0
$\begingroup$

In a previous question, link, I asked about how I could most effectively do a Fourier Transform of a radial function given at certain values and which we knew the asymptotical behaviour of. The Fourier transform reading $$\frac{4\pi}{q}\int^{\infty}_0 dr\, r \sin(qr) f(r).$$ I tried several ways and ended up choosing FFT, which approximates it by calculating the DFT over the interval $[0,\Delta r]$ with $N$ points. (I used more particularly the NAG subroutine C06FAF)

I still get some issue around $q=0$. Indeed, as can be seen on the figure this picture, I have some weird peak at very low frequencies. The black curve that is flat in q=0 is the analytical result while other curves are FFT calculations with increasing $\Delta r$. ($\color{blue}{\Delta r} >\color{red}{\Delta r} > {\bf \color{black}{\Delta r}} > \color{magenta}{\Delta r} $ and the dashed one being the highest $\Delta r$). As can be seen, this peak is narrower when $\Delta r$ gets larger.

The question is from where this peak does come ? And how could I get rid of it ? I already tried to take the mean of the function and substract it to the function but it does not change anything.

One subsidiary question is also the following : the subroutine calculates the integral. I then have to divide by $q$ to get this $4\pi/q$ factor. Though, at $q=0$, this can't be done for obvious reasons. So, what can be done instead ?

EDIT : the problem of the wide peak was simply due to the fact that I was doing a bad conversion between $q$ and the frequency from the routine. As far as the problem of the division by $q=0$ is concerned, I'm happy with Endulum's answer.

$\endgroup$
2
$\begingroup$

I would need additional information about your attempt (e.g., example code) to help with your first question, but here's my input for the division-by-zero question.

First, do you really need to divide by $q$? In the problems where these transforms come up, you can often get away with working with $rf(r)$ and $q\hat{f}(q)$ in their own right. Decide if you can put off dividing by $q$ until later on in the problem.

At such a time that you actually do need $q=0$, you can use the fact that $$\lim_{q\to 0} {\sin{(q r)} \over {q r}} = 1$$ for any finite $r$. Thus you have $$\hat{f}(0)\approx \lim_{q\to 0} 4\pi\int_0^R dr~{\sin{(q r)} \over {q r}} r^2 f(r) = 4\pi\int_0^R dr~r^2 f(r)$$ assuming $r^2 f(r)$ is well-behaved on $[0, R]$.

$\endgroup$
  • $\begingroup$ I found my problem, see EDIT. Thank you for your hint on the division by 0 :) $\endgroup$ – mwoua Sep 9 '16 at 12:13

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.