Iwasawa Coh Final

COATES-WILES HOMOMORPHISMS AND IWASAWA COHOMOLOGY
FOR LUBIN-TATE EXTENSIONS
PETER SCHNEIDER AND OTMAR VENJAKOB
To John Coates on the occasion of his 70th birthday
Abstract. For the p-cyclotomic tower of Qp Fontaine established a description of local

Iwasawa cohomology with coefficients in a local Galois representation V in terms of the
ψ-operator acting on the attached etale (ϕ, Γ)-module D(V ). In this article we generalize
Fontaine’s result to the case of arbitratry Lubin-Tate towers L∞ over finite extensions L of
Qp by using the Kisin-Ren/Fontaine equivalence of categories between Galois representations
and (ϕL , ΓL )-module and extending parts of [Her], [Sch]. Moreover, we prove a kind of
explicit reciprocity law which calculates the Kummer map over L∞ for the multiplicative
group twisted with the dual of the Tate module T of the Lubin-Tate formal group in terms of
Coleman power series and the attached (ϕL , ΓL )-module. The proof is based on a generalized
Schmid-Witt residue formula. Finally, we extend the explicit reciprocity law of Bloch and
Kato [BK] Thm. 2.1 to our situation expressing the Bloch-Kato exponential map for L(χrLT )
in terms of generalized Coates-Wiles homomorphisms, where the Lubin-Tate characater χLT
describes the Galois action on T.
1. Introduction
The invention of Coates-Wiles homomorphisms and Coleman [Col] power series - practically
at the same time, actually the Coleman power series for elliptic units show already up in [CW]
- were the starting point of a range of new and important developments in arithmetic and
they have not lost their impact and fascination up to now. In order to recall it we shall first
introduce some notation.
Consider a finite extension L of Qp and fix a Lubin-Tate formal group LT over the integers
oL (with uniformizer πL ). By t0 = (t0,n ) we denote a generator of the Tate module T of LT
as oL -module. The πLn -division points generate a tower of Galois extensions Ln = L(LT [πLn ])
of L the union of which we denote by L∞ with Galois group ΓL . Coleman assigned to any
norm compatible unit u = (un ) ∈ limn L× a Laurent series gu,t0 ∈ oL ((Z))× such that
←− n
gu,t0 (t0,n ) = un for all n. If ∂inv denotes the invariant derivation with respect to LT , then, for
r ≥ 1, the rth Coates-Wiles homomorphism is given by
1 r 1 r−1 ∂inv gu,t0 (Z)
r
ψCW : lim L×
n → L(χrLT ), u 7→ ∂inv log gu,t0 (Z)|Z=0 := ∂inv |Z=0 ,
←− r! r! gu,t0 (Z)
n
it is Galois invariant and satisfies at least heuristically - setting tLT = logLT (Z) - the equation
X
r
log gu,t0 (Z) = ψCW (u)trLT
r
1991 Mathematics Subject Classification. 11Sxx.

Key words and phrases. p-adic Hodge theory, explicit reciprocity law, Coates-Wiles homomorphisms, Lubin-
Tate formal groups, ramified Witt vectors, Artin-Schreier pairing, Schmidt-Witt residue formula, Coleman
power series.
1
2 PETER SCHNEIDER AND OTMAR VENJAKOB
the meaning of which in p-adic Hodge theory has been crucially exploited, e.g. [Fo2], [Co5],
[CC].
Explicitly or implicitly this mysterious map plays - classically for the multiplicative group
over Qp - a crucial role in the Bloch-Kato (Tamagawa number) conjecture [BK], in the study
of special L-values [CS], in explicit reciprocity laws [Kat], [Wi] and even in the context of the
cyclotomic trace map from K-theory into topological cyclic homology for Zp [BM].
In this context one motivation for the present work is to understand Kato’s (and hence
Wiles’s) explicit reciprocity law in terms of (ϕL , ΓL )-modules. Since in the classical situa-
tion a successful study of explicit reciprocity laws has been achieved by Colmez, Cherbon-
nier/Colmez, Benois and Berger using Fontaine’s work on (ϕ, Γ)-modules and Herr’s cal-
culation of Galois cohomology by means of them, the plan for this article is to firstly use
Kisin-Ren/Fontaine’s equivalence of categories (recalled in section 4) to find a description
of Iwasawa cohomology HIw i (L , V ) for the tower L
∞ ∞ and a (finitely generated oL -module)
representation V of GL in terms of a ψ-operator acting on the etale (ϕL , ΓL )-module DLT (V ).
To this aim we have to generalize parts of [Her] in section 3, in particular the residue pairing,
which we relate to Pontrjagin duality. But instead of using Herr-complexes (which one also
could define easily in this context) we use local Tate-duality HIwi (L , V ) ∼ H 2−i (L , V ∨ (1))
∞ = ∞
for V being an oL -module of finite length and an explicit calculation of the latter groups
in terms of (ϕL , ΓL )-modules inspired by [Sch] and [Fo1], [Fo2]. Using the key observa-
tion that DLT (V )∨ ∼ = DLT (V ∨ (χLT )) (due to the residue pairing involving differentials
1 ∼
Ω = DLT (oL )(χLT ) and the compatibility of inner Homs under the category equivalence) we
finally establish in Theorem 5.13 the following exact sequence
1 L ψ −1 2
0 −→ HIw (L∞ , V ) −→ DLT (V (τ )) −−−→ DLT (V (τ )) −→ HIw (L∞ , V ) −→ 0
as one main result of this article, where the twist by τ = χLT χ−1 cyc is a new phenomenon
(disappearing obviously in the cyclotomic case) arising from the joint use of Pontrjagin and
local Tate duality. The second main result is the explicit calculation of the twisted (by the
oL -dual T ∗ of T ) Kummer map
κ⊗ T∗
(L∞ , Zp (1)) ⊗Zp T ∗ ∼
Zp
lim L× ⊗ T ∗ −−−−−→ HIw
1 1
= HIw (L∞ , oL (τ ))
←− n Z
n
in terms of Coleman series (recalled in section 2) and (ϕL , ΓL )-modules, see Theorem 6.2,
which generalizes the explicit reciprocity laws of Benois and Colmez. Inspired by [Fo1], [Fo2]
we reduce its proof to an explicit reciprocity law, Proposition 6.3, in characteristic p, which
in turn is proved by the Schmid-Witt residue formula which we generalize to our situation
in section 7, see Theorem 7.16. In section 8 we generalize the approach sketched in [Fo2] to
prove in Theorem 8.7 a generalization of the explicit reciprocity law of Bloch and Kato [BK]
Thm. 2.1: again in this context the Bloch-Kato exponential map is essentally given by the
Coates-Wiles homomorphism. As a direct consequence we obtain as Corollary 8.8 a new proof
for a special case of Kato’s explicit reciprocity law for Lubin-Tate formal groups.
It is a great honour and pleasure to dedicate this work to John Coates who has been a
source of constant inspiration for both of us.
We thank L. Berger for pointing Lemma 8.3.i out to us and for a discussion about Prop.
1.1 and G. Kings for making available to us a copy of [Koe]. We also acknowledge support
by the DFG Sonderforschungsbereich 878 at Münster and by the DFG research unit 1920
“Symmetrie, Geometrie und Arithmetik” at Heidelberg/Darmstadt.
COATES-WILES HOMOMORPHISMS AND IWASAWA COHOMOLOGY 3
Notation
Let Qp ⊆ L ⊂ Cp be a field of finite degree d over Qp , oL the ring of integers of L, πL ∈ oL
a fixed prime element, kL = oL /πL oL the residue field, and q := |kL |. We always use the
absolute value | | on Cp which is normalized by |πL | = q −1 .
We fix a Lubin-Tate formal oL -module LT = LTπL over oL corresponding to the prime
element πL . We always identify LT with the open unit disk around zero, which gives us a
global coordinate Z on LT . The oL -action then is given by formal power series [a](Z) ∈ oL [[Z]].
For simplicity the formal group law will be denoted by +LT .
The power series ∂(X+ LT Y )
∂Y |(X,Y )=(0,Z)
is a unit in oL [[Z]] and we let gLT (Z) denote its
inverse. Then gLT (Z)dZ is, up to scalars, the unique invariant differential form on LT ([Haz]
§5.8). We also let logLT (Z) = Z + . . . denote the unique formal power series in L[[Z]] whose
formal derivative is gLT . This logLT is the logarithm of LT ([Lan] 8.6). In particular, gLT dZ =
d logLT . The invariant derivation ∂inv corresponding to the form d logLT is determined by
f 0 dZ = df = ∂inv (f )d logLT = ∂inv (f )gLT dZ
and hence is given by
−1 0
f:inv (1) ∂inv (f ) = gLT f .
For any a ∈ oL we have
f:dlog (2) logLT ([a](Z)) = a · logLT and hence agLT (Z) = gLT ([a](Z)) · [a]0 (Z)
([Lan] 8.6 Lemma 2).
Let T be the Tate module of LT . Then T is a free oL -module of rank one, and the action
of GL := Gal(L/L) on T is given by a continuous character χLT : GL −→ o× 0
L . Let T denote
the Tate module of the p-divisible group Cartier dual to LT , which again is a free oL -module
of rank one. The Galois action on T 0 is given by the continuous character τ := χcyc · χ−1 LT ,
where χcyc is the cyclotomic character.
For n ≥ 0 we let Ln /L n
S denote the extension (in Cp ) generated by the πL -torsion points of
LT , and we put L∞ := n Ln . The extension L∞ /L is Galois. We let ΓL := Gal(L∞ /L) and
∼
=
→ o×
HL := Gal(L/L∞ ). The Lubin-Tate character χLT induces an isomorphism ΓL − L.
2. Coleman power series

sec:Coleman
We recall the injective ring endomorphism
ϕL : oL [[Z]] −→ oL [[Z]]
f (Z) 7−→ f ([πL ](Z)) .
In order to characterize its image we let LT1 denote the group of πL -torsion points of LT .
According to [Col] Lemma 3 we have
im(ϕL ) = {f ∈ oL [[Z]] : f (Z) = f (a +LT Z) for any a ∈ LT1 }.
This leads to the existence of a unique oL -linear endomorphism ψCol of oL [[Z]] such that
X
ϕL ◦ ψCol (f )(Z) = f (a +LT Z) for any f ∈ oL [[Z]]
a∈LT1
([Col] Thm. 4 and Cor. 5) as well as of a unique multiplicative map N : oL [[Z]] −→ oL [[Z]]
such that
Y
ϕL ◦ N (f )(Z) = f (a +LT Z) for any f ∈ oL [[Z]]
a∈LT1
([Col] Thm. 11).

The group ΓL acts continuously on oL [[Z]] via
:Gamma-action (3) ΓL × oL [[Z]] −→ oL [[Z]]
(4) (γ, f ) 7−→ f ([χLT (γ)](Z))
([Col] Thm. 1).

Z-inverse Remark 2.1. i. ψCol ◦ ϕL = q.
ii. ψCol ([πL ] · f ) = ZψCol (f ) for any f ∈ oL [[Z]].
iii. N ([πL ]) = Z q .
Proof. Because of the injectivity of ϕL it suffices in all three cases to verify the asserted
identity after applying ϕL . i. We compute
X X
ϕL ◦ ψCol ◦ ϕL (f ) = (ϕL f )(a +LT Z) = f ([πL ](a +LT Z))
a∈LT1 a∈LT1
X
= f ([πL ](Z)) = ϕL (qf ).
a∈LT1
ii. We compute
X
(ϕL ◦ ψCol )([πL ]f ) = [πL ](a +LT Z)f (a +LT Z)
a∈LT1
X
= [πL ](Z) f (a +LT Z) = ϕL (Z)(ϕL ◦ ψCol )(f )
a∈LT1
= ϕL (ZψCol (f )) .
ii. We omit the entirely analogous computation.
We observe that for any f ∈ oL ((Z)) = oL [[Z]][Z −1 ] there is an n(f ) ≥ 1 such that
[πL ]n(f ) · f ∈ oL [[Z]]. The above remark therefore allows to extend ψCol to an oL -linear
endomorphism
ψCol : oL ((Z)) −→ oL ((Z))
f 7−→ Z −n(f ) ψCol ([πL ]n(f ) f )
and to extend N to a multiplicative map
N : oL ((Z)) −→ oL ((Z))
f 7−→ Z −qn(f ) N ([πL ]n(f ) f ) .
We choose an oL -generator t0 of T . This is a sequence of elements t0,n ∈ πLn oLn such that
[πL ](t0,n+1 ) = t0,n for n ≥ 1, [πL ](t0,1 ) = 0, and t0,1 6= 0.
Coleman Theorem 2.2 (Coleman). For any norm-coherent sequence u = (un )n ∈ limn L× there is a
←− n
×
unique Laurent series gu,t0 ∈ (oL ((Z)) )N =1 such that gu,t0 (t0,n ) = un for any n ≥ 1. This
defines a multiplicative isomorphism
∼
=
lim L× −−→ (oL ((Z))× )N =1
←− n
n
u 7−→ gu,t0 .
Proof. See [Col] Thm. A and Cor. 17.

∼
=
nge-generator Remark 2.3. i. The map (oL ((Z))× )N =1 −
→ kL ((Z))× given by reduction modulo πL
is an isomorphism; hence
∼
=
lim L× −−→ kL ((Z))×
←− n
n
u 7−→ gu,t0 mod πL
is an isomorphism of groups.
ii. If t1 = ct0 is a second oL -generator of T then gu,t1 ([c](Z)) = gu,t0 (Z) for any u ∈
limn L× .
←− n
Proof. i. [Col] Cor. 18. ii. This is immediate from the characterizing property of gu,t0 in the
theorem.
We now introduce the “logarithmic” homomorphism

∆LT : oL [[Z]]× −→ oL [[Z]]
∂inv (f ) 0
−1 f
f 7−→ = gLT ,
f f
whose kernel is o×
L.
Delta-N Lemma 2.4. i. ∆LT ◦ ϕL = πL ϕL ◦ ∆LT .

ii. ψCol ◦ ∆LT = πL ∆LT ◦ N .
Proof. We begin with a few preliminary observations. From (2) we deduce
[πL ]0
f:pre1 (5) gLT = ϕL (gLT ) .
πL
Secondly we have
f:pre2 (6) d
dZ ϕL (f (Z)) = d
dZ f ([πL ](Z)) = f 0 ([πL ](Z))[πL ]0 (Z) = ϕL (f 0 )[πL ]0
for any f ∈ oL [[Z]]. Finally, the fact that gLT (Z)dZ is an invariant differential form implies
that
d
f:pre3 (7) gLT = dZ logLT (a +LT Z) for any a ∈ LT1 .
For i. we now compute
d
∆LT ◦ ϕL (f ) = 1 dZ ϕL (f (Z)) = [πL ]0 ϕL (f 0 )
= πL ϕL (f 0 )
= πL ϕL ◦ ∆LT (f ) ,
gLT ϕL (f ) gLT ϕL (f ) ϕL (gLT ) ϕL (f )
where the second, resp. the third, identity uses (6), resp. (5). For ii. we compute
f0
X
1
ϕL ◦ ψCol ◦ ∆LT (f ) = gLT (a+LT Z) f (a +LT Z)
a∈LT1
d
dZ f (a+LT Z)
X
1 1
= gLT (a+LT Z) f (a+LT Z) d
dZ (a+LT Z)
a∈LT1
d
dZ f (a+LT Z)
X
1
= d f (a+LT Z)
a∈LT1 dZ logLT (a+LT Z)
X Y
= ∆LT (f (a +LT Z)) = ∆LT ( f (a +LT Z))
a∈LT1 a∈LT1
= ∆LT ◦ ϕL ◦ N (f ) = πL ϕL ◦ ∆LT ◦ N (f )
= ϕL (πL ∆LT ◦ N (f )) ,
where the fourth, resp. the seventh, identity uses (7), resp. part i. of the assertion.
It follows that ∆LT restricts to a homomorphism
∆LT : (oL [[Z]]× )N =1 −→ oL [[Z]]ψCol =πL .
Its kernel is the subgroup µq−1 (L) of (q − 1)th roots of unity in o×

L.
On the other hand ∆LT obviously extends to the homomorphism
∆LT : oL ((Z))× −→ oL ((Z))
−1 f0
f 7−→ gLT ,
f
with the same kernel o×
L.
ta-N-extended Lemma 2.5. The identity ψCol ◦ ∆LT = πL ∆LT ◦ N holds true on oL ((Z))× .
Proof. Let f ∈ oL ((Z)× be any element. It can be written f = Z −n f0 with f0 ∈ oL [[Z]]× .
Then
1
ψCol ◦ ∆LT (f ) = −nψCol ( ) + ψCol ◦ ∆LT (f0 )
ZgLT
and
πL ∆LT ◦ N (f ) = −nπL ∆LT (N (Z)) + πL ∆LT ◦ N (f0 ) .
The second summands being equal by Lemma 2.4.ii we see that we have to establish that
1
ψCol ( ) = πL ∆LT (N (Z)) .
ZgLT
d
[πL ] πL N (Z)
By definition the left hand side is Z −1 ψCol ( ZgLT
) and the right hand side is gLT
dZ
N (Z) . Hence
we are reduced to proving the identity
[πL ] d
gLT N (Z)ψCol ( ) = πL Z N (Z) ,
ZgLT dZ
which is an identity in oL [[Z]] and therefore can be checked after applying ϕL . On the left
hand side we obtain
Y X [πL ](b +LT Z)
ϕL (gLT ) (a +LT Z)
(b +LT Z)gLT (b +LT Z)
a∈LT1 b∈LT1
Y X 1
= ϕL (gLT )ϕL (Z) (a +LT Z)
(b +LT Z)gLT (b +LT Z)
a∈LT1 b∈LT1
πL Y X 1
= gLT 0
ϕL (Z) (a +LT Z) ,
[πL ] (b +LT Z)gLT (b +LT Z)
a∈LT1 b∈LT1
where the second equality uses (5). On the right hand side, using (6), we have
d πL d πL d
Y
πL ϕL (Z)ϕL ( dZ N (Z)) = ϕ L (Z) dZ ϕ L (N (Z)) = ϕL (Z) dZ (a +LT Z) .
[πL ]0 [πL ]0
a∈LT1
This further reduces us to proving that

dQ
1 (a +LT Z)
Q a∈LT1
X
dZ
gLT = .
(b +LT Z)gLT (b +LT Z) a∈LT1 (a +LT Z)
b∈LT1
The invariance of gLT (Z)dZ implies

d gLT (Z)
f:pre4 (8) dZ (a +LT Z) = .
gLT (a +LT Z)
We see that the above right hand side, indeed, is equal to
d Q X d (a +LT Z)
(a +LT Z) 1
Q a∈LT1
X
dZ dZ
= = gLT .
a∈LT1 (a +LT Z) a +LT Z
a∈LT1
(a +LT Z)gLT (a +LT Z)
a∈LT1

Hence we even have the homomorphism ∆LT : (oL ((Z))× )N =1 −→ oL ((Z))ψCol =πL with
kernel µq−1 (L).
3. Etale (ϕL , ΓL )-modules

sec:etale
We define the ring AL to be the πL -adic completion of oL [[Z]][Z −1 ] and we let BL :=
AL [πL−1 ] denote the field of fractions of AL . The ring endomorphism ϕL of oL [[Z]] maps Z to
[πL ](Z). Since [πL ](Z) ≡ Z q mod πL the power series [πL ](Z) is a unit in AL . Hence ϕL extends
to a homomorphism oL [[Z]][Z −1 ] −→ AL and then by continuity to a ring endomorphism
ϕL of AL and finally to an embedding of fields ϕL : BL −→ BL . Similarly the invariant
derivation ∂inv first extends algebraically to oL [[Z]][Z −1 ], then by continuity to AL , and
finally by linearity to BL . Evidently we still have (1) for any f ∈ BL .
basis Remark 3.1. 1, Z, . . . , Z q−1 is a basis of BL as a ϕL (BL )-vector space.
Proof. See [FX] Remark before Lemma 2.1 or [GAL] Prop. 1.7.3.
This remark allows us to introduce the unique additive endomorphism ψL of BL which
satisfies
ϕL ◦ ψL = πL−1 · traceBL /ϕL (BL ) .
By the injectivity of ϕL and the linearity of the field trace we have the projection formula
ψL (ϕL (f1 )f2 ) = f1 ψL (f2 ) for any fi ∈ BL .
as well as the formula

q
ψL ◦ ϕL = · id .
πL
Correspondingly, we consider the unique multiplicative map NL : BL −→ BL which satis-
fies
:normoperator (9) ϕL ◦ NL = NormBL /ϕL (BL ) .
psi Remark 3.2. i. ψL (AL ) ⊆ AL and NL (AL ) ⊆ AL .

ii. On oL ((Z)) we have ψL = πL−1 · ψCol and NL = N .
iii. ϕL ◦ ψL ◦ ∂inv = ∂inv ◦ ϕL ◦ ψL on BL .
iv. NL (f )([c](Z)) = NL (f ([c](Z))) for any c ∈ o×
L and f ∈ BL .
v. NL (f ) ≡ f mod πL AL for any f ∈ AL .
vi. If f ∈ AL satisfies f ≡ 1 mod πLm AL for some m ≥ 1 then NL (f ) ≡ 1 mod πLm+1 AL .
vii. (oL ((Z))× )N =1 = (AL× )NL =1 .
Proof. i. The homomorphism ϕL induces on AL /πL AL = kL ((Z)) the injective q-Frobenius

map. It follows that ϕ−1
L (AL ) = AL . Hence the assertion reduces to the claim that
f:tracemod (10) traceBL /ϕL (BL ) (AL ) ⊆ πL AL .
But the trace map traceBL /ϕL (BL ) induces the trace map for the purely inseparable extension
kL ((Z))/kL ((Z q )), which is the zero map.
ii. For any a ∈ LT1 we have the ring homomorphism
σa : oL [[Z]] −→ oL1 [[Z]] ⊆ AL1

f (Z) 7→ f (a +LT Z) .
Since σa (Z) = a +LT Z ≡ a + Z mod deg 2 we have σa (Z) ∈ AL×1 , so that σa extends to
oL [[Z]][Z −1 ]. By continuity σa further extends to AL and then by linearity to an embedding
of fields
σa : BL −→ BL1 = BL L1 .
Clearly these σa are pairwise different. Moreover, for any f ∈ oL [[Z]], we have
σa ◦ ϕL (f )(Z) = f ([πL ](a +LT Z)) = f (Z) .

