ORDINARITY OF LOCAL GALOIS REPRESENTATION

ARISING FROM DWORK MOTIVES

arXiv:2103.00106v1 [math.NT] 27 Feb 2021

LIE QIAN

1. Introduction
Let F be a characteristic 0 local field containing ζN whose residue characteristic
equals to p and does not divide N . We first introduce the Dwork motives.
Let T0 = P1 − ({∞} ∪ µN )/Z[1/N ] with coordinate t and Z ⊂ PN −1 × T0 be a
projective family defined by the following equation:
X1N + X2N + · · · + XN
N
= N tX1 X2 · · · XN
The map π : Z → T0 is a smooth of relative dimension N − 2. We will write Zs
for the fiber of this family at a point s. Let H = µN
N /µN and

H0 := {(ξ1 , . . . , ξN ) ∈ µN
N : ξ1 · · · ξN = 1}/µN ⊂ H

Over Z[1/N, ζN ] there is an H action on Z by:

(ξ1 , . . . , ξN )(X1 , . . . , XN , t) = (ξ1 X1 , . . . , ξN XN , (ξ1 · · · ξN )−1 t)
Thus H0 acts on every fibre Zs , and H acts on Z0 .
Fix χ a character H0 → µN of the form:
N
Y
χ ((ξ1 , . . . , ξN )) = ξiai
i=1
P
where (a1 , . . . , aN ) are N constants such that Ni=1 ai ≡ 0 mod N . Therefore
the character is well-defined.
We define the Dwork motive to be given by Z and the χ-eigenpart of the H0
group action. In concrete terms, its p-adic realization is defined below.
For any prime λ of Z[1/2N, ζN ] of residue characteristic p, we define the lisse
sheaf Vλ /(T0 × SpecZ[1/2N p, ζN ])et by:
Vλ = (RN −2 π∗ Z[ζN ]λ )χ,H0
here the χ, H0 in the supscript means the χ-eigenpart of the H0 action.
We let Vλ,t denote the fibre of the sheaf Vλ over t for a t ∈ T0 (F ′ ), where F ′ is
a finite extension of F . In other words, viewed as a GF ′ representation, Vλ,t is just
H N −2 (Zt , Z[ζN ]λ )χ,H0 , here t is the corresponding geometric point of t.
Fix the embedding τ : Q(ζN ) ֒→ C such that τ (ζN ) = e2πi/N . Let π̃ : Y (C) →
T0 (C) denote the base change of π along τ and VB be the locally constant sheaf
over T0 (C):
VB = (RN −2 π̃∗ Z[ζN ])χ,H0

Date: January 25th, 2021.

1
2 LIE QIAN

By standard comparison results(see e.g. [BLGHT09]), Vλ and VB are locally

constant and locally free of the same rank over Z[ζN ]λ and Z[ζN ]. This rank can
be computed by looking at the fibre over 0. Denote this rank by n.
Fix a nonzero base point s ∈ T0 (C). Now we have the monodromy representation:
ρs : π1 (T0 (C), s) → GL(VB,s )
Let γ∞ be the loop around ∞ as an element of π1 (T0 (C), s).
We can now state the main theorem of this paper.
Theorem 1.1. Under the assumption that for the motives defined by any fixed σχ,
where σ ∈ Gal(Q(µN )/Q) is arbitrary, ρs (γ∞ ) has minimal polynomial (X − 1)n ,
i.e. it is maximally unipotent, we have that the GF0 representation Vλ,t−1 is regular
ordinary for any finite extension F0 of F and t ∈ F0 with v(t) > 0, where v is the
valuation of F0 .
Remark 1.2. Note that by the argument in Page 12-13 of the author’s forthcoming
work Potential Automorphy for GLn , the assumption that ρs (γ∞ ) has minimal
polynomial (X − 1)n is satisfied when 0 ∈ {a1 , . . . , aN }.
The main application of our theorems would be in the proof of a potential auto-
morphy theorem claimed in a forthcoming paper by the author. There we choose
certain (a1 , . . . , aN ) so that the input condition of main theorem is satisfied. In
that paper we need certain local Galois representation to be regular and ordinary
in order to apply the automorphy lifting theorem from [ACC+ 18].
We shall briefly discuss the idea of the proof of the theorem. Since we may
replace F by F0 at the very beginning, we assume without loss of generality that t
is an F point T0 from now on.
The theorem will be implied by the following two theorems this paper is going
to establish, each possibly interesting in its own right.
The first theorem claims the existence of a semistable model for Zt with t of
negative valuation over a finite extension F ′ /F . The semistable model comes from
a series of blowup of the naive integral model given by the equation t−1 (X1N +X2N +
N
· · ·+ XN ) = N X1 X2 · · · XN in PN −1 × OF ′ for some extension F ′ /F . We first work
with the fundamental case where t is of (normalized) valuation −1. Then we use
the technique of toroidal embedding (in a mixed charateristic setting) established
by Mumford to go from this case to the general case. This is the main result of
section 2.
The second theroem gives a way to compute the log cristalline cohomology
∗
Hcrys ((Y1 , N1 )/(W (k), N, 1 7→ 0))
(See section 3 for a brief recall of the notation concerning log geometry) with its
N operator in terms of log de-Rham cohomology under the setting that (Y1 , N1 )
is log smooth over (k, N, 1 7→ 0) and can be lifted into a family with log structure
(Y, N )/(W (k)[T ], N, 1 7→ T ). Indeed, we use the log de Rham cohomology of the
characteristic 0 lift given by the fibre over 0 of (Y, N ) and the operator N is given
as the connecting homomorphism of certain long exact sequence that is similar to
the one used to define the Gauss-Manin connection. This is the first main result of
section 3.
From now on, when there are no ambiguity of the residue field k, we will use Wn
to denote Wn (k) and W to denote W (k).
We will now explain these 2 results in more detail.
 ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES3

1.1. Semistable Model. Let us recall what we mean by semistable throughout

this paper, following [HK94].
Definition 1.3. We say a scheme X over a discrete valuation ring A is with
semistable reduction if etale locally on X, there is a smooth morphism X →
A[T1 , . . . , Tr ]/(T1 · · · Tr − π) for some r ≥ 0, where π is a uniformizer of the ring A.
Theorem 1.4. For any t ∈ F such that v(t) = d > 0, there exists a totally ramified
extension F ′ /F generated by π 1/e , a choice of e-th root of π which is a uniformizer
of OF (thus F ′ is purely ramified of degree e over F with uniformizer π 1/e ), and a
semistable model Y over OF ′ of Zt−1 with compatible H0 action.
Here is a supplement of the above theorem. It gives some idea on how the
semistable model Y is constructed, and will be used as a link to reduce our compu-
tation to the situation over the unramified base ring W , where Hyodo-Kato’s log
geometry technique could be applied.
Let Spec W [S, U ±1 ]′ denote the open subscheme of Spec W [S, U ±1 ] defined by
(S U )N 6= 1 and let W [S, U ±1 ]′ denote the ring of regular function of this scheme.
de

Proposition 1.5. In the setting of the above theorem, we can actually find a
variety Z with H0 action over Spec W [S, U ±1 ]′ that is a blowup of the variety
U S de (X1N + X2N + · · · + XN N
) = N X1 X2 · · · XN over Spec W [S, U ±1 ]′ (the latter is
the base change of the previous U T (X1N + X2N + · · · + XN N
) = N X1 X2 · · · XN along
±1 ±1 ′ de
W [T, U ] → W [S, U ] , T 7→ S ) that is an isomorphism outside the closed
subscheme defined by S, such that locally Z admits an etale map to
±1
(1.1) W [U ±1 , S, Z1 , . . . , Zr , Zr+1 , . . . , Zn±1 ]/(U S − Z1 · · · Zr )
over W [S, U ±1 ]′
Choosing a uniformizer π of OF and write t = uπ d with u ∈ OF× , the Y in
the above theorem is obtained from base change of Z along W [S, U ±1 ]′ → OF ′ ,
S 7→ π 1/e , U 7→ u.
The link to a model over W will be stated in Remark 2.4.
1.2. Log Geometry. A general log scheme is usually denoted as (Z, M ) for a
scheme Z and a a sheaf of monoid M . However, we would sometime use a third
argument after M to show what the structure map M → OZ is. Again, we will
use Spec W [T ]′ to denote a fixed choice of affine open subscheme of Spec W [T ]
such that the closed subscheme defined by T = 0 is contained in Spec W [T ]′ . Let
Spec Wn [T ]′′ be the mod pn reduction of Spec W [T ]′ .
Let (Y, N ) be a log smooth scheme over (W [T ]′ , N, 1 7→ T ). Denote by (Yn , Nn )
(resp. (Yn , Nn )) the base change of (Y, N ) along (Wn [T ]′ , N, 1 7→ T ) → (W [T ]′ , N, 1 7→
T )(resp. (Wn , N, 1 7→ 0) → (W [T ]′ , N, 1 7→ T )). Let (Y , N ) be the base change of
(Y, N ) along (W, N, 1 7→ 0) → (W [T ]′ , N, 1 7→ T ). Note that the closed immersions
in the corresponding fibre diagram are all exact and the projection maps to the
base are all log smooth.
Hyodo-Kato define the i-th log cristalline cohomology of (Y1 , N1 ) as the limit
!

 1
(1.2) lim H i (Y1 , N1 )/(Wn , N, 1 7→ 0) crys , OY1 /Wn [ ]
←− p
n

and equip it with a nilpotent operator N .

4 LIE QIAN

Theorem 1.6. The above limit is isomorphic to Hi (dRY /(W,N,17→0) )[ p1 ] while the
operator N is given as the degree i boundary homomorphism of the long exact
sequence given by the following exact triangle
·dlog1
dRY /(W,N,17→0) [−1] dRY /(W,(0)) dRY /(W,N,17→0)
which is defined at each degree i by taking the i-th wedge power of locally split
exact sequence of locally free sheaves of modules
·d log 1
0 OY ωY1 /(W,(0)) ωY1 /(W,N,17→0) 0

given in Theorem 3.2.3 of [Ogu18] since we note that (Y , N ) is log smooth over
1
(W, N, 1 7→ 0) and we identify ω(W,N,17 ∼
→0)/(W,(0)) = W by d log 1 7→ 1.
Remark 1.7. In the setting of the above theorem, note also that the resulting exact
sequence
·d log 1
0 ωYi−1
/(W,N,17→0)
ωYi /(W,(0)) ωYi /(W,N,17→0) 0

at degree i is a locally split exact sequence of locally free sheaves of modules.

