Finite \(p\)-groups all of whose non-Abelian proper subgroups are metacyclic. (Q2502307)

The note is devoted to the proof of the following nice Main theorem. Suppose that all nonabelian maximal subgroups of a \(p\)-group \(G\) are metacyclic. Then one of the following holds: (1) \(G\) is minimal nonabelian. (2) \(G\) is minimal nonmetacyclic. (3) \(G=M\times K\), where \(K=\langle c\rangle\) and \(|c|=p\) and \(M\) is metacyclic minimal nonabelian. (4) \(G=\langle a,c\mid [a,c]=b,\;a^{p^n}=c^p=b^p,\;[c,b]=1,\;[a,b]=a^{p^{n-1}}\rangle\). The proof is fairly involved since it uses the classification of metacyclic \(2\)-groups. Below a shorter and more elementary proof follows. Assuming that \(G\) is neither minimal nonabelian nor minimal nonmetacyclic, one can choose a nonmetacyclic maximal subgroup \(A\) and a minimal nonabelian subgroup \(M\); then \(A\) is Abelian and \(M\) is metacyclic. Write \(E=\Omega_1(A)\); then \(G=ME\) since \(ME\) is neither Abelian nor metacyclic. All maximal subgroups of \(G\), containing \(E\), are Abelian, so \(G/E\) is of rank at most \(2\). (i) Let \(G/E\) be noncyclic. Then \(E\leq Z(G)\). If \(c\in E-M\), then \(G=M\times\langle c\rangle\) is a group from (3). (ii) Let \(G/E\) be cyclic. If \(|G/E|=p\), then \(G=MZ(G)\), and we obtain a group from (3) again. Now let \(|G/E|>p\). Then \(|M\cap E|=p^2\) so \(M\) is maximal in \(G\). Besides, \(M\) has a cyclic subgroup of index \(p\) and \(\Omega_1(G)=E\). If \(Z(G)\) is noncyclic, \(G\) is a group from (3) again. If \(Z(G)\) is cyclic, then all maximal subgroups of \(G\) that do not contain \(E\), are \(\cong M\), and it is easily seen that then \(G\) is a group from (4).