Izvestiya: Mathematics
RUS  ENG    JOURNALS   PEOPLE   ORGANISATIONS   CONFERENCES   SEMINARS   VIDEO LIBRARY   PACKAGE AMSBIB  
General information
Latest issue
Forthcoming papers
Archive
Impact factor
Guidelines for authors
Submit a manuscript

Search papers
Search references

RSS
Latest issue
Current issues
Archive issues
What is RSS



Izv. RAN. Ser. Mat.:
Year:
Volume:
Issue:
Page:
Find






Personal entry:
Login:
Password:
Save password
Enter
Forgotten password?
Register


Izvestiya: Mathematics, 2023, Volume 87, Issue 3, Pages 421–438
DOI: https://doi.org/10.4213/im9322e
(Mi im9322)
 

Ramification filtration and differential forms

V. A. Abrashkinab

a Department of Mathematical Sciences, University of Durham, United Kingdom
b Steklov Mathematical Institute of Russian Academy of Sciences, Moscow, Russia
References:
Abstract: Let $L$ be a complete discrete valuation field of prime characteristic $p$ with finite residue field. Denote by $\Gamma_{L}^{(v)}$ the ramification subgroups of $\Gamma_{L}=\operatorname{Gal}(L^{\mathrm{sep}}/L)$. We consider the category $\operatorname{M\Gamma}_{L}^{\mathrm{Lie}}$ of finite $\mathbb{Z}_p[\Gamma_{L}]$-modules $H$, satisfying some additional (Lie)-condition on the image of $\Gamma_L$ in $\operatorname{Aut}_{\mathbb{Z}_p}H$. In the paper it is proved that all information about the images of the groups $\Gamma_L^{(v)}$ in $\operatorname{Aut}_{\mathbb{Z}_p}H$ can be explicitly extracted from some differential forms $\widetilde{\Omega} [N]$ on the Fontaine etale $\phi $-module $M(H)$ associated with $H$. The forms $\widetilde{\Omega}[N]$ are completely determined by a canonical connection $\nabla $ on $M(H)$. In the case of fields $L$ of mixed characteristic, which contain a primitive $p$th root of unity, we show that a similar problem for $\mathbb{F}_p[\Gamma_L]$-modules also admits a solution. In this case we use the field-of-norms functor to construct the corresponding $\phi $-module together with the action of the Galois group of a cyclic extension $L_1$ of $L$ of degree $p$. Then our solution involves the characteristic $p$ part (provided by the field-of-norms functor) and the condition for a “good” lift of a generator of $\operatorname{Gal}(L_1/L)$. Apart from the above differential forms the statement of this condition uses the power series coming from the $p$-adic period of the formal group $\mathbb{G}_m$.
Keywords: local field, Galois group, ramification filtration.
Received: 10.02.2022
Revised: 02.11.2022
Russian version:
Izvestiya Rossiiskoi Akademii Nauk. Seriya Matematicheskaya, 2023, Volume 87, Issue 3, Pages 5–22
DOI: https://doi.org/10.4213/im9322
Bibliographic databases:
Document Type: Article
UDC: 512.625
MSC: 11S20, 11S15, 11R32
Language: English
Original paper language: English

Introduction

Let $L$ be a complete discrete valuation field with finite residue field of characteristic $p$. Let $\Gamma_L$ be the absolute Galois group of $L$. Let $\{\Gamma_L^{(v)}\}_{v\geqslant 0}$ be the filtration of $\Gamma_L$ by the ramification subgroups, [1]. This filtration provides $\Gamma_L$ with additional structure and allows us to introduce various classes of infinite field extensions (arithmetically profinite, deeply ramified etc.), which play an important role in modern arithmetic algebraic geometry. For $\Gamma_L$-modules $H$, the evaluation of $v_0(H)\in\mathbb{Q} $ such that $\Gamma_L^{(v)}$ act trivially on $H$ for $v>v_0(H)$, provides us with good estimates for discriminants of the fields of definition of $h\in H$. Such estimates are used very often to answer various number theoretic questions.

However, an explicit description of the structure of the ramification filtration for a very long time was known only at the level of the Galois groups of abelian field extensions. At the time when the structure of the Galois group $\Gamma_L$ was completely described (the case of the maximal $p$-extensions – Shafarevich [2], Demushkin [3], and the general case – Janssen–Wingberg [4]) it became clear that $\Gamma_L$ is a very weak invariant of the field $L$. The situation cardinally changed later when it was established (Mochizuki [5], author [6]) that taking $\{\Gamma_L^{(v)}\}_{v\geqslant 0}$ into account gives us absolute invariant of the field $L$. However, in order to work with this invariant we still need to know an explicit description of the filtration by the groups $\Gamma_L^{(v)}$.

I. R. Shafarevich always paid attention to this problem, for example, cf. Introduction to [7]. His motivation was the following: for every prime number $p$ there is only one such filtration and we know almost nothing about its structure. In 1990’s the author developed a nilpotent version of the Artin–Schreier theory and obtained an explicit description of the ramification filtration modulo the subgroup of $p$th commutators of $\Gamma_L$. Such description was obtained, first, in the characteristic $p$ case, [8]–[10], and then developed in the mixed characteristic case, [11]–[13]. These results play a crucial role in this paper, where we study the images of the ramification subgroups $\Gamma_L^{(v)}$ in the group of automorphisms $\operatorname{Aut}_{\mathbb{Z}_p}H$ of finite $\mathbb{Z}_p[\Gamma_L]$-modules $H$. Our main result states that this arithmetic structure can be completely described (under some additional condition) in purely geometric properties of Fontaine’s etale $\phi $-modules $M(H)$.

More precisely, if $\operatorname{char}L=p$ we construct differential forms $\widetilde{\Omega} [N]$, $N\in\mathbb{N} $, on an extension of scalars of $M(H)$, and specify the way how the image of the ramification filtration in $\operatorname{Aut}_{\mathbb{Z}_p}H$ can be recovered from these forms. Note that the definition of $\widetilde{\Omega} [N]$ depends only on a natural connection constructed on $M(H)$. If $\operatorname{char}L=0$ we assume that $L$ contains a $p$th root of unity $\zeta_1\ne 1$ and restrict ourselves to the case of Galois $\mathbb{F}_p$-modules. Then we apply the field-of-norms functor to reduce the situation to the characteristic $p$ case and use a characterization of “good” lifts of automorphisms of our cyclic field extension of $L$ from [11], [12]. This characterization uses again the differential forms $\widetilde{\Omega} [N]$ and a power series coming from the $p$-adic period of the formal group $\mathbb{G}_m$.

By our opinion this result establishes quite interesting link between the Galois theory of local fields and very popular area of $D$-modules, lifts of Frobenius, Higgs vector bundles etc.

§ 1. Statement of the main result

1.1. General notation

Everywhere in the paper $p$ is a fixed prime number. If $E_0$ is a field then $E_0^{\mathrm{sep}}$ is its separable closure in some algebraic closure $E_0^{\mathrm{alg}}$ of $E_0$. If $E$ is a field such that $E_0\subset E\subset E_0^{\mathrm{sep}}$ set $\Gamma_{E}=\operatorname{Gal}(E_0^{\mathrm{sep}}/E)$. The field $E_0^{\mathrm{sep}}$ will be considered as a left $\Gamma_{E_0}$-module, i. e., for any $\tau_1,\tau_2\in\Gamma_{E_0}$ and $o\in E_0^{\mathrm{sep}}$, $(\tau_1\tau_2)o=\tau_1(\tau_2o)$. If $\operatorname{char}E_0=p$ we set for any $a\in E_0^{\mathrm{sep}}$, $\sigma (a)=a^p$.

If $V$ is a module over a ring $R$ then $\operatorname{End}_RV$ is the $R$-algebra of $R$-linear endomorphisms of $V$. We always consider $V$ as a left $\operatorname{End}_RV$-module, i. e., if $l_1,l_2\in\operatorname{End}_RV$ and $v\in V$ then $(l_1l_2)v=l_1(l_2(v))$. We also consider $\operatorname{End}_RV$ as a Lie $R$-algebra with the Lie bracket $[l_1,l_2]=l_1l_2-l_2l_1$. If $S$ is an $R$-module then we often denote by $V_S$ the extension of scalars $V\otimes_RS$.

1.2. Functorial system of lifts to characteristic $0$

Suppose $K_0=k_0((t_0))$ is the field of formal Laurent series in a (fixed) variable $t_0$ with coefficients in a finite field $k_0$ of characteristic $p$. The uniformiser $t_0$ provides a $p$-basis for any field extension $ E$ of $ K_0$ in $ K_0^{\mathrm{sep}}$, i. e., the set $\{1,t_0,\dots,t_0^{p-1}\}$ is a $E^p$-basis of $ E$. We use this $p$-basis to construct a compatible system of lifts $O(E)$ of the fields $E$ to characteristic $0$. This is a special case of the construction of lifts from [14]; it can be explained as follows.

For all $M\in\mathbb{N} $, set $O_{M}(E)=W_{M}(\sigma^{M-1} E)[\overline t_0]$, where $\overline t_0=[t_0]$ is the Teichmuller representative of $t_0$ in the ring of Witt vectors $W_{M}(E)$. The rings $O_M(E)$ are the lifts of $E$ modulo $p^M$, i. e., they are flat $\mathbb{Z} /p^M$-algebras such that $O_M(E)\otimes_{\mathbb{Z} /p^M}\mathbb{Z} /p= E$. Note that the system

$$ \begin{equation*} \{O_M(E)\mid M\in\mathbb{N},\, K_0\subset E\subset K_0^{\mathrm{sep}}\} \end{equation*} \notag $$
is functorial in $M$ and $ E$. In particular, if $E/K_0$ is Galois then there is a natural action of $\operatorname{Gal}(E/ K_0)$ on $O_M(E)$ and $O_M(E)^{\operatorname{Gal}(E/K_0)}=O_M(K_0)$. The morphisms $W_M(\sigma)$ induce $\sigma $-linear morphisms on $O_M(E)$ which will be denoted again by $\sigma $. In particular, $\sigma (\overline t_0)=\overline t_0^{\,p}$.

Introduce the lifts of the above fields $E$ to characteristic $0$ by setting $O(E)=\varprojlim_M O_M(E)$. Then $O_M(E)=O(E)/p^M$ and $O(E)[1/p]$ is a complete discrete valuation field with uniformiser $p$ and the residue field $E$. Clearly, we have the induced morphism $\sigma $ on each $O(E)$. Also, if $ E/ K_0$ is Galois then $\operatorname{Gal}(E/ K_0)$ acts on $O(E)$ and $O(E)^{\operatorname{Gal}(E/ K_0)}=O(K_0)$. Notice that $O(K_0)=\varprojlim_M W_M(k_0)((\overline t_0))$ is the completion of the ring of formal Laurent series $W(k)((\overline t_0))$.

Set $O_{\mathrm{sep}}=O(K_0^{\mathrm{sep}})$.

