this would mean H = G.
For g ∈ G \ H, we have gV ∩ H = ∅. Otherwise there exists x ∈ gV ∩ H,
so x = gv ∈ H for some v ∈ V , so h = xv −1 ∈ H because x, v ∈ H. Thus
gV ⊂ G \ H, but gV is an open neighbourhood of g, so G \ H is open.
18.1 Topological groups

Example 18.1.1. Consider Mn (R), the set of all real n × n matrices. We can
view Mn (R) ∼
2
= Rn with the usual Euclidean topology.
Consider the subset GLn (R) of invertible n × n matrices with the subspace
topology.
18.2 Orbit spaces 59
Note that GLn (R) with the usual matrix multiplication is a group. In fact,
GLn (R) is a topological group, in other words the group operations are contin-
uous. First
m : Mn (R) × Mn (R) Mn (R)
(A, B) AB
is continuous since each entry of AB is a polynomial in the entries of A and B.

Hence it is also continuous when restricted to GLn (R).
For A 7→ A−1 , the entries aren’t quite polynomials in the entries of A,
but by Cramer’s rule it is a polynomial in the entries of A divided by det A.
1
A 7→ det(A) 7→ det(A) is also continuous, since in GLn (R) we avoid determinant
0. N

Remark 18.1.2. The map det : GLn (R) → R \ 0 is surjective, and by the
above discussion also continuous.
A continuous function should send compact sets to compact sets, likewise
connected, so GLn (R) is not compact and not connected.
18.2 Orbit spaces

Definition 18.2.1 (Orbit space). A topological group G acts as a group home-
omorphism on a set X if each g ∈ G induces a homeomorphism of X satisfying
(i) g1 g2 (x) = g1 (g2 (x)) for all g1 , g2 ∈ G and x ∈ X,
(ii) e(x) = x for all x ∈ X where e ∈ G is the identity of G, and
(iii) the map g : X → X, x 7→ g(x) is a homeomorphism.
We define an equivalence relation ∼ on X by x0 ∼ x1 if and only if x0 = g(x1 )

for some g ∈ G. In other words, x0 and x1 are in the same orbit.
We denote the equivalence classes by G \ X (if acting on the left) or X/G (if
acting on the right) and endow it with the quotient topology. This space G \ X
(or X/G) is called the orbit space.
Example 18.2.2. The integers (Z, +) is a group. Notice how Z acts on R by
Z×R R
(n, r) n + r.
Hence r1 ∼ r2 if and only if r1 − r2 ∈ Z, describing the orbit space R/Z. N
Definition 18.2.3. If G acts transitively on X, then X is called a homoge-

neous space of G.
The group G acting transitively means for any x, y ∈ X, there exists g ∈ G
such that g(x) = y.
Example 18.2.4. Let G be a topological group and let H < G be a subgroup.

We know
H ×G G
(h, g) gh−1
60 TOPOLOGICAL GROUPS
is continuous. For each h ∈ H,
G G
g gh−1
is a homeomorphism (it’s right-translation by h−1 ). So we have the orbit space

G/H. N
In general, G/H is not a group. Indeed G/H is a group if and only if H is

normal in G.

Example 18.2.5. Let H = x + iy ∈ C x, y ∈ R, y > 0 , the complex
upper half-plane. The special linear group SL2 (Z) acts on H by linear fractional
transformation,
a b az + b
z= .
c d cz + d
Hence SL2 (Z) \ H is an orbit space. N
Exercise 18.1. Verify that the linear fractional transformation actually defines
a group action.
Example 18.2.6. Consider the map f : SL2 (R) → H defined by

a b ai + b
g= 7 g·i=
→ .
c d ci + d
The map f is surjective since

1 y x
x + iy = · i.
y 1
So f induces a homeomorphism SL2 (R)/∼ ∼ = H where g1 ∼ g2 if and only if

f (g1 ) = f (g2 ). In other words if and only if g1 · i = g2 · i, so (g2−1 g1 ) · i = i. In
other words g2−1 g1 ∈ StabSL2 (R) (i), the stabiliser of i.
Here we mean

StabSL2 (R) (i) = g ∈ SL2 (R) g · i = i = SO(2),
the special orthogonal group.

So
H∼= SL2 (R)/ SO(2).
Hence also
SL2 (Z) \ H ∼
= SL2 (Z) \ SL2 (R)/ SO(2).
So a function on the upper half-plane invariant under the action of SL2 (Z) can
be thought of as a function on SL2 (R) that is left-invariant under the action of
SL2 (Z) and right-invariant under the action of SO(2). N
Exercise 18.2. Let G be a topological group and let H be a subgroup of G. Let

G/H denote the collection of left cosets with the quotient topology.
(a) Show that the projection map π : G → G/H is an open map.

HOMOTOPY THEORY 61
(b) Show that H is a closed set in G if and only if G/H is a T1 space.

Exercise 18.3. Let G be a topological group. A neighbourhood
V of the identity
element e is said to be symmetric if V = V −1 . Here V −1 = v −1 v ∈ V .

(a) Suppose that U is a neighbourhood of e. Show that there exists a sym-

metric neighbourhood of e such that V · V ⊂ U .
(b) Show that G is a regular topological space.
This finishes the discussion of general topology in this course.
Lecture 19 Homotopy theory

19.1 Homotopy of paths
Definition 19.1.1 (Homotopic). Let X and Y be topological spaces. Let
f, g : X → Y be continuous. We say f is homotopic to g, denoted by f ' g, if
there exists a continuous function F : X × [0, 1] → Y such that F (x, 0) = f (x)
and F (x, 1) = g(x) for all x ∈ X.
In other words, we can continuously move the image f (x) to the image g(x).
I.e., γt (x) = F (x, t) : X → Y for 0 ≤ t ≤ 1 is a family of continuous functions,
continuously deforming from f (x) to g(x).
Now let’s consider the special case where f and g are paths in X. Recall if
γ : [0, 1] → X is continuous, γ(0) = x0 and γ(1) = x1 , then γ is called a path
in X from x0 to x1 .
Definition 19.1.2 (Path homotopic). Two paths γ and γ 0 in X from x0 to x1
are path homotopic, denoted by γ 'p γ 0 , if there exists a continuous function
F : [0, 1]×[0, 1] → X such that F (s, 0) = γ(s) and F (s, 1) = γ 0 (s) (so homotopic)
and F (0, t) = x0 and F (1, t) = x1 for all 0 ≤ t ≤ 1 (so at every t it is still a
path from x0 to x1 ).
Lemma 19.1.3. The homotopy relation ' and the path homotopy relation 'p
are equivalence relations on

A = f : X → Y continuous
and

A(x0 , x1 ) = γ : [0, 1] → X continuous, γ(0) = x0 , γ(1) = x1
respectively.
Remark 19.1.4. If γ is a path, we denote its path homotopy equivalence class
by [γ].
Proof. We will show ' is an equivalence relation; path homotopy is very similar.
Reflexivity is obvious: f ' f by F (x, t) = f (x) for all x ∈ X and all t ∈ [0, 1].
For symmetry, suppose f ' g, say by a continuous F (x, t) such that F (x, 0) =
f (x) and F (x, 1) = g(x). Take G(x, t) = F (x, 1 − t). Then G(x, 0) = F (x, 1) =
g(x) and G(x, 1) = F (x, 0) = f (x). Hence g ' f .
62 HOMOTOPY THEORY
Finally, transitivity. Suppose f ' g and g ' h, say the first one by F (x, t)
and the second by G(x, t). Take
(
F (x, 2t), if 0 ≤ t ≤ 21 ,
H(x, t) =
G(x, 2t − 1), if 21 ≤ t ≤ 1.
Then H(x, 0) = F (x, 0) = f (x) and H(x, 1) = G(x, 1) = h(x). Moreover H is

continuous (only t = 21 needs to be checked); it is made up of two continuous
functions which agree on a closed set.
Example 19.1.5. Let γ1 and γ2 be two paths from x0 to x1 in R2 . Then
γ1 'p γ2 . For instance, by taking the convex combination of the two paths,
F (s, t) = (1 − t)γ1 (s) + tγ2 (s).
N
This argument works in slightly more generality:
Remark 19.1.6. Let γ1 , γ2 be two paths from x0 to x1 in a convex space X.
Then γ1 'p γ2 . (Since in a convex space the line segment connecting the two
at a fixed time is still in the space because of convexity.)
19.2 Fundamental group

Let γ0 be a path in X from x0 to x1 and let γ1 be a path in X from x1 to x2 .
Define γ0 ∗ γ1 to be the path from x0 to x2 given by
(
γ0 (2s), if 0 ≤ s ≤ 21 ,
γ0 ∗ γ1 (s) =
γ1 (2s − 1), if 12 ≤ s ≤ 1.
This induces an operation on the path homotopy classes:
[γ0 ] ∗ [γ1 ] := [γ0 ∗ γ1 ].
Exercise 19.1. Check that ∗ on the path homotopy classes is well-defined (i.e.,
doesn’t depend on the choice of representatives γ0 and γ1 ).
Proposition 19.2.1. (i) The operation ∗ is associative. In other words, let
γ0 be a path from x0 to x1 , γ1 be a path from x1 to x2 , and γ2 a path from
x2 to x3 . Then
([γ0 ] ∗ [γ1 ]) ∗ [γ2 ] = [γ0 ] ∗ ([γ1 ] ∗ [γ2 ]).
(ii) The operation ∗ has identities. Given x ∈ X, ex : [0, 1] → X, ex (s) = x

be the constant path. Let γ be a path from x0 to x1 . Then
[ex0 ] ∗ [γ] = [γ] = [γ] ∗ [ex1 ].
(iii) The operation ∗ has inverses. Let γ be a path from x0 to x1 . Let γ̄(s) :=
γ(1 − s). Then
[γ] ∗ [γ̄] = [ex0 ] and [γ̄] ∗ [γ] = [ex1 ].

