This paper investigates the structure and properties of unramified Iwasawa modules over multiple zp-extensions of CM-fields, providing formulas for Galois coinvariants and criteria for cyclicity in special cases.
Contribution
It offers new descriptions of Galois coinvariants and conditions for cyclicity of Iwasawa modules in multiple zp-extensions, extending prior understanding of their structure.
Findings
01
Order of Galois coinvariants expressed via characteristic power series
02
Necessary and sufficient conditions for cyclicity of Iwasawa modules
03
Results depend on splitting behavior of primes and Iwasawa invariants
Abstract
For a CM-field K and an odd prime number p, let K′ be a certain multiple Zp-extension of K. In this paper, we study several basic properties of the unramified Iwasawa module XK′ of K′ as a Zp[[Gal(K′/K)]]-module. Our first main result is a description of the order of a Galois coinvariant of XK′ in terms of the characteristic power series of the unramified Iwasawa module of the cyclotomic Zp-extension of K under a certain assumption on the splitting of primes above p. Second one is that if K is an imaginary quadratic field and p does not split in K, we give a necessary and sufficient condition for which XK is Zp[[Gal(K/K)]]-cyclic under several assumptions on the Iwasawa λ-invariant and the ideal class group of…
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAlgebraic Geometry and Number Theory · Advanced Algebra and Geometry · Homotopy and Cohomology in Algebraic Topology
Full text
Galois Coinvariants of the Unramified Iwasawa Modules
For a CM-field K and an odd prime number p,
let K′ be a certain multiple Zp-extension of K.
In this paper, we study several basic properties of the unramified Iwasawa module XK′ of K′
as a Zp[[Gal(K′/K)]]-module.
Our first main result is a description of
the order of a Galois coinvariant of XK′ in terms of
the characteristic power series of the unramified Iwasawa module of the cyclotomic Zp-extension of K
under a certain assumption on the splitting of primes above p.
Second one is that if K is an imaginary quadratic field and p does not split in K,
we give a necessary and sufficient condition for which XK is
Zp[[Gal(K/K)]]-cyclic
under several assumptions on the Iwasawa λ-invariant and the ideal class group of K,
where K is the Zp2-extension of K.
1 Introduction
1.1 The unramified Iwasawa modules
Let p be an arbitrary prime number, K a finite extension of the rational number field Q and K∞/K the cyclotomic Zp-extension.
For an arbitrary algebraic number field F, we denote by XF the Galois group of the maximal unramified abelian p-extension over F.
The module XK∞ is called the unramified Iwasawa module of K∞.
In the case where K is totally real, little is known about the structure of XK∞, although the Iwasawa main conjecture gives us highly nontrivial
information about
the minus part of XK∞ in the case where K is a CM-field.
Greenberg conjectured in [9] that XK∞ would be finite if K is totally real, which is called Greenberg’s conjecture.
A lot of efforts by a number of mathematicians have revealed that this conjecture holds true in many cases,
but it still remains unsolved (in general).
We consider the maximal multiple Zp-extension K/K and its unramified Iwasawa module XK.
It is known that XK is a finitely generated torsion Zp[[Gal(K/K)]]-module (see [8]).
There is a conjecture that XK would be pseudo-null as a Zp[[Gal(K/K)]]-module, which is called Greenberg’s generalized conjecture
(“pseudo-null” is defined
in §3).
Concerning this conjecture and its application, there are many studies
(Bleher et al. [1], Fujii [7], Itoh [11], Ozaki [16], and Minardi [14], etc.).
However, even if Greenberg’s generalized conjecture is true, it just states that the characteristic ideal of XK is trivial,
so that it seems difficult to consider any
analogues of the Iwasawa invariants and the Iwasawa main conjecture for XK.
Therefore, it is worthwhile to study not only Greenberg’s generalized conjecture, but also various basic properties of XK,
for example, the number of generators as a Zp[[Gal(K/K)]]-module, its Galois (co)invariants, and cohomogical properties.
In the following, we always assume that p is odd.
In this paper, we study the Galois coinvariants of the unramified Iwasawa modules of a certain multiple Zp-extension in a relatively general situation.
Roughly speaking, this paper consists of two parts
“Split case” (§2, 3)
and
“Non-split case” (§4, 5).
In Split case, we consider a certain multiple Zp-extension K′ of a CM-field K which satisfies the condition of Gross’s conjecture of rank one.
If K is an abelian extension in which p splits completely and the degree of K is coprime to p,
then K′ is coincide with K.
Our first main result is a description of
the order of a Galois coinvariant of XK′ in terms of
the characteristic power series of XK∞
(Theorem 1.1).
In addition, for an imaginary quadratic field K in which p splits as (p)=PP, we give a theorem which suggests that the characteristic ideal of P-ramified Iwasawa module of K relates to the structure of XK (Theorem 1.4).
On the other hand, in Non-split case, we consider K for an imaginary quadratic field K in which p does not split.
Our second main result is to give a necessary and sufficient condition for which XK is
Zp[[Gal(K/K)]]-cyclic
under several assumptions on the Iwasawa λ-invariant and the ideal class group of K
(Theorems 1.5 and 5.11).
Such XK will be useful for studying the Iwasawa theory of multiple Zp-extensions.
We remark that our results do not need the assumption that Greenberg’s generalized conjecture holds.
1.2 Notation
Throughout this paper, we use the following notation.
Let p be an odd prime number,
k a totally real number field,
K a CM-field such that K/k is a finite abelian extension of degree coprime to p,
K∞/K the cyclotomic Zp-extension, and K the maximal multiple Zp-extension over K.
For any (finite or infinite) extension F over Q,
we denote by LF and XF the maximal unramified abelian p-extension of F and
the Galois group of LF over F, respectively.
If F is a finite extension of Q, we denote by AF the p-Sylow subgroup of the ideal class group of F.
We identify
Zp[[Gal(K∞/K)]] with the ring of formal power series Zp[[S]]
by regarding a fixed generator of Gal(K∞/K) as 1+S.
For a character χ:Gal(K/k)→Qp×, we denote by Oχ the ring obtained from Zp by adjoining all values of χ.
For any Gal(K/k)-module M, put
Mχ:=M⊗Zp[Gal(K/k)]Oχ.
We denote by μp the set of all p-th roots of unity, and simply by X/a a quotient module X/aX .
Let Λ be the ring either
O[[S]] or O[[S,T]],
where O is Zp or Oχ.
For any finitely generated torsion Λ-module X,
we call a generator of the characteristic ideal of X
a characteristic power series of X, and denote it by charΛ(X)∈Λ,
which is determined up to Λ×.
1.3 Main theorems of “Split case”
Let χ:Gal(K/k)→Qp× be an odd character.
Assume that there is only one prime ideal p in k above p which satisfies that χ(p)=1.
Let K′/K be the maximal multiple Zp-extension such that K∞⊂K′ and K′/K∞ is unramified.
We know that Gal(K′/K) is a Gal(K/k)-module.
Note that if k=Q, then K=K′.
By [3] (as a consequence of Lemma 1.5 and Proposition 1.6),
we see that Gal(K′/K)χ≃Oχ.
Let Kχ be the subextension of K/K∞ such that
Gal(Kχ/K∞) is isomorphic to Gal(K′/K)χ.
We identify Zp[[Gal(Kχ/K)]]=Zp[[Gal(K∞/K)×Gal(Kχ/K∞)]] with Zp[[S,T1,…,Tdχ]] in a similar way as we did Zp[[Gal(K∞/K)]] with Zp[[S]], where dχ:=[Oχ:Zp].
Theorem 1.1**.**
Let p, k, K be as in §1.2, χ:Gal(K/k)→Qp× an odd character.
Assume that
μp⊂K
and that
there exists just only one prime ideal p in k above p which satisfies that χ(p)=1.
Then Gal(K/k)-module
(XKχ)Gal(Kχ/K)=XKχ/(S,T1,⋯,Tdχ)
satisfies
[TABLE]
where fχ∗ is the first non-vanishing coefficient of a characteristic power series of the Oχ[[S]]-module XK∞χ.
Moreover, assume that Leopoldt’s conjecture holds for the maximal totally real subfield K+ and p, then
there is a canonical isomorphism
[TABLE]
Remark 1.2**.**
Since Gross’s ‘order of vanishing conjecture’ (see Gross [10, Conjecture 1.15] or Dasgupta, Kakde, and Ventullo [4, Conjecture 1], for example) of rank one for K/k holds by the argument in [10, Proposition 2.13],
fχ∗ turns out to be the coefficient of degree 1.
**
Corollary 1.3**.**
Let p be an odd prime number, K an imaginary abelian finite extension over Q of degree coprime to p in which p splits completely.
Then, for any odd character χ of Gal(K/k), we have
[TABLE]
In §3, we will consider the case where K is an imaginary quadratic field such that p splits completely as (p)=PP and connect the above corollary with the Galois group XP(K) of the maximal abelian p-extension of K which is unramified outside all primes above P.
It is known that XP(K)
is finitely generated and torsion over Zp[[S,T]]=Zp[[Gal(K/K)]](T:=T1)
by [18, Theorem 5.3 (ii)].
Therefore, we can consider the characteristic ideal
of XP(K) as a Zp[[S,T]]-module,
which plays an important role in the Iwasawa main conjecture.
We denote by λ the Iwasawa λ-invariant of K∞/K.
Then it is known that XP(K) is generated by λ−1 elements as a Zp[[T]]-module.
Moreover, let Iλ−1 be the identity matrix of size λ−1 and A a matrix associated to multiplication by S on XP(K) whose entries are in Zp[[T]].
We will show the following theorem which suggests that the characteristic ideal of XP(K) relates to the structure of XK.
Theorem 1.4**.**
Let p be an odd prime number, K an imaginary quadratic field such that p splits completely as (p)=PP and K the unique Zp2-extension of K.
Assume that
LK⊂K
or that
the characteristic ideal
of XK∞
does not have any square factor.
Then the characteristic power series
charZp[[S,T]](XP(K))
of XP(K)
satisfies
[TABLE]
as ideals.
In particular, combining these with Corollary 1.3, we have
[TABLE]
1.4 Main theorems of “Non-split case”
In the latter half of this paper, we consider the case where K is an imaginary quadratic
field
such that p does not split.
Then it is well known that the number of generators of XK∞ as a Zp[[S]]-module is equal to dimFp(AK/p).
Our target is the number of generators of XK as a Zp[[S,T]]-module,
which is equal to dimFp(XK/(p,S,T)) by Nakayama’s lemma.
It is easy to see that
[TABLE]
(see §4.1).
In §4.1, we describe a system of generators of XK and show the following theorem which classifies dimFp(XK/(p,S,T)) in the case where AK is a cyclic abelian group.
Theorem 1.5**.**
Let p be an odd prime number and K an imaginary quadratic field such that p does not split.
(i)
(trivial case)*
Assume that LK∩K=K, then
dimFp(XK/(p,S,T))=dimFp(AK/p).*
(ii)
Suppose that LK∩K=K, and that dimFp(AK/p)=1.
(ii-a)
If λ=1, then
dimFp(XK/(p,S,T))=1.
(ii-b)
If λ≥2, then
[TABLE]
From the point of view of applications,
we are interested in the condition for XK to be Zp[[S,T]]-cyclic.
In §5, we classify when XK is Zp[[S,T]]-cyclic in a certain case with dimFp(AK/p)=1 and λ=2
(Theorem 5.11).
We are going to give some numerical examples about it and introduce a method of calculating these examples
in the forth coming paper.
For a module M and a morphism φ:M→M, we define M[φ]:=Ker(φ) and φM:=Im(M).
Proposition 2.1**.**
Let
[TABLE]
be an exact commutative diagram of modules.
Regard L as a submodule in M.
If γN=0 and #(M/M[β]+βM)<∞, then the following equation
[TABLE]
holds.
Proof.
We consider an exact commutative diagram
[TABLE]
where the vertical maps are defined by the difference between the two components, for example,
l:(x,y)↦x−y.
We will show Ker(n)≃Coker(n).
Since γN=0, the image of βM/αL by n is zero.
Also the natural map M[β]/L[α]→N[γ]=N is injective by the snake lemma.
Hence we obtain
Ker(n)=βM/αL.
On the other hand,
we see
[TABLE]
Hence an exact sequence
0→L+M[β]→M→ββM/αL→0
induces an isomorphism
βM/αL≃Coker(n).
Therefore we obtain Ker(n)≃Coker(n).
Applying the snake lemma to the first exact commutative diagram, we have an exact sequence
[TABLE]
From this and the assumption that #M/M[β]+βM<∞, we have the claim.
∎
Corollary 2.2**.**
For the exact commutative diagram (2.1) in Proposition 2.1,
suppose that #(M/M[β]+βM)<∞.
(i)
Assume that there exists some integer n>0 such that γnN=0.
Then the equation (2.2) holds.
In particular, if all modules in (2.1) are Zp[[S]]-modules, the maps α, β, γ are multiplication by S, and N is finite, then the equation (2.2) holds.
(ii)
Assume that γN=0, M[β]+βM/βM is torsion-free, and L/αL is finite.
Then
[TABLE]
Proof.
(i)
Put M0:=M,
Mi:=βiM+L (i≥0).
Note that Mn=L, since
0=γnN≃βnM+L/L.
Define Ni by the exact sequence
[TABLE]
Then we have βNi=0.
Therefore we can easily check that
#(Mi/Mi[β]+βMi)<∞ for i=0,…,n,
using the same method as in the proof of Proposition 2.1 and the induction on i.
Therefore we can apply Proposition 2.1 to the above exact sequence, so that
[TABLE]
Taking the products from i=0 to i=n−1, we have the claim since Mn=L.
(ii)
Note that βM⊂L and L[α]/L[α]∩αL is finite by the assumption.
Hence the exact sequence
[TABLE]
implies that
(M[β]∩βM)/(L[α]∩αL)≃L[α]+αL/αL,
since M[β]+βM/βM is torsion-free.
Therefore,
[TABLE]
by (i).
This completes the proof.
∎
Let O be the ring either Zp or Oχ.
Applying the corollary to O[[S]]-modules, we obtain the following corollary.
Let X be a finitely generated torsion
O[[S]]-module which has no nontrivial finite O[[S]]-submodules.
Then, by the structure theorem of finitely generated torsion
O[[S]]-modules
(see [19, Theorem 13.12]),
there is an exact sequence of
O[[S]]-modules
[TABLE]
such that E is described as E=⨁jO[[S]]/(fj(S)nj) and C is finite.
Here, each fj(S)∈O[S] is an irreducible distinguished polynomial or a uniformizer in O.
Corollary 2.3**.**
With the notation as above, we denote the first non-vanishing coefficient of ∏jfj(S)nj by f∗.
If S2 does not divide fj(S)nj for any j, then
[TABLE]
Proof.
In the case where E=O[[S]]/(f(S)n) with S∤f(S), we have
[TABLE]
And also, in the case where E=O[[S]]/(S), we see
[TABLE]
Hence, in both cases, E[S]∩SE=0.
Applying Corollary 2.2 (i) to the exact sequence (2.3), we have
#(X/X[S]+SX)=#(E/E[S]+SE)=#O/f∗.
This completes the proof.
∎
2.2 Galois coinvariants
We use the notation in §1.2.
Let χ:Gal(K/k)→Qp× be an odd character.
Suppose that the assumption in Theorem 1.1.
In other words,
we assume that μp⊂K and
there is only one prime ideal p lying above p in k which satisfies χ(p)=1.
Recall that K′ is the maximal multiple Zp-extension such that K∞⊂K′ and K′/K∞ is unramified.
Then there is a unique abelian extension Kχ/K contained in K′ such that
[TABLE]
as Gal(K/k)-modules by
[3, Lemma 1.5].
Also, recall that we identify
Zp[[Gal(Kχ/K)]] with Zp[[S,T1,…,Tdχ]],
where dχ:=[Oχ:Zp].
Since LKχ/k is a Galois extension because of the maximality of LKχ,
Gal(Kχ/k) acts on XKχ by the inner product:
α(x):=αxα−1 for α∈Gal(Kχ/k) and x∈XKχ, where α∈Gal(LKχ/k) is a fixed extension of α.
Note that the action is independent of the choice of extensions α.
By this action, XKχ becomes a Gal(Kχ/k)-module.
We have two exact sequences
[TABLE]
and
[TABLE]
Put Y:=XKχ/(T1,…,Tdχ).
The first part of the following lemma is a partial generalization of Ozaki [16, Lemma 1].
Lemma 2.4**.**
(i)
There is an exact sequence of Zp[[S]]-modules
[TABLE]
(ii)
Assume that Leopoldt’s conjecture holds for K+ and p, then the latter half of Theorem 1.1 holds, in other words,
there is a canonical isomorphism
((XK′)Gal(K′/K∞))χ≃Yχ.
Proof.
(i)
Taking the Hochschild-Serre spectral sequence of (2.4), we obtain an exact sequence
of Gal(K∞/k)-modules
[TABLE]
Since Gal(K/K+) acts trivially on
H2(Oχ,Zp)≃Oχ∧ZpOχ
(here, Gal(K/K+) acts on the right hand side diagonally), its χ-component is trivial.
Hence we obtain the claim.
(ii)
In the same way,
we have
[TABLE]
from (2.5).
Combining this with (i), we have only to show that
(Gal(K′/K∞)∧ZpGal(K′/K∞))−=0, since Gal(K′/K∞)χ=Gal(Kχ/K∞).
For this, it is enough to show that Gal(K′/K∞)+=0.
Assume that it does not hold.
Then Gal(K′/K∞)+ is nontrivial and torsion-free, which implies that there exists a Zp-extension of K+ different from the cyclotomic Zp-extension.
This contradicts our assumption that Leopoldt’s conjecture holds.
∎
It is known
that
(XK∞/XK∞[S]+SXK∞)χ is finite and
(XK∞[S]+SXK∞/SXK∞)χ is torsion-free
(see Kurihara [13], for example),
since Gross’s order of vanishing conjecture holds.
Applying Corollary 2.2 (ii) to the exact sequence (2.6), we have
[TABLE]
since the actions of Gal(K/k) and Gal(K∞/K) are commutative.
Again, since the Gross’s conjecture holds,
we can apply Corollary 2.3 to get
Under several assumptions, we can calculate the value f∗ in Corollary 1.3.
More precisely,
let K be an imaginary quadratic field and p an odd prime number.
We suppose that LK⊂K.
We also suppose that p≥5 if p does not split in K.
