2
$\begingroup$

Let $\Omega\subset \Bbb R^d$ be a bounded $C^1$ domain. Let $u:\Bbb R^d\to \Bbb R$ be a function in $C^2_b(\Bbb R^d)$. I would like to compute the following limit: for $x\in \partial \Omega$

$$L= \lim_{s\to 1}(1-s)\int_{\Omega}\frac{(u(x)-u(y))}{|x-y|^{d+2s}} d y. $$

Here is what I did so far:

Let $r>0$ be arbitrarily small enough. Then as $u$ is bounded, we have

$$\begin{align}&\lim_{s\to 1}(1-s)\int_{\Omega\cap \{|x-y|\geq r\}}\frac{|u(x)-u(y)|}{|x-y|^{d+2s}}dy\\&\leq C \lim_{s\to 1}(1-s)\int_{|x-y|\geq r}\frac{dy}{|x-y|^{d+2s}} \\&= Cc_d\lim_{s\to 1}(1-s) \int_r^\infty t^{-2s-1} dt= 0. \end{align}$$

so that

$$L= \lim_{s\to 1}(1-s)\int_{\Omega \cap B_r(x)}\frac{(u(x)-u(y))}{|x-y|^{d+2s}} d y. $$

Using the fundamental theorem of calculus and the relation $\nabla [|x|^\alpha]= \alpha x|x|^{\alpha-2}$,

$$\begin{align}&(1-s)\int_{\Omega \cap B_r(x)}\frac{(u(x)-u(y))}{|x-y|^{d+2s}} d y\\ & = -\int_0^1 dt (1-s)\int_{\Omega \cap B_r(x)}\frac{\nabla u(x+ t(y-x))\cdot (y-x)}{|x-y|^{d+2s}}dy\\&= \int_0^1 dt\frac{ (1-s)}{d-2(1-s))}\int_{\Omega \cap B_r(x)} \nabla u(x+ t(y-x))\cdot \nabla_y [|x-y|^{-d+2(1-s)}]dy \end{align}$$

Therefore my initial question could be resumed to computaion of

$$\lim_{s\to 1} (1-s)\int_{\Omega \cap B_r(x)} \nabla u(x+ t(y-x))\cdot \nabla_y [|x-y|^{-d+2(1-s)}]dy.$$

Any idea on to move further?

My feeling is the should be a multiple factor of $\frac{\partial u}{\partial n}(x)= \nabla u(x).n(x)$ where $n(x)$ is the normal derivative on $\partial \Omega$ at the point $x$. Don't be mind corrupted, I may have wrong expectation.

$\endgroup$
9
  • 2
    $\begingroup$ 1) In the last integral of the second formula, $t^{ 2s-1}$ is a typo for $t^{{\bf -}2s-1}$, isn't it? 2) If $u(x)$ is affine, say $u(y)=-y_1$ and $x=0$, the integrand in the first formula is $y_1|y|^{-d-2s}$ that is not locally integrable at $x$ for $s>1/2$ or am I wrong? $\endgroup$ Commented Jul 31, 2019 at 13:02
  • $\begingroup$ If you also assume $\nabla u(x)=0$, then I think you get $$L=-{\omega_d\over8}\Delta u(x) $$ where $\omega_d$ is the $d$-volume of the unit ball. $\endgroup$ Commented Jul 31, 2019 at 16:01
  • $\begingroup$ (as a consequence, $L$ should be infinite or undefined whenever $\nabla u(x)\neq0$ ) $\endgroup$ Commented Aug 1, 2019 at 11:58
  • $\begingroup$ Thanks, for reacting. the functons $u(y) =y_1 $ is not bounded as I am only working with such functions. $\endgroup$
    – Guy Fsone
    Commented Aug 1, 2019 at 16:03
  • 1
    $\begingroup$ @GuyFsone it is bounded if $\Omega$ is bounded and locally the integral still diverges. $\endgroup$ Commented Aug 1, 2019 at 16:26

0

You must log in to answer this question.