19.3 Changing base point 63

Remark 19.2.2. This means ∗ is a groupoid operation, but not a group oper-
ations, since the left and right identities are not necessarily equal.
On the other hand this means:
Remark 19.2.3. If we consider

A(x0 , x0 ) = γ : [0, 1] → X continuous, γ(0) = γ(1) = x0
then ∗ is a group operation on the path homotopy classes.
Definition 19.2.4. (i) Let X be a space and let x0 ∈ X. A path in X that
begins and ends at x0 is called a loop at x0 .
(ii) The set of path homotopy classes of loops based at x0 with the operation
∗ is called the fundamental group of X relative to the base point x0 ,
denoted by π1 (X, x0 ).

Example 19.2.5. For any x0 ∈ R2 , π1 (R2 , x0 ) = e = 0, the trivial group,
since all paths in R2 are path homotopic by Example 19.1.5. N
In general, if X is convex, then π1 (X, x0 ) = 0 for all x0 ∈ X. In particular,
π1 (Rn , x0 ) = 0.
19.3 Changing base point

Let x0 , x1 ∈ X and let α be a path from x0 to x1 . Then α induces a group
homomorphism
α̂ : π1 (X, x0 ) → π1 (X, x1 )
given by
α̂([γ]) = [ᾱ] ∗ [γ] ∗ [α] = [ᾱ ∗ γ ∗ α].
(Recall ᾱ(s) = α(1 − s) is α in reverse.)
Exercise 19.3. Verify that ᾱ is a group homomorphism.
Theorem 19.3.1. Even better: α̂ is a group isomorphism.
Proof. Let β(s) = ᾱ(s). This is a path from x1 to x0 . Then β̂ : π1 (X, x1 ) →
π1 (X, x0 ) is a group homomorphism, and α̂ and β̂ are each others inverses.
Lecture 20 Covering spaces

20.1 Simply connected space
Corollary 20.1.1. If X is path connected, then for any x0 , x1 ∈ X, we have
π1 (X, x0 ) ∼
= π1 (X, x1 ).
Remark 20.1.2. This isomorphism depends on the path α from x0 to x1 .
Two different paths may induce different isomorphisms.
Definition 20.1.3 (Simply connected). A space X is simply connected if X
is path connected and π1 (X, x0 ) = 0.
Note that since the space is path connected, the fundamental group π1 (X, x0 )
does not depend on the choice of x0 ∈ X in the first place.
Date: November 3rd, 2020.
64 COVERING SPACES
20.2 Induced homomorphism

Let h : X → Y be a continuous map such that h(x0 ) = y0 . Write h : (X, x0 ) →
(Y, y0 ). Then h induces a homomorphism h∗ : π1 (X, x0 ) → π1 (Y, y0 ) given by
h∗ ([γ]) = [h ◦ γ].
Exercise 20.1. Verify that h∗ is a group homomorphism.
Theorem 20.2.1. Let h : (X, x0 ) → (Y, y0 ) and k : (Y, y0 ) → (Z, z0 ) be contin-
uous. Then (k ◦ h)∗ = k∗ ◦ h∗ .
Exercise 20.2. Prove Theorem 20.2.1.
Corollary 20.2.2. If h : (X, x0 ) → (Y, y0 ) is a homeomorphism, then
h∗ : π1 (X, x0 ) → π1 (Y, y0 )
is an isomorphism.
Hence the fundamental group π1 is a topological invariant.
20.3 Covering spaces

Definition 20.3.1 (Covering space). (i) Let p : E → X be a continuous sur-
jective map. An open set U ⊂ X is evenly covered by p if p−1 (U ) is a
union of disjoint open subsets Vα ⊂ E such that

pV : Vα → U
α
is a homeomorphism for all α.

(ii) Let p : E → X be a continuous surjective map. If each x ∈ X has a
neighbourhood U that is evenly covered by p, then p is called a covering
map, and E is called a covering space of X.
Example 20.3.2. Consider p : R → S 1 defined by p(t) = e2πit . This p is a
covering map. N
Definition 20.3.3 (Fibre). Let p : E → X be a covering map. Let x ∈ X.
Then p−1 (x) is called the fibre over x.
Remark 20.3.4. The fibre p−1 (x) has the discrete topology, and for each x ∈
X there is an open neighbourhood U such that p−1 (U ) is homeomorphic to
p−1 (x) × U .
20.4 Lifting and universal covering spaces

Definition 20.4.1. Let p : E → X be a covering map. Let f : Y → X be a
continuous map. A continuous map g : Y → E is called a lift of f if p ◦ g = f .
In other words, a lift is a map making the diagram
E
g
p
Y f
X
commute.
20.5 Existence of path lifting 65
Lemma 20.4.2 (Uniqueness of lifts). Let p : E → X be a covering map and Y

be a connected space. Let f : Y → X be a continuous map and let g, h : Y → E
be two lifts of f .
Suppose g(y0 ) = h(y0 ) for some y0 ∈ Y . Then g(y) = h(y) for all y ∈ Y .

Proof. Let A = y ∈ Y g(y) = h(y) and B = y ∈ Y g(y) 6= h(y) .
Obviously y0 ∈ A, so A 6= ∅. Moreover Y = A t B.
We claim A and B are open. Thus A = Y since Y is connected and A 6= ∅.
In other words, g = h.
Let y ∈ Y and U be an open neighbourhood of f (y) that is evenly covered
by p. In other words G
p−1 (U ) = Vα
α

and pVα : Vα → U is a homeomorphism.
We have two cases: if y ∈ A, i.e., g(y) = h(y) ∈ p−1 (U ), then g(y) =
h(y) ∈ Vα for some α. Now g and h are continuous, so there exists an open
neighbourhood W of y such that g(W ) ⊂ Vα and h(W) ⊂ Vα .
For any z ∈ W , p ◦ g(z) = f (z) = p ◦ h(z). Since pVα is a homeomorphism,
it is in particular one-to-one, so g(z) = h(z). Thus y ∈ W ⊂ A, so A is open.
Second case, suppose y ∈ B, so g(y) 6= h(y). Since p ◦ g(y) = f (y) = p ◦ h(y),
the points g(y) and h(y) cannot be in the same Vα . Hence g(y) ∈ Vα and
h(y) ∈ Vβ for some α 6= β. Thus Vα ∩ Vβ = ∅.
Since g and h are continuous, there exists an open neighbourhood W of y
such that g(W ) ⊂ Vα and h(W ) ⊂ Vβ . So for z ∈ W , g(z) 6= h(z). Hence
W ⊂ B, so B is open.
This means that a map from a connected space has unique lifts, if a lift
exists at all. But do they exist?
20.5 Existence of path lifting

Theorem 20.5.1 (Path lifting theorem). Let p : E → X be a covering map.
Let γ : [0, 1] → X be a path and let e0 ∈ E such that p(e0 ) = γ(0). Then there
exists a unique path α : [0, 1] → E such that α(0) = e0 and p ◦ α = γ (so α is a
lift of γ).
Proof. The uniqueness follows from the lemma since we fixed the starting point
e0 .
On to existence. The idea is simple: pull back the path γ on the neighbour-
hoods on which we have one-to-one and onto, due to the covering map.
That is, for each x ∈ X, choose an open neighbourhood Ux of x that is
evenly covered by p. Then γ −1 (Ux ) x∈X is an open cover of [0, 1]. Since [0, 1]
is compact, by Lebesgue number lemma there exist 0 = s0 < s1 < s2 < · · · <
sm = 1 and U1 , U2 , . . . , Um such that [sj−1 , sj ] ⊂ γ −1 (Uj ).
p(e0 ) = γ(0) ∈ U1 , there is an open neighbourhood V1 of e0 such
Since
that pV1 : V1 → U is a homeomorphism. Define the lift α of γ on [0, s1 ] by

α(t) = (pV1 )−1 (γ(t)) for 0 ≤ t ≤ s1 .
Since γ([s1 , s2 ]) ⊂ U2 , we have γ(s1 ) ∈ U2. Set e1 = α(s1 ) ∈ E. There exists
an open neighbourhood V2 of e1 such that pV : V2 → U2 is a homeomorphism.
2
Define α on [s1 , s2 ] by α(t) = (pV )−1 (γ(t)) for s1 ≤ t ≤ s2 .
2
66 UNIVERSAL COVERING SPACE
Continuing this process (in finitely many steps since [0, 1] is compact), we
have a lift α of γ.
Lecture 21 Universal covering space

21.1 Lifting
Theorem 21.1.1 (Lifting of path homotopy). Let p : E → X be a covering
map. Let F : [0, 1] × [0, 1] → X be a continuous map. Let e0 ∈ E satisfy
p(e0 ) = F (0, 0) = x0 . Then there exists a unique lift G : [0, 1] × [0, 1] → E of F
such that G(0, 0) = e0 .
Moreover, if F is a path homotopy, then G is a path homotopy.
Proof. The uniqueness follows from Lemma 20.4.2.
Now existence. The idea is similar to path lifting in Theorem 20.5.1. Con-
sider the path s 7→ F (s, 0). By the path lifting theorem, there exists a unique
path α(s) in E such that p ◦ α(s) = F (s, 0) and α(0) = e0 .
Similarly, consider the path t 7→ F (0, t). Again there exists a unique lift β(t)
in E such that p ◦ β(t) = F (0, t) and β(0) = e0 .
Use the Lebesgue number lemma, so there exist 0 < s0 < s1 < s2 < · · · <
sm = 1 and 0 < t0 < t1 < t2 < · · · < tm = 1 such that
F ([si−1 , si ] × [tj−1 , tj ]) ⊂ Uij
in X that is evenly covered by p.

First consider F ([0, s1 ] × [0, t1 ]) ⊂ U00 . Let V00 ⊂ E such that pV : V00 →
00
U00 is a homeomorphism such that α([0, s1 ]) ⊂ V00 and β([0, t1 ]) ⊂ V00 .
Define the lift G on F on [0, s1 ] × [0, t1 ] by the pullback
G(s, t) = (pV00 )−1 (F (s, t)).

