Abstract:
Let ℓ=3, k=Q(√−3) and K=k(3√a), where a is a natural number such that a2≡1(mod9).
Under the assumption that there are exactly three places not over ℓ that ramify in the extension K∞/k∞, where k∞ and K∞ are cyclotomic Z3-extensions of the fields k and K, respectively, we study 3-class field towers for intermediate fields Kn of the extension K∞/K.
It is shown that for each Kn the 3-class field tower of the field Kn terminates already at the first step, which means that the Galois group of the extension Hℓ(Kn)/Kn, where Hℓ(Kn) is the maximal unramified ℓ-extension of the field Kn, is Abelian.
Bibliography: 7 titles.
Keywords:
Iwasawa theory, Tate module, extensions with bounded ramification, Riemann–Hurwitz formula, class field tower.
Let L be an algebraic number field, ℓ be an odd prime number, and Hℓ(L) be the ℓ-Hilbert class field of L, that is, the maximal Abelian unramified ℓ-extension of the field L. It is known that the Galois group G(Hℓ(L)/L) is canonically isomorphic to the ℓ-component Clℓ(L) of the class group of L.
Set F0=L, F1=Hℓ(L), and for n>1 let Fn=Hℓ(Fn−1). The sequence of fields Fi is called the ℓ-class field tower of the fieldL, and the field Hℓ(L)=⋃nFn is the maximal unramified ℓ-extension of L. According to the famous Golod–Shafarevich theorem, there exist fields L with an infinite ℓ-class field tower, that is, fields L for which Hℓ(L) has an infinite degree over L. Except for this fundamental fact, surprisingly little is known about ℓ-class field towers. We consider the problem of finiteness of the ℓ-class field tower in the following particular case.
Let ℓ=3, k=Q(√−3), and let K=k(3√a) be a cyclic cubic extension of the field k, where a∈Z and a is not a cube in Z. Thus, K is a Galois extension of the field Q with Galois group S3. In addition, we assume that K is Abelian over ℓ, that is, the completion Kv of the field K relative to any place v over ℓ=3 is an Abelian extension of the field Q3.
Let kn=Q(ζn), where ζn is a primitive ℓn+1th root of unity (thus, k=k0) and k∞=⋃nkn is a cyclotomic Zℓ-extension of k. Also set Kn=K⋅kn, and let K∞=⋃nKn be a cyclotomic Zℓ-extension of the field K.
The main result of this work consists in the proof of the following theorem.
Theorem. Let ℓ=3, let K have the form indicated above, and suppose that there are exactly three places not over ℓ that ramify the extension K∞/k∞. Then Hℓ(Kn)=Hℓ(Kn) for every n, which means that the 3-class field tower of the field Kn terminates already at the first step. Moreover, G(Hℓ(Kn)/Kn)≅Clℓ(Kn).
In § 2 we introduce the necessary definitions and recall the notation, which is basically the same as in [1]–[4].
In § 3, using an analogue of the Riemann–Hurwitz formula we study the case where the field K∞ has a finite Tate module (Iwasawa module). We show that for every n and any unramified ℓ-extension L/Kn the Galois group G(L/Kn) acts identically on the class group Clℓ(L) (Proposition 1). The definition of the Tate module Tℓ(K∞) and the related modules ¯Tℓ(K∞) and Rℓ(K∞) mentioned below is given in § 2.
In § 4, again with the use of the analogue of the Riemann–Hurwitz formula we consider the case where the module Tℓ(K∞) is infinite. It turns out again that in this case, for every n and any unramified Galois ℓ-extension L/Kn the Galois group G(L/Kn) acts identically on the class group Clℓ(L) (Proposition 5).
In § 5 we present a summary of the known results concerning fields K such that there are precisely three places ramifying the extension K∞/k∞ and concerning the class groups Clℓ(Kn) for such fields.
Note that in the case ℓ=3 the field K is in a certain sense similar to the field of rational functions on an elliptic curve. Namely, any unramified ℓ-extension L of the field Kn, where Kn is an intermediate subfield of the cyclotomic Zℓ-extension K∞/K, is Abelian, similarly to the case of an elliptic curve. However, the proof of this fact is based on quite different arguments than the ones employed in the geometric situation.
