Documentation

TauCeti.Algebra.Lie.Weights.Chevalley.Automorphism

What an automorphism does to a Chevalley system #

Let x be a family of root vectors normalised against the coroots, and let σ be an automorphism of L normalising the splitting Cartan subalgebra H. Then σ permutes the roots, and since the root spaces are lines it carries x α to a nonzero multiple of x (σ α):

σ (x α) = ε α • x (σ α).

Two constraints pin the scalars. The normalisation ⁅x α, x (-α)⁆ = α^∨ is preserved by σ, because σ sends α^∨ to (σ α)^∨, and this forces ε α * ε (-α) = 1. If σ moreover commutes with the Chevalley involution ω of the system, then applying both sides of ω (x α) = -x (-α) gives ε α = ε (-α). Together ε α ^ 2 = 1, so every scalar is 1 or -1.

That conclusion is the general-root form of a diagram automorphism of a split semisimple Lie algebra. A graph symmetry of the Dynkin diagram is pinned by its action on the simple root vectors, where the parameter is unchanged; on a general root vector a sign appears, and the signs cannot all be normalised away. The type-A graph automorphism X ↦ -J Xᵀ J⁻¹ of sl n already exhibits them. What is proved here is that a sign is all that appears.

The hypothesis that σ commutes with ω is not automatic and is not a weakening: the Chevalley involution attached to a normalised family is unique (TauCeti.IsChevalleySystem.unique), but σ moves the family, so nothing forces the two automorphisms to commute. For the diagram automorphism of a Serre presentation the commutation is visible on the generators, and it is available as TauCeti.GraphTwistedIndex.serreChevalleyInvolution_comm_serreGraphAut.

Nothing here identifies the signs, and nothing here constructs an automorphism: the input is an automorphism already in hand.

The transport of the underlying normalised family lives one file up, in TauCeti/Algebra/Lie/Weights/Automorphism.lean; only the consequences of the Chevalley involution are proved here. Throughout this file map_ is reserved, as elsewhere in the TauCeti.IsChevalleySystem namespace, for the involution ω, and the automorphism σ is called aut.

Main results #

References #

This is the general-root sign statement that milestone L1 of TauCetiRoadmap/CFSGStatement/README.md asks to record as a consequence of the pinned construction, and it supplies part of the diagram-automorphism data for Layer 9 of TauCetiRoadmap/ReductiveGroups/README.md.

theorem TauCeti.IsChevalleySystem.eq_of_aut_apply_eq_smul {K : Type u} {L : Type v} [Field K] [CharZero K] [LieRing L] [LieAlgebra K L] [LieAlgebra.IsKilling K L] [FiniteDimensional K L] {H : LieSubalgebra K L} [H.IsCartanSubalgebra] [LieModule.IsTriangularizable K (↥H) L] (σ : L ≃ₗ⁅K⁆ L) (hσ : LieSubalgebra.map σ.toLieHom H = H) {ω : L ≃ₗ⁅K⁆ L} {x : LieModule.Weight K (↥H) L → L} (hx : IsChevalleySystem ω x) (hcomm : ∀ (z : L), σ (ω z) = ω (σ z)) {α : LieModule.Weight K (↥H) L} (hα : α.IsNonZero) {c d : K} (hc : σ (x α) = c • x ((weightPerm σ hσ) α)) (hd : σ (x (-α)) = d • x (-(weightPerm σ hσ) α)) :
c = d

If a normalising automorphism commutes with the Chevalley involution of the system, the scalars it produces at α and at -α agree. Applying σ to ω (x α) = -x (-α) and ω to σ (x α) = c • x (σ α) gives two expressions for the same vector.

theorem TauCeti.IsChevalleySystem.aut_apply_eq_or_eq_neg {K : Type u} {L : Type v} [Field K] [CharZero K] [LieRing L] [LieAlgebra K L] [LieAlgebra.IsKilling K L] [FiniteDimensional K L] {H : LieSubalgebra K L} [H.IsCartanSubalgebra] [LieModule.IsTriangularizable K (↥H) L] (σ : L ≃ₗ⁅K⁆ L) (hσ : LieSubalgebra.map σ.toLieHom H = H) {ω : L ≃ₗ⁅K⁆ L} {x : LieModule.Weight K (↥H) L → L} (hx : IsChevalleySystem ω x) (hcomm : ∀ (z : L), σ (ω z) = ω (σ z)) (α : LieModule.Weight K (↥H) L) :
σ (x α) = x ((weightPerm σ hσ) α) ∨ σ (x α) = -x ((weightPerm σ hσ) α)

A normalising automorphism commuting with the Chevalley involution moves each root vector to the root vector of the permuted root, up to sign.

The two scalars attached to α and -α multiply to one and are equal, so each is a square root of one in a field. No sign is claimed for any particular root: the type-A graph automorphism already shows that they cannot all be made +1.

theorem TauCeti.IsChevalleySystem.exists_sign_aut_apply {K : Type u} {L : Type v} [Field K] [CharZero K] [LieRing L] [LieAlgebra K L] [LieAlgebra.IsKilling K L] [FiniteDimensional K L] {H : LieSubalgebra K L} [H.IsCartanSubalgebra] [LieModule.IsTriangularizable K (↥H) L] (σ : L ≃ₗ⁅K⁆ L) (hσ : LieSubalgebra.map σ.toLieHom H = H) {ω : L ≃ₗ⁅K⁆ L} {x : LieModule.Weight K (↥H) L → L} (hx : IsChevalleySystem ω x) (hcomm : ∀ (z : L), σ (ω z) = ω (σ z)) :
∃ (ε : LieModule.Weight K (↥H) L → K), (∀ (α : LieModule.Weight K (↥H) L), ε α = 1 ∨ ε α = -1) ∧ ∀ (α : LieModule.Weight K (↥H) L), σ (x α) = ε α • x ((weightPerm σ hσ) α)

The signs of TauCeti.IsChevalleySystem.aut_apply_eq_or_eq_neg, collected into one function. This is the form γ (x_α) = ε_α • x_{γ α} in which a graph automorphism of a pinned group is usually stated.