20

How can I find a closed form for the following integral $$\int_0^{\frac12}(2x - 1)^6\log^2(2\sin\pi x)\,dx?$$

user91500
  • 5,606

6 Answers6

11

Here is a proof of Cleo's claim that \begin{align} \int^\frac{1}{2}_0(2x-1)^6\ln^2(2\sin(\pi x))\ {\rm d}x=\boxed{\displaystyle\frac{11\pi^2}{360}+\frac{60}{\pi^4}\zeta^2(3)-\frac{720}{\pi^6}\zeta(3)\zeta(5)} \end{align} It doesn't take much to show that the integral is equivalent to $$\mathscr{J}=\frac{64}{\pi^7}\int^\frac{\pi}{2}_0x^6\ln^2(2\cos{x})\ {\rm d}x$$ Use the identity $\displaystyle\ln^2(2\cos{x})={\rm Re}\ln^2(1+e^{i2x})+x^2$ to get \begin{align} \color{red}{\Large{\mathscr{J}}} &=\frac{\pi^2}{72}+\frac{64}{\pi^7}{\rm Re}\int^\frac{\pi}{2}_0x^6\ln^2(1+e^{i2x})\ {\rm d}x\\ &=\frac{\pi^2}{72}+\frac{1}{2\pi^7}{\rm Im}\int^{-1}_1\frac{\ln^{6}{z}\ln^2(1+z)}{z}{\rm d}z\\ &=\frac{\pi^2}{72}-\frac{3}{\pi^6}\int^1_0\frac{\ln^5{z}\ln^2(1-z)}{z}{\rm d}z+\frac{10}{\pi^4}\int^1_0\frac{\ln^3{z}\ln^2(1-z)}{z}{\rm d}z-\frac{3}{\pi^2}\int^1_0\frac{\ln{z}\ln^2(1-z)}{z}{\rm d}z\\ &=\frac{\pi^2}{72}+\frac{720}{\pi^6}\sum^\infty_{n=1}\frac{H_n}{(n+1)^7}-\frac{120}{\pi^4}\sum^\infty_{n=1}\frac{H_n}{(n+1)^5}+\frac{6}{\pi^2}\sum^\infty_{n=1}\frac{H_n}{(n+1)^3}\\ &=\frac{\pi^2}{72}+\frac{720}{\pi^6}\left(\frac{\pi^8}{7560}-\zeta(3)\zeta(5)\right)-\frac{120}{\pi^4}\left(\frac{\pi^6}{1260}-\frac{1}{2}\zeta^2(3)\right)+\frac{6}{\pi^2}\left(\frac{\pi^4}{360}\right)\\ &=\boxed{\displaystyle\color{red}{\Large{\frac{11\pi^2}{360}+\frac{60}{\pi^4}\zeta^2(3)-\frac{720}{\pi^6}\zeta(3)\zeta(5)}}} \end{align} Generalized Euler sum $\sum^\infty_{n=1}\frac{H_n}{n^q}$

M.N.C.E.
  • 10,439
10

If you need only the closed form:

$$\frac{11}{60}\zeta(2)+\frac{60}{\pi^4}\zeta^2(3)-\frac{720}{\pi^6}\zeta(3)\zeta(5)$$

Cleo
  • 117
  • 8
    is this your account? – Norbert Dec 19 '13 at 14:05
  • 4
    Quote from bounty description: The question is widely applicable to a large audience. A detailed canonical answer is required to address all the concerns. – Norbert Dec 19 '13 at 15:32
  • 2
    may I ask... without being vulgar, How the hell did you come up with this. Cuz that is simply amazing – Sidharth Ghoshal Dec 22 '13 at 02:40
  • The term "closed form" is so misleading. This answer was probably generated using wolfram alpha, maybe? – Squirtle Dec 26 '13 at 08:45
  • "Remember, you are not locked into a single axiom system. You may invent your own, whenever you wish — just use your intuition and imagination. - Cleo" May I ask for an example where this helps you solve a certain problem? – Akiva Weinberger Aug 01 '14 at 23:05
4

I could have sworn when I began my answer there was a square in the integrand and not a 6 power on the 2x-1 term. Anyway, this is for $(2x-1)^{2}$. But, the same idea can be used on the 6 power, but it will be very long and messy. It may even make tech grunt and groan.

