Open and Closed Sets in Metric Spaces

Open and closed sets are the foundational concepts of metric space topology. A set is open if every point has a small ball entirely contained in the set; a set is closed if it contains all its limit points (equivalently, its complement is open). These notions formalize "nearness" without coordinates and underpin continuity, convergence, and compactness. Originating with Cantor, Fréchet, and Hausdorff (1880s–1914), they apply across analysis, geometry, and topology.

In a metric space $(X, d)$ , distances let us define a ball around each point, and a set is open when it is a neighborhood of each of its points — there is always wiggle room. A set is closed precisely when its complement is open, which is equivalent to containing all its limit points. These two classes of sets are not opposites: a set can be both open and closed (clopen), or neither (like a half-open interval). The surprising fact is that the entire edifice of continuity, compactness, and connectedness can be rebuilt using only open sets — this realization birthed point-set topology, where metrics disappear entirely and open sets become the primitive notion.

The Core Insight

Openness is local: a set is open iff around every one of its points you can fit a tiny ball staying inside. Closedness is limit-stability: a set is closed iff no sequence in it can sneak out to a limit outside. The duality $closed = complement of open$ ties these together and lets every topological statement be phrased two ways.

Formal Statement

Metric Space

A metric space is a pair $(X, d)$ where $X$ is a set and $d : X \times X \to [0, \infty)$ is a metric satisfying, for all $x, y, z \in X$ :

d (x, y) = 0 ⟺ x = y

⟳This is the identity-of-indiscernibles axiom; nonnegativity is built into the codomain

[0, \infty)

d (x, y) = d (y, x)

⟳Symmetry — distance does not depend on direction of measurement.

d (x, z) \leq d (x, y) + d (y, z)

⟳The triangle inequality; a detour through

y

is never shorter than the direct route.

Here $X$ is the underlying set, $d$ the distance function, and the points $x, y, z$ range over $X$ .

Open Ball

For $x_{0} \in X$ and radius $r > 0$ , the open ball of radius $r$ centered at $x_{0}$ is

B (x_{0}, r) = {x \in X : d (x, x_{0}) < r}

⟳Collect all points strictly closer than

r

to the center; strictness is what makes the ball "open."

The closed ball is $\overset{ˉ}{B} (x_{0}, r) = {x : d (x, x_{0}) \leq r}$ (note: $\overset{ˉ}{B}$ need not equal the closure of $B$ ).

Open Set

A set $U \subseteq X$ is open if every point of $U$ is an interior point:

U open ⟺ \forall x \in U, \exists r > 0 such that B (x, r) \subseteq U

⟳"Open" means room to wiggle — each point carries a ball of slack inside

U

. The radius

r

may depend on

x

Limit Point and Closed Set

A point $p \in X$ is a limit point (accumulation point) of $A \subseteq X$ if every open ball around $p$ contains a point of $A$ other than $p$ :

p limit point of A ⟺ \forall r > 0, (B (p, r) ∖ {p}) \cap A \neq = \emptyset

⟳No matter how small the ball, you still capture some point of

A

distinct from

p

— so

A

accumulates at

p

A set $C \subseteq X$ is closed if it contains all its limit points; equivalently:

C closed ⟺ X ∖ C open

⟳Complement duality — a point outside

C

that is not a limit point has a ball missing

C

entirely, witnessing openness of the complement.

Axioms of Open Sets

The collection $τ$ of open sets in a metric space satisfies:

\emptyset \in τ, X \in τ

⟳The empty condition is vacuously true; every ball around any point lies in

X

U_{α} \in τ (α \in I) ⟹ α \in I ⋃ U_{α} \in τ

⟳Arbitrary unions — a point in the union lies in some

U_{α}

, whose witnessing ball still sits inside the union.

U_{1}, \dots, U_{n} \in τ ⟹ k = 1 ⋂ n U_{k} \in τ

⟳Finite intersections — take the minimum of the finitely many witnessing radii. "Finite" is essential.

Dual Axioms for Closed Sets

By De Morgan applied to [8]–[10]: $\emptyset$ and $X$ are closed; arbitrary intersections of closed sets are closed; finite unions of closed sets are closed.

(α ⋃ U_{α})^{c} = α ⋂ U_{α}^{c}, (k = 1 ⋂ n U_{k})^{c} = k = 1 ⋃ n U_{k}^{c}

⟳De Morgan's laws convert each open-set axiom into its closed-set dual by complementation.

Why It Holds / Derivation

Every open ball is an open set

We show $B (x_{0}, r)$ is open per definition [5].

