2
$\begingroup$

Consider two functions $\alpha,\beta: \mathbb{R}^2 \to \mathbb{R}$, where $\alpha$ is given and we look for solutions $\beta$ such that

$$*(d\alpha \wedge d\beta) = \lambda \beta$$

for some $\lambda \in \mathbb{R}$ and where $d$ is the exterior derivative, $*$ is the Hodge star operator, which turns the 2-form $d\alpha \wedge d\beta$ into a 0-form.

Formulated differently, I am looking for eigensolutions of the operator

$$D_{\alpha}(\cdot) := *(d\alpha \wedge d(\cdot))$$

on the space of 0-forms (possibly under some restriction for the domain).

Is this a well posed problem?

If so, what are methods to find solutions in this language of exterior calculus?

Can someone give an example for a simple, but non-trivial $\alpha$?

$\endgroup$
2
  • $\begingroup$ Sorry, since $\alpha$ and $\beta$ map into $\mathbb{R}^3$, then how do you take the exterior derivative? Aren't they $0$-forms on $\mathbb{R}$? Do you want $\alpha,\beta:\mathbb{R}^3 \mapsto \mathbb{R}$? $\endgroup$
    – k3thomps
    May 29, 2014 at 18:49
  • 3
    $\begingroup$ On a domain where $\alpha$ is a constant, this is, of course, the zero operator. On a domain $D$ on which $\alpha:D\to\mathbb{R}$ can be taken to be the first entry of a coordinate chart $(x,y)$ with $\ast(\mathrm{d}x\wedge\mathrm{d}y)=1$ and image a coordinate box (which is locally always possible on neighborhoods of points where $\mathrm{d}\alpha\not=0$), your operator is simply $D_\alpha(\beta) = \beta_y$, and this is always locally solvable. The eigenvalue problem has solutions $\beta = f(x)\ e^{\lambda y}$ where $f$ is arbitrary. $\endgroup$ May 29, 2014 at 19:36

1 Answer 1

2
$\begingroup$

No, the problem as stated is not well posed. You need initial data.

We can rewrite the above equation in coordinates to get $$ \partial_x \alpha \partial_y \beta - \partial_y \alpha \partial_x \beta = \lambda \beta $$ This is a first order PDE and so one can use method of characteristics to solve this. I know this isn't in the language of exterior calculus like you want, but it shows that solutions can be found.

To see that it is not well posed take $\alpha(x,y) = x$ then we have $$\partial_y \beta = \lambda \beta$$ which has the solution $D(x) e^{\lambda y}$ (as pointed out by Robert Bryant in the comments above) where $D(x)$ is any function. This destroys uniqueness for you. You need some sort of initial data to rectify this.

$\endgroup$
2
  • $\begingroup$ Thanks! Is there a coordinate free language of dealing with such operators or would one always go back to local coordinates for their analysis? $\endgroup$
    – madison54
    May 29, 2014 at 20:10
  • $\begingroup$ To be quite honest, I don't know. Personally I don't see what the advantage to doing anything in a coordinate free manner is useful. Aesthetically it might be nice, but I don't see what it buys you. Now, I may just be mathematically immature and just because I don't know why it is useful doesn't mean it isn't. $\endgroup$
    – k3thomps
    May 30, 2014 at 13:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.