# Inertial and Hodge–Tate weights of crystalline representations

- 84 Downloads

## Abstract

Let *K* be an unramified extension of \({\mathbb {Q}}_p\) and \(\rho :G_K \rightarrow {\text {GL}}_n(\overline{{\mathbb {Z}}}_p)\) a crystalline representation. If the Hodge–Tate weights of \(\rho \) differ by at most *p* then we show that these weights are contained in a natural collection of weights depending only on the restriction to inertia of \({\overline{\rho }} = \rho \otimes _{\overline{{\mathbb {Z}}}_p} \overline{{\mathbb {F}}}_p\). Our methods involve the study of a full subcategory of *p*-torsion Breuil–Kisin modules which we view as extending Fontaine–Laffaille theory to filtrations of length *p*.

## Mathematics Subject Classification

11S20 11F80## 1 Introduction

Let \(K/{\mathbb {Q}}_p\) be a finite unramified extension with residue field *k*. In this paper we show that if the Hodge–Tate weights of a crystalline representation \(\rho \) of \(G_K\) are sufficiently small then these weights are encoded in an explicit way by the reduction of \(\rho \) modulo *p*. Using Fontaine–Laffaille theory this is known for Hodge–Tate weights differing by at most \(p-1\); we will treat weights differing by at most *p*. Our techniques are local and involve the study of a full subcategory of *p*-torsion Breuil–Kisin modules, which we view as extending (*p*-torsion) Fontaine–Laffaille theory to filtrations of length *p*.

Throughout Hodge–Tate weights are normalised so that the cyclotomic character has weight \(-\,1\).

### Theorem 1.0.1

When \(n=2\) and \(p>2\) the result is a theorem of Gee–Liu–Savitt [9]. When \(n=2\) and \(p=2\) the result is due to Wang [14]. In this paper we extend their methods to higher dimensions.

As already mentioned, when \(\lambda _{n,\tau } - \lambda _{1,\tau } \le p-1\), Theorem 1.0.1 is a straightforward consequence of Fontaine–Laffaille theory, so the main content of our result is that it applies to Hodge–Tate weights differing by *p*. On the other hand Theorem 1.0.1 does not hold if the condition \(\lambda _{n,\tau } - \lambda _{1,\tau } \le p\) is relaxed. For example, there exist irreducible two dimensional crystalline representations \(\rho \) of \(G_{{\mathbb {Q}}_p}\), with Hodge–Tate weights \((-p-1,0)\), whose reduction modulo *p* have the form \({\overline{\rho }} = ({\begin{matrix} \chi _{{\text {cyc}}} &{} * \\ 0 &{} \chi _{{\text {cyc}}} \end{matrix}})\), see [3, Théorème 3.2.1]. Here \(\chi _{{\text {cyc}}}\) denotes the cyclotomic character. It is easy to check that \((-p-1,0)\) is not an element of \({\text {Inert}}({\overline{\rho }})\).

*p*representations associated to automorphic representations on unitary groups of rank

*n*. To be more precise let

*F*be an imaginary CM field in which

*p*is unramified and fix an isomorphism \(\iota :\overline{{\mathbb {Q}}}_p \cong {\mathbb {C}}\). Attached to any RACSDC (regular, algebraic, conjugate self dual, and cuspidal) automorphic representation \(\varPi \) of \({\text {GL}}_n({\mathbb {A}}_F)\) there is a continuous irreducible \(r_{\iota ,p}(\varPi ):G_F \rightarrow {\text {GL}}_n(\overline{{\mathbb {Q}}}_p)\), cf. the main result of [5]. If \(\varPi \) is unramified above

*p*then \(r_{\iota ,p}(\varPi )\) is crystalline above

*p*, and if \(\lambda = (\lambda _\kappa )_{\kappa } \in ({\mathbb {Z}}_+^n)^{{\text {Hom}}(F,{\mathbb {C}})}\) is the weight of \(\varPi \) then the \(\kappa \)-Hodge–Tate weights

^{1}of \(r_{\iota ,p}(\varPi )\) equal

### Corollary 1.0.2

*p*, has weight \(\lambda \), and is such that \({\overline{r}}_{\iota ,p}(\varPi ) \cong {\overline{r}}\). Then

We point out that, while Corollary 1.0.2 involves only distinct Hodge–Tate weights, due to the regularity assumptions on our automorphic representations, Theorem 1.0.1 does not require such distinctness.

If \({\overline{r}}\) is assumed to arise from some potentially diagonalisable RACSDC automorphic representation (a notion introduced in [2]) and if we assume \({\overline{r}}_v\) is semi-simple for each \(v \mid p\) then, under a Taylor–Wiles hypothesis, the inclusion in the Corollary 1.0.2 is an equality. This follows from [1, Theorem 3.1.3].

To conclude this introduction we briefly explain our proof of the theorem; let us do this by sketching the content of the various sections in this paper. In the first two sections we recall some basic notions; in Sect. 2 we define the set \({\text {Inert}}({\overline{\rho }})\) and in Sect. 3 we give some elementary results on filtered modules. In Sect. 4 we recall the notion of a Breuil–Kisin module, and recall how to associate to them Galois representations. Breuil–Kisin modules killed by *p* admit a natural set of weights and in Sect. 5 we define what it means for a *p*-torsion Breuil–Kisin module to be strongly divisible; it’s weights must be contained in [0, *p*] and a certain explicit condition on its \(\varphi \) must be satisfied. We view the category of strongly divisible Breuil–Kisin modules \({\text {Mod}}^{{\text {SD}}}_k\) as an extension of *p*-torsion Fontaine–Laffaille theory to filtrations of length *p*. We establish two important properties of \({\text {Mod}}^{{\text {SD}}}_k\). The first main property (Proposition 5.4.7) is shown in Sect. 5 and states that \({\text {Mod}}^{{\text {SD}}}_k\) is stable under subquotients, and that weights behave well along short exact sequences. The second main property (Proposition 6.7.1) is proved in Sect. 6 and concerns the structure of simple objects in \(M \in {\text {Mod}}^{{\text {SD}}}_k\). We show that for such *M* the weights of *M* coincide with the inertial weights of the associated Galois representation. These two properties mirror the situation for Fontaine–Laffaille theory. However, unlike in Fontaine–Laffaille theory, it is not the case that simple \(M \in {\text {Mod}}^{{\text {SD}}}_k\) are determined by their weights together with their associated Galois representation. This complicates the proofs considerably. Thus, while there are similarities between \({\text {Mod}}^{{\text {SD}}}_k\) and Fontaine–Laffaille theory in some respects, the former category is more complicated, reflecting the fact that the reduction of crystalline representations with Hodge–Tate weights in [0, *p*] is genuinely more subtle than for weights in the Fontaine–Laffaille range. In the final section we recall a theorem of Gee–Liu–Savitt [9] which relates \({\text {Mod}}^{{\text {SD}}}_k\) with the reduction modulo *p* of those crystalline representations with Hodge–Tate weights contained in [0, *p*]. Using this, and the two properties of \({\text {Mod}}^{{\text {SD}}}_k\) described above, it is straightforward to deduce Theorem 1.0.1.

This work was supported by the Engineering and Physical Sciences Research Council [EP/L015234/1] and the EPSRC Centre for Doctoral Training in Geometry and Number Theory (The London School of Geometry and Number Theory), University College London, and the Max Planck Institute for Mathematics, Bonn.

### 1.1 Notation

Throughout we let *k* denote a finite field of characteristic \(p> 0\) and write \(K_0 = W(k)[\frac{1}{p}]\). In the introduction we took \(K = K_0\); however some of our constructions are valid for arbitrary finite extensions so now allow *K* to denote a totally ramified extension of \(K_0\) of degree *e*, with ring of integers \({\mathcal {O}}_K\). At certain points it will be necessary to assume \(K = K_0\).

Let *C* denote the completion of an algebraic closure \({\overline{K}}\) of *K* and let \({\mathcal {O}}_C\) be its ring of integers, with residue field \({\overline{k}}\). We write \(G_K = {\text {Gal}}({\overline{K}}/K)\) and \(v_p\) for the valuation on *C* normalised so that \(v_p(p) =1\).

We fix a uniformiser \(\pi \in K\) and a compatible system \(\pi ^{1/p^n} \in {\overline{K}}\) of \(p^n\)th roots of \(\pi \). Many constructions in this paper depend upon these choices. Set \(K_\infty = K(\pi ^{1/p^\infty })\) and \(G_{K_\infty } = {\text {Gal}}({\overline{K}}/K_\infty )\).

Let \(E/{\mathbb {Q}}_p\) denote a finite extension with ring of integers \({\mathcal {O}}\) and residue field \({\mathbb {F}}\). We assume throughout that \(K_0 \subset E\). This will be our coefficient field in which the representations we consider will be valued.

If *A* is any ring of characteristic *p* we let \(\varphi :A \rightarrow A\) denote the homomorphism \(x \mapsto x^p\). If *A* is perfect (i.e. \(\varphi \) is an automorphism) we let *W*(*A*) denote the ring of Witt vectors of *A* and write \(\varphi : W(A) \rightarrow W(A)\) for the automorphism lifting \(\varphi \) on *A*.

## 2 Inertial weights

In this section we recall the structure of irreducible torsion representations of \(G_K\) and \(G_{K_\infty }\). We then define the set \({\text {Inert}}({\overline{\rho }})\) from the introduction.

### 2.1 Tame ramification

*K*respectively. Set \(I^{{\text {t}}} = {\text {Gal}}(K^{{\text {t}}}/K^{{\text {ur}}})\). As in [12, Proposition 2] there is an isomorphism

*l*runs over finite extensions of

*k*with transition maps given by the norm maps. This isomorphism sends \(\sigma \mapsto (s(\sigma )_{l})_{l}\) where \(s(\sigma )_{l}\) is the image in the residue field of \(K^{{\text {t}}}\) of the \({\text {Card}}(l^\times )\)th root of unity

*s*we define the fundamental character

### Lemma 2.1.1

Any continuous \(\chi :I^{{\text {t}}} \rightarrow \overline{{\mathbb {F}}}_p^\times \) extends to a continuous character of \({\text {Gal}}(K^{{\text {t}}}/K)\) if and only if there exist integers \((r_\tau )_{\tau \in {\text {Hom}}_{{\mathbb {F}}_p}(k,\overline{{\mathbb {F}}}_p)}\) such that \(\chi = \prod _{\tau } \omega _\tau ^{r_\tau }\).

### Proof

Since \(1 \rightarrow I^{{\text {t}}} \rightarrow {\text {Gal}}(K^{{\text {t}}}/K) \rightarrow G_k \rightarrow 1\) is split, \(\chi \) extends to \({\text {Gal}}(K^{{\text {t}}}/K)\) if and only if \(\chi \) is stable under the conjugation action of \(G_k\) on \(I^{{\text {t}}}\). Via *s* this action is given by the natural action of \(G_k\) on \(\varprojlim l^\times \), and so \(\chi \) extends if and only if \(\chi ^{p^{[k:{\mathbb {F}}_p]}} = \chi \). After [12, Proposition 5] this is equivalent to asking that \(\chi \) be a power of \(\omega _k\), thus a product as in the lemma. \(\square \)

In particular we see each \(\omega _{l}\) extends to a character of \(G_{L}\) where *L* / *K* is the unramified extension with residue field *l*. Such an extension is well defined only up to twisting by an unramified character. Our fixed choice of uniformiser \(\pi \in K\) allows us to define a canonical choice of extension by sending \(\sigma \in G_L\) onto the image in the residue field of the element \(\sigma (\pi ^{1/{\text {Card}}(l^\times )}) / \pi ^{1/{\text {Card}}(l^\times )} \in K^{{\text {t}}}\) where \(\pi ^{1/{\text {Card}}(l^\times )}\) is a \({\text {Card}}(l^\times )\)th root of \(\pi \). We shall denote this character again by \(\omega _l:G_L \rightarrow \overline{{\mathbb {F}}}_p^\times \). Also, for \(\theta \in {\text {Hom}}_{{\mathbb {F}}_p}(l,\overline{{\mathbb {F}}}_p)\), we write \(\omega _{\theta } = \theta \circ \omega _{l}\) as characters of \(G_L\).

For an extension *L* / *K* write \({\text {Ind}}_L^{K} V\) in place of \({\text {Ind}}_{{\text {Gal}}({\overline{K}}/L)}^{{\text {Gal}}({\overline{K}}/K)} V\).

### Lemma 2.1.2

If *V* is a continuous irreducible representation of \(G_K\) on a finite dimensional \(\overline{{\mathbb {F}}}_p\)-vector space then \(V \cong {\text {Ind}}_L^K \chi \), where *L* / *K* is an unramified extension of degree \({\text {dim}}_{{\mathbb {F}}} V\) and \(\chi :G_L \rightarrow \overline{{\mathbb {F}}}_p^\times \) is a continuous character.

### Proof

As *V* is irreducible the \(G_K\)-action factors through \(G = {\text {Gal}}(K^{{\text {t}}}/K)\) by [12, Proposition 4]. Since \(I^{{\text {t}}}\) is abelian of order prime to *p*, \(V|_{I^{{\text {t}}}}\) is a sum of \(\overline{{\mathbb {F}}}_p^\times \)-valued characters. If \(\gamma \in G_k\) and \(\chi :I^{{\text {t}}} \rightarrow \overline{{\mathbb {F}}}_p^\times \) is a character define a new character by \(\chi ^{(\gamma )}(\sigma ) = \chi (\gamma ^{-1}\sigma \gamma )\). If \(I^{{\text {t}}}\) acts on \(v \in V|_{I^{{\text {t}}}}\) by \(\chi \) then \(I^{{\text {t}}}\) acts on \(\gamma (v)\) by \(\chi ^{(\gamma )}\); thus \(G_k\) acts on the set of \(\chi \) appearing in \(V|_{I^{{\text {t}}}}\). Fix \(\chi \) appearing in \(V|_{I^{{\text {t}}}}\) and let \(H \subset G\) be the normal subgroup containing \(I^{{\text {t}}}\) and corresponding to the stabiliser of \(\chi \) in \(G_k\). By the orbit-stabiliser theorem \([G:H] \le {\text {dim}}_{\overline{{\mathbb {F}}}_p} V\).

Frobenius reciprocity gives a non-zero map \(V|_H \rightarrow {\text {Ind}}_{I^{{\text {t}}}}^H \chi \). If *L* / *K* is the unramified extension corresponding to *H* then, since the image of *H* in \(G_k\) stabilises \(\chi \), this character can be extended to *H* as in Lemma 2.1.1. Thus \({\text {Ind}}_{I^{{\text {t}}}}^H \chi = \chi \otimes {\text {Ind}}_{I^{{\text {t}}}}^H \mathbb {1}\). Since \({\text {Ind}}_{I^{{\text {t}}}}^H \mathbb {1}\) is a discrete *H*-module we can find a finite dimensional sub-representation \(R \subset {\text {Ind}}_{I^{{\text {t}}}}^H \mathbb {1}\) so that \(V|_H\) is mapped into \(\chi \otimes R\). As \({\text {Gal}}(L^{{\text {ur}}}/L)\) is abelian *R* admits a composition series \(0 = R_n \subset \cdots \subset R_0 = R\) such that each \(R_i/R_{i+1}\) is one-dimensional. If *i* is the largest integer such that \(V|_{H} \rightarrow {\text {Ind}}^H_{I^{{\text {t}}}} V\) factors through \(\chi \otimes R_i\) then \(V|_H \rightarrow \chi \otimes R_i/R_{i+1}\) is non-zero. Frobenius reciprocity gives a non-zero map \(V \rightarrow {\text {Ind}}_L^K (\chi \otimes R_i/R_{i+1})\) which, *V* being irreducible, is injective. Thus \([G:H] = {\text {dim}}_{\overline{{\mathbb {F}}}_p} {\text {Ind}}_L^K(\chi \otimes R_i/R_{i+1})\) is \(\ge {\text {dim}}_{\overline{{\mathbb {F}}}_p} V\). The inequality of the first paragraph implies \([G:H] = {\text {dim}}_{\overline{{\mathbb {F}}}_p} V\) and so this map is an isomorphism. \(\square \)

### Definition 2.1.3

*n*-dimensional \(\overline{{\mathbb {F}}}_p\)-vector space. After Lemma 2.1.2 there exist continuous characters \(\zeta :G_{L_\zeta } \rightarrow \overline{{\mathbb {F}}}_p^\times \), with \(L_\zeta / K\) finite unramified, such that

We remark that for a given \(\zeta \) there always exists a unique tuple \((r_{\theta ,\zeta })_{\theta \in {\text {Hom}}_{{\mathbb {F}}_p}(l_\zeta ,\overline{{\mathbb {F}}}_p)}\) as above such that each \(r_{\theta ,\zeta } \in [0,p-1]\) with not all \(r_{\theta ,\zeta }\) equal to \(p-1\). However if we drop the restriction that \(r_{\theta ,\zeta } \in [0,p-1]\) then there will be many different such tuples.

It is easy to check that \({\text {Inert}}({\overline{\rho }})\) depends only on \({\overline{\rho }}^{{\text {ss}}}|_{I^{{\text {t}}}}\).

### 2.2 \(G_{K_\infty }\)-representations

### Lemma 2.2.1

Let \(K_\infty ^{{\text {t}}} = K_\infty K^{{\text {t}}}\). Then restriction defines an isomorphism \({\text {Gal}}(K_\infty ^{{\text {t}}}/K_\infty ) \rightarrow {\text {Gal}}(K^{{\text {t}}}/K)\). If *L* / *K* is a tamely ramified extension this isomorphism identifies \({\text {Gal}}(L_\infty /K_\infty )\) with \({\text {Gal}}(L/K)\) where \(L_\infty = L K_\infty \).

### Proof

Since \(K_\infty /K\) is totally wildly ramified we have \(K_\infty \cap K^{{\text {t}}} = K\). The lemma then follows from Galois theory. \(\square \)

### Corollary 2.2.2

If *V* is as in Lemma 2.1.2 then \(V|_{G_{K_\infty }} \cong {\text {Ind}}_{L_\infty }^{K_\infty } \chi |_{G_{L_\infty }}\) where \(L_\infty = LK_\infty \).

## 3 Filtrations

*A*and a collection of ideals \((F^iA)_{i \in {\mathbb {Z}}}\) satisfying

*A*-modules consists of

*A*-modules

*M*equipped with a collection of

*A*-sub-modules \((F^i M)_{i \in {\mathbb {Z}}}\) satisfying

*A*-modules such that \(f(F^iM) \subset F^iN\) for all

*i*. If

*M*is an object of \({\text {Fil}}(A)\) we set \({\text {gr}}(M) = \bigoplus _i {\text {gr}}^i (M)\) where \({\text {gr}}^i (M) = F^i M / F^{i+1}M\). The module \({\text {gr}}(A)\) admits an obvious structure of a ring and each \({\text {gr}}(M)\) admits the structure of a module over \({\text {gr}}(A)\).

### 3.1 Strict maps

If *M* is an object of \({\text {Fil}}(A)\) and \(N \subset M\) is an *A*-sub-module the induced filtration on *N* is that given by \(F^i N = N \cap F^i M\). If \(f:M \rightarrow N\) is a surjective *A*-module homomorphism the quotient filtration on *N* is that given by \(F^i N = f(F^i M)\).

### Remark 3.1.1

*N*and

*M*respectively.

### Definition 3.1.2

A morphism \(f: M \rightarrow N\) in \({\text {Fil}}(A)\) is strict if \(F^i N \cap f(M) = f(F^i M)\) for all \(i \in {\mathbb {Z}}\). Equivalently *f* is strict if \({\text {coim}}(f) \rightarrow {\text {im}}f\) is an isomorphism in \({\text {Fil}}(A)\).

### Notation 3.1.3

*A*induces the structure of a topological ring on

*A*; the \(F^i A\) form a basis of open neighbourhoods of zero. Similarly the filtration on an object

*M*of \({\text {Fil}}(A)\) gives

*M*the structure of a topological

*A*-module. Then

*M*is discrete if and only if \(F^i M = 0\) for \(i>>0\);*M*is Hausdorff if and only if \(\cap F^iM = 0\);*M*is complete if and only if the natural map \(M \rightarrow \varprojlim M/F^iM\) is an isomorphism.

### Lemma 3.1.4

*A*-modules.

- 1.
Then

*f*is an isomorphism in \({\text {Fil}}(A)\) if and only if \({\text {gr}}^i (M)\rightarrow {\text {gr}}^i (N)\) is injective for all*i*. - 2.
If

*M*is complete and*N*Hausdorff then*f*is an isomorphism in \({\text {Fil}}(A)\) if and only if \({\text {gr}}^i (M) \rightarrow {\text {gr}}^i (N)\) is surjective for all*i*.

### Proof

*A*-modules the leftmost and central vertical arrows are injective. For (1) use the snake lemma to obtain an exact sequence \(0 \rightarrow {\text {ker}}\, c \rightarrow {\text {coker}}(a) \rightarrow {\text {coker}}(b) \rightarrow {\text {coker}}(c)\). One proves \(F^i M \rightarrow F^i N\) is surjective by increasing induction on

*i*; using as the base case the fact that \(F^i M \rightarrow F^iN\) is surjective for \(i<<0\), since \(F^iM = M\) for \(i<<0\). For (2) argue as in [13, Proposition 6]. \(\square \)

### Lemma 3.1.5

- 1.
*f*is strict; - 2.
\({\text {gr}}\, ({\text {ker}}(f)) \rightarrow {\text {gr}}(M) \rightarrow {\text {gr}}\, (N)\) is exact;

- 3.
\(0 \rightarrow {\text {gr}}\, ({\text {ker}}(f)) \rightarrow {\text {gr}}(M) \rightarrow {\text {gr}}\, (N) \rightarrow {\text {gr}}({\text {coker}}(f)) \rightarrow 0\) is exact.

*M*is complete and

*N*is Hausdorff then the same is true with (2) replaced by

- \((2')\)
\( {\text {gr}}\, (M)\rightarrow {\text {gr}}\, (N) \rightarrow {\text {gr}}({\text {coker}}(f))\) is exact for all

*i*;

### Proof

It is straightforward to check (without any conditions on *M* and *N*) that (2) is equivalent to \({\text {gr}}^i{\text {coim}}(f) \rightarrow {\text {gr}}^i {\text {im}}(f)\) being injective for all *i*, that (\(2'\)) is equivalent to this map being surjective for all *i*, and that (3) is equivalent to this map being an isomorphism for all *i*. Thus \((1) \Leftrightarrow (2) \Leftrightarrow (3)\) follows from (1) of Lemma 3.1.4 applied to the morphism \({\text {coim}}(f) \rightarrow {\text {im}}(f)\). Similarly \((1) \Leftrightarrow (2') \Leftrightarrow (3)\) follows from (2) of Lemma 3.1.4, noting that *M* being complete implies \({\text {coim}}(f)\) is complete and *N* being Hausdorff implies \({\text {im}}(f)\) is Hausdorff. \(\square \)

### Corollary 3.1.6

*M*be a Hausdorff object of \({\text {Fil}}(A)\) with

*A*complete. Suppose \((m_j)\) is a finite collection of elements of

*M*and suppose that there are integers \(r_j\) such that \(m_j \in F^{r_j} M\). Let \({\overline{m}}_j\) denote the image of \(m_j\) in \({\text {gr}}^{r_j}(M)\). If the \({\overline{m}}_j\) generate \({\text {gr}}(M)\) over \({\text {gr}}(A)\) then

*M*is complete and the \(m_j\) generate

*M*. Further

*A*-basis of

*M*.

### 3.2 Adapted bases

We now put ourselves in the following situation. Let \(a \in A\) be a nonzerodivisor and equip *A* with the *a*-adic filtration (so \(F^iA = a^iA\)). Let *M* be a finite free *A*-module and let \(N \subset M[\frac{1}{a}]\) be a finitely generated *A*-sub-module with \(N[\frac{1}{a}] = M[\frac{1}{a}]\). Make *N* into an object of \({\text {Fil}}(A)\) by setting \(F^i N = a^iM \cap N\).

### Lemma 3.2.1

Suppose that *A* is complete. Give *N* / *a* the quotient filtration and suppose that a finite collection \((g_i)\) of elements of *N* is given, along with integers \((r_i)\), such that \(g_i \in F^{r_i} N\). If the images of \(g_i\) in \({\text {gr}}^{r_i}(N/a)\) form a \({\text {gr}}(A/a) = A/a\)-basis of \({\text {gr}}(N/a)\) then the \((g_i)\) form a basis of *N* and the \((a^{-r_i}g_i)\) form a basis of *M*.

### Proof

*aN*of \(N \rightarrow N/a\) is given by \(F^i(aN) = aN \cap F^iN = aF^{i-1} N\) (because

*a*is not a zerodivisor). Lemma 3.1.5 implies there is an exact sequence

*N*is finitely generated

*N*is Hausdorf and so we may apply Corollary 3.1.6 to deduce that the \((g_i)\) form an

*A*-basis of

*N*and that

*A*-linearly independent the \((a^{-r_i}g_i)\) are

*A*-linearly independent. To show they generate

*M*take \(m \in M\) and

*n*large enough that \(a^nm \in N\). Then \(a^nm \in F^nN\) and so \(a^nm = \sum a_i g_i\) with \(a_i \in F^{n-r_i}A\). It follows that \(m = \sum (a^{r_i-n}a_i) (a^{-r_i}g_i)\) and so, since \((a^{r_i-n}) F^{n-r_i}A \subset A\), we are done. \(\square \)

### 3.3 Filtered vector spaces

Finally we give criteria to determine when two filtrations on a vector space are the same.

### Lemma 3.3.1

*V*be an

*k*-vector space equipped with two discrete filtrations \(G^iV \subset F^i V\). Then

### Proof

*F*is replaced by the filtration

*G*. As \(G^i V \subset F^i V\), \({\text {dim}}_k G^i V \le {\text {dim}}_k F^iV\). The desired inequality follows. This inequality is an equality if and only if \({\text {dim}}_k G^iV = {\text {dim}}_k F^iV\) for all

*i*, i.e. if and only if \(G = F\). \(\square \)

### Notation 3.3.2

Say that a sequence of morphisms \(M \rightarrow N \rightarrow P\) in \({\text {Fil}}(A)\) is exact if it is exact as a sequence of *A*-modules and if \(M \rightarrow N\) is strict. Lemma 3.1.5 implies that a sequence \(0 \rightarrow M \rightarrow N \rightarrow P \rightarrow 0\) in \({\text {Fil}}(A)\) which is exact in the category of *A*-modules is exact in \({\text {Fil}}(A)\) if and only if \(0 \rightarrow {\text {gr}}(M) \rightarrow {\text {gr}}(N) \rightarrow {\text {gr}}(P) \rightarrow 0\) is an exact sequence of *A*-modules.

### Corollary 3.3.3

*k*-vector spaces. If

*f*(respectively

*g*) is strict then

*f*or

*g*is strict then equality implies the sequence is exact in \({\text {Fil}}(k)\).

### Proof

*P*is discrete we can apply Lemma 3.3.1 to deduce that

*g*is strict. If

*f*is strict Lemma 3.1.5 tells us that \(0 \rightarrow {\text {gr}}(M) \rightarrow {\text {gr}}(N) \rightarrow {\text {gr}}({\text {coker}}(f)) \rightarrow 0\) is exact, and so

*f*is strict. If

*g*is strict one argues similarly, applying Lemma 3.3.1 to the map \(M \rightarrow {\text {ker}}(g)\). \(\square \)

## 4 Breuil–Kisin modules

### 4.1 Etale \(\varphi \)-modules

First we recall the description of \(G_{K_\infty }\)-representations given by etale \(\varphi \)-modules.

### Definition 4.1.1

*p*. There is a multiplicative identification \({\mathcal {O}}_{C^\flat } = \varprojlim {\mathcal {O}}_C\) (the limit again taken with respect to the transition maps \(x \mapsto x^p\)) given by

### Notation 4.1.2

Let \({\mathfrak {S}} = W(k)[[u]]\) and \(A_{{\text {inf}}} = W({\mathcal {O}}_{C^\flat })\). Both rings are equipped with a \({\mathbb {Z}}_p\)-linear endomorphism \(\varphi \); on \(A_{{\text {inf}}}\) this is the usual Witt vector Frobenius and on \({\mathfrak {S}}\) it is given by \(\sum a_iu^i \mapsto \sum \varphi (a_i)u^{ip}\). The system \(\pi ^{1/p^n}\), fixed in Sect. 1.1, defines an element \(\pi ^\flat = (\pi ,\pi ^{1/p},\ldots ) \in {\mathcal {O}}_{C^\flat }\) and we embed \({\mathfrak {S}} \rightarrow A_{{\text {inf}}}\) by mapping \(u \mapsto [\pi ^\flat ]\) (where \([\cdot ]\) denotes the Teichmuller lifting). This embedding is compatible with \(\varphi \). Let \({\mathcal {O}}_{{\mathcal {E}}}\) denote the *p*-adic completion of \({\mathfrak {S}}[\frac{1}{u}]\). Then \(\varphi \) on \({\mathfrak {S}}\) extends to \({\mathcal {O}}_{{\mathcal {E}}}\) and the embedding \({\mathfrak {S}} \rightarrow A_{{\text {inf}}}\) extends to a \(\varphi \)-equivariant embedding \({\mathcal {O}}_{{\mathcal {E}}} \rightarrow W(C^\flat )\).

By functoriality there are \(\varphi \)-equivariant \(G_K\)-actions on \(A_{{\text {inf}}} = W({\mathcal {O}}_{C^\flat })\) and \(W(C^\flat )\) lifting those modulo *p*.

### Definition 4.1.3

### Construction 4.1.4

### Proposition 4.1.5

### Proof

*p*to an inclusion of

*k*((

*u*)) in \(C^\flat \). The completion of \(K_\infty \) is a perfectoid field in the sense of [11], whose tilt is the completed perfection of \(k((u)) \subset C^\flat \). It follows from [11, Theorem 3.7] that the action of \(G_{K_\infty }\) on \(C^\flat \) identifies \(G_K = G_{k((u))}\). Let \({\mathcal {O}}_{\widehat{{\mathcal {E}}^{{\text {ur}}}}}\) be the

*p*-adic completion of the Cohen ring (i.e. the discrete valuation ring of characteristic zero with uniformizer

*p*) with residue field \(k((u))^{{\text {sep}}}\). Then \({\mathcal {O}}_{\widehat{{\mathcal {E}}^{{\text {ur}}}}}\) may be identified as a subring of \(W(C^\flat )\) stable under the action of \(G_{K_\infty }\) and \(\varphi \). The proposition with \(T(M^{{\text {et}}})\) replaced by \(T'(M^{{\text {et}}}) := (M^{{\text {et}}} \otimes _{{\mathcal {O}}_{{\mathcal {E}}}} {\mathcal {O}}_{\widehat{{\mathcal {E}}^{{\text {ur}}}}})^{\varphi =1}\) follows from [8, Proposition 1.2.6] applied with \(E = k((u))\). It therefore suffices to show the inclusion \(T'(M^{{\text {et}}}) \subset T(M^{{\text {et}}})\) is an equality. Since we know there are \(\varphi \)-equivariant identifications

### 4.2 Breuil–Kisin modules

Breuil–Kisin modules appear as special \({\mathfrak {S}}\)-lattices inside etale \(\varphi \)-modules.

### Definition 4.2.1

*M*equipped with an isomorphism

### Notation 4.2.2

### Construction 4.2.3

*E*(

*u*) is a unit in \({\mathcal {O}}_{{\mathcal {E}}}\). Thus, if \(M \in {\text {Mod}}^{{\text {BK}}}_K\) then \(M \otimes _{{\mathfrak {S}}} {\mathcal {O}}_{{\mathcal {E}}}\) is an etale \(\varphi \)-module and

### Remark 4.2.4

Kisin [10, Proposition 2.1.12] has shown \(M \mapsto T(M)\) is fully faithful when restricted to Breuil–Kisin modules which are free over \({\mathfrak {S}}\). However if one does not restrict to Breuil–Kisin modules which are free over \({\mathfrak {S}}\) then this is not true.

### Construction 4.2.5

### 4.3 Coefficients

In practice we are interested in representations valued in extensions of \({\mathbb {Z}}_p\). For this reason we introduce a variant of \({\text {Mod}}^{{\text {BK}}}_K\).

### Definition 4.3.1

*M*equipped with an isomorphism

### Remark 4.3.2

By functoriality \(M \mapsto T(M)\) induces an exact functor from \({\text {Mod}}^{{\text {BK}}}_K({\mathcal {O}})\) into the category of continuous representations of \(G_{K_\infty }\) on finitely generated \({\mathcal {O}}\)-modules.

### Construction 4.3.3

### Construction 4.3.4

### Corollary 4.3.6

- 1.
If \(M \in {\text {Mod}}^{{\text {BK}}}_K({\mathcal {O}})\) is free as an \({\mathfrak {S}}\)-module then it is free as an \({\mathfrak {S}} \otimes _{{\mathbb {Z}}_p} {\mathcal {O}}\)-module.

- 2.
Let \(\varpi \in {\mathcal {O}}\) be a uniformiser and suppose \(M \in {\text {Mod}}^{{\text {BK}}}_K({\mathcal {O}})\) is \(\varpi \)-torsion. If

*M*is free as an \({\mathfrak {S}} / p = k[[u]]\)-module then it is free as a module over \(k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\).

### Proof

If *M* is free over \({\mathfrak {S}}\) then each \(M_\tau \) is free over \({\mathcal {O}}[[u]]\). By (4.3.5) the rank of \(M_\tau \) over \({\mathcal {O}}[[u]]\) does not depend on \(\tau \) so \(M = \prod _\tau M_\tau \) is free over \({\mathfrak {S}} \otimes _{{\mathbb {Z}}_p} {\mathcal {O}}\). (2) follows similarly. \(\square \)

## 5 Strongly divisibility

### 5.1 Torsion Breuil–Kisin modules

### Definition 5.1.1

Denote by \({\text {Mod}}^{{\text {BK}}}_{k} \subset {\text {Mod}}^{{\text {BK}}}_K\) the full subcategory whose objects are modules which are free over \({\mathfrak {S}}/ p = k[[u]]\).

