Lie Groups and Lie Algebras

version 1.6d — 24.11.
2008
Lie Groups and Lie Algebras

Lecture notes for 7CCMMS01/CMMS01/CM424Z
Ingo Runkel
j j
E
Contents
1 Preamble 3
2 Symmetry in Physics 4
2.1 Definition of a group . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Rotations and the Euclidean group . . . . . . . . . . . . . . . . . 6
2.3 Lorentz and Poincaré transformations . . . . . . . . . . . . . . . 8
2.4 (*) Symmetries in quantum mechanics . . . . . . . . . . . . . . . 9
2.5 (*) Angular momentum in quantum mechanics . . . . . . . . . . 11
3 Matrix Lie Groups and their Lie Algebras 12

3.1 Matrix Lie groups . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 The exponential map . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 The Lie algebra of a matrix Lie group . . . . . . . . . . . . . . . 15
3.4 A little zoo of matrix Lie groups and their Lie algebras . . . . . . 17
3.5 Examples: SO(3) and SU (2) . . . . . . . . . . . . . . . . . . . . 19
3.6 Example: Lorentz group and Poincaré group . . . . . . . . . . . 21
3.7 Final comments: Baker-Campbell-Hausdorff formula . . . . . . . 23
4 Lie algebras 24
4.1 Representations of Lie algebras . . . . . . . . . . . . . . . . . . . 24
4.2 Irreducible representations of sl(2, C) . . . . . . . . . . . . . . . . 27
4.3 Direct sums and tensor products . . . . . . . . . . . . . . . . . . 29
4.4 Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.5 The Killing form . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5 Classification of finite-dimensional, semi-simple, complex Lie

algebras 37
5.1 Working in a basis . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Cartan subalgebras . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3 Cartan-Weyl basis . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.4 Examples: sl(2, C) and sl(n, C) . . . . . . . . . . . . . . . . . . . 47
5.5 The Weyl group . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.6 Simple Lie algebras of rank 1 and 2 . . . . . . . . . . . . . . . . . 50
5.7 Dynkin diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.8 Complexification of real Lie algebras . . . . . . . . . . . . . . . . 56
6 Epilogue 62
A Appendix: Collected exercises 63

A.1 Hints and solutions for exercises . . . . . . . . . . . . . . . . . . 77
2
Exercise 0.1 :
I certainly did not manage to remove all errors from this script. So the first
exercise is to find all errors and tell them to me.
1 Preamble
The topic of this course is Lie groups and Lie algebras, and their representations.
As a preamble, let us have a quick look at the definitions. These can then again
be forgotten, for they will be restated further on in the course.
Definition 1.1 :
A Lie group is a set G endowed with the structure of a smooth manifold and of a
group, such that the multiplication · : G×G → G and the inverse ( )−1 : G → G
are smooth maps.
This definition is more general than what we will use in the course, where
we will restrict ourselves to so-called matrix Lie groups. The manifold will then
always be realised as a subset of some Rd . For example the manifold S 3 , the
three-dimensional sphere, can be realised as a subset of R4 by taking all points of
R4 that obey x12 +x22 +x32 +x42 = 1. [You can look up ‘Lie group’ and ‘manifold’
on eom.springer.de, wikipedia.org, mathworld.wolfram.org, or planetmath.org.]
In fact, later in this course Lie algebras will be more central than Lie groups.
Definition 1.2 :
A Lie algebra is a vector space V together with a bilinear map [ , ] : V ×V → V ,
called Lie bracket, satisfying
(i) [X, X] = 0 for all X ∈ V (skew-symmetry),
(ii) [X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y ]] = 0 for all X, Y, Z ∈ V (Jacobi identity).
We will take the vector space V to be over the real numbers R or the complex
numbers C, but it could be over any field.
Exercise 1.1 :
Show that for a real or complex vector space V , a bilinear map b(·, ·) : V ×V → V
obeys b(u, v) = −b(v, u) (for all u, v) if and only if b(u, u) = 0 (for all u). [If you
want to know, the formulation [X, X] = 0 in the definition of a Lie algebra is
preferable because it also works for the field F2 . There, the above equivalence
is not true because in F2 we have 1 + 1 = 0.]
Notation 1.3 :
(i) “iff” is an abbreviation for “if and only if”.
(ii) If an exercise features a “*” it is optional, but the result may be used in the
3
following. In particular, it will not be assumed in the exam that these exercises
have been done. (This does not mean that material of these exercises cannot
appear in the exam.)
(iii) If a whole section is marked by a “*”, its material was not covered in the
course, and it will not be assumed in the exam that you have seen it before. It
will also not be assumed that you have done the exercises in a section marked
by (*).
(iv) If a paragraph is marked as “Information”, then as for sections marked by
(*) it will not be assumed in the exam that you have seen it before.
2 Symmetry in Physics
The state of a physical system is given by a collection of particle positions and
momenta in classical mechanics or by a wave function in quantum mechanics.
A symmetry is then an invertible map f on the space of states which commutes
with the time evolution (as given Newton’s equation mẍ = −∇V (x), or the
∂
Schrödinger equation i~ ∂t ψ = Hψ)
state at time 0
evolve / state at time t (2.1)
f = f

state0 at time 0
evolve / state0 at time t
Symmetries are an important concept in physics. Recent theories are almost

entirely constructed from symmetry considerations (e.g. gauge theories, super-
gravity theories, two-dimensional conformal field theories). In this approach one
demands the existence of a certain symmetry and wonders what theories with
this property one can construct. But let us not go into this any further.
2.1 Definition of a group

Symmetry transformations like translations and rotations can be composed and
undone. Also ‘doing nothing’ is a symmmetry in the above sense. An appropri-
ate mathematical notion with these properties is that of a group.
Definition 2.1 :
A group is a set G together with a map · : G × G → G (multiplication) such
that
(i) (x · y) · z = x · (y · z) (associativity)
(ii) there exists e ∈ G s.t. e · x = x = x · e for all x ∈ G (unit law)
(iii) for each x ∈ G there exists an x−1 ∈ G such that x · x−1 = e = x−1 · x
(inverse)
4
Exercise 2.1 :
Prove the following consequences of the group axioms: The unit is unique. The
inverse is unique. The map x 7→ x−1 is invertible as a map from G to G.
e−1 = e. If gg = g for some g ∈ G, then g = e. The set of integers together with
addition (Z, +) forms a group. The set of integers together with multiplication
(Z, ·) does not form a group.
Of particular relevance for us will be groups constructed from matrices. De-
note by Mat(n, R) (resp. Mat(n, C)) the n×n-matrices with real (resp. complex)
entries. Let
GL(n, R) = {M ∈ Mat(n, R)| det(M ) 6= 0} . (2.2)
Together with matrix multiplication (and matrix inverses, and the identity ma-
trix as unit) this forms a group, called general linear group of degree n over R.
This is the basic example of a Lie group.
Exercise 2.2 :
Verify the group axioms for GL(n, R). Show that Mat(n, R) (with matrix mul-
tiplication) is not a group.
Definition 2.2 :
Given two groups G and H, a group homomorphism is a map ϕ : G → H such
that ϕ(g · h) = ϕ(g) · ϕ(h) for all g, h ∈ G.
Exercise 2.3 :
Let ϕ : G → H be a group homomorphism. Show that ϕ(e) = e (the units in G
and H, respectively), and that ϕ(g −1 ) = ϕ(g)−1 .
Here is some more vocabulary.
Definition 2.3 :
A map f : X → Y between two sets X and Y is called injective iff f (x) =
f (x0 ) ⇒ x = x0 for all x, x0 ∈ X, it is surjective iff for all y ∈ Y there is a x ∈ X
such that f (x) = y. The map f is bijective iff it is surjective and injective.
Definition 2.4 :
An automorphism of a group G is a bijective group homomorphism from G to
G. The set of all automorphisms of G is denoted by Aut(G).
Exercise 2.4 :
Let G be a group. Show that Aut(G) is a group. Show that the map ϕg : G → G,
ϕg (h) = ghg −1 is in Aut(G) for any choice of g ∈ G.
Definition 2.5 :
Two groups G and H are isomorphic iff there exists a bijective group homomor-
phism from G to H. In this case we write G ∼
= H.
5
2.2 Rotations and the Euclidean group
A typical symmetry is invariance under rotations and translations, as e.g. in the
Newtonian description of gravity. Let us start with rotations.
Take Rn (for physics probably n = 3) with the standard inner product
n
X
g(u, v) = ui vi for u, v ∈ Rn . (2.3)
i=1
A linear map T : Rn → Rn is an orthogonal transformation iff
g(T u, T v) = g(u, v) for all u, v ∈ Rn . (2.4)
Denote by ei = (0, 0, . . . , 1, . . . , 0) the i’th basis vector of Rn , so that in com-

ponent notation (T ei )k = Tki . Evaluating the above condition in this basis
gives
X
lhs = g(T ei , T ej ) = Tki Tkj = (T t T )ij , rhs = g(ei , ej ) = δij . (2.5)
k
Hence T is an orthogonal transformation iff T t T = 1 (here 1 denotes the unit

matrix (1)ij = δij ).
What is the geometric meaning of an orthogonal transformation? p The con-
dition g(T u, T v) = g(u, v) shows that it preserves the length |u| = g(u, u) of
a vector as well as the angle cos(θ) = g(u, v)/(|u| |v|) between two vectors.
However, T does not need to preserve the orientation. Note that
T t T = 1 ⇒ det(T t T ) = det(1) ⇒ det(T )2 = 1 ⇒ det(T ) ∈ {±1} (2.6)
The orthogonal transformations T with det(T ) = 1 preserve orientation. These

are rotations.
Definition 2.6 :
(i) The orthogonal group O(n) is the set
O(n) = {M ∈ Mat(n, R)|M t M = 1} (2.7)
with group multiplication given by matrix multiplication.

(ii) The special orthogonal group SO(n) is given by those elements M of O(n)
with det(M ) = 1.
For example, SO(3) is the group of rotations of R3 .
Let us check that O(n) is indeed a group.
(a) Is the multiplication well-defined?
Given T, U ∈ O(n) we have to check that also T U ∈ O(n). This follows from
(T U )t T U = U t T t T U = U t 1U = 1.
(b) Is the multiplication associative?
6
The multipication is that of Mat(n, R), which is associative.
(c) Is there a unit element?
The obvious candidate is 1, all we have to check is if it is an element of O(n).
But this is clear since 1t 1 = 1.
(d) Is there an inverse for every element?
For an element T ∈ O(n), the inverse should be the inverse matrix T −1 . It
exists because det(T ) 6= 0. It remains to check that it is also in O(n). To this
end note that T t T = 1 implies T t = T −1 and hence (T −1 )t T −1 = T T −1 = 1.
Definition 2.7 :
A subgroup of a group G is a non-empty subset H of G, s.t. g, h ∈ H ⇒ g ·h ∈ H
and g ∈ H ⇒ g −1 ∈ H. We write H ≤ G for a subgroup H of G.
From the above we see that O(n) is a subgroup of GL(n, R).
Exercise 2.5 :
(i) Show that a subgroup H ≤ G is in particular a group, and show that it has
the same unit element as G.
(ii) Show that SO(n) is a subgroup of GL(n, R).
The transformations in O(n) all leave the point zero fixed. If we are to de-
scribe the symmetries of euclidean space, there should be no such distinguished
point, i.e. we should include translations. It is more natural to consider the
euclidean group.
Definition 2.8 :
The euclidean group E(n) consists of all maps Rn → Rn that leave distances
fixed, i.e. for all f ∈ E(n) and x, y ∈ Rn we have |x − y| = |f (x) − f (y)|.
The euclidean group is in fact nothing but orthogonal transformations com-
plemented by translations.
Exercise 2.6 :
Prove that
(i*) for every f ∈ E(n) there is a unique T ∈ O(n) and u ∈ Rn , s.t. f (v) = T v+u
for all v ∈ Rn .
(ii) for T ∈ O(n) and u ∈ Rn the map v 7→ T v + u is in E(n).
The exercise shows that there is a bijection between E(n) and O(n) × Rn
as sets. However, the multiplication is not (T, x) · (R, y) = (T R, y + x). Instead
one finds the following. Writing fT,u (v) = T v + u, we have
fT,x (fR,y (v)) = fT,x (Rv + y) = T Rv + T y + x = fT R,T y+x (v) (2.8)
so that the group multiplication is
(T, x) · (R, y) = (T R, T y + x) . (2.9)
7
Definition 2.9 :
Let H and N be groups.
(i) The direct product H × N is the group given by all pairs (h, n), h ∈ H, n ∈ N
with multiplication and inverse
(h, n) · (h0 , n0 ) = (h · h0 , n · n0 ) , (h, n)−1 = (h−1 , n−1 ) . (2.10)
(ii) Let ϕ : H → Aut(N ), h 7→ ϕh be a group homomorphism. The semidirect
product H nϕ N (or just H n N for short) is the group given by all pairs (h, n),
h ∈ H, n ∈ N with multiplication and inverse
(h, n) · (h0 , n0 ) = (h · h0 , n · ϕh (n0 )) , (h, n)−1 = (h−1 , ϕh−1 (n−1 )) . (2.11)
Exercise 2.7 :
(i) Starting from the definition of the semidirect product, show that H nϕ N is
indeed a group. [To see why the notation H and N is used for the two groups,
look up “semidirect product” on wikipedia.org or eom.springer.de.]
(ii) Show that the direct product is a special case of the semidirect product.
(iii) Show that the multiplication rule (T, x) · (R, y) = (T R, T y + x) found in the
study of E(n) is that of the semidirect product O(n) nϕ Rn , with ϕ : O(n) →
Aut(Rn ) given by ϕT (u) = T u.
Altogether, one finds that the euclidean group is isomorphic to a semidirect
product
E(n) ∼= O(n) n Rn . (2.12)
2.3 Lorentz and Poincaré transformations

The n-dimensional Minkowski space is Rn together with the non-degenenerate
bilinear form
η(u, v) = u0 v0 − u1 v1 − · · · − un−1 vn−1 . (2.13)
Here we labelled the components of a vector u starting from zero, u0 is the
‘time’ coordinate and u1 , . . . , un−1 are the ‘space’ coordinates.
The symmetries of Minkowski space are described by the Lorentz group, if
one wants to keep the point zero fixed, or by the Poincaré group, if just distances
w.r.t. η should remain fixed.
Definition 2.10 :
(i) The Lorentz group O(1, n−1) is defined to be
O(1, n−1) = {M ∈ GL(n, R)|η(M u, M v) = η(u, v) for all u, v ∈ Rn } . (2.14)
(ii) The Poincaré group P (1, n−1) [there does not seem to be a standard symbol;
we will use P ] is defined to be
P (1, n−1) = {f : Rn → Rn | η(x−y, x−y) = η(f (x)−f (y), f (x)−f (y))
for all x, y ∈ Rn } .
(2.15)
8
Exercise 2.8 :
Show that O(1, n−1) can equivalently be written as
O(1, n−1) = {M ∈ GL(n, R)|M t JM = J}
where J is the diagonal matrix with entries J = diag(1, −1, . . . , −1).
Similar to the euclidean group E(n), an element of the Poincaré group can
be written as a composition of a Lorentz transformation Λ ∈ O(1, n−1) and a
translation.
Exercise 2.9 :
(i*) Prove that for every f ∈ P (1, n−1) there is a unique Λ ∈ O(1, n−1) and
u ∈ Rn , s.t. f (v) = Λv + u for all v ∈ Rn .
(ii) Show that the Poincaré group is isomorphic to the semidirect product
O(1, n−1) n Rn with multiplication
(Λ, u) · (Λ0 , u0 ) = (ΛΛ0 , Λu0 + u) . (2.16)
2.4 (*) Symmetries in quantum mechanics

In quantum mechanics, symmetries are at their best [attention: personal opin-
ion]. In particular, the representations of symmetries on vector spaces play an
important role. We will get to that in section 4.1.
Definition 2.11 :
Given a vector space E and two linear maps A, B ∈ End(E) [the endomorphisms
of a vector space E are linear maps from E to E], the commutator [A, B] is
[A, B] = AB − BA ∈ End(E) . (2.17)
Lemma 2.12 :
Given a vector space E, the space of linear maps End(E) together with the
commutator as Lie bracket is a Lie algebra. This Lie algebra will be called
gl(E), or also End(E).
The reason to call this Lie algebra gl(E) will become clear later. Let us use
the proof of this lemma to recall what a Lie algebra is.
Proof of lemma:
Abbreviate V = End(E).
(a) [ , ] has to be a bilinear map from V × V to V .
Clear.
(b) [ , ] has to obey [A, A] = 0 for all A ∈ V (skew-symmetry).
Clear.
(c) [ , ] has to satisfy the Jacobi identity [A, [B, C]] + [B, [C, A]] + [C, [A, B]] = 0
for all A, B, C ∈ V .
This is the content of the next exercise. It follows that V is a Lie algebra.
9
Exercise 2.10 :
Verify that the commutator [A, B] = AB − BA obeys the Jacobi identity.
The states of a quantum system are collected in a Hilbert space H over C.
[Recall: Hilbert spacep = vector space with inner product (·, ·) which is complete
w.r.t. the norm |u| = (u, u).] The time evolution is described by a self-adjoint
operator H [i.e. H † = H] on H. If ψ(0) ∈ H is the state of the system at time
zero, then at time t the system is in the state
2
t t 1 t
ψ(t) = exp( i~ H)ψ(0) = 1 + i~ H + 2! i~ H + . . . ψ(0) . (2.18)
[One should worry if the infinite sum converges. For finite-dimensional H it

always does, see section 3.2.] Suppose we are given a self-adjoint operator A
which commutes with the Hamiltonian,
[A, H] = 0 . (2.19)
Consider the family of operators UA (s) = exp(isA) for s ∈ R. The UA (s) are
unitary (i.e. UA (s)† = UA (s)−1 ) so they preserve probabilities (write U = UA (s))
|(U ψ, U ψ 0 )|2 = |(ψ, U † U ψ 0 )|2 = |(ψ, ψ 0 )|2 . (2.20)
Further, they commute with time-evolution
ψ
evolve / exp( t H)ψ (2.21)
i~
UA (s) = UA (s)

t
evolve / UA (s) exp( i~ H)ψ
UA (s)ψ t
= exp( i~ H)UA (s)ψ
The last equality holds because A and H commute. Thus from A we obtain a
continous one-parameter family of symmetries.
Some comments:
The operator A is also called generator of a symmetry. If we take s to be
very small we have UA (s) = 1 + isA + O(s2 ), and A can be thought of as an
infinitesimal symmetry transformation.
The infinitesimal symmetry transformations are easier to deal with than the
whole family. Therefore one usually describes continuous symmetries in terms
of their generators.
The relation between a continuous family of symmetries and their generators
will in essence be the relation between Lie groups and Lie algebras, the latter
are an infinitesimal version of the former. It turns out that Lie algebras are
much easier to work with and still capture most of the structure.
10
2.5 (*) Angular momentum in quantum mechanics
Consider a quantum mechanical state ψ in the position representation, i.e. a
wave function ψ(q). It is easy to see how to act on this with translations,
(Utrans (s)ψ)(x) = ψ(q + s) . (2.22)
So what is the infinitesimal generator of translations? Take s to be small to find

∂
(Utrans (s)ψ)(q) = ψ(q) + s ψ(q) + O(s2 ) , (2.23)
∂q
so that (the convention ~ = 1 is used)
∂ ∂
Utrans (s) = 1 + s + O(s2 ) = 1 + isp + O(s2 ) with p = −i . (2.24)
∂q ∂q
The infinitesimal generators of rotations in three dimensions are
3
X
Li = εijk qj pk , with i = 1, 2, 3 , (2.25)
j,k=1
and where εijk is antisymmetric in all indices and ε123 = 1.
Exercise 2.11 :
(i) Consider a rotation around the 3-axis,
(Urot (θ)ψ)(q1 , q2 , q3 ) = ψ(q1 cos θ − q2 sin θ, q2 cos θ + q1 sin θ, q3 ) (2.26)
and check that infinitesimally
Urot (θ) = 1 + iθL3 + O(θ2 ) . (2.27)
(ii) Using [qr , ps ] = iδrs (check!) verify the commutator

3
X
[Lr , Ls ] = i εrst Lt . (2.28)
t=1
P3
(You might need the relation k=1 εijk εlmk = δil δjm − δim δjl (check!).)
The last relation can be used to define a three-dimensional Lie algebra: Let
V be the complex vector space spanned by three generators `1 , `2 , `3 . Define
the bilinear map [ , ] on generators as
3
X
[`r , `s ] = i εrst `t . (2.29)
t=1
This turns V into a complex Lie algebra:

- skew-symmetry [x, x] = 0 : ok.
11
- Jacobi identity : turns out ok.
We will later call this Lie algebra sl(2, C).
This Lie algebra is particularly important for atomic spectra, e.g. for the hy-
drogen atom, because the electrons move in a rotationally symmetric potential.
This implies [Li , H] = 0 and acting with one of the Li on an energy eigenstate
gives results in an energy eigenstate of the same energy. We say: The states
at a given energy have to form a representation of sl(2, C). Representations of
sl(2, C) are treated in section 4.2.
3 Matrix Lie Groups and their Lie Algebras

3.1 Matrix Lie groups
Definition 3.1 :
A matrix Lie group is a closed subgroup of GL(n, R) or GL(n, C) for some n ≥ 1.
Comments:
‘closed’ in this definition stands for ‘closed as a subset of the topological space
GL(n, R) (resp. GL(n, C))’. It is equivalent to demanding that given a sequence
An of matrices belonging to a matrix subgroup H s.t. A = limn→∞ An exists
and is in GL(n, R) (resp. GL(n, C)), then already A ∈ H.
A matrix Lie group is a Lie group. However, not every Lie group is isomorphic
to a matrix Lie group. We will not prove this. If you are interested in more
details, consult e.g. [Baker, Theorem 7.24] and [Baker, Section 7.7].
So far we have met the groups
invertible linear maps GL(n, R) and GL(n, C). In general we set GL(V ) = {
invertible linear maps V → V }, such that GL(n, R) = GL(Rn ), etc.
Some subgroups of GL(n, R), namely O(n) = {M ∈ Mat(n, R)|M t M = 1}
and SO(n) = {M ∈ O(n)| det(M ) = 1}.
Some semidirect products, E(n) ∼= O(n) n Rn and P (1, n−1) ∼= O(1, n−1) n
n
R .
All of these are matrix Lie groups, or isomorphic to matrix Lie groups:
For O(n) and SO(n) we already know that they are subgroups of GL(n, R).
It remains to check that they are closed as subsets of GL(n, R). This follows
since for a continuous function f and any sequence an with limit limn→∞ an = a
we have limn→∞ f (an ) = f (a). The defining relations M 7→ M t M and M 7→
det(M ) are continuous functions.
Alternatively one can argue as follows: The preimage of a closed set under a
continuous map is closed. The one-point sets {1} ⊂ Mat(n, R) and {1} ⊂ R are
closed.
For the groups E(n) and P (1, n−1) we use the following lemma.
12
Lemma 3.2 :
Let ϕ : Mat(n, R) × Rn → Mat(n+1, R) be the map

M v
ϕ(M, v) = . (3.1)
0 1
(i) ϕ restricts to an injective group homomorphism from O(n)nRn to GL(n+1, R),
and from O(1, n−1) n Rn to GL(n+1, R).
(ii) The images ϕ(O(n) n Rn ) and ϕ(O(1, n−1) n Rn ) are closed subsets of
GL(n+1, R).
Proof:
(i) We need to check that
ϕ((R, u) · (S, v)) = ϕ(R, u) · ϕ(S, v) . (3.2)
The lhs is equal to

RS Rv + u
ϕ((R, u) · (S, v)) = ϕ(RS, Rv + u) = (3.3)
0 1
while the rhs gives

R u S v RS Rv + u
ϕ(R, u) · ϕ(S, v) = = , (3.4)
0 1 0 1 0 1
so ϕ is a group homomorphism. Further, it is clearly injective.
(ii) The images of O(n) n Rn and O(1, n−1) n Rn under ϕ consist of all matrices

R u
(3.5)
0 1
with u ∈ Rn and R an element of O(n) and O(1, n−1), respectively. This is a
closed subset of GL(n+1, R) since O(n) (resp. O(1, n−1)) and Rn are closed.
Here are some matrix Lie groups which are subgroups of GL(n, C).
Definition 3.3 :
On Cn define the inner product
n
X
(u, v) = (uk )∗ vk . (3.6)
k=1
Then the unitary group U (n) is given by

U (n) = {A ∈ Mat(n, C)|(Au, Av) = (u, v) for all u, v ∈ Cn } (3.7)
and the special unitary group SU (n) is given by
SU (n) = {A ∈ U (n)| det(A) = 1} . (3.8)
13
Exercise 3.1 :
(i) Show that U (n) and SU (n) are indeed groups.
(ii) Let (A† )ij = (Aji )∗ be the hermitian conjugate. Show that the condition
(Au, Av) = (u, v) for all u, v ∈ Cn is equivalent to A† A = 1, i.e.
U (n) = {A ∈ Mat(n, C) | A† A = 1} .
(iii) Show that U (n) and SU (n) are matrix Lie groups.
Definition 3.4 :
For K = R or K = C, the special linear group SL(n, K) is given by
SL(n, K) = {A ∈ Mat(n, K)| det(A) = 1} . (3.9)
3.2 The exponential map
Definition 3.5 :
The exponential of a matrix X ∈ Mat(n, K), for K = R or C, is
∞
X 1 k
exp(X) = X = 1 + X + 12 X 2 + . . . (3.10)
k!
k=0
Lemma 3.6 :
The series defining exp(X) converges absolutely for all X ∈ Mat(n, K).
Proof:
Choose your favorite norm on Mat(n, K), say
n
X
kXk = |Akl | . (3.11)
k,l=1
P∞ 1 k
The series exp(X) converges absolutely if the series of norms k=0 k! kX k
converges. This in turn follows since kXY k ≤ kXk kY k and since the series ea
converges for all a ∈ R.
The following exercise shows a convenient way to compute the exponential
of a matrix via its Jordan normal form (→ wikipedia.org, eom.springer.de).
Exercise 3.2 :
(i) Show that for λ ∈ C,

λ 1 1 1
exp = eλ · .
0 λ 0 1
14
(ii) Let A ∈ Mat(n, C). Show that for any U ∈ GL(n, C)
U −1 exp(A)U = exp(U −1 AU ) .
(iii) Recall that a complex n × n matrix A can always be brought to Jordan

normal form, i.e. there exists an U ∈ GL(n, C) s.t.
 
J1 0
U −1 AU = 
 ..  ,

.
0 Jr
where each Jordan block is of the form

 
λk 1 0
 .. .. 
 . . 
Jk =   , λk ∈ C .
 .. 
 . 1 
0 λk
In particular, if all Jordan blocks have size 1, the matrix A is diagonalisable.

Compute
0 t 5 9
exp and exp .
−t 0 −1 −1
Exercise 3.3 :
Let A ∈ Mat(n, C).
(i) Let f (t) = det(exp(tA)) and g(t) = exp(t tr(A)). Show that f (t) and g(t)
both solve the first order DEQ u0 = tr(A) u.
(ii) Using (i), show that
det(exp(A)) = exp(tr(A)) .
Exercise 3.4 :
Show that if A and B commute (i.e. if AB = BA), then exp(A) exp(B) =
exp(A + B).
3.3 The Lie algebra of a matrix Lie group

In this section we will look at the relation between matrix Lie groups and Lie
algebras. As the emphasis on this course will be on Lie algebras, in this section
we will state some results without proof.
Defintion 1.1:
A Lie algebra is a vector space V together with a bilinear map [ , ] : V ×V → V ,
called Lie bracket, satisfying
15
(i) [X, X] = 0 for all X ∈ V (skew-symmetry),
(ii) [X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y ]] = 0 for all X, Y, Z ∈ V (Jacobi identity).
If you have read through sections 2.4 and 2.5 you may jump directly to
definition 3.7 below. If not, here are the definition of a commmutator and of
the Lie algebra gl(E) ≡ End(E) restated.
Defintion 2.11:
Given a vector space E and two linear maps A, B ∈ End(E) [the endomorphisms
of a vector space E are the linear maps from E to E], the commutator [A, B] is
[A, B] = AB − BA ∈ End(E) . (3.12)
Lemma 2.12:
Given a vector space E, the space of linear maps End(E) together with the
commutator as Lie bracket is a Lie algebra. This Lie algebra will be called
gl(E), or also End(E).
Proof:
Abbreviate V = End(E).
(a) [ , ] has to be a bilinear map from V × V to V .
Clear.
(b) [ , ] has to obey [A, A] = 0 for all A ∈ V (skew-symmetry).
Clear.
(c) [ , ] has to satisfy the Jacobi identity [A, [B, C]] + [B, [C, A]] + [C, [A, B]] = 0
for all A, B, C ∈ V .
This is the content of the next exercise. It follows that V is a Lie algebra.
Exercise 2.10:
The above exercise also shows that the n×n matrices Mat(n, K), for K = R
or K = C, form a Lie algebra with the commutator as Lie bracket. This Lie
algebra is called gl(n, K). (In the notation of lemma 2.12, gl(n, K) is the same
as gl(Kn ) ≡ End(Kn ).)
Definition 3.7 :
A Lie subalgebra h of a Lie algebra g is a a sub-vector space h of g such that
whenever A, B ∈ h then also [A, B] ∈ h.
Definition 3.8 :
Let G be a matrix Lie group in Mat(n, K), for K = R or K = C.
(i) The Lie algebra of G is
g = {A ∈ Mat(n, K)| exp(tA) ∈ G for all t ∈ R} . (3.13)
(ii) The dimension of G is the dimension of its Lie algebra (which is a vector
space over R).
16
The following theorem justifies the name ‘Lie algebra of a matrix Lie group’.
We will not prove it, but rather verify it in some examples.
Theorem 3.9 :
The Lie algebra of a matrix Lie group, with commutator as Lie bracket, is a Lie
algebra over R (in fact, a Lie subalgebra of gl(n, R)).
What one needs to show is that first, g is a vector space, and second, that
for A, B ∈ g also [A, B] ∈ g. The following exercise indicates how this can be
done.
Exercise* 3.5 :
Let G be a matrix Lie group and let g be the Lie algebra of G.
(i) Show that if A ∈ g, then also sA ∈ g for all s ∈ R.
(ii) The following formulae hold for A, B ∈ Mat(n, K): the Trotter Product
Formula, n
exp(A + B) = lim exp(A/n) exp(B/n) ,
n→∞
and the Commutator Formula,

n2
exp([A, B]) = lim exp(A/n) exp(B/n) exp(−A/n) exp(−B/n) .
n→∞
(For a proof see [Baker, Theorem 7.26]). Use these to show that if A, B ∈ g,
then also A + B ∈ g and [A, B] ∈ g. (You will need that a matrix Lie group is
closed.) Note that part (i) and (ii) combined prove Theorem 3.9.
3.4 A little zoo of matrix Lie groups and their Lie algebras
Here we collect the ‘standard’ matrix Lie groups (i.e. those which are typically
mentioned without further explanation in text books). Before getting to the
table, we need to define one more matrix Lie algebra.
Definition 3.10 :
Let 1n×n be the n × n unit matrix, and let

0 1n×n
Jsp = ∈ Mat(2n, R) .
−1n×n 0
The set
SP (2n) = {M ∈ Mat(2n, R)|M t Jsp M = Jsp }
is called the 2n × 2n (real) symplectic group.
17
Exercise 3.6 :
Prove that SP (2n) is a matrix Lie group.
Mat. Lie gr. Lie algebra of the matrix Lie group Dim. (over R)
GL(n, R) gl(n, R) = Mat(n, R) n2

GL(n, C) gl(n, C) = Mat(n, C) 2n2
SL(n, R) sl(n, R) = {A ∈ Mat(n, R)|tr(A) = 0} n2 − 1
SL(n, C) sl(n, C) = {A ∈ Mat(n, C)|tr(A) = 0} 2n2 − 2
t 1
O(n) o(n) = {A ∈ Mat(n, R)|A + A = 0} 2 n(n − 1)
SO(n) so(n) = o(n)
SP (2n) sp(2n) = {A ∈ Mat(2n, R)|Jsp A + At Jsp = 0} n(2n + 1)
U (n) u(n) = {A ∈ Mat(n, C)|A + A† = 0} n2
SU (n) su(n) = {A ∈ u(n)|tr(A) = 0} n2 − 1
Let us verify this list.

