Integral Padic Hodge

Journal of Number Theory 147 (2015) 721748
Contents lists available at ScienceDirect
Journal of Number Theory

www.elsevier.com/locate/jnt
On some applications of integral p-adic Hodge

theory to Galois representations
Go Yamashita a,,1 , Seidai Yasuda b,2
a
Toyota Central R&D Labs. Inc., 41-1, Yokomichi, Nagakute, Aichi, 480-1192,
Japan
b
Department of Mathematics, Graduate School of Science, Osaka University,
Osaka, 560-0043, Japan
a r t i c l e
i n f o
Article history:
Received 5 November 2012
Received in revised form 18 July
2014
Accepted 18 July 2014
Available online 5 October 2014
Communicated by D. Burns
Keywords:
Automorphy (modularity) lifting
theorems
Integral p-adic Hodge theory
Wach module
Moduli of Wach modules
a b s t r a c t
We explicitly construct an analytic family of n-dimensional
crystalline representations by using integral p-adic Hodge
theory. This is a generalization of results by Berger, Li, and
Zhu and by Dousmanis. We show, by using Kisins method,
that the part of a universal deformation ring related to
the above constructions is connected. From this we obtain
an explicitly described subclass of potentially diagonalizable
representations in the sense of Barnet-Lamb, Gee, Geraghty
and Taylor. This yields automorphy lifting theorem and
potential automorphy theorem, in which the condition at p
is weakened.
2014 The Authors. Published by Elsevier Inc. This is an
open access article under the CC BY license
(http://creativecommons.org/licenses/by/3.0/).
* Corresponding author.
E-mail addresses: gokun@kurims.kyoto-u.ac.jp, gokun@mosk.tytlabs.co.jp (G. Yamashita),
s-yasuda@math.sci.osaka-u.ac.jp (S. Yasuda).
1
Supported by 21st Century COE Program in Kyoto University Formation of an international center of
excellence in the frontiers of mathematics and fostering of researchers in future generations, EPSRC grant
EP/E049109/1 in England, and TOYOTA Central R&D Labs., Inc.
2
Partially supported by JSPS Grant-in-Aid for Scientic Research 21540013, 24540018.
http://dx.doi.org/10.1016/j.jnt.2014.07.026
0022-314X/ 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/3.0/).
G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748
722
Contents
1.
2.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A family of Wach modules . . . . . . . . . . . . . . . . . . . . .
2.1. Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Review of the theory of Wach modules and Wach
2.3. Some elementary properties of A+
Qp . . . . . . . . . .
2.4. The matrix P ((Xi,j, )i,j[1,n], ) . . . . . . . . . .
(r)
......
......
......
lattices
......
......
2.5. The matrix G . . . . . . . . . . . . . . . . . . . . . . . .

2.6. The matrix G . . . . . . . . . . . . . . . . . . . . . . . . .
3. The representation of type (w, (ri, ), (i, )) . . . . . . . . . .
3.1. Semi-induced representations . . . . . . . . . . . . . . .
4. Local deformation rings and moduli of Wach modules . . .
5. Applications to automorphy lifting theorem and potential
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
722
724
724
725
728
729
...............
...............
...............
...............
...............
automorphy theorem
...............
...............
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
731
734
738
739
743
746
747
747
1. Introduction
Kisin introduced a new technique on automorphy (modularity) lifting theorems. The
technique is to study local deformation rings by using a moduli of semi-linear algebraic
objects (S-modules, and Wach modules) in the integral p-adic Hodge theory, and to
show Rred = T by using the properties of local deformation rings. In the GL2 case, this
method had an important application to Serres conjecture.
On the other hand, Clozel, Harris and Taylor proved some automorphy lifting theorems for unitary groups in [CHT]. Their method has been improved in [G,T,BLGHT]
(in [T], Taylor used Kisins technique so-called in = p case) and has produced an
important application to SatoTate conjecture [HSBT,BLGHT]. However, the automorphy lifting theorem has not been established in the best possible form, in the sense that
they require certain local conditions of the involved Galois representations. A main difculty in removing such local conditions lies in the study of local deformation rings in
the case where the coecient eld has the same characteristic p as the residue eld of
the considered local eld (so-called in = p case).
Currently, one of the strongest available statements is given in [BLGGT], where an
automorphy lifting theorem has been established under some mild condition in the case
where the Galois representation, restricted to the decomposition group at each prime
dividing p, is potentially diagonalizable in the sense of [BLGGT]. However, the notion
of potential diagonalizability is dened in terms of a universal deformation ring whose
structure has not yet been well-understood. Owing to this, potential diagonalizability is
in some sense like a tautological synonym for the local condition with which we can manage to show an automorphy of a global Galois representation by the currently available
methods, and it seems that we do not have a description of the potentially diagonalizable
representations which is concrete enough for applications.
In this paper we construct, for any integer n 2 and for any nite unramied extension K of Qp , an explicitly described subclass of potentially diagonalizable n-dimensional
723
p-adic representations (depending on a choice of some extra parameters w, r, which we

introduce in Denition 2.7) of the absolute Galois group GK of K, by using the method
of Berger, Li and Zhu [BLZ], and Kisin [K1,K2] (see Proposition 5.2). More precisely,
First, we construct an explicit analytic family of Wach modules by the method of
Berger, Li and Zhu (see Proposition 3.12 and Corollary 3.13).
Next, we show a property on the connected components of local deformation rings
by a method similar to that of Kisin (see Proposition 4.2).
The above extra parameters dene a p-adic representation of GK which is a direct sum
of representations induced from characters of open subgroups of GK corresponding to
nite unramied extensions of K. The subclass consists of the representations which are
p-adically near this direct sum. We call the members of the subclass the absolutely
nearly semi-induced representations. From our construction, combined with the result
in [BLGGT], we deduce an automorphy lifting and a potentially automorphy result for
an essentially conjugate self-dual p-adic representation r of the absolute Galois group
of a CM eld F , such that F is unramied at p and for each prime v of F dividing p,
the restriction of r to the decomposition group at v is absolutely nearly semi-induced
or a successive extension of one-dimensional representations (under some other mild
conditions for r). See Corollary 5.3 and Corollary 5.4 for the precise statements. We
emphasize the following two things:
The notion of the potentially diagonalizability in [BLGGT] means the representations
which are geometrically near to the induced representations. On the other hand,
the notion of absolutely nearly semi-induced representations in this paper means
the representations which are p-adically near to (the direct sums of) the induced
representations.
Our results are quantitative. We gave a p-adic estimate in which the automorphic
lifting machinery works (see also Denition 3.11).
The extra parameter r stands for the HodgeTate weights and we can construct the subclass of representations starting from any xed HodgeTate weights (however, when we
apply [BLGGT] after that, we have to assume that it has distinct HodgeTate weights).
However our results still remain partial and unfortunately they are not sucient to cover
all the n-dimensional crystalline representations of GK with xed HodgeTate weights,
except for the case where the HodgeTate weights are in an interval of length p 1,
and for some very limited case where n = 2 and the dierence of any two HodgeTate
weights is in {p, 0, p}. It seems dicult3 to extend our method, even when n = 2 and
the extra parameter w is not the identity, to cover the case where one of the eigenvalues
3
After a submission of this paper, we found a way of extending our method by using the hypergeometric
polynomials when n = 2 and K = Qp [YY]. We expect a possibility that this new technique is generalized to
the case where n 2 and K is an unramied extension of Qp .
724
of crystalline Frobenius is not close to zero. We note that, for such cases, a partial result
has been obtained in [K2] when K = Qp and n = 2, based on the computation by Berger
and Breuil of the reductions of some crystalline representations using the compatibility of the mod p and the p-adic local Langlands correspondences for GL2 (Qp ) (see also
Remark 3.14).
After a submission of our paper, a referee kindly informed us of the references [D,Z,
B3,GL] whose subjects are related to that in our paper. All of these papers are concerned
with the computation of the reduction of a crystalline representation V of GK with K/Qp
unramied. We briey explain their relations to our paper. The main results of [D] and
the calculations of the reductions in [D] are exclusively for the case where V is two dimensional. However some two dimensional cases not covered in our paper are treated there.
Moreover in Section 4 of [D], Dousmanis gives a criterion for a pair (P, G) of matrices
to be an approximation of the pair of the matrices of and on some family of Wach
modules. The approximation treats not only the case of the direct sum of the induced
representations like us, but also the case of more general ones. However he does not explicitly construct the family of Wach modules when the rank is greater than two. In this
sense, our paper is a generalization of [D]. In [Z] and [B3] they proved, among some other
results, that any crystalline representation V in an explicitly given p-adic neighbourhood
of V has the same reduction as V . However they do not give an explicit computation
of the reduction. (They also treat other topics. However, they have little relation to our
paper.) In [GL], they treat only the case where the length of the Hodge ltration is less
than or equal to p 1. In this sense, our paper is a generalization of [GL] as well.
Let us give a brief summary of the structure of the paper. In Section 2, we construct a
certain family of Galois representations by using the theory of Wach modules developed
in [B1]. Our family is a generalization of the family constructed in [BLZ]. In Section 3,
we introduce the notion of nearly semi-induced representations. In Section 4, we recall
the deformation rings introduced by Kisin [K2,K3] and check that the family of nearly
semi-induced representations are contained in one connected component of the local
deformation space. In Section 5, we describe some applications to automorphy lifting
and potential automorphy.
2. A family of Wach modules
Let n 1 be a positive integer. In this section, we construct an analytic family
of n-dimensional crystalline representations by using the theory of Wach modules, and
calculate its mod p reduction. This is a generalization of the work by Berger, Li and Zhu
[BLZ].
2.1. Notation
In this section, we x a rational prime p. Let K be a nite unramied extension
of Qp . Let val : K Z {} denote the normalized valuation on K, and let W K
725
be the valuation ring of K with respect to the valuation val. Let denote the Frobenius
on W , that is, the unique automorphism of the Zp -algebra W which induces the p-power
Frobenius automorphism on W/pW . Since K = W [1/p], the automorphism on W is
uniquely extended to the automorphism of K, which we denote by the same symbol by
abuse of notation. Fix an algebraic closure K of K and put GK := Gal(K/K). We put
K := Gal(K(p )/K) and let : K Zp denote the p-adic cyclotomic character.

Let us consider the ring K[[]] of formal power series with coecients in K with
respect to the formal variable . Let : K[[]] K[[]] denote the -semi-linear,
-adically continuous endomorphism of the K-algebra K[[]] which sends to (1+)p 1.
For K , let : K[[]] K[[]] denote the K-linear, -adically continuous endomorphism of the K-algebra K[[]] which sends to
(1 + )() 1 =
n1

(() i) n
i=0
n=1
n!
This gives an action of the group K on the K-algebra K[[]].

2.2. Review of the theory of Wach modules and Wach lattices
Let A+
K := W [[]] K[[]] denote the subring of formal power series with coecients
+
+
+
in W and put B+
K := AK [1/p]. The rings AK and BK are subrings of K[[]] which are
stable under the actions of and . We recall the denition of Wach modules and Wach
+
lattices over A+
K and over BK (cf. [B1,BLZ]).
Let
+

i
AK := W [[]][1/] =
ai ai W, val(ai ) as i
i=
denote the p-adic completion of the ring A+

K [1/]. The ring AK is a discrete valuation
ring and p is a uniformizer of AK . Let BK := AK [1/p] be the eld of fractions of AK .
The endomorphism of A+
K has a unique extension to a -semilinear endomorphism
of AK , which we also denote by , such that the latter is continuous with respect to
the p-adic topology. The action of the group K on A+
K has a unique extension to the
action of K on AK such that for each K , the automorphism : AK AK of
AK given by the action of is continuous with respect to the p-adic topology.
Let R be a nite Zp -algebra. We extend the endomorphism and the action of K to
(AK )R := RZp AK by R-linearity. A (, K )-module over (AK )R is a pair D = (D, D )
of a nitely generated free (AK )R -module D and a -semi-linear endomorphism D of D,
equipped with a continuous (with respect to the p-adic topology) semi-linear action of
the group K such that for each K the semi-linear automorphism of D given by
the action of commutes with the endomorphism D . We say that a (, K )-module D
over (AK )R is tale if the subset D (D) D generates D over (AK )R .
726
Let E is a nite extension of Qp and let OE denote its ring of integers. We extend
the endomorphism and the action of K to (BK )E := E Zp AK by E-linearity.
A (, K )-module over (BK )E is a pair D = (D, D ) of a nitely generated free
(BK )E -module D and a -semi-linear endomorphism D of D, equipped with a continuous (with respect to the p-adic topology) semi-linear action of the group K such
that for each K the semi-linear automorphism of D given by the action of commutes with the endomorphism D . We say that a (, K )-module D over (BK )E is tale
if there exists an (AK )OE -lattice D0 D stable under the actions of and K such
that D0 is an tale (, K )-module over (AK )OE .
For a nite Zp -algebra R (resp. a nite extension R of Qp ), a free R-representation of
GK is a nitely generated free R-module T equipped with a continuous, R-linear action
: GK EndR (T ) where we endow EndR (T ) with the p-adic topology. In [F], Fontaine
constructs a functor D from the category of free Zp -representations of GK to the category
of tale (, K )-modules over AK , and shows that the functor D gives an equivalence of
the categories. By extending this functor by R-linearity, we have a functor, also denoted
by D, which gives an equivalence between the category of free R-representations of GK
to the category of tale (, K )-modules over (AK )R (resp. (BK )R ).
From now on we x a nite extension E of Qp and let OE denote its ring of integers.
Let mE denote the maximal ideal of OE , and let F be the residue eld OE /mE . Below we
use the subscripts ()E and ()OE which respectively mean the operations EQp () and
OE Zp (). Since W is unramied over Zp , the OE -algebra WOE = OE Zp W is equal to

the direct product WOE = W , where for each , the OE -algebra W is the ring
of integers of a nite unramied extension of E, and the index set = E is a nite set
whose cardinality c is equal to the greatest common divisor of [F : Fp ] and [K : Qp ]. The
Frobenius automorphism = idOE of WOE induces a cyclic permutation :
=
of order c, and a -linear isomorphism W
W () of WOE -algebras for each .

For a WOE -algebra R and for , we put R = W WOE R so that R = R . The
+
=
endomorphism on (A+
(A+
K )OE induces an isomorphism : (AK )OE ,
K )OE , ()
for each .
In [B1] (see also [BLZ]), Berger gives a criterion for a free E-representation of GK
to be crystalline which is described purely in terms of its associated (, K )-module.
p
Let us recall the statement of the criterion. We put q := ()/ = i=1 pi i1 which
+
we regard an element in (B+
K )E = (AK [1/p])E . For an invertible element x = (x )

in (BK )E = (BK )E, and for a tuple a = (a ) of integers indexed by ,
a
we let xa denote the element (x ) in (BK )E . Let V be a free E-representation
of GK of dimension d := dimE V and let a = (a ) and b = (b ) be two tuples of integers indexed by such that a b holds for every . Then
a free E-representation V of GK is crystalline with HodgeTate weights in the tuple [a, b] := ([a , b ]) of intervals if and only if the associated (, K )-module
(D(V ), D(V ) ) over (BK )E has a (B+
K )E -submodule N(V ) D(V ) satisfying the following conditions:
(1)
(2)
(3)
(4)
727
The module N(V ) is free of rank d over (B+

K )E .
The action of K on D(V ) preserves N(V ).
The action of K on N(V ) induces a trivial action on N(V )/N(V ).
We have D(V ) ( b N(V )) b N(V ) and the quotient b N(V )/D(V ) ( b N(V )) is
b
killed by q ba . Here D(V ) ( b N(V )) denotes the (B+
K )E -submodule of N(V ) generated by the subset D(V ) ( b N(V )) b N(V ).
Moreover, if V is a free E-representation of GK which is crystalline with HodgeTate

weights in [a, b], then a (B+
K )E -submodule N(V ) D(V ) satisfying the above conditions
is unique and the condition (4) holds even when we replace a, b with any tuples of integers
a = (a ), b = (b ) with a b for every such that the HodgeTate weights of
V are in the tuple [a , b ] of intervals. Hence for a crystalline free E-representation V of
GK , the notation N(V ) makes sense even if we do not specify a range of its HodgeTate
weights.
=
When V is crystalline, Berger [B1] also gives an isomorphism N(V )/N(V )
Dcrys (V )E of ltered -modules with coecient in E. Here we endow N(V )/N(V )

with the structure of a ltered -module in the following way. The endomorphism of
N(V )/N(V ) is the one induced from the endomorphism (q/p)b of N(V ) for a suciently large integer b. We lter N(V ) by putting Fili N(V ) := {x N(V ) | D(V ) (x)
q i N(V )} for i Z, and endow the E-vector space N(V )/N(V ) with the ltration
induced by the ltration (Fili N(V ))iZ on N(V ).
+
A Wach module over (B+
K )E with HodgeTate weights in [a, b] is a free (BK )E -module
N of nite rank equipped with a semi-linear action K AutE (N ) of K which is
continuous with respect to the p-adic topology of AutE (N ) and with a -semi-linear homomorphism N : (B+
N (B+
N which commutes
K )E [1/] (B+
K )E [1/()] (B+
K )E
K )E
with the action of K such that the following conditions hold:
(1) The action of K on N induces a trivial action on N/N .
(2) We have N ( b N ) b N and the quotient b N/N ( b N ) is killed by q ba . Here
b
b
N ( b N ) denotes the (B+
K )E -submodule of N generated by the subset N ( N )
b
N.
If (N, N ) is a Wach module over (B+
K )E with HodgeTate weights in [a, b], then the
action of K on N (resp. the homomorphism N ) induces a semi-linear action of K
on the (BK )E -module D(N ) = (BK )E (B+ )E N (resp. a -semi-linear endomorphism
K
D(N ) of D(N )), and the (BK )E -modules together with the endomorphism D(N ) and
the action of K form an tale (, K )-module over (BK )E .
Let V be a crystalline free E-representation of GK with HodgeTate weights in a
tuple [a, b] of intervals. If T is a GK -stable OE -lattice in V , then N(T ) := D(T )
N(V ) is an (A+
K )OE -lattice in N(V ). The map T N(T ) gives a bijection between
the GK -stable OE -lattices in V and the (A+
K )OE -submodules N of N(V ) satisfying the
following conditions:
728
(1) The (A+

