This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Anticyclotomic main conjecture and the non-triviality of Rankin–Selberg LL-values in Hida families

Chan-Ho Kim Center for Mathematical Challenges, Korea Institute for Advanced Study, 85 Hoegiro, Dongdaemun-gu, Seoul 02455, Republic of Korea chanho.math@gmail.com  and  Matteo Longo Dipartimento di Matematica, Università di Padova, Via Trieste 63, 35121 Padova, Italy mlongo@math.unipd.it
Abstract.

We prove the two-variable anticyclotomic Iwasawa main conjecture for Hida families and discuss its arithmetic application to a definite version of the horizontal non-vanishing conjecture, which is formulated in [LV11]. Our approach is based on the two-variable anticyclotomic control theorem for Selmer groups and the relation between the two-variable anticyclotomic LL-function for Hida families built out of pp-adic families of Gross points on definite Shimura curves studied in [CL16] and [CKL17] and the self-dual twist of the specialisation to the anticyclotomic line of the three-variable pp-adic LL-function of Skinner–Urban [SU14].

Key words and phrases:
Anticyclotomic main conjecture, Rankin–Selberg LL-values, Hida families, control theorem, Iwasawa theory
2020 Mathematics Subject Classification:
11R23, 11F33,11F67

1. Introduction

To state our main results, fix a prime p5p\geq 5 and an integer NN with pNp\nmid N, and let \mathcal{R} be a primitive branch of a Hida family of pp-adic modular forms of tame conductor NN; more precisely, \mathcal{R} is a noetherian domain, finite and flat over the Iwasawa algebra Λ=𝒪[[1+pp]]\Lambda=\mathcal{O}[[1+p\mathbb{Z}_{p}]], where 𝒪\mathcal{O} is the valuation ring of a fixed finite extension of p\mathbb{Q}_{p}. We assume throughout that \mathcal{R} is a regular ring. Let 𝐟=n1𝐚nqn[[q]]\mathbf{f}=\sum_{n\geq 1}\mathbf{a}_{n}q^{n}\in\mathcal{R}[[q]] denote the Hida family of pp-adic modular forms associated with \mathcal{R}. For each arithmetic prime κ:Fκ¯p\kappa:\mathcal{R}\rightarrow F_{\kappa}\subseteq\bar{\mathbb{Q}}_{p}, where FκF_{\kappa} is a finite extension of p\mathbb{Q}_{p}, the specialisation fκ=n1κ(𝐚n)qnFκ[[q]]f_{\kappa}=\sum_{n\geq 1}\kappa(\mathbf{a}_{n})q^{n}\in F_{\kappa}[[q]] of 𝐟\mathbf{f} at κ\kappa is a pp-ordinary cuspform of level Γ1(Npsκ)\Gamma_{1}(Np^{s_{\kappa}}), weight kκk_{\kappa} and character ψκ\psi_{\kappa}, for an integer sκ1s_{\kappa}\geq 1 and an integer kκ2k_{\kappa}\geq 2; see §3.1 for a more accurate exposition.

Let 𝐓\mathbf{T}^{\dagger} denote the self-dual twist of Hida’s big Galois representation attached to \mathcal{R}; therefore, 𝐓\mathbf{T}^{\dagger} is a free \mathcal{R}-module of rank 22, and for each arithmetic prime κ:Fκ\kappa:\mathcal{R}\rightarrow F_{\kappa}, the specialisation Vκ=𝐓,κFκV_{\kappa}^{\dagger}=\mathbf{T}^{\dagger}\otimes_{\mathcal{R},\kappa}F_{\kappa} at κ\kappa is isomorphic to the base change to FκF_{\kappa} of the self-dual twist of Deligne’s Galois representation attached to the modular form fκf_{\kappa}. Let 𝐀=𝐓\mathbf{A}^{\dagger}=\mathbf{T}^{\dagger}\otimes_{\mathcal{R}}\mathcal{R}^{\vee}. Here, ()(-)^{\vee} means the Pontryagin dual.

Let KK be a quadratic imaginary field of discriminant prime to NpNp, and write the factorisation N=N+NN=N^{+}N^{-} where a prime divisor \ell of NN divides N+N^{+} if and only if it is split in KK. Throughout the paper, we place ourselves in the definite setting; more precisely, we assume that

  • NN^{-} is a square-free product of an odd number of distinct primes; we also assume that pp is split in KK.

If fκf_{\kappa} has trivial character, the order of vanishing at s=k/2s=k/2 of the LL-series L(fκ/K,s)L(f_{\kappa}/K,s) of fκf_{\kappa} over KK is even by the assumption above on NN^{-}, and therefore, in light of an analogue in this setting of Greenberg’s conjecture, one expects that these LL-values do not generically vanish when the order of vanishing of the LL-series L(fκ,s)L(f_{\kappa},s) of fκf_{\kappa} over \mathbb{Q} is also even. As a consequence of the Tamagawa number conjecture of Bloch–Kato, one expects that the Bloch–Kato Selmer groups of VκV_{\kappa}^{\dagger} over KK is generically trivial in the same setting, so one expects that Nekovář’s extended Selmer group of 𝐀\mathbf{A}^{\dagger} over KK is a cotorsion \mathcal{R}-module. when the order of vanishing of the LL-series L(fκ,s)L(f_{\kappa},s) of fκf_{\kappa} over \mathbb{Q} is even. See [LV11, Conjecture 9.5] for a more detailed discussion of this heuristic.

Denote by KK_{\infty} the anticyclotomic p\mathbb{Z}_{p}-extension of KK, and denote Γ=Gal(K/K)p\Gamma_{\infty}=\operatorname{Gal}(K_{\infty}/K)\simeq\mathbb{Z}_{p}. In [LV11] a big theta element Θ(𝐟)\Theta_{\infty}(\mathbf{f}) in [[Γ]]\mathcal{R}[[\Gamma_{\infty}]] is constructed by means of compatible families of Gross points on towers of Shimura curves associated with the definite quaternion algebra BB ramified at all primes dividing NN^{-} and Eichler orders of increasing pp-power level; we also define the two-variable anticyclotomic pp-adic LL-function

Lp(𝐟/K)=Θ(𝐟)Θ(𝐟),L_{p}(\mathbf{f}/K)=\Theta_{\infty}(\mathbf{f})\cdot\Theta_{\infty}(\mathbf{f})^{*},

where xxx\mapsto x^{*} is the involution of [[Γ]]\mathcal{R}[[\Gamma_{\infty}]] defined by γγ1\gamma\mapsto\gamma^{-1} on group-like elements. The construction Θ(𝐟)\Theta_{\infty}(\mathbf{f}), Lp(𝐟/K)L_{p}(\mathbf{f}/K), and the notion of compatibility of Gross points on towers of Shimura curves, are reviewed in §3.3.

The element Lp(𝐟/K)L_{p}(\mathbf{f}/K) is the analogue of the pp-adic LL-function Lp(E/K)L_{p}(E/K) introduced by Bertolini–Darmon [BD96a], [BD96b] and [BD05] using Gross points on definite Shimura curves to study the arithmetic of elliptic curves over KK in a similar definite setting. In particular, if the conductor NN of EE admits the same factorisation N=N+NN=N^{+}N^{-} as above and pp is a prime of good ordinary reduction for EE which splits in KK, then, under mild technical hypotheses, it is known that the pp-adic LL-function Lp(E/K)L_{p}(E/K) is a non-trivial element of p[[Γ]]\mathbb{Z}_{p}[[\Gamma_{\infty}]], the Pontryagin dual of the Selmer group of the pp-power torsions of EE over KK_{\infty} is a torsion p[[Γ]]\mathbb{Z}_{p}[[\Gamma_{\infty}]]-module and its characteristic ideal of equal to the ideal generated by the pp-adic LL-function (see [BD05, Theorem 1] and [SU14, Theorem 3.37]). Therefore, it is natural to expect a similar main conjecture holds for families. In other words, if [[Γ]]\mathcal{R}[[\Gamma_{\infty}]] is the Iwasawa algebra of Γ\Gamma_{\infty} with coefficients in \mathcal{R}, then one expects that the two-variable pp-adic LL-function Lp(𝐟/K)L_{p}(\mathbf{f}/K) is a non-zero element of [[Γ]]\mathcal{R}[[\Gamma_{\infty}]], the Selmer group of 𝐀\mathbf{A}^{\dagger} over KK_{\infty} is a cotorsion [[Γ]]\mathcal{R}[[\Gamma_{\infty}]]-module, and its characteristic ideal is equal to the ideal generated by the pp-adic LL-function. See [LV11, §9.3] for a more detailed discussion of this topic. The proof of these assertions is one of the main result of this paper, from which we also deduce some results on the Selmer group of 𝐓\mathbf{T}^{\dagger} over KK.

Let H~f1(K,𝐀)\tilde{H}^{1}_{f}(K_{\infty},\mathbf{A}^{\dagger}) be Nekovář’s extended Selmer group of 𝐀\mathbf{A}^{\dagger} over KK_{\infty} and H~f1(K,𝐀)\tilde{H}^{1}_{f}(K_{\infty},\mathbf{A}^{\dagger})^{\vee} its Pontryagin dual, which are discrete and compact [[Γ]]\mathcal{R}[[\Gamma_{\infty}]]-modules, respectively. Before stating our main results, we fix our assumptions. We suppose that there exists an arithmetic prime κ0\kappa_{0} such that f0=fκ0=n1anqnSk0(Γ0(Np))f_{0}=f_{\kappa_{0}}=\sum_{n\geq 1}a_{n}q^{n}\in S_{k_{0}}(\Gamma_{0}(Np)) is an ordinary pp-stabilised newform of weight k02k_{0}\geq 2 with k02(modp1)k_{0}\equiv 2\pmod{p-1} and trivial nebentypus character. We denote ρ¯f0\bar{\rho}_{f_{0}} the residual representation attached to f0f_{0}.

Our first main result is the two-variable anticyclotomic Iwasawa main conjecture for Hida families, which proves [LV11, Conjecture 9.12].

Theorem 1.1.

We assume the following statements.

  • NN^{-} is a square-free product of an odd number of distinct primes.

  • The residual representation ρ¯f0\bar{\rho}_{f_{0}} is absolutely irreducible, pp-distinguished, and ramified at all primes N\ell\mid N^{-}.

  • pp is a non-anomalous prime for ρ¯f0\bar{\rho}_{f_{0}} when k=2k=2, i.e. ap(f0)±1a_{p}(f_{0})\not\equiv\pm 1 modulo the maximal ideal of 𝒪\mathcal{O}.

  • pp is split in KK.

Then H~f1(K,𝐀)\tilde{H}^{1}_{f}(K_{\infty},\mathbf{A}^{\dagger})^{\vee} is a cotorsion [[Γ]]\mathcal{R}[[\Gamma_{\infty}]]-module, and its characteristic ideal is equal to the ideal generated by the two-variable pp-adic LL-function Lp(𝐟/K)L_{p}(\mathbf{f}/K).

We also deduce a result on the arithmetic of 𝐟\mathbf{f} over KK, which is an definite analogue of the horizontal non-vanishing conjecture [LV11, Conjecture 9.5]. Define 𝒥0=χtriv(Θ(𝐟))\mathcal{J}_{0}=\chi_{\mathrm{triv}}\left(\Theta_{\infty}(\mathbf{f})\right), where χtriv\chi_{\mathrm{triv}} is the trivial character of Γ\Gamma_{\infty} and let H~f1(K,𝐓)\tilde{H}^{1}_{f}(K,\mathbf{T}^{\dagger}) denote Nekovář’s extended Selmer group of 𝐓\mathbf{T}^{\dagger} over KK_{\infty}.

Theorem 1.2.

Under the same assumptions in Theorem 1.1, if H~f1(K,𝐓)\tilde{H}^{1}_{f}(K,\mathbf{T}^{\dagger}) is a torsion \mathcal{R}-module, then 𝒥00\mathcal{J}_{0}\neq 0.

The proofs of these results are the combination of the following ingredients.

  • A control theorem for Selmer groups of Hida’s big Galois representations over the antiyclotomic p\mathbb{Z}_{p}-extension, similar to analogous results for the cyclotomic p\mathbb{Z}_{p}-extension by Ochiai [Och06], which we prove in §2.6 of this paper;

  • The results from [CL16], [CKL17] and [KL22] proving a close relation between Lp(𝐟/K)L_{p}(\mathbf{f}/K) and the self-dual twist of the specialisation to the anticyclotomic line of the three-variable pp-adic LL-functions of Skinner–Urban [SU14];

  • The three-variable Iwasawa main conjecture proved by Skinner–Urban [SU14].

As hinted at in the lines above, the proof of the three-variable main conjecture in [SU14] has a prominent role in our argument; however, the careful comparison of the two setting is required, for which we use the results from [CL16], [CKL17] and [KL22].

Acknowledgements

We thank Francesc Castella and Stefano Vigni for useful discussions. Kim was partially supported by a KIAS Individual Grant (SP054103) via the Center for Mathematical Challenges at Korea Institute for Advanced Study and by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2018R1C1B6007009). Longo was partially supported by PRIN 2017 Geometric, algebraic and analytic methods in arithmetic and INDAM GNSAGA.

2. Selmer groups over ordinary deformation rings and their control theorem

In this section, we first review Iwasawa algebras over complete noetherian regular local rings of Krull dimension 1\geq 1 and Selmer groups of ordinary Galois representations over such rings. Then we prove a general control theorem for these Selmer groups and relate them with classical Selmer groups via Shapiro’s lemma. This generality certainly includes the case of Hida deformations. The notation of this section is independent of the notation of the other sections of the paper. Some of the arguments are similar to those in [Och00], [Och01], and [Och06].

