What kind of substitution should I use to obtain the following integrals? $$\begin{align} \int_0^1 \ln \ln \left(\frac{1}{x}\right)\,dx &=\int_0^\infty e^{-x} \ln x\,dx\tag1\\ &=\int_0^\infty \left(\frac{1}{xe^x} - \frac{1}{e^x-1} \right)\,dx\tag2\\ &=-\int_0^1 \left(\frac{1}{1-x} + \frac{1}{\ln x} \right)\,dx\tag3\\ &=\int_0^\infty \left( e^{-x} - \frac{1}{1+x^k} \right)\,\frac{dx}{x},\qquad k>0\tag4\\ \end{align}$$

This is not homework problems and I know that the above integrals equal to negative of the Euler–Mascheroni constant. I got these integrals while reading this Wikipedia page: The Euler–Mascheroni constant. According to Wikipedia, the Euler–Mascheroni constant is defined as the limiting difference between the harmonic series and the natural logarithm: $$\gamma=\lim_{N\to\infty} \left(\sum_{k=1}^N \frac{1}{k} - \ln N\right)$$ but I don't know why can this definition be associated to the above integrals?

I can obtain the equation $(1)$ using substitution $t=\ln \left(\frac{1}{x}\right)\rightarrow x=e^{-t} \rightarrow dx=-e^{-t}\,dt$ and I know that $$\int_0^\infty e^{-x} \ln x\,dx=\Gamma'(1)=\Gamma(1)\psi(1)=-\gamma$$ but I can't obtain the rest. Any idea? Any help would be appreciated. Thanks in advance.


I am not sure if you can obtain the other three from the first integral through some substitution.

The $(3)$ one can be obtained from the definition of constant..

Notice that $$\sum_{i=1}^N\frac{1}{i}=\int_0^1 \frac{1-t^N}{1-t}\,dt$$ and $$\ln N=\int_0^1 \frac{t^{N-1}-1}{\ln t}\,dt$$ The above can be proved using Frullani's integral. Therefore $$\begin{aligned} \lim_{N\rightarrow \infty} \left(\sum_{i=1}^N\frac{1}{i}-\ln N\right) &=\lim_{N\rightarrow \infty}\int_0^1\left(\frac{1-t^N}{1-t}-\frac{t^{N-1}-1}{\ln t}\,dt\right)\,dt \\ &=\int_0^1 \left(\frac{1}{1-t}+\frac{1}{\ln t}\right)\,dt \end{aligned}$$ Make the substitution $\ln x=-t$ to obtain the $(2)$.

  • $\begingroup$ Thanks for your answer. +1 $\endgroup$ – Venus Oct 19 '14 at 11:58
  • $\begingroup$ +1. BTW, long time no see you here Natsu Dragneel, what's up? (❛‿❛✿̶̥̥) $\endgroup$ – Anastasiya-Romanova 秀 Oct 19 '14 at 12:10
  • $\begingroup$ @Anastasiya-Romanova: Thanks. I really wish I could be more active here but college stuff doesn't leave me with any free time these days. $\endgroup$ – Pranav Arora Oct 19 '14 at 15:13

For the fourth one, express the integral as an iterated integral and then switch the order of integration.

$$ \begin{align} \int_{0}^{\infty} \left( e^{-x}- \frac{1}{1+x^{k}} \right) \frac{dx}{x} &= \int_{0}^{\infty} \int_{0}^{\infty} \left(e^{-x} - \frac{1}{1+x^{k}} \right) e^{-xt} \ dt \ dx \\ &= \int_{0}^{\infty} \int_{0}^{\infty} \left(e^{-x} - \frac{1}{1+x^{k}} \right) e^{-tx} \ dx \ dt \\ &= \int_{0}^{\infty} \left(\int_{0}^{\infty} e^{-(t+1)x} \ dx - \int_{0}^{\infty} \frac{e^{-tx}}{1+x^{k}} \ dx \right) \ dt \\ &= \int_{0}^{\infty} \left( \frac{1}{t+1} - \int_{0}^{\infty} \frac{e^{-u}}{1+ (\frac{u}{t})^{k}} \frac{du}{t} \right) \ dt \\ &= \int_{0}^{\infty} \left(\int_{0}^{\infty}\frac{e^{-u}}{t+1} \ du - \int_{0}^{\infty} \frac{t^{k-1}e^{-u}}{t^{k}+u^{k}} \ du \right) \ dt \tag{1} \\ &= \int_{0}^{\infty} \int_{0}^{\infty} \left(\frac{1}{t+1} - \frac{t^{k-1}}{t^{k}+u^{k}} \right)e^{-u} \ du \ dt \\ &= \int_{0}^{\infty} e^{-u} \int_{0}^{\infty} \left(\frac{1}{t+1} - \frac{t^{k-1}}{t^{k}+u^{k}} \right) \ dt \ du \\ &= \int_{0}^{\infty} e^{-u} \ln \left(\frac{1+t}{(t^{k}+u^{k})^{1/k}} \right) \Bigg|^{t=\infty}_{t=0} \ du \\ &= \int_{0}^{\infty} e^{-u} \ln u = -\gamma \end{align}$$

$(1)$ Since $\int_{0}^{\infty} e^{-u} \ du = 1$

$ $

Since I'm pretty sure I saw sos440 use this approach on another site to show that $$ \int_{0}^{\infty} \left(\cos (x) - \frac{1}{1+x^{2}} \right) \frac{dx}{x} = - \gamma,$$ I'm going to post this answer as a community wiki.


