Closed, Open, Clopen

by Marcus Math

🔍

## Closed + Open = Clopen?

In many branches of mathematics, we think in terms of distances and measurements. Topology takes a different approach. Instead of asking *how far apart* two points are, it asks questions about the structure of sets and the behavior of points near them:

* Which points belong to the **interior** of a set?
* Which points lie on its **boundary**?
* Which points are arbitrarily close to the set, even if they are not contained in it?
* Can a set be split into separate pieces, or is it all connected?

To answer these questions, mathematicians developed several fundamental concepts:

* **Open Sets** 
    - Sets that contain a neighborhood around each of their points.
* **Closure**
    - The process of adding all points that lie arbitrarily close to a set.
* **Closed Sets** 
    - Sets that already contain all of their limit points.
* **Clopen Sets** 
    - Sets that are simultaneously open and closed.
* **Connected Sets** 
    - Sets that cannot be separated into disjoint pieces.

Together, these concepts form the foundation of topology and provide the language used to study continuity, convergence, compactness, and many of the central ideas of real analysis.

In this post, we'll build these ideas from the ground up and discover why a set can sometimes be both **closed** and **open** at the same time.

*Note* please read the blog post [Metric Spaces vs Inner Product Spaces](https://www.noteblogdoku.com/blog/metric-space-vs-inner-product-space) as a prerequisite.

## Open Sets

In a metric space $(X, d)$, where $d$ is a distance function acting on set $X$, a set $U \subseteq X$ is open if:

For every point $x \in U$, $ \exists \varepsilon > 0$ such that the **open ball**
$$B(x, \varepsilon) = \{ y \in X \mid d(x, y) < \varepsilon \} $$
is entirely contained in $U$.

In other words, around every point in $U$, there is a small "neighborhood" (open ball) that still lies completely inside $U$. This also means that no matter how close we approach the border of the set, that we can always make a ball around the subset with non-zero radius, such that all points lie within $X$.

Intuitively, an open set $U$ is a set that doesn't include its boundary. Whenever we are within $U$, we are allowed to get arbitrarily close to the edge and still stay within $U$.

### Examples

#### Example 1: The interval $(0,1)$ is open in $(\mathbb{R},d)$ with $d(x,y) = |x-y|$

Let $x \in (0,1)$ be arbitrary, so $0 < x < 1$. Define
$$\varepsilon = \min(x,\ 1-x).$$
Since $x > 0$ and $1 - x > 0$, we have $\varepsilon > 0$.

We claim that 
$$B(x,\varepsilon) = (x-\varepsilon,\ x+\varepsilon) \subseteq (0,1)$$

Take any $y \in B(x,\varepsilon)$, i.e. $|y - x| < \varepsilon$, which means $x - \varepsilon < y < x + \varepsilon$. Then:

- Lower bound: $y > x - \varepsilon \ge x - x = 0$, using $\varepsilon \le x$. So $y > 0$.
- Upper bound: $y < x + \varepsilon \le x + (1 - x) = 1$, using $\varepsilon \le 1 - x$. So $y < 1$.

Thus $0 < y < 1$, so $y \in (0,1)$. Since $y$ was arbitrary, $B(x,\varepsilon) \subseteq (0,1)$.

So every $x \in (0,1)$ has an open ball around it contained in $(0,1)$, so $(0,1)$ is open.

<center>
$\textbf{Figure 1}$: The Interval $(0,1)$ which is Open in $\mathbb{R}$ since for every Point we choose inside, there exists an Open Ball in $(0,1)$
</center>

### Interior Points

The definition of an interior point is a point $x \in S$ is an interior point of set $S$ if there exists an open set $U$ such that 
$$ x \in U \subset S$$

In other words, there is some open set that contains $x$ and is entirely contained inside $S$. So basically given that we have a set $S$, an interior point has a ball (or some other set around it) such that no matter how arbitrarily close we get to the edge of $U$ we never reach it. Since $U \subset S$, this means that we never reach the edge of $S$ either.

