Bourbaki

§ 6. Coxeter transformation

In this paragraph, V denotes a real vector space of finite dimension l and W denotes a finite subgroup of GL(V)\mathbf{GL}(V), generated by reflections and essential (§ 3, no. 7). Provide V with a scalar product (x∣y)(x|y) invariant under W. Denote by H\mathfrak{H} the set of hyperplanes H of V such that the corresponding orthogonal reflection sHs_H belongs to W.

1. Definition of Coxeter transformations

An ordered chamber relative to W is a pair consisting of a chamber C determined by H\mathfrak{H} and a bijection i↦Hii \mapsto H_i from {1,2,…,l}\{1, 2, \ldots, l\} onto the set of walls of C (cf. § 3, no. 9, Prop. 7).

Definition 1

The Coxeter transformation determined by an ordered chamber* (C,(Hi)1≤i≤l)(C, (H_i)_{1 \leq i \leq l}) is the element c=sH1sH2⋯sHlc = s_{H_1} s_{H_2} \cdots s_{H_l} of W.

Proposition 1

All Coxeter transformations are conjugate in W*.

Since W permutes the chambers determined by H\mathfrak{H} transitively (§ 3, no. 2, Th. 1), we are reduced to proving the following: let (C,(Hi)1≤i≤l)(C, (H_i)_{1 \leq i \leq l}) be an ordered chamber and π∈Sl\pi \in S_l; then sH1sH2⋯sHls_{H_1} s_{H_2} \cdots s_{H_l} and sHπ(1)sHπ(2)⋯sHπ(l)s_{H_{\pi(1)}} s_{H_{\pi(2)}} \cdots s_{H_{\pi(l)}} are conjugate in W. Taking into account § 4, no. 8, Prop. 8, this will follow immediately from the following lemma:

Lemma 1

Let X be a finite forest, and* x↦gxx \mapsto g_x a map from X to a group Γ\Gamma such that gxg_x and gyg_y are conjugate whenever xx and yy are not linked in X. Let T\mathscr{T} be the set of total orderings on X. For all ξ∈T\xi \in \mathscr{T}, let pξp_\xi be the product in Γ\Gamma of the sequence (gx)x∈X(g_x)_{x \in X} defined by ξ\xi. Then the elements pξp_\xi are conjugate in Γ\Gamma.

1) We proceed by induction on n=Card Xn = \mathrm{Card}\, X. The case n=1n = 1 is immediate, so assume that n⩾2n \geqslant 2. There exists in XX a terminal vertex aa (Chap. IV, Appendix, no. 3, Prop. 2). Let b∈X−{a}b \in X - \{a\} be a vertex linked to aa if one exists; if aa is not linked to any vertex in X−{a}X - \{a\}, let bb in X−{a}X - \{a\} be arbitrary. In all cases, gag_a commutes with gxg_x for x≠bx \neq b. Let η∈T\eta \in \mathscr{T} be such that aa is the largest element of XX and bb the largest element of X−{a}X - \{a\}; we let ξ∈T\xi \in \mathscr{T} and prove that pξ,pηp_{\xi}, p_{\eta} are conjugate.

2) Assume first that, for ξ\xi, aa is the largest element of XX and bb the largest element of X−{a}X - \{a\}. Let X′X' be the full subgraph X−{a}X - \{a\}, which is a forest. Define a map x↦gx′x \mapsto g'_x from X′X' to Γ\Gamma by putting gx′=gxg'_x = g_x if x≠bx \neq b, gb′=gbgag'_b = g_bg_a. Let ξ′,η′\xi', \eta' be the restrictions of ξ,η\xi, \eta to X′X'. The induction hypothesis applies, so pξ′p_{\xi'} and pη′p_{\eta'} are conjugate. But it is clear that pξ′=pξ,pη′=pηp_{\xi'} = p_{\xi}, p_{\eta'} = p_{\eta}, proving the lemma in this case.

