Iwasawa theory and \(F\)-analytic Lubin-Tate \((\varphi,\Gamma)\)-modules (Q1677376)

From the introduction: ``Let \(K\) be a finite extension of \(\mathbb Q_p\) and \(G_K=\mathrm{Gal}(\mathbb Q_p/K)\). We use the theory of \((\varphi,\Gamma)\)-modules in the Lubin-Tate setting to construct some classes in \(H^1(K,V)\), for ``\(F\)-analytic'' representations \(V\) of \(G_K\). If in addition \(V\) is crystalline, we describe these classes explicitly using Bloch and Kato's exponential maps and generalize Perrin-Riou's period map to the Lubin-Tate setting.'' Let us now describe these constructions in more detail. The first step is a classification of overconvergent Lubin-Tate \((\varphi,\Gamma)\)-modules. For a finite extension \(F\) of \(\mathbb Q_p\), with uniformizer \(\pi\), let LT be the Lubin-Tate formal group attached to \(\pi\), with a fixed coordinate \(T\), so that for each \(a\in\mathcal O_F\), the multiplication by a is given by a power series \([a](T)=aT+O(T^2)\in\mathcal O_F [[T]]\). Let \(F_n=F(\mathrm{LT}[\pi^n])\), \(F_\infty=\cup_{n\geq 1}F_n\) and \(\Gamma_F=\mathrm{Gal}(F_\infty/F)\). Let \(\mathbf A_F\) denote the set of power series \(\sum_{i\in\mathbb Z}a_iT^i\) such that \(a_i\to 0\) as \(i\to-\infty\), and let \(\mathbf B_F=\mathbf A_F[1/\pi]\), which is a field endowed with a Frobenius map \(\varphi_q:f(T)\to f([\pi](T))\) and an action of \(\Gamma_F\) given by \(g:f(T)\to f([\chi_\pi (g)](T))\), where \(q\) is the cardinal of the residual field of \(F\) and \(\chi_\pi:G_F\to\mathcal O_F^\times\) is the Lubin-Tate character. If \(K\) is a finite extension of \(F\), \(\mathbf B_K\) is a finite unramified extension \(\mathbf B_K\) of \(\mathbf B_F\). Recall that a \((\varphi,\Gamma)\)-module over \(\mathbf B_K\) is a finite dimensional \(\mathbf B_K\)-vector space endowed with a compatible Frobenius map \(\varphi_q\) and action of \(\Gamma_K\), and it is called étale if it has a basis in which \(\mathrm{Mat}(\varphi_q)\in\mathrm{GL}_d(\mathbf A_K)\). A fundamental theorem of Fontaine asserts the existence of an equivalence of categories between the category of \(F\)-linear representations of \(G_K\) and the category of étale \((\varphi,\Gamma)\)-modules over \(\mathbf B_K\). A \((\varphi,\Gamma)\)-module over \(\mathbf B_K\) is called overconvergent if it has a basis in which \(\mathrm{Mat}(\varphi_q)\in\mathrm{GL}_d(\mathbf A_K)\) and \(\mathrm{Mat}(g)\in\mathrm{GL}_d(\mathbf A_K)\) for all \(g\in\Gamma_K\). If \(F=\mathbb Q_p\), every étale \((\varphi,\Gamma)\)-module over \(\mathbf B_K\) is overconvergent, but this does not hold in general. Let us say that an \(F\)-linear representation \(V\) of \(G_K\) is \textit{\(F\)-analytic} if for all embeddings \(\tau:F\to\bar{\mathbb Q}_p\), \(\tau\neq Id\), the representation \(V\otimes^\tau_F\mathbb C_p\) is trivial. The first author [Duke Math. J. 165, No. 18, 3567--3595 (2016; Zbl 1395.11084)] has shown that every \(F\)-analytic representation of \(G_K\) is overconvergent. Here the authors prove a conjecture of Colmez which describes all overconvergent representations of \(G_K\), more precisely: if \(V\) is an overconvergent representation of \(G_K\), there exists an \(F\)-analytic representation \(X_{\mathrm{an}}\) of \(G_K\), a representation \(Y_\Gamma\) of \(G_K\) which factors through \(\Gamma_K\) (hence is overconvergent), and a surjective \(G_K\)-equivariant map \(X_{\mathrm{an}}\otimes Y_\Gamma\to V\). The authors next focus on \(F\)-analytic representations. For an \(F\)-linear representation \(V\) of \(G_K\), let \(D(V)\) denote the \((\varphi,\Gamma)\)-module over \(\mathbf B_K\) attached to \(V\). If \(V\) is