Taking the limit $$\lim_{p\rightarrow \infty} \left( \frac{\|f\|_\infty}{\|f\|_p}\right)^p$$

First I think the expression after taking the limit will depend on the function $f$.

In my attempt, because it is in the form $``1^\infty"$, I tried to use L'Hopital's rule. And we can calculate the limit (assuming the integrals are defined and finite, I just want to see what the limit might look like). \begin{align*} \lim_{p\rightarrow \infty} \left(\frac{\|\nabla u \|_\infty}{\|\nabla u\|_p}\right)^{p} &= \lim_{p\rightarrow \infty} \exp\left( p\log\left(\frac{\|\nabla u \|_\infty}{\|\nabla u\|_p}\right)\right)\\ &=\lim_{p\rightarrow \infty} \exp\left( \frac{\log\left(\frac{\|\nabla u \|_\infty}{\|\nabla u\|_p}\right)}{\frac{1}{p}}\right)\\ &=\lim_{p\rightarrow \infty} \exp\left( \frac{\frac{d}{dp} \left[-\log\left(\frac{\|\nabla u \|_p}{\|\nabla u\|_\infty}\right)\right]}{\frac{-1}{p^2}}\right)\\ &=\lim_{p\rightarrow \infty} \exp\left( \frac{\left(\frac{\|\nabla u \|_\infty}{\|\nabla u\|_p}\right)\frac{\frac{d}{dp}\left[\|\nabla u\|_p \right]}{\|\nabla u \|_\infty}\}}{\frac{1}{p^2}}\right)\\ \end{align*} where \begin{align*} \frac{d}{dp}\left[\|\nabla u\|_p \right] &= \frac{d}{dp}\left[\left(\int_{\mathbb R^N} |\nabla u |^p dx \right)^{1/p} \right] \\ &=\frac{d}{dp}\left[\exp\left(\frac{1}{p} \log\left(\int_{\mathbb R^N} |\nabla u |^p dx \right)\right)\right] \\ &=\|\nabla u \|_p \left\{\frac{-1}{p^2}\log\left(\int_{\mathbb R^N} |\nabla u |^p dx \right) + \frac{1}{p} \frac{1}{\int_{\mathbb R^N} |\nabla u |^p dx }\int_{\mathbb R^N} |\nabla u|^p \log(|\nabla u |) dx \right\} \end{align*}

Putting it back into the limit we get $$\lim_{p\rightarrow \infty} \exp \left(-\log\left(\int_{\mathbb R^N} |\nabla u |^p dx \right) + p \frac{1}{\int_{\mathbb R^N} |\nabla u |^p dx }\int_{\mathbb R^N} |\nabla u|^p \log(|\nabla u |) dx \right)$$ which simplifies to $$\lim_{p\rightarrow \infty} \frac{\exp \left(p\frac{\int_{\mathbb R^N} |\nabla u|^p \log(|\nabla u |) dx}{\int_{\mathbb R^N} |\nabla u |^p dx } \right)}{\int_{\mathbb R^N} |\nabla u |^p dx }$$

And I am stuck. Is this a correct approach?

Thank you very much!

  • 3
    $\begingroup$ I think your original instinct is right. Consider $L^p[0,1]$. Then the limit for $f(x) = x$ is $\infty$; the limit for $f(x) = 1$ is $1$; and the limit for $f(x) = 1 + x$ is zero. So given that, it seems unlikely to me that there will be a simplier expression for that limit. $\endgroup$ – Simon S Aug 12 '15 at 13:10
  • $\begingroup$ Try using $\lim_{n\to\infty}(n+\frac1n)^n=e$ $\endgroup$ – Rowan Aug 12 '15 at 13:20
  • 1
    $\begingroup$ @user229922 : You mean $(1+1/n)^n$? ;) $\endgroup$ – Patrick Da Silva Aug 12 '15 at 13:23
  • $\begingroup$ @PatrickDaSilva Yes. So sorry $\endgroup$ – Rowan Aug 12 '15 at 13:34
  • $\begingroup$ @SimonS For the case $f(x) = x+1,$ the limit is $\infty.$ $\endgroup$ – zhw. Aug 12 '15 at 19:04

Suppose $f\in L^\infty([0,1])$ and $\|f\|_\infty> 0$. Let $E = \{x:|f(x)|=\|f\|_\infty\}.$ Then

$$\lim_{p\to \infty} \left( \frac{\|f\|_\infty}{\|f\|_p}\right)^p = \frac{1}{m(E)}.$$

(If $m(E)=0,$ the conclusion is that the limit is $\infty.$)

Proof: Let $M= \|f\|_\infty$. Then the expression equals

$$\frac{M^p}{M^p\cdot m(E) + M^p\int_{[0,1]\setminus E}|f/M|^p}.$$

Cancel the $M^p$ terms and then apply the dominated convergence theorem to see the integral in the denominator $\to 0.$ That gives the result.

  • $\begingroup$ Nice answer and a nice result! $\endgroup$ – Simon S Aug 12 '15 at 19:41
  • $\begingroup$ thank you for the answer, but I feel like this result depends on the fact that the domain is $[0,1]$ therefore we always have $\|f\|_p \leq \|f\|_\infty$. $\endgroup$ – Xiao Aug 13 '15 at 7:18
  • $\begingroup$ Well it would work the same on any set of finite measure. In fact, it should work on any measurable set $A$ with one small difference. Let $E$ be as above. If $m(E) = \infty,$ then $\|f\|_p = \infty, 0<p<\infty,$ and all the quotients are $0.$ If $m(E) <\infty,$ then the proof should work as before and the answer should be $1/m(E).$ $\endgroup$ – zhw. Aug 13 '15 at 8:51

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.