Poitou-Tate duality over extensions of global fields

Meng Fai Lim

arXiv:1104.1484v1 [math.NT] 8 Apr 2011

April 11, 2011

Abstract
In this paper, we are interested in the Poitou-Tate duality in Galois cohomology. We will
formulate and prove a theorem for a nice class of modules (with a continuous Galois action)
over a pro-p ring. The theorem will comprise of the Tate local duality, Poitou-Tate duality
and the Poitou-Tate’s exact sequence.

The classical Poitou-Tate duality is a duality principle for a local-global statement, namely it
relates the kernels of the localization maps. Using compactly supported cohomology groups,
one can give a cleaner formulation of the statement which we now do. Let F be a global field
with characteristic not equal to p, and let S be a finite set of primes of F containing all primes
above p and all archimedean primes of F . We let GF,S denote the Galois group Gal(FS /F ) of
the maximal unramified outside S extension FS of F inside a fixed separable closure of F . In its
usual formulation, Poitou-Tate duality relates the kernels of the localization maps on the GF,S -
cohomology of a module and the Tate twist of its Pontrjagin dual. For simplicity, we assume
that p is odd if F has any real places.
The nth compactly supported GF,S -cohomology group Hc,cts n (G
F,S , M ) with coefficients in a
topological GF,S -module M is defined as the nth cohomology group of the complex
 
resS M
Cone Ccts
·
(GF,S , M ) −→ ·
Ccts (GFv , M ) [−1],
v∈Sf

where GFv is the absolute Galois group of the completion of F at v, and resS is the sum of
restriction maps on the continuous cochain complexes. It therefore fits in a long exact sequence
M
n n n n+1
· · · → Hc,cts (GF,S , M ) → Hcts (GF,S , M ) → Hcts (GFv , M ) → Hc,cts (GF,S , M ) → · · · .
v∈S

We now let R denote a commutative complete Noetherian local ring with finite residue field of
characteristic p. Then we have the following formulation of Poitou-Tate duality due to Nekovář
[Ne, Prop. 5.4.3(i)].

1
Theorem. Let T be a finitely generated R-module with a continuous (R-linear) GF,S -action.
Then there are isomorphisms
n (G ∼ 3−n ∨ ∨
Hcts F,S , T ) −→ Hc,cts (GF,S , T (1))
∼ 3−n
Hc,cts (GF,S , T ) −→ Hcts (GF,S , T (1))∨
n ∨

of R-modules for all n, where T ∨ = Homcts (T, Qp /Zp ).

We now recall some notation from the language of derived categories. We denote by
D(M odR ) the derived category of R-modules which is obtained from from the category Ch(M odR )
of chain complexes of R-modules by inverting the quasi-isomorphisms, i.e., the maps of com-
plexes that induce isomorphisms on cohomology. We have the derived functors RHomR (−, −),
RΓcts (GF,S , −) and RΓc,cts (GF,S , −) that are obtained from HomR (−, −), Ccts (GF,S , −) and
Cc,cts (GF,S , −). Then Poitou-Tate duality can be reformulated as the following isomorphisms
 
∼
RΓcts (GF,S , T ) −→ RHomZp RΓc,cts (GF,S , T ∨ (1)), Qp /Zp [−3]
 
∼
RΓc,cts (GF,S , T ) −→ RHomZp RΓcts (GF,S , T ∨ (1)), Qp /Zp [−3]

in D(M odR ).
Now suppose that F∞ is a p-adic Lie extension of F contained in FS . We denote by Γ the
Galois group of the extension F∞ /F , and we let Λ = RJΓK denote the resulting Iwasawa algebra
over R. Let T be a finitely generated R-module with a continuous (R-linear) GF,S -action, and
let A be a cofinitely generated R-module with a continuous (R-linear) GF,S -action. The Λ-
modules of interest are the following direct and inverse limits of cohomology groups (and their
counterparts with compact support)
n n
lim Hcts (Gal(FS /Fα ), A) and lim Hcts (Gal(FS /Fα ), T ),
−→ ←−
Fα Fα

where the limits are taken over all finite Galois extensions Fα of F which are contained in F∞ .
By an application of Shapiro’s lemma, one can show that they are respectively isomorphic to
n (G
Hcts n
F,S , FΓ (A)) and Hcts (GF,S , FΓ (T )),

where the Λ-modules FΓ (A) and FΓ (T ) are defined by

lim HomR (R[Gal(Fα /F )], A) and lim R[Gal(Fα /F )] ⊗R T

−→ ←−
Fα Fα

respectively. Therefore, we can reduce the question of finding dualities on the Iwasawa modules
of interest to that of obtaining dualities over GF,S , but with R replaced by Λ.
In his monograph [Ne], Nekovář considers the above situation over a commutative p-adic
Lie extension (e.g., a Zrp -extension) and develops an extension of Poitou-Tate global duality for

2
the above cohomology groups. In view of the vast activity in the study of noncommutative
generalizations of the main conjecture of Iwasawa theory [CFKSV, FK, Kak, RW], one would
like to extend the above theory to the noncommutative setting.
In fact, in this paper, we study generalizations of the above duality of Poitou-Tate over
a general pro-p ring Λ (not necessarily commutative). Together with the module theory, we
carefully develop the theory of continuous group cohomology in our setting. From there, we are
able to state and prove our duality theorem (cf. Theorem 3.2.6).

Theorem. Let M be a bounded complex of objects that are profinite Λ-modules with a continuous
(Λ-linear) GF,S -action. Then we have the following isomorphism
M M

RΓ(Gv , M )[−1] / RHomZp RΓ(Gv , M ∨ (1)), Qp /Zp [−3]
v∈S v∈S


/ RHom

∨


RΓc (GF,S , M ) Zp RΓ(GF,S , M (1)), Qp /Zp [−3]


/ RHom

∨


RΓ(GF,S , M ) Zp RΓ c (GF,S M (1)), Qp /Zp [−3]
,

of exact triangles in D(M odΛ ).

We now give a brief description of the contents of each section of the paper. In Section 1,
we introduce notations and results from homological algebra required for the paper. Section
2 is about the discussion of profinite rings and their topological modules. We also introduce
continuous cohomology groups with coefficients in compact modules and discrete modules. In
Section 3, we will formulate and prove our duality theorems. In Section 4, we will apply the
duality theorems proved in Section 3 to extensions of global fields.

Acknowledgments. The material presented in this article forms a generalized version of part
of the author’s Ph.D. thesis [Lim]. The author would like to thank his Ph.D. supervisor Romyar
Sharifi for his advice and encouragement, for without which, this paper would not have been
possible. The author would also like to thank Manfred Kolster for his encouragement.

1 Preliminaries
We begin by reviewing certain objects and notation that will be used in this write-up. Most of
the material presented in this section can be found in [Hart, Ne, Wei]. Throughout the paper,
every ring is associative and has a unit.
Fix an abelian category C and denote the category of (cochain) complexes of objects in C
by Ch(C). We also denote the category of bounded below complexes, bounded above complexes

3
and bounded complexes by Ch+ (C), Ch− (C) and Chb (C) respectively. For each n ∈ Z, the
translation by n of a complex X is given by
n+i
X[n]i = X n+i , diX[n] = (−1)n dX .

If f : X −→ Y is a morphism of complexes, then f [n] : X[n] −→ Y [n] is given by f [n]i = f n+i .

If X is a complex, we have the following truncations of X:

σ≤i X = [· · · −→ X i−2 −→ X i−1 −→ X i −→ 0 −→ 0 −→ · · · ]

τ≤i X = [· · · −→ X i−2 −→ X i−1 −→ ker(diX ) −→ 0 −→ 0 −→ · · · ]
σ≥i X = [· · · −→ 0 −→ 0 −→ X i −→ X i+1 −→ X i+2 −→ · · · ]
τ≥i X = [· · · −→ 0 −→ 0 −→ coker(di−1
X ) −→ X
i+1 −→ X i+2 −→ · · · ].

The cone of a morphism f : X −→ Y is defined by Cone(f ) = Y ⊕ X[1] with differential

!
i diY f i+1
dCone(f ) = : Y i ⊕ X i+1 −→ Y i+1 ⊕ X i+2 .
0 −di+1
X

There is an exact sequence of complexes

j p
0 −→ Y −→ Cone(f ) −→ X[1] −→ 0,

where j and p are the canonical inclusion and projection respectively. The corresponding bound-
ary map
δ : H i (X[1]) = H i+1 (X) −→ H i+1 (Y )
is induced by f i+1 .
If X is a complex and x ∈ X i , we write x̄ = i for the degree.
Let Λ, S and T be rings. Let M (resp., N ) be a left Λ-S-bimodule (resp., a Λ-T -bimodule).
Then HomΛ (M, N ) is taken to be the S-T -bimodule of all left Λ-module homomorphisms from
M to N , where the left S-action is given by (s · f )(m) = f (ms) and the right T -action is given
by (f · t)(m) = f (m)t for f ∈ HomΛ (M, N ), m ∈ M, s ∈ S and t ∈ T . If M • is a complex
of Λ-S-bimodules and N • a complex of Λ-T -bimodules, we define a complex Hom•Λ (M • , N • ) of
S-T -bimodules by Y
HomnΛ (M • , N • ) = HomΛ (M i , N i+n )
i∈Z

with differentials defined as follows: for f ∈ HomΛ (M i , N i+n ), we have

df = di+n n i−1
N ◦ f + (−1) f ◦ dM .

In the case when S = T , we have a similar definition for the complexes Hom•Λ−S (M • , N • ) of
abelian groups, where HomΛ−S (M, N ) is the group of all of Λ-S-bimodule homomorphisms from

4
M to N . It follows immediately from the definition that for an element f ∈ Hom0Λ−S (M • , N • ),
we have f ∈ HomCh(Λ−S) (M • , N • ) if and only if df = 0. Here Ch(Λ − S) denotes the category
of complexes of Λ-S-bimodules.
Suppose that M • is a complex of Λ-S-bimodules and L• a complex of S-T -bimodules. We
define the complex M • ⊗S L• of Λ-T -bimodules by
M
(M • ⊗S L• )n = M i ⊗S Ln−i
i∈Z

with differentials
d(m ⊗ l) = dm ⊗ l + (−1)m̄ m ⊗ dl.
We end the section by collecting some technical results which will be used in the paper.

Lemma 1.1. The following formulas define isomorphisms of complexes:

Hom•Λ (M • , N • )[n] ∼
= Hom•Λ (M • , N • [n])
f 7→ f
(M • [n]) ⊗S L• ∼ = (M • ⊗S L• )[n]
m ⊗ l 7→ m ⊗ l
M ⊗S (L• [n]) ∼
•
= (M • ⊗S L• )[n]
m ⊗ l 7→ (−1)nm̄ m ⊗ l.

Proof. This follows from a straightforward verification of the definition of translation and the
sign conventions.

