---
title: S-split p-Hilbert Class Field Tower
url: https://www.emergentmind.com/topics/s-split-p-hilbert-class-field-tower
type: topic
---

# S-split p-Hilbert Class Field Tower

Searching arXiv for the primary paper and closely related class field tower references.
arxiv_search.query{"search_query":"all:\"S-split p-Hilbert class field tower\" OR ti:\"On S-Split p-Hilbert Class Field Towers with Prescribed Galois Groups\"","start":0,"max_results":10}
Searching for Ozaki-style prescribed \(p\)-group realizations and related class field tower work.
arxiv_search.query{"search_query":"ti:\"On Ozaki's theorem realizing prescribed p-groups as p-class tower groups\" OR au:Ozaki OR all:\"prescribed p-groups\" \"class tower groups\"","start":0,"max_results":10}
Searching for stability and Minkowski-unit methods in \(p\)-class towers.
arxiv_search.query{"search_query":"Hajir Maire Ramakrishna class field tower Minkowski units stability","start":0,"max_results":10}
An \(S\)-split \(p\)-Hilbert class field tower is the variant of the \(p\)-class field tower in which the maximal pro-\(p\) extension is required not only to be unramified everywhere, but also to be totally decomposed at a prescribed finite set \(S\) of primes. For a number field \(K\) and a prime \(p\), this construction refines the ordinary Hilbert \(p\)-class field tower by imposing local decomposition constraints inside a global unramified pro-\(p\) extension. In recent work, Maire–Sankara proved that these additional splitting conditions do not obstruct the realization of arbitrary finite \(p\)-groups: given a finite \(p\)-group \(G\), a number field \(K\) with trivial \(p\)-class group, and a finite set of primes \(S\) of \(K\), there exists a finite extension \(F/K\) such that the \(S\)-split \(p\)-Hilbert class field tower of \(F\) has Galois group \(G\) [2508.07946].

## 1. Definition and basic formalism

For a number field \(K\) and a prime \(p\), the ordinary \(p\)-Hilbert class field tower is built by iterating maximal unramified abelian \(p\)-extensions. Writing \(K(0)=K\) and letting \(K(i+1)\) be the \(p\)-Hilbert class field of \(K(i)\), one obtains the maximal unramified pro-\(p\) extension
\[
L_p(K),
\]
whose Galois group is
\[
G_K:=\mathrm{Gal}(L_p(K)/K).
\]
Class field theory identifies the first step through the Artin map:
\[
\mathrm{Gal}(K(i+1)/K(i)) \simeq \mathrm{Cl}_p(K(i)),
\]
and the tower is trivial exactly when
\[
L_p(K)=K \iff \mathrm{Cl}_p(K)=1
\]
[2508.07946].

Given a finite set \(S\) of primes of \(K\), the \(S\)-split \(p\)-Hilbert class field tower is the maximal pro-\(p\) extension
\[
L_p^S(K)/K
\]
that is unramified everywhere and totally decomposed at all primes in \(S\). Thus the adjective “\(S\)-split” refers to complete splitting, not to ramification or partial decomposition. Equivalently, \(L_p^S(K)\) is the largest normal subextension of \(L_p(K)/K\) fixed by the decomposition groups above the primes in \(S\). Its Galois group
\[
G^S:=\mathrm{Gal}(L_p^S(K)/K)
\]
is therefore a quotient of the ordinary tower group \(G_K\) [2508.07946].

This distinction is conceptually important. The ordinary tower is defined purely by the absence of ramification, whereas the \(S\)-split tower combines global unramifiedness with local splitting constraints. The same framework admits a more general \(T\)-ramified, \(S\)-split tower \(L_T^S(K)\), but the principal developments discussed here concern the unramified case \(T=\varnothing\).

## 2. Structural relation to ordinary \(p\)-class towers

The \(S\)-split tower sits naturally inside the classical theory of Hilbert \(p\)-class towers. In the ordinary setting, the maximal unramified pro-\(p\) extension controls the \(p\)-class tower group, its finite quotients, and tower length. In the \(S\)-split setting, one passes from \(G_K\) to a quotient obtained by forcing decomposition groups at the primes of \(S\) to act trivially. This produces a refined inverse Galois problem: not merely which finite \(p\)-groups occur as \(p\)-tower groups, but which occur under prescribed local splitting conditions [2508.07946].

The formal similarity with ordinary tower theory remains strong. For example, the first layer of an ordinary tower is governed by \(\mathrm{Cl}_p(K)\), and the maximal unramified pro-\(p\) extension can be studied through group-theoretic invariants of \(G_K\). In work on ordinary \(p\)-class towers, abelianization data such as the IPAD (index-\(p\) abelianization data) and iterated IPADs constrain the second Hilbert \(p\)-class field and, in many cases, the full tower group [1502.03388]. This suggests that analogous invariants may be informative in the \(S\)-split setting whenever the local splitting condition can be encoded through suitable quotients of the unramified tower group, although the cited IPAD work is not formulated explicitly in \(S\)-split language.

A related point of terminology is that “\(S\)-split” should not be conflated with “unramified outside \(S\).” Several class field tower constructions in the literature are phrased in terms of restricted ramification sets, but the \(S\)-split tower imposes the opposite type of local condition at \(S\): complete decomposition rather than allowed ramification.

## 3. Prescribed Galois groups and realization theorems

The central existence theorem for \(S\)-split towers states that prescribed finite \(p\)-groups remain realizable despite the added splitting constraint. In the form given in Theorem A / Theorem 2.1, one starts from a number field \(K\) with finite \(p\)-tower
\[
L_p(K)/K \text{ finite}, \qquad G:=\mathrm{Gal}(L_p(K)/K).
\]
Writing
\[
h^i=\dim_{\mathbf F_p}H^i(G,\mathbf F_p)
\]
and letting \(e_G\) denote the exponent of \(G\), the theorem asserts that if
\[
r_{K,1}+r_{K,2} > h^1 + h^2,
\]
or equivalently \(A_K\ge h^1\), then for every finite set \(S\) of primes of \(K\) there exists a tamely ramified extension \(F/K\) of degree \(p^m\) such that
\[
L_p(F)=L_p^S(F),
\]
\[
\mathrm{Gal}(L_p^S(F)/F)\simeq G,
\]
\(F/K\) is ramified at exactly \(m\) primes, and
\[
m\le e_G
\]
[2508.07946].

A particularly clean consequence is Corollary B: if \(K\) has trivial \(p\)-class group \(\mathrm{Cl}_p(K)=1\), then for any finite set \(S\) of primes of \(K\) and any finite \(p\)-group \(G\), there exists a tamely ramified extension \(F/K\), unramified at infinity, such that
\[
\mathrm{Gal}(L_p^S(F)/F)\simeq G
\]
[2508.07946]. Since \(\mathrm{Cl}_p(K)=1\) implies \(L_p(K)=K\), the starting tower is trivial, and the construction becomes maximally flexible.

This theorem extends ordinary realization results. Ozaki’s theorem established that every finite \(p\)-group can be realized as the Galois group of a \(p\)-class field tower over some number field, and later work gave a streamlined proof in arbitrary signature under the hypothesis that the class number of the base field is prime to \(p\), together with explicit bounds on degree and number of tame ramified primes [2204.08408]. The \(S\)-split theory strengthens this by adding decomposition requirements at a finite prescribed set \(S\) while preserving the ability to realize arbitrary finite \(p\)-groups [2508.07946].

## 4. Minkowski units, governing fields, and stability

The proof of the \(S\)-split realization theorem relies on a combination of cohomological control and explicit class field theory. A central invariant is the number \(A_K\) of Minkowski units in the tower. Using the \(\mathbf F_p[G_K]\)-module
\[
E_{L_p(K)}/(E_{L_p(K)})^p,
\]
where \(E_{L_p(K)}\) is the unit group of \(L_p(K)\), one has a decomposition
\[
M \simeq \mathbf F_p[G]^{A_K}\oplus N
\]
with \(N\) torsion, because \(\mathbf F_p[G]\) is a Frobenius algebra. The integer \(A_K\) measures the free part. Its growth under suitable \(\mathbf Z/p\)-extensions is controlled by
\[
A_F = A_K + (p-1)(r_{K,1}+r_{K,2})
\]
when \(F/K\) is a \(\mathbf Z/p\)-extension unramified at infinity with stable tower, and in general
\[
A_F\ge A_K
\]
[2508.07946].

The local splitting and ramification constraints are organized through the group
\[
V_S=\{x\in K^\times : (x)\in I_K^p(S)\},
\]
which fits into the exact sequence
\[
1\to E_S/(E_S)^p \to V_S/(K^\times)^p \to \mathrm{Cl}_S[p]\to 1.
\]
Here \(E_S\) is the group of \(S\)-units and \(\mathrm{Cl}_S\) is the \(S\)-class group. The associated governing field is
\[
\mathrm{Gov}_S := K'(\sqrt[p]{V_S}),
\]
where \(K'=K(\mu_p)\). Its Galois group
\[
M_S:=\mathrm{Gal}(\mathrm{Gov}_S/K')
\]
is an elementary abelian \(p\)-group, and Frobenius elements in this field determine whether one can construct \(\mathbf Z/p\)-extensions with prescribed ramification and splitting behavior [2508.07946].

A key input from Hajir–Maire–Ramakrishna is the stability theorem: if \(A_K\ge h^2\) and a tame prime \(q\) is chosen so that its Frobenius in the governing field matches a suitable element from the free part of the Minkowski-unit module, then there exists a \(\mathbf Z/p\)-extension \(N/K\) exactly ramified at \(q\) such that
\[
L_p(N)=N L_p(K).
\]
This means that the \(p\)-tower is stable under the base change \(K\mapsto N\). The \(S\)-split construction uses such extensions repeatedly while forcing each prime in \(S\) to become progressively more split in the tower [2508.07946].

## 5. Inductive construction and local splitting mechanism

The inductive argument begins by choosing a tame prime \(q\) such that the resulting \(\mathbf Z/p\)-extension \(N/K\) is ramified only at \(q\), inert at each prime in \(S\), and stable in the sense above. If a prime \(\mathfrak p\in S\) is not yet totally split in the new tower, then its residue degree drops by a factor of \(p\). Repeating this finitely many times forces every prime in \(S\) to split completely in the final field \(F\), yielding
\[
L_p(F)=L_p^S(F),
\]
while preserving the prescribed Galois group throughout the process [2508.07946].

This mechanism makes clear why the result is more refined than ordinary tower realizations. The extension \(F/K\) is not chosen merely to realize a group \(G\); it is constructed so that local decomposition at all primes in \(S\) is neutralized inside the maximal unramified pro-\(p\) extension. In that sense, the theorem solves an inverse Galois problem with both global and local specifications.

Related work shows that this local-global interaction persists in other parts of the subject. One study of class field towers in \(\mathbb Z_p\)-extensions defines \(F^{[0]}(L_n)\) as the maximal unramified pro-\(p\) extension of \(L_n\) in which the primes in \(S\) are completely split, and proves asymptotic lower bounds for growth invariants of the corresponding Galois groups [2309.03745]. Another line of work defines an \(S\)-class number as the degree of the maximal abelian extension unramified everywhere in which all places in \(S\) split completely, and uses Golod–Shafarevich methods to extract finite subextensions with infinite Hilbert \(p\)-class field towers [1904.07062]. These developments are not identical to the prescribed-group theorem, but they show that \(S\)-splitting is compatible both with explicit realizability and with asymptotic or infinitude phenomena.

## 6. Position within class field tower theory

The \(S\)-split \(p\)-Hilbert class field tower belongs to the broader program of controlling maximal pro-\(p\) extensions by simultaneously imposing global ramification conditions and local decomposition conditions. In ordinary tower theory, major themes include realization of finite \(p\)-groups, determination of tower length, detection of infinite towers, and analysis of low-level quotients via transfer data. The \(S\)-split theory adds a local decomposition layer to each of these questions.

In the finite-direction, the prescribed-group theorem shows that local splitting constraints do not destroy realizability [2508.07946]. In the ordinary setting, explicit realizations with controlled tame ramification were already known from Ozaki’s theorem and its later refinements [2204.08408]. In the diagnostic direction, IPADs and iterated IPADs provide a powerful bridge from class-group data to tower-group structure in the unramified setting [1502.03388]. This suggests a possible future interface between transfer-based diagnostics and \(S\)-split quotients of tower groups.

In the infinite-direction, classical and modern results on ordinary \(p\)-class towers show that infinitude can coexist with strong local restrictions. For example, for every prime \(p\) there exists a solvable number field ramified only at \(\{p,\infty\}\) whose Hilbert \(p\)-class field tower is infinite [1904.07062]. In quadratic imaginary settings, cohomological criteria involving generator and relation ranks, Zassenhaus filtration, and triple Massey products govern the difficult boundary between finite and infinite towers [1008.3003]. A plausible implication is that analogous higher-order cohomological obstructions may eventually play a role in deciding when an \(S\)-split tower is finite, infinite, or admits a prescribed finite quotient.

The principal conceptual contribution of the \(S\)-split perspective is therefore not a replacement of ordinary class field tower theory, but a refinement of it. It shows that maximal unramified pro-\(p\) extensions can be shaped by decomposition constraints at finitely many chosen primes without losing access to deep inverse Galois, cohomological, and tower-stability phenomena.

Source: https://www.emergentmind.com/topics/s-split-p-hilbert-class-field-tower