1
$\begingroup$

I feel like this should be obvious to me but I'm blanking.

Let $\Omega$ be an open bounded subset of $\mathbb R^n$. Let $f_n\in C^\infty_c(\Omega)$ and $f\in L^\infty(\Omega)$ be such that $$ f_n\overset*\rightharpoonup f \text{ in }L^\infty .$$ Now let $\rho\in C^\infty_c(\mathbb R^n)$. Extending all functions by $0$ to functions defined on $\mathbb R^n$, is it true that (up to a subsequence, if you must) $$ \rho*f_n \to \rho *f \text{ in } L^2?$$ I suspect this is indeed true. Although it is not quite my setting, if I use the Fourier basis $e_n(\theta) := e^{in\theta}$ of $\mathbb T = \mathbb R/2\pi \mathbb Z$ which converges weak* to $0$ in $L^\infty$ (which is just Riemann-Lebesgue lemma), then for $\rho\in C^\infty(\mathbb T)$, $$ \rho* e_n (\theta) = \int_{\mathbb T}\rho(\alpha)e^{in(\theta-\alpha)} d\alpha = e^{in\theta} \hat \rho (n) \to 0 \text{ uniformly in $\theta$,} $$ and therefore in $L^2$ (and other $L^p$ too.) But I do not know how to prove a general result. By routine symbol pushing, $\rho*f_n$ converges weakly* in $L^\infty$ and weakly in $L^2$. (Sketches or counterexamples are welcome)

$\endgroup$
4
  • 1
    $\begingroup$ @squid thanks for the bounty; is there something you want answered or improved? $\endgroup$ Nov 22 at 2:16
  • $\begingroup$ "Extending all functions by 0" so that one has $f_n\in C_c^\infty(\mathbb{R}^n)$ and $f\in L^\infty(\mathbb{R}^n)$, correct? I am curious what the motivation is for these two modes of convergence. $\endgroup$
    – user995162
    Nov 22 at 15:17
  • 1
    $\begingroup$ I have just earned back 50 pts to be able to leave a comment:-) I thought this well-written question interesting and hence the bounty. $\endgroup$
    – user995162
    Nov 22 at 15:24
  • $\begingroup$ I can answer that. I was reading a paper (specifically De Lellis and Székelyhidi’s “the Euler equations as a differential inclusion”) and this is part of getting a limit which solves the equation in a distributional sense. PS this is the first time I can guess where the “-10: user was removed” is from :) $\endgroup$ Nov 23 at 1:40
3
$\begingroup$

Let $f=0$. Then we want to prove $$ \int_{\mathbb R^n} \left(\int_\Omega \rho(x-y)f_n(y)dy\right)^2 dx \to0. $$ Since $\Omega$ is bounded, $f_n\rightharpoonup0$ in $L^2(\Omega)$, so $\int_\Omega \rho(x-y)f_n(y)dy \to 0$ for all $x$. That is, the integrand in the integral above converges pointwise to zero. In addition, $$ |\int_\Omega \rho(x-y)f_n(y)dy| \le |\Omega|\|\rho\|_{L^\infty} \sup_n \|f\|_{L^\infty} $$ for all $x$, which is a square-integrable pointwise upper bound. Then convergence follows by dominated convergence.

This is some kind of compactness result for the simple integral operator $f\mapsto \rho *f$ from $L^2(\Omega)$ to $L^2(\Omega)$.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.