Chapter X. Structure of Reductive Groups

FC16 — Humphreys, Linear Algebraic Groups.

This chapter contains 11 definitions, 13 theorems, 5 propositions, 7 lemmas, 7 corollaries, 3 constructions and 3 other statements.

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

Definition: abstract root system and abstract Weyl group

FC16-C10-U001 · §27.1; p. 163 · after FC16-C09-U031

A finite spanning subset Φ of a real vector space E, not containing 0, is a root system when only ±α are scalar multiples in Φ, each root has an associated reflection preserving Φ, and every reflection difference s_α(β)−β is an integral multiple of α. W(Φ) is generated by the root reflections.

Theorem: root system of a semisimple algebraic group

FC16-C10-U002 · §27.1–27.2; pp. 163–165 · after FC16-C10-U001, FC16-C09-U031, FC16-C09-U006

For semisimple G and E=R⊗_Z X(T), Φ(G,T) is an abstract root system of rank rank G, and its abstract Weyl group identifies with W(G,T).

Proposition: root subgroup shifts weights along root strings

FC16-C10-U003 · §27.2; p. 165 · after FC16-C09-U033, FC16-C06-U026, FC16-C10-U002

For a rational representation ρ:G→GL(V) with kernel contained in Z(G), if V_λ is a T-weight space then ρ(U_α)V_λ lies in the sum of V_{λ+kα} over integral k.

Definition/result: base, simple roots and positive roots

FC16-C10-U004 · §27.3; pp. 165–166 · after FC16-C10-U002

A base Δ is a basis of the ambient root space such that every root has a unique integral expansion with coefficients all of one sign; its elements are simple roots and determine the positive roots. Bases exist and W acts simply transitively on them.

Comparison: bases, abstract chambers, cocharacter chambers and Borel subgroups

FC16-C10-U005 · §27.3; p. 166 · after FC16-C10-U004, FC16-C09-U026, FC16-C09-U004

For fixed maximal T, bases of Φ, Weyl chambers in the root and cocharacter spaces, and Borel subgroups containing T correspond naturally and are each permuted simply transitively by W.

Theorem: generation by simple root rank-one subgroups

FC16-C10-U006 · §27.3; p. 166 · after FC16-C10-U004, FC16-C09-U031, FC16-C09-U033

For a base Δ, G is generated by Z_α for α∈Δ, equivalently by T together with U_{±α} for simple roots α.

Definition: diagram-stabilizing automorphism subgroup

FC16-C10-U007 · §27.4; p. 166 · after FC16-C10-U004, FC16-C08-U004, FC16-C08-U006

For semisimple G with fixed T⊆B, D≤Aut G consists of automorphisms stabilizing T and B; each induces an automorphism of the based root system, hence a Dynkin-diagram automorphism.

Theorem: automorphisms of a semisimple group modulo inner automorphisms

FC16-C10-U008 · §27.4; pp. 166–167 · after FC16-C10-U007, FC16-C10-U006, FC16-C08-U005

Aut G=(Int G)D, and the natural map from D to diagram automorphisms induces an injection Aut G/Int G into the finite diagram-automorphism group.

Theorem: minimal normal factors of a semisimple group

FC16-C10-U009 · §27.5; pp. 167–168 · after FC16-C05-U012, FC16-C07-U004, FC16-C10-U008, FC16-C10-U006

The minimal positive-dimensional closed connected normal subgroups G_i of semisimple G are finite in number, centralize one another, multiply onto G with finite kernel, and every closed connected normal subgroup is the product of the G_i it contains; moreover G=(G,G) and is generated by its root groups.

Corollary: simple factors and irreducible root-system components

FC16-C10-U010 · §27.5; p. 168 · after FC16-C10-U009, FC16-C10-U002

The almost-direct decomposition of semisimple G into its minimal connected normal factors corresponds exactly to the decomposition of Φ into irreducible root systems.

