This question follow up from a discussion on this post Prove that there exists a constant $c>0$ such that for $R>0$ and $t\in \Bbb R$ we have, $$\left|\int_{|x|<R} \frac{\sin x}{x}e^{-ixt}dx\right|<c$$

This aim to rpove that the Fourier trnasform of $g: x\mapsto \frac{\sin x}{x}$

Given by $$\widehat{g}(t) = \lim_{R\to \infty}\int_{|x|<R} \frac{\sin x}{x}e^{-ixt}dx$$ is bounded. Please I would like to see a prove with elementary tools. I mean a prove that can ignore the Fourier analysis here.

  • 2
    $\begingroup$ Why would you want to ignore Fourier analysis in a question about a Fourier transform? $\endgroup$ – Daniel Fischer Dec 1 '17 at 16:53
  • $\begingroup$ @DanielFischer from discussion [here][math.stackexchange.com/questions/2545959/…} it seems to be obvious using plancherel. Although I don't see why? $\endgroup$ – Guy Fsone Dec 1 '17 at 16:59
  • 2
    $\begingroup$ Up to constant factors (depending on the normalisation of the Fourier transform), $\frac{\sin x}{x}$ is the Fourier transform of the characteristic function of a bounded interval. By the inversion theorem [for the Fourier transform on $L^2$], the Fourier transform of $\frac{\sin x}{x}$ is [a multiple of] the characteristic function of a bounded interval. $\endgroup$ – Daniel Fischer Dec 1 '17 at 17:05
  • $\begingroup$ @DanielFischer that is really the point. I don't want to use that argument. since we don't how the Fourrier transform on $L^2$ look like. similar can be found here math.stackexchange.com/questions/2535865/… $\endgroup$ – Guy Fsone Dec 1 '17 at 17:08
  • 1
    $\begingroup$ The $L^2$ Fourier transform (i) is an isometry on $L^2$ and (ii) agrees with the $L^1$ Fourier transform on $L^2\cap L^1$. That tells us everything we need to know about what it looks like. $\endgroup$ – David C. Ullrich Dec 1 '17 at 21:17

$$\int_{-R}^R 2\frac{\sin x}{x} e^{-ixt}dx =\int_{-R}^R \int_{-1}^{1}e^{i x u}du e^{-ixt}dx =\int_{-1}^{1}\int_{-R}^R e^{ix(u-t)}dx du = \int_{-1}^1 2\frac{\sin (R(u-t))}{u-t}du=\int_{-1-t}^{1-t} 2\frac{\sin (Rv)}{v}dv=\int_{-R(1+t)}^{R(1-t)} 2\frac{\sin w}{w}dw$$

Thus for $|t| \ne 1$ writing $$\int_{-R(1+t)}^{R(1-t)} (...)= 1_{|t|\le 1}\int_{-R(1+t)}^{R(1-t)}(...)+ 1_{|t|>1}\int_{-R(1+t)}^{R(1-t)}(...)$$ we have, $$ \lim_{R \to \infty} 1_{|t|>1}\int_{-R(1+t)}^{R(1-t)}(...)=0 $$ Since $|t|>1$ implies $t+1$ and $t-1$ are of the same sign.

On the other hand $|t|<1$ we have, $1+t>0$ and $1-t>0$. whence, $$\lim_{R \to \infty} \int_{-R}^R 2\frac{\sin x}{x} e^{-ixt}dx = \lim_{R \to \infty} \int_{-R(1+t)}^{R(1-t)} 2\frac{\sin w}{w}dw \\= 1_{|t|<1}\lim_{R \to \infty} \int_{-R(1+t)}^{R(1-t)} 2\frac{\sin w}{w}dw = 1_{|t|<1}\int_{-\infty}^{\infty} 2\frac{\sin w}{w}dw\\ =2\pi1_{|t|<1}$$

The same method works for showing the Fourier inversion theorem for any $f \in L^1,f' \in L^1$ or $f'$ piecewise $L^1$ as here.

  • $\begingroup$ How do you obtain the first equality? also do you directly consider $t>0?$ $\endgroup$ – Guy Fsone Dec 1 '17 at 17:22
  • $\begingroup$ By integrating $e^{ixu}$ $\endgroup$ – reuns Dec 1 '17 at 17:23
  • $\begingroup$ Then it should be $e^{-ixu}$ $\endgroup$ – Guy Fsone Dec 1 '17 at 17:24
  • $\begingroup$ What did you use in the two last equalities? $\endgroup$ – Guy Fsone Dec 1 '17 at 17:34
  • 1
    $\begingroup$ What do you not understand in David's answer ?.. $\endgroup$ – reuns Dec 1 '17 at 17:40

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.