3) Assume that aa is the largest element of XX for ξ\xi. Let X1X_1 (resp. X2X_2) be the set of elements of X−{a}X - \{a\} strictly larger (resp. smaller) than bb; let ξi\xi_i be the restriction of ξ\xi to XiX_i. Then

pξ=pξ1gbpξ2ga=pξ1gbgapξ2, p_{\xi} = p_{\xi_1} g_b p_{\xi_2} g_a = p_{\xi_1} g_b g_a p_{\xi_2},

and this element is conjugate to pξ2pξ1gbgap_{\xi_2} p_{\xi_1} g_b g_a. We are thus reduced to case 2).

4) In the general case, let X3X_3 (resp. X4X_4) be the set of elements of XX strictly larger (resp. smaller) than aa; let ξi\xi_i be the restriction of ξ\xi to XiX_i. Then pξ=pξ3gapξ4p_{\xi} = p_{\xi_3} g_a p_{\xi_4}, and this element is conjugate to pξ4pξ3gap_{\xi_4} p_{\xi_3} g_a. We are thus reduced to case 3).

It follows from Prop. 1 that all the Coxeter transformations are of the same order h=h(W)h = h(W). This number is called the Coxeter number of WW.

Remark

Let W1,…,WmW_1, \ldots, W_m be essential finite groups acting in the spaces V1,…,VmV_1, \ldots, V_m and generated by reflections. Let CjC_j be a chamber relative to WjW_j. Let WW be the group W1×⋯×WmW_1 \times \cdots \times W_m acting in the space V1×⋯×VmV_1 \times \cdots \times V_m. Then C1×⋯×CmC_1 \times \cdots \times C_m is a chamber relative to WW. The Coxeter transformations of WW defined by CC are the products c1c2…cmc_1 c_2 \ldots c_m, where cjc_j is a Coxeter transformation of WjW_j defined by CjC_j.

2. Eigenvalues of a Coxeter transformation

Since all Coxeter transformations are conjugate (no. 1, Prop. 1), they all have the same characteristic polynomial P(T)P(T). Let hh be the Coxeter number of WW. Then

P(T)=∏j=1l(T−exp⁡2iπmjh), P(T) = \prod_{j=1}^l \left( T - \exp \frac{2i \pi m_j}{h} \right),

where m1,m2,…,mlm_1, m_2, \ldots, m_l are integers such that

0≤m1≤m2≤⋯≤ml<h. 0 \leq m_1 \leq m_2 \leq \cdots \leq m_l < h.

Definition 2

The integers* m1,m2,…,mlm_1, m_2, \ldots, m_l are called the exponents of W.

Let C be a chamber determined by H,H1,…,Hl\mathfrak{H}, H_1, \ldots, H_l its walls, and put si=sHis_i = s_{H_i}. Denote by eie_i the unit vector orthogonal to HiH_i and on the same side of HiH_i as C. By Prop. 2 of the Appendix of Chap. IV, we can assume that the HiH_i are numbered so that e1,e2,…,ere_1, e_2, \ldots, e_r are pairwise orthogonal and er+1,er+2,…,ele_{r+1}, e_{r+2}, \ldots, e_l are pairwise orthogonal. Then s′=s1s2…srs' = s_1 s_2 \ldots s_r is the orthogonal symmetry with respect to the subspace

V′=H1∩H2∩⋯∩Hr, V' = H_1 \cap H_2 \cap \cdots \cap H_r,

s′′=sr+1sr+2…sls'' = s_{r+1} s_{r+2} \ldots s_l is the orthogonal symmetry with respect to the subspace

V′′=Hr+1∩Hr+2∩⋯∩Hl, V'' = H_{r+1} \cap H_{r+2} \cap \cdots \cap H_l,

and c=s′s′′c = s' s'' is a Coxeter transformation. Since (e1,…,el)(e_1, \ldots, e_l) is a basis of V, V is the direct sum of V′V' and V′′V''.

