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

Local ε\varepsilon-conjecture and pp-adic differential equations

Tetsuya Ishida and Kentaro Nakamura
Abstract.

Laurent Berger attached a pp-adic differential equation 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M) with a Frobenius structure to an arbitrary de Rham (φ,Γ)(\varphi,\Gamma)-module MM over a Robba ring. In this article, we compare the local epsilon conjecture for the cyclotomic deformation of MM with that of 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M). We first define an isomorphism between the fundamental lines of their cyclotomic deformations using the second author’s results on the big exponential map. As a main result of the article, we show that this isomorphism enables us to reduce the local epsilon conjecture for the cyclotomic deformation of MM to that of 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M).

Key words and phrases:
(φ,Γ)(\varphi,\Gamma)-module, pp-adic Hodge theory

1. Introduction/Notation

In [Kat93a], Kato formulated a conjecture called the generalized Iwasawa main conjecture, which is a vast generalization of the Iwasawa main conjecture and Bloch-Kato conjecture. It claims the existence of so-called zeta isomorphisms for any family of pp-adic Galois representations of G𝐐G_{\mathbf{Q}}, interpolating the zeta elements of geometric pp-adic Galois representations. Note that a similar conjecture was formulated by Fontaine and Perrin-Riou in [FP94]. Since the zeta elements are conjectural bases in (the determinants of) the Galois cohomologies and closely related to the LL-functions, it is natural to regard the zeta isomorphisms as algebraic counterparts of the LL-functions. In [Kat93b] and [FK06], Kato’s local and global ε\varepsilon-conjectures are formulated as algebraic analogue of the functional equations of LL-functions; the local ε\varepsilon-conjecture claims the existence of the local ε\varepsilon-isomorphisms, the algebraic analogue of local ε\varepsilon-factors for families of pp-adic representations of G𝐐lG_{\mathbf{Q}_{l}}, and the global ε\varepsilon-conjecture states that the zeta isomorphisms satisfies the functional equations whose local factors are the local ε\varepsilon-isomorphisms.

The local ε\varepsilon-conjecture for lpl\neq p is proved [Yas09], [Kak14]. But for the case l=pl=p, which we treat in this paper, the existence of the local ε\varepsilon-isomorphisms are proved for limited families and the conjecture is still open. In particular, by generalizing the conjecture for (φ,Γ)(\varphi,\Gamma)-modules over relative Robba rings, the second author proves the existence of ε\varepsilon-isomorphisms for trianguline representations. The conjecture has turned out to be closely related to the Coleman isomorphisms [Kat93b] [Ven13], the Perrin-Riou maps [BB08] [LVZ13], and also the pp-adic local Langlands correspondence [Nak17b] [RJ18].

Our main theorem compares the local ε\varepsilon-isomorphisms of the following different objects. Let MM be an arbitrary de Rham (φ,Γ)(\varphi,\Gamma)-module over a Robba ring. The first object is the cyclotomic deformation of MM. The second one is the cyclotomic deformation of 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M), where 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M) is the pp-adic differential equation attached to MM by Laurent Berger. We remark that the existences of their local ε\varepsilon-isomorphisms are still conjectural. The main theorem claims that the difference of their local ε\varepsilon-isomorphisms is written as the generalized Perrin-Riou map defined by the second author in [Nak14].

To make the statement of the main theorem more precise, we recall (φ,Γ)(\varphi,\Gamma)-modules over Robba rings and the local ε\varepsilon-conjecture for them.

A (φ,Γ)(\varphi,\Gamma)-module DD is a module equipped with a suitable endomorphism φ:DD\varphi:D\to D and a continuous group action of Γ=Gal(𝐐p(μp)/𝐐p)\Gamma=\mathrm{Gal}({\mathbf{Q}_{p}}(\mu_{p^{\infty}})/{\mathbf{Q}_{p}}), where μp\mu_{p^{\infty}} is the group of pp-power roots of unity in 𝐐¯p\overline{\mathbf{Q}}_{p}. There are several specific rings over which (φ,Γ)(\varphi,\Gamma)-modules are useful to study pp-adic representations. An important case is the Robba rings L\mathcal{R}_{L} with their coefficients in local fields LL; by results of Fontaine [Fon90], Cherbonnier and Colmez [CC99] and Kedlaya [Ked08], the category of pp-adic representations over LL can be embedded fully and faithfully into the one of (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L}. A lot of important notions of pp-adic Hodge theory can be generalized to (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L}, such as the functors 𝐃cris\mathbf{D}_{{\operatorname{cris}}} and 𝐃dR\mathbf{D}_{{\operatorname{dR}}} [Ber02], or Bloch-Kato’s exponential maps [Ber03], [Nak14]. Another important feature is that, when a (φ,Γ)(\varphi,\Gamma)-module MM is de Rham, Berger attached to MM a pp-adic differential equation 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M) with Frobenius structure; as its application, one can prove the pp-adic monodromy theorem for pp-adic representations by reducing it to that for pp-adic differential equations, or Colmez-Fontaine’s theorem [Ber02], [Ber08].

In [Nak17a], the second author formulated the local ε\varepsilon-conjecture for (φ,Γ)(\varphi,\Gamma)-modules over relative Robba rings, generalizing the Kato’s conjecture for pp-adic representations. We recall only the conjecture for the cyclotomic deformations of de Rham (φ,Γ)(\varphi,\Gamma)-modules, since it is the case we treat in this paper. Let LL be a finite extension of 𝐐p{\mathbf{Q}_{p}}, and MM be a (φ,Γ)(\varphi,\Gamma)-module over the Robba ring L\mathcal{R}_{L} with coefficients in LL. Then, one can attach to MM a (graded) invertible module ΔL(M)\Delta_{L}(M) over LL and ΔLIw(M)\Delta^{\operatorname{{Iw}}}_{L}(M) over L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)} for a (φ,Γ)(\varphi,\Gamma)-module MM over L\mathcal{R}_{L}, where we put L+(Γ)=Γ(𝒲,𝒪𝒲)\mathcal{R}^{+}_{L}(\Gamma)=\Gamma(\mathcal{W},\mathcal{O}_{\mathcal{W}}) and 𝒲\mathcal{W} the Berthelot generic fiber of the Iwasawa algebra 𝒪L[[Γ]]\mathcal{O}_{L}[[\Gamma]]. When MM is de Rham, he constructed a canonical trivialization isomorphism

εLdR(M):LΔL(M).\varepsilon_{L}^{{\operatorname{dR}}}(M):L\xrightarrow{\sim}\Delta_{L}(M).

Its definition involves a lot of notions of pp-adic Hodge theory, such as the theory of local constants (ε\varepsilon-constants and LL-constants), Bloch-Kato’s exponential and dual exponential maps, Hodge-Tate weights. Then the local ε\varepsilon-conjecture in this situation claims that, there exists a unique isomorphism

εLIw(M):L+(Γ)ΔLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M):{\mathcal{R}^{+}_{L}(\Gamma)}\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L}(M)

interpolating εLdR(M(δ))\varepsilon^{\operatorname{dR}}_{L}(M(\delta)) for any de Rham character δ:ΓL×\delta:\Gamma\to L^{\times}, i.e. any character of the form δ=χkδ~\delta=\chi^{k}\tilde{\delta} for k𝐙k\in\mathbf{Z} and a finite character δ~\tilde{\delta}, where χ\chi is the cyclotomic character. More precisely, εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M) is required to make the following diagram

ΔLIw(M)fδL\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(M)\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}εLIw(M)id\scriptstyle{\varepsilon^{\operatorname{{Iw}}}_{L}(M)\otimes{\mathrm{id}}}ΔL(M(δ))\textstyle{\Delta_{L}(M(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLdR(M(δ))\scriptstyle{\varepsilon^{\mathrm{dR}}_{L}(M(\delta))}L+(Γ)fδL\textstyle{{\mathcal{R}^{+}_{L}(\Gamma)}\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces}can\scriptstyle{{\mathrm{can}}}L\textstyle{L}

commute for any de Rham character δ\delta of Γ\Gamma, where fδ:L+(Γ)L+(Γ)f_{\delta}:{\mathcal{R}^{+}_{L}(\Gamma)}\to{\mathcal{R}^{+}_{L}(\Gamma)} is a continuous homomorphism of LL-algebras given by [g]δ(g)1[g]\mapsto\delta(g)^{-1} and evδ{\mathrm{ev}}_{\delta} is a canonical isomorphism induced by the specialization a fδf_{\delta}. In the original article of Kato [Kat93b], he predicts the conjectural base ε𝒪LIw(T)\varepsilon^{\operatorname{{Iw}}}_{\mathcal{O}_{L}}(T) of an invertible 𝒪L[[Γ]]\mathcal{O}_{L}[[\Gamma]]-module Δ𝒪LIw(T)\Delta^{\operatorname{{Iw}}}_{\mathcal{O}_{L}}(T) similarly defined for any 𝒪L\mathcal{O}_{L}-representation TT of G𝐐pG_{\mathbf{Q}_{p}}. In [Nak17a], the second author predicts the equality ε𝒪LIw(T)id=εLIw(𝐃rig(T[1/p]))\varepsilon^{\operatorname{{Iw}}}_{\mathcal{O}_{L}}(T)\otimes{\mathrm{id}}=\varepsilon^{\operatorname{{Iw}}}_{L}(\mathbf{D}^{\dagger}_{\mathrm{rig}}(T[1/p])), that is, the right hand side has an integral structure in the étale case.

The following is the main theorem of this paper, which can be regarded as an extension of the studies in [Nak14] and [Nak17a]. It roughly states that, for a general de Rham (φ,Γ)(\varphi,\Gamma)-module MM over L\mathcal{R}_{L} and the pp-adic differential equation 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M) attached to MM, the differences of εL(M(δ))\varepsilon_{L}(M(\delta)) and εL(𝐍rig(M)(δ))\varepsilon_{L}(\mathbf{N}_{{\operatorname{rig}}}(M)(\delta)) for the de Rham characters δ\delta of Γ\Gamma are interpolated by the generalized Perrin-Riou map in [Nak14].

Theorem.

There exists an L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-linear isomorphism

Exp(M):ΔLIw(𝐍rig(M))ΔLIw(M)\mathrm{Exp}(M):\Delta^{\operatorname{{Iw}}}_{L}(\mathbf{N}_{{\operatorname{rig}}}(M))\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L}(M)

whose specialization at any de Rham character δ\delta of Γ\Gamma makes the following diagram

1L\textstyle{1_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLdR(𝐍rig(M)(δ))\scriptstyle{\varepsilon^{{\operatorname{dR}}}_{L}(\mathbf{N}_{{\operatorname{rig}}}(M)(\delta))}εLdR(M(δ))\scriptstyle{\varepsilon^{{\operatorname{dR}}}_{L}(M(\delta))}ΔL(𝐍rig(M)(δ))\textstyle{\Delta_{L}(\mathbf{N}_{{\operatorname{rig}}}(M)(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M)δ\scriptstyle{\mathrm{Exp}(M)_{\delta}}ΔL(M(δ))\textstyle{\Delta_{L}(M(\delta))}

commute, where the isomorphism Exp(M)δ\mathrm{Exp}(M)_{\delta} is defined by the following commutative diagram

ΔL(𝐍rig(M)(δ))\textstyle{\Delta_{L}(\mathbf{N}_{{\operatorname{rig}}}(M)(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M)δ\scriptstyle{\mathrm{Exp}(M)_{\delta}}ΔL(M(δ))\textstyle{\Delta_{L}(M(\delta))}ΔLIw(𝐍rig(M))fδL\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\Delta^{\operatorname{{Iw}}}_{L}(\mathbf{N}_{{\operatorname{rig}}}(M))\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}Exp(M)id\scriptstyle{\mathrm{Exp}(M)\otimes{\mathrm{id}}}ΔLIw(M)fδL.\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(M)\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces.}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}

In particular, if εLIw(𝐍rig(M))\varepsilon^{{\operatorname{{Iw}}}}_{L}(\mathbf{N}_{{\operatorname{rig}}}(M)) exists, then εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M) also exists and is written as

εLIw(M)=Exp(M)εLIw(𝐍rig(M)).\varepsilon^{\operatorname{{Iw}}}_{L}(M)=\mathrm{Exp}(M)\circ\varepsilon^{{\operatorname{{Iw}}}}_{L}(\mathbf{N}_{{\operatorname{rig}}}(M)).

We remark that our theorem can be regarded as a refined interpolation formula for Bloch-Kato morphisms. The isomorphism Exp(M)\mathrm{Exp}(M) is obtained by the generalized Perrin-Riou’s big exponential map

ExpM,h:Hψ1(𝐍rig(M))Hψ1(M)\mathrm{Exp}_{M,h}:\mathrm{H}_{\psi}^{1}(\mathbf{N}_{{\operatorname{rig}}}(M))\to\mathrm{H}_{\psi}^{1}(M)

of [Nak14] for de Rham (φ,Γ)(\varphi,\Gamma)-module MM, in conjunction with one of the main results, theorem δ(D)\delta(D). The big exponential maps are first introduced by Perrin-Riou [Per94] for crystalline representations and used essentially in her study of pp-adic LL-functions, and then generalized to de Rham representations [Col98] and to de Rham (φ,Γ)(\varphi,\Gamma)-modules [Nak14]. Their key feature is that they interpolate the Bloch-Kato’s morphisms of twists expM(χkδ~)\exp_{M(\chi^{k}\tilde{\delta})} and expM(χkδ~)\exp^{*}_{M(\chi^{k}\tilde{\delta})} for suitable k𝐙k\in\mathbf{Z}. The theorem can be seen as a refinement of such interpolation formulae; our big exponential map Exp(M)\mathrm{Exp}(M) interpolates, at any twists δ=χkδ~\delta=\chi^{k}\tilde{\delta} for any k𝐙k\in\mathbf{Z} and δ~\tilde{\delta}, not only the maps expM(δ)\exp_{M(\delta)}and expM(δ)\exp^{*}_{M(\delta)} but also another exponential map expf,M(δ):𝐃cris(M(δ))Hφ,γ1(M(δ))\exp_{f,M(\delta)}:\mathbf{D}_{{\operatorname{cris}}}(M(\delta))\to\mathrm{H}_{\varphi,\gamma}^{1}(M(\delta)), which is closely related with the exceptional zeros for pp-adic LL-functions. We note that, even when MM comes from a crystalline pp-adic representation, the map expf,M(δ)\exp_{f,M(\delta)} is non-zero in general and we can obtain its information via our refined formula.

We also remark a relation of our theorem to the local ε\varepsilon-conjecture itself. The local ε\varepsilon-conjecture for the cyclotomic deformation of a general de Rham (φ,Γ)(\varphi,\Gamma)-module is not proved yet, and only the following special cases are proved.

  • The case of rank 11 Galois representations (i.e. rank 11 étale (φ,Γ)(\varphi,\Gamma)-modules) is proved by Kato in [Kat93b] (proofs taking account of signs is given in [FK06] briefly and in [Ven13] in detail.)

  • The case of crystalline representations is proved by Benois and Berger in [BB08], which is generalized by Loeffler, Venjakob, and Zerbes in [LVZ13], and by Bellovin and Venjakob in [BV19].

  • The case of trianguline (φ,Γ)(\varphi,\Gamma)-modules over relative Robba rings, including all semi-stable representations and also the representations associated to finite slope overconvergent modular forms, is proved by the second author in [Nak17a].

  • The case of rank 22 Galois representations is proved by the second author in [Nak17b] in almost all cases and completed by Rodrigues Jacinto [RJ18], by showing its close relation to the pp-adic local Langlands conjecture for GL2(𝐐p)\mathrm{GL}_{2}({\mathbf{Q}_{p}}).

By the last assertion of the theorem, we can reduce the local ε\varepsilon-conjecture for the cyclotomic deformation of arbitrary de Rham (φ,Γ)(\varphi,\Gamma)-module MM to that of 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M). This reduction seems a useful approach, since 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M) is relatively simple (all of its Hodge-Tate weights are zero) and also has an additional structure of a pp-adic differential equation with a Frobenius structure so that we can utilize the theory of pp-adic differential equations. We note that such a reduction is implicitly used to prove the trianguline case, and this theorem is stated as a conjecture [Nak17a, Remark 4.15]; see also Remark 4.2.2.

The structure of the paper is as follows. In section 2, we recall definitions about (φ,Γ)(\varphi,\Gamma)-modules over Robba rings and prove the key lemma Lemma 2.2.5 on a relation of Bloch-Kato’s morphisms and distributions. In section 3, we recall (a special case of) the local ε\varepsilon-conjecture for (φ,Γ)(\varphi,\Gamma)-modules studied in [Nak17a], introduce the pp-adic differential equation 𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M) for a de Rham (φ,Γ)(\varphi,\Gamma)-module MM, and construct our big exponential map Exp(M):ΔLIw(𝐍rig(M))ΔLIw(M)\mathrm{Exp}(M):\Delta^{\operatorname{{Iw}}}_{L}(\mathbf{N}_{{\operatorname{rig}}}(M))\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L}(M); it is induced by distribution, and the construction depends heavily on [Nak14]. In section 4, we state our main theorem and prove it, by introducing the notion of genericity, deducing the proof of the general case to the case of generic, and proving the generic case by applying the key lemma.

Notation. Let pp be a prime number. We fix the algebraic closure 𝐐¯p\overline{\mathbf{Q}}_{p} of the pp-adic number field 𝐐p\mathbf{Q}_{p}. Let LL be a finite extension of 𝐐p\mathbf{Q}_{p}. Let μp\mu_{p^{\infty}} denote the group of pp-power roots of unity in 𝐐¯p\overline{\mathbf{Q}}_{p}. We fix primitive pnp^{n}-th roots of unity ζpnμp\zeta_{p^{n}}\in\mu_{p^{\infty}} such that ζpn+1p=ζpn\zeta^{p}_{p^{n+1}}=\zeta_{p^{n}} for any n𝐙1.n\geqslant\mathbf{Z}_{\geqslant 1}. The set Γ=Gal(𝐐p(μp)/𝐐p)\Gamma=\mathrm{Gal}(\mathbf{Q}_{p}(\mu_{p^{\infty}})/{\mathbf{Q}_{p}}). Let ΔΓ\Delta\subseteq\Gamma be the pp-torsion subgroup of Γ\Gamma and put pΔ=1|Δ|σΔσp_{\Delta}=\frac{1}{|\Delta|}\sum_{\sigma\in\Delta}\sigma. We fix an element γΓ\gamma\in\Gamma whose image in Γ/Δ\Gamma/\Delta is a topological generator. The cyclotomic character on Γ\Gamma is denoted by χ:Γ𝐙p×\chi:\Gamma\xrightarrow{\sim}\mathbf{Z}_{p}^{\times}, which is characterized by γ(ζ)=ζχ(γ)\gamma(\zeta)=\zeta^{\chi(\gamma)} for all ζμp\zeta\in\mu_{p^{\infty}} and γΓ\gamma\in\Gamma. For a ring RR, the objects of the category of graded invertible RR-modules are written as the pairs (,r)(\mathcal{L},r) of an invertible RR-module \mathcal{L} and a continuous function r:Spec(R)𝐙r:\mathrm{Spec}(R)\to\mathbf{Z}, and the product \boxtimes is defined by (1,r1)(2,r2):=(1R2,r1+r2)(\mathcal{L}_{1},r_{1})\boxtimes(\mathcal{L}_{2},r_{2}):=(\mathcal{L}_{1}\otimes_{R}\mathcal{L}_{2},r_{1}+r_{2}). We put 1R(R,0)1_{R}\coloneqq(R,0).

2. Review of the theory of (φ,Γ)(\varphi,\Gamma)-modules over Robba rings

In this section, we first recall the definition of (φ,Γ)(\varphi,\Gamma)-modules over Robba rings, their cohomologies, and some notions of pp-adic Hodge theory. Then, we study several kinds of morphisms defined by a distribution. Theorem 2.2.5 is the key result, which describes a relation between such morphisms and Bloch-Kato’s morphisms.

2.1. (φ,Γ)(\varphi,\Gamma)-modules over Robba rings

For each integer n𝐙1n\in\mathbf{Z}_{\geqslant 1}, put

L(n)={i𝐙aiTi|aiL,i𝐙aiTi is convergent on |ζpn1||T|<1}.\mathcal{R}_{L}^{(n)}=\set{\sum_{i\in\mathbf{Z}}a_{i}T^{i}}{a_{i}\in L,\sum_{i\in\mathbf{Z}}a_{i}T^{i}\text{ is convergent on }|\zeta_{p^{n}}-1|\leqslant|T|<1}.

We put L=n1L(n)\mathcal{R}_{L}=\cup_{n\geqslant 1}\mathcal{R}_{L}^{(n)}, with which we can equip a canonical LF-topology and we call the Robba ring over LL. Put t=log(1+T)Lt=\log(1+T)\in\mathcal{R}_{L}. There is an operator φ:LL\varphi:\mathcal{R}_{L}\to\mathcal{R}_{L} and a group action of Γ\Gamma on L\mathcal{R}_{L}, both of which are continuous and linear over LL satisfying

φ(T)=(1+T)p1,γ(T)=(1+T)χ(γ)1\varphi(T)=(1+T)^{p}-1,\,\gamma(T)=(1+T)^{\chi(\gamma)}-1

for any γΓ\gamma\in\Gamma. A tuple ((1+T)i)i=0,,p1((1+T)^{i})_{i=0,\dots,p-1} is a basis of L\mathcal{R}_{L} over φ(L)\varphi(\mathcal{R}_{L}), and we can define a map ψ:LL\psi:\mathcal{R}_{L}\to\mathcal{R}_{L} by

ψ(i=0p1φ(fi)(1+T)i)=f0\psi\quantity(\sum_{i=0}^{p-1}\varphi(f_{i})(1+T)^{i})=f_{0}

for fiLf_{i}\in\mathcal{R}_{L}. Then ψ\psi-operator turns out to be continuous and commutes with Γ\Gamma.

For each n𝐙1n\in\mathbf{Z}_{\geqslant 1}, set Ln=𝐐p(ζpn)𝐐pLL_{n}={\mathbf{Q}_{p}}(\zeta_{p^{n}})\otimes_{\mathbf{Q}_{p}}L. Then one has a continuous Γ\Gamma-equivariant homomorphism

ιn:L(n)Ln[[t]]\iota_{n}:\mathcal{R}_{L}^{(n)}\to L_{n}[[t]]

of LL-algebras such that

ιn(T)=ζpnexp(tpn)1,\iota_{n}(T)=\zeta_{p^{n}}\exp\quantity(\frac{t}{p^{n}})-1,

which satisfies the following commutative diagram

L(n)\textstyle{\mathcal{R}^{(n)}_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ιn\scriptstyle{\iota_{n}}φ\scriptstyle{\varphi}Ln[[t]]\textstyle{L_{n}[[t]]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}incl\scriptstyle{{\mathrm{incl}}}L(n+1)\textstyle{\mathcal{R}^{(n+1)}_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ιn+1\scriptstyle{\iota_{n+1}}Ln+1[[t]].\textstyle{L_{n+1}[[t]].}
Definition 2.1.1.

A (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L} is a free L\mathcal{R}_{L}-module DD of finite rank equipped with a semilinear endomorphism φ:DD\varphi:D\to D over satisfying φD=D\varphi^{*}D=D and a continuous Γ\Gamma-action commuting to φ\varphi.

The following lemma is Theorem 1.3.3 of [Ber08].

Lemma 2.1.2.

Let DD be a (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L}. Then, there exists an integer n1n\geqslant 1 such that there exists a unique Γ\Gamma-stable L(m)\mathcal{R}^{(m)}_{L}-submodule D(m)DD^{(m)}\subseteq D for each mnm\geqslant n such that for any mnm\geqslant n we have Lm+1Lm,φD(m)=D(m+1)\mathcal{R}^{m+1}_{L}\otimes_{\mathcal{R}^{m}_{L},\varphi}D^{(m)}=D^{(m+1)} and LLmD(m)=D\mathcal{R}_{L}\otimes_{\mathcal{R}^{m}_{L}}D^{(m)}=D.

The smallest integer nn satisfying the property in Lemma 2.1.2 is denoted as n(D)n(D).

For a (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L}, one can define ψ\psi-operator on DD by ψ(φ(x)f)=xψ(f)\psi(\varphi(x)\otimes f)=x\otimes\psi(f) for xDx\in D and fLf\in\mathcal{R}_{L}, which turns out to be well-defined, continuous and LL-linear.

For each nn(D)n\geqslant n(D), define

𝐃dif,n+(D)=D(n)ιn,L(n)Ln[[t]].\mathbf{D}_{{\operatorname{dif}},n}^{+}(D)=D^{(n)}\otimes_{\iota_{n},\mathcal{R}_{L}^{(n)}}L_{n}[[t]].

We put ιn:D(n)𝐃dif,n+(D):xx1\iota_{n}:D^{(n)}\to\mathbf{D}_{{\operatorname{dif}},n}^{+}(D):x\mapsto x\otimes 1 and

can=cann:𝐃dif,n+(D)𝐃dif,n+1+(D):f(t)xf(t)ιn+1(φ(x)){\mathrm{can}}={\mathrm{can}}_{n}:\mathbf{D}_{{\operatorname{dif}},n}^{+}(D)\to\mathbf{D}_{{\operatorname{dif}},n+1}^{+}(D):f(t)\otimes x\mapsto f(t)\cdot\iota_{n+1}(\varphi(x))

for f(t)Ln[[t]]f(t)\in L_{n}[[t]] and xD(n)x\in D^{(n)}. We set

𝐃dif,n(D)=𝐃dif,n+(D)[1/t],𝐃dif(+)(D)=limn𝐃dif,n(+)(D),\mathbf{D}_{{\operatorname{dif}},n}(D)=\mathbf{D}_{{\operatorname{dif}},n}^{+}(D)[1/t],\mathbf{D}_{{\operatorname{dif}}}^{(+)}(D)=\varinjlim_{n}\mathbf{D}_{{\operatorname{dif}},n}^{(+)}(D),

here the injective limit is taken over (cann)nn(D)({\mathrm{can}}_{n})_{n\geqslant n(D)}. On these modules, we define Γ\Gamma-actions diagonally.

One can consider several complexes which are related to pp-adic Hodge theory. Let RR be a topological ring, XX a topological module over RR. If XX is equipped with a continuous RR-linear Γ\Gamma-action, then we put as a complex of RR-modules

Cγ(X)[XΔγ1XΔ]C^{\bullet}_{\gamma}(X)\coloneqq[X^{\Delta}\xrightarrow{\gamma-1}X^{\Delta}]

concentrated in degree [0,1][0,1]. If XX is furthermore equipped with a continuous RR-linear action of φ\varphi or ψ\psi commuting the Γ\Gamma-action, then we put, again as a complex of RR-modules,

C,γ(X)[XΔ(γ1,1)XΔXΔ(1)(1γ)XΔ]C^{\bullet}_{*,\gamma}(X)\coloneqq[X^{\Delta}\xrightarrow{(\gamma-1,*-1)}X^{\Delta}\oplus X^{\Delta}\xrightarrow{(*-1)\oplus(1-\gamma)}X^{\Delta}]

concentrated in degree [0,2][0,2] for =φ,ψ*=\varphi,\psi, and

Cψ(X)[XΔψ1XΔ]C^{\bullet}_{\psi}(X)\coloneqq[X^{\Delta}\xrightarrow{\psi-1}X^{\Delta}]

concentrated in degree [1,2][1,2]. For each complex C(X)C^{\bullet}_{\square}(X) above, its ii-th cohomology group is denoted as Hi\mathrm{H}^{i}_{\square}. For a (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L} of rank rr, we use the following special notations

𝐃cris(D)=Hγ0(D[1/t]),𝐃dR(D)=Hγ0(𝐃dif(D)).\mathbf{D}_{{\operatorname{cris}}}(D)=\mathrm{H}_{\gamma}^{0}(D[1/t]),\quad\mathbf{D}_{{\operatorname{dR}}}(D)=\mathrm{H}_{\gamma}^{0}(\mathbf{D}_{{\operatorname{dif}}}(D)).

These spaces are of dimension r\leqslant r over LL, and we say DD is crystalline (resp. de Rham) if dimL(𝐃cris(D))=r\dim_{L}(\mathbf{D}_{{\operatorname{cris}}}(D))=r (resp. dimL(𝐃dR(D))=r\dim_{L}(\mathbf{D}_{{\operatorname{dR}}}(D))=r.) For i𝐙i\in\mathbf{Z}, we also define 𝐃dRi(D)=𝐃dR(D)ti𝐃dif+(D)\mathbf{D}_{{\operatorname{dR}}}^{i}(D)=\mathbf{D}_{{\operatorname{dR}}}(D)\cap t^{i}\mathbf{D}_{{\operatorname{dif}}}^{+}(D) and t(D)=𝐃dR(D)/𝐃dR0(D)t(D)=\mathbf{D}_{{\operatorname{dR}}}(D)/\mathbf{D}_{{\operatorname{dR}}}^{0}(D). When DD is de Rham, then we say that h𝐙h\in\mathbf{Z} is a Hodge-Tate weight of DD if 𝐃dRh(D)/𝐃dRh+1(D)0\mathbf{D}_{{\operatorname{dR}}}^{-h}(D)/\mathbf{D}_{{\operatorname{dR}}}^{-h+1}(D)\neq 0, and refer its dimension as the multiplicity of hh. We put hMh_{M} as the sum of the Hodge-Tate weights of MM with multiplicity.

2.2. Morphisms induced by distributions

Let 𝒲\mathcal{W} be the Berthelot generic fiber of the Iwasawa algebra 𝐙p[[Γ]]{\mathbf{Z}_{p}}[[\Gamma]], and define the distribution algebra L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)} as the global section Γ(𝒲,𝒪)\Gamma(\mathcal{W},\mathcal{O}). In this subsection, we consider several morphisms induced by an element of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}. Then we prove a theorem about relationships between such morphisms and Bloch-Kato morphisms, which will be used as a key ingredient for our main theorem.

We recall natural L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-actions on several objects related to a (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L}. For each nn(D)n\geqslant n(D), we can equip with D(n)D^{(n)} and 𝐃dif,n+(D)\mathbf{D}_{{\operatorname{dif}},n}^{+}(D) natural L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-actions. As in [KPX14], for each nn(D)n\geqslant n(D), we can equip D(n)D^{(n)}, D[1/t](n)D[1/t]^{(n)} with natural L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-actions, which extends to DD and D[1/t]D[1/t]. Also, for each nn(D)n\geqslant n(D), we can equip 𝐃dif,n+(D)\mathbf{D}_{{\operatorname{dif}},n}^{+}(D) with a natural L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-action. In fact, for any n1n\geqslant 1, we can equip with a natural L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-action a finite generated Ln[[t]]L_{n}[[t]]-module XX with semilinear and continuous Γ\Gamma-action with respect to the canonical Frechét topology as follows. Since one has X=limnX/tnXX=\varprojlim_{n}X/t^{n}X with the quotient X/tnXX/t^{n}X is a finite dimensional LL-vector space with LL-linear continuous Γ\Gamma-action, it suffices to define a natural L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-action on arbitrary finite dimensional LL-vector space MM with an LL-linear continuous Γ\Gamma-action. First, it is easy to see that Γ\Gamma-action on MM naturally extends to a continuous 𝒪L[[Γ]]\mathcal{O}_{L}[[\Gamma]]-action. Since MM is finite dimensional LL-vector space, the the action of 𝒪L[[Γ]][1/p]\mathcal{O}_{L}[[\Gamma]][1/p]-factors through a quotient R0R_{0} of 𝒪L[[Γ]][1/p]\mathcal{O}_{L}[[\Gamma]][1/p] of finite length. Since the maximal ideals of 𝒪L[[Γ]][1/p]\mathcal{O}_{L}[[\Gamma]][1/p] bijectively correspond to closed maximal ideals of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}, R0R_{0} is also a quotient of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}. i.e. the natural quotient map 𝒪L[[Γ]][1/p]R0\mathcal{O}_{L}[[\Gamma]][1/p]\rightarrow R_{0} factors through the inclusion 𝒪L[[Γ]][1/p]L+(Γ)\mathcal{O}_{L}[[\Gamma]][1/p]\hookrightarrow{\mathcal{R}^{+}_{L}(\Gamma)}.

From now until the end of this section, we consider the following situation. Let D,DD,D^{\prime} be (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L} such that D[1/t]=D[1/t]D[1/t]=D^{\prime}[1/t]. Then, we remark that one has 𝐃dif,m(D)=𝐃dif,m(D)\mathbf{D}_{{\operatorname{dif}},m}(D)=\mathbf{D}_{{\operatorname{dif}},m}(D^{\prime}), 𝐃dR(D)=𝐃dR(D)\mathbf{D}_{{\operatorname{dR}}}(D)=\mathbf{D}_{{\operatorname{dR}}}(D^{\prime}) and 𝐃cris(D)=𝐃cris(D)\mathbf{D}_{{\operatorname{cris}}}(D)=\mathbf{D}_{{\operatorname{cris}}}(D^{\prime}). Let λL+(Γ)\lambda\in{\mathcal{R}^{+}_{L}(\Gamma)} be any distribution. We assume that, there exists some nmax{n(D),n(D)}n\geqslant\max\set{n(D),n(D^{\prime})} such that we have

λ(𝐃dif,m+(D))𝐃dif,m+(D)\lambda(\mathbf{D}_{{\operatorname{dif}},m}^{+}(D))\subseteq\mathbf{D}_{{\operatorname{dif}},m}^{+}(D^{\prime})

in 𝐃dif,m(D)=𝐃dif,m(D)\mathbf{D}_{{\operatorname{dif}},m}(D)=\mathbf{D}_{{\operatorname{dif}},m}(D^{\prime}) for all mnm\geqslant n.

Proposition 2.2.1.

For any mnm\geqslant n, we have λ(D(m))(D)(m)\lambda(D^{(m)})\subseteq(D^{\prime})^{(m)}. In particular, we have λ(D)D\lambda(D)\subseteq D^{\prime}.

Proof.

By [Ber08, II.1], the submodule (D)(m)(D)(m)[1/t]=D(m)[1/t](D^{\prime})^{(m)}\subseteq(D^{\prime})^{(m)}[1/t]=D^{(m)}[1/t] can be written as

(D)(m)={xthD(m):ιm(x)𝐃dif,m+(D) for all mm},(D^{\prime})^{(m)}=\set{x\in t^{-h}D^{(m)}:\iota_{m^{\prime}}(x)\in\mathbf{D}^{+}_{{\operatorname{dif}},m^{\prime}}(D^{\prime})\text{ for all $m^{\prime}\geqslant m$}},

where h𝐙>0h\in\mathbf{Z}_{>0} is a sufficient large integer. Since D(m)D^{(m)} is an L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-module and ιm\iota_{m} commutes with L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-action for any mn(D)m\geqslant n(D), one has

ιm(λx)=λιm(x)𝐃dif,m+(D)\iota_{m^{\prime}}(\lambda x)=\lambda\iota_{m^{\prime}}(x)\in\mathbf{D}^{+}_{{\operatorname{dif}},m^{\prime}}(D^{\prime})

for each xD(m)x\in D^{(m)} and mmm^{\prime}\geqq m by our assumption λ(𝐃dif,m+(D))𝐃dif,m+(D)\lambda(\mathbf{D}_{{\operatorname{dif}},m}^{+}(D))\subseteq\mathbf{D}_{{\operatorname{dif}},m}^{+}(D^{\prime}) for all mnm\geqslant n, which shows that λx(D)(m)\lambda x\in(D^{\prime})^{(m)}. ∎

The following corollary is fundamental.

Corollary 2.2.2.

Multiplying by λ\lambda induces morphisms of complexes

Cφ,γ(D)Cφ,γ(D),Cψ,γ(D)Cψ,γ(D)C^{\bullet}_{\varphi,\gamma}(D)\to C^{\bullet}_{\varphi,\gamma}(D^{\prime}),\quad C^{\bullet}_{\psi,\gamma}(D)\to C^{\bullet}_{\psi,\gamma}(D^{\prime})

of LL-vector spaces,

Cψ(D)Cψ(D)C^{\bullet}_{\psi}(D)\to C^{\bullet}_{\psi}(D^{\prime})

of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-modules, and

Cγ(𝐃dif,m(+)(D))Cγ(𝐃dif,m(+)(D))C^{\bullet}_{\gamma}(\mathbf{D}_{{\operatorname{dif}},m}^{(+)}(D))\to C^{\bullet}_{\gamma}(\mathbf{D}_{{\operatorname{dif}},m}^{(+)}(D^{\prime}))

of Lm[[t]]L_{m}[[t]]-modules for each mnm\geqslant n.

Proof.

Since the operators φ,ψ\varphi,\,\psi are continuous so that they commute with the L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-action, Proposition 2.2.1 gives our assertion. ∎

By abuse of notation, we use the same expression ×λ\times\lambda for the morphisms defined in Proposition 2.2.1, the ones in Corollary 2.2.2, and the induced ones between their cohomologies, which will cause no confusion. We remark that the action ×λ\times\lambda on Hγi(𝐃dif(D))=Hγi(𝐃dif(D))\mathrm{H}^{i}_{\gamma}(\mathbf{D}_{{\operatorname{dif}}}(D))=\mathrm{H}^{i}_{\gamma}(\mathbf{D}_{{\operatorname{dif}}}(D^{\prime})), 𝐃dR(D)=𝐃dR(D)\mathbf{D}_{{\operatorname{dR}}}(D)=\mathbf{D}_{{\operatorname{dR}}}(D^{\prime}) and 𝐃cris(D)=𝐃cris(D)\mathbf{D}_{{\operatorname{cris}}}(D)=\mathbf{D}_{{\operatorname{cris}}}(D^{\prime}) is just the multiplication by λ(𝟙)L\lambda(\mathbb{1})\in L. Here, for any λL+(Γ)\lambda\in{\mathcal{R}^{+}_{L}(\Gamma)}, we denote by λ(𝟏)L\lambda(\mathbf{1})\in L the image of λ\lambda by the map f𝟏:L+(Γ)L:[γ]1f_{\mathbf{1}}:{\mathcal{R}^{+}_{L}(\Gamma)}\rightarrow L:[\gamma]\mapsto 1 (γΓ\gamma\in\Gamma).

Recall the following morphisms defined in [Nak14]:

can:Hφ,γ1(D)Hγ1(𝐃dif(D)):[(x,y)][ιn(x)],{\mathrm{can}}:\mathrm{H}_{\varphi,\gamma}^{1}(D)\to\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D)):\,[(x,y)]\mapsto[\iota_{n}(x)],
gD:𝐃dR(D)=Hγ0(𝐃dif(D))Hγ1(𝐃dif(D)):α[logχ(γ)α].g_{D}:\mathbf{D}_{{\operatorname{dR}}}(D)=\mathrm{H}_{\gamma}^{0}(\mathbf{D}_{{\operatorname{dif}}}(D))\to\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D)):\,\alpha\mapsto[\log\chi(\gamma)\alpha].

Since they commute with L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-action, we immediately obtain the following lemma.

Lemma 2.2.3.

The action ×λ\times\lambda induces the following commutative diagrams ::

Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ\scriptstyle{\times\lambda}can\scriptstyle{{\mathrm{can}}}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}can\scriptstyle{{\mathrm{can}}}Hγ1(𝐃dif(D))\textstyle{\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ(𝟙)\scriptstyle{\times\lambda(\mathbb{1})}Hγ1(𝐃dif(D)),\textstyle{\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D^{\prime})),}𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ(𝟙)\scriptstyle{\times\lambda(\mathbb{1})}gD\scriptstyle{g_{D}}𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}gD\scriptstyle{g_{D^{\prime}}}Hγ1(𝐃dif(D))\textstyle{\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ(𝟙)\scriptstyle{\times\lambda(\mathbb{1})}Hγ1(𝐃dif(D)).\textstyle{\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D^{\prime})).}

We next introduce a morphism of LL-vector spaces

expD:t(D)Hφ,γ1(D)\exp_{D}:t(D)\to\mathrm{H}_{\varphi,\gamma}^{1}(D)

called the Bloch-Kato’s exponential map, and if DD is de Rham, then we have another one

expD:Hφ,γ1(D)𝐃dR0(D)\exp^{*}_{D}:\mathrm{H}_{\varphi,\gamma}^{1}(D)\to\mathbf{D}_{{\operatorname{dR}}}^{0}(D)

called the Bloch-Kato’s dual exponential map that is the Tate dual of expD\exp_{D^{*}}. They are characterized by the following explicit formulae.

Theorem 2.2.4.

Let DD be a (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L}.

  1. (1)

    For x𝐃dR(D)x\in\mathbf{D}_{{\operatorname{dR}}}(D), there exists nn(D)n\geqslant n(D) and x~D(n)[1/t]Δ\tilde{x}\in D^{(n)}[1/t]^{\Delta} such that for any mnm\geqslant n we have

    ιm(x~)x𝐃dif,m+(D).\iota_{m}(\tilde{x})-x\in\mathbf{D}_{{\operatorname{dif}},m}^{+}(D).

    Using such an element x~\tilde{x}, we can calculate the value expD(x)\exp_{D}(x) as

    expD(x)=[(γ1)x~,(φ1)x~].\exp_{D}(x)=[(\gamma-1)\tilde{x},(\varphi-1)\tilde{x}].
  2. (2)

    We assume that DD is de Rham. Then gDg_{D} is an isomorphism and expD\exp^{*}_{D} is characterized by the following commutative diagram

    𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}gD\scriptstyle{g_{D}}expD\scriptstyle{\exp^{*}_{D}}Hγ1(𝐃dif(D))\textstyle{\mathrm{H}_{\gamma}^{1}(\mathbf{D}_{{\operatorname{dif}}}(D))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}can\scriptstyle{{\mathrm{can}}}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}=\scriptstyle{=}Hφ,γ1(D).\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D).}
Proof.

See [Nak14, Section 2.3, 2.4] or [Nak17a, Section 2B] for the definition of expD,expD\exp_{D},\exp^{*}_{D} and the proofs of the above formulae. We note that our notation expD\exp^{*}_{D} corresponds to expD\exp^{*}_{D^{*}} in those papers. ∎

To state a relation between ×λ\times\lambda and Bloch-Kato morphisms, we need some preparation.

For any γΓΓtor\gamma\in\Gamma\setminus\Gamma_{\mathrm{tor}}, we set

ω1log(χ(γ))d[γ][γ]=1log(χ(γ))d([γ]1)[γ]ΩL+(Γ)/L1,anΓ(𝒲,Ω𝒲/L1).\omega\coloneqq\frac{1}{\mathrm{log}(\chi(\gamma))}\frac{d[\gamma]}{[\gamma]}=\frac{1}{\mathrm{log}(\chi(\gamma))}\frac{d([\gamma]-1)}{[\gamma]}\in\Omega^{1,\mathrm{an}}_{{\mathcal{R}^{+}_{L}(\Gamma)}/L}\coloneqq\Gamma(\mathcal{W},\Omega^{1}_{\mathcal{W}/L}).

This is independent of the choice of γ\gamma since one has

1log(χ(γa))d[γ]a[γ]a=1alog(χ(γ))ad[γ][γ]=1log(χ(γ))d[γ][γ]\frac{1}{\mathrm{log}(\chi(\gamma^{a}))}\frac{d[\gamma]^{a}}{[\gamma]^{a}}=\frac{1}{a\mathrm{log}(\chi(\gamma))}a\frac{d[\gamma]}{[\gamma]}=\frac{1}{\mathrm{log}(\chi(\gamma))}\frac{d[\gamma]}{[\gamma]}

for any non zero apa\in\mathbb{Z}_{p}. Then, one has ΩL+(Γ)/L1,an=L+(Γ)ω\Omega^{1,\mathrm{an}}_{{\mathcal{R}^{+}_{L}(\Gamma)}/L}={\mathcal{R}^{+}_{L}(\Gamma)}\omega, which is a free L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-module of rank one. For each λL+(Γ)\lambda\in{\mathcal{R}^{+}_{L}(\Gamma)}, we define dλωL+(Γ)\frac{d\lambda}{\omega}\in{\mathcal{R}^{+}_{L}(\Gamma)} by dλ=dλωωd\lambda=\frac{d\lambda}{\omega}\cdot\omega. Explicitly, if γΓfree\gamma\in\Gamma_{\mathrm{free}} is a topological generator and λ\lambda is of the form λ=yf([γ]1)\lambda=y\otimes f([\gamma]-1) with yL[Γtor]y\in L[\Gamma_{\mathrm{tor}}] and f(T)L+f(T)\in\mathcal{R}_{L}^{+}, then one has dλω=log(χ)(γ)ydfdT([γ]1)\frac{d\lambda}{\omega}=\mathrm{log}(\chi)(\gamma)y\otimes\frac{df}{dT}([\gamma]-1).

In the following theorem, we shall compare the Bloch-Kato’s morphisms of DD and DD^{\prime} using λ\lambda. It is the key lemma to prove our main theorem.

Theorem 2.2.5.
  1. (1)

    The diagram

    𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ(𝟏)\scriptstyle{\times\lambda(\mathbf{1})}expD\scriptstyle{\exp_{D}}𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}expD\scriptstyle{\exp_{D^{\prime}}}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ\scriptstyle{\times\lambda}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D^{\prime})}

    commutes.

  2. (2)

    Assume that DD or DD^{\prime} is (thus both are) de Rham. Then the diagram

    𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ(𝟏)\scriptstyle{\times\lambda(\mathbf{1})}expD\scriptstyle{\exp^{*}_{D}}𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}expD\scriptstyle{\exp^{*}_{D^{\prime}}}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ\scriptstyle{\times\lambda}Hφ,γ1(D),\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D^{\prime}),}

    commutes.

  3. (3)

    Assume further that λ(𝟏)=0\lambda(\mathbf{1})=0 Then the diagram

    𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×dλω(𝟏)\scriptstyle{\times\frac{d\lambda}{\omega}(\mathbf{1})}expD\scriptstyle{\exp^{*}_{D}}𝐃dR(D)\textstyle{\mathbf{D}_{{\operatorname{dR}}}(D^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}expD\scriptstyle{\exp_{D^{\prime}}}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×λ\scriptstyle{\times\lambda}Hφ,γ1(D)\textstyle{\mathrm{H}_{\varphi,\gamma}^{1}(D^{\prime})}

    commutes.

Proof.

First we prove (1). Let α\alpha be an element of 𝐃dR(D)\mathbf{D}_{{\operatorname{dR}}}(D). By Theorem 2.2.4(1), there exist an integer nmax{n(D),n(D)}n\geqslant\max\{n(D),n(D^{\prime})\} and an element xD(n)[1/t]Δx\in D^{(n)}[1/t]^{\Delta} such that

ιm(x)α𝐃dif,m+(D)\iota_{m}(x)-\alpha\in\mathbf{D}^{+}_{{\operatorname{dif}},m}(D)

for any mnm\geqslant n. Then one has

expD(α)=[(γ1)x,(φ1)x]Hφ,γ1(D).\exp_{D}(\alpha)=[(\gamma-1)x,(\varphi-1)x]\in\mathrm{H}_{\varphi,\gamma}^{1}(D).

Thus, its image under the map ×λ:Hφ,γ1(D)Hφ,γ1(D)\times\lambda:\mathrm{H}_{\varphi,\gamma}^{1}(D)\to\mathrm{H}_{\varphi,\gamma}^{1}(D^{\prime}) is equal to

[λ((γ1)x),λ((φ1)x)]=[(γ1)(λx),(φ1)(λx)]Hφ,γ1(D).[\lambda((\gamma-1)x),\lambda((\varphi-1)x)]=[(\gamma-1)(\lambda x),(\varphi-1)(\lambda x)]\in\mathrm{H}_{\varphi,\gamma}^{1}(D^{\prime}).

This is nothing but expD(λα)\exp_{D^{\prime}}(\lambda\alpha) because λxD(n)[1/t]Δ\lambda x\in D^{\prime(n)}[1/t]^{\Delta} satisfies

ιm(λx)λ(α)=λ(ιm(x)α)𝐃dif,m+(D).\iota_{m}(\lambda x)-\lambda(\alpha)=\lambda(\iota_{m}(x)-\alpha)\in\mathbf{D}^{+}_{{\operatorname{dif}},m}(D^{\prime}).

for any mnm\geqslant n by Proposition 2.2.1.

(2) follows immediately by Theorem 2.2.4(2) and Lemma 2.2.3.

We shall prove (3). Assume that DD and DD^{\prime} are de Rham and λ(𝟏)=0\lambda(\mathbf{1})=0. We remark that the latter implies that one can write λpΔ=(γ1)λ0\lambda p_{\Delta}=(\gamma-1)\lambda_{0} for some λ0L+(Γ)\lambda_{0}\in{\mathcal{R}^{+}_{L}(\Gamma)}. Let [x,y]Hφ,γ1(D)[x,y]\in\mathrm{H}^{1}_{\varphi,\gamma}(D) and put α=expD([x,y])𝐃dR0(D)\alpha=\exp^{*}_{D}\quantity([x,y])\in\mathbf{D}_{{\operatorname{dR}}}^{0}(D). By replacing nn larger if necessary, we may assume that x(D(n))Δx\in(D^{(n)})^{\Delta}. Take mnm\geqslant n arbitrary. By Theorem 2.2.4(2), one has

[ιm(x)]=[logχ(γ)α]Hγ1(𝐃dif,m+(D)),[\iota_{m}(x)]=[\log\chi(\gamma)\alpha]\in\mathrm{H}^{1}_{\gamma}(\mathbf{D}_{{\operatorname{dif}},m}^{+}(D)),

and hence one obtains

ιm(x)logχ(γ)α(γ1)𝐃dif,m+(D)Δ.\iota_{m}(x)-\log\chi(\gamma)\alpha\in(\gamma-1)\mathbf{D}_{{\operatorname{dif}},m}^{+}(D)^{\Delta}.

Applying λpΔ/(γ1)=λ0L+(Γ)\lambda p_{\Delta}/(\gamma-1)=\lambda_{0}\in{\mathcal{R}^{+}_{L}(\Gamma)} on the both sides gives

ιm(λpΔγ1(x))log(χ(γ))λpΔγ1αλpΔ(𝐃dif,m+(D)Δ)𝐃dif,m+(D)Δ.\iota_{m}\quantity(\frac{\lambda p_{\Delta}}{\gamma-1}(x))-\mathrm{log}(\chi(\gamma))\frac{\lambda p_{\Delta}}{\gamma-1}\alpha\in\lambda p_{\Delta}(\mathbf{D}_{{\operatorname{dif}},m}^{+}(D)^{\Delta})\subseteq\mathbf{D}_{{\operatorname{dif}},m}^{+}(D^{\prime})^{\Delta}.

Since one has

log(χ(γ))λpΔγ1α=log(χ(γ))λpΔγ1(𝟏)α=dλω(𝟏)α,\mathrm{log}(\chi(\gamma))\frac{\lambda p_{\Delta}}{\gamma-1}\alpha=\mathrm{log}(\chi(\gamma))\frac{\lambda p_{\Delta}}{\gamma-1}(\mathbf{1})\alpha=\frac{d\lambda}{\omega}(\mathbf{1})\alpha,

we obtain

ιm(λpΔγ1(x))dλω(𝟏)α𝐃dif,m+(D)Δ.\iota_{m}\quantity(\frac{\lambda p_{\Delta}}{\gamma-1}(x))-\frac{d\lambda}{\omega}(\mathbf{1})\alpha\in\mathbf{D}_{{\operatorname{dif}},m}^{+}(D^{\prime})^{\Delta}.

Since λpΔγ1(x)(D(n))Δ(D(n)[1/t])Δ=(D(n)[1/t])Δ\frac{\lambda p_{\Delta}}{\gamma-1}(x)\in(D^{(n)})^{\Delta}\subseteq(D^{(n)}[1/t])^{\Delta}=(D^{\prime(n)}[1/t])^{\Delta} and we have taken mnm\geqslant n arbitrary, the explicit formula for expD\exp_{D^{\prime}} gives that

expD(dλω(𝟏)α)\displaystyle\exp_{D^{\prime}}\quantity(\frac{d\lambda}{\omega}(\mathbf{1})\alpha) =[(γ1)λpΔγ1(x),(φ1)λpΔγ1(x)]\displaystyle=\quantity[(\gamma-1)\frac{\lambda p_{\Delta}}{\gamma-1}(x),(\varphi-1)\frac{\lambda p_{\Delta}}{\gamma-1}(x)]
=[λx,λpΔγ1(φ1)(x)]\displaystyle=[\lambda x,\frac{\lambda p_{\Delta}}{\gamma-1}(\varphi-1)(x)]
=[λx,λpΔγ1(γ1)(y)]\displaystyle=[\lambda x,\frac{\lambda p_{\Delta}}{\gamma-1}(\gamma-1)(y)]
=λ[x,y],\displaystyle=\lambda[x,y],

which proves (3). ∎

3. Big exponential maps in the Local ε\varepsilon-conjecture for (φ,Γ)(\varphi,\Gamma)-modules

In this section, we first recall briefly the definition of the de Rham ε\varepsilon-isomorphisms for (φ,Γ)(\varphi,\Gamma)-modules, and state the local ε\varepsilon-conjecture for cyclotomic deformations. Then, we define the big exponential maps and study their several properties.

3.1. de Rham ε\varepsilon-isomorphisms for (φ,Γ)(\varphi,\Gamma)-modules

We recall de Rham ε\varepsilon-isomorphisms over Robba rings following [Nak17a].

First, for each local field L/𝐐pL/{\mathbf{Q}_{p}} and each (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L}, we define a graded line ΔL(D)\Delta_{L}(D) over LL called the fundamental line attached to DD as follows.

By [Liu08], the complex Cφ,γ(D)C^{\bullet}_{\varphi,\gamma}(D) is a perfect complex of LL-vector spaces, and we put

ΔL,1(D)=DetL(Cφ,γ(D)),\Delta_{L,1}(D)=\mathrm{Det}_{L}(C^{\bullet}_{\varphi,\gamma}(D)),

which is a graded line over LL. Here, DetL\mathrm{Det}_{L} is the determinant functor defined by Knudsen-Mumford [KM76]. We define another graded LL-vector space ΔL,2(D)\Delta_{L,2}(D) as follows. By the classification of rank 11 (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L} [Col08, Proposition 3.1], there exists a unique continuous homomorphism δDetL(D):𝐐p×L×\delta_{\mathrm{Det}_{\mathcal{R}_{L}}(D)}:{\mathbf{Q}_{p}}^{\times}\to L^{\times} such that there exists an isomorphism DetL(D)L(δDetL(D))\mathrm{Det}_{\mathcal{R}_{L}}(D)\cong\mathcal{R}_{L}(\delta_{\mathrm{Det}_{\mathcal{R}_{L}}(D)}), and we define

L(D)={xDetL(D)|φ(x)=δDetL(D)(p)x,γ(x)=δDetL(D)(χ(γ))x(γΓ)},\mathcal{L}_{L}(D)=\set{x\in\mathrm{Det}_{\mathcal{R}_{L}}(D)}{\varphi(x)=\delta_{\mathrm{Det}_{\mathcal{R}_{L}}(D)}(p)x,\gamma(x)=\delta_{\mathrm{Det}_{\mathcal{R}_{L}}(D)}(\chi(\gamma))x\ (\gamma\in\Gamma)},

which turns out to be an LL-vector space of dimension 11. We then define an graded line over LL

ΔD,2(D)=(L(D),rD),\Delta_{D,2}(D)=(\mathcal{L}_{L}(D),r_{D}),

where we put rD=rankL(D)r_{D}=\rank_{\mathcal{R}_{L}}(D). Finally, we define a graded line ΔL(D)\Delta_{L}(D) over LL called its fundamental line by

ΔL(D)=ΔL,1(D)ΔL,2(D).\Delta_{L}(D)=\Delta_{L,1}(D)\boxtimes\Delta_{L,2}(D).

We also define the fundamental line ΔLIw(D)\Delta^{\operatorname{{Iw}}}_{L}(D) for the cyclotomic deformation of a general (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L}. By [KPX14], the complex Cψ(D)C^{\bullet}_{\psi}(D) is perfect, thus we may define

ΔL,1Iw(D)DetL+(Γ)(Cψ(D)).\Delta^{{\operatorname{{Iw}}}}_{L,1}(D)\coloneqq\mathrm{Det}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(C^{\bullet}_{\psi}(D)).

We also define

ΔL,2Iw(D)ΔL,2(D)LL+(Γ),\Delta^{{\operatorname{{Iw}}}}_{L,2}(D)\coloneqq\Delta_{L,2}(D)\otimes_{L}{\mathcal{R}^{+}_{L}(\Gamma)},

and define the fundamental life for the cyclotomic deformation

ΔLIw(D)ΔL,1Iw(M)ΔL,2Iw(M).\Delta^{{\operatorname{{Iw}}}}_{L}(D)\coloneqq\Delta^{{\operatorname{{Iw}}}}_{L,1}(M)\boxtimes\Delta^{{\operatorname{{Iw}}}}_{L,2}(M).

Recall that for any continuous character δ:ΓL×\delta:\Gamma\to L^{\times}, we can consider a (φ,Γ)(\varphi,\Gamma)-module D(δ)=DeδD(\delta)=De_{\delta} with a formal element eδe_{\delta} on which we have

φ(xeδ)=φ(x)eδ,γ(xeδ)=δ(γ)γ(x)eδ.\varphi(xe_{\delta})=\varphi(x)e_{\delta},\,\gamma(xe_{\delta})=\delta(\gamma)\gamma(x)e_{\delta}.

In particular, we put D=HomL(D,L)(χ)D^{*}=\mathrm{Hom}_{\mathcal{R}_{L}}(D,\mathcal{R}_{L})(\chi).

As studied in Subsection 4A of [Nak17a], one has canonical isomorphisms

evδ,j:ΔL,jIw(D)fδLΔL,j(D(δ)),canδ,j:ΔL,jIw(D)gδL+(Γ)ΔL,jIw(D(δ)){\mathrm{ev}}_{\delta,j}:\Delta^{{\operatorname{{Iw}}}}_{L,j}(D)\otimes_{f_{\delta}}L\xrightarrow{\sim}\Delta_{L,j}(D(\delta)),\,{\mathrm{can}}_{\delta,j}:\Delta^{\operatorname{{Iw}}}_{L,j}(D)\otimes_{g_{\delta}}{\mathcal{R}^{+}_{L}(\Gamma)}\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L,j}(D(\delta))

for j=1,2,j=1,2,\emptyset, where fδ:L+(Γ)Lf_{\delta}:{\mathcal{R}^{+}_{L}(\Gamma)}\to L (resp. gδ:L+(Γ)L+(Γ)g_{\delta}:{\mathcal{R}^{+}_{L}(\Gamma)}\to{\mathcal{R}^{+}_{L}(\Gamma)}) is the continuous homomorphism of LL-algebra extending :γδ1(γ):\gamma\mapsto\delta^{-1}(\gamma) (resp. γδ1(γ)γ\gamma\mapsto\delta^{-1}(\gamma)\gamma).

The local ε\varepsilon-conjecture concerns canonical bases of the fundamental lines for de Rham (φ,Γ)(\varphi,\Gamma)-modules, which we recall briefly as follows: see [Nak17a] for the precise definition. Let MM be a de Rham (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L}. Set 1L=(L,0)1_{L}=(L,0) as the trivial line. We define the following two isomorphisms

θdR(M)\displaystyle\theta_{{\operatorname{dR}}}(M) :1LΔL,1(M)DetL(𝐃dR(M)),\displaystyle:1_{L}\xrightarrow{\sim}\Delta_{L,1}(M)\boxtimes\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M)),
fM\displaystyle f_{M} :ΔL,2(M)DetL(𝐃dR(M)).\displaystyle:\Delta_{L,2}(M)\xrightarrow{\sim}\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M)).

To define the isomorphism θdR(M)\theta_{{\operatorname{dR}}}(M), we first recall that there exist exact sequences of LL-vector spaces

C1(M):0Hφ,γ0(M)𝐃cris(M)x((1φ)x,x¯)𝐃cris(M)t(M)Hφ,γ1(M)f0,C^{\bullet}_{1}(M):0\to\mathrm{H}_{\varphi,\gamma}^{0}(M)\to\mathbf{D}_{{\operatorname{cris}}}(M)\xrightarrow{x\mapsto((1-\varphi)x,\bar{x})}\mathbf{D}_{{\operatorname{cris}}}(M)\oplus t(M)\to\mathrm{H}_{\varphi,\gamma}^{1}(M)_{f}\to 0,
C2(M):0Hφ,γ1(M)/f𝐃cris(M)𝐃dR0(M)𝐃cris(M)Hφ,γ2(M)0C^{\bullet}_{2}(M):0\to\mathrm{H}_{\varphi,\gamma}^{1}(M)_{/f}\to\mathbf{D}_{{\operatorname{cris}}}(M^{*})^{\lor}\oplus\mathbf{D}_{{\operatorname{dR}}}^{0}(M)\to\mathbf{D}_{{\operatorname{cris}}}(M^{*})^{\lor}\to\mathrm{H}_{\varphi,\gamma}^{2}(M)\to 0

obtained by the Bloch-Kato’s fundamental sequences and Tate duality, both of which are concentrated in [0,4][0,4]. We define the canonical isomorphism θdR(M)\theta_{{\operatorname{dR}}}(M) as the inverse of the isomorphism

θdR(M):ΔL,1(M)DetL(𝐃dR(M))()DetL(C1(M))[1]DetL(C2(M))()1L1Lcan1L,\theta_{{\operatorname{dR}}}(M):\Delta_{L,1}(M)\boxtimes\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M))\\ \xrightarrow{(\sharp)}\mathrm{Det}_{L}(C^{\bullet}_{1}(M))^{[-1]}\boxtimes\mathrm{Det}_{L}(C^{\bullet}_{2}(M))\xrightarrow{(\flat)}1_{L}\boxtimes 1_{L}\xrightarrow{{\mathrm{can}}}1_{L},

where the isomorphism ()(\sharp) is defined by cancellation XX11L:aff(a)X\boxtimes X^{-1}\xrightarrow{\sim}1_{L}:a\otimes f\mapsto f(a) for each graded invertible line XX and the one ()(\flat) is by the trivializations via the determinant functor.

Next we define the isomorphism fMf_{M}. Since MM is de Rham, we have 𝐃dif+(M)=𝐃dR(M)LL[[t]]\mathbf{D}_{{\operatorname{dif}}}^{+}(M)=\mathbf{D}_{{\operatorname{dR}}}(M)\otimes_{L}L_{\infty}[[t]] where L[[t]]=n1Ln[[t]]L_{\infty}[[t]]=\cup_{n\geqslant 1}L_{n}[[t]]. By Lemma.3.4 of [Nak17a], a map

L(M)𝐃dif,n(detL(M)):x1εL(W(M))1thMx\mathcal{L}_{L}(M)\to\mathbf{D}_{{\operatorname{dif}},n}(\mathrm{det}_{\mathcal{R}_{L}}(M)):x\mapsto\frac{1}{\varepsilon_{L}(W(M))}\frac{1}{t^{h_{M}}}x

for sufficient large nn induces an isomorphism fM:ΔL,2(M)DetL(𝐃dR(M))f_{M}:\Delta_{L,2}(M)\xrightarrow{\sim}\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M)). Here, the constant εL(W(M))L\varepsilon_{L}(W(M))\in L_{\infty} is defined by using the Weil-Deligne representation W(M)W(M) attached to MM and the fixed basis (ζpn)n𝐙p(1)(\zeta_{p^{n}})_{n}\in{\mathbf{Z}_{p}}(1), via the theory of ε\varepsilon-constants of Deligne-Langlands [Del73], Fontaine-Perrin-Riou [FP94].

Using θdR(M)\theta_{\operatorname{dR}}(M) and fMf_{M}, we define

εLdR(M)=(idfM1)(Γ(M)θdR(M)):1LΔL(M)\varepsilon^{\operatorname{dR}}_{L}(M)=({\mathrm{id}}\boxtimes f_{M}^{-1})\circ(\Gamma(M)\theta_{\operatorname{dR}}(M)):1_{L}\xrightarrow{\sim}\Delta_{L}(M)

and call it the de Rham ε\varepsilon-isomorphism for MM. Here, the Γ\Gamma-constant Γ(M)\Gamma(M) for MM is defined by Γ(M)=1irΓ(hi)1\Gamma(M)=\prod_{1\leqslant i\leqslant r}\Gamma^{*}(h_{i})^{-1}, where for r𝐙r\in\mathbf{Z} we put

Γ(r)={(r1)!(r1)(1)r(r)!(r0).\Gamma^{*}(r)=\begin{cases}(r-1)!&(r\geqslant 1)\\ \dfrac{(-1)^{r}}{(-r)!}&(r\leqslant 0).\end{cases}

Now we can state the local ε\varepsilon-conjecture for cyclotomic deformation for (φ,Γ)(\varphi,\Gamma)-modules.

Conjecture 3.1.1.

For each finite extension L/𝐐pL/{\mathbf{Q}_{p}} and each de Rham (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L}, there exists an isomorphism

εLIw(D):1L+(Γ)ΔLIw(D)\varepsilon^{\operatorname{{Iw}}}_{L}(D):1_{\mathcal{R}^{+}_{L}(\Gamma)}\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L}(D)

satisfying the following commutative diagram

ΔLIw(D)fδL\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(D)\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}εLIw(D)id\scriptstyle{\varepsilon^{\operatorname{{Iw}}}_{L}(D)\otimes{\mathrm{id}}}ΔL(D(δ))\textstyle{\Delta_{L}(D(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLdR(D(δ))\scriptstyle{\varepsilon^{\mathrm{dR}}_{L}(D(\delta))}1L+(Γ)fδL\textstyle{1_{\mathcal{R}^{+}_{L}(\Gamma)}\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces}can\scriptstyle{{\mathrm{can}}}1L.\textstyle{1_{L}.}

for arbitrary de Rham continuous characters δ:ΓL×\delta:\Gamma\to L^{\times}.

Since the set of all the de Rham characters is Zariski dense in the weight space 𝒲\mathcal{W}, the isomorphism εLIw(D)\varepsilon^{\operatorname{{Iw}}}_{L}(D) is uniquely determined (if it exists).

