Chapter IX. Centralizers of Tori

FC16 — Humphreys, Linear Algebraic Groups.

This chapter contains 8 definitions, 4 theorems, 6 propositions, 3 lemmas, 8 corollaries, 4 constructions and 1 other statements.

Nobody has checked this chapter against Lean yet, so the statements below say nothing about whether they are formalized.

Definition: regular and singular tori

FC16-C09-U001 · §24 opening; p. 147 · after FC16-C08-U032, FC16-C06-U016

For a torus S, let 𝔅^S be the Borel subgroups containing S. S is regular when 𝔅^S is finite and singular otherwise.

Definition: Weyl group relative to a torus

FC16-C09-U002 · §24.1; p. 147 · after FC16-C06-U025, FC16-C02-U023, FC16-C08-U006

W(G,S)=N_G(S)/C_G(S), finite by torus rigidity. For maximal tori these groups are all isomorphic and are called the Weyl group of G.

Lemma: centralizer of a maximal torus lies in every containing Borel

FC16-C09-U003 · §24.1; p. 147 · after FC16-C08-U021, FC16-C08-U024

If T is maximal and C=C_G(T), then C is contained in every Borel subgroup containing T.

Proposition: simple transitivity of Weyl group on Borel subgroups through T

FC16-C09-U004 · §24.1; pp. 147–148 · after FC16-C09-U002, FC16-C09-U003, FC16-C08-U027

For maximal T, W(G,T) acts simply transitively on 𝔅^T; hence |𝔅^T|=|W| and T is regular.

Proposition: Weyl groups and Borel sets under epimorphisms

FC16-C09-U005 · §24.1; p. 148 · after FC16-C08-U011, FC16-C09-U004

An epimorphism G→G′ with T↦T′ induces surjections 𝔅T→𝔅{T′} and W(G,T)→W(G′,T′), bijective if the kernel lies in every Borel subgroup.

Construction: Weyl group action on characters, cocharacters and roots

FC16-C09-U006 · §24.1; pp. 148–149 · after FC16-C09-U002, FC16-C06-U021, FC16-C06-U027

N_G(T) acts on X(T),Y(T), descends to W, preserves the pairing, and permutes roots/root spaces by w·α(t)=α(n^{-1}tn), Ad n(g_α)=g_{wα}.

Proposition: regular torus criterion by solvability of its centralizer

FC16-C09-U007 · §24.2; p. 149 · after FC16-C09-U001, FC16-C08-U020, FC16-C08-U024

A torus S is regular iff C_G(S) is solvable. Then C_G(S) lies in every Borel containing S, and for every maximal T⊇S one has 𝔅S=𝔅T.

Definition/construction: intersection subgroup I(T) and exterior roots

FC16-C09-U008 · §24.3; pp. 149–150 · after FC16-C09-U003, FC16-C06-U026, FC16-C08-U032

For maximal T, I(T) is the identity component of the intersection of all Borel subgroups containing T. Splitting g=Lie(I(T))⊕⊕_{α∈Ψ}g_α defines the finite set Ψ of root weights outside Lie(I(T)).

Definition: root-kernel torus

FC16-C09-U009 · §24.3; p. 150 · after FC16-C09-U008, FC16-C06-U012

For α∈Ψ, T_α=(Ker α)° is a codimension-one subtorus of T.

Proposition: singular subtori detected by roots

FC16-C09-U010 · §24.3; p. 150 · after FC16-C09-U007, FC16-C09-U009, FC16-C07-U021

For S⊆T, S is singular iff S is contained in T_α for some α∈Ψ; in particular the maximal singular subtori are the T_α.

Corollary: rank-one semisimple quotient of a root centralizer

FC16-C09-U011 · §24.3; p. 150 · after FC16-C09-U010, FC16-C07-U030, FC16-C08-U007

For α∈Ψ and Z_α=C_G(T_α), the quotient Z_α/R(Z_α) is semisimple of rank one.

Definition: regular cocharacter

FC16-C09-U012 · §24.4; pp. 150–151 · after FC16-C06-U021, FC16-C09-U001

A one-parameter subgroup λ∈Y(T) is regular when λ(G_m) is a regular torus; Y(T)_reg denotes the regular cocharacters.

