Lecture Notes

Reection Groups
Dr Dmitriy Rumynin
Anna Lena Winstel
Autumn Term 2010
Contents
1 Finite Reflection Groups 3
2 Root systems 6
3 Generators and Relations 14
4 Coxeter group 16
5 Geometric representation of W(m
ij
) 21
6 Fundamental chamber 28
7 Classification 34
8 Crystallographic Coxeter groups 43
9 Polynomial invariants 46
10 Fundamental degrees 54
11 Coxeter elements 57
1 Finite Reflection Groups
V = (V, , ) - Euclidean Vector space where V is a nite dimensional vector space over R and
, : V V R is bilinear, symmetric and positiv denit.
Example: (R
n
, ) : (
i
), (
i
) =
n

i=1

i
Gram-Schmidt theory tells that for all Euclidean vector spaces, there exists an isometry (linear
bijective and x, y V : T(x) T(y) = x, y) T : V R
n
.
In (V, , ) you can
measure length: [[x[[ =
_
x, x
measure angles: xy = arccos
_
x,y)
[[x[[[[y[[
_
talk about orthogonal transformations
O(V ) = T GL(V ) : x, y V : Tx, Ty = x, y GL(V )
T GL(V ). Let V
T
= v V : Tv = v the xed points of T or the 1-eigenspace.
Denition. T GL(V ) is a reection if T O(V ) and dim V
T
= dim V 1.
Lemma 1.1. Let T be a reection, x (V
T
)

= v : w V
T
: v, w = 0, x ,= 0. Then
1. T(x) = x
2. z V : T(z) = z 2
x,z)
x,x)
x
Proof.
1. Pick v V
T
= Tv = v = Tx, v = Tx, Tv = x, v = 0. Hence Tx (V
T
)

.
Since dim (V
T
)

= dim V dim V
T
= 1 : Tx = x for some R. Then

2
x, x = x, x = Tx, Tx = x, x
Since x, x , = 0,
2
= 1 = 1, 1.
If = 1 = Tx = x = x V
T
(V
T
)

= x = 0 which is a contradiction.
So = 1.
2.
_
z
x,z)
x,x)
x, x
_
= z, x
x,z)
x,x)
x, x = 0.
So z
x,z)
x,x)
x x

= V
T
and T(z
x,z)
x,x)
x) = z
x,z)
x,x)
x. Hence
T(z) = T((z
x, z
x, x
x) +
x, z
x, x
x) = T(z
x, z
x, x
x) +
x, z
x, x
T(x)
= z
x, z
x, x
x
x, z
x, x
x = z 2
x, z
x, x
x
3
2
For each x V, x ,= 0 dene S
x
(z) = z 2
x,z)
x,x)
x.
Lemma 1.1 implies that
1. any reection T is equal to S
x
for x determined up to a scalar. Any such x V is called
a root of T.
2. x V, x ,= 0 : S
x
is a reection.
3. any reection T satises T
2
= I.
Denition. A nite reection group is a pair (G, V ) where V is Euclidean space, G is a nite
subgroup of O(V ) and G = S
x
: S
x
G, generated by all reections in G.
Generation means: if G X, then X generates G if G = X where X is dened as one of the
following equivalent denitions:
1. (semantic) X =

GHX
H
2. (syntactic) X = 1 a
1
1
a
1
2
a
1
n
: a
i
X
Equivalence:
(G
1
, V
1
) (G
2
, V
2
) if there is an isometry : V
1
V
2
s.t. G
1

1
. .
T
1
: TG
1

= G
2
.
Problem: Classify all nite reection groups (G, V ) up to .
Example: of nite reection groups
1.
y
x

We want to know S
x
, S
y

det(S
x
, S
y
) = 1 = S
x
S
y
SO
2
(R) = S
x
S
y
= Rot

.
Since S
x
(S
y
(y)) = S
x
(y) = S
x
S
y
= 2.

/ Q = [S
x
S
y
[ = = S
x
, S
y
is not nite

=
m
n
Q (for m, n rel. prime) = [S
x
S
y
[ = [Rot
2m
n
[ = n = S
x
, S
y
= D
2n
Dihedral
I
2
(n) = (D
2n
, R
2
), [I
2
(n) = 2n[
2. S
n
-symmetric group acting on 1, . . . , n extend to action of S
n
on R
n
.
If e
i
basis of R
n
, if S
n
, then
T

: e
i
e
(i)
, T

O
n
(R)
4
x =
_
_
_
_
_
1
1
.
.
.
1
_
_
_
_
_
has the property that forall : T

(x) = x = x (R
n
)
S
n
,
x

=
_
_
_

1
.
.
.

n
_
_
_
:

i
= 0
A
n1
= (S
n
, x

)
Reection since S
n
= (i, j) and T
(i,j)
= S
e
i
e
j
because T
ij
(e
i
e
j
) = (e
i
e
j
) and
y =
_
_
_

1
.
.
.

n
_
_
_
(e
i
e
j
)

i
=
j
T
(i,j)
(y) = y
In particular, [A
m
[ = (m+ 1)! and A
2
I
2
(3)
3. F =
Z
/
2Z
= 0, 1 - eld of two elements. Consider action of S
n
on F
n
,
1
, . . . ,
n
basis
of F
n
.
S
n
: t

(
i
) =
(i)
Consider semidirect product S
n
F
n
. As a set: S
n
F
n
= S
n
F
n
, the product is
(, a) (, b) = (, t

1(a) +b)
S
n
F
n
acts on R
n
:
T
(,)
: e
i
(1)
a
i
e

i
Let us check that this is the action of S
n
F
n
:
T
(1,)
(T
(,0)
(e
i
)) = T
(1,)
(e
(i)
)
= (1)
a
(i)
e
(i)
= (1)
[t

1
(a)]
i
e
(i)
= T
(,t

1
(a))
(e
i
)
B
n
= (S
n
F
n
, R
n
)
It is reection since S
n
F
n
= ((i, j), 0), (1,
i
) and T
((i,j),0)
= S
e
i
e
j
, T
(1,
i
)
= S
e
i
.
[B
n
[ = n!2
n
, B
1
A
1
, B
2
I
2
(4)
4. In Example 3, let F
n
0
=
_
_
_

1
.
.
.

n
_
_
_
F
n
:

i
= 0 codim-1 subspace in F
n
or index 2
subgroup.
S
n
, a F
n
0
= t

(a) F
n
0
, so S
n
acts on F
n
0
and
D
n
= (S
n
F
n
0
, R
n
)
It is a reection since S
n
F
n
0
= ((i, j), 0), (1,
i
+
j
),
T
((i,j),0)
= S
e
i
e
j
, T
(1,
i
+
j
)
= S
e
i
+e
j
.
[D
n
[ = n! 2
n1
, D
1
is trivial (1, R), D
2
I
2
(2), D
3
A
3
5
2 Root systems
Let (G, V ) be a nite reection group. The root system of (G, V ) is

(G,V )
:= x V : |x| = 1, S
x
G
has the following properties:
1. x =Rx = x, x
2. [[ = 2( number of reections in G)
3. T G, x = T(x)
Property 3 follows from
Lemma 2.1. T O(V ), x V 0 = S
T(x)
= TS
x
T
1
.
Proof. RHS: T(x) TS
x
T
1
Tx = TS
x
x = T(x) = T(x)
RHS : T(x)

y TS
x
(T
1
y) = T(T
1
y 2

x, T
1
y
_
x, x
x)
= T(T
1
y 2
Tx, y
x, x
x)
= T(T
1
y)
= y
Hence RHS = S
T(x)
. 2
Example: I
2
(3) A
2
-dihedral group of order 6, symmetry of regular triangle.
6 roots sitting at the vertices of a regular hexagon.
Denition. A root system is a nite subset V s.t.
1. 0
V
/
2. x =Rx = x, x
3. x, y = S
x
(y)
6
Example:
1.
(G,V )
where (G, V ) is a nite reection group
2. =
(1,V )
3. is a root system, not
(G,V )
because there are vectors of 2 dierent length. Chop-
ping long vectors by

2 gives
B
2
.
Let V be a root system. Denition. A simple subsystem is s.t.
1. is linearly independent
2. x : x =

y
y where either all
y
0 or all
y
0
Example: In

, is a simple system but

is not simple since

.
Lemma 2.2. simple system in a root system. Then x, y , x ,= y = x, y 0
(angles in a simple system are obtuse)
Proof. Suppose x ,= y, x, y > 0.
S
x
(y) = y2
x,y)
x,x)
x = y+x, < 0. Since is lin. independent and x ,= y, S
x
(y) = y+x
is the only way to write S
x
(y) as a linear combination of elements of and both positive and
negative coecients are present. This is a contradiction. 2
Denition. A total order on R-vector space V is a linear order on V s.t. (-order, x > y if
x y and x ,= y)
1. x y = x +z y +z
2. x y, > 0 = x y
3. x y, < 0 = x y
Example: Phonebook-order on R
n
:
_
_

n
_
_
>
_
_
_

1
.
.
.

n
_
_
_
k 1, . . . , n s.t.
i
=
i
i < k and
k
>
k
Given an ordered basis e
1
, . . . , e
n
of V we get phonebook order on V by writing

i
e
i
=
_
_

n
_
_
.
If is a total order on V then
x V 0 = either x > 0 > x or x < 0 < x
and V = V
+

0

where
7
V
+
= x V : x > 0
V

= x V : x < 0
Denition. A positive system in a root system is a subset s.t. total order on V
s.t. = V
+
.
Example:
1. Box

. Let be the total order associated to the ordered basis , . Then

2 + > + > > > 0 > > > > 2
So , , +, 2 + is a positive system.
2. I
2
(n)-symmetry of n-gon, n reections, so 2n roots, at vertices of regular 2n gon.
(, ) is simple the cone

, =

n


, =
(n 1
n

Every root lies in 2 simple systems
number of simple systems = n
positive systems = vectors between some and where

, =
(n1
n

3. A
n
:
T
(a
1
,...,a
k
)
(z) = (z
k
1)(z 1)
m
(minimal polynomial of T
(a
1
,...,a
k
)
is (z
k
1).
T

reection
T

= (1 +z)(1 z)
m
= (i, j)
Reections are T
(i,j)
,
n(n1)
n
of them and = e
i
e
j
: i ,= j 1, . . . , n + 1.
A typical simple system is
e
1
e
2
, e
2
e
3
, . . . , e
n1
e
n
For i < j we have: e
i
e
j
= (e
i
e
i+1
) + (e
i+1
e
i+2
) +. . . + (e
j1
e
j
).
Notation: - root system,
+
- positive system.
Denition.
+
is a quasisimple system if it satises:
1. x
+

t
0 s.t. x =

t
t
2. No proper subset of satises (1).
Hint: simplicity quasisimplicity
Lemma 2.3. Let
+
be a quasisimple system. Then x, y , x ,= y = x, y 0.
Proof. Suppose x ,= y and x, y > 0. Then:
S
x
(y) = y 2
x,y)
x,x)
x = y x, > 0. Consider two cases:
8
Case 1 S
x
(y)
+
. By (1): y x =

t
t,
t
0. What is
y
?
If
y
< 1 then (1
y
)y = x +

t,=y

t
t. So y is a positive linear combination of elements
of y. Hence y satises (1), contradiction to (2).
Hence
y
1, then 0 = x+(
y
1)y +

t,=y

t
t, coecients in the RHS 0. Hence RHS
0. Since RHS = 0 = = 0 which is a contradiction.
Case 2 S
x
(y)

= y x =

t
(
t
)t,
t
0.
= x y =

t
t. Similarly to case 1:

x
< = (
x
)x = y +

t,=x

t
t contradicts (2)

x
= 0 = y + (
x
)x +

t,=x

t
t. RHS has a positive coecient, hence > 0
which is a contradiction.
2
Theorem 2.4. 1. Every positive system contains a unique simple system.
2. Every simple system is contained in an unique positive system.
Hence there is a natural bijection between
: is simple and
+
: is positive
Proof.
1. Let
+
be a positive system. Consider all subsets of
+
that satisfy condition (1)
of the denition of a quasisimple system. A minimal subset among them must satisfy
(2), so is a quasisimple system in
+
. To prove that is a simple system, it suces
to prove that is linearly independent.
Suppose

t

t
t = 0. Sorting positive and negative coecients to two dierent sides,
v =

t
t =

t
t,
t
,
t
0
Now we consider
v, v =
_

t
t,

t
_
=

s,t

s
..
0
t, s
..
0
0
Hence v = 0 =
i
= 0 =
t
=
t
= 0. Hence is linearly independent and simple.
Let ,
t
be two dinstinct simple systems. Wlog x
t
.
Since x
+
, x =

y
y,
y
0. We know
y =

z
y
z,
z
y
0
9
So
x =

_
_
_

z
y
. .
0
_
_
_
z
= y
0

t
:
y
0
,= 0 = at least two z
1
, z
2
s.t.
z
1
y
0
,= 0,
z
2
y
0
,= 0 which is a
contradiction with the lineary independece of
t
.
2. Let be a simple system. is linearly independent = we can extend to a basis
B of V . Choose an order on B and consider the phonebook total order on V . Clearly,
V
+
=
+
= V
+
. Uniqueness follows from the denition of a simple
system: Every x is a nonnegative or nonpositive linear combination of elements
of = nonnegative linear combinations V
+
, nonpositive linear combinations V

