1
$\begingroup$

It seems I am too fast to ask the question without thinking carefully.

The equation I consider is find $w:\Omega \to R$ satisfying $$\frac{Du}{\sqrt{1+|\nabla u|^2}}=Dw. \tag{*}$$ Where $u:\Omega \to R$ is given and $u=0$ on $\partial \Omega$.

It is obvious this equation could not be solved for general $u$, the central obstacle is the Frobenius condition (integrable condition). But I still wish to get some more information about this equation, in particular my eager could divide into two parts:

  1. Could the $u$ for which $(*)$ is solvable be dense in some suitable space, maybe $C^{2}(\Omega)$?

  2. For the $u$ which make $(*)$ locally solvable, what information of $w$ could we get from $u$?

Background: I am try to get some geometric explanation for certain PDE similar to mean curvature equation.

I will be appreciate to any related answer and remark.

$\endgroup$
1
  • $\begingroup$ You need to be more specific about your second question: right now it is too broad to be answerable. $\endgroup$ Commented Nov 21, 2017 at 14:41

1 Answer 1

4
$\begingroup$

The Frobenius condition which is necessary and sufficient for local integrability is that $Du$ and $D(|\nabla u|^2)$ are parallel. So going back through my answer to this question you see that $u$ must be a function whose gradient descent curves are geodesics.

This also tells you that such functions cannot be dense in $C^2$, answering in the negative your question (1). For example, if you start with a function like $$ f(x,y) = x + y^2 $$ then $Df = (1,2y)$ and $D |\nabla f|^2 = (0, 8y)$ can never be made to be parallel with a small $C^2$ perturbation.

Now, supposing we have a function $u$ such that $D u$ is parallel to $D |\nabla u|^2$. Define $v = |\nabla u|^2$. Then $v$ and $u$ have the same level sets, and thus away from the level sets of $u$ (where $v = 0$) we can write $v = v(u)$ (at least locally).

Letting $F(s)$ be a real valued function such that $F'(s) = \frac{1}{\sqrt{1 + v^2(s)}}$, you see then that the function $F\circ u$ satisfies $$ D(F\circ u) = F'(u) Du = \frac{Du}{\sqrt{1 + v^2(u)}} $$ and so away from the critical points of $u$ this gives you the solution to your equation (*).

$\endgroup$
2
  • $\begingroup$ A more question, is the linear space span by $w$ which is a solution of $(*)$ dense in $C^2(\Omega)$? I am sorry I do not exactly express the meaning of what I want to ask, this is the thing I really want to ask. $\endgroup$
    – Hu xiyu
    Commented Nov 21, 2017 at 16:59
  • 1
    $\begingroup$ I am not sure; on a hunch I would say that it is likely, as I think (haven't checked the details so not 100% certain) that there exists bump function solutions. $\endgroup$ Commented Nov 21, 2017 at 22:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .