M2201

Metric Topology

Lecture notes of Prof. Hicham Gebran

[email protected]

Lebanese University, Fanar, Fall 2024-2025

https://hichamgebran.wordpress.com
2

Introduction and orientation

• In elementary school geometry, you studied regular shapes like straight lines, triangles, circles,
polygons and planes. At some point you considered transformations that conserve these shapes
like central and axial symmetries, rotations, translations and homotheties. What if we allow more
general shapes and more general elastic transformations that can distort the geometric objects (like
twisting, bending, squeezing and stretching) but do not break them? Such transformations are said
to be continuous.
• Also, instead of working with geometric objects in the plane or space, we can work with objects
in n-dimensional space Rn . We can also work in the more general context of metric spaces and
topological spaces. Topological spaces are a generalization of metric spaces which are themselves
a generalization of n-dimensional space. We can say that topology is the study of continuity in a
general context. Topology is sometimes called ”rubber-sheet geometry”. In this course, we shall
focus on metric spaces leaving the study of topological spaces to the course M3315.
• In group theory or vector space theory, a fundamental concept is that of isomorphism. Two
groups are isomorphic if they have the same algebraic structure; in this case, we can identify them.
The analogous concept in topology is that of homeomorphism. A homeomorphism is a continuous
bijection whose inverse is also continuous. For example, a circle, an ellipse, a square and a triangle
are all homeomorphic, that is, they are topologically the same. However, a line segment and a circle
are not homeomorphic. A fundamental problem of topology is to distinguish between two given
topological spaces, that is, to decide whether they are homeomorphic or not. Another fundamental
problem is the classification of certain important spaces called manifolds up to a homeomorphism.
Manifolds are a generalization of curves and surfaces in higher dimensions. We can also say that
topology studies those properties of a space (or shape) that persist when the space undergoes a
homeomorphism (a bi-continuous deformation).
• Topology leads to existence and non-existence theorems that are important in other fields of
Mathematics like Analysis and Geometry. Certain mathematical objects may exist or not for a
topological reason. We will illustrate this during and by the end of the course.
• Topology allows one to go from the local to the global. We will explain that by the end of the
course.
• Topology has three big subfields.

▶ General Topology or point-set topology (this course and M3315).

▶ Algebraic topology (Math 413).

▶ Differential topology.

All questions, comments, remarks and suggestions are welcome.

Some references

James Munkres, Topology (Pearson, 2000).

Gustave Choquet, Topology (Academic Press, 1966).
Colin Adams and Robert Franzosa, Introduction to topology, pure and applied (Pearson,
2007).
Lynn Arthur Steen and J. Arthur Seebach, Counterexamples in Topology (Dover, 1978).
Contents

0 Preliminaries 5
0.1 Some fundamental properties of the real line . . . . . . . . . . . . . . . . . . . . . . 5
0.2 Sets and functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1 Metric spaces and topological spaces 11

1.1 Definitions and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2 Open sets and related concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3 Topology and topologically equivalent distances . . . . . . . . . . . . . . . . . . . . 24
1.4 Accumulation points and subsequences . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.5 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.6 Product spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2 Continuous functions and homeomorphisms 31

2.1 Continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2 Homeomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Compactness 39
3.1 Definitions, examples and basic properties . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Other forms of compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Compact subsets of Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.4 Compactness and uniform continuity . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4 Connectedness 51
4.1 Definitions, examples and properties . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Path connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3 Connected components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5 Complete metric spaces 61

A Topological existence and non-existence theorems 65

B From the local to the global 67

3
4 CONTENTS
Chapter 0

Preliminaries

0.1 Some fundamental properties of the real line

We start by formulating two fundamental properties of the set of real numbers that we shall use in
the sequel. This set is denoted by R and it is naturally identified to a line. An upper bound of a
subset E ⊂ R is a real number M such that x ≤ M for all x ∈ E. The set E is called bounded
from above if it has an upper bound. A lower bound of a subset E ⊂ R is a real number m such
that m ≤ x for all x ∈ E. The set E is called bounded from below if it has a lower bound.

Definition 0.1 Let E ⊂ R be bounded from above. The smallest upper bound of E is called the
least upper bound of E or the supremum of E. It is denoted by sup E. If E is not bounded from
above, we write sup E = +∞.

Proposition 0.1 Let E ⊂ R be bounded from above. Then α = sup E if and only if the following
hold.
i) x ≤ α for all x ∈ E.
ii) ∀ε > 0 ∃y ∈ E such that α − ε < y.

The first condition means that α is an upper bound of E; the second condition means that α − ε
is not an upper bound of E no matter how ε is small.

Definition 0.2 Let E ⊂ R be bounded from below. The biggest lower bound of E is called the
greatest lower bound of E or the infimum of E. It is denoted by inf E. If E is not bounded from
below, we write inf E = −∞.

Proposition 0.2 Let E ⊂ R be bounded from below. Then α = inf E if and only if the following
hold.
i) x ≥ α for all x ∈ E.
ii) ∀ε > 0 ∃y ∈ E such that α + ε > y.

The first condition means that α is a lower bound of E; the second condition means that α + ε is
not an lower bound of E no matter how ε is small.

Proposition 0.3 Let ∅ ̸= A ⊂ B ⊂ R. Then sup A ≤ sup B and inf A ≥ inf B.

Exercise. Let A and B be two nonempty subsets of R. If x ≤ y for all x ∈ A and all y ∈ B, then
sup A ≤ inf B.

Now we can state the two fundamental properties of R.

5
6 CHAPTER 0. PRELIMINARIES

Theorem 0.1 Any nonempty subset of R which is bounded from above has a least upper bound.
Any nonempty subset of R which is bounded from below has a greatest lower bound.

Any proof of this theorem involves going back to the construction of the real numbers from the
rational numbers. We do not address this issue here. Note that the above theorem
√ is not√true for
the set of rational numbers Q. For example, the supremum of set {x ∈ Q | x2 < 2} is 2 ∈ / Q.
This is one of the reasons we prefer to work in the R rather than in Q.
Recall now that a subset E ⊂ R is an interval if and only if [x, y] ⊂ E whenever x and y belong
to E. An interval has thus one of the following eleven forms ∅, {a}, ]a, b[, ]a, b], [a, b[, [a, b],
] − ∞, a], ] − ∞, a[, ]a, ∞[, [a, ∞[ and R. An interval is called trivial if it is empty or a singleton.

Theorem 0.2 Any nontrivial interval of the real line contains rational as well as irrational numbers.

This theorem will be proved in the exercises. We end with a useful result.

Proposition 0.4 Let a and b be given real numbers. If a ≤ b + ε for all ε > 0 then a ≤ b.
a−b
Proof. Otherwise, taking ε = 2 , we reach a contradiction.

0.2 Sets and functions

1. Given a set X we denote by P(X) or 2X the set of all subsets of X. The reason for the notation
2X is that if X is finite and contains n elements, then there are 2n subsets of X. We use small letters
to denote elements of a set (a, b, c, x, y, z...), capital letters to denote sets (A, B, C, X, Y, Z...) and
calligraphic letters to denote sets of sets (A, B, C, X , Y, Z...). The terms ”set of sets”, ”collection
of sets” and ”family of sets” all mean the same thing. A set containing only one element is called
a singleton or a one-point set.

2. Let X be a set and let A ⊂ X. The characteristic function of A is the function χA defined by
®
1 if x ∈ A
χA (x) =
0 if x ∈
/ A.

It is also denoted by 1A .

3. The complement of A in X is the set of points in X that are not in A. It is denoted by one of
the following notations
∁A , ∁AX , X\A, X − A, A .
c

4. Given a family of sets (Ai )i∈I , we set

[
Ai = {x | ∃j ∈ I such that x ∈ Aj },
i∈I
\
Ai = {x | x ∈ Ai , ∀i ∈ I}.
i∈I

For example, [ [
[−x, x] = ] − x, x[= R
x>0 x>0

and \ \
[−x, x] = ] − x, x[= {0}.
x>0 x>0
0.2. SETS AND FUNCTIONS 7

Recall De Morgan’s laws

[ \
X\ Ai = (X\Ai )
i∈I i∈I
\ [
X\ Ai = (X\Ai ),
i∈I i∈I

and the distributivity laws

[ [
A∩ Ai = (A ∩ Ai )
i∈I i∈I
\ \
A∪ Ai = (A ∪ Ai ).
i∈I i∈I

5. Let A and B be two subsets of a set X. Then

• A ⊂ B ⇐⇒ A ∩ B = A ⇐⇒ A ∪ B = B.

• A ∩ B = ∅ ⇐⇒ A ⊂ X − B ⇐⇒ B ⊂ X − A.

• B − (B ∩ A) = B ∩ (X − A).

6. The terms ”function”, ”application”, ”map”, ”mapping”, ”transformation” all mean the same
thing. A map f : X → Y between two sets associates to every element x ∈ X an element f (x) ∈ Y .
If f is not defined for every x ∈ X, we write f : D(f ) ⊂ X → Y where D(f ) is the domain of
definition of f . A map is characterized by three elements: a source set X, a target set Y and a
graph G(f ) = {(x, f (x)); x ∈ X} which is a subset of the Cartesian product X × Y which is the
set of couples (x, y) where x ∈ X and y ∈ Y .
Remark. There are several ways of thinking about the concept of function (which is central in all
modern mathematics). Here are some of them.

(a) A function may be thought of as an analytic expression like y = x2 +ln(1+x) or z = sin(x2 +y).
Mathematicians living around the time of Newton had this point of view. However, this point
of view is very restrictive because not all functions are given by such formulas.

(b) For a physicist or engineer, a function may represent a message or signal. This point of view is
beneficial in signal processing where one applies Fourier analysis in order to extract information
form this signal by transforming it into another signal.

(c) For me, a function is a general transformation that deforms objects. The objects may be
numbers, geometric shapes, sequences or even functions. In set theory, the nature of these
objects is irrelevant. Thinking of functions as transformations may give you a geometric insight
and stretches your imagination. For example, the function tan represents an infinite stretching
of the interval ] − π2 , π2 [. The function f : [0, 2π] → R2 given by f (t) = (cos t, sin t) transforms
the interval [0, 2π] into the unit circle by bending it and pasting the two endpoints. Also, instead
of thinking of a matrix as a table of numbers, it would be more beneficial to think of a matrix
as a transformation which may do the following: stretching and contracting in certain directions
(called eigenvectors), rotating about certain axes and reflecting about certain axes or planes.

(d) As we said, a function has three features: a source set, a target set and a graph. This lead
Bourbaki to define a function as a triple (X, Y, G) where X, Y are sets and G ⊂ X × Y satisfies
the condition: if (x, y) and (x, z) are in G, then z = y (this means that x has a unique image
under f ).
8 CHAPTER 0. PRELIMINARIES

Let f : X → Y be a map and let A ⊂ X. The (direct) image of A under f is

f (A) = {y ∈ Y ; ∃x ∈ X such that y = f (x)}.

Let B ⊂ Y , the inverse image of B under f is

f −1 (B) = {x ∈ X; f (x) ∈ B}.

Do not confuse the direct image with the inverse image. The direct image lives in the target set,
whereas the inverse image lives in the source set. Regardless of whether f is a bijection or not, f −1
defines a map

f −1 : P(Y ) → P(X)
B 7→ f −1 (B).

▶ This map satisfies the following properties.

(i) f −1 (∅) = ∅, f −1 (Y ) = X, and f −1 (f (X)) = X.

(ii) f −1 (Y − B) = X − f −1 (B).

(iii) If B ⊂ C ⊂ Y , then f −1 (B) ⊂ f −1 (C).

(iv) f −1 −1 (B ).
S
S
i∈I Bi = i∈I f i

(v) f −1 −1 (B ).
T
T
i∈I Bi = i∈I f i

(vi) f (f −1 (B)) ⊂ B and equality holds if f is surjective.

(vii) A ⊂ f −1 (f (A)) and equality holds if f is injective.

(viii) If B ⊂ Y , then f −1 (B ∩ f (X)) = f −1 (B). In particular, if B ∩ f (X) = ∅ then f −1 (B) = ∅.

▶ Note that if (Ai )i∈I is a family of subsets of X, then

!
[ [
f Ai = f (Ai ),
i∈I i∈I

but we only have !

\ \
f Ai ⊂ f (Ai ),
i∈I i∈I

and equality holds if f is injective.

▶ Without any additional assumption, there is no inclusion relation between f (X\A) and Y \f (A).
However,

(a) If f is injective, then, f (X\A) ⊂ Y \f (A).

(b) If f is surjective, then, Y \f (A) ⊂ f (X\A).

Therefore f (X\A) = Y \f (A) if f is bijective.

▶ When we restrict f to a subset A ⊂ X, we get a map denoted by f|A : A → X. Note that

−1
f|A (B) = A ∩ f −1 (B).

8. Given two sets X and Y , the projection onto X is the map π1 : X × Y → X given by
π1 (x, y) = x. It is also denoted by p1 , πX or pX . The projection onto Y is the map π2 : X ×Y → X
0.2. SETS AND FUNCTIONS 9

given by π2 (x, y) = y. It is also denoted by p2 , πY or pY . If A ⊂ X, then π1−1 (A) = A × Y . If

B ⊂ Y , then π2−1 (B) = X × B.

9. Let f : X → Y × Z be a map. So for every x ∈ X, we can write f (x) = (f1 (x), f2 (x)). This
defines two maps f1 : X → Y and f2 : X → Z. f1 and f2 are called the components of f . If π1
and π2 are the projections from Y × Z onto Y and Z respectively, we can write f1 = π1 ◦ f and
f2 = π2 ◦ f . If U ⊂ Y and V ⊂ Z, then

f −1 (U × V ) = f1−1 (U ) ∩ f2−1 (V ).

10. If f : X × Y → Z is a map of two variables and x ∈ X, the map g : Y → Z, given by

g(y) = f (x, y) is called the partial map with respect to the second variable and it is denoted by
f (x, ·). We define similarly the partial map f (·, y) : X → Z for y ∈ Y .
10 CHAPTER 0. PRELIMINARIES
Chapter 1

Metric spaces and topological spaces

1.1 Definitions and examples

As we said, topological spaces are a generalization of metric spaces which are themselves a gener-
alization of R and RN . A metric space is a set equipped with a distance. The notion of distance
gives a meaning to the closeness of two objects: two objects are close if their distance is small. So
how to define the notion of a distance in a general context?
Consider the usual Euclidean distance between two points in a plane equipped with an or-
thonormal coordinate system (so we identify the plane with R2 ). Let d(x, y) denote the Euclidean
distance between
p two points x = (x1 , x2 ) and y = (y1 , y2 ). Note that by the Pythagorean theorem,
d(x, y) = (x1 − y1 )2 + (x2 − y2 )2 . It is clear (geometrically at least) that

(i) d(x, y) ≥ 0 for all x, y ∈ R2 .

(ii) d(x, y) = 0 if and only if x = y.

(iii) d(x, y) = d(y, x) for all x, y ∈ X (symmetry).

(iv) d(x, z) ≤ d(x, y) + d(y, z) for all x, y, z ∈ R2 . This inequality is known as the triangle
inequality: in the triangle xyz, the length of an edge is less than or equal to the sum of the
lengths of the other edges.

These properties of the Euclidean distance are used as an axiomatic definition of a distance over an
arbitrary set.
Definition. Let X be a non empty set. A distance or metric over X is a function d : X × X → R
that satisfies the following properties

(i) d(x, y) ≥ 0 for all x, y ∈ X.

(ii) d(x, y) = 0 if and only if x = y.

(iii) d(x, y) = d(y, x) for all x, y ∈ X (symmetry).

(iv) d(x, z) ≤ d(x, y) + d(y, z) for all x, y, z ∈ X (Triangle inequality).

A metric space is a couple (X, d) where d is a distance on X.

Example 1. Let d(x, y) = |x − y| for x, y ∈ R. Then d is a distance called the usual or standard
distance on R.
Example 2. Let ρ(x, y) = |x3 − y 3 | for x, y ∈ R. Then ρ is a distance on R. Indeed,

(i) ρ(x, y) ≥ 0 because it is an absolute value.

11
12 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

(ii) ρ(x, y) = 0 ⇐⇒ |x3 − y 3 | = 0 ⇐⇒ x3 = y 3 ⇐⇒ x = y.

(iii) ρ(x, y) = |x3 − y 3 | = |y 3 − x3 | = ρ(y, x).

(iv) ρ(x, z) = |x3 − z 3 | = |x3 − y 3 + y 3 − z 3 | ≤ |x3 − y 3 | + |y 3 − z 3 | = ρ(x, y) + ρ(y, z).

Remark. The function d defined by d(x, y) = |x2 − y 2 | for x, y ∈ R is NOT a distance. (Why
not?)
Example 3. Let γ(x, y) = min(1, |x − y|) for x, y ∈ R. Then γ is a distance on R. Before we
prove that, let us observe the following.
®
1 if |x − y| ≥ 1
1. γ(x, y) =
|x − y| if |x − y| ≤ 1.

2. For all x, y ∈ R, we have γ(x, y) ≤ 1 and γ(x, y) ≤ |x − y|.

Now, we prove that γ is a distance.

(i) Since |x − y| ≥ 0 and 1 > 0, then their minimum is ≥ 0. Thus γ(x, y) ≥ 0.

(ii) If x = y, then, |x − y| = 0 and so γ(x, y) = min(1, 0) = 0. Conversely suppose that

γ(x, y) = 0. Now, γ(x, y) is either equal to 1 or to |x − y|. Since it cannot be equal to 1, we
necessarily have γ(x, y) = |x − y| and so |x − y| = 0. Therefore x = y.

(iii) γ(y, x) = min(1, |y − x|) = min(1, |x − y|) = γ(x, y).

(iv) Let x, y, z be three real numbers. We distinguish between two cases.

Case 1. Either |x − y| ≥ 1 or |y − z| ≥ 1. Then γ(x, y) = 1 or γ(y, z) = 1. Therefore,
γ(x, y) + γ(y, z) ≥ 1. On the other hand, γ(x, z) ≤ 1. Hence the triangle inequality in this
case.
Case 2. |x − y| < 1 and |y − z| < 1. Then γ(x, y) = |x − y| and γ(y, z) = |y − z|. Therefore,
γ(x, z) ≤ |x − z| ≤ |x − y| + |y − z| = γ(x, y) + γ(y, z).

Example 4. More generally, let (X, d) be a metric space and set d(x, y) = min(1, d(x, y)). Then
d is a distance on X. The proof is similar to the proof above.
Example 5. Let a < b and X = C([a, b]) be the set of continuous functions f : [a, b] → R. Let
d(f, g) = supx∈[a,b] |f (x) − g(x)| (draw a figure). Then d is a distance on X. Observe first that for
every x ∈ [a, b], we have |f (x) − g(x)| ≤ d(f, g). Now,
(i) since |f (x)−g(x)| ≥ 0 for all x ∈ [a, b], we have supx∈[a,b] |f (x)−g(x)| ≥ 0 and so d(f, g) ≥ 0.

