This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

\urladdr

perso.ens-lyon.fr/laurent.berger/

The perfectoid commutant
of Lubin-Tate power series

Laurent Berger UMPA de l’ENS de Lyon
UMR 5669 du CNRS
laurent.berger@ens-lyon.fr
Abstract

Let LT\mathrm{LT} be a Lubin-Tate formal group attached to a finite extension of 𝐐p\mathbf{Q}_{p}. By a theorem of Lubin-Sarkis, an invertible characteristic pp power series that commutes with the elements Aut⁑(LT)\operatorname{Aut}(\mathrm{LT}) is itself in Aut⁑(LT)\operatorname{Aut}(\mathrm{LT}). We extend this result to perfectoid power series, by lifting such a power series to characteristic zero and using the theory of locally analytic vectors in certain rings of pp-adic periods. This allows us to recover the field of norms of the Lubin-Tate extension from its completed perfection.

Key words and phrases:
Lubin-Tate group; field of norms; pp-adic period; locally analytic vector; pp-adic dynamical system; perfectoid field
1991 Mathematics Subject Classification:
11S; 12J; 13J

Introduction

Let FF be a finite extension of 𝐐p\mathbf{Q}_{p}, with ring of integers π’ͺF\mathcal{O}_{F} and residue field kk. Let q=Card⁑(k)q=\operatorname{Card}(k) and let Ο€\pi be a uniformizer of π’ͺF\mathcal{O}_{F}. Let LT\mathrm{LT} be the Lubin-Tate formal π’ͺF\mathcal{O}_{F}-module attached to Ο€\pi. Let F∞=F​(LT​[Ο€βˆž])F_{\infty}=F(\mathrm{LT}[\pi^{\infty}]) denote the extension of FF generated by the torsion points of LT\mathrm{LT}, and let Ξ“F=Gal⁑(F∞/F)\Gamma_{F}=\operatorname{Gal}(F_{\infty}/F). The Lubin-Tate character χπ\chi_{\pi} gives rise to an isomorphism χπ:Ξ“Fβ†’π’ͺFΓ—\chi_{\pi}:\Gamma_{F}\to\mathcal{O}_{F}^{\times}.

The field of norms ([Win83]) 𝐄F\mathbf{E}_{F} of the extension F∞/FF_{\infty}/F is a local field of characteristic pp, endowed with an action of Ξ“F\Gamma_{F}, that can be explicitly described as follows. We choose a coordinate TT on LT\mathrm{LT}, so that for each a∈π’ͺFa\in\mathcal{O}_{F} we get a power series [a]​(T)∈π’ͺF​[[T]][a](T)\in\mathcal{O}_{F}[\![T]\!]. We then have 𝐄F=k​((Y))\mathbf{E}_{F}=k(\!(Y)\!), on which Ξ“F\Gamma_{F} acts via the formula γ​(f​(Y))=f​([χπ​(Ξ³)]​(Y))\gamma(f(Y))=f([\chi_{\pi}(\gamma)](Y)). In pp-adic Hodge theory, we consider the field 𝐄~F\widetilde{\mathbf{E}}_{F}, which is the YY-adic completion of the maximal purely inseparable extension βˆͺnβ©Ύ0𝐄Fqβˆ’n\cup_{n\geqslant 0}\mathbf{E}_{F}^{q^{-n}} of 𝐄F\mathbf{E}_{F} inside an algebraic closure. The action of Ξ“F\Gamma_{F} extends to the field 𝐄~F\widetilde{\mathbf{E}}_{F}. If fβˆˆπ„~Ff\in\widetilde{\mathbf{E}}_{F} and Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}, we still have γ​(f​(Y))=f​([χπ​(Ξ³)]​(Y))\gamma(f(Y))=f([\chi_{\pi}(\gamma)](Y)). The question that motivated this paper is the following.

Question.

Can we recover 𝐄F\mathbf{E}_{F} from the data of the valued field 𝐄~F\widetilde{\mathbf{E}}_{F} endowed with the action of Ξ“F\Gamma_{F}?

If a∈π’ͺFΓ—a\in\mathcal{O}_{F}^{\times}, then u​(Y)=[a]​(Y)u(Y)=[a](Y) is an element of 𝐄F\mathbf{E}_{F} of valuation 11 that satisfies the functional equation u∘[g]​(Y)=[g]∘u​(Y)u\circ[g](Y)=[g]\circ u(Y) for all g∈π’ͺFΓ—g\in\mathcal{O}_{F}^{\times}. Conversely, we prove the following theorem, which answers the question, as it allows us to find a uniformizer of 𝐄F\mathbf{E}_{F} from the data of the valued field 𝐄~F\widetilde{\mathbf{E}}_{F} endowed with the action of Ξ“F\Gamma_{F}.