overconvergent, there is a well defined \((\varphi,\Gamma)\)-module \(D^\dagger (V)\) attached to \(V\) over \(\mathbf B_K^\dagger\), such that \(D(V)=\mathbf B_K\otimes_{\mathbf B^\dagger_K}D^\dagger (V)\). Here \(\mathbf B_F^\dagger\) denotes the set of power series \(f(T)\in\mathbf B_F\) with a non-empty domain of convergence. Call \(D^\dagger_{\mathrm{rig}}(V)\) the \((\varphi,\Gamma)\)-module attached to \(V\) over the Robba ring \(\mathbf B_{\mathrm{rig},K}^\dagger\) (which contains \(\mathbf B_K^\dagger\)), given by \(D^\dagger_{\mathrm{rig}}(V)=\mathbf B_{\mathrm{rig},K}^\dagger\otimes_{B^\dagger_K}D^\dagger (V)\). The ring \(\mathbf B_{\mathrm{rig},K}^\dagger\) is a free \(\varphi_q(\mathbf B_{\mathrm{rig},K}^\dagger)\)-module of degree \(q\), which allows to define a map \(\Psi_q:D^\dagger_{\mathrm{rig}} (V)\to D^\dagger_{\mathrm{rig}}(V)\) that is a \(\Gamma_K\)-equivariant left inverse of \(\varphi_q\). Starting from this, the authors' main result is the construction, for an \(F\)- analytic representation \(V\) of \(G_K\), of a collection of maps \(h^1_{K_n,V}:D^\dagger_{\mathrm{rig}} (V)^{\Psi_q=1}\to H^1 (K_n,V)\) with good functorial properties, for instance compatibility with corestriction. If moreover \(K=F\) and \(V\) is a crystalline representation of \(G_F\), the authors give explicit formulas for \(h^1_{F_n,V}\) using Bloch and Kato's exponential map. More precisely, given \(V\) as above, let \(D_{\mathrm{cris}}(V)=(\mathbf B_{\mathrm{cris},F}\otimes_FV)^{G_F}\) and \(t_\pi=\log_{\mathrm{LT}}(T)\). Let \(u=\{u_n\}_{n\geq 0}\) be a compatible system of primitive \(\pi^n\)-torsion points of LT. If \(n\geq 0\), one has a map \(\varphi_q^{-n}:\mathbf B_{\mathrm{rig},F}\to F_n[[t_\pi]]\), where \(\mathbf B_{\mathrm{rig},F}\) denotes the positive part of the Robba ring. The authors show the existence of a natural \((\varphi,\Gamma)\)-equivariant inclusion \(D _{\mathrm{rig}}^\dagger(V)^{\Psi_q=1}\to\mathbf B^+_{\mathrm{rig},F}[1/t_\pi]\otimes_FD_{\mathrm{cris}}(V)\). This produces, by composition, maps \(\varphi_q^{-n}:D^\dagger{\mathrm{rig}}(V)^{\Psi_q=1}\to F_n((t_\pi ))\otimes_FD_{\mathrm{cris}}(V)\) and \(\partial_V\circ\varphi_q^{-n}:D^\dagger _{\mathrm{rig}} (V) ^{\Psi_q=1}\to F_n\otimes_FD_{\mathrm{cris}}(V)\), where \(\partial_V\) is the ``coefficient of \(t_\pi^0\)'' map. The authors can explicitly express in terms of \(\partial_V\circ\varphi_q^{-n}\) the values \(\exp^\ast{F_n,V^\ast(1)}(h^1_{F_n,V}(y))\) for \(y\in D^\dagger_{\mathrm{rig}}(V)^{\Psi_q=1}\) and \(h^1_{F_n,V}(\nabla_{h-1}\circ\dots\circ\nabla_0(y))\) for \(y\in(\mathbf B_{\mathrm{rig},F}\otimes_FD_{\mathrm{cris}}(V))^{\Psi_q=1}\) (here \(\nabla=t_\pi.d/dt_\pi\), \(\nabla_i=\nabla-i\) if \(i\in\mathbb Z\), and \(h\geq 1\) is such that \(\mathrm{Fil}^{-h}D_{\mathrm{cris}}(V)=D_{\mathrm{cris}}(V))\). These two results give a Lubin-Tate analogue of Perrin-Riou's ``big exponential map.'' The specialization of the second to the setting \(V=F(\chi_\pi)\) corresponds to ``Lubin-Tate Kummer theory'', more specifically: let \(S\) denote the set of sequences \(x=\{x_n\}_{n\geq 1}\), with \(x_n\in (\pi_{F_n})\) and such that \(\mathrm{Tr}^{\mathrm{LT}}_{F_{n+1}/F_n}(x_{n+1})=[q/\pi](x_n)\) for \(n\geq 1\), and let \(\delta:\mathrm{LT}(\pi_{F_n})\to H^1(F_n, F(\chi_\pi))\) be the Kummer map. For \(x\in S\), there exists a power series \(f(T)\in (\mathbf B_{\mathrm{rig},F})^{\Psi_q=1/\pi}\) such that \(f(u_n)=\log_{\mathrm{LT}}(x_n)\) for \(n\geq 1\). Then \(h^1_{F_n,F(\chi_\pi)} (d/dt_\pi (f(T)).u)=(q/\pi)^{-n}.\delta(x_n)\) (this is the analogue of the theory of Coleman power series in the cyclotomic case).