Equivalently: there exists $  \varepsilon > 0  $ such that $  B(x, \varepsilon) \subseteq S  $.

From our previous example, we can see that every $x \in \mathbb{R}$ is an interior point of $\mathbb{R}$.

<img src="/media/images/20260601_230730_06_01_2026_15_34_02_open_set_animation.gif" alt="06 01 2026 15 34 02 open set animation" style="max-width: min(700px, 80%); height: auto;">
<center>
$\textbf{Figure 2}$: Animation of an Open Set $X$ and Recursively Approaching the Edge Arbitrarily Closely yet Not Leaving $X$.

$p$ is an Interior Point of $X$
</center>

## Closure 
In a metric space, the concept of closure is a prerequisite for understanding closed sets and completeness. To understand closure however we first need to define **Limit Points** of a set.

### Limit Points

Let $(X,d)$ be a metric space (set $X$, distance function $d$) with $A \subseteq X$. A point $x \in X$ is a **limit point** (aka **accumulation point**) of $A$ if every open ball around $x$ contains at least one point of $A$ different from $x$.

Formally $x$ is a limit point of $A$ if for every $r > 0$, the open ball 
$$B(x,r) = \{y \in X | d(x,y) < r\}$$
satisfies 
$$B(x,r) \cap (A \setminus \{x\} \neq \emptyset $$

I.e. no matter how small a neighborhood you draw around $x$, you will always find points of $A$ (besides $x$) inside it. The point $x$ is the limit the set $A$ approaches. You can find infinitely many points of $A$ near $x$, or at least points other than $x$ getting arbitrarily close to $x$.

Intuitively, they're called **limit** points, since in the **limit** of a sequence of points, you approach the **limit** point $x$.

<center>
$\textbf{Figure 3}$: Illustration of a Limit Point for some Arbitrary Sequence approaching $x \in X$
</center>

By contrast an **Isolated Point** is a point $x \in A$ that's not a limit point. For said point $x$, there exists some ball around $x$ that contains no other points of $A$.

An intuitive example of this is imagine a donut (a circle or a toroidal surface) in the plane or in space, and then you add one extra point floating in the middle of the hole, not touching the donut. That lone point in the center is an isolated point of the set.

An example for an isolated point is the set $A = \{0\} \cup [1,2]$ in $\mathbb{R}$. The point 0 is clearly isolated here since there's no way to approach this point from any other points within the set $A$. However, all points in $[1,2]$ are limit points.

<img src="/media/images/20260602_003625_Screenshot_2026-06-01_at_5.34.18_PM.png" alt="Screenshot 2026 06 01 at 5.34.18 PM" style="max-width: min(500px, 80%); height: auto;">
<center>
$\textbf{Figure 4}$: Isolated Point Graphical Example showing the Example Described
</center>

### Closure

Great, now we can define what a closure is in a metric space. For a given metric space $(X,d)$ with $A \subseteq X$ the closure of $A$, denoted $cl(A)$ is the set of all points in $X$ that are 
- Points in $A$ or
- Limit points of $A$

Thus formally we have 
$$ cl(A) = A \cup limit(A)$$
where $limit(A)$ is the set of all limit points of $A$.

Equivalently, we can use the Neighborhood definition. $x \in cl(A)$ if and only if every open ball around $x$ intersects $A$. I.e.
$$ \forall \varepsilon > 0 \quad B(x,\varepsilon) \cap A \neq 0$$

Or we could use the sequential definition stating that $x \in cl(A)$ if and only if there exists a sequence $(x_n) \in A$ such that $x_n$ approaches $x$ in the limit.

Intuitively, the closure $cl(A)$ is the set $A$ with all of its holes filled in, and includes every point that can be approached by points from $A$. It is the smallest closed set that contains $A$, and if $A$ is closed then $cl(A) = A$.

In this definition, limit points $limit(A)$ are points in the closure that may or may not be in $A$, and isolated points are points in $A$ that are in the closure, but are not limit points. So we have that

$$ cl(A) = \text{Isolated points of } A \cup \text{Limit points of } A$$

### Examples

#### Example 1: $A = (0,1) \subset \mathbb{R}$
For the open interval $A$, if we add the two endpoints 0,1, which are limit points we get the closure of $A$, thus
$$ cl(A) = [0,1] $$

#### Example 2: $A = \{(x,y) \in \mathbb{R}^2 \mid x^2 + y^2 = 1\} \ \cup \ \{(0,0)\}$

As we can see $A$ is the unit circle in $\mathbb{R}^2$ unioned with the origin, similar to what we described previously with the donut. We can basically break this into the two parts, the circle and the origin.

For the circle $C = \{(x,y) \in \mathbb{R}^2 \mid x^2 + y^2 = 1\}$, we have that this is a closed set in $\mathbb{R}^2$ since every point in the circle is a limit point of the circle. So the closure of the circle alone is the circle: $cl(C) = C$.

The origin $\{0,0\}$ is isolated. We have that $\{0,0\} \in A$, but it's not a limit point of $A$. We know this is the case since we can draw a ball around the origin such that the ball lies entirely around the hole of the donut, and contains no points of $C$. Thus there's a neighborhood of the origin that intersects $A$ only at the origin (definition of isolated point).

Thus we can see that $cl(A) = A$ and it already contains all its limit points.

## Closed Sets

The most used definition for a closed set, is a set $A$ in a metric space $(X,d)$ such that its complement $X \setminus A$ is an open set in $X$.

An example of this in $\mathbb{R}$ is the interval $[0, 1]$. $[0,1]$ is closed because its complement $ (-\infty, 0) \cup (1, \infty)$ is open.

Another example is the set $\mathbb{Z}$ (integers). $\mathbb{Z}$ is closed in $\mathbb{R}$ because its complement is the union of open intervals between integers.

Of note the following theorems are quite useful for open and closed sets, but we will not prove them here. These theorems would be used for proving that $\mathbb{Z}$ is closed since the complement of $\mathbb{Z}$ is a union of open sets, which is open (see 2).

Let $(X,d)$ be a metric space, then:

1. If $V_1, V_2, ... V_k$ are open subsets of $X$, then 
$$ \cap_{j=1}^k V_j$$
is also open (note finite $k$). A finite intersection of open sets is open.

2. If $\{ V_{\lambda} \}_{\lambda \in I}$ is an arbitrary collection of open subsets of $X$ (not necessarily finite) then
$$ \cup_{\lambda \in I} V_{\lambda} $$
is also open. A union of open sets is open.

3. If $\{E_{\lambda}\}_{\lambda \in I}$ is an arbitrary collection of closed subsets of $X$ then $\cap_{\lambda \in I} E_\lambda$ is also closed. 
An intersection of closed sets is closed.

4. If $E_1, E_2, ... E_k$ are closed subsets of $X$, then 
$$ \cup_{j=1}^k E_j $$
is also closed. A finite union of closed sets is closed.

*Fun fact* you can easily prove 3,4 from 1,2 using De Morgan's Laws and some basic rules about sets! I could put the proofs for these in the Appendix, but this blog is already quite long, so I recommend you look at **Basic Analysis 1** (Introduction to Real Analysis, Volume 1) by *Jiri Lebl*, which is well written.

Another equivalent definition for a closed set is that a set $A$ is closed if and only if it contains all its limit points. This means that the set can have isolated points too, but it's closed as long as it has all its limit points.

Referring to the example in **Figure 4** where we have a donut with a single point in the middle, we showed in the last section that this set contains all its limit points. Thus we can say that this set is closed!

For intuition, closed sets have "no holes" at their boundary — any point you can approach arbitrarily closely by points in $A$ must already be in $A$.

<img src="/media/images/20260602_081806_Screenshot_2026-06-02_at_1.17.53_AM.png" alt="Screenshot 2026 06 02 at 1.17.53 AM" style="max-width: min(500px, 80%); height: auto;">
<center>
$\textbf{Figure 5}$: Example showing Closed Set with Boundary Points (Outline) where Outline are Perimeter Limit Points 
</center>

Finally, we have that $A$ is closed if and only if $A = cl(A)$.

