Fourier transform of convolution of functions in Schwartz Space

286 Views Asked by At

I am trying to prove $\widehat{f*g}=\hat f*\hat g$ for $f,g$ in the Schwartz Space, i.e $\sup(|x|^nf^{(k)})<\infty$ for all $n,k$.

$$\widehat{f*g}=\int_{-\infty}^{\infty}\int_{-\infty}^{\infty}f(x-t)g(t)e^{-2\pi ix\xi}dtdx.$$ I am trying to apply the interchange of integrals.

So we have:

\begin{align}\widehat{f*g} &=\int_{-\infty}^{\infty}\int_{-\infty}^{\infty}f(x-t)e^{-2\pi ix\xi}dx g(t)dt \\&=\int_{-\infty}^{\infty}\int_{-\infty}^{\infty}f(x)e^{-2\pi ix\xi}dx e^{-2\pi it\xi}g(t)dt \\&=\hat{f}\hat{g} \end{align}

However, to apply the interchange of integral, I have to make sure $|f(x-t)g(t)e^{-2\pi ix\xi}|=|f(x-t)g(t)|$ is of moderate decrease, i.e $|f(x-t)g(t)|\le\frac{A}{(1+x^2)(1+t^2)}$. I know it is not true that $\exists A>0$ such that $|f(x-t)|\le\frac{A}{1+x^2}$ for all $x,t$. Since here $A$ is independent of $t$, and for each $A$ we can always shift $f$ by $t$ to make some point of $f(x-t)$ strictly larger than $\frac{A}{1+x^2}$. So we cannot prove $f(x-t),g(t)$ separately. We have to estimate them together to get the decay. Any help will be appreciated.

1

There are 1 best solutions below

8
On BEST ANSWER

You can show that bound, but you do not need to in order to show that the product is integrable.

First the easy way. $|f(x-t)g(t)|\le \|f\|_{L^\infty} \|g\|_{L^\infty} \lesssim 1$, and similarly $ |(x-t)^2 f(x-t)g(t)| \lesssim 1.$ Together, $$ |(1+(x-t)^2)f(x-t)g(t)| \lesssim 1$$ and similarly $$ |(1+(x-t)^2)(1+t^2)f(x-t)g(t)| \lesssim 1$$

Now note $$ \int_{\mathbb R^2 } \frac1{(1+(x-t)^2)(1+t^2)}dxdt = \int_{\mathbb R^2 } \frac1{(1+y^2)(1+t^2)}dydt < \infty$$ So $f(x-t)g(t)\in L^1(\mathbb R^2)$ and Fubini's theorem applies.


If you really want to prove the bound: Note that if $|x-t|\ge |x|/2$, then of course $4(1+(x-t)^2) \ge 1+x^2$. So for such $x,t$, $$|(1+x^2)(1+t^2)f(x-t)g(t)| \le 4|(1+(x-t)^2)(1+t^2)f(x-t)g(t)|\lesssim 1.$$ But if instead $|x-t|<|x|/2$ then $$|x|\le |x-t|+|t| < \frac{|x|}2 + |t|$$ so that $$|x| \le 2|t|.$$ This implies $$ |(1+x^2)(1+t^2)f(x-t)g(t)| \le 4 |(1+t^2)^2f(x-t)g(t)| \lesssim 1,$$ showing the moderate decrease.