Inverse Fourier Tranform of $e^{-|\xi|^{2s}}$ is in $L^p$

108 Views Asked by At

Problem

We wish to show that $ \mathcal{F}^{-1}[ e^{-|\xi|^{2s}}] \in L^{p}(\mathbb{R^2}) $, for all $ p \in [1,\infty], \ s\in(0,1) $.

That is, we wish to show that $ (\int_{\mathbb{R}^2} | \int_{\mathbb{R}^2} e^{2 \pi i x\cdot\xi} e^{-|\xi|^{2s}} \text{d}\xi |^p \text{d}x )^{1/p} < \infty $, for $p \in [1,\infty)$.

My Attempt

My thinking was that we could simply observe that $ e^{-|\xi|^{2s}} \leq 1 $, and then only consider:

$ (\int_{\mathbb{R}^2} | \int_{\mathbb{R}^2} e^{2 \pi i x\cdot\xi} \text{d}\xi |^p \text{d}x )^{1/p} = (\int_{\mathbb{R}^2} | \int_{-\infty}^{\infty} e^{2 \pi i x_1 \xi_1} \text{d}\xi_1 \int_{-\infty}^{\infty} e^{2 \pi i x_2 \xi_2} \text{d}\xi_2 |^p \text{d}x )^{1/p} $

We then use the fact that $ \int_{-\infty}^{\infty} e^{2 \pi i x_1 \xi_1} \text{d}\xi_1 = \frac{1}{2 \pi} \delta(x_1) $.

Now, I believed that, by the usual properties of the Dirac-Delta Distribution, we could say that:

$ \int_{-\infty}^{\infty} \frac{1}{2 \pi} \delta(x_1) \text{d}x_1 = \frac{1}{2\pi} $

and thus $ (\int_{-\infty}^{\infty} |\frac{1}{2 \pi} \delta(x_1)| \text{d}x_1)^{1/p} = \frac{1}{2\pi} $.

However, after some research online, it's come to my attention that this is NOT true, and in fact the dirac delta distribution is not in $L^p$, for any $p$.

As such, I am now lost as to how I can prove the above claim. Any hints/advice appreciated. Thank you.

Update

The case $p = \infty$ is actually quite easy to prove, so I only need help with the case $p \in [1,\infty)$.

1

There are 1 best solutions below

12
On BEST ANSWER

Let $f(\xi)=e^{-|\xi|^{2s}}$, where $0<s<1$. Since $f\in L^q(\mathbb{R}^n)$ for any $q\in[1,\infty]$, we have $$ F=\mathcal{F}^{-1}(f)\in L^p(\mathbb{R}^n) \qquad\textrm{for}\quad p\in[2,\infty], $$ by the Hausdorff-Young inequality.

So far, we have only used the decay of $f$. Now we want to exploit the regularity of $f$, which is related to the decay of $F$. We start by noting that $$ |x|^2F=c\mathcal{F}^{-1}(\Delta f), $$ where $c$ is a constant (involving $i$, $\pi$ etc.). Compute the Laplacian of $f$ as $$ \Delta f(\xi)=e^{-|\xi|^{2s}}(c_1|\xi|^{4s-2}+c_2|\xi|^{2s-2}), $$ where $c_1$ and $c_2$ are again constants. Since $s>0$, it is easy to see that $\Delta f\in L^{1+\varepsilon}(\mathbb{R}^2)$ for some $\varepsilon>0$. This gives us $|x|^2F\in L^q(\mathbb{R}^2)$ for some $q<\infty$ by Hausdorff-Young, and thus $$ F\in L^p(\mathbb{R}^2) \qquad\textrm{for}\quad p\geq1. $$ Let me clarify the last step a bit further. Since $F$ is smooth, it suffices to show that $$ \|F(x)\|_{L^1(Q)}<\infty \qquad\textrm{with}\quad Q=\{|x|>1\}. $$ Thus we compute $$ \int_Q|F(x)|dx=\int_Q|F(x)||x|^2|x|^{-2}dx \leq \||x|^2F\|_{L^q(Q)}\||x|^{-2}\|_{L^{r}(Q)}, $$ where $r$ is the Hölder conjugate of $q$. The crucial point is that $r>1$, and hence $|x|^{-2}\in L^r(Q)$, because $q<\infty$.