One way to go about it is to use the identity:

$\displaystyle \int_{0}^{\frac{\pi}{2}}x\cos^{p-1}(x)\sin(ax)dx=\frac{\pi}{2^{p+1}}\Gamma(p)\left[\frac{\psi(\frac{p+a+1}{2})-\psi(\frac{p-a+1}{2})}{\Gamma(\frac{p+a+1}{2})\Gamma(\frac{p-a+1}{2})}\right]$

Since you have the 2 inside your log term, you could split your integrand up and write as

$\displaystyle \int_{0}^{1/2}(2x-1)^{2}\left[\ln^{2}(2)+2\ln(2)\ln(\sin(\pi x))+\ln^{2}(\sin(\pi x))\right]dx$

Then, using the above identity, do each separately.

The first integration is straightforward. Take the middle one:

$\displaystyle 2\ln(2)\int_{0}^{1/2}(2x-1)^{2}\ln(\sin(\pi x))dx$

Let $\displaystyle u=2x-1, \;\ dx=1/2du$

$\displaystyle \ln(2)\int_{-1}^{0}u^{2}\ln(\cos(\pi u/2))du$

Now, let $\displaystyle t=\frac{\pi u}{2}, \;\ du=\frac{2}{\pi}dt$.

Due to symmetry, we can write the limits as:

$\displaystyle \left(\frac{2}{\pi}\right)^{3}\ln(2)\int_{0}^{\pi/2}t^{2}\ln(\cos(t))dt$

Now, using the identity I mentioned at the top, diff once w.r.t p, then once w.r.t a.

Then, let p=1 and a=0. As I said, diffing the right side of the above identity may prove messy. So, use some sort of tech. For the last integral on the end, you can make the same subs to get the pi/2 limit, but then diff twice w.r.t p in order to get the squared log term.

So, for $\displaystyle \int_{0}^{\frac{\pi}{2}}x^{2}\ln(\cos(x))dx=\frac{-\pi}{4}\zeta(3)-\frac{\ln(2)}{24}\pi^{3}$

Using this method with that 6 will be brutal...even with tech. So, how about keeping the 2?. As Norbert alluded, 'getting off the horse in the middle of the stream' can be a little annoying.

BTW, that link that Norbert provided is very nice. It outlines how to handle log trig integrals using contours. You do not see this in many places. Check it out.

Cody
  • 14,054
3

WARNING: This is the answer for the previous version of the question

After substittion $s=\frac{\pi}{2}-\pi x$ we get that our integral equals to $$ I=\frac{4}{\pi^3}\int_0^{\pi/2} s^2\log^2(2\cos s)ds $$ From this answer it follows that $$ I=\frac{4}{\pi^3}\frac{11\pi^5}{1440}=\frac{11\pi^2}{360} $$

Norbert
  • 56,803
1

$\newcommand{\+}{^{\dagger}}% \newcommand{\angles}[1]{\left\langle #1 \right\rangle}% \newcommand{\braces}[1]{\left\lbrace #1 \right\rbrace}% \newcommand{\bracks}[1]{\left\lbrack #1 \right\rbrack}% \newcommand{\ceil}[1]{\,\left\lceil #1 \right\rceil\,}% \newcommand{\dd}{{\rm d}}% \newcommand{\ds}[1]{\displaystyle{#1}}% \newcommand{\equalby}[1]{{#1 \atop {= \atop \vphantom{\huge A}}}}% \newcommand{\expo}[1]{\,{\rm e}^{#1}\,}% \newcommand{\fermi}{\,{\rm f}}% \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,}% \newcommand{\half}{{1 \over 2}}% \newcommand{\ic}{{\rm i}}% \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow}% \newcommand{\isdiv}{\,\left.\right\vert\,}% \newcommand{\ket}[1]{\left\vert #1\right\rangle}% \newcommand{\ol}[1]{\overline{#1}}% \newcommand{\pars}[1]{\left( #1 \right)}% \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}}% \newcommand{\root}[2][]{\,\sqrt[#1]{\,#2\,}\,}% \newcommand{\sech}{\,{\rm sech}}% \newcommand{\sgn}{\,{\rm sgn}}% \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}}% \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert}$ $\ds{% {\cal I} \equiv \int_{0}^{1/2}\left(2x - 1\right)^{6}\ \log^{2}\left(2\sin\left(\pi x\right)\right)\,{\rm d}x}$

