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

On the μ\mu equals zero Conjecture for Fine Selmer groups in Iwasawa theory

Shaunak V. Deo Department of Mathematics, Indian Institute of Science, Bangalore 560012, India [email protected] Anwesh Ray Centre de recherches mathématiques, Université de Montréal, Pavillon André-Aisenstadt, 2920 Chemin de la tour, Montréal (Québec) H3T 1J4, Canada [email protected]  and  R. Sujatha Department of Mathematics
University of British Columbia
Vancouver BC, Canada V6T 1Z2
[email protected]
Abstract.

We study the Iwasawa theory of the fine Selmer group associated to Galois representations arising from modular forms. The vanishing of the μ\mu-invariant is shown to follow in some cases from a natural property satisfied by Galois deformation rings. We outline conditions under which the μ=0\mu=0 conjecture is shown to hold for various Galois representations of interest.

Key words and phrases:
Iwasawa μ\mu-invariant, fine Selmer groups, adjoint representations, deformations of Galois representations.
2010 Mathematics Subject Classification:
11R23 (primary); 11F80, 11G05, 11F11 (secondary)

1. Introduction

Let FF be a number field and pp be an odd prime. The pp-primary roots of unity, considered as a module over the absolute Galois group of FF, is denoted by μp.\mu_{p^{\infty}}. Set FnF_{n} to be the unique field contained in F(μp)F(\mu_{p^{\infty}}) such that [Fn:F]=pn[F_{n}\mathrel{\mathop{\mathchar 58\relax}}F]=p^{n}. The cyclotomic p\mathbb{Z}_{p}-extension FcycF_{cyc} of FF is the subextension of F(μp)F(\mu_{p^{\infty}}) obtained as the union of the fields Fn.F_{n}. In his seminal work [Iwa73], Iwasawa studied the growth of the pp-part of the class groups over certain towers of number fields. In particular, for the cyclotomic p\mathbb{Z}_{p}-extension of FF, Iwasawa proved that there are invariants μ,λ0\mu,\lambda\in\mathbb{Z}_{\geq 0} and ν\nu\in\mathbb{Z} such that

#Clp(Fn)=ppnμ+λn+ν for n0,\#\text{Cl}_{p}(F_{n})=p^{p^{n}\mu+\lambda n+\nu}\text{ for }n\gg 0,

where Clp(Fn)\text{Cl}_{p}(F_{n}) is the pp-Sylow subgroup of the class group of FnF_{n}. The invariants μ\mu and λ\lambda are the Iwasawa invariants associated to the p\mathbb{Z}_{p}-extension Fcyc/FF_{\operatorname{cyc}}/F.

The genesis of Iwasawa theory arose from the study of these invariants for a broad class of p\mathbb{Z}_{p}-extensions of a number field. Iwasawa formulated various conjectures about these invariants. Amongst them is a famous conjecture, henceforth referred to as Iwasawa’s μ=0\mu=0 conjecture, which asserts that the μ\mu-invariant of the cyclotomic p\mathbb{Z}_{p}-extension Fcyc/FF_{\operatorname{\mathrm{cyc}}}/F of any number field FF is zero. Let \mathcal{F} be the maximal abelian, unramified, pro-pp extension of FcycF_{\operatorname{\mathrm{cyc}}}, let X(Fcyc)=Gal(/Fcyc)X(F_{\operatorname{\mathrm{cyc}}})=\operatorname{Gal}(\mathcal{F}/F_{\operatorname{\mathrm{cyc}}}) and let Γ=Gal(Fcyc/F)\Gamma=\operatorname{Gal}(F_{\operatorname{\mathrm{cyc}}}/F). Then an equivalent formulation of this conjecture is the statement that X(Fcyc)X(F_{\operatorname{\mathrm{cyc}}}) is a finitely generated p\mathbb{Z}_{p}-module. When FF is an abelian number field, the μ\mu-invariant is known to vanish by the work of Ferrero and Washington [FW79]. We remark that no general result in this direction is known for any broader class of non-abelian number fields.

In the modern language of Galois representations and arithmetic geometry, Iwasawa’s μ=0\mu=0 conjecture can be interpreted in terms of structural invariants associated to certain modules over the Iwasawa algebra of Gal(Fcyc/F)\operatorname{Gal}(F_{\operatorname{cyc}}/F). These modules arise naturally in the context of the Galois representation associated to the Tate motive μp.\mu_{p^{\infty}}. Subsequently, Iwasawa theory evolved to cover the study of a range of modules over Iwasawa algebras that arise from other Galois representations. An important instance of this evolution is the work of Mazur [Maz72] who initiated the Iwasawa theory of elliptic curves and abelian varieties. In this pathbreaking work, Mazur studied the pp-primary Selmer group of an elliptic curve (or abelian variety) E/FE_{/F} which has good, ordinary reduction at all primes dividing pp over FcycF_{\operatorname{cyc}}, and defined analogous Iwasawa μ\mu and λ\lambda-invariants in this context. The μ\mu-invariant of the pp-primary Selmer group need not vanish. For example, for E/=X0(11)E_{/\mathbb{Q}}=X_{0}(11), the μ\mu-invariant of the 55-primary Selmer group is positive. Given the analogy between class groups and Selmer groups of elliptic curves, it is natural to frame an analogue of Iwasawa’s μ=0\mu=0 conjecture in this setting which we describe below.

Let EE be an elliptic curve defined over a number field FF and pp be a prime above which EE has potential good reduction. Set Γ:=Gal(Fcyc/F)\Gamma\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Gal}(F_{\operatorname{cyc}}/F) and let Λ\Lambda denote the corresponding Iwasawa algebra, which is defined to be the completed group algebra Λ:=limnp[Γ/Γpn]\Lambda\mathrel{\mathop{\mathchar 58\relax}}=\varprojlim_{n}\mathbb{Z}_{p}[\Gamma/\Gamma^{p^{n}}]. Then, the pp-primary fine Selmer group consists of the Galois cohomology classes in H1(Gal(F¯/Fcyc),E[p])H^{1}\left(\operatorname{Gal}(\bar{F}/F_{\operatorname{cyc}}),E[p^{\infty}]\right) that are trivial at all primes above pp and the primes at which EE has bad reduction. The fine Selmer group is believed to be cotorsion as an Iwasawa module and this is known to be true for elliptic curves over \mathbb{Q} (see [CS05]). In contrast, the usual Selmer group need not satisfy this condition unless the elliptic curve is assumed to have ordinary reduction at all primes above pp.

The properties of the fine Selmer group were systematically studied by Coates and the third named author who showed that the Iwasawa μ\mu-invariant of the fine Selmer group is related to the classical Iwasawa μ\mu-invariant of the cyclotomic p\mathbb{Z}_{p}-extension of the number field F(E[p])F(E[p]) cut out by the pp-torsion in EE. We refer to [CS05, Corollary 3.5] for the precise statement of this result. This relationship shows that the vanishing of the μ\mu-invariant of the fine Selmer group is a consequence of Iwasawa’s classical μ=0\mu=0 conjecture. This led Coates and the third named author to formulate the μ=0\mu=0 conjecture for the fine Selmer group of an elliptic curve (see [CS05, Conjecture A]), as the analogue of Iwasawa’s μ=0\mu=0 conjecture for Galois representations associated to elliptic curves.

Iwasawa theory of pp-adic Galois representations is currently a central theme in number theory. The fine Selmer group can be defined in the more general context of pp-adic Galois representations (see Definition 2.1 and Definition 2.2). We remark that the μ=0\mu=0 conjecture can be formulated in this broader framework (see Conjecture 2.6 and [LS18, Conjecture A, p.74]). It has interesting applications to the study of Selmer groups associated to Galois representations of interest. The goal of this article is to study the μ=0\mu=0 conjecture in this broader setting. The main results formulate explicit conditions under which the μ\mu-invariant vanishes for the fine Selmer groups associated to important classes of pp-adic Galois representations. The cases of interest are the Galois representations arising from elliptic curves defined over number fields, Hecke eigencuspforms of weight at least 22, Tate twists of Artin representations and Tate twists of the adjoint representations associated to modular Galois representations.

In the study of Galois representations, the theme of pp-adic variation plays a key role. More precisely, the deformation theory of Galois representations, introduced and developed by Mazur in [Maz89], allows us to study pp-adic families of lifts of a fixed mod-pp Galois representation. Hida [Hid00] studied the relationship between certain modular deformation rings defined over the cyclotomic p\mathbb{Z}_{p}-extension of \mathbb{Q} and the Iwasawa theory of the classical Selmer group of the adjoint representation of a modular form. In this article, the μ=0\mu=0 conjecture for the fine Selmer group of the first Tate twist of the adjoint representation of a pp-adic Galois representation ρ\rho is related to unobstructedness of the residual representation of ρ\rho (see Definition 4.1). Note that the notion of unobstructedness appears naturally in Galois deformation theory in connection with the structure of the universal deformation ring (in the sense of Mazur). This provides a new approach in investigating the fine Selmer group. It is our belief that the linkage established here will help advance the current state of art in both the Iwasawa theory of the fine Selmer group as well as deformation theory of Galois representations. In particular, our methods provide a large class of new examples of fine Selmer groups of Galois representations for which the μ\mu-invariant vanishes.

Let us now briefly outline the results proved in this article. Section 2 is preliminary in nature. In section 3, we establish a criterion for the vanishing of the μ\mu-invariant of the fine Selmer group of a Galois representation (see Theorem 3.7). To be precise, we prove that the μ\mu-invariant of the fine Selmer group of a pp-adic Galois representation ρ\rho vanishes if the second global Galois cohomology group of the first Tate twist of the dual of the residual representation of ρ\rho vanishes. Thus, this criterion is determined purely in terms of the residual representation. Theorem 3.8 gives explicit conditions for the vanishing of the μ\mu-invariant of the fine Selmer group, which we specialize to the case of elliptic curves in Theorem 3.9.

Section 4 establishes the linkage between deformation theory of Galois representations and the vanishing of the μ\mu-invariant of the fine Selmer group. We study the fine Selmer group of the first Tate twist of the n2n^{2}-dimensional adjoint Galois representation Adρ\operatorname{Ad}\rho associated to an nn-dimensional Galois representation ρ\rho. An important class of such Galois representations arises naturally from automorphic forms. More precisely, Theorem 4.3 shows that the μ\mu-invariant vanishes for the fine Selmer group of Adρ(1)\operatorname{Ad}\rho(1) if ρ\rho is unobstructed. This concept was initially introduced by Mazur in [Maz89], and it was shown by Weston in [Wes04] that Galois representations arising from modular forms are unobstructed under some explicit conditions.

Let ff be a normalized Hecke newform of weight k2k\geq 2. Given a prime 𝔭\mathfrak{p} in the field of Fourier coefficients of ff, let ρf,𝔭\rho_{f,\mathfrak{p}} be the associated 22-dimensional 𝔭\mathfrak{p}-adic Galois representation. Theorem 4.6 shows that the μ\mu-invariant of the fine Selmer group of Adρf,𝔭(1)\operatorname{Ad}\rho_{f,\mathfrak{p}}(1) vanishes for a set of primes 𝔭\mathfrak{p} of Dirichlet density 11. Moreover, if k>2k>2, then it is shown to vanish for all but finitely many primes 𝔭\mathfrak{p}. Furthermore, if the level of ff is squarefree, then there is an explicit set of primes 𝔭\mathfrak{p} outside of which the associated μ\mu-invariant is 0. When k=2k=2 and ff has trivial nebentypus, our results imply that the μ\mu-invariant of the fine Selmer group of Sym2ρf,𝔭\operatorname{Sym}^{2}\rho_{f,\mathfrak{p}} vanishes for a set of primes 𝔭\mathfrak{p} of Dirichlet density 11. We also prove a similar result in the setting of Hilbert modular forms (see Theorem 4.11). On the other hand, we prove that if EE is a rational elliptic curve with squarefree conductor, then there are infinitely many primes pp such that the μ\mu-invariant of the fine Selmer group of Ad0ρE,p\operatorname{Ad}^{0}\rho_{E,p} vanishes (see Theorem 4.12). Here, ρE,p\rho_{E,p} is the 22-dimensional pp-adic Galois representation attached to EE and Ad0ρE,p\operatorname{Ad}^{0}\rho_{E,p} is the subrepresentation of AdρE,p\operatorname{Ad}\rho_{E,p} consisting of matrices with trace 0.

We then turn our attention to Artin representations. Note that we can realize an Artin representation as a pp-adic Galois representation for all primes pp. We prove that if ρ\rho is an Artin representation such that H0(Gal(/),ρ)=0H^{0}(\operatorname{Gal}(\mathbb{C}/\mathbb{R}),\rho)=0, then the μ\mu-invariants of the fine Selmer groups of the first Tate twists of the pp-adic realizations of the dual of ρ\rho vanish for all but finitely many primes pp (see Theorem 4.14).

We also connect the μ=0\mu=0 conjecture with the notion of neatness introduced by Mazur in [Maz89] to prove that the μ=0\mu=0 conjecture is true for first Tate twist of the adjoint representation of a pp-adic Galois representation ρ\rho whose residual representation is neat (see Corollary 4.19). Leveraging a construction of Mazur and Böckle, we end section 4 with explicit families of odd and even S3S_{3}-representations such that the μ=0\mu=0-conjecture holds for the fine Selmer groups of first Tate twists of all their characteristic 0 lifts.

In section 5, we study the vanishing of the μ\mu-invariant of the fine Selmer group associated to 22-dimensional residually dihedral Galois representations. The conditions are stated purely in terms of the residual representation and the set of primes at which the characteristic zero representation ramifies. In Theorem 5.6 we give explicit conditions for the vanishing of the μ\mu-invariant and obtain a large class of examples for which our results apply. Finally, we remark that our results are illustrated with concrete numerical examples.

Acknowledgements

The first named author would like to thank Gabor Wiese for helping with computation of examples presented in §5. The second named author is grateful to Ravi Ramakrishna for introducing him to the fascinating subject of Galois deformation theory, and would also like to thank Tom Weston and Jeffrey Hatley for helpful discussions along the way. His research is supported by the CRM-Simons bridge postdoctoral fellowship. The third named author gratefully acknowledges support from NSERC Discovery grant 2019-03987. The first and third named authors would also like to thank the online program on ‘Elliptic curves and the special values of LL-functions’ (code: ICTS/ECL2021/8) held at ICTS Bangalore, for providing the motivation and opportunity to begin discussions relevant to the work presented in this article. We would also like to thank the anonymous referee for providing many comments and suggestions which helped in improving the exposition.

2. Preliminaries

Throughout, pp is an odd prime number and FF is a number field. Set SpS_{p} to be the set of primes of FF above pp. Let SS be a finite set of primes of FF and assume that SS contains SpS_{p} and all archimedean primes of FF. Throughout, fix an algebraic closure F¯\bar{F} of FF. Let FSF_{S} be the maximal extension of FF contained in F¯\bar{F} in which all primes vSv\notin S are unramified. Set GF,S\operatorname{G}_{F,S} to be the Galois group Gal(FS/F)\operatorname{Gal}(F_{S}/F), identified with the maximal quotient of Gal(F¯/F)\operatorname{Gal}(\bar{F}/F) in which the primes vSv\notin S are unramified. Given a field LFSL\subset F_{S}, we set

Hi(FS/L,):=Hi(Gal(FS/L),).H^{i}\left(F_{S}/L,\cdot\right)\mathrel{\mathop{\mathchar 58\relax}}=H^{i}\left(\operatorname{Gal}(F_{S}/L),\cdot\right).

For a number field LL contained in FSF_{S}, set

Kvi(/L):=w|vHi(Lw,),K^{i}_{v}(\cdot/L)\mathrel{\mathop{\mathchar 58\relax}}=\bigoplus_{w|v}H^{i}(L_{w},\cdot),

where ww runs through the primes of LL above vv. The fine Selmer group associated with a Galois module is obtained by imposing the strict local condition at each prime that lies above the primes in SS.

Definition 2.1.

Let MM be a discrete pp-primary GF,S\operatorname{G}_{F,S} module and LL be a number field contained in FSF_{S}. The fine Selmer group over LL associated to MM is defined as follows

p(M/L):=ker{H1(FS/L,M)vSKv1(M/L)}.\mathcal{R}_{p^{\infty}}(M/L)\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{ker}\left\{H^{1}(F_{S}/L,M)\longrightarrow\bigoplus_{v\in S}K^{1}_{v}(M/L)\right\}.

Given an infinite extension FS\mathcal{L}\subset F_{S}, set

p(M/):=limLp(M/L),\mathcal{R}_{p^{\infty}}(M/\mathcal{L})\mathrel{\mathop{\mathchar 58\relax}}=\varinjlim_{L}\mathcal{R}_{p^{\infty}}(M/L),

where LL runs over all number fields contained in \mathcal{L}.

Note that, in general, the definition of p(M/)\mathcal{R}_{p^{\infty}}(M/\mathcal{L}) depends on the choice of SS. However, if \mathcal{L} contains the cyclotomic p\mathbb{Z}_{p}-extension of FF, then p(M/)\mathcal{R}_{p^{\infty}}(M/\mathcal{L}) does not depend on the choice of SS (see [SW18, Lemma 3.2]).

For every integer n1n\geq 1, let μpn\mu_{p^{n}} be the group of pnp^{n}-th roots of unity contained in F¯\bar{F} and μp:=n1μpn\mu_{p^{\infty}}\mathrel{\mathop{\mathchar 58\relax}}=\cup_{n\geq 1}\mu_{p^{n}}. Let F(μpn)F(\mu_{p^{n}}) (resp. F(μp)F(\mu_{p^{\infty}})) be the field extension of FF generated by μpn\mu_{p^{n}} (resp. μp\mu_{p^{\infty}}). The cyclotomic p\mathbb{Z}_{p}-extension of FF is the unique p\mathbb{Z}_{p}-extension of FF contained in F(μp)F(\mu_{p^{\infty}}), and is denoted by FcycF_{\operatorname{cyc}}. Let KK be a finite extension of p\mathbb{Q}_{p} and 𝒪\mathcal{O} be its valuation ring. Given an integral Galois representation, ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}), let 𝐓ρ\mathbf{T}_{\rho} be the underlying 𝒪\mathcal{O}-lattice on which GF,S\operatorname{G}_{F,S} acts by ρ\rho and set 𝐀(ρ):=𝐓ρ𝒪(K/𝒪)\mathbf{A}(\rho)\mathrel{\mathop{\mathchar 58\relax}}=\mathbf{T}_{\rho}\otimes_{\mathcal{O}}\left(K/\mathcal{O}\right).

Definition 2.2.

With respect to notation above, the fine Selmer group over FcycF_{\operatorname{cyc}} associated with ρ\rho is p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}).

Thus [SW18, Lemma 3.2] implies that the definition of p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) does not depend on the choice of SS.