Let $y \in B (x_{0}, r)$ , so $d (y, x_{0}) < r$ by [4].
Define the slack radius $s = r - d (y, x_{0}) > 0$ .
Claim $B (y, s) \subseteq B (x_{0}, r)$ . Take any $z \in B (y, s)$ , so $d (z, y) < s$ .
By the triangle inequality [3]: $d (z, x_{0}) \leq d (z, y) + d (y, x_{0}) < s + d (y, x_{0}) = r$ .
Hence $z \in B (x_{0}, r)$ , proving the inclusion.
Since every $y \in B (x_{0}, r)$ admits such a ball $B (y, s)$ , the set is open by [5]. $■$

Finite intersection of open sets is open

Let $U_{1}, \dots, U_{n}$ be open and $x \in ⋂_{k} U_{k}$ .

For each $k$ , since $U_{k}$ is open and $x \in U_{k}$ , there is $r_{k} > 0$ with $B (x, r_{k}) \subseteq U_{k}$ .
Set $r = min {r_{1}, \dots, r_{n}} > 0$ (finiteness guarantees the minimum is positive).
Then $B (x, r) \subseteq B (x, r_{k}) \subseteq U_{k}$ for every $k$ .
Therefore $B (x, r) \subseteq ⋂_{k} U_{k}$ , so the intersection is open. $■$

The failure for infinite intersections: $⋂_{n = 1}^{\infty} (- \frac{1}{n}, \frac{1}{n}) = {0}$ , which is not open in $R$ .

Closed sets are exactly limit-point-containing sets

We verify the equivalence of [7] (complement open) with "contains all limit points."

( $\Rightarrow$ ) Suppose $X ∖ C$ is open. Let $p$ be a limit point of $C$ ; assume for contradiction $p \in / C$ , i.e. $p \in X ∖ C$ .
Openness gives $r > 0$ with $B (p, r) \subseteq X ∖ C$ , so $B (p, r) \cap C = \emptyset$ , contradicting [6]. Hence $p \in C$ .
( $\Leftarrow$ ) Suppose $C$ contains all its limit points. Take $p \in X ∖ C$ . Since $p \in / C$ , $p$ is not a limit point of $C$ (else $p \in C$ ).
By negating [6], there exists $r > 0$ with $B (p, r) \cap C \subseteq {p}$ ; since $p \in / C$ , in fact $B (p, r) \cap C = \emptyset$ , so $B (p, r) \subseteq X ∖ C$ .
Thus $X ∖ C$ is open, i.e. $C$ is closed. $■$

Felix Hausdorff | 1868–1942

Hausdorff's Grundzüge der Mengenlehre (1914) consolidated topology around neighborhood axioms, building on Maurice Fréchet's 1906 thesis introducing metric spaces (he called them classes (E)). Hausdorff isolated the separation property now bearing his name (distinct points have disjoint neighborhoods) and gave open/closed sets their modern axiomatic footing. A profound analyst and topologist, he died by suicide in 1942 to escape deportation by the Nazi regime. His work transformed "nearness" from a metaphor into a rigorous lattice of open sets.

Building Intuition

Think of an open set as a country with no fence at the border: from any town inside, you can walk a little in any direction and still be inside. There is no "edge town" — every resident has neighbors all around them in the country. A closed set is the opposite: it owns its boundary. If a sequence of towns marches toward a destination, that destination is also part of the country.

The clopen/neither distinction confuses beginners. The interval $[0, 1)$ in $R$ is neither: at $0$ you cannot wiggle left (so not open), and the limit point $1$ is missing (so not closed). Openness and closedness are not negations of each other — that's a property of complements, not of the same set.

Open ball B(y,s) nested inside B(x₀,r): the geometric heart of openness

Worked Examples

The half-open interval is neither open nor closed

Given:

A = [0, 1) \subset R

with

d (x, y) = ∣ x - y ∣

. Not open: Test the point

0 \in A

. Any ball

B (0, r) = (- r, r)

contains

- \frac{r}{2} < 0

, which is not in

A

. So no ball around

0

fits inside

A

; openness fails at

0

. Not closed: The point

1

is a limit point of

A

: for any

r > 0

, the point

1 - \frac{r}{2}

lies in

B (1, r) \cap A

and differs from

1

. But

1 \in / A

, so

A

omits a limit point. Boxed answer:

[0, 1) is neither open nor closed in R

Clopen sets in the integers with the standard metric

Given:

X = Z

as a subspace of

R

, metric

d (m, n) = ∣ m - n ∣

. Consider the singleton

{5}

. Step 1 (open): In the subspace,

B (5, \frac{1}{2}) = {n \in Z : ∣ n - 5∣ < \frac{1}{2}} = {5}

. So a ball of radius

\frac{1}{2}

around

5

equals

{5}

, fitting inside

{5}

. Hence

{5}

is open in

Z

. Step 2 (closed): Its complement

Z ∖ {5}

is a union of singletons

⋃_{n \neq = 5} {n}

, each open by the same argument; by [9] the union is open. Therefore

{5}

is closed. Step 3 (count of clopen sets): Since every singleton is clopen and unions of opens are open, every subset of

