36

I found the following integral as a by product of another one.

It has a nice closed form.

$$ \int_{0}^{\pi} \arctan\left(\ln\left(\sin x \right) \over x\right)\,{\rm d}x $$

Mathematica and Maple fail to give the answer. Could you find it?

Hint 1:

The closed form is

$$ -\pi\arctan \left(2\ln 2 \over \pi\right) $$

Hint 2:

The following integral may help

$$ \int_{0}^{\pi}{x \over x^{2} + \ln^{2}\left(\alpha\sin x \right)} \,{\rm d}x $$

(see this post).

Olivier Oloa
  • 122,789
  • I upvoted, but the people who downvoted probably expect you to post the solution if you have it. – user111187 Jul 25 '14 at 23:17
  • Wolfram Alpha spits out a number. However which base is $\log$ and is it indeed $\log(\sin(x))/x$? – mvw Jul 25 '14 at 23:21
  • @mvw $\log (e)=\ln e = 1$ – Olivier Oloa Jul 25 '14 at 23:32
  • What is the closed form @OlivierOloa – ClassicStyle Jul 25 '14 at 23:37
  • @TylerHG OK. Please see the question ... – Olivier Oloa Jul 25 '14 at 23:38
  • If one applies the transformation $x\mapsto x-\frac{\pi}{2}$, the integrand looks fairly close to the closed form. But I'm stuck at that point. – apnorton Jul 26 '14 at 00:18
  • @anorton Good remark! So you may think of ... I will give another hint. – Olivier Oloa Jul 26 '14 at 00:20
  • I did integration by parts and got that form in the hint.. – recmath Jul 26 '14 at 01:02
  • I do not see why people would cast downvotes and close votes to this thread. – Pranav Arora Jul 26 '14 at 10:05
  • Since nobody has posted an answer, perhaps you could post your solution as an answer. – user111187 Jul 27 '14 at 21:59
  • 1
    @user111187 Yes, later. Thank you for your interest. – Olivier Oloa Jul 27 '14 at 22:03
  • @OlivierOloa Bonsoir, Je me permets de commenter votre question, dans le but de vous demander, quels livres conseillez-vous pour le calcul integral ? (cours, méthodes, exemples..) En français si possible. Merci d'avance. –  Jun 01 '15 at 18:15
  • @Gato Je n'avais pas beaucoup de livres lorsque j'étais étudiant, mais je lisais beaucoup les périodiques, notamment le mensuel 'American Mathematical Monthly' qui regorge d'intégrales intéressantes, de séries, de belles démonstrations, etc. Sinon, j'aime bien les ouvrages suivants (surtout en analyse) : Arnaudies-Fraysse, J.-M. Monier, Ramis-Deschamp-Odoux, Donnedu, ... Bien cordialement. – Olivier Oloa Jun 01 '15 at 20:13
  • 1
    @OlivierOloa Merci pour cette réponse, je regarderai le périodique et les livres cités. Merci encore. –  Jun 01 '15 at 23:19