Q
We conclude, by continuity, that σa |ϕL (BL ) = id. It follows that a∈LT1 (X − σa (f )), for any
f ∈ BL , is the characteristic polynomial of f over ϕL (BL ). Hence
X
f:trace (11) traceBL /ϕL (BL ) (f ) = σa (f ) ,
a∈LT1
which proves the assertion for f ∈ oL [[Z]]. For general f ∈ oL ((Z)), using the notation and
definition before Thm. 2.2 we compute

ϕL ◦ ψCol (f ) = ϕL Z −n(f ) ψCol ([πL ]n(f ) f )
X
= ϕL (Z)−n(f ) σa ([πL ]n(f ) f )
a∈LT1
X
= σa (f ) = traceBL /ϕL (BL ) (f )
a∈LT1
= ϕL ◦ πL ψL (f ) .
The proof for NL is completely analogous.

iii. By the invariance of ∂inv we have ∂inv ◦ σa = σa ◦ ∂inv on oL [[Z]][Z −1 ], whence on BL
by continuity and linearity. Therefore, the claim follows from (11).
iv. We compute
ϕL (NL (f )([c](Z))) = NL (f )([c]([πL ](Z))) = NL (f )([πL ]([c](Z)))

Y
= ϕL (NL (f ))([c](Z)) = σa (f )([c](Z))
a∈LT1
Y Y
= σ[c−1 ](a) (f ([c](Z))) = σa (f ([c](Z)))
a∈LT1 a∈LT1
= ϕL (NL (f ([c](Z)))) .
v. We have
ϕL ◦ NL (f ) mod πL AL = NormkL ((Z))/kL ((Z q )) (f mod πL AL ) ≡ f q mod πL AL

= ϕL (f ) mod πL AL .
vi. Let f = 1 + πLm g with g ∈ AL . We compute

Y X
ϕL (NL (1 + πLm g)) = 1 + πLm σa (g) ≡ 1 + πLm ( σa (g)) mod πLm+1 AL
a∈LT1 a∈LT1
≡ 1 mod πLm+1 AL ≡ ϕL (1) mod πLm+1 AL
where the third identity uses (10). The assertion follows since ϕL remains injective modulo
πLj for any j ≥ 1.
vii. We have the commutative diagram
⊆
(oL ((Z))× )N =1 / (A × )NL =1
L
∼
= ' x
kL ((Z))×
where the oblique arrows are given by reduction modulo πL . The left one is an isomorphism
by Remark 2.3.i. The right one is injective as a consequence of the assertion vi. Hence all
three maps must be bijective.
Due to Remark 3.2.vii we may view the Coleman isomorphism in Thm. 2.2 as an isomor-
phism
∼
=
f:Coleman (12) lim L× −−→ (AL× )NL =1 .
←− n
n
We always equip AL with the weak topology, for which the oL -submodules πLm AL +
Z mo L [[Z]], for m ≥ 1, form a fundamental system of open neighbourhoods of zero. The
weak topology on any finitely generated AL -module M is defined to be the quotient topology,
with respect to any surjective homomorphism ALn M , of the product topology on ALn ;
this is independent of the choice of this homomorphism. We have the following properties (cf.
[SV] Lemmas 8.2 and 8.22 for a detailed discussion of weak topologies):
– AL is a complete Hausdorff topological oL -algebra (with jointly continuous multipli-
cation).
– AL induces on oL [[Z]] its compact topology.
– M is a complete Hausdorff topological module (with jointly continuous scalar multi-
plication).
– M/πLm M , for any m ≥ 1, is locally compact.
phi-weakcont Remark 3.3. The endomorphisms ϕL and ψL of AL are continuous for the weak topology.
Proof. For ϕL see [GAL] Prop. 1.7.8.i. For ψL see [FX] Prop. 2.4(b) (note that their ψ is our
πL
q ψL ).
Let Ω1 = Ω1AL = AL dZ denote the free rank one AL -module of differential forms. Obviously
the residue map
Res : Ω1 −→ oL
X
( ai Z i )dZ 7−→ a−1
i
is continuous. Later on in section 7 it will be a very important fact that this map does not
depend on the choice of the variable Z. For the convenience of the reader we explain the
argument (cf. [Fo1] A2.2.3). First of all we have to extend the maps d and Res by linearity
to maps
d Res
BL −→ Ω1BL := L ⊗oL Ω1AL −−→ L .
Only for the purposes of the subsequent remark we topologize BL by taking as a fundamental
system of open neighbourhoods of zero the oL [[Z]]-submodules
πLm AL + L ⊗oL Z m oL [[Z]]) for m ≥ 1.
Using the isomorphism Ω1BL = BL dZ ∼ = BL we also make Ω1BL into a topological oL -module.
It is easy to see that the maps d and Res are continuous.
Res-variable Remark 3.4. i. d(BL ) is dense in ker(Res).
ii. Res = ResZ does not depend on the choice of the variable Z, i.e., if Z 0 is any element
in AL whose reduction modulo πL is a uniformizing element in k((Z)), then ResZ (ω) =
ResZ 0 (ω) for all ω ∈ Ω1BL .
Proof. i. On the one hand L[Z, Z −1 ] ∩ ker(Res) is dense in ker(Res). On the other hand we
have L[Z, Z −1 ] ∩ ker(Res) ⊆ d(BL ). ii. As a consequence of i. both maps ResZ and ResZ 0 have
0
the same kernel. It therefore suffices to show that ResZ ( dZ 0
Z 0 ) = 1. We have Z = cZη(1 + πL α)
×
with c ∈ oL , η ∈ 1 + ZoL [[Z]], and α ∈ AL . Hence
dZ 0 dZ dη d(1+πL α)
Z0 = Z + η + 1+πL α .
Clearly ResZ ( dη m
η ) = 0. Furthermore, if m ≥ 1 is sufficiently big, then log(1 + πL β), for any
β ∈ AL , converges in AL . Since (1 + π AL )/(1 + π AL ) ∼
j
L
j+1
= AL /πL AL , for any j ≥ 1, we
L
m m β)
d(1+πL
have (1 + πL α)p = 1 + πLm β for some β ∈ AL . It follows that pm d(1+πL α)
1+πL α = 1+πLmβ =
d(log(1 + πLm β)) and therefore that ResZ ( d(1+π L α)
1+πL α ) = 0.
Since Ω1 is a topological AL -module it follows that the residue pairing
sidue-pairing (13) AL × Ω1 −→ oL
(f, ω) 7−→ Res(f ω)
is jointly continuous. It induces, for any m ≥ 1, the continuous pairing
AL /πLm AL × Ω1 /πLm Ω1 −→ L/oL
(f, ω) 7−→ πL−m Res(f ω) mod oL
and hence (cf. [B-TG] X.28 Thm. 3) the continuous oL -linear map
f:dual (14) Ω1 /πLm Ω1 −→ HomcoL (AL /πLm AL , L/oL )
ω 7−→ [f 7→ πL−m Res(f ω) mod oL ] ,
where HomcoL denotes the module of continuous oL -linear maps equipped with the compact-
open topology. For the convenience of the reader we recall the following well known fact.
dual Lemma 3.5. The map (14) is an isomorphism of topological oL -modules.
Proof. Let R := oL /πLm oL . It is convenient to view the map in question as the map
R((Z))dZ −→ HomcR (R((Z)), R)
ω 7−→ `ω (f ) := Res(f ω) .
One easily checks that ω = i `ω (Z −i−1 )Z iP
P
dZ. Hence injectivity is clear. If ` is an arbitrary
element in the right hand side we put ω := i `(Z −i−1 )Z i dZ. The continuity of ` guarantees
that `(Z i ) = 0 for any sufficiently big i. Hence ω is a well defined preimage of ` in the left
hand side. Finally, the map is open since
{f ∈ R((Z)) : Res(f Z n R[[Z]]dZ) = 0} = Z −n R[[Z]]
is compact for any n ≥ 1.
For an arbitrary AL -module N we have the adjunction isomorphism
∼
=
f:adjunction (15) HomAL (N, HomoL (AL , L/oL )) −−→ HomoL (N, L/oL )
F 7−→ F (.)(1) .
op-adjunction Lemma 3.6. For any finitely generated AL /πLm AL -module M the adjunction (15) together
with (14) induces the topological isomorphism
∼
=
HomAL (M, Ω1 /πLm Ω1 ) −−→ HomcoL (M, L/oL )
F 7−→ πL−m Res(F (.)) mod oL .
Proof. It is clear that (15) restricts to an injective homomorphism

op-adjunction (16) HomAL (M, HomcoL (AL /πLm AL , L/oL )) −→ HomcoL (M, L/oL ) .
The inverse of (15) sends ` ∈ HomoL (M, L/oL ) to F (m)(f ) := `(f m) and visibly restricts to
an inverse of (16). By inserting (14) we obtain the asserted algebraic isomorphism. To check
that it also is a homeomorphism we first clarify that on the left hand side we consider the
weak topology of HomAL (M, Ω1 /πLm Ω1 ) as a finitely generated AL -module. The elementary
divisor theorem for the discrete valuation ring AL implies that M is isomorphic to a finite
direct product of modules of the form AL /πLn AL with 1 ≤ n ≤ m. It therefore suffices to
consider the case M = AL /πLn AL . We then have the commutative diagram of isomorphisms
HomAL (AL /πLn AL , Ω1 /πLm Ω1 )
=
,
πLm−n Ω1 /πLm Ω1 HomcoL (AL /πLn AL , L/oL )
O 2
m−n
πL ·
(14)
Ω1 /πLn Ω1
By Lemma 3.5 all maps in this diagram except possibly the upper oblique arrow, which is
the map in the assertion, are homeomorphisms. Hence the oblique arrow must be a homeo-
morphism as well.
The ΓL -action (3) on oL [[Z]] extends, by the same formula, to a ΓL -action on AL which,
moreover, is continuous for the weak topology (see [GAL] Prop. 1.7.8.ii).
def:modR Definition 3.7. A (ϕL , ΓL )-module M (over AL ) is a finitely generated AL -module M to-
gether with
– a ΓL -action on M by semilinear automorphisms which is continuous for the weak
topology 1, and
– a ϕL -linear endomorphism ϕM of M which commutes with the ΓL -action.
It is called etale if the linearized map
∼
=
M : AL ⊗AL ,ϕL M −
ϕlin −→ M
f ⊗ m 7−→ f ϕM (m)
is bijective. We let Met (AL ) denote the category of etale (ϕL , ΓL )-modules M over AL .
milinear-cont Remark 3.8. Let α : AL −→ AL be a continuous ring homomorphism, and let β : M −→ M
be any α-linear endomorphism of a finitely generated AL -module M ; then β is continuous for
the weak topology on M .
Proof. The map
β lin : AL ⊗AL ,α M −→ M
f ⊗ m 7−→ f β(m)
1In case L = Q we have automatic continuity. The simplest instance of this is the fact that any abstract
p
group homomorphism ΓQp −→ Z× p is continuous. By restricting to sufficiently small open subgroups this
reduces to the claim that any abstract group homomorphism Zp −→ Zp is continuous, i.e., is determined by
its value in 1. This follows from the triviality of any group homomorphism Zp /Z −→ Zp . The latter holds
because, by the surjectivity of the projection map Z −→ Zp /pZp , the group Zp /Z is p-divisible.
is AL -linear. We pick a free presentation λ : ALn M . Then we find an AL -linear map βe

such that the diagram
αn βe
ALn / A n = AL ⊗A ,α A n /An
L L L L
λ id ⊗λ λ

m7→1⊗m β lin
M / AL ⊗A ,α M /7 M
L
is commutative. All maps except possibly the lower left horizontal arrow are continuous.
The universal property of the quotient topology then implies that β must be continuous as
well.
Remarks 3.3 and 3.8 imply that the endomorphism ϕM of a (ϕL , ΓL )-module M is contin-
uous.
On any etale (ϕL , ΓL )-module M we have the oL -linear endomorphism
(ϕlin )−1
M
ψM : M −−−−−→ AL ⊗AL ,ϕL M −→ M
f ⊗ m 7−→ ψL (f )m ,
which, by construction, satisfies the projection formulas
ψM (ϕL (f )m) = f ψM (m) and ψM (f ϕM (m)) = ψL (f )m ,
for any f ∈ AL and m ∈ M , as well as the formula
q
ψM ◦ ϕM = · idM .
π
Remark 3.3 is easily seen to imply that ψM is continuous for the weak topology.
For technical purposes later on we need to adapt part of Colmez’s theory of treillis to our
situation. We will do this in the following setting. Let M be a finitely generated AL -module
(always equipped with its weak topology) such that πLn M = 0 for some n ≥ 1; we also assume
that M is equipped with a ϕL -linear endomorphism ϕM which is etale, i.e., such that ϕlin M is
bijective.
Definition 3.9. A treillis N in M is an oL [[Z]]-submodule N ⊆ M which is compact and
such that its image in M/πL M generates this kL ((Z))-vector space.
treillis Remark 3.10. i. If e1 , . . . , ed are AL -generators of M then oL [[Z]]e1 + . . . + oL [[Z]]ed
is a treillis in M .
ii. A compact oL [[Z]]-submodule N of M is a treillis if and only if it is open.
iii. For any two treillis N0 ⊆ N1 in M the quotient N1 /N0 is finite; in particular, any
intermediate oL [[Z]]-submodule N0 ⊆ N ⊆ N1 is a treillis as well.
Proof. Part i. is obvious from the compactness of oL [[Z]]. For ii. and iii. see [Co3] Prop.
I.1.2(i).
Following Colmez we define
i→∞
M ++ := {m ∈ M : ϕiM (m) −−−→ 0}.
Since oL [[Z]] is compact it is easily seen that M ++ is an oL [[Z]]-submodule of M . Obviously
M ++ is ϕM -invariant.
M++ Lemma 3.11. i. M ++ is a treillis.

ii. ϕM − 1 is an automorphism of M ++ .
Proof. i. Using Remark 3.10.i/iii this follows from [Co3] Lemma II.2.3. This lemma is stated
and proved there in the cyclotomic situation. But the only property of ϕL , besides being etale,
which is used is that ϕL (Z) ∈ Z 2 oL [[Z]] + πL ZoL [[Z]].
ii. Obviously m = 0 is the only element in M ++ which satisfies ϕM P(m) = m. Now let
m ∈ M ++ be an arbitrary element. Since M is complete the series m0 := i ϕiM (m) converges
and satisfies (ϕM − 1)(−m0 ) = m. But M ++ is open and hence closed in M so that −m0 ∈
M ++ .
The following lemma is a slight generalization of a result of Fontaine (cf. [Her] Prop. 2.4.1).
top-strict Lemma 3.12. On any etale (ϕL , ΓL )-module M such that πLn M = 0 for some n ≥ 1 the map
ϕM − 1 is open and, in particular, is topologically strict.
Proof. As M ++ , by Lemma 3.11.i and Remark 3.10.ii, is compact and open in M we first see,
using Lemma 3.11.ii, that ϕM − 1 is a homeomorphism on M ++ and then that ϕM − 1 is an
open map.
The category Met (AL ) has an internal Hom-functor. For any two modules M and N in
Met (AL ) the AL -module HomAL (M, N ) is finitely generated. It is a (ϕL , ΓL )-module with
respect to
γ(α) := γ ◦ α ◦ γ −1 and ϕHomA (M,N ) (α) := ϕlin lin −1
N ◦ (idAL ⊗ α) ◦ (ϕM )
L
for any γ ∈ ΓL and any α ∈ HomAL (M, N ). We need to verify that the ΓL -action, indeed, is
continuous. This is a consequence of the following general facts.
e-convergence Remark 3.13. For any two finitely generated AL -modules M and N we have:
i. The weak topology on HomAL (M, N ) coincides with the topology of pointwise conver-
gence.
ii. The bilinear map
HomAL (M, N ) × M −→ N
(α, m) 7−→ α(m)
is continuous for the weak topology on all three terms.
Proof. Since any finitely generated module over the discrete valuation ring AL is a direct sum
of modules of the form AL or AL /πLj AL for some j ≥ 1, it suffices to consider the case that
M and N both are such cyclic modules. In fact, we may even assume that M = AL and
N = AL =: AL /πL∞ AL or N = AL /πLj AL . We then have the isomorphism of AL -modules
∼
=
ev1 : HomAL (AL , AL /πLj AL ) −−→ AL /πLj AL
α 7−→ α(1) .
For i. we have to show that this map is a homeomorphism for the topology of pointwise
convergence and the weak topology on the left and right term, respectively. The topology of
pointwise convergence is generated by the subsets C(f, U ) := {α ∈ HomAL (AL , AL /πLj AL ) :
α(f ) ∈ U }, where f ∈ AL and where U runs over open subsets U ⊆ AL /πLj AL (for the weak
topology). For the open subset Uf := {n ∈ AL /πLj AL : f n ∈ U } we have C(f, U ) = C(1, Uf ).
We see that ev1 (C(f, U )) = Uf .
Using the topological isomorphism ev1 the bilinear map in ii. becomes the multiplication
map AL /πLj AL ×AL −→ AL /πLj AL . It is continuous since, as noted earlier, AL is a topological
algebra for the weak topology.
Let (γi )i∈N in ΓL , resp. (αi )i∈N in HomAL (M, N ), be a sequence which converges to γ ∈
ΓL , resp. to α ∈ HomAL (M, N ) for the weak topology. We have to show that the sequence
(γi (αi ))i converges to γ(α) for the weak topology. By Remark 3.13.i it, in fact, suffices to
check pointwise convergence. Let therefore m ∈ M be an arbitrary element. As ΓL acts
continuously on M , we have limi7→∞ γi−1 (m) = γ −1 (m). The Remark 3.13.ii then implies that
limi7→∞ αi (γi−1 (m)) = α(γ −1 (m). By the continuity of the ΓL -action on N we finally obtain
that limi7→∞ γi (αi (γi−1 (m))) = γ(α(γ −1 (m))).
In order to check etaleness we use the linear isomorphisms ϕlin lin
M and ϕN to identify
HomAL (M, N ) and HomAL (AL ⊗AL ,ϕL M, AL ⊗AL ,ϕL N ) = HomAL (M, AL ⊗AL ,ϕL N ). Then
the linearized map ϕlin HomA (M,N ) becomes the map
L
AL ⊗AL ,ϕL HomAL (M, N ) −→ HomAL (M, AL ⊗AL ,ϕL N )

f ⊗ α 7−→ [m 7→ f ⊗ α(m)] .
To see that the latter map is bijective we may use, because of the flatness of ϕL as an injective
ring homomorphism between discrete valuation rings, a finite presentation of the module M in
order to reduce to the case M = AL , in which the bijectivity is obvious. Hence HomAL (M, N )
is an etale (ϕL , ΓL )-module (cf. [Fo1] A.1.1.7). One easily checks the validity of the formula
f:inner (17) ϕHomA (M,N ) (α)(ϕM (m)) = ϕN (α(m)) .

L
As a basic example we point out that Ω1 naturally is an etale (ϕL , ΓL )-module via
γ(dZ) := [χLT (γ)]0 (Z)dZ = d[χLT (γ)](Z) and ϕΩ1 (dZ) := πL−1 [πL ]0 (Z)dZ = πL−1 d[πL ](Z) .
Note that the congruence [πL ](Z) ≡ πL Z + Z q mod πL indeed implies that the derivative
[πL ]0 (Z) is divisible by πL . The simplest way to see that Ω1 is etale is to identify it with
another obviously etale (ϕL , ΓL )-module.
If χ : ΓL −→ o× L is any continuous character with representation module Wχ = oL tχ then,
for any M in Met (AL ), we have the twisted module M (χ) in Met (AL ) where M (χ) := M ⊗oL
Wχ as AL -module, ϕM (χ) (m⊗w) := ϕM (m)⊗w, and γ|M (χ)(m⊗w) := γ|M (m)⊗γ|Wχ (w) =
χ(γ) · γ|M (m) ⊗ w for γ ∈ ΓL . It follows that ψM (χ) (m ⊗ w) = ψM (m) ⊗ w. For the character
χLT we take WχLT = T = oL t0 and Wχ−1 = T ∗ = oL t∗0 as representation module, where T ∗
LT
denotes the oL -dual with dual basis t∗0 of t0 .
mega-as-twist Lemma 3.14. The map

∼
=
AL (χLT ) −−→ Ω1
f ⊗ t0 7−→ f d logLT = f gLT dZ
is an isomorphism of (ϕL , ΓL )-modules.
Proof. Since gLT is a unit in oL [[Z]] it is immediately clear that the map under consideration
is well defined and bijective. The equivariance follows from (2).
si-invariance Remark 3.15. For later applications we want to point out that for û ∈ (oL ((Z))× )N =1 the
differential form dûû is ψΩ1 -invariant: In fact using Lemma 3.14, Remark 3.2.ii, and Lemma
2.5 for the second, fourth, and fifth identity, respectively, we compute
dû
ψΩ1 ( ) = ψΩ1 (∆LT (û)d logLT ) = ψΩ1 (∆LT (û)ϕΩ1 (d logLT ))
û
= ψL (∆LT (û))d logLT = πL−1 ψCol (∆LT (û))d logLT
dû
= ∆LT (û)d logLT = .
û
d-phi-psi Lemma 3.16. The map d : AL −→ Ω1 satisfies:
i. πL · ϕΩ1 ◦ d = d ◦ ϕL ;
ii. γ ◦ d = d ◦ γ for any γ ∈ ΓL ;
iii. πL−1 · ψΩ1 ◦ d = d ◦ ψL .
Proof. i. For f ∈ AL we compute
ϕΩ1 (df ) = ϕΩ1 (f 0 dZ) = πL−1 f 0 ([πL ](Z))[πL ]0 (Z)dZ = πL−1 d(f ([πL ](Z))) = πL−1 d(ϕL (f )) .
ii. The computation is completely analogous to the one for i.
iii. Since ϕΩ1 is injective, the asserted identity is equivalent to ϕΩ1 ◦ ψΩ1 ◦ d = d ◦ ϕL ◦ ψL
by i. Lemma 3.14 implies that (ϕL ◦ ψL (f ))gLT dZ = ϕΩ1 ◦ ψΩ1 (f gLT dZ). Using this, (1), and
Remark 3.2.iii in the second, first and fourth, and third identity, respectively, we compute
ϕΩ1 ◦ ψΩ1 (df ) = ϕΩ1 ◦ ψΩ1 (∂inv (f )gLT dZ)
= (ϕL ◦ ψL (∂inv (f ))gLT dZ = ∂inv (ϕL ◦ ψL (f ))gLT dZ
= d(ϕL ◦ ψL (f )) .

