MATHEMATICAL

STRUCTURES
of

QUANTUM
MECHANICS

8259hc.9789814366588-tp.indd 1 10/12/11 4:25 PM

This page intentionally left blank
 MATHEMATICAL
STRUCTURES
of

QUANTUM
MECHANICS

Kow Lung CHANG

National Taiwan University, Taiwan

World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI

8259hc.9789814366588-tp.indd 2 10/12/11 4:25 PM

 Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

MATHEMATICAL STRUCTURES OF QUANTUM MECHANICS

Copyright © 2012 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.

ISBN-13 978-981-4366-58-8
ISBN-10 981-4366-58-7

Printed in Singapore.

Kostas - Math'l Structures of Quan Mech.pmd 1 8/22/2011, 10:39 AM

TO FEZA AND SUHA
This page intentionally left blank
Preface

During the past few years, after a couple of weeks of lecturing the
course of quantum mechanics that I offered at the Physics Department,
National Taiwan University, some students would usually come to ask
me as to what extent they had to refurbish their mathematical back-
ground in order to follow my lecture with ease and confidence. It was
hard for me to provide a decent and proper answer to the question, and
very often students would show reluctance to invest extra time on sub-
jects such as group theory or functional analysis when I advised them
to take some advanced mathematics courses. All these experiences that
I have encountered in my class eventually motivated me to write this
book.

The book is designed with the hope that it might be helpful to those
students I mentioned above. It could also serve as a complementary text
in quantum mechanics for students of inquiring minds who appreciate
the rigor and beauty of quantum theory.

Assistance received from many sources made the appearance of this

book possible. I wish to express here my great appreciation and grati-
tude to Dr. Yusuf Gürsey, who painstakingly went through the manu-
script and responded generously by giving very helpful suggestions and
comments, and made corrections line by line. I would also like to thank
Mr. Paul Black who provided me with cogent suggestions and criticism
of the manuscript, particularly in those sections on quantum uncer-
tainty. I am indebted as well to Mr. Chih Han Lin who, with immense
patience, compiled the whole text and drew all the figures from my sug-
gestions. All his hard work and attention resulted in the present form
of this book.

Taipei, Taiwan
March, 2011 Kow Lung Chang

vii
This page intentionally left blank
Contents

1 Postulates and Principles of Quantum Mechanics 1

1.1 Vector space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Linearly dependent and linearly independent. . . . . . . . . . 2
1.1.2 Dimension and basis . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Inner product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Schwarz inequality . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Gram-Schmidt orthogonalization process . . . . . . . . . . . . 4
1.3 Completeness and Hilbert space . . . . . . . . . . . . . . . . . . . 6
1.3.1 Norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Cauchy sequence and convergent sequence . . . . . . . . . . . 7
1.3.3 Complete vector space . . . . . . . . . . . . . . . . . . . . . . 8
1.3.4 Hilbert space . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Linear operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1 Bounded operator . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Continuous operator . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.3 Inverse operator . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.4 Unitary operator . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.5 Adjoint operator . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.6 Hermitian operator . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.7 Projection operator . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.8 Idempotent operator . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 The postulates of quantum mechanics . . . . . . . . . . . . . . . . 12
1.6 Commutability and compatibility of dynamical observables. . . . . 16
1.6.1 Compatible observables . . . . . . . . . . . . . . . . . . . . . 16

ix
x Contents
1.6.2 Intrinsic compatibility of the dynamical observables and the
direct product space . . . . . . . . . . . . . . . . . . . . . . . 19
1.6.3 3rd postulate of quantum mechanics and commutator algebra 20
1.7 Non-commuting operators and the uncertainty principle . . . . . . 22
1.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2 Space-Time Translation, Quantum Dynamics

and Various Representations in Quantum Mechanics 27
2.1 Vector space and dual vector space . . . . . . . . . . . . . . . . . . 27
2.2 q-representation and p-representation in quantum mechanics . . . . 33
2.3 Harmonic oscillator revisited . . . . . . . . . . . . . . . . . . . . . 40
2.3.1 Creation and annihilation operators . . . . . . . . . . . . . . 41
2.4 N-representation and the Rodrigues formula of Hermite
polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.5 Two-dimensional harmonic oscillation and direct product of
vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.6 Elastic medium and the quantization of scalar field . . . . . . . . . 56
2.7 Time evolution operator and the postulate of quantum dynamics . 64
2.7.1 Time evolution operator and Schrödinger equation . . . . . . 64
2.7.2 Time order product and construction of time evolution
operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.8 Schrödinger picture vs. Heisenberg picture. . . . . . . . . . . . . . 68
2.9 Propagator in quantum mechanics . . . . . . . . . . . . . . . . . . 71
2.10 Newtonian mechanics regained in the classical limit . . . . . . . . 73
2.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

3 Symmetry, Transformation and Continuous Groups 83

3.1 Symmetry and transformation . . . . . . . . . . . . . . . . . . . . 83
3.1.1 Groups and group parameters. . . . . . . . . . . . . . . . . . 83
3.2 Lie groups and Lie algebras . . . . . . . . . . . . . . . . . . . . . . 87
3.3 More on semisimple group . . . . . . . . . . . . . . . . . . . . . . 94
3.3.1 Cartan’s criteria of the semisimple group. . . . . . . . . . . . 94
3.3.2 Casimir operator . . . . . . . . . . . . . . . . . . . . . . . . . 96
Contents xi

3.4 Standard form of the semisimple Lie algebras . . . . . . . . . . . . 98

3.5 Root vector and its properties . . . . . . . . . . . . . . . . . . . . 101
3.6 Vector diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.7 PCT: discrete symmetry, discrete groups. . . . . . . . . . . . . . . 107
3.7.1 Parity transformation . . . . . . . . . . . . . . . . . . . . . . 107
3.7.2 Charge conjugation and time reversal transformation . . . . . 110
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

4 Angular Momentum 117

4.1 O(3) group, SU(2) group and angular momentum . . . . . . . . . . 117
4.1.1 O(3) group . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.1.2 U(2) group and SU(2) group . . . . . . . . . . . . . . . . . . 124
4.2 O(3)/SU(2) algebras and angular momentum . . . . . . . . . . . . 128
4.3 Irreducible representations of O(3) group and spherical harmonics . 135
4.4 O(4) group, dynamical symmetry and the hydrogen atom . . . . . 141
4.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

5 Lorentz Transformation, O(3,1)/SL(2,C) Group and

the Dirac Equation 151
5.1 Space-time structure and Minkowski space. . . . . . . . . . . . . . 151
5.1.1 Homogeneous Lorentz transformation and SO(3,1) group . . . 152
5.2 Irreducible representation of SO(3,1) and Lorentz spinors. . . . . . 156
5.3 SL(2,C) group and the Lorentz transformation . . . . . . . . . . . 160
5.4 Chiral transformation and spinor algebra . . . . . . . . . . . . . . 163
5.5 Lorentz spinors and the Dirac equation . . . . . . . . . . . . . . . 168
5.6 Electromagnetic interaction and gyromagnetic ratio of the electron 177
5.7 Gamma matrix algebra and PCT in Dirac spinor system . . . . . . 179
5.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

Bibliography 189

Index 191
Chapter 1

Postulates and Principles of Quantum

Mechanics

As with many fields in physics, a precise and rigorous description

of a given subject requires the use of some mathematical tools. Take
Lagrange’s formulation of classical mechanics for instance, one needs the
basic knowledge of variational calculus in order to derive the equations
of motion for a system of particles in terms of generalized coordinates.
To formulate the postulates of quantum mechanics, it would also be
necessary to acquire some knowledge on vector space in general, and
Hilbert space in particular. It is in this chapter that we shall provide
the minimum but essential mathematical preparation that allows one to
perceive and understand the general framework of quantum theory and
to appreciate the rigorous derivation of the quantum principles.

1.1 Vector space

A vector space V is a set of elements, called vectors, with the

following 2 op erations:

• An operation of addition, which for each pair of vectors ψ and φ,

corresponds to a new vector ψ + φ ∈ V, called the sum of ψ and
φ.

• An operation of scalar multiplication, which for each vector ψ and

a number a, specifies a vector aψ, such that (assuming a, b are
numbers and ψ, φ and χ are vectors)

1
2 Chapter 1

ψ + φ = φ + ψ, (1.1a)
ψ + (φ + χ) = (ψ + φ) + χ, (1.1b)
ψ + 0 = ψ, 0 is null vector, (1.1c)
a(ψ + φ) = aψ + aφ, (1.1d)
(a + b)ψ = aψ + bψ, (1.1e)
a(bψ) = (ab)ψ, (1.1f)
1 · ψ = ψ, (1.1g)
0 · ψ = 0, (1.1h)

where if a, b are real numbers, we call this vector space the real vector
space, and denote it by Vr . On the other way, complex vector space Vc
means a, b are complex numbers.

Example
We take n-dimensional Euclidean space, Rn -space, as an exmple. It
is a vector space with the vectors ψ and φ specified as ψ = (x1 , x2 , . . . ,
xi , . . . , xn ) and φ = (y1 , y2 , . . . , yi , . . . , yn ), where xi and yi (i = 1, 2, . . . ,
n) are all taken as real numbers. The sum of ψ and φ becomes (x1 +
y1 , x2 +y2 , . . . , xi +yi , . . . , xn +yn ) and aψ = (ax1 , ax2 , . . . , axi , . . . , axn ).
If a and xi are taken as complex numbers, then ψ is a vector in C n -space;
a n-dimensional complex vector space.
It is easily understood that a set of the continuous functions f (x)
for a 6 x 6 b forms a vector space, namely L2 (a, b)-space.
Before leaving this section, we also introduce some terminologies
in the following subsections that will be frequently referred to in later
chapters.

1.1.1 Linearly dependent and linearly independent

Consider a set of m vectors {ψ1 , ψ2 , . . . , ψm }, and we construct the
linear combination of these m vectors as follows:

m
X
ai ψi . (1.2)
i=1
1.2 Inner product 3

This linear combination of m vectors is of course a vector. It becomes

a null vector if and only if all the coefficient ai = 0 for i = 1, 2, . . . , m,
then the set of m vectors {ψ1 , ψ2 , . . . , ψm } is called linearly independent.
If at least one of the coefficient al 6= 0 such that m
P
i=1 a i ψi = 0 , then
the set {ψ1 , ψ2 , . . . , ψm } is called linearly dependent.

1.1.2 Dimension and basis

The maximum number of linearly independent vectors in V is called
the dimension of V. Any n-linearly independent vectors in vector space
V of n-dimension form the basis of the vector space.

1.2 Inner product

An inner product, or sometimes called scalar product in vector

space, is a numerically valued function of the ordered pair of vectors ψ
and φ, denoted by (ψ, φ), and for a scalar a, such that
(ψ, φ + χ) = (ψ, φ) + (ψ, χ), (1.3a)
(ψ, aφ) = a(ψ, φ), (1.3b)
∗
(ψ, φ) = (φ, ψ) , (1.3c)
(ψ, ψ) > 0, (ψ, ψ) = 0 if and only if ψ is a null vector. (1.3d)

Two vectors ψ and φ are said to be orthogonal to each other if their

corresponding inner product vanishes, namely (ψ, φ) = 0.
For example, let us consider the vectors in C n -space ψ = (x1 , x2 , . . . ,
xn ) and φ = (y1 , y2 , . . . , yn ) where xi and yi are complex numbers. The
inner product of ψ and φ written as

n
X
(ψ, φ) = x∗i yi = x∗1 y1 + x∗2 y2 + · · · + x∗n yn . (1.4)
i=1

Consider the set of continuous function of f (x) where a 6 x 6 b.

An ordered pair of functions f (x) and g(x) define the inner product as
Rb
(f (x), g(x)) = a f (x)∗ g(x)dx. This vector space is called L2 (a, b)-space
when |f (x)|2 and |g(x)|2 = finite.
4 Chapter 1

1.2.1 Schwarz inequality

We are now in the position to prove the Schwarz inequality.
Let ψ and φ be any two vectors. The Schwarz inequality reads as
p p p
|(ψ, φ)| = (ψ, φ)(φ, ψ) 6 (ψ, ψ) (φ, φ). (1.5)

Proof
Since (ψ + αφ, ψ + αφ) > 0, where α = ξ + iη is a complex number.
Regard this inner product (ψ + αφ, ψ + αφ) = f (ξ, η) as a function of
two variables ξ and η. Then
f (ξ, η) = (ψ, ψ) + |α|2 (φ, φ) + α(ψ, φ) + α∗ (φ, ψ), (1.6)

which is positive definite. Let us look for the minimum of f (ξ, η) at

ξ0 , η0 by solving

∂f (ξ, η) ∂f (ξ, η)
= = 0, (1.7)
∂ξ ∂η
ξ0 ,η0 ξ0 ,η0

and we obtain

1 (ψ, φ) + (φ, ψ) i (ψ, φ) − (φ, ψ)

ξ0 = , η0 = − . (1.8)
2 (φ, φ) 2 (φ, φ)

Therefore

(ψ, φ)(φ, ψ)
f (ξ0 , η0 ) = (ψ, ψ) − > 0, (1.9)
(φ, φ)

that can be cast into the familiar expression of Schwarz inequality.

1.2.2 Gram-Schmidt orthogonalization process

The inner product we have been considering can be applied to the
orthogonalization of the basis in the n-dimensional vector space. Let
1.2 Inner product 5

{ψ1 , ψ2 , . . . , ψn } ∈ V be the set of n-linearly independent vectors.

Since (ψi , ψj ) 6= 0 in general, we can construct a new set of vectors
{ψ10 , ψ20 , . . . , ψn0 } such that (ψi0 , ψj0 ) = 0 for all i and j unless i = j,
namely ψi0 and ψj0 are orthogonal to each other for i 6= j by the follow-
ing procedure:
First take ψ10 = ψ1 and construct ψ20 = ψ2 + αψ10 . In order to force
ψ2 to be orthogonal to ψ10 , we solve the α as to meet the condition
0

(ψ20 , ψ10 ) = 0, i.e.

(ψ20 , ψ10 ) = (ψ2 , ψ10 ) + α∗ (ψ10 , ψ10 ) = 0, (1.10)

and we obtain α = −(ψ2 , ψ10 )∗ /(ψ10 , ψ10 ) = −(ψ10 , ψ2 )/(ψ10 , ψ10 ), hence

(ψ10 , ψ2 )
ψ20 = ψ2 − ψ10 . (1.11)
(ψ10 , ψ10 )

The same procedure can be performed repeatedly to reach ψ30 =

ψ3 + αψ20 + βψ10 which guarantees (ψ30 , ψ10 ) = (ψ30 , ψ20 ) = 0 with α =
−(ψ20 , ψ3 )/(ψ20 , ψ20 ) and β = −(ψ10 , ψ3 )/(ψ10 , ψ10 ). In general,

0 ,ψ )
(ψi−1 0 ,ψ )
(ψi−2 0
i i 0 (ψ1 , ψi )
ψi0 = ψi − ψi−1
0
0 , ψ0 ) − ψ 0
i−2 0 , ψ0 ) − · · · − ψ1 .
(ψi−1 i−1 (ψi−2 i−2 (ψ10 , ψ10 )

(1.12)

The set of orthogonal basis {ψ10 , ψ20 , . . . , ψn0 } can be normalized im-
mediately by multiplying the inverse square root of the corresponding
inner product, i.e.

ψi0
ψ̃i = p , (1.13)
(ψi0 , ψi0 )

and {ψ˜1 , ψ˜2 , . . . , ψ˜n } becomes the orthonormal set of the basis in the
vector space. From now on we shall take the basis to be orthonormal
without mentioning it particularly.
6 Chapter 1

Example
Consider the following set of continuous functions in C(−∞, ∞)
 2
x
fn (x) = xn exp − , n = 0, 1, . . . . (1.14)
2

We construct the new set of orthogonal vectors by applying the Gram-

Schmidt process and obtain:
 x
f00 (x) = f0 (x) = exp − , (1.15)
2

f 0 (f 0 , f1 ) x2
 
f10 (x) = f1 − 0 0 0 0 = f1 (x) = x exp , (1.16)
(f0 , f0 ) 2

f 0 (f 0 , f2 ) f 0 (f 0 , f2 )
   2
1 x
f20 (x) = f2 − 1 0 1 0 − 0 0 0 0 = 2
x − exp − . (1.17)
(f1 , f1 ) (f0 , f0 ) 2 2

Similarly we have f30 (x) = (x3 − 3x/2) exp(−x2 /2). The orthonormal
functions can be calculated according to

fn0 (x)
 2
1 x
f˜n (x) = p =p √ exp − Hn (x), (1.18)
0 0
(fn (x), fn (x)) n
2 n! π 2

where Hn (x) are called Hermite polynomials. One also recognizes that
f˜n (x) are in fact, the eigenfunctions of the Schrödinger equation for
one-dimension harmonic oscillation.

1.3 Completeness and Hilbert space

Let us introduce some other terminologies in discussing Hilbert space.

1.3.1 Norm
A norm on a vector space is a non-negative real function such that,
if ψ, φ are vectors, the norm of ψ is written as kψk, satisfying:
1.3 Completeness and Hilbert space 7

kψk > 0, kψk = 0 iff ψ is null vector, (1.19a)

kaψk = |a| · kψk, (1.19b)
kψ + φk 6 kψk + kφk. (1.19c)

Example

If f (x) ∈ C(a, b), namely if f (x) is a continuous function for a vari-

able that lies between a and b, the norm of f (x) can be defined either
as kf (x)k = Max{|f (x)|, a 6 x 6 b} or as the inner product of f (x), i.e.

Z b
2
kf (x)k = (f (x), f (x)) = |f (x)|2 dx.
a

1.3.2 Cauchy sequence and convergent sequence

Consider an infinite dimensional vector space and denote P the basis

by {φ1 , φ2 , φ3 , . . .}. We construct the partial sum ψN = i ai φi , where
i runs from 1 to N , and obtain . . . , ψj , ψj+1 , . . . , ψm , ψm+1 , . . . , ψn , . . .
for increasing values in N that forms an infinite sequence. The sequence
is called a Cauchy sequence if

lim ψn = lim ψm ,
n→∞ m→∞

or more precisely to put in terms of norm, i.e, lim kψn − ψm k = 0.

n,m→∞

It is said that a vector ψm converges to ψ if

lim ψm = ψ, or lim kψm − ψk = 0,

m→∞ m→∞

then {. . . , ψm−1 , ψm , . . .} is called a convergent sequence.

It is easily concluded that every convergent sequence is a Cauchy
sequence. Yet it is not necessary true conversely. Namely a Cauchy
sequence is not always a convergent sequence.
8 Chapter 1

1.3.3 Complete vector space

A vector space, in which every Cauchy sequence of a vector ψm
converges to a limiting vector ψ, is called a complete vector space.

1.3.4 Hilbert space

A Hilbert space is a complete vector space with norm defined as the
inner product. A Hilbert space, finite dimensional or infinite dimen-
sional, is separable if its basis is countable.

1.4 Linear operator

A linear operator A on a vector space assigns to each vector ψ a

new vector, i.e. Aψ = ψ 0 such that

A(ψ + φ) = Aψ + Aφ, A(αψ) = αAψ. (1.20)

Two operators A, B are said equal if Aψ = Bψ for all ψ in the

vector space.
For convenience in later discussion, we denote

• O: null operator such that Oψ = 0 for all ψ, and 0 is the null

vector.

• I: unit operator or identity operator such that Iψ = ψ.

The sum of the operators A and B is an operator, such that (A+B)ψ =

Aψ + Bψ. The product of operators A and B is again an operator that
one writes as A · B or AB such that (AB)ψ = A(Bψ).
The order of the operators in the product matters greatly. It is
generally that AB 6= BA. The associative rule holds for the product of
the operators A(BC) = (AB)C.

1.4.1 Bounded operator

An operator A is called a bounded operator if there exists a
positive number b such that

kAψk 6 bkψk, for any vector ψ in the vector space.

1.4 Linear operator 9

The least upperbound (supremum) of A, namely the smalleast num-

ber of b for a given operator A and for any ψ in V, is denoted by

 
kAψk
kAk = sup , ψ 6= 0 , (1.21)
kψk

then kAψk 6 kAkkψk.

We are now able to show readily that kA + Bk 6 kAk + kBk.

Proof

Let us denote kAψk 6 kAkkψk and kBψk 6 kBkkψk. Then

   
k(A + B)ψk kAψ + Bψk
kA + Bk = sup , ψ 6= 0 = sup , ψ 6= 0
kψk kψk
   
kAψk kBψk
6 sup , ψ 6= 0 + sup , ψ 6= 0 = kAk + kBk.
kψk kψk

Similarly, we have kABk 6 kAkkBk.

1.4.2 Continuous operator

Consider the convergent sequence {. . . , ψm , ψm+1 , . . . , ψn , . . .} such

that lim kψn − ψk = 0. If A is a bounded operator, then {. . . , Aψm ,
n→∞

Aψm+1 , . . . , Aψn , . . .} is also a convergent sequence because

lim kAψn − Aψk 6 kAk lim kψn − ψk = 0.

n→∞ n→∞

We call operator A the continuous operator.

10 Chapter 1

1.4.3 Inverse operator

An operator A has an inverse operator if there exists BR such that
ABR = I, then we call operator BR the right inverse of A. Similarly
an operator BL such that the product operator BL A = I, then we call
operator BL the left inverse of A. In fact, the left inverse operator
is always equal to the right inverse operator for a given operator A,
because

BL = BL I = BL (ABR ) = (BL A)BR = IBR = BR . (1.22)

The inverse operator of a given operator A is also unique. If opera-

tors B and C are all inverse operators of A, then C = CI = C(AB) =
(CA)B = B.
The implication of uniqueness of the inverse operator of operator A
allows us to write it in the form A−1 , namely AA−1 = A−1 A = I. It
is easily verified that (AB)−1 = B−1 A−1 .

1.4.4 Unitary operator

An operator U is unitary if kUψk = kψk. A unitary operation
preserves the invariant of the inner product of any pair of vectors, i.e.
(Uψ, Uφ) = (ψ, φ). This can be proved as follows:
Let χ = ψ + φ and we have

(Uχ, Uχ) = (U(ψ + φ), U(ψ + φ))

= (Uψ, Uψ) + (Uψ, Uφ) + (Uφ, Uψ) + (Uφ, Uφ)
= kUψk2 + kUφk2 + 2<{(Uψ, Uφ)},

and on the other hand,

(Uχ, Uχ) = (ψ + φ, ψ + φ)
= (χ, χ) = (ψ, ψ) + (ψ, φ) + (φ, ψ) + (φ, φ)
= kψk2 + kφk2 + 2<{(ψ, φ)}.

Since kUψk = kψk, kUφk = kφk, we have <{(Uψ, Uφ)} = <{(ψ, φ)}.
Similarly if χ0 = ψ + iφ, we obtain (Uχ0 , Uχ0 ) = (χ0 , χ0 ), that implies
={(Uψ, Uφ)} = ={(ψ, φ)}, therefore (Uψ, Uφ) = (ψ, φ).
1.4 Linear operator 11

1.4.5 Adjoint operator

Consider the inner product of (ψ, Aφ) where A is a given linear op-
erator of interest. This numerically scalar quantity certainly is a func-
tion of operator A and the pair of vectors ψ and φ, namely (ψ, Aφ) =
F (A, ψ, φ) is a scalar quantity.
Instead of performing the above inner product straightforwardly, we
shall obtain the very same scalar of (ψ, Aφ) by forming the following
inner product (A† ψ, φ) such that (ψ, Aφ) ≡ (A† ψ, φ). The operator A†
is called the adjoint operator of A. The following relations can be
easily established (proofs left to readers):

(A + B)† = A† + B† , (1.23a)

(αA)† = α∗ A† , (1.23b)

(AB)† = B† A† , (1.23c)

(A† )† = A, (1.23d)

(A† )−1 = (A−1 )† . (1.23e)

It can also be shown that A† is a bounded operator if A is bounded

and their norms are equal, i.e. kAk = kA† k.
To prove the above equality, let us consider kA† ψk2 = (A† ψ, A† ψ),
namely

kA† ψk2 = (A† ψ, A† ψ) = (AA† ψ, ψ) 6 kψkkAA† ψk 6 kψkkAkkA† ψk,

therefore kA† ψk 6 kAkkψk, and we have kA† k 6 kAk.

On the other hand, we have kAψk2 = (Aψ, Aψ) = (A† Aψ, ψ) 6
kψkkA† kkAψk, which implies kAk 6 kA† k. Therefore kAk = kA† k is
established.
12 Chapter 1

1.4.6 Hermitian operator

When an operator is self-adjoint, namely an adjoint operator A†

equals to operator A itself, i.e. A = A† , then we call A a Hermitian
operator.

1.4.7 Projection operator

Let H be a Hilbert space in which we consider a subspace M and
its orthogonal complement space M⊥ such that for each vector ψ in
H = M ⊕ M⊥ that are decomposed into unique vectors ψM in M and
ψM⊥ in M⊥ such that ψ = ψM + ψM⊥ , and (ψM , ψM⊥ ) = 0.
The projection operator PM when acting upon vector ψ onto a
subspace results in PM ψ = ψM . It is obvious that PM ψ = ψ if ψ ∈ M
and PM ψ = 0 if ψ ∈ M⊥ .
One can also be easily convinced that

(ψ, PM φ) = (ψ, φM ) = (ψM + ψM⊥ , φM ) = (ψM , φM )

= (ψM , φ) = (PM ψM , φ) = (PM ψ, φ).

Therefore PM is also a Hermitian operator, i.e. P†M = PM .

Similarly we define PM⊥ such that PM⊥ ψ = ψM⊥ and the sum of
PM and PM⊥ becomes an identity operator, i.e.

PM + PM⊥ = I.

1.4.8 Idempotent operator

The projection operator is an idempotent operator, namely
P2M = PM because P2M ψ = PM ψM = PM ψ.

1.5 The postulates of quantum mechanics

We start to formulate the postulates of quantum mechanics. We

shall treat the first three postulates in this chapter, and leave the 4th
postulate for the next chapter when we investigate the time evolution
of a quantum system.
1.5 The postulates of quantum mechanics 13

1st postulate of quantum mechanics:

For every physical system, there exists an abstract entity, called
the state (or the state function or wave function that shall be
discussed later), which provides the information of the dynamical
quantities of the system; such as coordinates, momenta, energy,
angular momentum, charge or isospin, etc. All the states for a
given physical system are elements of a Hilbert space, i.e.

physical system ←→ Hilbert space H

physical state ←→ state vector ψ in H

Furthermore for each physical observable, such as the 3rd compo-

nent of the angular momentum or the total energy of the system
and so forth, there associates a unique Hermitian operator in the
Hilbert space, i.e.

physical (dynamical) corresponding

observable hermitean operator
total energy E ←→ H = H†
coordinate ~x ←→ X = X†
~
angular momentum l ←→ L = L†

The physical quantity measured in the system for the corresponding

observable is obtained by taking the inner product of the pair ψ and
Aψ, i.e.

hAi = (ψ, Aψ), (1.24)

which is called the expectation value of dynamical quantity A for the

system in the state ψ, which is normalized, i.e. kψk = 1.
Since the action of operator A upon the vector ψ changes it into
another vector φ, which implies that the action of the measurement
of the dynamical quantity in a certain state usually would disturb the
physical system and the original state is changed into another state due
14 Chapter 1

to the external disturbance accompanying the measurement.

In particular, if an operator A such that Aψa = aψa , i.e. when A
acts upon a particular physical state ψa , the resultant state is the same
as the one before, then it is said that the physical state is prepared for
the measurement of the dynamical observable associated with the
operator A. We shall name:

• ψa : the state particularly prepared in the system for the mea-

surement of the dynamic quantity, called the eigenstate of the
operator A.

• a : the value of the measurement of the dynamical quantity in the

particular prepared state, called the eigenvalue of the operator
A.

We shall now explore some properties concerning the eigenvectors

and the eigenvalues through a few propositions.

Proposition 1.
The eigenvalues for a Hermitian operator are all real.

Let Aψa = aψa and A† ψa = aψa , and consider the inner product
hAiψa = (ψa , Aψa ) = (ψa , aψa ) = a(ψa , ψa ) = a. On the other hand, we
have hAiψa = (A† ψa , ψa ) = (Aψa , ψa ) = a∗ (ψa , ψa ) = a∗ which implies
a = a∗ if ψa is not a null vector.

Proposition 2.
Two eigenvectors of a Hermitian operator are orthogonal to each
other if the corresponding eigenvalues are unequal.

Let Aψa = aψa and Aψb = bψb where ψa 6= ψb , and since

(ψa , Aψb ) = b(ψa , ψb ) = (A† ψa , ψb ) = a∗ (ψa , ψb ) = a(ψa , ψb ),
1.5 The postulates of quantum mechanics 15

therefore (a − b)(ψa , ψb ) = 0. That implies (ψa , ψb ) = 0 if a 6= b.

It often occurs that there exists more than one eigenvector of an
operator with the same eigenvalue. Consider the Hermitian operator
C, such that

Cψci = ci ψci , (1.25)

where c1 = c2 = . . . = cm = c, and (ψc1 , ψc2 , . . . , ψcm ) are linearly

independent.
The eigenvalue c is called m-fold degenerate if there are m linearly
independent eigenvectors corresponding to the same eigenvalue c of the
operator C.

Proposition 3.
If the eigenvalue c of the operator C is degenerate, any linear
combination of the linearly independent eigenvectors is also an
eigenvector.

P
Due to the linearity of operator C, the linear combination αi ψci is
also the eigenvector of C with the eigenvalue c. By means of the Gram-
Schmidt orthogonalization process, one is able to easily construct a new
set of orthonormal vectors {ψ̃c1 , ψ̃c2 , . . . , ψ̃cm } out of the previous m
linearly independent set {ψc1 , ψc2 , . . . , ψcm }, such that

(ψ̃ci , ψ̃cj , ) = δij . (1.26)

Following the results of Propositions 2 and 3, we conclude that

(ψ̃ai , ψ̃bj , ) = δab δij , (1.27)

where a, b refer to the eigenvalues and i, j refer to the index of degener-

acy. We shall drop ˜ on top of the orthonormal basis without mentioning
it further.
16 Chapter 1

2nd postulate of quantum mechanics:

The set of eigenvectors ψai of a given Hermitian operator corre-
sponding to a physical observable form the basis of a Hilbert space.
Any state in the physical system can be denoted by a vector
P in the
Hilbert space as a linear combination of ψai , i.e. ψ = αai ψai ,
where (ψai , ψa0j ) = δaa0 δij .

The coefficient αai is obtained by taking the inner product

(ψai , ψ) = αai . (1.28)

The formulation of the 2nd postulate of quantum mechanics is purely

artificial. In fact, it has been proved and well studied that a function
space can be spanned by the eigenvectors of a Hermitian Sturm-Liouville
operator. Since the dynamical operators in quantum system are not
confined to those of the Sturm-Liouville form, we would formulate on
purpose the second postulate of quantum mechanics in order to build
and integrate the whole mathematical structure and the logical devel-
opment of the quantum theory on a solid and self consistent ground.

1.6 Commutability and compatibility of dynamical

observables

We shall introduce in the following subsections some terminologies

which will help to differentiate various types of the compatible observ-
able.

1.6.1 Compatible observables

If there exist a complete set of linearly independent vectors ψai which

are eigenstates of both operators R and S, then the two physical ob-
servables corresponding respectively to Hermitian operators R and S
are said to be compatible.
1.6 Commutability and compatibility of dynamical observables 17

Proposition 4.
If two observables are compatible, their corresponding operators
R and S commute, i.e. [R, S] = 0.

It is obvious because R and S have the following properties:

Rψa = ra ψa , and Sψa = sa ψa ,

which lead to
(RS − SR)ψa = [R, S]ψa = 0 ,

if we define the commutator of R and S as [R, S] = RS − SR.

Therefore [R, S]ψ = 0 for any ψ, and Proposition 4 is established.

Proposition 5.
If R and S are operators corresponding to two compatible ob-
servables, and if ψr are eigenvectors of R, then
(ψr , Sψr0 ) = 0, for r 6= r0 . (1.29)

The proof of Proposition 5 is straightforward, i.e.

r0 (ψr , Sψr0 ) = (ψr , SRψr0 ) = (ψr , RSψr0 ) = (Rψr , Sψr0 ) = r(ψr , Sψr0 ).

We have (r0 − r)(ψr , Sψr0 ) = 0. Hence Proposition 5 is proved.

Proposition 6.
If Pr is the projection operator onto subspace with vectors ψr ,
then
[Pr , S] = 0.

The proof is again straightforward. For

18 Chapter 1

Pr ψr0 = δrr0 ψr0 , (1.30)

we have

(ψr0 , [Pr , S]ψr00 ) = (ψr0 , (Pr S − SPr )ψr00 )

= (Pr ψr0 , Sψr00 ) − (ψr0 , SPr ψr00 )
= (δrr0 − δrr00 )(ψr0 , Sψr00 ) ≡ 0, (prove it)

that leads to (ψ, [Pr , S]ψ) = 0 for all ψ, hence

[Pr , S] = 0. (1.31)

Proposition 7.
If R and S are two commuting Hermitian operators, there exists
a complete set of states which are simultaneously eigenvectors of
R and S.

(s)
Let us construct a vector φr , projected by Ps upon the vector ψr
which is the eigenvector of the Hermitian operator R, i.e.

φ(s)
r = Ps ψr . (1.32)

(s)
It is obvious that φr is automatically the eigenvector of the Her-
mitian operator S with eigenvalue s, namely

Sφ(s) (s)
r = sφr . (1.33)

(s)
On the other hand, Proposition 6 ensures that φr is also an eigen-
vector of the operator R , because

Rφ(s) (s)
r = RPs ψr = Ps Rψr = rPs ψr = rφr . (1.34)

Hence Proposition 7 is proved.

1.6 Commutability and compatibility of dynamical observables 19

1.6.2 Intrinsic compatibility of the dynamical observables and

the direct product space
We have seen in the last section that if two operators corresponding
respectively to two dynamical observables commute with each other,
there always exists a complete set of states which are simultaneously
eigenvectors of these two operators. The construction of the simulta-
neous eigenvectors could be simplified even further in some particular
cases in which the compatibility of these two operators is solely based
upon the first and the second fundamental commutation relations,
i.e. [qi , qj ] = [pi , pj ] = 0, without making use of the third fundamen-
tal commutation relation. The dynamical observables corresponding to
the commuting operators in this particular category are said to be in-
trinsic compatible. The construction of the simultaneous eigenvectors
of the intrinsic compatible observables is formulated in the following
proposition.

Proposition 8.
Let A and B be two operators corresponding respectively to two
intrinsic compatible observables, and let ψai and ϕbj be the eigen-
vectors of A and B with the eigenvalues ai and bj respectively,
then the direct product of ψai and ϕbj , denoted by ψai ⊗ ϕbj is the
eigenvector of the operator F(A, B) with the eigenvalue F (ai , bj ).

It can be easily shown that ψai ⊗ ϕbj is the simultaneous eigenvector

of A and B with eigenvalues ai and bj respectively if we make the
following identifications:
A = F(A, I), and B = F(I, B).
Two operators in different Hilbert spaces are always of intrinsic com-
patibility. Consider the system of a particle with spin, the physical ob-
servable associated with the configuration space is compatible with the
physical observable in the spin space. Therefore one is able to express
the quantum state as the direct product of a vector in the configuration
space and another vector in spin space.
20 Chapter 1

We shall leave it for the reader to verify that Propositions 5 through

7 are consistent with the above formulation. We will also elaborate
more on the algebra of the direct product space in Section 2.5 in order
to provide a rigorous proof for Proposition 8.

1.6.3 3rd postulate of quantum mechanics and

commutator algebra

3rd postulate of quantum mechanics:

Every Poisson bracket in classical mechanics for canonical vari-
ables (pi , qj ) is replaced by the commutator of the corresponding
operators with the following relations:
Classical mechanics Quantum mechanics
[qi , qj ] = 0 → [Qi , Qj ] = 0
[pi , pj ] = 0 → [Pi , Pj ] = 0
~
[pi , qj ] = δij → [Pi , Qj ] = δij
i

where h = 2π~ is Planck’s constant.

We discuss the commutator algebra for further applications in later

chapters. The commutator of operators A and B, as it is defined pre-
viously

[A, B] = AB − BA, (1.35)

then we have

[A, B] = −[B, A], (1.36a)

[A, A] = O, (1.36b)
[A, B + C] = [A, B] + [A, C], (1.36c)
[A, [B, C]] + [B, [C, A]] + [C, [A, B]] = O. (1.36d)
1.6 Commutability and compatibility of dynamical observables 21

An operator C is called a constant operator if it commutes with

any operator corresponding to the dynamical observables. Obviously
any real number times unit operator I is a Hermitian constant operator.
When an operator is exponentiated, namely eA , it is defined as the
usual sense of the exponential function, i.e.

1 2 1
eA = I + A + A + A3 + . . . . (1.37)
2! 3!

We shall now show a useful identity

1 1
eA Be−A ≡ B + [A, B] + [A, [A, B]] + . . . . (1.38)
1! 2!

Proof

Let f (λ) = eλA Be−λA and expand the function f (λ) in terms of
power series of λ at λ = 0, i.e.

λ 0 λ2
f (λ) = f (0) + f (0) + f 00 (0) + . . . .
1! 2!

Since,

f 0 (λ) = Af (λ) − f (λ)A = [A, f (λ)],

we can evaluate each order of the derivatives of f (λ) at λ = 0, i.e.

f (0) = B,
f 0 (0) = [A, B],
f 00 (0) = [A, [A, B]],

and thus

λ 0 λ2
f (λ) = f (0) + f (0) + f 00 (0) + . . .
1! 2!
22 Chapter 1

becomes

λ2
f (λ) = eλA Be−λA = B + λ[A, B] + [A, [A, B]] + . . . .
2!

Therefore we reach the identity by setting λ = 1 in the function

f (λ), i.e.

1
f (1) = eA Be−A ≡ B + [A, B] + [A, [A, B]] + . . . . (1.39)
2!

1.7 Non-commuting operators and the uncertainty

principle

The uncertainty in the simultaneous measurement of the coordinate

and its conjugate momentum is due to the dual nature of a quantum
particle, namely the particle-wave duality. From the mathematical
point of view, the uncertainty originates from the non-commutability
of the two operators that correspond respectively to these two phys-
ical observables. In fact, uncertainty will arise for any simultaneous
measurement of a pair of dynamical quantities if their corresponding
quantum operators do not commute. A demonstration is in order to re-
establish Heisenberg’s famous uncertainty relations. For simplicity, we
only consider the one-dimensional case and denote ∆q and ∆p to be the
uncertainty in coordinate and momentum simultaneous measurement
respectively. ∆q and ∆p are also called the variance, or the deviation
of q and p correspondingly defined as

(∆q)2 = (ψ, (Q − hQiI)2 ψ), (1.40)

(∆p)2 = (ψ, (P − hPiI)2 ψ), (1.41)

where ψ is the quantum state of interest, hQiI and hPiI are both con-
stant operators obtained by means of multiplying respectively the ex-
pectation value hQi and hPi by a unit operator.
Let us introduce deviation operators Qd and Pd as follows:
1.7 Non-commuting operators and the uncertainty principle 23

Qd = Q − hQiI, (1.42)
Pd = P − hPiI, (1.43)
and denote
φ1 = Qd ψ, (∆q)2 = (φ1 , φ1 ), (1.44)

φ2 = Pd ψ, (∆p)2 = (φ2 , φ2 ). (1.45)

Apply the Schwarz inequality for φ1 and φ2 , i.e.
(φ1 , φ1 )(φ2 , φ2 ) > |(φ1 , φ2 )|2 . (1.46)
Thus we have
(∆q)2 (∆p)2 > |(Qd ψ, Pd ψ)|2 = |(ψ, Qd Pd ψ)|2 . (1.47)
Let us introduce an anticommutator of Qd and Pd , defined as
{Qd , Pd } = Qd Pd + Pd Qd . (1.48)
It can be readily shown that the expectation value of an anticommu-
tator with respect to the state ψ is always real, say α = (ψ, {Qd , Pd }ψ).
Furthermore since Qd Pd = 12 [Qd , Pd ] + 12 {Qd , Pd }, and its expectation
value is calculated as

1 1 i 1
(ψ, Qd Pd ψ) = (ψ, [Qd , Pd ]ψ) + (ψ, {Qd , Pd }) = ~ + α. (1.49)
2 2 2 2

Hence Heisenberg’s uncertainty relation in the one dimensional

case is derived, i.e.

~2 α2 ~2 1
(∆q)2 (∆p)2 > + > , or ∆q∆p > ~. (1.50)
4 4 4 2

The minimum uncertainty relation holds if the following conditions

are fulfilled:

(a) Pd ψ = λQd ψ, namely Pd ψ and Qd ψ are linearly dependent.

(b) α = 0.
24 Chapter 1

The coefficient λ can be evaluated by taking the expectation value of

the anticommutator 21 {Qd , Pd } with respect to the state ψ, and putting
it equal to zero. We reach

1
λ(∆q)2 + (∆p)2 = 0.
λ

On the other hand, the expectation value of the commutator

~ 1 ~
[Qd , Pd ] = leads to λ(∆q)2 − (∆p)2 = , hence
i λ i

i~
λ= . (1.51)
2(∆q)2

An explicit expression of the state function in terms of the least

uncertainty ∆q shall be postponed until we cover the materials on q-
representation in quantum mechanics.

1.8 Exercises

Ex 1.8.1

Consider the following set of continuous functions fn (x) = xn , x ∈

[−1, 1] which spans a L2 (−1, 1)-space. Find explicitly the first 3 or-
thonormal functions by the Gram-Schmidt process. What are those
functions that occur to your mind?

Ex 1.8.2

Prove that the Minkowski inequality holds in Hilbert space, i.e.

kψ + φk 6 kψk + kφk.

(Hint: take the square of either side.)

1.8 Exercises 25

Ex 1.8.3
Prove the law of parallelogram holds in Hilbert space, i.e.
kψ + φk2 + kψ − φk2 = 2(kψk2 + kφk2 ).

Ex 1.8.4
Prove that every finite dimensional vector space is complete. (Hint:
since the real and complex numbers are complete.)

Ex 1.8.5
Show that the kABk 6 kAkkBk if A, B ∈ bounded operator.

Ex 1.8.6
Show that the expectation value of any dynamical observable in the
physical system is always real.

Ex 1.8.7
Prove that the adjoint conjugate can equivalently be defined as
(Aψ, φ) = (ψ, A† φ) if (A† ψ, φ) = (ψ, Aφ).

Ex 1.8.8
Let {φ1 , φ2 , . . . , φn } be a set of an orthonormal basis. Prove that
operator U is unitary if {Uφ1 , Uφ2 , . . . , Uφn } is also a set of an or-
thonormal basis.

Ex 1.8.9
Prove that if A and B are two operators that both commute with
their commutator, then
eA eB = eA+B+[A,B]/2 .

df (λ)
(Hint: let f (λ) = eλA eλB e−λ(A+B) , and obtain dλ = λ[A, B]f (λ)
by Eq. (1.39). Then integrate it.)
26 Chapter 1

Ex 1.8.10
Consider a Hilbert space spanned by a Hermitian operator A.

Y
(a) Prove that (A − aI) is a null operator if Aψa = aψa .
a

Y A − aI
(b) What is the significance of the following operator, ?
0
a0 − a
a6=a

Ex 1.8.11
If two observables A1 and A2 are not compatible, but their corre-
sponding operators both commute with the Hamiltonian operator H,
i.e.
[A1 , H] = [A2 , H] = 0.
Show that the energy eigenstates are in general degenerate.

Ex 1.8.12
Consider a one-dimensional Hamiltonian

1 2
H= P + V (Q)
2m

and use the fact that the commutator of Q and [Q, H] is a constant
operator, to show that

X ~2
(Ek − Es )|Qsk |2 = ,
2m
k

which is referred to as the Thomas-Reiche-Kuhn sum rule, where Qsk =

(ψs , Qψk ) and ψs is the eigenstate of H with eigenvalue Es , i.e. Hψs =
Es ψs .
Chapter 2

Space-Time Translation, Quantum Dynamics

and Various Representations in Quantum
Mechanics

2.1 Vector space and dual vector space

We start from C n -space as an example. In matrix notation, a vector x̄

in C n -space can be represented by a vertical array of n complex numbers
as follows

 
x1
x2
 
 
  ..
.
 
 
 
x̄ =  xi 

, (2.1)
xi+1 
 
 . 
 . 
 . 
xn

which is called a column matrix, where xi is the ith-component of the

vector. Each component is labeled by a subscript i. Matrix algebra
ensures that the two basic operations of vectors, as we explained at
the beginning of the first chapter, can be reproduced. The orthonormal
basis of the n-dimensional vector space e¯i is also represented by a column

24:22.
28 Chapter 2

matrix. Take the ith-basis as an example:

 
0
0
 
 
 .. 
.
 
 
 
e¯i = 
 1 ,
 (2.2)
0
 
 
 .. 
.
 
 
0

that allows one to express the vector x̄ as the following linear combina-
tion

n
X
x̄ = xi e¯i . (2.3)
i=1

To perform the inner product of an ordered pair of vectors in C n , one

needs to introduce another vector space ∗ C n , a dual space to C n -space.
A vector in ∗ C n is denoted by a row matrix as follows
 
x̃ = x∗1 x∗2 . . . x∗i x∗i+1 . . . x∗n ,

which is exactly the adjoint conjugate, also called the Hermitian con-
jugate of vector x̄ in C n -space, namely by taking the complex conjugate
of each entry and transposing the column matrix into a row matrix in
∗ C n -space. One justifies immediately that ∗ C n is also a vector space.

Let us see how to perform an inner product by means of matrix

algebra. Take the ordered pair of vectors x̄ and ȳ as an example. We
construct a dual vector x̃, which is an adjoint conjugate of the first
vector x̄. Then the inner product of vectors x̄ and ȳ is obtained from
the multiplication of the row matrix x̃ in ∗ C n -space by the column matrix
ȳ in C n -space, i.e.

24:22.
2.1 Vector space and dual vector space 29

 
y1
 y2 

 
x̃ȳ = x∗1 x∗2 ∗
. . . xn  . 

. (2.4)
 .. 
yn

The linear operator A in C n -space is represented by a square matrix

with matrix elements Aij for i, j = 1, 2, . . . , n, obtained by sandwiching
the operator A between the basis row vector e˜i and the basis column
vector e¯j , i.e.

e˜i Ae¯j = Aij . (2.5)

If we take the adjoint conjugate of the above equation, we have

e˜j A† e¯i = (A† )ji (2.6)

on the left hand side of Eq. (2.5), while on the right hand side of
Eq. (2.5), we have merely a complex number, because its adjoint conju-
gate is just the complex conjugate. Hence we have the following relation

(A† )ji = A∗ij . (2.7)

A matrix A which equals to its adjoint conjugate matrix, namely

A = A† is called a Hermitian matrix. Then we have

Aij = (A† )ij = A∗ji , (2.8)

or A∗ij = Aji , by taking the complex conjugate on both sides.

The expectation value of an operator A with respect to the vector
x̄ is taken by multiplying the row vector x̃, the n by n matrix A, and
the column vector x̄ successively as follows

x̃Ax̄ = expetation value of A with respect to x̄.

If the operator stands for a physical observable, then it is a Hermitian

matrix and the expectation value is real because

x̃Ax̄ = x̃A† x̄ = (x̃Ax̄)† = (x̃Ax̄)∗ . (2.9)

24:22.
30 Chapter 2

The matrix algebra also ensures that n × n square matrices follow

the rules of being linear operators.
Matrix representation for C n -space allows us to construct the pro-
jection operator readily. A projection operator that projects a vector
x̄ onto a subspace spanned by basis e¯i is a square matrix formed by
putting a column basis before a row basis, i.e.

Pi = e¯i e˜i , (2.10)

or more specifically,

   
0 0 ... 0 0 ... 00
 .
0 . . .. 
0
 
   .
  ..  .. . .. 
 
. . .
 . .
  
  
Pi =  1  0 0 . . . 1 0 . . . 0 = 0 .
 
   1 0 
 0 
 
 . 

 0 0 0 

 .  . . .. 
 .  . . . .
.


0 0 0 ... 0 0 ... 0

The projection operator PMPonto a subspace M simply takes the

sum of the projection operators Pi , i.e.
X
PM = Pi for ei ∈ M. (2.11)

n
X
It is obvious that the sum of the projection operators Pi over
the basis in the entire space is an identity operator, i.e. i=1

n
X
Pi = I. (2.12)
i=1

It is advantageous in this matrix representation that it allows one

to easily construct any projection operator. If one wants to project a

24:22.
2.1 Vector space and dual vector space 31

vector onto a particular vector ȳ, the corresponding projection operator

can then be expressed as

Py = ȳ ỹ. (2.13)

The Hilbert space corresponding to a physical system is often infinite

dimensional and also not necessarily separable. It becomes awkward or
even impossible to use the matrix formulation in those cases.
To overcome the inconvenience and the difficulty, Dirac introduced
new notations for the vectors in Hilbert space. A vector in H is denoted
by the symbol |αi, called a ket vector, or simply a ket. While a vector
in the adjoint conjugate dual space ∗ H is denoted by the symbol hα|,
called a bra vector, or simply a bra. Similar to the case in C n -space,
to construct the inner product of an ordered pair of vectors in Dirac
notation, say the vectors |αi and |βi, we put the adjoint conjugate of
|αi, namely the bra vector hα| in front of the ket vector |βi as follows

hα||βi ≡ hα|βi, (2.14)

which reaffirms the definition of the adjoint conjugates of operator A.

The square norm of a vector |αi is defined as the inner product of a
pair of same vectors, i.e.

k|αik2 = hα|αi. (2.15)

A linear operator A which transforms a ket vector |αi into another

ket vector |α0 i is written as

|α0 i = A|αi. (2.16)

The inner product of the ordered pair of vectors of |αi and |β 0 i is

written as follows

hα|β 0 i = hα| (A|βi) ≡ hα|A|βi. (2.17)

Generally we omit writing the parenthesis as it appears in the second

term of last equation.
On the other hand, if the inner product is constructed for an ordered
pair of vectors |α0 i and |βi, in which the vector |α0 i is a transformed
vector of |αi by operator A, then the corresponding inner product is
written as

24:22.
32 Chapter 2

hα0 |βi = hα|A† |βi. (2.18)

Similar to the case in C n -space, we conclude that

hα|A† |βi = hβ|α0 i∗ = hβ|A|αi∗ , (2.19)

and the definition of adjoint conjugate of A is regained.

Dirac notations in vector space have the advantage of constructing
the projection operator readily. Consider a Hermitian operator A, and
the eigenvector |ai with the corresponding eigenvalue a, i.e.

A|ai = a|ai. (2.20)

The set of eigenvectors {|a1 i, |a2 i, . . . , |ai i, . . .} forms the basis of the
Hilbert space H.
The projection operator onto the subspace |αi is designed by putting
a ket vector and a bra vector together in successive order, i.e.

Pα = |αihα|. (2.21)

Any vector |αi in the Hilbert space is expressed in a linear combi-

nation of its basis as follows
X
|αi = αj |aj i, (2.22)
j

where the coefficient αj is obtained through the inner product, i.e.

X X
hai |αi = hai | αj |aj i = αj δji = αi . (2.23)
j j

Projection of the vector onto the subspace spanned by the basis |ai i
is then read as

Pi = |ai ihai |. (2.24)

A projection operator onto subspace M can also be constructed by

X X
PM = Pi = |ai ihai |, |ai i ∈ M. (2.25)
i i

24:22.
2.2 q-representation and p-representation in quantum mechanics 33

By extending subspace M to cover the whole Hilbert space, we reach

the closure relation,
X X
Pi = |ai ihai | = I. (2.26)

In the case of a separable Hilbert space with degeneracy, and denot-

ing the basis by |a, ii, the closure relation reads as
X X
Pa,i = |a, iiha, i| = I. (2.27)
a,i a,i

For the case that the eigenvalues become continuous, the eigenvec-
tors |ξi are no longer denumerable. The basis vector is characterized
by the eigenvalue ξ, which is a continuum and the inner product is
normalized to be a delta function, i.e.

hξ|ξ 0 i = δ(ξ − ξ 0 ), (2.28)

and the closure relation in this case takes the form as

Z Z
Pξ dξ = |ξidξhξ| = I. (2.29)

2.2 q-representation and p-representation in quantum

mechanics

Consider the Hilbert space spanned by the eigenvectors of the Hermi-

tian operator X corresponding to a dynamical observable in coordinate
x. For simplicity, we limit ourselves to the one-dimensional case. The
eigenvalue equation reads as

X|xi = x|xi. (2.30)

If we make a transport of the physical system to the right by a

distance ξ, namely the spatial translation of the system from x to x + ξ,
it is interesting to observe the change of the eigenvector in the new
system.

24:22.
34 Chapter 2

Let us first introduce a unitary operator, called the translational

operator, defined as
i
U(P; ξ) = e− ~ ξP , (2.31)

where
ξ = real parameter, namely the distance of the spatial translation,
P = momentum operator,

and U(P; ξ) is unitary because

i
U† (P; ξ) = e ~ ξP = U−1 (P; ξ).

Proposition 1.
U(P; ξ)|xi is an eigenvector of operator X with eigenvalue x + ξ.

To prove the above statement, let us start with the commutation

relation of X and P, i.e. [X, P] = i~I, that enables us to evaluate the
following commutators
[X, Pn ] = Pn−1 [X, P] + [X, Pn−1 ]P = in~Pn−1 , (2.32)

and
" #
X 1  i m
[X, U] = i~ X, − ξ Pm
m
m! ~

i m−1 m−1
 
X 1
=ξ − ξ P = ξU, (2.33)
m
(m − 1)! ~

or
XU = U(X + ξI). (2.34)

Construct the new state

24:22.
2.2 q-representation and p-representation in quantum mechanics 35

|αi = U|xi.

Since

X|αi = XU|xi = U(X + ξI)|xi = U(x + ξ)|xi = (x + ξ)|αi, (2.35)

it implies that |αi is also the eigenvector of operator X, with eigenvalue

x + ξ. Therefore we cast vector |αi as

|αi = c|x + ξi = |x + ξi.

The constant c is just a pure phase factor that is chosen to be a unit

without losing its generality.

Proposition 2.
A state function or wave function as it is generally called, is
defined as the inner product of the vector |xi and the state vector
|ψi of interest, i.e. ψ(x) = hx|ψi.

Furthermore, sandwiching the momentum operator in between

bra vector hx| and state vector |ψi results in differentiating the
wave function, or more precisely,
~ ∂
hx|P|ψi = ψ(x).
i ∂x

We consider an infinitesimal translation ξ, then

i
U(P; ξ) = 1 − ξP,
~

and

24:22.
36 Chapter 2

 
i
hx|U(P; ξ)|x i = hx|x + ξi = hx| 1 − ξP |x0 i,
0 0
~

i
δ(x − x0 ) − δ(x − x0 − ξ) = ξhx|P|x0 i,
~

~ d ~
hx|P|x0 i = δ(x − x0 ) = δ 0 (x − x0 ),
i dx i

where δ 0 (x − x0 ) is the derivative of the delta function with respect to

its argument, i.e.

d d
δ 0 (x − x0 ) = δ(x − x0 ) = δ(x − x0 ),
dx d(x − x0 )

and it has the following property:

Z
f (x)δ 0 (x − a)dx = −f 0 (a) (2.36)

for any function f (x). With this property, we are able to derive that

Z Z
0 0 0 ~ 0
hx|P|ψi = hx|P|x idx hx |ψi = δ (x − x0 )ψ(x0 )dx0
i
Z
~ d ~ d
=− 0
δ(x0 − x)ψ(x0 )dx0 = ψ(x). (2.37)
i dx i dx

We are now in the position to make a comparison between the ab-

stract state and the corresponding q-representation.

24:22.
2.2 q-representation and p-representation in quantum mechanics 37

Hilbert space q-representation

|ψi hx|ψi = ψ(x)
X|ψi hx|X|ψi = xψ(x)
~ d
P|ψi hx|P|ψi = ψ(x)
i dx
~ d
F (P, X)|ψi F( , x)ψ(x)
i dx

ψ(x) in the above table is called the wave function or the state
function, and the absolute square of ψ(x), i.e. |ψ(x)|2 is commonly
referred to as the probability density of the system.
We particularly consider the eigenvalue equation for the energy of
the system. The Hamiltonian operator reads

1
H= P + V (X), (2.38)
2m

and the eigenvalue equation

H|ψi = E|ψi, (2.39)

when it is expressed in terms of q-representation, becomes

hx|H|ψi = Ehx|ψi, (2.40)

or explicitly, the time independent Schrödinger equation as

~2 d2
 
− + V (x) ψ(x) = Eψ(x). (2.41)
2m dx2

T h e p -r ep r es entation ofth e s tate f u n ction can be f or mu lated p ar allel

to th at of th e q -r ep r es entation w ith th e h elp ofhte f ollow in g p r op os ition .

24:22.
38 Chapter 2

Proposition 3.
Q-representation of the eigenstate of the momentum operator is
a plane wave, namely
1 i
hx|pi = √ e ~ px
2π~
where |pi is the eigenstate of P, i.e,with the eigenvalue p, i.e.
with P|pi = p|pi.

|x0 idx0 hx0 |

R
Let us consider hx|P|pi and insert the identity operator
after the momentum operator P. We have
Z Z
~ 0 ~ d
hx|P|pi = hx|P|x0 idx0 hx0 |pi = δ (x − x0 )hx0 |pidx0 = hx|pi,
i i dx

or

~ d
phx|pi = hx|pi.
i dx

Solving this differential equation, we obtain

i 1 i
hx|pi = Ae ~ px = √ e ~ px . (2.42)
2π~

The normalization constant can be verified as follows,

Z
0 0
hx|x i = δ(x − x ) = hx|pidphp|x0 i

Z
i 0
= |A| 2
e ~ p(x−x ) dp = 2π~|A|2 δ(x − x0 ).

That implies |A|2 = 1/2π~.

24:22.
2.2 q-representation and p-representation in quantum mechanics 39

Similar to the case in the q-representation, the wave function in

p-space reads as

hp|ψi = ψ(p). (2.43)

In fact ψ(p) is the inverse Fourier transform of ψ(x), i.e.

Z Z
1 i 0
ψ(p) = hp|ψi = hp|x0 idx0 hx0 |ψi = √ e ~ px dx0 ψ(x0 ). (2.44)
2π~

The Schrödinger equation in the p-representation can be written as

1 2
hp|H|ψi = hp| P + V (X)|ψi. (2.45)
2m

The first term is trivially obtained,

1 2 1 2
hp| P |ψi = p ψ(p) (2.46)
2m 2m

while the second term can be evaluated by the closure relations,

Z
hp|V (X)|ψi = hp|V (X)|p0 idp0 hp0 |ψi,

where

ZZ
0
hp|V (X)|p i = hp|x0 idx0 hx0 |V (X)|x00 idx00 hx00 |p0 i

ZZ
1 i 0 00 −px0 )
= e ~ (p x V (x0 )δ(x0 − x00 )dx0 dx00
2π~
Z
1 i 0
= e ~ (p −p)x V (x)dx = Ṽ (p − p0 ), (2.47)
2π~

24:22.
40 Chapter 2

is the Fourier transform of the potential in p-space. Finally we reach

the Schrödinger equation in the p-representation as follows

p2
Z
ψ(p) + Ṽ (p − p0 )ψ(p0 )dp0 = Eψ(p). (2.48)
2m

Before leaving this section, let us go back to the q-representation

of the state with the minimum quantum uncertainty ∆p∆q = ~/2 dis-
cussed in Section 1.7, where |ψi satisfies the following condition

i~
Pd |ψi = Qd |ψi. (2.49)
2(∆q)2

The q-representation of the above equation reads as

 
~ ∂ i~
− hpi ψ(x) = (x − hxi)ψ(x), (2.50)
i ∂x 2(∆x)2

which allows us to obtain the normalized wave function of minimum

uncertainty as

(x − hxi)2 ihpix
 
1
ψ(x) = exp + . (2.51)
[2π(∆x)2 ]1/4 4(∆x)2 ~

2.3 Harmonic oscillator revisited

There exists very rich structure in dealing with the harmonic oscil-
lation mathematically on a quantum level. We shall attack the problem
by means of the number operator through the creation operator and
the annihilation operator rather than solving the Schrödinger equa-
tion in the q-representation, in which the quantization of the energy
levels is derived from the boundary conditions of the wave functions.

24:22.
2.3 Harmonic oscillator revisited 41

The technique developed in this section can even be applied to the sys-
tem of linear oscillators in the limit of continuum, which gives rise to
the quantization of the field.

2.3.1 Creation and annihilation operators

The Hamiltonian operator for the one-dimensional harmonic oscil-
lator reads as

1 2 1
H= P + mω 2 X2 . (2.52)
2m 2

Let us introduce the new variables defined as follows:

r
1 mω
pξ = √ P, ξ= X. (2.53)
mω~ ~

The Hamiltonian operator takes simpler form,

 
1 2 1 2
H = ~ω p + ξ = ~ωH, (2.54)
2 ξ 2

with H defined as H = (p2ξ + ξ 2 )/2.