3 Answers3

30

Consider the integral of more general form $$I(\alpha)=\int_0^\pi \arctan\left(\frac{\ln(\alpha \sin x)}{x}\right)\,dx,\qquad 0<\alpha\leq 1.$$ Then for $\alpha\in(0,1)$ $$I'(\alpha)=\frac{1}{\alpha}\int_0^\pi \frac{x}{x^2+\ln^2(\alpha\sin x)}\,dx,$$ as in the hint. To calculate the last integral we use the following identity, mentioned by Jack D'Aurizio in his comment to this question: $$\frac{b}{a^2+b^2}=\int_0^{+\infty} e^{-ay}\sin by \,dy,\quad a>0.\quad (*)$$ Setting $a=-\ln(\alpha\sin x)>0$ and $b=x$, we get $$I'(\alpha)=\frac{1}{\alpha}\int_0^\pi \left(\int_0^{+\infty} e^{y\ln(\alpha\sin x)}\sin xy \,dy\right)\,dx=\frac{1}{\alpha}\int_0^{+\infty}\left(\int_0^\pi (\alpha\sin x)^y\sin xy \,dx\right)\,dy.$$ (Changing order of integration is legitimate, since $|e^{y\ln(\alpha\sin x)}\sin xy|\leq e^{y\ln \alpha}$, so integral $\int_0^{+\infty} e^{y\ln(\alpha\sin x)}\sin xy \,dy$ converges uniformly by $x\in[0,\pi]$.) Therefore $$I'(\alpha)=\frac{1}{\alpha}\int_0^{+\infty}\left(\frac{\alpha}{2}\right)^y\left(\int_0^\pi (2\sin x)^y\sin xy \,dx\right)\,dy.$$ Now we need to deal with $$J=\int_0^\pi (2\sin x)^y\sin xy \,dx.$$ Changing the variable $x=t+\frac{\pi}{2}$ yields $$J=\int_{-\pi/2}^{\pi/2}(2\cos x)^y\left(\sin ty\cos\frac{\pi y}{2}+\cos ty\sin\frac{\pi y}{2}\right)\,dt=\sin\frac{\pi y}{2}\int_{-\pi/2}^{\pi/2}(2\cos t)^y\cos ty\,dt.$$ For the last integral we observe that $$\int_{-\pi/2}^{\pi/2}(2\cos t)^y\cos ty\,dt=\int_{-\pi/2}^{\pi/2}e^{y(\ln(2\cos t)-it)}\,dy=\int_{-\pi/2}^{\pi/2}(1+e^{-2it})^y\,dt=$$ $$=-\frac{1}{2i}\int_{-\pi/2}^{\pi/2}(1+e^{-2it})^y\frac{de^{-2it}}{e^{-2it}}=-\frac{1}{2i}\int_{C^-}\frac{(1+z)^y}{z}\,dz,$$ where $C^-$ is the unit circle (clockwise). Since $f(z)=\frac{(1+z)^y}{z}$ has just one simple pole $z=0$ inside $C^-$ with residue $\mathop{\mathrm{Res}}\limits_{z=0}f(z)=\lim\limits_{z\to 0}(1+z)^y=1$, we get $$-\frac{1}{2i}\int_{C^-}\frac{(1+z)^y}{z}\,dz=-\frac{1}{2i}(-2\pi i\cdot 1)=\pi$$ and $$J=\pi\sin\frac{\pi y}{2}.$$ Using $(*)$ one more time, we get $$I'(\alpha)=\frac{\pi}{\alpha}\int_0^{+\infty}\left(\frac{\alpha}{2}\right)^y\sin\frac{\pi y}{2}\,dy=\frac{\pi^2}{2\alpha(\ln^2\frac{\alpha}{2}+\frac{\pi^2}{4})}.$$ Now we can restore $I(\alpha)$ from its derivative: $$I(\alpha)=\frac{\pi^2}{2}\int\frac{d\alpha}{\alpha(\ln^2\frac{\alpha}{2}+\frac{\pi^2}{4})}=\frac{\pi^2}{2}\frac{2}{\pi}\arctan\left(\frac{2}{\pi}\ln\frac{\alpha}{2}\right)+c=\pi\arctan\left(\frac{2}{\pi}\ln\frac{\alpha}{2}\right)+c.$$

The next step is to show that $c=0$. For that purpose we observe that $$\ln(\alpha\sin x)\leq\ln\alpha\quad\Rightarrow\quad -\frac{\pi}{2}<\arctan\left(\frac{\ln(\alpha \sin x)}{x}\right)\leq \arctan\left(\frac{\ln\alpha}{x}\right)\quad\Rightarrow$$ $$-\frac{\pi^2}{2}\leq I(\alpha)\leq \int_0^\pi \arctan\left(\frac{\ln\alpha}{x}\right)\,dx\to -\frac{\pi^2}{2}$$ as $\alpha\to 0+$, so $\lim\limits_{\alpha\to 0+}I(\alpha)=-\frac{\pi^2}{2}$. Also it has to equal to $$\lim\limits_{\alpha\to 0+}\left(\pi\arctan\left(\frac{2}{\pi}\ln\frac{\alpha}{2}\right)+c\right)=-\frac{\pi^2}{2}+c,\quad\Rightarrow \quad c=0.$$ By now we have established that $$I(\alpha)=\int_0^\pi \arctan\left(\frac{\ln(\alpha \sin x)}{x}\right)\,dx=\pi\arctan\left(\frac{2}{\pi}\ln\frac{\alpha}{2}\right),\qquad 0<\alpha<1.$$ Letting $\alpha\to1-0$ we get the desired value $$I(1)=-\pi\arctan\left(\frac{2}{\pi}\ln2\right).$$ (we can change here limit and integral, since our integral is just proper).

CuriousGuest
  • 4,341
  • 2
    I'm surprised your answer only has 3 upvotes (including mine). It deserves a lot more. – Random Variable Jul 30 '14 at 16:25
  • 1
    @RandomVariable Thanks, I really appreciate your upvote and comment, they mean a lot. Also cheers to Olivier Oloa for the problem, I really enjoyed solving it. – CuriousGuest Jul 30 '14 at 17:06
  • This comment is 11 years after the fact, but I just wanted to add that you don’t need a restriction on alpha here believe it or not. You can extend this definition for all positive values if alpha, with the most interesting value occurring when alpha equals 2. – Jessie Christian Jan 24 '25 at 03:02
15

