An Introduction to Point-Set Topology

Andrew Pease

December 2020

0. Introduction

Overview 0.1: This document will contain many of the definitions that are included in a
standard introductory topology course. It will cover the typical types of topologies, continuous
functions and metric spaces, compactness and connectedness, and the separation axioms. In
addition to the definitions, I have included proofs to various exercises and theorems as well as
some commentary explaining the intuition behind the definitions/theorems/propositions. I found
the theorem-proof style of writing insufficient for building a strong understanding of the math, so
I wanted to include my own explanations in addition to the theorems and definitions.

Prerequisites 0.2: The only prerequisites for these notes is elementary set theory that you learn
in a discrete math class.

1. Topological Spaces

Definition 1.1: A topology on a set X is some collection 𝒯 of subsets of X such that

(1) ∅ , 𝑋 ∈ 𝒯
(2) The intersection of elements of any finite subcollection of 𝒯 is in 𝒯
(3) The union of arbitrarily many elements of 𝒯 is in 𝒯

Intuitively, a topology is a structure that we impose on some set, much like the
algebraic structure of numbers like integers and the reals. Just as numbers can
seem meaningless without operations defined on them, sets without structure like
a topology are, in a sense, vacuous.
Figure 1.2:

Exercise 1.3: Let X be a set, and let 𝒯 = {𝑈 | 𝑋 − 𝑈 𝑖𝑠 𝑐𝑜𝑢𝑛𝑡𝑎𝑏𝑙𝑒 𝑜𝑟 𝑖𝑠 𝑎𝑙𝑙 𝑜𝑓 𝑋} be some

collection of subsets of X. Does this collection form a topology on X?

Proof: If we let 𝑈 = 𝑋, then 𝑋 − 𝑈 = ∅ and the empty set is countable, so it is in 𝒯; let 𝑈 = ∅,

then 𝑋 − 𝑈 = 𝑋, and by definition of 𝒯, 𝑋 ∈ 𝒯 as well, so we have verified property (1) of the
definition of a topology. Now suppose {𝑈1 , 𝑈2 , . . . , 𝑈𝑛 } is some collection of sets 𝑈𝑖 ∈ 𝒯. Then,
to show that the finite intersection of elements of 𝒯 are in 𝒯, we show that

X − ∩𝑛𝑖=1 (𝑈𝑖 ) = ∪𝑛𝑖=1 (𝑋 − 𝑈𝑖 )

Note, any finite union of countable sets is also countable, so the latter set is countable, and thus
in 𝒯.
To show that arbitrary union of sets in 𝒯 is in 𝒯, we take an indexed family of sets {𝑈𝑖 }, 𝑖 ∈ 𝐼 for
some index family I. Then, we show that

𝑋 − ∪ (𝑈𝑖 ) = ∩ (𝑋 − 𝑈𝑖 )

Now, if 𝑈𝑖 ∈ 𝒯, then 𝑋 − 𝑈𝑖 is countable. The arbitrary intersection of countable sets is

countable, and thus the latter set is in 𝒯, and we are finished.

Definition 1.4: The discrete topology on a set X is defined as 𝒯 = 𝒫(𝑋), the power set of X.
The indiscrete topology or trivial topology on a set X is defined as 𝒯 = {∅, 𝑋}.

Definition 1.5: An open set A of some set X with topology 𝒯, is defined precisely as a subset of
X, as long as A is in 𝒯. If A is not in 𝒯, then A is not an open set of X. A set B of X is closed if
𝑋 − 𝐵 ∈ 𝒯, i.e., its complement is open.
 Note, this requires that both the whole set and the empty set be both open and
closed in any arbitrary topology. It seems counterintuitive, but a set being open is
not the negation of a set being closed (sometimes, you can even have a set that is
neither open nor closed).

Exercise 1.6: Let X be a topological space; let A be a subset of X. Suppose that for each 𝑥 ∈ 𝐴,
there is an open set U, such that 𝑥 ∈ 𝑈, 𝑈 ⊂ 𝐴. Show that A is open in X.

Proof: The set A is composed of arbitrarily many points 𝑥𝑖 . By hypothesis, for every 𝑥𝑖 , there is
an open set 𝑈𝑖 such that 𝑥𝑖 ∈ 𝑈𝑖 , 𝑈𝑖 ⊂ 𝐴. Then, 𝐴 = ∪ (𝑈𝑖 ), a union of open sets, and is thus
open.

If you want to show something is open or closed, you must use some set theory to
manipulate what you’re given to show that it is in the topology (or its complement
is). This previous example was quite simple, but the ones you might see in the
future can be more involved. If you’re ever uncertain, start from the basic
definition of open, and work from there.

Definition 1.7: A basis for a topology on a set X is some collection 𝔅 of subsets of X such that
(1) For every 𝑥 ∈ 𝑋, there is an element 𝐵 ∈ 𝔅 such that 𝑥 ∈ 𝐵 and
(2) If 𝑥 ∈ 𝐵1 and 𝑥 ∈ 𝐵2 , then there is an element 𝐵3 such that 𝑥 ∈ 𝐵3 ⊂ 𝐵1 ∩ 𝐵2

The topology generated by 𝔅 is defined as: for every open set 𝑈 ⊂ 𝑋 and ∀𝑥 ∈ 𝑈, there is a
basis element 𝐵 ∈ 𝔅, such that 𝑥 ∈ 𝐵 ⊂ 𝑈.

The topological definition of basis is, in a way, quite similar to the one used in
linear algebra. Just as every element in some vector space can be written as a
linear combination of basis vectors, every open set in some topological space can
be written as a union of basis elements. It is analogous to being the spanning set,
though not necessarily minimal, as in linear algebra. Bases are very valuable, as
they can describe a topology with relatively little information itself (refer to Def.
1.14). The motivation behind the idea of a basis was to find a way to encode the
structure of a set without enumerating (which is sometimes impossible!) all the
elements of the topology itself.

Below is a diagram detailing the relationship of a basis (with elements B) relative

to a topology (with elements U).
Figure 1.8:

Definition 1.9: Let X and Y be topological spaces. The product topology on 𝑋 × 𝑌 has as a
basis the collection 𝔅 = {𝑈 × 𝑉: 𝑈 𝑜𝑝𝑒𝑛 𝑖𝑛 𝑋, 𝑉 𝑜𝑝𝑒𝑛 𝑖𝑛 𝑌}.