We first set some general conventions. Let RR be a complete noetherian regular local ring with maximal ideal 𝔪R\mathfrak{m}_{R}, of Krull dimension d1d\geq 1, with finite residue field k=R/𝔪RRk=R/\mathfrak{m}_{R}R of characteristic pp, a prime number. For any ideal IRI\subseteq R, and any RR-module MM, denote M[I]M[I] the II-torsion RR-submodule of MM and MIM_{I} the localization of MM at II. Denote

M=HomR(M,R)M^{*}=\operatorname{Hom}_{R}(M,R)

the RR-linear dual of MM (where HomR\operatorname{Hom}_{R} denotes RR-linear homomorphisms) and

M=Homcont(M,p/p)M^{\vee}=\operatorname{Hom}_{\mathrm{cont}}(M,\mathbb{Q}_{p}/\mathbb{Z}_{p})

the Pontryagin dual of MM (where Homcont\operatorname{Hom}_{\mathrm{cont}} denotes continuous group homomorphisms). By [Nek06, §2.9.1, §2.9.2],

M=D(M)=HomR(M,R)M^{\vee}=D(M)=\operatorname{Hom}_{R}(M,R^{\vee})

under our assumptions for any RR-module MM of finite type, hence compact, or any RR-module MM of cofinite type equipped with the discrete topology. Following [Nek06, §0.4], define

Φ(M)=MRR.\Phi(M)=M\otimes_{R}R^{\vee}.

In particular, (M)Φ(M)(M^{*})^{\vee}\simeq\Phi(M) and (Φ(M))M(\Phi(M))^{\vee}\simeq M^{*} for any RR-module MM of finite type ([Nek06, (0.4.4)]). Further, by basic properties of Pontryagin duality, (M[𝔭])M/𝔭M(M[\mathfrak{p}])^{\vee}\simeq M^{\vee}/\mathfrak{p}M^{\vee} and, if MM is a GG-module for some profinite group GG, we have (MG)(M)G(M^{G})^{\vee}\simeq(M^{\vee})_{G}.

2.1. Iwasawa algebras over regular local rings

Fix a complete noetherian regular local ring RR, with maximal ideal 𝔪R\mathfrak{m}_{R}, of Krull dimension d1d\geq 1, and finite residue field k=R/𝔪Rk=R/\mathfrak{m}_{R} of characteristic pp, a prime number. Let F/FF_{\infty}/F be a p\mathbb{Z}_{p}-extension of FF, unramified outside pp and totally ramified at pp, and define G=Gal(F/F)pG_{\infty}=\operatorname{Gal}(F_{\infty}/F)\simeq\mathbb{Z}_{p}. Let FnF_{n} be the subfield of FF_{\infty} such that Gn=Gal(Fn/F)/pnG_{n}=\operatorname{Gal}(F_{n}/F)\simeq\mathbb{Z}/p^{n}\mathbb{Z} and define

ΛR=R[[G]]=limnR[Gn].\Lambda_{R}=R[\![G_{\infty}]\!]=\mathop{\varprojlim}\limits_{n}R[G_{n}].

We recall briefly some properties of ΛR\Lambda_{R} and finitely generated ΛR\Lambda_{R}-modules. We begin with the following standard fact.

Lemma 2.1.

The ring ΛR\Lambda_{R} is isomorphic to the power series ring R[[X]]R[[X]] via the map which sends a topological generator γ\gamma of GG_{\infty} to X1X-1.

Since RR is a complete noetherian regular local ring, thanks to Lemma 2.1 we see that ΛR\Lambda_{R} is also a complete noetherian regular local ring with maximal ideal 𝔪ΛR=(𝔪R,γ1)\mathfrak{m}_{\Lambda_{R}}=(\mathfrak{m}_{R},\gamma-1) of ΛR\Lambda_{R} ([Mat89, Theorem 3.3, Exercise 8.6, Theorem 19.5]). In particular, since RR and ΛR\Lambda_{R} are regular local ring, they are also UFD by Auslander–Buchsbaum Theorem ([Mat89, Theorems 20.3 and 20.8]), and therefore every prime ideal of height 11 of RR and ΛR\Lambda_{R} is principal ([Mat89, Theorem 20.1]), and RR and ΛR\Lambda_{R} are integrally closed ([Mat89, §9, Example 1]).

Recall that a ΛR\Lambda_{R}-module XX is said to be pseudo-null if its support SuppΛR(X)\mathrm{Supp}_{\Lambda_{R}}(X) contains only prime ideals of height at least 22, and that two ΛR\Lambda_{R}-modules XX and YY are said to be pseudo-isomorphic if there exists an exact sequence

0AXYB00\longrightarrow A\longrightarrow X\longrightarrow Y\longrightarrow B\longrightarrow 0

where AA and BB are pseudo-null ΛR\Lambda_{R}-modules ([Bou98, Chapter VII, §4, no.4, Definitions 2 and 3]). Since ΛR\Lambda_{R} is noetherian and integrally closed, we see from [Bou98, Chapter VII, §4, no.4, Theorem 4] that every finitely generated ΛR\Lambda_{R}-module MM is pseudo-isomorphic to a ΛR\Lambda_{R}-module of the form T×QT\times Q, where TT is the maximal torsion ΛR\Lambda_{R}-submodule of MM and QQ is a free ΛR\Lambda_{R}-module. By [Bou98, Chapter VII, §4, no.4, Theorem 5], we know that TT is isomorphic to i=1tΛR/𝔭ini\oplus_{i=1}^{t}\Lambda_{R}/\mathfrak{p}_{i}^{n_{i}} for suitable height 1 prime ideals 𝔭i\mathfrak{p}_{i} of ΛR\Lambda_{R} and integers ni1n_{i}\geq 1; moreover, since every prime ideal of ΛR\Lambda_{R} is principal, there are prime (hence irreducible) elements giΛRg_{i}\in\Lambda_{R} such that Ti=1sΛR/giniΛRT\simeq\oplus_{i=1}^{s}\Lambda_{R}/g_{i}^{n_{i}}\Lambda_{R}. Define the characteristic ideal CharΛR(M)\operatorname{Char}_{\Lambda_{R}}(M) of MM to be 0 if Q0Q\neq 0 and

CharΛR(M)=(i=1sgini)\operatorname{Char}_{\Lambda_{R}}(M)=\left(\prod_{i=1}^{s}g_{i}^{n_{i}}\right)

otherwise.

Lemma 2.2.

Let QQ be a finitely generated ΛR\Lambda_{R}-module and 𝔭=(g)\mathfrak{p}=(g) a principal prime ideal of ΛR\Lambda_{R}. Assume that Q/𝔭QQ/\mathfrak{p}Q is pseudo-null. Then the 𝔭\mathfrak{p}-torsion ΛR\Lambda_{R}-submodule Q[𝔭]Q[\mathfrak{p}] of QQ is isomorphic to S[𝔭]S[\mathfrak{p}], where SS is the maximal pseudo-null ΛR\Lambda_{R}-submodule of QQ.

Proof.

The theory of ΛR\Lambda_{R}-modules recalled above shows the existence of an exact sequence

(1) 0SQM=U(i=1sΛR/giniΛR)B00\longrightarrow S\longrightarrow Q\longrightarrow M=U\oplus\left(\bigoplus_{i=1}^{s}\Lambda_{R}/g_{i}^{n_{i}}\Lambda_{R}\right)\longrightarrow B\longrightarrow 0

where giΛRg_{i}\in\Lambda_{R} are irreducible elements, ni1n_{i}\geq 1 are integers, UU is free over ΛR\Lambda_{R}, and SS and BB are pseudo-null. It suffices to show that the multiplication by g map is injective on MM.

Since UU is torsion-free, the multiplication by gg map is injective on UU.

We now make the following observation. Suppose that gigg_{i}\mid g for some ii. Then 𝔭=(g)=(gi)\mathfrak{p}=(g)=(g_{i}) since gg is irreducible. This, the quotient ring ΛR/(g,gini)ΛR\Lambda_{R}/(g,g_{i}^{n_{i}})\Lambda_{R} is isomorphic to ΛR/𝔭ΛR\Lambda_{R}/\mathfrak{p}\Lambda_{R}, which is not pseudo-null, and therefore M/𝔭MM/\mathfrak{p}M is not pseudo-null. Since subquotients of pseudo-null ΛR\Lambda_{R}-modules are again pseudo-null, from (1) we have a pseudo-isomorphism between the pseudo-null ΛR\Lambda_{R}-module Q/𝔭QQ/\mathfrak{p}Q and the ΛR\Lambda_{R}-module M/𝔭MM/\mathfrak{p}M which is not pseudo-null. Hence, gigg_{i}\nmid g for every ii under our assumption.

We now study the multiplication by gg map on the torsion ΛR\Lambda_{R}-submodule of MM. Suppose that g[m]=0g\cdot[m]=0 for some class [m]ΛR/giniΛR[m]\in\Lambda_{R}/g_{i}^{n_{i}}\Lambda_{R}, where mΛRm\in\Lambda_{R}. Then gmg\cdot m belongs to ginig_{i}^{n_{i}}. Since ginigmg_{i}^{n_{i}}\mid g\cdot m and gigg_{i}\nmid g, we conclude that ginimg_{i}^{n_{i}}\mid m, so [m]=0[m]=0. Thus the multiplication by gg map is injective on ΛR/giniΛR\Lambda_{R}/g_{i}^{n_{i}}\Lambda_{R}. We conclude that the multiplication by gg map M×gMM\overset{\times g}{\rightarrow}M is injective, and therefore the 𝔭\mathfrak{p}-torsion ΛR\Lambda_{R}-submodule Q[𝔭]Q[\mathfrak{p}] of QQ is isomorphic to the 𝔭\mathfrak{p}-torsion ΛR\Lambda_{R}-submodule of SS, as was to be shown. ∎

2.2. Selmer groups over Iwasawa algebras

Let FF be an algebraic number field. For each place vv of FF, denote FvF_{v} the completion of FF at vv and 𝒪Fv\mathcal{O}_{F_{v}} the valuation ring of FvF_{v}. Define GF=Gal(F¯/F)G_{F}=\operatorname{Gal}(\bar{F}/F) and GFv=Gal(F¯v/Fv)G_{F_{v}}=\operatorname{Gal}(\bar{F}_{v}/F_{v}). Let IFvI_{F_{v}} the inertia subgroup of GFvG_{F_{v}}. We will also write 𝒪v=𝒪Fv\mathcal{O}_{v}=\mathcal{O}_{F_{v}}, Iv=IFvI_{v}=I_{F_{v}} and Gv=GFvG_{v}=G_{F_{v}} when the fields involved are clear from the context. Recall that F/FF_{\infty}/F is a fixed p\mathbb{Z}_{p}-extension of FF, unramified outside pp and totally ramified at pp, G=Gal(F/F)G_{\infty}=\operatorname{Gal}(F_{\infty}/F) and FnF_{n} is the subfield of FF_{\infty} such that Gn=Gal(Fn/F)/pnG_{n}=\operatorname{Gal}(F_{n}/F)\simeq\mathbb{Z}/p^{n}\mathbb{Z}; finally, recall that ΛR=R[[G]]\Lambda_{R}=R[\![G_{\infty}]\!].

Let 𝐓\mathbf{T} be finite free ΛR\Lambda_{R}-module equipped with a continuous action of GFG_{F}, and fix a prime number pp a prime number. Let Σp\Sigma_{p} denote the set of places of FF dividing pp, and let Σ\Sigma be a finite set of places of FF containing Σp\Sigma_{p}. We assume that 𝐓\mathbf{T} is unramified outside Σ\Sigma. Moreover, for each vpv\mid p a prime of the ring of integers 𝒪\mathcal{O} of FF, we suppose given a filtration

(2) 0Fv+(𝐓)𝐓Fv(𝐓)00\longrightarrow F_{v}^{+}(\mathbf{T})\longrightarrow\mathbf{T}\longrightarrow F^{-}_{v}(\mathbf{T})\longrightarrow 0

of Gv=Gal(F¯v/Fv)G_{v}=\operatorname{Gal}(\bar{F}_{v}/F_{v})-modules.

Remark 2.3.

For the moment, we do not impose any condition to the filtration (2), but of course the structure of the Selmer group defined below depends on this choice. In the applications, the filtration (2) is made of ΛR\Lambda_{R}-modules Fv+(𝐓)F_{v}^{+}(\mathbf{T}) and Fv(𝐓)F^{-}_{v}(\mathbf{T}) which are both free of rank 11, and the Galois action on each of them is characterised by a pair of characters, one unramified and the other factorising through the cyclotomic p\mathbb{Z}_{p}-extension of FF. See §2.4 for details.

Taking Φ\Phi (i.e. tensoring over ΛR\Lambda_{R} with ΛR\Lambda_{R}^{\vee}) we also get a filtration

0Fv+(𝐀)𝐀Fv(𝐀)0.0\longrightarrow F_{v}^{+}(\mathbf{A})\longrightarrow\mathbf{A}\longrightarrow F^{-}_{v}(\mathbf{A})\longrightarrow 0.

Define the Greenberg Selmer group of 𝐀\mathbf{A} (relative to the chosen filtrations (2)) by

Sel(F,𝐀)=ker(H1(F,𝐀)vΣpH1(Iv,𝐀)×vΣpH1(Iv,𝐀/Fv+(𝐀)))\operatorname{Sel}(F,\mathbf{A})=\ker\left(H^{1}(F,\mathbf{A})\longrightarrow\prod_{v\not\in\Sigma_{p}}H^{1}(I_{v},\mathbf{A})\times\prod_{v\in\Sigma_{p}}H^{1}(I_{v},\mathbf{A}/F^{+}_{v}(\mathbf{A}))\right)

and the strict Greenberg Selmer group of 𝐀\mathbf{A} (relative to the chosen filtrations (2)) by

Selstr(F,𝐀)=ker(H1(F,𝐀)vΣpH1(Iv,𝐀)×vΣpH1(Fv,𝐀/Fv+(𝐀)))\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A})=\ker\left(H^{1}(F,\mathbf{A})\longrightarrow\prod_{v\not\in\Sigma_{p}}H^{1}(I_{v},\mathbf{A})\times\prod_{v\in\Sigma_{p}}H^{1}(F_{v},\mathbf{A}/F^{+}_{v}(\mathbf{A}))\right)

where IvI_{v} is the inertia subgroup of GvG_{v}.

