romanSoQM 2015

Symmetries of Quantum Mechanics
Roman Zwicky, University of Edinburgh
March 26, 2015

Contents
1 Preliminaries 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Basic mathematical notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Mini-revision of linear algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Some properties of matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Group theory 10
2.1 Basics of group theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Group presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Notions of group theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Cosets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Group action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.3 Normal subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.4 Direct and semidirect products of groups . . . . . . . . . . . . . . . . . . . . . 21
2.3 Permutation groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 The symmetric permutation group Sn . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.2 The alternating group An . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Platonic solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3 Representation theory 29
3.1 Representations the pedestrian way . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Basic definitions of representation theory . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Basics of representation theory Maschke’s thm, Schur’s Lemma and decomposability . 32
4 Representation theory of finite groups 34

4.1 Character theory for finite groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Character tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2.1 Character table of S3 as an illustration . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Restriction to a subgroup – branching rules . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.1 Complex and real representations . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.2 Branching rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
i
CONTENTS ii
4.4 Constructing representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.4.1 A short notice on the irreducible representations of Sn . . . . . . . . . . . . . . 47
4.4.2 Direct product representations – Kronecker product . . . . . . . . . . . . . . . 49
4.5 A few sample applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5.1 Distortion of lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5.2 Fermi and bose statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5 Lie groups: U (1) ' SO(2), SO(3) ' SU (2)/Z2 56

5.1 Basic definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.2 The two abelian groups: U (1) and SO(2) . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3 Some generalities of (non-abelian) simple compact Lie groups . . . . . . . . . . . . . . 63
5.4 The two non abelian groups: SU (2) and SO(3) . . . . . . . . . . . . . . . . . . . . . . 66
5.4.1 Lie Algebras of SO(3) and SU (2) . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.4.2 Irreducible representations of SU (2) (and hence SO(3)) . . . . . . . . . . . . . 68
5.4.3 Clebsch-Gordan series of SU (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6 Applications to quantum mechanics 74

6.1 A short synopsis of quantum mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.2 SU (2)-group theory and the physics of angular momentum . . . . . . . . . . . . . . . 77
6.3 Selection rules I: Wigner-Eckart theorem for SU (2) (and SO(3)) . . . . . . . . . . . . 80
6.4 Wigner-Eckart applied: the Landé g-factor . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.5 Spherical harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.6 Selection rules II: parity and superselection . . . . . . . . . . . . . . . . . . . . . . . . 90
6.7 Wigner-Eckart applied: electric dipole selection rules . . . . . . . . . . . . . . . . . . . 91
6.8 Pauli’s Hydrogen atom and SU (2) ⊗ SU (2)-symmetry . . . . . . . . . . . . . . . . . . 92
Bibliography
[Arm88] M.A. Armstrong. Groups and Symmetry. Springer Verlag, New York, USA, 1988.
Undergraduate text on groups theory with focus in finite groups.
[Ful91] W.Fulton and J.Harris Representation theory - A first course. Springer Verlag, Graduate text
in mathematics, New York, USA, 1991.
Graduate text. More advanced. Pedagogical text, especially beginning of the book has a nice
discussion on representation theory of finite groups. For the student who wants to know more.
[Gil08] R.Gilmore Lie Groups, Physics, and Geometry: An Introduction for Physicists, Engineers
and Chemists. Cambdridge University Press 2008
An inspiring book with lots of good examples.
[Tun85] Wu-Ki Tung. Group theory in physics. World Scientific, Singapore, 1985.
A widely used textbook for the subject.
[Atk08] Peter Atkins and Julio de Paula Atkins’ Physics Chemistry. Oxford University Press, ninth
edition , 2010
Efficient use of group theory.
[Ham62] Morton Hammermesh Group Theory and its Application to Physical Problems. Dover Pub-
lications Inc. New York, 1962
A readable classic on the subject.
[Corn97] J.F. Cornwell Group Theory in Physics. Academic Press, 1997
A widely used textbook.
[Ell79] J.P.Elliott and P.G.Dawber Symmetry in Physics. MacMillan press LTD, 1979
Very well known among particle physicists.
[Geo99] H.Georgi Lie Algebras in Particle Physics. Frontiers in Physics 2nd ed, 1999
Idem.
[Gre97] W.Greiner and B.Müller Quantum Mechanics: Symmetries Greiner’s book are usually ap-
preciated for their explicitly working out many examples. Springer 1997
Chapter 1
Preliminaries
1.1 Introduction
The theory of finite groups has historically arisen as an effort (by Galois) to classify different solutions
of algebraic equations. The theory of Lie Groups was founded by Lie in an attempt to classify solutions
of differential equations. Lie groups turned out to be continuous groups with a natural manifold
structure. Both types will be discussed in this course.
Generally group theory is a standard algebraic structure which applies in many fields of mathe-
matics and applied sciences. In physics, symmetries are naturally described by groups. The power
of symmetries relies on the fact that they, partially, solve the dynamics; the symmetry restricts the
solutions. This happens at the more sophisticated level of the celebrated Wigner-Eckart theorem (to
be discussed in these lectures) as well as in simple integrals where the symmetries of the integrand
restrict the form of the solutions.
An example is the following integral
Z
Iij = d3 kf (k · p, k 2 , p2 )ki kj = A(p2 )δij + B(p2 )pi pj , (1.1)
where ki , pi are vectors in three dimensional space and δij is the Kronecker symbol. The integral of a
type Iij is an integral that describes quantum correction, be it in quantum field theory or statistical
mechanics. No matter how complicated the function f is, the solution has the form on the right hand
side since the integrand is covariant under the three dimensional euclidian rotation group. If the
symmetry is not assumed then one ought to write 3 · 3 = 9 coefficients instead of the two coefficients
A and B above.
At the level of equations of motion, or more precisely the Lagrangian, symmetries are related
to conservation laws by virtue of the famous Noether theorem. Most famously: time symmetry
⇔ energy conservation, space translation symmetry ⇔ momentum conservation, rotational symmetry
⇔ spatial angular momentum conservation. The result can often be turned around in the sense that an
unexpected symmetry in the result suggests that a symmetry in the Lagrangian has been overlooked.
Groups are abstract objects in principle. When realised on vector spaces they are referred to
as representations. As quantum theory is described by Hilbert spaces which are specific, at times
1
CHAPTER 1. PRELIMINARIES 2
infinite dimensional, vector spaces representation theory plays a special rôle.1 An example, that you
should have seen in a course on quantum mechanics, is the Hydrogen atom for which the potential
V ∼ 1/r has spherical symmetry (symmetry under the rotation group). The energy solutions are
given by En ∼ 1/n2 which are in particular independent of the so-called third quantum numbers m
(z-projection of the angular momentum) which singles out a direction in the physical space. Once an
external magnetic field (an example of hyperfine splitting) is applied to the hydrogen atom, a direction
is singled out. In this case, as expected, the energy depends on the magnetic quantum number m.
The course consists essentially of four parts.2 An introduction into abstract group theory with
focus on finite groups in chapter 2, generic remarks on representation theory 3, representation theory
of finite groups and of continuous groups in chapters 4 and 5 respectively followed by applications
to quantum mechanics in chapter 6. On the subject of continuous groups special focus is given on
U (1) (the symmetry group of quantum electrodynamics which is associated with charge conservation),
SO(3) (the rotation group) and SU (2) (the rotation group of half integer spin objects e.g. the electron
if spin 1/2).
Proofs are usually given or at least sketched. The proofs aim to give you insight about the
theorems. More elaborate proofs which give only limited insight are not given but can looked up in
textbooks on group theory. Throughout the text there are plenty of footnotes. Almost all of them
could be omitted but they, hopefully, provide the reader with more insight on alternative views or a
broader scope on the topic. In my personal experience this leads to higher appreciation of the topic
in general. Comments on possible applications in physics, of the underlying mathematical topic, are
given in italics. Please do not hesitate to give me comments on these notes either in person or via
e-mail. Your effort is much appreciated, also by future students attending this course.
1.2 Basic mathematical notions

The aim of this section is to introduce the basic notions of mathematics used throughout the text.
1.2.1 Notation
Basic notation used throughout the course:
- The integer numbers Z ≡ {.., −2, −1, 0, 1, 2, ..}
- The rational numbers Q ≡ {p/q | p, q ∈ Z}
- The unit circle in n + 1 dimensions is denoted by S n = {~x ∈ Rn+1 | x21 + x22 + .. + x2n+1 = 1}.
- Mn (F ) corresponds to the n × n-matrices over the set F which is most often R or C in this
course. In denotes the unit n × n matrix.
- GL(n, C) = {A ∈ Mn (C) | det(A) 6= 0} sometimes also denoted by GL(V ) where dim(V ) = n
is an n-dimensional vector space.
1 Also for mathematics according to this quote: “all of mathematics is a form a representation theory ” (Israel
Gelfand (1913-2009)).
2 The section on basic mathematical notions 1.2 will only be briefly discussed and mainly serves the purpose of
having a common language. The basic revision of linear algebra will be discussed before starting with the chapter on
representation theory 3 where it is really needed. Abstract group theory as presented in chapter 2 is independent of
linear algebra.
- Einstein summation convention: when two identical indices (on the same side of the equation)
are left P
”open”, then they are meant to be summed over unless otherwise stated. For example
x∗i yi → i x∗i yi .
1.2.2 Definitions
• injective-, surjective- and bijective-maps. Illustrated in Fig. 1.1.
– A map ϕ : A → B is surjective if ∀b ∈ B, ∃a ∈ A s.t. ϕ(a) = b.
– A map ϕ : A → B is injective if ∀a1 , a2 ∈ A : ϕ(a1 ) = ϕ(a2 ) ⇒ a1 = a2 .
– A map ϕ : A → B is bijective if and only if it is injective and surjective. Bijective is also
known as one-to-one or invertible.
Figure 1.1: Injective (left), surjective (middle) and bijective (right) maps as explained in the text.
• The kernel of a map ϕ : A → B is the subset of A which maps to the trivial element of B.
(which could be 0 for a vector space or the identity for a group). The kernel measures the degree
to which ϕ is injective.
• The image of a map ϕ : A → B is ϕ(A) ⊆ B.
• The empty set is denoted by ∅.
• The Kronecker symbol δij is widely used throughout the mathematical literature:

1 i=j
δij = (1.2)
0 i 6= j
In some sense it is a unit matrix.

• The Levi-Civita tensor in n-indices is the completely antisymmetric tensor which is fully
defined by
123..n = 1 . (1.3)
Antisymmetry implies for instance a minus sign, 213..n = −1, when the first two indices are
permuted.
• An equivalence relation: is a relation between two elements, say a and b denoted by a ∼ b

which satisfies the following three properties
1. a ∼ a (reflexivity)
2. a ∼ b implies b ∼ a (symmetry)
3. a ∼ b and b ∼ c implies a ∼ c (transitivity)
The equivalence class of an element, say a, is the set of all elements which are equivalent to
a. It is denoted by [a] = {x | a ∼ x}. Any element of [a] is said to be a representative of the
equivalence class. An example for an equivalence relation is: two vectors being parallel.
1.3 Mini-revision of linear algebra

In this section we revise some notions from linear algebra which are particularly useful for the course.
• Vector space is a collection of elements (vectors) which add and are multiplied with objects of
a field. A field is an algebraic body with certain properties. For the purpose of the course it is
sufficient to think of the field either as the real numbers or the complex numbers. Let ~u, ~v and
~ be such vectors (think of them as vectors in Cn for example) and k and k 0 complex numbers
w
then the following properties define a vector space:
(1) (~u + v) + w
~ = ~u + (~v + w)
~ , (5) k(~u + ~v ) = k~u + k~v ,
(2) (~u + 0) = ~u , (6) (k + k 0 )~u = k~u + k 0 ~u ,
(3) ~u + (−~u) = 0 , (7) (kk 0 )~u = k(k 0 ~u) ,
(4) ~u + ~v = ~v + ~u , (8) 1~u = ~u . (1.4)
In particular the collection of vectors has a neutral element 0 (axiom (2)) and the field has a
unit operators 1 (axiom (8)).
• Hilbert space is a vector space (see above) with a scalar or inner product, denoted here by (, ),
from V × V → C which satisfies the following axioms:
(1) (~u, k~v + k 0 w)

~ = k(~u, ~v ) + k 0 (~u, w)
~ , linearity ,
(2) (~u, ~v ) = (~v , ~u)∗ , conjugation ,
(3) (~u, ~u) ≥ 0 , positivity , (1.5)
where the equality (3) holds P ⇔ ~u = 0. As an example we mention the vector space Cn over C
n
with scalar product: (~u, ~v ) = i=1 u∗i vi .
We ought to mention the bra and ket notation (introduced by P.A.M. Dirac and widely used
in physics) which amounts to the notational identification,
u ↔ |ui , (u, v) ↔ hu|vi , (1.6)
which we shall use at times when illustrating examples.

• A linear operator is an operator which is compatible with linearity:
A(k|ui + k 0 |vi) = kA|ui + k 0 A|vi . (1.7)
In a finite dimensional vector space all linear operators can be represented by matrices. This
will be of importance when discussing finite representation theory.
In order to illustrate the remaining notations we will consider a finite dimensional complex vector
space, denoted by V = CN where N = dim V . Then there are N linearly independent vectors
{|1 i, |2 i, ..., |N i}3 which form a basis in the sense that each |vi ∈ V can be written as
N
X
|vi = vi |i i , vi ∈ C . (1.8)
i=1
The inner product (, ) : V × V → C is written in the Dirac notation as follows: hv|wi ≡ (|vi, |wi).
The combination of a vector space and an inner product are also known as a Hilbert space.4 A basis
{|e1 i, |e2 i, ..., |eN i} orthonormal to the inner product can be chosen:
hei |ej i = δij . (1.9)

PN PN
The scalar product of two vectors |vi = i=1 vi |ei i and |wi = i=1 wi |ei i is given by:
N
X
hv|wi = vi∗ wi . (1.10)
i=1
An orthonormal set provides a decomposition of the identity

N
X
1N = |ei ihei | . (1.11)
i=1
PNj
A partition of the identity is a collection of projectors Pj ≡ i=1 |ei ihei | satisfying:
NP
Pj†
X
Pi Pj = δij Pi , = Pj , 1N = Pj . (1.12)
j=1
PNP
Note: N = j=1 Nj . A linear operator A : V → V (A(a|vi + b|wi) = a(A|vi) + b(B|wi) for a, b ∈ C)5
can be represented as a matrix 6 and its matrix elements are given by
Aij = hei |Aej i ≡ hei |A|ej i , (1.13)

3 We shall use Dirac’s bra and ket notation throughout this section.
4 Subtleties arise in the case of infinite dimensional vector spaces which are beyond the scope of this course.
5 An operator is anti-linear if A(a|vi + b|wi) = a∗ (A|vi) + b∗ (B|wi). The operator that changes t → −t, denoted by
T (with t being the time), is an example of an anti-linear operator.

6 In this section matrix stands for a square matrix.
where the notation on the right hand side is the one used in quantum mechanics (it is understood
that A acts on the right i.e. |ej i). Hence a matrix A may be written as:
N
X
A= Aij |ei ihej | . (1.14)
i,j=1
A fundamental fact is that the eigenvalue equation:
A|vi i = λi |vi i , (1.15)
admits N solutions since the characteristic polynomial P (t) = det(A − t1) = 0 admits N solutions by
virtue of the fundamental theorem of algebra. The λi are said to be eigenvalues and the |vi i are the
corresponding eigenvectors.
1.3.1 Some properties of matrices

• The transpose of a matrix is (AT )ij = Aji . A matrix is (anti)symmetric ⇔ AT = ±A (with
symmetric for plus sign).
• The hermitian conjugate of a matrix is (A† )ij = A∗ji . A matrix is hermitian ⇔ A† = A.
• The inverse of a matrix A−1 is a matrix that satisfies: A−1 A = AA−1 = 1N .

• A unitary matrix is a matrix whose hermitian conjugate is its inverse A† = A−1 ⇔ AA† =
A† A = 1 N .
P
• The trace of matrix is Tr[A] = hei |A|ei i where |ei i is an orthonormal basis.
• The determinant of a matrix is det A = i1 ....iN A1i1 ....AN iN where summation over the indices
ix is implied and 123..N = 1 is the completely antisymmetric Levi-Civita tensor. The definition,
which is not as simple as other definitions in this list, is given for completeness only. You should
be familiar with its basic properties given in section 1.3.1.
• A matrix is diagonal ⇔ Aij = 0 for i 6= j.

• A matrix A is block diagonal if it can be written as:

A1 0m×n
A= (1.16)
0n×m A2
where A1 and A2 are n × n and m × m matrices. The symbol 0i×j stands for a i × j matrix
with entries equal to zero.
• The direct sum of two vector spaces V ⊕ W has the following meaning: To each |vi i ∈ V and
|wa i ∈ W we associate a state |vi i ⊕ |wa i ≡ |vi ⊕ wa i (the last equality is non-standard notation)
and the inner product is extended to:
hvi ⊕ wa |vj ⊕ wb iV ⊕W = hvi |vj iV + hwa |wb iW (1.17)

The dimension is dim(V ⊕ W ) = v + w (v ≡ dim(V ) and w ≡ dim(W )) Given A ∈ GL(V ) and

B ∈ GL(W ) then
(A ⊕ B)|vi i ⊕ |wa i = (A|vi i) ⊕ (B|wa i) . (1.18)
Hence the direct sum may be thought of as a v + w-dimensional vector space where operators
are represented as (v + w) × (v + w) matrices in block diagonal form:

A 0v×w
A⊕B = (1.19)
0w×v B
The inverse problem, can a vector space be written as a direct sum, plays an important rôle in
discussing the irreducibility of a representation.
• The direct product of two vector spaces V ⊗ W has the following meaning: To each |vi i ∈ V
and |wa i ∈ W we associate a state a state |vi i⊗|wa i ≡ |vi ⊗wa i (the last equality is non-standard
notation) and the inner product is extended to:
hvi ⊗ wa |vj ⊗ wb iV ⊗W = hvi |vj iV · hwa |wb iW (1.20)
The dimension is dim(V ⊗ W ) = v · w. The direct product of the two operators from the
proceeding item acts on the space as follows:
(A ⊗ B)|vi i ⊗ |wa i = A|vi i ⊗ B|wa i . (1.21)
An example in quantum mechanics is the spatial wave function of the electron Ψ(x) times (⊗)
its spin 1/2 part. Direct products play a rôle in representation theory in terms of, what we shall
call, Kronecker products. They are widely used in physics. For example by taking Kronecker
products (tensor products of representations to be defined throughout the course) the so-called
fundamental representations all other representations of the group can be obtained.
Properties associated with definitions above:

1. For a hermitian matrix the eigenvalues are real (λi ∈ R) and the eigenvectors can be chosen to
be orthonormal. In this basis the matrix A is diagonal. For a two 2 × 2 hermitian matrix in the
orthonormal eigenbasis reads:

λ1 0
A= ≡ diag(λ1 , λ2 ) , (1.22)
0 λ2
with obvious generalisation to n × n-matrices.

It is straightforward to construct the explicit transformation matrix. Suppose A is a hermitian
and a non-diagonal matrix and λi and |vi i are its orthonormal eigenvectors, then
N
X
A= λl |vl ihvl | (1.23)
l=1
is diagonalised by the following unitary basis transformation7

N
X
diag(λ1 , ..., λN ) = V † AV , V = |vl ihel | . (1.24)
l=1
2. The inverse of a matrix A exists ⇔ A has no zero eigenvalues. For a hermitian matrix, the
inverse in the orthonormal eigenbasis reads:
−1
−1 λ1 0
A = . (1.25)
0 λ−1
2
PN
3. The trace of a matrix is the sum of its eigenvalues Tr[A] = i=1 λi .
4. The trace is cyclic Tr[AB] = Tr[BA].8

QN
5. The determinant of a matrix is the product of its eigenvalues det(A) = i=1 λi .
6. The determinant of a product of matrices is the product of its determinants.
det(AB) = det(A) · det(B)(= det(BA)) . (1.26)
7. The following relation holds:

hv|Awi = hA† v|wi (1.27)
This implies that for a unitary matrix U :
hU v|U wi = hv|wi . (1.28)
Hence unitary transformations are the natural objects to implement symmetry transformations
on Hilbert spaces since they preserve the inner product (probability measure).
Exercise 1.3.1 Basic linear algebra:

a) Make sure you understand all the definitions in the list given in this section and that the
properties 1) to 7) are familiar to you.
b) Show that SL(V ) = {A ∈ GL(V ) | det(A) = 1} is a group. The letter “S” stands for
special since SL(V ) is a special linear group.
c) Verify that the determinant is a group homomorphism: det : GL(V ) → C∗ .
7 In the case where A is not hermitian it can happen that the eigenvectors, because of linear dependence, cannot be
brought into a orthonormal form. In such cases a so-called Jordan normal form can be obtained with eigenvalues on
the diagonal and entries equal to one for A(i+1)(i−1) -elements on subspaces of degenerate eigenvalues. For more details
consider textbooks as this topic is beyond the scope of this course.
8 The trace is also linear and hA|Bi = Tr[A† B] is an inner product on the vector space of matrices.
d) Show that
hv|Awi∗ = hw|A† vi . (1.29)
e) As additional exercises you might want to show that:

– det(A−1 ) = det(A)−1 , det(aA) = aN det(A) where a ∈ C.
PN
– Tr[AT ] = Tr[A], Tr[An ] = i=1 λni .
– The trace and the determinant are invariants under similarity transformations.
– Properties of hermitian conjugation (aA)† = a∗ A† for a ∈ C, (A + B)† = A† + B † ,
(AB)† = B † A† , (A† )† = A, (a|αihβ|)† = a∗ |βihα|.
Chapter 2
Group theory
This chapter aims to familiarise the student with the basic notions of group theory. This part resembles
an undergraduate course on group theory in algebra except that the proofs are not as detailed and
that comments on applications in physics are inserted here and there. Most proofs are straightforward
from the viewpoint of the mathematician. The exceptions are Cauchy’s theorem and the theorem on
the alternating groups. In those cases previous results (isomorphism and orbit-stabiliser theorem)
give rise to more elegant proofs.
2.1 Basics of group theory

Let us start from the formal definition of a group .
Definition 2.1.1 Group:
Let G be a set on which a binary operation G × G → G is defined. G is called a group if and only if
(g1 · g2 ) · g3 = g1 · (g2 · g3 ) ∀g1 , g2 , g3 ∈ G , product rule is associative

e·g =g·e=g , ∃e ∈ G : ∀g ∈ G , there is an identity element in G.
−1 −1 −1
g·g =g ·g =e, ∀g ∈ G, ∃g ∈ G , every element of G has an inverse in G.
Exercise 2.1.1 To be discussed in the workshop

a) Show that (a · b)−1 = b−1 · a−1 .
Examples 2.1.1 (basic ones)

- The integers Z form a group with respect to the binary operation of addition, denoted by (Z, +).
The unit element is given by 0. The inverse of n is −n.
10
CHAPTER 2. GROUP THEORY 11
- The integers Z are not a group with respect to the binary operation of multiplication. First the
0 element 0 · Z = 0 does not comply with the group axioms. Taking Z∗ ≡ Z\{0} is not a group
either since the inverse of say 2, which is 1/2, is not contained in the set.
- The rational numbers Q∗ (with notation understood as above), the real numbers R∗ or the
complex numbers C∗ form a group with respect to the binary operation of multiplication, which
we denote by (Q∗ , ·) and so on.
- The group of permutation of n elements, denoted by Sn , forms a group.1 The permutation
groups will be discussed in a separate section 2.3.
- The group Zn (cyclic group of order n) with n elements e2πi(m/n) (0 ≤ m < n integer) forms a
group under multiplication.
- The rotation group of three dimensional space O(3) = {O ∈ M3 (R) | OT O = I3 }. More precisely
this corresponds to the so-called fundamental representation of O(3). This viewpoint will be
discussed at length in chapter 3 on representation theory.
The examples are concrete examples of structures that you might already have known. Below we
give a few basic definitions that help to characterise a group. This can be helpful for example as
specific statements, e.g. theorems, might only be valid for certain classes of groups.
Definition 2.1.2 Basic defintions:

• A group is called finite, discrete or continuous depending on whether the set G is finite,
discrete (isomorphic to Z) or uncountable.
• The order of a group G (denoted by |G|) is the number of elements of the group. Thus the
order of a group is a sensible concept for a finite group.
• The order of a group element g ∈ G is the smallest integer n for which g n = e.
• A group G is called abelian ⇔ ∀g1 , g2 ∈ G : g1 · g2 = g2 · g1 .2 The elements g1 and g2 are
said to commute with each other.
• A group H is called subgroup of G (denoted by H ⊂ G) ⇔ H is a group and its elements are
a subset of the elements of G. A proper subgroup is a subgroup which is not the group itself
nor the group with one element (often called the trivial subgroup).
• Two elements g1 , g2 ∈ G are conjugate to each other ⇔ ∃g ∈ G s.t. g1 = g · g2 · g −1 .
• The generators of a group are a set of elements from which all other elements can be generated
through the group multiplication law. A set of generators is not unique in general. In particular
the number of elements of a generating set might differ. The minimal number of elements that
generate a certain group is called the rank of a group.3 It is a fact that all groups which are of
rank 1 are isomorphic to the cyclic group Zn mentioned above.
1 You may think of rearranging n objects with different colours which are placed on one line. The group concept
seems very natural for investigating this object.