GL(n, R):
We have to find all elements A ∈ Mat(n, R) such that exp(sA) ∈ GL(n, R) for
all s ∈ R. But exp(sA) is always invertible. Hence the Lie algebra of GL(n, R)
is just Mat(n, R) and its real dimension is n2 .
GL(n, C):
Along the same lines as for GL(n, R) we find that the Lie algebra of GL(n, C)
is Mat(n, C). As a vector space over R it has dimension 2n2 .
SL(n, R):
What are all A ∈ Mat(n, R) such that det(exp(sA)) = 1 for all s ∈ R? Use
det(exp(sA)) = es tr(A) (3.14)
to see that tr(A) = 0 is necessary and sufficient. The subspace of matrices with
tr(A) = 0 has dimension n2 − 1.
O(n):
What are all A ∈ Mat(n, R) s.t. (exp(sA))t exp(sA) = 1 for all s ∈ R? First,
suppose that M = exp(sA) has the property M t M = 1. Expanding this in s,
1 = (1 + sAt )(1 + sA) + O(s2 ) = 1 + s(At + A) + O(s2 ) , (3.15)
which shows that At + A = 0 is a necessary condition for A to be in the Lie

algebra of O(n). Further, it is also sufficient since A + At = 0 implies
(exp(sA))t exp(sA) = exp(sAt ) exp(sA) = exp(−sA) exp(sA) = 1 . (3.16)
In components, the condition At + A = 0 implies Aii = 0 and Aij = −Aji . Thus

only the entries Aij with 1 ≤ i < j ≤ n can be choosen freely. The dimension
of o(n) is therefore 21 n(n − 1).
18
SO(n):
What are all A ∈ Mat(n, R) s.t. exp(sA) ∈ O(n) and det(exp(sA)) = 1 for all
s ∈ R? First, exp(sA) ∈ O(n) (for all s ∈ R) is equivalent to A + At = 0.
Second, as for SL(n, K) use
1 = det(exp(sA)) = es tr(A) (3.17)
to see that further tr(A) = 0 is necessary and sufficient. However, A + At = 0

already implies tr(A) = 0. Thus SO(n) and O(n) have the same Lie algebra.
SU (n):
Here the calculation is the same as for SO(n), except that now A† + A = 0 does
not imply that tr(A) = 0, so this is an extra condition.
Exercise 3.7 :
In the table of matrix Lie algebras, verify the entries for SL(n, C), SP (2n),
U (n) and confirm the dimension of SU (n).
A Lie algebra probes the structure of a Lie group close to the unit element.
If the Lie algebras of two Lie groups agree, the two Lie groups look alike in a
neighbourhood of the unit, but may still be different. For example, even though
o(n) = so(n) we still have O(n) SO(n).
Information 3.11 :
This is easiest to see via topological considerations (which we will not treat in
this course). The group SO(n) is path connected, which means that for any
p, q ∈ SO(n) there is a continuous map γ : [0, 1] → SO(n) such that γ(0) = p
and γ(1) = q [Baker, section 9]. However, O(n) cannot be path connected.
To see this choose p, q ∈ O(n) such that det(p) = 1 and det(q) = −1. The
composition of a path γ with det is continuous, and on O(n), det only takes
values ±1, so that it cannot change from 1 to −1 along γ. Thus there is no
path from p to q. In fact, O(n) has two connected components, and SO(n) is
the connected component containing the identity.
3.5 Examples: SO(3) and SU (2)

SO(3)
We will need the following two notations. Let E(n)ij denote the n × n-matrix
which has only one nonzero matrix element at position (i, j), and this matrix
element is equal to one,
E(n)ij kl = δik δjl . (3.18)
For example,  
0 1 0
E(3)12 = 0 0 0  . (3.19)
0 0 0
19
If the value of n is clear we will usually abbreviate Eij ≡ E(n)ij .
Let εijk , i, j, k ∈ {1, 2, 3} be totally anti-symmetric in all indices, and let
ε123 = 1.
Exercise 3.8 :
(i) Show that Eab Ecd = δbc Ead .
P3
(ii) Show that x=1 εabx εcdx = δac δbd − δad δbc .
The Lie algebra so(3) of the matrix Lie group SO(3) consists of all real, anti-
symmetric 3×3-matrices. The following three matrices form a basis of so(3),
     
0 0 0 0 0 −1 0 1 0
J1 =  0 0 1  , J2 =  0 0 0  , J3 =  −1 0 0  .
0 −1 0 1 0 0 0 0 0
(3.20)
Exercise 3.9 :
(i) Show that the generators J1 , J2 , J3 can also be written as
3
X
Ja = εabc Ebc ; a ∈ {1, 2, 3} .
b,c=1
P3
(ii) Show that [Ja , Jb ] = − c=1 εabc Jc
(iii) Check that R3 (θ) = exp(−θJ3 ) is given by
 
cos(θ) − sin(θ) 0
R3 (θ) =  sin(θ) cos(θ) 0  .
0 0 1
This is a rotation by an angle θ around the 3-axis. Check explicitly that R3 (θ) ∈
SO(3).
SU(2)
The Pauli matrices are defined to be the following elements of Mat(2, C),

0 1 0 −i 1 0
σ1 = , σ2 = , σ3 = . (3.21)
1 0 i 0 0 −1
Exercise 3.10 : P
Show that for a, b ∈ {1, 2, 3}, [σa , σb ] = 2i c εabc σc .
The Lie algebra su(2) consists of all anti-hermitian, trace-less complex 2 × 2
matrices.
Exercise 3.11 :
(i) Show that the set {iσ1 , iσ2 , iσ3 } is a basis of su(2) as a real vector space.
Convince yourself that the set {σ1 , σ2 , σ3 } does not form a basis of su(2) as a
real vector space. P3
(ii) Show that [iσa , iσb ] = −2 c=1 εabc iσc .
20
so(3) and su(2) are isomorphic
Definition 3.12 :
Let g, h be two Lie algebras.
(i) A linear map ϕ : g → h is a Lie algebra homomorphism iff
ϕ([a, b]) = [ϕ(a), ϕ(b)] for all a, b ∈ g .
(ii) A Lie algebra homomorphism ϕ is a Lie algebra isomorphism iff it is invert-

ible.
If we want to emphasise that g and h are Lie algebras over R, we say that
ϕ : g → h is a homomorphism (or isomorphism) of real Lie algebras. We also
say complex Lie algebra for a Lie algebra whose underlying vector space is over
C.
Exercise 3.12 :
Show that so(3) and su(2) are isomorphic as real Lie algebras.
Also in this case one finds that even though so(3) ∼
= su(2), the Lie groups
SO(3) and SU (2) are not isomorphic.
Information 3.13 :
This is again easiest seen by topological arguments. One finds that SU (2)
is simply connected, i.e. every loop embedded in SU (2) can be contracted to a
point, while SO(3) is not simply connected. In fact, SU (2) is a two-fold covering
of SO(3).
3.6 Example: Lorentz group and Poincaré group

Commutators of o(1, n−1).
Recall that the Lorentz group was given by
O(1, n−1) = {M ∈ GL(n, R)|M t JM = J} (3.22)
where J is the diagonal matrix with entries J = diag(1, −1, . . . , −1), and that
these linear maps preserve the bilinear form
η(x, y) = x0 y0 − x1 y1 − · · · − xn−1 yn−1 (3.23)

n n
on
P R . Let e0 , e1 , . . . , en−1 be the standard basis of R (i.e. x = (x0 , . . . , xn−1 ) =
k xk ek ). We will use the numbers
ηkl = η(ek , el ) = Jkl . (3.24)
21
Exercise 3.13 :
Show that the Lie algebra of O(1, n−1) is
o(1, n−1) = {A ∈ Mat(n, R)|At J + JA = 0} .
If we write the matrices A ∈ o(1, n−1) in block form, the condition At J +

JA = 0 becomes
a ct

1 0 1 0 a b
+
bt D t 0 −1 0 −1 c D
t
(3.25)
a −c a b
= + = 0
bt −Dt −c −D
where a ∈ C and D ∈ Mat(n−1, R). Thus a = 0, c = bt and Dt = −D.
Counting the free parameters gives the dimension to be
dim(o(1, n−1)) = n−1 + 12 (n−1)(n−2) = 21 n(n−1) . (3.26)
Consider the following elements of o(1, n−1)),
Mab = ηbb Eab − ηaa Eba a, b ∈ {0, 1, . . . , n−1} . (3.27)
These obey Mab = −Mba and the set {Mab |0 ≤ a < b ≤ n−1} forms a basis of
o(1, n−1)).
Exercise 3.14 :
Check that the commutator of the Mab ’s is
[Mab , Mcd ] = ηad Mbc + ηbc Mad − ηac Mbd − ηbd Mac .
Commutators of p(1, n−1).
In lemma 3.2 we found an embedding of the Poincaré group P (1, n−1) into
Mat(n+1, R). Let us denote the image in Mat(n+1, R) by P̃ (1, n−1). In the
same lemma, we checked that P̃ (1, n−1) is a matrix Lie group. Let us compute
its Lie algebra p(1, n−1).
Exercise 3.15 :
(i) Show that, for A ∈ Mat(n, R) and u ∈ Rn ,
A ∞
A u e Bu X 1 n−1
exp = , B= A .
0 0 0 1 n!
n=1
[If A is invertible, then B = A−1 (eA − 1).]

(ii) Show that the Lie algebra of P̃ (1, n−1) (the Poincaré group embedded in
Mat(n+1, R)) is
n A x o
p(1, n−1) = A ∈ o(1, n−1) , x ∈ Rn .

0 0
22
Let us define the generators Mab for a, b ∈ {0, 1, . . . , n−1} as before and set
in addition
Pa = Ean , a ∈ {0, 1, . . . , n−1} . (3.28)
Exercise 3.16 :
Show that, for a, b, c ∈ {0, 1, . . . , n−1},
[Mab , Pc ] = ηbc Pa − ηac Pb , [Pa , Pb ] = 0 .
We thus find that altogether the Poincaré algebra p(1, n−1) has basis
{Mab |0 ≤ a < b ≤ n−1} ∪ {Pa |0 ≤ a ≤ n−1} (3.29)
which obey the commutation relations

[Mab , Mcd ] = ηad Mbc + ηbc Mad − ηac Mbd − ηbd Mac ,
[Mab , Pc ] = ηbc Pa − ηac Pb , (3.30)
[Pa , Pb ] = 0 .
3.7 Final comments: Baker-Campbell-Hausdorff formula

Here are some final comments before we concentrate on the study of Lie algebras.
Let g be the Lie algebra of a matrix Lie group G.
For X, Y ∈ g close enough to zero, we have
exp(X) exp(Y ) = exp(X ? Y ) , (3.31)
where
X ? Y = X + Y + 21 [X, Y ] + 1
12 [X, [X, Y ]] + 1
12 [Y, [Y, X]] + ... (3.32)
can be expressed entirely in terms of commutators (which we will not prove).

This is known as the Baker-Campbell-Hausdorff identity. For a proof, see [Bour-
baki “Groupes et algèbres de Lie” Ch. II § 6 n◦ 2 Thm. 1], and for an explicit
formula [n◦ 4] of the same book.
Thus the Lie algebra g encodes all the information (group elements and their
multiplication) of G in a neighbourhood of 1 ∈ G.
Exercise 3.17 :
There are some variants of the BCH identity which are also known as Baker-
Campbell-Hausdorff formulae. Here we will prove some.
Let ad(A) : Mat(n, C) → Mat(n, C) be given by ad(A)B = [A, B]. [This is
called the adjoint action.]
(i) Show that for A, B ∈ Mat(n, C),
f (t) = etA Be−tA and g(t) = etad(A) B
23
both solve the first order DEQ
d
u(t) = [A, u(t)] .
dt
(ii) Show that
eA Be−A = ead(A) B = B + [A, B] + 21 [A, [A, B]] + . . .
(iii) Show that

eA eB e−A = exp(ead(A) B) .
(iv) Show that if [A, B] commutes with A and B,
eA eB = e[A,B] eB eA .
(v) Suppose [A, B] commutes with A and B. Show that f (t) = etA etB and
1 2 d
g(t) = etA+tB+ 2 t [A,B] both solve dt u(t) = (A + B + t[A, B])u(t). Show further
that
1
eA eB = eA+B+ 2 [A,B] .
4 Lie algebras
In this course we will only be dealing with vector spaces over R or C. When a
definition or statement works for either of the two, we will write K instead of
R or C. (In fact, when we write K below, the statement or definition holds for
every field.)
4.1 Representations of Lie algebras
Definition 4.1 :
Let g be a Lie algebra over K. A representation (V, R) of g is a K-vector
space V together with a Lie algebra homomorphism R : g → End(V ). The
vector space V is called representation space and the linear map R the action
or representation map. We will sometimes abbreviate V ≡ (V, R).
In other words, (V, R) is a representation of g iff
R(x) ◦ R(y) − R(y) ◦ R(x) = R( [x, y] ) for all x, y ∈ g . (4.1)
Exercise 4.1 :
It is also common to use ‘modules’ instead of representations. The two concepts
are equivalent, as will be clear by the end of this exercise.
Let g be a Lie algebra over K. A g-module V is a K-vector space V together
with a bilinear map . : g × V → V such that
[x, y].w = x.(y.w) − y.(x.w) for all x, y ∈ g, w ∈ V . (4.2)
24
(i) Show that given a g-module V , one gets a representation of g by setting
R(x)w = x.w.
(ii) Given a representation (V, R) of g, show that setting x.w = R(x)w defines
a g-module on V .
Given a representation (V, R) of g and elements x ∈ g, w ∈ V , we will
sometimes abbreviate x.w ≡ R(x)w.
Definition 4.2 :
Let g be a Lie algebra.
(i) A representation (V, R) of g is faithful iff R : g → End(V ) is injective.
(ii) An intertwiner between two representations (V, RV ) and (W, RW ) is a linear
map f : V → W such that
f ◦ RV (x) = RW (x) ◦ f . (4.3)
(iii) Two representations RV and RW are isomorphic if there exists an invertible

intertwiner f : V → W .
In particular, two representations whose representation spaces are of dif-
ferent dimension are never isomorphic. There are two representations one can
construct for any Lie algebra g over K.
The trivial representation is given by taking K as representation space (i.e.
the one-dimensional K-vector space K itself) and defining R : g → End(K) to
be R(x) = 0 for all x ∈ g. In short, the trivial representation is (K, 0).
The second representation is more interesting. For x ∈ g define the map
adx : g → g as
adx (y) = [x, y] for all y ∈ g . (4.4)
Then x 7→ adx defines a linear map ad : g → End(g). This can be used to define
a representation of g on itself. In this way one obtains the adjoint representation
(g, ad). This is indeed a representation of g because
(adx ◦ ady − ady ◦ adx )(z) = [x, [y, z]] − [y, [x, z]]
(4.5)
= [x, [y, z]] + [y, [z, x]] = −[z, [x, y]] = ad[x,y] (z) .
Exercise 4.2 :
Show that for the Lie algebra u(1), the trivial and the adjoint representation
are isomorphic.
Given a representation R of g on Kn we define the dual representation R+
via
R+ (x) = −R(x)t for all x ∈ g . (4.6)
That is, for the n × n matrix R+ (x) ∈ End(Kn ) we take minus the transpose of
the matrix R(x).
25
Exercise 4.3 :
Show that if (Kn , R) is a representation of g, then so is (Kn , R+ ) with R+ (x) =
−R(x)t .
The dual representation can also be defined for a representation R on a
vector space V other then Kn . One then takes R+ to act on the dual vector
space V ∗ and defines R+ (x) = −R(x)∗ , i.e. (V, R)+ = (V ∗ , −R∗ ).
Definition 4.3 :
Let g be a Lie-algebra and let (V, R) be a representation of g.
(i) A sub-vector space U of V is called invariant subspace iff x.u ∈ U for all
x ∈ g, u ∈ U . In this case we call (U, R) a sub-representation of (V, R).
(ii) (V, R) is called irreducible iff V 6={0} and the only invariant subspaces of
(V, R) are {0} and V .
Exercise 4.4 :
Let f : V → W be an intertwiner of two representations V, W of g. Show that
the kernel ker(f ) = {v ∈ V |f (v) = 0} and the image im(f ) = {w ∈ W |w =
f (v) for some v ∈ V } are invariant subspaces of V and W , respectively.
Recall the following result from linear algebra.
Lemma 4.4 :
A matrix A ∈ Mat(n, C), n > 0, has at least one eigenvector.
This is the main reason why the treatment of complex Lie algebras is much
simpler than that of real Lie algebras.
Lemma 4.5 :
(Schur’s Lemma) Let g be a Lie algebra and let U , V be two irreducible repre-
sentations of g. Then an intertwiner f : U → V is either zero or an isomorphism.
Proof:
The kernel ker(f ) is an invariant subspace of U . Since U is irreducible, either
ker(f ) = U or ker(f ) = {0}. Thus either f = 0 or f is injective. The image
im(f ) is an invariant subspace of V . Thus im(f ) = {0} or im(f ) = V , i.e. either
f = 0 or f is surjective. Altogether, either f = 0 or f is a bijection.
Corollary 4.6 :
Let g be a Lie algebra over C and let U , V be two finite-dimensional, irreducible
representations of g.
(i) If f : U → U is an intertwiner, then f = λidU for some λ ∈ C.
(ii) If f1 and f2 are nonzero intertwiners from U to V , then f1 = λf2 for some
λ ∈ C× = C − {0}.
26
Proof:
(i) By lemma 4.4, f has an eigenvalue λ ∈ C. Note that the linear map hλ =
f − λidU is an intertwiner from U to U since, for all x ∈ g, u ∈ U ,
hλ (x.u) = f (x.u) − λx.u = x.f (u) − x.(λu) = x.hλ (u) .
Let u 6= 0 be an eigenvector, f u = λu. Then hλ (u) = 0 so that hλ is not an

isomorphism. By Schur’s lemma hλ = 0 so that f = λidU .
(ii) By Schur’s Lemma, f1 and f2 are isomorphisms. f2−1 ◦ f1 is an intertwiner
from U to U . By part (i), f2−1 ◦ f1 = λidU , which implies f1 = λf2 . As f1 6= 0
we also have λ 6= 0.
4.2 Irreducible representations of sl(2, C)

Recall that
sl(2, C) = {A ∈ Mat(2, C)|tr(A) = 0} . (4.7)
In section 3.4 we saw that this, understood as a real Lie algebra, is the Lie algebra
of the matrix Lie group SL(2, C). However, since Mat(2, C) is a complex vector
space and since the condition tr(A) = 0 is C-linear, we can also understand
sl(2, C) as a complex Lie algebra. We should really use a different symbol for
the two, but by abuse of notation we (and everyone else) will not.
In this section, by sl(2, C) we will always mean the complex Lie algebra. The
aim of this section is to prove the following theorem.
Theorem 4.7 :
The dimension gives a bijection

finite dim. irreducible repns
dim : −→ {1, 2, 3, . . . } . (4.8)
of sl(2, C) up to isomorphism
All matrices A in sl(2, C) are of the form

a b
A= for a, b, c ∈ C . (4.9)
c −a
A convenient basis will be

1 0 0 1 0 0
H= , E= , F = . (4.10)
0 −1 0 0 1 0
Exercise 4.5 :
Check that for the basis elements of sl(2, C) one has [H, E] = 2E, [H, F ] = −2F
and [E, F ] = H.
27
Exercise 4.6 :
Let (V, R) be a representation of sl(2, C). Show that if R(H) has an eigenvector
with non-integer eigenvalue, then V is infinite-dimensional.
Hint: Let H.v = λv with λ ∈ / Z. Proceed as follows.
1) Set w = E.v. Show that either w = 0 or w is an eigenvector of R(H) with
eigenvalue λ + 2.
2) Show that either V is infinite-dimensional or there is an eigenvector v0 of
R(H) of eigenvalue λ0 ∈ / Z such that E.v0 = 0.
m
3) Let vm = F .v0 and define v−1 = 0. Show by induction on m that
H.vm = (λ0 − 2m)vm and E.vm = m(λ0 − m + 1)vm−1 .
4) Conclude that if λ0 ∈
/ Z≥0 all vm are nonzero.
Corollary 4.8 :
(to exercise 4.6) In a finite-dimensional representation (V, R) of sl(2, C) the
eigenvalues of R(H) are integers.
Exercise 4.7 :
The Lie algebra h = CH is a subalgebra of sl(2, C). Show that h has finite-
dimensional representations where R(H) has non-integer eigenvalues.
Next we construct a representation of sl(2, C) for a given dimension.
Lemma 4.9 :
Let n ∈ {1, 2, 3, . . . } and let e0 , . . . , en−1 be the standard basis of Cn . Set
e−1 = en = 0. Then
H.em = (n − 1 − 2m)em
E.em = m(n − m)em−1 (4.11)
F.em = em+1
defines an irreducible representation Vn of sl(2, C) on Cn .
Proof: To see that this is a representation of sl(2, C) we check the definition
explicitly. For example
[E, F ].em = H.em = (n − 1 − 2m)em (4.12)
and
E.(F.em ) − F.(E.em ) = (m + 1)(n − m − 1)em − m(n − m)em
(4.13)
= (n − 1 − 2m)em = [E, F ].em .
To check the remaining conditions is the content of the next exercise.

Irreducibility can be seen as follows. Let W be a nonzero invariant subspace
of Cn . Then R(H)|W has an eigenvector v ∈ W . But v is also an eigenvector of
28
R(H) itself, and (because the em are a basis consisting of eigenvectors of H with
distinct eigenvalues) has to be of the form v = λem , for some m ∈ {0, . . . , n − 1}
and λ ∈ C. Thus W contains in particular the vector em Starting from em one
can obtain all other ek by acting with E and F . Thus W has to contain all ek
and hence W = Cn .
Exercise 4.8 :
Check that the representation of sl(2, C) defined in the lecture indeed also obeys
[H, E].v = 2E.v and [H, F ].v = −2F.v for all v ∈ Cn .
Proof of Theorem 4.7, part I:
Lemma 4.9 shows that the map dim( ) in the statement of Theorem 4.7 is
surjective.
Exercise 4.9 :
Let (W, R) be a finite-dimensional, irreducible representation of sl(2, C). Show
that for some n ∈ Z≥0 there is an injective intertwiner ϕ : Vn → W .
Hint: (recall exercise 4.6)
1) Find a v0 ∈ W such that E.v0 = 0 and H.v0 = λ0 v0 for some h ∈ Z.
2) Set vm = F m .v0 . Show that there exists an n such that vm = 0 for m ≥ n.
Choose the smallest such n.
3) Show that ϕ(em ) = vm for m = 0, . . . , n − 1 defines an injective intertwiner.
Proof of Theorem 4.7, part II:
Suppose (W, R) is a finite-dimensional irreducible representation of sl(2, C). By
exercise 4.9 there is an injective intertwiner ϕ : Vn → W . By Schur’s lemma, as
ϕ is nonzero, it has to be an isomorphism. This shows that the map dim( ) in
the statement of Theorem 4.7 is injective. Since we already saw that it is also
surjective, it is indeed a bijection.
4.3 Direct sums and tensor products
Definition 4.10 :
Let U, V be two K-vector spaces.
(i) The direct sum of U and V is the set
U ⊕ V = { (u, v) | u ∈ U, v ∈ V } (4.14)
with addition and scalar multiplication defined to be
(u, v) + (u0 , v 0 ) = (u + u0 , v + v 0 ) and λ(u, v) = (λu, λv) (4.15)
for all u ∈ U , v ∈ V , λ ∈ K. We will write u ⊕ v ≡ (u, v).

(ii) The tensor product of U and V is the quotient vector space
U ⊗ V = spanK ((u, v)| u ∈ U, v ∈ V )/W (4.16)
29
where W is the K-vector space spanned by the vectors
(λ1 u1 + λ2 u2 , v) − λ1 (u1 , v) − λ2 (u2 , v) , λ1 , λ2 ∈ K , u1 , u2 ∈ U , v ∈ V .
(u, λ1 v1 + λ2 v2 ) − λ1 (u, v1 ) − λ2 (u, v2 ) , λ1 , λ2 ∈ K , u ∈ U , v1 , v2 ∈ V .
The equivalence class containing (u, v) is denoted by (u, v) + W or by u ⊗ v.

What the definition of the tensor product means is explained in the following
lemma, which can also be understood as a pragmatic definition of U ⊗ V .
Lemma 4.11 :
(i) Every element of U ⊗ V can be written in the form u1 ⊗ v1 + · · · + un ⊗ vn .
(ii) In U ⊗ V we can use the following rules
(λ1 u1 + λ2 u2 ) ⊗ v = λ1 u1 ⊗ v + λ2 u2 ⊗ v , λ1 , λ2 ∈ K , u1 , u2 ∈ U , v ∈ V .
u ⊗ (λ1 v1 + λ2 v2 ) = λ1 u ⊗ v1 + λ2 u ⊗ v2 , λ1 , λ2 ∈ K , u ∈ U , v1 , v2 ∈ V .
Proof:
(ii) is an immediate consequence of the definition: Take the first equality as
an example. The difference between the representative (λ1 u1 + λ2 u2 , v) of the
equivalence class on the lhs and the representative λ1 (u1 , v) + λ2 (u2 , v) of the
equivalence class on rhs lies in W , i.e. in the equivalence class of zero.
(i) By definition, any q ∈ U ⊗ V is the equivalence class of an element of the
form
q = λ1 (u1 , v1 ) + · · · + λn (un , vn ) + W (4.17)
for some n > 0. But this is just the equivalence class denoted by
q = λ1 · u1 ⊗ v1 + · · · + λn · un ⊗ vn . (4.18)
By part (ii), we in particular have λ(u ⊗ v) = (λu) ⊗ v so that the above vector
can be written as
q = (λ1 u1 ) ⊗ v1 + · · · + (λn un ) ⊗ vn , (4.19)
which is of the desired form.
Exercise* 4.10 :
Let U, V be two finite-dimensional K-vector spaces. Let u1 , . . . , um be a basis
of U and let v1 , . . . , vn be a basis of V .
(i) [Easy] Show that
{uk ⊕ 0|k = 1, . . . , m} ∪ {0 ⊕ vk |k = 1, . . . , n}
is a basis of U ⊕ V .
(ii) [Harder] Show that
{ui ⊗ vj |i = 1, . . . , m and j = 1, . . . , n}
30
is a basis of U ⊗ V .
This exercise shows in particular that
dim(U ⊕V ) = dim(U )+dim(V ) and dim(U ⊗V ) = dim(U ) dim(V ) . (4.20)
Definition 4.12 :
Let g, h be Lie algebras over K. The direct sum g ⊕ h is the Lie algebra given
by the K-vector space g ⊕ h with Lie bracket
[x ⊕ y, x0 ⊕ y 0 ] = [x, x0 ] ⊕ [y, y 0 ] for all x, x0 ∈ g, y, y 0 ∈ h . (4.21)
Exercise 4.11 :
Show that for two Lie algebras g, h, the vector space g ⊕ h with Lie bracket as
defined in the lecture is indeed a Lie algebra.
Definition 4.13 :
Let g be a Lie algebra and let U, V be two representations of g.
(i) The direct sum of U and V is the representation of g on the vector space
U ⊕ V with action
x.(u ⊕ v) = (x.u) ⊕ (x.v) for all x ∈ g, u ∈ U, v ∈ V . (4.22)
(ii) The tensor product of U and V is the representation of g on the vector space
U ⊗ V with action
x.(u ⊗ v) = (x.u) ⊗ v + u ⊗ (x.v) for all x ∈ g, u ∈ U, v ∈ V . (4.23)
Exercise 4.12 :
(i) Show that the vector spaces U ⊕ V and U ⊗ V with g-action as defined in
the lecture are indeed representations of g.
(ii) Show that the vector space U ⊗ V with g-action x.(u ⊗ v) = (x.u) ⊗ (x.v) is
not a representation of g.
Exercise 4.13 :
Let Vn denote the irreducible representation of sl(2, C) defined in the lecture.
Consider the isomorphism of vector spaces ϕ : V1 ⊕ V3 → V2 ⊗ V2 given by
ϕ(e0 ⊕ 0) = e0 ⊗ e1 − e1 ⊗ e0 ,
ϕ(0 ⊕ e0 ) = e0 ⊗ e0 ,
ϕ(0 ⊕ e1 ) = e0 ⊗ e1 + e1 ⊗ e0 ,
ϕ(0 ⊕ e2 ) = 2e1 ⊗ e1 ,
31
(so that V1 gets mapped to anti-symmetric combinations and V3 to symmetric
combinations of basis elements of V2 ⊗ V2 ). With the help of ϕ, show that
V1 ⊕ V3 ∼
= V2 ⊗ V2
as representations of sl(2, C) (this involves a bit of writing).
4.4 Ideals
If U, V are sub-vector spaces of a Lie algebra g over K we define [U, V ] to be
the sub-vector space

[U, V ] = spanK [x, y]|x ∈ U, y ∈ V ⊂ g . (4.24)
Definition 4.14 :
(i) A sub-vector space h ⊂ g is an ideal iff [g, h] ⊂ h.
(ii) An ideal h of g is called proper iff h 6= {0} and h 6= g.
Exercise 4.14 :
(i) Show that a sub-vector space h ⊂ g is a Lie subalgebra of g iff [h, h] ⊂ h.
(ii) Show that an ideal of g is in particular a Lie subalgebra.
(iii) Show that for a Lie algebra homomorphism ϕ : g → g 0 from g to a Lie
algebra g 0 , ker(ϕ) is an ideal of g.
(iv) Show that [g, g] is an ideal of g.
(v) Show that if h and h0 are ideals of g, then their intersection h ∩ h0 is an ideal
of g.
Lemma 4.15 :
If g is a Lie algebra and h ⊂ g is an ideal, then quotient vector space g/h is a
Lie algebra with Lie bracket
[x + h, y + h] = [x, y] + h for x, y ∈ g . (4.25)
Proof:
(i) The Lie bracket is well defined: Let π : g → g/h, π(x) = x + h be the
canonical projection. For a = π(x) and b = π(y) we want to define
[a, b] = π([x, y]) . (4.26)
For this to be well defined, the rhs must only depend on a and b, but not on
the specific choice of x and y. Let thus x0 , y 0 be two elements of g such that
32
π(x0 ) = a, π(y 0 ) = b. Then there exist hx , hy ∈ h such that x0 = x + hx and
y 0 = y + hy . It follows that
π([x0 , y 0 ]) = π([x + hx , y + hy ])
(4.27)
= π([x, y]) + π([hx , y]) + π([x, hy ]) + π([hx , hy ]) .
But [hx , y], [x, hy ] and [hx , hy ] are in h since h is an ideal, and hence
0 = π([hx , y]) = π([x, hy ]) = π([hx , hy ]) . (4.28)
It follows π([x0 , y 0 ]) = π([x, y]) + 0 and hence the Lie bracket on g/h is well-
defined.
(ii) The Lie bracket is skew-symmetric, bilinear and solves the Jacobi-Identity:
Immediate from definition. E.g.
[x + h, x + h] = [x, x] + h = 0 + h . (4.29)

Exercise 4.15 :
Let g be a Lie algebra and h ⊂ g an ideal. Show that π : g → g/h given
by π(x) = x + h is a surjective homomorphism of Lie algebras with kernel
ker(π) = h.
Definition 4.16 :
A Lie algebra g is called
(i) abelian iff [g, g] = {0}.
(ii) simple iff it has no proper ideal and is not abelian.
(iii) semi-simple iff it is isomorphic to a direct sum of simple Lie algebras.
(iv) reductive iff it is isomorphic to a direct sum of simple and abelian Lie
algebras.
Lemma 4.17 :
If g is a semi-simple Lie algebra, then [g, g] = g.
Proof:
Suppose first that g is simple. We have seen in exercise 4.14(iv) that [g, g] is
an ideal of g. Since g is simple, [g, g] = {0} or [g, g] = g. But [g, g] = {0} implies
that g is abelian, which is excluded for simple Lie algebras. Thus [g, g] = g.
Suppose now that g = g1 ⊕ · · · ⊕ gn with all gk simple Lie algebras. Then

[g, g] = spanK [gk , gl ]|k, l = 1, . . . , n = spanK [gk , gk ]|k = 1, . . . , n
(4.30)
= spanK gk |k = 1, . . . , n = g
where we first used that [gk , gl ] = {0} for k 6= l and then that [gk , gk ] = gk since
gk is simple.
33
Exercise 4.16 :
Let g, h be Lie algebras and ϕ : g → h a Lie algebra homomorphism. Show that
if g is simple, then ϕ is either zero or injective.
4.5 The Killing form
Definition 4.18 :
Let g be a finite-dimensional Lie algebra over K. The Killing form κ ≡ κg on g
is the bilinear map κ : g × g → K given by
κ(x, y) = tr(adx ◦ ady ) for x, y ∈ g . (4.31)
Lemma 4.19 :
The Killing form obeys, for all x, y, z ∈ g,
(i) κ(x, y) = κ(y, x) (symmetry)
(ii) κ([x, y], z) = κ(x, [y, z]) (invariance)
Proof:
(i) By cyclicity of the trace we have
κ(x, y) = tr(adx ◦ ady ) = tr(ady ◦ adx ) = κ(y, x) . (4.32)
(ii) From the properties of the adjoint action and the cyclicity of the trace we
get
κ([x, y], z) = tr(ad[x,y] adz ) = tr(adx ady adz − ady adx adz )
(4.33)
= tr(adx ady adz − adx adz ady ) = κ(x, [y, z]) .
Exercise 4.17 :
(i) Show that for the basis of sl(2, C) used in exercise 4.5, one has
κ(E, E) = 0 , κ(E, H) = 0 , κ(E, F ) = 4 ,
κ(H, H) = 8 , κ(H, F ) = 0 , κ(F, F ) = 0 .
Denote by Tr the trace of 2×2-matrices. Show that for sl(2, C) one has κ(x, y) =
4 Tr(xy).
(ii) Evaluate the Killing form of p(1, 1) for all combinations of the basis elements
M01 , P0 , P1 (as used in exercises 3.14 and 3.16). Is the Killing form of p(1, 1)
non-degenerate?
34
Exercise 4.18 :
(i) Show that for gl(n, C) one has κ(x, y) = 2n Tr(xy) − 2Tr(x)Tr(y), where Tr
is the trace of n×n-matrices.
Hint: Use the basis Ekl to compute the trace in the adjoint representation.
(ii) Show that for sl(n, C) one has κ(x, y) = 2n Tr(xy).
Exercise 4.19 :
Let g be a finite-dimensional Lie algebra and let h ⊂ g be an ideal. Show that
h⊥ = {x ∈ g|κg (x, y) = 0 for all y ∈ h}
is also an ideal of g.
The following theorem we will not prove.
Theorem 4.20 :
If g is a finite-dimensional complex simple Lie algebra, then κg is non-degenerate.
Information 4.21 :
The proof of this (and the necessary background) needs about 10 pages, and
can be found e.g. in [Fulton, Harris “Representation Theory” Part II Ch. 9 and
App. C Prop. C.10]. It works along the following lines. One defines
g {0} = g , g {1} = [g {0} , g {0} ] , g {2} = [g {1} , g {1} ] , ... (4.34)
and calls a Lie algebra solvable if g {m} = {0} for some m. The hard part then
is to prove Cartan’s criterion for solvability, which implies that if a complex,
finite-dimensional Lie algebra g has κg = 0, then g is solvable. Suppose now
that g is simple. Then [g, g] = g, and hence g is not solvable (as g {m} = g for
all m). Hence κg does not vanish. But the set
g ⊥ = {x ∈ g|κg (x, y) = 0 for all y ∈ g} (4.35)
is an ideal (see exercise 4.19). Hence it is {0} or g. But g ⊥ = g implies κg = 0,

which cannot be for g simple. Thus g ⊥ = {0}, which precisely means that κg is
non-degenerate.
Lemma 4.22 :
Let g be a finite-dimensional Lie algebra. If g contains an abelian ideal h (i.e.
[h, g] ⊂ h and [h, h] = 0), then κg is degenerate.
Exercise 4.20 :
Show that if a finite-dimensional Lie algebra g contains an abelian ideal h, then
the Killing form of g is degenerate. (Hint: Choose a basis of h, extend it to a
basis of g, and evaluate κg (x, a) with x ∈ g, a ∈ h.)
35
Exercise 4.21 :
Let g = g1 ⊕· · ·⊕gn , for finite-dimensional Lie algebras gi . Let x = x1 +· · ·+xn
and y = y1 + · · · + yn be elements of g such that xi , yi ∈ gi . Show that
n
X
κg (x, y) = κgi (xi , yi ) .
i=1
Theorem 4.23 :
For a finite-dimensional, complex Lie algebra g, the following are equivalent.
(i) g is semi-simple.
(ii) κg is non-degenerate.
Proof:
(i) ⇒ (ii): We can write
g = g1 ⊕ · · · ⊕ gn (4.36)
for gk simple Lie algebras. If x, y ∈ gk , then κg (x, y) = κgk (x, y), while if x ∈ gk
and y ∈ gl with k 6= l, we have κg (x, y) = 0. Let x = x1 + · · · + xn 6= 0 be an
element of g, with xk ∈ gk . There is at least one xl 6= 0. Since gl is simple, κgl
is non-degenerate, and there is a y ∈ gl such that κgl (xl , y) 6= 0. But
κg (x, y) = κgl (xl , y) 6= 0 . (4.37)
Hence κg is non-degenerate.
(ii) ⇒ (i):
g is not abelian (or by lemma 4.22 κg would be degenerate). If g does not
contain a proper ideal, then it is therefore simple and in particular semi-simple.
Suppose now that h ⊂ g is a proper ideal and set X = h ∩ h⊥ . Then X is
an ideal. Further, κ(a, b) = 0 for all a ∈ h and b ∈ h⊥ , so that in particular
κ(a, b) = 0 for all a, b ∈ X. But then, for all a, b ∈ X and for all x ∈ g,
κ(x, [a, b]) = κ([x, a], b) = 0 (since [x, a] ∈ X as X is an ideal). But κ is
non-degenerate, so that this is only possible if [a, b] = 0. It follows that X
is an abelian ideal. By the previous lemma, then X = {0} (or κ would be
degenerate).
In exercise 4.22 you will prove that, since κg is non-degenerate, dim(h) +
dim(h⊥ ) = dim(g). Since [h, h⊥ ] = {0} and h∩h⊥ = {0}, we have g = h⊕h⊥ as
Lie algebras. Apply the above argument to h and h⊥ until all summands contain
no proper ideals. Since g is finite-dimensional, this process will terminate.
Exercise 4.22 :
Let g be a finite-dimensional Lie algebra with non-degenerate Killing form. Let
h ⊂ g be a sub-vector space. Show that dim(h) + dim(h⊥ ) = dim(g).
Exercise 4.23 :
Show that the Poincaré algebra p(1, n−1), n ≥ 2, is not semi-simple.
36
Definition 4.24 :
Let g be a Lie algebra over K. A bilinear form B : g × g → K is called invariant
iff B([x, y], z) = B(x, [y, z]) for all x, y, z ∈ g.
Clearly, the Killing form is an invariant bilinear form on g, which is in
addition symmetric. The following theorem shows that for a simple Lie algebra,
it is unique up to a constant.
Theorem 4.25 :
Let g be a finite-dimensional, complex, simple Lie algebra and let B be an
invariant bilinear form. Then B = λκg for some λ ∈ C.
The proof will be given in the following exercise.
Exercise 4.24 :
In this exercise we prove the theorem that for a finite-dimensional, complex,
simple Lie algebra g, and for an invariant bilinear form B, we have B = λκg for
some λ ∈ C.
(i) Let g ∗ = {ϕ : g → C linear} be the dual space of g. The dual representation
of the adjoint representation is (g, ad)+ = (g ∗ , −ad). Let fB : g → g ∗ be given
by fB (x) = B(x, ·), i.e. [fB (x)](z) = B(x, z). Show that fB is an intertwiner
from (g, ad) to (g ∗ , −ad).
(ii) Using that g is simple, show that (g, ad) is irreducible.
(iii) Since (g, ad) and (g ∗ , −ad) are isomorphic representations, also (g ∗ , −ad)
is irreducible. Let fκ be defined in the same way as fB , but with κ instead of
B. Show that fB = λfκ for some λ ∈ C.
(iv) Show that B = λκ for some λ ∈ C.
5 Classification of finite-dimensional, semi-simple,

complex Lie algebras
In this section we will almost exclusively work with finite-dimensional semi-
simple complex Lie algebras. In order not to say that too often we abbreviate
fssc = finite-dimensional semi-simple complex .
5.1 Working in a basis

Let g be a finite-dimensional Lie algebra over K. Let {T a |a = 1, . . . , dim(g)} be
a basis of g. Then we can write
X
[T a , T b ] = f abc T c , f abc ∈ K . (5.1)
c
The constants f abc

are called structure constants of the Lie algebra g. (If g is
infinite-dimensional, and we are given a basis, we will also call the f abc structure
constants.)
37
Exercise 5.1 :
Let {T a } be a basis of a finite-dimensional Lie algebra g over K. For x ∈ g, let
M (x)ab be the matrix of adx in that basis, i.e.
X XX
adx ( vb T b ) = ( M (x)ab vb )T a .
b a b
Show that M (T a )cb = f abc , i.e. the structure constants give the matrix elements
of the adjoint action.
Exercise* 5.2 :
A fact from linear algebra: Show that for every non-degenerate symmetric bilin-
ear form b : V × V → C on a finite-dimensional, complex vector space V there
exists a basis v1 , . . . , vn (with n = dim(V )) of V such that b(vi , vj ) = δij .
If g is a fssc Lie algebra, we can hence find a basis {T a |a = 1, . . . , dim(g)}
such that
κ(T a , T b ) = δab . (5.2)
In this basis the structure constants can be computed to be
X
κ(T c , [T a , T b ]) = f abd κ(T c , T d ) = f abc . (5.3)
d
Exercise 5.3 :
Let g be a fssc Lie algebra and {T a } a basis such that κ(T a , T b ) = δab . Show
that the structure constants in this basis are anti-symmetric in all three indices.
Exercise 5.4 :
Find a basis {T a } of sl(2, C) s.t. κ(T a , T b ) = δab .
5.2 Cartan subalgebras
Definition 5.1 :
An element x of a complex Lie algebra g is called ad-diagonalisable iff adx : g → g
is diagonalisable, i.e. iff there exists a basis T a of g such that [x, T a ] = λa T a ,
λa ∈ C for all a.
Lemma 5.2 :
Let g be a fssc Lie algebra g.
(i) Any x ∈ g with κ(x, x) 6= 0 is ad-diagonalisable.
(ii) g contains at least one ad-diagonalisable element.
Proof:
(i) Let n = dim(g). The solution to exercise 5.2 shows that we can find a basis
38
{T a | a = 1, . . . , n} such that κ(T a , T b ) = δab and such that x = λT 1 for some
λ ∈ C× . From exercise 5.1 we know that Mba ≡ M (T 1 )ba = f 1ab are the matrix
elements of adT 1 in the basis {T a }. Since f is totally antisymmetric (see exercise
5.3), we have
Mba = f 1ab = −f 1ba = −Mab , (5.4)
i.e. M t = −M . In particular, [M t , M ] = 0, so that M is normal and can be
diagonalised. Thus T 1 is ad-diagonalisable, and with it also x.
(ii) Exercise 5.2 also shows that (since κ is symmetric and non-degenerate) one
can always find an x ∈ g with κ(x, x) 6= 0.
Definition 5.3 :
A sub-vector space h of a fssc Lie algebra g is a Cartan subalgebra iff it obeys
the three properties
(i) all x ∈ h are ad-diagonalisable.
(ii) h is abelian.
(iii) h is maximal in the sense that if h0 obeys (i) and (ii) and h ⊂ h0 , then
already h = h0 .
Exercise 5.5 :
Show that the diagonal matrices in sl(n, C) are a Cartan subalgebra.
The dimension r = dim(h) of a Cartan subalgebra is called the rank of g. By
lemma 5.2, r ≥ 1. It turns out (but we will not prove it in this course, but
see [Fulton,Harris] § D.3) that r is independent of the choice of h and hence the
rank is indeed a property of g.
Let H 1 , . . . , H r be a basis of h. By assumption, adH i can be diagonalised for
each i. Further adH i and adH j commute for any i, j ∈ {1, . . . , r},
[adH i , adH j ] = ad[H i ,H j ] = 0 . (5.5)
Thus, all adH i can be simultaneously diagonalised.

Let y ∈ g be a simultaneous eigenvector for all H ∈ h,
adH (y) = αy (H)y , for some αy (H) ∈ C . (5.6)
The αy (H) depend linearly on H. Thus we obtain a function
αy : h → C , (5.7)
i.e. αy ∈ h∗ , the dual space of h. Conversely, given an element ϕ ∈ h∗ we set
gϕ = {x ∈ g|[H, x] = ϕ(H)x for all H ∈ h} . (5.8)
39
Definition 5.4 :
Let g be a fssc Lie algebra and h a Cartan subalgebra of g.
(i) α ∈ h∗ is called a root of g (with respect to h) iff α 6= 0 and gα 6= {0}.
(ii) The root system of g is the set
Φ ≡ Φ(g, h) = {α ∈ h∗ |α is a root} . (5.9)
Decomposing g into simultaneous eigenspaces of elements of h we can write
M
g = g0 ⊕ gα . (5.10)
α∈Φ
(This is a direct sum of vector spaces only, not of Lie algebras.)
Lemma 5.5 :
(i) [gα , gβ ] ⊂ gα+β for all α, β ∈ h∗ .
(ii) If x ∈ gα , y ∈ gβ for some α, β ∈ h∗ s.t. α + β 6= 0, then κ(x, y) = 0.
(iii) κ restricted to g0 is non-degenerate.
Proof:
(i) Have, for all H ∈ h, x ∈ gα , y ∈ gβ ,
(1)
adH ([x, y]) = [H, [x, y]] = −[x, [y, H]] − [y, [H, x]]
(5.11)
= β(H)[x, y] − α(H)[y, x] = (α + β)(H) [x, y]
where (1) is the Jacobi identity. Thus [x, y] ∈ gα+β .
(ii) Let H ∈ h be such that α(H) + β(H) 6= 0 (H exists since α + β 6= 0). Then
(α(H) + β(H))κ(x, y) = κ(α(H)x, y) + κ(x, β(H)y)
(1)
= κ([H, x], y) + κ(x, [H, y]) = −κ([x, H], y) + κ(x, [H, y]) (5.12)
(2)
= −κ(x, [H, y]) + κ(x, [H, y]) = 0
where (1) uses that x ∈ gα and y ∈ gβ , and (2) that κ is invariant. Thus
κ(x, y) = 0.
(iii) Let y ∈ g0 . Since κ is non-degenerate, there is an x ∈ g s.t. κ(x, y) 6= 0.
Write X
x = x0 + xα where x0 ∈ g0 , xα ∈ gα . (5.13)
α∈Φ
Then by part (ii), κ(x, y) = κ(x0 , y). Thus for all y ∈ g0 we can find an x0 ∈ g0
s.t. κ(x0 , y) 6= 0.
Exercise 5.6 :
Another fact about linear algebra: Let V be a finite-dimensional vector space
and let F ⊂ V ∗ be a proper subspace (i.e. F 6= V ∗ ). Show that there exists a
nonzero v ∈ V such that ϕ(v) = 0 for all ϕ ∈ F .
40
Lemma 5.6 :
Let g be a fssc Lie algebra and h a Cartan subalgebra. Then
(i) the Killing form restricted to h is non-degenerate.
(ii) g0 = h.
(iii) g0∗ = spanC (Φ).
Proof:
(i) Since for all a, b ∈ h, ada (b) = [a, b] = 0 we have h ⊂ g0 . Suppose there is
an a ∈ h such that κ(a, b) = 0 for all b ∈ h. Then in particular κ(a, a) = 0. As
κ is non-degenerate on g0 , there is a z ∈ g0 , z ∈ / h, such that κ(a, z) 6= 0. If
κ(z, z) 6= 0 set u = z. Otherwise set u = a + z (then κ(u, u) = κ(a + z, a + z) =
2κ(a, z) 6= 0). In either case u ∈ / h and κ(u, u) 6= 0. By lemma 5.2, u is ad-
diagonalisable. Also [b, u] = 0 for all b ∈ h (since u ∈ g0 ). But then spanC (h, u)
obeys conditions (i),(ii) in the definition of a Cartan subalgebra and contains h
as a proper subspace, which is a contradiction to h being a Cartan subalgebra.
Hence κ has to be non-degenerate on h.
(ii) By part (i) we have subspaces
h ⊂ g0 ⊂ g (5.14)
and κ is non-degenerate on h, g0 , g. It is therefore possible to find a basis {T a }

of g s.t.
κ(T a , T b ) = δab
T a ∈ h for a = 1, . . . , dim(h) and T a ∈ g0 for a = 1, . . . , dim(g0 ).
Let X = T a with a = dim(g0 ). We have [H, X] = 0 for all H ∈ h (since X ∈ g0 ).
Further, X is ad-diagonalisable (since K(X, X) 6= 0, see lemma 5.2 (i)). Thus
the space spanC (h, X) obeys (i) and (ii) in the definition of a Cartan subalgebra,
and hence by maximality of h we have h = spanC (h, X). Thus X ∈ h and hence
dim(g0 ) = dim(h).
(iii) Suppose that spanC (Φ) is a proper subspace of g0∗ . By exercise 5.6 there
exists a nonzero element H ∈ g0 s.t.
α(H) = 0 for all α ∈ Φ . (5.15)
Since g0 = h we have, for all α ∈ Φ and all x ∈ gα , [H, x] = α(H)x = 0.

Thus [H, x] = 0 for all x ∈ g. But then adH = 0, in contradiction to κ being
non-degenerate (have κ(y, H) = 0 for all y ∈ g).
Exercise 5.7 :
Let g be a fssc Lie algebra and let h ⊂ g be sub-vector space such that
(1) [h, h] = {0}.
(2) κ restricted to h is non-degenerate.
(3) if for some x ∈ g one has [x, a] = 0 for all a ∈ h, then already x ∈ h.
Show that h is a Cartan subalgebra of g if and only if it obeys (1)–(3) above.
41
5.3 Cartan-Weyl basis
Definition 5.7 :
(i) For ϕ ∈ g0∗ let H ϕ ∈ g0 be the unique element s.t.
ϕ(x) = κ(H ϕ , x) for all x ∈ g0 . (5.16)
(ii) Define the non-degenerate pairing (·, ·) : g0∗ × g0∗ → C via
(γ, ϕ) = κ(H γ , H ϕ ) . (5.17)
Information 5.8 :
We will see shortly that (α, α) > 0 for all α ∈ Φ. Since Φ is a finite set, there
is a θ ∈ Φ such that (θ, θ) is maximal. Some texts (such as [Fuchs,Schweigert]
√ ( · , · ). This is
Sect. 6.3) use a a rescaled version of the Killing form κ to define
done to impose the convention that the longest root lengths is 2, i.e. (θ, θ) = 2,
which leads to simpler expressions in explicit calculations. But it also makes the
exposition less clear, so we will stick to κ (as also done e.g. in [Fulton,Harris]
§ 14.2.)
Exercise 5.8 :
Let {H 1 , . . . , H r } ⊂ g0 be a basis of g0 such that κ(H i , H j ) = δij P
(recall that r =
r
dim(g0 ) is the rankP of g). Show that for γ, ϕ ∈ g0∗ one has H γ = i=1 γ(H i )H i ,
r i i ϕ
as well as (γ, ϕ) = i=1 γ(H )ϕ(H ) and (γ, ϕ) = γ(H ).
Lemma 5.9 :
Let α ∈ Φ. Then
(i) −α ∈ Φ.
(ii) If x ∈ gα and y ∈ g−α then [x, y] = κ(x, y)H α .
(iii) (α, α) 6= 0.
Proof:
(i) For x ∈ gα , the Killing form κ(x, y) can be nonzero only for y ∈ g−α (lemma
5.5(ii)). Since κ is non-degenerate, g−α cannot be empty, and hence −α ∈ Φ.
(ii) Since [gα , g−α ] ⊂ g0 we have [x, y] ∈ g0 . Note that for all H ∈ g0 ,
κ(H, [x, y]) = κ([H, x], y) = α(H)κ(x, y) = κ(H α , H)κ(x, y)

(5.18)
= κ(H, κ(x, y)H α ) .
Since κ is non-degenerate, this implies [x, y] = κ(x, y)H α .

(iii) By exercise 5.8 we have (α, α) = α(H α ). We will show that α(H α ) 6= 0.
42
Since α 6= 0 have H α 6= 0. Since g0∗ = spanC (Φ), there exists a β ∈ Φ s.t.
β(H α ) 6= 0. Consider the subspace
M
U= gβ+mα . (5.19)
m∈Z
For x ∈ gβ+mα have [H , x] = (β(H α ) + mα(H α ))x so that the trace of adH α
α
over U is
X
trU (adH α ) = (β(H α ) + mα(H α )) dim(gβ+mα ) . (5.20)
m∈Z
Choose a nonzero x ∈ gα . There is a y ∈ g−α s.t. κ(x, y) 6= 0; we can

choose y such that κ(x, y) = 1. Then [x, y] = H α . Since adx : gγ → gγ+α and
ady : gγ → gγ−α , both, adx and ady map U to U . Then we can also compute
trU (adH α ) = trU (ad[x,y] ) = trU (adx ady − ady adx ) = 0 , (5.21)
by cyclicity of the trace.
Together with the previous expression for trU (adH α ) this implies
X X
α(H α ) m dim(gβ+mα ) = −β(H α ) dim(gβ+mα ) (5.22)
m∈Z m∈Z
The rhs is nonzero (as β(H α ) 6= 0 by construction, and dim(gβ ) 6= 0 since β is

a root), and hence the lhs as to be nonzero. In particular, α(H α ) 6= 0.
Recall the standard basis E, F, H of sl(2, C) we introduced in section 4.2.
Theorem and Exercise 5.9 :

Let g be a fssc Lie algebra and g0 a Cartan subalgebra. Let α ∈ Φ(g, g0 ). Choose
2
e ∈ gα and f ∈ g−α such that κ(e, f ) = (α,α) . Show that ϕ : sl(2, C) → g, given
by
2
ϕ(E) = e , ϕ(F ) = f , ϕ(H) = Hα ,
(α, α)
is an injective homomorphism of Lie algebras.
This implies in particular that g can be turned into a finite-dimensional
representation (g, Rϕ ) of sl(2, C) via
Rϕ (x)z = adϕ(x) z for all x ∈ sl(2, C) , z ∈ g , (5.23)
i.e. by restricting the adjoint representation of g to sl(2, C). For z ∈ gβ we find
2 2 2(α, β)
Rϕ (H)z = [H α , z] = β(H α )z = z . (5.24)
(α, α) (α, α) (α, α)
From corollary 4.6 we know that in a finite-dimensional representation of sl(2, C),
all eigenvalues of Rϕ (H) have to be integers. Thus
2(α, β)
∈Z for all α, β ∈ Φ . (5.25)
(α, α)
43
Theorem 5.10 :
Let g be a fssc Lie algebra and g0 a Cartan subalgebra. Then
(i) if α ∈ Φ and λα ∈ Φ for some λ ∈ C, then λ ∈ {±1}.
(ii) dim(gα ) = 1 for all α ∈ Φ.
The proof will be given in the following exercise.
Exercise* 5.10 :
In this exercise we will show that dim(gα ) = 1 for all α ∈ Φ. On the way we
will also see that if α ∈ Φ and λα ∈ Φ for some λ ∈ C, then λ ∈ {±1}.
(i) Choose α ∈ Φ. Let L = {m ∈ Z|mα ∈ Φ}. Since Φ is a finite set, so is
L. Let n+ be the largest integer in L, n− the smallest integer in L. Show that
n+ ≥ 1 and n− ≤ −1.
(ii) We can assume that n+ ≥ |n− |. Otherwise we exchange α for −α. Pick
2
e ∈ gα , f ∈ g−α s.t. κ(e, f ) = (α,α) and define ϕ : sl(2, C) → g as in exercise
5.9. Show that M
U = CH α ⊕ gmα
m∈L
is an invariant subspace of the representation (g, Rϕ ) of sl(2, C).

(iii) Show that for z ∈ gmα one has Rϕ (H)z = 2mz.
(iv) By (ii), (U, Rϕ ) is also a representation of sl(2, C). Show that V = Ce ⊕
CH α ⊕ Cf is an invariant subspace of (U, Rϕ ). Show that the representation
(V, Rϕ ) is isomorphic to the irreducible representation V3 of sl(2, C).
(v) Choose an element v0 ∈ gn+ α . Set vk+1 = Rϕ (F )vk and show that
W = spanC (v0 , v1 , . . . , v2n+ )
is an invariant subspace of U .
(vi) (W, Rϕ ) is isomorphic to the irreducible representation V2n+ +1 of sl(2, C).
Show that the intersection X = V ∩ W is an invariant subspace of V and W .
Show that X contains the element H α and hence X 6= {0}. Show that X = V
and X = W .
We have learned that for any choice of v0 in gn+ α we have V = W . This can
only be if n+ = 1 and dim(gα ) = 1. Since 1 ≤ |n− | ≤ n+ , also n− = 1. Since
κ : gα × g−α is non-degenerate, also dim(g−α ) = 1.
Definition 5.11 :
Let g be a fssc Lie algebra. A subset
{H i , i = 1, . . . , r} ∪ {E α |α ∈ Φ}
∗
of g, for Φ a finite subset of spanC (H 1 , . . . , H r ) −{0}, is called a Cartan-Weyl
basis of g iff it is a basis of g, and [H i , H j ] = 0, [H i , E α ] = α(H i ) E α ,

0
 ; α + β 6∈ Φ
α β α+β
[E , E ] = Nα,β E ;α + β ∈ Φ
 2 Hα

; α = −β
(α,α)
44
where Nα,β ∈ C are some constants.
In the above definition, it is understood that (α, α) and H α are computed
with respect to the Cartan subalgebra h = spanC (H 1 , . . . , H r ) of g. That h is
indeed a Cartan subalgebra is the topic of the next exercise.
Exercise 5.11 :
Let {H i } ∪ {E α } be a Cartan-Weyl basis of a fssc Lie algebra g. Show that
(i) spanC (H 1 , . . . , H r ) is a Cartan subalgebra of g.
(ii) κ(E α , E −α ) = (α,α) 2
.
The analysis up to now implies the following theorem.
Theorem 5.12 :
For any fssc Lie algebra g, there exists a Cartan-Weyl basis.
Lemma 5.13 P :
(i) κ(G, H) = P α∈Φ α(G)α(H) for all G, H ∈ g0 .
(ii) (λ, µ) = α∈Φ (λ, α)(α, µ) for all λ, µ ∈ g0∗ .
(iii) (α, β) ∈ R and (α, α) > 0 for all α, β ∈ Φ.
Proof:
(i) Let {H i } ∪ {E α } be a Cartan-Weyl basis of g. One computes
r
X ∗ X α∗ X
κ(G, H) = H i [G, [H, H i ]] + E [G, [H, E α ]] = α(H)α(G) .
i=1 α∈Φ α∈Φ
(5.26)
(ii) Using part (i) we get
X X
(λ, µ) = κ(H λ , H µ ) = α(H λ )α(H µ ) = (α, µ)(α, λ) (5.27)
α∈Φ α∈Φ
(iii) Using part (ii) we compute

X
(α, α) = (α, β)(α, β) . (5.28)
β∈Φ
Multiplying both sides by 4/(α, α)2 yields
4 X 2(α, β) 2
= . (5.29)
(α, α) (α, α)
β∈Φ
We have already seen that 2(α,β)

(α,α) ∈ Z. Thus the rhs is real and non-negative.
Since (α, α) 6= 0 (see lemma 5.9 (iii)) it follows that (α, α) > 0. Together with
2(α,β)
(α,α) ∈ Z this in turn implies that (α, β) ∈ R.
45
Exercise 5.12 :
Let g = g1 ⊕ g2 with g1 , g2 fssc Lie algebras. For k = 1, 2, let hk be a Cartan
subalgebra of gk .
(i) Show that h = h1 ⊕ h2 is a Cartan subalgebra of g.
(ii) Show that the root system of g is Φ(g, h) = Φ1 ∪ Φ2 ⊂ h∗1 ⊕ h∗2 where
Φ1 = {α ⊕ 0|α ∈ Φ(g1 , h1 )} and Φ2 = {0 ⊕ β|β ∈ Φ(g2 , h2 )}.
(iii) Show that (α, β) = 0 for all α ∈ Φ1 and β ∈ Φ2 .
Lemma 5.14 :
Let g be a fssc Lie algebra and g0 be a Cartan subalgebra. The following are
equivalent.
(i) g is simple.
(ii) One cannot write Φ(g, g0 ) = Φ1 ∪ Φ2 where Φ1 , Φ2 are non-empty and
(α, β) = 0 for all α ∈ Φ1 , β ∈ Φ2 .
Proof:
¬(i)⇒ ¬(ii): This amounts to exercise 5.12.
¬(ii)⇒ ¬(i): Let Φ ≡ Φ(g, g0 ) = Φ1 ∪ Φ2 with the properties stated in (ii).
If α ∈ Φ1 , β ∈ Φ2 then α + β ∈/ Φ2 , since
(α, α + β) = (α, α) 6= 0 . (5.30)
Similarly, since (β, α + β) 6= 0 we have α + β ∈

/ Φ1 . Thus α + β ∈
/ Φ.
If α, β ∈ Φ1 , α 6= −β, then
0 6= (α + β, α + β) = (α + β, α) + (α + β, β) (5.31)
so that α + β ∈/ Φ2 .
Let {H i } ∪ {E α } be a Cartan-Weyl basis of g. Let h1 = spanC (H α |α ∈ Φ1 }
and M
g1 = h1 ⊕ CE α . (5.32)
α∈Φ1
Claim: g1 is a proper ideal of g. The proof of this claim is the subject of the
next exercise. Thus g has a proper ideal and hence is not simple.
Exercise 5.13 :
Let g be a fssc Lie algebra and g0 be a Cartan subalgebra. Suppose Φ(g, g0 ) =
Φ1 ∪ Φ2 where Φ1 , Φ2 are non-empty and (α, β) = 0 for all α ∈ Φ1 , β ∈ Φ2 . Let
{H i } ∪ {E α } be a Cartan-Weyl basis of g. Show that
M
g1 = spanC (H α |α ∈ Φ1 } ⊕ CE α
α∈Φ1
is a proper ideal in g.
The following theorem we will not prove. (For a proof see [Fulton,Harris]
§ 21.3.)
46
Theorem 5.15 :
Let g, g 0 be fssc Lie algebras with Cartan subalgebras g0 ⊂ g and g00 ⊂ g 0 . Then
∗
g and g 0 are isomorphic iff there is an isomorphism g0 ∗ → g00 of vector spaces
0 0
that preserves (·, ·) and maps Φ(g, g0 ) to Φ(g , g0 ).
Definition 5.16 :
Let g be a fssc Lie algebra g with Cartan subalgebra g0 . The root space is the
real span
R ≡ R(g, g0 ) = spanR (Φ(g, g0 )) . (5.33)
[The term root spaces is also used for the spaces gα , so one has to be a bit
careful.]
In particular, R is a real vector space.
Exercise 5.14 :
Let R the root space of a fssc Lie algebra with Cartan subalgebra g0 .
(i) Show that the bilinear form (·, ·) on g0∗ restricts to a real valued positive
definite inner product on R.
(ii) Use the Gram-Schmidt procedure to find an orthonormal basis {ε1 , . . . , εm }
of R (over R). Show that m = r (where r = dim(g0 ) is the rank of g) and that
{ε1 , . . . , εr } is a basis of g0∗ (over C).
(iii) Show that there exists a basis {H i |i = 1, . . . , r} of g0 such that α(H i ) ∈ R
for all i = 1, . . . , r and α ∈ Φ.
The basis {ε1 , . . . , εr } provides an identification of R and Rr , whereby
Pr the
inner product ( · , · ) on R becomes the usual inner product g(x, y) = i=1 xi yi
on Rr . [In other words, R and Rr are isomorphic as inner product spaces.]
5.4 Examples: sl(2, C) and sl(n, C)

Recall the basis
H = E11 − E22 , E = E12 , F = E21 (5.34)
of sl(2, C), with Lie brackets
[E, F ] = H , [H, E] = 2E , [H, F ] = −2F . (5.35)
Define the linear forms ω1 and ω2 on diagonal 2×2-matrices as
ωi (Ejj ) = δij . (5.36)
Fix the Cartan subalgebra h = CH. Define α ≡ α12 = ω1 − ω2 ∈ h∗ . Then α is

a root since
α(H) = (ω1 −ω2 )(E11 −E22 ) = 2 and [H, E] = 2E = α(H)E . (5.37)
47
Let us now work out H α . After a bit of staring one makes the ansatz
1
H α = E11 − E22 .

