# Evaluation of $4\int_0^{+\infty} \frac{\left(\sinh\left(\frac{x}{8}\right)\right)^2}{x(e^x-1)}dx$

From the well-known formula,

For $s$, such that $\Re(s)>1$,

$\displaystyle \zeta(s)=\dfrac{1}{\Gamma(s)}\int_0^{+\infty} \dfrac{x^{s-1}}{e^x-1}dx$

it follows that,

$\displaystyle \sum_{k=1}^{\infty} \frac{\zeta(2k)}{k16^{k}}=4\int_0^{+\infty} \dfrac{\left(\sinh\left(\tfrac{x}{8}\right)\right)^2}{x(e^x-1)}dx$

Is it possible to evaluate directly the latter integral?

• Looks like this with $a= i/8.$ – nospoon Jul 18 '16 at 13:39
• try abel plana! – tired Jul 18 '16 at 14:01
• @FDP Please, can you explain what's wrong with the answers below? – Olivier Oloa Feb 1 '18 at 18:58

An approach.

Hint. One may set $$f(s):=4\int_0^\infty \frac{\left(\sinh\left(sx\right)\right)^2}{x(e^x-1)}dx, \quad 0<s<\frac12. \tag1$$ By differentiating with respect to $s$ one gets \begin{align} f'(s)&=4\int_0^\infty \frac{\sinh\left(2sx\right)}{e^x-1}dx \\\\&=2\int_0^\infty\left(e^{2sx}-e^{-2sx}\right)\sum_{n=1}^\infty e^{-nx}dx \\\\&=2\sum_{n=1}^\infty\int_0^\infty\left(e^{2sx}-e^{-2sx}\right) e^{-nx}dx \\\\&=\sum_{n=1}^\infty\frac{8s}{n^2-4s^2} \\\\&=\frac1s-2\pi\cot(2\pi s) \tag2 \end{align} Integrating $(2)$, with the fact that, as $s \to 0$, $f(s) \to 0$, we get

$$\int_0^\infty \frac{\left(\sinh\left(sx\right)\right)^2}{x(e^x-1)}dx=\log \left(\frac{2\pi s}{\sin(2\pi s)}\right), \quad 0<s<\frac12, \tag4$$

from which you deduce the value of the given integral by putting $s:=\color{blue}{\frac18}$, which is $\color{blue}{\ln(\pi)-\frac32\ln 2}$.

Remark. With the general parameter $s$, we have obtained a little bit more than the initial integral.

• Whoever downvoted, please indicate the reason. I upvoted, because I don't see anything wrong with this answer – Yuriy S Jul 18 '16 at 14:15
• @You're In My Eye and Jack D'Aurizio good Job (+1) for all – Behrouz Maleki Jul 18 '16 at 14:25
• So he differentiates the integrand. Gets an integral that evaluates to a function that has pole at the value he wants to integrate to and then goes and just permutes the derivative with the integral. This needs some analysis to justify the permutation of the derivative with the integral. It's far from obvious. – zzchan Jul 18 '16 at 14:30
• @zzchan Do you even know what does a "hint" even mean? Clearly, this is far from a hint. It's an answer. The series $\displaystyle\sum_{n=1}^\infty\frac{1}{n^2-a^2}$ is very typical and common. – Anastasiya-Romanova 秀 Jul 18 '16 at 14:51
• Nice to see you around Anastasiya-Romanova :) – FDP Jul 18 '16 at 15:00

The identity: $$\sum_{k\geq 1}\frac{\zeta(2k)}{k}x^{2k} = \log\left(\frac{\pi x}{\sin (\pi x)}\right) \tag{1}$$ holds for any $x$ such that $|x|<1$ by the Weierstrass product for the sine function.
By setting $x=\frac{1}{4}$ we recover the wanted result.
