38

Inspired by Mr. Olivier Oloa in this question. Does the following integral admit a closed form?

\begin{align} \mathscr{R}=\int_0^{\Large\frac{\pi}{2}}\sin^2x\,\ln\big(\sin^2(\tan x)\big)\,\,dx \end{align}

It will be my last question before I take a long break from my activity on Mathematics StackExchange. So, please be nice. No more downvotes for no reason because this is a challenge problem.

Edit :

I am also interested in knowing the numerical value of $\mathscr{R}$ to the precision of at least $50$ digits. If you use Mathematica to find its numerical value, please share your method & the code.

Archer
  • 6,051
  • Using Matlab, the result seems to be $\mathscr{R}\approx -0.825490903211824$ at least that was the result of my computation (just as a first check, since this was already different from @Omran Kouba's result). – Thomas Nov 07 '14 at 18:31
  • 9
    Guys, what are the downvotes for? – Gennaro Marco Devincenzis Nov 09 '14 at 12:42
  • 2
    Please don't create a meta tag "challenge problem". Meta tags are very strongly discouraged in general - see http://math.stackexchange.com/help/tagging . There is a thread on meta about this one at http://meta.math.stackexchange.com/a/17275/630 – Carl Mummert Nov 09 '14 at 17:29
  • 7
    Prof. @CarlMummert But WHY??? I have a privilege for doing so. Lots of questions on Math SE can be covered by using this tag. I think it is about time we have this tag on Math SE. Why don't you guys just leave me alone? Please don't bother my posts again. They're all legit. I didn't do something wrong here – Anastasiya-Romanova 秀 Nov 09 '14 at 17:46
  • 11
    @Anastasiya-Romanova Please try to keep a professional tone on this site. People on this site aren't out to get you. – MT_ Nov 12 '14 at 18:00
  • I have changed the formatting of the title so as to make it take up less vertical space -- this is a policy to ensure that the scarce space on the main page is distributed evenly over the questions. See here for more information. Please take this into consideration for future questions. Thanks in advance. – GNUSupporter 8964民主女神 地下教會 Mar 12 '18 at 16:55

5 Answers5

29

Here's a slightly different approach. Using the power reduction formula and getting rid of the second exponent of the sine using the properties of the logarithm the integral becomes:

$$I=\int_0^{\frac{\pi}{2}} (1-\cos 2x) \log (\sin (\tan x))dx$$

Using the substitution $2x=u$ we have:

$$I = \frac{1}{2}\int_0^\pi (1-\cos u) \log\bigg(\sin \bigg(\tan \bigg(\frac{u}{2}\bigg)\bigg)\bigg)du$$

We have the integral ready for a Weierstrass substitution, after which it becomes:

$$I = \int_0^\infty\bigg(1-\frac{1-t^2}{1+t^2}\bigg) \log(\sin (t)) \frac{1}{1+t^2}dt$$

Or:

$$I = \int_0^\infty \frac{2t^2}{(1+t^2)^2} \log(\sin (t))dt$$

From now on I'll use $x$ again. The Fourier series of $\log (\sin x)$ is well known and it is:

$$\log(\sin x)= -\log 2 -\sum_{n=1}^\infty \frac{\cos(2nx)}{n}$$

So the integral, exchanging integration and summation, becomes:

$$I=-2\log 2 \int_0^\infty \frac{x^2}{(1+x^2)^2}dx-2\sum_{n=1}^\infty \frac{1}{n} \int_0^\infty \frac{x^2\cos(2nx)}{(1+x^2)^2}$$

These are both easy integrals from the point of view of residue calculus. The final result is:

$$I=-\frac{\pi \log 2}{2} -\frac{\pi}{2}\sum_{n=1}^\infty \frac{e^{-2n}}{n}+\pi \sum_{n=1}^\infty e^{-2n}$$

These sums can be evaluated using the geometric series and its integral. So we have:

$$I=-\frac{\pi \log 2}{2}+\frac{\pi}{e^2-1}-\frac{\pi}{2}(2-\log(e^2-1))$$

Simplifying:

$$I=\frac{\pi}{2}\log \bigg( \frac{e^2-1}{2} \bigg) +\pi\bigg(\frac{2-e^2}{e^2-1}\bigg)$$

21

The answer is $$ \mathscr{R}=\frac{\pi }{2} \left(\log \left(\frac{e^2-1}{2} \right)-\frac{2 \left(e^2-2\right)}{e^2-1}\right) $$ As Kirill proved we have $$\eqalign{ \mathscr{R}&=\frac{1}{2}\int_{-\infty}^\infty\frac{u^2}{(1+u^2)^2}\log(\sin^2u)du\cr &=\frac{1}{2}\int_{0}^\pi\left(\sum_{k\in\mathbb{Z}}\frac{(u+k\pi)^2}{(1+(u+k\pi)^2)^2}\right)\log(\sin^2u)du } $$ Now, the function $$ F(u)=\sum_{k\in\mathbb{Z}}\frac{(u+k\pi)^2}{(1+(u+k\pi)^2)^2} $$ is $\pi$-periodic and even function. It is not difficult to calculate its Fourier cosine coefficients $a_n$ such that $$ F(u)=\frac{a_0}{2}+\sum_{n=1}a_n\cos(2n u) $$ with, $$a_n=\frac{2}{\pi}\int_0^\pi F(u)\cos(2n u)du =\frac{2}{\pi}\int_{-\infty}^\infty \frac{u^2}{(1+u^2)^2}\cos(2n u)du= e^{-2 n} (1-2 n)$$ The last equality is obtained by a simple residue calculus.