Z

is clopen. Boxed answer:

{5} is clopen in Z; indeed every subset of Z is clopen

Common Misconceptions

"Closed = not open." False. These are independent properties. In

R

(0, 1)

is open-not-closed,

[0, 1]

is closed-not-open,

[0, 1)

is neither, and

\emptyset

and

R

are both. The correct duality is closed

=

complement of open, not not open. "The closed ball is the closure of the open ball." Not always. In a discrete metric space (

d (x, y) = 1

for

x \neq = y

B (x, 1) = {x}

which is already closed, so its closure is

{x}

. But

\overset{ˉ}{B} (x, 1) = X

(every point is within distance

1

). So

\overline{B (x, 1)} = {x} \neq = X = \overset{ˉ}{B} (x, 1)

Openness depends on the ambient space

The set $[0, 1)$ is not open in $R$ . But consider $X = [0, \infty)$ as a metric space in its own right. Now the point $0$ has $B (0, \frac{1}{2}) = [0, \frac{1}{2}) \subseteq [0, 1)$ inside $X$ , because negative numbers no longer exist. So $[0, 1)$ is open in $[0, \infty)$ . Moral: "open" is not absolute — it is always open in $X$ , governed by the subspace topology.

Connections

Open and closed sets are the gateway to Continuity in Metric Spaces: a map is continuous iff preimages of open sets are open (the topological definition), generalizing the $ε$ – $δ$ criterion. They underlie Compactness, where the open-cover definition (every open cover has a finite subcover) supersedes sequential compactness in general spaces. The closure, interior, and boundary operators built from these sets feed into Closure and Interior Operators. Abstracting away the metric and keeping only axioms [8]–[10] yields a Topological Space, the central object of Point-Set Topology. Completeness and Cauchy sequences interact with closedness via Complete Metric Spaces (closed subsets of complete spaces are complete).

Real-World Applications

Numerical analysis & optimization: Open feasible regions versus closed constraint sets determine whether minima are attained — the Extreme Value Theorem requires closed and bounded domains in $R^{n}$ .
Control theory: Stability regions in state space are described as open/closed sets; a system is stable if trajectories remain in a closed invariant set.
Computer graphics & GIS: Region-filling and point-in-polygon tests implement the interior/closure distinction; boundary pixels are handled via closedness.
Machine learning (topological data analysis): Persistent homology builds open-ball covers (Vietoris–Rips/Čech complexes) to detect "holes" in data clouds.
Signal processing: Open neighborhoods in frequency space define passbands; continuity of filters is the open-preimage condition.
Robotics motion planning: Configuration-space obstacles are closed sets; free space is their open complement, where collision-free paths live.
Economics: Budget sets are typically closed and bounded (compact), guaranteeing existence of utility-maximizing bundles via continuity on compact sets.

Historical Development

1872 — Georg Cantor introduces limit points and derived sets while studying convergence of Fourier series, planting the seed of accumulation points.
1883 — Cantor defines closed sets via derived sets ( $A$ closed iff $A^{'} \subseteq A$ ) in his work on point sets in $R^{n}$ .
1893 — Camille Jordan formalizes notions of boundary and interior for sets in the plane (Jordan curve theorem context).
1906 — Maurice Fréchet's doctoral thesis introduces abstract metric spaces (classes (E)), defining convergence and limit points without coordinates.
1910 — F. Riesz and others develop neighborhood-based topology, anticipating axiomatic open sets.
1914 — Felix Hausdorff's Grundzüge der Mengenlehre gives the neighborhood-axiom foundation and the Hausdorff separation axiom; metric topology becomes systematic.
1922 — Kazimierz Kuratowski axiomatizes topology via the closure operator, an alternative to open-set axioms.
1925–1930 — Pavel Urysohn and Pavel Alexandrov develop metrization theory, asking when a topology comes from a metric.
1937 — Bourbaki begins codifying general topology, cementing open sets as the primitive notion in modern texts.

Subtleties & Limits

Metrization asks: which topological spaces arise from some metric? The Urysohn, Nagata–Smirnov, and Bing metrization theorems answer this for large classes (a regular space is metrizable iff it has a

σ

-locally-finite base), but characterizing metrizability remains subtle for exotic spaces. Beyond this, the interaction of closedness with completeness is delicate: a set can be closed yet not complete in a non-complete ambient space, and the choice of equivalent metric can change whether balls are convex even while keeping the same open sets. In infinite dimensions, closed bounded sets need not be compact (failure of Heine–Borel), driving the entire field of functional analysis. The general question of when topological structure determines analytic structure underlies ongoing research in descriptive set theory and Borel complexity hierarchies.