3.

Closed sets, closures, and density

1 Motivation
Up to this point, all we have done is define what topologies are, define a way of comparing
two topologies, define a method for more easily specifying a topology (as a collection of sets
generated by a basis), and investigated some simple properties of bases.
At this point, we will start introducing some more interesting definitions and phenomena
one might encounter in a topological space, starting with the notions of closed sets and closures.
Thinking back to some of the motivational concepts from the first lecture, this section will
start us on the road to exploring what it means for two sets to be “close” to one another, or
what it means for a point to be “close” to a set. We will draw heavily on our intuition about
convergent sequences in Rn when discussing the basic definitions in this section, and so we begin
by recalling that definition from calculus/analysis.

Definition 1.1. A sequence {xn }∞ n

n=1 is said to converge to a point x ∈ R if for every  > 0
there is a number N ∈ N such that xn ∈ B (x) for all n > N .

Remark 1.2. It is common to refer to the portion of a sequence {xn }∞ n=1 after some index
∞
N —that is, the sequence {xn }n=N +1 —as a tail of the sequence. In this language, one would
phrase the above definition as “for every  > 0 there is a tail of the sequence inside B (x).”

Given what we have established about the topological space Rnusual and its standard basis
of -balls, we can see that this is equivalent to saying that there is a tail of the sequence inside
any open set containing x; this is because the collection of -balls forms a basis for the usual
topology, and thus given any open set U containing x there is an  such that x ∈ B (x) ⊆ U .
This observation will inform the definition of sequence convergence in a general topological
space, and in particular the definition of the “closure” of a set, as we will see shortly.
A note about the order in which we will cover this material: most textbooks on the subject
define what it means for a set to be “closed” first, then define closures of sets. We will use the
idea of a “closure” as our a priori definition, because the idea is more intuitive.

2 Closures
Definition 2.1. Let (X, T ) be a topological space, and let A ⊆ X. We define the closure of A
in (X, T ), which we denote with A, by:

x ∈ A if and only if for every open set U containing x, U ∩ A 6= ∅.

Or, in symbols:

A = { x ∈ X : ∀U ∈ T such that x ∈ U, U ∩ A 6= ∅ } .

2018–
c Ivan Khatchatourian 1
3. Closed sets, closures, and density 3.2. Closures

When there can be no confusion about the topological space with respect to which a closure is
being considered, we will simply write A without specifying “... in (X, T ).”

In words, x ∈ A if and only if any open set containing x also contains an element of A. After
we define what it means for a set to be closed, we will be able to present an alternate way of
defining the closure of a set.
Before going any further with examples we examine some elementary properties of closures,
the proofs of which use only the definition of closure and absolutely no cleverness or new ideas.

Proposition 2.2. Let (X, T ) be a topological space and let A, B ⊆ X. Then:

1. A ⊆ A.

2. A = A. (A mathematician interested in using fancy words would say that “taking closures
is an idempotent operation”.)

3. A ∪ B = A ∪ B.

4. If X \ A is open, then A = A.

5. Trivially, ∅ = ∅ and X = X.

Proof.

1. Let x ∈ A. Then clearly any open set U that contains x intersects A, since at least
x ∈ U ∩ A, and therefore x ∈ A.

2. By (1), A ⊆ A. On the other hand, let x ∈ A and let U be an open set containing x. We
need to show that U ∩ A 6= ∅. By assumption U ∩ A 6= ∅, so there is at least one point y
in this intersection. Since U is an open set containing an element y in the closure of A,
U ∩ A 6= ∅ by definition of A.

3. Exercise.

4. Assume X \ A is open. Again by (1) we know that A ⊆ A, so it remains to show that

A ⊆ A. Proceeding by contradiction, suppose x ∈ A\A ⊆ X \A. Since X \A is an open set
containing the element x of A, by definition of closure we must have that (X \ A) ∩ A 6= ∅,
which is obviously impossible. This means there can be no points in A \ A.

As promised, here are some examples in familiar settings.

Example 2.3.

2018–
c Ivan Khatchatourian 2
3. Closed sets, closures, and density 3.2. Closures

1. Working in Rusual , the closure of an open interval (a, b) is the corresponding “closed”
interval [a, b] (you may be used to calling these sorts of sets “closed intervals”, but we have
not yet defined what that means in the context of topology).
To see this, by 2.2.1 we have that (a, b) ⊆ (a, b). We first show that a, b ∈ (a, b). Let U
be an open set containing a. Then there is an  > 0 such that B (a) = (a − , a + ) ⊆ U .
Let δ = min{, b − a}. Then a + 2δ ∈ U ∩ (a, b), as required. The proof for b is essentially
the same. This establishes that [a, b] ⊆ (a, b).
Finally, if x ∈ R \ [a, b], the set (−∞, a) ∪ (b, ∞) is an open set containing x and disjoint
from (a, b), witnessing that x ∈ / (a, b).

2. Again in Rusual , show the following straightforward facts for any a < b < c:

(a) {a} = {a}.