Finally, let (Y C , N C ) be a proper log scheme smooth over (C, N, 1 7→ 0) and
(Y an , N an ) be the analytic log scheme associated to it which is also smooth over
the analytic point (pt, N, 1 7→ 0). We would like to reduce the computation to the
analytic setting by the following theorem. It will be proved by GAGA.
Theorem 1.8. There exists an N equivariant isomorphism
Hi (dRY C /(C,N,17→0) ) ∼
= Hi (dRY an /(pt,N,17→0) )
where the N on each hypercohomology is defined by degree i boundary morphism of
the exact triangles obtained similar to the one in Theorem 1.6
·dlog1
dRY C /(C,N,17→0) [−1] dRY C /(C,(0)) dRY C /(C,N,17→0)
·dlog1
dRY an /(pt,N,17→0) [−1] dRY an /(pt,(0)) dRY an /(pt,N,17→0)

Acknowledgements. I would like to first thank Richard Taylor for encouraging

me to think about the subject of this paper. I also want to thank him for all the
helpful comments on the draft of this paper. I am grateful to Richard Taylor, Brian
Conrad, Weibo Fu, Ravi Vakil, Bogdan Zavyalov for many interesting conversation
during the preparation of this text.

2. Existence of Semistable Blowup

In this section we always assume char(k) ∤ N and N > 1. Again, we will use Wn
to denote Wn (k) and W to denote W (k)
We first prove a lemma that basically settles the case where t is a uniformizer in
1.4 via base change along W [T, U ±1 ]′ → OF , T 7→ t, U 7→ u, where Spec W [T, U ±1 ]′
is the open subscheme of Spec W [T, U ±1 ] defined by (T U )N 6= 1. The idea of this
blowup process comes from a construction of Nick Shepherd-Barron.
Lemma 2.1. There exists a blowup X of the variety U T (X1N + X2N + · · · + XN
N
)=
±1 ′
N X1 X2 · · · XN over Spec W [T, U ] with an action of H0 such that the blowdown
map is H0 -equivariant and etale locally X admits an etale map to
±1
W [T, U ±1 ]′ [Z1 , . . . , Zr , Zr+1 , . . . , Zn±1 ]/(T U − Z1 · · · Zr )
ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES5

over Spec W [T, U ±1 ]′

Proof. The sequence of blowup is as the following. Initially we have divisors Di

defined by Xi and D defined by X1N + X2N + · · · + XNN
. Denote the original variety
as YN , we blowup along D ∩ DN to get a variety YN −1 , as well as the divisors
(N −1)
Di and D(N −1) as the strict transform of Di and D respectively. In general,
(j)
by induction we will have the variety Yj at the (N − j)-th stage, with divisors Di
(j+1)
and D(j) as the strict transform of Di and D(j+1) , ∀i ∈ {1, . . . , n}, then we
(j) (j) (j−1)
blowup along D ∩ Dj , we get the variety Yj−1 with the divisors Di and
(1)
D(j−1) . The final step is that we let X := Y0 be the blowup of Y1 along D(1) ∩ D1 .
Note that for each blowup Yj−1 → Yj , we can associate an H0 action on Yj−1 such
that the morphism is H0 -equivariant because we may see by induction that the
(j)
locus D(j) ∩ Dj we are blowing up along is H0 -stable.
We check the desired local property of Y0 by hand. Fix an i ∈ {1, . . . , N } from
now on and we will only work with the preimage of the affine chart Xi 6= 0 and
we will still use Yk to denote the primage of this chart in the original Yk . In each
step of the blowups, we might write the blowup as the union of two affine open
subschemes given by the complement of the strict transform of the two divisors
whose intersection we are blowing up along. And note that the affine open given by
the complement of D(j) will be isomorphic to its preimage under all further blowup,
since it has empty intersection with the locus we are blowing up along.
More precisely, let l be the function defined on {1, . . . , î, . . . , N } by the rule
l(x) = x + 1 if x 6= i − 1 while l(i − 1) = i + 1. For 0 < k ≤ N with k 6= i, let Uk
denote the variety
Y
Spec (W [T, U ±1 ]′ [x1 , . . . , x̂i , . . . , xN , b′k , bl(k) ]/(N b′k xj − U T,
1≤j≤k−1,j6=i

Y X
b′k bl(k) − xk , bl(k) xj − xN
j − 1)
l(k)≤j≤N,j6=i 1≤j≤N,j6=i

X
here xj are the affine coordinates Xji , with the understanding that when k = N ,
Q P
the product l(k)≤j≤N,j6=i xj = 1, and thus bl(N ) = 1≤j≤N,j6=i xN j + 1.
And for 0 < k < N with k 6= i, let Vk denote the variety
Y
Spec (W [T, U ±1 ]′ [x1 , . . . , x̂i , . . . , xN , bl(k) ]/(N xj − U T bl(k) ,
1≤j≤k,j6=i

Y X
bl(k) xj − xN
j − 1)
l(k)≤j≤N,j6=i 1≤j≤N,j6=i

It can be seen inductively for k, the following properties hold.

• Yk = Vk ∪ ∪l(k)≤j≤N,j6=i Uj as affine opens for 0 ≤ k ≤ N, k 6= i, Yi = Yi+1 .
(k)
• for any m ≤ k Dm is the divisor given by xm in the affine open Vk and Uj
for all l(k) ≤ j ≤ N, j 6= i.
(k)
• for m ≥ l(k), Dm is the divisor given by b′m in Um and xm in Uj for all
j > m and it has empty intersection with the rest of the affine opens.
• D(k) is given by the divisor bl(k) in the affine open Vk only.
6 LIE QIAN

Thus the blowup is an isomorphism over each Uk in the affine charts of Yj and
maps Vj−1 ∪ Uj to Vj in the next step. From this, it suffice to verify all Uk satisfy
the local property. Now Uk can be written as
Y
Spec W [U ±1 , x1 , . . . , x̂i , xˆk , . . . , xN , b′k , bl(k) ]/(bl(k) xj −
l(k)≤j≤N,j6=i
X
xN ′ N
j − (bk bl(k) ) − 1)
1≤j≤N,j6=i,k
We first prove Uk is regular by applying Jacobian criterion to this variety. We
fix a closed point q of this variety giving rise to a k-point of the form
x1 7→ a1 , . . . , xN 7→ aN , b′k 7→ u, bl(k) 7→ v, U 7→ w,
for some constant aj , u, v, w ∈ k. Let m be the maximal ideal in
W [U ±1 , x1 , . . . , x̂i , xˆk , . . . , xN , b′k , bl(k) ]. It suffice to check that the defining equa-
tion is not 0 in m/m2 ⊗k(m) k, which has a basis dp, dx1 , . . . , dxN , db′k , dbl(k) , dU .
Let f be the defining equation
Y X
bl(k) xj − xN ′ N
j − (bk bl(k) ) − 1
l(k)≤j≤N,j6=i 1≤j≤N,j6=i,k
P
of Uk . Then df = j6=i,k fj dxj + fu db′k + fv dbl(k) where

fj = −N aN −1
if j < k
Q j
fj = v l(k)≤s≤N,s6=i,j as − N aN
j
−1
if j ≥ l(k)
N −1 N
fu = −N u v Q
fv = −N uN v N −1 + l(k)≤s≤N,s6=i as
If all of the coefficients are 0, we see that aj = 0, ∀j < k from fj = 0 since l ∤ N .
From fu = 0 we see u = 0 or v = 0, either implies by fv = 0 that as = 0 for some
l(k) ≤ s ≤ N, s 6= i. This implies aj = 0, ∀j 6= i, k, s by the fj = 0, which in turn
implies as = 0 by the fs = 0. But substituting the known zeroes into f we see that
this can never happen. Thus Uk is regular.
To verify the local property, fix a Uk to work in.
(1) If fj 6= 0 for some j ≥ l(k), then we claim that the map
Uk → W [U ±1 , x1 , . . . , x̂i , xˆj , xˆk , . . . , xN , b′k , bl(k) ] given by corresponding coordinate
is etale near the point q. Let n denote the maximal ideal corresponding to the image
of q under this map. Then the claim follows because
m/m2 ⊗k(m) k ∼ = kdp ⊕ kdx1 ⊕ · · · ⊕ kdxN ⊕ kdb′k ⊕ kdbl(k) ⊕ kdU/
 
X
 fj dxj + fu db′k + fv dbl(k) 
j6=i,k

∼ [ ⊕ kdx ⊕ kdb′ ⊕ kdb

= kdp ⊕ kdx1 ⊕ · · · ⊕ kdxj N k l(k) ⊕ kdU
∼ n/n ⊗k(n) k
= 2

here the middle isomorphism follows because fj 6= 0, and
we define the struc-
ture map Spec W [U ±1 , x1 , . . . , x̂i , xˆj , xˆk , . . . , xN , b′k , bl(k) ] → W [T, U ±1 ]′ by T 7→
Q
N U −1 b′k 1≤m≤k−1,m6=i xm , so that the morphism is a morphism of W [T, U ±1 ]′
scheme.
 ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES7

