This question has been post on MO a week ago. I move it here to get more luck.

Let $\Omega\subset \mathbb R^N$ be open bounded, smooth boundary. Let $u_1$, $u_2\in H^{1}(\Omega)$ such that $T[u_1]=T[u_2]=T[\omega]$ where $T$ stands for the trace operator and $\omega\in H^1(\Omega)$ is a fixed function.

Define $$ F(u):=\inf_{v\in\mathcal V}\left\{\int_\Omega |\nabla u|^2v^2dx + \int_\Omega |\nabla v|^2+(1-v)^2dx \right\}, $$ where $\mathcal V:=\{v\in H^1(\Omega),\,0\leq v\leq 1\}$.

Question: does there exist a path $a(t): [0,1]\to H^1(\Omega)$ between $u_1$ and $u_2$ satisfies the following conditions?

  1. $a(0)=u_1$, $a(1)=u_2$
  2. $T[a(t)]=T[\omega]$ for all $t\in (0,1)$
  3. $a(t)$ is continuous in $L^2$ sense, i.e., if $t\to t_0$, then $a(t)\to a(t_0)$ in $L^2$.
  4. $F(a(t))\leq \max\{F(u_1),F(u_2)\}$, for all $t\in (0,1)$.

Any help, hint, or reference would be really welcome!

Update: based on @Jason's answer, I wrote, for arbitrary $v\in\mathcal V$, $$ F(a(t))\leq G(a(t),v)\leq tG(u_1,v)+(1-t)G(u_2,v). $$ Let's denote by $v_1$ and $v_2$ that $F(u_1)=G(u_1,v_1)$ and $F(u_2)=G(u_2,v_2)$, such $v_1$ and $v_2$ exists and unique by the properties of $H^1$ function.

I understand that $v$ on the right hand side is arbitrary so we may replace it with either $v_1$ or $v_2$ and we have $$ F(a(t))\leq tG(u_1,v_1)+(1-t)G(u_2,v_1)=t F(u_1)+(1-t)G(u_2,v_1)\tag 1 $$ or $$ F(a(t))\leq tG(u_1,v_2)+(1-t)F(u_2) $$ But we may can not go further from here. Take, from example, $(1)$. We can not switch $v_1$ by $v_2$ here since if we do, we will change $F(u_1)$ as well.


My first instinct at looking at this problem was to take a convex combination of $u_1$ and $u_2$, and this turns out to be the right thing to do. So let


Properties 1,2 and 3 all follow fairly immediately, so it remains to check property 4. First, it will be helpful to write

$$G(u,v)=\int_\Omega\bigg(|\nabla u|^2v^2+|\nabla v|^2+(1-v)^2\bigg)\,\mathrm dx$$

and note that it is sufficient to prove $G(a(t),v)\le\max\{G(u_1,v),G(u_2,v)\}$ for all $v\in\mathcal V$. So fix $v\in\mathcal V$ and observe that $F$ and $G$ are both now fairly irrelevant since we need only prove that the function $t\mapsto|(1-t)\nabla u_1+t\nabla u_2|^2$ does not take a maximum in $(0,1)$. It is fairly straightforward to show that the given map is convex (indeed it is smooth, so look at the second derivative) which implies the result.

A few little extra things to think about:

  • Property 3 is redundant given that $a$ is continuous into $H^1$;
  • Your 'relaxed' version of property 4 is actually equivalent to it;
  • One can show that $F(u)$ is attained as a minimum. This isn't particularly useful to this problem, but perhaps may be helpful in the wider context of where this question was found.
  • $\begingroup$ Thx for your answer! However, if you click the link I provided, you will see an already tried to use convex combination, but it does not work, or at least we can't prove it works... $\endgroup$
    – spatially
    Dec 18 '15 at 7:16
  • $\begingroup$ But I did prove it works! There's an answer there that provides a proof, but you don't even need all of the details of that proof. Are there any steps of my proof you don't agree with? $\endgroup$
    – Jason
    Dec 18 '15 at 15:31
  • $\begingroup$ I wrote some details according to your answer, as well as my concern and problem. Can you have a look? Thank you! $\endgroup$
    – spatially
    Dec 18 '15 at 18:26

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.