51
$\begingroup$

I encountered this integral in my calculations:

$$\int_0^\infty\frac{\log\left(1+\frac{\pi^2}{4\,x}\right)}{e^{\sqrt{x}}-1}\mathrm dx=2\int_0^\infty\frac{x\log\left(1+\frac{\pi^2}{4\,x^2}\right)}{e^x-1}\mathrm dx=6.041880938342236884944983747836284...,$$

but could not find a closed-form representation for it. I tried to replace a constant factor $\frac{\pi^2}4$ with a parameter and take a derivative, that made the integral look simpler, but I still was not successful in solving it.

I also tried to find possible closed forms using Inverse Symbolic Calculator and WolframAlpha but they did not find anything.

Could you please help me to find a closed form (even using non-elementary special functions), if it exists?

$\endgroup$
  • $\begingroup$ why do you need a close form so badly, may I ask? why do you think such a thing exists? $\endgroup$ – Lost1 May 19 '13 at 22:21
  • 14
    $\begingroup$ I am not sure it exists, but if it exists I want to know it. Closed forms are easier to manipulate, sometimes closed forms of different integrals or sums contain terms that cancel each other etc. I found this paper very useful and interesting: J. M. Borwein, R. E. Crandall, "Closed Forms: What They Are and Why We Care". $\endgroup$ – Oksana Gimmel May 19 '13 at 22:34
  • 10
    $\begingroup$ Plus, there is always the frivolous answer that they are fun and challenging to arrive at. $\endgroup$ – Jon Claus May 19 '13 at 22:55
  • 1
    $\begingroup$ A solution to the same integral was given on another forum, here - integralsandseries.prophpbb.com/post456.html#p456 $\endgroup$ – Shobhit Bhatnagar Jun 19 '13 at 4:35
34
$\begingroup$

A closed form indeed exists for this integral: $$\int_0^\infty\frac{\log\left(1+\frac{\pi^2}{4\,x}\right)}{e^{\sqrt{x}}-1}\mathrm dx=\pi^2\left(\log\frac{\pi\,A^9\sqrt{2}}{\Gamma(\frac{1}{4})^2}-\frac{9}{8}\right)+2\,\pi\,C,$$ where $A$ is the Glaisher-Kinkelin constant and $C$ is the Catalan constant.

A more general result: $$\int_0^\infty\frac{\log\left(1+\frac{a}{x}\right)}{e^{\sqrt{x}}-1}\mathrm dx=8\,\pi^2\psi^{(-2)}\left(\frac{\sqrt{a}}{2\pi}\right)-\frac{a}2\left(1+\log\frac{4\pi^2}{a}\right)-2\pi\sqrt{a}\left(1+2\log\Gamma\left(\frac{\sqrt{a}}{2\pi}\right)\right),$$ where $\psi^{(-2)}(z)$ is the generalized polygamma function.

The proof is based on Binet's second formula, but I still need to sort out some details.

$\endgroup$

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.