Theorem 1.10: If 𝔅 is a basis for X, and 𝒞 is a basis for Y, then 𝔇 = {𝐵 × 𝐶 | 𝐵 ∈ 𝔅, 𝐶 ∈ 𝒞}

is a basis for the product topology on 𝑋 × 𝑌

Proof: Suppose that 𝑥 × 𝑦 are some points in the open set 𝑈 × 𝑉 of 𝑋 × 𝑌, 𝑈 ⊂ 𝑋, 𝑉 ⊂ 𝑌. By

definition of the product topology, U and V are open in X and Y, respectively. Then, by
hypothesis, there is some 𝐵 ∈ 𝔅, 𝐶 ∈ 𝒞 such that 𝑥 ∈ 𝐵 ⊂ 𝑈, 𝑦 ∈ 𝐶 ⊂ 𝑌. Thus, there is some
𝐵 × 𝐶 such 𝑥 × 𝑦 ∈ 𝐵 × 𝐶 ⊂ 𝑈 × 𝑉. Therefore, 𝔇 is a basis for the product topology on 𝑋 × 𝑌.

Definition 1.11: Let 𝜋1 ∶ 𝑋 × 𝑌 → 𝑋 be the projection map defined as 𝜋1 (𝑥, 𝑦) = 𝑥.

Intuitively, what this map does is project some box in a “coordinate plane” onto
one of the axes. You’ve probably dealt with similar maps in vector calculus or
number theory, and this one behaves very similarly.

Definition 1.12: A map 𝑓 ∶ 𝑋 → 𝑌 is an open map if for every open set 𝑈 ∈ 𝑋, 𝑓(𝑈) is open in
𝑌.

Proposition 1.13: The projection map 𝜋1 ∶ 𝑋 × 𝑌 → 𝑋 is an open map

Proof: Recall, by Definition 1.9, an open set in 𝑋 × 𝑌 with the product topology has a basis of
the form 𝐴 × 𝐵, where 𝐴 is open in 𝑋, and 𝐵 is open in 𝑌. Thus, for any open set 𝑈 × 𝑉 ∈ 𝑋 × 𝑌
𝑓(𝑈 × 𝑉) = 𝑈, which is, by definition, open in X.

Definition 1.14: The standard topology on ℝ has, as a basis, the collection of all open intervals
on the real line:
(𝑎, 𝑏) = {𝑥 ∈ ℝ | 𝑎 < 𝑥 < 𝑏}
Typically, this is the topology that we use when discussing the real number line (hence the
name), unless otherwise specified.

Up to this point, you might (or maybe not) have begun to wonder what happens to
a topology on a space when we look at some subset of the space. That is, if we
begin to “zoom in” on, or cut out a subset of a space, what happens to the
topology? This natural question leads us to our next definition.

Definition 1.15: Let 𝑋 be a topological space with topology 𝒯, and let 𝑌 ⊂ 𝑋. Then, the
collection
𝒯Y = {U ∩ Y | U open in X}

forms what we call the subspace topology on Y. Additionally, when Y inherits this topology from
X, we call it the subspace of X.

Theorem 1.16: If A is a subspace of X, and B is a subspace of Y, then the product topology on

𝐴 × 𝐵 is the same as the topology 𝐴 × 𝐵 inherits as a subspace of 𝑋 × 𝑌.

Proof: Suppose A is a subspace of X and B is a subspace of Y. A and B have the topologies

𝒯𝐴 = {U ∩ 𝐴 | U open in X}

and
𝒯𝐵 = {𝑉 ∩ 𝐵 | V open in Y}

respectively. Then the product topology on 𝐴 × 𝐵 has as a basis, by definition,

𝔅 = {𝑊 × 𝑍 | 𝑊 𝑜𝑝𝑒𝑛 𝑖𝑛 𝐴, 𝑍 open in 𝐵}

or equivalently,

𝔅 = {(𝑈 ∩ 𝐴) × (𝑉 ∩ 𝐵) | 𝑈 open in 𝑋, 𝑉 open in 𝑌}

Consider the basis for the topology 𝐴 × 𝐵 inherits as a subspace of 𝑋 × 𝑌,

𝒞 = {(𝐶 × 𝐷) ∩ (𝐴 × 𝐵) | 𝐶 × 𝐷 𝑜𝑝𝑒𝑛 𝑖𝑛 𝑋 × 𝑌}

But, since 𝐶 × 𝐷 is open in 𝑋 × 𝑌 with the product topology, 𝐶 is open in X and D is open in Y.
Then, note that (𝐶 × 𝐷) ∩ (𝐴 × 𝐵) = (𝐶 ∩ 𝐴) × (𝐷 ∩ 𝐵), so

𝒞 = {(𝐶 ∩ 𝐴) × (𝐷 ∩ 𝐵) | 𝐶 𝑜𝑝𝑒𝑛 𝑖𝑛 𝑋, 𝐷 𝑜𝑝𝑒𝑛 𝑖𝑛 𝑌}

Thus, 𝒞 is equivalent to 𝔅, and we are done.

Definition 1.17: Let A be a subset of a topological space X. The closure (denoted 𝐴) of a set A
is the intersection of all closed sets that contain A; or equivalently, it is the minimal closed set
containing A. The interior (denoted 𝐼𝑛𝑡 𝐴) of a set A is the union of all open sets contained in A.

The relationship between 𝐼𝑛𝑡 𝐴, 𝐴, 𝐴 is 𝐼𝑛𝑡 𝐴 ⊂ 𝐴 ⊂ 𝐴. If A is open, then

𝐼𝑛𝑡 𝐴 = 𝐴; if A is closed, then 𝐴 = 𝐴. You might be wondering “what happens if
A is neither open nor closed?”, as this definition doesn’t make it clear what it
means to be neither open nor closed. This idea will motivate an equivalent
definition of closure.

Definition 1.18: Let A be a subset of a topological space X. A point 𝑥 ∈ 𝐴 is a limit point of A,

if every open set containing x intersects A in a point different from x (another term for an open
set containing x is a neighborhood of x). The closure of a set A is 𝐴 = 𝐴 ∪ 𝐴′ , where 𝐴′ is the set
containing all the limit points of A.

Suppose we have some circle A defined as

𝐴 = {𝑥, 𝑦 ∈ ℝ | 𝑥 2 + 𝑦 2 < 1}

The limit points of A are

𝐴′ = {𝑥, 𝑦 ∈ ℝ | 𝑥 2 + 𝑦 2 ≤ 1}

Suppose we adjoin the point (0,1) to A; then 𝐴 ≠ 𝐼𝑛𝑡𝐴, and 𝐴 ≠ 𝐴. We can

intuitively understand the limit points of a set more easily now; in this case, if we
take any open ball around the limit points of A, it will necessarily intersect the
inside of the circle, thus verifying our definition for A’. This intuition won’t
always make sense, but it’s strong when it does.

Note: Points in the interior can be limit points but need not be limit points. Try
and find an example where some point in the interior of a set is not a limit point
(what topology contains a set where a point is its own neighborhood?).

