27

Show that $$\int_0^{\pi/3}\arccos^2(2\sin^2 x-\cos x)\mathrm dx=\frac{19\pi^3}{135}$$

Wolfram strongly suggests that it's true.

Here is the graph of $y=\arccos^2(2\sin^2 x-\cos x)$.

enter image description here

Context

I stumbled upon this when I was trying to answer a similar question: Show that $\int_0^{\pi/3}\arccos(2\sin^2 x-\cos x)\mathrm dx=\frac{\pi^2}{5}$.

My attempt

Using the first three substitutions from the answer to the similar question, I got:

$$\int_0^\frac{\pi}{3}\arccos^2(2\sin^2 x-\cos x)\mathrm dx$$

$$\overset{\large\tan\frac{x}{2}\to x}=2\int_0^\frac{1}{\sqrt3}\frac{\arccos^2\left(\frac{-1+8x^2+x^4}{(1+x^2)^2}\right)}{1+x^2}\mathrm dx$$

$$=8\int_0^\frac{1}{\sqrt 3}\frac{\arctan^2\left(\frac{1}{x}\sqrt{\frac{1-3x^2}{5+x^2}}\right)}{1+x^2}\mathrm dx$$

$$\overset{\large\frac{1}{x}\to\sqrt x}=4\int_3^\infty\frac{\arctan^2\left(\sqrt x\sqrt{\frac{x-3}{5x+1}}\right)}{\sqrt{x}(1+x)}\mathrm dx$$

$$\overset{\large \frac{x-3}{5x+1}\to x}=16\int_0^\frac{1}{5} \frac{1}{(1-x)(1-5x)} \frac{\arctan^2\left(\sqrt x\sqrt{\frac{3+x}{1-5x}}\right)}{\sqrt{\frac{3+x}{1-5x}}}\mathrm dx$$

But this doesn't seem to lead anywhere.

Dan
  • 35,053

2 Answers2

33

Computing the integral

We can start by rewriting the integral in terms of its inverse, similarly as seen in this answer.

$$I=\int_0^{\frac{\pi}{3}}\arccos^2(2\sin^2 x-\cos x) dx \overset{x \to \pi-x}= \int_{\frac{2\pi}{3}}^\pi\arccos^2(2\sin^2 x+\cos x) dx$$

$$\small =\pi^3-\int_0^{\pi^2} \arccos\left(\frac14-\frac14 \sqrt{17-8\cos \sqrt{y}}\right)dy\overset{y\to y^2}=\pi^3-2\int_0^{\pi} y\arccos\left(\frac14-\frac14 \sqrt{17-8\cos y}\right)dy$$

Next, we should notice that the $\arccos$ argument is the "$-$" solution of a quadratic equation with $b=-\frac12$ and $c=\frac{\cos y}{2}-1$ in $y_{\pm}=\frac{-b\pm\sqrt{b^2-4c}}{2}$. Using Vieta's relations we can easily simplify $\arccos y_-\pm\arccos y_+=\arccos\left(y_-y_+\mp\sqrt{(1+y_-y_+)^2-(y_-+y_+)^2}\, \right)$, and therefore, it makes sense to also consider the "$+$" solution like this:

$$\mathcal J_{\pm}=\int_0^\pi y\arccos\left(\frac14\pm\frac14 \sqrt{17-8\cos y}\right)dy$$

Here it should be noted that the "$+$" solution is a purely imaginary number, as the $\arccos$ argument is outside the $(-1,1)$ range, thus we can subtract it from the original integral and directly take the real part.

