In any [metric space](/page/Metric%20Space), a convergent sequence has exactly one limit: if $x_k \to x$ and $x_k \to y$ with $x \neq y$, then for $r = d(x,y)/2 > 0$, the open balls $B(x,r)$ and $B(y,r)$ are disjoint, and the sequence cannot eventually belong to both. This argument uses nothing about the specific metric — only that distinct points can be separated by disjoint open sets. In a general [topological space](/page/Topology), this separation property can fail, and when it does, the consequences are severe: limits become non-unique, compact subsets need not be closed, and continuous bijections from compact spaces need not be homeomorphisms. The Hausdorff condition is the axiom that prevents these pathologies, and it is the minimal assumption under which the standard constructions of analysis remain valid. [example: The Line with Two Origins] The following space demonstrates what goes wrong without the Hausdorff axiom. Let $X = (\mathbb{R} \setminus \{0\}) \cup \{p, q\}$, where $p$ and $q$ are two distinct points meant to serve as "competing origins." Define a topology on $X$ as follows: a subset $U \subset X$ is open if - the intersection $U \cap (\mathbb{R} \setminus \{0\})$ is open in the standard topology on $\mathbb{R} \setminus \{0\}$, and - if $p \in U$, then $(-\varepsilon, 0) \cup (0, \varepsilon) \cup \{p\} \subset U$ for some $\varepsilon > 0$, and - if $q \in U$, then $(-\delta, 0) \cup (0, \delta) \cup \{q\} \subset U$ for some $\delta > 0$. In other words, $p$ and $q$ each behave like the origin — they have the same collection of "punctured" neighbourhoods — but they are distinct points. Consider the sequence $x_k = 1/k$ for $k \in \mathbb{N}$. Every open set containing $p$ includes an interval $(-\varepsilon, 0) \cup (0, \varepsilon) \cup \{p\}$, so $x_k = 1/k \in U$ for all $k > 1/\varepsilon$. Thus $x_k \to p$. By the identical argument with $q$ in place of $p$, we also have $x_k \to q$. The sequence converges to two distinct limits simultaneously. The underlying problem is that $p$ and $q$ cannot be separated by disjoint open sets: any open set containing $p$ intersects every open set containing $q$ (both must contain points of $(0, \varepsilon)$ for small $\varepsilon$). The Hausdorff axiom, stated below, prohibits exactly this failure. [/example] ## Definition The pathology above — the inability to separate distinct points by disjoint open sets — motivates the following axiom. [definition: Hausdorff Space] A [topological space](/page/Topology) $(X, \tau)$ is a **Hausdorff space** (or **$T_2$ space**) if for every pair of distinct points $x, y \in X$, there exist open sets $U, V \in \tau$ such that \begin{align*} x \in U, \quad y \in V, \quad U \cap V = \varnothing. \end{align*} [/definition] When we say that $x$ and $y$ can be "separated by open sets," we mean precisely that such disjoint open neighbourhoods exist. A topological space satisfying this condition is also called **separated**. Every [metric space](/page/Metric%20Space) is Hausdorff: given distinct points $x \neq y$, set $r = d(x,y)/2 > 0$ and take $U = B(x,r)$, $V = B(y,r)$. The triangle inequality guarantees $U \cap V = \varnothing$, since any $z \in U \cap V$ would satisfy $d(x,y) \le d(x,z) + d(z,y) < r + r = d(x,y)$, a contradiction. In particular, $\mathbb{R}^n$ with the Euclidean topology, every [Banach space](/page/Banach%20Space) with the norm topology, and every [manifold](/page/Smooth%20Manifold) with its standard topology are all Hausdorff. The Hausdorff condition sits in the middle of a hierarchy of separation axioms. To understand its role, we need to see what it gives beyond weaker axioms and what stronger axioms provide beyond it. ## The Hierarchy of Separation Axioms A basic challenge in point-set [topology](/page/Topology) is deciding how much separation between points