Counting Hopf-Galois structures on cyclic field extensions of squarefree degree (Q1675062): Difference between revisions

This paper is an interesting contribution to the growing literature about the counting of Hopf Galois structures on a given Galois extension \(L/K\) with group \(\Gamma\). These structures are given by the action of a \(K\)-Hopf algebra \(H\) on \(L\), satisfying certain natural axioms. The so-called classical Hopf Galois structure is given by \(H=K[\Gamma]\) acting in the obvious way. The \textit{type} of \(H\) is the unique group \(G\) such that \(H\) and \(K[G]\) become isomorphic over the algebraic closure of \(K\). It is pretty obvious that the given group \(\Gamma\) and the type group \(G\) must have the same order. One of the basic insights of Hopf Galois theory is that usually \(G\) and \(\Gamma\) need not be isomorphic. This has opened up an entire domain of research: given \(\Gamma\), find the possible type groups \(G\), and for each such \(G\), find out how many Hopf Galois structures there are, according to the precise definition of equivalence of such structures. For a fixed group \(G\), there may be non-isomorphic Hopf algebras \(H\) with type \(G\), but even isomorphic ones can give rise to non-equivalent Hopf Galois structures. This classification comes down to a problem in pure group theory, as shown by \textit{B. Pareigis} and the reviewer [J. Algebra 106, 239--258 (1987; Zbl 0615.12026)]. It looks elementary but it may be intractable anyway. For brevity, we only sketch the simplified form [\textit{L. N. Childs}, Commun. Algebra 17, No. 4, 809--825 (1989; Zbl 0692.12007)] [\textit{N. P. Byott}, Commun. Algebra 24, No. 10, 3217--3228 (1996; Zbl 0878.12001)], which is now called ``Byott's translation''. Given any candidate \(G\) for a type group, one needs to find and classify (or exclude the existence of) copies \(\Gamma^*\) of \(\Gamma\) which are regular subgroups in the holomorph \(\text{Hol}(G)\), the semidirect product of \(G\) (acting as translations) and \(\Aut(G)\) inside \(\text{Perm}(G)\). The introduction of the paper under review explains very nicely what had been known: e.g. if \(\Gamma\) is cyclic of order \(p^{m}\), then the count of Hopf Galois structures gives \(p^{m-1}\); if \(\Gamma\) is \(p\)-elementary requiring many generators, the problem appears to be wild. A complete list of previous results may be found in the MR review MR2715201 of the same paper, written by L. Childs; not only for this reason, it is highly worthwile to consult this review as well. The paper under review achieves a complete treatment of the case where \(\Gamma\) is cyclic of square-free order \(n\). Surprisingly, every group \(G\) of order \(n\) can occur as a type group, and the first main result determines exactly how often. The technical core of the proof is a thorough analysis of the holomorph of \(G\). The number of Hopf Galois structures of type \(G\) is obtained as a slightly complicated function of \(G\), which involves, among other quantities, the order \(z\) of the center of \(G\), the order \(g\) of the commutator group of \(G\), and the quotient \(d:=n/gz\). For squarefree \(n\), the quantity \(d\) is always an integer, as can be seen from the classification of \textit{M. R. Murty} and \textit{V. K. Murty} [Math. Ann. 267, 299--309 (1984; Zbl 0531.10048)], which is a crucial tool in the proof. (Side remark: For the group \(G=\mathrm{SL}(2,7)\) of order 336, we have \(z=2\) and \(g=|G|\), so \(d\) is not an integer; but of course 336 is not squarefree.) This classification says: Every group of order \(n\) can be obtained as the semidirect product of two cyclic groups \(N\) and \(H\), \(H\) acting on \(N\), and one can even assume that the action is faithful. This is a strong result; naively one might guess that one can iterate the semidirect product construction any number of times, and the resulting groups should require many generators. The truth is different: two generators always suffice, and one may assume that the first generator generates a normal subgroup. As their second main result, the authors prove an explicit formula for the total number of Hopf Galois structures, by summing the previous formula over all \(G\) of (squarefree) order \(n\). The paper concludes with a well-compiled list of examples. One extreme case occurs when \(n\) is a so-called Burnside number, that is, \(n\) is coprime to \(\phi(n)\). Note that this forces \(n\) to be squarefree. Then all groups of order \(n\) are cyclic, and the total count is 1 (so there are no Hopf Galois structure besides the classical one; this had been known). Then the authors discuss cases of increasing complexity, concluding with a partial discussion of the case that \(n\) has four prime factors. This paper is a very nice combination of thorough theoretical deductions and an instructive range of examples.