Set Γ:=Gal(Fcyc/F)\Gamma\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Gal}(F_{\operatorname{cyc}}/F) and let

Λ:=𝒪Γ=limn𝒪[Γ/Γpn]\Lambda\mathrel{\mathop{\mathchar 58\relax}}=\mathcal{O}\llbracket\Gamma\rrbracket=\varprojlim_{n}\mathcal{O}[\Gamma/\Gamma^{p^{n}}]

be the associated Iwasawa algebra. Given a finitely generated Λ\Lambda-module 𝔐\mathfrak{M}, there is a map of Λ\Lambda-modules

𝔐Λα(i=1sΛ/(pμi))(j=1tΛ/(fj(T)))\mathfrak{M}\longrightarrow\Lambda^{\alpha}\oplus\left(\bigoplus_{i=1}^{s}\Lambda/(p^{\mu_{i}})\right)\oplus\left(\bigoplus_{j=1}^{t}\Lambda/(f_{j}(T))\right)

with finite kernel and cokernel. Here, α=rankΛ𝔐\alpha=\operatorname{rank}_{\Lambda}\mathfrak{M} and each power series fj(T)f_{j}(T) is a distinguished polynomial. In other words, fj(T)Λf_{j}(T)\in\Lambda is a monic polynomial all of whose non-leading coefficients are non-units in 𝒪\mathcal{O}. The μ\mu-invariant is the sum iμi\sum_{i}\mu_{i}. Given an integral Galois representation

ρ:GF,SGLn(𝒪),\rho\mathrel{\mathop{\mathchar 58\relax}}G_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}),

set μfn(ρ)=μfn(ρ/Fcyc)\mu^{\operatorname{fn}}(\rho)=\mu^{\operatorname{fn}}(\rho/F_{\operatorname{cyc}}) to be the μ\mu-invariant of Y(𝐀(ρ)/Fcyc)Y(\mathbf{A}(\rho)/F_{\operatorname{cyc}}), the Pontryagin dual of the fine Selmer group p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}). The following is Conjecture A in [CS05].

Conjecture 2.3 (Coates-Sujatha).

Let FF be a number field, EE be an elliptic curve over FF, and SS be a finite set of primes of FF containing SpS_{p}, all archimedean primes and all primes at which EE has bad reduction. If

ρ:GF,SGL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{F,S}\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p})

is the pp-adic Galois representation arising from EE, then μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

The following result provides evidence for the above conjecture.

Theorem 2.4 (Coates-Sujatha).

Let FF be a number field, SS be a finite set of primes of FF containing SpS_{p} and all archimedean primes and

ρ:GF,SGL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{F,S}\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p})

be the pp-adic Galois representation associated to an elliptic curve E/FE_{/F}. Suppose that there is a number field L/FL/F such that

  1. (1)

    LF(E[p])L\subset F(E[p^{\infty}]),

  2. (2)

    the extension F(E[p])/LF(E[p^{\infty}])/L is pro-pp.

Then, μfn(ρ)\mu^{\operatorname{fn}}(\rho) is equal to 0 if and only if the Iwasawa’s μ=0\mu=0 conjecture holds for the cyclotomic p\mathbb{Z}_{p}-extension Lcyc/LL_{\operatorname{cyc}}/L.

Proof.

This result is [CS05, Corollary 3.5]. ∎

Theorem 2.5.

Let FF be an abelian extension of \mathbb{Q}, and

ρ:Gal(¯/)GL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q})\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p})

be the pp-adic Galois representation associated to an elliptic curve E/E_{/\mathbb{Q}}. Assume that the residual representation ρ¯\bar{\rho} is reducible, i.e., there are characters

φ1,φ2:Gal(¯/)𝔽p×\varphi_{1},\varphi_{2}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q})\rightarrow\mathbb{F}_{p}^{\times}

such that ρ¯=(φ10φ2)\bar{\rho}=\left({\begin{array}[]{cc}\varphi_{1}&\ast\\ 0&\varphi_{2}\\ \end{array}}\right). Let SS be a finite set of primes of FF containing SpS_{p}, all archimedean primes and all primes of bad reduction of E/FE_{/F}. Let ρF,S\rho_{F,S} be ρ|GF,S\rho|_{G_{F,S}} i.e. ρF,S\rho_{F,S} is the representation ρ\rho when viewed as a representation GF,SG_{F,S}. Then, μfn(ρF,S/Fcyc)\mu^{\operatorname{fn}}(\rho_{F,S}/F_{\operatorname{cyc}}) is equal to 0.

Proof.

Let (φi)\mathbb{Q}(\varphi_{i}) be the extension of \mathbb{Q} which is fixed by kerφi\operatorname{ker}\varphi_{i}. Note that (φi)\mathbb{Q}(\varphi_{i}) is an abelian extension of \mathbb{Q}. Let LL be the composite of the fields FF, (φ1)\mathbb{Q}(\varphi_{1}) and (φ2)\mathbb{Q}(\varphi_{2}). Since it is a composite of abelian extensions of \mathbb{Q}, it follows that L/L/\mathbb{Q} is abelian. By the result of Ferrero and Washington [FW79], the Iwasawa μ\mu-invariant vanishes for any abelian extension. Note that by construction, the extension F(E[p])/LF\left(E[p^{\infty}]\right)/L is pro-pp. Since the conditions of Theorem 2.4 are met, the result follows. ∎

We remark that the representation ρ¯\bar{\rho} in the statement of Theorem 2.5 is reducible if EE admits a rational pp-isogeny. In particular, if E()[p]0E(\mathbb{Q})[p]\neq 0, then the residual representation is of the form ρ¯=(10χ¯p)\bar{\rho}=\left({\begin{array}[]{cc}1&\ast\\ 0&\bar{\chi}_{p}\\ \end{array}}\right), where χ¯p\bar{\chi}_{p} is the mod-pp cyclotomic character. Although the μ=0\mu=0 conjecture for fine Selmer groups was originally stated in [CS05] for the 22-dimensional Galois representations associated to elliptic curves, the following generalized conjecture is expected to hold (cf. [LS18, Conjecture A, p.74]):

Conjecture 2.6.

Let pp be a prime number, FF be a number field and 𝒪\mathcal{O} be the ring of integers of a finite extension of p\mathbb{Q}_{p}. Let SS be a finite set of primes of FF containing the primes that lie above pp and ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be an integral Galois representation. Then, μfn(ρ)\mu^{\operatorname{fn}}(\rho) is equal to 0.

3. A Criterion for μ=0\mu=0 for the fine Selmer group of a Galois representation

In this section, we shall discuss various criteria for the vanishing of the μ\mu-invariant of the fine Selmer group associated to a Galois representation. As in the previous section, let pp be an odd prime, FF be a number field and SS be a finite set of primes containing all archimedean primes and the primes above pp. Let 𝒪\mathcal{O} be the valuation ring of a finite extension K/pK/\mathbb{Q}_{p} with uniformizer ϖ\varpi and let 𝔽:=𝒪/ϖ\mathbb{F}\mathrel{\mathop{\mathchar 58\relax}}=\mathcal{O}/\varpi be its residue field. Fix an integral Galois representation ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) on an 𝒪\mathcal{O}-lattice 𝐓ρ\mathbf{T}_{\rho}. Let ρ¯\bar{\rho} denote the residual representation ρmodϖ\rho\mod{\varpi} and set 𝐕ρ¯=𝐓ρ𝒪𝔽\mathbf{V}_{\bar{\rho}}=\mathbf{T}_{\rho}\otimes_{\mathcal{O}}\mathbb{F} to be the underlying vector space on which GF,S\operatorname{G}_{F,S} acts via ρ¯\bar{\rho}. Setting 𝐕ρ=𝐓ρ𝒪K\mathbf{V}_{\rho}=\mathbf{T}_{\rho}\otimes_{\mathcal{O}}K, recall that 𝐀(ρ)=𝐕ρ/𝐓ρ\mathbf{A}(\rho)=\mathbf{V}_{\rho}/\mathbf{T}_{\rho}. We identify 𝐕ρ¯\mathbf{V}_{\bar{\rho}} with 𝐀(ρ)[ϖ]\mathbf{A}(\rho)[\varpi]. Denote by ρ\rho^{\wedge} the dual of the representation ρ\rho and by χp\chi_{p} be the pp-adic cyclotomic character. We set ρ\rho^{*} to denote the representation ρχp\rho^{\wedge}\otimes\chi_{p}. Recall that χ¯p\bar{\chi}_{p} is the mod-pp cyclotomic character, and ρ¯\bar{\rho}^{\wedge} is the dual representation to ρ¯\bar{\rho}. Identify ρ¯\bar{\rho}^{*} with ρ¯χ¯p\bar{\rho}^{\wedge}\otimes\bar{\chi}_{p}. We note that 𝐓ρ\mathbf{T}_{\rho^{*}} is isomorphic to the twist 𝐓ρ𝒪𝒪(χp)\mathbf{T}_{{\rho}^{\wedge}}\otimes_{\mathcal{O}}\mathcal{O}(\chi_{p}). Therefore, to summarize, ρ\rho and ρ\rho^{*} are integral representations acting on the 𝒪\mathcal{O}-modules 𝐓ρ\mathbf{T}_{\rho} and 𝐓ρ:=Hom(𝐓ρ,𝒪(χp))\mathbf{T}_{\rho^{*}}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Hom}\left(\mathbf{T}_{\rho},\mathcal{O}(\chi_{p})\right), respectively. The representations ρ¯\bar{\rho} and ρ¯\bar{\rho}^{*} act on the 𝔽\mathbb{F}-vector spaces 𝐕ρ¯\mathbf{V}_{\bar{\rho}} and 𝐕ρ¯:=Hom(𝐕ρ¯,𝔽(χ¯p))\mathbf{V}_{\bar{\rho}^{*}}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Hom}\left(\mathbf{V}_{\bar{\rho}},\mathbb{F}(\bar{\chi}_{p})\right), respectively. We also note that 𝐀(ρ)\mathbf{A}(\rho) is a pp-divisible K/𝒪K/\mathcal{O}-module which is identified with 𝐓ρ𝒪K/𝒪\mathbf{T}_{\rho}\otimes_{\mathcal{O}}K/\mathcal{O}.

Conjecture 3.1 (Weak Leopoldt Conjecture for ρ\rho).

With respect to notation above, H2(FS/Fcyc,𝐀(ρ))=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho))=0.

We refer the reader to [PR00, Appendix B] for a list of cases where the Weak Leopoldt Conjecture is known.

We prove a criterion which shows that the vanishing of μfn(ρ)\mu^{\operatorname{fn}}(\rho) depends only on the residual representation ρ¯\bar{\rho}. In the special case when ρ\rho is the Galois representation on the pp-adic Tate module of an elliptic curve, a brief sketch of the proof is given in [Suj10, Proposition 4.6].

Theorem 3.2.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation such that the weak Leopoldt conjecture holds for ρ\rho^{*}. Then, the following are equivalent.

  1. (1)

    The fine Selmer group p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda and μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

  2. (2)

    The cohomology group H2(FS/Fcyc,𝐕ρ¯)H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{{\bar{\rho}}^{*}}) is equal to 0.

The proof of the above theorem is provided later in this section, following Lemma 3.5. The vanishing of H2(FS/Fcyc,𝐕ρ¯)H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{{\bar{\rho}}^{*}}) is studied in greater detail in this article. Given a GF,S\operatorname{G}_{F,S}-module MM, the Iwasawa cohomology group HIwi(FS/Fcyc,M)H^{i}_{\operatorname{Iw}}\left(F_{S}/F_{\operatorname{cyc}},M\right) is defined to be the inverse limit limnHi(FS/Fn,M)\varprojlim_{n}H^{i}\left(F_{S}/F_{n},M\right) with respect to corestriction maps. Given a Λ\Lambda-module MM, we denote the Pontryagin dual by M:=Homp(M,p/p)M^{\vee}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Hom}_{\mathbb{Z}_{p}}\left(M,\mathbb{Q}_{p}/\mathbb{Z}_{p}\right). Recall that the fine Selmer group of 𝐀(ρ)\mathbf{A}(\rho) over FcycF_{\operatorname{cyc}} is p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}), and Y(𝐀(ρ)/Fcyc):=p(𝐀(ρ)/Fcyc)Y(\mathbf{A}(\rho)/F_{\operatorname{cyc}})\mathrel{\mathop{\mathchar 58\relax}}=\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})^{\vee}.

Theorem 3.3.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation such that the weak Leopoldt conjecture holds for ρ\rho^{*}. Then, H1(FS/Fcyc,𝐀(ρ))H^{1}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho^{*}))^{\vee} contains no non-zero finite Λ\Lambda-submodule (i.e., Λ\Lambda-submodule having finite cardinality).

Proof.

The result is well known for Galois representations which are ordinary at pp (see [Gre89, Proposition 5]) and the general case follows from [OV02, Theorem 4.7] specialized to the cyclotomic p\mathbb{Z}_{p}-extension FcycF_{\operatorname{cyc}}. ∎

Corollary 3.4.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation and assume that the weak Leopoldt conjecture holds for ρ\rho^{*}. Then, the following are equivalent:

  1. (1)

    H2(FS/Fcyc,𝐕ρ¯)H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}}) is finite,

  2. (2)

    H2(FS/Fcyc,𝐕ρ¯)=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}})=0.

Proof.

For ease of notation, denote 𝐀(ρ)\mathbf{A}(\rho^{*}) by 𝐀\mathbf{A} throughout the proof. From the long exact sequence in cohomology associated with

0𝐀[ϖ]𝐀ϖ𝐀0,0\rightarrow\mathbf{A}[\varpi]\rightarrow\mathbf{A}\xrightarrow{\varpi}\mathbf{A}\rightarrow 0,

we have

H2(FS/Fcyc,𝐀)H2(FS/Fcyc,𝐀[ϖ])H1(FS/Fcyc,𝐀).\dots\rightarrow H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A})^{\vee}\rightarrow H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi])^{\vee}\rightarrow H^{1}(F_{S}/F_{\operatorname{cyc}},\mathbf{A})^{\vee}.

From our hypothesis on Weak Leopoldt conjecture for ρ\rho^{*}, we obtain that H2(FS/Fcyc,𝐀[ϖ])H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi])^{\vee} is a Λ\Lambda-submodule of H1(FS/Fcyc,𝐀)H^{1}(F_{S}/F_{\operatorname{cyc}},\mathbf{A})^{\vee}. By Theorem 3.3, H1(FS/Fcyc,𝐀)H^{1}(F_{S}/F_{\operatorname{cyc}},\mathbf{A})^{\vee} does not contain any non-zero finite Λ\Lambda-submodule which proves the corollary. ∎

From the Poitou-Tate sequence associated to 𝐀(ρ)\mathbf{A}(\rho), we obtain the following exact sequences (see [PR00, Section 1.3.1] for more details):

(3.1) 0H0(Fcyc,𝐀(ρ))vSKv0(𝐀(ρ)/Fcyc)HIw2(FS/Fcyc,𝐓ρ)Rp(𝐀(ρ)/Fcyc)0,0Rp(𝐀(ρ)/Fcyc)H1(FS/Fcyc,𝐀(ρ))vSKv1(𝐀(ρ)/Fcyc)HIw1(FS/Fcyc,𝐓ρ)H2(FS/Fcyc,𝐀(ρ)).\begin{split}0&\rightarrow H^{0}\left(F_{\operatorname{cyc}},\mathbf{A}(\rho)\right)\rightarrow\bigoplus_{v\in S}K_{v}^{0}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})\rightarrow H^{2}_{\operatorname{Iw}}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})^{\vee}\rightarrow\\ &R_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})\rightarrow 0,\\ 0&\rightarrow R_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})\rightarrow H^{1}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho))\rightarrow\bigoplus_{v\in S}K^{1}_{v}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})\rightarrow\\ &H_{\operatorname{Iw}}^{1}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})^{\vee}\rightarrow H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho)).\end{split}
Lemma 3.5.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation. The following statements are equivalent

  1. (1)

    Y(𝐀(ρ)/Fcyc)Y(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) is Λ\Lambda-torsion with μ=0\mu=0,

  2. (2)

    HIw2(FS/Fcyc,𝐓ρ)H^{2}_{\operatorname{Iw}}\left(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}}\right) is Λ\Lambda-torsion with μ=0\mu=0.

Proof.

From the Poitou-Tate sequence, we get the following exact sequence

0Y(𝐀(ρ)/Fcyc)HIw2(FS/Fcyc,𝐓ρ)vSKv0(𝐀(ρ)/Fcyc).0\rightarrow Y(\mathbf{A}(\rho)/F_{\operatorname{cyc}})\rightarrow H^{2}_{\operatorname{Iw}}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})\rightarrow\bigoplus_{v\in S}K_{v}^{0}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})^{\vee}.

Every non-archimedean prime vv is finitely decomposed in FcycF_{\operatorname{cyc}}, hence, it is not hard to see that Kv0(𝐀(ρ)/Fcyc)K_{v}^{0}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})^{\vee} is finitely generated as a p\mathbb{Z}_{p}-module. As a Λ\Lambda-module, Kv0(𝐀(ρ)/Fcyc)K_{v}^{0}(\mathbf{A}(\rho)/F_{\operatorname{cyc}})^{\vee} is thus clearly torsion with μ\mu-invariant equal to 0. The result is seen to follow from this. ∎

We now give a proof of Theorem 3.2. Denote the mod-pp Iwasawa algebra by Ω:=Λ/ϖΛ\Omega\mathrel{\mathop{\mathchar 58\relax}}=\Lambda/\varpi\Lambda.

Proof of Theorem 3.2.

For ease of notation, denote 𝐀(ρ)\mathbf{A}({\rho^{*}}) by 𝐀\mathbf{A} throughout the proof. Identify 𝐀[ϖ]\mathbf{A}[\varpi] with 𝐕ρ¯\mathbf{V}_{\bar{\rho}^{*}}. Note that HIw2(FS/Fcyc,𝐓ρ)H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}}) is finitely generated as a Λ\Lambda-module. This follows easily from the Poitou-Tate sequence (3.1) together with the proof of Lemma 3.5. According to the Lemma 3.5, condition (1) is equivalent to HIw2(FS/Fcyc,𝐓ρ)H^{2}_{\operatorname{Iw}}\left(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}}\right) being a finitely generated 𝒪\mathcal{O}-module. From the short exact sequence

0𝐓ρϖ𝐓ρ𝐀[ϖ]0,{0\rightarrow\mathbf{T}_{\rho^{*}}\xrightarrow{\varpi}\mathbf{T}_{\rho^{*}}\rightarrow\mathbf{A}[\varpi]\rightarrow 0},

we obtain

HIw2(FS/Fcyc,𝐓ρ)ϖHIw2(FS/Fcyc,𝐓ρ)HIw2(FS/Fcyc,𝐀[ϖ])0.H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})\xrightarrow{\varpi}H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})\rightarrow H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi])\rightarrow 0.