(5.38)
4
As h is one-dimensional, to verify this it is enough to check that κ(H α , H) =
α(H). Recall from exercise 4.17 that κ(x, y) = 4Tr(xy). Thus
κ(H α , H) = 4 Tr(H α H) = Tr (E11 − E22 )(E11 − E22 ) = 2 = α(H) . (5.39)

In the same way one gets (α, α) = κ(H α , H α ) = 21 . Finally, note that
2
[E, F ] = H = Hα . (5.40)
(α, α)
Altogether this shows that {H, E α ≡ E, E −α ≡ F } already is a Cartan-Weyl
basis of sl(2, C). We can draw the following picture,
−α α
Such a picture is called a root diagram.√The real axis is identified with the root
space R, and the root α has length 1/ 2.
Exercise 5.15 :
In this exercise we construct a Cartan-Weyl basis for sl(n, C). As Cartan sub-
algebra h we take the trace-less diagonal matrices.
(i) Define the linear forms ωi , i = 1, . . . , n on diagonal n×n-matrices as ωi (Ejj ) =
δij . Define αkl = ωk − ωl . Show that for k 6= l, P αkl is a root.
n Pn
Hint: Write a general element H ∈ h as H = k=1 ak Ekk with k=1 ak = 0.
Show that [H, Ekl ] = αkl (H)Ekl .
1
(ii) Show that H αkl = 2n (Ekk − Ell ).
Hint: Use exercise 4.18 to verify κ(H αkl , H) = αkl (H) for all H ∈ h.
(iii) Show that (αkl , αkl ) = 1/n and that [Ekl , Elk ] = 2/(αkl , αkl ) · H αkl
(iv) Show that, with Φ = {αkl | k, l = 1, . . . , n , k 6= l} and E αkl = Ekl ,
α
H k,k+1 k = 1, . . . , n−1 ∪ E α α ∈ Φ

is a Cartan-Weyl basis of sl(n, C).

(v) Show that the root diagram of sl(3, C) is
α23 α13
α12
√
where each arrow has length 1/ 3 and the angle between the arrows is 60◦ .
48
5.5 The Weyl group
Let g be a fssc Lie algebra with Cartan subalgebra g0 . For each α ∈ Φ(g, g0 ),
define a linear map
(α, λ)
sα : g0∗ −→ g0∗ , sα (λ) = λ − 2 α . (5.41)
(α, α)
The sα , α ∈ Φ are called Weyl reflections.
Exercise 5.16 :
Let sα be a Weyl reflection. Show that
sα (α) = −α , (α, λ) = 0 ⇒ sα (λ) = λ , sα ◦ sα = id , s−α = sα .
Thus sα is indeed a reflection, we have the following picture,

α
sα (λ)
And in sl(3, C) we find
α2 α3
α1
So in this example, it seems that roots get mapped to roots under Weyl reflec-
tions. This is true in general.
Theorem 5.17 :
Let g be a fssc Lie algebra with Cartan subalgebra g0 . If α, β ∈ Φ, then also
sα (β) ∈ Φ.
Proof:
Let α ∈ Φ. Recall the injective homomorphism of Lie algebras ϕ ≡ ϕα :
sl(2, C) → g,
ϕ(H) = 2
(α,α) H
α
, ϕ(E) = E α , ϕ(F ) = E −α . (5.42)
49
This turns g into a representation (g, Rϕ ) of sl(2, C). For a root β ∈ Φ have
2(α,β) β
Rϕ (H)E β = 2 α β
(α,α) [H , E ] = (α,α) E = mE β (5.43)
for some integer m. We may assume m ≥ 0 (otherwise replace α → −α and

start again). The representation theory of sl(2, C) tells us that then also −m
has to be an eigenvalue of Rϕ (H) with eigenvector
v = (Rϕ (F ))m E β 6= 0 . (5.44)
But
v = (Rϕ (F ))m E β = [E −α , [. . . , [E −α , E β ] . . . ] ∈ gβ−mα . (5.45)
Since v 6= 0 have gβ−mα 6= {0} so that β − mα ∈ Φ.

Now evaluate the Weyl reflection
(α, β)
sα (β) = β − 2 α = β − mα , (5.46)
(α, α)
which we have shown to be a root. Thus α, β ∈ Φ implies that sα (β) ∈ Φ.
Definition 5.18 :
Let g be a fssc Lie algebra with Cartan subalgebra g0 . The Weyl group W of g
is the subgroup of GL(g0∗ ) generated by the Weyl reflections,
W = {sβ1 · · · sβm | β1 , . . . , βm ∈ Φ, m = 0, 1, 2, . . . } . (5.47)
Exercise 5.17 :
Show that the Weyl group of a fssc Lie algebra is a finite group (i.e. contains
only a finite number of elements).
5.6 Simple Lie algebras of rank 1 and 2

Let g be a fssc Lie algebra and let R = spanR (Φ) be the root space. Recall that
on R, (·, ·) is a positive definite inner product.
Suppose g has rank 1, i.e. R = spanR (Φ) is one-dimensional. By theorem
5.10, if α ∈ Φ and λα ∈ Φ, then λ ∈ {±1}. Hence the root system has to be
−α α
This is the root system of sl(2, C). Thus by theorem 5.15 any fssc Lie algebra
of rank 1 is isomorphic to sl(2, C). This Lie algebra is also called A1 .
To proceed we need to have a closer look at the inner product of roots. By
the Cauchy-Schwartz inequality,
for all u, v ∈ R , (u, v)2 ≤ (u, u)(v, v) . (5.48)
50
Also, (u, v)2 = (u, u)(v, v) iff u and v are colinear.
For two roots α, β ∈ Φ, β 6= ±α, this means (α, β)2 < (α, α)(β, β), i.e.
(α,β)
p·q <4 where p = 2 (α,α) ∈ Z and q = 2 (α,β)
(β,β) ∈ Z . (5.49)
The angle between two roots α and β is
(α, β)2 pq
cos(θ)2 = = . (5.50)
(α, α)(β, β) 4
If (α, β) 6= 0 we can also compute the ratio of length between the two roots,
(β, β) p
= . (5.51)
(α, α) q
If (α, β) = 0 then p = q = 0 and we obtain no condition for the length ratio.

Now suppose
(α, β) 6= 0 (i.e. α and β are not orthogonal)
β 6= ±α (i.e. α and β are not colinear)
Then we can in addition assume
(α, β) > 0 (otherwise replace α → −α)
(β, β) ≥ (α, α) (otherwise exchange α ↔ β)
Then p ≥ q > 0. Altogether, the only allowed pairs (p, q) are
(β,β)
p q (cos θ)2 = pq/4 (α,α) = p/q
3 1 3
4 (θ = ±30◦ ) 3
2 1 1
2 (θ = ±45◦ ) 2
1 1 1
4 (θ = ±60◦ ) 1
◦
0 0 0 (θ = ±90 ) no cond.
Consider now a Lie algebra of rank 2. Let θm be the smallest angle between
two distinct roots that occurs. The following are all possible root systems:
θm = 90◦ . The root system is
By lemma 5.14 this Lie algebra is not simple. It has to be the direct sum of two
rank 1 Lie algebras. Up to isomorphism, there is only one such algebra, hence
g = sl(2, C) ⊕ sl(2, C) . (5.52)
51
θm = 60◦ . Let α1 , α3 be two roots with this angle. Then by the above table,
α1 and α3 have the same length,
α2 α3
α1
The root α2 has been obtained by completing this picture with respect to Weyl
reflections. Also for each root α, a root −α has been added. There can be no
further roots, or one would have a minimum angle less than 60◦ . Thus
g = sl(3, C) . (5.53)
This Lie algebra is also called A2 .
θm = 45◦ . Let α, β be two roots with this angle. Then by the above table,
(β, β) = 2(α, α),
Again, the root system has been completed with respect to Weyl reflections and
α → −α. This Lie algebra is called B2 . [It is a complexification of so(5), see
section 5.8.]
θm = 30◦ . Let α, β be two roots with this angle. Then by the above ta-
ble, (β, β) = 3(α, α), and the roots system (completed with respect to Weyl
reflections and α → −α) is
52
This Lie algebra is called G2 . [See [Fulton-Harris, chapter 22] for more on G2 .]
Exercise 5.18 :
Give the dimension (over C) of all rank two fssc Lie algebras as found in section
5.6.
5.7 Dynkin diagrams

Let g be a fssc Lie algebra and R = spanR (Φ). Pick a vector n ∈ R such that
the hyperplane
H = {v ∈ R|(v, n) = 0} (5.54)
does not contain an element of Φ (i.e. H ∩ Φ = ∅). Define
positive roots Φ+ = {α ∈ Φ|(α, n) > 0}
negative roots Φ− = {α ∈ Φ|(α, n) < 0}
Exercise 5.19 :
Let g be a fssc Lie algebra and let Φ+ and Φ− be the positive and negative roots
with respect to some hyperplane. Show that
(i) Φ = Φ+ ∪ Φ− .
(ii) α ∈ Φ+ ⇔ −α ∈ Φ− and |Φ+ | = |Φ− | (the number of elements in a set S is
denoted by |S|).
(iii) spanR (Φ+ ) = spanR (Φ).
For example, for sl(3, C) we can take
n
α2 α3
α1
(5.55)
Let Φs be all elements of Φ+ that cannot be written as a linear combination

of elements of Φ+ with positive coefficients and at least two terms [this is to
exclude the trivial linear combination α = 1 · α]. The roots Φs ⊂ Φ+ are called
simple roots. For example, for sl(3, C) (with the choice of n as in (5.55)) we get
Φs = {α1 , α2 }.
Properties of simple roots (which we will not prove):
Φs is a basis of R. [It is easy to see that spanR (Φs ) = R, linear independence
not so obvious.]
Let Φs = {α(1) , . . . , α(r) }. If i 6= j then (α(i) , α(j) ) ≤ 0. [Also not so obvious.]
53
Definition 5.19 :
Let g be a fssc Lie algebra and let Φs ⊂ Φ be a choice of simple roots. The
Cartan matrix is the r × r matrix with entries
2(α(i) , α(j) )
Aij = . (5.56)
(α(j) , α(j) )
Exercise 5.20 :
Using the properties of simple roots stated in the lecture, prove the following
properties of the Cartan matrix.
(i) Aij ∈ Z .
(ii) Aii = 2 and Aij ≤ 0 if i 6= j.
(iii) Aij Aji ∈ {0, 1, 2, 3} if i 6= j.
For sl(3, C) we can choose Φs = {α1 , α2 }. The off-diagonal entries of the
Cartan matrix are
2(α1 , α2 ) 2 · (−1)
A12 = = = −1 = A21 . (5.57)
(α2 , α2 ) 2
Thus the Cartan matrix of sl(3, C) is

2 −1
A= . (5.58)
−1 2
A Dynkin diagram is a pictorial representation of a Cartan matrix obtained
as follows:
Draw dots (called vertices) labelled 1, . . . , r.
For i 6= j draw Aij Aji lines between the vertices i and j.
If |Aij | > |Aji | draw an arrowhead ‘>’ on the lines between i and j pointing
from i to j.
Remove the labels 1, . . . , r.
Notes:
(1) If there is an arrow from node i to node j e i e then
i j
(i) (i) ij
(α , α ) |A |
= ji > 1 , (5.59)
(α(j) , α(j) ) |A |
i.e. the root α(i) is longer than the root α(j) .
(2) When giving a Dynkin diagram, we will often also include a labelling of
the vertices. However, this is not part of the definition of a Dynkin diagram.
Instead, it constitutes an additional choice, namely a choice of numbering of the
simple roots.
For sl(3, C) we get the Dynkin diagram (together with a choice of labelling
for the vertices, which is not part of the Dynkin diagram)
e e (5.60)
1 2
54
Exercise 5.21 :
A Dynkin diagram is connected if for any two vertices i 6= j there is a sequence of
nodes k1 , . . . , km with k1 = i, km = j such that Aks ,ks+1 6= 0 for s = 1, . . . , m−1.
[In words: one can go from any vertex to any other by walking only along lines.]
Show that if the Dynkin diagram of a fssc Lie algebra is connected, then the
Lie algebra is simple. [The converse follows from the classification theorem
of Killing and Cartan, which shows explicitly that simple Lie algebras have
connected Dynkin diagrams.]
The following two theorems we will not prove [but at least we can understand
their contents]. (For a proof see [Fulton,Harris] § 21.2 and § 21.3.)
Theorem 5.20 :
Two fssc Lie algebras g, g 0 are isomorphic iff they have the same Dynkin dia-
gram.
Theorem 5.21 :
(Killing, Cartan) Let g be a simple finite-dimensional complex Lie algebra.
The Dynkin diagram of g is one of the following:
Ar = e e e ··· e for r ≥ 1
1 2 3 r
Br = e e e ··· ei e for r ≥ 2
1 2 3 r−1 r
Cr = e e e ··· eh e for r ≥ 3
1 2 3 r−1 r
er
Dr = e e e ··· e for r ≥ 4
1 2 3 r−2@
@ er−1
e6
E6 = e e e e e
1 2 3 4 5
e7
E7 = e e e e e e
1 2 3 4 5 6
e8
E8 = e e e e e e e
1 2 3 4 5 6 7
F4 = e ei e e
1 2 3 4
G2 = ei e
1 2
55
(The names of the Dynkin diagrams above are also used to denote the corre-
sponding Lie algebras. The choice for the labelling of vertices made in the list
above is the same as e.g. in [Fuchs, Schweigert, Table IV].)
Exercise 5.22 :
Compute the Dynkin diagrams of all rank two fssc Lie algebras using the root
diagrams obtained in section 5.6.
Exercise 5.23 :
The Dynkin diagram (together with a choice for the numbering of the vertices)
determines the Cartan matrix uniquely. Write out the Cartan matrix for the
Lie algebras A4 , B4 , C4 , D4 and F4 .
5.8 Complexification of real Lie algebras

In sections 3.3–3.6 we studied the Lie algebras of matrix Lie groups. Those were
defined to be real Lie algebras. In this section we will make the connection to
the complex Lie algebras studied chapter 5.
Definition 5.22 :
Let V be a real vector space. The complexification VC of V is defined as the
quotient
spanC ((λ, v)|λ ∈ C, v ∈ V )/W (5.61)
where W is the vector space spanned (over C) by the vectors
(λ, r1 v1 + r2 v2 ) − λr1 (1, v1 ) − λr2 (1, v2 ) (5.62)
for all λ ∈ C, r1 , r2 ∈ R, v1 , v2 ∈ V . [This is nothing but to say that VC =

C ⊗R V .]
Elements of VC are equivalence classes. The equivalence class containing
the pair (λ, v) will be denoted (λ, v) + W , as usual. VC is a complex vector
space. All elements of VC are complex linear combinations of elements of the
form (λ, v) + W . But
(λ, v) + W = λ(1, v) + W , (5.63)
so that all elements of VC are linear combinations of elements of the form (1, v)+
W . We will use the shorthand notation v ≡ (1, v) + W .
Exercise* 5.24 :
Let V be a real vector space. Show that every v ∈ VC can be uniquely written
as v = (1, a) + i(1, b) + W , with a, b ∈ V , i.e., using the shorthand notation,
v = a + ib.
56
Exercise 5.25 :
Let V be a finite-dimensional, real vector space and let {v1 , . . . , vn } be a basis
of V . Show that {(1, v1 ) + W, . . . , (1, vn ) + W } is a basis of VC . (You may want
to use the result of exercise 5.24.)
Remark 5.23 :
The abbreviation v ≡ (1, v) + W has to be used with some care. Consider the
complex numbers C as two-dimensional real vectors space. Every element of C
can be written uniquely as a+ib with a, b ∈ R. Thus a basis of C (over R) is given
by e1 = 1 and e2 = i. The complexification CC therefore has the basis {e1 , e2 }
(over C). In particular, in CC we have ie1 6= e2 (or e1 and e2 are not linearly
independent). The shorthand notation might suggest that ie1 = i(1) = (i) = e2 ,
but this is not true, as in full notation

ie1 = i (1, 1) + W = (i, 1) + W 6= (1, i) + W = e2 . (5.64)
Definition 5.24 :
Let h be a real Lie algebra.
(i) The complexification hC of h is the complex vector space hC together with
the Lie bracket
[λx, µy] = λµ · [x, y] for all λ, µ ∈ C , x, y ∈ h . (5.65)
(ii) Let g be a complex Lie algebra. h is called a real form of g iff hC ∼

= g as
complex Lie algebras.
Exercise 5.26 :
Let h be a finite-dimensional real Lie algebra and let g be a finite-dimensional
complex Lie algebra. Show that the following are equivalent.
(1) h is a real form of g.
(2) There exist bases {T a |a = 1, . . . , n} of h (over R) and {T̃ a |a = 1, . . . , n} of
g (over C) such that
n
X n
X
[T a , T b ] = f ab
c T
c
and [T̃ a , T̃ b ] = f ab
c T̃
c
c=1 c=1
with the same structure constants f ab

c.
The following is an instructive example.
Lemma 5.25 :
(i) su(2) is a real form of sl(2, C).
(ii) sl(2, R) is a real form of sl(2, C).
(iii) su(2) and sl(2, R) are not isomorphic as real Lie algebras.
57
Proof:
(i) Recall that
su(2) = {M ∈ Mat(2, C)|M + M † = 0, tr(M ) = 0} ,

(5.66)
sl(2, C) = {M ∈ Mat(2, C)|tr(M ) = 0} .
In both cases the Lie bracket is given by the matrix commutator. A basis of
su(2) (over R) is

1 i 0 2 0 1 3 0 i
T = , T = , T = . (5.67)
0 −i −1 0 i 0
By exercise 5.25, the vectors T a also provide a basis (over C) of su(2)C . A basis
of sl(2, C) (over C) is

i 0 0 1 0 i
T̃ 1 = , T̃ 2 = , T̃ 3 = . (5.68)
0 −i −1 0 i 0
Since these are the the same matrices, their matrix commutator agrees, and
hence ϕ : su(2)C → sl(2, C), ϕ(T a ) = T̃ a is an isomorphism of complex Lie
algebras.
(ii) The proof goes in the same way as part (i). Recall that
sl(2, R) = {M ∈ Mat(2, R)|tr(M ) = 0} . (5.69)
A basis of sl(2, R) (over R) is

1 0 0 1 0 0
T1 = , T2 = , T3 = . (5.70)
0 −1 0 0 1 0
A basis of sl(2, C) (over C) is

1 1 0 2 0 1 3 0 0
T̃ = , T̃ = , T̃ = . (5.71)
0 −1 0 0 1 0
As before, ϕ : sl(2, R)C → sl(2, C), ϕ(T a ) = T̃ a an isomorphism of complex Lie

algebras.
(iii) This is shown in the next exercise.
Exercise 5.27 :
Show that the Killing form of su(2) is negative definite (i.e. κ(x, x) < 0 for all
x ∈ su(2)) and that the one of sl(2, R) is not. Conclude that there exists no
isomorphism ϕ : sl(2, R) → su(2) of real Lie algebras.
Here is a list of real Lie algebras whose complexifications give the simple
complex Lie algebras Ar , Br , Cr and Dr . As we have seen, several real Lie
algebras can have the same complexification, so the list below is only one possible
choice.
58
real Lie algebra h su(r+1) so(2r+1) sp(2r) so(2r)
complex Lie algebra g ∼
= hC Ar Br Cr Dr
Lemma 5.26 :
Let g be a finite-dimensional complex Lie algebra and let h be a real form of g.
Then κg is non-degenerate iff κh is non-degenerate.
Proof:
Let T a be a basis of h and T̃ a be a basis of g such that
n
X n
X
[T a , T b ] = f ab
cT
c
c T̃
c
. (5.72)
c=1 c=1
Then X X
κh (T a , T b ) = (T c )∗ ([T a , [T b , T c ]]) = f bcd f ad
c
c c,d
X
c ∗
(5.73)
= (T̃ ) ([T̃ , [T̃ b , T̃ c ]]) = κg (T̃ a , T̃ b ) .
a
This shows that in the bases we have chosen, the matrix elements of κh and κg
agree. In particular, the statement that κh is non-degenerate is equivalent to
the statement that κg is non-degenerate.
Exercise 5.28 :
Show that o(1, n−1)C ∼
= so(n)C as complex Lie algebras.
Recall that the Lie algebra of the four-dimensional Lorentz group is o(1, 3).
The exercise shows in particular, that o(1, 3) has the same complexification as
so(4). Since the Killing form of so(4) is non-degenerate, we know that so(4)C
is semi-simple.
Lemma 5.27 :
so(4) ∼
= su(2) ⊕ su(2) as real Lie algebras.
Proof:
Consider the following basis for so(4),
X1 = E23 − E32 , X2 = E31 − E13 , X3 = E12 − E21 ,

(5.74)
Y1 = E14 − E41 , Y2 = E24 − E42 , Y3 = E34 − E43 ,
i.e. in short
3
X
Xa = εabc Ebc , Ya = Ea4 − E4a , for a = 1, 2, 3 . (5.75)
b,c=1
59
The X1 , X2 , X3 are just the basis of the so(3) subalgebra of so(4) obtained
by considering only the upper left 3 × 3-block. Their commutator has been
computed in exercise 3.9,
3
X
[Xa , Xb ] = − εabc Xc . (5.76)
c=1
To obtain the remaining commutators we first compute

3
X 3
X
Xa Yb = εacd Ecd (Eb4 − E4b ) = εacb Ec4 ,
c,d=1 c=1
X 3 X3 (5.77)
Yb Xa = εacd (Eb4 − E4b )Ecd = εabd E4d ,
c,d=1 d=1
Ya Yb = (Ea4 − E4a )(Eb4 − E4b ) = −Eab − δab E44 ,
and from this, and exercise 3.8(ii), we get

3
X 3
X
[Xa , Yb ] = εabc (−Ec4 + E4c ) = − εabc Yc ,
c=1 c=1
3
X
[Ya , Yb ] = −Eab + Eba = − (δac δbd − δad δbc )Ecd (5.78)
c,d=1
3
X 3
X
=− εabx εxcd Ecd = − εabx Xx .
c,d,x=1 x=1
Set now
Ja+ = 21 (Xa + Ya ) , Ja− = 12 (Xa − Ya ) . (5.79)
Then
[Ja+ , Jb+ ] = 1

4 [Xa , Xb ] + [Xa , Yb ] + [Ya , Xb ] + [Ya , Yb ]
3
X 3
X
= − 14 εabc (Xc + Yc + Yc + Xc ) = − εabc Jc+ ,
c=1 c=1
(5.80)
[Ja+ , Jb− ] = · · · = 0 ,
3
X
[Ja− , Jb− ] = · · · = − εabc Jc− .
c=1
We see that we get two commuting copies of so(3), i.e. we have shown (note
that we have only used real coefficients)
so(4) ∼
= so(3) ⊕ so(3) (5.81)
as real Lie algebras. Together with so(3) ∼

= su(2) (as real Lie algebras) this
implies the claim.
60
Since su(2)C ∼= sl(2, C) (as complex Lie algebras), complexification immedi-
ately yields the identity
so(4)C ∼
= sl(2, C) ⊕ sl(2, C) . (5.82)
It follows that for the complexification of the Lorentz algebra o(1, 3) we equally
get o(1, 3)C ∼
= sl(2, C) ⊕ sl(2, C).
Information 5.28 :
This isomorphism is used in theoretical physics (in particular in the contexts of
relativistic quantum field theory and of supersymmetry) to describe representa-
tions of the Lorentz algebra o(1, 3) in terms of representations of sl(2, C) ⊕
sl(2, C). Irreducible representations Vd of sl(2, C) are labelled by their di-
mension d. It is also customary to denote representations of sl(2, C) by their
‘spin’ (this comes from sl(2, C) ∼ = su(2)C ); Vd then has spin s = (d − 1)/2, i.e.
V1 , V2 , V3 , . . . have spin 0, 21 , 1, . . . . Representations of o(1, 3)C are then labelled
by a pair of spins (s1 , s2 ).
61
6 Epilogue
A natural question for a mathematician would be “What are all Lie groups?
What are their representations?”. A physicist would ask the same question, but
would use different words: “What continuous symmetries can occur in nature?
How do they act on the space of quantum states?”
In this course we have answered neither of these questions, but we certainly
went into the good direction.
• In the beginning we have studied matrix Lie groups. They are typically
defined by non-linear equations, and it is easier to work with a ‘linearised
version’, which is provided by a Lie algebra. To this end we have seen that
a (matrix) Lie group G gives rise to a real Lie algebra g.
We have not shown, but it is nonetheless true, that a representation of a
Lie group also gives rise to a representation of the corresponding real Lie
algebra.
• So before addressing the question “What are all Lie groups?” we can try to
answer the simpler question “What are all real Lie algebras?” However, it
turns out that it is much simpler to work with complex Lie algebras than
with real Lie algebras. To obtain a complex Lie algebra we used that
every real Lie algebra g gives rise to a complex Lie algebra gC .
For representations one finds (but we did not) that a (real) representation
of a real Lie algebra gives rise to a (complex) representation of its complex-
ification.
• To classify all complex Lie algebras is still to hard a problem. But if
one demands two additional properties, namely that the complex Lie al-
gebra is finite-dimensional and that its Killing form is non-degenerate (these
were precisely the fssc Lie algebras), then a complete classification can be
achieved.
All finite-dimensional simple complex Lie algebras are classified in
the Theorem of Killing and Cartan.
It turns out (but we did not treat this in the course) that one can equally
classify all finite-dimensional representations of fssc Lie algebras.
One can now wonder what the classification result, i.e. the answer to “What are
all finite-dimensional simple complex Lie algebras?”, has to do with the original
question “What are all Lie groups?”. It turns out that one can retrace one’s steps
and arrive instead at an answer for the question “What are all compact connected
simple Lie groups?” (A group is called simple if has no normal subgroups other
than {e} and itself, and if it is not itself the trivial group. A connected Lie group
is called simple if it does not contain connected normal subgroups other than
{e} and itself.) A similar route can be taken to obtain the finite-dimensional
representations of a compact simple Lie group.
62
A Appendix: Collected exercises
Exercise 0.1:
I certainly did not manage to remove all errors from this script. So the first
exercise is to find all errors and tell them to me.
Exercise 1.1:
Show that for a real or complex vector space V , a bilinear map b(·, ·) : V ×V → V
obeys b(u, v) = −b(v, u) (for all u, v) if and only if b(u, u) = 0 (for all u). [If you
want to know, the formulation [X, X] = 0 in the definition of a Lie algebra is
preferable because it also works for the field F2 . There, the above equivalence
is not true because in F2 we have 1 + 1 = 0.]
Exercise 2.1:
Prove the following consequences of the group axioms: The unit is unique. The
inverse is unique. The map x 7→ x−1 is invertible as a map from G to G.
e−1 = e. If gg = g for some g ∈ G, then g = e. The set of integers together with
addition (Z, +) forms a group. The set of integers together with multiplication
(Z, ·) does not form a group.
Exercise 2.2:
Verify the group axioms for GL(n, R). Show that Mat(n, R) (with matrix mul-
tiplication) is not a group.
Exercise 2.3:
Let ϕ : G → H be a group homomorphism. Show that ϕ(e) = e (the units in G
and H, respectively), and that ϕ(g −1 ) = ϕ(g)−1 .
Exercise 2.4:
Show that Aut(G) is a group.
Exercise 2.5:
(i) Show that a subgroup H ≤ G is in particular a group, and show that it has
the same unit element as G.
(ii) Show that SO(n) is a subgroup of GL(n, R).
Exercise 2.6:
Prove that
(i*) for every f ∈ E(n) there is a unique T ∈ O(n) and u ∈ Rn , s.t. f (v) = T v+u
for all v ∈ Rn .
(ii) for T ∈ O(n) and u ∈ Rn the map v 7→ T v + u is in E(n).
Exercise 2.7:
(i) Starting from the definition of the semidirect product, show that H nϕ N is
indeed a group. [To see why the notation H and N is used for the two groups,
look up “semidirect product” on wikipedia.org or eom.springer.de.]
63
(ii) Show that the direct product is a special case of the semidirect product.
(iii) Show that the multiplication rule (T, x) · (R, y) = (T R, T y + x) found in the
study of E(n) is that of the semidirect product O(n) nϕ Rn , with ϕ : O(n) →
Aut(Rn ) given by ϕT (u) = T u.
Exercise 2.8:
Show that O(1, n−1) can equivalently be written as
O(1, n−1) = {M ∈ GL(n, R)|M t JM = J}
where J is the diaognal matrix with entries J = diag(1, −1, . . . , −1).
Exercise 2.9:
(i*) Prove that for every f ∈ P (1, n−1) there is a unique Λ ∈ O(1, n−1) and
u ∈ Rn , s.t. f (v) = Λv + u for all v ∈ Rn .
(ii) Show that the Poincaré group is isomorphic to the semidirect product
O(1, n−1) n Rn with multiplication
(Λ, u) · (Λ0 , u0 ) = (ΛΛ0 , Λu0 + u) .
Exercise 2.10:
Exercise 2.11:
(i) Consider a rotation around the 3-axis,
(Urot (θ)ψ)(q1 , q2 , q3 ) = ψ(q1 cos θ − q2 sin θ, q2 cos θ + q1 sin θ, q3 )
and check that infinitesimally
Urot (θ) = 1 + iθL3 + O(θ2 ) .
(ii) Using [qr , ps ] = iδrs (check!) verify the commutator

3
X
[Lr , Ls ] = i εrst Lt .
t=1
P3
(You might need the relation k=1 εijk εlmk = δil δjm − δim δjl (check!).)
Exercise 3.1:
(i) Show that U (n) and SU (n) are indeed groups.
(ii) Let (A† )ij = (Aji )∗ be the hermitian conjugate. Show that the condition
(Au, Av) = (u, v) for all u, v ∈ Cn is equivalent to A† A = 1, i.e.
U (n) = {A ∈ Mat(n, C) | A† A = 1} .
(iii) Show that U (n) and SU (n) are matrix Lie groups.
64
Exercise 3.2:
(i) Using the definition of the matrix exponential in terms of the infinite sum,
show that for λ ∈ C,

λ 1 λ 1 1
exp =e · .
0 λ 0 1
(ii) Let A ∈ Mat(n, C). Show that for any U ∈ GL(n, C)
U −1 exp(A)U = exp(U −1 AU ) .
(iii) Recall that a complex n × n matrix A can always be brought to Jordan

normal form, i.e. there exists an U ∈ GL(n, C) s.t.
 
