betaFeedback

WorksMathematics

Half-factorial generating sets and the realization of Dedekind class groups

Abstract

We prove a cardinality obstruction to realizing an abelian group as the ideal class group of a half-factorial Dedekind domain. If an abelian pp-group PP satisfies ⋂r≥1prP=0\bigcap_{r\ge1}p^rP=0 and admits a half-factorial generating subset, then ∣P∣≤max⁡{ℵ0,∣P/pP∣}|P|\le\max\{\aleph_0,|P/pP|\}. The proof separates generators of equal order by finite quotients. Its group-theoretic ingredient is a direct summand generated by selected members of an arbitrary finite generating set. The torsion subgroup of ∏j≥1Z/pjZ\prod_{j\ge1}\mathbb{Z}/p^j\mathbb{Z} is uncountable and pp-adically separated, but has countable quotient modulo pp; hence it has no half-factorial generating subset. This gives a negative answer to the universal half-factorial Dedekind realization problem for every prime. Pure subgroups give counterexamples of every uncountable cardinality at most the continuum, so the least cardinality of a counterexample is ℵ1\aleph_1. We also study free-kernel lifting, order-preserving lifts across torsion kernels, character-compatible generators, and the failure of compatibility under unions and contraction. These results distinguish global realizability from half-factoriality on bounded subgroups and from integrality of cross numbers.

Introduction

Factorization into irreducibles is unique in a factorial domain, but uniqueness of the number of factors is a substantially weaker condition. A domain is half-factorial if every nonzero nonunit factors into irreducibles and all such factorizations have the same length. Carlitz’s theorem characterizes half-factorial rings of integers of algebraic number fields by class number at most two [1]. The corresponding question for arbitrary Dedekind domains is different because the distribution of prime ideals among the ideal classes is no longer fixed by the number-field setting.

Claborn proved that every abelian group is the ideal class group of a Dedekind domain [3]. Grams and subsequent realization theorems permit prescribed classes of prime divisors under the appropriate generating condition [11, 2]. Thus the abstract class group does not by itself determine the factorization lengths of a particular Dedekind domain. Nevertheless, it is natural to ask whether every abelian group admits some prime-divisor distribution for which the resulting domain is half-factorial. Equivalently, one asks:

Is every abelian group isomorphic to the ideal class group of a half-factorial Dedekind domain?

This is Problem 15 in the survey of Geroldinger, Kim, and Loper [10 §12.1].

The group-theoretic formulation uses zero-sum sequences. For S⊆GS\subseteq G, let B(S)\mathcal{B}(S) be the monoid of finite sequences over SS whose sum is zero. The classical realization criterion says that GG occurs as the class group of a half-factorial Dedekind domain precisely when it has a monoid-generating subset SS for which B(S)\mathcal{B}(S) is half-factorial [2 Theorem 3.6]. Group generation and monoid generation agree for torsion groups. For such a group, the Skula–Śliwa cross-number criterion characterizes half-factoriality by k(U)=1k(U)=1 for every minimal nonempty zero-sum sequence UU, where k(U)=∑svs(U)/ord⁡(s)k(U)=\sum_s v_s(U)/\operatorname{ord}(s) [7 Lemma 3.2].

Michel and Steffan obtained early positive realization results [12]. Geroldinger and Göbel proved that every Warfield group admits half-factorial generators and, in particular, that every countable abelian group does so [8 Theorems 5.1 and 5.3]. They also reduced the universal problem to reduced uncountable unbounded primary groups and singled out the torsion-completion of ⨁j≥1Cpj\bigoplus_{j\geq1} C_{p^j} as an unresolved example [8 pp. 855–856]. The positive lifting and reduction statements are established results; we use them with explicit attribution rather than treating them as contributions of this paper.

Our main contribution is a restriction on arbitrary half-factorial generating sets. It does not require generators of the socle, compatible generators of bounded layers, or a prescribed character. Say that an abelian pp-group is pp-adically separated if ⋂r≥1prP=0\bigcap_{r\geq1}p^rP=0. The principal conclusions are as follows.

Main Theorem. Let pp be a prime.

(1) If a pp-adically separated abelian pp-group PP has a half-factorial generating subset, then

∣P∣≤max⁡{ℵ0,∣P/pP∣}.|P| \leq\max\{\aleph_0, |P/pP|\}.