Next consider
F ([s1 , s2 ] × [0, t1 ]) ⊂ U10 .
Let V10 ⊂ E such that
pV : V10 → U10
10
is a homeomorphism and α([s1 , s2 ]) ⊂ V10 and β([0, t1 ]) ⊂ V10 . Define G on

[s1 , s2 ] × [0, t1 ] by the pullback
G(s, t) = (pV10 )−1 (F (s, t))

Continue this process for each block [si−1 , si ] × [ti−1 , ti ] to fill out [0, 1] × [0, 1].
We get a lift G of F defined on [0, 1]×[0, 1] such that p◦G = F and G(0, 0) = e0 .
Now assume F is a path homotopy. So say F (0, t) = x0 and F (1, t) = x1 for
all t ∈ [0, 1].
Since p(G(0, t)) = F (0, t) = x0for all t, G( 0 × [0, 1]) ⊂ p−1 (x0 ). On the

one hand, p−1 (x0 ) is discrete, but 0 × [0, 1]is connected

and G is continuous,
so the left-hand side is continuous. Hence G( 0 × [0, 1]) is a single point set,
so G(0, t) = G(0, 0) = e0 for all t ∈ [0, 1].
Similarly, G(1, t) = G(1, 0) for all t, so G is a path homotopy.
Date: November 5th, 2020.
21.1 Lifting 67
Remark 21.1.2. Let p : (E, e0 ) → (X, x0 ) be a covering map.

First, let γ : [0, 1] → X be a loop based at x0 . (I.e., γ(0) = γ(1) = x0 .)
By the Path lifting theorem, there exists a unique α : [0, 1] → E such that
p ◦ α = γ and α(0) = e0 .
Now p ◦ α(1) = γ(1) = x0 , so α(1) ∈ p−1 (x0 ). However α(1) need not be e0 .
For example, p(t) = e2πit from the previously example applied to γ(t) = e2πit
has α(0) = 0 and α(1) = 1.
Second, suppose two loops γ1 , γ2 : [0, 1] → X based at x0 are homotopic
Then their lifts α1 , α2 : [0, 1] → E are path homotopic. Thus α1 (1) = α2 (1)
(but we don’t know if they’re loops).
So we can define a function Φ : π1 (X, x0 ) → p−1 (x0 ) by [γ] 7→ α(1) where α
is a lift of γ. Note by the previous paragraph Φ is well-defined.
Theorem 21.1.3. Let p : (E, e0 ) → (X, x0 ) be a covering map. Suppose E is
simply connected. Then Φ : π1 (X, x0 ) → p−1 (x0 ) is one-to-one and onto.
Proof. Let’s start with onto. For y ∈ p−1 (x0 ), let α be a path in E from e0 to y
(which exists since E is simply connected, hence path connected). Let γ = p ◦ α,
then γ is a loop in X based at x0 , and α is a lift of γ such that α(0) = e0 . Thus
Φ([γ]) = α(1) = y, so Φ is onto p−1 (x0 ).
Next one-to-one. Suppose γ0 and γ1 are two loops in X based at x0 such
that Φ([γ0 ]) = Φ([γ1 ]). We want to show [γ0 ] = [γ1 ], in other words γ0 ' γ1 .
Let α0 and α1 be lifts of γ0 and γ1 respectively, such that α0 (0) = α1 (0) = e0 .
Thus α0 ∗ ᾱ1 is a loop in E based at e0 .
Since E is simply connected, there exists a homotopy F : [0, 1] × [0, 1] → E
such that F (0, t) = α0 ∗ ᾱ1 (t) and F (1, t) = e0 .
Projecting this homotopy, p ◦ F : [0, 1] × [0, 1] → X is a homotopy of the loop
γ0 ∗ γ̄1 and the constant map x0 . Hence
[γ0 ∗ γ̄1 ] = [x0 ]
and [x0 ] is the identity in π1 (X, x0 ). But [γ̄1 ] = [γ1 ]−1 , so [γ0 ] = [γ1 ]. Hence Φ
is one-to-one.
Definition 21.1.4 (Universal covering space). Suppose p : (E, e0 ) → (X, x0 )
is a covering map. If E is simply connected, then E is called a universal
covering space of X.
Remark 21.1.5. Not every space has a universal covering space.
Consequently, a question: When does a universal covering space exist? The
answer is fairly technical and we won’t go into it here beyond stating the result:
Definition 21.1.6 (Semilocally simply connected). A space X is called semilo-
cally simply connected if for each x0 ∈ X there exists a neighbourhood U
of x such that the homomorphism i∗ : π1 (U, x0 ) → π1 (X, x0 ) induced by the
inclusion map i : U ,→ X is trivial.
Theorem 21.1.7. A space X has a universal covering space if and only if X
is path connected, locally path connected, and semilocally simply connected.
The point is this: if there is a universal covering space of X, then the
fundamental group π1 (X, x0 ) is in one-to-one correspondence with the fibre
p−1 (x0 ).
We’ll use this to show π1 (S 1 , 1) ∼
= Z.
68 HOMOTOPY TYPE
Exercise 21.1. Let p : (E, e0 ) → (X, x0 ) be a covering map and let E be path-
connected. Prove the following:
(a) p∗ : π1 (E, e0 ) → π1 (X, x0 ) is injective.
(b) There is a bijection φ : H \ π1 (X, x0 ) → p−1 (x0 ) where H = p∗ (π1 (E, e0 ))

and H \ π1 (X, x0 ) is the collection of right cosets of H in π1 (X, x0 ).
Lecture 22 Homotopy type

22.1 Fundamental group of the unit circle
Example 22.1.1. We will show π1 (S 1 , 1) ∼ = Z as groups.
Consider the covering map p : (R, 0) → (S 1 , 1) given by p(t) = e2πit .
Note that since R is simply connected (since R is convex), p is a universal
covering map. Secondly, p−1 (1) = Z.
So Φ : π1 (S 1 , 1) → π −1 (1) = Z is one-to-one and onto. We also want to
show, then, that Φ is a group isomorphism.
Let γ be a loop in S 1 based at 1. Let α be a lift of γ in R with α(0) = 0.
Then p ◦ α(1) = γ(1) = 1, so in other words e2πiα(1) = 1, so α(1) ∈ Z.
Define the index of γ by
ind(γ) = α(1) ∈ Z,
so the index counts how many times γ makes full revolutions (with sign). Let
γ1 and γ2 be two loops in S 1 based at 1. Since Φ is one-to-one, [γ1 ] = [γ2 ] if
and only if ind(γ1 ) = ind(γ2 ).
Now we claim Φ : π1 (S 1 , 1) → Z = p−1 (1) given by [γ] = ind(γ) is a group
isomorphism.
In other words, we want to show
ind(γ1 ∗ γ2 ) = ind(γ1 ) + ind(γ2 ).
Let α1 and α2 be lifts of γ1 and γ2 respectively, such that α1 (0) = α2 (0) = 0.

Thus p ◦ α1 = γ1 and p ◦ α2 = γ2 , or on other words γ1 (t) = e2πiα1 (t) and
γ2 (t) = e2πiα2 (t) .
We want to compute ind(γ1 ∗ γ2 ), so first we need a lift of γ1 ∗ γ2 . Define
β(t) as
(
α1 (2t), if 0 ≤ t ≤ 21 ,
β(t) =
α1 (1) + α2 (2t − 1), if 12 ≤ t ≤ 1.
Then β(t) is continuous, and indeed β is a lift of γ1 ∗ γ2 and β(0) = α1 (0) = 0.

Then
ind(γ1 ∗ γ2 ) = β(1) = α1 (1) + α2 (1) = ind(γ1 ) + ind(γ2 ). N

22.2 Homotopy type 69
22.2 Homotopy type

Recall that f : (X, x0 ) → (Y, y0 ), continuous, induces a group homomorphism
f∗ : π1 (X, x0 ) → π1 (Y, y0 ) given by [γ] 7→ [f ◦ γ].
Recall also that if X is path connected, then α̂ : π1 (X, x0 ) → π1 (X, x1 ) is a
(noncanonical) isomorphism.
Theorem 22.2.1. Let F : X×[0, 1] → Y be a homotopy of two maps f : (X, x0 ) →
(Y, y0 ) and g : (X, x0 ) → (Y, y1 ). (I.e., F (x, 0) = f (x) and F (x, 1) = g(x).)
Let α(t) = F (x0 , t), 0 ≤ t ≤ 1, a path from y0 to y1 .
Then g∗ = α̂ ◦ f∗ . In other words, the diagram
f∗
π1 (X, x0 ) π1 (Y, y0 )
α̂ isomorphism
g∗
π1 (Y, y1 )
commutes. In particular, f∗ is an isomorphism if and only if g∗ is an isomor-

phism.
Proof. Let γ : [0, 1] → X be a loop based at x0 . We need to show α̂(f∗ ([γ])) =
g∗ ([γ]). In other words, show
[ᾱ] ∗ [f ◦ γ] ∗ [α] = [g ◦ γ].
Note that f can be moved to g by the homotopy F . Consider therefore the map
G : [0, 1] × [0, 1] → Y defined by
G(s, t) = F (γ(s), t).
Let β1 , β2 , β3 , β4 be the paths along the edges of [0, 1] × [0, 1]: β1 to the right
along the bottom, β2 going up along the left-hand side, β4 going up along the
right-hand side, and β3 going to the right along the top.
So G ◦ β1 = F (γ(s), 0) = f ◦ γ, G ◦ β2 = F (γ(0), t) = F (x0 , t) = α. We have
G ◦ β3 = F (γ(s), 1) = g ◦ γ, and finally G ◦ β4 = F (γ(1), t) = F (x0 , t) = α.
Note that [0, 1] × [0, 1] is convex. Hence any path in the square is homotopic
to any other path with the same start and end points. Hence β3 ' β̄2 ∗ β1 ∗ β4 .
Compose with G, so
G ◦ β3 ' G ◦ (β̄2 ∗ β1 ∗ β4 ),
so in other words
[g ◦ γ] = [ᾱ ∗ (f ◦ γ) ∗ α].
Definition 22.2.2 (Homotopy type). Let X and Y be topological spaces.
(i) A map f : X → Y is a homotopy equivalence if there exists a map
g : Y → X such that f ◦ g ' IdY and g ◦ f ' IdX .
The map g is called a homotopy inverse of f .
(ii) The spaces X and Y are homotopy equivalent (or have the same ho-
motopy type) if there is a homotopy equivalence between X and Y .
Theorem 22.2.3. Let f : (X, x0 ) → (Y, y0 ) be a homotopy equivalence. Then
the induced map f∗ : π1 (X, x0 ) → π1 (Y, y0 ) is an isomorphism.
70 HOMOTOPY TYPE
Proof. Let g be a homotopy inverse of f , and x1 = g(y0 ), i.e. g : (Y, y0 ) →

(X, x1 ).
So
f g
(X, x0 ) (Y, y0 ) (X, x1 )
g◦f 'IdX
Then
(g ◦ f )∗ = g∗ ◦ f∗ : π1 (X, x0 ) → π1 (Y, y0 ) → π1 (X, x1 ),
but (IdX )∗ = Idπ1 (X,x0 ) .

