Originally asked on MSE.

Let $E$ be a metrizable locally convex topological vector space and let $E^{*}$ be its dual space endowed with the strong topology = topology of uniform convergence on (closed convex balanced) bounded subsets of $E$. Let $F\subset E^{*}$ be a linear subspace.

Is it true that $\overline{F}=F_1$, where $F_1$ is the set of all linear functionals on $E$, whose restrictions on every (closed convex balanced) bounded subset of $E$ are continuous in the weak $\sigma(E,F)$ topology?

It seems that I can adapt the proof of Grothendieck's completion theorem to prove this. Indeed, every linear functional continuous on a closed set $B$ in $E$ can be uniformly approximated on $B$ by an element of $E^{*}$ (I guess this is Grothendieck's lemma). Hence, we only need to show that $F_1\subset E^{*}$. But this follows from the fact, that since $E$ is metrizable, $E^{*}$ is complete, and so it is a closed subset of $E'$ (the algebraic dual), andowed with the uniformity of uniform convergence on bounded sets.

If this is correct, I hope this result is contained in some textbook on locally convex spaces, and so a reference would be highly appreciated.

  • $\begingroup$ Is $F$ intended to be a linear subspace? The definition of $F_1$ implies that it is a linear space. $\endgroup$ – Robert Furber Apr 29 '18 at 0:40
  • $\begingroup$ @RobertFurber Yes, thank you, I'll add that. $\endgroup$ – erz Apr 29 '18 at 6:32

I don't know a reference, but I think you can prove this without using the algebraic dual: To see $F_1\subseteq E^*$ you can use the fact that metrizable spaces are bornological, i.e., a linear map on $E$ is continuous if all restrictions to bounded sets are continuous. This shows that $F_1$ is a closed subspace of $(E^*, \beta(E^*,E))$ which clearly contains $F$.

For the other inclusion, given $f\in F_1$ and a $0$-neighbourhood $U$ of $(E^*, \beta(E^*,E))$, there is a (closed absolutely convex) bounded set $B$ in $E$ with $B^\circ \subseteq U$. Since $f|_B$ is $\sigma(E,F)$-continuous there is a finite set $M\subseteq F$ with $f\in \left(B\cap M^\bullet\right)^\circ$ (where the $\bullet$ indicates the polar in $E$). The theorem of bipolars then gives $f\in B^\circ + M^{\bullet\circ} = B^\circ + \Gamma(M) \subseteq U+F$ (because the absolutely convex hull $\Gamma(M)$ of a finite set is compact and hence closed). This shows $f\in \overline{F}^{\beta(E^*,E)}$.

This proof uses metrizability only to obtain bornologicity. The statement thus holds e.g. for strong duals of distinguished Frechet spaces.

  • $\begingroup$ Thank you for your answer, but I don't see major differences between your proof and mine. In fact, your second paragraph is very similar to the proof of Grothendieck's lemma. I was aware that metrizability was only used to get bornologicity, just thought (perhaps mistakingly), that it is better to state the question this way. $\endgroup$ – erz Apr 29 '18 at 15:46

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.