3
$\begingroup$

Let $X$ be a noncompact locally compact topological space, let $H$ by a complex Hilbert space and let $\boldsymbol f:X\to H$ be a continuous function vanishing at infinity whose support is equal to $X$.

It can be shown that closed convex hull of $\boldsymbol f(X)$ in $H$ is equal to the set of all the vectors $\int_X\boldsymbol f\,d\mu$ as $\mu$ ranges over the set of all Radon probability measures on $X$ (the proof can be adapted from Bourbaki INT III, § 3, prop. 5). So the linear subspace of $H$ spanned by the closed convex balanced hull of $\boldsymbol f(X)$ in $H$ is equal to the set of all the vectors $\int_X\boldsymbol f\,d\mu$ as $\mu$ ranges over the set of all bounded complex Radon measures on $X$.

For Baire reasons, if $\boldsymbol f(X)$ spans an infinite-dimensional linear subspace of $H$, then the closed linear subspace of $H$ it spans is strictly larger than the linear subspace of $H$ spanned by its closed convex balanced hull. Hence, I would like to know if the entirety of the closed linear span of $\boldsymbol f(X)$ in $H$ can be obtained by using unbounded complex Radon measures.

More precisely, my question is:

If $\xi$ is a vector in the closed linear space of $\boldsymbol f(X)$ in $H$, does there necessarily exist a complex Radon measure $\mu$ on $X$ such that $\boldsymbol f$ is Pettis-$\mu$-integrable and such that $\xi=\int_X\boldsymbol f\,d\mu$ ?

$\endgroup$

1 Answer 1

3
$\begingroup$

Let $X=\{1/n:n\in\mathbb{N}\}\cup\{0\}$ with the obvious topology. As $X$ is compact, any Radon measure is finite. Let $(e_n)_{n\in\mathbb N}$ be a ONB of a Hilbert space $H$. Define $f:X\to H$ by $f(n)=\frac1{n^2}e_n$, $f(0)=0$ and consider the vector $v=\sum_{n=1}^\infty \frac1ne_n$. If $v=\int_Xf(x)\ d\mu(x)$, then $\mu(n)=n$ which is not a Radon measure.

$\endgroup$
3
  • 1
    $\begingroup$ Oops sorry! My question had a mistake: the "For Baire reasons […]" paragraph already provides a negative answer when $X$ is compact. I added $X$ noncompact in the assumptions. $\endgroup$ Commented Sep 20 at 13:36
  • 1
    $\begingroup$ You can easily make this example non-compact by taking a coproduct (disjoint sum) with any non-compact space and setting $f=0$ there. $\endgroup$ Commented Sep 20 at 14:52
  • $\begingroup$ I did not think enough about my question… Of course, if it doesn't work if $X$ is compact, it doesn't work either if $\boldsymbol f$ has compact support. I added $\boldsymbol f$ having full support in the assumptions. Though maybe I should've asked about the concrete situation I'm interested in in the first place instead of general results… $\endgroup$ Commented Sep 20 at 16:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .