Journal of Number Theory 130 (2010) 1914–1924

Contents lists available at ScienceDirect

Journal of Number Theory

www.elsevier.com/locate/jnt

The algebraic functional equation of Selmer groups

for CM fields ✩
Ming-Lun Hsieh
Department of Mathematics, National Taiwan University, Taipei, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: Because the analytic functional equation holds for Katz p-adic
Received 31 August 2009 L-function for CM fields, the algebraic functional equation of the
Revised 6 February 2010 Selmer groups for CM fields is expected to hold. In this note
Available online 26 March 2010
we prove it following the specialization principle developed by
Communicated by David Goss
T. Ochiai (2005) in [Och05].
MSC: © 2010 Elsevier Inc. All rights reserved.
11R23

Keywords:
Selmer groups
Iwasawa main conjectures

1. Introduction

The aim of this note is to prove the algebraic functional equation of the Selmer groups for CM
fields which is predicted by the analytic functional equation for Katz p-adic L-function and Iwasawa
main conjecture for CM fields. The idea is to use the specialization principle developed by T. Ochiai
in [Och05].
Let us briefly recall Iwasawa main conjecture for CM fields. Let F be a totally real subfield of
degree d over Q and K be a totally imaginary quadratic extension of F and let c be the com-
plex conjugation, the unique nontrivial element in Gal(K/F ). Let p be an odd rational prime. The
main conjecture for CM fields states equality of two ideals generated by p-adic L-functions and the
characteristic power series of Selmer groups respectively. To introduce them, we make the ordinary
assumption (Ord) as follows:

every prime of F above p splits in K. (Ord)

✩
This work is partially supported by National Science Council grant 98-2115-M-002-017-MY2.
E-mail address: [email protected].

0022-314X/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jnt.2010.02.003
 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924 1915

Let S p be the set of places of K above p. Then (Ord) is equivalent to the existence of a p-adic CM
type Σ which is a subset in S p such that Σ and its complex conjugation Σ c form a partition of S p .
Namely

Σ ∩ Σ c = ∅, Σ  Σc = S p.

By definition Σ c is also a p-adic CM type.

Let K∞ + be the cyclotomic Z -extension and K− be the anticyclotomic Zd -extensions of K respec-
p ∞ p
tively. Let K∞ = K∞ + K− . If one assumes Leopoldt’s conjecture, then K
∞ ∞ would be the composition
of all Z p -extensions of K. Let Γ := Gal(K∞ /K) be a free Z p -module of rank 1 + d. Let K be a finite
abelian extension of K which is linearly disjoint from K∞ and  = Gal(K /K). Let K∞  = K K
∞ and
 ∼ ×
G = Gal(K∞ /K) =  × Γ . We let ψ : G → C p be a continuous p-adic character. Let D w be the de-
composition group of a place w. We further assume that ψ| D w for all w | p are locally algebraic. Let
Oψ = Z p [Im ψ] be the ring of values of ψ and let O be a complete discrete valuation ring which is
finite flat over Oψ .

1.1. p-adic L-functions and Selmer groups for CM fields

We shall formulate the main conjecture for CM fields from the p-adic Galois representation point
of view [Gre94]. We begin with some notation. Let G K = Gal(K/K) be the absolute Galois group of K
and Λ = O JΓ K. Then Λ is an Iwasawa algebra in d + 1-variables. Define Ψ : G K → Λ× the universal
Λ-adic character associated to ψ by

Ψ : G K −→ Λ× ,
g −→ ψ( g ) g |K∞ .

On the analytic side, one has the p-adic L-function for CM fields, L p (Ψ, Σ) ∈ Λ, which is con-
structed by Katz [Kat78] if the conductor of ψ divides p ∞ and by Hida and Tilouine [HT93] in general.
Roughly L p (Ψ, Σ) interpolates Hecke L-values for K p-adically. Moreover L p (Ψ, Σ) satisfies a func-
tional equation (cf. Theorem 2 [HT93]).
On the algebraic side, one has the Selmer group for CM fields. We recall its definition after in-
troducing some notation. For M a locally compact topological abelian group, we denote by M ∗ the
Pontryagin dual of M. Then Λ∗ = Homcont (Λ, Q p /Z p ) has a natural discrete Λ-module structure given
by a · f (x) := f (xa), a, x ∈ Λ. Let w be a place of K. We write I w for the inertia group at w. Denote by
F w a geometric Frobenius in D w . Let c be the prime-to-p conductor of ψ and S ψ be the set of finite
places dividing c. Let S ⊃ S ψ be a finite set of prime-to-p places of K and let K S be the maximal
S ∪ S p -ramified algebraic extension of K. Define the Selmer group SelK (Ψ, Σ) associated to (Ψ, Σ) by

  
1
 ∗
 1
 ∗

SelK (Ψ, Σ) = ker H K S /K , Ψ ⊗ Λ → H Iw,Ψ ⊗ Λ . (1.1)
w ∈ S Σ c

It follows from Theorem 1.2.2 [HT94] that SelK (Ψ, Σ) is a cofinitely generated and cotorsion dis-
crete Λ-module. Let F (Ψ, Σ) denote the characteristic power series of SelK (Ψ, Σ) which is unique up
to Λ-units (see Definition 2.1(1)). Let ht1 (Λ) be the set of height one primes of Λ. For P ∈ ht1 (Λ),
we let ord P be the valuation at P . Then Iwasawa main conjecture for CM fields is stated as follows.

Conjecture 1 (Iwasawa main conjecture for CM fields). For every P ∈ ht1 (Λ), we have

   
ord P L p (Ψ, Σ) = ord P F (Ψ, Σ) .
1916 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924

We also define the non-primitive p-adic L-function L Sp (Ψ, Σ) by

  
L pS (Ψ, Σ) = L p (Ψ, Σ) · 1 − Ψ (F w ) .
w∈S\Sψ

Similarly, the non-primitive Selmer group is defined by

  
   
SelK (Ψ, Σ) = ker H K S /K, Ψ ⊗ Λ∗ →
S 1
H 1
Iw,Ψ ⊗ Λ ∗
. (1.2)
w ∈ S ψ Σ c

S
Let F S (Ψ, Σ) be the characteristic power series of SelK (Ψ, Σ). We also consider the dual version of
the main conjecture for CM fields which has the advantage of including non-primitive p-adic L-func-
tions and Selmer groups. In the case of main conjecture for totally real fields, such a dual version is
proposed by R. Greenberg in [Gre77].
Let ε be the p-adic cyclotomic character of G K and ω : G K → μ p −1 be the p-adic Teichmüller
character. Define Ψ D the Cartier dual character of Ψ by Ψ D = Ψ −1 ε . Then the dual version of the
main conjecture for CM fields is stated as follows.

Conjecture 2. For every P ∈ ht1 (Λ), we have

    
ord P L pS (Ψ, Σ) = ord P F S Ψ D , Σ c .

1.2. Main result

Our main result is as follows.

Theorem 1.1 (Algebraic functional equation). For any P ∈ ht1 (Λ), we have

    
ord P F (Ψ, Σ) = ord P F Ψ D , Σ c .

Remark 1.

(1) Theorem 1.1 is an immediate consequence of the main conjecture for CM fields (Conjecture 1)
combined with the functional equation of Katz p-adic L-functions.
(2) The general functional equation of the Selmer groups associated to the cyclotomic deformation of
p-adic Galois representations is proved by R. Greenberg in [Gre89].

This theorem has the following corollary.

Corollary 1.2. If ψ| = 1, Conjecture 1 is equivalent to Conjecture 2.

2. The proof

2.1. Notation and definitions

We first prepare some notation and definitions.