(2) If fv 6= 0, then same argument as that in (1) gives that the map Uk →
W [U ±1 , x1 , . . . , x̂i , xˆk , . . . , xN , b′kQ
] is etale near at the point q. And again define the
structure map by T 7→ N U −1 b′k 1≤m≤k−1,m6=i xm .
Note now that if we try to use similar argument to "kill" the variable xj for some
j < k, then it is hard to define a structure map on the target scheme because xj
appears in the expression of T in the original Uk . Hence we use the following trick.
(3) If aN j 6= (uv)N for some j < k, then we consider the map
Uk → W [U ±1 , x1 , . . . , x̂i , xˆj , xˆk , . . . , xN , bˆ′k , bl(k) , cj ] where every variable maps to
the corresponding one in the structure ring of Uk except cj 7→ b′k xj , thus dcj 7→
udxj + aj db′k under the pullback of the map mentioned above. The condition
aN
j 6= (uv)
N
gives that
 
u aj
fj fu
is nondegenerate, and hence

m/m2 ⊗k(m) k ∼
= kdp ⊕ kdx1 ⊕ · · · ⊕ kdxN ⊕ kdb′k ⊕ kdbl(k) ⊕ kdU/
 
X
 fj dxj + fu db′k + fv dbl(k) 
j6=i,k

∼ [ ⊕ kdx ⊕ kdb
= kdp ⊕ kdx1 ⊕ · · · ⊕ kdx
d′ ⊕ kdb ′
j N k l(k) ⊕ kdU ⊕ k(udxj + aj dbk )
∼ n/n2 ⊗k(n) k
=

where n denote the maximal ideal corresoonding to the image of q under the map.
And we may take the structure map W [U ±1 , x1 , . . . , x̂i , xˆj , xˆk , . . . , xN , bˆ′k , bl(k) , cj ] →
Q
W [T, U ±1 ]′ to be T 7→ N U −1 cj 1≤m≤k−1,m6=i,j xm . It could be checked that the
map is a map of W [T, U ±1 ]′ scheme.
We conclude that this exhausts all possibilities if we impose the condition that
q lies in the closed subscheme defined by T : If all of the three conditions above do
not hold, then aN j = (uv)
N
for all j < k, fj = 0 for all j ≥ l(k) and fv = 0.
We claim for all j > l(k), aj 6= 0: If aj = 0 for some j ≥ l(k), then checking the
condition fj ′ = 0 for all j ′ ≥ l(k), j ′ 6= j, gives aj = 0 for all j ≥ l(k). Putting
that into fv = 0 gives u = 0 or v = 0. Putting that once more into the condition
aNj = (uv)
N
gives aj = 0 for all j < k. Thus, we see that all fj , j 6= i, k, fu and
fv are 0. This have been shown to be impossible by the argument for regularity,
hence we prove the claim that for all j > l(k), aj 6= 0.
1 Q
Multiplying each fj = 0 by aj gives aN N
l(k) = · · · = aN = N v l(k)≤s≤N,s6=i as and
thus v 6= 0. Hence it also follows from fv = 0 that u 6= 0. The condition aN j = (uv)
N
−1 ′
Q
then gives aj 6= 0 for all j < k. Therefore, T = N U bk 1≤m≤k−1,m6=i am is not
0 at q because u, am for 1 ≤ m ≤ k − 1, m 6= i are all nonzeron as we have seen.
Similar consideration shows that for characteristic 0 points, the only peculiartiy
can happen in the open locus T 6= 0.
Now over the locus T 6= 0, the blowdown map is an isomorphism since the locus
we are blowing up against is always a codimension 1 irreducible subvariety. Hence
it suffices to show that the original projective variety U T (X1N + X2N + · · · + XN N
)=
N X1 X2 · · · XN admits such an etale map locally. Again we only work with the
affine chart Xi 6= 0 and writing the intersection with this chart as the affine variety
8 LIE QIAN

U0
Y
Spec W [T, U ±1 ]′ [x1 , . . . , x̂i , . . . , xN ]/(U T (1 + xN ˆN N
1 + · · · + xi + · · · + xN ) − N xj )
j6=i

Take a geometric point q with coordinates xj sent to bj in some algebraically closed

field. Denote the defining equation by g and its derivatives with respect to each
variable xj evaluated at q as gj . Similar to the argument in (1), the map U0 →
Spec W [T, U ±1 ]′ [x1 , . . . , x̂i , xˆj , . . . , xN ] is etale at q if gj 6= 0. Hence such a map
(clearly over W [T, U ±1 ]′ ) exist as long as one of the gj 6= 0.
Q
If all gj = 0, then U T bN j
−1
= k6=j,i bk for any j 6= i. In other words, bN 1 = ··· =
N −1 −1
Q
bN = U T j6=i bj . Denote this constant by C. Also the defining equation gives
Q
us 1 + b + · · · + bˆN + · · · + bN = N U −1 T −1
1
N
i N j6=i j b . Hence 1 + (N − 1)C = N C
and it follows that C = 1. Putting it back to the equations give that U T = ζN for
an N -th root of unity. We have excluded this from the base, thus we conclude that
the local expression over W [T, U ±1 ]′ exists near all points on X.

Remark 2.2. Under the notation of the proof above, we can concretely describe
ξ
the action of H0 on each Uk as the following: (ξ1 , . . . , ξN ) acts by xj 7→ ξji xj
Q 
ξi
for 1 ≤ j ≤ N, j 6= i, k, T 7→ T , b′k 7→ ′
1≤j≤k−1,j6=i ξj bk and bl(k) 7→
Q 
ξi
l(k)≤j≤N,j6=i ξj bl(k) .
Thus we may also make all the local etale maps given in the proof to be H0 -
equivariant with the H0 action on the target be the corresponding multiplication
on each coordinates.
We have seen there is a variety denoted by X, that is a blowup with H0 action of
the variety U T (X1N + X2N + · · · + XN
N
) = N X1 X2 · · · XN over Spec W [T, U ±1 ]′ and
is an isomorphism outside the closed subscheme defined by T . X locally admits an
etale map to the form given as in 2.1. Thus X further satisfies the property that
the base change of it along W [T, U ±1 ]′ → OF , T 7→ π, U 7→ u is a semistable(in
the sense of Definition 1.3) model of its generic fibre, with H0 action. The generic
fibre is the underlying motive of the l-adic representation Vλ,(uπ)−1 . That follows
because the generic fibre factor through the open locus of X where T 6= 0. We can
use this semistable model and take Z = X to prove Theorem 1.4 and Proposition
1.5 in the case d = 1.
Base changing the blowdown map we get from Lemma 2.1 along W [T, U ±1 ]′ →
W [R, U ±1 ]′ , T 7→ Rd , where Spec W [R, U ±1 ]′ is the open subscheme of Spec W [R, U ±1 ]
given by (Rd U )N 6= 1, we immediately get a blowup Xd of the projective variety
U Rd (X1N + X2N + · · · + XN N
) = N X1 X2 · · · XN over Spec W [R, U ±1 ]′ with an ac-
tion of H0 , such that the blowdown map is H0 -equivariant and Xd locally admits
an etale map to some
±1
W [R, U ±1 ]′ [Z1 , . . . , Zr , Zr+1 , . . . , Zn±1 ]/(U Rd − Z1 · · · Zr )
Note that to proceed now, we cannot simply base change X along
W [R, U ±1 ]′ → OF , R 7→ π, U 7→ u, since then the model has etale locally the
form
±1
OF [Z1 , . . . , Zr , Zr+1 , . . . , Zn±1 ]/(uπ d − Z1 · · · Zr )
which is not semistable.
ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES9

We state the theorem we would like to prove in this section below.

Theorem 2.3. For any t ∈ F such that v(t) = d > 0, there exists a (e and a
extension) totally ramified extension F ′ /F generated by π 1/e , a choice of e-th root
of π which is a uniformizer of OF (thus F ′ is purely ramified of degree e over F
with uniformizer π 1/e ), and a semistable model Y over OF ′ of Zt−1 (defined in the
first page) with compatible H0 action.
The idea of the proof goes as the following. Since the Xd above is a blowup
of the projective variety U Rd (X1N + X2N + · · · + XN N
) = N X1 X2 · · · XN over
±1 ′ ±1 ′
Spec W [R, U ] , its base change along W [R, U ] → F , R 7→ π, U 7→ u is iso-
morphic to the underlying motive of Vλ,t−1 where t = uπ d . We will use the theory
of toroidal embedding to construct a blowup Z of Xde for some e along some locus
contained in the closed subscheme defined by S, where Xde is the base change of
Xd along W [R, U ±1 ]′ → W [S, U ±1 ]′ , T → S e (again Spec W [S, U ±1 ]′ is defined by
the condition (S de U )N 6= 1 in Spec W [S, U ±1 ] as mentioned in the introduction).
We want our Z to admit an etale morphism to
±1
(2.1) W [S, U ±1 , Z1 , . . . , Zr , Zr+1 , . . . , Zn±1 ]/(U S − Z1 · · · Zr )
Zariski locally so that we see from construction that its base change along
W [S, U ±1 ] → OF ′ , S 7→ π 1/e , U 7→ u gives our desired Y.
We will use the theory of toroidal embeddings in the mixed characteristic case
from [KKMSD73b] to reduce construction to a combinatorial problem of subdivid-
ing conical complexes, which is also resolved in [Knu73]. So we will sketch the main
idea and notation from [KKMSD73a] [KKMSD73c] [Knu73] [KKMSD73b] first.
Let η be the generic point of Spec W and we will work with the base Spec W .
A torus embedding is an irreducible normal variety X of finite type over Spec W
with inclusion Tη ֒→ X, here Tη is the generic fiber, with T acting on X extending
the action of T on Tη .
Let M (T ), N (T ) be the character and cocharacter group of T . Let M f(T ) (N
e (T ))
be Z × M (T ) (Z × N (T )) and NR (T ) = N (T ) ⊗ R (N eR (T ) = N
e (T ) ⊗ R). Like the
usual case over a field, affine torus embeddings all come from cones σ in R≥0 ×NR (T )
not containing any linear subspace in NR (T ). The association is as the following:
Xσ = Spec R[· · · , π k · Xα , · · · ](k,α)∈M
f(T ), h(k,α),vi≥0,∀v∈σ

with the orbits of Tη in Xη corresponding to the faces of σ in 0 × NR (T ) and the

