I'm learning about distributions and tempered distributions. From what I understand, by "enlarging" the space of test functions $\mathcal{D}$ to the Schwarz space $\mathcal{S}$ and correspondingly "shrinking" the space of distributions $\mathcal{D}'$ (continuous linear functionals on $\mathcal{D}$) to $\mathcal{S}'$, we get a space which is closed under Fourier transforms. It's clear that $\mathcal{D}$ is a linear subspace of $\mathcal{S}$, and $\mathcal{S}'$ is a linear subspace of $\mathcal{D}'$ (correct me if I'm wrong).

Question: Is it true that $\mathcal{D}$ is a topological subspace of $\mathcal{S}$? And is $\mathcal{S}'$ a topological subspace of $\mathcal{D}'$ if both have the weak-$*$ topology? Additionally, are the inclusion maps $\mathcal{D} \hookrightarrow \mathcal{S}$ and $\mathcal{S}' \hookrightarrow \mathcal{D}'$ continuous?

I'm using the definitions of this Wikipedia page which I believe are fairly standard. I'm not well-versed in the theory of topological vector spaces (yet) so I don't quite understand the definitions of the topologies on $\mathcal{D}$ and $\mathcal{S}$ in terms of semi-norms. However, I'm interested in knowing the answer to this.

My work: My guess is that the answer is no since it seems to me that you can find a sequence $(\varphi_k)$ of functions in $\mathcal{D}$ (which are then also in $\mathcal{S}$) which converge e.g. to the $0$ function in the topology of $\mathcal{S}$ but not in the topology of $\mathcal{D}$. This is because in order to converge in $\mathcal{D}$, there must exist a compact set $K$ containing the support of each $\varphi_k$ and the sequence must converge to $0$ on $K$. However, to converge to $0$ in $\mathcal{S}$ we require only that $$\lim_{k \to \infty} \sup \{|x^{\alpha} D^{\beta} \varphi_k| \;/\; x \in \mathbb{R^n} \} = 0 $$ for every $\alpha$, $\beta$. I don't think this implies that there exists a compact set $K$ containing the support of each $\varphi$. For example, letting $$ \gamma(x) = \begin{cases} e^{-(1-|x|^2)^{-1}} & \text{if } |x| < 1\\ 0 & \text{if } |x| \geq 1 \end{cases}$$ and $\varphi_k(x) = e^{-k} \gamma(\frac{x}{k})$, we get a sequence of smooth functions with compact support which converges in $\mathcal{S}$ but not in $\mathcal{D}$ because the support gets increasingly larger. Does this imply that the subspace topology of $\mathcal{D}$ on the set $\mathcal{S}$ is finer than the usual topology of $\mathcal{S}$?

Intuitively, I think that a positive/negative answer to the question for $\mathcal{D}$ and $\mathcal{S}$ should imply the same answer for the dual spaces but I could be wrong.


You have answered almost everything by yourself. Let $\tau$ be the usual topology on $\mathcal{D}$ and $\sigma$ the subspace topology of $\mathcal{S}$ on $C_c^\infty$. You are asking for $\tau=\sigma$.

You constructed a sequence in $\mathcal{D}$ converging with respect to $\sigma$ but not with respect to $\tau$. This shows $\tau\nsubseteq\sigma$.

To see $\sigma\subseteq\tau$, you need to show the continuity of the inclusion map $\iota\colon\mathcal{D}\to\mathcal{S}$. But as $\tau$ is the final topology of the $\mathcal{D}_K$, where $K$ runs through the compact sets, $\iota$ is continuous iff each $\iota_K\colon\mathcal{D}_K\to\mathcal{S}$ is continuous. And this is immediate to see.

For the dual space, take a non-tempered distribution $T$ and $\psi_n$ a function identical $1$ on the ball with radius $n$ with support in the ball with radius $n+1$. Define $S_n=\psi_nT$. Then each $S_n$ is tempered (as it has compact support), $(S_n)_n\to T$ in $\mathcal{D}'$, but as the limit is non-tempered, it cannot converge in the weak-star topology of $\mathcal{S}'$.


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.