§ 2. Notation and definitions
We follow basically the notation of [1]–[4]. Let ℓ be a regular odd prime number and ζn be a primitive ℓn+1th root of unity. Set k=Q(ζ0) and k∞=⋃∞n=1kn, where kn=k(ζn). Let K=k(ℓ√a), where a is a natural number such that a place v of the field k over ℓ splits completely in the extension K/k. This means that aℓ−1≡1(modℓ2). Moreover, we assume that there are exactly three places that ramify in the extension K∞/k∞. Accordingly, we assume that either a=pr11pr22pr33, or a=prqs. In the first case p1,p2 and p3 are prime numbers remaining prime in the extension k∞/Q; in the second case p splits in the unique quadratic subfield F of the field k into the product (p)=p1p2 and each of the divisors pi remains prime in the extension k∞/F, and q=(q) remains prime in the extension k∞/Q. Accordingly, following the terminology of [1], we refer to these as extensions of type 2.1 and extensions of type 2.2. There also exist extensions of type 2.4. These are fields K of the form K=k(3√p), where p≡8,17(mod27).
We let G denote the Galois group G(K/Q), H denote the group G(K/k) and Δ denote the group G(k/Q). Thus, G is a semidirect product of H and Δ, and the group Δ acts on H in accordance with the Teichmüller character ω:Δ→(Z/ℓZ)×.
We let Fℓ(i) denote the group Z/ℓZ on which Δ acts in accordance with the rule ωi. The index i is defined modulo ℓ−1. Let A be a finite G-module which is cyclic as an H-module and satisfies NH(A)=0, where NH=∑h∈Hh is the norm operator. Let A=A0⊇A1⊇⋯⊇An=0 be the lower central series of the H-module A. If A0/A1≅Fℓ(i) and An−1≅Fℓ(j), then we say that A starts with Fℓ(i) and terminates with Fℓ(j). In this case, by Lemma 3.2 in [1] we have Ak/Ak+1≅Fℓ(i+k) for every k<n. If |A|=ℓr, then r≡j−i+1(modℓ−1).
In this work we use an analogue of the Riemann–Hurwitz formula, which was derived by this author in [5] (a refined proof can be found in [6]). This formula, which we present in § 3, relates some linear combinations of the Iwasawa λ-invariants of certain Galois modules associated with a finite ℓ-extension of the fields L′/L. The definition of these Galois modules is given below. Note that all fields considered in this work are ℓ-extensions of the field k, and therefore all modules under study have zero Iwasawa μ-invariants.
Let L be an arbitrary algebraic number field and L∞ be the cyclotomic Zℓ-extension of L. Let ¯N denote the maximal Abelian unramified ℓ-extension of the field L∞ and N be the maximal subfield of ¯N such that all places in S (the set of all places over ℓ) split completely in N/L∞. We let Tℓ(L∞) and ¯Tℓ(L∞) denote the Galois groups of the extensions N/L∞ and ¯N/L∞, respectively. These groups are compact Noetherian modules under the action of the group Γ=G(L∞/L), where Γ≅Zℓ. We fix some topological generator γ0 in Γ. Accordingly, these modules are acted upon by the Iwasawa algebra Λ=Zℓ[[Γ]]=lim←Zℓ[Γ/Γn], where Γn is the unique subgroup of index ℓn in the group Γ. We let Rℓ(L∞) denote the kernel of the natural map ¯Tℓ(L∞)→Tℓ(L∞), which means that Rℓ(L∞) is the subgroup of the group ¯Tℓ(L∞) generated by the decomposition subgroups of all places over S.
Let M be the maximal Abelian ℓ-extension of the field L∞ that is unramified outside S, and let X(L∞)=G(M/L∞). Then X(L∞) is a Λ-module, whose submodule of Λ-torsion is denoted by TorsX(L∞). The natural maps X(L∞)→¯Tℓ(L∞) and X(L∞)→Tℓ(L∞) induce e maps TorsX(L∞)→¯Tℓ(L∞) and TorsX(L∞)→Tℓ(L∞). Their images are denoted by ¯T′ℓ(L∞) and T′ℓ(L∞), respectively. Set Tℓ″, and let \lambda'(L_\infty) and \lambda ''(L_\infty) denote the \lambda-invariants of the modules T'_\ell(L_\infty) and T''_\ell(L_\infty), respectively.
