1.

Introduction
1 Motivation and foreshadowing
Most of the fundamental concepts in this course are generalizations of concepts with which you
are familiar from previous courses in analysis and/or linear algebra. For example, you have
likely encountered of at least a few of the following concepts:

• The notions of open intervals of the form (a, b) and closed intervals of the form [a, b] in R.
More generally, the notion of an open subset of R2 or Rn , of whose definitions we will see
in more detail later.
The field of point-set topology (on which this course focuses) generalises this concept to a
much broader context.

• The definition of a continuous function. For example you probably recall that a function
f : R2 → R is continuous at a ∈ R2 if for every ε > 0 there is a δ > 0 such that kx − ak < δ
implies |f (x) − f (a)| < ε.
This definition is much more concisely and usefully stated in terms of open sets. We will
see this early on in this course.

• The idea of “closeness”. This is somewhat vague at the moment, but for example what
does it mean to say that two points in Rn are close to one another? What about two
subsets of Rn being close to another one? A point and a set?
In the context of Rn we are used to using expressions like “kx−ak < ε” to denote something
like “x is close to a”, but this requires some more context to make sense (namely, what is
ε?).
What if we want to discuss two points being close together in a context in which there is
no notion of distance?

• The notion of convergence of sequences. For example you probably recall that a sequence
{xn }∞ n n
n=1 in R is said to converge to a point x ∈ R if for every ε > 0 there is a N ∈ N
such that n > N implies kxn − xk < ε.
This is a concept we will be able to generalize significantly. Most notably we will see how
to define sequence convergence without a notion of distance.

• What does it mean for a subset of Rn , for example, to be “nice”? We recall a number
of theorems from first and second year calculus and/or analysis that require assumptions
about the domains of functions; for example the Extreme Value Theorem from single vari-
able calculus says that a continuous function defined on a closed and bounded interval

2018–
c Ivan Khatchatourian 1
1. Introduction 1.2. Topologies in a familiar setting

achieves a minimum and maximum value. Can we generalise this theorem after we gen-
eralise the definition of a continuous function, as we hinted at earlier? And if so, what is
the analogue of a “closed and bounded” domain in this more general setting?

These are some of the questions we will seek to answer in this course.

2 Topologies in a familiar setting

Broadly speaking, a topology on a given set X is simply a specified collection of subsets of X,
which we will call open sets, that satisfies certain properties.
Before we define what a topology is in general, we will define them in the familiar context
of Rn . Later in the course we will see that Rn with its “usual topology” is a very well-behaved
topological space (a metric space). For the moment, we will just present the basic idea and some
intuition behind it.
Throughout the following section, let x = (x1 , . . . , xn ) and y = (y1 , . . . , yn ) denote points in
Rn .
Recall the notion of an ε-ball around a point x in Rn :

Bε (x) = { y ∈ Rn : d(x, y) < ε }

p
where d(x, y) = (x1 − y1 )2 + · · · + (xn − yn )2 is the usual Euclidean distance between x and
y. (At the beginning of this note we used the notation kx − yk to refer to the same quantity.
The notation d(x, y) is preferable for us, because it will generalize later.) Using the idea of an
ε-ball, we had a definition of what it meant for a subset of Rn to be open:
Definition 2.1. A set U ⊆ Rn is called open if for every x ∈ U there is an ε > 0 such that
Bε (x) ⊆ U .
This is a fine definition, but it is very dependent on the specific structure of Rn . Most
notably it makes use of the fact that Rn has a notion of distance between points. However if we
were to carefully go through some of the theorems that have to do with open subsets of Rn , we
would find that we do not have to busy ourselves with the distance function. In fact, we would
find that much of what we need to show about open sets follows from these three properties:
Fact 2.2.
1. ∅ and Rn are both open.
2. The union of any collection of open sets (finite or infinite) is again open.
3. The intersection of finitely many open sets is open.
It turns out that these three are the fundamental properties that open sets should satisfy
to allow us to do useful things with them, like defining continuous functions, convergence of
sequences, and so on. Everything we do in this course will flow from asking what happens when
we fix a collection of sets satisfying these three properties.