Theorem A.

If uβˆˆπ„~Fu\in\widetilde{\mathbf{E}}_{F} is such that valY⁑(u)=1\operatorname{val}_{Y}(u)=1 and u∘[g]=[g]∘uu\circ[g]=[g]\circ u for all g∈π’ͺFΓ—g\in\mathcal{O}_{F}^{\times}, then there exists a∈π’ͺFΓ—a\in\mathcal{O}_{F}^{\times} such that u​(Y)=[a]​(Y)u(Y)=[a](Y).

In particular, 𝐄F=k​((u))\mathbf{E}_{F}=k(\!(u)\!) for any uu as in theorem A. The main difficulty in the proof of theorem A is to prove that if uu is as in the statement of theorem A, then there exists nβ©Ύ0n\geqslant 0 such that uβˆˆπ„Fqβˆ’nu\in\mathbf{E}_{F}^{q^{-n}}. If F=𝐐pF=\mathbf{Q}_{p} and Ο€=p\pi=p, namely in the cyclotomic situation, this follows from the main result of [BR22]. However, a crucial ingredient in that paper does not generalize to F≠𝐐pF\neq\mathbf{Q}_{p}. In order to go beyond the cyclotomic case, we instead use a result of Colmez ([Col02]) to lift uu to an element u^\hat{u} of a ring and F+\and_{F}^{+} (the Witt vectors over the ring of integers of 𝐄~F\widetilde{\mathbf{E}}_{F}, as well as a completion of βˆͺnβ©Ύ0Ο†qβˆ’n​(π’ͺF​[[Y^]])\cup_{n\geqslant 0}\varphi_{q}^{-n}(\mathcal{O}_{F}[\![\widehat{Y}]\!]), where Ο†q​(Y^)=[Ο€]​(Y^)\varphi_{q}(\widehat{Y})=[\pi](\widehat{Y})), that will satisfy a similar functional equation. In particular, u^\hat{u} is a locally analytic element of a suitable ring of pp-adic periods. By previous results of the author ([Ber16]), u^\hat{u} belongs to Ο†qβˆ’n​(π’ͺF​[[Y^]])\varphi_{q}^{-n}(\mathcal{O}_{F}[\![\widehat{Y}]\!]) for a certain nn. This allows us to prove that there exists nβ©Ύ0n\geqslant 0 such that uβˆˆπ„Fqβˆ’nu\in\mathbf{E}_{F}^{q^{-n}}. By replacing uu with upku^{p^{k}} for a well chosen kk, we are led to the study of elements of Yβ‹…k​[[Y]]Y\cdot k[\![Y]\!] under composition. We prove that uu is invertible for composition, and to conclude we use a theorem of Lubin-Sarkis ([LS07]) saying that if an invertible series commutes with a nontorsion element of Aut⁑(LT)\operatorname{Aut}(\mathrm{LT}), then that series is itself in Aut⁑(LT)\operatorname{Aut}(\mathrm{LT}). We finish this paper with an explanation of why the β€œTate traces” on 𝐄~F\widetilde{\mathbf{E}}_{F} used in [BR22] don’t exist if F≠𝐐pF\neq\mathbf{Q}_{p}.

1. Locally analytic vectors

We use the notation that was introduced in the introduction. In order to apply lemma 9.3 of [Col02], we assume that the coordinate TT on LT\mathrm{LT} is chosen such that [Ο€]​(T)[\pi](T) is a monic polynomial of degree qq (for example, we could ask that [Ο€]​(T)=Tq+π​T[\pi](T)=T^{q}+\pi T).

Let F0=𝐐punr∩FF_{0}=\mathbf{Q}_{p}^{\mathrm{unr}}\cap F. Let 𝐄~F+\widetilde{\mathbf{E}}_{F}^{+} denote the ring of integers of 𝐄~F\widetilde{\mathbf{E}}_{F} and let and F+=π’ͺFβŠ—π’ͺF0W​(𝐄~F+)\and^{+}_{F}=\mathcal{O}_{F}\otimes_{\mathcal{O}_{F_{0}}}W(\widetilde{\mathbf{E}}_{F}^{+}) be the π’ͺF\mathcal{O}_{F}-Witt vectors over 𝐄~F+\widetilde{\mathbf{E}}^{+}_{F}.

\propname \the\smf@thm.