For equation $(2)$, make substitution $t=e^{-x}\,\Rightarrow\, x=-\ln t\,\Rightarrow\,dx=-\dfrac{dt}{t}$, then we get equation $(3)$ after $t\mapsto x$. \begin{equation} \int_0^\infty \left(\frac{1}{xe^x} - \frac{1}{e^x-1} \right)\,dx=\int_0^\infty \left(\frac{e^{-x}}{x} - \frac{e^{-x}}{1-e^{-x}} \right)\,dx=\int_0^1 \left(\frac{1}{1-x} \color{red}{+} \frac{1}{\ln x} \right)\,dx \end{equation} Indeed equation $(3)$ equals $-\gamma$ and here is the proof. We have \begin{equation} -\int_0^1 \left(\frac{1}{1-x} \color{red}{+} \frac{1}{\ln x} \right)\,dx=-\int_0^1 \frac{\ln x+1-x}{(1-x)\ln x}\,dx \end{equation} I check on the cited link (Wikipedia), there is a minor typo in $(3)$.

Proposition :

\begin{equation} I(s)=\int_0^1 \frac{s\ln x+1-x^s}{(1-x)\ln x}\,dx=\ln\Gamma(s+1) +\gamma s \end{equation}

Proof :

Let \begin{equation} I(s)=\int_0^1 \frac{s\ln x+1-x^s}{(1-x)\ln x}\,dx \end{equation} then \begin{align*} I'(s)&=\int_0^1 \frac{1-x^s}{1-x}\,dx\\ I''(s)&=-\int_0^1 \frac{x^s\ln x}{1-x} \,dx\\ &=-\int_0^1\sum_{n=0}^\infty x^{n+s}\ln x\,dx\\ &=-\sum_{n=0}^\infty\partial_s\int_0^1 x^{n+s}\,dx\\ &=-\sum_{n=0}^\infty\partial_s\left[\frac{1}{n+s+1}\right]\\ &=\sum_{n=0}^\infty\frac{1}{(n+s+1)^2}\\ &=\psi_1(s+1)\\ I'(s)&=\int\psi_1(s+1)\,ds\\ I'(s)&=\int\frac{\partial}{\partial s}\bigg[\psi(s+1)\bigg]\,ds\\ I'(s)&=\psi(s+1)+C\\ \end{align*} For $s=0$, we have $I'(0)=0$. Implying $C=-\psi(1)=\gamma$, then \begin{align*} I'(s)&=\psi(s+1)+\gamma\\ I(s)&=\int \psi(s+1)\,ds +\gamma s+C\\ &=\int \frac{\partial}{\partial s}\bigg[\ln\Gamma(s+1)\bigg]\,ds +\gamma s+C\\ &=\ln\Gamma(s+1) +\gamma s+C\\ \end{align*} For $s=0$, we have $I(0)=0$. Implying $C=0$, then \begin{equation} I(s)=\int_0^1 \frac{s\ln x+1-x^s}{(1-x)\ln x}\,dx=\ln\Gamma(s+1) +\gamma s\qquad\qquad\square \end{equation}

For $s=1$, we have

\begin{equation} -I(1)=-\int_0^1 \left(\frac{1}{1-x} + \frac{1}{\ln x} \right)\,dx=-\int_0^1 \frac{\ln x+1-x}{(1-x)\ln x}\,dx=-\gamma \end{equation}

  • 1
    $\begingroup$ Thanks for your answer. +1 $\endgroup$ – Venus Oct 19 '14 at 11:58

The easiest way to obtain this result is to use a regularization procedure. I will explain how to do the integral associated with a previous response, $$\int_{0}^{\infty}\left(\mathrm{cos}\space x-\frac{1}{1+x^2}\right)\frac{dx}{x}=-\gamma$$

This integral is correct. Easiest way to show is to regularize the integral by adding the small real number $\delta$ to the exponent of $x^{-1}$. This, after the $u$ substitution $u=ix$, gives $$\mathrm{Re}\left[\Gamma(\delta)\left(\frac{1}{i}\right)^\delta\right]-\frac{\pi}{2}\mathrm{csc}\left(\frac{\pi\delta}{2}\right).$$ Now, we just use the recursion identity $\delta\Gamma(\delta)=\Gamma(\delta+1)$ along with the Taylor series for $\Gamma(1+\delta)$ and sin($x$), to show that the integral is given by $-\gamma$. There are two $\frac{1}{\delta}$ terms that cancel, and the remaining contribution when $\delta$ is very small is $-\gamma$. This is a very neat integral for $\gamma$!
I'd like to see the reason why $$\int_{0}^{\infty}\frac{\mathrm{ln}(1+x)}{\mathrm{ln}^2(x)+\pi^2}\frac{dx}{x^2}=\gamma$$ I don't have any 'neat tricks' for this expression... would appreciate any!


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.