2 In honour of the Norwegian mathematician Niels Henrik Abel (1802-1829).
3 For Lie groups this notion is extended to the Lie Algebra which is often finite dimensional unlike the group itself.
• The center of a group is the subset of elements which commute with all elements of the group:
Z(G) = {z ∈ G | ∀g ∈ G : z · g = g · z}

a) Which groups in examples 2.1 are finite, discrete or continuous?
b) Give an element in O(3) (c.f. example 2.1 last one) which is of order 2. That is to say write
down a 3 × 3 matrix in O(3), other than the identity, which squares to one.
c) Give one example of a proper and continuous subgroup of O(3). Give an example of a
proper subgroup of Z4 .
d) Give all minimal set of generators of Z4 and Z7 using the representation e2πi(n/4) and
e2πi(n/7)
e) Show that conjugation within a group forms an equivalence relation. This fact will become
particularly important when discussing the representation theory of finite groups.
f) Show that for a group where each element, except the identity, is of order 2 is abelian.
Theorem 2.1.1 subgroup criteria

A subset H of a group G is a subgroup ⇔ g1 · g2−1 ∈ H for g1 , g2 ∈ H.
The proof of theorem 2.1.1 does not provide any insight and so we shall skip it.
Finite groups can be represented in terms of so-called multiplication tables which simply list
all possible multiplications of the groups in terms of a rows and columns system.
G e a b ..
e e a b ..
(2.1)
a a a2 a · b ..
.. .. .. ..
Examples of the cyclic group Z2 and the permutation group S3 are given in tables 2.1 and 2.2
respectively. The multiplications are such that column multiplies the row which is relevant in the case
where the two elements do not commute. One may therefore envisage the program of writing down all
multiplication tables at any finite order and verify that the result satisfies the group axioms. Whereas
in principle possible you might guess that this is not the most efficient nor elegant way to characterise
groups. In essence this amounts to finding the subset of permutations which is compatible with the
group axioms. In addition in each row and column one find each elements exactly once as otherwise
the inverse would not be unique. For example in the table if a2 = a · b then by the uniqueness of the
inverse of a we would conclude a = b which is clearly wrong. This simple fact is sometimes referred
to as the rearrangement theorem throughout the literature. This insight makes the following theorem
obvious:
Theorem 2.1.2 Cayley’s thm:
Any finite group is isomorphic (to be defined below) to a subgroup of the symmetric group Sn .
S3 () (1,2) (2,3) (1,3) (1,2,3) (1,3,2)

() () (1,2) (2,3) (1,3) (1,2,3) (1,3,2)
(1,2) (1,2) () (1,2,3) (1,3,2) (2,3) (1,3)
Z2 1 -1 (2,3) (2,3) (1,3,2) () (1,2,3) (1,3) (1,2)
1 1 -1 (1,3) (1,3) (1,2,3) (1,3,2) () (1,2) (2,3)
-1 -1 1 (1,2,3) (1,2,3) (1,3) (1,2) (2,3) (1,3,2) ()
(1,3,2) (1,3,2) (2,3) (1,3) (1,2) () (1,2,3)
Table 2.1: Cyclic group Z2 as
described in Examples 2.1. Table 2.2: The group of three permutation S3 . The element () denotes
the identity, (1, 2) permutes or cycles the first two elements and (1, 2, 3) is
understood as a cyclic permutation in the three elements. The operation
(a, b) is called a transposition
Remark: this theorem is reminiscent of the Whitney embedding theorem that states that any manifold
can be embedded into a euclidian space Rn for suitable n.
Exercise 2.1.3 Exercises on permutation groups and multiplication tables (more: section 2.3)
a) Show that S3 is a non-abelian group (by using the multiplication table 2.2 for example). It
is sufficient to find one example for which a · b 6= b · a.
b) What is the order of the group Sn ?
c) Show the group of four elements {e, g1 , g2 , g3 } (with |G| = 4), where all elements, except
the element e, are of order two is uniquely determined. Hint: Build up the multiplication
table. This group is known as the Vierergruppe: Z2 × Z2 (after Felix Klein).
Definition 2.1.3 group morphisms

• A group homomorphism is a map between two groups that respects the group structure.
More precisely let G1 and G2 be two groups then φ : G1 → G2 is a group homomorphism ⇔
∀a1 , b1 ∈ G1 : φ(a1 · b1 ) = φ(a1 ) · φ(b1 ) (2.2)
Note that the “·” on the left and right hand side operate on G1 and G2 respectively.
• A group isomorphism is a bijective group homomorphism (one-to-one correspondence or
invertible). Notation: G1 ' G2 means that the two groups G1 and G2 are isomorphic to each
other.
An isomorphism really means that two groups are the same and the classification of groups are
up to isomorphisms. We shall see that the symmetry groups of the platonic solids are isomorphic
to permutation groups. A major application of these notions is the first isomorphism theorem 2.2.8
which we are going to discuss once the notation of a normal subgroup has been introduced.

a) Two finite groups which admit an isomorphism, obviously, have the same number of ele-
ments. For continuous groups the issue is more subtle as is often the case. Can you give an
isomorphism between the integer number Z and the even integers 2Z?
b) Show that for g ∈ G, g → g −1 (inversion) defines an isomorphism ⇔ G is abelian.
2.1.1 Group presentation

One may define groups from generators and their relation in a completely algebraic way. More
precisely consider a set of generators {a, b, c, ..}, then the group is defined as all words (arrangement
of letters) that can be formed modulo certain constraints e.g. a2 = e, b3 = e, abc = e, ... This way
of defining a group is called the group presentation. The presentation, for a given group, is not
unique. For example S3 can be defined as:
S3 = < a, b, c | a2 = b2 = c3 = abc = e > (2.3)
or
S3 = < A, B | A2 = B 2 = (AB)3 = e > (2.4)
Presentations, at least for finite groups, will not play a major part in this course. Yet it seems good
practice to do the following exercise:

a) Identify candidates for generators a, b, c and A, B in Eqs. (2.3) and (2.4) (by using the
multiplication table 2.2) respectively.
2.2 Notions of group theory

2.2.1 Cosets
Definition 2.2.1 (left)-coset:
Given a subgroup H ⊂ G one can associate to each g ∈ G a left-coset gH = {g · h | h ∈ H}
Remarks and properties:

• one can define in completely analogous manner a right-coset. Below we shall continue to
reference left cosets only.
• Only for a certain subclass of groups, called normal (to be defined), one can define a group
multiplication on cosets. Hence cosets do in general not admit a group structure and are
therefore called cosets rather than “cogroups”.
Definition 2.2.2 partition
A partition of a set is a decomposition into (non-empty) subsets which do not intersect. In terms of
symbols: Pi ⊂ X form a partition of X if X = P1 ∪ P2 ∪ .. and Pi ∩ Pj = ∅ for i 6= j.
Theorem 2.2.1 (a) The left- gH and also the right-cosets Hg partition G. (b) the partition sets are
of equal size: |gi H| = |gj H| = |H|.
Proof: (a) We need to show that for g1 ∈ / g2 H, g1 H ∩g2 H = ∅. Ad absurdum: suppose g1 H ∩g2 H 6= ∅
then there ∃h1 , h2 ∈ H s.t. g1 · h1 = g2 · h2 , but then g1 = g2 · h2 · h−1
1 ∈ g2 H which contradicts our
original assumption. Therefore gi H partition G. (b) The map g1 H → g2 H: g1 · h → g2 · h for h ∈ H
is clearly invertible and thus |g1 H| = |g2 H|. The assertion for the right cosets is obviously true as all
steps in the proof follow in complete analogy. q.e.d.
Thus gH partition G in sets of equal size as illustrated in Fig. 2.1. From this observation follows the
Figure 2.1: The partition of G into its left-cosets gi H as written symbolically in Eq. (2.5).
famous Lagrange theorem:

Theorem 2.2.2 Lagrange’s thm:
If H is a subgroup of G (H ⊂ G) ⇒ |G|/|H| is an integer.
According to theorem 2.2.1 the integer corresponds to the number of non-identical set gH which
partition G. Summarising the most important content of this section in terms of a formula:
|G|/|H|
G = ∪i=1 gni H , (2.5)
with ni an appropriate map from integers to integers.4

The collection of all gi H is called the coset space and often denoted by G/H. We will see in section
2.2.3 that this space has a group structure ⇔ the group H has a property which is called normal. In
4 Note the converse problem of theorem 2.2.2: Is there a subgroup of order p if |G|/p is integer is related to the
so-called Sylow theorems and are one of the tools used to classify finite groups. The study thereof is beyond the scope
of this course.
particle physics and statistical mechanics cosets naturally arise when a continuous global symmetry G
is broken by the vacuum state to a subgroup H. Then one can formulate an effective theory of the low
lying spectrum on the coset space.
Exercise 2.2.1 Exercises on Cosets and Lagrange’s thm.

a) Why can Z7 not be a subgroup of S6 ? (you need a result from exercises 2.1.3).
b) Consider H = {e, (2, 3)} subgroup of S3 . Find the other partitions by choosing elements
which are not in H and then form the left and right cosets.
(Observe in particular that the right and left cosets are not equal to each other. This will
be relevant in a forthcoming exercise.)
2.2.2 Group action

Definition 2.2.3 Group action:
A group action is a map of the group on a set compatible with the group structure.
More precisely, let G be a group and X a set then φ : G × X → X, denoted by φ(g, x) → g.x for
short5 (with g ∈ G and x ∈ X), is a group action6 ⇔
(1) e.x = x , identity

(2) (g1 · g2 ).x = g1 .(g2 .x) ,compatibility (g1 , g2 ∈ G) .
Let us illustrate this concept with a few examples:
Examples 2.2.1 (basic ones)

• Trivial action g.x = x.
• Group acts on itself (X = G) by:
– left(right) multiplication g.x = g · x (g.x = x · g).

– conjugation g.x = g · x · g −1 .
• The group action on a linear space is called representation and plays a distinct rôle in physics.
For example the Hilbert space of quantum mechanics and quantum field theories are linear spaces.
In mathematical language particles corresponds to representations of the Lorentz groups and
internal symmetry groups.
• The permutation group Sn acts on the set {1, ..., n} by permutation.7
5 You must pay attention to distinguish the group multiplication ”·” from the group action ”.”.
6 An alternative definition (for finite groups): A group action is a group homomorphism of G into S|X| .
7 Or one may work in Rn and associate permutations of unit vectors e = (1, 0.., 0) ∈ Rn with the permutation group.
1
One would be tempted to think that this is a good start for the representation theory of Sn . We will see in chapter 3
that this representation is not the smallest building block (irreducible) of the representations of Sn .
• The group (Z, +) acts on the set R as follows: m ∈ Z and r ∈ R : m.r → (−1)m r.
In gauge theories such as electrodynamics the gauge symmetry, which is U (1) (its elements are eiφ(x)
and the binary operation is just the ordinary multiplication) for electrodynamics, acts on the gauge
potential Aµ (x) → Aµ (x) + ∂x∂ µ φ(x). All the terminology in this section can be found in the physics
literature. Some further definitions are useful for group actions:
Figure 2.3: The orbit-stabiliser theorem 2.2.3 states

Figure 2.2: Group acting on a point x ∈ X defines the that there is a one-to-one correspondence between each
corresponding orbit denoted by G(x) ⊆ X. point on the orbit (c.f. Fig. 2.2) and the left (right)
cosets gGx (Gx g) of the stabiliser subgroup.
Definition 2.2.4 further terminology

The orbit of x ∈ X, illustrated in Fig. 2.2, is the set of elements that are reached by the group
acting on x: G(x) = {g.x | g ∈ G} ⊂ X.89 The stabiliser (subgroup) 10 of x ∈ X is the subset in G
which leaves x invariant under the group action Gx = {g ∈ G | g.x = x}.
Examples 2.2.2 orbit of O(2)
• Let us consider the action of the two dimensional group O(2) = {O ∈ M2 (R) | OT O = I2 }
(analogous to O(3) in examples 2.1) on a two dimensional plane around a point a. Then the
orbit of a point b in the plane is the circle of radius |a − b| centred around a. The point a is a
fixed point of the orbit.
Theorem 2.2.3 Orbit-Stabilizer thm:
The map g.x → gGx defines a bijection (one-to-one correspondence).
The theorem states that each point in the orbit corresponds to a coset gGx of the stabiliser subgroup.
Proof: The map is surjective by definition since gGx partitions G by virtue of theorem 2.2.1. To
show that the map is injective, and therefore bijective, we need to show that g1 Gx = g2 Gx implies
g1 .x = g2 .x. If g1 Gx = g2 Gx then ∃g ∈ Gx s.t. g1 = g2 g. But then g1 .x = (g2 · g).x = g2 .(g.x) = g2 .x
and therefore the map is also injective. q.e.d.
8A group action is transitive if there is only a single orbit.
9 Note that the orbit corresponds to an equivalence class under the group action.
10 The stabiliser subgroup is also known as the isotropy group.
Corollary 2.2.1 If G is a finite group then the size of each orbit is a divisor of the order of the group.
I.e. |Gx | = |G|/|G(x)| is an integer.
Proof: By theorem 2.2.3 there is a one-to-one correspondence between the size of the orbit and the
number of left (right) cosets i.e. |G(x)| = |G|/|gGx |. Hence |G|/|G(x)| = |Gx | is an integer and |G(x)|
a divisor of the order of the group. q.e.d.
As one application we shall consider the proof of the famous Cauchy theorem which reads:
Theorem 2.2.4 Cauchy’s thm:
If |G|/p is an integer and p is prime then there exists an element of order p. (∃g ∈ G s.t. g p = e.)
Proof: Let (x1 , x2 , ....xp ) ∈ X (with xi ∈ G) be the set of ordered strings for which x1 · x2 .. · xp = e.
Note that the number of such strings is |G|p−1 since we may freely choose p−1 elements out of G
and then fix the last one to be the inverse. In particular |X|/p is an integer since |G|/p is. Next,
we define a group action of Zp on X; m ∈ Zp acts on X by cyclic translation: m.(x1 , x2 , .., xp ) =
(xm+1 , .., xp , x1 , x2 , .., xm ). Since the size of the orbit divides the order of the group which is |Zp | = p
in the case at hand, the size of each orbit is either 1 (all elements are equal) or p (not all elements
are equal). An example of the former is (e, .., e). But it cannot be the only example as otherwise |X|
is not divisible by p (that is to say np + 1 for n integer is not divisible by p). Hence there exists a
configuration for which x1 = .. = xp 6= e. This implies that x1 (6= e) ∈ G with xp1 = e and ends the
proof. q.e.d.
This is proof is elegant and short and provides an example of the power of the notions introduced
previously in this section. Trying to understand is rewarding!
Corollary 2.2.2 A group G of prime order |G| = p (p is prime) is isomorphic to the cyclic group Zp .
Before presenting the proof let us refine the notion of the cyclic group Zp by stating its presentation
and representations. The latter notion will be discussed throughout this lecture in more detail for
generic groups for which it is more complicated. The cyclic group has the group presentation,
Zp = ha | ap = ei . (2.6)
The element a can be represented by any of the p − 1: a = e2πim/p for 0 ≤ m < p.

Proof: By Cauchy’s thm a group of order p has an element of order p i.e. ∃a ∈ G s.t. ap = e. (And
since p is prime am = e for m < p cannot exist since m does not divide p.) Hence the group, i.e. all
multiplication rules, are completely determined by ap = e c.f. (2.6) q.e.d.
Let us summarise to this end some of the notions introduced. Let there be a group G a set X and
a group action ”.” then the orbit of a an element x ∈ X is G(x) = {g.x | g ∈ G} ⊂ X a subset of X,
Gx = {g ∈ G | g.x = x} is as subgroup called the stabiliser subgroup. c.f. definition 2.2.4. The orbit
stabiliser thm states that to each element g.x there corresponds one and only one left coset gGx (or
right coset Gx g).
Exercise 2.2.2 Cauchy’s thm and orbits

a) The cyclic group Z6 is a subgroup of O(2) in example 2.2.2. Consider Z6 acting on points
on the two dimensional plane by rotation centred at a point a. Draw the two characteristic
orbits and give the corresponding stabiliser subgroups. Give the size of the orbits and check
that they satisfy corollary 2.2.1.
b) Find the order of a proper subgroup H of a group G (|G| = 100) given the following hints:
1) H is not cyclic 2) H has got no element which is its own inverse.
c) Following up on a previous exercise 2.1.3, show that the order of a group of four elements
{e, g1 , g2 , g3 } admits only four different multiplication tables. Argue (or show) that three of
those are isomorphic to each other. Hence there are two inequivalent groups of order four.
They are isomorphic to Z4 and Z2 × Z2 where the latter is the Vierergruppe from exercise
2.1.3 as introduced by Felix Klein.
d) Given a group G of order 52, what is the order of the proper non-abelian subgroup? The
claim is that the order is unique. Hint: you will need to use exercise c).
2.2.3 Normal subgroups

Definition 2.2.5 Normal subgroup = invariant subgroup:
A normal subgroup N is a subgroup of a group G, denoted by N C G, which is invariant under the
group action of conjugation. I.e.
N = {ni ∈ G | g · ni · g −1 = nj , ∀g ∈ G} . (2.7)
Theorem 2.2.5 Left- and right-cosets of subgroup H are identical ⇔ H is normal in G.
We shall omit the proof of this theorem. Comment: To some extent a normal subgroup is a subgroup
which is invariant under relabelling (conjugation is a kind of relabelling of the group when it comes
to representations at least). For the permutation groups this is most obvious. Having this in mind it
is not surprising that a normal subgroup can be factored retaining a group structure:
Theorem 2.2.6 left (right) cosets of normal subgroup allow a group structure.
Given N C G one can define a so-called quotient group (factor group) G/N of order |G|/|N | by
imposing the following binary operation (group multiplication):
g1 N · g2 N = g1 · g2 N . (2.8)
Another way of writing (2.8) is g1 N ·G/N g2 N := (g1 ·G g2 )N where ·G emphases in which group the
product is to be understood. In other words it is the group product in G that is used to define the
group product in G/N .
(2.8)
Proof: G/N is a group: Since g1 N, g2 N ∈ G/N ⇒ (g1 N ) · (g2−1 N ) = (g1 g2−1 )N ∈ G/N and
thus G/N is a subgroup by virtue of theorem 2.1.1. It remains to be shown that (2.8) is well defined
multiplication law. Step 1: show that g1 N · g2 N ⊇ g1 · g2 N : g1 · g2 N 3 g1 · g2 · n for n ∈ N . The
latter may be written as (g1 · e) · (g2 · n) which is clearly contained in (g1 N ) · (g2 N ). Step 2: show that
g1 N · g2 N ⊆ g1 · g2 N : Let g1 · n1 · g2 · n2 be an element of g1 N · g2 N . This element can be written as
(2.7)
(g1 · g2 ) · (g2−1 · n1 · g2 ) · n2 = g1 · g2 · (nj · n2 ) for some nj ∈ N . This proves the assertion that (2.8)
is a well-defined multipilciation law and ends the proof. q.e.d.
In essence in the quotient group the group elements are the cosets, the sets shown Fig. 2.1, and a
group multiplication is defined by (2.8). The identity corresponds to N (which is H is Fig. 2.1).
Definition 2.2.6 Simple group = no normal subgroups
A group which has no other normal subgroups other than the identity and the group itself is called a
simple group.11 The term simple is appropriate, in view of theorem 2.2.6. The simple groups are to
the classification of groups what the prime numbers are to the integer numbers. Namely a structure
that cannot be divided any further.
Theorem 2.2.7 If H is a subgroup of G and |G|/|H| = 2 then H is a normal subgroup.
Proof: Left and right cosets partition the group (theorem 2.2.1) ,
G = H ∪ gH = H ∪ Hg , (2.9)
see also (2.5). Hence left and right cosets are the same and theorem 2.2.5 implies that H is normal
in G. Note that this implies that the quotient group G/H is isomorphic to Z2 since the latter is the
only group of order two.
Theorem 2.2.8 First isomorphism thm: 12 Let there be homomorphism ϕ : G → G̃ then the
following statements are true:
a) (the image) Im(ϕ) is a subgroup of G̃.
b) (the kernel) K = Ker(ϕ)13 is a normal subgroup of G

c) G/K is isomorphic to Im(ϕ)
Proof: We shall content ourselves in proving b) from which a) at least is obvious. b) Show that K is
a subgroup. Assume k1 , k2 ∈ K then k1 · k2−1 ∈ K since ϕ(k1 · k2−1 ) = ϕ(k1 ) · ϕ(k2−1 ) = e · e = e.14 By
theorem 2.1.1 K is a subgroup. Show that K is normal: For k1 ∈ K and g ∈ G, then ϕ(g · k1 · g −1 ) =
ϕ(g) · ϕ(k1 ) · ϕ(g −1 ) = ϕ(g) · e · ϕ(g −1 ) = ϕ(g · g −1 ) = e and hence K is normal. q.e.d.
The isomorphism theorem is illustrated in Fig. 2.4. The essence of the first isomorphism theorem is
that to show that a group, say Ni is a normal subgroup, it is often simplest to find the homomor-
phism where the kernel corresponds to Ni . The theorem will be applied to show that even and off
permutations are well defined (c.f. section 2.3.2).
11 An interesting fact that emerged from the classification of finite groups is that all simple finite groups admit a set
of generators which consists of at most two elements. From the simple groups all other finite groups can be built.
12 There are two further isomorphism theorems frequently stated in textbook which we do not discuss in this course.
13 For a group homomorphism the Ker = {g ∈ G|φ(g) = e }.
G̃
14 In the last step we, silently, used ϕ(k −1 ) = ϕ(k )−1 . You are going to fill in this step in exercise 2.2.3.
2 2
Figure 2.4: The isomorphism theorem 2.2.8 illustrated.
Exercise 2.2.3 normal subgroups

a) Is {e, (1, 2)} a normal subgroup of S3 ? To assess this question you could compare left and
right cosets, which you have assessed in a previous problem for an equivalent situation, and
use a theorem of current section.
b) Show that (nZ, +) is a normal subgroup of (Z, +)
c) Show that the factor group Z/nZ is isomorphic to Zn . It is sufficient to state the isomor-
phism (that is to say the map).
d) Show that for a homomporphism, ϕ : G → G̃, the following relation holds: ϕ(g −1 ) = ϕ(g)−1 .
This fills a gap in the proof of thm 2.2.8.
2.2.4 Direct and semidirect products of groups

Definition 2.2.7 Direct products:
The direct product of two groups H and J, denoted by H × J, is the set
H × J = {(h, j) | h ∈ H , j ∈ J}
with the binary operation
(h, j) · (h0 , j 0 ) = (h · h0 , j · j 0 ) , h, h0 ∈ H , j, j 0 ∈ J .
Note that H × J is a group of order |H × J| = |H||J|. For finite groups you may think of the size of
the group H × J as the size of the rectangle with lines |H| and |J|.
Definition 2.2.8 Semidirect products:

Let φ be a group action of a group H on J then one can define a semi-direct product denoted by
H n J. H n J is the set H × J with the following binary operation:
(h, j) · (h0 , j 0 ) = (h · h0 , j · φ(h, j 0 )) ≡ (h · h0 , j · (h.j 0 )) , h, h0 ∈ H , j, j 0 ∈ J .
The order of H n J is |H| · |J| in close analogy to order of the direct product groups.
Examples 2.2.3 of semidirect products
- An example of the semidirect product is the group of euclidian isometries which consist of a
rotation R and a translation ~a which act on some vector space. Two consecutive applications
give rise to
((R0 , ~a0 ) · (R, ~a)) ◦ ~x = (R0 , ~a0 ) ◦ ((R, ~a) ◦ ~x) = (R0 , ~a0 ) ◦ (R~x + ~a) = R0 R~x + R0~a + ~a0 (2.10)
and hence (R0 , ~a0 ) · (R, ~a) = (R0 R, R0~a + ~a0 ) 6= (R0 R, ~a + ~a0 ) is not equal to a direct product
structure. One writes ISO(n) = O(n) n Rn . In physics rotation and boosts plus space-time
translations correspond to the so-called Poincaré group whose invariants, the mass and the spin,
characterise a particle. When the translations are omitted the group is referred to as the Lorentz
group which is obviously a subgroup of the Poincaré group. Hence the Poincaré group is the
semidirect product of the Lorentz and the time-space translation group.
Figure 2.6: Triangle with D3 -dihedral symmetry. The

dihedral symmetry Dn can be understood as the sym-
metry of the regular n-polygon.
Figure 2.5: C2 H6 -molecule with D3 -dihedral symme-

try. Taken from [Ham62].
• Another important example comes from chemistry where many molecules (c.f. Fig. 2.5) are
characterised by what is called the dihedral group, denoted by Dn , of order 2n which is a
semidirect product Zn o Z2 .15 The group homomorphism of Z2 on Zn is given by inversion.16

In Fig. 2.5 the inversion is literally the inversion through the centre point of the molecule.
Alternatively the dihedral group Dn is the symmetry group of a regular n-polygon as illustrated
for the triangle in Fig.2.6. The inversion (operation s in Fig. 2.6) corresponds to rotating
the polygon by 180◦ around a straight line from a vertex through its centre. Examples for
(ri , si ) ∈ Zn o Z2 are
(e2πi3/n , 1) · (e2πi1/n , 1) = (e2πi4/n , 1)

(e2πi3/n , −1) · (e2πi1/n , 1) = (e2πi3/n e2πi(−1/n) , −1) = (e2πi2/n , −1) , (2.11)
and hopefully clarify the text above. A presentation is given by,
Dn = < r, s | rn = s2 = s−1 rsr = e > . (2.12)
The group element r corresponds to a rotation of 360◦ /n around the axis of the molecule and
s corresponds to the rotation around 180◦ of the centre of the molecule. For D3 the operations
are illustrated in Fig. 2.6. As already mentioned the dihedral group is the symmetry group of
molecules. As another example we shall mention the symmetry of a snowflake which corresponds
to D6 (hexagon).
Exercise 2.2.4 Products of groups.

a) Show that S3 is isomorphic to D3 by finding the elements of S3 (from the multiplication
tables) that satisfy the presentation 2.12. You could also attempt a geometric approach
and identify the symmetries in either the C2 H6 -molecule Fig. 2.5 or Fig. 2.6.
b) As a matter of fact (up to isomorphisms) there are two groups of order six. One of
them is D3 . Can you find or give the other one?
c) The Vierergruppe Z2 × Z2 has got the following group presentation:
Z2 × Z2 = ha, b | a2 = b2 = (ab)2 = ei .
Write down explicit generators for a, b and ab in terms of the direct product notation:
(e2πim/2 , e2πin/2 ).
2.3 Permutation groups