Lemma 1.19: Let A be a subset of some space X. Then, 𝑥 ∈ 𝐴 if and only if every open set U
containing x intersects A.

Proposition 1.20: Let 𝐴, 𝐵 be subsets of some space X. If 𝐴 ⊂ 𝐵, 𝑡ℎ𝑒𝑛 𝐴 ⊂ 𝐵.

Proof: Suppose 𝐴 ⊂ 𝐵, and for the sake of contradiction 𝐴 is not a subset of 𝐵. Then ∃𝑥 ∈ 𝐴
such that 𝑥 ∉ 𝐵. If 𝑥 ∈ 𝐴, then every open set U containing x intersects A. But, by hypothesis,
𝐴 ⊂ 𝐵, so if U intersects A, then U also intersects B. Thus, by Lemma 1.19, 𝑥 ∈ 𝐵; this
contradicts our supposition, so no such x exists. Therefore, if 𝐴 ⊂ 𝐵, 𝑡ℎ𝑒𝑛 𝐴 ⊂ 𝐵.

Proposition 1.21: Let 𝐴, 𝐵 be subsets of some space X. Then, 𝐴 ∪ 𝐵 = 𝐴 ∪ 𝐵

Proof: Consider some arbitrary 𝑥 ∈ 𝐴 ∪ 𝐵. By Lemma 1.19, every open set U containing x
intersects 𝐴 ∪ 𝐵, or equivalently, U intersects A or B. We know that for any set 𝐶, 𝐶 ⊂ 𝐶, so if U
intersects 𝐴 𝑜𝑟 𝐵, U intersects 𝐴 𝑜𝑟 𝐵. Thus, by Lemma 1.19, 𝑥 ∈ 𝐴 𝑜𝑟 𝑥 ∈ 𝐵. Note, though, that
𝐴 = 𝐴 𝑎𝑛𝑑 𝐵 = 𝐵. So, if 𝑥 ∈ 𝐴 or 𝑥 ∈ 𝐵, 𝑥 ∈ 𝐴 or 𝑥 ∈ 𝐵. Therefore 𝑥 ∈ 𝐴 ∪ 𝐵, so we may say
𝐴 ∪ 𝐵 ⊂ 𝐴 ∪ 𝐵. For the other direction, consider some arbitrary 𝑥 ∈ 𝐴 ∪ 𝐵. Then, we say that
𝑥 ∈ 𝐴 𝑜𝑟 𝑥 ∈ 𝐵. Without loss of generality, assume 𝑥 ∈ 𝐴. Then, every open set U containing x
intersects A; and thus, U intersects 𝐴 ∪ 𝐵. But then, by Lemma 1.19, 𝑥 ∈ 𝐴 ∪ 𝐵, so we know
𝐴 ∪ 𝐵 ⊃ 𝐴 ∪ 𝐵. Thus, by showing containment both ways, we know that 𝐴 ∪ 𝐵 = 𝐴 ∪ 𝐵.

2. Continuous Functions and Metric Spaces

Definition 2.1: Let X and Y be topological spaces. A function 𝑓: 𝑋 → 𝑌 is said to be continuous

if for every open subset 𝑉 ⊂ 𝑌, 𝑓 −1 (𝑉) is open in X.

This is a natural generalization of continuous functions in calculus/analysis. In

calculus you are introduced to the epsilon-delta definition of continuity; for every
n-dimensional ball of radius epsilon (no matter how small) around a point in the
range, there is always a delta in the domain that maps to that ball. Under the
standard topology on ℝ, this definition of continuity coincides with the epsilon-
delta one we are all familiar with. However, we will see that this definition works
in all topological spaces, even those without a metric. We will also see the
epsilon-delta generalizes to all metric-spaces.

Proposition 2.2: Let X and Y be topological spaces. A function 𝑓: 𝑋 → 𝑌 is continuous if and

only if for each 𝑥 ∈ 𝑋 and each neighborhood V of 𝑓(𝑥), there is a neighborhood 𝑈 of x such
that 𝑓(𝑈) ⊂ 𝑉.

Proof: Suppose 𝑓: 𝑋 → 𝑌 is continuous, and for the sake of contradiction, there exists some
neighborhood V of 𝑓(𝑥) such that there does not exist a neighborhood U of x where 𝑓(𝑈) ⊂ 𝑉.
Then, there does not exist some neighborhood 𝑊 in X such that 𝑊 = 𝑓 −1 (𝑉). However, this
contradicts the definition of continuity, so no such V exists. In the other direction, suppose that
for each neighborhood V of 𝑓(𝑥), there is a neighborhood U of x such that 𝑓(𝑈) ⊂ 𝑉. Let V be
some neighborhood of 𝑓(𝑥). Then, 𝑥 ∈ 𝑓 −1 (𝑉). Choose some neighborhood 𝑈𝑖 such that 𝑥 ∈
𝑈𝑖 , 𝑓(𝑈𝑖 ) ⊂ 𝑉, which we can do by hypothesis. Then for all points 𝑥 ∈ 𝑓 −1 (𝑉), there exist
corresponding open sets containing them. Take their union, and you have an open set

∪ {𝑈𝑖 }𝑖∈𝐼 = 𝑓 −1 (𝑉)

This definition of continuity coincides a lot with our intuition of open balls that
we see in the epsilon-delta definition of metric spaces. That is, for every
 neighborhood (open ball) A of points in the range, there is a corresponding
neighborhood (open ball) in the domain that maps into entirely A.

Exercise 2.3: Prove that for functions 𝑓: ℝ → ℝ, the epsilon-delta definition of continuity
implies the open set definition.

Proof: Consider some function 𝑓: ℝ → ℝ; the epsilon-delta definition of continuity states that for
any 𝜀 > 0, ∃𝛿 > 0 such that 0 < |𝑥 − 𝑥0 | < 𝛿 ⇒ |𝑓(𝑥) − 𝑓(𝑥0 )| < 𝜀. Now, note that this
follows directly from Proposition 2.2, just with open intervals instead of neighborhoods.
Rewritten in terms of open intervals, for some 𝜀 > 0, there is a neighborhood 𝑈 =
(𝑥0 − 𝛿, 𝑥0 + 𝛿) of 𝑥0, such that ∀𝑥 ∈ 𝑈, 𝑓(𝑥) ∈ 𝑉, where 𝑉 = (𝑦0 − 𝜀, 𝑦0 + 𝜀) is some
neighborhood of 𝑓(𝑥0 ), and 𝑓(𝑥0 ) = 𝑦0. With less notation, for some neighborhood V of any
𝑓(𝑥0), there will always exist a neighborhood U of 𝑥0, such that 𝑓(𝑈) ⊂ 𝑉. By Proposition 2.2,
this is equivalent to the open set definition and we are done.