orbit of T0 in X0 corresponding to the faces of σ in R>0 × NR (T ). We will mostly
deal with the simple case where σ has the form R≥0 × σ ′ for some cone σ ′ ⊂ NR (T ).
In that case the orbits of Tη in Xη correspond naturally to the orbits of T0 in X0
via specialization.
The above is somehow the local picture. We will call an irreducible normal
scheme X of finite type over R with an open U ⊂ Xη a toroidal embedding if Zariski
locally near a point x, there exists an etale morphism ρ from (U, X) to some torus
embedding (Tη , Zρ(x) ) as pairs of schemes. This gives a global stratification of X −U
as the following. The existence of local models gives that X − U = ∪i∈I Ei with Ei
normal irreducible subscheme(so no self intersection for these Ei ) of codimension 1.
The strata are defined as the connected components of ∩i∈J Ei − ∪i∈J / Ei for various
subsets J ⊂ I, and for J = ∅, we take the strata to be U . Locally under the etale
maps to the torus embedding models Zρ(x) , these strata admit etale maps into the
strata of the local models Zρ(x) , hence the original strata are regular because those
10 LIE QIAN

of Zρ(x) are, due to the explicit description. The point of this definition is that we
may assign a canonical topological space ∆ that is the union of the cones obtained
locally from the models. The cones are glued together via inclusion contravariantly
with respect to the specialization of strata. The canonical topological space ∆
reflects lots of properties of X. The precise definition of the conical complex ∆
is in page 196 of [KKMSD73b]. It has the structure of a topological space and
each cone in it admits an injection into some Rm with integral structure. We just
stress that the main idea is that the character group M (T ) of the local model can
be canonically defined by our toroidal embedding (U, X) as the group of Cartier
divisors on StarY supported in StarY − U for a strata Y (Star means the union of
all the strata that specialize to it) and the cone can be canonically described as the
dual cone of the effective Cartier divisors.
Here is the crucial property we use to reduce our construction to the problem of
subdividing a complex: Any subdivision ∆′ of ∆ determines a morphism (U, X ′ ) →
(U, X) such that (U, X ′ ) is a toroidal embedding with the associated complex ∆′ and
that the associated map of complexes ∆′ → ∆ is the natural inclusion map. Indeed,
the map is a blowdown for an ideal sheaf on X supported in X − U completely
determined by the subdivision ∆′ .
We also have interpretation in terms of the complex ∆ of the properties that
the toroidal embedding variety X being regular and certain Cartier divisor being a
sum of irreducible Weil divisors without repetition. In the case over a field, for a
Cartier divisor S, by restricting to a Cartier divisor in StarY we can view it as an
element in the character group M associated to this stratum and hence a function
on the cone σ associated to this stratum. Compatibility gives us that S induce a
globally defined function on X, still denoted by S. Assume the hyperplane S = 1
in ∆ defined by this function meets every nonzero faces in ∆, then X is regular
and S vanishes to order 1 on all irreducible components of X − U if and only if the
intersection of S = 1 with any face τα of ∆ has vertices with integral coordinates
and the volume of the above intersection equals to 1/dα ! for dα the dimension of
τα . We will use similar argument to Mumford’s in the proof of this fact to prove
a similar result in the mixed characteristic case. It essentially reduces to the same
proof by adjoining the extra R factor.
Proof. To apply the theory of toroidal embeddings to get a blowup of Xd or Xde ,
we first need to check Xd and Xde have the structure of toroidal embeddings. Our
setting is as the following: let Ud ⊂ Xd,η be the open subvariety defined by R 6= 0.
We know Xd locally admits an etale map to
±1
(2.2) W [R, U ±1 ]′ [Z1 , . . . , Zr , Zr+1 , . . . , Zn±1 ]/(U Rd − Z1 · · · Zr )
which is an affine torus embedding of the generic torus Tη of the split torus T of
the form
W [R±1 , U ±1 , Z1±1 , . . . , Zr±1 , Zr+1
±1
, . . . , Zn±1 ]/(U Rd − Z1 · · · Zr )
with T action given by multiplying corresponding coordinates. And clearly Ud is
the preimage of the generic torus via R 6= 0. Let ei denote the character given
by the regular function Zi in X ∗ (T ), eu denote the character given by the regular
function U and similarly fi and fu in X∗P f(T ) is identified with
(T ), then the lattice M
r
Ze1 + . . .+ Zen + Zeu + Ze0 + Z d1 (−eu + j=1 ej ), where e0 corresponds to the extra
factor of Z as in page 191 of [KKMSD73b]. N eR (T ) is spanned by f1 , . . . , fn , fu f0 ,
 ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
11

and the cone σ giving the above toroidal embedding is just R≥0 f1 + . . . R≥0 fr +
R≥0 f0 . The structure of toroidal embedding on Xde is defined similarly with the
open set Ude given by S 6= 0.
We now give some description of the local picture and the relationship between
the complex associated to Xd and Xde . Consider the stratification on X mentioned
before the proof, as in page 195 of [KKMSD73b]. It suffices to look at a stratum
Yd of Xd supported in the closed subscheme defined by π because for any strata Yd′
not supported in the (π) we may intersect the closure of the strata with the closed
subscheme defined by π and get a strata Yd which is a specialisation of Yd′ and we
have StarYd′ naturally injects into StarYd . We may thus suppose that StarYd locally
admits an etale map to a scheme of the form 2.2. We have a bijective correspondance
between the strata on the varieties Xd and Xde because the fiber over R = 0
and S = 0 are isomorphic under pullback. Denote the stratum corresponding
to Yd by Yde . Similar local presentation as 2.2 holds for the strata Yde of Xde .
We have the intrinsically defined M fYd , MfYde , Ne Yd , N
e Yde , σ
e Yd , σ
eYde as in page 196
Yd
of [KKMSD73b] (their notation without the tilde). All σ e glued into a conical
complex ∆ e d and all σ eYde glued into a conical complex ∆ e de . Note that we have a
globally compatible decomposition M fYd = Z ⊕ M Yd , M fYde = Z ⊕ M Yde with the
first factor given by (π) and the second factor generated by the irreducible Cartier
divisors supported in StarYd − Ud (StarYde − Ude ) not supported in (π). Similarly
we have the decomposition N e Yd = Z ⊕ N Yd , N e Yde = Z ⊕ N Yde , σ eYd = R≥0 ⊕ σ Yd ,
e
σ Yde
= R≥0 ⊕ σ , ∆ Yde e de = R≥0 ⊕ ∆de , ∆ e d = R≥0 ⊕ ∆d . Indeed, if Yd is a
connected component of (π) ∩ ∩i∈I Ei \ ∪i∈I / Ei with I a subset of the irreducible
divisors not supported in π, and let Yd,η be the (unique, by local form) strata
of the generic fibre Xη in ∩i∈I Ei \ (∪i∈I / Ei ∪ (π)) that specializes to Yd , then the
M Yd , M Yde , N Yd , N Yde , σ Yd , σ Yde , ∆d , ∆de are the corresponding lattice, cones and
conical complexes associated to the strata Yd,η of the (field case) toroidal embedding
Ud ֒→ Xd,η defined as in page 59 of [KKMSD73c].
We now use the subdivision given in [Knu73] to construct a blowup of Xde .
Since S = R1/e is globally defined, it gives a Cartier divisor in Xde and hence by
the process described in the last paragraph before the proof, it lies in each M Yde
compatibly by restriction of Cartier divisors. Hence it defines a function on each
cone σ eYde compatibly and hence a function on ∆ e de . Denote this function by lde , so
e ∗ e Yde
it defines a closed subset ∆de = {x ∈ ∆ |lde (x) = 1}. Clearly ∆ e ∗ = R≥0 ⊕ ∆∗
de de
∗
where ∆de ⊂ ∆de also defined by lde = 1 and it is a compact convex polyhedral
set because the hyperplane lde = 1 meets every positive dimensional face of ∆de .
e ∗ and ∆∗ has integral structure given by various lattices M
∆ fYd and M Yd . The
de de
upshot of the discussion from page 105-108 of [KKMSD73c] and [Knu73] is that
we may find an e and a subdivision {σα } of ∆∗de into convex polyhedral sets such
that all vertices of all σα lie in (∆∗de )Z , the lattice given by the integral structure
mentioned above and that the volume (also given by the integral structure as in
page 95 of [KKMSD73c]) of each σα is 1/(dα )! where dα is the dimension of σα .
Adjoining the origin to the vertices gives a conical decomposition {τα } of ∆de
associated to {σα }. Now by part (d) of page 197 of [KKMSD73b], if we let ∆ e ′ be
de
the subdivided complex associated to the f.r.p.p decomposition {R≥0 ⊕ τα } of ∆ e de ,
it gives a toroidal embedding (Ude , Z) with a map Z → Xde respecting the inclusion
12 LIE QIAN

of Ude . Moreover, the associated map of the conical complex is just ∆ e′ → ∆ e de ,