2.3.1 The symmetric permutation group Sn
We have already mentioned the permutation group Sn in examples 2.1 and shown the multiplication
table of S3 in table 2.2. In this chapter we shall discuss some more details.
15 This is why some authors also denote the dihedral group of 2n elements by D2n .
16 Recall from exercises 2.1.4 that inversion, itself, is a group homomorphism for abelian groups which is clearly the
case for Zn .
Consider a set of n elements X = {x1 , ..., xn }, then all possible permutations of this set, obviously,
define a group which is denoted by Sn and called the symmetric permutation group.1718 ..Sn
leaves symmetric polynomials P (x1 , x2 , ..xn ) The number of elements of this group corresponds to the
number of possible permutations which is n!. The elements of the group can be denoted by brackets,
as in table 2.2, (a1 , a2 , a3 ..) with 1 ≤ a−i ≤ n which correspond to cyclic permutations; best explained
by example:
(1, 4, 2) · (1, 2, 3) ◦{xa , xb , xc , xd } = (1, 4, 2) ◦ {xc , xa , xb , xd } = {xa , xd , xb , xc } . (2.13)
| {z }
=(2,3,4)
Once understood we can define multiplication directly within the group without reference to the set
X, as indicated in Eq.(2.13). Other example which you may want to verify are:
(1, 2) · (2, 3) = (1, 2, 3) , (2, 3) · (1, 2) = (1, 3, 2) . (2.14)
A cycle (a1 , .., ak ) is referred to as a k-cycle and a 2-cycle is called a transposition. Without proof:
All elements of Sn can be written as the product of disjoint cycles. By two disjoint cycles we mean
cycles (a1 , .., an ) and (b1 , .., bm ) for which ai 6= bj for 1 ≤ i ≤ n and 1 ≤ j ≤ m. any which do not
have any permutation element in common. Note that disjoint cycles commute with each other e.g.
(1, 2)·(3, 4) = (3, 4)·(1, 2). Hence whenever two cycles are not disjoint one can commute them through
other elements and then multiply them into a single object as done above (2.14). You will provide
some more evidence in exercise 2.4.1. Problems such as the question of how many m-cycles there are
in Sn are very typical combinatorial problems that are best broken into parts. For example we may
first look at how many ways there are to get m distinct numbers out of n which is n!/(n − m)!/(m!).
Second we take into account the ordering. We may freely choose one element to be in the first entry
by convention but then the order of all the others matters and there are (m − 1)! possibilities. Hence
the answer is n!/(n − m)!/m. For example there 4!/(4 − 3)!/3 = 8 3-cycles in S4 :
{(1, 2, 3), (1, 3, 2), (1, 2, 4), (1, 4, 2), (1, 3, 4), (1, 4, 3), (2, 3, 4), (2, 4, 3)} . (2.15)
Theorem 2.3.1 The transpositions (2-cycles) generate Sn .
Proof: The elements of Sn consist of products of k-cycles with 1 ≤ k ≤ n. An arbitrary k-cycle can
be obtained from the following product of transpositions.
(a1 , a2 , ..., ak ) = (a1 , ak ) · (a1 , ak−1 ) · ... · (a1 , a2 ) , q.e.d.
More specifically
Theorem 2.3.2 (a) The transpositions {(1, 2), (1, 3), .., (1, n)} generate Sn .
(b) The transpositions {(1, 2), (2, 3), .., (n − 1, n)} generate Sn .
(c) The transposition (1, 2) and the n-cycle (1, 2, .., n) generate Sn .
Proof: (a) Note that (a1 , a2 ) = g · (1, a2 ) · g −1 , g = (1, a1 ) and then use theorem 2.3.1. (b)
Note that (1, k) = g · (1, 2) · g −1 , g = (k − 1, k) · .. · (3, 4) · (2, 3) and then use (a). (c) Note that
(k, k + 1) = g · (1, 2) · g −1 where g = (1, 2, .., n)k and then use (b) q.e.d.
The proof of (b) and (c) are in essence implementing a change of basis (relabelling the permutation
elements).
17 A more abstract definition for Sn are the maps that are bijections (one-to-one) on a set of n-elements.
18 S is the group which leaves so-called symmetric polynomials in n variables invariant.
n
2.3.2 The alternating group An

We have learned that any element of the symmetric group can be written as a product of k-transpositions.
The number of transposition that occur are unambiguously even or odd and are called even- and
odd-permutations respectively. This can be seen by using the first isomorporhism theorem 2.2.8.
Consider the Vandermonde polynomial P (x1 , .., xn ) = (x1 −x2 )·..·(x1 −xn )·(x2 −xn )·..·(xn−1 −xn )19
of all possible products of (xi − xj ) with i 6= j. Then the map φ : acting on P/|P | by permutation
is a group homomorphism from Sn to Z2 since the image of Sn acting on P is ±P and therefore
isomorphic to Z2 . The kernel of this map are the even permutations i.e. An . The latter are therefore,
according to theorem 2.2.8, a normal subgroup (i.e. An C Sn ) of order |An | = |Sn |/|Z2 | = n!/2.20
Stated as a theorem:
Theorem 2.3.3 alternating group An : The even-permutations generate a normal subgroup of Sn
of order n!/2
Proof: Given above.
Theorem 2.3.4 The 3-cycles generate An
Proof: According to theorem 2.3.2 Sn is generated {(1, 2), (1, 3)..}. In the case of An each element,
consists of an even number of transpositions. Since (1, a1 ) · (1, a2 ) = (1, a2 , a1 ) the statement follows.
q.e.d.
2.4 Platonic solids

A platonic solid is a regular, convex geometric body in three dimensions which consists of faces
of a single polygon. There are five solids which meet this criteria, each named after the number
of polygons: tetrahedron, hexahedron (cube), octahedron dodecahedron, icosahedron, c.f. Fig. 2.7.
Further information can be found in table 2.3. The symmetry of the platonic solids seems predestined
to be explained by group theory. The symmetry groups are given in table 2.3 on the right and
correspond all to groups of the An - and Sn -type. An obvious start to characterise the symmetry
group is to note that any symmetry permutes the vertices of the body. Hence the symmetry group
of a platonic solid with n vertices is a subgroup of Sn (in accordance with Cayley’s theorem 2.1.2).
Intuitively the reason why it is not necessarily Sn is that the platonic solids have a certain rigidity so
that one can not arbitrarily permute the edges.
Moreover, platonic solids which are equivalent under interchange of vertices and surfaces are dual
to each other; they have the same symmetry group. As can be seen from table 2.3 the tetrahedron is
dual to itself and the hexahedron and octahedron (c.f. Fig. 2.8 in addition) as well as the dodecahedron
and icosahedron are dual to each other respectively.
We would like to illustrate the symmetry group by looking at one specific example that is the
tetrahedron. The symmetry operations are shown in Fig. 2.9. They consist of a rotation symmetry
of 120◦ (denoted by r) around the axis of a vertex through the centre of the opposite face and a
180◦ rotation through around an axis that goes through the middle of two opposite edges. From the
figure it is easily deduced that r and s correspond to (2, 3, 4)- and (1, 4)(2, 3)-permutation respectively.
19 This is a so-called alternating polynomial. The name derives from the fact that the permutation of any two variables
alternates the sign and this is what gave the alternating group its name c.f. definition above.
20 A C S also follows from theorem 2.2.7.
n n
Figure 2.7: Platonic solids – photocopied from [Arm88].
solid faces vertices edges polygon symmetry group

tetrahedron 4 4 6 triangle A4
hexahedron 6 8 12 square S4
octahedron 8 6 12 triangle S4
dodecahedron 12 20 30 pentagon A5
icosahedron 20 12 30 triangle A5
Table 2.3: Platonic solids (c.f. Fig.2.7) with number of faces (F), vertices (V) and edges E), the polygon
and their symmetry group. The former are related by the famous Euler formula: V − E + F = 2 which can
be proven in many ways. The hexahedron and octahedron as well as the dodecahedron and icosahedron are
dual under interchange of the number of faces and vertices. This is the reason they admit the same symmetry
group.
Those permutations are even and provide some evidence that the subgroup could be A4 . You will
complete a more thorough proof in the exercises below.
Exercise 2.4.1 Permutation groups.

a) In the notes it was stated that any element of Sn can be written as a product of disjoint
cycles for which we have not given a proof. This exercise aims to give some evidence by
example.
Take (1, 2, 3) and (2, 4, 5) (common element 2) and write down the the 5-cycle corresponding
to (1, 2, 3) · (2, 4, 5). Hint: In order to establish the multiplication rule let the cycles act on
a set of five ordered elements.
b) Exercise a) will help you to answer the following question: What is the largest order of an
element in S10 ?
Figure 2.8: Octahedron and cube being dual under in-

terchange of faces and vertices. Figure from [Arm88].
Figure 2.9: Tetrahedron (photocopied out [Arm88])
c) The symmetry group of the tetrahedron is isomorphic to A4 . A presentation of A4 is given

by:
A4 = < a, b | a2 = b3 = (ab)3 = e > . (2.16)
Show that r = (2, 3, 4) and s = (1, 4)(2, 3), as defined above these exercises, satisfy the
presentation relation in (2.16).
d) The symmetry group of the cube is isomorphic to S4 . For the cube it is appropriate to
consider the space diagonals as the group elements (They retain the rigidity of the cube.
Think of the space diagonals as a scaffolding of the cube.) According to theorem 2.3.2 S4 is
generated by (1, 2) and (1, 2, 3, 4). Identify those symmetry operations geometrically. Hint:
Whereas the operation (1, 2, 3, 4) should be immediate, (1, 2) could be a bit more difficult.
You could try to look for a symmetry operation which is of order two since (1, 2)2 = e.
e) List all 24 elements of S4 in the form (), (x, y), (x, y, z), (x, y, z, w) and (x, y) · (z, w) with
x, y, z, w ∈ {1, 2, 3, 4}.
f) Let S4 act on the sets 1) x1 = {a, a, a, a} 2) x2 = {a, a, a, b}, 3) x3 = {a, a, b, c} and 4)
x4 = {a, b, c, d} by permutation. Give the stabiliser subgroup and the size of the orbit (write
down all elements of the orbits) for each set x1,2,3,4 and verify corollary 2.2.1.
2.5 Applications
Most applications of finite group theory involve representation theory which are going to treat in
subsequent chapters. We shall content ourselves alluding at one (two) example(s) from chemistry
which can be found in [Atk08] chapter 11.
Figure 2.10: Water molecule H2 0 with Z2 symmetry. The water molecule has a permanent electric dipole
moment along the symmetry axis as further discussed in examples 2.5.1.
Examples 2.5.1 Polarity and chirality of molecules..
• A polar molecule is a molecule with a permanent electric dipole. The basic idea is that the
molecule has to be asymmetric to a certain degree for it to admit an asymmetric charge distri-
bution. The latter is the source of a permanent electric dipole. For instance any molecule that
has a Zn≥2 -symmetry cannot have an electric dipole that is perpendicular to the symmetry axis.
So it appears at first that C2 H6 Fig. 2.5 and H2 O (water) in Fig.2.10 could have an electric
dipole along the rotation axis of Z3 and Z2 respectively. For C2 H6 the s-symmetry operation
Fig.2.6 takes one end of the molecule to the other and this obstructs the formation of a electric
dipole. Hence only H2 0, and not C2 H6 , has got a permanent electric dipole.
As a matter of fact only molecules with Zn -symmetry types (c.f. [Atk08] for more details)
posses a permanent electric dipole along the rotation axis.
• Another example is the chirality of a molecule (mirror image not the same as the original); c.f.
[Atk08]. A molecule may be chiral if does not admite an improper rotation axis. An improper
rotation, is a rotation around an axis followed by an inversion.21 Since we did not discuss those
kind of symmetry groups we shall not go into any further details here.
21 These groups are denoted, somewhat unfortunately, by S in the Schoenfliess notation used in chemistry literature.
n
They are of order n and not to be confused with the permutation group of order n!.
Chapter 3
Representation theory
Representations are realisation of groups on linear vector spaces. Applications within mathematics,
physics, computer science and so on are enormous. Especially from the viewpoints of sciences, other
than pure mathematics, representation are natural since computations are performed in concrete
settings such as linear spaces. From your course on quantum theory you are familiar with the
concept of discussing physics (equivalently) in different linear spaces. E.g. the harmonic oscillator can
be discussed in the coordinate space |xi, momentum space |pi or the number of particle representation
|ni. Each one of them has its own merits. When in addition the system displays a symmetry, e.g.
rotational symmetry of the potential, then the symmetry transformation acts as a representation on
the physical eigenstates of the Hilbert space. This will become clearer when we proceed to examples.
Discussion of finite representation theory, which consists to a large extent of character theory,
is presented in chapter 4; followed by chapter 5 on Lie groups. In the current chapter we give the
definition of a representation and discuss it in the, by now, familiar setting of S3 . Basic definitions
are given in section section 3.2. Section 3.3 discusses Schur’s lemma and decomposability, which are
relevant both to finite representation theory of chapter 4 and the Lie groups of chapter 5. Applications
are presented at the end of the chapters.
Definition 3.0.1 representation:
A representation ρ is a homomorphism (c.f. definition 2.1.3) from a group G into the set of linear
(invertible) operators on a complex vector space V (the latter are denoted by GL(V ) and form a
group structure themselves under composition).
ρ: G −→ GL(V )
g −→ ρ(g) .
Since ρ is a homomorphism the map is compatible with the group structure, (c.f. Eq. (2.2)),1
ρ(g1 · g2 ) = ρ(g1 )ρ(g2 ) . (3.1)
The relevant2 representations of finite groups are finite dimensional and therefore ρ(g)’s are nothing
but matrices acting on a finite dimensional vector space; i.e. V = Cn with n = dim V .
1 Alternatively we may think of a representation as a group action on the linear space V .
2 i.e. irreducible, a terms to defined shortly hereafter
29
CHAPTER 3. REPRESENTATION THEORY 30
3.1 Representations the pedestrian way

Going back to the permutation group of three elements S3 . According to theorem 2.3.2 a), S3 is
generated by the elements (1, 2), (1, 3). Associating the permutation with permutation of unit vectors
of a three dimensional space we may represent those elements by:
   
0 1 0 0 0 1
ρ((1, 2)) = 1 0 0 , ρ((1, 3)) = 0 1 0 . (3.2)
0 0 1 1 0 0
All other elements may be generated from (3.2). Whereas this is certainly a representation, known as
the permutation representation, it does nor answer the following two questions: i) are there any
other representations? ii) can the representation be further reduced? We shall see that representation
theory of finite groups gives concrete and definite criteria for answering questions i) and ii). Before
presenting the theory let us point out that one can see that the representation (3.2) can be further
reduced. Both matrices, and therefore all group elements, have the eigenvector ~v = (1, 1, 1)T , with
eigenvalue λ~v = 1, in common:
ρ((1, 2))~v = ~v , ρ((1, 3))~v = ~v . (3.3)
Hence one can construct a change of basis, ρ → ρ0 , such that the matrices are block diagonal
   
1 0 0 1 0 0
ρ0 ((1, 2)) = 0 a1 a2  , ρ0 ((1, 3)) = 0 b1 b2  . (3.4)
0 a3 a4 0 b3 b4
It is noted that the original three dimensional representation has split into a one and a two dimensional
representation. Representations which can be split as the one in Eq. (3.2) are called reducible. A
representation which cannot be split any further is called an irreducible representation. An aim of
representation theory is to find, and if possible construct, all irreducible representation of the group.
It is now time to proceed to the more formal representation theory.
Exercise 3.1.1 Representations the pedestrian way

a) Verify through explicit matrix multiplication that the product ρ((1, 2))ρ((1, 3)) in (3.2)
generates an element which can be interpreted as ρ((1, 3, 2)) in terms of permutation of the
unit vectors.
3.2 Basic definitions of representation theory

Representations have been defined in definition 3.0.1 and we have seen a first example in section 3.1.
Below ρ(g) denotes a representation of g ∈ G on the vector space V .
Definition 3.2.1 of representation theory
• A representation is faithful if the homomorphism ρ is injective (inverse map from image is

defined). If the latter is not the case the representation is called unfaithful accordingly. If
ρ(g) = ρ(e) for ∀g ∈ G then the representation is called the trivial representation.
• A unitary representation is a representation for which ρ(g) is a unitary operator on the
vector space.
• Two representations ρ and ρ0 are called equivalent if they can be mapped into each other by
a similarity transformation. I.e. if there exists S ∈ GL(V ) s.t. ∀g ∈ G, ρ(g) = Sρ0 (g)S −1 .3
• An invariant subspace is a subspace W ⊂ V which is left invariant by the representation
acting on it i.e.
~ ⊂ W , ∀w
ρ(g)w ~ ∈ W , ∀g ∈ G . (3.5)
• A representation is irreducible if there are no invariant subspaces.
Fact: If there are invariant subspaces one can find a basis where ρ(g) can be written as

ρ1 (g) 0
ρ(g) = , ∀g ∈ G , (3.6)
0 ρ2 (g)
in a block diagonal form.4 If a representation can be written in the form (3.6) the representation
is said to be reducible.
In essence a representation is either irreducible or reducible. For finite groups (c.f. theorem
3.3.3) and continuous groups which are compact the reducible representations are a direct sum
of the irreducible representations.5
• A real representation is a representation which can be brought (similarity transformation)
into a form where the matrix entries are all real numbers. If this is not the case we speak of a
complex representation.6 The complex conjugate representation of ρ is denoted by ρ̄ = ρ∗
and is given by complex conjugation (of the matrix entries).
Unitary representations are of importance for quantum physics. Unitary transformations are the
natural language to describe symmetries on the physical Hilbert space since they preserve the inner
product. Irreducible representation are not only the atomic objects of mathematical representation the-
ory but they are also the ”atoms” in physics. According to Wigner an elementary particle corresponds
to an irreducible representation of the Poincaré group (i.e. space-time symmetry group including
translations.) In physics, or particle physics, complex representation are associated with charges; real
representations carry no charge.
Exemplifying some of these terms for the representation discussed in (3.2). The representation
(3.2) is reducible since it can be written in block diagonal form (3.4). The two representations (3.2)
and (3.4) are equivalent. The representation ρ0 (and therefore also ρ since it is equivalent ) contain
the trivial representation which is in particular unfaithful. The representation ρ is real since the
corresponding matrices have real entries. The representation ρ0 is real since it is equivalent to the
representation ρ which is real.
3 As the wording suggest the equivalence of representations is an equivalence relation.
4 We omit the proof of this statement.
5 We shall give an example of a non-reducible representation for the translation group (which is non-compact) in
chapter 5.
6 For SU (2) we will encounter a refinement of this notion to so-called pseudo-real representations.
3.3 Basics of representation theory Maschke’s thm, Schur’s

Lemma and decomposability
In this section we state a few facts about representation theory of finite groups which are also relevant
for representation theory of the Lie groups (compact Lie groups U (1), SO(2), SO(3) and SU (2))
which are going to discuss in the chapter 5. The theorems are stated for finite groups.
Theorem 3.3.1 Maschke’s thm:
Any representation of a finite group is equivalent to a unitary representation.
Proof: Suppose (x, y) is the inner product of the vector space. One can define a new inner product
1 X
(x, y)G ≡ (ρ(g)x, ρ(g)y) (3.7)
|G|
g∈G
by averaging over the group (Weyl’s trick). It is easily verified (exercises) that ρ(g) is unitary i.e.
(ρ(g)x, ρ(g)y)G = (x, y)G with respect to the new inner product. q.e.d.
Theorem 3.3.2 Schur’s Lemma
Let ρ and ρ0 be two irreducible representations of vector spaces V and V 0 and T : V → V 0 be a linear
map satisfying ρ0 T = T ρ then
a) T is either an isomorphism, or T = 0 is trivial.
b) If V = V 0 then T = λ · 1 (for λ ∈ C and 1 the identity map on V ).
Schur’s lemma states in essence that there is no room for any non-trivial homomorphisms between
irreducible representations of a group. Schur’s Lemma is the fundament of representation theory. It
is the essential ingredient to many powerful theorems. Often you will find Schur’s lemma stated as
follows: let ρ be a matrix irreducible representation and T a matrix in the same space with [T, ρ(g)] = 0
for ∀g ∈ G then T = λ1. This is clearly equivalent to theorem 3.3.2b).
Proof: a) The first statement follows from the fact that Ker(T ) and Im(T ) are invariant subspaces (to
be discussed below) and therefore Ker(T ) = 0, V and Im(T ) = 0, V 0 since irreducible representation
do not admit non-trivial invariant spaces. If Ker(T ) = 0 then Im(T ) = V 0 and T is an isomorphism
by virtue of the first isomorphism theorem 2.2.8. If Ker(T ) = V then Im(T ) = 0 and T = 0 (maps all
elements into zero vector of V 0 .)
Remains to be shown that Ker(T) = {v0 ∈ V | T v0 = 0} is an invariant subspace i.e. T ρ(g)v0 = 0.
This follows from T ρ(g)v0 = ρ0 (g)T v0 = 0 by using the premise T ρ(g) = ρ0 (g)T .
b) If T = 0 then λ = 0. If T 6= 0, T must have at least one eigenvalue λ. Hence for U ≡ T − λ1,
dim[ker(U )] ≥ 1 and U is not an isomorphism; by virtue of a) U = 0; i.e. T = λ1 q.e.d.
Theorem 3.3.3 Decomposability thm
a) Any reducible representation of a finite group decomposes into irreducible representations,
ρ(g) = m1 ρ1 (g) ⊕ .. ⊕ mk ρk (g) , (3.8)
with multiplicities mi ∈ Z+ .7
7A representation is simply reducible if mi = 0, 1.
b) The decomposition into k factors is unique whereas the decomposition into a direct sum of mi
copies is not.
Proof: a) We content ourselves with the remark that the proof uses the same inner product as in
theorem 3.3.1. b) Consider another representation of ρ0 = m01 ρ01 ⊕ .. ⊕ m0K ρ0K of G in V and T a
map between those representations. Then it follows from Schur’s lemma that T must map ρi to ρ0i
which are isomorphic to each other. (One cannot map a representations of different dimensions to
each other as otherwise T would not correspond to an isomorphism.) q.e.d.
Exercise 3.3.1 a) Argue that a representation ρ(g) is unitary with respect to the inner prod-
uct (3.7). I.e. (ρ(g)x, ρ(g)y)G = (x, y)G .
Chapter 4
Representation theory of finite

groups
4.1 Character theory for finite groups

The character theory is the heart of the representation theory of finite groups.
Definition 4.1.1 The trace of a representation ρ is called the character χ
χ(g) ≡ Tr[ρ(g)] . (4.1)
Hence to each group element a character is associated. The following properties are immediate:
Corollary 4.1.1 (of definition 4.1.1):
a) The characters of two equivalent representation (c.f. 3.2.1) are equal:
χ(Sρ(g)S −1 ) = Tr[Sρ(g)S −1 ] = Tr[ρ(g)] = χ(g) . (4.2)
b) The characters of conjugate elements are identical for the same reason as in a) (replace S → ρ(g 0 )
above). It is therefore sufficient to consider conjugacy classes only. The character is said to be
a class function (its value is constant on any conjugation class of a representation).
c)
χ(ρa ⊕ ρb ) = χ(ρa ) + χ(ρb ) , (4.3)
χ(ρa ⊗ ρb ) = χ(ρa ) · χ(ρb ) , (4.4)
−1 ∗
χ(ρ(g )) = χ(ρ(g)) . (4.5)
Above the index a and b are labels referring to a type of representation (e.g. trivial, permutation
representation in the case of Sn ). The first two properties follow from (1.18) and (1.20) and
the definition of the trace respectively. The last property (4.5) is a consequence of Maschke’s
theorem 3.3.1:
exercise(2.2.3) Tr[AT ]=Tr[A]
χ(ρ(g −1 ) = χ(ρ(g)−1 ) = χ(ρ(g)† ) = χ(ρ(g)∗ ) = χ(ρ(g))∗ . (4.6)
34
CHAPTER 4. REPRESENTATION THEORY OF FINITE GROUPS 35
Maschke’s thm was used in the second equality.

• d) For one dimensional representations the character is equal to the representation matrix since
the latter is a 1 × 1-matrix. Note that a one dimensional representation is not necessarily the
trivial representation.
The characters characterise groups to a large extent. In fact the mathematician Gerhard Brauer
asked the question whether the characters together with the order of the elements would completely
determine a finite group. It took decades to find a counterexample and this shows how important
character tables are.
Definition 4.1.2 One can define an inner product on the space of characters as follows:
1 X
hχ(ρa )|χ(ρb )i ≡ χ(ρa (g))∗ χ(ρb (g))
|G|
g∈G
Nc
1 X
= ck χ∗ak χbk . (4.7)
|G|
k=1
Above χak = χ(ρa (gk )) with gk belonging to the k th -class with ck elements. The number Nc denotes
the total number of classes of the group. The second line exploits the fact that the character is a
class function. The character χ(g) can be thought of as a vector in CNc where each class represent a
linearly independent direction:
χ(ρ) = (χ(ρ(e)), ..., χ(ρ(gNc )) ∈ CNc , (4.8)
where gNc , as above, is a representative of the Nc th -class. Note, the first class is usually assigned to
the identity element.
Theorem 4.1.1 Orthogonality theorem

In terms of the inner product (4.7) the irreducible representations of G are orthonormal, i.e.
hχ(ρa )|χ(ρb )i = δab . (4.9)
The proof is given for the interested reader only as it is too abstract for a first course on represen-
tation theory. Let us just mention that the main ingredient is Schur’s lemma.
for the interested reader (not part of the course)
Proof: Let us assume that Va and Vb are irreducible representations of G. Then the space
of all homomorphism from Va → Vb is equal to Hom(Va , Vb ) = Va∗ ⊗ Vb where Va∗ is the dual
vector space of Va (going from ket to bra in Dirac notation). Let V0 denote the vector space
of trivial representations i.e. ρ(g)|V0 = 1V0 , , ∀g ∈ G.
The rest of the proof builds on three observations: i) Schur’s lemma states that the only
non-trivial homomorphism, compatible with the group structure (i.e. invariant), there can
be is an isomorphism. Hence dim(Hom(Va , Vb )0 ) = δab where the subscript zero denotes the
trivial representation subspace as before. ii) The character of the Va∗ ⊗ Vb representation is
given by
(4.4)
χ(ρVa∗ ⊗Vb ) = χ(ρVa∗ ) · χ(ρVb ) = χ(ρVa )∗ · χ(ρVb ) , (4.10)
1
P
where ρVa ≡ ρa in the statement of the theorem. iii) The operator φ = g∈G ρV (g) is a
|G|
2 ⊥ ⊥
projector (φ = φ) on V0 . I.e. φv0 = v0 , ∀v0 ∈ V0 and φV0 = 0 where V0 ⊕ V0 = V . To
see this note that applying φ to a vector results in an invariant vector and the only invariant
vectors corresponds to the subspace of trivial representation. Moreover, a second application
of φ just results in a rearrangement which has though no effect since it is averaged over the
group and therefore φ2 = φ. Hence the number of trivial representation is given by
1 X
dim(V0 ) = Tr[φ] = χ(ρV (g)) . (4.11)
|G|
g∈G
Assembling
i) iii) 1 X
δab = dim(Hom(Va , Vb )0 ) = dim((Va∗ ⊗ Vb )0 ) = χ(ρVa∗ ⊗Vb (g))
|G|
g∈G
ii) 1 X def
= χ(ρVa (g))∗ · χ(ρVb (g)) = hχ(ρVa )|χ(ρVb )i , q.e.d. (4.12)
|G|
g∈G
Corollary 4.1.2 (of the orthogonality theorem 4.1.1)

Let ρ = m1 ρ1 ⊕ ..mk ρk stand for a generic representation.
a) A representation is fully characterised by its character since the distinct ρi ’s correspond to
(4.3)
linearly independent vectors in χ(ρ(g)) = m1 χ(ρ1 (g)) + .. + mk χ(ρk (g)). You should think of
χ(ρ(g)) as a vector in CNc .
b) The multiplicity ma of ρa in ρ is given by
ma = hχ(ρa )|χ(ρ)i . (4.13)
c) The norm of a representation is

X
||χ(ρ)||2 = hχ(ρ)|χ(ρ)i = m2i . (4.14)
i
Hence ρ is irreducible ⇔ ||χ(ρ)||2 = 1.

d) The character of the identity element corresponds to the dimension of the representation:
χ(ρi (e)) = dim Vi (4.15)
e) The characters of the trivial representation are all one. I.e.

χ(ρ1 ) = (1, .., 1) ∈ CNc . (4.16)
Exercise 4.1.1 Characters:

a) Verify statements b),c), d) and e) in corollary 4.1.2
b) Obtain all six matrices of the representation (3.2). Two are given, one is trivial and one
of them you have already computed in exercise 3.1.1. Compute the remaining two and
compute the norm of the character using the inner product (4.7).
c) According to the decomposability theorem 3.3.3 the permutation representation is either a
direct sum of i) three one dimensional or ii) one one dimensional and a two dimensional
irreducible representation. I.e. ρperm = ρ1 +ρ10 +ρ100 or ρperm = ρ1 +ρ2 where the subscript
denotes the dimension of the representation. Using the result in b) and a statement in
corollary 4.1.2, show that ii) is the case.
d) Identify the equivalence classes of S3 . You may for instance guess them and verify them
either through 1) explicit computation of the matrices 2) the multiplication laws from the
multiplication table 2.2 or 3) you may proceed along the lines of proof of theorem 2.3.2 and
find the equivalence transformations. You might even want to try more than one method
to gain some confidence in the formalism. Note for S3 members of an equivalence class
corresponds to a renumbering of the elements so that guessing the equivalence classes is
possible.
In order to proof the powerful dimensionality theorem we are going to introduce the regular
representation
Definition 4.1.3 The regular representation corresponds to the left-action of the group on itself.
More precisely to each element g ∈ G we associate a vector ~eg (declaring ~eg to be linearly independent
of ~eg0 for g 6= g 0 ) and the representation acts as follows:
ρR (g)~eg0 = ~eg·g0 . (4.17)
The regular representation has got the following property

0 g 6= e
χ(ρR (g)) = , (4.18)
|G| g = e
which follows from the fact the ρR (e) = 1|G| and that no other element than ρR (e) has any entry on
the diagonal (to be verified in the exercises).
This definition together with the observation (4.18) is sufficient to proof the dimensionality theo-
rem:
Theorem 4.1.2 Dimensionality theorem
The sum of the dimension squared of all irreducible representations of a finite group is equal to the
order of the group
Nc
X
|G| = dim(Vi )2 . (4.19)
i=1
Proof: First we observe that

(4.13) def 1 X (4.18) (4.15)
mi = hχ(ρi )|χ(ρR )i = χ(ρi (g))∗ χ(ρR (g)) = χ(ρi (e)) = dim(Vi ) . (4.20)
|G|
g∈G
Second by virtue of the decomposability theorem 3.3.3 we may write

X (4.20) X
χ(ρR (g)) = mi χ(ρi (g)) = dim(Vi )χ(ρi (g)) (4.21)
i i
and choosing g = e with Eqs. (4.18) and (4.15) we get

X
|G| = dim Vi2
i
which completes the proof q.e.d..
Note, had we known the dimensionality theorem earlier, the reducibility of the representation
(3.2) would have been clear. Moreover exercise 4.1.1 could have been solved more efficiently. You can
complete the argument in exercise 4.1.2 a).
Exercise 4.1.2 dimensionality thm

a) Use the dimensionality theorem to argue that the largest irreducible representation of S3 is
of dimension two.
b) What is the dimension of the permutation and the regular representation of Sn ?

c) Is the regular representation always reducible?
4.2 Character tables

The information on characters is commonly presented in terms of so called character tables; e.g.
table 4.1 (a few pages ahead). First all group elements are divided into equivalence classes under con-
jugation. Note the identity is always a conjugation class on its own. At the top row a representative
of each equivalence class is listed. Sometimes a second row is present stating the order of the repre-
sentative. To each successive row an irreducible representation is associated where the corresponding
characters are listed. The fact that the table, without counting the top rows, can be interpreted as a
unitary square matrix (to be made precise) is going to be one of the main results of this chapter.
The rows in the character tables can be thought of as linearly independent vectors by virtue of
the orthogonality theorem 4.1.1. Each conjugation class defines a linearly independent direction.
Hence it is clear that the number of irreducible representations of G is less or equal to the number of
equivalence classes as otherwise the rows would not be linearly independent. In fact they are equal:
Theorem 4.2.1 Character table is a square: The number of irreducible representations of G is
equal to the number of conjugacy classes of G.
An equivalent statement is that the characters form an orthonormal basis for the class function space.
Much alike exp(2πix/Ln), for n ∈ Z, form a basis for periodic function f (x + L) = f (x) in Fourier
theory. The proof of this theorem will be presented below but is not really part of the course and is
given for the interested reader only. It is powerful but a bit abstract for an introductory course on
the subject.

Proof: Assume that there was a class function f : G → C (c.f. definition 4.1.1) s.t.
hf |χ(ρi )i = 0 (4.22)
on all irreducible representations. If we can show that f = (0, ..., 0) ∈ CNc , then the theorem
is proven as this implies that the characters form a complete set of vectors on the class function
space. Consider the following map,
X
φ= f (g)∗ · ρi (g) : Vi → Vi , (4.23)
g∈G
on the vector space Vi of the irreducible representation ρi . From ρi (g 0 )φ = φρi (g 0 ) and Schur’s
lemma (theorem 3.3.2) it follows that φ = λ1 and we may write:
X (4.22)
λ dim(Vi ) = Tr[φ] = f (g)∗ χ(ρi (g)) = |G|hf |χ(ρi )i = 0 . (4.24)
g∈G
Hence either λ = 0 (i.e. f (g) = 0 for ∀g ∈ G) or g∈G f ∗ (g)ρi (g) = 0 has to hold on any
P
irreducible representation and by virtue of decomposability on any representation; in particu-
lar the regular representation ρR (g) (which is a direct sum of the irreducible representations).
Since all ρR (g) are linearly independent by definition this enforces f (g) = 0 for all g ∈ G and
this completes the proof q.e.d.Note, we have therefore shown that no f other than f = (0, .., 0)
obeys (4.22).
Another powerful theorem, somewhat reminiscent, of Lagrange’s theorem is:

Theorem 4.2.2 The dimension of any irreducible representation divides the order of the group:
|G|
∈ Z+ , ρi is an irreducible representation on Vi (4.25)
dim Vi
We shall omit the proof1 and state simple corollary:
Corollary 4.2.1 All irreducible representations of an abelian group are one dimensional.
Proof: First we notice that for an abelian group each element forms a conjugacy class on its own since
a = g · a · g −1 ⇔ a · g = g · a. Hence for an abelian group there are as many irreducible representations
as there are elements by virtue of theorem 4.2.2. In order to obey the constraint (4.19) all those
P|G| P|G|
irreducible representations are necessarily one dimensional i=1 dim(Vi )2 = i=1 12 = |G| q.e.d.
Exercise 4.2.1 Representations of abelian groups By virtue of corollary 4.2.1 representa-

tions of abelian groups are rather straightforward. You might find the following exercises useful
as they clarify previous matters.
a) Write down all irreducible representations of Z7 . By virtue of corollary 4.2.1 they are one
dimensional and in fact you have been using them in a previous exercise.
b) How many of the irreducible representations of the previous exercise are faithful?
c) How many irreducible representations of Z4 are faithful?
The amount of a priori information on the dimension of the irreducible representations is rather
impressive and is largely grounded in the Schur’s important Lemma (theorem 3.3.2). In summary,
given a finite group its irreducible representation ρi on Vi are constrained as follows:
Diophantine equations for dim Vi of G

(1) |G| = 1 + dim V2 2 + .. + dim Vk 2 , (theorem 4.1.2)
(4.26)
(2) k = Nc ≡ number of conjugacy classes (theorem 4.2.1)
(3) |G|/ dim(Vi ) ∈ Z+ , (theorem 4.2.2)
Note the one in |G| = 1 + .. stands for the trivial representation which is always present. The
equations above, diophantine nature, are in many cases sufficient to obtain the dimensions of the
irreducible representations (without working them out explicitly). From the orthogonality 4.1.1 and
the fact that the character table is essentially a unitary matrix follows a second orthogonality theorem:
1 In fact an even stronger statement applies: |G/Z(G)|/ dim V is an integer where Z(G) is the centre of the group
i
which is always normal in the group itself.
Theorem 4.2.3 Second orthogonality thm: We can define an inner product on the class space,
with symbols as in (4.7), as follows:
Nc
c[g] X
hχ(ρ([g])|χ(ρ([g 0 ])iC = χ(ρi ([g])∗ χ(ρi ([g 0 ]) . (4.27)
|G| i=1
which is orthonormal,
hχ(ρ([g])|χ(ρ([g 0 ])iC = δ[g][g0 ] , (4.28)
in the class directions. The subscript C stands for class and helps to avoid confusion with the previous
inner product (4.7). Note, the slightly asymmetric definition in (4.27) anticipates the result (4.28).
Proof: The proof is constructive and uses the fact that for a unitary n × n matrix U :
U U † = U † U = 1n . (4.29)
The second equation follows from the first by hermitian conjugation. The matrix U in our case is
1/2
ck
Ua[g] = χ(ρa ([g]) , (4.30)
|G|
where we remind the reader that a and [g] are indices referring to an irreducible representation and
a class representative respectively. The first and second orthogonality relations (4.9) and (4.28) can
be written as
X
hχ(ρa )|χ(ρb )i = (U U † )ab = ∗
Ua[g] Ub[g] = δab ,
[g]
Nc
X
hχ(ρ([g]))|χ(ρ([g 0 ]))i = (U † U )[g][g0 ] = ∗
Ua[g] Ua[g0 ] = δ[g][g0 ] , (4.31)
a=1
where we implied the notation (U † )ab = Uba ∗

. Hence (4.30) guaranties that the two orthogonality
relations imply each other. Stated differently: (4.9) ⇔ (4.28) q.e.d.
4.2.1 Character table of S3 as an illustration

We shall exemplify the character table of S3 . The conjugacy classes correspond, for any permutation
group Sn , to the number of ways in which n can be written in terms of integers. The latter is known
as a partition of an integer number n = λ1 + ....λk , λ1 ≥ .. ≥ λk ≥ 1. For S3 we have got:
3=1+1+1 ↔ [()] , 1 element , (4.32)
3=2+1 ↔ [(1, 2)] , 3 elements (4.33)
3=3 ↔ [(1, 2, 3)] , 2 elements . (4.34)
Hence there are three irreducible representations. We can start building the character table:
S3 class () (1, 2) (1, 2, 3)
dim class 1 3 2
trivial 1 1 1 (4.35)
alternating 1 −1 1
.. .. .. ..
S3 class () (1, 2) (1, 2, 3)

dim class 1 3 2
trivial 1 1 1
alternating 1 −1 1
standard 2 0 −1
Table 4.1: Character table of S3 (which is isomorphic to D3 ).
With the class in the first row and the dimension of the class in the second row. The latter is often
omitted. The trivial representation is always present and all its entries are equal to 1. For the
permutation groups Sn there is always a second one dimensional representation which acts as follows:
ρalt (g)~v = sgn(g)~v , (4.36)
where sgn(g) = ±1 depending on whether g is an even or odd permutation. There are many ways
to know that the remaining irreducible representation is two dimensional c.f. exercise 4.2.2. In fact
the permutation representation (3.2) decomposes into the trivial and the two dimensional irreducible
representation : ρperm = ρtriv ⊕ρstand (permutation, trivial and standard have been abbreviated.) The
standard representation for the permutation groups is the non-trivial irreducible representation
contained in the permutation representation.
In the case of S3 the standard representation is two dimensional (c.f. exercise 4.1.2a) ). By virtue
of exercise 4.2.2 the character of the permutation representation is equal to the number of elements
that the operation leaves in invariant it follows that χ(ρperm. ) = (3, 1, 0) in the order of the classes as
in the character table. Therefore
(4.3)
S3 : (3, 1, 0) = χ(ρperm. ) = χ(ρtriv. ) + χ(ρstand ) = (1, 1, 1) + χ(ρstand ) , (4.37)
which yields χ(ρstand ) = (2, 0, −1) and completes the S3 character table 4.1.
Exercise 4.2.2 characters and orthogonality

a) Let ρperm be the permutation representation, as given for S3 in (3.2). Argue that χ(ρperm (g))
corresponds to the number of elements that are left invariant by ρperm (g) acting on its vector
space. The answer can be as short as one sentence.
b) Verify the second orthogonality theorem 4.2.3 on the example of the character table of S3
given in table 4.1.
Exercise 4.2.3 Step by step construction of the S4 character table.

a) Find all the character classes of S4 following the same recipe as for S3 . Hint: You should
find that there are five of them.
b) Argue that the irreducible representations satisfy the following equation:
24 = 12 + 12 + d23 + d24 + d25 (4.38)
Hint: the two 12 correspond to two one dimensional irreducible representations. Which
ones?
c) Find the dimension d3,4,5 of the remaining three irreducible representations. The equation
are “diophantine enough” to be solved.
d) Begin to build up the character table by writing down the first two irreducible representa-
tions.
e) Find the character of the standard representation using the same method as for S3 (recall:
ρperm = ρtriv ⊕ ρstand ).
f) You can find another representation by multiplying the standard representation by the
non-trivial one dimensional representation.
g) You can find the character entries of the last irreducible representation by making sure that
it is orthogonal to the other entries. Hint: the class-characters assume the following values:
{0, 0, −1, 2, 2}.
4.3 Restriction to a subgroup – branching rules

Before addressing the problem of restriction to a subgroup it will prove useful (in this context as well
as for the rest of the course) to expand on the definitions of real and complex representations given
in definition 3.2.1.
4.3.1 Complex and real representations

For a complex representation the complex conjugate representation ρ̄a = ρ∗a is inequivalent. For a real
representation one can find a basis where ρa is real and therefore two complex conjugate representation
are equivalent to each other. This knowledge gives a simple criterion, which we call the complex-real
criterion, for determining the type of representation2

1 X 0 ρa complex
hχ(ρa )|χ(ρ̄a )i = χ(ρa (g)2 )∗ = . (4.39)
|G| 1 ρa real
g∈G
The first equality is by no means trivial but follows from the decomposition of V ⊗ V into a symmetric
and antisymmetric part as discussed in the footnote. In practice representations are often denoted by
their dimension and if there is ambiguity then one uses prime or other symbols to distinguish them.
In this spirit we write the complex conjugate representation as
complex conjugate representation of [3] ≡ 3̄ , (4.40)
where we have taken a three dimensional representation as an example for the sake of explicitness.
4.3.2 Branching rules

When a representation ρ(g) for g ∈ G is restricted to a subgroup H ⊂ G, ρ(h) certainly forms a
representation since ρ is still a group homomorphism on a linear space. The crucial point is though
that it is not necessarily an irreducible representation. By virtue of the decomposability theorem the
representation of G decomposes into irreducible representations of H
ρ(g)|H = m1 ρ1 (h) ⊕ ... ⊕ mk ρk (h) . (4.41)
where k is the number of equivalence classes of H.3

Branching rules can be constructed systematically from the character tables of the group G and
H. In some cases, rather often, one is able to guess the branching rules with less information. We
illustrate this after pursuing the systematic construction.
We consider the example of Z3 ⊂ S3 and identify the branching rules. In doing so many concepts
of the previous sections get illustrated. First of all we note that since Z3 is abelian, c.f. corollary
4.2.1, the irreducible representations of Z3 are all one dimensional. The group Z3 is generated by
either (1, 2, 3) or (1, 3, 2) and has therefore no transpositions (2-cycles). We denote the irreducible
2 For completeness (non-examinable) let us mention that there is in addition there is a pseudo-real (or quaternionic)
type representation where ρ̄a = S −1 ρa S exists but it P is not a real representation. Fortunately there is a criterion for
1 2
pseudo-reality: ρ(g) is pseudo-real if and only if |G| g∈G χ(ρ(g) ) = −1. Note hχ(ρ)|χ(ρ̄)i = 1 for a pseudo-real
representation. Pseudo-real representations will be discussed in the context of the Lie group SU (2) in the following
chapter. In case you want to understand of why a pseudo-real representations are a logical possibility you may want to
consider exercise 3.38 in [Ful91]. The essential idea is that if the V is the vector space of a non-complex representation
then V ⊗V contains a trivial representation. The tensor product V ⊗V = Sym(V ⊗V )⊕Asym(V ⊗V ) decomposes into a
symmetric and antisymmetric subspace similar to the case of a matrix that can be decomposed uniquely into a symmetric
plus a antisymmetric matrix. In the case where the trivial representation falls into the Sym(V ⊗ VP ) the representation
1 2
is real and if it falls into Asym(V ⊗ V ) it is pseudo-real. The criterion for pseudo-reality is |G| g∈G (ρ(g )) = −1.
An example is given by the two dimensional representation of the Quaternions. The Quaternions are defined through
the presentation Q8 =< I, J, K, −e | I 2 = J 2 = K 2 = IJK = −e, (−e)2 = e >. The two dimensional representation is
given by ρ2 (I) = iσ3 , ρ2 (J) =Piσ2 and ρ2 (K) = iσ1 where σi are the well-known Pauli-matrices to be defined later on.
1 2 1
One readily obtains that |G| g∈G χ(ρ2 (g) ) = 8 (2 · 2 + 6 · (−2)) = −1.
3 In the physics literature such decompositions are frequently called branching rules. Branching rules are important
in the case where a large symmetry breaks into a smaller one such as in grand unification e.g. SU (5) → SU (3) ⊗
SU (2) ⊗ U (1) with the latter being the gauge group of the so-called Standard Model.
representations by 1Z3 , 10Z3 , 100Z3 . The irreducible representations of 10Z3 , 100Z3 are easily guessed to be
ρ(1, 2, 3)|10Z = e2πi/3 ⇒ ρ(1, 3, 2)|10Z = e−2πi/3 ,

3 3
ρ(1, 2, 3)|100Z = e−2πi/3 ⇒ ρ(1, 3, 2)|100Z = e2πi/3 . (4.42)

3 3
Notice that the 10Z3 , 100Z3 are complex representations (c.f. definition 3.2.1). In the case under discussion
they are necessarily complex conjugate representations of each other since they originate from S3 whose
representations are real. In terms of equations, using the notation introduced in Eq. (4.40),
0 00
1̄Z3 = 100Z3 , 1̄Z3 = 10Z3 , (4.43)
where the second equation is just the complex conjugate of the first one. This knowledge can help in
guessing the branching rules as we shall see below.
The characters are equal to the expressions in (4.42) since the representation is one dimensional.
Below we give the character table for Z3 and reprint the character table of S3 in tables 4.2 and
2.2 respectively. Note the elements (1, 2, 3) and (1, 3, 2) which are in the same equivalence class in
S3 constitute different equivalence classes in Z3 . This happens frequently and is due, in this case,
to the fact that the transpositions (x, y), which guarantee the equivalence under S3 are not in Z3 !
Given both character tables one can compute the branching rules by taking scalar products of the
Z3 class () (1, 2, 3) (1, 3, 2) S3 class () (1, 2) (1, 2, 3)

dim class 1 1 1 dim class 1 3 2
1 1 1 1 trivial (1) 1 1 1
10 1 e2πi/3 e−2πi/3 alternating (10 ) 1 −1 1
standard (2) 2 0 −1
100 1 e−2πi/3 e2πi/3
Table 4.3: Character table of S3 reprinted from table
Table 4.2: Character table of Z3 .
2.2 for the readers convenience
corresponding characters. The question of how many times a representation j of Z3 is contained in

the representation i of S3 , M
ρ(S3 )i |Z3 = mij ρ(Z3 )j (4.44)
j
can be obtained by the scalar product of the characters restricted to the subgroup Z3 :
1 X
mij = hχ(ρ(S3 )i |χ(ρ(Z3 )j iZ3 = χ(ρ(S3 )i (h))∗ χ(ρ(Z3 )j (h)) , (4.45)
|Z3 |
h∈Z3
where the right hand side resolves the potential ambiguity in notation between ρ(S3 ) and ρ(Z3 ) in the
same scalar product. This leads to the following branching rules:
branching rules: ρ1S3 |Z3 = 1 · ρ1Z3 , 1S3 → 1Z3

ρ10S |Z3 = 1 · ρ1Z3 , 10S3 → 1Z3
3
ρ2S3 |Z3 = 1 · ρ10Z ⊕ 1 · ρ100Z , 2S3 → 10Z3 ⊕ 100Z3 , (4.46)

3 3
since the only non-vanishing scalar products are (to be verified in exercise 4.3.1):
hχ(ρ1S3 )|χ(ρ1Z3 )i = 1 , hχ(ρ10S )|χ(ρ1Z3 )i = 1 ,

3
hχ(ρ2S3 )|χ(ρ10Z )i = 1 , hχ(ρ2S3 )|χ(ρ100Z )i = 1 . (4.47)

3 3
As promised, when character tables are known the computation of branching rules is straightforward.
The notation on the right hand side of Eq.(4.46) is the one used in the physics literature. The notation
on the left hand side is a bit extensive but presumably good for the sake of clarity and connected to
the initial one in Eq. (4.41).
Alternative reasoning: Often it is possible to guess the branching rules in indirect ways. We shall
illustrate this below.
• General rule: the trivial representation is always mapped to the trivial representation. Hence:
1S3 → 1Z3 .
• The alternating representation 10S3 is distinct from the trivial representation by a minus sign on
odd permutations. Z3 is generated by (1, 2, 3) or (1, 3, 2) (as stated above), which are even and
hence the trivial and the alternating representation are the same when restricted to Z3 . Hence:
10S3 → 1Z3 .
• The standard representation 2S3 has to branch into two one dimensional representations of Z3 .
Given the fact that 2S3 is a real representation it can either split into 2 · 1 or 10 + 100 of Z3 .
0
The split 1 + 10 is not possible because it is not a real combination: (1Z3 + 10Z3 )∗ = 1̄Z3 + 1̄Z3 =
00
1Z3 + 1Z3 . A split into two identities would indicate that 2S3 is not a faithful representation.
This cannot be the case as one representation has got to be faithful (1 and 10 of S3 aren’t) and
therefore 2S3 → (10 + 100 )Z3 is unavoidable.
Exercise 4.3.1 branching rules
a) Compute or confirm the scalar products in (4.47).

b) Make sure you understand that (4.42) are complex representations. To do so you should
consult the definition.
c) Why does a complex representation have at least one character which is complex valued
(by which we mean that it is not real)? Hint: A representation and its complex conjugate
representation are inequivalent to each other i.e. hρa |ρ̄a i = 0.
4.4 Constructing representations

Constructing all irreducible representations is not the main aim of this course. An appreciation of some
of the tools is nevertheless very helpful and this is what we aim for in this section. For example direct
product representations. The latter naturally appear in problems of physics e.g. when two particle in
given representations form bound states. The latter behaves like the irreducible representations of the
product representations. The diophantine rules (4.26) are of great help in starting to construct the
character table as illustrated in section 4.2.1 for S3 . As we shall see for Sn there is a beautiful one to
one correspondence between so-called Young diagrams and all irreducible representations. The magic
of Young tableau’s also migrates, for example, into Lie group representations of SU (n).
4.4.1 A short notice on the irreducible representations of Sn

We are going to state a few basic facts about the irreducible representation of the symmetric group
in this section without giving further proofs. The simplicity of the results and their appeal hint at a
much deeper underlying structure which is beyond the scope of this course.
Figure 4.1: Partition of 5 on the left and in the middle the associated Young tableaus including the hook
length of each square. (The number of vertical boxes in a row corresponds to a summand of the partition.
From top to bottom the size of the corresponding decreases. The hook length is explained in the text). On the
right the dimension of the irreducible representation is computed according to formula (4.48). A non-trivial
check is the verification of the dimensionality theorem in (4.49).
In section 4.2.1 we have stated that the number of the conjugacy classes corresponds to the number
of ways one can partition an integer. By virtue of theorem 4.2.1 this number corresponds to the number
of irreducible representations. In fact to each partition one can associate a so-called Young tableau,
illustrated for S5 in Fig. 4.1. In addition a number, called the hook length, is associated to each
square. The hook length counts the number of squares to the right and below the square including
the square. The amazing property is that the dimension of the irreducible representation corresponds
to
n!
dim Vi [Sn ] = Q , (4.48)
hook lengths
where Π stands for the product (of all hook lengths). All of which, the partition, the Young tableau,
hooks and formula (4.48), are illustrated in Fig. 4.1 for S5 . A non-trivial check is whether it satisfies
the dimensionality theorem 4.1.2:
|S5 | = 5! = 120 = 2 · 12 + 2 · 42 + 2 · 52 + 62 . (4.49)
Indeed it does! Certain symmetries are apparent in Fig. 4.1 and demand a few explanations. Repre-
sentations whose Young tableau can be mapped into each other by a mirror reflection at an axis going
from top left to bottom right corner correspond to the parity partner representation. For example
n = n, which always corresponds to the trivial representation, is mirror dual to n = 1 + .. + 1 which
corresponds to the alternating representation. The complete set of relations is given by:
mirror
5=5 ↔ 5=1+1+1+1+1,
mirror
5=4+1 ↔ 5=2+1+1+1,
mirror
5=3+2 ↔ 5=2+2+1. (4.50)
Hence the characters of mirror dual representation are
χ(ρi (g)) = sgn(g)χ(ρmirror(i) (g)) (4.51)
with sgn(g) denoting the parity g ∈ Sn . Further we note that the 5 = 3 + 1 + 1 is mirrored to
itself (self-dual in a certain sense). With all this information we could in principle (re)construct the
character table of S5 without too much effort. In view of limited time we shall not do so and move
on, but the interested reader is referred to [Ful91] chapter 3. The program of this section is to be
exercised for S4 , by you, in exercise 4.4.1.
Exercise 4.4.1 Young tableau for S4 etc
a) Repeat the program of determining the number and dimensions of the irreducible represen-
tations for S4 (or any other Sn6=5 ) as outlined in this section. i) write down all partitions
ii) draw the corresponding Young tableaus with hook lengths iii) compute the dimension
according to formula (4.48) and then, finally, check the dimensionality theorem.
b) Argue that χ(ρa (g)) = 0 for g ∈
/ An (i.e. g odd permutation), if ρa is a self-dual (identical
mirror image of Young tableau) irreducible representation of Sn . Hint: If this was not the
case then one could construct a further irreducible representation.
4.4.2 Direct product representations – Kronecker product