By Theorem 22.2.1, g∗ ◦f∗ = IdX so f∗ is one-to-one and g∗ is onto. Similarly,
consider f ◦ g, saying f∗ is onto and g∗ is one-to-one.
Remark 22.2.4. If X and Y are homeomorphic, then X and Y have the same
homotopy type.
The converse is not true: for instance a single point x0 is homotopy
equivalent to R, but they are certainly not homeomorphic.
X is contractible if X is homotopy
Definition 22.2.5 (Contractible). A space
equivalent to a single-point space Y = y0 .
Corollary 22.2.6. A contractible space is simply connected.
Exercise 22.1. Prove that if X is contractible and Y is path connected, then

any two maps from X to Y are homotopic.
Exercise 22.2. Let X be a path-connected space and let x0 , x1 ∈ X. Show that
π1 (X, x0 ) is abelian if and only if for any paths α, β from x0 to x1 , we have
α̂ = β̂.
Exercise 22.3. Let X be a topological space and let x0 ∈ X. Suppose that there
is a continuous map F : X × [0, 1] → X such that
F (x, 0) = x0 , x ∈ X,
F (x, 1) = x, x ∈ X,
F (x0 , t) = x0 , 0 ≤ t ≤ 1.
(a) Show that X is path connected.
(b) Show that X is simply connected.
Exercise 22.4. A subset A of Rn is called star convex with respect to a0 ∈ A if

all the line segments joining a0 to any other points of A lie in A.
(a) Find a star convex set that is not convex.
(b) Show that a star convex set is simply connected.
Exercise 22.5. Let X be a simply connected topological space. Let x0 , x1 ∈ X.

Show that any two paths from x0 to x1 are homotopic.
FUNDAMENTAL GROUP CALCULATIONS 71
Lecture 23 Fundamental group calculations

23.1 Fundamental group of the punctured plane
Theorem 23.1.1. Let x0 ∈ S 1 . Let 0 = (0, 0) ∈ R2 . The inclusion map
1 2 1 2
, x0 ) → (R \ 0 , x0 ) induces an isomorphism ι∗ : π1 (S , x0 ) → π1 (R \
ι: (S
0 , x0 ).
Hence π1 (R2 \ 0 , x0 ) = Z.

Remark 23.1.2. Since R2 \ 0 is path-connected, we can replace x0 by any
base point x 6= 0 above.
to show ι is an homotopy
Proof. It suffices
x
equivalence. Define the continuous
map r : R2 \ 0 → S 1 by r(x) = kxk . Then r ◦ ι = IdS 1 quite obviously. We
need to also show
r ι
ι ◦ r : R2 \ 0 S1 R2 \ 0
is homotopic to Id . Note
R2 \ 0
x x
ι ◦ r(x) = ι
= .
kxk kxk

Define F : R2 \ 0 × [0, 1] → R2 \ 0 by
x
F (x, t) = t + (1 − t)x.
kxk
x
Then F (x, 0) = x = Id 2 (x) and F (x, 1) = kxk = ι ◦ r(x), so ι ◦ r '
R \ 0
Id .
R2 \ 0
Using the same argument, we can prove

Theorem
23.1.3. Let x0 ∈ S n−1 . The inclusion map ι : (S n−1 n
, x0 ) → (R \
n−1 n
0 , x0 ) induces an isomorphism ι∗ : π1 (S , x0 ) → π1 (R \ 0 , x0 ).
This raises the question: what is (S n−1 , x0 ) for n ≥ 3? The answer is 0. We
will show this later.
23.2 Fundamental group of S n

Theorem 23.2.1 (van Kampen theorem (special case)). Let X = U ∪ V where
U and V are open sets in X. Suppose U and V are both simply connected and
U ∩ V is path connected. Let x0 ∈ U ∩ V . Then π1 (X, x0 ) = 0. (So X is simply
connected.)
Sketch of proof. Let f : [0, 1] → X be a loop based at x0 . We need to show f is
path-homotopic to the constant map x0 .
Step 1: By the Lebesgue number lemma there exists a subdivision of [0, 1],
say
0 = a0 < a1 < a2 < · · · < an = 1
72 FUNDAMENTAL GROUP CALCULATIONS
such that f ([ai , ai+1 ]) ⊂ U or V (since U ∪ V is an open cover of the compact

set f ([0, 1])). We can choose ai such that f (ai ) ∈ U ∪ V for all i. (Namely,
remove ai and extent the subsegments until this happen; since the list is finite
this is doable.)
Step 2: Now all ai ∈ U ∩ V , and U ∩ V is path connected. Hence we can
find a path r1 from x0 = f (a0 ) to f (a1 ) in U ∩ V .
Then we have two paths from the same start and endpoints in U or V , which
are both simply connected, so they’re homotopic.
Repeat this for all ai , and we find that f is homotopic to a loop r1 ∗r2 ∗· · ·∗rn
in U ∩ V . Hence it is homotopic to a loop strictly in U or V , both of which are
simply connected, so f is homotopic to the constant map x0 . Hence π1 (X, x0 ) =
0.
Theorem 23.2.2. For n ≥ 2, the n-sphere S n is simply connected. I.e.,
π1 (S n , x0 ) = 0.
Proof. Let p = (0, 0, . . . , 0, 1) ∈ Rn+1 be the north pole in S n and let q =
(0, 0, . . . , 0, −1) ∈ Rn+1 n
be the south pole innS .
n
Step 1: S \ p is homeomorphic to R . This is classic: place the sphere
on top of the origin on Rn and project from the north pole via a ray through
the sphere, onto the plane. In particular, the map is f : S n \ p → Rn by
1
f (x) = f (x1 , . . . , xn+1 ) := (x1 , x2 , . . . , xn )
1 − xn+1
for x = (x1 , . . . , xn+1 ). Since xn+1 6= 1 (we removed
the north pole), f is
continuous. The inverse map of f is g : Rn → S n \ p given by
g(y) = g(y1 , . . . , yn ) = (t(y)y1 , t(y)y2 , . . . , t(y)yn , 1 − t(y))
where
2
t(y) = .
1 + kyk2

Remark 23.2.3. Similarly, S n \ q is homeomorphic to S n \ p by the re-
flection map (x1 , . . . , xn , xn+1 ) 7→ (x1 , . . . , xn , −xn+1 ). Hence S n \ q is also
homeomorphic Rn .

Step 2: Let U = S n \ p and V = S n \ q . Then
Sn = U ∪ V
and
U ∩ V = S n \ p, q .

Note that π1 (U ) = π1 (V ) = π 1 (Rn ) = 0 (since the latter is convex). Also, U ∩V

is homeomorphic to Rn \ 0 which is path connected for n ≥ 2.
By Theorem 23.2.1, π1 (S n ) = 0 for n ≥ 2.

Corollary 23.2.4. For n ≥ 3, Rn \ 0 is simply connected.
Remark 23.2.5. Since π1 (S 1 ) = Z and π1 (S n ) = 0 for n ≥ 2, S 1 is not homeo-
morphic to S n for n ≥ 2.
Question: the fundamental group π1 distinguishes S 1 and S n , n ≥ 2. But
how to distinguish say S 2 and S 3 ?
23.3 nth homotopy group πn (X) 73
23.3 nth homotopy group πn (X)

Recall π1 (X) is the set of homotopy classes of loops f : [0, 1] → X with f (0) =
f (1). In other words, we can view f : S 1 → X.
Definition 23.3.1. We let πn (X) be the set of homotopy classes of f : S n → X.

We call πn (X) the nth homotopy group of X.
Theorem 23.3.2. We have

(
n Z, if n = m
πn (S ) =
0 if n < m.
We won’t explore higher homotopic theory here, so we omit the proof. But
it does tell us all S n are not homeomorphic.
It does raise a question though: what is πn (S m ) for n > m? In general, we
don’t know, there are only some partial results (mostly by J. P. Serre, who in
their PhD thesis studied this problem and won the Fields medal).
23.4 Fundamental group of the torus

We’d like to work out the fundamental group of the torus T = S 1 × S 1 . We get
immediately that it is Z × Z from the following:
Theorem 23.4.1. Let X and Y be topological spaces and let x0 ∈ X and

y0 ∈ Y . Then π1 (X × Y, (x0 , y0 )) ∼
= π1 (X, x0 ) × π1 (Y, y0 ).
Proof. Let p : X × Y → X and q : X × Y → Y be the natural projections. They
induce homomorphisms p∗ : π1 (X × Y, (x0 , y0 )) → π1 (X, x0 ) and q∗ : π1 (X ×
Y, (x0 , y0 )) → π1 (Y, y0 ).
Define Φ : π1 (X × Y, (x0 , y0 )) → π1 (X, x0 ) × π1 (Y, y0 ) by
Φ([γ]) = (p∗ ([γ]), q∗ ([γ])).
Then Φ is a group homomorphism (since componentwise it is a group homo-

morphism).
It remains to show Φ is one-to-one and onto.
First, onto: Let [α] ∈ π1 (X, x0 ) and [β] ∈ π1 (Y, y0 ). Define γ : [0, 1] → X ×Y
by γ(y) = (α(t), β(t)). Then γ is a loop based at (x0 , y0 ). Moreover
Φ([γ]) = (p∗ ([γ]), q∗ ([γ])) = ([p ◦ γ], [q ◦ γ]) = ([α], [β]),
so Φ is onto.
Second, one-to-one. Suppose [γ] ∈ ker Φ. Then Φ([γ]) = ([p ◦ γ], [q ◦ γ]) = 0,
so p◦γ ' x0 and q◦γ ' y0 . Let G and H be the corresponding path homotopies.
Define F : [0, 1] × [0, 1] → X × Y by
F (s, t) = (G(s, t), H(s, t)).