Res-phi-psi Proposition 3.17. The residue map Res : Ω1 −→ L satisfies:
i. Res ◦ ϕΩ1 = πL−1 q · Res;
ii. Res ◦ γ = Res for any γ ∈ ΓL ;
iii. Res ◦ ψΩ1 = Res.
Proof. Of course, exact differential forms have zero residue. Let now α denote any of the
endomorphisms ϕΩ1 , γ, or ψΩ1 of Ω1 . Using Lemma 3.16 we have (m + 1)Res(α(Z m dZ)) =
Res(α(d(Z m+1 ))) = πL Res(dα(Z m+1 )) = 0 with ∈ {−1, 0, 1} and hence Res(α(Z m dZ)) = 0
for any m 6= −1. Since Res is continuous it follows that α preserves the kernel of Res. This
reduces us to showing the asserted identities on the differential form Z −1 dZ. In other words
we have to check that Res(α(Z −1 dZ)) = πL−1 q, 1, 1, respectively, in the three cases.
0
i. We have ϕΩ1 (Z −1 dZ) = πL−1 [π L ] (Z)
[πL ](Z) dZ. But [πL ](Z) = Z (1 + πL v(Z)) with v ∈ AL .
q
Hence ϕΩ1 (Z −1 dZ) = πL−1 qZ −1 dZ + πL−1 d(1+π L v)

1+πL v . In the proof of Remark 3.4.ii we have seen
d(1+πL v)
that 1+πL v has zero residue.
[χLT (γ)]0 (Z)
ii. Here we have γ(Z −1 dZ) = [χLT (γ)](Z) dZ. But [χLT (γ)](Z) = Zu(Z) with a unit u ∈
0
oL [[Z]]× .
It follows that γ(Z dZ) = Z −1 dZ + uu dZ.
−1 The second summand has zero residue,
of course.
iii. The identity in i. implies that Res ◦ ϕΩ1 = πL−1 q · Res = Res ◦ ψΩ1 ◦ ϕΩ1 . Hence the
identity in iii. holds on the image of ϕΩ1 (as well as on the kernel of Res). But in the course
of the proof of i. we have seen that Z −1 dZ ∈ im(ϕΩ1 ) + ker(Res).
phi-psi-dual Corollary 3.18. The residue pairing satisfies

Res(f ψΩ1 (ω)) = Res(ϕL (f )ω) for any f ∈ AL and ω ∈ Ω1 .
Proof. By the projection formula the left hand side of the asserted equality is equal to
Res(ψΩ1 (ϕL (f )ω)). Hence the assertion follows from Prop. 3.17.iii.
For a finitely generated AL /πLn AL -module M the isomorphism in Lemma 3.6 induces the
following pairing
[ , ] = [ , ]M : M × HomAL (M, Ω1 /πLn Ω1 ) −→ L/oL
(m, F ) 7−→ πL−n Res(F (m)) mod oL .
Since M is locally compact it is (jointly) continuous by [B-TG] X.28 Thm. 3. Note that this
pairing (and hence also the isomorphism in Lemma 3.6 ) is ΓL -invariant by Prop. 3.17.ii.
∼
=
The map HomAL (AL /πLn AL , Ω1 /πLn Ω1 ) − → Ω1 /πLn Ω1 which sends F to F (1) is an isomor-
phism of (etale) (ϕL , ΓL )-modules. Cor. 3.18 then implies that
[ϕAL /πLn AL (f ), F ]AL /πLn AL = [f, ψHomA n A ,Ω1 /π n Ω1 ) (F )]A /π n A
(AL /πL L L L L L
L
for all f ∈ AL /πLn AL and F ∈ HomAL (AL /πLn AL , Ω1 /πLn Ω1 ). More generally, we show:
pairing] Proposition 3.19. Let M be an etale (ϕL , ΓL )-module such that πLn M = 0 for some n ≥ 1;
we have:
i. The operator ψM is left adjoint to ϕHomA (M,Ω1 /πLn Ω1 ) under the pairing [ , ], i.e.,
L
[ψM (m), F ] = [m, ϕHomA n Ω1 ) (F )]

(M,Ω1 /πL
L
for all m ∈ M and all F ∈ HomAL (M, Ω1 /πLn Ω1 );

ii. the operator ϕM is left adjoint to ψHomA (M,Ω1 /πLn Ω1 ) under the pairing [ , ], i.e.,
L
[ϕM (m), F ] = [m, ψHomA n Ω1 ) (F )]

(M,Ω1 /πL
L
for all m ∈ M and all F ∈ HomAL (M, Ω1 /πLn Ω1 ).

Proof. For notational simplicity we abbreviate the subscript HomAL (M, Ω1 /πLn Ω1 ) to Hom.
i. Since M is etale it suffices to check the asserted identity on elements of the form f ϕM (m)
with f ∈ AL and m ∈ M . By the projection formula for ψM the left hand side then becomes
[ψL (f )m, F ]. We compute the right hand side:
[f ϕM (m), ϕHom (F )] ≡ πL−n Res(ϕHom (F )(f ϕM (m)))
≡ πL−n Res(f ϕHom (F )(ϕM (m)))
≡ πL−n Res(f ϕΩ1 /πLn Ω1 (F (m)))
≡ πL−n Res(ψΩ1 /πLn Ω1 (f ϕΩ1 /πLn Ω1 (F (m))))
≡ πL−n Res(ψL (f )F (m))
≡ πL−n Res(F (ψL (f )m))
≡ [ψL (f )m, F ] mod oL ;
here the first and last identities are just the definition of [ , ], the second and sixth use AL -
linearity, the third the formula (17), the fourth Prop. 3.17 iii., and the fifth the projection
formula for ψΩ1 /πLn Ω1 .
ii. Correspondingly we compute

[ϕM (m), f ϕHom (F )] ≡ πL−n Res(f ϕHom (F )(ϕM (m)))
≡ πL−n Res(f ϕΩ1 /πLn Ω1 (F (m)))
≡ πL−n Res(ψΩ1 /πLn Ω1 (f ϕΩ1 /πLn Ω1 (F (m))))
≡ πL−n Res(ψL (f )F (m))
≡ πL−n Res((ψHom (f ϕHom (F )))(m))
≡ [m, ψHom (f ϕHom (F ))] mod oL .

airing-rangle Remark 3.20. Similarly, for any etale (ϕL , ΓL )-module M such that πLn M = 0 one can
consider the ΓL -invariant (jointly) continuous pairing
[ , i = [ , iM : M × HomAL (M, AL (χLT )/πLn AL (χLT )) −→ L/oL
(m, F ) 7−→ πL−n Res(F (m)d logLT ) mod oL
which arises from [ , ]M by plugging in the isomorphism from Lemma 3.14. Clearly, it has
adjointness properties analogous to the ones in Prop. 3.19.
4. The Kisin-Ren equivalence

sec:KR
Let E e + := lim oC /poC with the transition maps being given by the Frobenius φ(a) = ap .
←− p p
We may also identify E e + with lim oC /πL oC with the transition maps being given by the
←− p p
q-Frobenius φq (a) = aq . Recall that E e + is a complete valuation ring with residue field Fp and
its field of fractions E
e = lim Cp being algebraically closed of characteristic p. Let m e denote
←− E
the maximal ideal in E e +.
The q-Frobenius φq first extends by functoriality to the rings of the Witt vectors W (E e +) ⊆
W (E)e and then oL -linearly to W (E e + )L := W (E e + ) ⊗o oL ⊆ W (E) e L := W (E)e ⊗o oL , where
L0 L0
L0 is the maximal unramified subextension of L. The Galois group GL obviously acts on E e

and W (E)L by automorphisms commuting with φq . This GL -action is continuous for the weak
e
topology on W (E) e L (cf. [GAL] Lemma 1.5.3). Let ML denote the ideal in W (E e + )L which is
the preimage of mE e under the residue class map.
Evaluation of the global coordinate Z of LT at πL -power torsion points induces a map
(not a homomorphism in the naive sense) ι : T −→ E e + . Namely, if t = (zn )n≥1 ∈ T with
q
[πL ](zn+1 ) = zn and [πL ](z1 ) = 0, then zn+1 ≡ zn mod πL and hence ι(t) := (zn mod πL )n ∈
e +.
E
iota Lemma 4.1. The image of the map ι is contained in mE
e . The map
ιLT : T −→ ML
t 7−→ lim ([πL ] ◦ φ−1 n
q ) ([ι(t)])) ,
n→∞
where [ι(t)] denotes the Teichmüller representative of ι(t), is well defined and satisfies:
a. [a](ιLT (t)) = ιLT (at) for any a ∈ oL ;
b. φq (ιLT (t)) = ιLT (πL t) = [πL ](ιLT (t));
c. σ(ιLT (t)) = [χLT (σ)](ιLT (t)) for any σ ∈ GL .
Proof. This is [KR] Lemma 1.2, which refers to [Co1] Lemma 9.3. For full details see [GAL]
§2.1.
As before we fix an oL -generator t0 of T and put ωLT := ιLT (t0 ). By sending Z to ωLT we
obtain an embedding of rings
e + )L .
oL [[Z]] −→ W (E
As explained in [KR] (1.3) it extends to embeddings of rings
AL −→ W (E)
e L and BL −→ L ⊗oL W (E)
e L.
The left map, in fact, is a topological embedding for the weak topologies on both sides ([GAL]
Prop. 2.1.14.i). The Galois group GL acts through its quotient ΓL on BL by (σ, f ) 7−→
f ([χLT (σ)](Z)). Then, by Lemma 4.1.c, the above embeddings are GL -equivariant. Moreover,
the q-Frobenius φq on L ⊗oL W (E) e L , by Lemma 4.1.b, restricts to the endomorphism f 7−→
f ◦ [πL ] of BL which we earlier denoted by ϕL .
We define AL to be the image of AL in W (E) e L . It is a complete discrete valuation ring
with prime element πL and residue field the image EL of kL ((Z)) ,→ E. e As a consequence
of Lemma 4.1.a this subring AL is independent of the choice of t0 . As explained above each
choice of t0 gives rise to an isomorphism
∼
=
f:isoVariousA (18) (AL , ϕL , ΓL , weak topology) −−→ (AL , φq , ΓL , weak topology)
between the o-algebras AL and AL together with their additional structures. By literally
repeating the Def. 3.7 we have the notion of (etale) (φq , ΓL )-modules over AL as well as the
category Met (AL ). In the same way as for Remark 3.2.i we may define the operator ψL on
AL and then on any etale (φq , ΓL )-module over AL . The above algebra isomorphism gives
rise to an equivalence of categories
'
quCatvariable (19) Met (AL ) −−→ Met (AL ) ,
which also respects the ψL -operators. Using the norm map for the extension AL /φq (AL ) we
define, completely analogously as in (9), a multiplicative norm operator N : AL −→ AL .
Then, using also Lemma 4.1.b, NL and N correspond to each other under the isomorphism
(18). In particular, (18) (for any choice of t0 ) induces an isomorphism
∼
=
f:N-iso (20) (AL× )NL =1 −−→ (A×
L)
N =1
.
We form the maximal integral unramified extension (= strict Henselization) of AL inside
W (E)
e L . Its p-adic completion A still is contained in W (E) e L . Note that A is a complete
discrete valuation ring with prime element πL and residue field the separable algebraic closure
Esep
L of EL in E. By the functoriality properties of strict Henselizations the q-Frobenius φq
e
preserves A. According to [KR] Lemma 1.4 the GL -action on W (E) e L respects A and induces
∼
=
an isomorphism HL = ker(χLT ) − → Autcont (A/AL ).
Let RepoL (GL ) denote the abelian category of finitely generated oL -modules equipped with
a continuous linear GL -action. The following result is established in [KR] Thm. 1.6.
KR-equiv Theorem 4.2. The functors
V 7−→ DLT (V ) := (A ⊗oL V )ker(χLT ) and M 7−→ (A ⊗AL M )φq ⊗ϕM =1
are exact quasi-inverse equivalences of categories between RepoL (GL ) and Met (AL ).
For the convenience of the reader we discuss a few properties, which will be used later on,
of the functors in the above theorem.
First of all we recall that the tensor product M ⊗oL N of two linear-topological o-modules
M and N is equipped with the linear topology for which the oL -submodules
im(UM ⊗oL N → M ⊗oL N ) + im(M ⊗oL UN → M ⊗oL N ) ⊆ M ⊗oL N ,
where UM and UN run over the open submodules of M and N , respectively, form a funda-
mental system of open neighbourhoods of zero. One checks that, if a profinite group H acts
linearly and continuously on M and N , then its diagonal action on M ⊗oL N is continuous
as well.
In our situation we consider M = A with its weak topology induced by the weak topology
of W (E)
e L and N = V in Repo (GL ) equipped with its πL -adic topology. The diagonal action
L
of GL on A ⊗oL V then, indeed, is continuous. In addition, since the πL -adic topology on A
is finer than the weak topology any open oL -submodule of A contains πLj A for a sufficiently
big j. Hence {im(U ⊗oL V → A ⊗oL V ) : U ⊆ A any open oL -submodule} is a fundamental
system of open neighbourhoods of zero in A ⊗oL V . This implies that the tensor product
topology on A ⊗oL V is nothing else than its weak topology as a finitely generated A-module.
induces-weak Remark 4.3. For any V in RepoL (GL ) the tensor product topology on A ⊗oL V induces the
weak topology on DLT (V ). In particular, the residual ΓL -action on DLT (V ) is continuous.
Proof. The finitely generated AL -module DLT (V ) is of the form DLT (V ) ∼ n
= ⊕ri=1 AL /πLi AL
with 1 ≤ ni ≤ ∞. Using the isomorphism in the subsequent Prop. 4.4.ii we obtain that
A ⊗oL V ∼ n
= ⊕ri=1 A/πLi A. We see that the inclusion DLT (V ) ⊆ A ⊗oL V is isomorphic to the
direct product of the inclusions AL /πLni AL ⊆ A/πLni A, which clearly are compatible with the
weak topologies.
exact Proposition 4.4. i. The functor DLT is exact.
∼
=
ii. For any V in RepoL (GL ) the natural map A ⊗AL DLT (V ) −−→ A ⊗oL V is an isomor-
phism (compatible with the GL -action and the Frobenius on both sides).
Proof. We begin with three preliminary observations.
1) As A is oL -torsion free, the functor A ⊗oL − is exact.
2) The functor DLT restricted to the full subcategory of finite length objects V in RepoL (GL )
is exact. This follows immediately from 1) and the vanishing of H 1 (HL , A ⊗oL V ) in Lemma
5.2 below.
3) For any V in RepoL (GL ) we have DLT (V ) = limn DLT (V /πLn V ). To see this we compute
←−
n
lim DLT (V /πL V ) = lim(A ⊗oL V /πL V )HL = (lim(A ⊗oL V /πLn V ))HL
n
←− ←− ←−
n n n
= (A ⊗oL lim V /πLn V )HL = (A ⊗oL V )HL
←−
n
= DLT (V ) .
here the third identity becomes obvious if one notes that V as an oL -module is a finite direct
sum of modules of the form oL /πLj oL with 1 ≤ j ≤ ∞. In this case limn (A ⊗oL oL /πLj+n oL ) =
←−
limn A/πLj+n A = A/πLj A = A ⊗oL oL /πLj oL .
←−
i. Let
0 −→ V1 −→ V2 −→ V3 −→ 0
be an exact sequence in RepoL (GL ). By 2) we obtain the exact sequences of projective systems
of finitely generated AL -modules
DLT (ker(πLn |V3 )) −→ DLT (V1 /πLn V1 ) −→ DLT (V2 /πLn V2 ) −→ DLT (V3 /πLn V3 ) −→ 0 .
Since AL is a noetherian pseudocompact ring taking projective limits is exact. By 3) the

resulting exact sequence is
lim DLT (ker(πLn |V3 )) −→ DLT (V1 ) −→ DLT (V2 ) −→ DLT (V3 ) −→ 0 .
←−
n
But since the torsion subgroup of V3 is finite and the transition maps in the projective system
(ker(πLn |V3 ))n are multiplication by πL , any composite of sufficiently many transition maps in
this projective system and hence also in the projective system (DLT (ker(πLn |V3 )))n is zero. It
follows that limn DLT (ker(πLn |V3 )) = 0.
←−
ii. The compatibility properties are obvious from the definition of the map. To show its
bijectivity we may assume, by devissage and 3), that πL V = 0. In this case our assertion
reduces to the bijectivity of the natural map Esep sep Gal(Esep
L /EL ) −→ E
sep
L ⊗EL (EL ⊗k V ) L ⊗k
V . But this is a well known consequence of the vanishing of the Galois cohomology group
H 1 (Gal(Esep sep
L /EL ), GLd (EL )) where d := dimk V .
innerhom Lemma 4.5. The above equivalence of categories DLT is compatible with the formation of
inner Hom-objects, i.e., there are canonical isomorphisms
HomAL (DLT (V1 ), DLT (V2 )) = DLT (HomoL (V1 , V2 ))
for every V1 , V2 in RepoL (GL ). We also have
ψHomA (DLT (V1 ),DLT (V2 )) = ψDLT (HomoL (V1 ,V2 )) .