The nonvanishing canonical quantization relation of P and X, i.e.
[P, X] = ~I/i can also be simplified as

[pξ , ξ] = −iI. (2.55)

Define the annihilation operator and the creation operator as follows

1 1
a = √ (ξ + ipξ ), and a† = √ (ξ − ipξ ), (2.56)
2 2

or reversely

24:22.
42 Chapter 2

1
ξ = √ (a† + a), (2.57)
2

i
pξ = √ (a† − a). (2.58)
2

The commutation relations become:

[a, a] = [a† , a† ] = O, [a, a† ] = I. (2.59)

Then H can be cast into H = (a† a + aa† )/2 = N + I/2 where

N = a† a, a Hermitian operator called the number operator, which
is treated as a new dynamical observable.
Consider the Hilbert space spanned by the eigenvectors of the num-
ber operator. The properties concerning the eigenvalues as well as the
eigenvectors are summarized in the following propositions.

Proposition 4.
If ν and |νi are respectively the eigenvalue and eigenvector of
number operator N , then ν is positive definite.

For the eigenvalue equation of operator N ,

N |νi = ν|νi, (2.60)

where the eigenvector is labeled by the eigenvalue and denoted it by |νi.

Let us define the vector |αi = a|νi. The positive definite of the norm
of the vector |αi requires that

hα|αi = hν|a† a|νi = νhν|νi > 0.

Therefore ν > 0, namely ν is positive definite.

24:22.
2.3 Harmonic oscillator revisited 43

Proposition 5.
The state with zero eigenvalue i.e. |0i, then |0i will be annihi-
lated into the null vector by a, the annihilation operator, namely
a|0i = 0 .

Let |αi = a|0i, and the norm square of |αi reads as

hα|αi = h0|a† a|0i = h0|N |0i = 0h0|0i = 0, (2.61)

which implies that a|0i is a null vector.

Proposition 6.
a† |νi is an eigenvector of N with eigenvalue ν + 1.

This can be proved with the commutation relation

[N , a† ] = a† , or N a† = a† (N + I). (2.62)

For |νi 6= 0 and letting |αi = a† |νi, then

N |αi = N a† |νi = a† (N + I)|νi = a† (ν + 1)|νi = (ν + 1)|αi. (2.63)

Proposition 7.
For |νi 6= 0 , then a|νi is the eigenvector of N with eigenvalue
ν − 1.

Since [N , a] = −a, or N a = a(N − I), and letting |αi = a|νi, then

N |αi = N a|νi = a(N − I)|νi = a(ν − 1)|νi = (ν − 1)|αi. (2.64)

24:22.
44 Chapter 2

Proposition 8.
The eigenvalues of the number operator N take only non-negative
integers, i.e. ν = n = 0, 1, 2, . . . .

The proof can be facilitated with Proposition 7. If ν 6= integer,

then vectors a|νi, a2 |νi, a3 |νi, . . . , as |νi, are all eigenvectors of N re-
spectively with the eigenvalues ν − 1, ν − 2, . . . , ν − s. Let |αi = a|ν − si,
we calculate the square norm of |αi, i.e.

hα|αi = (ν − s)hν − s|ν − si.

This does not guarantee that hα|αi is positive definite if s is chosen

larger than ν. Positive definite of the norms of all eigenvectors generated
by means of applying as on |νi can only be achieved if ν takes non-
negative integers, namely ν = n = 0, 1, 2, . . ..
Therefore the eigenvectors of the number operator N , as well as
that of the new Hamiltonian are written as

|0i, |1i, |2i, . . . , |ni, . . .

with the normalization hn|n0 i = δnn0 .

The eigenvector with zero eigenvalue, i.e. |0i is called the ground
state of N or H. The eigenvector |1i is called the one particle state for
operator N , or the first excited state of the Hamiltonian operator H
and so forth. This is the reason that a† is called the creation operator
or the promotion operator, while a is called the annihilation operator
or the demotion operator. a† acts on state |ni to produce the state
c|n + 1i, i.e.

a† |ni = c|n + 1i,

so that the action of a† will increase the n-particle system into the
(n+1)-particle system or will promote the n-th excited state
√ into (n+1)-
th excited state. The constant c is determined to be n + 1 when all
eigenstates are orthonormalized, i.e.
√
a† |ni = n + 1|n + 1i. (2.65)

24:22.
2.3 Harmonic oscillator revisited 45

Similarly operator a acts upon state |ni and it becomes the state
c0 |n − 1i, i.e.

a|ni = c0 |n − 1i.

The action of a decreases the n-particle system into the (n − 1)-

particle system or demotes the n-th excited state into the (n − 1)-th
√
excited state. Again the constant c0 can be evaluated to be n, i.e.

√
a|ni = n|n − 1i. (2.66)

With the help of the creation operator, one is able to construct the
orthonormal basis of the Hilbert space spanned by the eigenvectors of
the number operator N . We start from the ground state or zero particle
states and build the higher excited states or many particle state by
acting the creation operator upon it successively. Then we have

|0i,

|1i = a† |0i,

1
|2i = √ (a† )2 |0i,
2!

..
.

1
|ni = √ (a† )n |0i. (2.67)
n!

Conversely, we are also able to construct the lower excited states by

acting the annihilation operator upon the n-particle states successively,
i.e.

24:22.
46 Chapter 2

|0i = a|1i,

1
|1i = √ a|2i,
2

1
|2i = √ a|3i,
3

..
.

1
|ni = √ a|n + 1i. (2.68)
n+1

2.4 N-representation and the Rodrigues formula of

Hermite polynomials

It is interesting to investigate how the eigenstate |ni is to be realized

in the q-representation. Let us consider the eigenvalue equation of the
operator ξ, i.e.

ξ|ξi = ξ|ξi. (2.69)

One can also easily verify that

d
hξ|pξ |ξ 0 i = −i δ(ξ − ξ 0 ),
dξ

with the normalization hξ|ξ 0 i = δ(ξ − ξ 0 ).

The inner product of |ξi and |ni defines the q-representation of the n-
particle state, namely the n-particle wave function or the wave function
of the n-th excited state, which reads as

ψn (ξ) = hξ|ni.

24:22.
2.4 N-representation and the Rodrigues formula of Hermite polynomials 47

1
Since a† = √ (ξ − ipξ ), the wave function ψn (ξ) becomes
2
1 n
 
1 † n 1
ψn (ξ) = hξ|ni = √ hξ|(a ) |0i = √ √ hξ|(ξ − ipξ )n |0i.
n! n! 2

By Proposition 2, we have

d n d n
   
n
hξ|(ξ − ipξ ) |0i = ξ − hξ|0i = ξ − ψ0 (ξ).
dξ dξ

To obtain the 0-particle wave function ψ0 (ξ), we take the q-representa-

tion of a|0i = 0 , i.e.
hξ|a|0i = 0,
namely that

1
hξ| √ (ξ + ipξ )|0i = 0,
2

or
 
d
ξ+ ψ0 (ξ) = 0. (2.70)
dξ

Therefore we can solve the equation and obtain the normalized wave
function as

1 − 1 ξ2
ψ0 (ξ) = √
4
e 2 , (2.71)
π

and the wave function ψn (ξ) becomes

n
d n − 1 ξ2
  
1 1 1
ψn (ξ) = √ √ √
4
ξ− e 2 . (2.72)
n! 2 π dξ

24:22.
48 Chapter 2

Applying the following identity

d 1 2 d 1 2
ξ− ≡ (−1)e 2 ξ e− 2 ξ ,
dξ dξ

we conclude that

d n
   n
n 21 ξ 2 d 1 2
ξ− = (−1) e e− 2 ξ .
dξ dξ

Then

n 1 2
(−1)n e− 2 ξ

1 2 d −ξ 2
ψn (ξ) = p √ e2ξ e =p √ Hn (ξ), (2.73)
2n n! π dξ 2n n! π

and the Rodrigues formula of Hermite polynomials is identified as

 n
n ξ2 d 2
Hn (ξ) = (−1) e e−ξ . (2.74)
dξ

2.5 Two dimensional harmonic oscillation and direct

product of vector spaces

Let us consider the system of a two-dimensional isotropic harmonic

oscillator whose Hamiltonian takes the following expression

1 m
H= (P2 + P2y ) + ω(X2 + Y2 ). (2.75)
2m x 2

Similar to the case of one-dimensional harmonic oscillation, we define

the creation operators and the annihilation operators as follows:

24:22.
2.5 Two-dimensional harmonic oscillation and direct product of vector spaces 49
1
a† = √ (ξ − ipξ ), (2.76a)
2

1
a = √ (ξ + ipξ ), (2.76b)
2

1
b† = √ (η − ipη ), (2.76c)
2

1
b = √ (η + ipη ), (2.76d)
2

where ξ, η, pξ and pη are defined as

r
mω 1
ξ= X, pξ = √ Px , (2.77a)
~ m~ω
r
mω 1
η= Y, pη = √ Py . (2.77b)
~ m~ω

The fundamental commutation relations can be simplified as

[ξ, η] = [ξ, pη ] = O,
[pξ , η] = [pξ , pη ] = O,
[pξ , ξ] = [pη , η] = −iI,

which lead to the following commutation relations among the annihila-

tion operators and the creation operators, namely
[a, b] = [a, b† ] = [a† , b] = [a† , b† ] = O, (2.78)

[a, a† ] = [b, b† ] = I. (2.79)

The Hamiltonian can therefore be written as
   
† 1 † 1
H = a a + I ~ω + b b + I ~ω,
2 2

24:22.
50 Chapter 2

or H = Ha + Hb if we put

   
† 1 1
Ha = a a + I ~ω = Na + I ~ω,
2 2
   
† 1 1
Hb = b b + I ~ω = Nb + I ~ω.
2 2

To solve the Schrödinger equation of the two dimensional harmonic

oscillation, we shall make use of Proposition 8 in Section 1.6.2. Eval-
uating the commutator of the operator Na and the operator Nb , we
find that only the first and the second fundamental commutation rela-
tions are required to show that [Na , Nb ] = O. Therefore we recognize
that Na and Nb are, in fact, the intrinsic compatible observables. The
eigenvectors |ψi in the following Schrödinger equation

    
1 1
Na + I ~ω + Nb + I ~ω |ψi = E|ψi, (2.80)
2 2

can be expressed as the direct product of the two eigenvectors in one-

dimensional case, i.e.

|ψi = |na i ⊗ |nb i ≡ |na ; nb i, (2.81)

with the corresponding eigenvalues given as

E = (na + nb + 1)~ω = (n + 1)~ω. (2.82)

The energy of the two-dimensional harmonic oscillator is also quan-

tized to be (n + 1)~ω, with degeneracy if na + nb = n > 0.
We are now in the position to discuss the algebra of the direct vector
space. A direct product space of the Hilbert spaces Ha and Hb is also
a Hilbert space, denoted as follows

H = Ha ⊗ H b . (2.83)

24:22.
2.5 Two-dimensional harmonic oscillation and direct product of vector spaces 51

Let Ha be the Hilbert space spanned by |ai i, the eigenvectors of the

dynamical operator A, then the basis of Ha takes the following set of
vectors,
|a1 i, |a2 i, . . . |ai i, . . .
Similarly, if the Hilbert space Hb is spanned by another set of basis,
denoted by
|b1 i, |b2 i, . . . |bj i, . . .
which are the eigenvectors of another dynamical operator B. The prod-
uct space H = Ha ⊗ Hb is then spanned by the set of direct product of
a base vector |ai i in Ha and another base vector |bj i in Hb , denoted by
|ai i ⊗ |bj i ≡ |ai ; bj i. (2.84)
This very direct product vector |ai ; bj i is also the eigenvector of
any linear operator in H if the operator is constructed in the form of
a function of two operators A and B, denoted by F(A, B), and the
eigenvalue equation for the operator F(A, B) reads as follows
F(A, B)|ai ; bj i = F (ai , bj )|ai ; bj i, (2.85)
where F (ai , bj ) is the eigenvalue of F(A, B).
The dimension of the product space H equals to the product of
the dimensions of the two individual spaces Ha and Hb , namely d =
da db with da , db and d standing for the dimensions of Ha ,Hb and H
respectively.
The linear operator in H also takes the form of direct product of two
linear operators, denoted by R and S, where R is a linear operator in the
space Ha and S is a linear operator in the space Hb . The direct product
operator is commonly denoted by R ⊗ S, whose algebraic operational
rule reads as
R ⊗ S|ai ; bj i = R|ai i ⊗ S|bj i. (2.86)
We summarize the properties of the direct product operators as fol-
lows:

24:22.
52 Chapter 2

R ⊗ O = O ⊗ S = O, (2.87a)
I ⊗ I = I, R ⊗ I = R, I ⊗ S = S, (2.87b)
R ⊗ (S1 + S2 ) = R ⊗ S1 + R ⊗ S2 , (2.87c)
(R1 + R2 ) ⊗ S = R1 ⊗ S + R2 ⊗ S, (2.87d)
αR ⊗ βS = αβR ⊗ S, (2.87e)
−1 −1 −1
(R ⊗ S) =R ⊗S , (2.87f)
R1 R2 ⊗ S1 S2 = (R1 ⊗ S1 )(R2 ⊗ S2 ), (2.87g)

where α and β are scalars.

To demonstrate the difference between the compatible observables
and the intrinsic compatible ones, we take the motion of a charged par-
ticle in the uniform magnetic field as an example. Consider a uniform
~ y) can be ex-
magnetic field of H0 in z-axis, the vector potential A(x,
pressed as

~ y) = 1 H0 (−y î + xĵ).
A(x,
2

The principle of minimal interaction allows us to cast the Hamiltonian

of a particle, with the mass m and the charge q moving in the uniform
field H0 , into the following form

1  q 2
H= P− A
2m c
( 2  2 )
1 1q 1q
= Px + H0 Y + Py − H0 X . (2.88)
2m 2c 2c

The sum of the two square terms in the Hamiltonian of last equation
ensures that the total energy of the system is always positive definite.
Namely when we solve the following Schrödinger equation

H|ψi = E|ψi, (2.89)

we shall obtain the positive definite eigenvalue E.

24:22.
2.5 Two-dimensional harmonic oscillation and direct product of vector spaces 53

By expanding the quadratic terms of the above Hamiltonian, we

obtain the following expression, i.e.

1 2 m  ωc 2 2 1 2 m  ωc 2 2 1
H= Px + X + P + Y − ωc Lz , (2.90)
2m 2 2 2m y 2 2 2

where ωc is the cyclotron frequency and Lz stands for the 3rd com-
ponent of the angular momentum operator which can be calculated
explicitly as follows:

qH0
ωc = , Lz = XPy − YPx .
mc

One recognizes immediately that combining the first two terms of

Eq. (2.90) is , in fact, the Hamiltonian of a two dimensional harmonic

m  ωc 2
oscillator with the force constant k = , i.e.
2 2
(2) 1 m  ωc 2 2
Hh.o. = (P2x + P2y ) + (X + Y2 ). (2.91)
2m 2 2

(2)
It can be easily verified that Hh.o. and Lz commute with each other
i.e.
(2)
[Hh.o. , Lz ] = O. (2.92)
We will run into a dilemma if we apply the method of Proposition
8 to calculate the eigenvalues of the following Hamiltonian

(2) 1
H = Hh.o. − ωc Lz , (2.93)
2

which is the sum of two compatible observables. The eigenvalue equa-

tions for these two observables read as

24:22.
54 Chapter 2

(2) 1
Hh.o. |n1 ; n2 i = (n1 + n2 + 1)~ωc |n1 ; n2 i, (2.94)
2

Lz |mi = m~|mi, (2.95)

where m takes the integers, while n1 and n2 take only the positive ones.
Of course we have made use of the result that the angular momentum Lz
is quantized as it is expressed in Eq. (2.95), which we shall investigate
in the next chapter.
Should we try to evaluate the total energy of the system by applying
(2)
H = Hh.o. −ωc Lz /2 to act upon the direct product vector |n1 ; n2 i⊗|mi,
we would end up in the following eigenvalue equation

 
(2) 1
Hh.o. − ωc Lz |n1 ; n2 i ⊗ |mi = E|n1 ; n2 i ⊗ |mi, (2.96)
2

with E = (n1 + n2 + 1)~ωc /2 − m~ωc /2, which is no longer a positive

definite quantity if m > (n1 + n2 + 1). Namely the total energy of
the system E obtained in this way does not guarantee to be positive
definite, which contradicts our previous assertion.

(2)
The contradiction arises from the fact that the observable Hh.o. and
the observable Lz are not intrinsic compatible even though they com-
mute with each other. One needs, in fact, all the three fundamental
(2)
commutation relations in order to prove that the commutator of Hh.o.
and Lz vanishes, i.e.

(2)
[Hh.o. , Lz ] = O. (2.97)

The dilemma can be resolved by introducing another set of linear

operators in terms of the original four dynamical observables with the
following relations:

24:22.
2.5 Two-dimensional harmonic oscillation and direct product of vector spaces 55

r r 
1 mωc 2
c= √ (X + iY) + i (Px − iPy ) , (2.98a)
2 2~ m~ωc
r r 
† 1 mωc 2
c =√ (X − iY) − i (Px + iPy ) , (2.98b)
2 2~ m~ωc
r r 
1 mωc 2
d= √ (X − iY) + i (Px + iPy ) , (2.98c)
2 2~ m~ωc
r r 
† 1 mωc 2
d =√ (X + iY) − i (Px − iPy ) . (2.98d)
2 2~ m~ωc

It can be verified that the commutation relations among c, c† , d and

d† become as follows:

[c, d] = [c, d† ] = O, (2.99a)

[c† , d] = [c† , d† ] = O, (2.99b)

[c† , c] = [d† , d] = I. (2.99c)

If we construct two Hermitian number operators defined as Nc = c† c

and Nd = d† d. It is obvious that Nc and Nd become the intrinsic com-
patible observables because that [Nc , Nd ] = O is derived only based
upon the first two sets of the commutation relations, i.e. Eqs. (2.99a)
and (2.99b). Therefore we can construct the direct product vector
|nc ; nd i = |nc i ⊗ |nd i, where |nc i and |nd i stand for the eigenvectors of
the operators Nc and Nd respectively. Using the property that |nc ; nd i
is automatically the eigenvector of the of the operator F(Nc , Nd ) with
the eigenvalue F (nc , nd ), we are able to calculate the eigenvalue E of
the Hamiltonian operator H in the following equation

 
(2) 1
Hh.o. − ωc Lz |nc ; nd i = E|nc ; nd i. (2.100)
2

24:22.
56 Chapter 2

(2)
After some algebra, the two terms Hh.o. and Lz in the Hamiltonian
above can be expressed in terms of the new dynamical operators Nc
and Nd as follows:

(2) 1
Hh.o. = (Nc + Nd + 1)~ωc , (2.101)
2

1 1
− ωc Lz = (Nc − Nd )~ωc , (2.102)
2 2

which allow us to take F(Nc , Nd ) as follow

 
(2) 1 1
F(Nc , Nd ) = Hh.o. − ωc Lz = Nc + ~ωc , (2.103)
2 2

and the eigenvalue E is evaluated as

 
1
E = F (nc , nd ) = nc + ~ωc , (2.104)
2

which is positive definite as we have expected at the beginning of our

discussion on the system of a charged particle moving in the uniform
magnetic field.

2.6 Elastic medium and the quantization of scalar field

Consider the linear chain of N identical particles of mass m, with

the springs of force constant α attached to both sides of the nearest
mass points. Each mass point is also attached with another spring of
force constant β to the equilibrium positions which are of equal spacing
along the chain, as shown in Figure 2.1. The Hamiltonian of the system
reads as

N N N
1 X 2 αX 2 βX
H= pr + (qr − qr+1 ) + qr qr , (2.105)
2m 2 2
r=1 r=1 r=1

24:22.
2.6 Elastic medium and the quantization of scalar field 57

where qr is the displacement of the r-th mass point away from its equi-
librium position, and pr is the corresponding canonical conjugate mo-
mentum. The commutation relations are expressed as

[qr , qr0 ] = [pr , pr0 ] = O, [qr , pr0 ] = i~δrr0 I.

The Hamiltonian can be simplified tremendously if we introduce

the normal coordinates denoted by Qs through the following Fourier
expansion

r
~ X 2πi rs
qr = e N Qs . (2.106)
Nm s

Here we assume the total number of mass points to be N , and

(
takes from s = − 21 (N − 1) to s = 12 (N − 1),
s=
they are integers if N is odd, and half odd integer if N is even.

The condition that qr is real will lead us to take Q−s = Q†s . Before
we try to decouple the mixing terms in the Hamiltonian, we establish
the following formula:

Figure 2.1: Linear chain of identical particles

24:22.
58 Chapter 2

Proposition 9.
For r = integers, and s − s0 = integers, then
N
1 X 2πi r(s−s0 )
eN = δss0 .
N
r=1

2πi
(s−s0 )
Denoting z = e N , we have

1 X r 1 z(1 − z N )
F (z) = z = (z + z 2 + . . . + z N ) = , (2.107)
N r N N (1 − z)

which implies that for the case s − s0 = 0,

F (z) = F (1) = 1.

For the case that s − s0 = non-zero integers, say s − s0 = m then

z = exp(2πiN/m), which is one of the roots for the equation z N −1 = 0.
Therefore

1
F (z) = (1 + z + . . . + z N −1 ) = 0,
N

and one reaches the conclusion

F (z) = δss0 . (2.108)

This proposition allows us to invert qr into Qs . We leave it as an

exercise to invert Eq. (2.106) and obtain that

r
m X − 2πi rs
Qs = e N qr . (2.109)
N~ r

Similarly the Fourier expansion for the momentum is expressed as

24:22.
2.6 Elastic medium and the quantization of scalar field 59

r
m~ X 2πi rs
pr = e N Ps , (2.110)
N s

with

P−s = Ps† ,

and

1 X 2πi
Ps = √ e− N rs pr . (2.111)
mN ~ r

It also leaves you as an exercise to show that each term in the Hamil-
tonian operator can be calculated as follows,

1 X 2 ~X
p = Ps Ps† , (2.112)
2m r r 2 s

 
αX α~ X 2πs
(qr − qr+1 )2 = 4 sin2 Qs Q†s , (2.113)
2 r 2m s N

βX β~ X
qr qr = Qs Q†s . (2.114)
2 r 2m s

Therefore the Hamiltonian takes the following expression

~X
H= [Ps Ps† + ω 2 (s)Qs Q†s ], (2.115)
2

where ω 2 (s) is defined as

 
2 4α 2πs β
ω (s) = sin2 + , (2.116)
m N m

24:22.
60 Chapter 2

and ω(s) is an even function in s, i.e.

ω(−s) = ω(s).

It can be verified that the commutation relations for the operators

expressed in terms of normal coordinate and momentum become

[Qs , Qs0 ] = [Ps , Ps0 ] = O, [Qs , Ps0 ] = iδss0 I. (2.117)

Let us investigate the system in the Heisenberg picture in which the

operators are regarded as time dependent. The equations of motion
for Qs , Q†s , Ps and Ps† take the identical expressions as the canonical
equations of motion in classical mechanics, namely,

∂H
= Q̇†s , or Ps = Q̇†s ,
∂Ps†

∂H
= −Ṗs† , or ω 2 (s)Qs = −Ṗs† .
∂Q†s

Hence we have

Q̈s + ω 2 (s)Qs = 0, (2.118)

and the solution is cast into

1  
Qs = p as e−iω(s)t + a†−s eiω(s)t , (2.119)
2ω(s)

where as / 2ω(s) and a†−s / 2ω(s), playing the roles of the constants
p p

of integration, are chosen such as to meet the condition Qs = Q†−s .

Similarly we have the solution for Ps as

r
ω(s)  
Ps = i −as e−iω(s)t + a†−s eiω(s)t . (2.120)
2

24:22.
2.6 Elastic medium and the quantization of scalar field 61

Let us define bs = as e−iω(s)t and b†−s = a†−s eiω(s)t . Then the normal
coordinate and the normal momentum become as follows

1
Qs = p (bs + b†−s ),
2ω(s)
r
ω(s)
Ps = i (−bs + b†−s ),
2

which enables us to write the commutation relations among the opera-

tors as

[bs , bs0 ] = [b†s , b†s0 ] = O, [bs , b†s0 ] = δss0 I.

The Hamiltonian operator can then be expressed as

~X X 1
H= ω(s)(bs b†s + b†s bs ) = ~ ω(s)(b†s bs + I)
2 s s
2

X 1
=~ ω(s)(Ns + I), (2.121)
s
2

where Ns = b†s bs is the boson number operator of s-th mode.

Let us investigate the system of an elastic medium in the limit of
continuum. Take the length of the linear chain be l with N mass points
equally spaced when they are at equilibrium. Then the position for the
r-th particle is given as

l
xr = r . (2.122)
N

We also define the wave number

2π
ks = s, (2.123)
l

24:22.
62 Chapter 2

and rescale the displacement away from the r-th equilibrium position
at the time t defined as

r r r
mN mN ~ X 2πi rs
Φ(xr , t) = qr = e N Qs
l l mN s

r
~ X 2πi rs 1  
= eN p as e−iω(s)t + a†−s eiω(s)t
l s 2ω(s)
r
~X 1  
= p as ei(ks xr −ω(s)t) + a†s e−i(ks xr −ω(s)t) ,
l s 2ω(s)
(2.124)

and the corresponding momentum as

r r
~ X ω(s)  
Π(xr , t) = i −as ei(ks xr −ω(s)t) + a†s e−i(ks xr −ω(s)t) ,
l s 2
(2.125)

where the last term in Eqs. (2.124) and (2.125) are obtained by making
use of ks = −k−s and ω(s) = ω(−s).
The commutation relation between Φ(xr , t) and Π(xr0 , t) can be
evaluated as

~
[Φ(xr , t), Π(xr0 , t)] = N δrr0 I.
l

Let us take the limit as N → ∞, the positions xr = rl/N become

continuous, and the subscript r can be removed from xr , namely that

l
x = lim xr = lim r . (2.126)
N →∞ N →∞ N

24:22.
2.6 Elastic medium and the quantization of scalar field 63

We also simplify the wave number ks into k, therefore

Φ(x, t) = lim Φ(xr , t)

N →∞

r
~X 1  
= p ak ei(kx−ω(k)t) + a†k e−i(kx−ω(k)t) , (2.127)
l 2ω(k)
k

Π(x, t) = lim Π(xr , t)

N →∞

r r
~X ω(k)  
= i −ak ei(kx−ω(k)t) + a†k e−i(kx−ω(k)t) ,
l 2
k
(2.128)

and the commutation relation between Φ(xr , t) and Π(xr , t) reduces to

[Φ(x, t), Π(x0 , t)] = lim [Φ(xr , t), Π(xr0 , t)]

N →∞

N
= i~ lim δrr0 I = i~δ(x − x0 )I. (2.129)
N →∞ l

It is on the quantum level that the continuous elastic medium is

treated as a space being filled by the quantized scalar field or boson
field.
The linear system can be generalized into the three-dimensional case.
The field and its conjugate momentum take the expressions as

r
~X 1  
Φ(r, t) = p ak ei(kr−ω(k)t) + a†k e−i(kr−ω(k)t) , (2.130)
l 2ω(k)
k

r r
~X ω(k)  
Π(r, t) = i −ak ei(kr−ω(k)t) + a†k e−i(kr−ω(k)t) ,
l 2
k
(2.131)

24:22.
64 Chapter 2

and the commutation rules read as follows:

[Φ(r, t), Φ(r0 , t)] = [Π(r, t), Π(r0 , t)] = O, (2.132)

0 0
[Φ(r, t), Π(r , t)] = i~δ(r − r )I. (2.133)

2.7 Time evolution operator and the postulate of quantum

dynamics

The universe evolves in time and so does the physical system. Any
quantum state in the quantum mechanical system always changes in
time, just like the motion of the particles in classical mechanics changes
its configuration according to laws of motion. All the physical states as
well as the dynamical observables we have dealt with in the first three
postulates of quantum mechanics were expressed in such a way that
time is totally absent from the equations. In fact, the physical quantities
we had met or dealt with are time dependent implicitly, namely that
we set our observation of the system at a convenient time scale, say
t = 0. Choosing the time scale is entirely arbitrary . We would like to
express the physical state at the particular time t = t0 to perform the
observation or to make the measurement. Then the state will be denoted
by |ψ; t0 i. All the contents in our previous discussion do not change and
all the theories and conclusions we have derived so far remain the same
if we replace |ψi by |ψ; t0 i.

2.7.1 Time evolution operator and Schrödinger equation

What is more interesting and challenging for us is how the state
|ψ; t0 i will change from the moment t0 to any future time t. Let us
denote a time evolution operator by U(t, t0 ). It is a linear operator
whose action upon a physical state |ψ; t0 i will evolve the state into a new
state |ψ; ti governed by the law of quantum dynamics. Mathematically
it is expressed by

|ψ; ti = U(t, t0 )|ψ; t0 i. (2.134)

Conservation of the norm square at any time, or conservation of the

probability when they are expressed in the q-representation, requires
that the time evolution operator be unitary, namely,

24:22.
2.7 Time evolution operator and the postulate of quantum dynamics 65

hψ; t|ψ; ti = hψ; t0 |U† (t, t0 )U(t, t0 )|ψ; t0 i = hψ; t0 |ψ; t0 i (2.135)

or

Z Z Z
2
hψ; t|xidxhx|ψ, ti = |ψ(x; t)| dx = |ψ(x; t0 )|2 dx.

Therefore

U† (t, t0 )U(t, t0 ) = I or U† (t, t0 ) = U−1 (t, t0 ). (2.136)

Let us investigate a state at time t, which evolves into state at time

t + δt, an infinitesimal time interval δt after t, i.e.

|ψ; t + δti = U(t + δt, t)|ψ; ti. (2.137)

Since U(t, t) = I, the time evolution operator U(t + δt, t), of course
is very close to identity operator, can be approximated up to first order
in δt by the following expression

U(t + δt, t) = I − iΩδt, (2.138)

where the operator Ω is Hermitian because

U−1 (t + δt, t) = I + iΩ† δt = I + iΩδt,

or Ω† = Ω.
It is the operator Ω that dictates the change of the state, and this
forms the last postulate of quantum mechanics, namely the law of quan-
tum dynamics, in which Ω is taken to be H(P, Q; t)/~.

24:22.
66 Chapter 2

4th Postulate of Quantum Mechanics:

The evolution of the quantum state is governed by the following
dynamical law,
 
i
|ψ; t + δti = U(t, t + δt)|ψ; ti = I − H(P, Q; t)δt |ψ; ti
~
(2.139)
or to rewrite it in the familiar expression of the Schrödinger equa-
tion,
1 ∂
i~ lim (|ψ; t + δti − |ψ; ti) = i~ |ψ; ti = H(P, Q; t)|ψ; ti.
δt→0 δt ∂t
(2.140)

In other words, the total energy of the system is the generator that
drives the quantum state into evolution. Therefore the Schrödinger
equation is referred to as the law of quantum dynamics.

