I recently read the PDE book of L. Evans, and in its chapter 6 some kinds of regularities of second-order elliptic equations were discussed. My question is about its proof of interior smooth regularity (thm 3 in Ch. 6.3.1).

The theorem asserts that if a second-order elliptic PDE $Lu=f$ has smooth coefficients and admits an $H^1$ weak solution $u$ on a bounded domain $U\subseteq \mathbb R^n,$ then $u$ is smooth.

In the statement of the theorem, no condition is assumed about $\partial U,$ but in the book, the author would like to apply the general Sobolev inequality to assure $u\in C^\infty(U)$ if $u\in H^m_{\text{loc}}(U)$ for all $m\in\mathbb N,$ which can be obtained by the theorem 6 in its chapter 5.6.3, but with $\partial U$ is $C^1.$

My problem is whether the regular assumption on the boundary is necessary. However, without the condition $\partial U$ being $C^1,$ I only know some Sobolev inequality about $W^{k,p}_0(U).$ I am not sure if these are sufficient to derive the desired conclusion (since I don't know the behavior of $u$ near the boundary a priori). Perhaps there are alternatives to this problem, or the conclusion is just wrong without the boundary assumption?

Thanks in advance!


I assume that you require $f\in C^\infty(U)$. You do not need regularity of the boundary of $U\subset \mathbb{R}^N$. The condition $u\in H^m_{loc}(U)$ is equivalent with $\widetilde{\phi u}\in H^m(\mathbb{R}^N)$, for any $\phi\in C^\infty_0(U)$. Here $\widetilde{g}$ denotes the extension of the function $g:U\to\mathbb{R}$ by zero outside $U$. Apply the Sobolev embedding theorems to $\widetilde{\phi u}$.

I strongly recommend opening Brezis' book on functional anaylsis and pde's.

  • $\begingroup$ Thanks! You are right that I forgot to say $f$ is smooth. This is the answer I am looking for! Is this a common technique as dealing with such a problem? $\endgroup$ – tommy xu3 May 12 at 11:50
  • $\begingroup$ Yes. For another approach in the general case check Sec 10.3.2. of these notes www3.nd.edu/~lnicolae/Lectures.pdf $\endgroup$ – Liviu Nicolaescu May 12 at 12:12
  • $\begingroup$ This helps me a lot, and thanks for your suggestions about good references! $\endgroup$ – tommy xu3 May 12 at 12:24

If I understand your question correctly, you speak about interior regularity. Let me quote a classical result for linear elliptic equations with $C^\infty$ coefficients, even true for pseudo-differential equations.

Let $P$ be an elliptic differential operator with $C^\infty$ coefficients in an open subset $\Omega$ of $\mathbb R^N$. Then for $u$ a distribution on $\Omega$, $$ Pu\in C^\infty(\Omega)\Longrightarrow u\in C^\infty(\Omega). $$ You may refine that result in the Sobolev scale with $$ Pu\in H^s_{loc}(\Omega)\Longrightarrow u\in H^{s+m}_{loc}(\Omega), $$ where $m$ is the order of $P$. If you like the wave-front-set, you have $$WF(Pu)\subset WF(u)\subset WF(Pu)\cup \text{char}P, $$ and in the elliptic case $\text{char}P=\emptyset$. You can also formulate a result on the $H^s$ wave-front-set.

  • $\begingroup$ Thank you! So this result doesn't require any assumption on the boundary of $\Omega?$ $\endgroup$ – tommy xu3 May 12 at 11:04
  • $\begingroup$ No, you do not need any assumption on the boundary for this result. $\endgroup$ – Bazin May 12 at 16:43
  • $\begingroup$ Thank you. I will also try to see more general theorems like what you mentioned! $\endgroup$ – tommy xu3 May 17 at 4:53

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.