Definition: simple/almost-simple algebraic group

FC16-C10-U011 · §27.5; p. 168 · after FC16-C10-U009

A noncommutative connected algebraic group with no proper nontrivial closed connected normal subgroup is called simple, or almost simple when emphasizing that its abstract group may have a finite center.

Corollary/result: reductive group as center times derived group

FC16-C10-U012 · §27.5; p. 168 · after FC16-C10-U009, FC16-C07-U031

A reductive G is the product Z(G)°·(G,G), with finite intersection; its semisimple derived part decomposes according to the irreducible components of Φ.

Definition: directly spanned by closed subgroups

FC16-C10-U013 · §28.1; pp. 169–170 · after FC16-C01-U045

H is directly spanned by H_1,…,H_r in the given order when the product morphism H_1×⋯×H_r→H is bijective.

Proposition: closed T-stable subgroups of a Borel unipotent radical

FC16-C10-U014 · §28.1; pp. 169–171 · after FC16-C09-U033, FC16-C10-U013, FC16-C09-U032

If H is a closed T-stable subgroup of U=R_u(B), then H is connected and is directly spanned, in any order, by exactly the root subgroups U_α whose root spaces lie in Lie(H).

Construction/result: root-sign factorization U=Uw Uw^+

FC16-C10-U015 · §28.1; p. 171 · after FC16-C10-U014, FC16-C10-U002

For w∈W, the positive roots split according to the sign of w^{-1}α; the corresponding T-stable root-subgroup products give complementary directly spanning subgroups of U, with U equal to their product in either order.

Proposition: Bruhat decomposition in semisimple rank one

FC16-C10-U016 · §28.2; pp. 171–172 · after FC16-C09-U022, FC16-C10-U015

If G has semisimple rank one and s is the nontrivial Weyl element, then G=B ⊔ BsB. After fixing a representative n(s), every element of the nontrivial cell has a unique factorization u n(s)t u′ with the indicated rank-one unipotent factors and t∈T.

Theorem: Bruhat decomposition

FC16-C10-U017 · §28.3; pp. 172–173 · after FC16-C10-U006, FC16-C10-U016, FC16-C09-U002

For reductive G, G is the disjoint union ⨆_{w∈W}BwB, and two B-double cosets are equal iff their Weyl elements are equal.

Corollary: multiplication of a Bruhat cell by a simple reflection

FC16-C10-U018 · §28.3; p. 173 · after FC16-C10-U017, FC16-C10-U016

For simple reflection s and w∈W, left multiplication of BwB by the rank-one cell BsB is contained in the union BwB∪BswB.

Corollary: intersection of two Borel subgroups

FC16-C10-U019 · §28.3; p. 173 · after FC16-C10-U017, FC16-C10-U014, FC16-C08-U005

The intersection of any two Borel subgroups is connected and contains a maximal torus.

Theorem: unique Bruhat normal form

FC16-C10-U020 · §28.4; p. 173 · after FC16-C10-U017, FC16-C10-U015

After choosing representatives n(w)∈N, every x∈G has a unique expression x=u′n(w)tu with w∈W, t∈T, u∈U and u′ in the root-sign subgroup U_w determined by w.

Proposition: big cell

FC16-C10-U021 · §28.5; p. 174 · after FC16-C09-U032, FC16-C10-U017, FC16-C10-U020

Multiplication U^-×B→G is a bijection onto an open dense subset Ω, the big cell.

Result: big-cell product is an isomorphism of varieties

FC16-C10-U022 · §28.5; p. 174 · after FC16-C10-U021, FC16-C01-U088, FC16-C01-U076

The product U^-×B→Ω is separable and birational and, using its equivariance, is an isomorphism of varieties.

Recorded result: ordered products of positive root groups

FC16-C10-U023 · §28.5; p. 174 · after FC16-C10-U014, FC16-C10-U022

For any ordering α_1,…,α_r of the positive roots, multiplication U_{α_1}×⋯×U_{α_r}→U is an isomorphism of varieties.