$$\Rightarrow I= \pi^3-2 \mathcal J_{-}=\pi^3-2\Re\left(\mathcal J_{-}-\mathcal J_{+}\right)= \pi^3-2\Re \int_0^\pi y \arccos\left(-1+\frac12e^{iy}\right)dy$$

$$=4\Re \int_0^\pi y\arcsin\left(\frac12 e^{iy/2}\right)dy=2\Re\sum_{n=0}^\infty \frac{(2n)!}{2^{4n} (n!)^2(2n+1)} \int_0^\pi y e^{iy(n+1/2)}dy$$

$$=4\pi \sum_{n=0}^\infty \frac{(-1)^n(2n)!}{2^{4n} (n!)^2 (2n+1)^2}-8 \sum_{n=0}^\infty \frac{(2n)!}{2^{4n} (n!)^2 (2n+1)^3}$$

$$=8\pi\int_0^1\frac{\operatorname{arcsinh}\left(\frac{y}{2}\right)}{y}dy+16 \int_0^1 \frac{\arcsin\left(\frac{y}{2}\right)\ln y}{y}dy$$

$$=\frac{2\pi^3}{5} - \frac{7\pi^3}{27}=\boxed{\frac{19\pi^3}{135}}$$

Above we used the power series of the $\arcsin$ function (twice) as well as the one for $\operatorname{arcsinh}$. Also, the first integral can be found here and the second one here (after integrating it by parts).


Generalization

It might be worth to mention that we can further attempt to generalize the integral as:

$$I(k)=\int_0^\frac{\pi}{3} \arccos^k\left(2\sin^2 x- \cos x \right) dx=2k\Re \int_0^\pi y^{k-1}\arcsin\left(\frac{1}2 e^{iy/2}\right)dy$$

$$=2k\sum_{n=0}^\infty \frac{(2n)!}{2^{4n} (n!)^2(2n+1)} \Re\left(\frac{1}{i^{k-1}}\frac{d^{k-1}}{dn^{k-1}}\frac{i+e^{in\pi}}{2n+1}\right)$$

For example, with $k=3$ we'll obtain:

$$\small I(3)=\int_0^\frac{\pi}{3} \arccos^3\left(2\sin^2 x- \cos x \right) dx=6\pi^2\sum_{n=0}^\infty \frac{(-1)^n(2n)!}{2^{4n} (n!)^2(2n+1)^2}-48\sum_{n=0}^\infty \frac{(-1)^n(2n)!}{2^{4n} (n!)^2(2n+1)^4}$$

$$\small =12\pi^2\int_0^1\frac{\operatorname{arcsinh}\left(\frac{y}{2}\right)}{y}dy-48 \int_0^1 \frac{\operatorname{arcsinh}\left(\frac{y}{2}\right)\ln^2 y}{y}dy=\frac{3\pi^4}{5}-48_5F_4\left(\begin{array}{c}\tfrac12,\tfrac12,\tfrac12,\tfrac12,\tfrac12\\ \tfrac32,\tfrac32,\tfrac32,\tfrac32\end{array}\middle|-\tfrac14\right)$$

Above there's a huge chance that the last hypergeometric function can be simplified, as we have the closed form for a similar one here:

$$\small _5F_4\left(\begin{array}{c}\tfrac12,\tfrac12,\tfrac12,\tfrac12,\tfrac12\\ \tfrac32,\tfrac32,\tfrac32,\tfrac32\end{array}\middle|\color{red}{\tfrac14}\right) = \frac{\pi}{12}\zeta(3)+\frac{27}{32}\operatorname{Cl}_4\big(\tfrac23\pi\big)$$

Update: Thanks to KStar's efforts, we can indeed obtain a closed form for $I(3)$.

$$I(3)=\boxed{\frac{4\pi^4}{15}-24\zeta(3)\ln(\varphi)+2\pi^2\ln(\varphi)^2-5\ln(\varphi)^4-15\operatorname{Li}_4\left(\frac{1}{\varphi^2}\right)}$$

However, starting from $I(4)$ it seems that the integrals are no longer reducible in closed form, without introducing new special functions.

