Nagata-Smirnov Metrization Theorem
The Nagata-Smirnov metrization theorem provides a complete characterization of metrizable spaces, removing the second-countability assumption of the Urysohn theorem. It states that a space is metrizable if and only if it is and has a -locally finite basis.
Statement
A topological space is metrizable if and only if is (regular and Hausdorff) and has a -locally finite basis, i.e., a basis where each is a locally finite family of open sets.
Motivation and Context
The Urysohn metrization theorem states: regular + second-countable metrizable. Every countable family is -locally finite (it is a single locally finite family if it is locally finite, or a countable union of singletons). So the Nagata-Smirnov theorem is strictly more general.
The key insight is that second-countability (a countable basis) is replaced by the weaker condition of a -locally finite basis, which allows for uncountably many basis elements organized into countably many locally finite layers.
Proof Outline
(: Metrizable implies with -locally finite basis.)
Every metrizable space is (Theorem 7.1). For the -locally finite basis: for each , the collection of open balls covers but may not be locally finite. However, using paracompactness (every metrizable space is paracompact), we can refine to a locally finite open cover with each having diameter at most . Then is a -locally finite basis.
(: with -locally finite basis implies metrizable.)
This is the deeper direction. The proof proceeds in several steps:
Step 1: is normal. (Regular + paracompact implies normal, and the existence of a -locally finite basis implies paracompactness.)
Step 2: For each , use the Urysohn lemma to construct a continuous function with and (well, actually for a suitable ).
Step 3: Use the family of functions from Step 2 to embed into a metrizable space. The functions separate points and closed sets. Define an embedding by . With a carefully chosen metric on the codomain that respects the -locally finite structure, this becomes an embedding into a metric space.
Step 4: Verify that is a topological embedding (injective, continuous, open onto its image).
Applications
-
Discrete spaces of any cardinality: A discrete space has basis . This is a single locally finite family (each point has a neighborhood meeting only one basis element). So discrete spaces are metrizable (with the discrete metric for ).
-
for uncountable : The Hilbert space with uncountable index set is metrizable (normed spaces are always metrizable) and has a -locally finite basis, but is not second-countable.
-
Disjoint unions: A disjoint union of metrizable spaces is metrizable (when the union topology is regular and has a -locally finite basis).
Related Metrization Theorems
A topological space is metrizable if and only if it is and has a -discrete basis (i.e., a basis that is a countable union of discrete families, where a family is discrete if every point has a neighborhood meeting at most one member).
| Theorem | Condition | Strength | |---|---|---| | Urysohn | + second-countable | Sufficient, not necessary | | Nagata-Smirnov | + -locally finite basis | Necessary and sufficient | | Bing | + -discrete basis | Necessary and sufficient |
The Nagata-Smirnov and Bing theorems give equivalent necessary and sufficient conditions. The Urysohn theorem is a corollary (every countable basis is -locally finite).
A topological space is metrizable if and only if it is paracompact Hausdorff and locally metrizable.
Smirnov's theorem adds the condition of local metrizability: every point has a metrizable neighborhood. Combined with paracompactness and Hausdorff, this gives global metrizability. This is particularly useful for manifolds: a second-countable Hausdorff manifold is paracompact and locally metrizable (locally homeomorphic to ), hence metrizable.