J1 0
U −1 AU = 
 ..  ,

.
0 Jr
where each Jordan block is of the form

 
λk 1 0
 .. .. 
 . . 
Jk =   , λk ∈ C .
 .. 
 . 1 
0 λk
In particular, if all Jordan blocks have size 1, the matrix A is diagonalisable.

Compute
0 t 5 9
exp and exp .
−t 0 −1 −1
Exercise 3.3:
Let A ∈ Mat(n, C).
(i) Let f (t) = det(exp(tA)) and g(t) = exp(t tr(A)). Show that f (t) and g(t)
both solve the first order DEQ u0 = tr(A) u.
(ii) Using (i), show that
det(exp(A)) = exp(tr(A)) .
Exercise 3.4:
Show that if A and B commute (i.e. if AB = BA), then exp(A) exp(B) =
exp(A + B).
Exercise* 3.5:
Let G be a matrix Lie group and let g be the Lie algebra of G.
(i) Show that if A ∈ g, then also sA ∈ g for all s ∈ R.
65
(ii) The following formulae hold for A, B ∈ Mat(n, K): the Trotter Product
Formula, n
exp(A + B) = lim exp(A/n) exp(B/n) ,
n→∞
and the Commutator Formula,

n2
exp([A, B]) = lim exp(A/n) exp(B/n) exp(−A/n) exp(−B/n) .
n→∞
(For a proof see [Baker, Theorem 7.26]). Use these to show that if A, B ∈ g,
then also A + B ∈ g and [A, B] ∈ g. (You will need that a matrix Lie group is
closed.) Note that part (i) and (ii) combined prove Theorem 3.9.
Exercise 3.6:
Prove that SP (2n) is a matrix Lie group.
Exercise 3.7:
In the table of matrix Lie algebras, verify the entries for SL(n, C), SP (2n),
U (n) and confirm the dimension of SU (n).
Exercise 3.8:
(i) Show that Eab Ecd = δbc Ead .
P3
(ii) Show that x=1 εabx εcdx = δac δbd − δad δbc .
Exercise 3.9: P3
(i) Show that the generators J1 , J2 , J3 can also be written as Ja = b,c=1 εabc Ebc ,
a ∈ {1, 2, 3}.
P3
(ii) Show that [Ja , Jb ] = − c=1 εabc Jc
(iii) Check that R3 (θ) = exp(−θJ3 ) is given by
 
cos(θ) − sin(θ) 0
R3 (θ) =  sin(θ) cos(θ) 0  .
0 0 1
This is a rotation by an angle θ around the 3-axis. Check explicitly that R3 (θ) ∈
SO(3).
Exercise 3.10: P
Show that for a, b ∈ {1, 2, 3}, [σa , σb ] = 2i c εabc σc .
Exercise 3.11:
(i) Show that the set {iσ1 , iσ2 , iσ3 } is a basis of su(2) as a real vector space.
Convince yourself that the set {σ1 , σ2 , σ3 } does not form a basis of su(2) as a
real vector space. P3
(ii) Show that [iσa , iσb ] = −2 c=1 εabc iσc .
Exercise 3.12:
Show that so(3) and su(2) are isomorphic as real Lie algebras.
66
Exercise 3.13:
Show that the Lie algebra of O(1, n−1) is
o(1, n−1) = {A ∈ Mat(n, R)|At J + JA = 0} .
Exercise 3.14:
Check that the commutator of the Mab ’s is
[Mab , Mcd ] = ηad Mbc + ηbc Mad − ηac Mbd − ηbd Mac .
Exercise 3.15:
(i) Show that, for A ∈ Mat(n, R) and u ∈ Rn ,
∞
eA

A u Bu X 1 n−1
exp = , B= A .
0 0 0 1 n!
n=1
[If A is invertible, then B = A−1 (eA − 1).]

(ii) Show that the Lie algebra of P̃ (1, n−1) (the Poincaré group embedded in
Mat(n+1, R)) is
n A x o
p(1, n−1) = A ∈ o(1, n−1) , x ∈ Rn .

0 0
Exercise 3.16:
Show that, for a, b, c ∈ {0, 1, . . . , n−1},
[Mab , Pc ] = ηbc Pa − ηac Pb , [Pa , Pb ] = 0 .
Exercise 3.17:
There are some variants of the BCH identity which are also known as Baker-
Campbell-Hausdorff formulae. Here we will prove some.
Let ad(A) : Mat(n, C) → Mat(n, C) be given by ad(A)B = [A, B]. [This is
called the adjoint action.]
(i) Show that for A, B ∈ Mat(n, C),
f (t) = etA Be−tA and g(t) = etad(A) B
both solve the first order DEQ

d
u(t) = [A, u(t)] .
dt
(ii) Show that
eA Be−A = ead(A) B = B + [A, B] + 21 [A, [A, B]] + . . .
67
(iii) Show that
eA eB e−A = exp(ead(A) B)
(iv) Show that if [A, B] commutes with A and B,
eA eB = e[A,B] eB eA .
(v) Suppose [A, B] commutes with A and B. Show that f (t) = etA etB and
1 2 d
g(t) = etA+tB+ 2 t [A,B] both solve dt u(t) = (A + B + t[A, B])u(t). Show further
that
1
eA eB = eA+B+ 2 [A,B] .
Exercise 4.1:
It is also common to use ‘modules’ instead of representations. The two concepts
are equivalent, as will be clear by the end of this exercise.
Let g be a Lie algebra over K. A g-module V is a K-vector space V together
with a bilinear map . : g × V → V such that
[x, y].w = x.(y.w) − y.(x.w) for all x, y ∈ g, w ∈ V . (A.1)
(i) Show that given a g-module V , one gets a representation of g by setting

R(x)w = x.w.
(ii) Given a representation (V, R) of g, show that setting x.w = R(x)w defines
a g-module on V .
Exercise 4.2:
Show that for the Lie algebra u(1), the trivial and the adjoint representation
are isomorphic.
Exercise 4.3:
Show that if (Kn , R) is a representation of g, then so is (Kn , R+) with R+ (x) =
−R(x)t .
Exercise 4.4:
Let f : V → W be an intertwiner of two representations V, W of g. Show that
ker(f ) = {v ∈ V |f (v) = 0} and im(f ) = {w ∈ W |w = f (v) for some v ∈ V } are
invariant subspaces of V and W , respectively.
Exercise 4.5:
Check that for the basis elements of sl(2, C) one has [H, E] = 2E, [H, F ] = −2F
and [E, F ] = H.
Exercise 4.6:
Let (V, R) be a representation of sl(2, C). Show that if R(H) has an eigenvector
with non-integer eigenvalue, then V is infinite-dimensional.
Hint: Let H.v = λv with λ ∈ / Z. Proceed as follows.
1) Set w = E.v. Show that either w = 0 or w is an eigenvector of R(H) with
68
eigenvalue λ + 2.
2) Show that either V is infinite-dimensional or there is an eigenvector v0 of
R(H) of eigenvalue λ0 ∈/ Z such that E.v0 = 0.
3) Let vm = F m .v0 and define v−1 = 0. Show by induction on m that
H.vm = (λ0 − 2m)vm and E.vm = m(λ0 − m + 1)vm−1 .
4) Conclude that if λ0 ∈
/ Z≥0 all vm are nonzero.
Exercise 4.7:
The Lie algebra h = CH is a subalgebra of sl(2, C). Show that h has finite-
dimensional representations where R(H) has non-integer eigenvalues.
Exercise 4.8:
Check that the representation of sl(2, C) defined in the lecture indeed also obeys
[H, E].v = 2E.v and [H, F ].v = −2F.v for all v ∈ Cn .
Exercise 4.9:
Let (W, R) be a finite-dimensional, irreducible representation of sl(2, C). Show
that for some n ∈ Z≥0 there is an injective intertwiner ϕ : Vn → W .
Hint: (recall exercise 4.6)
1) Find a v0 ∈ W such that E.v0 = 0 and H.v0 = λ0 v0 for some h ∈ Z.
2) Set vm = F m .v0 . Show that there exists an n such that vm = 0 for m ≥ n.
Choose the smallest such n.
3) Show that ϕ(em ) = vm for m = 0, . . . , n − 1 defines an injective intertwiner.
Exercise* 4.10:
Let U, V be two finite-dimensional K-vector spaces. Let u1 , . . . , um be a basis
of U and let v1 , . . . , vn be a basis of V .
(i) [Easy] Show that
{uk ⊕ 0|k = 1, . . . , m} ∪ {0 ⊕ vk |k = 1, . . . , n}
is a basis of U ⊕ V .
(ii) [Harder] Show that
{ui ⊗ vj |i = 1, . . . , m and j = 1, . . . , n}
is a basis of U ⊗ V .
Exercise 4.11:
Show that for two Lie algebras g, h, the vector space g ⊕ h with Lie bracket as
defined in the lecture is indeed a Lie algebra.
Exercise 4.12:
(i) Show that the vector spaces U ⊕ V and U ⊗ V with g-action as defined in
69
the lecture are indeed representations of g.
(ii) Show that the vector space U ⊗ V with g-action x.(u ⊗ v) = (x.u) ⊗ (x.v) is
not a representation of g.
Exercise 4.13:
Let Vn denote the irreducible representation of sl(2, C) defined in the lecture.
Consider the isomorphism of vector spaces ϕ : V1 ⊕ V3 → V2 ⊗ V2 given by
ϕ(e0 ⊕ 0) = e0 ⊗ e1 − e1 ⊗ e0 ,
ϕ(0 ⊕ e0 ) = e0 ⊗ e0 ,
ϕ(0 ⊕ e1 ) = e0 ⊗ e1 + e1 ⊗ e0 ,
ϕ(0 ⊕ e2 ) = 2e1 ⊗ e1 ,
(so that V1 gets mapped to anti-symmetric combinations and V3 to symmetric

combinations of basis elements of V2 ⊗ V2 ). With the help of ϕ, show that
V1 ⊕ V3 ∼
= V2 ⊗ V2
as representations of sl(2, C) (this involves a bit of writing).
Exercise 4.14:
(i) Show that a sub-vector space h ⊂ g is a Lie subalgebra of g if and only if
[h, h] ⊂ h.
(ii) Show that an ideal of g is in particular a Lie subalgebra.
(iii) Show that for a Lie algebra homomorphism ϕ : g → g 0 from g to a Lie
algebra g 0 , ker(ϕ) is an ideal of g.
(iv) Show that [g, g] is an ideal of g.
(v) Show that if h and h0 are ideals of g, then their intersection h ∩ h0 is an ideal
of g.
Exercise 4.15:
Let g be a Lie algebra and h ⊂ g an ideal. Show that π : g → g/h given
by π(x) = x + h is a surjective homomorphism of Lie algebras with kernel
ker(π) = h.
Exercise 4.16:
Let g, h be Lie algebras and ϕ : g → h a Lie algebra homomorphism. Show that
if g is simple, then ϕ is either zero or injective.
Exercise 4.17:
(i) Show that for the basis of sl(2, C) used in exercise 4.5, one has
κ(E, E) = 0 , κ(E, H) = 0 , κ(E, F ) = 4 ,
κ(H, H) = 8 , κ(H, F ) = 0 , κ(F, F ) = 0 .
70
Denote by Tr the trace of 2×2-matrices. Show that for sl(2, C) one has κ(x, y) =
4 Tr(xy).
(ii) Evaluate the Killing form of p(1, 1) for all combinations of the basis elements
M01 , P0 , P1 (as used in exercises 3.14 and 3.16). Is the Killing form of p(1, 1)
non-degenerate?
Exercise 4.18:
(i) Show that for gl(n, C) one has κ(x, y) = 2n Tr(xy) − 2Tr(x)Tr(y), where Tr
is the trace of n×n-matrices.
Hint: Use the basis Ekl to compute the trace in the adjoint representation.
(ii) Show that for sl(n, C) one has κ(x, y) = 2n Tr(xy).
Exercise 4.19:
Let g be a finite-dimensional Lie algebra and let h ⊂ g be an ideal. Show that
h⊥ = {x ∈ g|κg (x, y) = 0 for all y ∈ h}
is also an ideal of g.
Exercise 4.20:
Show that if a finite-dimensional Lie algebra g contains an abelian ideal h, then
the Killing form of g is degenerate. (Hint: Choose a basis of h, extend it to a
basis of g, and evaluate κg (x, a) with x ∈ g, a ∈ h.)
Exercise 4.21:
Let g = g1 ⊕· · ·⊕gn , for finite-dimensional Lie algebras gi . Let x = x1 +· · ·+xn
and y = y1 + · · · + yn be elements of g such that xi , yi ∈ gi . Show that
n
X
κg (x, y) = κgi (xi , yi ) .
i=1
Exercise 4.22:
Let g be a finite-dimensional Lie algebra with non-degenerate Killing form. Let
h ⊂ g be a sub-vector space. Show that dim(h) + dim(h⊥ ) = dim(g).
Exercise 4.23:
Show that the Poincaré algebra p(1, n − 1), n ≥ 2, is not semi-simple.
Exercise 4.24:
In this exercise we prove the theorem that for a finite-dimensional, complex,
simple Lie algebra g, and for an invariant bilinear form B, we have B = λκg for
some λ ∈ C.
(i) Let g ∗ = {ϕ : g → C linear} be the dual space of g. The dual representation
of the adjoint representation is (g, ad)+ = (g ∗ , −ad). Let fB : g → g ∗ be given
by fB (x) = B(x, ·), i.e. [fB (x)](z) = B(x, z). Show that fB is an intertwiner
from (g, ad) to (g ∗ , −ad).
71
(ii) Using that g is simple, show that (g, ad) is irreducible.
(iii) Since (g, ad) and (g ∗ , −ad) are isomorphic representations, also (g ∗ , −ad)
is irreducible. Let fκ be defined in the same way as fB , but with κ instead of
B. Show that fB = λfκ for some λ ∈ C.
(iv) Show that B = λκ for some λ ∈ C.
Exercise 5.1:
Let {T a } be a basis of a finite-dimensional Lie algebra g over K. For x ∈ g, let
M (x)ab be the matrix of adx in that basis, i.e.
X XX
adx ( vb T b ) = ( M (x)ab vb )T a .
b a b
Show that M (T a )cb = f abc , i.e. the structure constants give the matrix elements
of the adjoint action.
Exercise* 5.2:
A fact from linear algebra: Show that for every non-degenerate symmetric bilin-
ear form b : V × V → C on a finite-dimensional, complex vector space V there
exists a basis v1 , . . . , vn (with n = dim(V )) of V such that b(vi , vj ) = δij .
Exercise 5.3:
Let g be a fssc Lie algebra and {T a } a basis such that κ(T a , T b ) = δab . Show
that the structure constants in this basis are anti-symmetric in all three indices.
Exercise 5.4:
Find a basis {T a } of sl(2, C) s.t. κ(T a , T b ) = δab .
Exercise 5.5:
Show that the diagonal matrices in sl(n, C) are a Cartan subalgebra.
Exercise 5.6:
Another fact about linear algebra: Let V be a finite-dimensional vector space
and let F ⊂ V ∗ be a proper subspace (i.e. F 6= V ∗ ). Show that there exists a
nonzero v ∈ V such that ϕ(v) = 0 for all ϕ ∈ F .
Exercise 5.7:
Let g be a fssc Lie algebra and let h ⊂ g be sub-vector space such that
(1) [h, h] = {0}.
(2) κ restricted to h is non-degenerate.
(3) if for some x ∈ g one has [x, a] = 0 for all a ∈ h, then already x ∈ h.
Show that h is a Cartan subalgebra of g if and only if it obeys (1)–(3) above.
Exercise 5.8:
Let {H 1 , . . . , H r } ⊂ g0 be a basis of g0 such that κ(H i , H j ) = δij P
(recall that r =
r
dim(g0 ) is the rankP of g). Show that for γ, ϕ ∈ g0∗ one has H γ = i=1 γ(H i )H i ,
r
as well as (γ, ϕ) = i=1 γ(H i )ϕ(H i ) and (γ, ϕ) = γ(H ϕ ).
72
Exercise 5.9:
Let g be a fssc Lie algebra and g0 a Cartan subalgebra. Let α ∈ Φ(g, g0 ). Choose
2
e ∈ gα and f ∈ g−α such that κ(e, f ) = (α,α) . Show that ϕ : sl(2, C) → g given
by
2
ϕ(E) = e , ϕ(F ) = f , ϕ(H) = Hα
(α, α)
is an injective homomorphism of Lie algebras.
Exercise* 5.10:
In this exercise we will show that dim(gα ) = 1 for all α ∈ Φ. On the way we
will also see that if α ∈ Φ and λα ∈ Φ for some λ ∈ C, then λ ∈ {±1}.
(i) Choose α ∈ Φ. Let L = {m ∈ Z|mα ∈ Φ}. Since Φ is a finite set, so is
L. Let n+ be the largest integer in L, n− the smallest integer in L. Show that
n+ ≥ 1 and n− ≤ −1.
(ii) We can assume that n+ ≥ |n− |. Otherwise we exchange α for −α. Pick
2
e ∈ gα , f ∈ g−α s.t. κ(e, f ) = (α,α) and define ϕ : sl(2, C) → g as in exercise
5.9. Show that M
U = CH α ⊕ gmα
m∈L
is an invariant subspace of the representation (g, Rϕ ) of sl(2, C).

(iii) Show that for z ∈ gmα one has Rϕ (H)z = 2mz.
(iv) By (ii), (U, Rϕ ) is also a representation of sl(2, C). Show that V = Ce ⊕
CH α ⊕ Cf is an invariant subspace of (U, Rϕ ). Show that the representation
(V, Rϕ ) is isomorphic to the irreducible representation V3 of sl(2, C).
(v) Choose an element v0 ∈ gn+ α . Set vk+1 = Rϕ (F )vk and show that
W = spanC (v0 , v1 , . . . , v2n+ )
is an invariant subspace of U .
(vi) (W, Rϕ ) is isomorphic to the irreducible representation V2n+ +1 of sl(2, C).
Show that the intersection X = V ∩ W is an invariant subspace of V and W .
Show that X contains the element H α and hence X 6= {0}. Show that X = V
and X = W .
We have learned that for any choice of v0 in gn+ α we have V = W . This can
only be if n+ = 1 and dim(gα ) = 1. Since 1 ≤ |n− | ≤ n+ , also n− = 1. Since
κ : gα × g−α is non-degenerate, also dim(g−α ) = 1.
Exercise 5.11:
Let {H i } ∪ {E α } be a Cartan-Weyl basis of a fssc Lie algebra g. Show that
(i) spanC (H 1 , . . . , H r ) is a Cartan subalgebra of g.
(ii) κ(E α , E −α ) = (α,α) 2
.
Exercise 5.12:
Let g = g1 ⊕ g2 with g1 , g2 fssc Lie algebras. For k = 1, 2, let hk be a Cartan
subalgebra of gk .
73
(i) Show that h = h1 ⊕ h2 is a Cartan subalgebra of g.
(ii) Show that the root system of g is Φ(g, h) = Φ1 ∪ Φ2 ⊂ h∗1 ⊕ h∗2 where
Φ1 = {α ⊕ 0|α ∈ Φ(g1 , h1 )} and Φ2 = {0 ⊕ β|β ∈ Φ(g2 , h2 )}.
(iii) Show that (α, β) = 0 for all α ∈ Φ1 and β ∈ Φ2 .
Exercise 5.13:
Let g be a fssc Lie algebra and g0 be a Cartan subalgebra. Suppose Φ(g, g0 ) =
Φ1 ∪ Φ2 where Φ1 , Φ2 are non-empty and (α, β) = 0 for all α ∈ Φ1 , β ∈ Φ2 . Let
{H i } ∪ {E α } be a Cartan-Weyl basis of g. Show that
M
g1 = spanC (H α |α ∈ Φ1 } ⊕ CE α
α∈Φ1
is a proper ideal in g.
Exercise 5.14:
Let R the root space of a fssc Lie algebra with Cartan subalgebra g0 .
(i) Show that the bilinear form (·, ·) on g0∗ restricts to a real valued positive
definite inner product on R.
(ii) Use the Gram-Schmidt procedure to find an orthonormal basis {ε1 , . . . , εm }
of R (over R). Show that m = r (where r = dim(g0 ) is the rank of g) and that
{ε1 , . . . , εr } is a basis of g0∗ (over C).
(iii) Show that there exists a basis {H i |i = 1, . . . , r} of g0 such that α(H i ) ∈ R
for all i = 1, . . . , r and α ∈ Φ.
Exercise 5.15:
In this exercise we construct a Cartan-Weyl basis for sl(n, C). As Cartan sub-
algebra h we take the trace-less diagonal matrices.
(i) Define the linear forms ωi , i = 1, . . . , n on diagonal n×n-matrices as ωi (Ejj ) =
δij . Define αkl = ωk − ωl . Show that for k 6= l, P αkl is a root.
n Pn
Hint: Write a general element H ∈ h as H = k=1 ak Ekk with k=1 ak = 0.
Show that [H, Ekl ] = αkl (H)Ekl .
1
(ii) Show that H αkl = 2n (Ekk − Ell ).
Hint: Use exercise 4.18 to verify κ(H αkl , H) = αkl (H) for all H ∈ h.
(iii) Show that (αkl , αkl ) = 1/n and that [Ekl , Elk ] = 2/(αkl , αkl ) · H αkl
(iv) Show that, with Φ = {αkl | k, l = 1, . . . , n , k 6= l} and E αkl = Ekl ,
α
H k,k+1 k = 1, . . . , n−1 ∪ E α α ∈ Φ

is a Cartan-Weyl basis of sl(n, C).
74
(v) Show that the root diagram of sl(3, C) is
α23 α13
α12
√
where each arrow has length 1/ 3 and the angle between the arrows is 60◦ .
Exercise 5.16:
Let sα be a Weyl reflection. Show that
sα (α) = −α , (α, λ) = 0 ⇒ sα (λ) = λ , sα ◦ sα = id , s−α = sα .
Exercise 5.17:
Show that the Weyl group of a fssc Lie algebra is a finite group (i.e. contains
only a finite number of elements).
Exercise 5.18:
Give the dimension (over C) of all rank two fssc Lie algebras as found in section
5.6.
Exercise 5.19:
Let g be a fssc Lie algebra and let Φ+ and Φ− be the positive and negative roots
with respect to some hyperplane. Show that
(i) Φ = Φ+ ∪ Φ− .
(ii) α ∈ Φ+ ⇔ −α ∈ Φ− and |Φ+ | = |Φ− | (the number of elements in a set S is
denoted by |S|).
(iii) spanR (Φ+ ) = spanR (Φ).
Exercise 5.20:
Using the properties of simple roots stated in the lecture, prove the following
properties of the Cartan matrix.
(i) Aij ∈ Z .
(ii) Aii = 2 and Aij ≤ 0 if i 6= j.
(iii) Aij Aji ∈ {0, 1, 2, 3} if i 6= j.
Exercise 5.21:
A Dynkin diagram is connected if for any two vertices i 6= j there is a sequence of
nodes k1 , . . . , km with k1 = i, km = j such that Aks ,ks+1 6= 0 for s = 1, . . . , m−1.
[In words: one can go from any vertex to any other by walking only along lines.]
75
Show that if the Dynkin diagram of a fssc Lie algebra is connected, then the
Lie algebra is simple. [The converse follows from the classification theorem
of Killing and Cartan, which shows explicitly that simple Lie algebras have
connected Dynkin diagrams.]
Exercise 5.22:
Compute the Dynkin diagrams of all rank two fssc Lie algebras using the root
diagrams obtained in section 5.6.
Exercise 5.23:
The Dynkin diagram (together with a choice for the numbering of the vertices)
determines the Cartan matrix uniquely. Write out the Cartan matrix for the
Lie algebras A4 , B4 , C4 , D4 and F4 .
Exercise* 5.24:
Let V be a real vector space. Show that every v ∈ VC can be uniquely written
as v = (1, a) + i(1, b) + W , with a, b ∈ V , i.e., using the shorthand notation,
v = a + ib.
Exercise 5.25:
Let V be a finite-dimensional, real vector space and let {v1 , . . . , vn } be a basis
of V . Show that {(1, v1 ) + W, . . . , (1, vn ) + W } is a basis of VC . (You may want
to use the result of exercise 5.24.)
Exercise 5.26:
Let h be a finite-dimensional real Lie algebra and let g be a finite-dimensional
complex Lie algebra. Show that the following are equivalent.
(1) h is a real form of g.
(2) There exist bases {T a |a = 1, . . . , n} of h (over R) and {T̃ a |a = 1, . . . , n} of
g (over C) such that
n
X n
X
[T a , T b ] = f ab
c T
c
c T̃
c
c=1 c=1
with the same structure constants f ab

c.
Exercise 5.27:
Show that the Killing form of su(2) is negative definite (i.e. κ(x, x) < 0 for all
x ∈ su(2)) and that the one of sl(2, R) is not. Conclude that there exists no
isomorphism ϕ : sl(2, R) → su(2) of real Lie algebras.
Exercise 5.28:
Show that o(1, n−1)C ∼
= so(n)C as complex Lie algebras.
76
A.1 Hints and solutions for exercises
Solution for exercise 1.1:
⇒: Setting u = v in b(u, v) = −b(v, u) gives b(u, u) = −b(u, u), i.e. 2b(u, u) = 0
and therefore b(u, u) = 0.
⇐: By assumption, b(u + v, u + v) = 0. Since b is bilinear, this is equivalent to
b(u, u) + b(u, v) + b(v, u) + b(v, v) = 0. The first and last term are again zero by
assumption, so b(u, v) + b(v, u) = 0.
Hints for exercise 2.1:
Suppose e0 is another unit. Then, since e is a unit, e · e0 = e0 , and, since e0 is
a unit, e · e0 = e. Thus e = e0 .
If x−1 and x0 are inverses of x, consider x−1 · x · x0 to show that x−1 = x0 .
Since the inverse is unique, x 7→ x−1 does indeed define a map. Show that
(x−1 )−1 = x to see that the map squares to the identity.
To see gg = g ⇒ g = e, multiply both sides by g −1 .
For (Z, ·) note that 0 and 2, 3, 4, . . . do not have inverses in Z with respect to
multiplication of whole numbers.
Hint for exercise 2.2:
Regarding Mat(n, R), what is the inverse of the matrix with all entries zero?
Since ϕ is a group homomorphism, we have ϕ(e) · ϕ(e) = ϕ(e · e) = ϕ(e).
Multiplying both sides with ϕ(e)−1 gives ϕ(e) = e. To show that ϕ(g −1 ) is
inverse to ϕ(g), by uniqueness of the inverse, it is enough to prove ϕ(g −1 )·ϕ(g) =
e = ϕ(g) · ϕ(g −1 ). This follows since ϕ(g −1 ) · ϕ(g) = ϕ(g −1 · g) = ϕ(e) = e and
similarly for the condition that ϕ(g −1 ) is right-inverse to ϕ(g).
To show that Aut(G) is a group we first have to show that the composition φ ◦ ψ
of two group automorphisms φ, ψ ∈ Aut(G) is again in Aut(G). Certainly, φ ◦ ψ
is again a bijection from G to G. It remains to show that it is a group homo-
morphism. This follows from φ(ψ(g · h)) = φ(ψ(g) · ψ(h)) = φ(ψ(g)) · φ(ψ(h)).
Now we can proceed to check the three group properties associativity, unit
and inverse. The first is clear since composition of maps is always associative:
φ ◦ (ψ ◦ η) and (φ ◦ ψ) ◦ η, applied to an element g, are both defined to mean
φ(ψ(η(g))). The unit is provided by the identity map id : G → G. The inverse
of an element φ ∈ Aut(G) is the inverse map φ−1 since by definition of the
inverse map φ(φ−1 (g)) = g = φ−1 (φ(g)), i.e. φ ◦ φ−1 = id = φ−1 ◦ φ.
(i) By definition of a subgroup, H comes with a multiplication · : H × H → H,
given by the restriction of the multiplication on G. Also by definition of a sub-
group, the multiplication on H is associative (since that of G is) and has an
inverse. It remains to check that there is a unit for H. This can be seen as
follows. Pick an element h in H. Then also h−1 and h · h−1 are in H. But
h · h−1 is the unit e of G. Thus H contains the unit of G, which is hence also a
unit for H. In particular, H is a group.
77
(ii) Let U, T ∈ SO(n). Then in particular, U, T ∈ O(n). Since O(n) is a sub-
group of GL(n, R) we have U T ∈ O(n). To see that also U T ∈ SO(n) it remains
to check that det(U T ) = 1. But this is true since det(U T ) = det(U ) det(T )
and both, det(U ) and det(T ) are equal to one. For the inverse of an element
U ∈ SO(n), we again have U −1 ∈ O(n) since O(n) is a subgroup of GL(n, R).
Further, det(U −1 ) = (det(U ))−1 = 1, so that also U −1 ∈ SO(n).
Part of solution for exercise 2.6:
(i) Let f ∈ E(n). Define h(v) = f (v) − f (0) so that now h(0) = 0. Clearly,
h preserves distances. The only difficulty is that we are not allowed to assume
that h is linear. Start from the standard basis ei of Rn and set e0i = h(ei ). Since
|ei − 0| = |h(ei ) − h(0)| = |h(ei )| we see that e0i has length one. Further, for
i 6= j,
2 = |ei − ej |2 = |h(ei ) − h(ej )|2 = g(e0i − e0j , e0i − e0j ) = 1 − 2g(e0i , e0j ) + 1 ,
which implies g(e0i , e0j ) = 0. Hence the e0i also form an orthonormal basis, and
there is a unique map T ∈ O(n) such that e0i = T ei . A point in Rn is uniquely
fixed once we know its distance to zero and to the n points ei . It follows
that knowing the e0i completely fixes the map h. Since further the orthogonal
transformation T preserves distances, we have h(v) = T v, and f is indeed of
the form f (v) = T v + u.
(i) Verify the group axioms. The ‘·’ will not be written explicitly.
Associativity: On the one hand,

(h1 , n1 )(h2 , n2 ) (h3 , n3 ) = h1 h2 , n1 ϕh1 (n2 ) (h3 , n3 )

= h1 h2 h3 , n1 ϕh1 (n2 )ϕh1 h2 (n3 ) ,
and on the other hand,

(h1 , n1 ) (h2 , n2 )(h3 , n3 ) = (h1 , n1 ) h2 h3 , n2 ϕh2 (n3 )

= h1 h2 h3 , n1 ϕh1 (n2 ϕh2 (n3 )) .
To see that these two are equal, note that
n1 ϕh1 (n2 ϕh2 (n3 )) = n1 ϕh1 (n2 )ϕh1 (ϕh2 (n3 )) = n1 ϕh1 (n2 )ϕh1 h2 (n3 )) ,
where we first used that ϕh is an automorphism of N (and hence ϕh (ab) =

ϕh (a)ϕh (b)), and second that ϕ itself is a group homomorphism from H to
Aut(N ) (and hence ϕgh (a) = ϕg (ϕh (a))).
Unit: Have
(e, e)(h, n) = (eh, eϕe (n)) = (h, n) ,
where we used that ϕe is the identity in Aut(N ), and
(h, n)(e, e) = (he, nϕh (e)) = (h, n) ,
78
where we used that a group homomorphism takes the unit to the unit.
Inverse: Have
(h, n)(h, n)−1 = (h, n)(h−1 , ϕh−1 (n−1 )) = (hh−1 , nϕh (ϕh−1 (n−1 ))) = (e, e) ,
where we used that ϕh (ϕh−1 (a)) = ϕhh−1 (a) = ϕe (a) = a, and
(h, n)−1 (h, n) = (h−1 , ϕh−1 (n−1 ))(h, n) = (h−1 h, ϕh−1 (n−1 )ϕh−1 (n))
= (e, ϕh−1 (n−1 n)) = (e, e) ,
where we used that ϕh−1 (ab) = ϕh−1 (a)ϕh−1 (b) and that ϕh−1 (e) = e.
(ii) The map ϕ : H → Aut(N ) defined by ϕh = idN , i.e. the map taking all
elements of H to the unit in Aut(N ), is a group homomorphism. With this
map, multiplication and inverse of H nϕ N read
(h, n)(h0 , n0 ) = (hh0 , nϕh (n0 )) = (hh0 , nn0 )
and
(h, n)−1 = (h−1 , ϕh−1 (n−1 )) = (h−1 , n−1 )
which is precisely the definition of the direct product H × N .
(iii) The map ϕT (u) = T u is invertible, since T is, and it obeys ϕT (u + v) =
T (u + v) = T u + T v = ϕT (u) + ϕT (v), where ‘+’ is the group operation on
Rn . Thus indeed ϕT ∈ Aut(Rn ). Further, ϕRS (u) = RSu = ϕR (ϕS (u)) so
that ϕ : O(n) → Aut(Rn ) is a group homomorphism. The multiplication on
O(n) nϕ Rn reads, writing again ‘+’ for the group operation on Rn ,
(T, x)(R, y) = (T R, x + ϕT (y)) = (T R, T y + x) .

