betaFeedback

WorksMathematics

Galois embedding problems and number field counting

Abstract

Let pp be a prime number and 1→H→G→qT→11\to H\to G\xrightarrow{q}T\to1 be an exact sequence of finite groups, where HH is a nontrivial pp-group. Fix a continuous surjection \pi:\GQ\twoheadrightarrow T and finitely many HH-conjugation-invariant local conditions \cD. Let N_{G,\pi,\cD}(X) count the continuous surjections \rho:\GQ\twoheadrightarrow G satisfying q∘ρ=πq\circ\rho=\pi and these local conditions, with ∣ΔFρ∣≤X|\Delta_{F_\rho}|\leq X, where F_\rho=\overline{\Q}^{\ker\rho}. If at least one such lift exists, we prove that, for every ε>0\varepsilon>0, X^{p/(|G|(p-1))} \ll_{G,\pi,\cD} N_{G,\pi,\cD}(X) \ll_{G,\pi,\cD,\varepsilon} X^{p/(|G|(p-1))+\varepsilon} for all sufficiently large XX. This establishes the power of XX predicted by Alberts's conjecture for every finite pp-group kernel, without any solvability assumption on TT. If every element of order pp in HH lies in Z(H)Z(H), then every nonempty family satisfies the asymptotic predicted by Alberts: N_{G,\pi,\cD}(X) \sim C_{G,\pi,\cD} X^{p/(|G|(p-1))}(\log X)^{b_V-1}, with C_{G,\pi,\cD}>0. As an application, for odd pp and m≥2m\geq2, we obtain discriminant asymptotics for surjective lifts of a fixed residual Galois representation to \GL_d(\Z/p^m\Z), subject to finitely many local conditions, whenever such a lift exists.

Introduction

Malle’s conjectures concern the number of extensions of a number field with prescribed Galois group and bounded discriminant [12, 13]. Let kk be a number field and let 1≠G⊆Sn1 \ne G \subseteq S_n be a transitive permutation group. Denote by Nk,G(X)N_{k,G}(X) the number of kk-isomorphism classes of degree-nn extensions E/kE/k such that Nk/Q(dE/k)≤XN_{k/\mathbb{Q}}(\mathfrak{d}_{E/k}) \le X and the Galois group of the Galois closure of E/kE/k, acting on the kk-embeddings of EE, is permutation isomorphic to GG. Here dE/k\mathfrak{d}_{E/k} is the relative discriminant ideal. For g∈Gg \in G, define ind⁡(g)=n−#({1,…,n}/⟨g⟩)\operatorname{ind}(g) = n - \#(\{1,\ldots,n\}/\langle g\rangle), and put a(G)=min⁡1≠g∈Gind⁡(g)a(G) = \min_{1\ne g\in G}\operatorname{ind}(g). Thus ind⁡(g)\operatorname{ind}(g) is nn minus the number of orbits of gg.

The weak form of Malle’s conjecture [12] predicts that, for every ε>0\varepsilon> 0,

X1/a(G)≪k,GNk,G(X)≪k,G,εX1/a(G)+εX^{1/a(G)} \ll_{k,G} N_{k,G}(X) \ll_{k,G,\varepsilon} X^{1/a(G)+\varepsilon}

for all sufficiently large XX.

The strong form also specifies a logarithmic factor. Let Cmin⁡(G)C_{\min}(G) be the set of conjugacy classes of elements of index a(G)a(G). Set M=∣G∣M = |G|, write Gk=Gal⁡(k‾/k)G_k = \operatorname{Gal}(\overline{k}/k), and let χM:Gk→(Z/MZ)×\chi_M : G_k \to(\mathbb{Z}/M\mathbb{Z})^\times be the mod-MM cyclotomic character. The action σ⋅[g]=[gχM(σ)−1]\sigma\cdot[g] = [g^{\chi_M(\sigma)^{-1}}] preserves Cmin⁡(G)C_{\min}(G). Set b(k,G)=#(Gk\Cmin⁡(G))b(k,G) = \#(G_k\backslash C_{\min}(G)). The original strong form [13 Conjecture 1.1] predicts a constant ck,G>0c_{k,G} > 0 such that

Nk,G(X)∼ck,GX1/a(G)(log⁡X)b(k,G)−1N_{k,G}(X) \sim c_{k,G}X^{1/a(G)}(\log X)^{b(k,G)-1}

as X→∞X \to\infty. Both invariants depend on the chosen permutation representation. Klüners [10] showed that the predicted logarithmic exponent is incorrect in general. His example is compatible with the weak form of the conjecture.

Throughout this paper, all homomorphisms between profinite groups are continuous, and finite groups have the discrete topology. We consider finite groups in their regular permutation representation. A surjection ρ:GQ↠G\rho: G_{\mathbb{Q}} \twoheadrightarrow G cuts out the field Fρ=Qker⁡ρF_\rho= \mathbb{Q}^{\ker\rho}, which is Galois over Q\mathbb{Q} with Galois group isomorphic to GG. Conversely, each GG-Galois extension gives ∣Aut⁡(G)∣|\operatorname{Aut}(G)| such surjections. An element g∈Gg \in G of order rr has ∣G∣/r|G|/r orbits in the regular action. Consequently, ind⁡reg(g)=∣G∣(1−ord⁡(g)−1)\operatorname{ind}_{\mathrm{reg}}(g) = |G|(1-\operatorname{ord}(g)^{-1}) and areg(G)=∣G∣(1−ℓG−1)a_{\mathrm{reg}}(G) = |G|(1-\ell_G^{-1}), where ℓG\ell_G is the smallest prime divisor of ∣G∣|G|.

Let pp be a prime number and fix a surjection ρˉ:GQ↠GL⁡d(Fp)\bar{\rho} : G_{\mathbb{Q}} \twoheadrightarrow\operatorname{GL}_d(\mathbb{F}_p) and an integer m≥2m \ge2. One can ask how many surjective lifts to GL⁡d(Z/pmZ)\operatorname{GL}_d(\mathbb{Z}/p^m\mathbb{Z}) have splitting field of discriminant at most XX. Deformation theory provides methods for constructing lifts with prescribed properties. More generally, suppose that

1⟶H⟶G→qT⟶1(1)1 \longrightarrow H \longrightarrow G \xrightarrow{q} T \longrightarrow1 \tag*{(1)}

is an exact sequence of finite groups in which HH is a nontrivial pp-group. Fix a surjection π:GQ↠T\pi: G_{\mathbb{Q}} \twoheadrightarrow T, and put L=Qker⁡πL = \mathbb{Q}^{\ker\pi}. A surjection ρ:GQ↠G\rho: G_{\mathbb{Q}} \twoheadrightarrow G satisfying q∘ρ=πq \circ\rho= \pi is called a proper solution of the embedding problem determined by (1) and π\pi. For a proper solution ρ\rho, the equality ker⁡π=ρ−1(H)\ker\pi= \rho^{-1}(H) gives ker⁡ρ⊆ker⁡π\ker\rho\subseteq\ker\pi. Hence L⊆FρL \subseteq F_\rho, Gal⁡(Fρ/Q)≃G\operatorname{Gal}(F_\rho/\mathbb{Q}) \simeq G, and Gal⁡(Fρ/L)≃H\operatorname{Gal}(F_\rho/L) \simeq H. In particular, [Fρ:Q]=∣G∣[F_\rho: \mathbb{Q}] = |G| and [Fρ:L]=∣H∣[F_\rho: L] = |H|.

We order proper solutions by the absolute discriminant ∣ΔFρ∣|\Delta_{F_\rho}| of the full splitting field. We find that

∣ΔFρ∣=∣ΔL∣∣H∣NL/Q(DFρ/L),|\Delta_{F_\rho}| = |\Delta_L|^{|H|}N_{L/\mathbb{Q}}(\mathfrak{D}_{F_\rho/L}),

where DFρ/L\mathfrak{D}_{F_\rho/L} is the relative discriminant ideal. Since LL and ∣H∣|H| are fixed, this is equivalent (up to a fixed scaling) to ordering by the norm of the relative discriminant.

We first recall Alberts’s conjecture [2 Conjecture 3.10]. Assume that a proper solution ρ0:GQ↠G\rho_0 : G_{\mathbb{Q}} \twoheadrightarrow G exists. Let NG,π(X)N_{G,\pi}(X) denote the number of surjections ρ:GQ↠G\rho: G_{\mathbb{Q}} \twoheadrightarrow G satisfying q∘ρ=πq \circ\rho= \pi and ∣ΔFρ∣≤X|\Delta_{F_\rho}| \le X. For each nontrivial normal subgroup N⊴GN \trianglelefteq G contained in HH, set

a(N)=min⁡1≠n∈Nind⁡reg(n).a(N) = \min_{1 \ne n \in N}\operatorname{ind}_{\mathrm{reg}}(n).

All indices here are computed in the regular permutation representation of GG. Let Cmin⁡(N)C_{\min}(N) be the set of NN-conjugacy classes of elements attaining this minimum. A continuous homomorphism φ:GQ→G\varphi: G_{\mathbb{Q}} \to G defines an action on Cmin⁡(N)C_{\min}(N) by

σ⋆[n]N=[φ(σ)nχM(σ)−1φ(σ)−1]N.\sigma\star[n]_N = [\varphi(\sigma)n^{\chi_M(\sigma)^{-1}}\varphi(\sigma)^{-1}]_N.

Denote the number of orbits by bφ(N)b_\varphi(N).

Following [2 Definition 3.8], define

βρ0=max⁡1≠N⊴G, N⊆Ha(N)=a(H) max⁡φ:GQ→Gφ≡ρ0(mod⁡N)bφ(N).\beta_{\rho_0} = \max_{\substack{1 \ne N \trianglelefteq G,\ N \subseteq H\\a(N)=a(H)}} \ \max_{\substack{\varphi:G_{\mathbb{Q}}\to G\\\varphi\equiv\rho_0\;(\operatorname{mod}N)}} b_\varphi(N).

The homomorphisms in the inner maximum are continuous and need not be surjective. The congruence means equality after composition with G→G/NG \to G/N.

Alberts’s conjecture predicts a constant cG,π>0c_{G,\pi} > 0 such that

NG,π(X)∼cG,πX1/a(H)(log⁡X)βρ0−1as X→∞.(2)N_{G,\pi}(X) \sim c_{G,\pi}X^{1/a(H)}(\log X)^{\beta_{\rho_0}-1} \quad\text{as } X \to\infty. \tag*{(2)}

Thus the predicted power of XX is determined by the minimum index among the nonidentity elements of the kernel HH. We now specialize Alberts’s predicted exponent to our pp-group kernel. In the regular representation, ind⁡reg(h)=∣G∣(1−ord⁡(h)−1)\operatorname{ind}_{\mathrm{reg}}(h) = |G|(1-\operatorname{ord}(h)^{-1}). Since every nonidentity element of HH has pp-power order and HH contains an element of order pp, the minimum is

a(H)=∣G∣(1−p−1).a(H)=|G|(1-p^{-1}).

