Documentation

LeanPool.Ado.Algebra.Lie.Killing.BaseChange

The Killing property under base change #

LieAlgebra.IsKilling R L says that the Killing form of L is nonsingular. This file proves that for a finite free Lie algebra over an integral domain the property descends along any base change into a second integral domain, and that an injective base change loses nothing either:

IsKilling A (A ⊗[R] L) → IsKilling R L,  and  IsKilling A (A ⊗[R] L) ↔ IsKilling R L.

Both halves are useful. The descent direction transports the Killing property down from a large coefficient ring to a small one, which is what a construction over ℚ needs when the available theorem is stated over an algebraically closed field; it asks nothing of the structure map, so it also covers reduction ℤ → 𝔽ₚ. The ascent direction transports the property up, which is what a base-changed Lie algebra needs before Mathlib's IsKilling machinery applies to it, and there injectivity is genuinely needed.

The mathematical content is entirely in the Gram matrix. Mathlib's LieModule.traceForm_baseChange already says that the Killing form of A ⊗[R] L is the base change of the Killing form of L, and Ado.nondegenerate_of_nondegenerate_baseChange and Ado.nondegenerate_baseChange_iff say how a base change of integral domains reflects and preserves nondegeneracy of a bilinear form on a finite free module. What remains is to read IsKilling as nondegeneracy of the Killing form in both directions; Ado.isKilling_of_killingForm_nondegenerate is the direction Mathlib does not already provide.

Nothing here needs the coefficients to be a field, and no characteristic hypothesis appears: the counterexamples to HasTrivialRadical → IsKilling in positive characteristic concern the passage between semisimplicity and the Killing property, not the base change of the Killing form itself.

Main declarations #

Roadmap #

This is a prerequisite for Layer 9 of TauCetiRoadmap/ReductiveGroups/README.md, "The Chevalley--Demazure construction. For each root datum, an explicitly constructed split reductive group scheme over ℤ realizing it, via a Chevalley basis and the Kostant ℤ-form of the enveloping algebra", which is consumed by milestone L0 of TauCetiRoadmap/CFSGStatement/README.md. The declaration that will consume it is Ado.IsChevalleySystem, whose ambient Lie algebra carries [LieAlgebra.IsKilling K L], applied to Ado.DynkinType.lieAlgebra. That Lie algebra is defined over ℚ, and TauCeti/LinearAlgebra/RootSystem/SimplyConnectedRootDatum/LieAlgebra.lean records that it "is not asserted to be semisimple: Mathlib derives that from Geck's construction over an algebraically closed field, and ℚ is not one", Mathlib's RootPairing.GeckConstruction.instHasTrivialRadical carrying an [IsAlgClosed K] hypothesis. This file supplies the general half of closing that gap; the remaining half is the identification of AlgebraicClosure ℚ ⊗[ℚ] Ado.DynkinType.lieAlgebra with Geck's Lie algebra over AlgebraicClosure ℚ.

References #

Descent of the Killing property. A finite free Lie algebra over an integral domain is Killing as soon as some base change of it into an integral domain is; the structure map is unrestricted, so this covers reduction of an integral form as well as extension of scalars.

@[simp]

The Killing property under an injective base change. For a finite free Lie algebra over an integral domain, and an injective structure map into a second integral domain, the base change is Killing exactly when the original is.

The intended application is a field extension, where every hypothesis above is automatic.