(2) The group Pp=t(∏j≥1Z/pjZ)P_p=t\left(\prod_{j\geq1}\mathbb{Z}/p^j\mathbb{Z}\right), where tt denotes the torsion subgroup, has no half-factorial generating subset. Consequently it is not the ideal class group of a half-factorial Dedekind domain, nor the divisor class group of a half-factorial Krull monoid.

(3) For every uncountable cardinal κ≤2ℵ0\kappa\leq2^{\aleph_0}, there is a reduced abelian pp-group of cardinality κ\kappa with no half-factorial generating subset. The least cardinality of an abelian group not realizable as the class group of a half-factorial Dedekind domain is therefore ℵ1\aleph_1.

The proof of the cardinality bound is finite at its decisive step. Given a finite abelian pp-group GG, a generating set SS, and r≥1r\geq1, we select A⊆SA\subseteq S such that K=⟨A⟩K=\langle A\rangle is a direct summand with prG⊆Kp^rG\subseteq K and K[p]⊆prGK[p]\subseteq p^rG. Two distinct generators of the same order that become equal modulo KK without losing order yield a minimal zero-sum sequence of cross number greater than one. It follows that, for a half-factorial generating set of any abelian pp-group,

ord⁡(s)=ord⁡(t)=pn,s−t∈prP,pn−1s∉prP⟹s=t.\operatorname{ord}(s)=\operatorname{ord}(t)=p^n,\quad s-t\in p^rP,\quad p^{n-1}s\notin p^rP\quad\Longrightarrow\quad s=t.

Stratifying the generators by nn and rr gives the cardinality bound. This separates the present argument from obstructions to particular extension procedures: the necessary subgroup is selected from the very generating set being tested.

The complementary results address how half-factorial generating sets behave under algebraic constructions. We extend the Geroldinger–Göbel free-kernel decomposition to quotients that may contain elements of infinite order. For torsion kernels, we prove that half-factoriality of an order-preserving lifted generating set is equivalent to splitting along those representatives. We also construct weakly half-factorial generators for every torsion abelian group, impose a prescribed character in bounded primary groups, and give explicit obstructions to thinning, adjoining roots, and repeated multiplication by a prime. The weak character criterion and its finite-group antecedents belong to the established theory [7, 13, 14]; the additional constructions and proofs are presented separately from the main nonrealization argument.

Section 2 fixes conventions and recalls the realization and cross-number criteria. Section 3 proves the Main Theorem and its cardinal refinements. Section 4 treats lifting, the classical reduction, and pure extensions. Section 5 studies characters and finite obstructions. Section 6 concludes with socle compatibility, contraction, and the limitations of the cardinality criterion.

Zero-sum sequences and realization

This section establishes the conventions used throughout the paper, records the exact connection with Dedekind class groups, and gives the elementary positive examples needed later. We distinguish the ambient group of labels from the intrinsic divisor class group of an unlabelled block monoid. Standard references for the group and factorization theory are [5, 6, 9].

Conventions and factorization lengths

We write N={1,2,…}\mathbb{N} = \{1, 2, \ldots\} and N0=N∪{0}\mathbb{N}_0 = \mathbb{N} \cup\{0\}. All groups are abelian and written additively. All monoids are commutative and cancellative and have an identity. A monoid is reduced if its only unit is its identity; an atom is a nonunit that is not a product of two nonunits. A monoid is atomic if every nonunit is a finite product of atoms. Write A(M)\mathcal{A}(M) for the set of atoms of a monoid MM and q(M)\mathrm{q}(M) for its quotient group. For an atomic monoid MM and a∈Ma \in M, let LM(a)\mathsf{L}_M(a) be the set of lengths of its factorizations into atoms, and set LM(a)={0}\mathsf{L}_M(a) = \{0\} for a unit. The monoid is half-factorial if every such set is a singleton. A domain has the corresponding property when its multiplicative monoid of nonzero elements does. All rings are commutative with 1≠01 \ne0, and ring homomorphisms preserve the identity.