.
Hence
+
=

t
t :
t
0.
2
Proposition 2.5. Let
+
be a root system, positive system, simple system. Then
x , y
+
, x ,= y:
S
x
(y)
+
Note: x ,= y is essential since S
x
(x) = x

.
Proof. Let y =

t
t,
t
0. y ,= x = t
0
s.t. t
0
,= x,
t
0
> 0.
Hence S
x
(y) =

t
t x with
t
0
> 0. Hence S
x
(y)
+
. 2
Example: Box: simple: , simple:
g G, let L(g) = x
+
: g(x)

=
+
g
1
(

). We know that L(1) =

, L(S
x
)
x
= x.
Dene function l : G Z
0
called length by l(g) = [L(g)[. In particular l(1) = 0, l(S
x
) = 1.
Proposition 2.6. The following statements about l( ) hold:
1. l(g) = l(g
1
)
2. l(S
x
g) =
_
l(g) + 1 if g
1
(x)
+
l(g) 1 if g
1
(x)

3. l(gS
x
) =
_
l(g) + 1 if g(x)
+
l(g) 1 if g(x)

Proof.
1. x g(x) is a bijection between L(g) and L(g
1
)
x
+
x = g
1
(g(x))

g(x)

g(x)
+
10
2. By 2.5, x is the only root that changes the sigh under S
x
.
g
1
(x)
g
x
S
x
x
if g
1
(x)
+
= g
1
(x) / L(g) but g
1
(x) L(S
x
g). So L(S
x
g) = L(g) g
1
(x).
if g
1
(x)

= g
1
(x) L(g) but not in L(S
x
g) and L(S
x
g) = L(g) g
1
(x).
2
G acts on V, , : is simple .
Theorem 2.7. Let ,
t
be two simple systems. Then g G s.t.

t
= g() = gx : x
Note: such g is unique but we need some more tools to prove it.
Proof. Let
+
,
t
+
be the corresponding positive systems,

,
t

the corresponding negative

systems. Procced by induction on [
+

t

[ (distance between and

t
).
If [
+

t

[ = 0 =
+
=
t
+
= =
t
.
[
+

t

[ = k suppose proved.
[
+

t

[ = k + 1. Since k + 1 1 = ,=
t
= ,
t
+
= x
t

. Consider

= S
x
() corresponding positive system is

+
= (
+
x) x
Hence

+

t

= (
+

t

) x. In particular [

+

t

[ = k and by induction assumption

there is a g s.t.
t
= g(

), hence

t
= g(S
x
()) = gS
x
()
2
Dene a height function h : R by
h
_

t
t
_
=

t
Theorem 2.8. G, as before.
1. x y g G s.t. x = gy.
2. G = S
x

x
Proof. Let H = S
x

x
G. Pick any t let

t
= Ht
+
11
If
t
,= (obvious if t
+
) then z
t
with minimal height h(z) in
t
.
Let z =

s
s,
s
0. Then
0 < z, z =

s
..
0
s, z
Hence x s.t. x, z > 0. S
x
(z) =

s
s 2
x, z
x, x
. .
>0
x hence height goes down. So
h(S
x
(z)) < h(z). Moreover S
x
(z) Ht. By minimality of heigt of z:
S
x
(z) /
t
= S
x
(z)

By 2.4: z = x = gt
. .
Ht
for some g H. Hence S
x
= gS
t
g
1
= S
t
= g
1
S
x
g H
t = g
1
x Hx H G
(1) follows because if x is positive this proof shows that x G, if x is negative we use t = x
to show that Sx H, t = x G. Hence x = S
x
(x) G.
Finally we have shown that all reections S
x
are in H, so H = G. 2
l : G Z
0
. Now since G = S
x

x
, is l related to group theoretic length in these
generators?
Theorem 2.9. YES
g G : l(g) = mink : x
1
, . . . , x
k
: g = S
x
1
S
x
2
S
x
k

Proof. 2.6: l(S

x
h) l(h) + 1 = if g = S
x
1
S
x
k
then l(g) k. Hence l(g) RHS.
Note: the opposite direction is based on deletion principle.
Assume that g = S
x
1
S
x
k
, x
i
but l(g) < k. Hence j s.t.
l(S
x
1
S
x
j
= j but l(S
x
1
S
x
j+1
= j 1
(we pick the position where the length goes down for the rst time)
= S
x
1
S
x
j
(x
j+1
)

and x
j+1

+
. Find largest i s.t. i j and
S
x
i
(y) = S
x
i
S
x
j
(x
j+1
)

, y = S
x
i+1
S
x
j
(x
j+1
)
+
= y = x
i
= S
x
i+1
S
x
j
. .
T
(x
j+1
) = T(x
j+1
). Hence S
x
i
= TS
x
j+1
T
1
. Then
g = S
x
1
S
x
i
S
x
j
S
x
k
= S
x
1
S
x
i1
S
x
i
TS
x
j+1
S
x
k
= S
x
1
S
x
i1
(TS
x
j+1
T
1
TS
x
j+1
)S
x
j+2
S
x
k
= S
x
1
S
x
i1
TS
x
j+2
S
x
k
2
12
Theorem 2.10. If ,
t
are two simple systems then ! g G s.t.
t
= g().
Proof. Existence is Theorem 2.7. Suppose g, h G satisfy
t
= g() = h(). Then
= h
1
g =
+
= h
1
g(
+
) hence l(h
1
g) = 0 = h
1
g = 1. 2
Corollary 2.11. Consider action of G and X = : is simple . For all X, the
orbit map

: G X : g g()
is a bijection.
Proof. See orbit-stabiliser Theorem. 2
13
3 Generators and Relations
X-free group on a set X
= C
1
x

= Z-innite cyclic group
X = all nite words (including ) in alphabet x
+1
, x
1
as x X without subworts
x
+1
x
1
or x
1
x
+1
.
v w = Reduction of concenated word vw
It is a Theorem that is well-dened.
Universal property:
X
f


_

G
X
!

groups G functions f : X G ! group homomorphism

: X G s.t. a X : (a) = f(a)
X [ R: R is a set of relations i.e. either words in alphabet x
1
: x X or u = v where
u, v are two words. Relation between words and equations:
u u = 1, uv
1
uv
1
= 1 u = v
E.g. dihedral group I
2
(n) =
_
a
..
rot
, b
..
refl.
: b
2
, a
n
, bab
1
= a
1
_
.
Denition. X [ R =
X
/
R
. If the relations are words u
1
, u
2
, . . . then R =

X)H, all u
i
H
H.
Universal property of X [ R:
X [ R
!

X
q

G
functions f : X G s.t. relations in R hold in G for elements f(x), x X then ! group
homomorphism : X [ R G s.t. x X: (xR) = f(x).
14
Example: B(2, 5) =

x, y [ w
5
_
where w is any word in the alphabet x
1
, y
1
.
Universal property: groups G s.t. x G : x
5
= 1, a, b G ! : B(2, 5) G s.t.
(x) = a, (y) = b.
Known: B(2, 5) is either innite or of order 5
34
.
15
4 Coxeter group
A Coxeter graph is a non-oriented graph without loops or double edges s.t. each edge is marked
by a symbol m Z
3
.
Convention: Draw instead of
3

We will study only graphs with nitely many vertices.

Denition. A Coxeter matrix (on set X) is an X X matrix (m
ij
)
i,jX
s.t.
m
ij
Z
1

m
ij
= m
ji
m
ij
= 1 i = j
There is a 1:1 correspondance between Coxeter graphs and Coxeter matrices
Graph Matrix (c
ij
)
X = vertices of the graph
i
m
ij
j c
ij
= m
ij
i
j c
ij
= 2
Example:
1

b
b
b
b
b
b
b
2

77
b
b
b
b
b
b
b
3 4

_
_
_
_
1 3 2
3 1 3 77
3 1 2
2 77 1 1
_
_
_
_
Denition 4.1. The Coxeter group of a Coxeter graph (or corresponding XX-matrix (m
ij
))
is
W = W(graph) = X [ (ab)
m
ab
= 1
Remark 4.2. 1. m
ab
= = no relation
2. a = b, m
aa
= 1 = (aa)
1
= 1 = a
2
= 1
3. no edge m
ab
= 2 = (ab)
2
= 1 ab = b
1
a
1
= ba Therefore:
no edge a, b commute
Lemma 4.3. Let G be a group generated by a
1
, . . . , a
n
, all of order 2. Then G is a quotient of
W(m
ij
= [a
i
a
j
[).
16
Proof. f : 1, . . . , n G where f(i) = a
i
is a function. Then f(i)
2
= a
2
i
= 1 and
(f(i)f(j))
m
ij
= (a
i
a
j
)
[a
i
a
j
[
= 1. So f(i) satisfy relations of the Coxeter group and ! group
homomorphism : W(m
ij
) G s.t. (i) = a
i
(by the universal property). Im a
i
=
is surjective. 2
Big monster M is the largest sporadic simple group,
M 10
53
, M

=
W
_
_
_
_
/
H
Theorem 4.4. G is a nite reection group -simple system in the root system. Then
the natural surjection for -matrix m
x,y
= [S
x
S
y
[, W((m
x,y
)) G : x S
x
is an
isomorphism.
Example: O
2
(R) I
2
() =
_
S
a
..
order 2
, Rot
. .
inf. order
_
. .

=C
2
C

/ Q
W(

) S
a
, S
b
where

ab

/ Q,

= S
a
, Rot

action of W(

) = I
2
() or R
2
is a mess.
But I
2
() has action on R
1
with geometric meaning.
-1 0 1
S
0
S
1
S
0
(x) = x, S
1
(x) = 2x, S
1
S
0
(x) = 2+x translation by 2. So S
1
, S
0
is W(

) inside
the group of motions on R. C

is a subgroup of translations (x x+2n). As C

2
W(

)
you can choose any 1, S
n
where S
n
(x) = 2n x. The fundamental domain is [0, 1], I
2
()
tesselates R
1
by interval [0, 1].
Example: W
_
_
_
_
_
b
c
a
_
_
_
_
_

sidereections across the sides of a triangle with angles

a
,

b
,

c
_
.

where =
a

, =
b

, = c.
1
a
+
1
b
+
1
c
> 1 spherical
= 1 euclidean
< 1 hyperbolic
17
tesselation of the space.
In particular, W
_
_
_
b
c
a
_
_
_
triangles in tesselation
Lemma 4.5. (Deletion Condition)
G - nite reection group, - root system, - simple system.
If x = S
a
1
S
a
n
, a
i
and l(x) < n then i < j s.t.
x = S
a
1


S
a
i


S
a
j
S
a
n
Proof. already proved. 2
Proof. of Theorem 4.3

W(m
ab
)

G
a a
..
we will work in here
S
a
It suces to show that is injective.
Let w = a
1
a
2
a
n
Ker (a
1
2
= 1). Then (w) = S
a
1
S
a
n
= 1. Since det S
a
i
=
1, (1)
n
= 1 n is even, write n = 2k.
Proceed by induction on k =
n
2
k = 1 n = 2 w = a
1
a
2
1 = (w) = S
a
1
S
a
2
S
a
1
= S
a
2
a
1
= a
2
a
i