So far we have been concerned with finding the smallest building block of representation theory: the
irreducible representations. Tensor product representations ρa ⊗ ρb acting on Va ⊗ Vb is a straightfor-
ward way to generate new representations which are generically reducible but decompose, by virtue
of theorem 3.3.3, into irreducible representations.
First we notice that the direct product of representations is a representation since the (direct)
product of two group homomorphism is still a group homomorphism. By theorem 3.3.3 we have:
M
ρa ⊗ ρb = m1ab ρ1 ⊕ .. ⊕ mkab ρk = mcab ρc , (4.52)
c
where mcab ∈ Z+ are the multiplicities. Direct product of representations are known as Kronecker
products or Clebsch-Gordan series. The former refers more to the left hand side and the latter
to the right hand side respectively. By virtue of properties of direct products and sums (c.f. section
1.3) we have the following sum rule:
X
dimVa · dimVb = mcab dimVc , (4.53)
c
which is yet another diophantine equation.
Motivation for decomposition of direct product representations

We shall motivate the fact that direct products decompose into smaller representations in two dif-
ferent ways: one formal argument (for finite groups) and one physical argument (addition of angular
momenta).
• the largest irreducible representation: The direct product of the largest irreducible representation
of a finite group has to decompose as otherwise it is not the largest irreducible representation.
In concrete cases conflicts with the dimensionality theorem could arise.
• addition of angular momenta: From your course on quantum mechanics you should know that
when two particles of a certain angular momentum form a bound state, the angular momenta
add to a total angular momentum. We shall illustrate this below with two states of angular
momentum l = 1, i.e. states which transform as vectors under SO(3) = {O ∈ M3 (R) | OT O =
I3 , det O = 1}. Strictly speaking we should be addressing this example in the next chapter,
where we discuss continuous groups, but I believe there is some merit in doing it here.
Consider three dimensional vectors ~v ∈ R3 which transforms under a l = 1 irreducible represen-
tations of the rotation group SO(3).4 The direct product of two such irreducible representations,
say ~v and w~ can be written,5
δij δij
vi wj = Aij + Bij , Aij = (vi wj − ~v · w)
~ , Bij = ~v · w
~, i, j = 1..3 , (4.54)
3 3
unambiguously as a sum of a trace free (Aij ) and a trace part (Bij ). Intuitively it is clear that
the trace part, the term on the righthand side, does not transform under rotation since it is
4 Be warned: physicists will often state that ~v in the case above is in a l = 1 irreducible representation of SO(3).
5 We omit the ⊕ sign in the equation below. Strictly speaking one should write vi ⊗ wj etc.
a scalar product. Formally this follows from δij being just a unit matrix which multiplied by
any matrix (representation) yields back the matrix. Hence the term is a scalar (= invariant)
corresponding to l = 0 irreducible representation of SO(3).6
Multiplicities from the scalar product on characters

The multiplicity mcab ∈ Z+ , according to corollary 4.1.2 b), is is given by
1 X
mcab = hχ(ρa ⊗ ρb )|χ(ρc )i = χ(ρa (g) ⊗ ρb (g))∗ χ(ρc (g)) (4.55)
|G|
g∈G
(4.4) 1 X
= χ(ρa (g))∗ χ(ρb (g))∗ χ(ρc (g)) . (4.56)
|G|
g∈G
Now follows a corollary of immense practical use:

Corollary 4.4.1 The Kronecker product of two irreducible representations ρa,b contains the identity
(with multiplicity one),
ρa ⊗ ρb = 1 · ρ1 ⊕ ... (4.57)
if and only if ρb = ρ̄a (ρb is the complex conjugate representation of ρa ). Note: if the ρa,b are not
irreducible then the multiplicity of ρ1 can be larger than one.
Proof:
(4.55) 1 X Cor.4.1.2e) 1 X
m1ab = χ(ρa (g))∗ χ(ρb (g))∗ χ(ρ1 (g)) = χ(ρa (g))∗ χ(ρb (g))∗ = 1
|G| |G|
g∈G g∈G
Cor.4.1.2c)
⇔ χ(ρb (g))∗ = χ(ρa (g)) , ∀g ∈ G , (4.58)
where the last line is the very definition of the complex conjugate representation. q.e.d.
Generalisations of corollary 4.4.1 to reducible representations are immediate and left to the reader.
Remark: To this end we note that Kronecker products allow us to understand what we called mirror
representation in section 4.4.1 in the following manner:
ρmirror(a) = ρ10 ⊗ ρa , G = Sn (4.59)
where 10 is the alternating representation ρ(g)10 = sgn(g).
6 Risking an outlook towards the next chapter: the full story is that in the trace free part there is a l = 1 (antisym-
metric) and l = 2 (symmetric) irreducible representation . In fact we know that one can get a vector from two vectors
x = ~v × w
via the cross product ~ ~ and this corresponds to the l = 1 part. The qualitative discussion above is also coherent
with the formula of addition of angular momenta |~l1 + ~l2 | = |l1 − l2 |, ...l1 + l2 = 0, 1, 2 for l1 = l2 = 1. The reason
why angular momenta add when representations are multiplied will become clear when we discuss the corresponding
Lie group and Lie Algebra of SO(3).
Exercise 4.4.2 Exercises on Kronecker products

a) Check that
10Z3 ⊗ 100Z3 = 1Z3 (4.60)
0
and hence 1̄Z3 = 100Z3 and vice versa.
In an earlier version of the notes we did state the Wigner-Eckart theorem (valid for any group
with finite dimensional representation) at this stage. In order to reduce the level of abstraction we
shall solely state the Wigner-Eckart theorem for SU (2)(SO(3)) in section 6.3 with brief comments on
the general case stated thereafter.
4.5 A few sample applications

4.5.1 Distortion of lattices
Consider a model of a solid with a cubic lattice where you can imagine the atoms to be located, in
equilibrium, at the corners of a cube c.f. figure 4.2 According to section 2.4 and table 2.3 such a
lattice exhibits a S4 symmetry instead of the SO(3)-rotation symmetry of the free space.
Let us consider a cartesian coordinate system whose coordinates are aligned with the sides of the
cube. We consider two type of distortions for which we are interested in the reduction of the symmetry
group S4 → H.
• A distortion in the (1, 0, 0)-direction: S4 → D4 . The effect on the symmetry is like losing
a direction and hence the symmetry reduces to the one of the planar square. The latter (c.f.
figure 2.6) corresponds to the dihedral group D4 of eight elements.
• A distortion in the (1, 1, 1)-direction along the diagonal: S4 → D3 . The effect is similar to the
previous case though not quite as obvious. Imagine looking at a cube onto a vertex then you
would see the three faces of that cube. You can convince yourself that the symmetry corresponds
to the symmetry of a triangle. The latter corresponds to the dihedral group D3 of six elements
according to c.f. figure 2.6.
4.5.2 Fermi and bose statistics

Identical particles in quantum theory give rise to rather curious phenomenon. In classical theory we
may follow the trajectory of each particle and one may therefore attach a definite number to each
particle.
In quantum theory this is not possible anymore. Let us imagine an ensemble of non-interacting
identical particles. States of that space may be written as tensor product of single particle states.
E.g. |φ1 ⊗ .. ⊗ φN i with φi being a single particle state. The generic state vector can then a
priori be in any irreducible representation of the symmetric permutation group SN . How do we
know which representation describes which type of particle configuration? In particular if N is large
Figure 4.3: Possible arrangements (irreducible representations) of an identi-

Figure 4.2: Lattice with cu- cal particle state. (right) trivial representations as chosen by identical bosons.
bic symmetry with only one shell (middle) mixed representations do not appear in nature or are equivalent to
shown. boson and fermion statistics times a global symmetry, (left) alternating rep-
resentation as chosen by identical fermions.
(N ' 1023 Avogadro number for systems of macroscopic size) then the number of choices of irreducible
representations is rather large.
A very important fact of quantum theory: If the identical particles are bosons (i.e. particles of
integer spin s = 0, 1, 2...) then the wave function is in the trivial representation and if the identical
particles are fermions (i.e. particles of half integer spin s = 1/2, 3/2...) then the wave function is in
the alternating representation. For N = 2 we have got:
1
Φ12 |bosons = √ (|φ1 ⊗ φ2 i + |φ2 ⊗ φ1 i) ,
2
1
Φ12 |fermions = √ (|φ1 ⊗ φ2 i − |φ2 ⊗ φ1 i) (4.61)
2
This is the essence of it. A few comments seem in order:
• This implies that no two fermions can be in the same state as otherwise the wave function
vanishes on grounds of antisymmetry. This entails Pauli’s famous exclusion principle that
was postulated to give the atomic a shell model a more solid foundation within non-relativistic
quantum mechanics. The so-called occupation number nEα of a given fermion state of fixed
energy Eα is either nfermion
Eα = 0, 1 whereas for bosons nboson
Eα = 0, 1, 2, 3, .. it can be any integer.
• The fact that all the bosons can condense into (or occupy) the same state at low energies has
spectacular consequences and is known as Bose-Einstein condensation. The crucial bit is that
the bosons are all coherent in the ground state and act as one entity. This is the basis of such
phenomenon as superfluidity, superconductivity and many others.
For fermions the fact that the occupation number per state is either one or zero leads to the so-
called zero point pressure. In neutronstars (neutrons are fermions) it is the zero point pressure
of the fermions that balances the gravitational attraction and avoids the star imploding into a
black hole.
• The general connection between spin and statistics is called, not surprisingly, the spin statis-
tics theorem. It’s proof is deep down in the structure of quantum field theory. If one tries to
go the other way then one can show (Pauli 1940) that this leads to violation of Einstein causality
(commutators of quantum field have support for space-like separation points). The most rigor-
ous proof is found in axiomatic quantum field theory (Jost 1957) and uses a complexification
(analytic continuation) of the Lorentz symmetry group SO(3, 1).
• In fact at first it might seem somewhat surprising that of the many representation of the symme-
try group of SN (for N large) there are only two types which are realised for identical particles.
The other representation (mixed representations) were also considered, and referred to as paras-
tatistics, but were found to be equivalent to the trivial and alternating representation times
a global symmetry and therefore do not seem to bring in a new element. Note that adding
fermions and bosons coherently is not possible because the boson and fermion number is a con-
served quantum number. This is known as a superselection rule and will be briefly discussed
towards the end of the course.
Exercise 4.5.1 Invariants and mixed representations.
a) The group S4 is generated by

     
0 1 0 0 0 1 1 0 0
A = 0 0 1 , B = 0 1 0 , C = 0 −1 0 . (4.62)
1 0 0 1 0 0 0 0 −1
Note, in order to check invariance of a potential invariant it is sufficient to check invariance

for a set of group generators. Check that A, B, C both leave I2 = x2 + y 2 + z 2 I3 = xyz
invariant.
b) Explicit construction of mixed representation We shall construct the standard representation

of S3 (which corresponds to a mixed representation) explicitly. The construction is similar
to the permutation representation although it is algebraically more involved. This exercise
also contains the seeds to understand the hook formula (4.48) for the Sn .
The representations (4.61) correspond to the trivial and alternating representation of S2 :
S2 : 1: , 10 :
Let us consider the situation for three particles and construct the mixed representation
(parastatistics) of S3 :
2 1
S3 : 2 : 3
The representation is two dimensional and the numbers are not the hook lengths but an
enumeration to be used below.
The meaning of the arrangement of the boxes for Sn is as follow: First take a vector with n
directions say |φi i i = 1..n and then take a tensor product of n of them |φi1 ⊗..⊗φin i and the
boxes give the rules of anti(symmetrisation). Indices associated with rows are symmetrised
and indices which are in one column are antisymmetrised over.
Let us implement this on the following state: |φ1 ⊗ φ2 ⊗ φ3 i (or you can use the notation
|φa ⊗ φb ⊗ φc i if you find it less confusing) and call the state e1 :
e1 = |φ1 ⊗ φ2 ⊗ φ3 i + |φ2 ⊗ φ1 ⊗ φ3 i − |φ1 ⊗ φ3 ⊗ φ2 i − |φ2 ⊗ φ3 ⊗ φ1 i . (4.63)
A second basis vector might be obtained by starting with |φ1 ⊗ φ3 ⊗ φ2 i
e2 = |φ1 ⊗ φ3 ⊗ φ2 i + |φ3 ⊗ φ1 ⊗ φ2 i − |φ1 ⊗ φ2 ⊗ φ3 i − |φ3 ⊗ φ2 ⊗ φ1 i . (4.64)
The action of ρstand. (1, 2) = ρ(1, 2) on e1 , e2 is then just to interchange labels 1 and 2 (or a
and b in the alternative notation).
S4 class [()] [(1, 2)] [(1, 2)(3, 4)] [(1, 2, 3)] [(1, 2, 3, 4)]
size class 1 6 3 8 6
χ(ρ1 ) 1 1 1 1 1
χ(ρ11 ) 1 -1 1 1 -1
χ(ρ31 ) 3 1 -1 0 -1
χ(ρ32 ) 3 -1 -1 0 1
χ(ρ2 ) 2 0 2 -1 0
Table 4.4: Character table of S4 as obtained in a previous exercise. Note 10 ≡ 11 as with regard to
previously used notation.
– Show that this leads to

.. −1
ρ(1, 2) = (4.65)
.. −1
with dots to be completed by you. Obtain ρ(2, 3) (which takes on a rather simple form
by construction) and ρ(1, 2, 3). Verify that

.. 1
ρ(1, 2, 3) = ρ(1, 2)ρ(2, 3) = (4.66)
.. 0
is indeed satisfied by explicit matrix multiplication.

– (optional) For those who feel that they need further practice can obtain the other
elements by matrix multiplication using the rules in the multiplication table given in
the lecture and then check that the representation is indeed irreducible.
c) Practice Kronecker products. Verify a few of the following Kronecker products
31 ⊗ 31 = (1 ⊕ 2 ⊕ 31 )s ⊕ (32 )a ,
31 ⊗ 2 = 31 ⊕ 32 ,
31 ⊗ 32 = 11 ⊕ 2 ⊕ 31 ⊕ 32 ,
using the S4 character table 4.4. Note the subscripts s and a stand for the symmetric and
antisymmetric subspace and are not of importance for this exercise per se.
Chapter 5
Lie groups: U (1) ' SO(2),

SO(3) ' SU (2)/Z2
Lie groups were discovered by the Norwegian mathematician Sophus Lie (1842-1899) in attempting
to find a Galois theory (solving algebraic equations) for differential equations. Technically they
correspond to manifolds which are compatible with a group structure. By no means does every
manifold admit a group structure.
In some sense Lie groups are more complicated than finite groups but the classification of the
simple and compact Lie groups due to Cartan and Dynkin is remarkably simple and ingenious at the
same time. The rigidity of the group axioms and the manifold structure are restrictive enough for a
short and concise classification unlike the thousand of research papers that lead to the classification of
simple finite groups. The key observation is that the linearisation of the Lie group, which corresponds
to the so-called Lie Algebra, is sufficient to study many of the properties of the Lie groups. Hence
the apparatus of linear algebra comes into play and allows to make strong statements.
We will only be concerned with the four Lie groups U (1), SO(2), SU (2) and SO(3) which are
related, as groups as follows:
U (1) ' SO(2) , SO(3) ' SU (2)/Z2 , (5.1)
where the symbol ' stands for isomorphic. In the first section 5.1 we shall, nevertheless, keep
the definitions general so that you can apply them in your further studies to more complicated Lie
groups. Lie groups are of major importance in particle physics in term of gauge symmetries (the
Standard Model, that is to say the most fundamental theory of particles, has a gauge symmetry
U (1) ⊗ SU (2) ⊗ SU (3)) or so-called global symmetries (e.g. isospin SU (2) the approximate symmetry
under interchange of up and down quarks).
5.1 Basic definitions

Definition 5.1.1 Continuous group: a continuous group is a group G with an infinite number
of elements which depends continuously on some parameters {α} ≡ {α1 , .., αdimG }. The number of
continuous parameters dimG is known as the dimension of the group.
56
CHAPTER 5. LIE GROUPS: U (1) ' SO(2), SO(3) ' SU (2)/Z2 57
Note that it does not make sense to talk of the order of the group as it is simply infinite.
Definition 5.1.2 Lie group: a Lie group is a continuous group which admits an analytic structure
in the parameters {α}. (Therefore corresponds to a manifold).
Examples of Lie groups are:

Examples 5.1.1 Lie groups acting on ~x ∈ R3 1
a) Spatial translations, xi 7→ xi + ai ; dimG = 3 (three parameters ai correspond to the αi above).

b) Spatial rotations (O(3)), xi 7→ Rij (θ)xj ; RT R = I; dimG = 3 (three Euler angles θi correspond
to the αi above).
Definition 5.1.3 Basic characterisation of Lie groups: Lie groups are distinguished by several
characteristic properties, which we list in the table below. In addition we indicate the properties of
the groups we will be concerned with in this course.
properties U (1) SO(2) SU (2) SO(3)
finite or infinite dimensional (as a manifold) f f f f
real or complex (as a manifold) r r r r
compact or non-compact (as a manifold) c c c c
simply-connected, connected or disconnected (as a manifold) c c sc c
simple, semi-simple or composite s s s s
Of these properties the first two should be clear and the last three demand some explanation. Further
the statement that SU (2) is complex is refined to SU (2) being pseudo-real later on.
Compactness On a metric space (a space with a distance) compact is equivalent to closed2 and
bounded (finite distance). For example [a, b] is compact whereas the intervals [a, b[, ]a, b] or ]a, b[ are
not compact. More concretely and relevant for our discussion: the circle S 1 is compact where the real
line R is not. Hence example 5.1.1a) is not a compact Lie group whereas example 5.1.1b) is a compact
Lie group. We will only be concerned with compact Lie groups in this course. The important point
about compact Lie groups are that all the theorems in chapter 3, such as Maschke’s theorem (unitary
representations),
P decomposability etc, apply.R This is the case since the replacement of the discrete
sum g∈G with the so called Haar measure d{α} is unproblematic, in the sense, that all proofs are
essentially the same. For pedagogic reason let us give a counterexample for a non-compact group.
Examples 5.1.2 A non-compact non-decomposable Lie group Consider the translation group
R then the following representation,

1 a
ρ(a) = , ρ(a)ρ(b) = ρ(a + b) (5.2)
0 1
but does not decompose subspaces. The direction ~ex = (1, 0) is an invariant subspace but the
orthogonal complement ~ey = (0, 1) is not. This is due to the fact that the translation group is
non-compact!
1 Be warned the fact that dimG = 3 and dimR3 = 3 is not a rule (though certainly related). The rotation group in
two (O(2)) and four dimension (O(4)) have got dimO(2) = 1 and dimO(4) = 6 respectively.
2 Every converging sequence converges to a value in the set: C closed, a ∈ C, a = lim
n n→∞ an ∈ C.
Connectedness A set is connected if there is a smooth path from each point of the set to any
other point of the set. It is in particular simply connected if all loops can be contracted in a smooth
fashion to a point. A set which is not connected is disconnected. In Fig. 5.1 three illustrative and
hopefully clarifying examples are given. We shall only be concerned with connected Lie groups. We
Figure 5.1: (left) simply connected set (middle) connected set which is not simply connected (right) discon-
nected set
will briefly discuss an example of a non-connected group in the next chapter in example 5.2.1 so that
you will get an idea of what is going on. The fact that SU (2) is simply connected and SO(3) only
connected is related to the existence of fermions.
Simplicity A simple Lie group is a connected non-abelian Lie group which does not admit (non-
trivial) connected normal subgroups. This connects to the definition of a simple group for finite
groups. Similarly simple Lie groups are the atoms of Lie group since no (Lie) quotient groups can be
defined. Semi-simple Lie groups are direct products of simple Lie groups. Composite Lie groups are
composed of simple Lie groups through semi-direct products or more complicated operations and are
beyond the scope of this course. We shall only be concerned with simple Lie groups3 in this course.
Some examples of simple and compact Lie groups used in physics are:
Examples 5.1.3 Examples of finite, simple and compact Lie groups are:
- Special orthogonal groups are real.4
SO(n) = {O ∈ GL(n, R) | OT O = In , det O = 1} (5.3)
- Special unitary groups are complex.
SU (n) = {U ∈ GL(n, C) | U † U = In , det U = 1} (5.4)
- Symplectic groups are complex.

T 0 In
U Sp(2n) ≡ Sp(n) = {S ∈ GL(2n, R) | S JS = J , J = } (5.5)
−In 0
play a role in Hamiltonian dynamics. Note J 2 = −I2n and hence J is in some sense a square
root of minus one.
3 Including the abelian Lie groups in the simple category.
4 The S-denotes the determinant condition.
G dim rank
An [SU (n + 1)] (n + 1)2 − 1 n
Bn [SO(2n + 1)] (2n + 1)n n
Cn [U Sp(2n)] n(2n + 1) n
Dn [SO(2n)] n(2n − 1) n
E6 78 6
E7 133 7
E8 248 8
F4 52 4
G2 14 2
Table 5.1: Dimension and rank (which is the dimension of the Lie Algebra, c.f. section 5.3) of simple and
compact Lie groups. The so-called exceptional groups E6,7,8 , F4 and G2 are mentioned for completeness only
and do not play any rôle in the remainder of the course. They are called exceptional because they do not fall
into any of four first series. The corresponding Dynkin diagrams are shown on the right for completeness only
without any in depth explanation other than the remark that the number of circles corresponds to the rank.
Symmetries can be inferred from the Dynkin diagrams. For example some types of low ranks are degenerate.
E.g. A1 ' B1 (SU (2) ' SO(3) which is relevant for the course). Or D2 ' A1 ⊕ A1 (SO(4) ' SU (2) ⊗ SU (2))
which is the only example of a semi-simple type in this list.
The groups U (n) and O(n) are the same as SU (n) and SO(n) except that the determinant condition
is not imposed. (The “S” stands for special). A relation between the two is given in (5.17) later
on. The classification of the simple compact Lie groups has been done by Cartan and Dynkin and is
astonishingly beautiful and simple. The study of which is beyond the scope of this course. In table 5.1
we have listed all groups for the sake of completeness. The dimension of the defining representations
(5.3,5.4,5.5) is given in the middle column and the rank (to be defined in section 5.3) is given in the
right column. For pedagogical reason let us give an example of a non-compact Lie group of importance
in physics:
Examples 5.1.4 The Lorenzt-group SO(3, 1): a non-compact Lie group
SO(3, 1) = {Λ ∈ M4 (R) | ΛT ηΛ = η , det Λ = 1} , η = diag(1, −1, −1, −1) . (5.6)
A way to see that they are not compact is to note that they contain the boost to any reference frame
of relative speed v ∈ [0, c[ where c is the speed of light. This is a non-compact interval and therefore
SO(3, 1) cannot be a compact group.5
Much of the power of Lie groups relies on the realisation that, for many aspects, it is sufficient
to linearise the problem in the parameters {α} mentioned earlier. By linearising one departs from
the Lie group to the Lie Algebra which may be thought of as the tangent space of the manifold.
By exponentiating the Lie Algebra elements one gets back the connected part of the Lie group (by
which we mean the part connected to the identity). These concepts will be illustrated through the
simple example of U (1) in the next section. The end this introduction with the following mnemonic:
Linearisation: Lie group ⇒ Lie Algebra ,
Exponentiation: Lie Algebra ⇒ connected part of Lie group (5.7)
5 One consequence of this fact is that there do not exist any finite dimensional unitary representations of the Lorentz
group. (For the experts: this is why Ψ̄ = Ψ† γ0 is necessary to define invariant of the form Ψ̄Ψ where Ψ is a spinor.)
Exercise 5.1.1 Generalities on compact simple Lie groups

a) Check that the dimension of the matrices SO(n) and SU (n) are indeed n2 −1 and n(n−1)/2
in accordance with table 5.1.
Hint: Use the fact that U = exp iH and O = exp iA where H and A are hermitian traceless
and antisymmetric respectively. Note an arbitrary Mn (C)-matrix has got 2n2 parameters.
b) Using the formula det(exp(tr[A])) = exp(tr[A]) check that the determinant condition is
satisfied if H from the previous exercise are traceless.
5.2 The two abelian groups: U (1) and SO(2)

The groups U (1) and SO(2) can be defined as:
U (1) = {z ∈ C | |z| = 1} , SO(2) = {O ∈ M2 (R) | OT O = I2 , det(O) = 1} . (5.8)
They are rather simple, as abelian groups tend to be, and in particular isomorphic to each other.
They correspond to an angle of rotation around 2π. Representation matrices are given by

cos(nα) sin(nα)
ρn (α)|U (1) = einα , ρn (α)|SO(2) = , n ∈ Z , α ∈ [0, 2π[ (5.9)
− sin(nα) cos(nα)
which makes the isomorphism U (1) ' SO(2) explicit. Intuitively they correspond to rotation around
a single axis. All irreducible representations of U (1) are given by one dimensional representation on
C. The irreducible representations of SO(2) are given by two dimensional representations on R2 .
Note that the cyclic groups Zk are finite subgroups of U (1). The Kronecker product (Clebsch-Gordan
series) of two representations is given by
U (1) : ρn ⊗ ρm = ρn+m (5.10)
for U (1) (and therefore for SO(2) as well.)

In principle we could end here but we consider it useful to illustrate the notions of exponentiation,
Lie Algebra and connectedness in this simplified context. Let us make the following ansatz for
O ∈ SO(2)6
O = exp(iαT ) , T ∈ M2 (R) , (5.11)
where the exponential map of matrices can be understood in the sense of a power series or any other
well known limiting process
X An n
A
exp(A) = = lim 1 + . (5.12)
n! n→∞ n
n≥0
6 Mathematicians often omit the factor of i in the exponential and absorb it into the generator. This is in some
ways more convenient as you will notice. We will stick to the convention below which is the one adopted in the physics
literature.
The exponential has the following property: det(exp(A)) = exp(Tr(A)) which can be shown by
considering the eigenvalue expression of the determinant and the trace.
Note that the defining condition (5.9) (using exp(A)T = exp(AT ) which follows from (5.12))
I2 = OT O = exp(iαT T ) exp(iαT )) = I2 + iα(T T + T ) + O(α2 ) ⇔ T T = −T , (5.13)
implies that T has got to be an antisymmetric matrix and this condition immediately generalises to
SO(n).7 The following matrix,

0 1
T = −i , T 2 = I2 , eiαT = I2 cos(α) + iT sin(α) = ρ(α)SO(2) , (5.14)
−1 0
leads to the representation (5.9). The generator T emerges from the linearisation of eiαT = 1 +
iαT + O(α2 ). In the case at hand the Lie Algebra is rather trivial since it has only a single element T
and no further structure. Nevertheless it serves us well in illustrating or hinting at the connection with
differential geometry in terms of manifolds (surfaces correspond to manifolds in two dimensions). In
Fig. 5.2 we illustrate the manifold nature of SO(2) and its tangent space. It is meant as an illustration
only and is of no importance for the rest of the course other than broadening the viewpoint.
Figure 5.2: SO(2)-manifold S 1 embedded into R2 . (S 1 corresponds to a circle). Shown are two maps from
intervals on R which cover the manifolds. An explicit realisation of f1 is given by f1 (θ) = (cos(θ), sin(θ) for
example. It is explicitly seen that T , the (Lie Algbra) generator of SO(2), corresponds to the tangent vector.
Illustrated for (x, y) = (−1, 0) (in green) which leads to T (−1, 0) = (0, −1).
Another element which wish to illustrate is (dis)connectedness.