(b) [a, b) = [a, b].
(c) (a, b) ∪ (b, c) = [a, c].
(d) As you might expect from the name “closed interval”, [a, b] = [a, b].
(e) Let A = n1 : n ∈ N . Then A = A ∪ {0}.


(Hint: Use the fact that the elements of A, as a sequence in the obvious order,
converge to 0 in the sense you are familiar with as defined in the first section.)
(f) More generally, if {xn }∞n=1 is a sequence that converges to a point x ∈ R (in the
first-year calculus sense), then { xn : n ∈ N } = { xn : n ∈ N } ∪ {x}.
(By the way, note that { xn : n ∈ N } and {xn }∞ n=1 are different objects. The former
is a set of numbers, while the latter is an ordered list of numbers or more properly a
function.)

3. Generalizing our first example, let x ∈ Rn and  > 0. Then the closure of an -ball B (x)
in Rnusual is { y ∈ Rn : d(x, y) ≤  }.

4. It may seem as though any singleton is its own closure, but this need not be true in all
topological spaces. For example, let X = {0, 1} and let T = {∅, X, {1}}. Then (X, T ) is
a topological space, and {1} = X. Check this.

5. For a less trivial example of the same idea, let (R, Tray ) be the reals with the ray topology
defined in the first set of notes (in which the open sets are ∅, R, and rays of the form
(a, ∞)). Then {7} = (−∞, 7].
To see this, first note that if x ∈ (7, ∞), then (7, ∞) is itself an open set that is disjoint
from {7}, so x ∈ / {7}. If x ≤ 7, then any open set containing x is of the form (7 − t, ∞)
for some t > 0, and in particular any such set contains 7, showing that x ∈ {7}.

2018–
c Ivan Khatchatourian 3
3. Closed sets, closures, and density 3.3. Closed sets

We will see later in the course that the property “singletons are their own closures” is a
very weak example of what is called a “separation property”. Topological spaces that do
not have this property, like in this and the previous example, are pretty ugly.

6. In (X, Tdiscrete ), for any A ⊆ X, A = A. In other words, every set is its own closure.

7. In (X, Tindiscrete ), for any nonempty A ⊆ X, A = X.

8. In the Sorgenfrey line:

(a) Show that singletons are their own closures. For example, show that {7} = {7}.
(b) For any a < b ∈ R, show that [a, b] = [a, b] and [a, b) = [a, b). This latter fact, which
in this context says that basic open sets are their own closures, is weird. Keep it in
mind for much later in the course.
(c) On the other hand, (a, b) = [a, b). Indeed, first of all if x ≥ b, then [x, x + 1) is an
open set containing x and disjoint from (a, b), so x ∈ / (a, b). Second of all if x < a,
a−x
then [x − 1, x + 2 ) is an open set containing x and disjoint from (a, b), so x ∈/ (a, b).
Finally, a ∈ (a, b) since if U is an open set containing a, then there is an  > 0 such
that [a, a + ) ⊆ U , and [a, a + ) ∩ (a, b) 6= ∅.
(d) Show that if A = n1 : n ∈ N , then A = A ∪ {0}. This is the same as it was in


Rusual
(e) On the other hand, if B = − n1 : n ∈ N , show that B = B. Once we have a


general topological definition of sequence convergence, we will be able to see that this
sequence does not converge in this topological space.

3 Closed sets
In this section we finally introduce the definition we have been tiptoeing around.

Definition 3.1. A subset A of a topological space X is said to be closed if X \ A is open.

Caution: “Closed” is not the opposite of “open” in the context of topology. A

subset of a topological space can be open and not closed, closed and not open, both open and
closed, or neither. We will see some examples to illustrate this shortly.
Cautionary note aside, the definition seems simple enough right? Again, here are some
observations that follow immediately from this definition and previous definitions in the course,
along with De Morgan’s laws.

Proposition 3.2. Let (X, T ) be a topological space.

1. X and ∅ are both closed.

2. The union of finitely many closed sets is closed.

2018–
c Ivan Khatchatourian 4
3. Closed sets, closures, and density 3.3. Closed sets

3. An arbitrary intersection of closed sets is closed.

Proof. The proofs of (1) and (3) are left as exercises.

For (2), let C1 , . . . , CN be closed subsets of X. We have to show that
N
[
X\ Ck
k=1

is open. By De Morgan’s laws, we immediately have

N
[ N
\
X\ Ck = (X \ Ck ),
k=1 k=1

which is open since it is the intersection of finitely many open sets.

As you might suspect from this proposition, or indeed from the definition of a closed set alone,
one can completely specify a topology by specifying the closed sets rather than by specifying the
open sets as we have been doing thus far. To be more precise, one can “recover” all the open
sets in a topology from the closed sets, by taking complements. There are equivalent notions of
“basic closed sets”, and so on. For example, given a set X we can define the co-finite topology
on X equivalently as the topology in which the closed sets are precisely the finite sets. We will
not be defining topologies in this way much (if at all) but it is nice to know it can be done, and
is sometimes more convenient.
One question that is still lingering: What do closed sets have to do with closures of sets?

