How large can a "small" set be? More precisely: if we decompose a space into countably many pieces, each of which is "thin" in the sense that it is dense nowhere, can these thin pieces exhaust the entire space? In finite-dimensional Euclidean spaces and in the function spaces of analysis, intuition suggests that countably many negligible sets should not fill up everything — but intuition alone does not settle the matter. The rational numbers $\mathbb{Q}$, for instance, can be written as a countable union of singletons, each nowhere dense, and these singletons do exhaust $\mathbb{Q}$. What property of $\mathbb{Q}$ permits this, and what property of $\mathbb{R}$ prevents it?

The answer is [completeness](/page/Complete%20Metric%20Space). The Baire Category Theorem asserts that in a [complete metric space](/page/Complete%20Metric%20Space) (or, more generally, in a [locally compact](/page/Locally%20Compact%20Space) [Hausdorff space](/page/Hausdorff%20Space)), the space cannot be written as a countable union of nowhere dense sets. Equivalently, the countable intersection of dense open sets is dense. This result, due to René-Louis Baire (1899), is one of the most powerful existence tools in analysis: it proves that objects with certain properties exist — often in overwhelming abundance — without constructing a single example.

[example: The Rationals Are Exhaustible] The [metric space](/page/Metric%20Space) $(\mathbb{Q}, |\cdot|)$ is countable, so we may enumerate it as $\mathbb{Q} = \{q_1, q_2, q_3, \ldots\}$. Each singleton $\{q_k\}$ is [nowhere dense](/page/Nowhere%20Dense%20Set) in $\mathbb{Q}$: its closure in $\mathbb{Q}$ is $\{q_k\}$ itself, which has empty interior because every open interval in $\mathbb{Q}$ contains infinitely many rationals besides $q_k$. Thus $\mathbb{Q} = \bigcup_{k=1}^\infty \{q_k\}$ is a countable union of nowhere dense sets. Now consider the same decomposition attempt in $\mathbb{R}$. Each singleton $\{q_k\}$ is still nowhere dense in $\mathbb{R}$, so $\mathbb{Q} = \bigcup_{k=1}^\infty \{q_k\}$ is a countable union of nowhere dense sets in $\mathbb{R}$. But $\mathbb{Q}$ is far from all of $\mathbb{R}$ — its complement, the set of irrationals $\mathbb{R} \setminus \mathbb{Q}$, is dense. The Baire Category Theorem guarantees that this must be the case: $\mathbb{R}$, being complete, cannot equal a countable union of nowhere dense sets. [/example]

The distinction between $\mathbb{Q}$ and $\mathbb{R}$ in this example is not accidental — it is the entire point. The Baire Category Theorem draws a sharp line between spaces where countable decompositions into thin sets are possible (incomplete spaces) and spaces where they are not (complete or locally compact Hausdorff spaces). This line has far-reaching consequences: it is the foundation of the [Uniform Boundedness Principle](/page/Uniform%20Boundedness%20Principle), the [Open Mapping Theorem](/page/Open%20Mapping%20Theorem), and the [Closed Graph Theorem](/page/Closed%20Graph%20Theorem), and it provides the standard method for proving that "generic" objects in function spaces have extreme pathological properties.

## Definition

The Baire Category Theorem concerns the distinction between "large" and "small" subsets of a topological space, where "small" is measured not by cardinality or measure but by topological density. The central definitions formalise what it means for a set to be topologically negligible.

A set is [nowhere dense](/page/Nowhere%20Dense%20Set) when its closure fails to contain any open ball — it is "thin" in the sense that it is not dense in any region of the space. A countable union of such thin sets is called **meagre** (or of **first category**), and a set that is not meagre is of **second category**. These terms, due to Baire, classify subsets by their topological thickness.

[definition: Baire Space] Let $X$ be a [topological space](/page/Topology). We say that $X$ is a **Baire space** if whenever $(G_n)_{n=1}^\infty$ is a countable family of [dense](/page/Dense%20Subset) [open](/page/Topology) subsets of $X$, the intersection $\bigcap_{n=1}^\infty G_n$ is dense in $X$. Equivalently, $X$ is a Baire space if every [meagre](/page/Nowhere%20Dense%20Set) subset of $X$ (every countable union of [nowhere dense](/page/Nowhere%20Dense%20Set) sets) has empty [interior](/page/Interior). Equivalently again, $X$ is a Baire space if no nonempty open subset of $X$ is meagre. [/definition]

The three characterisations are logically equivalent (by taking complements and using the definition of density), but they serve different purposes. The dense $G_\delta$ formulation — "countable intersections of dense open sets are dense" — is the tool for proving **existence** results: if each $G_n$ captures functions satisfying some approximation property, the intersection captures functions satisfying all such properties simultaneously, and it is nonempty (in fact dense). The category formulation — "no nonempty open set is meagre" — is the tool for proving **impossibility** results: a Baire space cannot be exhausted by countably many thin pieces.

