A torus theorem for homotopy nilpotent loop spaces (Q1749365)

Nilpotency of a discrete group can be understood as a measure of how far the group fails to be commutative. So a \(1\)-nilpotent group is abelian, and \(n\)-nilpotency implies \((n+1)\)-nilpotency for general \(n\). A way to formalize this is by counting how many central extensions are necessary to build the group, or equivalently how many iterated commutators must always vanish. \textit{I. Berstein} and \textit{T. Ganea} [Ill. J. Math. 5, 99--130 (1961; Zbl 0096.17602)] introduced the notion of nilpotency in the homotopical context by taking the commutators point of view. They called a loop space for which the \((n+1)\)-st commutator map is homotopy trivial \textit{\(n\)-nilpotent}. \textit{B. Chorny} and \textit{J. Scherer} [Forum Math. 27, No. 1, 119--130 (2015; Zbl 1311.55019)] proved that an \(n\)-nilpotent space in this sense is precisely a \(\mathrm{Nil}_n\)-algebra in the homotopy category of pointed spaces, where \(\mathrm{Nil}_n\) is the algebraic theory for nilpotent groups. \textit{G. Biedermann} and \textit{W. G. Dwyer} [Algebr. Geom. Topol. 10, No. 1, 33--61 (2010; Zbl 1329.55008)] had previously considered a stronger notion of nilpotency by introducing a simplicial algebraic theory \(\mathcal G_n\) whose homotopy category is \(\mathrm{Nil}_n\), and calling \(n\)-nilpotent a loop space that admits the structure of a \(\mathcal G_n\)-algebra. In this paper, the authors refine further the comparison of Chorny-Scherer between nilpotency in the sense of Berstein-Ganea and Biedermann-Dwyer, by introducing an intermediate invariant, called ``extension by principal fibrations length'' or epfl for short, and a corresponding intermediate notion of nilpotency. As suggested by the name, epfl counts how many principal fibrations (which play the role of central extensions in the homotopical context) are necessary to build a loop space. They prove that epfl is bounded below by Berstein-Ganea nilpotency and bounded above by Biedermann-Dwyer nilpotency, and that this (non-strict) inequality specializes to an equality for discrete groups and for infinite loop spaces (i.e., \(1\)-nilpotent loop spaces). A direct consequence is a more direct proof of the vanishing of iterated Whitehead products in values of excisive functors, which had been previously proven by Chorny-Scherer and Eldred with different techniques. Moreover, the authors study the effect of certain homotopical localization functors on loop spaces with finite epfl, and use it to show that connected \(p\)-compact groups with finite Biedermann-Dwyer nilpotency are precisely \(p\)-compact tori, and connected finite loop spaces with finite Biedermann-Dwyer nilpotency have the homotopy type of a torus. This is an extension of Hubbuck's Torus Theorem to loop spaces that are nilpotent, but not necessarily infinite loop spaces.