On the Galois correspondence for Hopf Galois structures (Q510727)

In this short paper the author explains a way to measure the failure of the Fundamental Theorem of Galois Theory (FTGT) in the context of Hopf Galois extensions of fields. Let us give a little context: If \(L/K\) is \(H\)-Galois (\(H\) a \(K\)-Hopf algebra), then one has a canonical map from the lattice of \(K\)-sub-Hopf algebras of \(H\) to the lattice of intermediate fields in \(L/K\). This map is still inclusion-reversing and injective; the interesting question is whether it is surjective. The answer has turned out to be No so often that it is natural to wonder to what extent surjectivity fails. The author restricts attention to Hopf algebras \(H\) which become isomorphic to the group ring \(\bar L[G]\) of an elementary abelian \(p\)-group \(G\) over the algebraic closure \(\bar L\) of \(L\). Using the basic description of Hopf Galois extensions due to the reviewer and Pareigis, the reinterpretation by Nigel Byott, and finally a cute result of \textit{S. C. Featherstonhaugh} et al. [Trans. Am. Math. Soc. 364, No. 7, 3675--3684 (2012; Zbl 1287.12002)], he establishes that every such \(H\) can be alternatively described by imposing a certain multiplication \(\cdot\) on \(G\) (whose given group structure is taken as addition), such that \(G\) becomes a nilpotent ring. This ring will be written \(A\) (suppressing \(G\)). The core point of the present paper is Theorem 3.1: In this setting, the sub-Hopf algebra lattice of \(H\) is isomorphic to the lattice of ideals of \(A\). Since one can also show that the subfields of \(L/K\) are in correspondence with the additive subgroups of \(A\), it becomes patent that there should be, in general, many more subfields than sub-Hopf algebras. This instinctive conclusion is confirmed by a nice theorem (4.1) which says: The only case where the correspondence of FTGT is also surjective is the extreme case where the multiplication map on \(A\times A\) is identically zero; if one unravels things, one sees that this means \(H\) is a group ring itself and the ``Hopf'' Galois structure is classical. The author goes on to give a more quantitative result (Theorem 4.2 plus discussion after it). We will not elaborate on this, because there is now a kind of sequel paper by the same author and the reviewer, which goes considerably farther. It gives fairly low upper bounds on the ratio between the number of ideals and the number of additive subgroups of \(A\), showing that this ratio tends to be extremely small. In other words, FTGT tends to fail quite badly in this setting. This paper [\textit{L. N. Childs} and the reviewer, Publ. Math. 92, No. 3--4, 495--516 (2018; Zbl 1399.13026)].