Documentation

LeanPool.CompositionAlgebrasOverRings.Composition.Basic

Composition algebras are alternative (GPR 19.3) #

GPR 19.3 says that a composition algebra is alternative, and that its norm is the unique conic norm on it. This file proves the alternativity statement in the two cases where the hypothesis GPR 19.3 actually consumes — vanishing of the radical of the norm — is available from non-singularity: over a field, unconditionally, and over a ring in which 2 is invertible.

The mathematical content is entirely in Conic/Kirmse.lean, where the Kirmse defect (conj x)(xy) - n(x)·y is shown to lie in the radical of the norm with no hypothesis at all. All that happens here is the passage from the composition-algebra hypothesis, QuadraticForm.IsGPRNonsingular, to QuadraticForm.IsGPRNondegenerate.

⚠ That passage is not free over a ring. Non-singularity does not imply that the radical vanishes: QuadraticForm.rankOne 1 over 𝔽₂[t]/(t²) is non-singular but has t in its radical. Over a field, non-singularity implies a vanishing radical, because it includes the trivial base change (QuadraticForm.radical_eq_bot_of_radical_baseChangeOfProjective_eq_bot) — only that direction: x² + t y² over 𝔽₂(t) has zero radical but acquires a radical vector over 𝔽₂(√t). Over a ring with 2 invertible, non-singular implies regular (GN (C.8.2)) and hence injective polar form. The general ring case is not a descent from the residue fields (Conic/Kirmse.lean explains why that fails); alternativity over an arbitrary ring is proved instead in Composition/Alternative.lean.

Main results #

References #

When 2 ∈ kˣ, the radical of the norm of a composition algebra vanishes: non-singular implies regular (GN (C.8.2)), and with 2 invertible an injective polar form is the same thing as a vanishing radical.

GPR 19.3 over a commutative ring in which 2 is invertible: a composition algebra is alternative.

Over a field, the radical of the norm of a composition algebra vanishes.

Non-singularity quantifies over every field extension of the base, so it applies in particular to the base field itself, and QuadraticForm.radical_eq_bot_of_radical_baseChangeOfProjective_eq_bot brings the conclusion back down along K → K ⊗[K] A. No hypothesis on 2, and none on the rank: in characteristic 2 the polar form may well have a kernel, but the radical — which constrains n as well as n(·,·) — does not.

GPR 19.3 over a field: every composition algebra is alternative.

No hypothesis on 2 and none on the rank. The characteristic-2 case, where the polar form of the norm can be degenerate, is covered: Kirmse's identities need the radical to vanish, not the polar form to be injective (Conic/Kirmse.lean).