Cauchy's theorem on subgroup existence

https://arbital.com/p/cauchy_theorem_on_subgroup_existence

by Patrick Stevens Jun 18 2016 updated Jun 30 2016

Cauchy's theorem is a useful condition for the existence of cyclic subgroups of finite groups.


Cauchy's theorem states that if G is a finite Group and p is a prime dividing |G| the order of G, then G has a subgroup of order p. Such a subgroup is necessarily cyclic (proof).

Proof

The proof involves basically a single magic idea: from thin air, we pluck the definition of the following set.

Let X={(x1,x2,,xp):x1x2xp=e} the collection of p-[-tuple]s of elements of the group such that the group operation applied to the tuple yields the identity. Observe that X is not empty, because it contains the tuple (e,e,,e).

Now, the cyclic group Cp of order p acts on X as follows: (h,(x1,,xp))(x2,x3,,xp,x1) where h is the generator of Cp. So a general element hi acts on X by sending (x1,,xp) to (xi+1,xi+2,,xp,x1,,xi).

This is indeed a group action (exercise).

%%hidden(Show solution):

Now, fix ˉx=(x1,,xp)X.

By the Orbit-Stabiliser theorem, the orbit OrbCp(ˉx) of ˉx divides |Cp|=p, so (since p is prime) it is either 1 or p for every ˉxX.

Now, what is the size of the set X? %%hidden(Show solution): It is |G|p1.

Indeed, a single p-tuple in X is specified precisely by its first p elements; then the final element is constrained to be xp=(x1xp1)1. %%

Also, the orbits of Cp acting on X partition X (proof). Since p divides |G|, we must have p dividing |G|p1=|X|. Therefore since |OrbCp((e,e,,e))|=1, there must be at least p1 other orbits of size 1, because each orbit has size p or 1: if we had fewer than p1 other orbits of size 1, then there would be at least 1 but strictly fewer than p orbits of size 1, and all the remaining orbits would have to be of size p, contradicting that p|X|. [todo: picture of class equation]

Hence there is indeed another orbit of size 1; say it is the singleton {ˉx} where ˉx=(x1,,xp).

Now Cp acts by cycling ˉx round, and we know that doing so does not change ˉx, so it must be the case that all the xi are equal; hence (x,x,,x)X and so xp=e by definition of X.