Arakelov Ray Class Groups

• Arakelov ray class groups are compact abelian groups that unify finite congruence conditions with archimedean data via a fiber product over ℝ>0.
• They extend traditional ray class groups by incorporating a toric unit component, fitting into an exact sequence that underpins Cohen–Lenstra heuristics and capitulation results.
• This framework enables explicit class field generation, probabilistic modeling of arithmetic statistics, and analytic applications in prime ideal distribution and p-adic integration.

Arakelov ray class groups are compact abelian refinements of ray class groups that incorporate archimedean data into the class field theoretic modulus. For a number field $F$ and modulus $\fm=(\fm_{\mathrm f},\fm_\infty)$, the ordinary ray class group is

$\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$

where $P_F^1(\fm)$ is generated by principal ideals $(\alpha)$ with prescribed congruence at $\fm_{\mathrm f}$ and prescribed positivity at $\fm_\infty$. The Arakelov refinement introduced in recent heuristic work is

$\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$

with $\overline{F_\mathbf R^\times}$ the quotient of $(F\otimes_\mathbf Q\mathbf R)^\times$ by its maximal compact subgroup; it is a compact abelian group (Bartel et al., 24 Sep 2025). This construction places finite congruence conditions, sign conditions at real places, and unit data in a single object, and it connects the finite ray class groups of classical class field theory to Arakelov class groups, narrow class groups, explicit class field constructions, capitulation theorems, and probabilistic models for arithmetic statistics (Bartel et al., 24 Sep 2025).

1. Definition and exact-sequence structure

The starting point is the ordinary ray class exact sequence. For $\fm=(\fm_{\mathrm f},\fm_\infty)$0, $\fm=(\fm_{\mathrm f},\fm_\infty)$1 is defined by quotienting ideals prime to $\fm=(\fm_{\mathrm f},\fm_\infty)$2 by principal ideals generated by elements congruent to $\fm=(\fm_{\mathrm f},\fm_\infty)$3 modulo $\fm=(\fm_{\mathrm f},\fm_\infty)$4 and positive at the real places in $\fm=(\fm_{\mathrm f},\fm_\infty)$5 (Bartel et al., 24 Sep 2025). In the Arakelov refinement, one replaces the finite group $\fm=(\fm_{\mathrm f},\fm_\infty)$6 by a compact group obtained from a fiber product over $\fm=(\fm_{\mathrm f},\fm_\infty)$7, using the ideal norm on $\fm=(\fm_{\mathrm f},\fm_\infty)$8 and the absolute value of the $\fm=(\fm_{\mathrm f},\fm_\infty)$9-algebra norm on $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$0 (Bartel et al., 24 Sep 2025).

The resulting group $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$1 fits into an exact sequence whose left term is built from $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$2, the sign group $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$3, and the roots of unity $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$4, and whose quotient is the ordinary Arakelov class group $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$5 (Bartel et al., 24 Sep 2025). This is the structural reason the Arakelov ray class group is better suited to Cohen–Lenstra-type heuristics: when $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$6 is generated by a fixed rational integer and $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$7, the left term depends only on the splitting behavior of $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$8 in $\Cl_F(\fm)=\Id_F(\fm)/P_F^1(\fm),$9 and on $P_F^1(\fm)$0, so natural families can be partitioned into subfamilies with constant local term (Bartel et al., 24 Sep 2025).

There is also a direct comparison with the ordinary ray class group: $P_F^1(\fm)$1 Thus $P_F^1(\fm)$2 is an extension of the finite ray class group by a torus coming from the $P_F^1(\fm)$3-units modulo $P_F^1(\fm)$4 (Bartel et al., 24 Sep 2025). In this form, the Arakelov ray class group simultaneously contains the finite ray class part and the archimedean toric part contributed by units.

This packaging continues the general Arakelov philosophy developed for class groups without modulus. In that setting, the Pontryagin dual $P_F^1(\fm)$5 fits into

$P_F^1(\fm)$6

so torsion is the class group and the torsion-free part is the $P_F^1(\fm)$7-dual of the unit group (Bartel et al., 2020). A plausible implication is that Arakelov ray class groups should be viewed as the corresponding mod-$P_F^1(\fm)$8 enlargement, with ordinary ray class groups recovered as finite quotients.

