12

Here's the integral that I would like to solve. Purely for recreational purposes:

$$I=\int_{0}^{\frac{\pi}{2}} \ln(1+\sin^3x)\text{d}x$$

Here's my shot at it. I would like to stick to this method if possible. Let $I(\alpha)$ be defined as follows:

$$I(\alpha) = \int_{0}^{\frac{\pi}{2}} \ln(1+\alpha\sin^3 x)\text{d}x$$

$$\implies I'(\alpha) = \int_{0}^{\frac{\pi}{2}} \frac{\sin^3x}{1+\alpha \sin^3x}\text{d}x$$

Let $t = \tan(\frac{x}{2})$, and we have

$$I'(\alpha) = \int_{0}^{1}\frac{\frac{8t^3}{(1+t^2)^3}}{1 + \alpha\frac{8t^3}{(1+t^2)^3}}\cdot\frac{2}{1+t^2}\text{d}t$$

I will now get rid of the fraction

$$I'(\alpha) = \int_{0}^{1}\frac{\frac{8t^3}{(1+t^2)^3}}{1 + \alpha\frac{8t^3}{(1+t^2)^3}}\cdot\frac{2}{1+t^2}\cdot\frac{(1+t^2)^3}{(1+t^2)^3}\text{d}t = \int_{0}^{1} \frac{16t^3}{(1+t^2)(t^6 + 3t^4 + 8\alpha t^3 + 3t^2 + 1)}\text{d}t$$

Perform partial fractions (which was so time consuming):

$$I'(\alpha) = \frac{-2}{\alpha}\int_{0}^{1} \left(\frac{(t^2+1)^2}{t^6+3t^4+8\alpha t^3+3t^2+1}-\frac{1}{1+t^2}\right)\text{d}t$$

The second integral is inverse tangent. How would I go about doing the first integral? It's a $4^{th}$-degree polynomial over a $6^{th}$ degree polynomial, but parts of the polynomials look kind of simple.

Thomas Andrews
  • 177,126
Moni145
  • 2,142
  • The denominator is $(t^2+1)^3+8\alpha t^3,$ so the roots of the denominator are where $$\left(t +\frac{1}{t}\right)^3=-8\alpha.$$ There will be two value $t$ for each of the three complex cube root of $-8\alpha.$ So you can use partial fractions. It’s gonna get ugly, thought. – Thomas Andrews Jul 09 '21 at 00:53
  • 1
    I would suggest using the fact that $I'(\alpha)$ was even about the point $x=\frac{\pi}{2}$ to extend the integral to $\frac{1}{2}\int_0^\pi \cdots$. That way, when you do the tangent half angle substitution the integral will be from $0$ to $\infty$. The reason this may be a good idea is because there are a lot of integral tricks for rational functions involving the substitution $t \leftrightarrow \frac{1}{t}$ – Ninad Munshi Jul 09 '21 at 01:23
  • @ThomasAndrews This was some nasty partial fractions work with $8$ equations $8$ unknowns. I did $\frac{16t^3}{(1+t^2)(t^6 + 3t^4 + 8\alpha t^3 + 3t^2 + 1} = \frac{At + B}{1+t^2} + \frac{Ct^5 + Dt^4 + Et^3 + Ft^2 + Gt + H}{t^6 + 3t^4 + 8\alpha t^3 + 3t^2 + 1}.$ I got $(A,B,C,D,E,F,G,H) = (0,\frac{-2}{\alpha},0,\frac{-2}{\alpha},0,\frac{-4}{\alpha},0,\frac{-2}{\alpha})$. I have factored out $\frac{-2}{\alpha}$. – Moni145 Jul 09 '21 at 01:26
  • Ah, this missing $-2/\alpha$ was why my sanity check was failing. I missed it. – Thomas Andrews Jul 09 '21 at 01:28
  • @NinadMunshi I will try this. – Moni145 Jul 09 '21 at 01:29
  • You can also try factoring $a^3+1=(a+1)(a^2-a+1)$ and use this factoring to split the logarithms. You can also try finding the roots of the quadratic to factor further? Then use this crude answer for your integral. – Тyma Gaidash Jul 09 '21 at 01:41
  • From introducing the substitution $x = \arcsin (u)$, one can obtain that $I = \frac{2}{3} {}{5} F{4} \left( \frac{1}{2}, \frac{2}{3},1,1,\frac{4}{3} ; \frac{5}{6}, \frac{7}{6}, \frac{3}{2}, \frac{3}{2} ; 1 \right) - \frac{5 \pi}{64} {}{5} F{4} \left( 1,1, \frac{7}{6}, \frac{3}{2}, \frac{11}{6} ; \frac{4}{3}, \frac{5}{3}, 2, 2 ; 1 \right)$ where $F$ is the hypergeometric function. – KStarGamer Jul 11 '21 at 14:06

1 Answers1

8

Not a finished answer, just the beginning of a variation of the $I(\alpha)$ approach from the OP.

