1
$\begingroup$

Let $U \subset \Bbb R ^n$ be an open subset and let $\mathcal C$ be the space of the smooth functions on $U$ that vanish at infinity, endowed with the seminorms $p_\alpha (f) = \sup \limits _{x \in U} |\partial _\alpha f|$ (or any other equivalent combinations of them), $\alpha$ being a multiindex. Is its dual a well-known space?

I could also make $\mathcal C$ a Banach space by including in the definition the requirement $\sup \limits _\alpha \sup \limits _{x \in U} |\partial _\alpha f| < \infty$. Would this make the dual more convenient to work with?

Or maybe I should just ask: what topology to put on $\mathcal C$ (if any) in order to make it a civilized, well-studied space? I am asking this because its dual would share a number of nice properties with the space of distributions, and might prove to be the correct choice in those contexts where smooth functions with compact support are not an option.

$\endgroup$
3
  • 3
    $\begingroup$ I think a better choice would be the space of functions which not only vanish at infinity themselves but also have all their derivatives vanishing at infinity. Then this becomes a certain subspace of the usual $C^\infty$-space of the one-point-compactification of $U$ (which is a manifold in many interesting cases so it makes sense to talk about its $C^\infty$ space) $\endgroup$ Commented Sep 23, 2015 at 21:29
  • $\begingroup$ @JohannesHahn: Indeed, maybe this is what I should look for. Nevertheless, even though the compatification of $U$ might turn out to be a manifold itself, it will not be a Riemannian one, since the restriction of the Euclidean metric does not vanish at infinity on $U$. $\endgroup$
    – Alex M.
    Commented Sep 24, 2015 at 3:11
  • 1
    $\begingroup$ The dual spaces of the first two suggested spaces are just certain spaces of distributions on $U\,$. What is "convenient to work with", depends on what one aims to work for. This does not become clear enough from the question, and so there is no proper answer. $\endgroup$
    – TaQ
    Commented Sep 24, 2015 at 4:17

0

You must log in to answer this question.