Lemma 1.2. The adjunction morphisms define morphisms

Hom•Λ−T (M • ⊗S L• , N • ) −→ Hom•Λ−S M • , Hom•T o (L• , N • )


f 7→ m 7→ (l 7→ f (m ⊗ l))


Hom•Λ−T (M • ⊗S L• , N • ) −→ Hom•S−T L• , Hom•Λ (M • , N • )


f 7→ l 7→ (m 7→ (−1)m̄l̄ f (m ⊗ l))

of complexes and morphisms

HomCh(Λ−T ) (M • ⊗S L• , N • ) −→ HomCh(Λ−S) M • , Hom•T o (L• , N • )


HomCh(Λ−T ) (M • ⊗S L• , N • ) −→ HomCh(S−T ) L• , Hom•Λ (M • , N • )

of abelian groups. All of these maps are monomorphisms; they are isomorphisms if M • and L•
are bounded above and N • is bounded below.

Lemma 1.3. Given the following data:

(1) Complexes A1 , B1 of Λ-S-modules, complexes A2 , B2 of S-T -modules, and complexes
A3 , B3 of Λ-T -modules.

5
 (2) Morphisms of complexes fj : Aj −→ Bj preserving the respective module structures.
(3) Morphisms of complexes of Λ-T -modules

∪A : A1 ⊗S A2 −→ A3
∪B : B1 ⊗S B2 −→ B3

such that f3 ◦ ∪A = ∪B ◦ (f1 ⊗ f2 ). For j = 1, 2, 3, define Ej to be the complex

fj

Cone Aj −→ Bj [−1].

Then we have morphisms of complexes

∪0 , ∪1 : E1 ⊗S E2 −→ E3

given by the formulas

(a1 , b1 ) ∪0 (a2 , b2 ) = a1 ∪A a2 , (−1)ā1 f1 (a1 ) ∪B b2


(a1 , b1 ) ∪1 (a2 , b2 ) = a1 ∪A a2 , b1 ∪B f2 (a2 ) ,

and the formula

s (a1 , b1 ) ⊗ (a2 , b2 ) = 0, (−1)ā1 b1 ∪B b2
defines a homotopy s : ∪1 ∪0 .

Proof. This is a special case of [Ne, Prop. 1.3.2].

2 Profinite rings
Completed group algebras of profinite groups arise naturally in the study of Iwasawa theory,
and such rings are profinite rings. In this section, we shall study the properties of profinite rings
and their (topological) modules. We will also develop a cohomological theory over such rings.
Throughout the section, Λ will always denote a profinite ring, and I is a directed fundamental
system of open neighborhoods of zero consisting of two-sided ideals of Λ.

2.1 Topological Λ-modules

In this section, we will study the topological modules over a profinite ring Λ. These are Hausdorff
topological abelian groups with a continuous Λ-action. In particular, we are interested in the
following two classes of topological Λ-modules.

Definition 2.1.1. We say that a topological Λ-module M is a compact (resp., discrete) Λ-

module if its underlying topology is compact (resp., discrete). The category of compact Λ-
modules (resp., discrete Λ-modules) is denoted by CΛ (resp., DΛ ).

6
 The following proposition records some of the properties of the above two categories, whose
proof can be found in [RZ, Chap. 5].
Proposition 2.1.2. (i) Every compact Λ-module is a projective limit of finite modules and has
a fundamental system of neighborhoods of zero consisting of open submodules. In particular, it
is an abelian profinite group.
(ii) Every discrete Λ-module is the direct limit of finite Λ-modules. In particular, it is an
abelian torsion group.
(iii) Pontryagin duality induces a duality between the category CΛ of compact Λ-modules and
the category DΛo of discrete Λo -modules.
(iv) The category CΛ is abelian and has enough projectives and exact inverse limits. The
category DΛ is abelian and has enough injectives and exact direct limits.

We give another description of discrete Λ-modules. If M is a Λ-module and A is a two-sided

ideal of Λ, we define
M [A] = {x ∈ M | A ⊆ Ann(x)}.
With this, we have the following lemma.
Lemma 2.1.3. Let M be an abstract Λ-module. Then M is a discrete Λ-module (i.e., the
Λ-action is continuous with respect to the discrete topology on M ) if and only if
[
M= M [A].
A∈I

Proof. Suppose that M is a discrete Λ-module. Let x ∈ M . Then by the continuity of the
Λ-action, there exists A ∈ I such that A · x = 0. This implies that x ∈ M [A].
Conversely, suppose that [
M= M [A].
A∈I
We shall show that the action
θ : Λ × M −→ M
is continuous, where M is given the discrete topology. In other words, for each x ∈ M , we
need to show that θ −1 (x) is open in Λ × M . Let (λ, y) ∈ θ −1 (x). Then y ∈ M [A] for some
A ∈ I. Therefore, we have (λ, y) ∈ (λ + A) × {y}, and the latter set is an open set contained in
θ −1 (x).

When working with topological Λ-modules, one will have to consider continuous homomor-
phisms between the modules. In general, an abstract homomorphism of modules may not be
continuous. In the next lemma, we record a few situations where every abstract homomorphism
is continuous. We say that a topological Λ-module M is endowed with the I-adic topology if
the collection {AM }A∈I forms a fundamental system of neighborhoods of zero.

7
Lemma 2.1.4. Let M and N be two topological Λ-modules. Suppose one of the following cases
holds.
(1) Both M and N have the I-adic topology.
(2) Both M and N have the discrete topology.
(3) M is a finitely generated Λ-module endowed with the I-adic topology, and N is a compact
Λ-module.
(4) M is a finitely generated Λ-module endowed with the I-adic topology, and N is a discrete
Λ-module.
Then every abstract Λ-homomorphism is continuous. In other words, we have

HomΛ, cts (M, N ) = HomΛ (M, N ).

Proof. (1) and (2) are straightforward.

(3) Suppose M is generated by e1 , ..., er . Let f : M −→ N be an abstract Λ-homomorphism,
and for each i, set xi = f (ei ). Let V be an open Λ-submodule of N . By continuity of the
Λ-action on N , for each i, there exists Ai ∈ I such that Ai · xi ⊆ V . Since I is directed, we can
find A ∈ I such that A ⊆ Ai for all i. It follows that f (AM ) ⊆ V , establishing the continuity of
f.
(4) We retain the notation in (3). By Lemma 2.1.3, for each i, there exists Ai ∈ I such that
xi ∈ N [Ai ]. Since I is directed, we can find A ∈ I such that A ⊆ Ai for all i, and f (AM ) = 0.

Corollary 2.1.5. Let M be a compact Λ-module. Then every finitely generated abstract Λ-
submodule of M is a closed subset of M . In particular, every finitely generated left (or right)
ideal of Λ is closed in Λ.

Proof. Let N be a Λ-submodule of M generated by x1 , ..., xr . By Lemma 2.1.4(3), the following

Λ-homomorphism L
φ : ri=1 Λ −→ M
ei 7→ xi
Lr
is continuous. Since i=1 Λ is compact, so is its image N .

Corollary 2.1.6. Let M be a finitely presented abstract Λ-module. Then M is a compact Λ-

module.
f
Proof. Since M is finitely presented, we have an exact sequence Λr −→ Λs −→ M −→ 0 for
some integers r and s. By Lemma 2.1.4(1), the map f is a continuous Λ-homomorphism of
compact Λ-modules. Since the category CΛ is abelian by 2.1.2(iv), it follows that M is an object
in CΛ .

In view of Corollary 2.1.6, one may ask the following two questions. The first is if one can say
anything about the I-adic topology on an abstract Λ-module M . In general, it is not even clear

8
whether this topology is Hausdorff. The second question that one may ask is if there are other
ways to endow a finitely presented Λ-module with a topology such that it becomes a compact
Λ-module. In response to these two questions, we have the following proposition. In fact, as we
shall see, if M is already a compact Λ-module, the I-adic topology is Hausdorff, and it is the
only one with which one can endow a finitely presented Λ-module in order to make it into a
compact Λ-module. One may compare the following proposition with [NSW, Prop. 5.2.17].

Proposition 2.1.7. Let M be a compact Λ-module. Then the I-adic topology is finer than the
original topology of M , and the canonical homomorphism

α : M −→ lim M/AM
←−
A∈I

of Λ-modules is injective. Furthermore, if M is a finitely generated Λ-module, then the topologies

coincide, and the above homomorphism is a continuous isomorphism of compact Λ-modules.

Proof. : Let N be an open submodule of M . Then by continuity, for each x ∈ M , there exists
a neighborhood Vx of x and Ax ∈ I such that Ax Vx ⊆ N . Since M is compact, it is covered by
finitely many such sets, say Vx1 , Vx2 , ..., Vxr . Choose A ∈ I such that A ⊆ Ai for all i = 1, ..., r.
Then we have AM ⊆ N , and this shows the first assertion. Since M is Hausdorff under its
original topology, it follows that M is Hausdorff under the I-adic topology and so
\
ker α = AM = 0.
A∈I

Now if M is finitely generated, we have a surjection

Λm ։ (M with I−adic topology),

which is continuous by Lemma 2.1.4(1). This implies that M with the I-adic topology is
compact. By the first assertion, the identity map

(M with I−adic topology) −→ M

is continuous. This in turn gives a continuous bijection between compact spaces and is therefore
a homeomorphism. If M is given the I-adic topology, then the image of α is dense in lim M/AM ,
←−
A∈I
and so is surjective since M is compact.

We conclude with a description of projective objects in CΛ that are finitely generated over
Λ.

Proposition 2.1.8. Let P be a projective object in CΛ that is finitely generated over Λ. Then
P is a projective Λ-module. Conversely, let P be a finitely generated projective Λ-module. Then
P , endowed with the I-adic topology, is a compact Λ-module and is a projective object in CΛ .

9
Proof. Let P be a projective object in CΛ that is finitely generated over Λ. Then there is a
surjection f : Λr ։ P of Λ-modules. By Proposition 2.1.7, the topology on P is precisely the
I-adic topology, and it follows from Lemma 2.1.4(1) that f is a continuous homomorphism of
compact Λ-modules. Now since P is a projective object in CΛ , the map f has a continuous
Λ-linear section. In particular, this implies that we have an isomorphism Λr ∼= P ⊕ (ker f ) of
Λ-modules. Hence P is a projective Λ-module.
Conversely, suppose that P is a finitely generated projective Λ-module. Then there exists a
finitely generated projective Λ-module Q such that P ⊕ Q is a free Λ-module of finite rank. We
then have a surjection π : Λn ։ Q, and this gives a finite presentation

Λn −→ P ⊕ Q −→ P −→ 0

of P where the first map sends an element x of Λn to (0, π(x)) and the second map is the
canonical projection. It then follows from Proposition 2.1.6 that P is a compact Λ-module
under the I-adic topology. Now suppose we are given the following diagram

P
α
ε 
M // // N

of compact Λ-modules and continuous Λ-homomorphisms. Since P is a projective Λ-module,

there is an abstract Λ-homomorphism β : P → M such that εβ = α. On the other hand,
it follows from Lemma 2.1.4(3) that β is also continuous. Therefore, this shows that P is a
projective object of CΛ .

2.2 Continuous cochains

Definition 2.2.1. Let G be a profinite group. We define CΛ,G to be the category where the
objects are compact Λ-modules with a continuous Λ-linear G-action and the morphisms are
continuous Λ[G]-homomorphisms. Similarly, we define DΛ,G to be the category where the objects
are discrete Λ-modules with a continuous Λ-linear G-action and the morphisms are (continuous)
Λ[G]-homomorphisms.

