I know this question has been widely answered here, but without using Fourier analysis. Also there is a video referring to this trick but I want to use a different Fourier series.

First of Parseval's Theorem states: $\displaystyle{\dfrac{1}{\pi}\int_{-\pi}^{\pi}[f(x)]^2 = [a_0]^2+\sum_{n=1}^{\infty}[a_n]^2+[b_n]^2}$.

I calculate $a_0$ like $\frac{1}{2\,\pi}\,\int_{-\pi}^{\pi}f(x)\,\mathrm{dx}$ instead of $\frac{1}{\pi}\,\int_{-\pi}^{\pi}f(x)\,\mathrm{dx}$, so no need for $\frac{1}{2}\,a_0$

Here I'd like to involve the series of $\cosh(x)$ whose partition I have been asked for in a prior question.

It should be: $\displaystyle{\cosh(x) = \underbrace{\dfrac{\sinh(\pi)}{\pi}}_{a_0}+\sum_{n = 1}^{\infty}\underbrace{\color{red}{2}\,\dfrac{1}{\pi}\,\dfrac{\sinh(\pi)}{1+n^2}\,(-1)^n\,\cos(n\,x)}_{a_n}}$

Plugging those terms into the original theorem:

$$\begin{align} &\dfrac{1}{\pi}\int_{-\pi}^{\pi}\cosh^2(\pi) = \left(\dfrac{\sinh(\pi)}{\pi}\right)^2+\sum_{n=1}^{\infty}\left(\color{red}{2}\,\dfrac{1}{\pi}\,\dfrac{\sinh(\pi)}{1+n^2}\,(-1)^n\right)^2 \\\\ & 1+\dfrac{\sinh(2\,\pi)}{2\,\pi}= \dfrac{\sinh^2(\pi)}{\pi^2}\,\left[1+\sum_{n=1}^{\infty}\color{red}{4}\,\left(\dfrac{1}{1+n^2}\right)^2\right]\\\\ &\sqrt{\left(\dfrac{\pi^2}{\sinh^2(\pi)}+\dfrac{\sinh(2\,\pi)\,\pi}{2\,\sinh^2(\pi)}-1\right)\,\color{red}{\dfrac{1}{4}}} = \sum_{n=1}^{\infty}\dfrac{1}{1+n^2} \quad ? \end{align}$$

Actually the value is coming close to the approximated sum, but it's not exactly the same result...


I fixed the Fourier Series highlighting the missing term in red. On the other hand I eradicated some factors, not sure about making it worse. It's still differing from the approximation.

approximation $\approx 1,0767$

Therefore it works exactly like suggested by Stefan Lafon in the remarks: setting $x=\pi$

$\begin{align} &\cosh(\pi) = \dfrac{\sinh(\pi)}{\pi}+\displaystyle{\sum_{n=1}^{\infty}2\,\dfrac{1}{\pi}\,\dfrac{\sinh(\pi)}{1+n^2}}\\\\ &\Rightarrow \quad \dfrac{\cosh(\pi)\,\pi}{2\,\sinh(\pi)}-\dfrac{1}{2} =\displaystyle{\sum_{n=1}^{\infty}\dfrac{1}{1+n^2}} \end{align}$

It just remains a mystery why the method on top fails

  • 5
    $\begingroup$ I don't think you should be using Parseval. Rather, use the Fourier series expansion that you have for $\cosh(x)$, and evaluate at $x=\pi$. $\endgroup$ Jul 25, 2021 at 21:32

1 Answer 1


Note that when you did Parseval, you had (ignoring other pieces in the expression)

$$\sum_{n=1}^{\infty} \frac{1}{(1+n^2)^{\color{red}2}}.$$

Note the exponent in red. This is not $\displaystyle\sum_{n=1}^{\infty} \frac{1}{1+n^2}$ and is the source of the issues you are having.

Per Stefan Lafon (for posterity and the purposes of having an answer to this question): simply evaluate your Fourier series at $x=\pi$ to get the correct expression which is

$$ \frac{\pi}{2} \operatorname{coth}(\pi) - \frac{1}{2}.$$

  • $\begingroup$ I thought taking the square root would have fixed the issue but apparently it's changing the sum. Anyway the other method has worked. $\endgroup$
    – Leon
    Jul 27, 2021 at 19:36
  • 1
    $\begingroup$ Recall that $(a^2 + b^2)^{1/2} \neq a + b$ in general. This is just an extension of that fact. $\endgroup$ Jul 27, 2021 at 19:43

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.