6
$\begingroup$

While investigating the function $$A(z)=\int_0^\frac{\pi}{2} \frac{\sin(zx)}{\sin(x)}dx$$ I stumbled upon the integral $$\int_0^{\frac{\pi}{2}}x^{2n+1}\cot(x)dx$$ when attempting to calculate the taylor series of $A(z)$ at $z=1$. As the coefficients of the even powers in the series reduce to integrating over a polynomial which is fairly trivial, the only real problem I have is in determining the the coefficients of the odd powers as I cannot seem to find a pattern between the coefficients.

Wolfram Alpha evaluates the first couple of integrals as:

\begin{align*} \int_0^\frac{\pi}{2} x\cot(x)dx&=\frac{\pi\ln(2)}{2}\\ \int_0^\frac{\pi}{2}x^3\cot(x)dx&=\frac{1}{16}(\pi^3\ln(4)-9\pi\zeta(3))\\ \int_0^\frac{\pi}{2}x^5\cot(x)dx&=\frac{1}{64}(-3\pi^3\zeta(3)+225\pi\zeta(5)+\pi^5\ln(4)) \end{align*}

and in general it seems that higher powers could also be calculated in terms of the zeta function, multiples of $\pi$, and $\ln(2)$. So far I have been unsuccessful in determining a pattern for these integrals but if anyone has any ideas I would be very grateful for any help on this.

$\endgroup$
6
$\begingroup$

Denote your integral as $\mathfrak{I}(n)$ and apply IBP by choosing $u=x^{2n+1}$ and $\mathrm dv=\cot(x)\mathrm dx$ to get

\begin{align*} \mathfrak{I}(n)&=\int_0^{\pi/2}x^{2n+1}\cot(x)\mathrm dx =\underbrace{\left[(2n+1)\cdot x^{2n}\log(\sin x)\right]_0^{\pi/2}}_{\to0}-(2n+1)\int_0^{\pi/2}x^{2n}\log(\sin x)\mathrm dx\\ &=-(2n+1)\int_0^{\pi/2}x^{2n}\log(\sin x)\mathrm dx \end{align*}

Now utilizing the well-known Fourier series expansion of $\log(\sin x)$, which converges within $[0,\pi]$, and switching the order of summation and integration further gives us

\begin{align*} \mathfrak{I}(n)&=-(2n+1)\int_0^{\pi/2}x^{2n}\log(\sin x)\mathrm dx\\ &=-(2n+1)\int_0^{\pi/2}x^{2n}\left[-\log(2)-\sum_{k=1}^\infty\frac{\cos(2kx)}k\right]\mathrm dx\\ &=\log(2)\left(\frac\pi2\right)^{2n+1}+(2n+1)\sum_{k=1}^\infty\frac1k\underbrace{\int_0^{\pi/2}x^{2n}\cos(2kx)\mathrm dx}_{=J} \end{align*}

The integral $J$ can be computed via IBP again which explains the connection to values of the Riemann Zeta Function hence for integer $n$ every IBP step produces another reciprocal power of $n$ which overall combines to sums that can be expressed with the help of the Riemann Zeta Function.

As one may see the values for $n=0$ and $n=1$ can be easily verfied since for $n=0$ $J$ is overall $0$ aswell whereas for $n=1$ the latter integral can be expressed using the Dirichlet Eta Function. To be precise we got

\begin{align*} n=0:~~~\mathfrak{I}(0)&=\log(2)\left(\frac\pi2\right)^{1}+(1)\sum_{k=1}^\infty\frac1k\underbrace{\int_0^{\pi/2}\cos(2kx)\mathrm dx}_{=0}\\ &=\frac{\pi\log(2)}2 \end{align*}

