1
$\begingroup$

Let $X$ be a normed space and let $X^\ast$ denote its continuous dual. There is a norm on $X^\ast$ defined by $\|\varphi\|=\sup_{\|x\|=1}|\varphi(x)|$. The weak star topology on $X^\ast$ is defined to be the weakest (=coarsest) topology such that the maps $e_x: \varphi \mapsto \varphi(x)$ are continuous.

The Banach-Alaoglu theorem states that $D=\{\varphi \in X^\ast : \|\varphi\|\le1\}$ is compact in the weak star topology.

I hope everything I wrote so far is correct. I believe the diagram here should apply to $X^\ast$ in that the norm topology on $X^\ast$ is finer than the weak star topology. But if it was so then Banach-Alaoglu would follow immediately since of course $D$ is compact in the norm topology and sets that are compact in a stronger topology are automatically compact in a weaker topology, that is, if $C$ is compact in $T$ and $T'\subseteq T$ then $C$ is also compact in $T'$ where $T,T'$ are topologies.

Therefore the weak star topology cannot be weaker than the norm topology. Is this correct and if so could someone please provide an example of a set that is open in the weak star topology but not open in the norm topology?

$\endgroup$
  • $\begingroup$ If $X$ is infinite dimensional, then $D$ is not compact in the norm topology. (If $X$ is finite dimensional, then the norm and weak-star topologies are the same.) $\endgroup$ – Nate Eldredge Apr 7 '14 at 16:30
2
$\begingroup$

of course $D$ is compact in the norm topology

That is your error. When $X$ is infinite dimensional, $D$ is not compact in the norm topology.

The norm topology on $X^*$ is always at least as fine (strong) as the weak-* topology. When $X$ is finite dimensional, both topologies are the same, and when $X$ is infinite dimensional, the norm topology is strictly finer.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Great, thanks. Now I also see which property fails: it's not totally bounded. $\endgroup$ – Student Apr 7 '14 at 16:55
  • $\begingroup$ If you don't mind -- please could you add an example of an open cover of the closed unit ball that does not admit a finite subcover? $\endgroup$ – Student Apr 8 '14 at 12:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.