Maximal tame extensions over Hopf orders in rings of integers of \(\mathfrak p\)-adic number fields. (Q1886834): Difference between revisions

This article deals with integral Hopf Galois theory. More precisely, one fixes an odd prime power \(p^n\) and a Kummer extension \(K/k\) which is cyclic of degree \(p^n\) (so \(k\) contains a primitive \(p^n\)-th root \(\zeta\) of unity), and then one looks for \(O_k\)-orders \(S\) in \(O_K\) and \(O_k\)-orders \(H\) in \(k[G]\) (where \(G=\text{Gal}(K/k)\)), such that \(S\) is Hopf Galois (or at least tame) under the action of \(H\). Putting aside some easy cases, it is already a nontrivial task to check whether a given \(H\) acts on a given \(S\) at all; usually this comes down to cumbersome divisibility questions. The case \(n=1\) was entirely done by \textit{L.~N.~Childs} [Trans. Am. Math. Soc. 304, 111-140 (1987; Zbl 0632.12013)], who even gave a complete list of possibilities for \(S\) and \(H\). There is a largest eligible order \(S\), and it corresponds to the smallest possible \(H\). The details of the description of \(H\) and \(S\) depend on the ramification number of \(K/k\) in a fairly straightforward way. The paper under review proves comparable (but weaker) results for general \(n\). The chief restriction is that one only looks for \(H\)'s which are dual Larson orders. (This is no restriction at all in the case \(n=1\), but a rather severe one in general.) A somewhat milder restriction is the existence of a Kummer generator \(\alpha\) of \(K/k\) such that \(v_K(\alpha-1)\) is positive and coprime to \(p\). (Again, this is almost no restriction for \(n=1\).) In a first step, the author explicitly calculates the dual of an arbitrary Larson order via generators. This looks somewhat familiar for \(n\leq 2\), but as far as the reviewer can tell, this has not been done in print yet for all \(n\). The final result is the construction of an extremal pair \((S,H(S))\) where \(S\) is \(H(S)\)-Galois and \(H(S)\) is a dual Larson order. However, extremality is in a weaker sense than for \(n=1\) and just means \(S\) cannot be enlarged any more. Note in this context that for any \(S\), there can be at most one Hopf order \(H\) in \(k[G]\) making it \(H\)-Galois. The author also explains the link to work of \textit{N.~P.~Byott} [Math. Z. 220, No. 4, 495-522 (1995; Zbl 0841.16021)] and obtains a precise criterion, in terms of a ramification number, for \(S\) to coincide with the full ring of integers \(O_K\). Roughly speaking, this happens very rarely. Reviewer's remarks: In the last line of the abstract, read ``among such orders''. In the last sentence of the introduction, read ``a criterion for \(S\) coinciding with \(O_K\)'' instead of ``the condition \dots''. The often-used rings of \(O_kG\)-endomorphisms are called ``associated orders'' by many authors. It would be interesting to check whether a direct discriminant argument might suffice to deduce Theorem 1 from Proposition 1. The paper under review is quite technical; it does prove interesting results in local Galois module theory.