2. Narrow, oriented, and archimedean variants

For real quadratic fields, the relevant object is often the narrow version $P_F^1(\fm)$9, corresponding to the modulus whose infinite part contains both real places (Bartel et al., 24 Sep 2025). In this setting, the connected components of $(\alpha)$0 are the narrow ideal classes, while the connected component itself is a $(\alpha)$1-dimensional torus-like object controlled by the fundamental unit (Bartel et al., 24 Sep 2025). After identifying Schoof’s oriented Arakelov class group with $(\alpha)$2, the connected-components map isometrically to closed geodesics on

$(\alpha)$3

(Bartel et al., 24 Sep 2025). This gives the archimedean part a concrete geometric interpretation.

At finite level, the narrow ray class group already incorporates the sign conditions at all real embeddings. For a fixed number field $(\alpha)$4 and integral ideal $(\alpha)$5, the narrow ray class group is

$(\alpha)$6

where $(\alpha)$7 consists of principal fractional ideals $(\alpha)$8 with $(\alpha)$9 and $\fm_{\mathrm f}$0 for every real embedding $\fm_{\mathrm f}$1 (Xie, 29 Jun 2026). The paper on products of prime ideals emphasizes that this is exactly the standard setting in which archimedean sign conditions are built into the modulus, and it explicitly observes that the same distributional results naturally apply to the corresponding Arakelov ray class groups (Xie, 29 Jun 2026).

The distinction between narrow finite quotients and compact Arakelov extensions is conceptually important. Narrow ray class groups are finite and encode prescribed positivity and congruence conditions in the sense of global class field theory. Arakelov ray class groups refine them by adjoining a torus $\fm_{\mathrm f}$2. This suggests that the narrow groups are the discrete connected-component data, while the Arakelov group remembers the continuous archimedean contribution.

3. Capitulation, tame ramification, and principalization

A major structural result for ray class groups is the extension of Bosca’s principalization method from ordinary ideal class groups to tame ray class groups (Jaulent, 2018). In that work, for a squarefree divisor $\fm_{\mathrm f}$3 of a number field $\fm_{\mathrm f}$4, the ray class group is written

$\fm_{\mathrm f}$5

where $\fm_{\mathrm f}$6 is the group of ideals prime to $\fm_{\mathrm f}$7 and $\fm_{\mathrm f}$8 is generated by principal ideals $\fm_{\mathrm f}$9 with $\fm_\infty$0 (Jaulent, 2018). The squarefree hypothesis is the condition that keeps the ramification tame; the paper explicitly says that from the class field theory viewpoint this “revient à accepter de la ramification modérée” (Jaulent, 2018).

The main capitulation theorem states that for every extension $\fm_\infty$1 in which at least one infinite place splits completely, and every finite set $\fm_\infty$2 of noncomplex places of $\fm_\infty$3, there exists a finite abelian extension $\fm_\infty$4, completely split at all infinite places, such that the ray class group of $\fm_\infty$5 modulo

$\fm_\infty$6

capitulates in $\fm_\infty$7; moreover there are infinitely many such $\fm_\infty$8, and they may be chosen unramified at any prescribed finite set of places (Jaulent, 2018). The splitting-at-infinity hypothesis is the cohomological input that allows the unit module to contain an augmentation-character piece via Herbrand’s theorem on units (Jaulent, 2018).

The proof is a ray-class analogue of Chevalley’s ambiguous class number method. One reduces to the $\fm_\infty$9-primary part, represents a given class by a prime ideal $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$0, constructs a cyclic $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$1-extension $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$2 ramified only at a carefully chosen prime $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$3, and then shows in the compositum $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$4 that the chosen class becomes principal because it is represented by a sufficiently high power of an ambiguous ideal (Jaulent, 2018). The ambiguous ray-class formula, the Herbrand quotient, and Hilbert 90 play the same roles as in the classical ideal-class case.

The appendix exhibits a sharp limitation. If one allows wild ramification, the analogue can fail: when the modulus contains primes above $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$5, the relevant $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$6-group of infinitesimal classes may be infinite, and under Leopoldt-type assumptions the capitulation map can even be injective (Jaulent, 2018). In Arakelov-style language, the tame theory behaves like the ordinary ideal-class theory, but the wild theory introduces genuinely different infinitesimal phenomena.

4. Explicit models and class field generation

Several strands of explicit class field theory realize ray class groups through more concrete objects, and these constructions are closely aligned with the Arakelov viewpoint because they combine finite congruence data with positivity or archimedean geometry.

