Dense sets, small sets, and the question
Definition 9.1.1
(Dense, nowhere dense and meager sets)
A subset of a topological space is dense if every nonempty open set meets . It is nowhere dense if the interior of its closure is empty. A set is meager if it is a countable union of nowhere dense sets. These definitions describe topology, not length or probability: a dense set may still be meager. For example, the rational numbers are a countable union of singletons, each nowhere dense in .
Two dense open subsets have dense open intersection: every nonempty open meets the first in a nonempty open set, which meets the second. The key question is whether a countable intersection remains dense. Completeness supplies the missing step.
The Baire theorem and proof
Theorem 9.1.2
(Baire category theorem)
Let be a complete metric space and let be open dense subsets. Then is dense in . The empty space is included vacuously.
Proof
(Nested closed balls)
Fix a nonempty open . Density of makes nonempty and open. Choose a closed ball contained in it with . Here ; this closed ball may be larger than the closure of the corresponding open ball, which does not affect the proof.
Suppose has been selected. Its open ball is nonempty, so its intersection with is nonempty and open. Choose a closed ball satisfying
Such a ball exists because a sufficiently small closed ball around any point of an open set lies inside it. For , nestedness places in , so . Thus is Cauchy. Completeness gives a limit , and every closed ball contains all sufficiently late centers, hence contains . Consequently
Since was arbitrary, the intersection is dense. The proof uses completeness to keep the limit in ; mere density of each does not do that.
Proof X-ray
What each hypothesis does in the nested-ball proof
Each ingredient is used once. Density of makes the next open set nonempty, and openness lets a closed ball fit inside it, so a point lying in every chosen ball lies in every . Radii tending to zero make the centers Cauchy, and completeness supplies their limit inside . Closed balls are what capture that limit: each closed ball is a closed set containing all later centers, so it contains their limit. A limit of points of an open ball can lie on its boundary, outside the ball; choosing each closed ball inside the previous open ball, and inside , removes that danger. Starting inside an arbitrary is what turns the conclusion into density.
Corollary 9.1.3
(Meager sets have empty interior)
In a complete metric space every meager subset has empty interior. In particular, a nonempty complete metric space is not a countable union of closed nowhere dense subsets.
Proof
(Pass to closures and complements)
Let with every nowhere dense; a finite union is included by adding empty sets. Each closure is closed and nowhere dense, so is open and dense. If is open and , then misses , which is dense by the theorem. Hence .
Worked examples and a necessary hypothesis
Worked example 9.1.4
(Irrationals meet every interval)
In , write . Each is open and dense. Baire's theorem shows that the irrational numbers meet every nonempty open interval. It also implies is not countable. This does not say the rationals are not dense: is both dense and meager.
Worked example 9.1.5
(Empty interior is not nowhere dense)
The set has empty interior, since every interval contains irrationals. Its closure, however, is , whose interior is , so is not nowhere dense. Nowhere density is a condition on the closure. The set is nevertheless meager, as a countable union of nowhere dense singletons.
Worked example 9.1.6
(Nowhere density depends on the ambient space)
In each singleton is closed with empty interior, hence nowhere dense. In the subspace with the usual metric, is open, so it is not nowhere dense in . The space is complete, because a Cauchy sequence of integers is eventually constant, and it is a countable union of singletons. This is consistent with Baire's theorem precisely because those singletons are not nowhere dense in .
Counterexample mode
An incomplete space where the intersection is empty
Claim: the conclusion of Baire's theorem holds in every metric space. Give its usual subspace metric and enumerate it as . Each is open in , since a singleton is closed, and dense, because no rational point is isolated. Yet . The nested-ball proof breaks at its last step: the centers form a Cauchy sequence of rationals, which can converge to an irrational number outside the space. Completeness is the repair. This shows failure in one incomplete space, not that every incomplete metric space fails the theorem.
Common mistake
(A dense need not be open)
A countable intersection of dense open sets need not itself be open. In , let . The intersection contains , but every neighborhood of contains some excluded .
Common mistake
(Meager does not mean countable)
The middle-thirds Cantor set is closed and contains no interval, since after removal stages it lies in intervals of length ; hence it is nowhere dense and in particular meager. Yet is uncountable: its points are exactly the numbers with a ternary expansion using only the digits and , and these correspond bijectively to infinite binary sequences. Category measures topological smallness, not cardinality.
Exercises with solutions
Quick check
(Is nowhere dense in ?)
Solution · Answer
Yes. It is closed and has empty interior; every open interval contains points outside it.
Quick check
(Why does the density of not contradict its being meager?)
Solution · Answer
Density asks whether every nonempty open interval meets the union. Meagerness asks for a countable decomposition into nowhere dense sets. Different conditions can hold simultaneously.
Quick check
(If is closed and nowhere dense in a complete metric space , can a nonempty open set be contained in ?)
Solution · Answer
No. Their complements are open dense; their intersection must meet .
Quick check
(In the nested-ball construction, why does an arbitrary metric space not suffice?)
Solution · Answer
The centers are Cauchy, but without completeness their limit may lie outside , leaving the nested closed balls with empty intersection in .