Definition 2.1. Let R be a compact normal Noetherian domain and ht1 ( R ) be the set of height one
primes of R. For P ∈ ht1 ( R ), let R P be the localization of R at P . Let S be a cofinitely generated
R-module and let S ∗ be the Pontryagin dual of S .
 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924 1917

(1) If S is cotorsion, define the characteristic ideal char R S by


P (S )
char R S = P ,
P ∈ht1 ( R )

where P (S ) = length R P (S ∗ ⊗ R R P ). The characteristic power series is a generator of char R S .

(2) S is said to be pseudo-null if S ∗ is a pseudo-null R-module.
(3) If S is a finite Z p -module, we put

p (S ) = lengthZ p (S ).

(4) Denote by Snull the maximal pseudo-null R-module quotient of S .

(5) For S and S  two discrete cofinitely generated R-modules, we say S ∼ S  if there exists a R-
module morphism S → S  such that the kernel and cokernel are pseudo-null.

The following observation is useful.

Lemma 2.2. Suppose that R is UFD and S is cotorsion. Let f ∈ R be prime to char R S . Then

∼
S ⊗ R R /( f ) → Snull ⊗ R R /( f ).

In particular if R ∼
= O J T K, p (S ⊗ R R /( f )) is uniformly bounded for all f ∈ R prime to char R S .

Proof. The Pontryagin dual S ∗ of S is a finitely generated torsion R-module. By the structure theorem
for finitely generated torsion modules over a normal domain [Bou65, §4.4, Théorème 5, p. 253], there
exist R-module morphisms

K → S∗ → E  C ,
∼ 
where K and C are pseudo-nulland E → i R /( g i ). Note that by definition the Pontryagin dual
K ∗ of K is Snull and char R S = ( i g i ). Thus if f is prime to char R S , then K [ f ] = S ∗ [ f ] and hence
∼
S ⊗ R R /( f ) → Snull ⊗ R R /( f ). 2

Definition 2.3. Put G S = Gal(K S /K) and G = G S ψ for brevity. Let A be a discrete G S -module and L w
be a Λ-submodule in H 1 ( D w , A ) for each w ∈ S ∪ S p . Then H L
1
( A ) the Selmer group associated to
the local condition L = {L w } w ∈ S ∪ S p is defined by
  
1 1 H 1(D w , A)
H L ( A ) = ker H (G S , A ) → .
Lw
w∈S∪S p

Define the local condition L (Σ) = {L (Σ) w } w ∈ S ∪ S p by

⎧
⎨ H 1 ( D w , A ), w ∈ Σ,
L (Σ) w = 0, w ∈ Σc, (2.1)
⎩ 1
H ( D w / I w , A I w ), w ∈ S .

We define Selstr
K (Ψ, Σ) the strict Selmer group by

∗
 
Selstr 1
K (Ψ, Σ) := H L (Σ) Ψ ⊗ Λ .

This definition is independent of the choice of S ⊃ S ψ .

1918 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924

If A is a finite Z p -module, we set h1L ( A ) = 1

p ( H L ( A )),

   
hi ( A ) = p H i (G , A ) i
and hΣ ( A) = p H 1(D w , A)
w ∈Σ

for i = 0, 1, 2. We write χ (G , A ) (resp. χ ( D w , A )) for the global (resp. local) Euler-characteristics.

Lemma 2.4. Let A = Ψ ⊗ R ∗ for R a finite quotient of Λ and let A D = Ψ D ⊗ R ∗ be the Cartier dual of A. We
have the following long exact sequence:

 H 1 (D w , A)
1 1 D ∗
0 HL ( A) H 1 (G , A ) Lw HL ⊥(A )
w∈Sψ ∪S p


H 2(D w , A)
0 H 0 (G , A D )∗ w∈Sψ ∪S p
H 2 (G , A )


where L ⊥ is the orthogonal complement of L in w∈S p ∪Sψ H 1 ( G , A D ).