Proposition 2.2.2. (i) The category CΛ,G is abelian, has enough projectives and exact inverse
limits.
(ii) The category DΛ,G is abelian, has enough injectives and exact direct limits.
(iii) The Pontryagin duality induces a contravariant equivalence between CΛ,G and DΛo ,G
(resp. CΛo ,G and DΛ,G ).

Proof. We shall prove (iii) first. By Proposition 2.1.2, it suffices to show that if M (resp., N ) is
an object of CΛ,G (resp., DΛo ,G ), then M ∨ (resp., N ∨ ) is an object of DΛo ,G (resp., CΛ,G ). We

10
define a G-action on M ∨ by σ · f (m) = f (σ −1 m) for f ∈ M ∨ , σ ∈ G and m ∈ M . This is clearly
Λo -linear, and since G is profinite, we may apply [F, Prop. 3] to conclude that the G-action is
continuous. The same argument works for N . Hence we have proven (iii). It remains to prove
(ii), since (i) will follow from (ii) and (iii).
To prove (ii), we note that it is clear that DΛ,G is abelian and has exact direct limits. It
remains to show that it has enough injectives. By the lemma to follow, we see that the functor
[[
U
M 7→ M [A] : M odΛ[G] −→ DΛ,G
A∈I U

is right adjoint to an exact functor, and so preserves injectives by [Wei, Prop. 2.3.10]. Since
M odΛ[G] has enough injectives, it follows that DΛ,G also has enough injectives.

Lemma 2.2.3. An abstract Λ[G]-module N is an object in DΛ,G if and only if

[[
U
N= N [A] ,
A∈I U

where U runs through all the open subgroups of G. Moreover, if M is an abstract Λ[G]-module,
then [[
U
M [A] ,
A∈I U

is an object of DΛ,G , and there is a canonical isomorphism

 [[
U 
HomΛ[G],cts N, M [A] ∼
= HomΛ[G] (N, M )
A∈I U

for every N ∈ DΛ,G .

Proof. Suppose N is an object in DΛ,G . Then, in particular, it is a discrete Λ-module. By

S
Lemma 2.1.3, we have N = A∈I N [A]. Let x ∈ N [A]. Then by continuity of the G-action,
there exists an open subgroup U of G such that U · x = x.
Conversely, suppose that [[
U
N= N [A] .
A∈I U
S
Clearly this implies that N = A∈I N [A], and so N is a discrete Λ-module. It remains to show
that the G-action
θ : G × N −→ N
is continuous. Let x ∈ N , and let (σ, y) ∈ θ −1 (x). Then y ∈ N [A]U for some A ∈ I and open
subgroup U . In particular, we have (σ, y) ∈ σU × {y} ⊆ θ −1 (x). Therefore, this proves the first
assertion. The second assertion is an immediate consequence of the first.

11
Lemma 2.2.4. Let M be an object of CΛ,G . Then M has a fundamental system of neighborhoods
of zero consisting of open Λ[G]-submodules.

Proof. Let N be an open Λ-submodule of M . Then for each g ∈ G, there exists an open Λ-
submodule Ng of M and an open subgroup Ug of G such that gUg · Ng ⊆ N . Since G is compact,
it is covered by finite number of such cosets, say g1 Ug1 , ..., gr Ugr . Set N0 = ∩ri=1 Ngi . This is an
open Λ-submodule of M . Then Λ[G] · N0 is a Λ[G]-submodule of M which contains N0 and is
contained in N .

For the remainder of the subsection, we will be studying the continuous cochain complex
(and its cohomology) of G with coefficients in certain classes of topological Λ-modules.

Definition 2.2.5. Let M be a topological Λ-module with a continuous Λ-linear G-action. The
(inhomogeneous) continuous cochains Ccts i (G, M ) of degree i ≥ 0 on G with values in M are

defined to be the left Λ-module of continuous maps Gi → M with the usual differential
i
X
i
(δ c)(g1 , ..., gi+1 ) = g1 c(g2 , ..., gi+1 ) + (−1)j c(g1 , ..., gj gj+1 , ..., gi+1 ) + (−1)i−1 c(g1 , ..., gi ),
j=1

i (G, M ) to C i+1 (G, M ). It then follows that

which maps Ccts α cts α

i M i+1 δi
· · · −→ Ccts (G, M ) −→ Ccts (G, M ) −→ · · ·

i (G, M ). The following

is a complex of Λ-modules and its ith cohomology group is denoted by Hcts
lemma is a standard result (cf. [NSW, Lemma 2.7.2]).

Lemma 2.2.6. Let

α β
0 −→ M ′ −→ M −→ M ′′ −→ 0
be a short exact sequence of topological Λ-modules with a continuous Λ-linear G-action such that
the topology of M ′ is induced by that of M and such that β has a continuous (not necessarily
Λ-linear) section. Then

• α β∗
0 → Ccts (G, M ′ ) →∗ Ccts
• •
(G, M ) → Ccts (G, M ′′ ) → 0

is an exact sequence of complexes of Λ-modules.

We are particularly interested in the case when M is an object of CΛ,G or DΛ,G . We now
discuss cohomology and limits.

12
Proposition 2.2.7. Let N = lim Nα be an object of DΛ,G , where Nα ∈ DΛ,G . Then we have an
−→
α
isomorphism
i
Ccts (G, N ) ∼
= lim C i (G, Nα )
−→ cts
α

of continuous cochain groups which induces an isomorphism

i
Hcts (G, N ) ∼
= lim H i (G, Nα )
−→ cts
α

of cohomology groups.

Proof. The first isomorphism is immediate and the second follows from the first since direct
limit is exact.

In the next proposition, we shall examine the relationship between cohomology and inverse
limit. We shall denote lim(i) to be the ith derived functor of lim.
←− ←−
Proposition 2.2.8. Let M = lim Mα be an object in CΛ,G , where each Mα is finite. Then we
←−
α
have an isomorphism
Ccts (G, M ) ∼
= lim C (G, Mα )
← − cts
α

of complexes of Λ-modules and a spectral sequence

j i+j
lim(i) Hcts (G, Mα ) =⇒ Hcts (G, M ).
←−
α

m (G, N ) is finite for all finite discrete Λ-modules

Suppose further that G has the property that Hcts
N with a continuous commuting G-action and for all m ≥ 0. Then
i
Hcts (G, M ) ∼
= lim H i (G, Mα ).
←− cts
α

Proof. The first assertion is immediate from the definition. The second assertion follows from
a similar argument as in [Ne, Prop. 8.3.5]. We consider the two hypercohomology spectral
i (G, M ):
sequences for the functor lim and the inverse system Ccts
←− α

lim(j) Ccts
i (G, M ) =⇒ H i+j
α
← −
α
j
lim(i) Hcts (G, Mα ) =⇒ H i+j .
←−
α

For each i, it is clear that

i i
lim Ccts (G, Mα ) −→ Ccts (G, Mα )
←−
α

13
 i (G, M ) is “weakly flabby” in the sense of
is surjective for every α, and so the inverse system Ccts α
[Jen, Lemma 1.3]. Therefore, by [Jen, Thm. 1.8], we have that lim(j) Ccts i (G, M ) = 0 for j > 0.
α
←−
α
Hence, the first spectral sequence degenerates and we obtain



H i = H i lim Ccts (G, Mα ) ∼= H i Ccts (G, M ) = Hcts
i
(G, M ).
←−
α

For the last assertion, the additional assumption allows one to invoke [Jen, Cor. 7.2] to
j
conclude that lim(i) Hcts (G, Mα ) = 0 for i > 0.
← −
α

For the remainder of the subsection, we let C denote either CΛ,G or DΛ,G . Let M • be a
complex of objects in C with differentials denoted by diM . We define Ccts
• (G, M • ) by
M j
n
Ccts (G, M • ) = Ccts (G, M i ).
i+j=n
i+j i+j j i
Its differential δM • is determined as follows: restriction of δM • to Ccts (G, M ) is the sum of

j j
(diM )∗ : Ccts (G, M i ) −→ Ccts (G, M i+1 )
and
j j j+1
(−1)i δM i i
i : Ccts (G, M ) −→ Ccts (G, M ).

We denote its ith cohomology group by H i (G, M • ).

α β
Proposition 2.2.9. Let 0 → M ′ → M → M ′′ → 0 be an exact sequence of objects in C. Then
• α β∗
0 → Ccts (G, M ′ ) →∗ Ccts
• •
(G, M ) → Ccts (G, M ′′ ) → 0
is an exact sequence of complexes of Λ-modules. The statement also holds true if we replace
M ′ , M, M ′′ by complexes of objects in C.
Proof. By Lemma 2.2.6, it suffices to show that β has a continuous section. If C = DΛ,G , this is
obvious. In the case when C = CΛ,G , since every compact Λ-module is profinite by Proposition
2.1.2, every continuous surjection has a continuous section.

Let M • be a complex of objects in C. The filtration τ≤j M • induces a filtration

•
τ≤j Ccts (G, M • ) = Ccts
•
(G, τ≤j M • )
on the cochain groups which fit into the following exact sequence of complexes
•
0 −→ Ccts (G, τ≤j M • ) −→ Ccts
•
(G, τ≤j+1 M • ) −→ τ≤j+1 Ccts
•
(G, M • )/τ≤j Ccts
•
(G, M • ) −→ 0
by Proposition 2.2.9. This filtration gives rise to the following hypercohomology spectral se-
quence
i

i+j
Hcts G, H j (M • ) =⇒ Hcts (G, M • ),
which is convergent if M • is cohomologically bounded below.

14
Lemma 2.2.10. Let f : M • −→ N • be a quasi-isomorphism of cohomologically bounded below
complexes of objects in C. Then the induced map
•
f∗ : Ccts (G, M • ) −→ Ccts
•
(G, N • )

is also a quasi-isomorphism.
Proof. The map f induces isomorphisms
i

∼

Hcts i
G, H j (M • ) −→ Hcts G, H j (N • ) .

By convergence of the above spectral sequence, this implies that the induced maps
i
Hcts (G, M • ) −→ Hcts
i
(G, N • )

are isomorphisms.

Hence we can conclude the following.

Proposition 2.2.11. The functor
•
Ccts (G, −) : Ch+ (C) −→ Ch+ (M odΛ )

preserves homotopy, exact sequences and quasi-isomorphisms, hence induces the following exact
derived functors
RΓcts (G, −) : Db (CΛ,G ) −→ D+ (M odΛ )
RΓcts (G, −) : D+ (DΛ,G ) −→ D+ (M odΛ ).
Proof. This proposition follows from what we have done so far. The only subtlety lies in the fact
that CΛ,G does not necessarily have enough injectives and therefore we do not know if D+ (CΛ,G )
exists. However, we know that CΛ,G has enough projectives. Therefore, D− (CΛ,G ) exists, and
we may apply Lemma 2.2.10 to Db (CΛ,G ).

We now like to extend Proposition 2.2.8 to the case of complexes. Before that, we first prove
a lemma which will be required in our discussion.
Lemma 2.2.12. Let f : M −→ N be a morphism of objects in CΛ,G . Then there exists a directed
indexing set I with the following properties:
(1) There exist a fundamental system {Ui } (resp., {Vi }) of neighborhoods of zero consisting
of open Λ[G]-submodules of M (resp., N ).
(2) For each i ∈ I, there is a Λ[G]-homomorphism fi : M/Ui −→ N/Vi which fits into the
following commutative diagram
f
M // N

 fi 
M/Ui // N/Vi

15
where the vertical morphisms are the canonical quotient map.
(3) One has f = lim fi .
←−
i

Proof. Let {Uα }α∈IM (resp., {Vβ }β∈IN ) be a system of neighborhoods of zero consisting of open
Λ[G]-submodules of M (resp., N ). Then we set I = IM ×IN , Uα,β = Uα ∩f −1 (Vβ ) and Vα,β = Vβ .
It is then straightforward to verify that f factors through M/Uα,β to give a Λ[G]-homomorphism
fα,β : M/Uα,β −→ N/Vα,β and f = lim fα,β .
←−
α,β

In view of the above lemma, we say that a morphism f : M −→ N in CΛ,G is compatible

with a directed indexing set I if the conclusion in the lemma holds. By the lemma, we have
that for every morphism f : M −→ N in CΛ,G , there exists a directed indexing set I such that f
is compatible with I. In particular, if M is a bounded complex in CΛ,G , we can find a directed
indexing set I such that the differentials are compatible with I.
Proposition 2.2.13. Suppose that G has the property that Hcts m (G, N ) is finite for all finite

discrete Λ-modules N with a continuous commuting G-action and for all m ≥ 0. Let M • =
lim Mi• be a bounded complex of objects in CΛ,G with I-compatible differentials. Then we have
←−
i∈I
the following isomorphism
n
Hcts (G, M • ) ∼
= lim H n (G, Mi• )
←− cts
i
of hypercohomology groups for each n.
Proof. The canonical chain map M • −→ Mi• induces the following morphism of (convergent)
spectral sequences
r

r+s
Hcts G, H s (M • ) =⇒ Hcts (G, M • )


r G, H s (M • ) =⇒ H r+s (G, M • )
Hcts i cts i

which is compatible with i. By the hypothesis, the bottom spectral sequence is a spectral
sequence of finite Λ-modules. Therefore, the inverse limit is compatible with the inverse system
of the spectral sequences, and we have the following morphism
r

r+s
Hcts G, H s (M • ) =⇒ Hcts (G, M • )

r

 r+s
lim Hcts G, H s (Mi• ) =⇒ lim Hcts (G, Mi• )
←− ←−
i i

of (convergent) spectral sequences. By Proposition 2.2.8, we have the isomorphisms

r
Hcts G, H s (M • ) ∼= lim r
Hcts G, H s (Mi• ) .
←−
i

16
Hence, by the convergence of the spectral sequences, we obtain the required isomorphism.

For ease of notation, we will drop the ‘•’ for complexes. We also drop the notation ‘cts’.
Therefore, we write C(G, M ) as the complex of continuous cochains and RΓ(G, M ) for its
derived functor. Its ith cohomology group is then written as H i (G, M ).

2.3 Total cup products

We first recall the definition for topological G-modules (in other words, abelian Hausdorff topo-
logical groups with a continuous G-action).

Definition 2.3.1. (Cup products) Let A, B and C be topological G-modules. Suppose

h , i : A × B −→ C

is a continuous map satisfying σha, bi = hσa, σbi for a ∈ A, b ∈ B and σ ∈ G. Then we define
the cup product on the cochain groups

C i (G, A) × C j (G, B) −→ C i+j (G, C)

as follows: for α ∈ C i (G, A), β ∈ C j (G, B) and σ1 , ..., σi+j ∈ G, we have

D E
(α ∪ β)(σ1 , ..., σi+j ) = α(σ1 , ..., σi ), σ1 · · · σi β(σi+1 , ..., σi+j ) .

The cup product satisfies the following relation

δC (α ∪ β) = (δA α) ∪ β + (−1)i α ∪ (δB β)

and induces a pairing

H i (G, A) × H j (G, B) −→ H i+j (G, C)
on the cohomology groups.

Now fix a prime p. For the remainder of the paper, we shall assume that our profinite ring
Λ is pro-p. In other words, for each A ∈ I, the ring Λ/A is finite of a power of p. Let M and N
be objects in CΛ,G and DΛo ,G respectively, and let A be a topological G-module. Suppose there
is a continuous pairing
h , i : N × M −→ A
such that
(1) σhy, xi = hσy, σxi for x ∈ M, y ∈ N and σ ∈ G, and
(2) hyλ, xi = hy, λxi for x ∈ M, y ∈ N and λ ∈ Λ.

17
 As before, condition (1) will give rise to the cup product

C i (G, N ) × C j (G, M ) −→ C i+j (G, A),

which is Λ-balanced by condition (2). The cup product induces a group homomorphism

C i (G, N ) ⊗Λ C j (G, M ) −→ C i+j (G, A)

which gives rise to the following morphism

C(G, N ) ⊗Λ C(G, M ) −→ C(G, A)

of complexes of abelian groups. Taking the adjoint, we have a morphism

C(G, M ) −→ HomZp C(G, N ), C(G, A)

of complexes of Λ-modules.

Lemma 2.3.2. Suppose we are given another continuous pairing

( , ) : N ′ × M ′ −→ A

such that (1) σ(y ′ , x′ ) = (σy ′ , σx′ ) for x′ ∈ M ′ , y ′ ∈ N ′ and σ ∈ G;

(2) (y ′ λ, x′ ) = (y ′ , λx′ ) for x′ ∈ M ′ , y ′ ∈ N ′ and λ ∈ Λ, and
(3) there are morphisms f : N ′ −→ N in DΛo ,G and g : M −→ M ′ in CΛ,G such that the
following diagram
id⊗g
N ′ ⊗Λ M // N ′ ⊗ M ′
Λ

f ⊗id ( , )
 h , i 
N ⊗Λ M // A

commutes. Then we have the following commutative diagram

C(G, M ) // HomZ C(G, N ), C(G, A)
p

g∗ f∗
 

C(G, M ′ ) // HomZ C(G, N ′ ), C(G, A)
p

of complexes of Λ-modules.

Proof. It follows from a direct calculation that following diagram

id⊗g
C(G, N ′ ) ⊗Λ C(G, M ) // C(G, N ′ ) ⊗Λ C(G, M ′ )

f ⊗id ∪( , )
 ∪h , i

C(G, N ) ⊗Λ C(G, M ) // C(G, A)

18
is commutative, where ∪( , ) and ∪h , i are the cup products induced by the pairings ( , ) and
h , i respectively. By taking the adjoint and another straightforward calculation, we have the
commutative diagram in the lemma.

Now let M and N be bounded complexes of objects in CΛ,G and DΛo ,G respectively, and let
A be a bounded complex of topological G-modules. Suppose there is a collection of continuous
pairings
h , ia,b : N a × M b −→ Aa+b
where each pairing satisfies conditions (1) and (2), and the following hold:


(a) hdaN y, xia+1,b = da+b
A hy, xia,b for y ∈ N a and x ∈ M b , and


(b) (−1)a hy, dbM xia,b+1 = da+b
A hy, xia,b for y ∈ N a and x ∈ M b .
For each pair (a, b), we have a morphism

∪ab i a j b
ij : C (G, N ) ⊗Λ C (G, M ) −→ C
i+j
(G, Aa+b )

of abelian groups induced by the cup product. Then the total cup product

∪ : C(G, N ) ⊗Λ C(G, M ) −→ C(G, A)

is a morphism of complexes of Zp -modules given by the collection ∪ = (−1)ib ∪ab ij . The
definition given for the total cup products follows that in [Ne, 3.4.5.2]. We also have an analogous
result to Lemma 2.3.2 for complexes.

Lemma 2.3.3. Suppose we are given another collection of continuous pairings

h , ia,b : N ′a × M ′b −→ A′a+b

as above. Then we have the following commutative diagram

C(G, M ) // HomZ C(G, N ), C(G, A)
p

g∗ f∗
 

C(G, M ′ ) // HomZ C(G, N ′ ), C(G, A)
p

of complexes of Λ-modules.

2.4 Tate cohomology groups

We shall now describe the Tate cochain complexes of a finite group G. We begin by giving an
alternative description of the (inhomogeneous) cochain complexes. Throughout this subsection,

19
G will always denote a finite group. Consider the standard Z[G]-resolution (in inhomogenous
form) of Z (cf. [Wei, Sect. 6.5])1 ,

X0 ←− X1 ←− X2 ←− · · ·

where X0 = Z[G] and, for n ≥ 1, Xn is the free Z[G]-module generated by the set of all symbols
(g1 , ..., gn ) with gi ∈ G, and the differentials are given by the formula
n−1
X
∂ n (g1 , ..., gn ) = g1 (g2 , ..., gn ) + (−1)j (g1 , ..., gj gj+1 , ..., gn ) + (−1)n (g1 , ..., gn−1 ).
j=1

For any Z[G]-module M , there is a natural isomorphism C i (G, M ) −→ HomZ[G] (Xi , M ) which
is compatible with the differentials, thus giving an identification of complexes. Furthermore, if
M is a Λ[G]-module, the above identification is an isomorphism of complexes of Λ-modules.
We now construct the complete cochain groups. For a Z[G]-module A, we write A∗ =
HomZ (A, Z). Note that this is a Z[G]-module in a natural way. Applying HomZ (−, Z) to the
long exact sequence
0 ←− Z ←− X0 ←− X1 ←− X2 ←− · · · ,
we obtain the following long exact sequence

0 −→ Z −→ X0∗ −→ X1∗ −→ X2∗ −→ · · · ,

since each Xi is a free Z-module. Splicing the two long exact sequence and applying HomZ[G] (−, M )
to the resulting long exact sequence, we obtain the following complex

· · · −→ HomZ[G] (X1∗ , M ) −→ HomZ[G] (X0∗ , M ) −→ HomZ[G] (X0 , M ) −→ · · ·

The completed cochain complex Ĉ i (G, M ) are defined by

(
C i (G, M ) ∼
= HomZ[G] (Xi , M ) if i ≥ 0,
Ĉ i (G, M ) =
∗
HomZ[G] (X−1−i , M) if i ≤ −1.

Following [Ne, 5.7.2], we may extend the above definition to a complex M • of G-modules by
setting M
Ĉ n (G, M • ) = Ĉ i (G, M j )
i+j=n

with differential defined using the sign conventions of the previous sections. As before, for ease
of notation, we will drop the ‘•’ for complexes. The usual cup product for Tate cohomology
groups (cf. [NSW, Prop. 1.4.6]) extends to a total cup product with the same sign convention
as in the preceding section.
1
Weibel called this the unnormalized bar resolution.

20
3 Duality over pro-p rings
Let p be a fixed prime. Throughout the section, our profinite ring Λ will always be pro-p. In this
section, we will formulate and prove Tate’s (and Poitou’s) local and global duality theorems.

