3

I am trying to find a function with Fourier coefficients $\frac{1}{ 1 + (2 \pi n)^2}$. How do I go about this problem?

user82261
  • 1,237
  • 1
    I tried using the Fourier Inversion Theorem. I'd get the answer if I can explicitly sum up $\sum_{n \in \mathbb{Z}} \frac{e^{2 \pi i x n}}{ 1 + (2 \pi n)^2}$ but I don't know how to. – user82261 Apr 03 '23 at 15:09
  • 1
    One approach is to look at tables of Fourier series to see if any of them do the trick. Based on that, the Fourier series for hyperbolic cosine may be relevant: https://math.stackexchange.com/q/2370397/137524 – Semiclassical Apr 03 '23 at 15:38

2 Answers2

3

We can explicitly sum up $\displaystyle S(x)=\sum_{n \in \mathbb{Z}} \frac{e^{2 \pi i x n}}{ 1 + (2 \pi n)^2}$, using integration in the complex plane.

First, we note that $S(x)$ is a periodic function with the period $1$, so we can consider only $x\in[0;1]$.

Next, we construct the function $\displaystyle g(z)=\frac{2\pi i}{e^{2\pi iz}-1}\,\frac{e^{2\pi ixz}}{1+(2\pi z)^2}$ and it integrate in the complex plane along the circle with the radius $R\to\infty$.

On the one hand, for $x\in [0;1]$ the integral along the circle $\to0$, as $R\to\infty$.

On the other hand, $$\oint g(z)dz=\frac{i}{2\pi}\oint_{C_R}\frac{1}{e^{2\pi iz}-1}\,\frac{e^{2\pi ixz}}{(z+\frac i{2\pi})(z-\frac i{2\pi})}dz=2\pi i\sum Res \,g(z)$$ $$=2\pi i\left(S(x)+\frac i{2\pi}\frac{e^{- x}}{e^{-1}-1}\frac{2\pi}{2i}-\frac i{2\pi}\frac{e^{ x}}{e^{1}-1}\frac{2\pi}{2i}\right)=0$$ $$\Rightarrow\,S(x)=\frac12\left(\frac{e^{ x-\frac12}}{e^\frac12-e^{-\frac12}}-\frac{e^{- x+\frac12}}{e^{-\frac12}-e^\frac12}\right)=\frac{\cosh(x-\frac12)}{2\sinh\frac12}$$

Svyatoslav
  • 15,657
  • Thanks for the answer. I suspected residue theory may have to be used to sum up the answer.

    I asked about this computation because it arose as part of this question I am trying to solve:https://math.stackexchange.com/questions/4672017/find-element-in-riesz-representation-theorem-in-a-sobolev-space.

    – user82261 Apr 03 '23 at 21:44
  • Do you think what I have on this link is correct? – user82261 Apr 03 '23 at 21:45
  • Yes, it seems correct – Svyatoslav Apr 03 '23 at 22:11
  • Someone else suggested this answer: https://math.stackexchange.com/questions/4672017/find-element-in-riesz-representation-theorem-in-a-sobolev-space/4673203?noredirect=1#comment9879255_4673203. What’s going on? Any ideas? Only one answer can be correct. Is my solution in this post linked above wrong then? I don’t see why though. – user82261 Apr 06 '23 at 09:05
  • Regarding the sum - I made a mistake (an extra-$\pi$ in the nominator and denominator (correcting) And I'm sure that the sum $S(x)$ is correct (for example, the check for $x=0$. But I cannot find a pitfall in your original solution. – Svyatoslav Apr 06 '23 at 12:11
0

given the series $$\sum_{n=-\infty}^{\infty} \frac{e^{2\pi i n x}}{1+(\pi n)^2}$$ we can immediately turn this into the cosine series (since the sine terms cancel each other out):

$$\sum_{n=-\infty}^{\infty} \frac{\cos({2\pi n x})}{1+(\pi n)^2}$$ now since cos is even: $$1+2\sum_{n=1}^{\infty} \frac{\cos({2\pi n x})}{1+(\pi n)^2}$$

now using the fourier series of cosh(ax), we can evaluate this to be: $$\frac{\cosh(2x-1)}{\sinh(1)}$$

for x $\in (0,1)$

in case anyone is reading this, I'm wondering about the sine series: $\sum_{n=1}^{\infty}\frac {sin(2\pi n x)}{(a^2+n^2)}$ and what it evaluates to.

Control
  • 162
  • At $x=0$ we get https://www.wolframalpha.com/input?i=sum_%7Bn+%3D-%5Cinfty%7D+%5E%5Cinfty%5Cfrac%7B1%7D%7B+1+%2B+%282+%5Cpi+n%29%5E2%7D – Svyatoslav May 01 '23 at 12:58
  • 1
    @Svyatoslav oh my god, i was working with the wrong coefficient, it should have been $2 \pi$ and not just $\pi$ – Control May 01 '23 at 13:35