For a group GG and S⊆GS \subseteq G, write ⟨S⟩\langle S\rangle for the generated subgroup and [S][S] for the generated submonoid, including the empty sum. Thus [S]=G[S] = G requires nonnegative coefficients. If GG is torsion, then [S]=⟨S⟩[S] = \langle S\rangle. Write Cn=Z/nZC_n = \mathbb{Z}/n\mathbb{Z} and identify Cp∞C_{p^\infty} with the pp-primary subgroup of Q/Z\mathbb{Q}/\mathbb{Z}. An abelian pp-group is a group all of whose elements have finite pp-power order. For m≥1m \ge1, put G[m]={g∈G:mg=0}G[m] = \{g \in G : mg = 0\}. The subgroup P[p]P[p] of a pp-group is its socle, an Fp\mathbb{F}_p-vector space. A group is divisible if multiplication by every positive integer is surjective, and reduced if it contains no nonzero divisible subgroup. A subgroup H⊆GH \subseteq G is pure if H∩mG=mHH \cap mG = mH for every m≥1m \ge1. A pp-group is bounded if pNP=0p^N P = 0 for some NN, and unbounded otherwise. The axiom of choice is used for bases and cardinal comparisons.

Let F(S)\mathcal{F}(S) be the free commutative monoid on SS. Its members are finite sequences, written B=∏s∈Ssvs(B)B = \prod_{s \in S} s^{v_s(B)} with finite support. The sequence length and sum are ∣B∣=∑s∈Svs(B)|B| = \sum_{s \in S} v_s(B) and σ(B)=∑s∈Svs(B)s\sigma(B) = \sum_{s \in S} v_s(B)s. A subsequence means a divisor in F(S)\mathcal{F}(S), including multiplicities. A zero-sum sequence over an additive abelian group GG is a finite sequence of elements of GG, with repetitions allowed and order disregarded, whose sum is 00. It is minimal if it is nonempty and no proper nonempty subsequence has sum 00. The block monoid is

B(S)={B∈F(S):σ(B)=0}.\mathcal{B}(S) = \{B \in\mathcal{F}(S) : \sigma(B) = 0\}.

Its identity is the empty sequence. Its atoms are exactly the minimal nonempty zero-sum sequences; induction on ∣B∣|B| proves atomicity. A subset SS is half-factorial if B(S)\mathcal{B}(S) is half-factorial. We write H\mathcal{H} for the class of groups possessing a half-factorial subset that generates them as a monoid. The empty set is allowed for the trivial group.

Adding or deleting the letter zero does not affect generation or half-factoriality: if 0∉S0 \notin S, then B(S∪{0})≅B(S)×N0\mathcal{B}(S \cup\{0\}) \cong\mathcal{B}(S) \times\mathbb{N}_0. Unless stated otherwise, generating sets omit zero. We use the usual convention ord⁡(0)=1\operatorname{ord}(0) = 1 when zero occurs in a sequence. Half-factoriality has finite character: a set is half-factorial precisely when each of its finite subsets is half-factorial [7 Lemma 3.1]. In particular, subsets and unions of nested chains of half-factorial sets are half-factorial. Indeed, any failure is witnessed by two factorizations using only finitely many letters. This observation concerns chains of sets, not arbitrary increasing unions of the groups they generate.

The realization and cross-number criteria

A Dedekind domain is a Noetherian integrally closed domain of Krull dimension at most one. Its ideal class group is the group of nonzero fractional ideals modulo nonzero principal fractional ideals. Let HH be a commutative cancellative monoid and let F=F(P)F = \mathcal{F}(P) be the free commutative monoid on a set PP. A monoid homomorphism φ:H→F\varphi: H \to F is a divisor homomorphism if, for all a,b∈Ha, b \in H, φ(a)∣φ(b)\varphi(a) \mid\varphi(b) implies a∣Hba \mid_H b. It is a divisor theory if, in addition, for every f∈Ff \in F there exist m≥1m \ge1 and a1,…,am∈Ha_1, \ldots, a_m \in H such that

f=gcd⁡F(φ(a1),…,φ(am)).f = \gcd_F(\varphi(a_1), \ldots, \varphi(a_m)).

Here greatest common divisors in FF are computed by taking the minimum exponent of each generator in PP. A commutative cancellative monoid is a Krull monoid if it admits a divisor theory. For a divisor theory φ:H→F\varphi: H \to F, its divisor class group is

C(H)=q(F)/q(φ(H)).C(H)=\mathsf{q}(F)/\mathsf{q}(\varphi(H)).

This group is independent, up to isomorphism, of the chosen divisor theory; see [9]. We recall the realization theorem in the form needed here. The existence statement is classical [11, 2]; its formulation in terms of generating subsets is also recorded in [7 p. 598] and [8 p. 842].

