MATH 5345H --- Week 6: Metric topologies, metrizability, and sequences

← Back to Lecture Notes

0.1  (§20) The metric topology

  • Def. (X,𝒯) is metrizable if 𝒯=𝒯d for some metric d.

  • Def. (X,d) bounded if dM throughout; diam(X)=sup{d(x,y)}.

  •  Thm. d¯(x,y)=min{d(x,y),1} is a bounded metric inducing the same topology as d.

    Proof. Triangle inequality by cases. Same topology because ε-balls with ε<1 already form a basis, and there d and d¯ agree. Moral: boundedness is metric data, not topological data.

  • Def. On n:  x=x12++xn2, d(x,y)=yx (Euclidean); x=maxi|xi|, ρ(x,y)=yx (square metric).

  • Thm. d and ρ induce the same topology on n.

  • Def. J={x:J};  ω for J=.

    Note.Neither x nor x is defined for all of ω — hence the need to truncate with d¯.

  • Def. Uniform metric on XJ, (X,d) metric: ρ¯(x,y)=sup{d¯(xα,yα)αJ}. The induced topology is the uniform topology.

  • Thm. On J: uniform topology is finer than the product topology (strictly, for J infinite).

    Proof. Given a basis box B constrained at α1,,αn, take ε=miniεi; then Bρ¯(x,ε)B.

  •  Thm. The product topology on ω is metrizable: D(x,y)=sup{d¯(xn,yn)/n|n} induces it.

    Note.The 1/n damping makes all but finitely many coordinates irrelevant at scale ε — exactly matching “Un for finitely many n”.

0.2  (§21) The metric topology, continued

  • Def. X has a countable basis at x if there are neighborhoods {Bn}n=1 of x such that every neighborhood of x contains some Bn. X is first-countable if this holds at every point. May assume B1B2 (replace Bn by B1Bn).

  • Lem. Every metric space is first-countable  — take Bd(x,1/n).

  •  Lem. (Sequence lemma.) If some sequence in A converges to x, then xA¯. If X is metrizable, the converse holds.

    Proof. Converse: pick xnABd(x,1/n).

  •  Thm. f continuous (xnx gives f(xn)f(x)). If X is metrizable, the converse holds.

    Proof. Converse: show f(A¯)f(A)¯ using the sequence lemma.

  •  Lem. J with J uncountable is not metrizable.

    Proof. Let A={xxα1 for finitely many α} and x=0. Then 0A¯, but no sequence in A converges to 0: the union of countably many finite “support” sets misses some βJ, and πβ1(12,12) separates. So the sequence lemma fails.