Then we have
[TABLE]
where
D is the decomposition group in Gal(K/K) of a prime lying above p (Murakami [15, Proposition 3.4]).
**
We consider the case where K is an imaginary quadratic field such that p splits completely as (p)=PP.
Then K′=K.
Let Λ be the ring either Zp[[S]] or Zp[[S,T]].
Recall that, for any finitely generated torsion Λ-module X, we chose a characteristic power series charΛ(X)∈Λ of X.
A Λ-module X is called pseudo-null if X has two relatively prime annihilators in Λ.
Lemma 3.1**.**
Let
0→X→E→C→0
be an exact sequence of finitely generated torsion Λ-modules.
For a Λ-module M, denote by AnnΛ(M) the annihilator ideal of M.
(i)
If E has no non-trivial pseudo-null Λ-submodules and C is a pseudo-null Λ-module, then we have
AnnΛ(X)=AnnΛ(E).
(ii)
Furthermore, suppose that
E=⨁i=1sΛ/pini
where each pi is a height one prime ideal with pi=pj for i=j.
Then we have
AnnΛ(X)=(charΛ(X)).
Proof.
(i)
The inclusion
AnnΛ(X)⊃AnnΛ(E)
is obvious.
We will show the other inclusion.
Let f∈Λ.
Then we have an exact sequence
[TABLE]
Suppose that f∈AnnΛ(X).
Then we have X/f=X, hence we have an commutative diagram
[TABLE]
By snake lemma, we have
0→fE→fC→0, hence fE is a pseudo-null Λ-module.
Since E has no non-trivial pseudo-null Λ-submodules, we have fE=0.
Thus we have f∈AnnΛ(E), hence we have proved AnnΛ(X)⊂AnnΛ(E).
(ii)
This follows easily since AnnΛ(E)=(charΛ(X)) in this case.
∎
Let XP(K) be the Galois group of the maximal p-extension of K unramified outside all primes above P.
Put
[TABLE]
for simplicity.
We may assume that F(S) is a distinguished polynomial with degree λ, where λ is the Iwasawa λ-invariant of K∞/K.
From Lemma 2.4 (i), we have the exact sequence
[TABLE]
Therefore F(S) is written as F(S)=SF∗(S) for some distinguished polynomial F∗(S) which is coprime to S with degree λ−1,
and XK/T has no nontrivial finite Zp[[S]]-submodules.
On the other hand, by Fujii [6, Lemma 3],
XP(K) has no nontrivial pseudo-null submodules and
[TABLE]
since K is an imaginary quadratic field.
Therefore XP(K)/T is Zp[[S]]-torsion.
This yields that XP(K)[T] is a pseudo-null Zp[[S,T]]-module by
Perrin-Riou [17, Chapitre I Lemme 4.2 in page 12],
so that XP(K)[T]=0.
Hence we have
[TABLE]
as ideals in Zp[[S]], again by the same lemma in [17].
By (3.2),
we can take λ−1 generators
ξ1,…,ξλ−1
of XP(K) as a Zp[[T]]-module.
Define a matrix A and f(S,T)∈Zp[[T]][S] by
[TABLE]
and f(S,T)=det(S⋅Iλ−1−A), respectively.
Then f(S,T) annihilates XP(K), and also XK.
Hence, f(S,0) annihilates XK/T.
Note that the degree of f(S,T) with respect to S is λ−1.
Now we show Theorem 1.4.
Theorem 3.2**.**
(Theorem 1.4)*
Suppose that
LK⊂K
or F(S) does not have any square factor.
Then the following equalities*
[TABLE]
as ideals
hold.
Proof.
First, suppose that F(S) does not have any square root.
Combining the equation (F∗(S))=(FP(S,0)) with this assumption,
we see that FP(S,T) does not have any square factor.
By Lemma 3.1 (ii),
[TABLE]
for some g(S,T)∈Zp[[S,T]].
Since degF∗(S)=λ−1=degf(S,0) and f(S,0) is monic, we obtain that
F∗(S) is equal to f(S,0)=FP(S,0)g(S,0) up to multiplication by unit.
Hence, g(S,T)∈Zp[[S,T]]×, which yields the claim.
Next, suppose that LK⊂K.
Then it is known that K coincides with the maximal abelian p-extension of K which is unramified all primes outside above p.
Hence XK∞/S≃Zp.
Applying the snake lemma to (3.1),
we obtain an exact sequence
Zp→XK/(S,T)→XK∞/S→Zp→0.
Hence XK is Zp[[S,T]]-cyclic and so XP(K) is.
This means that there are surjections
Zp[[S,T]]/f(S,T)↠XP(K)
and
Zp[[S]]/f(S,0)↠XK/T.
So, in the same way as above, we obtain the claim.
∎
First of all, we show a lemma from group theory.
Let p be an arbitrary prime number and G a finite abelian p-group such that dimFpG/p=g.
For x∈G with order pn, ⟨x⟩n denotes the cyclic group generated by x
(we also use the notation ⟨x⟩ instead of ⟨x⟩n).
Lemma 4.1**.**
With the notation as above, let H be a subgroup of G such that G/H is cyclic, then
there exists a minimal system of generators x1,x2…,xg∈G such that the following holds:
(i)
G/H* is generated by the image of x1 under the projection.
Moreover, x1 has the minimum order among such elements.*
(ii)
x2,…,xg∈H, if g≥2.
Proof.
We denote the image of x∈G in G/H by x.
First, we take a direct sum decomposition
[TABLE]
with
ni1≤ni2≤…≤nigi for each 1≤i≤r
and
[TABLE]
in G/H.
For simplicity, we put xi:=xi1 and ni:=ni1(1≤i≤r).
For any i and j with 1≤i≤r and 2≤j≤gi,
there exist hij∈H and p∤aij
such that
xi=hij+aijxij,
since ⟨xi⟩mi=⟨xij⟩mi.
Then, we can easily check that
⟨xi⟩n1⊕⟨xij⟩nij=⟨xi⟩n1⊕⟨hij⟩nij.
Hence, we may change ⟨xij⟩nij in the above decomposition with ⟨hij⟩nij.
Therefore, by changing the names appropriately, we obtain a direct sum decomposition
[TABLE]
for some subset H′ in H.
Next, for any i with 2≤i≤r,
there exist hi∈H and p∣ai
such that
xi=hi+aix1,
since ⟨x1⟩m1⊋⟨xi⟩mi.
Then
[TABLE]
and elements in ⋃i=2r{hi}∪H′ are linearly independent mod pG.
Finally, we show that the above x1 can be taken such that it has the minimum order among the elements whose images generate G/H.
Let x1′∈G be such an element.
Then we have x1′=h′+a′x1 for some h′∈H and p∤a′.
We also obtain
h′=b1x1+∑i=2rbihi+∑h∈H′bhh
for some b1,…br,hh∈Z.
Then b1x1∈⟨x1⟩∩H, so that pm1∣b1.
Since x1′ has the form x1′=(a′+b1)x1+∑i=2rbihi+∑h∈H′bhh,
we obtain that
G=⟨x1′⟩+⟨h2,…,hr⟩+⨁h∈H′⟨h⟩.
∎
Let p be an odd prime number, K an imaginary quadratic field such that p does not split.
We use the notation in §1.2 for such p and K, K, K∞, XK, etc.
Put g:=dimFp(AK/p).
We use Ozaki’s exact sequence in the case where p does not split:
Lemma 4.2**.**
([16, Lemma 1])*
There is an exact sequence of Zp[[S]]-modules*
[TABLE]
Lemma 4.3**.**
There is a canonical isomorphism
[TABLE]
Proof.
Let L′ be the maximal abelian subextension in LK∞/K.
We see that L′=K∞LK,
since LK is the fixed field by the inertia group of a prime lying above p in Gal(L′/K) and L′/K∞LK is unramified.
If we show that
[TABLE]
then we obtain
Gal(LK∞∩K/K∞)=Gal(K∞(LK∩K)/K∞)≃Gal(LK∩K/K).
Let us show (4.1).
We have
LK∩K⊂LK⊂K∞LK⊂LK∞, so that
[TABLE]
Since LK∞∩K is abelian over K and unramified over K∞,
we see that
LK∞∩K⊂L′=K∞LK.
This yields
Now, we show Theorem 1.5 (i).
If m=0, i.e., LK∩K=K, then
XK/T≃XK∞,
so that
dimFp(XK/(p,S,T))=g.
This completes the proof of Theorem 1.5 (i).
In the following, we assume that m>0.
We fix certain generators x1,…,xg of XK∞ as a Zp[[S]]-module as follows.
Applying Lemma 4.1 to Gal(LK/K) and its quotient Gal(LK∩K/K),
we can choose the basis of Gal(LK/K) which satisfies the conditions in Lemma 4.1.
Moreover,
applying Nakayama’s lemma to this basis,
then we can choose x1,…,xg in XK∞ which satisfy the conditions bellow.
•
x1,…,xg generate XK∞.
•
The image of x1 generates Gal(LK∩K/K).
Also the images of x2,…,xg
become [math] in Gal(LK∩K/K).
•
Moreover, if
Gal(LK/K)≃Gal(LK∩K/K)⊕Gal(LK/LK∩K)
and the exponent of Gal(LK/LK∩K) is equal to or less than pm, then
x1 can be replaced modulo (x2,…,xg)Zp[[S]]
(this fact is useful in §5).
Note that x1,…,xg are defined modulo SXK∞.
In the following, we assign the sum “∑j=2g” to [math] if g=1.
For any a1,…,am∈XK∞, (a1,…,am)X denotes the Zp[[S]]-submodule of XK∞ generated by a1,…,am.
Put
[TABLE]
By Lemma 4.2, we identify XK/T
with
the submodule in XK∞.
Proposition 4.4**.**
We have the following.
(i)
XK/T=M+N.
(ii)
(M+L/L)∩(N+L/L)=0, and hence XK/(p,S,T)≃(M+L/L)⊕(N+L/L).
(iii)
N+L/L≃Fp⊕(g−1).
Proof.
(i)
By
Lemmas 4.2 and 4.3,
we see
x2,…,xg∈XK/T, and also
[TABLE]
since Gal(LK∩K/K) has order pm and
the
trivial action by Gal(K∞/K).
Hence
M+N⊂XK/T.
Since [XK∞:M+N]≤pm, we have the claim.
(ii)
Any elements in (M+L/L)∩(N+L/L) are represented
by some elements in XK∞ of the form
[TABLE]
The left hand side is congruent to [math] modulo (p,S)XK∞.
Since x1,…,xg are linearly independent in AK/p, we see
b2,…,bg=0.
This implies that (M+L/L)∩(N+L/L)=0.
(iii)
By an exact sequence
N/(p,S)→XK∞/(p,S)→(XK∞/N)/(p,S)→0,
we have
N/(p,S)≃(Zp/p)⊕(g−1).
Therefore we have only to show N∩L⊂(p,S)N, since
[N:N∩L]≤pg−1.
Any elements in N∩L are represented
by some elements in XK∞ of the form
[TABLE]
Since x1,…,xg are linearly independent modulo (p,S)XK∞, we see
B2′′,…,Bg′′∈(p,S)Zp[[S]].
This implies that N∩L⊂(p,S)N.
∎
We denote the image of x∈XK∞ in Gal(LK/K) by x.
The following lemma gives criterions in the case where ⟨x1⟩ is a direct summand of Gal(LK/K).
We will use the lemma in §4.2 and §5.
Lemma 4.5**.**
Denote the order of x1 by ord(x1).
(i)
If pm=ord(x1),
in other words,
Gal(LK/K)≃Gal(LK∩K/K)⊕Gal(LK/LK∩K),
then
[TABLE]
(ii)
Suppose that
Gal(LK/K)=⟨x1⟩⊕⟨x2,…,xg⟩
and
0<pm<ord(x1), then
pmx1∈L.
In particular, dimFp(XK/(p,S,T))≥g.
Proof.
(i)
By the assumption, we have pmx1≡0 mod SXK∞.
Therefore, there exist some A1,…,Ag∈Zp[[S]] such that
[TABLE]
The second term in the right hand side is in (p,S)N⊂L.
This implies that M+L/L is generated by Sx1+L.
Moreover, we see that pm+1x1 is written as a linear form of pSx1 and elements in (p,S)N.
We claim that Sx1∈L if and only if Sx1∈(p,S)N,
which implies the conclusion by Proposition 4.4.
Assume that Sx1∈L.