Then F is a path homotopy between γ and (x0 , y0 ). Hence ker Φ = 0 and Φ
is one-to-one.
Corollary 23.4.2. Let T = S 1 × S 1 , the torus. Then π1 (T ) = Z × Z.

74 RETRACTION
Lecture 24 Retraction
24.1 Fundamental group of projective space
Definition 24.1.1 (Real projective plane). The real projective plane RP 2 =
S 2 /∼ where x ∼ −x in S 2 . That is, antipodal points on the sphere are identified.
This is sort of hard to imagine, because the real projective plane RP 2 can
not be embedded into R3 .
Remark 24.1.2. Since p : S 2 → S 2 /∼ and S 2 is compact, RP 2 is compact.
The quotient map p is a covering map: for [x] ∈ RP 2 , take a small neigh-
bourhood U , small enough that the pullback around x and −x don’t meet.
Proposition 24.1.3. π1 (RP 2 , x0 ) ∼

= Z/2Z.
Proof. Since p : S 2 → RP 2 is a covering map, and π1 (S 2 ) = 0, this is a universal

covering. Hence Φ : π1 (RP 2 , x 0 ) → π −1 (x0 ) is one-to-one and onto.
Note p−1 (x0 ) = x0 , −x0 . Thus π1 (RP 2 ) is a group of order 2. But there’s
only one group of order 2, so π1 (RP 2 , x0 ) ∼ = Z/2Z.
Remark 24.1.4. Taking RP n = S n /∼ where x ∼ −x in S n is the real projective

n-space. Again p : S n → RP n is a covering map, and π1 (S n ) = 0 for n ≥ 2, so
it is a universal cover, so by the same argument π1 (RP n , x0 ) ∼
= Z/2Z.
For n = 1 this argument doesn’t work, since π1 (S 1 ) = Z 6= 0. In this case
p : S 1 → RP 1 is a homeomorphism, so π1 (RP 1 ) = Z.
Exercise 24.1. For n ≥ 1, define the real projective space of dimension n by
RP n = S n /∼, where the equivalence relation is defined by x ∼ y if and only if
x = y or x = −y.
Remark 24.1.5. In other words, RP n is obtained from S n by identifying pairs
of antipodal points. It can be regarded as the set of lines in Rn+1 which pass
through the origin.
(a) RP n is a compact Hausdorff space.
(b) The projection π : S n → RP n is a local homeomorphism, that is, each

x ∈ S n has an open neighbourhood that is mapped homeomorphically by
π onto an open neighbourhood of π(x).
∼ B n /S n−1 ∼
(c) RP 1 is homeomorphic to S 1 . (In fact, RP n = = S n for n =
1.)
24.2 Brouwer fixed-point theorem

Definition 24.2.1 (Retract). Let A ⊂ X. A retraction of X onto A is a
continuous map r : X → A such that rA = IdA . If such a map r exists, we call
A a retract of X.
Lemma 24.2.2. Suppose A is a retract of X. Let j : A ,→ X be the inclusion.

Then j∗ : π1 (A, a) → π1 (X, a) is one-to-one.
24.2 Brouwer fixed-point theorem 75
Proof. Suppose r : X → A is a retraction. Then r ◦ j = IdA . Hence (r ◦ j)∗ =

Idπ1 (A,a) = r∗ ◦ j∗ . Since the identity map is one-to-one (and onto), j∗ is one-
to-one.
In fact,
Exercise 24.2. Let r be a retraction of X onto A and let x0 ∈ A. Let i : A → X
be the inclusion map. Prove the following:
(a) i∗ : π1 (A, x0 ) → π1 (X, x0 ) is one-to-one.
(b) r∗ : π1 (X, x0 ) → π1 (A, x0 ) is onto.
x

Example 24.2.3. The map r : R2 \ 0 → S 1 given by r(x) = is a retrac-
kxk
tion. N

Theorem
24.2.4.
There is no retraction of the closed unit ball B 2 = x ∈
R kxk ≤ 1 onto S 1 .
2
Proof. Suppose there is a retraction r : B 2 → S 1 . Let j : S 1 ,→ B 2 be then

inclusion. Then r ◦ j : S 1 ,→ B 2 → S 1 is the identity map. Hence r∗ ◦ j∗ =
Id : π1 (S 1 ) ,→ π1 (B 2 ) → π1 (S 1 ). But π1 (S 2 ) = Z and π1 (B 2 ) = 0, so this is a
contradiction.
Lemma 24.2.5. Let h : S 1 → X be a continuous map. Then the following are
equivalent:
(i) h is nullhomotopic (i.e., h ' x0 , a constant map).
(ii) h extends to a continuous map k : B 2 → X.
(iii) h∗ : π1 (S 1 ) → π1 (X) is the trivial map (i.e., h∗ = 0 sends everything to
the identity).
Proof. First, (i) implies (ii). Assume h ' x0 , and let H : S 1 × [0, 1] → X be a
homotopy between h and x0 . In other words, H(s, 0) = h(s) for all s ∈ S 1 and
H(s, 1) = x0 for all s ∈ S 1 .
Now imagine projecting S 1 × [0, 1] to B 2 . So define π : S 1 × [0, 1] → B 2 by
π(s, t) = (1 − t)s. (So identify the top circle of the cylindrical shell S 1 × [0, 1]
as the centre of B 2 ).
Notice that π is continuous and onto, and that π(s, 0) = s for all s ∈ S 1 ,
and π(s, 1) = 0.
Now define k : B 2 → X by k(b) = H(b0 ) for any b0 ∈ π −1 (b). Aside from the
origin, π −1 (b) is only one point, so no trouble there, and for the origin b = 0, the
pullback is a circle, but that circle is homotopic to x0 so it remains well-defined.
Since π(s, 0) = s, k extends h. (Try drawing the graph of this situation—it
makes it clear.)
Next (ii) implies (iii): Assume h extends to k : B 2 → X. Let j : S 1 ,→ B 2
be the inclusion. Then
j k
S1 B2 X.
Then
h∗ = k∗ ◦ j∗ : π1 (S 1 ) → π1 (B 2 ) → π1 (X),
76 BROUWER FIXED-POINT THEOREM
but π1 (B 2 ) = 0 and k∗ is a homomorphism, so 0 must map to 0. Thus h∗ = 0.

Finally, (iii) implies (i). Assume h∗ = 0. Note h∗ : π1 (S 1 ) = Z → π1 (X).
Let p0 : [0, 1] → S 1 , p0 (r) = e2πir . Then [p0 ] generates π1 (S 1 ) (it is the single
anticlockwise loop, corresponding to 1 ∈ Z).
Fix a point s0 ∈ S 1 and let x0 = h(s0 ).
Since h∗ = 0, f := h ◦ p0 represents the identity in π1 (X, x0 ), since h∗ (p0 ) =
[h ◦ p0 ] by definition.
Hence the image of f is homotopic to x0 , so h is homotopic to x0 . In other
words there is a path homotopy F in X between f and x0 , i.e., F : [0, 1]×[0, 1] →
X with F (s, 0) = x0 and F (s, 1) = f (s).
Note that since F is a path homotopy, F (0, t) = F (1, t) = x0 for all t. Hence
F decomposes as
p0 ×I G
F : [0, 1] × [0, 1] S 2 × [0, 1] X
where G is a homotopy between x0 and h.

Corollary 24.2.6. (i) The inclusion j : S 1 ,→ R2 \ 0 is not nullhomo-
topic.
(ii) The identity map i : S 1 → S 1 is not nullhomotopic.

x

Proof. (i) Define the retraction r : R2 \ 0 → S 1 by r(x) = kxk . Then
j r
S1 R2 \ 0 S1.
Id
So r∗ ◦ j∗ = Idπ1 (S 1 ) . Thus j∗ is one-to-one. If j is nullhomotopic, then by

Lemma 24.2.5(i), j∗ = 0, which is a contradiction. Hence j is not nullhomotoic.
(ii) More or less the same argument, i : S 1 → S 1 means i∗ = Idπ1 (S 1 ) so i is

not nullhomotopic.
Theorem 24.2.7. For any nonvanishing continuous vector field on B 2 , there

exists a point on S 1 where the vector field points directly inward and a point on
S 1 where the vector field points directly outward.
Remark 24.2.8. A continuous vector field on B 2 is a pair (x, v(x)) where

x ∈ B 2 and v(x) ∈ R2 , with v(x) continuous.
That the vector field is nonvanishing means v(x) 6= 0 for all x.
This theorem is equivalent to the Brouwer fixed-point theorem.
Lecture 25 Brouwer fixed-point theorem

