Commutator theory for loops. (Q397857)

Loop theory is influenced by group theory and by universal algebra. The usual definitions of normality, center, central nilpotency and commutators are very similar to those for groups. The authors suggest a different approach to commutator theory, which seems more adapted for loops. They use the Freese-McKenzie commutator theory for congruence modular varieties to develop a new commutator theory for the variety of loops. The main theorem relates generators of the congruence commutator to generators of the total inner mapping group, more precisely, it describes the commutator \([\alpha,\beta]\) of two congruences \(\alpha,\beta\) using only the terms resulting from any set \(\mathcal W\) of words that generate total inner mapping groups. The main theorem can be used to obtain various descriptions of the commutator of two congruences. It turns out that the generating set \(\mathcal W\) in the main theorem can be simplified in some cases, in particular when it is specialized into several varieties of loops, such as inverse property loops, commutative loops, and groups. Moreover, the authors present a systematic definition of all possible associators and commutators in any loop. All these associators and commutators are associated with certain inner mappings and thus can be used to obtain a generating set of the congruence commutator. The main theorem is re-stated for commutators \([A,B]_Q\) of two normal subloops \(A,B\) of a loop \(Q\). Also in this case simplifications of the generating set \(\mathcal W\) are studied. It is shown that the center as defined in universal algebra and the center as defined in loop theory agree. Thus the concept of nilpotency based on the commutator theory yields the same concept as central nilpotency usually used in loop theory. On the other hand, the concept of solvability derived from solvability in group theory does not agree with the universal algebraic solvability. The authors argue that the usual definition of solvability in loops may be too weak. They call a loop \textit{congruence solvable}, if there is a chain \(1=Q_0\leq Q_1\leq\cdots\leq Q_n=Q\) of normal subloops \(Q_i\) of \(Q\) such that every factor \(Q_{i+1}/Q_i\) is abelian in \(Q/Q_i\). Then they show that every congruence solvable loop is solvable in the usual sense, whereas the converse is not true, as is illustrated by a counterexample.