KEVOS
ArticlesServicesCase studiesAboutContact
ArticlesServicesCase studiesAboutContact
← ArticlesClass Group and Unit Computation: the Computational ProblemEngineering · Engineering MathematicsLesson 800/884← PrevNext →
ArticlePublished 7 Aug 20262 min readBy Kevin Joginclass groupunitsregulatorrelation collection
On this page

Ask about this page

KEVOS AIClass Group and Unit Computation: the Computational Problem

KEVOS knowledge first · trusted web sources when needed

Class Groups, Units and Regulators

Class Group and Unit Computation: the Computational Problem

Why class group and unit computation are a single problem, what the combined algorithm produces, and how the results are verified.

Engineering / MathematicsClass Groups, Units and Regulators2 min readKV-MATH-0598

Class group and unit group are computed together, not separately. The same relation collection produces both, and separating them would duplicate the expensive part of the work.

Why they are one problem

Relations among ideal classes are collected by finding elements whose principal ideals factor over the factor base. A relation that turns out to be trivial in the class group corresponds to a principal ideal, and the generator of a principal ideal that is trivial in every coordinate is a unit.

Collect relations→Smith normal form→Torsion gives the class group→Kernel gives units

Key point

The relation lattice's torsion quotient is the class group; its kernel corresponds to the units. One matrix, two answers. This is the structural reason the two computations are inseparable.

The combined output

What a complete run produces

Class groupInvariant factors plus generating ideals
Class numberProduct of the invariant factors
Unit rankKnown in advance from the signature
Fundamental unitsRecovered from trivial relations
RegulatorCovolume of the unit lattice
Roots of unityComputed separately and cheaply
ConditionalityWhether GRH was assumed

The pipeline

The combined class group and unit computation

  1. Build the factor basePrime ideals of norm below a chosen bound — see factor base selection.
  2. Collect relationsFind smooth principal ideals — see relation matrix construction.
  3. Reduce the matrixStructured elimination, then Smith normal form.
  4. Read the class groupFrom the invariant factors — see group structure.
  5. Recover unitsFrom the kernel — see regulator recovery.
  6. VerifyAgainst the analytic class number formula.

Two failure modes

How a class group computation can be wrong
FailureEffectDetection
Too few relationsComputed class number is a multiple of the truthAnalytic formula comparison
Factor base does not generateComputed group is a quotient of the truthAnalytic formula comparison; GRH bound rules it out

Caution

Neither failure is visible from the relation matrix itself. Both produce a self-consistent and entirely plausible answer. The external check is not optional.

Cost

Cost

This is the dominant cost in number field computation, running in sub-exponential time in the discriminant. It sets the practical limit on field size — see Buchmann's algorithm and, for the quadratic case, quadratic class groups.

Source. Henri Cohen, A Course in Computational Algebraic Number Theory, Springer GTM 138 — 4.9.3. Structural reference unverified: the source file was not available during authoring; chapter and section numbers are taken from the published edition and have not been checked against a physical copy.

Related pages

  • Learning Pathways Through Computational Number Theory
  • The Four Core Computational Tasks of Number Fields
  • Recovering Abelian Group Structure from a Relation Matrix
  • Buchmann's Sub-exponential Algorithm: Overview
  • Relation Matrix Construction

Continue learning

Minkowski and Bach BoundsArticle · Engineering MathematicsNEXT LESSON →Quadratic Field Discriminants and Integral BasesArticle · Engineering MathematicsThe Regulator: Definition and ComputationArticle · Engineering MathematicsPrime Decomposition in Quadratic FieldsArticle · Engineering Mathematics
KEVOS · Engineering, manufacturing and project improvement
ArticlesServicesCase studiesAboutContact
© 2026 KEVOS®