Baire Category Theorem [dark version]

TL;DR

In a complete metric space, the intersection of countably many dense open sets is dense, provided openness is included.

Briefing Cornell Notes

Briefing

Baire category theorem turns a topological “largeness” idea into a practical existence tool: in a complete metric space, you can’t cover a nonempty open set with countably many nowhere-dense pieces. The theorem matters because it guarantees that certain “bad” behaviors (sets that are small in the category sense) cannot exhaust the space—so “good” objects must exist without explicitly constructing them.

A common metric-space version starts with a complete metric space X and a sequence of subsets Q_j that are both dense and open. “Dense” means the closure of Q_j is all of X, so every point can be approximated by points from Q_j. When each Q_j is open and dense, the countable intersection ⋂_{j=1}^∞ Q_j remains dense in X. The openness condition is essential: without it, completeness alone does not prevent pathological counterexamples.

To state broader forms, the theorem is reframed using topological notions. A set M is nowhere dense in X if the closure of M has empty interior; equivalently, M is so thin that no open ball sits entirely inside its closure. Sets that are “small” in this sense are called meager (or first category): a meager set can be written as a countable union of nowhere-dense sets. The complement notion—sets that are not meager—are called non-meager (or second category). With these definitions, the theorem becomes: in a complete metric space X, every nonempty open set is non-meager. Put differently, no nonempty open set can be expressed as a countable union of nowhere-dense sets.

Completeness is the dividing line. The rationals Q inside the reals R illustrate why: Q is countable and can be written as a countable union of single points, yet Q is not “large” in the complete space R. In contrast, the real line (with its usual metric) is complete, and the Baire category conclusion holds.

The transcript then connects the theorem to functional analysis. Consider the space C[0,1] of continuous functions on the unit interval, equipped with a complete metric (typically the supremum norm). The strategy is to partition C[0,1] into two parts: a countable union of sets A_j designed so that functions in ⋃ A_j are differentiable at least at one point, and a remaining set B that consists of functions that are nowhere differentiable (at least a subset of them). Applying the Baire category theorem shows that C[0,1] is non-meager, so B cannot be meager; in fact, B is dense in C[0,1]. The payoff is existence and abundance: nowhere-differentiable continuous functions not only exist, but they form a dense (hence “large”) subset of C[0,1], achieved without constructing any specific example.

Cornell Notes

Baire category theorem says that in a complete metric space, “small” sets built from nowhere-dense pieces cannot cover nonempty open sets. A dense open set remains dense under countable intersections, and the more general language uses category: nowhere-dense sets have closures with empty interior; meager sets are countable unions of nowhere-dense sets. In a complete metric space, every nonempty open set is non-meager (second category), so it cannot be written as a countable union of nowhere-dense sets. Completeness is crucial: the rationals inside the reals fail the conclusion because R is complete but Q is not. In functional analysis, this yields that nowhere-differentiable continuous functions on [0,1] form a dense, non-meager subset of C[0,1].

What does “dense” mean in the Baire category theorem’s metric formulation, and why does it matter?

A subset Q_j of X is dense if its closure equals the whole space: cl(Q_j)=X. That means every point of X can be approximated by points from Q_j. This approximation property is what allows the theorem to preserve “largeness” when taking intersections of many sets.

Why must the sets be open in the statement “a countable intersection of dense open sets is dense”?

Openness prevents the intersection from losing density due to boundary effects. The transcript emphasizes that openness is an “important ingredient” and that dropping it can break the conclusion, even if each set is dense. The theorem’s safe form is specifically for dense open sets in a complete metric space.

How is “nowhere dense” defined, and how does it relate to “meager” sets?

A set M is nowhere dense in X if the interior of its closure is empty: int(cl(M))=∅. Intuitively, no open region of X lies entirely within cl(M). A meager (first category) set is then a countable union of nowhere-dense sets, so it is “small” in the category sense.

What is the key Baire category conclusion in a complete metric space?

Every nonempty open set in a complete metric space is non-meager (second category). Equivalently, no nonempty open set can be expressed as a countable union of nowhere-dense sets. This is the version most useful for existence arguments.

How does the theorem imply that nowhere-differentiable functions are dense in C[0,1]?

The space C[0,1] is complete under a standard metric such as the supremum norm. One partitions C[0,1] into a countable union ⋃ A_j (constructed so that functions in ⋃ A_j are differentiable at least at one point) and a leftover set B containing functions that are nowhere differentiable. Since C[0,1] is non-meager, B cannot be meager; the transcript notes that non-meager implies B is dense in C[0,1]. Thus nowhere-differentiable continuous functions are not just present—they are dense.

Review Questions

In a complete metric space, what prevents a nonempty open set from being written as a countable union of nowhere-dense sets?
State the definition of nowhere dense using closure and interior, and explain how it differs from merely being “not dense.”
How does the Baire category theorem convert a problem about differentiability into an existence result for nowhere-differentiable continuous functions?

Key Points

1
In a complete metric space, the intersection of countably many dense open sets is dense, provided openness is included.
2
A set is nowhere dense when the interior of its closure is empty, meaning it cannot fill any open region even after closing.
3
Meager (first category) sets are countable unions of nowhere-dense sets; non-meager (second category) sets are those that cannot be so decomposed.
4
Baire category theorem: every nonempty open set in a complete metric space is non-meager, so it cannot be covered by countably many nowhere-dense sets.
5
Completeness is essential; the rationals inside the reals illustrate how the conclusion can fail without completeness.
6
In C[0,1] (a complete metric space), a Baire-category partition shows that nowhere-differentiable continuous functions form a dense, non-meager subset, yielding existence without explicit construction.

Highlights

A nonempty open set in a complete metric space cannot be expressed as a countable union of nowhere-dense sets.

Nowhere dense means even the closure has no interior points—no open ball can hide inside it.

Completeness is the safeguard: Q can be written as a countable union of single points, but the Baire conclusion relies on working in a complete space.

Using C[0,1] and a countable decomposition tied to differentiability, the “leftover” set of nowhere-differentiable functions is dense.

The theorem often proves existence by showing the “bad” set must be non-meager, without constructing any specific example.

Topics

Baire Category Theorem
Complete Metric Spaces
Nowhere Dense Sets
Meager Sets
Nowhere Differentiable Functions