Proposition: root pairing criterion for regular cocharacters

FC16-C09-U013 · §24.4; p. 151 · after FC16-C09-U012, FC16-C09-U010, FC16-C06-U021

λ∈Y(T) is regular iff ⟨α,λ⟩≠0 for every α∈Ψ.

Construction/result: extension of a cocharacter orbit in projective space

FC16-C09-U014 · §25.1; pp. 151–153 · after FC16-C06-U026, FC16-C01-U092, FC16-C06-U021

For a torus representation with weights χ_i and λ∈Y(T), the orbit map t↦λ(t)[v] extends to P¹; the 0 and ∞ limits are obtained by retaining the minimum and maximum λ-weight components and are λ(G_m)-fixed.

Result: cocharacters with separated weights have the same projective fixed points as T

FC16-C09-U015 · §25.1; p. 153 · after FC16-C09-U014, FC16-C06-U019

Choosing λ so that distinct T-weights have distinct pairings makes the fixed points of λ(G_m) in P(V) coincide with the T-fixed points.

Lemma: hyperplane section of a projective variety

FC16-C09-U016 · §25.2; p. 153 · after FC16-C01-U092, FC16-C01-U061

If irreducible projective X⊄P(W) with W codimension one and dim X>0, then X∩P(W) is nonempty and every irreducible component has codimension one.

Theorem: fixed points of a torus on an invariant projective variety

FC16-C09-U017 · §25.2; pp. 153–154 · after FC16-C09-U015, FC16-C09-U016, FC16-C09-U014

For torus T acting on irreducible closed X⊆P(V): if dim X≥1 the source obtains at least two T-fixed points; if dim X≥2 it obtains at least three.

Corollary: torus fixed points on G/P

FC16-C09-U018 · §25.2; p. 154 · after FC16-C09-U017, FC16-C08-U008, FC16-C08-U032

For proper parabolic P and torus T, T fixes at least two points of G/P, and at least three when dim G/P>1; a maximal torus therefore lies in at least two/three corresponding Borel subgroups.

Corollary: lower bounds for the Weyl group from dim G/B

FC16-C09-U019 · §25.2; p. 154 · after FC16-C09-U018, FC16-C09-U004

By simple transitivity, dim G/B=1 forces |W|=2, while higher-dimensional cases give the corresponding ≥3 lower bound recorded in the source.

Corollary: Borel subgroups through a maximal torus generate G

FC16-C09-U020 · §25.2; p. 154 · after FC16-C09-U018, FC16-C08-U009, FC16-C08-U030

For maximal T, G is generated by all Borel subgroups containing T.

Definition: semisimple and reductive rank

FC16-C09-U021 · §25.3; p. 154 · after FC16-C08-U007, FC16-C07-U029

rank_s G=rank(G/R(G)) and rank_red G=rank(G/R_u(G)).

Theorem: equivalent descriptions of semisimple rank one

FC16-C09-U022 · §25.3; pp. 154–155 · after FC16-C09-U021, FC16-C09-U019, FC16-C01-U094, FC16-C08-U032

For connected G and maximal T, rank_sG=1 iff |W|=2 iff |𝔅^T|=2 iff dim G/B=1 iff G/B≅P¹ iff there is an epimorphism G→PGL_2 with kernel identity component R(G).

Corollary: structure of reductive groups of semisimple rank one

FC16-C09-U023 · §25.3; pp. 155–156 · after FC16-C09-U022, FC16-C07-U031, FC16-C07-U004

If G is reductive of semisimple rank one, (G,G) is 3-dimensional semisimple, G=(G,G)Z(G) with finite intersection, C_G(T)=T, and any epimorphism to PGL_2 has kernel Z(G).

Construction: rank-one Weyl reflection attached to a singular torus

FC16-C09-U024 · §25.4; p. 156 · after FC16-C09-U011, FC16-C09-U022, FC16-C09-U002

For α∈Ψ, Z_α=C_G(T_α) has Weyl group of order two; its nontrivial element s_α embeds canonically in W(G,T) and exchanges its two Borel subgroups through T.