We set R'(L_\infty)=R(L_\infty)\cap\overline T_\ell'(L_\infty) and R''(L_\infty)=R(L_\infty)/R'_\ell(L_\infty). Let r(L_\infty), r'(L_\infty) and r''(L_\infty) denote the \lambda-invariants of the modules R(L_\infty), R'(L_\infty) and R''(L_\infty), respectively.
In addition, we need one more invariant, denoted by d(L_\infty), which is a \lambda-invariant of the module D(L_\infty). This module is defined in the case where the field L_\infty is Abelian over \ell and k\subset L. The definition of D(L_\infty) can be found in the work [1], § 6, which also contains the definition of the modules V(L_\infty), V^+(L_\infty) and V^-(L_\infty) involved in the definition of D(L_\infty), and a detailed description of the structure of this module and its properties can be found in [6]. Here we only note that D(L_\infty) is defined as V(L_\infty)/(V^+(L_\infty)\oplus V^-(L_\infty)), where V(L_\infty), V^+(L_\infty) and V^-(L_\infty) are free \Lambda-modules. It should also be noted that we often use without special mention the additive notation for the operation of multiplication since the operation of addition is not used here.
§ 3. An analogue of the Riemann–Hurwitz formula
This formula was first obtained by this author in [5]. A refined proof can be found in [6]. Let L'/L be a finite \ell-extension of algebraic number fields (here \ell is an arbitrary prime number). Let L'_\infty and L_\infty be cyclotomic \mathbb Z_\ell-extensions of the fields L' and L, respectively, where the fields L and L' satisfy some additional constraints, namely, L and L' are Abelian over \ell. This means that for any place v over \ell the completion L_v of the field L (or the completion L'_v of L') is an Abelian extension of the field \mathbb Q_\ell. We also assume that the field L contains a primitive \ellth root \zeta_0 of unity, which means that L\supseteq k and all Galois modules considered below have zero Iwasawa \mu-invariants. The last condition holds for k_\infty when \ell=3, since 3 is a regular prime number, and therefore it holds for any finite \ell-extension of the field k_\infty. In particular, it holds for L_\infty and L'_\infty.
The analogue of the Riemann–Hurwitz formula provides a relation between the Iwasawa \lambda-invariants of the Galois modules defined in § 2 for the fields L_\infty' and L_\infty, where L'/L is a finite \ell-extension of algebraic number fields and the fields L' and L satisfy some additional conditions which were indicated above.
The analogue of the Riemann–Hurwitz formula relates the invariant 2g(L_\infty) to an invariant 2g(L'_\infty) defined similarly for the finite \ell-extension L'/L. Namely, under the above assumptions about the fields L_\infty and L'_\infty the following relation, which is an analogue of the Riemann–Hurwitz formula, holds:
where v ranges over all the places of the field L_\infty' that are not over \ell and e_v means the ramification index of the place v in the extension L'_\infty/L_\infty.
Remark. It is obvious that g(k_\infty)=0. In the case \ell=3 an application of the above formula to the extension K_\infty/k_\infty, where K_\infty/k_\infty is a cyclic extension of degree \ell such that there are precisely three places not over \ell ramifying in K_\infty/k_\infty, gives g(K_\infty)=1.
As explained in [4], Proposition 3.1, a finite unramified \ell-extension L_\infty/K_\infty for {\ell=3} obeys the relation g(L_\infty)=1 and one of the following two options holds:
(A) d(L_\infty)=2(\ell-1) and \lambda'(L_\infty)=\lambda''(L_\infty)=r''(L_\infty)=0;
(B) d(L_\infty)=0, \lambda'(L_\infty)=\ell-1 and \lambda''(L_\infty)=r''(L_\infty)=0.
If L_\infty/K_\infty is a finite unramified extension, then either both fields pertain to type (A), or both fields pertain to type (B). In this section we consider Case (A) in detail.
If L/K_n is a finite unramified Galois extension, then the Galois group G(L_\infty/K_n) is isomorphic to the direct product G(L/K_n)\times G(K_\infty/K_n) and the group G(L/K_n) acts on the \Lambda_n-modules D(L_\infty), T_\ell(L_\infty) and the other \Lambda-modules defined in § 2.
Proposition 1. The group G(L/K_n) acts identically on the module D(L_\infty).
Proof. In Case (A) we have d(L_\infty)=d(K_\infty)=2(\ell-1)=4. The inclusion K_\infty\hookrightarrow L_\infty induces a natural mapping i\colon D(K_\infty)\to D(L_\infty), and the sequence of the norm mappings N_m\colon L_m^\times\to K_m^\times, where L_m=K_m\cdot L, induces a mapping N\colon D(L_\infty)\to D(K_\infty). It is obvious that N\circ i=[L:K_n]. Since D(K_\infty) and D(L_\infty) are free \mathbb Z_\ell-modules of the same rank, this means that i maps D(K_\infty) isomorphically onto a submodule of finite index in D(L_\infty). The mappings i and N are G(L/K_n)-homomorphisms and the group G(L/K_n) acts identically on D(K_\infty). Hence the group G(L/K_n) also acts identically on D(L_\infty).
The proof of the proposition is complete.
Proposition 2. In Case ( A) the module \overline T_\ell(L_\infty) is a finite group.
Proof. It is sufficient to verify that the groups T_\ell(L_\infty) and R_\ell(L_\infty) are finite. The finiteness of the group T_\ell(L_\infty) follows from the fact that in Case (A) we have \lambda'(L_\infty)=\lambda''(L_\infty)=0.
Since in Case (A) we have r''(L_\infty)=0, to prove that the group R_\ell(L_\infty) is finite it is sufficient to establish the equality r'(L_\infty)=0. Note that all the places of the field L over \ell are purely ramified in the extension L_\infty/L, and therefore the group \Gamma_n=G(L_\infty/L) acts identically on R_\ell(L_\infty) and, by implication, on R'(L_\infty). Since there exists an epimorphism \operatorname{Tors}X(L_\infty)\to R'(L_\infty), it is sufficient to verify that
where \gamma_n=\gamma_0^{\ell^n} is a topological generator of the group \Gamma_n. If this rank were positive, then the field L would have ‘superfluous’ \Gamma-extensions, which would mean that it disobeys the Leopoldt conjecture, contrary to the finiteness of the module T_\ell(L_\infty) (see [1], Proposition 5.1, or [7], Theorem 4.1).
The proof of the proposition is complete.
If L is an extension of the field K_n such that L\cap K_\infty=K_n, then all Galois modules connected with the extension L_\infty/L are \Gamma_n-modules or \Lambda_n-modules, where \Lambda_n=\mathbb Z_\ell[[\Gamma_n]].
Let \mathscr F(V(L_\infty)/V^+(L_\infty)) be a minimal free \Lambda_n-module containing the torsion-free \Lambda_n-module V(L_\infty)/V^+(L_\infty) as a submodule of finite index. Then there exists a natural embedding of the free \Lambda_n-module V^-(L_\infty) into the free module \mathscr F(V(L_\infty)/V^+(L_\infty)) and the quotient module D'(L_\infty)=\mathscr F(V(L_\infty)/V^+(L_\infty))/V^-(L_\infty) has no nonzero finite submodules, which means that it is a free \mathbb Z_\ell-module containing D(L_\infty) as a submodule of finite index. Consequently, the Galois group G(L/K_n) also acts identically on D'(L_\infty). Set E'(L_\infty)=D'(L_\infty)/D(L_\infty). Then the group G(L/K_n) acts identically on E'(L_\infty).
In a similar way we set D''(L_\infty)=\mathscr F(V(L_\infty)/V^-(L_\infty)) and E''(L_\infty)=D''(L_\infty)/D(L_\infty). Again, we show that the group G(L/K_n) acts identically on E''(L_\infty). The groups E'(L_\infty) and E''(L_\infty) are isomorphic as Abelian groups (the isomorphism is induced by the skew automorphism \psi; [1], Theorem 6.1). In addition, D'(L_\infty)\cap D''(L_\infty)= D(L_\infty) ([6], Proposition 1.7), and therefore the group E'(L_\infty)\oplus E''(L_\infty) is embedded into (D(L_\infty)\otimes \mathbb Q_\ell)/D(L_\infty) and the minimum number of generators of the group E'(L_\infty) does not exceed \ell-1=2.