2.7.2 Time order product and construction of time

evolution operator
The product of two time evolution operators is again a time evolution
operator, given by

U(t, t0 ) = U(t, t0 )U(t0 , t0 ). (2.141)

Hence we have

 
i
U(t + δt, t0 ) = U(t + δt, t)U(t, t0 ) = I − H(P, Q; t) δt. (2.142)
~

From which the differential equation for U(t, t0 ) is obtained as follows

∂U(t, t0 ) 1 i
= lim (U(t + δt, t0 ) − U(t, t0 )) = − H(P, Q; t)U(t, t0 ).
∂t δt→0 δt ~
(2.143)

24:22.
2.7 Time evolution operator and the postulate of quantum dynamics 67

The solution for U(t, t0 ) is trivial for the conserved system in which the
Hamiltonian operator does not depend upon time explicitly. U(t, t0 )
then takes the following expression
i
U(t, t0 ) = e− ~ H(P,Q)(t−t0 ) . (2.144)

If the Hamiltonian operator becomes time dependent, we can solve

the evolution operator by iteration, i.e.

 Z t
i
U(t, t0 ) = I + − H(P, Q; t0 )U(t0 , t0 )dt0 ,
~ t0

or with the abbreviation H(P, Q; t) = H(t), then

Z t0
i 2 t 0
 Z t   Z
i 0 0 0
U(t, t0 ) = I+ − dt H(t )+ − dt H(t ) dt00 H(t00 )+. . . .
~ t0 ~ t0 t0

Let us define the time order product as follows

(
A(t1 )B(t2 ) if t1 > t2 ,
T (A(t1 )B(t2 )) =
B(t2 )A(t1 ) if t2 > t1 ,

then

Z t Z t0 ZZ t
1
dt0 H(t0 ) dt00 H(t00 ) = dt0 dt00 T (H(t0 )H(t00 )),
t0 t0 2! t0

and U(t, t0 ) can be cast into more compact expression as

24:22.
68 Chapter 2

i 2
 Z t   ZZ t
i 0 0 1
U(t, t0 ) = I + − dt H(t ) + − dt0 dt00 T (H(t0 )H(t00 ))
~ t0 2! ~ t0

 3 Z Z Z t
1 i
+ − dt0 dt00 dt000 T (H(t0 )H(t00 )H(t000 )) + . . .
3! ~ t0

n i Rt o
− dtH(P,Q;t)
=T e ~ t0 . (2.145)

2.8 Schrödinger picture vs. Heisenberg picture

Measurement of the physical quantities differs from time to time in

the dynamical system. The expectation values of the dynamical ob-
servables in a quantum system correspond to the data of the measure-
ments, whose values depend upon the time at which the measurements
are performed. Therefore, when we take the expectation value of a dy-
namical observable A with respect to the state |ψ; ti, i.e. the value
hAi = hψ; t|A|ψ; ti that we obtain, in fact, is time dependent. From
now on we shall denote the implicitly time dependent expectation value
of the dynamical operator A with respect to the state |ψ; ti by
hAit = hψ; t|A|ψ; ti. (2.146)
Taking the time derivative of the above equation, we have the following
relation

d d d
hAit = hψ; t|A|ψ; ti = hψ; t0 |U† (t, t0 )AU(t, t0 )|ψ; t0 i
dt dt dt

i i
= hψ; t0 |U† (HA − AH)U|ψ; t0 i = hψ; t|[H, A]|ψ; ti.
~ ~
(2.147)

If we take A to be the position operator Q, the equation becomes

24:22.
2.8 Schrödinger picture vs. Heisenberg picture 69

d i
hQit = hψ; t|[H, Q]|ψ; ti. (2.148)
dt ~

The commutator of H and Q can be evaluated according to the

fundamental commutation relations among coordinate and momentum,
and we obtain

~ ∂H
[H, Q] = ,
i ∂P

hence

 
d ∂H
hQit = . (2.149)
dt ∂P t

Similarly in the case for the momentum operator P,

 
d ∂H
hPit = − . (2.150)
dt ∂Q t

It is in this Schrödinger picture that treating the quantum state

as time dependent while keeping the dynamical observables as fixed
quantum operators, Hamilton’s equations of motion are regained as in
the following table:

Quantum equations of motion classical equations of motion

in Schrödinger picture
 
d ∂H ∂Hc
hQit = , q̇c (t) = ,
dt ∂P t ∂pc (t)
 
d ∂H ∂Hc
hPit = − , ṗc (t) = − ,
dt ∂Q t ∂qc (t)

24:22.
70 Chapter 2

where Hc = H(pc (t), qc (t)).

There also exists another picture to look at the quantum equations of
motion, namely the Heisenberg picture. Instead of regarding the state
vector as time dependent, one takes the new dynamical observables
by the unitary transformation of the old ones, namely taking the new
dynamical operator as

Ã(t) = U† (t, 0)AU(t, 0) = U† (t)AU(t), (2.151)

here for simplicity and clarity, we take t0 = 0 and put U(t, t0 ) = U(t).
Since the commutation relations are invariant under the unitary
transformation, the fundamental quantization rules remain the same
as before, provided that all new operators are of at equal time, i.e.

~
[P̃(t), Q̃(t)] = I, (2.152)
i

which allows us to establish the following commutation relations

~ ∂F
[Q̃(t), F (P̃(t), Q̃(t))] = − , (2.153)
i ∂ P̃(t)

~ ∂G
[P̃(t), G(P̃(t), Q̃(t))] = . (2.154)
i ∂ Q̃(t)

Then the quantum equations of motion take the following expression

∂ Q̃(t) i ∂H
= U† (t)[H, Q]U(t) = , (2.155)
∂t ~ ∂ P̃(t)

∂ P̃(t) i ∂H
= U† (t)[H, P]U(t) = − , (2.156)
∂t ~ ∂ Q̃(t)

and we have the following table:

24:22.
2.9 Propagator in quantum mechanics 71

Quantum equations of motion classical equations of motion

in Heisenberg picture

∂ Q̃(t) ∂H ∂Hc
= , q̇c (t) = ,
∂t ∂ P̃(t) ∂pc (t)
∂ P̃(t) ∂H ∂Hc
=− , ṗc (t) = − .
∂t ∂ Q̃(t) ∂qc (t)

2.9 Propagator in quantum mechanics

Let us consider a one-dimensional system in which the Hamiltonian

does not depend on time explicitly. The time evolution operator there-
fore takes the simple form,
i
U(t, t0 ) = e− ~ H(P,Q)(t−t0 ) = U (t − t0 ).

The wave function in the Schrödinger picture can be expressed as

ψ(x; t) = hx|ψ; ti = hx|U(t − t0 )|ψ; t0 i
Z Z
− ~i H(t−t0 ) 0 0 0 0
= hx|e |x idx hx |ψ; t i = K(x, t; x0 , t0 )ψ(x0 , t0 )dx0 ,

and the equality of the last equation defines the propagator

i 0
K(x, t; x0 , t0 ) = hx|e− ~ H(t−t ) |x0 i. (2.157)

It is in fact the propagator that is taken as the matrix element of

the time evolution operator U(t, t0 ) between the states |xi and |x0 i. The
propagator brings the state function at spacetime point (x0 , t0 ) to the
state function at another spacetime point (x, t). It is obvious that when
t = t0 , the propagator reduces to the delta function, i.e.
i
K(x, t; x0 , t) = hx|e− ~ H(t−t) |x0 i = hx|x0 i = δ(x − x0 ). (2.158)

We shall explore the property of the propagator by putting it in a

different form. Let us insert an identity projection operator formed by

24:22.
72 Chapter 2

energy eigenstates in front of state |x0 i, then we express the propagator

in the following form
X i 0
X i 0
K(x, t; x0 , t0 ) = hx|e− ~ H(t−t ) |αihα|x0 i = hx|e− ~ Eα (t−t ) |αihα|x0 i.
α α
(2.159)

The time derivative of the propagator can be performed as

∂ X i 0
i~ K(x, t; x0 , t0 ) = Eα hx|αie− ~ Eα (t−t ) hα|x0 i
∂t α

 
X ~ ∂ i 0
= H , x hx|αie− ~ Eα (t−t ) hα|x0 i
α
i ∂x

 
~ ∂
=H , x K(x, t; x0 , t0 ). (2.160)
i ∂x

If we introduce the step function

(
0 1, for t > t0 ,
η(t − t ) =
0, for t < t0 ,

and make use of the property

d
η(t − t0 ) = δ(t − t0 ),
dt

then the time derivative for the product of η(t − t0 ) and K(x, t; z 0 , t0 )
can be calculated as
∂
i~ (η(t − t0 )K(x, t; x0 , t)) = η(t − t0 )HK + i~δ(t − t0 )K
∂t
= η(t − t0 )HK + i~δ(t − t0 )δ(x − x0 ),
(2.161)

24:22.
2.10 Newtonian mechanics regained in the classical limit 73

where the last term is obtained by the property that δ(x − a)f (x) =
δ(x − a)f (a). Hence

 
∂
i~ − H η(t − t0 )K(x, t; x0 , t0 ) = i~δ(t − t0 )δ(x − x0 ). (2.162)
∂t

Therefore the propagator is also referred to as Green’s function or kernel.

A simple demonstration is in order to evaluate the propagator for a
free particle system in which we take |αi = |pi, and summation over α
is replaced by integration over the momentum p, then

Z
i p2 0
0 0
K(x, t; x , t ) = hx|pie− ~ 2m (t−t ) hp|x0 idp

Z
1 i 0 i p2 0
= e ~ p(x−x ) e− ~ 2m (t−t ) dp
2π~

i m(x − x0 )2
r  
m
= exp .
2πi~(t − t0 ) 2~ (t − t0 )

This formula reminds us the familiar form of the delta function

r
(x − a)2
 
1
δ(x − a) = lim exp − , (2.163)
→0 π 

and lim0 K(x, t; x0 , t0 ) = δ(x − x0 ) is reproduced.

t→t

2.10 Newtonian mechanics regained in the classical limit

Modern physics is treated usually as a new branch of physics. New

theories are developed to analyze the new discoveries and phenomena
in physical systems, particularly in the microscopic systems of atoms
or sub-atoms. Quantum theory rose after the end of the 19th century

24:22.
74 Chapter 2

an d is com m on ly b elieved to b e a cor rect th eor y f or inves tigatin g th e

phenomena that were discovered in microscopic quantum systems in the
modern era. It is a new theory that replaced Newtonian theory which
was successful and widely accepted in the studies of classical dynamics.
Yet if we carefully examine the propagator we have treated in the
last section, and take the very same theory in the limit that ~ tends
to zero, we are able to recover the whole of the theories developed in
classical dynamics.
In order to treat the spatial coordinates and the temporal ones on
equal footing, we introduce a new notation as follows
i
|x, ti = e ~ Ht |xi. (2.164)

The time dependent state function and the Schrödinger equation we had
before can be re-expressed respectively as
i
ψ(x; t) = hx, t|ψi = hx|e− ~ Ht |ψi, (2.165)

∂ ∂ i
i~ ψ(x; t) = i~ hx|e− ~ Ht |ψi = Hhx, t|ψi = Hψ(x; t). (2.166)
∂t ∂t

Furthermore, the propagator from the initial spacetime point (x0 , t0 ) to

final spacetime point (x, t) takes the following inner product,
i
hx, t|x0 , t0 i = K(x, t; x0 , t0 ) = hx|e− ~ H(t−t0 ) |x0 i, (2.167)

which in fact is the transition probability for a quantum object at space-

time (x0 , t0 ) to another spacetime (x, t). Since the closure relation holds
at any instant, we have,

Z
|x0 , t0 idx0 hx0 , t0 | = I,

that can be inserted in between the state hx, t| and the state |x0 , t0 i at
the instant t0 . If we consider the case that the transition takes place in
N steps from the initial spacetime (x0 , t0 ) to the final spacetime (x, t),
and let

24:22.
2.10 Newtonian mechanics regained in the classical limit 75

t − t0 = N , tn = t0 + n, where n = 1, 2, . . . , N − 1,

with xN = x and tN = t as shown in Figure 2.2, then the propagator

can be expressed as in the following form
Z Y
hx, t|x0 , t0 i = dxi hx, t|xN −1 , tN −1 i . . . hxn , tn |xn−1 , tn−1 i . . .
i

× hx1 , t1 |x0 , t0 i.
(2.168)

A typical term hxn , tn |xn−1 , tn−1 i in the above formula shall be ex-
plored in detail as follows
i i 1 2 +V
hxn , tn |xn−1 , tn−1 i = hxn |e− ~ H |xn−1 i = hxn |e− ~ ( 2m )P (X))
|xn−1 i
Z
i 2 i
= hxn |e− 2m~ P  |ξidξhξ|e− ~ V (X) |xn−1 i. (2.169)

We have made the approximation in the last equality by neglecting the

i
term higher than order 2 . The term hxn | exp − 2m~ P2  |ξi can be

Figure 2.2: Spacetime diagram.

24:22.
76 Chapter 2

calculated explicitly by Fourier transform as in the case of K(x, t; x0 , t0 )

for a free particle system and we obtain,
Z
i 2 i 2
hxn |e− 2m~ P  |ξi = hxn |e− 2m~ P  |pidphp|ξi

m(xn − ξ)2
Z r  
1 i
− 2m~ p2  i
p(xn −ξ) m
= e e ~ dp = exp i .
2π~ 2πi~ 2~

The typical term hxn ; tn |xn−1 , tn−1 i can finally be cast into the ex-
pression,
( "  #)
i m xn − xn−1 2
r 
m
hxn ; tn |xn−1 , tn−1 i = exp  − V (xn−1 )
2πi~ ~ 2 

r  
m i
= exp Ln , (2.170)
2πi~ ~

where the discrete form of Lagrangian Ln takes the following form

 2
m xn − xn−1
Ln = − V (xn−1 ),
2 

and the propagator of Eq. (2.167) is eventually evaluated as

 m N/2 Z Y  
i
hx, t|x0 , t0 i = dxn exp Ln .
2πi~ ~

Instead of performing the multiple integration for all steps, we let

N tend to infinity, and the summation over the discrete Lagrangian in
the exponent becomes integration over time. This new way of perform-
ing the integration has a special name called path integration, after
Feynman. Therefore we express

24:22.
2.10 Newtonian mechanics regained in the classical limit 77

 m N/2 Z Y  
i
hx, t|x0 , t0 i = lim dxn exp Ln
→0 2πi~ ~
Z  Z t 
i
= Dx(τ ) exp L(x(τ ), ẋ(τ ))dτ , (2.171)
~ t0

where  and xn (or xn−1 ) are replaced by dτ and x(τ ) respectively, and

   
m xn − xn−1
L(x(τ ), ẋ(τ )) = lim Ln = lim − V (xn−1 ) .
→0 →0 2 

In the classical limit when ~ → 0, the path integration over the

ex pon ential f u n ction wou ld os cillate s o r ap id ly an d th e rpop agator ,
the transition probability, would become negligibly small. If we regard
the initial spacetime as the location of a particle at the beginning time
t0 , the particle ends up at the final spacetime, the final position x at
time t, along the path determined by laws of classical dynamics, then we
shall get the maximum transition probability. This can only be achieved
if the phase in the exponential appearing in the integrant of the path
integration is most stable, namely the variation of the phase be zero.
Mathematically it is expressed by

Z t
δS = δ L(x(τ ), ẋ(τ ))dτ, (2.172)
t0

which is exactly the principle of least action in classical mechanics that

leads to Lagrange’s equations of motion, namely

 
d ∂L ∂L
− = 0, (2.173)
dτ ∂ ẋ(τ ) ∂x(τ )

and classical mechanics is regained.

24:22.
78 Chapter 2

2.11 Exercises

Ex 2.11.1
Show that delta function δ(x) can be expressed as

1 sin N x
(a) δ(x) = lim or
π N →∞ x

1 d2
(b) δ(x) = |x|.
2 dx2

Ex 2.11.2
Show that

 2
~0
2
hx|P |x i = δ 00 (x − x0 ),
i

and hence that

 
0 ~ d
hx|F (P)|x i = F δ(x − x0 ).
i dx

Ex 2.11.3
Consider a physical system of one dimension that is translated by a
distance ξ to the right, the wave function then becomes

hx|U(P; ξ)|ψi.

Evaluate the translated wave function explicitly and interpret the

result. (Hint: take adjoint conjugate of U† (P; ξ)|xi as hx|U(P; ξ).)

24:22.
2.11 Exercises 79

Ex 2.11.4

Show that the p-representation of the Schrödinger equation can also

be expressed as

p2
 
~ d
ψ(p) + V − ψ(p) = Eψ(p).
2m i dp

Ex 2.11.5

Solve the Schrödinger equation for the potential energy

V (x) = −gδ(x − a).

with g positive definite, in q-representation and p-representation.

Ex 2.11.6

Find the one dimensional quantum state of minimum uncertainty in

the p-representation.

Ex 2.11.7

Find the quantum uncertainty ∆ξ∆pξ in the n particle state.

Ex 2.11.8

2
Show that h0|eikx |0i = exp(− k2 h0|X2 |0i), where |0i is the ground
state of the one dimensional harmonic oscillator, and X is the position
operator.

Ex 2.11.9

Prove that

24:22.
80 Chapter 2

1 X ~X
pr pr = Ps Ps† ,
2m 2 s

βX β~ X
qr qr = Qs Q†s ,
2 2m s

if we define

r
~ X 2πi rs
pr = e N Ps ,
Nm s

r
m~ X 2πi rs
qr = e N Qs .
N s

(Hint: by means of Proposition 9.)

Ex 2.11.10

Prove that

 
αX α~ X 2πs
(qr − qr+1 )2 = 4 sin2 Qs Q†s .
2 r 2m s N

(Hint: also by Proposition 9.)

Ex 2.11.11

By using Mehler’s formula

" #
1 x2 + y 2 − 2xyξ 2 +y 2 )
X ξn
exp − = e−(x Hn (x)Hn (y),
2n n!
p p
1 − ξ2 1 − ξ2 n

24:22.
2.11 Exercises 81

prove that the propagator in the case of the one dimensional harmonic
oscillator can be expressed as (Hint: use Eq. (2.159) for |αi = |ni.),
r
mω
K(x, t; x0 , t0 ) =
2πi~ sin ω(t − t0 )
  
imω 2
× exp [(x + x20 ) cos ω(t − t0 ) − 2xx0 ] .
2~ sin ω(t − t0 )

Ex 2.11.12
For a dynamical observable A(t) which is time dependent explicitly,
the time derivative of the expectation value takes the form as

d D ∂A E
i~ hA(t)it = h[A, H]it + i~ .
dt ∂t t

Use the above formula repeatedly and show that

hpi2
 
2 1
h∆xi2t = h∆xi20 + hxp + pxi0 − hxi0 hpi0 t + 20 t2 ,
m 2 m

for the case of a free particle in one-dimension.

Ex 2.11.13
Prove that the one dimensional Schrödinger equation in free space

∂ ~2 ∂ 2
i~ ψ(x, t) = − ψ(x, t)
∂t 2m ∂x2

is invariant under Galilean transformation:

x0 = x − v0 t,
t0 = t.

24:22.
82 Chapter 2

(Hint: let ψ 0 (x0 , t0 ) = f (x, t; v0 )ψ(x, t).)

Ex 2.11.14
Derive explicitly the propagator in the p-representation, i.e.

Z  Z t 
− ~i H(t−t0 ) 0 i
hx|e |x i = DxDp exp L̄(τ )dτ ,
~ t0

where

Y  dxn dpn  Z t
DxDp = lim , L̄(τ )dτ = lim L̄n ,
→0
n
2π~ t0 →0

with
L̄n = pn ẋn − H(xn , pn ),

1
ẋn = (xn − xn−1 ).


24:22.
Chapter 3

Symmetry, Transformation and Continuous

Groups

3.1 Symmetry and transformation

Nature shows symmetries in physics, not only in the geometrical

sense; it also contains a wide range of deeper implications among the
invariance and the conservation laws in physics. Let us take space as an
example. If we assume the homogeneity in spatial property, a translation
along any direction in this space will not change the description of the
system. The physical quantities involved will be invariant under the
spatial transformation, or briefly, it is said to be translational invariant,
and hence the total momentum of the system is conserved. So as to the
isotropy property of the space, the total angular momentum is conserved
due to the rotational symmetry.
Transformations in physics often form a group which is character-
ized by a set of continuous parameters, called the group parameters.
Henceforth we call it the continuous group. We shall start with some
simple examples in the following section.

3.1.1 Groups and group parameters

Let us consider the translation in n-dimensional space. A vector x
in Rn -space is transformed into a vector x0 by a translation specified
with n continuous parameters denoted by the contravariant components
ξ = (ξ 1 , ξ 2 , . . . , ξ n ). We shall write the transformation as

25:23.
84 Chapter 3

x0 = x + ξ or x0i = xi + ξ i for i = 1, 2, . . . , n. (3.1)

Further translation from the vector x0 to the vector x00 by another dis-
placement η which is characterized by another set of n continuous pa-
rameters η = (η 1 , η 2 , . . . , η n ) is written as
x00 = x0 + η or x00i = x0i + η i = xi + (ξ i + η i ). (3.2)

It is obvious that all the translations form a group, called the n-

dimensional translational group. If we denote the group elements
simply by ξ and η for the previous two consecutive translations, and put
a dot · in between ξ and η, then the group operation for the consecutive
translation is expressed as
γ = η · ξ,

and the composition rules for the group parameters are just the addition,
i.e.
γ = (γ 1 , γ 2 , . . . , γ n ), γ i = ξi + ηi. (3.3)

Hence the transformation of the space translation satisfies the group

postulates.

(a) γ is again a translation with the group element specified by the

set of n continuous parameters, i.e. x00 = x + γ, γ i = ξ i + η i .

(b) There exists the identity element e = (0, 0, . . . , 0) such that e · ξ =

ξ · e = ξ.

(c) The inverse element of ξ, denoted by ξ −1 exists, such that

ξ −1 · ξ = ξ · ξ −1 = e,

that implies

ξ −1 = (−ξ 1 , −ξ 2 , . . . , −ξ n ). (3.4)

Instead of performing the displacement upon the vectors in Rn -

space, if we consider the transformation of the function F (x) by making
the displacement in the coordinates system, then the expression for

25:23.
3.1 Symmetry and transformation 85

the group elements will be drastically changed. When the function

F (x) = F (x1 , x2 , . . . , xn ) is transformed by a displacement into F (x0 ) =
F (x + ξ) = F (x1 + ξ 1 , x2 + ξ 2 , . . . , xn + ξ n ), we can expand the function
in terms of a Taylor series and put it in a more compact expression by
summing over the repeated indices as
F (x0 ) = F (x + ξ)

∂ 1 i j ∂ ∂ ξi ∂ i
= F (x) + ξ i F (X) + ξ ξ F (x) + . . . = e ∂x F (x).
∂xi 2! ∂xi ∂xj
(3.5)

One notices that the group elements for translation take a completely
different form.

The group element of the displacement is written as
exp ξ i ∂/∂xi instead of ξ. Therefore the group operation is the usual
multiplication, namely the successive displacements
of ξ and η is written
as the product of two group elements exp η i ∂/∂xi and exp ξ i ∂/∂xi ,

and reaching to another element as

       
i ∂ i ∂ i i ∂ i ∂
exp η · exp ξ = exp (ξ + η ) i = exp γ ,
∂xi ∂xi ∂x ∂xi

and the addition of the group parameters γ i = ξ i + η i is regained in

the exponent but in totally different context. Namely the fact that the
partial derivatives with respect to two different coordinates commute
with each other ensure us to have γ i = ξ i + η i .
The parameters in the translational group form an n-dimensional
Euclidean space itself. This Rn -space is called the group parameter
space, or group manifold in short.
Let us consider the linear transformation on n-dimensional vector
space as another example. Vector x when expressed by a column matrix,
is changed into vector x0 by an n × n matrix A, i.e.
x0 = Ax,
or if we express it in terms of the matrix element as
x0i = Aij xj = (δji + aij )xj , where Aij = δji + aij .

25:23.
86 Chapter 3

The linear transformations form a group provided that all the trans-
formation matrices are of nonvanishing determinants. Consider two suc-
cessive transformations and x0 = Ax, x00 = Bx0 , then we conclude:

(a) x00 = Bx0 = BAx = Cx and C = BA with property det C 6= 0

because det A 6= 0 and det B 6= 0.

(b) The identity matrix exists by choosing aij = 0.

(c) Since det A 6= 0, of course A−1 exists.

The group element A of this linear transformation consists of n2

parameters, namely aij for i, j = 1, 2, . . . , n. Therefore its group param-
2
eter space or the group manifold is a Rn -space, a Euclidean space of
n2 -dimension, and this group is called the GL(n, R) group.
The composition rules for the group parameters in GL(n, R) are
more complicated than in the case of spatial translation. The group
parameters of the group element C = BA obtained from the successive
transformations A and B take the following composition rules

cij = (δli + bil )(δjl + alj ), (3.6)

or

cij = δji + bij + aij + bil alj . (3.7)

Different groups dictate different sets of the composition rules for

the group parameters. We shall consider the general case of r-parameter
group and denote the element simply by the set of r parameters. Let
us denote the group elements by

a, b for a = (a1 , a2 , . . . , ar ), b = (b1 , b2 , . . . , br ).

For the product of a and b, i.e. c = ba = f (ai , bj ) = f (a, b), we can

express the group parameters of the resultant element cl = f l (ai , bj )
with the composition functions f l (ai , bj ) being infinitively differentiable
with respect to ai and bj .
Denote e as the identity element and identify its position at the
origin in the parameter space, i.e. e = (0, 0, . . . , 0), then we have

25:23.
3.2 Lie groups and Lie algebras 87

f (a, e) = f (e, a) = a. (3.8)

Also if we denote the inverse of element a by a−1 , then the following

relations hold
f (a, a−1 ) = f (a−1 , a) = e or f l (a, a−1 ) = f l (a−1 , a) = 0. (3.9)

Continuous groups were systematically investigated by Lie who par-

ticularly concentrated on the behavior of function f (a, b) when both ele-
ments a and b are at the vicinity of the identity transformation, namely
putting elements a and b close to identity element e, and looking into
the functional behavior of f l (ai , bj ) for infinitesimal of ai and bj . We
shall summarize the important results in the next section.

3.2 Lie groups and Lie algebras

Let us start with the composition rules of the group parameters

c = f (a, b) or cl = f l (ai , bj ).

Since the elements a = (a1 , a2 , . . . , ar ) and b = (b1 , b2 , . . . , br ) are close

to the identity element e = (0, 0, . . . , 0), we perform Taylor’s expansion
of the functions f l (ai , bj ) at the origin of the group parameter space in
power series of ai and bj up to second order, namely,

∂f l (e, e) i ∂f l (e, e) i ∂ 2 f l (e, e) i j

cl = f l (ai , bj ) = f l (e, e)+ a+ b+ a b +O(3).
∂ai ∂bi ∂ai ∂bj

This equation can be further simplified by the following relations:

f l (e, e) = 0, f l (a, e) = al , f l (e, b) = bl ,

and hence
cl = f l (ai , bj ) = al + bl + fijl ai bj + O(3),

with

∂ 2 f l (e, e)
fijl = = constant coefficient. (3.10)
∂ai ∂bj

25:23.
88 Chapter 3

The above formula enables us to calculate the group parameters of

the inverse element a−1 = ((a−1 )1 , (a−1 )2 , . . . , (a−1 )n ). When a ∼ e, we
evaluate f (a, a−1 ) as
0 = cl = f l (a, a−1 ) = al + (a−1 )l + fijl ai (a−1 )j + O(3),
therefore, up to O(3) we have,
(a−1 )l = −al + fijl ai aj + O(3).
Let us consider the transformation of a vector in n-dimensional
space, as shown in the Figure 3.1, such that the transformation forms a
group G(a), i.e.

G(a)
x 7−→ x0 = h(x, a),

Figure 3.1: Transformation of a vector in n-dimensional space.

where G(a) is the r-parameter group, and a is the group element.

Let us perform an infinitesimal transformation from x0 to x00 by δa,
an element close to identity element e, right after the first transformation
by a, as follows
x00 = x0 + dx0 = h(x0 , δa).
The changes in the vector can be expressed as

∂h(x0 , e)
dx0 = δa = u(x0 )δa,
∂(δa)

where u(x0 ) stands for a n × r matrix.

The combined transformation can also be obtained through a single
transformation by the group element a+da with the following relations:

25:23.
3.2 Lie groups and Lie algebras 89

x00 = h(x, a + da), a + da = f (a, δa).

Therefore da can also be expressed as the first order approximation in

δa as follows

∂f (a, e) ∂f l (a, e) m
da = δa, or dal = l
δa = fm (a)δam , (3.11)
∂b ∂bm

where ∂f (a, e)/∂b is a r × r square matrix evaluated at b = e or if we

express δa in terms of da by inverting the matrix ∂f (a, e)/∂b, namely

 −1
∂f (a, e)
δa = da = ψ(a)da, (3.12)
∂b

where ψ(a) again is a r×r square matrix, the inverse matrix of ∂f (a, e)/∂b.
Therefore
dx0 = u(x0 )ψ(a)da,

or when it is written in terms of the components as follows

dx0m = um i j
i ψj da .

Let us consider the transformation for an arbitrary function F (x) =

F (x1 , x2 , . . . , xn ) in the vector space by a group element at the vicinity
of identity e. The change in F (x) is calculated as
 
∂F (x) ∂F (x) i m ∂
dF (x) = dx = u(x)δa = δa ui F (x) = δai Xi F (x).
∂x ∂x ∂xm

We define the differential operator

∂
Xi = um
i , (3.13)
∂xm

which is called the generator of the transformation, or the infinitesimal

generator of the group, or simply group generator.

25:23.
90 Chapter 3

Let us go back to the space translation discussed in previous chapter.

A vector x in R3 -space is displaced by a group element ξ = (ξ 1 , ξ 2 , ξ 3 ),
namely

T (ξ)
x 7−→ x0 = x + ξ,
or expressed in components
x0m = xm + ξ m .
The generators of the 3-dimensional translational group can be calcu-
lated as follows

∂ ∂x0m ∂ ∂ ∂
Xi = um
i m
= i m
= δim m = ,
∂x ∂ξ ∂x ∂x ∂xi

which is related to the i-th component of the momentum operators in

q-representation in quantum mechanics, i.e.

∂ i
i
= Pi , (3.14)
∂x ~

i ∂ i ~~
which allows us to express U (P, −ξ) = eξ ∂xi = e ~ ξ·P as the group
element for a coordinate transformation to the right by xi → xi + ξ i ,
which corresponds to the transformation of the physical state to the left
by ξ i .
We take the rotation about the azimuthal axis as another example.
It is a one-parameter group defined by
x01 = x1 cos θ − x2 sin θ, (3.15)

x02 = x1 sin θ + x2 cos θ, (3.16)

that allows us to calculate

25:23.
3.2 Lie groups and Lie algebras 91

∂x01
u11 = = −x2 , (3.17)
∂θ θ=0

∂x02
u21 = = x1 , (3.18)
∂θ θ=0

and we obtain the generator of the rotational group about z-axis as

follows

∂ ∂ ∂ i
X3 = ui1 i
= −x2 1 + x1 2 = L3 , (3.19)
∂x ∂x ∂x ~

which is a familiar expression related to the q-representation of the 3rd

component of the angular momentum operator.
The angle θ of the rotation in fact is just the group parameter. The
parameter takes the value 0 6 θ < 2π. We call it the O(2) group, an
orthogonal transformation in R2 -space. A continuous group with a fi-
nite number of parameters that takes a bounded domain in the group
parameter space is called compact Lie group. Otherwise we call them
non-compact ones. The translational group in the previous example is
a non-compact 3-parameter group, while the O(2) group is a compact
1-parameter group.

