Mendelsohn designs associated with a class of idempotent quasigroups (Q1842156)

Mendelsohn introduced the concept of perfect cyclic design, now generally known as Mendelsohn design. A Mendelsohn design can be thought of as a decomposition of a \(\lambda\)-fold complete, directed, loopless graph into a collection of directed cycles of length \(k\) such that the edge set of the graph is partitioned. Mendelsohn gave a technique for constructing perfect designs, with \(\lambda= 1\), using quasigroups derived from Galois fields. He defined an operation \(*\) on \(\text{GF}(g)\) by \(x* y= -ux+ (1+ u)y\), where \(u\) is a primitive \(k\)th root of unity, and showed that the cycles of the form \((a, b, a* b, b* (a* b),a* (b* (a* b)),\dots)\), where the vertices of the graph are labelled with the elements of the field, yield a perfect \(k\)-cycle system of order \(q\) with \(\lambda= 1\). The case \(u= -{1\over 2}\) is of particular interest, because the quasigroup so obtained is commutative, and when \(q\) is a prime, the obtained quasigroup is the idempotent commutative quasigroup which plays an important role in the construction of Steiner triple systems. The above quasigroup can be defined on \(\mathbb{Z}_ n\) for any odd \(n\) and, in general, the operation \(*\) is a quasigroup operation on \(\mathbb{Z}_ n\). Both \(-u\) and \(1+ u\) are units in \(\mathbb{Z}_ n\) (if \(n\) is odd). This paper investigates the designs obtained in this more general situation, determines the cycle length, and proves the resolvability. C. C. Lindner and N. S. Mendelsohn considered another interesting case, also over a finite field, namely \(u= 1\). In this situation \(x* y= -x+ 2y\), and they have cycles of length \(p\). Since \((x* y) *y= x\), each cycle occurs with its reverse. Thus they considered the system as consisting of undirected cycles.