Proposition 3. The group G(L/K_n) acts identically on the group \operatorname{Cl}_\ell(L).
Proof. By Theorem 3.1 in [2] the Galois module E'(L_\infty) contains a submodule E'_2(L_\infty) naturally isomorphic to the module \overline T_\ell(L_\infty) (note that the proof of that theorem is based only of the fact that \overline T_\ell(L_\infty) is finite, so its claim is not only true for K_\infty, but also for the field L_\infty). Then it follows from Proposition 2 that G(L/K_n) acts identically on \overline T_\ell(L_\infty).
Since the extension L_\infty/L is purely ramified in the divisors of \ell, the natural mapping \overline T_\ell(L_\infty)\to \operatorname{Cl}_\ell(L) is epimorphic. Consequently, G(L/K_n) also acts identically on \operatorname{Cl}_\ell(L).
The proof of is complete.
§ 4. The Riemann–Hurwitz formula (Case (B))
Assume that the field L satisfies the conditions of Case (B).
Proposition 4. In Case (B) \overline T_\ell(L_\infty)=T_\ell(L_\infty) and T_\ell(L_\infty)\cong\mathbb Z_3^2.
Proof. Let H :=G(K_\infty/k_\infty). Then T_\ell(K_\infty) is a cyclic H-module ([1], Theorem 4.1) annihilated by the operator N_H. With regard to the condition \lambda'(L_\infty)=2 this means that T_\ell(K_\infty)\cong \mathbb Z_\ell^2 as a \mathbb Z_\ell-module. Suppose that \gamma_0 acts on roots of unity \zeta_n by the rule \gamma_0(\zeta_n)=\zeta_n^{\varkappa(\gamma_0)}, \varkappa(\gamma_0)\in\mathbb Z_\ell. Then by [1], Theorem 5.1, \gamma_0 acts on T_\ell(K_\infty) by multiplication by \sqrt{\varkappa(\gamma_0)}. In the case \ell=3 we assume that \gamma_0 is defined by the condition \varkappa(\gamma_0)=4, where \sqrt{\varkappa(\gamma_0)}=-2.
Then, as in the proof of Proposition 1, the embedding of fields K_\infty\hookrightarrow L_\infty induces the embedding i\colon T_\ell(K_\infty)\hookrightarrow T_\ell(L_\infty) with a finite cokernel. Let T_\ell^0(L_\infty) be the maximal \mathbb Z_\ell-free quotient module of the module T_\ell(L_\infty). Then T_\ell^0(L_\infty) is acted upon by the topological generator \gamma_n=\gamma_0^{\ell^n} of the group \Gamma_n and \gamma_n multiplies T_\ell^0(L_\infty) by -2^{\ell^n}. The group G(L/K_n) acts identically on T_\ell^0(L_\infty).
As in the proof of Proposition 2, we use Theorem 4.1 in [7] to show that any field L_n obeys the Leopoldt conjecture and thus r'(L_\infty)=0 and R(L_\infty) is a finite group. However, by [4], Proposition 3.3, the module \overline T_\ell(L_\infty) has no nontrivial finite submodules. This means that \overline T_\ell(L_\infty)=T_\ell(L_\infty)=T_\ell^0(L_\infty)\cong \mathbb Z_3^2.
The proof of the proposition is complete.
Proposition 5. In Case (B) the group G(L/K_n) acts identically on the group \operatorname{Cl}_\ell(L).
Proof. There is a natural mapping \varphi\colon \overline T_\ell(L_\infty)\!\to\! \operatorname{Cl}_\ell(L), which is a G(L/K_n)-homomorphism. Since the extension L_\infty/L is purely ramified at the places lying over \ell, \varphi is epimorphic. By Proposition 4 the group G(L/K_n) acts identically on \overline T_\ell(L_\infty). Consequently, it also acts identically on \operatorname{Cl}_\ell(L) as well.
The proof of the proposition is complete.
To use Propositions 3 and 5 in the proof of our theorem we need the following simple statement concerning finite \ell-groups.
Proposition 6. Let G be a finite \ell-group such that for each normal subgroup H of G the quotient group G/H acts identically on the G/H-module H^{\mathrm{ab}}=H/[H,H]. Then G is an Abelian group.