Thus the predicted power is XαX^\alpha, where

α=p∣G∣(p−1).\alpha=\frac{p}{|G|(p-1)}.

The logarithmic invariant also simplifies in this setting. Let Cp(H)C_p(H) be the set of HH-conjugacy classes of elements of order pp. The fixed quotient π\pi defines an action on this set by σ⋆[h]H=[gσhχˉp(σ)−1gσ−1]H\sigma\star[h]_H=[g_\sigma h^{\bar{\chi}_p(\sigma)^{-1}}g_\sigma^{-1}]_H, where q(gσ)=π(σ)q(g_\sigma)=\pi(\sigma) and χˉp:GQ→Fp×\bar{\chi}_p:G_{\mathbb{Q}}\to\mathbb{F}_p^\times is the mod-pp cyclotomic character. The action is independent of the chosen lift gσg_\sigma, since conjugation by HH acts trivially on HH-conjugacy classes. Put bπ=#(GQ\Cp(H))b_\pi=\#(G_{\mathbb{Q}}\backslash C_p(H)). Then

Bρ0=bρ0(H)=bπ.B_{\rho_0}=b_{\rho_0}(H)=b_\pi.

Alberts’s prediction specializes to

NG,π(X)∼cG,πXp/(∣G∣(p−1))(log⁡X)bπ−1.N_{G,\pi}(X)\sim c_{G,\pi}X^{p/(|G|(p-1))}(\log X)^{b_\pi-1}.

Under the additional hypothesis that every element of order pp in HH is central in HH, the group V={h∈H:hp=1}V=\{h\in H:h^p=1\} is elementary abelian, and Cp(H)C_p(H) identifies with V∖{0}V\setminus\{0\}. In this case, Bρ0=bπ=bVB_{\rho_0}=b_\pi=b_V, where

bV=#(GQ\(V(−1)∖{0}))b_V=\#(G_{\mathbb{Q}}\backslash(V(-1)\setminus\{0\}))

for the action σ⋆v=χˉp(σ)−1π(σ)(v)\sigma\star v=\bar{\chi}_p(\sigma)^{-1}\pi(\sigma)(v).

Alberts [2 Theorem 1.4 and Corollary 1.5] used the Greenberg–Wiles formula to prove the matching bounds NG,π(X)≍Xα(log⁡X)bπ−1N_{G,\pi}(X)\asymp X^\alpha(\log X)^{b_\pi-1} for abelian HH. Alberts and O’Dorney [4, 5] subsequently obtained the full asymptotic. Alberts [1] studied upper bounds obtained by successive abelian extensions. His results give the weak Malle upper bound for nilpotent groups. Koymans and Pagano [11] prove the strong form of Malle’s conjecture in the regular representation for nilpotent groups whose elements of order ℓ\ell are central, where ℓ\ell is the smallest prime divisor of the group order. Alberts, Lemke Oliver, Wang and Wood [3] give an inductive framework for summing fixed-quotient counting formulas. Their applications are restricted to abelian normal subgroups and normal subgroups isomorphic to products of S3S_3. In a related counting problem, lower bounds for unrestricted GL2(Fℓ)\mathrm{GL}_2(\mathbb{F}_\ell)-extensions were obtained from elliptic curves in [18].

Main results

Fix a prime pp and a finite group GG with a nontrivial normal pp-subgroup HH. Put T=G/HT=G/H, and let q:G↠Tq:G\twoheadrightarrow T be the quotient map. Fix a continuous surjection π:GQ↠T\pi:G_{\mathbb{Q}}\twoheadrightarrow T, and set L=Qker⁡πL=\mathbb{Q}^{\ker\pi}. Fix embeddings of decomposition groups GQv⊂GQG_{\mathbb{Q}_v}\subset G_{\mathbb{Q}}, and let SS be a finite set of places containing pp, ∞\infty, and every prime ramified in LL. For each v∈Sv\in S, choose a subset Dv\mathcal{D}_v of the continuous homomorphisms ψ:GQv→G\psi:G_{\mathbb{Q}_v}\to G satisfying q∘ψ=π∣GQvq\circ\psi=\pi|_{G_{\mathbb{Q}_v}}. Assume that Dv\mathcal{D}_v is stable under conjugation by HH, and write D=(Dv)v∈S\mathcal{D}=(\mathcal{D}_v)_{v\in S}. There are no local restrictions outside SS.

Define

NG,π,D(X)=#{ρ∈Sur⁡(GQ,G):q∘ρ=π,ρ∣GQv∈Dv for every v∈S,∣ΔFρ∣≤X}.(3)N_{G,\pi,\mathcal{D}}(X)=\#\left\{\rho\in\operatorname{Sur}(G_{\mathbb{Q}},G): \begin{aligned} &q\circ\rho=\pi,\\ &\rho|_{G_{\mathbb{Q}_v}}\in\mathcal{D}_v\text{ for every }v\in S,\\ &|\Delta_{F_\rho}|\leq X \end{aligned} \right\}. \tag*{(3)}

Here Sur⁡\operatorname{Sur} denotes continuous surjective homomorphisms. All the data GG, HH, π\pi, SS, and D\mathcal{D} remain fixed as X→∞X\to\infty.

Theorem 1.1. Put B=∣H[L:Q]∣−1B=|H^{[L:\mathbb{Q}]}|-1. There is a constant C>0C>0 such that

NG,π,D(X)≤CXα(log⁡(X))BN_{G,\pi,\mathcal{D}}(X)\leq CX^\alpha(\log(X))^B

for every X≥2X \ge2. If a proper solution satisfying the local conditions exists, then there are a constant c>0c > 0 and an integer b≥1b \ge1 such that

NG,π,D(X)≥cXα(log⁡X)b−1N_{G,\pi,\mathcal{D}}(X) \ge cX^\alpha(\log X)^{b-1}

for all sufficiently large XX. In particular, every nonempty fiber satisfies

lim⁡X→∞log⁡NG,π,D(X)log⁡X=p∣G∣(p−1).\lim_{X\to\infty}\frac{\log N_{G,\pi,\mathcal{D}}(X)}{\log X}=\frac{p}{|G|(p-1)}.

For the asymptotic, suppose that every element of order pp in HH belongs to Z(H)Z(H). Then

V={h∈H:hp=1}.(4)V=\{h\in H:h^p=1\}. \tag*{(4)}

Regard VV as an Fp\mathbb{F}_p-vector space, and let V(−1)V(-1) denote the Galois module with action

σ⋆v=χ‾p(σ)−1π(σ)(v).(5)\sigma\star v=\overline{\chi}_p(\sigma)^{-1}\pi(\sigma)(v). \tag*{(5)}

Here π(σ)(v)\pi(\sigma)(v) denotes conjugation by any lift of π(σ)\pi(\sigma) to GG. This is independent of the lift because HH centralizes VV. Recall that bV=#(GQ\(V(−1)∖{0}))b_V=\#(G_{\mathbb{Q}}\backslash(V(-1)\setminus\{0\})).

Theorem 1.2. Suppose that every element of order pp in HH is central in HH. If the fiber in (3) contains a proper solution, then

NG,π,D(X)∼CG,π,DXp/(∣G∣(p−1))(log⁡X)bV−1(6)N_{G,\pi,\mathcal{D}}(X)\sim C_{G,\pi,\mathcal{D}}X^{p/(|G|(p-1))}(\log X)^{b_V-1} \tag*{(6)}

for a finite positive constant CG,π,DC_{G,\pi,\mathcal{D}}.

For T=1T=1 and no local restrictions, our proof of Theorem 1.2 simplifies the proof of the pp-group case over Q\mathbb{Q} of Koymans and Pagano’s theorem [11 Theorem 1.1]. Our arguments do not rely on their coordinate parametrization. Instead it uses the abelian counting theorem of Alberts and O’Dorney applied to a fibrew-ise argument.

The principal application concerns Gj=GL⁡d(Z/pjZ)G_j=\operatorname{GL}_d(\mathbb{Z}/p^j\mathbb{Z}) for j≥1j\ge1, with m≥2m\ge2 and a fixed residual representation ρ‾:GQ↠G1=GL⁡d(Fp)\overline{\rho}:G_{\mathbb{Q}}\twoheadrightarrow G_1=\operatorname{GL}_d(\mathbb{F}_p). For odd pp, every element of order pp in Hm=ker⁡(Gm→G1)H_m=\ker(G_m\to G_1) belongs to Vm=ker⁡(Gm→Gm−1)V_m=\ker(G_m\to G_{m-1}). This subgroup is central in HmH_m and is isomorphic to Md(Fp)M_d(\mathbb{F}_p).

Write Nρ‾,m,D(X)N_{\overline{\rho},m,\mathcal{D}}(X) for the count in (3) with G=GmG=G_m, H=HmH=H_m, T=G1T=G_1, and π=ρ‾\pi=\overline{\rho}, where the quotient map Gm↠G1G_m\twoheadrightarrow G_1 is reduction modulo pp. Thus it counts surjective lifts of ρ‾\overline{\rho} satisfying D\mathcal{D} whose splitting fields have absolute discriminant at most XX.

Corollary 1.3. Let pp be odd. If a surjective lift of ρ‾\overline{\rho} to GmG_m satisfying the finite local conditions exists, then

Nρ‾,m,D(X)∼Cρ‾,m,DXαm(log⁡X)bρ‾−1,N_{\overline{\rho},m,\mathcal{D}}(X)\sim C_{\overline{\rho},m,\mathcal{D}}X^{\alpha_m}(\log X)^{b_{\overline{\rho}}-1},

where αm=p/(∣Gm∣(p−1))\alpha_m=p/(|G_m|(p-1)) and bρ‾b_{\overline{\rho}} is the number of orbits on Md(Fp)∖{0}M_d(\mathbb{F}_p)\setminus\{0\} for the action σ⋆A=χ‾p(σ)−1ρ‾(σ)Aρ‾(σ)−1\sigma\star A=\overline{\chi}_p(\sigma)^{-1}\overline{\rho}(\sigma)A\overline{\rho}(\sigma)^{-1}. In particular, bρ‾b_{\overline{\rho}} is independent of mm.