Thus we have that all of the following are equivalent ways to defined closed sets:
- Complement of the set is open 
- The set contains all limit points 
- The set equals its own closure

## Boundary Points

<img src="/media/images/20260602_202927_Screenshot_2026-06-02_at_1.28.49_PM.png" alt="Screenshot 2026 06 02 at 1.28.49 PM" style="max-width: min(700px, 80%); height: auto;">
<center>
$\textbf{Figure 6}$: Interior, Boundary, and Exterior Points with respect to Set A
</center>

So far we have built two complementary constructions. The interior comes from the open set side, where we defined $x$ as an interior point of $A$ when some ball $B(x,\varepsilon)$ sits entirely inside $A$. The closure comes from the closed set side $x \in cl(A)$ when every ball around $x$ meets $A$.

The boundary is the object that sits between these two.

### The Interior of a Set
Collecting all the interior points of $A$ into one set gives the interior, written $int(A)$

$$ int(A) = \{\, x \in A \mid \exists\, \varepsilon > 0 \text{ such that } B(x,\varepsilon) \subseteq A \,\}$$

Equivalently, $int(A)$ is the largest open set contained in $A$. This is the mirror of the closure, which was the smallest closed set containing $A$.

Interior and closure are dual notions: closure grows $A$ outward to the smallest closed set around it, interior shrinks $A$ inward to the largest open set inside it.

Note that $A$ is open when $A=int(A)$ i.e. $\forall x \in A$ has an open ball around it, and $A$ is closed when $A = cl(A)$.

### Exterior and Boundary Points

The interior asks if there's an open ball inside $A$. Conversely, we can also ask if there's a ball inside the complement $X \setminus A$?. A point $x$ is an exterior point of $A$ if some ball $B(x,\varepsilon) \subseteq X \setminus A$. I.e. there exists a neighborhood around $x$ that has no points inside $A$. The exterior is the interior of the complement:

$$ ext(A) = int(X \setminus A)$$

But what about the points that are neither inside nor outside $A$? A point $x$ is a **boundary point** of $A$ if every ball around $x$ meets both $A$ and $X \setminus A$

$$ \forall \varepsilon >0: \quad B(x,\varepsilon) \cap A \neq \emptyset \quad and \quad B(x,\varepsilon) \cap (X \setminus A) \neq \emptyset$$

The **boundary** is the set of all such points, denoted $\partial A$. No matter how small you take the radius around a boundary point, there will always exist at least one point that's in $A$ and one point that's not in $A$.

These three categories, interior, exterior, boundary, are mutually exclusive and cover all cases and no point can be any two or more categories at once. Thus we can define the entire space $X$ as:

$$ X = int(A) \cup \partial(A) \cup ext(A) $$

The boundary condition treats $A$ and $X \setminus A$ equivalently, and we have that a set and its complement have the same boundary. Boundary points can be seen to be the points that exist in the overlap between the closures of $A$ and its complement, since these are the points that are approached from both sides.

$$ \partial A = cl(A) \cap cl(X \setminus A) $$

### Closure = Interior $\cup$ Boundary

Recall the neighborhood definition from the **Closure** section $x \in cl(A)$ if and only if every ball around $x$ meets $A$. In terms of our three categories of points: 
- An interior point has a ball inside $A$, so every ball meets $A$. Interior points lie in $cl(A)$
- A boundary point meets $A$ in every ball by definition. Boundary points lie in $cl(A)$
- An exterior point has at least one ball that does not have any points from $A$, so its not in $cl(A)$

Thus, the closure is exactly the interior unioned with the boundary
$$ cl(A) = int(A) \cup \partial A $$

Equivalently 
$$ int(A) = A \setminus \partial A$$

the closure adds the boundary, and the interior removes it.

## Clopen Sets

<img src="/media/images/20260602_220545_Screenshot_2026-06-02_at_3.05.34_PM.png" alt="Screenshot 2026 06 02 at 3.05.34 PM" style="max-width: min(750px, 80%); height: auto;">
<center>
$\textbf{Figure 7}$: A Clopen Set Example showing how a Set can be both Closed and Open
</center>

We now reach the strange object promised in the title. A set that is **open and closed at the same time** sounds like a contradiction. One can assume a set either includes its edge or it does not?