Let 𝔮=(g)ΛR\mathfrak{q}=(g)\subseteq\Lambda_{R} be a principal ideal and assume that ΛR/𝔮ΛR\Lambda_{R}/\mathfrak{q}\Lambda_{R} is finite and flat over RR. Since ΛR/𝔮ΛR\Lambda_{R}/\mathfrak{q}\Lambda_{R} is flat over RR, tensoring over RR with ΛR/𝔮ΛR\Lambda_{R}/\mathfrak{q}\Lambda_{R} we also have a filtration

0Fv+(𝐓/𝔮𝐓)𝐓/𝔮𝐓Fv(𝐓/𝔮𝐓)00\longrightarrow F_{v}^{+}(\mathbf{T}/\mathfrak{q}\mathbf{T})\longrightarrow\mathbf{T}/\mathfrak{q}\mathbf{T}\longrightarrow F^{-}_{v}(\mathbf{T}/\mathfrak{q}\mathbf{T})\longrightarrow 0

where 𝐓/𝔮𝐓=𝐓ΛRΛR/𝔮ΛR=TRΛR/𝔮ΛR\mathbf{T}/\mathfrak{q}\mathbf{T}=\mathbf{T}\otimes_{\Lambda_{R}}\Lambda_{R}/\mathfrak{q}\Lambda_{R}=T\otimes_{R}\Lambda_{R}/\mathfrak{q}\Lambda_{R}. We also have a filtration

0Fv+(𝐀[𝔮])𝐀[𝔮]Fv(𝐀[𝔮])00\longrightarrow F_{v}^{+}(\mathbf{A}[\mathfrak{q}])\longrightarrow\mathbf{A}[\mathfrak{q}]\longrightarrow F^{-}_{v}(\mathbf{A}[\mathfrak{q}])\longrightarrow 0

where Fv+(𝐀[𝔮])=A[𝔮]Fv+(𝐀)F_{v}^{+}(\mathbf{A}[\mathfrak{q}])=A[\mathfrak{q}]\cap F^{+}_{v}(\mathbf{A}). Define the Greenberg Selmer group of 𝐀[𝔮]\mathbf{A}[\mathfrak{q}] (relative to the chosen filtrations (2)) by

Sel(F,𝐀[𝔮])=ker(H1(F,𝐀[𝔮])vΣpH1(Iv,𝐀[𝔮])×vΣpH1(Iv,𝐀[𝔮]/Fv+(𝐀[𝔮])))\operatorname{Sel}(F,\mathbf{A}[\mathfrak{q}])=\ker\left(H^{1}(F,\mathbf{A}[\mathfrak{q}])\longrightarrow\prod_{v\not\in\Sigma_{p}}H^{1}(I_{v},\mathbf{A}[\mathfrak{q}])\times\prod_{v\in\Sigma_{p}}H^{1}(I_{v},\mathbf{A}[\mathfrak{q}]/F^{+}_{v}(\mathbf{A}[\mathfrak{q}]))\right)

and the strict Greenberg Selmer group of 𝐀\mathbf{A} (relative to the chosen filtrations (2)) by

Selstr(F,𝐀[𝔮])=ker(H1(F,𝐀[𝔮])vΣpH1(Iv,𝐀[𝔮])×vΣpH1(Fv,𝐀/Fv+(𝐀[𝔮]))).\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}[\mathfrak{q}])=\ker\left(H^{1}(F,\mathbf{A}[\mathfrak{q}])\longrightarrow\prod_{v\not\in\Sigma_{p}}H^{1}(I_{v},\mathbf{A}[\mathfrak{q}])\times\prod_{v\in\Sigma_{p}}H^{1}(F_{v},\mathbf{A}/F^{+}_{v}(\mathbf{A}[\mathfrak{q}]))\right).

2.3. The control theorem

Let the notation be as in §2.2. Let 𝔮=(g)ΛR\mathfrak{q}=(g)\subseteq\Lambda_{R} be a principal ideal and assume that ΛR/𝔮ΛR\Lambda_{R}/\mathfrak{q}\Lambda_{R} is finite and flat over RR. Then we have canonical maps

r𝔮:Sel(F,𝐀[𝔮])Sel(F,𝐀)[𝔮],r_{\mathfrak{q}}:\operatorname{Sel}(F,\mathbf{A}[\mathfrak{q}])\longrightarrow\operatorname{Sel}(F,\mathbf{A})[\mathfrak{q}],
r𝔮str:Selstr(F,𝐀[𝔮])Selstr(F,𝐀)[𝔮].r_{\mathfrak{q}}^{\mathrm{str}}:\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}[\mathfrak{q}])\longrightarrow\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A})[\mathfrak{q}].
Proposition 2.4.

Assume that H0(F,𝐀[𝔮])H^{0}(F,\mathbf{A}[\mathfrak{q}])^{\vee} is a pseudo-null ΛR\Lambda_{R}-module. Then ker(r𝔮)\mathrm{ker}(r_{\mathfrak{q}})^{\vee} and ker(r𝔮str)\mathrm{ker}(r_{\mathfrak{q}}^{\mathrm{str}})^{\vee} are also pseudo-null ΛR\Lambda_{R}-modules, and are contained in the 𝔮\mathfrak{q}-torsion subgroup of the maximal pseudo-null ΛR\Lambda_{R}-submodule of (𝐓)GF(\mathbf{T}^{*})_{G_{F}}.

Proof.

We explain the argument only for r𝔮r_{\mathfrak{q}}; the case of r𝔮strr_{\mathfrak{q}}^{\mathrm{str}} is verbatim.

We have the following commutative diagram:

Sel(F,𝐀[𝔮])\textstyle{\operatorname{Sel}(F,\mathbf{A}[\mathfrak{q}])\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}r𝔮\scriptstyle{r_{\mathfrak{q}}}Sel(F,𝐀)[𝔮]\textstyle{\operatorname{Sel}(F,\mathbf{A})[\mathfrak{q}]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H0(F,𝐀)/𝔮H0(F,𝐀)\textstyle{H^{0}(F,\mathbf{A})/\mathfrak{q}H^{0}(F,\mathbf{A})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H1(F,𝐀[𝔮])\textstyle{H^{1}(F,\mathbf{A}[\mathfrak{q}])\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H1(F,𝐀)[𝔮]\textstyle{H^{1}(F,\mathbf{A})[\mathfrak{q}]}

therefore it is enough to show that H0(F,𝐀)/𝔮H0(F,𝐀)H^{0}(F,\mathbf{A})/\mathfrak{q}H^{0}(F,\mathbf{A}) is a pseudo-null ΛR\Lambda_{R}-module, and that it is contained in the 𝔮\mathfrak{q}-torsion subgroup of the maximal pseudo-null ΛR\Lambda_{R}-submodule of (𝐓)GF(\mathbf{T}^{*})_{G_{F}}.

Note that H0(F,𝐀[𝔮])=H0(F,𝐀)[𝔮]H^{0}(F,\mathbf{A}[\mathfrak{q}])=H^{0}(F,\mathbf{A})[\mathfrak{q}] is the Pontryagin dual of (𝐓)GF/𝔮(𝐓)GF(\mathbf{T}^{*})_{G_{F}}/\mathfrak{q}(\mathbf{T}^{*})_{G_{F}}, and that H0(F,𝐀)/𝔮H0(F,𝐀)H^{0}(F,\mathbf{A})/\mathfrak{q}H^{0}(F,\mathbf{A}) is the Pontryagin dual of (𝐓)GF[𝔮](\mathbf{T}^{*})_{G_{F}}[\mathfrak{q}]. Since H0(F,𝐀[𝔮])H^{0}(F,\mathbf{A}[\mathfrak{q}])^{\vee} is a pseudo-null ΛR\Lambda_{R}-module by assumption, applying Lemma 2.2 to the ΛR\Lambda_{R}-module (𝐓)GF(\mathbf{T}^{*})_{G_{F}} we see that H0(F,𝐀)/𝔮H0(F,𝐀)H^{0}(F,\mathbf{A})/\mathfrak{q}H^{0}(F,\mathbf{A}) has also pseudo-null Pontryagin dual, contained in the 𝔮\mathfrak{q}-torsion subgroup of the maximal pseudo-null ΛR\Lambda_{R}-submodule of (𝐓)GF(\mathbf{T}^{*})_{G_{F}}. ∎

For vΣv\in\Sigma, define

Cv={ΛR-torsion submodule of the module ((𝐓/Fv+(𝐓)))Iv if vΣp,ΛR-torsion submodule of the module (𝐓)Iv if vΣΣp.C_{v}=\begin{cases}\text{$\Lambda_{R}$-torsion submodule of the module $((\mathbf{T}/F_{v}^{+}(\mathbf{T}))^{*})_{I_{v}}$ if $v\in\Sigma_{p}$},\\ \text{$\Lambda_{R}$-torsion submodule of the module $(\mathbf{T}^{*})_{I_{v}}$ if $v\in\Sigma-\Sigma_{p}$.}\end{cases}
Cvstr={ΛR-torsion submodule of the module ((𝐓/Fv+(𝐓)))Gv if vΣp,ΛR-torsion submodule of the module (𝐓)Iv if vΣΣp.C_{v}^{\mathrm{str}}=\begin{cases}\text{$\Lambda_{R}$-torsion submodule of the module $((\mathbf{T}/F_{v}^{+}(\mathbf{T}))^{*})_{G_{v}}$ if $v\in\Sigma_{p}$},\\ \text{$\Lambda_{R}$-torsion submodule of the module $(\mathbf{T}^{*})_{I_{v}}$ if $v\in\Sigma-\Sigma_{p}$.}\end{cases}

Denote FΣF_{\Sigma} the maximal extension of FF which is unramified outside Σ\Sigma.

Proposition 2.5.

Assume that

  • The ΛR\Lambda_{R}-module Cv/𝔮CvC_{v}/\mathfrak{q}C_{v} (Cvstr/𝔮CvstrC_{v}^{\mathrm{str}}/\mathfrak{q}C_{v}^{\mathrm{str}}, respectively) is pseudo-null for each vΣv\in\Sigma;

  • H0(FΣ/F,𝐀[𝔮])H^{0}(F_{\Sigma}/F,\mathbf{A}[\mathfrak{q}])^{\vee} is pseudo-null.

Then coker(r𝔮)\mathrm{coker}(r_{\mathfrak{q}})^{\vee} (coker(r𝔮str)\mathrm{coker}(r_{\mathfrak{q}}^{\mathrm{str}})^{\vee}, respectively) is a pseudo-null ΛR\Lambda_{R}-module.

Proof.

We do the proof only for r𝔮r_{\mathfrak{q}}; the case of r𝔮strr_{\mathfrak{q}}^{\mathrm{str}} is verbatim.

Recall that

H1(FΣ/F,𝐀)=ker(H1(F,𝐀)vΣH1(Iv,𝐀))H^{1}(F_{\Sigma}/F,\mathbf{A})=\ker\left(H^{1}(F,\mathbf{A})\longrightarrow\prod_{v\not\in\Sigma}H^{1}(I_{v},\mathbf{A})\right)

as a submodule of H1(F,𝐀)H^{1}(F,\mathbf{A}). It follows that there exists a commutative diagram:

0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Sel(F,𝐀[𝔮])\textstyle{\operatorname{Sel}(F,\mathbf{A}[\mathfrak{q}])\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}r𝔮\scriptstyle{r_{\mathfrak{q}}}H1(FΣ/F,𝐀[𝔮])\textstyle{H^{1}(F_{\Sigma}/F,\mathbf{A}[\mathfrak{q}])\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}γ𝔮\scriptstyle{\gamma_{\mathfrak{q}}}s𝔮\scriptstyle{s_{\mathfrak{q}}}pΣpH1(Iv,(𝐀/Fv+(𝐀))[𝔮])×vΣΣpH1(Fv,𝐀[𝔮])\textstyle{\prod_{p\in\Sigma_{p}}H^{1}(I_{v},(\mathbf{A}/F^{+}_{v}(\mathbf{A}))[\mathfrak{q}])\times\prod_{v\in\Sigma-\Sigma_{p}}H^{1}(F_{v},\mathbf{A}[\mathfrak{q}])\ignorespaces\ignorespaces\ignorespaces\ignorespaces}t𝔮\scriptstyle{t_{\mathfrak{q}}}0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Sel(F,𝐀)[𝔮]\textstyle{\operatorname{Sel}(F,\mathbf{A})[\mathfrak{q}]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H1(FΣ/F,𝐀)[𝔮]\textstyle{H^{1}(F_{\Sigma}/F,\mathbf{A})[\mathfrak{q}]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}pΣpH1(Iv,(𝐀/Fv+(𝐀)))[𝔮]×vΣΣpH1(Fv,𝐀)[𝔮]\textstyle{\prod_{p\in\Sigma_{p}}H^{1}(I_{v},(\mathbf{A}/F^{+}_{v}(\mathbf{A})))[\mathfrak{q}]\times\prod_{v\in\Sigma-\Sigma_{p}}H^{1}(F_{v},\mathbf{A})[\mathfrak{q}]}

where the vertical arrows are restriction maps. The multiplication by gg map induces an exact sequence

0𝐀[𝔮]𝐀𝑔𝔮𝐀00\longrightarrow\mathbf{A}[\mathfrak{q}]\longrightarrow\mathbf{A}\overset{g}{\longrightarrow}\mathfrak{q}\mathbf{A}\longrightarrow 0

which shows that map s𝔮s_{\mathfrak{q}} is surjective. Therefore by the snake lemma the cokernel of r𝔮r_{\mathfrak{q}} is a subquotient of the kernel of t𝔮t_{\mathfrak{q}}. Therefore, it is enough to show that the Pontryagin dual ker(t𝔮)\ker(t_{\mathfrak{q}})^{\vee} of ker(t𝔮)\ker(t_{\mathfrak{q}}) is pseudo-null. The module ker(t𝔮)\ker(t_{\mathfrak{q}}) is isomorphic to

ker(t𝔞)\displaystyle\ker(t_{\mathfrak{a}}) coker(pΣpH0(Iv,𝐀/Fv+(𝐀))×vΣΣpH0(Fv,𝐀)𝑔pΣpH0(Iv,𝐀/Fv+(𝐀))×vΣΣpH0(Fv,𝐀))\displaystyle\simeq\mathrm{coker}\left(\prod_{p\in\Sigma_{p}}H^{0}(I_{v},\mathbf{A}/F^{+}_{v}(\mathbf{A}))\times\prod_{v\in\Sigma-\Sigma_{p}}H^{0}(F_{v},\mathbf{A})\overset{g}{\longrightarrow}\prod_{p\in\Sigma_{p}}H^{0}(I_{v},\mathbf{A}/F^{+}_{v}(\mathbf{A}))\times\prod_{v\in\Sigma-\Sigma_{p}}H^{0}(F_{v},\mathbf{A})\right)
ker(vΣΣp(𝐓)Iv×vΣp((𝐓/Fv+(𝐓)))Gv𝑔vΣp((𝐓/Fv+(𝐓)))Gv×vΣΣp(𝐓)Iv).\displaystyle\simeq\ker\left(\prod_{v\in\Sigma-\Sigma_{p}}(\mathbf{T}^{*})_{I_{v}}\times\prod_{v\in\Sigma_{p}}((\mathbf{T}/F^{+}_{v}(\mathbf{T}))^{*})_{G_{v}}\overset{g}{\longrightarrow}\prod_{v\in\Sigma_{p}}((\mathbf{T}/F^{+}_{v}(\mathbf{T}))^{*})_{G_{v}}\times\prod_{v\in\Sigma-\Sigma_{p}}(\mathbf{T}^{*})_{I_{v}}\right)^{\vee}.