Let us make some remarks on the nonemptiness hypothesis and how it relates to well established lifting theorems for Galois representations. Ramakrishna [16] developed a method for constructing characteristic-zero lifts by allowing auxiliary ramification. In dimension two, his large-image theorem [17 Theorem 1.1] shows that, for p≥7p\ge7, every surjective GL⁡2(Fp)\operatorname{GL}_2(\mathbb{F}_p)-representation admits a lift with image containing SL⁡2(Zp)\operatorname{SL}_2(\mathbb{Z}_p). A refined argument yields a lift surjective onto GL⁡2(Zp)\operatorname{GL}_2(\mathbb{Z}_p); see Proposition 8.3. Reduction modulo pmp^m therefore removes the nonemptiness hypothesis from Corollary 1.3.

Related constructions were developed by Khare, Larsen and Ramakrishna [9], who construct pp-adic lifts with prescribed Frobenius properties. Fakhruddin, Khare and Patrikis [8] subsequently established lifting theorems for irreducible odd representations valued in general reductive groups over totally real fields, assuming sufficiently large residual characteristic and suitable local lifts. Whenever a lifting theorem supplies a surjective lift realizing the prescribed local conditions, its reductions provide the proper solutions required by Corollary 1.3.

Declarations

Conflict of interest: Not applicable; there is no conflict of interest to report.

Availability of data and materials: No data was generated or analyzed in obtaining the results in this article.

Funding: No funding was received during the preparation of this article.

Use of artificial intelligence: The author used ChatGPT 6 Astra as an exploratory tool to study related papers and to assist in correcting mistakes in the article. It was not used to generate the mathematical ideas or the main results of the article.

Preliminary facts

In this short preliminary section we identify Alberts’s logarithmic invariant, bound the number of surjections defining a fixed field, and recall the tame discriminant formula. Recall that α=p/(∣G∣(p−1))\alpha= p/(|G|(p-1)), and write a=α−1a = \alpha^{-1}. Let Bρ0\mathcal{B}_{\rho_0}, bρ0(H)b_{\rho_0}(H) and bπb_{\pi} be the invariants defined in the introduction.

Lemma 2.1. If a proper solution ρ0\rho_0 exists, then Bρ0=bρ0(H)=bπ\mathcal{B}_{\rho_0} = b_{\rho_0}(H) = b_{\pi}.

Proof. Every nontrivial subgroup N⊴GN \trianglelefteq G contained in HH contains an element of order pp, and all its elements have pp-power order. Thus a(N)=a(H)a(N) = a(H). If φ≡ρ0(modN)\varphi\equiv\rho_0 \pmod N, their conjugation actions on NN-conjugacy classes agree. Their twisted actions also agree, so bφ(N)=bρ0(N)b_{\varphi}(N) = b_{\rho_0}(N).

Put K=ker⁡χ‾pK = \ker\overline{\chi}_p. Since ρ0\rho_0 is surjective, ρ0(K)\rho_0(K) is normal in GG and G/ρ0(K)G/\rho_0(K) is a quotient of GQ/KG_{\mathbb{Q}}/K. Its order therefore divides p−1p-1. The image of the pp-group HH in this quotient is trivial, so H⊆ρ0(K)H \subseteq\rho_0(K).

Let ΩN\Omega_N be the set of elements of order pp in NN. The action

σ⋆x=ρ0(σ)xχ‾p(σ)−1ρ0(σ)−1\sigma\star x = \rho_0(\sigma)x^{\overline{\chi}_p(\sigma)^{-1}}\rho_0(\sigma)^{-1}

on ΩH\Omega_H preserves ΩN\Omega_N. For every h∈Hh \in H, choose σ∈K\sigma\in K with ρ0(σ)=h\rho_0(\sigma) = h. The action of this σ\sigma is conjugation by hh. In particular, each NN-conjugacy class in ΩN\Omega_N lies in a single GQG_{\mathbb{Q}}-orbit. Consequently, passing to NN-conjugacy classes induces a bijection on GQG_{\mathbb{Q}}-orbit sets, and bρ0(N)=#(GQ\ΩN)b_{\rho_0}(N) = \#(G_{\mathbb{Q}}\backslash\Omega_N).

Since ΩN\Omega_N is an invariant subset of ΩH\Omega_H, bρ0(N)≤bρ0(H)b_{\rho_0}(N) \leq b_{\rho_0}(H). The subgroup N=HN = H attains equality. The two maxima defining Bρ0\mathcal{B}_{\rho_0} therefore give bρ0(H)b_{\rho_0}(H). On HH-conjugacy classes, the action through ρ0\rho_0 agrees with the action through π\pi defined in the introduction. Thus bρ0(H)=bπb_{\rho_0}(H) = b_{\pi}. □

Lemma 2.2. A field F/QF/\mathbb{Q} occurs as FρF_{\rho} for at most ∣Aut⁡(G)∣|\operatorname{Aut}(G)| of the surjections in (1.3).

Proof. Every surjection with kernel field FF induces an isomorphism Gal⁡(F/Q)≃G\operatorname{Gal}(F/\mathbb{Q}) \simeq G. Fix one such isomorphism ι\iota. Every other one is uniquely γ∘ι\gamma\circ\iota with γ∈Aut⁡(G)\gamma\in\operatorname{Aut}(G). □

Lemma 2.3. Let F/QF/\mathbb{Q} be Galois of degree gg. At a tamely ramified rational prime ℓ\ell with inertia order ee,

vℓ(ΔF)=g(1−e−1).(7)v_{\ell}(\Delta_F) = g(1-e^{-1}). \tag*{(7)}

For a proper solution of (1.1), inertia at a ramified prime ℓ∉S\ell\notin S is tame and cyclic of order pp, with r≥1r \geq1. In particular, vℓ(ΔFρ)≥av_{\ell}(\Delta_{F_{\rho}}) \geq a.

Proof. The first assertion follows from [7 Corollary 4.16], since F/QF/\mathbb{Q} is Galois.

Now suppose that ℓ∉S\ell\notin S ramifies in FρF_{\rho}. Since L/QL/\mathbb{Q} is unramified at ℓ\ell, we have π(Iℓ)=1\pi(I_{\ell}) = 1. The equality φ∘ρ=π\varphi\circ\rho= \pi gives ρ(Iℓ)⊆H\rho(I_{\ell}) \subseteq H.

The image of wild inertia has ℓ\ell-power order and is contained in the pp-group HH. Since ℓ≠p\ell\ne p, this image is trivial. Thus ρ(Iℓ)\rho(I_{\ell}) is a finite cyclic quotient of tame inertia. It is nontrivial because ℓ\ell ramifies, so its order is prp^{r} for some r≥1r \ge1. Applying (7) gives

vℓ(ΔFρ)=∣G∣(1−p−r)≥∣G∣(1−p−1)=a.v_{\ell}(\Delta_{F_{\rho}})=|G|(1-p^{-r})\ge|G|(1-p^{-1})=a.

A fixed-field upper bound

In this section, we bound the number of proper solutions with prescribed ramification support. This gives the upper bound in Theorem 1.1.

Let KK be a number field and let Σ\Sigma be a finite set of its finite places containing the places above pp. Let KΣK_{\Sigma} be the maximal extension of KK unramified outside Σ\Sigma and the infinite places. Set GK,Σ=Gal⁡(KΣ/K)G_{K,\Sigma}=\operatorname{Gal}(K_{\Sigma}/K). Let PK,ΣP_{K,\Sigma} be its maximal pro-pp quotient, and put XK,Σ=Hom⁡(GK,Σ,Fp)=Hom⁡(PK,Σ,Fp)X_{K,\Sigma}=\operatorname{Hom}(G_{K,\Sigma},\mathbb{F}_p)=\operatorname{Hom}(P_{K,\Sigma},\mathbb{F}_p). Here Fp\mathbb{F}_p is additive, with trivial action. Ramification at real places is allowed.

For a positive squarefree integer RR, let ΣR\Sigma_R be obtained from Σ\Sigma by adjoining the places above the prime divisors of RR. Write ω(R)\omega(R) for the number of these prime divisors, and d(P)d(P) for the minimum number of topological generators of a pro-pp group PP.

Lemma 3.1. Let PP be a pro-pp group. If r=dim⁡FpHom⁡(P,Fp)<∞r=\dim_{\mathbb{F}_p}\operatorname{Hom}(P,\mathbb{F}_p)<\infty, then PP is finitely generated and d(P)=rd(P)=r.

Proof. For the trivial PP-module Fp\mathbb{F}_p, we have H1(P,Fp)=Hom⁡(P,Fp)H^1(P,\mathbb{F}_p)=\operatorname{Hom}(P,\mathbb{F}_p). The assertion follows from [15 Proposition 3.9.1].

Lemma 3.2. The group PK,ΣP_{K,\Sigma} is finitely generated. If cK,Σ=d(PK,Σ)c_{K,\Sigma}=d(P_{K,\Sigma}), then

d(PK,ΣR)≤cK,Σ+[K:Q]ω(R).(8)d(P_{K,\Sigma_R})\le c_{K,\Sigma}+[K:\mathbb{Q}]\omega(R). \tag*{(8)}

Proof. By [15 Theorem 8.3.20(i)], with the infinite places included, the spaces XK,ΣX_{K,\Sigma} and XK,ΣRX_{K,\Sigma_R} are finite. Lemma 3.1 therefore shows that the corresponding pro-pp groups are finitely generated and that their generator ranks equal the dimensions of these spaces.

Inflation identifies XK,ΣX_{K,\Sigma} with a subspace of XK,ΣRX_{K,\Sigma_R}. Restriction to inertia induces an injection

XK,ΣR/XK,Σ↪⨁l∈ΣR∖ΣHom⁡(Il,Fp).X_{K,\Sigma_R}/X_{K,\Sigma}\hookrightarrow\bigoplus_{\mathfrak l\in\Sigma_R\setminus\Sigma}\operatorname{Hom}(I_{\mathfrak l},\mathbb{F}_p).

Indeed, a character trivial on inertia at every new place is unramified outside Σ\Sigma and the infinite places, and hence factors through GK,ΣG_{K,\Sigma}.

Every new place has residue characteristic different from pp. Its wild inertia has trivial image in Fp\mathbb{F}_p, and the maximal pro-pp quotient of tame inertia is procyclic. Thus each summand has dimension at most one. There are at most [K:Q]ω(R)[K:\mathbb{Q}]\omega(R) new places, so

d(PK,ΣR)=dim⁡FpXK,ΣR≤dim⁡FpXK,Σ+#(ΣR∖Σ)≤cK,Σ+[K:Q]ω(R).\begin{aligned} d(P_{K,\Sigma_R})&=\dim_{\mathbb{F}_p}X_{K,\Sigma_R}\\ &\le\dim_{\mathbb{F}_p}X_{K,\Sigma}+\#(\Sigma_R\setminus\Sigma)\\ &\le c_{K,\Sigma}+[K:\mathbb{Q}]\omega(R). \end{aligned}

This proves (8).

