Scribd
Upload
160 views

Root Data For Reductive Groups

This document provides an explanation of how reductive Lie groups are classified by root data. It begins with defining a root datum and outlining its properties. It then defines reductive groups in terms of a structure theorem. Finally, it describes the construction of a complex reductive Lie group from a given root datum in four steps: defining the associated Lie algebra, exponentiating to a 1-connected group, realizing this as a quotient, and verifying the correct root datum.

Uploaded by

bhaumiksaurav
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
160 views

Root Data For Reductive Groups

This document provides an explanation of how reductive Lie groups are classified by root data. It begins with defining a root datum and outlining its properties. It then defines reductive groups in terms of a structure theorem. Finally, it describes the construction of a complex reductive Lie group from a given root datum in four steps: defining the associated Lie algebra, exponentiating to a 1-connected group, realizing this as a quotient, and verifying the correct root datum.

Uploaded by

bhaumiksaurav
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

ROOT DATA AND REDUCTIVE GROUPS

CALDER DAENZER
This is an explanation of how reductive Lie groups are classied by root data. The aim is to
make the explanation concise, but to at least outline all of the steps that are involved in the
classication. For further details and proofs of the theorems I recommend [Ser], [Kna], and
[Mil].
Let us start right o with the denition of a root datum:
Denition 0.1. A (reduced) root datum is a triple:
C , R

, k
where is a lattice (that is, a nitely generated free Z-module),

= Hom(, Z) is the dual


lattice, C and R are nite sets (called the coroots and roots of the root datum), k is a
bijection
C
k
R, h

,
and the following conditions are satised:
(1) (Reduced) For every R, is the only multiple of which also lies in R.
(2) For each R, the map


s
(h

) preserves R and sends to .


Furthermore, the collection s

generates a nite group W (called the Weyl group).


The bijection k will often be implicit and we will assume without mention that root data
are reduced. A root datum is called semi-simple if the lattice
C
generated by C has nite
index in .
Now on to reductive groups. Rather than give the original denition, I will dene reductive
groups in terms of a structure theorem:
Structure theorem for reductive Lie groups. The (split) reductive algebraic groups are precisely
the nite quotients of groups of the form (G
m
)
n
G
1
G
k
, where G
m
= GL
1
is the
multiplicative group and G
i
are simple algebraic groups. Note that simple algebraic groups
are (by denition) connected, so all reductive groups are connected.
Remark 0.2. By taking the C-points of a reductive algebraic group, one recovers the complex
structure theorem: the complex reductive Lie groups are those isomorphic to nite quotients
of (C

)
n
G
1
G
k
, with G
i
being simple complex Lie groups. It can be proved that every
complex reductive group is in fact algebraic, so algebraic reductive groups may be dened via
complex Lie groups.
Complex reductive groups are also in correspondence with compact connected Lie groups.
Indeed, the maximal compact subgroups of a complex reductive group (all of which are conju-
gate to one another) are compact connected Lie group, and the complexication of any compact
connected Lie group is reductive. These correspondences can be codied as equivalences of
1
2 CALDER DAENZER
categories (see [Ser] Chap VIII for the semisimple case) :
compact connected Lie groups reductive complex Lie groups
split reductive algebraic groups .
Root data of a complex reductive Lie group. Let G be a complex reductive Lie group with
chosen maximal torus T. The root datum of (G,T) is the triple
!(G, T) := C , R

, k
dened as follows. First, set
= Hom(C

, T) and

=

T = Hom(T, C

).
Next, the roots R

of (G, T) are dened to be the nonzero characters of the adjoint


representation G
Ad
GLg. More precisely,
Ad[
T
: T GL(g)
is diagonalizable, which gives rise to a decomposition g = g
0
(

T\0
g

) in which Ad [
T
acts on g

by the character . The roots are those



T for which g

,= (0).
The coroots of (G, T) are dened as
C := h Hom(C

, T) [ h = [
C
for some Hom(SL
2
, G) .
To describe the bijection between roots and coroots one can use the Killing form,
K(X, Y ) := trace(ad(X) ad(Y )) X, Y g,
in the following way: The Lie algebra g decomposes as
g = (a h) (

T\0
g

)
where a is the center of g, and the Killing form restricts to a nondegenerate form on h.
View characters and cocharacters as elements of the Lie algebra and dual Lie algebra of T by
dierentiating at the identity:
Hom(C

, T) Hom(C, t) = t, Hom(T, C

) t

.
Then in fact C h t and R t

, and it is a theorem that the coroots C h satisfy


h

C
k
:= 2
K(, h

)
K(h

, h

)
R.
It is not trivial to show, but !(G, T) does satisfy the axioms of Denition (0.1) (see [Ser] Chap.
VI.1-2).
So we have described the root datum of a group. Next up is to show that this construction
induces a 1-to-1 correspondence (up to isomorphism) between reduced root data and reductive
Lie groups. To that end, we will outline the construction of a complex reductive Lie group G
which has a prescribed root datum.
The reductive Lie group of a root datum. This will proceed in 4 steps: rst dene a Lie al-
gebra g associated to a root datum, second, exponentiate to a 1-connected Lie group