3.2. Big exponential maps

Throughout this section, let LL be a finite extension of 𝐐p{\mathbf{Q}_{p}}, MM a de-Rham (φ,Γ)(\varphi,\Gamma)-modules of rank rr over L+\mathcal{R}^{+}_{L}, and N=𝐍rig(M)N=\mathbf{N}_{{\operatorname{rig}}}(M) its associated pp-adic differential equation define by Berger in [Ber08]. Note that NN is characterized as the (φ,Γ)(\varphi,\Gamma)-module in M[1/t]M[1/t] satisfying N[1/t]=M[1/t]N[1/t]=M[1/t] and

𝐃dif,n+(N)=Ln[[t]]L𝐃dR(M)\mathbf{D}_{{\operatorname{dif}},n}^{+}(N)=L_{n}[[t]]\otimes_{L}\mathbf{D}_{{\operatorname{dR}}}(M)

for a sufficient large nn.

In this subsection, we construct the big exponential map of MM

Exp(M):ΔLIw(N)ΔLIw(M),\mathrm{Exp}(M):\Delta^{{\operatorname{{Iw}}}}_{L}(N)\xrightarrow{\sim}\Delta^{{\operatorname{{Iw}}}}_{L}(M),

and prove its properties. Its construction involves the theory of big exponential map, especially the δ(D)\delta(D)-theorem studied in [Nak14], which generalizes the original δ(V)\delta(V)-conjecture in [Per94].

First, we shall construct

Exp1:ΔL,1Iw(N)ΔL,1Iw(M)\mathrm{Exp}_{1}:\Delta^{\operatorname{{Iw}}}_{L,1}(N)\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L,1}(M)

as follows.

Since all the complexes Hψ1(D)tors[0]\mathrm{H}_{\psi}^{1}(D)_{\operatorname{tors}}[0], Hψ1(D)[0]\mathrm{H}_{\psi}^{1}(D)[0] and Hψ2(D)[0]\mathrm{H}_{\psi}^{2}(D)[0] are perfect, there exist canonical isomorphisms

DetL+(Γ)(Cψ(D))DetL+(Γ)(Hψ1(D)[0])1DetL+(Γ)(Hψ2(D)[0])\displaystyle\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(C^{\bullet}_{\psi}(D))\cong\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\mathrm{H}_{\psi}^{1}(D)[0])^{-1}\boxtimes\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\mathrm{H}_{\psi}^{2}(D)[0])
DetL+(Γ)(Hψ1(D)free[0])1DetL+(Γ)(Hψ1(D)tors[0])1DetL+(Γ)(Hψ2(D)[0]),\displaystyle\cong\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\mathrm{H}_{\psi}^{1}(D)_{{\operatorname{free}}}[0])^{-1}\boxtimes\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\mathrm{H}_{\psi}^{1}(D)_{\operatorname{tors}}[0])^{-1}\boxtimes\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\mathrm{H}_{\psi}^{2}(D)[0]),

where we set Hψ1(D)free=Hψ1(D)/Hψ1(D)tors\mathrm{H}_{\psi}^{1}(D)_{{\operatorname{free}}}=\mathrm{H}_{\psi}^{1}(D)/\mathrm{H}_{\psi}^{1}(D)_{\operatorname{tors}}. Extending the coefficients to the total fraction ring Q(L+(Γ))Q({\mathcal{R}^{+}_{L}(\Gamma)}) of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}, we have

DetL+(Γ)(Cψ(D))L+(Γ)Q(L+(Γ))\displaystyle\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(C^{\bullet}_{\psi}(D))\otimes_{{\mathcal{R}^{+}_{L}(\Gamma)}}Q({\mathcal{R}^{+}_{L}(\Gamma)})
DetQ(L+(Γ))(Hψ1(D)freeQ(L+(Γ))[0])1(1Q(L+(Γ)))11Q(L+(Γ))\displaystyle\cong\mathrm{Det}_{Q({\mathcal{R}^{+}_{L}(\Gamma)})}\quantity(\mathrm{H}_{\psi}^{1}(D)_{{\operatorname{free}}}\otimes Q({\mathcal{R}^{+}_{L}(\Gamma)})[0])^{-1}\boxtimes(1_{Q({\mathcal{R}^{+}_{L}(\Gamma)})})^{-1}\boxtimes 1_{Q({\mathcal{R}^{+}_{L}(\Gamma)})}
=(r(Hψ1(D)freeL+(Γ)Q(L+(Γ)))1,r),\displaystyle=\quantity(\bigwedge^{r}\quantity(\mathrm{H}_{\psi}^{1}(D)_{{\operatorname{free}}}\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}Q({\mathcal{R}^{+}_{L}(\Gamma)}))^{-1},\,-r),

under which the image of DetL+(Γ)(Cψ(D))\mathrm{Det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(C^{\bullet}_{\psi}(D)) is calculated as

rHψ1(D)free1charL+(Γ)(Hψ1(D)tors)charL+(Γ)(Hψ2(D))1.\bigwedge^{r}\mathrm{H}_{\psi}^{1}(D)_{{\operatorname{free}}}^{-1}\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}(\mathrm{H}_{\psi}^{1}(D)_{\operatorname{tors}})\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{2}(D))^{-1}.

On the other hand, let h>0h>0 be a sufficient large integer satisfying 𝐃dRh(M)=𝐃dR(M)\mathbf{D}_{{\operatorname{dR}}}^{-h}(M)=\mathbf{D}_{{\operatorname{dR}}}(M). Then by [Nak14, Lemma 3.6] we have (i=0h1i)(𝐃dif,m+(N))𝐃dif,m+(M)\left(\prod^{h-1}_{i=0}\nabla_{i}\right)(\mathbf{D}_{{\operatorname{dif}},m}^{+}(N))\subseteq\mathbf{D}_{{\operatorname{dif}},m}^{+}(M) for any mn(M)m\geqslant n(M), and we define a morphism Exp(h)(M):Hψ1(N)Hψ1(M)\mathrm{Exp}_{(h)}(M):\mathrm{H}_{\psi}^{1}(N)\to\mathrm{H}_{\psi}^{1}(M) as the induced one by Corollary 2.2.2;

Exp(h)(M)×i=0h1i:Hψ1(N)Hψ1(M).\mathrm{Exp}_{(h)}(M)\coloneqq\times\prod^{h-1}_{i=0}\nabla_{i}:\mathrm{H}_{\psi}^{1}(N)\to\mathrm{H}_{\psi}^{1}(M).

It induces an injective map Exp(h)¯(M):Hψ1(N)freeHψ1(M)free\overline{\mathrm{Exp}_{(h)}}(M):\mathrm{H}_{\psi}^{1}(N)_{\operatorname{free}}\to\mathrm{H}_{\psi}^{1}(M)_{\operatorname{free}}, which turns out to be injective. Since iL+(Γ)\nabla_{i}\in{\mathcal{R}^{+}_{L}(\Gamma)} is a non-zero-divisor for any integer ii, we can define a modified map Exp¯(M):rHψ1(N)freeL+(Γ)Q(L+(Γ))rHψ1(M)freeL+(Γ)Q(L+(Γ))\overline{\mathrm{Exp}}(M):\wedge^{r}\mathrm{H}_{\psi}^{1}(N)_{\operatorname{free}}\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}Q({\mathcal{R}^{+}_{L}(\Gamma)})\to\wedge^{r}\mathrm{H}_{\psi}^{1}(M)_{\operatorname{free}}\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}Q({\mathcal{R}^{+}_{L}(\Gamma)}) by

Exp¯(M)=rExp(h)¯(M)1i=1rji=1hhi1hi+jiidQ(L+(Γ)),\overline{\mathrm{Exp}}(M)=\bigwedge^{r}\overline{\mathrm{Exp}_{(h)}}(M)\otimes\frac{1}{\displaystyle\prod^{r}_{i=1}\prod^{h-h_{i}-1}_{j_{i}=1}\nabla_{h_{i}+j_{i}}}\cdot{\mathrm{id}}_{Q({\mathcal{R}^{+}_{L}(\Gamma)})},

where h1,,hrh_{1},\dots,h_{r} are the Hodge-Tate weights of MM with multiplicity. Note that the right hand side doesn’t depend on hh, which justifies our notation Exp¯(M)\overline{\mathrm{Exp}}(M).

To define Exp1(M)\mathrm{Exp}_{1}(M), the main part of Exp(M)\mathrm{Exp}(M), the following theorem is essential. It is nothing but theorem δ(D)\delta(D) in the context of the local ε\varepsilon-conjecture.

Theorem 3.2.1.

Exp¯(M)\overline{\mathrm{Exp}}(M) is an isomorphism of Q(L+(Γ))Q({\mathcal{R}^{+}_{L}(\Gamma)})-modules. Moreover, by restriction, it induces an isomorphism of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-modules

rHψ1(N)freecharL+(Γ)(Hψ1(N)tors)1charL+(Γ)(Hψ2(N))\displaystyle\bigwedge^{r}\mathrm{H}_{\psi}^{1}(N)_{\operatorname{free}}\otimes{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{1}(N)_{\operatorname{tors}})^{-1}\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{2}(N))
rHψ1(M)freecharL+(Γ)(Hψ1(M)tors)1charL+(Γ)(Hψ2(M)).\displaystyle\longrightarrow\bigwedge^{r}\mathrm{H}_{\psi}^{1}(M)_{\operatorname{free}}\otimes{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}(\mathrm{H}_{\psi}^{1}(M)_{\operatorname{tors}})^{-1}\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{2}(M)).
Proof.

Since Exp¯(M)\overline{\mathrm{Exp}}(M) is a multiplication of a product of non-zero divisors i±Q(L+(Γ))\nabla^{\pm}_{i}\in Q({\mathcal{R}^{+}_{L}(\Gamma)}) with i𝐙i\in\mathbf{Z}, it is an isomorphism as Q(L+(Γ))Q({\mathcal{R}^{+}_{L}(\Gamma)})-modules.

For the latter assertion, we first remark that in rHψ1(M)freeL+(Γ)Q(L+(Γ))\wedge^{r}\mathrm{H}_{\psi}^{1}(M)_{\operatorname{free}}\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}Q({\mathcal{R}^{+}_{L}(\Gamma)}) we have

Exp¯(M)(rHψ1(N)free)=Exp(h)¯(M)(rHψ1(N)free)(i=1rji=1hhi1hi+ji)1=detL+(Γ)(Exp(h)¯(M))rHψ1(M)free(i=1rji=1hhi1hi+ji)1,\overline{\mathrm{Exp}}(M)\quantity(\bigwedge^{r}\mathrm{H}_{\psi}^{1}(N)_{\operatorname{free}})=\overline{\mathrm{Exp}_{(h)}}(M)\quantity(\bigwedge^{r}\mathrm{H}_{\psi}^{1}(N)_{\operatorname{free}})\otimes\quantity(\displaystyle\prod^{r}_{i=1}\prod^{h-h_{i}-1}_{j_{i}=1}\nabla_{h_{i}+j_{i}})^{-1}\\ =\mathrm{det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\overline{\mathrm{Exp}_{(h)}}(M))\cdot\bigwedge^{r}\mathrm{H}_{\psi}^{1}(M)_{\operatorname{free}}\otimes\quantity(\displaystyle\prod^{r}_{i=1}\prod^{h-h_{i}-1}_{j_{i}=1}\nabla_{h_{i}+j_{i}})^{-1},

where detL+(Γ)(Exp(h)¯(M))L+(Γ)\mathrm{det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\overline{\mathrm{Exp}_{(h)}}(M))\subseteq{\mathcal{R}^{+}_{L}(\Gamma)} is the determinant ideal of Exp(h)¯(M)\overline{\mathrm{Exp}_{(h)}}(M). Therefore, the claim is equivalent to the equality

(i=1rji=1hhi1hi+ji)1detL+(Γ)(Exp(h)¯(M))charL+(Γ)(Hψ1(N)tors)1charL+(Γ)(Hψ1(M)tors)=charL+(Γ)(Hψ2(M))charL+(Γ)(Hψ2(N))1\quantity(\displaystyle\prod^{r}_{i=1}\prod^{h-h_{i}-1}_{j_{i}=1}\nabla_{h_{i}+j_{i}})^{-1}\mathrm{det}_{{\mathcal{R}^{+}_{L}(\Gamma)}}(\overline{\mathrm{Exp}_{(h)}}(M))\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{1}(N)_{\operatorname{tors}})^{-1}\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{1}(M)_{\operatorname{tors}})\\ ={\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{2}(M))\cdot{\mathrm{char}}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{2}(N))^{-1}

of fractional ideals in Q(L+(Γ))Q({\mathcal{R}^{+}_{L}(\Gamma)}), which is proved as the theorem δ(D)\delta(D) [Nak14, Theorem 3.14.]. ∎

Definition 3.2.2.

We define an isomorphism

Exp1(M):ΔL,1Iw(N)ΔL,1Iw(M)\mathrm{Exp}_{1}(M):\Delta^{\operatorname{{Iw}}}_{L,1}(N)\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L,1}(M)

as the isomorphism corresponding to the one appearing Theorem 3.2.1 under the functor [1][-1].

Second, we shall define

Exp2(M):ΔL,2Iw(N)ΔL,2Iw(M).\mathrm{Exp}_{2}(M):\Delta^{\operatorname{{Iw}}}_{L,2}(N)\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L,2}(M).
Lemma 3.2.3.

Under the canonical identification of detL(M)[1/t]\mathrm{det}_{\mathcal{R}_{L}}(M)[1/t] and detL(N)[1/t]\mathrm{det}_{\mathcal{R}_{L}}(N)[1/t], we have

detL(N)=𝐍rig(detL(M))=thMdetL(M).\mathrm{det}_{\mathcal{R}_{L}}(N)=\mathbf{N}_{{\operatorname{rig}}}(\mathrm{det}_{\mathcal{R}_{L}}(M))=t^{-h_{M}}\mathrm{det}_{\mathcal{R}_{L}}(M).
Proof.

The first equality follows from detL(𝐃dR(M))=𝐃dR(detL(M))\mathrm{det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M))=\mathbf{D}_{{\operatorname{dR}}}(\mathrm{det}_{\mathcal{R}_{L}}(M)). The second one follows from the fact that for a general 11-dimensional (φ,Γ)(\varphi,\Gamma)-module DD corresponding a continuous character δ\delta, we have 𝐍rig(D)=thDD\mathbf{N}_{{\operatorname{rig}}}(D)=t^{-h_{D}}D. This shows the second equality. ∎

Lemma 3.2.3 justifies the following definition.

Definition 3.2.4.

We define the isomorphism

Exp2(M):ΔL,2Iw(N)ΔL,2Iw(M)\mathrm{Exp}_{2}(M):\Delta^{\operatorname{{Iw}}}_{L,2}(N)\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L,2}(M)

as the scalar extension of the isomorphism

2(N)2(M);x(t)hMx.\mathcal{L}_{2}(N)\xrightarrow{\sim}\mathcal{L}_{2}(M);\,x\mapsto(-t)^{h_{M}}x.

We define the big exponential map of MM as follows.

Definition 3.2.5.

We define the isomorphism

Exp(M):ΔLIw(N)ΔLIw(M)\mathrm{Exp}(M):\Delta^{\operatorname{{Iw}}}_{L}(N)\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L}(M)

as the product Exp(M)Exp1(M)Exp2(M)\mathrm{Exp}(M)\coloneqq\mathrm{Exp}_{1}(M)\boxtimes\mathrm{Exp}_{2}(M) and call it the big exponential map of MM.

We also define relative big exponential maps, which are useful to prove our main theorem.

Definition 3.2.6.

Let MM^{\prime} be another (φ,Γ)(\varphi,\Gamma)-module such that M[1/t]=M[1/t]M[1/t]=M^{\prime}[1/t]. We define the isomorphism

Expj(M,M):ΔL,jIw(M)ΔL,jIw(M)\mathrm{Exp}_{j}(M,M^{\prime}):\Delta^{\operatorname{{Iw}}}_{L,j}(M)\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L,j}(M^{\prime})

as the composition Expj(M)Expj(M)1\mathrm{Exp}_{j}(M^{\prime})\circ\mathrm{Exp}_{j}(M)^{-1} for each j=1,2,j=1,2,\emptyset.

We note that the definition of Expj(M,M)\mathrm{Exp}_{j}(M,M^{\prime}) is justified by the equality 𝐍rig(M)=𝐍rig(M)\mathbf{N}_{{\operatorname{rig}}}(M)=\mathbf{N}_{{\operatorname{rig}}}(M^{\prime}).

The following proposition is used when we reduce the proof of our main theorem to the generic case.

Proposition 3.2.7.