The system of lifts $O(E)$, $ E\subset K_0^{\mathrm{sep}}$, can be extended to the system of lifts of all extensions of $ K_0$ in $K_0^{\mathrm{alg}}$. Indeed, note that $K_0^{\mathrm{alg}}=\bigcup_{n\geqslant 0}K_0(t_n)^{\mathrm{sep}}$, where $t_n^{p^n}=t_0$. Then $t_n$ gives the $p$-basis $\{1,t_n,\dots,t_n^{p-1}\}$ for all separable extensions $E$ of $K_0(t_n)$ in $K_0^{\mathrm{alg}}$ and we obtain (as earlier) the corresponding lifts $O(E)$. The system of lifts $O(E)$ is functorial in $E\subset K_0^{\mathrm{alg}}$ (use that any separable extension $ E_n$ of $ K_0(t_n)$ appears uniquely as the composite $ E_0 K_0(t_n)$, where $ E_0/ K_0$ is separable). In particular, consider $K_0^{\mathrm{rad}}=\bigcup_{N\in\mathbb{N}} K_0^{\mathrm{ur}}(t_0^{1/N})$. Then for the above defined lift of $ K_0^{\mathrm{rad}}$ we have $O(K_0^{\mathrm{rad}})=\bigcup_{N\in\mathbb{N}}O(K_0^{\mathrm{ur}})[\overline t_0^{1/N}]$, where $K_0^{\mathrm{ur}}=\overline k_0((t_0))$ is the maximal unramified extension of $ K_0$.

1.3. Equivalence of the categories of $p$-groups and Lie algebras [15]

Let $L$ be a finitely generated Lie $\mathbb{Z}_p$-algebra of nilpotent class $<p$, i. e., the ideal of $p$ th commutators $C_p(L)$ of $L$ is equal to zero. Let $A$ be an enveloping algebra of $L$. Then the elements of $L\subset A$ generate the augmentation ideal $J$ of $A$. There is a morphism of $\mathbb{Z}_p$-algebras $\Delta\colon A\to A\otimes A$ uniquely determined by the condition $\Delta (l)=l\otimes 1+1\otimes l$ for all $l\in L$. Then the set ${\exp}(L)\ \operatorname{mod}J^p$ is identified with the group of all “diagonal elements modulo degree $p$” consisting of $a\in 1+J\ \operatorname{mod}J^p$ such that $\Delta (a) \equiv a\otimes a\ \operatorname{mod}(J\otimes 1+1\otimes J)^p$.

In particular, there is a natural embedding $L\subset A/J^p$ and the identity

$$ \begin{equation*} \exp(l_1)\cdot \exp(l_2)\equiv \exp(l_1 \circ l_2)\ \operatorname{mod}J^p \end{equation*} \notag $$
induces the Campbell–Hausdorff composition law
$$ \begin{equation*} (l_1,l_2)\mapsto l_1 \circ l_2= l_1+l_2+\frac{1}{2}[l_1,l_2]+\cdots,\qquad l_1,l_2\in L. \end{equation*} \notag $$
This composition law provides the set $L$ with a group structure. We denote this group by $G(L)$. Clearly $G(L)\simeq {\exp}(L)\ \operatorname{mod}J^p$.

With the above notation the functor $L\mapsto G(L)$ determines the equivalence of the categories of finitely generated $\mathbb{Z}_p$-Lie algebras and profinite $p$-groups of nilpotence class $<p$. Note that a subset $I\subset L$ is an ideal in $L$ if and only if $G(I)$ is a normal subgroup in $G(L)$.

1.4. Lie-condition

For any finite field extension $ K$ of $ K_0$ in $ K_0^{\mathrm{sep}}$, let $\operatorname{M\Gamma}_{K}$ be the category of finitely generated $\mathbb{Z}_p$-modules $H$ with continuous left action of $\Gamma_{K}$. Each element $h\in H$ is defined over some finite extension $K(h)$ of $K$. In some sense the family of these fields determines “arithmetic” properties of $H$. More detailed information about the fields $K(h)$ can be obtained from the knowledge of the images of the ramification subgroups in upper numbering $\Gamma_{K}^{(v)}$, $v>0$, in $\operatorname{Aut}_{\mathbb{Z}_p} H$. For example, the minimal number $v_0(H)\in\mathbb{Q} $ such that all $\Gamma_{K}^{(v)}$ with $v>v_0(H)$ act trivially on $H$ provides us with upper estimates for the discriminants of the fields of definition of $h\in H$ (cf. [6]).

Let $ H_0\in\operatorname{M\Gamma}_{K_0}$ and let $\pi_{H_0}\colon\Gamma_{K_0}\to \operatorname{Aut}_{\mathbb{Z}_p} H_0$ be the group homomorphism which determines the $\Gamma_{K_0}$-module structure on $ H_0$. Consider the full subcategory $\operatorname{M\Gamma}_{K_0}^{\mathrm{Lie}}$ in $\operatorname{M\Gamma}_{K_0}$ which consists of modules $ H_0$ satisfying the following condition.

Condition (Lie). The image $I(H_0):=\pi_{H_0}(\mathcal I)\subset \operatorname{Aut}_{\mathbb{Z}_p}(H_0)$ of the wild inertia subgroup $\mathcal I\subset\Gamma_{K_0}$ appears in the form $\exp(L(H_0))$, where $L(H_0)\subset\operatorname{End}_{\mathbb{Z}_p}H_0$ is a Lie subalgebra such that $L(H_0)^p=0$.

The condition $L(H_0)^p=\{l_1\cdots l_p\mid l_1,\dots,l_p\in L(H_0)\}=0$ (the product is taken in $\operatorname{End}_{\mathbb{Z}_p} H_0\supset L(H_0)$) implies that $L(H_0)$ is a finitely generated nilpotent $\mathbb{Z}_p$-algebra Lie of nilpotence class $<p$. This gives the group isomorphism $\exp\colon G(L(H_0)) \simeq I(H_0)$. Note that any normal subgroup of $I(H_0)$ appears in the form $\exp G(J)$, where $J$ is a Lie ideal of $L(H_0)$.

Remark 1.1. If $p H_0=0$ and $\operatorname{dim}_{\mathbb{F}_p} H_0\leqslant p$ then $ H_0$ is of the Lie type.

1.5. The first main result: the characteristic $p$ case

Suppose $ H_0\,{\in}\operatorname{M\Gamma}_{K_0}^{\mathrm{Lie}}$. Our target is to determine for all $v> 0$, the images $\pi_{H_0}(\Gamma_{K_0}^{(v)})$ of the ramification subgroups $\Gamma_{K_0}^{(v)}$ via an explicit construction of the ideals $L(H_0)^{(v)}\subset L(H_0)$ such that $\exp (L(H_0)^{(v)})= \pi_{H_0}(\Gamma_{K_0}^{(v)})$.

Our approach uses Fontaine’s “analytical” description of the Galois modules $H_0\in\operatorname{M\Gamma}_{K_0}^{\mathrm{Lie}}$ in terms of etale $(\phi,O(K_0))$-modules $M(H_0)$. A geometric nature of $M(H_0)$ is supported by the existence of an analogue of the classical connection $\nabla\colon M(H_0)\to M(H_0)\otimes_{O(K_0)}\Omega^1_{O(K_0)}$ (cf. [16]). (This map is uniquely characterized by the condition $\nabla\phi =(\phi \otimes\phi)\cdot \nabla $.) The required information about the behaviour of ramification subgroups can be then extracted from some differential forms

$$ \begin{equation*} \widetilde{\Omega}[N]\in M(H_0)_{O(K_0^{\mathrm{rad}})} \otimes_{O(K_0)} \Omega^1_{O(K_0)}. \end{equation*} \notag $$
The construction of these differential forms is given completely in terms of the above connection $\nabla $ and can be explained as follows.

Let $ K\subset K_0^{\mathrm{sep}}$ be a fixed tamely ramified finite extension of $K_0$ such that $\pi_{H_0}(\Gamma_{K})=I(H_0)$. Then $H:=H_0|_{\Gamma_{K}}$ can be described via an etale $(\phi,O(K))$-module $M(H)=M(H_0)\otimes_{O(K_0)}O(K)$. Recall that $M(H)=(H\otimes_{\mathbb{Z}_p}O_{\mathrm{sep}})^{\Gamma_{K}}$ is a finitely generated $O(K)$-module with a $\sigma $-linear morphism $\phi\colon M(H)\to M(H)$ such that the image $\phi(M(H))$ generates $M(H)$ over $O(K)$. This allows us to identify the elements of $H$ with a set of $O_{\mathrm{sep}}$-solutions of a suitable system of equations with coefficients in $O(K)$.

We establish below the construction of $M(H)$ by introducing a $\mathbb{Z}_p$-linear embedding $\mathcal F\colon H\to M(H)$ which induces by extension of scalars the identification $H_{O(K)}\simeq M(H)$. (We denote this identification by the same symbol $\mathcal F$.)

Now let $\widetilde{B}$ be (a inique) $O(K)$-linear operator on $M(H)$ such that for any $m\in \mathcal F(H)$, $\nabla (m)=\widetilde{B}(m)\,d{\overline t}/\overline t$. Then for every $N\in\mathbb{Z}_{\geqslant 0}$ we introduce the differential forms

$$ \begin{equation*} \widetilde{\Omega}[N]=\phi^N\widetilde{B}\phi^{-N}\, \frac{d\overline t}{\overline t}\in \operatorname{End}M(H)_{O(K^{\mathrm{rad}})}\otimes_{O(K)}\Omega^1_{O(K)}. \end{equation*} \notag $$

Now we can use the identification $\mathcal F\colon H_{O(K)}\simeq M(H)$ to obtain the corresponding differential forms $\Omega [N]$ on $\operatorname{End}(H)_{O(K^{\mathrm{rad}})}$ and to verify that

$$ \begin{equation*} \Omega [N]\in L(H)_{O(K^{\mathrm{rad}})} \otimes_{O(K)}\Omega^1_{O(K)}= L(H_0)_{O(K_0^{\mathrm{rad}})} \otimes_{O(K_0)}\Omega^1_{O(K_0)}. \end{equation*} \notag $$

Remark 1.2. Our differential forms will usually appear in the form $\Omega =F\cdot d\overline t_0/\overline t_0$, where $F\in L(H)_{K_0^{\mathrm{rad}}}$. Then we set by definition

$$ \begin{equation*} (\operatorname{id}_{L(H)}\otimes \, \sigma)\Omega = (\operatorname{id}_{L(H)}\otimes\, \sigma)F\cdot \frac{d\overline t_0}{\overline t_0}. \end{equation*} \notag $$

Our first main result can be stated as follows.

Theorem 1.1. Suppose $H_0\in\operatorname{M\Gamma}_{K_0}^{\mathrm{Lie}}$ is finite. Then there is $N_0(H_0)\in\mathbb{N}$ such that for any (fixed) $N\geqslant N_0(H_0)$ the following property holds:

if $(\operatorname{id}_{L(H)}\otimes\,\sigma^{-N})\Omega [N]= \sum_{r\in\mathbb{Q}}\overline t_0^{\,-r}l_{r}\,d\overline t_0/\overline t_0$, where all $l_r\in L(H_0)_{W(\overline{k}_0)}$, then the ideal $L(H_0)^{(v)}$ is the minimal ideal in $L(H_0)$ such that for all $r\geqslant v$, $l_r\in L(H_0)^{(v)}_{W(\overline{k}_0)}$.