The isomorphism

HIw2(FS/Fcyc,𝐀[ϖ])HIw2(FS/Fcyc,𝐓ρ)ϖHIw2(FS/Fcyc,𝐓ρ)H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi])\simeq\frac{H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})}{\varpi H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},{\mathbf{T}_{\rho^{*}}})}

shows that (1) is equivalent to the finiteness of HIw2(FS/Fcyc,𝐀[ϖ])H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi]).

We show that (2) is also equivalent to the finiteness of HIw2(FS/Fcyc,𝐀[ϖ])H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi]). Given a finitely generated Ω\Omega-module MM, we set Ej(M)E^{j}(M) to denote the Ext-group ExtΩj(M,Ω)\operatorname{Ext}^{j}_{\Omega}(M,\Omega). Setting Ei,j:=Ej(Hi(FS/Fcyc,𝐀[ϖ]))E^{i,j}\mathrel{\mathop{\mathchar 58\relax}}=E^{j}\left(H^{i}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi])^{\vee}\right), we find that the Iwasawa cohomology group HIw2(FS/Fcyc,𝐀[ϖ])H_{\operatorname{Iw}}^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi]) is related to the cohomology groups Hi(FS/Fcyc,𝐀[ϖ])H^{i}\left(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi]\right)^{\vee} via Jannsen’s spectral sequence [Jan13, Corollary 13]. The spectral sequence states that

Ei,jHIwi+j(FS/Fcyc,𝐀[ϖ]).E^{i,j}\Rightarrow H_{\operatorname{Iw}}^{i+j}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi]).

Letting WiW_{i} to be the cohomology group Hi(FS/Fcyc,𝐀[ϖ])H^{i}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi])^{\vee}, note that for j>0j>0, the adjoint Iwasawa module Ej(Wi)E^{j}(W_{i}) is Ω\Omega-torsion. Recall that condition (2) is the requirement that W2=0W_{2}=0. According to Corollary 3.4, W2=0W_{2}=0 if and only if W2W_{2} is finite. On the other hand, it is clear that W2W_{2} is finite if and only if E2,0=HomΩ(W2,Ω)E^{2,0}=\operatorname{Hom}_{\Omega}(W_{2},\Omega) is 0. Therefore, W2=0W_{2}=0 (i.e., (2) holds) if and only if E2,0=0E^{2,0}=0. From the above spectral sequence, it follows that (2) is equivalent to the finiteness of HIw2(FS/Fcyc,𝐀[ϖ])H^{2}_{\operatorname{Iw}}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}[\varpi]). ∎

Lemma 3.6.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation. With notation as above, suppose that H2(FS/Fcyc,𝐕ρ¯)=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}})=0. Then, the weak Leopoldt conjecture is true for ρ\rho^{*} and p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

Proof.

Consider the short exact sequence of Galois modules

0𝐕ρ¯𝐀(ρ)ϖ𝐀(ρ)0,0\rightarrow\mathbf{V}_{\bar{\rho}^{*}}\rightarrow\mathbf{A}(\rho^{*})\xrightarrow{\varpi}\mathbf{A}(\rho^{*})\rightarrow 0,

from which we find that H2(FS/Fcyc,𝐕ρ¯)H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}}) surjects onto H2(FS/Fcyc,𝐀(ρ))[ϖ]H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho^{*}))[\varpi]. From the assumption H2(FS/Fcyc,𝐕ρ¯)=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}})=0, it follows that H2(FS/Fcyc,𝐀(ρ))[ϖ]=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho^{*}))[\varpi]=0. The vanishing of H2(FS/Fcyc,𝐀(ρ))[ϖ]H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho^{*}))[\varpi] implies that H2(FS/Fcyc,𝐀(ρ))=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{A}(\rho^{*}))=0. Thus, it follows that the weak Leopoldt conjecture holds for ρ\rho^{*}. It follows from Theorem 3.2 that p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0. ∎

The following criterion will be used in establishing new cases of the μ=0\mu=0 conjecture for the fine Selmer group.

Theorem 3.7.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{F,S}\to\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation as above. Suppose that H2(FS/F,𝐕ρ¯)=0H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}})=0. Then, p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0. Furthermore, the weak Leopoldt conjecture holds for ρ\rho^{*}.

Proof.

Since the pp-cohomological dimension of Γ=Gal(Fcyc/F)\Gamma=\operatorname{Gal}(F_{\operatorname{\mathrm{cyc}}}/F) is 11, by the Hochschild-Serre spectral sequence, the restriction map

H2(FS/F,𝐕ρ¯)H2(FS/Fcyc,𝐕ρ¯)ΓH^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}})\rightarrow H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}})^{\Gamma}

is surjective. Hence, if H2(FS/F,𝐕ρ¯)=0H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}})=0, then,

H2(FS/Fcyc,𝐕ρ¯)Γ=0.H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}})^{\Gamma}=0.

Since Γ\Gamma is pro-pp, this implies that H2(FS/Fcyc,𝐕ρ¯)=0H^{2}(F_{S}/F_{\operatorname{cyc}},\mathbf{V}_{\bar{\rho}^{*}})=0, and the result follows from Lemma 3.6. ∎

In the remainder of this section, we outline sufficient conditions for the vanishing of H2(FS/F,𝐕ρ¯)H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}}). We shall illustrate these conditions for Galois representations that arise from elliptic curves. In section 5, it is shown that these conditions are satisfied for a number of examples of dihedral Galois representations of interest. Given i0i\geq 0, we let (𝐕ρ¯)Si{}^{i}_{S}(\mathbf{V}_{\bar{\rho}}) be defined as follows

(𝐕ρ¯)Si:=ker{Hi(FS/F,𝐕ρ¯)vSHi(Fv,𝐕ρ¯)}.{}^{i}_{S}(\mathbf{V}_{\bar{\rho}})\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{ker}\left\{H^{i}(F_{S}/F,\mathbf{V}_{\bar{\rho}})\longrightarrow\bigoplus_{v\in S}H^{i}(F_{v},\mathbf{V}_{\bar{\rho}})\right\}.

Set LL to be the Galois extension of FF cut out by ρ¯\bar{\rho}. In other words, it is the extension of FF fixed by the kernel of ρ¯\bar{\rho}. Set HLH_{L} to be the mod-pp class group HL:=Cl(L)𝔽pH_{L}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Cl}(L)\otimes\mathbb{F}_{p}. Denote by HLH_{L}^{\prime} the maximal quotient of HLH_{L} such that the primes of LL that lie above SS are split in the corresponding subextension of the Hilbert class field of LL. Note that HLH_{L}^{\prime} is a stable with respect to the natural action of G=Gal(L/F)G=\operatorname{Gal}(L/F) on HLH_{L}. Given a prime vv of FF, set Gv:=Gal(F¯v/Fv)\operatorname{G}_{v}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Gal}(\bar{F}_{v}/F_{v}), which is viewed as a subgroup of Gal(F¯/F)\operatorname{Gal}(\bar{F}/F) after fixing an embedding F¯Fv¯\bar{F}\to\bar{F_{v}}. Set ρ|v\rho_{|v} (resp. ρ¯|v\bar{\rho}_{|v}) to denote the restriction of ρ\rho (resp. ρ¯\bar{\rho}) to Gv\operatorname{G}_{v}.

Theorem 3.8.

Let ρ:GF,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a Galois representation and ρ¯:GF,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{n}(\mathbb{F}) be the associated residual representation. With respect to notation above, assume that the following conditions are satisfied:

  1. (1)

    H1(G,𝐕ρ¯)=0H^{1}(G,\mathbf{V}_{\bar{\rho}})=0,

  2. (2)

    HomG(HL,𝐕ρ¯)=0\operatorname{Hom}_{G}(H_{L}^{\prime},\mathbf{V}_{\bar{\rho}})=0.

  3. (3)

    At each prime vSv\in S, the local representation ρ¯|v\bar{\rho}_{|v} does not have the trivial 11-dimensional representation as a subrepresentation.

Then, p(𝐀(ρ)/Fcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0. Furthermore, the weak Leopoldt conjecture is true for ρ\rho^{*}.

Proof.

According to Theorem 3.7, if H2(FS/F,𝐕ρ¯)=0H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}})=0, then the assertions follow. We show that H2(FS/F,𝐕ρ¯)=0H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}})=0, thus proving the result. Note that H2(FS/F,𝐕ρ¯)H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}}) fits into an exact sequence

0(𝐕ρ¯)S2H2(FS/F,𝐕ρ¯)vSH2(Fv,𝐕ρ¯).0\rightarrow{}^{2}_{S}(\mathbf{V}_{\bar{\rho}^{*}})\rightarrow H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}})\rightarrow\bigoplus_{v\in S}H^{2}(F_{v},\mathbf{V}_{\bar{\rho}^{*}}).

By Global duality of -groups [NSW08, Theorem 8.6.7],

(𝐕ρ¯)S2(𝐕ρ¯)S1.{}^{2}_{S}(\mathbf{V}_{\bar{\rho}^{*}})\simeq{}^{1}_{S}\left(\mathbf{V}_{\bar{\rho}}\right)^{\vee}.

By local Tate duality [Ser13, pp. 91-92], H2(Fv,𝐕ρ¯)H0(Fv,𝐕ρ¯)H^{2}(F_{v},\mathbf{V}_{\bar{\rho}^{*}})\simeq H^{0}(F_{v},\mathbf{V}_{\bar{\rho}})^{\vee}, and hence, is trivial since we are assuming that the trivial representation does not occur as a subrepresentation of ρ¯|v\bar{\rho}_{|v}.

The hypothesis H1(G,𝐕ρ¯)=0H^{1}(G,\mathbf{V}_{\bar{\rho}})=0, together with the inflation-restriction sequence applied to the extensions FLFSF\subset L\subset F_{S}, implies that (𝐕ρ¯)S1{}^{1}_{S}\left(\mathbf{V}_{\bar{\rho}}\right) injects into HomG(HL,𝐕ρ¯)\operatorname{Hom}_{G}(H_{L}^{\prime},\mathbf{V}_{\bar{\rho}}). Since it is assumed that HomG(HL,𝐕ρ¯)=0\operatorname{Hom}_{G}(H_{L}^{\prime},\mathbf{V}_{\bar{\rho}})=0, it follows that (𝐕ρ¯)S1=0{}^{1}_{S}\left(\mathbf{V}_{\bar{\rho}}\right)=0, and hence (𝐕ρ¯)S2=0{}^{2}_{S}(\mathbf{V}_{\bar{\rho}^{*}})=0 as well. Thus, putting everything together, we have shown that H2(FS/F,𝐕ρ¯)H^{2}(F_{S}/F,\mathbf{V}_{\bar{\rho}^{*}}) =0=0. ∎

Let EE be an elliptic curve defined over a number field FF and let pp be an odd prime number. Let ρ:GF,SGL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{F,S}\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p}) be the Galois representation on the pp-adic Tate module of EE and let ρ¯\bar{\rho} be its residual representation. Let GG denote the Galois group Gal(L/F)\operatorname{Gal}(L/F). For a prime vv of FF, denote by FvF_{v} the completion of FF at vv and by 𝔽v\mathbb{F}_{v} the residue field of FvF_{v}.

Theorem 3.9.

Let E/FE_{/F} be an elliptic curve. Let SS be the set of primes of FF that consists of primes SpS_{p}, all archimedean primes and the primes at which EE has bad reduction. Let ρ:GF,SGL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{F,S}\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p}) be the pp-adic Galois representation attached to EE. With respect to above notation, assume that the following conditions are satisfied

  1. (1)

    the residual representation ρ¯\bar{\rho} is irreducible,

  2. (2)

    HomG(HL,E[p])=0\operatorname{Hom}_{G}\left(H_{L}^{\prime},E[p]\right)=0,

  3. (3)

    for every prime vSv\in S, we assume that E(𝔽v)[p]=0E(\mathbb{F}_{v})[p]=0.

Then, the following assertions hold:

  1. (a)

    the weak Leopoldt conjecture is true for ρ\rho,

  2. (b)

    p(E[p]/Fcyc)\mathcal{R}_{p^{\infty}}(E[p^{\infty}]/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

Proof.

We verify that the conditions of Theorem 3.8. Clearly, the conditions (2) and (3) are satisfied. Note that 𝐕ρ¯\mathbf{V}_{\bar{\rho}} is identified with E[p]E[p]. That condition (1) is satisfied follows from [PS21, Lemma 2.2]. This requires the assumption that ρ¯\bar{\rho} is irreducible. Hence, Theorem 3.8 implies that the weak Leopoldt conjecture is true for ρ\rho^{*} and p(E[p]/Fcyc)\mathcal{R}_{p^{\infty}}(E[p^{\infty}]/F_{\operatorname{cyc}}) is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0. As ρ\rho is the pp-adic Galois representation attached to an elliptic curve, ρρ\rho\simeq\rho^{*}. This finishes the proof of the theorem. ∎

We note here that for elliptic curves E/E_{/\mathbb{Q}} with Mordell Weil rank 0 and the Tate-Shafarevich group (E/)[p]=0\Sh(E/\mathbb{Q})[p]=0, the upper bound given by [PS21, Theorem 4.2] provides sufficient conditions for the vanishing of HomG(HL,E[p])\operatorname{Hom}_{G}\left(H_{L},E[p]\right). To illustrate this, consider the following example.

Example 3.10.

Let EE be the elliptic curve over F=F=\mathbb{Q} given by 11a2 and Weierstrass equation y2+y=x3x27820x263580y^{2}+y=x^{3}-x^{2}-7820x-263580 and set p=7p=7. According to the data provided by the L-functions and Modular forms database [LMF22],

  1. (1)

    rankE()=0\operatorname{rank}E(\mathbb{Q})=0,

  2. (2)

    EE has conductor 1111 and thus, has good reduction at 77,

  3. (3)

    EE has split multiplicative reduction at 1111, thus, #E(𝔽11)=10\#E(\mathbb{F}_{11})=10,

  4. (4)

    #E(𝔽7)=8a7(E)=10\#E(\mathbb{F}_{7})=8-a_{7}(E)=10,

  5. (5)

    the mod-77 representation is irreducible,

  6. (6)

    the Tamagawa product is 11,

  7. (7)

    (E/)=0\Sh(E/\mathbb{Q})=0.

Set S={7,11,}S=\{7,11,\infty\}. The conditions (1) and (3) are satisfied. In order to see that condition (2) is satisfied, it suffices to show that the stronger vanishing condition HomG(HL,E[p])=0\operatorname{Hom}_{G}\left(H_{L},E[p]\right)=0. We refer to the notation in [PS21, Theorem 4.2]. Since the Mordell-Weil rank of EE is 0 and (E/)=0\Sh(E/\mathbb{Q})=0, it follows that the mod-pp Selmer group defined in loc.cit. is 0. Furthermore, since 7#E(𝔽11)7\nmid\#E(\mathbb{F}_{11}) the set \mathcal{I} in loc.cit. is empty. Thus, by the upper bound of [PS21, Theorem 4.2], we find that HomG(HL,E[p])=0\operatorname{Hom}_{G}\left(H_{L},E[p]\right)=0, and in particular, (2) is also satisfied for E/E_{/\mathbb{Q}} at p=7p=7.

We note here that our condition HomG(HL,E[p])=0\operatorname{Hom}_{G}\left(H_{L}^{\prime},E[p]\right)=0 is a weaker condition than HomG(HL,E[p])=0\operatorname{Hom}_{G}\left(H_{L},E[p]\right)=0. Since explicit computation with class groups of fields generated by torsion points of elliptic curves is difficult, we do not provide an example to illustrate this.

4. The Adjoint representation of a modular form and μ=0\mu=0 for the fine Selmer group

In this section, we prove results for the vanishing of the μ\mu-invariant of the fine Selmer group for the first Tate-twist of the adjoint Galois representation associated with a newform. Here, by newform, we mean a new cuspidal eigenform. We will use this convention throughout the article. When the newform has weight 22 and trivial nebentypus, the symmetric square representation is a direct summand of this Tate-twist, and hence, in this case, we effectively obtain results for the symmetric square representation as well. Moreover, given a rational elliptic curve EE of squarefree conductor, we prove results for the vanishing of the μ\mu-invariant of the fine Selmer group for Ad0(ρE,p)\operatorname{Ad}^{0}(\rho_{E,p}). Here ρE,p\rho_{E,p} is the pp-adic Galois representation attached to EE and Ad0(ρE,p)\operatorname{Ad}^{0}(\rho_{E,p}) is the subrepresentation of the adjoint representation of ρE,p\rho_{E,p} consisting of matrices with trace 0.

The study of these adjoint Galois representations is of special significance in deformation theory. We shall introduce ideas from deformation theory to study the vanishing of the μ\mu-invariant of the fine Selmer group. Our methods also extend to Artin representations and neat representations (in the sense of Mazur) and we prove results for the vanishing of the μ\mu-invariants of the fine Selmer groups of their first Tate-twists.

For a number field KK, we will be studying continuous mod pp representations of Gal(K¯/K)\operatorname{Gal}(\overline{K}/K) which are unramified outside a finite set of primes of KK and their lifts to characteristic 0. Our focus will be on μ\mu-invariants of fine Selmer groups of such representations. Of particular significance are the Galois representations arising from modular eigenforms.

4.1. Adjoints of Galois representations

The notion of unobstructedness arises naturally in the study of Galois deformation theory. It was initially introduced by Mazur in [Maz89] and is recalled below. Let 𝔽\mathbb{F} be a finite field of characteristic pp, KK be a number field and SS be a finite set of primes of KK containing all primes dividing pp and all archimedean primes. Let ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{K,S}\to\operatorname{GL}_{n}(\mathbb{F}) be a continuous representation. Recall, from § 2, that KSK_{S} is the maximal extension of KK unramified at all primes of KK lying outside SS and GK,S\operatorname{G}_{K,S} is the corresponding Galois group Gal(KS/F)\operatorname{Gal}(K_{S}/F).

