19
$\begingroup$

I came across the follwing integral: $$\int_0^{\pi/2}x^2\sqrt{\cot x} \ dx=\frac{\sqrt2}{8}\left( \frac{5\pi^{3}}{12}-\pi^2\ln2-\pi \ln^22 \right)$$ I try to do it like the following:

Consider $$I(a,b)=\int_0^{\pi} \frac{\cos ax}{\sin^b x}\ dx=2\int_0^{\pi/2} \frac{\cos 2ax}{\sin^b 2x}\ dx$$ Then $$I''(a,b)=-8\int_0^{\pi/2}x^2 \frac{\cos 2ax}{\sin^b 2x}\ dx$$ Let $a=1/2,b=1/2$ $$I''(1/2,1/2)=-8\int_0^{\pi/2}x^2 \frac{\cos x}{\sqrt{\sin 2x}}\ dx=-\frac{8}{\sqrt2}\int_0^{\pi/2}x^2\sqrt{\cot x} \ dx $$ Back to $I(a,b),\quad I(a,b) $can be expressed by beta fuction by Wolfram Mathematica. $$I(a,b)=\int_0^{\pi} \frac{\cos ax}{\sin^b x}\ dx=\frac{\pi \cdot 2^b\cdot\cos (\pi a/2)\cdot \Gamma(1-b) }{\Gamma(a/2-b/2+1) \cdot \Gamma(-a/2-b/2+1)} dx$$ Then we can get $I''(a,b)$ approach from other way. Finally,we can get $\int_0^{\pi/2}x^2\sqrt{\cot x} \ dx$,but it seems a little complex.

For a similar integral: $$\int_0^{\pi/2}x\cdot\tan^p x \ dx=\frac{\pi}{4\sin (p\pi/2)}\left(\Psi\left(\frac{1}{2}\right)-\Psi\left(\frac{1-p}{2}\right) \right)$$ The above integral can be solved by method of parametric development.Let $p=-\frac{1}{2}$,We can get $$\int_0^{\pi/2}x\sqrt{\cot x} \ dx=\frac{\pi\left(\pi-2\ln 2\right)}{4\sqrt2}$$

But to this one, it seems to be difficult with method of parametric development. Could you suggest any ideas how to prove this?

$\endgroup$
  • 3
    $\begingroup$ May I ask where you did find the result for the integral ? $\endgroup$ – Claude Leibovici Sep 10 '16 at 11:43
  • $\begingroup$ Mathematica confirms the numerical values for 50 digits at least. But I'm curious about the source of this result too $\endgroup$ – Yuriy S Sep 10 '16 at 11:55
  • $\begingroup$ @ClaudeLeibovici Hi, I got the integral from one of my friends with result, and he want to know how to prove it. $\endgroup$ – gcy-rolle Sep 10 '16 at 12:12
  • 1
    $\begingroup$ The Wikipedia Article on the Trigamma Function has some known values which seem to match up with the desired values of your attempt. $\endgroup$ – Jack Lam Sep 11 '16 at 5:43
  • $\begingroup$ Mathematica seems able to simplify your expression to yield the answer. $\endgroup$ – Sangchul Lee Sep 11 '16 at 5:44
6
$\begingroup$

A slightly easier form for $I(a, \frac{1}{2})$ is

$$I(a, \frac{1}{2}) = 2^{a-\frac{3}{2}} \beta\left(\frac{1+2a}{4}\right) (1 + \sin \pi a + \cos \pi a), $$

where $\beta(p) = \Gamma(p)^2/\Gamma(2p)$ is the central beta function. (This can be easily obtained by applying the Euler reflection formula and the Legendre duplication formula to OP's representation.) Now from the relation

$$ \frac{d}{dp} \log \beta(p) \bigg|_{p=\frac{1}{2}} = 2\psi(1/2) - 2\psi(1) = 2 \sum_{n=0}^{\infty} \left( \frac{1}{n+1} - \frac{1}{n+\frac{1}{2}} \right) = -4 \log 2 $$

and

$$ \frac{d^2}{dp^2} \log \beta(p) \bigg|_{p=\frac{1}{2}} = 2\psi^{(1)}(1/2) - 4\psi^{(1)}(1) = \frac{\pi^2}{3}, $$

we can compute $\frac{\partial^2 I}{\partial a^2}(\frac{1}{2}, \frac{1}{2})$ as follows:

\begin{align*} &\frac{\partial^2 I}{\partial a^2} (1/2, 1/2) \\ &= \frac{1}{4} \beta''(1/2) - \left(\frac{\pi}{2}-\log 2\right) \beta'(1/2) - \left(\frac{\pi^2}{2} + \pi\log 2-\log^2 2\right) \beta(1/2) \\ &= \frac{1}{4} \pi \left( 16\log^2 2 + \frac{\pi^2}{3} \right) \\ &\qquad - \left(\frac{\pi}{2}-\log 2\right) (-4\pi \log 2) - \left(\frac{\pi^2}{2} + \pi\log 2-\log^2 2\right) \pi \\ &= -\left(\frac{5\pi^3}{12} - \pi^2\log2 - \pi \log^2 2 \right). \end{align*}


I am also trying a more direct approach. Using contour integral, we can check that

$$ \int_{0}^{\frac{\pi}{2}} x^2 \sqrt{\cot x} \, \mathrm{d}x = \frac{1}{3\sqrt{2}} \left( \frac{\pi^3}{4} - \frac{3\pi}{2} \int_{0}^{1} \frac{\operatorname{artanh}^2 t}{\sqrt{t}} \, \mathrm{d}t \right). $$

So I am tackling the last integral, but have no good news at this point.

