4
$\begingroup$

I asked this question on Math StackExchange recently but the only useful comment I got was that this could be a good question for Math Overflow. Here it goes:

Consider the Gaussian $G(x):=e^{-x^2}$ on the real line, and localize it to the region $|x|\sim 2^k$ by multiplying it by an appropriate smooth cut-off. More precisely, take $\phi\in C_0^\infty(\mathbb{R})$ supported in the region $$\{x\in\mathbb{R}: \frac{1}{2}<|x|\leq2\}$$ such that $0\leq\phi\leq 1,$ and let $\phi_k(x):=\phi(2^{-k}x)$. Consider: $$G_k(x):=\phi_k(x)G(x).$$ It is straightforward to check that $\|G_k\|_{L^1}\lesssim 2^ke^{-4^k}$, which tends (very quickly) to $0$ as $k\rightarrow\infty$. Also, using Young's convolution inequality one can easily show that $\|\widehat{G_k}\|_{L^1}\leq \|\phi\|_1\|G\|_1$, but this gives no decay in terms of $k$.

My question is: does $\|\widehat{G_k}\|_{L^1}$ decay as $k\rightarrow\infty$? If so, how fast? Can you prove sharp bounds (in $k$)?

Thank you.

$\endgroup$
3
  • $\begingroup$ Original MSE question: math.stackexchange.com/questions/72097 $\endgroup$
    – Yemon Choi
    Oct 13, 2011 at 21:37
  • 1
    $\begingroup$ Note that much of the phrasing above comes from DJC's answer to that MSE question $\endgroup$
    – Yemon Choi
    Oct 13, 2011 at 21:40
  • $\begingroup$ @Yemon: Yes, I should have mentioned that. Thanks! $\endgroup$
    – user17240
    Oct 13, 2011 at 22:27

2 Answers 2

1
$\begingroup$

If I didn't miss anything, the dominated convergence argument turns out to work well. The following shows that $\widehat{G_k}$ can be dominated by a rapidly decaying function.

Denote $\psi=\widehat{\phi}, \psi_k=\widehat{\phi_k}$, then $\psi_k(y)=2^k\psi(2^ky), \int \psi_k=0$ and $\int |\psi_k|=\int |\psi|$, hence

$$ |\widehat{G_k}(-x)| =|\widehat{\phi_k}\ast G(-x)| =|\int \psi_k(y)G(y+x)dy| =|\int \psi_k(y)[G(y+x)-G(x)]dy|$$ $$=|\int_{|y|\le |x|/2} \psi_k(y)[G(y+x)-G(x)]dy|+|\int_{|y|\ge |x|/2} \psi_k(y)[G(y+x)-G(x)]dy|$$ $$\le \frac{|x|}{2}\sup_{B(x,|x|/2)}|G'|\cdot \int |\psi|+C\int_{|y|\ge |x|/2}|\psi|dy$$

The last line is a function in $x$ decaying at any rate.

In view of the above argument, the function $\phi$ and $G$ can be replaced by any Schwartz functions where in addition $\phi$ vanishes at $0$, and we will still have $\|\widehat{G_k}\|_1 \rightarrow 0$. An estimate of the rate of decay follows in the specified case.

$\endgroup$
0
$\begingroup$

Since $\phi$ is $C^\infty_c$ we know that $\widehat{\phi} \in \mathcal{S}(\mathbb{R})$, in particular $\widehat{\phi}(\xi) \lesssim (1+|\xi|)^{-100}$. On the other hand $\widehat{\phi_k}(\xi) = 2^k \widehat{\phi}(\xi 2^k)$ which by our previous estimate satisfies $|\widehat{\phi_k}(\xi)| \lesssim 2^k (1+|2^k \xi|)^{-100}$. Now, $\widehat{G_k} = G \ast \widehat{\phi_k}$ so by Cauchy-Schwarz we can estimate the $L^\infty$ norm of $\widehat{G_k}$ by $\lesssim 2^{-90k} \|G\|_{L^2}$. So by the dominated convergence theorem $\|\widehat{G_k}\|_{L^1} \to 0$ as $k \to \infty$.

$\endgroup$
5
  • $\begingroup$ I agree that the $L^\infty$ norm of $\widehat{G_k}$ is small (it is bounded by $\|G_k\|_{L^1}$). But how will this imply bounds on the $L^1$ norm of $\widehat{G_k}$? I don't think dominated convergence will do it... $\endgroup$
    – user17240
    Oct 26, 2011 at 1:04
  • $\begingroup$ it doesn't imply bounds, just decay to 0 $\endgroup$ Oct 26, 2011 at 22:34
  • $\begingroup$ it is a qualitative statement $\endgroup$ Oct 26, 2011 at 23:05
  • $\begingroup$ Well, "decay to 0" is equivalent to some kind of bound. My second sentence can be rephrased as "how will this imply that the $L^1$ norm of $\widehat{G_k}$ tends to $0$ as $k\rightarrow\infty$?". $L^1$ and $L^\infty$ norms are in general not comparable... $\endgroup$
    – user17240
    Oct 27, 2011 at 0:06
  • $\begingroup$ i dont think you can justify the use of dom conv in the last line $\endgroup$ Nov 8, 2011 at 20:18

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.