Local Leopoldt's problem for rings of integers in Abelian \(p\)-extensions of complete discrete valuation fields (Q1591984)

This article proves some remarkable and novel results in local Galois module theory. One central result is the following: Suppose that \(K/k\) is a totally ramified Abelian \(p\)-extension of \(p\)-adic fields with Galois group \(G\) satisfying two properties: (1) \(K/k\) is not almost maximally ramified, and (2) the different ideal \(D_{K/k}\) is induced from an ideal of \(O_k\). As usual, the associated order \(A=A_{K/k}(O_K)\) is defined as the subring of \(K[G]\) consisting of all \(z\) with \(zO_K \subset O_K\). Then freeness of \(O_K\) over \(A\) already implies that: (H) the extension \(O_K/O_k\) is Hopf Galois with Hopf algebra \(A\). (The converse is well known, and the condition (2) is known to be necessary for (H) to hold.) Actually more is proved; see below. We will try to give an idea of some of the major ideas involved in the proofs. One idea is to narrow down the choice for a generator \(a\) of \(O_K\) as a module over its associated order \(A\). This has already been done in a more specific context (see work of M.~Taylor), and one is led to consider elements \(a\) of \(K\)-valuation \([K:k]-1\). More precisely, one says that \(K/k\) fulfils condition (17) (numbering of the paper) if there is \(a\in O_K\) with \(v_K=[K:k]-1\) and \(\xi\in A\) such that \(\xi(a)\) has \(K\)-valuation 1. Then Theorem 2.4.2 states that (17) is equivalent to the freeness of \(O_K\) over \(A\) (\(a\) becoming a generator), under a mild extra hypothesis. A very important tool is a surprising new multiplication on \(A\) which may be briefly explained as follows: There is a sort of canonical \(K\)-linear isomorphism \(c: K[G] \to K^G\) from the group ring to its dual \(K^G=\text{Map}(G,K)\), sending \(\sigma\) to the idempotent labeled with \(\sigma\). Now suppose the different of \(K/k\) is generated by \(\delta\in O_k\). What the author proves amounts to saying that the image of \(A\) under \(\delta\cdot c\) is a subring (with unit) of \(K^G\). It is instructive to note that the preimage of 1 is \(\delta^{-1}\)tr which is indeed in \(A\). Now one may pull the multiplication on \(K^G\) back to \(A\) through \(\delta\cdot c\), to obtain a new multiplication on \(A\) denoted \(*^{\Delta}\) by the author. In the later part of the paper, Theorem A is proved which is even stronger than the result mentioned at the beginning of this review: under the same hypotheses, \(O_K\) is \(A\)-free if and only if \(K/k\) is Kummerian for a formal group \(F\) defined over \(k\). This is a strong result indeed, and the construction of \(F\) proceeds through various rather technical stages. A few minor remarks: ``Correctly defined'' means ``well defined'' (in Russian ``korrektno opredeleno''). In the statement of Proposition 2.3.1, the intended meaning is: the image of \(A_{K/k}(O_K)(a)\) in \(O_K/M^n\) under the canonical map. On the same page 671, the element \(a\) in formula (13) has nothing to do with the one from (12). In Remark 2.3.2, the congruence probably means equality of images. In the proof of 3.4.1, by a standard trick, it would not be necessary to check that \(A\) is stable under the antipode, once one knows that it is stable under \(\Delta\). On p.680, unnumbered formula just above formula (26): \(\psi\) should probably be \(\xi\). Two lines onward: \(\alpha\) is \(\phi^{-1}(\delta\xi)\), as can be inferred from (22) three pages earlier and a belated definition on p.689. On p.690, formula just after (48): just before last \(=\) sign, \(\sigma_\tau\) is probably \(\sigma\tau\), and it seems the \(\sigma\) just after \(=\) shouldn't be there. The reviewer does not want to pretend he has worked through all the details, though. The paper under review is strong because of the fresh ideas it introduces, and the results it proves; as Bondarko unambiguously comments in Remark 0.3.4, some of the results were not even conjectured to hold. The paper is likely to generate a lot of follow-up research.