= a
1
= a
2
w = a
1
2
= 1
k m1: done (induction assumption)
k = m : n = 2m, w = a
1
a
2m
, 1 = (w) = S
a
1
S
a
2m
. Consider
w
1
= a
1
1
wa
1
= a
1
wa
1
= a
2
a
3
a
2m
a
1
Repeating conjugation w.r.t. the rst symbol, we can consider two cases:
Case 1 w = abab ab for some a, b , w = (ab)
m
.
1 = (w) = (S
a
S
b
)
m
= [S
a
S
b
[ = m
ab
[m = w = (ab)
m
= 1 W
Case 2 there is a v W s.t. w
2
= v
1
wv = b
1
b
2
b
3
b
2m
with all b
i
and b
1
,= b
3
.
Let us play the trick: Observe that
x = S
b
1
S
b
m+1
= S
b
2m
S
b
m+2
= y
Indeed y
1
= S
b
m+2
S
b
2m
and xy
1
= S
b
1
S
b
2m
= (w) = 1. Hence x = y.
By deletion principle (note that l(x) m1) there are 1 i < l m+ 1 s.t.
x = S
b
1

S
b
i


S
b
j
S
b
m+1
Consider two cases:
18
Case 2.1 (i, j) ,= (1, m+ 1)
S
b
i
S
b
j
= S
b
i+1
S
b
j1
S
b
i
S
b
j
S
b
j1
S
b
i+1
= 1
Hence w
3
= b
i
b
j
b
j1
b
i+1
ker . It has < 2m terms hence by induction
assumption w
3
= 1. Hence
b
i
b
j
= b
i+1
b
j1
So w
2
= b
1

b
i

b
j
b
2m
. By induction assumption w
2
= 1 and therefore
w = vw
2
v
1
= vv
1
= 1
Case 2.2 (i, j) = (1, m+ 1)
Then w
3
= b
1
b
2
b
m+1
b
m
b
2
Ker has length 2m and we seem to be
stuck. (Hint: eventual contradiction is with b
1
,= b
3
).
Let w
4
= b
1
1
w
2
b
1
= b
2
b
n
b
1
ker . Let us play the trick on
x
1
= S
b
2
S
b
m+1
S
b
m+2
= S
b
1
S
b
m+3
Deletion principle x
1
= S
b
2

S
b

i


S
b

j
S
b
m+2
. Get 2 cases:
Case 2.2.1 (i
t
, j
t
) ,= (2, m+ 2)
S
b

i
S
b

j
= S
b
i

+1
S
b
j

1
S
b

i
S
b

j
S
b
j

1
S
b
i

+1
= 1
w
5
= b
i
b
j
b
j

1
b
i

+1
ker . By induction assumption w
5
= 1
b
i
b
j
= b
i

+1
b
j

1
and w
2
= b
1

b
t
i

b
t
j
b
2m
. By induction assumption w
2
= 1 = w =
vw
2
v
1
= 1.
Case 2.2.2 (i
t
, j
t
) = (2, m+ 2)
w
5
= b
2
b
m+1
b
m+2
b
m+1
b
m
b
3
ker
but of length n. Let w
6
= b
3
w
5
b
3
1
= b
3
b
2
b
3
b
m+2
b
m+1
b
4
ker
and repeat the trick on (w
6
) = 1.
x
2
= S
b
3
S
b
2
S
b
3
S
b
m+1
= S
b
4
S
b
m+2
can delete at i
tt
, j
tt
. Get 2 cases:
Case 2.2.2.1 (i
tt
, j
tt
) ,= (1st incidence of 3, m+ 1)
w
7
= b
i
b
j
b
j

1
b
i

+1
ker
Induction assumption = w
7
= 1 = w
6
= 1 = w
5
= 1 = w
2
=
1 = w = 1.
19
Case 2.2.2.2 (i
tt
, j
tt
) = (first 3, m+ 1)
w
7
= b
3
b
2
b
3
b
m+1
b
m
b
2
ker
Remember w
3
= b
1
b
2
b
3
b
m+1
b
m
b
2
. Then
w
7
= b
3
b
1
w
3
b
3
b
1
= w
7
w
1
3
ker
S
b
3
S
b
1
= 1 = S
b
3
= S
b
1
= b
3
= b
1
= b
3
= b
1
But this is a contradiction!
2
20
5 Geometric representation of W(m
ij
)
Recall some geometry of a vector space V with symmetric bilinear form , . (the eld is R,
dim V - any, , - any)
Still W V we have
W

= x V : a W : a, w = 0
(W

W
Example: R
2
,
__
x
y
_
,
_
a
b
__
= xa yb
W = R
_
1
1
_
= W

= W = W +W

= W ,= R
2
.
Vectors v with x, x = 0 are called isotropic. If x, x , = 0 (x is not isotropic) then
V = Rx x

and
S
x
: V V, S
x
(a) = a 2
a, x
x, x
x
is well-dened and has usual properties
S
x
(x) = x
S
x
[
x
b
ot
= id
S
2
x
= id
S
x
O(V, , )
Start with X X Coxeter matrix (m
ab
) or its Coxeter graph. Let
V (m
ab
) =

aX

a
e
a
:
a
R, finitely many
a
,= 0
e
a
, a X form a basis of V (m
ab
). Symmetric bilinear form , : V (m
ab
) V (m
ab
) R is
dened on the basis by
e
a
, e
b
= cos
_

m
ab
_
a = b = m
ab
= 1 = e
a
, e
a
= 1
m
ab
= 2 = e
a
, e
b
= 0
m
ab
= = e
a
, e
b
= cos
_

_
= cos 0 = 1
21
Lemma 5.1. V (
m
) is euclidean m ,= .
Proof. e
1
, e
2
= cos

m
. If x = e
1
+e
2
then
x, x =
2
e
1
, e
1
+ 2 e
1
, e
2
+
2
e
2
, e
2

=
2
+
2
2 cos
_

m
_
= ( +)
2
2(1 + cos
_

m
_
)
0
and
x, x = 0 = = 0 OR m = , =
2
Lemma 5.2. (V, , ) R-vector space with symmetric bilinear form. Let U V be a nite
dimensional subspace s.t. U U

= 0. Then
V = U U

Note: it breaks down if dim U = .

Example: (V, , ) Hilbert space with Hilbert base e
1
, . . . , U = span(e
i
) V dense.
Then U

= 0, so U U

= 0 but U U

= U ,= V .
Proof. Pick v V , basis e
1
, . . . , e
n
of U.
It suces to nd x
1
, . . . , x
n
R s.t.
v

x
i
e
i
U

j :
_
v

x
i
e
i
, e
j
_
= 0 v, e
j

i
x
i
e
i
, e
j
= 0
The system of n equations

i
x
i
e
i
, e
j
= v, e
i
j = 1, . . . , n has a solution. It holds
true because U U

= 0 which implies that , [

U
is non-degenerate which is equivalent
to det e
i
, e
j
, = 0. 2
Coxeter graph (matrix) = (m
ab
)
a,bX
Coxeter group W() = X : (ab)
m
ab
= 1
vector space with bilinear form
V () =

aX
Re
a
, e
a
, e
b
= cos
_

m
ab
_
Operators

a
= S
e
a
: V () V () : x x 2 e
a
, x e
a
Since : a(e
a
) = e
a
,
a
[
e

a
= I, [
a
[ = 2.
Proposition 5.3.
[
a

b
[ = m
ab
22
Proof.
Case 1 m
ab
,=
5.1
= U = Re
a
Re
b
is euclidean vector space w.r.t , . In particular
U U

= 0 = V () = U U

. Since U

a
,
a
[
U
= I. Similarly,
b
[
U
= I
Since
b
(U

) U

,
a

b
[
U
=
a
[
U

b
[
U
= I I = I. On U, we can do explicit
calculation.
e

m
ab
Hence
a
[
U

b
[
U
= Rot by
2
m
ab
and [
a

b
[ = [
a

b[U
[ = m
ab
Case 2 Exercise: show that in e
a
, e
b
basis

a
[
U

b
[
U
=
_
3 2
2 1
_

_
1 1
0 1
_
so [
a

b[U
[ = [
a

b
[ = .
2
Hence we have a group homomorphism
: W() GL(V ()) : a (a X)
a
is called geometric representation of W().
Suppose =
1

2
(disjoint union of 2 graphs). On the level of Coxeter matrices
X = X
1
X
2
, X
1
X
2
=
and for all a X
1
, b X
2
: m
ab
= 2.
Proposition 5.4. W()

= W(
1
) W(
2
)
Proof. Construct inverse group homomorphism : W() W(
1
) W(
2
), : W(
1
)
W(
2
) W().
If a X, let
(a) =
_
(a, 1) if a X
1
(1, a) if a X
2
Since all relations of W() hold in W(
1
) W(
2
) such uniquely extends to a group homo-
morphism.
If a X
i
X,
i
(a) = a extends to a group homomorphism
i
: W(
i
) W() (all
relations of W(
i
) hold in W() )
Notice that generators of the image of
1
commute with generators of Im
2
. Hence subgroups
Im
1
, Im
2
commute. Hence (x, y) =
1
(x)
2
(y) is a well-dened group homomorphism.
Since a W(), ((a)) = a, = I
W()
. Similarly = I
W(
1
)W(
2
)
. 2
23
is connected if =
1

2
=
1
= or
2
= .
Each has connected components
1
,
2
, . . . s.t.
=
1

3

. . .
Corollary 5.5. In this case W()

= W(
1
) W(
2
) . . . and V () = V (
1
) V (
2
) . . .
with V (
i
)V (
j
) for i ,= j.
Proof. Exercise (use transnite induction). 2
A representation of a group G is the pair (V, ), where V is a vector space and : G GL(V ).
Example: Take a Reection group (G, V ), then (V, i : G GL(V )) is a representation.
If is a Coxeter graph, then (V (), ) is a representation of W().
A subrepresentation of (V, ) is a subspace U V ) s.t. g G : (g)(U) U. So (U, [
GL(U)
)
is a representation.
V is irreducible if V ,= 0 and 0, V are the only subrepresentations. V is called completely
irreducible if for all subrepresentations U V there exists a subrepresentation W V s.t.
V = U W.
Convention: The Coxeter graph is not connected.
Proposition 5.6. If is connected, then any proper W()-subrepresentation of V () lies in
(V ())

.
Exercise: (V ())

is a subrepresentation of V ().
Proof. Let U V = V () be a subrepresentation. Then a X ( = (m
ab
)
a,bX
),
a
(U)
U. Observe that e
a
/ U. Otherwise: b X, m
ab
,= 2 holds:
U
b
(e
a
) = e
a
+ 2 cos

m
ab
. .
,=0
e
b
= e
b
U
Since is connected, this would imply that e
b
U for all b X. So U = V and not a proper
subrepresentation which is a contradiction.
By 5.?, V = Re
a
Re

a
. Pick any z U and write z = e
a
+z
t
where z
t
Re

a
. Then
U
a
(z) = z 2 e
a
, z e
a
= e
a
+z
t
2e
a
, e
a
e
a
2

e
a
, z
t
_
e
a
= e
a
+z
t
Hence z
t
=
1
2
(z +
a
(z)) U and 2e
a
U = = 0. Hence z = z
t
Re

a
and therefore
U Re

a
and U

Re

a
= V

. 2
Corollary 5.7. If is connected and V () nonsingular ( V ()

= 0), then V () is irreducible.

Theorem 5.8. (Maschke)
If G is a nite group and F a eld s.t. char F ,[ [G[ and V is a representation of G over F,
then V is completely irreducible.
24
Proof. Pick a subrepresentation U V .
Observe that direct sum representations V = U W are in bijection with linear operators
p End
F
(V ) s.t. p
2
= p and Im(p) = U:
V = U W p(v) = u where v = u +w is the unique decomposition with u U, w W
V = Im(p) ker(p) p
Pick any vector space decomposition and let p be the corresponding operator, : G GL(X).
Dene p =
1
[G[

xG
(x)p(x
1
)
if t U
p(t) =
1
[G[

xG
(x)(p((x
1
)(t)
. .
U
))
p[
U
=I
=
1
[G[

xG
(x)(x
1
)(t) = t
So p[
U
= I.
since p(t) = t for t U, Im( p) U
by denition of p, Im( p) Im(p) U. Hence Im( p) = U = p
2
= p
p is a homomorphism of representations (commutes with any (x), x G):
(x) p =
1
[G[

yG
(x)(y)p(y
1
)
=
1
[G[

yG
(xy)p((xy)
1
)(x)
=
1
[G[

zG
(z)p(z
1
)(x)
= p(x)
Hence ker p is a G-subrepresentation and V = U ker p.
2
Corollary 5.9. If is connected and [W()[ < then V () is nonsingular (w.r.t , ) and
irreducible (as W()-representation).
Proof. Irreducibility follows from nonsingularity by Corollary 5.?. Suppose V

,= 0 (V =
V ()). By 5.8, V = V

U for some G-subrepresentation U. By 5.6, U V

= U = 0 =
V = V

= , = 0. Nonsence since e
a
, e
a
= 1 and is non-empty. 2
Proposition 5.10. (Schurs Lemma)
If is connected and [W()[ < then End
W()
V () = R.
25
Proof. Pick End
W()
V (). Then a X :
a
=
a
. Therefore: z V () :
(
a
(z)) =
a
((z)).
(z) 2 e
a
, z (e
a
) = (z 2 e
a
, z e
a
) = (z) 2 e
a
, (z) e
a
Hence e
a
, z (e
a
) = e
a
, (z) e
a
. Let z = e
a
, then (e
a
) = e
a
, (e
a
)
. .

e
a
. Therefore, has
an -eigenvector, I End
W()
V ().
0 ,= ker I is a subrepresentation of V (). By 5.9,
ker I = V () = I = 0 = = I
2
Theorem 5.11. If [W()[ < then (V (), , ) is euclidean.
Proof. [W()[ < = X is nite = dim V () = [X[ < .
If =

T
i
, T
i
connected, then V () =

i
V (
i
) with V (
i
)V (
j
), i ,= j. Hence it suces
to consider a case of connected .
Pick any Euclidean form ,
1
over V (). Let us integrate ,
1
over W() so let
u, v
2
=

xW()

x
(u),
x
(v)
1
,
2
is symmetric since ,
1
is
u ,= 0, u, u
2
=

x
(u),
x
(u)
1
. .
>0
hence ,
2
is euclidean.
y W():

y
(u),
y
(v)
2
=

y
(u),
x

y
(v)
1
=

xy
(u),
xy
(v)
1
=

zW()

z
(u),
z
(v)
1
= u, v
2
Hence all
y
O(V (), ,
2
). Remember that all
u
O(V (), , ) and consider
: V V dened by
(u), v
2
= u, v v
End
W()
V () since for all x W():

x
((u)), v
2
=

(u),
1
x
(v)
_
2
=

u,
1
x
(v)
_
2
=
x
(u), v
= (
x
(u)), v
2
26
This implies
x
=
x
. By 5.10, = I for some R, so for all u, v:
u, v
2
= u, v
Since e
a
, e
a
= 1 = e
a
, e
a

2
= =
1
e
a
,e
a
)
2
> 0. 2
Section 4 = Finite Reection group

= Coxeter group
Section 5 = Finite Coxeter group Reection group (V (), (W()))
Problem: could have a kernel. Section 6 rules this out.
Example: Exercise: D
4n

= C
2
D
2n
when n is odd.
Hence distinct reection groups I
2
(2n) = W
_
2n
_
and A
1
I
2
(n) = W
_
_
n
_
_
are
isomorphic as groups.
27
6 Fundamental chamber
Aim: : W() Im() is isomorphic (at least for nite Coxeter groups)
Dene l : W() Z
0
: l(x) = minn : x = a
1
a
n
, a
i
X().
In particular: l(1) = 0, l(a
i
) = 1.
Lemma 6.1. For all a X, g W holds: l(ag) is either l(g) + 1 or l(g) 1.
Proof. g = a
1
a
n
with l(g) = n, then ag = aa
1
a
n
so l(ag) l(g) + 1. Since g =
a(ag), l(g) l(ag) + 1 so l(ag) l(g) 1.
Need to rule out l(ag) = l(g). Consider sign representation:
sgn : W GL
1
(R) : sgn(a) = (1)
Since relations of W hold in GL
1
(R) ((ab)
m
ab
= 1 ((1)(1))
m
ab
= 1 ok) sgn is well-
dened.
If l(x) = n then x = a
1
a
n
and sgn(x) = sgn(a
1
) sgn(a
n
) = (1)
n
= (1)
l(x)
.
If l(ag) = l(g) then sgn(ag) = sgn(g), sgn(a)sgn(g) = sgn(g) = sgn(a) = 1 which is a
contradiction. 2
For each a X() consider half spaces
H
+
a
= u V () : u, e
a
> 0
H

a
= u V () : u, e
a
< 0
the dividing hyperplane H
a
= u V () : u, e
a
= 0
and the cone C =

aX()
H
+
a
which we call the fundamental chamber.
Idea: Let x W :
x
i
(H
a
)

x
(C)
C
28
l(x) is the minimum of hyperplanes one has to cross to walk from C to
x
(C).
Let = (m
ab
)
a,bX
and denote W = W(), V = V ()
Lemma 6.2. [W[ < , a, b X. Then g

a, b
_
W : either
g(H
+
a
H
+
b
) H
+
a
OR
g(H
+
a
H
+
b
) H

a
and l(ag) = l(g) 1
Proof. If g(H
+
a
H
+
b
) , H
+
a
and g(H
+
a
H
+
b
) , H

a
then by convexity of g(H
+
a
H
+
b
):
g(H
+
a
H
+
b
) H
a
,=
Pick z g(H
+
a
H
+
b
) H
a
. Then z H
a
= z, e
a
= 0
z g(H
+
a
H
+
b
) = g
1
(z) H
+
a
H
+
b
=

g
1
(z), e
a
_
> 0,

g
1
(z), e
b
_
> 0
On the other hand

g
1
(z), g
1
(e
a
)
_
= z, e
a
= 0.
As g

a, b
_
:
g
1
(e
a
) = e
a
+e
b
, R
We denote by () a root system of (W) V () and since e
a
, e
b
form a simple system of
() span(e
a
, e
b
), either
0, 0 s.t. not both are 0 and then

g
1
(z), g
1
(e
a
)
_
=

g
1
(z), e
a
+e
b
_
=

g
1
(z), e
a
_
+

g
1
(z), e
b
_
> 0
which is a contradiction
OR 0, s.t. not both are 0 and therefore

g
1
(z), g
1
(e
a
)
_
< 0
which is a contradiction too.
It remains to prove the length statement if g(H
+
a
H
+
b
) H

a
. Inspect U = span(e
a
, e
b
) and
/ = U H
+
a
H
+
b
. This is a dihedral group of order 2m
ab
.
e
a
H
b
U
e
b
abab.../
b/
ba/
/
H
+
a
H

a
g/ H

a
U g = a or g = ab or . . . or g = abab . . . b
. .
m
ab
All such g admit a reduced expression starting with a. Hence l(ag) = l(g) 1. 2
29
Remark 6.3. There are 2 niteness assumptions:
1. W is nite
2. is nite, i.e. [X[ < and a, b X : m
ab
< .
Exercise: W is nite = is nite.
Hint: a ,= b X =
a
,=
b
GL(V ()) = a ,= b W(). Then one needs to show that
m
ab
< . This follows from
a

b
[
span(e
a
,e
b
)
=
_
1 1
0 1
_
.
Note: Lemma 6.2 holds under assumption that is nite.
Proposition 6.4. Let be nite and let C be the fundamental chamber. Then:
1. For all w W and for all u X: either
w(C) H
+
a
OR
w(C) H

a
and l(aw) = l(w) 1
2. for all w W and for all a ,= b X there is a g

a, b
_
s.t.
w(C) g(H
+
a
H
+
b
) and l(w) = l(g) +l(g
1
w)
Proof. Simultaneous induction on l(w):
Base l(w) = 0 = w = 1 = w(C) = C H
+
a
and g = 1 works for (2).
Assumption if l(w) = k 1 then (1) and (2) are assumed to have been proved
Step Let l(w) = k.
1. Write w = dw
t
, d X, l(w
t
) = k 1. Consider two cases:
Case 1: a = d.
Statement (1) for w
t
gives either w
t
(C) H
+
a
or w
t
(C) H

a
.
w
t
(C) H

a
= w(C) = a(w
t
(C)) a(H

a
) = H
+
a
w
t
(C) H
+
a
= w(C) = a(w
t
(C)) a(H
+
a
) = H

a
Case 2: a ,= d.
By (2) for w
t
there is a g

a, b
_
s.t. w
t
(C) g(H
+
a
H
+
b
) and l(w
t
) =
l(g) +l(g
1
w). Then
w(C) = d(w
t
(C)) dg(H
+
a
H
+
d
)
By Lemma 6.2 either dg(H
+
a
H
+
d
) H
+
a
(= w(C) H
+
a
) or
dg(H
+
a
H
+
d
) H

a
and l(adg) = l(dg) 1.
Then w(C) dg(H
+
a
H
+
d
) H

a
and
l(aw) = l(adgg
1
w
t
) l(adg) +l(g
1
w
t
) l(g) +l(g
1
w
t
) = l(w
t
) = k 1
= l(aw) = l(w) 1
30
2. Now assume (1) and (2) for l(w) = k1 and note that (1) for l(w) = k is established.
Consider two Cases:
Case 1: wC H
+
a
H
+
b
. Then g = 1 satises (2).
Case 2: wC , H
+
a
H
+
b
.
H
a
H
b
H
+
a
H
+
b
H
+
a
H

b
H

a
H

b
H

a
H
+
b
Observe that wC H
a
= . wC w(H
+
a
H
+
b
) is in either H
+
a
or H

a
by 6.2.
Wlog wC H

a
.
By part (1): l(aw) = l(w) 1.
Use part (2) on w
t
= aw (by induction assumption) then there is g
t

a, b
_
s.t. w
t
C g
t
(H
+
a
H
+
b
) and l(w
t
) = l(g
t
) +l((g
t
)
1
w).
Claim: g = ag
t
satises (2) for w.
wC = aw
t
C ag
t
(H
+
a
H
+
b
) = g(H
+
a
H
+
b
)
l(w) = 1 +l(aw) = 1 +l(g
t
) +l((g
t
)
1
w
t
)
= 1 +l(g
t
) +l(g
1
w) l(ag
t
) +l(g
1
w)
= l(g) +l(g
1
w) l(gg
1
w) = l(w)
Hence l(w) = l(g) +l(g
1
w).
2
Theorem 6.5. 1. W is nite = V =

gW
gC
2. is nite, then: g, x W : gC xC ,= = g = x
Note: If W is innite

gW
gC is called Tits cone which is a proper subset of V .
Proof.
1. C = W : a X : , a 0.
Recall the height function h : V R : v =

aX

a
e
a


aX

a
.
Pick any v V . Since W is nite, the height achieves a maximum on the set
Wv = g(v) : g W. Let z = g(v) be such a maximum.
31
Claim: z C and this implies that v = g
1
(z) g
1
(C) and V =

gC.
Indeed, otherwise there is an a X s.t. z, e
a
< 0. Then:

a
(z) = h(z) 2 z, e
a
e
a
has hight
h(
a
(z)) = h(z) 2 z, e
a
> h(z)
which is a contradiction to the maximality.
2. Assume gC xC ,= . Then C g
1
xC ,= . If g
1
x = 1 then g = x and we are done.
If g
1
x ,= 1 then there is an a X s.t. l(ag
1
x) = l(g
1
x) 1.
Set w = ag
1
x, then l(aw) = l(w) + 1. By 6.3 ag
1
xC = wC H
+
a
= g
1
xC H

a
.
Then C g
1
xC H
+
a
H

a
= which is a contradiction.
2
Corollary 6.6. is nite. Then : W GL(V ) is injective.
Proof. Suppose (g) = (w). Then gC = wC. Hence g = w by 6.4. 2
Let (G, V ) be a reection group, G O(V ), V is Euclidean. Let V
1
= span(x : x (G)).
Say (G, V ) is essential if V
1
= V .
Corollary 6.7.
(W(), V ())
is a bijection between isomorphism classes of Coxeter graphs with nite W() and equivalence
classes of essential reection groups.
Proof.
(W(), V ())
X = is a simple system in (G), m
ab
= [S
a
S
b
[ (G, V )
By denition, (W(), V ()) is essential. Going (W(), V ()) (X with m
ab
= [S
a
S
b
[)
gives back by 6.5 and 4.3 (?).
(G, V ) = (X, m
ab
) (W(), V ()). Then
V () V

xX

x
e
x

xX

x
x
W() G
a
1
a
2
a
n
S
a
1
S
a
n
This is an isomorphism by 4.3 and 6.5. 2
Some fun bits:
R
n
R
n1
: disjoint union of 2 contractible half-spaces
32
C
n
C
n1
: is connected with fundamantal group Z
Try
C
C = C
R
V
C
H
a
= C
R
H
a
. Then
C
V
0
=
C
V

gW, aX
g(
C
H
a
) is connected.
W acts freely on
C
V
0
and the universal cover of
C
V
0
is simply connected (if W is nite).
One can draw the fundamental group of
C
V
0
/
W
:
H
a
x

a
(x)
T
a
C
C
V
0
[x] = [
a
(x)]
C
V
0
/
W
.
T
a
is a loop in
C
V
0
/
W
. In fact:
B()
. .
Braid group
=
1
_
C
V
0
/
W
_
=
_
T
a
, a X : T
a
T
b
T
a

. .
m
ab
= T
b
T
a
T
b

. .
m
ab
_
V
0

V
0
/
W
gives
1 PB()
. .
pure Braid group
B() W 1
33
7 Classification
Let (V, , ) be an Euclidean vector space. Therefore it has a norm
|v| =
_
v, v
It induces a norm on Lin
R
(V, V ):
|T| = sup
vV \0
|Tv|
/
|v|
1. |Tx| |T| |x|
2. |T| 0 and [|T| = 0 T = 0]
3. |T +S| |T| +|S|
4. T O(V ) = |Tv| = |v| = |T| = 1
5. O(V ) is a closed bounded subset of Lin
R
(V, V ).
Theorem 7.1. Let be nite. Then W() is nite V () is Euclidean.
Proof.
"=" Theorem 5.9
"=" short version: By 6.4 W() acts discretely on the unit sphere in V (). Hence it mus
be nite.
Long version: pick v C s.t. |v| = 1. Pick > 0 s.t. B

(v) C.
g ,= x W = gC xC = = B

(gv) B

(xv) =
= |gv xv| > = |(g) (x)| >
By fact 5, O(V ) is compact. Suppose W is innite, W O(V ) compact. Then there is
a convergent sequence T
1
, T
2
, . . . , T
n
in W s.t. T
i
,= T
j
for i ,= j.
Hence T
i
T O(V ), so there is a N s.t i, j > N : |T
i
T
j
| < .
But T
i
, T
j
(W) with T
i
,= T
j
, so this is a contradiction.
2
is positive denite if V () is Euclidean. This is equivalent to W() being nite. From what
we proved, we may conclude that any nite reection group has a form
(W(
1
) . . . W(
n
), V (
1
) . . . V (
n
) R
k
)
where the
i
are connected positive denite Coxeter graphs and each W(
i
) acting trivially on
V (
j
) with i ,= j and R
k
.
Problem: What are the connected positive denite Coxeter graphs?
34
Proposition 7.2. A =
_
_
_
a
11
a
1n
.
.
.
.
.
.
a
n1
a
nn
_
_
_
symmetric matrix over R.
A is positive denite j 1, . . . , n : det
_
_
_
a
11
a
1j
.
.
.
.
.
.
a
j1
a
jj
_
_
_
> 0.
Proof. Induction on n:
n=1 obvious
n-1 we assume to be done
n "=" A =
_
_
_
_
_
_
a
1n
B
.
.
.
.
.
.
a
n1
a
nn
_
_
_
_
_
_
where B is (n 1) (n 1) matrix.
B represents the same bilinear form restricted to U =
_
_
_
_
_

.
.
.
0
_
_
_
_
_
=
_

_
_
_
_

1
.
.
.

n
_
_
_
:
n
= 0
_

_
in the standard basis of U.
In particular: for all x U 0 : x
T
Bx > 0. By induction assimption:
_
_
_
a
11
a
1j
.
.
.
.
.
.
a
j1
a
jj
_
_
_
> 0
for j < n. Since there exists an orthogonal Q s.t.
QAQ
1
=
_
_
_

1
.
.
.

n
_
_
_
and A is positive denite all
i
> 0, we conclude that det(A) =
1

n
> 0.
"=" By induction assumption B is positive denite, so (U, x, y = x
T
By) is Eu-
clidean. By Lemma ??:
R
n
= U U

..
under x
T
Ay
Any x R
n
can be written as x = y +z, y U, z U

and
x, x = y +z, y +z = y, y +z, z
So it suces to check that for z U

, z ,= 0 and z, z > 0.
Let f
1
, . . . , f
n1
be an orthonormal basis of U, f
n
U

, f
n
,= 0. x
T
Ay in the basis
35
f
1
, . . . , f
n
will look like
C =
_
_
_
_
_
0
I
n
.
.
.
0
0 0 f
n
, f
n

_
_
_
_
_
Hence A is positive denite C is positive denite f
n
, f
n
> 0.
2
Let = (X, m
ab
). We say that
t
= (Y, n
ab
) is a subgraph of if
Y X
a, b X : n
ab
m
ab
Lemma 7.3. If
t
is a Coxeter subgraph of and is positive denite then
t
is positive
denite.
Proof.
2 n
ab
m
ab
= cos

n
ab
cos

m
ab
. Pick any v =

aY

a
e
a
V (
t
) and suppose v, v 0. Let w =

aY
[
a
[e
a
. Then:
w, w
V ()
=

a,bY
[
a
[ [
b
[(cos

m
ab
)

a,bY
[
a
[ [
b
[(cos

n
ab
)

a,bY

b
(cos

n
ab
)
0
Since V () is Euclidean, w = 0. Then all
a
= 0 and v = 0. 2
Let d() = det(2 e
a
, e
n

. .
C()
)
XX
= det(2 cos

m
ab
)
a,bX
.
Note: d() is independent of the order on X one chooses to write a matrix.
Lemma 7.4. Let be of the following form:

1
n
n-1

2
Then d() = 2d(
1
) d(
2
).
36
Proof. Let [X[ = n, call the two vertices n and n 1. Then the matrix C() looks like
_
_
_
_
_
_
_
0
C(
2
)
.
.
.
.
.
.
0
2 1
0 0 1 2
_
_
_
_
_
_
_
Expanding the last row:
d() = 2d(
1
) + (1)
n+(n1)
(1)det
_
_
_
_
_
0
C(
2
)
.
.
.
0
1
_
_
_
_
_
= 2d(
1
) d(
2
)
2
Lemma 7.5. The following Coxeter graphs have d() > 0:
A
n

B
n

4

q
q
q
q
q
D
n

w
w
w
w
w
E
6

E
7

E
8

F
4

4

H
2

5

H
3

5

H
4

5

I
2
(m)
m

Note: A
2
= I
2
(3), B
2
= I
2
(4), H
2
= I
2
(5). Sometimes G
2
= I
2
(6) is used.
Proof. Let x
n
denote d(X
n
).
i
2
(m) = det
_
2 2 cos

m
2 cos

m
2
_
= 4 4(cos

m
)
2
> 0
37
a
1
= det(2) = 2
a
2
= i
2
(3) = 4 4(cos

3
)
2
= 3
Lemma 7.4 allows us to compute a
n
by


A
n

A
n2

A
n1

Then a
n
= 2(a
n1
) a
n2
and by induction a
n
= n + 1.
b
2
= i
2
(4) = 2


B
3

4

A
1

B
2

b
3
= 2b
2
a
1
= 2 2 2 = 2
And for B
n
:


B
n

4

B
n2

B
n1

= b
n
= 2b
n1
b
n2
= b
n
= 2
Before computing D
n
notice that:
d(
1

2
) = d(
1
) d(
2
)
The decomposition for D
n
is:

H
H
H
H
H

H
H
H
H
H

A
1

A
n3

A
n1

Therefore d
n
= 2a
n1
a
1
a
n3
= 2n 2(n 3 + 1) = 4.
38
Similarly: e
n
= 2a
n1
a
2
a
n4
= 9 n.
Hence e
n
is positive for n 8.
h
2
= i
2
(5) = 4 4 cos
2
5
=
5

5
2
h
3
= 2h
2
a
1
= 2
5

5
2
= 3

5 > 0
h
4
= 3h
3
h
2
= 2(3

5)
5

5
2
=
73

5
2
> 0
Notice that h
5
= 3h
4
h
3
= 5 2

5 < 0.
2
Lemma 7.6. The following Coxeter graphs have d() < 0:

A
n

z
z
z
z

h
h
h
h

z
z
z
z

h
h
h
h

B
n

4

z
z
z
z

C
n

4

4

h
h
h
h

z
z
z
z

B
n

z
z
z
z

h
h
h
h

h
h
h
h

E
6


z
z
z
z

h
h
h
h

E
7

z
z
z
z

h
h
h
h

E
8

z
z
z
z
39

F
4

4

G
2

6

H
3

5

H
4
= H
5

5

Proof.
We have computed e
8
= e
9
= 9 9 = 0.

h
4
= h
5
= 4 2

5 < 0
a
n
= det
_
_
_
_
_
_
_
_
2 1 1
1 2 1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1 1 2
_
_
_
_
_
_
_
_
= 0 because the sum of the rows is zero

b
n
= 2b
n
a
1
b
n2
= 4 4 = 0

d
n
= 2d
n
a
1
d
n2
= 2 4 2 4 = 0
e
6
= 2e
6
a
5
= 0
e
7
= 2e
7
d
6
= 0

f
4
= 2f
4
b
3
= 0
g
2
= 2i
2
(6) a
1
= 0

h
3
= 2h
3
a
2
= 3 2

5 < 0
c(

C
n
) looks like
_
_
_
_
_
_
0
c(B
n1
) 0
0
2 2
0 2 2
_
_
_
_
_
_
.
Similarly to Lemma 7.4, expanding the last row gives
d(

C
n
) = 2d(B
n
) 2d(B
n1
) = 2 2 2 2 = 0
2
40
Theorem 7.7. The connected Coxeter graphs which give a nite W() are precisely the
graphs listed in Lemma 7.5.
Proof. The graphs on 7.5 are closed under taking connected Coxeter subgraphs. Therefore
7.2 tells us that all c() are positive denite in 7.5 and 7.1 tells us that all W() are nite in
7.5.
Now let be a connected graph with nite W(). Then c() is positive denite by 7.1. In
particular, contains no subgraphs in 7.6
No

A
n
= has no cycles
No

D
4
= has no vertex with 4 edges
No

D
n
, n 5 = contains at most one vertex with 3 edges.
Case 1 No vertex has 3 edges, hence looks like this:

x
1

x
2

x
m

If has 2 vertices it is either A

1
or I
2
(m), both in Lemma 7.5.
Let contain 3 vertices (m 2).
No

G
3
= edges may have multiplicity at most 5
No

C
n
= at most one edge has multplicity 4.
Case 1.1 No edge of multiplicity 4 = = A
n
Case 1.2 There is one edge of multiplicity 4:
if it is on a side, = B
n
if it is in the middle then no

F
4
= = F
4
Case 1.3 There is one edge of multiplicity 5:
No

H
3
= the multiple edge is on a side and no

H
4
= = H
3
or H
4
Case 2 there is a vertex with 3 edges. Then look like this:

Let b denote the vertices going up, a the vertices going left and c the vertices going
right. Then we have a +b +c + 1 vertices in total and wlog a b c.
No

B
n
= has no multiple edges.
No

E
6
= c = 1
No

E
7
= b 2
if b = c = 1 = = D
a+3
if b = 2, c = 1, then no

E
8
= a < 5. As a b = 2 there are 3 cases:
1. a = 2 = = E
6
41
2. a = 3 = = E
7
3. a = 4 = = E
8
2
crystallographic NON crystallographic
A
n
, B
n
, D
n
, F
4
, E
n
, G
2
= I
2
(6) I
2
(m) for m = 5, m 7, H
3
, H
4
d() Z d() / Z
the character table of is integral non-integral
all but one representation of W()
can be realised by integral matrices
42
8 Crystallographic Coxeter groups
Let V be a vector space over R.
A lattice in V is an abelian subgroup L of V s.t. L = e
1
, . . . , e
n
where e
1
, . . . , e
n
is a basis of
V .
Informally, L is Z
n
inside R
n
.
Denition. W() is crystallographic if there is a lattice L V () s.t.
g W() : g(L) L
Informally this means that there is a basis e
1
, . . . , e
n
of V () s.t. (g) is an integer matrix for
each g W().
Lemma 8.1. Let [X[ < . If W() is crystallographic then a, b X :
m
ab
2, 3, 4, 6,
Proof. Assume W() is crystallographic. Then g W : (g) can be written as an integer
matrix, in particular Tr(()) Z.
If m
ab
= then we are done.
Suppose m
ab
= = U = span(e
a
, e
b
) is Euclidean = W() = U U

.
In the basis e
a
, e
b
, any basis of U

:
(S
a
S
b
) =
_
_
cos
2
m
ab
sin
2
m
ab
0
sin
2
m
ab
cos
2
m
ab
0
0 0 I
_
_
Tr((S
a
S
b
)) = [X[ 2 + 2 cos
2
m
ab
hence
2 cos
2
m
ab
Z cos
2
m
ab

