Varieties of hexagonal quasigroups (Q687662)

The decomposition of a complete graph into disjoint cycles can be used to define a binary operation \(\star\) on the vertices of the graph -- if a cycle is \((\dots, a, b, c, \dots)\) then \(a \star b = c\) and \(c \star b = a\). In general the groupoid thus obtained is not a quasigroup, but when the decomposition satisfies an extra condition, known as ``2- perfectness'', a quasigroup is obtained. It has been known for some time that the classes of quasigroups thus obtained form varieties in the cases where the cycle sizes are 3, 5 or 7 [see \textit{C. C. Lindner}, Matematiche 45, No. 1, 83-118 (1990; Zbl 0735.05065)], but the question of what happens in the case of cycle size 6, where the obvious conditions involve an implication (\((xy = yx) \to (x = y)\)) was open. In this paper it is proved that this class, \(\mathcal H\), of quasigroups is not a variety, but that \(\mathcal H\) contains a subvariety, eventually cospectral with \(\mathcal H\). A different proof of the first result was given by \textit{D. E. Bryant} [Des. Codes Cryptogr. 2, 159-168 (1992; Zbl 0767.05065)] and he and \textit{C. C. Lindner} have since shown that 3, 5 and 7 are the only cycle sizes for which the corresponding class of quasigroups forms a variety [Algebra Univers. (to appear)].