de
and the integral structure is preserved.
We may associate an H0 action on (Ude , Z) such that the map Z → Xde is H0 -
equivariant for the following reason: Part (e) in Page 198 of [KKMSD73b] gives that
this map we constructed via a subdivision of the conical complex ∆ e ′ is a blowup
de
along an ideal sheaf I over Xde . The ideal I over StarYde can be written as the
sum of OXde (−D) for a collection of Cartier divisors D ∈ M fYde that satisfy certain
condition relevant to the subdivision of the cone σ eYde . Now we see from the local
form of Xde that the H0 action fix every Cartier divisor D ∈ M fYde (see Remark
2.2), and hence preserve the sum of these OXde (−D). The asserted H0 -equivariance
thus follows.
We now use the information on the evaluation of S on ∆ e ′ to prove the crite-
de
rion for regularity and the divisor S vanishing to the order 1 on any irreducible
Weil divisors, as well as giving a local model over W [S, U ±1 ]′ . By the two con-
dition on each σα , we see that each τα is generated as a cone by a set of vectors
{fα,i }1≤i≤dα +1 ⊂ (σα )Z ⊂ (τα )Z given by the vertices of σα , and by Lemma 4 on
Page 96 of [KKMSD73c], if we denote the linear subspace generated by τα as (τα )R ,
then {fα,i }1≤i≤dα +1 is a basis for the lattice of (τα )R given by the restriction of the
integral structure M Yde for some Yde such that τα ⊂ (N Yde )R . Thus it can be ex-
panded into a basis {fα,i }1≤i≤n of N Yde . Thus by changing the corresponding dual
basis as well, we write the torus Tη in the local model of the strata corresponding to
the cone R≥0 ⊕ τα as W [ p1 ][T1±1 , . . . , Tn±1 ] with the the Ti ∈ X ∗ (T ) being the dual
Pdα +1
basis of fα,i . Since the cone R≥0 ⊕ τα is now generated by R≥0 f0 + i=1 R≥0 fα,i ,
±1
the local model is by definition 2 isomorphic to W [T1 , . . . , Tdα +1 , Tdα +2 , . . . , Tn±1 ]
and hence smooth over W . Now the globally defined U ∈ X ∗ (T ) is a unit in
the coordinate ring W [T1 , . . . , Tdα +1 , Td±1 α +2
, . . . , Tn±1 ], hence U can be written as
cdα +2 cn
Tdα +2 · · · Tn . We also know that U is non-divisible in the finite free abelian group
X ∗ (T ) as this only involves the integral structure and hence could be checked
over Xde where an explicit local model gives the result. It follows that there are
no nontrivial common divisor of the cdα +2 , . . . , cn , and hence we can change co-
ordinate and write the coordinate ring as W [T1 , . . . , Tdα +1 , Td′±1 α +2
′±1
, . . . , Tn−1 , U ±1 ]
over W [U ±1 ]. Furthermore, since S ∈ M fYde evaluated at each fα,i , 1 ≤ i ≤
dα + 1 is 1 and evaluated at the basis of the first copy of R≥0 is 0, we see that
Qdα +1 Qn−1 ′d
S = i=1 Ti · j=dα +2 Tj j · U s , and hence over W [S, U ±1 ], we may change co-
Qn−1 ′d
ordinate by T1′ 7→ T1 j=dα +2 Tj j · U s+1 , Ti′ 7→ Ti for 2 ≤ i ≤ dα + 1, and
Tj′ 7→ Tj′ for dα + 2 ≤ j ≤ n − 1, we have that locally Z admits an etale map to
Qdα +1 ′
W [S, U ±1 , T1′ , . . . , Td′ α +1 , Td′±1
α +2
′±1
, . . . , Tn−1 ]/(SU − i=1 Ti ) over W [S, U ±1 ].
We deduce from the local description above that the base change of Z over
W [S, U ±1 ]′ along W [S, U ±1 ]′ → OF ′ : S 7→ π 1/e , U 7→ u, which we denote by Y,
Q α +1
admits an etale map to OF ′ [T1′ , . . . , Td′ α +1 , Td′±1 α +2
′±1
, . . . , Tn−1 ]/(π 1/e u − di=1 Ti ).
Since π 1/e is a uniformizer of OF ′ , we see that Y is regular and semistable over
OF ′ . We may further identify its generic fibre with the projective variety over F ′
defined by the equation

uπ d (X1N + · · · + XN
N
) = N X1 · · · XN
ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
13

compatible with the H0 action because the generic fibre of Y is also the base change
of Zη along W [ p1 ][S, U ±1 ] → F ′ : S 7→ π 1/e , U 7→ u. The map Zη → Xde,η is an
isomorphism outside S = 0, and hence the above base change of Z is isomorphic
to the fibre over S = π 1/e of Xde,η , which has the above form from its definition
and corresponding property of Xd,η . The H0 compatibility follows from the H0
compatibility of the map Z → Xde .

We now analyze the special fibre of Y because we want to apply Hyodo-Kato’s
semistable comparison theorem.
Remark 2.4. Denote the canonical log structure on Y given by the Cartier divisor
(π 1/e ) by N , the special fibre with pullback log structure by (Y, N ), the log struc-
ture on Z given by the divisor (S) by M . Picking a u′ ∈ W × that have the same
reduction as u in k × , we denote the base change of Z with pullback log structure
from M along W [S, U ±1 ]′ → W [T ]′ : U → u′ , S 7→ T by (Y, N ) (here Spec W [T ]′
is the open subscheme of Spec W [T ] defined by (T de u′ )N 6= 1), and further base
change along W [T ] → k, T 7→ 0 with pullback log structure by (Y1 , N1 ). Then we
have that as log schemes over k with H0 action, (Y, N ) ∼ = (Y1 , N1 ) (because they
are both base change of (Z, M ) by W [S, U ±1 ]′ → k : S 7→ 0, U 7→ u) and Y is a
blowup of the projective variety u′ T de (X1N + X2N + · · · + XN N
) = N X1 X2 · · · XN
′
over Spec W [T ] with equivariant H0 -action.

3. Comparison Theorem of Log Geometry

We start by introducing some notation we will use in log geometry, and then
we will specialize to the case we work in. We will denote a general log scheme by
(X, M ) where X is a usual scheme and M denotes the log structure. When M is
a monoid, it is to be interpretted as a chart for this log structure. In particular,
let (X, N, 1 7→ f ) be the log scheme determined by the chart N with 1 sent to the
f ∈ OX (X) under the log structure map. And let (X, (0)) denote the log scheme
X with trivial log structure. When there is only one log structure mentioned on a
scheme X, we will simply use X to refer to the log scheme (X, M ).
We also record the following lemma from [Ogu18] Chapter 4 Proposition 1.3.1:
Lemma 3.1. Let
g
(X ′ , M ′ ) (X, M )
′
f f
h
(Y ′ , N ′ ) (Y, N )
be a Cartesian diagram of log schemes. Then we have a natural isomorphism
g∗ω1 ∼ ω1 ′ ′
=
(X,M)/(Y,N ) (X ,M )/(Y ,N )
′ ′

We will work with (Y, N ) being a log smooth scheme over (W [T ]′ , N, 1 7→ T ),

where Spec W [T ]′ is a fixed choice of affine open subscheme of Spec W [T ] such that
the closed subscheme defined by T = 0 is contained in Spec W [T ]′ . In particular,
we will apply the theory to the (Y, N ) we get from last section in Remark 2.4
over the particular base (W [T ]′ , N, 1 7→ T )(although same notation). Denote by
(Yn , Nn ) (resp. (Yn , Nn )) the base change of (Y, N ) along (Wn [T ]′ , N, 1 7→ T ) →
14 LIE QIAN

(W [T ]′ , N, 1 7→ T )(resp. (Wn , N, 1 7→ 0) → (W [T ]′ , N, 1 7→ T )). Let (Y , N ) be

the base change of (Y, N ) along (W, N, 1 7→ 0) → (W [T ]′ , N, 1 7→ T ). Note that
the closed immersions in the corresponding fibre diagram are all exact and the
projection maps to the base are all log smooth.
For a base 4-ple (S, L, I, γ), where S is a scheme with OS killed by some integer
pn , L is a fine log structure on it (see [HK94] (2.6)), I is a quasi-coherent ideal on
S and γ is a PD structure on I, and a fine log scheme (X, M ) over (S, L) such that
γ extends to X, we may define the cristalline site and cristalline cohomology as in
[HK94] (2.15).
Take (X, M ) = (Y1 , N1 ) and (S, L) = (Wn , N, 1 7→ 0) with the ideal I = (p) and
its usual PD structure, [HK94] define the i-th log cristalline cohomology of (Y1 , N1 )
as the limit
1
lim H i (((Y1 , N1 )/(Wn , N, 1 7→ 0))crys , OY1 /Wn )[ ]
←− p
n
and they also equip it with a nilpotent operator N that will be the same as the N
given by p-adic Hodge theory when identified by the comparison isomorphism.
Theorem 3.2. The above limit is isomorphic to Hi (dRY /(W,N,17→0) )[ p1 ] while the
operator N is given as the degree i boundary homomorphism of the long exact
sequence given by the following exact triangle
·dlog1
dRY /(W,N,17→0) [−1] dRY /(W,(0)) dRY /(W,N,17→0)
which is defined at each degree i by taking the i-th wedge power of the following
locally split exact sequence of locally free sheaf of modules
·d log 1
0 OY ωY1 /(W,(0)) ωY1 /(W,N,17→0) 0

by Theorem 3.2.3 of [Ogu18] since we note that (Y , N ) is log smooth over (W, N, 1 7→
1 ∼
0) and we identify ω(W,N,17 →0)/(W,(0)) = W by d log 1 7→ 1.

Proof. We will prove the theorem in 3 steps.

(i) We prove that there exists an N -equivariant isomorphism
(3.1) H i (((Y1 , N1 )/(Wn , N, 1 7→ 0))crys , OY1 /Wn ) ∼
= Hi (dRYn /(Wn ,N,17→0) )
with the operator N on the right hand side given by the degree i boundary homo-
morphism of the following exact triangle
·dlog1
dRYn /(Wn ,N,17→0) [−1] dRYn /(Wn ,(0)) dRYn /(Wn ,N,17→0)

obtained via various wedge power of the following exact sequence as the process
above
·d log 1
0 OYn ωY1 ωY1 0
n /(Wn ,(0)) n /(Wn ,N,17→0)

The transition maps in the limit

lim H i (((Y1 , N1 )/(Wn , N, 1 7→ 0))crys , OY1 /Wn )
←−
n

is compatible with the pullback homomorphism of cohomology

Hi (dRYn+1 /(Wn+1 ,N,17→0) ) → Hi (dRYn /(Wn ,N,17→0) ) under the isomorphism 3.1.
ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
15

(ii) We will prove there exists commuting short exact sequences of W modules
(3.2)
0 Hi (dRY /(W,N,17→0) )/pn+1 Hi (dRYn+1 /(Wn+1 ,N,17→0) ) Hi+1 (dRY /(W,N,17→0) )[pn+1 ] 0
·p

0 Hi (dRY /(W,N,17→0) )/pn Hi (dRYn /(Wn ,N,17→0) ) Hi+1 (dRY /(W,N,17→0) )[pn ] 0

where the middle vertical map is the pullback map, the left vertical map is natural
reduction and the right vertical map is multiplication by p. All maps in the left
commutative square are N equivariant.
(iii) Conclude Theorem 3.2.

We first prove (i).

