4
$\begingroup$

Consider the heat equation $$ u_t = u_{xx} + f, $$ on the circle, and for a finite time interval. From Duhamel's principle one can deduce that $u\in L^\infty H^2$ if for instance $f\in L^\infty H^s$ for some $s>0$, and if the initial condition is smooth. I am wondering if one can get this result by energy method. A naive application gives estimates of the form $$ \frac{\mathrm{d}}{\mathrm{d}t}\|\partial^2 u\|^2 + \|\partial^3u\|^2 \lesssim |\langle \partial^2f,\partial^2u\rangle|\leq \|\partial f\|\|\partial^3 u\|, $$ where the norms are the $L^2$-norms, and $\partial$ denotes a generic derivative. This seems to require that $f\in H^1$ rather than $f\in H^s$ with some $s>0$ (which can be smaller than $1$). Is there a trick to get around this in the energy method?

$\endgroup$
2
  • 2
    $\begingroup$ I think the energy method is proving something slightly different, namely that $F \in L^2_t H^1_x$ implies $u \in L^\infty_t H^2_x$. Note that the energy method can certainly get the right fixed-time $H^s_x \to H^2_x$ bounds for the homogeneous heat propagator $e^{t\Delta}$, at which point Duhamel finishes the job; but it appears to be crucial that the energy method is applied to a homogeneous component of the solution rather than the solution itself. $\endgroup$
    – Terry Tao
    Jul 31, 2012 at 6:26
  • $\begingroup$ @Terry: Thanks a lot! It was very helpful. $\endgroup$
    – timur
    Jul 31, 2012 at 6:44

0

Your Answer

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

Browse other questions tagged or ask your own question.