2
$\begingroup$

If $\mathcal{X}$ is a smooth cutoff near 0 in $\mathbb{R}^n$, then $M_0 = \mathcal{X}(-\Delta+Id)\mathcal{X}$ is a self-adjoint operator in $L^2(\mathbb{R}^n)$. Because $M_0$ is semi-positive and the spectral theory of self-adjoint operator, we can define $M_0^{n/4+\epsilon}$. I need an inequality $$\|M_0^{n/4+\epsilon}u\|_{L^2(\mathbb{R}^n)}\geq C\epsilon^{-1/2}|u(0)|\ \,\,\,\,\ \ \forall\epsilon>0$$

The difficulty of this inequality is local and fractional. And the power of $\epsilon$ can be gotten by using Fourier transform and suppose $\mathcal{X}\equiv1$. I met this inequality in a paper of P.Lax, he uses this to get a global inequality in manifold.

$\endgroup$
1
  • $\begingroup$ The trace operator is bounded from $H^{\frac{1}{2}+\epsilon}(\Omega)$ to $L^2(\partial \Omega)$, but your point evaluation is not, unless $n=1$, so in general your inequality will likely not hold (I am sure about the case $n=2$). Roughly speaking, $\{0\}$ is "too small" (google for "capacity of a set") to matter. $\endgroup$ Dec 10, 2014 at 9:27

1 Answer 1

1
$\begingroup$

For $\epsilon >0$, $u\in H^{\frac{n}{2}+2\epsilon}$, $N_0=-\Delta+1$, $$ \Vert N_0^{\frac{n}{4}+\epsilon} u\Vert_{L^2}=\Vert u\Vert_{H^{\frac{n}{2}+2\epsilon}}\ge c_{n,\epsilon} \Vert u\Vert_{L^\infty}. $$ This implies for $u\in H^{\frac{n}{2}+2\epsilon}_{loc}$ and $\chi_1$ smooth compactly supported, $$ \Vert N_0^{\frac{n}{4}+\epsilon} \chi_1 u\Vert_{L^2}\ge c_{n,\epsilon} \Vert \chi_1 u\Vert_{L^\infty}. $$ Now if $\chi, \chi_1\in C^\infty_c$, $\chi_1=1$ on the support of $\chi$, we get $\chi_1\chi=\chi$ and $$ \Vert \underbrace{(1-\chi_1)N_0^{\frac{n}{4}+\epsilon} \chi}_{\text{Pseudo with order $-\infty$}} u\Vert_{L^2}+ \Vert \chi_1 N_0^{\frac{n}{4}+\epsilon} \chi_1 \chi u\Vert_{L^2}\ge c_{n,\epsilon} \Vert \chi u\Vert_{L^\infty}. $$ This not exactly what you are asking, but it could give you the information you are looking for.

$\endgroup$

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.