If uβˆˆπ„~F+u\in\widetilde{\mathbf{E}}_{F}^{+} is such that γ​(u)=[χπ​(Ξ³)]​(u)\gamma(u)=[\chi_{\pi}(\gamma)](u) for all Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}, then uu has a lift u^∈ and F+\hat{u}\in\and^{+}_{F} such that γ​(u^)=[χπ​(Ξ³)]∘u^\gamma(\hat{u})=[\chi_{\pi}(\gamma)]\circ\hat{u} for all Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}.

Proof.

By lemma 9.3 of [Col02], there is a unique lift u^∈ and F+\hat{u}\in\and^{+}_{F} of uu such that Ο†q​(u^)=[Ο€]​(u^)\varphi_{q}(\hat{u})=[\pi](\hat{u}) (in ibid., this element is denoted by {u}\{u\}). If Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}, then both γ​(u^)\gamma(\hat{u}) and [χπ​(Ξ³)]​(u^)[\chi_{\pi}(\gamma)](\hat{u}) are lifts of uu that are compatible with Frobenius as above. By unicity, they are equal. ∎

Let logLT⁑(T)\log_{\mathrm{LT}}(T) and expLT⁑(T)\exp_{\mathrm{LT}}(T) be the logarithm and exponential series for LT\mathrm{LT}. Write expLT⁑(T)=βˆ‘nβ©Ύ1en​Tn\exp_{\mathrm{LT}}(T)=\sum_{n\geqslant 1}e_{n}T^{n} and expLT(T)j=βˆ‘nβ©Ύjej,nTn\exp_{\mathrm{LT}}(T)^{j}=\sum_{n\geqslant j}e_{j,n}T^{n} for jβ©Ύ1j\geqslant 1.

\lemmname \the\smf@thm.

We have valπ⁑(ej,n)β©Ύβˆ’n/(qβˆ’1)\operatorname{val}_{\pi}(e_{j,n})\geqslant-n/(q-1) for all j,nβ©Ύ1j,n\geqslant 1.

Proof.

Fix Ο–βˆˆπΒ―p\varpi\in\overline{\mathbf{Q}}_{p} such that valπ⁑(Ο–)=1/(qβˆ’1)\operatorname{val}_{\pi}(\varpi)=1/(q-1) and let K=F​(Ο–)K=F(\varpi). Recall that logLT⁑(T)=limnβ†’+∞[Ο€n]​(T)/Ο€n\log_{\mathrm{LT}}(T)=\lim_{n\to+\infty}[\pi^{n}](T)/\pi^{n}. If zβˆˆπ‚pz\in\mathbf{C}_{p} and valπ⁑(z)β©Ύ1/(qβˆ’1)\operatorname{val}_{\pi}(z)\geqslant 1/(q-1), then valπ⁑([Ο€]​(z))β©Ύvalπ⁑(z)+1\operatorname{val}_{\pi}([\pi](z))\geqslant\operatorname{val}_{\pi}(z)+1. This implies that 1/Ο–β‹…logLT⁑(ϖ​T)∈T+T2​π’ͺK​[[T]]1/\varpi\cdot\log_{\mathrm{LT}}(\varpi T)\in T+T^{2}\mathcal{O}_{K}[\![T]\!]. Its composition inverse 1/Ο–β‹…expLT⁑(ϖ​T)1/\varpi\cdot\exp_{\mathrm{LT}}(\varpi T) therefore also belongs to T+T2​π’ͺK​[[T]]T+T^{2}\mathcal{O}_{K}[\![T]\!]. This implies the claim for j=1j=1. The claim for jβ©Ύ1j\geqslant 1 follows easily. ∎

We use a number of rings of pp-adic periods in the Lubin-Tate setting, whose construction and properties were recalled in Β§3 of [Ber16]. Proposition 1 gives us an element Y^∈ and F+\hat{Y}\in\and_{F}^{+} (denoted by uu in ibid.). Let 𝐁~F+= and F+​[1/Ο€]\widetilde{\mathbf{B}}_{F}^{+}=\and_{F}^{+}[1/\pi]. Given an interval I=[r;s]βŠ‚[0;+∞[I=[r;s]\subset[0;+\infty[, a valuation V​(β‹…,I)V(\cdot,I) on 𝐁~F+​[1/Y^]\widetilde{\mathbf{B}}^{+}_{F}[1/\hat{Y}] is constructed in ibid., as well as various completions of that ring. We use 𝐁~FI\widetilde{\mathbf{B}}_{F}^{I}, the completion of 𝐁~F+​[1/Y^]\widetilde{\mathbf{B}}^{+}_{F}[1/\hat{Y}] for V​(β‹…,I)V(\cdot,I) and 𝐁~rig,F†,r=lim←sβ©Ύr⁑𝐁~F[r;s]\widetilde{\mathbf{B}}^{\dagger,r}_{\mathrm{rig},F}=\varprojlim_{s\geqslant r}\widetilde{\mathbf{B}}_{F}^{[r;s]}. Inside 𝐁~rig,F†,r\widetilde{\mathbf{B}}^{\dagger,r}_{\mathrm{rig},F}, there is the ring 𝐁rig,F†,r\mathbf{B}^{\dagger,r}_{\mathrm{rig},F} of power series f​(Y^)f(\hat{Y}) with coefficients in FF, where f​(T)f(T) converges on a certain annulus depending on rr.