\begin{align} {\cal I}&=2^{6}\int_{-1/2}^{0}x^{6}\ \bracks{% \log\pars{2} + \log\pars{\cos\left(\pi x\right)}}^{2}\,{\rm d}x \\[3mm]&={1 \over \pi}\,\pars{2 \over \pi}^{6}\int_{0}^{\pi/2}x^{6}\ \bracks{% \log^{2}\pars{2} + 2\log\pars{2}\log\pars{\cos\left(x\right)} + \log^{2}\pars{\cos\left(x\right)}}\,{\rm d}x \\[3mm]&= {\log^{2}\pars{2} \over \pi}\,\pars{2 \over \pi}^{6}\int_{0}^{\pi/2}x^{6}\,\dd x + \log\pars{2}\pars{2 \over \pi}^{7}\int_{0}^{\pi/2}x^{6}\log\pars{\cos\pars{x}}\,\dd x \\[3mm]&\phantom{=}+ {1 \over \pi}\,\pars{2 \over \pi}^{6} \int_{0}^{\pi/2}x^{6}\log^{2}\pars{\cos\pars{x}}\,\dd x \\[3mm]&= {\log^{2}\pars{2} \over 14} + \log\pars{2}\pars{2 \over \pi}^{7}\ \overbrace{\int_{0}^{\pi/2}x^{6}\log\pars{\cos\pars{x}}\,\dd x} ^{\ds{\equiv\ {\cal J}_{1}}}\ +\ {1 \over \pi}\,\pars{2 \over \pi}^{6}\ \overbrace{\int_{0}^{\pi/2}x^{6}\log^{2}\pars{\cos\pars{x}}\,\dd x} ^{\ds{\equiv\ {\cal J}_{2}}} \end{align}

${\cal J}_{1}$ is evaluated by Mathematica in terms of the logarithm and zeta functions. In addition, it yields the numerical value ${\cal J}_{1} = -7.25681$. However, for ${\cal J}_{2}$, it just yields the numerical value ${\cal J}_{2} = 20.6214$

Felix Marin
  • 89,464
-3

I tried the partial integration, with

$\frac{d \log^2(2\sin(\pi x))}{dx} = \frac{1}{3}\log^3(2\sin(\pi x))\frac{1}{\frac{1}{2\sin(\pi x)}2cos(\pi x)\pi} = \frac{1}{3}\log^3(2\sin(\pi x))\frac{sin(\pi x)}{\pi cos(\pi x)} = \frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}$

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

$[(2x-1)^2\frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}]_0^{\frac{1}{2}} - \int_0^{\frac{1}{2}}(4x-2)\log^2(2\sin(\pi x))\,dx =$

$[(2x-1)^2\frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}]_0^{\frac{1}{2}} + [(2 - 4x)\frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}]_0^{\frac{1}{2}} + 4\int_0^{\frac{1}{2}}\log^2(2\sin(\pi x))\,dx =$

$[(2x-1)^2\frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}]_0^{\frac{1}{2}} + [(2 - 4x)\frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}]_0^{\frac{1}{2}} + 4[\frac{\log^3(2\sin(\pi x))sin(\pi x)}{3\pi cos(\pi x)}]_0^{\frac{1}{2}}$

  • of course it's possible I messed it up somewhere
  • with $\cos(\frac{\pi}{2})$ you have a $0$ in the denominator which makes it rather hard to evaluate the integral.