3.1 Tate’s local duality

Let F be a nonarchimedean local field with characteristic not equal to p. Fix a separable closure
F sep of F . Set GF = Gal(F sep /F ).
Lemma 3.1.1. We have
(
Qp /Zp if j = 2,
H j (GF , Qp /Zp (1)) ∼
=
0 if j > 2.

Proof. For j > 2, the conclusion follows from the fact that GF has p-cohomological dimension 2
(see [NSW, Thm. 7.1.8(i)]). By [NSW, Thm. 7.1.8(ii)], we have H 2 (Gv , Z/pr (1)) ∼
= Z/pr . The
assertion now follows by taking direct limits.
i
By the preceding lemma, we have a quasi-isomorphism Qp /Zp [−2] −→ τ≥2 C(GF , Qp /Zp (1))
of complexes of Zp -modules. Since Qp /Zp is an injective Zp -module, the map i has a homotopy
inverse. We shall fix one such map

r : τ≥2 C(GF , Qp /Zp (1)) −→ Qp /Zp [−2].

This gives a morphism

r
θ : C(GF , Qp /Zp (1)) −→ τ≥2 C(GF , Qp /Zp (1)) −→ Qp /Zp [−2]

of complexes of Zp -modules.
Let M be a bounded complex of objects in CΛ,GF . The obvious pairing

M ∨ (1) ⊗Λ M −→ Qp /Zp

induces the total cup product

C GF , M ∨ (1) ⊗Λ C(GF , M ) −→ C GF , Qp /Zp (1) .

Suppose that N is another bounded complex of objects in CΛ,GF , and there is a morphism
f : M −→ N of complexes in CΛ,GF . Then we have the following commutative diagram
id⊗f
N ∨ (1) ⊗Λ M // N ∨ (1) ⊗Λ N

f ∨ ⊗id
 
M ∨ (1) ⊗Λ M // Qp /Zp (1)

21
with the obvious pairings. Applying cochains and θ, we obtain the following commutative
diagram

id⊗f

C GF , N ∨ (1) ⊗Λ C(GF , M ) // C GF , N ∨ (1) ⊗Λ C(GF , N )

f ∨ ⊗id

 
C GF , M ∨ (1) ⊗Λ C(GF , M ) // Qp /Zp [−2]

which induces the following commutative diagram

αM


C(GF , M ) // HomZ C GF , M ∨ (1) , Qp /Zp [−2]
p

 αN
 


C(GF , N ) // HomZ C GF , N ∨ (1) , Qp /Zp [−2]
p

of complexes of Λ-modules by Lemma 2.3.2. We are now able to prove the following formulation
of Tate’s local duality.

Theorem 3.1.2. Let M be a bounded complex of objects in CΛ,GF . Then we have the following
isomorphism  


RΓ(GF , M ) −→ RHomZp RΓ GF , M ∨ (1) , Qp /Zp [−2]

in D(M odΛ ).

Proof. We shall show that αM (in the above diagram) is a quasi-isomorphism. Now if A −→
B −→ C −→ A[1] is an exact triangle 2 in Db (CΛ,GF ), we then have a morphism

αA


RΓ(GF , A) // RHomZ RΓ GF , A∨ (1) , Qp /Zp [−2]
p

 αB
 


RΓ(GF , B) // RHomZp RΓ GF , B ∨ (1) , Qp /Zp [−2]

  


αC
RΓ(GF , C) // RHomZp RΓ GF , C ∨ (1) , Qp /Zp [−2]

of exact triangles. Therefore, if any two of the morphisms αA , αB and αC are isomorphisms, so
is the third. For a bounded complex M in CΛ,GF , we have the following exact triangle

σ≤i−1 M −→ σ≤i M −→ M i [−i].

2
We write an exact triangle A → B → C → A[1] more compactly as A → B → C throughout.

22
Therefore, by induction, we are reduced to showing that αM is a quasi-isomorphism in the case
when M is a single module. Write M = lim Mβ , where each Mβ is a finite module. By the
←−
β
functoriality of α, we have the following commutative diagram
αM 

C(GF , M ) // Hom ∨
Zp C GF , M (1) , Qp /Zp [−2]

u v

lim αMβ  


 ←− ∨
lim C(GF , Mβ ) // lim Hom
Zp C GF , M β (1) , Q p /Zp [−2]
←− ←−
β β

of complexes of Λ-modules. By Proposition 2.2.7 and Proposition 2.2.8, we have that u and
v in the above diagram are isomorphisms of complexes, and the vertical maps in the following
commutative diagram
(αM )∗ 

H i (GF , M ) // Hom
Zp H 2−i G , M ∨ (1) , Q /Z
F p p

u∗ v∗

lim(αMβ )∗  


←−

i 2−i
lim H (GF , Mβ ) // lim Hom
Zp H GF , Mβ∨ (1) , Qp /Zp
←− ←−
β β

are isomorphisms. Since each (αMβ )∗ is an isomorphism by Tate local duality [NSW, Thm.
7.2.6], we have the required conclusion.

3.2 Global duality over pro-p rings

Let F be a global field with characteristic not equal to p, and let S be a finite set of primes of
F containing all primes above p and all archimedean primes of F (if F is a number field). Let
Sf (resp., SR ) denote the collection of non-archimedean primes (resp., real primes) of F in S.
Fix a separable closure F sep of F . Set GF,S = Gal(FS /F ), where FS is the maximal
subextension of F sep /F unramified outside S. For each v ∈ Sf , we fix a separable closure
Fvsep of Fv and an embedding F sep ֒→ Fvsep . This induces a continuous group homomorphism
Gv := Gal(Fvsep /Fv ) → GF,S . If v is a real prime, we also write Gv for Gal(C/R).
If M is a complex in CΛ,GF,S (resp., DΛ,GF,S ), then we can view M as a complex in CΛ,Gv
(resp., DΛ,Gv ) via the continuous homomorphism Gv −→ GF,S . Therefore, the cochain complexes
C(GF,S , M ) and C(Gv , M ) can be defined. Recall that for v ∈ Sf , we have the restriction map

resv : C(GF,S , M ) −→ C(Gv , M )

induced by the group homomorphism Gv −→ GF,S . For a real prime v, we have the following

resv : C(GF,S , M ) −→ C(Gv , M ) ֒→ Ĉ(Gv , M ).

23
To shorten notation in what follows, for v ∈ SR , we will abuse notation and use C(Gv , M ),
H i (Gv , M ), and RΓ(Gv , M ) to denote the Tate cochains Ĉ(Gv , M ), its cohomology groups, and
its derived object. We now make the following definition.

Definition 3.2.1. Let M be a complex in CΛ,GF,S or DΛ,GF,S . The complex of continuous

cochains of M with compact support is defined as
!
resS M
Cc (GF,S , M ) = Cone C(GF,S , M ) −→ C(Gv , M ) [−1] ,
v∈S

where the elements of

!
M
Cci (GF,S , M ) = C i (GF,S , M ) ⊕ C i−1 (Gv , M )
v∈S

have the form (a, aS ) with a ∈ C i (GF,S , M ), aS = (av )v∈S , av ∈ C i−1 (Gv , M ), and the differential
is given by
d(a, aS ) = (da, −resS (a) − daS ).
The ith cohomology group of Cc (GF,S , M ) is denoted by Hci (GF,S , M ).

Remark. If F is a function field in one variable over a finite field or F is a totally imaginary
number field, then SR is empty, and the cone is given by
 
resSf M
Cone C(GF,S , M ) −→ C(Gv , M ) [−1].
v∈Sf

Now suppose that p is odd and F is a number field with at least one real prime. Let v ∈ SR .
Then Ĥ i (Gv , M ) = 0 for every M in CΛ,GF,S (resp., DΛ,GF,S ) and for all i, since Gv is a finite
group of order 2 and M is a inverse limit of finite p-groups (resp., direct limit of finite p-groups).
Therefore, it follows that the canonical map
 
resSf M
Cone C(GF,S , M ) −→ C(Gv , M ) [−1] −→ Cc (GF,S , M )
v∈Sf

is a quasi-isomorphism. Therefore, one may take the above cone as a definition of the complex
of continuous cochains with compact support in this case.

Proposition 3.2.2. The functor

Cc (GF,S , −) : Ch+ (CΛ,GF,S ) −→ Ch(M odΛ )

resp., Cc (GF,S , −) : Ch+ (DΛ,GF,S ) −→ Ch(M odΛ )

24
preserves homotopy, exact sequences and quasi-isomorphisms, hence induces the following exact
derived functors
RΓc (GF,S , −) : Db (CΛ,GF,S ) −→ D(M odΛ )


resp., RΓc (GF,S , −) : D+ (DΛ,GF,S ) −→ D(M odΛ )
such that for M in Db (CΛ,GF,S ) or D+ (DΛ,GF,S ), we have the following exact triangle
M
RΓc (GF,S , M ) −→ RΓ(GF,S , M ) −→ RΓ(Gv , M )
v∈Sf

in D(M odΛ ) and the following long exact sequence

M
· · · −→ Hci (GF,S , M ) −→ H i (GF,S , M ) −→ H i (Gv , M ) −→ Hci+1 (GF,S , M ) −→ · · · .
v∈S

Proof. This is immediate from the definition of the cone.

By [NSW, Thm. 7.1.8(iii), Thm. 8.3.19], Proposition 2.2.8 can be applied to GF,S and Gv ,
where v ∈ Sf . For v ∈ SR , Gv is a finite group of order 2, and so the finiteness hypothesis in
Proposition 2.2.8 is satisfied, so the conclusion also holds in this case. The following analogous
statement to Proposition 2.2.8 for cohomology groups with compact support will now follow from
the definition of the cone and the long exact sequence of cohomology groups in the preceding
proposition.

Proposition 3.2.3. The functor Cc (GF,S , −) preserves direct limits in DΛ,GF,S . Moreover,
if M = lim Mα is an object in CΛ,GF,S , where each Mα is finite, then we have the following
←−
α
isomorphism
Cc (GF,S , M ) ∼
= lim C (G , M )
←− c F,S α
α
of complexes and isomorphisms

Hci (GF,S , M ) ∼
= lim H i (G , M )
←− c F,S α
α

of cohomology groups.

Lemma 3.2.4. We have

(
Qp /Zp if j = 3,
Hcj (GF,S , Qp /Zp (1)) ∼
=
0 if j > 3.

Proof. By the long exact sequence of Poitou-Tate [NSW, 8.6.13], we have the following exact
sequence M
H 2 (GF,S , Z/pn Z(1)) −→ H 2 (Gv , Z/pn Z(1)) −→ Z/pn Z −→ 0
v∈S

25
and an isomorphism
res
M
H 3 (GF,S , Z/pn Z(1)) −→ Ĥ 3 (Gv , Z/pn Z(1)).
v∈SR

By the definition of continuous cochains with compact support and the fact that cdp (Gv ) = 2 for
v ∈ Sf , we have Hc3 (GF,S , Z/pn Z(1)) ∼
= Z/pn Z. The remainder of the lemma will then follow
from a similar argument to that in Lemma 3.1.1.