Definition 2.4: Let X and Y be topological spaces. A function 𝑓: 𝑋 → 𝑌 is a homeomorphism if

𝑓 𝑎𝑛𝑑 𝑓 −1 are continuous and bijective. If X and Y have a homeomorphism between them, they
are homeomorphic

A homeomorphism is a structure preserving map, i.e., the topological version of

an isomorphism. Notice, a homeomorphism requires 𝑓 and 𝑓 −1 to be continuous.
Bijective functions that are continuous preserve a certain type of structure: the
topology. If a set is open in X, it is mapped to an open set in Y and vice versa.
This is analogous to, for example, a ring isomorphism, where you want 𝑓(0) = 0
(part of the structure of a ring are its identity elements, so this must also be
preserved). Just like any isomorphism, topological spaces that are homeomorphic
are, for all intents, the same spaces with different notation.

Proposition 2.5: The subspace (𝑎, 𝑏) of ℝ is homeomorphic with (0,1).

𝑥−𝑎
Proof: For (𝑎, 𝑏), let 𝑓: (𝑎, 𝑏) → (0,1) be defined as 𝑓(𝑥) = 𝑏−𝑎. Now, to show 𝑓 and 𝑓 −1 are
bijective and continuous. It suffices to show that 𝑓 is bijective, as the bijectivity of 𝑓 −1 will
necessarily follow. To show surjectivity choose some 𝑦 ∈ (0,1). We must find an 𝑥 ∈ (𝑎, 𝑏)
such that 𝑓(𝑥) = 𝑦. Define 𝑥 = 𝑦(𝑏 − 𝑎) + 𝑎; this is a valid definition, as 0 < 𝑦 < 1, so we see
𝑎 < 𝑦(𝑏 − 𝑎) + 𝑎 < 𝑏, for any 𝑦. Thus, there will always be an 𝑥 ∈ (𝑎, 𝑏) that satisfies
𝑦(𝑏−𝑎)+𝑎−𝑎
𝑦(𝑏 − 𝑎) + 𝑎. Then, 𝑓(𝑥) = 𝑓(𝑦(𝑏 − 𝑎) + 𝑎) = (𝑏−𝑎) = 𝑦. Now for injectivity, choose
1 𝑥 −𝑎
some 𝑓(𝑥1 ), 𝑓(𝑥2 ) such that 𝑓(𝑥1) = 𝑓(𝑥2 ); we must show that 𝑥1 = 𝑥2. 𝑓(𝑥1) = (𝑏−𝑎) and
𝑥 −𝑎
2 𝑥 −𝑎
1 𝑥 −𝑎
2
𝑓(𝑥2) = (𝑏−𝑎); therefore (𝑏−𝑎) = (𝑏−𝑎), 𝑥1 − 𝑎 = 𝑥2 − 𝑎, 𝑥1 = 𝑥2, which is what we desired. We
know 𝑓 and 𝑓 −1 are bijective, and now must show that they’re continuous.

Let’s start with 𝑓. Choose some 𝑥 ∈ (𝑎, 𝑏), and some neighborhood (𝑐, 𝑑) of 𝑓(𝑥). Then, there
exist some 𝑦, 𝑧 ∈ (𝑎, 𝑏), such that 𝑦 < 𝑥 < 𝑧 and 𝑓(𝑦), 𝑓(𝑧) ∈ (𝑐, 𝑑) by definition of bijectivity
and the function 𝑓. Then, let (𝑦, 𝑧) be a neighborhood of 𝑥. It is obvious that 𝑓((𝑦, 𝑧)) ⊂ (𝑐, 𝑑),
so we can conclude that 𝑓 is continuous. An identical argument follows for 𝑓 −1 , so we are done.
Therefore, the subspace (𝑎, 𝑏) of ℝ is homeomorphic to (0,1).

Proposition 2.6: Let 𝑓: 𝐴 → 𝐵 𝑎𝑛𝑑 𝑔: 𝐶 → 𝐷 be continuous functions. Define 𝑓 × 𝑔: 𝐴 × 𝐶 →

𝐵 × 𝐷 as
(𝑓 × 𝑔)(𝑎 × 𝑐) = 𝑓(𝑎) × 𝑔(𝑐)
Then 𝑓 × 𝑔 is continuous.

Proof: The basis for the product topology of 𝑋 × 𝑌 is 𝔅 = {𝑀 × 𝑁 | 𝑀 𝑜𝑝𝑒𝑛 𝑖𝑛 𝑋, 𝑁 open in 𝑌}.
Take some set W open in 𝐵 × 𝐷; we need to show (𝑓 × 𝑔)−1 (𝑊) is open in 𝐴 × 𝐶. Let W be
defined as

W =∪(𝑖,𝑗)∈𝐼×𝐽 (𝑈𝑖 × 𝑉𝑗 )

i.e., some arbitrary union of basis elements. Then,

W = (∪𝑖∈𝐼 𝑈𝑖 ) × (∪𝑗∈𝐽 𝑉𝑗 ) = U × 𝑉

where 𝑈𝑖 is open in 𝐵, and 𝑉𝑗 is open in 𝐷. Then,

(𝑓 × 𝑔)−1 (𝑊) = (𝑓 × 𝑔)−1 (𝑈 × 𝑉) = 𝑓 −1 (𝑈) × 𝑔−1 (𝑉)

But 𝑓: 𝐴 → 𝐵 and 𝑔: 𝐶 → 𝐷 are continuous, so 𝑓 −1 (𝑈) and 𝑔−1 (𝑉) are open. Consequently,
(𝑓 × 𝑔)−1 (𝑊) is also open as it is a product of open sets; therefore 𝑓 × 𝑔 is continuous.

Definition 2.7: A metric 𝑑 on some space is a function 𝑑: 𝑋 × 𝑋 → ℝ having the properties for
each 𝑥, 𝑦, 𝑧 ∈ 𝑋

(1) 𝑑(𝑥, 𝑦) ≥ 0, 𝑑(𝑥, 𝑥) = 0

(2) 𝑑(𝑥, 𝑦) = 𝑑(𝑦, 𝑥)
(3) 𝑑(𝑥, 𝑦) + 𝑑(𝑦, 𝑧) ≥ 𝑑(𝑥, 𝑧)

A metric is a generalization of distance. Essentially, we chose the most important

and general characteristics of distance to define a metric. Metric spaces are
studied frequently in real analysis and won’t be covered in great detail here. The
principal question arising from this section is demonstrating that a metric induces
a topology identical to ones previously encountered.