Corollary 1.1. If $v_0(H_0)=\max\{r\mid l_r\ne 0\}$ then the ramification subgroups $\Gamma_{K_0}^{(v)}$ act trivially on $H_0$ if and only if $v>v_0(H_0)$.

Remark 1.3. The construction of $\Omega [N]$ almost does not depend on the choice of the tamely ramified finite field extension $ K$ of $ K_0$. It depends essentially on the choice of the uniformising element $t_0$ in $ K_0$ and a compatible system of $\alpha (k)\in W(k)$, where $[k:k_0]<\infty $, such that the trace of $\alpha (k)$ in the field extension $W(k)[1/p]/{K}_0$ equals $1$.

Remark 1.4. If $H_0$ is not $p$-torsion Theorem 1.1 can be applied to the factors $H_0/p^M$ and our result describes the structure of the images of $L(H_0)^{(v)}$ in all $L(H_0)/p^M$.

1.6. The second main result: the mixed characteristic case

Let $E_0$ be a finite field extension of $\mathbb{Q}_p$ with residue field $k_0$ and a uniformizing element $\pi_0$. Assume that $E_0$ contains a $p$th primitive root of unity $\zeta_1$. Consider the category $\operatorname{M\Gamma}_{E_0,1}^{\mathrm{Lie}}$ of finitely generated $\mathbb{F}_p[\Gamma_{E_0}]$-modules which satisfy a direct analog of the Lie condition from § 1.4.

Take the infinite arithmetically profinite field extension $\widetilde{E}_0$ obtained from $E_0$ by joining all $p$-power roots of $\pi_0$. Then the theory of the field-of-norms functor $X$ provides us with the complete discrete valuation field of characteristic $p$, $X(\widetilde{E}_0)\,{=}\,K_0$, which has the same residue field and the uniformizing element $t_0$ obtained from the $p$-power roots of $\pi_0$. The functor $X$ also provides us with the identification of Galois groups $\Gamma_{K_0}=\Gamma_{\widetilde{E}_0}\subset \Gamma_{E_0}$.

If $H_{E_0}\in\operatorname{M\Gamma}^{\mathrm{Lie}}_{E_0,1}$ then we obtain $H_0:=H_{E_0}|_{\Gamma_{K_0}} \in\operatorname{M\Gamma}_{K_0}^{\mathrm{Lie}}$. As earlier, take a finite tamely ramified extension $K$ of $K_0$ (it corresponds to a unique tamely ramified extension $E$ of $E_0$ with uniformizer $\pi $ such that $\pi^{e_0}=\pi_0$, where $e_0$ is the ramification index of $E/E_0$), and construct $(\phi,O(K))$-module $M(H)$. This module inherits the action of $\operatorname{Gal}(E(\sqrt[p]{\pi})/E)= \langle\tau_0\rangle^{\mathbb{Z} /p}$. (Here $\tau_0$ is such that $\tau_0(\sqrt[p]{\pi})= \zeta_1\sqrt[p]{\pi}$.) This situation was considered in all details in the papers [11], [12]. In particular, in those papers we gave a characterization of the “good” lifts $\widehat\tau_0$ of $\tau_0$. By definition, $\widehat\tau_0\in\Gamma_{E}$ is “good” if its restriction to $H_{E_0}$ belongs to the image of the ramification subgroup $\Gamma_{E}^{(e^*)}$. Here $e^*:=pe/(p-1)$ and $e=e(E/\mathbb{Q}_p)$ is the ramification index of $E/\mathbb{Q}_p$. (This makes sense because $\tau_0\in \operatorname{Gal}(E(\sqrt[p]{\pi}/E)^{(e^*)})$.) Note that the field-of-norms functor is compatible with ramification filtrations on $\Gamma_{E_0}$ and $\Gamma_{K_0}$. Therefore, the knowledge of “good lifts” $\widehat\tau_0$ together with Theorem 1.1 gives a complete description of the image of the ramification filtration of $\Gamma_{E_0}$ in $\operatorname{Aut}H_{E_0}$.

In [12] we proved that the action of $\widehat\tau_0$ appears from an action of a formal group scheme of order $p$. As a result, the lift $\widehat\tau_0$ is completely determined by the value $d\widehat\tau_0(0)\in L(H)$ of its differential at $0$, and we can use the characterization of differentials of “good” lifts from [12], Theorem 5.1.

Namely, let us first specify our $p$th root of unity

$$ \begin{equation*} \zeta_1=1+\sum_{j\geqslant 0}[\beta_j] \pi^{(e^*/p)+j}\ \operatorname{mod}p \end{equation*} \notag $$
(here all $[\beta_j]$ are the Teichmuller representatives of elements from the residue field of $E$). Then we introduce the power series $\omega (t)\in O(K)$ such that
$$ \begin{equation*} 1+\sum_{j\geqslant 0}\beta_j^pt^{e^*+pj}= \widetilde{\exp}(\omega (t)^p) \end{equation*} \notag $$
(here $\widetilde{\exp}$ is the truncated exponential). The series $\omega (t)^p$ is a kind of approximation of the $p$-adic period of the formal multiplicative group, which appears usually in explicit formulas for the Hilbert symbol (e.g., [17]). In other words, we obtain another geometric condition characterizing “arithmetic” of the $\Gamma_{E_0}$-module $H_{E_0}$.

Theorem 1.2. The lift $\widehat\tau_0$ is “good ” if and only if

$$ \begin{equation*} d\widehat\tau_0(0)\equiv \sum_{m\geqslant 0} \operatorname{Res} \bigl(\omega (t)^{p^{m+1}}\Omega [m]\bigr) \ \operatorname{mod}L(H)_{k}^{(e^*)}. \end{equation*} \notag $$

Remark 1.5. Notice that the power series $\omega (t)^p$ has non-zero coefficients only for the powers $t^{e^*+pj}$ and all these exponents $\geqslant e^*$. Therefore, the differential forms $\Omega [m]$ contribute to the right-hand side only via the images of $\mathcal F^0_{e^*+pj,-m}t^{-(e^*+pj)}$ in $L(H)_k$. But for $m\gg 0$, these images belong to the images of the ramification ideals $\mathcal L^{(e^*)}_k$ and, therefore, disappear modulo $L(H)_{k}^{(e^*)}$, and the sum in the right-hand side is, as a matter of fact, finite.

§ 2. $\phi$-module $M(H)$

2.1. Specification of $\log \pi_{H}\colon \Gamma_{K}\to G(L(H))$

As earlier, $H_0\in\operatorname{M\Gamma}_{K_0}^{\mathrm{Lie}}$, $K$ is a finite tamely ramified extension of $K_0$ in $K_0^{\mathrm{sep}}$ such that $\pi_{H_0}(\Gamma_{K})=I(H_0)$, $H=H_0|_{\Gamma_{K}}$. Set $L(H)=L(H_0)$, $\pi_{H}=\pi_{H_0}|_{\Gamma_{K}}$.

Consider the continuous group epimorphism $l_{H}:=\log(\pi_{H})\colon \Gamma_{K}\to G(L(H))$. Since the $p$-group $G(L(H))$ has nilpotence class $<p$ this epimorphism can be described in terms of the covariant version of the nilpotent Artin–Schreier theory from [9]. Namely, there are $e\in L(H)_{O(K)}$ and $f\in L(H)_{O_{\mathrm{sep}}}$ such that $(\operatorname{id}_{L(H)}\otimes\,\sigma)(f)= e \circ f$ and for any $\tau\in\Gamma_{K}$, $l_{H}(\tau)=(-f) \circ (\operatorname{id}_{L(H)}\otimes\tau)f$.

It could be easily verified that $l_{H}$ is a group homomorphism. Indeed,

$$ \begin{equation*} \begin{aligned} \, l_{H}(\tau_1\tau_2) &= (-f) \circ (\operatorname{id}_{L(H)}\otimes\, \tau_1\tau_2)f \\ &=(-f) \circ (\operatorname{id}_{L(H)}\otimes\,\tau_1)f \circ (-f) \circ (\operatorname{id}_{L(H)}\otimes\,\tau_2)f= l_{H}(\tau_1)\circ l_{H}(\tau_2), \end{aligned} \end{equation*} \notag $$
because $(\operatorname{id}_{L(H)}\otimes\, \tau_1)l_{H}(\tau_2)=l_{H}(\tau_2)$.

Notation. We will use below the following notation: $\sigma_H=\operatorname{id}_H\otimes\,\sigma $ and $\sigma_{L(H)}=\operatorname{id}_{L(H)}\otimes\, \sigma $. For example, if $u=\sum_{\alpha}h_{\alpha}\otimes\, o_{\alpha}$, where all $h_{\alpha}\in H$ and $o_{\alpha}\in O(K)$ then $\sigma_H(u)=\sum_{\alpha}h_{\alpha}\otimes \sigma (o_{\alpha})$. Or, if $X$ is a linear operator on $L(H)_{O(K)}$ then $\sigma_{L(H)}X$ is also a linear operator such that

$$ \begin{equation*} \sigma_{L(H)}X\biggl(\sum_{\alpha}h_{\alpha}\otimes o_{\alpha}\biggr)= \sum_{\alpha}\sigma_H(X(h_{\alpha}))(1\otimes o_{\alpha}). \end{equation*} \notag $$
In addition, $\mathcal X:=X\cdot\sigma_H$ is a unique $\sigma$-linear operator such that $\mathcal X|_H=X|_H$, and we have the following identity: $\sigma_H \cdot X=\sigma_{LH}(X)\cdot\sigma_H$. If there is no risk of confusion we will use just the notation $\sigma$.

Remark 2.1. Originally we developed in [9] the contravariant version of the nilpotent Artin–Schreier theory, cf. the discussion in [18]. The contravariant version uses similar relations $\sigma_{L(H)}(f)= f \circ e$ and the map $l_{H}$ was defined via $\tau\mapsto (\operatorname{id}_{L(H)}\otimes\,\tau) f \circ (-f)$. In this case $l_{H}$ determines the group homomorphism from $\Gamma_{K}$ to the opposite group $G^0(L(H))$ (this group is isomorphic to $G(L(H))$ via the map $g\mapsto g^{-1}$). The results from the papers [8]–[10] were obtained in terms of the contravariant version, but the results from [11]–[13], [19] used the covariant version. We can easily switch from one theory to another via the automorphism $-\operatorname{id}_{L(H)}$.

We consider $O_{\mathrm{sep}}$ as a left $\Gamma_{K}$-module via the action $o\mapsto \tau (o)$ with $o\in O_{\mathrm{sep}}$ and $\tau\in\Gamma_{K}$. This corresponds to our earlier agreement about the left action of the elements of $\Gamma_{K}$ as endomoprhisms of $O_{\mathrm{sep}}$. As a result we obtain the left $\Gamma_{K}$-module structure on $H_{O_{\mathrm{sep}}}$ by the use of the (left) action of $\Gamma_{K}$ on $H$ via $h\mapsto l_{H}(\tau)(h)$.