K )OE -module N is free of rank d = dimE V .
(2) The action of K on N(V ) preserves N .
(3) We have D(V ) ( b N ) b N and the quotient b N/D(V ) ( b N ) is killed by q ba .
b
Here D(V ) ( b N ) denotes the (B+
K )E -submodule of N generated by the subset
b
b
D(V ) ( N ) N .
We call an (A+
K )OE -submodule N of N(V ) satisfying the above conditions a Wach lattice
in N(V ).
2.3. Some elementary properties of A+
Qp
In this subsection, we assume that K = Qp . For later use we state the following two
statements as lemmas. Proofs of the statements are easy and are left to the reader.
Lemma 2.1. For any Qp , the element ()/ is invertible in A+
Qp .
Lemma 2.2. The sequence {i (1 + )}i0 converges to 1 in A+

Qp with respect to the
(p, )-adic topology on A+
.
2
Qp
Lemma 2.3. The sequence {i (q)}i0 converges to p in A+
Qp with respect to the (p, )-adic
topology.
Proof. Since i (q) =
p1
j=0
i ((1 + )j ), the claim follows from Lemma 2.2. 2
We dene a subring RQp of Qp [[]] as follows:

RQp :=

i0

ci Qp [[]] val(ci ) +
i
i
0 for all i 0
p1
The subring RQp is stable under and the action of K . We endow the ring RQp with
the linear topology such that the set

pi RQp + RQp j Qp [[]] i, j Z0
of ideals in RQp forms a fundamental system of neighbourhoods of 0. We easily see

that the ring RQp is complete with respect to this topology and that the element q/p is
invertible in RQp .
Lemma 2.3 has the following corollary:
Corollary 2.4. The sequence of invertible elements {i (q)/p}i0 converges to 1 in RQp .
The following lemma can be veried by direct calculation:
729
Lemma 2.5.
(1) The ring A+
Qp is a closed subring in RQp .
+
(2) The topology on A+
Qp induced from that on RQp by the inclusion AQp RQp coincides
the (p, )-adic topology on A+
Qp .
By combining Corollary 2.4 with Lemma 2.1, we get the following:
Corollary 2.6. For any Qp , the sequence {i (q/(q))}i0 converges to 1 in A+
Qp
with respect to the (p, )-adic topology.
For a map f : Z0 Z, we dene an element (f ) RQp to be the innite product
(f ) :=

i (q)/p
f (i)
i0
It follows from Corollary 2.4 that this innite product converges in RQp .
It is straightforward to check the element (f ) RQp has the following properties:
We have (f ) RQ
.
p
For two maps f1 , f2 : Z0 Z, we have (f1 + f2 ) = (f1 )(f2 ).
For a map f : Z0 Z, let Sf : Z0 Z denote the map
Sf (i) :=
0
f (i 1)
if i = 0,
if i 1.
Then we have ((f )) = (Sf ).

2.4. The matrix P ((Xi,j, )i,j[1,n], )
Let us return to the case for a general nite unramied extension K of Qp . Fix an
integer n 1 and put [1, n] := {1, . . . , n}. For a ring A, let Mn (A) denote the ring of
n-by-n matrices with entries in A. Let E be a nite extension of Qp and let OE denote
h
its ring of integers. We x 0 and put h =
(0 ) for h = 1, . . . , c 1. We x an
OE -basis (e ) of W0 , where the index set is a nite set of cardinality [K : Qp ]/c.
c
c
Let (c) = (,
of W0
), denote the matrix of the OE -linear automorphism
with respect to the basis (e ) .
We let

A+
K

OE

[[X]] := A+
Xi,j, (1 i, j n, )
K O
E
denote the ring of formal power series with n2 [K : Qp ]/c variables over (A+
K )OE .
+
+
For , we put (AK )OE , [[X]] = (AK )OE , [[Xi,j, (1 i, j n, )]] so that

+
(A+
K )OE [[X]] =
(AK )OE , [[X]].
730
Denition 2.7. We dene a set DataK,E,n to be the set of triples (w, r, ) of an element w
in the n-th symmetric group Aut([1, n]), a tuple r = (ri, )(i,)[1,n] of integers ri, Z
indexed by the pairs (i, ) [1, n] , and a tuple = (i, )(i,)[1,n] of elements
i, W indexed by the pairs (i, ) [1, n] . We call an element in DataK,E,n a
basic datum of degree n. We say that a basic datum (w, (ri, ), (i, )) is eective if ri,
is non-negative for every (i, ) [1, n] . We say that a basic datum (w, (ri, ), (i, ))
has distinct HodgeTate weights if the integers r1, , . . . , rn, are distinct for every .
Fix an eective basic datum (w, r, ) = (w, (ri, ), (i, )) DataK,E,n of degree n.
The goal of this subsection is to introduce a matrix P (X) Mn ((A+
K )OE [[X]]), which
depends on the basic data (w, r, ).
We put r,min = mini[1,n] ri, and r,max = maxi[1,n] ri, . Let w
denote the auto1
morphism of the set [1, n] which sends (i, ) [1, n] to (i, ()) if = 0 and
1
which sends (i, 0 ) to (w(i),
(0 )). Let us consider the following three conditions on
the data w and (ri, , i, )(i,)[1,n] :
There exists a pair (i, ) [1, n] such that rw k (i,) = rk (),max holds for any
k 0.
There exists a pair (i, ) [1, n] such that rw k (i,) = rk (),min holds for any
k 0.
r,max r,min is independent of .
We dene to be := 1 if all of the conditions hold, and := 0 otherwise. For
example, if w
has no xed points and if the integers (ri, )(i,)[1,n] are dierent with
each others, then neither of conditions holds and hence = 0 in this case. We put
r := + max(r,max r,min ).
(2.1)
We need some more notation before introducing the matrix P (X). For and for
+

f (A+
K )OE , [[X]], let (f ) (AK )OE , () [[X]] denote the formal power series
obtained by applying on each coecient of f . For f = (f ) (A+
K )OE [[X]],
let (f ) denote the element in (A+
)
[[X]]
whose
component
at
is
equal to
K OE

(f1 () ) for = 0 and whose component at 0 is equal to

(c)
(f1 (0 ) )
, Xi,j,
i,j,

A+
K O
E ,0
[[X]].
For f (A+
K )OE , [[X]] and for K , let (f ) denote the formal power series obtained by applying on each coecient of f . For a matrix M in Mn (K[[]]) or in
Mn ((A+
K )OE [[X]]), let (M ) (resp. (M )) denote the matrix obtained by applying
(resp. ) on the entries of M .
731
Let 1n be the identity n-by-n matrix, and let C be the permutation matrix corresponding to w1 , that is, C is the n-by-n matrix whose (i, j)-entry is w(i),j where ,
is the Kroneckers delta.
For each (i, ) [1, n] , let fi, : Z0 Z0 denote the map which sends j Z0
to
fi, (j) := rw j (i,) .
For i, j [1, n] and for , we write
(fi, )/(fj, ) =
(k)
zi,j, k .
k0
(k)
Let m denote the smallest non-negative integer such that pm zi,j, Zp for every i, j
[1, n], for every , and for every 0 k r 1. Since (fi, )/(fj, ) RQp , it follows
from the denition of RQp that m (r 1)/(p 1).
Example 1. One of the simplest cases where we cannot apply FontaineLaaille theory is
r = p. In this case, if ri, = 0 or p for any i [1, n], , then we have m = 0. However,
unfortunately for n 3, this condition (ri, = 0, or p for any i [1, n], ) cannot
hold with distinct HodgeTate weights. So, if r = p with distinct HodgeTate weights,
and n 3, then we have m = 1. For n = 2, it can hold with distinct HodgeTate weights
and we have m = 0 (see [BLZ, Remark 4.1.2, 1]).
(k)
(k)
We write zi,j, = pm zi,j, and put

zi,j, :=
r1
(k)
zi,j, k ,
k=0
which is, by the denition of m, an element in A+

Qp .
For , let Z denote the matrix in Mn (W [[]][[X]]) whose (i, j)-entry is zi,j,

Xi,j, . For , dene a matrix P (X) Mn (W [[]][[X]]) to be
P (X) :=
(1n + Z0 ) diag(1,0 q r1,0 , . . . , n,0 q rn,0 ) C, if = 0 ,

if = 0
(1n + Z ) diag(1, q r1, , . . . , n, q rn, ),
r1,
and put P (X) := (P (X)) Mn ((A+
, . . . , n, q rn, )
K )OE [[X]]). Here diag(1, q
r1,
denotes the diagonal matrix whose diagonal entries are 1, q , . . . , n, q rn, .
(r)
2.5. The matrix G
For , we let G Mn (RQp ) denote the matrix

G := diag (f1, ), . . . , (fn, ) .
732
It follows from the properties described in Section 2.3 that the matrix G is invertible
(as a matrix with entries in RQp ).
Lemma 2.8. We have

G = diag (q/p)r1, , . . . , (q/p)rn, (G1 () )
for = 0 and

G0 = diag (q/p)r1,0 , . . . , (q/p)rn,0 C(G1 (0 ) )C 1 .
Proof. Let i [1, n]. We have fw(i,)

(j 1) = fi, (j) for j 1. Hence the equality

(fi, ) = (q/p)ri, (fw(i,)
)

holds from which the claim immediately follows for = 0 . The claim for = 0 follows
from the equality

C(G1 (0 ) )C 1 = diag (fw(1),1 (0 ) ) , . . . , (fw(n),1 (0 ) ) .
(r)
For and for K , we let G, denote the n-by-n matrix

G (G )1 = diag (f1, )/ (f1, ) , . . . , (fn, )/ (fn, ) .
(r)
It follows from Corollary 2.6 that the matrix G, is an invertible matrix with entries
(r)
(r)
in A+ . We put G = (G, ) which we regard an element in Mn ((A+

K )OE ) =
Qp
+
M
((A
)
).
n
K OE ,
Lemma 2.9. We have

P (0) G(r)
= G(r)
P (0) .

+
Proof. For a matrix M Mn ((A+
K )OE ) =
Mn ((AK )OE , ) and for , let M
denote the component of M at . By Lemma 2.8, we have the equalities

r

r

(r)
G, = diag q/(q) 1, , . . . , q/(q) n, G(r)
for = 0 and

r

r

(r)
G,0 = diag q/(q) 1,0 , . . . , q/(q) n,0 C G(r)
C 1 .
On the other hand, the equalities

r

r

1
diag q/(q) 1, , . . . , q/(q) n, = P (0)
P (0)
733
for = 0 and

r

r

1
diag q/(q) 1,0 , . . . , q/(q) n,0 = C P (0)0
P (0)0 C 1
follow from the denition of P (0). By combining these four equalities, we have

1

1

1 C
G(r) C
G(r)
P (0)C
= C P (0)
Mn (WO ) =
where C
E
Mn (W ) denotes the matrix whose component at 0 is
equal to C and whose other components are equal to the identity matrix. The three
(r) 1
1 , C(G

matrices C(P
commute with each other since they are
(0))1 , P (0)C
)C
diagonal matrices. Hence the right hand side of the above formula can be rewritten as

1

1 C
G(r) C
C
P (0) 1 = P (0) G(r) P (0) 1 .
P (0)C
Thus we have G = P (0)(G )(P (0))1 , which proves the claim. 2

(r)
(r)
Proposition 2.10. The entries of the matrix

1
P (X) G(r)
P (X)
G(r)
are in r (A+
K )OE [[X]].
Proof. By the denition of P (X), we have
P (X)P (0)1 = 1n + Z,
(2.2)

+
where Z Mn ((A+
K )OE [[X]]) =
Mn ((AK )OE , [[X]]) denotes the matrix whose
component at is equal to Z for every . The matrix 1n +Z is congruent to 1n mod1
ulo the maximal ideal of (A+
K )OE [[X]]. Hence by (2.2), the matrix (P (X))(P (0))
is invertible in Mn ((A+
K )OE [[X]]). Thus to prove the claim, it suces to show that the
entries of the matrix

1

1
Y := P (X) G(r)
P (X)
G(r)
P (X) P (0)
are in r (A+
K )OE [[X]].
By (2.2), we have

1
Y = P (X) G(r)
G(r)
P (0)
P (X)

1
= P (X) G(r)
P (0)
G(r)
(1n + Z)
(r)
1
= (1n + Z)P (0) G P (0)
G(r)
(1n + Z).
734
We also have P (0)(G )(P (0))1 = G , by using Lemma 2.9. Thus we have
(r)
(r)
(r)
Y = (1n + Z)G(r)
G (1n + Z)
(r)
= ZG(r)
G (Z).
(r)
For h = 0, . . . , c 1, it follows from the denitions of G and Z that the component at

h of the (i, j)-entry of Y is

zi,j,h (fj,h )/ (fj,h ) (zi,j,h )(fi,h )/ (fi,h )

multiplied by Xi,j, . Let Kh denote the eld of fractions of Wh . Then by the
denition of zi,j,h , the element zi,j,h pm (fi,h )/(fj,h ) is in r Kh [[]]. Therefore, the component at h of the (i, j)-entry of Y is an element of r Kh [[]]

Wh [[]] = r Wh [[]] multiplied by
Xi,j, . In particular, the (i, j)-entry of Y
lies in r (A+
)
[[X]].
We
are
done.
2
K OE
2.6. The matrix G
Proposition 2.11. For each K , there exists a unique matrix G Mn ((A+
K )OE [[X]])
satisfying the following conditions:
(r)
G G mod r (A+
K )OE [[X]], and
P (X)(G ) = G (P (X)).
Before going to the proof, we will introduce some notation. We put P0 :=
+
P (X) mod (A+
K )OE [[X]] and Z0 := Z mod (AK )OE [[X]]. They are n-by-n matri+
ces with entries in (AK )OE /() = WOE [[X]]. For , we let D0, denote the diagonal
matrix diag(1 pr1, , . . . , n prn, ) Mn (W ) and put D0 = (D0, ) Mn (WOE ). It
follows from the denition of P0 that the equality

P0 = (1n + Z0 )D0 C
(2.3)
holds, which shows that the matrix P0 is invertible as a matrix with entries in
WOE [[]][1/p].
For and for f0, W [[X]], let (f0, ) W () [[X]] denote the formal
power series obtained by applying on each coecient of f0, . For f0 = (f0, )
WOE [[X]], let (f0 ) denote the element in WOE [[X]] whose component at is equal
to (f0,1 () ) for = 0 and whose component at 0 is equal to

(f0,1 (0 ) )
(c)
W0 [[X]].
, Xi,j,
i,j,
735
For M0 Mn (WOE [[X]]), let (M0 ) denote the matrix obtained by applying on the
entries of M0 .
Let M0 be an element in WOE [[X]] (resp. in Mn (WOE [[X]])). For any M
+
(A+
K )OE [[X]] (resp. for any M Mn ((AK )OE [[X]])) satisfying M0 M mod
(A+
K )OE [[X]], the following properties are easily checked:
(M0 ) (M ) mod (A+
K )OE [[X]], and
+
(M0 ) (M ) mod (AK )OE [[X]] for any K .
For M0 Mn (WOE [[X]]), put
(M0 ) := pr P0 (M0 )P01 .
By (2.1) and (2.3), the component (M0 ) of (M0 ) at is equal to

1 pr,max D1 (1n + Z0 )1 ,
(M0 ) = p (1n + Z0 ) pr,min D0 C(M
0 )C
0
(2.4)
where = r (r,max r,min ). Thus (M0 ) lies in Mn (WOE [[X]]), and any entries of
(M0 ) are divisible by p if = 1. Let m be the Jacobson radical of WOE [[X]]. It is clear
that the operation M0 (M0 ) is -semilinear. Hence induces a semilinear operation
0 := mod m on Mn (WOE [[X]]/m) = Mn (WOE OE F).
The following is a key lemma for the proof of Proposition 2.11:
Lemma 2.12.
(1) The operation 0 is nilpotent, i.e., there exists an integer i 1 such that 0i = 0.
Furthermore, we can take such an i satisfying i nc.
(2) The operation is m-adically nilpotent in the following sense: for any integer k 1,
there exists an integer i, such that the entries of i (M0 ) are in mk for any M0
Mn (WOE [[X]]).
Proof. First we prove (1). Suppose that = 1. Since 0 = 0 by (2.4), the claim is clear in
this case. Let us assume = 0. For i, j [1, n] and for , let Ei,j, Mn (WOE [[X]])
denote the matrix such that the component at of the (i, j)-entry is equal to 1 and that
the other components and other entries are equal to 0. Let denote the subset
of the elements satisfying = 0. Let I denote the subset of [1, n] which
consists of the elements (i, ) [1, n] satisfying ri, = r,min , and let J denote
the subset of [1, n] which consists of the elements (i, ) [1, n] satisfying
ri, = r,max .
We consider the reductions modulo m of the both sides of (2.4). We have Z0 0
mod m, and (pr,min D0, ) mod m is a diagonal matrix such that the component at
of the i-th diagonal entry is invertible if (i, ) I and is zero otherwise. On the other
736
1
hand, (pr,max D0,
) mod m is a diagonal matrix such that the component at of the
i-th diagonal entry is invertible if (i, ) J and is zero otherwise. Thus it follows from
(2.4) that 0 (Ei,j, ) is a multiple of Ei,j, () for () = 0 , and of Ew1 (i),w1 (j),0
for () = 0 by a constant in W [[X]]/m, and 0 (Ei,j, ) = 0 unless w
1 (i, ) I
and w
1 (j, ) J. (Note that w
1 (i, ) is equal to (i, ()) for () = 0 , and to
1
(w (i), 0 ) for () = 0 .) Since we have assumed = 0, there exists, for any i, j [1, n]
k (i, )
/ I or w
k (j, )
/ J. Hence
and for any , an integer k nc such that w
nc
nc
we have 0 (Ei,j, ) = 0 for any i, j [1, n] and for any . Therefore 0 = 0 which
prove the claim (1).
We prove the claim (2) by induction on k. If k = 1, the claim follows from (1).
Suppose that k 2, and assume that the statement holds for k 1. By the induction