2018–
c Ivan Khatchatourian 2
1. Introduction 1.3. Topologies and topological spaces

3 Topologies and topological spaces

Definition 3.1. Let X be a set. A collection T ⊆ P(X) of subsets of X is called a topology on
X provided that the following three properties are satisfied:

1. ∅ ∈ T and X ∈ T .
2. T is closed under finite intersections. That is, given any finite collection U1 , . . . , Un of
sets in T , their common intersection U1 ∩ · · · ∩ Un is also an element of T .
3. T is closed under arbitrary unions. That is, if { Uα : α ∈ I } is a family of sets in T
S
(here I is some indexing set, which may be infinite), then their union α∈I Uα is also an
element of T .

Given a set X and a topology T on X, the pair (X, T ) is called a topological space. We will
often conflate a topological space (X, T ) with its underlying set X if the topology in question is
clear from context.
The elements U ∈ T of a topology on X are called open subsets of X, or simply open sets.

This is our fundamental definition. By only dealing with open sets in topological spaces, we
will be able to solidify all of the concepts mentioned earlier in the introduction. Before going
any further, however, let’s see some examples.

Example 3.2. Let X = Rn , and let

Tusual := { U ⊆ Rn : ∀x ∈ U, ∃ε > 0 such that Bε (x) ⊆ U } .

As we already noted above, Tusual forms a topology on Rn . We call this the usual topology
on Rn , and refer to the topological space (Rn , Tusual ) as Rn with the usual topology. We will
often simply write Rnusual for convenience.
Of particular note is the case where n = 1. In this case our definition above takes the
following form:

Tusual = { U ⊆ R : ∀x ∈ U, ∃δ > 0 such that (x − δ, x + δ) ⊆ U } .

The nonempty open subsets of Rusual are precisely the open intervals you have been working
with since first year calculus—those intervals of the form (a, b), (a, ∞), (−∞, b), and (−∞, ∞) =
R for any real numbers a < b—along with arbitrary unions of them.

Exercise 3.3. Fix real numbers a < b. Explicitly show that the interval (a, b) is open in Rusual .
Show that the interval [a, b) is not open in Rusual .

Note that while the usual topology on R contains, for example, sets like (0, 1) ∪ (7, 8) ∪
(100, 127), we can describe all of these sets by specifying the usual open intervals, then allowing
all unions of them. The idea that an entire topology can be specified by some smaller collection

2018–
c Ivan Khatchatourian 3
1. Introduction 1.3. Topologies and topological spaces

of special open sets (open intervals in this case) along with arbitrary unions of them is an
important one. We will return to it in the next section.

The following two examples are somehow trivial sorts of topologies. In fact, the second of
them is often called exactly that! They are the simplest topologies we can define, and we can
define both of them on any nonempty set.

Example 3.4. Let X be any set. Define Tdiscrete := P(X). That is, Tdiscrete is the collection of
all subsets of X. Then Tdiscrete is called the discrete topology on X.

Example 3.5. Let X be any nonempty set. Define Tindiscrete := {∅, X}. Then Tindiscrete is called
the indiscrete topology on X, or sometimes the trivial topology on X.

These two are the “extremal” topologies on a given set, in the sense that they have the most
open sets and the fewest open sets, respectively; it is not possible to have more open sets than
the discrete topology, or to have fewer open sets than the indiscrete topology. There is quite a
substantial difference in their usefulness in mathematics, however.
The discrete topology comes up relatively frequently. It is even a metric space (which for
now you should just read as “very nice space”). That said, it still has some weird properties that
might make you uneasy. For example, every function whose domain is a discrete topological space
is continuous. For another example, the only sequences that converge in a discrete topological
space are the (eventually) constant sequences. Even though we have not yet formally defined
sequence convergence or continuity in a general topological context, these should strike you as
weird.
The indiscrete topology, on the other hand, practically never comes up while doing mathe-
matics. It is a barren, sad place. We only mention it when we have to, and we never go there if
we can help it. To give you a feeling for how awful indiscrete spaces are, we will later discover
that in an indiscrete topological space, every sequence converges to every point in the space. For
example in (R, Tindiscrete ), the constant sequence 0, 0, 0, 0, . . . converges to π. It also converges
to 7. What a mess!