Construction: Borel subgroup B(λ) from a regular cocharacter

FC16-C09-U025 · §25.4; pp. 156–157 · after FC16-C09-U014, FC16-C09-U012, FC16-C08-U032, FC16-C09-U006

A regular λ∈Y(T) selects a canonical T-fixed point on G/B by the ∞-limit construction, hence a Borel subgroup B(λ) containing T; the assignment is W-equivariant and onto 𝔅^T.

Definition: Weyl chamber

FC16-C09-U026 · §25.4; p. 157 · after FC16-C09-U025, FC16-C09-U004

For B∈𝔅^T, C(B)={λ∈Y(T)_reg | B(λ)=B}; these nonempty sets partition the regular cocharacters and are permuted simply transitively by W.

Proposition: root sign detects the Borel in a rank-one centralizer

FC16-C09-U027 · §25.4; p. 157 · after FC16-C09-U024, FC16-C09-U025, FC16-C09-U026

For α∈Ψ, the two Borels B_α,B′_α of Z_α through T can be labeled so that ⟨α,λ⟩>0 iff B(λ)∩Z_α=B_α.

Lemma: regular cocharacter separates nonproportional roots

FC16-C09-U028 · §26.1; p. 158 · after FC16-C09-U013, FC16-C06-U021

For nonproportional α,β∈Ψ there is λ∈Y(T)_reg with ⟨α,λ⟩>0 and ⟨β,λ⟩<0.

Theorem: unipotent radical from intersection of Borels through T

FC16-C09-U029 · §26.1; pp. 158–159 · after FC16-C09-U008, FC16-C09-U028, FC16-C09-U020, FC16-C07-U029

If I(T) is the identity component of the intersection of B∈𝔅^T, then R_u(G) equals the unipotent part I(T)_u.

Corollary: torus centralizers in reductive groups

FC16-C09-U030 · §26.2; pp. 159–160 · after FC16-C09-U029, FC16-C07-U030, FC16-C09-U007

If G is reductive and S⊆T is a torus, C_G(S) is reductive. If S is regular then C_G(S)=T; hence Cartan subgroups are maximal tori and Z(G)⊆T.

Corollary: root-space, root-centralizer, center and root-lattice structure of a reductive group

FC16-C09-U031 · §26.2; p. 160 · after FC16-C09-U030, FC16-C09-U008, FC16-C09-U010, FC16-C09-U023, FC16-C09-U020, FC16-C09-U021

For reductive G: Ψ=Φ(G,T) and Φ=−Φ; g=t⊕⊕{α∈Φ}g_α with dim g_α=1; the codimension-one singular tori are T_α; each Z_α=C_G(T_α) is reductive of semisimple rank one with Lie algebra t⊕g_α⊕g{−α}, and the Z_α generate G; Z(G)=∩_{α∈Φ}Ker α; and the root lattice in X(T) has rank rank_sG.

Corollary/definition: opposite Borel subgroup

FC16-C09-U032 · §26.2; p. 160 · after FC16-C09-U025, FC16-C09-U031, FC16-C09-U027

For reductive G, every B∈𝔅^T has a unique opposite B^- with B∩B^-=T; writing B=TU and B-=TU- gives g=u^-⊕t⊕u.

Theorem/definition: root subgroups

FC16-C09-U033 · §26.3; pp. 160–162 · after FC16-C09-U031, FC16-C07-U042, FC16-C09-U006

For each α∈Φ there is a unique connected T-stable one-dimensional unipotent subgroup U_α with Lie algebra g_α; W-conjugation sends U_α to U_{wα}, and there is an admissible isomorphism x_α:G_a→U_α satisfying t x_α(c)t^{-1}=x_α(α(t)c). Together T and all U_α generate G.

Definition: admissible root-group parametrization

FC16-C09-U034 · §26.3; p. 162 · after FC16-C09-U033

An isomorphism x_α:G_a→U_α satisfying the T-equivariance formula is called admissible; it is unique up to nonzero scalar.