Examples 5.2.1 O(2): A disconnected Lie group. The manifold SO(2) is connected as can be
seen from the figure but O(2) is not. Note that det(O) = ±1 follows from OT O = I2 . Since O is
differentiable, one cannot go in a continuous fashion from a point in the set det(O) = 1 to a point of
7 Note the last step on the left hand side in (5.13) is only valid if T commutes with its transpose as otherwise the
Baker–Campbell–Hausdorff formula (to be discussed soon) has to be applied. Since T is an antisymmetric matrix it
does commute with its transpose and so the steps are finally consistent.
the set det O = −1 and hence O(2) is not connected.8 For example

1 0
P = , P ∈ O(2) , P ∈ / SO(2) (5.15)
0 −1
is an example of a matrix which is in O(2) but not in SO(2). One may think of P as a parity operator.
In fact P generates the cyclic group Z2 which is a normal subgroup9 of O(2) and the following relation
holds:
SO(2) ' O(2)/Z2 , (5.16)
and slightly more generally
SO(n) ' O(n)/Z2 , SU (n) ' U (n)/U (1) . (5.17)
The symbol ' stands for isomorphic.
It seems worthwhile to pause for a moment and try to reconnect to the chapter of finite groups.
First we note that there ought to be, in general, an infinite amount of irreducible representations by
virtue of a limiting process of the dimensionality theorem 4.1.2 (|G| → ∞ implies infinitely many
irreducible representations in Eq. (4.19)). For U (1) the series is given in Eqs. (5.9). For U (1) the
Haar measure.10 is rather simple. An inner product is defined as follow11 :
Z 2π Z 2π
dα dα iα(m−n)
hχn |χm i|U (1) ≡ χ(ρn (α))∗ χ(ρm (α)) = e = δmn , (5.18)
0 2π 0 2π
under which the χn are orthonormal. The characters will not play a pivotal rôle in the analysis of of
the representations as we will find that the so-called highest weight representations of Lie Algebra to
be a powerful tool to construct representations.
Exercise 5.2.1 U (1) and SO(2)
a) Check that ρn (α1 )SO(2) ρn (α2 )SO(2) = ρn (α1 + α2 )SO(2) using the representation (5.9).
Trigonometric identities can easily derived using Euler’s formula eia = cos(a) + i sin(a).
b) Using Euler’s formula verify (5.14).
c) Conceptual question: what would happen in case n is not chosen to be an integer in (5.9)?
8 If one were to complexify the group which merely amounts to allow for complex {α} then the determinant would take
on any complex value and one could go from any point to another in a continuous fashion. Hence the complexification
of O(2) is connected. This general knowledge is encapsulated in Jacques Hadamard’s (1865-1963) famous quote: “the
shortest path between two truths in the real domain passes through the complex domain”.
9 In particular any abelian subgroup is normal. Hence Z , which is abelian, is normal.
2
10 The Haar measure generalises 1
P
|G| g∈G from finite groups.
11 The second orthogonality relation becomes hχ[α]|χ[α0 ]i | 1 P in(α−α0 ) = δ(α − α0 ) for α, α0 ∈ [0, 2π[.
C U (1) = 2π n∈Z e
5.3 Some generalities of (non-abelian) simple compact Lie

groups
In this section we would like to clarify and elaborate on the aspect of the necessity of the exponential
map and the structure of the Lie Algebra. The exponentiation of the latter gives rise to the group
structure.
Theorem 5.3.1 exponential map
Each one-parameter subgroup of GL(n, C) is given by a matrix exponential ρT (t) = exp(tT ) where
T ∈ Mn (C) 6= 0 is the generator.
Proof: Consider a one-parameter homomorphism ρT (t) : R → GL(n, C) defined by:
d
ρT (t)|t=0 = T , (5.19)
dt
with boundary condition ρT (0) = I. Since ρT is a group homomorphism
ρT (s + t) = ρT (s)ρT (t) . (5.20)
Differentiating w.r.t. to s and setting s = 0, leads to

d
ρT (t) = T ρT (t) ⇒ ρT (t) = exp(tT ) , (5.21)
dt
a first order linear differential equation with unique solution as indicated q.e.d.
Next we want to explore what constraints the group axioms impose on the structure of the gen-
erator T in general. Higher dimensional groups are parameterised by several parameters as discussed
at the beginning of the chapter (e.g. the three-dimensional rotation group SO(3) is described by the
three Euler angles.). Motivated by the exponential form we start with the following ansatz:
1
ρ1 = I + iαa Ta + (iαa Ta )2 + O(α3 ) ,
2
1
ρ2 = I + iβ b Tb + (iβ b Tb )2 + O(β 3 ) ,
2
1
ρ3 = I + iγ c Tc + (iγ c Tc )2 + O(γ 3 ) , (5.22)
2
where summation over repeated indices a, b and c is understood a = 1..dLA . (In the case of SO(3),
αa correspond to the three Euler angles.) Further we note that the precoefficient of the quadratic
term follows from ρ(α)2 = ρ(2α) which in turn derives from (5.20) or generally from theorem 5.3.1.
The matrices Ta are the Lie Algebra generators.
Multiplication of two exponential of matrices is different from normal multiplication of two expo-
nentials exp(a) exp(b) = exp(a + b) for a, b ∈ C in that the two matrices do not necessarily commute.
The formula governing the arithmetics is the famous Baker–Campbell–Hausdorff formula:
1 1 1
eA eB = eA+B+ 2 [A,B]+ 3! 2 ([A,[A,B]]+[[A,B],B])+··· , (5.23)
where [A, B] = AB − BA denotes the commutator. In order to assess the structure of the Lie Algebra
we are going to investigate under what conditions
ρ3 = ρ1 ρ2 ρ−1 −1
1 ρ2 , (5.24)
is consistent with the linear ansatz in (5.22). Eq. (5.24) can be considered as the definition of the
element ρ3 for instance. In order to determine the conditions on the αa , β b , γ c ’s we substitute (5.22)
into (5.24) which leads to
[αa Ta , β b Tb ] = −iγ c Tc + O(α3 , β 3 , γ 3 ) , (5.25)
where all lower order terms cancel. The solution is given by γ c = −fab c αa β b provided that the
generators satisfy the so-called Lie Algebra
Generic Lie Algebra: [Ta , Tb ] = ifab c Tc . (5.26)
Summation over c is again understood and fab c are known as the structure constants of the Lie
Algebra which are in particular antisymmetric in the indices a and b. This answers the question what
kind of structure the generators have to satisfy in order to be compatible with the group structure.
Crucially the Lie Algebra only depends on the commutator, as is inherent in the Baker–Campbell–
Hausdorff formula (5.23), and not on the anti-commutator. The latter depends on the representation
of the Lie Algebra.12
An important notation is the rank of a Lie Algebra , which is the maximal number of Lie Algebra
generators that commute with each other (and not to be confused with the dimension of the Lie
Algebra dLA ≥ rank). For the simple compact groups this number is indicated in table 5.1 on the
right hand side and corresponds to the number of dots in the Dynkin diagrams. Its instrumental in
constructing the representations. Note, the rank of SO(3) and SU (2) is one. Hence they are the
simplest non-abelian Lie groups but nevertheless rich enough to keep us busy until the end of the
course! We note that the matrix T (5.14) plays, the slightly degenerate, rôle of the Lie Algebra of
SO(2).
Theorem 5.3.2 Lie Algebra ↔ unique simply connected Lie group
To each given Lie Algebra corresponds a unique Lie Group which is simply connected. This group is
known as the universal covering group.13
We will not present the proof. The impotent point is that for identical Lie Algrabra’s correspond in
general several Lie groups. Only under the restriction to simply connectedness the exponential map
is 1-to-1 (bijective). The other not simply connected groups can be obtained as quotient groups.
Theorem 5.3.3 Lie Algebra generators of a compact group are hermitian
For compact Lie groups there exist a Lie Algebra basis where the generators are hermitian (Ta† = Ta ).
Proof: (sketch) The part connected to the identity matrix can be written as ρ = exp(iαa Ta ) and
since compact groups admit finite dimensional unitary representations,
In = exp(iαa Ta )† exp(iαa Ta ) = exp(−iαa Ta† ) exp(iαa Ta ) (5.27)
a
= In + iα (Ta − Ta† ) 2
+ O(α ) ⇔ †
T =T , (5.28)
expansion in the parameter α-reveals the condition.
12 For SU (2) the anti-commutator vanishes for all representations and is related to the fact that SU (2) is of rank one.
13 A term which derives from the field of topology.
Corollary 5.3.1 The Hilbert-Schmidt inner product on Mn (C) assumes the form
thm 5.3.3
hTa |Tb i ≡ Tr[Ta† Tb ] = Tr[Ta Tb ] . (5.29)
Theorem 5.3.4 For compact, semi-simple Lie algebras there exists an orthogonal basis,
hTa |Tb i = Tr[Ta Tb ] = 2kR δab . (5.30)
The number kR is the called the Dynkin index and depends on the representation (hence the
subscript R). In this basis (5.30) the structure constants fab c = fabc are completely antisymmetric in
all three indices (c.f. box below).
We shall omit the proof. Below we give a slightly more precise statement which is though beyond the
scope of this course and targeted at the interested reader only. Yet we shall use the result as a matter
of convenience to lower all indices.14

Theorem 5.3.5 Killing metric
One can define a unique symmetric tensor (Killing metric)
Tr[Ta Tb ] = kR kab , (5.31)
invariant under the action of the Lie group
0 = δTc Tr[Ta Tb ] = Tr[[Tc , Ta ]Tb ] + Tr[Tb [Tc , Tb ]] = 2ikR (fca b + fcb a ) . (5.32)
The number kR is the Dynkin index which is dependent on the representation. Note the
(complete) antisymmetry of fab c follows from the invariance of the killing metric (5.32).
For semi-simple Lie groups the killing metric is non-singular and if in addition the group is
compact then there exists a basis where
kab = 2δab . (5.33)
Proof omitted. Note the Killing metric can be used to lower and raise indices. Since for
semi-simple groups the Killing metric is the identity matrix we shall take a rather relaxed
attitude towards upper and lower indices.
Theorem 5.3.6 General properties of structure constants fabc for Lie groups
The structure constant satisfy the Jacobi identity :
faed fbce + fbed fcae + fced fabe = 0 . (5.34)
Proof: (sketch) Any three matrices A, B, C satisfy (exercises):
[A, [B, C]] + [B, [C, A]] + [C, [A, B]] = 0 , (5.35)
14 For semi-simple but non-compact groups the metric hT |T i = k
a b ab is non-degenerate and can be used to lower
indices.
and so does
(Taadj )bc = ifabc , (5.36)
interpreted as a matrix q.e.d.The label adj(oint) is explained in the preceding corollary.
Corollary 5.3.2 The adjoint representation is real
a) Eq. (5.36) defines a representation called the adjoint representation.15
b) The structure constants are real. (fabc ∈ R)
Proof: The show that (5.36) is a representation is left as an (optional) exercise. Reality of fabc can
be seen as follows. If Ta generates a representation of the Lie Algebra so does (−Ta∗ ), the complex
conjugate representation.16 By Taking the complex conjugate of (5.26) one gets:
[−Ta∗ , −Tb∗ ] = ifabc
∗
(−Tc∗ ) . (5.37)
∗
Since the complex conjugate corresponds to the same group this implies fabc = fabc q.e.d.
Exercise 5.3.1 Generalities on compact simple Lie groups

a) Check that Eq. (5.35) is correct for any set of matrices.
b) Check that ρ(α)2 = ρ(2α) and ρ(α)−1 = ρ(−α) using equations in the proof of theorem
5.3.1.
c) Verify Eq. (5.25).

d) (optional) Using the Jacobi identity (5.34) show that (5.36) is indeed a Lie Algebra repre-
sentation of the form (5.26).
5.4 The two non abelian groups: SU (2) and SO(3)

5.4.1 Lie Algebras of SO(3) and SU (2)
After the preparatory steps of the preceding chapter it is not difficult to write down the Lie Algebra
of SO(3) (usually denoted by so(3) lower case letters). From exercise 5.1.1 we know that any three
linearly independent antisymmetric matrices will generate the algebra. A basis for which the structure
constant is completely antisymmetric, which is a possibility according to theorem 5.3.4, is given by:
     
0 0 0 0 0 −1 0 1 0
T1 = −i  0 0 1  , T2 = −i  0 0 0  , T3 = −i  −1 0 0  . (5.38)
0 −1 0 1 0 0 0 0 0
15 The adjoint representation is reminiscent of the regular representation in that the element of the set are declared
to be linearly independent vectors defining a vector space. The adjoint representation acts on the vector space of the
Lie Algebra as follows: (Taadj )Td = [Ta , Td ] = ifade Te .
16 The minus sign originates from the factor i in the parameterisation exp(iαa T ) of Lie group elements.
a
The first two indices of T3 are equal to the SO(2) rotation T in (5.14) and corresponds to a rotation
around the z-axis. The remaining T1,2 are just permutations thereof. The Lie Algebra is readily
computed to be
Lie Algebra: [Ta , Tb ] = iabc Tc , (5.39)
where the structure constant fabc = abc is the completely antisymmetric Levi-Civita tensor (123 = 1).
In accordance with theorem 5.3.6 the structure constant is real and totally antisymmetric as claimed
above. From Tr[Ta Tb ](5.38) = 2δab , (kR )(5.38) = 1 by (5.30).
The Lie Algebra of SU (2) does not follow, at least in the first instance, in such a geometric manner.
We know though that the Lie Algebra generators of SU (n) are hermitian and traceless matrices
(according to exercise 5.1.1). The group SU (2) is of dimension three and the three generators can be
chosen to be
σa
(Ta )SU (2) = , (5.40)
2
proportional to the famous Pauli matrices

0 1 0 −i 1 0
σ1 = σ2 = σ3 = . (5.41)
1 0 i 0 0 −1
This implies that (5.40) satisfies the Lie Algebra (5.39). Yet Tr[Ta Tb ](5.40) = 1/2δab and hence the
Dynkin index (kR )(5.40) = 1/4. What is going on? This means that the Lie groups SU (2) and SO(3)
are locally isomorphic (same Lie Algebra) but are globally different (unequal Dynking index). This
is indeed possible as they are of the same rank; namely of rank one. By virtue of theorem 5.3.2 there
is a unique simply connected Lie group associated with each Lie Algebra. In the case at hand this is
SU (2) as we wish to demonstrate. Consider the mapping from R4 → SU (2):

x0 + ix3 ix1 + x2
φ : (x0 , x1 , x2 , x3 ) = x0 I2 + x1 iσ1 + x2 iσ2 + x3 iσ3 = =U (5.42)
ix1 − x2 x0 − ix3
The two defining conditions of SU (2) lead to:
U U † = I2 , det U = 1 ⇔ x20 + x21 + x22 + x23 = 1 . (5.43)
It follows that SU (2) is topologically equivalent to the three sphere S 3 . The latter is simply connected
and therefore SU (2) is the unique simply connected Lie group corresponding to the Lie Algebra (5.39).
Hence if SO(3) is not identical to SU (2), which it is clearly not, then there must exist a n-to-1
homomorphism from SU (2) to SO(3). The latter is given by the so-called Weyl-homomorphism
h(~x) = x1 σ1 + x2 σ2 + x3 σ3 , (5.44)
which obeys a a
U h(~x)U † = h(O~x) , U = eiα (Ta )SU (2)
, O = eiα (Ta )SO(3)
. (5.45)
Hence the Weyl homomorphism is a 2-to-1 map ρSU (2) → ρSO(3) : U → O. The kernel of this
homomorphism is given by {±I} ' Z2 and by virtue of the first isomorphism theorem 2.2.8 the
following isomorphism holds
SO(3) ' SU (2)/Z2 . (5.46)
We note that SO(3) is not simply connected since SU (2) is the unique simply connected Lie group
associated to (5.39). This has profound implications and is related to the existence of fermions as we
shall see.
Exercise 5.4.1 SO(3) and SU (2) Lie algebra

a) Verify that (5.40), (5.38) satisfy the Lie Algebra (5.39).
b) Verify that the representations in a) satisfy (5.30) with kR = (1/4, 1) respectively.
5.4.2 Irreducible representations of SU (2) (and hence SO(3))

In view of the 2-to-1 Weyl homomorphism it is sufficient to construct the irreducible representations
of SU (2) from which the irreducible representations of SO(3) will follow. We shall apply the very
general method of highest weight representation which can be applied to any of the Lie Algebra’s of the
groups given in table 5.1. The complication raises with the rank. of the irreducible representations.
In fact to each r = rank is associated a so-called Casimir operator which is an operator constructed
from the Lie Algebra element which commutes with the Lie Algebra. Irreducible representations are
labeled by r labels.
The group SU (2) is of rank one and has therefore got one Casimir operator. In this section we
shall denote, because of reasons of familiarity, the generators of the SU (2) Lie Algebra by
Ja ≡ Ta of Eq. (5.39) (5.47)
The quadratic Casimir operator, for any Lie group, is given by T 2 ≡ k ab Ta Tb (with summation over
indices a and b understood); when applied to SU (2) one gets:
J 2 = J12 + J22 + J32 , [J 2 , Ji ] = 0 (5.48)
in the basis (5.31). The Casimir operator is of the form,
J 2 = C2 (R)Idim VR , (5.49)
by virtue of Schur’s lemma! The eigenvalue C2 (R) (which will be identified by j(j +1) latter on) serves
to label the irreducible representation. As a set of maximally commuting operators we choose,17
{J 2 , J3 } . (5.50)
The eigenvalue of J3 , denoted by m, labels the states in a given irreducible representation:
J3 em = mem , m∈R, (5.51)
where em denotes a basis state in a given representation. From the Lie Algebra the following algebraic
relations are immediate:
[J3 , J± ] = ±J± , J± ≡ J1 ± iJ2 ((J± )† = J∓ ) (5.52)
2
J = J32 + J3 + J− J+ , (5.53)
2
J = J32 − J3 + J+ J− . (5.54)
17 In a quantum system that exhibits rotational symmetry, one would add the Hamiltonian to this set of commuting
observables. Here we try to extract the group theoretic aspects only.

One concludes that with J± one may construct,

J3 (J± em ) = (m ± 1)(J± em ) , (5.55)
and therefore em±1 ∝ (J± em ) is an orthogonal vector (since J3 is a hermitian operator). That is
to say provided that (5.55) does not yield zero. The operators J± are usually referred to as ladder
operators for the obvious reason that they allow to go up and down in the representation space as
illustrated in Fig. 5.3.
Figure 5.3: Graphic illustration of highest weight rep-

resentation of SU (2). The J± ladder operator allow to Figure 5.4: Geometric interpretation 1of angular mo-
move up and down in representation space. Eq. (5.56) mentum acting on a vector operator vi where the one
corresponds to the definition of the highest weight state stands for j = 1.
m = j(−m∗ ). Other states are obtained through
Eq. (5.69) by applying J− .
We have got all the tools to construct the highest weight representation. Our goal is to construct
a finite dimensional representation. For the latter there exists a maximal m∗ (top of the ladder) for
which
J+ em∗ = 0 . (5.56)
This state is called the highest weight state. Acting with (5.53) on this state one gets
J 2 em∗ = m∗ (m∗ + 1)em∗ . (5.57)
This relation remains true for all other states em∗ −n ∝ (J− )n em∗ since [J 2 , J− ] = 0 as J 2 is a Casimir
operator. Hence we write
J 2 em∗ −n = m∗ (m∗ + 1)em∗ −n . (5.58)
Since the representation is finite dimensional there must exists a positive integer n∗ for which
J− em∗ −n∗ = 0 . (5.59)
2
Applying J (5.54) on this state and using the fact that
(5.54)
J 2 em∗ −n∗ = (m∗ − n∗ )((m∗ − n∗ ) − 1)em∗ −n∗ ,
(5.58)
J 2 em∗ −n∗ = m∗ (m∗ + 1)em∗ −n∗ , (5.60)
once concludes that

2m∗ = n∗ ∈ Z+ , (5.61)
the irreducible representations are labelled by half-integer.
Going over to more standard notation we choose j ≡ m∗ and denote the state vectors by
em → |j, mi , (5.62)
where we have introduced in addition the letter j which labels the irreducible represenation. In
summary the irreducible representations of SU (2) are labelled by half integer j and they are (2j + 1)
dimensional with m ranging from m = {−j, −(j − 1), .., (j − 1), j}.
In particular we conclude that the SO(3) representation (5.38) corresponds to j = 1 (note (2 · (j =
1) + 1) = 3 is three dimensional) and the SU (2) representation (5.40) corresponds to j = 1/2.
Hence (5.40) is the smallest irreducible representation for SU (2) and (5.38) is the smallest irreducible
representation of SO(3). The matrix elements of the highest weight representation are given by:

(j)
J3 = hj, m|J3 |j, m0 i = mδmm0 , (5.63)
mm0

(j)
J± = hj, m|J± |j, m0 i = N± (j, m)δm,m0 ±1 , (5.64)
mm0
with normalisation factor

1/2 1/2
N± (j, m) ≡ [(j ∓ m)(j ± m + 1)] = [j(j + 1) − m(m ± 1)] . (5.65)
The equations above fully specify the representations of the algebra. The representation ρ themselves
are given by a
α)mm0 = hj, m|eiα Ta |j, m0 i ,
ρj (~ (5.66)
which connects us back to the way we have been handling representations for finite groups. Note j
denotes the representation, α “enumerates” the group elements and m, m0 parameterise the (2j + 1) ×
(2j+1)-dimensional matrix in the so-called highest weight representation. A specific parameterisation,
0
Dj (α, β, γ)mm0 = hj, m|e−iαJz e−iβJy e−iγJz |j, m0 i ≡ e−i(mα+m γ) dj (β)mm0 (5.67)
is known as the Wigner D-matrix (djmm0 (β) is sometimes called the little Wigner D-matrix). It
is widely used in the literature and ties in nicely with calculation of angular decay distribution
amplitudes. Note Jx does not need to be parameterised since it is implicitly generated through
commutators of Jy and Jz .
The quadratic Casimir and the Dynkin index are given by:
(2j + 1)
C2 (j) = j(j + 1) , kj = j(j + 1) , (5.68)
6
where the first equation follows from (5.58) with m∗ = j and the second one is left as an exercise.
Exercise 5.4.2 Irreps of SU (2)

a) Derive Eqs. (5.52,5.53,5.54) using the commutation relations (5.39).

d) Verify the eigenvalue equation for the ladder states in Eq. (5.55).
c) Prove that (5.65) is the correct normalisation factor of
J± |j, mi = N± (j, m)|j, m ± 1i , (5.69)
up to an absolute value. Hint: You may

P want to use: (5.53) or (5.54) in doing so as well as
the completeness relation in I2j+1 = m |j, mihj, m|.
d) Obtain Jx,y,z from Eqs. (5.63,5.64) for j = 1. Check that [Jx , Jy ] = iJz by explicit matrix
multiplication. How do the rotation matrices compare with (5.38)?
e) (optional)
– Take the trace of the Casimir
T 2 = δab T a T b = C2 (R)Idim VR (5.70)
to establish18
2kR dim Vadj = C2 (R) dim VR . (5.71)
– For SU (2) establish (5.68):
(2j + 1)
kj = j(j + 1) (5.72)
6
and hence (k1/2 , k1 , k3/2 , ..) = (1/4, 1, 3/2, ..). Hint: C2 (j) = j(j+1), dim Vj = (2j+1),
dimVadj = dimV1 = 3. Make sure you understand all the equations in the hint.
5.4.3 Clebsch-Gordan series of SU (2)

From your course on quantum mechanics you know that when two particles of angular momentum j1
and j2 are combined then the total angular momentum ranges from |j1 − j2 | to j1 + j2 in integer steps.
The addition of angular momentum corresponds to taking tensor product representations of states19 .
The product on the level of the group corresponds by virtue of the exponential map to addition at
the level of the Lie Algebra.
Theorem 5.4.1 The Clebsch-Gordan series for SU (2) (or Kronecker product) is given by
jM
1 +j2
ρj1 ⊗ ρj2 = ρ|j1 −j2 | ⊕ ρ|j1 −j2 |+1 ⊕ ... ⊕ ρj1 +j2 = ρJ . (5.73)
J=|j1 −j2 |
where two representations ρj are in an irreducible representation with Casimir value j(j + 1).
19 Strictly speaking one assumes, in this simplified line of argument, that the mutual interaction will not change this
picture. In fact it doesn’t. The explanation of which is beyond the scope of this course.
A few explanations are in order. We see that the multiplicity of each representation is one which
implies that SU (2) is a simply reducible group (c.f. definition 6.3.2) as stated earlier on. A non-
trivial check of this formula is given by matching up the dimensions,
(2j1 + 1)(2j2 + 1) = (2|j1 − j2 | + 1) + (2(|j1 − j2 | + 1) + 1) + ...(2(j1 + j2 ) + 1) , (5.74)
which is left as an exercise. A more substantial proof, from which we will use some results later on,
is given below.
Proof: Let us compute the character of a Wigner rotation matrix. First consider a rotation around a
single angle. This is in fact sufficient since any other rotation is equivalent by similarity transformation
and has therefore identical character. Let us choose (α, β, γ) = (θ, 0, 0) for which the rotation matrix
(5.67) is diagonal:
Dj (θ, 0, 0)mm0 = e−imθ δmm0 . (5.75)
From (5.75) one obtains (last step is left as an exercise):
j j 2j
X
j
X
imθ −ijθ
X m sin(j + 1/2)θ
eiθ

χ(ρj (θ)) ≡ χj (θ) = D (θ, 0, 0) mm = e =e = , (5.76)
m=−j m=−j m=0
sin θ/2
where
χ1/2 (θ) = 2 cos(θ/2) , χ1 (θ) = 2 cos θ + 1 , (5.77)
are special cases. Using Eq. (5.76), we can prove the following identity:
χj1 (θ)χj2 (θ) = χj1 +j2 (θ) + χj1 −1/2 (θ)χj2 −1/2 (θ) . (5.78)
Without loss of generality we mat assume that j1 ≥ j2 . Then this equation can be iterated 2j2 − 1
more times to
χj1 χj2 = χj1 +j2 + χj1 +j2 −1 + . . . + χj1 −j2 , (5.79)
and this shows that (5.73) is correct and therefore proves the theorem. q.e.d.
Corollary 5.4.1 Fundamental representation