Proof. This lemma follows from Poitou–Tate duality (cf. Theorem 4.50(4) [Hid00]). 2

Definition 2.5. Let P ∈ ht1 (Λ) and m be a positive integer. Let R be a quotient ring of Λ with π R :
Λ → R such that J R := π R ( P m ) = 0. The Pontryagin dual ( R / J R )∗ of R / J R is a discrete G-module
on which G acts via π R ◦ Ψ . Define (Ψ1 , Σ 1 ) = (Ψ, Σ c ) and (Ψ2 , Σ 2 ) = (Ψ D , Σ). For • = 1, 2, we put
A • = Ψ• ⊗ ( R / J R )∗ and define

   
S • ( R ) = ker H 1 (G , A • ) → H 1(I w , A•)D w × H 1(D w , A•) ,
w∈Sψ w ∈Σ •

T 1 ( R ) = H 0 (G , A 2 ) × H 0 ( D w , A 1 ),
w ∈Σ 1

T 2 ( R ) = H 0 (G , A 1 ) × H 0 ( D w , A 2 ),
w ∈Σ 2

H 1P ,m ( R ) = H 1(I w , A1)D w ,
w∈Sψ

H 2P ,m ( R ) = H 0 ( D w , A 2 ),
w∈Sψ

M P ,m ( R ) = S • ( R ) ⊕ T • ( R ).
•

By definition, S 1 (Λ) = Selstr 2 str

K (Ψ, Σ) and S (Λ) = SelK (Ψ , Σ ).
D c

2.2. Proof of the theorem

Notations are as in the previous subsection. We begin with the following key proposition.
 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924 1919

Proposition 2.6. Suppose

(1) R ∼= OJ X 1 , . . . , Xn K for some n  1.

(2) H 1P ,m ( R ) and H 2P ,m ( R ) are pseudo-null R-modules.

Then we have

char R M 1P ,m ( R ) = char R M 2P ,m ( R ).

Proof. We will use the specialization principle developed by T. Ochiai in [Och05] and then proceed
by induction on n. To simplify the notation we suppress the subscript and write M • ( R ) (resp. H • ( R ))
for M •P ,m ( R ) (resp. H •P ,m ( R )), • = 1, 2. We first assume n = 1 and R ∼
= OJ T K. By Theorem 2.3(1) and
(2) [Och08], it suffices to show

   
p M 1 ( R )[ f ] − p M 2 ( R )[ f ] (2.2)

is uniformly bounded from above and below for all f ∈ R prime to J R . Set

 
H0 ( A • ) = H 0(I w , A•) × H 0 ( D w , A • ).
w∈Sψ w ∈Σ •

For f ∈ R prime to J R , we consider the following exact sequence:

0 0

γ1  
H 1 ( I w , A • )[ f ] × H 1 ( D w , A • )[ f ]
H 1 (G , A • )[ f ] w∈Sψ w ∈Σ •

γ2  
H 1 ( I w , A • [ f ]) × H 1 ( D w , A • [ f ])
H 1 (G , A • [ f ]) w∈Sψ w ∈Σ •

γ3
H 0 (G , A • ) ⊗ R /( f ) H0 ( A • ) ⊗ R /( f )

0 0

Then S • ( R )[ f ] = ker γ1 and S • ( R /( f )) = ker γ2 . By the snake lemma, we have

 
Ker γ3 → S • R /( f ) → S • ( R )[ f ] → Coker γ3 .

In addition, since H0 ( A • ) and H 0 (G , A • ) are annihilated by J R , f is prime to the characteristic ideal

of H 0 (G , A • ) and H0 ( A • ). By Lemma 2.2 we deduce that p (Ker γ3 ) and p (Coker γ3 ) are uniformly
bounded. Thus p (S • ( R /( f ))) − p (S • ( R )[ f ]) is uniformly bounded. On the other hand, it is clear
that T • ( R )[ f ] = T • ( R /( f )). Therefore the uniform boundedness of (2.2) is equivalent to showing
1920 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924

     
p M 1 R /( f ) − p M 2 R /( f ) (2.3)

is uniformly bounded for all f ∈ R prime to J R .

