Explicit units and the equivariant Tamagawa number conjecture. II (Q2466954)

The authors present a nice and rather explicit consequence of the validity of (a particular case of) the Equivariant Tamagawa Number Conjecture ETNC. In more detail, they consider a cyclic extension \(K/k\) of prime power degree \(\ell\) together with a finite set \(S=\{v_0,\ldots,v_n\}\) of places of \(k\), containing all archimedean and all ramified places. It is further assumed that exactly \(r\) places in \(S\) are totally split with \(r>0\), and that all other places of \(S\) do not split at all in \(K/k\). In this situation it is shown (assuming ETNC) that the Rubin-Stark ``unit'' \(\eta_{K,S}\) is the \((\sigma-1)^{n-r}\)-th power of another ``unit'' \(\varepsilon\) (for the exact statement see Theorem 3.1), where \(\sigma\) is a chosen generator of \(G\), the Galois group of \(K/k\). (The quotation marks are there to point out that one does not have units in the usual sense but elements of \(r\)-fold exterior powers of unit groups.) Moreover the authors obtain information on the norm of \(\varepsilon\) down to \(k\), in terms of the Rubin-Stark unit of \(k\) and certain Artin symbols. Finally they prove an important statement about Fitting ideals which is a little too involved to be given here. These results are quite interesting, given the twofold context in which they arise. On the one hand, there is a vastly more general theory of Burns (see his recent paper in [Invent. Math. 169, No. 3, 451--499 (2007; Zbl 1133.11063)]), where the relation between Rubin-Stark ``units'' at intermediate levels in a given abelian extension is elucidated. The other context is also interesting but different in flavor, being much more explicit: Radan Kučera and the reviewer constructed the root \(\varepsilon\) of the Stark unit in the case where the bottom field is \(\mathbb Q\) and \(r=1\) by combinatorial methods. See [\textit{C. Greither} and \textit{R. Kučera}, Acta Arith. 112, No. 2, 177--198 Zbl 1065.11089)]. Here only the infinite place splits, and all the ramified places do not split at all, so \(r=1\) and we do not need quotation marks. Moreover, the Stark unit is essentially given by the conductor-level cyclotomic unit. The approach that Burns and Hayward use to recover this result (note ETNC is a theorem if the base field is \(\mathbb Q\)) is more abstract and elegant: for instance, the determinants that came up in the work just quoted via combinatorics on trees now arise in a more natural way by evaluating a certain wedge product of homomorphisms at a wedge product of elements. But the results in the paper under review do require an effort as well, building on previous results of \textit{W. Bley} and \textit{D. Burns} [Am. J. Math. 123, No. 5, 931--949 (2001; Zbl 0984.11055)], and rather circumstantial arguments involving four-term exact sequences, base change, and Fitting ideals. The results on Fitting ideals (which we did not explain here) almost completely settle a question raised in the work of Kučera and the reviewer. At the end, the authors show that the root extraction as above is in general not possible if the Galois group is not cyclic. (A similar, perhaps slightly simpler argument occurs in the paper ``Annihilators for the class group of a cyclic field of prime power degree. II'' by \textit{R. Kučera} and the reviewer [Can. J. Math. 58, No. 3, 580--599 (2006; Zbl 1155.11054)]. This paper by Burns and Hayward is another piece of evidence for the suppleness and power of ETNC.