3
$\begingroup$

Consider the semilinear energy-critical parabolic PDE in $\mathbb{R}^3$ \begin{align} \partial_t u &= \Delta u + |u|^{4/(n-2)}u = \Delta u + u^5\\ u(0,x) &= u_0\in \smash{\dot{H}}^1(\mathbb{R}^n). \end{align}

I am trying to understand the smoothing effects of the above flow. Brezis–Cazenave here proved

Theorem: Assume $q>N(p-1) / 2$ (resp. $q=N(p-1) / 2$ ) and $q \geq 1$ (resp. $q>1), N \geq 1$. Given any $u_0 \in L^q(\Omega)$, there exist a time $T=T\left(u_0\right)>0$ and a unique function $u \in C\left([0,T], L^q(\Omega)\right)$ with $u(0)=u_0$, which is a classical solution of $\partial_t u = \Delta u + |u|^{p-1}u$ on $(0, T) \times\bar{\Omega}$ (in the Duhamel sense).

Moreover, we have:

(i) Smoothing effect and continuous dependence, namely

$$ \|u(t)-v(t)\|_{L^q}+t^{N / 2 q}\|u(t)-v(t)\|_{L^{\infty}} \leq C\left\|u_0-v_0\right\|_{L^q} $$

for all $t \in(0, T]$ where $T=\min \left\{T\left(u_0\right), T\left(v_0\right)\right\}$ and $C$ can be estimated in terms of $\left\|u_0\right\|_{L^a}$ and $\left\|v_0\right\|_{L^q}$.

(ii) $\lim_{t \downarrow 0} t^{N / 2 q}\|u(t)\|_{L^{\infty}}=0$.

(iii) If $u_0 \geq 0$, then $u(t) \geq 0$ for all $t \in\left[0, T\left(u_0\right)\right]$.

Furthermore, for any bounded set (resp. compact set) ${K}$ in $L^q(\Omega)$, there is a (uniform) time $T=T({K})$ such that for any $u_0 \in \mathcal{K}$ the solution of the nonlinear heat equation exists on $[0, T]$.

I would like to understand if one can show further gain in the regularity of solutions to the above PDE. For instance, is it true that the following integral is finite

$$\int_t^T \int_{\mathbb{R}^n} |u|^{2(n+2)/(n-2)} dx dt = \int_t^T \int_{\mathbb{R}^n} |u|^{10} dx dt < +\infty$$

for $0<t<T?$.

Edit: Thanks to the comment below, when $T<+\infty$ this follows from the smoothing effect which can be used to show that $$\int_{t}^{T} |u|^{2p} dx dt \leq C \log(T/t).$$ However, when the flow is global, i.e. $T=+\infty$ I am not sure how to extend the above argument since $1/t$ is not integrable at $+\infty.$

$\endgroup$
7
  • $\begingroup$ Roughly speaking you gain two space and one time derivative when you solve the heat equation. For details, you can look at Krylov's 2008 book on elliptic and parabolic PDEs in Lebesgue and Sobolev spaces. // For fixed $t$ the claimed inequality is certainly true; but I suspect you actually want the bound to be uniform as $t\searrow 0$? $\endgroup$ Commented Oct 4 at 14:20
  • $\begingroup$ @WillieWong Thank you for the comment! In my application, I would be happy to have that estimate hold when $t>t_0>0$ uniformly. In particular I would like $\int_t^T \int_{\mathbb{R}^n} |u|^{2p} \to 0$ as $t\to T.$ $\endgroup$
    – Student
    Commented Oct 4 at 17:04
  • 1
    $\begingroup$ Doesn't Brezis-Cazenave already give that $\|u(t)\|_{L^6}$ is uniformly bounded? Then since $L^{10}$ is between $L^{6}$ and $L^\infty$, just interpolate to get uniform boundedness of $\|u(t)\|_{L^{10}}$ when $t\in [t_0, T)$, from which what you want follow. $\endgroup$ Commented Oct 4 at 18:32
  • $\begingroup$ Thank you for pointing that out! It seems to only work when $T < +\infty.$ In that case one can bound $\int_{t_0}^{T} \int_{\mathbb{R}^n} |u|^{10} \leq C \log(T/t_0)$ since $u$ decays like $t^{-1/(p-1)}=t^{-1/4}$ and so in particular as $t_0 \to T$ this goes to zero. Do you think there is a way to make this work also when $T = + \infty$? My impression was that the heat flow would make the solution more regular at later times $t_0$ which could then give us more decay at later times. $\endgroup$
    – Student
    Commented Oct 4 at 21:50
  • $\begingroup$ Can you include a statement of the Brezis-Cazenave theorem in your question statement? I don't have access to it at the moment. $\endgroup$ Commented Oct 7 at 20:21

0

You must log in to answer this question.