Groups of class \(2n\) in which all proper subgroups have class at most \(n\) (Q1694578)

The paper under review studies the nilpotency class of a nilpotent group in relation to the nilpotency class of its proper subgroups. Fitting's theorem states that the product of two normal nilpotent subgroups of nilpotency class \(c\) and \(d\) is again nilpotent of class \(\leq c + d\). As a consequence, if a nilpotent group contains only proper subgroups of nilpotency class \(\leq n\), the group itself has nilpotency class \( \leq 2n\). Similarly, if a torsion-free nilpotent group has only subgroups of infinite index for which the nilpotency class is \(\leq n\), then the group itself has nilpotency class \(\leq 2n\). This leads to the question whether these bounds are optimal, i.e. whether there exist groups of nilpotency class \(2n\) for which every proper or infinite index subgroup has nilpotency class at most \(n\). It is well-known that these bounds are sharp for \(n= 1\) and \(n=3\), whereas for \(n=2\) every group of nilpotency class \(4\) has a proper subgroup of nilpotency class \(3\). The authors improve these results by constructing for every \(n \geq 3\) a nilpotent Lie algebra of nilpotency class \(2n\) for which every proper subalgebra has nilpotency class \(\leq n\). The construction of this Lie algebra starts from a free Lie algebra on two generators and then takes a certain quotient of nilpotency class \(2n\). The main difficulty is showing that that the nilpotency class of the quotient is indeed \(2n\), for which the authors make some clever computations and use the action of a permutation group on the free Lie algebra. Starting from the constructed nilpotent Lie algebras, the authors then apply the Mal'cev and the Lazard correspondence to show that the aforementioned bounds are sharp for both finite and torsion-free nilpotent groups.