The resolution is that *open* and *closed* are not opposites but two separate conditions. Maybe we should've named these conditions differently.

### Definition

Let $(X,d)$ be a metric space. A set $A \subseteq X$ is **clopen** if it is both open and closed in $X$. The word is a portmanteau of *closed* and *open*.

Two sets are clopen in every metric space:

- The whole space $X$ is open (every ball lies in $X$ trivially) and closed (its complement $\emptyset$ is open).
- The empty set $\emptyset$ is open (the open-set condition holds vacuously, since there are no points to check) and closed (its complement $X$ is open).

We call $\emptyset$ and $X$ the **trivial** clopen sets. The interesting question is whether a space has others.

### The Empty Boundary Characterization

In the previous section we read open and closed straight off the boundary:

- $A$ is open $\iff A$ contains none of its boundary, $A \cap \partial A = \emptyset$.
- $A$ is closed $\iff A$ contains all of its boundary, $\partial A \subseteq A$.

A clopen set satifies both conditions: it contains both all of $\partial A$ and none of $\partial A$. 
The only way a set can contain all and none of something is for that something to be empty.

Therefore
$$ A \text{ is clopen} \iff \partial A = \emptyset. $$

Intuitively, a clopen set has **no edge anywhere**. There is no point in the entire space where you can stand and see both $A$ and its complement arbitrarily close by. $A$ and $X \setminus A$ are separate with empty space between them.

This is why clopen sets may feel paradoxical in familiar spaces, such as $\mathbb{R}^n$.
A blob in the plane, or an interval on the line, always has a visible edge, so its only hope of being clopen would be to have no boundary at all.

### Note: clopen depends on the ambient space

Whether a set is clopen is not just a property of the set, it also depends on the space $X$ the set lives in. The same collection of points can be clopen in one space and not in another, because the question "is there a ball that stays inside?" depends on which points are around in the first place. The examples below make this concrete.

### Examples

#### Example 1: In $\mathbb{R}$, the only clopen sets are $\emptyset$ and $\mathbb{R}$

Take any set $A \subseteq \mathbb{R}$ that is neither empty nor all of $\mathbb{R}$. Then $A$ contains some point and excludes some other point, and because the real line has no gaps, somewhere between those two points the set has to go from inside to outside. That point where we go from inside the set to outside is a boundary point, so $\partial A \neq \emptyset$ and $A$ is not clopen. On the real line, the trivial clopen sets are the only clopen sets. (We will make the "no gaps" argument precise in the next section; this is exactly the statement that $\mathbb{R}$ is connected.)

#### Example 2: A nontrivial clopen set in a disconnected space

Now let $X = [0,1] \cup [2,3]$, viewed as a metric space with the usual distance inherited from $\mathbb{R}$. Consider $A = [0,1]$ (this is the situation drawn in Figure 7).

- **$A$ is open in $X$.** Take any $x \in [0,1]$ and the ball $B(x, \tfrac{1}{2})$ measured inside $X$. It can only reach points within distance $\tfrac12$ of $x$, so it never gets past $1.5$, and every point of $[2,3]$ sits at distance at least $1$ from $[0,1]$. The ball therefore meets only $[0,1]$, even at the endpoints $0$ and $1$. So a small ball around any point of $A$ stays inside $A$ (any radius below $1$ works).
- **$A$ is closed in $X$.** Its complement in $X$ is $[2,3]$, which is open in $X$ by the identical argument.

So $A = [0,1]$ is both open and closed in $X$: a nontrivial clopen set. And indeed $\partial A = \emptyset$ in $X$ the endpoint $1$ is not a boundary point here, because every small ball around it (inside $X$) meets $[0,1]$ but never reaches $[2,3]$. The empty gap between the two pieces is precisely what kills the boundary.