Again we recall the base 4-ple (S, L, I, γ) being a fine pn torsion log scheme with
a quasi-coherent ideal I and a PD structure on it. In the discussion below, we will
always take I = (p) and the usual PD structure on it. This PD structure extends
to all schemes X over S. Recall the following theory of log cristalline cohomology
of crystals. This is the basic case in (2.18) of [HK94] combined with (2.17). We
will not recall the notions in the proposition but refer to the same reference.

Proposition 3.3. For (X, M )/(S, L) log schemes, if there exists a log smooth
(Z, N )/(S, L) with a closed immersion i : (X, M ) ֒→ (Z, N ). Denote the (log)
PD envelope of i : (X, M ) ֒→ (Z, N ) by (D, MD ), then there exists a complex of
OZ -module CX,Z/S of the form:
∇ 1 ∇ 2 ∇
OD OD ⊗OZ ωZ/S OD ⊗OZ ωZ/S ···
1
where ωZ/S is the log differential module, and Leibniz rules are satisfied:
i j
∇(m ⊗ ω) = ∇(m) ⊗ ω + m ⊗ dω (m ∈ OD ⊗OZ ωZ/S , ω ∈ ωZ/S )

such that the cohomology of this complex as an object in D(Sh(Xet )) computes the
log cristalline cohomology H · (((X, M )/(S, L))crys , OX/S ).

Remark 3.4. [HK94] (2.18) also gives that the association of the complex CX,Z/S
is natural with respect to the system (X, M ) → (Z, N ) → (S, L, I, γ).

Apply this proposition to (X, M ) = (Y1 , N1 ), (S, L) = (Wn , N, 1 7→ 0) and the

closed immersion i : (Y1 , N1 ) → (Yn , Nn ) whose ideal of definition is generated by
p, hence (D, MD ) = (Yn , Nn ) and by Leibniz rule we see that CY1 ,Yn /(Wn ,N,17→0) =
dRYn /(Wn ,N,17→0) and hence the abstract isomorphism in 3.1 follows. We still need
to show it’s N -equivariant.
There is also a description of the operator N on the log cristalline cohomology in
terms of the complexes mentioned above, see [HK94] (3.6). We simplify it a little
bit.

Proposition 3.5. Suppose there exist a (log) closed immersion (X, M ) ֒→ (Z, N )
over (Wn [T ]′ , N, 1 7→ T ) with (Z, N ) log smooth over (Wn [T ]′ , N, 1 7→ T ). Then
there exist an exact triangle

CX,Zhi/(Wn hT i′ ,N,17→T ) [−1] → CX,Z/(Wn ,(0)) → CX,Zhi/(Wn hT i′ ,N,17→T ) →

16 LIE QIAN

where Wn hT i′ denotes the usual PD envelope of Wn [T ]′ along the ideal (t) and Zhi
is the base change of Z along Wn [T ]′ → Wn hT i′ with the pullback log structure. We
refer the reader to [HK94] (3.6) for the definition of this exact triangle.
Tensoring the above exact triangle by ⊗LWn hT i′ Wn , we get:

CX,Z/(Wn ,N,17→0) [−1] → CX/(Wn ,(0)) ⊗LWn hT i′ Wn → CX,Z/(Wn ,N,17→0) →

where Z is the base change of Z along Wn [T ]′ → Wn , t 7→ 0 with the pullback log

structure, so that the operator N on H i (((X, M )/(Wn , N))crys , OX/Wn ) is given by
the connecting homomorphism of degree i of this exact triangle under the identifi-
cation given after Proposition 3.3.
In fact, we will use this proposition in our case for (X, M ) = (Y1 , N1 ) and
(Z, N ) = (Yn , Nn ). The first is the base change of the second along (Wn , N, 1 7→
0) → (Wn [T ]′ , N, 1 7→ T ). Thus Zhi = (Yn hi, Nn hi) as log schemes, the latter of
which we define to be the base change of (Yn , Nn ) → (Wn [T ]′ , N, 1 7→ T ) along
(Wn hT i′ , N, 1 7→ T ). Also Z = (Yn , Nn ) as log schemes. We would identify the
objects appearing in the above proposition in more concrete terms.
Here is a diagram illustrating the situation:
(Y1 , N1 ) (Yn hi, Nn hi) (Yn , Nn )

(Wn hT i′ , N, 1 7→ T ) (Wn [T ]′ , N, 1 7→ T )

(Wn , (0))
Apply Proposition 3.3 to (X, M ) = (Y1 , N1 ), (S, L) = (Wn hT i′ , N, 1 7→ T ) and
the closed immersion (Y1 , N1 ) → (Yn hi, Nn hi). Since the closed immersion is ex-
act and the ideal of definition is generated by p and all the divided powers of t,
(D, MD ) = (Yn hi, Nn hi) in this case. Proposition 3.3 gives that CY1 ,Yn hi/(Wn hT i′ ,N,17→T ) ∼
=
dRYn hi/(Wn hT i′ ,N,17→T )
Apply Proposition 3.3 to (X, M ) = (Y1 , N1 ), (S, L) = (Wn , (0)) and the closed
immersion (Y1 , N1 ) → (Yn , Nn ). (Yn , Nn ) is log smooth over (Wn , (0)) because
it is log smooth over (Wn [T ]′ , N, 1 7→ T ) and (Wn [T ]′ , N, 1 7→ T ) is log smooth
over (Wn , (0)) by the smoothness criterion (2.9) in [HK94]. In this case (D, MD ) =
(Yn hi, Nn hi) since the ideal of deinition given by the closed immersion i is generated
by p, t. Proposition 3.3 shows that CY1 ,Yn /(Wn ,(0)) is given by a complex
∇ ∇ ∇
OYn hi OYn hi ⊗OYn ωY1 n /(Wn ,(0)) OYn hi ⊗OYn ωY2 n /(Wn ,(0)) ···
Via the application of Proposition 3.3 indicated above, and identifying
= OYn hi ⊗OYn ωYi n /(Wn [T ]′ ,N,17→T ) ∼
ωYi n hi/(Wn hT i′ ,N,17→T ) ∼ = Wn hT i′ ⊗Wn [T ]′ ωYi n /(Wn [T ]′ ,N,17→T ) ,
one can verify that using the recipe given in (3.6) of [HK94], the exact triangle se-
quence
CY1 ,Yn hi/(Wn hT i′ ,N,17→T ) [−1] → CY1 ,Yn /(Wn ,(0)) → CY1 ,Yn hi/(Wn hT i′ ,N,17→T ) →
is given by actual chain maps
(3.3)
·d log T
0 OYn hi ⊗OYn ωYi−1
n /(Wn [T ] ,N,17→T )
′ OYn hi ⊗OYn ωYi n /(Wn ,(0)) OYn hi ⊗OYn ωYi n /(Wn [T ]′ ,N,17→T ) 0
 ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
17

which is obtained by tensoring OYn hi with the i-th wedge power of the locally split
short exact sequence
·d log T
0 OYn ωY1 n /(Wn ,(0)) ωY1 n /(Wn [T ]′ ,N,17→T ) 0

Now we claim that the following diagram commutes and the three vertical arrow
are isomorphisms :
·dlogT
dRYn hi/(Wn hT i′ ,N,17→T ) ⊗LWn hT i′ Wn [−1] CY1 ,Yn /(Wn ,(0)) ⊗LWn hT i′ Wn dRYn hi/(Wn hT i′ ,N,17→T ) ⊗LWn hT i′ Wn

·dlog1
dRYn /(Wn ,N,17→0) [−1] dRYn /(Wn ,(0)) dRYn /(Wn ,N,17→0)
The commutativity is straightforward. The isomorphisms of left and right col-
umn just follows form the locally freeness of each modules of differentials and the
pullback result Lemma 3.1 for modules of differentials.
For the middle vertical arrow, we know that (OYn hi ⊗OYn ωYi n /(Wn ,(0)) ) ⊗Wn hT i′
Wn ∼= ωYi n /(Wn ,(0)) and since ωYi n /(Wn ,(0)) → ωYi n /(Wn ,(0)) is already surjective and
we know the ∇ in CY1 ,Yn /(Wn ,(0)) satisfy Leibniz rules and hence the differential
applied to the image of ωYi n /(Wn ,(0)) in OYn hi ⊗OYn ωYi n /(Wn ,(0)) is just log differential,
we see that the differentials in the complex after tensoring are exactly the log de-
Rham differentials.
Thus, by the last part of Proposition 3.5, we see that the operator N on
i
Hcrys ((Y1 , N1 )/(Wn , N), OY1 /Wn ) is given by the degree i connecting homomor-
phism of the long exact sequence associated to the lower row of the above diagram.
Thus (i) is proved.

Now we prove (ii).

By proper base change theorem in the context of derived category of coherent
sheaves(Section 7.7 of [Gro63]) applied to the Cartesian square
in
Yn Y
πn π∞

Wn W
and the exact triangle
·dlog1
(3.4) dRY /(W,N,17→0) [−1] dRY /(W,(0)) dRY /(W,N,17→0)

,we see that

Rπ∞,∗ dRY /(W,N,17→0) ⊗LW Wn [−1] Rπ∞,∗ dRY /(W,(0)) ⊗LW Wn Rπ∞,∗ dRY /(W,N,17→0) ⊗LW Wn
∼
= ∼
= ∼
=

Rπn,∗ dRYn /(Wn ,N,17→0) [−1] Rπn,∗ dRYn /(Wn ,(0)) Rπn,∗ dRYn /(Wn ,N,17→0)

(Note that each term of the de Rham complexes in 3.4 (as actual chain complexes) is
locally free by the log smoothness and that Li∗n dRY /(W,N,17→0) = dRYn /(Wn ,N,17→0) ,
Li∗n dRY /(W,(0)) = dRYn /(Wn ,(0)) by applying usual pullback at each degree and 3.1).
18 LIE QIAN

Applying Rπ∞,∗ dRY /(W,N,17→0) ⊗LW to the exact triangle

·pn
W W Wn
and then taking the long exact sequence, we get a short exact sequence
(3.5)
0 Hi (dRY /(W,N,17→0) )/pn Hi (dRYn /(Wn ,N,17→0) ) Hi+1 (dRY /(W,N,17→0) )[pn ] 0

where the inclusion is N -equivariant since it is induced by the above diagram from
the connecting homomorphisms of the diagram

Rπ∞,∗ dRY /(W,N,17→0) ⊗LW W [−1] Rπ∞,∗ dRY /(W,(0)) ⊗LW W Rπ∞,∗ dRY /(W,N,17→0) ⊗LW W

Rπ∞,∗ dRY /(W,N,17→0) ⊗LW Wn [−1] Rπ∞,∗ dRY /(W,(0)) ⊗LW Wn Rπ∞,∗ dRY /(W,N,17→0) ⊗LW Wn

Applying Rπ∞,∗ dRY /(W,N) ⊗LW to the morphism between exact triangles

·pn+1
W W Wn+1
·p =
·pn
W W Wn
and take long exact sequence will give us the desired commutative diagram

(3.6)
0 Hi (dRY /(W,N,17→0) )/pn+1 Hi (dRYn+1 /(Wn+1 ,N,17→0) ) Hi+1 (dRY /(W,N,17→0) )[pn+1 ] 0
·p

0 Hi (dRY /(W,N,17→0) )/pn Hi (dRYn /(Wn ,N,17→0) ) Hi+1 (dRY /(W,N,17→0) )[pn ] 0

in (ii) where the left commutative square is N equivariant. Thus (ii) is proved.