Hence, the module ker(t𝔮)\ker(t_{\mathfrak{q}}) is equal to (vΣCv[𝔮])\left(\oplus_{v\in\Sigma}C_{v}[\mathfrak{q}]\right)^{\vee}, by definition. On the other hand, vΣCv/𝔮Cv\oplus_{v\in\Sigma}C_{v}/\mathfrak{q}C_{v} is pseudo-null by assumption, and therefore Lemma 2.2 applied to the module vΣCv[𝔮]\oplus_{v\in\Sigma}C_{v}[\mathfrak{q}] completes the proof. ∎

Theorem 2.6.

Let 𝔮=(g)\mathfrak{q}=(g) be a principal ideal of ΛR\Lambda_{R}. Assume that H0(FΣ/F,𝐀[𝔮])H^{0}(F_{\Sigma}/F,\mathbf{A}[\mathfrak{q}])^{\vee} is pseudo-null and that the ΛR\Lambda_{R}-module Cv/𝔮CvC_{v}/\mathfrak{q}C_{v} ( Cvstr/𝔮CvstrC_{v}^{\mathrm{str}}/\mathfrak{q}C_{v}^{\mathrm{str}}, respectively) is pseudo-null for each vΣv\in\Sigma. Then ker(r𝔮)\mathrm{ker}(r_{\mathfrak{q}})^{\vee} and coker(r𝔮)\mathrm{coker}(r_{\mathfrak{q}})^{\vee} (ker(r𝔮str)\mathrm{ker}(r_{\mathfrak{q}}^{\mathrm{str}})^{\vee} and coker(r𝔮str)\mathrm{coker}(r_{\mathfrak{q}}^{\mathrm{str}})^{\vee}, respectively) are pseudo-null ΛR\Lambda_{R}-modules.

Proof.

Observe that if H0(FΣ/F,𝐀[𝔮])H^{0}(F_{\Sigma}/F,\mathbf{A}[\mathfrak{q}])^{\vee} is pseudo-null the same is true for H0(F,𝐀[𝔮])H^{0}(F,\mathbf{A}[\mathfrak{q}])^{\vee}. The result then follows combining Proposition 2.4 and Proposition 2.5.∎

2.4. Shapiro’s Lemma

Let the notation be as in §2.2; thus, FF is a number field and F/FF_{\infty}/F is a p\mathbb{Z}_{p}-extension, with finite layers FnF_{n}, totally ramified at pp and unramified outside pp. Let TT be a finite free RR-module equipped with a continuous action of GF=Gal(F¯/F)G_{F}=\operatorname{Gal}(\bar{F}/F) and fix a filtration

(3) 0Fv+(T)TFv(T)00\longrightarrow F_{v}^{+}(T)\longrightarrow T\longrightarrow F^{-}_{v}(T)\longrightarrow 0

of Gv=Gal(F¯v/Fv)G_{v}=\operatorname{Gal}(\bar{F}_{v}/F_{v})-modules, where FvF_{v} is the completion of FF at vv. Denote Fv(μp)/FvF_{v}(\mu_{p^{\infty}})/F_{v} be the cyclotomic extension of FvF_{v}, where μp\mu_{p^{\infty}} is the pp-divisible group of roots of unity in F¯v\bar{F}_{v}. Let Σp\Sigma_{p} denote the set of places of FF dividing pp, and let Σ\Sigma be a finite set of places of FF containing Σp\Sigma_{p}; denote FΣ/FF_{\Sigma}/F the maximal extension of FF which is unramified outside Σ\Sigma.

Assumption 2.7.

We suppose that the following conditions are satisfied.

  1. (1)

    TT is unramified outside Σ\Sigma.

  2. (2)

    H0(FΣ/Fn,A)H^{0}(F_{\Sigma}/F_{n},A)^{\vee} is pseudo-null.

  3. (3)

    Both F+(T)F^{+}(T) and F(T)F^{-}(T) are free RR-modules.

  4. (4)

    For each vpv\mid p, there are characters δv,θv:GvR×\delta_{v},\theta_{v}:G_{v}\rightarrow R^{\times} such that

    • δv\delta_{v} is unramified and takes the Frobenius Frobv\operatorname{Frob}_{v} to δv(Frobv)=uv\delta_{v}(\operatorname{Frob}_{v})=u_{v} with uv1u_{v}\not\equiv 1 modulo the maximal ideal 𝔪R\mathfrak{m}_{R} of RR.

    • θv\theta_{v} factors through GvGal(Fv(μp)/Fv)G_{v}\rightarrow\operatorname{Gal}(F_{v}(\mu_{p^{\infty}})/F_{v}).

    • GvG_{v} acts on Fv(T)F^{-}_{v}(T) via multiplication by the product δvθv\delta_{v}\cdot\theta_{v}.

Define

A=Φ(T)=TRR.A=\Phi(T)=T\otimes_{R}R^{\vee}.

The filtration Fv+(T)TF^{+}_{v}(T)\subseteq T induces a filtration Fv+(A)AF^{+}_{v}(A)\subseteq A of AA. For each integer n0n\geq 0 and any prime ideal vv of FnF_{n}, let Fn,vF_{n,v} be the completion of FnF_{n} at vv. Denote Σn,p\Sigma_{n,p} the set of places of Fn,vF_{n,v} above pp and define

Selstr(Fn,A)=ker(H1(Fn,A)vΣn,pH1(In,v,A)×vΣn,pH1(Fn,v,A/Fv+(A)))\operatorname{Sel}_{\mathrm{str}}(F_{n},A)=\ker\left(H^{1}(F_{n},A)\longrightarrow\prod_{v\not\in\Sigma_{n,p}}H^{1}(I_{n,v},A)\times\prod_{v\in\Sigma_{n,p}}H^{1}(F_{n,v},A/F^{+}_{v}(A))\right)

and

Selstr(F,A)=limnSelstr(Fn,A).\operatorname{Sel}_{\mathrm{str}}(F_{\infty},A)=\mathop{\varinjlim}\limits_{n}\operatorname{Sel}_{\mathrm{str}}(F_{n},A).

For any character χ:GB×\chi:G_{\infty}\rightarrow B^{\times}, where BB is a ring, and any BB-module MM, let M(χ)M(\chi) denote the BB-module MM equipped with GG_{\infty}-action given by gm=χ(g)mg\cdot m=\chi(g)m. Let κ:GΛR×\kappa:G_{\infty}\rightarrow\Lambda_{R}^{\times} be the tautological character. Note in particular that ΛR(κ)\Lambda_{R}(\kappa) is just ΛR\Lambda_{R} as ΛR\Lambda_{R}-module, but we prefer to keep the notation ΛR(κ)\Lambda_{R}(\kappa) to stress that we are considering ΛR\Lambda_{R} as a ΛR\Lambda_{R}-module and not as a ring. Define the ΛR\Lambda_{R}-module

𝐓=TRΛR(κ1).\mathbf{T}=T\otimes_{R}\Lambda_{R}(\kappa^{-1}).

Since the extension of rings ΛR/R\Lambda_{R}/R is flat (by Lemma 2.1 and [Mat89, Exercise 7.4]) then, tensoring (2) over RR with ΛR\Lambda_{R} we also have a filtration

0Fv+(𝐓)𝐓Fv(𝐓)00\longrightarrow F_{v}^{+}(\mathbf{T})\longrightarrow\mathbf{T}\longrightarrow F^{-}_{v}(\mathbf{T})\longrightarrow 0

where Fv±(𝐓)=Fv±(T)RΛR(κ1)F^{\pm}_{v}(\mathbf{T})=F^{\pm}_{v}(T)\otimes_{R}\Lambda_{R}(\kappa^{-1}). Define

𝐀=Φ(𝐓)=𝐓ΛRΛR.\mathbf{A}=\Phi(\mathbf{T})=\mathbf{T}\otimes_{\Lambda_{R}}\Lambda_{R}^{\vee}.

We observe that (cf. [Nek06, §2.9.1])

ΛR=Homcont(ΛR,p/p)HomR(ΛR,R).\Lambda_{R}^{\vee}=\operatorname{Hom}_{\mathrm{cont}}(\Lambda_{R},\mathbb{Q}_{p}/\mathbb{Z}_{p})\simeq\operatorname{Hom}_{R}(\Lambda_{R},R^{\vee}).

Moreover, it we stress the structure of ΛR\Lambda_{R}-modules, we have

ΛR(κ)HomR(ΛR(κ1),R),ΛR(κ1)HomR(ΛR(κ),R).\Lambda_{R}^{\vee}(\kappa)\simeq\operatorname{Hom}_{R}(\Lambda_{R}(\kappa^{-1}),R^{\vee}),\qquad\Lambda_{R}^{\vee}(\kappa^{-1})\simeq\operatorname{Hom}_{R}(\Lambda_{R}(\kappa),R^{\vee}).

where we use the standard action of ΛR\Lambda_{R} on HomR(ΛR,R)\operatorname{Hom}_{R}(\Lambda_{R},R^{\vee}) given by (λφ)(x)=φ(λ1x)(\lambda\cdot\varphi)(x)=\varphi(\lambda^{-1}x) for λΛR\lambda\in\Lambda_{R} and φHom(ΛR,R)\varphi\in\operatorname{Hom}(\Lambda_{R},R^{\vee}).

Note that, since TT is a free RR-module, we have isomorphisms of ΛR\Lambda_{R}-modules:

𝐀=𝐓ΛRΛR=(TRΛR(κ1))ΛRHomcont(ΛR(κ1),p/p)=(TRΛR(κ1))ΛRHomR(ΛR(κ1),R)=TRHomR(ΛR,R)=HomR(ΛR,A)\begin{split}\mathbf{A}&=\mathbf{T}\otimes_{\Lambda_{R}}\Lambda_{R}^{\vee}\\ &=(T\otimes_{R}\Lambda_{R}(\kappa^{-1}))\otimes_{\Lambda_{R}}\operatorname{Hom}_{\mathrm{cont}}(\Lambda_{R}(\kappa^{-1}),\mathbb{Q}_{p}/\mathbb{Z}_{p})\\ &=(T\otimes_{R}\Lambda_{R}(\kappa^{-1}))\otimes_{\Lambda_{R}}\operatorname{Hom}_{R}(\Lambda_{R}(\kappa^{-1}),R^{\vee})\\ &=T\otimes_{R}\operatorname{Hom}_{R}(\Lambda_{R},R^{\vee})\\ &=\operatorname{Hom}_{R}(\Lambda_{R},A)\end{split}

We now concentrate on ideals 𝔮n\mathfrak{q}_{n} generated by elements ωn=γpn1\omega_{n}=\gamma^{p^{n}}-1:

𝔮n=(ωn)=(γpn1),\mathfrak{q}_{n}=(\omega_{n})=(\gamma^{p^{n}}-1),

where γ\gamma is a topological generator of GG_{\infty}. We have isomorphisms of ΛR/𝔮nΛR\Lambda_{R}/\mathfrak{q}_{n}\Lambda_{R}-modules

𝐀[𝔮n]=HomR(ΛR(κ),A)[𝔮n]=HomR(ΛR(κ)/𝔮nΛR(κ),A)HomR(R[Gn],A)\begin{split}\mathbf{A}[\mathfrak{q}_{n}]&=\operatorname{Hom}_{R}(\Lambda_{R}(\kappa),A)[\mathfrak{q}_{n}]\\ &=\operatorname{Hom}_{R}(\Lambda_{R}(\kappa)/\mathfrak{q}_{n}\Lambda_{R}(\kappa),A)\\ &\simeq\operatorname{Hom}_{R}(R[G_{n}],A)\end{split}
Lemma 2.8.

For each integer n0n\geq 0 we have Selstr(F,𝐀[𝔮n])Selstr(Fn,A)\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}[\mathfrak{q}_{n}])\simeq\operatorname{Sel}_{\mathrm{str}}(F_{n},A). Moreover, we have Selstr(F,𝐀)Selstr(F,A)\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A})\simeq\operatorname{Sel}_{\mathrm{str}}(F_{\infty},A).