Because $L(H)_{O_{\mathrm{sep}}}\subset \operatorname{End}_{O_{\mathrm{sep}}}(H_{O_{\mathrm{sep}}})$ we can introduce for any $h\in H$, $\mathcal F(h):=\exp (-f)(h)\in H_{O_{\mathrm{sep}}}$.

Proposition 2.1. For any $h\in H$, $\mathcal F(h)\in (H_{O_{\mathrm{sep}}})^{\Gamma_{K}}$.

Proof. Suppose $f=\sum_{\alpha} l_{\alpha}\otimes o_{\alpha}$, where all $l_{\alpha}\in L(H)$ and $o_{\alpha}\in O_{\mathrm{sep}}$.

If $\tau\in\Gamma_{K}$ then

$$ \begin{equation*} \begin{aligned} \, \tau (\mathcal F(h)) &=(\tau\otimes\operatorname{id}_{O_{\mathrm{sep}}})\biggl(\exp \biggl( -\sum_{\alpha}l_{\alpha}\otimes \tau (o_{\alpha})\biggr)(h)\biggr) \\ &=(\tau\otimes\operatorname{id}_{O_{\mathrm{sep}}})\bigl(\exp \bigl( (\operatorname{id}_{L(H)}\otimes\tau)(-f)\bigr)(h)\bigr) \\ &=(\tau\otimes\operatorname{id}_{O_{\mathrm{sep}}})\bigl(\exp \bigl((-l_{H}(\tau)) \circ (-f)\bigr)(h)\bigr) \\ &=(\tau\otimes\operatorname{id}_{O_{\mathrm{sep}}})\bigl(\bigl(\exp (-l_{H}(\tau))\cdot \exp (-f)\bigr)h \bigr) \\ &=(\tau\otimes\operatorname{id}_{O_{\mathrm{sep}}})(\pi_{H}(\tau^{-1})\otimes \operatorname{id}_{O_{\mathrm{sep}}})\mathcal F(h) \\ &=\bigl(\tau \cdot \pi_H(\tau^{-1})\otimes \operatorname{id}_{O_{\mathrm{sep}}} \bigr) \mathcal F(h)=\mathcal F(h). \end{aligned} \end{equation*} \notag $$
The proposition is proved.

The elements $e$ and $f$ are not determined uniquely by $l_{H}$. A pair $e'\in L(H)_{O(K)}$ and $f'\in L(H)_{O_{\mathrm{sep}}}$ give the same group epimorphism $l_{H}$ if and only if there is $x\in L(H)_{O(K)}$ such that $e'=\sigma (x) \circ e \circ (-x)$ and $f'=x \circ f$.

2.2. Special choice of $e\in L(H)_{O(K)}$

We can always assume (by replacing, if necessary, $ K$ by its finite unramified extension) that a uniformiser $t$ in $ K$ is such that $t^{e_0}=t_0$, where $e_0$ is the ramification index for $ K/ K_0$. Then $O(K)=\varprojlim_M W_M(k)((\overline t))$, where $\overline t^{\,e_0}=\overline t_0$ and $\overline t$ is the Teichmuller representative of $t$. We denote by $k$ the residue field of $ K$ and fix a choice of $\alpha_0=\alpha (k)\in W(k)$ such that its trace in the field extension $W(k)[1/p]/\mathbb{Q}_p$ equals $1$.

Let $\mathbb{Z}^+(p):=\{a\in\mathbb{N} \mid \operatorname{gcd}(a,p)=1\}$ and $\mathbb{Z}^0(p)=\mathbb{Z}^+(p)\cup\{0\}$.

Definition 2.1. An element $e\in L(H)_{O(K)}$ is special if $e=\sum_{a\in\mathbb{Z}^0(p)}\overline t^{\,-a}l_{a0}$, where $l_{00}\in\alpha_0L(H)$ and for all $a\in\mathbb{Z}^+(p)$, $l_{a0}\in L(H)_{W(k)}$.

Lemma 2.1. Suppose $e\in L(H)_{O(K)}$. Then there is $x\in L(H)_{O(K)}$ such that $\sigma (x) \circ e \circ (-x)$ is special.

Proof. Use induction on $s$ to prove lemma modulo the ideals of $s$th commutators $C_s(L(H)_{O(K)})$.

If $s=1$ there is nothing to prove.

Suppose lemma is proved modulo $C_s(L(H)_{O(K)})$.

Then there is $x\in L(H)_{O(K)}$ such that

$$ \begin{equation*} \sigma (x) \circ e \circ (-x) = \sum_{a\in\mathbb{Z}^0(p)}\overline t^{\,-a}l_{a0}+l, \end{equation*} \notag $$
where $l\in C_s(L(H)_{O(K)})$. Using that
$$ \begin{equation} O(K)=(\sigma - \operatorname{id}_{O(K)})O(K)\oplus (\mathbb{Z}_p\alpha_0) \oplus \biggl(\sum_{a\in\mathbb{Z}^+(p)}W(k)\overline t^{\,-a}\biggr), \end{equation} \tag{2.1} $$
we obtain the existence of $x_s\in C_s(L(H)_{O(K)})$ such that
$$ \begin{equation*} l=\sigma (x_s)-x_s+\sum_{a\in\mathbb{Z}^0(p)}\overline t^{\,-a}l_a, \end{equation*} \notag $$
where $l_0\in\alpha_0L$ and all remaining $l_a\in L_{W(k)}$. Then we can take $x'=x-x_s$ to obtain our statement modulo $C_{s+1}(L(H)_{O(K)})$.

Lemma 2.2. Suppose $e\in L_{O(K)}$ is special and $x\in L_{O(K)}$. Then the element $\sigma (x) \circ e \circ (-x)$ is special if and only if $x\in L$ (or, equivalently, if $\sigma x=x$).

Proof. Use relation (cf. [20])
$$ \begin{equation*} \operatorname{Ad}(\exp (X))\exp (Y)=\exp \biggl(\sum_{n\geqslant 0} \frac{1}{n!}\operatorname{ad}^n(X)(Y)\biggr) \ \operatorname{mod} (\operatorname{deg} p), \end{equation*} \notag $$
where $\operatorname{Ad}(U)(V)=UVU^{-1}$ and $\operatorname{ad}(U)(V)=[U,V]$. Indeed, if $X=x\in L(H)$ and $Y=e$ then $\sum_{n\geqslant 0}\operatorname{ad}^n(x)(e)/n!$ is also special.

When proving the inverse statement we can use induction modulo the ideals $C_s(L(H))_{O(K)}$, $s\geqslant 1$, as follows.

Assume the lemma is proved modulo $C_s(L(H)_{O(K)})$. Then using the if part we can assume that $x\in C_s(L(H)_{O(K)})$. Therefore, $e+\sigma (x)-x$ is special modulo $C_{s+1}(L(H))_{O(K)}$, i. e.,

$$ \begin{equation*} \sigma (x)-x\in \alpha_0C_s(L)+\sum_{a\in\mathbb{Z}^+(p)} t^{-a}C_s(L)_{W(k)} \end{equation*} \notag $$
modulo $C_{s+1}(L(H)_{O(K)})$. By (2.1) this implies the congruence
$$ \begin{equation*} \sigma (x)\equiv x\ \operatorname{mod}C_{s+1}(L(H))_{O(K)}, \end{equation*} \notag $$
i. e., $x\in C_s(L(H))\ \operatorname{mod}C_{s+1}(L(H)_{O(K)})$.

The lemma is proved.

2.3. Construction of the $\phi $-module $M(H)$

Note that $\pi_{H}=\exp (l_{H})$, and therefore for all $\tau\in\Gamma_{K}$, it holds

$$ \begin{equation*} \pi_{H}(\tau)= \exp (-f)\cdot \exp (\operatorname{id}_{L(H)}\otimes \,\tau)f, \end{equation*} \notag $$
where $f\,{\in}\, L(H)_{O_{\mathrm{sep}}}\,{\subset} \operatorname{End}_{O_{\mathrm{sep}}}(H_{O_{\mathrm{sep}}})$, $\sigma_{L(H)}(f)\,{=}\,e\circ f$ and $\exp f\,{\in}\operatorname{Aut}_{O_{\mathrm{sep}}}(H_{O_{\mathrm{sep}}})$.

Let $\operatorname{MF}^{\mathrm{et}}_{K}$ be the category of etale $\phi $-modules over $O(K)$. Recall that its objects are $O(K)$-modules of finite type $M$ together with a $\sigma $-linear morphism $\phi\colon M\to M$ such that its $O(K)$-linear extension $\phi_{O(K)}\colon M\otimes_{O(K),\sigma_{\vphantom{q}}}O(K)\to M$ is isomorphism. The correspondence $H\mapsto M(H):=(H\otimes_{\mathbb{Z}_p} O_{\mathrm{sep}})^{\Gamma_{K}}$, where $\phi\colon M(H)\to M(H)$ comes from the action of $\sigma $ on $O_{\mathrm{sep}}$, determines the equivalence of the categories $\operatorname{M\Gamma}_{K}$ and $\operatorname{MF}^{\mathrm{et}}_{K}$.

Consider the $\mathbb{Z}_p$-linear embedding $\mathcal F\colon H\to H_{O_{\mathrm{sep}}}$ from § 2.1. Let $M(H)=\mathcal F(H)_{O(K)}$. By extension of scalars we obtain natural isomorphisms (use that $O(K)$ and $O_{\mathrm{sep}}$ are flat $\mathbb{Z}_p$-modules):

$$ \begin{equation*} \mathcal F\otimes\operatorname{id}_{O_{\mathrm{sep}}}\colon H_{O_{\mathrm{sep}}} \simeq M(H)_{O_{\mathrm{sep}}},\qquad \mathcal F\otimes \operatorname{id}_{O(K)}\colon H_{O(K)}\simeq M(H), \end{equation*} \notag $$
which will be denoted for simplicity just by $\mathcal F$.

Note that by Proposition 2.1, $M(H)= (H_{O_{\mathrm{sep}}})^{\Gamma_{K}}$.

The $O(K)$-module $M(H)$ is provided with the $\sigma $-linear morphism $\phi \colon M(H)\to M(H)$ uniquely determined for all $h\in H$ via

$$ \begin{equation*} \phi (\mathcal F(h))=\exp (-\sigma_{L(H)}f)(h) =(\exp (-f) \cdot \exp (-e))(h)=\mathcal F(\exp (-e)(h)). \end{equation*} \notag $$

Consider the $O(K)$-linear operator

$$ \begin{equation*} A:=\exp (-e)\in \exp (L(H)_{O(K)})\subset \operatorname{Aut}_{O(K)} H_{O(K)}. \end{equation*} \notag $$
Then $\mathcal A:=A\cdot \sigma_H$ will be a unique $\sigma $-linear operator on $L(H)_{O(K)}$ such that $\mathcal A|_{H}= A|_{H}$. Clearly, for any $u\in H_{O(K)}$,
$$ \begin{equation*} \phi (\mathcal F(u))=\mathcal F(\mathcal A(u)), \end{equation*} \notag $$
and $M(H)$ is etale $\phi $-module associated with the $\mathbb{Z}_p[\Gamma_{K}]$-module $H$.

