I have tried to calculate $$ \int_{-\infty}^{\infty}\left[\arctan\left(x\right) \over x\right]^{2}\,{\rm d}x $$ with integration by parts and that didn't work.

I looked up the indefinite integral and found it contained a polylogarithm which I don't know how to use so I tried contour integration but got stuck.

$${\tt\mbox{Wolfram Alpha said the answer is}}\,\,\,{\large \pi\log\left(4\right)}$$

Can anyone show me how to do this integral ?.


$$ \begin{align} \int_{-\infty}^\infty\left(\frac{\arctan(x)}{x}\right)^2\mathrm{d}x &=2\int_0^\infty\left(\frac{\arctan(x)}{x}\right)^2\mathrm{d}x\tag{1}\\ &=2\int_0^\infty\left(\frac\pi2-\arctan(u)\right)^2\mathrm{d}u\tag{2}\\ &=2\int_0^{\pi/2}\left(\frac\pi2-v\right)^2\,\mathrm{d}\tan(v)\tag{3}\\ &=4\int_0^{\pi/2}\tan(v)\left(\frac\pi2-v\right)\,\mathrm{d}v\tag{4}\\ &=4\int_0^{\pi/2}w\cot(w)\,\mathrm{d}w\tag{5}\\ &=4\int_0^{\pi/2}w\,\mathrm{d}\log(\sin(w))\tag{6}\\ &=-4\int_0^{\pi/2}\log(\sin(w))\,\mathrm{d}w\tag{7} \end{align} $$ Explanation:
$(1)$: integrand is even, halve the domain and double
$(2)$: substitute $x=1/u$
$(3)$: substitute $u=\tan(v)$
$(4)$: integrate by parts
$(5)$: substitute $v=\pi/2-w$
$(6)$: $\mathrm{d}\log(\sin(w))=\cot(w)\,\mathrm{d}w$
$(7)$: integrate by parts

Steven Stadnicki suggests that we should mention
$\ \ $In the explanation of $(4)$, note that $\lim\limits_{v\to\pi/2}\left(\frac\pi2-v\right)^2\tan(v)=0$
$\ \ $In the explanation of $(7)$, note that $\lim\limits_{w\to0}w\log(\sin(w))=0$

Note that $$ \begin{align} &\int_0^{\pi/2}\log(\sin(w))\,\mathrm{d}w\\ &=\int_0^{\pi/2}\log(2\sin(w/2)\cos(w/2))\,\mathrm{d}w\\ &=\frac\pi2\log(2)+\int_0^{\pi/2}\log(\sin(w/2))\,\mathrm{d}w+\int_0^{\pi/2}\log(\cos(w/2))\,\mathrm{d}w\\ &=\frac\pi2\log(2)+2\int_0^{\pi/4}\log(\sin(w))\,\mathrm{d}w+2\int_0^{\pi/4}\log(\cos(w))\,\mathrm{d}w\\ &=\frac\pi2\log(2)+2\int_0^{\pi/4}\log(\sin(w))\,\mathrm{d}w+2\int_{\pi/4}^{\pi/2}\log(\sin(w))\,\mathrm{d}w\\ &=\frac\pi2\log(2)+2\int_0^{\pi/2}\log(\sin(w))\,\mathrm{d}w\tag{8} \end{align} $$ Therefore, $$ \int_0^{\pi/2}\log(\sin(w))\,\mathrm{d}w=-\frac\pi2\log(2)\tag{9} $$ Combining $(7)$ and $(9)$ yields $$ \int_{-\infty}^\infty\left(\frac{\arctan(x)}{x}\right)^2\mathrm{d}x =2\pi\log(2)\tag{10} $$

  • $\begingroup$ One tiny additional thing I'd add: in going from $(3)$ to $(4)$, you have to do a little bit of massage to get rid of the 'non-integral' term in the integration by parts (i.e., the $(\frac\pi2-v)^2\tan(v)$ term); it's obviously $0$ at $v=0$, but you have to show that the limit as $v\to\frac\pi2$ is 0 (which is easy because $\tan(v)$ goes to infinity as $(\frac\pi2-v)^{-1}$ but maybe deserves an explicit note). $\endgroup$ – Steven Stadnicki Aug 8 '14 at 20:22
  • $\begingroup$ I considered that simply part of the mechanics of integration by parts, but if we add that for $(4)$, we also need to add a similar comment for $(7)$. $\endgroup$ – robjohn Aug 8 '14 at 20:52
  • $\begingroup$ Good point; I'd missed that one. I consider it trivial when the result is well-defined at both inputs, but since there's a necessary limit here (you can't simply evaluate the result at the endpoint) I like calling it out explicitly. (And since I didn't note it before, this is really an excellent answer. I virtually never bother answering integration questions here; I'm not nearly as good at analytics in general, and your answers are consistently among the highest-quality on this site.) $\endgroup$ – Steven Stadnicki Aug 8 '14 at 21:15
  • $\begingroup$ I have not seen a cleaner answer on Math.SE. +1. $\endgroup$ – Ahaan S. Rungta Aug 10 '14 at 1:44

You can use the following way to evaluate. It is pretty neat and simple. Let $$ I(a,b)=\int_{-\infty}^\infty\frac{\arctan(ax)\arctan(bx)}{x^2}dx. $$ Clearly $I(0,b)=I(a,0)=0$ and $I(1,1)=I$. Now \begin{eqnarray} \frac{\partial^2I(a,b)}{\partial a\partial b}&=&\int_{-\infty}^\infty\frac{1}{(1+a^2x^2)(1+b^2x^2)}dx\\ &=&\frac{1}{a^2-b^2}\int_{-\infty}^\infty\left(\frac{a^2}{1+a^2x^2}-\frac{b^2}{1+b^2x^2}\right)dx\\ &=&\frac{1}{b^2-a^2}\pi(a-b)\\ &=&\frac{\pi}{a+b}. \end{eqnarray} Hence $$ I=\int_0^1\int_0^1\frac{\pi}{a+b}dadb=2\pi\ln2.$$

  • $\begingroup$ Nice! I think there's an algebraic error though? the second line should be $\frac{1}{a^2-b^2}\int_{-\infty}^\infty\left(\frac{a^2}{1+a^2x^2}-\frac{b^2}{1+b^2x^2}\right)$, and the third line adjusted also. $\endgroup$ – Holographer Aug 8 '14 at 12:39
  • $\begingroup$ @Holographer, you are right. Thanks. $\endgroup$ – xpaul Aug 8 '14 at 12:56
  • $\begingroup$ +1, very slick trick; convert a single integral to a double integral to make it easier. $\endgroup$ – robjohn Aug 8 '14 at 18:41
  • $\begingroup$ @robjohn, thanks. $\endgroup$ – xpaul Aug 10 '14 at 23:49

\begin{align} \int^{\infty}_{-\infty}\frac{(\arctan{x})^2}{x^2}dx &=2\int^{\infty}_{0}\frac{(\arctan{x})^2}{x^2}dx\tag1\\ &=2\int^{\pi/2}_0u^2\csc^2{u} \ du\tag2\\ &=4\int^{\pi/2}_0u\cot{u} \ du\tag3\\ &=-4\int^{\pi/2}_0\ln\sin{u} \ du\tag4\\ &=2\pi\ln{2}\tag5\\ \end{align} Explanation:
$1)$ Integrand is even
$2)$ Use the substitution $x=\tan{u}$
$3), 4)$ Integrate by parts
$5)$ See here


Since you mentioned you tried contour integration, I'll add an approach using contour integration.

Consider $ \displaystyle f(z) = \frac{\log^{2}(1-iz)}{z^{2}}$ where the branch cut runs down the imaginary axis from $z=-i$.

And notice that the singularity at the origin is removable.

Now integrate around a contour that consists of the line segment $[-R,R]$ and the upper half of the circle $|z|=R$.

As $R \to \infty$, $ \displaystyle \int f(z) \ dz$ vanishes along the upper half of $|z|=R$.


$$ \int_{-\infty}^{\infty} \frac{\big( \log|1-ix| - i \arctan (x) \big)^{2}}{x^{2}} \ dx = 0 $$

which implies $$\int_{-\infty}^{\infty} \frac{\frac{1}{4} \log^{2}(1+x^{2}) -i \log(1+x^{2}) \arctan (x) - \arctan^{2}(x) }{x^{2}} =0 . $$

And equating the real parts on both sides of the equation we have $$ \begin{align} \int_{-\infty}^{\infty} \frac{\arctan^{2} (x)}{x^{2}} \ dx &= \frac{1}{4} \int_{-\infty}^{\infty} \frac{\log^{2} (1+x^{2})}{x^{2}} \ dx \\ &= \frac{1}{2} \int_{0}^{\infty} \frac{\log^{2}(1+x^{2})}{x^{2}} \ dx \\ &= \frac{1}{4} \int_{0}^{\infty} \frac{\log^{2}(1+u)}{u^{3/2}} \ du \\ &= \frac{1}{4} (8 \pi \log 2)\tag{1} \\ &= 2 \pi \log 2 . \end{align}$$

$(1)$ Integrate $ \int_0^\infty \frac{ \ln^2(1+x)}{x^{3/2}} dx=8\pi \ln 2$


As was stated in Gennaro Marco Devincenzis' answer, integrating by parts shows that $$\int_{-\infty}^{\infty} \frac{\arctan^{2}(x)}{x^{2}} \ dx = 2 \int_{-\infty}^{\infty} \frac{\arctan x}{x(1+x^{2})} \ dx. $$

And that integral as well can be evaluated using contour integration, which is probably preferable since it's just a residue calculation.

