3
$\begingroup$

I would like to ask the following problem.

Let $\Omega$ be a $C^{r+1,\alpha}$ domain, $r\in \mathbb{N}, 0<\alpha<1.$ We denote $$C^{r,\alpha}_{0}(\overline{\Omega})=\{f\in C^{r,\alpha}(\overline{\Omega}): f=0 \mbox{ on }\partial \Omega\},$$ here $C^{r,\alpha}(\overline{\Omega})$ is Holder spaces. Is $C^{r+1,\alpha}_0(\overline{\Omega})$ dense in $C^{r,\alpha}_0(\overline{\Omega})?$

We see that when $r\geq 2$, the answer is positive. For any $u\in C^{r,\alpha}_{0}(\overline{\Omega}),$ we have $$\Delta u := f\in C^{r-2,\alpha}_{0}(\overline{\Omega}).$$ Then there exists a sequence $f_n\in C^{r-1,\alpha}_{0}(\overline{\Omega})$ such that $f_n\rightarrow f$ in $C^{r-2,\alpha}_{0}(\overline{\Omega}).$ With each $f_n,$ there exists unique $u_n\in C^{r+1,\alpha}_{0}(\overline{\Omega})$ such that

$$\left\{\begin{array}{ll}\Delta u_n=f_n &\mbox{ in }\Omega\\ u_n =0 &\mbox{ on }\partial \Omega. \end{array}\right.$$ Therefore, by eliptic regularity, we obtain that $||u_n-u||_{C^{2,\alpha}}\leq C||f_n-f||_{C^{r-2,\alpha}}.$ It implies the conclution. It seems that we can not apply the above method for the case $r=1.$

$\endgroup$
2
  • $\begingroup$ So, is the question about the applicability of this method for $r=1$? $\endgroup$
    – Alex M.
    Commented Jul 13, 2018 at 14:29
  • $\begingroup$ No, just, Does the density property hold right for r=1? $\endgroup$ Commented Jul 13, 2018 at 14:39

2 Answers 2

1
$\begingroup$

This is already answered in this post, I think, by a simple sandwich argument.

Smooth $C^\infty_c(\Omega)$ functions are dense in $C^{r+1,\alpha}_0(\Omega)$ and $C^{r,\alpha}_0(\Omega)$ and $C^\infty_c(\Omega)\subset C^{r+1,\alpha}_0(\Omega)\subset C^{r+1,\alpha}_0(\Omega)$, so there.

$\endgroup$
0
$\begingroup$

Too long for a comment: the parametrix of the Dirichlet problem is a pseudo-differential operator with order $-2$ and thus sends $C^s$ into $C^{s+2}$ for any non-integer $s$ where $$ C^s=B^s_{\infty,\infty} \quad\text{is the Besov space}. $$ If $\Omega$ is a smooth open set, your argument should work with the estimate $$ \Vert u-u_n\Vert_{C^{r,\alpha}}\lesssim \Vert f-f_n\Vert_{C^{r-2,\alpha}}. $$ Note on Besov spaces. Using a Littlewood-Paley decomposition, $ 1=\sum_{k\ge 0}\phi_k(\xi), $ where the smooth compactly supported $\phi_k$ have support in $\{2^{k-1}\le\vert \xi\vert\le 2^{k+1}\}$ for $k\ge 1$, we get that, with $s=r+\alpha$, $$C^{r,\alpha}_0(\overline\Omega)=\{u\in L^{\infty}(\mathbb R^d),\ \sup_{k\ge 0} 2^{s k}\Vert \phi_k(D) u\Vert_{L^\infty(\mathbb R^d)}<+\infty,\ \text{supp} u\subset \overline\Omega\}=B^{r+\alpha}_{\infty,\infty}\cap \mathscr E'_{\overline\Omega}.$$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.