1
2
Z m
ab
2, 3, 4, 6
2
Note: = is true but the proof is harder unless W() is nite.
Theorem 8.2. Let W() be a nite Coxeter group. Then W() is crystallographic a, b
X : m
ab
2, 3, 4, 6.
Proof.
"=" Lemma 8.1
43
"=" Let us choose a new basis
a
= c
a
e
a
, a X, c
a
R 0. Then

a
(
b
) =
b
2
b
, e
a
e
a
=
b
2

a
,
b

a
,
a

a
Since
a
= S
e
a
= S

a
and (a
1
a
n
) =
a
1

a
n
it suces to ensure that for all
a, b X :
2

a
,
b

a
,
a

Z
Wlog is connected.
m
ab
= 2 =
a
,
b
= 0 = 2

a
,
b
)

a
,
a
)
= 0
m
ab
= 3 =
a
,
b
= c
a
c
b
e
a
, e
b
= c
a
c
a
cos

m
ab
. .
=
1
2
=
c
a
c
b
2
= 2

a
,
b

a
,
a

=
c
a
c
b
c
a
c
a
=
c
b
c
a
Need in this case c
a
= c
b
m
ab
= 4 =
a
,
b
= c
a
c
b
(cos

4
) =
c
a
c
b

2
= 2

a
,
b

a
,
a

2
c
b
c
a
Need c
a
=

2c
b
or c
b
=

2c
a
m
ab
= 6 =
a
,
b
=

3
2
c
a
c
b
= 2

a
,
b

a
,
a

= 2
c
a
c
b

3
2
c
a
c
a
=

3
c
b
c
a
Need c
a
=

3c
b
or c
b
=

3c
a
In each component of we choose
Case 1 A
n
, D
n
, E
n
: choose all c
a
= 1.
Case 2 B
n
, F
4

1
..
choose c
a
=1
4

2
..
choose c
a
=

2
Case 3 In G
2
= I
2
(6):
a
6
b

a
=

3e
a
,
b
= e
b
2
Corollary 8.3. The nite crystallographic reection groups are W() where is either
44
A
n
, n 1
B
n
, n 2
D
n
, n 4
E
n
, 6 n 8
F
4
G
2
Note: In B
n
, n 3 there are 2 non-isomorphic choices of crystallographic structure
1
4
2 3
n

1
= e
1
,
2
=

2e
2
, . . . ,
k
=

2e
k
which is called the crystallographic reection group
B
n

1
=

2e
1
,
2
= e
2
, . . . ,
k
= e
k
which is called the crystallographic reection group C
n
semisimple Lie algebra
Weyl group and Root lattice

crystallographic reflection group

Serre Theorem

KacMoody algebra

Nakajima construction
1
2

We have seen A
n
, B
n
, D
n
, I
2
(m) explicitely.
Can we compute [W()[, [()[ for all ?
45
9 Polynomial invariants
Let G be a nite group acting on V
R
, dim V < . Let S = R[V ], algebra of all polynomial
functions on V . Let e
1
, . . . , e
n
be a basis of V and X
i
: V R the linear maps which form
the dual basis. Then X
i
(e
j
) =
ij
. In fact, S = R[X
1
, . . . , X
n
].
G acts on the algebra S: g G, F S, v V : g F(v) = F(g
1
v).
Exercise: check that G acts on S by algebra automorphisms.
Note: for F S the following are equivalent:
1. g G : gF = F
2. F is constant on G-orbits in V .
We call such functions invariant and they form a subalgebra S
G
.
We will prove that S
G
= R[z
1
, . . . , z
k
] (when G is a reection group). In fact
S
G

= R[z
1
, . . . , z
k
] G is a reection group
(no proof)
S is a graded algebra, i.e. S =

k=0
S
k
where S
k
consists of polynomials where each monomial
has degree k.
Example:
2010 S
0
x
2
1
+x
2
2
+x
1
x
3
S
2
1 +x
1
S
0
S
1
but not in any S
i
If F S
i
we say deg F = i.
Lemma 9.1. Let F S
G
, F =
m

j=0
F
j
, F
j
S
j
. Then j: F
j
S
G
.
Proof. GR

acts on V : (g, )v = g(v) = g(v). Hence GR

also acts on S:
(g, ) F : v F(g
1

1
v)
In particular, f S
j
: (1, ) f =
j
f.
Since actions of G and R

commutes, R

:
(1, )F =
m

j=0

j
F
j
S
G
1, . . . , m+ 1 : m+ 1 dierent real numbers
46
1 F = F
0
+. . . +F
m+1
S
G
2 F = F
1
+
1
2
F
1
+. . . +
1
2
m+1
F
m+1
S
G
. . .
(m+ 1) F = F
0
+
1
m+1
F
1
+. . . +
1
(m+1)
m+1
F
m+1
S
G
If M =
_
_
_
_
_
_
1 1
1
2

_
1
2
_
m+1
.
.
.
.
.
.
1
m+1

_
1
m+1
_
m+1
_
_
_
_
_
_
R
(m+1)(m+1)
is a VanDerMonde matrix then
M
_
_
_
F
0
.
.
.
F
m
_
_
_
. .
S
m+1

_
S
G
_
m+1
Since M is invertible,
_
_
_
F
0
.
.
.
F
m
_
_
_
= M
1
_
_
_
H
0
.
.
.
H
M
_
_
_
where H
i
S
G
.
Hence all F
i
S
G
. 2
Lemma 9.1 means that S
G
is a graded subalgebra of S, i.e.
S
G
=

k=0
(S
G
S
k
) and S
G
k
= S
G
S
k
Let S
G
+
=

k=1
S
G
k
be the invariants of positive degree. Let I =
_
S
G
+
_
be the ideal of S generated
by S
G
+
. By Hilberts Basis Theorem, any ideal in S is generated by nitely many elements.
Say I = (f
1
, . . . , f
k
)
S
, f
i
S
G
. Write f
i
=

j
f
ij
where f
ij
S
j
. With Lemma 9.1:
f
ij
S
G
j
, j < 0.
Hence I = (I
1
, . . . , I
r
)
S
where I
j
S
G
d
j
. If for any k: I
k
(I
1
, . . . , I
k1
, I
k+1
, . . . , I
r
)
S
we throw
I
k
away. Wlog, I
1
, . . . , I
r
is a minimal system of such generators.
Denition. I
1
, . . . , I
r
are called fundamental invariants if they are a minimal system of
homogeneous invariant generators of (S
G
+
)
S
.
Coinvariants: S
G
=
S
/
I
Integral:
_
G
: S S
G
_
G
F =
1
[G[

gG
gF
47
Note that F S
G
= gF = F =
_
G
F = F and H S :
_
G
(FH) =
1
[G[

g
(gF)(gH) = F
1
[G[

g
gH = F
_
G
H
So
_
G
is an S
G
-linear map S S
G
.
Proposition 9.2. Let G be any nite group acting on V , I
1
, . . . , I
r
S
G
= R[V ]
G
fundamental
invariants. Then:
F S
G
p(z
1
, . . . , z
r
) R[z
1
, . . . , z
r
] s.t. F = p(I
1
, . . . , I
r
)
Proof. Lemma 9.1: we may restrict to homogeneous F. We do induction on d = deg(F).
d = 0: = F = const = p = const 1 does the job
d < N: is done
d = N > 0: F S
G
= F (S
G
)
S
= F =
r

j=1
H
j
I
j
, H
j
S. Since d = deg(F), d
j
=
deg(I
j
):
F =
r

j=1
(H
j
)
dd
j
I
j
and therefore
F =
_
G
F =
r

j=1
I
j
_
G
(H
j
)
dd
j
By induction assumption j p
j
R[z
1
, . . . , z
r
] s.t.
_
G
(H
j
)
dd
j
= p
j
(I
1
, . . . , I
r
). Hence
p =
r

j=1
z
j
p
j
does the job.
2
So for any real representation V of any nite group G the algebra map
R[z
1
, . . . , z
r
] S
G
, z
i
I
i
is surjective.
1. Injectivity holds only for reection groups
2. Hilberts 14-th problem is that S
G
is a nitely generated algebra for any group
3. S
G
0
= R 1 constant functions
48
4. f S
G
1
, f ,= 0 = V = Ker(f) R. Hence S
G
1
,= 0
G
V has a trivial constituent
5. if , is a Euclidean form on V then [x, y] =
_
G
x, y makes G orthogonal. Hence
F(x) = [x, x] dened F S
G
2
.
Lemma 9.3. Let P S, L S
1
0, if x V : L(x) = 0 = P(x) = 0 then L[P.
Proof. S = R[x
1
, . . . , x
n
]. Wlog L = x
n
+ b(x
1
, . . . , x
n1
) with ,= 0. Let us divide P by
L with remainder by replacing every x
n
with
1

b(x
1
, . . . , x
n1
). x
n
=
1

L
1

b(x
1
, . . . , x
n1
).
Example: L = x
1
x
2
P = x
2
1
= x
1
x
1
= x
1
(L +x
2
)
= x
1
L +x
1
x
2
= x
1
L + (L +x
2
)x
2
= (x
1
+x
2
)L +x
2
2
P = AL + B, A R[x
1
, . . . , x
n
], B R[x
1
, . . . , x
n1
]. If B = 0 we are done. If B ,= 0 pick
a
1
, . . . , a
n1
R s.t. B(a
1
, . . . , a
n1
) ,= 0.
Let a
n
=
1

b(a
1
, . . . , a
n1
) = L(a
1
, . . . , a
n
) = 0 = P(a
1
, . . . , a
n
) = 0.
But P(a
1
, . . . , a
n
) = B(a
1
, . . . , a
n1
) ,= 0 which is a contradiction. 2
Proposition 9.4. Let G = W be a nite reection group. Let J
1
, . . . , J
k
S
W
s.t.
J
1
/ (J
2
, . . . , J
k
)
S
W. If P
i
S
b
i
s.t.
k

i=1
P
i
J
i
= 0 then P
1
(S
W
+
)
S
.
Proof. Observe that J
1
/ (J
2
, . . . , J
k
)
S
: if J
1
=
k

t=2
H
t
J
t
with H
t
S then
J
1
=
_
J
1
=

t
J
t
_
H
t
(J
2
, . . . , J
k
)
S
W
Proceed by induction on deg(P
1
):
deg(P
1
) = 0: = P
1
= c 1 = c J
1
=
k

t=2
(P
t
)J
t
= c = 0 because J
1
/ (J
2
, . . . , J
k
)
S
=
P
1
= 0 (S
W
+
)
S
deg(P
1
) < N: done
deg(P
1
) = N: Let W be the root system of G. V let L

S
1
be L

(v) = , v.
L

= 0 = v = S

(v) = v
= j : (S

P
j
P
j
)(v) = P
j
(S
1

v) P
j
(v) = 0
With Lemma 9.3: S

P
j
P
j
= L

P
j
for some P
j
S.
Note that deg(P
j
) < deg(P
j
)
P
1
J
1
+. . . +P
k
J
k
= 0
(S

P
1
)J
1
+. . . + (S

P
k
)J
k
= 0
49
(P
1
S

P
1
)J
1
+. . . + (P
k
S

P
k
)J
k
= 0
Hence L

(P
1
J
1
+. . .+P
k
J
k
) = 0. Since S is a domain P
1
J
1
+. . . P
k
J
k
= 0. By induction
assumption, P
1
(S
W
+
)
S
. Hence S

P
1
P
1
= L

P
1
(S
W
+
)
S
.
For each polynomial F, denote

F = F + (S
W
+
)
W
S
W
. Hence : S

P
1
=

P
1
.
Since W is generated by S

, g W : g

P
1
=

P
1
. Hence
_
W

P
1
=

P
1
(W acts on S
W
and
_

F =

_
F)
So P
1

_
W
P
1
+ (S
W
+
)
S
(S
W
+
)
S
.
2
Lemma 9.5. (Eulers formula)
F S
k
= R[x
1
, . . . , x
n
]
k
, then
n

j=1
x
j
F
x
j
= kF
Proof. Both sides are linear in F, hence it suces to verify the formula on monomials. Let
F = x
a
1
1
x
a
n
n
,

a
i
= k. Then
x
j
F
x
j
= a
j
F
Hence
n

j=1
F
x
j
= (

a
i
)F = kF. 2
Proposition 9.6. I
1
, . . . , I
r
fundamental invariants for a (nite) reection group (W, V ). If
P(I
1
, . . . , I
r
) = 0 for some P R[z
1
, . . . , z
r
] then P = 0.
Note: I
1
, . . . , I
r
are algebraically independent.
Proof. Consider a grading on R[z
1
, . . . , z
r
] with deg z
i
= d
i
= deg I
i
. Let P =