and closed sets a topological space should guarantee. The axioms of a topology are deliberately minimal — they ensure that unions and finite intersections of open sets are open, but say nothing about how effectively the topology distinguishes points from each other or from closed sets. Different levels of separation are captured by the $T_i$ axioms, and the choice of which axiom to impose determines how much of classical analysis can be carried out. ### The $T_0$ and $T_1$ Axioms The weakest useful separation axiom asks only that the topology can tell points apart *somehow*. [definition: $T_0$ Space] A topological space $(X, \tau)$ is **$T_0$** (or **Kolmogorov**) if for every pair of distinct points $x, y \in X$, there exists an open set $U \in \tau$ that contains one of the two points but not the other. [/definition] The $T_0$ axiom says the topology is fine enough to distinguish points, but it does not say which point gets the separating open set. This asymmetry is a real limitation: in a $T_0$ space, it is possible for $x$ to have no open neighbourhood avoiding $y$, even though $y$ has an open neighbourhood avoiding $x$. The $T_1$ axiom eliminates this asymmetry. [definition: $T_1$ Space] A topological space $(X, \tau)$ is **$T_1$** (or **Fr\'{e}chet**) if for every pair of distinct points $x, y \in X$, there exists an open set $U \in \tau$ with $x \in U$ and $y \notin U$. [/definition] The $T_1$ condition is equivalent to requiring that every singleton $\{x\}$ is a [closed set](/page/Closed%20Set). To see this, note that if $(X, \tau)$ is $T_1$, then for every $y \neq x$, there is an open set $U_y$ containing $y$ but not $x$. The union $\bigcup_{y \neq x} U_y = X \setminus \{x\}$ is open, so $\{x\}$ is closed. Conversely, if every singleton is closed, then for $x \neq y$, the set $U = X \setminus \{y\}$ is open, contains $x$, and excludes $y$. However, the $T_1$ axiom is still insufficient for analysis. In a $T_1$ space, limits of sequences can fail to be unique, because the axiom provides an open set containing $x$ but not $y$, and a separate open set containing $y$ but not $x$ — but these sets need not be disjoint. Without disjointness, a sequence can belong eventually to both. [example: The Cofinite Topology on an Infinite Set] Let $X$ be any infinite set and define the **cofinite topology** $\tau = \{U \subset X : X \setminus U \text{ is finite}\} \cup \{\varnothing\}$. This topology is $T_1$: for any distinct $x, y \in X$, the set $U = X \setminus \{y\}$ is open (its complement is the finite set $\{y\}$), contains $x$, and excludes $y$. However, $(X, \tau)$ is not Hausdorff. Suppose $U$ and $V$ are nonempty open sets. Then $X \setminus U$ and $X \setminus V$ are both finite, so $X \setminus (U \cap V) = (X \setminus U) \cup (X \setminus V)$ is a finite union of finite sets, hence finite. Since $X$ is infinite, $U \cap V \neq \varnothing$. Therefore no two nonempty open sets are disjoint, and no pair of distinct points can be separated. As a consequence, every sequence in $X$ whose range is infinite converges to every point of $X$. Indeed, if $\{x_k\}_{k=1}^\infty$ takes infinitely many distinct values and $p \in X$ is any point, then for any open set $U$ containing $p$, the complement $X \setminus U$ is finite, so at most finitely many terms $x_k$ lie outside $U$. Thus $x_k$ is eventually in $U$, and $x_k \to p$. Limits are completely meaningless in this topology. [/example] The Hausdorff axiom ($T_2$) resolves this by requiring that the separating open sets be disjoint, which — as demonstrated in the opening — is exactly the condition needed to guarantee uniqueness of limits. ### Regularity and Normality Beyond the Hausdorff condition, two stronger separation axioms play central roles in topology and analysis. [definition: Regular Space] A topological space $(X, \tau)$ is **regular** (or **$T_3$**) if it is $T_1$ and for every closed set $C \subset X$ and every point $x \notin C$, there exist disjoint open sets $U, V \in \tau$ with $x \in U$ and $C \subset V$. [/definition] Regularity upgrades the Hausdorff condition from separating points from points to separating points from closed sets. Every metric space is regular: given a closed set $C$ and a point $x \notin C$, set $r = \operatorname{dist}(x, C) > 0$ (which is positive since $C$ is closed) and take $U = B(x, r/2)$ and $V = \bigcup_{c \in C} B(c, r/2)$. [definition: Normal Space] A topological space $(X, \tau)$ is **normal** (or **$T_4$**) if it is $T_1$ and for every pair of disjoint closed sets $C_0, C_1 \subset X$, there exist disjoint open sets $U_0, U_1 \in \tau$ with $C_0 \subset U_0$ and $C_1 \subset U_1$. [/definition] The significance of normality lies in [Urysohn's Lemma](/theorems/887): a space is normal if and only if disjoint closed sets can be separated by a continuous function $f: X \to [0,1]$ with $f|_{C_0} \equiv 0$ and $f|_{C_1} \equiv 1$. This functional separation is essential for [partition of unity](/page/Partition%20of%20Unity) constructions and for the [Urysohn Metrization Theorem](/theorems/992), which characterises metrizability in terms of separation axioms. The implications among these axioms are strict: \begin{align*} \text{normal} \implies \text{regular} \implies \text{Hausdorff} \implies T_1 \implies T_0. \end{align*} None of these implications reverses. The [Sorgenfrey line](/page/Metrizable%20Space) — $\mathbb{R}$ equipped with the topology generated by half-open intervals $[a, b)$ — is normal but not metrizable. The line with two origins from the opening example is $T_1$ but not Hausdorff. The key point for analysis is that **Hausdorff is the weakest axiom guaranteeing uniqueness of limits**, and it is therefore the standard assumption in functional analysis, differential geometry, and measure theory. ## Uniqueness of Limits and Closedness of Points The two most immediate and frequently used consequences of the Hausdorff property are the uniqueness of limits and the fact that compact subsets are well-behaved with respect to closedness. Both rely on the same underlying mechanism: the ability to "separate and shrink." ### Uniqueness of Limits In analysis, the construction of objects — solutions to differential equations, fixed points, approximations — typically proceeds by producing a convergent sequence or net and identifying its limit. If limits were not unique, this strategy would be meaningless: we could not speak of "the limit" of a sequence, and continuity could not be defined in terms of preservation of limits. [quotetheorem:291] The argument is the one given in the opening paragraph: if $x \neq y$, choose disjoint open sets $U \ni x$ and $V \ni y$. By convergence, the net is eventually in $U$ and eventually in $V$. But "eventually" for a net means "for all sufficiently large indices," and the intersection of two cofinal sets in a directed set is cofinal. So the net is eventually in $U \cap V = \varnothing$, which is impossible. This theorem would be false in any non-Hausdorff space — the line with two origins and the cofinite topology examples above both demonstrate this. The statement is also noteworthy for its generality: it applies to nets, not just sequences. In spaces that are not [first-countable](/page/Metrizable%20Space), sequences alone do not determine the topology, and nets (or equivalently, [filters](/page/Nets%20and%20Filters)) are the correct notion of convergence. The Hausdorff property ensures uniqueness of limits for all of these generalized convergence notions simultaneously. ### Closedness of Singletons and the Diagonal Characterisation The Hausdorff condition is equivalent to several other properties that are useful in different