Wintenberger's functor for abelian extensions (Q988069): Difference between revisions

Fix a finite field \(k\) of characteristic \(p\). Let \(\mathcal{A}\) be the category whose objects are totally ramified abelian extensions \(E/F\) where \(F\) is a local field with residue field \(k\), and \(E/F\) is infinite if \(\text{char}F=0\); a morphism from \(E/F\) to \(E'/F'\) is a continuous embedding \(E\to E'\) satisfying several conditions. Let \(\mathcal{B}\) be the category whose objects are pairs \((K,A)\) where \(K\) is a local field of characteristic \(p\) and residue field \(k\), and \(A\) is a closed abelian subgroup of \(\Aut_k(K)\); a morphism \((K,A)\to (K',A')\) is a continuous embedding \(K\to K'\) satisfying several conditions. \textit{J.-P. Wintenberger} [C. R. Acad. Sci., Paris, Sér. A 290, 201--203 (1980; Zbl 0428.12012)] used the field of norms construction to give an equivalence \(\mathcal{F}\) of the categories \(\mathcal{A}_{Lie}\) and \(\mathcal{B}_{Lie}\), obtained respectively by restricting to abelian \(p\)-adic Lie extensions \(E/F\) and abelian \(p\)-adic subgroups \(A\). The paper under review extends this equivalence to the case of arbitrary abelian groups, i.e. to the categories \(\mathcal{A}\) and \(\mathcal{B}\). Here is a sketch of the proof of the essential surjectivity to indicate the main ideas. The closed subgroup \(A\) of \(\Aut_k(K)\) is a \(p\)-adic Lie group if and only if it is (topologically) finitely generated, so we may suppose that it is not. Let \(l_0<l_1<\dots\) be the positive lower ramification breaks of \(A\), and \(A_0\leq A_1\leq \) an increasing tower of finitely generated subgroups of \(A\) satisfying \(A[l_n]A_n=A\) for all \(n\). Then each \(A_n\) is a \(p\)-adic Lie group, so (by Wintenberger's result) is isomorphic to \(\mathcal{F}(L_n/F_n)\) for some \(L_n/F_n\in\mathcal{A}\). The fact that \(A\) is not finitely generated, combined with Wintenberger's result, implies that \(\Aut_k(F_n)\) is not finitely generated, whence \(\text{char}\,F_n=p\); now we may identify \(F_n\) with \(k((T))\) and thus view each \(L_n\) as an extension of the same field. Let \(E_n\) be the fixed field of \(\text{Gal}(L_n/k((T)))[l_n]\) and set \(E=\bigcup_{n\geq 0} E_n\); then \((E/k((T)))\in\mathcal{A}\) and in fact \((K,A)\cong\mathcal{F}(E/k((T)))\).