hypothesis, there exists an integer i such that the entries of 0i (M0 ) are in mk1 for
any M0 Mn (WOE [[X]]). Thus there exist an integer 0, elements x1 , . . . , x
m
1 ,

and N1 , . . . , N
Mn (WOE [[X]]) such that 0i (M0 ) can be written as a nite sum
0i (M0 ) =
xj Nj .
j=1
Then, for i := i + nc, we obtain
0i (M0 ) =
nc (xj )0nc (Nj ).
j=1
Hence it follows from (1) that the entries of 0i (M0 ) are in mk . Thus the statement holds
for k. 2
Proof of Proposition 2.11. First we prove the uniqueness of G . Suppose that two matrices G and G , with G = G , satisfy the conditions of Proposition 2.11. Let k be
the maximum non-negative integer satisfying G G mod k (A+
K )OE [[X]]. We have
+

r
k r, since G G mod (AK )OE [[X]] by assumption. Put G G = k H,
k
k
where H Mn ((A+
K )OE [[X]]). By assumption, we have P (X)( H) = H(P (X)),
that is,

q k P (X)(H) = H P (X) .
(2.5)
+
Put H0 := H mod (A+
K )OE [[X]]. Then H0 is in Mn ((AK )OE [[X]]/()) =
Mn (WOE [[X]]), and H0 = 0 by the denition of k. By (2.5), we have pk P0 (H0 ) = H0 P0 .
Thus we have
H0 = pkr (H0 ).
(2.6)
737
By applying (2.6) repeatedly, we have

H0 = pi(kr) i (H0 )
for any i 1. It follows from Lemma 2.12 that the entries of the right hand
side of the above formula converge to 0 with respect to the m-adic topology of
WOE [[X]]. Hence we have H0 = 0, which gives a contradiction. This proves that
the matrix G satisfying the conditions of Proposition 2.11 is unique if it exists.
Next we will show the existence of G . It suces to construct Gk Mn ((A+
K )OE [[X]])
for each k r satisfying
(k)
(r)
G G mod r (A+
K )OE [[X]],
(k)
(k1)
k1
G G
mod
(A+
K )OE [[X]] for k > r, and
(k)
(k)
P (X)(G )(P (X))1 G mod k (A+
K )OE [[X]],
(k)
(k)
since we can dene G as the limit of G . We construct such a G by induction on k.

(k)
(r)
(k1)
For k = r, we put G := G . We assume that k > r and that G
is constructed.
(k)
To construct G , it is enough to nd a matrix H Mn ((A+
)
[[X]])
satisfying
O
E
K

1

P (X) G(k1) + k1 H P (X)
G(k1) + k1 H mod k A+
K O [[X]].
E
(2.7)
By
assumption,
there
exists
Y
Mn ((A+
satisfying
K )OE [[X]])
(k1)
(k1)
1
k1
P (X)(G
)(P (X)) G
=
Y . Put H1 := H mod (A+
K )OE [[X]],
+
Y0 := Y mod (AK )OE [[X]]. The equality (2.7) is equivalent to
H1 pk1 P0 (H1 )P01 = Y0 ,
that is,
H1 pkr1 (H1 ) = Y0 .
(2.8)
Let us consider an innite sum

H1 :=
pi(kr1) i (Y0 ).
i0
Lemma 2.12 shows that this innite sum m-adically converges, and satises (2.8). Take
an arbitrary lift H Mn ((A+
K )OE [[X]]) of H1 . Then H satises (2.7). Therefore the
(k)
(k1)
matrix G := G
+ k1 H satises the conditions listed in the previous paragraph.
This completes the proof of the existence of G . 2
738
3. The representation of type (w, (ri, ), (i, ))

We put

Tuple :=
mE .
i,j[1,n],
For a = (ai,j, ) Tuple, we dene an OE -algebra homomorphism eva : (A+

K )OE [[X]]
+
+
(AK )OE to be the homomorphism which sends f (X) = (f (X)) (AK )OE [[X]] to
(fh (ai,j, h (e )))0hc1 (A+
K )OE . It is easy to check that the homomorphism eva
+
is compatible with the endomorphisms of (A+
K )OE [[X]] and (AK )OE , and with the
+
+
actions of K on (AK )OE [[X]] and (AK )OE .
Let n 1 and let (w, r, ) DataK,E,n be an eective basic datum of degree n. We put
r,min = mini[1,n] ri, and r,max = maxi,[1,n] ri, . Let P (X) and G for K be
as in the previous section. For a Tuple, we put Pa := eva (P (X)) and G,a := eva (G ).
n
+
Let Na,OE = (A+
K )OE be a free (AK )OE -module of rank n. Let N,a be the -semi-linear
endomorphism of Na,OE whose matrix with respect to the standard basis is equal to
Pa . For K , let a be the -semi-linear endomorphism of Na,OE whose matrix with
respect to the standard basis is equal to G,a . The map a gives a semi-linear action
of the group K on Na,OE . We put Na = E OE Na,OE which is a free (B+
K )E -module of
rank n. We extend the endomorphism N,a of Na,OE (resp. the action of K on Na,OE )
to that of Na (resp. that on Na ) by E-linearity. The endomorphism N,a on Na induces
a -semi-linear homomorphism (B+
Na (B+
Na
K )E [1/] (B+
K )E [1/()] (B+
K )E
K )E
which we denote by the same symbol N,a by abuse of notation. The following lemma
is easily checked:
Lemma 3.1. The (B+
K )E -module Na together with the endomorphism N,a and the action of K form a Wach module over (B+
K )E with the HodgeTate weights in the tuple
([r (),max , r (),min ]) of intervals and the (A+
K )OE -module Na,OE is a Wach
lattice of the tale (, K )-module Na (B+ )E (BK )E over (BK )E . 2
K
Let Va be the free E-representation of GK corresponding to the tale (, K )-module

Na (B+ )E (BK )E over (BK )E . Then V is crystalline with the HodgeTate weights in
K
([r (),max , r (),min ]) . Let Ta Va be the GK -stable OE -lattice corresponding
to the Wach lattice Na,OE . The following lemma follows from the argument in [BLZ,
Theorem 4.1.1].
Lemma 3.2. For a Tuple, there is a canonical isomorphism F OE Ta
= F OE T0 of
free F-representation of GK where 0 = (0)i,j[1,n], Tuple.
We need the following lemma.
739
Lemma 3.3. Let and let s 1 be a positive integer. Let x be an element in (B+
K )E,
+
+
s
s
satisfying (x) q (BK )E, . Then x (BK )E, .
Proof. Take a primitive p-th root p of unity in an algebraic closure K of K = EOE W
and let W = W [p ]. Let g be the E -linear (, p)-adically continuous automorphism of
W [[]] which sends to p (1 + ) 1.
Since g((1 + )p ) = (1 + )p , we have g = on W [[]]. Hence (x) = g((x))
g(q)s W [[]]. Since g(q) = q/(p (1 + ) 1) is divisible by , we have (x) (B+
K )E,
+
s
s W [[]] = s (B+
)
.
This
shows
x
(B
)
.
2
K E,
K E,
Corollary 3.4. Let r, s 0 be two non-negative integers and let . Then the set

r

+
s

x B+
K E, q (x) q BK E,
sr
is equal to (B+
(B+
K )E, if r s, and is equal to
K )E, if r < s.
Corollary 3.5. For each integer i 0, the set

Fili Na = x Na N,a (x) q i Na
n
is equal to the set of the elements t ((x1, ) , . . . , (xn, ) ) Na = (B+
such that
K )E
+
irj,
xj,1 ()
(BK )E, for every (j, ) [1, n] with = 0 and i > rj, , and that
xw(j),1 (0 ) irj,0 (B+

K )E,0 for every j [1, n] with i > rj,0 . 2
3.1. Semi-induced representations

Denition 3.6. We say that a free E-representation V is semi-induced if V is isomorphic

to a direct sum i Vi of free E-representations of GK such that for each i, there exists
a nite extension Ei of E such that Ei E Vi is induced from an Ei -valued continuous
character of an open subgroup of GK .
Let L be a nite unramied extension of K and let OL denote its ring of integers. For
, the OE Zp OL -algebra W W OL is equal to a direct product W W OL =