Then there exist some A,A′,Bj,Bj′∈Zp[[S]] (j=2,…,g) such that
[TABLE]
so that
[TABLE]
This implies that Sx1∈(p,S)N since 1−Ap−A′S∈Zp[[S]]×.
The converse is trivial.
(ii)
Assume that pmx1∈L.
Then, in a similar way as (i), we see that there exists some a∈Zp such that
[TABLE]
since
Gal(LK/K)=⟨x1⟩⊕⟨x2,…,xg⟩.
This implies that pmx1 equals [math] in AK, since 1−ap∈Zp×, which is a contradiction.
∎
4.2 Classification in the case where AK is cyclic
Now, we show Theorem 1.5 (ii).
Suppose that LK∩K=K and that dimFp(AK/p)=1, in other words, g=1.
Then, N=0 and there is an isomorphism
XK∞≃Zp[[S]]/F(S),
where F(S)∈Zp[S] is
the distinguished polynomial generating the characteristic ideal
of XK∞,
since AK is cyclic and XK∞ has no nontrivial finite Zp[[S]]-submodules.
Note that we have Sx1=0, since S dose not divide F(S).
In this case, we can apply Lemma 4.5.
Proof of (ii-a).
Assume that λ=1.
If pm=#AK, then
dimFp(XK/(p,S,T))=1 by Lemma 4.5 (i).
On the other hand, assume that 0<pm<#AK.
Since
F(S)x1=0 in XK∞
and F(0)x1∈L, we obtain Sx1∈L, which implies that dimFp(XK/(p,S,T))=1 by
Proposition 4.4 (ii), (iii).
Proof of (ii-b).
Assume that λ≥2.
If pm=#AK, then
dimFp(XK/(p,S,T))=1.
On the other hand, assume that 0<pm<#AK.
Since we see that F(S)≡F(0) mod (pS,S2) from degF(S)≥2, we obtain
[TABLE]
This completes the proof of (ii).
5 Conditions of XK to be cyclic in the case where λ=2
5.1 Setting and Methods
Let p be an odd prime number, K an imaginary quadratic field such that p does not split.
We use the notation in §1.2 for such p and K.
We consider conditions that
XK
becomes
Zp[[S,T]]-cyclic.
In the case where dimFp(AK/p)=1, Theorem 1.5 gives the condition, so that we have only to consider the case where dimFp(AK/p)=2.
In this section, we treat the case where
dimFp(AK/p)=2 and Gal(LK∩K/K) is a direct summand of Gal(LK/K)≃AK.
Moreover, we add more assumptions that λ=2 and that
the roots α,β∈Qp of the distinguished polynomial generating the characteristic ideal
of XK∞ satisfy α=β.
We remark that the latter part of this assumption is expected to be held.
Define O:=Zp[α,β] and Λ:=O[[S]], and let
π
be a uniformizer in O.
Then, by the above assumption, the characteristic ideal
of XK∞⊗ZpO is described as
[TABLE]
Then, by Koike [12], there exist an integer k with 0≤k≤ordπ(β−α)
and an O-basis e1,e2 of XK∞⊗ZpO such that the homomorphism of Λ-modules
[TABLE]
is injective.
Note that k depends only on the isomorphism class of XK∞.
We regard XK∞⊗ZpO
as
a Λ-submodule of Λ/(S−α)⊕Λ/(S−β) by the above injection.
We can express the action of S by
[TABLE]
For convenience, we regard XK∞⊂XK∞⊗ZpO by the injection x↦x⊗1, and
put
γ:=(β−α)π−k and ord(a):=ordπ(a) for a∈O.
We take such generators x1,x2 of XK∞ as in §4.1.
They are represented as
[TABLE]
for some λij∈O.
We have AK⊗ZpO=(O/πN1)x1⊕(O/πN2)x2 and
Gal(LK∩K/K)⊗ZpO=(O/πN1)x1
for some N1,N2∈Z.
We denote
by
⟨a1,…,al⟩Λ
(resp. ⟨a1,…al⟩O)
the Λ-submodule (resp. O-submodule) in Λ/(S−α)⊕Λ/(S−β) generated by a1,…,al.
Lemma 5.1**.**
We have an isomorphism
[TABLE]
Proof.
We have
[TABLE]
Since AK⊗ZpO≃(XK∞⊗ZpO)/S, this yields the claim.
∎
Lemma 5.2**.**
The following conditions hold:
[TABLE]
Proof.
Since πN1x1∈SXK∞⊗ZpO, we have
πN1(λ11e1+λ12e2)∈⟨αe1+γe2,βe2⟩O.
Hence, there exist some s,t∈O such that λ11πN1=sα, λ12πN1=sγ+tβ.
They induce the first relations.
The rest are verified in the same way.
∎
Recall that we defined N as the Zp[[S]]-submodule of XK∞ generated by x2 and that Lemma 4.5 (i) gives a criterion whether XK is Zp[[S,T]]-cyclic or not.
In the same way as the proofs of Proposition 4.4 and Lemma 4.5, we can show the following
Lemma 5.3**.**
XK* is Zp[[S,T]]-cyclic if and only if Sx1∈(π,S)N⊗ZpO.*
Proof.
Note that there is nothing to show if O/Zp is unramified.
And also, note that
XK is Zp[[S,T]]-cyclic if and only if XK⊗ZpO is O[[S,T]]-cyclic by Nakayama’s lemma.
Put
L′:=(π,S)(M⊗ZpO+N⊗ZpO).
Then, in the same way as the proofs of Proposition 4.4 (ii) and (iii), we can show that
XK⊗ZpO≃(M⊗ZpO+L′/L′)⊕(N⊗ZpO+L′/L′)
and
(N⊗ZpO+L′)/L′≃Fp.
Moreover, in the same way as the proof of Lemma 4.5 (i), we can show that
Sx1∈(π,S)N⊗ZpO
if and only if
Sx1∈L′.
∎
Lemma 5.4**.**
XK* is Zp[[S,T]]-cyclic if and only if
there exist some f(S),g(S)∈Λ such that*
[TABLE]
Proof.
We have
Sx1=[λ11αλ11β+λ12βπk]
and
[TABLE]
Then
the
claim follows immediately by Lemma 5.3.
∎
Now, we consider the case
where
N1≥N2.
Then, as we mentioned in §4.1,
x1 is defined modulo (x2,SXK∞)⊂XK∞,
so that we may change x1 up to
x2Λ.