Proof. Assume that G is not Abelian, that is, [G,G]\!\neq\! 1. Then {[G,G]/[G,[G,G]]\!\neq\! 1}. Let \varphi be an epimorphism of the group [G,G]/[G,[G,G]] onto A\cong \mathbb F_\ell. Then the group extension
where B=G/\ker \varphi. Thus, [B,B]=A and there exist elements x,y\in B such that [x,y]\neq 1. Let B_1 be the subgroup of B generated by x^\ell and y. Then B_1 is an Abelian subgroup and x acts nontrivially on B_1. Set F=B/B_1. Then there exists a group extension
such that F acts nontrivially on the Abelian kernel B_1.
The group F is a quotient group of G with some kernel H, hence there exists a commutative diagram of group extensions
Consequently, F acts nontrivially on H/[H,H], contrary to the hypothesis of the proposition.
The proof of the proposition is complete.
Proof of the theorem. Assume that there exists a finite unramified Galois \ell-extension M/K_n with non-Abelian Galois group G. Suppose that the subfield L\subset M is a Galois extension, that is, the group H=G(M/L) is a normal subgroup of G. Set L_1=M^{[H,H]}. Then L_1 is an Abelian unramified \ell-extension of the field L and, consequently, there exists a canonical epimorphism \operatorname{Cl}_\ell(L)\to G(L_1/L)=H/[H,H].
The Galois group G(L/K_n) acts identically on the group \operatorname{Cl}_\ell(L) by Propositions 3 and 5, and therefore it also acts identically on H/[H,H]. Thus, all the hypotheses of Proposition 6 are satisfied, and so the group G is Abelian, which means that there can exist only Abelian unramified 3-extensions over K_n. In turn, this means that the claim of the theorem holds for K_n.
§ 5. The form of the field K and the structure of the class groups of the fields K_n
For the reader’s convenience we give a brief survey of the results obtained in [1]–[4] and concerning the field K and the class groups \operatorname{Cl}_\ell(K_n).
If \ell=3 and K is a cubic extension of the field k such that K is a Galois extension of \mathbb Q with Galois group S_3, K is Abelian over \ell, and there are precisely three places not over \ell that ramify in the extension K_\infty/k_\infty, then K pertains to one of the following three types, which, as in [1], are referred to as Cases 2.1, 2.2, and 2.4 (see [1], Propositions 2.1, 2.2 and 2.4). The work [1] also contains one more case, namely, Case 2.3, but it cannot occur for \ell=3.
Case 2.1. K=k(\sqrt[3]{a}), a\in\mathbb Z, a^{\ell-1}\equiv 1\pmod{\ell^2} and a=p_1^{r_1}p_2^{r_2}p_3^{r_3}, where p_1, p_2 and p_3 are distinct prime numbers not equal to \ell. Here r_1r_2r_3\not\equiv 0\pmod\ell and the principal divisors (p_i) remain prime in K_\infty for i=1,2,3. The last condition means that the p_i are primitive roots modulo \ell^2.
In Case 2.1 the module \overline T_\ell(K_\infty) starts with \mathbb F_\ell(1), and there can be three subcases.
Subcase 2.1.a. |T_\ell(K_\infty)|=3, |\overline T_\ell(K_\infty)|=27, \operatorname{Cl}_\ell(K)\cong (\mathbb Z/3\mathbb Z)^2 and \operatorname{Cl}_\ell(K_n)\cong \mathbb Z/3\mathbb Z\oplus \mathbb Z/9\mathbb Z for n>0.
Subcase 2.1.b. |T_\ell(K_\infty)|=3^r, |\overline T(K_\infty)|=3^{r+2} for some odd r>1, and \operatorname{Cl}_\ell(K_n)\cong (\mathbb Z/3^{n+1})^2 for n<n_0=(r-1)/2 and \operatorname{Cl}_\ell(K_n)\cong \mathbb Z/3^{n_0+1}\mathbb Z\oplus \mathbb Z/3^{n_0+2}\mathbb Z for n\geqslant n_0.
Subcase 2.1.c. T_\ell(K_\infty)\cong \mathbb Z_3^2 and \operatorname{Cl}_\ell (K_n)\cong (\mathbb Z/3^{n+1}\mathbb Z)^2 for n\geqslant 0.