We defined O(1, n−1) = {M ∈ GL(n, R)|η(M u, M v) = η(u, v) for all u, v ∈
Rn }. Consider the standard basis e0 , . . . , en−1 of Rn (with non-standard la-
belling). It is easy to see that η(ei , ej ) = Jij . Further,
X X X
η(M ei , M ej ) = η( Mki ek , Mlj el ) = Mki Mlj η(ek , el )
X k l k,l
= Mki Jkl Mlj = [M t JM ]ij
k,l
so that η(M u, M v) = η(u, v) for all u, v ∈ Rn is equivalent to M t JM = J.

(i) Similar to part (i) of exercise 2.6.
(ii) Let ϕ : O(1, n−1) → Aut(Rn ) be given by ϕΛ (u) = Λu. That this is a
group homomorphism is checked in the same way as in exercise 2.7(iii). For
79
Λ ∈ O(1, n−1) and x ∈ Rn , let fΛ,x : Rn → Rn be given by fΛ,x (u) = Λu + x.
Since
η(fΛ,x (u) − fΛ,x (v), fΛ,x (u) − fΛ,x (v)) = η(Λu − Λv, Λu − Λv)
= η(Λ(u − v), Λ(u − v)) = η(u − v, u − v)
we have fΛ,x ∈ P (1, n−1). Consider the map F : O(1, n−1) nϕ Rn → P (1, n−1)
given by F (Λ, x) = fΛ,x . The map F is clearly injective. By part (i) it is
also surjective, so that it is a bijection. It remains to check that it is a group
homomorphism. To this end note that for all u ∈ Rn ,
F ((Λ, x)(Λ0 , x0 ))(u) = F (ΛΛ0 , Λx0 + x)(u) = fΛΛ0 ,Λx0 +x (u) = ΛΛ0 u + Λx0 + x
and
(F (Λ, x)F (Λ0 , x0 ))(u) = fΛ,x (fΛ0 ,x0 (u)) = fΛ,x (Λ0 u + x0 ) = Λ(Λ0 u + x0 ) + x ,
so that indeed F ((Λ, x)(Λ0 , x0 )) = F (Λ, x)F (Λ0 , x0 ).

Just substitute the definition of the commutator and check that in the resulting
sum all terms cancel.
(i) Show that they are subgroups of GL(n, C), similar to the argument for O(n)
and SO(n).
(ii) First convince yourself that for A, B ∈ Mat(n, C) the statements A = B and
(u, Av) = (u, Bv) ∀u, v ∈ Cn are equivalent. Next note that by definition
n
X n
X
(u, Av) = u∗k Akl vl = (A∗kl uk )∗ vl = (A† u, v) .
k,l=1 k,l=1
Then
(Au, Av) = (u, v) ∀u, v ∈ Cn ⇔ (u, A† Av) = (u, v) ∀u, v ∈ Cn ⇔ A† A = 1 .
(iii) From (i) we know that U (n) and SU (n) are subgroups of GL(n, C). It
remains to show that they are closed. The maps f : Mat(n, C) → Mat(n, C) and
g : Mat(n, C) → Mat(n, C) × C given by f (A) = A† A and g(A) = (A† A, det(A))
are both continuous. Thus the preimage of the closed set {1} under f and of
{(1, 1)} under g will also be closed. But f −1 ({1}) = U (n) and g −1 ({(1, 1)}) =
SU (n), proving that these two subsets of GL(n, C) are closed.
(i) By recursion on n one checks
n n
nλn−1

λ 1 λ
= .
0 λ 0 λn
80
The infinite sum is then easy to evaluate,
∞ n X ∞ n
nλn−1 eλ eλ

X 1 λ 1 1 λ
= = ,
n! 0 λ n! 0 λn 0 eλ
n=0 n=0
which proves the assertion.

(ii) Note that
U −1 An U = U −1 AA · · · AU = U −1 AU U −1 AU · · · U −1 AU = (U −1 AU )n .
Then
∞ ∞
X 1 −1 n X 1 −1
U −1 exp(A)U = U A U= (U AU )n = exp(U −1 AU ) .
n=0
n! n=0
n!
(iii) Have

−1 0 t it 0 −i i
U U= for U= .
−t 0 0 −it 1 1
Then
eit

0 t it 0 −1 0
exp = U exp U =U U −1
−t 0 0 −it 0 e−it

cos(t) sin(t)
= .
− sin(t) cos(t)
Have

−1 5 9 2 1 3 1
U U= for U = .
−1 −1 0 2 −1 0
Then

5 9 2 1 −1 2 1 1 −1 2 4 9
exp = U exp U = e ·U U =e ·
−1 −1 0 2 0 1 −1 −2

(i) First note that det(1 + εA) is a polynomial in ε. From the explicit form of
the determinant one finds det(1 + εA) = 1 + εtr(A) + O(ε2 ). Next consider the
functions f (t) = det(exp(tA)), g(t) = exp(t tr(A)). To find the derivative of f
with respect to t note
det(exp((t + ε)A)) = det(exp(tA)) det(exp(εA))
= det(exp(tA))(1 + εtr(A) + O(ε2 )) ,
so that f 0 (t) = tr(A) det(exp(tA)) = tr(A)f (t). The derivative of g(t) is

straightforward, g 0 (t) = tr(A) exp(ttr(A)) = tr(A)g(t). Thus both, f (t) and
81
g(t) solve u0 (t) = tr(A)u(t).
(ii) The solution to the first order DEQ u0 (t) = tr(A)u(t) is uniquely determined
by the initial condition u(0). Since f (0) = 1 = g(0), it follows that f (t) = g(t).
In particular f (1) = g(1), which proves the assertion.
Have
∞
X 1
eA+B = (A + B)n .
n=0
n!
Now since A and B commute,
n
n
X n
(A + B) = Ak B n−k .
k
k=0
Substituting this results in the same sum as for eA eB .

(i) This is just part of the definition of a Lie algebra g of a matrix Lie group G:
if A ∈ g then also tA ∈ g for all t ∈ R.
(ii) Since A, B ∈ g, it follows that exp(tA) and exp(tB) are in G for all t ∈ R,
in particular for t = 1/n for n ∈ Z>0 . Since G is a group, also the prod-
uct exp(tA) exp(tB) is in G. Hence Mn = exp(A/n) exp(B/n) is a sequence
of elements in G. By the Trotter product formula, the n → ∞ limit exists
in Mat(n, R) and is given by exp(A + B). Since G is closed, it follows that
exp(A + B) ∈ G. Substituting A = tX and B = tY we see that if X, Y ∈ g then
exp(t(X + Y )) ∈ G for all t ∈ R and hence X + Y ∈ g.
Regarding the commutator, if A, B ∈ g then Mn = exp(A/n) exp(B/n) exp(−A/n)
exp(−B/n) is a sequence of elements in G. By the commutator formula its limit
exists in Mat(n, R) and is given by exp([A,
√ B]). Since√G is closed, it follows that
exp([A, B]) ∈ G. Substituting A = tX and B = ± tY we see that if X, Y ∈ g
then exp(±t[X, Y ]) ∈ G for all t > 0 and hence [X, Y ] ∈ g.
This follows in the same way as exercise 3.1 (iii), i.e. by using that SP (2n) is
defined via the inverse image of a closed set under a continuous function.
SL(n, C): We have to find all A ∈ Mat(n, C) such that det(exp(tA)) = 1 for
all t ∈ R. From det(exp(tA)) = et tr(A) we see that tr(A) = 0 is necessary and
sufficient. The subspace of complex matrices with tr(A) = 0 has real dimension
2(n2 − 1).
SP (2n): Suppose exp(sA), with A ∈ Mat(2n, R) and s ∈ R, obeys (exp(sA))t Jsp exp(sA) =
Jsp . To first order in s this implies
Jsp + s(At Jsp + Jsp A) + O(s2 ) = Jsp ,
82
i.e. Jsp A + At Jsp = 0. Conversely, suppose that Jsp A + At Jsp = 0. Then, for
any s ∈ R,
−1
(exp(sA))t Jsp exp(sA) = Jsp Jsp exp(sAt )Jsp exp(sA)
(1) −1 t (2) −1
= Jsp exp(sJsp A Jsp ) exp(sA) = Jsp exp(−sJsp Jsp A) exp(sA)
(3)
= Jsp exp(−sA) exp(sA) = Jsp ,
where (1) makes use of exercise 3.2 (ii), (2) follows since by assumption At Jsp =
−Jsp A, and (3) amounts to exercise 3.4 for B = −A. We see that Jsp A+At Jsp =
0 is necessary and sufficient for exp(sA) ∈ SP (2n) for all s ∈ R.
To find the dimension of sp(2n), write A ∈ Mat(2n, R) in terms of four n×n-
matrices E, F, G, H as
E F
A= .
G H
One computes
Et Gt −Gt Et

0 1
At Jsp = =
Ft Ht −1 0 −H t Ft
and similarly
G H
Jsp A = .
−E −F
The condition Jsp A + At Jsp = 0 is thus equivalent to
Gt = G , E t = −H , Ft = F .
This shows that G and F have to be symmetric, while H is arbitrary, and E is

fixed uniquely by H. The (real) dimension of sp(2n) is therefore
dimR (sp(2n)) = 21 n(n+1) + 21 n(n+1) + n2 = n(2n+1) .
U (n): Suppose M = exp(tA), with M, A ∈ Mat(n, C) and t ∈ R, obeys

M † M = 1. To first order in t this implies
1 + t(A† + A) + O(t2 ) = 1 ,
i.e. A† + A = 0. Conversely, suppose that A† + A = 0. Then, for any t ∈ R,
(exp(tA))† exp(tA) = exp(tA† ) exp(tA) = exp(−tA) exp(tA) = 1 .
Thus A† + A = 0 is necessary and sufficient for exp(tA) ∈ U (n) for all t ∈ R.

A matrix A ∈ Mat(n, C) obeying A† + A = 0 will have purely imaginary diag-
onal entries, and its lower triangular part is determined in terms of its upper
triangular part. Hence A is determined by n real and 21 n(n − 1) complex pa-
rameters. The real dimension of u(n) is therefore n + n(n − 1) = n2 .
83
SU (n): In the calculation for U (n) we have seen that an element A of u(n)
has entries ir1 , . . . , irn on the diagonal, with rk ∈ R. The condition tr(A) = 0
fixes rn in terms of r1 , . . . , rn−1 so that the real dimension of su(n) is that of
u(n) less one: dimR (su(n)) = n2 − 1.
(i) We verify the identity Eab Ecd = δbc Ead componentwise. By matrix multipli-
cation
n
X n
X

Eab Ecd ij = Eab ik Ecd kj = δai δbk δck δdj = δai δbc δdj = δbc Ead ij .
k=1 k=1
(ii) We want to show that

3
X
εabx εcdx = δac δbd − δad δbc . (∗)
x=1
Suppose a = b or c = d: By anti-symmetry, either εabx = 0 or εcdx = 0, so that

the lhs of (*) is zero. On the rhs, both terms cancel, so in this case (*) reads
0 = 0 and is true.
Suppose a 6= b and c 6= d: There is a unique value xab such that εabxab 6= 0. So
only one term (x = xab ) contributes to the lhs of (*), which now reads
εabxab εcdxab = δac δbd − δad δbc if a 6= b and c 6= d . (∗∗)
In particular, if the two (unordered) sets {a, b} and {c, d} are not equal, then
both sides of (**) are zero. If {a, b} = {c, d}, then either a = c and b = d in
which case the lhs is (±1)2 and the rhs is 1, or a = d and b = c in which case
the lhs is (±1)(∓1) and the rhs is −1.
Hence indeed (*) is true. (Of course one could also have checked all 34 = 81
possible index combinations explicitly.)
(i)
P This can be seen by explicit verification of all three cases. For example,
a,b ε1ab Eab = ε123 E23 + ε132 E32 = E23 − E32 . This is precisely the matrix given
for J1 .
(ii) We have
X X X
Ja Jb = εars εbtu Ers Etu = εars εbtu δst Eru = εars εubs Eru
X r,s,t,u r,s,t,u r,s,u
= (δau δrb − δab δru )Eru
r,u
and
X X
[Ja , Jb ] = (δau δrb − δab δru − δbu δra + δba δru )Eru = (δau δrb − δbu δra )Eru
X r,u X r,u
= εabc εurc Eru = − εabc Jc
r,u,c c
84
(iii) Use the method from exercise 3.2 (iii) to compute the exponential. For the
explicit check that R3 (θ) ∈ SO(3) compute det(R3 (θ)) and R3 (θ)t R3 (θ) using
sin(θ)2 + cos(θ)2 = 1.
Because of anti-symmetry of the commutator it is enough to verify the identity
in the cases [σ1 , σ2 ] = 2iε123 σ3 , [σ1 , σ3 ] = 2iε132 σ2 and [σ2 , σ3 ] = 2iε231 σ1 . For
example

0 1 0 −i i 0 −i 0
σ1 σ2 = = and σ2 σ1 =
1 0 i 0 0 −i 0 i
so that indeed [σ1 , σ2 ] = σ1 σ2 − σ2 σ1 = 2iσ3 .

(i) An element A ∈ su(2) obeys A† + A = 0 and tr(A) = 0. It is therefore a
complex 2 × 2 matrix with entries

ix y + iz
A= where x, y, z ∈ R .
−y + iz −ix
Each such matrix can be written as a real linear combination of the three ma-
trices

0 i 0 1 i 0
b1 = , b2 = , b3 = .
i 0 −1 0 0 −i
We have b1 = iσ1 , b2 = iσ2 and b3 = iσ3 , which demonstrates the claim. (In
particular, σ1 , σ2 , σ3 do not form a basis of su(2) as, first of all, they are not
actually in su(2), and second, one would need complex linear combinations to
reach elements of su(2)).
(ii) Have
3
X 3
X
[iσa , iσb ] = −[σa , σb ] = −2i εabc σc = −2 εabc (iσc ) .
c=1 c=1

Define the linear map ϕ : so(3) → su(2) by prescribing its values on a basis.
For so(3) we use the basis {J1 , J2 , J3 } from section 3.5 and for su(2) we use the
basis {iσ1 , iσ2 , iσ3 }. We set ϕ(Jk ) = 21 iσk , k ∈ {1, 2, 3}. As ϕ takes a basis to
a basis it is an isomorphism of real vector spaces.
To establish that so(3) is isomorphic to su(2) as a real Lie algebra, it is now
enough to verify that ϕ is a Lie algebra homomorphism. We find
3
X 3
X 3
X
ϕ([Ja , Jb ]) = ϕ(− εabc Jc ) = − εabc ϕ(Jc ) = − 12 εabc (iσc )
c=1 c=1 c=1
85
and
3
X
[ϕ(Ja ), ϕ(Jb )] = [ 12 iσa , 12 iσb ] = 14 [iσa , iσb ] = 14 (−2) εabc (iσc ) .
c=1
Thus indeed ϕ([Ja , Jb ]) = [ϕ(Ja ), ϕ(Jb )].

We need to find all A ∈ Mat(n, R) such that exp(tA)t J exp(tA) = J for all
t ∈ R. To first order in t this implies
1 + tAt + O(t2 ) J 1 + tA + O(t2 ) = J

⇒ J + t(At J + JA) + O(t2 ) = J ⇒ At J + JA = 0 .
Conversely, note that JJ = 1 so that J −1 = J. Thus At J +JA = 0 is equivalent

to At = −JAJ and we have
exp(tA)t J exp(tA) = exp(tAt )J exp(tA) = exp(−tJAJ)J exp(tA)
= J exp(−tA)JJ exp(tA) = J exp(−tA) exp(tA) = J .
Altogether, we see that exp(tA)t J exp(tA) = J for all t ∈ R is equivalent to

At J + JA = 0, as was to be shown.
Note that
[Eab , Ecd ] = Eab Ecd − Ecd Eab = δcb Ead − δad Ecb .
Using this and ηab = ηaa δab , we find
[Mab , Mcd ] = [ηbb Eab − ηaa Eba , ηdd Ecd − ηcc Edc ]

= ηbb ηdd δcb Ead − δad Ecb − ηbb ηcc δdb Eac − δac Edb

− ηaa ηdd δca Ebd − δbd Eca + ηaa ηcc δda Ebc − δbc Eda

= ηad ηcc Ebc − ηbb Ecb + ηbc ηdd Ead − ηaa Eda

− ηac ηdd Ebd − ηbb Edb − ηbd ηcc Eac − ηaa Eca
= ηad Mbc + ηbc Mad − ηac Mbd − ηbd Mac .

(i) By induction on n one can show that for n ≥ 1,
n n
An−1 u

A u A
= .
0 0 0 0
86
Using this to evaluate the exponential gives the desired answer.
(ii) Consider exp(tC) for C ∈ Mat(n+1, R). To first order in t, exp(tC) =
1 + tC + O(t2 ). For this to be in P̃ (1, n−1) we need

2 M (t) u(t)
1 + tC + O(t ) =
0 1
for some M (t) ∈ O(1, n−1) and u(t) ∈ Rn . This requires C to be of the form

A x
C= with A ∈ o(1, n−1) .
0 0
Conversly, if C is of the above form, then part (i) shows that

tA
e Bx
exp(tC) =
0 1
for some B ∈ Mat(n, R). Since A ∈ o(1, n−1), by definition exp(tC) ∈ O(1, n−1)
so that indeed p(1, n−1) has the claimed form.
Writing out the definition gives
[Mab , Pc ] = [ηbb Eab − ηaa Eba , Ecn ]

= ηbb δbc Ean − δan Ecb − ηaa δac Ebn − δbn Eca
= ηbc Ean − ηac Ebn = ηbc Pa − ηac Pb ,
where we used that δan = 0 and δbn = 0 as a, b ∈ {0, 1, . . . , n − 1}. Further,
[Pa , Pb ] = [Ean , Ebn ] = δnb Ean − δan Ebn = 0 .

(i) Calculate
d
f (t) = AetA Be−tA − etA BAe−tA = Af (t) − f (t)A = [A, f (t)]
dt
where we used that [A, etA ] = 0. Further
∞
d X ntn−1
g(t) = ad(A)n B = ad(A)etad(A) B = [A, g(t)] .
dt n=0
n!
(ii) Since f (t) and g(t) solve the same first order DEQ and since f (0) = g(0) =
B, we have f (t) = g(t) for all t. In particular f (1) = g(1), which is the identity
that was to be shown.
(iii) Recall that for any invertible U we have U eA U −1 = exp(U AU −1 ). Hence
eA eB e−A = exp(eA Be−A ) = exp(ead(A) B) .
87
(iv) If [A, B] commutes with A it follows that
ead(A) B = B + [A, B]
as all higher terms vanish. By part (ii) we have
eA eB = eB+[A,B] eA .
But B and [A, B] commute by assumption, so by exercise 3.4 we have
eB+[A,B] = e[A,B] eB .
which implies the claim.

(v) For the derivative of f (t) we compute
f 0 (t) = AetA etB + etA BetB = (A + etA Be−tA )etA etB = (A + B + t[A, B])f (t) ,
where in the last step we used part (ii) and that assumption that [A, B] com-
mutes with A. To compute the derivative of g(t) we write out the exponen-
tial, differentiate each power with respect to t, and use that by assumption
A+B+ 21 t[A, B] and A+B+t[A, B] commute to get g 0 (t) = (A+B+t[A, B])g(t).
Now f (0) = g(0) = 1, so that f (t) = g(t). For t = 1 this implies the second
claim.
(i) We are given a g-module V . For x ∈ g we define the linear map R(x) : V → V
as R(x)w = x.w for all w ∈ V . We now check that R : g → End(V ) is a Lie
algebra homomorphism. For all x, y ∈ g and w ∈ V we have
R([x, y])w = [x, y].w = x.(y.w) − y.(x.w) = R(x)R(y)w − R(y)R(x)w
= [R(x), R(y)]w .
Thus (V, R) is a representation of g.

(ii) We are given a representation (V, R) of g. Consider the bilinear map . :
g × V → V given by x.w = R(x)w for all x ∈ g, w ∈ V . We need to check that
this defines a g-module on V . For all x, y ∈ g and w ∈ V we have
[x, y].w = R([x, y])w = R(x)R(y)w − R(y)R(x)w = x.(y.w) − y.(x.w) .
Thus V with the g-action . : g × V → V defined above is a g-module.

The (real) Lie algebra u(1) is the vector space R with Lie bracket [x, y] = 0 for all
x, y ∈ R. Let (R, Rtr ) be the trivial representation (i.e. Rtr = 0), and let (R, ad)
be the adjoint representation. Define f : R → R to be just the identity f (x) = x.
This is an isomorphism, and it provides an intertwiner (R, ad) → (R, Rtr ) since
Rtr (x)(f (y)) = 0 and f (adx y) = [x, y] = 0 .
88
Have
R+ ([x, y]) = −R([x, y])t = −(R(x)R(y) − R(y)R(x))t
= −(R(y)t R(x)t − R(x)t R(y)t )
= (−R(x)t )(−R(y)t ) − (−R(y)t )(−R(x)t )
= R+ (x)R+ (y) − R+ (y)R+ (x) ,

thus (Kn , R+ ) is a representation of g.
Suppose v ∈ ker(f ). We need to show that x.v ∈ ker(f ) for all x ∈ g. Have
f (x.v) = x.f (v) = 0 ,
where we first used that f is an intertwiner and then that v ∈ ker(f ). It follows
that indeed x.v ∈ ker(f ), and hence that ker(f ) is an invariant subspace.
Next suppose that w ∈ im(f ) and x ∈ g. We need to show that also x.w ∈ im(f ).
By assumption there is a v ∈ V such that w = f (v). Then
x.w = x.f (v) = f (x.v) ∈ im(f) .
Thus also im(f ) is an invariant subspace.

Just check all equations explicitly. For example

1 0 0 1 0 1 1 0 0 2
[H, E] = − = = 2E .
0 −1 0 0 0 0 0 −1 0 0

1) We have H.w = H.(E.v) = [H, E].v + E.(H.v) = 2E.v + λE.v = (λ + 2)w.
Thus if w 6= 0 then w is an eigenvector of R(H) with eigenvalue λ + 2.
2) Suppose E m .v 6= 0 for all m > 0. Then they form an infinite set of eigen-
vectors of R(H) with distinct eigenvalues, and hence an infinite set of linearly
independent vectors in V . Therefore V is infinite-dimensional.
Suppose next that E n .v = 0 for some n ≥ 0. Choose the smallest such n. Then
v0 = E n−1 .v is nonzero (by definition) and E.v0 = 0. Continuing the argument
of 1) by induction we have H.(E m .v) = (λ + 2m)E m .v and so H.v0 = λ0 v0 with
λ0 = λ + 2n − 2. In particular, if λ 6= Z also λ0 6= Z.
3) For m = 0 the equations are clearly true. Assume now that we have shown
them for a given m. Then for m + 1 we compute
H.vm+1 = H.(F.vm ) = [H, F ].vm + F.(H.vm ) = −2F.vm + (λ0 − 2m)F.vm
= (λ0 − 2m − 2)vm+1 ,
89
and
E.vm+1 = E.(F.vm ) = [E, F ].vm + F.(E.vm ) = H.vm + m(λ0 − m + 1)F.vm−1

= λ0 − 2m + m(λ0 − m + 1) vm = (m + 1)(λ0 − m)vm .
4) We proceed again by induction. v0 6= 0 by construction. Suppose now

that vm−1 6= 0 for some m > 0. Then E.vm = m(λ0 − m + 1)vm−1 . But
m(λ0 −m+1) 6= 0 as λ0 ∈ / Z and m > 0, and vm−1 6= 0 by induction assumption.
Therefore E.vm 6= 0 and hence also vm 6= 0.
To complete the proof of the statement in the exercise it now suffices to notice
that the infinite set of vectors {vm | m ∈ Z≥0 } are all eigenvectors of R(H)
with distinct eigenvalues, and hence linearly independent. Thus V is infinite-
dimensional.
Take V to be a one-dimensional vector space with basis v. Define R on a basis
via R(H)v = λv for an arbitrary λ ∈ C. Since h is one-dimensional, the relation
[R(x), R(y)] = R([x, y]) for all x, y ∈ h amounts to [R(H), R(H)] = R([H, H])
which holds trivially. Therefore (V, R) is a finite-dimensional representation of
h such that R(H) has an eigenvector with eigenvalue λ ∈ C which need not be
an integer.
It is again useful to define e−1 = en = 0. For m = 0, . . . , n − 1 have
[H, E].em = 2E.em = 2m(n − m)em−1
and
H.(E.em ) − E.(H.em ) = m(n − m)H.em−1 − (n − 1 − 2m)E.em
= m(n − m)(n − 1 − 2m + 2)em−1 − (n − 1 − 2m)m(n − m)em−1
= m(n − m)(n − 1 − 2m + 2 − n + 1 + 2m)em−1 = 2m(n − m)em−1 .
Further,
[H, F ].em = −2F.em = −2em+1
and
H.(F.em ) − F.(H.em ) = H.em+1 − (n − 1 − 2m)F.em
= (n − 1 − 2m − 2)em+1 − (n − 1 − 2m)em+1 = −2em+1 .