Let 𝒪\mathcal{O} be the ring of integers of a finite extension of p\mathbb{Q}_{p} with residue field 𝔽\mathbb{F}. Let CNL𝒪\operatorname{CNL}_{\mathcal{O}} be the category of complete local noetherian 𝒪\mathcal{O}-algebras RR with residue field isomorphic to 𝔽\mathbb{F}. Given RCNL𝒪R\in\operatorname{CNL}_{\mathcal{O}}, let 𝔪\mathfrak{m} be the maximal ideal of RR and fix a residual isomorphism of 𝒪\mathcal{O}-algebras R/𝔪𝔽R/\mathfrak{m}\xrightarrow{\sim}\mathbb{F}. As RR is noetherian, there exist α1,,αmR\alpha_{1},\cdots,\alpha_{m}\in R such that 𝔪=(α1,,αm)\mathfrak{m}=(\alpha_{1},\cdots,\alpha_{m}). Since RR is a complete local 𝒪\mathcal{O}-algebra, there exists a morphism ϕ:𝒪X1,,XmR\phi\mathrel{\mathop{\mathchar 58\relax}}\mathcal{O}\llbracket X_{1},\cdots,X_{m}\rrbracket\to R which sends XiX_{i} to αi\alpha_{i} for all 1im1\leq i\leq m (see [Eis95, Theorem 7.16 (a)]). Note that the map 𝒪R/𝔪\mathcal{O}\to R/\mathfrak{m} induced by ϕ\phi is surjective and α1,,αm\alpha_{1},\cdots,\alpha_{m} generate 𝔪\mathfrak{m}. Therefore, it follows, from [Eis95, Theorem 7.16 (b)], that ϕ\phi is surjective. Hence, we get a presentation for RR of the form

R𝒪X1,,Xm(f1,,fn).R\simeq\frac{\mathcal{O}\llbracket X_{1},\dots,X_{m}\rrbracket}{\left(f_{1},\dots,f_{n}\right)}.

An RR-lift of ρ¯\bar{\rho} is a Galois representation ρR:GK,SGLn(R)\rho_{R}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(R) such that ρ¯=ρRmod𝔪\bar{\rho}=\rho_{R}\mod{\mathfrak{m}}. The subgroup of matrices AGLn(R)A\in\operatorname{GL}_{n}(R) that reduce to the identity modulo the maximal ideal of RR is denoted by GL^n(R)\widehat{\operatorname{GL}}_{n}(R). Two RR-lifts ρR\rho_{R} and ρR\rho_{R}^{\prime} of ρ¯\bar{\rho} are strictly equivalent if ρR=AρRA1\rho_{R}=A\rho_{R}^{\prime}A^{-1} for some matrix AGL^n(R)A\in\widehat{\operatorname{GL}}_{n}(R). An RR-deformation of ρ¯\bar{\rho} is a strict equivalence class of RR-lifts.

Given a Galois representation ρR:GK,SGLn(R)\rho_{R}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(R), set AdρR\operatorname{Ad}\rho_{R} to be the adjoint representation of ρR\rho_{R}. As an RR-module, AdρR\operatorname{Ad}\rho_{R} consists of all n×nn\times n matrices with entries in RR. The Galois action on AdρR\operatorname{Ad}\rho_{R} is defined by the adjoint action

σv=ρR(σ)vρR(σ)1,\sigma\cdot v=\rho_{R}(\sigma)v\rho_{R}(\sigma)^{-1},

where, σGK,S\sigma\in\operatorname{G}_{K,S} and vAdρRv\in\operatorname{Ad}\rho_{R}. Thus, we may view AdρR\operatorname{Ad}\rho_{R} as a Galois representation

AdρR:GK,SGLn2(R).\operatorname{Ad}\rho_{R}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n^{2}}(R).

In particular, specializing this construction to the case R=𝔽R=\mathbb{F} and ρR=ρ¯\rho_{R}=\bar{\rho}, we get the adjoint representation of ρ¯\bar{\rho}:

Adρ¯:GK,SGLn2(𝔽).\operatorname{Ad}\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n^{2}}(\mathbb{F}).
Definition 4.1.

A representation ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathbb{F}) is said to be unobstructed if H2(KS/K,Adρ¯)=0H^{2}(K_{S}/K,\operatorname{Ad}\bar{\rho})=0. Following this definition, we say that ρ:GK,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) is unobstructed if the corresponding residual representation ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathbb{F}) (obtained by composing ρ\rho with the natural surjective map 𝒪𝔽\mathcal{O}\to\mathbb{F}) is unobstructed.

Associated to ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{K,S}\to\operatorname{GL}_{n}(\mathbb{F}), consider a functor of deformations

Defρ¯:CNL𝒪Sets,\operatorname{Def}_{\bar{\rho}}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{CNL}_{\mathcal{O}}\rightarrow\operatorname{Sets},

which takes RCNL𝒪R\in\operatorname{CNL}_{\mathcal{O}} to the set Defρ¯(R)\operatorname{Def}_{\bar{\rho}}(R) consisting of RR-deformations of ρ¯\bar{\rho} considered as a representation of the group GK,S\operatorname{G}_{K,S}. When ρ¯\bar{\rho} is absolutely irreducible, there is a universal Galois representation unramified away from SS (see [Maz89, Proposition 1]), which we denote by

ρuniv:GK,SGLn(Rρ¯).\rho^{\operatorname{univ}}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}\left(R_{\bar{\rho}}\right).

Here, Rρ¯R_{\bar{\rho}} is the universal deformation ring associated to ρ¯\bar{\rho} and the group GK,S\operatorname{G}_{K,S} (the set SS being suppressed in our notation). It is universal in the sense that for any RR-deformation ρR:GK,SGLn(R)\rho_{R}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(R) of ρ¯\bar{\rho}, there exists a unique homomorphism of complete noetherian local 𝒪\mathcal{O}-algebras ψ:Rρ¯R\psi\mathrel{\mathop{\mathchar 58\relax}}R_{\bar{\rho}}\rightarrow R such that the following diagram commutes

GLn(Rρ¯){\text{GL}_{n}(R_{\bar{\rho}})}GK,S{\operatorname{G}_{K,S}}GLn(R),{\text{GL}_{n}(R),}ρuniv\scriptstyle{\rho^{\operatorname{univ}}}ρR\scriptstyle{\rho_{R}}

where the vertical map is induced by ψ\psi.

In general, finding the explicit structure of the universal deformation ring Rρ¯R_{\bar{\rho}} is a very difficult problem. However, if ρ¯\bar{\rho} is unobstructed, then we know that

Rρ¯𝒪X1,,XdR_{\bar{\rho}}\simeq\mathcal{O}\llbracket X_{1},\cdots,X_{d}\rrbracket

where d=dim𝔽(H1(GK,S,Adρ¯))d=\dim_{\mathbb{F}}(H^{1}(G_{K,S},\operatorname{Ad}\bar{\rho})) (see [Maz89, Proposition 2]). Indeed, [Maz89, Section 1.6] implies that Rρ¯R_{\bar{\rho}} is a quotient of 𝒪X1,,Xd\mathcal{O}\llbracket X_{1},\cdots,X_{d}\rrbracket. Since ρ¯\bar{\rho} is unobstructed, [Maz89, Proposition 2] implies that the Krull dimension of Rρ¯R_{\bar{\rho}} is at least d+1d+1. This allows us to conclude that Rρ¯R_{\bar{\rho}} is isomorphic to 𝒪X1,,Xd\mathcal{O}\llbracket X_{1},\cdots,X_{d}\rrbracket. In particular, specializing to the case K=K=\mathbb{Q} and n=2n=2, we get:

Proposition 4.2.

Suppose that ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(\mathbb{F}) is absolutely irreducible and unobstructed. Then:

  1. (1)

    If ρ¯\bar{\rho} is odd, then Rρ¯R_{\bar{\rho}} is isomorphic (as a complete noetherian local 𝒪\mathcal{O}-algebra) to the formal power series ring 𝒪X1,X2,X3\mathcal{O}\llbracket X_{1},X_{2},X_{3}\rrbracket.

  2. (2)

    If ρ¯\bar{\rho} is even, then Rρ¯R_{\bar{\rho}} is isomorphic (as a complete noetherian local 𝒪\mathcal{O}-algebra) to the formal power series ring 𝒪X\mathcal{O}\llbracket X\rrbracket.

Proof.

If ρ¯\bar{\rho} is odd, then the global Euler characteristic formula implies that dim𝔽(H1(G,S,Adρ¯))=3\dim_{\mathbb{F}}(H^{1}(G_{\mathbb{Q},S},\operatorname{Ad}\bar{\rho}))=3. If ρ¯\bar{\rho} is even, then the global Euler characteristic formula implies that dim𝔽(H1(G,S,Adρ¯))=1\dim_{\mathbb{F}}(H^{1}(G_{\mathbb{Q},S},\operatorname{Ad}\bar{\rho}))=1. The result now follows from [Maz89, Proposition 2] and the explanations given above. ∎

Let ρ:G,SGL2(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathcal{O}) be a representation. As we are assuming that pp is odd, we get a direct sum decomposition

Adρ=Ad0ρ𝒪\operatorname{Ad}\rho=\operatorname{Ad}^{0}\rho\oplus\mathcal{O}

as G,SG_{\mathbb{Q},S}-representations with summands Ad0ρ\operatorname{Ad}^{0}\rho and 𝒪\mathcal{O} corresponding to trace 0 matrices and scalars, respectively. We have a similar decomposition for Adρ¯\operatorname{Ad}\bar{\rho} as well. Recall that we denoted the pp-adic cyclotomic character by χp\chi_{p}. So we have

Adρ(1)=Ad0ρ(1)𝒪(χp)=Sym2(ρ)det(ρ)1χp𝒪(χp).\operatorname{Ad}\rho(1)=\operatorname{Ad}^{0}\rho(1)\oplus\mathcal{O}(\chi_{p})=\operatorname{Sym}^{2}(\rho)\otimes\det(\rho)^{-1}\chi_{p}\oplus\mathcal{O}(\chi_{p}).

Thus, if det(ρ)=χp\det(\rho)=\chi_{p}, then Sym2(ρ)\operatorname{Sym}^{2}(\rho) is a subrepresentation of Adρ(1)\operatorname{Ad}\rho(1). We obtain the following criterion for the vanishing of the μ\mu-invariant of the fine Selmer group of the first Tate-twist of the adjoint representation (without assuming dim(ρ)=2\dim(\rho)=2):

Theorem 4.3.

Let ρ:GK,SGLn(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be a continuous Galois representation. If ρ\rho is unobstructed, then the following assertions hold:

  1. (1)

    the weak Leopoldt conjecture is true for Adρ\operatorname{Ad}\rho.

  2. (2)

    the fine Selmer group p(𝐀(Adρ(1))/Kcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\operatorname{Ad}\rho(1))/K_{\operatorname{cyc}}) associated to Adρ(1)\operatorname{Ad}\rho(1) is cotorsion over Λ\Lambda with μfn(Adρ(1))=0\mu^{\operatorname{fn}}(\operatorname{Ad}\rho(1))=0.

  3. (3)

    If n=2n=2 and r=Sym2(ρ)(det(ρ))1χpr=\operatorname{Sym}^{2}(\rho)\otimes(\det(\rho))^{-1}\chi_{p}, then the fine Selmer group p(𝐀(r)/Kcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(r)/K_{\operatorname{cyc}}) associated to rr is cotorsion over Λ\Lambda with μfn(r)=0\mu^{\operatorname{fn}}(r)=0.

Proof.

Observe that (Adρ(1))=Adρ\left(\operatorname{Ad}\rho(1)\right)^{*}=\operatorname{Ad}\rho. Unobstructedness of ρ\rho implies that H2(KS/K,Adρ¯)=0H^{2}(K_{S}/K,\operatorname{Ad}\bar{\rho})=0. Hence, the assertions (1) and (2) above follow directly from Theorem 3.7. For (3), consider the case when n=2n=2 and r=Sym2(ρ)(det(ρ))1χpr=\operatorname{Sym}^{2}(\rho)\otimes(\det(\rho))^{-1}\chi_{p}. In this case, we find that r=Ad0ρr^{*}=\operatorname{Ad}^{0}\rho. Hence, H2(KS/K,𝐕r¯)H^{2}(K_{S}/K,\mathbf{V}_{\bar{r}^{*}}) is a direct summand of H2(KS/K,Adρ¯)H^{2}(K_{S}/K,\operatorname{Ad}\bar{\rho}), and therefore is equal to 0. The assertion in this case follows once again from Theorem 3.7. ∎

The unobstructedness of modular Galois representations ρ\rho is studied in greater detail in [Wes04]. Before recalling the main result of [Wes04], we briefly describe its setup. Let ff be a newform (i.e. a new cuspidal eigenform) of weight k2k\geq 2 on Γ1(N)\Gamma_{1}(N), where N1N\in\mathbb{Z}_{\geq 1}. Let SS be a finite set of primes of \mathbb{Q} containing the primes dividing NN and \infty. Denote by FF the field of Fourier coefficients of ff and let 𝔭\mathfrak{p} be a prime of 𝒪F\mathcal{O}_{F}. Set 𝒪\mathcal{O} to be the completion of 𝒪F\mathcal{O}_{F} at 𝔭\mathfrak{p}. Fix a uniformizer ϖ\varpi of 𝒪\mathcal{O} and set 𝔽\mathbb{F} to denote the residue field 𝒪/ϖ\mathcal{O}/\varpi. Let pp be the prime of \mathbb{Q} lying below 𝔭\mathfrak{p} and let S𝔭:=S{p}S_{\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}=S\cup\{p\}. Let

ρf,𝔭:G,S𝔭GL2(𝒪)\rho_{f,\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S_{\mathfrak{p}}}\rightarrow\operatorname{GL}_{2}(\mathcal{O})

be an integral Galois representation associated to ff and the prime 𝔭\mathfrak{p} by the construction of Eichler-Shimura and Deligne. Let ρ¯f,𝔭:G,S𝔭GL2(𝔽)\bar{\rho}_{f,\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S_{\mathfrak{p}}}\rightarrow\operatorname{GL}_{2}(\mathbb{F}) be the residual representation obtained by reducing ρf,𝔭\rho_{f,\mathfrak{p}} modulo ϖ\varpi. Note that the construction of Eichler-Shimura and Deligne gives a Galois representation over the fraction field of 𝒪\mathcal{O} and the Galois representation ρf,𝔭\rho_{f,\mathfrak{p}} arises from a choice of a Galois stable 𝒪\mathcal{O}-lattice 𝐓\mathbf{T} in it. When the residual representation is absolutely irreducible, there is a unique choice of 𝐓\mathbf{T}. In all the examples that we consider, the residual representation will be absolutely irreducible. We will mostly be working with this setup in the rest of the section.

Denote by Obs(f)\operatorname{Obs}(f) the set of primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F} at which ρf,𝔭\rho_{f,\mathfrak{p}} is obstructed (i.e., not unobstructed). An important notion in this context is that of a congruence prime.

Definition 4.4.

Let dd be a divisor of NN. We say that 𝔭\mathfrak{p} is a congruence prime of level dd for ff if there is a newform ff^{\prime} such that

  • ff^{\prime} has weight kk and level Γ1(d)\Gamma_{1}(d);

  • ff^{\prime} is not Galois conjugate to ff;

  • ρ¯f,𝔭¯ρ¯f,𝔭¯\bar{\rho}_{f^{\prime},\bar{\mathfrak{p}}}\simeq\bar{\rho}_{f,\bar{\mathfrak{p}}} for some prime 𝔭¯\bar{\mathfrak{p}} of ¯\overline{\mathbb{Q}} above 𝔭\mathfrak{p}.

Denote by Cong(f)\operatorname{Cong}(f) the set of all congruence primes of ff (as dd ranges through all divisors of NN).

Since there are only finitely many newforms whose level is a divisor of NN, it is easy to see, by comparing their Hecke eigenvalues which are not equal, that Cong(f)\operatorname{Cong}(f) is finite. Let φ\varphi denote Euler’s totient function.

The following explicit result is due to Weston and has been subsequently generalized in various directions.

Theorem 4.5 (Weston).

Let ff be a Hecke newform of weight k2k\geq 2 on Γ1(N)\Gamma_{1}(N) and let MM be the conductor of the nebentypus of ff. Consider two cases.

  1. (1)

    First, consider the case k>2k>2. Then, for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}, the Galois representation ρf,𝔭\rho_{f,\mathfrak{p}} is unobstructed. Furthermore, if NN is assumed to be squarefree, we have that

    Obs(f){𝔭|p;pk+1}{𝔭|p;p|(Nφ(N)|(N/M)(+1))}Cong(f).\operatorname{Obs}(f)\subseteq\left\{\mathfrak{p}|p;p\leq k+1\right\}\cup\left\{\mathfrak{p}|p;p|\left(N\varphi(N)\prod_{\ell|(N/M)}(\ell+1)\right)\right\}\cup\operatorname{Cong}(f).
  2. (2)

    Consider the case k=2k=2. Then, the set Obs(f)\operatorname{Obs}(f) of obstructed primes 𝔭\mathfrak{p} has Dirichlet density zero.

Proof.

The first part is obtained by combining [Wes04, Theorem 5.4] with [Wes05, Theorem 1]. The second part follows directly from [Wes04, Theorem 5.5] (see also [Maz97, Corollary 2] when ff is a non-CM modular newform having weight 22, trivial nebentypus and rational Hecke eigenvalues). ∎

Combining the above theorem with Theorem 4.3, one obtains an immediate application towards the vanishing of the μ\mu-invariant of the fine Selmer group.

Theorem 4.6.

Let ff be a Hecke newform of weight k2k\geq 2 on Γ1(N)\Gamma_{1}(N) and let MM be the conductor of the nebentypus ϵ\epsilon of ff. Let FF be the number field generated by its Hecke eigenvalues and let 𝒪F\mathcal{O}_{F} be the ring of integers of FF. Consider two cases.

  1. (1)

    If k>2k>2, then, for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}, the following assertions hold:

    1. (a)

      the weak Leopoldt conjecture holds for Adρf,𝔭\operatorname{Ad}\rho_{f,\mathfrak{p}}.

    2. (b)

      Let rr denote any one of the representations Sym2(ρf,𝔭)(det(ρf,𝔭))1χp\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes(\det(\rho_{f,\mathfrak{p}}))^{-1}\chi_{p} or Adρf,𝔭(1)\operatorname{Ad}\rho_{f,\mathfrak{p}}(1). Then the fine Selmer group associated to rr is cotorsion over Λ\Lambda with μfn(r)=0\mu^{\operatorname{fn}}(r)=0.

    Furthermore, if NN is squarefree, then the above assertions hold for all primes outside the finite set

    Obs(f){𝔭|p;pk+1}{𝔭|p;p|(Nφ(N)|(N/M)(+1))}Cong(f).\operatorname{Obs}(f)\subseteq\left\{\mathfrak{p}|p;p\leq k+1\right\}\cup\left\{\mathfrak{p}|p;p|\left(N\varphi(N)\prod_{\ell|(N/M)}(\ell+1)\right)\right\}\cup\operatorname{Cong}(f).
  2. (2)

    Consider the case k=2k=2. Then, for a set of primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F} of Dirichlet density one, the following assertions hold:

    1. (a)

      the weak Leopoldt conjecture is true for Adρf,𝔭\operatorname{Ad}\rho_{f,\mathfrak{p}}.

    2. (b)

      For r{Sym2(ρf,𝔭)ϵ1,Adρf,𝔭(1)}r\in\{\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes\epsilon^{-1},\operatorname{Ad}\rho_{f,\mathfrak{p}}(1)\}, the fine Selmer group associated to rr is cotorsion over Λ\Lambda with μfn(r)=0\mu^{\operatorname{fn}}(r)=0.