Definition: Tits system (BN-pair)

FC16-C10-U024 · §29.1; p. 176 · after FC16-C10-U017, FC16-C10-U018

A Tits system (G,B,N,S) has G generated by B,N; T=B∩N normal in N; W=N/T generated by involutions S; and axioms (T1) sBw⊆BwB∪BswB and (T2) sBs is not contained in B. W is its Weyl group and |S| its rank.

Definition: length and standard subsets in a Tits system

FC16-C10-U025 · §29.2; p. 177 · after FC16-C10-U024

Length ℓ(w) is minimal word length in S; for I⊆S let W_I=⟨I⟩ and P_I=BW_IB.

Theorem: Bruhat decomposition for a Tits system

FC16-C10-U026 · §29.2; p. 177 · after FC16-C10-U024, FC16-C10-U025

For every I⊆S, P_I is a subgroup; in particular P_S=G. The double cosets BwB are pairwise distinct.

Lemma: length change under a simple generator

FC16-C10-U027 · §29.3; pp. 177–178 · after FC16-C10-U026, FC16-C10-U025

For s∈S,w∈W, ℓ(sw)=ℓ(w)±1. The sign determines whether sBw lies in BswB alone or meets BwB, giving the basic double-coset length calculus.

Lemma: generation of PI from a reduced word

FC16-C10-U028 · §29.3; p. 178 · after FC16-C10-U027, FC16-C10-U026

If a reduced expression for w uses simple generators in I, then P_I is generated by B together with wBw^{-1}, equivalently by B and representatives of those simple generators.

Lemma: intrinsic characterization of simple reflections

FC16-C10-U029 · §29.3; p. 178 · after FC16-C10-U028, FC16-C10-U026

S is exactly the set of w for which B∪BwB is a subgroup; consequently S is a minimal generating set of W.

Lemma: conjugate-standard-parabolic containment

FC16-C10-U030 · §29.3; pp. 178–179 · after FC16-C10-U028, FC16-C10-U026

If wP_Iw^{-1}⊆P_J, then every representative of w in N lies in P_J.

Theorem: parabolic subgroups of a Tits system

FC16-C10-U031 · §29.3; p. 179 · after FC16-C10-U030, FC16-C10-U029

The subgroups containing B are exactly P_I. Standard parabolics are conjugate only when equal, are self-normalizing, and inclusions P_I⊆P_J and W_I⊆W_J are equivalent to I⊆J.

Definition: Coxeter group presentation

FC16-C10-U032 · §29.4; pp. 179–180 · after FC16-C10-U025

A Coxeter group has generators s_i and defining relations (s_is_j)^{m(i,j)}=1, with m(i,i)=1 and symmetric m(i,j)≥2 or ∞.

Lemma: exchange condition

FC16-C10-U033 · §29.4; p. 180 · after FC16-C10-U027, FC16-C10-U032

If w=s_1⋯s_r is reduced and ℓ(sw)<ℓ(w), then sw is obtained by deleting one generator from the reduced word.

Theorem: Weyl group of a Tits system is a Coxeter group

FC16-C10-U034 · §29.4; pp. 180–181 · after FC16-C10-U033, FC16-C10-U032

If m(i,j) is the order of s_is_j in W, the canonical epimorphism from the Coxeter group on S with these exponents to W is an isomorphism.

Lemma: normal subgroups and decompositions of the Coxeter generators

FC16-C10-U035 · §29.5; p. 181 · after FC16-C10-U031, FC16-C10-U034

For a normal subgroup H◁G there is a partition S=I⊔J with I and J commuting elementwise and HB=P_I.

Definition: irreducible Weyl group of a Tits system

FC16-C10-U036 · §29.5; p. 181 · after FC16-C10-U035

W is irreducible when S has no nontrivial partition into two elementwise commuting subsets.

Theorem: Tits simplicity criterion

FC16-C10-U037 · §29.5; p. 182 · after FC16-C10-U036, FC16-C10-U035

If W is irreducible, G is generated by conjugates of a normal solvable U≤B, and G=(G,G), then G/Z is simple or trivial, where Z is the intersection of all conjugates of B.