For example, suppose $pH=0$ and $\{h_i\mid 1\leqslant i\leqslant N\}$ is $\mathbb{F}_p$-basis of $H$. Then $\{\mathcal F(h_i)\mid 1\leqslant i\leqslant N\}$ is a $ K$-basis for $M(H)$. If $A(h_i)=\exp (-e)(h_i)=\sum_ja_{ij}h_j$ with all $a_{ij}\in K$ then $\phi (\mathcal F(h_i))=\sum_ja_{ij}\mathcal F(h_j)$, and $((a_{ij}))$ appears as the corresponding “Frobenius matrix”.

It can be easily seen also that if $\{h_i\mid 1\leqslant i\leqslant N\}$ is a minimal system of $\mathbb{Z}_p$-generators in $H$ then $\{\mathcal F(h_i)\mid 1\leqslant i\leqslant N\}$ is a minimal system of $O(K)$-generators in $M(H)$.

2.4. The connection $\nabla $ on $M(H)$

The $(\phi, O(K))$-module $M(H)$ can be provided with a connection $\nabla \colon M(H)\to M(H)\otimes_{O(K)}\Omega^1_{O(K)}$, [16]. This is an additive map uniquely determined by the properties:

a) for any $o\in O(K)$ and $m\in M(H)$, $\nabla (mo)=\nabla (m)o+m\otimes d(o)$;

b) $\nabla\cdot \phi =(\phi\otimes\phi) \cdot \nabla$.

By a), $\nabla $ is uniquely determined by its restriction to $\mathcal F(H)$ (use that $M(H)=\mathcal F(H)_{O(K)}$). Let $\widetilde{B}$ be a unique $O(K)$-linear operator on $M(H)$ such that for any $m\in\mathcal F(H)$, $\nabla (m)=\widetilde{B}(m)\,d\overline t/\overline t$. Consider the $O(K)$-linear operator $B\in\operatorname{End}H_{O(K)}$ such that for all $u\in H_{O(K)}$, $\widetilde{B}(\mathcal F(u))=\mathcal F(B(u))$. Obviously, $\widetilde B$ and $B$ can be recovered one from another.

Note that for any $\mathbb{Z}_p$-module $\mathcal C$, the elements $c\in \mathcal C_{O(K)}$ appear uniquely in the form $c=\sum_n c_n\otimes \overline t^{\,n}$ with all $c_n\in \mathcal C_{W(k)}$. Therefore, the map $\operatorname{id}_{\mathcal C}\otimes\, \partial_{\overline t}\colon \mathcal C_{O(K)} \to \mathcal C_{O(K)}$ such that $c\mapsto \sum_n c_n\otimes n\overline t^{\,n}$ is well-defined. If $\mathcal C\subset \mathcal C_1$ is an embedding of $\mathbb{Z}_p$-modules then we have $(\operatorname{id}_{\mathcal C_1}\otimes\, \partial_{\overline t})|_{\mathcal C_{O(K)}}=\operatorname{id}_{\mathcal C}\otimes\, \partial_{\overline t}$.

With the above notation:

1) for all $m\in M(H)$, $\nabla (m)=(\widetilde{B} +\operatorname{id}_{\mathcal F(H)}\otimes\, \partial_{\overline t}) (m)\,d\overline t/\overline t$;

2) for all $u\in H_{O(K)}$ and $X\in \operatorname{End}H_{O(K)}$,

$$ \begin{equation*} (\operatorname{id}_H\otimes\, \partial_{\overline t})(X (u))= (\operatorname{id}_{\operatorname{End} H}\otimes\, \partial_{\overline t})(X)(u) + X \bigl((\operatorname{id}_{H}\otimes\, \partial_{\overline t})u\bigr). \end{equation*} \notag $$

Proposition 2.2. Let $C=-(\operatorname{id}_{\operatorname{End}H}\otimes\, \partial_{\overline t})A A^{-1}$ and let for any $n\geqslant 1$, $D^{(n)}=A \cdot \sigma_{\operatorname{End}H}(A) \cdots \sigma_{\operatorname{End}H}^{n-1}(A)$. Then

$$ \begin{equation*} B=\sum_{n\geqslant 0}p^n \operatorname{Ad}(D^{(n)})\sigma_{\operatorname{End}H}^n(C). \end{equation*} \notag $$

Remark 2.2. In § 4 we will prove that $C,B\in L(H)_{O(K)}\subset\operatorname{End}H_{O(K)}$. In particular, $\sigma_{\operatorname{End}H}C=\sigma_{L(H)}C$ and the correspondence $u\mapsto (B+\operatorname{id}_{L(H)}\otimes\, \partial_{\overline t})(u)\,d\overline t/\overline t$ gives a connection on $L(H)_{O(K)}$.

Proof of Proposition 2.2. Indeed, for any $u\in H_{O(K)}$, it holds
$$ \begin{equation*} \begin{aligned} \, (\nabla \cdot \phi)(\mathcal F(u)) &=(\nabla \cdot \mathcal F\cdot \mathcal A)u=\bigl((\widetilde B+\operatorname{id}_{\mathcal F(H)}\otimes\, \partial_{\overline t}) \cdot \mathcal F \cdot \mathcal A\bigr)(u)\, \frac{d\overline t}{\overline t} \\ &=\bigl(\mathcal F\cdot (B+\operatorname{id}_H\otimes\, \partial_{\overline t})\cdot \mathcal A\bigr)(u)\, \frac{d\overline t}{\overline t}. \end{aligned} \end{equation*} \notag $$

On the other hand,

$$ \begin{equation*} \begin{aligned} \, (\phi\otimes\phi)(\nabla (\mathcal F(u))) &= \phi \bigl(\mathcal F(B+\operatorname{id}_H\otimes\, \partial_{\overline t})u\bigr)\phi \biggl(\frac{d\overline t}{\overline t}\biggr) \\ &= p\bigl(\mathcal F\cdot \mathcal A \cdot (B+\operatorname{id}_H\otimes\, \partial_{\overline t})\bigr)(u)\, \frac{d\overline t}{\overline t}. \end{aligned} \end{equation*} \notag $$

Equivalently, we have the following identity on $H_{O(K)}$,

$$ \begin{equation*} (B\cdot A+(\operatorname{id}_H\otimes\, \partial_{\overline t}) \cdot A) \cdot \sigma_H= pA \cdot \sigma_H \cdot (B+\operatorname{id}_H \otimes\, \partial_{\overline t}). \end{equation*} \notag $$
Rewrite this equality as follows
$$ \begin{equation*} (B \cdot A-pA \cdot \sigma_{\operatorname{End}H}(B)) \cdot \sigma_H= \bigl(-(\operatorname{id}_H\otimes\, \partial_{\overline t}) \cdot A+ A\cdot \sigma_H\cdot (\operatorname{id}_H\otimes\, p\,\partial_{\overline t}) \cdot \sigma_H^{-1}\bigr) \cdot \sigma_H. \end{equation*} \notag $$

Notice that on $H_{O(K)}$ we have $\sigma_H \cdot (\operatorname{id}_H\otimes\, p\,\partial_{\overline t}) \cdot \sigma_H^{-1}= \operatorname{id}_H\otimes\, \partial_{\overline t}$. As a result, the right-hand side equals $-(\operatorname{id}_{\operatorname{End}H}\otimes\, \partial_{\overline t})(A)\cdot \sigma_H$.

From $\sigma_H|_H=\operatorname{id}_H$ it follows that by restriction on $H$ we have

$$ \begin{equation*} B \cdot A-pA \cdot \sigma_{\operatorname{End}H}(B) =-(\operatorname{id}_{\operatorname{End}H}\otimes\, \partial_{\overline t})A. \end{equation*} \notag $$

By $O(K)$-linearity this identity holds on the whole $H_{O(K)}$.

As a result, our identity appears in the form

$$ \begin{equation*} \bigl(\operatorname{id}_H-p\operatorname{Ad}(A) \cdot \sigma_{\operatorname{End}H} \bigr)B= -(\operatorname{id}_H\otimes\, \partial_{\overline t})(A) \cdot A^{-1}. \end{equation*} \notag $$
It remains to recover $B$ using that
$$ \begin{equation*} (\operatorname{id}_{H}-p\operatorname{Ad}A \cdot \sigma_{\operatorname{End}H})^{-1}= \sum_{n\geqslant 0} p^n\operatorname{Ad} (D^{(n)}) \cdot \sigma_{\operatorname{End}H}^n. \end{equation*} \notag $$

§ 3. Ramification filtration modulo $p$th commutators

Recall that $K=k((t))\subset K_0^{\mathrm{tr}}$, $\pi_{H}(\Gamma_{K})=\exp (L(H))=I(H)\subset\operatorname{Aut}_{\mathbb{Z}_p}(H)$. Note also that $O(K)=W(k)((\overline t))$, where ${\overline t}^{\,e_0}=\overline t_0$ and $e_0$ is the ramification index of $K/K_0$.

3.1. Lie algebra $\mathcal L$ and identification $\eta_{<p}$

Let $ K_{<p}$ be the maximal $p$-extension of $ K$ in $ K_0^{\mathrm{sep}}$ with the Galois group of nilpotent class $<p$. Then $\mathcal G_{<p}:=\operatorname{Gal}(K_{<p}/ K)= \varprojlim_M \Gamma_{K}/\Gamma_{K}^{p^M}C_p(\Gamma_{K})$.

Let $\widetilde{\mathcal L}_{W(k)}$ be a profinite free Lie $W(k)$-algebra with the set of topological generators $\{D_{0}\}\cup\{D_{an}\mid a\in\mathbb{Z}^+(p), \, n\in\mathbb{Z}/N_0\}$. Let $\mathcal L_{W(k)}=\widetilde{\mathcal L}_{W(k)}/C_p(\widetilde{\mathcal L}_{W(k)})$, where $C_p(\widetilde{\mathcal L}_{W(k)})$ is the ideal of $p$th commutators. Define the $\sigma $-linear action on $ \mathcal L_{W(k)}$ via $D_{an}\mapsto D_{a,n+1}$ and $D_0\mapsto D_0$, denote this action by the same symbol $\sigma $, and set $ \mathcal L= \mathcal L_{W(k)}|_{\sigma =\operatorname{id}}$.

Fix $\alpha_0\in W(k)$ such that the trace of $\alpha_0$ in the field extension $W(k)[1/p]/\mathbb{Q}_p$ equals $1$. For any $n\in\mathbb{Z} /N_0$, set $D_{0n}=(\sigma^n\alpha_0)D_0$.

