Enumerating dihedral Hopf-Galois structures acting on dihedral extensions (Q2333344)

This is another interesting paper concerned with the counting of Hopf Galois structures on a given Galois extension \(L/K\) with group \(G\). (These structures are given by the action of a \(K\)-Hopf algebra \(H\) on \(L\), satisfying certain natural axioms.) The classical Hopf Galois structure is given by \(H=K[G]\) acting in the obvious way, and the \textit{type} of \(H\) is the unique group \(N\) such that \(H\) and \(K[N]\) become isomorphic over \(\bar K\). The present article gives a complete count in the case where both \(G\) and \(N\) are the dihedral group \(D_n\) of order \(2n\ge 6\). We remark that of course \(G\) and \(N\) must have the same order, but they need not be isomorphic in general; this is a hypothesis imposed in the present paper, to focus the search. We give the result in the simplest case, when \(n\) is odd. Then the number of Hopf Galois structures is exactly 2, and (reviewer's remark) they are readily identified: one is the classical H.-G. structure, and the other is the ``anti-classical'' structure, which always exists and is not the classical one as soon as \(G\) is not abelian. The answer in general involves the number of prime factors of \(n\), via the 2-torsion in the multiplicative group \((\mathbb Z/n)^*)\), and the congruence class of \(n\) modulo 8. The key of the classification is the original description of H.-G. structures [\textit{C. Greither} and \textit{B. Pareigis}, J. Algebra 106, 239--258 (1987; Zbl 0615.12026)], and the nice observation that the choice of subgroup \(N \cong D_n\) in \(B=\operatorname{Perm}(D_n)\) can be significantly narrowed down as follows: if \(n\) is odd, then the action of \(N\) respects the obvious partition of \(D_n\) in two blocks of size \(n\) (the cyclic subgroup \(X\) of order \(n\), and its nontrivial coset); for even \(n\), a variant of this remains true which is technically a bit more involved. The author uses a previous result of his [Commun. Algebra 43, No. 10, 4290--4304 (2015; Zbl 1342.20001)] on a so-called multiple holomorph. More precisely, this result determines the index of \(\operatorname{Hol}(G)\) in its \(B\)-normalizer, where \(G=D_n\), and \(\operatorname{Hol}(G)\) is the so-called holomorph of \(G\), embedded in \(B=\operatorname{Perm}(G)\). Again, in the simplest case \(n\) odd this index is 2. The methods are elementary and the arguments are partly a little involved and technical, but everything is laid down in very clear detail. (Cor.~3.12 also admits a quick proof by looking at the quotient of \(D_n\) modulo commutators; this is of order 2 for \(n\) odd and a Klein four group for \(n\) even.)