### Remark 5.1.2

An \(M \in {\text {Mod}}^{{\text {BK}}}_k\) is the same thing as a *k*[[*u*]]-lattice inside an etale \(\varphi \)-module over \({\mathcal {O}}_{{\mathcal {E}}}/p = k((u))\) because \(E(u) \equiv u^e\) modulo *p*. In particular, there are many *p*-torsion Breuil–Kisin modules giving rise to the same etale \(\varphi \)-module. This is in contrast to the integral situation described in Remark 4.2.4.

### Lemma 5.1.3

### Proof

If *T* is an \({\mathbb {F}}_p\)-representation of \(G_{K_\infty }\) then there exists a *p*-torsion \(M^{{\text {et}}} \in {\text {Mod}}^{{\text {et}}}_K\) such that \(T(M^{{\text {et}}}) = T\). Remark 5.1.2 shows that any *k*[[*u*]]-lattice \(M \subset M^{{\text {et}}}\) is an object of \({\text {Mod}}^{{\text {BK}}}_k\) with \(T(M) = T\).

For the second part, the functor from Proposition 4.1.5 is an exact equivalence and so there exists an exact sequence \(0 \rightarrow M_1^{{\text {et}}} \rightarrow M^{{\text {et}}} \rightarrow M_2^{{\text {et}}} \rightarrow 0\) recovering \(0 \rightarrow T_1 \rightarrow T \rightarrow T_2 \rightarrow 0\) after applying \(T(-)\). Then *M* is a *k*[[*u*]]-lattice inside \(M^{{\text {et}}}\) and we must have \(M_2 = {\text {Im}}(M) \subset M^{{\text {et}}}_2\) and \(M_1 = M \cap M_1^{{\text {et}}}\). \(\square \)

### Construction 5.1.4

*M*is a filtration

*k*[[

*u*]]-torsion-free) of \({\text {Mod}}^{{\text {BK}}}_k\). Lemma 5.1.3 implies being irreducible is equivalent to asking that \(T(M_i/M_{i+1})\) is an irreducible \(G_{K_\infty }\)-representation. Lemma 5.1.3 also implies that composition series for

*M*are in bijection with composition series for

*T*(

*M*).

### Warning 5.1.5

*M*. On the other hand if \(x +1 = \alpha \) then

*u*.

### 5.2 Strong divisibility

In this subsection we define a full-subcategory \({\text {Mod}}^{{\text {SD}}}_k \subset {\text {Mod}}^{{\text {BK}}}_k\) which we view as an extension of *p*-torsion Fontaine–Laffaille theory to filtrations of length *p*.

### Construction 5.2.1

Let *M* be an object of \({\text {Mod}}^{{\text {BK}}}_k\). Recall \(M^\varphi \) is the *k*[[*u*]]-sub-module of \(M[\frac{1}{u}]\) generated by \(\varphi (M)\). Equip \(M^\varphi \) with the filtration \(F^i M^\varphi = M^\varphi \cap u^iM\). Let \(M^\varphi _k = M^\varphi /u\). We equip this *k*-vector space with the quotient filtration.

### Definition 5.2.2

*i*with multiplicity

### Construction 5.2.3

*M*with a filtration by setting \(F^iM = \lbrace m \in M \mid \varphi (m) \in u^iM \rbrace \). The semilinear injection

*k*-vector space with the quotient filtration.

### Lemma 5.2.4

*k*-semilinear bijection of vector spaces

### Proof

All that needs to be checked is that \(\varphi :M \rightarrow M^\varphi \) induces a *k*-semilinear bijection \(M_k \rightarrow M_{k}^\varphi \). As \(M_k\) and \(M^\varphi _k\) have the same dimension over *k* we only need to check surjectivity. As \(M^\varphi \) is the *k*[[*u*]]-module generated by \(\varphi (M) \subset M[\frac{1}{u}]\) surjectivity follows because \(\varphi \) is an automorphism on \(k = k[[u]]/u\). \(\square \)

### Lemma 5.2.5

*M*be an object of \({\text {Mod}}^{{\text {BK}}}_k\). The following are equivalent:

- 1.
The map \(M_k \rightarrow M_k^\varphi \) is an isomorphism of filtered modules.

- 2.
There exists a

*k*[[*u*]]-basis \((f_i)\) of*M*and integers \((r_i)\) such that \((u^{r_i}f_i)\) is a \(k[[u^p]]\)-basis of \(\varphi (M)\).

### Proof

Suppose \(M_k \rightarrow M^\varphi _k\) is an isomorphism of filtered modules. We can find integers \(r_i\) and elements \(g_i \in F^{r_i} M\) whose images in \({\text {gr}}(M_k)\) form a *k*-basis. As the induced map \({\text {gr}}(M_k) \rightarrow {\text {gr}}(M^\varphi _k)\) is an isomorphism it follows that the images of the \(\varphi (g_i) \in \varphi (M)\) in \({\text {gr}}(M_k^\varphi )\) form a *k*-basis. Applying Lemma 3.2.1 with \(M = M\), \(N = M^\varphi \) and \(a \in A\) equal to \(u \in k[[u]]\) proves that (1) implies (2) with \(f_i = u^{-r_i}\varphi (g_i)\).

*k*[[

*u*]] every \(m \in M^\varphi \) can be written as \(m = \sum \alpha _i (u^{r_i}f_i)\) with \(\alpha _i \in k[[u]]\). If \(m \in F^jM^\varphi \) then \(\alpha _i \in u^{{\text {max}} \lbrace j-r_i,0 \rbrace }k[[u]] = F^{j-r_i}k[[u]]\) since the \(f_i\) form a basis of

*M*. Hence

### Remark 5.2.6

Note that if we have a basis as in (2) of Lemma 5.2.5 then the above proof shows that \({\text {gr}}^j(M_k^\varphi ) = \sum _{r_i = j} k {\overline{f}}_i\). Thus the multiset \(\lbrace r_i \rbrace \) is equal to \({\text {Weight}}(M)\).

### Remark 5.2.7

*k*((

*u*)) can be written as \(C_1 \varLambda C_2\) where \(\varLambda = {\text {diag}}(u^{r_i})\) and \(C_i \in {\text {GL}}_n(k[[u]])\).

If

*M*is an object of \({\text {Mod}}^{{\text {BK}}}_k\) corresponding under (5.2.8) to a \(\varphi \)-conjugacy class represented by \(C_1\varLambda C_2\) then the \((r_i) = {\text {Weight}}(M)\).The isomorphism classes of Breuil–Kisin modules satisfying the equivalent conditions of Lemma 5.2.5 identify, via (5.2.8), with \(\varphi \)-conjugacy classes represented by matrices \(C_1\varLambda \) with \(C_1 \in {\text {GL}}_n(k[[u]])\) and \(\varLambda = {\text {diag}}(u^{r_i})\). Indeed, if \((f_i)\) is a

*k*[[*u*]]-basis as in Lemma 5.2.5(2) then there exists \(C \in {\text {GL}}_n(k[[u]])\) so that \((u^{r_1}f_1,\ldots ,u^{r_n}f_n) = (\varphi (f_1),\ldots ,\varphi (f_n)) \varphi (C)\) and so \(\varphi ((f_1,\ldots ,f_n)C) = (f_1,\ldots ,f_n)C C^{-1}{\text {diag}}(u^{r_i})\).

### Definition 5.2.9

Let \({\text {Mod}}^{{\text {SD}}}_k \subset {\text {Mod}}^{{\text {BK}}}_k\) denote the full subcategory whose objects satisfy the equivalent conditions of Lemma 5.2.5 and have \({\text {Weight}}(M) \subset [0,p]\). We say such *M* are strongly divisible.

### 5.3 Strong divisibility with coefficients

We reproduce the previous subsection allowing \({\mathcal {O}}\)-coefficients.

### Definition 5.3.1

Let \({\text {Mod}}^{{\text {BK}}}_k({\mathcal {O}})\) denote the full subcategory of \({\text {Mod}}^{{\text {BK}}}_K({\mathcal {O}})\) whose objects are finite free over \(k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\). This is equivalent to being free over *k*[[*u*]] and killed by \(\varpi \) by Corollary 4.3.6.

### Remark 5.3.2

*M*is by \(k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\)-sub-modules this is a decomposition of filtered modules. Thus \(M_k = \prod _{\tau } M_{k,\tau }\) as filtered modules (each \(M_{k,\tau }\) being a filtered \({\mathbb {F}}\)-vector space). Analogous statements hold for \(M^\varphi \) and \(M^\varphi _k\).

### Definition 5.3.3

*i*with multiplicity equal to

The following is a version of Lemma 5.2.5 for objects of \({\text {Mod}}^{{\text {BK}}}_k({\mathcal {O}})\) and is proved in exactly the same fashion.

### Lemma 5.3.4

*M*be an object of \({\text {Mod}}^{{\text {BK}}}_k({\mathcal {O}})\). Then the following are equivalent:

- 1.
The semilinear map \(M_k \rightarrow M^\varphi _k\) is an isomorphism of filtered modules.

- 2.
For \(\tau \in {\text {Hom}}_{{\mathbb {F}}_p}(k,{\mathbb {F}})\) there exists an \({\mathbb {F}}[[u]]\)-basis \((f_i)\) of \(M_\tau \) and integers \((r_i)\) such that \((u^{r_i}f_i)\) is an \({\mathbb {F}}[[u^p]]\)-basis of \(\varphi (M)_\tau \).

### Remark 5.3.5

As in Remark 5.2.6, if bases as in (2) of Lemma 5.3.4 exist then the multiset \(\lbrace r_{i,\tau } \rbrace \) equals \({\text {Weight}}_\tau (M)\).

### Remark 5.3.6

*f*-tuples of matrices satisfy \((A_\tau ) \sim (B_\tau )\) if there exist \(C_{\tau } \in {\text {GL}}_n({\mathbb {F}}[[u]])\) such that \(A_\tau = C_\tau ^{-1}B_{\tau }\varphi (C_{\tau \circ \varphi })\) for all \(\tau \). Each \(A_{\tau }\) can be written as \(C_{\tau }\varLambda _\tau C'_\tau \) with \(C_\tau ,C_{\tau }' \in {\text {GL}}_n({\mathbb {F}}[[u]])\) and \(\varLambda _\tau = {\text {diag}}(u^{r_{i,\tau }})\).

The multiset \(\lbrace r_{i,\tau } \rbrace \) is the multiset \({\text {Weight}}_\tau (M)\).

The

*M*which satisfy Lemma 5.3.4 correspond to classes represented by an*f*-tuple of matrices \((A_\tau )\) such that each \(A_{\tau } = C_{\tau } \varLambda _\tau \).

### Definition 5.3.7

Let \({\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}}) \subset {\text {Mod}}^{{\text {BK}}}_k({\mathcal {O}})\) denote the full subcategory whose objects are strongly divisible when viewed as objects of \({\text {Mod}}^{{\text {BK}}}_k\).

### 5.4 Subquotients

We now show \({\text {Mod}}^{{\text {SD}}}_k\) and \({\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) are closed under subquotients.

### Remark 5.4.1

### Lemma 5.4.2

- 1.
The map \(N \rightarrow P\) is strict when viewed as a map of filtered modules if and only if \(0 \rightarrow M_k \rightarrow N_k \rightarrow P_k \rightarrow 0\) is an exact sequence in \({\text {Fil}}(k)\) in the sense of Notation 3.3.2.

- 2.
The map \(N^\varphi \rightarrow P^\varphi \) is strict if and only if \(0 \rightarrow M_k^\varphi \rightarrow N_k^\varphi \rightarrow P_k^\varphi \rightarrow 0\) is exact in \({\text {Fil}}(k)\)

- 3.
Statement (2) is equivalent to \(M_k^\varphi \rightarrow N_k^\varphi \) being strict, which is equivalent to \(N_k^\varphi \rightarrow P_k^\varphi \) being strict.