25.1 Brouwer fixed-point theorem
Proof of Theorem 24.2.7. Suppose v(x) does not point directly inward for any
x ∈ S 1 . Since the vector field is nonvanishing, v : B 2 → R2 \ 0 . Let w = v S 1 .
25.2 Application of the Brouwer fixed-point theorem 77
Then by construction v is an extension of w to B 2 , so by Lemma 24.2.5, w is

nullhomotopic.
On the other hand, let j : S 1 ,→ R2 \ 0 be the inclusion map. We claim w
is homotopic to j. Then w ' 0 and w ' j, so 0 ' j, but this is a contradiction
since j 6' 0.
Let F : S 1 × [0, 1] → R2 \ 0 ,
F (x, t) = tx + (1 − t)w(x),
so the sort of convex, straight-line homotopy between w(x) and j(x) = x (think-
ing of j(x) as the outward unit normal of S 1 ) since if t = 0, F (x, 0) = w(x). If
t = 1, F (x, 1) = x = j(x). To make sure this is actually a homotopy, we must
make sure F (x, t) 6= 0 for all x and t.
Suppose F (x, t) = 0 for some 0 < t < 1. Then tx + (1 − t)w(x) = 0, or
t
w(x) = − x.
1−t
But x is radius of the unit circle, so this w(x) points directly inward at x, which
is a contradiction.
For w(x) pointing directly outward at some point x ∈ S 1 , we can consider
the vector field (x, −v(x)), where the − switches directly inward for directly
outward.
Theorem 25.1.1 (Brouwer fixed-point theorem). Let f : B 2 → B 2 be a con-

tinuous function. Then there exists a point x ∈ B 2 such that f (x) = x (i.e., a
fixed-point).
Proof. Suppose f (x) 6= x for all x ∈ B 2 . Let v(x) := f (x) − x. Then (x, v(x))
is a nonvanishing continuous vector field on B 2 .
By Theorem 24.2.7 there exists an x ∈ S 1 such that v(x) = f (x) − x = kx
points directly outward, so k > 0.
Hence f (x) = (k + 1)x ∈ B 2 , and f : B 2 → B 2 , but
kf (x)k = (k + 1)kxk = k + 1 > 1
which is a contradiction.
Remark 25.1.2. In general, every continuous map f : B n+1 → B n+1 , n ≥ 1, has

a fixed point, i.e., there exists x ∈ B n+1 such that f (x) = x. This is the general
form of the Brouwer fixed-point theorem. (It’s also true for B 1 → B 1 , but here
it’s much easier: it’s just the intermediate value theorem.)
This is a consequence of there in general being no retraction r : B n+1 → S n
for n ≥ 2. But it takes more work to show this, requiring higher homotopy, or
homology.
But once we have such a result, the proof of the n = 1 case can be generalised.
25.2 Application of the Brouwer fixed-point theorem

Corollary 25.2.1. Let A ∈ M3 (R) with positive entries. Then A has a positive
real eigenvalue.
78 BROUWER FIXED-POINT THEOREM
Proof. Define T : R3 → R3 be the corresponding linear map to A, T (v) = Av.

Let B = S 2 ∩ (x1 , x2 , x3 ) x1 , x2 , x3 ≥ 0 . (So the part of the unit sphere in
the first orthant.)
Exercise 25.1. B is homeomorphic to B 2 .
Hence the Brouwer fixed-point theorem holds for any continuous map f : B →
B.
Consider in particular f : B → B defined by f (x) = kTT (x)
(x)k ∈ B (since both
x and A have all positive entries).
Hence there exists some x0 ∈ B such that f (x0 ) = x0 . In other words,
T (x0 ) = kT (x0 )kx0 ,
so kT (x0 )k > 0 is an eigenvalue of T , so of A, since x0 6= 0.
25.3 The Borsuk–Ulam theorem

Definition 25.3.1. The antipode of x ∈ S n is the point −x ∈ S n .
A map h : S n → S n is antipode-preserving if h(−x) = −h(x) for all
x ∈ Sn.
Theorem 25.3.2. Suppose h : S 1 → S 1 is a continuous and antipode-preserving
map. Then h is not nullhomotopic.
Proof. Let S0 = (1, 0) ∈ S 1 . Let ρ : S 1 → S 1 be a rotation such that ρ(h(s0 )) =
s0 .
Note that ρ is antipode-preserving. Then
ρ ◦ h(−x) = ρ(h(−x)) = ρ(−h(x)) = −ρ(h(x)).
Hence the composition ρ ◦ h is also antipode-preserving.

If H is a homotopy between h and a constant map (so if h is nullhomotopic),
then ρ ◦ H is a homotopy between ρ ◦ h and a constant map.
So we may assume h(s0 ) = s0 (else rotate via ρ as described).
Step 1: Define q : S 1 → S 1 by q(z) = z 2 , thinking of z ∈ C, i.e., double the
angle of z thought of in polar representation on S 1 .
Note that q is a quotient
map, i.e., q is continuous, onto, and open. Note
also that q −1 (a) = z, −z , where q(z) = z 2 = a.
Second, q(h(−z)) = q(−h(z)) = q(h(z)). Consider the diagram
h
S1 S1
q q
S1 k
S1.
where we define k : S 1 → S 1 such that k(a) = g(h(z)) for z 2 = a; since h is

antipode-preserving, this is well-defined. Then
k ◦ q = q ◦ h,
so the diagram commute.

Note s0 = 1 ∈ Z, then q(s0 ) = s0 and h(−s0 ) = −h(s0 ) = −s0 ). Hence
k(s0 ) = s0 .
THE BORSUK–ULAM THEOREM 79
Step 2: we claim the induced map h∗ : π1 (S 1 , s0 ) → π1 (S 1 , s0 ) is nontrivial.

Note that q is a covering map. Let f˜ be any path in S 1 from s0 to −s1 .
Then f := q ◦ f˜ is a loop based at s0 . Compute
k∗ ([f ]) = [k ◦ f ] = [k ◦ q ◦ f˜] = [q ◦ h ◦ f˜] 6= 0
since (h ◦ f˜) is a path from s0 to −s0 too, so mapped through q it’s a loop
covering the whole circle. Thus k∗ is nontrivial.
Step 3: We claim h∗ is nontrivial. Hence by Lemma 24.2.5, h 6' 0.
We have k∗ : π1 (S 1 ) = Z → π1 (S 1 ) = Z. The kernel ker(k∗ ) is a subgroup
of (Z, +). If the kernel is nonzero, ker(k∗ ) = mZ since Z is cyclic. Thus
Z/mZ ∼ = Im(k∗ ). But Im(k∗) is a subgroup of Z, and Z/mZ is finite, but the
only finite subgroup of Z is 0 , contradicting ker(k∗ ) 6= 0.
Hence ker(k∗ ) = 0, so k∗ is one-to-one.
Note q∗ is also one-to-one; q∗ : π1 (S 1 ) → π1 (S 1 ) is the map m 7→ 2m.
Thus
h∗
π1 (S 1 ) π1 (S 1 )
q∗ q∗
h∗
π1 (S 1 ) π1 (S 1 )
and since every step is one-to-one, h∗ cannot be trivial.
Lecture 26 The Borsuk–Ulam theorem

26.1 The Borsuk–Ulam theorem
Theorem 26.1.1. There is no continuous antipode-preserving map g : S 2 → S 1 .
Proof. Suppose there exists such a g : S 2 → S 1 which is continuous and antipode-

preserving.
Take S 1 to be the equator of S 2 . Then h := g S 1 : S 1 → S 1 is continuous
and antipode-preserving. By Theorem 25.3.2, h is not nullhomotopic. But h
has an extension g the upper hemisphere E of S 2 .
Moreover, E ' B 2 , so by Lemma 24.2.5, h is nullhomotopic, which is a
contradiction.
Theorem 26.1.2 (Borsuk–Ulam theorem). For any continuous map f : S 2 →

R2 , there exists x ∈ S 2 such that f (x) = f (−x).
Proof. Suppose f (x) 6= f (−x) for all x ∈ S 2 . Define g : S 2 → S 1 by
f (x) − f (−x)
g(x) = ,
kf (x) − f (−x)k
which is defined since the denominator is never zero. Then g is continuous, and
g(−x) = −g(x) for all x ∈ S 2 .
Hence g : S 2 → S 1 is continuous and antipode-preserving, contradiction The-
orem 26.1.1.
Date: December 1st, 2020.
80 THE BORSUK–ULAM THEOREM
Remark 26.1.3. In general, for any continuous map f : S n+1 → Rn+1 , n ≥ 0,

there exists x ∈ S n+1 such that f (x) = f (−x). That is, Borsuk–Ulam theorem
is true in higher dimension.
The proof in general is similar, but it is much more difficult to show that
if h : S n+1 → S n+1 is continuous and antipode-preserving, then h is not null-
homotopic (that is, generalising Theorem 25.3.2). This requires either higher
homotopy theory or homology.
Given this, the proof of Borsuk–Ulam theorem for n = 1 generalises to n ≥ 2.
26.2 Applications
A question: Given a bounded region A ⊂ R2 , can you find a line L that bisects
A?
The answer is yes: consider horizontal slices of the set at y = c, and let f (c)
denote the area of the part of A below y = c.
Then f (c) is a continuous function that varies from 0 to the area of A, so by
the intermediate value theorem, it must attain half the area of A for some c.
Now a trickier question: Given two bounded regions A and B in R2 , can you
find a line L that bisects both A and B simultaneously?
The answer is still yes, but it is quite a bit less trivial.
Theorem 26.2.1 (Ham-sandwich theorem). Given any two bounded regions

A, B ⊂ R2 , there exists a line L in R2 that bisects both A and B.

Proof. Take any two bounded regions A and B in the plane R2 × 1 , and treat
this as though it is in R3 .
Given u ∈ S 2 (i.e., corresponding to a unit vector in R3 ), let p = p(u) be
the plane in R3 passing through
the origin with unit normal vector u.
Then the plane R2 × 1 and p(u) intersect in a line (unless u is vertical).
Hence let L = p ∩ (R2 × 1 ) be this line of intersection.
Let f1 (u) denote the area of the part of A that lies on side of L (say, on the
same side as u). Similarly, let f2 (u) be the area of the part of B that lies on
the same side as u.
Note that if we replace u by −u, we get the same plane, only with the
opposite normal. Hence fi (−u) is the area of the other side of A or B.
Hence f1 (u) + f1 (−u) is the area of A, and f2 (u) + f2 (−u) is the area of B.
Now consider a map F : S 2 → R2 given by F (u) = (f1 (u), f2 (u)), which
is continuous By Borsuk–Ulam theorem, there exists some u ∈ S 2 such that
F (u) = F (−u). Hence f1 (u) = f1 (−u), half the area of A, and f2 (u) = f2 (−u),
of B, meaning we have bisected A and B with the line L =
half the area
p ∩ (R2 × 1 ).
26.3 Invariance of domain

This is a theorem of Brouwer.
Theorem 26.3.1 (Invariance of domain). Let U ⊂ Rn be an open subset. Let

f : U → Rn be continuous and one-to-one. The f (U ) is open in Rn .
We will prove this for n = 2.