P
j
, P
j
homogeneous of degree j. Then
P(I
1
, . . . , I
r
) =

j
P
j
(I
1
, . . . , I
r
)
. .
S
j
= 0
Hence P
j
(I
1
, . . . , I
r
) = 0. WLOG, we assume P is homogeneous.
Let P
(j)
=
P
z
j
and f
j
= P
(j)
(I
1
, . . . , I
r
) S
W
is homogeneous. Let us ddle around with the
ideal
K = (f
1
, . . . , f
r
)
S
W S
W
WLOG, f
1
, . . . , f
m
is a minimal system of generators of K, m r. Note that deg f
i
=
deg P deg I
i
and f
i
is homogeneous. If i m + 1: f
i
K = (f
1
, . . . , f
m
)
S
W so there are
Q
ij
S
W
, Q
ij
homogeneous of degree deg I
i
deg I
j
s.t. f
i
=
m

j=1
Q
ij
f
j
.
P(I
1
, . . . , I
r
) = 0 = 0 =
P(I
1
, . . . , I
r
)
x
k
=
r

i=1
P
z
i
(I
1
, . . . , I
r
)
I
i
x
k
k 1, . . . , n
50
Hence:
k :
m

i=1
f
i
I
i
x
k
+
r

i=m+1
m

j=1
Q
ij
f
j
I
i
x
k
= 0
k :
m

i=1
f
i
_
_
I
i
x
k
+
r

j=m+1
Q
ji
I
j
x
k
_
_
= 0
By 9.4, k i 1, . . . , m :
I
I
x
k
+
r

j=m+1
Q
ji
I
j
x
k
(S
W
+
)
S
Hence for some H
k
t
S :
i, k :
I
I
x
k
+
r

j=m+1
Q
ji
I
j
x
k
=
r

t=1
I
t
H
k
t
Multiply each equality by x
k
and sum over k. By Eulers formula:
i : d
i
I
i
+
r

j=m+1
d
j
Q
ji
I
j
=
r

t=1
I
t

H
t
where

H
t
have no free terms (

H
t
=

x
k
H
k
t
). I
i
enters both sides of the equality. If we consider
the homogeneous degree d
i
parts of both sides, then
d
i
I
i
+
r

j=m+1
c
j
I
j
=

t
I
t
b
t
where b
i
has to be zero because

H
i
has no free terms. Therefore
I
i
=

j,=i
I
j
G
j
= I
i
(I
1
, . . . , I
i1
, I
i+1
, . . . , I
r
)
which is a contradiction. Therefore all
P
x
k
= 0 = P = 0. 2
Example: S
n
acting on R
n
by permutations.
R
n
= U R where U =
_

_
_
_
_

1
.
.
.

n
_
_
_
:

i
= 0
_

_
and R = R
_
1
.
.
.1
_
(S
n
, U) is then a type A
n1
essential reection group with elementary symmetric functions:

1
(x
1
, . . . , x
n
) = x
1
+. . . +x
n

2
(x
1
, . . . , x
n
) = x
1
x
2
+x
1
x
3
+. . . +x
n1
x
n
. . .

n
(x
1
, . . . , x
n
) = x
1
x
n
51
is a system of fundamental invariants. The power functions
j
= x
j
1
+. . . +x
j
n
, j = 1, . . . , n is
another fundamental system.
A
n1
:
1
[
U
= 0

2
[
U
, . . . ,
n
[
U
are fundamental invariants with fundamental degrees 2, 3, . . . , n.
B
n
: group has more transformations x
i
x
i
, S
W(B
n
)
= (S
S
n
)
C
n
2
= R[
1
, . . . ,
n
]
C
n
2
.
Hence the fundamental invariants are
j
(x
2
1
, . . . , x
2
n
) with degrees 2 1, 2 2, . . . , 2 n.
D
n
: in D
n
, instead of the full group C
n
2
, it is a subgroup of index 2. So
n
= x
1
x
n
S
W(D
n
)
.
So the fundamental invariants are
1
(x
2
i
), . . . ,
n1
(x
2
i
),
n
(x
i
) with degrees 2 1, 2
2, . . . , 2 (n 1), n.
Theorem 9.7. Let (G, V ) be a nite reection group, n = dim V, S = R[V ], I
1
, . . . , I
r
fundamental invariants. Then
S
G
= R[I
1
, . . . , I
r
] and r = n
Proof. The natural algebra homomorphism
: R[z
1
, . . . , z
r
] S
G
, F(z
1
, . . . , z
r
) F(I
1
, . . . , I
r
)
is surjective by 9.2 and injective by 9.6. Hence is an isomorphism. It remains to show that
n = r.
Consider the eld of rational functions
F = R(x
1
, . . . , x
n
) =
f
g
: f, g S, g ,= 0
Transcendence degree is trdeg
R
F = n. The smaller eld A = R(I
1
, . . . , I
r
) F, trdeg
R
A = r.
From Galois theory, trdeg
A
F = trdeg
R
Ftrdeg
R
A = nr. It suces to show that trdeg
A
F =
0, i.e. every element of F is algebraic over A.
Pick f F, G acts on F with F
G
= A. Consider h(z) F[z] dened by
h(z) =

gG
(z g.f)
Since t G, t.h =

gG
(z tg.f) = h = h A[z].
h(f) =

gG
(f g.f) = (f f)

g,=1
(f g.f) = 0
Hence f is algebraic over A. 2
Let d
1
, . . . , d
n
with d
i
= deg I
i
be the fundamental degrees.
Theorem 9.8. If I
t
1
, . . . , I
t
n
is another system of fundamental invariants with degrees d
t
1
, . . . , d
t
n
, d
t
j
=
deg I
t
j
then S
n
s.t.
j : d
j
= d
t
(j)
52
Proof. By 9.7 there are F
1
, . . . , F
n
, G
1
, . . . , G
n
R[z
1
, . . . , z
n
] s.t.
I
j
= F
j
(I
t
1
, . . . , I
t
n
) and I
t
j
= G
j
(I
1
, . . . , I
n
) j
Then j : I
j
= F
j
(G
1
(I
1
, . . . , I
n
), . . . , G
n
(I
1
, . . . , I
n
)). Hence, since I
1
, . . . , I
n
are algebraically
independent:
z
j
= F
j
(G
1
(z
1
, . . . , z
n
), . . . , G
n
(z
1
, . . . , z
n
))
and therefore
z
j
z
k
=
n

s=1
F
j
z
s
(G
1
(z
1
, . . . , z
n
), . . . , G
n
(z
1
, . . . , z
n
))
G
s
z
k
(z
1
, . . . , z
n
)
And on matrix level:
I
n
=
_
F
i
z
s
_
I

1
,...,I

_
G
s
z
k
_
I
1
,...,I
n
Mat
n
(S)
Hence det
_
F
i
z
s
_
is invertible in S = R[x
1
, . . . , x
n
]. Hence S
n
s.t.
F
1
z
(1)

F
n
z
(n)
= c 1 + higher degree terms, c ,= 0
Since deg I
t
i
= deg z
i
= d
t
i
, deg F
i
= deg I
i
= d
i
:
i : d
i
= d
t
(i)
2
53
10 Fundamental degrees
Discuss the role of d
1
, . . . , d
n
.
Lemma 10.1. A nite group G acts in a nite dimensional vectorspace over C
(by : G GL
n
(C)). Then x G : (x) is diagonalisable with eigenvalues of absoulute
value 1.
Proof. [G[ = n < = x
n
= 1 = (x)
n
= I
n
.
Hence the minimal poynomial of (x),
x
(z) divides z
n
1. Therefore eigenvalues satisfy

n
1 = 0, hence
n
= 1 = [[
n
= 1 = [[ = 1.
Further, since z
n
1 has no multiple rootes,
x
(z) has no multiple roots and all Jordan blocks
of (x) have size 1. 2
Let C[[t]] =

n=0

n
z
n
be the ring of formal power series. If V =

n=0
V
n
is a graded vector
space with dimV
n
< , the "right" notion of dimension of V is the Poincar polynomial
p
V
(t) =

n=0
dimV
n
t
n
Example:
p
C[z]
(t) = 1 +t
d
+t
2d
+t
3d
+. . . =
1
1t
d
where deg z = d.
p
V W
(t) = p
V
(t) +p
W
(t)
p
V W
(t) = p
V
(t)p
W
(t)
p
C[z
1
,...,z
n
]
=

j
p
C[z
j
]
(t) =
1
1t
d
1

1
1t
d
n
where deg z
i
= d
i
, C[z
1
, . . . , z
n
] = C[z
1
] . . .
C[z
n
].
If d
1
= . . . = d
n
= 1 then p
C[z
1
,...,z
n
]
(t) =
1
(1t)
n
.
Example: If G acts on V , then x G let

x
(t) = det(I
V
t(x))
1
=

j
1
1
j
t
=

j
(1+
j
t+
2
j
t
2
+. . .) =

j=0
t
j
(

a
1
+...+a
n
=j

a
1
1

a
n
n
)
where
1
, . . . ,
n
are the eigenvalues of (x).
Theorem 10.2. Let (G, V ) be a nite reection group, d
1
, . . . , d
n
the fundamental degrees.
Then
1
[G[

xG

x
(t) =
n

j=1
1
1 t
d
j
C[[t]]
54
Proof. By 9.7 and the Example before, the RHS is p
S
G(t).
x[
S
1
belongs to
_
_
_

1
0
.
.
.
0
n
_
_
_
in some basis e
1
, . . . , e
n
of S
1
. Then e
a
1
1
e
a
n
n
with
a
1
+. . . +a
n
= j form a basis of S
j
and
xe
a
1
1
e
a
n
n
=
a
1
1

a
n
n
e
a
1
1
e
a
n
n
hence
tr(x[
S
j
) =

a
1
+...+a
n
=j

a
1
1

a
n
n
Hence
LHS =
1
[G[

xG,j=0
tr(x[
S
j
)t
j
Remember
j
=
_
G
: S
j
S
G
j
,
j
(v) =
1
[G[

gG
gv. In particular LHS =

j=0
tr(
j
)t
j
.

2
j
=
j
=
j

_
I 0
0 0
_
and tr
j
= rk
j
= dimS
G
j
Hence LHS = p
S
G(t). 2
Corollary 10.3. [G[ = d
1
d
n
.
Corollary 10.4. d
1
+. . . +d
n
= n +
1
2
[(G)[
. .
number of reflections
.
Proof. Multiply 10.2 by (1 t)
n
:
1
[G[

xG

x
(t)(1 t)
n
=
n

j=1
1 t
1 t
d
j
1
[G[

xG
(1 t)
n
det(1 tg)
=
n

j=1
1
1 +t +t
2
+. . . +t
d
j
1
. .
=:(t)
=
1
[G[
( 1
..
g=1
+
1
2
[[
1 t
1 +t
. .
gS
x
, x
+ h(t) (1 t)
2
. .
others have 1 as eigenvalue of mult.2
)
[h(1) is dened ]
Plug in t = 1, then
1
[G[
=
n

j=1
1
d
j
= [G[ = d
1
d
n
Apply

t
:
[[
2[G[
2
(1 +t)
2
+

h(t)(1t) =
n

j=1
(t)(1+t+. . . +t
d
j
1
)
_
1
1 + 2t + 3t
2
+. . . + (d
j
1)t
d
j
2
(1 +t +. . . +t
d
j
1
)
2
_
55
Plug in t = 1:

[[
4[G[
=
n

j=1
d
j
d
1
d
n
_

(d
j
1)d
j
2d
2
j
_
As [G[ = d
1
d
n
[[
4
=
n

j=1
d
j
1
2
=
1
2
((

d
j
) n)
2
Example:
I
2
(m) has fundamental degrees 2, m.
[G[ = 2m, [[ = 2((m+ 2) 2) = 2m
A
n
has fundamental degrees 2, 3, . . . , n + 1
[G[ = (n+1)!, [[ = 2((2+3+. . .+n+1)n) = 2
_
n(n + 3)
2
n
_
= n
2
+3n2n = n
2
+n
56
11 Coxeter elements
Let (G, V ) be a nite reection group, V a root system.
A Coxeter element is w = S

1
S

n
for some simple system =
1
, . . . ,
n
.
There are two choices involved:
simple system
order on
In S
n
the choices will lead to
S