We are going to apply the profinite version of the covariant nilpotent Artin–Schreier theory to the Lie algebra $\mathcal L$ and $e_{<p}=\sum_{a\in\mathbb{Z}^0(p)} \overline t^{\,-a}D_{a0}\in \mathcal L\mathbin{\widehat\otimes} O(K)$. In other words, if we fix

$$ \begin{equation*} f_{<p}\in \{f\in \mathcal L\mathbin{\widehat\otimes} O_{\mathrm{sep}}\mid \sigma_{\mathcal L}(f)= e_{<p} \circ f\}\ne\varnothing, \end{equation*} \notag $$
then the map $\eta_{<p}:=\pi_{f_{<p}}(e_{<p})$ given by $\tau\mapsto (-f_{<p}) \circ (\operatorname{id}_{\mathcal L}\otimes\,\tau)f_{<p}$ induces the group isomorphism $\overline\eta_{<p}\colon\Gamma_{<p}\simeq G(\mathcal L)$.

The following property is an easy consequence of the above construction.

Proposition 3.1. Suppose $e\in L(H)_{O(K)}$ is special and given with notation from the definition from § 2.2. Then the map $\log\pi_{H}\colon \mathcal G_{<p}\to G(L(H))$ is given via the correspondences $D_{a0}\mapsto l_{a0}$ (and $D_{an}\mapsto \sigma_{L(H)}^n(l_{a0})$) for all $a\in\mathbb{Z}^0(p)$.

3.2. The ramification ideals $\mathcal L^{(v)}$

For $v\geqslant 0$, denote by $\mathcal G^{(v)}_{<p}$ the image of $\Gamma_{K}^{(v)}$ in $\mathcal G_{<p}$. Then $\overline{\eta}_{<p}(\mathcal G^{(v)}_{<p})=G(\mathcal L^{(v)})$, where $\mathcal L^{(v)}$ is an ideal in $\mathcal L$. The images $\mathcal L^{(v)}(M)$ of the ideals $\mathcal L^{(v)}$ in the quotients $\mathcal L/p^M\mathcal L$ for all $M\in\mathbb{N}$ were explicitly described in [10]. By going to the projective limit on $M$ this description can be presented as follows.

Definition 3.1. Let $\overline n=(n_1,\dots,n_s)$ with $s\geqslant 1$. Suppose there is a partition $0=i_0<i_1<\dots <i_r=s$ such that for $i_j<u\leqslant i_{j+1}$, it holds $n_u=m_{j+1}$ and $m_1>m_2>\dots >m_r$. Then set

$$ \begin{equation*} \eta (\overline n)=\frac{1}{(i_1-i_0)!\cdots (i_r-i_{r-1})!}. \end{equation*} \notag $$
If such a partition does not exist we set $\eta (\overline n)=0$.

For $s\in\mathbb{N} $, $\overline a=(a_1,\dots,a_s)\in\mathbb{Z}^0(p)^s$ and $\overline n=(n_1,\dots,n_s)\in\mathbb{Z}^s$, set

$$ \begin{equation*} [D_{\overline a,\overline n}]=[\dots, [D_{a_1n_1},D_{a_2n_2}],\dots,D_{a_sn_s}]. \end{equation*} \notag $$

For $\alpha\geqslant 0$ and $N\in\mathbb{Z}_{\geqslant 0} $, introduce $\mathcal F^0_{\alpha,-N}\in{L}_{W(k)}$ such that

$$ \begin{equation*} \mathcal F^0_{\alpha,-N}=\sum_{\substack {1\leqslant s<p\\ \gamma (\overline a, \overline n)=\alpha}} a_1\eta (\overline n)p^{n_1}[D_{\overline a,\overline n}]. \end{equation*} \notag $$

Here $n_1\geqslant 0$, all $n_i\geqslant -N$ and $\gamma (\overline a,\overline n)= a_1p^{n_1}+a_2p^{n_2}+\dots +a_sp^{n_s}$.

Note that the non-zero terms in the above expression for $\mathcal F^0_{\alpha, -N}$ can appear only if $n_1\geqslant n_2\geqslant\dots \geqslant n_s$ and $\alpha $ has at least one presentation in the form $\gamma (\overline a,\overline n)$.

Denote by $\mathcal I^{(v)}[N]$ the minimal closed ideal in $\mathcal L$ such that its extension of scalars $\mathcal I^{(v)}[N]_{W(k)}$ contain all $\mathcal F^0_{\alpha,-N}$ with $\alpha\geqslant v$.

Our result from [10] about explicit generators of the ideal ${\mathcal L}^{(v)}$ can be stated in the following form.

Theorem 3.1. For any $v>0$ and $M\in\mathbb{N} $, there is $\widetilde{N}(v,M)\in\mathbb{N} $ such that if $N\geqslant \widetilde{N}(v,M)$, then the images of the ideals $\mathcal L^{(v)}$ and $\mathcal I^{(v)}[N]$ in $\mathcal L/p^M$ coincide.

3.3. Some relations

Let $A(\mathcal L)$ be the enveloping algebra of $ \mathcal L$ and $\widetilde{A}(\mathcal L)= A(\mathcal L)/J(\mathcal L)^p$, where $J(\mathcal L)$ is the augmentation ideal in $A(\mathcal L)$. Note that there is a natural embedding of $\mathbb{Z}_p$-modules $\mathcal L\subset \widetilde{A}(\mathcal L)$.

Let $A_{<p}=\exp (-e_{<p})\in \widetilde{A}(\mathcal L)_{O(K)}$ and $C_{<p}=- (\operatorname{id}_{\widetilde{A}(\mathcal L)}\otimes\, \partial_{\overline t})A_{<p} \cdot A_{<p}^{-1}$.

For $s\geqslant 1$, set $\overline 0_s={\displaystyle(\,\underbrace{0,\dots,0}_{s\text{ times}}\,)}$.

Proposition 3.2. Let $D^{(m)}_{<p}:=A_{<p} \cdot \sigma_{\widetilde{A}(\mathcal L)}(A_{<p}) \cdots \sigma_{\widetilde{A}(\mathcal L)}^{m-1}(A_{<p})$, where $m\,{\geqslant}\, 1$. Then we have the following relations:

$$ \begin{equation} C_{<p}= \sum_{s\geqslant 1,\overline a} a_1\eta (\overline 0_s) [D_{\overline a, \overline 0_s}] \overline t^{\,-\gamma (\overline a, \overline 0_s)}, \end{equation} \tag{3.1} $$
$$ \begin{equation} B_{<p} :=\sum_{n\geqslant 0}p^n \operatorname{Ad}(D^{(n)}_{<p})(\sigma_{\mathcal L}^n(C_{<p}))= \sum_{\alpha >0}\mathcal F^0_{\alpha,0}\overline t^{\,-\alpha}, \end{equation} \tag{3.2} $$
$$ \begin{equation} \operatorname{Ad}\sigma_{\widetilde{A}(\mathcal L)}^{-m}(D^{(m)}_{<p})(B_{<p})= \sum_{\alpha >0}\mathcal F^0_{\alpha,-m}\overline t^{\,-\alpha}. \end{equation} \tag{3.3} $$

Proof. For (3.1) use (cf. [20], Theorem 4.22), to obtain
$$ \begin{equation*} \smash{d\exp (-e_{<p}) \cdot \exp (e_{<p})= \sum_{k\geqslant 1}\frac{1}{k!} (-\operatorname{ad} e_{<p})^{k-1}(-de_{<p})} \end{equation*} \notag $$
and note that
$$ \begin{equation*} \begin{aligned} \, (-\operatorname{ad}e_{<p})^{k-1}(de_{<p}) &= (-1)^{k-1}\bigl[\underbrace{e_{<p},\dots, [e_{<p}}_{k-1 \text{ times}},de_{<p}],\dots\bigr] \\ &=\bigl[\dots, [de_{<p},\underbrace{e_{<p}],\dots, e_{<p}}_{k-1\text{ times}}\bigr]. \end{aligned} \end{equation*} \notag $$

For (3.2) we need the following relation (cf. [20], § 4.4)

$$ \begin{equation*} \exp (X) \cdot Y \cdot \exp (-X)= \sum_{n\geqslant 0}\frac{1}{n!}\operatorname{ad}^n(X)(Y). \end{equation*} \notag $$

After applying this relation to the summand with $m=1$ we obtain

$$ \begin{equation*} \begin{aligned} \, &p\operatorname{Ad}D_{<p}^{(1)}(\sigma_{\mathcal L}C_{<p}) = \exp (-e_{<p})\cdot \sigma_{\mathcal L}(C_{<p}) \cdot \exp (e_{<p}) \\ &\qquad=\sum_{s\geqslant 0}\eta (\overline 0_s)(-1)^s\operatorname{ad}^s(e_{<p})(C_{<p}) = \sum_{\overline n\geqslant 0, \overline a}\,\sum_{n_1=0}^{n_1=1} \eta (\overline n)p^{n_1}[D_{\overline a,\overline n}] \overline t^{\,-\gamma (\overline a, \overline n)}. \end{aligned} \end{equation*} \notag $$
Repeating this procedure we obtain relation (3.2).

Similar calculations prove the remaining item (3.3). Proposition 3.2 is proved.

§ 4. Proof of Theorem 1.1

Recall briefly what we’ve already achieved.

The field $K=k((t))$ is tamely ramified extension of $K_0=k_0((t_0))$, where $t^{e_0}\,{=}\, t_0$, $H:=H_0|_{\Gamma_{K}}$ and the corresponding group epimorphism $\pi_H\colon \Gamma_K\to I(H)\subset \operatorname{Aut}_{\mathbb{Z}_p}H$ is such that $I(H)=\exp (L(H))$, where $L(H)\subset\operatorname{End}_{\mathbb{Z}_p}H$ and $L(H)^p=0$.

Applying formalism of nilpotent Artin–Schreier theory we obtained a special $e=\sum_{a\in\mathbb{Z}^0(p)}\overline t^{\,-a}l_{a0}\in L(H)_{O(K)}$ and $f\in L(H)_{O_{\mathrm{sep}}}$ such that

$$ \begin{equation*} \exp (\sigma_{L(H)}f)=\exp (e) \cdot \exp (f) \end{equation*} \notag $$
and for any $\tau \in \Gamma_{K}$, $\pi_H(\tau) =\exp (-f) \cdot \exp(\operatorname{id}_{L(H)}\otimes \tau)f$.

We used $O(K)$-linear operator $\mathcal F\,{=}\exp (-f)$ to introduce $O(K)$-module $M(H):=\mathcal F(H_{O(K)})$. Let $A=\exp (-e)$ and let ${\mathcal A}$ be a unique $\sigma $-linear operator on $H_{O(K)}$ such that for any $h\in H$, $\mathcal A(h)=A(h)$. Then the $\sigma $-linear $\phi \colon M(H)\to M(H)$ is such that for any $u\in L(H)_{O(K)}$, $\phi (\mathcal F(u))=\mathcal F(\mathcal A(u))$. As a result, we obtain the structure of etale $(\phi, O(K))$-module on $M(H)$ related to the $\Gamma_K$-module $H$.