For imaginary quadratic fields $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$7, the extended form class group

$\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$8

defined using primitive positive definite binary quadratic forms of discriminant $\Pic_F^0(\fm)=\big(\Id_F(\fm)\times_{\mathbf R_{>0}}\overline{F_\mathbf R^\times}\big)\big/\operatorname{im}(F^1(\fm)),$9 with $\overline{F_\mathbf R^\times}$0 and the congruence subgroup $\overline{F_\mathbf R^\times}$1, is canonically isomorphic to the ray class group $\overline{F_\mathbf R^\times}$2, hence to $\overline{F_\mathbf R^\times}$3 for $\overline{F_\mathbf R^\times}$4 (Eum et al., 2017). The group law is transported from the ray class group and is “basically the Dirichlet composition” (Eum et al., 2017). Passing to inverse limits gives a form-theoretic model for $\overline{F_\mathbf R^\times}$5, the maximal abelian extension unramified outside primes dividing $\overline{F_\mathbf R^\times}$6 (Eum et al., 2017).

For CM-fields $\overline{F_\mathbf R^\times}$7 with maximal real subfield $\overline{F_\mathbf R^\times}$8 of class number one, a form class group $\overline{F_\mathbf R^\times}$9 of primitive positive definite binary quadratic forms over $(F\otimes_\mathbf Q\mathbf R)^\times$0 is constructed and shown to be isomorphic to the ray class group $(F\otimes_\mathbf Q\mathbf R)^\times$1 via

$(F\otimes_\mathbf Q\mathbf R)^\times$2

(Jung et al., 2019). Under the additional assumption that the narrow class number of $(F\otimes_\mathbf Q\mathbf R)^\times$3 is one, every class has a totally positive representative, and singular values of Hilbert modular functions at the CM points $(F\otimes_\mathbf Q\mathbf R)^\times$4 generate a class field of the reflex field, with Galois group identified as

$(F\otimes_\mathbf Q\mathbf R)^\times$5

(Jung et al., 2019). The authors explicitly remark that this positivity theory is reminiscent of the role of positivity in narrow and Arakelov class groups.

Explicit CM generation results furnish another realization. For an imaginary quadratic field $(F\otimes_\mathbf Q\mathbf R)^\times$6 with $(F\otimes_\mathbf Q\mathbf R)^\times$7 and an integer $(F\otimes_\mathbf Q\mathbf R)^\times$8 divisible by $(F\otimes_\mathbf Q\mathbf R)^\times$9, the ray class field $\fm=(\fm_{\mathrm f},\fm_\infty)$00 modulo $\fm=(\fm_{\mathrm f},\fm_\infty)$01 is generated by the coordinates of a specific $\fm=(\fm_{\mathrm f},\fm_\infty)$02-torsion point on a normalized elliptic curve attached to $\fm=(\fm_{\mathrm f},\fm_\infty)$03: $\fm=(\fm_{\mathrm f},\fm_\infty)$04 (Koo et al., 2010). Under a degree inequality and for odd $\fm=(\fm_{\mathrm f},\fm_\infty)$05, the singular value $\fm=(\fm_{\mathrm f},\fm_\infty)$06 alone generates $\fm=(\fm_{\mathrm f},\fm_\infty)$07, giving a partial result toward the Lang–Schertz conjecture (Koo et al., 2010). Related work compares the full torsion field $\fm=(\fm_{\mathrm f},\fm_\infty)$08 of a CM elliptic curve $\fm=(\fm_{\mathrm f},\fm_\infty)$09 with the ray class field $\fm=(\fm_{\mathrm f},\fm_\infty)$10, proving for $\fm=(\fm_{\mathrm f},\fm_\infty)$11 that

$\fm=(\fm_{\mathrm f},\fm_\infty)$12

(Jung et al., 2020). These results concern finite ray class fields rather than compact Arakelov groups, but they provide explicit generators for the finite quotients that the Arakelov refinement extends.

