1
$\begingroup$

Consider the integral

$$2^n\int_{0}^{1}{x^{2^n-1} \over 1+x^{2^n}}\ln{(-\ln{x})}\mathrm dx=-{2n+1\over 2}\cdot\ln^2{2}\tag1$$ $n\ge0$

An attempt:

$u=x^{2^n}$ $\implies$ $du=2^n\cdot x^{2^n-1}dx$

$(1)$ becomes

$$\int_{0}^{1}{\mathrm du\over 1+u}\cdot\ln{\left(-{1\over 2^n}\ln{u}\right)}\tag2$$

Another sub:

$v=-{1\over 2^n}\ln{u}$ $\implies$ $-2^nudv=du$

$(2)$ becomes

$$2^n\int_{0}^{\infty}{\mathrm dv\over 1+e^{2^nv}}\cdot\ln{v}\tag3$$

Maybe I could apply geometric series and integrate term by term?

How else can we confirm the result of $(1)?$

$\endgroup$
  • $\begingroup$ Confirm numerically first ? $\endgroup$ – Gabriel Romon Mar 8 '17 at 19:58
  • $\begingroup$ @Dr.MV: sorry Mark, I do not get it: why is $(2)$ incorrect? $\endgroup$ – Jack D'Aurizio Mar 8 '17 at 21:13
  • $\begingroup$ @jackd'aurizio Well, I guess it is I who have the bad eyes here. I read the numerator of $(1)$ as $x{2^n}-1$ and not $x^{2^n-1}$. $\endgroup$ – Mark Viola Mar 8 '17 at 21:24
2
$\begingroup$

Starting from your $(3)$ and using the substitution $v=\frac{u}{2^n}$ it is enough to compute the integrals $$ I=\int_{0}^{+\infty}\frac{du}{1+e^u}=\log(2),\qquad J=\int_{0}^{+\infty}\frac{\log u}{1+e^u}\,du.\tag{4} $$ Since $$ g(\alpha)=\int_{0}^{+\infty}\frac{u^\alpha}{1+e^u}\,du = (1-2^{-\alpha})\,\Gamma(1+\alpha)\,\zeta(1+\alpha) \tag{5}$$ for any $\alpha>-1$, we have: $$ J=g'(0) = g(0)\cdot\left.\frac{d}{d\alpha}\log g(\alpha)\right|_{\alpha=0}\tag{6}$$ and the claim follows from known values for $\Gamma,\zeta,\Gamma'$ and $\zeta'$ or simply $\eta$ and $\eta'$. Alternative methods for the evaluation of $\int_{0}^{+\infty}\frac{\log x}{e^x+1}\,dx$ are given here, here and here - thanks to Hazem Orabi.

$\endgroup$
  • 1
    $\begingroup$ $$\int_{0}^{\infty}\frac{\log{v}}{e^{2^n v}+1}\,2^n dv = \int_{0}^{\infty}\frac{\log\left(u/2^n\right)}{e^u+1}\,du = -n\log{2}\int_{0}^{\infty}\frac{du}{e^u+1}+\int_{0}^{\infty}\frac{\log{u}}{e^u+1}\,du$$ And for the record, alternative methods of computing: $\displaystyle\,\int_{0}^{\infty}\frac{\log(x)}{e^{x}+1}\,dx=-\frac{\log^{2}2}{2}\,$ are available in (1) and (2). (+1) $\endgroup$ – Hazem Orabi Mar 8 '17 at 21:57

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.