Let C:0M1M2M30C^{\bullet}:0\to M_{1}\to M_{2}\to M_{3}\to 0 be an exact sequence of de Rham (φ,Γ)(\varphi,\Gamma)-modules, and 𝐍rig(C):0N1N2N30\mathbf{N}_{{\operatorname{rig}}}(C^{\bullet}):0\to N_{1}\to N_{2}\to N_{3}\to 0 the exact one corresponding CC^{\bullet} via the functor 𝐍rig\mathbf{N}_{{\operatorname{rig}}}. Then we have

ΔL,jIw(M1)ΔL,jIw(M3)\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M_{1})\boxtimes\Delta^{\operatorname{{Iw}}}_{L,j}(M_{3})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Expj(M1)Expj(M3)\scriptstyle{\mathrm{Exp}_{j}(M_{1})\boxtimes\mathrm{Exp}_{j}(M_{3})}ΔL,jIw(M2)\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Expj(M2)\scriptstyle{\mathrm{Exp}_{j}(M_{2})}ΔL,jIw(N1)ΔL,jIw(N3)\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(N_{1})\boxtimes\Delta^{\operatorname{{Iw}}}_{L,j}(N_{3})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ΔL,jIw(N2),\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(N_{2}),}

for j=1,2,j=1,2,\emptyset, where the horizontal isomorphisms are induced by CC^{\bullet} and 𝐍rig(C)\mathbf{N}_{{\operatorname{rig}}}(C^{\bullet}) respectively.

Proof.

The Hodge-Tate weights of M2M_{2} is the same as the union of the ones of M1M_{1} and M3M_{3} with multiplicity, and thus we have hM2=hM1+hM3h_{M_{2}}=h_{M_{1}}+h_{M_{3}}. This gives the commutativity for each j=1,2j=1,2 by the definition of Expj\mathrm{Exp}_{j} and so for j=j=\emptyset. ∎

Big exponential maps are compatible with twists by characters on Γ\Gamma as follows.

Lemma 3.2.8.

Let δ:ΓL×\delta:\Gamma\to L^{\times} be a de Rham character. Then the diagram

ΔL,jIw(M)gδL+(Γ)\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M)\otimes_{g_{\delta}}{\mathcal{R}^{+}_{L}(\Gamma)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}canδ,j\scriptstyle{{\mathrm{can}}_{\delta,j}}Expj(M,M)id\scriptstyle{\mathrm{Exp}_{j}(M,M^{\prime})\otimes{\mathrm{id}}}ΔL,jIw(M(δ))\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Expj(M(δ),M(δ))\scriptstyle{\mathrm{Exp}_{j}(M(\delta),M^{\prime}(\delta))}ΔL,jIw(M)gδL+(Γ)\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M^{\prime})\otimes_{g_{\delta}}{\mathcal{R}^{+}_{L}(\Gamma)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}canδ,j\scriptstyle{{\mathrm{can}}_{\delta,j}}ΔL,jIw(M(δ)),\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M^{\prime}(\delta)),}

commutes for j=1,2,j=1,2,\emptyset.

Proof.

The case j=2j=2 can be checked easily by definition. The case j=1j=1 follows from the facts that the first term of a big exponential map are induced by the multiplication of a product of i±\nabla^{\pm}_{i} for i𝐙i\in\mathbf{Z} by definition, and that

gδ(i)=gδ(logγlogχ(γ)i)=logδ1(γ)γlogχ(γ)i=(logγlogχ(γ)k)i=i+k,g_{\delta}(\nabla_{i})=g_{\delta}\quantity(\frac{\log\gamma}{\log\chi(\gamma)}-i)=\frac{\log\delta^{-1}(\gamma)\gamma}{\log\chi(\gamma)}-i=\quantity(\frac{\log\gamma}{\log\chi(\gamma)}-k)-i=\nabla_{i+k},

where kk is the Hodge-Tate weight of δ\delta. ∎

4. Interpolation formula of Exp(M)\mathrm{Exp}(M) for local ε\varepsilon-isomorphisms

In this section, we first state the main result and its corollary. Its proof will be divided into the next three subsections. We utilize an explicit construction of εLIw\varepsilon^{\operatorname{{Iw}}}_{L}-isomorphisms for rank 11 (φ,Γ)(\varphi,\Gamma)-modules, which is one of the main results in [Nak17a].

4.1. Statement of main result

Let LL be a finite extension of 𝐐p{\mathbf{Q}_{p}}, MM a de-Rham (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L}, and N=𝐍rig(M)N=\mathbf{N}_{{\operatorname{rig}}}(M) the pp-adic differential equation corresponding to MM. For any character δ:ΓL×\delta:\Gamma\to L^{\times}, we denote Expj(M)δ\mathrm{Exp}_{j}(M)_{\delta} as the isomorphism commuting the diagram

ΔL,j(N(δ))\textstyle{\Delta_{L,j}(N(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Expj(M)δ\scriptstyle{\mathrm{Exp}_{j}(M)_{\delta}}ΔL,j(M(δ))\textstyle{\Delta_{L,j}(M(\delta))}ΔL,jIw(N)fδL\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\Delta^{\operatorname{{Iw}}}_{L,j}(N)\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}Expj(M)id\scriptstyle{\mathrm{Exp}_{j}(M)\otimes{\mathrm{id}}}ΔL,jIw(M)fδL\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M)\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}

for each j=1,2,j=1,2,\emptyset.

The following is the main theorem of this paper.

Theorem 4.1.1.

For any de Rham character δ:ΓL×\delta:\Gamma\to L^{\times}, the diagram

1L\textstyle{1_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLdR(N(δ))\scriptstyle{\varepsilon^{{\operatorname{dR}}}_{L}(N(\delta))}εLdR(M(δ))\scriptstyle{\varepsilon^{{\operatorname{dR}}}_{L}(M(\delta))}ΔL(N(δ))\textstyle{\Delta_{L}(N(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M)δ\scriptstyle{\mathrm{Exp}(M)_{\delta}}ΔL(M(δ))\textstyle{\Delta_{L}(M(\delta))}

commutes.

Since de Rham ε\varepsilon-isomorphisms are compatible with base change, a similar statement for any de Rham character δ:Γ𝐐¯p×\delta:\Gamma\to\overline{\mathbf{Q}}_{p}^{\times} is deduced from the above case by enlarging LL if necessary.

Remark 4.1.2.

Since εdR\varepsilon^{\operatorname{dR}}-isomorphisms consist particularly of Bloch-Kato’s exponential maps and dual exponential maps, Theorem 4.1.1 can be regarded as a generalized interpolation formula of big exponential maps in the context of the local ε\varepsilon-conjecture; our theorem treats general de Rham (φ,Γ)(\varphi,\Gamma)-modules and covers all of the twists by de Rham characters on Γ\Gamma, that is, χkδ~\chi^{k}\tilde{\delta} for any k𝐙k\in\mathbf{Z} and any finite character δ~\tilde{\delta}.

We also remark that, in a case such as 𝐃cris(M)φ=10\mathbf{D}_{{\operatorname{cris}}}(M^{*})^{\varphi=1}\neq 0, our theorem gives a non-trivial information of another exponential map expf,M:𝐃cris(M)Hφ,γ1(M)\exp_{f,M}:\mathbf{D}_{{\operatorname{cris}}}(M)\to\mathrm{H}_{\varphi,\gamma}^{1}(M), by which we can study exceptional zeros of pp-adic LL-functions (See [Ben14] for example.)

The following corollary is an important consequence.

Corollary 4.1.3.

The existence of εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M) is equivalent to that of εLIw(N)\varepsilon^{{\operatorname{{Iw}}}}_{L}(N) for N=𝐍rig(M)N=\mathbf{N}_{{\operatorname{rig}}}(M). More precisely, if one of them exists, then the other one also exists and we have the following commutative diagram:

1L+(Γ)\textstyle{1_{{\mathcal{R}^{+}_{L}(\Gamma)}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLIw(N)\scriptstyle{\varepsilon^{{\operatorname{{Iw}}}}_{L}(N)}εLIw(M)\scriptstyle{\varepsilon^{{\operatorname{{Iw}}}}_{L}(M)}ΔLIw(N)\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(N)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M)\scriptstyle{\mathrm{Exp}(M)}ΔLIw(M).\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(M).}
Proof.

If εLIw(N)\varepsilon^{{\operatorname{{Iw}}}}_{L}(N) (resp. εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M)) exists, then we define εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M) (resp. εLIw(N)\varepsilon^{{\operatorname{{Iw}}}}_{L}(N)) by

εLIw(M):=Exp(M)εLIw(N)(resp.εLIw(N):=Exp(M)1εLIw(M)).\varepsilon^{{\operatorname{{Iw}}}}_{L}(M):=\mathrm{Exp}(M)\circ\varepsilon^{{\operatorname{{Iw}}}}_{L}(N)\ \ (\text{resp}.\varepsilon^{{\operatorname{{Iw}}}}_{L}(N):=\mathrm{Exp}(M)^{-1}\circ\varepsilon^{{\operatorname{{Iw}}}}_{L}(M)).

Since the isomorphism εLIw(N)\varepsilon^{{\operatorname{{Iw}}}}_{L}(N) (resp. εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M)) satisfies the commutative diagram in Conjecture 3.1.1 for arbitrary de Rham character δ\delta by assumption, the isomorphism εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M) (resp. εLIw(N)\varepsilon^{{\operatorname{{Iw}}}}_{L}(N)) also satisfies the commutative diagram for arbitrary de Rham δ\delta (in Conjecture) by Theorem 4.1.1, which shows that εLIw(M)\varepsilon^{{\operatorname{{Iw}}}}_{L}(M) (resp. εLIw(N)\varepsilon^{{\operatorname{{Iw}}}}_{L}(N)) satisfies the conjecture. ∎

By this corollary, the conjecture for all the de Rham (φ,Γ)(\varphi,\Gamma)-modules is reduced to that for de Rham (φ,Γ)(\varphi,\Gamma)-modules with a structure of pp-adic differential equation (equivalently, de Rham (φ,Γ)(\varphi,\Gamma)-modules with all Hodge-Tate weights 0). This equivalence was in fact effectively used to prove the conjecture for rank 11 case in [Nak17a] (see also Remark 4.2.2).

Remark 4.1.4.

Assume that MM is crystalline. We remark that Theorem 4.1.3 gives an alternative construction of εLIw(M)\varepsilon^{\operatorname{{Iw}}}_{L}(M) (cf. [Nak17a]). In this case, the canonical map

LL𝐃cris(M)N\mathcal{R}_{L}\otimes_{L}\mathbf{D}_{{\operatorname{cris}}}(M)\to N

is an isomorphism as (φ,Γ)(\varphi,\Gamma)-modules, and we can easily construct εLIw(N)\varepsilon^{\operatorname{{Iw}}}_{L}(N); its scalar extension with respect to the canonical homomorphism L+(Γ)L(Γ){\mathcal{R}^{+}_{L}(\Gamma)}\to\mathcal{R}_{L}(\Gamma) is induced by a composition of isomorphisms

L(Γ)L𝐃cris(M)Lψ=0L𝐃cris(M)(LL𝐃cris(M))ψ=0Nψ=0Nψ=1L+(Γ)L(Γ),\mathcal{R}_{L}(\Gamma)\otimes_{L}\mathbf{D}_{{\operatorname{cris}}}(M)\cong\mathcal{R}_{L}^{\psi=0}\otimes_{L}\mathbf{D}_{{\operatorname{cris}}}(M)\\ \cong\quantity(\mathcal{R}_{L}\otimes_{L}\mathbf{D}_{{\operatorname{cris}}}(M))^{\psi=0}\cong N^{\psi=0}\cong N^{\psi=1}\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}\mathcal{R}_{L}(\Gamma),

where the first isomorphism is obtained by the one

L(Γ)(L)ψ=0;λλ((1+T)1),\mathcal{R}_{L}(\Gamma)\cong(\mathcal{R}_{L})^{\psi=0};\lambda\mapsto\lambda\quantity((1+T)^{-1}),

and the last isomorphism is obtained by the map

1φ:Nψ=1Nψ=0.1-\varphi:N^{\psi=1}\to N^{\psi=0}.

Thus, using Theorem 4.1.3, we obtain another construction of εLIw(M)\varepsilon^{\operatorname{{Iw}}}_{L}(M).

Before proving the main theorem, we shall state an equivalent version of Theorem 4.1.1. As before, let LL be a finite extension of 𝐐p{\mathbf{Q}_{p}} and M,MM,M^{\prime} de-Rham (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L} with M[1/t]=M[1/t]M[1/t]=M^{\prime}[1/t]. For any character δ:ΓL×\delta:\Gamma\to L^{\times}, we denote Expj(M,M)δ\mathrm{Exp}_{j}(M,M^{\prime})_{\delta} as the isomorphism commuting the diagram

ΔL,j(M(δ))\textstyle{\Delta_{L,j}(M(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Expj(M,M)δ\scriptstyle{\mathrm{Exp}_{j}(M,M^{\prime})_{\delta}}ΔL,j(M(δ))\textstyle{\Delta_{L,j}(M^{\prime}(\delta))}ΔL,jIw(M)fδL\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\Delta^{\operatorname{{Iw}}}_{L,j}(M)\otimes_{f_{\delta}}L\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}Expj(M,M)id\scriptstyle{\mathrm{Exp}_{j}(M,M^{\prime})\otimes{\mathrm{id}}}ΔL,jIw(M)fδ\textstyle{\Delta^{\operatorname{{Iw}}}_{L,j}(M^{\prime})\otimes_{f_{\delta}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}evδ\scriptstyle{{\mathrm{ev}}_{\delta}}

for each j=1,2,j=1,2,\emptyset.

Theorem 4.1.5.

For any de Rham character δ:ΓL×\delta:\Gamma\to L^{\times}, the diagram

1L\textstyle{1_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLdR(M(δ))\scriptstyle{\varepsilon^{{\operatorname{dR}}}_{L}(M(\delta))}εLdR(M(δ))\scriptstyle{\varepsilon^{{\operatorname{dR}}}_{L}(M^{\prime}(\delta))}ΔL(M(δ))\textstyle{\Delta_{L}(M(\delta))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M,M)δ\scriptstyle{\mathrm{Exp}(M,M^{\prime})_{\delta}}ΔL(M(δ))\textstyle{\Delta_{L}(M^{\prime}(\delta))}

commutes.

We shall prove Theorem 4.1.1 in the rest of the paper as follows. In the subsection 4.2, we prove Theorem 4.1.1 for rank 11 (φ,Γ)(\varphi,\Gamma)-modules. In the subsection 4.3, we introduce a special class of (φ,Γ)(\varphi,\Gamma)-modules called generic, and reduce the proof of 4.1.1 for general (φ,Γ)(\varphi,\Gamma)-modules to that for generic ones using the result for rank 11 case. In the final subsection 4.4, we complete the proof of Theorem 4.1.1 by proving Theorem 4.1.5 for generic (φ,Γ)(\varphi,\Gamma)-modules.

4.2. Proof for rank one case

We prove Theorem 4.1.1 when MM is of rank 11. We utilize the explicit construction of εLIw(M)\varepsilon^{\operatorname{{Iw}}}_{L}(M) obtained in [Nak17a].

Theorem 4.2.1.

When MM is of rank 11, the diagram of Theorem 4.1.1 commutes.

Proof.

By Theorem 3.11 of [Nak17a], the isomorphisms εLIw(M)\varepsilon^{\operatorname{{Iw}}}_{L}(M) and εLIw(N)\varepsilon^{\operatorname{{Iw}}}_{L}(N) exist. Moreover, since we have N=thMMN=t^{-h_{M}}M, it suffices to show that the diagram

1L+(Γ)\textstyle{1_{{\mathcal{R}^{+}_{L}(\Gamma)}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εLIw(M)\scriptstyle{\varepsilon^{{\operatorname{{Iw}}}}_{L}(M)}εLIw(tM)\scriptstyle{\varepsilon^{{\operatorname{{Iw}}}}_{L}(tM)}ΔLIw(M)\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(M)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M,tM)\scriptstyle{\mathrm{Exp}(M,tM)}ΔLIw(tM)\textstyle{\Delta^{\operatorname{{Iw}}}_{L}(tM)}

commutes.