$\newcommand{\+}{^{\dagger}} \newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\down}{\downarrow} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\isdiv}{\,\left.\right\vert\,} \newcommand{\ket}[1]{\left\vert #1\right\rangle} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large A}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert} \newcommand{\wt}[1]{\widetilde{#1}}$ $\ds{\int_{0}^{\pi}\arctan\pars{\ln\pars{\sin\pars{x}} \over x}\,\dd x =-\pi\,\arctan\pars{2\ln\pars{2} \over \pi}}$

With $\ds{{\large\tt 0 < \mu < 1}}$: \begin{align} \mbox{Lets define}\quad{\cal F}\pars{\mu}&\equiv \int_{0}^{\pi}\arctan\pars{\ln\pars{\mu\sin\pars{x}} \over x}\,\dd x\quad \mbox{such that} \\[3mm]{\cal F}'\pars{\mu}&=\int_{0}^{\pi} {1 \over \bracks{\ln\pars{\mu\sin\pars{x}}/x}^{2} + 1} \,{1 \over x}\,{1 \over \mu\sin\pars{x}}\,\sin\pars{x}\,\dd x \\[3mm]&={1 \over \mu}\int_{0}^{\pi}{x \over \ln^{2}\pars{\mu\sin\pars{x}} + x^{2}} \,\dd x =-\,{1 \over \mu}\,\Im\int_{0}^{\pi}{\dd x \over \ln\pars{\mu\sin\pars{x}} + x\ic} \end{align}

$$ \mbox{We are interested in}\quad{\cal F}\pars{1^{-}}:\ {\large ?}. \quad\mbox{Note that}\quad{\cal F}\pars{0^{+}} = -\,{\pi^{2} \over 2}\tag{1} $$

\begin{align} {\cal F}'\pars{\mu}&= -\,{1 \over \mu}\,\Im \int_{\verts{z}\ =\ 1\atop{\vphantom{\Huge A}0\ <\ {\rm Arg}\pars{z}\ <\ \pi}} {1 \over \ln\pars{\mu\bracks{z^{2} - 1}/\bracks{2\ic z}} + \ln\pars{z}}\,{\dd z \over \ic z} \\[3mm]&={1 \over \mu}\,\Re \int_{\verts{z}\ =\ 1\atop{\vphantom{\Huge A}0\ <\ {\rm Arg}\pars{z}\ <\ \pi}} {1 \over \ln\pars{\mu\bracks{1 - z^{2}}\ic/2}}\,{\dd z \over z} \end{align}

We 'close' the contourn with the line segment $\ds{\braces{\pars{x,0}\ \mid\ x \in \pars{-1,1}}}$. The segment is indented, with arcs of radius $\ds{\epsilon}$ such that $\ds{0 < \epsilon < 1}$, around $\ds{z = -1}$, $\ds{z = 0}$ and $\ds{z = 1}$. It turns out that the contributions from the 'indented points' at $\ds{z = \pm 1}$ vanishes out in the limit $\ds{\epsilon \to 0^{+}}$. We are left with a principal value along $\ds{\pars{-1,1}}$ and the contribution from the 'indented point' at $\ds{z = 0}$. The above mentioned principal value vanishes out $\ds{\pars{~\mbox{its integrand is odd in}\ \pars{-1,1}~}}$ such that the whole contribution to $\ds{{\cal F}'\pars{\mu}}$, in the limit $\ds{\epsilon \to 0^{+}}$, arises 'curiously and amusing' just from the 'indented point' at $\ds{z = 0}$.

It's shown as follows: \begin{align} {\cal F}'\pars{\mu}&=\left.-\,{1 \over \mu}\,\Re\int_{\pi/2}^{0} {1 \over \ln\pars{\mu\bracks{1 - z^{2}}\ic/2}}\,{\dd z \over z} \right\vert_{\,z\ =\ -1\ +\ \epsilon\expo{\ic\theta}} \\[3mm]&\phantom{=}-\,{1 \over \mu}\,\Re\int_{-1 + \epsilon}^{\epsilon} {1 \over \ln\pars{\mu\bracks{1 - x^{2}}\ic/2}}\,{\dd x \over x} \left.-\,{1 \over \mu}\,\Re\int_{\pi}^{0} {1 \over \ln\pars{\mu\bracks{1 - z^{2}}\ic/2}}\,{\dd z \over z} \right\vert_{\,z\ =\ \epsilon\expo{\ic\theta}} \\[3mm]&\phantom{=}-\,{1 \over \mu}\,\Re\int_{\epsilon}^{1 - \epsilon} {1 \over \ln\pars{\mu\bracks{1 - x^{2}}\ic/2}}\,{\dd x \over x} \\[3mm]&\phantom{=}\left.-\,{1 \over \mu}\,\Re\int_{\pi}^{\pi/2} {1 \over \ln\pars{\mu\bracks{1 - z^{2}}\ic/2}}\,{\dd z \over z} \right\vert_{\,z\ =\ 1\ +\ \epsilon\expo{\ic\theta}} \end{align}