We put A = A 1 [ f ] = Ψ ⊗ ( R /( f ))∗ . Then R /( f ) = O J T K/( J R , f ) is a finite ring as f is prime
to J R . Then A is a finite Z p -module and the Cartier dual A D of A is A 2 [ f ]. Let L = L (Σ) be the
local condition defined in (2.1). Then it is well known that the orthogonal complement L ⊥ of L is
L (Σ c ). Note that h1L ( A ) = p (S 1 ( R /( f ))) and h1L ⊥ ( A D ) = p (S 2 ( R /( f ))). We also note that

H 1(D w , A)
= H 1(I w , A)D w , w ∈ Sψ (2.4)
Lw

because D w / I w ∼
= Ẑ has cohomological dimension one. Put
   
Df = p H 2(D w , A) − p H 1(I w , A)D w .
w∈Sψ

By Tate’s formula of local and global Euler characteristics (cf. Theorems 2.8 and 5.1 [Mil06]), we find
that χ (G , A ) = −[F : Q] p ( A ) and χ ( D w , A ) = −[K w : Q p ] p ( A ) for w ∈ S p (K is a CM field). It
follows from the ordinary assumption (Ord) that [F : Q] = w ∈Σ c [K w : Q p ]; hence

χ (G , A ) = χ ( D w , A ). (2.5)
w ∈Σ c

We also have the equality h2Σ ( A ) = h0Σ ( A D ) by Tate local duality (cf. Corollary 2.3 [Mil06]).
Now by Lemma 2.4, (2.4) and (2.5) we have

   
h1L ( A ) − h1L ⊥ A D = −χ (G , A ) + h0 ( A ) − h0 A D + χ ( D w , A ) − h0Σ c ( A ) + h2Σ ( A ) + D f
w ∈Σ c
0 0
 D
  
= h ( A) − h A − hΣ c ( A ) + hΣ A D + D f .
0 0

Then

     
h1L ( A ) + h0Σ c ( A ) + h0 A D = h1L ⊥ A D + h0Σ A D + h0 ( A ) + D f . (2.6)

It is clear that (2.6) implies p ( M 1 ( R /( f ))) − p ( M 2 ( R /( f ))) = D f .

Now we show D f is uniformly bounded. For w ∈ S ψ ,

        
p H 2(D w , A) = p H0 D w, AD = p H 0 ( D w , A 2 )[ f ]  p H 0(D w , A2) . (2.7)

Also from the exact sequence

 D w
0 → H 0 ( I w , A 1 ) ⊗ R /( f ) → H 1(I w , A)D w → H 1(I w , A1)D w [ f ]

and Lemma 2.2, we find that

     
p H 1(I w , A)D w  p H 0 ( D w , A 1 )null + p (I w , A1)D w . (2.8)

By the assumption, H 1 ( R ) and H 2 ( R ) are pseudo-null and hence finite. It follows from (2.7) and (2.8)
that
 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924 1921

     
− p H 1(R ) − p H 0 ( D w , A 1 )null  D f  p H 2(R) .
w∈Sψ

We conclude that D f is uniformly bounded. This completes the proof when n = 1.

If n  2, then by Lemma 2.5 [Och08] there exists a pseudo-null R-module N such that for any
(n) (n) (n) (n)
linear element l ∈ LO ( M 1 ( R )) ∩ LO ( M 2 ( R )) ∩ LO ( N ) ∩ LO ( H 1 ( R ) ⊕ H 2 ( R )) (for the definitions of
(n)
linear element and LO , see Definition 2.2 [Och08]) and l is prime to J R , the kernels and cokernels
of the natural R /(l)-module morphisms

 
H 1 G , A • [l] → H 1 (G , A • )[l],
 
H 1 D w , A • [l] → H 1 ( D w , A • )[l], w ∈ S p,
 
H I w , A • [l] → H 1 ( I w , A • )[l],
1
w ∈ Sψ

are pseudo-null R /(l)-modules, • = 1, 2. Therefore we can deduce that

 
char R /(l) M • R /(l) = char R /(l) M • ( R )[l].