Proof.

Shapiro’s Lemma shows the the first of the following isomorphism

H1(Fn,A)H1(F,Hom(R[Gn],A))H1(F,𝐀[𝔮n]),H^{1}(F_{n},A)\simeq H^{1}(F,\operatorname{Hom}(R[G_{n}],A))\simeq H^{1}(F,\mathbf{A}[\mathfrak{q}_{n}]),

while the second follows from the previous discussion. Taking direct limits over nn, we also see that

H1(F,A)=limnH1(Fn,A)limnH1(F,HomR(R[Gn],A))H1(F,limnHomR(R[Gn],A))H1(F,HomR(ΛR(κ),A))H1(F,𝐀)\begin{split}H^{1}(F_{\infty},A)&=\mathop{\varinjlim}\limits_{n}H^{1}(F_{n},A)\\ &\simeq\mathop{\varinjlim}\limits_{n}H^{1}(F,\operatorname{Hom}_{R}(R[G_{n}],A))\\ &\simeq H^{1}(F,\mathop{\varprojlim}\limits_{n}\operatorname{Hom}_{R}(R[G_{n}],A))\\ &\simeq H^{1}(F,\operatorname{Hom}_{R}(\Lambda_{R}(\kappa),A))\\ &\simeq H^{1}(F,\mathbf{A})\end{split}

where the first and the last isomorphism follow from the previous discussion. We need to show that, under these isomorphisms, Selstr(Fn,A)\operatorname{Sel}_{\mathrm{str}}(F_{n},A) corresponds to Selstr(F,𝐀[𝔮n])\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}[\mathfrak{q}_{n}]) and Selstr(F,A)\operatorname{Sel}_{\mathrm{str}}(F_{\infty},A) corresponds to Selstr(F,𝐀)\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}).

Put Cn(M)=HomR(R[Gn],M)C_{n}(M)=\operatorname{Hom}_{R}(R[G_{n}],M) for any RR-module MM. Let Σn\Sigma_{n} be the set of places of FnF_{n} above places in Σ\Sigma. We have a commutative diagram:

Selstr(Fn,A)\textstyle{\operatorname{Sel}_{\mathrm{str}}(F_{n},A)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}rn\scriptstyle{r_{n}}H1(FΣ,Fn,A)\textstyle{H^{1}(F_{\Sigma},F_{n},A)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}sn\scriptstyle{s_{n}}vΣn,vpH1(Iw,A)×vΣn,vpH1(Fn,w,A/Fw+(A))\textstyle{\prod_{v\in\Sigma_{n},v\nmid p}H^{1}(I_{w},A)\times\prod_{v\in\Sigma_{n},v\mid p}H^{1}(F_{n,w},A/F^{+}_{w}(A))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tn\scriptstyle{t_{n}}Selstr(F,Cn(A))\textstyle{\operatorname{Sel}_{\mathrm{str}}(F,C_{n}(A))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}H1(FΣ/F,Cn(A))\textstyle{H^{1}(F_{\Sigma}/F,C_{n}(A))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}vΣ,vpH1(Iv,Cn(A))×vΣ,vpH1(Fv,Cn(A/Fv+(A)))\textstyle{\prod_{v\in\Sigma,v\nmid p}H^{1}(I_{v},C_{n}(A))\times\prod_{v\in\Sigma,v\mid p}H^{1}(F_{v},C_{n}(A/F^{+}_{v}(A)))}

where Selstr(F,Cn(A))\operatorname{Sel}_{\mathrm{str}}(F,C_{n}(A)) is defined by the exactness of the lower horizontal arrow. We claim that the vertical arrow tnt_{n} is injective. To show this, note that the map tnt_{n} is the product local maps tn,vt_{n,v} for all vΣnv\in\Sigma_{n}, so we study first these maps tn,vt_{n,v}. If wpw\nmid p, then Iw=IvI_{w}=I_{v} because Fn/FF_{n}/F is unramified outside pp; the map tn,vt_{n,v} defined by

tn,v:wvH1(Iw,A)=H1(Iv,A)wvH1(Iv,Cn(A))H1(Iv,HomR(R,A)){wv}.t_{n,v}:\prod_{w\mid v}H^{1}(I_{w},A)=H^{1}(I_{v},A)^{\sharp{w\mid v}}\longrightarrow H^{1}(I_{v},C_{n}(A))\simeq H^{1}(I_{v},\operatorname{Hom}_{R}(R,A))^{\sharp\{w\mid v\}}.

It follows that tn,vt_{n,v} is injective. The map tn,vt_{n,v} for vpv\mid p is defined by

tn,v:H1(Fn,w,A/Fv+(A))H1(Fv,HomR(R[Gn],A/Fv+(A)))t_{n,v}:H^{1}(F_{n,w},A/F^{+}_{v}(A))\longrightarrow H^{1}(F_{v},\operatorname{Hom}_{R}(R[G_{n}],A/F_{v}^{+}(A)))

which are all isomorphisms by Shapiro’s Lemma because, being pp totally ramified in the extension Fn/FF_{n}/F, we have Gal(Fn/F)Gn\operatorname{Gal}(F_{n}/F)\simeq G_{n}. We therefore conclude that tnt_{n} is injective. Since sns_{n} is an isomorphism, the map rnr_{n} is an isomorphism too, showing the result. ∎

Lemma 2.9.

Let MM be an RR-module equipped with a GG_{\infty}-action, denote MR-torsM_{R\text{-{tors}}} the RR-torsion submodule of MM and let

N=MR-torsRΛR(κ).N=M_{R\text{-tors}}\otimes_{R}\Lambda_{R}(\kappa).

Then the quotient N/(γpn1)NN/(\gamma^{p^{n}}-1)N is a pseudo-null ΛR\Lambda_{R}-module for each integer n1n\geq 1.

Proof.

Set I=(γpn1)I=(\gamma^{p^{n}}-1) for convenience. The support of N/INN/IN consists of the prime ideals of ΛR\Lambda_{R} containing II. Fix such a height one prime ideal 𝔞=(a)\mathfrak{a}=(a). Then aa is an irreducible factor of γpn1\gamma^{p^{n}}-1. Therefore, 𝔞R=0\mathfrak{a}\cap R=0. Thus, we have 𝔪R{0}(ΛR)𝔞×\mathfrak{m}_{R}\setminus\{0\}\subseteq(\Lambda_{R})^{\times}_{\mathfrak{a}}. It implies that the localization (MR-tors)𝔞(M_{R\text{-tors}})_{\mathfrak{a}} of MR-torsM_{R\text{-tors}} at 𝔞\mathfrak{a} is trivial. It follows that no height one prime ideal of ΛR\Lambda_{R} lies in the support of N/INN/IN, and therefore N/INN/IN is pseudo-null over ΛR\Lambda_{R}. ∎

Proposition 2.10.

The Pontryagin duals of the kernel and cokernel of the restriction

resF/Fn:Selstr(Fn,A)Selstr(F,A)Gal(F/Fn)\mathrm{res}_{F_{\infty}/F_{n}}:\operatorname{Sel}_{\mathrm{str}}(F_{n},A)\longrightarrow\operatorname{Sel}_{\mathrm{str}}(F_{\infty},A)^{\operatorname{Gal}(F_{\infty}/F_{n})}

are pseudo-null ΛR\Lambda_{R}-modules.

Proof.

We only need to check that the assumptions in Theorem 2.6 are satisfied. If so, the result follows by taking 𝔮n=(γpn1)\mathfrak{q}_{n}=(\gamma^{p^{n}}-1) in Theorem 2.6, and by using Lemma 2.8 to identify Selstr(F,𝐀[𝔮n])\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}[\mathfrak{q}_{n}]) and Selstr(F,𝐀)\operatorname{Sel}_{\mathrm{str}}(F,\mathbf{A}) with Selstr(Fn,A)\operatorname{Sel}_{\mathrm{str}}(F_{n},A) and Selstr(F,A)\operatorname{Sel}_{\mathrm{str}}(F_{\infty},A), respectively.

By Shapiro’s Lemma, we have H0(FΣ/F,𝐀[𝔮n])H0(FΣ/Fn,A)H^{0}(F_{\Sigma}/F,\mathbf{A}[\mathfrak{q}_{n}])\simeq H^{0}(F_{\Sigma}/F_{n},A), and therefore the first assumption in Theorem 2.6 is equivalent to (2) in Assumption 2.7.

We first consider Cvstr/(γpn1)CvstrC_{v}^{\mathrm{str}}/(\gamma^{p^{n}}-1)C_{v}^{\mathrm{str}} for vpv\nmid p. The action of IvI_{v} on ΛR(κ1)\Lambda_{R}(\kappa^{-1}) trivial since all the primes outside pp are unramified in FF_{\infty}. Therefore, (𝐓)Iv=(T)IvRΛR(κ)(\mathbf{T}^{*})_{I_{v}}=(T^{*})_{I_{v}}\otimes_{R}\Lambda_{R}(\kappa), and

Cvstr=((T)Iv)R-torsΛR(κ)C_{v}^{\mathrm{str}}=((T^{*})_{I_{v}})_{R\text{-tors}}\otimes\Lambda_{R}(\kappa)

where ((T)Iv)R-tors((T^{*})_{I_{v}})_{R\text{-tors}} is the RR-torsion submodule of (T)Iv(T^{*})_{I_{v}}. Thus, for vpv\nmid p, the statement in the assumption of Theorem 2.6 is equivalent to that

((T)Iv)R-torsΛR(κ)/(γpn1)((T)Iv)R-torsΛR(κ)((T^{*})_{I_{v}})_{R\text{-tors}}\otimes\Lambda_{R}(\kappa)/(\gamma^{p^{n}}-1)((T^{*})_{I_{v}})_{R\text{-tors}}\otimes\Lambda_{R}(\kappa)

is pseudo-null, which follows from Lemma 2.9 applied to M=(T)IvM=(T^{*})_{I_{v}}.

We now consider Cvstr/(γpn1)CvstrC_{v}^{\mathrm{str}}/(\gamma^{p^{n}}-1)C_{v}^{\mathrm{str}} for vpv\mid p. Since ΛR\Lambda_{R} is flat over RR, we have

(𝐓/Fv+(𝐓))=(T/Fv+(T))RΛR(κ)=Fv(T)RΛR(κ).(\mathbf{T}/F_{v}^{+}(\mathbf{T}))^{*}=(T/F^{+}_{v}(T))^{*}\otimes_{R}\Lambda_{R}(\kappa)=F^{-}_{v}(T)^{*}\otimes_{R}\Lambda_{R}(\kappa).

We have

(T/Fv+(T))RΛR(κ)((T/Fv+(T))RR(θv))R(ΛR(κ)RR(θv1))((T/Fv+(T))RR(θv1))R(ΛR(κθv1))(Fv(T)RR(θv1))R(ΛR(κθv1)).\begin{split}(T/F^{+}_{v}(T))^{*}\otimes_{R}\Lambda_{R}(\kappa)&\simeq\left((T/F^{+}_{v}(T))^{*}\otimes_{R}R(\theta_{v})\right)\otimes_{R}\left(\Lambda_{R}(\kappa)\otimes_{R}R(\theta_{v}^{-1})\right)\\ &\simeq\left((T/F^{+}_{v}(T))\otimes_{R}R(\theta_{v}^{-1})\right)^{*}\otimes_{R}\left(\Lambda_{R}(\kappa\cdot\theta_{v}^{-1})\right)\\ &\simeq\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}\otimes_{R}\left(\Lambda_{R}(\kappa\cdot\theta_{v}^{-1})\right).\end{split}

The action of IvI_{v} on Fv(T)R(θv1)F^{-}_{v}(T)\otimes R(\theta_{v}^{-1}) is trivial by (4) in Assumption 2.7, and therefore the IvI_{v}-coinvariant of (𝐓/Fv+(𝐓))(\mathbf{T}/F_{v}^{+}(\mathbf{T}))^{*} is

(Fv(T)RR(θv1))R(ΛR(κθv1))Iv.(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1}))^{*}\otimes_{R}\left(\Lambda_{R}(\kappa\cdot\theta_{v}^{-1})\right)_{I_{v}}.

Since δv\delta_{v} is unramified by (4) in Assumption 2.7, the coinvariant of the action of Gv/IvG_{v}/I_{v} on (Fv(T)RR(θv1))\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*} is given by

(Fv(T)RR(θv1))(Frobv1)(Fv(T)RR(θv1))(Fv(T)RR(θv1))(uv1)(Fv(T)RR(θv1))(Fv(T)RR(θv1))((uv1)Fv(T)RR(θv1))\begin{split}\frac{\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}}{(\operatorname{Frob}_{v}-1)\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}}&\simeq\frac{\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}}{(u_{v}-1)\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}}\\ &\simeq\frac{\left(F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}}{\left((u_{v}-1)F^{-}_{v}(T)\otimes_{R}R(\theta_{v}^{-1})\right)^{*}}\end{split}

By (4) in Assumption 2.7, uvu_{v} is not congruent to 1 modulo the maximal ideal of RR, so uv1R×u_{v}-1\in R^{\times}, and therefore (uv1)Fv(T)=0(u_{v}-1)F_{v}^{-}(T)=0. Moreover, δv\delta_{v} acts trivially on (ΛR(κθv1))Iv\left(\Lambda_{R}(\kappa\cdot\theta_{v}^{-1})\right)_{I_{v}}. Therefore, the GvG_{v}-coinvariant of (𝐓/Fv+(𝐓))(\mathbf{T}/F_{v}^{+}(\mathbf{T}))^{*} is trivial, and it follows in particular that the assumption on Cvstr/(γpn1)CvstrC_{v}^{\mathrm{str}}/(\gamma^{p^{n}}-1)C_{v}^{\mathrm{str}} for vpv\mid p in Theorem 2.6 is satisfied. ∎

Lemma 2.11.