From the representations j = 1/2 (5.40) [and j = 1 (5.38)], known as fundamental representa-
tions 20 of SU (2) [and SO(3) respectively], all other finite dimensional irreducible representations can
be generated through Kronecker products (direct product representations).
The proof of this corollary is given by Eq. (5.73). The corollary makes it clear that the irreducible
representations (j = 0, 1, 2, ..) of SO(3) are just a sub-series of irreducible representations (j =
0, 1/2, 1, 3/2, 2, ..) of SU (2). The identification follows through the representation dependent Casimir
operator and or the Dynkin index (5.68).
A curious fact is that in particle physics the smallest constituent of matter are all in funda-
mental representations. Surely when particles organise themselves in representations other than the
20 The term fundamental representations originates from the Dynkin-Cartan classification. A fundamental represen-
tation is a representation whose highest weight is a fundamental weight. For some groups there can be more than one
representation which is fundamental e.g. for SO(2n). For SU (2) and SO(3) there is a unique one.
fundamental ones then there is a good chance that they are made up of constituents of smaller rep-
resentations. For example the particles π + , π 0 , π − are in the j = 1 representation of SU (2). Indeed
they are made out of more fundamental particles called quark for which the up and down quarks are in
the fundamental representation. There are also physicists extending this pattern to “real” (i.e. space-
time) spin in saying that the Higgs boson, discovered at the LHC, might not be a fundamental particle
since it is a scalar particle J = 0. It could be made out of two spin 1/2 (fundamental representation)
constituents much alike the pions.
The representations of SO(3) are manifestly real but the SU (2) representations aren’t. As hinted
at earlier on they are pseudo-real (quaternionic).
The essence of pseudo-real is that the complex conjugate representation is not an inequivalent
representation and yet no basis can be found where the representation matrices are real. This in
particular implies that:
ρj1 ⊗ ρ̄j2 ' ρj1 ⊗ ρj2 , (5.80)
and hence we do not need to consider the complex conjugate representation of SU (2). Stated more
precisely:
Theorem 5.4.2 Pseudo-reality of SU (2)
The representation (5.40) is pseudo real. I.e. the complex conjugate representation is equivalent
through conjugation At the level of the Lie Algebra the equivalence relation is given by:
(−Ti∗ ) = STi S −1 , S = σ2 = 2T2 . (5.81)
The proof is left as an exercise. Note the minus sign on the left hand side is due to the convention
exp(iαT ).
Exercise 5.4.3 Irreps of SU (2)

a) Verify the correctness of Eq. (5.74). Hint: without loss of generality assume that j1 ≥ j2 ,
as then you can work without absolute values.
PN i
b) Verify the steps in Eq. (5.76) using the geometric progression formula: i=0 z = (1 −
z N +1 )/(1 − z).
c) Show that Eq. (5.81) holds.
d) Investigate the limit θ → 0 of Eq. (5.76) and give an interpretation of your result. Stated
in other words: could you have predicted the result?
Chapter 6
Applications to quantum mechanics
6.1 A short synopsis of quantum mechanics

Quantum mechanics was developed, in its current form, from 1925 onwards within a short amount of
time. Different versions exists such as the Schrödinger equation (an equation based on energy conser-
vation) and Heisenberg’s abstract matrix mechanics. Dirac showed that (the) different representations
are connected by unitary transformations.
Abstract rules
A quantum mechanical state is described by a state in a Hilbert space (linear space with an inner
product as described in section 1.3) for which we use the ket notation:
Physical state: |Ψi (6.1)
The latter corresponds to some physical situation which is described by a dynamical equation. For
example the Schrödinger equation which is an eigenvalue equation for the Hamiltonian:
H|Ψi = EΨ |Ψi (6.2)
The Hamiltonian (as all observables) are hermitian operators H † = H, and EΨ are therefore a real
number since:
EΨ hΨ|Ψi = hΨ|H|Ψi = hΨ|H † |Ψi = hΨ|H|Ψi∗ = (EΨ )∗ hΨ|Ψi∗ = (EΨ )∗ hΨ|Ψi , (6.3)
which implies EΨ = (EΨ )∗ ∈ R. Furthermore states with different eigenvalues are orthogonal to each
other since:
EΨ hΨ0 |Ψi = hΨ0 |H|Ψi = hΨ0 |H † |Ψi = hΨ|H|Ψ0 i = EΨ0 hΨ|Ψ0 i , (6.4)
which implies hΨ|Ψ0 i = 0 if EΨ 6= EΨ0 (with both of them being real). The basic rule of quantum
mechanics is that everything is determined up to a probabilities only. The probability of Ψ becoming
Ψ0 is given by:
P = |AΨ→Ψ0 |2 , AΨ→Ψ0 ≡ hΨ0 |Ψi . (6.5)
For the probability to be properly normalised one chooses hΨ|Ψi = hΨ0 |Ψ0 i = 1.
74
CHAPTER 6. APPLICATIONS TO QUANTUM MECHANICS 75
From Hamiltonian to the Schrödinger equation in coordinate space

The Hamiltonian, in many cases, can be obtained from classical mechanics up to ambiguities of
ordering . In classical mechanics quantities like position and momentum can be known to an arbitrary
precision whereas in quantum mechanics this os not the case as encapsulated in the commutation
relations of the operators1 , and described by the uncertainty relations. The most famous ones are the
Heisenberg commutation relations for the position operator x̂ and the momentum operator p̂,
[x̂i , p̂j ] = i~δij , [x̂i , x̂j ] = 0 , [p̂i , p̂j ] = 0 , (6.6)
where [A, B] = AB − BA, i, j = 1..3 and ~ ≡ h/(2π) = 1.055 · 10−34 Js. Let us note that from these
relations one can derive the famous Heisenberg uncertainty relations. We shall not do so as this is
not central to the course.
The Schrödinger equation is an equation for a single non-relativistic particle of mass m subdue to
a potential V (x)
p̂2
H = Tkin + V (x̂) = + V (x̂)
2m
The question on how this is turned into a computational scheme is answered by going to a certain
representation which constitutes a set of maximally commuting operators. The canonical choice is
the |xi representation for which:
x̂i |xi = xi |xi (6.7)
and Z
Ix = d3 x|xihx| , hx|x0 i = δ (3) (~x − ~x0 ) . (6.8)
Most importantly the substation for the canonically conjugate momentum operator is as follows:
∂
p̂j |x−space → −i~∂j , ∂j ≡ . (6.9)
∂xj
Now we have all the tools for bridging the gap between the abstract statement of the energy
equation (6.2) and the Schrödinger equation in its conventional form:
Z
d3 x0 hx|H|x0 ihx0 |Ψi = Ehx|Ψi . (6.10)
The latter is equivalent to the time independent Schrödinger equation 2 :

~2

− ∆ + V (x) Ψ(x) = EΨ(x) , Ψ(x) ≡ hx|Ψi , (6.11)
2m
with ∆ = ∂x2 + ∂y2 + ∂z2 upon using Eqs.(6.8) and (6.9).
A typical range of our application involves the transitions matrix elements of the form:
AΨ→Ψ0 = hΨ0 |Htrans |Ψi , (6.12)
where Htrans is some transition Hamiltonian which are prone to a application of the Wigner-Eckart
theorem. In the case where H = H0 +Htrans and Ψ and Ψ0 are eigenstates of H0 , then (6.12) describes
the first order effect and is known as Fermi’s golden rule. It is self-understood that the formula
gives good results in the case where the perturbation can be considered to be small.
1 The latter can be abstracted from the Poisson brackets of Hamiltonian dynamics as emphasised by Dirac in his
transformation theory approach.

2 To get the time dependent Schrödinger equation one substitutes E → −i~∂
t
Symmetries and conservation laws

From course on Hamiltonian dynamics (known as the Noether theorem) you might be familiar with the
concept that symmetries lead to conservations laws and vice versa. The same connection is observed
in quantum mechanics.
Suppose we have a system which is governed by a Hamiltonian H. The transition matrix elements
(6.12) involve the Hamiltonian and if the system is to be invariant under a symmetry |Ψi → U |Ψi
then this implies,
hΨ0 |H|Ψi = hU Ψ0 |H|Ψi = hΨ0 |U † HU |U Ψi ⇔ [U, H] = 0 , (6.13)
that the unitary symmetry transformation3 commutes with the Hamiltonian H. The unitary trans-
formation written in terms of a generator G as U = exp(iαG) then implies that the generator has to
commute with the Hamiltonian, [H, G] = 0.
Furthermore the Heisenberg equation of motion of an operator G, independent of time (∂t G = 0),
reads:
d
hGi = ih[H, G]i , (6.14)
dt
where h..i denotes the expectation value on any state. Hence the quantity hGi is a conserved quantity
if it commutes with the Hamiltonian. Therefore:
symmetry generated by G ⇔ conservation law for G (6.15)
A list of symmetries and conservations laws are given below:

• Space-time symmetry: The Poincaré group4 SO(3, 1) n R4 which consists of the Lorentz group
SO(3, 1) and space-time translations. We shall discuss them one by one:
– In a rest-frame this reduces to: Rotation group SO(3) (or SU (2)) and time and space
translation
∗ SO(3) rotational symmetry ↔ conservation of angular momentum
∗ space translations ↔ conservation of momentum
∗ time translation ↔ conservation of energy
– When one considers different inertial frames then there are in addition rotations, so to
say, between the three space and the time components. Thos three additional symmetries
correspond to boosts between in the three space directions. Note dim SO(3) = 3 and
dim SO(3, 1) = 6 and this explains the additional three symmetries.
• Extension of space-time symmetries are C, P and T :
– parity symmetry P : (x, y, z, t) → (−x, −y, −z, t) conserved up to the weak interactions.
– time symmetry T : (x, y, z, t) → (x, y, z, −t) small violations in weak interactions (presum-
ably responsible for matter anti-matter asymmetry in the Universe)
3 Besides unitarity symmetry transformation there is second possibility according to a famous result by Wigner:
anti-unitary transformations for which hAΨ|AΨ0 i = hΨ0 |Ψi. We shall not discuss them in this course but mention that
the time inversion operator is an anti-unitary transformation.
4 The Poincaré group is a group of rank two. Its Casimir operators correspond to the spin and the mass of a particle
which are therefore, besides the interaction, the primary characterisation of a particle.
– charge symmetry C: exchange of particle and anti-particles conserved up to weak interac-

tions
– the combined symmetry of CP T is within the foundation of quantum field theory. Viola-
tions thereof would lead to a revolution in the field. In this sense violation of T -symmetry
is the same as violation of the combined CP -symmetry.
• Gauge symmetries (local symmetries) of Lie groups lead to conservation of charges. The Stan-
dard Model exhibits, at high energy, a SU (3)c × SU (2)L × U (1)Y symmetry. The c stands for
the colour charge, L for left handed and Y for something called hypercharge. At low energies
the group reduces, through the Higgs mechanism, the SU (3)c × U (1)em . The U (1)em formally
describes the conservation of electric charge (and em stand for the electromagnetic).
• Global flavour symmetries such as described in exercise 6.7.1 a) which help to classify bound
states of the strong interactions.
Exercise 6.1.1 Basic quantum mechanics

a) Verify the Heisenberg commutation rules in x-space by using the substitution rule (6.9) for
the momentum.
b) Complete the steps from (6.10) to (6.11).
6.2 SU (2)-group theory and the physics of angular momentum

The goal of this section is to connect results on SU (2)-group theory with the physics of angular
momentum. The bottom line is as follows: According to the principle of the previous section (6.15),
rotational symmetry implies conservation of angular momentum and vice versa. More precisely the
Casimir number C2 (j) = j(j + 1) (or equivalently j) as well as the projection on one component,
say m (magnetic quantum number Jz ), are conserved. The former being independent of the frame
whereas the latter is frame dependent (since Jz is not a Casimir operator).
Before we outline a few important points below let us be clear about the notation used throughout.
~ = ~x × p~ (or La in components) denotes the orbital angular momentum.
• L
~ denotes the intrinsic angular momentum also known as spin.5
• S
• J~ denotes the total angular momentum, i.e. the sum of all angular moment involved in the
~ or no S
physical system. Note, if there is either no L ~ involved then J~ can also stand for S
~ or L
~
alone.
5 The spin is a relativistic concept and entered quantum mechanics, at the formal level, through Dirac’s relativistic
equation. One can nevertheless incorporate it into the non-relativistic Schrödinger equation in a heuristic manner.
i) Can we make the symmetry conservation principle (6.15) more concrete?

One may simply observe that the quantum mechanical angular momentum,6
La = (~x × p~)a = abc x̂b p̂c |x−space → −i~abc xb ∂c , (6.16)
satisfies the SU (2) Lie Algebra (5.39) with an additional, but trivial, extra factor ~ on the right
hand side.
ii) What is the meaning of angular momentum j?
A system with angular momentum j has to be rotated by 360◦ /j in order to be identical to
its initial configuration. This follows from (5.75) Dj (θ, 0, 0)mm0 = e−imθ δmm0 and remembering
that |m|max = j, and hence Dj (2π/j, 0, 0) = I2j+1 . We observe that:

1 j integer bosons
Dj (2π, 0, 0)mm0 = (−1)2m = (−1)2j = . (6.17)
−1 j half-integer fermions
Hence if a fermion is rotated by 360◦ -degrees then it picks up a minus sign. This seems odd at
first but in fact this sign does disappear when one considers matrix elements since the fermion
number is a superselection rule (c.f. also section 6.6) by virtue the nature SU (2) Clebsch-Gordan
series. The crucial point is that in order to obtain a scalar (invariant) one needs an even number
of fermionic objects. Any Kronecker product of an odd number of half-integer representations
does not give rise j = 0 object. To be discussed in section 6.6. The crucial differences between
bosons and fermions have been briefly described by in section 4.5.2. It seems remarkable that
the fact that SO(3) (the rotation group of our 3D-space) is not simply connected opens the gate
to half-integer spin and fermions (through the universal covering group SU (2)).
iii) Total angular momentum: addition of angular momentum
Suppose there is a quantum mechanical system with many subsystems then a change of coordi-
nate frame e.g. rotation must affect all its constituents. Hence the generator G in (and around)
(6.13) corresponds to the total angular momentum.7 This applies at the level of two particles
which form a bound state J~tot = J~1 + J~2 or for a single particle with orbital angular momentum
~ and spin S
L ~ to J~tot = L ~ Furthermore: unless otherwise stated, two angular momenta J (a)
~ + S.
i
(b) (a) (b)
and Ji are compatible in the sense that [Ji , Jj ] = 0 for (a 6= b).
An example: a total wave function could be made out of spatial wave function Ψl which is in a
l = 1 state and a spin s = 1/2 state. Schematically and according theorem 5.4.1,
|Ψl=1 i ⊗ |s = 1/2i = |j = 1/2i ⊕ |j = 3/2i . (6.18)
Crucially there is no j = 1 by virtue of the superselection rule for bosons and fermions.
Exercise 6.2.1 The physics of angular momentum.
6 Note, there’s no problem of normal ordering by going to the quantum system.

7 It’s the same pattern with energy and momentum. The total sum of energy and momentum, in a specific process,
is conserved and not the individual momentum of say one particle. Otherwise the world would not be as interesting as
it is!
a) Verify that the representation of the angular momentum (6.16) satisfies the SU(2) Lie
Algebra (5.39) . Do this by considering just [L1 , L2 ] = i123 L3 commutator. The general
derivation, handling the Levi-Civita tensor, is trickier than it looks and I do not recommend
to do it unless you know the right kind of identities.
b) Consider j = 1 representation. The state space is spanned by |1, −1i, |1, 0i and |1, 1i. These
states will mix into each other under a generic rotation. First state why this has to happen
on physical grounds. Second provide a more formal argument why it happens. Should you
find it difficult to disentangle the two points then that is fine as it could mean that you
have already internalised the notation!
Figure 6.1: Taken from the Particle Data Group Book (booklets available for free) from http://pdg.lbl.gov/.
In the Condon Shortly phase convention the coefficients are real so that one does not need to worry about
complex conjugation.
6.3 Selection rules I: Wigner-Eckart theorem for SU (2) (and

SO(3))
In order to state the celebrated Wigner Eckart theorem we have to introduce the definition of tensor
operator and Clebsch-Gordan coefficient. After stating the theorem for SU (2) we shall give some
general remarks on the general version which mainly has to do with the case when the group is not
simply reducible.
Definition 6.3.1 Tensor operator: A tensor operator 8 is an operator that transforms under
an irreducible representation of a group G. More specifically let ρ(g) be a representation on the
c
vector space under consideration then Tm c
is a tensor operator in the irreducible representation c if
it transforms as follow,
ρ(g)Tmc
c
ρ(g)† = (ρc (g))mc m0c Tm
c
0 ,
c
(6.19)
with summation over the index m0c implied.
An example of a tensor operator is given by the momentum p~ which transforms under the vector
representation of the SO(3). The type of transformation law in Eq. (6.19) is also known as the
operator or quantum mechanical implementation of a symmetry transformation.
Definition 6.3.2 Clebsch-Gordan coefficient Since the direct product of irreducible representa-
tions decomposes into a direct sum of irreducible representations (4.52), the following relations must
exist, X
|a, ma ⊗ b, mb i = hc(i), mc |a, ma ⊗ b, mb i |c(i), mc i , (6.20)
| {z }
c,mc ,i c(i)ab
Cmc ma mb
with i = 1..mabc denoting the multiplicity. Groups for which mcab = 0, 1 are called simply reducible
and for the latter we may write:
X
simply reducible: |a, ma ⊗ b, mb i = hc, mc |a, ma ⊗ b, mb i |c, mc i . (6.21)
c,m
| {z }
c cab
Cm c ma m b
The groups SO(3) and SU (2) are simply reducible. An efficient way to picture Eqs. (6.20,6.21) is to
note that: X
Ia⊗b = |c(i), mc ihc(i), mc | . (6.22)
c,mc ,i
c(i)ab
The matrix element Cmc ma mb is known as a Clebsch-Gordan coefficient. For fixed a, b, c, i, ma , mb , mc
it is indeed a coefficient. In this course we shall be concerned with simple reducible cases only.
Theorem 6.3.1 Wigner-Eckart theorem (1930-1931) (for simply reducible groups)
The Wigner-Eckart theorem states that the matrix element of two states and a tensor operator is
proportional to the Clebsch-Gordan coefficient,
a cab
hc, mc |Tm a
|b, mb i = Cm c ma mb
hc||T a ||bi (6.23)
times a reduced matrix element hc||T a ||bi which is independent of the directions mabc .
8 The word tensor, presumably, originates from the Cauchy stress tensor which is a tensor with two vector indices
under SO(3).
a
Proof: (sketch) The state Tm a
|b, mb i transforms like the direct product representation. This can be
seen as follows:
(6.19)
a
ρ(g)Tm a
a
|b, mb i = ρ(g)Tm a
ρ(g)† ρ(g)|b, mb i = ρa (g)ma m0a ρb (g)mb m0b Tm
a 0
0 |b, mb i .
a
(6.24)
Therefore we may use (6.21) to deduce:

X 0
a c ab 0 0
Tm a
|b, mb i ∝ Cm 0 m m |c , mc i .
c a b
(6.25)
c0 ,m0c
Eq. (6.23) follows by applying a bra state hc, mc | and using hj, m|j 0 , m0 i = δjj 0 δmm0 q.e.d.
A few remarks:
• The moral of the Wigner-Eckart theorem is that for a given representation one only needs
to compute one matrix element. The rest of them follows from group theory only; namely
the Clebsch-Gordan coefficients. The Clebsch-Gordan coefficients are standard devices which
are tabulated in books. One may not even want or be able to compute the reduced matrix
element but parameterise it by a number and consider ratios of matrix elements only. The latter
gives relative decay times in the case the particles are part of a representation of an internal
symmetry group. This is the case for the so-called SU (2) isospin symmetry (exchange of up and
down quarks) for the pions. Be warned that the phases of the Clebsch-Gordan coefficients are
dependent on the phase convention of |a, ma i. The Wigner-Eckart theorem is a device that gives
immediate access to so-called selection rules. By selection rules physicists mean that certain
class of matrix elements are zero which could be related to certain atomic or decay transitions
being forbidden (by symmetry).
• We consider it worthwhile to briefly restate the essence in a colloquial style. The Wigner-
Eckart theorem and Clebsch-Gordan coefficient construction states: i) that a matrix element
a
hc, mc |Tm a
|b, mb i 6= 0 only if the irreducible representation c appears in the Clebsch-Gordan
series of a ⊗ b (= mcab c ⊕ ... with mcab 6= 0) ii) the direction of the matrix element is completely
determined through Clebsch-Gordan coefficient. I have tried to summarise the essence of all
this in the diagram in Fig. 6.2.
Figure 6.2: The Wigner-Eckart theorem illustrated in a figure.

The term selection rule refers to matrix elements which are zero on grounds of vanishing
Clebsch-Gordan coefficient. The term originates from looking at atomic transitions and the
observation that the symmetry selects only certain transitions. Disregarding specific selection
rules on the coefficients ma , mb and mc , a first selection rule is provided by the necessity of
mcab 6= 0: i.e. the Clebsch-Gordan series of ρa ⊗ ρb must contain ρc at least once as otherwise
c(t)ab
the Cmc ma mb = 0. We will discuss applications of the Wigner-Eckart theorem towards the end
of the course in the context of SO(3).
• It ought to be clear that none of the arguments above rely on the SU (2)-representation theory.
Thus the Wigner-Eckart theorem applies to any group with finite dimensional representation
straightforwardly. In the case where the group is not simply reducible the general Clebsch-
Gordan coefficients have to be employed and the formulae complicate but no new ideas are
needed.
We consider it worthwhile to restate it here in the notation used for SU (2) because of its numerous
applications. Furthermore we will restate the tensor operator condition (for l = 1) at the level of the
Lie Algebra.
The Wigner-Eckart theorem for SU(2)-notation reads:
j1 Jj1 j2 j1
hJ, M |Tm 1
|j2 , m2 i = CM m1 m2 hJ||T ||j2 i (6.26)
Jj1 j2
with Clebsch-Gordan coefficient CM m1 m2 ≡ hJ, M |j1 , m1 ⊗ j2 , m2 i explicitly given in Fig. 6.1 (bor-
rowed from the Particle Data Group book) in the so-called Condon-Shortly convention which is
rather standard throughout the literature. Equation (6.26), unmasks the selection rules imposed by
2
the Wigner-Eckart theorem on SU (2), as conservation of the Jtot and (Jtot )z quantum numbers
selection rules : j1 + j2 ≥ J ≥ |j1 − j2 | , M = m1 + m2 . (6.27)

j
Theorem 6.3.2 The definition 6.3.1 of a tensor operator ρ(g)Tm j
ρ(g)† = (ρj (g))mm0 Tm 0 (6.19) is
written in infinitesimal form as:

[Ja , Tb1 ] = iabc Tc1 , (6.28)
with summation over repeated indices understood. Note in particular this shows that Ja itself is a
tensor operator j = 1, i.e. a vector, since Tc1 = Jc is obeys (6.28) by virtue of the SU (2) Lie Algebra
equation. Angular momentum is a vector! The proof is left as an exercise.
Clebsch-Gordan coefficients in direct products In practice a direct product of two SU (2)

representations can be expressed as follows:
X Jj1 j2 ∗
|J, M i = (CM m1 m2 ) |j1 , m1 ⊗ j2 , m2 i (6.29)
j1 ,m1 ,j2 ,m2
using the fact that X

Ij1 ⊗j2 = |j1 , m1 ⊗ j2 , m2 ihj1 , m1 ⊗ j2 , m2 | . (6.30)
j1 ,m1 ,j2 ,m2
As an example of j = 1/2, 3/2 state on the right hand side of (6.18) let us quote:
r r
1 2
|3/2, 1/2i = |1, 1i ⊗ |1/2, −1/2i + |1, 0i ⊗ |1/2, 1/2i ,
3 3
r r
2 1
|1/2, 1/2i = |1, 1i ⊗ |1/2, −1/2i − |1, 0i ⊗ |1/2, 1/2i , (6.31)
3 3
where the Clebsch-Gordan coefficients are taken from Fig. 6.1. The verification of Eq. (6.31) is left
as an exercise.
Geometric interpretation of infinitesimal version of the j = 1 representation Expanding

the left hand side of (6.19) (left as an exercise) to first order one gets
eiαa Ja Tk1 e−iαb Jb = Tk1 + alk αa Tl1 + O(α2 ) . (6.32)
The geometric interpretation of the O(α)-term in (6.32) is:
α × T~ )k = sin(γ)|~
alk αa Tl1 = (~ α||T~ |~ek . (6.33)
Above ~ek is a unit vector orthogonal to the plane spanned by α~ and T~ . It is nothing but the cross
product (familiar from electrodynamics for example) which is the infinitesimal version of a rotation.
The reader is encouraged to consider Fig. 5.4 to convince him or herself.
Exercise 6.3.1 Wigner-Eckart theorem (The proof of theorem 6.3.2)
a) Using the commutator (6.28) derive (6.32).

b) Let us consider α3 = α and α1,2 = 0 in which case the rotation matrix j = 1 assumes the
following form:
   
cos(α) sin(α) 0 0 α 0
ρj=1 (α) − sin(α) cos(α) 0 = I3 + −α 0 0 + O(α2 ) (6.34)
0 0 1 0 0 0
Show that (6.32) and (6.70) are consistent with the definition of a tensor operator
j
j
ρ(α)Tm ρ(α)† = (ρj (α))mm0 Tm 0 .
c) Verify Eq. (6.31) using the table in Fig. 6.1.

d) Check in addition that the two states Eq. (6.31) are orthogonal to each other.
e) Verify (6.29). All the necessary equations are given in the corresponding section above.
6.4 Wigner-Eckart applied: the Landé g-factor

The H-atom with with Coulomb potential exhibits spherical symmetry. This symmetry is lifted by
the application of a homogeneous magnetic field. The relevant part of the perturbing Hamiltonian
reads9
eB
H 0 = x(L3 + 2S3 ) = x(J3 + S3 ) , x ≡ − , (6.35)
2me c
with B the magnetic field in 3-direction, me , e the electron mass and charge and c the speed of
light. The operators J3 and S3 are the total angular momentum and (intrinsic) spin operator in the
3-direction. We assume to know the states of the hydrogen atom including spin-orbit coupling in
terms of |n, l, j, mi quantum numbers. Using first order perturbation theory10 the in H 0 leads to a
correction energy of the following form:
E 0 = hn, l, j, m|H 0 |n, l, j, mi = x(~m + hn, l, j, m|S3 |n, l, j, mi) . (6.36)
Note that m is the magnetic quantum number of the total angular momentum J. ~ Hence the task is
to compute the matrix element on the right hand side with the complication that the state |n, l, j, mi
is not obviously related to the spin operator S3 . The solution of this problem goes with a classic
application of the Wigner-Eckart theorem.
In the derivation below the quantum numbers nlj are kept fixed and we shall use the condensed
notation |n, l, j, mi → |mi. By virtue of the Wigner-Eckart theorem we may write,
hm|Si |m0 i = ahm|Ji |m0 i (6.37)
~ and J~ are vector operators. The symbol a is the ratio of reduced matrix elements which
since both S
we do not need to specify any further.11 Note, once a is known, E 0 is known and the problem is
solved.
The rest of the derivation involves the combination of three equations. First the product J~ · S
~ can
be written in terms of three Casimir operators
−2J~ · S
~ = (J~ − S)
~ 2 − J~2 − S
~2 = L
~ 2 − J~2 − S
~2 . (6.38)
Second one may write,

X
hm|J~2 |mi = hm|Ji |m0 ihm0 |Ji |mi ,
m0
X
hm|J~ · S|mi
~ = hm|Ji |m0 ihm0 |Si |mi (6.39)
m0
9 This derives from the magnetic moment interaction with the magnetic field H = −~ µ·B ~ with µ e~ ~
~ |orbital = 2m L
ec
and µ ~ The latter differs by a factor of 2 which can be derived from the Dirac equation and was a bit of a
~ |spin = me~c S.
e
surprise. The spin is a relativistic effect and has to be introduced into a non-relativistic framework like Schrödinger’s
equation in an ad-hoc way.
10 If higher orders were to be considered then we would need to take into account the backreaction of H 0 on the states
|n, l, j, mi amongst other effects.

