3
$\begingroup$

Consider a $C^2$ bounded domain $D$ of $\mathbb{R}^d$. Let $b \subset \partial D$ a non-empty part of the boundary. Let $n(x)$ be the unit outward vector on $\partial D$.

Is there any smooth function $f \in C^\infty(\overline{D})$ such that $f=\partial_n f=0$ on $b$.

PS: It was further required the following condition $f=1$ on $\partial D \setminus b$, but it seems to be impossible in this case.

If there is no direct answer, it would be more helpful to mention a reference where the existence of such a function is studied.

$\endgroup$
7
  • $\begingroup$ Such a $b$ must be both open and closed in the boundary, so be a union of components. If the boundary is connected, that means that you ask $f$ to be $0$ or $1$ on all of $\partial D$, so you may just take it to have that same value on all of $\overline D$. $\endgroup$
    – LSpice
    Commented Jun 21, 2020 at 14:53
  • $\begingroup$ @Lspice I don't see why it must be open and closed. I can assume it to be open w.r.t $\partial D$. The same value on whole domain is not what I need. $\endgroup$
    – MathGeo
    Commented Jun 21, 2020 at 15:37
  • 1
    $\begingroup$ According to your requirements, unless I am misreading, $b = f^{-1}(0) \cap \partial D$ and $\partial D \setminus b = f^{-1}(1) \cap \partial D$, both of which are closed in $\partial D$. $\endgroup$
    – LSpice
    Commented Jun 21, 2020 at 15:39
  • $\begingroup$ You're right. I think I should at least modify the second condition. Now, I consider just the first one. $\endgroup$
    – MathGeo
    Commented Jun 21, 2020 at 15:55
  • $\begingroup$ Does $f=0$ satisfy your conditions? $\endgroup$
    – Ben McKay
    Commented Jun 23, 2020 at 6:23

1 Answer 1

3
$\begingroup$

From your assumptions, you have a $C^2$ function $\rho:\mathbb R^d\rightarrow \mathbb R$, such that $$ D=\{x\in \mathbb R^d, \rho(x)<0\}, \quad \partial D=\{x\in \mathbb R^d, \rho(x)=0\}, $$ and $ x\in \partial D\Longrightarrow d\rho(x)\not=0. $ As a result, locally the set $D$ is given by an inequality $x_d<\phi(x')$ and the boundary by the equality $x_d=\phi(x')$ where $\phi$ is a $C^2$ function on $\mathbb R^{d-1}$. Now the function $f$ given by $$ f(x) =\bigl(x_d-\phi(x')\bigr)^2 $$ satisfies your requirements, but is only $C^2$. Going back to the function $\rho$, you can use a theorem of H.Whitney saying that given the closed set $\partial D$, you can find a $C^\infty$ function $f$ positive on the complement of $\partial D$ and vanishing on $\partial D$. Then of course $df$ must vanish on $\partial D$ since $f$ is non-negative.

$\endgroup$
6
  • $\begingroup$ Thank you. Could you give your opinion about the condition I removed, If we can get something close? $\endgroup$
    – MathGeo
    Commented Jun 21, 2020 at 20:20
  • 1
    $\begingroup$ That removed condition would make the smoothness impossible: the most caricatural case would be the situation where $b$ is reduced to a single point of the boundary. $\endgroup$
    – Bazin
    Commented Jun 21, 2020 at 21:58
  • $\begingroup$ Why can your locally defined function be pieced together to a global function? $\endgroup$
    – LSpice
    Commented Jun 22, 2020 at 4:03
  • 1
    $\begingroup$ Partition of unity. $\endgroup$
    – Bazin
    Commented Jun 22, 2020 at 22:44
  • 1
    $\begingroup$ Well, near every point of the boundary, you have a local function $\rho$, so that you can cover the boundary by a finite collection of open sets $\{U_k\}_{1\le k\le N}$ to which you add $U_0=\Omega$. Using a partition of unity (see e.g. Section 1.4 in Hörmander's ALPDO I) you can construct a global $\rho$. $\endgroup$
    – Bazin
    Commented Jul 10, 2020 at 15:38

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .