21

Some rumours point out that the integral you see might be evaluated in a straightforward way.
But rumours are sometimes just rumours. Could you confirm/refute it?

$$ \int_0^{\infty}\left[\frac{\log\left(x\right)\arctan\left(x\right)}{x}\right]^{2} \,{\rm d}x $$

EDIT W|A tells the integral evaluates $0$ but this is not true. Then how do I exactly compute it?

beep-boop
  • 11,595
user68326
  • 251
  • Wow - even though, as Isaac says, it's non-negative (zero at $x=1$) even Wolfram|Alpha says so! – wchargin Mar 24 '13 at 21:06
  • 12
    The function is non-negative, positive and continuous (except perhaps at the origin). I don't think that the integral should be zero, even if that is what WA is indicating. Then again, if you change the integral from 0 to 5, for example, you get a positive number. Perhaps this is an error on WA? – Elchanan Solomon Mar 24 '13 at 21:08
  • @Isaac - I suppose so, and it doesn't have the "Show step-by-step solution" like it usually does – wchargin Mar 24 '13 at 21:11
  • @IsaacSolomon: W|A insists is $0$. Why? Then if it's not $0$ what is its real value? – user68326 Mar 24 '13 at 21:11
  • Also if you just write "integral of (the function)" into W|A, under "Definite integrals" it says the value is 6.200200821... – wchargin Mar 24 '13 at 21:12
  • 5
    Mathematica agrees that the definite integral should be zero (yet clearly it should not be, as Isaac's comment explains). Cool! I've seen this before - usually it is due to Wolfram's integration methods (complex functions are employed). Report it to Wolfram, sometimes you can get a free T-shirt out of it! – icurays1 Mar 24 '13 at 21:15
  • 2
    @user68326 I would follow the suggestion in this comment if I were you. – Git Gud Mar 24 '13 at 21:15
  • 1
    @icurays1 I smile at this coincidence. Yet I weep knowing I haven't won the lottery yet. – Git Gud Mar 24 '13 at 21:16
  • 3
    I don't understand why everyone is upvoting this question. It was posed as a riddle, without any indication that the "rumours" refer to Wolfram|Alpha; if WChargin hadn't told us about the Wolfram|Alpha bug, the question would have made no sense whatsoever. I downvoted. – joriki Mar 24 '13 at 21:16
  • @IsaacSolomon You should turn your comment into an answer. – Git Gud Mar 24 '13 at 21:23
  • 2
    I originally upvoted, @joriki, but agree now that the question was deceptive in light of how it all turned out. My vote is locked in, though. +1 for your comment. – Lepidopterist Mar 24 '13 at 21:25
  • 2
    @user68326 your phrasing in terms of "rumours" without attributing it to WA was deceptive. calling him frustrated is rude on top of that. why don't you prove it, or show why you think this had a hope of being evaluated in a straightforward manner. – Lepidopterist Mar 24 '13 at 21:30
  • @Lepidopterist: do you know what is rude? To know nothing about mathematics but to spread your opinions here as they are valuable opinions. This is rude. Perhaps you're another frustrated. – user68326 Mar 24 '13 at 21:34
  • 3
    @user68326 Regardless of how good at maths people here are, there's no need for personal attacks. I suggest you delete your offensive comments towards other users. On a side note: I sure hope people here are good at maths, because if they aren't, then I must be a very, very, very dumb person as I know close to nothing when compared to the average user here. – Git Gud Mar 24 '13 at 21:36
  • 2
    Who wants a Wolfram tee-shirt anyway? – Julien Mar 24 '13 at 22:15
  • @julien: I offer the Wolfram tee-shirt to you. Just send them an email with the issue. – user68326 Mar 24 '13 at 22:24

3 Answers3

41

Related problems: (I), (II), (III). Denoting our integral by $J$ and recalling the mellin transform $$ F(s)=\int_{0}^{\infty}x^{s-1} f(x)\,dx \implies F''(s)=\int_{0}^{\infty}x^{s-1} \ln(x)^2\,f(x)\,dx.$$

Taking $f(x)=\arctan(x)^2$, then the mellin transform of $f(x)$ is

$$ \frac{1}{2}\,{\frac {\pi \, \left( \gamma+2\,\ln\left( 2 \right) +\psi \left( \frac{1}{2}+\frac{s}{2} \right)\right) }{s\sin \left( \frac{\pi \,s}{2} \right)}}-\frac{1}{2}\,{\frac {{\pi }^{2}}{s\cos\left( \frac{\pi \,s}{2} \right) }},$$

where $\psi(x)=\frac{d}{dx}\ln \Gamma(x)$ is the digamma function. Thus $J$ can be calculated directly as

$$ J= \lim_{s\to -1} F''(s) = \frac{1}{12}\,\pi \, \left( 3\,{\pi }^{2}\ln \left( 2 \right) -{\pi }^{2}+24 \,\ln \left( 2 \right) -3\,\zeta \left( 3 \right) \right)\sim 6.200200824 .$$

11

The function is positive and continuous (except perhaps at the origin) with isolated zeros. The integral is therefore some nonzero, positive quantity.

