Tsit-Yuen Lam
Finite groups that are embeddable in the multiplicative groups of division rings $K$ were completely determined by S. A. Amitsur in 1955. In case $K$ has characteristic $p>0$, the only possible finite subgroups of $K^*$ are cyclic groups, according to a theorem of I. N. Herstein. Thus, the only interesting case is when $K$ has characteristic $0$; that is, when $K\supseteq {\Bbb Q}$.
Herstein conjectured that odd-order subgroups of division rings $K$ were cyclic, and he proved this to be the case when $K$ is the division ring of the real quaternions. Herstein's conjecture was settled negatively by Amitsur. As part of his complete classification of finite groups in division rings, Amitsur showed that the smallest noncyclic odd-order group that can be embedded in a division ring is one of order $63$ (and this group is unique).
Amitsur's paper is daunting to read as it is long and technically complicated. In lecturing to a graduate class on division rings, I tried to find a simple reason for the ``first exceptional odd order'' $63$ (to Herstein's conjecture). After some work, I did come up with a reason that was simple enough to be explained to my class, without having to go through any part of Amitsur's paper. Furthermore, the method I used led easily to the second exceptional odd order, $117$ (which was not mentioned in Amitsur's paper). Since this line of reasoning did not seem to have appeared in the literature before, I record it in this short note. To better motivate the results discussed here, I have also included a quick exposition on the beginning part of the theory of finite subgroups of division rings.