Topics in Representation Theory: Roots and

Weights

1 The Representation Ring

Last time we defined the maximal torus T and Weyl group W (G, T ) for a
compact, connected Lie group G and explained that our goal is to relate the
representation theory of T to that of G. One aspect of the representation theory
of T and of G for which there is a simple relation is that of their representation
rings.
Consider representations π of a compact Lie group G on complex vector
spaces and recall that a representation is characterized up to isomorphism by
its character
χπ : g ∈ G → T r(π(g)) ∈ C
and that character functions are conjugation invariant

χ(hgh−1 ) = χ(g)

Since
χπ1 ⊕π2 = χπ1 + χπ2
and
χπ1 ⊗π2 = χπ1 χπ2
the characters generate an interesting ring of functions on G. One can define
Definition 1 (Character Ring). Let R(G) be the free abelian group generated
by characters of complex representations of G. This is called the character ring
of G.
The ring structure comes from the sum and product of representations. Very
explicitly one can think of this ring as what one gets by adding, subtracting and
multiplying character functions of irreducible representations. Elements of R(G)
are sometimes called virtual characters since, given any two representations
π1 and π2 , one has not only corresponding elements [π1 ] and [π2 ] in R(G)
corresponding to their characters, but also differences.

[π1 ] − [π2 ] ∈ R(G)

This kind of construction which starts with a semigroup (the irreducible repre-
sentations with direct sum) and produces a group (R(G)) which also happens
to be a ring appears in other contexts in mathematics. The simplest example is
the construction of the ring of integers from the semigroup of natural numbers.
A more sophisticated version of this is topological K-theory, which is the ring
you get starting from the semigroup of isomorphism classes of vector bundles
on a given topological space.

1
 The Peter-Weyl theorem implies that for compact G any conjugation invari-
ant function can by uniformly approximated by linear combinations of character
functions.
Since we have seen that any element g ∈ G can be conjugated into an element
of T , a conjugation invariant function such as χπ is determined by its values on
T . Associated to the inclusion map

i:T →G

there is a map on character rings induced by restriction of characters to T

i∗ : R(G) → R(T )

This map is injective and since the Weyl group W (G, T ) acts on T by conjuga-
tion, the image is in
R(T )W (G,T ) ⊂ R(T )
the W (G, T ) invariant character functions on T .

2 Weights
All the irreducible complex representations of T are one-dimensional. When we
restrict an n-dimensional representation π of G to T , it will break up into n
one-dimensional irreducible representations of T .
Definition 2 (Weights). A weight is an irreducible representation of T . For
any representation π of G, the weight space corresponding to a given weight is
the subspace of the representation space of π that transforms under T according
to the given weight.
There are various ways of explicitly labelling weights, and one can do so in
terms of either the group T or the Lie algebra t. This can be rather confusing,
so we’ll now belabor the rather trivial subject of the representation theory of
T . As we’ll see, there are many possible choices of convention. I’ll be following
to some extent [1].
A complex irreducible representation of S 1 can be chosen to be unitary and
is given by a homomorphism

θ : S 1 → U (1)

taking the differential of this map gives a map

θ∗ : R → R

on the Lie algebras. Such irreducible representations are labelled by an integer

n and the θ corresponding to n is

θn : [x] ∈ R/Z = S 1 → e2πinx

2
The map on Lie algebras is
(θn )∗ : x ∈ R → nx ∈ R
Note that (θn )∗ is an element of the dual space to R.
Now for T = (S 1 )k the irreducible representations are homomorphisms
θ : (S 1 )k → U (1)
These correspond to a choice of k integers n = (n1 , · · · , nk ) and
θn : ([x1 ], · · · , [xk ]) ∈ (R/Z)k → e2πi(n1 x1 +···+nk xk )
The corresponding Lie algebra maps are
(θn )∗ : (x1 , · · · , xk ) → n1 x1 + · · · + nk xk
and can be thought of as elements of the dual space to the Lie algebra of T , t∗ .
When we use the term weight, we will often mean an element of t∗ corresponding
to an irreducible T representation. These elements of t∗ are those that take
integer values on the integer lattice in Rk . More abstractly, the integer lattice
is the set of points
exp−1 (e) ⊂ t
So far we have been considering unitary representations of T on complex
vector spaces. We also need to consider representations of T on real vector
spaces. In this case irreducible representations are homomorphisms
θ : T → SO(2)
and there are two different sorts of irreducible representations:
1. The one-dimensional trivial representation on R.
2. Non-trivial representations on R2 labelled by k integers, not all zero.
As in the complex case a weight is one of these irreducible representations and
is given by an element
(θn )∗ : (x1 , · · · , xk ) → n1 x1 + · · · + nk xk
of t∗ taking integer values on the integer lattice in Rk . The corresponding T
representation is explicitly
 
