11

In my analysis lecture I am given a topology on the space of distributions as follows:

Let $u_k$ be a sequence in $\mathcal D'(u)$, $u \in \mathcal D'(u)$. We say $u_k \rightarrow u$, if $\forall \phi \in \mathcal D(u) : u_k(\phi) \rightarrow u(\phi)$.

This is the weak-$*$-topology on $\mathcal D'(u)$. It seems lecturers don't care too much about the topology of $\mathcal D'(u)$, hence I wonder whether there are stronger topologies on $\mathcal D'(u)$.

glS
  • 7,963
shuhalo
  • 8,084
  • 2
    For what its worth, I faintly remember from a class about topological vector spaces that the spaces of distributions are nuclear spaces, therefore they cannot be normed spaces (all normed nuclear spaces are finite dimensional). – Tim van Beek Mar 19 '11 at 09:48
  • 1
    I don't think that is technically the weak-$*$-topology, since $\mathcal{D}(U)$ isn't necessarily the dual space of $\mathcal{D}'(U)$. I think that only holds on bounded domains. – Brian Mar 19 '11 at 11:59
  • This is the most common topology on the space of distributions. Schwartz doesn't mention any other one. –  Mar 19 '11 at 13:52
  • Why do you want to have a stronger topology on $\mathcal D(U)$? – Dirk Mar 19 '11 at 14:49
  • 1
    @JonasTeuwen It seens Schwartz mentions two topologies on the space of distributions: the "strong" (p. 71) which is the main topology in the book, and the weak* star (p. 72) which you refer to and plays a secondary role in the book. – Pedro Jul 10 '17 at 12:37

2 Answers2

6

Certainly there exist stronger topologies on distributions, but as a practical matter the weak-* definition is the one that is interesting and I assume that was the direction of your question. There isn't the usual norm topology available on $\mathcal D(U)$, and per Tim's comment do not have a different norm topology either.

$\mathcal D(U)$ is a pretty strict space to be in and to converge in, so it isn't very demanding to be a distribution. The hard work is all put on the test functions, so to speak. Although there is a certain amount of interesting things you can do with distributions, practically distributions are a stepping stone for getting to more interesting spaces, such as using their differentiability properties to define Sobolev spaces.

Jason Knapp
  • 1,689
6

Some remarks:

$D(U)$ is reflexive, even a Montel space, so the dual of $D'(U)$ with weak or strong topology is again $D(U)$ [This is in contrast to Brian's remark].

A linear functional on $D(U)$ is continuous (i.e. a distribution) if and only if it is sequentially continuous. This is remarkable, as the space of test functions $D(U)$ is not metrizable, so sequential continuity is usually not sufficient.

A sequence of distributions is weakly convergent if and only if it is strongly convergent [i.e. uniformly on bounded subsets of $D(U)$].

The last remark is why usually only the weak topology is known. And Schwartz proved and mentioned this consequence quite often in his book.

t.b.
  • 80,986
Vobo
  • 3,055
  • Can you give reference for the second paragraph? – timur Dec 06 '11 at 07:05
  • See my answer I gave in this question http://math.stackexchange.com/questions/94897/sequential-and-topological-duals-of-test-function-spaces – Vobo Dec 29 '11 at 16:04
  • Thanks a lot. I also found Treves' book on TVS explains everything and is very readable. – timur Jan 01 '12 at 18:59