5

I am asked to prove that: $$ \int_{0}^{+\infty}\frac{\log z}{1+z^{\alpha}}\,dz = -\frac{\pi^2}{\alpha^2}\cdot\frac{\cos\frac{\pi}{\alpha}}{\sin^2\frac{\pi}{\alpha}},$$ provided that $\alpha > 1$, with a complex analytic method, i.e. contour integration.

However, I was not able to find a good candidate as a meromorphic function to integrate, neither a proper contour. Would you mind giving me a hand?

Ron Gordon
  • 138,521
  • The contour integral over which contour? And what is $\theta$ in your case? – Jack D'Aurizio Apr 06 '15 at 03:04
  • @jackd'aurizio The thing is I'm not sure which contour to take, the questions says it has to enclose one pole, include the positive real axis from ε<<1 to R (where R>>1) and asks us to prove that the integral of the function of interest from 0 to infinity has the form -(π^2)cos(π/β)/(β^2)sin^2(π/β)). – Maths Meister Apr 08 '15 at 18:40
  • in fact, everything you need is just to prove the $\Gamma$ reflection formula. Your integral can be derived from differentiation under the integral sign, see below. – Jack D'Aurizio Apr 08 '15 at 20:16

4 Answers4

8

To do the contour integration, use a circular wedge of radius $R$ and angle $2 \pi/\alpha$ in the complex plane. This wedge encloses the pole at $z=e^{i \pi/\alpha}$. The integral about the arc vanishes as $R \to \infty$. (We technically should have a small cutout of radius $\epsilon$ about the origin, but we may ignore that piece as there is no contribution.)

enter image description here

The integrals that remain are over the real axis and over the line in the complex plane that is at an angle $\alpha$ with respect tot he real axis. Thus, by the residue theorem,

$$\int_0^{\infty} dx \frac{\log{x}}{1+x^{\alpha}} - e^{i 2 \pi/\alpha} \int_0^{\infty} dx \frac{\log{x}+i 2 \pi/\alpha}{1+x^{\alpha}} = i 2 \pi \operatorname*{Res}_{z=e^{i \pi/\alpha}} \frac{\log{z}}{1+z^{\alpha}} $$

which becomes

$$\left ( 1-e^{i 2 \pi/\alpha} \right ) \int_0^{\infty} dx \frac{\log{x}}{1+x^{\alpha}} - i \frac{2 \pi}{\alpha} e^{i 2 \pi/\alpha} \int_0^{\infty} \frac{dx}{1+x^{\alpha}} = \frac{2 \pi^2}{\alpha^2} e^{i \pi/\alpha}$$

To evaluate the second integral on the LHS, use the same contour and pole:

$$\left ( 1-e^{i 2 \pi/\alpha} \right )\int_0^{\infty} \frac{dx}{1+x^{\alpha}} = i 2 \pi \operatorname*{Res}_{z=e^{i \pi/\alpha}} \frac{1}{1+z^{\alpha}} = -\frac{i 2 \pi}{\alpha} e^{i \pi/\alpha}$$

so that

$$\int_0^{\infty} \frac{dx}{1+x^{\alpha}} = \frac{\pi}{\alpha \sin{(\pi/\alpha)}}$$

Thus, we now have that

$$-i 2 \sin{(\pi/\alpha)}\int_0^{\infty} dx \frac{\log{x}}{1+x^{\alpha}} - i \frac{2 \pi^2}{\alpha^2} \frac{e^{i \pi/\alpha}}{\sin{(\pi/\alpha)}} = \frac{2 \pi^2}{\alpha^2} $$

Take real and imaginary parts of the second term, and the real part cancels the RHS. Thus, dividing by the factor in front of the integral, we are left with

$$\int_0^{\infty} dx \frac{\log{x}}{1+x^{\alpha}} = -\frac{\pi^2}{\alpha^2} \frac{\cos{(\pi/\alpha)}}{\sin^2{(\pi/\alpha)}} $$

as was to be shown.