contexts. [quotetheorem:1025] The equivalence between (1) and (2) is the most useful reformulation. In a Hausdorff space, if $(x, y) \notin \Delta$ (i.e., $x \neq y$), then there exist disjoint open sets $U \ni x$ and $V \ni y$, so $U \times V$ is an open neighbourhood of $(x,y)$ in the product topology that is disjoint from $\Delta$. Hence $X \times X \setminus \Delta$ is open, and $\Delta$ is closed. The equivalence with (3) is particularly important in algebraic topology and algebraic geometry. If $f, g: Y \to X$ are continuous maps into a Hausdorff space, the set where they agree is closed. This has an immediate consequence: if two continuous maps agree on a [dense subset](/page/Dense%20Subset), they agree everywhere. In analysis, this is the principle behind unique extension theorems — a bounded linear operator on a Hilbert space is determined by its values on any dense subspace. [example: Two Continuous Maps Agreeing on a Dense Subset] Let $X = \mathbb{R}$ with the standard topology (which is Hausdorff), and let $f, g: \mathbb{R} \to \mathbb{R}$ be continuous functions satisfying $f(q) = g(q)$ for every $q \in \mathbb{Q}$. The set $E = \{x \in \mathbb{R} : f(x) = g(x)\}$ is closed (by characterisation (3) above, since $\mathbb{R}$ is Hausdorff) and contains $\mathbb{Q}$, which is [dense](/page/Dense%20Subset) in $\mathbb{R}$. Therefore $E \supset \overline{\mathbb{Q}} = \mathbb{R}$, so $f = g$. This argument fails in non-Hausdorff spaces. On the cofinite topology on $\mathbb{R}$, every function $f: \mathbb{R} \to \mathbb{R}$ whose preimage of every cofinite set is cofinite is continuous. Many non-equal functions agree on dense subsets without being forced to agree everywhere, because the equaliser need not be closed. [/example] ## Compactness and the Hausdorff Property The interaction between [compactness](/page/Compact%20Space) and the Hausdorff condition produces some of the most powerful results in topology. In a Hausdorff space, compact sets inherit a rigidity that general closed sets do not possess: they are closed, they can be separated from points by open sets, and continuous maps from compact spaces to Hausdorff spaces are automatically well-behaved. The following results underpin much of analysis. In the theory of [Sobolev spaces](/page/Sobolev%20Space), the interplay between weak compactness and the Hausdorff property of the norm topology is what makes the direct method in the calculus of variations work. In differential geometry, the compact-Hausdorff structure of manifolds guarantees that partitions of unity exist and that local constructions can be globalised. ### Compact Subsets Are Closed In a general topological space, there is no relationship between compactness and closedness. A [compact](/page/Compact%20Space) subset of a topological space need not be closed, and a closed subset of a compact space need not be compact (the latter does hold, however — closed subsets of compact spaces are always compact). The Hausdorff property rectifies the first of these failures. [quotetheorem:307] The idea of the argument reveals a general technique. Fix a point $x \in X \setminus K$. For each $y \in K$, the Hausdorff property provides disjoint open sets $U_y \ni x$ and $V_y \ni y$. The collection $\{V_y\}_{y \in K}$ is an open cover of $K$, and compactness extracts a finite subcover $V_{y_1}, \ldots, V_{y_m}$. The finite intersection $U = U_{y_1} \cap \cdots \cap U_{y_m}$ is an open neighbourhood of $x$ disjoint from $V_{y_1} \cup \cdots \cup V_{y_m} \supset K$. Since $x$ was arbitrary, $X \setminus K$ is open, so $K$ is closed. This argument illustrates the "separate and shrink" method that recurs throughout the theory: use the Hausdorff axiom to produce separating open sets for individual points, then use compactness to reduce an infinite collection to a finite