5. Statistics, heuristics, and random exact sequences

The statistical theory of Arakelov ray class groups was formulated for real quadratic fields in terms of random extensions rather than random finite groups (Bartel et al., 24 Sep 2025). Fix a finite set $\fm=(\fm_{\mathrm f},\fm_\infty)$13 of odd primes and a local algebra $\fm=(\fm_{\mathrm f},\fm_\infty)$14 determined by the finite modulus. Writing

$\fm=(\fm_{\mathrm f},\fm_\infty)$15

the heuristic studies short exact sequences

$\fm=(\fm_{\mathrm f},\fm_\infty)$16

and postulates that, locally at $\fm=(\fm_{\mathrm f},\fm_\infty)$17, the $\fm=(\fm_{\mathrm f},\fm_\infty)$18-torsion of the Arakelov ray class sequence of a real quadratic field is distributed according to a Cohen–Lenstra-type measure in which a sequence appears with probability proportional to $\fm=(\fm_{\mathrm f},\fm_\infty)$19 (Bartel et al., 24 Sep 2025). The paper emphasizes that this is the right object to randomize because it encodes not only the class-group part but also the reduction of units modulo the finite modulus (Bartel et al., 24 Sep 2025).

Several consequences are derived. The heuristic implies the Cohen–Lenstra heuristics on class groups of real quadratic fields; it predicts equidistribution of the fundamental unit modulo an integer; it predicts asymptotic independence of the class-group part and the unit-reduction part; and it recovers equidistribution statements for ray class extension classes in $\fm=(\fm_{\mathrm f},\fm_\infty)$20 that are compatible with earlier work of Pagano–Sofos on imaginary quadratic ray class groups (Bartel et al., 24 Sep 2025). The same paper checks compatibility with Varma’s results on average sizes of $\fm=(\fm_{\mathrm f},\fm_\infty)$21-torsion subgroups of ray class groups of quadratic fields (Bartel et al., 24 Sep 2025).

At finite level, Varma studies $\fm=(\fm_{\mathrm f},\fm_\infty)$22 for quadratic fields $\fm=(\fm_{\mathrm f},\fm_\infty)$23 and fixed integral conductor $\fm=(\fm_{\mathrm f},\fm_\infty)$24, under the admissibility condition that $\fm=(\fm_{\mathrm f},\fm_\infty)$25 is squarefree away from $\fm=(\fm_{\mathrm f},\fm_\infty)$26 and $\fm=(\fm_{\mathrm f},\fm_\infty)$27 if $\fm=(\fm_{\mathrm f},\fm_\infty)$28 (Varma, 2016). The $\fm=(\fm_{\mathrm f},\fm_\infty)$29-torsion is related to counts of cubic fields with prescribed ramification, and the resulting averages are given by explicit Euler products and powers of $\fm=(\fm_{\mathrm f},\fm_\infty)$30; for $\fm=(\fm_{\mathrm f},\fm_\infty)$31 coprime to $\fm=(\fm_{\mathrm f},\fm_\infty)$32, the counting function has a secondary term

$\fm=(\fm_{\mathrm f},\fm_\infty)$33

(Varma, 2016). The paper also proves positive-proportion results for trivial $\fm=(\fm_{\mathrm f},\fm_\infty)$34-torsion in certain families (Varma, 2016). Its language is finite and non-Arakelov, but it explicitly situates these ray class groups as the finite arithmetic objects governing abelian extensions with prescribed local conditions.

Pagano–Sofos go further by treating the ray class sequence itself as the random object. For imaginary quadratic fields and conductor $\fm=(\fm_{\mathrm f},\fm_\infty)$35, they study the exact sequence

$\fm=(\fm_{\mathrm f},\fm_\infty)$36

as a sequence of $\fm=(\fm_{\mathrm f},\fm_\infty)$37-modules, and they prove the conjectural $\fm=(\fm_{\mathrm f},\fm_\infty)$38-primary distribution at the level of the $\fm=(\fm_{\mathrm f},\fm_\infty)$39-rank statistic by asymptotically evaluating mixed moments of the corresponding cohomology map (Pagano et al., 2017). This finite exact-sequence viewpoint is explicitly close in spirit to the later Arakelov ray class formulation, where the ordinary ray class sequence is recovered as a quotient of the compact Arakelov sequence (Bartel et al., 24 Sep 2025).

This probabilistic program extends a broader correction to Cohen–Lenstra–Martinet heuristics in which Arakelov class groups, rather than class groups alone, are treated as the fundamental random objects because they package units and class groups in a single Galois module (Bartel et al., 2020, Bartel et al., 2018).

