7
$\begingroup$

can you help me with this question find $\sin$ Fourier transform of $\frac{\sin{x}}{x}$

$\endgroup$
0
10
$\begingroup$

I write the Fourier transform as

$$\hat{f}(v) = \int_{-\infty}^{\infty} dx \: \frac{\sin{x}}{x} e^{-i 2 \pi v x} $$

Consider, rather, the integral

$$ \frac{1}{i 2} \int_{-\infty}^{\infty} dx \: \frac{e^{i x}-e^{-i x}}{x} e^{-i 2 \pi v x} $$

$$ = \frac{1}{i 2} \int_{-\infty}^{\infty} dx \: \frac{e^{i (1-2 \pi v) x}}{x} - \frac{1}{i 2} \int_{-\infty}^{\infty} dx \: \frac{e^{-i (1+2 \pi v) x}}{x} $$

Consider the following integral corresponding to the first integral:

$$\oint_C dz \: \frac{e^{i (1-2 \pi v) z}}{z} $$

where $C$ is the contour defined in the illustration below:

integration contour

This integral is zero because there are no poles contained within the contour. Write the integral over the various pieces of the contour:

$$\int_{C_R} dz \: \frac{e^{i (1- 2 \pi v)z}}{z} + \int_{C_r} dz \: \frac{e^{i (1- 2 \pi v) z}}{z} + \int_{-R}^{-r} dx \: \frac{e^{i (1- 2 \pi v) x}}{x} + \int_{r}^{R} dx \: \frac{e^{i (1- 2 \pi v) x}}{x} $$

Consider the first part of this integral about $C_R$, the large semicircle of radius $R$:

$$\int_{C_R} dz \: \frac{e^{i (1- 2 \pi v)z}}{z} = i \int_0^{\pi} d \theta e^{i (1-2 \pi v) R (\cos{\theta} + i \sin{\theta})} $$

$$ = i \int_0^{\pi} d \theta e^{i (1-2 \pi v) R \cos{\theta}} e^{-(1- 2 \pi v) R \sin{\theta}} $$

By Jordan's lemma, this integral vanishes as $R \rightarrow \infty$ when $1-2 \pi v > 0$. On the other hand,

$$ \int_{C_r} dz \: \frac{e^{i (1-2 \pi v) z}}{z} = i \int_{\pi}^0 d \phi \: e^{i (1-2 \pi v) r e^{i \phi}} $$

This integral takes the value $-i \pi$ as $r \rightarrow 0$. We may then say that

$$\begin{align} & \int_{-\infty}^{\infty} dx \: \frac{e^{i (1-2 \pi v) x}}{x} = i \pi & 1-2 \pi v > 0\\ \end{align}$$

When $1-2 \pi v < 0$, Jordan's lemma does not apply, and we need to use another contour. A contour for which Jordan's lemma does apply is one flipped about the $\Re{z}=x$ axis. By using similar steps as above, it is straightforward to show that

$$\begin{align} & \int_{-\infty}^{\infty} dx \: \frac{e^{i (1-2 \pi v) x}}{x} = -i \pi & 1-2 \pi v < 0\\ \end{align}$$

Using a similar analysis as above, we find that

$$\int_{-\infty}^{\infty} dx \: \frac{e^{-i (1+2 \pi v) x}}{x} = \begin{cases} -i \pi & 1+2 \pi v < 0 \\ i \pi & 1+2 \pi v >0 \\ \end{cases} $$

We may now say that

$$\hat{f}(v) = \int_{-\infty}^{\infty} dx \: \frac{\sin{x}}{x} e^{-i 2 \pi v x} = \begin{cases} \pi & |v| < \frac{1}{2 \pi} \\ 0 & |v| > \frac{1}{2 \pi} \\ \end{cases} $$

$\endgroup$
2
  • $\begingroup$ What about when $\vert v \vert = 1/2\pi$? $\endgroup$
    – Max
    Apr 21 '20 at 3:27
  • $\begingroup$ @Max The value at any specific point can be arbitrarily set, since the Fourier Transform does not distinguish between functions that coincide almost everywhere, see the accepted answer of math.stackexchange.com/questions/2172510/… $\endgroup$ Dec 16 '20 at 9:38
4
$\begingroup$

Note that the sinc function is not in $L^1$; therefore you cannot compute its Fourier transform in the "obvious" way by means of the residue theorem or something like that. But this function is in $L^2$, so it is indeed Fourier related to some other function $f\in L^2$. If we can guess an $f$ such that its inverse Fourier transform is sinc we are done.

Now we know from Fourier series as well as from Fourier transform that sinc turns up when processing a rectangle pulse. Therefore we try $$f(\omega):=\cases{c\quad &$\bigl(|\omega|< \Omega\bigr)$ \cr 0&$\bigl(|\omega|> \Omega\bigr)$ \cr}$$ with parameters $c$ and $\Omega$ yet to be determined. At any rate $${\cal F}^{-1}f(x)=\int_{-\infty}^\infty f(\omega)e^{i\omega x}\ d\omega=c\int_{-\Omega}^\Omega e^{i\omega x}\ d\omega={2c\sin(\Omega x)\over\Omega}\ .$$ Now adjust the parameters $c$ and $\Omega$ to your needs. (Note that your definition of ${\cal F}$, resp. ${\cal F}^{-1}$, might lead to slightly different values.)

$\endgroup$
4
  • $\begingroup$ I don't understand your first sentence. While the function $f(x) = \frac{\sin{x}}{x}$ is not $\in L^1$, it doesn't matter with respect to being able to compute the Fourier transform $\hat{f}(v)$ by contour integration; see my solution. $\endgroup$
    – Ron Gordon
    Jan 19 '13 at 13:04
  • 2
    $\begingroup$ @rigordonna: Your calculation is alright; but you have computed the Fourier transform as a principal value of an improper integral, not as an $L^1$-integral. $\endgroup$ Jan 19 '13 at 16:48
  • $\begingroup$ correct. But what does that have to do with an "obvious" use of the Residue theorem? I understand that you are referring to the fact that the "singularity" lies on the integration contour. But, as you point out, we avoid it using a PV-like avoidance. I could have also included the "pole" within the contour and got the same result. $\endgroup$
    – Ron Gordon
    Jan 19 '13 at 16:54
  • $\begingroup$ @rigordonma: The word "obvious" referred to calculating the Fourier transform in $L^1$. Note that your approach wasn't here when I posted mine. Otherwise I would have posted mine anyway, but avoiding the word "obvious". $\endgroup$ Jan 19 '13 at 18:34
3
$\begingroup$

You've got:

$$F(w) = \int \frac{\sin x}{ x} exp(-iw x) dx$$

Differentiate wrt $w$ to get:

$$F'(w) = -i \int \sin x exp(- iwx)dx $$

You know how to find fourier transform of sine and then you should integrate your result.

$\endgroup$

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.