We deduce first that 1 is not an eigenvalue of c. For if x∈Vx \in V is such that c(x)=xc(x) = x, then s′(x)=s′′(x)s'(x) = s''(x), so x−s′(x)=x−s′′(x)x - s'(x) = x - s''(x) is orthogonal to V′V' and V′′V'', and hence is zero. Thus, x=s′(x)=s′′(x)∈V′∩V′′={0}x = s'(x) = s''(x) \in V' \cap V'' = \{0\}.

Consequently,

0<m1≤m2≤⋯≤ml<h. 0 < m_1 \leq m_2 \leq \cdots \leq m_l < h.
(1)

The characteristic polynomial of c has real coefficients. Thus, for all jj, the power of T−exp⁡2iπmjhT - \exp \frac{2i\pi m_j}{h} in P(T) is equal to that of T−exp⁡2iπ(h−mj)hT - \exp \frac{2i\pi (h - m_j)}{h}. Hence

mj+ml+1−j=h(1≤j≤l). m_j + m_{l+1-j} = h \quad (1 \leq j \leq l).
(2)

Adding the equalities (2) we obtain

m1+m2+⋯+ml=12lh. m_1 + m_2 + \cdots + m_l = \frac{1}{2} lh.
(3)

Lemma 2

Assume that* W is irreducible and that dim⁡V≥2\dim V \geq 2. With the preceding notation, there exist two linearly independent vectors z′,z′′z', z'' such that (i) the plane P generated by z′,z′′z', z'' is stable under s' and s''; (ii) s'|P and s''|P are orthogonal reflections with respect to Rz′Rz' and Rz′′Rz''; (iii) z′,z′′∈C‾z', z'' \in \overline{C}, and P∩CP \cap C is the set of linear combinations of z′,z′′z', z'' with coefficients > 0.

Let (e1,…,el)(e^1, \ldots, e^l) be the basis of V such that (ei∣ej)=δij(e^i|e_j) = \delta_{ij}. Then C is the open simplicial cone determined by the eie^i (§ 3, no. 9, Prop. 7). It is clear that V′V' is generated by er+1,…,ele^{r+1}, \ldots, e^l and V′′V'' by e1,…,ere^1, \ldots, e^r. Let q be the endomorphism of V such that q(e1)=e1,…,q(el)=elq(e^1) = e_1, \ldots, q(e^l) = e_l. Its matrix with respect to (e1,…,el)(e^1, \ldots, e^l) is Q=((ei∣ej))Q = ((e_i|e_j)). We have (ei∣ej)≤0(e_i|e_j) \leq 0 for i≠ji \neq j (§ 3, no. 4, Prop. 3). Since W is irreducible, there does not exist any partition {1,2,…,l}=I1∪I2\{1, 2, \ldots, l\} = I_1 \cup I_2 such that (ei∣ej)=0(e_i|e_j) = 0 for i∈I1i \in I_1 and j∈I2j \in I_2. Thus (§ 3, no. 5, Lemma 4) Q has an eigenvector (a1,…,al)(a_1, \ldots, a_l) all of whose coordinates are > 0; let a be the corresponding eigenvalue. Put

z=a1e1+⋯+alel, z = a_1 e^1 + \cdots + a_l e^l,
z′′=a1e1+⋯+arer∈V′′∩C‾, z'' = a_1 e^1 + \cdots + a_r e^r \in V'' \cap \overline{C},
z′=ar+1er+1+⋯+alel∈V′∩C‾, z' = a_{r+1} e^{r+1} + \cdots + a_l e^l \in V' \cap \overline{C},

and let PP be the plane generated by z′z' and z′′z''. Then P∩CP \cap C is the set of linear combinations of z′z' and z′′z'' with coefficients >0> 0. The relation q(z)=azq(z) = az gives ∑j=1lajej=∑j=1laajej\sum_{j=1}^l a_j e_j = \sum_{j=1}^l aa_j e^j; scalar multiplying by eke_k (where k≤rk \leq r) gives