W, , where for each , the OE -algebra W, is the ring of integers of a
nite unramied extension of K , and the index set is a nite set whose cardinality
is equal to the greatest common divisor of [F : Fp ]/c and [L : K].
Proposition 3.7. Let V be a crystalline free E-representation V of GK of dimension n. Let
Dcrys (V )E be the ltered -module over EQp K associated to V . Then V is semi-induced
if there exist a basic datum (w, r, ) = (w, (ri, ), (i, )) DataK,E,n of degree n and an
E Qp K-basis d1 , . . . , dn of Dcrys (V )E which satisfy the following properties:
740
(1) For each i Z, the submodule Fili Dcrys (V )E is equal to the set of elements
n
1
j=1 xj dj Dcrys (V )E such that the component of xj at () is equal to zero
for every (j, ) [1, n] with = 0 and i > rj, , and that the component of xw(j)
1
at
(0 ) is equal to zero for every j [1, n] with i > rj,0 .
(2) For , the component at of the matrix of with respect to the basis d1 , . . . , dn
is equal to

diag 1, pr1, , . . . , n, prn,
for = 0 and is equal to

diag 1, pr1, , . . . , n, prn, C
for = 0 , where C denotes the permutation matrix corresponding to w1 .
Moreover the converse is true if V has distinct HodgeTate weights or if we replace V
by E E V for some nite extension E of E.
Denition 3.8. For a basic datum (w, r, ) DataK,E,n and for a crystalline free
E-representation V of GK , we say that a V is semi-induced of type (w, r, ) if V satises
the conditions in Proposition 3.7 for (w, r, ).
Remark 3.9. We remark that, for a free E-representation V of GK which is crystalline
and semi-induced, a basic data (w, r, ) as in Proposition 3.7 is not necessarily unique.
To prove Proposition 3.7 we need the following lemma.
Lemma 3.10. A free E-representation V of GK is semi-induced and crystalline if and

only if V is isomorphic to a direct sum
i Vi of free E-representations such that for
each i there exist a nite extension Ei of E, a nite unramied extension Ki of K, and
a crystalline character i : GKi Ei such that Ei E Vi is isomorphic to the induced
representation IndK
Ki i .
Proof. The if part of the claim is clear. We prove the only if part. Suppose that

V is semi-induced and crystalline. Then V is isomorphic to a direct sum i Vi of free
E-representations such that for each i, there exist a nite extension Ei of E and a nite
extensions Ki of K and a continuous character i : GKi Ei such that Ei E Vi
K
is isomorphic to the induced representation IndK
Ki i . Since V is crystalline, IndKi i is
crystalline for every i. This in particular shows that i is crystalline. Since the inertia
subgroup of GK trivially acts on Dpst (IndK
Ki i )Ei , the extension Ki /K is unramied.
This proves the claim. 2
Proof of Proposition 3.7. Below we prove the converse part. The other part is easier
to prove and follows by reversing the argument below.
741
Suppose that V is crystalline and semi-induced. By Lemma 3.10, V is isomorphic to

a direct sum
i Vi of free E-representations such that for each i there exist a nite
extension Ei of E and a nite unramied extensions Ki and a crystalline character
i : GKi Ei such that Ei E Vi is isomorphic to the induced representation IndK
Ki i .

Let us x i. By enlarging Ei if necessary, we may assume that the residual degree of
Ei /Qp is divisible by [Ki : Qp ]. Let rec : Ki Gab
Ki denote the reciprocity map in
the local class eld theory which sends a uniformizer in Ki to a lift of the geometric
Frobenius, and let I(Ki ) denote the set of embeddings Ki Ei of Qp -algebras. It
follows from the classication of crystalline characters that for each I(Ki ), there
exists an integer r Z such that the character i rec is equal to the tensor product
of an unramied quasi-character i : Ki Ei and the quasi-character which sends
x Ki to the product

(x)r Ei .
I(Ki )

The tensor product Ei Qp Ki is a product Ei Qp Ki = I(Ki ) Ei of copies of Ei
where x y Ei Qp Ki in the left hand side corresponds to ((x)y)I(Ki ) in the right
hand side. Let i denote the Frobenius automorphism of Ki . The composition with i1
induces a cyclic permutation of the set I(Ki ) which we denote by I(Ki ) : I(Ki ) I(Ki ).
With the above notation, the ltered -module Dcrys (i )Ei over Ei Qp Ki has the following description. For I(Ki ), let Dcrys (i ) denote the component at of Dcrys (i )Ei .
Then for j Z and for I(Ki ), the component at of Filj Dcrys (i )Ei is equal to
Dcrys (i ) for j r , and is equal to zero for j > r . For each I(Ki ), take an Ei -basis
1
d of Dcrys (i ) . Let I(Ki ) and put = I(K
(). Let b Ei denote the unique ele i)
ment satisfying (d ) = b d . Then we have I(Ki ) b = i (p). Since the character i

b
takes values in OE
i (p) is in OE
, the element p
. Hence by changing the basis d ,
i
i
r
we may assume that b p OE for each I(Ki ).
i
Then the claim follows from the observation that Dcrys (IndK
Ki i )Ei is equal to

Dcrys (i )Ei regarded as a ltered -module over Ei Qp K. 2
Denition 3.11. For a basic datum (w, r = (ri, ), = (i, )) DataK,E,n and for a
crystalline free E-representation V of GK , we say that V is nearly semi-induced of type
(w, r, ) if the ltered -module Dcrys (V )E over E Qp K associated to V has a basis
d1 , . . . , dn over E Qp K satisfying the following conditions:
(1) For each i Z, the submodule Fili Dcrys (V )E is equal to the set of elements
n
1
j=1 xj dj Dcrys (V )E such that the component of xj at () is equal to zero for
every (j, ) [1, n] with = 0 and i > rj, , and that the component of xw(j) at
1
(0 ) is equal to zero for every j [1, n] with i > rj,0 .

(2) Let m be the integer introduced in Section 2.4. (Note that we have m (r 1)/
(p 1).) Then for each there exists a matrix U GLn (W ) which is congruent
742
to the identity matrix modulo pm mE W such that the component at of the matrix
of the endomorphism of Dcrys (V )E with respect to the basis d1 , . . . , dn is equal to

U diag 1, pr1, , . . . , n, prn,
for = 0 and is equal to

U0 diag 1, pr1, , . . . , n, prn, C
for = 0 .
The following Proposition is an immediate consequence of Corollary 3.5.
Proposition 3.12. Let (w, r, ) DataK,E,n be an eective basic datum of degree n and
let V be a free E-representation of GK which is crystalline and nearly semi-induced of
type (w, r, ). Then there exists a tuple a Tuple such that the Wach module N(V ) over
(B+
K )E is isomorphic to the Wach module Na . 2
The following Corollary is an immediate consequence of the Proposition 3.12 and
Lemma 3.2.
Corollary 3.13. Let V be as in Proposition 3.12. Then the semi-simplication of the reduction mod mE of any GK -stable OE -lattice of V is isomorphic to the semi-simplication
of F OE T0 as a free F-representation of GK .
Remark 3.14. The condition U 1 (mod pm mE W ) in Denition 3.11 implies that, for
the representation V in Proposition 3.12 and Corollary 3.13, the characteristic polynomial ch(Frob) of the crystalline Frobenius Frob := [K:Qp ] on Dcrys (V )E is p-adically
close to the characteristic polynomial ch(Qw,r, ) of the matrix
Qw,r, := diag

1, pr1, , . . . ,

n, prn, w1 .
It is remarkable that for some two dimensional crystalline representations of GQp , for
which ch(Frob) are not p-adically close to ch(Qw,r, ) for any (w, r, ) DataQp ,E,2 , the
semi-simplication of its reduction mod mE is explicitly obtained in [BB] and [BG] by
a completely dierent method, based on the compatibility of the p-adic and the mod p
local Langlands correspondences. The latter method is eective for the other extreme
case where ch(Frob) are nearly p-adically farthest from ch(Qw,r, ) for any (w, r, ).
The limitations n = 2 and K = Qp in the latter method come from the fact that the
p-adic local Langlands correspondence has been established only under these conditions.
Hopefully they will be relaxed once the p-adic local Langlands correspondence will be
743
established for other cases. However no promising method seems to be known4 except
for these two extreme cases.
4. Local deformation rings and moduli of Wach modules
In this section, we study universal family of Wach modules, local deformation rings,
and subfamily corresponding to the family constructed in the previous section.
Let AROE denote the category of Artinian local OE -algebra with residue eld F. Let VF
be a free F-representation of GK of dimension n. We consider the functor DVF from AROE
the category of groupoids which sends an object A in AROE to the groupoid DVF (A) of
deformations of VF to A, that is, the groupoid DVF (A) such that the objects in DVF are
=
the pairs of a free A-representation VA of GK and an isomorphism VA A F
VF of free
F-representations of GK , and for two objects VA,1 and VA,2 of DVF (A), the morphisms
=
from VA,2 to VA,2 are the isomorphisms VA,1
VA,2 of free A-representations of GK
which are compatible with the given isomorphisms VA,i A F
= VF for i = 1, 2. Fix an
2
F-basis of VF . We also consider the functor DVF from AROE to the category of sets which
sends an object A in AROE to the set DV2F (A) of framed deformations (or, liftings in
the terminology of [CHT]) of VF to A, that is, the set DV2F (A) of isomorphism classes
of triples of a free A-representation VA of GK , an isomorphism VA A F
= VF of free
F-representations of GK , and a lifting to VA of the xed basis of VF . Here two such
=
triples VA,1 and VA,2 are called isomorphic if there exists an isomorphism VA,1
VA,2
which transports the given basis of VA,1 to the given basis of VA,2 . The functor DV2F is
pro-representable by a complete local Noetherian OE -algebra RV2F .
Take an object A in AROE , and let VA be an object of the groupoid DVF (A). Let
MA := D(VA ) be the tale (, K )-module over (AK )A associated to the A-linear dual
VA of VA . For a tuple r = (r ) of non-negative integers indexed by , let us consider
the functor WVA ,r from the category of A-algebras to the category of sets, which sends
an A-algebra B to the set of Wach lattices in MA A B with HodgeTate weights in the
tuple [0, r] of intervals. (see [K2, 3] for the denition of Wach lattices in MA A B).
By [K2, Proposition (3.5)] the functor WVA ,r has the following properties: The functor
WVA ,r is represented by a projective A-scheme WVA ,r . If R is any complete local
Noetherian ring with residue eld F, and VR a deformation of VF to R, then there exists
a projective R-scheme R : WVR ,r Spec R such that
(1) If A is in AROE and R A is a homomorphism of local OE -algebras inducing the
=
identity on residue elds, then there exists a canonical isomorphism WVR ,r R A
WVA ,r , where VA = VR R A.
4
After a submission of this paper, we found a way of extending our method by using the hypergeometric
polynomials when n = 2, K = Qp and ch(Frob) are between p-adically close to ch(Qw,r, ) and nearly
p-adically farthest from ch(Qw,r, ), which covers all the intermediate range if the width of Hodge ltration
2
is less than p 2+1 [YY]. We also expect a possibility that this new technique is generalized to the case where
n 2 and K is an unramied extension of Qp .
744
(2) The map R : WVR ,r Spec R becomes a closed embedding after inverting p. If A
is a nite E-algebra, then a homomorphism R A of OE -algebras factors through
the coordinate ring of the scheme theoretic image of R if and only if VA = VR R A
is a crystalline representation with HodgeTate weights in [0, r].
For R = RV2F , let RV2,r
denote the coordinate ring of the scheme theoretic image of
F
RV2 . For a nite local Qp -algebra A and a free A-representation VA of GK which is
F
crystalline with non-negative HodgeTate weights, we put