Using $$I(\alpha)=\int_{0}^{\pi/2} \log(1+\alpha^3\sin^3x)\,dx$$

Then use $$1+z^3=(1+z)\left(1+\omega z\right) \left(1+\omega^2z\right)$$ where $\omega=e^{2\pi i/3}.$

Then, for $k=0,1,2,$ let$$I_k(\alpha)=\int\log(1+\omega^k\alpha\sin x)\,dx.$$

Then $$\begin{align}I_k’(\alpha)&=\omega^k\int_{0}^{\pi/2}\frac{\sin x}{1+\omega^k\alpha \sin x}\,dx\\ &=\omega^k\int_0^2 \dfrac{\frac{2t}{1+t^2}}{1+\omega^k\alpha\frac{2t}{1+t^2}}\cdot\dfrac2{1+t^2}\,dt\\ &=-\frac{2}{\alpha}\int_0^1\left(\frac{1}{t^2+2\omega^k\alpha t+1}-\frac{1}{1+t^2}\right)\,dt \end{align}$$

Now:

$$\int_0^1\frac{1}{1+t^2}\,dt =\arctan(t)\Bigg\vert_0^1=\frac{\pi}4.$$

So now you need to compute:

$$I_k’(\alpha)-\frac\pi{2\alpha}=-\frac2{\alpha}\int_{0}^1\frac{1} {t^2+2\omega^k\alpha t+1}\,dt$$

The roots of the denominator are $$r_k^{\pm}=-\omega^k\alpha \pm\sqrt{\omega^{2k}\alpha^2-1}$$

And we get the partial fractions for this integrand:

$$2\sqrt{\omega^{2k}\alpha^2-1}\left(\frac1{t-r_k^+}-\frac1{t-r_k^-}\right)$$

So

$$\int_0^1\frac{dt}{1+2\omega^k\alpha t+t^2}\\=2 \sqrt{\omega^{2k}\alpha^2-1} \left(\log(t-r_k^+)-\log(t-r_k^-)\right)\Bigg\vert_0^1\\ =2 \sqrt{\omega^{2k}\alpha^2-1} \left(\log(1-r_k^- )-\log(1-r_k^+)\right) $$

A cursory look indicates it might be tough to take the antiderivative of the resulting expression. Remember, the $r_k^\pm$ are functions of $\alpha.$

When $k=0$ and $\alpha=\cos\theta$ then $r_k^\pm=e^{\pm i\theta}$ and $$-\frac{2}{\alpha}\int_0^1\frac{dt}{1+2\alpha t+t^2} =4i\theta\tan(\theta)$$

But there isn’t any nice way that I can see to get $k=1,2,$ and not seeing much nice cancellation when summing, either.

Thomas Andrews
  • 177,126
  • Accepting as the answer for the nice approach, but... looks like this isn't getting me anywhere. I tried your method by factoring the $6^{th}$ degree polynomial and partial fractions, and am POSITIVE that integrating to get $I(\alpha)$ is impossible. So my method isn't working. I'm going to try both of the other approaches in the comments. – Moni145 Jul 09 '21 at 05:25
  • @Luna145 I feel like Tyma's approach is nice looking, because the $\ln$ breaks into two parts. One of them is $\ln(1+\sin x)$ but $1+\sin x = 2\sin^2(\frac{x}{2} + \frac{\pi}{4})$ so it comes down to $\ln(\sin^2 x)$ which can be done. The other side is $\ln(\sin^2 x - \sin x + 1)$ which perhaps you can try the parametric approach you were thinking about. It seems easier. I have a question : If I do manage to work out Tyma's approach, then I'm not sure it will answer this question, right? So in that case, should I leave pointers for you as comments? – Sarvesh Ravichandran Iyer Jul 09 '21 at 06:59
  • 1
    @TeresaLisbon You can go ahead and post it if you want, or leave comments. Doesn't matter to me either way. – Moni145 Jul 09 '21 at 07:14