Open and Closed Sets in Metric Spaces

⚡ Key Takeaways

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.

💡 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 $\text{closed} = \text{complement of open}$ ties these together and lets every topological statement be phrased two ways.

DefinitionMetric 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 $x,y,z\in X$ :

1

\quad d(x,y)=0 \iff x=y

Derivation hint: This is the identity-of-indiscernibles axiom; nonnegativity is built into the codomain $[0,\infty)$ .

2

\quad d(x,y)=d(y,x)

Derivation hint: Symmetry — distance does not depend on direction of measurement.

3

\quad d(x,z)\le d(x,y)+d(y,z)

Derivation hint: 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 $X$ .

DefinitionOpen Ball

For $x_0\in X$ and radius $r>0$ , the open ball of radius $r$ centered at $x_0$ is

4

\quad B(x_0,r)=\{x\in X : d(x,x_0)<r\}

Derivation hint: Collect all points strictly closer than $r$ to the center; strictness is what makes the ball "open." The closed ball is $\bar B(x_0,r)=\{x : d(x,x_0)\le r\}$ (note: $\bar B$ need not equal the closure of $B$ ).

DefinitionOpen Set

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

5

\quad U \text{ open} \iff \forall x\in U,\ \exists\, r>0 \text{ such that } B(x,r)\subseteq U

Derivation hint: "Open" means room to wiggle — each point carries a ball of slack inside $U$ . The radius $r$ may depend on $x$ .

DefinitionLimit 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 $p$ :

6

\quad p \text{ limit point of } A \iff \forall r>0,\ \big(B(p,r)\setminus\{p\}\big)\cap A \neq \varnothing

Derivation hint: 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:

7

\quad C \text{ closed} \iff X\setminus C \text{ open}

Derivation hint: Complement duality — a point outside $C$ that is not a limit point has a ball missing $C$ entirely, witnessing openness of the complement.

TheoremAxioms of Open Sets

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

8

\quad \varnothing\in\tau,\quad X\in\tau

Derivation hint: The empty condition is vacuously true; every ball around any point lies in $X$ .

9

\quad U_\alpha\in\tau\ (\alpha\in I)\ \Longrightarrow\ \bigcup_{\alpha\in I}U_\alpha\in\tau

Derivation hint: Arbitrary unions — a point in the union lies in some $U_\alpha$ , whose witnessing ball still sits inside the union.

10

\quad U_1,\dots,U_n\in\tau\ \Longrightarrow\ \bigcap_{k=1}^{n}U_k\in\tau

Derivation hint: Finite intersections — take the minimum of the finitely many witnessing radii. "Finite" is essential.

TheoremDual Axioms for Closed Sets

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

11

\quad \Big(\bigcup_\alpha U_\alpha\Big)^c=\bigcap_\alpha U_\alpha^c,\qquad \Big(\bigcap_{k=1}^n U_k\Big)^c=\bigcup_{k=1}^n U_k^c

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

ProofEvery open ball is an open set

We show $B(x_0,r)$ is open per definition [5]. 1. Let $y\in B(x_0,r)$ , so $d(y,x_0)<r$ by [4]. 2. Define the slack radius $s=r-d(y,x_0)>0$ . 3. Claim $B(y,s)\subseteq B(x_0,r)$ . Take any $z\in B(y,s)$ , so $d(z,y)<s$ . 4. By the triangle inequality [3]: $d(z,x_0)\le d(z,y)+d(y,x_0)<s+d(y,x_0)=r$ . 5. Hence $z\in B(x_0,r)$ , proving the inclusion. 6. Since every $y\in B(x_0,r)$ admits such a ball $B(y,s)$ , the set is open by [5]. $\blacksquare$

∎

ProofFinite intersection of open sets is open

Let $U_1,\dots,U_n$ be open and $x\in\bigcap_k U_k$ . 1. For each $k$ , since $U_k$ is open and $x\in U_k$ , there is $r_k>0$ with $B(x,r_k)\subseteq U_k$ . 2. Set $r=\min\{r_1,\dots,r_n\}>0$ (finiteness guarantees the minimum is positive). 3. Then $B(x,r)\subseteq B(x,r_k)\subseteq U_k$ for every $k$ . 4. Therefore $B(x,r)\subseteq\bigcap_k U_k$ , so the intersection is open. $\blacksquare$ The failure for infinite intersections: $\bigcap_{n=1}^\infty(-\tfrac1n,\tfrac1n)=\{0\}$ , which is not open in $\mathbb R$ .

