Frobenius groups of automorphisms with almost fixed point free kernel (Q1755572)

Consider the following assertion: \par ``For each positive integer $n$, if \begin{itemize} \item[1.] $G$ is a finite solvable group, \item[2.] $A=FH$ is a Frobenius group with kernel $F$ and complement $H$, \item[3.] $A$ acts coprimely by group automorphisms on $G$, and \item[4.] $C_G(H)$, the subgroup of $G$ consisting of the common fixed points of $H$, has Fitting length $n$, \end{itemize} then the index in $G$ of the $n$-th Fitting subgroup of $G$ is bounded from above in terms of \begin{itemize} \item[1.] $|C_G(F)|$, the number of common fixed points of $F$ in $G$, and \item[2.] $|F|$.'' \end{itemize} For $n=1$, this is a theorem of \textit{E. I. Khukhro} and \textit{N. Yu. Makarenko} [J. Algebra 386, 77--104 (2013; Zbl 1293.20020), Theorem 2.1]. In its full strength, the assertion is the main result of the paper at hand. In what follows, we give a sketch of the authors' proof of this result. \par 1. The proof is by induction on $n$. \par 2. In the induction step, the authors start with several structural simplifications on $G$ (such as $\mathrm{Frat}(G)=\{1\}$) that are without loss of generality. \par 3. To reduce to the induction hypothesis, consider the quotient $\overline{G}=G/\mathrm{Fit}(G)$. \par 4. $\mathrm{Fit}(G)$, which is elementary abelian by the assumption $\mathrm{Frat}(G)=\{1\}$, can be written as a direct sum of irreducible $GA$-modules $M_1,\ldots,M_k$. \par 5. By distinguishing between two kinds of $M_i$, namely those that contain nontrivial common fixed points of $F$ versus those that do not, one can write $\overline{G}$ as a subdirect product of two quotients $\overline{G}/\overline{X}$ and $\overline{G}/\overline{Y}$, and it suffices to bound their respective $(n-1)$-th Fitting subgroup indices. \par 6. For $\overline{G}/\overline{X}$, one can actually bound $|\overline{G}/\overline{X}|$ as a whole using a theorem of \textit{B. Hartley} and \textit{I. M. Isaacs} [J. Algebra 131, No. 1, 342--358 (1990; Zbl 0703.20023), Theorem B]. \par 7. Moreover, the index of the $(n-1)$-th Fitting subgroup of $\overline{G}/\overline{Y}$ is bounded as required according to the induction hypothesis unless the Fitting length of $C_{\overline{G}/\overline{Y}}(H)$ is $n$.\newline 8. Assuming that $C_{\overline{G}/\overline{Y}}(H)$ is of Fitting length $n$, the authors can produce, via a length $n$ irreducible tower in $C_{\overline{G}/\overline{Y}}(H)$ in the sense of \textit{A. Turull} [J. Algebra 86, 555--566 (1984; Zbl 0526.20017), Definitions 1.1 and 1.2], a so-called $\mathcal{F}$-chain of length $n$ in $C_{\overline{G}/\overline{Y}}(H)$, an auxiliary notion introduced by the authors in their Definition 2.4.\newline 9. This, in turn, allows the authors to produce an $\mathcal{F}$-chain of length $n+1$ in $C_G(H)$, forcing (by the authors' Lemma 2.5) $C_G(H)$ to have Fitting length at least $n+1$, the desired contradiction concluding the proof.