I need to compute for a positive integer the following integral: $$I=\int\limits_{S^2}(x^k+y^k+z^k)dS$$ where $S^2$ is the unit sphere in $\mathbb{R}^3$.

I'll decribe my attempt.

First, by symmetry, $$\int\limits_{S^2}(x^k+y^k+z^k)dS=3\int\limits_{S^2}z^kdS$$ (The right integral is the same for a replacement of $z$ with any of the other 2 variables) - I'm quite certain about that.

Now, we can use the map $r: (0, 2\pi)\times (0, \pi)\rightarrow S^2$ defined by shperical coordinates: $$r(\varphi, \theta)=(\cos\varphi \sin\theta, \sin\varphi \sin\theta, \cos\theta)$$ This map covers all the sphere, except for a neasure-zero set w.r.t to $S^2$.
In addition, our lecturer showed us that $$\sqrt{\Gamma \left ( \frac{\partial r}{\partial \varphi}, \frac{\partial r}{\partial \theta} \right )}= \sin^2\theta$$ so I have no doubts about that either.
Therefore, by definition of integration on manifolds, $$I=3\int _{(0, 2\pi)\times (0, \pi)}(\cos\theta)^k(\sin\theta)^2 d\varphi d\theta=3\cdot 2\pi \int \limits_{0}^{\pi}(\cos\theta)^k(\sin\theta)^2 d\theta$$

And that's not right. It can be seen thorugh plugging $k=2$. $I$ should be the surface area of the unit sphere which is $4\pi$, but the obtained result gives something far different.

Where did I go wrong?


I think you've just missed a square root: the area measure on the unit sphere is $\sin{\theta } \, d\theta \, d\phi$.

Note that $$r_{\theta} = (\cos{\phi}\cos{\theta},\sin{\phi}\cos{\theta},-\sin{\theta}),\\ r_{\phi} = (-\sin{\phi}\sin{\theta},\cos{\phi}\sin{\theta},0), $$ so $g_{\theta\theta} = 1$, $g_{\theta\phi}=g_{\phi\theta}=0$, $g_{\phi\phi} = \sin^2{\theta}$, and so $\sqrt{\det{g}} = \sin{\theta}$ since $0<\theta<\pi$.

  • $\begingroup$ Oh dear, you're right. Thank you! $\endgroup$ – 35T41 May 18 '17 at 20:04
  • $\begingroup$ As confirmation of that, this changes the linked integral to $2/3$ which rectified the error. (Incidentally, the resulting integral is actually easier than the original since $\frac{d}{d\theta}\cos\theta=-\sin \theta$.) $\endgroup$ – Semiclassical May 18 '17 at 20:07

I would avoid spherical coordinates and just apply Cavalieri's principle. Given some $a\in[-1,1]$ the section $S^2\cap\{z=a\}$ is a circle with length $2\pi\sqrt{1-a^2}$, hence

$$ I(k)=\int_{S^2}x^k+y^k+z^k\,d\mu = 3\int_{-1}^{1}2\pi\sqrt{1-a^2}a^k\,da $$ clearly equals zero if $k$ is odd, while assuming $k=2m$ we have: $$ I(k)=12\pi \int_{0}^{1}\sqrt{1-a^2}a^{2m}\,da = 6\pi\int_{0}^{1}(1-u)^{1/2} u^{m-1/2}\,du = 6\pi\cdot\frac{\Gamma\left(\frac{3}{2}\right)\Gamma\left(m+\frac{1}{2}\right)}{\Gamma(m+2)}$$ by Euler's Beta function.


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.