An easy and quick question:

Consider a function $u\in C(\Omega)$, where $\Omega$ is a bounded domain in $\mathbb{R}^n$.

Define a function $Q$ that measures the distance of a point $(x,y) \in\mathbb{R}^{n+1}$ from the graph of $u$. That is, consider the function $$Q(x,y)=\inf_{z\in \mathrm{graph}(u)} d_z (x,y)$$

where $d_z(\zeta)=|\zeta - z|$.

My question is:

  • why the function $Q$ should be the limit of smooth approximations $Q_p \geq Q$ given by $$Q_p(x,y)=\bigg[\int_\Omega \{ d_{(\xi,u(\xi))}(x,y)\}^{-p} \mathrm{d}\xi \bigg]^{-1/p}\quad \textrm{for } (x,y) \notin \mathrm{graph}(u)$$
  • why the derivative is $$ DQ_p(x,y)=(Q_p(x,y) )^{p+1} \int_\Omega |(x-\xi, y-u(\xi)|^{-p-2}(x-\xi,y-u(\xi))\,\mathrm{d}\xi$$

I just tried to show the uniformly convergence. Can somebody check this following calculation?

Let $d_\xi (x,y) = d_{(\xi,u(\xi))}(x,y)\$. One have

$$d_\xi (x,y) \leq \sup_{\xi} d_\xi (x,y) \rightarrow \dfrac{1}{d_\xi (x,y) } \geq \dfrac{1}{\sup_{\xi} d_\xi (x,y)}$$ so applying the integral $$\int_\Omega \bigg( \dfrac{1}{d_\xi (x,y) } \bigg)^p \geq \bigg( \dfrac{1}{\sup_{\xi} d_\xi (x,y)} \bigg) ^p |\Omega|$$ and so we have $$ \bigg[\int_\Omega \{ d_{(\xi,u(\xi))}(x,y)\}^{p} \mathrm{d}\xi \bigg]^{1/p}\geq \dfrac{1}{\sup_{\xi} d_\xi (x,y)} |\Omega|^{1/p}$$ and accordingly $$\dfrac{1}{\bigg[\int_\Omega \{ d_{(\xi,u(\xi))}(x,y)\}^{p} \mathrm{d}\xi \bigg]^{1/p}} \leq \dfrac{1}{\bigg( \dfrac{1}{\sup_{\xi} d_\xi (x,y)} \bigg) |\Omega|^{1/p}} $$ Now to prove the uniformly convergence, one need $$\lim_{p\rightarrow \infty } \sup_{(x,y)} | Q_p(x,y) - Q(x,y)| \rightarrow 0$$ By the calculation above, we have $$\lim_{p\rightarrow \infty } \sup_{(x,y)} | Q_p(x,y) - Q(x,y)|\leq \lim_{p\rightarrow \infty } \sup_{(x,y)} \bigg| \dfrac{1}{\bigg( \dfrac{1}{\sup_{\xi} d_\xi (x,y)} \bigg) |\Omega|^{1/p}} - \inf_\xi d_\xi(x,y)\bigg|$$ and using the fact that, in general, $ \dfrac{1}{\sup_x \dfrac{1}{f(x)}} \geq \inf_x f(x)$ follows $$\lim_{p\rightarrow \infty } \sup_{(x,y)} | Q_p(x,y) - Q(x,y)|\leq \lim_{p\rightarrow \infty } \sup_{(x,y)} \bigg| \big(\dfrac{1}{|\Omega|^{1/p}} -1\big) \dfrac{1}{\sup_{\xi} d_\xi (x,y)}\bigg|$$

Now one can put $\big(\dfrac{1}{|\Omega|^{1/p}} -1\big)$ out of the supremum and calculete the limit for $p\rightarrow \infty$ the result is $0$.

Is this right?


The second identity appears to follow simply from the chain rule applied to $Q_p(\eta) = f_p(g_p(\eta))$, where \begin{align*} f_p(t) &=t^{-1/p},\\ g_p(x,y) &= \int_\Omega \left(d_{(\xi,u(\xi))}(x,y)\right)^{-p}\,\mathrm{d}\xi,\\ d_{(\xi,u(\xi))}(x,y) &= |(x-\xi,y-u(\xi))| \end{align*} by definition.

As for the first identity, it looks like $Q_p$ measures the inverse of the distance function $d$ in an $L^p$ norm, and thus the limit $p\to\infty$ should give the supremum norm for this inverse, hence the infimum of the distance itself.

| cite | improve this answer | |
  • $\begingroup$ Thank you @gmvh . I wanted a more formal proof of the first question, for example show that $Q_p(x,y)$ converges uniformly to $ Q(x,y)$...any idea? $\endgroup$ – Jason Mar 28 at 16:00
  • $\begingroup$ While I believe that "can you check this"-type questions are strongly discouraged on MO, your step-by-step derivation of a more formal statement of what I stated informally looks fine to me (but then I don't know what level of rigour you're aiming for). $\endgroup$ – gmvh Apr 4 at 9:29

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.