Let $1\le p \le \infty$. Prove that for all $\epsilon >0$ there exists a constant $C>0$ such that $$\|u'\|_{L^p(\mathbb{R})}\le \epsilon \|u''\|_{L^p(\mathbb{R})}+C\|u\|_{L^p(\mathbb{R})} \;\; \; \forall u\in W^{2,p}(\mathbb{R}).$$ This is a special case of an interpolation inequality, see here http://conteudo.icmc.usp.br/pessoas/andcarva/sobolew.pdf , theorem 2.

I need help to prove this. In any case the first step is to choose a continuous representative function for u. Then, I have the following 2 ideas:

1) To do it similarly as to prove theorem 2. Apply the mean value theorem to u to receive an estimation for $|u'|$. But the first problem I have here is, that the mean value theorem is local. But here the domain is $\mathbb{R}$. Therefore, i don't know how to proceed.

2) To use the fundamental theorem of anlysis applied to $u'$, to receive something like $$u'(\delta+x)=u'(x)+\int_x^{x+\delta} u''(z)dz.$$ Furthermore, convexity of $|\cdot |^p$ and Jensens inequality https://de.wikipedia.org/wiki/Jensensche_Ungleichung might be important, but I am not completely sure how to math that.

The second idea may be more promising. I appreciant any help.

Edit: There are other variants of this inequality available and meanwhile I have discovered Interpolation inequality and Proving an interpolation inequality for $C^2_b$ functions which seem to be a variant of the inequality above (however, the norms are different).

  • $\begingroup$ Should it not rather be $W^{2,p}$ otherwise the summands don’t Need to exist? $\endgroup$ – Idun E. Jan 16 at 2:59
  • $\begingroup$ thank you. I fixed it $\endgroup$ – TheAppliedTheoreticalOne Jan 16 at 10:32
  • $\begingroup$ The first pdf you link gives a proof (they even prove it first for the one-dimensional case). Is there anything you don't understand? $\endgroup$ – MaoWao Jan 16 at 15:42
  • $\begingroup$ @MaoWao in the meantime I have been succesful proving it with idea 2! I am going to post my solution on MSE as soon as possible $\endgroup$ – TheAppliedTheoreticalOne Jan 16 at 16:15

In fact, I think you only need to prove for $u\in C_c^{\infty}(\mathbb{R})$ which is a dense subset of $W^{1,2}(\mathbb{R})$ (But NOT dense in $W^{1,2}[0,100]$). Then the desired result holds by pushing limit. I hope I remembered it right. :)

  • $\begingroup$ Could you explain how you would push the Limits? Since $C^\infty_c(\mathbb{R})$ is dense, we finde a converving subsequence with regards to $W^{2,k}$. But Since $\mathbb{R}$ is not bounded, this does not imply convergence in $L^p$, if I remeber it correctly. So how does ist work? $\endgroup$ – Idun E. Jan 16 at 11:51
  • $\begingroup$ @IdunE. $C_c^\infty(\mathbb{R})$ is dense in $W^{2,p}(\mathbb{R})$. Since both sides of the inequality are continuous w.r.t. to the $W^{2,p}$-norm, it suffices to prove it on a dense subset. The case $p=\infty$ has to be treated separately, I think (if it's even true at all). $\endgroup$ – MaoWao Jan 16 at 15:44

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.