Understanding Fourier Coefficients in the Fourier Transform of a Function

member 428835
Homework Statement
Let ##f:\mathbb R \to \mathbb C## be a smooth function such that ##|f| < C/(1+|x|^2)## for some ##C##. Put ##\phi(x) = \sum_{n = -\infty}^\infty f(x+n)##. Find Fourier coefficients of ##\phi## on [0,1] in terms of the Fourier transform of ##f##.
Relevant Equations
Fourier transform ##\hat f## of ##f## is ##\hat f(k) = 1/\sqrt{2\pi}\int_\mathbb R f(x) e^{-i k x}##
Hi PF!

Unsure how to begin. Fourier transform of ##f## I've given as an equation. I'm not sure what is meant by Fourier coefficients. Fourier coefficients of what?
 
Physics news on Phys.org
It seems like it is asking you to find the Fourier coefficients of ##\phi(x)##. Fourier coefficients are found by integrating the multiple of a function ##f(x)## and ##sin## or ##cos( \frac {2\pi rt}{T})## with respect to time and where ##r## is the order number of the coefficient. So ##r=4## would be the fourth harmonic.
 
Usually there are integration tables for different functions that help.
 
First, note that you don't have a formula for ##f(x)##, just an inequality that probably helps convergence. So you aren't going to do any integrations. So you have $$\phi(x) = \sum_{k=-\infty}^\infty f(x+k),~x\in [0,1]$$When you are given a function defined on ##[0,1]## and asked to find its Fourier coefficients, I guess you would assume its period is ##1## and extend it periodically. So the complex form of the Fourier coefficients would be$$
c_n =\frac 1 2\int_{-\frac 1 2}^{\frac 1 2} \sum_{k=-\infty}^\infty f(x+k)e^{-i n \pi x}~dx$$Now, I haven't worked with this stuff for years, so you should check my formulas. I am guessing that if you swap the sum and integral and make appropriate variable changes you might be able to make it look like a Fourier transform. I haven't worked it but that's what I would try. Good luck with it.
 
  • Like
Likes member 428835
LCKurtz said:
First, note that you don't have a formula for ##f(x)##, just an inequality that probably helps convergence. So you aren't going to do any integrations. So you have $$\phi(x) = \sum_{k=-\infty}^\infty f(x+k),~x\in [0,1]$$When you are given a function defined on ##[0,1]## and asked to find its Fourier coefficients, I guess you would assume its period is ##1## and extend it periodically. So the complex form of the Fourier coefficients would be$$
c_n =\frac 1 2\int_{-\frac 1 2}^{\frac 1 2} \sum_{k=-\infty}^\infty f(x+k)e^{-i n \pi x}~dx$$Now, I haven't worked with this stuff for years, so you should check my formulas. I am guessing that if you swap the sum and integral and make appropriate variable changes you might be able to make it look like a Fourier transform. I haven't worked it but that's what I would try. Good luck with it.
Thanks for the responses everyone, and especially this insight!
 
So to summarize; since we are given a function ##[0,1]##, the period must be 1. If we extend the function periodically, the complex form of the Fourier coefficients is
$$ c_k = \frac 1 2\int_{-\frac 1 2}^{\frac 1 2} \sum_{n=-\infty}^\infty f(x+n)e^{-i k \pi x}\,dx\\ = \frac 1 2 \sum_{n=-\infty}^\infty \int_{-\frac 1 2}^{\frac 1 2} f(x+n)e^{-i k \pi x}\,dx. $$
Let ##u = x + n##. Then
$$ c_k = \frac 1 2 \sum_{n=-\infty}^\infty \int_{-\frac 1 2 + n}^{\frac 1 2 + n} f(u)e^{-i k \pi (u-n)}\,du\\ = \frac 1 2 \sum_{n=-\infty}^\infty e^{i k \pi n} \hat f(k) : \hat f(k) \equiv \int_{-\frac 1 2 + n}^{\frac 1 2 + n} f(u)e^{-i k \pi u}\,du. $$
Note that ##\hat f(k)## is actually the definition of a Fourier transform of ##f(k)##. But now I'm not really sure what to do. Any ideas?
 
What happens if you write out a few terms of$$
\sum_{n=-\infty}^\infty \int_{-\frac 1 2 +n}^{\frac 1 2 + n}$$and combine the integrals? Starting at ##0## and working both ways.
 
  • Like
Likes member 428835
LCKurtz said:
What happens if you write out a few terms of$$
\sum_{n=-\infty}^\infty \int_{-\frac 1 2 +n}^{\frac 1 2 + n}$$and combine the integrals? Starting at ##0## and working both ways.
The sum looks something like
$$
\frac 1 2\exp\left( -2 i k \pi \right) \int_{-5/2}^{-3/2} f(u)\exp(-i k \pi u)\, du + \frac 1 2\exp\left( -i k \pi \right) \int_{-3/2}^{-1/2} f(u)\exp(-i k \pi u)\, du+\\
\frac 1 2 \int_{-1/2}^{-1/2} f(u)\exp(-i k \pi u)\, du + \frac 1 2\exp\left( i k \pi \right) \int_{1/2}^{3/2} f(u)\exp(-i k \pi u)\, du + \\\frac 1 2\exp\left( i 2 k \pi \right) \int_{3/2}^{5/2} f(u)\exp(-i k \pi u)\, du =\\
\frac 1 2 \int_{-\infty}^\infty f(u) \exp(-i k \pi u)\, du
$$

since ##\exp(i k \pi n) = 1##.
 
So doesn't that pretty much solve your problem? That's a multiple of ##\hat f(k)## isn't it? Do you see why the given bound on ##|f(x)|## helps you to get convergence for ##\hat f##? And you probably better check your advanced calculus book to see whether swapping the sum and integral was legit.
 
  • Like
Likes member 428835
  • #10
Thanks a ton! Yep, everything makes good sense. I appreciate it!
 
Back
Top