Definition 2.8: Let X be a metric space, and let d be the metric on X. The 𝜀-ball centered at x is
the set of all points such that, for some 𝜀 > 0, 𝑑(𝑥, 𝑦) < 𝜀. We denote this set

𝐵𝑑 (𝑥, 𝜀) = {𝑦 | 𝑑(𝑥, 𝑦) < 𝜀}

Definition 2.9: The metric topology induced by d is the topology with basis 𝐵𝑑 (𝑥, 𝜀) =
{𝑦 | 𝑑(𝑥, 𝑦) < 𝜀}, for all 𝑥 ∈ 𝑋, 𝜀 > 0.
 This definition of a metric topology looks similar to the standard topology on ℝ.
In fact, we can see that, given what’s called the standard metric on ℝ, 𝑑(𝑥, 𝑦) =
|𝑥 − 𝑦|, we can see that the definition of an open ball in ℝ, and an open interval
in ℝ are identical; that is, the metric topology induced by d and the standard
topology are the same.

Exercise 2.10: The metric 𝑑′(𝒙, 𝒚) = |𝑥1 − 𝑦1 | + ⋯ + |𝑥𝑛 − 𝑦𝑛 | induces the regular topology
on ℝ𝑛

Proof: First, let’s show that 𝑑 ′ is a metric. The first two properties are fulfilled trivially. To show
the triangle inequality, let 𝒙, 𝒚, 𝒛 ∈ ℝ𝑛 . Then, we see that:

𝑑′(𝒙, 𝒛) = |𝑥1 − 𝑧1 | + ⋯ + |𝑥𝑛 − 𝑧𝑛 |

and

𝑑′(𝒙, 𝒚) + 𝑑′(𝒚, 𝒛) = |𝑥1 − 𝑦1 | + ⋯ + |𝑥𝑛 − 𝑦𝑛 | + |𝑦1 − 𝑧1 | + ⋯ + |𝑦𝑛 − 𝑧𝑛 |

or equivalently

𝑑′(𝒙, 𝒚) + 𝑑′(𝒚, 𝒛) = |𝑥1 − 𝑦1 | + |𝑦1 − 𝑧1 | + ⋯ + |𝑥𝑛 − 𝑦𝑛 | + |𝑦𝑛 − 𝑧𝑛 |

But term by term, |𝑥𝑖 − 𝑦𝑖 | + |𝑦𝑖 − 𝑧𝑖 | ≥ |𝑥𝑖 − 𝑧𝑖 |, as in general, |𝑎 − 𝑏| + |𝑏 − 𝑐| ≥ |𝑎 − 𝑐|.

Therefore, as each term is in 𝑑′(𝒙, 𝒚) + 𝑑′(𝒚, 𝒛) greater than or equal to the corresponding term
of 𝑑′(𝒙, 𝒛), we may say that 𝑑′(𝒙, 𝒚) + 𝑑′(𝒚, 𝒛) ≥ 𝑑′(𝒙, 𝒛).

Now to show that 𝑑 ′ induces the standard topology on ℝ𝑛 . Consider some open set 𝑈 of ℝ𝑛 and
let 𝒙 ∈ 𝑈. We wish to find some 𝐵𝑑′ (𝒙, 𝜀) ⊂ 𝑈. Recall that U = ∏𝑛𝑖=1(𝑦𝑖 , 𝑧𝑖 ) under the product
topology of ℝ𝑛 . Choose an 𝜀, such that 𝜀 = 𝑚𝑖𝑛{𝜀1 , . . . , 𝜀𝑛 }, where 𝜀𝑖 = 𝑚𝑖𝑛{𝑥𝑖 − 𝑦𝑖 , 𝑥𝑖 − 𝑧𝑖 }.
Then, it is clear that 𝐵𝑑′ (𝒙, 𝜀) ⊂ 𝑈. For the other direction, I will only verbally explain proof.
Consider some 𝐵𝑑′ (𝒙, 𝜀); we wish to find an open set 𝑉 such that 𝑉 ⊂ 𝐵𝑑′ (𝒙, 𝜀). For any 𝜀 > 0,
there certainly exists a 𝛿 > 0, such that (𝑥1 − 𝛿, 𝑥1 + 𝛿) ×. . .× (𝑥𝑛 − 𝛿, 𝑥𝑛 + 𝛿) is “inside” the
𝐵𝑑′ (𝒙, 𝜀). Constructing such a 𝛿 is less trivial though. Regardless, once you find such 𝛿, the
proof is complete, and 𝑑 ′ induces the standard topology on ℝ𝑛 .

3. Connectedness and Compactness

Definition 3.1: Let X be a topological space. A separation of X is some nonempty open sets
𝑈, 𝑉 ⊂ 𝑋 such that 𝑈 ∩ 𝑉 = ∅ and 𝑈 ∪ 𝑉 = 𝑋. If a separation exists for a space X, then that
space is said to be disconnected. If no such separation exists, then that space is connected.
 The idea of connectedness is quite intuitive. It’s easy to imagine some space X
being composed of open sets, and if there is some partition of X, then the space
can be divided cleanly, i.e., the word disconnected. However, if every open set is
intertwined with another open set, like links in a chain, then we say that is
connected.

Theorem 3.2: The union of a collection of connected subspaces of X that share a point is
connected.

Proof: Suppose we have some family {𝐴𝑛 } of connected subspaces and let x be a point in ∩ 𝐴𝑛 .
Suppose for the sake of contradiction that ∪ 𝐴𝑛 is not connected. Then, there are nonempty open
sets C and D such that 𝐶 ∪ 𝐷 =∪ 𝐴𝑛 𝑎𝑛𝑑 𝐶 ∩ 𝐷 = ∅. However, if C and D have an empty
intersection, then either C or D do not contain any elements from {𝐴𝑛 }, as, by hypothesis, x is a
point in ∩ 𝐴𝑛 (if they both had an element from {𝐴𝑛 }, 𝑥 ∈ 𝐶 ∩ 𝐷). But that cannot be the case,
as C and D are nonempty. Thus, with the discovery of a contradiction, we can conclude that no
such C and D exist, and ∪ 𝐴𝑛 is connected.

Proposition 3.3: Let {𝐴𝑛 } be a sequence of connected subspaces such that 𝐴𝑛 ∩ 𝐴𝑛+1 ≠ ∅.
Then ∪ {𝐴𝑛 } is connected.