Lie algebra is defined as the commutator of the group generators.

Let us consider the commutator of Xi and Xj , i.e.

∂ m ∂
∂um
j ∂ l
m ∂ui ∂
[Xi , Xj ] = [uli , u ] = ul
− u
∂xl j ∂xm i
∂xl ∂xm j
∂xm ∂xl

∂um m
 
l j l ∂ui ∂
= ui l
− uj l
. (3.20)
∂x ∂x ∂xm

Yet the last parenthesis can be simplified as

25:23.
92 Chapter 3

∂um
j
m
l ∂ui
uli − uj = ckij um
k , (3.21)
∂xl ∂xl

where ckij = −ckji is a constant coefficient and anti-symmetric with re-

spect to i and j.
The proof, though a little tedious, goes as follows:

∂xm
Since dxm = um k n
k ψn da , we have = um k
k ψn . With the property
∂an
of infinite differentiability of the transformation functions h(x, a) with
respect to the group parameters, namely that

∂ 2 xm ∂ 2 xm
= , (3.22)
∂al ∂an ∂an ∂al

we reach at

∂ψlk ∂ψnk ∂um ∂um

 
um − + k
ψ k
− k
ψ k = 0. (3.23)
k
∂an ∂al ∂an l ∂al n

∂um
k ∂um
k l α
Putting = u ψ , the above equation takes the expression
∂an ∂xl α n
as

∂ψlk ∂ψnk ∂um ∂um

 
k α β k
um − + k α β k
u ψ n ψ − u ψ ψ = 0,
k
∂an ∂al ∂xα β l
∂xα β l n

or cast into the following expression

∂ψlk ∂ψnk ∂um ∂um

   
β α k
um − = uk − u ψlα ψnβ . (3.24)
k
∂an ∂al ∂xk α ∂xk β

By inverting ψ into f , the above equation can be written as

25:23.
3.2 Lie groups and Lie algebras 93

∂ψlk ∂ψnk ∂um ∂um

   
j
um − fil fjn = k
u − i k
u . (3.25)
k
∂an ∂al ∂xk i ∂xk j

The term on the right hand side of the last equation depends upon
the coordinate only, while the left hand side is product of a coordinate
dependent function umk and a group parameter dependent factor
 k k

∂ψl ∂ψn l n
∂an − ∂al fi fj . The condition can only be satisfied if,

∂ψlk ∂ψnk
 
− fil fjn = ckij = −ckji = constant coefficient.
∂an ∂al

Therefore we have,

∂um
j k ∂um
ui − i
uk = ckij um
k , (3.26)
∂xk ∂xk j

and hence the commutator of Xi and Xj can then be expressed as

∂um m
 
j k ∂ui ∂ ∂
[Xi , Xj ] = uki − uj = ckij um
k = ckij Xk , (3.27)
∂xk ∂xk ∂xm ∂xm

and they are called Lie algebras. ckij are called the structure con-
stants.

A few terminologies concerning Lie groups are introduced as follows:

(a) A Lie group is said to be Abelian if all the structure constants are
zero, i.e.

ckij = 0, i, j, k 6 r,

which imply that the generators of the group commute with each
other.

25:23.
94 Chapter 3

(b) A subset of the elements of a group G is called a subgroup of G

if those elements satisfy the group postulates. The generators of
the subgroup, X1 , X2 , . . . , Xi , . . . , Xp for p < r are closed, namely
they form a sub-Lie algebra by themselves, and

[Xi , Xj ] = ckij Xk , with ckij = 0 if i, j 6 p, k > p.

(c) It is called an invariant subgroup H of G, if the generators in

H commute with the generators outside H, namely that

[Xi , Xj ] = ckij Xk , with ckij = 0 if i 6 p, k > p.

(d) A group is called a simple group if it contains no invariant sub-

group. It is called a semi-simple if it contains no invariant, Abelian
subgroup.

The generators of the group also satisfy the Jacobi identity, i.e.
[[Xi , Xj ], Xk ] + [[Xj , Xk ], Xi ] + [[Xk , Xi ], Xj ] = 0, (3.28)

which leads to the following relation on the structure constants,

clij cm l m l m
lk + cjk cli + cki clj = 0. (3.29)

3.3 More on semisimple group

We shall confine ourselves to semisimple Lie group from now on for

further investigation. A symmetric g tensor is introduced for the discus-
sion on Cartan’s criteria of the semisimple group, and also the inverse
of the g tensor, i.e. g −1 , another symmetric one will be constructed for
the discussion on the Casimir operator.

3.3.1 Cartan’s criteria of the semisimple group

With the structure constants we construct a symmetric tensor g
defined as
gij = clim cm
jl = gji , (3.30)

25:23.
3.3 More on semisimple group 95

which is called the metric tensor or g tensor or sometimes by the name

of Killing form. We formulate the criteria given by Cartan in charac-
terizing the semisimple group in the following proposition.

Proposition 1.
A group G is semisimple if and only if det|g| 6= 0.

The condition det|g| 6= 0 for a semisimple group is necessary and

sufficient. Let us assume a Lie group contains an Abelian invariant sub-
group. We denote the index of the generators in the Abelian invariant
subalgebra by putting a bar on the top, i.e. ī, j̄ and k̄. It can be easily
demonstrated that det|g| of the group vanishes, i.e.

det|g| = 0, (3.31)

if the group contains an Abelian invariant subgroup. Let us calculate

the element gij̄ of the g tensor as follows

gij̄ = clim cm l m̄ l̄ m̄
j̄l = cim̄ cj̄l = cim̄ cj̄ l̄ = 0, (3.32)

which implies that the elements of the whole column of the g tensor
equal to zero. Hence the determinant of the g tensor is zero, i.e.

det|g| = 0,

and the sufficient condition of Cartan’s criteria is reached.

The condition of zero in determinant of the g tensor is also necessary.
Consider a system of r simultaneous homogeneous equations as follows

gij xj = 0. (3.33)

The condition det|g| = 0 also ensures the existence of the nontrivial

solutions of xj . Let us construct a generator X, with the coefficients
taken from the nontrivial solutions xj . The generators then form an
invariant subgroup as demonstrated in the following proposition.

25:23.
96 Chapter 3

Proposition 2.
If xj are the nontrivial solutions of the following simultaneous
homogeneous equations,
gki xi = 0
and Xi is the generators of the group G, then the generators xi Xi
form an invariant subalgebra of G.

The proof is quite straightforward by considering the following com-

mutator

[xi Xi , Xj ] = xi ckij Xk = y k Xk , (3.34)

where y k = xi ckij .

If we contract gik with y k , we obtain the following result

gik y k = gik xl cklj = xl clji = cnji gnl xl = 0, (3.35)

which implies that y k is also the solution of Eq. (3.33), and hence
the generators xi Xi form an invariant subalgebra. While in reaching
Eq. (3.35), we have made use of the cyclic properties of the indices in
clji = cjil which will be shown later in Proposition 3.

3.3.2 Casimir operator

The nonzero determinant of g, i.e. det|g| 6= 0, enables us to construct
the inverse of g, a tensor denoted by g −1 , with elements g ij such that

g il glj = δji ,

then g −1 will allow us to form a quadratic of the generators of the Lie

group, called Casimir operator defined as

C = g ij Xi Xj , (3.36)

25:23.
3.3 More on semisimple group 97

which shall commute with any generator of the group as can be verified
by taking the commutator of C with, say, the generator Xk in the
following expression
[C, Xk ] = g ij [Xi Xj , Xk ] = g ij clik (Xl Xj + Xj Xl ). (3.37)
Let us define a 3rd rank tensor through the following relation
clik = g lj cjik .
We shall prove the anti-symmetric property of cijk in the following
proposition.

Proposition 3.
The 3rd rank tensor cijk is anti-symmetric with respect to the
interchange of any pair of the indices.

The proof goes as follows

cijk = clij glk = clij cnlm cm
kn

= −cljm cnli cm l n m l n m l n m
kn − cmi clj ckn = cjm cil ckn + cmi clj cnk . (3.38)
By summing over the repeated indices, the last line of the above
equation becomes invariant under the cyclic permutation of i, j and k.
Therefore the conclusion of Proposition 3 is reached. Eqation (3.37) can
then be expressed as
[C, Xk ] = g ij g lm cmik (Xl Xj + Xj Xl )

= g ml g ij cmik (Xl Xj + Xj Xl ) = 0. (3.39)

Since the last term in the last equation is symmetric with respect to l and
j, so is the upper indices m and i, yet cmik is anti-symmetric with respect
to the lower indices m and i, the summation leads to [C, Xk ] = 0, which
concludes that the Casimir operator of the group commutes with all the
group generators. We shall make use of this property to investigate the
theory of angular momentum in the following chapter.

25:23.
98 Chapter 3

3.4 Standard form of the semisimple Lie algebras

Following the approach of Cartan and Weyl, let us consider the

following equation

[A, X] = ρX, (3.40)

where A is an arbitrary linear combination of the generators Xi with

the coefficient ai , i.e.

A = ai Xi , (3.41)

and X is another linear combination of the generators Xi with the co-

efficients xi , namely

X = xi Xi , (3.42)

such that Eq. (3.40) can be met.

We shall give a special name to X of the eigenvalue equation in
Eq. (3.40) and call X the eigengenerator, while ρ is also called the
eigenvalue for this particular form of the eigenvalue equation. If we
express Eq. (3.40) explicitly in terms of the structure constants ckij ,
then we have the following simultaneous homogeneous equations, i.e.

(ai ckij − ρδjk )xj = 0, (3.43)

with the secular equation given by

det|ai ckij − ρδjk | = 0. (3.44)

For the r-parameter Lie group, there are r roots of the solution of
ρ in Eq. (3.44). We shall state the conclusions from Cartan’s without
providing the argument in details that if ai is chosen such that the
secular equation has the maximum numbers of distinct roots, then only
the eigenvalue ρ = 0 becomes degenerate. Let l be the multiplicity of
the degenerate roots of the secular equation, then l is said to be the
rank of the semisimple Lie algebra. The eigengenerator corresponding
to the degenerate eigenvalue ρ = 0 is denoted by Hi (i = 1, 2, . . . , l),
which commutes with each other, i.e.

[Hi , Hj ] = 0, (3.45)

25:23.
3.4 Standard form of the semisimple Lie algebras 99

with the structure constant cµij = 0 where i, j 6 l and µ 6 r.

The eigenvalue equations with nonzero eigenvalues read as

[A, Eα ] = αEα , (3.46)

where we have used the eigenvalue α to label the eigengenerator as Eα .

Since A commutes with Hi , it allows us to express A as the linear
combination of Hi , namely

A = λi Hi . (3.47)

Let us investigate the commutator of A and [Hi , Eα ] as follows

[A, [Hi , Eα ]] = [A, Hi Eα ] − [A, Eα Hi ]

= α(Hi Eα − Eα Hi ) = α[Hi , Eα ], (3.48)

which implies that there exist l eigengenerators corresponding to the

same eigenvalue α in the eigenvalue equation of Eq. (3.40). It is the re-
sult contradictory to the assumption that α is not degenerate. Therefore
we conclude that [Hi , Eα ] must be proportional to Eα , i.e.

[Hi , Eα ] = αi Eα , (3.49)

where the structure constant cβiα = αi δαβ , with α and β taking all the
distinct nonzero eigenvalues. One also proves readily that

α = λ i αi (3.50)

by making use of Eqs. (3.46), (3.47) and (3.49).

To further our investigation, let us make use of the Jacobi identity
and we find that

[A, [Eα , Eβ ]] = −[Eα , [Eβ , A]] − [Eβ , [A, Eα ]]

= (α + β)[Eα , Eβ ], (3.51)

which means that [Eα , Eβ ] is the eigengenerators of A with eigenvalue

α + β if and only if α + β 6= 0. Hence we can express the commutator
of Eα and Eβ as another eigengenerator as follows

[Eα , Eβ ] = Nαβ Eα+β , (3.52)

25:23.
100 Chapter 3

with the structure constant cγαβ = Nαβ δα+β

γ
.
For the case that α + β = 0, the commutator [Eα , E−α ] shall be
proportional to the linear combination of Hi . Therefore we reach the
following conclusions:
[Eα , E−α ] = ciα−α Hi , (3.53)
[Eα , Eβ ] = Nαβ Eα+β . (3.54)

Let us evaluate the elements of the g tensor in the standard form

and divide them into three categories, namely gij , giα and gαβ , where
the indices in the subscripts take the following values:


 i, j 6 l,

α, β takes the values, r − l in total numbers, corresponding to

the roots of non-zero eigenvalues.


The evaluation of gij is the simplest as follows, for

gij = cαiβ cβjα = αi αj . (3.55)

We shall leave them as an exercise to show that

giα = 0. (3.56)

Finally let us calculate the element gαβ as follows

gαβ = cναµ cµβν = cααγ cγβα + cµα−α c−α γ
X
βµ + cα+γ
αγ cβ(α+γ) , (3.57)
γ6=−α

where µ, ν takes the values from 1 to l as well as all those corresponding

to the nonzero eigenvalues.
With the properties of Eqs. (3.49), (3.53) and Eq. (3.54), we reach the
conclusion that gαβ is 0 unless β = −α, namely
gαβ = 0 if α + β 6= 0. (3.58)

Suppose that α is the eigenvalue of Eq. (3.40), but −α is not, then

the whole column in the g tensor becomes 0 because gαβ = 0. The deter-
minant of g will vanish in this case, and Cartan’s criteria for semisimple

25:23.
3.5 Root vector and its properties 101

group will be violated. Therefore the roots α and −α of the secular

equation Eq. (3.44) must appear in pair and gα−α will be different from
0. We shall choose the nonzero gα−α = 1, due to the fact that the
eigengenerator Eα is linear on both sides of the eigenvalue equation of
Eq. (3.40). This degree of freedom in the normalization of gij allows us
to evaluate ciα−α as follows
ciα−α = g ij cα−αj = g ij cjα−α = g ij cβjα gβ−α

= g ij cαjα · 1 = g ij αj = αi . (3.59)
Hence we can rewrite Eq. (3.53) as follows
[Eα , E−α ] = αi Hi . (3.60)
Here we summarize the algebra of the semisimple group in the fol-
lowing standard form
[Hi , Hj ] = 0, (3.45)
[Hi , Eα ] = αi Eα , (3.49)
i
[Eα , E−α ] = α Hi , (3.60)
[Eα , Eβ ] = Nαβ Eα+β , if α + β 6= 0. (3.54)

3.5 Root vector and its properties

Let us regard the root α as an element in l-dimensional vector space

with its covariant components αi , i.e.
α = (α1 , α2 , . . . , αl ), (3.61)
where a bar is placed under the root, written as α to indicate it as a
covariant vector. While the contravariant vector, denoted by α, with
the component αi , which has been derived previously, can be expressed
as
α = (α1 , α2 , . . . , αl ). (3.62)
If we denote the scalar product of two roots α and β as follows
(α, β) = αi β i = αi βi , (3.63)

25:23.
102 Chapter 3

then we obtain the following proposition.

Proposition 4.
If α and β are two roots, then
(α, β) 1
= integers. (3.64)
(α, α) 2

We start the proof by taking the commutator of the eigengenerators

E−α and Eγ , namely [E−α , Eγ ]. Since we enjoy the degree of freedom
to rescale the eigengenerators, the structure constants can be chosen to
be one, i.e.
[E−α , Eγ ] = Eγ−α . (3.65)
Let us assume that γ is a root but not γ + α. Then we can construct a
series of roots by performing the commutator successively of E−α and
a general eigenvector Eγ−jα with different j, i.e.
[E−α , Eγ−jα ] = Eγ−(j+1)α . (3.66)
The series of roots will start from γ and terminate at γ − gα, or
explicitly as follows
γ, γ − α, γ − 2α, . . . , β, . . . , γ − gα, (3.67)
which is called α-string, in which the root β stands for any root in the
string.
The finite length of the α-string comes from the fact that it is a
r-parameter group, the finite number of roots prohibit us to perform
the commutator indefinitely. Let us assume the string stops at the root
γ − gα, then the commutator vanishes, i.e.
[E−α , Eγ−gα ] = Eγ−(g+1)α = 0. (3.68)
Now let us take the commutator of eigengenerators Eα and any
generator, say Eγ−(j+1)α , then we obtain as follows
[Eα , Eγ−(j+1)α ] = µj+1 Eγ−jα , (3.69)

25:23.
3.5 Root vector and its properties 103

where the structure constant µj+1 can not be normalized further for
the reason that all eigengenerators corresponding to the roots in α-
string have been rescaled already. Applying the Jacobi identity for the
generators Eα , E−α and Eγ−jα , which leads to the following relation

µj+1 Eγ−jα = [Eα , [E−α , Eγ−jα ]]

= −[E−α , [Eγ−jα , Eα ]] − [Eγ−jα , [Eα , E−α ]]

= −[Eγ−jα , αi Hi ] + µj [E−α , Eγ−(j−1)α ]

= αi [Hi , Eγ−jα ] + µj Eγ−jα ,

and we arrive at a recursion formula for the structure constant µj+1 as

follows

µj+1 = (γ, α) − j(α, α) + µj , (3.70)

where j takes the integer with µ0 = 0.

The structure constant µj then can be evaluated as follows

1
µj = j(γ, α) − j(j − 1)(α, α). (3.71)
2

Since j terminates at g, therefore µg+1 = 0. It follows that

1
(γ, α) = g(α, α). (3.72)
2

Let us now take β = γ −jα as any root in the α-string, and eliminate
γ in Eq. (3.72), then we reach

1
(α, β) = (g − 2j)(α, α). (3.73)
2

which proves Proposition 4.

25:23.
104 Chapter 3

3.6 Vector diagrams

We are now in the position to investigate the graphic representations

of the root vectors. As we have shown in the construction of α-string
that

1
(α, β) = m(α, α), (3.74)
2

where m takes the integer. Similarly, it follows that

1
(α, β) = n(β, β), (3.75)
2

if the β-string is taken into consideration. Therefore we obtain the

following condition

(α, β)(β, α) 1
cos2 ϕ = = mn, (3.76)
(α, α)(β, β) 4

and cos2 ϕ takes only the value 0, 1/4, 1/2, and 1.

As we have demonstrated that the roots α and −α come in pair, we
need only to consider the positive values of cos ϕ, which lead us to take
the following possible angles for ϕ, i.e.
π π π π
ϕ = 0, , , , .
6 4 3 2

The length of the root vector α and β can then be calculated according
to the various combinations in m and n as follows:

r
p 2(α, β)
|α| = (α, α) = ,
m
r
p 2(α, β)
|β| = (β, β) = .
n

25:23.
3.6 Vector diagrams 105

With the method introduced by Van der Waerden, we summarize the

results in 5 cases as follows:
(1) for ϕ = 0, thus we have the trivial result that two vectors α and
β are identical.
(2) for ϕ = π/6, m = 1 or 3 and n = 3 or 1 respectively, and the ratio
of the length square, i.e. (β, β)/(α, α) = 1/3 or 3.
(3) for ϕ = π/4, m = 1 or 2 and n = 2 or 1 respectively, and the ratio
of the length square, i.e. (β, β)/(α, α) = 1/2 or 2.
(4) for ϕ = π/3, m = 1 and n = 1 respectively, and the ratio of the
length square, i.e. (β, β)/(α, α) = 1.
(5) for ϕ = π/2, the scalar product vanishes and the ratio of the
length is undetermined.
It is easy for us to demonstrate the graphic representation in two
dimensional case, namely for rank two Lie groups. The root vectors can
be drawn on a two dimensional space. For the above case 2 to case 4,
the vector diagrams are:

A. ϕ = π/6

(a) (β, β)/(α, α) = 3 (b) (β, β)/(α, α) = 1/3

The group is called G2 group after Cartan, an exceptional group

containing 2 null root vectors and another 12 root vectors.

25:23.
106 Chapter 3

B. ϕ = π/4

(c) (β, β)/(α, α) = 2 (d) (β, β)/(α, α) = 1/2

(e) (β, β)/(α, α) = 1 (f) (β, β)/(α, α) = 1

Figure 3.2: Vector diagram of G2 , B2 , C2 , A2 and D2 group.

They are called B2 group and C2 group, which has 2 null root vectors
and 8 root vectors and are associated with SO(5) group and Sp(4) group,
the symplectic group in 4-dimension, respectively.

C. ϕ = π/3

It is called A2 group, a group associated with SU(3), which contains

6 root vectors and 2 null root vectors.

25:23.
3.7 PCT: discrete symmetry, discrete groups 107

D. ϕ = π/2
We shall consider another one called D2 group, with two pairs of
mutually orthogonal root vectors. D2 group is commonly referred to as
SO(4), which is isomorphic to the direct product of two SO(3) groups.

The groups of general rank l take the names Al , Bl , Cl and Dl

corresponding to SU(l + 1), SO(2l + 1), Sp(2l) and SO(2l) respectively.
Further investigation in those groups mentioned above as well as the
exceptional Lie groups G2 , F4 , E6 , E7 and E8 shall be left to the readers
to consult with more advanced books and journals.

3.7 PCT: discrete symmetry, discrete groups

Besides the continuous groups we have investigated in the past sec-

tions, physics involves another category of symmetry, called the dis-
crete symmetry. The groups associated with discrete transformations
are called the discrete groups. Parity, charge conjugation and time re-
versal transformation are among the most frequently discussed subjects
in quantum mechanics, particularly in particle physics. We shall devote
ourselves to study of PCT in this section.

3.7.1 Parity transformation

Let us start with parity transformation. To investigate the phe-
nomenon of any physical system, it is imperative to use a coordinate
system. Yet as far as the motion of the particle is concerned, physi-
cists do not care if a right-handed coordinate system is used or if a
left-handed one is adopted. The reason is simple. For the motion of a
system of particles in the right-handed coordinate space, there always
exists an identical motion in the space of its mirror image (3-dimensional
one, of course). In other words, the motion of the particles is invari-
ant under the mirror reflections. It is said that the equation of motion
in classical mechanics is invariant under the parity transformation. It
can be visualized as the symmetry between a physical phenomenon in
space and its image phenomenon in mirror space. This is contrary to
our previous understanding of the symmetry discussed in the past few
sections, which are restricted solely to continuous transformations such

25:23.
108 Chapter 3

as the spatial translation or rotation. The symmetry arising from the

parity transformation on the other hand, is discrete. For this very rea-
son, we coin a term, discrete symmetry. The immediate consequence of
this discrete symmetry is that we are unable to reorient continuously a
right-handed coordinate system into a left-handed one and vice versa.
Therefore parity transformation falls into the category of discrete trans-
formation.
Quantum mechanically, the parity transformation can be represented
by an operator P, which changes the position ~x into −~x, i.e.
P.T.
P : ~x 7−→ ~xp = −~x, (3.77)

or in general,
P.T.
P : f (~x) 7−→ f (~xp ) = f (−~x). (3.78)

It is obvious that

P 2 = I, (3.79)

because that two successive parity transformations leave the position

vector unchanged.
The parity transformation of a dynamical observable V is expressed
as follows

PVP −1 = Vp . (3.80)

We shall introduce two terms which are relevant to parity transfor-

mation.
An operator V is called polar vector operator, or simply vector
operator if it anti-commutes with the parity operator, i.e.

{P, V} = 0, (3.81)

or more explicitly that

PVP −1 = Vp = −V. (3.82)

The position operator and the momentum operator are of this kind, i.e.

25:23.
3.7 PCT: discrete symmetry, discrete groups 109

PXP −1 = −X, (3.83a)

PPP −1 = −P. (3.83b)

On the other hand, an operator A which commutes with the parity,

is called the axial vector operator, namely

[P, A] = 0, (3.84)

or

PAP −1 = Ap = A. (3.85)

One can easily prove that the angular momentum L is an axial vector
operator.
It is of interest to investigate a quantum system with symmetry
under parity transformation. We shall summarize a few remarkable
results in the following proposition.

Proposition 5.
If a dynamical operator S is invariance under the parity trans-
formation, then there exists a pair of vectors which are the si-
multaneous eigenvectors of S and P with the eigenvalues s and
±1 respectively.

Let us consider the following eigenvalue equation

S|si = s|si.

By applying the parity operator on both sides of the last equation, we

can easily show that

PS|si = PSP −1 P|si = Sp |sip = S|sip = s|sip , (3.86)

where Sp = S and P|si = |sip .

25:23.
110 Chapter 3

This implies that |sip is also the eigenvector of S with the eigenvalue
s, namely that the eigenvectors are of 2-fold degeneracy. Therefore it
allows one to construct a pair of vectors

1
|s± i = √ (|si ± |sip ), (3.87)
2

with the following properties

P|s± i = ±|s± i, (3.88)

which proves Proposition 5.

If we take S to be the Hamiltonian operator of a central forces
system, i.e.

1 ~2 ~
S=H= P + V (|X|). (3.89)
2m

It is obvious that PHP −1 = H. Hence we have

d
i~ hPi = h[P, H]i = 0. (3.90)
dt

Therefore the parity is conserved, or in another way of saying, sym-

metry is preserved under parity transformation.

3.7.2 Charge conjugation and time reversal transformation

There exists another discrete transformation called the charge con-
jugation. A charge conjugation is denoted by C, which takes a particle
into its antiparticle, namely C changes the charge of a particle q into
the charge of its antiparticle −q. We construct a one-dimensional charge
space in which all the particles take their positions on a line or an axis,
the coordinate system of the charge space, according to the amount of
the electric charge they bear. Particles with positive charge and par-
ticles with the negative one are respectively located on the right hand
side and the left hand side of the origin of this one-dimensional frame of

25:23.
3.7 PCT: discrete symmetry, discrete groups 111

reference. The operation of the charge conjugation can then be regarded

mathematically as mapping a point q into the mirror image point −q,
a reflection with respect to the origin, i.e.
C.T
C : q 7−→ −q. (3.91)

Hence the charge conjugation operator C is isomorphic onto the parity

operator P of the space reflection in 1-dimensional space.
If we denote the charge operator Q associated with the physical ob-
servable of the electric charge of a particle, then the charge conjugation
C takes Q into −Q in the following similarity transformation
C : CQC −1 = −Q, (3.92)

or {C, Q} = 0. (3.93)

It implies that the charge conjugation operator C always anti-

commutes with the charge operator Q.
The one-to-one correspondence between the charge conjugation and
the parity transformation in one dimension allows one to apply the
results derived from the parity operation to the case of the charge con-
jugation.
Let us now proceed to explore the third discrete transformation T ,
called the time reversal transformation. We start with the following
time dependent Schrödinger equation
 
∂ ~
i~ ψ(~x, t) = H ∇, ~x ψ(~x, t). (3.94)
∂t i

If we take the complex conjugate on both sides of the equation above,

and replace t by −t, then

 
∂ ∗ ~
i~ ψ (~x, −t) = H − ∇, ~x ψ ∗ (~x, −t). (3.95)
∂t i

It occurs often that the Hamiltonian H(−(~/i)∇, ~x) = H((~/i)∇, ~x)

if H contains only the quadratic term in P~ . Then the Schrödinger equa-

25:23.
112 Chapter 3

tion becomes

 
∂ ∗ ~
i~ ψ (~x, t) = H ∇, ~x ψ ∗ (~x, −t). (3.96)
∂t i

This implies that both ψ(~x, t) and ψ ∗ (~x, −t) are solutions of the
Schrödinger equation.
The appearance of the complex conjugate as well as the replacement
of t by −t in the wave function of the last equation allows one to con-
struct the time reversal operator which bears some sort of property in
antilinearity. If we decompose T into the product of K and T , respec-
tively standing for an antilinear operator and a temporal reflection
operator, then we reach the conclusions in the following proposition.

Proposition 6.
If the time reversal operator T takes as product of K and T , i.e.

T = KT ,
(
K = antilinear operator,
where
T = temporal reflection operator,

then both ψ(~x, t) and Kψ(~x, −t) are solutions of the time depen-
dent Schrödinger equation as long as KHK −1 = H. Moreover,
K 2 ψ = ηψ for any wave function ψ and unimodulus factor η.

The proof goes as follows. Let us consider the transformation of H

and ψ,

 
−1 ~
T HT = KH ∇, ~x, −t K −1 ,
i

T ψ(~x, t) = Kψ(~x, −t).

25:23.
3.7 PCT: discrete symmetry, discrete groups 113

Then the Schrödinger equation takes the expression as

∂
i~ Kψ(~x, −t) = (KHK −1 )Kψ(~x, −t). (3.97)
∂t

For the stationary system with Hamiltonian containing the quadratic

term in linear momentum P~ , then KH(~∇/i, ~x)K −1 = H(~∇/i, ~x).
Therefore we have

 
∂ ~
i~ Kψ(~x, −t) = H ∇, ~x Kψ(~x, −t). (3.98)
∂t i

This proves the first part of the proposition.

If we take the inner product of any pair of vectors ψ and ϕ, then by
the antiunitary property of the operator K, we have

(ψ, ϕ) = (Kϕ, Kψ) = (K 2 ψ, K 2 ϕ) = |η|2 (ψ, ϕ), (3.99)

which implies that η is unimodulus, i.e. |η| = 1. Hence we complete the

proof of Proposition 6.
An immediate consequence of the second part of Proposition 6 is
the existence of Kramers degeneracy if the quantum system is invariant
under time reversal transformation. We derive the result as an example.

Example

Since ψ and Kψ are solutions of the Schrödinger equation, and are

also orthogonal to each other as can be proved in the following line

(ψ, Kψ) = (K 2 ψ, Kψ) = η(ψ, Kψ) = −(ψ, Kψ) = 0, (3.100)

by taking η = −1. Therefore the existence of Kramers degeneracy is

established.

25:23.
114 Chapter 3

3.8 Exercises

Ex 3.8.1

Consider the following transformation,

x0 = a1 x + a2 .

(a) What are the conditions if the transformation forms a 2 parame-

ters group?

(b) Denote the group element by a = (a1 , a2 ). Find the inverse ele-
ment of a, namely, a−1 = ((a−1 )1 , (a−1 )2 ).

(c) What are the composition rules of the group parameters, i.e. try
to find c = ba for cl = cl (ai , bj ) = f l (ai , bj )?

(d) Can you construct the 2 × 2 matrix representation of the group to

justify your answer in (a), (b) and (c)?

Ex 3.8.2

Consider the rotational transformation about the azimuthal axis

with an angle θ, namely,

x01 = x1 cos θ − x2 sin θ,

x02 = x1 sin θ + x2 cos θ,

x03 = x3 ,

then the function F (x) = F (x1 , x2 , x3 ) will be changed into

F (x0 ) = F (x1 cos θ − x2 sin θ, x1 sin θ + x2 cos θ, x3 ).

