Nonexistence for small data

74 Views Asked by At

For context of this question, please refer to page 689 of PDE Evans, 2nd edition. My question is at the bottom.

As an example, consider this initial-value problem in three space dimensions: $$\begin{cases}u_{tt}-\Delta u - |u|^p = 0 & \text{in }\mathbb{R}^3 \times (0,\infty) \\ \qquad \, \, \, \, \,u=g,u_t = h & \text{on }\mathbb{R}^2 \times \{t=0\}. \end{cases}\tag{8}$$ Take any smooth initial data with $$\int_{\mathbb{R}^3} g \, dx > 0, \quad \int_{\mathbb{R}^3} h \, dx > 0 \tag{9}$$ and $\operatorname{spt}g,\operatorname{spt}h \subset B(0,R)$. We will deduce from these conditions alone that there is no solution, provided the exponent $p > 1$ is small enough.

In case $\text{spt}g$ is unfamiliar to some, it means "support", which is also written conventionally as $\operatorname{supp}g$, for example.

THEOREM 2 (Nonexistence for small data). Assume that $$1 < p < 1+\sqrt{2}.$$ Then under the above conditions on $g$ and $h$, the initial-value problem $(8)$ does not have a smooth solution $u$ existing for all times $t \ge 0$.

Here is a fragment of the textbook's proof:

Proof. 1. We assume to the contrary that $u$ is in fact a solution and derive a contradiction. Since the initial data have support within $B(0,R)$, Theorem 3 from §12.1.2 implies that $u(\cdot,t)$ is supported within the ball $B(0,R+t)$.

$\quad$ Put $$I(t) = \int_{\mathbb{R}^3} u \, dx.$$ Then $$I'' = \int_{\mathbb{R}^3} u_{tt} \, dx = \int_{\mathbb{R}^3} |u|^p \, dx.$$

This proof continues, but my question concerns the expression $I''$. I understand the first equality, but how did we obtain the second equality? Namely, why is $$\int_{\mathbb{R}^3} u_{tt} \, dx = \int_{\mathbb{R}^3} |u|^p \, dx$$ when the PDE in $(8)$ gives $u_{tt}=|u|^p+\Delta u$? Because integrating this gives $$I'' = \int_{\mathbb{R}^3} u_{tt} \, dx = \int_{\mathbb{R}^3} |u|^p + \Delta u \, dx.$$ For some reason, however, the textbook "omits" the $\Delta u$ term...

1

There are 1 best solutions below

2
On BEST ANSWER

I think it is just an application of (one of) Green's formulas.

As $u(\cdot, t)$ has support in $B(0,R+t)$, then $u\equiv 0$ outside $B(0,R+t)$; in particular $\nabla u \equiv 0$ on $\partial B(0,r)$ for large enough $r > R + t$.

So \begin{equation} \int_{B(0,r)}\Delta u d x = \int_{\partial B(0,r)}\nabla u \cdot \hat{n}\,d S = 0. \end{equation} Therefore \begin{equation} \int_{\mathbb{R}^3}\Delta u\, dx = \lim_{r \to \infty}\int_{B(0,r)}\Delta u\, d x = \lim_{r \to \infty} \int_{\partial B(0,r)}\nabla u \cdot \hat{n}\,d S = 0. \end{equation}