1
$\begingroup$

Let $G:\mathbb{R}^3\rightarrow\mathbb{R}^3$ be smooth vector field over $\mathbb{R}^3$. For which vector fields $F:\mathbb{R}^3\rightarrow\mathbb{R}^3$ does the PDE

$$\dfrac{\partial}{\partial t}\textbf{v} + \textbf{Curl}(F(\textbf{v})) = \textbf{G}$$ have solutions $\textbf{v}(x,t):\mathbb{R}^3\times (0,\infty) \rightarrow\mathbb{R}^3$ for all time $t>0$, given smooth initial condition $$\textbf{v}(x,0) = \textbf{i}(x)?$$

It's not hard to see F has to be smooth. Are there any other implications we can make? If I give you $F$ can you find $\textbf{v}$ every time? Are the solutions unique?

$\endgroup$
4
  • 1
    $\begingroup$ The way you ask this question is unusual. Analysts prefer to choose a PDE (or a system of PDEs), the lanscape, and then solve it if possible, for every data $(G,i)$. $\endgroup$ Jan 24, 2023 at 7:27
  • 1
    $\begingroup$ Do you want to assume also that $\mathbf{i}(x)$ is smooth? I can imagine that if $F$ is close to the identity, you might want some control on some function space norm of $\mathbf{i}(x)$, but for more general $\mathbf{F}$, I can't imagine how you could get any control on the growth of $\mathbf{v}(x,t)$. $\endgroup$
    – Ben McKay
    Jan 24, 2023 at 13:28
  • 1
    $\begingroup$ @BenMcKay yes the initial condition should be smooth. $\endgroup$
    – MrPie
    Jan 24, 2023 at 16:14
  • 1
    $\begingroup$ @DenisSerre yes of course but I am not doing that sorry that it is unusual. My point here is to try to see the constraint $F$ has on this type of system and see if there is universal technique to solve for arbitrary fields $F$ and $G$. I would prefer that $F$ doesn't really matter, but I cannot make that conclusion myself. $\endgroup$
    – MrPie
    Jan 24, 2023 at 16:16

1 Answer 1

1
$\begingroup$

This is an example of a first order system. A necessary condition for well-posedness of initial value problems is hyperbolicity. But even if F is the identity, the simplest case, this system is not hyperbolic.

$\endgroup$
2
  • 1
    $\begingroup$ @MichealRenardy when $F$ is the identity I believe this problem has been solved. $\endgroup$
    – MrPie
    Jan 24, 2023 at 16:18
  • 1
    $\begingroup$ Take the case where F is the identity and G=0. We have $\partial v/\partial t=curl \,v$, which implies $\partial^2 v/\partial t^2=curl\, curl\, v=-\Delta\, v+\nabla(div\, v)$. If div v is zero initially, this persists, so we have $\partial^2 v/\partial t^2=-\Delta\, v$, which is an elliptic equation, and therefore ill-posed. $\endgroup$ Jan 24, 2023 at 22:00

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.