The other night before turning in, I read Tim Gowers’ post on the orbit-stabilizer theorem. It starts with an invitation to count the rotational symmetries of a cube. He suggests a natural approach: consider the number of ways a rotation could act on a face, on a vertex, on an edge, etc. A face, for instance, can be sent to any one of the six faces. Having chosen one to send it to, there are four places to send one of its corners. Once that’s decided the rotation is determined, so there are altogether rotations.
Similarly, there are eight places to send a vertex. Having chosen one, there are two places to send an adjacent vertex. Once that’s decided, the rotation is determined, so there are altogether rotations.
(A few details are deliberately suppressed in the examples above. In particular, you may want to convince yourself that a rotation followed by a rotation is a rotation and that each sequence of choices discussed above does indeed fix the rotation.)
The general principle is that when a (finite) group acts on a set , the number of elements to which can be sent (the order of its orbit) times the number of elements that fix (the order of its stabilizer) is equal to the order of :
Gowers gives at least four proofs, two of which are reproduced here.
Proof 1. Let be in the orbit of and define the set consisting of all such that . We show that . The result follows by letting run through .
Pick some and define a function (henceforth just ) that sends to . Now consider the function which sends to . One readily verifies that is the identity on and that is the identity on . Therefore, since is invertible, and the result follows.
The next proof is apparently quite common.
Proof 2. We show that the elements of are in one-to-one correspondence with the left cosets of . The naive choice which sends to works. First, it’s well defined- if , then for some , so . It’s also invertible: let send to . This is also well-defined- if then so that , i.e. . Then is the identity on and is the identity the set of left cosets of . Therefore, since is invertible, .