∎

ProofClosed sets are exactly limit-point-containing sets

We verify the equivalence of [7] (complement open) with "contains all limit points." 1. ( $\Rightarrow$ ) Suppose $X\setminus C$ is open. Let $p$ be a limit point of $C$ ; assume for contradiction $p\notin C$ , i.e. $p\in X\setminus C$ . 2. Openness gives $r>0$ with $B(p,r)\subseteq X\setminus C$ , so $B(p,r)\cap C=\varnothing$ , contradicting [6]. Hence $p\in C$ . 3. ( $\Leftarrow$ ) Suppose $C$ contains all its limit points. Take $p\in X\setminus C$ . Since $p\notin C$ , $p$ is not a limit point of $C$ (else $p\in C$ ). 4. By negating [6], there exists $r>0$ with $B(p,r)\cap C\subseteq\{p\}$ ; since $p\notin C$ , in fact $B(p,r)\cap C=\varnothing$ , so $B(p,r)\subseteq X\setminus C$ . 5. Thus $X\setminus C$ is open, i.e. $C$ is closed. $\blacksquare$

∎

mathematician · 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.

🧠 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 $\mathbb R$ 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.

Figure

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

ExampleThe half-open interval is neither open nor closed

Given: $A=[0,1)\subset\mathbb R$ with $d(x,y)=|x-y|$ .

Not open: Test the point $0\in A$ . Any ball $B(0,r)=(-r,r)$ contains $-\tfrac r2<0$ , 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 $1-\tfrac r2$ lies in $B(1,r)\cap A$ and differs from $1$ . But $1\notin A$ , so $A$ omits a limit point.

Boxed answer: $\boxed{[0,1)\text{ is neither open nor closed in }\mathbb R}$

ExampleClopen sets in the integers with the standard metric

Given: $X=\mathbb Z$ as a subspace of $\mathbb R$ , metric $d(m,n)=|m-n|$ . Consider the singleton $\{5\}$ .

Step 1 (open): In the subspace, $B(5,\tfrac12)=\{n\in\mathbb Z : |n-5|<\tfrac12\}=\{5\}$ . So a ball of radius $\tfrac12$ around $5$ equals $\{5\}$ , fitting inside $\{5\}$ . Hence $\{5\}$ is open in $\mathbb Z$ .

Step 2 (closed): Its complement $\mathbb Z\setminus\{5\}$ is a union of singletons $\bigcup_{n\neq5}\{n\}$ , 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 $\mathbb Z$ is clopen.

Boxed answer: $\boxed{\{5\}\text{ is clopen in }\mathbb Z;\ \text{indeed every subset of }\mathbb Z\text{ is clopen}}$

⚠️ Common mistake

"Closed = not open." False. These are independent properties. In $\mathbb R$ : $(0,1)$ is open-not-closed, $[0,1]$ is closed-not-open, $[0,1)$ is neither, and $\varnothing$ and $\mathbb R$ are both. The correct duality is closed $=$ complement of open, not not open.

⚠️ Common mistake

"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 $\bar B(x,1)=X$ (every point is within distance $1$ ). So $\overline{B(x,1)}=\{x\}\neq X=\bar B(x,1)$ .

CounterexampleOpenness depends on the ambient space

The set $[0,1)$ is not open in $\mathbb R$ . But consider $X=[0,\infty)$ as a metric space in its own right. Now the point $0$ has $B(0,\tfrac12)=[0,\tfrac12)\subseteq[0,1)$ 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.

connection

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 $\varepsilon$ – $\delta$ 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).

🔬 Where this appears

⏳ History

- 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

\mathbb 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.

open-problem

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 $\sigma$ -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.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also