Proof.

The result follows directly from Theorems 4.3 and 4.5. ∎

Remark 4.7.

If ff is a newform of weight 22 with trivial nebentypus, then, under the notation established above, Theorem 4.6 implies that for a set of primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F} of Dirichlet density one, the fine Selmer group associated to the symmetric square representation Sym2(ρf,𝔭)\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}}) is co-torsion over Λ\Lambda with μfn(Sym2(ρf,𝔭))=0\mu^{\operatorname{fn}}(\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}}))=0.

Remark 4.8.

For Galois representations associated to elliptic curves of rank 0 (not the adjoint or symmetric square representation, but the representation itself), similar results can be proved for a set of primes 𝔭\mathfrak{p} of Dirichlet density 11, see [Wut07]. On the other hand, for modular forms of weight k>2k>2, Theorem 4.6 proves that the μ\mu-invariant of the first Tate twist of the adjoint representation vanishes for all but finitely many primes, which is stronger than showing that it vanishes for a set of primes of Dirichlet density 11. Moreover, the set of primes outside which the μ\mu-invariant is known to vanish, is made explicit in the squarefree level case by the above result. Also, in the weight 22 case, when ff coincides with an abelian variety of GL2\operatorname{GL}_{2}-type, there is no assumption made on the rank of this abelian variety.

Remark 4.9.

Hatley [Hat16] has generalized [Wes05, Theorem 1] to modular newforms of arbitrary level. Thus, combining [Hat16, Theorem 3.6] with Theorem 3.7, one can explicitly describe a finite set of primes of 𝒪F\mathcal{O}_{F} outside of which μfn(Sym2(ρf,𝔭)(det(ρf,𝔭))1χp)\mu^{\operatorname{fn}}({\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes(\det(\rho_{f,\mathfrak{p}}))^{-1}\chi_{p}}) and μfn(Adρf,𝔭(1))\mu^{\operatorname{fn}}(\operatorname{Ad}\rho_{f,\mathfrak{p}}(1)) vanish even when the level of ff is not squarefree. Since the description of this set is tedious, we will not give it here. We refer the reader to [Hat16, Theorem 3.6] for more details.

Specializing to the case when N=1N=1, S={}S=\{\infty\} and ff is the unique normalized cusp form of level 11 and weight k{12,16,18,20,22,26}k\in\{12,16,18,20,22,26\}, we get, for every prime pp, the pp-adic Galois representation

ρf,p:G,{p,}GL2(p)\rho_{f,p}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},\{p,\infty\}}\to\operatorname{GL}_{2}(\mathbb{Z}_{p})

attached to ff. Let

ρ¯f,p:G,{p,}GL2(𝔽p)\bar{\rho}_{f,p}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},\{p,\infty\}}\to\operatorname{GL}_{2}(\mathbb{F}_{p})

be the reduction of ρ\rho modulo pp.

Combining [Wes04, Theorem 2] with Theorem 4.3, we get:

Theorem 4.10.

Let ff be the unique normalized cusp form of level 11 and weight k{12,16,18,20,22,26}k\in\{12,16,18,20,22,26\}. If p>k+1p>k+1 is a prime such that ρ¯f,p\bar{\rho}_{f,p} is absolutely irreducible, then

  1. (1)

    the weak Leopoldt conjecture is true for Adρf,p\operatorname{Ad}\rho_{f,p}.

  2. (2)

    For r{Sym2(ρf,p)(2k),Adρf,p(1)}r\in\{\operatorname{Sym}^{2}(\rho_{f,p})(2-k),\operatorname{Ad}\rho_{f,p}(1)\}, the fine Selmer group associated to rr is cotorsion over Λ\Lambda with μfn(r)=0\mu^{\operatorname{fn}}(r)=0.

Proof.

In the setting of the theorem above, Weston obtains [Wes04, Theorem 2] by proving that ρf,p\rho_{f,p} is unobstructed i.e. H2({p,}/,Ad(ρ¯f,p))=0H^{2}(\mathbb{Q}_{\{p,\infty\}}/\mathbb{Q},\operatorname{Ad}(\bar{\rho}_{f,p}))=0 (see the proof of [Wes04, Theorem 5.6] for more details). Combining this with Theorem 4.3 proves the theorem. Note that, in the cases at hand, det(ρf,p)=χpk1\det(\rho_{f,p})=\chi_{p}^{k-1}. Hence, we have Sym2(ρf,𝔭)(det(ρf,𝔭))1χp=Sym2(ρf,p)(2k)\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes(\det(\rho_{f,\mathfrak{p}}))^{-1}\chi_{p}=\operatorname{Sym}^{2}(\rho_{f,p})(2-k). ∎

If ff is one of the newforms considered in Theorem 4.10, then the set of primes pp such that ρ¯f,p\bar{\rho}_{f,p} is not absolutely irreducible is given in [Wes04, Section 5.4]. In particular if f=Δf=\Delta, then the hypotheses of Theorem 4.10 are satisfied by primes p17p\geq 17 and p691p\neq 691.

Weston’s work on unobstructedness of modular Galois representations has been generalized in several directions by various authors. In the setting of Hilbert modular forms, such results have been obtained by Gamzon [Gam16]. Using his main result, we get:

Theorem 4.11.

Let KK be a totally real field, ff be a Hilbert modular newform over KK and SS be a finite set of primes of KK containing all prime divisors of the level of ff and all archimedean primes of KK. Let FF be the number field generated by its Hecke eigenvalues and let 𝒪F\mathcal{O}_{F} be the ring of integers of FF. For a prime 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}, let pp be the rational prime lying below 𝔭\mathfrak{p}, S𝔭S_{\mathfrak{p}} be S{primes of K lying above p}S\cup\{\text{primes of }K\text{ lying above }p\} and

ρf,𝔭:GK,S𝔭GL2(𝒪F,𝔭)\rho_{f,\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}G_{K,S_{\mathfrak{p}}}\to\operatorname{GL}_{2}(\mathcal{O}_{F,\mathfrak{p}})

be the corresponding Galois representation attached to ff. Suppose the following hypotheses hold:

  1. (1)

    ff has no CM,

  2. (2)

    ff is not a twist of a base change of a Hilbert newform over a proper subfield EE of KK,

  3. (3)

    All weights of ff are greater than 22.

Then, for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}, the following assertions hold:

  1. (a)

    the weak Leopoldt conjecture is true for Ad0ρf,𝔭\operatorname{Ad}^{0}\rho_{f,\mathfrak{p}}.

  2. (b)

    The fine Selmer group associated to Sym2(ρf,𝔭)(det(ρf,𝔭))1χp\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes(\det(\rho_{f,\mathfrak{p}}))^{-1}\chi_{p} is cotorsion over Λ\Lambda with μfn(Sym2(ρf,𝔭)(det(ρf,𝔭))1χp)=0\mu^{\operatorname{fn}}(\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes(\det(\rho_{f,\mathfrak{p}}))^{-1}\chi_{p})=0.

Proof.

Note that, under the hypotheses of the theorem above, [Gam16, Theorem 1.1] implies that H2(KS𝔭/K,Ad0ρ¯f,𝔭)=0H^{2}(K_{S_{\mathfrak{p}}}/K,\operatorname{Ad}^{0}\bar{\rho}_{f,\mathfrak{p}})=0 for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}. We note here that (Sym2(ρf,𝔭)(det(ρf,𝔭))1χp)\left(\operatorname{Sym}^{2}(\rho_{f,\mathfrak{p}})\otimes(\det(\rho_{f,\mathfrak{p}}))^{-1}\chi_{p}\right)^{*} is identified with Ad0ρf,𝔭\operatorname{Ad}^{0}\rho_{f,\mathfrak{p}}. Theorem 3.7 gives the result. ∎

The analogous problem for GSp4\operatorname{GSp}_{4}-representations is studied by Broshi, Mullath, Sorensen and Weston [BMSW20]. On the other hand, Guiraud in [Gui20] has established a related generalization (in the spirit of Gamzon) in the setting of regular algebraic conjugate self-dual cuspidal (RACSDC) automorphic representations (see [Gui20, Theorem 1.2] for more details). We expect that Theorem 4.6 can be suitably generalized to such settings.

We end this subsection by proving an analogue of Theorem 4.6 for the vanishing of μ\mu-invariant of Ad0ρ\operatorname{Ad}^{0}\rho, where ρ\rho is the Galois representation associated to an elliptic curve.

Theorem 4.12.

Let EE be an elliptic curve over \mathbb{Q} with squarefree conductor NN and let SS be the set of primes dividing NN and \infty. For a prime pp, let

ρE,p:G,SpGL2(p)\rho_{E,p}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S_{p}}\to\operatorname{GL}_{2}(\mathbb{Z}_{p})

be the pp-adic Galois representation attached to EE. Then, for infinitely many primes pp, the following assertions hold:

  1. (1)

    the weak Leopoldt conjecture holds for Ad0ρE,p(1)=Sym2(ρE,p)\operatorname{Ad}^{0}\rho_{E,p}(1)=\operatorname{Sym}^{2}(\rho_{E,p}).

  2. (2)

    The fine Selmer group associated to Ad0ρE,p\operatorname{Ad}^{0}\rho_{E,p} is cotorsion over Λ\Lambda with μfn(Ad0ρE,p)=0\mu^{\operatorname{fn}}(\operatorname{Ad}^{0}\rho_{E,p})=0.

Proof.

Recall that (Ad0ρE,p)=Ad0ρE,p(1)(\operatorname{Ad}^{0}\rho_{E,p})^{*}=\operatorname{Ad}^{0}\rho_{E,p}(1). So, by Theorem 3.7, it suffices to prove that for infinitely many primes p, we have H2(Sp/,Ad0ρ¯E,p(1))=0H^{2}(\mathbb{Q}_{S_{p}}/\mathbb{Q},\operatorname{Ad}^{0}\bar{\rho}_{E,p}(1))=0. Using Poitou-Tate duality, we get the following exact sequence (see [Was97, Proposition 10]):

(4.1) 0H0(Sp/,Ad0ρ¯E,p)SH0(G,Ad0ρ¯E,p)H0(Gp,Ad0ρ¯E,p)H2(Sp/,Ad0ρ¯E,p(1))(Ad0ρ¯E,p)Sp10.0\to H^{0}(\mathbb{Q}_{S_{p}}/\mathbb{Q},\operatorname{Ad}^{0}\bar{\rho}_{E,p})\to\oplus_{\ell\in S}H^{0}(G_{\ell},\operatorname{Ad}^{0}\bar{\rho}_{E,p})\oplus H^{0}(G_{p},\operatorname{Ad}^{0}\bar{\rho}_{E,p})\to\\ H^{2}(\mathbb{Q}_{S_{p}}/\mathbb{Q},\operatorname{Ad}^{0}\bar{\rho}_{E,p}(1))^{*}\to{}^{1}_{S_{p}}(\operatorname{Ad}^{0}\bar{\rho}_{E,p})\to 0.

Hence, we conclude that H2(Sp/,Ad0ρ¯E,p(1))=0H^{2}(\mathbb{Q}_{S_{p}}/\mathbb{Q},\operatorname{Ad}^{0}\bar{\rho}_{E,p}(1))=0 if the following conditions hold:

  1. (1)

    H0(G,Ad0ρ¯E,p)=0H^{0}(G_{\ell},{\operatorname{Ad}^{0}\bar{\rho}_{E,p}})=0 for all S\ell\in S,

  2. (2)

    H0(Gp,Ad0ρ¯E,p)=0H^{0}(G_{p},{\operatorname{Ad}^{0}\bar{\rho}_{E,p}})=0,

  3. (3)

    (Ad0ρ¯E,p)Sp1=0{}^{1}_{S_{p}}(\operatorname{Ad}^{0}\bar{\rho}_{E,p})=0.

Suppose ρ¯E,p\bar{\rho}_{E,p} is absolutely irreducible. Hence, it follows that H0(Sp/,Ad0ρ¯E,p)=0H^{0}(\mathbb{Q}_{S_{p}}/\mathbb{Q},\operatorname{Ad}^{0}\bar{\rho}_{E,p})=0. Note that, in [Wes05], Weston defines a suitable Selmer group H1(G,𝐀(Ad0ρE,p))H^{1}_{\emptyset}(G_{\mathbb{Q}},\mathbf{A}(\operatorname{Ad}^{0}\rho_{E,p})) of 𝐀(Ad0ρE,p)\mathbf{A}(\operatorname{Ad}^{0}\rho_{E,p}) (see [Wes05, p. 204]). This Selmer group is also extensively studied in [DFG05]. Since H0(Sp/,Ad0ρ¯E,p)=0H^{0}(\mathbb{Q}_{S_{p}}/\mathbb{Q},\operatorname{Ad}^{0}\bar{\rho}_{E,p})=0, we get, using the proof of [Wes05, Lemma 7] and [Wes05, Equation 2.3], that (Ad0ρ¯E,p)Sp1{}^{1}_{S_{p}}(\operatorname{Ad}^{0}\bar{\rho}_{E,p}) is a subgroup of H1(G,𝐀(Ad0ρE,p))H^{1}_{\emptyset}(G_{\mathbb{Q}},\mathbf{A}(\operatorname{Ad}^{0}\rho_{E,p})). Let fEf_{E} be the weight 22 modular form associated to EE. From [DFG05, Theorem 3.7], we get that the length of H1(G,𝐀(Ad0ρE,p))H^{1}_{\emptyset}(G_{\mathbb{Q}},\mathbf{A}(\operatorname{Ad}^{0}\rho_{E,p})) is the pp-valuation of the congruence ideal ηfE\eta_{f_{E}}^{\emptyset} defined in [DFG05, Section 1.7.3]. As ηfE\eta_{f_{E}}^{\emptyset} is an ideal of \mathbb{Z}, we conclude that H1(G,𝐀(Ad0ρE,p))=0H^{1}_{\emptyset}(G_{\mathbb{Q}},\mathbf{A}(\operatorname{Ad}^{0}\rho_{E,p}))=0 for all but finitely many primes pp (see the proofs of [Wes04, Theorem 5.4, Theorem 5.6] for more details). Note that ρ¯E,p\bar{\rho}_{E,p} is absolutely irreducible for all but finitely many primes pp. Hence, we conclude that (Ad0ρ¯E,p)Sp1=0{}^{1}_{S_{p}}(\operatorname{Ad}^{0}\bar{\rho}_{E,p})=0 for all but finitely many primes pp.

Now EE is a rational elliptic curve of conductor NN. So it follows, from the modularity of rational elliptic curves, that there exists a newform fEf_{E} of level Γ0(N)\Gamma_{0}(N) and weight 22 with rational Fourier coefficients such that ρE,p=ρfE,p\rho_{E,p}=\rho_{f_{E},p} for all primes pp.

Now, in addition to irreducibility of ρ¯E,p\bar{\rho}_{E,p}, suppose p>N2p>N^{2} and H0(G,Ad0ρ¯E,p)0H^{0}(G_{\ell},\operatorname{Ad}^{0}\bar{\rho}_{E,p})\neq 0 for some S\ell\in S. As NN is squarefree and fEf_{E} has trivial nebentypus, it follows that ρ¯E,p\bar{\rho}_{E,p} is reducible, semi-simple and hence, unramified at \ell (see the proof of [Wes05, Lemma 11] for more details). So, we conclude, from [Edi97, (B) of p. 221], that pp is a congruence prime of level dividing N\dfrac{N}{\ell} (see Definition 4.4 for the definition of congruence primes). Since ρ¯E,p\bar{\rho}_{E,p} is absolutely irreducible for all but finitely many primes pp and there are only finitely many congruence primes, we conclude that for all but finitely many primes pp, H0(G,Ad0ρ¯E,p)=0H^{0}(G_{\ell},\operatorname{Ad}^{0}\bar{\rho}_{E,p})=0 for all S\ell\in S.

Note that if pp is a supersingular prime of EE, then the restriction of ρ¯E,p\bar{\rho}_{E,p} to GpG_{p} is absolutely irreducible and hence, H0(Gp,Ad0ρ¯E,p)=0H^{0}(G_{p},\operatorname{Ad}^{0}\bar{\rho}_{E,p})=0. A celebrated theorem of Elkies ([Elk87, Theorem 2]) implies that EE has infinitely many supersingular primes. Combining this with all the analysis given above proves the theorem. ∎

4.2. Artin representations

Note that, in Theorem 4.5, the weight of the modular newforms is always assumed to be greater than 11. Therefore, it does not shed any light on the unobstructedness of Artin representations as they arise only in the setting of weight 11 modular forms. Moreover, an Artin representation can be considered as a pp-adic representation for every prime pp (as we will see below). So it is easy to formulate the question studied by Weston [Wes04] for Artin representations of arbitrary dimensions. This question is studied by Böckle, Guiraud, Kalyanswamy and Khare in [BGKK18] in more generality. In particular, they focus on the vanishing of H2H^{2} for arbitrary Artin representations rather than restricting to the case of adjoint of Artin representations. We will now briefly describe their setup following [BGKK18, Section 6.1].

Let ρ:Gal(¯/)GLn()\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q})\to\mathrm{GL}_{n}(\mathbb{C}) be a non-trivial irreducible Artin representation (which is not necessarily odd). So it is a continuous, irreducible representation with finite non-trivial image. Let EE be the finite extension of \mathbb{Q} fixed by ker(ρ)\ker(\rho). Let G=Gal(E/)G=\operatorname{Gal}(E/\mathbb{Q}) and hEh_{E} be the class number of EE. Let SS be the set of primes of \mathbb{Q} consisting of primes which are ramified in EE and \infty.

As Im(ρ)\operatorname{Im}(\rho) is finite, there exists a number field FF such that, under a suitable basis, Im(ρ)GLn(F)\operatorname{Im}(\rho)\subset\operatorname{GL}_{n}(F). Let 𝒪F\mathcal{O}_{F} be the ring of integers of FF and 𝔭\mathfrak{p} be a prime of 𝒪F\mathcal{O}_{F}. Let F𝔭F_{\mathfrak{p}} be the completion of FF at 𝔭\mathfrak{p}, 𝒪\mathcal{O} be its ring of integers and 𝔽\mathbb{F} be its residue field. Let pp be the rational prime lying below 𝔭\mathfrak{p} and let S𝔭:=S{p}S_{\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}=S\cup\{p\}. Then, under a suitable basis, the representation ρ𝔭:G,S𝔭GLn(F𝔭)\rho_{\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S_{\mathfrak{p}}}\to\operatorname{GL}_{n}(F_{\mathfrak{p}}), obtained by composing ρ:G,S𝔭GLn(F)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S_{\mathfrak{p}}}\to\operatorname{GL}_{n}(F) with the map FF𝔭F\to F_{\mathfrak{p}} induced by completion, takes values in GLn(𝒪)\operatorname{GL}_{n}(\mathcal{O}).