Theorem 2.1 (Classical realization criterion). Let GG be an abelian group. The following are equivalent:

  1. GG has a half-factorial subset SS with [S]=G[S]=G;

  2. GG is isomorphic to the ideal class group of a half-factorial Dedekind domain;

  3. GG is isomorphic to the divisor class group of a half-factorial Krull monoid.

More precisely, for every nonempty S⊆GS\subseteq G with [S]=G[S]=G, there is a Dedekind domain whose class group is identified with GG and whose classes containing nonzero prime ideals are exactly SS. Its factorization lengths are transferred to B(S)\mathcal{B}(S), so it is half-factorial precisely when SS is half-factorial.

Explanation of the equivalence. The prescribed prime-class realization theorem supplies the domain. If (a)(a) factors into prime ideals, replace each prime ideal by its class to obtain a block over SS. A zero-sum subsequence corresponds to a principal subproduct; its generator gives the corresponding element factor. Partitioning the prime-ideal occurrences lifts factorizations, including their lengths. Conversely, irreducible elements give minimal zero-sum subsequences under this transfer. These are the standard block-transfer properties [2, 9].

The same argument applies to a divisor theory of a Krull monoid. The classes of its prime divisors generate the class group as a monoid, and the block transfer preserves sets of lengths. Thus (3) implies (1), while (1) implies (2) by realization and (2) implies (3) by the divisor theory of a Dedekind domain. The trivial group may be realized by k[X]k[X] for any field kk. □\square

For a subset TT of a torsion group and B∈F(T)B\in\mathcal{F}(T), define its cross number by

k(B)=∑t∈Tvt(B)ord⁡(t).\mathsf{k}(B)=\sum_{t\in T}\frac{\mathsf{v}_t(B)}{\operatorname{ord}(t)}.

The following criterion is due to Skula and Śliwa; see [7 Lemma 3.2] and [14 Theorem 4.1]. We include its short proof because it is the only numerical factorization criterion used in the nonrealization argument.

Proposition 2.2 (Cross-number criterion). Let GG be a torsion abelian group and let T⊆GT\subseteq G, with ord⁡(0)=1\operatorname{ord}(0)=1 if 0∈T0\in T. Then TT is half-factorial if and only if k(U)=1\mathsf{k}(U)=1 for every U∈A(B(T))U\in\mathcal{A}(\mathcal{B}(T)).

Proof. If every atom has cross number one, additivity of k\mathsf{k} shows that all factorizations of a fixed block have the same length. Conversely, assume half-factoriality and let UU be an atom. Choose a positive integer mm divisible by the orders of all elements in its support. The block UmU^m has a factorization of length mm using UU, and a factorization using the pure atoms tord⁡(t)t^{\operatorname{ord}(t)} of length ∑tmvt(U)/ord⁡(t)=mk(U)\sum_t m\mathsf{v}_t(U)/\operatorname{ord}(t)=m\mathsf{k}(U). Equality of these lengths gives k(U)=1\mathsf{k}(U)=1. □\square

Labels and the intrinsic class group

The quotient of the ambient free group by the group of blocks need not be the intrinsic divisor class group of an unlabelled block monoid. The distinction is relevant when applying a realization theorem; compare the divisor-theoretic discussion in [15 §2].

Example 2.3. Let G=Cn=⟨g⟩G=C_n=\langle g\rangle, with n≥2n\ge2, and let S={g}S=\{g\}. Then [S]=G[S]=G and B(S)={gnk:k∈N0}\mathcal{B}(S)=\{g^{nk}:k\in\mathbb{N}_0\} is a free monoid on one generator. Its intrinsic divisor class group is therefore zero. On the other hand, q(F(S))/q(B(S))≅Z/nZ\mathsf{q}(\mathcal{F}(S))/\mathsf{q}(\mathcal{B}(S))\cong\mathbb{Z}/n\mathbb{Z}. The inclusion B(S)⊂F(S)\mathcal{B}(S)\subset\mathcal{F}(S) is saturated, but is not a divisor theory. Indeed, gg is not a greatest common divisor of finitely many members of B(S)\mathcal{B}(S).

