C^\infty versus semiclassical wavefront sets - MathOverflow most recent 30 from http://mathoverflow.net2013-05-23T11:01:49Zhttp://mathoverflow.net/feeds/question/114950http://www.creativecommons.org/licenses/by-nc/2.5/rdfhttp://mathoverflow.net/questions/114950/c-infty-versus-semiclassical-wavefront-setsC^\infty versus semiclassical wavefront setsKelly2012-11-30T01:57:20Z2012-12-03T21:35:08Z
<p>Zworski states that if $u$ is a compactly supported distribution, independent of the semiclassical parameter $h$, then the relationship between the $C^\infty$ and semiclassical wavefront sets of $u$ is given by:</p>
<p>$WF_h(u)=(\mathrm{supp}(u)\times\{0\})\cup WF(u)$.</p>
<p>Would someone please explain this relationship? Thanks!</p>
http://mathoverflow.net/questions/114950/c-infty-versus-semiclassical-wavefront-sets/115008#115008Answer by Jeff for C^\infty versus semiclassical wavefront setsJeff2012-11-30T17:50:26Z2012-12-03T21:35:08Z<p>Heuristically, this result follows because the semiclassical wavefront set measures oscillations at frequency $\frac{\xi}{h}$. To understand this, observe that since $u$ does not depend on $h$, if $u$ is smooth, it has small high frequency oscillations, and hence, for $\xi\neq 0$, and hence points where $u$ is smooth and $\xi\neq 0$ do not appear in $WF_h(u)$. However, where $u$ is not smooth, $u$ has high frequency oscillations and so these points do appear.</p>
<p>Below you will find a proof of the statement.</p>
<p>To see that $WF_h(u)\cap \mathbb{R}^d\times {0}=$ supp $(u)\times{0}$ observe that
$$\mathcal{F}_h(\chi u)(0)=\langle u, \chi\rangle \neq 0$$
for $\chi$ supported near $x_0\in $ supp $(u)$ and
$$\mathcal{F}_h(\chi u)(0)=0 $$
for $\chi$ supported near $x_0\notin $ supp $(u)$. </p>
<p>To see the second part of $WF_h(u)$, simply use the characterization of $WF(u)$ that for $\xi\neq 0$, $(x_0,\xi_0)\notin WF(u)$ if and only if for $\chi$ supported sufficiently close to $x_0$, all $N>0$ and $\xi$ in a conic neighborhood of $\xi_0$,
$$|\mathcal{F}(\chi u)(\xi)|\leq C_N\langle \xi \rangle ^{-N}.$$</p>
<p>Thus, $(x_0,\xi_0)\notin WF(u)$ if and only if for all $N>0$ and all $\xi$ in a neighborhood of $\xi_0$,
$$|\mathcal{F}_h(\chi u)(\xi)|\leq C_N\langle \xi/h\rangle^{-N}.$$
But, since $\xi_0\neq 0$, this gives $(x_0,\xi_0)\notin WF(u)$ if and only if for all $N>0$ and $\xi$ in a neighborhood of $\xi_0$,
$$|\mathcal{F}_h(\chi u)(\xi)|\leq C_Nh^N$$</p>
<p>and hence if and only if $(x_0,\xi_0)\notin WF_h(u)$. </p>