Groups with all quotient groups Lagrangian (Q1345841)

It is common practise to call a finite group \(G\) lagrangian or a Lagrange group if for every number \(k\) dividing the order of \(G\) there is a subgroup of index \(k\) in \(G\). All finite supersoluble groups are lagrangian; it is even true that the direct product of any soluble group and a supersoluble group of the same order is lagrangian. Consequently, the class of lagrangian groups is not closed with respect to subgroups or quotient groups. \textit{G. Zappa} [Duke Math. J. 6, 511-512 (1940; Zbl 0023.21203)] has shown, that the class of groups with all subgroups lagrangian coincides with the class of supersoluble groups. Results of \textit{C. De Vivo, G. Giordano, F. Tuccillo} [Ric. Mat. 33, 121-131 (1984; Zbl 0578.20012)] and of \textit{F. Tuccillo} [Port. Math. 51, No. 2, 205-215 (1994; Zbl 0839.20027)] show that the situation is not equally simply described for the class of groups with all quotient groups lagrangian. This class is the object of this article. It has been conjectured (see for instance \textit{F. Tuccillo} [loc. cit.]), that the groups of this class are \(p\)-supersoluble for all primes \(p\) different from 2 and 3. This explains why the consideration here is restricted to \((2,3)\)-groups. We will also consider a generalization in the following manner. A group \(G\) will be called \(p\)-lagrangian, if for every \(p\)-power \(q\) dividing the order of \(G\) there is a subgroup of index \(q\). It is well-known that subgroups of coprime index intersect in a subgroup of index the product of the two indices. It is therefore apparent that a group is lagrangian if and only if it is \(p\)-lagrangian for all primes \(p\) dividing the order of the group. This allows to begin with ``local'' statements which will be pieced together at the end. The conditions imposed for the results are of such a nature that it follows immediately that the group classes are formations, unlike the class of all \((2,3)\)-groups with all quotient groups lagrangian -- for this see the counterexample at the end of this paper. The results given here lead to groups \(G\) which have Fitting length 4, with \((G/F(G))'''\) of order 2 at most (Theorem 2).