\lemmname \the\smf@thm.

If sβ©Ύ0s\geqslant 0, then 𝐁rig,F†,s∩ and F+=𝐀F+\mathbf{B}^{\dagger,s}_{\mathrm{rig},F}\cap\and^{+}_{F}=\mathbf{A}^{+}_{F}.

Proof.

Take f​(Y^)∈𝐁rig,F†,sf(\hat{Y})\in\mathbf{B}^{\dagger,s}_{\mathrm{rig},F}, tβ©Ύst\geqslant s and let I=[s;t]I=[s;t]. We have V​(f,I)β©Ύ0V(f,I)\geqslant 0, so that ff is bounded by 11 on the corresponding annulus. This is true for all tt, so that f∈𝐁F†,sf\in\mathbf{B}^{\dagger,s}_{F}. We now have f∈𝐁F†,s∩ and F+=𝐀F+f\in\mathbf{B}^{\dagger,s}_{F}\cap\and^{+}_{F}=\mathbf{A}_{F}^{+}. ∎

Let WW be a Banach space with a continuous action of Ξ“F\Gamma_{F}. The notion of locally analytic vector was introduced in [ST03]. Recall (see for instance Β§2 of [Ber16]; the definition given there is easily seen to be equivalent to the following one) that an element w∈Ww\in W is locally FF-analytic if there exists a sequence {wk}kβ©Ύ0\{w_{k}\}_{k\geqslant 0} of WW such that wkβ†’0w_{k}\to 0, and an integer nβ©Ύ1n\geqslant 1 such that for all Ξ³βˆˆΞ“F\gamma\in\Gamma_{F} such that χπ​(Ξ³)=1+pn​c​(Ξ³)\chi_{\pi}(\gamma)=1+p^{n}c(\gamma) with c​(Ξ³)∈π’ͺFc(\gamma)\in\mathcal{O}_{F}, we have γ​(w)=βˆ‘kβ©Ύ0c​(Ξ³)k​wk\gamma(w)=\sum_{k\geqslant 0}c(\gamma)^{k}w_{k}. If W=lim←i⁑WiW=\varprojlim_{i}W_{i} is a FrΓ©chet representation of Ξ“F\Gamma_{F}, we say that w∈Ww\in W is pro-FF-analytic if its image in WiW_{i} is locally FF-analytic for all ii.

\propname \the\smf@thm.

If rβ©Ύ0r\geqslant 0 and x∈ and F+x\in\and^{+}_{F} is such that valY⁑(xΒ―)>0\operatorname{val}_{Y}(\overline{x})>0 and γ​(x)=[χπ​(Ξ³)]​(x)\gamma(x)=[\chi_{\pi}(\gamma)](x) for all Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}, then xx is a pro-FF-analytic element of 𝐁~rig,F†,r\widetilde{\mathbf{B}}^{\dagger,r}_{\mathrm{rig},F}.

Proof.

We prove that for all sβ©Ύrs\geqslant r, xx is a locally FF-analytic vector of 𝐁~F[r;s]\widetilde{\mathbf{B}}_{F}^{[r;s]}. The proposition then follows, since 𝐁~rig,F†,r=lim←sβ©Ύr⁑𝐁~F[r;s]\widetilde{\mathbf{B}}^{\dagger,r}_{\mathrm{rig},F}=\varprojlim_{s\geqslant r}\widetilde{\mathbf{B}}_{F}^{[r;s]} as FrΓ©chet spaces.

