• Tidak ada hasil yang ditemukan

Weyl Group

Dalam dokumen Graduate Texts in Mathematics 255 (Halaman 147-151)

3.1 Roots and Weights

3.1.1 Weyl Group

LetGbe a connected classical group and letHbe a maximal torus inG. Define the normalizerofHinGto be

NormG(H) ={g∈G:ghg1∈Hfor allh∈H},

and define the Weyl group WG=NormG(H)/H. Since all maximal tori ofGare conjugate, the groupWGis uniquely defined (as an abstract group) byG, and it acts by conjugation as automorphisms ofH. We shall see in later chapters that many aspects of the representation theory and invariant theory for Gcan be reduced to questions about functions on H that are invariant underWG. The success of this approach rests on the fact thatWG is a finite group of known structure (either the symmetric group or a somewhat larger group). We proceed with the details.

SinceHis abelian, there is a natural homomorphismϕ:WG //Aut(H)given byϕ(sH)h=shs1fors∈NormG(H). This homomorphism gives an action ofWG on the character groupX(H), where forθ∈X(H)the characters·θis defined by

s·θ(h) =θ(s1hs), forh∈H.

(Note that the right side of this equation depends only on the coset sH.) Writing θ=eλforλ∈P(G)as in Section 2.4.1, we can describe this action as

s·eλ=es·λ , where hs·λ,xi=hλ,Ad(s)1xi forx∈h. This defines a representation ofWGonh.

Theorem 3.1.1.WGis a finite group and the representation of WGonhis faithful.

Proof. Lets∈NormG(H). Supposes·θ=θ for allθ∈X(H). Thens1hs=hfor allh∈H, and hences∈Hby Theorem 2.1.5. This proves that the representation of WGonhis faithful.

To prove the finiteness ofWG, we shall assume thatG⊂GL(n,C)is in the matrix form of Section 2.4.1, so thatHis the group of diagonal matrices inG. In the proof of Theorem 2.1.5 we noted thath∈Hacts on the standard basis forCnby

heii(h)eifori=1, . . . ,n,

whereθi∈X(H)andθi6=θjfori6=j. Lets∈NormG(H). Then

hsei=s(s1hs)ei=sθi(s1hs)ei= (s·θi)(h)sei. (3.1) Hencesei is an eigenvector forhwith eigenvalue (s·θi)(h). Since the characters s·θ1, . . . ,s·θnare all distinct, this implies that there is a permutationσ∈Snand there are scalarsλi∈C×such that

seiieσ(i) fori=1, . . . ,n. (3.2)

3.1 Roots and Weights 129

Sincesheiiθi(h)eσ(i) forh∈H, the permutationσ depends only on the coset sH. Ift∈NormG(H)andteiieτ(i)withτ∈Snandµi∈C×, then

tseiiteσ(i)σ(i)λieτ σ(i).

Hence the maps7→σis a homomorphism from NormG(H)intoSnthat is constant on the cosets ofH. If σ is the identity permutation, thenscommutes withH and therefores∈H by Theorem 2.1.5. Thus we have defined an injective homomor- phism fromWGintoSn, soWGis a finite group. ut We now describe WG for each type of classical group. We will use the em- bedding ofWG intoSn employed in the proof of Theorem 3.1.1. Forσ ∈Sn let sσ∈GL(n,C)be the matrix such thatsσei=eσ(i)fori=1, . . . ,n. This is the usual representation ofSnonCnaspermutation matrices.

Suppose G=GL(n,C). Then H is the group of all n×n diagonal matrices.

Clearlysσ ∈NormG(H)for everyσ ∈Sn. From the proof of Theorem 3.1.1 we know that every coset inWG is of the form sσH for someσ ∈Sn. HenceWG∼= Sn. The action ofσ∈Snon the diagonal coordinate functionsx1, . . . ,xnforH is σ·xi=xσ1(i).

LetG=SL(n,C). NowH consists of all diagonal matrices of determinant 1.