ak+∑j=r+1laj(ej∣ek)=aak; thus a_k + \sum_{j=r+1}^l a_j (e_j|e_k) = aa_k; \text{ thus}
(a−1)z′′=∑k=1r(∑j=r+1laj(ej∣ek))ek (a-1)z'' = \sum_{k=1}^r \left( \sum_{j=r+1}^l a_j (e_j|e_k) \right) e^k
=∑j=r+1laj(∑k=1r(ej∣ek)ek) = \sum_{j=r+1}^l a_j \left( \sum_{k=1}^r (e_j|e_k) e^k \right)
=∑j=r+1laj(−ej+∑k=1l(ej∣ek)ek) = \sum_{j=r+1}^l a_j \left( -e^j + \sum_{k=1}^l (e_j|e_k) e^k \right)
=−z′+∑j=r+1lajej. = -z' + \sum_{j=r+1}^l a_j e_j.

Thus, (a−1)z′′+z′(a-1)z'' + z' is orthogonal to e1,…,ere^1, \ldots, e^r, that is, to V′′V''. Hence, s′′s'' leaves stable the plane generated by z′′z'' and (a−1)z′′+z′(a-1)z'' + z', that is, PP. Similarly, s′s' leaves PP stable. Since z′∈P∩V′z' \in P \cap V' and z′′∈P∩V′′z'' \in P \cap V'', s′∣Ps'|P and s′′∣Ps''|P are the reflections with respect to Rz′Rz' and Rz′′Rz''.

Theorem 1

Assume that W is irreducible. Then:* (i) m1=1,ml=h−1m_1 = 1, m_l = h-1. (ii) Card(H)=12lh\mathrm{Card}(\mathfrak{H}) = \frac{1}{2} lh.

We retain the preceding notation. The restriction of c=s′s′′c = s's'' to PP is the rotation with angle 2(z′′,z′^)2(\widehat{z'', z'}) (§ 2, no. 5, Cor. of Prop. 6). Since cc is of order hh, the hh elements 1,c,…,ch−11, c, \ldots, c^{h-1} of WW are pairwise distinct; the elements s′,s′c,…,s′ch−1s', s'c, \ldots, s'c^{h-1} are thus pairwise distinct, and are distinct from the preceding elements since ci∣Pc^i|P is a rotation and s′cj∣Ps'c^j|P is a reflection. The set

{1,c,…,ch−1,s′,s′c,…,s′ch−1} \{1, c, \ldots, c^{h-1}, s', s'c, \ldots, s'c^{h-1}\}
is the subgroup W′W' of WW generated by s′s' and s′′s'', and induces on PP the group W′′W'' generated by the orthogonal reflections with respect to Rz′,Rz′′Rz', Rz''. The transform of CC by an element of W′W' is either disjoint from −C-C or equal to −C-C. Thus, the transform of P∩CP \cap C by an element of W′′W'' is either disjoint from −(P∩C)-(P \cap C) or equal to −(P∩C)-(P \cap C). Hence, for a suitable orientation of P, there exists an integer m>0m > 0 such that (z′′‾,z′)=πm(\overline{z''}, z') = \frac{\pi}{m} (§ 2, no. 5, Cor. of Prop. 7). Moreover, the sets g′(C)g'(C), for g′∈W′g' \in W', are pairwise disjoint; the sets g′′(P∩C)g''(P \cap C), for g′′∈W′′g'' \in W'', are thus pairwise disjoint; so W′′W'' is of order 2h2h. Hence m=hm = h. By definition, c∣Pc|P is a rotation with angle 2πm\frac{2\pi}{m}, and so has eigenvalues exp⁡2iπh\exp \frac{2i\pi}{h}, exp⁡2iπ(h−1)h\exp \frac{2i\pi(h-1)}{h}. This proves that m1=1,ml=h−1m_1 = 1, m_l = h - 1.

The transforms of Rz′Rz' and Rz′′Rz'' by W′W' are hh lines D1,…,DhD_1, \ldots, D_h of P, and the points of P−(D1∪⋯∪Dh)P - (D_1 \cup \cdots \cup D_h) are transforms by the elements of W′W' of points of P∩CP \cap C. Thus, a hyperplane of H\mathfrak{H} necessarily cuts P along one of the lines DiD_i, and consequently is a transform by an operation of W′W' of a hyperplane of H\mathfrak{H} containing Rz′Rz' or Rz′′Rz''.