Let M be a bounded complex in CΛ,GF,S . For each v ∈ S, we define a morphism ∪v of

complex of Zp -modules to be




C Gv , M ∨ (1) ⊗Λ C Gv , M ) −→ C Gv , Qp /Zp (1) −→ Cc GF,S , Qp /Zp (1) [1],

where the first map is the total cup product, and the second is the natural morphism arising
i
from the definition of the cone. By Lemma 3.2.4, we have a quasi-isomorphism Qp /Zp [−3] −→
τ≥3 Cc (GF,S , Qp /Zp (1)) of complexes of Zp -modules. Since Qp /Zp is an injective Zp -module, the
map i has a homotopy inverse. We shall fix one such map

r : τ≥3 Cc (GF,S , Qp /Zp (1)) −→ Qp /Zp [−3],

and this induces the following morphism

r
ϑ : Cc (GF,S , Qp /Zp (1)) −→ τ≥3 Cc (GF,S , Qp /Zp (1)) −→ Qp /Zp [−3]

of complexes of Zp -modules. Combining this with ∪v , we obtain a morphism

∪v
ϑ[1]
C Gv , M ∨ (1) ⊗Λ C Gv , M ) −→ Cc GF,S , Qp /Zp (1) [1] −→ Qp /Zp [−2]

of complexes of Zp -modules. For v ∈ Sf , this is essentially the morphism constructed in Section

3.1, which will give the Tate local duality as in Theorem 3.1.2. We also have the following.

Theorem 3.2.5. Let p = 2, and let v ∈ SR . For a bounded complex M of objects in CΛ,Gv , we
have the following isomorphism


d v , M ) −→ RHomZp RΓ
RΓ(G d Gv , M ∨ (1) , Q2 /Z2 [−2]

in D(M odΛ ).

Proof. By a similar argument to that in Theorem 3.1.2, it suffices to consider a finite module
M , and the conclusion then follows from [NSW, Thm. 7.2.17].

26
 L
For ease of notation, we shall write P (GF,S , −) for v∈S C(Gv , −). We now define a mor-
phism ∪S by

ν
M
P

P GF,S , M ∨ (1) ⊗Λ P (GF,S , M ) −→ Cc GF,S , Qp /Zp (1) −→ Cc GF,S , Qp /Zp (1) ,
v∈S

where ν(aS , bS ) = (av ∪v bv )v∈S .

We now construct the total cup products for the compactly supported cochain groups. Since
these are defined as cones, it follows from Lemma 1.3 that there are two morphisms
∪0 , ∪1 : Cc (GF,S , M ∨ (1)) ⊗Λ Cc (GF,S , M ) −→ Cc (GF,S , Qp /Zp (1))
of complexes of Zp -modules given by
(a, aS ) ∪0 (b, bS ) = (a ∪ b, (−1)ā resS (a) ∪S bS )
(a, aS ) ∪1 (b, bS ) = (a ∪ b, aS ∪S resS (b))
where ∪ is the total cup product
C(GF,S , M ∨ (1)) ⊗Λ C(GF,S , M ) −→ C(GF,S , Qp /Zp (1)).
The morphisms ∪0 and ∪1 induce the following morphisms

∪c : C(GF,S , M ∨ (1)) ⊗Λ Cc (GF,S , M ) −→ Cc (GF,S , Qp /Zp (1))

∨
c ∪ : Cc (GF,S , M (1)) ⊗Λ C(GF,S , M ) −→ Cc (GF,S , Qp /Zp (1))
of complexes of abelian groups which are given by the following respective formulas (see also
[Ne, 5.3.3.2, 5.3.3.3])
a ∪c (b, bS ) = (a ∪ b, (−1)ā resS (a) ∪S bS )
(a, aS ) c ∪ b = (a ∪ b, aS ∪S resS (b)).
All of these fit into the following diagram
Cc (GF,S , M ∨ (1)) ⊗Λ Cc (GF,S , M ) // C(GF,S , M ∨ (1)) ⊗Λ Cc (GF,S , M )

∪c
 
c∪
Cc (GF,S , M ∨ (1)) ⊗Λ C(GF,S , M ) // C(GF,S , Qp /Zp (1))

which is commutative up to homotopy by Lemma 1.3. Also, the following diagrams

C(GF,S , M ∨ (1)) ⊗Λ P (Gv , M )[−1] // P G , M ∨ (1) ⊗ P (G , M )[−1]
F,S Λ F,S

t


 
∨
P GF,S , M (1) ⊗Λ P (GF,S , M ) [−1]

∪S [−1]
 ∪c


C(GF,S , M ∨ (1)) ⊗Λ Cc (GF,S , M ) // C G , Q /Z (1)
c F,S p p

27
 P (GF,S , M ∨ (1))[−1] ⊗Λ C(GF,S , M ) // P (G , M ∨ (1))[−1] ⊗ P (G , M )
F,S Λ F,S

t′


P (GF,S , M ∨ (1)) ⊗Λ P (GF,S , M ) [−1]
∪S [−1]
 c∪


Cc (GF,S , M ∨ (1)) ⊗Λ C(GF,S , M ) // C G , Q /Z (1)
c F,S p p

are commutative, where t and t′ are the morphisms defined as in Lemma 1.1. These in turn
induce the following morphism of exact triangles in K(M odΛ ).

M 

P (GF,S , M )[−1] // HomZp C(Gv , M ∨ (1)), Cc GF,S , Qp /Zp (1) [−3]
v∈S

  


Cc (GF,S , M ) // Hom C(G , M ∨ (1)), C G , Q /Z (1) [−3]
Zp F,S c F,S p p

  


C(GF,S , M ) // Hom ∨
Zp Cc (GF,S , M (1)), Cc GF,S , Qp /Zp (1) [−3]

Combining this with the morphism

r
ϑ : Cc (GF,S , Qp /Zp (1)) −→ τ≥3 Cc (GF,S , Qp /Zp (1)) −→ Qp /Zp [−3],

we obtain the following morphism of exact triangles

M M

RΓ(Gv , M )[−1] // RHomZp RΓ(Gv , M ∨ (1)), Qp /Zp [−3]
v∈S v∈S

 
∨


RΓc (GF,S , M ) // RHom
Zp RΓ(GF,S , M (1)), Qp /Zp [−3]

 
∨


RΓ(GF,S , M ) // RHom
Zp RΓc (GF,S , M (1)), Qp /Zp [−3]

in D(M odΛ ).

Theorem 3.2.6. For any bounded complex M in CΛ,GF,S , the above morphism of exact triangles
is an isomorphism.

28
Proof : The top morphism is an isomorphism by Theorem 3.1.2. It remains to show that the
middle morphism is an isomorphism. By a similar argument (using Proposition 3.2.3 for the
limiting argument for the compactly supported cohomology) to that of Theorem 3.1.2, we can
reduce to the case that M is a single finite module. The conclusion then follows from the usual
Poitou-Tate duality (cf. [NSW, 8.6.13]). ✷

Remark. Theorem 3.1.2 and Theorem 3.2.6 are stated in [FK] for the case that Λ is a profinite
ring with a basis of neighborhoods consisting of powers of the Jacobson radical of Λ, and M is
a finitely generated projective Λ-module.

4 Iwasawa modules
In this section, we will introduce certain modules over an Iwasawa algebra. The next two
paragraphs will introduce some notations which will be adhered to throughout this section.
Fix a prime p. Let R be a commutative pro-p ring with a directed fundamental system I of
neighborhoods of zero consisting of open ideals. Let G and Γ be two profinite groups such that
there is continuous homomorphism π : G −→ Γ of profinite groups. Set Λ = RJΓK. We now
describe the natural profinite topology on Λ (see [RZ, Sect. 5.3]). Let U be the collection of
open normal subgroups of Γ, and consider the following family of two-sided ideals:

aΛ + I(U ), a ∈ I, U ∈ U .

Here I(U ) denotes the kernel of the map Λ ։ R[Γ/U ]. We take these ideals as a fundamental
system of neighborhoods of zero.
We have a map ι : Λ → Λ which sends γ to γ −1 . Note that this is only a homomorphism of
R-modules. It is a ring homomorphism if and only if Γ is abelian. Denote by
π
ρ = ρΓ : G −→ Γ ⊆ Λ×

the tautological one-dimensional representation of G over Λ.

Remark. In most situations, the ring R is usually a commutative complete Noetherian lo-
cal ring with finite residue field of characteristic p, and the group Γ is a compact p-adic Lie
group. However, despite motivated by the above situation, we shall consider the theory in more
generality.

4.1 Induced modules

For a given Λ-module M , we define a Λo -module M ι by the formula m ·ι λ := ι(λ)m for λ ∈
Λ, m ∈ M . Similarly, if N is a Λo -module, we define a Λ-module, which is also denoted as N ι ,
by λ ·ι m := mι(λ).
We have the following lemma.

29
Lemma 4.1.1. (a) If M is a Λ[G]-module, then M ι is a Λo [G]-module.
(b) If M is a Λ[G]-Λ-module (not necessarily balanced), then M ι is a Λo [G]-Λ-module (not
necessarily balanced).

Proof. (a) Let g ∈ G, λ ∈ Λ and m ∈ M ι . Then we have

(gm) ·ι λ = ι(λ)gm = g(ι(λ)m) = g(m ·ι λ).

(b) Similar argument as above.

For a given U ∈ U and a given R[G]-module M , we define two Λ[G]-Λ-modules as follows:

UM = HomR (R[Γ/U ], M )
MU = R[Γ/U ]ι ⊗R M,

where G acts on R[Γ/U ] via ρΓ/U and Λ acts on R[Γ/U ] via the canonical projection Λ ։ R[Γ/U ].
Note that the Λ[G]-Λ-modules defined above are balanced as Λ-Λ-modules. They are balanced
as Λ[G]-Λ-modules if Γ/U is abelian.
Let V ∈ U with U ⊆ V . Then there is a canonical surjection pr : R[Γ/U ] ։ R[Γ/V ] and a
map Tr : R[Γ/V ] −→ R[Γ/U ] given by
X
gU 7→ gvU.
v∈V /U

These in turn induce the following maps.

pr∗ : V M −→ U M
pr∗ : MU −→ MV
Tr∗ : U M −→ V M
Tr∗ : MV −→ MU

Denote by δβ : G/U → Z the Kronecker delta-function

(
1 if β = β ′ ,
δβ (β ′ ) ∼
=
0 if β 6= β ′ .

The next two lemmas then follow from a straightforward calculation.

Lemma 4.1.2. We have the following isomorphism of R[G]-modules

∼
MU −→ U M
X X
β ⊗ xβ 7→ xβ δβ
β∈G/U β∈G/U

30
which is functorial in M . Moreover, if V is another open normal subgroup of G such that U ⊆ V ,
then the isomorphism fits into the following commutative diagrams.
∼ ∼
MU / UM MV /VM

pr∗ Tr∗ Tr∗ pr∗

 ∼
  ∼

MV / V M MU / UM

Lemma 4.1.3. We have the following equalities of Λo [G]-modules:

(U M )ι = HomR (R[Γ/U ]ι , M ),
(MU )ι = R[Γ/U ] ⊗R M.