one, and finally take a finite intersection (which is still open) to obtain a single separating set. Without the Hausdorff hypothesis, the theorem fails. In the line with two origins $(X, \tau)$, the subset $\{p\}$ is compact (it is a single point), but it is not closed: its complement $X \setminus \{p\} = (\mathbb{R} \setminus \{0\}) \cup \{q\}$ is not open, because every open set containing $q$ must also contain points of $\mathbb{R} \setminus \{0\}$ that are "shared" with $p$'s neighbourhood system. ### Separation of Compact Sets The Hausdorff axiom allows us to separate points from compact sets. With slightly more work, we can separate compact sets from each other. [quotetheorem:1026] Part (1) strengthens the Hausdorff axiom from point-point separation to point-compact separation — it says that compact Hausdorff spaces are regular. Part (2) further upgrades to compact-compact separation. These results are notable because they show that compactness acts as a "finiteness proxy" that amplifies the Hausdorff axiom's pointwise separation to set-level separation. An important consequence is that every [compact](/page/Compact%20Space) Hausdorff space is normal: disjoint closed subsets of a compact Hausdorff space are compact (since closed subsets of compact spaces are compact), and part (2) provides the required disjoint open neighbourhoods. This means that [Urysohn's Lemma](/theorems/887) applies in any compact Hausdorff space, guaranteeing the existence of continuous functions separating disjoint closed sets — a fact essential for the [Stone-Weierstrass Theorem](/theorems/886) and the [Riesz-Markov-Kakutani Representation Theorem](/theorems/976). ### The Closed Map Lemma and Automatic Homeomorphisms One of the most useful results in point-set topology is that continuous maps from compact spaces to Hausdorff spaces are automatically closed maps. This is the content of the [Closed Map Lemma](/theorems/317). [quotetheorem:317] The mechanism is clean: a closed subset of a compact space is compact (since $X$ is compact), so $C$ is compact; the continuous image of a compact set is compact, so $f(C)$ is compact in $Y$; and a compact subset of a Hausdorff space is closed (by the theorem above), so $f(C)$ is closed in $Y$. The Closed Map Lemma has an immediate and powerful corollary: a continuous bijection from a compact space to a Hausdorff space is a homeomorphism. [quotetheorem:318] This result is used constantly in practice. Whenever one constructs a continuous bijection from a compact space to a Hausdorff space — by quotienting, by restricting a known homeomorphism, or by composing continuous maps — the inverse continuity is free. The compactness of the domain and the Hausdorff property of the codomain together do the work that would otherwise require an explicit verification of inverse continuity. [example: Continuous Bijection That Is Not a Homeomorphism] The hypothesis that $Y$ is Hausdorff is essential, but the compactness of $X$ is equally critical. Consider the map \begin{align*} f: [0, 2\pi) &\to S^1 \\ t &\mapsto (\cos t, \sin t), \end{align*} where $S^1 = \{(x,y) \in \mathbb{R}^2 : x^2 + y^2 = 1\}$ is the unit circle with the subspace topology inherited from $\mathbb{R}^2$. This map is continuous (as a composition of continuous functions) and bijective (each point on the circle corresponds to a unique angle in $[0, 2\pi)$). The codomain $S^1$ is Hausdorff (as a subspace of $\mathbb{R}^2$). However, $f$ is **not** a homeomorphism. To see the failure, note that $[0, 2\pi)$ is not compact: it is not closed in $\mathbb{R}$. The inverse $f^{-1}: S^1 \to [0, 2\pi)$ is discontinuous at the point $(1, 0) = f(0)$. Consider the sequence $z_k = (\cos(2\pi - 1/k), \sin(2\pi - 1/k))$ on the circle, which converges to $(1,0)$ in $S^1$. The preimage $f^{-1}(z_k) = 2\pi - 1/k$ converges to $2\pi$, but $f^{-1}(1, 0) = 0$. Since $2\pi - 1/k \not\to 0$, the inverse is discontinuous. By contrast, the restriction $g: [0, \pi] \to S^1_+$ to the upper semicircle is a continuous bijection from a **compact** space to a Hausdorff space, so it is automatically a homeomorphism. [/example] ### Compact Hausdorff Spaces When a space is both compact and Hausdorff, the two properties interact to produce an especially rigid topological structure. The compact subsets are exactly the closed subsets (a closed subset of a compact space is compact; a compact subset of a Hausdorff space is closed). This means the topology is "self-dual" with respect to compactness and closedness. [quotetheorem:1027] The normality of compact Hausdorff spaces connects the theory to some of the deepest results in topology and functional analysis. By [Urysohn's Lemma](/theorems/887), disjoint closed subsets of a compact Hausdorff space can be separated by continuous functions. By the [Tietze Extension Theorem](/page/Topology), continuous real-valued functions on closed subsets extend to the whole space. By the [Stone-Weierstrass Theorem](/theorems/886), subalgebras of $C(K)$ that separate points are dense. And by the [Riesz-Markov-Kakutani Representation Theorem](/theorems/976), the dual of $C(K)$ is the space of signed Borel measures on $K$. ## Products, Subspaces, and Quotients A natural question for any topological property is its permanence: does it pass to subspaces, products, and quotients? For the Hausdorff condition, the answers are yes, yes, and emphatically no — and the failure for quotients is the primary source of non-Hausdorff spaces encountered in practice. ### Subspaces The Hausdorff property passes to all subspaces. If $(X, \tau)$ is Hausdorff and $A \subset X$, then the [subspace topology](/page/Subspace%20Topology) on $A$ is Hausdorff: distinct points $x, y \in A$ can be separated by disjoint open sets $U, V$ in $X$, and then $U \cap A$ and $V \cap A$ are disjoint open sets in $A$ separating $x$ from $y$. ### Products The Hausdorff property is preserved under arbitrary products. [quotetheorem:1028] The forward direction is immediate from the subspace property: each $X_\alpha$ embeds homeomorphically into the product (by fixing all other coordinates), and a subspace of a Hausdorff space is Hausdorff. For the converse, let $x = (x_\alpha)$ and $y = (y_\alpha)$ be distinct points in the product. Then $x_\beta \neq y_\beta$ for some index $\beta$. Since $X_\beta$ is Hausdorff, there exist disjoint open sets $U_\beta, V_\beta \subset X_\beta$ with $x_\beta \in U_\beta$ and $y_\beta \in V_\beta$. The sets $\pi_\beta^{-1}(U_\beta)$ and $\pi_\beta^{-1}(V_\beta)$ are disjoint open sets in the product separating $x$ from $y$. This result is important for functional analysis, where many spaces arise as products. The [weak topology](/page/Weak%20Topology) on a Banach space $X$, for instance, is the initial topology induced by the family of evaluation functionals $\{f \in X^* : f \mapsto f(x)\}$. Since $\mathbb{R}$ is Hausdorff, the weak topology — which embeds $X$ into a product of copies of $\mathbb{R}$ — is Hausdorff. ### Quotients: The Primary Source of Non-Hausdorff Spaces The Hausdorff property is **not** preserved under quotient maps, and this is the principal mechanism by which non-Hausdorff spaces arise in mathematics. Given a Hausdorff space $X$ and an equivalence relation $\sim$ on $X$, the quotient space $X/{\sim}$ with the quotient topology need not be Hausdorff, even if $X$ is as well-behaved as $\mathbb{R}$. [example: Collapsing the Rationals] Let $X = \mathbb{R}$ with the standard topology and define the equivalence relation $x \sim y$ if and only if $x - y \in \mathbb{Q}$. Each equivalence class is a coset of $\mathbb{Q}$ in $\mathbb{R}$, and by the density of $\mathbb{Q}$, every