Prove explicitly that the transformed function can be obtained by

25:23.
3.8 Exercises 115

  
2 ∂ 1 ∂
exp(θX3 )F (x) = exp θ −x +x F (x1 , x2 , x3 )
∂x1 ∂x2

= F (x01 , x02 , x03 ).

(Hint: use exp(θX3 ) = exp (θ∂/∂ϕ) and write

F (x) = F (r cos ϕ, r sin ϕ, x3 ).)

Ex 3.8.3
A conjugate subgroup is defined as a−1 Ha if H is a subgroup, where
a ∈ G. Show that the self-conjugate subgroup can alternatively be
defined as the invariant subgroup.

Ex 3.8.4
Verify that the Jacobi identity leads to following relation among the
structure i.e.

clij cm l m l m
lk + cjk cli + cki clj = 0.

Ex 3.8.5
Show that the elements giα of the g tensor in the standard form of
Lie algebra vanish.

Ex 3.8.6
Give the argument to verify the elements gαβ of the g tensor in the
standard form of Lie algebra also vanish if α + β 6= 0.

Ex 3.8.7
Let the parity operator be defined as (Hint: by using Eq. (1.39).)

25:23.
116 Chapter 3

π ~ ~ ~ ~
P = e 2 (P ·X+X·P ) .
Show that
~ −1 = −X,
P XP ~

P P~ P −1 = −P~ .

25:23.
Chapter 4

Angular Momentum

4.1 O(3) group, SU(2) group and angular momentum

Rotational transformation is one of the most commonly studied sub-

jects in physics. Rotational transformation is an orthogonal transfor-
mation, that leaves the norm of the vector invariant. Orthogonal trans-
formations of the coordinate system of a three-dimensional vector space
form a group of O(3). Yet as far as the invariance of the norm is con-
cerned, there exists another transformation, namely a unitary transfor-
mation, the SU(2) group, which is homomorphic onto SO(3) group, and
transformation also leaves the norm of a vector in C 2 -space invariant.
We shall review the orthogonal transformation in three-dimensional vec-
tor space and investigate the group structure and properties of the O(3)
group, as well as the SU(2) group which are utilized as the tools for the
discussion of the theory of angular momentum in the following section.

4.1.1 O(3) group

Let us perform the rotational transformation in R3 -space, i.e. the
three-dimensional vector space, about a fixed direction denoted by a unit
vector n̂ as the axis of rotation with an infinitesimal angle dθ. It can also
be written as a vector representing the direction of the rotational axis
as well as the magnitude of the rotation as expressed in the following
form
dθn̂ = dθ~ = (dθ1 , dθ2 , dθ3 ), (4.1)

25:31.
118 Chapter 4

where we decompose the infinitesimal rotation dθ~ into its components

in the last term of the above equation. For a finite rotation with an
angle θ~ about the fixed axis n̂ can be also written as

θn̂ = θ~ = (θ1 , θ2 , θ3 ). (4.2)

Henceforth, each rotation will be represented by a point at (θ1 , θ2 , θ3 )

in the three-dimensional parameter space.
The infinitesimal rotation will change the vector ~x into ~x0 , expressed
in the usual vectorial algebraic notation in the following form as

~x0 − ~x = d~x = dθ~ × ~x, (4.3)

where dθ~ = (dθ1 , dθ2 , dθ3 ) is an infinitesimal vector in the parameter

space of the rotational group at the vicinity of the origin. The parameter
space is also a three-dimensional Euclidean space that is isomorphic onto
the previous vector space described by the coordinate system.
To express the change in the vector due to the infinitesimal rotation
in terms of the vector components, one has

dxi = ijk dθj xk ,

that allows us to calculate the matrix elements uli in the rotational group
O(3), i.e.

∂xl
uli = = lik xk .
∂θi

∂
Since Xi = uli , we obtain that
∂xl

∂
Xi = lik xk , (4.4)
∂xl

and the Lie algebra of the O(3) group is given as follows

25:31.
4.1 O(3) group, SU(2) group and angular momentum 119

 
m ∂ k ∂ ∂
[Xi , Xj ] = lim njk x ,x = (kim ljk + kmj lik )xm l
∂xl ∂xn ∂x

∂
= kij lkm xm = −kij Xk , (4.5)
∂xl

where the fourth term of the above equation is obtained by the Jacobi
identity,
kim ljk + kmj lik + kji lmk = 0. (4.6)
Hence we have the following Lie algebra of the rotational transfor-
mation in 3-dimensional vector space,
[Xi , Xj ] = −kij Xk , (4.7)
that reproduce the commutation relations for the quantum operators of
the angular momentum, i.e.
[Li , Lj ] = i~kij Lk , (4.8)

i
if we put Xi = Li .
~
There are various representations for the generators of the symme-
try transformation as well as for the Lie algebras. The representations
of the group generators and the Lie algebras we adopted so far are ex-
pressed with the purpose to investigate the properties in the symmetry
transformation of the functions. We refer this representation as the
canonical formulation. If we focus our attention on the transformation
of the vectors directly, we attain completely different representations.
Take the rotational transformation of a vector in matrix notation as
follows

x01
   
x1
x0 = Rx
 02   2
or x  = R x  , (4.9)
x03 x3

25:31.
120 Chapter 4

where R is a 3×3 square matrix, which is called an orthogonal matrix

because the rotational transformation leaves the norm of the vector
invariant, i.e.

x0T x0 = xT RT Rx or RT R = I. (4.10)

Orthogonality imposes the following 6 conditions on the matrix ele-

ments of R, i.e.

RT ij Rkj = δki or Rij Rkj = δik . (4.11)

The above orthogonality conditions, asides from having the following

relation

det RT R = (det R)2 = 1,

will also reduce the independent parameters in the matrix R from 9 to

3. It is obvious that the orthogonal matrices R form a group and that
is the very reason to name it the O(3) group, the orthogonal group in
3-dimension.
To investigate the matrix representations of the group generators,
let us construct the orthogonal matrix R = R(θ1 , θ2 , θ3 ), where θ1 , θ2
and θ3 stand for the group parameters. If we take the parameters small
enough and keep only the first order in θi , the matrix can be expressed
as follows

R(θ1 , θ2 , θ3 ) = I + A(θ1 , θ2 , θ3 ),

where the matrix A(θ1 , θ2 , θ3 ) is close to a null matrix, namely that

A(0, 0, 0) = 0. (4.12)

Since the orthogonality condition of the matrix R leads, to first

order approximation, the following relation

RT R = (I + AT )(I + A) ' I + AT + A = I,

which implies that matrix A is antisymmetric The generators of the

rotational group O(3) are obtained by taking the derivative of matrix
A(θ1 , θ2 , θ3 ) with respect to the corresponding group parameter, i.e.

25:31.
4.1 O(3) group, SU(2) group and angular momentum 121

∂
Xi = A(θ1 , θ2 , θ3 ) , (4.13)
∂θi θi =0

which again is an antisymmetric matrix as well.

Let us consider the rotational transformation about the 3rd axis
with an angle ϕ, the orthogonal matrix R(0, 0, ϕ) can, up to first order
in ϕ, be approximated as

   
0 − sin ϕ 0 0 −ϕ 0
R = I + A(0, 0, ϕ) = I + sin ϕ 0 0 ' I + ϕ 0 0 ,
   

0 0 0 0 0 0

which allows us to obtain the generator X3 of O(3) group by taking the

derivative of matrix A respect to ϕ, namely

 
0 −1 0
∂A 
X3 = = 1 0 0 . (4.14)

∂ϕ
0 0 0

The reason that X3 is called the generator of the group is due to

the retrievability of the rotational matrix R(0, 0, ϕ), even with the finite
angle ϕ, by exponentiating the matrix ϕX3 and obtaining the following
result

 
cos ϕ − sin ϕ 0
R(0, 0, ϕ) = eϕX3 =  sin ϕ cos ϕ 0 . (4.15)
 

0 0 1

Similar to the derivation of the group generator X3 , the other two

generators can also be calculated and given as

25:31.
122 Chapter 4

   
0 0 1 0 0 0
X2 =  0 0 0 , X1 = 0 0 −1 , (4.16)
   

−1 0 0 0 1 0

by performing the infinitesimal rotations about the axis accordingly.

These three matrices are referred to as the generators of O(3) group in
Cartesian bases.
Let us construct the orthogonal matrix R in terms of the Euler
angles ϕ, θ and χ. It is obtained by performing a rotation about the
3rd axis with an angle ϕ followed by another rotation about the 2nd
axis with an angle θ and then making a third rotation about the 3rd
axis with an angle χ to achieve the matrix R as follows
R(ϕ, θ, χ) = eχX3 eθX2 eϕX3 . (4.17)

It can be easily verified that

R(ϕ, θ, χ + 2mπ) = R(ϕ, θ, χ), (4.18)

for m = integers. That implies only a finite domain in the parameter

space is enough to exhaust all the elements of the O(3) group. Therefore
the parameters take the following ranges
0 6 ϕ 6 2π, 06θ6π and 0 6 χ 6 π. (4.19)

Furthermore the orthogonality condition is automatically built-in be-

cause of the antisymmetric property in the group generators, namely
that
RT = e−ϕX3 e−θX2 e−ξX3 = R−1 . (4.20)

It is interesting to observe that

det R = 1,

which takes only the positive root in the equation (det R)2 − 1 = 0.
This can be understood that the matrix R(ϕ, θ, χ) is reached from the
identity element I = R(0, 0, 0) with the determinant being equal to one,
by varying the group parameters continuously away from the origin of

25:31.
4.1 O(3) group, SU(2) group and angular momentum 123

Figure 4.1: Unit vector n̂(θ, ϕ).

the group parameter space. The elements of this particular orthogonal

matrices form a group which is called the group of SO(3). Since the
Euler angles take a finite range of the values, therefore the group pa-
rameters of SO(3) are confined within the domain of a sphere with the
radius π in the parameter space. If we assign χ as the angle of rotation
about a unit vector n̂ = (sin θ cos ϕ, sin θ sin ϕ, cos θ) which serves as the
axis of rotation as shown in Figure 4.1, all points within the sphere of
radius π in the parameter space, representing all the elements of the
rotational group, can be reached by choosing suitable n̂ and χ, only if
the two diametrically opposite points, namely points of antipode on the
surface of the sphere in the group manifold, are identified.
Let us abbreviate the unit vector as n̂(θ, ϕ), and denote the group
generators as X ~ = (X1 , X2 , X3 ), then the group elements can be ex-
pressed as follows
~
R(ϕ, θ, χ) = eχn̂(θ,ϕ)·X . (4.21)

The elements of the O(3) group consist of R(ϕ, θ, χ) and −R(ϕ, θ, χ)

and −R(ϕ, θ, χ), the mirror images of the elements R(ϕ, θ, χ), namely
taking spatial reflection followed by a rotation. In the other word,

25:31.
124 Chapter 4

O(3) group contains the elements R(ϕ, θ, χ) as well as the elements

−R(ϕ, θ, χ).

4.1.2 U(2) group and SU(2) group

Let us now consider the complex vector space of 2 dimensions, i.e.

C 2 - space, in which a vector ξ is represented by a column matrix with
two components of complex numbers as follows

!
ξ1
ξ= . (4.22)
ξ2

A linear transformation on the vector by a matrix M is expressed

as

! ! !
0 ξ 01 α β ξ1
ξ = Mξ, or = , (4.23)
ξ 02 γ δ ξ2

where the matrix elements of M are all of complex numbers.

If we restrict the transformation to be unitary, which leaves the norm
of the vector invariant, then

ξ 0† ξ 0 = ξ † M† Mξ = ξ † ξ, (4.24)

and we reach that

M† M = I or M† = M−1 and |det M|2 = 1. (4.25)

The transformation stated above is called the U(2) group. If we

take det M = 1 as an extra restriction on matrix M, then matrix M
is not only unitary but also unimodular, and the transformations form
the group of SU(2).
From now on we shall confine ourselves in discussing SU(2), in which
the condition M† = M−1 can be expressed explicitly as follows

25:31.
4.1 O(3) group, SU(2) group and angular momentum 125

! !
α∗ β ∗ 1 δ −β
= . (4.26)
γ ∗ δ∗ det M −γ α

Therefore we have the following relations

α∗ = δ, β = −γ ∗ , (4.27)

and the matrix M takes the expression as

!
α β
M= ,
−β α∗
∗

which contains only 3 independent parameters due to the condition

det M = |α|2 + |β|2 = 1.
Let us investigate the group property of the matrix M when it is
close to the identity element I, and if we put α = 1 + ξ, then the matrix
M can be rewritten as

!
ξ β
M=I+ = I + D.
−β ξ ∗
∗

The unitarity condition of matrix M, at the vicinity of the identity

element, becomes

M† M = (I + D† )(I + D) ' I + D† + D = I,

which implies D† +D = 0, or more specifically that ξ+ξ ∗ = 0. Therefore

we are able to rewrite the matrix M in the following form

!
i i
2c 2 (a− ib)
M=I+ i
,
2 (a + ib) − 2i c

25:31.
126 Chapter 4

where a, b and c, all real, are the three parameters of the SU(2) group.
The generators of the group are obtained as follows
! !
i 1
∂M 0 2 ∂M 0 2
Y1 = = i
, Y2 = = ,
∂a 2 0 ∂b − 12 0
!
i
∂M 2 0
Y3 = = . (4.28)
∂c 0 − 2i

We can verify immediately that the commutators among the gener-

ators take the general expression
[Yi , Yj ] = −kij Yk , (4.29)
which is exactly the same as the Lie algebras in the case of the O(3)
group.
The generators of the SU(2) group are related to Pauli matrices
as follows

i i i
Y1 = σ 1 , Y2 = σ2 , Y3 = σ3 . (4.30)
2 2 2

The identical Lie algebras for both groups O(3) and SU(2) do not
necessary have the same domain in the group manifold. Let us construct
the group elements, i.e. the matrix M by exponentiating the group
generators as follows
~
M = eχn̂·Y , (4.31)
where n̂ = n̂(θ, ϕ) has the same definition as that of O(3) group, yet the
parameter χ takes different domain in the group manifold for the reason
that both matrix M and matrix −M fulfill the unitarity condition of
the transformation. Therefore the group elements of SU(2) will fill the
sphere of radius 2π, namely it takes all the points in a domain of the
sphere with radius 2π in the parameter space to exhaust the group
elements. The parameters ϕ, θ and χ range as follows

25:31.
4.1 O(3) group, SU(2) group and angular momentum 127

0 6 ϕ 6 2π, 0 6 θ 6 π, 0 6 χ 6 2π, (4.32)

with again the points of antipode on the surface of the sphere are iden-
tified.
It will be left to you as an exercise to show that

~ ) = cos χ χ
M(χn̂ · Y + in̂ · ~σ sin , (4.33)
2 2

which enables us to conclude that

~ ) = (−1)m M(χn̂ · Y
M((χ + 2mπ)n̂ · Y ~ ). (4.34)

While in the case of the O(3) group, in which R(ϕ, θ, χ + 2mπ) =

R(ϕ, θ, χ), one concludes that elements M and −M in SU(2) correspond
to an element R in O(3). It is said that the SU(2) group is homomorphic
onto the O(3) group.
To obtain a deeper understanding of the homeomorphism of SU(2)
onto O(3), let us construct a 2 × 2 matrix X defined as follows

!
i x3 x1 − ix2
X = x σi = . (4.35)
x1 + ix2 −x3

A transformation in the coordinate system performed by the rotational

matrix R(ϕ, θ, χ) can equivalently be achieved by the unitary trans-
formation of the matrix X which is sandwiched in between the SU(2)
~ ) and its adjoint conjugate M† = exp(−χn̂ · Y
matrix M = exp(χn̂ · Y ~)
as follows
~ ~
X0 = M† XM = e−χn̂·Y Xeχn̂·Y . (4.36)

The above transformation leaves the norm of the vector invariant and is
justified by taking the determinant on both sides of the above equation,
i.e.

−x0i x0i = det X0 = (det M† )(det X)(det M) = −xi xi |det M|2 = −xi xi .
(4.37)

25:31.
128 Chapter 4

Let us take the rotation about the 3rd axis with angle ϕ as an
example and evaluate the matrix M = eϕY3 explicitly as follows

!
x03 x01 − ix02
X0 =
x01 + ix02 −x03

i i
! ! !
e− 2 ϕ 0 x3 x1 − ix2 e2ϕ 0
= i i
0 e2ϕ x1 + ix2 −x3 0 e− 2 ϕ
!
x3 e−iϕ (x1 − ix2 )
= . (4.38)
eiϕ (x1 + ix2 ) −x3

Equating the matrix elements on both sides, we obtain

x01 = x1 cos ϕ − x2 sin ϕ, (4.39)

x02 = x1 sin ϕ + x2 cos ϕ, (4.40)

x03 = x3 , (4.41)

and the coordinate transformation by rotation is reproduced.

4.2 O(3)/SU(2) algebras and angular momentum

Let us consider the Lie algebras of the O(3) group or the SU(2)
group, in which we take the generators as follows

~ ~
Ji = Xi , or Ji = Yi , (4.42)
i i

in order to be in conformity with the notations used in quantum me-

chanics. The algebra then has the following usual commutation relations

[Ji , Jj ] = i~kij Jk . (4.43)

25:31.
4.2 O(3)/SU(2) algebras and angular momentum 129

Let us now construct Cartan’s standard form of the algebra by solv-

ing the eigenvalue equation of the following type

[A, X] = ρX, (4.44)

where A is a linear combination of the generators with arbitrary choice

of the coefficients ai , and X which is also a linear combination of the
generators with the unknown coefficients xi to be determined in order
to satisfy the above eigenvalue equation.
Denote A and X as follows

A = ai Ji , X = xi Ji , (4.45)

and the eigenvalue equation becomes as

[A, X] = i~ai xj kij Jk = ρδjk xj Jk . (4.46)

Therefore we reach the following set of 3 homogeneous equations as

(i~ai kij − ρδjk )xj = 0. (4.47)

For the existence of the non-trivial solutions in xj , one imposes that

det(i~ai kij − ρδjk ) = 0, (4.48)

or more explicitly that

−ρ −i~a3 i~a2
i~a3 −ρ −i~a1 = 0. (4.49)
−i~a2 i~a1 −ρ

Hence, we obtain the 3 eigenvalues given as

µ0 = 0, (4.50a)
p
µ+ = ~ (a1 )2 + (a2 )2 + (a3 )2 , (4.50b)
p
µ− = −~ (a1 )2 + (a2 )2 + (a3 )2 . (4.50c)

25:31.
130 Chapter 4

It is the very example of 3-parameter Lie group of rank equaling to

one.
If we denote J0 , J+ and J− , the eigengenerators corresponding re-
spectively to the eigenvalues µ0 , µ+ and µ− , we find that J0 is propor-
tional to A. By setting

ai Ji A
J0 = p =p , (4.51)
1 2 2 2 3
(a ) + (a ) + (a )2 (a ) + (a2 )2 + (a3 )2
1 2

we obtain the following standard form for the commutator of J0 and J±

in Lie algebras of the O(3)/SU(2) group, [J0 , J± ] = ±~J± .
Let us compute the commutator of the generators J+ and J− , by
applying the Jacobi identity, one finds that

[[J+ , J− ], J0 ] = 0, (4.52)

which implies that the commutator of J+ and J− is proportional to J0 ,

i.e.

[J+ , J− ] = α~J0 = ~J0 , (4.53)

where we take α to be one due to the fact that the eigenvalue equation is
linear in X on both sides of the equation, that allows the eigengenerators
J+ and J− to absorb any arbitrary factor. Henceforth the Lie algebra
of the O(3)/SU(2) group is summarized as follows

[J0 , J± ] = ±~J± , (4.54)

[J+ , J− ] = ~J0 , (4.55)

with the hermiticity property given as

J0 = J0† , †
J+ = J− , †
J− = J+ . (4.56)

The calculation of the Casimir operator of the group, we shall leave

as an exercise, and it is found that

C = 2g ij Ji Jj . (4.57)

The factor 2 is introduced with the purpose to identify C with the total
angular momentum operator J 2 , i.e.

25:31.
4.2 O(3)/SU(2) algebras and angular momentum 131

C = J 2 = J+ J− + J− J+ + J02 . (4.58)
Since C commutes with the generators of the group, it is then invariant
under the transformation. Furthermore the eigenvalues of the Casimir
operator will be used to label the dimension of the irreducible represen-
tation of the group.
Let us construct the eigenvectors of both the operators C and J0 ,
and denote it as |c, νi, referred to commonly as the irreducible repre-
sentations of the O(3)/SU(2) group, hence we have
C|c, νi = c|c, νi, (4.59a)
J0 |c, νi = ν~|c, νi. (4.59b)
With the commutation relations,[J0 , J± ] = ±~J± , one is now ready
to show the following proposition.

Proposition 1.
If |c, νi is the eigenvector of the Casimir operator C and the op-
erator J0 with the eigenvalues c and ν~ respectively, then J± |c, νi
are also the eigenvectors of operators C and J0 with the eigen-
values c and (ν ± 1)~ respectively.

The vectors J± |c, νi can be easily proved to be also the eigenvectors

of operator C because [C, J± ] = 0, which leads to the following relation
CJ± |c, νi = J± C|c, νi = cJ± |c, νi. (4.60)
To show that J± |c, νi are the eigenvectors of the operator J0 with
the eigenvalues (ν ± 1)~, let us apply the operator J0 uponJ± |c, νi and
make use of the commutation relation of Eq. (4.53), then we find that
J0 J± |c, νi = (J± J0 ± ~J± )|c, νi = (ν ± 1)~J± |c, νi,
which imply that J± |c, νi are the eigenvectors of J0 corresponding to
the eigenvalues (ν ± 1)~ respectively. Proposition 1 is then established.
Let us now apply Proposition 1 to generate the eigenvectors |c, νi
as follows:
J+ |c, νi = α|c, ν + 1i, J− |c, νi = β|c, ν − 1i.

25:31.
132 Chapter 4

Therefore we are able to build a sequence of vectors by applying J+ and

J− on |c, νi consecutively, and obtain

|c, ν− i, . . . , |c, ν−qi, . . . , |c, ν−1i, |c, νi, |c, ν+1i, . . . , |c, ν+pi, . . . , |c, ν+ i.

The sequence must terminate somewhere on both ends, otherwise we

will encounter the dilemma of having the negative norm of the eigen-
vector. To see this, let us take the highest positive value ν of the eigen-
vector, such that the norm of the vector J+ |c, ν+ i vanishes, namely

1
hc, ν+ |J− J+ |c, ν+ i = hc, ν+ |(C − J02 − ~J0 )|c, ν+ i = 0,
2
or

c − ~2 ν+ (ν+ + 1) = 0, (4.61)

where we have made use of the relation J− J+ = C −J02 −~J0 . Therefore

one proves that the eigenvector |c, αi with α > ν+ , will lead to negative
norm. Similarly, J− |c, ν− i = 0, if we denote ν− to be the lowest value
of ν, and we have the relation

c − ~2 ν− (ν− − 1) = 0. (4.62)

Solving the above two quadratic equations to obtain values of ν+

and ν− , we take the difference of ν+ and ν− , which must be an integer,
because it is the total number of steps in applying J+ or J− to form the
nonvanishing eigenvectors in the sequence. Therefore we have

r
1 c
ν+ − ν− = 2 + 2 − 1 = 2j, (4.63)
4 ~

where j is taken as half integers. It also allows one to express the

eigenvalue of the Casimir operator as

c = j(j + 1)~2 , (4.64)

which helps one to calculate that

ν+ = −ν− = j. (4.65)

25:31.
4.2 O(3)/SU(2) algebras and angular momentum 133

It is the construction of the irreducible representation of the O(3)/SU(2)

group that enables us to quantize the dynamical observables C and J0 .
In order to express the eigenvectors of J 2 and J0 in the usual form
appearing in quantum mechanics, we shall replace vector |c, νi with the
familiar form of the vector |j, mi from now on, and it is normalized as
hj, m|j 0 , m0 i = δjj 0 δmm0 ,
with −j 6 m 6 j, −j 0 6 m0 6 j 0 .

One also has the following relations

~ p
J+ |j, mi = √ (j − m)(j + m + 1)|j, m + 1i, (4.66a)
2

~ p
J− |j, mi = √ (j + m)(j − m + 1)|j, m − 1i. (4.66b)
2

Different values of j form different irreducible representations of the

group. The 1-dimensional representation is a trivial one by taking j = 0.
For the case that j = 1/2, it forms a two-dimensional representation,
called the spinor representation. If we identify the vectors |1/2, 1/2i and
|1/2, −1/2i with the 2-component bases column matrices respectively as
follows
! !
1 1E 1 1 1E 0
, −→ , ,− −→ ,
2 2 0 2 2 1

then the matrix representation of the generators can be constructed

(1/2)
with the matrix element Jmm0 = h1/2, m|J|1/2, m0 i. We shall leave
them as an exercise again to show that

25:31.
134 Chapter 4

! ! !
(1) ~ 0 1 (1) ~ 0 0 (1) ~ 1 0
J+2 =√ , J−2 =√ , J0 2 = ,
2 0 0 2 1 0 2 0 −1

(4.67)

which allows us to cast the spin matrices as the following expression

!
1 (1) (1) ~ 0 1
Sx = √ (J+2 + J−2 ) = , (4.68a)
2 2 1 0
!
1 (1) (1) ~ 0 −i
Sy = √ (J+2 − J−2 ) = , (4.68b)
2i 2 i 0
!
(1) ~ 1 0
Sz = J0 2 = . (4.68c)
2 0 −1

In the case of the 3-dimensional representation, of which j = 1, the

generators can be calculated as follows:

 
0 1 0
~ 
Jx(1) = √ 1 0 1 , (4.69a)

2
0 1 0

 
0 −i 0
~ 
Jy(1) = √ i 0 −i , (4.69b)

2
0 i 0

 
1 0 0
Jz(1) = ~ 0 0 0 , (4.69c)
 

0 0 −1

which take completely different expression from the 3 generators of the

25:31.
4.3 Irreducible representations of O(3) group and spherical harmonics 135

rotational group of the Cartesian form obtained in previous Subsection

4.1. In fact the generators of 3-dimensional irreducible representation
are related to those of the Cartesian form through an unitary transfor-
mation which need to be elaborated more as an exercise by the readers.

4.3 Irreducible representations of O(3) group and spherical

harmonics

Let us consider the irreducible representations of the O(3) group

|l, mi, in which l takes only the positive integers l and m lies in between
l and −l, i.e. l = 1, 2, . . . and −l 6 m 6 l. Then the q-representation
of the vector |l, mi in the spherical coordinate system can be expressed
as hθ, ϕ|l, mi, which is exactly the spherical harmonics Ylm (θ, ϕ), i.e.
Ylm (θ, ϕ) = hθ, ϕ|l, mi. (4.70)

Instead of expressing the operators as J± and J0 , we shall denote the

orbital angular momentum operators by L± and L3 . It is algebraically
elegant to derive all the spherical harmonics of a given degree l and the
order m from the spherical harmonics of the same degree l but with
a lower order m − 1, namely building from Ylm (θ, ϕ) based upon the
results we have obtained in last section. We shall demonstrate these
relations in the following few propositions.

Proposition 2.

Let Ylm and Ylm−1 be two spherical harmonics of the same degree
l with the order m and m−1 respectively, then they are connected
by the following relation
Ylm (θ,ϕ)
√  
2 ∂ ∂
=p eiϕ
+ i cot θ Ylm−1 (θ, ϕ).
(l + m)(l − m − 1) ∂θ ∂ϕ
(4.71)

The proof is straightforward. Take the q-representation of the fol-

25:31.
136 Chapter 4

lowing equation
√
2
hθ, ϕ|l, mi = p hθ, ϕ|L+ |l, m − 1i,
(l + m)(l − m − 1)~

where if the q-representation of operator L+ is expressed by

 
iϕ ∂ ∂
L+ = e + i cot θ ~, (4.72)
∂θ ∂ϕ

then we arrive at the conclusion of Proposition 2.

Proposition 3.
All the spherical harmonics takes the form of the product of eimϕ ,
an exponential function of ϕ and another function of f (θ), i.e.
Ylm (θ, ϕ) = eimϕ f (θ). (4.73)

The proof goes by taking the q-representation of the equation as

follows
hθ, ϕ|L3 |l, mi = m~hθ, ϕ|l, mi.
Express the q-representation of the operator L3 by (~/i)∂/∂ϕ, then we
have the following differential equation

∂ m
Y (θ, ϕ) = imYlm (θ, ϕ),
∂ϕ l

and the solution of the above equation leads to Proposition 3, i.e.

Ylm (θ, ϕ) = eimϕ f (θ).
The property of the function f (θ) can be analyzed in the following
proposition.

25:31.
4.3 Irreducible representations of O(3) group and spherical harmonics 137

Proposition 4.
Function f (θ) satisfies the associated Legendre’s differential
equation, i.e.
m2
   
d 2 d
(1 − µ ) f (µ) + l(l + 1) − f (µ) = 0, (4.74)
dµ dµ 1 − µ2
or more explicitly by writing f (θ) in terms of the associated
Legendre polynomial of degree l and order m, i.e.
f (θ) ∝ Plm (cos θ) = Plm (µ), (4.75)
where we put µ = cos θ, and identify f (θ) = f (µ).

We start the proof by taking the following equation in q-representation,

namely
hθ, ϕ|L2 |l, mi = hθ, ϕ|l(l + 1)~2 |l, mi.
Denote the q-representation of the total angular momentum opera-
tor by

1 ∂2
   
1 ∂ ∂
L2 = −~2 sin θ +
sin θ ∂θ ∂θ sin2 θ ∂ 2 ϕ

which can be evaluated by means of coordinate transformation from

the Cartesian system to the spherical one. Then we reach the following
equation

1 ∂2
   
1 ∂ ∂
− sin θ + Ylm (θ, ϕ) = l(l + l)Ylm (θ, ϕ).
sin θ ∂θ ∂θ sin2 θ ∂ 2 ϕ

Putting µ = cos θ, the above equation can be reduced to the follow-

ing associated Legendre’s differential equation as

m2
   
d 2 d
(1 − µ ) f (µ) + l(l + 1) − f (µ) = 0,
dµ dµ 1 − µ2

25:31.
138 Chapter 4

and the solution of it is referred to as the associated Legendre polyno-

mials. Therefore Proposition 4 is proved, i.e.
f (µ) = APlm (µ).
Proposition 2 can be facilitated to construct the spherical harmonics
by avoiding the painstaking methods of solving the associated Legen-
dre’s differential equation. Ylm (θ, ϕ) can be formulated easily by succes-
sive applications of Proposition 2 from the spherical harmonics of lowest
order, namely from Yl−l (θ, ϕ). Let us first evaluate Yl−l (θ, ϕ) by taking
the q-representation of the following equation
hθ, ϕ|L− |l, −li = 0,
or
   
∂ ∂ ∂ ∂
− i cot θ Yl−l (θ, ϕ) = − i cot θ e−ilϕ f (θ)
∂θ ∂ϕ ∂θ ∂ϕ
 
−ilϕ ∂
=e − l cot θ f (θ) = 0,
∂θ

which leads to the solution of f (θ) as

f (θ) = A sinl θ.
The normalization constant A can be calculated by taking the inner
product as follows
Z Z π
hl, −l|l, −li = |Yl−l (θ, ϕ)|2 dΩ = 2π|A| 2
sin2l+1 θdθ = 1.
0

Hence A reads as
r
(2l + 1)! 1
A= .
4π 2l l!