\begin{align} {\cal F}'\pars{\mu}&= -\,{1 \over \mu}\,\Re\pp\ \overbrace{\int_{-1}^{1} {1 \over \ln\pars{\mu\bracks{1 - x^{2}}\ic/2}}\,{\dd x \over x}}^{\ds{=\ 0}}\ -\,{1 \over \mu}\,\Re\int_{\pi}^{0}{\ic\,\dd\theta \over \ln\pars{\mu\ic/2}} \\[3mm]&=-\,{\pi \over \mu}\,\Im\bracks{1 \over \ln\pars{\mu\ic/2}} \end{align}

By using the boundary condition $\pars{1}$: \begin{align} {\cal F}\pars{1^{-}}& =\int_{0}^{\pi}\arctan\pars{\ln\pars{\sin\pars{x}} \over x}\,\dd x =-\,{\pi^{2} \over 2} -\pi\,\Im\int_{0^{+}}^{1^{-}}{1 \over \mu}\,{\dd\mu \over \ln\pars{\mu\ic/2}} \\[3mm]&=-\,{\pi^{2} \over 2} -\pi\,\Im\int_{0^{+}}^{1^{-}}{\dd\mu/\mu \over \ln\pars{\mu/2} + \pi\ic/2} =-\,{\pi^{2} \over 2} -\pi\,\Im\int_{-\infty}^{-\ln\pars{2^{+}}}{\dd t \over t + \pi\ic/2} \\[3mm]&=-\,{\pi^{2} \over 2} +\pi\int_{-\infty}^{-\ln\pars{2^{+}}}{\pi\,\dd t/2 \over t^{2} + \pars{\pi/2}^{2}} =-\,{\pi^{2} \over 2} +\left.\pi\arctan\pars{2t \over \pi}\right\vert_{\,-\infty}^{\,-\ln\pars{2^{+}}} \\[3mm]&=-\,{\pi^{2} \over 2} + \pi\bracks{% \arctan\pars{-\,{2\ln\pars{2} \over \pi}} + {\pi \over 2}} \end{align}

$$\color{#66f}{\large% \int_{0}^{\pi}\arctan\pars{\ln\pars{\sin\pars{x}} \over x}\,\dd x =-\pi\,\arctan\pars{2\ln\pars{2} \over \pi}} \approx {\tt -1.3055} $$

Felix Marin
  • 94,079
10

$\def\Im{\mathrm{Im}}$ I hope it is not too late to present an alternative solution.

Let \begin{align} I(s)=\int^\pi_0\arctan\left(\frac{\ln(s\sin{x})}{x}\right)\mathrm{d}x \end{align} Differentiate under the integral sign to get \begin{align} I'(s) &=\frac{1}{s}\int^\pi_0\frac{x}{x^2+\ln^2(s\sin{x})}\mathrm{d}x\\ &=-\frac{1}{s}\Im\int^\pi_0\frac{1}{\ln\left(\frac{se^{i2x}-s}{2i}\right)}\mathrm{d}x\\ &=-\frac{1}{s}\Im\int_{|z|=1}\frac{1}{\ln\left(\frac{sz-s}{2i}\right)}\frac{\mathrm{d}z}{2iz}\\ &=-\frac{1}{s}\Im\frac{\pi}{\ln\left(-\frac{s}{2i}\right)}\\ &=-\frac{1}{s}\Im\frac{\pi}{\ln\left(\frac{s}{2}\right)+\frac{\pi i}{2}}\frac{\ln\left(\frac{s}{2}\right)-\frac{\pi i}{2}}{\ln\left(\frac{s}{2}\right)-\frac{\pi i}{2}}\\ &=\frac{1}{2s}\frac{\pi^2}{\ln^2\left(\frac{s}{2}\right)+\frac{\pi^2}{4}} \end{align} where the fourth equality follows from the residue theorem and the fact that the indent around the branch point $z=1$ produces no contribution as $\epsilon\to0$.

Integrating back, \begin{align} I(1) &=I(\infty)+\frac{\pi^2}{2}\int^{s=1}_{s=\infty}\frac{1}{\ln^2\left(\frac{s}{2}\right)+\frac{\pi^2}{4}}\mathrm{d}\ln\left(\frac{s}{2}\right)\\ &=\frac{\pi^2}{2}+\left.\frac{\pi^2}{2}\frac{2}{\pi}\arctan\left(\frac{2\ln\left(\frac{s}{2}\right)}{\pi}\right)\right|^1_\infty\\ &=-\pi\arctan\left(\frac{2\ln{2}}{\pi}\right) \end{align} as desired.

M.N.C.E.
  • 10,779