### Proof

Note that \(M \rightarrow N\) is strict as a map of filtered modules. To see this suppose \(m \in M \cap F^i N\). Then \(\varphi (m) \in \varphi (M) \cap u^iN \subset M[\frac{1}{u}] \cap u^iN\). Since \(M \rightarrow N\) has *u*-torsion-free cokernel \(M[\frac{1}{u}] \cap u^iN = u^iM\). Thus \(m \in F^iM\). Similarly \(M^\varphi \rightarrow N^\varphi \) is strict. Hence, Lemma 3.1.5 implies \(N \rightarrow P\) is strict if and only if \(0 \rightarrow {\text {gr}}^i(M) \rightarrow {\text {gr}}^i(N) \rightarrow {\text {gr}}^i(P) \rightarrow 0\) is exact for each *i*. Likewise, \(N^\varphi \rightarrow P^\varphi \) is strict if and only if \(0 \rightarrow {\text {gr}}^i(M^\varphi ) \rightarrow {\text {gr}}^i(N^\varphi ) \rightarrow {\text {gr}}^i(P^\varphi ) \rightarrow 0\) is exact.

*i*(for small enough

*i*the left column will be zero). This proves (1). The same argument, but with the diagram replaced with the diagram obtained by considering the first exact sequence of Remark 5.4.1, proves (2) also.

It remains to show that if \(M_k^\varphi \rightarrow N_k^\varphi \) or \(N_k^\varphi \rightarrow P_k^\varphi \) is strict then \(0 \rightarrow M_k^\varphi \rightarrow N_k^\varphi \rightarrow P_k^\varphi \rightarrow 0\) is exact. It suffices to show that \(\sum _{i \in {\text {Weight}}(M)} i + \sum _{i \in {\text {Weight}}(P)} i = \sum _{i \in {\text {Weight}}(N)} i\) after Corollary 3.3.3. Remark 5.2.7 says that \(\sum _{i \in {\text {Weight}}(M)} i\) equals the *u*-adic valuation of the determinant of \(\varphi :M \rightarrow M[\frac{1}{u}]\) (in any choice of basis). Since this is clearly additive on exact sequences the lemma follows. \(\square \)

### Lemma 5.4.3

Let \(0 \rightarrow M \rightarrow N \rightarrow P \rightarrow 0\) be an exact sequence in \({\text {Mod}}^{{\text {BK}}}_k\). Suppose *M* and *P* satisfy the equivalent conditions of Lemma 5.2.5. If \(N \rightarrow P\) is strict then *N* satisfies the equivalent conditions of Lemma 5.2.5 also.

### Proof

### Lemma 5.4.4

Let \(0 \rightarrow M \rightarrow N \rightarrow P \rightarrow 0\) be an exact sequence in \({\text {Mod}}^{{\text {BK}}}_k\). Suppose that *N* satisfies the equivalent conditions of Lemma 5.2.5 and that \(M_k \rightarrow N_k\) is strict. Then \(N \rightarrow P\) is strict and *M* and *P* also satisfy the equivalent conditions of Lemma 5.2.5.

### Proof

*k*-vector spaces the horizontal arrows are injective and the vertical arrows are isomorphisms. By assumption the maps \(M_k \rightarrow N_k\) and \(N_k \rightarrow N^\varphi _k\) are strict. It follows that \(M^\varphi _k \rightarrow N^\varphi _k\) and \(M_k \rightarrow M^\varphi _k\) are strict also.

*k*-vector spaces the vertical maps are isomorphisms and the horizontal arrows are surjections. By assumption the leftmost vertical arrow is strict. Using part (3) of Lemma 5.4.2, \(M_k^\varphi \rightarrow N^\varphi _k\) being strict implies \(N^\varphi _k \rightarrow P^\varphi _k\) is strict. It follows that \(P_k \rightarrow P^\varphi _k\) and \(N_k \rightarrow P_k\) are strict. Thus

*M*and

*P*are as in Lemma 5.2.5 and, after (1) of Lemma 5.4.2, we know \(N \rightarrow P\) is strict. \(\square \)

### Lemma 5.4.5

Suppose *N* is strongly divisible. If \(0 \rightarrow M \rightarrow N \rightarrow P \rightarrow 0\) is an exact sequence in \({\text {Mod}}^{{\text {BK}}}_k\) then \(M_k \rightarrow N_k\) is strict.

### Proof

*i*.

For injectivity of \(\alpha \) when \(i< p\) we argue as follows. As \({\text {Weight}}(N) \subset [0, p]\), and because \(N_k \cong N^\varphi _k\), we have \({\text {gr}}^i(N_k) = 0\) for \(i<0\). Hence \({\text {gr}}^i(N) = {\text {gr}}^{i-p}(N)\) for \(i<0\). This implies \({\text {gr}}^i(N) = 0\) for \(i<0\) because \(F^i N = N\) for small enough *i*. Using the diagram we deduce that \({\text {gr}}^i(M) = 0\) for \(i<0\) also, and that for \(i<p\) we have \({\text {gr}}^i(M) = {\text {gr}}^i(M_k)\) and \({\text {gr}}^i(N) = {\text {gr}}^i(N_k)\). This proves \(\alpha \) is injective when \(i<p\).

For injectivity of \(\alpha \) when \(i \ge p\) it suffices to show \(F^i N_k = 0\) for \(i> p\) (because then \(F^iM_k = 0\) for \(i>p\) so \(\alpha \) is just the zero map when \(i>p\) and when \(i=p\), \(\alpha \) is the inclusion \(F^iM_k \rightarrow F^iN_k\)). By the strong divisibility of *N* this is equivalent to showing \(F^i N_k^\varphi = 0\) for \(i >p\). Since \({\text {Weight}}(N) \subset [0,p]\) we have \(F^{p+1} N_k^\varphi = 0\) which completes the argument. \(\square \)

Putting all this together we deduce the following.

### Proposition 5.4.6

- 1.If \(N \in {\text {Mod}}^{{\text {SD}}}_k\) then
*M*and*P*are strongly divisible and the sequenceis exact in \({\text {Fil}}(k)\). In particular, \({\text {Weight}}(N) = {\text {Weight}}(M) \cup {\text {Weight}}(P)\).$$\begin{aligned} 0 \rightarrow M^\varphi _k \rightarrow N^\varphi _k \rightarrow P^\varphi _k \rightarrow 0 \end{aligned}$$ - 2.
If \(P, M \in {\text {Mod}}^{{\text {SD}}}_k\) then \(N \in {\text {Mod}}^{{\text {SD}}}_k\) if and only if \(N \rightarrow P\) is strict.

### Proposition 5.4.7

- 1.
If \(N \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) then

*M*and*P*are both strongly divisible and, for each \(\tau \in {\text {Hom}}_{{\mathbb {F}}_p}(k,{\mathbb {F}})\), \({\text {Weight}}_\tau (N) = {\text {Weight}}_\tau (M) \cup {\text {Weight}}_\tau (P)\). - 2.
If \(M,P \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) then \(N \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) if and only if \(N \rightarrow P\) is strict.

### Proof

## 6 Irreducible objects

Provided \({\mathbb {F}}\) is sufficiently large, irreducible \({\mathbb {F}}\)-representations of \(G_K\) and \(G_{K_\infty }\) are induced from characters, see Lemma 2.1.2. In this section and the next we investigate the extent with which this is true for objects of \({\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\). Throughout assume \(k \subset {\mathbb {F}}\).

### 6.1 Rank ones

Recall from Construction 4.3.4 how \({\mathfrak {S}} \otimes _{{\mathbb {Z}}_p} {\mathcal {O}}\) is made into an \({\mathcal {O}}[[u]]\)-algebra. Then \(k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\) becomes an \({\mathbb {F}}[[u]]\)-algebra. Also let \(e_\tau \in k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\) denote the image of the idempotent \({\widetilde{e}}_\tau \in {\mathfrak {S}} \otimes _{{\mathbb {Z}}_p} {\mathcal {O}}\) defined in Construction 4.3.4. Thus \(\varphi (e_{\tau \circ \varphi }) = e_\tau \).

The next lemma is proven by an easy change of basis argument (see [9, Lemma 6.2]).

### Lemma 6.1.1

*M*is isomorphic to a Breuil–Kisin module

### Remark 6.1.2

If *N* is as in Lemma 6.1.1 then \({\text {Weight}}_\tau (N) = \lbrace r_\tau \rbrace \). Note also that *N* satisfies the equivalent conditions of Lemma 5.3.4. Thus \(N \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) if and only if \(r_{\tau } \in [0,p]\).

### Proposition 6.1.3

*N*is as in Lemma 6.1.1 then the \(G_{K_\infty }\)-action on

*T*(

*N*) is through the restriction to \(G_{K_\infty }\) of the character

*x*, and the \(\omega _\tau \) are the characters defined in the paragraph after the proof of Lemma 2.1.1.

### Proof

This is [9, Proposition 6.7]. However note that in *loc. cit.* they contravariantly associate a \(G_{K_\infty }\)-representation to Breuil–Kisin module; this is why the character appearing here is the inverse of that in *loc. cit*. \(\square \)

### 6.2 Induction and restriction

### Notation 6.2.1

*L*/

*K*be the unramified extension corresponding to a finite extension

*l*/

*k*, and let \(L_\infty = K_\infty L\). Set \({\mathfrak {S}}_L = W(l)[[u]]\). Extension of scalars along the inclusion \(f:{\mathfrak {S}} \rightarrow {\mathfrak {S}}_L\) describes a functor

### Notation 6.2.2

*f*induces a functor

### Lemma 6.2.3

### Proof

The standard adjunction between \(f^*\) and \(f_*\) provides functorial \({\mathfrak {S}}\)-linear isomorphisms \({\text {Hom}}_{{\mathfrak {S}}}(M,f_*N) \rightarrow {\text {Hom}}_{{\mathfrak {S}}_L}(f^*M,N)\). Explicitly, this map sends \(\alpha \) onto the homomorphism \(m \otimes s \mapsto s \alpha (m)\). As this is \(\varphi \)-equivariant we get isomorphisms as claimed. \(\square \)

### Lemma 6.2.4

### Proof

*p*-adic completion of \({\mathfrak {S}}_L[\frac{1}{u}]\). The map \(f: {\mathfrak {S}} \rightarrow {\mathfrak {S}}_L\) extends to a map \(f: {\mathcal {O}}_{{\mathcal {E}}} \rightarrow {\mathcal {O}}_{{\mathcal {E}},L}\) and so we can make sense of the operations \(f^*\) and \(f_*\) on etale \(\varphi \)-modules. Write \(M^{{\text {et}}} = M \otimes _{{\mathfrak {S}}} {\mathcal {O}}_{{\mathcal {E}}}\) and \(N^{{\text {et}}} = N \otimes _{{\mathfrak {S}}_L} {\mathcal {O}}_{{\mathcal {E}},L}\). Then clearly \(f^* (M^{{\text {et}}}) = (f^*M)^{{\text {et}}}\) and, because \({\mathcal {O}}_{{\mathcal {E}},L} = {\mathcal {O}}_{{\mathcal {E}}} \otimes _{{\mathfrak {S}}} {\mathfrak {S}}_L\), we also have that \(f_*(N^{{\text {et}}}) = (f_*N)^{{\text {et}}}\). We obtain maps

*T*. The analogue of Lemma 6.2.3 in the setting of etale \(\varphi \)-modules is proved in exactly the same way, and the obtained identification is compatible with the maps above. Thus, to prove the lemma we may replace \({\text {Hom}}_{{\text {BK}}}\) with \({\text {Hom}}_{{\text {et}}}\) (homsets in the category of etale \(\varphi \)-modules) and

*M*and

*N*with \(M^{{\text {et}}}\) and \(N^{{\text {et}}}\) in the diagram of the lemma.