Let S​(X,Y)=βˆ‘i,jsi,j​Xi​Yj∈π’ͺF​[[X,Y]]S(X,Y)=\sum_{i,j}s_{i,j}X^{i}Y^{j}\in\mathcal{O}_{F}[\![X,Y]\!] be the power series that gives the addition in LT\mathrm{LT}. We have logLT⁑(x)∈𝐁~F[r;s]\log_{\mathrm{LT}}(x)\in\widetilde{\mathbf{B}}_{F}^{[r;s]}. Take nβ©Ύ1n\geqslant 1 such that V​(pnβˆ’1​logLT⁑(x),[r;s])>0V(p^{n-1}\log_{\mathrm{LT}}(x),[r;s])>0. We have [a]​(T)=expLT⁑(a​logLT⁑(T))[a](T)=\exp_{\mathrm{LT}}(a\log_{\mathrm{LT}}(T)), so that [1+pn​c]​(T)=S​(T,expLT⁑(pn​c​logLT⁑(T)))[1+p^{n}c](T)=S(T,\exp_{\mathrm{LT}}(p^{n}c\log_{\mathrm{LT}}(T))). If χπ​(Ξ³)=1+pn​c​(Ξ³)\chi_{\pi}(\gamma)=1+p^{n}c(\gamma), then

γ​(x)\displaystyle\gamma(x) =βˆ‘kβ©Ύ0c(Ξ³)kβˆ‘jβ©½kpn​kej,klogLT(x)kβˆ‘iβ©Ύ0si,jxi\displaystyle=\sum_{k\geqslant 0}c(\gamma)^{k}\sum_{j\leqslant k}p^{nk}e_{j,k}\log_{\mathrm{LT}}(x)^{k}\sum_{i\geqslant 0}s_{i,j}x^{i}
=βˆ‘kβ©Ύ0c​(Ξ³)kβ€‹βˆ‘jβ©½kpk​ej,kβ‹…(pnβˆ’1​logLT⁑(x))kβ‹…βˆ‘iβ©Ύ0si,j​xi.\displaystyle=\sum_{k\geqslant 0}c(\gamma)^{k}\sum_{j\leqslant k}p^{k}e_{j,k}\cdot(p^{n-1}\log_{\mathrm{LT}}(x))^{k}\cdot\sum_{i\geqslant 0}s_{i,j}x^{i}.

We have pk​ej,k∈π’ͺFp^{k}e_{j,k}\in\mathcal{O}_{F} by lemma 1, V​(pnβˆ’1​logLT⁑(x),[r;s])>0V(p^{n-1}\log_{\mathrm{LT}}(x),[r;s])>0 by hypothesis, si,j∈π’ͺFs_{i,j}\in\mathcal{O}_{F} and V​(x,[r;s])>0V(x,[r;s])>0. This implies the claim. ∎

\propname \the\smf@thm.

If r>0r>0 and x∈ and F+x\in\and^{+}_{F} is a pro-FF-analytic element of 𝐁~rig,F†,r\widetilde{\mathbf{B}}^{\dagger,r}_{\mathrm{rig},F}, then there exists nβ©Ύ0n\geqslant 0 such that xβˆˆΟ†qβˆ’n​(𝐀F+)x\in\varphi_{q}^{-n}(\mathbf{A}^{+}_{F}).

Proof.

By item (3) of theorem 4.4 of [Ber16] (applied with K=FK=F), there exists nβ©Ύ0n\geqslant 0 and s>0s>0 such that xβˆˆΟ†qβˆ’n​(𝐁rig,F†,s)x\in\varphi_{q}^{-n}(\mathbf{B}^{\dagger,s}_{\mathrm{rig},F}). The proposition now follows from lemma 1 applied to Ο†qn​(x)\varphi_{q}^{n}(x). ∎

2. Composition of power series

Recall that a power series f​(Y)∈k​[[Y]]f(Y)\in k[\![Y]\!] is separable if f′​(Y)β‰ 0f^{\prime}(Y)\neq 0. If f​(Y)∈Yβ‹…k​[[Y]]f(Y)\in Y\cdot k[\![Y]\!], we say that ff is invertible if f′​(0)∈kΓ—f^{\prime}(0)\in k^{\times}, which is equivalent to ff being invertible for composition (denoted by ∘\circ). We say that w​(Y)∈Yβ‹…k​[[Y]]w(Y)\in Y\cdot k[\![Y]\!] is nontorsion if w∘n​(Y)β‰ Yw^{\circ n}(Y)\neq Y for all nβ©Ύ1n\geqslant 1. If w​(Y)=βˆ‘iβ©Ύ0wi​Yi∈k​[[Y]]w(Y)=\sum_{i\geqslant 0}w_{i}Y^{i}\in k[\![Y]\!] and mβˆˆπ™m\in\mathbf{Z}, let w(m)​(Y)=βˆ‘iβ©Ύ0wipm​Yiw^{(m)}(Y)=\sum_{i\geqslant 0}w_{i}^{p^{m}}Y^{i}. Note that (w∘v)(m)=w(m)∘v(m)(w\circ v)^{(m)}=w^{(m)}\circ v^{(m)}.