Lemma 3.3. Put A=∣H∣[L:Q]A=|H|^{[L:\mathbb{Q}]}. The number of proper solutions unramified outside S∪supp⁡(R)S\cup\operatorname{supp}(R) is at most

CAω(R),(9)CA^{\omega(R)}, \tag*{(9)}

where CC depends only on the fixed data. If V⊆HV\subseteq H is normal in GG, the same assertion holds for G/VG/V, with AV=∣H/V∣[L:Q]≤AA_V=|H/V|^{[L:\mathbb{Q}]}\le A in place of AA.

Proof. Let Σ\Sigma consist of the finite places of LL above SS. For every solution under consideration, Fρ/LF_\rho/L is an HH-extension unramified outside ΣR\Sigma_R and the infinite places. Hence it is the kernel field of a surjection PL,ΣR↠HP_{L,\Sigma_R} \twoheadrightarrow H.

A continuous homomorphism from a group with dd topological generators to HH is determined by the dd generator images. Indeed, two such homomorphisms agreeing on the generators agree on a dense subgroup and therefore everywhere. Thus there are at most ∣H∣d|H|^d homomorphisms. Distinct kernel fields give distinct kernels, so Lemma 3.2 bounds the number of possible fields by

∣H∣cL,Σ+[L:Q]ω(R)=∣H∣cL,ΣAω(R).|H|^{c_{L,\Sigma}+[L:\mathbb{Q}]\omega(R)}=|H|^{c_{L,\Sigma}}A^{\omega(R)}.

This counts all HH-extensions of LL with the stated ramification. Requiring Galoisness over Q\mathbb{Q} only decreases their number. Lemma 2.2 bounds the number of surjections for each remaining field by ∣Aut⁡(G)∣|\operatorname{Aut}(G)|, proving (9). For G/VG/V, the kernel over LL is H/VH/V. The same proof applies with H/VH/V in place of HH. If H/V=1H/V=1, the only quotient field is LL, so the assertion also holds in this case. □

Proposition 3.4. The upper bound in Theorem 1.1 holds without a nonemptiness hypothesis.

Proof. For a proper solution put Q(ρ)=∏ℓ∉S, ℓ ramified in FρℓQ(\rho)=\prod_{\ell\notin S,\,\ell\ \mathrm{ramified\ in}\ F_\rho}\ell. Lemma 2.3 gives Q(ρ)a≤∣ΔFρ∣Q(\rho)^a\leq|\Delta_{F_\rho}|. Thus ∣ΔFρ∣≤X|\Delta_{F_\rho}|\leq X implies Q(ρ)≤Y=X1/aQ(\rho)\leq Y=X^{1/a}. Partitioning by Q(ρ)Q(\rho) and applying Lemma 3.3 gives

NG,π,D(X)≤C∑n≤Yμ2(n)Aω(n),(10)N_{G,\pi,D}(X)\leq C\sum_{n\leq Y}\mu^2(n)A^{\omega(n)}, \tag*{(10)}

where μ\mu is the Möbius function.

For the positive integer AA, let τA(n)\tau_A(n) count ordered AA-tuples of positive integers whose product is nn. When nn is squarefree, each prime factor is assigned to one of the AA entries, so τA(n)=Aω(n)\tau_A(n)=A^{\omega(n)}. For all nn, it follows that μ2(n)Aω(n)≤τA(n)\mu^2(n)A^{\omega(n)}\leq\tau_A(n).

We prove by induction on AA that

DA(Y):=∑n≤YτA(n)≪AY(log⁡(2Y))A−1(11)D_A(Y):=\sum_{n\leq Y}\tau_A(n)\ll_A Y(\log(2Y))^{A-1} \tag*{(11)}

where Y≥1Y\geq1. For A=1A=1, this is ⌊Y⌋≤Y\lfloor Y\rfloor\leq Y. If A≥2A\geq2, the convolution identity τA=1∗τA−1\tau_A=1*\tau_{A-1} and the inductive bound give

DA(Y)=∑m≤YDA−1(Y/m)D_A(Y)=\sum_{m\leq Y}D_{A-1}(Y/m)
≪AY∑m≤Y(log⁡(2Y/m))A−2m\ll_A Y\sum_{m\leq Y}\frac{(\log(2Y/m))^{A-2}}{m}
≪AY(log⁡(2Y))A−2∑m≤Y1m\ll_A Y(\log(2Y))^{A-2}\sum_{m\leq Y}\frac{1}{m}
≪AY(log⁡(2Y))A−1.\ll_A Y(\log(2Y))^{A-1}.

Here ∑m≤Ym−1≤1+log⁡Y≪log⁡(2Y)\sum_{m\leq Y}m^{-1}\leq1+\log Y\ll\log(2Y). This proves (11).

Combining (10) and (11), and using 1/a=α1/a=\alpha and log⁡(2X1/a)≪alog⁡(X)\log(2X^{1/a})\ll_a\log(X), yields

NG,π,D(X)≪Xα(log⁡(X))A−1.N_{G,\pi,D}(X)\ll X^\alpha(\log(X))^{A-1}.

Since A−1=BA-1=B, this is the asserted upper bound. □

Abelian fibers and the lower bound

In this section, we parametrize lifts with an elementary abelian kernel in terms of Galois cocycles and apply the counting results of Alberts and O’Dorney. Choosing such a kernel inside Z(H)Z(H) then gives the lower bound in Theorem 1.1. Let VV be any nonzero GG-normal elementary abelian subgroup of Z(H)Z(H). Conjugation on VV factors through TT, so π\pi makes VV a Galois module. We set

b(V)=#(GQ\(V(−1)∖{0})).b(V)=\#\left(G_{\mathbb{Q}}\backslash\left(V(-1)\setminus\{0\}\right)\right).

This equals bVb_V when VV is the subgroup in (1.4).

Let qV:G/V↠Tq_V:G/V\twoheadrightarrow T be the map induced by qq. Fix a surjection ψ:GQ↠G/V\psi:G_{\mathbb{Q}}\twoheadrightarrow G/V satisfying qV∘ψ=πq_V\circ\psi=\pi. Define Nψ(X)N_\psi(X) to be the number of surjections ρ\rho counted by (1.3) whose composition with the quotient map G→G/VG\to G/V equals ψ\psi.

Lemma 4.1. Let ρ0:GQ→G\rho_0:G_{\mathbb{Q}}\to G lift ψ\psi. The assignment

f⟼ρfdefined byρf(σ)=f(σ)ρ0(σ),(12)f\longmapsto\rho_f\quad\text{defined by}\quad\rho_f(\sigma)=f(\sigma)\rho_0(\sigma), \tag*{(12)}

is a bijection from Z1(Q,V)Z^1(\mathbb{Q},V) to the lifts of ψ\psi. It induces a bijection from H1(Q,V)H^1(\mathbb{Q},V) to their VV-conjugacy classes. Every class contains

∣B1(Q,V)∣=∣V∣∣VGQ∣(13)\left|B^1(\mathbb{Q},V)\right|=\frac{|V|}{|V^{G_{\mathbb{Q}}}|} \tag*{(13)}

actual homomorphisms. These assertions also hold locally upon replacing GQG_{\mathbb{Q}} by any decomposition group.

Proof. The proof is straightforward and thus is omitted. □

Proposition 4.2. If ψ\psi has a surjective lift satisfying the local conditions, then there is a constant cψ>0c_\psi>0 such that

Nψ(X)∼cψXα(log⁡X)b(V)−1.(14)N_\psi(X)\sim c_\psi X^\alpha(\log X)^{b(V)-1}. \tag*{(14)}

Otherwise Nψ(X)=0N_\psi(X)=0 for every XX.

Proof. Choose a surjective lift ρ0\rho_0 satisfying the local conditions. We apply the Alberts–O’Dorney theorem [6 Theorem 1.1 and Corollary 1.2]. We check the local conditions, the ordering, and surjectivity.

For v∈Sv\in S, let LvL_v be the subset of H1(Qv,V)H^1(\mathbb{Q}_v,V) consisting of the classes whose twists of ρ0∣GQv\rho_0|_{G_{\mathbb{Q}_v}} belong to DvD_v. This is well defined since changing a cocycle by a coboundary conjugates its twist by an element of V⊂HV\subset H. For v∉Sv\notin S, take Lv=H1(Qv,V)L_v=H^1(\mathbb{Q}_v,V). Let HL1(Q,V)H^1_L(\mathbb{Q},V) denote the global classes satisfying these restrictions.

Set g=∣G∣g=|G|, and let AfA_f be the degree-gg étale algebra associated to ρf\rho_f in the regular permutation representation. Define D([f])=∣disc⁡(Af)∣D([f])=|\operatorname{disc}(A_f)|. Conjugation does not change AfA_f, so this depends only on [f][f]. To describe it explicitly, set Uf=ρf(GQ)U_f=\rho_f(G_{\mathbb{Q}}) and Kf=Qker⁡ρfK_f=\mathbb{Q}^{\ker\rho_f}. The action of UfU_f on GG is free, with g/∣Uf∣g/|U_f| orbits. We find that

Af≃∏j=1g/∣Uf∣KfA_f\simeq\prod_{j=1}^{g/|U_f|}K_f

and consequently,

D([f])=∣ΔKf∣g/∣Uf∣.D([f])=\left|\Delta_{K_f}\right|^{g/|U_f|}.

Enlarge SS to a finite set S0S_0 containing the primes ramified in ρ0\rho_0. For ℓ∉S0\ell\notin S_0, both ρ0\rho_0 and VV are unramified, and ℓ≠p\ell\ne p. Hence ρf(Iℓ)=f(Iℓ)⊂V\rho_f(I_\ell)=f(I_\ell)\subset V. Wild inertia has trivial image in VV. Its nontrivial tame image is cyclic of order pp. Such an inertia generator acts on GG as g/pg/p cycles of length pp. The tame discriminant formula gives $v_\ell(D([f]))=a for a ramified class and 0 for an unramified class. Thus the minimum exponent in the counting theorem is a=α−1a=\alpha^{-1}.

We also compute its logarithmic exponent. Inflation–restriction at ℓ∉S0\ell\notin S_0 gives

0⟶Hur1(Qℓ,V)⟶H1(Qℓ,V)⟶V(−1)Frob⁡ℓ⟶0.0\longrightarrow H^1_{\mathrm{ur}}(\mathbb{Q}_{\ell},V)\longrightarrow H^1(\mathbb{Q}_{\ell},V)\longrightarrow V(-1)^{\operatorname{Frob}_{\ell}}\longrightarrow0.