*V*on a finitely generated \({\mathbb {Z}}_p\)-module. As (6.2.5) is functorial in

*V*, Yoneda’s lemma provides the isomorphism \(\iota _N\). As (6.2.5) is functorial in

*N*we see that \(\iota _N\) is functorial. \(\square \)

### Lemma 6.2.6

- 1.If \(M \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) then \(f^*M \in {\text {Mod}}^{{\text {SD}}}_l({\mathcal {O}})\) and for each \(\theta \in {\text {Hom}}_{{\mathbb {F}}_p}(l,{\mathbb {F}})\) we have$$\begin{aligned} {\text {Weight}}_\theta (f^*M) = {\text {Weight}}_{\theta |_k}(M) \end{aligned}$$
- 2.If \(N \in {\text {Mod}}^{{\text {SD}}}_l({\mathcal {O}})\) then \(f_*N \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) and$$\begin{aligned} {\text {Weight}}_\tau (f_*N) = \bigcup _{\theta |_k = \tau } {\text {Weight}}_\theta (N) \end{aligned}$$

### Proof

By functoriality both \(f^*\) and \(f_*\) preserve \({\mathcal {O}}\)-actions. Note that the inclusion \(k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}} \rightarrow l[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\) sends \(e_\tau \mapsto \sum _{\theta |_k = \tau } e_\theta \). Thus \((f^*M)_\theta = M_{\theta |_k}\) and \((f_*N)_\tau = \prod _{\theta |_k = \tau } N_\theta \). Both (1) and (2) then follow by verifying the second condition of Lemma 5.3.4. \(\square \)

### 6.3 Approximation by induced Breuil–Kisin modules

We consider the situation from Notation 6.2.1. Thus *L* / *K* is a finite unramified extension, corresponding to an extension *l* / *k* of residue fields, and \(L_\infty = L(\pi ^{1/p^\infty })\). We also have the map \(f:{\mathfrak {S}} \rightarrow {\mathfrak {S}}_L\).

### Lemma 6.3.1

*u*.

### Proof

*T*, induces the identification \(T(N) = T'\). Thus \(M \rightarrow f_*N\) becomes an isomorphism after inverting

*u*and is, in particular, injective. Lemma 6.2.6 implies \(f^*M \in {\text {Mod}}^{{\text {SD}}}_l({\mathcal {O}})\), since \(M \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\). Therefore \(N \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) by Proposition 5.4.7. \(\square \)

When *T*(*M*) is irreducible and \({\mathbb {F}}\) is sufficiently large *T*(*M*) is induced from a character. Thus, Lemma 6.3.1 produces an inclusion \(M \hookrightarrow f_*N\) with *N* of rank one. Lemma 6.1.1 allows us to describe *N* explicitly. In this case we would like to know which submodules of \(f_*N\) arise in this way. The following example shows that there are non-trivial (i.e. \(M \ne f_*N\)) possibilities.

### 6.4 An example

*L*/

*K*be of degree 5 with residue extension

*l*/

*k*. Let \(N \in {\text {Mod}}^{{\text {SD}}}_l({\mathcal {O}})\) be the rank one object defined by

### 6.5 Irreducibility and strong divisibility

*L*/

*K*,

*l*/

*k*and \(L_\infty /K_\infty \) be as in Notation 6.2.1; we obtain \(f:{\mathfrak {S}} \rightarrow {\mathfrak {S}}_L\). Let \(N \in {\text {Mod}}_l^{{\text {SD}}}({\mathcal {O}})\) be the rank one object given by

*N*is as in Lemma 6.1.1, except we’ve fixed \(x = 1\). This is to simplify notation (it will be easy to reduce from the general case to this one). The following proposition describes which Breuil–Kisin modules embed into \(f_*N\) as in Lemma 6.3.1.

### Proposition 6.5.1

- 1.
If \(m \in M\) then \(\varphi (m) \in M\) and if \(m \in f_*N\) and \(\varphi (m) \in uM\) then \(m \in M\).

- 2.
If \(m \in f_*N\) then \(um \in M\).

- 3.If \(\sum \alpha _\theta e_\theta \in M\) with \(\alpha _{\theta } \in {\mathbb {F}}\) thenfor every \(0 \le r \le p\).$$\begin{aligned} \sum _{r_\theta \equiv r~{\text { mod}}p} \alpha _{\theta } e_{\theta } \in M \end{aligned}$$

### Proof (that SD implies (1), (2) and (3))

Any \(m \in M\) with \(\varphi (m) \in u^{p+1}M\) must be zero in \(M_k\), and so is contained in

*uM*.

*T*(

*N*) is the trivial character. In this case \(T(f_*N)\) is not irreducible.

Now we can deduce the second part of (1). If \(m \in f_*N\) and \(\varphi (m) \in uM\) then \(\varphi (um) \in u^{p+1}M\); as (2) holds we have \(um \in M\) and so the above bullet point give \(um \in uM\). Hence \(m \in M\).

- There exist \(\widetilde{\alpha _{\theta ,r}} \in {\mathbb {F}}[[u]]\) such that \(\sum \widetilde{\alpha _{\theta ,r}} e_\theta \in M\), \(u^{r-1} \sum \widetilde{\alpha _{\theta ,r}} e_\theta \in M^\varphi \) and$$\begin{aligned} \widetilde{\alpha _{\theta ,r}} \equiv {\left\{ \begin{array}{ll} \alpha _{\theta }~{\text {mod}} u &{} \text {if }r_\theta \ne r, \text { except possibly if }r_\theta = 0\text { and }r = p \\ 0~{\text {mod}} u &{} \text {if }r_\theta = r \end{array}\right. } \end{aligned}$$

*M*is strongly divisible, the map \(M_k \rightarrow M^\varphi _k\) is an isomorphism of filtered modules. As \(u^r\sum \alpha _{\theta } e_\theta \in F^rM^\varphi \) it follows that there exists an element \(\beta \in F^r M\) such that \(\varphi (\beta ) - u^r\sum \alpha _{\theta } e_\theta \in uM^\varphi \). If \(\beta = \sum \beta _\theta e_{\theta \circ \varphi }\) then

*u*-adic valuation. If \(r_\theta = r\) this inequality implies \(\alpha _{\theta } \equiv \varphi (\beta _{\theta })\) modulo

*u*, and so we can write \(\varphi (\beta _{\theta }) = \alpha _{\theta } + u \gamma _{\theta }\) for some \(\gamma _{\theta } \in {\mathbb {F}}[[u]]\). If \(r > r_\theta \) the inequality implies \(\varphi (\beta _{\theta }) \equiv 0\) modulo

*u*, and so we can write \(\varphi (\beta _{\theta }) = u^p\gamma _{\theta }\) for some \(\gamma _{\theta } \in {\mathbb {F}}[[u]]\). If \(r_\theta > r\) then we simply write \(\varphi (\beta _{\theta }) = \gamma _{\theta }\). Dividing (6.5.2) by \(u^r\) we therefore see that

### 6.6 Finishing the proof of Proposition 6.5.1

Let *N* be as in the previous subsection and suppose that \(M \subset f_*N\) is a free \(k[[u]] \otimes _{{\mathbb {F}}_p} {\mathbb {F}}\)-module with \(M[\frac{1}{u}] = (f_*N)[\frac{1}{u}]\). Assume that *M* satisfies conditions (1), (2) and (3) from Proposition 6.5.1. We are going to prove that \(M \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\). Along the way we shall describe the weights of *M* in terms of the \(r_\theta \).

### Construction 6.6.1

*X*depends upon the choice of \(\lambda \).

### Lemma 6.6.3

- 1.
If \(\sum _{\kappa \in X} \alpha _{\kappa }e_\kappa \) is an \({\mathbb {F}}\)-linear combination contained in

*M*then \(\sum _{\kappa \in X} \alpha _{\kappa } e_\kappa = 0\). - 2.If \(\theta \not \in X\) there exists a unique \({\mathbb {F}}\)-linear combinationin which the sum runs over \(\kappa \in X\) satisfying ($$\begin{aligned} e_{\theta } + \sum \alpha _{\kappa } e_{\kappa } \in M, \quad \alpha _{\kappa } \in {\mathbb {F}} \end{aligned}$$
*i*) \(\kappa <_\lambda \theta \) (ii) \(r_{\kappa } \equiv r_\theta \) modulo*p*and (iii) \(\kappa |_k = \theta |_k\). In particular, the element lies in \(M_{\theta |_k}\).

### Proof

(1) If \(\sum _{\kappa \in X} \alpha _{\kappa } e_{\kappa } \ne 0\) then there exists a maximal \(\kappa \) (with respect to \(<_{\lambda }\)) with \(\alpha _{\kappa } \ne 0\). From \(\sum _{\kappa \in X} \alpha _{\kappa } e_\kappa \in M\) it follows this maximal \(\kappa \) is not contained in *X*, a contradiction.

(2) As \(\theta \not \in X\), there exists \(e_\theta + \sum \alpha _{\kappa } e_{\kappa } \in M\) with the sum running over \(\kappa <_\lambda \theta \). Arguing inductively one shows there exists such a sum running only over those \(\kappa <_\lambda \theta \) with \(\kappa \in X\). There can be at most one sum of this form; indeed their difference would be a sum as in (1) and so would be zero. Condition (3) of Proposition 6.5.1 therefore implies the sum may be taken to run over \(\kappa \) additionally satisfying (ii). As \(M = \prod _{\tau \in {\text {Hom}}_{{\mathbb {F}}_p}(k,{\mathbb {F}})} M_{\tau }\) we also have (iii). \(\square \)

### Definition 6.6.4

Note that when there exists a \(\theta \) such that \(e_\theta \in M\) then the minimal elements are simply scalar multiples of \(e_\theta \) for any \(\theta \) with \(e_\theta \in M\).

### Lemma 6.6.6

If (6.6.5) is a minimal sum then \(r_{\iota \circ \varphi ^j} = r_\iota \) whenever \(\alpha _j \ne 0\) and \(j \le I\).

### Proof

*p*. Since each \(r_{\iota \circ \varphi ^j} \in [0,p]\) this will be an equality, except possibly if \(r_\iota = 0\) or

*p*. In this case set

*M*contradicting the minimality of (6.6.5). \(\square \)

The next proposition is where we use that \(T(M)= T(f_*N)\) is irreducible.

### Proposition 6.6.7

- 1.
If \(\theta \in X\) and \(\theta \circ \varphi \not \in X\) then \(r_\theta >0\).

- 2.
If \(\theta \not \in X\) and \(\theta \circ \varphi \in X\) then \(r_{\theta } = 0\).

- 3.
If \(\theta \in X\) and \(e_{\theta \circ \varphi } \not \in M\) then \(0 \le r_\theta \le 1\). In particular this holds if \(\theta \circ \varphi \in X\).

### Proof

First observe (3) holds for any choice of \(\lambda \). Indeed, if \(e_{\theta \circ \varphi } \not \in M\) then condition (1) of Proposition 6.5.1 would imply \(\varphi (e_{\theta \circ \varphi }) = u^{r_\theta } e_\theta \not \in uM\). If \(r_\theta \ge 2\) then \(u^{r_\theta } e_\theta \in u^2f_*N\), which is contained in *uM* by (2) of Proposition 6.5.1.