PVA (T ) := det KQp A T [K:Qp ] Dcrys VA ,
and let cVA ,i denote the coecient of T i in PVA (T ) for i = 0, . . . , n 1. It follows from the
argument in [K2, Corollary (3.7)] that the OE -algebra RV2,r
has the following properties.
F
(1) RV2,r
[1/p] is formally smooth over E.
F
(2) For i = 0, . . . , n 1, there is a unique ci RV2,r
[1/p] such that for any nite
F
E-algebra A, and any OE -algebra homomorphism h : RV2F A, the representation
VA obtained from the universal representation over RV2F by specialization by h satises cVA ,i = h(ci ). Moreover, the element ci is contained in the normalization of
Im{RV2,r
RV2,r
[1/p]}.
F
F
Lemma 4.1. For any nc-tuple r = (ri, )(i,)[1,n] of integers and for every , there
exists a unique quotient RV2,r
of RV2F with the following properties:
F
(1) RV2,r
is p-torsion free, and if RV2,r
= 0, then RV2,r
[1/p] is formally smooth over E
F
F
F

2
of pure dimension n + ([K : Qp ]/c) m where for , let m denote the
number of pairs (i, j) [1, n] [1, n] satisfying ri, > rj, .
(2) If E is a nite extension of E, x : RVF E a morphism of E-algebras, and Vx the
representation of GK obtained by specializing the universal RVF -representation by x,
then x factors through RV2,r
if and only if Vx is crystalline with HodgeTate weights
F
(r1,z() , . . . , rn,z() )E where z : E = E denotes the map induced by the
inclusion WOE WOE .
Proof. This is a special case of [K3, Theorem (3.3.8)].
Fix an eective basic datum (w, r, ) = (w, (ri, ), (i, )) DataK,E,n . We restrict
ourselves to the case where VF is the F-linear dual of FOE T0 where T0 is as in Lemma 3.2.
We x an F-basis of VF . Let a Tuple. By Lemma 3.2, we have a canonical isomorphism
F OE Ta
= VF . Let B be an OE -basis of the OE -dual Ta of Ta which is compatible with
the xed basis of VF under the above isomorphism F OE Ta
= VF . The pair (Ta , B) gives
2,r
a homomorphism RVF OE of local OE -algebras. Let xa,B : RV2,r
[1/p] E denote its
F
base change to E.
745
Proposition 4.2. The connected component of Spec RV2,r

[1/p] which contains the
F
E-rational point xa,B does not depend on the choice of the pair (a, B).
Proof. First we prove that the connected component does not depend on the choice
of B. Let (a, B1 ) and (a, B2 ) be two pairs with the common rst component. Then
there exists an element g in the kernel of GLn (OE ) GLn (F) such that B2 = gB1 ,
that is, the matrix of the identity endomorphism of Ta with respect to the basis B2
in the source and the basis B1 in the target is equal to g. We x a uniformizer of
E and write g = 1n + Y . Let R := OE [[t]] be the ring of formal power series with
coecients in OE . For an object A in AR(OE ) and for a homomorphism f : R A
of local OE -algebras, the pair (A OE Ta , (1 + f (t)Y )B1 ) denes an element in DV2F (A).
The association f (A OE Ta , (1 + f (t)Y )B1 ) yields a homomorphism : RVF R
of complete local OE -algebras. We claim that the homomorphism factors through
the quotient RV2,r
and hence induces a homomorphism : RV2,r
R. For a m,
F
F
let fa : R E denote the continuous homomorphism of OE -algebras which sends
T to a. It follows from the construction of the homomorphism that the composite
fa factors through the quotient RV2,r
for any a m. Then the claim follows since
F

Weierstrass preparation theorem implies that the homomorphism R am E given
by fa is injective. It is clear that the homomorphism xa,B1 (resp. xa,B2 ) is equal to the
composite
RV2,r
[1/p] R[1/p] E
F
where the second map is the continuous homomorphism which sends t to 0 (resp. 1).
Since Spec R[1/p] is irreducible, the two points xa,B1 and xa,B2 lie in the same irreducible
component of Spec RV2,r
[1/p].
F
Let R be a quotient of the ring OE [[X]] = OE [[Xi,j, (i, j [1, n], )]] such
that R is a nite OE -algebra. Let sR : OE [[X]] R denote the canonical surjection.
Let P (X)R and G,R denote the matrices obtained by applying the homomorphism
+
+
id sR : (A+
K )OE [[X]] = OE [[X]] Zp AK R Zp AK to the entries of P (X) and G
respectively. Then the pair (P (X)R , (G,R ) ) gives a structure of tale (, K )-module
on (RZp AK )n . Let TR denote the free R-representation of GK associated to this tale
(, K )-module. Let TR denote the R-linear dual of TR . The collection (TR )R , where R
runs over the quotients of the ring OE [[X]] which is nite over OE , forms a projective
system. We put TO E [[X]] =
lim
T and take an OE [[X]]-basis BOE [[X]] which lifts the
R R
given basis of TF = VF . The pair (TO E [[X]] , BOE [[X]] ) gives a framed deformation of
VF to OE [[X]]. We put r,max = maxi[1,n] ri, for and rmax = (r,max ) .
It follows from the construction of the OE [[X]]-scheme W = WTO [[X]] ,rmax that the
E
matrices P (X) and G give an OE [[X]]-rational point of W . Hence the homomorphism
max
RV2F OE [[X]] which comes from the universality factors through the quotient RV2,r
.
F
Since the set of E-rational points of a rigid analytic open unit disc over E is not contained
in any proper analytic closed subvariety, a technique similar to that in the proof of the
746
claim in the previous paragraph shows that the homomorphism RVF OE [[X]] factors
through the quotient RV2,r
. Since the scheme Spec OE [[X]][1/p] is irreducible, the claim
F
follows. 2
5. Applications to automorphy lifting theorem and potential automorphy theorem
Denition 5.1. We say that a crystalline free E-representation V of GK is absolutely
nearly semi-induced if there exist a nite extension E of E and a basic datum (w, r, ) =
(w, (ri, ), (i, )) DataK,E ,n such that E E V is nearly semi-induced of type (w, r, ).
We say that a free E-representation V of GK is upper-triangular if there is a ltration
on V by E-subrepresentations whose graded pieces are one-dimensional.
Proposition 4.2 together with Lemma 1.4.1 and Lemma 1.4.3 of [BLGGT] gives the
following result.
Proposition 5.2. Let V be a free E-representation of GK which is either crystalline
and absolutely nearly semi-induced or potentially crystalline and upper-triangular. Then
V E OQp is potentially diagonalizable in the sense of [BLGGT, 1.4]. 2
This proposition, together with the main results of [BLGGT, Theorem 4.2.1, Corollary
4.5.2], we have the following automorphy lifting and potential automorphy results.
Corollary 5.3. Let F be an imaginary CM eld and let F + denote its maximal totally
real subeld. Let n 1 be an integer and let p be a prime number such that F is
unramied at p, and that F does not contain a primitive p-th root of unity. Fix an
=
isomorphism : Qp
C. Let E be a nite extension of Qp and let (r, ) be the pair of
a continuous absolutely irreducible representation r : GF GLn (E) and a continuous
character : GF + E such that rc is isomorphic to r where c Gal(F/F + )
denotes the complex conjugation. Let r denote the reduction of r and assume that the
following conditions are satised:
(1) r is unramied outside a nite set of primes.
(2) The restriction of r to GF (p ) is irreducible.
(3) p 2(d + 1) where d is the maximal dimension of an irreducible subrepresentation
of the restriction of r to the closed subgroup of GF generated by all Sylow pro-p
subgroups.
(4) For all prime v of F dividing p, the restriction r|GFv is potentially crystalline with
distinct HodgeTate weights and is either crystalline and absolutely nearly semiinduced or upper-triangular.
(5) There is a RAECSDC automorphic representation (, ) of GLn (AF ) whose associated pair (rp, (), rp, ()) of a p-adic Galois representation rp, () and a p-adic
character rp, () satises the following conditions:
747
(r, ) is isomorphic to (rp, (), rp, ()).