Indeed, inertia acts trivially on VV, its maximal pro-pp tame quotient is Zp(1)\mathbb{Z}_p(1), and the unramified quotient has pp-cohomological dimension one. The inertia value ww satisfies Frob⁡ℓ(w)=ℓw\operatorname{Frob}_{\ell}(w)=\ell w, which gives the indicated inverse cyclotomic twist.

Every inertia value has ∣Hur1(Qℓ,V)∣|H^1_{\mathrm{ur}}(\mathbb{Q}_{\ell},V)| preimages. This kernel is V/(Frob⁡ℓ−1)VV/(\operatorname{Frob}_{\ell}-1)V, whose order is ∣VFrob⁡ℓ∣=∣H0(Qℓ,V)∣|V^{\operatorname{Frob}_{\ell}}|=|H^0(\mathbb{Q}_{\ell},V)|. The normalized local factor is consequently

1∣H0(Qℓ,V)∣∑[fℓ]∈H1(Qℓ,V)ℓ−svℓ(Dℓ([fℓ]))=1+(∣V(−1)Frob⁡ℓ∣−1)ℓ−as.\frac{1}{|H^0(\mathbb{Q}_{\ell},V)|} \sum_{[f_{\ell}]\in H^1(\mathbb{Q}_{\ell},V)} \ell^{-s v_{\ell}(D_{\ell}([f_{\ell}]))} =1+\left(|V(-1)^{\operatorname{Frob}_{\ell}}|-1\right)\ell^{-as}.

Here DℓD_{\ell} is the local discriminant formed by twisting ρ0∣GQℓ\rho_0|_{G_{\mathbb{Q}_{\ell}}}.

The action on V(−1)V(-1) factors through Γ=Gal⁡(L(μp)/Q)\Gamma=\operatorname{Gal}(L(\mu_p)/\mathbb{Q}). Thus these factors are Frobenian. The logarithmic parameter of the counting theorem is the Chebotarev average of their coefficients at ℓ−as\ell^{-as}. Burnside’s lemma identifies it with

1∣Γ∣∑γ∈Γ(∣V(−1)γ∣−1)=b(V).\frac{1}{|\Gamma|}\sum_{\gamma\in\Gamma}\left(|V(-1)^{\gamma}|-1\right)=b(V).

Theorem 1.1 of [6] now gives a constant Cψ>0C_{\psi}>0 such that

#{[f]∈HL1(Q,V):D([f])≤X}∼CψXα(log⁡X)b(V)−1.\#\{[f]\in H^1_L(\mathbb{Q},V):D([f])\leq X\}\sim C_{\psi}X^{\alpha}(\log X)^{b(V)-1}.

Since the chosen ρ0\rho_0 is surjective, [6], Corollary 1.2(i)(a) states that the proportion of these classes whose twists are surjective converges to a positive number ϑψ\vartheta_{\psi}.

Surjectivity and the local conditions are invariant under VV-conjugation. Apply Lemma 4.1 to convert the count of surjective cohomology classes to actual homomorphisms. This gives (14) with cψ=∣V∣ϑψCψ/∣VGQ∣>0c_{\psi}=|V|\vartheta_{\psi}C_{\psi}/|V^{G_{\mathbb{Q}}}|>0. If there is no surjective lift compatible with local conditions, the defining set for Nψ(X)N_{\psi}(X) is empty. □

Lemma 4.3. There is a nonzero GG-normal elementary abelian subgroup of Z(H)Z(H).

Proof. The proof of this lemma is straightforward and thus is left to the reader. □

Proof of the lower bound in Theorem 1.1. Choose a compatible proper solution ρ0\rho_0, and choose VV as in Lemma 4.3. Put ψ=ρ0 mod V\psi=\rho_0\bmod V. Every proper lift of ψ\psi has quotient π\pi, so Proposition 4.2 gives

NG,π,D(X)≥Nψ(X)∼cψXα(log⁡X)b(V)−1.N_{G,\pi,D}(X)\geq N_{\psi}(X)\sim c_{\psi}X^{\alpha}(\log X)^{b(V)-1}.

The set V(−1)∖{0}V(-1)\setminus\{0\} is nonempty, so b(V)≥1b(V)\geq1. Take b=b(V)b=b(V) and c=cψ/2c=c_{\psi}/2 for all sufficiently large XX.

Together with Proposition 3.4, the lower bound gives log⁡NG,π,D(X)=αlog⁡X+O(log⁡log⁡(X))\log N_{G,\pi,D}(X)=\alpha\log X+O(\log\log(X)). Divide by log⁡X\log X to obtain the stated limit. □

A uniform estimate for abelian fibers

In this section, we prove an upper bound for Nψ(X)N_{\psi}(X) whose constant is independent of ψ\psi. Assume the hypothesis of Theorem 1.2, and use VV from (4). In particular, VV is normal in GG, is elementary abelian, and is centralized by HH. Its GQG_{\mathbb{Q}}-action therefore depends only on π\pi. Set

a′=∣G∣(1−p−2)anda′a=1+1p.(15)a'=|G|(1-p^{-2})\qquad\text{and}\qquad\frac{a'}{a}=1+\frac{1}{p}. \tag*{(15)}

All constants in this section may depend on the fixed data and on SS, but not on the quotient representation ψ\psi. We let GQ,SG_{\mathbb{Q},S} be the Galois group of the maximal extension of Q\mathbb{Q} unramified at all finite primes outside SS.

Lemma 5.1. The group Z1(GQ,S,V)Z^{1}(G_{\mathbb{Q},S}, V) is finite.

Proof. Since SS is finite, p∈Sp \in S, and VV is a finite pp-primary GQ,SG_{\mathbb{Q},S}-module, [15 Theorem 8.3.20(i)] shows that H1(GQ,S,V)H^{1}(G_{\mathbb{Q},S}, V) is finite.

The coboundary group B1(GQ,S,V)B^{1}(G_{\mathbb{Q},S}, V) is an image of VV, so it is finite. Since H1(GQ,S,V)=Z1(GQ,S,V)/B1(GQ,S,V)H^{1}(G_{\mathbb{Q},S}, V) = Z^{1}(G_{\mathbb{Q},S}, V)/B^{1}(G_{\mathbb{Q},S}, V), the cocycle group is also finite. □

For each finite ℓ∉S\ell\notin S, choose an arithmetic Frobenius ϕℓ\phi_{\ell} in the fixed decomposition group. Also choose τℓ∈Iℓ\tau_{\ell} \in I_{\ell} whose image generates the Zp\mathbb{Z}_{p}-quotient of tame inertia. Since ℓ≠p\ell\ne p, every continuous homomorphism Iℓ→HI_{\ell} \to H kills wild inertia and factors through this procyclic quotient. It is therefore determined by the image of τℓ\tau_{\ell}.

In the tame quotient of the decomposition group, conjugation by ϕℓ\phi_{\ell} sends an inertia element to its ℓ\ellth power. We use this relation only after applying a representation that kills wild inertia. The inertia action on VV is trivial. Define

Wℓ={v∈V:ϕℓ(v)=ℓv},nℓ=∣Wℓ∣−1.(16)W_{\ell} = \{v \in V : \phi_{\ell}(v) = \ell v\}, \qquad n_{\ell} = |W_{\ell}| - 1. \tag*{(16)}

Since χ‾p(ϕℓ)=ℓ mod p\overline{\chi}_{p}(\phi_{\ell}) = \ell\bmod p, the number nℓn_{\ell} is the number of fixed points of ϕℓ\phi_{\ell} on V(−1)∖{0}V(-1) \setminus\{0\}.

Lemma 5.2. For every Y≥1Y \ge1,

∑Q≤YQ squarefree(Q,S)=1∏ℓ∣Qnℓ≪Y(log⁡(2Y))bV−1.(17)\sum_{\substack{Q \le Y\\ Q\ \mathrm{squarefree}\\ (Q,S)=1}} \prod_{\ell\mid Q} n_{\ell} \ll Y(\log(2Y))^{b_V-1}. \tag*{(17)}

Here (Q,S)=1(Q,S) = 1 means that no finite prime in SS divides QQ. The summand for Q=1Q = 1 is one.

Proof. Put E=L(μp)E = L(\mu_p) and Γ=Gal⁡(E/Q)\Gamma= \operatorname{Gal}(E/\mathbb{Q}). The action on V(−1)V(-1) factors through Γ\Gamma, and every prime ramified in EE belongs to SS. Decompose the finite Γ\Gamma-set as

V(−1)∖{0}=∐j=1bVΓ/Jj,V(-1) \setminus\{0\} = \coprod_{j=1}^{b_V} \Gamma/J_j,

and put Ej=EJjE_j = E^{J_j}. These fields remain fixed throughout the argument.

Let aK(n)a_K(n) be the number of nonzero integral ideals of a number field KK with absolute norm nn. For ℓ∉S\ell\notin S, the cycles of ϕℓ\phi_{\ell} on Γ/Jj\Gamma/J_j correspond to the primes of EjE_j above ℓ\ell. Their lengths are the residue degrees. Thus a fixed point corresponds to a prime ideal of norm ℓ\ell, and

nℓ=∑j=1bVaEj(ℓ).(18)n_{\ell} = \sum_{j=1}^{b_V} a_{E_j}(\ell). \tag*{(18)}

For fixed KK, put AK(Y)=∑n≤YaK(n)A_K(Y) = \sum_{n \le Y} a_K(n). The ideal-counting estimate is AK(Y)≪KYA_K(Y) \ll_K Y; see [14]. Partial summation gives

∑n≤YaK(n)n=AK(Y)Y+∫1YAK(t)t2 dt≪Klog⁡(2Y).(19)\sum_{n \le Y} \frac{a_K(n)}{n} = \frac{A_K(Y)}{Y} + \int_1^Y \frac{A_K(t)}{t^2}\,dt \ll_K \log(2Y). \tag*{(19)}

For 1≤r≤bV1 \le r \le b_V, let cr=aE1∗⋯∗aErc_r = a_{E_1} * \cdots* a_{E_r}, where ∗* denotes Dirichlet convolution. We claim that

∑n≤Ycr(n)≪Y(log⁡(2Y))r−1.(20)\sum_{n \le Y} c_r(n) \ll Y(\log(2Y))^{r-1}. \tag*{(20)}

For r=1r=1 this is the ideal-counting estimate. Suppose r≥2r \ge2 and that the assertion holds for r−1r-1. Then

∑n≤Ycr(n)=∑m≤YaEr(m)∑n≤Y/mcr−1(n)≪Y∑m≤YaEr(m)m(log⁡(2Y/m))r−2≤Y(log⁡(2Y))r−2∑m≤YaEr(m)m≪Y(log⁡(2Y))r−1.\begin{aligned} \sum_{n\le Y} c_r(n)=\sum_{m\le Y}a_{E_r}(m)\sum_{n\le Y/m}c_{r-1}(n) \ll Y\sum_{m\le Y}\frac{a_{E_r}(m)}{m}\left(\log(2Y/m)\right)^{r-2} \\ &\le Y(\log(2Y))^{r-2}\sum_{m\le Y}\frac{a_{E_r}(m)}{m} \\ &\ll Y(\log(2Y))^{r-1}. \end{aligned}

The last estimate is (19). This proves (20) by induction.