Recall that we have fixed pp to be an odd prime. Moreover, assume that p|G|.hEp\nmid|G|.h_{E}, pp is unramified in EE and pmax{|G|S}p\geq\text{max}\{\ell^{|G|}\mid\ell\in S\}. Then, the residual representation ρ¯𝔭:G,S𝔭GLn(𝔽)\bar{\rho}_{\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S_{\mathfrak{p}}}\to\mathrm{GL}_{n}(\mathbb{F}) obtained by reducing ρ𝔭:G,S𝔭GLn(𝒪)\rho_{\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S_{\mathfrak{p}}}\to\operatorname{GL}_{n}(\mathcal{O}) modulo the maximal ideal of 𝒪\mathcal{O} is absolutely irreducible (see [BGKK18, Section 6.1] for more details). Thus it is natural to ask whether H2(S𝔭/,ρ¯𝔭)=0H^{2}(\mathbb{Q}_{S_{\mathfrak{p}}}/\mathbb{Q},\bar{\rho}_{\mathfrak{p}})=0 for all but finitely many such primes 𝔭\mathfrak{p}. This question is studied in [BGKK18].

To be precise, in the setup described above, they prove:

Theorem 4.13 (Böckle-Guiraud-Kalyanswamy-Khare).

Let ρ:Gal(¯/)GLn()\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q})\to\operatorname{GL}_{n}(\mathbb{C}) be a non-trivial irreducible Artin representation and let FF be a number field over which ρ\rho is defined as above. Let 𝒪F\mathcal{O}_{F} be the ring of integers of FF. Let SS be the set of primes of \mathbb{Q} consisting of primes at which ρ\rho is ramified and \infty. If H0(Gal(/),ρ)=0H^{0}(\operatorname{Gal}(\mathbb{C}/\mathbb{R}),\rho)=0, then H2(S𝔭/,ρ¯𝔭)=0H^{2}(\mathbb{Q}_{S_{\mathfrak{p}}}/\mathbb{Q},\bar{\rho}_{\mathfrak{p}})=0 for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}.

Proof.

This is a part of Proposition 6.66.6 of [BGKK18]. ∎

Combining this theorem with Theorem 3.7, we get:

Theorem 4.14.

Let ρ:Gal(¯/)GLn()\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q})\to\operatorname{GL}_{n}(\mathbb{C}) be a non-trivial irreducible Artin representation and let FF be a number field over which ρ\rho is defined as above. Let 𝒪F\mathcal{O}_{F} be the ring of integers of FF. Suppose H0(Gal(/),ρ)=0H^{0}(\operatorname{Gal}(\mathbb{C}/\mathbb{R}),\rho)=0. Then, for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}, the following assertions hold:

  1. (a)

    the weak Leopoldt conjecture is true for ρ𝔭\rho_{\mathfrak{p}}.

  2. (b)

    The fine Selmer group p(𝐀(ρ𝔭)/cyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho_{\mathfrak{p}}^{*})/\mathbb{Q}_{\operatorname{cyc}}) associated to ρ𝔭\rho_{\mathfrak{p}}^{*} is cotorsion over Λ\Lambda with μfn(ρ𝔭)=0\mu^{\operatorname{fn}}(\rho_{\mathfrak{p}}^{*})=0.

Remark 4.15.

Note that, in [BGKK18, Proposition 6.6], it is also proved that if

0<dim(H0(Gal(/),ρ))<dim(ρ)0<\dim(H^{0}(\operatorname{Gal}(\mathbb{C}/\mathbb{R}),\rho))<\dim(\rho)

and Heuristic 6.56.5 of [BGKK18] holds, then H2(S𝔭/,ρ¯𝔭)=0H^{2}(\mathbb{Q}_{S_{\mathfrak{p}}}/\mathbb{Q},\bar{\rho}_{\mathfrak{p}})=0 for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F}. Combining this result with Theorem 3.7, we conclude that μfn(ρ𝔭)=0\mu^{\operatorname{fn}}(\rho_{\mathfrak{p}}^{*})=0 for all but finitely many primes 𝔭\mathfrak{p} of 𝒪F\mathcal{O}_{F} in these cases as well.

Note that the same question for Artin representations of arbitrary number fields is analyzed in Section 6.26.2 of [BGKK18].

4.3. Neat Representations

Let KK be a number field, 𝔽\mathbb{F} be a finite field of characteristic pp and SS be a finite set of primes of KK containing all primes of KK dividing pp and all archimedean primes. Let ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{K,S}\to\operatorname{GL}_{n}(\mathbb{F}) be a continuous, absolutely irreducible representation. Let 𝒪\mathcal{O} be the ring of integers of a finite extension of p\mathbb{Q}_{p} with residue field 𝔽\mathbb{F}. Fix a uniformizer ϖ\varpi of 𝒪\mathcal{O}.

Suppose ρ¯\bar{\rho} is unobstructed and let d=dim𝔽(H1(KS/K,Adρ¯))d=\dim_{\mathbb{F}}(H^{1}(K_{S}/K,\operatorname{Ad}\bar{\rho})). Then, from [Maz89, Proposition 2], it follows that the residual representation ρ¯\bar{\rho} lifts to the universal representation unramified away from SS

ρuniv:GK,SGLn(𝒪X1,,Xd),\rho^{\operatorname{univ}}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}\left(\mathcal{O}\llbracket X_{1},\cdots,X_{d}\rrbracket\right),

which we view as a family of Galois representations. Indeed, given any dd-tuple a=(a1,,ad)𝒪da=(a_{1},\cdots,a_{d})\in\mathcal{O}^{d}, the 𝒪\mathcal{O}-valued homomorphism φa:𝒪X1,,Xd𝒪\varphi_{a}\mathrel{\mathop{\mathchar 58\relax}}\mathcal{O}\llbracket X_{1},\cdots,X_{d}\rrbracket\rightarrow\mathcal{O} sending XiX_{i} to ϖai\varpi a_{i} gives rise to a Galois representation ρa\rho_{a} given by the composite

(4.2) ρa:GK,SρunivGLn(𝒪X1,,Xd)φaGLn(𝒪),\rho_{a}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\xrightarrow{\rho^{\operatorname{univ}}}\operatorname{GL}_{n}\left(\mathcal{O}\llbracket X_{1},\cdots,X_{d}\rrbracket\right)\xrightarrow{\varphi_{a}^{*}}\operatorname{GL}_{n}(\mathcal{O}),

where the second homomorphism is induced by φa\varphi_{a}. Since ρ¯\bar{\rho} is unobstructed, it follows from Theorem 4.3 that the weak Leopoldt conjecture holds for Adρa\operatorname{Ad}\rho_{a}, and moreover, μfn(Adρa(1))=0\mu^{\operatorname{fn}}(\operatorname{Ad}\rho_{a}(1))=0. Thus, we get a family of Galois representations parametrized by 𝒪d\mathcal{O}^{d} for which the μ\mu-invariant of the fine Selmer group of the adjoint representation vanishes.

Recall that if K=K=\mathbb{Q} and n=2n=2, then we get that d=3d=3 if ρ¯\bar{\rho} is odd and d=1d=1 if ρ¯\bar{\rho} is even (see Proposition 4.2).

In [Maz89], Mazur introduced the notion of neatness and constructed odd Galois representations of Gal(¯/)\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q}) that are unobstructed and unramified away from a single prime pp. Böckle [B9̈9] gave a similar construction for even, unobstructed Galois representations of Gal(¯/)\operatorname{Gal}(\bar{\mathbb{Q}}/\mathbb{Q}). We relate the notion of neatness to vanishing of the μ\mu-invariant of fine Selmer groups.

Fix an absolutely irreducible residual representation ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathbb{F}) as above, let L/KL/K be the splitting field cut out by ρ¯\bar{\rho}. Denote by SS^{\prime} the set of primes of LL that lie above the non-archimedean primes of SS and set GG to be the Galois group G=Gal(L/K)G=\operatorname{Gal}(L/K). For a prime wSw\in S^{\prime}, denote by LwL_{w} the completion of LL at ww. Denote the subgroups of L×L^{\times} and Lw×L_{w}^{\times} consisting of of pp-th roots of unity by μp(L)\mu_{p}(L) and μp(Lw)\mu_{p}(L_{w}), respectively. Denote the ring of integers of LL and LwL_{w} by 𝒪L\mathcal{O}_{L} and 𝒪Lw\mathcal{O}_{L_{w}}, respectively.

An 𝔽p[G]\mathbb{F}_{p}[G]-module MM is said to be coprime to Adρ¯\operatorname{Ad}\bar{\rho} if the tensor product M𝔽p(Adρ¯)M\otimes_{\mathbb{F}_{p}}(\operatorname{Ad}\bar{\rho})^{\vee} does not contain the identity representation.

Definition 4.16.

We define three Galois modules M1M_{1}, M2M_{2} and M3M_{3} associated to LL as follows:

  1. (1)

    M1:=coker{μp(L)wSμp(Lw)}M_{1}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{coker}\left\{\mu_{p}(L)\rightarrow\bigoplus_{w\in S^{\prime}}\mu_{p}(L_{w})\right\};

  2. (2)

    M2:=ker{𝒪L/(𝒪L)pwS𝒪Lw/(𝒪Lw)p}M_{2}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{ker}\left\{\mathcal{O}_{L}^{*}/(\mathcal{O}_{L}^{*})^{p}\rightarrow\bigoplus_{w\in S^{\prime}}\mathcal{O}_{L_{w}}^{*}/(\mathcal{O}_{L_{w}}^{*})^{p}\right\};

  3. (3)

    M3:=Cl(L)𝔽pM_{3}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Cl}(L)\otimes\mathbb{F}_{p}, where Cl(L)\operatorname{Cl}(L) is the class group of LL.

Definition 4.17 (Mazur).

With notation as above, ρ¯\bar{\rho} is said to be neat if pp does not divide the cardinality of Im(ρ¯)\operatorname{Im}(\bar{\rho}) and M1M_{1}, M2M_{2} and M3M_{3} are coprime to Adρ¯\operatorname{Ad}\bar{\rho}.

See [Maz89, Section 1.12] for more details.

Theorem 4.18 (Mazur).

Let 𝔽\mathbb{F} be a finite field of characteristic pp, KK be a number field and SS be a set of primes of KK containing all primes above pp and all archimedean primes. Let ρ¯:GK,SGLn(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathbb{F}) be a neat residual representation. Then ρ¯\bar{\rho} is unobstructed.

Proof.

The reader is referred to the proof of [Maz89, Proposition 7]. ∎

Corollary 4.19.

Let ρ¯\bar{\rho} be a neat Galois representation as in Theorem 4.18, let d=dim𝔽(H1(KS/K,Adρ¯))d=\dim_{\mathbb{F}}(H^{1}(K_{S}/K,\operatorname{Ad}\bar{\rho})) and let ρa:GK,SGLn(𝒪)\rho_{a}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S}\rightarrow\operatorname{GL}_{n}(\mathcal{O}) be the 𝒪\mathcal{O}-deformation of ρ¯\bar{\rho} associated to a=(a1,,ad)𝒪da=(a_{1},\cdots,a_{d})\in\mathcal{O}^{d} as above (see (4.2) for the definition of ρa\rho_{a}). Then, the following assertions hold:

  1. (1)

    the weak Leopoldt conjecture is true for Adρa\operatorname{Ad}\rho_{a},

  2. (2)

    the fine Selmer group p(𝐀(Adρa(1))/Kcyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\operatorname{Ad}\rho_{a}(1))/K_{\operatorname{cyc}}) associated to Adρa(1)\operatorname{Ad}\rho_{a}(1) is cotorsion over Λ\Lambda with μfn(Adρa(1))=0\mu^{\operatorname{fn}}(\operatorname{Ad}\rho_{a}(1))=0. If n=2n=2, then the same assertion holds for Sym2(ρa)(det(ρa))1χp\operatorname{Sym}^{2}(\rho_{a})\otimes(\det(\rho_{a}))^{-1}\chi_{p}.

Proof.

The result is a direct consequence of Theorem 4.18 and Theorem 4.3. ∎

In section 1.13 of loc.cit., Mazur constructs examples of 22-dimensional neat, odd Galois representations associated to certain S3S_{3}-extensions of L/L/\mathbb{Q} ( which means d=3d=3). The examples give a family of odd representations ρ¯\bar{\rho} for which the above Corollary applies. More specifically, let pp be a prime number which can be represented as 27+4a327+4a^{3}, where aa is an integer. Let LL be the splitting field of f(x):=x3+ax+1f(x)\mathrel{\mathop{\mathchar 58\relax}}=x^{3}+ax+1, the Galois group Gal(L/)\operatorname{Gal}(L/\mathbb{Q}) is isomorphic to S3S_{3}. The discriminant of f(x)f(x) is p-p and therefore, pp is the only prime that is ramified in LL. There is a natural inclusion of S3S_{3} into GL2(𝔽p)\operatorname{GL}_{2}(\mathbb{F}_{p}), via which we obtain a Galois representation ρ¯:G,{p,}GL2(𝔽p)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},\{p,\infty\}}\rightarrow\operatorname{GL}_{2}(\mathbb{F}_{p}). We refer to loc. cit. for further details. Such Galois representations are referred to as special S3S_{3}-representations since they are unobstructed, i.e., H2(G,{p,},Adρ¯)=0H^{2}(\operatorname{G}_{\mathbb{Q},\{p,\infty\}},\operatorname{Ad}\bar{\rho})=0. Since ρ¯\bar{\rho} is an S3S_{3}-representation, it is easy to see that ρ¯\bar{\rho} is a direct summand of Adρ¯\operatorname{Ad}\bar{\rho}. Hence, it follows that H2(G,{p,},ρ¯)=0H^{2}(\operatorname{G}_{\mathbb{Q},\{p,\infty\}},{\bar{\rho}})=0. Note that ρ¯\bar{\rho} is self-dual i.e. ρ¯=ρ¯\bar{\rho}^{\vee}=\bar{\rho}. Therefore, it follows from Theorem 3.7 that for any representation ρ:G,{p,}GL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},\{p,\infty\}}\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p}) that lifts ρ¯\bar{\rho}, the following assertions hold:

  1. (1)

    the weak Leopoldt conjecture is true for ρ\rho,

  2. (2)

    the fine Selmer group p(𝐀(ρ(1))/cyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho(1))/\mathbb{Q}_{\operatorname{cyc}}) associated to ρ(1)\rho(1) is cotorsion over Λ\Lambda with μfn(ρ(1))=0\mu^{\operatorname{fn}}(\rho(1))=0.

Similarly in [B9̈9, Section 3.1], Böckle constructs examples of 22-dimensional neat, even Galois representations associated to certain totally real S3S_{3}-extensions of L/L/\mathbb{Q} (which means d=1d=1). The examples give a family of even representations ρ¯\bar{\rho} for which the above Corollary applies. To be precise, let pp be a prime number which can be represented as 4a6274a^{6}-27, where a2a\geq 2 is an integer and LL be the splitting field of f(x):=x3a2x1f(x)\mathrel{\mathop{\mathchar 58\relax}}=x^{3}-a^{2}x-1. So the Galois group Gal(L/)\operatorname{Gal}(L/\mathbb{Q}) is isomorphic to S3S_{3}. The discriminant of f(x)f(x) is pp and therefore, pp is the only prime that is ramified in LL. Using the arguments of the previous paragraph, we obtain a Galois representation ρ¯:G,{p,}GL2(𝔽p)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},\{p,\infty\}}\rightarrow\operatorname{GL}_{2}(\mathbb{F}_{p}). In [B9̈9, Theorem 3.1], Böckle proves that ρ¯\bar{\rho} is neat and hence, unobstructed if the following hypotheses are satisfied:

  1. (1)

    pp satisfies the Ankeney-Artin-Chowla conjecture,

  2. (2)

    (u(32a2)1)(2a3+9)/6(u(332a3)1)1243(4a627)(9/4+a3)(mod(4a627)2)\left(u\left(\dfrac{3}{2a^{2}}\right)-1\right)(2a^{3}+9)/6-\left(u\left(\dfrac{3}{3-2a^{3}}\right)-1\right)\not\equiv\dfrac{1}{243}(4a^{6}-27)(9/4+a^{3})\pmod{(4a^{6}-27)^{2}}, where for xpx\in\mathbb{Z}_{p}, u(x):=x/xpu(x)\mathrel{\mathop{\mathchar 58\relax}}=x/x^{p}.

It is verified in loc.cit. that these hypotheses are satisfied for all 108108 primes pp of the form 4a6274a^{6}-27 with 2a10002\leq a\leq 1000 (see [B9̈9, Section 3.1] for more details). As noted above, ρ¯\bar{\rho} is a direct summand of Adρ¯\operatorname{Ad}\bar{\rho} in this case as well. Therefore, we conclude, using the same arguments as above, that for any representation ρ:G,{p,}GL2(p)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},\{p,\infty\}}\rightarrow\operatorname{GL}_{2}(\mathbb{Z}_{p}) that lifts ρ¯\bar{\rho}, the following assertions hold:

  1. (1)

    the weak Leopoldt conjecture is true for ρ\rho.

  2. (2)

    The fine Selmer group p(𝐀(ρ(1))/cyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho(1))/\mathbb{Q}_{\operatorname{cyc}}) associated to ρ(1)\rho(1) is cotorsion over Λ\Lambda with μfn(ρ(1))=0\mu^{\operatorname{fn}}(\rho(1))=0.

5. Residually dihedral representations

In this section, we study residually dihedral Galois representations that arise from modular forms to illustrate Theorem 3.8.

To be precise, let pp be an odd prime, 𝔽\mathbb{F} be a finite field of characteristic pp, SS be a finite set of primes of \mathbb{Q} containing pp and \infty and ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathbb{F}) be a continuous representation. We assume that ρ¯\bar{\rho} is dihedral, which is to say that there exists a quadratic extension KK of \mathbb{Q} and a character ψ:GK,S𝔽×\psi\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{K,S^{\prime}}\rightarrow\mathbb{F}^{\times}, where SS^{\prime} is the set of primes of KK that lie above the primes in SS, such that ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(\mathbb{F}) is isomorphic to the induced representation IndGK,SG,S(ψ)\operatorname{Ind}_{\operatorname{G}_{K,S^{\prime}}}^{\operatorname{G}_{\mathbb{Q},S}}(\psi). For the ease of notation, we will denote this representation by IndGKG(ψ)\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi).

Let 𝒪\mathcal{O}^{\prime} be the ring of integers of a finite extension of p\mathbb{Q}_{p} with residue field 𝔽\mathbb{F} and fix a uniformizer ϖ\varpi of 𝒪\mathcal{O}^{\prime}. We say that a representation ρ:G,SGL2(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathcal{O}^{\prime}) is residually dihedral if the residual representation ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathbb{F}) obtained by reducing ρ\rho modulo ϖ\varpi is dihedral as described above. Note that a residually dihedral representation ρ\rho is not necessarily dihedral.