Ron Gordon
  • 138,521
2

We may notice that: $$\frac{\log z}{1+z^\alpha}=\frac{d}{d\alpha}\log\left(\frac{z^\alpha}{1+z^\alpha}\right)\tag{1}$$ hence it is enough to compute: $$ I(\alpha)=\int_{0}^{+\infty}\log\left(\frac{z^\alpha}{1+z^\alpha}\right)\,dz = -\frac{1}{\alpha}\int_{0}^{+\infty}z^{\frac{1}{\alpha}-1}\log\left(1+\frac{1}{z}\right)\,dz$$ or: $$ I(\alpha) = -\frac{1}{\alpha}\int_{0}^{+\infty}\frac{\log(1+z)}{z^{1+\frac{1}{\alpha}}}\,dz = -\frac{1}{\alpha}\left.\frac{d}{d\beta}\int_{0}^{+\infty}\frac{(z+1)^\beta}{z^{1+\frac{1}{\alpha}}}\,dz\,\right|_{\beta=0^+}\tag{2}$$ that exploiting the Euler beta function is just: $$ I(\alpha) = -\frac{1}{\alpha}\lim_{\beta\to 0^+}\frac{d}{d\beta}\left(\frac{\Gamma\left(-\frac{1}{\alpha}\right)\Gamma\left(\frac{1}{\alpha}-\beta\right)}{\Gamma\left(-\beta\right)}\right)=-\frac{\pi}{\sin\frac{\pi}{\alpha}}\tag{3}$$ by the $\Gamma$ reflection formula. If we differentiate $(3)$ with respect to $\alpha$ we get: $$ \int_{0}^{+\infty}\frac{\log z}{1+z^{\alpha}}\,dz = -\frac{\pi^2}{\alpha^2}\cdot\frac{\cos\frac{\pi}{\alpha}}{\sin^2\frac{\pi}{\alpha}}\tag{4}$$ as wanted.

Jack D'Aurizio
  • 353,855
2

This is similar to Jack D'Aurizio's solution, but proceeds by making use of the derivative $\displaystyle \frac{dx^b}{db}=\log(x)x^b$.

So, we begin by writing the integral of interest as

$$\begin{align} \int_0^\infty \frac{\log(z)}{1+z^\alpha}\,dz&=\left.\frac{d}{db}\left(\int_0^\infty \frac{x^b}{1+x^\alpha}\,dx\right)\right|_{b=0}\tag1 \end{align}$$

Enforcing the substitution $x\to x^{1/\alpha}$ in $(1)$ reveals

$$\begin{align} \int_0^\infty \frac{\log(z)}{1+z^\alpha}\,dz&=\left.\frac{d}{db}\left(\frac1\alpha\int_0^\infty \frac{x^{(b+1)/\alpha-1}}{1+x}\,dx\right)\right|_{b=0}\tag2 \end{align}$$

Using the integral representation $\displaystyle B(x,y)=\int_0^\infty\frac{t^{x-1}}{(1+t)^{x+y}}\,dt$ of the Beta function, the relationship $\displaystyle B(x,y)=\frac{\Gamma(x)\Gamma(y)}{\Gamma(y)}$, and the reflection formula $\displaystyle \Gamma(x)\Gamma(1-x)=\frac{\pi }{\sin(\pi x)}$ in $(2)$ (See This Answer for Development of these relationships), we find that $(2)$ can be expressed as

$$\begin{align} \int_0^\infty \frac{\log(z)}{1+z^\alpha}\,dz&=\left.\frac{d}{db}\left(\frac1\alpha B\left(\frac{b+1}{\alpha},1-\frac{b+1}{\alpha}\right)\right)\right|_{b=0}\\\\ &=\frac1\alpha \left.\frac{d}{db}\left(\Gamma\left(\frac{b+1}{\alpha}\right)\Gamma\left(1-\frac{b+1}{\alpha}\right)\right)\right|_{b=0}\\\\ &=\frac1\alpha \left.\frac{d}{db}\left(\frac{\pi}{\sin\left(\pi(b+1)/\alpha\right)}\right)\right|_{b=0}\\\\ &=-\frac{\pi^2}{\alpha^2}\frac{\cos(\pi \alpha)}{\sin^2(\pi \alpha)} \end{align}$$

as was to be shown!