The functions aEja_{E_j} are nonnegative and multiplicative. Their convolution is therefore nonnegative and multiplicative. By (18), cbV(ℓ)=nℓc_{b_V}(\ell)=n_\ell for ℓ∉S\ell\notin S. Hence cbV(Q)=∏ℓ∣Qnℓc_{b_V}(Q)=\prod_{\ell\mid Q}n_\ell for every squarefree QQ prime to SS. The sum in (17) is at most ∑n≤YcbV(n)\sum_{n\le Y}c_{b_V}(n), and (20) proves the lemma. □

For a surjection ψ:GQ↠G/V\psi:G_{\mathbb{Q}}\twoheadrightarrow G/V above π\pi, write

R(ψ)=∏ℓ∉Sψ(Iℓ)≠1ℓ.(21)R(\psi)=\prod_{\substack{\ell\notin S\\\psi(I_\ell)\ne1}}\ell. \tag*{(21)}

This is squarefree and finite. As in § 4, Nψ(X)N_\psi(X) counts the proper lifts of ψ\psi satisfying the prescribed local conditions and the discriminant bound.

Proposition 5.3. There is a constant CC such that, for every ψ\psi and every X≥2X\ge2,

Nψ(X)≤C∣V∣ω(R(ψ))R(ψ)−a′/aX1/a(log⁡(X))bV−1.(22)N_\psi(X)\le C|V|^{\omega(R(\psi))}R(\psi)^{-a'/a}X^{1/a}(\log(X))^{b_V-1}. \tag*{(22)}

The constant is independent of ψ\psi and of its restrictions at the places in SS.

Proof. Write R=R(ψ)R=R(\psi). For a lift ρ\rho counted by Nψ(X)N_\psi(X), put

Q(ρ)=∏ℓ∉S,ℓ∤Rρ(Iℓ)≠1ℓ.Q(\rho)=\prod_{\substack{\ell\notin S,\ell\nmid R\\\rho(I_\ell)\ne1}}\ell.

Then Q(ρ)Q(\rho) is squarefree and prime to RR. By Lemma 2.3, each inertia image outside SS is a cyclic pp-subgroup of HH generated by ρ(τℓ)\rho(\tau_\ell).

If ℓ∣R\ell\mid R, the image of ρ(τℓ)\rho(\tau_\ell) in G/VG/V is nontrivial. Thus ρ(τℓ)∉V\rho(\tau_\ell)\notin V. Every element of order pp in HH belongs to VV, so ord⁡(ρ(τℓ))≥p2\operatorname{ord}(\rho(\tau_\ell))\ge p^2. The tame discriminant formula gives vℓ(ΔFρ)≥∣G∣(1−p−2)=a′v_\ell(\Delta_{F_\rho})\ge|G|(1-p^{-2})=a'.

If ℓ∣Q(ρ)\ell\mid Q(\rho), the inertia image is a nontrivial cyclic subgroup of VV. It has order pp, so vℓ(ΔFρ)=av_\ell(\Delta_{F_\rho})=a. Consequently,

∣ΔFρ∣≥Ra′Q(ρ)a.(23)|\Delta_{F_\rho}|\ge R^{a'}Q(\rho)^a. \tag*{(23)}

If Ra′>XR^{a'}>X, no lift is counted. Otherwise, set

Y=(X/Ra′)1/a≥1.(24)Y=(X/R^{a'})^{1/a}\ge1. \tag*{(24)}

Every counted lift satisfies Q(ρ)≤YQ(\rho)\le Y.

Fix a possible value QQ of Q(ρ)Q(\rho). We first count the tuples (ρ(τℓ))ℓ∉S(\rho(\tau_\ell))_{\ell\notin S}. For ℓ∣R\ell\mid R, the element ψ(τℓ)\psi(\tau_\ell) is fixed. Its inverse image under G→G/VG\to G/V has ∣V∣|V| elements, so there are at most ∣V∣|V| choices for ρ(τℓ)\rho(\tau_\ell). For ℓ∤R\ell\nmid R outside SS, put vℓ=ρ(τℓ)∈Vv_\ell=\rho(\tau_\ell)\in V, using additive notation. The representation ρ\rho kills wild inertia at ℓ\ell. Applying it to the relation in the tame quotient gives

ρ(ϕℓ)ρ(τℓ)ρ(ϕℓ)−1=ρ(τℓ)ℓ.\rho(\phi_\ell)\rho(\tau_\ell)\rho(\phi_\ell)^{-1}=\rho(\tau_\ell)^\ell.

Conjugation on VV factors through π\pi, so this identity says ϕℓ(vℓ)=ℓvℓ\phi_\ell(v_\ell)=\ell v_\ell. Thus vℓ∈Wℓv_\ell\in W_\ell. It has at most nℓn_\ell choices if ℓ∣Q\ell\mid Q, and is zero if ℓ∤QR\ell\nmid QR. The number of possible tuples is therefore at most

∣V∣ω(R)∏ℓ∣Qnℓ.(25)|V|^{\omega(R)}\prod_{\ell\mid Q}n_\ell. \tag*{(25)}

Fix one tuple and one lift ρ1\rho_1 having that tuple, if a lift exists. By Lemma 4.1, every other lift ρ2\rho_2 of ψ\psi determines a unique f∈Z1(Q,V)f\in Z^1(\mathbb{Q},V) satisfying ρ2(σ)=f(σ)ρ1(σ)\rho_2(\sigma)=f(\sigma)\rho_1(\sigma) for all σ∈GQ\sigma\in G_{\mathbb{Q}}. The module action here is the fixed action through π\pi; it does not depend on ψ\psi or ρ1\rho_1.

Equal tuples give f(τℓ)=0f(\tau_\ell)=0 for every ℓ∉S\ell\notin S. Since inertia acts trivially on VV, the restriction of ff to IℓI_\ell is a homomorphism. It kills wild inertia, whose finite quotients have ℓ\ell-power order, and factors through the Zp\mathbb{Z}_p-quotient of tame inertia. Since τℓ\tau_\ell generates that quotient, f∣Iℓ=0f|_{I_\ell}=0.

Let KSK_S be the kernel of GQ→GQ,SG_{\mathbb{Q}}\to G_{\mathbb{Q},S}. The action on VV factors through GQ,SG_{\mathbb{Q},S}. If i∈Iℓi\in I_\ell and σ∈GQ\sigma\in G_{\mathbb{Q}}, the cocycle identity therefore gives

f(σiσ−1)=f(σ)+σf(i)+σif(σ−1)=σf(i)=0.f(\sigma i\sigma^{-1})=f(\sigma)+\sigma f(i)+\sigma i f(\sigma^{-1})=\sigma f(i)=0.

Here the first and last terms cancel because ii acts trivially on VV. The restriction f∣KSf|_{K_S} is a homomorphism. The group KSK_S is the closed normal subgroup generated by the inertia groups outside SS, so continuity implies f∣KS=0f|_{K_S}=0. Thus ff descends to Z1(GQ,S,V)Z^1(G_{\mathbb{Q},S},V).

It follows that each tuple occurs for at most CS:=∣Z1(GQ,S,V)∣C_S:=|Z^1(G_{\mathbb{Q},S},V)| global lifts. This is finite by Lemma 5.1. The estimate counts all lifts of ψ\psi having the tuple, so it also applies after imposing surjectivity, the local conditions, and the discriminant bound. By (25),

Nψ(X)≤CS∣V∣ω(R)∑Q≤YQ squarefree(Q,RS)=1∏ℓ∣Qnℓ.(26)N_\psi(X)\le C_S|V|^{\omega(R)} \sum_{\substack{Q\le Y\\ Q\ \mathrm{squarefree}\\ (Q,RS)=1}} \prod_{\ell\mid Q}n_\ell. \tag*{(26)}

Dropping the coprimality condition with RR and applying Lemma 5.2 gives

Nψ(X)≪∣V∣ω(R)R−a′/aX1/a(log⁡(2Y))bV−1.N_\psi(X)\ll|V|^{\omega(R)}R^{-a'/a}X^{1/a}(\log(2Y))^{b_V-1}.

Since 1≤Y≤X1/a1\le Y\le X^{1/a}, we have log⁡(2Y)≤max⁡{1,a−1}log⁡(X)\log(2Y)\le\max\{1,a^{-1}\}\log(X). Finally, bV≥1b_V\ge1 because V≠0V\ne0. This proves (22). □

Summing the fibers

In this section, we use the uniform estimate from § 5 to prove Theorem 1.2.

Proof of Theorem 1.2. The subgroup VV in (4) satisfies the hypotheses of Proposition 5.3. Indeed, its elements are central in HH, so (vw)p=vpwp=1(vw)^p=v^pw^p=1 for v,w∈Vv,w\in V. It is therefore an abelian subgroup of exponent pp. Cauchy’s theorem shows that it is nontrivial. Every automorphism of HH preserves VV, and HH is normal in GG, so VV is normal in GG. It contains every element of order pp in HH.

Let P\mathcal{P} be the set of surjections ψ:GQ↠G/V\psi:G_{\mathbb{Q}}\twoheadrightarrow G/V whose further quotient is π\pi. For each squarefree positive integer RR, Lemma 3.3, applied with kernel H/VH/V, gives

#{ψ∈P:R(ψ)=R}≤C0A0ω(R),(27)\#\{\psi\in\mathcal{P}:R(\psi)=R\}\le C_0A_0^{\omega(R)}, \tag*{(27)}

where C0C_0 and A0≥1A_0\ge1 are independent of RR. If H=VH=V, then P={π}\mathcal{P}=\{\pi\} and we take A0=1A_0=1. Every ψ\psi is ramified at only finitely many primes, so R(ψ)R(\psi) is finite. There are finitely many integers R≤ZR \le Z, and (27) bounds each corresponding set. Consequently P\mathcal{P} is countable and {ψ∈P:R(ψ)≤Z}\{\psi\in\mathcal{P}: R(\psi) \le Z\} is finite for every ZZ.