Since R /(l) is a (n − 1)-variable Iwasawa algebra. By the choice of l, H 1 ( R /(l)) and H 2 ( R /(l)) are
also pseudo-null R /(l)-modules. Therefore the assertion follows from the induction hypothesis and
Proposition 3.6 [Och05]. 2

Lemma 2.7. Let w ∈

/ S p . We have

(1) H 1 ( D w / I w , Ψ ⊗ Λ∗ ) = 0 for w ∈
/ Sψ .
(2) H 1 ( I w , Ψ ⊗ Λ∗ ) = 0 for w ∈ S ψ .

Proof. We first prove (1). Since w ∈ / S ψ ∪ S p , Ψ is unramified at w, and Ψ w ( F w ) = 1 because K∞

contains the cyclotomic Z p -extension. Λ∗ is divisible, and hence

   
H 1 D w / I w , Ψ ⊗ Λ ∗ = Λ ∗ / Ψ ( F w ) − 1 Λ ∗ = 0.

Next we prove (2). Let I tw be the maximal tame pro-p quotient of I w and I  be the kernel of
the quotient map I w  I tw . As w is prime to p, it is well known that I tw is also the maximal pro-p
quotient of I w and I tw ∼
= Z p (1). Let γt be a generator. Because Ψ | I w = 1 for w ∈ S ψ and Λ∗ is divisible,
we have

    I    I     I 
H 1 I w , Ψ ⊗ Λ∗ = H 1 I tw , Ψ ⊗ Λ∗ = Λ∗ / Ψ (γt ) − 1 Λ∗ = 0. 2

Now we are ready to prove our main result.

Theorem 2.8. For every P ∈ ht1 (Λ), we have

 
charΛ SelK (Ψ, Σ) = charΛ SelK Ψ D , Σ c .

Proof. By definition, it is equivalent to

    
P SelK (Ψ, Σ) = P SelK Ψ D , Σ c
1922 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924

for every P ∈ ht1 (Λ). If d = 1, the theorem is a consequence of the Iwasawa main conjecture for
imaginary quadratic fields proved by K. Rubin [Rub91]. Now we assume d > 1 and S = S ψ . Put A =
Ψ ⊗ Λ∗ . By Lemma 2.7(2), we have the exact sequence
  
0 → Selstr
K (Ψ, Σ) → SelK (Ψ, Σ) → H 1 D w /I w , A I w . (2.9)
w ∈Σ c

For w ∈ S p , let J w be the ideal generated by Ψ ( D w ) − 1. Then J w has height greater than one if
d > 1 by Lemma 4.2 [HT91]. It follows that H 1 ( D w / I w , A I w ) is a pseudo-null Λ-module because it
is annihilated by J w . Therefore by (2.9) we conclude that

Selstr
K (Ψ, Σ) ∼ SelK (Ψ, Σ). (2.10)

Note that (2.10) implies Selstr

K (Ψ, Σ) is a cotorsion Λ-module.
Now let m be a positive integer and let S • and T • be as in Definition 2.5. For w ∈ S p , the ideal
J w also annihilates A D w , so A D w is a pseudo-null Λ-module. It follows that P ( T • (Λ)) = 0 and by
Lemma 2.7(2)

 •  m 
S • (Λ) ∼ Selstr
K Ψ ,Σ P , • = 1, 2.

Note that for any discrete cotorsion Λ-module N, P (N ) = P (N [ P

m
]) for m  0. We deduce that for
m0

         
P SelK Ψ • , Σ = P Selstr •
K Ψ ,Σ = P S • (Λ) = P M •P ,m (Λ) .