By the explicit construction in Section 4A of [Nak17a], for a general (φ,Γ)(\varphi,\Gamma)-module DD of rank 11, the isomorphism

εLIw(D)L+(Γ)idL(Γ):1L(Γ)ΔLIw(D)L+(Γ)L(Γ)\varepsilon^{\operatorname{{Iw}}}_{L}(D)\otimes_{{\mathcal{R}^{+}_{L}(\Gamma)}}{\mathrm{id}}_{\mathcal{R}_{L}(\Gamma)}:1_{\mathcal{R}_{L}(\Gamma)}\xrightarrow{\sim}\Delta^{\operatorname{{Iw}}}_{L}(D)\otimes_{{\mathcal{R}^{+}_{L}(\Gamma)}}{\mathcal{R}_{L}(\Gamma)}

are obtained by the isomorphisms

θ1=1φ:ΔL,1Iw(M)L+(Γ)L(Γ)((Leδ)ψ=0,1)1,\theta_{1}=1-\varphi:\Delta^{\operatorname{{Iw}}}_{L,1}(M)\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}\mathcal{R}_{L}(\Gamma)\cong\quantity((\mathcal{R}_{L}e_{\delta})^{\psi=0},1)^{-1},
θ2:L(Γ)LLeδD(Leδ)ψ=0;λeδDλ((1+X)1eδD),\theta_{2}:\mathcal{R}_{L}(\Gamma)\otimes_{L}Le_{\delta_{D}}\xrightarrow{\sim}(\mathcal{R}_{L}e_{\delta})^{\psi=0};\lambda\otimes e_{\delta_{D}}\mapsto\lambda((1+X)^{-1}e_{\delta_{D}}),

where we put δD:𝐐p×L×\delta_{D}:{\mathbf{Q}_{p}}^{\times}\to L^{\times} as the character corresponding to DD. Since Exp1(M,tM)\mathrm{Exp}_{1}(M,tM) is induced by multiplying hM\nabla_{h_{M}}, and we can calculate

hM(λ((1+X)1eδM))\displaystyle\nabla_{h_{M}}(\lambda((1+X)^{-1}e_{\delta_{M}})) =λ((0((1+X)1eδM))hM(1+X)1eδM)\displaystyle=\lambda((\nabla_{0}((1+X)^{-1}e_{\delta_{M}}))-h_{M}(1+X)^{-1}e_{\delta_{M}})
=λ(t(1+X)1eδM+(1+X)1(hMeδM)hM(1+X)1eδM)\displaystyle=\lambda(-t(1+X)^{-1}e_{\delta_{M}}+(1+X)^{-1}(h_{M}e_{\delta_{M}})-h_{M}(1+X)^{-1}e_{\delta_{M}})
=λ((1+X)1teδM),\displaystyle=-\lambda((1+X)^{-1}te_{\delta_{M}}),

our assertion follows from the equality Exp2(M,tM)(eδM)=teδM\mathrm{Exp}_{2}(M,tM)(e_{\delta_{M}})=-te_{\delta_{M}}. ∎

Remark 4.2.2.

Theorem 4.2.1 shows that our main theorem exactly generalizes the Proposition 4.13 in [Nak17a], which is proved in a different way and used in the proof of the local ε\varepsilon-conjecture for rank 11 (φ,Γ)(\varphi,\Gamma)-modules.

4.3. Reduction to generic case

In this subsection, we define genericity of a (φ,Γ)(\varphi,\Gamma)-module and reduce the proof of our main theorem for the general case to that the generic case.

Definition 4.3.1.

A (φ,Γ)(\varphi,\Gamma)-module DD over L\mathcal{R}_{L} is generic if for any character δ:Γ𝐐¯p×\delta:\Gamma\to\overline{\mathbf{Q}}_{p}^{\times} we have 𝐃cris(D(δ))=0\mathbf{D}_{{\operatorname{cris}}}(D(\delta))=0 and 𝐃cris(D(δ))=0.\mathbf{D}_{{\operatorname{cris}}}(D(\delta)^{*})=0.

For general (φ,Γ)(\varphi,\Gamma)-modules D,DD,D^{\prime} with D[1/t]=D[1/t]D[1/t]=D^{\prime}[1/t], DD is generic if and only if DD^{\prime} is generic by definition of 𝐃cris\mathbf{D}_{{\operatorname{cris}}}. In particular, if a generic (φ,Γ)(\varphi,\Gamma)-module MM is de Rham, then the attached pp-adic differential equation N=𝐍rig(M)N=\mathbf{N}_{{\operatorname{rig}}}(M) is also generic.

Lemma 4.3.2.

Let DD be a generic (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L} of rank rr. Then we have Hψ2(D)=Hψ2(D)=0\mathrm{H}_{\psi}^{2}(D)=\mathrm{H}_{\psi}^{2}(D^{*})=0.

Proof.

We shall show Hψ2(D)=0\mathrm{H}_{\psi}^{2}(D)=0. For any continuous character δ:Γ𝐐¯p×\delta:\Gamma\to\overline{\mathbf{Q}}_{p}^{\times}, we have Hφ,γ0(D(δ))(D(δ))Γ𝐃cris(D(δ))=0\mathrm{H}_{\varphi,\gamma}^{0}(D(\delta)^{*})\subseteq(D(\delta)^{*})^{\Gamma}\subseteq\mathbf{D}_{{\operatorname{cris}}}(D(\delta)^{*})=0, and the Tate duality gives Hφ,γ2(D(δ))=0\mathrm{H}_{\varphi,\gamma}^{2}(D(\delta))=0. On the other hand, we have Hψ2(D)L+(Γ)(L+(Γ)/𝔪δ)DΔ/(ψ1,𝔪δ)Hφ,γ2(D(δ1))=0\mathrm{H}_{\psi}^{2}(D)\otimes_{\mathcal{R}^{+}_{L}(\Gamma)}({\mathcal{R}^{+}_{L}(\Gamma)}/\mathfrak{m}_{\delta})\cong D^{\Delta}/(\psi-1,\mathfrak{m}_{\delta})\cong\mathrm{H}_{\varphi,\gamma}^{2}(D(\delta^{-1}))=0 where 𝔪δL+(Γ)\mathfrak{m}_{\delta}\subseteq{\mathcal{R}^{+}_{L}(\Gamma)} is the corresponding maximal ideal. Since the set {𝔪δ}δ:Γ𝐐¯p×\{\mathfrak{m}_{\delta}\}_{\delta:\Gamma\to\overline{\mathbf{Q}}_{p}^{\times}} coincides with the set of all the maximal ideals of L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)} and Hψ2(D)\mathrm{H}_{\psi}^{2}(D) is a torsion coadmissible L+(Γ){\mathcal{R}^{+}_{L}(\Gamma)}-module, it gives Hψ2(D)=0\mathrm{H}_{\psi}^{2}(D)=0. ∎

The cohomologies of a (φ,Γ)(\varphi,\Gamma)-module whose second ψ\psi-cohomology and that of its dual vanish are quite simple.

Lemma 4.3.3.

Let DD be a (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L} of rank rr such that Hψ2(D)=Hψ2(D)=0\mathrm{H}_{\psi}^{2}(D)=\mathrm{H}_{\psi}^{2}(D^{*})=0. Then one has Hφ,γi(D)=0\mathrm{H}_{\varphi,\gamma}^{i}(D)=0 for i=0,2i=0,2, and dimL(Hφ,γ1(D))=r\dim_{L}(\mathrm{H}_{\varphi,\gamma}^{1}(D))=r. Also, the first ψ\psi-cohomology Hψ1(D)\mathrm{H}_{\psi}^{1}(D) is free of rank rr over RL+(Γ)R_{L}^{+}(\Gamma) and for any continuous character δ:ΓL×\delta:\Gamma\to L^{\times}, one has a canonical isomorphism Hψ1(D)fδLHφ,γ1(M(δ))\mathrm{H}_{\psi}^{1}(D)\otimes_{f_{\delta}}L\cong\mathrm{H}_{\varphi,\gamma}^{1}(M(\delta)).

Proof.

For example, see [KPX14, Section 5]. ∎

We reduce the proof of our main theorem to the generic case via the next proposition.

Proposition 4.3.4.

Assume that Theorem 4.1.1 holds for any LL and for all of the de Rham generic (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L}. Then, Theorem 4.1.1 holds unconditionally.

Proof.

Let MM be a de Rham (φ,Γ)(\varphi,\Gamma)-module over L\mathcal{R}_{L} of rank rr. We prove Theorem 4.1.1 for MM and N=𝐍rig(M)N=\mathbf{N}_{{\operatorname{rig}}}(M) by induction on rr.

The base case r=1r=1 has been proved as Theorem 4.2.1.

Suppose that r2r\geqslant 2. We assume that Theorem 4.1.1 holds for all the de Rham (φ,Γ)(\varphi,\Gamma)-modules over L\mathcal{R}_{L} of rank r1\leqq r-1. If MM is not generic, we have 𝐃cris(M(δ))0\mathbf{D}_{{\operatorname{cris}}}(M(\delta))\neq 0 for some character δ:Γ𝐐¯p×\delta:\Gamma\to\overline{\mathbf{Q}}_{p}^{\times}. Extending LL if necessary, we may assume that δ(Γ)L×\delta(\Gamma)\subseteq L^{\times} and that there is a nonzero φ\varphi-eigen vector xeδ𝐃cris(M(δ))=(M(δ)[1/t])Γx\otimes e_{\delta}\in\mathbf{D}_{{\operatorname{cris}}}(M(\delta))=(M(\delta)[1/t])^{\Gamma} with xM[1/t]x\in M[1/t]. Then, the submodule L[1/t]xM[1/t]\mathcal{R}_{L}[1/t]x\subseteq M[1/t] is stable under (φ,Γ)(\varphi,\Gamma)-actions. Since L\mathcal{R}_{L} is a Bézout domain, it turns that out that its saturation ML[1/t]xMM(δ)M^{\prime}\coloneqq\mathcal{R}_{L}[1/t]x\cap M\subsetneq M(\delta) and the quotient M(δ)/MM(\delta)/M^{\prime} are (φ,Γ)(\varphi,\Gamma)-modules. Therefore, by considering an exact sequence

0MM(δ)M(δ)/M00\to M^{\prime}\to M(\delta)\to M(\delta)/M^{\prime}\to 0

of de Rham (φ,Γ)(\varphi,\Gamma)-modules, Lemma 3.2.7 gives our assertion. ∎

4.4. Proof for generic case

We continue to use the same notation as in the previous section.

For a technical reason, we introduce another (φ,Γ)(\varphi,\Gamma)-module M+M^{+}. Let h1hrh_{1}\leqslant\dots\leqslant h_{r} be the Hodge-Tate weights of MM. Let α1,,αr𝐃dR(M)\alpha_{1},\dots,\alpha_{r}\in\mathbf{D}_{{\operatorname{dR}}}(M) be a basis; taking along the filtration of 𝐃dR(M)\mathbf{D}_{{\operatorname{dR}}}(M), we may assume that thiαi𝐃dif+(M)t^{h_{i}}\alpha_{i}\in\mathbf{D}_{{\operatorname{dif}}}^{+}(M) for each ii, and that (thiαi)1ir(t^{h_{i}}\alpha_{i})_{1\leqslant i\leqslant r} is a basis of 𝐃dif+(M)\mathbf{D}_{{\operatorname{dif}}}^{+}(M). Then Theorem II.1.2 in [Ber08] gives that there exists a unique (φ,Γ)(\varphi,\Gamma)-module M+MM^{+}\subseteq M such that

𝐃dif,n(M)+(M+)=Ln(M)[[t]]th1+1α1(2irLn(M)[[t]]thiαi).\mathbf{D}^{+}_{{\operatorname{dif}},n(M)}(M^{+})=L_{n(M)}[[t]]\cdot t^{h_{1}+1}\alpha_{1}\oplus\quantity(\bigoplus_{2\leqslant i\leqslant r}L_{n(M)}[[t]]\cdot t^{h_{i}}\alpha_{i}).

Note that, since the big exponential maps are transitive by definition, thNM,Mt^{h}N\subseteq M,M^{\prime} for a sufficient large h𝐙>0h\in\mathbf{Z}_{>0}, and thNt^{h}N is obtained by the above procedure repeatedly starting from MM, it suffices to prove the case M=M+M^{\prime}=M^{+}. Moreover, by Lemma 3.2.8, we may assume that δ=𝟏\delta=\mathbf{1}.

In summary, it is sufficient to prove that the diagram

1L\textstyle{1_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}εdR(M)\scriptstyle{\varepsilon_{{\operatorname{dR}}}(M)}εdR(M+)\scriptstyle{\varepsilon_{{\operatorname{dR}}}(M^{+})}ΔL(M)\textstyle{\Delta_{L}(M)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp(M,M+)𝟏\scriptstyle{\mathrm{Exp}(M,M^{+})_{\mathbf{1}}}ΔL(M+).\textstyle{\Delta_{L}(M^{+}).}

commutes.

Lemma 4.4.1.

The diagram

DetL(𝐃dR(M))\textstyle{\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}×(1)\scriptstyle{\times(-1)}fM\scriptstyle{f_{M}}DetL(𝐃dR(M+))\textstyle{\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M^{+}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}fM+\scriptstyle{f_{M^{+}}}ΔL,2(M)\textstyle{\Delta_{L,2}(M)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp2(M,M+)𝟏\scriptstyle{\mathrm{Exp}_{2}(M,M^{+})_{\mathbf{1}}}ΔL,2(M+),\textstyle{\Delta_{L,2}(M^{+}),}

commutes.

Proof.

This follows from the direct calculation

fM+(Exp2(M,M+)𝟏(x))\displaystyle f_{M^{+}}(\mathrm{Exp}_{2}(M,M^{+})_{\mathbf{1}}(x)) =fM+(tx)\displaystyle=f_{M^{+}}(-tx)
=1ε(M+)1thM+φn(tx)\displaystyle=-\frac{1}{\varepsilon(M^{+})}\frac{1}{t^{h_{M^{+}}}}\otimes\varphi^{n}(tx)
=1ε(M+)tthM+φn(x)\displaystyle=-\frac{1}{\varepsilon(M^{+})}\frac{t}{t^{h_{M^{+}}}}\otimes\varphi^{n}(x)
=1ε(M+)1thMφn(x)\displaystyle=-\frac{1}{\varepsilon(M^{+})}\frac{1}{t^{h_{M}}}\otimes\varphi^{n}(x)
=fM(x),\displaystyle=-f_{M}(x),

where xL(M)x\in\mathcal{L}_{L}(M) is any element and nmax{n(M),n(M+)}n\geqslant\max\set{n(M),\,n(M^{+})}. We note that the last equality follows from the fact that for two de Rham (φ,Γ)(\varphi,\Gamma)-modules D,DD,D^{\prime} with D[1/t]=D[1/t]D[1/t]=D^{\prime}[1/t], the corresponding filtered (φ,N,G𝐐p)(\varphi,\,N,\,G_{{\mathbf{Q}_{p}}})-modules are the same, so are the attached ε\varepsilon-constants. ∎

Thus, the main theorem is deduced from the following lemma.

Lemma 4.4.2.

The diagram

1L\textstyle{1_{L}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Γ(M)θ(M)\scriptstyle{\Gamma(M)\theta(M)}Γ(M+)θ(M+)\scriptstyle{\Gamma(M^{+})\theta(M^{+})}ΔL,1(M)LDetL(𝐃dR(M))\textstyle{\Delta_{L,1}(M)\boxtimes_{L}\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Exp1(M,M+)𝟏Lid\scriptstyle{-\mathrm{Exp}_{1}(M,M^{+})_{\mathbf{1}}\otimes_{L}{\mathrm{id}}}ΔL,1(M+)LDetL(𝐃dR(M+))\textstyle{\Delta_{L,1}(M^{+})\boxtimes_{L}\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M^{+}))}

commutes.

Proof.

By Lemma 4.3.4, we may assume that MM is generic, which implies M+M^{+} is also generic. In the following, we use a letter DD to denote a general generic de Rham (φ,Γ)(\varphi,\Gamma)-module. We say =Exp1(M,M+)𝟏Lid\mathcal{E}=\mathrm{Exp}_{1}(M,M^{+})_{\mathbf{1}}\otimes_{L}{\mathrm{id}} for short.

We first give explicit descriptions of the isomorphisms appearing in the diagram. By Lemma 4.3.3, we have canonical quasi-isomorphisms Cψ(D)Hψ1(D)[1]C^{\bullet}_{\psi}(D)\cong\mathrm{H}_{\psi}^{1}(D)[1] and Cφ,γ(D)Hφ,γ1(D)[1]C^{\bullet}_{\varphi,\gamma}(D)\cong\mathrm{H}_{\varphi,\gamma}^{1}(D)[1]. The canonical base change isomorphism

ΔL,1Iw(D)f𝟏LΔL,1(D)\Delta^{\operatorname{{Iw}}}_{L,1}(D)\otimes_{f_{\mathbf{1}}}L\xrightarrow{\sim}\Delta_{L,1}(D)

is thus the image under [1][-1]-functor of the isomorphism

DetL+(Γ)(Hψ1(D))f1LDetL(Hψ,γ1(D)):(rxi)1r[p1plogχ(γ)pΔ(xi),0].\mathrm{Det}_{\mathcal{R}^{+}_{L}(\Gamma)}\quantity(\mathrm{H}_{\psi}^{1}(D))\otimes_{f_{1}}L\xrightarrow{\sim}\mathrm{Det}_{L}(\mathrm{H}_{\psi,\gamma}^{1}(D)):(\wedge^{r}x_{i})\otimes 1\mapsto\bigwedge^{r}\quantity[\frac{p-1}{p}\log\chi(\gamma)p_{\Delta}(x_{i}),0].