Furthermore, since λ11λ22−λ12λ21∈O×,
at least one (possibly both) of λ21, λ22 is in O×.
When λ21∈O×
(resp. λ22∈O×),
we may assume that λ21=1, λ11=0, and λ12=1
(resp. λ22=1, λ12=0, and λ11=1).
Thus, if N1≥N2, we are reduced to only two cases:
(I)
λ21=1, λ11=0, and λ12=1.
In other words,
x1=e2 and x2=e1+λ22e2.
(II)
λ22=1, λ11=1, and λ12=0.
In other words,
x1=e1 and x2=λ21e1+e2.
Corollary 5.5**.**
Suppose that N1≥N2.
We have the following.
(i)
In the case of (I),
XK is Zp[[S,T]]-cyclic if and only if
there exists some x∈O such that
[TABLE]
(ii)
In the case of (II),
XK is Zp[[S,T]]-cyclic if and only if
there exist some f(S),g(S)∈Λ such that
[TABLE]
Proof.
There is nothing to show for (ii)
by Lemma 5.4.
We prove (i).
First, assume that f(S),g(S)∈Λ satisfy (5.2).
Then f(α)=0.
Therefore, f(S) is divisible S−α by division lemma
(see [2, Chapter VII §3.8 Proposition 5]).
Put x:=f(β)π/(β−α)+g(β).
Then we can easily check x∈O and
β=(1+λ22πk)γx.
Conversely, assume that there exist such x∈O.
Put f(S):=0, g(S):=x
Then we can also easily check that they satisfy (5.2).
∎
5.2 The case where k>0
Proposition 5.6**.**
If k>0 and
ord(γ)<min{ord(α),ord(β)},
then XK is Zp[[S,T]]-cyclic.
Proof.
In this case, AK⊗ZpO≃O/γ⊕O/(αβ/γ) by Lemma 5.1.
First, we suppose that N1=ord(γ), N2=ord(αβ/γ).
Then N1<N2
(note that we cannot apply Corollary 5.5 in this case).
Then, πN1/α∈/O, so that π∣λ11 by Lemma 5.2.
Hence, λ12∈O×.
We may assume that λ12=1 and also that λ21=1.
By Lemma 5.2, we see that
λ12−λ11γ/α
must be divided by π,
so that we obtain
ord(λ11)=ord(α/γ).
Put
[TABLE]
Then g(S)∈O, since 1+λ22πk∈O× and ord(β−α)−k=ord(γ)<min{ord(α),ord(β)}.
They satisfy (5.2).
Second, we suppose that N1=ord(αβ/γ), N2=ord(γ).
Then N1>N2 so that we can apply Corollary 5.5.
By Lemma 5.2, we know that λ21γ/α∈O,
and hence ord(λ21)>0.
This implies that λ11,λ22∈O×, so we are reduced only to the case of (II).
Thus, we may assume that
λ11=λ22=1 and λ12=0.
Put
δ:=(λ22−λ21γ/α)/β=(1−λ21γ/α)/β.
Then
[TABLE]
Note that ord(δ)≥−ord(γ)>−min{ord(α),ord(β)} by Lemma 5.2,
so that both 1−αδ and 1−βδ are units.
Now, put
[TABLE]
Here, the sum is convergent since ord(αδ)>0 and ord(βδ)>0.
Hence g(S)∈O, and they satisfy the condition in Corollary 5.5 (ii).
∎
The following proposition does not need the assumption that k>0.
Proposition 5.7**.**
If ord(γ)>min{ord(α),ord(β)},
then XK is not Zp[[S,T]]-cyclic.
Proof.
Note that ord(α)=ord(β)=N1=N2 in this case, so that we can apply the criterion in Corollary 5.5.
Assume that XK is Zp[[S,T]]-cyclic.
In the case of (I), the condition
that
(1+λ22πk)−1β/γ∈O
contradicts the inequality ord(γ)>ord(β).
We consider
the case of (II).
By ord(γ)>ord(α),
we have
s:=1−λ21γ/α∈O×.
By the above assumption and Corollary 5.5 (ii), there are some f(S),g(S)∈Λ such that
α=f(α)λ21π and
[TABLE]
Now, f(β) has a form f(β)=f(α)+(β−α)A(β) with some A(S)∈Λ
by division lemma,
and f(α)π=α/λ21=γ(1−s)−1.
Therefore,
[TABLE]
So we have
[TABLE]
The order of the left hand side is equal to or less than k, since s∈O×.
On the other hand, the order of 1−αs/β is greater than [math].
In fact, s has a form s=1+πt with some t∈O, so that
[TABLE]
since ord(γ)>ord(β).
This is a contradiction.
∎
Proposition 5.8**.**
Suppose that k>0 and ord(γ)=min{ord(α),ord(β)}.
Then XK is Zp[[S,T]]-cyclic if and only if λ21∈O×.
Proof.
Note that ord(γ)=ord(α)=ord(β)=N1=N2, since k>0.
Therefore we can apply the criterion in Corollary 5.5.
In the case of (I), we see that (1+λ22πk)−1β/γ∈O, which implies
that
XK is Zp[[S,T]]-cyclic.
We consider
the case of (II).
In the case where λ21∈O×, we are reduced to
the case of (I),
so we may assume that π∣λ21.
Then we have only to show that XK is not Zp[[S,T]]-cyclic.
Assume the contrary.
Note that s:=1−λ21γ/α∈O×.
In the same way as Proposition 5.7, we get a contradiction,
since (β−α)/β≡0 mod π by k>0.
∎
5.3 The case where k=0
Suppose that
k=0.
Then XK∞⊗ZpO≃Λ/(S−α)⊕Λ/(S−β).
We use the standard basis
[10],
[01],
instead of
[11],
[01].
If ord(γ)>min{ord(α),ord(β)}, then XK is not Zp[[S,T]]-cyclic by Proposition 5.7.
So, in the following, we consider the case where
ord(γ)=min{ord(α),ord(β)}.
Moreover, we may assume that ord(α)≤ord(β).
Express the generators x1, x2 as
[TABLE]
for some μij∈O.
Then Sx1∈(π,S)N⊗ZpO if and only if
there exist some f(S),g(S)∈Λ such that
[TABLE]
Proposition 5.9**.**
Assume that ord(β−α)=ord(α)≤ord(β) and that N1<N2.
