\(p\)-adic Galois representations and \((\varphi,\tau)\)-modules (Q383723)

Let \(p\) be an odd prime and \(K\) be a \(p\)-adic field, i.e. a field of characteristic zero, complete w.r.t. a discrete valuation, with perfect residue field k of characteristic \(p\). Fix an algebraic closure \(\bar{K}\) of \(K\) and put \(G_K = \mathrm{Gal}(\bar{K}/K)\). One powerful tool for the study of the \(p\)-adic representations of \(G_K\) is Fontaine's theory of \((\varphi, \Gamma)\)-modules, in which the \(p\)-cyclotomic extension \(K(\zeta_{p^\infty})\) plays an essential role. One first crucial step consists in showing that the \(\mathbb Q_p\)-representations (resp. the free or torsion \(\mathbb Z_p\)-representations) of finite type of the Galois group \(H_K = \mathrm{Gal}(\bar{K}/K(\zeta_{p^\infty}))\) are classified by certain \(\varphi\)-modules (where \(\varphi\) is an endomorphism which is semi-linear w.r.t. the Frobenius automorphism) over the field \(\mathcal E\) (resp. the ring \(\mathcal E^{\mathrm{int}}\)) defined as follows: \[ \mathcal E = \left\{ \sum_{i\in\mathbb Z} a_i u^i\mid a_i \in W[1/p], (a_i) \text{ bounded, }\lim_{i\to-\infty}a_i = 0 \right\} , \] \[ \mathcal E^{\mathrm{int}}= \left\{\sum_{i\in\mathbb Z} a_i u^i\mid a_i \in W, \lim_{i\to-\infty}a_i = 0\right\} \] where \(W\) is the ring of Witt vectors with coefficients in \(k\). One recovers the complete action of \(G_K\) by adding to these \(\varphi\)-modules an action of \(\Gamma= \mathrm{Gal}(K(\zeta_{p^\infty})/K)\), the final objects being called \((\varphi, \Gamma)\)-modules. In the particular case of semi-stable representations of \(G_K\), some recent work of \textit{C. Breuil} [Math. Ann. 307, No.2, 191--224 (1997; Zbl 0883.11049)] and \textit{M. Kisin} [Algebraic geometry and number theory. Prog. Math. 253, 457--496 (2006; Zbl 1184.11052)] has shown that it would be more adequate to replace \(K(\zeta_{p^\infty})\) by the extension \(K_\infty\) of \(K\) obtained by adding a compatible system of \(p^n\)-th roots of a fixed uniformizing element \(\pi\) of \(K\). The main goal of the present paper is to develop an analog of the theory of \((\varphi, \Gamma)\)-modules starting from the extension \(K_\infty\). As before, the first step consists in classifying the \(p\)-adic representations of \(G_\infty = \mathrm{Gal}(\bar{K}/K_\infty)\) by means of \(\varphi\)-modules over \(\mathcal E\) or \(\mathcal E^{\mathrm{int}}\). But a difficulty arises when trying to replace the action of \(\Gamma\) by an action of ``\(\mathrm{Gal}(K_\infty/K)\)'' since the extension \(K_\infty/K\) is not galois. Following \textit{T. Liu} [Ann. Sci. Éc. Norm. Supér. (4) 40, No. 4, 633--674 (2007; Zbl 1163.11043)], one considers an element \(\tau\in \mathrm{Gal}(\bar{K}/K(\zeta_{p^\infty}))\) such that \(\tau\) and \(G_\infty\) generate \(G_K\). The additional action of \(\tau\) on the \(\varphi\)-module \(M\) corresponds to a semi-linear endomorphism of \(\mathcal E_\tau^{\mathrm{int}} \otimes_{\mathcal E^{\mathrm{int}}} M \), where \(\mathcal E_\tau^{\mathrm{int}}\) is a certain \(\mathcal E^{\mathrm{int}}\)-algebra on which \(G_K\) acts. A \((\varphi,\tau)\)-module over \((\mathcal E^{\mathrm{int}},\mathcal E_\tau^{\mathrm{int}})\) (resp. over \((\mathcal E,\mathcal E _\tau)\), where \(\mathcal E_\tau= \mathcal E_\tau^{\mathrm{int}} [1/p]\)) is then given by a \(\varphi\)-module \(M\) of finite type over \(\mathcal E^{\mathrm{int}}\) (resp. \(\mathcal E\)) equipped with an additional action \(\tau: \mathcal E_\tau^{\mathrm{int}} \otimes_{\mathcal E^{\mathrm{int}}} M \to \mathcal E_\tau^{\mathrm{int}} \otimes_{\mathcal E^{\mathrm{int}}} M\) satisfying certain technical conditions. The first main result of this paper establishes an equivalence between the category of \(\mathbb Q_p\)-representations of finite dimension of \(G_K\) and the category of \((\varphi,\tau)\)-modules over \((\mathcal E,\mathcal E _\tau)\); analogous statements are available for finite type free or torsion \(\mathbb Z_p\)-representations. But one important difficulty lies in the technical complexity of the ring \(\mathcal E_\tau\), whose elements are power series in infinitely many variables requiring subtle convergence conditions. A greater control in terms of ramification can be exerted on torsion \(\mathbb Z_p\)-representations of \(G_K\) by introducing the notion of ``\((\varphi,\tau)\)-lattices of height dividing \(U\)'' (we don't recall the definitions here): if the restriction to \(G_\infty\) of such a representation \(T\) corresponds to a \(\varphi\)-module over \(\mathcal E^{\mathrm{int}}\) containing such a lattice, there exists a constant \(d(K)\) depending only on \(K\), \(U\) and the exponent of \(T\), such that the upper ramification group \(G_K^{(\mu)}\) acts trivially on \(T\) for \(\mu> d(K)\). Finally, let us restrict our interest to semi-stable representations. Breuil and Kisin have shown that they have ``finite \(E(u)\)-height'' (again, we don't recall the precise definition). The author proves the following partial converse: suppose that \(K\) contains maximally a primitive \(p^s\)-th root of unity ; then any representation of \(G_K\) which has finite \(E(u)\)-height becomes semi-stable when restricted to the subgroup \(\mathrm{Gal}(\bar{K}/K(\root{p^s}\of{\pi}))\).