Answers to questions by Dénes on Latin power sets
The i-th power, Li, of a Latin square L is that matrix
obtained by replacing each row permutation in L by its i-th power. A
Latin power set of cardinality m≥2 is a set of Latin squares
{A,A2,A3,...,Am}. We prove some basic
properties of Latin power sets and use them to resolve questions asked
by Denes and his various collaborators.
Denes has used Latin power sets in an attempt to settle a conjecture
by Hall and Paige on complete mappings in groups. Denes suggested
three generalisations of the Hall-Paige conjecture. We refute all
three with counterexamples.
Elsewhere, Denes et al. unsuccessfully tried to construct three
mutually orthogonal Latin squares of order 10 based on a Latin power
set. We confirm as a result of an exhaustive computer search that
there is no Latin power set of the kind sought. However we do find a
set of four mutually orthogonal 9×10 Latin rectangles.
We also show the non-existence of a 2-fold perfect (10,9,1)-Mendelsohn
design which was conjectured to exist by Denes. Finally, we prove a
conjecture originally due to Denes and Keedwell and show that two others
of Denes and Owens are false.