Crucially, $[0,1]$ is not clopen in $\mathbb{R}$. There it is closed but not open, since any ball around $1$ spills out past $1$. The clopen-ness appeared only because we discarded the points between 1 and 2, splitting the space into two separated pieces. This is the relativity warning in action.

#### Example 3: The discrete metric, where everything is clopen

At the opposite extreme, take any set $X$ with the discrete metric, $d(x,y) = 1$ if $x \neq y$ and $0$ otherwise. Here the ball $B\!\left(x, \tfrac{1}{2}\right) = \{x\}$, so every singleton is open. Every subset of $X$ is a union of singletons, so *every* subset is open, and since complements are subsets too, every subset is closed as well. In a discrete space every set is clopen, and every boundary is empty. The space has been shattered into isolated points, each one its own separated piece.

### Looking ahead

These examples trace out a spectrum. At one end, $\mathbb{R}$ has only the two trivial clopen sets, it is all "one piece". At the other end, a discrete space has nothing but clopen sets, it is "all pieces". The pattern is that nontrivial clopen sets exist exactly when the space can be broken into separated chunks. Pinning that down is the job of the final section: a space is connected if and only if its only clopen subsets are $\emptyset$ and $X$.

## Connected Sets

We can generalize the idea of an interval to general metric spaces. One of the key features of an interval in $\mathbb{R}$ is that it's connected: we can move between any two of its points without ever leaving the interval or jumping across a gap. In $\mathbb{R}$ we study functions on intervals, and in more general metric spaces we study functions on connected sets, so it pays to make "all one piece" precise.

Formally, a nonempty metric space $(X,d)$ is **connected** if the only subsets of $X$ that are both open and closed (clopen) are

- $\emptyset$, and
- $X$ itself.

If a nonempty space $X$ is not connected, it is **disconnected**.

Intuitively, a nonempty $X$ is connected if whenever we write $X = X_1 \cup X_2$ with $X_1 \cap X_2 = \emptyset$ and $X_1, X_2$ open, then either $X_1 = \emptyset$ or $X_2 = \emptyset$. So to show that $X$ is disconnected, we need to find two nonempty disjoint open sets $X_1, X_2$ whose union is $X$. Such a pair is called a **separation** of $X$.

We often want to ask whether a subset $S \subseteq X$ is connected, even when $S$ is not the whole space. The definition is the same idea, carried out with the open sets of $X$ cut down to $S$.

Formally, let $(X,d)$ be a metric space. A nonempty set $S \subseteq X$ is **disconnected** if and only if there exist open sets $U_1, U_2 \subseteq X$ such that $U_1 \cap U_2 \cap S = \emptyset$, $U_1 \cap S \neq \emptyset$, $U_2 \cap S \neq \emptyset$, and
$$ S = (U_1 \cap S) \cup (U_2 \cap S). $$
The pieces $U_1 \cap S$ and $U_2 \cap S$ are exactly the open subsets of $S$ in its own right (its subspace topology), so this is just the separation idea applied to $S$ instead of all of $X$.

### Why the Two Definitions Agree

We have now described connectedness two ways — "no nontrivial clopen set" and "no separation" — so we should check that they say the same thing. Happily, we did most of the work back in the **Clopen Sets** section.

- Suppose $X$ has a clopen set $A$ with $\emptyset \neq A \neq X$. Since $A$ is closed, its complement $X \setminus A$ is open, and since $A$ is open too, the pair $U_1 = A$ and $U_2 = X \setminus A$ is made of two open sets that are nonempty, disjoint, and cover $X$. That is a separation. A nontrivial clopen set is a separation.
- Conversely, suppose $X = U_1 \cup U_2$ is a separation. Then $U_1 = X \setminus U_2$ is the complement of an open set, hence closed, and it is also open. So $U_1$ is clopen, nonempty, and not all of $X$ (because $U_2$ is nonempty). That is a nontrivial clopen set.

So $X$ has a nontrivial clopen subset if and only if $X$ has a separation. The two definitions are genuinely the same, which is why connectedness is the natural sibling of the clopen story: a connected space is one that refuses to come apart, and "coming apart" is exactly the same as owning a clopen piece.