(ii) If f = g, then f (x) = g(x) for all x ∈ [a, b] and so d(f, g) = supx∈[a,b] |f (x) − g(x)| = 0.
Conversely if d(f, g) = 0, then 0 ≤ |f (x) − g(x)| ≤ d(f, g) = 0 for all x ∈ [a, b]. Therefore
f (x) = g(x) for all x ∈ [a, b] and so f = g.

(iii) d(g, f ) = supx∈[a,b] |g(x) − f (x)| = supx∈[a,b] |f (x) − g(x)| = d(f, g).

(iv) Let f, g, h ∈ C[a, b] and let x ∈ [a, b]. Then,

|f (x) − h(x)| ≤ |f (x) − g(x)| + |g(x) − h(x)| ≤ d(f, g) + d(g, h).

Since x was arbitrary, we get

d(f, h) = sup |f (x) − h(x)| ≤ d(f, g) + d(g, h).

x∈[a,b]
1.1. DEFINITIONS AND EXAMPLES 13

Exercise. Let a = 0, b = π2 , f (x) = x and g(x) = sin x. Find d(f, g).

Z b
Example 6. Again consider the set X = C([a, b]) and set d(f, g) = |f (x) − g(x)| dx (draw a
a
figure). Then d is a distance on X. Indeed,

(i) Since |f (x) − g(x)| ≥ 0 for all x ∈ [a, b], and integration conserves inequalities, we have
d(f, g) ≥ 0.

(ii) If f = g, then |f (x) − g(x)| = 0 for all x ∈ [a, b] and by integration we get d(f, g) = 0.
Z b
Conversely, suppose that d(f,g)=0. Then |f (x) − g(x)| dx = 0. Since |f − g| is continuous
a
and has a constant sign, a fundamental result in integration1 implies that |f (x) − g(x)| = 0
for all x ∈ [a, b] and so f = g.
Z b Z b
(iii) d(g, f ) = |g(x) − f (x)| dx = |f (x) − g(x)| dx = d(f, g).
a a

(iv) Let f, g, h ∈ C[a, b]. We have |f (x) − h(x)| ≤ |f (x) − g(x)| + |g(x) − h(x)|. Integrating this
inequality and using the linearity of the integral, we get d(f, h) ≤ d(f, g) + d(g, h).

Exercise. Let a = 0, b = π2 , f (x) = x and g(x) = sin x. Find d(f, g).

Example 7. Let X be a set. The discrete distance d on X is defined by
®
0 if x = y
d(x, y) =
1 ̸ y.
if x =

The first three properties should be clear. Let x, y, z ∈ X. To prove the triangle inequality, we
distinguish between two cases.

Case 1. x = z. Then d(x, z) = 0 ≤ d(x, y) + d(y, z).

Case 2. x ̸= z. Then either y ̸= x or y ̸= z. Therefore either d(x, y) = 1 or d(y, z) = 1. Therefore

d(x, z) = 1 ≤ d(x, y) + d(y, z).

Norms and distances on RN

Let N be a positive integer. The set of all N −tuples of real numbers (x1 , x2 , . . . , xN ) is denoted
by RN . Otherwise stated, RN is the Cartesian product of N copies of R. Elements of RN are
considered as vectors with N components. The origin of RN is the vector (0, . . . , 0) denoted simply
by 0.
A norm on RN is a way to measure the length of a vector (x1 , . . . , xN ). It is the analog of the
absolute value of a real number. The norm of a vector x ∈ RN is a number denoted by ∥x∥ that
satisfies the following properties.

(i) ||x|| ≥ 0.

(ii) ∥x∥ = 0 if and only if x = 0.

(iii) ∥λx∥ = |λ|∥x∥ for any real number λ and any x ∈ RN .

(iv) ∥x + y∥ ≤ ∥x∥ + ∥y∥ for all x, y ∈ RN (the triangle inequality).

1
Rb
If h is continuous, has a constant sign on [a, b] and a
h(x) dx = 0, then h = 0 on [a, b].
14 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

The most natural norm on RN is the Euclidean norm. Let x = (x1 , . . . , xN ) ∈ RN . The
Euclidean norm of x is defined by

N
!1/2
X
2
∥x∥ = |xi | .
i=1

Other norms can also be defined on RN . The sup norm or infinity norm of a vector x is defined
by
∥x∥∞ = max(|x1 |, . . . , |xN |).

It is easy to check that this norm satisfies the four properties of a norm.
The 1−norm is defined by
XN
∥x∥1 = |xi |.
i=1

Here again it is not difficult to prove the four properties of a norm. More generally, let p ≥ 1. The
p−norm of x is
N
!1/p
X
||x||p = |xi |p .
i=1

It can be shown that this is indeed a norm (the triangle inequality is not trivial to prove; it follows
from another inequality known as Hölder’s inequality). Note finally that all these norms coincide
when N = 1.
Now a norm on RN defines a metric by setting d(x, y) = ||x − y|| (check that).

Balls in metric spaces

The open ball of center x and radius r is the set

B(x, r) = {y ∈ X; d(y, x) < r}.

The closed ball of center x and radius r is the set

B ′ (x, r) = {y ∈ X; d(y, x) ≤ r}.

Remark. If there are more than one distance on a set X, we need to distinguish between balls
with respect to each distance. The open ball of center x and radius r with respect to a distance d
will then be denoted by Bd (x, r). The corresponding closed ball will be denoted by Bd′ (x, r).
Example 1. The open ball B(a, r) in R (with the usual distance) is the interval ]a − r, a + r[.
Indeed,

x ∈ B(a, r) ⇐⇒ |x − a| < r ⇐⇒ −r < x − a < r ⇐⇒ a − r < x < a + r.

The closed ball B ′ (a, r) in R (with the usual distance) is the interval [a − r, a + r].
Example 2. In (R2 , || · ||2 ) the open ball B(a, r) is the usual open disc of center a and radius r.
Indeed, let a = (a1 , a2 ), then

1/2
(x1 , x2 ) ∈ B(a, r) ⇐⇒ |x1 − a1 |2 + |x2 − a2 |2 < r ⇐⇒ |x1 − a1 |2 + |x2 − a2 |2 < r2 .
1.1. DEFINITIONS AND EXAMPLES 15

x2

x1

In (R2 , || · ||2 ) the closed ball B ′ (a, r) is the usual closed disc of center a and radius r.

x2

x1

Example 3. In (R3 , || · ||2 ), the open ball B(a, r) is the usual geometric open ball of center a and
radius r. Same thing for the closed ball.
Example 4. In (R2 , || · ||∞ ) the open ball B(0, 1) is the full open square of vertices (-1,-1), (1,-1)
(1,1) and (-1,1). Indeed,

(x1 , x2 ) ∈ B(0, 1) ⇐⇒ max(|x1 − 0|, |x2 − 0|) < 1 ⇐⇒ max(|x1 |, |x2 |) < 1
⇐⇒ |x1 | < 1 and |x2 | < 1 ⇐⇒ −1 < x1 < 1 and − 1 < x2 < 1.

x2

x1

In (R2 , || · ||∞ ) the closed ball B ′ (0, 1) is the full closed square of vertices (-1,-1), (1,-1) (1,1) and
(-1,1).

x2

x1
16 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

Example 5. 5) In (R2 , || · ||1 ), the ball B(0, 1) is the square of vertices (1,0), (0,-1) (-1,0) and
(0,1). Indeed, first, (x1 , x2 ) ∈ B(0, 1) ⇐⇒ |x1 | + |x2 | < 1. We distinguish between 4 cases.
(i) x1 ≥ 0 and x2 ≥ 0 (first quadrant). In this case we have x1 + x2 < 1. So this is the triangle
below the line of equation x2 = 1 − x1 .
x2

x1

(ii) x1 ≤ 0 and x2 ≥ 0 (second quadrant). In this case we have −x1 + x2 < 1 and so x2 < x1 + 1.
So this the triangle below the line of equation x2 = x1 + 1.
x2

x1

(iii) x1 ≤ 0 and x2 ≤ 0 (third quadrant). In this case we have −x1 − x2 < 1 and so x2 > −x1 − 1.
So this the triangle above the line of equation x2 = −x1 − 1.

x2

x1

(iv) x1 ≥ 0 and x2 ≤ 0 (fourth quadrant). In this case we have x1 − x2 < 1 and so x2 > x1 − 1.
So this the triangle above the line of equation x2 = x1 − 1.
x2

x1
1.2. OPEN SETS AND RELATED CONCEPTS 17

Putting everything together we get the rotated square

x2

x1

In (R2 , || · ||1 ) the closed ball B ′ (0, 1) is the full closed rotated square.

x2

x1

Example 6. Let d be the discrete distance on a set X. Then B(a, 1) = {a} and B ′ (a, 1) = X.
Indeed, since d takes just two values 0 and 1, we have

x ∈ B(a, 1) ⇐⇒ d(x, a) < 1 ⇐⇒ d(x, a) = 0 ⇐⇒ x = a ⇐⇒ x ∈ {a}.

Next for every x ∈ X, we have d(x, a) ≤ 1. Therefore x ∈ B ′ (a, 1). This means that X ⊂ B ′ (a, r).
Since B ′ (a, r) ⊂ X, we get equality.

1.2 Open sets and related concepts

Open subsets of a metric space are a generalization of open intervals and open subsets of RN . The
adjective ”open” probably comes from an analogy with some situation in real life. For example a
container is open if it has no cap. The cap may be defined as a boundary. So a set is open if it
does not contain any part of its boundary (we will make this clear once we introduce the necessary
definitions). When the set does not meet its boundary, it has the property that whenever a point
x is in the set, all points that are sufficiently close to x must belong to set. Why are we interested
in open sets? Here’s one reason. The derivative of a function f : D(f ) ⊂ R → R at a point x is
defined by
f (x + h) − f (x)
f ′ (x) = lim (if the limits exists).
h→0 h
But this requires that x + h belong to the domain of definition D(f ) for all h small enough (say
|h| < r). The same situation happens if we want to define the partial derivatives or the total
derivative of a function of several variables.
Therefore if f : D(f ) ⊂ RN → R must be differentiable everywhere, its domain of definition
should satisfy one the following equivalent properties.

• ∀x ∈ D(f ), ∃r > 0, ||h|| < r ⇒ x + h ∈ D(f ).

18 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

• ∀x ∈ D(f ), ∃r > 0, ||x − y|| < r ⇒ y ∈ D(f ).

• ∀x ∈ D(f ), ∃r > 0, B(x, r) ⊂ D(f ).

Remark. When the domain of f does not satisfy this property, one should be careful in defining
the notion of differentiability. For example we say that f : [a, b] → R is differentiable at b if the left
limit exists
f (b + h) − f (b)
f ′ (b) = lim .
h→0,h<0 h
Another way to define differentiability is the following. We say that f : A → R is differentiable if f
can be extended to a differentiable function g : O → R where O is an open set containing A.

Definition. Let X be a metric space. A subset O ⊂ X is said to be open if for every x ∈ O there
is an r > 0 such that B(x, r) ⊂ O.

Convention. Unless otherwise stated, R will always be equipped with the usual distance d(x, y) =
|x − y|.
Examples. The interval ]0, 1[ is open in R but ]0, 1] is not. Any interval of the form ]a, b[, ]a, ∞[,
] − ∞, a[ is open. We proved that in M1106.
Remark. Since B(x, r) ⊂ B ′ (x, r) and B ′ (x, 2r ) ⊂ B(x, r), the condition of openness is equivalent
to the following
∀x ∈ O ∃r > 0 B ′ (x, r) ⊂ O.

Proposition 1.1. An open ball is an open set.

Remark. This proposition is not a tautology because the adjective open has different meanings in
”open ball” and ”open set”. An open ball is a set of the form B(a, r) = {x; d(x, a) < r}, whereas
an open set is a set having the property that if a point belongs to this set, then a whole ball around
this point is also in the set. So we have to prove that an ”open ball” has this property.
Proof. Consider an open ball B(a, r) and let x ∈ B(a, r).

a
r
1.2. OPEN SETS AND RELATED CONCEPTS 19

Then d(x, a) < r. Set ρ = r − d(x, a). We claim that B(x, ρ) ⊂ B(a, r). Indeed, let y ∈ B(x, ρ).
Then d(x, y) < ρ = r − d(x, a). Therefore d(x, a) + d(x, y) < r. By the triangle inequality,
d(y, a) < r. This means that y ∈ B(a, r). The claim implies that B(a, r) is open. □
Exercise. Prove that an open set of a metric space is a union of open balls.
Proposition 1.2. Let X be a metric space. Then the following properties hold.

(i) X and ∅ are open.

(ii) An arbitrary union of open sets is open.

(iii) A finite intersection of open sets is open.

Proof. (i). Let x ∈ X. Then indeed B(x, 1) ⊂ X. Therefore X is open. The empty set is open
because the condition x ∈ ∅ is never satisfied.
S
(ii). Let (Oλ )λ∈L be a collection of open subsets of X and let x ∈ Oλ . Then x ∈ OSµ for some
µ ∈ L. Since Oµ is open, there exists r > 0 such that B(x, r) ⊂ Oµ and so B(x, r) ⊂ Oλ .
(iii). It is enough to prove that the intersection of two open sets is open because the result will
follow by induction on the number of open sets. Let therefore O1 and O2 be open sets and let
x ∈ O1 ∩ O2 . Since O1 is open and x ∈ O1 , there exists r1 > 0 such that B(x, r1 ) ⊂ O1 . Similarly
there exists r2 > 0 such that B(x, r2 ) ⊂ O2 . It follows that B(x, r1 ) ∩ B(x, r2 ) ⊂ O1 ∩ O2 . But
B(x, r1 ) ∩ B(x, r2 ) = B(x, r) where r = min(r1 , r2 ). This means that O1 ∩ O2 is open. □
Remark. An arbitrary intersection of open sets need not be open. For example,
\ 1 1
− , = {0}
n n
n≥1

is not open.

Definition. A neighborhood of a point x in a metric space X is a set V which contains an open

set containing x. The set of neighborhoods of a point x is usually denoted by U(x).
Remark. Note that the whole space is a neighborhood of x, therefore U(x) is not empty. Note
also that if V ∈ U(x) and V ⊂ U then U ∈ U(x).
Example. [−1, 1] is a neighborhood of any point x ∈] − 1, 1[. However it is not a neighborhood
of 1 nor of −1.
Proposition 1.3. A set is open if and only if it is a neighborhood of all of its points.
Proof. Suppose first that A is open and let x ∈ A. Then A contains an open set (itself) containing
x. This means that A is a neighborhood of x. Conversely, suppose that A is a neighborhood of all
its points. This means that for every x ∈ A, there exists an open set Ox such that x ∈ Ox ⊂ A.
Then, we can write A = ∪x∈A Ox . Therefore A is open as a union of open sets. □

Remark. For some mathematicians, a neighborhood of a point is an open set containing that
point. This makes a little difference because in practice we can always assume that a neighborhood
is open. But in this case Proposition 1.3. is a tautology.

Definition. The interior of a set A is the union of all open sets that are contained in A. It is
◦
therefore the biggest open set contained in A. It is denoted by A or intA.

Examples.
20 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

◦
a) [0,
¯ 1] =]0, 1[. Indeed, ]0,1[ is open and contained in [0,1]. Therefore it is contained in the interior
◦ ◦
of [0,1]. So we have ]0, 1[⊂ [0,
¯ 1] ⊂ [0, 1]. Therefore [0,
¯ 1] is either ]0,1[, ]0,1], [0,1[ or [0,1]. But
◦ ◦
the only open set among these sets is ]0,1[. Hence the result. Similarly, [0,
¯ 1[ = ]0,
¯ 1] =]0, 1[.
◦
b) Q = ∅. This is because any open interval contains points outside Q and so Q cannot contain
◦
any open interval and therefore cannot contain any nonempty open set. Similarly, R\Q
¯ = ∅.
◦
c) Similarly, Z = ∅.

Proposition 1.4. A set is open if and only if it is equal to its interior.

◦ ◦
Proof. Suppose first A = A. But A is open. Therefore A is open. Conversely, suppose that A
◦ ◦
is open. But A contains A. Since A is the biggest open set containing A, we have A ⊂ A. But
◦
A ⊂ A. Hence the equality. □

Closed sets
Definition. Let X be a metric space. A subset F ⊂ X is called closed if its complement is open.
Examples.
1) [a, b] is closed because its complement ] − ∞, a[ ∪ ]b, +∞[ is open.
2) [a, +∞[ and ] − ∞, a] are closed.
3) In a metric space singletons are closed. Indeed, let {a} be a singleton (one point set) and let
x ∈ X − {a}. Then x ̸= a and so d(x, a) > 0. We claim that B(x, d(x, a)) ⊂ X − {a}. Indeed,
let y ∈ B(x, d(x, a)). Then d(x, y) < d(x, a). This implies that y ̸= a because otherwise we would
have d(x, a) < d(x, a). This means that y ∈ X − {a}. Therefore etc.

Proposition 1.5. A closed ball is a closed set.

This proposition is not a tautology.
Proof. Let B ′ (a, r) be a closed ball in a metric space X. We have to prove that X\B ′ (a, r) is open.
Let x ∈ X\B ′ (a, r). Then d(x, a) > r. Set ρ = d(x, a) − r. We claim that B(x, ρ) ⊂ X\B ′ (a, r).
Indeed, let y ∈ B(x, ρ). Then d(x, y) < ρ = d(x, a) − r. Therefore r < d(x, a) − d(x, y). By the
triangle inequality, d(x, a) − d(x, y) ≤ d(y, a). Hence d(y, a) > r and so y ∈ X\B ′ (a, r). This
proves the claim. The claim implies that X\B ′ (a, r) is open. □

Proposition 1.6. In a metric space X, the following properties hold.

(i) X and ∅ are closed.

(ii) An arbitrary intersection of closed sets is closed.

(iii) A finite union of closed sets is closed.

Proof. Take complements in Proposition 1.2. □

Consequence. Let A and B be two subsets of a metric space.

(i) If A is open and B is closed then A\B is open; in words when we remove a closed set from
an open set we get an open set.
1.2. OPEN SETS AND RELATED CONCEPTS 21

(ii) If A is closed and B is open then A\B is closed; in words when we remove an open set from
a closed set we get an closed set.

Proof. We can write A\B = A ∩ (X\B).

