Does this $\int_{0}^{\infty}(\frac{\log x}{e^x})^n dx$ always have a closed form for $n$ being positive integer ? what about its irrationality

closed-formeuler-mascheroni-constantimproper-integralsintegrationirrational-numbers

It is known that $\int_{0}^{\infty}\left(\frac{\log x}{e^x}\right)^n dx=-\gamma$ for $n=1$ and for $n=2$ we have :$\frac{1}{12}(\pi^2+6(\gamma+\log 2)^2)$ and for $n=3$ we have this form ,

What I have noted is that for $n$ odd the integrand is negative and for $n$ even the integrand is positive , Now my question here is : How I prove that integral always have a closed form for any fixed integer $n$ ? And can we expect a general formula or any reccurence relation for that integrand for aribitrary integer $n$ ? Also what about its irrationality ?

For attempt: I have tried to use this method for $n=1$ in order to generalise it for any fixed $n$ :
\begin{equation}
\int\limits_{0}^{\infty} \mathrm{e}^{-x} x^k \mathrm{d} x = \Gamma(k+1)
\end{equation}

Differentiate with respect to $k$
\begin{equation}
\int\limits_{0}^{\infty} \mathrm{e}^{-x} x^k \mathrm{ln}(x) \mathrm{d} x = \frac{d\Gamma(k+1)}{dt} = \Gamma(k+1) \psi^{(0)}(k+1)
\end{equation}

Taking the limit $k \to 0$ yields
\begin{equation}
\int\limits_{0}^{\infty} \mathrm{e}^{-x} \mathrm{ln}(x) \mathrm{d} x = \Gamma(1) \psi^{(0)}(1) = -\gamma
\end{equation}

Now for $n=2$ I should use the integration by part I can up to the result but by this way it would be two long and complicated , Then I want probably an abreviate path to deduce any general formula for any arbitary $n$ ?

it is clear only it is irrational for $n=1,2$

Best Answer

Not a complete answer, but I'm sure it'll help:

Let $$I(a) = \int_0^\infty e^{-nx}x^a\,dx$$ $$\implies \frac{d^nI(a)}{da^n} = \int_0^\infty e^{-nx}x^a(\ln x)^n\,dx$$ Put $nx \rightarrow v$ in the first integral to get: $$I(a) = \frac1{n^{1+a}}\int_0^\infty e^{-v}v^a\,dv$$ $$\implies I(a) = \frac{\Gamma(1+a)}{n^{1+a}}$$ Now $$\implies \frac{d^nI(a)}{da^n}\bigg|_{a=0} = \frac{d^n}{da^n}\left(\frac{\Gamma(1+a)}{n^{1+a}}\right)\bigg|_{a=0}$$ Which evaluates to: $$\frac1{n}\sum_{k=0}^n(-1)^k\binom{n}{k}\Gamma^{(n-k)}(1+a)\ln^k(n)\bigg|_{a=0}$$ Where $\Gamma^{(n-k)}(1+a)$ is the $(n-k)$th derivative of the Gamma function.