either is -ordinary in the sense of [G, Denition 5.1.2.], or for every v|p, the
restriction rp, ()|GFv is either crystalline and absolutely nearly semi-induced or
potentially crystalline and upper-triangular.
Then (r, ) is automorphic of level potentially prime to p.
Corollary 5.4. Let F + be a totally real eld. Let n 1 be an integer. Let p be a prime
number such that p 2(n + 1) and p is unramied in F + . Let E be a nite extension
of Qp and let (r, ) be the pair of a continuous absolutely irreducible representation
r : GF + GLn (E) and a continuous character : GF + E such that the pair (r, )
is a totally odd, essentially conjugate self dual representation. Let r denote the reduction
of r and assume that the following conditions are satised:
(1) r is unramied outside a nite set of primes.
(2) The restriction of r to GF + (p ) is irreducible.
(3) For all prime v of F + dividing p, the restriction r|GF + is potentially crystalline with
v
distinct HodgeTate weights and is either crystalline and absolutely nearly semiinduced or upper-triangular.
Then there exists a Galois, totally real extension F + /F + such that restriction
(r|GF + , |GF + ) is automorphic of level prime to p.
Acknowledgments
This paper is a generalized version of the preprint (2008) of the rst author. He thanks
the second author for discussions and for extending the construction of the family of Wach
modules in the previous version, and for checking and correcting the other part. (He also
thanks him for precisely citing the recent preprints to remove some parts of the previous
version. Especially, by citing [BLGGT], we could remove the part showing Rred = T for
GLn case.) He thanks Ivan Fesenko in Nottingham University for the hospitality from
April/2008 to March/2010.
He sincerely thanks TOYOTA Central R&D Labs., Inc. for oering him a special
position in which he can concentrate on pure mathematical research. The president of
the company at that time told him that if they forced him to do something else, then
it would be against the policy of the company, and it would be against the philosophy
of the founder of the company as well. He also sincerely thanks Sakichi Toyoda for his
philosophy, and the executives for inheriting it from him for 80 years after his death.
References
[BLGGT] T. Barnet-Lamb, T. Gee, D. Geraghty, R. Taylor, Potential automorphy and change of weight,
Ann. of Math. (2) 179 (2014) 501609.
748
[BLGHT] T. Barnet-Lamb, D. Geraghty, M. Harris, R. Taylor, A family of CalabiYau varieties and

potential automorphy II, Publ. Res. Inst. Math. Sci. 47 (1) (2011) 2998.
[B1]
L. Berger, Limites de reprsentations cristallines, Compos. Math. 140 (2004) 14731498.
[B2]
L. Berger, Reprsentations modulaires de GL2 (Qp ) et reprsentations galoisiennes de dimension 2, Astrisque 330 (2010) 263279.
[B3]
L. Berger, Local constancy for the reduction mod p of 2-dimensional crystalline representations, Bull. Lond. Math. Soc. 44 (3) (2012) 451459.
[BB]
L. Berger, C. Breuil, Sur la rduction des reprsentations cristallines de dimension 2 en poid
moyens, unpublished note, which is contained in [B2].
[BLZ]
L. Berger, H. Li, H.J. Zhu, Construction of some families of 2-dimensional crystalline representations, Math. Ann. 329 (2) (2004) 365377.
[BG]
K. Buzzard, T. Gee, Explicit reduction modulo p of certain two-dimensional crystalline representations, Int. Math. Res. Not. IMRN (12) (2009) 23032317.
[CHT]
L. Clozel, M. Harris, R. Taylor, Automorphy for some -adic lifts of automorphic mod Galois
representations, Publ. Math. Inst. Hautes Etudes Sci. 108 (2008) 1181.
[D]
G. Dousmanis, On reductions of families of crystalline Galois representations, Doc. Math. 15
(2010) 873938.
[F]
J.-M. Fontaine, Reprsentations p-adiques des corps locaux I. The Grothendieck Festschrift,
vol. II, Progr. Math., vol. 87, Birkhuser Boston, Boston, MA, 1990, pp. 249309.
[GL]
H. Gao, T. Liu, A note on potential diagonalizability of crystalline representations, preprint,
http://arxiv.org/abs/1204.4704.
[G]
D. Geraghty, Modularity lifting theorems for ordinary Galois representations, PhD thesis,
Harvard University, ISBN 978-1124-07900-4, 2010, 131 pp., ProQuest LLC.
[HSBT] M. Harris, N. Shepherd-Barron, R. Taylor, A family of CalabiYau varieties and potential
automorphy, Ann. of Math. 171 (2010) 779813.
[K2]
M. Kisin, Modularity for some geometric Galois representations, in: L-Functions and Galois
Representations, Durham, 2004, pp. 438470.
[K3]
M. Kisin, Potentially semistable deformation rings, J. Amer. Math. Soc. 103 (2008) 513546.
[K1]
M. Kisin, Moduli of nite at group schemes and modularity, Ann. of Math. 170 (2009)
10851180.
[T]
R. Taylor, Automorphy for some -adic lifts of automorphic mod Galois representations II,
Publ. Math. Inst. Hautes Etudes Sci. 108 (2008) 183239.
[YY]
G. Yamashita, S. Yasuda, Reduction of two dimensional crystalline representations and Hypergeometric polynomials, in preparation.
[Z]
H.J. Zhu, Crystalline representations of GQpa with coecients, preprint, http://arxiv.org/
abs/0807.1078.

Integral Padic Hodge

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Integral Padic Hodge

Hochgeladen von

Copyright:

Verfügbare Formate

Journal of Number Theory 147 (2015) 721748

Contents lists available at ScienceDirect

Journal of Number Theory

On some applications of integral p-adic Hodge

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

2.5. The matrix G . . . . . . . . . . . . . . . . . . . . . . . .

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

p-adic representations (depending on a choice of some extra parameters w, r, which we

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

K := Gal(K(p )/K) and let : K Zp denote the p-adic cyclotomic character.

This gives an action of the group K on the K-algebra K[[]].

denote the p-adic completion of the ring A+

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

The module N(V ) is free of rank d over (B+

Moreover, if V is a free E-representation of GK which is crystalline with HodgeTate

Dcrys (V )E of ltered -modules with coecient in E. Here we endow N(V )/N(V )

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

(1) The (A+

Lemma 2.2. The sequence {i (1 + )}i0 converges to 1 in A+

i ((1 + )j ), the claim follows from Lemma 2.2. 2

We dene a subring RQp of Qp [[]] as follows:

of ideals in RQp forms a fundamental system of neighbourhoods of 0. We easily see

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

Then we have ((f )) = (Sf ).

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

We write zi,j, = pm zi,j, and put

which is, by the denition of m, an element in A+

(1n + Z0 ) diag(1,0 q r1,0 , . . . , n,0 q rn,0 ) C, if = 0 ,

2.5. The matrix G

For , we let G Mn (RQp ) denote the matrix

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

Proof. Let i [1, n]. We have fw(i,)

For and for K , we let G, denote the n-by-n matrix

in A+ . We put G = (G, ) which we regard an element in Mn ((A+

Lemma 2.9. We have

On the other hand, the equalities

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

Thus we have G = P (0)(G )(P (0))1 , which proves the claim. 2

Proposition 2.10. The entries of the matrix

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

For h = 0, . . . , c 1, it follows from the denitions of G and Z that the component at

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

Then, for i := i + nc, we obtain

nc (xj )0nc (Nj ).

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

By applying (2.6) repeatedly, we have

since we can dene G as the limit of G . We construct such a G by induction on k.

Let us consider an innite sum

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

3. The representation of type (w, (ri, ), (i, ))

For a = (ai,j, ) Tuple, we dene an OE -algebra homomorphism eva : (A+

Let Va be the free E-representation of GK corresponding to the tale (, K )-module

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

Corollary 3.5. For each integer i 0, the set

xw(j),1 (0 ) irj,0 (B+

3.1. Semi-induced representations

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

G. Yamashita, S. Yasuda / Journal of Number Theory 147 (2015) 721748

Suppose that V is crystalline and semi-induced. By Lemma 3.10, V is isomorphic to

(0 ) is equal to zero for every j [1, n] with i > rj,0 .