See here for the details.

  • $\begingroup$ (+1) log and arctan are nicely related via complex analysis. $\endgroup$ – robjohn Aug 8 '14 at 18:54
  • $\begingroup$ Good point about integrating by parts then using contour integration on it. The second contour integral looks much easier. $\endgroup$ – Maxwell Aug 8 '14 at 23:01

Disclaimer: It's one of the first times I evaluate an integral whose primitive cannot be found in terms of elementary functions. God it's a drug.

Integrating by parts

$$\int_{-\infty}^{+\infty} \frac{\arctan^2 x}{x^2}dx=\bigg[-\frac{\arctan^2 x}{x}\bigg]_{-\infty}^{+\infty}+2\int_{-\infty}^{+\infty}\frac{\arctan x}{(1+x^2)x}dx$$ The first part yields $0$. For the second consider: $$I(t)=\int_{-\infty}^{+\infty}\frac{\arctan tx}{x(1+x^2)}dx$$ Differentiating under the integral sign: $$I'(t)=\int_{-\infty}^{+\infty}\frac{1}{(1+t^2x^2)(1+x^2)}dx=\frac{\pi}{t+1}$$ $$I(t)=\pi\log(t+1)+C$$ So: $$\int_{-\infty}^{+\infty}\frac{\arctan tx}{x(1+x^2)}dx=\pi\log(t+1)+C$$ Letting $t=0$ we recover $C=0$. So, letting t=1: $$\int_{-\infty}^{+\infty}\frac{\arctan x}{x(1+x^2)}dx=\pi\log(2)$$ Since there was a $2$ in the integration by parts we have: $$\int_{-\infty}^{+\infty} \frac{\arctan^2 x}{x^2}dx=2\int_{-\infty}^{+\infty}\frac{\arctan x}{x(1+x^2)}dx=2\pi\log2=\pi\log4$$

  • 1
    $\begingroup$ I have always liked the differentiating under the integral sign trick. Sweet. $\endgroup$ – Asier Calbet Aug 8 '14 at 18:15
  • $\begingroup$ Me too! But it's one of the first times I succeed in using it effectively. :P $\endgroup$ – Gennaro Marco Devincenzis Aug 8 '14 at 18:17
  • 1
    $\begingroup$ There are tons of classical examples online if you want to practice using it. ;) $\endgroup$ – Asier Calbet Aug 8 '14 at 18:19
  • 2
    $\begingroup$ I am pretty annoyed at myself for not using differentiation under the integral sign. The success I have had with contour integration lately made me over look it. This is great. $\endgroup$ – Maxwell Aug 8 '14 at 22:59

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

\begin{align}&\color{#66f}{\large\int_{-\infty}^{\infty}% \bracks{\arctan\pars{x} \over x}^{2}\,\dd x} =\int_{x\ \to\ -\infty}^{x\ \to\ \infty} \arctan^{2}\pars{x}\,\dd\pars{-\,{1 \over x}} \\[3mm]&=\int_{-\infty}^{\infty} {1 \over x}\bracks{2\arctan\pars{x}\,{1 \over 1 + x^{2}}}\,\dd x =2\,\Im\ \overbrace{\int_{-\infty}^{\infty}{\ln\pars{1 + x\ic} \over x\pars{1 + x^{2}}} \,\dd x}^{\ds{\mbox{Set}\ 1 + x\ic \equiv t\ \imp\ x = \pars{1 - t}\ic}} \\[3mm]&=2\,\Im\int_{1 - \infty\ic}^{1 + \infty\ic} {\ln\pars{t} \over \pars{1 - t}\ic\pars{2 - t} t}\,\pars{-\ic\,\dd t} =2\,\Im\int_{1 - \infty\ic}^{1 + \infty\ic} {\ln\pars{t} \over t\pars{t - 1}\pars{t - 2}}\,\dd t \\[3mm]&=2\,\Im\bracks{2\pi\ic\,{\ln\pars{2} + \ic 0^{+} \over 2\pars{2 - 1}}} =\color{#66f}{\large 2\pi\ln\pars{2}} \approx {\tt 4.3552} \end{align}

In the above calculation we took the $\ds{\ln}$-branch cut: $$ \ln\pars{z} = \ln\pars{\verts{z}} + {\rm Arg}\pars{z}\ic\,,\quad\verts{{\rm Arg}\pars{z}} < \pi\,,\quad z \not= 0 $$ The integration path is closed with the arc $\ds{\braces{\pars{x,y}\quad \mid\quad \pars{x - 1}^{2} + y^{2} = R^{2}\,,\quad x \geq 1\,,\quad R > 1}}$ such that the 'arc contribution' vanishes out in the limit $\ds{R \to \infty}$.


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.