\begin{align*} n=1:~~~\mathfrak{I}(1)&=\log(2)\left(\frac\pi2\right)^{3}+(2+1)\sum_{k=1}^\infty\frac1k\int_0^{\pi/2}x^2\cos(2kx)\mathrm dx\\ &=\log(2)\left(\frac\pi2\right)^{3}+3\sum_{k=1}^\infty\frac1k\left[\frac\pi4\frac{\cos(\pi k)}{k^2}\right]_0^{\pi/2}\\ &=\log(2)\left(\frac\pi2\right)^{3}-\frac{3\pi}4\sum_{k=1}^\infty\frac{(-1)^{k+1}}{k^3}\\ &=\log(2)\left(\frac\pi2\right)^{3}-\frac{3\pi}4\eta(3)\\ &=\log(2)\left(\frac\pi2\right)^{3}-\frac{9\pi}{16}\zeta(3)\\ &=\frac1{16}(\pi^3\log(4)-9\pi\zeta(3)) \end{align*}

Note that we used the relation $\eta(s)=(1-2^{1-s})\zeta(s)$. Similiar can be done for all integer $n$. So as at least close to a closed-form I can offer the following formula

$$\therefore~\mathfrak{I}(n)~=~\log(2)\left(\frac\pi2\right)^{2n+1}+(2n+1)\sum_{k=1}^\infty\frac1k\int_0^{\pi/2}x^{2n}\cos(2kx)\mathrm dx$$

$\endgroup$
  • 2
    $\begingroup$ I have discussed a generalised version of these integrals in this question. Using your method I have obtained (detailed in my answer to this question) $$\mathfrak{I}(n) = \frac{(2n+1)!}{2^{2n+1}} \sum \limits_{l=0}^n (-1)^l \frac{\pi^{2(n-l)+1}}{(2(n-l)+1)!} \eta(2l+1) \, , \, n \in \mathbb{N}_0 \, , $$ for this special case. $\endgroup$ – ComplexYetTrivial Feb 5 at 20:05
  • $\begingroup$ @ComplexYetTrivial I have to thank you for the given links aswell as for your formula which is even more a closed-form than mine! I was not sure how to evaluate the given integral in a general form without invoking some kind of recursion formula. $\endgroup$ – mrtaurho Feb 5 at 21:52
1
$\begingroup$

Continuing off of @mrtaurho's excellent answer, we may find another form for $$C_k(n)=\int_0^{\pi/2}x^{2n}\cos(2kx)\mathrm dx$$ for $n\in \Bbb N$. First, we note that $$C_k(n)=\frac1{(2k)^{2n+1}}\int_0^{k\pi}x^{2n}\cos(x)\mathrm dx$$ Then we integrate by parts with $\mathrm dv=\cos(x)\mathrm dx$: $$C_k(n)=\frac1{(2k)^{2n+1}}x^{2n}\sin(x)\big|_0^{k\pi}-\frac{2n}{(2k)^{2n+1}}\int_0^{k\pi}x^{2n-1}\sin(x)\mathrm dx$$ $$C_k(n)=-\frac{2n}{(2k)^{2n+1}}\int_0^{k\pi}x^{2n-1}\sin(x)\mathrm dx$$ IBP once again, $$C_k(n)=-\frac{2n}{(2k)^{2n+1}}\left[-x^{2n-1}\cos(x)\big|_0^{k\pi}+(2n-1)\int_0^{k\pi}x^{2n-2}\sin(x)\mathrm dx\right]$$ $$C_k(n)=(-1)^k\frac{2n(k\pi)^{2n-1}}{(2k)^{2n+1}}-\frac{2n(2n-1)}{(2k)^{2n+1}}C_k(n-1)$$ $$C_k(n)=(-1)^k\frac{n\pi^{2n-1}}{2^{2n}k^2}-\frac{2n(2n-1)}{(2k)^{2n+1}}C_k(n-1)$$ So we have that $$\mathfrak{I}(n)=\left(\frac\pi2\right)^{2n+1}\log2-\frac{n(2n+1)}{4^n}\pi^{2n-1}\eta(3)-\frac{n(2n+1)(2n-1)}{4^n}\sum_{k\geq1}\frac{C_k(n-1)}{k^{2n+2}}$$ Which doesn't seem to give any sort of recurrence relation... :(

If I think of any new approaches I'll update my answer.

$\endgroup$

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.