Now, any H∈HH \in \mathfrak{H} that contains Rz′Rz' is one of the hyperplanes H1,…,HrH_1, \ldots, H_r. Indeed, let eHe_H be the unit vector orthogonal to H and on the same side of H as C. Then eH=λ1e1+⋯+λlele_H = \lambda_1 e_1 + \cdots + \lambda_l e_l with the λi\lambda_i all ≥0\geq 0 (§ 3, no. 5, Lemma 6, (i)). Now 0=(eH∣z′)=λr+1ar+1+⋯+λlal0 = (e_H|z') = \lambda_{r+1} a_{r+1} + \cdots + \lambda_l a_l, so

λr+1=⋯=λl=0,andeH=λ1e1+⋯+λrer. \lambda_{r+1} = \cdots = \lambda_l = 0, \quad \text{and} \quad e_H = \lambda_1 e_1 + \cdots + \lambda_r e_r.
Suppose that two of the λi\lambda_i were non-zero, for example λ1\lambda_1 and λ2\lambda_2; since e1,…,ere_1, \ldots, e_r are pairwise orthogonal, we would have
s1(eH)=−λ1e1+λ2e2+⋯+λrer s_1(e_H) = -\lambda_1 e_1 + \lambda_2 e_2 + \cdots + \lambda_r e_r
and the coordinates of s1(eH)s_1(e_H) would not be of the same sign, which is absurd (loc. cit.). Thus, eHe_H is proportional to one of the vectors e1,…,ere_1, \ldots, e_r, which proves our assertion. Similarly, any H∈HH \in \mathfrak{H} that contains Rz′′Rz'' is one of the hyperplanes Hr+1,…,HlH_{r+1}, \ldots, H_l.

The number of elements of H\mathfrak{H} containing Rz′Rz' or Rz′′Rz'' is therefore ll. If hh is even, Card(H)\mathrm{Card}(\mathfrak{H}) is thus equal to h2l\frac{h}{2} l. If hh is odd, Card(H)\mathrm{Card}(\mathfrak{H}) is equal to h−12l+r\frac{h-1}{2} l + r, and also to h−12l+(l−r)\frac{h-1}{2} l + (l - r); hence r=l−rr = l - r, so r=l2r = \frac{l}{2}, and Card(H)=h−12l+l2=h2l\mathrm{Card}(\mathfrak{H}) = \frac{h-1}{2} l + \frac{l}{2} = \frac{h}{2} l.

Remark

Retain the notation of the preceding proof. Let c′c' be the C\mathbf{C}-linear extension of cc to V⊗RCV \otimes_{\mathbf{R}} \mathbf{C}, and c′′c'' the restriction of c′c' to P⊗RCP \otimes_{\mathbf{R}} \mathbf{C}. From our study of c∣Pc|P, c′′c'' has an eigenvector xx corresponding to the eigenvalue exp⁡2iπh\exp \frac{2i\pi}{h}, and this eigenvector does not belong to any of the sets D⊗RCD \otimes_{\mathbf{R}} \mathbf{C}, where D denotes a line of P (since D is not stable under c). Now, for any H∈HH \in \mathfrak{H}, we have seen that H∩PH \cap P is a line; thus, x∉H⊗RCx \notin H \otimes_{\mathbf{R}} \mathbf{C}.

Corollary

Let R0R_0 be the set of unit vectors of VV orthogonal to an element of H\mathfrak{H}. If WW is irreducible,

∑u∈R0(x∣u)2=h(x∣x) \sum_{u \in R_0} (x|u)^2 = h(x|x)

for all x∈Vx \in V.

