MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider the linear (geometric) wave equation in dimension (3+1) with non smooth background metric $g$ say $g \in L^\infty_t H^3_x$ and $\partial_t g \in L^\infty H^2_x$, then energy estimates enable to propagate sobolev regularity of the initial data set till $H^3\times H^2$, is this sharp in the following sense : can you find a metric with this regularity and an initial data set $u_0, u_1$ in say $H^4 \times H^3$ s.t. the corresponding solution of the linear wave equation does not lie even locally in $L^\infty_t H^4_x$ but only in $L^\infty_t H^3_x$ ? Seems to me true but i fail to construct a counterexample. (I would like to apologize for not understanding how to reedit from another computer and a unregistered account, sorry).

share|cite|improve this question
I don't think your scaling argument is correct. The fundamental energy estimate for the linear wave equation $g^{ij}\partial^2_{ij}\phi = 0$ gives that $$ E[\phi](t) - E[\phi](0) \leq \int |\partial g| |\partial \phi|^2 dx dt$$ so the more natural regularity assumption on the metric should be $\partial g\in L^1_tL^\infty_x$. Just $g\in L^\infty_tH^3_x$ does not give you any energy estimate, since in the energy estimate you need to take a time derivative of the metric. If you use the $L^2$ Sobolev scale, you also have to specify the number of spatial dimensions. – Willie Wong Aug 25 '11 at 18:05
up vote 0 down vote accepted

-> question reedited

share|cite|improve this answer
Actually, for spatial dimension $n$, the general energy estimate is $$\|(u, \partial_tu)\|_{H^{k+1}\times H^k} \lesssim E_0 \exp \int_0^t \|\partial g\|_{\infty} + \|g\|_{W^{k,p}} ds $$ where $p = (k-1)/n$ if $n \geq 2(k-1)$ or $p = 2$ otherwise if you use a Moser type product estimate. (BTW, this is what gives the LWP for quasilinear wave equations: if $g = g(u)$ you get LWP in $H^k$ for $k > n/2 + 1$ and if $g = g(u,\partial u)$ you need $n/2 + 2$.) The "covariant" doesn't matter at all since it just adds a term $\partial g \partial u$ to the RHS. – Willie Wong Aug 25 '11 at 19:25
Thanks. Downvote removed. – Deane Yang Aug 25 '11 at 21:04

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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