(i) If A is open and B is closed then X\B is open and so A ∩ (X\B) is open (as a finite intersection
of open sets).
(ii) If A is closed and B is open then X\B is closed and so A ∩ (X\B) is closed (as an intersection
of two closed sets). □
Corollary 1.1. In a metric space every finite set is closed.
Remark 1 . An arbitrary union of closed sets need not be closed. For example for each n ≥ 2,
the set [ n1 , 1 − n1 ] is closed. However
∞
[ 1 1
,1 − =]0, 1[
n n
n=2

is not closed.
Remark 2. A set which is not open is not necessarily closed. For example ]0,1] is neither open nor
closed. Also a set can be both open and closed. Indeed, in any metric space X, ∅ and X are closed
an open. Here is another example. Let d be the discrete distance on a set X containing more than
one point. Then, B(a, 1) = {a}. Hence {a} is open. But we know that {a} is also closed. In fact,
every subset of X is open (because a subset is a union of singletons). But if every subset of X is
open, then, every subset is also closed. A subset which is both closed and open is sometimes called
clopen. If X and ∅ are the only clopen subsets of X, we say that X is connected (chapter 4).

Definition. The closure of a set A is the intersection of all closed sets that contain A. It is
therefore the smallest closed set containing A. It is usually denoted by A and sometimes by cl(A).
Examples.
a) ]0, 1[ = [0, 1]. Indeed, [0,1] is a closed set containing ]0,1[, therefore it contains ]0, 1[. Thus
]0, 1[⊂ ]0, 1[ ⊂ [0, 1]. Therefore ]0, 1[ is either ]0,1[, ]0,1], [0,1[ or [0,1]. However, the only closed
set among these four sets is [0,1]. Hence the result. Similarly, ]0, 1] = [0, 1[ = [0, 1].
b) We shall see below that Q = R. We say that Q is dense in R. We shall also prove that the
irrationals are dense in R.
c) We shall see in the exercises that if A is the open disk x2 + y 2 < 1, then, the closure of A is the
closed disk x2 + y 2 ≤ 1.
Proposition 1.7. A set is closed if and only if it is equal to its closure.
Proof. Suppose first that A = A. Since A is closed, A is closed. Conversely, suppose that A is
closed. Since A contains A and A is the smallest closed set containing A, we have A ⊂ A. But
A ⊂ A. Hence the equality. □

Convergence
In a metric space, we can define the notion of convergence. A sequence (xn ) converges to x in a
metric space, if xn can be made arbitrarily close to x for all n large enough. This idea of convergence
is therefore captured by the following definition.
Definition. Let (xn ) be a sequence in a metric space (X, d). We say that (xn ) converges to a point
x and write xn → x if the following condition holds.

∀ε > 0 ∃n0 ∈ N ∀n ≥ n0 d(xn , x) < ε.

22 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

If you studied calculus, this condition is equivalent to the condition that the sequence of real numbers
(d(xn , x)) converges to 0 in R.
Remark. The following conditions are equivalent.
(i) xn → x.

(ii) ∀ε > 0 ∃n0 ∈ N ∀n ≥ n0 d(xn , x) < ε.

(iii) ∃ε0 > 0 ∀ε > 0 ε ≤ ε0 ∃n0 ∈ N ∀n ≥ n0 d(xn , x) < ε.

(iv) ∀ε > 0 ∃n0 ∈ N ∀n ≥ n0 xn ∈ B(x, ε).

(v) ∀U ∈ U(x) ∃n0 ∈ N ∀n ≥ n0 xn ∈ U .

In words, the last condition states that every neighborhood of x contains xn for all n large enough.

Proposition 1.8. Let X be a metric space. Let A ⊂ X and let x ∈ X. Then, the following
conditions are equivalent.
(i) x ∈ A.

(ii) Every neighborhood of x intersects A.

(iii) There is a sequence in A which converges to x.

Proof. We will prove that ¬(i)⇔ ¬(ii) and (ii)⇔(iii).

¬(i) ⇒ ¬(ii). Suppose that x ∈ / A. Then, the set U := X\A is an open neighborhood of x that
does not meet A. This is ¬(ii).
¬(ii) ⇒ ¬(i). Let U be an open neighborhood of x that does not meet A. Then, A ⊂ X\U . Since
X\U is closed, we get A ⊂ X\U . Since x ∈ U , x ∈
/ X\U and so x ∈ / A.
(ii) ⇒ (iii). For every positive integer n the open ball B(x, n1 ) intersects A. Choose accordingly
for each n, an element xn in the intersection. This defines a sequence of points in A which clearly
converges to x.
(iii) ⇒ (ii). Let U be a neighborhood of x. Since (xn ) converges to x, we have that xn ∈ U for n
large enough. This means that U intersects A. □

Remark. The above conditions are also equivalent to the following one.
(iv) There exists ε0 > 0 such that for every ε > 0 such that ε ≤ ε0 we have B(x, ε) ∩ A ̸= ∅; in
words, every small enough neighborhood of x meets A (or every small enough ball of center
x meets A).
Proof. It is enough to prove that (iv) ⇒ (ii). So let ε0 > 0 be as in condition (iv). Let ε > 0 be
arbitrary. If ε ≤ ε0 , then, by assumption B(x, ε)∩A ̸= ∅. If ε > ε0 , then, B(x, ε0 )∩A ⊂ B(x, ε)∩A.
By assumption, B(x, ε0 ) ∩ A ̸= ∅. Therefore B(x, ε) ∩ A ̸= ∅. □

Corollary 1.2. Let X be a metric space. Let A ⊂ X and let (xn ) be sequence in A which
converges to some x. Then, x ∈ A.
In particular, if A is closed and (xn ) is a sequence in A converging to x, then, x ∈ A. We can say
the set is ”closed” under the operation of taking limits (we cannot escape the set no matter how
close to the boundary we get).

Definition. Let X be a metric space. A subset A ⊂ X is called dense if A = X.

Corollary 1.3. Let X be a metric space and A ⊂ X. Then, the following are equivalent.
1.2. OPEN SETS AND RELATED CONCEPTS 23

(i) A is dense in X.

(ii) Every nonempty open set of X meets A.

(iii) For every x ∈ X, there exists a sequence in A that converges to x.

Corollary 1.4. The set of rational numbers and the set of irrational numbers are dense in R.

Proof. Let O be an open nonempty subset of R. Then, O contains an open interval I. But any
open interval of R contains rational and irrational numbers. Therefore O meets Q as well as R\Q.
□

Other examples.
a) R∗ = R\{0} is dense in R.
b) R\Z is dense in R. This is because Z has an empty interior.

Limit points and isolated points

Let A be a subset of a metric space X and let x ∈ A. We know from Proposition 1.8 that every
neighborhood of x meets A. There are two possible mutually exclusive cases.

1. There exists a neighborhood U of x such that U ∩ A = {x}. Said differently, some neighborhood
of x meets A only at x. In this case, x ∈ A.

2. For every neighborhood U of x, we have U ∩A−{x} = ̸ ∅. Otherwise stated, every neighborhood

of x meets A at a point different from x. In this case, we need not have x ∈ A.

Definition. Let x ∈ A. In the first case above, we say that x is an isolated point of A. In the
second case, we say that x is a limit point or an accumulation point of A.

Examples.

• Let A =]0, 1[∪{2}. Then, any point in [0, 1] is a limit point of A, whereas 2 is an isolated
point of A.

• Every finite set in a metric space consists of isolated points. Indeed, let A = {x1 , . . . , xn }
be a finite set of n distinct points. For each i = 1, . . . , n, let ri = minj̸=i d(xi , xj ). Then,
B(xi , ri ) ∩ A = {xi }.

• Let A = { n1 | n = 1, 2, 3, . . .}. Then, 0 is a limit point of A, and every point in A is isolated.

• Z has no limit points in R. Therefore all points of Z are isolated.

• All points of R are limit points of Q.

Proposition 1.9. Let A be a subset of a metric space X and let x be a limit point of A. Then,
every neighborhood of x contains infinitely many points of A.
24 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

The boundary of set

Definition. Let A be a subset of a metric space X. The boundary of A denoted by ∂A is the set
◦
∂A = A\A.

Examples.
• ∂[0, 1] = [0, 1]\]0, 1[= {0, 1}. Similarly, ∂]0, 1] = ∂[0, 1[= {0, 1}.
◦
• ∂Q = Q\Q = R.
•We shall see in the exercises that if A is the open disk x2 +y 2 < 1, then, ∂A is the circle x2 +y 2 = 1.

Remark 1. The boundary of a set is always a closed set. Why?

Remark 2. ∂A = ∅ ⇔ A is clopen.

1.3 Topology and topologically equivalent distances

If you studied carefully the previous section and did the exercises, you observe that the majority
of concepts and results in metric spaces rely only on the notion of an open set. Therefore we can
extend the results and concepts of the previous to a still more general situation. The properties of
open sets in a metric space were recorded in Proposition 1.2. which is a starting point towards more
generality.
Definition. Let X be a set. A family T of subsets of of X is called a topology on X provided
the following conditions hold.

(i) ∅, X ∈ T .

(ii) An arbitrary union of elements of T belongs to T (we say that T is stable under arbitrary
unions).

(iii) A finite intersection of elements of T belongs to T (we say that T is stable under finite
intersections).

The elements of T are called open sets of X. A topological space is a couple (X, T ). Thus, a
metric space is a topological space but the converse is not true. Most of the concepts and results
of the previous section and the exercises still hold in topological spaces. You will study topological
spaces in the course M3315.
If (X, d) is a metric space, then. the topology generated (or induced) by d is the collection of
all open subsets of X. We will denote it by Td . Thus,

Td = {open subsets of (X, d)}


= O ⊂ X; ∀x ∈ O, ∃r > 0, Bd (x, r) ⊂ O .

Definition. Let d and d′ be two distances on a set X. We say that d and d′ are topologically
equivalent if Td = Td′ , that is, if they generate the same topology.
We give some examples.
Proposition 1.10 Consider on RN the three norms

N N
!1/2
X X
∥x∥1 = |xi |, ||x||2 = |xi |2 , ∥x∥∞ = max(|x1 |, . . . , |xN |).
i=1 i=1
1.3. TOPOLOGY AND TOPOLOGICALLY EQUIVALENT DISTANCES 25

Let d1 , d2 and d∞ be respectively the distances associated with these norms. Then, d1 , d2 and d∞
are topologically equivalent.
Proof. We prove that d2 and d∞ are equivalent. In the exercises, you are asked to prove that d1
and d∞ are equivalent. It is not difficult to see that
√
||x||∞ ≤ ||x||2 ≤ N ||x||∞ .

It follows that √
d∞ (x, y) ≤ d2 (x, y) ≤ N d∞ (x, y).
Therefore,
r
Bd2 (x, r) ⊂ Bd∞ (x, r) and Bd∞ (x, √ ) ⊂ Bd2 (x, r).
N
Now let O be an open set for the topology generated by d∞ and let x ∈ O. Then, there exists
r > 0 such that Bd∞ (x, r) ⊂ O. But Bd2 (x, r) ⊂ Bd∞ (x, r) and so Bd2 (x, r) ⊂ O. This means
that O is open for the topology generated by d2 .
Conversely, let O be an open set for the topology generated by d2 and let x ∈ O. Then, there
exists r > 0 such that Bd2 (x, r) ⊂ O. But Bd∞ (x, √rN ) ⊂ Bd2 (x, r) and so Bd∞ (x, √rN ) ⊂ O.
This means that O is open for the topology generated by d∞ . □

Remark 1.1 (Methodology) Let d and d′ be two two distances on the same set X. Then, Td′ ⊂ Td
(we say that d generates a finer topology than d′ ) if and only if the following condition holds:

∀x ∈ X ∀r > 0 ∃δ > 0 Bd (x, δ) ⊂ Bd′ (x, r).

Therefore d and d′ are topologically equivalent if and only the following two conditions hold

∀x ∈ X ∀r > 0 ∃δ > 0 Bd (x, δ) ⊂ Bd′ (x, r)

∀x ∈ X ∀r > 0 ∃δ > 0 Bd′ (x, δ) ⊂ Bd (x, r).

Suppose that there is a constant k such that d′ (x, y) ≤ kd(x, y) for all x, y ∈ X. Then, the first
condition above holds with δ = kr (check that). So we have

d′ (x, y) ≤ kd(x, y) ⇒ Td′ ⊂ Td .

But the converse is not true.

Remark 1.2 There is another notion of equivalence between distances. Let d and d′ be two distances
on a set X. We say that d and d′ are Lipschitz equivalent or strongly equivalent if there are two
constants α and β such that

αd(x, y) ≤ d′ (x, y) ≤ βd(x, y) ∀x, y ∈ X.

According to what we said above

Lipschitz equivalence ⇒ Topological equavlence

but the converse is not true as shown by the next example.

Proposition 1.11. Let (X, d) be a metric space. Set ρ(x, y) = min(1, d(x, y)). Then, ρ and d
are topologically equivalent.
Proof. We know that ρ is a metric. Let x ∈ X. Observe that Bd (x, r) ⊂ Bρ (x, r) since
ρ(x, y) ≤ d(x, y). By the previous remark, d generates a finer topology than ρ.
Conversely, let x ∈ X and r > 0 be given. Let δ = min(1, r). Then, Bρ (x, δ) ⊂ Bd (x, r).
Indeed, let y ∈ Bρ (x, δ). Then, ρ(x, y) < δ ≤ 1. It follows that ρ(x, y) = d(x, y) (because
ρ(x, y) ̸= 1) and so d(x, y) < δ ≤ r; that is, y ∈ Bd (x, r). □
26 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

Remark 1.3 Observe that the distance ρ is always bounded. So the above result is saying that we
ca always replace a distance by a topologically equivalent bounded distance.

Remark 1.4 Take in the above proposition X = R with the usual distance d(x, y) = |x − y|. Then,
there is no constant k such that d(x, y) ≤ kρ(x, y). If not, then, |x − y| ≤ k for all x, y ∈ R because
ρ ≤ 1. This is impossible. It follows that d and ρ are not Lipschitz equivalent.

1.4 Accumulation points and subsequences

Recall that a sequence (xn ) of a metric space is convergent to a limit x if every neighborhood of x
contains xn for all n large enough, that is except for a finite set of indices. Not all sequences are
convergent. A weaker notion than that of limit is that of accumulation point or cluster point.
Definition. Let (xn ) be a sequence of a topological space. We say that x is an accumulation
point or a cluster point of (xn ) if every neighborhood of x contains xn for infinitely many indices
n. In symbols
∀U ∈ U (x) ∀n ∈ N ∃m ≥ n such that xm ∈ U.
This is equivalent to saying that the set {n ∈ N; xn ∈ U } is infinite.
Compare with the condition of convergence

∀U ∈ U (x) ∃n0 ∈ N m ≥ n0 ⇒ xm ∈ U.

Examples. 1) It should be clear that a limit of a convergent sequence is an accumulation point of

the sequence. The converse however is not true as testified by the next example.
2) Let xn = (−1)n . Then, (xn ) is not convergent in R but has two accumulation points 1 and −1.
3) The sequence (1, 2, 1 + 21 , 2 − 12 , 1 + 13 , 2 − 31 , . . .) has two accumulation points 1 and 2.

Proposition 1.12. Let A be the set of accumulation points of a sequence (xn ). Then,
∞
\ ∞
\
A= {xk ; k ≥ n} = {xn , xn+1 , . . .}.
n=1 n=1

Proof. Let Fn = {xn , xn+1 , . . .}. Then,

x ∈ A ⇐⇒ ∀U ∈ U (x) ∀n ∈ N∗ ∃m ≥ n xm ∈ U
∗
⇐⇒ ∀n ∈ N ∀U ∈ U (x) U ∩ Fn ̸= ∅
∗
⇐⇒ ∀n ∈ N x ∈ F¯n
⇐⇒ x ∈ ∩F¯n .

□
Definition. Let (xn ) be a sequence of some set and let n1 < n2 < · · · < ni < · · · be an increasing
sequence of positive integers. Then, the sequence (yi ) defined by yi = xni is called a subsequence
of the sequence (xn ).
Here’s is way to picture a subsequence. Write the original sequence in a list, and as you go
through the list, delete certain terms (those that do not correspond to the indices {n1 , n2 , . . .}).
What remains is the subsequence.
Examples. a) Let xn = n. Then, (x2n+1 ) is a subsequence of (xn ).
b) Let xn = n. Then, the sequence 2, 4, 8, 16, 32, 64, . . ., i.e. (x2n ) is a subsequence of (xn ).
Proposition 1.13. Let (xn ) be a sequence of a metric space (X, d). Then, the following conditions
are equivalent.
1.5. SUBSPACES 27

(i) x is an accumulation of (xn ).

(ii) There is a subsequence of (xn ) that converges to x.

Proof. (ii) ⇒ (i). Let (xnk ) be a subsequence of (xn ) that converges to x. Let U be a
neighborhood of x. Then, there exists k0 such that xnk ∈ U for all k ≥ k0 . This means that
xm ∈ U for all m ∈ {nk0 , nk0 +1 , nk0 +2 . . .} and this set of integers is infinite.
(i) ⇒ (ii). In the condition

∀U ∈ U (x) ∀n ∈ N ∃m ≥ n such that xm ∈ U,

take first U = B(x, 1) and n = 1. Then, there exists n1 ≥ 1 such that xn1 ∈ B(x, 1). Next
take U = B(x, 21 ) and n = n1 + 1. Then, there exists n2 ≥ n1 + 1 such that xn2 ∈ B(x, 21 ). At
the kth step, take U = B(x, k1 ) and n = nk−1 + 1. Then, there exists nk ≥ nk−1 + 1 such that
xnk ∈ B(x, k1 ). This procedure defines a subsequence (xnk ) that converges to x. □

1.5 Subspaces
Let (X, d) be a metric space and let Y ⊂ X. Let dY denote the restriction of d to Y × Y . Then,
dY is a metric on Y which therefore generates a topology on Y . How to characterize the open
subsets of Y ? If x ∈ Y , we denote by BY (x, r) the ball in Y of center and radius r. It coincides
with Y ∩ B(x, r).
Proposition 1.14. Under the above assumptions, let A ⊂ Y . Then, A is open in Y if and only if
A = Y ∩ O where O is open in X.
Proof. Suppose first that A = Y ∩ O where O is open in X. Let x ∈ A. Then, x ∈ O. Since O
is open in X, there exists r > 0 such that B(x, r) ⊂ O. Then, B(x, r) ∩ Y ⊂ O ∩ Y = A, that is,
BY (x, r) ⊂ A. This means that A is open in Y .
Conversely, suppose that A is open in Y . Then, for each x ∈ A, there exists rx > 0 such that
BY (x, rx ) ⊂ A. It follows that A = ∪x∈A BY (x, rx ). Let now O = ∪x∈A B(x, rx ). Then, O is open
in X and O ∩ Y = A. □
Examples. 1) Consider [0, ∞[ as a subspace of R with the usual distance. Then, although [0, 1[
is not open in R, it is open in [0, ∞[ because for example [0, 1[= [0, ∞[∩] − 1, 1[.
2) Consider Z as a subspace of R with the usual distance. Then, each singleton {n} is open in Z
because {n} =]n − 1, n + 1[∩Z. It follows that that every subset of Z is open in Z.
This lead us to the following definition .