Suppose that MM is a pseudo-null ΛR\Lambda_{R}-module. Then for each integer n0n\geq 0, M/(γpn1)MM/(\gamma^{p^{n}}-1)M is torsion over R[Gn]ΛR/(γpn1)ΛRR[G_{n}]\simeq\Lambda_{R}/(\gamma^{p^{n}}-1)\Lambda_{R}.

Proof.

Suppose M/(γpn1)MM/(\gamma^{p^{n}}-1)M is not a torsion R[Γn]R[\Gamma_{n}]-module, and take a copy NN of R[Γn]R[\Gamma_{n}] in M/(γpn1)MM/(\gamma^{p^{n}}-1)M. Take any height one prime ideal 𝔞=(a)\mathfrak{a}=(a) of ΛR\Lambda_{R} such that (a,γpn1)=1(a,\gamma^{p^{n}}-1)=1. Then N𝔞0N_{\mathfrak{a}}\neq 0. In particular, M𝔞/(γpn1)M𝔞0M_{\mathfrak{a}}/(\gamma^{p^{n}}-1)M_{\mathfrak{a}}\neq 0 so M𝔞0M_{\mathfrak{a}}\neq 0, which contradicts the assumption that MM is a pseudo-null ΛR\Lambda_{R}-module. ∎

Corollary 2.12.

The Pontryagin duals of the kernel and cokernel of the restriction

resF/Fn:Selstr(Fn,A)Selstr(F,A)Gal(F/Fn)\mathrm{res}_{F_{\infty}/F_{n}}:\operatorname{Sel}_{\mathrm{str}}(F_{n},A)\longrightarrow\operatorname{Sel}_{\mathrm{str}}(F_{\infty},A)^{\operatorname{Gal}(F_{\infty}/F_{n})}

are cotorsion R[Gn]R[G_{n}]-modules.

Proof.

It follows from Proposition 2.10 and Lemma 2.11. ∎

3. Anticyclotomic Iwasawa theory for Hida families

3.1. Ordinary families of modular forms

Let f0=n=1anqnSk0(Γ0(Np))f_{0}=\sum_{n=1}^{\infty}a_{n}q^{n}\in S_{k_{0}}(\Gamma_{0}(Np)) an ordinary pp-stabilized newform (in the sense of [GS93, Def. 2.5]) of weight k02k_{0}\geq 2 and trivial nebentypus, defined over a finite extension L/pL/\mathbb{Q}_{p}. Let 𝒪=𝒪L\mathcal{O}=\mathcal{O}_{L} be the valuation ring of LL and ap𝒪×a_{p}\in\mathcal{O}^{\times}, and f0f_{0} is either a newform of level NpNp, or arises from a newform of level NN. Denote

ρf0:G:=Gal(¯/)GL2(𝒪)\rho_{f_{0}}:G_{\mathbb{Q}}:={\rm Gal}(\overline{\mathbb{Q}}/\mathbb{Q})\longrightarrow{\rm GL}_{2}(\mathcal{O})

the Galois representation associated with f0f_{0}. Since f0f_{0} is ordinary at pp, the restriction of ρf0\rho_{f_{0}} to a decomposition group DpGD_{p}\subset G_{\mathbb{Q}} is upper-triangular. We also denote k=kLk=k_{L} the residue field of LL and

ρ¯f0:GGL2(k)\bar{\rho}_{f_{0}}:G_{\mathbb{Q}}\longrightarrow{\rm GL}_{2}(k)

the residual representation obtained by reduction modulo the maximal ideal 𝔪=𝔪L\mathfrak{m}=\mathfrak{m}_{L} of 𝒪\mathcal{O}.

Assumption 3.1.

The representation ρ¯f0\bar{\rho}_{f_{0}} is absolutely irreducible, and pp-distinguished, i.e., writing ρ¯f0|Dp(ε¯0δ¯)\bar{\rho}_{f_{0}}|_{D_{p}}\sim\left(\begin{smallmatrix}\bar{\varepsilon}&*\\ 0&\bar{\delta}\end{smallmatrix}\right), we have ε¯δ¯\bar{\varepsilon}\neq\bar{\delta}.

Let 𝔥ord\mathfrak{h}^{\mathrm{ord}} be the Hida ordinary Hecke algebra of tame level Γ0(N)\Gamma_{0}(N), and let \mathcal{R} be the branch of 𝔥ord\mathfrak{h}^{\mathrm{ord}} passing through f0{f_{0}}. If Λ:=𝒪[[Γ]]\Lambda:=\mathcal{O}[[\Gamma]], where Γ=1+pp\Gamma=1+p\mathbb{Z}_{p}, then \mathcal{R} is a finite flat extension of Λ\Lambda (the structure of Λ\Lambda-algebra in 𝔥ord\mathfrak{h}^{\mathrm{ord}} is given by the action of diamond operators in Γ\Gamma). The eigenform f0f_{0} defines an 𝒪L\mathcal{O}_{L}-algebra homomorphism λf0:𝒪\lambda_{f_{0}}:\mathcal{R}\rightarrow\mathcal{O}, which is called arithmetic. More generally, an arithmetic point of \mathcal{R} is a continuous 𝒪L\mathcal{O}_{L}-algebra homomorphism 𝜅¯p\mathcal{R}\overset{\kappa}{\rightarrow}\overline{\mathbb{Q}}_{p} such that the composition

ΓΛ×𝜅¯p×\Gamma\longrightarrow\Lambda^{\times}\longrightarrow\mathcal{R}\xrightarrow{\;\kappa\;}\overline{\mathbb{Q}}_{p}^{\times}

is given by γψ(γ)γk2\gamma\mapsto\psi(\gamma)\gamma^{k-2}, for some integer k2k\geq 2 and some finite order character ψ:Γ¯p×\psi:\Gamma\rightarrow\overline{\mathbb{Q}}_{p}^{\times}. We then say that κ\kappa has weight kk, character ψ\psi, and wild level pmp^{m}, where m>0m>0 is such that ker(ψ)=1+pmp{\rm ker}(\psi)=1+p^{m}\mathbb{Z}_{p}. Denote by 𝒳()\mathcal{X}(\mathcal{R}) the set of continuous 𝒪\mathcal{O}-algebra homomorphisms from \mathcal{R} into 𝒪\mathcal{O}, and by 𝒳arith()\mathcal{X}_{\rm arith}(\mathcal{R}) the subset of 𝒳()\mathcal{X}(\mathcal{R}) consisting of arithmetic primes. For each κ𝒳arith()\kappa\in\mathcal{X}_{\rm arith}(\mathcal{R}), let FκF_{\kappa} be the residue field of ker(κ){\rm ker}(\kappa)\subset\mathcal{R}, which is a finite extension of p\mathbb{Q}_{p}.

For each n1n\geq 1, let 𝐚n\mathbf{a}_{n}\in\mathcal{R} be the image of Tn𝔥ordT_{n}\in\mathfrak{h}^{\rm ord} under the natural projection 𝔥ord\mathfrak{h}^{\rm ord}\rightarrow\mathcal{R}, and form the qq-expansion

𝐟=n=1𝐚nqn[[q]].\mathbf{f}=\sum_{n=1}^{\infty}\mathbf{a}_{n}q^{n}\in\mathcal{R}[[q]].

By [Hid86, Thm. 1.2], if κ𝒳arith()\kappa\in\mathcal{X}_{\rm arith}(\mathcal{R}) is an arithmetic prime of weight k2k\geq 2, character ψ\psi, and wild level pmp^{m}, then

fκ:=n=1κ(𝐚n)qnFκ[[q]]f_{\kappa}:=\sum_{n=1}^{\infty}\kappa(\mathbf{a}_{n})q^{n}\in F_{\kappa}[[q]]

is (the qq-expansion of) an ordinary pp-stabilized newform in Sk(Γ0(Npm),ωk0kψ)S_{k}(\Gamma_{0}(Np^{m}),\omega^{k_{0}-k}\psi) of level Γ0(Npn)\Gamma_{0}(Np^{n}), character ωk0kψ\omega^{k_{0}-k}\psi and weight kk, where ω:(/p)×p×\omega:(\mathbb{Z}/p\mathbb{Z})^{\times}\rightarrow\mathbb{Z}_{p}^{\times} is the Teichmüller character.

3.2. Critical characters

Following [How07, Def. 2.1.3], factor the pp-adic cyclotomic character as

εcyc=εtameεwild:Gp×𝝁p1×Γ,\varepsilon_{\rm cyc}=\varepsilon_{\rm tame}\cdot\varepsilon_{\rm wild}:G_{\mathbb{Q}}\longrightarrow\mathbb{Z}_{p}^{\times}\simeq\boldsymbol{\mu}_{p-1}\times\Gamma,

and define the critical character Θ:G×\Theta:G_{\mathbb{Q}}\rightarrow\mathcal{R}^{\times} by

(4) Θ(σ)=εtamek022(σ)[εwild1/2(σ)],\Theta(\sigma)=\varepsilon_{\rm tame}^{\frac{k_{0}-2}{2}}(\sigma)\cdot[\varepsilon^{1/2}_{\rm wild}(\sigma)],

where εtamek022:G𝝁p1\varepsilon_{\rm tame}^{\frac{k_{0}-2}{2}}:G_{\mathbb{Q}}\rightarrow\boldsymbol{\mu}_{p-1} is any fixed choice of square-root of εtamek02\varepsilon_{\rm tame}^{k_{0}-2} (see [How07, Rem. 2.1.4]), εwild1/2:GΓ\varepsilon_{\rm wild}^{1/2}:G_{\mathbb{Q}}\rightarrow\Gamma is the unique square-root of εwild\varepsilon_{\rm wild} taking values in Γ\Gamma, and []:ΓΛ××[\cdot]:\Gamma\rightarrow\Lambda^{\times}\rightarrow\mathcal{R}^{\times} is the map given by the inclusion as group-like elements.

Define the character θ:p××\theta:\mathbb{Z}_{p}^{\times}\rightarrow\mathcal{R}^{\times} by the relation Θ=θεcyc,\Theta=\theta\circ\varepsilon_{\rm cyc}, and for each κ𝒳arith()\kappa\in\mathcal{X}_{\rm arith}(\mathcal{R}), let θκ:pׯp×\theta_{\kappa}:\mathbb{Z}_{p}^{\times}\rightarrow\overline{\mathbb{Q}}_{p}^{\times} be the composition of θ\theta with κ\kappa. If κ\kappa has weight k2k\geq 2 and character ψ\psi, then

(5) θκ2(z)=zk2ωk0kψ(z)\theta_{\kappa}^{2}(z)=z^{k-2}\omega^{k_{0}-k}\psi(z)

for all zp×z\in\mathbb{Z}_{p}^{\times}.

3.3. pp-adic LL-functions

Let K/K/\mathbb{Q} be an imaginary quadratic field of discriminant prime to NpNp. Write N=N+NN=N^{+}N^{-}, where all primes dividing N+N^{+} are split in KK, and all primes dividing NN^{-} are inert in KK. We will work under the following

Assumption 3.2.
  1. (1)

    NN^{-} is a square-free product of an odd number of primes.

  2. (2)

    The residual representation ρ¯f0\bar{\rho}_{f_{0}} is ramified at all primes N\ell\mid N^{-}.

  3. (3)

    ap±1a_{p}\not\equiv\pm 1 modulo the maximal ideal of 𝒪\mathcal{O} (we say that pp is a non-anomalous prime for ρ¯f0\bar{\rho}_{f_{0}} in this case).

  4. (4)

    pp is split in KK.

Let BB be the definite quaternion algebra over \mathbb{Q} of discriminant NN^{-}. For each prime N\ell\nmid N^{-}, fix isomorphisms ι:BM2()\iota_{\ell}:B\otimes_{\mathbb{Q}}\mathbb{Q}_{\ell}\simeq\operatorname{M}_{2}(\mathbb{Q}_{\ell}). Let mRmm\mapsto R_{m}, for m0m\geq 0 an integer, be the sequence of Eichler orders of level N+pmN^{+}p^{m}, defined by the condition that ι(Rm)\iota_{\ell}(R_{m}\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}) consists of the matrices in M2()\operatorname{M}_{2}(\mathbb{Z}_{\ell}) which are upper triangular modulo val(N+pm)\ell^{\mathrm{val}_{\ell}(N^{+}p^{m})} for all primes N\ell\nmid N^{-} (thus, in particular, Rm+1RmR_{m+1}\subseteq R_{m} for all integers m0m\geq 0). For a ring AA, denote A^\hat{A} its profinite completion. Let UmR^m×U_{m}\subset\widehat{R}_{m}^{\times} be the compact open subgroup defined by

Um:={(xq)qR^m×|ip(xp)(10)(modpm)}.U_{m}:=\left\{(x_{q})_{q}\in\widehat{R}_{m}^{\times}\;\;|\;\;i_{p}(x_{p})\equiv\left(\begin{array}[]{cc}1&*\\ 0&*\end{array}\right)\pmod{p^{m}}\right\}.

Consider the double coset spaces

X~m(K)=B×\(Hom(K,B)×B^×)/Um,\widetilde{X}_{m}(K)=B^{\times}\big{\backslash}\bigl{(}\operatorname{Hom}_{\mathbb{Q}}(K,B)\times\widehat{B}^{\times}\bigr{)}\big{/}U_{m},

where bB×b\in B^{\times} act on left on (Ψ,g)Hom(K,B)×B^×(\Psi,g)\in\operatorname{Hom}_{\mathbb{Q}}(K,B)\times\widehat{B}^{\times} by b(Ψ,g)=(bgb1,bg),b\cdot(\Psi,g)=(bgb^{-1},bg), and UmU_{m} acts on B^×\widehat{B}^{\times} by right multiplication and on Hom(K,B)\operatorname{Hom}_{\mathbb{Q}}(K,B) trivially. The space X~m(K)\widetilde{X}_{m}(K) is equipped with a nontrivial Galois action defined as follows: If σGal(Kab/K)\sigma\in{\rm Gal}(K^{\rm ab}/K) and PX~m(K)P\in\widetilde{X}_{m}(K) is the class of a pair (Ψ,g)(\Psi,g), then Pσ:=[(Ψ,gΨ^(a))],P^{\sigma}:=[(\Psi,g\widehat{\Psi}(a))], where aK×\K^×a\in K^{\times}\backslash\widehat{K}^{\times} is such that recK(a)=σ{\rm rec}_{K}(a)=\sigma, and we extend this to an action of GKG_{K} by letting each σGK\sigma\in G_{K} act on X~m(K)\widetilde{X}_{m}(K) as σ|Kab\sigma|_{K^{\rm ab}}. The space X~m(K)\widetilde{X}_{m}(K) is also equipped with standard action of Hecke operators TT_{\ell} for Np\ell\nmid Np, UpU_{p} and diamond operators d\langle d\rangle for dp×d\in\mathbb{Z}_{p}^{\times}.

