Higher wild kernels and divisibility in the \(K\)-theory of number fields (Q2491741)

For a number field \(F,\) the author proposes to define the higher wild kernel \(K^w_{2i}(F),\) for \(i \neq 1,\) as the kernel of \(\lambda_F = \oplus \lambda_v\;:\;K_{2i} (F) \to (\mathbb Z/2 \mathbb Z)^{r_1}\oplus \coprod_{v \text{ finite}}\;\mu^{\otimes i} (F_v),\) where \(\lambda_v\;:\;K_{2i} (F) \to K_{2i}(F_v) \to H^2 \Bigl( F_v, \mu^{\otimes i+1}\Bigl) \simeq \mu^{\otimes i}(F_v)\) is the Dwyer-Friedlander map (this generalizes the classical wild kernel \(K^w).\) It is obvious that div \(K_{2i} (F) \subseteq K^w_{2i}(F) \subseteq K_{2i} ({\mathcal O}_F),\) and the main result of this paper asserts that \([K^w_{2i} (F)\;:\text{div}\, K_{2i} (F)]\) is equal to 2 if \(i\) is odd and \(F\) is special, to 1 otherwise. Recall that a number field \(F\) is called non-exceptional if every cyclotomic extension of \(F\) is cyclic, and special [\textit{K. Hutchinson}, J. Number Theory 107, 368--391 (2004; Zbl 1082.11076)] if \(F\) is exceptional and for every place \(v\mid 2,\) there is a 2-primary root of unity \(\xi\) such that \(\xi + \xi^{-1}\) belongs to \(F_v\) but not to \(F.\) Note that Hutchinson has also obtained essentially the same result as here (preprint 2004). He used an Iwasawa theoretic approach and the present author argues at ``finite level'', but the common starting point is to admit the Quillen-Lichtenbaum conjecture (which should be a theorem of Voevodsky-Rost) and to proceed by co-descent, i.e., by taking co-invariants on Galois cohomology. In this setting, the odd part -- and the 2-primary part if \(F\) is non-exceptional -- of the theorem go back to \textit{P. Schneider} [Math. Z. 168, 181--205 (1979; Zbl 0421.12024)]. At the prime 2, two kinds of difficulty occur: -- the Quillen-Lichtenbaum no longer gives an isomorphism, only an exact sequence [\textit{I. Rognes} and \textit{C. Weibel}, J. Am. Math. Soc. 13, 1-54 (2000; Zbl 0934.19001)] -- if \(F\) is real, co-descent works well only for a modified cohomology called positive cohomology [which was also used in the same context by \textit{B. Kahn}, \(K\)-Theory 7, 55--100 (1993; Zbl 0780.12007)].