If w ∈ S ψ , Ψ ( F w ) − 1 = 0 is prime to p in Λ and the ideal generated by Ψ ( I w ) − 1 is (Ψ ( I w ) − 1) =

(ψ( I w ) − 1) = ( p r ) for some integer r. Then H 1 ( I w , A [ P m ]) D w and H 0 ( D w , A D [ P m ]) are annihilated
by ( p r ) and Ψ ( F w ) − 1. It follows that

    D w    
H 1P ,m (Λ) = H1 Iw, A Pm and H 2P ,m (Λ) = H0 D w, A Pm
w∈Sψ w∈Sψ

are pseudo-null Λ-modules. The theorem follows from Proposition 2.6 directly. 2

2.3. Proof of the corollary

Now we prove Corollary 1.2. By Theorem 1.1, it is equivalent to the following proposition.

Proposition 2.9. If ψ| = 1, we have

 S
      
P SelK Ψ D, Σc = P SelK Ψ D , Σ c + ord P 1 − Ψ (F w )
w∈S\Sψ

for every P ∈ ht1 (Λ).

Note that we have the following exact sequence by Lemma 2.7:

    γ   
S
0 → SelK Ψ D , Σ c → SelK Ψ D, Σc → H 1 D w , Ψ D ⊗ Λ∗ . (2.11)
w∈S\Sψ
 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924 1923

/ S ψ ∪ S p be a finite prime of K. Then for any P ∈ ht1 (Λ), we have

Lemma 2.10. Let w ∈

    
P H 1 D w , Ψ D ⊗ Λ∗ = ord P 1 − Ψ ( F w ) .

Therefore by (2.11) we have

 S
      
P SelK Ψ D, Σc  P SelK Ψ D , Σ c + ord P 1 − Ψ ( F w ) .
w∈S\Sψ

Proof. Let A = Ψ ⊗ Λ∗ and B = Ψ D ⊗ Λ∗ . As w ∈ / S p ∪ S ψ , I w acts on A and B trivially. For n  1,

we put An = A [ P n ] and Bn = B[ P n ]. Then An and Bn are cofinitely generated and cotorsion Λ-
modules. Because D w / I w is topologically generated by the Frobenius F w , we have H i ( D w / I w , Bn ) =
H i (Z p , Bn ) = 0 for i > 1. By Hochschild–Serre exact sequence, we have

0 −→ H 1 ( D w / I w , Bn ) −→ H 1 ( D w , Bn ) −→ H 1 ( I w , Bn ) D w / I w −→ 0. (2.12)

Recall that I tw is the maximal tame pro-p quotient of I w . As w is prime to p, we have the following
isomorphisms as D w / I w -modules:

   
H 1 ( I w , Bn ) = H 1 I tw , Bn = Hom Z p (1), Bn ∼
= An . (2.13)

In addition, H 1 ( D w / I w , Bn ) = Bn /( F w − 1)Bn , and hence

   
P H 1 ( D w / I w , Bn ) = P H 0 ( D w , Bn ) . (2.14)

Put (2.12), (2.13) and (2.14) together, we obtain that

     
P H 1 ( D w , Bn ) = P H 0 ( D w , Bn ) + P H 0 ( D w , An ) . (2.15)

By the equality P (H
0
( D w , B) ⊗ Λ/ P n ) = P (H
0
( D w , Bn )), (2.15) and the exact sequence
 
0 −→ H 0 ( D w , B) ⊗ Λ/ P n −→ H 1 ( D w , Bn ) −→ H 1 ( D w , B) P n −→ 0,

we deduce that for all n  1

       
P H 1 ( D w , B) P n = P H 0 ( D w , An ) = P H 0(D w , A ) P n . (2.16)

Since K∞ contains the cyclotomic Z p -extension, Ψ | D w = 1, ω . It follows that H 1 ( D w , B) and

H 0 ( D w , A ) are cotorsion. By (2.16), we have

      
P H 1 ( D w , B) = P H 1 ( D w , B) P ∞ = P H 0(D w , A ) .