Proof: Note, this follows almost directly from Theorem 3.2, just in an inductive case. Consider
𝐴1 𝑎𝑛𝑑 𝐴2 as our base case. From Theorem 3.2, 𝐴1 ∪ 𝐴2 is connected. For our inductive
hypothesis, assume 𝐴1 ∪ … ∪ 𝐴𝑛 is connected. Then, as 𝐴𝑛 ∩ 𝐴𝑛+1 ≠ ∅, (𝐴1 ∪. . .∪ 𝐴𝑛 ) ∩
𝐴𝑛+1 ≠ ∅. Thus, when we invoke our inductive hypothesis, there cannot exist a separation of
𝐴1 ∪ … ∪ 𝐴𝑛 ∪ 𝐴𝑛+1, so it is connected. Our induction is complete, and we know that for any
sequence {𝐴𝑛 } of connected subspaces such that 𝐴𝑛 ∩ 𝐴𝑛+1 ≠ ∅, ∪ {𝐴𝑛 } is connected.

Proposition 3.4: A space is totally disconnected if the only connected subspaces are singletons.
Then, some space X with the discrete topology is totally disconnected, but the converse is not
true.

Proof: Recall that a set X with the discrete topology 𝒯 has as a topology the collection 𝒫(𝑋).
Suppose that A is an open set of X such that it contains at least 2 points. Select some one point
𝑥 ∈ 𝐴, and denote all the other points in A as {𝑦𝑖 } ∈ 𝐴. By the definition of a power set, there
exists a 𝐵 ∈ 𝒯 such that {𝑥} = 𝐵 and 𝐶 ∈ 𝒯 such that {{𝑦𝑖 }} = 𝐶. Then, from the definition of
subspace topology, 𝐵 ∩ 𝐴 = 𝐵, 𝐵 ∈ 𝒯𝐴 and 𝐶 ∩ 𝐴 = 𝐶, 𝐶 ∈ 𝒯𝐴 . Therefore, there exist 2 open sets
in A such that 𝐵 ∪ 𝐶 = 𝐴, 𝐵 ∩ 𝐶 = ∅. Consequently, any subspace of X with more than one
element is disconnected. Thus, any space X with the discrete topology is totally disconnected.
However, not every totally disconnected set has the discrete topology. At the end of this section,
we will consider the Cantor set, which is totally disconnected, but does not have the discrete
topology.

Proposition 3.5: Let 𝑌 ⊂ 𝑋, and let X and Y be connected. If A and B form a separation of 𝑋 −
𝑌, then 𝑌 ∪ 𝐴, 𝑌 ∪ 𝐵 are connected.
Proof: Suppose that X and Y are connected, A and B form a separation of 𝑋 − 𝑌, and, for the sake
of contradiction, 𝑌 ∪ 𝐴 or 𝑌 ∪ 𝐵 is not connected. Without loss of generality, suppose 𝑌 ∪ 𝐴 is
not connected. Then, there exist sets 𝑈, 𝑉 such that 𝑈 ∪ 𝑉 = 𝑌 ∪ 𝐴, 𝑈 ∩ 𝑉 = ∅. Note, U and
V are disjoint from B, as Y and A are disjoint from B. Define 𝑊 = 𝑈 ∪ 𝐵. Then, we can see that
𝑊 ∪ 𝑉 = 𝑋, (as 𝑊 ∪ 𝑉 = 𝑉 ∪ 𝑈 ∪ 𝐵 = 𝑌 ∪ 𝐴 ∪ 𝐵 = 𝑋). But W and V are a separation
of X: V is disjoint from U and B, so V is also disjoint from their union. This contradicts our
assumption that X is connected, and we thus know that 𝑌 ∪ 𝐴 is connected. Because we proved
this for a general set A, we also know that 𝑌 ∪ 𝐵 is connected, and thus we are finished.

Definition 3.6: Let X be a topological space. A covering 𝒞 of X is a collection of subsets of X

such that ∪ (𝐴 ∈ 𝒞) = 𝑋. If 𝒞 contains only open sets, then it is called an open covering.

Definition 3.7: Let X be a topological space. X is compact if for every open cover 𝒞 of X, there
is a finite subcover that also covers X.

Compactness is less intuitive than connectedness. In a way, it conveys

information about the “density” (dense means something else in topology but
ignore that for now) of a space. Imagine you’re a rich planter and you want to
protect your crops from thieves. Obviously, you hire guards to protect your estate,
but you can’t hire too many guards or else you’ll run out of money. So, you hire a
finite number of guards to watch your crops. Now, your guards can only protect
what they see. If, for some reason, some land is not watched, you will lose crops
and ultimately run out of business. Note, though, you can have an infinite number
of crops, they just have to be sufficiently “dense” such that a finite number of
watchmen can protect them. To be compact is to have a set that is capable of
being completely “watched” (covered) by a finite number of “watchmen” (open
sets).

Proposition 3.8: The finite union of compact subspaces of X is compact.

Proof: Suppose that we have finite open subspaces of X, 𝐴1 , … , 𝐴𝑛 . Then, for each cover of 𝐴𝑖 ,
there exist finite subcovers {𝐶𝑖 }𝑖∈𝐼 such that Ai =∪𝑖∈𝐼 (𝒞𝒾 ). Note, any covering of 𝐴1 ∪ … ∪ 𝐴𝑛
must be the union of covers of each 𝐴𝑖 . Note, though, every cover of each 𝐴𝑖 has a finite
subcover, so, for any finite union of any combination of covers, there exist a finite union of finite
subcovers of 𝐴1 ∪ … ∪ 𝐴𝑛 , which is finite. Thus, 𝐴1 ∪ … ∪ 𝐴𝑛 must be compact as every cover
has a finite subcover (defined as the finite union of finite subcovers of 𝐴𝑖 ).

Definition 3.9: A set S in a metric space X is totally disconnected if for any distinct 𝑥, 𝑦 ∈ 𝑆,
there exist separated sets A and B such that 𝑥 ∈ 𝐴, 𝑦 ∈ 𝐵, 𝑎𝑛𝑑 𝐴 ∪ 𝐵 = 𝑆.

Definition 3.10: Let X be a topological space. A point 𝑥 ∈ 𝑋 is an isolated point if {𝑥} is open
in X.

Theorem 3.11: A subset of ℝ𝑛 is compact if and only if it is closed and bounded.

Theorem 3.12: Let X be a non-empty compact Hausdorff space. If X has no isolated points, then
X is uncountable

Definition 3.13: Let 𝐴0 be the closed interval [0,1] in ℝ. Define

1 + 3𝑘 2 + 3𝑘
𝐴𝑛 = 𝐴𝑛−1 −∪∞
𝑘=0 ( , )
3𝑛 3𝑛

Let C =∩𝑛∈Z+ 𝐴𝑛 , which we call the Cantor set, or the Cantor ternary set.

Proposition 3.14: The Cantor set is totally disconnected.