Put f(x)=∑u∈R0(x∣u)2f(x) = \sum_{u \in R_0} (x|u)^2. It is clear that ff is a positive quadratic form invariant under WW, and non-degenerate since the eie_i form a basis of VV. Since

W is irreducible, there exists a constant β\beta such that f(x)=β(x∣x)f(x) = \beta(x|x) (§ 2, no. 1, Prop. 1). If (xi)1≤i≤l(x_i)_{1 \leq i \leq l} is an orthonormal basis of V for the scalar product (x∣y)(x|y), then

βl=∑i=1lβ(xi∣xi)=∑i=1lf(xi)=∑i=1l∑u∈R0(xi∣u)2=∑u∈R01=Card(R0)=2Card(H)=hl. \beta l = \sum_{i=1}^l \beta(x_i|x_i) = \sum_{i=1}^l f(x_i) = \sum_{i=1}^l \sum_{u \in R_0} (x_i|u)^2 \\ = \sum_{u \in R_0} 1 = \mathrm{Card}(R_0) = 2\mathrm{Card}(\mathfrak{H}) = hl.
Thus β=h\beta = h, which proves (4).

Proposition 2

If W is irreducible and h is even, the unique element of W that transforms C to −C is ch/2c^{h/2}*.

We use the notation in the proof of Th. 1. Since c∣Pc|P is a rotation through an angle 2πh\frac{2\pi}{h}, ch/2c^{h/2} transforms z′z' to −z′-z', z′′z'' to −z′′-z'', and hence z′+z′′=zz' + z'' = z to −z-z. Now z∈Cz \in C, so the chamber ch/2(C)c^{h/2}(C) is necessarily −C-C.

Proposition 3

Assume that W is irreducible. Let u1,…,ulu_1, \ldots, u_l be homogeneous elements of the symmetric algebra S=S(V)S = S(V), algebraically independent over R\mathbf{R} and generating the algebra of elements of S invariant under W (§ 5, no. 3, Th. 3). If pjp_j is the degree of uju_j, the exponents of W are p1−1,…,pl−1p_1 - 1, \ldots, p_l - 1*.

Put V′=V⊗RCV' = V \otimes_{\mathbf{R}} \mathbf{C}, S′=S(V′)=S⊗RCS' = S(V') = S \otimes_{\mathbf{R}} \mathbf{C}, and extend the scalar product on V to a hermitian form on V′V'. If cc is a Coxeter transformation of W, there exists an orthonormal basis (Xi)1≤i≤l(X_i)_{1 \leq i \leq l} of V′V' consisting of eigenvectors of c⊗1c \otimes 1 (Algebra, Chap. IX, § 7, no. 3, Prop. 4); moreover, we can assume that, for 1≤j≤l1 \leq j \leq l, XjX_j corresponds to the eigenvalue exp⁡2iπmjh\exp \frac{2i\pi m_j}{h} of c⊗1c \otimes 1. It is clear that S′S' can be identified with the algebra C[X1,…,Xl]\mathbf{C}[X_1, \ldots, X_l], and that we can write uj⊗1=fj(X1,…,Xl)u_j \otimes 1 = f_j(X_1, \ldots, X_l), where fjf_j is a homogeneous polynomial of degree pjp_j in C[X1,…,Xl]\mathbf{C}[X_1, \ldots, X_l]. Put Dj=∂∂XjD_j = \frac{\partial}{\partial X_j} and J=det⁡(Dkfj)J = \det(D_k f_j). Recall (§ 5, no. 4, Prop. 5) that J(X1,…,Xl)J(X_1, \ldots, X_l) is proportional to the product in S′S' of Card(H)\mathrm{Card}(\mathfrak{H}) vectors yky_k of V each of which is orthogonal to a hyperplane of H\mathfrak{H}. Since we can assume that X1∉H⊗CX_1 \notin H \otimes \mathbf{C} for all H∈HH \in \mathfrak{H} (Remark), the X1X_1 component of each of the vectors yky_k is non-zero, so J(1,0,…,0)≠0J(1, 0, \ldots, 0) \neq 0. The rule for expanding determinants now proves the existence of a permutation σ\sigma of {1,2,…,l}\{1, 2, \ldots, l\} such that (Dσ(j)fj)(1,0,…,0)≠0(D_{\sigma(j)} f_j)(1, 0, \ldots, 0) \neq 0 for all jj. Since Dσ(j)fjD_{\sigma(j)} f_j is homogeneous of degree pj−1p_j - 1, the coefficient of X1pj−1Xσ(j)X_1^{p_j - 1} X_{\sigma(j)} in fj(X1,…,Xl)f_j(X_1, \ldots, X_l) is non-zero. Now fj(X1,…,Xl)f_j(X_1, \ldots, X_l) is invariant under c⊗1c \otimes 1, and

