I'm working on the following assignment:

Define the weighted $L^2$-space $L^2_s$ on $\mathbb{R}^n$ to be the completion of $C_{\text{comp}}^{\infty}(\mathbb{R}^n)$ with the norm $$ \|f\|_{L^2_s}=\left(\int_{\mathbb{R}^n}(1+|x|^2)^s|f(x)|^2d^nx\right)^{1/2}. $$ Show that $H^1\subset L^2_s$ is a compact embedding when $s<0$.

Here "compact embedding" means that every uniformly bounded sequence in $H^1$ has a subsequence which is Cauchy in $L^2_s$. But since $L^2_s$ is just $L^2(\mathbb{R}^n,\mu_s)$ with the measure $d\mu_s=(1+x^2)^sd\mathcal{L}^n$ (here $\mathcal{L}^n$ is Lebesgue measure), $L^2_s$ is complete and this is the same as saying every uniformly bounded sequence in $H^1$ has a convergent subsequence in $L^2_s$.


I've spent the last 2-3 days thinking about this problem, but to no avail. It's driving me insane. Here are a couple of my thoughts/observations:

  • For small enough $n$ and $|s|$ large enough, $L^2_s=L^2(\mathbb{R}^n,\mu_s)$ is a finite measure space since $s<0$.
  • The Fourier transform is a unitary isomorphism from $L^2_s=L^2(\mathbb{R}^n,\mu_s)$ to $H^s$.
  • $\|f\|_{L^2_s}\leq\|f\|_{L^2}\leq\|f\|_{H^1}$, so given a uniformly bounded subsequence $\{f_k\}\subset H^1$, $\big\{\|f_k\|_{L^2_s}\big\}$ is uniformly bounded and hence we can find a subsequence $\{f_{k_j}\}$ such that $\big\{\|f_{k_j}\|_{L^2_s}\big\}_{j\in\mathbb{N}}$ converges, but the problem is this doesn't necessarily mean that $\{f_{k_j}\}$ converges in $L^2_s$
  • I wanted to use the Rellich compactness theorem, but this only holds when there exists a compact $K\subset\mathbb{R}^n$ such that $\text{supp }f_k\subset K$ for all $k$. I thought using the theorem on closed balls centered at zero whose radii go to infinity and then diagonalizing, but I don't believe this should work.
  • For simplicity, I was thinking about the case where $n=1$. In this case I have the Sobolev embedding theorem which shows that $H^1(\mathbb{R}^n)\subset C_0(\mathbb{R}^n)$, where $C_0(\mathbb{R}^n)$ is the family of continuous functions on $\mathbb{R}^n$ which vanish at infinity. Morrey's inequality also shows here that the elements of $H^1$ are H$\ddot{\text{o}}$lder continuous, so a uniformly bounded sequence in $H^1$ would be equicontinuous, but I don't think necessarily bounded, so this ruins trying to use Arzela-Ascoli in any significant way.

I feel like, as with most things in my life, I'm making this more complicated than it needs to be. I feel as though a push in the right direction would hugely beneficial. Any help is greatly appreciated. Please only hints as this is an assignment for class.


Break the integral into $\mathbb{R}^{n}\setminus B(0,R)$ where $R$ is very large and make the integral over $\mathbb{R}^{n}\setminus B(0,R)$ as small as you want using the fact that $\frac1{(1+R^{2})^{-s}}\to 0$ as $R\to\infty$ and your sequence is bounded in $L^2$ with no weights. In $B(0,R)$ use Rellich theorem and use the fact that $\frac{1}{(1+|x|^{2})^{-s}}\leq1$.

PS: I can post the full answer once the homework has been handed in. I already typed it before I read the last part of your post (which I really appreciated. Thanks for the honesty).

  • $\begingroup$ Thank you for the help. I believe I've figured it out. I actually feel kind of embarrassed; the problem is a lot easier than I was thinking. I'll award the bounty in 9 hours when it lets me. $\endgroup$ – Blake Mar 26 '17 at 21:27
  • $\begingroup$ you are welcome! And thanks. I liked the problem! $\endgroup$ – Gio67 Mar 27 '17 at 0:42

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.