\propname \the\smf@thm.

Let w​(Y)∈Y+Y2β‹…k​[[Y]]w(Y)\in Y+Y^{2}\cdot k[\![Y]\!] be an invertible nontorsion series, and let f​(Y)∈Yβ‹…k​[[Y]]f(Y)\in Y\cdot k[\![Y]\!] be a separable power series. If w(m)∘f=f∘ww^{(m)}\circ f=f\circ w, then ff is invertible.

Proof.

This is a slight generalization of lemma 6.2 of [Lub94]. Write

f​(Y)\displaystyle f(Y) =fn​Yn+O⁑(Yn+1)\displaystyle=f_{n}Y^{n}+\operatorname{O}(Y^{n+1})
f′​(Y)\displaystyle f^{\prime}(Y) =gk​Yk+O⁑(Yk+1)\displaystyle=g_{k}Y^{k}+\operatorname{O}(Y^{k+1})
w​(Y)\displaystyle w(Y) =Y+wr​Yr+O⁑(Yr+1),\displaystyle=Y+w_{r}Y^{r}+\operatorname{O}(Y^{r+1}),

with fn,gk,wrβ‰ 0f_{n},g_{k},w_{r}\neq 0. Since ww is nontorsion, we can replace ww by w∘pβ„“w^{\circ p^{\ell}} for ℓ≫0\ell\gg 0 and assume that rβ©Ύk+1r\geqslant k+1. We have

w(m)∘f\displaystyle w^{(m)}\circ f =f​(Y)+wr(m)​f​(Y)r+O⁑(Yn​(r+1))\displaystyle=f(Y)+w_{r}^{(m)}f(Y)^{r}+\operatorname{O}(Y^{n(r+1)})
=f​(Y)+wr(m)​fnr​Yn​r+O⁑(Yn​r+1).\displaystyle=f(Y)+w_{r}^{(m)}f_{n}^{r}Y^{nr}+\operatorname{O}(Y^{nr+1}).

If k=0k=0, then n=1n=1 and we are done, so assume that kβ©Ύ1k\geqslant 1. We have

f∘w\displaystyle f\circ w =f​(Y+wr​Yr+O⁑(Yr+1))\displaystyle=f(Y+w_{r}Y^{r}+\operatorname{O}(Y^{r+1}))
=f​(Y)+wr​Yr​f′​(Y)+O⁑(Y2​r)\displaystyle=f(Y)+w_{r}Y^{r}f^{\prime}(Y)+\operatorname{O}(Y^{2r})
=f​(Y)+wr​gk​Yr+k+O⁑(Yr+k+1).\displaystyle=f(Y)+w_{r}g_{k}Y^{r+k}+\operatorname{O}(Y^{r+k+1}).

This implies that n​r=r+knr=r+k, hence (nβˆ’1)​r=k(n-1)r=k, which is impossible if r>kr>k unless n=1n=1. Hence n=1n=1 and ff is invertible. ∎