INVARIANCE OF DOMAIN 81
Lemma 26.3.2 (Homotopy extension lemma). Let X be a space such that

X × [0, 1] is normal. Let A ⊂ X be a closed subset. Let Y ⊂ Rn .
Let f : A → Y be continuous. Suppose f is null-homotopic. Then f can be
extended to a continuous map g : X → Y which is also nullhomotopic.
Proof. Let F : A × [0, 1] → Y be a homotopy between f and a constant map y0 .
∈ A, and f (a, 1) = y0 for all a ∈ A.
Hence F (a, 0) = f (a) for all a
First extend F to X × 1 by setting F (x, 1) = y0 for all x ∈ X. So F is
continuous on (A × [0, 1]) ∪ (X × 1 ).
Note that this is a closed subset of X × [0, 1], since A is closed, and this
space is normal.
Use Tietze extension theorem to show that F can be extended to a continous
map G : X × [0, 1] → Rn .
Exercise 26.1. Actually do this; note that our version of Tietze extension theo-
rem only maps to R.
Now G(x, 0) is an extension of f (x), but it maps X to Rn , not necessarily
to Y ⊂ Rn . We need to fix this.
Let U = G−1 (Y ) ⊂ X × [0, 1]. Then (A × [0, 1]) ∪ (X × 1 ) ⊂ U .

Since [0, 1] is compact, the tubular neighbourhood theorem implies there

is an open set W ⊂ X containing A such that W × [0, 1] ⊂ U .
Thus G(W × [0, 1]) ⊂ Y . Note X ' X × 0 ⊂ X × [0, 1] is a closed subset
of an normal set, so X is normal.
Hence by Urysohn’s lemma there exists a continuous φ : X → [0, 1] such that
φ = 0 on A and φ = 1 on X \ W .
Then x 7→ (x, φ(x)) ∈ (W × [0, 1]) ∪ (X × 1 ).
Thus defining g(x) := G(x, φ(x)) we have g : X → Y .
For x ∈ A, φ(x) = 0, so g(x) = G(x, 0) = f (x), meaning g is the desired
extension of f .
Finally, we need to show g ' 0. Define H : X × [0, 1] → Y by H(x, t) =
G(x, (1 − t)φ(x) + t). Then H(x, 0) = G(x, φ(x)) = g(x), and H(x, 1) =
G(x, 1) = F (x, 1) = y0 .
Lecture 27 Invariance of domain

27.1 Invariance of domain
Lemma 27.1.1 (Borsuk
lemma). Let a, b ∈ S 2 . Let A be a compact space. Let
2
f : A → S \ a, b be a continuous injective map.
Suppose f is nullhomotopic. Then a and b lie in the same component of
S 2 \ f (A).
Proof. Note that since A is compact and S 2 is Hausdorff, because f is one-to-
one, A is homeomorphic to f (A) (this is Theorem8.1.1).

So we may assume (or think of)
f : A ,→ S 2 \ a, b is the inclusion map.
Note also that S 2 ' R2 ∪ ∞ , so let us say a = 0 and b = ∞.
2
Hence we reduce to the following: Let
A be a compact subset of R \ 0 .
2
Suppose the inclusion j : A ,→ R \ 0 is nullhomotopic. Then 0 lies in the
unbounded component of R2 \ A.
Date: December 3rd, 2020.
82 INVARIANCE OF DOMAIN
To this end, let C be the connected component of R2 \ A containing 0. We

need to show that C is unbounded.
Let D be the union of the other components of R2 \ A. Then R2 \ A = C t D.
Notice the left-hand side is open (since A is closed), and R2 is locally path
connected, meaning that both C and D are open.
Suppose C is bounded. We assume j ' x0 , Lemma 26.3.2, j can be
so by
extended to a continuous map k : C ∪ A → R2 \ 0 .

Next we extend k to h : R2 → R2 \ 0 by h(x) = x for all x ∈ D ∪ A. So
h(x) = x for all x ∈ R2 \ C.
Let B = x ∈ R2 kxk ≤ M , with M large enough that C ∪ A ⊂ int(B)
(which is doable since C is bounded and A is compact). Let g = hB : B →

R2 \ 0 . Then g(x) = x on ∂B.
g(x)
Thus we have a retraction ι : B → ∂B, x 7→ M kg(x)k . Note in particular
that ι(x) = x for x ∈ ∂B.
This is a contradiction, because there exists no retraction of B 2 to S 1 (this
is Theorem 24.2.4; in fairness our argument is scaled up, but it’s the same
argument).
Now we are ready to prove the n = 2 case of Invariance of domain:
Theorem 27.1.2 (Invariance of domain). Let U ⊂ R2 be an open subset.

Let f : U → R2 be continuous and one-to-one. The f (U ) is open in R2 and
f −1 : f (U ) → U is continuous. (So f is an open map.)
Proof. First we can replace R2 by S 2 ∼

= R2 ∪ ∞ . We need to show that, for
U ⊂ R2 open, if f : U → S 2 is continuous and one-to-one, then f (U ) ⊂ S 2 is
open and f −1 is continuous.
Step 1: For any closed ball B ⊂ U ⊂ R2 , f (B) does not separate S 2 . I.e.,
2
S \ f (B) only has one connected component.

Let a, b ∈ S 2 \ f (B). Let h = f B : B → S 2 . Then h is nullhomotopic
(consider F (x, y) = h(tx), x ∈ B, t ∈ [0, 1]).
The Borsuk lemma then says a and b lie in the same component of S 2 \h(B) =
2
S \ f (B).
Step 2: For any closed ball B ⊂ U ⊂ R2 , the image f (int(B)) is open in S 2 .
To prove this we need the following theorem:
Theorem 27.1.3 (Jordan separation theorem). Let C be a simple closed curve

in S 2 . Then C separates S 2 . (In general, S n separates S n+1 .)
Note that C := f (∂B) is a simple closed curve in S 2 , so C separates S 2 .

Let V be the connected component of S 2 \ C that contains f (int(B)) (which
is connected since int(B) is connected, and f is continuous).
Let W be the union of the other components of S 2 \ C.
Since C is closed and S 2 is locally path connected, V and W are both open
in S 2 .
We will show f (int(B)) = V (and hence open).
Suppose there exists some a ∈ V \ f (int(B)) and take any b ∈ W .
By step 1, f (B) does not separate S 2 , so a and b lie in the same component
of S 2 \ f (B) ⊂ S 2 \ C. Hence a and b lie in the same component of S 2 \ C,
which contradicts a ∈ V and b ∈ W .
27.1 Invariance of domain 83
Now for any ball B ⊂ U , f (int(B)) is open in S 2 . Hence f is an open map

