Generalised Kawada-Satake method for Mackey functors in class field theory (Q1999431)

Since Artin and Tate, the axioms of \textit{class formation} have been used to derive the existence theorem and the reciprocity homomorphism of class field theory (CFT for short). The authors propose to call such an approach a \textit{CFT mechanism}. Many CFT mechanisms are written down in a purely group-theoretical and Galois module-theoretical way, without using ring structures, and many of the best known go through a computation of the Brauer group. There exist however approaches which are explicit and essentially cohomology free (Hazewinkel, Neukirch). In particular, \textit{Y. Kawada} and \textit{I. Satake} [J. Fac. Sci., Univ. Tokyo, Sect. I 7, 353--389 (1956; Zbl 0101.02902)] derived the $p$-primary part of CFT of local and global fields of characteristic $p > 0$, using Witt theory which describes the $p$-part of the maximal abelian extension in terms of the direct limit of the truncated Witt vectors modulo the operator $\wp =$ Frobenius-Id. The main part of the present article presents a CFT mechanism in positive characteristic which can be viewed as a simplification and generalisation of the Kawada-Satake approach. After recalling basic facts on Witt theory, the authors formulate key properties/axioms which deal with more general objects than Galois modules and do not include any of the standard or modified class formation axioms, such as information on the index of the norm group and the kernel of the norm map. The novelty, already present in previous work (1992--1995) of I. Fesenko, is the introduction of special Mackey functors $\mathcal{A}$ from the category of finite left $G$- sets (where $G$ is a profinite group) to the category of abelian groups endowed with translation invariant topology (not necessarily topological groups) and continuous closed morphisms between them. Without recalling the five axioms defining $\mathcal{A}$ as a Mackey functor, nor the five additional special properties required here for A, one can give an idea of some of the main CFT results in positive characteristic which can be proved without cohomology. Notations: $k$ is a field of characteristic $p > 0$, $k^{sep,p}$ a maximal separable $p$-extension, $G = G^p(k)$ its Galois group over $k$; $W (k) =$ the $(p)$-Witt ring of $k$, which is the inverse limit of the rings of truncated Witt vectors $W_m(k)$, $m \geq 1$, $\mathcal{W}(k) = W (k)/\wp(W (k)) \otimes \mathbb{Q}_p /\mathbb{Z}_p$, $\mathcal{W}_m (k) = W_m (k)/\wp(W (k)) \otimes p^{-m} \mathbb{Z}/\mathbb{Z}$; $\mathcal{C}(H) = \mathcal{A}(H)/\bigcap p^m \mathcal{A}(H)$ for every open subgroup $H$ of $G$. Then: (1) If $k_H$ is the subfield of $k^{sep,p}$ fixed by a closed subgroup $H$ of $G$, one has a continuous and non-degenerate in each variable pairing $<. , . >_H : \mathcal{C}(H) \times \mathcal{W}(k_H) \to \mathbb{Q}_p /\mathbb{Z}_p$. It induces the ($p$-primary) $H$-reciprocity homomorphism $\Phi_H : \mathcal{C}(H) \to H^{\mathrm{ab}}$, which is injective, continuous, with dense image; (2) For a normal open subgroup $K$ of $H$, the index of the norm group $N_{H/K}\mathcal{C}(K)$ in $\mathcal{C}(H)$ is equal to the order of $(H/K)^{\mathrm{ab}}$ and the reciprocity map $\Phi_H$ induces the isomorphism $\Phi_{H/K} : \mathcal{C}(H)/N_{H/K}\mathcal{C}(K) \cong (H/K)^{\mathrm{ab}}$; (3) The lattice of open subgroups $K$ of $H^{\mathrm{ab}}$ (w.r.t. intersection and product) is in an order preserving bijection with the lattice of closed subgroups of finite index in $\mathcal{C}(H)$ (existence theorem). It remains to check the axioms of this CFT mechanism for a specific type of groups $G$, typically Galois groups of specific fields which appear in (more or less) classical CFT. This is done in the last part of the article: (i) In one-dimensional local CFT in positive characteristic, the base field is a complete discrete valuation $F$ of characteristic $p$ with a finite residue field $\mathbb{F}_q$. The Mackey functor $\mathcal{A}$ associates to an open subgroup $H$ of $G^p(F)$ the multiplicative group $F_H^*$. The main results above are valid, in particular the most difficult and central part, the $p$-primary part of the local reciprocity map $F^*\to G^p(F)$. To get the full local CFT for $F$, a little extra (but easy) work is needed to glue together the unramified $\hat{\mathbb{Z}}$-part and the CFT of the prime-to-$p$ kummerian part $F({}^{q-1}\sqrt{F^*}/F)$. In one-dimensional global CFT in positive characteristic, the base field $K$ is an algebraic function field in one variable over a finite field $\mathbb{F}_q$ which is algebraically closed in $K$. The Mackey functor associates to an open subgroup $H$ the idèle class group of $K_H$. Again, the above mechanism and an easy extra work give the full global CFT for $K$ in positive characteristic. (ii) In higher local CFT in positive characteristic, the objects in the image of the Mackey functor are the topological $K_n^{top}(F)$-groups. The base field $F$ is an $n$-dimensional local field of characteristic $p$, i.e. a complete discrete valuation field whose residual field is $(n-1)$-dimensional. Let $\mathbb{F}_q$ be the last of the chain of residual fields. The authors' CFT mechanism applies to produce the $p$-primary part of higher CFT in characteristic $p$. The full CFT for $F$ requires again to glue together the purely unramified $\hat{\mathbb{Z}}$-part and the higher tame reciprocity homomorphism $K_n^{top}(F)\to G(F({}^{q-1}\sqrt{F^*})/F)$ to get the full reciprocity homomorphism $\Phi_F :K_n^{top}(F)\to G(F^{\mathrm{ab}}/F)$.