Imagine we have a sphere. We will fix the orthogonal coordinate system $xyz$. First, we paint the lower hemisphere in some color(for lower I mean we place the center of sphere in the origin, and hemisphere with negative $y$-coordinate will be the lower). After that we rotate the sphere (randomly) and again paint the lower hemisphere. What is expected value for the number of such rotations for which sphere will be fully painted. First, I consider simpler case. Just circle. So, there is no information for distribution of random choose of rotation, so it will be $U[0;\pi]$ (uniform). So let $\xi \in U[0;\pi]$ the rotation angle (also, we have a freedom of choosing the clockwise rotation or not, but I couldn’t think of how I can handle it). Now, let $N_{\pi} =\{ min n : \sum_{i=0}^n \xi_i > \pi\}$. We can easily find the expected value of such random variable by recursion formula (I won’t provide it, since it is quite easy, and also we didn’t take into account choice of rotation direction). Even if I could solve the easier case. I don’t know how to apply it to original problem.
Asked
Active
Viewed 312 times
3
-
4Remark: this is equivalent to choosing points randomly on the sphere and asking how many points it takes before every hemisphere contains one of the points. – Greg Martin Aug 21 '21 at 01:10
-
But there are continuum points on sphere – BeesaFangirl DOTO Aug 21 '21 at 06:50
1 Answers
3
For a circle, the probability you need to paint exactly $n$ halves is $\frac{n-2}{2^{n-1}}$ for $n\ge 3$ and the expected number of painted halves is $5$. See the related questions Probability that n points on a circle are in one semicircle and Spinning a disk and painting one half - probability whole disk is painted after $n$ steps?
For a sphere, this was considered by Kevin Brown, where he gave the probability of not succeeding after $n$ halves (i.e. of all the points being in the same hemisphere, as pointed out by Greg Martin) of $\frac{n^2-n+2}{2^n}$. This would imply the probability of needing to paint exactly $n$ halves of $\frac{(n-3)(n-2)}{2^n}$ for $n\ge 4$ and an expected number of painted halves of $7$

Henry
- 157,058
-
I think I got missunderstanding. The given probability is of all the points get in one hemisphere. Then, you write "This would imply the probability of needing to paint exactly $n$ halves of". I don't get it. Also I can't understand the connection between this two problems. So we choose some amount of points and "throw" them on sphere. But how it is connected with random rotation + always painting lower hemisphere? – BeesaFangirl DOTO Aug 21 '21 at 12:12
-
(previous comment was too long, thats why I'm finishing my thoughts in two sections) At first, I thought that we have some amount of points (which we throw on sphere), which determine the "new-lower" hemisphere we should paint (i.e we rotate sphere so the points will be lying on lower hemisphere). But which amount we should take then? – BeesaFangirl DOTO Aug 21 '21 at 12:13
-
@BeesaFangirlDOTO If you have the probability that there is a hemisphere with all $n$ points, then in your version, that is also the probability that not all of the sphere is painted (@GregMartin's comment). So the probability it takes exactly $n$ points is the difference between the probability $n-1$ points are not enough and the probability $n$ points are not enough. You can then use that to find the mean. – Henry Aug 21 '21 at 12:52
-
Generalising to a $d$-sphere (the surface of a $d+1$-dimensional ball), the expectation is $2d+3$ – Henry Aug 21 '21 at 12:54