We now prove theorem A. Take uβˆˆπ„~Fu\in\widetilde{\mathbf{E}}_{F} such that valY⁑(u)=1\operatorname{val}_{Y}(u)=1 and u∘[g]=[g]∘uu\circ[g]=[g]\circ u for all g∈π’ͺFΓ—g\in\mathcal{O}_{F}^{\times}. By proposition 1, uu has a lift u^∈ and F+\hat{u}\in\and^{+}_{F} such that γ​(u^)=[χπ​(Ξ³)]∘u^\gamma(\hat{u})=[\chi_{\pi}(\gamma)]\circ\hat{u} for all Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}. By proposition 1, u^\hat{u} is a pro-FF-analytic element of 𝐁~rig,F†,r\widetilde{\mathbf{B}}^{\dagger,r}_{\mathrm{rig},F}. By proposition 1, there exists nβ©Ύ0n\geqslant 0 such that u^βˆˆΟ†qβˆ’n​(𝐀F+)\hat{u}\in\varphi_{q}^{-n}(\mathbf{A}^{+}_{F}). This implies that uβˆˆΟ†qβˆ’n​(𝐄F+)u\in\varphi_{q}^{-n}(\mathbf{E}^{+}_{F}). Hence there is an mβˆˆπ™m\in\mathbf{Z} such that f​(Y)=u​(Y)pmf(Y)=u(Y)^{p^{m}} belongs to Yβ‹…k​[[Y]]Y\cdot k[\![Y]\!] and is separable. Note that valY⁑(f)=pm\operatorname{val}_{Y}(f)=p^{m}. Take g∈1+π​π’ͺFg\in 1+\pi\mathcal{O}_{F} such that gg is nontorsion, and let w​(Y)=[g]​(Y)w(Y)=[g](Y) so that u∘w=w∘uu\circ w=w\circ u. We have f∘w=w(m)∘ff\circ w=w^{(m)}\circ f. By proposition 2, ff is invertible. This implies that valY⁑(f)=1\operatorname{val}_{Y}(f)=1, so that m=0m=0 and uu itself is invertible. Since u∘[g]=[g]∘uu\circ[g]=[g]\circ u for all g∈π’ͺFΓ—g\in\mathcal{O}_{F}^{\times}, theorem 6 of [LS07] implies that u∈Aut⁑(LT)u\in\operatorname{Aut}(\mathrm{LT}). Hence there exists a∈π’ͺFΓ—a\in\mathcal{O}_{F}^{\times} such that u​(Y)=[a]​(Y)u(Y)=[a](Y).

3. Tate traces in the Lubin-Tate setting

If F=𝐐pF=\mathbf{Q}_{p} and Ο€=p\pi=p (namely in the cyclotomic situation) the fact that, in the proof of theorem A, there exists nβ©Ύ0n\geqslant 0 such that uβˆˆΟ†qβˆ’n​(𝐄F+)u\in\varphi_{q}^{-n}(\mathbf{E}^{+}_{F}) follows from the main result of [BR22]. We now explain why the methods of ibid don’t extend to the Lubin-Tate case. More precisely, we prove that there is no Ξ“F\Gamma_{F}-equivariant kk-linear projector 𝐄~F→𝐄F\widetilde{\mathbf{E}}_{F}\to\mathbf{E}_{F} if F≠𝐐pF\neq\mathbf{Q}_{p}. Choose a coordinate TT on LT\mathrm{LT} such that logLT⁑(T)=βˆ‘nβ©Ύ0Tqn/Ο€n\log_{\mathrm{LT}}(T)=\sum_{n\geqslant 0}T^{q^{n}}/\pi^{n}, so that logLT′⁑(T)≑1modΟ€\log^{\prime}_{\mathrm{LT}}(T)\equiv 1\bmod{\pi}. Let βˆ‚=1/logLT′⁑(T)β‹…d/d​T\partial=1/\log^{\prime}_{\mathrm{LT}}(T)\cdot d/dT be the invariant derivative on LT\mathrm{LT}.

\lemmname \the\smf@thm.

We have d​γ​(Y)/d​Y≑χπ​(Ξ³)d\gamma(Y)/dY\equiv\chi_{\pi}(\gamma) in 𝐄F\mathbf{E}_{F} for all Ξ³βˆˆΞ“F\gamma\in\Gamma_{F}.

Proof.

Since logLT′≑1modΟ€\log^{\prime}_{\mathrm{LT}}\equiv 1\bmod{\pi}, we have βˆ‚=d/d​Y\partial=d/dY in 𝐄F\mathbf{E}_{F}. Applying βˆ‚βˆ˜Ξ³=χπ​(Ξ³)β€‹Ξ³βˆ˜βˆ‚\partial\circ\gamma=\chi_{\pi}(\gamma)\gamma\circ\partial to YY, we get the claim. ∎

\lemmname \the\smf@thm.

If Ξ³βˆˆΞ“F\gamma\in\Gamma_{F} is nontorsion, then 𝐄FΞ³=1=k\mathbf{E}_{F}^{\gamma=1}=k.

\propname \the\smf@thm.

If F≠𝐐pF\neq\mathbf{Q}_{p}, there is no Ξ“F\Gamma_{F}-equivariant map R:𝐄F→𝐄FR:\mathbf{E}_{F}\to\mathbf{E}_{F} such that R​(Ο†q​(f))=fR(\varphi_{q}(f))=f for all fβˆˆπ„Ff\in\mathbf{E}_{F}.

Proof.