On the other hand it is easy and well-known that $$ \log(\sin^2u)=-2\log 2-\sum_{n=1}^\infty\frac{2}{n}\cos(2nu) $$ So using Parseval's formula we get $$ \mathscr{R}=\frac{\pi}{2}\left(- \log 2-\frac{1}{2}\sum_{n=1}^\infty\frac{2}{n}e^{-2 n} (1-2 n)\right) $$ and this simplifies easily to the announced closed form.$\qquad\square$

Omran Kouba
  • 28,772
13

Write $\tan x = u$ to get the integral $$ \int_0^\infty \frac{u^2\,du}{ (1+u^2)^2} \log(\sin^2 u) $$ and split over periods of length $\pi$ by writing $u=\pi k+s$, $0<s<\pi$ so that the integral is $$ 2\int_0^\pi \sum_{k\geq0} \frac{u^2}{(1+u^2)^2}\log\sin s\,ds, \qquad u = \pi k+s. $$ The sum can be done explicitly in terms of polygamma functions: $$ \frac{1}{2\pi}\Im\psi\left(\frac{i+s}{\pi}\right) + \frac{1}{2\pi^2}\Re\psi_1\left(\frac{i+s}{\pi}\right), $$ so the integral is equal to $$ \frac1\pi \int_0^\pi \log(\sin s)\left(\Im\psi\left(\frac{i+s}{\pi}\right) + \frac{1}{\pi}\Re\psi_1\left(\frac{i+s}{\pi}\right) \right)\,ds$$ which is numerically $$ \begin{array}{rl} -0.8254932940&1920795045&3494583393&8145490721&3051472153&0015143141\\ 0263463153&8662518683&0960012709&2734877933&9171668805&2198716476\\ 0581876961&2557665495&3473838059&8389072188&9187974995&4963384740\\ 9429563810&6463831818&4148444098&6729534027&3239373746&3514130065&\ldots \end{array} $$

Kirill
  • 14,494
2

$$I = \int_{0}^{\frac{\pi}{2}} \sin^2(x) \ln(\sin^2(\tan(x))) \, dx \\ = \underbrace{2 \int_{0}^{\infty} \frac{t^2}{(1 + t^2)^2} \ln(\sin(t)) \, dt}_{\text{Let } t = \tan(x)} \\$$ $$= 2 \int_{0}^{\infty} \frac{t^2}{(1 + t^2)^2} \left(-\ln(2) - \sum_{n=1}^{\infty} \frac{\cos(2nt)}{n}\right) \, dt$$ $$= -2 \ln(2) \int_{0}^{\infty} \frac{t^2}{(1 + t^2)^2} \, dt - 2 \sum_{n=1}^{\infty} \frac{1}{n} \int_{0}^{\infty} \frac{t^2 \cos(2nt)}{(1 + t^2)^2} \, dt$$ $$= -\frac{\pi}{2} \ln(2) - \frac{1}{2} \sum_{n=1}^{\infty} e^{-2n} \left(\pi - 2\pi n\right) = -\frac{\pi}{2} \ln(2) + \frac{\pi}{e^2 - 1} - \frac{\pi}{2} \left(2 - \ln(e^2 - 1)\right)$$

Sbsty
  • 413
  • 11
1

After the substitution $\tan{x}\mapsto x$, we get \begin{align} \mathscr{R} =&\int^\infty_0\frac{x^2}{(1+x^2)^2}\ln(\sin^2{x})\ {\rm d}x\\ =&\Re\int^\infty_{-\infty}\frac{x^2\ln(1-e^{i2x})}{(1+x^2)^2}{\rm d}x-\ln{2}\underbrace{\int^\infty_{-\infty}\frac{x^2}{(1+x^2)^2}{\rm d}x}_{\frac{\pi}{2}} \end{align} Even though the function $\displaystyle f(z)=\frac{z^2\ln(1-e^{i2z})}{(1+z^2)^2}$ has infinitely many branch points at $z=n\pi$, once we close the contour along the upper half of $|R|$ and make semicircular bumps around the branch points, one may check (by letting $z=n\pi+ \epsilon e^{i\theta}$) that the contribution along the bumps vanishes. The integral along the big arc also tends to $0$. Hence \begin{align} \mathscr{R} =&2\pi i{\rm Res}(f,i)-\frac{\pi}{2}\ln{2}\\ \end{align} Using WolframAlpha to compute the residue and simplify terms, $$\mathscr{R}=\frac{\pi}{e^2-1}-\pi-\frac{\pi}{2}\ln{2}+\frac{\pi}{2}\ln(e^2-1)$$

Sbsty
  • 413
  • 11