To prove (iii), take inverse limit over the short exact sequences in 3.6 indexed
by n, we get

0 limn Hi (dRY /(W,N,17→0) )/pn limn Hi (dRYn /(Wn ,N,17→0) ) lim·p Hi+1 (dRY /(W,N,17→0) )[pn ]
←− ←− ←−
where the inclusion is N equivariant.
Note that all Hj (dRY /(W,N,17→0) ) are finitely generated over W (by truncation
and that proper pushforward preserve coherent sheaves). So
limn Hi (dRY /(W,N,17→0) )/pn ∼
= Hi (dRY /(W,N,17→0) ) and
←−
lim·p Hi+1 (dRY /(W,N,17→0) )[pn ] = 0. Thus (iii) is proved.
←−


We would like to prove another GAGA type result.

Let (X, M ) be a log scheme smooth over (C, N, 1 7→ 0) and proper as a scheme.
Let (X an , M an ) be the analytic log scheme associated to it which is also smooth
over the analytic point (pt, N, 1 7→ 0).
ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
19

Proposition 3.6. There exists an N equivariant isomorphism

Hi (dRX/(C,N,17→0) ) ∼
= Hi (dRX an /(pt,N,17→0) )
where the N on each hypercohomology is defined by degree i boundary morphism of
the exact triangles obtained similar to the one in Theorem 1.6
·dlog1
dRX/(C,N,17→0) [−1] dRX/(C,(0)) dRX/(C,N,17→0)
·dlog1
dRX an /(pt,N,17→0) [−1] dRX an /(pt,(0)) dRX an /(pt,N,17→0)

Proof. We can apply the exact GAGA functor F 7→ F an from the category of
sheaves of OX modules to the category of sheaves of OX an modules. Let λ be the
morphism X an → X. Thus it suffices to show there is a canonical isomorphism
between
·dlog1
λ∗ dRX/(C,N,17→0) [−1] λ∗ dRX/(C,(0)) λ∗ dRX/(C,N,17→0)
∼
= ∼
= ∼
=
·dlog1
dRX an /(pt,N,17→0) [−1] dRX an /(pt,(0)) dRX an /(pt,N,17→0)
Commutativity of diagram is clear from the definition of log1 as the image of 1
of the monoid in the differential modules. The isomorphism of complexes actually
holds termwise. Namely, we could prove the stronger statement that for any log
scheme (X, M )/(Z, L) over C, if we denote its analytification by (X an , M )/(Z an , L)
then λ∗ ω(X,M)/(Z,L)
i ∼ i
= ω(X an ,M)/(Z an ,L) for any non-negative integer i. It suffices
1
to prove for i = 1. We know from definition that ω(X,M)/(Z,L) is Ω1X/Z ⊕ M ⊗ OX
modulo the relation di(m) ⊕ −m ⊗ i(m) and 0 ⊕ l ⊗ 1 for all m ∈ M and l ∈
1
L where i is the structure morphism M → OX . Similarly ω(X an ,M)/(Z an ,L) is
1
ΩX an /Z an ⊕ M ⊗ OX an modulo the relation di(m) ⊕ −m ⊗ i(m) and 0 ⊕ l ⊗ 1 for all
m ∈ M and l ∈ L where i is the structure morphism M → OX an . We conclude the
case i = 1 by noting the fact that λ∗ Ω1X/Z ∼ = Ω1X an /Z an .


We apply this proposition to (X, M ) = (Y C , N C ) and its analytification (Y an , N an ).

Since Z is a blowup of the projective variety Xde (Under the notation of previous
section), as the base change of Z along W [T ]′ → C, T 7→ 0, Y C is projective as
well. Hence the condition of the proposition is satisfied and we may reduce the
computation of the operator N to the analytic setting.

4. Proof of Theorem 1.1

Recall the hypothesis that F is a characteristic 0 local field containing ζN whose
residue characteristic equals to p and does not divide N .
It suffices to prove that Vλ,t−1 is regular ordinary as a GF ′ representation for a
finite extension F ′ /F because of the following lemma.
Lemma 4.1. If a representation r : GF → GLn (Qp ) satisfy that its restriction
r|GF ′ is de Rham regular and ordinary for some finite extension F ′ /F , then the
same holds for r.
Proof. We may assume without loss of generality that F ′ /F is Galois. By the inter-
pretation of the regular and ordinary condition as in Theorem 6.1.2 of [ACC+ 18],
20 LIE QIAN

we see that  
ψ1 ∗ ∗ ∗
0 ψ2 ∗ ∗
 
r|GF ′ ∼ . .. .. 
 .. . . ∗
0 ··· 0 ψn
×
where for each i = 1, . . . , n the character ψi : GF ′ → Qp agrees with the character
Y
σ ∈ IF ′ 7→ τ (Art−1
F ′ (σ))
µτ,i

τ ∈Hom(F ′ ,Qp )
on an open subgroup of the inertia group IL , for some tuple µτ,i satisfying for each
fixed τ , µτ,1 < · · · < µτ,n . It suffices to show the GF (actually Gal(F ′ /F )) action
preserve the filtration given by the above triangular form and µτ,i = µτ σ,i for any
τ ∈ Hom(F ′ , Qp ) and σ ∈ Gal(F ′ /F ).
Take a σ ∈ Gal(F ′ /F ), consider the one dimensional subspace Qp e1 underlying
ψ1 and the subspace L1 = Qp σe1 . GF ′ acts on L1 with the τ HT weight −µτ σ,1 .
Now if there exists some τ such that µτ,1 6= µτ σ,1 , there must exist some τ such
that µτ σ,1 < µτ,1 , thus the τ HT weights of L1 is strictly greater than the τ HT
weights of ψi for any i = 1, . . . , n(Following from the strict ordering of the various
µτ,i ). From this we see Hom(L1 .ψi ) = 0 for any i = 1, . . . , n(otherwise their τ HT
weights must be same for all τ ) and thus Hom(L1 , V ) = 0 where V denotes the
underlying vector space of r, which is a contradiction.
Thus µτ,1 = µτ σ,1 for any τ ∈ Hom(F ′ , Qp ) and σ ∈ Gal(F ′ /F ). By similar
argument, we still have Hom(L1 .ψi ) = 0 for any i = 2, . . . , n, hence Hom(L1 , V ) =
Hom(L1 , ψ1 ). The left hand side is nonzero and L1 and the vector space underlying
ψ1 are both of 1 dimensional, thus we must have σ preserve the 1 dimensional vector
space underlying ψ1 . We proved that the first step of the filtration is preserved
under Gal(F ′ /F ) action. Quotienting out the subspace underlying ψ1 , we can use
the same argument to deduce the rest of the claim.

Note that the F ′ /F we will use is totally ramified, hence the residue field is still
k.
The following p-adic Hodge theoretic lemma will be used.
Lemma 4.2. Suppose that Fv is a characteristic 0 local field over Qp and
r : Gal(Fv /Fv ) → GLn (Qp )
is semi-stable. Suppose moreover that for every τ : Fv0 ֒→ Qp , the operator N on
the n-dimensional vector space
(r ⊗τ,Fv0 Bst )Gal(F v /Fv )
is maximally nilpotent, i.e. the smallest j such that N j = 0 is n. Then r is regular
and ordinary.
Proof. This Lemma 2.2 (2) of [BLGHT09]. 
Thus it suffices to prove that the operator N on the space Dst,σ (Vλ,t−1 ) :=
′
(Vλ,t−1 ⊗σ,(F ′ )0 Bst )Gal(F /F ) is maxinally nilpotent, where (F ′ )0 is the maximally
′

unramified subextension of F ′ and σ : (F ′ )0 ֒→ Q(ζN )λ is any fixed embedding of

ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
21

(F ′ )0 and λ is a place of Q(ζN ) over p. Applying the main comparison theorem of