L
Proof. We have
DLT (HomoL (V1 , V2 )) = (A ⊗oL HomoL (V1 , V2 ))HL

= HomA (A ⊗oL V1 , A ⊗oL V2 )HL
= HomA (A ⊗AL DLT (V1 ), A ⊗oL V2 )HL
= HomAL (DLT (V1 ), A ⊗oL V2 )HL
= HomAL (DLT (V1 ), (A ⊗AL V2 )HL )
= HomAL (DLT (V1 ), DLT (V2 )) .
Here the second identity is clear for V1 being free, the general case follows by choosing a finite
presentation of V1 (as oL -module neglecting the group action). The third identity uses Prop.
4.4.ii, while the fourth one comes from the adjointness of base extension and restriction. The
fifth one uses the fact that HL acts trivially on DLT (V1 ).
One easily checks that the above sequence of identities is compatible with the ΓL -actions
(which are induced by the diagonal GL -action on A ⊗oL −). The compatibility with Frobenius
can be seen as follows. First of all we abbreviate ϕDLT (Hom) := ϕDLT (HomoL (V1 ,V2 )) and ϕHom :=
ϕHomA (DLT (V1 ),DLT (V2 )) . An element β = i ai ⊗ αi ∈ (A ⊗oL HomoL (V1 , V2 ))HL becomes,
P
L
under the above identifications, the map ιβ : DLT (V1 ) → (A⊗oL V2 )HL which sends j cj ⊗vj
P
ai cj ⊗ αi (vj ). Assuming that cj = φq (c0j ) we compute

P
to i,j
X X
ιϕD (Hom) (β) (ϕDLT (V1 ) ( c0j ⊗ vj ) = ιPi φq (ai )⊗αi ( φq (c0j ) ⊗ vj )
LT
j j
X
= φq (ai )φq (c0j ) ⊗ αi (vj )
i,j
X
= ϕDLT (V2 ) ( ai c0j ⊗ αi (vj ))
i,j
X
= ϕDLT (V2 ) (ιβ ( c0j ⊗ vj ))
j
X
= ϕHom (ιβ )(ϕDLT (V1 ) ( c0j ⊗ vj )) ,
j
where the last identity comes from 17. Using the etaleness of DLT (V1 ) we deduce that
ιϕD (Hom) (β) = ϕHom (ιβ ) for any β ∈ DLT (HomoL (V1 , V2 )). The additional formula for the
LT
ψ-operators is a formal consequence of the compatibility of the ϕ-operators.
twist Remark 4.6. For any V in RepoL (GL ) and any continuous character χ : ΓL −→ o× L with
representation module Wχ = oL tχ the twisted GL -representation V (χ) is defined to be V (χ) =
V ⊗ Wχ as oL -module and σ|V (χ) (v ⊗ w) = σ|V (v) ⊗ σ|Wχ (w) = χ(σ) · σ|V (v) ⊗ w. One
easily checks that DLT (V (χ)) = DLT (V )(χ). If V = oL /πLn oL , 1 ≤ n ≤ ∞ is the trivial
representation, we usually identify V (χ) and Wχ . Recall that for the character χLT we take
WχLT = T = oL t0 and Wχ−1 = T ∗ = oL t∗0 as representation module, where T ∗ denotes the
LT
oL -dual with dual basis t∗0 of t0 .
Defining Ω1 := Ω1AL = AL dωLT any choice of t0 defines an isomorphism Ω1AL ∼ = Ω1AL by
1 1
sending f (Z)dZ to f (ωLT )dωLT ; moreover ΩAL and ΩAL correspond to each other under the
equivalence of categories (19). Due to the isomorphism (18) we obtain a residue pairing
due-pairingbA (21) AL × Ω1AL −→ oL
(f (ωLT ), g(ωLT )dωLT ) 7−→ Res(f (Z)g(Z)dZ)
which satisfies
:adjunctionbA (22) Res(f ψΩ1 (ω)) = Res(φq (f )ω) for any f ∈ AL and ω ∈ Ω1AL
(by Lemma 3.18) and which is independent of the choice of t0 , i.e., ωLT , by Remark 3.4.ii
and Lemma 4.1.a. In particular, we have a well defined map Res : Ω1AL → oL . In this context
Remark 3.15 together with Remark 3.2.vii tell us that
dû dû
-invariancebA (23) ψΩ1 ( )=
û û
holds true for every û ∈ (A×
L)
N =1 .
Moreover, Lemma 3.6 translates into the (existence of the) topological isomorphism
∼
=
3.6bA-Version (24) HomAL (M, Ω1AL /πLn Ω1AL ) −−→ HomcoL (M, L/oL )
F 7−→ πL−n Res(F (.)) mod oL ,
for any M annihilated by πLn . Lemma 3.14 implies the isomorphism

∼
=
AL (χLT ) = AL ⊗oL T −−→ Ω1AL
f (ιLT (t0 )) ⊗ t0 7−→ f (ιLT (t0 ))gLT (ιLT (t0 ))dιLT (t0 ) .
Using Lemma 4.1.a as well as the second identity in (2) one verifies that this isomorphism
(unlike its origin in Lemma 3.14) does not depend on the choice of t0 . We use it in order to
transform (24) into the topological isomorphism
∼
=
itybA-Version (25) HomAL (M, AL /πLn AL (χLT )) −−→ HomcoL (M, L/oL ).
Finally we obtain the following analogues of Prop. 3.19 and Remark 3.20.
nglebAVersion Remark 4.7. For any etale (ϕL , ΓL )-module M such that πLn M = 0 one has the ΓL -invariant
(jointly) continuous pairing
[ , i = [ , iM : M × HomAL (M, AL /πLn AL (χLT )) −→ L/oL
(m, F ) 7−→ πL−n Res(F (m)d logLT (ωLT )) mod oL
with adjointness properties analogous to the ones in Prop. 3.19. Again, it is independent of
the choice of t0 .
5. Iwasawa cohomology
For any V in RepoL (GL ) we also write H i (K, V ) = H i (GK , V ), for any algebraic extension
K of L, and we often abbreviate ϕ := ϕDLT (V ) .
phi-1 Remark 5.1. For any V in RepoL (GL ) the sequence
⊆ φq ⊗id −1
f:phi-1 (26) 0 −→ V −−→ A ⊗oL V −−−−−−→ A ⊗oL V −→ 0
is exact.
x7→xq −x
Proof. We clearly have the exact sequence 0 → kL → Esep sep
L −−−−−→ EL → 0. By devissage we
φq −1
deduce the exact sequence 0 → oL /πLn oL → A/πLn A −−−→ A/πLn A → 0 for any n ≥ 1. Since
the projective system {oL /πLn oL }n has surjective transition maps, passing to the projective
limit is exact and gives the exact sequence
φq −1
eier-sequence (27) 0 → oL → A −−−→ A → 0.
Finally, A is oL -torsion free and hence flat over oL . It follows that tensoring by V is exact.
Since A is the πL -adic completion of an unramified extension of AL with Galois group HL
the HL -action on A/πLn A, for any n ≥ 1, and hence on A ⊗oL V , whenever πLn V = 0 for some
n ≥ 1, is continuous for the discrete topology. We therefore may, in the latter case, pass from
(26) to the associated long exact Galois cohomology sequence with respect to HL .
phi-version Lemma 5.2. Suppose that πLn V = 0 for some n ≥ 1; we then have:
i. H i (HL , A ⊗oL V ) = 0 for any i ≥ 1;
ii. the long exact cohomology sequence for (26) gives rise to an exact sequence
ϕ−1 ∂ϕ
f:partial-phi (28) 0 −→ H 0 (L∞ , V ) −→ DLT (V ) −−−→ DLT (V ) −−→ H 1 (L∞ , V ) −→ 0 .
Proof. i. Since A ⊗oL V ∼

= A ⊗AL DLT (V ) by Prop. 4.4.ii, it suffices to show the vanishing
i n
of H (HL , A/πL A) = 0 for i ≥ 1. This reduces, by devissage, to the case n = 1, i.e., to
H i (HL , Esep
L ) = 0, which is a standard fact of Galois cohomology. ii. follows immediately
from i.
In order to derive from this a computation of Iwasawa cohomology in terms of (ϕL , ΓL )-
modules we first have to recall Pontrjagin duality and local Tate duality in our setting. The
trace pairing
L × L −→ Qp
(x, y) 7−→ TrL/Qp (xy)
gives rise to the inverse different
∼
=
D−1
L/Qp = {x ∈ L : TrL/Qp (xoL ) ⊆ Zp } −
−→ HomZp (oL , Zp )
y 7−→ [x 7→ TrL/Qp (yx)] .
Let DL/Qp = πLs oL . we fix once and for all the oL -linear isomorphism
∼
=
f:different (29) oL −−→ HomZp (oL , Zp )
y 7−→ [x 7→ TrL/Qp (πL−s xy)] .
By tensoring with Qp /Zp it induces the isomorphism of torsion oL -modules
Ξ : L/oL ∼
= HomZ (oL , Zp ) ⊗Z Qp /Zp ∼
p = HomZ (oL , Qp /Zp ) .
p p
Now let M be any topological oL -module. Since HomZp (oL , −) is right adjoint to scalar
restriction from oL to Zp and by using Ξ−1 in the second step, we have a natural isomorphism
f:LPontrjagin (30) HomZ (M, Qp /Zp ) ∼
p = Homo (M, HomZ (oL , Qp /Zp )) ∼
L p = Homo (M, L/oL ) .L
LPontrjagin Lemma 5.3. The isomorphism (30) restricts to an isomorphism

Homc (M, Qp /Zp ) ∼
Zp = Homc (M, L/oL ) oL
of topological groups between the subgroups of continuous homomorphisms endowed with the
compact-open topology.
Proof. Coming from an isomorphism between the targets the second isomorphism in (30)
obviously restricts to a topological isomorphism
HomcoL (M, HomZp (oL , Qp /Zp )) ∼
= HomcoL (M, L/oL ) .
The first isomorphism is induced by the homomorphism λ 7→ λ(1) between the targets and
therefore, at least restricts to a continuous injective map
HomcoL (M, HomZp (oL , Qp /Zp )) −→ HomcZp (M, Qp /Zp ) .
Let ` : M → Qp /Zp be a continuous homomorphism. Then the composite map
(m,a)7→am `
M × oL −−−−−−−→ M −→ Qp /Zp
is continuous. Therefore the preimage F` ∈ HomoL (M, HomZp (oL , Qp /Zp )) of `, which is
given by F` (m)(a) := `(am), is continuous by [B-TG] X.28 Thm. 3. Finally let A ⊆ M be any
compact subset and V ⊆ Qp /Zp be any subset. Define B := {a ∈ oL : aA ⊆ A}. Then 1 ∈ B,
and, since A is closed, also B is closed and hence compact. Put Ve := {λ ∈ HomZp (oL , Qp /Zp ) :
λ(B) ⊆ V }. One easily checks that `(A) ⊆ V if and only if F` (A) ⊆ Ve . This means that the
inverse bijection ` 7→ F` is continuous as well.
In the following we shall use the notation
M ∨ := HomcoL (M, L/oL ) ,
always equipped with the compact-open topology. The following version of Pontrjagin duality
should be well known. Since we could not find a reference we will sketch a proof for the
convenience of the reader.
jagin-duality Proposition 5.4 (Pontrjagin duality). The functor −∨ defines an involutory contravariant
autoequivalence of the category of (Hausdorff ) locally compact linear-topological oL -modules.
In particular, for such a module M , the canonical map
∼
=
M −−→ (M ∨ )∨
is an isomorphism of topological oL -modules.
Proof. We recall that a topological oL -module M is called linear-topological if it has a fun-
damental system of open zero neighbourhoods consisting of oL -submodules. If M is linear-
topological and locally compact one easily checks that it has a fundamental system of open
zero neighbourhoods consisting of compact open oL -submodules.
Classical Pontrjagin duality M 7−→ Homc (M, R/Z), the right hand side being the group
of all continuous group homomorphisms equipped with the compact-open topology, is an
autoequivalence of the category of locally compact abelian groups M . We first compare this,
for any locally compact linear-topological Zp -module M , with the group HomcZp (M, Qp /Zp ), as
always equipped with the compact-open topology. There is an obvious injective and continuous
map
f:Pontrjagin (31) HomcZp (M, Qp /Zp ) −→ Homc (M, R/Z) .
Step 1: The map (31) is bijective. Let ` : M → R/Z be any continuous group homomor-
phism. We have to show that im(`) ⊆ Qp /Zp and that ` is continuous for the discrete topology
on Qp /Zp . We fix a compact-open Zp -submodule U ⊆ M . Then `(U ) is a compact subgroup
of R/Z, and hence is either equal to R/Z or is finite. Since U is profinite the former cannot
occur. We conclude that `(U ) is a finite subgroup of Qp /Zp . In particular, there is an r ∈ N
such that pr · `|U = 0. The quotient module M/U is discrete and Zp -torsion. It follows that
pr · `(M ) ⊆ Qp /Zp and hence that `(M ) ⊆ Qp /Zp . Since R/Z induces the discrete topology
on the finite subgroup `(U ) the restricted homomorphism ` : U → Qp /Zp is continuous.
Since U is open in M this suffices for the continuity of ` : M → Qp /Zp . That ` then is a
homomorphism of Zp -modules is automatic.
Step 2: The map (31) is open. As usual, C(A, V ), for a compact subset A ⊆ M and an
arbitrary subset V ⊆ Qp /Zp , denotes the open subset of all ` ∈ HomcZp (M, Qp /Zp ) such
that `(A) ⊆ V . First one checks that the C(m + U, V ), for m ∈ M , U ⊆ M a compact-
open Zp -submodule, and V arbitrary, form a subbase of the compact-open topology. For any
`0 ∈ C(m+U, V ) we have `0 ∈ C(m+U, `0 (m)+`0 (U )) ⊆ C(m+U, V ), where `0 (U ) is a finite
subgroup of Qp /Zp . These observations reduce us to showing that the sets C(m+U, v+ p1n Z/Z)
are open in Homc (M, R/Z). We fix a point `0 ∈ C(m + U, v + p1n Z/Z), and we let pt be the
order of m modulo U . Note that we have `0 (m) ∈ v + p1n Z/Z and hence
1 1
f:ell0 (32) `0 (U ) ⊆ pn Z/Z and `0 (m) ∈ pn+t
Z/Z .
We use the open subsets

1 1 1

V1 (`0 ) := (− 2pn+t+1 , 2pn+t+1 )+ pn Z /Z ⊆ V2 (`0 ) := V1 (`0 ) − V1 (`0 ) ⊆ R/Z .
They satisfy:
1 1 1
a) V2 (`0 ) ∩ pn+t+1 Z/Z = V1 (`0 ) ∩ pn+t+1
Z/Z = pn Z/Z.
1
b) V1 (`0 ) + pn Z/Z ⊆ V1 (`0 ).
We claim that `0 ∈ C(m + U, v + V1 (`0 )) ⊆ C(m + U, v + p1n Z/Z). This means that in
Homc (M, R/Z) we have found an open neighbourhood of `0 which is contained in C(m +
U, v + p1n Z/Z). Hence C(m + U, v + p1n Z/Z) is open in Homc (M, R/Z). Since p1n Z/Z ⊆ V1 (`0 )
we certainly have `0 ∈ C(m + U, v + V1 (`0 )). Let now ` ∈ C(m + U, v + V1 (`0 )) be an arbitrary
element. We have `(U ) = p1j Z/Z for some j ≥ 0 and consequently `(m) ∈ pj+t 1
Z/Z.
1 1
Case 1: j ≤ n. Using (32) we then have `0 (m) − v ∈ pn Z/Z and `(m) − `0 (m) ∈ pn+t Z/Z
and hence
1
`(m) − v = `(m) − `0 (m) + `0 (m) − v ∈ pn+t Z/Z .
1
Since also `(m) − v ∈ V1 (`0 ) we deduce from a) that `(m) − v ∈ pn Z/Z and therefore
1 1 1
`(m + U ) = `(m) + `(U ) ⊆ v + pn Z/Z + pj
Z/Z =v+ pn Z/Z .
Case 2: j > n. We obtain
1 1
pn+1
Z/Z ⊆ pj
Z/Z = `(U ) ⊆ −`(m) + v + V1 (`0 )
= −(`(m) − `0 (m)) − `0 (m) + v + V1 (`0 )
1
⊆ −(`(m) − `0 (m)) + pn Z/Z + V1 (`0 )
⊆ −(`(m) − `0 (m)) + V1 (`0 )
1
⊆ −(V1 (`0 ) + pn Z/Z) + V1 (`0 )
⊆ V1 (`0 ) − V1 (`0 )
⊆ V2 (`0 ) ,
where the fourth and the sixth inclusion use b). This is in contradiction to b). We deduce
that this case, in fact, cannot occur.
At this point we have shown that (31) is a topological isomorphism of locally compact
abelian groups. From now on we assume that M is a locally compact linear-topological oL -
module. By combining this latter isomorphism with the isomorphism in Lemma 5.3 we obtain
a topological isomorphism of locally compact abelian groups
f:topiso (33) Homc (M, L/oL ) ∼
oL = Homc (M, R/Z) ,
which is natural in M . Of course, HomcoL (M, L/oL ) naturally is an oL -module again.
Step 3: The oL -module HomcoL (M, L/oL ) is linear-topological. It is straightforward to check
that the C(U, {0}) with U running over all compact open oL -submodules of M form a fun-
damental system of open zero neighbourhoods in M ∨ . Each such C(U, {0}) evidently is an
oL -submodule.
Hence the topological isomorphism (33) also applies to M ∨ instead of M . One checks that
under this isomorphism the natural map M → Homc (Homc (M, R/Z), R/Z) corresponds to
the natural map M → (M ∨ )∨ . We finally see that the classical Pontrjagin duality implies
∼
=
that M − → (M ∨ )∨ is a topological isomorphism; it, of course, is oL -linear.
α β
gin-exactness Remark 5.5. Let M0 − → M − → M1 be a sequence of locally compact linear-topological oL -
modules such that im(α) = ker(β) and β is topologically strict with closed image; then the
β∨ α∨
dual sequence M1∨ −−→ M ∨ −−→ M0∨ is exact as well, i.e., we have im(β ∨ ) = ker(α∨ ).
Proof. We have ker(α∨ ) = (M/ im(α))∨ ∼ = im(β)∨ , where the second isomorphism uses the
assumption that β is topologically strict. The assertion therefore reduces to the claim that
the closed immersion im(β) ⊆ M1 induces a surjection between the corresponding Pontrjagin
duals. For this see, for example, [HR] Thm. 24.11.
We recall that the weak topology on a finitely generated AL -module M is oL -linear; more-
over, it is locally compact if M is annihilated by some power of πL . Suppose that M is a
finitely generated AL /πLn AL -module. From (24) and (25) we have topological isomorphisms
M∨ ∼ = HomAL (M, Ω1 /πLn Ω1 ) ∼ = HomAL (M, AL /πLn AL (χLT )). By Prop. 5.4 they dualize
into topological isomorphisms M ∼ = HomAL (M, Ω1 /πLn Ω1 )∨ ∼ = HomAL (M, AL /πLn AL (χLT ))∨ .
If M actually is an etale (ϕq , ΓL )-module then we see that in the adjoint pairs of maps
(ψM , ϕHomA (M,Ω1 /πLn Ω1 ) ) and (ϕM , ψHomA (M,Ω1 /πLn Ω1 ) ) from Remark 4.7 each map deter-
L L
mines the other uniquely.
in-dual-inner Remark 5.6. Let V be an object in RepoL (GL ) of finite length. Then, the pairing [ , iDLT (V )
from Remark 4.7, the Remark 4.6, and the compatibility of the functor DLT (−) with internal
Hom’s by Lemma 4.5 induce, for n sufficiently large, a natural isomorphism of topological
groups
DLT (V )∨ =∼ HomA (DLT (V ), AL /π n AL (χLT ))
L L
∼
= HomAL (DLT (V ), DLT ((oL /πLn oL )(χLT )))
= DLT (HomoL (V, (oL /πLn oL )(χLT ))) = DLT (V ∨ (χLT )) ,
which is independent of the choice of n and under which ψDLT (V ∨ (χLT )) identifies with ϕ∨
DLT (V )
by Remark 4.7.
Tate-local Proposition 5.7 (Local Tate duality). Let V be an object in RepoL (GL ) of finite length, and
K any finite extension of L. Then the cup product and the local invariant map induce perfect
pairings of finite oL -modules
H i (K, V ) × H 2−i (K, HomZp (V, Qp /Zp (1))) −→ H 2 (K, Qp /Zp (1)) = Qp /Zp
and
H i (K, V ) × H 2−i (K, HomoL (V, L/oL (1))) −→ H 2 (K, L/oL (1)) = L/oL
where −(1) denotes the Galois twist by the cyclotomic character. In other words, there are
canonical isomorphisms
H i (K, V ) ∼
= H 2−i (K, V ∨ (1))∨ .
Proof. Note that the isomorphism H 2 (K, L/oL (1)) = L/oL arises from H 2 (K, Qp /Zp (1)) =
Qp /Zp by tensoring with oL over Zp . The first pairing is the usual version of local Tate duality
(cf. [Ser] II.5.2 Thm. 2). It induces the first isomorphism in
H i (K, V ) ∼
= HomZ (H 2−i (K, HomZ (V, Qp /Zp (1))), Qp /Zp )
p p
∼
= HomoL (H 2−i (K, HomZp (V, Qp /Zp (1))), L/oL )
∼ Homo (H 2−i (K, Homo (V, L/oL (1)))), L/oL ) ,
= L L
while the second and third are induced by Lemma 5.3. To obtain the second pairing it remains
to check that the above composite isomorphism is given by the cup product again. By the
functoriality properties of the cup product this reduces to the following formal fact. Let
ξ : L/oL −→ Qp /Zp be any group homomorphism. Then the diagram
H 2 (K, L/oL (1))

= / L/oL
H 2 (K,ξ(1)) ξ

H 2 (K, Qp /Zp (1))
= / Qp /Zp
commutes, where the horizontal maps are the local invariant maps. This in turn is an easy
consequence of the Zp -linearity of the local invariant map if one uses the following description
ξ
of ξ viewed as a map L/oL = Qp /Zp ⊗Zp oL →
− Qp /Zp . Let ζ : oL −→ HomZp (Qp /Zp , Qp /Zp ) =
Zp be the homomorphism which sends a to c 7→ ξ(c ⊗ a). Then ξ(c ⊗ a) = ζ(a)c.
For any V in RepoL (GL ) we define the generalized Iwasawa cohomology of V by

∗
HIw (L∞ /L, V ) := lim H ∗ (K, V )
←−
K
where K runs through the finite Galois extensions of L contained in L∞ and the transition
maps in the projective system are the cohomological corestriction maps. 2
Shapiro’s lemma for cohomology gives natural isomorphisms
H ∗ (K, V ) = H ∗ (GL , oL [GL /GK ] ⊗oL V )
where, on the right hand side, GL acts diagonally on the coefficients. In this picture the
corestriction map, for K ⊆ K 0 , becomes the map induced on cohomology by the map pr ⊗ idV :
oL [GL /GK 0 ] ⊗oL V −→ oL [GL /GK ] ⊗oL V .
Lemma 5.8. HIw ∗ (L /L, V ) = H ∗ (G , o [[Γ ]] ⊗
limit ∞ L L L oL V ) (where the right hand side refers to
cohomology with continuous cochains).
Proof. On the level of continuous cochain complexes we compute
lim C • (GL , oL [GL /GK ] ⊗oL V ) = C • (GL , lim(oL [GL /GK ] ⊗oL V ))
←− ←−
K K
•
= C (GL , oL [[ΓL ]] ⊗oL V ) .
The second identity comes from the isomorphism limK (oL [GL /GK ] ⊗oL V ) ∼ = oL [[ΓL ]] ⊗oL V
←− s r
which is easily seen by using a presentation of the form 0 → oL → oL → V → 0. Since the
transition maps in this projective system of complexes are surjective the first hypercohomology
spectral sequence for the composite functor lim ◦H 0 (GL , .) degenerates so that the second
←−
hypercohomology spectral sequence becomes
Ri lim H j (GL , oL [GL /GK ] ⊗oL V ) =⇒ H i+j (GL , oL [[ΓL ]] ⊗oL V ) .
←−
K
2Note that, for any finite extension K/L contained in L the definition HIw ∗
(L∞ /K, V ) :=
∞
∗ 0
limK⊆K 0 ⊆L H (K , V ) produces the same oL -modules. Our notation indicates that we always consider these
←− ∞
groups as ΓL -modules.
Due to the countability of our projective system we have Ri limK = 0 for i ≥ 2. Hence this
←−
spectral sequence degenerates into short exact sequences
0 −→ R1 lim H i−1 (GL , oL [GL /GK ] ⊗oL V ) −→