Definition. Let X be a topological space and let Y ⊂ X. The collection {Y ∩ O; O is open in X}

is a topology on Y (check that). It is called the subspace topology of Y or the topology inherited
from X.

Corollary 1.5. Let X a be a topological space, let Y be a subspace of X and let A ⊂ Y . If A is

open in Y and Y is open in X, then, A is open in X.
Proof. If A is open in Y , then, A = Y ∩ O where O is open in X. Since Y is also open in X,
then, so is Y ∩ O. □

Proposition 1.15. Let Y be a subspace of a topological space X and let A ⊂ Y . Then, A is closed
in Y if and only if A = Y ∩ F where F is closed in X.
Proof. Suppose first that A = Y ∩ F where F is closed in X. Then, we can write

Y \A = Y \(Y ∩ F ) = Y ∩ (X\F ).
28 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES

Let O = X\F . Then, O is open in X and Y \A = Y ∩ O. By the previous proposition, Y \A is

open in Y and so A is closed in Y .
Suppose conversely that A is closed in Y . Then, Y \A is open in Y and so Y \A = Y ∩ O where
O is open in X. Then, we can write

A = Y \(Y ∩ O) = Y ∩ (X\O).

Let F = X\O. Then, F is closed in X and A = Y ∩ F . □

Example. ]0,1] is closed in ]0, ∞[ since ]0, 1] =]0, ∞[∩[0, 1].

Corollary 1.6. Let X a be a topological space, let Y be a subspace of X and let A ⊂ Y . If A is

closed in Y and Y is closed in X, then, A is closed in X.

1.6 Product spaces

If X1 , X2 . . . , Xn are sets, Qthe set of n−tuples (x1 , . . . , xn ) where each xi ∈ Xi is denoted by
X1 × X2 × · · · × Xn or by ni=1 Xi .
Suppose now that X1 , . . . , Xn are topological spaces. If Ui is an open set of Xi for each
i = 1, . . . , n,Qtheir product U1 × U2 × · · · × Un will be called an open rectangle. The product
topology on ni=1 Xi is defined as follows. A set O ⊂ X1 × X2 × · · · × Xn is open if whenever
x = (x1 , . . . , xn ) ∈ O, there exist open sets Ui ⊂ Xi such that
n
Y
x∈ Ui ⊂ O.
i=1

We need to check first that this is indeed a topology.

(i) The empty set is open because it contains no point. The whole space is open because if
(x1 , . . . , xn ) is a point in the product, then, we can take Ui = Xi .

(ii) Let (Oλ )λ∈L be a collection of open sets in ni=1 Xi and let (x1 , . . . , xn ) ∈ ∪λ∈L OλQ
Q
. Then,
(x1 , . . . , xn ) ∈ Oµ for some µ ∈ L. Since Oµ is open, there exists an open rectangle ni=1 Ui
containing (x1 , . . . , xn ) and contained in Oµ . It is therefore contained in ∪λ∈L Oλ .

(iii) Let O1 and O2 beQopen in ni=1 Xi and let x = (x1 , . . . , xn ) ∈ O1 ∩ O2 . Then, there is
Q
n
an
Qnopen rectangle i=1 Ui containing x and contained in O1 and there is an open rectangle
Qn Vi containing x and contained in O2 . Now, the intersection of these two open rectangles
i=1
is i=1 (Ui ∩ Vi ) and is again an open rectangle containing x and contained in O1 ∩ O2 .

Exercise. Prove that an open element of the product topology is a union of open rectangles.

Proposition 1.16. The product of open sets is an open set. The product of closed sets is a closed
set.
Proof. Another way to express theQfirst statement is the following: an openQrectangle is an
is trivial. Indeed, let ni=1 Ui be an open rectangle and let x ∈ ni=1 Ui . Then,
openQset. This Q
x ∈ ni=1 Ui ⊂ ni=1 Ui .
For the second statement, it is enough to prove that the product of two closed sets is a closed
set (then, the result follows by induction).
S Let A ⊂ X1 and B ⊂ X2 be closed sets. Then,
X1 × X2 − A × B = (X1 − A1 ) × X2 X1 × (X2 − B) which is the union of two open sets. □
In the exercises, you are asked to prove the following proposition.
1.6. PRODUCT SPACES 29

Proposition 1.17. Let (X1 , d1 ), . . . , (Xn , dn ) be n metric spaces. Let X = X1 × · · · × Xn . For

x = (x1 , . . . , xn ) and y = (y1 , . . . , yn ), define
n
X
ρ1 (x, y) = di (xi , yi )
i=1
n
!1/2
X
ρ2 (x, y) = di (xi , yi )2
i=1
ρ∞ (x, y) = max di (xi , yi ).
i=1...,n

Then, ρ1 , ρ2 and ρ∞ are metrics on X that generate the product topology on X.

30 CHAPTER 1. METRIC SPACES AND TOPOLOGICAL SPACES
Chapter 2

Continuous functions and

homeomorphisms

2.1 Continuous functions

Let f : (X, dX ) → (Y, dY ) be a function between two metric spaces and let x ∈ X. We say that
f is continuous at x if f (y) can be made arbitrarily close to f (x) for y sufficiently close to x. In
metric spaces the concept of a distance gives a meaning to closeness. This leads to the following
definition. We can think of ε as the accuracy in the approximation.

Definition. Let f : (X, dX ) → (Y, dY ) be a function between two metric spaces and let x ∈ X.
f is said to be continuous at x if the following condition holds.

∀ε > 0 ∃δ > 0 such that dX (x, y) < δ ⇒ dY (f (x), f (y)) < ε.

Remark. In the definition above, one can replace ”<” by ”≤”.

Examples. 1) A constant function is continuous (given ε > 0, any δ > 0 would do).
2) The identity function is continuous (given ε > 0, we can take δ = ε).
3) Let (X, d) be a metric space and a ∈ X. The function x 7→ d(x, a) is continuous at every point
because |d(x, a) − d(y, a)| ≤ d(x, y) (we can take δ = ε). More generally, if A ⊂ X, then, the
function x 7→ d(x, A) is continuous at every point because |d(x, A) − d(y, A)| ≤ d(x, y).
4) All the elementary functions of calculus are continuous on their domains. The elementary functions
are the rational functions (ratio of polynomials), the logarithm, the exponential, the trigonometric
and inverse trigonometric functions.

Proposition 2.1. Let f : X → Y be a function between two metric spaces and let x ∈ X. Then,
the following are equivalent.

(i) f is continuous at x.

(ii) ∀ε > 0, ∃δ > 0 such that f (B(x, δ)) ⊂ B(f (x), ε).

(iii) ∀V ∈ U(f (x)), ∃U ∈ U(x) such that f (U ) ⊂ V .

(iv) If {xn } is a sequence of X that converges to x then {f (xn )} converges to f (x) (sequential
continuity at x).

(v) If {xn } is a sequence of X that converges to x then there exists a subsequence (xnk ) of (xn )
such that {f (xnk )} converges to f (x).

31
32 CHAPTER 2. CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS

Proof. We will prove that (i) ⇒ (ii) ⇒ (iii) ⇒ (iv) ⇒ (v) ⇒ (i).
(i) ⇒ (ii). (ii) is just a reformulation of (i). Indeed, let ε and δ be as in the definition of continuity.
Let y ∈ f (B(x, δ)). Then y = f (z) for some z ∈ B(x, δ). This means that d(z, x) < δ and
so by the condition of continuity, d(f (z), f (x)) < ε, that is, d(y, f (x)) < ε. This means that
y ∈ B(f (x), ε).
(ii) ⇒ (iii). Let V be a neighborhood of f (x). Then there exists a ball B(f (x), ε) ⊂ V . Take now
U = B(x, δ). Then U is a neighborhood of x. Now (ii) can be written as f (U ) ⊂ B(f (x), ε) ⊂ V .
(iii) ⇒ (iv). Let (xn ) be a sequence of X that converges to x. Let V be a neighborhood of f (x).
Then there exists a neighborhood U of x such that f (U ) ⊂ V . Now convergence means that
xn ∈ U for all n large enough. Therefore f (xn ) ∈ f (U ) ⊂ V for n large enough. This means that
(f (xn )) converges to f (x).
(iv) ⇒ (v) is clear.
(v) ⇒ (i). Suppose that f is not continuous at x. Then there exists ε > 0 such that for all δ > 0
there exists y with d(y, x) < δ and d(f (y), f (x)) ≥ ε. Taking δ = n1 we get a sequence (yn )
converging to x and such that no subsequence of (f (yn )) converges to f (x), a contradiction. □

Definition. Condition (iii) involves only neighborhoods and not balls and distances. Therefore, we
take it as a definition of continuity of a map between topological spaces.

Definition. A function f : X → Y between two topological spaces is said to be continuous if it

is continuous at every point of X.

Theorem 2.1. Let f : X → Y be a function between two topological spaces. Then the following
are equivalent.
(i) f is continuous.
(ii) The inverse image under f of any open subset of Y is open in X.
(iii) The inverse image under f of any closed subset of Y is closed in X.

Proof. (i) ⇒ (ii). Let O ⊂ Y be open and let x ∈ f −1 (O) (so that f (x) ∈ O). Since O is a
neighborhood of f (x), there exists according to the previous proposition, a neighborhood U of x
such that f (U ) ⊂ O. Therefore U ⊂ f −1 (f (U )) ⊂ f −1 (O). This means that f −1 (O) is open.
(ii) ⇒ (i). Let x ∈ X and let V be an open neighborhood of f (x). Set U = f −1 (V ). Then U is a
neighborhood of x (being an open set containing x) and f (U ) = f (f −1 (V )) ⊂ V . According to the
previous proposition, f is a continuous at x. Since x was arbitrary, this proves that f is continuous.
The equivalence (ii) ⇔ (iii) follows from the fact that f −1 (Y \O) = X\f −1 (O). □

Corollary 2.1. Let X be a topological space and f : X → R be a continuous function. Let r ∈ R.

Then
(i) The sets {x ∈ X; f (x) < r} and {x ∈ X; f (x) > r} are open.
(ii) The sets {x ∈ X; f (x) ≤ r}, {x ∈ X; f (x) ≥ r} and {x ∈ X; f (x) = r} are closed.

Proof. This is because we can write {x ∈ X; f (x) < r} = f −1 (] − ∞, r[), {x ∈ X; f (x) ≤ r} =

f −1 (] − ∞, r]) and so on. □

Example. Since in a metric space x 7→ d(x, a) is continuous, an open ball is open and a closed
ball is closed. The set {x ∈ X; d(x, a) > r} is open, the set {x ∈ X; d(x, a) ≥ r} is closed. The
sphere {x ∈ X; d(x, a) = r} is closed.

Proposition 2.2. The following are true.

2.1. CONTINUOUS FUNCTIONS 33

(i) The composition of two continuous functions is continuous.

(ii) The restriction of a continuous function is continuous.

(iii) If A ⊂ X, the inclusion map j : A → X is continuous.

(iv) If f : X → Y is continuous, then f : X → f (X) is also continuous.

Proof. (i). Let f : X → Y and g : Y → Z be two continuous functions. We have to prove that
g ◦ f : X → Z is continuous. Observe that (g ◦ f )−1 (O) = f −1 (g −1 (O)) for every subset O ⊂ Z.
In particular, if O is open, then g −1 (O) is open in Y because g is continuous. Since f is continuous,
we have f −1 (g −1 (O)) is open. Hence the result.
(ii). Let f : X → Y be continuous and let A ⊂ X. We have to prove that f|A : A → Y is also

−1
continuous. Let O be open in Y . Then f|A (O) = A ∩ f −1 (O) is open in A.
(iii). Let O be open in X. Then j −1 (O) = A ∩ O is open in A.
(iv). Let O be open in f (X). Then O = O′ ∩ f (X) where O′ is open in Y . But f −1 (O) = f −1 (O′ )
and f −1 (O′ ) is open in X. □

Proposition 2.3. For real valued functions, the sum and product of continuous functions are
continuous. If f : X → R is continuous, then 1/f is continuous on its domain, i.e. on the set where
f does not vanish.
Proof. (i). Let f, g : X → R be continuous at some point x0 . Then, given ε > 0, there exists a
neighborhood U1 of x0 such that |f (x) − f (x0 )| < ε/2. Also, there exists a neighborhood U2 of x0
such that |g(x)−g(x0 )| < ε/2. Then for all x ∈ U1 ∩U2 , we have |f (x)+g(x)−(f (x0 )+g(x0 )| < ε.
(ii). Write f (x)g(x) − f (x0 )g(x0 ) = f (x)g(x) − f (x0 )g(x) + f (x0 )g(x) − f (x0 )g(x0 ). Then, by
the triangle inequality

|f (x)g(x) − f (x0 )g(x0 )| ≤ |g(x)||f (x) − f (x0 )| + |f (x0 )||g(x) − g(x0 )|

Let ε > 0 be given. Continuity of g at x0 implies that there exists a neighborhood U1 of x0 such
that |g(x) − g(x0 )| < 2(|f (xε0 )|+1) for all x ∈ U1 . This implies that |f (x0 )||g(x) − g(x0 )| < ε/2
for all x ∈ U1 . Continuity of g at x0 also dictates that |g(x)| < |g(x0 )| + 1 for all x in some
neighborhood U2 of x0 . Finally, continuity of f at x0 dictates that there exists a neighborhood U3
of x0 such that |f (x) − f (x0 )| < 2(|g(xε0 )|+1) for all x ∈ U3 . Therefore for x ∈ U1 ∩ U2 ∩ U3 , we
have |f (x)g(x) − f (x0 )g(x0 )| < ε.
(iii). Let x0 be a a point where f (x0 ) ̸= 0. Then |f (x0 )| > 0. Then, there exists a neighborhood
U1 of x0 such that |f (x) − f (x0 )| < |f (x0 )|/2. Then |f (x)| > |f (x0 )|/2 and so |f (x)||f (x0 )| >
|f (x0 )|2 /2 for all x ∈ U1 . Given ε > 0 there exists a neighborhood U2 of x0 such that |f (x) −
2
f (x0 )| < ϵ|f (x20 )| for all x ∈ U2 . It follows that if x ∈ U1 ∩ U2 then

1 1 |f (x) − f (x0 )|
− = < ε. □
f (x) f (x0 ) |f (x)f (x0 )|

Given two sets X and Y , the projection π1 : X × Y → X is defined by π1 (x, y) = x. The

projection π2 : X ×Y → X is defined by π2 (x, y) = y. Observe that if U ⊂ X then π1−1 (U ) = U ×Y
and if V ⊂ Y then π2−1 (V ) = X × V . Therefore we have

Proposition 2.4. Let X and Y be topological spaces. Then the projections on X and Y are
continuous (X × Y is equipped with the product topology).
34 CHAPTER 2. CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS

Proposition 2.5. Let X, Y, Z be topological spaces and let f : X → Y × Z. Let f1 and f2 denote
the components of f . Then f is continuous if and only if f1 and f2 are continuous.
Proof. Suppose first that f is continuous. Observe that if πY and πZ denote the projections into
Y and Z respectively, then f1 = πY ◦ f and f2 = πZ ◦ f . Since the projection are continuous, it
follows that f1 and f2 are continuous.
Suppose conversely that f1 and f2 are continuous. We show that the inverse image under f of
an open rectangle U × V is open in X. This follows from the fact that

f −1 (U × V ) = f1−1 (U ) ∩ f2−1 (V ). □

Proposition 2.6. Let f : X × Y → Z be continuous. Then the partial map f (·, y) : X → Z and
f (x, ·) : Y → Z are continuous.
Proof. Let g(x) = f (x, y) for y fixed. Let α(x) = (x, y). Then α : X → X × Y is continuous
(because each component is continuous). Since g = f ◦ α, it follows that that g is continuous.
Similarly for the partial map f (x, ·). □

Remark. If f : X × Y → Z and the partial maps f (·, y) : X → Z and f (x, ·) : Y → Z are

continuous, it does not follow that f is continuous. For example, let f : R × R → R be defined by
f (x, y) = x2xy
+y 2
if (x, y) ̸= (0, 0) and f (0, 0) = 0. Then the partial maps are continuous. However
f is not continuous at (0,0) because it has not limit there.

2.2 Homeomorphisms
The word homeomorphism comes from the Greek; ”morph” means form and ”homeo” means similar.
Thus, two spaces are homeomorphic if they have a similar form or shape. Intuitively, this means we
can deform one of them into the other without any cut and go back also without any cut. The idea
of a deformation that does not break the shape is captured by the concept of a continuous map.
Hence the following definition.
Definition. A function f : X → Y between two topological spaces is called a homeomorphism if
it is bijective, continuous and its inverse is also continuous. Two spaces are homeomorphic if there
exists a homeomorphism between them.
The concept of homeomorphism is fundamental in topology as is the concept of isomorphism in
algebra. Recall that two isomorphic groups have the same structure and therefore may be identified.
Similarly, two homeomorphic spaces have exactly the same topological properties and therefore
can be identified. A fundamental problem in topology is to decide whether two given spaces are
homeomorphic or not. This may be a hard problem.
Examples. 1) The function cos : [0, π] → [−1, 1] is a homeomorphism whose inverse is arccos :
[−1, 1] → [0, π]. Continuity of these functions is proved in the first year.
2) The function sin : [− π2 , π2 ] → [−1, 1] is a homeomorphism whose inverse is arcsin : [−1, 1] →
[− π2 , π2 ].
3) The function tan :] − π2 , π2 [→ R is a homeomorphism whose inverse is arctan : R →] − π2 , π2 [.
4) The function f : R → R defined by f (x) = x3 is a homeomorphism whose inverse is given by
f −1 (x) = x1/3 .
Remark. A continuous bijection need not be a homeomorphism. Here is an example. Let S 1
denote the unit circle. The function f : [0, 2π[→ S 1 given by f (t) = (cos t, sin t) is a continuous
2.2. HOMEOMORPHISMS 35

bijection. However, it is not a homeomorphism. Indeed, solving the system x = cos t; y = sin t, we
find ®
arccos x if y ≥ 0
f −1 (x, y) =
π + arccos(−x) if y < 0.
Then
lim f −1 (x, y) = arccos(1) = 0,
(x,y)→(1,0+)

whereas
lim f −1 (x, y) = π + arccos(−1) = 2π.
(x,y)→(1,0−)

So f −1 is not continuous. Note that in contrast, f :]0, 2π[→ S 1 \{(1, 0)} is a homeomorphism (the
limits above do not belong to the space ]0, 2π[).
Geometrically, the map f transforms (by bending) the interval [0, 2π[ into S 1 and transforms the
interval ]0, 2π[ into S 1 − {1, 0)}. Therefore we can go from [0, 2π[ to S 1 in a continuous way but
if we want to go back, we will have to cut S 1 . This is the idea behind discontinuity of the inverse.
In contrast, we can bend ]0, 2π[ into S − {(1, 0)} and we can go back by flattening it.