Therefore we are able to express Yl−l (θ, ϕ) as follows

25:31.
4.3 Irreducible representations of O(3) group and spherical harmonics 139

r
(2l + 1)! 1 −ilϕ l
Yl−l (θ, ϕ) = e sin θ.
4π 2l l!

The spherical harmonics Ylm (θ, ϕ) can then be constructed by ap-

plying the L+ operator l + m times upon Yl−l (θ, ϕ), namely

s   l+m
(l − m)! ∂ ∂
Ylm (θ, ϕ) = eiϕ
+ i cot θ Yl−l (θ, ϕ)
(2l)!(l + m)! ∂θ ∂ϕ

s   l+m
1 (2l + 1)(l − m)! ∂ ∂
= l iϕ
e + i cot θ e−ilϕ sinl θ.
2 l! 4π(l + m)! ∂θ ∂ϕ

To simplify the calculation of the last term in the previous equation,

we apply the identity given in Proposition 5.

Proposition 5.
 
∂ ∂
Acting the operator eiϕ ∂θ + i cot θ ∂ϕ n times repeatedly upon
the function eipϕ f (θ) will result in another function of the fol-
lowing form, i.e.
  n
∂ ∂
eiϕ + i cot θ eipϕ f (θ)
∂θ ∂ϕ
 
n i(p+n)ϕ p+n d −p
= (−1) e sin θ sin θ f (θ).
d cos θ
(4.76)

It is just a pure algebraic manipulation to show the above identity.

Let us take the first step to calculate the following equation

25:31.
140 Chapter 4

 
iϕ ∂ ∂
e + i cot θ eipϕ f (θ)
∂θ ∂ϕ
 
i(p+1)ϕ d −p
= (−1)e sinp+1 θ sin θ f (θ). (4.77)
d cos θ

 
d −p
If we regard sinp+1 θ sin θ f (θ) as a new function of g(θ),
d cos θ

i.e.

 
d −p
g(θ) = sinp+1 θ sin θ f (θ),
d cos θ

and take the second step as follows

 
iϕ ∂ ∂
e + i cot θ ei(p+1)ϕ g(θ)
∂θ ∂ϕ
 
i(p+2)ϕ p+2 d −p−1
= (−1)e sin θ sin θ g(θ)
d cos θ

d2
 
2 i(p+2)ϕ p+2 −p
= (−1) e sin θ 2 sin θ f (θ) .
d cos θ

Repeating n times of the same operation, we have

  n
iϕ ∂ ∂
e + i cot θ eipϕ f (θ)
∂θ ∂ϕ
 
n i(p+n)ϕ p+n d −p
= (−1) e sin θ sin θ f (θ),
d cos θ

25:31.
4.4 O(4) group, dynamical symmetry and the hydrogen atom 141

which proves Proposition 5.

With all the preparations in the last few propositions, we are now
in the position to evaluate Ylm (θ, ϕ) by letting n = l + m, p = −l in
Proposition 5, then

s   l+m
(2l + 1)(l − m)! 1 ∂ ∂
Ylm (θ, ϕ) = eiϕ + i cot θ e−ilϕ sinl θ
4π(l + m)! 2l l! ∂θ ∂ϕ

s
(2l + 1)(l − m)! eimϕ m dl+m
= (−1) l+m
sin θ sin2l θ
4π(l + m)! 2l l! dl+m cos θ

s
(2l + 1)(l − m)! eimϕ 2 m dl+m
= (−1)l+m (1 − µ ) 2 (1 − µ2 )l
4π(l + m)! 2l l! dl+m µ

s
m (2l + 1)(l − m)! eimϕ m
= (−1) P (µ),
4π(l + m)! 2l l! l

where

(−1)l 2 m dl+m 2 m dm
Plm (µ) = (1 − µ ) 2 (1 − µ 2 l
) = (−1)l
(1 − µ ) 2 Pl (µ),
2l l! dl+m µ dm µ

1 dl
Pl (µ) = (1 − µ2 )l ,
2l l! dl µ

and Rodrigues’ formulas for Legendre polynomials Pl (µ) as well as for

the associated Legendre functions Plm (µ) are reproduced.

4.4 O(4) group, dynamical symmetry and the hydrogen

atom

Let us investigate the orthogonal group in 4-dimensional Euclidean

space, denoted by the O(4) group, in which the norm of the vector,

25:31.
142 Chapter 4

xi xi = invariant for i sums from 1 to 4,

under the orthogonal transformation.

The generators of the group in canonical form can be easily con-
structed by means of rotation on the (xi , xj )-plane similar to the case of
the O(3) group as we have done before, namely, they are the 6 generators
given as

∂ ∂
xi j
− xj i , for i, j = 1, 2, 3, 4.
∂x ∂x

If we denote

∂ ∂ ∂ ∂ ∂ ∂
M1 = x 2 − x3 2 , M2 = x3 − x1 3 , M3 = x1 − x2 1 ,
∂x3 ∂x ∂x1 ∂x ∂x2 ∂x

and

∂ ∂ ∂ ∂ ∂ ∂
N1 = x1 4
− x4 1 , N 2 = x2 4
− x4 2 , N 3 = x3 4
− x4 3 ,
∂x ∂x ∂x ∂x ∂x ∂x

then the Lie algebra of the O(4) group reads as

[Mi , Mj ] = −kij Mk , [Mi , Nj ] = −kij Nk , [Ni , Nj ] = −kij Mk . (4.78)

The above relations enable us to redefine the generators by the fol-

lowing combinations:

1 1
Ai = (Mi + Ni ), Bi = (Mi − Ni ), (4.79)
2 2

and the algebra for these redefined generators become

[Ai , Aj ] = −kij Ak , [Ai , Bj ] = 0, [Bi , Bj ] = −kij Bk . (4.80)

Among the Lie algebra of the O(4) group, we find that there exist
two sets of generators Ai and Bi whose commutators respectively form

25:31.
4.4 O(4) group, dynamical symmetry and the hydrogen atom 143

a sub-algebra of the O(3) group. The O(4) algebra becomes the direct
sum of two O(3) algebras and the O(4) group is then locally isomorphic
onto O(3) ⊗ O(3) group.
We are now in position to discuss the hydrogen-like atoms from
the dynamical symmetry properties. The Hamiltonian operator for the
hydrogen-like atoms takes the following expression

1 2 Ze2 1 2 k
H= P − = P − , (4.81)
2m r 2m r

where we have fixed the center of the Coulomb potential at the origin.
The space described by this coordinate system is isotropic with respect
to the rotation about any axis through the origin. It implies that the
angular momentum operators are conserved because they commute with
the Hamiltonian operator. Furthermore, the attractive Coulomb poten-
tial enables us to introduce another three conserved operators, called
Lenz operators. Classically, it is the vector, called the Lenz vector or
the Runge-Lenz vector, that originates from the center of the force and
points to the aphelion, as shown in Figure 4.2. It is a particular feature
that the orbit is closed and fixed in space for the case of the Coulomb
potential. If we construct a plane perpendicular to the constant angular
momentum through the origin, the closed orbit lies on the plane without

Figure 4.2: Hydrogen-like orbit.

25:31.
144 Chapter 4

precession of the major axis. Then the Lenz vector can be constructed
by means of the following elliptical orbit equation, i.e.

1 mk
= 2 (1 −  cos θ), (4.82)
r l

where l is the angular momentum of the system and  is the eccentricity

of the ellipse given as
r
2El2
= 1+ , with E < 0.
mk 2

Let us assume the Lentz vector

~ = α~r + β~l × p~.
R
~ we reach the following relations
By taking ~r · R,

l2 l2
 
~ = rR cos θ = r 1
~r · R 1− = αr2 − βl2 ,
mk(1 − )  mkr

or

l2 l2
   
αr − r − l2 β − 2 2
= 0,
mk(1 − ) m k (1 − )

that allows us to obtain

1 l2 l2
α= , β= ,
r mk(1 − ) m2 k 2 (1 − )

which leads to the classical Lenz vector as

l2
 
~ = ~r 1 ~
R + l × p~ . (4.83)
mk(1 − ) r mk

25:31.
4.4 O(4) group, dynamical symmetry and the hydrogen atom 145

To obtain the quantum Lenz operators, we replace the classical dy-

namical variables by the equivalent quantum operators in the above
equation. Then we have
( )
l2 Xi 1
Ri = p − ijk (Pj Lk − Lj Pk ) .
mk(1 − ) X12 + X22 + X32 2mk
(4.84)

If we take the commutator of Ri and Rj , though it is a tedious

computation, we find that

2H 1
[Ri , Rj ] = − i~ijk Lk .
mk  (1 − )2
2 2

By rescaling the operators in the following expression

r
mk 2
R̃i = (1 − )Ri ,
−2H

the Lie algebra of the O(4) group is regained, i.e.

[Li , Lj ] = i~ijk Lk , (4.85a)

[Li , R̃j ] = i~ijk R̃k , (4.85b)

[R̃i , R̃j ] = i~ijk Lk . (4.85c)

The dynamical symmetry of the Coulomb potential can also be ap-

plied to calculate the eigenenergy of the hydrogen like atoms. Let us
denote

1 1
Ji = (Li + R̃i ), Ki = (Li − R̃i ), (4.86)
2 2

and these operators are the generators of O(3) ⊗ O(3) group, namely,

25:31.
146 Chapter 4

[Ji , Jj ] = i~ijk Jk , (4.87a)

[Ji , Kj ] = 0, (4.87b)
[Ki , Kj ] = i~ijk Kk , (4.87c)

with the property J 2 = K 2 because of the relation Li R̃i = 0. Let us

calculate J 2 + K 2 , which is related to the Hamiltonian as follows

mk 2
 
2 1 2 2
J +K =− +~ , (4.88)
2 2H

or

mk 2
 
1
H=− . (4.89)
2 2(J 2 + K 2 ) + ~2

Since the hydrogen atom is subject to the condition J 2 = K 2 , the

eigenstates of it are then taken as the direct product of the irreducible
representations of O(3) ⊗ O(3) as follows
|j, mi ⊗ |k, m0 ij=k ≡ |(j, m); (k, m0 )ij=k = |(j, m); (j, m0 )i. (4.90)

They are also the eigenstates of Hamiltonian (why?), i.e.

H|(j, m); (j, m0 )i = E|(j, m); (j, m0 )i, (4.91)

with the eigenvalues

mk 2 1 mk 2 1
E=− 2 2
= − , (4.92)
2 4j(j + 1)~ + ~ 2 (2j + 1)2 ~2

or in terms of the principal quantum number n = 2j + 1, we reach the

well known eigenenergy of the hydrogen atom,

2
mk 2 1 mc2 e2 mc2 2 1

1
En = − 2 2 = − = − α 2,
2~ n 2 ~c n2 2 n

25:31.
4.5 Exercises 147

where α is the fine structure constant.

4.5 Exercises

Ex 4.5.1

Show that SU(2) matrix M (χn̂ · Y ~ ) = exp(χ i n̂ · ~σ ) corresponding to

2
a rotation about axis n̂ with an angle χ takes the following form,
M (χn̂ · Y~ ) = cos χ + in̂ · ~σ sin χ .
2 2

Ex 4.5.2
Taking σ i = σi and making use of the following properties of Pauli
matrices,

(a) hermiticity σ i† = σ i ,

(b) traceless Tr σ i = 0,

(c) {σ i , σ j } = 2δ ij I,

(d) Tr σ i σ j = 2δ ij ,

(e) det σ i = −1,

show that the rotational matrix element Rji is given by

1
Rji = Tr (σ i M † σj M ).
2

(Hint: by Eq. (4.36).)

Ex 4.5.3
Construct explicitly the matrix representations of the generators in
two dimensions.

25:31.
148 Chapter 4

Ex 4.5.4
Denote the 2 × 2 traceless matrices A and B by arbitrary parameters
as A = ~a · ~σ , B = ~b · ~σ . Show that AB = ~a · ~bI + i(~a × ~b) · ~σ .

Ex 4.5.5
Show that

ϕ)~σ U† (~
U(~ ϕ) = (n̂ · ~σ )n̂ + n̂ × ~σ sin ϕ − n̂ × (n̂ × ~σ ) cos ϕ,

ϕ) = exp(i ϕ2 n̂ · ~σ ).
where U(~

Ex 4.5.6
Show that the Casimir operator of the rotational group takes the
expression

(
Ji Ji in the Cartesian basis,
C= 2
J+ J− + J− J+ + J0 in the standard form.

Ex 4.5.7
Find the unitary matrix U which diagonalizes

   
0 −1 0 1 0 0
~ (1)
Jz(c) = 1 0 0 into J0 = ~ 0 0 0 ,
  
i
0 0 0 0 0 −1

and hence verify explicitly that

   
0 1 0 0 −i 0
~  ~ 
Jx(1) = √ 1 0 1  , Jy(1) = √ i 0 −i .
 
2 2
0 1 0 0 i 0

25:31.
4.5 Exercises 149

Ex 4.5.8
Show that [H, Ri ] = O for the hydrogen atom.

Ex 4.5.9
Prove that

2H 1
[Ri .Rj ] = − i~ijk Lk .
mk  (1 − )2
2 2

Ex 4.5.10
Prove that

mk 2
 
2 1 2 2
J +K =− +~ .
2 2H

(Hint: use the fact that [H, Ji ] = [H, Ki ] = O.)

25:31.
This page intentionally left blank

25:31.
Chapter 5

Lorentz Transformation, O(3,1)/SL(2,C) Group

and the Dirac Equation

5.1 Space-time structure and Minkowski space

Space and time had been treated separately since the birth of physics
as an experimental science in mediaeval time. As Newton stated in his
work, The Principia, absolute space, in its own nature, without relation
to anything external, remains always similar and immovable. Absolute,
true, and mathematical time, of itself, and from its own nature, flows
equably without relation to anything external. Physicists had been
confined by the concept of the absoluteness of space and time such
that the coordinate transformations from one inertial frame to another,
always kept time as an invariant parameter.
All this concept of absolute space and time resulted in the formu-
lation of Galilean relativity which has not been challenged until the
end of 19th century and the beginning of the 20th century. It was
the triumphal results that related the wave phenomena of light to the
electromagnetic theory in Maxwell equations, that space and time were
interwoven into the formulation of the theory and were treated as equals.
The development in special relativity afterwards led naturally to gen-
eralize the 3-dimensional coordinate space to the 4-dimensional one by
including the time-axis, an extra dimension. And so it is no longer the
Euclidean vector space, but a four-dimensional Minkowski space with
metric tensor defined as

35:17.
152 Chapter 5

gµν = 0 if µ 6= ν ; − g00 = g11 = g22 = g33 = 1. (5.1)

The zero-th axis is defined as the time axis, and x0 = ct. A vec-
tor in Minkowski space, also called a 4-vector, is denoted by x =
(x0 , x1 , x2 , x3 ) to specify the time and position. In order to avoid using
the term “norm”, we define the length square of the vector as
x2 = gµν xµ xν = −(x0 )2 + (x1 )2 + (x2 )2 + (x3 )2 . (5.2)
This is a postulate in special relativity that the laws of physics are
invariant under a homogeneous linear transformation on the space-time
4-vector if the transformation leaves the length square of a 4-vector in-
variant. This chapter is devoted to the exploration of the properties of
these transformations as well as their applications in relativistic quan-
tum wave equations.

5.1.1 Homogeneous Lorentz transformation and SO(3,1) group

Let us consider the following linear transformation in the matrix
form
x 7−→ x0 = Λx, (5.3)
where x and x0 are column matrices, and Λ is a 4 × 4 square matrix.
If we denote the Minkowski metric tensor also in matrix form as
 
−1 0 0 0
 
 0 1 0 0
gµν =
 0
, (5.4)
 0 1 0
0 0 0 1

the length square of the 4-vector x becomes

xT gx = −(x0 )2 + (x1 )2 + (x2 )2 + (x3 )2 . (5.5)
The invariance in the square length of x due to the isotropy prop-
erty of space-time restrains the Lorentz transformation matrix Λ to the
following conditions
x0T gx0 = xT ΛT gΛx = xT gx, (5.6)

35:17.
5.1 Space-time structure and Minkowski space 153

namely that

ΛT gΛ = g or ΛT g = gΛ−1 . (5.7)

This is the crucial difference between the Euclidean space and the
Minkowski space because the g matrix is involved in the latter case.
If det Λ 6= 0, then the transformation forms a group, called the ho-
mogeneous Lorentz group. If we take the determinant on both sides
of Eq. (5.7), we obtain (det Λ)2 = 1. We shall restrict ourselves to the
case det Λ = 1 with which the group elements can be reached continu-
ously from the identity element. The length invariance will provide 10
equations of constraints among the matrix elements of Λ, namely

Λµα gµν Λνβ = gαβ . (5.8)

Therefore there remain 6 independent parameters in matrix Λ. Fur-

thermore for the case α = β = 0, the constraint reads as
X
(Λ00 )2 − (Λi0 )2 = 1, (5.9)
i

which provides two choices, i.e. Λ00 > 1 and Λ00 6 1.

If all the group elements are continuously connected to the identity
element, Λ00 > 1 should be chosen, and the group is called proper
Lorentz group, or simply the Lorentz group, or the SO(3,1) group, to
which we shall confine ourselves for further investigation.
To construct the generators of the group, let us consider the boost
Lorentz transformation, in which one spatial coordinate system O0 is
travelling with velocity in a particular direction, say along the x01 -axis,
with respect to another spatial coordinate system O along the direction
of x1 -axis, of which both origins of the coordinate frames coincide at
x0 = x00 = 0. Then the boost matrix takes the following familiar
expression:

x00 = γx0 − βγx1 , (5.10a)

x01 = −βγx0 + γx1 , (5.10b)

35:17.
154 Chapter 5
p
where β = v/c and γ = 1/ 1 − β 2 .
It is interesting to visualize the above boost transformation as a
rotation on the (x0 , x1 ) plane with an imaginary angle, or hyperbolic
angle ξ defined by

ξ = tanh−1 β. (5.11)

Then the boost matrix becomes

 
cosh ξ − sinh ξ 0 0
 
− sinh ξ cosh ξ 0 0
Λ=
 0
. (5.12)
 0 1 0
0 0 0 1

The matrix representation of the generator for the boost transfor-

mation can then be easily obtained by taking the derivative of the above
matrix with respect to the group parameter ξ at the origin of the pa-
rameter space, i.e.

 
0 −1 0 0
 
−1 0 0 0
B1 = 
 . (5.13a)
 0 0 0 0
0 0 0 0

The other two generators of the boost transformation in x2 -axis and

x3 -axis can also be constructed with similar procedures. They read as:

   
0 0 −1 0 0 0 0 −1
   
 0 0 0 0  0 0 0 0
B2 = 
−1
, B3 =  . (5.13b,c)
 0 0 0
 0
 0 0 0
0 0 0 0 −1 0 0 0

35:17.
5.1 Space-time structure and Minkowski space 155

The remaining three generators of the SO(3,1) group are merely those
of the spatial rotational ones, expressed as:

 
0 0 0 0
 
0 0 0 0
A1 = 
0
 (5.14a)
 0 0 1
0 0 −1 0

   
0 0 0 0 0 0 0 0
   
0 0 0 −1 0 0 1 0
and A2 = 
0
, A3 = 
0 −1 0
. (5.14b,c)
 0 0 0  0
0 1 0 0 0 0 0 0

This allows us to obtain the Lie algebra of the SO(3,1) group as

follows:

[Ai , Aj ] = −kij Ak , [Ai , Bj ] = −kij Bk , (5.15)

but the commutators among the boost generators Bi take the different
sign to close the algebra, i.e.

[Bi , Bj ] = kij Ak . (5.16)

The algebra of the generators Ai closes by themselves, they form a

sub algebra of SO(3,1), which implies that SO(3,1) contains SO(3), the
rotational transformation, as its subgroup.
As in the case of rotational transformation, the Lorentz boost matrix
Λ of the relative motion along the x1 - axis can also be regained by
exponentiating the matrix generator B1 , namely

 
cosh ξ − sinh ξ 0 0
 
− sinh ξ cosh ξ 0 0
eξB1 =
 0
. (5.17)
 0 1 0
0 0 0 1

35:17.
156 Chapter 5

Therefore the group elements of the SO(3,1) can be expressed in the

general forms as
~ ~ ~ ~
Λ = eθ·A+ξ·B . (5.18)

Since the group parameters ξ = (ξ 1 , ξ 2 , ξ 3 ) are unbounded, hence the

Lorentz group, or the SO(3,1) group, is a non-compact Lie group.
The algebra of the SO(3,1) group can also be constructed by means
of the usual canonical formulation, in which we denote the contravariant
tensor generators as

∂ ∂
M µν = xµ − xν , (5.19)
∂xν ∂xµ

and one will immediately confirm that

[M µν , M αβ ] = −g νβ M µα − g µα M νβ + g να M µβ + g µβ M να . (5.20)

It is convenient to cast the algebra in terms of Ai and Bi , which will

reproduce the SO(3,1) algebra by the following identification

1
Bi = gij M 0j , Ai = ijk M jk . (5.21)
2

5.2 Irreducible representation of SO(3,1) and

Lorentz spinors

As in the case of the SO(4) group, let us take the linear combinations
of the generators of the SO(3,1) group and denote that

   
1 Ai 1 Ai
Li = + Bi , and Ri = − Bi . (5.22a,b)
2 i 2 i

Then the Lie algebra of the SO(3,1) group takes the expressions as those
of the SO(3) ⊗ SO(3) group, namely, i.e.

35:17.
5.2 Irreducible representation of SO(3,1) and Lorentz spinors 157

[Li , Lj ] = ikij Lk , (5.23a)

[Li , Ri ] = 0, (5.23b)

[Ri , Rj ] = ikij Rk . (5.23c)

Yet they are not exactly parallel to the case of the SO(4) group,
because the generators Li and Ri of the non-compact Lorentz group are
not Hermitian. The immediate consequence of these properties gives
rise to two options we have to face. Either we choose the finite dimen-
sional representations, in which case the unitarity condition shall be
abandoned or we choose to preserve the unitarity of the representation
by accepting the infinite dimensional representations.
Let us consider the finite dimensional irreducible representation, and
take the direct product |l, mil ⊗ |r, nir ≡ |l, m; r, ni as the bases, where
|l, mil is the eigenvector of the commuting operators L2 and L3 with the
eigenvalues l(l + 1) and m respectively, while |r, nir is the eigenvector
of another pair of commuting operators R2 and R3 with the eigenvalues
r(r + 1) and n respectively, such that

−l 6 m 6 l, − r 6 n 6 r, and l, r = half integers. (5.24)

Consider the simplest case of one dimension in which we take l =

r = 0, and label the representation by the symbol (0, 0). The matrix
representations of the generators are of one dimension, namely a 1 × 1
matrix with element as

h0, 0; 0, 0|Li |0, 0; 0, 0i = h0, 0; 0, 0|Ri |0, 0; 0, 0i = 0, (5.25)

and we have the trivial representation of the group, identity.

There are two distinct 2-dimensional representations, namely (1/2, 0)-
representation and (0, 1/2)-representation. We shall leave them as an
exercise for the readers to calculate that the generators shall take the
following forms:

35:17.
158 Chapter 5

( 1 ,0) 1 ( 1 ,0)
Li 2 = σi , Ri 2 = 0, (5.26a)
2

(0, 12 ) (0, 12 ) 1
Li = 0, Ri = σi , (5.26b)
2

which allow us to obtain

( 1 ,0) i ( 1 ,0) 1
Ai 2 = σi , Bi 2 = σi , (5.27a)
2 2

(0, 12 ) i (0, 12 ) 1
Ai = σi , Bi = − σi , (5.27b)
2 2

and obtain finally the 2-dimensional irreducible representation of the

Lorentz group as follows
 
( 21 ,0) ~ ξ)
~ = exp i
D (θ, ~σ · (θ~ − iξ)
~ , (5.28a)
2
 
(0, 12 ) ~ ξ)
~ = exp i ~ ~
D (θ, ~σ · (θ + iξ) . (5.28b)
2

As a quick check out of curiosity, one finds that

 
( 12 ,0)† ~ ~ i ~ ~ 1
~ −1 ,
~ ξ)
D (θ, ξ) = exp − ~σ · (θ + iξ) 6= D( 2 ,0) (θ,
2

or
 
(0, 21 )† ~ ~ i ~ ~ 1
~ −1 ,
~ ξ)
D (θ, ξ) = exp − ~σ · (θ − iξ) 6= D(0, 2 ) (θ,
2

namely that unitarity is lost in the finite dimensional irreducible repre-

sentation.

35:17.
5.2 Irreducible representation of SO(3,1) and Lorentz spinors 159

With the aim of investigating the relativistic wave equation of

spin 12 system, we shall limit ourselves by not going into further explo-
ration of the higher dimensional representations.
Let us perform the following identifications:

! !
1 1 1 1 1 0
| , il 7−→ = e1 , | , − il 7−→ = e2 , (5.29a,b)
2 2 0 2 2 1

as the two bases of the two-dimensional complex manifold, or V( 1 ,0) -

2

space, and construct the left-handed Lorentz spinor, or simply the left-
handed spinor as follows

!
ψl1 (x)
ψl (x) = ψla (x)ea = , (5.30)
ψl2 (x)

where ψl1 (x) or ψl2 (x) are all complex functions of space-time coordinate
x = (x0 , x1 , x2 , x3 ). Under the Lorentz transformation, the left-handed
spinor is transformed by applying the (1/2, 0) matrix representation of
the SO(3,1) group, i.e. Eq. (5.28a) upon the spinor as follows
1
ψl (x) 7−→ ψl0 (x0 ) = D( 2 ,0) (θ, ~ l (Λ−1 x0 ),
~ ξ)ψ (5.31)

where x0 = Λ−1 x implies that active transformation of the quantum

states equivalent to a passive Lorentz transformation of the space-time
vector. It should be emphasized that not only the components of the
spinor are transformed by the (1/2, 0) matrix , but that the space-time
coordinates in the argument of the complex functions ψl (x) are also
simultaneously transformed accordingly.
As for the Lorentz transformation of the right-handed spinor, one
finds that it can be expressed as

! ! !
1 0 ψr1 (x)
ψr (x) = ψra (x)fa = ψr1 (x) + ψr2 (x) = , (5.32)
0 1 ψr2 (x)

35:17.
160 Chapter 5

where ψra (x)(a = 1, 2) are the complex functions of the space-time co-
ordinates, and fa (a = 1, 2) are the bases column matrix of the right-
handed spinor space. Similar to the transformation in the left-handed
spinor space, the Lorentz transformation for ψr (x) can then be written
as
1
ψr (x) 7−→ ψr0 (x0 ) = D(0, 2 ) (θ, ~ r (Λ−1 x0 )).
~ ξ)ψ (5.33)

5.3 SL(2,C) group and the Lorentz transformation

1
~ ξ)
It is interesting to observe that both matrices D( 2 ,0) (θ, ~ and
1
~ ξ)
D(0, 2 ) (θ, ~ play more roles than just performing the Lorentz trans-
formation upon the spinor wave functions. Let us consider the linear
transformation on a two-dimensional complex vector space, in which a
vector ξ is acted upon by a 2×2 matrix L with complex matrix elements
as follows

! ! ! !
0 ξ 01 ξ1 a b ξ1
ξ = =L 2 = . (5.34)
ξ 02 ξ c d ξ2

Let us impose a condition such that

a b
det L = = ab − bc = 1, (5.35)
c d

which reduces the 8 parameters to 6 independent ones. Matrices with

the properties of such specification form a group of SL(2,C). If we ex-
press the SL(2,C) matrix by exponentiating an arbitrary 2 × 2 matrix
A, namely

L = eA , (5.36)

then we have the following proposition

35:17.
5.3 SL(2,C) group and the Lorentz transformation 161

Proposition 1.
If a matrix L can be expressed as L = eA , then
det L = eTr A
. (5.37)

Proof of the proposition is straightforward: we find the eigenvec-

tors of the matrix A as the new bases that form the matrix S, which
diagonalizes the matrix A, i.e.
A0 = S −1 AS. (5.38)

If we insert the matrix L between S −1 and S, we obtain the following

relation
! 1
!
S −1 AS A0 a1 0 ea 0
S −1 LS = e =e = exp = 2 . (5.39)
0 a2 0 ea

Therefore we conclude that

= eTr A0
= eTr S −1 AS
= eTr
1 +a2
det L = det (S −1 LS) = ea A
.
(5.40)

Let us go back to the two dimensional irreducible representations

1
of the Lorentz group, the matrix D( 2 ,0) (θ, ~ and matrix D(0, 12 ) (θ,
~ ξ) ~ ξ).
~
They are in fact the SL(2,C) matrices because
~ ξ)~ ~ ξ)~
= eTr
1
det D( 2 ,0) (θ, ~ = det e 2i ~σ·(θ−i
~ ξ) i
2
σ ·(θ−i
~
= 1, (5.41)

and similarly
~ ξ)~ ~ ξ)~
= eTr
1
det D(0, 2 ) (θ, ~ = det e 2i ~σ·(θ+i
~ ξ) i
2
σ ·(θ+i
~
= 1. (5.42)

The 6 group parameters are designated as θ~ = (θ1 , θ2 , θ3 ) and ξ~ =

(ξ 1 , ξ 2 , ξ 3 ),
which enable us to formulate the generators of the SL(2,C)
group in the following matrix representations, i.e.

35:17.
162 Chapter 5

i 1 i 1
Ai = σi , Bi = σi or Ai = σi , Bi = − σi , (5.43a,b)
2 2 2 2

with the same Lie algebra of the SO(3,1) group, i.e.