cos(2π(θn )∗ (x1 , · · · , xk )) − sin(2π(θn )∗ (x1 , · · · , xk ))
θn : ([x1 ], · · · , [xk ]) ∈ (R/Z)k →
sin(2π(θn )∗ (x1 , · · · , xk )) cos(2π(θn )∗ (x1 , · · · , xk ))
except for the special case of (θn )∗ = 0 where the representation is the trivial
one and not on R2 , but R.
Note that θn )∗ and −θn )∗ are isomorphic representations since you can go
from the matrix for one to the matrix for the other by conjugation by
 
0 1
1 0

3
3 Roots
We’ll be using some of the following terminology:
Definition 3. A subalgebra h of a Lie algebra g is an ideal if it satisfies [g, h] ⊂
h.
If the Lie algebra h of a subgroup H ⊂ G is an ideal in g, then H will be a
normal subgroup of G.
Definition 4. A Lie algebra g is called simple if it has no non-trivial proper
ideals and if it is not one-dimensional. A Lie group is called simple if its Lie
algebra is simple.
Definition 5. A Lie algebra g is called semi-simple if has no non-trivial abelian
ideals. A Lie group is called semi-simple if its Lie algebra is semi-simple.
Equivalently, a semi-simple Lie group is one with no non-trivial abelian
connected normal subgroups. Also equivalently, a semi-simple Lie algebra is
isomorphic with a product of simple Lie algebras. Examples of non-semi-simple
Lie groups and Lie algebras are G = U (n) and g = u(n). U (1) is not a simple Lie
group. Simple Lie groups and Lie algebras include G = SU (n) and g = su(n).
Products such as SU (n) × SU (m) are semi-simple Lie groups, direct sums such
as su(n) ⊕ su(n) are semi-simple Lie algebras.
The Lie algebras we will be concerned with are the Lie algebras of compact
Lie groups. These are real vector spaces, but for some purposes it is convenient
to complexify and consider complex Lie algebras. It turns out that the classi-
fication of the Lie algebras of semi-simple compact Lie groups is equivalent to
the classification of complex semi-simple Lie algebras since each complex semi-
simple Lie algebra has a unique associated real Lie algebra of a compact group.
The general theory of complex semi-simple Lie algebras will not be considered
in this course although we will be going a little more into the subject later on.
One important property of the real Lie algebra g is that the adjoint repre-
sentation is an orthogonal one

Ad : G → SO(dim g)

One can see this by observing that the adjoint representation has an invariant
positive definite inner product.
We won’t prove this, but it is true that for any semi-simple Lie algebra the
Killing form is non-degenerate. For those corresponding to compact groups it is
negative definite and its negative gives a positive definite inner product. Next
week we will examine in more detail the properties of the Killing form.
Definition 6 (Cartan Subalgebra). For a compact Lie group G, a Cartan
sub-algebra is a Lie subalgebra whose Lie group is a maximal torus T of G.
Note: One can equivalently define a Cartan sub-algebra as a maximal abelian
sub-algebra. There is a more general notion of Cartan subalgebra for general
complex semi-simple Lie algebras.

4
 Given the adjoint representation Ad of G on the real vector space g, one
can ask the question of what the weights of the T action are in this case. T
acts trivially on the Cartan subalgebra, so the trivial weight will appear with
multiplicity rank(G). The part of g orthogonal to the Cartan subalgebra will
break up into non-trivial two-dimensional irreducible orthogonal representations
of T , these are the roots:
Definition 7 (Roots). The roots of G are the non-trivial weights of the adjoint
representation on the real vector space g. More explicitly, the roots are non-zero
elements of α ∈ t∗ , taking integer values on the integer lattice.

Next week we will be studying further properties of roots, and classifying all
possible systems of roots, giving a classification of all compact, connected Lie
groups.

References
[1] Adams, J. Frank, Lectures on Lie Groups, University of Chicago Press,
1969.