Proposition 3.3. Let (X, T ) be a topological space, and let A ⊆ X. Then A is closed if and
only if A = A.

Proof. (⇒) Assume A is closed. We have already shown that A ⊆ A, so let x ∈ X \ A. But
then X \ A itself is an open set containing x and disjoint from A, so x ∈ / A. This shows that
X \ A ⊆ X \ A, or in other words that A ⊆ A.
(⇐) Assume A = A. We want to show that X \ A is open. Fix x ∈ X \ A = X \ A. Then by
definition of A there is an open set U containing x and disjoint from A. That is, x ∈ U ⊆ X \ A.
Therefore X \ A is open, as required.

Another connection between these two concepts is the following, which is often taken as the
definition of the closure of a set in contexts where “closed” is defined before “closure”.

Proposition 3.4. Let (X, T ) be a topological space, and let A ⊆ X. Then A is the intersection
of all closed subsets of X that contain A.

Proof. Exercise. (See the Big List.)

Remark 3.5. It should be noted here that there is a notion which is somehow dual to the idea
of the closure of a set called the interior of a set. You will explore this in a Big List problem.

2018–
c Ivan Khatchatourian 5
3. Closed sets, closures, and density 3.4. Density

On to some examples surrounding closed sets. We have done almost all of the work already,
so check all of these results yourself.

Example 3.6.

1. In (X, Tindiscrete ), the only closed sets are ∅ and X.

2. In (X, Tdiscrete ), every subset of X is closed.

3. In Rusual , let a < b. Then {a}, [a, b], [a, ∞), and (−∞, b] are closed. (a, b), [a, b) and
A = n1 : n ∈ N are not closed.


4. In (X, Tco-countable ), a subset A of X is closed if and only if A is countable or A = X.

This topology can just as easily be specified by saying that X is closed and all countable
subsets of X are closed.

5. Let X be a set and let p ∈ X. Consider Tp , the particular point topology on X. A proper
subset A of X is closed in this topology if and only if p ∈
/ A.

6. In the Sorgenfrey Line, let a < b. Then [a, b] and [a, b) are closed. This should feel
weird. Sets of the form [a, b) are basic open sets in the Sorgenfrey line, yet they are also
closed. Subsets of a topological space that are both open and closed are called clopen
sets (mathematicians are bad at naming things). Later in the course, when we talk about
a property called connectedness, we will see that a topological space that has a basis of
clopen sets is highly disconnected.
(a, b), (a, b], and A = n1 : n ∈ N are not closed in this space. B = − n1 : n ∈ N is
 

closed, however.

4 Density
By now you should have some intuition saying that the points in the closure of a set A are
“close” to A. Phrased the other way around, A is the collection of points that A is close to. In
this metaphorical framework, a “dense” set is one that is close to everything.
We will not have many applications for these ideas for the moment, but they will become
much more important later.

Definition 4.1. Let (X, T ) be a topological space. A subset D ⊆ X is said to be dense if D = X.

We can immediately rephrase this definition in an equivalent form that talks about open
sets.

Proposition 4.2. Let (X, T ) be a topological space, and let D ⊆ X. Then the following are
equivalent:

2018–
c Ivan Khatchatourian 6
3. Closed sets, closures, and density 3.4. Density

1. D is dense.

2. For every nonempty open U ⊆ X, D ∩ U 6= ∅.

Proof. Exercise. (See the Big List.)

We will henceforth treat both of the above properties as definitions of “dense”, and use them
interchangeably.

During the first lecture we said that one of the ideas we would try to formalize is the idea of
“largeness”, and density is one way to that.
Let’s see some examples. Most of the proofs are immediate or ones you have actually done
before, though you didn’t know the words at the time. Some others are left as exercises on the
Big List.

Example 4.3.

1. In any topological space (X, T ), X is dense in itself.

2. If (X, T ) is a topological space, D ⊆ X is dense, and D ⊆ A, then A is dense.

3. Working in Rusual :

(a) The set Q of rationals is dense. You proved this on the Big List. Notably this means
R has a countable dense subset. Topological spaces with this property are called
separable. We will investigate this property more later in the course.
(b) The set R \ Q of irrationals is dense.
(c) The set N is not dense.
(d) No finite set is dense.

4. Q is also dense in the Sorgenfrey Line. You also proved this on the Big List. The irrationals
are also dense here.

5. In (R, Tray ), N is dense. More generally, any subset of R that is not bounded above is
dense.

6. In (X, Tdiscrete ), the only subset of X that is dense is X itself.

7. In (X, Tindiscrete ), every nonempty subset of X is dense.

8. Let p ∈ X and consider the particular point space (X, Tp ). Then {p} is dense, but no other
singleton is dense. Try to come up with a condition on subsets of X that is equivalent to
being dense in this space.

9. If X is infinite, then any infinite subset of X is dense in (X, Tco-finite ). If X is uncountable,

then any uncountable subset of X is dense in (X, Tco-countable ).

2018–
c Ivan Khatchatourian 7