Note. A continuous bijection f : I → J between two intervals of R is automatically a homeomor-

phism. This will be proved in the exercises of chapter 3 and 4.

Now, we give more fundamental examples of homeomorphisms and homeomorphic spaces. Their
geometric significance should be clear but we write equations to eliminate any doubt.

More examples
1) All intervals of the real line of the form [a, b] with a < b are homeomorphic. Indeed, the map
x 7→ (x−a)/(b−a) is a homeomorphism between [a, b] and [0, 1]. This map is a translation followed
by a dilation or contraction.
2) All open intervals of the real line are homeomorphic. The map x 7→ (x − a)/(b − a) is a
homeomorphism between ]a, b[ and ]0, 1[. Therefore all intervals of the form ]a, b[ are homeomorphic
between each other. The function x 7→ tan x is a homeomorphism between ] − π/2, π/2[ and R
(this function represents an infinite stretching of the interval ] − π/2, π/2[). The interval ]a, +∞[ is
homeomorphic to R by x 7→ ln(x−a). The interval ]−∞, a[ is homeomorphic to R by x 7→ ln(a−x).
3) A circle and an ellipse are homeomorphic (for example by the map (x, y) 7→ (ax, by)).
4) Rotations, translations, homothecies and symmetries of the plane or space are homeomorphisms.
This should be intuitively clear but you can write equations for these maps.
5) The punctured disc {0 < x2 + y 2 < 1} and the annulus {1 < x2 + y 2 < 4} are homeomorphic.
X
For X = (x, y), let f (X) = X + ||X|| = ||X||+1
||X|| X. Then f
−1 (X) = ||X||−1 X. Then f is a
||X||
homeomorphism.
6) A circle and the boundary of a square are homeomorphic. To be specific consider the unit circle
S 1 defined by the equation x2 + y 2 = 1 and the square C, defined by the equation |x| + |y| = 1.
Let f : R2 → R2 be given by f (x, y) = (x|x|, y|y|). Then f is continuous because its components
√
are continuous. f is bijective
√ and its inverse is given by f −1 (x, y) = (g(x), g(y)) where g(x) = x
if x ≥ 0 and g(x) = − −x if x ≤ 0. It is easy to see that , g is continuous (draw the graph of g).
It follows that f −1 is also continuous. Thus, f is a homeomorphism.
Next, we prove that f (S 1 ) = C. Indeed, let first (x, y) ∈ S 1 , then x2 + y 2 = 1 and so
|x|x|| + |y|y|| = 1. This means that f (x, y) ∈ C and so f (S 1 ) ⊂ C. Conversely, let (u, v) ∈ C.
Then there exists (x, y) ∈ R2 such that f (x, y) = (x|x|, y|y|) = (u, v). Then x2 = |u| and y 2 = |v|.
Since |u| + |v| = 1, then x2 + y 2 = 1 i.e (x, y) ∈ S 1 .
36 CHAPTER 2. CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS

In fact, f maps the open unit disc onto the open square, and the exterior of the disc onto the
exterior of the square.
7) Equip R2 with the Euclidean distance

1/2
d((x, y), (x′ , y ′ )) = (x − x′ )2 + (y − y ′ )2

(but any other equivalent distance would do). Equip C with the distance

1/2
d(z, z ′ ) = |z − z ′ | = (x − x′ )2 + (y − y ′ )2 .

Then R2 is homeomorphic to C. Indeed, consider the function f : R2 → C defined by f (x, y) =

x + iy. Then f is a bijection. Next, f preserves the distances (we say that f is an isometry).
Therefore f is continuous. Since f −1 also preserves distances, it is continuous as well.
8) A circle with a point removed (or a punctured circle) is homeomorphic to R. Indeed, let S 1
denote the unit circle in R2 and let N denote the point with coordinates (0,1) (the north pole).
The stereographic projection with pole N associates to each point M ∈ S 1 − N the point M ′ which
is the point of intersection of the line (N M ) with the horizontal line {y = 0}.

M′
M′ x

If M has coordinates (x, y), then the point M ′ has coordinates ( 1−y
x
, 0). Indeed, let M ′ have
−→ −→
coordinates (x′ , 0). Since M ′ belongs to (N M ), there exists λ ∈ R such that N M ′ = λN M and
so (x′ − 0, 0 − 1) = λ(x − 0, y − 1). It follows that λ = 1/(1 − y) and so X = 1−y
x
. This defines a
1
continuous function F : S − N → R given by
x
F (x, y) = .
1−y
It inverse is given by
1
F −1 (t) = (2t, t2 − 1).
t2 +1
x 2t
Indeed, if t = 1−y , then t − x = ty. Squaring and using the fact that x2 + y 2 = 1, we get x = t2 +1
,
t2 −1
and then y = t2 +1
. It is clear that F −1 is continuous.

Geometric interpretation. F is projecting the punctured circle onto the real line and F −1 is
wrapping the real line onto the punctured circle.

Remark 1. There is another way to see that a circle with a point removed is a homeomorphic with
the the real line. We already know that S 1 − {(1, 0)} is homeomorphic with ]0, 2π[. But ]0, 2π[ is
homeomorphic with R. Therefore S 1 − {(1, 0)} and R are homeomorphic. Of course S 1 − {(1, 0)}
and S 1 − {(0, 1)} are homeomorphic by a rotation.
2.2. HOMEOMORPHISMS 37

Remark 2. For some mathematicians, the stereographic projection maps a circle to its tangent
opposite to the north pole N = (0, 2). This projection leads to the map F : S 1 − N → R
2x
F (x, y) = .
2−y

M′ x

9) A sphere with a point removed is homeomorphic to R2 . Let S 2 denote the unit sphere in R3
defined by x2 + y 2 + z 2 = 1 and let N denote the point with coordinates (0,0,1) (the north pole).
The stereographic projection with pole N associates to each point M ∈ S 2 − N the point M ′ which
is the point of intersection of the line (N M ) with the plane {z = 0}. If M has coordinates (x, y, z),
y
then the point M ′ has coordinates ( 1−z x
, 1−z , 0). Indeed, let M ′ have coordinates (x′ , y ′ , 0). Since
−→ −→
M ′ belongs to (N M ), there exists λ ∈ R such that P N = λN M and so (x′ − 0, y ′ − 0, 0 − 1) =
y
λ(x − 0, y − 0, z − 1). It follows that λ = 1/(1 − z) and so x′ = 1−z
x
and y ′ = 1−z . This defines a
2
homeomorphism F : S − N → R given by2
Å ã
x y
F (x, y, z) = ,
1−z 1−z
whose inverse is given by
1
F −1 (t, s) = (2t, 2s, t2 + s2 − 1).
t2 + s2 + 1

10) More generally the unit sphere S n (defined by x21 + x22 + · · · x2n+1 = 1) with the north pole
removed is homeomorphic to Rn via the stereographic projection
Å ã
x1 x2 xn
F (x1 , x2 , . . . , xn+1 ) = , ,··· , .
1 − xn+1 1 − xn+1 1 − xn+1
The inverse is given by
n
−1 1 X
F (X1 , X2 , . . . , Xn ) = Pn 2 (2X1 , 2X2 , · · · , 2Xn , Xi2 − 1).
i=1 Xi + 1 i=1

11) Choose a norm on Rn . Then, the open unit ball Dn := {x ∈ Rn ; ||x|| < 1} is homeomorphic
x
with the whole space Rn . One homeomorphism is given by f (x) = 1−||x|| . Another homeomorphism
π x


is given by g(x) = tan 2 ||x|| ||x|| for x ̸= 0 and g(0) = 0. Check that. These two homeomorphisms
represent an infinite stretching of the open unit ball; as x approaches the boundary, f (x) and g(x)
approach infinity.
38 CHAPTER 2. CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS

Imbeddings
Definition. Let X and Y be two topological spaces. A topological imbedding or simply an
imbedding of X in Y , is a map f : X → Y such that f : X → f (X) is a homeomorphism.
The existence of an imbedding of X in Y , means that we can place a copy of X inside Y .
Examples. 1) Let f : R → R2 be defined by f (x) = (x, 0). Then f is an imbedding. Here the
copy of R is a straight line.
2) The map g : R → R2 defined by g(x) = (x, x3 ) is an imbedding. Here the topological copy of
R is the curve of equation y = x3 .
3) The map g : R → R2 defined by g(x) = (e−x cos x, e−x sin x) is an imbedding. Here, the
topological copy of R is an infinite spiral.
1
4) The map g : [0, 1] →]0, 1[ defined by g(x) = 3 + 16 x is an imbedding.
Chapter 3

Compactness

• Compactness is a central concept in Topology and in modern Mathematical Analysis. It took

many years for mathematicians to formulate this concept. A topological space is said to be compact
if every open covering of X has a finite subcovering. Any interval [a, b] has this property. This was
discovered by Borel an Lebesgue in their work in measure theory. We may say that compactness is
a finteness condition: a compact space is topologically not too big.
• There is another related property which is easier to understand. Any sequence of [a, b] has a
convergent subsequence. This property was initially called compactness but ultimately it was called
sequential compactness. A topological space X is said to be sequentially compact if every sequence
of X has a convergent subsequence.
• Without any assumption on the space X, there is no relation between these two conditions.
However there are some interesting and important cases where these two conditions are equivalent.
This is the case when X is a metric space. This result is far from being trivial and it is rather a deep
theorem. In the graduate course on Functional Analysis, we will present another important situation
where the two conditions are equivalent. And again this equivalence is a deep theorem.

3.1 Definitions, examples and basic properties

Definition. Let X be a topological
S space. A collection of subsets (Oλ )λ∈L of X is called a
covering or cover of X if X = Oλ . The collection is called an open covering or open cover
λ∈L
if its elements are open.



Example 1. The collection [n, n + 1] n∈Z is a covering of R. The collection ]n, n + 1[ n∈Z is
not a covering of R.


Example 2. The collection ]n, n + 2[ n∈Z is an open covering of R.


Example 3. The collection ] − n, n[ n∈N is an open covering of R.

Example 4. The collection ] n1 , 1] n∈N∗ is an open covering of ]0, 1]. Recall that ] n1 , 1] is not

open in R but is open in ]0, 1].

Example 5. Let X be a metric space and let a ∈ X. The collection he (B(a, r))r>0 is an open
covering of X.


Example 6. Let Xbe a metric space and let α > 0. The collection B(x, α) x∈X is an open
covering of X.

Definition. Let (Oλ )λ∈L be a covering of a topological space X. A subcollection (Oλ )λ∈J that
also covers X is called a subcovering of the covering (Oλ )λ∈L (J ⊂ L).

39
40 CHAPTER 3. COMPACTNESS

Example 1. Consider the covering ] − x, x[ x∈R of R. Then the collection ] − x, x[ x∈N is a


subcovering of this covering. The collection ] − x, x[ x∈Q is also a subcovering.
Example
2. More generally, let X
be a metric space and let a ∈ X. The collection he
B(a, r) r∈N∗ is subcovering of B(a, r) r>0 .

Definition. A topological space X is said to be compact if every open covering of X has a finite
subcovering. Otherwise stated, if [
X= Oλ ,
λ∈L

where each Oλ is open in X, then there is a finite set J ⊂ L such that

[
X= Oλ ,
λ∈J

or equivalently there exist λ1 , . . . , λn ∈ L such that

n
[
X = Oλ1 ∪ · · · ∪ Oλn = Oλk .
k=1

Remark 1. The definition of compactness can also be rephrased as follows.

• X is compact if from every open covering of X, we can extract a finite subcovering.
• X is compact if every open covering of X contains a finite subcollection that also covers X.

Remark 2. An open covering of a space can be a very large collection often infinite. What
compactness means is that we don’t need all the elements of the collection to cover the space: a
few of them will do.

Remark 3. Instead of the indexed notation (Oλ )λ∈L or (Oi )∈I , coverings can be written in
calligraphic letters like A or B and elements of these covering are then written in capital letters like
A, B, O. So if A is covering of X, we can write
[
X= A.
A∈A

Example 1. Any finite space is compact. Indeed, letSX = {a1 , . . . , an } and let (Oλ )λ∈L be
an open covering of X. Let k = 1, . . . , n. Since ak ∈ λ∈L Oλ , there exists λk ∈ L such that
ak ∈ Oλk . It follows that
X = Oλ1 ∪ . . . ∪ Oλn .

Example 2. Consider X = {0} ∪ n ; n ∈ N∗ as a subspace of R. Then X is compact. Indeed

1

let (Oλ ) be an open covering of X. Then 0 belongs to some Oλ0 . Since ( n1 ) converges to 0, Oλ0
contains all n1 except finitely many terms 1, 21 , . . . , n10 . Now each i = 1, . . . , n0 , 1i belongs to some
open set Oλi . It follows that
X = Oλ0 ∪ Oλ1 ∪ · · · ∪ Oλn .
The argument that we used shows that more generally, if (xn ) is a sequence of a topological space
that converges to x, then the subspace {xn ; n ∈ N∗ } ∪ {x} is compact.


Example 3. The real line is not compact. Indeed, ] − n, n[ n∈N∗ is an open covering of R.


Suppose that there is a finite subcollection ] − n, n[ n∈J that covers R. Then
[
R= ]−n, n[ = ]−M, M [
n∈J
3.1. DEFINITIONS, EXAMPLES AND BASIC PROPERTIES 41

where M = max J, a contradiction. For a similar reason, RN is not compact.

Example 4. (0, 1] is not compact. Indeed, ] n1 , 1] n∈N∗ is an open covering of ]0, 1]. Suppose

that there is a finite subcollection ] n1 , 1] n∈J that covers ]0, 1]. Then

[ 1   1 
]0, 1] = ,1 = ,1 where m = max J.
n m
n∈J

Example 5. The subspace Y = n1 ; n ∈ N∗ of R is not compact. Indeed,

 1

n n∈N∗
is an
open covering of Y that has no finite subcollection that covers Y (because Y is infinite). Therefore,
removing a point from a compact space can destroy the property of compactness.
Remark. We said above that a compact space is in some sense not too ”big”. So how can
we understand that [0,1] is compact whereas ]0,1[ is not compact? Actually, ]0,1[ is topologically
big because it is homeomorphic to R whereas [0,1] is not that big. In ordinary language, compact
means small or reduced or tight. Therefore we may say that a compact space is topologically small
or reduced or tight.

Compactness can also be formulated in terms of closed sets.

Definition.
T A collection (Fλ )λ∈L of sets is said to have the finite intersection property if
λ∈J Fλ ̸= ∅ for every finite subcollection (Fλ )λ∈J .

Example. Set for λ ∈ R, Fλ = [λ, ∞[. Then the collection (Fλ )λ∈R has the finite intersection
property. Indeed, let J = {λ1 , . . . , λn } be a finite subset of R. Then
\
Fλ = Fλ1 ∩ Fλ2 · · · ∩ Fλn = [t, ∞[̸= ∅
λ∈J

where t = max(λ1 , . . . , λn ). Observe however that the whole intersection

\
Fλ = ∅
λ∈R

because there is no real number bigger than every other real number.

Proposition 3.1. Let X be a topological space. Then X is compact if and only if for every
collection
T (F )
λ λ∈L of closed sets of X having the finite intersection property, the whole intersection
λ∈L λF is nonempty.
T
Proof. Suppose that λ∈L Fλ ̸= ∅ for every collection (Fλ )λ∈L having the finite intersection
property. To show that X is compact, we consider an open covering (Oλ )λ∈L of X. This means
that [
X= Oλ .
λ∈L

By De Morgan’s law \
(X\Oλ ) = ∅.
λ∈L

This means that the collection of closed sets (X\Oλ )λ∈L has an empty intersection. According to
our assumption, it cannot have the finite intersection property. This means that there is a finite set
J ⊂ L such that \
(X\Oλ ) = ∅.
λ∈J
42 CHAPTER 3. COMPACTNESS

By De Morgan’s law again, [

X= Oλ .
λ∈J

The converse is left to you as an exercise to test your understanding. □

Corollary 3.1. Let X be a compact topological space and letT(Fn ) be a decreasing sequence of
nonempty closed subsets of X (that is, F1 ⊃ F2 ⊃ · · · ). Then Fn ̸= ∅.
̸
Proof. The collection (Fn ) has the finite intersection property because Fn1 ∩· · ·∩Fnk = Fmax ni =
∅. □

Now we prove some facts about subspaces.

Definition. Let Y be a subspace of a topological space X and let (Oλ )λ∈L be S a collection of
subsets of X. We say that (Oλ )λ∈L is a covering of Y (or that it covers Y ) if Y ⊂ λ∈L Oλ .
We have now a more general meaning of the notion of covering. There is no contradiction
between this definition and the first one. The first definition is a particular case of the second one.
Conversely if (Oλ )λ∈L Sis a covering of Y by subsets of Y according to the second definition, then,
on
S the one hand Y ⊂ λ∈L OSλ . On the other hand, by assumption Oλ ⊂ Y for all λ ∈ L and so
λ∈L Oλ ⊂ Y . Hence Y = λ∈L Oλ . This means that (Oλ )λ∈L is a covering of Y according to
the first definition.
Let us look at two slightly different conditions.

(i) Every covering of Y by open subsets of Y has a finite subcovering (this is the compactness
condition of Y ). In symbols
[
Y = Uλ ; Uλ open in Y =⇒ Y = Uλ1 ∪ · · · ∪ Uλn .
λ∈L

(ii) Every covering of Y by open subsets of X has a finite subcovering. In symbols

[
Y ⊂ Oλ ; Oλ open in X =⇒ Y ⊂ Oλ1 ∪ · · · ∪ Oλn .
λ∈L

Lemma 3.2. The two conditions above are equivalent.