An inclusion H⊆F(P)H \subseteq\mathcal{F}(P) is saturated if H=q(H)∩F(P)H = \mathsf{q}(H) \cap\mathcal{F}(P) inside q(F(P))\mathsf{q}(\mathcal{F}(P)); equivalently, the inclusion is a divisor homomorphism. For such an inclusion to be a divisor theory, it suffices that every p∈Pp \in P be the greatest common divisor of finitely many elements of HH. For such an inclusion the divisor class group is q(F(P))/q(H)\mathsf{q}(\mathcal{F}(P))/\mathsf{q}(H).

Proposition 2.4. Let GG be an abelian group and let S⊆GS \subseteq G satisfy [S]=G[S] = G. Set P=S×NP = S \times\mathbb{N}, let β:F(P)→G\beta: \mathcal{F}(P) \to G send (s,j)(s,j) to ss, and put H=β−1(0)H = \beta^{-1}(0). Then H⊆F(P)H \subseteq\mathcal{F}(P) is a divisor theory with class group GG. Forgetting the second coordinate gives a surjective homomorphism H→B(S)H \to\mathcal{B}(S) that preserves all sets of factorization lengths. In particular, HH is half-factorial if and only if SS is half-factorial.

References

  1. [1]L. Carlitz, A characterization of algebraic number fields with class number two, Proc. Amer. Math. Soc. 11 (1960), 391–392.DOI
  2. [2]S. T. Chapman and A. Geroldinger, Krull domains and monoids, their sets of lengths, and associated combinatorial problems, in Factorization in Integral Domains (D. D. Anderson, ed.), Lecture Notes in Pure and Appl. Math. 189, Marcel Dekker, New York, 1997, 73–112.
  3. [3]L. Claborn, Every abelian group is a class group, Pacific J. Math. 18 (1966), no. 2, 219–222.DOI
  4. [4]D. Dikranjan and A. Giordano Bruno, Functional topologies and finite-index subgroups of abelian groups, Topology Appl. 158 (2011), no. 17, 2391–2407.arxiv.org/abs/1103.4565
  5. [5]L. Fuchs, Infinite Abelian Groups, Vol. I, Pure Appl. Math. 36, Academic Press, New York–London, 1970.
  6. [6]L. Fuchs, Infinite Abelian Groups, Vol. II, Pure Appl. Math. 36-II, Academic Press, New York–London, 1973.
  7. [7]W. Gao and A. Geroldinger, Half-factorial domains and half-factorial subsets of abelian groups, Houston J. Math. 24 (1998), no. 4, 593–611.
  8. [8]A. Geroldinger and R. Göbel, Half-factorial subsets in infinite abelian groups, Houston J. Math. 29 (2003), no. 4, 841–858.
  9. [9]A. Geroldinger and F. Halter-Koch, Non-Unique Factorizations: Algebraic, Combinatorial and Analytic Theory, Pure Appl. Math. 278, Chapman & Hall/CRC, Boca Raton, FL, 2006.
  10. [10]A. Geroldinger, H. Kim, and K. A. Loper, On long-term problems in multiplicative ideal theory and factorization theory, in The Ideal Theory and Arithmetic of Rings, Monoids, and Semigroups (S. T. Chapman, ed.), Contemp. Math. 836, Amer. Math. Soc., Providence, RI, 2026, 1–33. doi:10.1090/conm/836/16701. Also available as arXiv:2502.21020v3.DOI
  11. [11]A. P. Grams, The distribution of prime ideals of a Dedekind domain, Bull. Austral. Math. Soc. 11 (1974), no. 3, 429–441.DOI
  12. [12]D. Michel and J.-L. Steffan, Répartition des idéaux premiers parmi les classes d’idéaux dans un anneau de Dedekind et équidécomposition, J. Algebra 98 (1986), no. 1, 82–94.
  13. [13]M. Radziejewski and W. A. Schmid, Weakly half-factorial sets in finite abelian groups, Forum Math. 19 (2007), no. 4, 727–747.DOI
  14. [14]W. A. Schmid, Half-factorial sets in finite abelian groups: a survey, Grazer Math. Ber. 348 (2005), 41–64.
  15. [15]W. A. Schmid, Higher-order class groups and block monoids of Krull monoids with torsion class group, J. Algebra Appl. 9 (2010), no. 3, 433–464.

Paper details

vlnc:W_rdrzrx6v
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

Not recorded for this paper

View statement
Looking for related papers…
Contents