←−
K
H ∗ (GL , oL [[ΓL ]] ⊗oL V ) −→ lim H i (GL , oL [GL /GK ] ⊗oL V ) −→ 0 .
←−
K
It is well known that the Galois cohomology groups H ∗ (GL , oL [GL /GK ] ⊗oL V ) are finitely
generated oL -modules (cf. [Ser] II.5.2 Prop. 14). It therefore follows from [Jen] Thm. 8.1 that
the above R1 limK -terms vanish.
←−
delta-functor ∗ (L /L, V ) is a δ-functor on Rep (G ).
Lemma 5.9. HIw ∞ oL L
Proof. Let 0 → V1 → V2 → V3 → 0 be a short exact sequence in RepoL (GL ). Then the

sequence of topological GL -modules 0 → oL [[ΓL ]]⊗oL V1 → oL [[ΓL ]]⊗oL V2 → oL [[ΓL ]]⊗oL V3 →
0 is short exact as well. In view of Lemma 5.8 our assertion therefore follows from [NSW]
Lemma 2.7.2 once we show that
1. the topology of oL [[ΓL ]] ⊗oL V1 is induced by the topology of oL [[ΓL ]] ⊗oL V2 and
2. the (surjective continuous) map oL [[ΓL ]] ⊗oL V2 → oL [[ΓL ]] ⊗oL V3 has a continuous
section as a map of topological spaces.
Each oL [[ΓL ]] ⊗oL Vi is a profinite (hence compact) abelian group with a countable base of
the topology, which therefore is metrizable by [B-TG] IX.21 Prop. 16. One easily deduces 1.
and that the map in 2. is open. The property 2. then follows from [Mic] Cor. 1.4.
omology-twist Remark 5.10. For any V0 in RepoL (GL ) which is oL -free and on which GL acts through its
∗ (L /L, V ⊗
factor ΓL there is a natural isomorphism HIw ∞
∼ ∗
oL V0 ) = HIw (L∞ /L, V ) ⊗oL V0 .
Proof. In view of Lemma 5.8 the asserted isomorphism is induced by the GL -equivariant
isomorphism on coefficients
∼
=
oL [[ΓL ]] ⊗oL V ⊗oL V0 −−→ oL [[ΓL ]] ⊗oL V ⊗oL V0
γ ⊗ v ⊗ v0 7−→ γ ⊗ v ⊗ γ −1 v0 ;
on the left GL acts diagonally on all three factors, whereas on the right it acts trivially on
the third factor.
TatelocalIw Remark 5.11. Let V be in RepoL (GL ) of finite length; in particular V is discrete. Then local
Tate duality (Prop. 5.7) induces an isomorphism
i
HIw (L∞ /L, V ) ∼
= H 2−i (L∞ , V ∨ (1))∨ .
Proof. Use Prop. 5.7 over the layers Ln and take limits.
We point out that HIw ∗ (L /L, V ) = H ∗ (G , o [[Γ ]] ⊗

∞ L L L oL V ) is a left oL [[ΓL ]]-module
−1
through the action of γ ∈ ΓL by right multiplication with γ on the factor oL [[ΓL ]].
vanishing Lemma 5.12. i. HIw∗ (L /L, V ) = 0 for ∗ =6 1, 2.
∞
2
ii. HIw (L∞ /L, V ) is finitely generated as oL -module.
1 (L /L, V ) is finitely generated as o [[Γ ]]-module.
iii. HIw ∞ L L
Proof. i. In case ∗ > 2 the assertion follows from the fact that the groups GK have cohomo-
0 (L /L, V ) = lim V GK is
logical p-dimension 2 ([Ser] II.4.3 Prop. 12). The vanishing of HIw ∞ ←−K
clear if V is finite. Hence we may assume that V is finitely generated free over oL . Note that
the identity HIw 0 (L /L, V ) = lim V GK shows that H 0 (L /L, V ) is a profinite o -module.
∞ ←−K Iw ∞ L
On the other hand we then have the exact sequence
0 Lπ ·0 0
0 −→ HIw (L∞ /L, V ) −−−→ HIw (L∞ /L, V ) −→ HIw (L∞ /L, V /πL V )
Since we observed already that the last term vanishes it follows that HIw 0 (L /L, V ) is an
∞
0
L-vector space. Both properties together enforce the vanishing of HIw (L∞ /L, V ).
ii. We have
[
2
HIw (L∞ /L, V ) = lim H 2 (K, V ) = lim H 0 (K, V ∨ (1))∨ = ( V ∨ (1)GK )∨ ,
←− ←−
K K K
which visible is a finitely generated oL -module.
1 (L /L, V ) = H 1 (L , V ∨ (1))∨ is the Pontrja-
iii. Case 1: V is finite. By Remark 5.11 HIw ∞ ∞
gin dual of a discrete torsion module and hence is a compact oL [[ΓL ]]-module. The compact
Nakayama lemma (cf. [NSW] Lemma 5.2.18) therefore reduces us to showing that the Pon-
trjagin dual (HIw1 (L /L, V ) )∨ = H 1 (L , V ∨ (1))Γ of the Γ-coinvariants of H 1 (L /L, V )
∞ Γ ∞ Iw ∞
is cofinitely generated over oL ; here Γ is a conveniently chosen open subgroup of ΓL . The
Hochschild-Serre spectral sequence for the extension L∞ /K, where K := LΓ∞ , gives us an
exact sequence
H 1 (K, V ∨ (1)) −→ H 1 (L∞ , V ∨ (1))Γ −→ H 2 (Γ, V ∨ (1)HL ) .
The first group is finite by local Galois cohomology. At this point we choose Γ to be isomorphic
to Zdp . Then H 2 (Γ, Fp ) is finite. Since Fp is the only simple Zp [[Γ]]-module it follows by
devissage that H 2 (Γ, V ∨ (1)HL ) is finite.
Case 2: V is oL -free. As pointed out above oL [[ΓL ]] ⊗oL V is a oL [[ΓL ]]-module, which is
finitely generated free and on which GL acts continuously and oL [[ΓL ]]-linearly. In view of
Lemma 5.8 we therefore may apply [FK] Prop. 1.6.5 and obtain that HIw ∗ (L /L, V ), as a
∞
oL [[ΓL ]]-module, is isomorphic to the cohomology of a bounded complex of finitely generated
projective oL [[ΓL ]]-modules. The ring oL [[ΓL ]] is noetherian. Hence the cohomology of such a
complex is finitely generated.
The general case follows by using Lemma 5.9 and applying the above two special cases to
the outer terms of the short exact sequence 0 → Vtor → V → V /Vtor → 0, where Vtor denotes
the torsion submodule of V .
psi-version Theorem 5.13. Let V in RepoL (GL ). Then, with ψ = ψDLT (V (τ −1 )) , we have a short exact
sequence
ψ−1
f:psi-version 1
(34) 0 −→ HIw (L∞ /L, V ) −→ DLT (V (χLT χ−1 −1
cyc )) −−−→ DLT (V (χLT χcyc ))
2
−→ HIw (L∞ /L, V ) −→ 0 ,
which is functorial in V .
Proof. In the sense of Remark 4.6 we take T ∗ ⊗Zp Zp (1) and T ⊗Zp HomZp (Zp (1), Zp ) as
representation module for τ and τ −1 , respectively.
We first assume that V has finite length. Then the exact sequence (28) is a sequence of
locally compact linear-topological oL -modules. In fact, the first term is finite and the last term
is cofinitely generated over oL . In particular, the first and the last term carry the discrete
topology and the first map is a closed immersion. Moreover, the map ϕ − 1 in the middle, by
Lemma 3.12, is topologically strict with open image. The latter implies that ∂ϕ induces an
∼
=
→ H 1 (L∞ , V ).
isomorphism of discretely topologized oL -modules DLT (V )/(ϕ − 1)DLT (V ) −
In particular, the map ∂ϕ is topologically strict as well.
Therefore, using Remark 5.5, we obtain that the dual sequence
ϕ∨ −1
0 −→ H 1 (L∞ , V )∨ −→ DLT (V )∨ −−−−→ DLT (V )∨ −→ H 0 (L∞ , V )∨ −→ 0
is exact. If we identify the terms in this latter sequence according to Remark 5.11 and Remark
5.6 then the result is the exact sequence in the assertion.
Now let V be arbitrary and put Vn := V /πLn V . We have the exact sequence of projective
systems
ψ−1
1
0 → HIw (L∞ /L, Vn ) → DLT (Vn (τ −1 )) −−−→ DLT (Vn (τ −1 )) → HIw
2
(L∞ /L, Vn ) → 0 .
Since the functor DLT is exact (Prop. 4.4.i) we have
lim DLT (Vn (τ −1 )) = lim DLT (V (τ −1 ))/πLn DLT (V (τ −1 )) = DLT (V (τ −1 )) .
←− ←−
n n
Moreover,
i
lim HIw (L∞ /L, Vn ) = lim lim H i (K, Vn ) = lim lim H i (K, Vn ) = lim H i (K, V )
←− ←− ←− ←− ←− ←−
n n K K n K
i
= HIw (L∞ /L, V ).
Therefore it remains to show that passing to the projective limit in the above exact se-
quence of projective systems is exact. For this it suffices to show that R1 lim of the two
1 (L /L, V )} and {(ψ − 1)D
←−
−1 ))} vanishes. Because of
projective systems {HIw ∞ n n LT (Vn (τ n
DLT (Vn (τ −1 )) = DLT (V (τ −1 ))/πLn DLT (V (τ −1 )) the transition maps in the second projective
system are surjective, which guarantees the required vanishing. For the first projective system
we choose an open pro-p subgroup Γ in ΓL , so that oL [[Γ]] is a complete local noetherian com-
mutative ring. From Lemma 5.12.iii we know that {HIw 1 (L /L, V )} is a projective system
∞ n n
of finitely generated oL [[Γ]]-modules. Hence [Jen] Thm. 8.1 applies and gives the required
vanishing.
-equivariance Remark 5.14. Each map in the exact sequence (34) is continuous and oL [[ΓL ]]-equivariant.
Proof. Continuity and ΓL -equivariance follow from the construction. Since the weak topology
on DLT (V ) is oL -linear and complete we may apply [Laz] Thm. II.2.2.6 (which is valid for any
profinite group) and obtain that the continuous ΓL -action extends, by continuity, uniquely to
an oL [[ΓL ]]-action on DLT (V ).
6. The Kummer map

sec:Kummer
We consider the Kummer isomorphism
pm ∼
=
κ : A(L∞ ) := lim L× /L× 1
−−→ HIw (L∞ , Zp (1)) .
←− n n
n,m
Recall that we have fixed a generator t0 of the Tate module T = oL t0 . Correspondingly we

have the dual generator t∗0 of the oL -dual T ∗ = oL t∗0 of T . This leads to the twisted Kummer
isomorphism
κ⊗ T∗
(L∞ , Zp (1)) ⊗Zp T ∗ ∼
Zp
A(L∞ ) ⊗Zp T ∗ −−−∼−−→ HIw
1 1
= HIw (L∞ , oL (τ ))
=
where the second isomorphism comes from Remark 5.10. On the other hand, by Thm. 5.13,
we have a natural isomorphism
∼
=
Exp∗ : HIw
1
(L∞ , oL (τ )) −−→ DLT (oL )ψ=1 = Aψ=1
L .
Finally we have the homomorphism
∇ : (lim L× ) ⊗Z T ∗ −→ Aψ=1
←− n L
n
∂inv (gu,t0 )
u ⊗ at∗0 7−→ a (ιLT (t0 )) .
gu,t0
It is well defined by Thm. 2.2, the last sentence in section 2, and Remark 3.2.ii.
ndependencet0 Remark 6.1. The map ∇ is independent of the choice of t0 .
Proof. Let u ⊗ at∗0 ∈ (limn L× ) ⊗Z T ∗ . We temporarily write ∇t0 instead of ∇, and we let t1
←− n
be a second generator of T , so that t1 = ct0 for some c ∈ o× ∗ ∗
L . Then u ⊗ at0 = u ⊗ act1 , and
by inserting the definitions in the first line we compute
0
gu,t (ιLT (t1 ))
ac
∇t1 (u ⊗ at∗0 ) = ∇t1 (u ⊗ act∗1 ) = 1
gLT (ιLT (t1 )) gu,t1 (ιLT (t1 ))

0
gu,ct0 (ιLT (ct0 ))
ac
=
gLT (ιLT (ct0 )) gu,ct0 (ιLT (ct0 ))
0
gu,ct ([c](ιLT (t0 )))
ac 0
=
gLT ([c](ιLT (t0 ))) gu,ct0 ([c](ιLT (t0 )))
a (gu,ct0 ◦ [c])0 (ιLT (t0 ))
=
gLT (ιLT (t0 )) (gu,ct0 ◦ [c])(ιLT (t0 ))
0
gu,t (ιLT (t0 ))
a 0
=
gLT (ιLT (t0 )) gu,t0 (ιLT (t0 ))
= ∇t0 (u ⊗ at∗0 ) ,
where we use Lemma 4.1.a for the fourth identity, (2) for the fifth one, and Remark 2.3.ii for
the sixth one.
Generalizing [CC] Prop. V.3.2.iii) (see also [Co2] Thm. 7.4.13) we will establish the following
kind of reciprocity law.
r-commutative Theorem 6.2. The diagram
−κ⊗T ∗
f:Kummer (35) (limn L× ) ⊗Z T ∗ / H 1 (L∞ , oL (τ ))
←− n Iw
∼
=
∇ Exp∗
& x
Aψ=1
L
is commutative.
3Colmez has no minus but which seems to be a mistake.
In a first step we consider, for any n ≥ 1, the diagram

[ , i
(AL /πLn AL )(χLT ) / im(ϕ − 1) (AL /πLn AL )ψ=1 / L/oL

×
O
∂ϕ ∼
= ∼
= Exp∗

H 1 (L∞ , oL /πLn oL (χLT )) 1 (L /L, o /π n o (χ
HIw −1 / L/oL
× ∞ L L L cyc χLT ))
O
−κ⊗Zp id
(limn L× ) ⊗Z oL /πLn oL (χ−1

LT )
←− n
rec⊗Zp id

Hom(HL , oL /πLn oL )(χLT ) HLab (p) ⊗Zp oL /πLn oL (χ−1 / L/oL ,
LT )
×
where the second pairing is induced by local Tate duality and the third pairing is the ob-
vious one. By rec : (limn L× ) −→ HLab (p) we denote the map into the maximal abelian
←− n
pro-p quotient HLab (p) of HL induced by the reciprocity homomorphisms of local class field
theory for the intermediate extensions Lm . Note that Gal(Lab ∞ /L∞ ) = lim Gal(Lab
m /L∞ ) =
←−m
ab ab
limm Gal(Lm /Lm ), where L? denotes the maximal abelian extension of L? . The upper half
←−
of the diagram is commutative by the construction of the map Exp∗ . The commutativity of
the lower half follows from [NSW] Cor. 7.2.13. All three pairings are perfect in the sense of
Pontrjagin duality.
In order to prove Thm. 6.2 we have to show that, for any u ∈ limn L× and any a ∈ oL , we
←− n
have
∂inv (gu,t0 )
[z ⊗ t0 , −Exp∗ (κ(u) ⊗ at∗0 )i ≡ Res(za d logLT ) mod πLn
gu,t0
for any z ∈ AL and any n ≥ 1. Due to the commutativity of the above diagram the left
hand side is equal to a∂ϕ (z ⊗ t0 )(rec(u) ⊗ t∗0 ) = a∂ϕ (z)(rec(u)). On the other hand the right
g0 d(g (ι (t )))
hand side, by (1), is equal to Res(za( gu,t 0 u,t0 LT 0
dZ)|Z=ιLT (t0 ) ) = Res(za gu,t ). By the oL -
u,t0 0 (ιLT (t0 ))
bilinearity of all pairings involved we may assume that a = 1. Hence we are reduced to proving
that
d(gu,t0 (ιLT (t0 )))
Res(z ) = ∂ϕ (z)(rec(u))
gu,t0 (ιLT (t0 ))
holds true for any z ∈ AL and u ∈ limn L× . According to the theory of fields of norms we
←− n×
×
have the natural identification limn Ln = EL (cf. [KR] Lemma 1.4). Under this identification,
←−
by [Lau] Thm. 3.2.2, rec(u) coincides with the image recEL (u) of u under the reciprocity
homomorphism recEL : E× ab
L −→ HL (p) in characteristic p. Furthermore, gu,t0 (ιLT (t0 )) ∈
× N =1
(AL ) , is, by Remark 2.3.i, Remark 3.2.vii, and (20), a lift of u ∈ E×L . This reduces the
proof of Thm. 6.2 further to the following proposition which generalizes the explicit reciprocity
law in [Fo1] Prop. 2.4.3.
ion-to-char-p Proposition 6.3. For any z ∈ AL and any u ∈ E× × N =1
L with (unique) lift û ∈ (AL ) we have
dû
Res(z
) = ∂ϕ (z)(recEL (u)) ,
û
where ∂ϕ is the connecting homomorphism in (28).
Obviously the connecting homomorphism ∂ϕ for V = oL induces, by reduction modulo

πno n
L , the corresponding connecting homomorphism for V = oL /πL oL . Hence we may prove
n
the identity in Prop. 6.3 as a congruence modulo π oL for any n ≥ 1. Recall that for û ∈
(A×L)
N =1 the differential form dû is ψ -invariant by (23). Hence, by the adjointness of ψ
û Ω1 Ω1
and ϕL (cf. 22), we obtain the equality
dû dû
Res(ϕm L (z) ) = Res(z )
û û
for any m ≥ 1. This reduces Prop. 6.3 and consequently Thm. 6.2 to proving the congruence
dû
:key-identity (36) Res(ϕn−1
L (z) ) ≡ ∂ϕ (z)(recEL (u)) mod πLn oL
û
for all n ≥ 1. This will be the content of the next section (cf. Lemma 7.18).
7. The generalized Schmid-Witt formula

sec:Witt
The aim of this section is to generalize parts of Witt’s seminal paper [Wit] (see also the
detailed accounts [Tho] and [Koe] of Witt’s original article) to the case of ramified Witt
vectors.
First of all we need to recall a few facts about ramified Witt vectors W (B)L for oL -
algebras B. Details of this construction can be found in [Haz]. But we will use [GAL] where a
much more straightforward approach is fully worked out. We denote by ΦB = (Φ0 , Φ1 , . . .) :
W (B)L −→ B N0 the homomorphism of oL -algebra, called the ghost map, given by the poly-
n n−1
nomials Φn (X0 , . . . , Xn ) = X0q + πL X1q + . . . πLn Xn . On the other hand, the multiplicative
Teichmüller map B −→ W (B)L is given by b 7→ [b] := (b, 0, . . .) (cf. [GAL] Lemma 1.1.15).
If B is a kL -algebra then the Frobenius endomorphism F = φq of W (B)L has the form
φq (b0 , . . . , bn , . . .) = (bq0 , . . . , bqn , . . .) (cf. [GAL] Prop. 1.1.18.i).
For a perfect kL -algebra B we have Wn (B)L = W (B)L /πLn W (B)L for any n ≥ 1 and, in
particular, ker(Φ0 ) = πL W (B)L ; moreover, any b = (b0 , b1 , . . .) ∈ W (B)L has the unique
−m
convergent expansion b = ∞ m q
P
m=1 πL [bm ] (cf. [GAL] Prop. 1.1.19).
s-map Proposition 7.1. Suppose that πL is not a zero divisor in B and that B has an endomor-
phism of oL -algebras σ such that σ(b) ≡ bq mod πL B for any b ∈ B. Then there is a unique
homomorphism of oL -algebras
sB : B −→ W (B)L such that Φi ◦ sB = σ i for any i ≥ 0.
Moreover, we have:
i. sB is injective;
ii. for any n ≥ 1 there is a unique homomorphism of oL -algebras sB,n : B/πLn B −→
Wn (B/πL B)L such that the diagram
sB W (pr)L
B / W (B)L / W (B/πL B)L
pr pr
sB,n
B/πLn B / Wn (B/πL B)L
is commutative;
iii. if B/πL B is perfect then sB,n , for any n ≥ 1, is an isomorphism.
Proof. See [GAL] Prop. 1.1.23.
phi-maps Lemma 7.2. For any perfect kL -algebra B we have:

i. The diagram
Φn−1
Wn (W (B)L )L / W (B)L
Wn (pr)L pr
φn−1
Wn (B)L
q
/ Wn (B)L
is commutative for any n ≥ 1.