Let Dm:=Div(X~m)𝒪LD_{m}:=\operatorname{Div}(\tilde{X}_{m})\otimes\mathcal{O}_{L} be the divisor group of X~m\tilde{X}_{m} and denote αm:DmDm1\alpha_{m}:D_{m}\twoheadrightarrow D_{m-1} the canonical projection. Passing to the ordinary part DmordD_{m}^{\mathrm{ord}} and tensoring with the primitive component \mathcal{R} gives Hecke modules 𝐃m\mathbf{D}_{m} (for m0m\geq 0) and, twisting the Galois action by Θ1\Theta^{-1}, Hecke modules 𝐃m\mathbf{D}_{m}^{\dagger}. The analogous Hecke modules obtained from the inverse limits of the divisor group DmD_{m} (with respect to the canonical projection maps αm\alpha_{m}) are the Hecke modules denoted 𝐃\mathbf{D} and 𝐃\mathbf{D}^{\dagger} in [LV11, §6.4]. Let eorde^{\mathrm{ord}} denote the ordinary projector. Denote Pic(X~m){\rm Pic}(\widetilde{X}_{m}) the Picard group of X~m\widetilde{X}_{m}. Define the Hecke modules Jmord:=eord(Pic(X~m)𝒪L)J_{m}^{\rm ord}:=e^{\rm ord}({\rm Pic}(\widetilde{X}_{m})\otimes_{\mathbb{Z}}\mathcal{O}_{L}), 𝐉m:=Jmord𝔥ord\mathbf{J}_{m}:=J_{m}^{\rm ord}\otimes_{\mathfrak{h}^{\rm ord}}\mathcal{R} and 𝐉m:=𝐉m\mathbf{J}_{m}^{\dagger}:=\mathbf{J}_{m}\otimes_{\mathcal{R}}\mathcal{R}^{\dagger}. Finally define 𝐉:=limm𝐉m\mathbf{J}^{\dagger}:=\varprojlim_{m}\mathbf{J}_{m}^{\dagger}. The projections Div(X~m)Pic(X~m){\rm Div}(\widetilde{X}_{m})\rightarrow{\rm Pic}(\widetilde{X}_{m}) induce a map

λ:𝐃𝐉.\lambda:\mathbf{D}^{\dagger}\longrightarrow\mathbf{J}^{\dagger}.

Thanks to Assumptions 3.1 and 3.2, we have dimk(𝐉/𝔪𝐉)=1{\rm dim}_{k_{\mathcal{R}}}(\mathbf{J}^{\dagger}/\mathfrak{m}_{\mathcal{R}}\mathbf{J}^{\dagger})=1 by [CKL17, Theorem 3.1]; here, 𝔪\mathfrak{m}_{\mathcal{R}} is the maximal ideal of \mathcal{R}, and k:=/𝔪k_{\mathcal{R}}:=\mathcal{R}/\mathfrak{m}_{\mathcal{R}} is its residue field. By [LV11, Prop. 9.3], we conclude that the module 𝐉\mathbf{J}^{\dagger} is free of rank one over \mathcal{R}. Fix an isomorphism

η:𝐉.\eta:\mathbf{J}^{\dagger}\simeq\mathcal{R}.

Let KK_{\infty} be the anticyclotomic p\mathbb{Z}_{p}-extension of KK, and define Γ=Gal(K/K)p\Gamma_{\infty}=\operatorname{Gal}(K_{\infty}/K)\simeq\mathbb{Z}_{p}. Denote KnK_{n} the subfield of KK_{\infty} such that Γn=Gal(Kn/K)/pn\Gamma_{n}=\operatorname{Gal}(K_{n}/K)\simeq\mathbb{Z}/p^{n}\mathbb{Z}. Define

Λ=[[Γ]]=limn[Γn].\Lambda_{\mathcal{R}}=\mathcal{R}[[\Gamma_{\infty}]]=\mathop{\varprojlim}\limits_{n}\mathcal{R}[\Gamma_{n}].

The paper [LV11] introduces for each integer n0n\geq 0 a sequence mP~pn,mm\mapsto\tilde{P}_{p^{n},m} of Gross-Heegner points in X~m(K)\tilde{X}_{m}(K) of conductor pm+np^{m+n}; these points satisfy norm-relations and allows to construct big theta elements Θn(𝐟)𝐃[Γn]\Theta_{n}(\mathbf{f})\in\mathbf{D}[\Gamma_{n}] by an inverse limit procedure inverting the UpU_{p} operator; we will view Θn(𝐟)\Theta_{n}(\mathbf{f}) as elements in [Γn]\mathcal{R}[\Gamma_{n}] by means of the map 𝐃𝜆𝐉𝜂\mathbf{D}\overset{\lambda}{\rightarrow}\mathbf{J}\overset{\eta}{\simeq}\mathcal{R}. The elements Θn(𝐟)\Theta_{n}(\mathbf{f}) are compatible under the natural maps [Γm][Γn]\mathcal{R}[\Gamma_{m}]\rightarrow\mathcal{R}[\Gamma_{n}] for all mnm\geq n, thus defining an element Θ(𝐟):=limnΘn(𝐟)\Theta_{\infty}(\mathbf{f}):=\varprojlim_{n}\Theta_{n}(\mathbf{f}) in the completed group ring Λ\Lambda_{\mathcal{R}}.

Definition 3.3.

The two-variable pp-adic LL-function attached 𝐟\mathbf{f} and KK is the element

Lp(𝐟/K):=Θ(𝐟)Θ(𝐟)Λ,L_{p}(\mathbf{f}/K):=\Theta_{\infty}(\mathbf{f})\cdot\Theta_{\infty}(\mathbf{f})^{*}\in\Lambda_{\mathcal{R}},

where xxx\mapsto x^{*} is the involution on [[Γ]]\mathcal{R}[[\Gamma_{\infty}]] given by γγ1\gamma\mapsto\gamma^{-1} on group-like elements.

3.4. Selmer groups of Hida families

Let 𝐓\mathbf{T} be Hida’s big Galois representation associated with \mathcal{R}. Then 𝐓\mathbf{T} is a free \mathcal{R}-module of rank 22, equipped with a continuous action of G=Gal(¯/)G_{\mathbb{Q}}=\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q}) and a filtration of [Gp]\mathcal{R}[G_{\mathbb{Q}_{p}}]-modules

0Fv+(𝐓)𝐓Fv(𝐓)00\longrightarrow F^{+}_{v}(\mathbf{T})\longrightarrow\mathbf{T}\longrightarrow F^{-}_{v}(\mathbf{T})\longrightarrow 0

where Gp=Gal(¯p/p)G_{\mathbb{Q}_{p}}=\operatorname{Gal}(\bar{\mathbb{Q}}_{p}/\mathbb{Q}_{p}) is a decomposition group of GG_{\mathbb{Q}} at pp. Both Fv+(𝐓)F^{+}_{v}(\mathbf{T}) and Fv(𝐓)F^{-}_{v}(\mathbf{T}) are free \mathcal{R}-modules of rank 11; GvG_{v} acts on Fv(𝐓)F^{-}_{v}(\mathbf{T}) via the unamified character ηv:Gv/Iv×\eta_{v}:G_{v}/I_{v}\rightarrow\mathcal{R}^{\times} which takes the arithmetic Frobenius to UpU_{p}, and GvG_{v} acts on Fv+(𝐓)F_{v}^{+}(\mathbf{T}) via ηv1εcyc[εcyc]\eta_{v}^{-1}\varepsilon_{\mathrm{cyc}}[\varepsilon_{\mathrm{cyc}}].

Denote 𝐓=𝕋Θ1\mathbf{T}^{\dagger}=\mathbb{T}\otimes\Theta^{-1} the critical twist of 𝐓\mathbf{T} corresponding to the choice of the critical character Θ\Theta in (4). For each arithmetic point, define Fκ=κ/ker(κ)κF_{\kappa}=\mathcal{R}_{\kappa}/\ker(\kappa)\mathcal{R}_{\kappa}, where κ\mathcal{R}_{\kappa} is the localisation of \mathcal{R} at κ\kappa. Then Vκ=𝐓FκV_{\kappa}^{\dagger}=\mathbf{T}^{\dagger}\otimes_{\mathcal{R}}F_{\kappa} is isomorphic to the self-dual twist of Deligne representation VfκV_{f_{\kappa}} attached to the eigenform fκf_{\kappa}. If 𝔭=𝔭κ=ker(κ)\mathfrak{p}=\mathfrak{p}_{\kappa}=\ker(\kappa), we also denote κ\mathcal{R}_{\kappa} by 𝔭\mathcal{R}_{\mathfrak{p}} and VκV_{\kappa}^{\dagger} by V𝔭V_{\mathfrak{p}}^{\dagger}. Moreover, we have a filtration [Gp]\mathcal{R}[G_{\mathbb{Q}_{p}}]-modules

0Fv+(𝐓)𝐓Fv(𝐓)00\longrightarrow F^{+}_{v}(\mathbf{T}^{\dagger})\longrightarrow\mathbf{T}^{\dagger}\longrightarrow F^{-}_{v}(\mathbf{T}^{\dagger})\longrightarrow 0

where GvG_{v} acts on Fv(𝐓)F^{-}_{v}(\mathbf{T}^{\dagger}) via the character ηvΘ1\eta_{v}\Theta^{-1} and GvG_{v} acts on Fv+(𝐓)F_{v}^{+}(\mathbf{T}^{\dagger}) via ηv1Θ1εcyc[εcyc]\eta_{v}^{-1}\Theta^{-1}\varepsilon_{\mathrm{cyc}}[\varepsilon_{\mathrm{cyc}}]. Let

𝐀=Φ(𝐓)=𝐓.\mathbf{A}^{\dagger}=\Phi(\mathbf{T}^{\dagger})=\mathbf{T}^{\dagger}\otimes_{\mathcal{R}}\mathcal{R}^{\vee}.

As in §2.2 we introduce strict Greenberg Selmer groups Selstr(Kn,𝐀)\operatorname{Sel}_{\mathrm{str}}(K_{n},\mathbf{A}^{\dagger}) and Selstr(K,𝐀)\operatorname{Sel}_{\mathrm{str}}(K_{\infty},\mathbf{A}^{\dagger}) and Selmer groups Sel(Kn,𝐀)\operatorname{Sel}(K_{n},\mathbf{A}^{\dagger}) and Sel(K,𝐀)\operatorname{Sel}(K_{\infty},\mathbf{A}^{\dagger}). Under our assumptions, by [CKL17, Theorem 4.1], we know that Selstr(Kn,𝐀)Sel(Kn,𝐀)\operatorname{Sel}_{\mathrm{str}}(K_{n},\mathbf{A}^{\dagger})\simeq\operatorname{Sel}(K_{n},\mathbf{A}^{\dagger}) and Selstr(K,𝐀)Sel(K,𝐀)\operatorname{Sel}_{\mathrm{str}}(K_{\infty},\mathbf{A}^{\dagger})\simeq\operatorname{Sel}(K_{\infty},\mathbf{A}^{\dagger}). We may also consider Nekovář’s extended Selmer groups H~f1(Kn,𝐀)\tilde{H}^{1}_{f}(K_{n},\mathbf{A}^{\dagger}) and H~f1(K,𝐀)\tilde{H}^{1}_{f}(K_{\infty},\mathbf{A}^{\dagger}). By [Nek06, Lemma 9.6.3] we have an exact sequence

H0(Kn,𝐀)H~f1(Kn,𝐀)Selstr(Kn,𝐀)0.H^{0}(K_{n},\mathbf{A}^{\dagger})\longrightarrow\tilde{H}^{1}_{f}(K_{n},\mathbf{A}^{\dagger})\longrightarrow\operatorname{Sel}_{\mathrm{str}}(K_{n},\mathbf{A}^{\dagger})\longrightarrow 0.
Lemma 3.4.

H0(Kn,𝐀)=0H^{0}(K_{n},\mathbf{A}^{\dagger})=0.

Proof.

Let M=H0(Kn,𝐀)M=H^{0}(K_{n},\mathbf{A}^{\dagger})^{\vee} be the Pontryagin dual of H0(Kn,𝐀)H^{0}(K_{n},\mathbf{A}^{\dagger}). By the topological Nakayama’s Lemma, it is enough to show that M/𝔪M=0M/\mathfrak{m}_{\mathcal{R}}M=0, where 𝔪\mathfrak{m}_{\mathcal{R}} is the maximal ideal of \mathcal{R}. For this, taking again Pontryagin duals, it is enough to show that

H0(Kn,𝐀)[𝔪R]=H0(Kn,𝐀[𝔪R])=0.H^{0}(K_{n},\mathbf{A}^{\dagger})[\mathfrak{m}_{R}]=H^{0}(K_{n},\mathbf{A}^{\dagger}[\mathfrak{m}_{R}])=0.