Proof: Let 𝑎, 𝑏 ∈ 𝐶 be distinct points in the Cantor set, and without loss of generality, let 𝑎 < 𝑏.
Then, consider the interval (𝑎, 𝑏). By the construction of the Cantor set, we remove some
nonzero length from this interval. Take some point c in this removed interval such that 𝑎 < 𝑐 <
𝑏; this 𝑐 ∉ 𝐶. Define 𝐴 = (−∞, 𝑐) ∩ 𝐶 and 𝐵 = (𝑐, ∞) ∩ 𝐶. Note, A and B are separated sets,
𝑎 ∈ 𝐴, 𝑏 ∈ 𝐵, and 𝐴 ∪ 𝐵 = 𝐶. Thus, by Definition 3.9, C is totally disconnected.

Alternatively, consider the sum of all intervals, or equivalently, measures, removed by the
process of constructing the Cantor set:
∞ ∞
2𝑛 1 2𝑛 1 1
∑ ( 𝑛+1 ) = ∑ ( ) ( 𝑛 ) = ( ) ( )=1
3 3 3 3 1−2
𝑛=0 𝑛=0 3

Note, the Lebesgue measure of [0,1] is 1, and the measure of the intervals removed is also 1. We
are left with a set of measure 0, or a set of only points. Consequently, we have 2 cases left
following from the definition of a totally disconnected set. There exists a subspace (singletons)
of C that is connected, or the subspaces (singletons) of C are disconnected. If any subspace of C
is connected, then C is totally disconnected because there isn’t a non-singleton set that is
connected. If the subspaces in C are not connected, then C is vacuously totally disconnected as
there is no connected subspace of C at all.

Proposition 3.15: The Cantor set is compact.

Proof: Recall that a set is compact in ℝ if and only if it is closed and bounded. C is bounded as
𝐶 ⊂ [0,1]. C is closed as it is an intersection of closed sets. Therefore, C is compact.

Proposition 3.16: The Cantor set has no isolated points.

Proof: It suffices to show that there is no open interval U of ℝ, such that, with respect to the
subspace topology on C, 𝑈 ∩ 𝐶 = {𝑥}, for some 𝑥 ∈ 𝐶

Recall the definition of a subspace topology. In this case, the topology on C is the intersection of
all open sets of ℝ with C. Let 𝑈 be some interval (𝑥 − 𝜀, 𝑥 + 𝜀) for some 𝜀 > 0 and 𝑥 ∈ 𝐶. The
point x must be some element of C, for if it isn’t, for some sufficiently small epsilon, 𝑈 ∩ 𝐶 = ∅.
To guarantee 𝑈 ∩ 𝐶 is non-empty, we require U to be centered at some point 𝑥 ∈ 𝐶. I assert that
for any such 𝑈, 𝑈 ∩ 𝐶 contains at least 2 points. We divide this proof into 2 cases: x is an
endpoint, or x is not an endpoint.

Case 1: Suppose 𝑥 ∈ 𝐶 is an endpoint of some 𝐴𝑘 . Then 𝑥 ∈ 𝑈 ∩ 𝐶; we need to find one more

point 𝑦 ∈ 𝑈 ∩ 𝐶. Recall that, in the construction of the Cantor set, the removed intervals have
1 1
length 3𝑛 for some 𝑛 ∈ ℕ. Choose some 𝑛 such that 3𝑛 < 𝜀; then in the n+1st step of constructing
𝐶 you will find some interval (𝑐, 𝑑) ⊂ (𝑥 − 𝜀, 𝑥 + 𝜀) removed in 𝐴𝑛 (draw a picture if this is not
obvious). From this, we see that c or d are in the intersection (𝑥 − 𝜀, 𝑥 + 𝜀) ∩ 𝐶. Without loss of
generality, assume c is in the intersection. Then 𝑥, 𝑐 ∈ (𝑥 − 𝜀, 𝑥 + 𝜀) ∩ 𝐶, for any 𝜀 > 0.
Therefore, in this case, there is no interval (𝑥 − 𝜀, 𝑥 + 𝜀) open in ℝ such that 𝑈 ∩ 𝐶 = {𝑥}, for
any endpoint 𝑥 ∈ ℝ.

Case 2: Suppose 𝑥 ∈ 𝐶 is not an endpoint. What we would like to find is an endpoint in 𝑈 ∩ 𝐶

for any 𝜀 > 0. If x is not an endpoint, then for any step n, 𝑥 ∈ 𝐴𝑛 . More specifically, there is an
interval [𝑎, 𝑏] ⊂ 𝐴𝑛 such that 𝑥 ∈ [𝑎, 𝑏]. Choose an 𝑛 ∈ ℕ such that for some interval 𝑥 ∈ [𝑎, 𝑏]
and [𝑎, 𝑏] ⊂ 𝑈. Then, 𝑎, 𝑏, 𝑥 ∈ 𝑈 ∩ 𝐶, for any 𝜀 > 0.

Because we showed that there is no singleton set open in C, C consequently has no isolated
points.

Proposition 3.17: The Cantor set is uncountable.

Proof: C is Hausdorff, has no isolated points, and is compact. From Theorem 3.12, C is
uncountable.

4. Separation Axioms

Definition 4.1: Let X be a topological space. It is called Hausdorff or 𝑇2 if for each distinct
𝑥, 𝑦 ∈ 𝑋, there exist disjoint sets A and B such that 𝑥 ∈ 𝐴 𝑎𝑛𝑑 𝑦 ∈ 𝐵.

One example of a Hausdorff space is ℝ. For any 𝑥, 𝑦 ∈ ℝ, there exist disjoint

|𝑥−𝑦| |𝑥−𝑦| |𝑥−𝑦| |𝑥−𝑦|
open sets (𝑥 − ,𝑥 + ) , (𝑦 − ,𝑦 + ) containing x and y,
2 2 2 2
respectively.

Theorem 4.2: A subspace of a Hausdorff space is Hausdorff.

Proof: Let X be a Hausdorff space. Let 𝑌 ⊂ 𝑋 be equipped with the subspace topology 𝒯𝑌 . Let
𝑥, 𝑦 ∈ 𝑌. Then, 𝑥, 𝑦 ∈ 𝑋. Because X is Hausdorff, there exist sets 𝐴 𝑎𝑛𝑑 𝐵 open in X such that
𝑥 ∈ 𝐴 𝑎𝑛𝑑 𝑦 ∈ 𝐵, 𝐴 ∩ 𝐵 = ∅. As 𝑥, 𝑦 ∈ 𝑌, then 𝐴 ∩ 𝑌, 𝐵 ∩ 𝑌 are open in Y. Note, though, that
(𝐴 ∩ 𝑌) ∩ (𝐵 ∩ 𝑌) = ∅. To show this, suppose that (𝐴 ∩ 𝑌) ∩ (𝐵 ∩ 𝑌) ≠ ∅. Then, as 𝑌 ⊂ 𝑋,
𝐴 ∩ 𝐵 ≠ ∅, which contradicts our assumption that X is Hausdorff. So, (𝐴 ∩ 𝑌) ∩ (𝐵 ∩ 𝑌) = ∅
and, 𝑥 ∈ 𝐴 ∩ 𝑌, 𝑦 ∈ 𝐵 ∩ 𝑌. Because we chose arbitrary points in Y, we know that Y is Hausdorff.