Every representation counted by (3) has a unique quotient modulo VV. Hence

NG,π,D(X)=∑ψ∈PNψ(X).(28)N_{G,\pi,D}(X)=\sum_{\psi\in\mathcal{P}}N_\psi(X). \tag*{(28)}

For each fixed XX, this sum has only finitely many nonzero terms, since the left side counts finitely many representations. Proposition 4.2 gives

lim⁡X→∞Nψ(X)X1/a(log⁡X)bV−1=cψ.(29)\lim_{X\to\infty}\frac{N_\psi(X)}{X^{1/a}(\log X)^{b_V-1}}=c_\psi. \tag*{(29)}

Here cψ=0c_\psi=0 for an empty fiber, and cψ>0c_\psi>0 whenever the fiber has a proper solution satisfying the local conditions. The action on VV factors through π\pi, so the integer bVb_V is the same in every fiber.

For X≥eX\ge e, write Fψ(X)=Nψ(X)/(X1/a(log⁡X)bV−1)F_\psi(X)=N_\psi(X)/(X^{1/a}(\log X)^{b_V-1}). Proposition 5.3 and a′/a=1+1/pa'/a=1+1/p give

0≤Fψ(X)≤C1∣V∣ω(R(ψ))R(ψ)1+1/p=:Mψ.(30)0\le F_\psi(X)\le C_1\frac{|V|^{\omega(R(\psi))}}{R(\psi)^{1+1/p}}=:M_\psi. \tag*{(30)}

The constant is independent of XX and ψ\psi: the only additional factor after normalization is (log⁡(X)/log⁡X)bV−1(\log(X)/\log X)^{b_V-1}, which is bounded for X≥eX\ge e.

Grouping by R(ψ)R(\psi) and using (27), we obtain

∑ψ∈PMψ≤C2∑R squarefree(A0∣V∣)ω(R)R1+1/p=C2∏ℓ(1+A0∣V∣ℓ1+1/p)<∞.(31)\sum_{\psi\in\mathcal{P}}M_\psi\le C_2\sum_{R\ \mathrm{squarefree}}\frac{(A_0|V|)^{\omega(R)}}{R^{1+1/p}}=C_2\prod_{\ell}\left(1+\frac{A_0|V|}{\ell^{1+1/p}}\right)<\infty. \tag*{(31)}

To justify the last two assertions, expand the product over a finite set of primes. Its terms are exactly those squarefree integers supported on that set. Increasing these finite sets gives the displayed equality by monotone convergence. The logarithm of each finite product is at most A0∣V∣∑ℓℓ−1−1/pA_0|V|\sum_\ell\ell^{-1-1/p}, which is finite because it is bounded by A0∣V∣∑n≥2n−1−1/pA_0|V|\sum_{n\ge2}n^{-1-1/p}.

Taking limits in (30) gives 0≤cψ≤Mψ0\le c_\psi\le M_\psi. Thus the sum CG,π,D:=∑ψ∈PcψC_{G,\pi,D}:=\sum_{\psi\in\mathcal{P}}c_\psi converges. We verify the passage to the limit in (28). Given ε>0\varepsilon>0, choose ZZ with ∑R(ψ)>ZMψ<ε\sum_{R(\psi)>Z}M_\psi<\varepsilon. For the finite set of ψ\psi satisfying R(ψ)≤ZR(\psi)\le Z, (29) implies

lim⁡X→∞∑R(ψ)≤Z(Fψ(X)−cψ)=0.\lim_{X\to\infty}\sum_{R(\psi)\le Z}\left(F_\psi(X)-c_\psi\right)=0.

For every X≥eX\ge e, the two tails are bounded by ∑R(ψ)>ZFψ(X)<ε\sum_{R(\psi)>Z}F_\psi(X)<\varepsilon and ∑R(ψ)>Zcψ<ε\sum_{R(\psi)>Z}c_\psi<\varepsilon. Therefore

∣∑ψ∈PFψ(X)−CG,π,D∣≤∣∑R(ψ)≤Z(Fψ(X)−cψ)∣+2ε.\left|\sum_{\psi\in\mathcal{P}}F_\psi(X)-C_{G,\pi,D}\right| \le\left|\sum_{R(\psi)\le Z}\left(F_\psi(X)-c_\psi\right)\right|+2\varepsilon.

First let X→∞X\to\infty, and then let ε→0\varepsilon\to0. Together with (28), this proves

lim⁡X→∞NG,π,D(X)X1/a(log⁡X)bV−1=CG,π,D.\lim_{X\to\infty}\frac{N_{G,\pi,D}(X)}{X^{1/a}(\log X)^{b_V-1}}=C_{G,\pi,D}.

Finally, the proper solution assumed in the theorem has a quotient ψ0∈P\psi_0\in\mathcal{P} with a nonempty proper-lift fiber satisfying the local conditions. Proposition 4.2 gives cψ0>0c_{\psi_0}>0. Hence CG,π,D≥cψ0>0C_{G,\pi,D}\ge c_{\psi_0}>0, which proves (6).

Lemma 2.1 gives bV=bπ=Bρ0b_V=b_\pi=B_{\rho_0} for every proper solution ρ0\rho_0. Thus the logarithmic exponent agrees with (2). □

General linear congruence groups

In this section, we give an application to the asymptotic counting of lifts of Galois representations. We retain the notation of Corollary 1.3. For 1≤r≤m1 \le r \le m, put Kr=ker⁡(Gm→Gr)K_r=\ker(G_m \to G_r), and set Kr=1K_r=1 for r≥mr \ge m. Thus K1=HmK_1=H_m and Km=1K_m=1.

Lemma 7.1. For 1≤r<m1 \le r < m, the map

Kr/Kr+1⟶Md(Fp),(Id+prA)Kr+1⟼A mod p(32)\begin{aligned} K_r/K_{r+1} &\longrightarrow M_d(\mathbb{F}_p),\\ (I_d+p^rA)K_{r+1} &\longmapsto A \bmod p \tag*{(32)} \end{aligned}

is an isomorphism of groups, where the target is additive. Conjugation on this quotient factors through G1G_1 and is given by A↦gˉAgˉ−1A \mapsto\bar{g}A\bar{g}^{-1} for g∈Gmg \in G_m. Moreover, for r,s≥1r,s \ge1 one has

[Kr,Ks]⊆Kr+s.(33)[K_r,K_s]\subseteq K_{r+s}. \tag*{(33)}

In particular, HmH_m is a nontrivial pp-group of order pd2(m−1)p^{d^2(m-1)}, and Vm=Km−1V_m=K_{m-1} is elementary abelian and central in HmH_m.

Proof. The proof is an exercise which is left to the reader. □

Lemma 7.2. Suppose that pp is odd. If g∈Hmg \in H_m is nontrivial and g=Id+prAg=I_d+p^rA with 1≤r<m1 \le r<m and A≢0(modp)A\not\equiv0\pmod p, then ord⁡(g)=pm−r\operatorname{ord}(g)=p^{m-r}. Consequently, {h∈Hm:hp=1}=Vm\{h\in H_m:h^p=1\}=V_m.

Proof. Choose a lift A~∈Md(Zp)\widetilde{A}\in M_d(\mathbb{Z}_p), and put g~=Id+prA~\widetilde{g}=I_d+p^r\widetilde{A}. The binomial formula applies because every term is a power of the same matrix. It gives

g~p=Id+pr+1A~+∑j=2p−1(pj)prjA~j+prpA~p.\widetilde{g}^{p}=I_d+p^{r+1}\widetilde{A}+\sum_{j=2}^{p-1}\binom{p}{j}p^{rj}\widetilde{A}^{j}+p^{rp}\widetilde{A}^{p}.

For 2≤j≤p−12\le j\le p-1, the corresponding term is divisible entrywise by p1+rjp^{1+rj}, hence by pr+2p^{r+2}. The last term is also divisible by pr+2p^{r+2}, since rp≥r+2rp\ge r+2 for odd pp and r≥1r\ge1. Therefore

g~p=Id+pr+1A1.(34)\widetilde{g}^{p}=I_d+p^{r+1}A_1. \tag*{(34)}

Here A1≡A~(modp)A_1\equiv\widetilde{A}\pmod p, so A1A_1 has an entry prime to pp.

The same calculation can now be applied to Id+pr+1A1I_d+p^{r+1}A_1. Induction gives matrices At∈Md(Zp)A_t\in M_d(\mathbb{Z}_p) such that g~pt=Id+pr+tAt\widetilde{g}^{p^t}=I_d+p^{r+t}A_t and At≡A~(modp)A_t\equiv\widetilde{A}\pmod p for every t≥0t\ge0. Thus the least valuation of an entry of g~pt−Id\widetilde{g}^{p^t}-I_d is exactly r+tr+t.

Reduction modulo pmp^m gives gpm−r=Idg^{p^{m-r}}=I_d and gpm−r−1≠Idg^{p^{m-r-1}}\ne I_d. The order of gg divides pm−rp^{m-r} and does not divide pm−r−1p^{m-r-1}, so it is pm−rp^{m-r}. Every nonidentity element of HmH_m admits such an expression by choosing rr to be the least valuation of an entry of g−Idg-I_d. It has order pp precisely when r=m−1r=m-1. Adjoining the identity proves the final assertion. □

Proof of Corollary 1.3. Apply Theorem 1.2 to the reduction map Gm→G1G_m\to G_1 and the fixed quotient ρˉ\bar{\rho}. Lemma 7.1 shows that its kernel HmH_m is a nontrivial pp-group. By Lemma 7.2, its elements of order dividing pp form VmV_m. This subgroup centralizes HmH_m by Lemma 7.1. The assumed proper solution satisfies the prescribed local conditions, so the nonemptiness hypothesis of Theorem 1.2 holds.

Under the identification in (32), conjugation on VmV_m is the action A↦ρˉ(σ)Aρˉ(σ)−1A\mapsto\bar{\rho}(\sigma)A\bar{\rho}(\sigma)^{-1}. After the inverse cyclotomic twist, it is exactly the action defining bˉρ\bar{b}_{\rho} in Corollary 1.3. Theorem 1.2 therefore gives the asserted asymptotic with a positive constant and exponent p/(∣Gm∣(p−1))p/(|G_m|(p-1)). The twisted action depends on ρˉ\bar{\rho} and χˉp\bar{\chi}_p, and hence its orbit number is independent of mm. □

An application in dimension two

We use Ramakrishna’s large-image theorem [17 Theorem 1.1] to construct surjective lifts of ρˉ\bar{\rho} to GL⁡2(Zp)\operatorname{GL}_2(\mathbb{Z}_p) when d=2d=2 and p≥7p\ge7. Reduction modulo pmp^m then removes the nonemptiness hypothesis from Corollary 1.3 for any finite local conditions satisfied by these reductions.

