Ideal class groups and Minkowski bounds

Nonzero fractional ideals in the integer ring form an abelian group under ideal multiplication, because every such ideal is invertible. Principal fractional ideals generated by nonzero field elements form a subgroup. Their quotient is the ideal class group. An integral ideal and a fractional ideal can represent the same class, so restricting the definition to principal integral divisibility would miss inverses. The identity class consists of principal fractional ideals; multiplication of classes is induced by ideal multiplication.

For a Dedekind domain, a trivial class group means every nonzero ideal is principal, hence the ring is a principal ideal domain and a unique factorization domain for elements. Conversely a unique factorization domain has trivial class group in this setting. A nonprincipal ideal whose square is principal therefore yields a class of order two if it is not already the identity. This is a group statement; a failure to find a small generator numerically does not prove nonprincipality without a norm or other obstruction.

Let K have degree n, r₁ real embeddings, and r₂ conjugate pairs of complex embeddings, so $n=r_1+2r_2$. Every ideal class contains a nonzero integral ideal with norm at most $(4/\pi)^{r_2}(n!/n^n)\sqrt{|d_K|}$. The signature-dependent factor and the field discriminant are essential. Replacing the field discriminant by an order discriminant can weaken or misidentify the calculation, while forgetting the factorial factor can greatly enlarge the apparent search space.

The proof embeds an ideal as a lattice in the product of real and complex embedding spaces. A convex-body argument finds a nonzero element with controlled absolute field norm. Combining its principal ideal with the inverse of the starting ideal produces a small integral representative of the inverse class; inversion permutes the classes, giving the bound for every class. This is a geometry-of-numbers theorem, not an assertion that every ideal itself has small norm or that every small ideal is principal.

Once the bound is evaluated, list all possible integral ideal norms up to it, then factor the relevant rational primes and determine which ideals actually have those norms. Norm one gives only the unit ideal. If every candidate ideal is principal, the class group is trivial. If one nonprincipal candidate survives, establish its relations and ensure that the list accounts for every class. The bound limits the candidates but does not classify them automatically from their numerical norms.

Norm equations are useful for ruling out principal generators. If an ideal of norm two were principal, some integer-ring element would have absolute field norm two. In an imaginary quadratic ring this becomes a positive-definite quadratic equation, often easy to exclude. In a real quadratic ring signs and infinitely many units complicate the search, so a bounded scan without justification can be inconclusive. Distinguish a complete theoretical elimination from a numerical experiment that has merely failed to find a solution.