6. Analytic applications and arithmetic constraints

The archimedean sign-conditioned ray class setting supports strong analytic results on prime ideals. For a fixed number field $\fm=(\fm_{\mathrm f},\fm_\infty)$40 and integral ideal $\fm=(\fm_{\mathrm f},\fm_\infty)$41, every class in the narrow ray class group modulo $\fm=(\fm_{\mathrm f},\fm_\infty)$42 is represented by a product of three prime ideals of norm at most

$\fm=(\fm_{\mathrm f},\fm_\infty)$43

for any $\fm=(\fm_{\mathrm f},\fm_\infty)$44, under the stated character-sum and bounded-order subconvexity hypotheses (Xie, 29 Jun 2026). Using Wu’s subconvexity bound, one may take $\fm=(\fm_{\mathrm f},\fm_\infty)$45, yielding the explicit exponent $\fm=(\fm_{\mathrm f},\fm_\infty)$46 (Xie, 29 Jun 2026). The same work proves that a positive proportion of ray classes are represented by products of two prime ideals (Xie, 29 Jun 2026). Since the group already includes all real places in the modulus, these are distribution results in the standard finite-level Arakelov setting.

Ray class groups also furnish the natural inverse-limit domain for $\fm=(\fm_{\mathrm f},\fm_\infty)$47-adic integration. In the theory of $\fm=(\fm_{\mathrm f},\fm_\infty)$48-adic $\fm=(\fm_{\mathrm f},\fm_\infty)$49-functions for automorphic forms on $\fm=(\fm_{\mathrm f},\fm_\infty)$50, compatible systems on finite ray class groups $\fm=(\fm_{\mathrm f},\fm_\infty)$51 are interpreted as locally constant distributions on a profinite global class group (Loeffler, 2013). If a ray class distribution has local growth parameters $\fm=(\fm_{\mathrm f},\fm_\infty)$52, then it induces a locally constant distribution of finite order

$\fm=(\fm_{\mathrm f},\fm_\infty)$53

and if $\fm=(\fm_{\mathrm f},\fm_\infty)$54 it extends uniquely to a finite-order distribution (Loeffler, 2013). In the split imaginary quadratic supersingular case $\fm=(\fm_{\mathrm f},\fm_\infty)$55, this yields a two-variable plus-minus decomposition and four signed bounded $\fm=(\fm_{\mathrm f},\fm_\infty)$56-adic $\fm=(\fm_{\mathrm f},\fm_\infty)$57-functions after division by Pollack half-logarithms, confirming a conjecture of B.D. Kim (Loeffler, 2013).

Arithmetic applications also appear in the study of minus parts of ray class groups of real quadratic fields. For primes $\fm=(\fm_{\mathrm f},\fm_\infty)$58 with $\fm=(\fm_{\mathrm f},\fm_\infty)$59, if $\fm=(\fm_{\mathrm f},\fm_\infty)$60 denotes the order of the minus part of the ray class group of $\fm=(\fm_{\mathrm f},\fm_\infty)$61 of modulus $\fm=(\fm_{\mathrm f},\fm_\infty)$62, then—assuming a non-emptiness hypothesis—the number of positive fundamental discriminants $\fm=(\fm_{\mathrm f},\fm_\infty)$63 with $\fm=(\fm_{\mathrm f},\fm_\infty)$64 split and $\fm=(\fm_{\mathrm f},\fm_\infty)$65 is bounded below by

$\fm=(\fm_{\mathrm f},\fm_\infty)$66

when $\fm=(\fm_{\mathrm f},\fm_\infty)$67 is a Sophie Germain prime, the existence input is unconditional (Lecouturier et al., 15 Jun 2026). Combined with results of Lecouturier–Wang, this has consequences for the $\fm=(\fm_{\mathrm f},\fm_\infty)$68-part of BSD for even quadratic twists of $\fm=(\fm_{\mathrm f},\fm_\infty)$69 (Lecouturier et al., 15 Jun 2026).

A persistent constraint is that tame and wild behavior diverge sharply. In capitulation theory, tame squarefree moduli support principalization theorems, whereas allowing wild ramification can make the relevant infinitesimal class group infinite and can force capitulation maps to be injective under Leopoldt-type assumptions (Jaulent, 2018). This contrast is one of the clearest indications that Arakelov ray class groups, despite their formal similarity to ordinary ray class groups, are sensitive to the interaction of finite ramification, archimedean conditions, and unit-theoretic infinitesimals in ways that are not captured by finite class groups alone.