### Examples

#### Example 1: $\mathbb{R}$ and its intervals are connected

The connected subsets of $\mathbb{R}$ are exactly the **intervals** (including rays, single points, and $\mathbb{R}$ itself). One direction is easy and worth seeing. Suppose $S \subseteq \mathbb{R}$ is *not* an interval. Then there are points $a < c < b$ with $a, b \in S$ but $c \notin S$. Take
$$ U_1 = (-\infty, c), \qquad U_2 = (c, \infty), $$
both open in $\mathbb{R}$. They are disjoint, the intersections $U_1 \cap S$ and $U_2 \cap S$ are nonempty (they contain $a$ and $b$ respectively), and since $c \notin S$ every point of $S$ lies strictly on one side or the other, so $S = (U_1 \cap S) \cup (U_2 \cap S)$. That is a separation, so $S$ is disconnected. A set with a hole in it always splits at the hole.

The converse — that every interval really is connected — takes a little more, and rests on the completeness of $\mathbb{R}$ (the least upper bound property). The idea is to take the supremum of how far $U_1$ reaches toward $b$ and check that this point cannot consistently be placed in either open set. A clean writeup is in Lebl's **Basic Analysis 1**.

#### Example 2: $X = [0,1] \cup [2,3]$ is disconnected

Let us run our recurring example through the formal definition. Work inside $\mathbb{R}$ and take
$$ U_1 = \left(-1, \tfrac{3}{2}\right), \qquad U_2 = \left(\tfrac{3}{2}, 4\right), $$
both open in $\mathbb{R}$. Then

- $U_1 \cap X = [0,1]$ and $U_2 \cap X = [2,3]$, so both pieces are nonempty;
- $U_1 \cap U_2 = \emptyset$, so in particular $U_1 \cap U_2 \cap X = \emptyset$;
- $(U_1 \cap X) \cup (U_2 \cap X) = [0,1] \cup [2,3] = X$.

All four conditions hold, so $X$ is disconnected. This is the same fact we met in the **Clopen Sets** section, where $[0,1]$ was a nontrivial clopen subset of $X$. The separation $U_1, U_2$ and the clopen set $[0,1]$ are just two views of the single gap between the pieces.

### What Connectedness Buys Us

Connectedness matters because it survives continuous functions, and that one fact powers a surprising amount of analysis.

**Theorem.** If $f : X \to Y$ is continuous and $X$ is connected, then the image $f(X)$ is connected.

In $\mathbb{R}$ this is the **Intermediate Value Theorem** in disguise. If $f : [a,b] \to \mathbb{R}$ is continuous, then $[a,b]$ is connected, so $f([a,b])$ is a connected subset of $\mathbb{R}$ — and by Example 1 that means an interval. An interval containing $f(a)$ and $f(b)$ must contain every value between them, so $f$ attains every intermediate value. The "no jumping" picture we started with turns out to be a theorem about connected sets.

### A Note on Path-Connectedness

The intuition of "moving continuously between points" is really a slightly stronger and more vivid notion called **path-connectedness**: $X$ is path-connected if any two of its points can be joined by a continuous path lying inside $X$. Every path-connected space is connected, and for intervals and ordinary regions in $\mathbb{R}^n$ the two notions coincide, so the picture in your head is reliable. They are not identical in general — the classic counterexample is the *topologist's sine curve*, which is connected but not path-connected — but that is a story for another post.

### Closing the Loop

This finally answers the question in the title. In $\mathbb{R}$, and in any connected space, the only clopen sets are $\emptyset$ and $X$, because a nontrivial clopen set would tear the space into two separated pieces. Clopen sets feel exotic on the real line precisely because the line is connected: there is no seam to cut along. As soon as we move to a space that *does* come apart, like $[0,1] \cup [2,3]$ or a discrete space, clopen sets show up everywhere. Open, closed, and clopen were never exotic at all — they are simply the language for describing where a set's edges lie, and whether the space has any seams to begin with.