
Gao's theorem for nonabelian groups
For every finite multiplicative group , let
(
) denote the smallest integer
so that every sequence of
elements of
has a subsequence of length
(length
) which has product equal to 1 in some order.


A beautiful theorem of Gao (previously conjectured by Caro) shows that the above property holds for all abelian groups. Rather surprisingly, almost all of the proof for the abelian case seems to work as well for the general case - only one rather innocent looking bit does not carry through. Next we explore this curiosity in detail, beginning with an easy observation.


To see this, choose a sequence of length of elements which has no nontrivial subsequence with product equal to 1 in any order. Now, append
copies of 1 to this sequence. The new sequence has length
and has no subsequence of length
with product 1 in any order.
So, the hard part of Gao's theorem is to prove , and we now have multiple proofs of this fact. One of the nicest arguments uses a theorem of Kempermann-Scherck, and can be split into the following two parts.














The lemma and observation now combine easily to show in abelian groups, since we may take any sequence
of length
and modify it by mutiplying each element by a fixed constant so that 1 is the most common element of
. The lemma shows that there is now a subsequence with product 1, and the observation shows that the corresponding subsequence has product 1 in the original. So, surprisingly, the Lemma - which includes all of the real difficutly - works just fine for general groups. The only place we required the assumption
is abelian is for the observation.