Next we show (2) holds whenever \(r_\lambda =0\). Suppose \(\theta \not \in X\) and \(r_\theta > 0\) (we’re assuming that \(r_\lambda = 0\) so \(\theta \ne \lambda \)). We show \(\theta \circ \varphi \not \in X\). Choose \(e_\theta + \sum _{\kappa <_\lambda \theta } \alpha _{\kappa }e_\kappa \in M\) as in Lemma 6.6.3. Set \(z = e_{\theta \circ \varphi } + \sum _{r_\kappa \ne 0} \alpha _{\kappa } e_{\kappa \circ \varphi } + u\sum _{r_\kappa = 0} \alpha _{\kappa } e_{\kappa \circ \varphi }\). Using that \(r_\theta \equiv r_\kappa \) modulo *p* and \(r_\theta >0\) we see that \(\varphi (z) = u^{r_\theta }( e_\theta + \sum \alpha _{\kappa }e_\kappa )\). Condition (1) of Proposition 6.5.1 implies \(z \in M\). Since \(\theta \ne \lambda \), if \(\kappa <_\lambda \theta \) then \(\kappa \circ \varphi <_\lambda \theta \circ \varphi \). Therefore *z* shows \(\theta \circ \varphi \not \in X\).

Now choose a minimal sum as in (6.6.5) (if none exists then we must have \(M = u(f_*N)\) and so \(X = {\text {Hom}}_{{\mathbb {F}}_p}(l,{\mathbb {F}})\), in which case conditions (1), (2), and (3) hold vacuously). We are going to show that either \(\lambda = \iota \) satisfies the conditions of the proposition or \(e_{\iota \circ \varphi } + \sum \alpha _{j} e_{\iota \circ \varphi ^{j+1}} \in M\). Let us explain why this implies the proposition. If there exists no \(\lambda \in {\text {Hom}}_{{\mathbb {F}}_p}(l,{\mathbb {F}})\) satisfying conditions (1)–(3) then it would follow that \(e_{\iota \circ \varphi ^i} + \sum \alpha _{j} e_{\iota \circ \varphi ^{j+i}} \in M\) for every \(i \ge 0\). Lemma 6.6.6 would then imply \(r_{\iota '} = r_{\iota ' \circ \varphi ^I}\) for every \(\iota ' \in {\text {Hom}}_{{\mathbb {F}}_p}(l,{\mathbb {F}})\). If \(\omega \) is the character through which \(G_{K}\) acts on *T*(*N*) we would then have \(\omega = \prod _\iota \omega _{\iota }^{-r_\iota } = \prod \omega _{\iota \circ \varphi ^I}^{-r_\iota } = \omega ^{p^I}\). If \(I >0\) this contradicts the irreducibility of \(T(M) = {\text {Ind}}_L^K T(N)\). If \(I = 0\) it would follow that every \(e_{\iota '} \in M\), so \(X = \emptyset \) and conditions (1)–(3) hold vacuously.

### Proof (End of the proof of Proposition 6.5.1)

*M*is strongly divisible. Fix \(\lambda \) as in Proposition 6.6.7 and for \(\theta \in {\text {Hom}}_{{\mathbb {F}}_p}(l,{\mathbb {F}})\) set

*W*. We see that \(\sum \alpha _{\theta }e_\theta - \sum _{\theta \not \in X} \alpha _{\theta }f_\theta \) is an \({\mathbb {F}}\)-linear combination of \(e_\theta \) with \(\theta \in X\), and is contained in

*M*. Such a linear combination must be zero (cf. the proof of Lemma 6.6.3) so \(W = M_\tau \), as claimed.

- Suppose \(\theta \not \in X\) and \(\theta \circ \varphi \in X\). Set \(h_\theta := f_\theta = e_\theta + \sum _{\kappa <_\lambda \theta , \kappa \in X} \alpha _{\kappa } e_{\kappa } \in M_{\theta |_k}\). (2) of Proposition 6.6.7 implies \(r_\theta = 0\), so each \(r_\kappa \), being congruent to \(r_\theta \) modulo
*p*, equals 0 or*p*. If \(r_\kappa = p\) then (3) of Proposition 6.6.7 implies \(e_{\kappa \circ \varphi } \in M\), and so \(\kappa \circ \varphi \not \in X\) and \(f_{\kappa \circ \varphi } = e_{\kappa \circ \varphi }\). If \(r_\kappa = 0\) then (1) of Proposition 6.6.7 implies \(\kappa \circ \varphi \in X\). Thusis such that \(\varphi (g_{\theta \circ \varphi }) = u^ph_\theta \). Since \(h_\theta \in M_{\theta }\) we must have \(g_{\theta \circ \varphi } \in M_{\theta \circ \varphi |_k}\).$$\begin{aligned} g_{\theta \circ \varphi } := f_{\theta \circ \varphi } + \sum _{\kappa <_\lambda \theta , \kappa \in X} \alpha _{\kappa } f_{\kappa \circ \varphi } \in M \end{aligned}$$ - Suppose \(\theta \not \in X\), \(\theta \circ \varphi \not \in X\) and \(r_\theta =0\). Set \(g_{\theta \circ \varphi } := f_{\theta \circ \varphi } = e_{\theta \circ \varphi } + \sum _{\kappa <_\lambda \theta \circ \varphi ,\kappa \in X} \alpha _{\kappa } e_{\kappa } \in M_{\theta \circ \varphi |_k}\). Since \(\kappa \in X\), if \(\kappa \circ \varphi ^{-1} \not \in X\) then \(r_{\kappa \circ \varphi ^{-1}} = 0\) by (2) of Proposition 6.6.7. By (3) of Proposition 6.6.7, if \(\kappa \circ \varphi ^{-1} \in X\) then \(r_{\kappa \circ \varphi ^{-1}} \in [0,1]\). Therefore the difference between \(\varphi (g_{\theta \circ \varphi })\) andis an \({\mathbb {F}}\)-linear combination of \(e_\kappa \) with \(\kappa \in X\). Since this \({\mathbb {F}}\)-linear combination is contained in$$\begin{aligned} h_\theta := f_{\theta } + \sum _{\kappa<_\lambda \theta \circ \varphi ,\kappa \circ \varphi ^{-1} \not \in X} \alpha _{\kappa } f_{\kappa \circ \varphi ^{-1}} + \sum _{\kappa <_\lambda \theta \circ \varphi ,\kappa \circ \varphi \in X,r_{\kappa \circ \varphi ^{-1}} = 1} \alpha _{\kappa } f_{\kappa \circ \varphi ^{-1}} \end{aligned}$$
*M*it must be zero by (1) of Lemma 6.6.3. Therefore \(\varphi (g_{\theta \circ \varphi }) = h_\theta \). Since \(g_{\theta \circ \varphi |_k}\) we must have \(h_\theta \in M_{\theta |_k}\). - Suppose \(\theta \not \in X\), \(\theta \circ \varphi \not \in X\) and \(r_\theta > 0\). Set \(h_\theta := f_\theta = e_\theta + \sum _{\kappa <_\lambda \theta ,\kappa \in X} \alpha _{\kappa }e_\kappa \in M_{\theta |_k}\). Each \(r_\kappa \equiv r_\theta \) modulo
*p*and so \(\varphi \) sendsonto \(u^{r_\theta } h_\theta \) (note the term \(u\sum _{r_\kappa = 0} \alpha _{\kappa } e_{\kappa \circ \varphi }\) appears only if \(r_\theta = p\)). As \(r_\theta >0\) this displayed sum is contained in$$\begin{aligned} e_{\theta \circ \varphi } + \sum _{r_\kappa > 0} \alpha _{\kappa }e_{\kappa \circ \varphi } + u\sum _{r_\kappa = 0} \alpha _{\kappa } e_{\kappa \circ \varphi } \end{aligned}$$*M*by condition (1) of Proposition 6.5.1. We claim this displayed sum is equal toTo see this note that, by (1) of Proposition 6.6.7, if \(r_\kappa = 0\) then \(\kappa \circ \varphi \in X\) and if \(r_{\kappa \circ \varphi } \not \in X\) then \(r_\kappa > 0\). From this it follows that the difference between these two sums, which is an element of$$\begin{aligned} g_{\theta \circ \varphi } := f_{\theta \circ \varphi } + \sum _{\kappa<_\lambda \theta , \kappa \circ \varphi \not \in X} \alpha _{\kappa } f_{\kappa \circ \varphi } + \sum _{\kappa <_\lambda \theta , \kappa \circ \varphi \in X, r_\kappa =0} \alpha _{\kappa } f_{\kappa \circ \varphi } \end{aligned}$$*M*, is an \({\mathbb {F}}\)-linear combination of \(e_\kappa \) with \(\kappa \in X\). This difference is therefore zero, and so \(\varphi (g_{\theta \circ \varphi }) = u^{r_\theta } h_\theta \). As \(h_\theta \in M_{\theta |_k}\) we have \(g_{\theta \circ \varphi } \in M_{\theta \circ \varphi |_k}\). - Suppose \(\theta \in X\) and \(\theta \circ \varphi \not \in X\). Set \(g_{\theta \circ \varphi } := f_{\theta \circ \varphi } = e_{\theta \circ \varphi } + \sum _{\kappa <_\lambda \theta \circ \varphi ,\kappa \in X} \alpha _{\kappa }e_{\kappa } \in M_{\theta \circ \varphi |_k}\), and setWe claim \(\varphi (g_{\theta \circ \varphi }) = u^{r_\theta -1}h_\theta \). If \(e_{\theta \circ \varphi } \in M\) then this is clear since \(g_{\theta \circ \varphi } = e_{\theta \circ \varphi }\) and \(h_\theta = ue_\theta \). If \(e_{\theta \circ \varphi } \not \in M\) then (1) and (3) of Proposition 6.6.7 implies \(r_\theta = 1\), so we have to show \(\varphi (g_{\theta \circ \varphi }) = h_\theta \). Proposition 6.6.7 tells us \(\kappa \in X\) and \(\kappa \circ \varphi ^{-1} \not \in X\) implies \(r_{\kappa \circ \varphi ^{-1}} = 0\), while if \(\kappa \in X\) and \(\kappa \circ \varphi ^{-1} \in X\) then \(r_{\kappa \circ \varphi ^{-1}} \in [0,1]\). Using these two facts we see that the difference between \(\varphi (g_{\theta \circ \varphi })\) and \(h_\theta \) is an \({\mathbb {F}}\)-linear combination of \(e_\kappa \) with \(\kappa \in X\). Since this difference is contained in$$\begin{aligned} h_\theta := f_\theta + \sum _{\kappa<_\lambda \theta \circ \varphi ,\kappa \circ \varphi ^{-1} \not \in X} \alpha _{\kappa } f_{\kappa \circ \varphi ^{-1}} + \sum _{\kappa <_\lambda \theta \circ \varphi ,\kappa \circ \varphi ^{-1} \in X, r_{\kappa \circ \varphi ^{-1}} = 1} \alpha _{\kappa } f_{\kappa \circ \varphi ^{-1}} \end{aligned}$$
*M*it must be zero. As \(g_{\theta \circ \varphi } \in M_{\theta \circ \varphi |_k}\) we have \(h_\theta \in M_{\theta |_k}\). Finally, if \(\theta \in X\) and \(\theta \circ \varphi \in X\) set \(g_{\theta \circ \varphi } := f_{\theta \circ \varphi } \in M_{\theta \circ \varphi |_k}\) and \(h_{\theta } := f_\theta \in M_{\theta |_k}\). Then \(\varphi (g_{\theta \circ \varphi }) = u^{r_\theta + p-1} h_\theta \).

