The Stickelberger splitting map and Euler systems in the \(K\)-theory of number fields (Q1932376)

Let \(K\) be a totally real field, \(F/K\) an abelian CM extension of conductor \(\mathfrak{f}\). For any integer \(n \geq 1\), for any integral ideal \(\mathfrak{p}\) of \(K\) coprime to \(\mathfrak{f}\) and to \(w_{n+1}(F):= \# H^0(F, \mathbb{Q}/\mathbb{Z}(n+1))\), one can define a higher Stickelberger element \(\Theta_n := \Theta_n(\mathfrak{p}, \mathfrak{f}) \in \mathbb{Z}[G(F/K)]\), and the Coates-Sinnott conjecture, say (CS), states that \(\Theta_n\) annihilates \(K_{2n}(o_F)\). Note that for \(n\) even, (CS) says nothing more than ``zero annihilates everything''. The main result of this paper is the proof of a weak form of the \(\ell\)-part (\(\mathrm {CS}_\ell\)) of (CS), more precisely: for any prime \(\ell\) and under certain conditions, \(\Theta_n\) annihilates the group \(div (K_{2n}(F)_\ell)\) of infinitely divisible elements in \(K_{2n}(F)_\ell\) (this subgroup is contained in \(K_{2n}(o_F)_\ell\)). Before giving an idea of the authors' method, let us recall the étale version (\(\mathrm {CS}_ {\text{ét}}\)) of (CS): if \(\mathfrak{p}\) is moreover coprime to \(\ell\), then \(\Theta_n\) annihilates \(K_{2n}^{\text{ét}}(o_F[1/\ell])\). From the Quillen-Lichtenbaum conjecture (now a theorem thanks to the work of Voevodsky, Rost and others), which asserts that \(K_{2n}(o_F)_\ell \overset{\sim} K_{2n}^{\text{ét}}(o_F[1/\ell])\), one deduces readily that (\(\mathrm {CS}_{\text{ét}}\)) implies the \(\ell\)-part (\(\mathrm {CS}_\ell\)) of (CS). Modulo the vanishing of the \(\mu\)-invariant \(\mu_{F, \ell}\) attached to the cyclotomic \(\mathbb{Z}_\ell\)-extension \(F_\infty/F\) the étale conjecture (\(\mathrm{CS}_{\text{ét}}\)) has been proved independently by \textit{D. Burns} and \textit{C. Greither} [Doc. Math., J. DMV Extra Vol., 157--185 (2003; Zbl 1142.11371)], assuming that \(F \cap K_\infty = K\), and by the reviewer [J. Théor. Nombres Bordx. 17, No. 2, 643--668 (2005; Zbl 1098.11054)], unconditionally (in this respect, the comment on ``extra hypotheses'' on page 843 of the present paper is misleading). See also the preprint by the second author and \textit{C. Greither} [``An equivariant main conjecture in Iwasawa theory and applications'', Preprint, \url{arXiv:1103.3069v1}]. The authors' approach here is to ignore the Quillen-Lichtenbaum conjecture and to prove their weaker version of (CS\(_\ell\)) by constructing ``Stickelberger pseudo-splitting maps'' for the étale and the Quillen \(K\)-theory of \(F\). More precisely, consider the boundary map \(\delta_F\) in Quillen's localization exact sequence \(0 \to K_{2n}(o_F)_\ell \to K_{2n}(F)_\ell \overset{\delta_F} \oplus_v K_{2n-1}(\mathfrak{k}_v)_\ell 0,\) where \(\mathfrak{k}_v\) is the residue field at \(v\). A Stickelberger pseudo-splitting map is by definition a homomorphism \(\Lambda : \oplus_v K_{2n-1}(\mathfrak{k}_v)_\ell \to K_{2n}(F)_\ell\) such that \(\delta_F \circ \Lambda\) is the multiplication by \(\Theta_n\). The existence of such a \(\Lambda\) easily implies that \(\Theta_n\) annihilates \(div (K_{2n}(F)_\ell)\). The (very technical) construction of \(\Lambda\) and its étale analogues \(\Lambda^{\acute{\text{ét}}}\) requires a starting hypothesis, say (\(H_m\)), that there exists an integer \(m \geq 1\) such that \(\Theta_m(\mathfrak{p}, \mathfrak{f}_k)\) annihilates \(K_{2m}(o_{F_k})_\ell\) for all \(k \geq 0\) (here \(F_k=F(\mu_{\ell^k})\) and \(\mathfrak{f}_k\) is the conductor of \(F_k\)). Compatibility properties between \(\Lambda\) and \(\Lambda^{\acute{e}t}\) must be taken great care of, because the starting hypothesis (\(H_m\)) will come from étale \(K\)-theory in the following way: supposing that \(\mu_{F, \ell}=0\), then (CS\(_\ell\)) holds true for \(\Theta_m\) and \(K_{2m}^{\text{ét}}(o_F[1/\ell])\) for all \(m \geq 1\), as was recalled above; take \(m=1\), in which case the Quillen-Lichtenbaum conjecture holds (Tate's theorem); since the vanishing of the \(\mu\)-invariant depends only on \(F\), not on \(F_k\), the hypothesis (\(H_1\)) is valid, and one can conclude that \(\Theta_n\) annihilates \(div (K_{2n}(F)_\ell)\) for any \(n \geq 1\). A by-product of the elaboration of the ``Stickelberger pseudo-splitting maps'' is the construction of the Euler system for the even Quillen \(K\)-theory of CM abelian extensions of totally real number fields. The authors promise to apply these Euler systems to study the structure of the group \(div (K_{2n}(F)_\ell)\) in forthcoming work. In the reviewer's opinion, this (if achieved) would be the genuine interest of their present approach, because as far as their weakened (CS\(_\ell\)) is concerned, here is a straightforward argument: it is known that the étale Chern class map is split surjective, (see e.g., \textit{B. Kahn} [``On the Lichtenbaum-Quillen conjecture'', in: Proceedings of the NATO Advanced Study Institute, Lake Louise, Alberta, Canada, December 12--16, 1991. Dordrecht: Kluwer Academic Publishers. NATO ASI Ser., Ser. C, Math. Phys. Sci. 407, 147--166 (1993; Zbl 0885.19004)]; one can even construct a splitting which sends the so-called étale wild kernel (= the kernel of the localization in \(H^2_{\text{ét}}(o_F[1/\ell], \mathbb{Z}_\ell(n+1))\)) onto \(div (K_{2n}(F)_\ell)\), see the reviewer [Sémin. Théor. Nombres Bordx., Sér. II 4, No. 2, 263--271 (1992; Zbl 0783.11042)]; then one just has to apply (\(\mathrm {CS}_{\text{ét}}\)).