equivalence class is [dense](/page/Dense%20Subset) in $\mathbb{R}$. The quotient space $Y = \mathbb{R}/{\sim}$ has the **indiscrete topology**: the only open sets are $\varnothing$ and $Y$ itself. To see this, let $U \subset Y$ be a nonempty open set. Its preimage $\pi^{-1}(U) \subset \mathbb{R}$ is open (by the definition of the quotient topology) and **saturated** — it is a union of equivalence classes, meaning that if $x \in \pi^{-1}(U)$ and $x \sim y$, then $y \in \pi^{-1}(U)$. Now, $\pi^{-1}(U)$ is a nonempty open subset of $\mathbb{R}$, so it contains some interval $(a, b)$. Let $[z]$ be any equivalence class. Since $[z] = z + \mathbb{Q}$ is dense in $\mathbb{R}$, the class $[z]$ intersects $(a, b)$: there exists $z + q \in (a, b)$ for some $q \in \mathbb{Q}$. This point $z + q$ lies in $\pi^{-1}(U)$, and since $\pi^{-1}(U)$ is saturated and $z \sim z + q$, the entire class $[z]$ is contained in $\pi^{-1}(U)$. Since $[z]$ was arbitrary, $\pi^{-1}(U) = \mathbb{R}$ and $U = Y$. In particular, $Y$ is not Hausdorff (in fact, it is not even $T_0$). This is the standard example showing that quotients of Hausdorff spaces need not be Hausdorff. [/example] The characterisation of when a quotient is Hausdorff uses the diagonal. Recall that $(X, \tau)$ is Hausdorff if and only if the diagonal $\Delta \subset X \times X$ is closed. For quotients, the analogous criterion is as follows. [quotetheorem:1029] This criterion explains why the quotient $\mathbb{R}/\mathbb{Q}$ fails to be Hausdorff: the equivalence relation $R = \{(x, y) : x - y \in \mathbb{Q}\}$ is dense in $\mathbb{R} \times \mathbb{R}$ (since $\mathbb{Q}$ is dense in $\mathbb{R}$), and a dense set is closed only if it equals the whole space. Since $R \neq \mathbb{R} \times \mathbb{R}$ (take $x = 0, y = \sqrt{2}$), $R$ is not closed. In practice, this criterion is used to verify that group quotients and orbit spaces are Hausdorff. If a topological group $G$ acts continuously on a Hausdorff space $X$, the orbit space $X/G$ is Hausdorff if and only if the set $\{(x, g \cdot x) : x \in X, g \in G\}$ is closed in $X \times X$. For compact group actions, this condition is automatically satisfied. ## Standard Arguments Using the Hausdorff Property The Hausdorff axiom is invoked in dozens of arguments throughout topology and analysis. Most of these arguments use a small number of recurring patterns, which we collect here. ### The "Separate and Shrink" Method This is the most fundamental technique. When two objects (points, a point and a compact set, two compact sets) need to be separated, the strategy is: 1. **Separate at each point.** For each "obstruction point" $y$, use the Hausdorff property to produce disjoint open sets $U_y \ni x$ and $V_y \ni y$. 2. **Extract a finite subcover.** If the collection of obstruction points forms a compact set, extract a finite subcover $V_{y_1}, \ldots, V_{y_m}$. 3. **Shrink by intersection.** The finite intersection $U = U_{y_1} \cap \cdots \cap U_{y_m}$ remains open (finite intersections of open sets are open) and is disjoint from the union $V_{y_1} \cup \cdots \cup V_{y_m}$. This method was used in the proof that compact subsets of Hausdorff spaces are closed, and it is the engine behind the separation of compact sets and the normality of compact Hausdorff spaces. The same pattern appears in the proof that locally compact Hausdorff spaces have "enough" compact neighbourhoods — a fact essential for the theory of $C_c(X)$ and the Riesz representation theorem. [example: Every Compact Hausdorff Space Is Regular] We illustrate the "separate and shrink" technique by proving that a compact Hausdorff space $(X, \tau)$ is regular: given a closed set $C \subset X$ and a point $x \notin C$, we produce disjoint open sets separating them. **Step 1: Separate at each point.