Therefore, the isomorphism Exp1(M,M+)𝟏:Δ1(M)Δ1(M+)\mathrm{Exp}_{1}(M,M^{+})_{\mathbf{1}}:\Delta_{1}(M)\xrightarrow{\sim}\Delta_{1}(M^{+}) is obtained as the image under [1][-1]-functor of the isomorphism

DetL(Hψ,γ1(M))DetL(Hψ,γ1(M+)):rxih1(rxi).\mathrm{Det}_{L}(\mathrm{H}_{\psi,\gamma}^{1}(M))\xrightarrow{\sim}\mathrm{Det}_{L}(\mathrm{H}_{\psi,\gamma}^{1}(M^{+})):\wedge^{r}x_{i}\mapsto\nabla_{h_{1}}(\wedge^{r}x_{i}).

Next we consider θ(D)\theta(D). Under the assumption of genericity, we have Hφ,γi(D)=0\mathrm{H}_{\varphi,\gamma}^{i}(D)=0 for i=0,2i=0,2 and 𝐃cris(D)=0\mathbf{D}_{{\operatorname{cris}}}(D)=0 again by Lemma 4.3.3, so θ(D)\theta(D) is simply obtained via the trivializations of the exact sequences

0t(D)3expDHφ,γ1(D)f,40,0\to t(D)_{3}\xrightarrow{\exp_{D}}\mathrm{H}_{\varphi,\gamma}^{1}(D)_{f,4}\to 0,
0Hφ,γ1(D)/f,1expD𝐃dR0(D)20,0\to\mathrm{H}_{\varphi,\gamma}^{1}(D)_{/f,1}\xrightarrow{\exp^{*}_{D}}\mathbf{D}_{{\operatorname{dR}}}^{0}(D)_{2}\to 0,
0𝐃dR0(D)1𝐃dR(D)2t(D)30,0\to\mathbf{D}_{{\operatorname{dR}}}^{0}(D)_{1}\to\mathbf{D}_{{\operatorname{dR}}}(D)_{2}\to t(D)_{3}\to 0,
0Hφ,γ1(D)f,1Hφ,γ1(D)2Hφ,γ1(D)/f,30,0\to\mathrm{H}_{\varphi,\gamma}^{1}(D)_{f,1}\to\mathrm{H}_{\varphi,\gamma}^{1}(D)_{2}\to\mathrm{H}_{\varphi,\gamma}^{1}(D)_{/f,3}\to 0,

where the index appearing at each space expresses its degree in the sequences and the last two sequences are canonical ones. More explicitly, θ(D):1L(DetL(Hψ,γ1(D)))1DetL(𝐃dR(D))\theta(D):1_{L}\xrightarrow{\sim}(\mathrm{Det}_{L}(\mathrm{H}_{\psi,\gamma}^{1}(D)))^{-1}\boxtimes\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(D)) is written as follows: if we put d0(D)=dimL(𝐃dR0(D))d_{0}(D)=\dim_{L}(\mathbf{D}_{{\operatorname{dR}}}^{0}(D)), then for any basis (βi)1ir𝐃dR(D)(\beta_{i})_{1\leqslant i\leqslant r}\in\mathbf{D}_{{\operatorname{dR}}}(D) such that (βi)d0(D)+1ir(\beta_{i})_{d_{0}(D)+1\leqslant i\leqslant r} spans 𝐃dR0(D)\mathbf{D}_{{\operatorname{dR}}}^{0}(D), θ(D)1\theta(D)^{-1} is described as

θ(D)1:[expD(β1¯)expD(βrd0(D)¯)βrd0(D)+1DβrD1](rβi)(1)d0(D),\theta(D)^{-1}:\quantity[\exp_{D}(\overline{\beta_{1}})\wedge\dots\wedge\exp_{D}(\overline{\beta_{r-d_{0}(D)}})\wedge\beta^{*_{D}}_{r-d_{0}(D)+1}\wedge\dots\wedge\beta^{*_{D}}_{r}\mapsto 1]\otimes(\wedge^{r}\beta_{i})\\ \mapsto(-1)^{d_{0}(D)},

where βiD\beta^{*_{D}}_{i} are any lifts of βi\beta_{i} with respect to expD\exp^{*}_{D}.

Using the above descriptions, we can say the asserted commutativity in more concrete form. We put an element XX of (DetL(Hψ,γ1(D(M))))1DetL(𝐃dR(M))(\mathrm{Det}_{L}(\mathrm{H}_{\psi,\gamma}^{1}(D(M))))^{-1}\boxtimes\mathrm{Det}_{L}(\mathbf{D}_{{\operatorname{dR}}}(M)) as

[expM(αr¯)expM(αd0+1¯)α1Mαd0M1](αrαd0+1α1αd0).\quantity[\exp_{M}(\overline{\alpha_{r}})\wedge\dots\wedge\exp_{M}(\overline{\alpha_{d_{0}+1}})\wedge\alpha^{*_{M}}_{1}\wedge\dots\wedge\alpha^{*_{M}}_{d_{0}}\mapsto 1]\otimes(\alpha_{r}\wedge\dots\wedge\alpha_{d_{0}+1}\wedge\alpha_{1}\wedge\dots\wedge\alpha_{d_{0}}).

Then, since XX is a basis by the definition of (αi)1ir(\alpha_{i})_{1\leqslant i\leqslant r}, our claim deduces to show the commutativity at XX, that is, the equality

Γ(M+)1θ(M+)1((X))=Γ(M)1θ(M)1(X),\Gamma(M^{+})^{-1}\theta(M^{+})^{-1}(-\mathcal{E}(X))=\Gamma(M)^{-1}\theta(M)^{-1}(X),

or, furthermore, by the description of θ(M)\theta(M) above, the equality

Γ(M+)1θ(M+)1((X))=(1)d0Γ(M)1.-\Gamma(M^{+})^{-1}\theta(M^{+})^{-1}(\mathcal{E}(X))=(-1)^{d_{0}}\Gamma(M)^{-1}.

By our construction of M+M^{+}, we have h1(𝐃dif,m+(M))𝐃dif,m+(M+)\nabla_{h_{1}}(\mathbf{D}_{{\operatorname{dif}},m}^{+}(M))\subseteq\mathbf{D}_{{\operatorname{dif}},m}^{+}(M^{+}) for all mn(M)m\geqslant n(M). Therefore, we can verify the above equality essentially by Lemma 2.2.5 as follows.

For the case h1<0h_{1}<0, Lemma 2.2.5 (i) gives that h1(α1M)=h1α1M+\nabla_{h_{1}}(\alpha^{*_{M}}_{1})=-h_{1}\alpha^{*_{M^{+}}}_{1}, so one obtains

(X)=(h1)1[expM+(αr¯)expM+(αd0+1¯)α1M+αd0M+1](αrαd0+1α1αd0).\mathcal{E}(X)=(-h_{1})^{-1}\quantity[\exp_{M^{+}}(\overline{\alpha_{r}})\wedge\dots\wedge\exp_{M^{+}}(\overline{\alpha_{d_{0}+1}})\wedge\alpha^{*_{M^{+}}}_{1}\wedge\dots\wedge\alpha^{*_{M^{+}}}_{d_{0}}\mapsto 1]\\ \otimes(\alpha_{r}\wedge\dots\wedge\alpha_{d_{0}+1}\wedge\alpha_{1}\wedge\dots\wedge\alpha_{d_{0}}).

Since we have 𝐃dR0(M+)=𝐃dR0(M)\mathbf{D}_{{\operatorname{dR}}}^{0}(M^{+})=\mathbf{D}_{{\operatorname{dR}}}^{0}(M), by the the description of θ(M+)\theta(M^{+}) we obtain

θ(M+)1((X))=(1)d0(M+)(h1)1=(1)d0+1h11.\theta(M^{+})^{-1}(\mathcal{E}(X))=(-1)^{d_{0}(M^{+})}(-h_{1})^{-1}=(-1)^{d_{0}+1}h^{-1}_{1}.

Thus, the desired equality is rewritten as

h11Γ(M+)1=Γ(M)1,h^{-1}_{1}\Gamma(M^{+})^{-1}=\Gamma(M)^{-1},

which clearly holds since Γ(D)1\Gamma(D)^{-1} is the product of the Hodge-Tate weights of DD with multiplicity and by the relation Γ(k+1)=kΓ(k)\Gamma^{*}(k+1)=k\cdot\Gamma^{*}(k) for any nonzero k𝐙k\in\mathbf{Z}.

The case h1>0h_{1}>0 follows similarly to the previous case h1<0h_{1}<0, by using Lemma 2.2.5 (ii) instead of Lemma 2.2.5 (i).

For the last case h1=0h_{1}=0, canceling Γ(M)=Γ(M+)\Gamma(M)=\Gamma(M^{+}) from the equality our assertion becomes the following one:

θ(M+)1((X))=(1)d0.-\theta(M^{+})^{-1}\quantity(\mathcal{E}(X))=(-1)^{d_{0}}.

Lemma 2.2.5 (iii) gives that 0(α1M)=expM+(α1¯)\nabla_{0}(\alpha^{*_{M}}_{1})=\exp_{M^{+}}(\overline{\alpha_{1}}), we obtain

(X)=[expM+(αr¯)expM+(αd0+1¯)expM+(α1¯)α2M+αd0M+1](αrαd0+1α1αd0).\mathcal{E}(X)=\quantity[\exp_{M^{+}}(\overline{\alpha_{r}})\wedge\dots\wedge\exp_{M^{+}}(\overline{\alpha_{d_{0}+1}})\wedge\exp_{M^{+}}(\overline{\alpha_{1}})\wedge\alpha^{*_{M^{+}}}_{2}\dots\wedge\alpha^{*_{M^{+}}}_{d_{0}}\mapsto 1]\\ \otimes(\alpha_{r}\wedge\dots\wedge\alpha_{d_{0}+1}\wedge\alpha_{1}\wedge\dots\wedge\alpha_{d_{0}}).

In this case, the elements α2,,αd0\alpha_{2},\dots,\alpha_{d_{0}} spans 𝐃dR0(M+)\mathbf{D}_{{\operatorname{dR}}}^{0}(M^{+}), thus we can use the previous explicit description of θ(M+)\theta(M^{+}) and obtain

θ(M+)1((X))=(1)d0(M+)=(1)d01,\theta(M^{+})^{-1}\quantity(\mathcal{E}(X))=(-1)^{d_{0}(M^{+})}=(-1)^{d_{0}-1},

which completes all the cases and finishes the proof.

Acknowledgements

This work was supported by JSPS KAKENHI Grant Number 22K03231.

References

  • [Ben14] D. Benois, On extra zeros of pp-adic LL-functions: the crystalline case, Iwasawa theory 2012, 65–133, Contrib. Math. Comput. Sci., 7, Springer, Heidelberg, 2014.
  • [BB08] D. Benois, L. Berger, Théorie d’Iwasawa des représentations cristallines. II, Comment. Math. Helv. 83, no. 3 (2008), 603–677.
  • [Ber02] L. Berger, Représentations pp-adiques et équations différentielles, Invent. Math. 𝟏𝟒𝟖\mathbf{148} (2002), 219–284.
  • [Ber03] L. Berger. Bloch and Kato’s exponential map: three explicit formulas, Doc. Math. 2003, Extra Vol., Kazuya Kato’s fiftieth birthday, 99–129 (electronic).
  • [Ber08] L. Berger, Équations différentielles pp-adiques et (φ\varphi,NN)-modules filtrés, Astérisque 319 (2008), 13–38.
  • [BV19] R. Bellovin and O. Venjakob, Wach modules, regulator maps, and ε\varepsilon-isomorphisms in families, Int. Math. Res. Not. IMRN 2019, no. 16, 5127–5204.
  • [CC99] F. Cherbonnier, P. Colmez, Théorie d’Iwasawa des représentations pp-adiques dun corps local, J. Amer. Math. Soc. 12 (1999), 241–268.
  • [Col98] P. Colmez,Théorie d’Iwasawa des représentations de de Rham d’un corps local, Ann. of Math. 148 (1998), 485–571.
  • [Col08] P. Colmez, Représentations triangulines de dimension 2, Astérisque 319 (2008), 213–258.
  • [Del73] P. Deligne, Les constantes des équations fonctionelles des fonctions LL, Modular functions of one variable, II (Proc. Internat. Summer School, Univ. Antwerp, Antwerp, 1972), pp. 501–597. Lecture Notes in Math., Vol. 349, Springer, Berlin, 1973.
  • [FP94] J.-M. Fontaine, B.Perrin-Riou, Autour des conjectures de Bloch et Kato; Cohomologie galoisienne et valeurs de fonctions LL. Motives (Seattle, WA, 1991), 599–706, Proc. Sympos. Pure Math., 55, Part 1, Amer. Math. Soc., Providence, RI, 1994.
  • [FK06] T. Fukaya, K. Kato, A formulation of conjectures on pp-adic zeta functions in non commutative Iwasawa theory, Proceedings of the St. Petersburg Mathematical Society. Vol. XII (Providence, RI), Amer. Math. Soc. Transl. Ser. 2, vol. 219, Amer. Math. Soc., 2006, pp. 1–85.
  • [Fon90] J.-M. Fontaine, Représentations pp-adiques des corps locaux, I in The Grothendieck Festschrift, Vol. 2, Progr. Math. 87, Birkhäuser, Boston, 1990, 249–309.
  • [Ked08] K.Kedlaya, Slope filtrations for relative Frobenius, Astérisque 319 (2008), 259–301.
  • [Kak14] M. Kakde, Kato’s local epsilon conjecture: lpl\neq p case, J. Lond. Math. Soc. (2) 90, no. 1 (2014), 287–308.
  • [Kat93a] K. Kato, Lectures on the approach to Iwasawa theory for Hasse-Weil LL-functions via BdRB_{dR}, Arithmetic algebraic geometry, Lecture Notes in Mathematics 1553, Springer-Verlag, Berlin, 1993, 50–63.
  • [Kat93b] K. Kato, Lectures on the approach to Iwasawa theory for Hasse-Weil LL-functions via BdRB_{{\operatorname{dR}}}. Part II. Local main conjecture, unpublished preprint.
  • [Ked04] K. Kedlaya, A pp-adic local monodromy theorem, Ann. of Math. (2) 160 (2004), 93–184.
  • [KPX14] K. Kedlaya, J. Pottharst, L. Xiao, Cohomology of arithmetic families of (φ,Γ)(\varphi,\Gamma)-modules, J. Amer. Math. Soc. 27 (2014), 1043–1115.
  • [KM76] F. Knudsen, D. Mumford, The projectivity of the moduli space of stable curves I, Math. Scand. 39 (1976), no. 1, 19–55.
  • [LVZ13] D. Loeffler, O. Venjakob, S. L. Zerbes, Local ε\varepsilon-isomorphism, Kyoto J. Math. 55, no. 1 (2015), 63–127.
  • [Liu08] R. Liu, Cohomology and duality for (φ,Γ\varphi,\Gamma)-modules over the Robba ring, Int. Math. Res. Not. IMRN, no. 3 (2008), Art. ID rnm150, 32.
  • [Nak14] K. Nakamura, Iwasawa theory of de Rham (φ,Γ)(\varphi,\Gamma)-modules over the Robba ring, J. Inst. Math. Jussieu, Volume 13, Issue 01 (2014), 65–118.
  • [Nak17a] K. Nakamura, A generalization of Kato’s local ε\varepsilon-conjecture for (φ,Γ)(\varphi,\Gamma)-modules over the Robba ring, Algebra and Number Theory, Volume 11, Number 2 (2017), 319–404.
  • [Nak17b] K. Nakamura, Local ε\varepsilon-isomorphisms for rank two pp-adic representations of Gal(¯p/p)\mathrm{Gal}(\overline{\mathbb{Q}}_{p}/\mathbb{Q}_{p}) and a functional equation of Kato’s Euler system, Cambridge Journal of Mathematics Volume 5, Number 3 (2017), 281–368.
  • [Pot13] J. Pottharst, Analytic families of finite-slope Selmer groups, Algebra and Number Theory 7-7 (2013), 1571–1612.
  • [Ven13] O. Venjakob, On Kato’s local ε\varepsilon-isomorphism conjecture for rank one Iwasawa modules, Algebra Number Theory 7 (2013), 2369–2416.
  • [Yas09] S. Yasuda, Local constants in torsion rings, J. Math. Sci. Univ. Tokyo 𝟏𝟔\mathbf{16} (2009), no. 2, 125–197.
  • [Per94] B. Perrin-Riou, Théorie d’Iwasawa des représentations pp-adiques sur un corps local. Invent. Math. 115 (1994) 81–161.
  • [RJ18] J. Rodrigues Jacinto, La conjecture ε\varepsilon locale de Kato en dimension 2, Math. Ann. 372 (2018), no. 3-4, 1277–1334.