$$\small I(4)=8\pi^3\sum_{n=0}^\infty \frac{(-1)^n (2n)!}{2^{4n} (n!)^2(2n+1)^2} - 192\pi\sum_{n=0}^\infty \frac{(-1)^n (2n)!}{2^{4n} (n!)^2(2n+1)^4} + 384\sum_{n=0}^\infty \frac{(2n)!}{2^{4n} (n!)^2(2n+1)^5}$$

$$\small = 16\pi^3 \int_0^1 \frac{\operatorname{arcsinh}\left(\frac{y}{2}\right)}{y} dy - 192\pi \int_0^1 \frac{\operatorname{arcsinh}\left(\frac{y}{2}\right) \ln^2 y}{y}dy - 128 \int_0^1 \frac{\arcsin\left(\frac{y}{2}\right) \ln^3 y}{y}dy $$

$$\small =\boxed{\frac{179}{243}\pi^5-96\pi \ln(\varphi)\zeta(3)+8\pi^3\ln^2(\varphi)-20\pi \ln^4(\varphi)-60\pi\operatorname{Li}_4\left(\frac{1}{\varphi^2}\right)-96\operatorname{Gl}_{4,1}\left(\frac{\pi}{3}\right)}$$

Above we introduced the multiple Glaisher function, as seen in $2$ and $5.1$ from this article. In our particular case there is $\operatorname{Gl}_{4,1}\left(\frac{\pi}{3}\right)=\sum_{n=1}^\infty \frac{H_{n-1}}{n^4}\sin\frac{n\pi}{3}$.

Zacky
  • 30,116
  • 7
    Incredible work! (+1) – David H Oct 02 '24 at 05:01
  • 3
    You really are the king of Ahmed-type integrals. (+1) – KStar Oct 02 '24 at 10:54
  • 3
    This is brilliant – polychroma Oct 03 '24 at 17:01
  • 3
    Mind blowing solution!!!(+1) – Martin.s Oct 07 '24 at 06:25
  • Thank you all for the nice words! – Zacky Oct 07 '24 at 08:44
  • 5
    A truly beautiful result: $$3\cdot {}_5F_4\left(\begin{array}{c}\tfrac12,\tfrac12,\tfrac12,\tfrac12,\tfrac12\ \tfrac32,\tfrac32,\tfrac32,\tfrac32\end{array}\middle|-\tfrac14\right)=\frac{\pi^4}{48}+\frac{3\zeta(3)}{2}\log(\varphi)-\frac{\pi^2}{8}\log(\varphi)^2+\frac{5}{16}\log(\varphi)^4+\frac{15}{16}\mathrm{Li}_4\left(\frac{1}{\varphi^2}\right),$$ where $\varphi :=\frac{1+\sqrt{5}}{2}$ is the golden ratio. I would be more than happy to prove this for you if interested! – KStar Oct 29 '24 at 18:29
  • @KStar very impressive! It would definitely be of great interest to see an approach of deriving that closed form, I'm sure your approach could serve as a reference for many other similar integrals in future. – Zacky Oct 30 '24 at 12:42
5

This is in specific response to the generalization mentioned in Zacky's (incredible) answer (I lack the necessary reputation to comment.)

Evaluating $$S_-=\sum_{n=1}^\infty\frac{\Gamma\left(n+\frac{1}{2}\right)}{(2n+1)^4\,(-4)^n\,n!}$$

should eventually give a very similar closed form for $$_5F_4\left(\begin{array}{c}\tfrac12,\tfrac12,\tfrac12,\tfrac12,\tfrac12\\ \tfrac32,\tfrac32,\tfrac32,\tfrac32\end{array}\middle|-\tfrac14\right)$$ as the one in that original question, although not without some effort.

I will be able to look at this more in a few days, but as of now I do not have the time, so I am putting this out here in case anyone wishes to try it themselves.

If someone wishes to make this a comment, somehow, I would be thankful.

  • This is just loose, actually there should some simpler sums that will result in the necessary function much easier... I just cannot think of one directly. – user_infty Oct 02 '24 at 05:19
  • 1
    Hey just to let you know, I have evaluated this series- see my comment on Zacky's answer above! – KStar Oct 29 '24 at 19:27