Givenσ∈Sn, we may pickλi∈C×such that the transformationsdefined by (3.2) has determinant 1 and hence is in NormG(H). To prove this, recall that every permu- tation is a product of cyclic permutations, and every cyclic permutation is a product of transpositions (for example, the cycle(1,2, . . . ,k)is equal to(1,k)···(1,3)(1,2)).

Consequently, it is enough to verify this whenσ is the transpositioni↔j. In this case we takeλj=−1 andλk=1 fork6=j. Since det(sσ) =−1, we obtain dets=1.

Thus the homomorphismWG //Sn constructed in the proof of Theorem 3.1.1 is surjective. HenceWG∼=Sn. Notice, however, that this isomorphism arises by choosing elements of NormG(H)whose adjoint action onhis given by permutation matrices; the group of all permutation matrices is not a subgroup ofG.

Next, consider the caseG=Sp(C2l,Ω), withΩ as in (2.6). Letsl∈GL(l,C) be the matrix for the permutation(1,l)(2,l−1)(3,l−2)···, as in equation (2.5).

Forσ ∈Sl let sσ ∈GL(l,C)be the corresponding permutation matrix. Clearly stσ=sσ1,so if we define

π(σ) = sσ 0

0 slsσsl

,

thenπ(σ)∈Gand henceπ(σ)∈NormG(H). Obviouslyπ(σ)∈Hif and only if σ=1, so we obtain an injective homomorphism ¯π:Sl //WG.

To find other elements ofWG, consider the transpositions(i,2l+1−i)inS2l, where 1≤i≤l. Setei=e2l+1i, where{ei}is the standard basis forC2l. Define τi∈GL(2l,C)by

τiei=ei, τiei=−ei, τiek=ekfork6=i,−i.

Since {τiej : j=±1, . . . ,±l} is an Ω-symplectic basis for C2l, we have τi ∈ Sp(C2l,Ω) by Lemma 1.1.5. Clearlyτi∈NormG(H)and τi2∈H. Furthermore, τiτjjτiif 1≤i,j≤l. GivenF⊂ {1, . . . ,l}, define

τF=

iF

τi∈NormG(H).

Then theH-cosets of the elements{τF}form an abelian subgroupTl∼= (Z/2Z)lof WG. The action ofτF on the coordinate functionsx1, . . . ,xl forH isxi7→xi 1for i∈Fandxj7→xjfor j∈/F. This makes it evident that

π(σ)τFπ(σ)1σF forF⊂ {1, . . . ,l}andσ∈Sl. (3.3) Clearly,Tl∩π¯(Sl)H={1}.

Lemma 3.1.2.For G=Sp(C2l,Ω), the subgroup Tl⊂WGis normal, and WGis the semidirect product of Tlandπ¯(Sl). The action of WGon the coordinate functions in O[H]is by xi7→ xσ(i)±1 (i=1, . . . ,l), for every permutationσand choice±1of exponents.

Proof. Recall that a groupKis asemidirect productof subgroupsLandMifMis a normal subgroup ofK,L∩M=1, andK=L·M.

By (3.3) we see that it suffices to prove that WG =Tlπ(Sl). Suppose s∈ NormG(H). Then there existsσ∈S2l such thatsis given by (3.2). Define

F={i:i≤landσ(i)≥l+1},

and letµ∈S2lbe the product of the transpositions interchangingσ(i)withσ(i)−l for i∈F. Then µ σ stabilizes the set {1, . . . ,l}. Let ν ∈Sl be the corre- sponding permutation of this set. Thenπ(ν)−1τFsei=±λieifori=1, . . . ,l. Thus

s∈τFπ(ν)H. ut

Now consider the caseG=SO(C2l+1,B), with the symmetric formBas in (2.9).

Forσ∈Sldefine

ϕ(σ) =

sσ 0 0

0 1 0

0 0slsσsl

.