** For each $y \in C$, the Hausdorff property yields disjoint open sets $U_y \ni x$ and $V_y \ni y$. **Step 2: Extract a finite subcover.** The collection $\{V_y\}_{y \in C}$ covers $C$. Since $C$ is closed in a compact space, $C$ is itself compact. Extract a finite subcover: $C \subset V_{y_1} \cup \cdots \cup V_{y_m}$. **Step 3: Shrink by intersection.** Set $U = U_{y_1} \cap \cdots \cap U_{y_m}$ and $V = V_{y_1} \cup \cdots \cup V_{y_m}$. Then $U$ is open (finite intersection), $V$ is open (finite union), $x \in U$, $C \subset V$, and $U \cap V = \varnothing$ since $U \subset U_{y_j}$ and $U_{y_j} \cap V_{y_j} = \varnothing$ for each $j$, so $U \cap V_{y_j} = \varnothing$ for each $j$. This is the standard argument; it appears in virtually every topology textbook. The structure — Hausdorff gives local separation, compactness finitises, intersection preserves openness — is the template for many more sophisticated arguments. [/example] ### The Compact-Hausdorff Duality In a compact Hausdorff space, the lattices of open sets and closed sets interact with compactness in a particularly symmetric way: closed $\Leftrightarrow$ compact. This duality simplifies many arguments: - To show a set is closed, show it is compact (e.g., the image of a compact set under a continuous map into a Hausdorff space). - To show a set is compact, show it is closed in a compact ambient space (this does not require the Hausdorff property, but the converse direction does). This duality is the reason why compact Hausdorff spaces have such clean structure theory. The [Cantor Intersection Theorem](/theorems/624) — a decreasing sequence of nonempty compact sets in a Hausdorff space has nonempty intersection — relies on this duality: the compact sets are closed, so the intersection of any finite subcollection is nonempty (it is a closed subset of a compact set, hence compact and nonempty by the finite intersection property), and the full intersection is nonempty by compactness. ### The Hausdorff Condition as a Uniqueness Tool Many uniqueness arguments in analysis reduce to the Hausdorff property. The pattern is: 1. Construct two candidates $x$ and $y$ (e.g., two limits of the same sequence, two fixed points of a contraction, two extensions of a continuous function). 2. Observe that both candidates arise as limits of the same net or as values of continuous maps that agree on a dense set. 3. Invoke the Hausdorff property (either directly via uniqueness of limits, or via the closed diagonal/equaliser characterisation) to conclude $x = y$. [example: Uniqueness of Extensions from Dense Subsets] Let $(X, \tau)$ be a topological space, let $D \subset X$ be [dense](/page/Dense%20Subset), and let $Y$ be a Hausdorff space. Suppose $f, g: X \to Y$ are continuous maps satisfying $f|_D = g|_D$. Define $E = \{x \in X : f(x) = g(x)\}$. The continuous map $(f, g): X \to Y \times Y$ defined by $x \mapsto (f(x), g(x))$ has $E = (f,g)^{-1}(\Delta_Y)$, where $\Delta_Y = \{(y,y) : y \in Y\}$ is the diagonal in $Y \times Y$. Since $Y$ is Hausdorff, $\Delta_Y$ is closed, and since $(f,g)$ is continuous, $E$ is closed. By hypothesis, $E \supset D$. Since $D$ is dense, $\overline{D} = X$, and since $E$ is closed and contains $D$, we obtain $E \supset \overline{D} = X$. Therefore $f = g$. This argument is the foundation of the [Continuous Extension Theorem](/theorems/964): if a uniformly continuous function is defined on a dense subset of a metric space and maps into a complete metric space, it has a unique continuous extension to the whole space. The uniqueness part is precisely the above argument; the existence part requires completeness. [/example] ## References Munkres, J., *Topology* (2000). Willard, S., *General Topology* (1970). Kelley, J., *General Topology* (1955). Engelking, R., *General Topology* (1989). Bourbaki, N., *General Topology, Chapters 1--4* (1995).