Let $\nabla $ be the connection on $M(H)$ from § 2.4, and let $\widetilde{B}$ be the $O(K)$-linear operator on $M(H)$ uniquely determined by the condition: for any $m\in\mathcal F(H)$, $\nabla (m)=\widetilde{B}(m)d\overline t/\overline t$. Then for any $u\in M(H)$,

$$ \begin{equation*} \nabla (u)=(\widetilde{B}+\operatorname{id}_{\mathcal F(H)}\otimes\, \partial_{\overline t})(u)\, \frac{d\overline t}{\overline t} \end{equation*} \notag $$
and we introduce the differential forms
$$ \begin{equation*} \widetilde{\Omega}[N]= \phi^N\widetilde{B}\phi^{-N}\, \frac{d\overline t}{\overline t} \in \operatorname{End} M(H)_{O(K^{\mathrm{rad}})}\otimes\Omega^1_{O(K)}. \end{equation*} \notag $$

Finally, define the $O(K)$-linear operator $B$ on $H$ by setting for any $u\in H_{O(K)}$, $\mathcal F(B(u))=\widetilde{B}(\mathcal F(u))$, and transfer $\widetilde{\Omega}[N]$ to $\operatorname{End}H_{O(K^{\mathrm{rad}})}$ in the following form

$$ \begin{equation*} \Omega [N]= \operatorname{Ad}({\mathcal A}^N)(B)\, \frac{d\overline t}{\overline t} \in \operatorname{End}(H)_{O(K^{\mathrm{rad}})}\otimes _{O(K)}\Omega^1_{O(K)}. \end{equation*} \notag $$

Remark 4.1. Obviously we have the following identification:

$$ \begin{equation*} \operatorname{End}(H)_{O(K^{\mathrm{rad}})}\otimes_{O(K)}\Omega^1_{O(K)}= \operatorname{End}(H_0)_{O(K_0^{\mathrm{rad}})}\otimes_{O(K_0)}\Omega^1_{O(K_0)}. \end{equation*} \notag $$

Recall that $\mathcal A=A\cdot \sigma_H$, where $A=\exp(-e)$.

Lemma 4.1. If $\mathcal D^{(N)}=\sigma_{\operatorname{End}H}^{-N}(A)\cdots \sigma_{\operatorname{End}H}^{-1}(A)$, then

$$ \begin{equation*} \bigl(\sigma_{\operatorname{End}H}^{-N} \cdot \operatorname{Ad}(\mathcal A^N)\bigr)(B)= \operatorname{Ad}\mathcal D^{(N)}(B). \end{equation*} \notag $$

Proof. Use induction on $N\geqslant 0$. If $N=0$ there is nothing to prove. Suppose lemma is proved for $N\geqslant 0$. Then
$$ \begin{equation*} \begin{aligned} \, &\bigl(\sigma_{\operatorname{End}H}^{-(N+1)}\cdot \operatorname{Ad}(\mathcal A^{N+1})\bigr)(B)= \sigma_{\operatorname{End}H}^{-(N+1)}\bigl(\mathcal A\cdot \operatorname{Ad}(\mathcal A^N) (B) \cdot \mathcal A^{-1}\bigr) \\ &\qquad =\sigma_{\operatorname{End}H}^{-(N+1)}\bigl(\mathcal A\cdot (\sigma_{\operatorname{End}H}^N \cdot \operatorname{Ad}(\mathcal D^{(N)})(B))\cdot \mathcal A^{-1}\bigr) \\ &\qquad=\sigma_{\operatorname{End}H}^{-(N+1)}\bigl(A \cdot \sigma_H \cdot \bigl(\sigma_{\operatorname{End}H}^{N}\operatorname{Ad}(\mathcal D^{(N)})(B)\bigr)\cdot \sigma_H^{-1} \cdot A^{-1}\bigr) \\ &\qquad=\sigma_{\operatorname{End}H}^{-(N+1)}(A \cdot \bigl(\sigma_{\operatorname{End}H}^{N+1}\operatorname{Ad}(\mathcal D^{(N)})(B)\bigr)\cdot A^{-1}) \\ &\qquad=\sigma_{\operatorname{End}H}^{-(N+1)}(A) \cdot \operatorname{Ad}(\mathcal D^{(N)})(B) \cdot \sigma_{\operatorname{End}H}^{-(N+1)}(A^{-1})= \operatorname{Ad}(\mathcal D^{(N+1)})(B). \end{aligned} \end{equation*} \notag $$
The lemma is proved.

Under the projection $\log\overline{\pi}_H\colon \mathcal G_{<p}\to G(L(H))$ we have:

$$ \begin{equation*} \begin{gathered} \, D_{an}\mapsto l_{an}=\sigma_{L(H)}^nl_{a0},\quad e_{<p}\mapsto e,\quad f_{<p}\mapsto f,\quad A_{<p}\mapsto A,\quad C_{<p}\mapsto C, \\ B_{<p}\mapsto B,\quad\text{and}\quad \sigma^{-m}_{\widetilde{\mathcal A}(\mathcal L)}D^{(m)}_{<p}\mapsto \mathcal D^{(m)}. \end{gathered} \end{equation*} \notag $$

Remark 4.2. Because $\log \overline\pi_H(\mathcal L_{<p})=L(H)$ we obtain the proof of the property stated in Remark 2.2.

As a result, our differential form appears as the image of $\sum \mathcal F^0_{\alpha,-N}\overline t^{\,-\alpha}\, d\overline t/\overline t$.

It remains to notice that when getting back to the field $K_0$, we have $d\overline t/\overline t=e_0^{-1}\, d\overline t_0/\overline t_0$, $\overline t^{\,-\alpha}=\overline t^{\,-\alpha /e_0}$, $\Gamma_K^{(\alpha)}=\Gamma_{K_0}^{(\alpha /e_0)}$ and $\pi_H|_{\mathcal I}=\pi_{H_0}|_{\mathcal I}$.

Theorem 1.1 is proved.

Remark 4.3. a) The conjugacy class of the differential form $\Omega [N]$ does not depend on a choice of a special form for $e$.

b) It would be very interesting to verify whether our results could be established in the case of $\Gamma_{K}$-modules which do not satisfy the $\operatorname{Lie}$ condition, e. g., for the $\Gamma_{K}$-module from [21] (the case $n=p$ in the notation of that paper).

§ 5. Mixed characteristic

Let $ E_0$ be a complete discrete valuation field of characteristic $0$ with finite residue field $k_0$ of characteristic $p$. Let $\overline E_0$ be an algebraic closure of $E_0$ and for any field $E$ such that $E_0\subset E\subset \overline E_0$, set $\operatorname{Gal}(\overline E_0/E)=\Gamma_E$. Suppose that $E_0$ contains a primitive $p$th root of unity $\zeta_1$.

We are going to develop an analog of the above characteristic $p$ theory in the context of finite $\mathbb{F}_p[\Gamma_{E_0}]$-modules $H_{E_0}$ satisfying an analogue of the Lie condition from § 2.1:

if $\pi_{H_{E_0}}\colon \Gamma_{E_0}\to\operatorname{Aut}_{\mathbb{F}_p}(H_{E_0})$ determines a $\Gamma_{E_0}$-action on $H_{E_0}$ then there is a Lie $\mathbb{F}_p$-subalgebra $L(H_{E_0})\subset \operatorname{End}_{\mathbb{F}_p}(H_{E_0})$ such that $L(H_{E_0})^p=0$ and $\exp (L(H_{E_0}))=\pi_{H_{E_0}}(I)$, where $I$ is the wild ramification subgroup in $\Gamma_{E_0}$.

Remark 5.1. Contrary to the characteristic $p$ case we restrict ourselves to the Galois modules killed by $p$ because the theory from [11], [12] is developed recently only under that assumption.

Fix a choice of a uniformising element $\pi_0$ in $E_0$.

Let $\widetilde{E}_0=E_0(\{\pi^{(n)}_0)\mid n\in\mathbb{Z}_{\geqslant 0}\}) \subset \overline E_0$, where $\pi^{(0)}_0=\pi_0$ and for all $n\in\mathbb{N} $, $\pi^{(n)p}_0=\pi^{(n-1)}_0$. The field-of-norms functor $X$ provides us with:

– a complete discrete valuation field $X(\widetilde{E}_0)=K_0$ of characteristic $p$ with residue field $k_0$ and a fixed uniformizer $t_0=\varprojlim \pi^{(0)}_n$;

– an identification of $\Gamma_{K_0}= \operatorname{Gal} (K_0^{\mathrm{sep}}/K_0)$ with ${\Gamma}_{\widetilde{E}_0} \subset\Gamma_{E_0}$.

Let $E$ be a finite tamely ramified extension of $E_0$ in $\overline E_0$ such that $\pi_{H_{E_0}} (\Gamma_E)=I(H_0)$. By replacing $E$ with a suitable finite unramified extension we can assume that $E$ has uniformiser $\pi $ such that $\pi^{e_0}=\pi_0$. Let $k$ be the residue field of $E$.

It is easy to see that the field $\widetilde{E}:=E\widetilde{E}_0$ appears in the form $E(\{\pi^{(n)}\mid n\geqslant 0\})$, where $\pi^{(0)}=\pi $, $\pi^{(n)p}=\pi^{(n-1)}$ and for all $n$, $\pi^{(n)e_0}=\pi^{(n)}_0$. In particular, $K:=X(\widetilde{E})=k((t))$, where $t=\varprojlim \pi^{(n)}$ is uniformiser such that $t^{e_0}=t_0$.

Let $\mathcal G_{<p}=\Gamma_K/\Gamma_K^pC_p(\Gamma_K)$ and $\Gamma_{<p}=\Gamma_E/\Gamma_E^pC_p(\Gamma_E)$. According to [11], [12] we have the following natural exact sequence:

$$ \begin{equation*} \mathcal G_{<p}\to \Gamma_{<p}\to \langle \tau_0 \rangle^{\mathbb{Z} /p}\to 1, \end{equation*} \notag $$
where $\tau_0\in\operatorname{Gal}(E(\pi^{(1)})/E)$ is such that $\tau_0(\pi^{(1)})=\zeta_1\pi^{(1)}$. We can use the identification $\overline\eta_{<p}\colon \mathcal G_{<p}\simeq G(\mathcal L)$ from § 3.1 obtained via the special element $e_{<p}$ and the corresponding $f_{<p}$ such that $\sigma_{L(H)}(f_{<p})=e_{<p}\circ f_{<p}$.

øverfullrule 0pt

Then we can use the equivalence of categories from § S1.3 to identify $\Gamma_{<p}$ with $G(L)$ where $L$ is a profinite Lie $\mathbb{F}_p$-algebra included into the following exact sequence:

$$ \begin{equation} \mathcal L\to L\to \mathbb{F}_p\tau_0\to 0. \end{equation} \tag{5.1} $$

When studying the structure of (5.1) in [12] we proved that $\tau_0$ can be replaced by a suitable $h_0\in\operatorname{Aut}K$. More precisely, suppose

$$ \begin{equation*} \zeta_1\equiv 1+\sum_{j\geqslant 0}[\beta_j]\pi_0^{(e_0^*/p)+j}\ \operatorname{mod}p \end{equation*} \notag $$
with Teichmüller representatives $[\beta_j]$ of $\beta_j\in k$ and $e^*=ep/(p-1)$, where $e$ is the ramification index for $E/\mathbb{Q}_p$. Then $h_{0}$ can be defined as follows: $h_{0}|_k=\operatorname{id}_k$ and
$$ \begin{equation*} h_{0}(t)=t\biggl(1+\sum_{j\geqslant 0}\beta_j^pt^{e^*+pj}\biggr)= t\,\widetilde{\mathrm{exp}}(\omega (t)^p), \end{equation*} \notag $$
where $\widetilde{\exp}$ is the truncated exponential and $\omega (t)\in t^{e^*/p}k[[t]]^*$.