Corollary: abstract simplicity modulo center of an almost-simple algebraic group

FC16-C10-U038 · §29.5; p. 182 · after FC16-C10-U037, FC16-C10-U011, FC16-C10-U009

If G is simple/almost simple as an algebraic group and Z its finite center, then the abstract group G/Z is simple.

Definition: standard parabolic subgroup

FC16-C10-U039 · §30.1; p. 183 · after FC16-C08-U008, FC16-C10-U031, FC16-C10-U004

Relative to B and base Δ, standard parabolics are the parabolic subgroups containing B, indexed by I⊆Δ as P_I=BW_IB.

Theorem: classification and roots of parabolic subgroups

FC16-C10-U040 · §30.1; p. 184 · after FC16-C10-U039, FC16-C10-U002

Every parabolic is conjugate to a unique standard P_I; its roots relative to T are all positive roots together with the negative roots in the root subsystem spanned by I.

Definition: Levi decomposition and Levi factor

FC16-C10-U041 · §30.2; pp. 184–185 · after FC16-C07-U029, FC16-C07-U030

For P with V=R_u(P), a factorization P=L⋉V with L reductive is a Levi decomposition and L is a Levi factor.

Construction: canonical Levi factor of a standard parabolic

FC16-C10-U042 · §30.2; p. 185 · after FC16-C10-U040, FC16-C09-U030, FC16-C10-U041

For standard P_I, if Z is the connected intersection of the kernels of roots in I, then L=C_G(Z) is reductive, has the root subsystem spanned by I, and P_I=L⋉R_u(P_I).

Theorem: existence and conjugacy of Levi factors

FC16-C10-U043 · §30.2; p. 185 · after FC16-C10-U042, FC16-C10-U041

Every parabolic subgroup has a Levi decomposition, and any two Levi factors are conjugate by an element of its unipotent radical.

Construction: iterated unipotent-normalizer sequence P(U)

FC16-C10-U044 · §30.3; p. 185 · after FC16-C07-U011, FC16-C07-U029, FC16-C02-U023

For closed unipotent U, repeatedly enlarge U_i by R_u(N_G(U_i)) and take the next normalizer until stabilization; the terminal normalizer is denoted P(U).

Lemma: connected solvable normalizer lies in a Borel containing the normalized subset

FC16-C10-U045 · §30.3; p. 185 · after FC16-C08-U003, FC16-C08-U004

If A⊆B for a Borel B of H and connected solvable S≤H normalizes A, then S and A are contained in some common Borel subgroup of H.

Proposition: parabolicity criterion for a unipotent normalizer

FC16-C10-U046 · §30.3; p. 186 · after FC16-C10-U044, FC16-C10-U045, FC16-C10-U019

If closed unipotent V lies in a Borel subgroup, N=N_G(V), and V contains R_u(N) with the stabilized identity-component condition of the construction, then N is parabolic and V=R_u(N).

Corollary: parabolic associated to a closed unipotent subgroup

FC16-C10-U047 · §30.3; p. 186 · after FC16-C10-U046, FC16-C10-U044

If closed unipotent U lies in a Borel, then P(U) is parabolic, N_G(U)⊆P(U), and the terminal unipotent group is R_u(P(U)).

Corollary: self-unipotent-radical normalizer is parabolic

FC16-C10-U048 · §30.3; p. 186 · after FC16-C10-U046

If connected unipotent U satisfies U=R_u(N_G(U)), then N_G(U) is parabolic.

Theorem: maximal subgroups and maximal unipotent subgroups of a reductive group

FC16-C10-U049 · §30.4; p. 187 · after FC16-C10-U048, FC16-C10-U047, FC16-C08-U008, FC16-C08-U004

If H is a maximal proper closed subgroup of reductive G, either H° is reductive or H is parabolic. Every unipotent subgroup lies in a Borel subgroup; maximal unipotent subgroups are exactly Borel unipotent radicals.