Lemma 8.1. Let p≥5p\ge5. The group SL⁡2(Zp)\operatorname{SL}_2(\mathbb{Z}_p) is generated by the matrices

u(t)=(1t01)andv(t)=(10t1)u(t)=\begin{pmatrix}1&t\\0&1\end{pmatrix} \qquad\text{and}\qquad v(t)=\begin{pmatrix}1&0\\t&1\end{pmatrix}

for t∈Zpt\in\mathbb{Z}_p and equals its commutator subgroup.

Proof. For a∈Zp×a\in\mathbb{Z}_p^\times, we find that

w(a)=u(a)v(−a−1)u(a)=(0a−a−10)w(a)=u(a)v(-a^{-1})u(a)=\begin{pmatrix}0&a\\-a^{-1}&0\end{pmatrix}

Thus w(a)w(−1)=diag⁡(a,a−1)w(a)w(-1)=\operatorname{diag}(a,a^{-1}) belongs to the subgroup generated by the elementary matrices.

Let A=(abcd)∈SL⁡2(Zp)A=\begin{pmatrix}a&b\\c&d\end{pmatrix}\in\operatorname{SL}_2(\mathbb{Z}_p). If aa is a unit, then ad−bc=1ad-bc=1 gives

A=v(c/a)(a00a−1)u(b/a).A=v(c/a)\begin{pmatrix}a&0\\0&a^{-1}\end{pmatrix}u(b/a).

If aa is not a unit, cc must be a unit. The upper left entry of w(1)Aw(1)A is then a unit, so the preceding factorization applies to w(1)Aw(1)A. Since w(1)w(1) and its inverse belong to the generated subgroup, so does AA.

Put D=diag⁡(2,2−1)D=\operatorname{diag}(2,2^{-1}) and use the convention [x,y]=xyx−1y−1[x,y]=xyx^{-1}y^{-1}. Direct calculation gives [D,u(t)]=u(3t)[D,u(t)]=u(3t) and [D,v(t)]=v(−3t/4)[D,v(t)]=v(-3t/4). The elements 33 and 44 are units in Zp\mathbb{Z}_p. Consequently, u(t)=[D,u(t/3)]u(t)=[D,u(t/3)] and v(t)=[D,v(−4t/3)]v(t)=[D,v(-4t/3)]. Every generator is a commutator, which proves the assertion. □

In the result below, no parity assumption on det⁡ρˉ\det\bar{\rho} is required.

Theorem 8.2 (Ramakrishna [17 Theorem 1.1]). Let p≥7p\ge7, and let ρˉ:GQ→GL⁡2(Fp)\bar{\rho}:G_{\mathbb{Q}}\to\operatorname{GL}_2(\mathbb{F}_p) be a continuous representation whose image contains SL⁡2(Fp)\operatorname{SL}_2(\mathbb{F}_p). Put A=Zp[[T1,T2,…]]\mathcal{A}=\mathbb{Z}_p[[T_1,T_2,\ldots]]. Then ρˉ\bar{\rho} admits a continuous lift r~:GQ→GL⁡2(A)\widetilde{r}:G_{\mathbb{Q}}\to\operatorname{GL}_2(\mathcal{A}) whose image contains SL⁡2(A)\operatorname{SL}_2(\mathcal{A}).

Proposition 8.3. Suppose that d=2d=2 and p≥7p\ge7. Then ρˉ\bar{\rho} admits a continuous surjective lift ρ∞:GQ↠GL⁡2(Zp)\rho_\infty:G_{\mathbb{Q}}\twoheadrightarrow\operatorname{GL}_2(\mathbb{Z}_p).

Proof. Theorem 8.2, followed by specialization Ti=0T_i=0, gives a continuous lift r:GQ→GL⁡2(Zp)r:G_{\mathbb{Q}}\to\operatorname{GL}_2(\mathbb{Z}_p) whose image contains SL⁡2(Zp)\operatorname{SL}_2(\mathbb{Z}_p).

Let ω\omega be the Teichmüller lift of det⁡ρˉ\det\bar{\rho}, and let χp,∞\chi_{p,\infty} be the pp-adic cyclotomic character. Put η=ωχp,∞p−1\eta=\omega\chi_{p,\infty}^{p-1}. This character lifts det⁡ρˉ\det\bar{\rho} and is surjective. Indeed, χp,∞p−1\chi_{p,\infty}^{p-1} remains surjective onto 1+pZp1+p\mathbb{Z}_p, when restricted to ker⁡ω\ker\omega. The character η(det⁡r)−1\eta(\det r)^{-1} takes values in 1+pZp1+p\mathbb{Z}_p. Since squaring is an automorphism of this group, it has a unique continuous square-root character θ\theta with θ≡1(modp)\theta\equiv1\pmod p. Set ρ∞=θr\rho_\infty=\theta r. Then ρ∞\rho_\infty lifts ρˉ\bar{\rho} and det⁡ρ∞=η\det\rho_\infty=\eta.

Put J=ρ∞(GQ)J=\rho_\infty(G_{\mathbb{Q}}). Since the twisting factors are scalar matrices, they cancel in commutators. Thus, for σ,τ∈GQ\sigma,\tau\in G_{\mathbb{Q}},

[r(σ),r(τ)]=[ρ∞(σ),ρ∞(τ)]∈J.[r(\sigma),r(\tau)]=[\rho_\infty(\sigma),\rho_\infty(\tau)]\in J.

Given A,B∈SL⁡2(Zp)A,B\in\operatorname{SL}_2(\mathbb{Z}_p), choose σ,τ∈GQ\sigma,\tau\in G_{\mathbb{Q}} with r(σ)=Ar(\sigma)=A and r(τ)=Br(\tau)=B. Such choices exist because the image of rr contains SL⁡2(Zp)\operatorname{SL}_2(\mathbb{Z}_p). The displayed equality gives $[A,B]\in J. By Lemma 8.1, every element of SL⁡2(Zp)\operatorname{SL}_2(\mathbb{Z}_p) is a finite product of such commutators. Since det⁡ρ∞\det\rho_{\infty} is surjective, it then follows that ρ∞\rho_{\infty} is surjective. □

Corollary 8.4. Suppose that d=2d=2 and p≥7p\geq7. For every m≥2m\geq2, the asymptotic in Corollary 1.3 holds with no local restrictions. More generally, let ρ∞\rho_{\infty} be a surjective lift as in Proposition 8.3, and put ρm=ρ∞ mod pm\rho_m=\rho_{\infty}\bmod p^m. The same asymptotic holds with finite local conditions D\mathcal{D} whenever ρm∣GQv∈Dv\rho_m|_{G_{\mathbb{Q}_v}}\in\mathcal{D}_v for every v∈Sv\in S.

Proof. Proposition 8.3 provides a surjective lift ρ∞\rho_{\infty}. Each reduction ρm\rho_m is surjective and, under the stated compatibility condition, satisfies D\mathcal{D}. Corollary 1.3 therefore applies. □

Chennai Mathematical Institute, H1, SIPCOT IT Park, Kelambakkam, Siruseri, Tamil Nadu 603103, India

Email address: [email protected]

References

  1. [1]Brandon Alberts, The weak form of Malle’s conjecture and solvable groups, Research in Number Theory 6 (2020), no. 1, Paper No. 10.arxiv.org/abs/1804.11318
  2. [2]———, Statistics of the first Galois cohomology group: A refinement of Malle’s conjecture, Algebra & Number Theory 15 (2021), no. 10, 2513–2569.DOI
  3. [3]Brandon Alberts, Robert J. Lemke Oliver, Jiuya Wang, and Melanie Matchett Wood, Inductive methods for counting number fields, Journal für die reine und angewandte Mathematik 835 (2026), 3–61.arxiv.org/abs/2501.18574
  4. [4]Brandon Alberts and Evan O’Dorney, Harmonic analysis and statistics of the first Galois cohomology group, Research in the Mathematical Sciences 8 (2021), no. 3, Paper No. 50.arxiv.org/abs/2102.11223
  5. [5]———, Corrigendum to “Harmonic analysis and statistics of the first Galois cohomology group”, Research in the Mathematical Sciences 10 (2023), no. 3, Paper No. 31.DOI
  6. [6]———, Harmonic analysis and statistics of the first Galois cohomology group, arXiv:2102.11223v2 [math.NT], 2023, Corrected version incorporating the corrigendum.arxiv.org/abs/2102.11223
  7. [7]Keith Conrad, The different ideal, Expository notes, https://kconrad.math.uconn.edu/blurbs/gradnumthy/different.pdf.
  8. [8]Najmuddin Fakhruddin, Chandrashekhar Khare, and Stefan Patrikis, Relative deformation theory, relative Selmer groups, and lifting irreducible Galois representations, Duke Mathematical Journal 170 (2021), no. 16, 3505–3599.arxiv.org/abs/1904.02374
  9. [9]Chandrashekhar Khare, Michael Larsen, and Ravi Ramakrishna, Constructing semisimple p-adic Galois representations with prescribed properties, American Journal of Mathematics 127 (2005), no. 4, 709–734.
  10. [10]Jürgen Klüners, A counter example to Malle’s conjecture on the asymptotics of discriminants, Comptes Rendus. Mathématique 340 (2005), no. 6, 411–414.DOI
  11. [11]Peter Koymans and Carlo Pagano, On Malle’s conjecture for nilpotent groups, Transactions of the American Mathematical Society, Series B 10 (2023), 310–354.DOI
  12. [12]Gunter Malle, On the distribution of Galois groups, Journal of Number Theory 92 (2002), no. 2, 315–329.DOI
  13. [13]———, On the distribution of Galois groups, II, Experimental Mathematics 13 (2004), no. 2, 129–135.DOI
  14. [14]Jürgen Neukirch, Algebraic number theory, Grundlehren der mathematischen Wissenschaften, vol. 322, Springer-Verlag, Berlin, 1999, Translated from the German by Norbert Schappacher.DOI
  15. [15]Jürgen Neukirch, Alexander Schmidt, and Kay Wingberg, Cohomology of number fields, 2 ed., Grundlehren der mathematischen Wissenschaften, vol. 323, Springer-Verlag, Berlin, 2008.DOI
  16. [16]Ravi Ramakrishna, Deforming Galois representations and the conjectures of Serre and Fontaine–Mazur, Annals of Mathematics. Second Series 156 (2002), no. 1, 115–154.DOI
  17. [17]———, Constructing Galois representations with very large image, Canadian Journal of Mathematics 60 (2008), no. 1, 208–221.DOI
  18. [18]Anwesh Ray, Lower bounds for the number of number fields with Galois group GL₂(Fℓ), Forum Mathematicum 37 (2025), no. 5, 1589–1596.DOI

Paper details

vlnc:W_nxbgj7sp
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

ChatGPT 6 Astra

View statement
Looking for related papers…
Contents