Let $\mathcal{E}'(\Omega)$ be the space of the distributions with compact support. Suppose that $u_n \rightarrow u$ in $\mathcal{E}'(\Omega)$. I'm trying to show that there exist a compact $K\subset \Omega$ such that $\operatorname{supp} u_n, \operatorname{supp} u \subset K$ for all $n \in \mathbb{N}$.

My attempt: Let us show that there exist $n_0 \in \mathbb{N}$ and a compact $K_1$ such that $\operatorname{supp} u_n, \operatorname{supp} u \subset K_1$ for all $n \geq n_0$. Then setting $K_2=\bigcup_{n=1}^{n_0-1} \operatorname{supp} u_n$ we obtain $K=K_1\cup K_2$ which is a compact subset set of $\Omega$ and $\operatorname{supp} u_n, \operatorname{supp} u \subset K$ for all $n \in \mathbb{N}$.

Indeed, since $u_n \rightarrow u$ in $\mathcal{E}'(\Omega)$, then $$\langle u_n, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)} \rightarrow \langle u, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)} \hbox{ for all } \varphi \in \mathcal{E}(\Omega)=C^\infty(\Omega).$$ Let $K_1=\operatorname{supp} u$ and $\varphi \in C_0^{\infty}(\Omega)$ such that $\operatorname{supp} \varphi \subset \Omega \setminus K_1$. Thus, for all $\varepsilon>0$ there exist $n_0(\varphi) \in \mathbb{N}$ such that $$|\langle u_n, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)} -\langle u, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)}|<\varepsilon \hbox{ for all } n \geq n_0.$$ Since $\operatorname{supp} u \cap \operatorname{supp} \varphi =\emptyset$, then $$\langle u, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)}=0.$$

Therefore, for all $\varepsilon>0$ there exist $n_0(\varphi) \in \mathbb{N}$ such that $$|\langle u_n, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)}|<\varepsilon \hbox{ for all } n \geq n_0.$$ From the arbitrariness of $\varepsilon>0$ we obtain $$\langle u_n, \varphi \rangle_{\mathcal{E}'(\Omega), \mathcal{E}(\Omega)}=0 \hbox{ for all } n \geq n_0.$$ Since the function $\varphi$ was taken arbitrarily, we have $U=\Omega \setminus K_1$ is a open subset of $\Omega$ such that $$u_n=0 \hbox{ in } U \hbox{ for all } n \geq n_0.$$

(I'm not sure about this part, because the constant $n_0$ depends on $\varphi$).

Therefore, $$ \operatorname{supp} u_n \subset \Omega \setminus U=K_1 \hbox{ for all } n \geq n_0.$$

Defining $K_2=\bigcup_{n=1}^{n_0-1} \operatorname{supp} u_n$ and $K=K_1\cup K_2$, it's easy to see that $K$ is a compact subset set of $\Omega$ and $\operatorname{supp} u_, \operatorname{supp} u \subset K$ for all $n \in \mathbb{N}$.


Say the $u_n$ are compactly supported distributions $ \in D'(\Bbb{R})$.

If they are not all supported on a common compact, looking at the smallest interval containing $supp(u_n)$ find $ n_j \to \infty,|k_j|>|k_{j-1}| +2$ such that $u_{n_j}$ is zero on $|x| >|k_j|+1$ but not on $(k_j,k_j+1)$,

Take $\psi_j\in C^\infty_c(k_j,k_j+1)$ such that $<u_{n_j},\psi_j>\ne 0$ and look at $$\Psi_0 = 0, \qquad\Psi_j = \Psi_{j-1} + \frac{2^j - <u_{n_j},\Psi_{j-1}>}{<u_{n_j},\psi_j>} \psi_j$$ $$\Psi= \sum_j \frac{2^j - <u_{n_j},\Psi_{j-1}>}{<u_{n_j},\psi_j>} \psi_j \in C^\infty(\Bbb{R})$$ Then $$\lim_{j \to \infty}<u_{n_j},\Psi>=\lim_{j \to \infty} 2^j \ \ne\ <u,\Psi>$$ It works the same way for $\mathcal{E}'(\Omega)$ and arbitrary open $\Omega$.

  • $\begingroup$ I did not understand the construction of the compacts and why the sequence $\Psi_j $ converges in $C^\infty(\mathbb{R})$. Could you give more details? $\endgroup$ Aug 28 '19 at 18:16
  • $\begingroup$ $\lim_{j \to \infty} \Psi_j$ converges because the $\psi_j$ are supported on distinct intervals. $\endgroup$
    – reuns
    Aug 28 '19 at 18:26
  • $\begingroup$ Sorry, the proof is very complicated for me. Could you explain why $\Psi= \sum_j \frac{2^j - <u_{n_j},\Psi_{j-1}>}{<u_{n_j},\psi_j>} \psi_j$ converges? I also don't know how to guarantee that $\langle u_{n_j}, \Psi \rangle= 2^j$, since we have a double limit. $\endgroup$ Aug 28 '19 at 18:46
  • $\begingroup$ The $\psi_j$ are smooth functions supported on disjoint intervals. On an interval $\Psi$ is given by finitely many $\psi_j$. That we obtain $<u_{n,j},\Psi>=2^j$ is because $<u_{n_j},\psi_{j+l}>=0$ for $l \ge 1$ $\endgroup$
    – reuns
    Aug 28 '19 at 18:57
  • $\begingroup$ You know what is $C^\infty_c(\Bbb{R})$ right ? The smooth functions on $\Bbb{R}$ vanishing for $|x|\ge r$. $\endgroup$
    – reuns
    Aug 28 '19 at 19:42

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.