$\endgroup$
  • $\begingroup$ Excellent answer, Thank you very much@Sangchul Lee. This question really confused me for a long time. And your form of $I(a,1/2)$ makes it much easrier. $\endgroup$ – gcy-rolle Sep 11 '16 at 6:41
2
$\begingroup$

$\newcommand{\angles}[1]{\left\langle\,{#1}\,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{\mathrm{i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\Li}[1]{\,\mathrm{Li}_{#1}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ It's quite clear the OP already evaluated the integral $\ds{\color{#f00}{\int_{0}^{\pi/2}x^{2}\root{\cot\pars{x}}\,\dd x}}$. However, the OP is $\texttt{still looking for a simple evaluation}$ which start from $\ds{\,\mrm{I}\pars{a,b}\ \pars{~\mbox{see below}~}}$. Hereafter we presente a 'relatively simple' evaluation of it.

$\ds{\mrm{I}\pars{a,b} = {\pi\,2^{b}\,\cos\pars{\pi a/2}\Gamma(1 - b) \over \Gamma\pars{a/2 - b/2 + 1}\Gamma\pars{-a/2 - b/2 + 1}}\,,\qquad\Gamma\pars{\half} = \root{\pi}}$

\begin{align} \mrm{I}\pars{a,\half} & = \root{2}\pi^{3/2}\,\, {\cos\pars{\pi a/2} \over \Gamma\pars{3/4 + a/2}\Gamma\pars{3/4 - a/2}} \end{align}


Note that $$ \color{#f00}{\int_{0}^{\pi/2}x^{2}\root{\cot\pars{x}}\,\dd x} = -\,{\root{2} \over 8}\braces{2\bracks{\epsilon^{2}}\mrm{I}\pars{\half + \epsilon,\half}} $$ where \begin{align} \mrm{I}\pars{\half + \epsilon,\half} & = \pi^{3/2}\,\,{\cos\pars{\pi\epsilon/2} - \sin\pars{\pi\epsilon/2} \over \Gamma\pars{1 + \epsilon/2}\Gamma\pars{1/2 - \epsilon/2}} = \pi^{3/2}\,\,{\cos\pars{\pi\epsilon/2} - \sin\pars{\pi\epsilon/2} \over \pars{\epsilon/2}!\pars{-1/2 - \epsilon/2}!} \\[5mm] & = \pi^{3/2}\,{1 \over \pars{-1/2}!}{-1/2 \choose \epsilon/2} \bracks{\cos\pars{\pi\epsilon/2} - \sin\pars{\pi\epsilon/2}} \\[5mm] & = \pi\,{-1/2 \choose \epsilon/2} \bracks{\cos\pars{\pi\epsilon \over 2} - \sin\pars{\pi\epsilon \over 2}} \end{align}

We just need the binomial and 'the inside brackets term' expansion up to $\ds{\epsilon^{2}}$. The binomial expansion, up to order $\ds{\epsilon^{2}}$, is simple but laborious: It is simplified by using the Digamma value $\ds{\Psi\pars{1/2} = -\gamma - 2\ln\pars{2}}$. $\ds{\gamma}$: Euler-Mascheroni Constant.

$$ \left\{\begin{array}{rcl} \ds{-1/2 \choose \epsilon/2} & \ds{=} & \ds{1 - \ln\pars{2}\,\epsilon + \bracks{\half\,\ln^{2}\pars{2} - {\pi^{2} \over 12}}\epsilon^{2} + \,\mrm{O}\pars{\epsilon^{3}}} \\[3mm] \ds{\cos\pars{\pi\epsilon \over 2} - \sin\pars{\pi\epsilon \over 2}} & \ds{=} & \ds{1 - {\pi \over 2}\,\epsilon - {\pi^{2} \over 8}\,\epsilon^{2} + \,\mrm{O}\pars{\epsilon^{3}}} \end{array}\right. $$


\begin{align} &\color{#f00}{\int_{0}^{\pi/2}x^{2}\root{\cot\pars{x}}\,\dd x} \\[5mm] = &\ -\,{\root{2} \over 8}\,\times 2\times \pi\braces{% 1\times\pars{-\,{\pi^{2} \over 8}} + \bracks{-\ln\pars{2}}\pars{-\,{\pi \over 2}} + \bracks{\half\,\ln^{2}\pars{2} - {\pi^{2} \over 12}}\times 1} \\[5mm] & = \color{#f00}{{\root{2} \over 8}\bracks{{5\pi^{3} \over 12} - \pi^{2}\ln\pars{2} - \pi\ln^{2}\pars{2}}} \approx 0.8077 \end{align}

$\endgroup$
  • $\begingroup$ Thank you very much for your detailed answer @Felix Marin. But I don't know the meaning of $\left[\epsilon^{2}\right]$, can you explain it ? $\endgroup$ – gcy-rolle Sep 13 '16 at 16:10
  • $\begingroup$ $\left[\epsilon^{2}\right]\mathrm{f}\left(\epsilon\right)$ is a notation for the coefficient of $\epsilon^{2}$ in the $\,\mathrm{f}\left(\epsilon\right)$ expansion in powers of $\left(\epsilon\right)$. Thanks. $\endgroup$ – Felix Marin Sep 13 '16 at 20:17
  • $\begingroup$ @ Felix Marin Yeah, I got it. Thanks. $\endgroup$ – gcy-rolle Sep 14 '16 at 4:22
  • $\begingroup$ @gcy-rolle You're welcome. Glad to see it was useful for you. $\endgroup$ – Felix Marin Sep 14 '16 at 4:23

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.