Let us assume $\Omega \subset \mathbb{R}^n$ is as nice as required. Let $f \in L^\infty((0,T)\times \Omega)$ and let $g \in L^\infty((0,T)\times \Omega)$ satisfy $$0 < a \leq g(x,t) \leq b < \infty\quad\text{for all $(x,t)$}$$ $$\frac{dg}{dt} \in L^\infty((0,T)\times \Omega).$$

We know that there exists a function $u \in L^2(0,T;H^1)\cap H^1(0,T;L^2(\Omega))$ satisfying $u(0) = u_0$ given $u_0 \in L^2(\Omega)$ such that $$\int_0^T \int_\Omega u'(t) \varphi(t) + \int_0^T \int_\Omega g(t)\nabla u(t) \nabla \varphi(t) = \int_0^T\int_\Omega f(t)\varphi(t)$$ for all $\varphi \in L^2(0,T;H^1(\Omega))$.

My question is, if we know that $u_0 \in L^\infty(\Omega)$, does it follow that $u' \in L^\infty((0,T) \times \Omega)$? I.e. do we have the expected regularity of $u$? We expect this because $f$ is appropriately regular and as is $u_0$ so we expect the same for $u'$.

If so can anyone refer me to a citation for this result? Thank you. If not, what additional smoothness do I need? I posted this on MSE too but did not receive any answers.

  • $\begingroup$ What can you say in the special case $n=1$, $\Omega=[-1,1]$, and $g$ is a $t$-independent step-function in $x$: $g(t,x)= a$ if $x<0$, and $g(t,x)=b$ if $x\geq 0$? In this case the equation looks like a parabolic problem with distributional coefficients. $\endgroup$ Mar 10, 2014 at 9:45
  • $\begingroup$ Same question with a more complicated $g$: $g(t,x)=a$ for $x\leq t$, $=b$ for $x>t$. In any case, I find your question intriguing. $\endgroup$ Mar 10, 2014 at 9:55
  • $\begingroup$ Good questions. I'm trying to find literature that address bounded coefficients.. $\endgroup$
    – student
    Mar 10, 2014 at 10:15
  • 1
    $\begingroup$ Note that in the first of above examples the coefficient of $\partial_x$ is a distribution, $(b-a)\delta_0(x)$, where $\delta_0(x)$ denotes Dirac's delta at $0$ in the $x$-direction. $\endgroup$ Mar 10, 2014 at 11:18
  • 1
    $\begingroup$ @student in elliptic case there is a difference for regularity between $n=2$ and higher dimensions. In particular solutions of equations in divergence form with measurable coefficients belong to $C^{1+\alpha_0}$ locally. In higher dimensional case to $C^{\alpha_0}$ for some $\alpha_0\in (0,1)$. And two dimensions in elliptic case corresponds to one space variable for parabolic equations. So the answer may differ for $n=1$ and $n\ge2$. $\endgroup$
    – Andrew
    Mar 10, 2014 at 18:05

1 Answer 1


Your variational formulation is for some parabolic equation in divergence form. The kind of regularity your are looking for can be deduced by time differentiating the equation and then applying a Moser-Alikakos-like iteration scheme for $L^p$-powers of the time derivative. Such arguments don't usually carry a reference since every parabolic equation is different and so it's yours. Plus you want an estimate that shows regularity for times $t>0$. On the other hand, there may be other ways to get what you want but it all depends for what purpose you wish to use this regularity result. In any event, I have performed many of such estimates for systems that were somewhat similar to yours, and the details are usually not straight-forward. This is again to emphasize the fact that you don't a priori know what assumptions these procedures will require although they will be close to what your are asking.


Your Answer

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

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