(c⊗1)(X1pj−1Xσ(j))=(exp⁡2iπh(pj−1+mσ(j)))(X1pj−1Xσ(j)). (c \otimes 1)(X_1^{p_j - 1} X_{\sigma(j)}) = \left( \exp \frac{2i\pi}{h} (p_j - 1 + m_{\sigma(j)}) \right) (X_1^{p_j - 1} X_{\sigma(j)}).
This proves that pj−1+mσ(j)≡0p_j - 1 + m_{\sigma(j)} \equiv 0 (mod. hh). Now h−mσ(j)h - m_{\sigma(j)} is an exponent (formula (2)). Permuting the uju_j if necessary, we can assume that pj−1≡mjp_j - 1 \equiv m_j (mod. hh) for all jj. Since pj−1≥0p_j - 1 \geq 0 and mj<hm_j < h, we have pj−1=mj+μjhp_j - 1 = m_j + \mu_j h with μj\mu_j an integer ≥0\geq 0. By § 5, Prop. 3, we see that

Card⁡(H)=∑j=1l(pj−1)=∑j=1lmj+h∑j=1lμj. \operatorname{Card}(\mathfrak{H}) = \sum_{j=1}^{l} (p_j - 1) = \sum_{j=1}^{l} m_j + h \sum_{j=1}^{l} \mu_j.

Taking into account formula (3) and Th. 1 (ii), we obtain h∑j=1lμj=0h \sum_{j=1}^{l} \mu_j = 0, so μj=0\mu_j = 0 for all jj, and finally pj−1=mjp_j - 1 = m_j for all jj.

Corollary 1

If* (mi)1≤i≤l(m_i)_{1 \leq i \leq l} is the increasing sequence of exponents of W, the order of W is equal to (m1+1)(m2+1)…(ml+1)(m_1 + 1)(m_2 + 1) \ldots (m_l + 1).

This follows from the relations mj+1=pjm_j + 1 = p_j and § 5, Cor. of Th. 3.

Corollary 2

If* c is a Coxeter transformation of W,

exp⁡(2iπh)andexp⁡(−2iπh) \exp \left( \frac{2i\pi}{h} \right) \quad \text{and} \quad \exp \left( \frac{-2i\pi}{h} \right)

are eigenvalues of c of multiplicity 1.

Otherwise, there would exist two non-proportional homogeneous invariants of degree 2 in S, and hence two non-proportional quadratic forms on V* invariant under W, contrary to § 2, no. 1, Prop. 1.

Corollary 3

The homothety with ratio* −1 of V belongs to W if and only if all the exponents of W are odd. In that case, h is even and ch/2=−1c^{h/2} = -1 for any Coxeter transformation c of W.

The first assertion follows from § 5, no. 3, Prop. 4. Assume that the exponents of W are odd. Then h is even by formula (2), and

(exp⁡2iπmjh)h/2=exp⁡(iπmj)=−1; \left( \exp \frac{2i\pi m_j}{h} \right)^{h/2} = \exp(i\pi m_j) = -1;

thus ch/2=−1c^{h/2} = -1 since c is a semi-simple automorphism of V (Algebra, Chap. IX, § 7, no. 3, Prop. 4).

Exercises

In the exercises below (with the exception of Exerc. 3), the assumptions and notation are those of § 6.

See the exercises for § 6.

Transcribed from the printed volume. Not translated.