Knowing that $\omega\Subset\Omega\subset\mathbb{R}^2$ (compactly included) are two open and bounded sets with $C^2$ boundary, is it true that for any function $\phi_0:\overline{\omega}\to\mathbb{R},\ \phi_0\in C^1(\overline{\omega})$ ($\overline{\omega}$ is the closure of $\omega$) we can find an extension $\phi:\Omega\to\mathbb{R}$ with $\phi\in C^1_{c}(\Omega)$ (compactly supported in $\Omega$)?


If this type of result is true than we can obtain simple formulas for perimeter of implicitly defined curves in $\mathbb{R}^2$, putting $\phi_0$ the unit outer normal vector to a regular curve (which is defined in a neighborhood of the boundary of $\omega$). See here: Perimeter continuity of $BV$ sets on any sequence from $W^{1,1}$

What did I do?

I proved by standard methods (using cut-off functions and convolution) that we can obtain a mollifying sequence $\phi_n,\ n\in\mathbb{N}^*$ compactly supported in $\Omega$ that tends to $\phi_0$ in $L^1(\Omega)$, but I cannot prove that we can indeed have an extension.


1 Answer 1


$\newcommand\de\delta\newcommand\Om\Omega\newcommand\om\omega\newcommand\R{\mathbb R}$The answer is yes. Indeed, by Whitney's theorem, there is a function $f\in C^1(\mathbb R^2)$ whose restriction to $\overline\omega$ is $\phi_0$. Now take any open set $\Omega_0$ such that $\omega\Subset\Omega_0\Subset\Omega$ and then any function $g\in C^1(\mathbb R^2)$ such that $g=1$ on $\overline\omega$ and $g=0$ on $\mathbb R^2\setminus\Omega_0$. It remains to let $\phi$ be the restriction of $gf$ to $\Om$. (The conditions that $\om$ and $\Om$ have $C^2$ boundaries and that $\Om$ be bounded are not needed. The condition that $\om$ be bounded follows from $\om\Subset\Om$.)

Detail: The set $\Om_0$ and the function $g$ can be constructed as follows. Since $\om\Subset\Om$, there is some real $\de>0$ such that the closure $\overline{\om_{2\de}}$ of the $(2\de)$-neighborhood $\om_{2\de}$ of $\om$ is contained in $\Om$. Let then $\Om_0:=\om_{2\de}$ and $g:=1_{\om_\de}*\psi_\de$, where $\psi_\de$ is any nonnegative function in $C^1(\R^2)$ supported on the ball of radius $\de$ in $\R^2$ centered at $0$ and such that $\int_{\R^2}\psi_\de=1$.

  • $\begingroup$ Is it true that the Whitney Extension of $\phi_0$ satisfies the inequality $|f|\leq |\phi_0|$ on $\overline{\omega}$? $\endgroup$
    – Bogdan
    Commented Jul 22, 2021 at 15:14
  • $\begingroup$ @Bogdan : Since $f$ and $\phi_0$ are continuous and $f=\phi_0$ on $\omega$, we have $f=\phi_0$ on $\overline\omega$. $\endgroup$ Commented Jul 22, 2021 at 15:42
  • $\begingroup$ Sorry. I mean in $|f|\leq sup_{x\in\overline{\omega}} |\phi_0|$ in $\Omega$. I found an article which says something like this but I do not understand exactly. Here it is: core.ac.uk/download/pdf/82133103.pdf (the proposition at page 326). Is it indeed true that $f$ can be chosen that way? $\endgroup$
    – Bogdan
    Commented Jul 22, 2021 at 17:09
  • 1
    $\begingroup$ @Bogdan : I think this is impossible in general. Consider e.g. $\phi_0(x_1,x_2):=x_1$ for $(x_1,x_2)\in\overline\omega$, where $\omega$ is the open unit disk. Then $\sup_{\overline\omega}|\phi_0|=1$, but $f(x_1,0)>1$ for all $x_1>1$ close enough to $1$. Perhaps, you misunderstood the proposition. $\endgroup$ Commented Jul 22, 2021 at 17:34

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.