[Tsu99] to the semistable model Y we get from Theorem 1.4, we have
∼ H N −2 ((Y, N )/(W, N, 1 7→ 0))[ 1 ]H0 ,σ−1 χ ⊗(F ′ )0 ,σ Q(ζN )
Dst,σ (Vλ,t−1 ) = crys λ
p
as filtered (φ, N )-modules, where the right hand side is Hyodo-Kato’s log cristalline
cohomology.
Now apply Proposition 1.5 and Remark 2.4 to get the variety Y with H0 action
over Spec W [T ] and the remark gives that
N −2 1 −1
∼
Hcrys ((Y, N )/(W, N, 1 7→ 0))[ ]H0 ,σ χ ⊗(F ′ )0 ,σ Q(ζN )λ =
p
N −2 1 −1
Hcrys ((Y1 , N1 )/(W, N, 1 7→ 0))[ ]H0 ,σ χ ⊗(F ′ )0 ,σ Q(ζN )λ
p
So we only need to identify the limit
1
lim H N −2 (((Y1 , N1 )/(Wn , N, 1 7→ 0))crys , OY1 /Wn )[ ]
←− p
n
and describe the operator N on its H0 eigenspace determined by the character
σ −1 χ.
Now by Theorem 1.6 applied to the log scheme (Y, N ) where Y is as above and
N is the log structure on it given by the divisor T over (W [T ], N, 1 7→ T ) and
i = N − 2, we may identify the above limit with the de Rham cohomology of
(Y , N )/(W, N, 1 7→ 0). And it suffice to prove that the operator N defined as the
degree N − 2 boundary homomorphism of the long exact sequence given by the
following exact triangle
·dlog1
dRY /(W,N,17→0) [−1] dRY /(W,(0)) dRY /(W,N,17→0)
is maximally nilpotent on the σ −1 χ eigenspace of the H0 action.
This algebraic statement can be verified over C. So let (Y C , N C ) denote the base
change of (Y , N ) along (W, N, 1 7→ 0) → (C, N, 1 7→ 0) for some fixed embedding
τ : W → C. We are reduced to show that the N on HN −2 (dRY C /(C,N,17→0) ) given
as the degree N − 2 boundary map of the long exact sequence given by the exact
triangle
·dlog1
dRY C /(C,N,17→0) [−1] dRY C /(C,(0)) dRY C /(C,N,17→0)
−1
is maximally nilpotent on the τ σ χ eigenspace of the H0 action.
Let (S, N, 1 7→ T ) be the analytic disc centered at 0 with radius 1, with coordinate
T in C and π an : Y an → S be the analytic space associated to Y pulled back to S
from the affine line. We have the log structure N an on Y an given by N on Y . We
write (Y an , N an ) as the base change of (Y an , N an ) along (pt, N, 1 7→ 0) → (S, N, 1 7→
T ), which is clearly also the analytification of (Y C , N C ). Now we apply Theorem
1.8 to see there exists an N and H0 equivariant isomorphism
HN −2 (dR )=∼ HN −2 (dR an )
Y C /(C,N,17→0) Y /(pt,N,17→0)

We are reduced to prove that the operator N on the σ −1 χ eigenspace of

HN −2 (dRY an /(pt,N,17→0) ) is maximally nilpotent.

Apply [Ill94] 2.2.2 to the setting X = Y an , S = S with their mentioned log

structure.
22 LIE QIAN

[Ill94] 2.2.2 says Ri π∗an dRY an /(S,N,17→T ) is locally free for all i(in particular for
i = N − 2) and

(4.1) RN −2 π∗an dRY an /(S,N,17→T ) ⊗OS C{0} ∼
= HN −2 (dRY an /(pt,N,17→0) )
(Here, we identify the ωY· in [Ill94] 2.2.2, which is defined there as the derived
pullback of dRY an /(S,N,17→T ) to the point 0, with dRY an /(pt,N,17→0) by Lemma 3.1
and that by locally freeness we can pullback termwise for dRY an /(S,N,17→T ) .)
We claim this isomorphism 4.1 is N equivariant, where the N on the left hand
side is the reduction of the Gauss-Manin connection
1
RN −2 π∗an dRY an /(S,N,17→T ) → RN −2 π∗an dRY an /(S,N,17→T ) ⊗OS ω(S,N,17 →T )/C

that is (see 2.2.1.2 and 2.2.1.3 of [Ill94]) the degree N −2 connecting homomorphism
of the following exact triangle (the definition is completely similar to that in step (2)
of section 3) when applying Rπ∗an , if we identify ω(S,N,17
1 ∼
→T )/C = OS by dlogT 7→ 1:
·dlogT
dRY an /(S,N,17→T ) [−1] dRY an /(pt,(0)) dRY an /(S,N,17→T )
and the N on the right hand side is defined in step (2) of section 3 and is the
operator we reduced to calculate.
Granting the following lemma, we are reduced to calculate the residue of the
Gauss-Manin connection at 0 on the locally free sheaf RN −2 π∗an dRY an /(S,N,17→T ) ,
and that is linked to monodromy of this locally free sheaf by [Ill94] (2.2.3). The
proof of the lemma is quite formal, the reader is recommended to skip it.
Lemma 4.3. The above isomorphism 4.1 is N -equivariant.
Proof. By proper base change applied to the cartesian diagram
i0
Y an Y an
π an

pt S
an
and the exact triangle of complexes over Y
·dlogT
dRY an /(S,N,17→T ) [−1] dRY an /(pt,(0)) dRY an /(S,N,17→T )
we get an isomorphism of exact triangles
H· (dRY an /(pt,N,17→0) )[−1] H· (dRY an /(pt,(0)) ) H· (dRY an /(pt,N,17→0) )

Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0} [−1] Rπ∗an dRY an /(pt,(0)) ⊗LOS C{0} Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0}

Thus HN −2 (dRY an /(pt,N,17→0) ) ∼

= HN −2 (Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0} ) is N
equivariant where the N on the right hand side is given as the degree N −2 boundary
momorphism of the lower exact triangle. Now to prove the lemma, it suffices to
prove that the surjection RN −2 π∗an dRY an /(S,N,17→T ) ։ HN −2 (dRY an /(pt,N,17→0) ) is
N equivariant, which factor as the composition of the (N − 2)-th cohomology of
the map Rπ∗an dRY an /(S,N,17→T ) → Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0} and the inverse
of the isomorphism Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0} → H· (dRY an /(pt,N,17→0) ) of the
left or right column in the above diagram
 ORDINARITY OF LOCAL GALOIS REPRESENTATION ARISING FROM DWORK MOTIVES
23

Now it is reduced to show the N equivariance of the natural map RN −2 π∗an dRY an /(S,N,17→T ) →
N −2
H (Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0} ). But this clearly follows from the commu-
tative diagram

Rπ∗an dRY an /(S,N,17→T ) [−1] Rπ∗an dRY an /(S,(0)) Rπ∗an dRY an /(S,N,17→T )

Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0} [−1] Rπ∗an dRY an /(pt,(0)) ⊗LOS C{0} Rπ∗an dRY an /(S,N,17→T ) ⊗LOS C{0}

Thus it now is reduced to show the residue N at 0 of the H0 -τ σ −1 χ eigenpart of

RN −2 π∗an dRY an /(S,N,17→T )is maximally nilpotent. But by Corollary 2.2.3 of [Ill94],
if we identify the fibre of the H0 -τ σ −1 χ eigenpart of RN −2 π∗an dRY an /(S,N,17→T ) at 0
with the fibre at s, here s 6= 0 and (sde u′ )N 6= 1, then the monodromy Ts at s around
0 is identified with exp(−2πiN ). We see from Proposition 1.5 and Remark 2.4 that
outside T = 0, the family Y an is just the projective variety u′ T de (X1N + X2N + · · · +
N
XN ) = N X1 X2 · · · XN over the punctured analytic disc with coordinate T , where
u should be viewed as its image in C. As Y an is (over the punctured unit disc) the
′

base change of the projective variety u′ T (X1N + X2N + · · · + XN N

) = N X1 X2 · · · XN
along S → S, T 7→ T , we know the monodromy Ts can be identified with ρs (γ∞ )de
de

which is again maximally unipotent from the condition of our Theorem 1.1 since the
coefficient field C is of characteristic 0. Now the maximal nilpotence of N follows
from the monodromy T1 = exp(−2πiN ) on the eigenspace of the H0 action being
maximally unipotent.

References
[ACC+ 18] Patrick B. Allen, Frank Calegari, Ana Caraiani, Toby Gee, David Helm, Bao V. Le
Hung, James Newton, Peter Scholze, Richard Taylor, and Jack A. Thorne, Potential
automorphy over CM fields, arXiv e-prints (2018), arXiv:1812.09999.
[BLGHT09] Tom Barnet-Lamb, David Geraghty, Michael Harris, and Richard Taylor, A family
of calabi–yau varieties and potential automorphy ii, Publications of the Research
Institute for Mathematical Sciences 47 (2009).
[Gro63] Alexander Grothendieck, éléments de géométrie algébrique : Iii. étude coho-
mologique des faisceaux cohérents, seconde partie, Publications Mathématiques de
l’IHÉS 17 (1963), 5–91 (fr). MR 163911
[HK94] Osamu Hyodo and Kazuya Kato, Exposé v : Semi-stable reduction and crystalline
cohomology with logarithmic poles, Périodes p-adiques - Séminaire de Bures, 1988
(Jean-Marc Fontaine, ed.), Astérisque, no. 223, Société mathématique de France,
1994, talk:5, pp. 221–268 (en). MR 1293974
[Ill94] Luc Illusie, Exposé i : Autour du théorème de monodromie locale, Périodes p-
adiques - Séminaire de Bures, 1988 (Jean-Marc Fontaine, ed.), Astérisque, no. 223,
Société mathématique de France, 1994, talk:1, pp. 9–57 (fr). MR 1293970
[KKMSD73a] G. Kempf, F. Knudsen, D. Mumford, and B. Saint-Donat, Equivariant embeddings
of tori, pp. 1–52, Springer Berlin Heidelberg, Berlin, Heidelberg, 1973.
[KKMSD73b] , Further applications, pp. 165–209, Springer Berlin Heidelberg, Berlin, Hei-
delberg, 1973.
[KKMSD73c] , Semi-stable reduction, pp. 53–108, Springer Berlin Heidelberg, Berlin, Hei-
delberg, 1973.
[Knu73] Finn F. Knudsen, Construction of nice polyhedral subdivisions, pp. 109–164,
Springer Berlin Heidelberg, Berlin, Heidelberg, 1973.
[Ogu18] Arthur Ogus, Lectures on logarithmic algebraic geometry, Cambridge Studies in
Advanced Mathematics, Cambridge University Press, 2018.
24 LIE QIAN

[Tsu99] Takeshi Tsuji, p-adic étale cohomology and crystalline cohomology in the semi-
stable reduction case, Inventiones mathematicae 137 (1999), no. 2, 233–411.

Department of Mathematics, Stanford University

Current address: Department of Mathematics, Stanford University
Email address: [email protected]