Suppose that such a map exists, and take Ξ³βˆˆΞ“F\gamma\in\Gamma_{F} nontorsion and such that χπ​(Ξ³)≑1modΟ€\chi_{\pi}(\gamma)\equiv 1\bmod{\pi}. We first show that if fβˆˆπ„Ff\in\mathbf{E}_{F} is such that (1βˆ’Ξ³)​fβˆˆΟ†q​(𝐄F)(1-\gamma)f\in\varphi_{q}(\mathbf{E}_{F}), then fβˆˆΟ†q​(𝐄F)f\in\varphi_{q}(\mathbf{E}_{F}). Write f=f0+Ο†q​(R​(f))f=f_{0}+\varphi_{q}(R(f)) where f0=fβˆ’Ο†q​(R​(f))f_{0}=f-\varphi_{q}(R(f)), so that R​(f0)=0R(f_{0})=0 and (1βˆ’Ξ³)​f0=Ο†q​(g)βˆˆΟ†q​(𝐄F)(1-\gamma)f_{0}=\varphi_{q}(g)\in\varphi_{q}(\mathbf{E}_{F}). Applying RR, we get 0=(1βˆ’Ξ³)​R​(f0)=g0=(1-\gamma)R(f_{0})=g. Hence g=0g=0 so that (1βˆ’Ξ³)​f0=0(1-\gamma)f_{0}=0. Since 𝐄FΞ³=1=k\mathbf{E}_{F}^{\gamma=1}=k by lemma 3, this implies f0∈kf_{0}\in k, so that fβˆˆΟ†q​(𝐄F)f\in\varphi_{q}(\mathbf{E}_{F}).

However, lemma 3 and the fact that χπ​(Ξ³)≑1modΟ€\chi_{\pi}(\gamma)\equiv 1\bmod{\pi} imply that γ​(Y)=Y+fγ​(Yp)\gamma(Y)=Y+f_{\gamma}(Y^{p}) for some fΞ³βˆˆπ„Ff_{\gamma}\in\mathbf{E}_{F}, so that γ​(Yq/p)=Yq/p+Ο†q​(gΞ³)\gamma(Y^{q/p})=Y^{q/p}+\varphi_{q}(g_{\gamma}). Hence (1βˆ’Ξ³)​(Yq/p)βˆˆΟ†q​(𝐄F)(1-\gamma)(Y^{q/p})\in\varphi_{q}(\mathbf{E}_{F}) even though Yq/pY^{q/p} does not belong to Ο†q​(𝐄F)\varphi_{q}(\mathbf{E}_{F}). Therefore, no such map RR can exist. ∎

\coroname \the\smf@thm.

If F≠𝐐pF\neq\mathbf{Q}_{p}, there is no Ξ“F\Gamma_{F}-equivariant kk-linear projector Ο†qβˆ’1​(𝐄F)→𝐄F\varphi_{q}^{-1}(\mathbf{E}_{F})\to\mathbf{E}_{F}. A fortiori, there is no Ξ“F\Gamma_{F}-equivariant kk-linear projector 𝐄~F→𝐄F\widetilde{\mathbf{E}}_{F}\to\mathbf{E}_{F}.

Proof.

Given such a projector TT, we could define RR as in prop 3 by R=Tβˆ˜Ο†qβˆ’1R=T\circ\varphi_{q}^{-1}. ∎

Acknowledgements. I thank Juan Esteban RodrΓ­guez Camargo for asking me the question that motivated both this paper and [BR22].

References

  • [Ber16] L. Berger – β€œMultivariable (Ο†,Ξ“)(\varphi,\Gamma)-modules and locally analytic vectors”, Duke Math. J. 165 (2016), no. 18, p. 3567–3595.
  • [BR22] L. Berger & S. Rozensztajn – β€œDecompletion of cyclotomic perfectoid fields in positive characteristic”, preprint, 2022.
  • [Col02] P. Colmez – β€œEspaces de Banach de dimension finie”, J. Inst. Math. Jussieu 1 (2002), no. 3, p. 331–439.
  • [LS07] J. Lubin & G. Sarkis – β€œExtrinsic properties of automorphism groups of formal groups”, J. Algebra 315 (2007), no. 2, p. 874–884.
  • [Lub94] J. Lubin – β€œNonarchimedean dynamical systems”, Compositio Math. 94 (1994), no. 3, p. 321–346.
  • [ST03] P. Schneider & J. Teitelbaum – β€œAlgebras of pp-adic distributions and admissible representations”, Invent. Math. 153 (2003), no. 1, p. 145–196.
  • [Win83] J.-P. Wintenberger – β€œLe corps des normes de certaines extensions infinies de corps locaux; applications”, Ann. Sci. Γ‰cole Norm. Sup. (4) 16 (1983), no. 1, p. 59–89.