Now the Galois representation 𝐀[𝔪R]\mathbf{A}^{\dagger}[\mathfrak{m}_{R}] is isomorphic to ρ¯f0\bar{\rho}_{f_{0}}, which is irreducible by assumption, and it follows from standard arguments (e.g. [LV17, Lemmas 3.9, 3.10]) that the KnK_{n}-invariants of 𝐀[𝔪]\mathbf{A}^{\dagger}[\mathfrak{m}_{\mathcal{R}}] are trivial. ∎

It follows from Lemma 3.4 that H~f1(Kn,𝐀)Selstr(Kn,𝐀)\tilde{H}^{1}_{f}(K_{n},\mathbf{A}^{\dagger})\simeq\operatorname{Sel}_{\mathrm{str}}(K_{n},\mathbf{A}^{\dagger}). Thus, summing up, we have

(6) Selstr(Kn,𝐀)Sel(Kn,𝐀)H~f1(Kn,𝐀)\operatorname{Sel}_{\mathrm{str}}(K_{n},\mathbf{A}^{\dagger})\simeq\operatorname{Sel}(K_{n},\mathbf{A}^{\dagger})\simeq\tilde{H}^{1}_{f}(K_{n},\mathbf{A}^{\dagger})

and, taking direct limits with respect to the canonical restriction maps,

(7) Selstr(K,𝐀)Sel(K,𝐀)H~f,Iw1(K,𝐀)=limnH~f1(Kn,𝐀).\operatorname{Sel}_{\mathrm{str}}(K_{\infty},\mathbf{A}^{\dagger})\simeq\operatorname{Sel}(K_{\infty},\mathbf{A}^{\dagger})\simeq\tilde{H}^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})=\mathop{\varinjlim}\limits_{n}\tilde{H}^{1}_{f}(K_{n},\mathbf{A}^{\dagger}).

3.5. Control theorems for Hida representations

Let InI_{n} be the kernel of the map Λ𝒪\Lambda\rightarrow\mathcal{O} which takes the topological generator γ\gamma of Γ\Gamma_{\infty} to γpn1\gamma^{p^{n}}-1. For an integer n0n\geq 0, define

Δn=Gal(K/Kn).\Delta_{n}=\operatorname{Gal}(K_{\infty}/K_{n}).

In particular, we have Γ/ΔnΓn\Gamma_{\infty}/\Delta_{n}\simeq\Gamma_{n}.

Theorem 3.5.

The kernel and cokernel of the map

H~f1(Kn,𝐀)H~f,Iw1(K,𝐀)Δn\tilde{H}^{1}_{f}(K_{n},\mathbf{A}^{\dagger})\longrightarrow\tilde{H}^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\Delta_{n}}

are cotorsion Λ/InΛ[Γn]\Lambda_{\mathcal{R}}/I_{n}\Lambda_{\mathcal{R}}\simeq\mathcal{R}[\Gamma_{n}]-modules.

Proof.

This follows from Corollary 2.12 and (6), (7) once we check that Assumption 2.7 of Proposition 2.10 are satisfied for T=𝐓T=\mathbf{T}^{\dagger} and R=R=\mathcal{R} in Assumption 2.7. We know that 𝐓\mathbf{T}^{\dagger} is free of rank 22 over \mathcal{R}, and is unramified over the set of places Σ\Sigma dividing NpNp; moreover, Fv+(𝐓)F^{+}_{v}(\mathbf{T}^{\dagger}) and Fv(𝐓F^{-}_{v}(\mathbf{T}^{\dagger} are free of rank 11 over \mathcal{R}, so both (1) and (3) are satisfied. For (2) we need to check that H0(KΣ/Kn,𝐀)H^{0}(K_{\Sigma}/K_{n},\mathbf{A}^{\dagger}) is a pseudo-null Λ\Lambda_{\mathcal{R}}-module. Since 𝐀\mathbf{A}^{\dagger} is unramified outside Σ\Sigma, the Galois group Gal(¯/KΣ)\operatorname{Gal}(\bar{\mathbb{Q}}/K_{\Sigma}) acts trivially on 𝐀\mathbf{A}^{\dagger}, so H0(KΣ/Kn,𝐀)=H0(Kn,𝐀)H^{0}(K_{\Sigma}/K_{n},\mathbf{A}^{\dagger})=H^{0}(K_{n},\mathbf{A}^{\dagger}), which is trivial by Lemma 3.4. Condition (2) is guaranteed by the fact that pp is non-anomalous in Assumption 3.2, after taking δv=ηv1\delta_{v}=\eta_{v}^{-1} and θv=Θ1εcyc[εcyc]\theta_{v}=\Theta^{-1}\varepsilon_{\mathrm{cyc}}[\varepsilon_{\mathrm{cyc}}], noting that θv\theta_{v} factors through the cyclotomic p\mathbb{Z}_{p}-extension of KK. ∎

4. Proofs of the main results

The following result proves [LV11, Conjecture 9.12], a definite version of the two-variable Iwasawa main conjecture for Hida families in the anticyclotomic context.

Theorem 4.1.

Suppose Assumptions 3.1 and 3.2 are satisfied, and that the Hida family 𝐟\mathbf{f} admits a specialisation fkf_{k} of weight k2(modp1)k\equiv 2\pmod{p-1} and trivial nebentypus. Then the group H~f,Iw1(K,𝐀)\tilde{H}^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger}) is a finitely generated cotorsion Λ\Lambda_{\mathcal{R}}-module and there is an equality

(Lp(𝐟/K))=CharΛ(Hf,Iw1(K,𝐀))\left(L_{p}(\mathbf{f}/K)\right)=\mathrm{Char}_{\Lambda_{\mathcal{R}}}\left(H^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\vee}\right)

of ideals in Λ\Lambda_{\mathcal{R}}.

Proof.

That H~f,Iw1(K,𝐀)\tilde{H}^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger}) is finitely generated follows easily from the topological Nakayama’s Lemma. The proof of [CKL17, Theorem 5.3] shows the inclusion of the characteristic ideal in the ideal generated by the pp-adic LL-function (see in particular the last displayed equation in the proof of [CKL17, Theorem 5.3]). More precisely, by [KL22, Theorem 11.1] we know that Lp(𝐟/K)L_{p}(\mathbf{f}/K) is equal, up to units of 𝕀\mathbb{I}, to the self-dual twist of the restriction of Skinner–Urban’s three-variable pp-adic LL-function to the anticyclotomic line (see [KL22, §4.4]). Combining [SU14, Theorem 3.26 ] and [Rub11, Lemma 1.2], we see that the inclusion of CharΛ(Hf,Iw1(K,𝐀))\mathrm{Char}_{\Lambda_{\mathcal{R}}}\left(H^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\vee}\right) in (Lp(𝐟/K))\left(L_{p}(\mathbf{f}/K)\right) holds. To get the equality, it suffices to establish equality for some classical specialisation, which follows in our setting from [CKL17, Corollary 3]. Finally, since Lp(𝐟/K)0L_{p}(\mathbf{f}/K)\neq 0, it follows that Hf,Iw1(K,𝐀)H^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger}) is Λ\Lambda_{\mathcal{R}}-cotorsion. ∎

As a corollary of Theorem 4.1, we obtain a result in the direction of [LV11, Conjecture 9.5], a definite version of the horizontal non-vanishing conjecture of Howard [How07, Conjecture 3.4.1]. Denote χtriv:[[Γ]]\chi_{\mathrm{triv}}:\mathcal{R}[[\Gamma_{\infty}]]\rightarrow\mathcal{R} the morphism associate with the trivial character of Γ\Gamma_{\infty}, and define

(8) 𝒥0=χtriv(Θ(𝐟)).\mathcal{J}_{0}=\chi_{\mathrm{triv}}\left(\Theta_{\infty}(\mathbf{f})\right).
Corollary 4.2.

Let the assumptions be as in Theorem 4.1. If H~f1(K,𝐓)\tilde{H}^{1}_{f}(K,\mathbf{T}^{\dagger}) is a torsion \mathcal{R} module, then 𝒥00\mathcal{J}_{0}\neq 0.

Proof.

Since H~f1(K,𝐓)\tilde{H}^{1}_{f}(K,\mathbf{T}^{\dagger}) is a torsion \mathcal{R}-module, it follows from [LV14, Corollary 5.5] that H~f1(K,Vfκ)=0\tilde{H}^{1}_{f}(K,V_{f_{\kappa}}^{\dagger})=0 for all but finitely many arithmetic character κ\kappa, where H~f1(K,Vfκ)\tilde{H}^{1}_{f}(K,V_{f_{\kappa}}^{\dagger}) is the extended Bloch–Kato Selmer group of VfκV_{f_{\kappa}}^{\dagger}. By [Nek06, Proposition 12.7.13.4(i)], this implies that H~f2(K,𝐓)\tilde{H}^{2}_{f}(K,\mathbf{T}^{\dagger}) is a torsion \mathcal{R}-module. Poitou–Tate global duality [Nek06, §0.1] implies then that Hf1(K,𝐀)H^{1}_{f}(K,\mathbf{A}^{\dagger})^{\vee} is also a torsion \mathcal{R}-module.

Let II be the kernel of χtriv\chi_{\mathrm{triv}}. By Theorem 3.5, the kernel and cokernel of the map

Hf,Iw1(K,𝐀)/IHf,Iw1(K,𝐀)Hf1(K,𝐀)H^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\vee}/IH^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\vee}\longrightarrow H^{1}_{f}(K,\mathbf{A}^{\dagger})^{\vee}

are torsion \mathcal{R}-modules. Since Hf1(K,𝐀)H^{1}_{f}(K,\mathbf{A}^{\dagger})^{\vee} is a torsion \mathcal{R}-module, it follows that

Hf,Iw1(K,𝐀)/IHf,Iw1(K,𝐀)H^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\vee}/IH^{1}_{f,\mathrm{Iw}}(K_{\infty},\mathbf{A}^{\dagger})^{\vee}

is also a torsion \mathcal{R}-module, and its characteristic power series is then a non-zero element of \mathcal{R}. By Theorem 4.1 we then have Lp(𝐟/K)(χtriv)0L_{p}(\mathbf{f}/K)(\chi_{\mathrm{triv}})\neq 0. The result follows now from Definition 3.3 and (8). ∎

References

  • [BD96a] M. Bertolini and H. Darmon, Heegner points on Mumford-Tate curves, Invent. Math. 126 (1996), no. 3, 413–456. MR 1419003 (97k:11100)
  • [BD96b] by same author, Heegner points on Mumford-Tate curves, Invent. Math. 126 (1996), no. 3, 413–456. MR MR1419003 (97k:11100)
  • [BD05] by same author, Iwasawa’s Main Conjecture for elliptic curves over anticyclotomic p\mathbb{Z}_{p}-extensions, Ann. of Math. (2) 162 (2005), no. 1, 1–64.
  • [Bou98] Nicolas Bourbaki, Commutative algebra. Chapters 1–7, Elements of Mathematics (Berlin), Springer-Verlag, Berlin, 1998, Translated from the French, Reprint of the 1989 English translation. MR 1727221
  • [CKL17] Francesc Castella, Chan-Ho Kim, and Matteo Longo, Variation of anticyclotomic Iwasawa invariants in Hida families, Algebra Number Theory 11 (2017), no. 10, 2339–2368. MR 3744359
  • [CL16] Francesc Castella and Matteo Longo, Big Heegner points and special values of LL-series, Ann. Math. Qué. 40 (2016), no. 2, 303–324. MR 3529184
  • [GS93] Ralph Greenberg and Glenn Stevens, pp-adic LL-functions and pp-adic periods of modular forms, Invent. Math. 111 (1993), no. 2, 407–447. MR 1198816 (93m:11054)
  • [Hid86] Haruzo Hida, Galois representations into GL2(𝐙p[[X]]){\rm GL}_{2}({\bf Z}_{p}[[X]]) attached to ordinary cusp forms, Invent. Math. 85 (1986), no. 3, 545–613. MR 848685 (87k:11049)
  • [How07] Benjamin Howard, Variation of Heegner points in Hida families, Invent. Math. 167 (2007), no. 1, 91–128. MR 2264805 (2007h:11067)
  • [KL22] Chan-Ho Kim and Matteo Longo, An explicit comparison of anticyclotomic pp-adic LL-functions for Hida families, preprint https://arxiv.org/pdf/2205.08080.pdf (2022).
  • [LV11] Matteo Longo and Stefano Vigni, Quaternion algebras, Heegner points and the arithmetic of Hida families, Manuscripta Math. 135 (2011), no. 3-4, 273–328. MR 2813438 (2012g:11114)
  • [LV14] by same author, Vanishing of central values and central derivatives in Hida families, Annali SNS Pisa (2014), no. 13, 859–888.
  • [LV17] by same author, A refined Beilinson-Bloch conjecture for motives of modular forms, Trans. Amer. Math. Soc. 369 (2017), no. 10, 7301–7342. MR 3683110
  • [Mat89] Hideyuki Matsumura, Commutative ring theory, second ed., Cambridge Studies in Advanced Mathematics, vol. 8, Cambridge University Press, Cambridge, 1989, Translated from the Japanese by M. Reid. MR 1011461 (90i:13001)
  • [Nek06] Jan Nekovář, Selmer complexes, Astérisque (2006), no. 310, viii+559. MR 2333680
  • [Och00] Tadashi Ochiai, Control theorem for Bloch-Kato’s Selmer groups of pp-adic representations, J. Number Theory 82 (2000), no. 1, 69–90. MR 1755154
  • [Och01] by same author, Control theorem for Greenberg’s Selmer groups of Galois deformations, J. Number Theory 88 (2001), no. 1, 59–85. MR 1825991
  • [Och06] by same author, On the two-variable Iwasawa main conjecture, Compos. Math. 142 (2006), no. 5, 1157–1200.
  • [Rub11] K. Rubin, Euler systems and Kolyvagin systems, Arithmetic of LL-functions, IAS/Park City Math. Ser., vol. 18, Amer. Math. Soc., Providence, RI, 2011, pp. 449–499.
  • [SU14] Christopher Skinner and Eric Urban, The Iwasawa main conjectures for GL2\rm GL_{2}, Invent. Math. 195 (2014), no. 1, 1–277. MR 3148103