Definition 4.3: Let X be a topological space and suppose that one-point sets are closed. It is
called regular or 𝑇3 if for each 𝑥 ∈ 𝑋, and closed set 𝑈 (such that 𝑥 ∉ 𝑈), there exist disjoint
open sets A and B such that 𝑥 ∈ 𝐴 𝑎𝑛𝑑 𝑈 ⊂ 𝐵.

ℝ is also regular (and normal) under the standard topology. A space equipped
with the discrete topology is regular (and normal). Every finite regular space is
also normal. The Sorgenfrey plane is regular but not normal (the proof for this is
challenging).

Definition 4.4: Let X be a topological space and suppose that one-point sets are closed. It is
called normal or 𝑇4 if for each distinct closed set, U and V, there exist disjoint open sets A and B
such that 𝑉 ⊂ 𝐴 𝑎𝑛𝑑 𝑈 ⊂ 𝐵.

Figure 4.5:

Lemma 4.6: Let X be a topological space, and let one-point sets be closed in X

1) X is regular if and only if given a point x of X and a neighborhood U of x, there is a

neighborhood V of x such that 𝑉 ⊂ 𝑈
2) X is normal if and only if given a closed set A and an open set U containing A, there is an
open set V containing A such that 𝑉 ⊂ 𝑈

Proposition 4.7: Let X be a regular space. Then for any distinct 𝑥, 𝑦 ∈ 𝑋, x and y have
neighborhoods whose closures are disjoint.

Proof: Let A be some neighborhood of x. Then, because X is regular, there exists some open sets
B and 𝐶 such that 𝑥 ∈ 𝐴 ⊂ 𝐵, 𝑦 ∈ 𝐶 𝑎𝑛𝑑 𝐵 ∩ 𝐶 = ∅. By Lemma 4.6, there exists some open set
D such that 𝑦 ∈ 𝐷 ⊂ 𝐶. Note that, because 𝐵 ∩ 𝐶 = ∅, 𝐴 ∩ 𝐷 = ∅. Then A and D are
neighborhoods of x and y such that their closures are disjoint.
Proposition 4.8: Let X be a normal space. Then for any disjoint closed sets 𝐴 𝑎𝑛𝑑 𝐵, 𝐴 𝑎𝑛𝑑 𝐵
have neighborhoods whose closures are disjoint.

Proof: Let A and B be disjoint closed sets. Then, there exists open sets U and V such that 𝐴 ⊂
𝑈, 𝐵 ⊂ 𝑉, 𝑈 ∩ 𝑉 = ∅. By Lemma 4.6, there also exist open sets 𝑊, 𝑍 such that 𝐴 ⊂ 𝑊 ⊂ 𝑈, and
𝐵 ⊂ 𝑍 ⊂ 𝑉. Then, as 𝑈 ∩ 𝑉 = ∅, 𝑊 ∩ 𝑍 = ∅. Therefore, for any disjoint closed set A and B,
there exist open neighborhoods W and Z such that 𝑊 ∩ 𝑍 = ∅.

Theorem 4.9: A closed subspace of a normal space is normal.

Proof: Let Y be a closed subspace of a normal space X. Take some disjoint closed sets 𝐴, 𝐵 ⊂ 𝑌;
then 𝐴 = 𝐴𝑋 ∩ 𝑌, 𝐵 = 𝐵𝑋 ∩ 𝑌. But 𝐴𝑋 and 𝐵𝑋 must be closed in X. Then, there exist sets 𝐶𝑋 , 𝐷𝑋
open in X such that 𝐴𝑋 ⊂ 𝐶𝑋 , 𝐵𝑋 ⊂ 𝐷𝑋 , 𝑎𝑛𝑑 𝐶𝑋 ∩ 𝐷𝑋 = ∅. Then define sets 𝐶 = 𝐶𝑋 ∩ 𝑌 and 𝐷 =
𝐷𝑋 ∩ 𝑌. C and D are open in Y and are disjoint open sets that separate A and B. Because we
showed this for arbitrary closed sets A and B, we may conclude that the closed subspace of a
normal space is normal.

Lemma 4.10: A subspace of a regular space is regular

Proof: Let X be a regular space and let 𝑌 ⊂ 𝑋. Choose some point 𝑥 ∈ 𝑌 and closed set 𝑈 ⊂ 𝑌;
then 𝑥 ∈ 𝑋, and U = V ∩ 𝑋, for some set V closed in X. Then, as X is regular, there exist disjoint
open sets 𝐴, 𝐵 such that 𝑥 ∈ 𝐴 𝑎𝑛𝑑 𝑉 ⊂ 𝐵. Then, 𝐴 ∩ 𝑌, 𝐵 ∩ 𝑌 are disjoint open sets that contain
x and U, respectively.

Theorem 4.11: X and Y are regular if and only if 𝑋 × 𝑌 is regular.

Proof: Suppose X and Y are regular. Choose some point 𝑥 × 𝑦 ∈ 𝑋 × 𝑌 and a neighborhood 𝑈 of
𝑥 × 𝑦. U may be written as a product of open sets 𝐴 × 𝐵, where 𝐴 is open in 𝑋, and 𝐵 is open in
𝑌. Note, 𝑥 ∈ 𝐴 and 𝑦 ∈ 𝐵. By Lemma 4.6, there exist open sets 𝐶, 𝐷 such that 𝑥 ∈ 𝐶 ⊂ 𝐴 and
𝑦 ∈ 𝐷 ⊂ 𝐵. Then, it is easy to see that 𝑥 × 𝑦 ∈ 𝐶 × 𝐷 ⊂ 𝐴 × 𝐵, which by Lemma 4.6 implies
that 𝑋 × 𝑌 is regular, as 𝐶 × 𝐷 = 𝐶 × 𝐷.

For the other direction, suppose 𝑋 × 𝑌 is regular. Then 𝑋 × {𝑦} is regular by Lemma 4.10; but it
can be readily shown that 𝑋 × {𝑦} is homeomorphic to 𝑋, so 𝑋 is regular. Again, take a subspace
𝑌 × {𝑥 }, this is regular and homeomorphic to Y, so Y is regular.

In fact, it can be proven that the arbitrary product of regular spaces is regular, not
just the finite product.
 5. References

All figures, exercises, theorems, and definitions come directly from Topology by Munkres.

• Munkres, James Raymond. Topology. Pearson, 2018.