Then XK is Zp[[S,T]]-cyclic if and only if
μ21 is in O×.
Proof.
In this case,
AK⊗ZpO=(O/πN1)x1⊕(O/πN2)x2≃O/α⊕O/β.
Hence N1=ord(α), N2=ord(β).
Therefore, we have ord(μ12)>0.
In fact, if μ12∈O×, then the order of x1 in AK⊗ZpO is #(O/πN2), which is a contradiction.
Thus, we may assume that μ11=μ22=1.
By (5.3), XK is Zp[[S,T]]-cyclic if and only if there exist some f(S),g(S)∈Λ such that
[TABLE]
Assume that ord(μ21)>0 and
that
XK is Zp[[S,T]]-cyclic.
By division lemma,
f(α)=f(β)+(α−β)A(α) with some A(S)∈Λ.
Then
[TABLE]
This is a contradiction.
Hence, if XK is Zp[[S,T]]-cyclic, then μ21∈O×.
Conversely,
let μ21 be in O×.
Put
[TABLE]
Then we can easily check that
μ21(πf(α)+αg(α))=α and that πf(β)+βg(β)=βμ12.
Hence XK is Zp[[S,T]]-cyclic.
∎
Proposition 5.10**.**
Assume that ord(β−α)=ord(α)≤ord(β) and that N1≥N2.
Then XK is Zp[[S,T]]-cyclic if and only if
it holds that μ21∈O× and ord(μ22)=ord(β)−ord(α).
Proof.
In this case, we may change x1 up to modulo x2Λ
as we mentioned after the proof of Lemma 4.3.
As in §5.1, we have only to consider the two cases:
(I’)
μ21=1, μ11=0, μ12=1.
(II’)
μ22=1, μ11=1, μ12=0.
First, we deal with the case of (I’).
Then ord(μ22)≥ord(β)−ord(α).
In fact, if ord(α)=ord(β), then the inequality is trivial.
And also, if ord(α)<ord(β), then N2=ord(α) by Lemma 5.1, so that
πN2x2∈SXK∞⊗ZpO=⟨[α0],[0β]⟩O.
This implies that μ22α/β∈O.
By (5.3),
XK is Zp[[S,T]]-cyclic if and only if there exist some f(S),g(S)∈Λ such that
[TABLE]
Assume that ord(μ22)>ord(β)−ord(α) and XK is Zp[[S,T]]-cyclic.
Since πf(β) has a form πf(β)=πf(α)+π(β−α)A(β)=−αg(α)+π(β−α)A(β) with some A(S)∈Λ,
[TABLE]
This is a contradiction.
Conversely,
assume that
ord(μ22)=ord(β)−ord(α).
Put
[TABLE]
Then we can easily check
πf(α)+αg(α)=0 and μ22(πf(β)+βg(β))=β, and hence XK is Zp[[S,T]]-cyclic.
Second, we deal with the case of (II’).
By (5.3), XK is Zp[[S,T]]-cyclic if and only if there exist some f(S),g(S)∈Λ such that
[TABLE]
The similar argument in the case of (I’) shows that ord(α)<ord(β) does not occur.
In fact, the assumption ord(α)<ord(β) yields μ22α/β∈O, which is a contradiction since μ22=1.
Hence ord(α)=ord(β).
Using this fact, in the same way as the case of (I’), we can show that if ord(μ21)>0, then XK is not Zp[[S,T]]-cyclic.
Conversely, if ord(μ21)=0, then we can prove that
XK
is
Zp[[S,T]]-cyclic by taking
[TABLE]
Then, the proof is completed.
∎
Summarizing all results in this section, we obtain the following.
Theorem 5.11**.**
Let p be an odd prime number, K an imaginary quadratic field such that p does not split.
Suppose that
•
dimFp(AK/p)=2* and Gal(LK∩K/K) is a direct summand of Gal(LK/K).*
•
The Iwasawa λ-invariant of K∞/K is 2.
•
Let α,β∈Qp be the roots of the distinguished polynomial generating the characteristic ideal
of XK∞.
Then α=β.
Put O:=Zp[α,β] and l:=min{ord(α),ord(β)}.
Let x2∈XK∞ be a preimage of a generator of Gal(LK/LK∩K).
Also, we denote by
[μ21μ22]
the image of x2⊗1 under the injective map
XK∞⊗ZpO↪O[[S]]/(S−α)⊕O[[S]]/(S−β),
which is defined by (5.1).
Then, XK is Zp[[S,T]]-cyclic if and only if one of the following holds:
[TABLE]
*where each n1 and n2 is defined by
pn1=#Gal(LK∩K/K) and pn2=#Gal(LK/LK∩K), respectively.
*
Bibliography19
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] F. Bleher, T. Chinburg, R. Greenberg, M. Kakde, G. Pappas, R. Sharifi, and M Taylor, Higher Chern Classes in Iwasawa Theory, ar Xiv:1512.00273.
2[2] N. Bourbaki, Commutative algebra. Chapters 1–7, Springer-Verlag, Berlin, 1998.
3[3] H. Darmon, S. Dasgupta and R.Pollack, Hilbert modular forms and the Gross-Stark conjecture, Ann. Math. 174 (2011), 439–484.
4[4] S. Dasgupta, M. Kakde and K. Ventullo, On the Gross-Stark conjecture, Ann. of Math. 188 (2018), 833–870.
5[5] S. Fujii, On a bound of λ 𝜆 \lambda and the vanishing of μ 𝜇 \mu of ℤ p subscript ℤ 𝑝 \mathbb{Z}_{p} -extensions of an imaginary quadratic field, J. Math. Soc. Japan 65 (2013), 277–298.
6[6] S. Fujii, On restricted ramifications and pseudo-null submodules of Iwasawa modules for ℤ p 2 superscript subscript ℤ 𝑝 2 \mathbb{Z}_{p}^{2} -extensions, J. Ramanujan Math. Soc. 29 (2014), 295–305.
7[7] S. Fujii, On Greenberg’s generalized conjecture for CM-fields, J. Reine Angew. Math. 731 (2017), 259–278.
8[8] R. Greenberg, The Iwasawa invariants of Γ Γ \Gamma -extensions of a fixed number field, Amer. J. Math. 95 (1973), 204–214.