Proof. We prove (ii) ⇒ (i) and leave the converse to you as an exercise. Let (Uλ )λ∈L be an open
covering of Y . By definition of the subspace topology, we can write Uλ = Y ∩ Oλ where Oλ is open
in X. Therefore [ [ [
Y = Uλ = Y ∩ Oλ = Y ∩ Oλ .
λ∈L λ∈L λ∈L

This implies that [

Y ⊂ Oλ .
λ∈L

Condition (ii) then dictates that

Y ⊂ Oλ1 ∪ · · · ∪ Oλn .
This implies that
Y = Y ∩ (Oλ1 ∪ · · · ∪ Oλn ) = Uλ1 ∪ · · · ∪ Uλn . □

Proposition 3.2. Every closed subspace of a compact space is compact.

Proof. Let Y be a closed subspace of a topological space X. Let (Oλ )λ∈L be a covering of Y
by open subsets of X. Add to this covering the open set X − Y . Otherwise stated let µ be an
3.1. DEFINITIONS, EXAMPLES AND BASIC PROPERTIES 43

element not in L and set Oµ = X − Y and consider the new collection (Oλ )λ∈L∪{µ} . Then this new
collection is an open covering of X. Compactness of X implies that a finite subcollection (Oλ )λ∈J
covers X and consequently also Y . Then the collection (Oλ )λ∈J−{µ} is a finite subcollection of
(Oλ )λ∈L that covers Y . □

Proposition 3.3. Every compact subset of a metric space is closed (in the space).
Proof. Let X be a metric space and let K ⊂ X be compact. We prove that X − K is open
by proving that it is a neighborhood of all its points. Let x ∈ X − K. For any y ∈ K, there is
S neighborhood Uy of x and an open neighborhood Vy of y such that Uy ∩ Vy = ∅. Now
an open
K⊂ Vy . Since K is compact, we can write K ⊂ Vy1 ∪ · · · ∪ Vyn . Then U = Uy1 ∩ · · · ∩ Uyn is
y∈K
a neighborhood of x not intersecting K. Therefore U ⊂ X − K. □

Let us establish some fundamental properties of compact spaces.

Proposition 3.4. Every compact metric space is bounded.

Proof. Fix a point a ∈ X and consider the collection (B(a, r))r>0 . This collection of open
sets coversSX. Compactness of X implies that finitely
Snmany balls B(a, r1 ), . . . , B(a, rn ) cover X,
n
i.e., X = k=1 B(a, rk ). But it is easy to see that k=1 B(a, rk ) = B(a, r) where r = max rk .
Therefore X = B(a, r) and so X is bounded. □

Theorem 3.1. The image of a compact space under a continuous map is compact. In particular
compactness is preserved by homeomorphisms. We say that compactness is a topological property.
Proof. Let f : X → Y be continuous with X compact. Let (Oλ )λ∈L be a covering of f (X) by
open subsets of Y . This means that [
f (X) ⊂ Oλ .
λ∈L

Therefore [
X = f −1 (f (X)) ⊂ f −1 Oλ .

λ∈L

Actually, we have equality, because f −1 (Oλ ) ⊂ X. Now observe that f −1 (Oλ ) is open because f is
continuous and Oλ is open in Y . This means that the collection (f −1 (Oλ ))λ is an open covering of
X. Compactness of X implies that finitely many subsets f −1 (Oλ1 ), · · · , f −1 (Oλn ) cover X, i.e.,
n
[
X= f −1 (Oλi ).
i=1

Therefore,
n
[ n

[
f (X) ⊂ f f −1 (Oλi ) ⊂ Oλi □.
i=1 i=1

Theorem 3.2. (Extreme value theorem). Let X be a topological space and let f : X → R be
continuous. If X is compact, then there exist points c and d in X such that f (c) ≤ f (x) ≤ f (d)
for every x ∈ X. Otherwise stated, a real valued continuous map has a maximum value and a and
minimum value on a compact space. We also say that a continuous map is bounded on a compact
space and achieves its upper an lower bounds.
Proof. Suppose that X is compact. Then f (X) is a compact subspace of R. Therefore it is
closed and bounded. Boundedness implies that f (X) has a lowest upper bound sup f (X) and a
greatest lower bound inf f (X). But we proved in the exercises that for a bounded subset A ⊂ R,
sup A ∈ A and inf A ∈ A. Therefore, if A is closed, it contains its sup and inf. Consequently,
44 CHAPTER 3. COMPACTNESS

sup f (X) ∈ f (X). This means that sup f (X) = f (d) for some d ∈ X. Similarly, inf f (X) = f (c)
for some c ∈ X. But inf f (X) ≤ f (x) ≤ sup f (X) for every x ∈ X. Hence the conclusion. □

Remark. sup f (X) is usually denoted by sup f (x) and inf f (X) is usually denoted by inf f (x).
x∈X x∈X

Here is a useful criterion involving compactness for proving that a continuous bijection is actually
a homeomorphism.

Theorem 3.3. Let f : X → Y be a continuous bijection between metric spaces. If X is compact,

then f is a homeomorphism.
Proof. We have to prove that f −1 is continuous and this is equivalent to proving that f maps
closed sets into closed sets (because f = (f −1 )−1 ). So let A ⊂ X be closed. By Proposition 3.2.,
A is compact. By Theorem 3.1., f (A) is compact. By Proposition 3.3., f (A) is closed in Y . □

3.2 Other forms of compactness

Theorem 3.4. (Total boundedness). Let X be a compact metric space. Then, for any ε > 0
n
S
there are finitely many points x1 , . . . , xn in X such that X = B(xi , ε). This property is called
i=1
total boundedness (in French précompacité).
Proof. Consider the open covering (B(x, ε))x∈X . Compactness implies that there exists a finite
subcollection {B(x1 , ε), . . . , B(xn , ε)} that covers X. This is the conclusion. □

Example/Exercise. Equip X =]0, 1[ with the usual distance. Let us show by going back to the
definition that X is totally bounded. Let ε > 0 be given. Choose an integer n bigger than 1ε and
n
i S
let xi = n+1 (i = 1, . . . , n). We claim that X = B(xi , ε). Here
i=1

B(xi , ε) = {x ∈ X; |x − xi | < ε} =]xi − ε, xi + ε[ ∩ ]0, 1[.

Let x ∈ X. Let j = ⌊nx + 1⌋. Check that 1 ≤ j ≤ n and that x ∈ B(xj , ε). This example shows
that a totally bounded metric space need not be compact.

Lemma 3.3. (The Lebesgue number lemma). Let (X, d) be a compact metric space and let
A be an open covering of X. If X is compact, then there exists δ > 0 such that each subset of X
having diameter less than δ is contained in some element of A. The number δ is called a Lebesgue
number of the covering A.
Proof. If X is an element of A, then any positive number is a Lebesgue number of A. So we
assume that X ∈
/ A. Choose a finite subcollection {A1 , . . . , An } of A that covers X. For each i,
set Ci = X − Ai and let
n
1X
f (x) = d(x, Ci ).
n
i=1

Then f (x) > 0 for all x ∈ X. Indeed, otherwise there exists x ∈ X such that d(x, Ci ) = 0 for all i.
Since Ci is closed, we get x ∈ Ci . Thus
n
\ n
\ n
[
x∈ Ci = (X − Ai ) = X − Ai = ∅, a contradiction.
i=1 i=1 i=1
3.2. OTHER FORMS OF COMPACTNESS 45

Since f is continuous, it has a minimum value δ > 0. We show that δ is the required number. Let
B be a subset of X of diameter less than δ. Choose a point x0 ∈ B. Then B ⊂ B(x0 , δ). Now, if
d(x0 , Cm ) is the largest of the numbers d(x0 , Ci ), then δ ≤ f (x0 ) ≤ d(x0 , Cm ). It follows that

B(x0 , δ) ⊂ B(x0 , d(x0 , Cm )) ⊂ X − Cm = Am

because if d(x0 , x) < d(x0 , Cm ), then x ∈

/ Cm . Consequently, B ⊂ Am . □

Definition. A topological space is said to be limit point compact if every infinite subset of X
has a limit point. A topological space is said to be sequentially compact if very sequence in X
has a convergent subsequence.
It turns out that if X is a metric space, then the three conditions of compactness are equivalent.

Theorem 3.5. Let X be a metric space. Then the following conditions are equivalent.

(i) X is compact.

(ii) X is limit point compact.

(iii) X is sequentially compact.

Remark. The implication (i) ⇒ (ii) is true for any topological space X. It follows from this
implication that a compact and discrete space is necessarily finite. This is an important fact in
Analysis.
Before we prove Theorem 3.5., let us deduce an important consequence.

Corollary 3.1. A finite product of compact metric spaces is compact.

Proof. It is enough to prove the result for the product of two metric spaces because the general
result follows by induction on the number of spaces. So let (X, dX ) and (Y, dY ) be two metric
spaces. Here X × Y is equipped with the product topology. We may equip X × Y with the distance

ρ (x, y), (x′ , y ′ ) = dX (x, x′ ) + dY (y, y ′ )

but this is irrelevant because compactness is a topological property and not a metric property. All
that we need to know about this topology is that (xn , yn ) → (x, y) in this topology if and only if
xn → x in X and yn → y in Y ; and this follows from the definition of the product topology (check
that).
Let ((xn , yn )) be a sequence of X ×Y . Then (xn ) is a sequence of X. Sequential compactness of
X implies that (xn ) has a convergent subsequence (xnk ). Sequential compactness of ÄY implies that


ä
the sequence (ynk ) has a convergent sequence ynkℓ . It follows that the subsequence xnkℓ , ynkℓ
is convergent in X × Y . □

Remark. The above result holds for the product of arbitrary topological spaces, but the proof is
harder.

Proof of Theorem 3.5.

(i) ⇒ (ii). Let A ⊂ X be infinite. Suppose by contradiction that no point of X is a limit point of
A.
Claim. Each point x ∈ X has an open neighborhood Vx intersecting A in at most one point.
46 CHAPTER 3. COMPACTNESS

Indeed, let x ∈ X. If x ∈
/ A, then by Proposition 1.8., there exists an open neighborhood Vx of
x not meeting A. If x ∈ A, since x is not a limit point of A, it must be an isolated point of A and
so there exists a neighborhood Vx of x such that Vx ∩ A = {x}. In both cases, Vx ∩ A ⊂ {x}.
S Sn
Sn we can write X = x∈X Vx . Compactness of X implies that X = i=1 Vxi . Therefore
Now,
A ⊂ i=1 Vxi . It follows that
n
[ n
[ n
[ 
A=A∩ Vxi = A ∩ V xi ⊂ xi .
i=1 i=1 i=1

This means that A is finite, contrary to our assumption.

(ii) ⇒ (iii). Let (xn ) be a sequence in X. Let E = {x1 , x2 , . . . , xn . . .} be the range of (xn ).
Suppose first that E is finite.
Claim. (xn ) contains a constant and therefore a convergent subsequence.
Indeed let E = {a1 , . . . , aℓ }. So for any n ∈ N∗ , xn takes some value ai . Therefore we can write
ℓ
[
n ∈ N∗ ; xn = ai .

N=
i=1

Since N∗ is infinite, at least one of these sets say {n ∈ N∗ ; xn = aj } is infinite. This means that
xn takes the value aj for infinitely many indices n and this is equivalent to saying that there exists
an increasing sequence of integers n1 < n2 < . . . < nk < · · · such that xnk = aj . This proves the
claim.
Suppose now that E is infinite. Then by assumption, E has a limit point x. In particular B(x, 1)
1


meets E at some point say xn1 . Next, B x, 2 ∩ E is infinite. In particular,

1

B x, ∩ E − xi ; i ≤ n1 ̸= ∅.
2
Choose accordingly an element in the above set; it is necessarily of the form xn2 with n2 > n1 . We
continue in this way. At the k th step, we choose an element in
1

B x, ∩ E − xi ; i ≤ nk−1 ̸= ∅.
k
By construction xnk ∈ B x, k1 . Therefore, xnk → x as k → ∞.

(iii) ⇒ (i). We prove this implication in three steps.

Step 1. We prove that if X is sequentially compact then the Lebesgue number lemma holds for
X. So let A be an open covering of X. Suppose that there is no Lebesgue number δ for this
covering. Then for each n ∈ N∗ , there exists a set Cn of diameter less than n1 and not contained
in any element of A. For each n, choose a point xn in Cn . By hypothesis, the sequence (xn )
has a subsequence (xnk ) that converges to some point x. Now x belongs to some element A of
the covering. Because A is open, there exists r > 0 such that B(x, r) ⊂ A. Now for all k large
enough, we have n1k < r/2. Then Cnk ⊂ B(xnk , r/2) (because the diameter of Cnk is less than
r/2). Also, for all k large enough, we have d(xnk , x) < r/2 and so by the triangle inequality
B(xnk , r/2) ⊂ B(x, r) ⊂ A. Thus, if k is large enough, then Cnk ⊂ A. But this contradicts our
hypothesis.
Step 2. We prove that if X is sequentially compact then it is totally bounded. So let ε > 0 be
given. Suppose that X cannot be covered by finitely many open ε−balls. We construct a sequence
(xn ) of X as follows. Let x1 be an arbitrary point in X. Since B(x1 , ε) does
 not cover X, choose
a point x2 ∈/ B(x1 , ε). Suppose that x1 , . . . , xn are constructed. Since B(x1 , ε), . . . , B(xn , ε)
3.3. COMPACT SUBSETS OF Rn 47

does not cover X, choose a point xn+1 not in the union of these balls. Then d(xn , xm ) ≥ ε for
all m ̸= n. This means that (xn ) cannot contain any convergent subsequence, contradicting our
assumption.
Step 3. We prove the implication (iii) ⇒ (i). Let A be an open covering of X. By the first
step, there exists a Lebesgue number δ > 0 for A. Let ε = δ/3. By the second step, there exist
finitely many B(x1 , ε), . . . , B(xn , ε) that cover X. Each of these balls B(xi , ε) has diameter at most
2ε = 2δ/3. Therefore it belongs to an element Ai of A. Then the finite collection {A1 , . . . , An }
covers X. □

3.3 Compact subsets of Rn

Theorem 3.6. Any closed interval [a, b] of the real line is compact.
S
Proof. Let [a, b] ⊂ Oλ where each Oλ is open in R. We need to prove that [a, b] is covered
λ∈L
by finitely many Oλ . We divide the proof into four steps.
Step 1. We show that if x ∈ [a, b[, then there is y ∈]x, b] such that [x, y] is covered by exactly
one Oµ . Thus let x ∈ [a, b[. Then x ∈ ∪λ∈L Oλ , and so there is an index µ ∈ L such that x ∈ Oµ .
But since Oµ is open, we can find an r > 0 such that ]x − r, x + r[⊂ Oµ . Let ε = min(b − x, r).
Choose now an element y ∈]x, x + ε[. Then y ∈]x, b] and [x, y] ⊂]x − ε, x + ε[⊂ Oµ . Thus, the
claim is proved.
Step 2. Let C be the set of all y ∈]a, b] such that [a, y] can be covered by finitely many Oλ .
Applying Step 1 to x = a, we see that a ∈ C and so C =
̸ ∅. Since C is bounded from above by b,
C has a least upper bound c ∈]a, b].
Step 3. We prove that c ∈ C. There is k ∈ L such that c ∈ Ok . Since Ok is open, there is δ > 0
such that ]c − δ, c + δ[⊂ Ok . By the property of the least upper bound, there is an element z ∈ C
such that c − δ < z ≤ c. Then [z, c] ⊂ Ok . Also, [a, z] can be covered by finitely many Oλ since
z ∈ C. Thus, [a, c] = [a, z] ∪ [z, c] can also be covered by finitely many Oλ .
Step 4. We show that c = b and the proof is finished. Suppose that c < b. Applying Step 1 to
x = c, we see that there is y ∈]c, b] such that [c, y] can be covered by one element Oj . But then,
[a, y] = [a, c] ∪ [c, y] can be covered by finitely many elements of the cover. This means that y ∈ C
and accordingly y should not be greater than c. We thus got a contradiction. □

Recall that we defined three topologically equivalent distances on Rn :

n
X
ρ1 (x, y) = |xi − yi |
i=1
n
!1/2
X
ρ2 (x, y) = |xi − yi |2 the Euclidean metric
i=1
ρ∞ (x, y) = max |xi − yi | the square metric.
i=1...,n

We also noticed that √

ρ∞ (x, y) ≤ ρ2 (x, y) ≤ nρ∞ (x, y).
and
ρ∞ (x, y) ≤ ρ1 (x, y) ≤ nρ∞ (x, y).
Recall that a subset A of a metric space X is said to be bounded if it is contained in a ball, or
equivalently if its diameter is finite. The inequalities between the usual distances on Rn imply that
boundedness in ρ2 is equivalent to boundedness for ρ∞ or boundedness for ρ1 .
48 CHAPTER 3. COMPACTNESS

Theorem 3.7. A subspace A of Rn is compact if and only if it is closed and bounded (in any of
the usual metrics).
Proof. Suppose first that A is compact. By Proposition 3.3, A is closed because Rn is a metric
space. By Proposition 3.4., A is bounded.
Conversely, suppose that A is closed and bounded for ρ∞ . Then ρ∞ (0, x) ≤ M for every x ∈ A.
This means that |xi | ≤ M for every x = (x1 , . . . , xn ) ∈ A and so A ⊂ [−M, M ]n . This last set is
compact as a product of compact spaces. Being closed, A is also compact. □

Remark. It is important to precise the distance for which a subspace is bounded. If d and d′ are
topologically equivalent distances, then boundedness for d′ does not imply boundedness for d. Here
is an example. Define on R, d′ (x, y) = min(1, |x−y|). Then d′ is a distance topologically equivalent
to the usual distance d on R. However, R is bounded for d′ and unbounded for d. Therefore, R
is closed and bounded in d′ , however, it is not compact (because (R, d) is not compact). As a
convention, when we do not mention the distance on Rn , it means that we consider any of the usual
distances.

Example 1. The unit sphere S n−1 and the closed ball B n in Rn are compact because they are
closed and bounded.
Example 2. The set A = (x, x1 ); 0 < x ≤ 1 is closed in R2 . However it is not compact because


it is not bounded.
Example 3. The set S = (x, sin x1 ); 0 < x ≤ 1 is bounded but not compact because it is not


closed. Can you find its closure?

3.4 Compactness and uniform continuity

Definition. A function f from a metric space (X, dX ) to a metric space (Y, dY ) is said to be
uniformly continuous if for every ε > 0, there exists δ > 0 such that for every points x, y in X,
dY (f (x), f (y)) < ε whenever dX (x, y) < δ. In symbols

∀ε > 0 ∃δ > 0 ∀x, y ∈ X dX (x, y) < δ ⇒ dY (f (x), f (y)) < ε.

Compare this definition with the definition of continuity

∀x ∈ X ∀ε > 0 ∃δ > 0 ∀y ∈ X dX (x, y) < δ ⇒ dY (f (x), f (y)) < ε.