ii. The composite map
sW (B) W (pr)L
W (B)L −−−−−L→ W (W (B)L )L −−−−−→ W (B)L
is the identity.
Proof. i. Let (b0 , . . . , bn−1 ) ∈ Wn (W (B)L )L with bj = (bj,0 , bj,1 , . . .). As ker(Φ0 ) = πL W (B)L
m m
we have bj ≡ [bj,0 ] mod πL W (B)L . Hence [GAL] Lemma 1.1.1 implies that bqj ≡ [bqj,0 ] mod
πLm+1 W (B)L for any m ≥ 0. Using this as well as [GAL] Lemma 1.1.13.i we now compute
n−1
n−1−m
X
Φn−1 (b0 , . . . , bn−1 ) = πLm bqm
m=0
n−1
n−1−m
πLm [bqm,0
X
≡ ] mod πLn W (B)L
m=0
n−1 n−1
= (bq0,0 , . . . , bqn−1,0 , 0, . . .)
= φn−1
q (b0,0 , . . . , bn−1,0 , 0, . . .) = φqn−1 ◦ Wn (pr)L (b0 , . . . , bn−1 ) .
ii. First of all we note that the Frobenius on W (B)L is the qth power map modulo πL .
Hence the homomorphism sW (B)L exists.
Let b = (b0 , . . . , bj , . . .) ∈ W (W (B)L )L with bj = (bj,0 , bj,1 , . . .) ∈ W (B)L be the image
under sW (B)L of some b = (b0 , b1 , . . .) ∈ W (B)L . We have to show that bi = bi,0 for any
i i
i ≥ 0. By the characterizing property of sW (B)L we have Φi (b) = (bq0 , . . . , bqj , . . .). On the
i i
other hand the computation in the proof of i. shows that Φi (b) = (bq0,0 , . . . , bqi,0 , . . .). Hence
i i
bqi = bqi,0 and therefore bi = bi,0 .
By construction (and [GAL] Prop. 1.1.26) we have AL ⊆ W (E)

e L . But there is the following
observation.
contained Remark 7.3. Let A ⊆ W (E) e L be a φq -invariant oL -subalgebra such that A/πL A ⊆ E;
e we
then have A ⊆ W (A/πL A)L .
Proof. We consider the diagram
sA W (pr)L
A / W (A)L / W (A/πL A)L
⊆ ⊆ ⊆
sW (E)
W (pr)L
L
/ W (W (E) / W (E)
e
W (E)
e L e L )L e L.
For the commutative left square we apply Prop. 7.1 (with σ := φq ). The right hand square
is commutative by naturality. By Lemma 7.2.ii the composite map in the bottom row is the
identity. Hence the composite map in the top row must be an inclusion.
This applies to AL and shows that

AL ⊆ W (EL )L ⊆ W (E)
e L
holds true. In particular, we have the commutative diagram (cf. Prop. 7.1.ii)
⊆
f:alpha-n (37) AL / W (EL )L
αn
pr pr
αn :=sAL ,n )
AL /πLn AL / Wn (EL )L
for any n ≥ 1, where αn by definition is the composite of the outer maps. For later use before
Lemma 7.18 we note that Remark 7.3 also applies to A showing that A ⊆ W (Esep L )L .
phi-maps2 Lemma 7.4. For any n ≥ 1 the diagram

Φn−1
Wn (AL )L / AL
Wn (pr)L αn
φqn−1
Wn (EL )L / Wn (EL )L
is commutative.
Proof. We consider the diagram
Φn−1
Wn (AL )L / AL
⊆
⊆ ' &
Φn−1
Wn (pr)L Wn (W (E)
e L )L / W (E)
e L
αn
Wn (pr)L
φn−1
Wn (EL )L
q
/ Wn (EL )L pr
⊆
⊆ ' &
φn−1
Wn (E)
e L q
/ Wn (E)e L.
The front square is commutative by Lemma 7.2.i. The top and bottom squares are commu-
tative by the naturality of the involved maps and the side squares for trivial reasons. Hence
the back square is commutative as claimed.
overlinealpha Lemma 7.5. For any n ≥ 1 the map αn : AL /πLn AL −→ Wn (EL )L is injective.
Proof. We have to show that Vn (EL )L ∩ AL = πLn AL . We know already that πLn AL ⊆
Vn (EL )L ∩ AL = Vn (E)e L ∩ AL = π n W (E)e L ∩ AL . But AL ⊆ W (E) e L both are discrete
L
valuation rings with the prime element πL . Therefore we must have equality.
The above two lemmas together with the surjectivity of Wn (α1 )L imply that, for any n ≥ 1,
there is a unique homomorphism of oL -algebras wn−1 : Wn (EL )L −→ AL /πLn AL such that
the diagram
Φn−1
f:A-omega (38) Wn (AL )L / AL
Wn (pr)L pr
wn−1
Wn (EL )L / AL /π n AL
L
is commutative. Furthermore, we have

f:alphaphiq (39) αn ◦ wn−1 = φn−1
q |W
and wn−1 ◦ αn = φqn−1 |A n .
n (EL )L L /πL AL
We also may apply Prop. 7.1 to oL itself (with σ := id) and obtain analogous commutative
diagrams as well as the corresponding maps
oL
αn
pr
αn
& w
oL /πLn oL / Wn (kL )L n−1 / oL /π n oL .
L
But here αn and wn−1 are isomorphisms which are inverse to each other (cf. Prop. 7.1.iii). Of
course these maps for oL and AL are compatible with respect to the inclusions oL ⊆ AL and
kL ⊆ EL .
For the rest of this section let K denote any local field isomorphic to k((Z)) with k = kL
(such an isomorphism depending on the choice of an uniformizing element Z of K), K sep
any separable closure of it and H = Gal(K sep /K) its Galois group. Furthermore we write
K for an algebraic closure of K sep , φq for the qth power Frobenius, and ℘ := φq − 1 for the
corresponding Artin-Schreier operator. By induction with respect to n one easily proves the
following fact.
rtin-Schreier Lemma 7.6. We have the short exact sequences
℘
0 −→ Wn (k)L −→ Wn (K sep )L −→ Wn (K sep )L −→ 0
and
℘
0 −→ Wn (k)L −→ Wn (K̄)L −→ Wn (K)L −→ 0 .
From the H-group cohomology long exact sequence associated with the first sequence above
we obtain a homomorphism
∂ rec∗
→ Homcont (K × , Wn (k)L ) ,
Wn (K)L = (Wn (K sep )L )H −−→ H 1 (H, Wn (k)L ) −−−K
which induces the generalized, bilinear Artin-Schreier-Witt pairing
[ , ) := [ , )K : Wn (K)L × K × −→ Wn (k)L
(x, a) 7−→ [x, a) := ∂(x)(recK (a)) ,
i.e., [x, a) = recK (a)(α) − α for any α ∈ Wn (K sep )L with ℘(α) = x. It is bilinear in the sense
that it si oL -linear in the first and additive in the second variable.
Krad Remark 7.7. Let K rad be the perfect closure of K in K. Then one can use the second exact
sequence to extend the above pairing to Wn (K rad )L × K × .
For a separable extension F of K we obtain similarly by taking Gal(K sep /F )-invariants

(instead of H-invariants) an Artin-Schreier-Witt-pairing for F
[ , )F : Wn (F )L × F × −→ Wn (k)L
(x, a) 7−→ [x, a) := ∂(x)(recF (a)) ,
(with respect to the same q!) satisfying
[x, a)F = [x, NormF/K (a))K for x ∈ Wn (K)L and a ∈ F ×
- and similarly for any pair of separable extensions F and F 0 - by the functoriality of class
field theory.
Although Wn (k)L is not a cyclic group in general, many aspects of Kummer/Artin-Schreier
theory still work. In particular, for any α = (α0 , . . . , αn−1 ) ∈ Wn (K sep )L with ℘(α) = x ∈
Wn (K)L the extension K(α) := K(α0 , . . . , αn−1 ) = (K sep )Hx = K(℘−1 (x)) of K is Galois
with abelian Galois group Gal(K(α)/K) contained in Wn (k)L via sending σ to χx (σ) :=
σ(α) − α; here Hx ⊆ H denotes the stabilizer of α, which also is the stabilizer of ℘−1 (x).
We also need the injective additive map
τ: Wn (B)L −→ Wn+1 (B)L
(x0 , . . . , xn−1 ) 7−→ (0, x0 , . . . , xn−1 )
induced in an obvious way by the additive Verschiebung V (cf. [GAL] Prop. 1.1.10). If B is
a kL -algebra then
au-additivity (40) τ ◦ φq = φq ◦ τ and, in particular, ℘◦τ =τ ◦℘
(cf. [GAL] Prop. 1.1.18.i).
properties[) Lemma 7.8. Let K ⊆ F ⊆ K sep be a finite extension. Then, for any a ∈ F × , x ∈ Wn (F )L ,
and α ∈ Wn (K sep )L with ℘(α) = x we have:
i. [τ x, a)F = τ [x, a)F (where we use the same notation for the pairing at level n + 1 and
n, respectively!);
n
ii. if a belongs to (F × )p , then [x, a)F = 0;
iii. [x, a)F = 0 if and only if a ∈ NormF (α)/F (F (α)× ).
Proof. i. By (40) we have ℘(τ α) = τ x. Therefore [τ x, a) = recK (a)(τ α)−τ α = τ (recK (a)(α)−
α) = τ [x, a). ii. Since pn Wn (k)L = 0 (this is not sharp with regard to pn !) this is immediate
from the bilinearity of the pairing. iii. Because of [x, a)F = χx (recF (a)) this is clear from local
class field theory.
For any subset S of an oL -algebra R we define the subset
Wn (S)L := {(s0 , . . . , sn−1 ) ∈ Wn (R)L : si ∈ S for all i}
of Wn (R)L as well as V Wn (S)L := V (Wn (S)L ). If I ⊆ R is an ideal then Wn (I)L is an ideal
in Wn (R)L , and we have the exact sequence
f:exact-ideal (41) 0 −→ Wn (I)L −→ Wn (R)L −→ Wn (R/I)L −→ 0 .
If R0 ⊆ R is an oL -subalgebra (not necessarily with a unit), then Wn (R0 )L ⊆ Wn (R)L forms
a subgroup and there is an exact sequence of abelian groups
Φ
f:exactR’ (42) 0 −→ V Wn (R0 )L −→ Wn (R0 )L −−→
0
R0 −→ 0 .
We apply this to R0 = ak[a] for a ∈ K × .
aka Proposition 7.9. For any x ∈ Wn (ak[a])L we have [x, a) = 0.

Proof. We prove by induction on n that for any finite separable extension F of K and for any
a ∈ F × the corresponding statement holds true with R0 = ak[a].
For both, n = 1 (trivially) and n > 1 (by induction hypothesis and Lemma 7.8.i), we know
the implication
x ∈ V Wn (ak[a])L =⇒ [x, a) = 0 .
Therefore, for arbitrary x ∈ Wn (ak[a])L we have [x, a) = [[x0 ], a) by the bilinearity of the
pairing and Lemma 1.1.13.i in [GAL] - we constantly will make use of the fact that the first
component of Witt vectors behaves additively. Moreover, again by the additivity of the pairing
and using (42) it suffices to prove (for all n > 0) that
[[ral ], a) = 0 for all r ∈ k × , l ≥ 1 .
Writing l = l0 pm with l0 and p coprime and denoting by r0 the pm th root of r we see that
0 m 0 0 m 0 0 m 0 m 0
l0 [[ral ], a) = [[(r0 al )p ], al ) = [[(r0 al )p ], al ) + [[(r0 al )p ], r0 ) = [[(r0 al )p ], r0 al )
n
by Lemma 7.8.ii because r0 ∈ (k × )p . Noting that l0 is a unit in Wn (k)L we are reduced to
m
the case x = [ap ] for m ≥ 0.
To this end let α0 ∈ F sep be in ℘−1 (a) and define α˜0 := ξ∈k/ im(χa ) (α0 + ξ). Then
Q
Y
NormF (α0 )/F (α˜0 ) = σ α˜0
σ∈Gal(F (α0 )/F )
Y Y
f:norm (43) = (α0 + ξ 0 + ξ)
ξ 0 ∈im(χa ) ξ∈k/im(χa )
Y
= (α0 + ξ) = a ,
ξ∈k
since α0 + ξ, ξ ∈ k, are precisely the zeros of X q − X − a.

m m
Now let β be in ℘−1 ([ap ]) with β0 = α0p . Then F (α0 )(β) = F (α0 , β0 , . . . , βn−1 ) =
F (α0 )(β − [β0 ]) and
m m
℘(β − [β0 ]) = [ap ] − ℘([α0p ])
m m m
belongs to V Wn (α0 k[α0 ])L because ap = (α0q −α0 )p belongs to α0 k[α0 ] as well as ℘([α0p ]) =
m m m m m
[(α0p )q ] − [α0p ] ∈ Wn (α0 k[α0 ])L and [ap ]0 = ap = ℘([α0p ])0 .
Note that for n = 1 we have F (α0 )(β) = F (α0 ) (i.e., the last consideration is not needed)
and since a is a norm with respect to the extension F (α0 )/F the claim follows from Lemma
7.8.iii.
Now let n > 1. Then, the induction hypothesis for F 0 := F (α0 ) and a0 := α0 implies that
m m
[[ap ] − ℘([α0p ]), α0 ) = 0, i.e., that α0 ∈ NormF (α0 )(β)/F (α0 ) (F (α0 )(β)× ) by Lemma 7.8.iii.
Replacing α0 by α0 + ξ, ξ ∈ k, we see that also α0 + ξ ∈ NormF (α0 )(β)/F (α0 ) (F (α0 )(β)× )
(note that F (α0 ) = F (α0 + ξ) and the composite F (α0 )(β) = F (α0 )F (β) does not depend
on the choices involved above). By the multiplicativity of the norm we obtain that α˜0 lies
in NormF (α0 )(β)/F (α0 ) (F (α0 )(β)× ), whence by transitivity of the norm and (43) a belongs to
NormF (α0 )(β)/F (F (α0 )(β)× ) and thus also to NormF (β)/F (F (β)× ) because F (β) ⊆ F (α0 )(β).
m
Thus [[ap ], a) = 0, again by Lemma 7.8.iii, as had to be shown.
Now we will define a second bilinear pairing

( , ) : Wn (K)L × K × −→ Wn (k)L
by using the residue pairing (cf. (13))
Res : AL × Ω1AL −→ oL .
To this end we choose an isomorphism AL /πL AL = k((Z)) ∼ = K and remark that our con-
struction will not depend on this choice of a prime element of K by Remark 3.4. Consider
the map d log : oL ((Z))× −→ Ω1AL , sending f to df
f . We define the upper pairing in
{ , }
Wn (AL )L × oL ((Z))× / oL
Φn−1 d log

AL × Ω1AL
Res / oL
via the commutativity of the diagram.

pairing2 Lemma 7.10. There is a unique well defined bilinear pairing ( , ) such that the diagram
{ , }
Wn (AL )L × oL ((Z))× / oL
Wn (α1 )L mod πL αn

( , )
Wn (K)L × K× / Wn (k)L ,
is commutative.
Proof. (Note that the reduction map oL ((Z))× → K × indeed is surjective.) We need to show
that
{ker(Wn (pr)L ), oL ((Z))× } ⊆ πLn oL and {Wn (AL )L , ker(oL ((Z))× → K × )} ⊆ πLn oL .
For a = (a0 , . . . , an−1 ) ∈ Wn (pr)L such that ai ∈ πL AL we obviously have Φn−1 (a) ∈ πLn AL .
Hence {a, oL ((Z))× } ⊆ πLn oL .
For the second inclusion we first observe that ker(oL ((Z))× → K × ) = 1 + πL oL [[Z]]. Hence
we have to prove that
f:claim (44) Res(Φn−1 (f )d log(1 + πL h)) ∈ πLn oL
holds true for all f = (f0 , . . . , fn−1 ) ∈ ALn and h ∈ oL [[Z]]. We observe that sending Z to
Z 0 := Z(1 + πL h) defines a ring automorphism first of oL [[Z]], then by localization of oL ((Z)),
and finally by πL -adic completion of AL . We write fi (Z) = gi (Z 0 ) and g := (g0 , . . . , gn−1 ),
and we compute
ResZ (Φn−1 (f )d log(1 + πL h)) = ResZ (Φn−1 (g(Z 0 ))d log(Z 0 )) − ResZ (Φn−1 (f (Z))d log(Z))
= ResZ 0 (Φn−1 (g(Z 0 ))d log(Z 0 )) − ResZ (Φn−1 (f (Z))d log(Z))
= ResZ ([Φn−1 (g(Z)) − Φn−1 (f (Z))]d log(Z)) .
Here the second equality uses the fact that the residue does not depend on the choice of the
variable (cf. Remark 3.4) while in the third equality we just rename the variable Z 0 into Z
both in the argument and the index of Res, which of course does not change the value. Now
note that, since Z 0 ≡ Z mod πL oL [[Z]], we have the congruences
fi (Z) = gi (Z 0 ) ≡ gi (Z) mod πL AL for any 0 ≤ i ≤ n − 1.
This implies that
Φn−1 (g(Z)) − Φn−1 (f (Z)) ≡ 0 mod πLn oL
(cf. [GAL] Lemma 1.1.2.i) whence the claim (44) by the oL -linearity of the residue.
Remark 7.11. Alternatively, one can define similarly a pairing by using the full ghost map
Φ = (Φ0 , . . . , Φn−1 ) via commutativity of the diagram
Wn (A L )L × oL ((Z))× / Wn (oL )L
_ _
Φ d log Φ

ALn × Ω1AL
Res / on
L
and by showing that for all f ∈ Wn (AL )L and h ∈ oL ((Z))× the residue vector (Res(Φi (f ) dh
h ))i
belongs to the image of (the right hand) Φ. We leave it to the interested reader to check that
this induces the same pairing as ( , ) above by applying Wn (pr)L to the target. For unramified
Witt vectors this is done in [Tho] Prop. 3.5. 4
Our aim is to show that the two pairings [ , ) and ( , ), in fact, coincide. This generalizes
a result of Witt ([Wit] Satz 18), which we learned from [Fo1, Fo2]. The strategy is to reduce
this to the comparison Pof the restrictions of the two pairings to Wn (k)L × K × .
For an element x = j xj Z j ∈ K with xj ∈ k and xj = 0 for j < vZ (x) (the valuation of
K) we set x+ := j≥1 xj Z j and x− := j<0 xj Z j . Then, for x = (x0 , . . . , xn−1 ) ∈ Wn (K)L ,
P P
with arbitrary n ≥ 1, we define iteratively elements (the ’constant term’ and the plus and
negative parts of x with respect to the variable Z)
ΩnZ (x) ∈ Wn (k)L , x+ ∈ Wn (Zk[[Z]])L , and x− ∈ Wn (Z −1 k[Z −1 ])L
such that
x = ΩnZ (x) + x+ + x− .
For n = 1 put
− − −
Ω1Z (x) := [x0 − x+ + +
0 − x0 ], x := [x0 ], and x := [x0 ]
and for n > 1 define
− − − −
ΩnZ (x) := [x0 − x+ n−1 + + +
0 − x0 ] + τ ΩZ (y), x := [x0 ] + τ y , and x := [x0 ] + τ y ,
− −
where y ∈ Wn−1 (K)L satisfies τ y = x − [x0 − x+ +
0 − x0 ] − [x0 ] − [x0 ].
4Another alternative formulation for the definition of ( , ) goes as follows: The residue pairing
Res : AL /πL
n
A × Ω1AL /πL n
n A −→ oL /πL oL
induces the pairing