Let M be a R[G]-module. We define two Λ[G]-Λ-modules as follows:

FΓ (M ) = lim U M,
−→
U ∈U
FΓ (M ) = lim MU ,
←−
U ∈U

where the transition maps are induced by the surjections R[Γ/U ] ։ R[Γ/V ] for U ⊆ V . Note
that the Λ[G]-Λ-modules defined above are balanced as Λ-Λ-modules. They are balanced as
Λ[G]-Λ-modules if and only if Γ is abelian. One easily sees from Lemma 3.1.3 that

FΓ (M )ι = lim HomR (R[Γ/U ]ι , M ) and

−→
U ∈U
FΓ (M )ι = lim (R[Γ/U ] ⊗R M ).
←−
U ∈U

We also have the following description of FΓ (A), when A is an object of DR,G .

Lemma 4.1.4. If A is an object of DR,G , then FΓ (A) is an object of DΛ,G and

FΓ (A) ∼
= HomR,cts (Λ, A).

Similarly, we have
FΓ (A)ι ∼
= HomR,cts (Λι , A).
If {Aα } is a direct system of objects in DR,G , then we have isomorphisms


FΓ (A) ∼
= lim FΓ (Aα ) resp., FΓ (A)ι ∼
= lim FΓ (Aα )ι
−→α
−→α

in DΛ,G (resp., in DΛo ,G ).

31
Proof. By Lemma 2.1.4(3), for each U ∈ U , we have

HomR (R[Γ/U ], A) = HomR,cts (R[Γ/U ], A).

Therefore, the lemma will now follow from [RZ, Prop. 5.1.4]

We would like to have a description of FΓ (T ), when T is an object in CR,G . Before we can

do this, we shall recall the notion of a complete tensor product from [RZ]. Let M be an object
in CΛ,G , and let N be an object in CR,G . The completed tensor product of M and N is taken to
be
M⊗ ˆ R N = lim M/U ⊗R N/V,
←−
U,V

where U (resp., V ) runs through the open Λ[G]-submodules of M (resp., open R[G]-submodules
of N ).

Lemma 4.1.5. Let M be an object in CΛ,G and N be an object in CR,G . Then the completed
tensor product M ⊗ ˆ R N is an object of CΛ,G , and coincides with the usual tensor product if N
is a finitely generated R-module. Moreover, as a functor, the completed tensor product is right
exact (in both variables) and preserves inverse limits.

Proof. It follows from [Wil, Lemma 7.7.2] that M ⊗ˆ R N is a compact Λ-module. By a similar
argument to that used in the proof of that lemma, we have that the G-action is continuous.

We are now in position to describe FΓ (T ).

Lemma 4.1.6. If T is an object of CR,G , then FΓ (T ) is isomorphic to Λι ⊗ ˆ R T and

ι
FΓ (T ) is isomorphic to Λ⊗ ˆ R T . If {Tα } is an inverse system of objects in CR,G such that
∼
T = lim Tα , then we have isomorphisms
←−
α


FΓ (T ) ∼
= lim FΓ (Tα ) resp., FΓ (T )ι ∼
= lim FΓ (Tα )ι
← −
α
← −
α

in CΛ,G (resp., in CΛo ,G ).

Proof. We have



ˆ RT ∼
FΓ (T ) = lim(R[Γ/U ]ι ⊗R T ) = lim R[Γ/U ]ι ⊗ = lim R[Γ/U ]ι ⊗ˆ RT ∼
= Λι ⊗
ˆ R T.
←− ←− ←−
U U U

Suppose T ∼
= lim T in CR,G . Then
←− α
α

FΓ (T ) ∼ ˆ RT ∼
= Λι ⊗ = lim ˆ R Tα ∼
Λι ⊗ = lim F (T ).
←− ←− Γ α
α α

32
 As a conclusion to the subsection, we record the following duality relation between the
modules we have defined.
Proposition 4.1.7. Let T be an object in CR,G . Then we have isomorphisms
 
FΓ (T )∨ ∼ F
= Γ (T ∨ ι
) resp., (FΓ (T )ι ∨ ∼
) F
= Γ (T ∨
)

in DΛo ,G (resp., in DΛ,G ).

Proof. We will prove the first isomorphism, the second will follow from a similar argument. This
follows by the following calculations:

FΓ (T )∨ ∼
= HomZp (Λι ⊗ ˆ R T, Qp /Zp ) (by Lemma 4.1.6)


∼
= HomR,cts Λι , HomZp ,cts (T, Qp /Zp ) (by [RZ, Prop. 5.5.4(c)])
∼
= FΓ (T ∨ )ι (by Lemma 4.1.4).

4.2 Shapiro’s lemma

As before, R denotes a commutative pro-p ring. Let G be a profinite group. Fix a closed normal
subgroup H of G and write Γ = G/H. Let π : G −→ Γ be the canonical quotient map. We
identify U as the collection of open normal subgroups of G containing H. Therefore, in this
context, for each U ∈ U , and an R[G]-module M , we have

UM = HomR (R[G/U ], M ),
MU = R[G/U ]ι ⊗R M.
We will apply Shapiro’s lemma to see that the direct limits and inverse limits of cohomology
groups over every intermediate field Fα can be viewed as cohomology groups of certain Λ-
modules. The results in this section can be found in [Ne, 8.2.2, 8.3.3-5, 8.4.4.2].

Lemma 4.2.1. Let U be an open normal subgroup of G, and let N be a bounded below complex
of objects of DR,G . Then we have a quasi-isomorphism
∼
C(G, U N ) −→ C(U, N )

of complexes of Λ-modules.
Proof. We first prove the lemma in the case that N is an object of DR,G . Then we may write
N = lim Nα , where Nα is a finite object of DR,G . The usual Shapiro’s lemma holds for such
−→
α
modules. Also, we note that U N ∼
= lim (N ). Hence, we have
−→ U α
α
 
sh
C(G, U N ) = C G, lim U (Nα ) ∼= lim C G, U (Nα ) −→ lim C(U, Nα ) = C(U, N )
−→ −→ −→
α α α

33
which gives the required conclusion for the case that N is an object of DR,G . For the case that
N is a bounded below complex of objects of DR,G , one can prove this by a spectral sequence
argument as used in Lemma 2.2.10.

Recall that if A is a complex in DR,G , then FΓ (A) = lim UA is a complex in DΛ,G by Lemma
−→
U ∈U
4.1.4. We then have the following proposition.

Proposition 4.2.2. Let A be a bounded below complex of objects of DR,G . Then the composite
morphism
∼ sh res
C(G, FΓ (A)) −→ lim C(G, U A) −→ lim C(U, A) −→ C(H, A)
−→ −→
U ∈U U ∈U

is a quasi-isomorphism of complexes of Λ-modules. In other words, we have an isomorphism

∼
RΓ G, FΓ (A) −→ RΓ(H, A)

in D(M odΛ ).

The next result will give a Shapiro-type relation for cohomology groups of objects (and
complexes of objects) in CR,G .

Lemma 4.2.3. Let U be an open normal subgroup of G. Then for any bounded complex M in
CR,G , we have a quasi-isomorphism
∼
C(G, MU ) −→ C(U, M )

of complexes of Λ-modules.

Proof. By the same argument as that in Lemma 4.2.1, it suffices to consider the case when M
is an object of CR,G . Then we have M = lim Mα , where Mα is a finite object in CR,G . Note that
←−
α
∼
MU = lim(Mα )U . Then we have morphisms
←−
α

sh
C(G, MU ) ∼
= lim C G, (Mα )U −→ lim C(U, Mα ) ∼
= C(U, M )
← −
α
←−
α

which induce a morphism

lim(i) H j G, (Mα )U ⇒ H i+j (G, MU )
← −
α


lim(i) H j U, Mα ⇒ H i+j (U, M )
← −
α

34
of convergent spectral sequences. Since Mα is finite, the usual Shapiro’s lemma implies that



H j G, (Mα )U ∼ = H j U, Mα

is an isomorphism. This in turn implies that

lim(i) H j G, (Mα )U ∼= lim(i) H j U, Mα .
← −
α
← −
α

By the convergence of the spectral sequences, we have isomorphisms

H n (G, MU ) ∼
= H n (U, M ),

as required.

Since inverse limits are not necessarily exact, we cannot always view inverse limits of coho-
mology groups over every intermediate field Fα as cohomology groups of certain Λ-modules in
general. However, we can say something if we impose an extra assumption on G.

Proposition 4.2.4. Let M be a bounded complex of objects in CR,G . Then we have the following
isomorphism
∼
C(G, FΓ (M )) −→ lim C(G, MU )
←−
U ∈U

of complexes of Λ-modules. Furthermore, if H m (G, N )

is finite for all finite discrete Λ-modules
N with a Λ-linear continuous G-action and all m ≥ 0, then we have

H j (G, FΓ (M )) ∼
= lim H j (U, M ).
←−
U ∈U

Proof. As before, it suffices to consider the case when M is an object of CR,G . Write M = lim Mα ,
←−
α
where each Mα is a finite object in CR,G . By Lemma 4.1.6, we have a continuous isomorphism

FΓ (M ) ∼
= lim F (M ) ∼ lim(M ) .
←− Γ α = ←− α U
α α,U

The second assertion now follows from Proposition 2.2.8 and Lemma 4.2.3.

4.3 Iwasawa setting

We now apply the discussion in Subsection 4.2 to the arithmetic situation. Let F∞ be a Galois
extension of F which is contained in FS . Write H = Gal(FS /F∞ ), and write Γ = Gal(F∞ /F ).
Let U denote the collection of open normal subgroups of GF,S containing H. For each U ∈ U ,
we let FU = (FS )U , and define SU to be the set of primes in FU above S. As before, we write
Λ = RJΓK, where R is a commutative pro-p ring. The following lemma is immediate from the
discussion in the preceding subsection.

35
Lemma 4.3.1. Let T be a bounded complex of objects in CR,GF,S . Then we have the following
isomorphisms

H j (GF,S , FΓ (T )) ∼
= lim H j (GF,S , TU ) ∼
= lim H j (GFU ,SU , T ),
←− ←−
U U
H j (GF,S , FΓ (T ∨ )) ∼
= lim H j (GF,S , U T ∨ ) ∼
= lim H j (GFU ,SU , T ∨ ) ∼
= H j (Gal(FS /F∞ ), T ∨ ).
−→ −→
U U

Let v ∈ Sf . Fix an embedding F sep ֒→ Fvsep , which induces a continuous group monomor-
phism
α = αv : Gv ֒→ GF ,
where GF = Gal(F sep /F ). Let X be a finite discrete R[GF ]-module. For a finite Galois extension
F ′ of F , write U = Gal(F sep /F ′ ) and XU = R[GF /U ] ⊗R X. The embedding F sep ֒→ Fvsep
determines a prime v ′ of F ′ above v such that Fv′ ′ is a finite Galois extension of Fv and Gv′ :=
Gal(Fvsep /Fv′ ′ ) = α−1 (U ).
Fix coset representatives σi ∈ GF of
[
GF /U = σi α(Gv /Gv′ ).
i

Then the set of distinct primes in F ′ above v is given by the (finite) collection {σi (v ′ )}. Then
by [Ne, 8.1.7.6, 8.5.3.1], we have a quasi-isomorphism
∼
M
C(Gv , XU ) −→ C(Gσi (v′ ) , X)
i

and isomorphisms M
H n (Gv , XU ) ∼
= H n (Gσi (v′ ) , X)
i

of cohomology groups for n ≥ 0.