Now suppose ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathbb{F}) is dihedral which means ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi) for some character ψ:GK,S𝔽×\psi\mathrel{\mathop{\mathchar 58\relax}}G_{K,S^{\prime}}\to\mathbb{F}^{\times}. Then the projective image of ρ¯\bar{\rho} (i.e. the image of ρ¯\bar{\rho} under the natural surjective map GL2(𝔽)PGL2(𝔽)\operatorname{GL}_{2}(\mathbb{F})\to\operatorname{PGL}_{2}(\mathbb{F}) obtained by going modulo scalars) is either a dihedral group or a cyclic group. Moreover, ρ¯\bar{\rho} is absolutely irreducible if and only if the projective image of ρ¯\bar{\rho} is a dihedral group.

In addition to being dihedral, now assume ρ¯\bar{\rho} is also odd and absolutely irreducible. Combining the work of Khare and Wintenberger on Serre’s conjecture ([KW09, Theorem 9.1]) and work of Kisin ([Kis09, Theorem 0.1]), we see that ρ¯\bar{\rho} lifts to a pp-adic modular Galois representation ρf,𝔭:G,SGL2(𝒪)\rho_{f,\mathfrak{p}}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(\mathcal{O}) without increasing the set SS of ramified primes for the representation. There is however, always a natural choice of lift ρ:G,SGL2(W(𝔽))\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(W(\mathbb{F})) (where W(𝔽)W(\mathbb{F}) is the ring of Witt vectors of 𝔽\mathbb{F}), letting ψ~\tilde{\psi} denote the Teichmüller lift of ψ\psi, we see that ρ=IndGKG(ψ~)\rho=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\tilde{\psi}) is a lift of ρ¯\bar{\rho}.

In the setting of dihedral representations ρ¯\bar{\rho} as above, we will provide explicit criteria for the conditions in Theorem 3.8 to be satisfied. As a result, we obtain that if ρ¯\bar{\rho} satisfies these criteria, then for all characteristic 0 lifts ρ\rho of ρ¯\bar{\rho}, and in particular for the lifts arising from modular newforms, the weak Leopoldt conjecture holds for ρ\rho^{*} and μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0 (see Theorem 5.8). Note that lifts of such representations arising from CM modular forms have been studied in [BN18].

Lemma 5.1.

Let 𝒪\mathcal{O} be the ring of integers of a finite extension of p\mathbb{Q}_{p} with residue field 𝔽\mathbb{F}. Let ρ:G,SGL2(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathcal{O}) be a Galois representation such that its residual representation ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi) is dihedral as above. Let LL be the extension of \mathbb{Q} cut out by ρ¯\bar{\rho} and let G=Gal(L/)G=\operatorname{Gal}(L/\mathbb{Q}). Then H1(G,𝐕ρ¯)=0H^{1}(G,\mathbf{V}_{\bar{\rho}})=0 (i.e. condition (1) in Theorem 3.8 is satisfied).

Proof.

Observe that the order of GG is coprime to pp. Therefore, it follows that H1(G,𝐕ρ¯)=0H^{1}(G,\mathbf{V}_{\bar{\rho}})=0. ∎

Let ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi) and let LL be the field extension of \mathbb{Q} which is fixed by the kernel of ρ¯\bar{\rho} as above. Note that KLK\subset L. Let K1K_{1} and L1L_{1} be the mod-pp Hilbert class fields of KK and LL, respectively. In other words, Gal(K1/K)\operatorname{Gal}(K_{1}/K) and Gal(L1/L)\operatorname{Gal}(L_{1}/L) are identified with the mod-pp quotients of the class groups of KK and LL, respectively. Now SS is a finite set of primes of \mathbb{Q} containing pp, \infty and the primes p\ell\neq p at which ρ¯\bar{\rho} is ramified. So SS contains the primes that ramify in LL. Set K1,SK_{1,S} (resp. L1,SL_{1,S}) to be the maximal subextension of K1K_{1} (resp. L1L_{1}) in which the primes of KK (resp. LL) above SS split completely. Thus, Gal(K1,S/K)\operatorname{Gal}(K_{1,S}/K) (resp. Gal(L1,S/L)\operatorname{Gal}(L_{1,S}/L)) is identified with the maximal quotient of Cl(K)/pCl(K)\operatorname{Cl}(K)/p\operatorname{Cl}(K) (resp. Cl(L)/pCl(L)\operatorname{Cl}(L)/p\operatorname{Cl}(L)) such that the primes of KK (resp. LL) above SS split completely in the corresponding subextension. The fields defined fit into the following diagram:

KKLLK1,SK_{1,S}L1,SL_{1,S}K1K_{1}L1L_{1}

Note that all fields in the diagram are Galois over \mathbb{Q} and that G=Gal(L/)G=\operatorname{Gal}(L/\mathbb{Q}) acts naturally on HL:=Gal(L1/L)H_{L}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Gal}(L_{1}/L) and HL,S:=Gal(L1,S/L)H_{L,S}\mathrel{\mathop{\mathchar 58\relax}}=\operatorname{Gal}(L_{1,S}/L). Here, the action is described as follows. Given vHLv\in H_{L} and σG\sigma\in G, pick a lift σ~\tilde{\sigma} of σ\sigma to Gal(L1/)\operatorname{Gal}(L_{1}/\mathbb{Q}). Since HLH_{L} is a normal subgroup of Gal(L1/)\operatorname{Gal}(L_{1}/\mathbb{Q}) it follows that σ~vσ~1\tilde{\sigma}v\tilde{\sigma}^{-1} belongs to HLH_{L}. Since HLH_{L} is abelian, σ~vσ~1\tilde{\sigma}v\tilde{\sigma}^{-1} is independent of the choice of lift σ~\tilde{\sigma} of σ\sigma. The action is defined by setting σv:=σ~vσ~1\sigma\cdot v\mathrel{\mathop{\mathchar 58\relax}}=\tilde{\sigma}v\tilde{\sigma}^{-1} for any choice of lift σ~\tilde{\sigma} of σ\sigma. The module HL,SH_{L,S} is a GG-stable quotient of HLH_{L}, i.e., a quotient by a GG-stable submodule.

Lemma 5.2.

Suppose that ρ¯=IndGKG(ψ):G,SGL2(𝔽)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi)\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathbb{F}) is a dihedral representation and let LL, K1,SK_{1,S} and L1,SL_{1,S} be the fields defined as above. Assume that K1,SL=L1,SK_{1,S}\cdot L=L_{1,S}. Let G=Gal(L/)G=\operatorname{Gal}(L/\mathbb{Q}) and MM be an irreducible 𝔽p[G]\mathbb{F}_{p}[G]-module such that dimM>1\operatorname{dim}M>1. Then, with respect to notation above,

HomG(HL,S,M)=0.\operatorname{Hom}_{G}(H_{L,S},M)=0.
Proof.

Since it is assumed that K1,SL=L1,SK_{1,S}\cdot L=L_{1,S}, it follows that the GG-action on HL,SH_{L,S} factors through an action of the quotient Gal(K/)\operatorname{Gal}(K/\mathbb{Q}). Indeed, both K1,SK_{1,S} and LL are abelian extensions of KK and hence, so is L1,S=L.K1,SL_{1,S}=L.K_{1,S}. Therefore, Gal(L/K)\operatorname{Gal}(L/K) acts trivially on Gal(L1,S/L)\operatorname{Gal}(L_{1,S}/L). As Gal(K/)\operatorname{Gal}(K/\mathbb{Q}) is abelian, it follows that any irreducible representation of Gal(K/)\operatorname{Gal}(K/\mathbb{Q}) is one-dimensional. The order of GG is coprime to pp and hence, it follows that HL,SH_{L,S} decomposes into 11-dimensional representations of GG that factor through the action of Gal(K/)\operatorname{Gal}(K/\mathbb{Q}). Combining this with the assumption that MM is an irreducible GG-module with dimM>1\operatorname{dim}M>1, we get that HomG(HL,S,M)=0\operatorname{Hom}_{G}(H_{L,S},M)=0. ∎

Lemma 5.3.

Assume that ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathbb{F}) is dihedral i.e. ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi), and that its projective image is a dihedral group. Let LL, K1,SK_{1,S} and L1,SL_{1,S} be the fields defined as above and let G=Gal(L/)G=\operatorname{Gal}(L/\mathbb{Q}). Moreover, assume that the condition K1,SL=L1,SK_{1,S}\cdot L=L_{1,S} of Lemma 5.2 is satisfied. Then,

HomG(HL,S,Vρ¯)=0,\operatorname{Hom}_{G}(H_{L,S},V_{\bar{\rho}})=0,

i.e., condition (2) of Theorem 3.8 is satisfied for ρ¯\bar{\rho}.

Proof.

Since the projective image of ρ¯\bar{\rho} is dihedral, it follows that ρ¯{\bar{\rho}} is an irreducible representation of GG of dimension 22. Hence, the lemma follows directly from Lemma 5.2. ∎

We shall provide some examples in which the condition K1,SL=L1,SK_{1,S}\cdot L=L_{1,S} is satisfied. First, we obtain sufficient conditions for condition (3) in Theorem 3.8 to hold. Let ν\nu be a generator of Gal(K/)\operatorname{Gal}(K/\mathbb{Q}) and ψν\psi^{\nu} be defined as follows:

ψν(x):=ψ(ν~xν~1),\psi^{\nu}(x)\mathrel{\mathop{\mathchar 58\relax}}=\psi(\tilde{\nu}x\tilde{\nu}^{-1}),

where ν~\tilde{\nu} is a lift of ν\nu in G,SG_{\mathbb{Q},S}. Note that in the above formula, ψ(ν~xν~1)\psi(\tilde{\nu}x\tilde{\nu}^{-1}) is independent of the choice of the lift ν~\tilde{\nu} of ν\nu.

Lemma 5.4.

Assume that ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi). Then the projective image of ρ¯\bar{\rho} is a dihedral group if and only if ψψν\psi\neq\psi^{\nu}. Moreover, the image of ρ¯\bar{\rho} is itself a non-abelian dihedral group if and only if ψ1=ψν\psi^{-1}=\psi^{\nu} and ψνψ\psi^{\nu}\neq\psi.

Proof.

Recall that the projective image of ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi) is dihedral if and only if ρ¯\bar{\rho} is irreducible.

Let VV be the vector space underlying ρ¯\bar{\rho}. As ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi), we get, after choosing a suitable basis of VV, that ρ¯(g)=(ψ(g)00ψν(g))\bar{\rho}(g)=\begin{pmatrix}\psi(g)&0\\ 0&\psi^{\nu}(g)\end{pmatrix} for all gGKg\in G_{K} and ρ¯(ν~)=(01a0)\bar{\rho}(\tilde{\nu})=\begin{pmatrix}0&1\\ a&0\end{pmatrix} for some a𝔽.a\in\mathbb{F}. Now G=GKν~GKG_{\mathbb{Q}}=G_{K}\cup\tilde{\nu}G_{K}. Therefore, ρ¯\bar{\rho} is reducible if and only if there exists a non-zero vVv\in V which is a common eigenvector for ρ¯(ν~)\bar{\rho}(\tilde{\nu}) and ρ¯(g)\bar{\rho}(g) for all gGKg\in G_{K}. If ψ=ψν\psi=\psi^{\nu}, then such an eigenvector clearly exists and hence, ρ¯\bar{\rho} is reducible.

If ψψν\psi\neq\psi^{\nu}, then there exists a g0GKg_{0}\in G_{K} such that ψ(g0)ψν(g0)\psi(g_{0})\neq\psi^{\nu}(g_{0}). So if a non-zero vector vVv\in V is an eigenvector for ρ¯(g0)\bar{\rho}(g_{0}), then either ρ¯(g0)v=ψ(g0)v\bar{\rho}(g_{0})v=\psi(g_{0})v or ρ¯(g0)v=ψν(g0)v\bar{\rho}(g_{0})v=\psi^{\nu}(g_{0})v. Let w=ρ¯(ν~)vw=\bar{\rho}(\tilde{\nu})v. Then w0w\neq 0. If ρ¯(g0)v=ψ(g0)v\bar{\rho}(g_{0})v=\psi(g_{0})v, then ρ¯(g0)w=ψν(g0)w\bar{\rho}(g_{0})w=\psi^{\nu}(g_{0})w. As ψ(g0)ψν(g0)\psi(g_{0})\neq\psi^{\nu}(g_{0}) and w0w\neq 0, it follows that ww is not a scalar multiple of vv. This means that vv is not an eigenvector of ρ¯(ν~)\bar{\rho}(\tilde{\nu}). If ρ¯(g0)v=ψν(g0)v\bar{\rho}(g_{0})v=\psi^{\nu}(g_{0})v, then we similarly conclude that vv is not an eigenvector of ρ¯(ν~)\bar{\rho}(\tilde{\nu}). Therefore, if ψψν\psi\neq\psi^{\nu}, then ρ¯\bar{\rho} is irreducible. This finishes the proof of the first part of the lemma.

Note that the image of ρ¯\bar{\rho} is a non-abelian dihedral group if and only if ρ¯(GK)\bar{\rho}(G_{K}) is a cyclic group of order >2>2 and ρ¯(ν~gν~1)=ρ¯(g1)\bar{\rho}(\tilde{\nu}g\tilde{\nu}^{-1})=\bar{\rho}(g^{-1}) for all gGKg\in G_{K}. Further, ρ¯(ν~gν~1)=ρ¯(g1)\bar{\rho}(\tilde{\nu}g\tilde{\nu}^{-1})=\bar{\rho}(g^{-1}) for all gGKg\in G_{K} if and only if ψν=ψ1\psi^{\nu}=\psi^{-1}. Moreover, if this holds, then ρ¯(GK)\bar{\rho}(G_{K}) is cyclic. If ψν=ψ1\psi^{\nu}=\psi^{-1}, then ρ¯(GK)\bar{\rho}(G_{K}) is cyclic of order >2>2 if and only if ψ1ψ\psi^{-1}\neq\psi. Therefore, we conclude that the image of ρ¯\bar{\rho} is a non-abelian dihedral group if and only if ψν=ψ1\psi^{\nu}=\psi^{-1} and ψνψ\psi^{\nu}\neq\psi. ∎

Lemma 5.5.

Assume that ρ¯=IndGKG(ψ)\bar{\rho}=\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi). Let \ell be a prime and LL be the extension of \mathbb{Q} cut out by ker(ρ¯)\ker(\bar{\rho}). Suppose that the trivial representation is a subrepresentation of ρ¯|\bar{\rho}_{|\ell}. Let ww\mid\ell be a choice of prime of KK that lies above \ell. Then, at least one of the following conditions hold:

  1. (1)

    \ell splits in KK and either ψ|w=1\psi_{|w}=1 or ψ|wν=1\psi^{\nu}_{|w}=1,

  2. (2)

    \ell is either inert or ramified in KK, and the unique prime ww of KK lying above \ell splits completely in LL.

Proof.

Assume that \ell splits in KK. Then, we have that ρ¯|w=(ψ|w00ψ|wν)\bar{\rho}_{|w}=\left({\begin{array}[]{cc}\psi_{|w}&0\\ 0&\psi^{\nu}_{|w}\\ \end{array}}\right), and the result follows in this case.

Now suppose \ell is either inert or ramified in KK. Then ww is the unique prime of KK lying above \ell and we can choose ν~G\tilde{\nu}\in G_{\ell}. As the trivial representation is a subrepresentation of ρ¯|\bar{\rho}_{|\ell}, it follows that either ψ|w=1\psi_{|w}=1 or ψ|wν=1\psi^{\nu}_{|w}=1.

As ν~G\tilde{\nu}\in G_{\ell} and GwG_{w} is normal in GG_{\ell}, it follows that ν~Gwν~1=Gw\tilde{\nu}G_{w}\tilde{\nu}^{-1}=G_{w}. Therefore, it follows that if ψ|w=1\psi_{|w}=1, then ψ|wν=1\psi^{\nu}_{|w}=1 and vice versa. This finishes the proof of the Lemma.

Combining Lemmas 5.1, 5.3 and 5.5, we obtain the following result.

Theorem 5.6.

Let 𝒪\mathcal{O} be the ring of integers in a finite extension of p\mathbb{Q}_{p} with residue field 𝔽\mathbb{F} and SS be a set of primes of \mathbb{Q} containing pp and \infty. Consider a Galois representation ρ:G,SGL2(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(\mathcal{O}) and let ρ¯\bar{\rho} be its residual representation. Assume that the following assertions hold:

  1. (1)

    For some quadratic field extension K/K/\mathbb{Q} and a character ψ:GK,S𝔽×\psi\mathrel{\mathop{\mathchar 58\relax}}G_{K,S^{\prime}}\rightarrow\mathbb{F}^{\times} (where SS^{\prime} is the set of primes of KK lying above primes in SS), the residual representation ρ¯\bar{\rho} is isomorphic to the induced representation IndGKG(ψ)\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi). In addition, we assume that the projective image of ρ¯\bar{\rho} is a non-abelian dihedral group.

  2. (2)

    With respect to above notation, we assume that K1,SL=L1,SK_{1,S}\cdot L=L_{1,S} (cf. Lemma 5.3).

  3. (3)

    At each prime S\ell\in S, the local representation ρ¯|\bar{\rho}_{|\ell} does not have the trivial representation as a subrepresentation.

Then, the following assertions hold:

  1. (a)

    the weak Leopoldt conjecture is true for ρ\rho^{*},

  2. (b)

    the fine Selmer group p(𝐀(ρ)/cyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/\mathbb{Q}_{\operatorname{cyc}}) associated to ρ\rho is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

Proof.

Note that condition (3) of Theorem 3.8 is satisfied by assumption (3) above. It follows from Lemmas 5.1 and 5.3 that the remaining conditions of Theorem 3.8 are satisfied, and thus the result follows. ∎

We reiterate that Lemma 5.5 gives sufficient conditions for condition (3) of Theorem 5.6 to hold.

Note that the above theorem involves conditions on the residual representation ρ¯\bar{\rho} and the set of primes SS. Via purely Galois theoretic computations, we construct a class of examples of residual representations ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(\mathbb{F}) satisfying the above conditions.

Note that the condition K1,SL=L1,SK_{1,S}\cdot L=L_{1,S} is satisfied if K1L=L1K_{1}\cdot L=L_{1}. Examples satisfying this condition are constructed in [DW20, section 5.2] (see [DW20, Table 1]). We will adapt the approach of [DW20, Section 5.2] to our context to obtain a class of examples satisfying the conditions of Theorem 5.6. We will now briefly outline our strategy to get examples.