n
im(Φn−1 ) + πL AL /πL
n
A × Ω1AL /πn A /d log(1 + πL oL [[Z]]) / oL /πLn oL
O L
O O
wn−1 wn−1
Wn (K)L × K×
( , )
/ Wn (k)L ,
where the middle vertical map is induced by d log and the inverse of the isomorphism oL ((Z))× /(1 +
πL oL [[Z]]) ∼
= K×.
unique Remark 7.12. x = ΩnZ (x) + x+ + x− is the unique decomposition of x ∈ Wn (K)L such that
the three summands lie in Wn (k)L , Wn (Zk[[Z]])L , and Wn (Z −1 k[Z −1 ])L , respectively.
Proof. Let x = a + a+ + a− be any decomposition such that a ∈ Wn (k)L , a+ ∈ Wn (Zk[[Z]])L ,
and a− ∈ Wn (Z −1 k[Z −1 ])L . Since the projection onto the zeroth component is additive we
− −
immediately obtain that a = [x0 − x+ +
0 − x0 ] + τ b, a+ = [x0 ] + τ b+ , and x− = [x0 ] +
τ b− for (uniquely determined) elements b ∈ Wn−1 (k)L , b+ ∈ Wn−1 (Zk[[Z]])L , and b− ∈
− −
Wn−1 (Z −1 k[Z −1 ])L . We put y := b + b+ + b− and obtain x = [x0 − x+ +
0 − x0 ] + [x0 ] + [x0 ] + τ y.
Hence y is the element in the above inductive construction for x. By induction with respect
to n we have b = Ωn−1 + − n
Z (y), b+ = y and b− = y . It follows that a = ΩZ (x), a+ = x , and
+
a− = x− .
omega Lemma 7.13. For any prime element Z in K and any x ∈ Wn (K)L we have
(x, Z) = ΩnZ (x) .
Proof. For x ∈ Wn (Zk[[Z]])L ∪ Wn (Z −1 k[Z −1 ])L we may choose the lift f of x to lie in
Wn (ZoL [[Z]])L and Wn (Z −1 oL [Z −1 ])L , respectively. It is straightforward to see that then
{f, Z} = Res(Φn−1 (f )d log Z) = Res(Φn−1 (f ) dZ n
Z ) = 0. By Remark 7.12 we have ΩZ (x) = 0
as well.
By the additivity of ( , ) in the first component it therefore remains to treat the case
that x ∈ Wn (k)L . Let x̃ ∈ Wn (W (k)L )L ⊆ Wn (AL )L be any lift of x. Then we have that
{x̃, Z} = Res(Φn−1 (x̃) dZ
Z ) = Φn−1 (x̃). But αn (Φn−1 (x̃)) = αn ◦ wn−1 (x) = φq
n−1 (x) = x
n
by (38) and (39) for oL . Hence (x, Z) = αn ({x̃, Z}) = αn (Φn−1 (x̃)) = x = ΩZ (x), the last
identity again by Remark 7.12.
constant Lemma 7.14. For any prime element Z in K and any x ∈ Wn (k)L we have [x, Z) = (x, Z).
Proof. We choose α ∈ Wn (k sep )L such that ℘(α) = x. Then K(α) ⊆ k(α)((Z)) is an un-
ramified extension of K = k((Z)). From local class field theory we therefore obtain that
recK (Z) = φq . It follows that [x, Z) = recK (Z)(α) − α = ℘(α) = x. On the other hand
Lemma 7.13 implies that (x, Z) = ΩnZ (x) = x as well.
constants Proposition 7.15. For any prime element Z in K, any a ∈ K × , and any x ∈ Wn (K)L we
have
[x, a) = [(x, a), Z) and (x, a) = ((x, a), Z) .
Proof. As ΩnZ ((x, a)) = (x, a) the second identity is a consequence of Lemma 7.13.
For the fist identity we first consider the special case a = Z. We will compare the decom-
positions
[x, Z) = [ΩnZ (x), Z) + [x+ , Z) + [−x− , Z −1 ) and
[(x, Z), Z) = [(ΩnZ (x), Z), Z) + [(x+ , Z), Z) + [(−x− , Z −1 ), Z)
term by term. By Lemma 7.13 the two first terms coincide and the remaining terms in the
second decomposition vanish. The last term in the first decomposition vanishes by Prop. 7.9.
Hence it remains to show that [x+ , Z) = 0. For this it suffices to check
P thati Wn (Zk[[Z]])L ⊆
℘(Wn (K)L ). Indeed, we claim that for y ∈ Wn (Zk[[Z]])L the series ∞ i=0 φq (y) converges in
Wn (k[[Z]])L (componentwise in the Z-adic topology). We observe that, for x ∈ Wn (k[[Z]])L
and z ∈ Wn (Z l k[[Z]])L with l ≥ 0, one has, by (41), the congruence
(x + z)i ≡ xi mod Z l k[[Z]]
for the components of thePrespective Witt vectors. It follows that each component of the
sequence of partial sums m i
i=0 φq (y) forms a Cauchy sequence. Since φq , being the compo-
nentwise qthPpoweri map, obviously is continuous for the topology under consideration we
obtain ℘(− ∞ i=0 φq (y)) = y.
For a general a we find a ν ∈ Z and another prime element Z 0 ∈ K such that a = Z ν Z 0 .
Using bilinearity and the special case (for Z as well as Z 0 ) we compute
[x, a) = ν[x, Z) + [x, Z 0 ) = ν[(x, Z), Z) + [(x, Z 0 ), Z 0 )
= [(x, Z), Z ν ) + [(x, Z 0 ), Z) = [(x, Z ν Z 0 ), Z)
= [(x, a), Z) ;
the third equality uses Lemma 7.14.
schmid-witt Theorem 7.16 (Schmid-Witt formula). The pairings [ , ) and ( , ) coincide.
Proof. This now is an immediate consequence of Lemma 7.14 and Prop. 7.15.
Witt Corollary 7.17. For all z ∈ Wn (K)L and û ∈ oL ((Z))× any lift of u ∈ K × we have
dû
Res(wn−1 (z) ) = wn−1 (∂(z)(recK (u))).
û
Proof. Let f ∈ Wn (AL )L be a lift of z. Thm. 7.16 implies that
dû
∂(z)(recK (u)) = αn (Res(Φn−1 (f ) ))
û
holds true. Applying wn−1 (for oL ) we obtain
dû
wn−1 (∂(z)(recK (u))) = Res(Φn−1 (f ) ) mod πLn oL .
û
On the other hand, by (38), the element Φn−1 (f ) module πLn is equal to wn−1 (z) (with wn−1
for AL ∼= AL ).
We finally are able to establish the congruence (36). First note that since A ⊆ W (Esep
L )L
we obtain the commutative diagram
φq −1
0 / oL /π n oL / A/π n A / A/π n A /0
L L L
∼ −1
= wn−1 =αn
φq −1
0 / Wn (k)L / Wn (Esep )L / Wn (Esep )L / 0.
L L
We recall that ∂ϕ and ∂ denote the connecting homomorphisms arising from the upper and
lower exact sequence, respectively. We obviously have the identity (αn )∗ ◦ ∂ϕ = ∂ ◦ αn for the
map αn which was defined in (37).
lemma-Witt Lemma 7.18. For any z ∈ AL and û ∈ oL ((ωLT ))× ⊆ A× ×
L any lift of u ∈ EL we have
dû
(45) Res(ϕn−1
L (z) ) ≡ ∂ϕ (z)(recEL (u)) mod πLn oL .
û
Proof. We use the identity in Cor. 7.17 for the element z 0 := αn (z) with respect to K = EL .
Since wn−1 ◦ αn = ϕn−1
L by (39) its left hand side becomes the left hand side of the assertion.
For the right hand sides we compute
wn−1 (∂(αn (z))(recEL (u))) = wn−1 ◦ αn (∂ϕ (z)(recEL (u))) = ∂ϕ (z)(recEL (u)) .
As explained at the end of section 6 this last lemma implies Prop. 6.3. The proof of Thm.
6.2 therefore now is complete.
8. Bloch and Kato’s as well as Kato’s explicit reciprocity law revisited

c:CoatesWiles
In this section we give a proof of a generalization of the explicit reciprocity law of Bloch and
Kato ([BK] Thm. 2.1) as well as of (a special case of) Kato’s explicit reciprocity law ([Kat]
Thm. II.2.1.7) replacing his method of syntomic cohomology by generalizing the method of
Fontaine in [Fo2] from the cyclotomic to the general Lubin-Tate case.
First we recall some definitions and facts from [Co1]. (This reference assumes that the power
series [πL ](Z) is a polynomial. But, by some additional convergence considerations, the results
can be seen to hold in general (cf. [GAL] §2.1 for more details).) The ideal IL ⊆ W (E e + )L
is defined to be the preimage of πL oCp under the surjective homomorphism of oL -algebras
θ : W (Ee + )L −→ oC (cf. [GAL] Lemma 1.4.17). [Co1] §8.5 introduces the πL -adic completion
p
e + )L [ 1 ] as well as B +
e + )L [ 1 IL ] ⊆ W (E 1
Amax,L of the subalgebra W (E πL πL max,L := Amax,L [ πL ] ⊆
+ +
BdR . The important point is that the Frobenius φq naturally extends to Amax,L ⊆ Bmax,L (but
+ +
not to B ). Let EL denote the algebraic closure of EL in E
dR
e and E its ring of integers. Setting
L
+ −1 +
ωL := ωLT , ω1 := φ−1
q (ωLT ) ∈ W (E )L , ξL := ωL ω1 ∈ W (EL )L and tL := logLT (ωL ) ∈
Amax,L we have the following properties by Prop.s 9.10 and 9.6 in loc. cit.:
+
f:invBmax (46) (Bmax,L )GL = L,
tL X −lq (k)
f:unitAmax (47) φq (tL ) = πL tL and =1+ ek ωLk−1 ∈ A×
max,L with ek ∈ πL oL ,
ωL
k≥2
where lq (k) denotes the maximal integer l such that q l ≤ k,

e + )L ξL = ker W (E θ
e + )L −→
f:xi (48) W (E oCp ,
e + )L ωL = {x ∈ W (E
W (E e + )L : θ(φi (x)) = 0 for all i ≥ 0},
q
+ +
f:tLBmax (49) Bmax,L tL = {x ∈ Bmax,L : θ(φiq (x)) = 0 for all i ≥ 0}.
Since ωL is a unit in W (EL )L we may define a φq - and ΓL -stable ring Asuf ⊆ W (EL )L as
+
the localization Σ−1 W (EL )L where Σ denotes the multiplicatively closed subset generated by
+ 1
{φm
q (ωL )}m≥0 . Setting Bmax,L := Bmax,L [ tL ] ⊆ BdR we obtain the following fact.
suf Lemma 8.1. Asuf ⊆ Bmax,L .

+ e + )L ⊆ Amax,L it suffices to prove that φm (ωL ) is invertible in
Proof. Since W (EL )L ⊆ W (E L q
Bmax,L . As ωL = tL a for some a ∈ A× max,L by (47) this is clear for m = 0 and follows for
m ≥ 1 because
φm m m m m
q (ωL ) = φq (tL )φq (a) = πL tL φq (a) .

By (47) we see that

−r
tL X
=: λm,r ωLm
ωL
m≥0
+
belongs to L[[ωL ]] ⊆ BdR for r ≥ 0.
m
P
vergenceBmax+ Lemma 8.2. i. m≥0 λm,r ωL converges in Amax,L .
P m−1 +
ii. m≥1 λm+r,r ωL converges in Bmax,L .
Proof. First of all we note that the πL -adic completion R of the polynomial ring oL [Z] is the
subring of all power series in oL [[Z]] whose coefficients tend to zero. Using the geometric series
we see that 1 + πL R ⊆ R× .
According to (47) and the proof of [Co1] Prop. 9.10 there exists a g(Z) ∈ R such that
r
tL
= 1 − πL g(u) ,
ωL
where u = ωπLL ∈ πL−1 IL for q =
6 2 and u = ωπL2 ∈ πL−2 I2L for q = 2, respectively. By the initial
L
observation we have X
(1 − πL g(Z))−1 = bm Z m ∈ R .
m≥0
Thus −r
tL X X
= bm um = λm,r ωLm
ωL
m≥0 m≥0
converges in Amax,L . For the second part of the assertion it remains to note that
−(r+1)
X X
λm+r,r ωLm−1 = πL bm+r+1 um .
m≥1 m≥0

Pr
Setting τr0 := m
m=0 λm,r ωL and τr := ωL−r τr0 ∈ L[ ω1L ] ⊆ W (EL )L [ π1L ] we have
f:taur (50) τr − t−r +
L ∈ L[[ωL ]] ⊆ BdR .
By [Co1] Prop. 9.25 (SEF 3E) we have the exact sequence
+
f:BmaxBdR (51) 0 −→ L −→ (Bmax,L )φq =1 −→ BdR /BdR −→ 0 .
We define
+ + +
F ilr Bmax,L := Bmax,L ∩ trL BdR for r ≥ 0.
FilBmaxexact Lemma 8.3. i. For r ≥ 1 the sequence
+ + π −r φq −1
0 −→ LtrL −→ F ilr Bmax,L −−L−−−−→ Bmax,L −→ 0
is exact.
ii. For r = 0 the sequence
+ φq −1
+
0 −→ L −→ Bmax,L −−−→ (φq − 1)Bmax,L −→ 0
+
is exact, and ((φq − 1)Bmax,L )GL = L.
+
iii. φq − 1 is bijective on ωL Bmax,L .
Proof. i. and ii. By [Co1] Prop. 9.22 we have, for any r ≥ 0, the exact sequence
+ + r +
0 −→ LtrL −→ (Bmax,L )φq =πL −→ BdR /trL BdR −→ 0 .
+ r + + r
First we deduce that LtrL = (Bmax,L )φq =πL ∩ trL BdR = (F ilr Bmax,L )φq =πL . Secondly it implies
+ + r + + + r +
that BdR = (Bmax,L )φq =πL + trL BdR and hence Bmax,L = (Bmax,L )φq =πL + F ilr Bmax,L . In the
+ + π −r φq −1
proof of [Co1] Prop. 9.25 it is shown that, for r ≥ 1, the map Bmax,L −−L−−−−→ Bmax,L is
+ −r r +
surjective. It follows that Bmax,L = (πL φq − 1)F il Bmax,L for r ≥ 1.
+
It remains to verify the second part of ii. By (46) we have ((φq − 1)Bmax,L )GL ⊆ L. For the
+ +
reverse inclusion it suffices to consider any a ∈ oL ⊆ W (EL )L . Since EL is integrally closed
+ + +
the map φq − 1 on W (EL )L is surjective. Hence we find a y ∈ W (EL )L ⊆ Bmax,L such that
(φq − 1)y = a.
+ + +
iii. First of all we note that φq (ωL Bmax,L ) ⊆ φq (ω1 ξ)Bmax,L ⊆ ωL Bmax,L , so that, in-
+ +
deed, φq − 1 restricts to an endomorphism of ωL Bmax,L . By ii. we have (ωL Bmax,L )φq =1 ⊆
+ +
(Bmax,L )φq =1 = L. But ωL Bmax,L ∩ L = 0 by (48). This proves the injectivity. It suffices to
establish surjectivity on ωL Amax,L . Let ωL a ∈ ωL Amax,L . Similarly as in the proof of Lemma
8.1 we let ωL = tL u with u ∈ A× max,L and compute
φnq (ωL a) = φnq (tL au) ∈ πLn tL Amax,L = ωL πLn Amax,L .

It follows that the series − n≥0 φnq (ωL a) converges (πL -adically) to some element ωL c ∈
P
ωL Amax,L such that (φq − 1)(ωL c) = ωL a.

The sequences in Lemma 8.3.i/ii induce, for any r ≥ 0, the connecting homomorphism in
continuous Galois cohomology
+ ∂r
L = (Bmax,L )GL −−→ H 1 (L, LtrL ) , 5
Note that as a GL -representation LtrL is isomorphic to V := L ⊗oL T (cf. Lemma 4.1.c and
(2)). We introduce the composite homomorphism
∂r res
δ r : L −−→ H 1 (L, LtrL ) −−→ HomΓL (HL , LtrL ) .
By Lemma 7.6 we also have the connecting homomorphism
H
∂ϕ : W (EL L )L [ π1L ] −→ H 1 (HL , L) .
Proposition 8.4. δ r (a) = ∂ϕ (τr a)trL for any a ∈ L.
Proof. Let n ≥ 0 such that πLn τr ∈ oL [ ω1L ], and assume without loss of generality that a
+
belongs to πLn oL , i.e., that τr0 a ∈ oL [ωL ] ⊆ W (EL )L . In order to compute δ r (a) we choose any
+
α ∈ F ilr Bmax,L such that (πL−r φq − 1)(α) = a. Then
δ r (a)(g) = (g − 1)α for all g ∈ HL .
5Setting Lr −r r +
adm := L ∩ (πL φq − 1)(F il Bmax,L ) we still may define
∂r
Lradm −→ H 1 (L, LtrL )
without knowing the right hand surjectivity in Lemma 8.3.i and define ∂ r with source Lradm instead. In the
course of the next Proposition one can then shown that Lradm = L.
In fact, choosing α is equivalent to choosing β := t−r −r + +

L α ∈ tL Bmax,L ∩ BdR such that
(φq − 1)(β) = b := t−r

L a .
+
On the other hand, to compute ∂ϕ (τr a) we note that τr0 a and ξLr belong to W (EL )L and that,
+ + +
since EL is integrally closed, the map φq − ξLr : W (EL )L → W (EL )L is surjective (argue
+
inductively with respect to the length of Witt vectors). Hence we find a y ∈ W (EL )L such
that
ϕL (y) − ξLr y = τr0 a .
By using (47) we see that ω1−r = ξLr ωL−r ∈ t−r + −r
L Bmax,L . It follows that the element β0 := ω1 y
belongs to Asuf ∩ t−r +
L Bmax,L ⊆ W (EL )L and satisfies
(φq − 1)(β0 ) = φq (ω1 )−r φq (y) − ω1−r y

= ωL−r (φq (y) − ξLr y)
= ωL−r τr0 a = τr a .
Hence
∂ϕ (τr a)(g) = (g − 1)β0 for all g ∈ HL .
+ + +
We observe that y ∈ W (EL )L ⊆ BdR , that ω1 is
a unit in BdR (since θ(ω1 ) 6= 0 by the first
−r + +
sentence in the proof of [Co1] Prop. 9.6), and hence that β0 ∈ tL Bmax,L ∩ BdR . At this point
+ H + −r + +
we are reduced to finding an element γ ∈ (ωL Bmax,L ) L ⊆ Bmax,L ⊆ tL Bmax,L ∩ BdR such
that (φq − 1)(γ) = (t−r
L − τr )a. We then put β := β 0 + γ and obtain
δ r (a)(g) = (g − 1)β ⊗ trL = (g − 1)(β0 + γ) ⊗ trL = (g − 1)β0 ⊗ trL = ∂ϕ (τr a)(g) ⊗ trL
for any g ∈ HL . In order to find γ it suffices, because of Lemma 8.3.iii, to observe that
X
t−r
L − τr = ωL (
+
λm+r,r ωLm−1 ) ∈ (ωL Bmax,L )HL
m≥1
by Lemma 8.2.ii.
Now we define the Coates-Wiles homomorphisms in this context for r ≥ 1 and m ≥ 0 by6
r
ψCW,m : lim o× −→ Lm
←− Ln
n
1 r−1

u 7−→ rm ∂inv ∆LT gu,t0 |Z=t0,m .
r!πL
Then the map
ΨrCW,m : lim o× −→ Lm trL
←− Ln
n
r
u 7−→ ψCW,m (u)trL .
r
is GL -equivariant (it depends on the choice of t0 ). In the following we abbreviate ψCW :=
r r r
ψCW,0 and ΨCW := ΨCW,0 . One might think about these maps in terms of the formal identity
X X
r
log gu,t0 (ωLT ) = ψCW (u)trL = ΨrCW (u) in L[[tL ]] ⊆ BdR .
r r
6For m > 0 one can extend the definition to lim L× while for m = 0 one cannot evaluate at t
n 0,0 = 0!
←−n
7
But instead of justifying in which sense me may insert gu,t0 (ωLT (tL )) into the logarithm
series, we shall only explain (and below use) the following identity
dgu,t0 (ωLT ) X r
d log gu,t0 (ωLT ) = = rψCW (u)tr−1
L dtL .
gu,t0 (ωLT )
r≥1
Indeed, tL = logLT (ωLT ) implies d

dtL ωLT = gLT (ωLT )−1 and hence
d d
f (ωLT ) = gLT (ωLT )−1 f (ωLT ) = ∂inv (f )(Z)|Z=ωLT .
dtL dωLT
We calculate

1 d r−1 1 dgu,t0 (ωLT ) 1 r−1
(r−1)! ( dtL ) gu,t0 (ωLT ) dtL
= (r−1)! ((∂inv ∆LT gu,t0 (Z))|Z=ωLT )|00 tL =000
|00 tL =000
1 r−1
= (r−1)! (∂inv ∆LT gu,t0 (Z))|Z=0
r
= rψCW (u) .
Proposition 8.5. For all a ∈ L, r ≥ 1, and u ∈ limn o× we have

←− Ln
r
arψCW (u) = ∂ϕ (τr a)(rec(u)) .
+
Proof. Using L[[tL ]]=L[[ωL ]]⊆ BdR we obtain from (50) that τr − t−r
L ∈ L[[tL ]]. By the dis-
cussion before Prop. 6.3 we therefore obtain
∂ϕ (τr a)(rec(u)) = ResωL (τr ad log gu,t0 (ωLT )) = RestL (τr ad log gu,t0 (ωLT ))
X
= RestL (at−r
L d log gu,t 0 (ω LT )) = RestL (at −r
L nψCWn
(u)tLn−1
dtL )
n≥1
r
= arψCW (u) .
With (51) also the sequence

diag
q φ =1
+ (x,y)7→x−y
f:BmaxplusBdR (52) 0 −→ L −−−→ Bmax,L ⊕ BdR −−−−−−−→ BdR −→ 0
is exact. Tensoring with V = L⊗oL T over L gives the upper exact sequence in the commutative
diagram
(53) 0 / V ⊗r / (B φq =1 ⊕ B + ) ⊗L V ⊗r / BdR ⊗L V ⊗r /0
f:Ltr O max,L
O dR
∼ ∼ −r −r ⊗r
j(atrL ):=at⊗r
0 = = (x,y)7→(xtL ,ytL )⊗t0
diag r (x,y)7→(x−y)t−r ⊗r
L ⊗t0
0 / Ltr / B φq =πL ⊕ tr B + / BdR ⊗L V ⊗r / 0.
L max,L L dR
7This power series has a constant term: see [Fo2] for a technical solution.
Passing to continuous GL -cohomology gives rise to the connecting isomorphism

1 ⊗r
exp:=expL,V ⊗r ,id 3 H (L,O V )
(BdR ⊗L V ⊗r )GL = tanL (V ⊗r ) ∼

=
+
H 1 (L, LtrL ),
which is the identity component (see Appendix) of the Bloch-Kato exponential map over L
for V ⊗r . We introduce the composite map
a7→at−r ⊗t⊗r expL,V ⊗r ,id res
L
expr : L −−−−−
∼
0
−−−→ tanL (V ⊗r ) −−−−−−−→ H 1 (L, V ⊗r ) −−→ HomΓL (HL , V ⊗r ) .
=
Proposition 8.6. For all a ∈ L we have

j −1 ◦ expr (a) = −δ r ((πL−r − 1)a) .
φq =π r +
Proof. By (53) we find (x, y) ∈ Bmax,LL ⊕ trL BdR such that x − y = a. Then
(j −1 ◦ expr (a))(g) = (g − 1)x = (g − 1)y for all g ∈ HL .
+ +
We claim that y belongs to F ilr Bmax,L . For this it suffices to prove that y lies in Bmax,L
+
(because it is contained in trL BdR by assumption). We know that y = x − a ∈ Bmax,L =
−s +
with respect to the property that y ∈ t−s +
S
t
s≥0 L B max,L . Let s be minimal L Bmax,L , i.e.,
+
that tsL y ∈ Bmax,L . We want to show that s = 0. Assume to the contrary that s > 0. Then
+ +
Bmax,L 3 φq (tL y) = πLis tsL φiq (y), for any i ≥ 0, belongs to F ils Bmax,L
i s ⊆ ker(θ) because
+
φiq (y) = φiq (x − a) = πLri x − a = πLri y + πLri a − a ∈ BdR .
+
By (49) we obtain tsL y = tL y 0 for some y 0 ∈ Bmax,L . Hence ts−1 +
L y already belongs to Bmax,L ,
which is a contradiction. The above claim follows.
In particular, by the definition of δ r and using that y = x − a we see that
(j −1 ◦ expr (a))(g) = (g − 1)y = δ r ((πL−r φq − 1)(y))(g) = −δ r ((πL−r − 1)a)(g) for g ∈ HL ,
r
φq =πL
because (πL−r φq − 1)(x) = 0 as x belongs to Bmax,L .
Putting the previous results together we obtain the following generalization of the explicit
reciprocity law of Bloch and Kato ([BK] Thm. 2.1) from the cyclotomic to the general Lubin-
Tate case. In particular, this confirms partly the speculations in [dS] §11: de Shalit had
suggested to find a replacement for Bmax,Qp (or rather Bcris which was used at that time) in
the context of general Lubin-Tate formal groups and it is precisely Colmez’ Bmax,L which has
this function (although the path in (loc. cit.) is slightly different from the one chosen here).
BK Theorem 8.7. For all u ∈ limn o× , a ∈ L, and r ≥ 1 we have the identities
←− Ln
δ (a)(rec(u)) = arΨrCW (u)
r
and
(j −1 ◦ expr (a))(rec(u)) = −(πL−r − 1)arΨrCW (u)
= 1
(r−1)! (1 − πL−r )a∂inv
r
log gu,t0 (Z)|Z=0 trL .
Finally we consider the following commutative diagram

