The automorphism tower problem revisited (Q1969247): Difference between revisions

If \(G\) is a group with trivial centre, then the group \(\operatorname{Inn}G\) of inner automorphisms is isomorphic to \(G\) and so \(G\) can be identified with \(\operatorname{Inn}G\) and is therefore a normal subgroup of \(\Aut G\). Now \(\Aut G\) has also trivial centre and is a subgroup of \(\Aut(\Aut G)\). Hence this gives the automorphism tower of \(G\) \[ G=G_0\trianglelefteq G_1\trianglelefteq G_2\trianglelefteq\cdots\trianglelefteq G_\alpha\trianglelefteq G_{\alpha+1}\trianglelefteq\cdots \] such that for each ordinal \(\alpha\) \(G_{\alpha+1}=\Aut G_\alpha\) and if \(\alpha\) is a limit ordinal, then \(G_\alpha=\bigcup_{\beta<\alpha}G_\beta\). The automorphism tower terminates if there is an ordinal such that \(G_{\alpha+1}=G_\alpha\). A classical result of \textit{H. Wielandt} [Math. Z. 45, 209-244 (1939; Zbl 0021.21003)] says that for a finite group the automorphism tower terminates after finitely many steps. This is not true in general for infinite groups and the problem was whether the automorphism tower terminates at all after finitely or countably or uncountably many steps. This problem was solved in full generality by \textit{S. Thomas} [Proc. Am. Math. Soc. 95, 166-168 (1985; Zbl 0575.20030)], who proved that the automorphism tower of an arbitrary centreless group eventually terminates; and that for each ordinal \(\alpha\) there exists a group whose automorphism tower terminates in exactly \(\alpha\) steps. The authors study the bound of the number of steps the automorphism tower terminates for a group of cardinality \(\kappa\). In their own words: It is well known that the automorphism towers of infinite centreless groups of cardinality \(\kappa\) terminate in less than \((2^\kappa)^+\) steps. But an easy counting argument shows that \((2^\kappa)^+\) is not the best possible bound. However, in this paper, we show that it is impossible to find an explicit better bound using ZFC. In the course of their study the authors introduce many terms and notations and prove many results which have their own significance. We mention only one: If \(\tau(G)\) is the least ordinal \(\alpha\) such that \(G_{\alpha+1}=G_\alpha\) for the centreless group \(G\), then for the infinite cardinal \(\kappa\) let \(\tau_\kappa\) be the least ordinal such that \(\tau(G)<\tau_\kappa\) for every centreless group of cardinality \(\kappa\). The authors prove that \(\tau_\kappa\) is increasing, i.e., if \(\omega\leq\theta<\kappa\), then \(\tau_\theta\leq\tau_\kappa\).