Exercise 3.6. Fix an arbitrary nonempty set X, and let Tdiscrete and Tindiscrete be as above.
Show that these are both topologies on X.

We will see many different topologies on many different sets throughout the course, so our
current catalogue of topological spaces is not very exhaustive. Here are just a few more examples,
and any proofs that may be required will be left to the Big List.

Example 3.7. Let X = {4, , ♦, ♥}. Define

T := {∅, X, {4, , ♦}, {♦, ♥}, {♦}}.

Then (X, T ) is a topological space.

2018–
c Ivan Khatchatourian 4
1. Introduction 1.4. Comparing topologies

Example 3.8. Working with R as the underlying set, define

Tray := { (a, ∞) : a ∈ R } ∪ {∅, R}.

Then Tray is a topology on R that we will call the “ray topology”, for obvious reasons.

Example 3.9. Let X be any nonempty set. Define

Tco-finite := { U ⊆ X : X \ U is finite } ∪ {∅}.

Then Tco-finite is called the co-finite topology on X.

Example 3.10. Let X be any nonempty set. Define

Tco-countable := { U ⊆ X : X \ U is countable } ∪ {∅}.

Then Tco-countable is called the co-countable topology on X.

Example 3.11. Let X be a nonempty set, and fix an element p ∈ X. Define

Tp := { U ⊆ X : p ∈ U } ∪ {∅}.

Then Tp is called the particular point topology at p on X.

These last three examples are not quite as useless in “real” mathematics as the indiscrete
topology, but they are close. However, they will serve as testing grounds for some of the
properties of topological spaces we will soon define, and may serve to illustrate what a topology
can do for us.
For example, we will learn that in the topological space (R, T0 ) (that is, the reals with the
particular point topology at 0), the sequence { n1 }∞ n=1 does not converge to 0. In fact the only
sequences that do converge to 0 in this topological space are eventually equal to the constant
sequence 0, 0, 0, 0, . . . . Try to guess what point or points, if any, the sequence 1, 1, 1, 1, . . . might
converge to in this topology.

Exercise 3.12. Under what assumption(s) on the set X does Tco-finite = Tco-countable ?

4 Comparing topologies
It is often useful to be able to compare two topologies on the same set. To this end, we have
the following terminology.

Definition 4.1. Let X be a set, and suppose T1 and T2 are topologies on X. We say that T1
refines T2 , or that T1 is finer than T2 , if T1 ⊇ T2 . In other words, a topology with more open
sets is finer than a topology with fewer open sets.

2018–
c Ivan Khatchatourian 5
1. Introduction 1.4. Comparing topologies

We can (and will) also express this same idea by saying T2 is refined by T1 , or by saying T2
is coarser than T1 .
You may find this definition somewhat confusing at first, but it will become very natural as
you strengthen your intuition about topological spaces and what they can do. For now, imagine
that a finer topology on a set X can make finer (in the colloquial sense) distinctions between
points in X than a coarser topology.

Example 4.2. Thus far we have discussed two topologies specific to the set of real numbers:
the usual topology and the ray topology. We have also discussed the discrete and indiscrete
topologies that any nonempty set can have. For the reals, we have:

Tdiscrete refines Tusual , which refines Tray , which refines Tindiscrete .

Exercise 4.3. Fix a nonempty set X. Show that Tco-finite is coarser than Tco-countable . This
exercise may be tricky at first.

Remark 4.4. Given two topologies T1 and T2 on a set X, it may be that neither one refines
the other. In this case, we sometimes say T1 and T2 are incomparable topologies.

Example 4.5. Tusual and T7 (the particular point topology at 7) are incomparable topologies
on R. To see this, observe that the interval (1, 2) is open in the usual topology, but it does not
contain 7 and so is not open in T7 . This shows that Tusual 6⊆ T7 , or in other words that the
particular point topology does not refine the usual topology.
On the other hand, the set {π, 7} is open in T7 since it contains 7, but is not open in Tusual .
This shows that T7 6⊆ Tusual , or in other words that the usual topology does not refine the
particular point topology.
(The number 7 is, of course, not really important in this example. Tusual and Tp are incom-
parable for any p ∈ R.)

2018–
c Ivan Khatchatourian 6