(on a basis, and so the whole space), and we are done.
Remark 27.1.4. There are two difficulties in generalising this for n ≥ 3. First,
there is no retraction of B n to S n−1 (for which, as should now appear natural,
we need higher homotopy or homology).
Second, we need the more general form of the Jordan separation theorem,
saying that S n−1 separates S n .
Given these two, the proof of Invariance of domain generalises nicely.
84 INDEX
Index
A norm . . . . . . . . . . . . . . . . . . . . . . 15
antipode . . . . . . . . . . . . . . . . . . . . . . . 78 even cover . . . . . . . . . . . . . . . . . . . . . 64
antipode-preserving . . . . . . . . . . . . 78
F
B fibre . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Baire space . . . . . . . . . . . . . . . . . . . . 50 first-countable . . . . . . . . . . . . . . . . . 34
ball . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 fixed-point . . . . . . . . . . . . . . . . . . . . . 77
Banach fixed-point theorem . . . . 32 fundamental group . . . . . . . . . . . . . 63
basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Bolzano–Weierstrass theorem . . 28, G
30 generated topology . . . . . . . . . . . . . . 5
bounded set . . . . . . . . . . . . . . . . . . . . 14
box topology . . . . . . . . . . . . . . . . . . . 12 H
Hausdorff space . . . . . . . . . . . . . . . . . 7
C homeomorphic spaces . . . . . . . . . . 10
Cauchy sequence . . . . . . . . . . . . . . . 49 homeomorphism . . . . . . . . . . . . . . . 10
closed set . . . . . . . . . . . . . . . . . . . . . . . 2 homogeneous space . . . . . . . . . . . . 59
closure . . . . . . . . . . . . . . . . . . . . . . . . . . 6 homotopic . . . . . . . . . . . . . . . . . . . . . 61
cofinite topology . . . . . . . . . . . . . . . . 2 homotopy equivalence . . . . . . . . . . 69
compact . . . . . . . . . . . . . . . . 17, 22, 28 homotopy equivalent . . . . . . . . . . . 69
compactification . . . . . . . . . . . . . . . 44 homotopy group . . . . . . . . . . . . . . . 73
induced . . . . . . . . . . . . . . . . . . . 45 homotopy inverse . . . . . . . . . . . . . . 69
complete space . . . . . . . . . . . . . . . . . 49 homotopy type . . . . . . . . . . . . . . . . . 69
completely regular . . . . . . . . . . . . . 44
connected I
component . . . . . . . . . . . . . . . . 20 indiscrete topology . . . . . . . . . . . . . . 1
space . . . . . . . . . . . . . . . . . . . . . . 18 interior . . . . . . . . . . . . . . . . . . . . . . . . . .6
subspace . . . . . . . . . . . . . . . . . . 19 isolated point . . . . . . . . . . . . . . . . . . . 6
continuity . . . . . . . . . . . . . . . . . . . . . . . 1
continuous function . . . . . . . . . . . . . 9 L
at a point . . . . . . . . . . . . . . . . . . 9 Lebesgue number lemma . . . . . . . 30
contractible . . . . . . . . . . . . . . . . . . . . 70 lift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
convergence . . . . . . . . . . . . . . . . . . . . . 6 limit point . . . . . . . . . . . . . . . . . . .6, 28
convex space . . . . . . . . . . . . . . . . . . . 62 limit point compact . . . . . . . . . . . . 28
coordinate functions . . . . . . . . . . . 11 locally compact . . . . . . . . . . . . . . . . 33
covering map . . . . . . . . . . . . . . . . . . 64 at a point . . . . . . . . . . . . . . . . . 33
covering space . . . . . . . . . . . . . . . . . 64 locally finite . . . . . . . . . . . . . . . . . . . 10
locally path-connected . . . . . . . . . 22
D loop . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
dense . . . . . . . . . . . . . . . . . . . . . . . . . . 35 lower limit topology . . . . . . . . . . . . . 4
diameter . . . . . . . . . . . . . . . . . . . . . . . 30
disconnected space . . . . . . . . . . . . . 18 M
discrete topology . . . . . . . . . . . . . . . . 1 maximal element . . . . . . . . . . . . . . . 26
dyadic rational number . . . . . . . . .38 metric . . . . . . . . . . . . . . . . . . . . . . . . . 13
metric space . . . . . . . . . . . . . . . . . . . 14
E metric topology . . . . . . . . . . . . . . . . 13
equivalent metrisability . . . . . . . . . . . . . . . . . . . 10
compactification . . . . . . . . . . . 45 metrisable . . . . . . . . . . . . . . . . . . . . . 14
INDEX 85
N T
neighbourhood . . . . . . . . . . . . . . . . . . 1 T2 -space . . . . . . . . . . . . . . . . . . . . . . . 35
norm . . . . . . . . . . . . . . . . . . . . . . . . . . .14 T2 -space . . . . . . . . . . . . . . . . . . . . . 7, 36
normal space . . . . . . . . . . . . . . . . . . . 36 T3 -space . . . . . . . . . . . . . . . . . . . . . . . 36
normed topology . . . . . . . . . . . . . . . 15 T3.5 -space . . . . . . . . . . . . . . . . . . . . . . 44
nullhomotopic . . . . . . . . . . . . . . . . . . 75 T4 -space . . . . . . . . . . . . . . . . . . . . . . . 36
topological group . . . . . . . . . . . . . . 57
O topological property . . . . . . . . 10, 14
one-point compactification . . . . . 33 topological space . . . . . . . . . . . . . . . . 1
open map . . . . . . . . . . . . . . . . . . . . . . 24 topologist’s sine curve . . . . . . . . . . 22
open set . . . . . . . . . . . . . . . . . . . . . . . . .1 topology . . . . . . . . . . . . . . . . . . . . . . . . 1
orbit . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 finer . . . . . . . . . . . . . . . . . . . . . . . . 3
orbit space . . . . . . . . . . . . . . . . . . . . . 59 strictly finer . . . . . . . . . . . . . . . . 3
totally bounded . . . . . . . . . . . . 31, 49
P totally disconnected . . . . . . . . 12, 19
partially ordered set . . . . . . . . . . . .25 totally ordered set . . . . . . . . . . . . . 26
path . . . . . . . . . . . . . . . . . . . . . . . 21, 61 transitive action . . . . . . . . . . . . . . . .59
path homotopic . . . . . . . . . . . . . . . . 61 triangle inequality . . . . . . . . . . 13, 14
path-connected . . . . . . . . . . . . . . . . .21 trivial topology . . . . . . . . . . . . . . . . . 1
product topology . . . . . . . . . . . . 7, 12 tubular neighbourhood theorem 81
Q U
quotient space . . . . . . . . . . . . . . . . . 55 uniform convergence . . . . . . . . . . . 18
quotient topology . . . . . . . . . . . . . . 55 uniformly continuous . . . . . . . . . . . 32
universal covering space . . . . . . . . 67
R upper bound . . . . . . . . . . . . . . . . . . . 26
real projective plane . . . . . . . . . . . 74
real projective space . . . . . . . . . . . 74 V
regular space . . . . . . . . . . . . . . . . . . . 36 vector field
relative topology continuous . . . . . . . . . . . . . . . . .76
see subspace topology . . . . . . .5 nonvanishing . . . . . . . . . . . . . . 76
retract . . . . . . . . . . . . . . . . . . . . . . . . . 74
retraction . . . . . . . . . . . . . . . . . . . . . . 74
Riemann sphere . . . . . . . . . . . . . . . . 34
S
second-countable . . . . . . . . . . . . . . . 34
separable . . . . . . . . . . . . . . . . . . . . . . 35
sequentially compact . . . . . . . . . . . 28
simply connected . . . . . . . . . . . . . . .63
semilocally . . . . . . . . . . . . . . . . 67
special orthogonal group . . . . . . . 60
stabiliser . . . . . . . . . . . . . . . . . . . . . . . 60
standard bounded metric . . . . . . . 42
standard topology . . . . . . . . . . . . . . . 4
Stone–Čech compactification 46, 48
subbasis . . . . . . . . . . . . . . . . . . . . . . . . . 5
subspace . . . . . . . . . . . . . . . . . . . . . . . . 5
subspace topology . . . . . . . . . . . . . . . 5
sup-norm . . . . . . . . . . . . . . . . . . . . . . 16

5 Lecture Notes in General Topology

Uploaded by

5 Lecture Notes in General Topology

Uploaded by

Lecture Notes in General Topology

Lectures by Dr Sheng-Chi Liu

Throughout these notes, signifies end proof, N signifies end of exam-

Lecture 1 Laying the groundwork 1

Lecture 3 Many basic notions 5

Lecture 5 Infinite product spaces 12

Lecture 6 Connectedness and compactness 16

Lecture 7 Different kinds of connectedness 20

Lecture 8 Tychonoff’s theorem 24

Lecture 9 Tychonoff’s theorem 27

Lecture 11 Countability and separation axioms 34

Lecture 12 Urysohn’s lemma 38

Lecture 13 Urysohn’s metrisation theorem 41

Lecture 14 Completely regular space and Stone–Čech compact-

Lecture 16 Baire category theorem 50

Lecture 17 Quotient spaces and topological groups 55

Lecture 18 Topological groups 58

Lecture 19 Homotopy theory 61

Lecture 20 Covering spaces 63

Lecture 21 Universal covering space 66

Lecture 22 Homotopy type 68

Lecture 23 Fundamental group calculations 71

Lecture 25 Brouwer fixed-point theorem 76

Lecture 26 The Borsuk–Ulam theorem 79

Lecture 27 Invariance of domain 81

Lecture 1 Laying the groundwork

1.1 Topological space

(i) ∅ and X are in T ,

(ii) any union of sets in T is in T , and

(iii) any finite intersection of sets in T is in T .

Definition 1.1.2 (Topological space, open set, (open) neighbourhood). A topo-

Some trivial examples.

Example 1.1.3 (Trivial topology). If T = ∅, X , then T is a topology on

X called the trivial topology or indiscrete topology.

Our first nontrivial example:

but there are more.

Definition 2.1.1 (Basis). Let X be a topological space. A collection B of

In other words, the basis generalises the set of open balls in R2 :

Theorem 2.1.2. A collection B of open subsets of a topological space X is a

where Vα ∈ B by the definition of basis. Hence, since x ∈ U , we must have

Theorem 2.1.3. A collection B of subsets of X is a basis for the topology on

(i) Each x ∈ X lies in at least one set in B.

(ii) If U, V ∈ B, and x ∈ U ∩ V , then there exists W ∈ B such that x ∈ W

so the first two conditions for a topology are satisfied.

so by definition U ∩ V ∈ T , being the union of sets in B.

(i) T 0 is finer than T (i.e., T ⊂ T 0 ),

(ii) For each x ∈ X and U ∈ B containing x, there exists U 0 ∈ B 0 such that

Example 2.1.4. Let X = R. Let B be the collection of all open intervals

Example 2.1.5. Let X = R and B 0 be the collection of all half-open intervals

Note that T 0 is strictly finer than T . I.e., T 0 ) T .

Lecture 3 Many basic notions

is a topology on Y called the subspace topology or relative topology.

is a basis for the subspace topology on Y .

3.2 Interior and closure

3.3 Hausdorff space

an arbitrary union of open sets, so X \ E is open, whence E is closed.

3.4 Finite product spaces

(x1 , x2 , . . . , xn ) ∈ (U1 × U2 × · · · × Un ) ∩ (V1 × V2 × · · · × Vn ),

then the point is also in (U1 ∩ V1 ) × · · · × (Un ∩ Vn ). Each of the constituent

Proposition 3.4.2. Let Bi be a basis for the topology on Xi , i = 1, 2, . . . , n.

is a basis for the product topology on X1 × X2 × · · · × Xn .

Proof. We apply Theorem 2.1.2. Let x = (x1 , x2 , . . . , xn ) ∈ X1 × X2 × · · · × Xn .

where V1 × V2 × · · · × Vn ∈ C . Hence by Theorem 2.1.2, C is a basis for the

Exercise 3.7. Let X and Y be topological spaces. Define the projections π1 : X ×

B = π1−1 (U ) U open in X ∪ π2−1 (V ) V open in Y

Definition 4.1.1 (Continuous function). Let X and Y be topological spaces.

a union of open sets, hence open. In addition, x ∈ f −1 (V ), x ∈ Ux ⊂ U ,

Definition 4.2.2 (Homeomorphic spaces). The topological spaces X and Y

Classifying objects and/or spaces as homeomorphic or not is an important

Definition 4.2.3 (Topological property). A property of a topological space is

Example 4.2.4. Metrisability (being able to assign a metric) is a topological

Example 4.2.5. Let f : (−1, 1) → R be defined by

This map is a homeomorphism, showing that (−1, 1) is homeomorphic to R. N

4.3 Constructing continuous functions

Proof. (i) Let U ⊂ Y be open. We have

is open in Z (using the subspace topology) since f 1 (U ) is open in X. Hence