HLab (p) ⊗Zp V ⊗−r × Homc (HL , V ⊗r ) /L
O O
rec⊗id
limn L× ⊗ V ⊗−r res

←− n Zp
cores(κ⊗id)

H 1 (L, V ⊗−r (1)) × H 1 (L, V ⊗r )
∪ / H 2 (L, L(1)) = L
O
∼ −1 cyc
exp∗ exp = c7→ctQp ⊗t0

0
DdR,L (V ⊗−r (1)) × tanL (V ⊗r ) / DdR,L (L(1))
O O O
a7→atrL t−1 ⊗−r
Qp ⊗(t0 ⊗tcyc ∼
0 ) =
∼ −r ⊗r
= b7→btL ⊗t0 ∼ −1 cyc
= c7→ctQp ⊗t0
(a,b)7→ab
L × L / L.
Here tcyc cyc

0 is a generator of the cyclotomic Tate module Zp (1), and tQp := logGm ([ι(t0 )+1]−1).
The commutativity of the upper part can be shown by taking inverse limits (on both sides) of
a similar diagram with appropriate torsion coefficients and afterwards tensoring with L over
oL . Its middle part is the definition of the dual exponential map exp∗ . The commutativity of
the lower part is easily checked. Note also that the composite of the middle maps going up is
nothing else than expr by definition. Thus setting dr := trL t−1 ⊗−r
Qp ⊗ (t0 ⊗ tcyc
0 ) we obtain the
following consequence.
Kato Corollary 8.8 (A special case of Kato’s explicit reciprocity law). For r ≥ 1 the diagram
limn o× ⊗ T ⊗−r
←− Ln Z
κ⊗id
00 (1−π −r )rψ r 00
1 (L , T ⊗−r (1)) L CW (− )dr
HIw ∞
cores
exp∗
&
H 1 (L, T ⊗−r (1)) / D 0 (V ⊗−r (1)) = Ldr ,
dR,L
commutes, i.e., the diagonal map sends u ⊗ at⊗−r

0 to
1 − πL−r r
a(1 − πL−r )rψCW
r
(u)dr = a ∂ log gu,t0 (Z)|Z=0 dr .
(r − 1)! inv
Appendix A. p-adic Hodge theory

For a continuous representation of GK on a finite dimensional Qp -vector space V we write
as usual
+
DdR,K (V ) := (BdR ⊗Qp V )GK ⊇ DdR,K
0
(V ) := (BdR ⊗Qp V )GK and
Dcris,K (V ) := (Bmax,Qp ⊗Qp V )GK .
0
The quotient tanK (V ) := DdR,K (V )/DdR,K (V ) is called the tangent space of V .
Henceforth we assume that V is de Rham. Then the usual Bloch-Kato exponential map
expK,V : tanK (V ) → H 1 (K, V ) can be defined as follows. Apply the tensor functor − ⊗Qp V
to the exact sequence
p φ =1 +
f:FESQp (54) 0 → Qp → Bmax,Qp
→ BdR /BdR →0
and take the (first) connecting homomorphism in the associated GK -cohomology sequence.8
+
Note that by [BK] Lemma 3.8.1 we have tanK (V ) = (BdR /BdR ⊗Qp V )GK . Furthermore, the
∗
dual exponential map expK,V is defined by the commutativity of the following diagram
exp∗K,V
(55) H 1 (K, V ) / D0
f:dualexp dR,K (V )
∼
= ∼
=
(expK,V ∗ (1) )∗
H 1 (K, V ∗ (1))∗ / (DdR,K (V ∗ (1))/D 0 ∗ ∗
dR,K (V (1))) ,
where the left, resp. right, perpendicular isomorphism comes local Tate duality, resp. from
the perfect pairing
f:pairingDdR (56) DdR,K (V ) × DdR,K (HomQ (V, Qp (1))) −→ DdR,K (Qp (1)) ∼
p = K,
0
in which the DdR,K -subspaces are orthogonal to each other. Note that the isomorphism K ∼
=
cyc
DdR,K (Qp (1)) sends a to at−1
Qp
∗
⊗ t0 . Also, (−) here means the Qp -dual.
Now assume that V is in RepL (GK ) and consider K = L in the following. Tensoring (29)
with Qp gives the isomorphism of L-vector spaces
Ξ̃ : L ∼
= HomZ (oL , Zp ) ⊗Z Qp ∼
p = HomQ (L, Qp ) .
p p
Since HomQp (L, −) is right adjoint to scalar restriction from L to Qp , and by using Ξ̃−1 in
the second step, we have a natural isomorphism
f:Ldual (57) ∼ HomL (V, HomQ (L, Qp ) ∼
HomQ (V, Qp ) = = HomL (V, L) .
p p
Combined with (56) we obtain the perfect pairing

iringDdRoverL (58) DdR,L (V ) × DdR,L (HomL (V, L(1))) −→ L
with an analogous orthogonality property. Furthermore, similarly as in Prop. 5.7 local Tate
duality can be seen as a perfect pairing of finite dimensional L-vector spaces
f:Tate-local (59) H i (K, V ) × H 2−i (K, HomL (V, L(1))) −→ H 2 (K, L(1)) = L .
Altogether we see that, for such a V , the dual Bloch-Kato exponential map can also be defined
by an analogous diagram as (55) involving the pairings (58) and (59) and in which (−)∗ means
taking the L-dual.
Since BdR contains the algebraic closure L of L we have the isomorphism
∼
=
Y
BdR ⊗Qp V = (BdR ⊗Qp L) ⊗L V −−→ BdR ⊗σ,L V
σ∈GQp /GL
which sends b⊗v to (b⊗v)σ . The tensor product in the factor BdR ⊗σ,L V is formed with respect
to L acting on BdR through σ. With respect to the GL -action the right hand side decomposes
8 It follows from [Co4, Prop. III.3.1] that this sequence splits in the category of topological Q -vector
ctssection p
spaces. Since the p-adic topology on Qp coincides with the induced topology from Bmax,Qp the existence of
the transition map is granted by [NSW, Lem. 2.7.2].
id
according to the double cosets in GL \GQp /GL . It follows, in particular, that DdR,L (V ) :=
+
G
(BdR ⊗L V ) L is a direct summand of DdR,L (V ). Similarly, tanL,id (V ) := (BdR /BdR ⊗L V )GL
is a direct summand of tanL (V ). We then have the composite map
⊆ expL,V
→ tanL (V ) −−−−→ H 1 (L, V ),
xpL,V,id : tanL,id (V ) −
eg
the identity component of expL,V . On the other hand, applying the tensor functor − ⊗L V to
the exact sequence (51)
φq =1 +
0 → L → Bmax,L → BdR /BdR →0
and taking the (first) connecting homomorphism9 in the associated GL -cohomology sequence
gives rise to a map
expL,V,id : tanL,id (V ) → H 1 (L, V ) .
Suppose that V is even L-analytic, i.e., that the Hodge-Tate weights of V at all embeddings
id 6= σ : L → L are zero. We then have tanL (V ) = tanL,id (V ) and the following fact.
id-component Proposition 1.1. If V is L-analytic, the maps expL,V , eg
xpL,V,id and expL,V,id coincide.
Because of this fact we call also expL,V,id the identity component of expL,V in the situation
of the Proposition. We remark that by [Se0] III.A4 Prop. 4 and Lemma 2(b) the character
V = L(χLT ) is L-analytic.
Proof of Prop. 1.1: Let L0 ⊆ L be the maximal unramified subextension and let f := [L0 : Qp ].
As explained at the beginning of §9.7 in [Co1] we have Bmax,L ⊆ Bmax,Qp ⊗L0 L and hence
q φ =1 p φf =1
Bmax,L ⊆ Bmax,Qp
⊗L0 L. We claim that
φ =1
q p φf =1p φ =1
Bmax,Qp
= Bmax,Qp
= Bmax,Qp
⊗Qp L0 .
The left hand side contains L0 . Let ∆ := Gal(L0 /Qp ) with Frobenius generator δ. For any
φfp =1 P −1
x ∈ Bmax,Q p
we have the finite dimensional L0 -vector space Vx := fi=0 L0 φip (x) on which
the Galois group ∆ acts semilinearly by sending δ to φp . Hilbert 90 therefore implies that
Vx = L0 ⊗Qp Vx∆ . This proves the claim.
It follows that we have the commutative diagram
0 /L / B φp =1 ⊗Q L / BdR /B + ⊗Q L /0
max,Qp p dR p
= mult

0 /C / B φq =1 ⊗L L / BdR /B + /0
O max,Qp 0
O dR
O
⊆ =
0 /L / B φq =1 / BdR /B + / 0,
max,L dR
in which the upper exact sequence is induced by tensoring (54) by L over Qp while the lower
one is (51). C is defined to be the kernel in the middle horizontal sequence which is therefore
also exact. Note that the two vertical maps L → C both coincide as their composites into
the middle term each sends l ∈ L to 1 ⊗ l. By tensoring this diagram with V over L and
9Analogous arguments as in footnote 8 grant the existence of this connecting homomorphism.
forming the cohomology sequences we conclude that the composites of expL,V and expL,V,id
with H 1 (GL , V ) → H 1 (GL , C ⊗L V ) coincide whence the claim shall follow from the injectivity
of the latter map. The snake lemma applied to the upper part of the diagram (tensored with
V ) leads to the exact sequence10
Y
+
0 → V → C ⊗L V → BdR /BdR ⊗σ,L V → 0 ,
σ6=id
which in turn, by forming continuous GL -cohomology, induces the exact sequence

0 = ker(tanL (V ) → tanL,id (V )) → H 1 (GL , V ) → H 1 (GL , C ⊗L V ) ,
i.e., we obtain the desired injectivity.
References
[Ben] Benois D.: On Iwasawa theory of crystalline representations. Duke Math. J. 104, 211-267 (2000)
[BM] Bentzen S., Madsen I.: Trace maps in algebraic K-theory and the Coates-Wiles homomorphism. J.
Reine Angew. Math. 411 , 171195 (1990)
[BK] Bloch S., Kato K.: L-functions and Tamagawa numbers of motives. The Grothendieck Festschrift,
Vol. I, 333-400, Progress Math., 86, Birkhäuser Boston 1990
[B-TG] Bourbaki N.: Topologie Générale. Chap. 1–4, 5–10. Springer 2007
[CC] Cherbonnier F., Colmez P.: Théorie d’Iwasawa des représentations p-adiques d’un corps local. J.
AMS 12, 241-268 (1999)
[CS] Coates J., Sujatha R.: Cyclotomic fields and zeta values. Springer 2006
[CW] Coates J., Wiles A.: On p-adic L-functions and elliptic units. J. Austral. Math. Soc. Ser. A 26 (1),
1-25 (1978)
[Col] Coleman R.: Division Values in Local Fields. Invent. math. 53, 91-116 (1979)
[Co1] Colmez P.: Espaces de Banach de dimension finie. J. Inst. Math. Jussieu 1, 331-439 (2002)
[Co2] Colmez P.: Fontaine’s rings and p-adic L-functions. Lecture Notes at Tsinghua Univ. 2004
[Co3] Colmez P.: (ϕ, Γ)-modules et représentations du mirabolique de GL2 (Qp ). In “Représentations p-
adiques de groupes p-adiques”, vol. II (Eds. Berger, Breuil, Colmez). Astérisque 330, 61-153 (2010)
[Co4] Colmez, P.: Théorie d’Iwasawa des représentations de de Rham d’un corps local. Ann. Math. 148
(2), 485-571 (1998)
[Co5] Colmez, P.: A generalization of Coleman’s isomorphism. In Algebraic number theory and related
topics (Kyoto, 1997). Srikaisekikenkysho Kkyroku 1026 , 110-112 (1998)
[dS] de Shalit, E.: The explicit reciprocity law of Bloch-Kato. Columbia University Number Theory
Seminar (New York, 1992). Astérisque 228, 4, 197-221 (1995)
[Fo1] Fontaine J.-M.: Répresentations p-adiques des corps locaux. In “The Grothendieck Festschrift”, vol.
II, 249-309, Birkhäuser 1990
[Fo2] Fontaine J.-M.: Appendice: Sur un théorème de Bloch et Kato (lettre à B. Perrin-Riou). Invent.
Math. 115, 151-161 (1994)
[FX] Fourquaux L., Xie B.: Triangulable OF -analytic (ϕq , Γ)-modules of rank 2. Algebra & Number Theory
7 (10), 2545-2592 (2013)
[FK] Fukaya T., Kato K.: A formulation of conjectures on p-adic zeta functions in non-commutative Iwa-
sawa theory. Proc. St. Petersburg Math. Soc., Vol. XII, AMS Transl. Ser. 2, vol. 219, 1-86 (2006)
[Haz] Hazewinkel M.: Formal Groups and Applications. Academic Press 1978
[Her] Herr L.: Sur la cohomologie galoisienne des corps p-adiques. Bull. Soc. Math. France 126, 563-600
(1998)
[HR] Hewitt E., Ross K.: Abstract Harmonic Analysis, Vol. I. Springer 1994
[Jen] Jensen C.U.: Les Foncteurs Dérivés de lim et leurs Applications en Théorie des Modules. Springer
←−
Lect. Notes Math., vol. 254 (1972)
10Using the facts from footnote 8 one checks that this sequence again satisfies the conditions of [NSW, Lem.
2.7.2] whence the existence of the long exact cohomology sequence below is granted.
[Kat] Kato K.: Lectures on the approach to Iwasawa theory for Hasse-Weil L-functions via BdR . I. Arith-
metic algebraic geometry (Trento, 1991), Springer Lect. Notes Math., vol. 1553, 50-163 (1993)
[KR] Kisin M., Ren W.: Galois representations and Lubin-Tate groups. Documenta Math. 14, 441-461
(2009)
[Koe] Kölcze P.: Ein Analogon zum Hilbertsymbol für algebraische Funktionen und Witt-Vektoren solcher
Funktionen. Diplomarbeit (Betreuer: J. Neukirch) Universität Regensburg 1990
[Lan] Lang S.: Cyclotomic Fields. Springer 1978
[Lau] Laubie F.: Extensions de Lie et groupes d’automorphismes de corps locaux. Compositio Math. 67,
165-189 (1988)
[Laz] Lazard M.: Groupes analytiques p-adiques. Publ. Math. IHES 26, 389-603 (1965)
[Mic] Michael E.: Continuous Selections II. Ann. Math. 64, 562-580 (1956)
[NSW] Neukirch J., Schmidt A., Wingberg K.: Cohomology of Number Fields. 2nd Ed., Springer 2008
[GAL] Schneider P.: Galois representations and (ϕ, Γ)-modules. Lecture Notes, Münster 2015,
http://wwwmath.uni-muenster.de/u/schneider/publ/lectnotes/index.html
[SV] Schneider P., Vigneras M.-F.: A functor from smooth o-torsion representations to (ϕ, Γ)-modules. In
“On Certain L-Functions” (Eds. Arthur, Cogdell,...). Clay Math. Proc., vol. 13, 525-601, AMS-CMI
2011
[Sch] Scholl A. J.: Higher fields of norms and (φ, Γ)-modules. Documenta Math. 2006, Extra Vol., 685-709
[Se0] Serre J.-P.: Abelian l-Adic Representations and Elliptic Curves. W.A. Benjamin 1968
[Ser] Serre J-P: Cohomologie Galoisienne. Springer Lect. Notes Math., vol. 5 (1973)
[Tho] Thomas L.: Ramification groups in Artin-Schreier-Witt extensions. J. Théorie des Nombres de Bor-
deaux 17, 689-720 (2005)
[Wi] Wiles, A.: Higher explicit reciprocity laws. Ann. Math. 107 (2), 235-254 (1978)
[Wit] Witt E.: Zyklische Körper und Algebren der Charakteristik p vom Grad pn . Struktur diskret bewerteter
perfekter Körper mit vollkommenem Restklassenkörper der Charakteristik p. J. Reine Angew. Math.
176, 126-140 (1936)
Universität Münster, Mathematisches Institut, Einsteinstr. 62, 48291 Münster, Germany,

http://www.uni-muenster.de/math/u/schneider/
E-mail address: [email protected]
Universität Heidelberg, Mathematisches Institut, Im Neuenheimer Feld 288, 69120 Heidel-

berg, Germany, http://www.mathi.uni-heidelberg.de/˜venjakob/
E-mail address: [email protected]

Iwasawa Coh Final

Uploaded by

Copyright:

Available Formats

Iwasawa Coh Final

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Iwasawa Coh Final

Uploaded by

Copyright:

Available Formats

COATES-WILES HOMOMORPHISMS AND IWASAWA COHOMOLOGY

FOR LUBIN-TATE EXTENSIONS

PETER SCHNEIDER AND OTMAR VENJAKOB

To John Coates on the occasion of his 70th birthday

Abstract. For the p-cyclotomic tower of Qp Fontaine established a description of local

1991 Mathematics Subject Classification. 11Sxx.

2. Coleman power series

([Col] Thm. 11).

([Col] Thm. 1).

Proof. See [Col] Thm. A and Cor. 17.

We now introduce the “logarithmic” homomorphism

Delta-N Lemma 2.4. i. ∆LT ◦ ϕL = πL ϕL ◦ ∆LT .

It follows that ∆LT restricts to a homomorphism

∆LT : (oL [[Z]]× )N =1 −→ oL [[Z]]ψCol =πL .

Its kernel is the subgroup µq−1 (L) of (q − 1)th roots of unity in o×

This further reduces us to proving that

The invariance of gLT (Z)dZ implies

3. Etale (ϕL , ΓL )-modules

ψL (ϕL (f1 )f2 ) = f1 ψL (f2 ) for any fi ∈ BL .

as well as the formula

:normoperator (9) ϕL ◦ NL = NormBL /ϕL (BL ) .

psi Remark 3.2. i. ψL (AL ) ⊆ AL and NL (AL ) ⊆ AL .

Proof. i. The homomorphism ϕL induces on AL /πL AL = kL ((Z)) the injective q-Frobenius

f:tracemod (10) traceBL /ϕL (BL ) (AL ) ⊆ πL AL .

σa : oL [[Z]] −→ oL1 [[Z]] ⊆ AL1

σa ◦ ϕL (f )(Z) = f ([πL ](a +LT Z)) = f (Z) .

The proof for NL is completely analogous.

ϕL (NL (f )([c](Z))) = NL (f )([c]([πL ](Z))) = NL (f )([πL ]([c](Z)))

ϕL ◦ NL (f ) mod πL AL = NormkL ((Z))/kL ((Z q )) (f mod πL AL ) ≡ f q mod πL AL

vi. Let f = 1 + πLm g with g ∈ AL . We compute

≡ 1 mod πLm+1 AL ≡ ϕL (1) mod πLm+1 AL

Proof. It is clear that (15) restricts to an injective homomorphism

is AL -linear. We pick a free presentation λ : ALn  M . Then we find an AL -linear map βe

M++ Lemma 3.11. i. M ++ is a treillis.

AL ⊗AL ,ϕL HomAL (M, N ) −→ HomAL (M, AL ⊗AL ,ϕL N )

f:inner (17) ϕHomA (M,N ) (α)(ϕM (m)) = ϕN (α(m)) .

mega-as-twist Lemma 3.14. The map

is an isomorphism of (ϕL , ΓL )-modules.

Hence ϕΩ1 (Z −1 dZ) = πL−1 qZ −1 dZ + πL−1 d(1+π L v)

phi-psi-dual Corollary 3.18. The residue pairing satisfies

[ψM (m), F ] = [m, ϕHomA n Ω1 ) (F )]

for all m ∈ M and all F ∈ HomAL (M, Ω1 /πLn Ω1 );

[ϕM (m), F ] = [m, ψHomA n Ω1 ) (F )]

for all m ∈ M and all F ∈ HomAL (M, Ω1 /πLn Ω1 ).

ii. Correspondingly we compute

4. The Kisin-Ren equivalence

L0 is the maximal unramified subextension of L. The Galois group GL obviously acts on E e

Since AL is a noetherian pseudocompact ring taking projective limits is exact. By 3) the

HomAL (DLT (V1 ), DLT (V2 )) = DLT (HomoL (V1 , V2 ))

for every V1 , V2 in RepoL (GL ). We also have

ψHomA (DLT (V1 ),DLT (V2 )) = ψDLT (HomoL (V1 ,V2 )) .

DLT (HomoL (V1 , V2 )) = (A ⊗oL HomoL (V1 , V2 ))HL

ai cj ⊗ αi (vj ). Assuming that cj = φq (c0j ) we compute

for any M annihilated by πLn . Lemma 3.14 implies the isomorphism

Proof. i. Since A ⊗oL V ∼

LPontrjagin Lemma 5.3. The isomorphism (30) restricts to an isomorphism

We use the open subsets

H 2 (K, L/oL (1))

is AL -linear. We pick a free presentation λ : ALn M . Then we find an AL -linear map βe