By the above equality we can deduce the lemma from the following the equality of Λ-modules:

 ∗  
H 0 ( D w , A )∗ = H 0 D w , Ψ ⊗ Λ∗ = Λ/ Ψ ( F w ) − 1 Λ. 2

In virtue of Lemma 2.10 and the exact sequence (2.11), to prove Proposition 2.9, it suffices to
prove the cokernel of the map γ in (2.11) is a pseudo-null Λ-module. This follows from the following
stronger proposition due to [GV00].
1924 M.-L. Hsieh / Journal of Number Theory 130 (2010) 1914–1924

Proposition 2.11. If ψ| = 1, the restriction map

       
H 1 G S , Ψ D ⊗ Λ∗ → H 1 D w , Ψ D ⊗ Λ∗ × H 1 I w , Ψ D ⊗ Λ∗
w∈S w ∈Σ c

has finite cokernel.

Proof. Since Selstr ∗

K (Ψ , Σ ) is a cotorsion Λ-module and H (K∞ , Ψ ⊗Λ ) = 0 if ψ| = 1, we can pro-
D c 0

ceed the proof as in Proposition 2.1 [GV00] (because K∞ contains the cyclotomic Z p -extension). 2

Acknowledgments

The author would like to thank Prof. Ochiai for many useful suggestions. Especially the idea of
using the specialization technique is suggested by him. The author also would like to thank Prof. Hida
for useful e-mail correspondence. The author also thanks the referee for the careful reading and the
suggestion on the improvement of this paper.

References

[Bou65] N. Bourbaki, Éléments de mathématique. Fasc. XXXI. Algèbre Commutative. Chapitre 7: Diviseurs, Actualités Scien-
tifiques et Industrielles, vol. 1314, Hermann, Paris, 1965.
[Gre77] Ralph Greenberg, On p-adic L-functions and cyclotomic fields. II, Nagoya Math. J. 67 (1977) 139–158.
[Gre89] Ralph Greenberg, Iwasawa theory for p-adic representations, in: Algebraic Number Theory, in: Adv. Stud. Pure Math.,
vol. 17, Academic Press, Boston, MA, 1989, pp. 97–137.
[Gre94] Ralph Greenberg, Iwasawa theory and p-adic deformations of motives, in: Motives, Seattle, WA, 1991, in: Proc. Sympos.
Pure Math., vol. 55, Amer. Math. Soc., Providence, RI, 1994, pp. 193–223.
[GV00] Ralph Greenberg, Vinayak Vatsal, On the Iwasawa invariants of elliptic curves, Invent. Math. 142 (1) (2000) 17–63.
[Hid00] Haruzo Hida, Modular Forms and Galois Cohomology, Cambridge Stud. Adv. Math., vol. 69, Cambridge University Press,
Cambridge, 2000.
[HT91] H. Hida, J. Tilouine, Katz p-adic L-functions, congruence modules and deformation of Galois representations, in: L-
Functions and Arithmetic, Durham, 1989, in: London Math. Soc. Lecture Note Ser., vol. 153, Cambridge Univ. Press,
Cambridge, 1991, pp. 271–293.
[HT93] H. Hida, J. Tilouine, Anti-cyclotomic Katz p-adic L-functions and congruence modules, Ann. Sci. École Norm. Sup.
(4) 26 (2) (1993) 189–259.
[HT94] H. Hida, J. Tilouine, On the anticyclotomic main conjecture for CM fields, Invent. Math. 117 (1) (1994) 89–147.
[Kat78] Nicholas M. Katz, p-adic L-functions for CM fields, Invent. Math. 49 (3) (1978) 199–297.
[Mil06] J.S. Milne, Arithmetic Duality Theorems, second ed., BookSurge, LLC, Charleston, SC, 2006.
[Och05] Tadashi Ochiai, Euler system for Galois deformations, Ann. Inst. Fourier (Grenoble) 55 (1) (2005) 113–146.
[Och08] Tadashi Ochiai, The algebraic p-adic L-function and isogeny between families of Galois representations, J. Pure Appl.
Algebra 212 (6) (2008) 1381–1393.
[Rub91] Karl Rubin, The “main conjectures” of Iwasawa theory for imaginary quadratic fields, Invent. Math. 103 (1) (1991)
25–68.