This allowed us to apply formalism of the nilpotent Artin–Schreier theory to specify “good” lifts $\tau_{<p}$ of $\tau_0$ to $L$ (what is equivalent to specifying “good lifts” $h_{<p}$ of $h_0$).

In particular, we obtained in [11], [12] the following description of the image $\overline L\subset L$ of $\mathcal L$ from exact sequence (5.1). Introduce the weight function $\operatorname{wt}$ on $\mathcal L_{k}$ by setting $\operatorname{wt}(D_{an})=s\in\mathbb{N} $ if and only if $(s-1)e^*\leqslant a<se^*$. Then

$$ \begin{equation*} \operatorname{Ker}(\mathcal L\to L)=\mathcal L(p)=\{l\in \mathcal L\mid \operatorname{wt}(l)\geqslant p\}, \end{equation*} \notag $$
$\overline L=\mathcal L/\mathcal L(p)$ and $\mathcal L\to \overline L$ is the natural projection.

If $h_{<p}$ is a lift of $h_0$ to $K_{<p}$ then it is uniquely determined by $c=c(h_{<p})\in L_{K}$ such that

$$ \begin{equation*} (\operatorname{id}_{\mathcal L_{<p}}\otimes h_{<p})f=c\circ \bigl(\operatorname{Ad}(h_{<p})\otimes\operatorname{id}_{K_{<p}}\bigr)f. \end{equation*} \notag $$
This allowed us to describe the corresponding action of the group $\langle h_{<p}\rangle^{\mathbb{Z} /p}$ on $f$ as an action of an infinitesimal group scheme of order $p$. The differential of this action is given by the “linear part” $c_1\in \overline L_{K}$ of $c$ which could be described by a suitable recurrent procedure. Finally, we proved that $c_1(0)\in \overline L_{k}$ (where $c_1=\sum_{n\in\mathbb{Z}}c_1(n)t^n$ with all $c_1(n)\in\overline L_{k}$) is an absolute invariant of the lift $h_{<p}$.

Consider the expansion $\omega (t)^p=\sum_{j\geqslant 0}A_jt^{e^*+pj}$, $A_j\in k$.

Denote by $\overline L^{(e^*)}\subset \overline L$ the image of the ramification subgroup $\mathcal L^{(e^*)}$ in $L$ (cf. (5.1)). Then Theorem 5.1 from [12] states:

$\bullet$ the lift $\tau_{<p}$ of $\tau_0$ is “good” if and only if the value $c_1(0)\in \mathcal L_{k}$ of the differential $d\tau_{<p}$ at $0$ satisfies the following congruence:

$$ \begin{equation*} c_1(0)\equiv \sum_{j\geqslant 0} \sum_{i\geqslant 0}\sigma^i(A_j \mathcal F^0_{e^*+pj,-i}) \ \operatorname{mod}\overline{L}_k^{(e^*)}. \end{equation*} \notag $$
It remains to note that for $i,j\gg 0$, $\mathcal F^0_{e^*+pj,-i}\in \overline L_k^{(e^*)}$ and the right-hand double sum contains only finitely many nonzero terms modulo $\overline L_{k}^{(e^*)}$. It can be rewritten also in the following form:
$$ \begin{equation*} \sum_{i,j\geqslant 0}\operatorname{Res}\bigl( \sigma^i(A_jt^{e^*+pj} \cdot \sigma^{-i}\Omega_{<p}[i])\bigr), \end{equation*} \notag $$
and the image of this expression in $L(H)_k$ equals
$$ \begin{equation*} \sum_{i\geqslant 0} \operatorname{Res}\bigl(\sigma^{i+1}\omega (t) \cdot \Omega [i]\bigr). \end{equation*} \notag $$

The author expresses his deep gratitude to Prof. E. Khukhro for helpful discussions.


Bibliography

1. J.-P. Serre, Local fields, Transl. from the French, Grad. Texts in Math., 67, Springer-Verlag, New York–Berlin, 1979  crossref  mathscinet  zmath
2. I. R. Šafarevič, “On $p$-extensions”, Mat. Sb., 20(62):2 (1947), 351–363  mathnet  mathscinet  zmath; English transl. Amer. Math. Soc. Transl. Ser. 2, 4, Amer. Math. Soc., Providence, RI, 1956, 59–72  crossref
3. S. P. Demushkin, “The group of a maximal $p$-extension of a local field”, Izv. Akad. Nauk SSSR Ser. Mat., 25:3 (1961), 329–346  mathnet  mathscinet  zmath
4. U. Jannsen and K. Wingberg, “Die Struktur der absoluten Galoisgruppe $\mathfrak p$-adischer Zahlkörper”, Invent. Math., 70:1 (1982/83), 71–98  crossref  mathscinet  zmath  adsnasa
5. Sh. Mochizuki, “A version of the Grothendieck conjecture for $p$-adic local fields”, Internat. J. Math., 8:4 (1997), 499–506  crossref  mathscinet  zmath
6. V. A. Abrashkin, “On a local analogue of the Grothendieck conjecture”, Internat. J. Math., 11:2 (2000), 133–175  crossref  mathscinet  zmath
7. H. Koch, Galois theory of $p$-extensions, Springer Monogr. Math., Springer-Verlag, Berlin, 2002  crossref  mathscinet  zmath
8. V. A. Abrashkin, “A ramification filtration of the Galois group of a local field”, Tr. St. Peterbg. Mat. Obs., 3, St. Petersburg Univ., St. Petersburg, 1994; English transl. Proceedings of the St. Petersburg Mathematical Society, v. III, Amer. Math. Soc. Transl. Ser. 2, 166, Amer. Math. Soc., Providence, RI, 1995, 35–100  mathscinet  zmath
9. V. A. Abrashkin, “Ramification filtration of the Galois group of a local field. II”, Number theory, algebra and algebraic geometry, Collection of articles. To seventieth birthday of Academician I. R. Shafarevich, Trudy Mat. Inst. Steklov., 208, Nauka, Fizmatlit, Moscow, 1995, 18–69  mathnet  mathscinet  zmath; English transl. Proc. Steklov Inst. Math., 208 (1995), 15–62
10. V. A. Abrashkin, “A ramification filtration of the Galois group of a local field. III”, Izv. Ross. Akad. Nauk Ser. Mat., 62:5 (1998), 3–48  mathnet  crossref  mathscinet  zmath; English transl. Izv. Math., 62:5 (1998), 857–900  crossref  adsnasa
11. V. Abrashkin, “Groups of automorphisms of local fields of period $p$ and nilpotent class $<p$. I”, Internat. J. Math., 28:6 (2017), 1750043  crossref  mathscinet  zmath
12. V. Abrashkin, “Groups of automorphisms of local fields of period $p$ and nilpotent class $<p$. II”, Internat. J. Math., 28:10 (2017), 1750066  crossref  mathscinet  zmath
13. V. Abrashkin, “Groups of automorphisms of local fields of period $p^M$ and nilpotent class $<p$”, Ann. Inst. Fourier (Grenoble), 67:2 (2017), 605–635  mathnet  crossref  mathscinet  zmath
14. P. Berthelot and W. Messing, “Théorie de Deudonné cristalline. III. Théorèmes d'équivalence et de pleine fidélité”, The Grothendieck festschrift, v. 1, Progr. Math., 86, Birkhäuser Boston, Boston, MA, 1990, 173–247  mathscinet  zmath
15. M. Lazard, “Sur les groupes nilpotents et les anneaux de Lie”, Ann. Sci. Ecole Norm. Sup. (3), 71:2 (1954), 101–190  crossref  mathscinet  zmath
16. J.-M. Fontaine, “Représentations $p$-adiques des corps locaux. I”, The Grothendieck festschrift, v. 2, Progr. Math., 87, Birkhäuser Boston, Boston, MA, 1990, 249–309  mathscinet  zmath
17. V. Abrashkin and R. Jenni, “The field-of-norms functor and the Hilbert symbol for higher local fields”, J. Théor. Nombres Bordeaux, 24:1 (2012), 1–39  crossref  mathscinet  zmath
18. V. Abrashkin, “Galois groups of local fields, Lie algebras and ramification”, Arithmetic and geometry, London Math. Soc. Lecture Note Ser., 420, Cambridge Univ. Press, Cambridge, 2015, 1–23  crossref  mathscinet  zmath
19. V. A. Abrashkin, “Ramification filtration via deformations”, Mat. Sb., 212:2 (2021), 3–37  mathnet  crossref  mathscinet  zmath; English transl. Sb. Math., 212:2 (2021), 135–169  crossref  adsnasa
20. A. Bonfiglioli and R. Fulci, Topics in noncommutative algebra, Lecture Notes in Math., 2034, Springer, Heidelberg, 2012  crossref  mathscinet  zmath
21. K. Imai, Ramification groups of some finite Galois extensions of maximal nilpotency class over local fields of positive characteristic, arXiv: 2102.07928

Citation: V. A. Abrashkin, “Ramification filtration and differential forms”, Izv. RAN. Ser. Mat., 87:3 (2023), 5–22; Izv. Math., 87:3 (2023), 421–438
Citation in format AMSBIB
\Bibitem{Abr23}
\by V.~A.~Abrashkin
\paper Ramification filtration and differential forms
\jour Izv. RAN. Ser. Mat.
\yr 2023
\vol 87
\issue 3
\pages 5--22
\mathnet{http://mi.mathnet.ru/im9322}
\crossref{https://doi.org/10.4213/im9322}
\mathscinet{http://mathscinet.ams.org/mathscinet-getitem?mr=4223960}
\adsnasa{https://adsabs.harvard.edu/cgi-bin/bib_query?2023IzMat..87..421A}
\transl
\jour Izv. Math.
\yr 2023
\vol 87
\issue 3
\pages 421--438
\crossref{https://doi.org/10.4213/im9322e}
\isi{https://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=Publons&SrcAuth=Publons_CEL&DestLinkType=FullRecord&DestApp=WOS_CPL&KeyUT=001063937600001}
\scopus{https://www.scopus.com/record/display.url?origin=inward&eid=2-s2.0-85171875943}
Linking options:
  • https://www.mathnet.ru/eng/im9322
  • https://doi.org/10.4213/im9322e
  • https://www.mathnet.ru/eng/im/v87/i3/p5
  • Citing articles in Google Scholar: Russian citations, English citations
    Related articles in Google Scholar: Russian articles, English articles
    Известия Российской академии наук. Серия математическая Izvestiya: Mathematics
    Statistics & downloads:
    Abstract page:306
    Russian version PDF:12
    English version PDF:50
    Russian version HTML:102
    English version HTML:112
    References:26
    First page:9
     
      Contact us:
     Terms of Use  Registration to the website  Logotypes © Steklov Mathematical Institute RAS, 2024