That Wolfram | Alpha indicates otherwise is quite possibly an error. If you change the upper limit of the integral from $\infty$ to, say $5$, Wolfram | Alpha gives you a positive quantity, further indicating that there may be an error in the way the site is evaluating this integral.

10

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

$$ \mbox{With the identity}\quad\int_{0}^{1}{\dd y \over x^{2}y^{2} + 1} ={\arctan\pars{x} \over x}\quad\mbox{we'll have} $$

\begin{align} &\int_{0}^{\infty}\bracks{\ln\pars{x}\arctan\pars{x} \over x}^{2}\,\dd x =\int_{0}^{\infty}\ln^{2}\pars{x} \int_{0}^{1}{\dd y \over x^{2}y^{2} + 1}\int_{0}^{1}{\dd z \over x^{2}z^{2} + 1}\,\dd x \\[3mm]&=\int_{0}^{1}\int_{0}^{1} \int_{0}^{\infty} {\ln^{2}\pars{x} \over \pars{x^{2}y^{2} + 1}\pars{x^{2}z^{2} + 1}}\,\dd x \,\dd y\,\dd z \\[3mm]&=\int_{0}^{1}\int_{0}^{1}\braces{ {y^{-2}z^{-2} \over z^{-2} - y^{-2}} \bracks{\int_{0}^{\infty} {\ln^{2}\pars{x} \over x^{2} + y^{-2}}\,\dd x -\int_{0}^{\infty}{\ln^{2}\pars{x} \over x^{2} + z^{-2}}\,\dd x}}\,\dd y\,\dd z\tag{1} \end{align}

However, with $\ds{a > 0}$: \begin{align} &\int_{0}^{\infty}{\ln^{2}\pars{x} \over x^{2} + a^{-2}}\,\dd x =a\int_{0}^{\infty}{\ln^{2}\pars{x/a} \over x^{2} + 1}\,\dd x =a\int_{0}^{\infty}{\bracks{\ln\pars{x} - \ln\pars{a}}^{2} \over x^{2} + 1}\,\dd x \\[3mm]&=a\bracks{\int_{0}^{\infty}{\ln^{2}\pars{x} \over x^{2} + 1}\,\dd x -2\ln\pars{a}\ \overbrace{\int_{0}^{\infty}{\ln\pars{x} \over x^{2} + 1}\,\dd x}^{\ds{=\ 0}}\ +\ \ln^{2}\pars{a}\int_{0}^{\infty}{\dd x \over x^{2} + 1}} \\[3mm]&={\pi^{3} \over 8}\,a + {\pi \over 2}\,a\ln^{2}\pars{a}\tag{2} \end{align} since $\ds{\int_{0}^{\infty}{\ln\pars{x} \over x^{2} + 1}\,\dd x = {\pi^{3} \over 8}}$ is a well known result.

Replacing $\pars{2}$ in $\pars{1}$: \begin{align} &\int_{0}^{\infty}\bracks{\ln\pars{x}\arctan\pars{x} \over x}^{2}\,\dd x \\[3mm]&=\int_{0}^{1}\int_{0}^{1}\braces{{1 \over y^{2} - z^{2}} \bracks{{\pi^{3} \over 8}\,\pars{y - z} + {\pi \over 2}\,y\ln^{2}\pars{y} -{\pi \over 2}\,z\ln^{2}\pars{z}}}\,\dd y\,\dd z \\[3mm]&={\pi^{3} \over 8}\int_{0}^{1}\int_{0}^{1}{\dd y\,\dd z \over y + z} +{\pi \over 2}\int_{0}^{1}\int_{0}^{1} {y\ln^{2}\pars{y} - z\ln^{2}\pars{z} \over y^{2} - z^{2}}\,\dd y\,\dd z \end{align} Both integrals can be trivially evaluated: \begin{align} \int_{0}^{1}\int_{0}^{1}{\dd y\,\dd z \over y + z} &=2\ln\pars{2} \\[3mm]\int_{0}^{1}\int_{0}^{1} {z\ln^{2}\pars{z} - y\ln^{2}\pars{y} \over y^{2} - z^{2}}\,\dd y\,\dd z&= 4\ln\pars{2} - {1 \over 6}\,\pi^{2} - \half\,\zeta\pars{3} \end{align}

Then, \begin{align}&\color{#66f}{\large \int_{0}^{\infty}\bracks{\ln\pars{x}\arctan\pars{x} \over x}^{2}\,\dd x} ={\pi^{3} \over 8}\,\bracks{2\ln\pars{2}} +{\pi \over 2}\bracks{4\ln\pars{2} - {1 \over 6}\,\pi^{2} - \half\,\zeta\pars{3}} \\[3mm]&=\color{#66f}{\large{1 \over 4}\bracks{\ln\pars{2} - {1 \over 3}}\pi^{3} + \bracks{2\ln\pars{2} - {1 \over 4}\,\zeta\pars{3}}\pi} \approx 6.200200822 \end{align}

Felix Marin
  • 89,464