3
$\begingroup$

Is there a way to show in a relatively simple manner that for a Lipschitz bounded, connected, open domain $\Omega\subset\mathbb{R}^N$ for any $f\in L^2(\Omega)$ the solution of the problem:

$$\begin{cases}\dfrac{\partial u}{\partial t}=\Delta u, & (t,x)\in (0,T)\times \Omega\\ \dfrac{\partial u}{\partial\nu}=0, & (t,x)\in (0,T)\times\partial\Omega \\ u(0,x)=f(x), & x\in\Omega\end{cases}$$

is bounded and $\Vert u(t,\cdot)\Vert_{L^{\infty}(\Omega)}\leq c\cdot t^{-N/4}\cdot \Vert f\Vert_{L^2(\Omega)}$ for any $t\in (0,T)$, for some $c$ depending only on $\Omega$?

$\endgroup$
9
  • $\begingroup$ Use the integral representation in terms of the fundamental solution? $\endgroup$ Commented Jan 12 at 20:35
  • 1
    $\begingroup$ In Theorem 6.3 and 6.4 this inequality in Ouhabaz (page 158) this inequality is equivalent with proving that $\Vert u\Vert^2_{\frac{2d}{d-2}}\leq c\cdot a(u,u),\ \forall u\in D(a)$. But this is not true in our case since $a(u,u)=\int_{\Omega}|\nabla u |^2\ dx$ and $D(a)=H^1(\Omega)$. For example $u=$constant do not satisfy such an inequality. $\endgroup$
    – Bogdan
    Commented Jan 13 at 4:40
  • 3
    $\begingroup$ One has to be a bit careful here, there a two (non-equivalent) forms of ultracontractivity. The one holds for the heat semigroup on full space and requires the $L^1$-$L^\infty$ bound from the question for all $t\geq 0$. As noted in Denis' answer, this cannot possible hold for the Neumann Laplacian on bounded domains. The second form of ultracontractivity requires the $L^1$-$L^\infty$ bound from the question only on bounded intervals with a constant $c$ depending on $T$. This form of ultracontractivity is true for the Neumann Laplacian on bounded domains with Lipschitz boundary. $\endgroup$
    – MaoWao
    Commented Jan 13 at 9:53
  • 3
    $\begingroup$ The equivalent characterization in terms of a Sobolev inequality takes the form $\lVert u\rVert_{2d/(d-1)}^2\leq c(a(u,u)+\lVert u \rVert_2^2)$, which is true for bounded domains with Lipschitz boundary. $\endgroup$
    – MaoWao
    Commented Jan 13 at 9:54
  • 1
    $\begingroup$ As explained by @MaoWao one has to apply Ouhabaz results to the form $a+1$ which gives global ultracontractivity to for the scaled semigroup and local for the original one. $\endgroup$ Commented Jan 13 at 12:05

1 Answer 1

4
$\begingroup$

The inequality is obviously false, because if $f\equiv c$ is a constant, then $u\equiv c$ remains constant. Since $\Omega$ is bounded, this $f$ is $L^2(\Omega)$, and yet $\|u(t)\|_\infty=|c|$ does not decay as $t\to+\infty$.

I suppose that you make a confusion with the heat equation in the hole space ($\Omega={\mathbb R}^N$, of course the boundary condition disappears). Then $$u(t)=H_t\star f$$ where the heat kernel is given by $$H_t(x)=\frac1{(2\pi t)^{N/2}}K\left(\frac x{\sqrt t}\right),\qquad K(y):=\exp\frac{-|y|^2}4.$$ From Young inequality (or just Hölder) $$\|u(t)\|_\infty\le\|H_t\|_2\|f\|_2=\frac c{T^{N/4}}\|f\|_2.$$

Turning back to the case of a bounded domain, the general rule is an exponential decay as $t\to+\infty$. For instance, the spectrum of $\Delta$ under the Dirichlet boundary condition $u|_{\partial\Omega}=0$ is real, bounded above by $-\mu_D<0$, and we get $\|u(t)\|_p=O(e^{-\mu_D t})$ for every $p\ge2$. In the case of the Neumann boundary condition, we must subtract the mean of the initial data: $$\|u(t)-\frac1{|\Omega|}\int_\Omega f(x)dx\|_p=O(e^{-\mu_N t}),\qquad\forall p\ge2.$$ Here $\mu_N$ is the least non-zero eigenvalue of $-\Delta$.

$\endgroup$

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.