Mark Viola
  • 179,405
  • It asked for a contour integral solution. Also the reflection formula for $\Gamma$ is far from being trivial – reuns May 16 '17 at 21:59
  • True, but why did you single out mine when another solution also proceeded without contour integration? Or are you going to post an analogous comment on that solution? And the OP seemed quite happy with the other solution. As for the reflection formula, See My Answer Here. – Mark Viola May 16 '17 at 22:00
  • @user1952009 I've embedded a link to an answer I posted that uses real analysis only to develop the relationships used herein. Note that the OP has assumed that $\alpha $ is real. – Mark Viola May 16 '17 at 22:08
  • because you are answering to a 2 yr old exercice.. – reuns May 16 '17 at 22:12
  • @user1952009 And is there anything inherently wrong with that? I am struggling to understand your point here. Or are you harassing me again? What rule of "law" are you applying to this? – Mark Viola May 16 '17 at 22:14
1

$\newcommand{\bbx}[1]{\,\bbox[8px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$

I start with the following rewriting $\ds{\pars{~z^{\alpha} \mapsto z~}}$

\begin{equation} \left.\int_{0}^{\infty}\!\!{\ln\pars{z} \over 1 + z^{\alpha}}\,\dd z\,\right\vert_{\ \alpha\ >\ 1} = {1 \over \alpha^{2}}\int_{0}^{\infty}\!\!{z^{1/\alpha - 1}\ln\pars{z} \over 1 + z}\,\dd z\label{1}\tag{1} \end{equation}


I'll consider $\ds{\,\mc{I} \equiv \oint_{\mc{C}}{z^{1/\alpha - 1}\ln\pars{z} \over 1 - z}\,\dd z}$. Both $\ds{z^{1/\alpha - 1}\ \mbox{and}\ \ln}$ are its principal branches. The contour $\ds{\,\mc{C}}$ is a key-hole one which 'takes care' of the above mentioned branch-cuts. Obviously, $\ds{\,\mc{I} = 0}$. \begin{align} 0 & \,\,\,\stackrel{\mrm{as}\ \epsilon\ \to\ 0^{+}}{\sim}\,\,\, -\int_{-\infty}^{-\epsilon} {\pars{-x}^{1/\alpha - 1} \exp\pars{\bracks{1/\alpha - 1}\ic\pi}\bracks{\ln\pars{-x} + \ic\pi} \over 1 - x}\,\dd x \\[2mm] & \phantom{\,\,\,\stackrel{\mrm{as}\ \epsilon\ \to\ 0^{+}}{\sim}\,\,\,\,\,\,} -\int_{-\epsilon}^{-\infty} {\pars{-x}^{1/\alpha - 1} \exp\pars{-\bracks{1/\alpha - 1}\ic\pi}\bracks{\ln\pars{-x} - \ic\pi} \over 1 - x}\,\dd x \end{align} The integral around an indented semicircle around the origin and around a circumference with a radius $\ds{R \to \infty}$ vanish out and they were omitted in the above expression. Then, as $\ds{\epsilon \to 0^{+}}$, \begin{align} 0 & = \expo{\ic\pi/\alpha}\int_{0}^{\infty} {x^{1/\alpha - 1}\bracks{\ln\pars{x} + \ic\pi} \over 1 + x}\,\dd x - \expo{-\ic\pi/\alpha}\int_{0}^{\infty} {x^{1/\alpha - 1}\bracks{\ln\pars{x} - \ic\pi} \over 1 + x}\,\dd x \\[5mm] & = 2\ic\sin\pars{\pi \over a} \int_{0}^{\infty}{x^{1/\alpha - 1}\ln\pars{x} \over 1 + x}\,\dd x + 2\pi\ic\cos\pars{\pi \over a} \int_{0}^{\infty}{x^{1/\alpha - 1} \over 1 + x}\,\dd x \end{align} \begin{equation} \bbx{\int_{0}^{\infty}{x^{1/\alpha - 1}\ln\pars{x} \over 1 + x}\,\dd x = -\pi\cot\pars{\pi \over a} \int_{0}^{\infty}{x^{1/\alpha - 1} \over 1 + x}\,\dd x}\label{2}\tag{2} \end{equation}
The remaining integral is evaluated by a similar procedure ( as the above one ). Namely: \begin{align} &\bbx{\int_{0}^{\infty}{x^{1/\alpha - 1} \over 1 + x}\,\dd x = \pi\csc\pars{\pi \over a}}\label{3}\tag{3} \end{align}

With \eqref{1}, \eqref{2} and \eqref{3}:

$$\bbox[15px,#ffe,border:1px dotted navy]{\ds{% \left.\int_{0}^{\infty}\!\!{\ln\pars{z} \over 1 + z^{\alpha}}\,\dd z\,\right\vert_{\ \alpha\ >\ 1} = -\pars{\pi \over \alpha}^{2}\cot\pars{\pi \over a}\csc\pars{\pi \over a}}} $$

Felix Marin
  • 89,464