For any pair of distinct prime numbers p_1, p_2 there exist infinitely many primes p_3 such that K pertains to Subcase 2.1.a, and infinitely many primes p_3 such that K pertains either to Subcase 2.1.b, or to Subcase 2.1.c, though there is no field K of the last type that can definitely be classified as pertaining to Subcase 2.1.b or to Subcase 2.1.c (see [2], Theorem 4.1 and Proposition 4.3, and [3], Theorem 6.2).
Case 2.2. In this case K=k(\sqrt[3]{a}), a^{\ell-1}\equiv 1\pmod{\ell^2} and a=p^r q^s, rs\not\equiv 0\ (\operatorname{mod}\ell), where p and q are distinct prime numbers not equal to \ell, and p splits into a product of two divisors \mathfrak p_1 and \mathfrak p_2 in the unique quadratic subfield F of k (note that for \ell=3 the fields F and k coincide) such that each divisor remains prime in the extension K_\infty/F.
The divisor (q) remains prime in the extension K_\infty/\mathbb Q. In this case \overline T_\ell(K_\infty) starts with \mathbb F_\ell(0) and the following three subcases are possible.
Subcase 2.2.a. T(K_\infty)=0 and \overline T_\ell(K_\infty)\cong (\mathbb Z/3\mathbb Z)^2. Then \operatorname{Cl}_\ell(K_n)\cong (\mathbb Z/3\mathbb Z)^2 for every n.
Subcase 2.2.b. T_\ell(K_\infty)\cong (\mathbb Z/3^{n_0}\mathbb Z)^2 for some index n_0 and |R_\ell(K_\infty)|=9. In this subcase \operatorname{Cl}_\ell(K_n)\cong (\mathbb Z/\ell^{n+1}\mathbb Z)^2 for n\leqslant n_0 and \operatorname{Cl}_\ell(K_n)\cong (\mathbb Z/\ell^{n_0}\mathbb Z)^2 for {n>n_0}.
Subcase 2.2.c. T_\ell(K_\infty)\cong \mathbb Z_3^2. In this case we have \operatorname{Cl}_\ell(K_n)\cong (\mathbb Z/3^{n+1}\mathbb Z)^2 for every n.
Again, it is known that there exist infinitely many fields K pertaining to Subcase 2.2.a and infinitely many fields K pertaining to Subcases 2.2.b and 2.2.c, but there is no field of the last type that can definitely be classified as pertaining to Subcase 2.2.b or to Subcase 2.2.c (see [4], Propositions 4.1–4.4). Note also that in Case 2.2 there exist fields K pertaining to none of the three subcases distinguished above.
Case 2.4. In this case we have K=k(\sqrt[3]{p}), where the prime number p satisfies the congruence p\equiv 8,17\pmod{27}. This case remains unexplored so far, and there is nothing definite about it yet.
Bibliography
1.
L. V. Kuz'min, “Arithmetic of certain \ell-extensions ramified at three places”, Proc. Steklov Inst. Math., 307 (2019), 65–84
2.
L. V. Kuz'min, “Arithmetic of certain \ell-extensions ramified at three places. II”, Izv. Math., 85:5 (2021), 953–971
3.
L. V. Kuz'min, “Arithmetic of certain \ell-extensions ramified at three places. III”, Izv. Math., 86:6 (2022), 1143–1161
4.
L. V. Kuz'min, “Arithmetic of certain \ell-extensions ramified at three places. IV”, Izv. Math., 88:2 (2024), 270–283
5.
L. V. Kuz'min, “An analog of the Riemann–Hurwitz formula for one type of l-extension of algebraic number fields”, Math. USSR-Izv., 36:2 (1991), 325–347
6.
L. V. Kuz'min, “New explicit formulas for the norm residue symbol, and their applications”, Math. USSR-Izv., 37:3 (1991), 555–586
7.
L. V. Kuz'min, “The Tate module for algebraic number fields”, Math. USSR-Izv., 6:2 (1972), 263–321
Citation:
L. V. Kuz'min, “On a family of algebraic number fields with finite 3-class field tower”, Sb. Math., 215:7 (2024), 911–919