Automorphisms of descending mod-\(p\) central series (Q2178513)

Let \(\Gamma\) be a free group of finite rank and \(\Gamma_k\) the \(k\)-th term of its lower central series. Let \(\mathcal{N}_k\) be the quotient \(\Gamma/\Gamma_{k+1}\) and \(\mathcal{L}_k\) the quotient \(\Gamma_k/\Gamma_{k+1}\). It is a well-known result of \textit{S. Andreadakis} [Proc. Lond. Math. Soc. (3) 15, 239--268 (1965; Zbl 0135.04502)] that for each \(k \in \mathbb{N}\), the natural homomorphism \(\mathrm{Aut}(\mathcal{N}_{k+1}) \to\mathrm{Aut}(\mathcal{N}_{k})\) is surjective. Later, it was proved by \textit{S. Morita} [Invent. Math. 111, No. 1, 197--224 (1993; Zbl 0787.57008)] that the kernel of this epimorphism gives a non-central extension \[ 1 \to\mathrm{Hom}(\mathcal{N}_1, \mathcal{L}_{k+1}) \to\mathrm{Aut}(\mathcal{N}_{k+1}) \to\mathrm{Aut}(\mathcal{N}_{k}) \to 1. \] Further, it is known due to \textit{W. Pitsch} [J. Algebra 249, No. 2, 512--527 (2002; Zbl 1007.20029)] that this extension does not split for all \(k \in \mathbb{N}\). The paper under review deals with the construction of analogous extensions by considering Zassenhaus and Stallings mod-\(p\) central series instead of the lower central series, and study their centrality and splitting. Given a free group \(\Gamma\) of finite rank \(n\) and a prime number \(p\), let \(\Gamma^{\bullet}_k\) be the \(k\)-th term of the Stallings (\(\bullet\) = S) or Zassenhaus (\(\bullet\) = Z) mod-\(p\) central series. Let \(\mathcal{N}^{\bullet}_k\) be the quotient \(\Gamma/\Gamma^{\bullet}_{k+1}\) and \(\mathcal{L}_k\) the quotient \(\Gamma^{\bullet}_k/\Gamma^{\bullet}_{k+1}\). A part of the main result of the paper is the following theorem. Theorem. There is a non-central extension of groups \[ 1\to\mathrm{Hom}(\mathcal{N}^{\bullet}_1,\mathcal{L}^{\bullet}_{k+1})\to\mathrm{Aut}(\mathcal{N}^{\bullet}_{k+1})\to \mathrm{Aut}(\mathcal{N}^{\bullet}_{k}) \to 1. \] Further, the sequence splits if and only if \(\bullet = Z\), \(k = 1\) and \(p\) is odd or \(\bullet= S\), \(k = 1\) and \((p, n) = (3, 2), (2, 2)\).