The Baire Category Theorem itself identifies two large and practically important classes of Baire spaces.

## The Baire Category Theorem

The fundamental question that the Baire Category Theorem answers is: which spaces are Baire spaces? For arbitrary topological spaces, the Baire property can fail — the rationals $\mathbb{Q}$ with their usual topology are not a Baire space, as the opening example shows. But the two most important classes of spaces in analysis — complete metric spaces and locally compact Hausdorff spaces — are always Baire.

[quotetheorem:1072]

[quotetheorem:1105]

The two versions are independent: neither hypothesis implies the other. The space $\mathbb{R}$ satisfies both (it is a complete metric space and locally compact Hausdorff). The space $\ell^2(\mathbb{N})$ — the Hilbert space of square-summable sequences — is a complete metric space but not locally compact (the closed unit ball is not compact in infinite dimensions). Conversely, the space $\mathbb{Q}_p$ of $p$-adic numbers is both complete and locally compact, but a non-complete locally compact Hausdorff space can be constructed by taking an open subset of a compact Hausdorff space that is not completely metrizable.

The key idea behind the proof of the complete metric space version deserves a brief sketch, since the technique — a **nested sequence of shrinking balls** — is itself a fundamental tool. Given a dense open set $G_1$ and any open ball $B(x_0, r_0)$, density of $G_1$ guarantees a point $x_1 \in G_1 \cap B(x_0, r_0)$; since $G_1$ is open, there exists $r_1 < r_0/2$ with $\overline{B}(x_1, r_1) \subset G_1 \cap B(x_0, r_0)$. Repeating with $G_2$, we find $x_2 \in G_2 \cap B(x_1, r_1)$ and $r_2 < r_1/2$ with $\overline{B}(x_2, r_2) \subset G_2 \cap B(x_1, r_1)$. The sequence $(x_n)$ is Cauchy because $d(x_n, x_m) \le 2r_n$ for $m \ge n$ (both points lie in $\overline{B}(x_n, r_n)$, and $r_n < 2^{-n} r_0 \to 0$). Completeness provides a limit $x = \lim x_n$, and the nested inclusion $x \in \overline{B}(x_n, r_n) \subset G_n$ for every $n$ shows $x \in \bigcap G_n \cap B(x_0, r_0)$. Since $B(x_0, r_0)$ was arbitrary, the intersection is dense.

The completeness hypothesis enters at exactly one point: to guarantee that the Cauchy sequence $(x_n)$ converges. Without completeness, the nested balls might "shrink to nothing" — the intersection could be empty. This is precisely what happens in $\mathbb{Q}$: one can construct nested closed balls in $\mathbb{Q}$ with radii tending to zero whose intersection is empty (choose the centres converging to an irrational number).

### The Hypothesis Cannot Be Dropped

The completeness (or local compactness) hypothesis is essential, and the failure modes are instructive.

[example: The Rationals Are Not a Baire Space] The space $\mathbb{Q}$ with the standard metric is not complete. We show directly that $\mathbb{Q}$ is not a Baire space by exhibiting it as a countable union of nowhere dense sets. Enumerate $\mathbb{Q} = \{q_1, q_2, \ldots\}$. Each singleton $\{q_k\}$ is closed in $\mathbb{Q}$ (since $\mathbb{Q}$ is a metric space, hence $T_1$). Each singleton has empty interior: for any $\varepsilon > 0$, the open ball $B(q_k, \varepsilon) \cap \mathbb{Q}$ contains infinitely many rationals besides $q_k$, so $\{q_k\}$ contains no nonempty open set. Therefore each $\{q_k\}$ is nowhere dense. Yet \begin{align*} \mathbb{Q} = \bigcup_{k=1}^\infty \{q_k\}, \end{align*} which shows that $\mathbb{Q}$ is meagre in itself. Since $\mathbb{Q}$ has nonempty interior in itself (every open interval is nonempty), the interior of $\mathbb{Q}$ in $\mathbb{Q}$ is all of $\mathbb{Q}$, which is not empty. So $\mathbb{Q}$ equals a meagre set that has nonempty interior — exactly the condition that the Baire property prohibits. [/example]