The number δ in the condition of continuity depends on ε and x, whereas in the uniform continuity
condition it depends only on ε, so it the same for all x. Hence the adjective uniform.

Remark. We can replace ’< δ’ by ’≤ δ’ and ’< ε’ by ’≤ ε’.

Example 1. Consider the function f : ]0, 1] → R defined by f (x) = x1 . We know that f is
δ
continuous. However, it is not uniformly continuous. Indeed, let δ > 0. Take x = 1+δ 2 and
2δ δ 2
1+δ
y = 1+δ 2 . Then x, y ∈]0, 1], |x − y| = 1+δ 2 < δ and |f (x) − f (y)| = 2δ ≥ 1. Therefore we have
proved
∃ε = 1 > 0 ∀δ > 0 ∃x, y |x − y| < δ but |f (x) − f (y)| ≥ 1.

Example 2. The function sin is uniformly continuous. This follows from the fact that | sin x −
sin y| ≤ |x − y| for every x, y ∈ R (one can take δ = ε in the condition of uniform continuity.

Exercise 1. Let f : X → Y be a map between two metric spaces. Prove that the following
conditions are equivalent.
3.4. COMPACTNESS AND UNIFORM CONTINUITY 49

(i) f is uniformly continuous.

(ii) For any any two sequences (xn ) and (yn ) of X, we have

d(xn , yn ) → 0 ⇒ d(f (xn ), f (yn ) → 0.

Therefore to prove that a function is not uniformly continuous, it is enough to construct two sequence
such that d(xn , yn ) → 0 and d(f (xn ), f (yn )) does not tend to 0.

Exercise 2. (a) Prove that the map f : R → R defined by f (x) = x2 is not uniformly continuous.
(b) Prove that the map f : R → R defined by f (x) = sin x2 is not uniformly continuous.

Now, we give a positive result.

Theorem 3.8. (Uniform continuity theorem). Let X and Y be metric spaces with X compact.
If f : X → Y is continuous, then f is uniformly continuous.
First proof. Let ε > 0 be given. Consider the open covering of Y by the balls B(b, ε/2), b ∈ Y .
Let A be the collection of f −1 (B(b, ε/2)). Then A is an open covering of X (the covering is
open because f is continuous). Let δ be a Lebesgue number of A. Let x, y be two points in X
such that d(x, y) < δ. Then the set {x, y} has diameter less that δ. Therefore it is contained in
some open set f −1 (B(b, ε/2)). This means that f (x), f (y) ∈ B(b, ε/2). By the triangle inequality
d(f (x), f (y)) < ε. □

Second proof. Suppose that f is not uniformly continuous. Then

∃ε0 > 0 ∀δ > 0 ∃x, y d(x, y) < δ but d (f (x), f (y)) ≥ ε0 .

Taking δ = 1, 21 , 31 , . . ., we get two sequence (xn ) and (yn ) in X such that

1
d(xn , yn ) < and d (f (xn ), f (yn )) ≥ ε0 .
n
Sequential compactness of X implies that (xn ) has a subsequence (xnk ) that converges to some
point x. Now, by the triangle inequality
1
d(ynk , x) ≤ d (ynk , xnk ) + d (xnk , x) < + d (xnk , x) .
nk

It follows that ynk → x as well. Continuity of f at x dictates that f (xnk ) → f (x) and f (ynk ) →
f (x). By the triangle inequality d (f (xnk ), f (ynk )) → 0 as k → ∞. But his contradicts the
inequality
d (f (xnk ), f (ynk )) ≥ ε0 . □
50 CHAPTER 3. COMPACTNESS
Chapter 4

Connectedness

4.1 Definitions, examples and properties

Connectedness is an important concept in topology. Intuitively, a set is connected if it cannot be
split into more than one part (it is topologically one piece). More precisely we have the following
definition.

Definition. Let X be a topological space. A separation of X is a pair {U, V } of nonempty

disjoint open subsets of X whose union is X. The space X is said to be connected if there does
not exist a separation of X.

Remarks. 1) If {U, V } is a separation of X, then V = X − U and so U and V are clopen.

2) A separation of a space is a partition of that space. However, not every partition is a separation;
we can always write X = A ∪ (X − A), but the point in the definition above is that A is clopen.

Examples.
 1) The space X = {x ∈ R; |x| > 1} (as a subspace of R) is not connected since
] − ∞, −1[, ]1, ∞[ is a separation of X.
2) The space R∗ is not connected since ] − ∞, 0[, ]0, ∞[ is a separation of R∗ .


3) We shall prove that R and all its intervals are connected.

4) The empty set is connected.
5) A singleton is connected.
6) The only nonempty connected subsets of Q are the singletons. Indeed, let Y ⊂ Q and suppose
that Y containstwo rational numbers a and b. Then, there exists an irrational number c between
a and b. Then, Y ∩] − ∞, c[, Y ∩]c, ∞[ is a separation of Y .
7) Let X be a set equipped with the indiscrete topology {∅, X}. Then, X is connected (there is
not enough open sets to separate X).

Proposition 4.1. Let X be a topological space. Then the following conditions are equivalent.

(i) X is not connected.

(ii) There exist two disjoint nonempty closed subsets of X whose union is X.

(iii) There exists a proper clopen subset of X.

(iv) There exists a non constant continuous function f : X → Z.

51
52 CHAPTER 4. CONNECTEDNESS

Proof. (i) ⇒ (ii). Let {A, B} be a separation of X. Then A is also closed because its complement
B is open. Similarly B is closed.
(ii) ⇒ (iii). If X = A ∪ B where A and B are closed disjoint nonempty subsets, then A and B are
also open. So A is a proper clopen subset of X.
(iii) ⇒ (iv). Let A be a proper clopen subset of X. Let f (x) = 1 if x ∈ A and f (x) = 0 if
x ∈ X − A (f is the characteristic function of A). Then, f is continuous because the inverse image
of a subset O of Z is either ∅ or A or X − A or X. In all cases, it is open. Now clearly f is not
constant.
(iv) ⇒ (i). Let a and b be two distinct values taken by a continuous non constant function f :
X → Z. Now {a} is open in Z. Therefore f −1 (a) is open in X and nonempty. On the other hand,
Z\{a} is also open and nonempty (because it contains b). Therefore, f −1 (Z\{a}) is open in X
and nonempty. Now, we can write

X = f −1 (a) ∪ f −1 (Z\{a}) where f −1 (a) ∩ f −1 (Z\{a}) = ∅. □

Corollary 4.1. Let X be a topological space. Then the following conditions are equivalent.
(i) X is connected.
(ii) There do not exist two disjoint nonempty closed subsets of X whose union is X.
(iii) The only clopen subsets of X are X and ∅.
(iv) If f : X → Z is continuous, then f is constant.

Remark. Let X be a topological space and let Y ⊂ X. We say that Y is connected if it is

connected with respect to the subspace topology. We have also the following characterization.
Lemma 4.1. Let X be a topological space and let Y ⊂ X. Then Y is not connected if and only
if there exist two nonempty sets A and B such that A ∪ B = Y and A ∩ B = A ∩ B = ∅.
Proof. Suppose first that Y is not connected, then there exist two disjoint nonempty sets A and
B which are clopen in Y and whose union is Y . We proved in the exercises that the closure of A in
Y is A ∩ Y where A is the closure of A in X. Since A is closed in Y , we have A = A ∩ Y . Then
A ∩ B = A ∩ B ∩ Y = A ∩ B = ∅. A similar argument shows that B ∩ A = ∅.
Now suppose conversely that A and B are two nonempty sets whose union is Y and such that
A ∩ B = A ∩ B = ∅. Then we claim that A ∩ Y = A and B ∩ Y = B. Indeed, first A ⊂ Y ∩ A.
Second, let x ∈ A ∩ Y . If x ∈/ A, then x ∈ B since A ∪ B = Y . Then x ∈ A ∩ B = ∅, a
contradiction. Therefore x ∈ A. The other equality is proved similarly. The claim means that A
and B are two disjoint nonempty closed subsets of Y whose union is Y . Thus Y is not connected.
□
√
Example.√ Consider Q as a subspace of R. We can write Q = A ∪ B where√A = Q∩] − ∞, √ 2[ and
B = Q∩] 2, +∞[. So {A, B} is a separation of Q. Note √that A =] − ∞, 2] and B = [ 2, +∞[.
So A ∩ B = A ∩ B = ∅. Note however that A ∩ B = { 2} = ̸ ∅.

The next lemma will be useful.

Lemma 4.2. If {C, D} is a separation of a topological space X and Y is a connected subspace of

X, then either Y ⊂ C or Y ⊂ D.

Proof. Since C and D are both open in X, the sets C ∩ Y and D ∩ Y are open in Y . These
two sets are disjoint and their union is Y . If they were both nonempty, they would constitute a
separation of Y . Therefore one of them is empty. Hence Y is contained in C or D. □
4.1. DEFINITIONS, EXAMPLES AND PROPERTIES 53

Theorem 4.1. Let X be topological space. The union of connected subspaces of X that have a
common point is connected.
T
Proof.S Let (Aλ ) be a collection of connected subspaces of X and let p ∈ Aλ . Suppose that
Y := Aλ is not connected. Then, there exists a separation {C, D} of Y . The point p is in one of
the sets C or D. We can assume that it is in C (otherwise relabel C and D). Since Aλ is connected,
then by lemma 4.2., it must be contained in either C or D; and it cannot be contained inSD because
it contains the point p of C. Hence Aλ ⊂ C. Since λ was arbitrary, we have Y = Aλ ⊂ C,
contradicting the fact that D is not empty. □

Theorem 4.2. Let A be a connected subspace of a X. If A ⊂ B ⊂ A, then B is also connected.

Said differently: if B is obtained from a connected subspace A by adjoining some or all its limits
points, then B is connected.
Proof. Suppose that {C, D} is a separation of B. By Lemma 4.2., A must lie either in C or D.
Suppose that A ⊂ C. Then B ⊂ A ⊂ C. But C ∩ D = ∅, therefore B cannot intersect D. This
contradicts the fact that D is a nonempty subset of B. □

Theorem 4.3. The image of a connected space under a continuous map is connected. In particular,
connectedness is preserved by a homeomorphism.
Proof. Let f : X → f (X) be continuous with X connected. Suppose that {A, B} is a separation
of f (X). Then f −1 (A) and f −1 (B) are open (because f is continuous and A, B are open) and
disjoint because A ∩ B = ∅. They are also nonempty because A and B are contained in the image
of f . Therefore {f −1 (A), f −1 (B)} is a separation of X, contradicting the fact that X is connected.
□

Theorem 4.4. A product of connected spaces is connected.

Proof. We prove the theorem for the product of two spaces X and Y . The result for an arbitrary
number of spaces will follow by induction.
Choose a base point (a, b) ∈ X × Y .
The ”horizontal slice” X × {b} is connected because it is home-
Y {x} × Y
omorphic with X. Also, each vertical slice {x} × Y is con-
nected, being homeomorphic with Y . As a result, the space (a, b)
b X × {b}
Tx = (X × {b}) ∪ ({x} × Y ) is connected as a union of two con-
nected subspaces
S having the point (x, b) in common. Now form
the union x∈X Tx of all these subspaces. This union is connected
because these subspaces have the point (a, b) in common. But
x a X
[
X ×Y = Tx .
x∈X

Therefore X × Y is connected. □

Theorem 4.5. A subset E ⊂ R is connected if and only if it is an interval.

Proof. Let E be connected and assume it is not an interval. Then, there are two points x and y
in E and z ∈]x, y[ such that z ∈
/ E. Then, E has the following decomposition
[
E = E ∩ ] − ∞, z[ E ∩ ]z, ∞[.

The two subsets are both open in E and nonempty since the first one contains x and the second
contains y. This means that E is not connected contrary to the assumption.
Conversely, let E be an interval and suppose that it is not connected. Then it is the union
of two disjoint nonempty sets A and B each of which is open in E. Choose a ∈ A and b ∈ B.
54 CHAPTER 4. CONNECTEDNESS

f
f (b)

f (a)

a b
X

We may assume that a < b (otherwise we relabel A and B). The interval [a, b] is contained in E
since E is an interval. Hence [a, b] is the union of the disjoint nonempty sets A0 = A ∩ [a, b] and
B0 = B ∩ [a, b]. Let c = sup A0 . We distinguish between two cases.
Case 1. c ∈ A0 . In this case, c ̸= b, because otherwise c ∈ A0 ∩ B0 . So either c = a or a < c < b.
In both cases, because A0 is open in [a, b], there is some interval [c, c + ε[ contained in A0 . In
particular, this means that there is z ∈ A0 such that c < z < c + ε. The fact that z ∈ A0 means
that z ≤ sup A0 = c. Thus, we reached a contradiction.
Case 2. c ∈ B0 . In this case, c ̸= a, so either c = b or a < c < b. In either case, it follows from
the fact that B0 is open in [a, b] that there is some interval of the form ]c − ε, c] contained in B0 .
But since c = sup A0 , the interval ]c − ε, c] contains an element of A0 . Again this is a contradiction
since A0 ∩ B0 = ∅. □

Theorem 4.6. (Intermediate value theorem). Let X be a connected topological space and let
f : X → R be a continuous function. If a and b are two points of X and r is a number between
f (a) and f (b), then there exists a point c ∈ X such that f (c) = r.
Proof. Since X is connected and f is continuous, then f (X) is connected and therefore an
interval. We can suppose that f (a) < f (b). Since f (a) and f (b) belong to the interval f (X),
[f (a), f (b)] ⊂ f (X). Hence any number between f (a) and f (b) is the image of an element of X.
□

4.2 Path connectedness

Now we come to a condition stronger than connectedness: path connectedness.

Definition. Given two points x and y of a topological space X, a path in X from x to y is a

continuous map f : [a, b] → X such that f (a) = x and f (b) = y. A space X is said to be path
connected if any two points of X can be joined by a path in X.
Note that since [a, b] is homeomorphic to [0, 1], the domain of any path can be taken as [0, 1]. This
is usually done in algebraic topology.

Examples. 1) Intervals of R are path connected.

2) A ball B(a, r) in RN is path connected. In fact it is convex, which means that any two points
in B can be joined by a straight line segment. Indeed let x and y be two points in B(a, r). Set
f (t) = (1 − t)x + ty for t ∈ [0, 1]. Then f is a path joining x to y which is contained in B(a, r).
Indeed
||f (t) − a|| = ||(1 − t)(x − a) + t(y − a)|| < (1 − t)r + tr = r.
Similarly, a closed ball is path connected.
3) The punctured Euclidean space is RN \{0}. If N > 1, then this space is path connected. Indeed,
let x and y be two points in this space. If the straight line segment joining them does not pass by
4.2. PATH CONNECTEDNESS 55

0, then this segment is a path in RN \{0} joining x to y. Otherwise, we can choose a point z not
on the line joining x to y. Then x and y can be joined by the broken line path from x to z and then
from z to y.
4) The continuous image of a path connected space is also path connected (why?). In particular,
the unit sphere S n−1 in Rn is path connected for n > 1. Indeed, let g : Rn \{0} → S n−1 be defined
x
by g(x) = ||x|| . Then g is continuous and surjective. Therefore the result follows from the previous
example.

Proposition 4.2. A path connected space is connected.

Proof. Let X be a path connected space. Suppose that {A, B} is a separation of X. Let x ∈ A
and y ∈ B. Then there exists a path f : [a, b] → X joining x to y. Now f ([a, b]) is connected as
the image of a connected space under a continuous function. It follows that f ([a, b]) is contained
in either A or B, which is impossible since f (a) = x ∈ A and f (b) = y ∈ B. □

The converse of this proposition is however not true as the following example shows.
Example. Let S = (x, sin x1 ), 0 < x ≤ 1 .


Then S is connected as the image of ]0, 1] under a continuous function. Therefore its closure S
is also connected. S is called the topologist’s sine curve. In fact

S = S ∪ {0} × [−1, 1].

Let us prove this.

⊃ Since S ⊃ S, we need to prove that S ⊃ {0} × [−1, 1]. Let (0, u) ∈ {0} × [−1, 1]. Then,
1
u = sin v for some v. Then the sequence ( v+2nπ , sin(v + 2nπ)) belongs to S and converges to
(0, u). It follows that (0, u) ∈ S.
⊂ Let (x, y) ∈ S − {0} × [−1, 1]. Then x > 0. If y ̸= sin x1 , separate y and sin x1 by two
neighborhoods V1 and V2 . Let g(x) = sin x1 and let U be a neighborhood of x such that g(U ) ⊂ V2
(g is continuous). Then U × V1 is a neighborhood of (x, y) that does not intersect S, contradicting
the fact that (x, y) ∈ S.
We now show that S is not path connected. Suppose there is a path f : [a, c] → S joining the
origin to a point in S. The set {t ∈ [a, c]; f (t) ∈ {0} × [−1, 1]} is closed and bounded from above.
Therefore it has a biggest element b. Then f : [b, c] → S is a path that maps b into the vertical
slice {0} × [−1, 1] and maps the other points of [b, c] to points of S. Let f (t) = (x(t), y(t)). Then
x(b) = 0 while x(t) > 0 and y(t) = sin(1/x(t)) for t > b. By continuity, limt→b+ x(t) = x(b) = 0
1
and so limt→b+ x(t) = +∞. But the sine function has no limit at infinity. Therefore y(t) has no
limit as t → b. However, it should have because f is continuous. This contradiction shows that
there is no path joining the origin to a point in S, and so S is not path connected.
Remark. To prove rigorously that y(t) has no limit as t → b, we may proceed as follows. Let
b < b′ < c. Then x([b, b′ ]) is an interval that contains 0 and a positive number x(b′ ). Therefore
56 CHAPTER 4. CONNECTEDNESS

[0, x(b′ )] ⊂ x([b, b′ ]). Now for every n large enough π

1
+2nπ ∈ [0, x(b′ )]. Therefore there exists a
2
sequence tn ∈ [b, b′ ] such that x(tn ) = π
1
+2nπ . Then y(tn ) = 1. Similarly, there exists a sequence
2
sn ∈ [b, b′ ]
such that x(sn ) = π
1
+(2n+1)πand so y(sn ) = −1. Passing to a subsequence, we may
2
assume that tn → t and sn → s. Then x(tn ) → x(t) = 0 and x(sn ) → x(s) = 0. Therefore
t = s = b because otherwise x(t) > 0 or x(s) > 0. So we have two sequences that converge to b,
yet their images under y converge to distinct limits.

There are some interesting situations where connectedness and path connectedness are equiv-
alent. Here’s one of them. In Rn , a polygonal path (or broken line path) is a juxtaposition of
straight-line paths.

Theorem 4.8. Let U be an open subset of Rn . Then the following conditions are equivalent.

(i) U is connected.

(ii) U is path connected.