Thenϕ(σ)∈Gand henceϕ(σ)∈NormG(H). Obviously,ϕ(σ)∈Hif and only if σ=1, so we get an injective homomorphism ¯ϕ:Sl //WG.

We construct other elements ofWG in this case by the same method as for the symplectic group. Set

ei=e2l+2i fori=1, . . . ,l+1.

For each transposition (i,2l+2−i) in S2l+1, where 1≤i≤l, we define γi ∈ GL(2l+1,C)by

3.1 Roots and Weights 131

γiei=e−i, γie−i=ei, γie0=−e0, and γiek=ek fork6=i,0,−i.

Thenγi∈O(B,C)and by makingγiact onel+1by−1 we obtain detγi=1. Hence γi ∈NormG(H). Furthermore,γi2∈H andγiγjjγi if 1≤i,j≤l. Given F ⊂ {1, . . . ,l}, we define

γF=

iF

γi∈NormG(H).

Then theH-cosets of the elements{γF}form an abelian subgroupTl∼= (Z/2Z)lof WG. The action ofγFon the coordinate functionsx1, . . . ,xlforO[H]is the same as that ofτFfor the symplectic group.

Lemma 3.1.3.Let G=SO(C2l+1,B). The subgroup Tl⊂WGis normal, and WGis the semidirect product of Tlandϕ(S¯ l). The action of WGon the coordinate functions inO[H]is by xi7→ xσ(i)±1

(i=1, . . . ,l), for every permutationσ and choice±1 of exponents.

Proof. Supposes∈NormG(H). Then there existsσ∈S2l+1such thatsis given by (3.2), withn=2l+1. The action ofsas an automorphism ofHis

s·diag[a1, . . . ,an]s1=diag[aσ1(1), . . . ,aσ1(n)].

Sinceal+1=1 for elements ofH, we must haveσ(l+1) =l+1. Now use the same

argument as in the proof of Lemma 3.1.2. ut

Finally, we consider the caseG=SO(C2l,B), withBas in (2.6). Forσ ∈Sl defineπ(σ)as in the symplectic case. Thenπ(σ)∈NormG(H). Obviously,π(σ)∈ Hif and only ifσ=1, so we have an injective homomorphism ¯π:Sl //WG. The automorphism ofHinduced byσ∈Slis the same as for the symplectic group.

We have slightly less freedom in constructing other elements ofWGthan in the case ofSO(C2l+1,B). Set

ei=e2l+1i fori=1, . . . ,l.

For each transposition (i,2l+1−i) in S2l, where 1≤i≤l, we define βi ∈ GL(2l,C)by

βiei=e−i, βie−i=ei, βiek=ek fork6=i,−i.

Then βi∈O(C2l,B)and clearly βii1=H. However, detβi=−1, soβi∈/ SO(C2l,B). Nonetheless, we still haveβi2∈Handβiβjjβiif 1≤i,j≤l. Given F⊂ {1, . . . ,l}, we define

βF=

iF

βi.

If Card(F)iseven,then detβF=1 and henceβF∈NormG(H). Thus theHcosets of the elements{βF : Card(F)even}form an abelian subgroupRlofWG.

Lemma 3.1.4.Let G=SO(C2l,B). The subgroup Rl⊂WGis normal, and WGis the semidirect product of Rl andπ¯(Sl). The action of WGon the coordinate functions inO[H]is by xi7→ xσ(i)±1 (i=1, . . . ,l), for every permutationσand choice±1 of exponents with anevennumber of negative exponents.

Proof. By the same argument as in the proof of Lemma 3.1.2 we see that the normal- izer ofHinO(C2l,B)is given by theHcosets of the elementsβFπ(σ)asσ ranges overSl andF ranges over all subsets of{1, . . . ,l}. Sinceπ(σ)∈NormG(H),we haveβFπ(σ)∈NormG(H)if and only if Card(F)is even. ut

Dalam dokumen Graduate Texts in Mathematics 255 (Halaman 147-151)