*H*is the \({\mathbb {F}}[[u]]\)-linear endomorphism of \(M_\tau \) sending \(f_\theta \) onto \(h_\theta \) then \(H - {\text {Id}}\) sends \(f_\theta \) onto \({\mathbb {F}}\)-linear combinations of \(f_{\kappa \circ \varphi ^{-1}}\) with \(\kappa <_\lambda \theta \circ \varphi \). Hence \(H- {\text {Id}}\) is nilpotent,

*H*is an automorphism, and the \(h_\theta \) form an \({\mathbb {F}}[[u]]\)-basis as claimed. A similar observation shows the \(g_{\theta \circ \varphi }\) also form an \({\mathbb {F}}[[u]]\)-basis. \(\square \)

Using Remark 5.2.6 we deduce:

### Corollary 6.6.10

### 6.7 Putting everything together

Applying what we’ve shown so far in this subsection gives:

### Proposition 6.7.1

*T*(

*M*) irreducible. Then there exist integers \(\widetilde{r_{\theta }}\) indexed over \(\theta \in {\text {Hom}}_{{\mathbb {F}}_p}(l,{\mathbb {F}})\) such that (

*i*):

*L*an unramified extension

*K*, and such that (

*ii*):

### Proof

*N*to equal 1. Let us explain how to reduce to this case. Let \({\text {ur}}_x \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) be the rank one object given by

*M*, and so we may assume \(x= 1\). Applying Corollary 6.6.10, \({\text {Weight}}_\tau (M) = \lbrace r_\theta + ps_{\theta \circ \varphi } - s_\theta \mid \theta |_k = \tau \rbrace \). On the other hand, \(\chi = T(N)\) and this equals \(\prod _\theta \omega _{\theta }^{r_\theta +ps_{\theta \circ \varphi } - s_\theta }\) by Proposition 6.1.3. Therefore, we can take \(\widetilde{r_{\theta }} = r_\theta + ps_{\theta \circ \varphi } - s_\theta \). \(\square \)

## 7 Crystalline representations

In this section we state the key results which relate \({\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) with crystalline representations. We then give a proof of the theorem from the introduction

### 7.1 Crystalline representations and Breuil–Kisin modules

*p*-adic periods, and \(B_{{\text {crys}}} \subset B_{{\text {dR}}}\) the ring of crystalline periods. As in [7] a

*p*-adic representation

*V*of \(G_K\) is crystalline if

*t*is any choice of uniformiser.

### Theorem 7.1.1

(Kisin) There is a fully faithful functor \(T \mapsto M(T)\) which sends a crystalline \({\mathbb {Z}}_p\)-lattice onto an object of \({\text {Mod}}^{{\text {BK}}}_K\) which is free over \({\mathfrak {S}}\). The Breuil–Kisin module *M*(*T*) is uniquely determined by the fact that \(T(M(T)) = T|_{G_{K_\infty }}\).

### Proof

This is the main result of [10]. The formulation we give here is taken from [4, Theorem 4.4]. \(\square \)

### Notation 7.1.2

A crystalline \({\mathcal {O}}\)-lattice is a \(G_K\)-stable \({\mathcal {O}}\)-lattice inside a continuous representation of \(G_K\) on a finite dimensional *E*-vector space which is crystalline when viewed as a \({\mathbb {Q}}_p\)-representation. By functoriality \(M \mapsto T(M)\) restricts to a functor from the category of crystalline \({\mathcal {O}}\)-lattices into \({\text {Mod}}^{{\text {BK}}}_K({\mathcal {O}})\).

### Definition 7.1.3

*V*is a crystalline representation on an

*E*-vector space then \(D_{{\text {crys}}}(V)\) is a free module over \(K_0 \otimes _{{\mathbb {Q}}_p} E\) of rank \({\text {dim}}_E V\) and so \(D_{{\text {crys}}}(V)_K\) is a free \(K_0 \otimes _{{\mathbb {Q}}_p} E\)-module of rank \(e{\text {dim}}_E V\). If \(K_0 \subset E\) then as in Construction 4.3.4 there is a decomposition

*E*-vector space of dimension \(e\,{\text {dim}}_EV\). Define the \(\tau \)th Hodge–Tate weights of

*V*to be the multiset \({\text {HT}}_\tau (V)\) which contains

*i*with multiplicity

*e*copies of \(-1\)).

### Theorem 7.1.4

(Gee–Liu–Savitt, Wang) Suppose \(K = K_0\). If \(p=2\) choose \(\pi \) so that \(K_\infty \cap K(\mu _{p^\infty }) = K\). If *T* is a crystalline \({\mathcal {O}}\)-lattice such that \({\text {HT}}_\tau (V) \subset [0,p]\) where \(V = T \otimes _{\mathcal {O}} E\), then \({\overline{M}} := M(T) \otimes _{{\mathcal {O}}} {\mathbb {F}} \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) and \({\text {Weight}}_\tau ({\overline{M}}) = {\text {HT}}_\tau (V)\).

### 7.2 Proof of main theorem

We can now give the proof of the theorem in the introduction. Assume \(K = K_0\). Recall that if \({\overline{\rho }}:G_K \rightarrow {\text {GL}}_n(\overline{{\mathbb {F}}}_p)\) is a continuous representation then in Definition 2.1.3 we defined the set \({\text {Inert}}({\overline{\rho }})\).

### Theorem 7.2.1

### Proof

Choose a coefficient field *E* so that \(\rho \) is defined over \({\mathcal {O}}\). Via a straightforward twisting argument we may suppose \({\text {HT}}_\tau (\rho ) \in [0,p]\). Let \(M(\rho ) \in {\text {Mod}}^{{\text {BK}}}_K({\mathcal {O}})\) be the associated Breuil–Kisin module. By Theorem 7.1.4, \({\overline{M}} = M(\rho ) \otimes _{{\mathcal {O}}} {\mathbb {F}} \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) and \({\text {HT}}_\tau (\rho ) = {\text {Weight}}_\tau ({\overline{M}})\).

Choose a \(G_K\)-composition series of \(\rho \otimes _{{\mathcal {O}}} {\mathbb {F}}\). Enlarging *E* if necessary we can suppose that Lemma 2.1.2 holds for each Jordan–Holder factor. Let \(0 = {\overline{M}}_n \subset \cdots \subset {\overline{M}}_0 = {\overline{M}}\) be the corresponding composition series of \({\overline{M}}\). By Proposition 5.4.7 each of \({\overline{M}}_i/{\overline{M}}_{i+1} \in {\text {Mod}}^{{\text {SD}}}_k({\mathcal {O}})\) and \({\text {Weight}}_{\tau }({\overline{M}}) = \bigcup _i {\text {Weight}}_\tau (M_i/M_{i+1})\). Lemma 2.2.2 implies \(T({\overline{M}}_i/{\overline{M}}_{i+1})\) is induced from a character \(\chi _i :L \rightarrow {\mathbb {F}}^\times \) for some unramified extension *L* / *K* (depending on *i*). Therefore Proposition 6.7.1 applies to \({\overline{M}}_i/{\overline{M}}_{i+1}\) and shows that the weights of \({\overline{M}}_i/{\overline{M}}_{i+1}\) are contained in \({\text {Inert}}({\text {Ind}}_L^K\chi _i)\). Since this is true for each *i* we deduce \((\lambda _\tau ) \in {\text {Inert}}({\overline{\rho }})\). \(\square \)

## Footnotes

- 1.
Using \(\iota \) we can identify \(\kappa \in {\text {Hom}}(F,{\mathbb {C}})\) with pairs \((v,{\widetilde{\tau }})\) where

*v*is a place of*F*above*p*and \({\widetilde{\tau }} \in {\text {Hom}}(F_v,\overline{{\mathbb {Q}}}_p)\). Since*p*is unramified in*F*, \({\widetilde{\tau }}\) can be identified with \(\tau \in {\text {Hom}}_{{\mathbb {F}}_p}(k_v,\overline{{\mathbb {F}}}_p)\) where \(k_v\) denotes the residue field of \(F_v\). The \(\kappa \)-th Hodge–Tate weights of \(r_{\iota ,\varPi }\) are then the \(\tau \)-th Hodge–Tate weights of \(r_{\iota ,p}(\varPi )\) at*v*. - 2.

## Notes

### Acknowledgements

Open access funding provided by Max Planck Society. This paper contains part of the my Ph.D. thesis, and I would like to thank my advisor Fred Diamond for his guidance and support. I would also like to thank Dougal Davis and Misja Steinmetz for helpful conversations, David Savitt for bringing our attention to the work of Wang [14], and Xavier Caruso for encouraging me to extend the main result (which was originally proved for distinct Hodge–Tate weights). Finally it is a pleasure to acknowledge the debt this paper owes to the work of Toby Gee, Tong Liu, and David Savitt. Funding was provided by Engineering and Physical Sciences Research Council (EP/L015234/1) and Max-Planck-Institut für Mathematik in den Naturwissenschaften.

## References

- 1.Barnet-Lamb, T., Gee, T., Geraghty, D.: Serre weights for \(U(n)\). J. Reine Angew. Math.
**735**, 199–224 (2018)MathSciNetCrossRefGoogle Scholar - 2.Barnet-Lamb, T., Gee, T., Geraghty, D., Taylor, R.: Potential automorphy and change of weight. Ann. Math. (2)
**179**(2), 501–609 (2014)Google Scholar - 3.Berger, L.: La correspondance de Langlands locale \(p\)-adique pour \({\rm GL}_2({\bf Q}_p)\), Astérisque, 339, Exp. No. 1017, viii, 157–180 (2011). (French, with French summary), Séminaire Bourbaki. Vol. 2009/2010. Exposés 1012–1026Google Scholar
- 4.Bhatt, B., Morrow, M., Scholze, P.: Integral \(p\)-adic Hodge theory. Preprint, arXiv:1602.03148 (2016)
- 5.Chenevier, G., Harris, M.: Construction of automorphic Galois representations. II. Camb. J. Math.
**1**(1), 53–73 (2013)MathSciNetCrossRefGoogle Scholar - 6.Fontaine, J.-M.: Le corps des périodes \(p\)-adiques. Astérisque
**223**, 59–111 (1994). (**(French) With an appendix by Pierre Colmez; Périodes p-adiques (Bures-sur-Yvette, 1988)**)Google Scholar - 7.Fontaine, J.-M.: Représentations \(p\)-adiques semi-stables. Astérisque
**223**, 113–184 (1994). (**(French) With an appendix by Pierre Colmez; Périodes p-adiques (Bures-sur-Yvette, 1988)**)Google Scholar - 8.Fontaine, J.-M.: Représentations \(p\)-adiques des corps locaux. I, The Grothendieck Festschrift, vol. II, Progr. Math., vol. 87, pp. 249–309. Birkhäuser, Boston (1990) (
**French**)Google Scholar - 9.Gee, T., Liu, T., Savitt, D.: The Buzzard-Diamond-Jarvis conjecture for unitary groups. J. Am. Math. Soc.
**27**(2), 389–435 (2014)MathSciNetCrossRefGoogle Scholar - 10.Kisin, M.: Crystalline Representations and \(F\)-crystals, Algebraic Geometry and Number Theory, Progr. Math., vol. 253, pp. 459–496. Birkhäuser, Boston (2006)Google Scholar
- 11.Scholze, P.: Perfectoid spaces. Publ. Math. Inst. Hautes Études Sci.
**116**, 245–313 (2012)MathSciNetCrossRefGoogle Scholar - 12.Serre, J.-P.: Propriétés galoisiennes des points d’ordre fini des courbes elliptiques. Invent. Math.
**15**(4), 259–331 (1972). (**French**)Google Scholar - 13.Serre, J.-P.: Local Algebra, Springer Monographs in Mathematics, pp. xiv+128. Springer, Berlin (2000). (
**Translated from the French by CheeWhye Chin and revised by the author**)Google Scholar - 14.Wang, X.: Weight elimination in dimensions 2 when \(p=2\). Preprint, arXiv:1711.09035 (2017)

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.