Thus indeed [H, E].v = 2E.v and [H, F ].v = −2F.v holds for a basis of Cn , and
therefore for all v ∈ Cn .
1) Since W is a finite-dimensional complex vector space, R(H) has an eigenvec-
tor v. The argument in part 2) of exercise 4.6 then shows that we can find an
eigenvector v0 of R(H) such that E.v0 = 0. Let λ0 be the eigenvalue of v0 .
90
2) From part 3) of exercise 4.6 we know that the nonzero vm are all eigenvectors
of R(H) with distinct eigenvectors. Since W is finite-dimensional, there can
only be a finite number of nonzero vm . Choose n such that v0 , . . . , vn−1 are
nonzero and vn = 0. As vm = F.vm−1 , clearly vm = 0 for m ≥ n.
3) From part 3) of exercise 4.6,
E.vn = n(λ0 − n + 1)vn−1 .
By construction, vn−1 6= 0 and vn = 0. This is only possible if λ0 = n − 1

(the possibility n = 0 is excluded because v0 6= 0). Comparing to the definition
in lemma 4.9 it is now straightforward to check that for x ∈ {E, H, F } and
m = 0, . . . , n − 1,
ϕ(x.em ) = x.vm
which implies ϕ(x.u) = x.ϕ(u) for all x ∈ sl(2, C) and u ∈ Cn . Thus ϕ is an
intertwiner. Since the em form a basis and the vm are linearly independent, ϕ
is injective.
(i) In the given bases of U and V , any element x of U ⊕ V can be written as
x = (a1 u1 + · · · + am um , b1 v1 + · · · + bn vn ) .
By the rules for addition and scalar multiplication, this is equal to
x = a1 (u1 , 0) + · · · + am (um , 0) + b1 (0, v1 ) + · · · + bn (0, vn ) .
Further, x = 0 implies that all ai and bj have to be zero (since the ui and vj
form bases), so that (ui , 0) ≡ ui ⊕ 0, i = 1, . . . , m together with (0, vj ) ≡ 0 ⊕ vj ,
j = 1, . . . , n does indeed form a basis of U ⊕ V .
(ii)
The ui ⊗ vj span U ⊗ V : It is enough to show that we can reach any element
of the form u ⊗ v (since any x ∈ U ⊗ V can be written as a sum of elements of
this form). In terms of the bases of U and V we have u = a1 u1 + · · · + am um
and v = b1 v1 + · · · + bn vn for some constants ai , bj . By the rules for the tensor
product we can write X
u⊗v = ai bj ui ⊗ vj .
i,j
The ui ⊗ vj are linear independent: Let S = spanK ((u, v)| u ∈ U, v ∈ V ), so

that U ⊗ V = S/W , with W as in the lecture. Let α : U → K be a linear
map. Define the map β̃ : S → V on the generators of S by β̃((u, v)) = α(u)v.
One checks that β̃ vanishes on W (by linearity), so that it gives rise to a well
defined map on the quotient space S/W : β(u ⊗ v) = α(u)v. Choose α(ui ) = δij .
Suppose now that for some constants λkl ,
m X
X n
λkl uk ⊗ vl = 0 .
k=1 l=1
91
Applying β gives X X
0= λkl α(uk )vl = λjl vl ∈ V
k,l l
Since the vl are linear independent, we find that λjl = 0 for l = 1, . . . , n. But j
was arbitrary, so that λkl = 0 for all k, l.
The Lie bracket on g ⊕ h is bilinear. For example,
[x ⊕ y + x0 ⊕ y 0 , p ⊕ q] = [(x + x0 ) ⊕ (y + y 0 ), p ⊕ q]
= [x + x0 , p] ⊕ [y + y 0 , q] = ([x, p] + [x0 , p]) ⊕ ([y, q] + [y 0 , q])
= [x, p] ⊕ [y, q] + [x0 , p] ⊕ [y 0 , q] = [x ⊕ y, p ⊕ q] + [x0 ⊕ y 0 , p ⊕ q] .
Skew-symmetry [x⊕y, x⊕y] = 0 and the Jacobi identity also follows immediately
form the corresponding properties of g and h.
(i) U ⊕ V is a representation of g, since the action is bilinear, and we have, for
all x, y ∈ g and u ⊕ v ∈ U ⊕ V ,
[x, y].u ⊕ v = ([x, y].u) ⊕ ([x, y].v)
and (not writing the ‘.’)
xy u ⊕ v − yx u ⊕ v = (xyu) ⊕ (xyv) − (yxu) ⊕ (yxv)
= (xyu − yxu) ⊕ (xyv − yxv)
which agrees with [x, y].u ⊕ v since U and V are representations of g.

U ⊗ V is a representation of g, since the action is bilinear, and we have, for
all x, y ∈ g and u ⊗ v ∈ U ⊗ V ,
[x, y].u ⊗ v = ([x, y].u) ⊗ v + u ⊗ ([x, y].v)
and (not writing the ‘.’)
xy u ⊗ v − yx u ⊗ v = (xyu) ⊗ v − (yxu) ⊗ v + u ⊗ (xyv) − u ⊗ yxv
= (xyu − yxu) ⊗ v + u ⊗ (xyv − yxv)
which agrees with [x, y].u ⊗ v.

(ii) The map (x, u ⊗ v) 7→ (xu) ⊗ (xv) from g × (U ⊗ V ) to U ⊗ V is not bilinear.
For example, (λx, u ⊗ v) gets mapped to λ2 (xu) ⊗ (xv) and not λ(xu) ⊗ (xv).
For e0 ⊕ 0 note that H, E, F annihilate this element, and also
H.(e0 ⊗ e1 − e1 ⊗ e0 ) = (He0 ) ⊗ e1 + e0 ⊗ (He1 ) − (He1 ) ⊗ e0 − e1 ⊗ (He0 ) = 0
92
and similar for E, F . For 0 ⊕ e0 , 0 ⊕ e1 , 0 ⊕ e2 we check
F.0 ⊕ e0 = 0 ⊕ e1 , F.0 ⊕ e1 = 0 ⊕ e2 , F.0 ⊕ e2 = 0 ,
and
F.e0 ⊗ e0 = (F e0 ) ⊗ e0 + e0 ⊗ (F e0 ) = e1 ⊗ e0 + e0 ⊗ e1
F.(e0 ⊗ e1 + e1 ⊗ e0 ) = 2e1 ⊗ e1
F.e1 ⊗ e1 = 0
So that ϕ(F.0 ⊕ ek ) = F.ϕ(0 ⊕ ek ). The same calculation can be repeated for
H and E. For example,
ϕ(H.(0 ⊕ e2 )) = ϕ(−2(0 ⊕ e2 )) = −4e1 ⊗ e1
and
H.ϕ(0 ⊕ e2 ) = H.(2e1 ⊗ e1 ) = 2(−e1 ⊗ e1 − e1 ⊗ e1 ) .
or
ϕ(E.(0 ⊕ e1 )) = ϕ(2(0 ⊕ e0 )) = 2e0 ⊗ e0
and
E.ϕ(0 ⊕ e1 ) = E.(e0 ⊗ e1 + e1 ⊗ e0 ) = e0 ⊗ e0 + e0 ⊗ e0 .

(i) By definition, a sub-vector space h ⊂ g is a Lie subalgebra if and only if for
all A, B ∈ h also [A, B] ∈ h.
Suppose [h, h] ⊂ h. Then for all A, B ∈ h also [A, B] ∈ h, so that h is a Lie
subalgebra.
Suppose that h is a Lie subalgebra. Then for all A, B ∈ h also [A, B] ∈ h.
Since h is a vector space, also all linear combinations of elements of the form
[A, B] are in h. Thus [h, h] ⊂ h.
(ii) For an ideal h ⊂ g we have [h, g] ⊂ h. Since [h, h] ⊂ [h, g], in particular
[h, h] ⊂ h. By part (i), h is a Lie subalgebra.
(iii) Let y ∈ ker(ϕ). Then for all x ∈ g,
ϕ([x, y]) = [ϕ(x), ϕ(y)] = [ϕ(x), 0] = 0 ,
so that also [x, y] ∈ ker(ϕ). Thus [g, ker(ϕ)] ⊂ ker(ϕ), i.e. ker(ϕ) is an ideal.
(iv) Let h = [g, g]. Then
[g, h] = [g, [g, g]] ⊂ [g, g] = h ,
i.e. h is an ideal.
(v) Let y ∈ h ∩ h0 . Then for any x ∈ g we have [x, y] ∈ h because h is an ideal,
and [x, y] ∈ h0 because h0 is an ideal. Thus [x, y] ∈ h ∩ h0 for all x ∈ g, i.e. h ∩ h0
is an ideal.
By definition, all elements of g/h are equivalence classes of the form x + h, for
93
x ∈ g. Since π(x) = x+h, π is surjective. Further π(x) = 0 ⇔ x+h = h ⇔ x ∈ h
thus ker(π) = h. Further, π is a homomorphism of Lie algebras because for all
x, y ∈ g, [π(x), π(y)] = [x + h, y + h] = [x, y] + h = π([x, y]).
By exercise 4.14(iii), ker(ϕ) is an ideal. Since g is simple, ker(ϕ) is either {0}
or equal to g. Thus ϕ is either injective or identically zero.
(i) Let E ∗ , F ∗ , H ∗ ∈ sl(2, C)∗ be the dual basis, i.e. E ∗ (E) = 1, E ∗ (F ) = 0,
E ∗ (H) = 0, etc. Then for all x, y ∈ sl(2, C)
κ(x, y) = tr(adx ady ) = E ∗ ([x, [y, E]]) + F ∗ ([x, [y, F ]]) + H ∗ ([x, [y, H]]) .
Computing all combinations results in
κ(E, E) = E ∗ ([E, [E, E]]) + F ∗ ([E, [E, F ]]) + H ∗ ([E, [E, H]])
= 0 + F ∗ ([E, H]) + H ∗ ([E, −2E]) = 0
κ(E, F ) = E ∗ ([E, [F, E]]) + F ∗ ([E, [F, F ]]) + H ∗ ([E, [F, H]])
= E ∗ ([E, −H]) + 0 + H ∗ ([E, 2F ]) = 2 + 0 + 2 = 4
κ(E, H) = E ∗ ([E, [H, E]]) + F ∗ ([E, [H, F ]]) + H ∗ ([E, [H, H]]) = 0
κ(F, F ) = E ∗ ([F, [F, E]]) + F ∗ ([F, [F, F ]]) + H ∗ ([F, [F, H]]) = 0
κ(F, H) = E ∗ ([F, [H, E]]) + F ∗ ([F, [H, F ]]) + H ∗ ([F, [H, H]]) = 0
κ(H, H) = E ∗ ([H, [H, E]]) + F ∗ ([H, [H, F ]]) + H ∗ ([H, [H, H]])
= E ∗ ([H, 2E]) + F ∗ ([H, −2F ]) + 0 = 4 + 4
The relation κ(x, y) = 4 Tr(xy) can also be verified by straightforward com-

putation. For example, HH is the identity matrix, so that Tr(HH) = 2 and
4 Tr(HH) = κ(H, H).
(ii) The non-vanishing commutators of M ≡ M01 , P0 and P1 are (recall that in
two dimensions, η = diag(1, −1))
[M01 , P0 ] = η10 P0 − η00 P1 = −P1 , [M01 , P1 ] = η11 P0 − η01 P1 = −P0 .
94
Denoting the dual basis by M ∗ , P0∗ , P1∗ , by the same method as above we find
κ(M, M ) = M ∗ ([M, [M, M ]]) + P0∗ ([M, [M, P0 ]]) + P1∗ ([M, [M, P1 ]])
= 0 + P0∗ ([M, −P1 ]) + P1∗ ([M, −P0 ]) = 2
κ(M, P0 ) = M ∗ ([M, [P0 , M ]]) + P0∗ ([M, [P0 , P0 ]]) + P1∗ ([M, [P0 , P1 ]]) = 0
κ(M, P1 ) = M ∗ ([M, [P1 , M ]]) + P0∗ ([M, [P1 , P0 ]]) + P1∗ ([M, [P1 , P1 ]]) = 0
κ(P0 , P0 ) = M ∗ ([P0 , [P0 , M ]]) + P0∗ ([P0 , [P0 , P0 ]]) + P1∗ ([P0 , [P0 , P1 ]]) = 0
It follows that κ is degenerate, since e.g. κ(x, P0 ) = 0 for all x ∈ p(1, 1).
(i) It is enough to verify the identity for x = Eab and y = Ecd . Use
[Eij , Ekl ] = δkj Eil − δil Ekj .
to check
[Eab , [Ecd , Eij ]] = δbc δdi Eaj + δad δcj Eib − δaj δdi Ecb − δbi δcj Ead .
One then computes

n
X
∗

κ(Eab , Ecd ) = Eij [Eab , [Ecd , Eij ]]
i,j=1
n
X
= δbc δdi δai + δad δcj δbj − δaj δdi δci δbj − δbi δcj δai δdj
i,j=1
= 2nδad δbc − 2δab δcd .
On the other hand,
Tr(Eab ) = δab , Tr(Ecd ) = δcd , Tr(Eab Ecd ) = δad δbc ,
from which the claim follows.

(ii) This is an immediate consequence of (i) together with the fact that the
matrices in sl(n, C) are trace-less.
We have to show that if y ∈ h⊥ , then for all x ∈ g also [x, y] ∈ h⊥ . This can be
seen by direct computation. Let z ∈ h be arbitrary. Then
κg ([x, y], z) = −κg ([y, x], z) = −κg (y, [x, z]) = 0 ,
95
since [x, z] ∈ h (h is an ideal) and κg (y, p) = 0 for any p ∈ h (as y ∈ h⊥ ). Thus
h⊥ is an ideal.
Let m = dim(h) and n = dim(g). Let {v1 , . . . , vm } be a basis of h and let
vm+1 , . . . , vn s.t. {v1 , . . . , vn } is a basis of g. Let vk∗ ∈ g ∗ , k = 1, . . . , n denote
the dual basis. For any a ∈ h and x ∈ g,
m
X n
X
vi∗ [x, [a, vi ]] + vj∗ [x, [a, vj ]]

κ(x, a) =
i=1 j=m+1
The first sum is zero, because for i = 1, . . . , m both vi and a are in h so that
[a, vi ] = 0. The second sum is zero, because [a, vj ] is in h (as h is an ideal),
and hence also [x, [a, vj ]] ∈ h; but for j = m + 1, . . . , n we have vj (b) = 0 for all
b ∈ h. Thus κ(x, a) = 0 for all x ∈ g and hence κ is degenerate.
(i) (i)
Let vα , α = 1, . . . , ni be a basis of gi . Then the set {vα | i = 1, . . . , n , α =
1, . . . , ni } provides a basis of g. Writing out the trace of the Killing form gives
ni
n X
X
(vα(i) )∗ [x, [y, vα(i) ]] .

κg (x, y) =
i=1 α=1
By definition of the Lie bracket on a direct sum of Lie algebras, if a ∈ gi and

b ∈ gj for i 6= j, then [a, b] = 0 in g. Thus
[x, [y, vα(i) ]] = [xi , [yi , vα(i) ]] .
This gives the desired result,

X ni
n X n n
X X
κg (x, y) = (vα(i) )∗ [xi , [yi , vα(i) ]] = trgi (adxi adyi ) = κgi (xi , yi ) .
i=1 α=1 i=1 i=1

Define a linear map f : g → g ∗ via f (x) = κ(x, · ). Suppose f (x) = 0 for some
x ∈ g. This means that (f (x))(y) = 0 for all y ∈ g, i.e. that κ(x, y) = 0 for
all y ∈ g. Since κ is non-degenerate, this implies x = 0. Thus f is injective.
But dim(g) = dim(g ∗ ) and so f is also surjective. The last statement is a
consequence of the rank-nullity theorem of linear algebra (→wikipedia) which
states that for a linear map F : V → W between to finite-dimensional vector
spaces V and W we have
dim(im F ) + dim(ker F ) = dim(V ) .
Since f is surjective, every linear map g → K is of the form κ(x, · ) for an

appropriate x. In particular, every linear map h → K can be written as κ(x, · )
for an appropriate x. Thus we get a surjective map f˜ : g → h∗ , also given by
96
f (x) = κ(x, · ), but now the second argument of κ has to be taken from h. For
the kernel of f˜ one finds
ker f˜ = x ∈ g κ(x, y) = 0 for all y ∈ h = h⊥ .

Furthermore, since f˜ is surjective, im f˜ = h∗ . Applying the rank-nullity theorem

to f˜ then gives dim(h∗ ) + dim(h⊥ ) = dim(g). Since dim(h∗ ) = dim(h), this
proves the statement.
We will use the basis Mab , Pa for p(1, n−1) as in exercises 3.14 and 3.16. Let
h = spanC (Pa |a = 0, . . . , n−1). From the commutation relations of p(1, n−1)
one reads off immediately that [u, v] = 0 for any u, v ∈ h (as [Pa , Pb ] = 0)
and [x, u] ∈ h for any x ∈ p(1, n−1) and u ∈ h (as [Pa , Pb ] = 0 ∈ h and
[Mab , Pc ] ∈ h). Thus h is an abelian ideal and by lemma 4.22, the Killing form
of p(1, n−1) is degenerate. This implies that p(1, n−1) is not semi-simple.
(i) Recall that by definition of the dual of a linear map, we have that for some
F : g → g, the dual F ∗ : g ∗ → g ∗ is given by [F ∗ ϕ](y) = ϕ(F y) where ϕ ∈ g ∗
and y ∈ g.
We have to show that for all x ∈ g,
fB adx = −ad∗x fB .
Both sides are maps from g to g ∗ . For any y, z ∈ g we compute, with F = −adx
and ϕ = fB (y),
[−ad∗x fB (y)](z) = [F ∗ ϕ](z) = ϕ(F z) = [fB (y)](−adx z)
= B(y, −[x, z]) = −B(y, [x, z]) = −B([y, x], z) = B([x, y], z) .
Furthermore,
[fB adx y](z) = fB ([x, y])(z) = B([x, y], z) .
Thus fB is indeed an intertwiner.
(ii) An invariant subspace h of (g, ad) has the property that for all x ∈ g, a ∈ h
one has ad(x)a ∈ h, i.e. that [x, a] ∈ h. Thus h is an ideal of g. Since g is simple,
its only ideals are {0} and g. Therefore the only invariant subspaces of (g, ad)
are {0} and g and hence the adjoint representation of a simple Lie algebra is
irreducible.
(iii) Since κ is also an invariant bilinear form, fκ : g → g ∗ is an intertwiner from
(g, ad) to (g, ad)+ . Further, κ is non-degenerate, so that fκ is invertible (and
hence in particular nonzero). Suppose that also fB is nonzero. Since (g, ad) to
(g, ad)+ are irreducible, by Schur’s lemma any two nonzero intertwiners between
them are multiples of each other, fB = λfκ for some λ ∈ C× . If fB is zero, then
fB = λfκ with λ = 0.
(iv) Since by part (iii) fB = λfκ , in particular [fB (y)](z) = [λfκ (y)](z) for all
y, z ∈ g, i.e. B(y, z) = λκ(y, z). Thus B = λκ.
97
Have X
adT a (T b ) = [T a , T b ] = f abc T c
c
and X XX
adT a ( vd T d ) = ( M (T a )cd vd )T c .
d c d
Comparing the two (set vd = δdb ) shows
M (T a )cb = f abc .

Let n = dim(V ). p
Pick any nonzero vector u ∈ V . If b(u, u) 6= 0 set v1 = u/ b(u, u) and go to
the next step. If b(u, u) = 0 choose a vector v ∈ V such thatpb(u, v) 6= 0 (which
exists since b is non-degenerate). If b(v, v) 6= 0 set v1 = v/ b(v, v) and go to
the next step. If b(v, v) = 0 then x = u + v has
b(x, x) = b(u, u) + 2b(u, v) + b(v, v) = 2b(u, v) 6= 0 .

p
Set v1 = x/ b(x, x).
We found v1 ∈ V with b(v1 , v1 ) = 1. Choose u2 , . . . , un s.t. {v1 , u2 , . . . , un } is
a basis of V . Set
wk = uk − b(uk , v1 )v1 , k = 2, . . . , n .
Then b(v1 , wk ) = b(v1 , uk ) − b(v1 , uk )b(v1 , v1 ) = 0. The set {v1 , w2 , . . . , wn } is

still a basis of V . Let W = spanC (w2 , . . . , wn ). Since b(v1 , w) = 0 for all w ∈ W ,
b restricted to W also has to be non-degenerate. Now start again at step 1 with
W instead of V .
We have seen that
f abc = κ(T c , [T a , T b ]) .
Clearly this is anti-symmetric in a and b, f abc = −f bac . For b and c note
f abc = κ(T c , [T a , T b ]) = κ([T c , T a ], T b ) = −κ(T b , [T a , T c ]) = −f acb .
But then also for a and c,
f abc = −f acb = f cab = −f cba .

Recall from exercise 4.17 that in the standard basis H, E, F of sl(2, C) one has
κ(E, F ) = 4, κ(H, H) = 8 and zero for all other pairings. We can thus choose
T1 = 1
√
2 2
H , T2 = 1
√
2 2
(E + F) , T3 = 1
√
2 2
(E − F) .
98
Then, e.g.
1 1
κ(T 2 , T 2 ) = κ(E + F, E + F ) = 2κ(E, F ) = 1 .
8 8

We have to verify properties (i)–(iii) of a Cartan subalgebra. Let
n
X n
X
h={ λk Ekk |λk ∈ C, λk = 0}
k=1 k=1
be the diagonal
Pn matrices in sl(n, C).
(i) Let H = k=1 λk Ekk ∈ h. Then
n
X
adH Eij = [H, Eij ] = λk (Ekk Eij − Eij Ekk ) = (λi − λj )Eij
k=1
so that adH acts diagonal on all Eij . Since

M
sl(n, C) = h ⊕ CEij
i6=j
we have found a basis in which adH is diagonal for all H ∈ h.

(ii) Diagonal matrices commute. Therefore [h, h] = 0.
(iii) To show maximality of h it is sufficient to show that [x, H] = 0 for all H ∈ h
implies xP∈ h (i.e. h is already maximal as an abelian subalgebra). To this end
n
let H = k=1 λk Ekk ∈ h with λi 6= λj for i 6= j. (This always exists, e.g. λk = k
Pn−1 P
for k = 1, . . . , n − 1 and λn = − k=1 λk .) Let M = i,j aij Eij ∈ sl(n, C).
Then X X
[H, M ] = aij [H, Eij ] = aij (λi − λj )Eij .
i,j i,j
Since λi 6= λj for i 6= j the commutator [H, M ] can only be zero if aij = 0 for
i 6= j. But then M ∈ h.
Let f1 , . . . , fm be a basis of F . By assumption, m < dim(V ∗ ) = dim(V ).
Define the a linear map A : V → Cm via A(v) = (f1 (v), . . . , fm (v)). By the
rank-nullity theorem of linear algebra, dim(im A) + dim(ker A) = dim(V ). But
dim(im A) ≤ dim(Cm ) = m and so dim(ker A) ≥ dim(V ) − m > 0. Choose any
nonzero v ∈ ker A. Then fk (v) = 0 for k = 1, . . . , m and since the fk are a
basis, ϕ(v) = 0 for all ϕ ∈ F .
Suppose h is a Cartan subalgebra. Then (1) is part of the definition of a
Cartan subalgebra, (2) was proved in the lecture, and for (3) note that, if for
some x ∈ g one has [x, a] = 0 for all a ∈ h, then by definition x ∈ g0 . We have
proved in the lecture that g0 = h, so that indeed x ∈ h.
99
Suppose h ⊂ g is a sub-vector space that obeys (1)–(3). Since by (2), κ re-
stricted to h is non-degenerate we can find a basis H i , i = 1, . . . , dim(h) of h
such that κ(H i , H j ) = δij . In particular κ(H i , H i ) 6= 0 so that (as proved in the
lecture) H i is ad-diagonalisable. Since [h, h] = {0}, all adH i , i = 1, . . . , dim(h)
can be simultaneously diagonalised. But then also each linear combination of
the H i is ad-diagonalisable. Thus h consists entirely of ad-diagonalisable ele-
ments and therefore obeys condition (i) in the definition of a Cartan subalgebra.
Condition (ii) is just property (1), and condition (iii) is implied by (3) (since
(iii) is stronger than (3)).
By definition, H γ is the unique element of g0 such that κ(H γ , x) = γ(x) for all
x ∈ g0 . In particular κ(H γ , H i ) = γ(H i ). On the other hand we can always
write
Xr
Hγ = ci H i for some ci ∈ C .
i=1
Because κ(H i , H j ) = δij it follows that ci = κ(H i , H γ ) and hence ci = γ(H i ).

Furthermore,
r
X r
X
(γ, ϕ) = κ(H γ , H ϕ ) = γ(H i )ϕ(H j )κ(H i , H j ) = γ(H i )ϕ(H i ) .
i,j=1 i=1
Finally, by definition of H γ ,
(γ, ϕ) = κ(H γ , H ϕ ) = γ(H ϕ ) .

We have shown in the lecture that for α ∈ Φ, (α, α) 6= 0, so that ϕ is well-defined.
We compute explicitly
2
[ϕ(E), ϕ(F )] = [e, f ] = κ(e, f )H α = H α = ϕ(H) = ϕ([E, F ])
(α, α)
2 2
[ϕ(H), ϕ(E)] = [H α , e] = α(H α )e = 2e = ϕ([H, E])
(α, α) (α, α)
2 2
[ϕ(H), ϕ(F )] = [H α , f ] = (−α(H α ))f = −2f = ϕ([H, F ])
(α, α) (α, α)
Thus ϕ is a homomorphism of Lie algebras. Finally, e, f, H α ∈ g are linearly
independent, so that ϕ is injective.
(i) Since α ∈ L we have n+ ≥ 1. As shown in the lecture, if α ∈ Φ, then also
−α ∈ Φ. Thus n− ≤ −1.
(ii) Abbreviate R ≡ Rϕ . We have already seen in the lecture that R(E) : gγ →
gγ+α and R(F ) : gγ → gγ−α . Further, R(H) : gγ → gγ . Thus clearly
M
Ũ = g0 ⊕ gmα
m∈L
100
is an invariant subspace of g. Now, for x ∈ gα and y ∈ g−α we have, by a lemma
shown in the lecture,
R(E)y = [e, y] = κ(e, y)H α and R(F )x = [f, x] = −[x, f ] = −κ(x, f )H α .
Further, R(H)H α = 0, so that indeed also U is an invariant subspace of g.
(iii) Have, for z ∈ gmα ,
2 2 2
R(H)z = [H α , z] = (mα)(H α )z = m(α, α)z = 2mz .
(α, α) (α, α) (α, α)
(iv) Clearly, V is a subspace of U . It is also the image of the Lie algebra

homomorphism ϕ : sl(2, C) → g and therefore, for given a ∈ V let y ∈ g such
that ϕ(y) = a. Then for all x ∈ (2, C),
R(x)a = [ϕ(x), a] = [ϕ(x), ϕ(y)] = ϕ([x, y]) ∈ V .
Further, to check that (V, Rϕ ) ∼= V3 as sl(2, C)-representations, first note that
ϕ : sl(2, C) → V is invertible, and that ϕ in fact gives an isomorphism from the
adjoint representation (sl(2, C), ad) to (V, Rϕ ) (verifying this just amounts to
the equation spelled out above). Finally, the adjoint representation of sl(2, C) is
isomorphic to V3 (via the intertwiner f : sl(2, C) → V3 , f (E) = v0 , f (H) = v1 ,
f (2F ) = v2 , check!).
(v) Since v0 ∈ gn+ α , by part (iii) we have R(H)v0 = 2n+ v0 . Since by con-
struction, gmα = {0} for all m > n+ we have R(E)v0 = 0. By the analysis
of sl(2, C)-representations in section 4.2, R(F )k v0 6= 0 for k = 0, . . . , 2n+ and
R(F )2n+ +1 v0 = 0. Since spanC (R(F )k v0 |k ∈ Z≥0 ) closes under the action of
R(E), R(F ) and R(H), it follows that W is an invariant subspace of U .
(vi) Both, V and W are invariant subspaces of the representation (U, Rϕ ) of
sl(2, C). The intersection of two invariant subspaces is again an invariant sub-
space, since for all x ∈ sl(2, C) and a ∈ V ∩ W we have Rϕ (x)a ∈ V (as V is
an invariant subspace) and Rϕ (x)a ∈ W (as W is an invariant subspace), i.e.
ϕ (x)a ∈ V ∩ W .
Next, clearly H α ∈ V . As for W , note that vn+ = R(F )n+ v0 is nonzero, and
obeys R(H)vn+ = (2n+ − 2n+ )vn+ = 0, so that vn+ = λH α for some nonzero
λ ∈ C.
But as (V, Rϕ ) is an irreducible representation, any invariant subspace is
either {0} or V . Since V ∩ W is not empty, we have V ∩ W = V . Similarly,
since (W, Rϕ ) is irreducible, we have V ∩ W = W . Thus V = W .
(i) We will show that h = spanC (H 1 , . . . , H r ) satisfies properties (i)–(iii) of def-
inition 5.3. By definition the H i are ad-diagonalisable, and so every H ∈ h is
ad-diagonalisable. Furthermore, again by definition, h is abelian. Thus proper-
ties (i) and (ii) hold. Now suppose there exists a K ∈ g such that spanC (h∪{K})
is a Cartan subalgebra. Use the Cartan-Weyl basis to write
r
X X
K= xj H j + yα E α , xi , yα ∈ C ,
j=1 α∈Φ
101
Pick an α ∈ Φ. To show that yα = 0 pick an H ∈ h such that α(H) 6= 0 (this
exists because α 6= 0). Then the condition that K commutes with H implies
X X
0 = [H, K] = yα [H, E α ] = yα α(H)E α .
α∈Φ α∈Φ
Since the E α are linearly independent, we see that yα = 0. Repeating the

argument for all α ∈ Φ shows that all yα = 0 and so K ∈ h. This establishes
property (iii) and thus h is a Cartan subalgebra of g.
(ii) We have
κ(E α , E −α ) = 1 α α
(α,α) κ([H , E ], E
−α
) = 1 α α
(α,α) κ(H , [E , E
−α
])
2 α α 2
= (α,α)2 κ(H , H ) = (α,α) .

(i) We will show that h = h1 ⊕ h2 satisfies properties (1)–(3) of exercise 5.7.
Property (1) is clear, as [hi , hj ] = {0} for i, j = 1, 2.
Property (2) can be seen as follows. κg restricted to g1 is equal to κg1 . Further,
κg1 restricted to the Cartan subalgebra h1 is non-degenerate. Therefore κg
restricted to h1 is non-degenerate. In the same way one sees that κg restricted
to h2 is non-degenerate. Thus, κg restricted to h1 ⊕ h2 is non-degenerate.
For property (3), let x = x1 ⊕ x2 and suppose that [x, c] = 0 for all c ∈ h.
Then in particular, for all a ∈ h1 , 0 = [x1 ⊕ x2 , a ⊕ 0] = [x1 , a] ⊕ 0. Thus x1 ∈ h1
as h1 is a Cartan subalgebra. Similarly it follows that x2 ∈ h2 . Thus x ∈ h.
This establishes that h1 ⊕ h2 is a Cartan subalgebra of g1 ⊕ g2 .
(ii) Let
{H1i |i = 1, . . . , r1 } ∪ {E1α |α ∈ Φ(g1 , h1 )}
and
{H2i |i = 1, . . . , r2 } ∪ {E2α |α ∈ Φ(g2 , h2 )}
be Cartan-Weyl bases of g1 and g2 , respectively. Then
( (
i
H 1 ⊕ 0 ; i = 1, . . . , r 1 E1α ⊕ 0 ; γ = α ⊕ 0 ∈ Φ1
H̃ i = , Ẽ γ =
0 ⊕ H2i−r1
; i = r1 +1, . . . , r1 +r2 0 ⊕ E2β ; γ = 0 ⊕ β ∈ Φ2
is a basis of g. Let a ⊕ b ∈ h be arbitrary. Then
[a ⊕ b, H1i ⊕ 0] = 0 , [a ⊕ b, 0 ⊕ H2j ] = 0 ,
[a ⊕ b, E1α ⊕ 0] = [a, E1α ] ⊕ 0 = (α(a)E1α ) ⊕ 0 = (α ⊕ 0)(a ⊕ b) · (E1α ⊕ 0) ,
[a ⊕ b, 0 ⊕ E2β ] = 0 ⊕ [b, E2β ] = 0 ⊕ (β(b)E2β ) = (0 ⊕ β)(a ⊕ b) · (0 ⊕ E2β ) .
This show that all elements of Φ1 ∪ Φ2 are indeed roots. It also shows that there
are no more roots, because all basis elements Ẽ γ are already accounted for by
102
the roots γ ∈ Φ1 ∪ Φ2 . Thus Φ(g, h) = Φ1 ∪ Φ2 .
(iii) By definition,
(λ, µ) = κg (H λ , H µ )
where H λ is defined by
λ(a) = κg (H λ , a) for all a ∈ h .
If λ = λ1 ⊕ 0 ∈ Φ1 we obtain in this way
H λ = H1λ1 ⊕ 0 ,
since for any a ⊕ b ∈ h,
κg (H λ , a ⊕ b) = κg (H1λ1 ⊕ 0, a ⊕ b) = κg1 (H1λ1 , a) = λ1 (a) = λ(a ⊕ b) .
Similarly, if µ = 0 ⊕ µ2 ∈ Φ2 we get
H µ = 0 ⊕ H2µ2 .
Thus, for λ ∈ Φ1 and µ ∈ Φ2 ,
κg (H λ , H µ ) = κg (H1λ1 ⊕ 0, 0 ⊕ H2µ2 ) = 0 .

