6

The weak* topology on the dual of a separable space is metrizable. On a Hilbert space, the weak topology and the weak* topology coincide, and the dual of the Hilbert space is itself. Thus, on a separable Hilbert space, the weak topology is metrizable.

What's the error in the reasoning here? There must be an error because of the example $\{\sqrt{n} e_n \}$ being a set whose weak closure includes $0$ but no sequence in the set converges weakly to $0$.

keej
  • 1,213
  • $e_n \to 0$ weakly but $\sqrt{n} e_n$ doesn't. Why is $0$ in the weak closure of $\sqrt{n} e_n$ ? – reuns Apr 17 '17 at 22:36
  • 1
    See https://math.stackexchange.com/questions/42337/on-the-weak-closure – keej Apr 17 '17 at 22:42
  • 2
    If $X$ is a separable Banach space the norm-closure of the unit ball of $X^$ is metrizable (and compact) in the weak$^$ topology. But in general the weak$^$ topology on $ X^$ is not metrizable – DanielWainfleet Apr 17 '17 at 23:23
  • @user1952009 A detailed proof: https://math.stackexchange.com/questions/231176/weak-closure-of-sqrt-n-e-nn-in-mathbb-n-and-metrizability-of-weak-topol?noredirect=1&lq=1 More generally, if you have an orthogonal sequence $(x_n)$ s.t. $0\leq |x_n|\leq \sqrt{n}$, for every $n$, then $0$ belongs in the weak closure of ${x_n:n\in \mathbb{N}}$. One can have the same result even if he substitutes $(\sqrt n)$ by a sequence $(a_n)$ satisfying an appropriate condition (by looking carefully at the proof for the $\sqrt{n}$ case, the condition that $(a_n)$ needs to satisfy becomes obvious). – tree detective Apr 17 '17 at 23:39

1 Answers1

4

It is not true that the weak* topology on the dual of a separable Banach space is metrizable. What is true that the weak* topology is metrizable on any bounded subset of the dual. Your set $\{\sqrt{n} e_n\}$ is not bounded, so it does not give a contradiction.

Eric Wofsey
  • 330,363