# Studying non-linear PDEs with manifolds

I'm sorry if this is an inappropriate forum to ask this question on, for I fear it is pretty undergraduate-level one :) I was contemplating on the study of non-linear PDEs. Is it possible to reduce a non-linear PDE on $\mathbb{R}^n$ to a distribution or a 'good' PDE on a smooth manifold? It seems to me like a natural step, but I don't know anything about it yet :(

-
too vague a 'question'. – Denis Serre Sep 22 '10 at 5:23
I agree with Dennis Serre. What do you mean by 'good'? Do you have a simple example or heuristic? – hce Sep 22 '10 at 7:21
I take 'good' to mean able to solve, intepret, or extract qualitative behaviour due to the reduction/transformation. I think the question makes sense and that some vagueness correlates well with how ad hoc the study of PDEs is. Perhaps writing a simple PDE on the sphere and then using stereographic projections to put it on R^2 could help to reverse-engineer an illustrative example. – Q.Q.J. Sep 22 '10 at 7:46
You may want to look up the formalism of exterior differential systems which has a bit of the flavor you are describing. In particular, there one can use the equations of a system of PDEs to formally cut out a submanifold of a jet bundle and then solutions to the PDE are in correspondence with subsets where certain differential forms vanish and others are non-vanishing. See e.g. "Cartan for Beginners" by Ivey and Landsberg. – j.c. Sep 22 '10 at 13:40
Oh, forgot to look at the tag. In that case, as jc pointed out, the thing to look for are exterior differential systems and jets. Another reference is the book "Symmetries and Conservation laws for Differential Equations of Mathematical Physics" from Krasilshchik Vinogradov et. al.. Despite the specialized sounding title it's an introduction to this geometric approach. – Michael Bächtold Sep 22 '10 at 15:25

The question is vague (Denis is right) but it does make some sense. When looking at explicit solutions of nonlinear PDEs one frequently encounters singularities that remind of projections from some higher dimensional manifold (e.g. in Burger's equation). But I think the OP does not have a clear picture. Locally, working 'on' a manifold instead than on $R^n$ is essentially like changing coordinates in the equation. So usually you do not resolve nonlinearities, you just modify the coefficients of the equation.