1
= (a
1
a
2
), S

2
= (a
2
a
3
) . . . S

n1
= (a
n1
a
n
)
where a
1
, . . . , a
n
= 1, . . . , n. Hence w = (a
1
. . . a
n
). In S
n
, Coxeter elements are cycles
of order n.
Lemma 11.1. Let G a group, X a nite forest (graph without cycles). Suppose f : V er(X)
G is a function s.t. if two vertices a, b V er(X) are not connected, then
f(a)f(b) = f(b)f(a)
Then for any choice of a total order on the vertices of X, say V er(X) = a
1
, . . . , a
n
, the
conjugacy class Cl
G
(g) of g = f(a
1
) f(a
n
) does not depend on the order chosen.
Proof. Induction on n = [V er(X)[.
n=1: g = f(a
1
), nothing to prove
n<k proved
n=k Easy case: X has no edges, then all f(a)f(b) = f(b)f(a), so g = f(a
1
) f(a
n
) is
independent of the choice of the order.
If X has an edge, there is a vertex b connected with exactly one other vertex a. For each
function : 1, . . . , n V er(X), let g

= f((1)) f((n)).
Pick a particular : 1, . . . , V er(X) s.t. (n) = b, (n 1) = a. It suces to nd
a s G s.t. g

= sg

s
1
.
Case 1 (n) = n, (n 1) = a.
Let Y = X b and f
t
: Y G be f
t
(c) =
_
f(a)f(b) c = a
f(c) c ,= a
. Then
g

= f((1)) f((n 2))f(a)f(b)

= f
t
((1)) f
t
((n 1))
By induction assumption it is conjugate to f
t
((1)) f
t
((n 1)) = g

.
57
Case 2 (n) = b, (j) = a, j < n 1.
Let X
0
= (k) : k < j, X
1
= (k) : j < k < n. Then dene

i
:
1
(X
i
) X
i
,
i
(c) = (c)
Then
g

= g

0
f(a)g

1
f(b)
= g

0
f(a)f(b)g

1
g

1
(g

0
f(a)f(b)g

1
)g
1

1
= g

1
g

0
f(a)f(b)
This is conjugate to g

by Case 1.
Case 3 (j) = b, j < n, X
0
, X
1
,
0
,
1
as above. Then
g

= g

0
f(b)g

1
g

1
g

0
f(b)
done by Case 2.
2
Proposition 11.2. Coxeter elements form a conjugancy class in G.
Proof. Let C = Coxeter elements . Then
xS

1
S

n
x
1
= (xS

1
x
1
) (xS

n
x
1
) = S
x(
1
)
S
x(
n
)
C
Hence xCx
1
= C, so C is the union of some conjugacy classes. Let w
1
= S

1
S

n
, w
2
=
S

1
S

n
C. Then there is a g G s.t. g(
1
), . . . , g(
n
) =
1
, . . . ,
n
. Hence gw
1
g
1
=
S
g(
1
)
S
g(
n
)
is conjugate to w
2
by Lemma 11.1. 2
Let W be a nite Coxeter group, w = a
1
a
n
a Coxeter element, a
1
, . . . , a
n
the vertices of
the Coxeter graph. Then h = [w[ is called the Coxeter number of W.
Let
1
, . . . ,
n
be the eigenvalues of w[
V (W)
. Since w
h
= 1,
h
k
= 1 =
k
= e
2im
k
h
for a
unique 0 m
k
< h.
We call m
1
, . . . , m
n
the exponents of W.
Finite Coxeter graphs are bipartite (X = X
1
X
2
) Example:
D
6
:

d
d
d
d
d
d
d

~
~
~
~
~
~
~

d
d
d
d
d
d
d

58
E
8
:

~
~
~
~
~
~
~
d
d
d
d
d
d
d

d
d
d
d
d
d
d


For a, b X
i
= ab = ba. Hence
j
=

aX
j
a does not depend on the order of X
j
. Moreover,

2
j
=

a
2
= 1.
=
1

2
is called canonical Coxeter element.
G =
1
,
2
is a dihedral group of order 2h.
Proposition 11.3. Let us order the exponents 0 m
1
. . . m
n
< h. Then
1. m
1
1
2. j : h m
j
is an exponent
3.
n

j=1
m
j
= h
n
2
Proof.
1. if m
1
= 0 then e
0
= 1 is a eigenvalue of . Hence x V (W) s.t. x ,= 0, x = x.

2
x =
1

2
x =
1
x =
1
x
Let a X
j
: S
a
(c) = a x = x 2 e
a
, x e
a
. Hence

1
x = x +

aX
1

a
e
q
=
2
x = x +

bX
2

b
e
b
=

aX
1

a
e
a
=

bX
2

b
e
b
and therefore all
a
,
b
= 0. So
1
x =
2
x = 0. Consider

1
[
V (W)
a
j
X
1
= S
a
1
S
a
k
:
_
V

1
y y
V
1
y y
where V
i
=

aX
j
Re
a
.
Hence x V

1
V

2
and V

1
V

2
= 0 because V = V
1
V
2
= V

1
V

2
since the form
is non-degenerate. But this is a contradiction (x = 0) proving that m
1
> 0.
59
2. G acts on V (W). Let f(z) be the characteristic polynomial of [
V (W)
. Then
f(z) = f
1
(z) f
s
(z)
where f
j
(z) R[z] and monic irreducible in R[z].
Case 1 f
j
(z) = z , R.
Since
h
= 1 =
h
= 1 = = 1. Note that 1 is not an eigenvalue of by
part (1), so = 1, h is even and = e
2i
h
h
2
,
h
2
is the exponent.
Case 2 f
j
(z) = (z )(z

), / R.
Then if = e
2i
h
m
k
then

= e

2i
h
m
k
= e
2i
h
(hm
k
)
.
3. This also gives part (3) because the average of exponents in each f
j
(z) is
h
2
.
2
Lemma 11.4. Let W be nite and connected. Then there are z
1
, z
2
V s.t.
1. z
1
, z
2
are linearly independent
2. P = Rz
1
Rz
2
is G = D
2h
-linear
3.
i
[
P
= S
z

i
4. z
i
C, the closure of the fundamental chamber
5. P C = R
>0
z
1
+R
>0
z
2
Proof.
a
i
is the dual basis to e
a
i
. That means

a
i
, e
a
j
_
=
ij
.
Easy claims:
C = x : x, e
a
0 =

a
R
0

a
V

j
=

a/ X
j
R
a
angle between
a
and
b
= angle between e
a
and e
b
The angles between
a
and
b
are accute and
a
,
b
0 and
a
,
b
= 0 e
a
, e
b
=
0 a and b commute.
Q = (e
a
, e
b
)
a,b
is the matrix of
symmetric positiv denite bilinear form , in the basis e
a
i
linear map q : V V, q(
a
) = e
a
in the basis
a
i
.
In particular, its eigenvalues are positive and real. Let be the largest eigenvalue of Q. Consider
the new form [ , ] given by QI in the basis e
a
i
.
Note: [x, x] 0 and [x, x] = 0 x ker[ , ] = K. K ,= 0, K = eigenspace of q.
Pick z ,= 0, z =

a
K and let z =

[
a
[
a
.
Now, the idea is to use that z
j
=

aX
j
[
a
[
a
, but we have no time to nish the proof. 2
60
Theorem 11.5. If W is nite connected of rank n (i.e. the number of vertices in the Coxeter
graph is n), then
1. m
1
= 1, m
n
= h 1
2. [[ = hn
Proof. (Idea)
H

P is G-conjugate to Rz
1
or Rz
2
.
if h is odd:
Rz
1
Rz
2
There are
n
2
hyperplanes H

s.t. H

P is one of these h lines

if h is even:
Rz
1
Rz
2
There are
h
2
lines with one intersection pattern and
h
2
lines with another intersection
pattern
2
Theorem 11.6. Let W be nite and connected of rank n, 1 m
1
. . . m
n
= h 1 the
fundamental exponents, 2 d
1
. . . d
n
the fundamental degrees. Then:
j : d
j
= m
j
+ 1
Proof. (idea)
Compute Jac = det
_
I
j
z
k
_
in two dierent ways. 2
Corollary 11.7. [W[ =

j
(m
j
+ 1), [[ = 2

j
m
j
.
61
Corollary 11.8. Let W be connected.
1 =
_
_
_
1
.
.
.
1
_
_
_
(W) all m
i
are odd
Proof.
"=" (1) I
j
= I
j
S
W
. Therefore (1) x
k
= x
k
= (1)I
j
= (1)
d
j
I
j
= (1)
d
j
= 1.
So all d
j
are even and therefore all m
j
are odd.
"=" m
1
= 1, m
n
= h 1 are both odd = h is even
([
V (W)
)
h
2

_
_
_
e
2i
h
m
k
.
.
.
e
2i
h
m
k
_
_
_
h
2
=
_
_
_
e
im
k
.
.
.
e
im
k
_
_
_
=
_
_
_
1
.
.
.
1
_
_
_
2
Proposition 11.9. Let W be nite, connected, crystallographic. Then m s.t. 1 m
h 1, gcd(m, h) = 1 then m is an exponent.
Proof. Cyclotomic polynomial

h
(z) =

1mh1
gcd m,h=1
(z e
2i
h
m
)
is irreducible in Z[z].
The characteristic polynomial f

(z) of [
V (W)
is in Z[z].
For m = 1: (z e
2i
h
)[f

(z) in C[z]. Therefore

h
(z)[f

(z) in Z[z]
and therefore all e
2i
h
m
are eigenvalues. 2
Application: E
8
has rank 8 and 240 roots. Hence h =
[[
n
=
240
8
= 30. The number of coprimes
of 30 is
(30) = (2)(5)(3) = 1 4 2 = 8
Hence the exponents are 1, 7, 11, 13, 17, 19, 23, 29 and so
[W[ = 2 8 12 14 18 20 24 30
[[ h =
[[
n
easy exponents missing exponents [W[
E
6
72 12 1, 5, 7, 11 4, 8
E
7
126 18 1, 5, 7, 11, 13, 17 9
E
8
240 30 1, 7, 11, 13, 17, 19, 23, 29 none
F
4
48 12 1, 5, 7, 11 none 2 6 8 12 = 2
7
3
2
H
3
30 10 1, 9 5 2 6 10 = 120
H
4
120 30 1, 29 11, 19 2 12 20 30 = (120)
2
62
Calculations:
Since exponents come in pairs (m, hm) the only stand alone exponent is
h
2
. This gives missing
E
7
and H
3
components.
E
6
: two exponents are missing. We know that z
12
1 is satised by Coxeter transformations.
z
12
1 =
12

1
if = e
2im
12
where m is a missing component, there are 4 possible cases:

2
() = 0 6, 6

4
() = 0 3, 9

3
() = 0 4, 8

6
() = 0 2, 10
One needs to write the Coxeter transformations itself to see that
3
() = 0.
H
4
: can be realised by matrices with coecients in Q(

5), z
30
1 =
30
. Over Q(

5),
30
=

(1)
30

(0)
30
, the product of 2 irreducibles of degree 4. Exponents correspond to the factor

(0)
30
and we know that
(0)
30
(e
2i
30
) = 0. The roots of
(0)
30
are:
e
2i
30
, e
2i
30
29
, e
2i
30
11
, e
2i
30
19
H
3
: H
3
= Sym(Icosahedron) = Sym(Dodecahedron).
Every reection of the Dodecahedron xes two opposite edges. Hence
[[ = 2[reections[ = 2
[edges[
2
= 30
and
[W[ = 20
..
[vertices[
6 = 120
63
F
4
, H
4
: W(F
4
) = Sym(24 cells), W(H
4
) = Sym(120 cells) = Sym(600 cells). We want
to use a trick here:
Let x, y H, then S
x
(y) = xyx. Hence if G H

, [G[ < , [G[ even = G is a root

system. Let q : |q| = 1 = U H

and consider
: U
SU
2
(C)
SO(3)(R) = SO(H
im
) : q (x qxq)
which is 2:1. Then
H
4
=
1
(Rot.Symm(Dodec)) = [
H
4
[ = 2 60 = 120.

F
4
=
1
(Rot.Symm.(Cube)) = [
F
4
[ = 2 24 = 48.
E
8
, E
7
: There exists a 8-dimensional algebra (octonions). If G

, [G[ < , [G[ is even

= G is a root system.

E
8
is the "group" of invertible octavian integers

E
7
are the elements of order 4 in it.
64