Let KK be an imaginary quadratic extension of \mathbb{Q} and pp be an odd prime. Assume that there is an odd prime qpq\neq p such that qq divides the class number of KK. Let NN be the maximal extension of KK contained in the Hilbert class field of KK with odd degree of extension [N:K][N\mathrel{\mathop{\mathchar 58\relax}}K]. By class field theory, NN is a Galois extension of KK and Gal(N/K)\operatorname{Gal}(N/K) is identified with the odd part of Cl(K)\operatorname{Cl}(K) i.e. with Cl(K)/Cl2(K)\operatorname{Cl}(K)/\operatorname{Cl}_{2}(K), where Cl2(K)\operatorname{Cl}_{2}(K) is the 22-Sylow subgroup of Cl(K)\operatorname{Cl}(K). Note that Cl2(K)\operatorname{Cl}_{2}(K) is stable under the action of Gal(K/)\operatorname{Gal}(K/\mathbb{Q}) on Cl(K)\operatorname{Cl}(K). Therefore, using class field theory, we conclude that NN is Galois over \mathbb{Q}.

Let ν\nu be a lift of the generator of Gal(K/)\operatorname{Gal}(K/\mathbb{Q}) in G0=Gal(N/)G_{0}=\operatorname{Gal}(N/\mathbb{Q}) with order 22. Since |Gal(N/K)||\operatorname{Gal}(N/K)| is odd, such a lift does exist. Let H0=Gal(N/K)H_{0}=\operatorname{Gal}(N/K).

Lemma 5.7.

Under the notation and set-up established above, νgν1=g1\nu g\nu^{-1}=g^{-1} for all gH0g\in H_{0}.

Proof.

As H0H_{0} is normal in G0G_{0}, νH0ν1=H0\nu H_{0}\nu^{-1}=H_{0}. Let H1={hH0νhν1=h}H_{1}=\{h\in H_{0}\mid\nu h\nu^{-1}=h\} and H2={hH0νhν1=h1}H_{2}=\{h\in H_{0}\mid\nu h\nu^{-1}=h^{-1}\}. So H1H_{1} and H2H_{2} are subgroups of H0H_{0} and moreover, they are normal in G0G_{0}. As H0H_{0} has odd order, we have H1H2={1}H_{1}\cap H_{2}=\{1\}. We now claim that H0=H1H2H_{0}=H_{1}H_{2}.

Let gH0g\in H_{0} and suppose νgν1=gH0\nu g\nu^{-1}=g^{\prime}\in H_{0}. Let g1=ggg_{1}=gg^{\prime} and g2=g(g)1g_{2}=g(g^{\prime})^{-1}. So g2=g1g2g^{2}=g_{1}g_{2}. As ν2=1\nu^{2}=1, we get that νg1ν1=g1\nu g_{1}\nu^{-1}=g_{1} and νg2ν1=g21\nu g_{2}\nu^{-1}=g_{2}^{-1}. So g1H1g_{1}\in H_{1} and g2H2g_{2}\in H_{2}. Thus g2H1H2g^{2}\in H_{1}H_{2} for all gH0g\in H_{0}. As H0H_{0} is abelian and has odd order, every element of H0H_{0} is a square and hence, H0=H1H2H_{0}=H_{1}H_{2}.

Suppose |H1|=k>1|H_{1}|=k>1. Note that G:=G0/H2G^{\prime}\mathrel{\mathop{\mathchar 58\relax}}=G_{0}/H_{2} is an abelian group of order 2|H1|=2k2|H_{1}|=2k. So if NN0N^{\prime}\subset N_{0} is the fixed field of H2H_{2}, then NN^{\prime} is an abelian extension of \mathbb{Q} with Gal(N/)=G\operatorname{Gal}(N^{\prime}/\mathbb{Q})=G^{\prime}. Note that NN^{\prime} is an unramified extension of KK. Therefore, a prime \ell of \mathbb{Q} is ramified in NN^{\prime} if and only if it is ramified in KK. As KK is quadratic over \mathbb{Q}, the order of the inertia group at any rational prime \ell in Gal(N/)\operatorname{Gal}(N^{\prime}/\mathbb{Q}) is at most 22.

As kk is odd and GG^{\prime} is abelian, GG^{\prime} contains a unique subgroup HH^{\prime} of order 22. Thus the subfield of NN^{\prime} fixed by HH^{\prime} is an unramified abelian extension of \mathbb{Q} of degree k>1k>1. This gives us a contradiction. Hence, |H1|=1|H_{1}|=1 which means H0=H2H_{0}=H_{2} proving the lemma. ∎

Let LL^{\prime} be a cyclic, unramified extension of KK with [L:K]=n>1[L^{\prime}\mathrel{\mathop{\mathchar 58\relax}}K]=n>1, nn odd (i.e. LNL^{\prime}\subset N) and pnp\nmid n. Note that such an extension exists as we are assuming that the class number of KK is divisible by an odd prime qpq\neq p. From Lemma 5.7, we conclude that LL^{\prime} is Galois over \mathbb{Q}, with Gal(L/)=Dn\operatorname{Gal}(L^{\prime}/\mathbb{Q})=D_{n}, the dihedral group with 2n2n elements. As nn is odd, Gal(L/)\operatorname{Gal}(L^{\prime}/\mathbb{Q}) is non-abelian.

Theorem 5.8.

Let 𝔽\mathbb{F} be a field of characteristic pp. Let KK be an imaginary quadratic field with class number divisible by an odd prime different from pp. Let LL^{\prime} be a cyclic extension of KK contained in the Hilbert class field of KK with [L:K]=n>1[L^{\prime}\mathrel{\mathop{\mathchar 58\relax}}K]=n>1, nn odd and pnp\nmid n. Let ψ:Gal(L/K)𝔽×\psi\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(L^{\prime}/K)\rightarrow\mathbb{F}^{\times} and η:G𝔽×\eta\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q}}\to\mathbb{F}^{\times} be non-trivial characters. Set ρ¯\bar{\rho} to denote the induced representation IndGKG(ψη)\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi\eta) of GG_{\mathbb{Q}}. Let LL be the extension of \mathbb{Q} cut out by ρ¯\bar{\rho}. Let SS be a set of prime numbers containing pp, \infty and the primes which are ramified in LL. Let ρ:G,SGL2(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}\operatorname{G}_{\mathbb{Q},S}\rightarrow\operatorname{GL}_{2}(\mathcal{O}) be a lift of ρ¯\bar{\rho}. Assume that the following conditions are satisfied:

  1. (1)

    If S\ell\in S, \ell is split in KK and ww is a fixed place of KK lying above \ell, then |ψ(Gw)||η(G)||\psi(G_{w})|\neq|\eta(G_{\ell})|,

  2. (2)

    If \ell is either inert or ramified in KK and ww is the unique place of KK lying above \ell, then η|\eta_{|\ell} is non-trivial and is not the quadratic character of GG_{\ell} corresponding to KwK_{w},

  3. (3)

    L1,S=LK1,SL_{1,S}=L\cdot K_{1,S}.

Then, the following assertions hold:

  1. (a)

    the weak Leopoldt conjecture is true for ρ\rho^{*},

  2. (b)

    the fine Selmer group p(𝐀(ρ)/cyc)\mathcal{R}_{p^{\infty}}(\mathbf{A}(\rho)/\mathbb{Q}_{\operatorname{cyc}}) associated to ρ\rho is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

Proof.

Note that, under the hypotheses of the theorem above, conditions (1) and (2) of Theorem 5.6 hold. Therefore, by Theorem 5.6, it suffices to check that no condition of Lemma 5.5 is satisfied. Let S\ell\in S and fix a place ww of KK lying above \ell.

Suppose \ell is split in KK. Now condition (1) implies that ψη|w1\psi\eta_{|w}\neq 1. As Gal(L/)\operatorname{Gal}(L^{\prime}/\mathbb{Q}) is a non-abelian dihedral group, it follows that ψν=ψ1\psi^{\nu}=\psi^{-1}. As η\eta is a character of GG_{\mathbb{Q}}, we get that ην=η\eta^{\nu}=\eta. Therefore, condition (1) implies that (ψη)|wν=ψ1η|w1(\psi\eta)^{\nu}_{|w}=\psi^{-1}\eta_{|w}\neq 1. Thus condition (1) of Lemma 5.5 is not satisfied.

Now suppose \ell is either inert or ramified in KK. As LL^{\prime} is an unramified extension of KK and KwK_{w} is a quadratic extension of \mathbb{Q}_{\ell}, it follows that the image of GG_{\ell} in Gal(L/)\operatorname{Gal}(L^{\prime}/\mathbb{Q}) is abelian. As Gal(L/)=Dn\operatorname{Gal}(L^{\prime}/\mathbb{Q})=D_{n} with nn odd, it follows that the image of GG_{\ell} in Gal(L/)\operatorname{Gal}(L^{\prime}/\mathbb{Q}) is an abelian group of order 22. Hence, ww is totally split in L/KL^{\prime}/K. Now as η|\eta_{|\ell} is not the quadratic character corresponding to KwK_{w} and η|1\eta_{|\ell}\neq 1, it follows that η|w1\eta_{|w}\neq 1 and hence, ww is not completely split in L/KL/K. Therefore, condition (2) of Lemma 5.5 is not satisfied. This finishes the proof of the theorem. ∎

Examples: Let p=3p=3 and K=(D)K=\mathbb{Q}(\sqrt{-D}) with

D{239,971,1259,2243,2699,2843}.D\in\{239,971,1259,2243,2699,2843\}.

We verify, using LMFDB [LMF22], that the class number of KK is 1515 (see Imaginary quadratic number fields with class number 15). Let LL^{\prime} be the unramified, abelian extension of KK of degree 55. Let η\eta be the mod 33 cyclotomic character of χ¯3\bar{\chi}_{3} of GG_{\mathbb{Q}}. So L=L(μ3)L=L^{\prime}(\mu_{3}). We verify, from LMFDB [LMF22], that L1=LK1L_{1}=L\cdot K_{1} which means condition (3) of Theorem 5.8 is satisfied for all finite set SS of primes of \mathbb{Q}. Let

𝒮1={ prime 1(mod3) and  is split in K but not totally split in L}\mathcal{S}_{1}=\{\ell\text{ prime }\mid\ell\equiv 1\pmod{3}\text{ and }\ell\text{ is split in }K\text{ but not totally split in }L^{\prime}\}

and

𝒮2={ prime 1(mod3) and  is not inert in K}.\mathcal{S}_{2}=\{\ell\text{ prime }\mid\ell\equiv-1\pmod{3}\text{ and }\ell\text{ is not inert in }K\}.

Let 𝒮=𝒮1𝒮2\mathcal{S}=\mathcal{S}_{1}\cup\mathcal{S}_{2}. Note that, by Chebotarev density theorem, 𝒮\mathcal{S} is an infinite set with Dirichlet density 9/209/20. Let SS^{\prime} be a finite subset of 𝒮\mathcal{S} and let S=S{3,D,}S=S^{\prime}\cup\{3,D,\infty\}. Then SS satisfies conditions (1) and (2) of Theorem 5.8. Let 𝔽\mathbb{F} be the finite field with 8181 elements and let ψ:Gal(L/K)𝔽×\psi\mathrel{\mathop{\mathchar 58\relax}}\operatorname{Gal}(L/K)\rightarrow\mathbb{F}^{\times} be a non-trivial character. Let ρ¯:G,SGL2(𝔽)\bar{\rho}\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathbb{F}) be the induced representation IndGKG(ψχ¯3)\operatorname{Ind}_{\operatorname{G}_{K}}^{\operatorname{G}_{\mathbb{Q}}}(\psi\bar{\chi}_{3}) of G,SG_{\mathbb{Q},S}. If ρ:G,SGL2(𝒪)\rho\mathrel{\mathop{\mathchar 58\relax}}G_{\mathbb{Q},S}\to\operatorname{GL}_{2}(\mathcal{O}) is a lift of ρ¯\bar{\rho}, then by Theorem 5.8, we conclude that

  1. (a)

    the weak Leopoldt conjecture is true for ρ\rho^{*},

  2. (b)

    the fine Selmer group 3(𝐀(ρ)/cyc)\mathcal{R}_{3^{\infty}}(\mathbf{A}(\rho)/{\mathbb{Q}_{\operatorname{cyc}}}) associated to ρ\rho is cotorsion over Λ\Lambda with μfn(ρ)=0\mu^{\operatorname{fn}}(\rho)=0.

Moreover, observe that ρ¯\bar{\rho} is self-dual i.e. ρ¯=ρ¯\bar{\rho}^{\vee}=\bar{\rho}. Hence, we conclude that the weak Leopoldt conjecture is true for ρ(1)\rho(1).

References

  • [B9̈9] G. Böckle. Explicit universal deformations of even Galois representations. Math. Nachr., 206:85–110, 1999.
  • [BGKK18] G. Böckle, D.-A. Guiraud, S. Kalyanswamy, and C. Khare. Wieferich primes and a mod pp leopoldt conjecture. arXiv preprint arXiv:1805.00131, 2018.
  • [BMSW20] M. Broshi, M. Z. Mullath, C. Sorensen, and T. Weston. Unobstructed deformation problems for GSp(4)\mathrm{GSp}(4). arXiv preprint arXiv:2009.06575, 2020.
  • [BN18] N. Billerey and F. A. E. Nuccio. Représentations galoisiennes diédrales et formes á multiplication complexe. Journal de Théorie des Nombres de Bordeaux, 30(2):651–670, 2018.
  • [CS05] J. Coates and R. Sujatha. Fine selmer groups of elliptic curves over pp-adic lie extensions. Mathematische Annalen, 331(4):809–839, 2005.
  • [DFG05] F. Diamond, M. Flach, and L. Guo. The tamagawa number conjecture of adjoint motives of modular forms. Annales scientifiques de l’École Normale Supérieure, Série 4, 37(5):663–727, 2005.
  • [DW20] S. V. Deo and G. Wiese. Dihedral universal deformations. Research in Number Theory, 6(3):37 pp., 2020.
  • [Edi97] B. Edixhoven. Serre’s conjecture. In Modular forms and Fermat’s last theorem (Boston, MA, 1995), pages 209–242. Springer, New York, 1997.
  • [Eis95] David Eisenbud. Commutative algebra, volume 150 of Graduate Texts in Mathematics. Springer-Verlag, New York, 1995. With a view toward algebraic geometry.
  • [Elk87] N. D. Elkies. The existence of infinitely many supersingular primes for every elliptic curve over \mathbb{Q}. Inventiones Mathematicae, 89:561–567, 1987.
  • [FW79] B. Ferrero and L. C. Washington. The Iwasawa invariant μp\mu_{p} vanishes for abelian number fields. Annals of Mathematics, pages 377–395, 1979.
  • [Gam16] A. Gamzon. Unobstructed Hilbert modular deformation problems. Journal de Théorie des Nombres de Bordeaux, 28(1):221–236, 2016.
  • [Gre89] R. Greenberg. Iwasawa theory for pp-adic representations. Advanced Studies in Pure Math, (17):97–137, 1989.
  • [Gui20] D. A. Guiraud. Unobstructedness of Galois deformation rings associated to regular algebraic conjugate self-dual cuspidal automorphic representations. Algebra & Number Theory, 14(6):1331–1380, 2020.
  • [Hat16] J. Hatley. Obstruction criteria for modular deformation problems. International Journal of Number Theory, 12(01):273–285, 2016.
  • [Hid00] H. Hida. Adjoint Selmer groups as Iwasawa modules. Israel Journal of Mathematics, 120(2):361–427, 2000.
  • [Iwa73] K. Iwasawa. On \mathbb{Z}_{\ell}-extensions of algebraic number fields. Annals of Mathematics, pages 246–326, 1973.
  • [Jan13] U. Jannsen. A spectral sequence for Iwasawa adjoints. arXiv preprint arXiv:1307.6547, 2013.
  • [Kis09] Mark Kisin. Modularity of 2-adic Barsotti-Tate representations. Invent. Math., 178(3):587–634, 2009.
  • [KW09] Chandrashekhar Khare and Jean-Pierre Wintenberger. Serre’s modularity conjecture. I. Invent. Math., 178(3):485–504, 2009.
  • [LMF22] The LMFDB Collaboration. The L-functions and modular forms database. http://www.lmfdb.org, 2022. [Online; accessed 7 April 2022].
  • [LS18] Meng Fai Lim and Ramdorai Sujatha. Fine selmer groups of congruent galois representations. Journal of Number Theory, 187:66–91, 2018.
  • [Maz72] B. Mazur. Rational points of abelian varieties with values in towers of number fields. Inventiones mathematicae, 18(3):183–266, 1972.
  • [Maz89] B. Mazur. Deforming Galois representations. In Galois Groups over \mathbb{Q}, pages 385–437. Springer, 1989.
  • [Maz97] B. Mazur. An “infinite fern” in the universal deformation space of Galois representations. volume 48, pages 155–193. 1997. Journées Arithmétiques (Barcelona, 1995).
  • [NSW08] J. Neukirch, A. Schmidt, and K. Wingberg. Cohomology of number fields, volume 323 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer-Verlag, Berlin, second edition, 2008.
  • [OV02] Y. Ochi and O. Venjakob. On the structure of Selmer groups over pp-adic lie extensions. Journal of Algebraic Geometry, 11(3):547–580, 2002.
  • [PR00] B. Perrin-Riou. pp-adic LL-functions and pp-adic representations, volume 3 of SMF/AMS Texts and Monographs. American Mathematical Society, Providence, RI; Société Mathématique de France, Paris, 2000. Translated from the 1995 French original by Leila Schneps and revised by the author.
  • [PS21] D. Prasad and S. Shekhar. Relating the Tate–Shafarevich group of an elliptic curve with the class group. Pacific Journal of Mathematics, 312(1):203–218, 2021.
  • [Ser13] J. P. Serre. Galois cohomology. Springer Science & Business Media, 2013.
  • [Suj10] R. Sujatha. Elliptic curves and Iwasawa’s μ\mu= 0 conjecture. In Quadratic forms, linear algebraic groups, and cohomology, pages 125–135. Springer, 2010.
  • [SW18] R. Sujatha and M. Witte. Fine Selmer groups and isogeny invariance. In Geometry, algebra, number theory, and their information technology applications, volume 251 of Springer Proc. Math. Stat., pages 419–444. Springer, Cham, 2018.
  • [Was97] Lawrence C. Washington. Galois cohomology. In Modular forms and Fermat’s last theorem (Boston, MA, 1995), pages 101–120. Springer, New York, 1997.
  • [Wes04] T. Weston. Unobstructed modular deformation problems. American Journal of Mathematics, 126(6):1237–1252, 2004.
  • [Wes05] T. Weston. Explicit unobstructed primes for modular deformation problems of squarefree level. Journal of Number Theory, 110(1):199–218, 2005.
  • [Wut07] C. Wuthrich. Iwasawa theory of the fine Selmer group. Journal of Algebraic Geometry, 16(1):83–108, 2007.