Bourbaki

Exercise 3, § 2

Lie Groups and Lie Algebras, IV

Let G be a group, B a subgroup of G and (C(w))_{w∈W} the family of double cosets of G with respect to B. Then B is called a Tits subgroup of G if there exists a subset S of W such that the following conditions are satisfied:

(1) the union of the C(s) for s ∈ S generates G; (2) for all s ∈ S, the set B ∪ C(s) is a subgroup of G and the index of B in B ∪ C(s) is ≥ 3; (3) for all s ∈ S and all w ∈ W, there exists an element w' ∈ W such that C(s).C(w) ⊂ C(w) ∪ C(w').

From now on we assume that B is a Tits subgroup of G and that we are given a subset S of W satisfying conditions (1), (2) and (3).

a) Show that C(s)^{-1} = C(s) and that C(s).C(s) = B ∪ C(s) for all s ∈ S. Show that, for all s ∈ S and all w ∈ W, there exists an element w'' ∈ W such that C(w).C(s) ⊂ C(w) ∪ C(w'').

b) If w ∈ W, the length of w, denoted by l(w), is the smallest integer n ≥ 0 for which there exist s_1, ..., s_n ∈ S with C(w) ⊂ C(s_1)...C(s_n). Show that l(w) is finite for all w ∈ W.

Let u, v ∈ W with l(u) < l(v) and let s ∈ S. Show that, if C(v) ⊂ C(u).C(s) (resp. C(v) ⊂ C(s).C(u)), then C(v) = C(u).C(s) (resp, C(v) = C(s).C(u)) (argue by induction on the length of u. If C(v) ≠ C(u).C(s), then C(u).C(s) = C(u) ∪ C(v). By using the induction hypothesis, show that there exist t ∈ S and w ∈ W such that

C(u)=C(t).C(w)withl(w)=l(u)−1. C(u) = C(t).C(w) \quad \text{with} \quad l(w) = l(u) - 1.

From the relations C(v) ⊂ C(u).C(s) = C(t).C(w).C(s) and C(t).C(w) = C(u) ≠ C(v), deduce the existence of an element w' ≠ w such that

C(w′)⊂C(w).C(s)C(w') \subset C(w).C(s) and C(v)⊂C(t).C(w′)C(v) \subset C(t).C(w'), so l(w′)≥l(v)−1>l(u)−1=l(w)l(w') \geq l(v) - 1 > l(u) - 1 = l(w). The induction hypothesis now implies that

C(w′)=C(w).C(s). C(w') = C(w).C(s).

Moreover,

C(t).C(u).C(s)=C(t).C(t).C(w).C(s)=C(t).C(w).C(s)∪C(w).C(s)=C(u).C(s)∪C(w′)=C(u)∪C(v)∪C(w′) \begin{align*} C(t).C(u).C(s) &= C(t).C(t).C(w).C(s) = C(t).C(w).C(s) \cup C(w).C(s) \\ &= C(u).C(s) \cup C(w') = C(u) \cup C(v) \cup C(w') \end{align*}

and also

C(t).C(u).C(s)=C(t).(C(u)∪C(v))⊃C(t).C(t).C(w)⊃C(w), C(t).C(u).C(s) = C(t).(C(u) \cup C(v)) \supset C(t).C(t).C(w) \supset C(w),

which is a contradiction since w≠u,v,w′w \neq u, v, w'.

c) Show that, for all w∈Ww \in W and all s∈Ss \in S, there exists a unique element, denoted by s.ws.w (resp, w∗sw * s), distinct from ww and such that

C(s.w)⊂C(s).C(w)⊂C(w)∪C(s.w)(resp. C(w∗s)⊂C(w).C(s)⊂C(w)∪C(w∗s)). \begin{align*} C(s.w) &\subset C(s).C(w) \subset C(w) \cup C(s.w) \\ (\text{resp. } C(w * s) &\subset C(w).C(s) \subset C(w) \cup C(w * s)). \end{align*}

(Show by induction on l(w)l(w) that C(s).C(w)≠C(w)C(s).C(w) \neq C(w). For this, write

C(w)=C(u).C(t) C(w) = C(u).C(t)

with t∈S,u∈Wt \in S, u \in W and l(w)=l(u)+1l(w) = l(u) + 1. If C(s).C(w)=C(w)C(s).C(w) = C(w), then

C(s).C(u).C(t)=C(u).C(t) C(s).C(u).C(t) = C(u).C(t)

and multiplying on the right by C(t)C(t), we obtain that C(u)∪C(w)=C(s).C(u)∪C(w)C(u) \cup C(w) = C(s).C(u) \cup C(w). Since C(u)≠C(s).C(u)C(u) \neq C(s).C(u) by the induction hypothesis, we have, by b),

C(s).C(u)=C(w), C(s).C(u) = C(w),

and hence

C(u)⊂C(s).C(s).C(u)=C(s).C(w)=C(w), C(u) \subset C(s).C(s).C(u) = C(s).C(w) = C(w),

which is absurd).

d) Let s∈Ss \in S. Show that the map ps:w↦s.wp_s : w \mapsto s.w (resp. qs:w↦w∗sq_s : w \mapsto w * s) is a permutation of WW and that ps2=Idp_s^2 = \mathrm{Id} (resp. qs2=Idq_s^2 = \mathrm{Id}). Show that ps∘qt=qt∘psp_s \circ q_t = q_t \circ p_s for all s,t∈Ss, t \in S (study the product C(s).C(w).C(t)C(s).C(w).C(t) for w∈Ww \in W and show that (s.w)∗t∈{w,s.w,w∗t,s.(w∗t)}(s.w)*t \in \{w, s.w, w*t, s.(w*t)\}; show that (s.w)∗t∉{s.w,w∗t}(s.w)*t \notin \{s.w, w*t\} and that if (s.w)∗t=w(s.w)*t = w then s.w=w∗ts.w = w*t and w=s.(w∗t)w = s.(w*t)).

e) Show that the group of permutations PP (resp. QQ) generated by the psp_s (resp. qsq_s) for s∈Ss \in S acts simply-transitively on WW (to show that PP is transitive, use (2) and argue as in Lemma 1 of no. 1; to show that PP is simply transitive, use d)). Deduce that WW has a unique group structure such that the map p↦p(e)p \mapsto p(e) (where e denotes the element of W such that B = C(e)) is an isomorphism from P to W. The map q ↦ q(e) is then also an isomorphism; moreover, s.w = w \* s = sw for all s ∈ S and all w ∈ W and C(w)^{-1} = C(w^{-1}).

f) Show that the pair (W, S) is a Coxeter system and generalize the results of no. 4.

g) Let X be a subset of S and WX the subgroup of W generated by X. Show that the union G_X of the C(w) for w ∈ WX is a subgroup of G and that Th. 3 of no. 5 is still true. Show that B is a Tits subgroup of G_X. Generalize Props. 2 and 3 of no. 5, and Def. 2, Prop. 4 and Th. 4 of no. 6. Show that S is the set of w ∈ W such that B ∪ C(w) is a subgroup of G distinct from B. The Coxeter system (W, S) and the group W (called the Weyl group of (G, B)) thus depend only on (G, B).

h) Let N be a subgroup of G such that B ∩ N is normal in N and such that the intersection with N of every double coset C(w) with respect to B is a double coset with respect to B ∩ N. Show that the group N/(B ∩ N) can be identified with W and that (G, B, N, S) is a Tits system.

Solution

No solution has been written yet.

Transcribed from the printed volume. Not translated.