G, third,
realize G as a quotient of

G by a central subgroup, and fourth, verify that G has the desired
root datum.
ROOT DATA AND REDUCTIVE GROUPS 3
Step 1: The Lie algebra. Choose a subset S =
1
, . . . ,
n
R (called a base of simple roots)
which is a basis of Span
C
(R) and such that for any R expressed in the basis: =

m
i

i
,
the coecients m
i
are either all non-negative integers or non-positive integers. (It is a theorem
that this can be done, and also that the Weyl group acts simply transitively on the collection
of such bases.)
For i, j = 1, . . . , n dene
ij
:=
j
(h

i
) Z. It is shown in [Ser] Chap V.7 that
ii
= 2 and

ij
0, 1, 2, 3 when i ,= j (this fact is important for the Lie algebra relations). Now we
form a complex Lie algebra g() by generators and relations:
Generators: h
i
, X
i
, Y
i

i=1,...,n
.
Relations: For all i, j:
[h
i
, h
j
] = 0, [h
i
, X
j
] =
ij
X
j
, [h
i
, Y
j
] =
ij
Y
j
, [X
i
, Y
j
] =
ij
H
j
and for i ,= j:
(ad X
i
)

ij
+1
X
j
= 0, (ad Y
i
)

ij
+1
Y
j
= 0.
It can be shown that g() is a nite dimensional semi-simple Lie algebra, that h := Span
C
h
i

is a maximal abelian subalgebra of g(), and that b := Span


C
h
i
, X
i
is a maximal solvable
Lie subalgebra of g(). When the root data is semi-simple, this is the desired Lie algebra. In
general, we have
h
Z
C =: g
0
.
It is not hard to check that h is complemented in g
0
by
a := Span
C
[ () = 0 for all R.
Then the desired Lie algebra is g = a g(), with bracket extended to A a by [A, X] = 0
for all X g (thus a is the center of g).
Step 2: The 1-connected group. Lies third theorem provides a unique up to isomorphism 1-
connected Lie group

G corresponding to g. It should be noted that this group, while complex,
need not be reductive. For instance it equals C
n
when the Lie algebra is abelian. Furthermore,
we will next make use of the exponential map which is not available in the category of ane
algebraic groups. Thus while this construction is quick and easy, it is not the one to use for
algebraic groups in general.
Step 3: Realize the group G as a quotient.
Denote by
C
, and
R
the lattices generated by C and R. Dene a set

R
by:

R
:= h g
0
[ (h) Z for all R,
and note that

R
contains the vector space a (for which (h) = 0 Z), and also contains the
lattice .
We claim that Z(

G) = exp(2i

R
). Indeed, the exponential map is surjective since

G is
connected, and for h g
0
, we have
exp(2ih) Z(

G) Ad(exp(2ih)) exp(2i ad(h)) = 1 (h) Z R.


Now we know that the set L := exp(2i) is a subgroup of Z(

G) (and is in particular normal


in

G). Dene
G =

G/L,
4 CALDER DAENZER
which is the desired reductive Lie group associated to the given root datum.
The maximal torus in G is
T := exp(g
0
)/L,
and one gets for free that Z(G)

R
/, which is nite if and only if the center a = (0).
Also,
1
(G) /
C
is not hard to verify: indeed,
1
(T) maps onto
1
(G), and an
element Hom(S
1
, T) goes to 0
1
(G) if and only if it lifts to an element of
1
(

G), if
and only if it factors through a map S
1


T g
0
/
C
, if and only if
C
.
Step 4: Verify that !(G, T) = C , R

:
First note that G has the correct roots and coroots, because the quotient

G G induces
an isomorphism of Lie algebras, and the roots and coroots only depend on the Lie algebra,
and g = Lie(

G) has (by construction) the correct roots and coroots. The only remaining
thing to check is that Hom(C

, T) for T = exp(g
0
/2i). This is clearly true as long as
g
0
/
exp(2i)
T is injective.
Remark 0.3. If !(G, T) = C , R

is a reduced root datum, then interchanging the


roles of roots and coroots produces a new root datum
!(G, T)

:= R

, C .
This leads to the denition of a Langlands dual group to G, which is by denition any
group G

satisfying
!(G

, T

) !(G, T)

.
Looking at how the center and fundamental group may be read o from root data, one sees
that for G semisimple,
1
(G) = Z(G

) and Z(G) =
1
(G

).
References
[Mil] Milne, J.S; Reductive Groups. Unnished course notes at http://www.jmilne.org/math/CourseNotes/RG.pdf.
[Kna] Knapp, Anthony W.; Lie groups beyond an introduction. Progress in Mathematics, 140. Birkhuser Boston,
Inc., Boston, MA, 1996.
[Ser] Serre, Jean-Pierre; Complex semisimple Lie algebras. Translated from the French by G. A. Jones. Springer-
Verlag, New York, 1987.
E-mail address: [email protected]

You might also like