Now suppose that p = 2 and F is a number field with at least one real prime. Let v ∈ SR .
Let F ′ be a finite Galois extension of F , and retain the above notations. Then the primes in F ′
above v are either all real or all complex. We first consider the case when all the primes above
F ′ are all real. Then Gv acts trivially on XU , and we have

C(Gv , XU ) = C(Gv , X)|GF /U |

L ′
which is precisely τ C(Gτ (v′ ) , X), where v is a prime above v and τ runs through a set of
coset representatives for GF /U . If all the primes above F ′ are complex, it remains to show
that Ĥ i (Gv , XU ) = 0 for all i. Since Gv is cyclic (of order 2), we are reduced to showing
this for i = 1, 2, which follows from Shapiro’s Lemma in the usual sense (since these are usual
cohomology groups)

36
 We shall apply the above discussion to finite discrete R[GF,S ]-modules, which we view as
R[GF ]-modules via the canonical quotient map GF ։ GF,S . By the compatibility of limits
and the groups of continuous cochains, we can apply the above results to objects in DR,GF,S or
N−fte
CR,G F,S
.

Lemma 4.3.2. Let T be a bounded complex of objects in CR,GF,S . Then for v ∈ Sf , we have the
following isomorphisms
M
H j (Gv , FΓ (T )) ∼
= lim H j (Gv , TU ) ∼
= lim H j (Gw , T ),
←− ←−
U U w|v
M
j ∨ ∼ j ∨ ∼
H (Gv , FΓ (T )) = lim H (Gv , U T ) = lim H j (Gw , T ∨ ).
−→ −→
U U w|v

The same conclusion holds for the case when p = 2 and v ∈ SR , if we replace the cohomology
groups by the completed cohomology groups as defined in Subsection 2.4.

We would like to derive an analogue of Shapiro’s lemma for compactly supported cohomology.
Let F ′ be a finite Galois extension of F which is contained in FS . Denote the set of primes of
F ′ above S by S ′ . Let X be a discrete R[GF,S ]-module. We write U = Gal(FS /F ′ ) and
XU = R[GF,S /U ] ⊗R X. By the discussion in the previous subsection and the above, we have
the following diagram

L ∼ L L
C(GF,S , XU ) //
v∈S C(Gv , XU ) // v∈S v′ |v C(Gv , R[Gv /Gv′ ] ⊗R X)

sh sh
 L 
C(GF ′ ,S ′ , X) //
v′ ∈S ′ C(Gv′ , X)

which commutes up to homotopy. By a similar argument to that in [Ne, 8.1.7.2.1, 8.5.3.2], this
in turn induces a quasi-isomorphism (functorial in X)

shc : Cc (GF,S , XU ) −→ Cc (GF ′ ,S ′ , X)

which fits into the following commutative (up to homotopy) diagram with exact rows.
L
0 //
v∈S C(Gv , XU )[−1] // Cc (GF,S , XU ) // C(GF,S , XU ) // 0

≀
L L 
0 // v∈S v′ |v C(Gv , R[Gv /Gv′ ] ⊗R X)[−1] // Cc (GF,S , XU ) // C(GF,S , XU ) // 0

sh[−1] shc sh
L   
0 //
v′ ∈S ′ C(Gv , X)[−1] // Cc (GF ′ ,S ′ , X) // C(GF ′ ,S ′ , X) // 0

37
 Suppose that F ′′ ⊆ FS is another finite Galois extension of F containing F ′ , and write S ′′
for the set of primes of F ′′ above S and V = Gal(FS /F ′′ ). Again by similar arguments to that
in [Ne, 8.5.3.4], we have the following morphisms

resc : Cc (GF ′ ,S ′ , X) −→ Cc (GF ′′ ,S ′′ , X)

corc : Cc (GF ′′ ,S ′′ , X) −→ Cc (GF ′ ,S ′ , X),

which are functorial in X and fit in the following diagrams, which are commutative up to
homotopy:

0 // ⊕v′ ∈S ′ C(Gv′ , X)[−1] // Cc (GF ′ ,S ′ , X) // C(GF ′ ,S ′ , X) // 0

res[−1] resc res

  
0 // ⊕v′′ ∈S ′′ C(Gv′′ , X)[−1] // Cc (GF ′′ ,S ′′ , X) // C(GF ′′ ,S ′′ , X) // 0

0 // ⊕v′′ ∈S ′′ C(Gv′′ , X)[−1] // Cc (GF ′′ ,S ′′ , X) // C(GF ′′ ,S ′′ , X) // 0

cor[−1] corc cor

  
0 // ⊕v′ ∈S ′ C(Gv′ , X)[−1] // Cc (GF ′ ,S ′ , X) // C(GF ′ ,S ′ , X) // 0

shc // Cc (GF ′ ,S ′ , X) shc // Cc (GF ′′ ,S ′′ , X)

Cc (GF,S , XU ) Cc (GF,S , XV )
Tr∗ resc pr∗ corc
 shc
  shc

Cc (GF,S , XV ) // Cc (GF ′′ ,S ′′ , X) Cc (GF,S , XU ) // Cc (GF ′ ,S ′ , X)

Since all the morphisms constructed above are functorial, they can be extended to complexes.
Hence, we may conclude the following.

Proposition 4.3.3. (a) For a bounded below complex A of objects in DR,GF,S , the canonical
morphism of complexes
∼
Cc (GF,S , FΓ (A)) −→ lim Cc (GF,S , U A)
−→
U,Tr

is an isomorphism.
(b) Let T be a bounded complex of objects in CR,GF,S . Then the canonical morphism of
complexes
∼
Cc (GF,S , FΓ (T )) −→ lim Cc (GF,S , TU )
←−
U,pr

is an isomorphism and induces isomorphisms

Hcj (GF,S , FΓ (T )) ∼
= lim H j (G , T ) ∼ lim H j (G ,T)
←− c F,S U = ←− c FU ,SU
U,pr U,corc

of cohomology groups for j ≥ 0.

38
4.4 Duality over extensions of a global/local field
We retain the notations introduced in the previous section. Let F∞ be a Galois extension of F
which is contained in FS . Write H = Gal(FS /F∞ ), and write Γ = Gal(F∞ /F ). As before, we
write Λ = RJΓK, where R is a commutative pro-p ring.
Applying Theorem 3.2.6 and Proposition 4.1.7, we obtain the following theorem. In the
d v , M ) for v ∈ SR .
theorem, we abuse notation and use RΓ(Gv , M ) to denote RΓ(G

Theorem 4.4.1. Then, for a bounded complex T in CR,GF,S , we have the following isomorphism
of exact triangles
M ∼ M

RΓ(Gv , FΓ (T ))[−1] // RHomZp RΓ(Gv , FΓ (T ∨ )ι (1)), Qp /Zp [−3]
v∈S v∈S

 ∼

∨ ι


RΓc (GF,S , FΓ (T )) // RHom
Zp RΓ(GF,S , FΓ (T ) (1)), Qp /Zp [−3]

 ∼

∨ ι


RΓ(GF,S , FΓ (T )) // RHom
Zp RΓc (GF,S , FΓ (T ) (1)), Qp /Zp [−3]

in D(M odΛ ).

We end by saying something about the situation over nonarchimedean local fields. Let F be
a nonarchimedean local field of characteristic not equal to p. Let F∞ be a Galois extension of
F with Galois group Γ. Write GE = Gal(F sep /E) for every Galois extension E/F . Recall that
by [NSW, Thm. 7.1.8(i)], we have cdp (GF ) = 2.
Let T be a bounded complex of objects in CR,GF . By Proposition 4.2.2 and Proposition
4.2.4, we have
∼
C GF , FΓ (T ∨ ) −→ lim C(GFα , T ∨ )

−→
H i GF , FΓ (T ∨ ) ∼= lim H i (GFα , T ∨ ) ∼
= H i (GF∞ , T ∨ )
−→
∼
C GF , FΓ (T ) −→ lim C(GFα , T )

←− i
H i GF , FΓ (T ) ∼ = lim H (GFα , T ),
←−
where Fα runs through all finite Galois extension of F∞ /F . Applying Theorem 3.1.2 and Propo-
sition 4.1.7, we obtain the following.

Theorem 4.4.2. Let T be a bounded complex of objects in CR,GF . Then we have the following
isomorphism


RΓ(GF , FΓ (T )) −→ RHomZp RΓ GF , FΓ (T ∨ )ι (1) , Qp /Zp [−2]

in D(M odΛ ).

39
References
[CFKSV] J. Coates, T. Fukaya, K. Kato, R. Sujatha and O. Venjakob, The GL2 main con-
jecture for elliptic curves without complex multiplication, Publ. Math. IHES, 101
(2005), 163-208.

[F] J. Flood, Pontryagin duality for topological modules, Proc. Amer. Math. Soc. vol.
75 (1979), No. 2, 329-333.

[FK] T. Fukaya and K. Kato, A formulation of conjectures on p-adic zeta functions in

noncommutative Iwasawa theory, Amer. Math. Soc. Transl. (2) vol. 219, (2006),
1-85.

[Hart] R. Hartshorne, Residues and Duality, Lect. Notes in Math. 20, Springer-Verlag,
Berlin, (1966).

[Jen] C. U. Jensen, Les Foncteurs Dérivés de lim et leurs Applications en Théorie des
←−
Modules, Lect. Notes in Math. 254, Springer-Verlag, Berlin, (1972).

[Kak] M. Kakde, The main conjecture of Iwasawa theory for totally real fields,
arXiv:1008.0142v1 [math.NT].

[Lam] T. Y. Lam, Lectures on Modules and Rings, Grad. Texts in Math. 189, Springer-
Verlag, New York, (1999).

[Lim] M. F. Lim, Duality over p-adic Lie extensions of global fields, Ph.D. thesis, McMaster
University, (2010).

[Ne] J. Nekovár̂, Selmer complexes, Astérisque 310 (2006).

[NSW] J. Neukirch, A. Schmidt and K. Wingberg, Cohomology of Number Fields,

Grundlehren der mathematischen Wissenschaften 323, Springer (2000).

[RW] J. Ritter and A. Weiss, Towards equivariant Iwasawa theory II, Indag. Mathem.,
N.S., 15(4) (2004), 549-572.

[RZ] L. Ribes and P. Zalesskii, Profinite Groups, Ergeb. Math. Grenzgeb. 3, Springer-
Verlag, Berlin, (2000).

[Wei] C. A. Weibel, An Introduction to Homological Algebra, Reprinted, Cambridge Stud.

Adv. Math. 38, Cambridge Univ. Press, Cambridge, UK, (1997).

[Wil] J. Wilson, Profinite Groups, London Mathematical Society Monographs New Series,
vol. 19, Oxford University Press, (1998).

40