We have g1 6= {0} as Φ1 6= {0} and g1 6= g as Φ2 6= {0}. Thus if g1 is
an ideal of g it is a proper ideal of g. Let h1 = spanC (H α |α ∈ Φ1 } and
h2 = spanC (H α |α ∈ Φ2 }. We have to show [g1 , g] ⊂ g. Now
M M M
CE γ , g0 ⊕ CE α ⊕ CE β

[g1 , g] = h1 ⊕
γ∈Φ1 α∈Φ1 β∈Φ2
= spanC [h1 , g0 ], [h1 , CE α ], [h1 , CE β ], [CE γ , g0 ],

[CE γ , CE α ], [CE γ , CE β ] α, γ ∈ Φ1 , β ∈ Φ2 .

Next we show that all terms enterning the span are already in g1 , so that the
span itself is in g1 .
[h1 , g0 ] = {0}.
[h1 , CE α ] ⊂ CE α as h1 ⊂ g0 and [H, E α ] = α(H)E α for all H ∈ g0 .
[h1 , CE β ] = 0 as any element of h1 can be written as a linear combination of
the H α , α ∈ Φ1 , and [H α , E β ] = β(H α )E β = (α, β)E β = 0 (by assumption,
(α, β) = 0 for α ∈ Φ1 and β ∈ Φ2 ).
[CE γ , g0 ] ⊂ CE γ as [H, E γ ] = α(H)E γ for all H ∈ g0 .
[CE γ , CE α ] ⊂ g1 , since as shown in the lecture, for γ, α ∈ Φ1 , γ + α ∈ / Φ2 .
this leaves γ = −α, in which case [E −α , E α ] = − (α,α)
2
H α ∈ g1 , or γ + α ∈ Φ1 , in
which case [E γ , E α ] = Nγ,α E γ+α ∈ g1 , or γ + α ∈
/ Φ, in which case [E γ , E α ] =
0 ∈ g1 .
[CE γ , CE β ] = {0} since we have seen in the lecture that if γ ∈ Φ1 and β ∈ Φ2 ,
103
then γ + β is not a root.
Altogether we see that [g1 , g] ⊂ g such that g1 is a proper ideal of g.
Solution for exercise 5.14: P P
(i) Let λ, µ ∈ R. We can then write λ = α∈Φ cα α, µ = α∈Φ dα α with
coeffiecients cα , dα ∈ R. Further,
X X
(λ, µ) = (λ, γ)(µ, γ) = cα dβ (α, γ)(β, γ) .
γ∈Φ α,β,γ∈Φ
All the factors on the rhs are real, thus (λ, µ) ∈ R, i.e. (·, ·) takes real values on
R. Further, it is positive definite, since for λ ∈ R nonzero,
X
(λ, λ) = (λ, γ)2 > 0,
γ∈Φ
as (λ, γ) cannot be zero for all γ ∈ Φ. [Otherwise λ = 0 since (·, ·) is non-

degenerate on g0∗ and Φ spans g0∗ over C.] Thus (λ, λ) > 0 for all λ ∈ R and
(λ, λ) = 0 ⇔ λ = 0, and hence (·, ·) is positive definite on R.
(ii) The Gram-Schmidt procedure can be used to produce from a set of vectors
{v1 , . . . , vm } in a real vector space V an orthonormal basis of spanR (v1 , . . . , vm )
with respect to a positive-definite symmetric bilinear form g(·, ·). It proceeds
recursively (compare to the solution p to exercise 5.2).
Start with v1 . Set u1 = v1 / g(v1 , v1 ) and uk = vk − g(u1 , vk )u1 for k =
2, . . . , m. Then, again for k = 2, . . . , m, g(u1 , uk ) = g(u1 , vk )−g(u1 , vk )g(u1 , u1 ) =
0.
Let V 0 = spanR (u2 , . . . , um ). Since g is non-degenerate, and g(u1 , v) = 0 for
all v ∈ V 0 , the restriction of g to V 0 is also non-degenerate. We can thus iterate
the procedure, using the vector space V 0 with basis {u2 , . . . , um } as a starting
point for the previous step.
Now pick a basis {v1 , . . . , vm } of R (over R). Use the Gram-Schmidt proce-
dure to obtain an orthonormal basis {ε1 , . . . , εm } of R. Since {ε1 , . . . , εm } is a
basis P of R over R, we know that if for any set of real coefficients c1 , . . . , cm such
m
that i=1 ci εi = 0 then we have ci = 0 for i = 1, . . . , m. NowPsuppose that
m
we have a set of complex coefficients λ1 , . . . , λm such that in g0∗ , i=1 λi εi = 0.
Taking the inner product with εi on both sides then shows that λi = 0 for
i = 1, . . . , m. Thus the {ε1 , . . . , εm } are linear independent also in g0∗ (over C).
Further, since all roots α ∈ Φ can be expressed as linear combinations of the εi ,
we have spanC (ε1 , . . . , εm ) = g0∗ . Thus {ε1 , . . . , εm } is a basis of g0∗ . As g0∗ has
dimension r, this shows also that m = r.
(iii) Set H i = H εi , i.e. the H i are defined via εi (a) = κ(H i , a) for all a ∈ g0 .
Then {H 1 , . . . , H r } is a basis of g0 because {ε1 , . . . , εr } is a basis of g0∗ , and for
all α ∈ Φ,
α(H i ) = α(H εi ) = (α, εi ) ∈ R
as both, α and εi lie in R.
104
Solution for exercise P5.15:
n
(i) Note that αkl (H) = i=1 ai (ωk − ωl )(Eii ) = ak − al . Then
n
X
[H, Ekl ] = ai [Eii , Ekl ] = (ak − al )Ekl = αkl (H) Ekl .
i=1
For k 6= l we have αkl 6= 0 and so αkl is a root.

(ii) Write a general H ∈ h as in (i). Then
n
X
κ(H αkl , H) = 2nTr(H αkl H) =

ai Tr(Ekk Eii ) − Tr(Ell Eii )
i=1
= ak − al = αkl (H) ,
as required.
(iii) By definition
(αkl , αkl ) = κ(H αkl , H αkl ) = 2nTr(H αkl H αkl )

1
Tr (Ekk −Ell )(Ekk −Ell ) = n1 ,

= 2n
and
[Ekl , Elk ] = Ekk − Ell = 2
(αkl ,αkl ) H αkl .
(iv) Follows from (i)–(iii).
(v) Compute as in part (iii),
(α12 , α23 ) = 6 Tr(H α12 H α23 ) = 16 Tr (E11 −E22 )(E22 −E33 ) = − 16 ,

(α12 , α13 ) = 16 Tr (E11 −E22 )(E11 −E33 ) = 16 ,

(α23 , α13 ) = 61 Tr (E22 −E33 )(E11 −E33 ) = 16 .

Together with (α12 , α12 ) = (α23 , α23 ) = (α13 , α13 ) = 1/3 this shows that the
angle between α12 and α23 is 120◦ , that the angle between α12 and α13 is 60◦ ,
etc. Together with αkl = −αlk this gives the root diagram as shown.
All identities follow from direct substitution into the definition
(α,λ)
sα (λ) = λ − 2 (α,α) α .
For example, for any λ ∈ g0∗ ,

(1)
(α,λ) (α,λ)
sα (sα (λ)) = sα λ − 2 (α,α) α = sα (λ) − 2 (α,α) sα (α)
(2) (α,λ) (α,λ)
= λ − 2 (α,α) α − 2 (α,α) (−α) = λ ,
where we used (1) that sα is a linear map and (2) that sα (α) = −α. It follows
that sα ◦ sα = id.
We have proved in the lecture that a Weyl reflection maps a root to a root.
Since an element of the Weyl group W is a finite composition of Weyl reflections,
we have
w(α) ∈ Φ for all w ∈ W , α ∈ Φ .
Let us number the roots in Φ by 1, 2, . . . , m (with m = dim(g) − r),
Φ = {α1 , α2 , . . . , αm } .
Let w ∈ W . For k = 1, . . . , m we know that w(αk ) is a root. Denote the number

of that root by πw (k). Then πw defines a map from the set {1, . . . , m} to itself.
Furthermore,
αk = w−1 (w(αk )) = w−1 (απw (k) ) = απw−1 (πw (k))
which shows that πw−1 (πw (k)) = k, i.e. πw is invertible and hence a permutation:
πw ∈ Sm .
Suppose that for w, x ∈ W we have πx = πw . Then, for k = 1, . . . , m,
w(αk ) = απw (k) = απx (k) = x(αk ) .
Since the roots span g0∗ , a linear map from g0∗ to g0∗ is uniquely determined once
we know its action on Φ. This implies that w = x.
The above argument shows that we obtain an injective map from W to Sm .
But Sm has a finite number of elements (in fact, m!), and hence also W can
only have a finite number of elements.
The decomposition of g into root spaces (simultaneous eigenspaces of elements
of the Cartan subalgebra) is
M
g = g0 ⊕ gα .
α∈Φ
We have proved that dim(gα ) = 1 for all α ∈ Φ. Further, g0 is the Cartan

subalgebra, and its dimension is by definition the rank. Thus
dim(g) = r + |Φ| ,
where |Φ| denote the number of elements in Φ. For the examples treated in
section 5.6 this gives:
θm 90◦ 60◦ 45◦ 30◦

dim(g) 6 8 10 14

(i) By definition, no root lies in the hyperplane H = {v ∈ R|(v, n) = 0}. Thus
(n, α) 6= 0 for all α ∈ Φ. Hence either (α, n) > 0 or (α, n) < 0 and therefore
106
each root is an element either of Φ+ or else of Φ− .
(ii) If α ∈ Φ+ , then by definition (α, n) > 0, where n is the normal defining
the hyperplane. This is equivalent to (−α, n) < 0, i.e. −α ∈ Φ− . Thus of each
pair of roots {α, −α} one element lies in Φ+ and one in Φ− . This implies that
|Φ+ | = |Φ− |.
(iii) Since Φ− = {−α|α ∈ Φ+ }, it is immediate that spanR (Φ+ ) = spanR (Φ+ ∪
Φ− ), which together with part (i) implies the claim.
(i) We have proved in the lecture that for any two roots α, β ∈ Φ we have
2(α, β)
∈Z .
(α, α)
This holds in particular for simple roots, and hence Aij ∈ Z.

(i)
,α(i) )
(ii) First, Aii = 2(α
(α(i) ,α(i) )
= 2 and second, if i 6= j then we have stated (but
not proved) in the lecture that (α(i) , α(j) ) ≤ 0. Since in addition (α, α) > 0 for
2(α(i) ,α(j) )
all α ∈ Φ we have, again for i 6= j, Aij = (α(j) ,α(j) )
≤ 0.
(i) (j)
(iii) By assumption i 6= j, so that α and α are not colinear (or they would
not form a basis). Then, by the Cauchy-Schwartz inequality,
(α(i) , α(j) )2 < (α(i) , α(i) )(α(j) , α(j) ) ,
or, equivalently, Aij Aji < 4. Since both Aij ≤ 0 and Aji ≤ 0 it follows that
Aij Aji ≥ 0. Since further Aij , Aji ∈ Z, also their product is an integer. This
only leaves the possibilites Aij Aji ∈ {0, 1, 2, 3}.
We will show that if g is not simple, then the Dynkin diagram is not connected.
Suppose thus that g is not simple. We have seen in the lecture, that in this
case we can decompose the root system Φ of g as Φ = Φ1 ∪ Φ2 with both
Φ1 and Φ2 not empty, and such that (α, β) = 0 for all α ∈ Φ1 and β ∈ Φ2 .
Set R = spanR (Φ) as well as R1 = spanR (Φ1 ) and R2 = spanR (Φ2 ). Then
R1 ∩ R2 = {0} (as any element λ in the intersection has to obey (λ, λ) = 0).
Let α(1) , . . . , α(r) be a choice of simple roots for g. Since the simple roots span
R (over R), they cannot all lie in R1 . Let us number the simple roots such
that α(1) , . . . , α(s) ∈ Φ1 and α(s+1) , . . . , α(r) ∈ Φ2 , with 1 ≤ s ≤ r − 1. Since
(α(i) , α(j) ) = 0 for i ≤ s and j > s, there is no line joining a vertex {1, . . . , s}
to a vertex {s + 1, . . . , r}. Thus the Dynkin diagram is not connected.
For each of the four cases, below a choice of hyperplane and the resulting set of
simple roots are given. From this the Cartan matrix and the Dynkin diagram
are computed.
107
θm = 90◦ :
n
α(2)
α(1)
Thus (α(1) , α(2) ) = 0 and
A11 A12

2 0
A= =
A21 A22 0 2
is the Cartan matrix. Consequently the Dynkin diagram is

e e
1 2
θm = 60◦ : This is treated in detail in section 5.7 of the lecture notes.

θm = 45◦ :
α(2) n
α(1)
Thus (α(2) , α(2) ) = 2 (longest root must have length squared 2) and (α(1) , α(1) ) =
1 (ratio of length squared is 2). The coordinates are then α(1) = (1, 0) and
α(2) = (−1, 1), so that (α(1) , α(2) ) = −1. Inserting this in the definition of the
Cartan matrix we find A12 = −1 and A21 = −2. Thus

2 −1
A=
−2 2
ei e
2 1
which is the Dynkin diagram of B2 (with a different convention for labelling

simple roots as compared to theorem 5.21, but the labelling is just a choice and
not part of the diagram).
108
θm = 30◦ :
n
α(2)
α(1)
Thus (α(2) , α(2) ) = 2 (longest root must have length squared 2) and (α(1) , α(1) ) =
2/3 (ratio of length squared is 3). We also know that the angle between α(1)
and α(2) is 150◦ , so that
q q √
(α(1) , α(2) ) = cos(150◦ ) · |α(1) | · |α(2) | = − 34 · 23 · 2 = −1
Inserting this in the definition of the Cartan matrix we find A12 = −1 and
A21 = −3. Thus
2 −1
A=
−3 2
ei e
2 1
which is the Dynkin diagram of G2 (with a different convention for labelling

simple roots as compared to theorem 5.21).
Using the numbering of nodes used in the lecture, as well as the standard con-
vention for matrices
 11
A12 A13 A14

A
 A21 A22 A23 A24 
A=  A31

A32 A33 A34 
A41 A42 A43 A44
109
we find
   
2 −1 0 0 2 −1 0 0
 −1 2 −1 0   −1 2 −1 0 
A(A4 ) =   , A(B4 ) =   ,
 0 −1 2 −1   0 −1 2 −2 
0 0 −1 2 0 0 −1 2
   
2 −1 0 0 2 −1 0 0
 −1 2 −1 0   −1 2 −1 −1 
A(C4 ) =   , A(D4 ) =   ,
 0 −1 2 −1   0 −1 2 0 
0 0 −2 2 0 −1 0 2
 
2 −1 0 0
 −1 2 −2 0 
A(F4 ) =   .
 0 −1 2 −1 
0 0 −1 2

We will start by defining the ‘real part’ of an element in VC . Let
X = spanC ((λ, v)|λ ∈ C, v ∈ V ) ,
so that VC = X/W with W as defined in the lecture. Define a function R̃ : X →

V as follows. For
u = λ1 (µ1 , v1 ) + · · · + λm (µm , vm )
with vk ∈ V and λk , µk ∈ C, set
R̃(u) = Re(λ1 µ1 )v1 + Re(λm µm )vm .
One can check that, for u, u0 ∈ X and r ∈ R,
R̃(u + u0 ) = R̃(u) + R̃(u0 ) and R̃(ru) = rR̃(u) .
Note further that, for any µ ∈ C,

R̃ µ(λ, r1 v1 + r2 v2 ) − µλr1 (1, v1 ) − µλr2 (1, v2 )
= Re(µλ)(r1 v1 + r2 v2 ) − Re(µλr1 )v1 − Re(µλr2 )v2 = 0 .
By definition, every element of W can be written as a sum of vectors of the

form as in the argument of R̃. Since R̃(u + u0 ) = R̃(u) + R̃(u0 ) it follows that
R̃(w) = 0 for all w ∈ W . In particular R̃(u + w) = R̃(u) for all u ∈ X, w ∈ W .
Thus it it possible to define a map R : VC → V as R(x + W ) = R̃(x), since the
rhs does not depened on the the choice of representative x of the equivalence
class x + W .
Suppose (1, a) + i(1, b) + W = 0 + W . Then also

0 = R (1, a) + i(1, b) + W = Re(1)a + Re(i)b = a
110
and, since also i((1, a) + i(1, b) + W ) = 0 + W ,

0 = R i(1, a) − (1, b) + W = −b ,
i.e. a = 0 and b = 0.
Any element u ∈ VC can be written as a finite sum
u = λ1 (µ1 , v1 ) + · · · + λm (µm , vm ) + W
Using the definition of W , we can rewrite u as
u = (λ1 µ1 , v1 ) + · · · + (λm µm , vm ) + W .
Suppose that λk µk = xk + iyk with xk , yk ∈ R. Then we can further rewrite
u = (x1 + iy1 , v1 ) + · · · + (xm + iym , vm ) + W
= x1 (1, v1 ) + · · · + xm (1, vm ) + iy1 (1, v1 ) + · · · + iym (1, vm ) + W
= (1, x1 v1 + · · · + xm vm ) + i(1, y1 v1 + · · · + ym vm ) + W
which is of the form (1, a) + i(1, b) + W . It remains to show that a and b are
unique, i.e. that if (1, a) + i(1, b) + W = (1, a0 ) + i(1, b0 ) + W , then a = a0 and
b = b0 . But this follows since we have seen that (1, a−a0 )+i(1, b−b0 )+W = 0+W
implies a − a0 = 0 and b − b0 = 0.
Let B = {(1, v1 ) + W, . . . , (1, vn ) + W }.
B spans VC : Let u ∈ VC . By exercise 5.24 we can write u = (1, a)+i(1, b)+W .
Since {v1 , . . . , vn } is a basis of V , we can write a = a1 v1 + · + an vn and b =
b1 v1 + · + bn vn for some ak , bk ∈ R. Thus
u = (1, a) + i(1, b) + W
= a1 (1, v1 ) + · · · + an (1, vn ) + ib1 (1, v1 ) + · · · + ibn (1, vn ) + W ,
which shows that B spans VC (over C).

The elements of B are linearly independent: Let λk ∈ C be such that
λ1 (1, v1 ) + · · · + λn (1, vn ) + W = 0 + W .
We have to show that this implies that all λk are zero. Write λk = xk + iyk
with xk , yk ∈ R. Then
n
X
0+W = xk (1, vk ) + iyk (1, vk ) + W
k=1
= (1, x1 v1 + · · · + xn vn ) + i(1, y1 v1 + · · · + yn vn ) + W
Since the decomposition of an element of VC as (1, a) + i(1, b) + W is unique,

and (1, 0) + i(1, 0) + W = 0 + W , it follows that x1 v1 + · · · + xn vn = 0 and
111
y1 v1 + · · · + yn vn = 0. But the {v1 , . . . , vn } are linearly independent. This
implies that xk = 0 and yk = 0 for k = 1, . . . , n.
(1)⇒(2): Let ϕ : hC → g be an isomorphism of complex Lie algebras. Let
n = dim(h) (over R) and let T a , a = 1, . . . , n be a basis of h. Then T a,
a = 1, . . . , n is a basis of hC (recall our shorthand notation v ≡ (1, v) + W ). Set
T̃ a = ϕ(T a ). Since ϕ is an isomorphism, T̃ a , a = 1, . . . , n is a basis of g. Let
f ab
c be the structure constants of h,
n
X
[T a , T b ] = f ab
c T
c
.
c=1
Then also
[T̃ a , T̃ b ] = [ϕ(T a ), ϕ(T b )] = ϕ([T a , T b ])
n
X n
X n
X
= ϕ( f ab c
c T )= f ab c
c ϕ(T ) = f ab
c T̃
c
.
c=1 c=1 c=1
(2)⇒(1): Define a linear map ϕ : hC → g on the given basis as ϕ(T a ) =

T̃ a . Since ϕ maps a basis to a basis it is an isomorphism of vector spaces.
Furthermore
n
X
[ϕ(T a ), ϕ(T b )] = [T̃ a , T̃ b ] = f ab
c T̃
c
c=1
n
X n
X
= f ab c
c ϕ(T ) = ϕ( f ab c a b
c T ) = ϕ([T , T ])
c=1 c=1
so that ϕ is also an isomorphism of Lie algebras.

Abbreviate κ ≡ κsu(2) . Recall from exercise 3.11 that a basis of su(2) is given
in terms of the Pauli matrices as {T 1 , T 2P
, T 3 } with T a = iσa , a = 1, 2, 3. The
a b
commutation relations are [T , T ] = −2 c εabc T c . The Killing form evaluated
on T a and T b reads
3
X 3
X
κ(T a , T b ) = (T c )∗ ([T a , [T b , T c ]]) = (−2εbcd )(−2εadc )
c=1 c,d=1
3
X
=4 (δbc δca − δba δcc ) = 4(δab − 3δba δcc ) = −8δa,b .
c=1
This implies that κ(x, x) < 0 for all nonzero x ∈ su(2).

Abbreviate κ ≡ κsl(2,R) . Consider the following basis of sl(2, R):

1 0 0 1 0 0
H= , E= , F = .
0 −1 0 0 1 0
112
Then a calculation identical to that of exercise 4.17 shows that
κ(E, E) = 0 , κ(E, H) = 0 , κ(E, F ) = 4 ,
κ(H, H) = 8 , κ(H, F ) = 0 , κ(F, F ) = 0 .
In particular, κ(H, H) > 0, so that the Killing form of sl(2, R) is not negative
definite. (As an aside, note that κ(E − F, E − F ) = −2κ(E, F ) = −8, so that
κsl(2,R) is not positive-definite either.)
Let ϕ : g → h be an isomorphism of Lie algebras, and let {T a } be a basis of
g. Then {T̃ a } with T̃ a = ϕ(T a ) is a basis of h. One computes, for all x, y ∈ g,
X
κh (ϕ(x), ϕ(y)) = (T̃ c )∗ ([ϕ(x), [ϕ(y), T̃ c ]])
X c X
= (T c )∗ (ϕ−1 ([ϕ(x), [ϕ(y), ϕ(T c )]])) = (T c )∗ ([x, [y, T c ]]) = κg (x, y) .
c c
In particular, a real Lie algebra with negative definite Killing form cannot be
isomorphic to a Lie algebra which contains an element H with κ(H, H) > 0.
First convince yourself that
so(n)C = {M ∈ Mat(n, C) | M t + M = 0} and
o(1, n−1)C = {M ∈ Mat(n, C) | M t J + JM = 0} .
Let A be the diagonal matrix A = diag(1, i, . . . , i). It has the properties
AJA = 1 , JA−1 = A = A−1 J .
Define a linear map ϕ : Mat(n, C) → Mat(n, C) as ϕ(M ) = A−1 M A and note

that ϕ is an isomorphism (what is the inverse?). Furthermore it is easy to see
that it preserves commutators (check!),
ϕ([M, N ]) = [ϕ(M ), ϕ(N )] .
Now
M t + M = 0 ⇔ AM t A + AM A = 0 ⇔ AM t A−1 J + JA−1 M A = 0
⇔ ϕ(M )t J + Jϕ(M ) = 0
Thus M ∈ so(n)C if and only if ϕ(M ) ∈ o(1, n−1)C . Therefore, ϕ also provides
an isomorphism so(n)C → o(1, n−1)C . We have already checked that it is
compatible with the Lie bracket.
113
Index
abelian, for Lie algebras, 33 group, definition, 4
action, of a Lie algebra, 24
ad-diagonalisable, 38 homomorphism, of Lie algebras, 21
adjoint representation, 25
Aut(G), 5 ideal, of a Lie algebra, 32
automorphism of a group, 5 iff, 3
image, of an intertwiner, 26
Baker-Campbell-Hausdorff identity, 23 injective map, 5
bijective map, 5 intertwiner, of representations, 25
invariant bilinear form, 37
Cartan matrix, 54 invariant subspace, 26
Cartan subalgebra, 39 irreducible representation, 26
Cartan-Weyl basis, 44 isomorphic, for groups, 5
commutator, 16 isomorphic, for representations, 25
complexification, of a real Lie alg., 57 isomorphism, of Lie algebras, 21
complexification, of a real vect.sp., 56
connected, for Dynkin diagrams, 55 Jacobi identity, 16
dimension, of a matrix Lie group, 16 kernel, of an intertwiner, 26

direct product, of groups, 8 Killing form, 34
direct sum, of Lie algebras, 31
direct sum, of representations, 31 Lie algebra, definition, 15
direct sum, of vector spaces, 29 Lie algebra, of a matrix Lie group, 16
dual representation, 25 Lie bracket, 15
Dynkin diagram, 54 Lie group, definition, 3
Lie subalgebra, 16
E(n), 7 Lorentz group, 8
E(n)ij , 19
εijk , 20 Mat(n, K), 5
End(E), 16 matrix Lie group, definition, 12
endomorphism, of a vector space, 16 Minkowski space, 8
euclidean group, 7 module, of a Lie algebra, 24
exponential of a matrix, 14
negative root, 53
faithful, for representations, 25
O(1, n−1), 8
fssc, 37
o(1, n−1), 22
general linear group, 5 O(n), 6
gl(E), 16 o(n), 18
gl(n, K), 16 orthogonal group, 6
GL(n, R), 5 orthogonal transformation, 6
GL(V ), 12
P (1, n−1), 8
group homomorphism, 5
p(1, n−1), 22
group of rotations of R3 , 6
114
Pauli matrices, 20 unitary group, 13
Poincaré group, 8
Poincaré algebra, 23 Weyl group, 50
positive root, 53 Weyl reflection, 49
proper ideal, 32
rank, of a Lie algebra, 39

real form, of a complex Lie algebra, 57
reductive, for Lie algebras, 33
representation map, 24
representation space, 24
representation, of a Lie algebra, 24
root diagram, 48
root space, 47
root system, 40
root, of a Lie algebra, 40
Schur’s Lemma, 26
semi-simple, for Lie algebras, 33
semidirect product, 8
simple root, 53
simple, for Lie algebras, 33
SL(n, K), 14
sl(n, K), 18
SO(n), 6
so(n), 18
SP (2n), 17
sp(2n), 18
special linear group, 14
special orthogonal group, 6
special unitary group, 13
structure constants, of a Lie alg., 37
SU (n), 13
su(n), 18
sub-representation, 26
subgroup, 7
surjective map, 5
symmetry, of physical system, 4
symplectic group, 17
tensor product, of representations, 31

tensor product, of vector spaces, 29
trivial representation, 25
U (n), 13
u(n), 18
115

Lie Groups and Lie Algebras - Ingo Runkel

Uploaded by

Copyright:

Available Formats

Lie Groups and Lie Algebras - Ingo Runkel

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lie Groups and Lie Algebras - Ingo Runkel

Uploaded by

Copyright:

Available Formats

What is the topic of the document?

What is the topic of the document?

What are Lie groups and Lie algebras?

What are Lie groups and Lie algebras?

version 1.6d — 24.11.

3 Matrix Lie Groups and their Lie Algebras 12

5 Classification of finite-dimensional, semi-simple, complex Lie

A Appendix: Collected exercises 63

Symmetries are an important concept in physics. Recent theories are almost

2.1 Definition of a group

A linear map T : Rn → Rn is an orthogonal transformation iff

g(T u, T v) = g(u, v) for all u, v ∈ Rn . (2.4)

Denote by ei = (0, 0, . . . , 1, . . . , 0) the i’th basis vector of Rn , so that in com-

Hence T is an orthogonal transformation iff T t T = 1 (here 1 denotes the unit

T t T = 1 ⇒ det(T t T ) = det(1) ⇒ det(T )2 = 1 ⇒ det(T ) ∈ {±1} (2.6)

The orthogonal transformations T with det(T ) = 1 preserve orientation. These

O(n) = {M ∈ Mat(n, R)|M t M = 1} (2.7)

with group multiplication given by matrix multiplication.

2.3 Lorentz and Poincaré transformations

2.4 (*) Symmetries in quantum mechanics

[One should worry if the infinite sum converges. For finite-dimensional H it

|(U ψ, U ψ 0 )|2 = |(ψ, U † U ψ 0 )|2 = |(ψ, ψ 0 )|2 . (2.20)

Further, they commute with time-evolution

(Utrans (s)ψ)(x) = ψ(q + s) . (2.22)

So what is the infinitesimal generator of translations? Take s to be small to find

and where εijk is antisymmetric in all indices and ε123 = 1.

(Urot (θ)ψ)(q1 , q2 , q3 ) = ψ(q1 cos θ − q2 sin θ, q2 cos θ + q1 sin θ, q3 ) (2.26)

and check that infinitesimally

Urot (θ) = 1 + iθL3 + O(θ2 ) . (2.27)

(ii) Using [qr , ps ] = iδrs (check!) verify the commutator

This turns V into a complex Lie algebra:

3 Matrix Lie Groups and their Lie Algebras

Then the unitary group U (n) is given by

SL(n, K) = {A ∈ Mat(n, K)| det(A) = 1} . (3.9)

3.2 The exponential map

(iii) Recall that a complex n × n matrix A can always be brought to Jordan

where each Jordan block is of the form

In particular, if all Jordan blocks have size 1, the matrix A is diagonalisable.

3.3 The Lie algebra of a matrix Lie group

[A, B] = AB − BA ∈ End(E) . (3.12)

g = {A ∈ Mat(n, K)| exp(tA) ∈ G for all t ∈ R} . (3.13)

and the Commutator Formula,

GL(n, R) gl(n, R) = Mat(n, R) n2

Let us verify this list.

det(exp(sA)) = es tr(A) (3.14)

1 = (1 + sAt )(1 + sA) + O(s2 ) = 1 + s(At + A) + O(s2 ) , (3.15)

which shows that At + A = 0 is a necessary condition for A to be in the Lie

(exp(sA))t exp(sA) = exp(sAt ) exp(sA) = exp(−sA) exp(sA) = 1 . (3.16)

In components, the condition At + A = 0 implies Aii = 0 and Aij = −Aji . Thus

1 = det(exp(sA)) = es tr(A) (3.17)

to see that further tr(A) = 0 is necessary and sufficient. However, A + At = 0

3.5 Examples: SO(3) and SU (2)

ϕ([a, b]) = [ϕ(a), ϕ(b)] for all a, b ∈ g .

(ii) A Lie algebra homomorphism ϕ is a Lie algebra isomorphism iff it is invert-

3.6 Example: Lorentz group and Poincaré group

O(1, n−1) = {M ∈ GL(n, R)|M t JM = J} (3.22)

η(x, y) = x0 y0 − x1 y1 − · · · − xn−1 yn−1 (3.23)

ηkl = η(ek , el ) = Jkl . (3.24)