(iii) Any two points of U can be joined by a polygonal path in U .

Proof. It should be clear that (iii) ⇒ (ii) ⇒ (i). To prove that (i) ⇒ (iii), we use what we call
a connectedness argument. We fix a point x0 ∈ U and we consider the set A of points in U that
can be joined to x0 by a polygonal path. Then we prove that A is not empty, open and closed.
Connectedness then implies that A = U and so all points of U can be joined to a by a polygonal
path. Since a was arbitrary, this means that any two points of U can be joined by a polygonal path.
Let us carry out this program.

1. A is not empty since x0 can be joined to itself by the constant path.

2. A is open. Indeed, let x ∈ A. Since U is open, there exists a ball B(x, r) centered at x and
contained in U .

x0

Now, every point y ∈ B(x, r) can be joined to x by a straight line segment and therefore can
be joined to x0 by a polygonal path (by composition of paths). This means that B(x, r) ⊂ A.

3. A is closed in U , or equivalently U − A is open. Let x ∈ U − A. Then, x cannot be joined to

x0 by a polygonal path. Now there exists a ball B(x, δ) ⊂ U .
4.3. CONNECTED COMPONENTS 57

x0

If some point y ∈ B(x, δ) can be joined to x0 by a polygonal path, then x could be joined to x0
by a polygonal path (through y), contrary to the assumption. This means that B(x, δ) ⊂ U − A.
□

4.3 Connected components

If a space is not connected, it can be divided into pieces that are connected. Indeed, given a
topological space X, define the binary relation on X by setting x ∼ y if there exists a connected
subset of X containing both x and y. This is an equivalence relation. Indeed, first, the relation
is reflexive: x ∼ x because {x} is connected. Next, symmetry is evident. Last (transitivity), if
x ∼ y and y ∼ z, then, there is a connected subset A containing x and y and a connected subset
B containing y and z. Then A ∪ B contains x and z and is connected because A an B have y in
common. Therefore x ∼ z.

Definition. The equivalence classes for the relation above are called the connected components
(or sometimes just components) of X.
Now recall that given an equivalence relation on a set, the equivalence class containing an
element x is the set of points y that are in relation with x. Two equivalence classes are either
disjoint or identical and the whole set is the union of the pairwise disjoint equivalence classes. It is
also possible that there is only one equivalence class (in this case, all elements are equivalent).
Examples. 1) The connected components of R∗ are ] − ∞, 0[ and ]0, ∞[.
2) The connected components of Z are the singletons {n}. A space whose components are singletons
is said to be totally disconnected.
3) Q is totally disconnected.
4) A connected space has only one component: the space itself.

Proposition 4.3. Let X be a topological space. Then

(i) Two connected components of X are either disjoint or identical and X is the union of the
pairwise disjoint components.

(ii) The connected components of X containing an element x is the biggest connected subset of
X that contains x. It is usually denoted by C(x) or [x].

(iii) The components of X are closed.

(iv) Every nonempty connected subset of X is contained in exactly one connected component.
58 CHAPTER 4. CONNECTEDNESS

Proof. (i) is true for any equivalence relation.

(ii) Let D(x) be the union of all connected sets containing x. Then D(x) is connected as a union of
connected sets having a common point. Now if y ∈ D(x), then x ∼ y because D(x) is a connected
set containing x and y. Therefore y ∈ C(x). This proves that D(x) ⊂ C(x). Conversely, let
y ∈ C(x). Then, there is a connected set A containing x and y. Then, A ⊂ D(x) and so y ∈ D(x).
This proves that C(x) ⊂ D(x). Now of course, the union of all connected sets that contain x is the
biggest connected set containing x.
(iii) Let C(x) be a connected component. Then C(x) is a connected set containing x. By the
previous point, it is contained in C(x). This means that C(x) is closed.
(iv) We already observed this. If A is a nonempty connected set that contains x, then, A ⊂ C(x).
Since distinct components are disjoint, A cannot meet another component. □

Remark. If X has finitely many components, then every connected component is also open
because its complement is a finite union of closed sets. In general, components need not be open.
For instance the components of Q are not open. However, if O is an open subset of Rn then the
connected components of O are open (try to prove that if you want).

Theorem 4.9. Let f : X → Y be a homeomorphism. Then for every x ∈ X, f (C(x)) = C(f (x)).
In words: a homeomorphism takes connected components into connected components. In particular,
the number of connected components of a space is a topological invariant.
Proof. f (C(x)) is a connected subset containing f (x). Therefore it is contained in C(f (x)).
Conversely, let y = f (z) ∈ C(f (x)). Then f (z) ∼ f (x) i.e., f (z) and f (x) belong to the connected
set C = C(f (x)). But then, z and x belong to the same connected set f −1 (C). This means that
z ∼ x and so z ∈ C(x). Therefore y = f (z) ∈ f (C(x)). □

Examples. 1) The spaces [0, 1] and S 1 are not homeomorphic. Indeed, if we remove 12 from [0, 1],
the resulting space is not connected; whereas if we remove any point from S 1 , the resulting space
is connected.
2) The shapes X and Y (as compact subspaces of R2 ) are not homeomorphic. Indeed, if we remove
the middle point from X, the resulting space has 4 connected components; whereas if we remove
any point from Y, the resulting space has at most 3 components.
3) The shapes A and ∞ are not homeomorphic. Indeed, if we remove from A the topological circle
that it contains, the resulting space has two connected components; whereas if we remove from ∞
any topological circle, the resulting space is a circle without a point (which is connected).
4) The sphere S 2 and the torus T 2 (surface of an American doughnut) are not homeomorphic.

Indeed, if we remove an appropriate circle from the torus the resulting space is connected,
whereas if we remove from S 2 any topological circle, the resulting space is not connected (this
intuitive result is actually a difficult theorem called Jordan’s curve theorem).
We end this chapter by a useful characterization of open sets of the real line.
4.3. CONNECTED COMPONENTS 59

Lemma 4.3. Let O be an open subset of R. Then the connected components of O are open in
R.
Proof. Let C be a connected component of O. Let x ∈ C. Since O is a neighborhood of x,
there exists an interval V containing x such that V ⊂ O. Since V is connected and contains x, it
is contained in C. This means that C is a neighborhood of each of its points. Therefore C is open.

Theorem 4.10. Every open set O of the real line is the union of a countable family of pairwise
disjoint open intervals.
Proof. Let (Iλ )λ∈L be the collection of connected components of O. It follows from the previous
lemma that each Iλ is an open interval. Since the rational numbers are dense in R, each Iλ contains
a rational number. By the axiom of choice, we can choose exactly one rational number in each Iλ .
This defines a function between L and Q which is one to one because two distinct components are
disjoint. Therefore L is countable. Finally we have indeed O = ∪λ∈L Iλ . □
60 CHAPTER 4. CONNECTEDNESS
Chapter 5

Complete metric spaces

Let (xn ) be a sequence of a metric space (X, d) that converges to x. Recall that this means that
for any ε > 0, we can find an integer n0 such that d(xn , x) < ε for all n ≥ n0 . Therefore, by
the triangle inequality, d(xn , xm ) < 2ε for all integers n and m greater or equal to n0 . Otherwise
stated, the distance between any two terms can be made arbitrary small when their indices are large
enough. We are thus led to formulate the following definition.

Definition. Let (X, d) be a metric space. A sequence (xn ) of X is called a Cauchy sequence if
for any ε > 0 there is a number n0 such that d(xn , xm ) < ε for all integers n, m ≥ n0 .

According to what we said, any convergent sequence is a Cauchy sequence. Is the converse true?
The converse need not be true. If it is true, the space is said to be complete.

Definition. A metric space (X, d) is said to be complete if every Cauchy sequence of X is

convergent.

Counterexamples. 1) The sequence

√ 1.4, 1.41, 1.414, 1.4142, 1.41421, . . . of rational numbers is a
Cauchy sequence whose limit 2 is not in Q. Thus, Q is not complete.
2) Consider the sequence (xn ) of ]0,2[ defined by xn = n1 . This sequence is a Cauchy sequence which
is not convergent in ]0,2[. Thus, ]0,2[ is not complete. This example also shows that completeness
is not a topological property (that is, not invariant under homeomorphisms) because ]0,2[ is not
complete whereas R as we shall see is complete.

Remark. The Cauchy condition can also be stated as

• lim d(xn , xm ) = 0, or
n,m→∞

• for every p ∈ N∗ , lim d(xn , xn+p ) = 0.

n→∞

Lemma 5.1. A Cauchy sequence is bounded.

Proof. Let (xn ) be a Cauchy sequence of a metric space (X, d). Taking ε = 1 in the definition
of a Cauchy sequence, we see that there exists n0 such that d(xn , xm ) < 1 for n, m ≥ n0 . Let
M = max{d(xn , xn0 ); n ≤ n0 }. Now, if n ≤ n0 , then, d(xn , xn0 ) ≤ M ≤ M + 1. If n > n0 , then,
d(xn , xn0 ) < 1 ≤ M + 1. Therefore, d(xn , xn0 ) ≤ M + 1 for all n ∈ N∗ . This means that the
sequence (xn ) belongs to the ball B ′ (xn0 , M + 1) and so it is bounded. □

Lemma 5.2. If a Cauchy sequence (xn ) contains a convergent subsequence, then (xn ) is conver-
gent.

61
62 CHAPTER 5. COMPLETE METRIC SPACES

Proof. Let (xni ) be a subsequence converging to x. Let ε > 0 be given. Then, there is an index
m1 such that d(xni , x) < ε/2 for i ≥ m1 . Since (xn ) is a Cauchy sequence and ni ≥ i, there is an
integer m2 such that d(xni , xi ) < ε/2 for i ≥ m2 . By the triangle inequality, d(xi , x) < ε for all
i ≥ max(m1 , m2 ). □

Corollary 5.1. A compact metric space is complete.

Proof. Let (xn ) be an arbitrary Cauchy sequence of a compact metric space. Compactness implies
that (xn ) has a convergent subsequence. By the previous lemma, the original sequence (xn ) is
convergent.

Corollary 5.2. Rn is a complete under any one of the usual distances.

Proof. Since a Cauchy sequence is bounded, it belongs to a compact set (a closed ball). Therefore,
it contains a convergent subsequence. According to the previous lemma it converges. □

Remark. Thus, we have established a new criterion for convergence. Note that, this criterion
permits us to decide whether a sequence is convergent without knowing its limit. This is very useful,
especially in theoretical considerations.

Proposition 5.1. Let (X, d) be a complete metric space and let F be a subset of X. Then F is
complete if and only if it is closed in X.
Proof. Suppose that F is complete (under d). Let x ∈ F . Then, there is a sequence (xn ) of F
that converges to x. Therefore (xn ) is a Cauchy sequence. Since F is complete, (xn ) converges to
some y ∈ F . But (xn ) converges to x. Therefore x = y ∈ F . This means that F is closed in X.
Conversely, suppose that F is closed in X. Let (xn ) be a Cauchy sequence of F . Since X is
complete, (xn ) has a limit x ∈ X. But closedness of F implies that x ∈ F . □

Definition. Let X be a set and f : X → X. A fixed point of f is a point a ∈ X such that

a = f (a).

Definition. Let (X, dX ) and (Y, dY ) be metric space. A function f : X → Y is called a

contraction if there exists a constant k < 1 such that dY (f (x), f (y)) ≤ kdX (x, y) for all x, y ∈ X.

2
Example. The function f : [−1, 1] → [−1, 1] defined by f (x) = x3 is a contraction. Indeed,
f ′ (x) = 23 x and so |f ′ | is bounded by 23 . By the mean value theorem, |f (x) − f (y)| ≤ 23 |x − y| for
all x, y ∈ [−1, 1].

The following theorem is fundamental in analysis.

Theorem 5.1. (The Banach fixed point theorem). Let X be a complete metric space and let
f : X → X be a contraction. Then f has a unique fixed point.
Proof. Let k < 1 satisfy d(f (x), f (y)) ≤ kd(x, y) for all x, y ∈ X. We first prove uniqueness.
Suppose that we have two fixed points a and b. Then d(f (a), f (b)) ≤ kd(a, b) and so d(a, b) ≤
kd(a, b). This implies that d(a, b) = 0 because otherwise, we get 1 ≤ k.
Next, we prove existence of a fixed point. Pick a point a ∈ X and define the sequence (xn ) by
x0 = a and xn+1 = f (xn ). Observe that for n ≥ 1,

d(xn+1 , xn ) = d(f (xn ), xn ) = d(f (xn ), f (xn−1 )) ≤ kd(xn , xn−1 ).

 63

It follows by induction that d(xn+1 , xn ) ≤ k n d(x1 , x0 ). Let now p ≥ 1. Using the triangle inequality
and the last estimate, we get

d(xn , xn+p ) ≤ d(xn , xn+1 ) + d(xn+1 , xn+2 ) · · · + d(xn+p−1 , xn+p )

≤ k n d(x1 , x0 ) + k n+1 d(x1 , x0 ) + · · · + k n+p−1 d(x1 , x0 )
= (k n + k n+1 + · · · + k n+p−1 )d(x1 , x0 )
1 − kp kn
= kn d(x1 , x0 ) ≤ d(x1 , x0 ).
1−k 1−k
Since k n → 0 as n → ∞, it follows that d(xn , xn+p ) → 0 as n → ∞. This means that (xn )
is a Cauchy sequence. Since X is complete (xn ) has a limit ℓ. Letting n → ∞ in the relation
xn+1 = f (xn ) and using the continuity of f we get ℓ = f (ℓ). □

Remark 1. It’s essential that f maps X into itself. First, if x is a fixed point of f , then x and
f (x) should belong to the same space. But even, if X ⊂ Y and f : √ X → Y is a contraction,√then f
need not have a fixed point. For example the map f : [0, 1] → [0, 2] defined by f (x) = 1 + x2
is a contraction but has no fixed point (check that). Of course a contraction has at most one fixed
point.

Remark 2. Completeness is also essential. For example let X = {x ∈ Q; 1 ≤ x ≤ 2} be equipped

with the usual distance. Let f be defined by f (x) = x2 + x1 . Then f is a contraction from X to
itself but it has no fixed point (check the details).
64 CHAPTER 5. COMPLETE METRIC SPACES
Appendix A

Topological existence and

non-existence theorems

We already encountered two topological existence theorems. The first one is the extreme value
theorem which states that a continuous real valued function has a maximum and a minimum value
on a compact space. The nature of this result is topological. Note that this theorem gives no
effective procedure or algorithm to find the minimum or maximum, it only ensures their existence.
But this can be an important information to have before we start looking at a minimum value. The
second result is the intermediate value theorem which states that a continuous real valued function
on a connected space achieves every value between two given values it takes. Here again, there is
no effective procedure for finding the points where the function takes the given value.
Note that we encountered another existence result in chapter 5: the Banach fixed point theorem.
This theorem is different from the first two. It is not a topological theorem but rather a metric
theorem and its proof is constructive. However, it is the first example of a fixed point theorem.
Here’s a fundamental topological fixed point theorem.
Theorem (Brouwer’s fixed point theorem). Let B n denote the closed unit ball of Rn . Then
any continuous function f : B n → B n has a fixed point.
The proof requires advanced tools from algebraic topology like homotopy or homology. But note
that when n = 1, the theorem is a direct consequence of the intermediate value theorem. This result
has many interesting applications in Analysis and Linear Algebra. It has also many generalizations
like the Schauder fixed point theorem which is very important especially in differential equations.
Another problem of topology is the problem of existence of continuous extensions of a continuous
function. More precisely, let X and Y be topological spaces, let A be a subset of X and let
f : A → Y be continuous. Does there exist a continuous map g : X → Y such that g(x) = f (x)
for all x ∈ A? One fundamental extension theorem is the Tietze extension theorem. Another one
is the Dugundji extension theorem. Their statements and proofs require more advanced tools from
point set topology.
We already encountered some non-existence results. The fact that certain spaces are not home-
omorphic means that there is no homeomorphism between them. We will now give a non-existence
topological theorem.
Definition. Let A be a subset of a topological space X. A retraction f : X → A is a continuous
map such that f (x) = x for all x ∈ A. If f is such a map, we say that A is a retract of X. Otherwise
stated, a retraction is a continuous surjective map that fixes the points of the retract. For example,
x
the map x 7→ ||x|| is a retraction from the punctured space Rn − {0} to the sphere S n . This means
that we can deform in a continuous way the punctured space to the sphere while fixing the sphere.
However,

65
66 APPENDIX A. TOPOLOGICAL EXISTENCE AND NON-EXISTENCE THEOREMS

Theorem (No retraction theorem). There is no retraction from Rn to S n or from B n to S n .

Intuitively, this means that we cannot deform in a continuous way the closed unit ball to its
boundary while fixing the boundary. Otherwise stated, if we deform the closed unit ball to its
boundary while fixing the boundary, then, we have to tear the unit ball. Actually the no retraction
theorem is equivalent to the Brouwer fixed point theorem (prove this as a challenging exercise).
Appendix B

From the local to the global

We illustrate on some examples how topology turns a local property into a global one. We already
encountered such situation in chapter 3.

• Let f : X → Y be a continuous map (a local property) between metric spaces. If X is

compact (a topological property), then f is uniformly continuous (a global property).

Here are more examples. A function f : X → Y is said to be locally constant if every point
x ∈ X has a neighborhood on which f is constant. A function f : X → R is said to be locally
bounded if every point x ∈ X has a neighborhood on which f is bounded.
Exercise. (a) Give an example of a locally constant function which is not globally constant.
(b) Give an example of a locally bounded function which is not globally bounded.

However, the following hold. Prove them as an exercise.

• Let f : X → Y be locally constant. If X is connected (a topological property), then f is

globally constant.

• Let f : X → R be locally bounded. If X is compact, then f is globally bounded.

In M3304, we shall prove the following.

• Let Ω be an open subset of the complex plane and let f : Ω → C be holomorphic. Then, f
has locally an antiderivative. If Ω is simply connected (a topological property), then f has
globally an antiderivative. A simply connected subset of C is a set with no holes.

This is similar to the following result.

• Let F : Ω ⊂ R2 → R2 be an irrotational vector field. Then F has locally a potential function.

If Ω is simply connected, then F has globally a potential function.

Remark. Algebraic topology was invented before point set topology. One of the big ideas of
algebraic topology is to understand the holes in manifolds (think of spheres, tori, double tori,...).
This was first done with the so called Betti numbers. These numbers are topological invariants,
meaning that two homeomorphic manifolds have the same Betti numbers. Later in the 1920’s, the
algebraist Emmy Nother suggested to use groups instead of numbers. This opened the way to the
definition of homology groups and homotopy groups (and these are topological invariants). So we
can say that topology is the science of holes!

67