11 More precisely a = h||S ||i/h||J ||i in our condensed notation. The ratio of Clebsch-Gordan coefficients is one and
i i
therefore disappears.
with summation over i implied. Now,

(6.37)
hm|J~ · S|mi
~ = ahm|J~2 |mi = a~2 j(j + 1) ,
(6.38) ~2
hm|J~ · S|mi
~ = − (l(l + 1) − j(j + 1) − s(s + 1)) , (6.40)
2
where S 2 = s(s + 1) etc. Now, we can solve for a in (6.37) and obtain E 0 as follows,
0 −e~mB 1 1
Ejlsm = g, g =1+ (j(j + 1) + s(s + 1) − l(l + 1)) , (6.41)
2me c 2 j(j + 1)
where g is called the Landé g-factor. The energy E 0 is a shift in energy that should be added to the
main part and it resolves the degeneracy in the spectrum. Such tests were of utmost importance in
developing an understanding of quantum mechanics.
Exercise 6.4.1 Applications of the Wigner-Eckart theorem

a) Wigner-Eckart theorem and nucleon-nucleon scattering.
This exercise is basically solved for as you read along. You should try to understand the
intermediate steps and provide the final step for completion.
Let us consider the following three processes:
(1) p + p → deut + π + ,
(2) p + n → deut + π 0 ,
(3) n + n → deut + π − , (6.42)
where deut, p and n are deuteron (pn), proton and neutron respectively. The π ±,0 are
pions of different charges as indicated. The processes in (6.42) are governed by the strong
force. The latter does not distinguish between so-called up (u) and down (d) quarks. This
can be formalised into a so-called SU (2)-isospin symmetry which acts on the vector (u, d).
Instead of |j, mi one uses the symbols |I, Iz i where I stands for Isospin. (Isospin should not
be confused with spin. The spin in isospin is borrowed because of the analogy that arises
through SU (2) in the mathematical sense.) The up and down quarks are associated with
|I, Iz i-states as follows:
u ∼ |1/2, 1/2i , d ∼ |1/2, −1/2i
– Convince yourself that charge conservation is obeyed in Eq. (6.42). This is (very) easy
if you know how it works.
– The protons and neutrons are made out of uud and udd quarks respectively. They are
associated with the following isospin states:
p ∼ |1/2, 1/2i , n ∼ |1/2, −1/2i . (6.43)
Out of three isospin I = 1/2 states one can either form a I = 1/2 or I = 3/2 state. (The
I = 3/2 states corresponds to the four ∆++ , ∆+ , ∆0 and ∆− which are degenerate
in mass (up to small electromagnetic corrections) and weigh about 1.2 the proton and
neutron mass.) Can you guess why the proton and neutron do not correspond to
I = 3/2 states?
– Out of protons and neutrons (6.43) four states can be formed,

√
|1, 1i ∼ |p ⊗ pi , |1, 0i ∼ 1/ 2(|p ⊗ ni + |n ⊗ pi) ,
√
|1, −1i ∼ |n ⊗ ni , |0, 0i ∼ 1/ 2(|p ⊗ ni − |n ⊗ pi) , (6.44)
one of isospin I = 0 (singlet) and three of isospin I = 1 (triplet). Show that the
relations are correct by using (6.29) and table in Fig. 6.1 .
– As a matter of fact we know that the deuteron corresponds to singlet state. In addition
the pions are made out of up and down quarks and corresponds to the following isospin
associations:
|1, 1i ∼ π + , |1, 0i ∼ π0 ,
i
|1, −1i ∼ π . (6.45)
√
From Eq.(6.44) we obtain |p ⊗ ni ∼ 1/ 2(|0, 0i + |1, 0i). Show that the relative size
of the amplitudes, A1 = hdeut π + |p ⊗ pi (and analogous for A2,3 ), is given by:
√
A1 : A2 : A3 = 1 : 1/ 2 : 1 . (6.46)
This implies that the cross section, which are proportional to the absolute value of
the amplitudes squared, obey the ratios: σ1 : σ2 : σ3 = 2 : 1 : 2 as they indeed do!
Historically such processes go the other way around. A pattern in rates and cross
section is observed and then possible symmetries of the underlying constituents are
identified.
Hint: You can ignore the deuteron as it is a isospin singlet. Hence the right hand side
is given in Eq. (6.45). In fact we have also obtained the left hand side completely (in
terms of isospin states).
0
b) Which quantum number in Ejlsm (6.41) indicates that spherical symmetry is broken?
c) (optional on popular demand) The usefulness of branching rules! (good revision exercise)
The energy levels of a free atom, which enjoys spherical symmetry, arrange themselves
in multiplets of the rotational symmetry. The s–wave (l = 0) corresponds to a 1–fold
degenerate state, the p–wave (l = 1) to a 3–fold degenerate level, and finally the d–wave
(l = 2) yields a 5–fold degenerate level.
Consider an atom in a crystal which has A4 (tetrahedral) symmetry. The symmetry
becomes more constraining and the mutiplets√branch into smaller representations. By
using the A4 (|A4 | = 12) character table (ω = ( 3 − 1)/2):
A4 class () (12)(34) (123) (132)

1 1 1 1 1
10 1 1 ω −1 − ω
100 1 1 −1 − ω ω
3 3 −1 0 0
Show that:
1. the 1–fold degeneracy of the s–wave is not lifted;

2. the 3–fold degeneracy of the p–wave is not lifted;
3. the 5–fold degeneracy of the p–wave is lifted , and we obtain three different eigenvalues,
which are 1–fold, 1–fold, and 3–fold degenerate respectively.
Hint: You need to construct the SO(3) character table restricted to the classes for A4 for
l = 0, 1, 2. In order to do this you need to identify the order of the class form which you
will get the rotation angle and then you can use the general formula χl (θ) obtained in the
lecture. In a second step you then form the scalar products from which you can read off the
branching rules: ρl |SO(3) → ⊕i mi ρi |A4 with i summing over the irreducible representations
of A4 and mi being the multiplicities.
6.5 Spherical harmonics

The Heisenberg commutation relations (6.6) indicate that the x-space representation is infinite di-
mensional since for any finite dimensional representation Tr[A, B] = 0 by the cyclicity of the trace.
How come the reprentations |l, mi are all finite dimensional. The resolution lies in the fact that the
Hilbert space on S 2 , HS2 ' ⊕l≥0 Hl , decomposes into an infinite sum of the finite dimensional Hilbert
spaces Hl,m (dim Hl = (2l + 1)). The group theoretic study of the spherical harmonics, which is goal
of this chapter, should clarify this statement.
Let us consider S 2 parameterised by the usual variables φ and θ as shown in Fig. 6.3. A state
|θ, φi obeys the following completeness and orthogonality relation, analogous to (6.8) for |xi,
Z
IS 2 = dΩ|θ, φihθ, φ| ↔ hθ, φ|θ0 , φ0 i = δ(φ − φ0 )δ(cos θ − cos θ0 ) ,
X X
= |l, mihl, m| hl, m||l, mi = δll0 δmm0 (6.47)
l≥0 m=−l
where Z Z 2π Z 1
dΩ = dφ dcosθ , (6.48)
0 −1
is the usual integration over the solid angle. The spherical harmonics Ylm are the projections of
the state |l, mi on |θ, φi:
Ylm (θ, φ) ≡ hθ, φ|l, mi . (6.49)
From (6.47) the following orthogonality and completeness relation for the Ylm are immediate:
Z
dΩ Yl∗0 m0 (θ, φ)Ylm (θ, φ) = δll0 δmm0 , (6.50)
l
X X
∗
Ylm (θ, φ)Ylm (θ0 , φ0 ) = δ(φ − φ0 )δ(cos θ − cos θ0 ) . (6.51)
l≥0 m=−l
The Casimir operator L2 on the two sphere S 2 is given by,
Figure 6.3: Coordinates on S 2 the unit sphere in three dimensions.
Z
1 1
∆S 2 = dΩhθ0 , φ0 |(−L2 /~2 )|θ, φi = ∂2 + ∂θ (sin θ∂θ ) , (6.52)
sin θ2 φ sin θ
and corresponds to the Laplacian on S 2 . Hence the Ylm satisfy,
∆S 2 Ylm = −L2 /~2 Ylm = −l(l + 1)Ylm , (6.53)
the harmonic equation and is why they are called spherical harmonics.12 After all a Laplacian is
invariant (under rotation) and of second order and since there is only one Casimir the connection is
a necessity rather than an accident!
Application to R3 ' S 2 ⊗ R+
The three dimensional Euclidian space R3 is decomposed by polar coordinates as follows:
R3 ' S 2 ⊗ R+ , (6.54)
where R+ parameterises the radial coordinate r. Since the Ylm are a complete set of functions on S 2
(6.51), any function f (x, y, z) on (6.54) may be written as:
l
XX X
f (x, y, z) = unlm (r)Ylm (θ, φ) , (6.55)
n l≥0 m=−l
where unlm (r) ∈ C is any complete set of functions on R+ . The Laplacian on R3 in polar coordinates
is given by:
1
∆R3 = 2 (∂r r2 ∂r + ∆S 2 ) . (6.56)
r
Hence we can write the Schrödinger equation (6.11) in terms of polar coordinates. We note that the so-
called solid harmonics Ylm = rl Ylm satisfy Laplace equation in spherical coordinates ∆R3 Ylm = 0.
12 The harmonics on euclidian space are sin and cos, as you know.
This becomes particularly useful in when the potential exhibits spherical symmetry V (x, y, z) =
V (r). In this case the solution does not depend on the magnetic quantum number m as the latter
singles out a direction in space. Hence unlm (r) → unl (r) in Eq. (6.55) and the Schrödinger equation
becomes an equation of one variable:
~2 1

2 (l)
− ∂r r ∂r + V eff (r) unl (r) = Enl unl (r) , (6.57)
2m r2
with
(l) ~2 l(l + 1)
Veff (r) = V (r) + , (6.58)
2m r2
and unl (r) being the eigenfunction of the operator on the left hand side rather than an arbitrary set
of complete function on R+ as previously described.
In this small mathematical aside we want to clarify why it is S 2 that appears as the natural
manifold of the SO(3) Lie group. Consider an arbitrary vector in R3 which we may well
choose to be v = (1, 0, 0). This vector has SO(2) acting on the coordinates y and z as a
stabiliser group. This signals redundancy of the problem and has to be divided out. The
relevant manifold is then indeed,
S 2 = SO(3)/SO(2) . (6.59)
Note, this is not a group manifold since SO(2) is not normal in SO(3)! Some of you might
already have been convinced by the parameterisation in Fig. 6.3 and it is indeed an equivalent
way to look at it for SO(3). One should be aware though that for groups which do not permit
such a simple graphical illustration one has to go the algebraic way as outlined in this small
note.
Exercise 6.5.1 Spherical harmonics

a) In this exercise we intend to clarify to what extent the highest weight representation tech-
nique applies to the spherical harmonics or solid harmonics respectively. In fact the most
efficient way to proceed is through cartesian and not spherical coordinates. The solid har-
monics for l = 1 are given by:
r r
∗ 1 3 1 3
Y11 = Y1−1 = − (x + iy) , Y10 = z. (6.60)
2 2π 2 π
First obtain L± = (Lx ± iLy ) from
Lx = −i(y∂z − z∂y ) , Ly = −i(z∂x − x∂z ) ,
then check that i) L+ Y11 = 0 (i.e. (5.56)) and ii) obtain Y10 ∝ L− Y11 (expression just
below (5.57)). Note the factor of proportionality is given by the normalisation factor N− (l =
1, m = 1).
b) Verify that the solid harmonics (see text) satisfy the Laplace equation ∆R3 Ylm = 0 with
∆R3 as in Eq. (6.56).
c) Verify the orthogonality and completeness relation (6.50) and (6.51) by using the definitions
given in the corresponding section.
6.6 Selection rules II: parity and superselection

Parity selection rule
Parity symmetry P , which was briefly mentioned in section 6.1, is the reflection of space (mirror):
P ◦ (x, y, z, t) → (−x, −y, −z, t) . (6.61)
It was believed to be a strictly conserved quantum number of nature until 1956 when theorists (T.D.Lee
and Yang) reexamined the grounds for this believe; coming to the conclusions that parity might be
violated (in the weak interactions). Indeed it was found that parity is violated, even maximally in
the weak interactions. This led Pauli’s to one of his most famous quotes (alluding to the fact that
weak interactions couple to left handed fermions in a definite way): “I can’t believe that god is a
weak left-hander”. Yet the weak interactions are weak and can be neglected when the transition is
induced by the strong interactions. Therefore parity, in practice, is a good quantum number as long
as the weak force does not play a major rôle in the decay.
By including parity we effectively extend the discussion from SO(3) to O(3) since every O(3)-
element can be written as a reflection times an SO(3)-element. The implementation of the parity
symmetry is unitary. Furthermore U (P )2 = eiη I13 is identical to the identity times a phase which is
conveniently chosen to be zero. Hence U (P ) = ±I.
The parity transformation of the state are given by,
U (P )|1/2, mi = |1/2, mi , U (P )|l, mi = (−)l |l, mi , l ∈ Z+ , (6.62)
where the later can be inferred from the transformation properties of the Ylm :
P ◦ Ylm (θ, φ) = Ylm (π − θ, φ + π) = (−1)l Ylm (θ, φ) . (6.63)
The transformation of the angles can be inferred from Fig. 6.3. Let us suppose that a transition is
governed by an operator X with parity U (P )XU (P )† = (−1)ηX X. Then the following selection rule
applies
hlm|X|l0 m0 i =
6 0 ⇒ l + l0 + ηX even . (6.64)
This is a selection rule on top of the ones quoted in (6.27) (the latter applies provided X is a tensor
operator).
Examples of operators with definite transformation properties are the coordinate ~x, momenta p~,
~ and magnetic B
electric E ~ field as well as angular momenta L:~
η~x = ηp~ = ηE~ = 1 , ηL

~ = ηB
~ =0. (6.65)
13 The very careful reader might notice that the phase can differ on each superselection sector.
The operators L ~ and B~ are even vectors under parity and are also known as pseudo vectors. For
the angular momentum this comes from P ◦ L ~ = P ◦ (~x × p~) = P ◦ ~x × P ◦ p~ = −~x × (−~ ~
p) = L.
At last we would like to mention that there is a subtlety that we have not discussed. There is
something called intrinsic parity which cannot be seen from l and s alone. As much as the B ~ and
~ are pseudo vectors particles can also have this pseudo-quality. For example the ρ-meson and the
L
A1 -meson are both vectors S = J = 1 but the former has minus parity and the other has plus parity.
Hence A1 is a pseudo vector particle in some sense. When you compute transitions between alike
particles then the intrinsic parity cancels. This is often the case in quantum mechanics which is the
scope of this course.
Superselection rules
Superselection rules were introduced by Wick, Wightman and Wigner in 1952 as a generalisation of
selection rules. Selection rules refer to a a certain Hamiltonian which only allows transitions between
states of restricted quantum numbers as formalised by the Wigner-Eckart theorem. Superselection
rules state that certain states vanish as matrix elements on all observables. I.e. the states |Ψ1 i and
|Ψ2 i are said to be separated by a superselection rule if for all observables O
hΨ1 |O|Ψ2 i = 0 , O any observale . (6.66)
This means that the relative phase between |Ψ1 i and |Ψ2 i is not observable and hence no coherent
state of the form a|Ψ1 i + b|Ψ2 i can be prepared (or exists). Hence the space of observables is smaller
than the space of hermitian operators (X = |Ψ1 ihΨ2 | + |Ψ2 ihΨ1 | = X † and hΨ1 |X|Ψ2 i 6= 0. X is
hermitian but not an observable as it contradicts (6.66).)
An example is given by the fermion boson quantum number. Suppose you had a coherent quantum
superposition of the form |ψi = a|j = 1/2i + b|j = 1i and subject it to 360◦ -rotation. This leads to
an expression,
(6.17)
U (360◦ )|ψi = −a|j = 1/2i + b|j = 1i , (6.67)
which is clearly non-sense, in the sense that the relative phase between a and b is unobservable. The
same happens for conserved charges; e.g. the electric charge for example. It does not make sense to
write |ψi = |e+ i+|e− i as they transform in complex conjugate representations which acquire different
phases under rotation of the group (in this case U (1)) associated with the conserved charge. What
about particles which transform under a real representations? Well they have no charges as previously
stated!
6.7 Wigner-Eckart applied: electric dipole selection rules

From electrodynamics we know that the (dominant mode) of interaction between an electric field
and a charge distribution is the dipole interaction H 0 = −d~e · E.
~ The electric dipole operator is
~
the first second term in the multiple interaction and defined as de = d3 xρ(x)~x (with ρ being the
R
charge density) and has the same transformation properties as ~x. Hence an applied external electric
field allows for new types of transitions. The ~x is clearly a j = 1 tensor operator and hence the
Wigner-Eckart selection rules apply.
Let us assume that the electric field is on the 3-direction which implies that H 0 ∼ z. We may
write the selection rules by writing
(6.60)
H 0 ∼ z ∼ Y10 (6.68)
and hence by virtue of the Wigner-Eckart theorem (6.26)

0
hlm|H 0 |l0 mi ∝ hlm|Y10 |l0 m0 i = Cm0m
l1l 0
0 hl||Y1 ||l i . (6.69)
Hence we have the following selection rules by virtue of Eq. (6.27):
m = m0 , ∆l ≡ |l − l0 | ≤ 1 , (6.70)
and in addition no transition if l = l0 = 0.

Eq. (6.70) can be refined by invoking the parity selection rule of the last section. By virtue of
η~x = −1 (6.65) and (6.63) this implies this forbids ∆l = 0 (i.e.∆l 6= 0 ) and hence ∆l ≤ 1 → ∆l = 1
under parity selection. This is in accordance with (6.64).
Exercise 6.7.1 Electric dipole moments

a) Suppose ~x was a pseudo vector, how would the selection rule change?
6.8 Pauli’s Hydrogen atom and SU (2) ⊗ SU (2)-symmetry

The quantum mechanical solution of the hydrogen atom with Coulomb potential, in units where
~ = c = 1, is given by
me e4
En = − , (6.71)
2n2
where n the principal quantum number. More precisely to each fixed n there corresponds l =
0, 1, 2..(n − 1) rotation representations (also called mutliplets). That is to say the total degeneracy is
n−1
X 2(n − 1)
(2l + 1) = n + n = n2 . (6.72)
2
l=0
What can be learned from this? Clearly we expect the states of fixed l to be (2l + 1)-degenerate
e.g. l = 1 m = 1, 0, −1 states. Yet for n = 2 there is l = 1 and l = 0 state and this should be
taken as a sign that there is more symmetry than meets the eye. In fact this has been known for a
long time (Kepler’s problem) that there is an additional conserved vector (Laplace, Runge and Lenz)
Ai = ijk pj Lk − me2 xri . For use in quantum mechanics the operator is symmetrised and becomes
xi
Ai = ijk pj Lk − me2 − ipi . (6.73)
r
The following relations are readily verified:
~ ·A
L ~=A
~·L
~ =0 (6.74)
[Li , Aj ] = iijk Ak (6.75)
2
pi pi e
[H, Ai ] = 0 , H= − (6.76)
2m r
~ is a conserved vector by virtue of Eqs.(6.14) and (6.28) respec-

where Eq. (6.76)(6.75) mean that A
tively. Hence a hidden symmetry has been identified!
Redefining14 Ãi = Ai (−2mH)−1/2 one gets
[Ãi , Ãj ] = iijk Ãk , (6.77)
which suggests to use,

1
Xi± = (Li ± Ãi ) , (6.78)
2
since
[Xi± , Xj± ] = iijk Xk± . (6.79)
This shows that there is a SU (2) ⊗ SU (2)-symmetry. An immediate question is what the two corre-
sponding Casimir operators are. It turns out that they are the same by virtue of Eq. (6.74):
C2+ = (X + )2 = (X − )2 = C2− (6.80)
and we shall parameterise it by C2 ≡ C2± = x(x + 1) with x = 0, 1/2, 1, 3/2, ...

A rather lengthy, but in principle straightforward computation yields the following relation:
Ai Ai = 2mH(Li Li + 1) + m2 e4 (6.81)
Dividing both sides by (−2mH) yields
me4 /2 me4 /2 me4 /2

H=− =− =− , x = 0, 1/2, 1, 3/2, .. . (6.82)
Li Li + Ãi Ãi + 1 4C2 + 1 (2x + 1)2
The orbital angular momentum is

Li = Xi+ + Xi−
and counting the degeneracy therefore amount to counting the states in the direct product represen-
tation ρj=x ⊗ ρj=x which is (2x + 1)2 Comparing with (6.71) suggest the following identification with
the principal quantum number n
n = (2x + 1) . (6.83)
Let us briefly pause and reflect what Pauli has done. He has solved the Hydrogen atom with group
theory alone. He has done this by rewriting the Hamiltonian, purely, in terms of the two Casimir
operators. Is this an accidental or not? Presumably not. In the absence of spin a single particle
state is described by three quantum numbers e.g. {x1 , x2 , x3 }, {p1 , p2 , p3 } or {n, l, m}. The spherical
symmetry guarantees that m does not enter and hence reduces it to two quantum numbers. So a rank
two symmetry (or two Casimir operators) can be expected to be sufficient to obtain the solutions with
group theory alone.
14 The Hamiltonian of a bound state problem usually has negative eigenvalues and is therefore chosen −H under the
square root.
Epilogue
I hope you have enjoyed this (brief) journey into group theory and some of its applications in quantum
mechanics. Group theory is an indispensable tool, as outlined at the beginning, as it often partly
solves complex dynamical equations. Even if you have not understood all the details of this course I
would hope that you take this on into your professional activities. There are various points where one
could have gone deeper in this course. For example a generic study of the simple and compact Lie
groups in Fig. 5.1. E.g. the classification of Dynkin and Cartan. Or the study of the representations
of the Poincaré group; in particular the so-called spinor calculus. The course as such certainly has a
lot of material already and hence it is time to close. Good luck with the exam!

romanSoQM 2015

Uploaded by

Copyright:

Available Formats

romanSoQM 2015

Uploaded by

Document Information

Original Title

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

romanSoQM 2015

Uploaded by

Copyright:

Available Formats

Symmetries of Quantum Mechanics

Roman Zwicky, University of Edinburgh

March 26, 2015

4 Representation theory of finite groups 34

4.4 Constructing representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5 Lie groups: U (1) ' SO(2), SO(3) ' SU (2)/Z2 56

6 Applications to quantum mechanics 74

1.2 Basic mathematical notions

In some sense it is a unit matrix.

• An equivalence relation: is a relation between two elements, say a and b denoted by a ∼ b

1.3 Mini-revision of linear algebra

(1) (~u, k~v + k 0 w)

u ↔ |ui , (u, v) ↔ hu|vi , (1.6)

which we shall use at times when illustrating examples.

• A linear operator is an operator which is compatible with linearity:

A(k|ui + k 0 |vi) = kA|ui + k 0 A|vi . (1.7)

hei |ej i = δij . (1.9)

An orthonormal set provides a decomposition of the identity

Aij = hei |Aej i ≡ hei |A|ej i , (1.13)

T (with t being the time), is an example of an anti-linear operator.

A fundamental fact is that the eigenvalue equation:

A|vi i = λi |vi i , (1.15)

1.3.1 Some properties of matrices

• The hermitian conjugate of a matrix is (A† )ij = A∗ji . A matrix is hermitian ⇔ A† = A.

• The inverse of a matrix A−1 is a matrix that satisfies: A−1 A = AA−1 = 1N .

• A matrix is diagonal ⇔ Aij = 0 for i 6= j.

hvi ⊕ wa |vj ⊕ wb iV ⊕W = hvi |vj iV + hwa |wb iW (1.17)

The dimension is dim(V ⊕ W ) = v + w (v ≡ dim(V ) and w ≡ dim(W )) Given A ∈ GL(V ) and

hvi ⊗ wa |vj ⊗ wb iV ⊗W = hvi |vj iV · hwa |wb iW (1.20)

(A ⊗ B)|vi i ⊗ |wa i = A|vi i ⊗ B|wa i . (1.21)

Properties associated with definitions above:

with obvious generalisation to n × n-matrices.

is diagonalised by the following unitary basis transformation7

4. The trace is cyclic Tr[AB] = Tr[BA].8

6. The determinant of a product of matrices is the product of its determinants.

det(AB) = det(A) · det(B)(= det(BA)) . (1.26)

7. The following relation holds:

hU v|U wi = hv|wi . (1.28)

Exercise 1.3.1 Basic linear algebra:

e) As additional exercises you might want to show that:

2.1 Basics of group theory

(g1 · g2 ) · g3 = g1 · (g2 · g3 ) ∀g1 , g2 , g3 ∈ G , product rule is associative

Exercise 2.1.1 To be discussed in the workshop

Examples 2.1.1 (basic ones)

Definition 2.1.2 Basic defintions:

seems very natural for investigating this object.

Exercise 2.1.2 To be discussed in the workshop

Theorem 2.1.1 subgroup criteria

S3 () (1,2) (2,3) (1,3) (1,2,3) (1,3,2)

Definition 2.1.3 group morphisms

∀a1 , b1 ∈ G1 : φ(a1 · b1 ) = φ(a1 ) · φ(b1 ) (2.2)

Exercise 2.1.4 To be discussed in the workshop

2.1.1 Group presentation

S3 = < a, b, c | a2 = b2 = c3 = abc = e > (2.3)

Exercise 2.1.5 To be discussed in the workshop

2.2 Notions of group theory

Remarks and properties:

famous Lagrange theorem:

with ni an appropriate map from integers to integers.4