Consider the minimizing problem of $$\mathcal{F}(u)=\int_0^1 F(x,u(x),u'(x)) \, \mathrm{d}x$$ on $W^{1,4}(I)$ with $I=(0,1)$. Do you have an idea how to show that there exists no minimizer of $\mathcal{F}(u)$ if we chose $$F(x,z,p)=(p^2-1)^2+z^4$$ where $p=u'(x)$ and $z=u(x)$?


Here, the equations are tedious but the idea I present below will hopefully be sufficient to convince you!

Heuristically, there are two contributions of "energy'' to $F$: $((u')^2-1)^2$ and $u^4$. The energy of $F$ is minimized when $u'(x)=\pm 1$ and when $|u(x)|=0$. With these ideas in mind, consider a sequence of functions $u_n(x)$, where $u_n$ is a sawtooth function (alternating piece of $u'(x)=1$ and $u'(x)=-1$) with more and more oscillations. Then, $((u'(x))^2-1)^2 =0$ for all $x$ (where $u'(x)$ is defined).

In addition, note that by taking more and more teeth in your sawtooth, you can make $\int (u_n)^4 dx$ arbitrarily small, since $u_n$ is converging uniformly to $0$. Thus, if a minimizer to $\mathcal{F}$ were to exist, it must satisfy

$$\min \mathcal{F} \leq \lim_{n\to \infty} \mathcal{F}(u_n) =0.$$

Of course, there is no hope for $\mathcal{F}(u)=0$ since for any non-zero function $u$ we have $\int u^4>0$ and for $u\equiv 0$ we have $\mathcal{F}(u)=1.$

  • $\begingroup$ Thank you for the answer! I have 2 questions. Do you mean by $ |u(x)|=0$ that $u(x)= 0$ almost everywhere? And what argument allows me to show that the sawtooth functions belong to $W^{1,4}(I)$? $\endgroup$ – derthomas Jan 31 '17 at 16:10
  • 1
    $\begingroup$ Hi @derthomas. You're welcome! Yes, that is correct: minimization of the functional requires, in particular, that $\int u^4 dx =0$, which happens if and only if $u=0$ almost everywhere. To see that the sawtooth functions belong to $W^{1,4}(I)$ you can use trivial bounds. Using that $|u_n(x)|\leq 1$ and $|\partial_x u_n(x)|=1$ (almost everywhere), we have $\|u_n\|_{W^{1,4}}^4 =\|u_n\|^4_{L^4} + \|\partial_x u_n\|^4_{L^4}\leq 1+1=2$ $\endgroup$ – Matt Jan 31 '17 at 16:27
  • $\begingroup$ That's very helpful! I have one last question: a necessary condition for a minimizer of $\mathcal{F}$ is, as you said, that (1) $u'(x)=^+_-1$ and (2) $u(x)=0$ almost everywhere. Could one argue that (1) and (2) can not be verified at the same time, hence a minimizer can not exist? $\endgroup$ – derthomas Jan 31 '17 at 19:37
  • 1
    $\begingroup$ @derthomas yup! Exactly correct! $\endgroup$ – Matt Jan 31 '17 at 19:41
  • $\begingroup$ Allright, you helped me to get a good view on the problem, thanks! Also, your argumentation with the toothsaw functions is very intuitive, so that helped a lot! It would be nice if you gave me an upvote on my question! ;-) $\endgroup$ – derthomas Jan 31 '17 at 20:17

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.