[Ai , Aj ] = −kij Ak , (5.44a)

[Ai , Bj ] = −kij Bk , (5.44b)

[Bi , Bj ] = kij Ak . (5.44c)

We conclude therefore that SL(2,C) is an isomorphism onto the SO(3,1)

group. To demonstrate the relation between the two groups with respect
to Lorentz transformation, we introduce the space-time matrix

!
µ −x0 + x3 x1 − ix2
X = x σµ = , (5.45)
x1 + ix2 −x0 − x3

where σ 0 = −σ0 = I, a unit matrix, and σi are the Pauli matrices. The
length of the space-time vector is related the determinant of the matrix,
i.e.

det X = (x0 )2 − ~x2 . (5.46)

Therefore the transformation of X into X 0 by multiplying SL(2,C)

1
matrices, D( 2 ,0) (θ, ~ and D( 12 ,0)† (θ,
~ ξ) ~ ξ)
~ on both sides of X in the follow-
ing construction
1 1
X 0 = D( 2 ,0) (θ,
~ ξ)XD
~ ( 2 ,0)† ~ ~
(θ, ξ), (5.47)

will leave the length of the vector x = (x0 , x1 , x2 , x3 ) invariant because

det X 0 = det X. (5.48)

35:17.
5.4 Chiral transformation and spinor algebra 163

Take a specific example of a Lorentz boost transformation along the

x3 -axis, in which θ~ = (0, 0, 0), and ξ~ = (0, 0, ξ), then we obtain

1 1
! !
0 1
σ ξ 1
σ ξ e2ξ 0 e2ξ 0
X =e 2 3 Xe 2 3 = − 12 ξ
X − 12 ξ
. (5.49)
0 e 0 e

Therefore we recover the usual Lorentz transformation by boosting the

coordinate frame along the x3 -axis if we compare the matrix elements
on both sides of the last equation and obtain the following relations:
x00 = cosh ξx0 − sinh ξx3 , (5.50a)

x01 = x1 , (5.50b)

x02 = x2 , (5.50c)

x03 = − sinh ξx0 + cosh ξx3 . (5.50d)

5.4 Chiral transformation and spinor algebra

As we have mentioned before, physics involves two distinct types of

transformations, the continuous transformations and the discrete ones.
Here in this section, we shall introduce another discrete symmetry trans-
formation. In addition to parity, charge conjugation or time rever-
sal transformations in quantum mechanics, the chiral transformation
is among the least discussed topics. The word “chirality” was derived
from the Greek χειρ, which means hand. Henceforth chirality is gen-
erally recognized as synonym to handedness in the geometric sense. It
is a discrete transformation from the right-handed irreducible repre-
sentation to the left-handed one of the Lorentz group, and vice versa.
In contrast to the usual parity transformation, an antilinear and an-
tiunitary operator is required in the chiral transformation in order to
obtain the self consistent theory, such that the Lorentz transformation
can equivalently be performed in the left-handed frame of reference as
well as the right-handed one. It also provides the interrelation between
the left-handed spinors and the right-handed spinors.

35:17.
164 Chapter 5

Let us denote the chiral operator K, an antilinear operator and for

each vector ψ, there exists a vector aψ + bϕ such that

K(aψ + bϕ) = a∗ Kψ + b∗ Kϕ, (5.51)

where a and b are complex numbers. The antiunitarity of K is defined

for each pair of vectors ψ and ϕ, such that the inner product takes the
following relation

(Kψ, Kϕ) = (ϕ, ψ) = (ψ, ϕ)∗ . (5.52)

The chiral transformation is neither the parity transformation on

the spatial coordinate system nor is it the time reversal transformation
on the temporal axis in Minkowski space. The six generators of the
Lorentz group therefore are invariant under the chiral transformation,
namely the rotational generators as well as the Lorentz boost generators
remain unchanged under K, i.e.

KAi K−1 = Ai , KBi K−1 = Bi . (5.53)

Yet the new set of the group generators Li and Ri are transformed
according to the following relations

 
−1 1 Ai
KLi K = K + Bi K−1 = −Ri , KRi K−1 = −Li , (5.54a,b)
2 i

because of the antilinearity property of operator K. Since chiral trans-

formation on L2 becomes R2 and reciprocally it holds for R2 , namely
the chiral transformation will interchange operator L2 and operator R2 ,
therefore we have the following proposition:

Proposition 2.
The vector KLjm is the eigenvector of R2 and R3 with the eigen-
values j(j + 1) and −m respectively. While the vector KRkn is
the eigenvector of L2 and L3 with the eigenvalues k(k + 1) and
−n respectively.

35:17.
5.4 Chiral transformation and spinor algebra 165

The proof goes as follows: let us use the brief notations Ljm , Rkn
instead of |j, mil and |k, nir respectively, and consider R2 KLjm =
KL2 Ljm = j(j + 1)KLjm , and R3 KLjm = −KL3 Ljm = −mKLjm
which implies that KLjm = γ(m)Rj,−m . Similarly that KRkn = δ(n)×
Lk,−n .
The antiunitarity of the operator K allows us to find the coefficients
γ(m) and δ(n) as follows
(KLjm , KLjm0 ) = γ ∗ (m)γ(m0 )(Rj,−m , Rj,−m0 ) = (Ljm0 , Ljm ), (5.55)

or γ ∗ (m)γ(m0 )δ−m,−m0 = δm0 m , that leads to |γ(m)|2 = 1.

The same argument can be applied to the right-handed case to ob-
tain |δ(n)|2 = 1. It is of interest to observe that the chiral transforma-
tion will also make the connection for the irreducible representation of
the left-handed ones to the right-handed ones. Let us take the spinor

2
( 1 ,0)
representation as an example by considering the matrix elementDmm 0

of the left-handed Lorentz transformation,

( 1 ,0) θ·A+ξ·B ~ ~ ~ ~ ~ ~ ~ ~
2
Dmm 0 = (L 1 ,m , e L 1 ,m0 ) = (Keθ·A+ξ·B L 1 ,m0 , KL 1 ,m )
2 2 2 2

~ ~ ~ ~ (0, 1 )∗
= γ ∗ (m0 )γ(m)(eθ·A+ξ·B R 1 ,−m0 , R 1 ,−m ) = γ(m)D−m,−m
2 ∗ 0
0 γ (m ),
2 2

or it can be cast into the matrix form as follows

! !
( 12 ,0) 0 γ( 12 ) 1 0 γ ∗ (− 21 )
D = D(0, 2 )∗ . (5.56)
1
γ(− 2 ) 0 γ ∗ ( 12 ) 0

Similarly the right-handed Lorentz matrix is related to the left-handed

one by

! !
(0, 12 ) 0 δ( 12 ) 1 0 δ ∗ (− 21 )
D = D( 2 ,0)∗ . (5.57)
1
δ(− 2 ) 0 δ ∗ ( 12 ) 0

35:17.
166 Chapter 5

By choosing the phase factors γ( 12 ) = δ( 12 ) = −γ(− 12 ) = −δ(− 21 ) = 1,

!
0 1
and defining a 2 × 2 matrix  as , one can easily prove for
−1 0
i = 1, 2 and 3 that
! !
0 1 0 −1
σi∗ −1 = σi∗ = −σi , (5.58)
−1 0 1 0

which allows one to reciprocate the following relation:

1 1
D(0, 2 ) = D( 2 ,0)∗ −1 , (5.59a)

1 1
D( 2 ,0) = D0,( 2 )∗ −1 , (5.59b)

We are now in the position to investigate further the connection

between the SL(2,C) transformation and the Lorentz transformation
with spinor algebra. Let us construct a dual space V( 1̇ ,0) to the spinor
2
space V( 1 ,0) , then V( 1̇ ,0) is called the co-left-handed spinor space, a
2 2

2-dimensional complex vector space with the basis ėa (a = 1, 2) which

can be identified with the row matrix as follows
ė1 = (1, 0), ė2 = (0, 1), (5.60a,b)

where a new set of notations are introduced in the subscript indices in

order to differentiate the co-left-handed spinor to the right-handed one.
A spinor in V( 1̇ ,0) is then expressed by
2

ψ̇ = ψ̇a ėa = (ψ̇1 , ψ̇2 ), (5.61)

where ψ̇ is related to the components of the left-handed spinor ψ =

(ψ 1 , ψ 2 )T by the following equation
ψ̇a = ψ b ba , (5.62)

or in matrix notation as
ψ̇ = ψ T T = (ψ̇1 , ψ̇1 ) = (ψ 1 , ψ 2 )T = (ψ 1 , ψ 2 )−1 . (5.63)

35:17.
5.4 Chiral transformation and spinor algebra 167

The Lorentz transformation on the co-left-handed spinor can be

found to be
L.T. 1
ψ̇ 7−→ ψ̇ 0 = ψ 0T −1 = ψ T DT ( 2 ,0) −1

1 1
= ψ̇(D( 2 ,0)∗ −1 )† = ψ̇D(0, 2 )† .

Similarly if we construct a co-right-handed spinor ϕ̇ = (ϕ̇1 , ϕ̇2 ), the

!
ϕ1
dual vector to the right-handed spinor ϕ = is defined as follows
ϕ2

ϕ̇ = ϕT T . (5.64)

Then the Lorentz transformation on ϕ̇ = (ϕ̇1 , ϕ̇2 ) can be expressed as

L.T. 1
ϕ̇ 7−→ ϕ̇0 = ϕ0T T = ϕ̇D( 2 ,0)† . (5.65)

Let us proceed to consider the irreducible representation denoted by

( 21 , 1̇2 ). It is a 4-dimensional vector space spanned by the direct product
of two spinor spaces, namely the left-handed spinor spaces and the co-
right-handed spinor space, with the bases given by ea f˙b . A general
element in ( 1 , 1̇ )-space can be expressed as ua ea f˙b , or if cast into the
2 2 ḃ
matrix formulation, it can be written as

!
( 12 , 1̇2 ) u11̇ u12̇
U = . (5.66)
u21̇ u22̇

A similar argument leads to express the ( 21̇ , 12 ) representation as follows

!
( 1̇2 , 12 ) u1̇1 u2̇1
U = . (5.67)
u1̇2 u2̇2

It is obvious that Lorentz transformation on vectors in ( 21 , 1̇2 )-space

and in ( 1̇2 , 12 )- space, takes respectively the following expressions:

35:17.
168 Chapter 5

1 1̇ L.T. 0 ( 1 , 1̇ ) 1 1
U ( 2 , 2 ) 7−→ U 2 2 = D( 2 ,0) U D( 2 ,0)† , (5.68a)

1̇ 1 L.T. 0 ( 1̇ , 1 ) 1 1
U ( 2 , 2 ) 7−→ U 2 2 = D(0, 2 ) U D(0, 2 )† . (5.68b)

This transformation property immediately convinces us that the space-

time matrix
!
µ −x0 + x3 x1 − ix2
X = x σµ = , (5.69)
x1 + ix2 −x0 − x3

which we have constructed in Section 5.3 is an operator belonging to

( 12 , 1̇2 )-representation.

5.5 Lorentz spinors and the Dirac equation

As we have discussed in Section 5.1, time and space are regarded

as a four vector in the Minkowski space. In fact, the energy and the
momentum of a mass point, can also be treated as a Minkowski four
vector in relativistic dynamics. Let us define the world line as the
trajectory of a mass point in the four-dimensional Minkowski space in
which the space-time coordinates are functions of a parameter τ , called
the proper time, a Lorentz invariant scalar whose differential is defined
as follows

1
(dτ )2 = − gµν dxµ dxν . (5.70)
c2

Consider the Lorentz frame which is boosted instantaneously with

the same velocity as that of the moving mass point. Then the time
measured in this frame where the mass point is instantaneously at rest,
namely that d~x0 = 0, is in fact the proper time because of the invariant
property,
 1
0 1 0µ 0ν
2
dτ = dτ = − 2 gµν dx dx = dt0 . (5.71)
c

35:17.
5.5 Lorentz spinors and the Dirac equation 169

Yet this invariant quantity when measured in the initial reference frame,
namely O-system, is given by

r
v2
 
1 µ ν 1
dτ = − 2 gµν dx dx 2 = 1 − 2 dt. (5.72)
c c

We will then introduce a velocity 4-vector uµ (τ ) defined as uµ (τ ) =

dxµ /dτ which is related to the energy-momentum 4-vector pµ by the
following equation

pµ = m0 uµ . (5.73)

The energy-momentum 4-vector for a mass point moving with ve-

locity ~v can be expressed in the reference frame as pµ = (p0 , p~), which
when evaluated in the usual relativistic form becomes:

dx0 m0 c
p0 = m 0 =p , (5.74a)
dτ 1 − β2

d~x m0 cβ~
p~ = m0 =p . (5.74b)
dτ 1 − β2

One can easily verify that the scalar product, or the length of the 4-
vector pµ is characterized by the rest mass of the moving particle, an
invariant under the Lorentz transformation, namely

pµ pµ = −m20 c2 . (5.75)

Let us construct a 2 × 2 energy-momentum matrix as follows

!
−p0 + p3 p1 − ip2
P = p µ σµ = . (5.76)
p1 + ip2 −p0 − p3

It transforms as the ( 21 , 1̇2 ) irreducible representation of the Lorentz

group, i.e.

35:17.
170 Chapter 5

1
L.T.
P 7−→ P 0 = D( 2 ,0) (θ, ~ D( 12 ,0)† (θ,
~ ξ)P ~ ξ).
~ (5.77)

As a particular case, let us consider a mass point at rest at the origin

of a frame O, and a frame O0 which is travelling along x3 - axis with the
relative velocity v. Then the 4-momentum in the O0 system can be
expressed as
1 1 †
P 0 = e 2 σ3 ξ P e 2 σ3 ξ , (5.78)

or in matrix notation,

1 1
! ! !
0 e2ξ 0 −m0 c 0 e2ξ 0
P = − 21 ξ 1
0 e 0 −m0 c 0 e− 2 ξ
!
−m0 ceξ 0
= . (5.79)
0 −m0 ce−ξ

We expect that the energy and the momentum of the mass point, when
measured in the O0 -system, will be respectively given as follows:
m0 c
p00 = m0 c cosh ξ = p , (5.80a)
1 − β2

−m0 v
p03 = −m0 c sinh ξ = p . (5.80b)
1 − β2

Let us consider the chiral conjugate (abbreviated as c.c.) of the

space-time matrix X defined as follows
c.c.
X 7−→ Xc = X ∗ −1 . (5.81)

The Lorentz transformation on Xc can be evaluated by the transforma-

tion property of X, namely that
1 1 1 1
Xc0 = X 0∗ −1 = (D( 2 ,0) XD( 2 ,0)† )∗ −1 = D(0, 2 ) Xc D(0, 2 )† . (5.82)

One recognizes that X and Xc belong respectively to the ( 21 , 1̇2 ) and the

35:17.
5.5 Lorentz spinors and the Dirac equation 171

( 1̇2 , 12 ) irreducible representations. Yet as far as Lorentz transformation

1
on the space-time coordinate is concerned, both matrices D( 2 ,0) and
1
D(0, 2 ) play equivalent roles but result in the very same Lorentz transfor-
mation, namely we have two different ways to perform the same Lorentz
transformation by means of employing two different SL(2,C)
1 1
matrices, i.e. D( 2 ,0) and D(0, 2 ) .

The space-time operator X is transformed as ( 12 , 21̇ ) representation

under the Lorentz group. If this tensor operator acts upon a vector
in the right-handed spinor space, i.e. Xϕ, it turns out that the new
transformed vector is a left-handed spinor , because the contraction of
the second indices of the ( 12 , 1̇2 )-tensor with the indices of the (0, 21 )-
tensor results in the ( 12 , 0)-tensor, namely, a left-handed spinor. We
shall summarize these properties in the following proposition:

Proposition 3.

An operator of the ( 12 , 1̇2 )-representation that acts upon a vector of

the (0, 21 )-representation yields a vector of ( 21 , 0)-representation.
Conversely an operator of the ( 1̇2 , 21 )-representation that acts upon
a vector of the ( 12 , 0)-representation will yield a vector of the
(0, 12 )-representation.

Let A be an operator of the ( 21 , 1̇2 )-representation, and ξ a vector of

1
(0,2 )-representation,
then if we denote η as follows
η = Aξ, (5.83)
then the Lorentz transformation on η becomes
L.T. 1 1 1 1
η 7−→ η 0 = A0 ξ 0 = D( 2 ,0) AD( 2 ,0)† D(0, 2 ) ξ = D( 2 ,0) η, (5.84)
which implies that η is a vector of the ( 12 , 0)-representation. Similarly

if we denote Ac , an operator of the ( 1̇2 , 21 )-representation, then ξ = Ac η

has the following transformation property,

35:17.
172 Chapter 5

L.T. 1 1 1 1
ξ 7−→ ξ 0 = A0c η 0 = D(0, 2 ) Ac D(0, 2 )† D( 2 ,0) ξ = D(0, 2 ) ξ, (5.85)

which states that ξ is the vector of the (0, 21 )-representation.

The above proposition allows us to conclude that Pc ψl and P ψr are
the right-handed spinor and the left-handed spinor respectively. If we
represent the quantum state of a free mass point with spin equal to 1/2
by the spinors, then we have

Pc ψl = m0 cψr , (5.86a)
P ψr = m0 cψl , (5.86b)

where we introduce the proportional constant m0 c in order to obtain

the same physical dimension on both sides of the equation.
If we take the direct sum of the right-handed spinor and the left-
handed spinor to form a 4-component spinor, called Dirac spinor, the
above two equations can be simplified in the following expression

! ! !
0 Pc ψr ψr
= m0 c , (5.87)
P 0 ψl ψl

or

!
0 Pc
ψd (x) = m0 cψd (x), (5.88)
P 0

where ψd (x) = ψr (x) ⊕ ψl (x) is the Dirac spinor.

The Dirac equation is usually expressed in the q-representation by
identifying that

~ ∂ ~
Pµ = µ
= ∂µ . (5.89)
i ∂x i

The Dirac equation then reads as

35:17.
5.5 Lorentz spinors and the Dirac equation 173

" ! #
0 iσcµ m0 c
∂µ + ψd (x) = 0, (5.90)
iσ µ 0 ~

or often expressed in covariant form as

 m0 c 
iγ µ ∂µ + ψd (x) = 0, (5.91)
~

where the 4×4 matrices γ µ are called Dirac matrices with the explicit
expression given as follows:

! ! !
0 σcµ 0 I 0 −σ i
γµ = or γ0 = , γi = . (5.92)
σµ 0 I 0 σi 0

By making use of the following property of the anti-commutator in

γ µ , namely that

{γ µ , γ ν } = γ µ γ ν + γ ν γ µ = −2g µν I, (5.93)

we are able to convert the Dirac equation which is of the first order
derivative in space and time, into the second order differential equation
of the Klein-Gordon one. Applying the factor iγ ν ∂ν once more upon
the Dirac equation, we then obtain the following relation

  m c 2 

ν µ m0 c ν  µν 0
iγ ∂ν iγ ∂µ + iγ ∂ν ψd (x) = g ∂µ ∂ν − ψd (x) = 0,
~ ~

or

  m c 2 
µ 0
∂µ ∂ − ψd (x) = 0, (5.94)
~

which implies that each component in the Dirac spinor satisfies the
Klein-Gordon equation.

35:17.
174 Chapter 5

The covariant formulation of the Dirac equation does not imply that
γ µ ∂µ is Lorentz invariant, namely that iγ µ ∂µ in the O-Lorentz frame
will not take the expression iγ 0µ ∂µ0 in the O0 -Lorentz frame. Once the
Dirac matrices are chosen in one Lorentz frame, they will in fact be
valid in all of the Lorentz frames. The universality of the γ µ matrices
is the unique feature of the Dirac equation. What has to be modified
in the Dirac equation under Lorentz transformation are the space-time
coordinates as well as the Dirac spinor components, namely the Lorentz
transformed Dirac equation in the O0 -system will take the following
form
 m0 c  0 0
iγ µ ∂µ0 + ψd (x ) = 0, (5.95)
~

with exactly the same γ µ matrices appearing as in the O-system. We

shall prove the statement in the following proposition.

Proposition 4.
The gamma matrices γ µ are universal in all Lorentz frame,
namely the Dirac equation in another Lorentz frame, i.e. the
O0 -system always takes the same gamma matrices γ µ used in O-
system. The equation in O0 -system is expressed as
 m0 c  0 0
iγ µ ∂µ0 + ψd (x ) = 0.
~

The proof of the statement goes as follows: let us perform the

Lorentz transformation of the Dirac equation in O-system by multiply-
~ ξ),
ing the following 4 × 4 matrix D(θ, ~ the direct sum of the irreducible

1
representation D(0, 2 ) (θ, ~ and D( 12 ,0) (θ,
~ ξ) ~ ξ),
~ i.e.

D(θ, ~ = D(0, 21 ) (θ,

~ ξ) ~ ⊕ D( 12 ,0) (θ,
~ ξ) ~ ξ)
~

~ iγ µ ∂µ + m0 c ψd (x) = 0.
 
~ ξ)
D(θ,
~

35:17.
5.5 Lorentz spinors and the Dirac equation 175

If we insert an identity I = D−1 D right after the γ µ matrix, then the

above equation reads as
 m0 c 
iDγ µ D−1 ∂µ + Dψd (x) = 0. (5.96)
~

We shall leave it as an exercise for the readers to show that

Dγ µ D−1 = Λµν γ ν , (5.97)

which allows us to express the Dirac equation in O0 -system as

 m0 c  0 0
iγ µ ∂µ0 + ψd (x ) = 0,
~

~ ξ)ψ
by identifying ψd0 (x0 ) ≡ D(θ, ~ d (x) = D(θ,
~ ξ)ψ
~ d (Λ−1 x0 ).

The representations of the Dirac matrices γ µ are not unique. During

the development of quantum theory, the formulation of γ µ matrices dif-
fered from one school to another. The γ µ matrices we adopt in Eq. (5.92)
are called the Weyl representation.
Let us define a new set of gamma matrices by means of similarity
transformation with the matrix S given as follows

!
1 I I
S=√ , (5.98)
2 −I I

and denote the transformed matrices by γ̃µ . Then the new set of Dirac
matrices reads as follows

γ̃ µ = Sγ µ S −1 , (5.99)

or explicitly:

! !
I 0 0 −σ i
γ̃ 0 = , γ̃ i = , (5.100)
0 I σi 0

35:17.
176 Chapter 5

which are commonly referred to as the standard representation of

the gamma matrices or the Dirac-Pauli representation as it is often
called.
The similarity transformation of the gamma matrices reshuffles the
components of a Dirac spinor in order to maintain exactly the same form
of the wave equation. Let us multiply the whole equation of Eq. (5.91)
by the matrix S, i.e.
 m0 c 
S iγ µ ∂µ + ψd (x) = 0.
~

If we insert an identity I = S −1 S right after γ µ in the last equation,

and denote the new Dirac spinor by ψ̃d (x) and γ̃ µ = Sγ µ S −1 , we regain
the Dirac equation, namely
 m0 c 
iγ̃ µ ∂µ + ψ̃d (x) = 0.
~

where ψ̃d (x) = Sψd (x).

As the new set of gamma matrices are obtained from the original set
by a similarity transformation, therefore the anti-commutation relation
of Eq. (5.93) is preserved, namely

{γ̃ µ , γ̃ ν } = −2g µν I, (5.101)

this enables us to show that each component in Dirac spinor ψ̃d also
satisfies the Klein-Gordon equation.
Let us return to the p-representation of the Dirac equation expressed
in Eq. (5.88), i.e.

!
0 Pc
ψd (x) − m0 cψd (x) = 0. (5.102)
P 0

In the limit of zero mass, the equation reduces to:

Pc ψl = 0, P ψr = 0. (5.103)

35:17.
5.6 Electromagnetic interaction and gyromagnetic ratio of the electron 177

Since in the zero mass limit, p0 = −p0 = |~

p| and the above equations
can be expressed in terms of the helicity operator as follows:
(σ · p̂)ψr = ψr , (σ · p̂)ψl = −ψl , (5.104)
where σ · p̂ = σ · p/|p| is the helicity operator.

The equations imply that the left-handed spinor and the right-handed
spinor are of opposite helicities.

5.6 Electromagnetic interaction and gyromagnetic ratio of

the electron

Let us investigate a system of a spin 1/2 charged particle moving in

the electromagnetic field of the 4-vector potential Aµ (x) = (ϕ(x), A(x)).
The principle of minimal interaction allows us to express the Dirac equa-
tion in the following form
n  q  o
γ µ Pµ − Aµ (x) − m0 c ψd (x) = 0. (5.105)
c

We shall adopt the Weyl representation of the gamma matrices and

abbreviate P − qA/c into the operator π. Then the above equation can
be separated into the following two coupled equations:
n q  o
P0 + ϕ − σ · π ψl = m0 cψr , (5.106a)
c
n q  o
P0 + ϕ + σ · π ψr = m0 cψl . (5.106b)
c

Let us apply the operator {(P0 + qϕ/c) − σ · π} upon both sides of

the last equation and make use of the Eq.(5.106a). We then obtain the
equation of the right-handed spinor as follows
n qϕ  o n qϕ  o
P0 + −σ·π P0 + + σ · π ψr = (m0 c)2 ψr . (5.107)
c c

Consider for the case of pure magnetic field, namely Aµ = (0, A),
then the stationary state solution of the last equation takes a much

35:17.
178 Chapter 5

simpler expression, i.e.

( 2 )
E
(σ · π)2 − − (m0 c)2 ψr = 0, (5.108)
c

where we replace P0 = −P 0 = −E/c. The first term (σ · π)2 in

Eq. (5.108) can be calculated as follows
 q   q 
(σ · π)2 = π 2 + iσ · p − A × p − A
c c

q~
= π2 − σ · (∇ × A), (5.109)
c

 2
E
and the second term − (m0 c)2 can be approximated as
c
 2  
E E
− (m0 c)2 ' 2m0 c − m0 c = 2m0 E 0 , (5.110)
c c

in the non-relativistic limit, where E 0 represents the total energy minus

the rest mass energy of the particle.
Finally Eq. (5.106a) reduces to the familiar expression of the
Schrödinger equation for a spin 21 charged particle coupled with the
magnetic field H, expressed as follows

 
1  q 2 q~
P− A − σ · H ψr = E 0 ψr . (5.111)
2m0 c 2m0 c

The second term of the last equation stands for the energy of the
intrinsic magnetic moment of the charged spin 1/2 particle µe coupled
to the magnetic field H, namely

q~ q
− σ·H =− S · H = −µe · H. (5.112)
2m0 c m0 c

35:17.
5.7 Gamma matrix algebra and PCT in Dirac spinor system 179

For the case of an electron, the gyromagnetic ratio γe , the ratio

of the magnetic moment µe to the intrinsic spin S, is related to the
g-factor in the following relation
e
µe = γe S = − ge S. (5.113)
2m0 c

Comparing Eq. (5.113) with Eq. (5.112), we reach a remarkable con-

clusion that, to the lowest order approximation, an electron has the
g-factor equaling to 2, namely

ge = 2. (5.114)

The precision in the measurement of the electron magnetic moment

had challenged theorists to higher order calculations in terms of fine
structure constant, namely the radiative corrections, which are beyond
the scope of our investigation.

5.7 Gamma matrix algebra and PCT in Dirac

spinor system

Before entering into the discussion on discrete symmetry transfor-

mation of the Dirac spinor, we shall explore the properties of gamma
matrices. As we have shown that the condition

{γ µ , γ ν } = −2g µν I

is valid for any representation of the Dirac matrix.

Out of these 4 Dirac matrices γ µ , let us construct 16 matrices of
4 × 4 and divide them into 5 classes as follows:

(1) I,

(2) γ µ ,

(3) γ 5 = γ 0 γ 1 γ 2 γ 3 ,

(4) γ 5 γ µ ,

(5) σ µν = 21 [γ m u, γ ν ].

35:17.
180 Chapter 5

Class 1
It contains only the identity matrix, obtained, up to a sign, by mak-
ing the square of any of γ µ matrix. It commutes with all the gamma
matrices in 5 classes.

Class 2
It contains 4 traceless matrices with the square equaling to ±I.

Class 3
It is a traceless, diagonal matrix obtained by successive multiplica-
tion of the 4 Dirac matrices. The matrix is denoted by γ 5 with the
following properties,
(γ 5 )2 = −I,

{γ 5 , γ µ } = 0.

Class 4
It contains 4 traceless matrices with the following properties,
{γ 5 γ µ , γ 5 γ ν } = 2g µν I.

Class 5
It contains 6 matrices, also traceless, by taking different values on µ
and ν. The square of the matrix is, up to a sign, a unit matrix.

With all the preparation on the properties of the gamma matrices,

we are now in position to discuss the parity transformation of the Dirac
spinor. Firstly let us consider the parity transformation of a 4-vector
Rµ in the Minkowski space expressed as follows
P.T
P: Rµ = (R0 , R) −→ Rpµ = (R0 , −R),

or P : PRµ P −1 = Rpµ . (5.115)

35:17.
5.7 Gamma matrix algebra and PCT in Dirac spinor system 181

If Rµ is an operator such that Rpµ = (R0 , −R), then we call Rµ a

4-vector in Minkowski space.
In fact, there exists another vector S µ called the axial 4-vector in
Minkowski space if the transformation property takes the following re-
lation
P.T
P: S µ = (R0 , S) −→ S pµ = (S 0 , S),

or P : PS µ P −1 = S pµ = (S 0 , S), (5.116)

when S µ is an operator.
It is of the outmost interest to look into the parity transformation of
a physical quantity S which is invariant under Lorentz transformation,
namely a Lorentz scalar. We name S a scalar if it is also invariant under
parity transformation, i.e.
P.T.
P: S −→ S p = S,

or if it is an operator, then

P: PSP −1 = S p = S. (5.117)

In contrast to the last expression, if a physical quantity P is trans-

formed by operator P as to meet the following relation, i.e.

PP P −1 = −P, (5.118)

then we call P a pseudoscalar.

Let us start to investigate the parity transformation property of the
Dirac spinor. For the sake of simplicity, we shall express the Dirac
equation in momentum operator as follows

(γ µ Pµ − m0 c)ψd = 0, (5.119a)

or (γ 0 P0 + γ · P − m0 c)ψd = 0. (5.119b)

Applying the parity operator P upon the Dirac equation above, and
inserting the identity I = P −1 P right after the momentum operator Pµ ,
we obtain

35:17.
182 Chapter 5

(γ µ PPµ P −1 − m0 c)Pψd (x) = 0, (5.120a)

or (γ 0 P0 − γ · P − m0 c)ψdp (x) = 0, (5.120b)

where ψdp (x) = Pψd (x).

Therefore we have the following proposition:

Proposition 5.
Let ψd (x) be the solution of the Dirac equation, and Pψd (x) be
the parity transformed Dirac spinor, i.e. ψdp (x) = Pψd (x) =
ψ(x0 , −x), then γ 0 ψ p (x) is the solution of the same Dirac equa-
tion.

The above conclusion is obvious if the Eq. (5.120b) is multiplied by

γ 0 from the right, and making use of the relation γ 0 γ = −γγ 0 , then we
arrive at
(γ µ Pµ − m0 c)γ 0 ψdp (x) = 0. (5.121)

Let us return to the 5 classes of the gamma matrices given at the

beginning of this section. With the definition ψ̄(x) = ψ † (x)γ 0 , we shall
investigate the parity transformation properties of the bilinear spinor
combinations given as follows:
ψ̄(x)ψ(x), ψ̄(x)γ µ ψ(x), ψ̄(x)γ 5 ψ(x), ψ̄(x)γ 5 γ µ ψ(x), ψ̄(x)σ µν ψ(x).

To simplify the demonstration, let us take the solution to be non-

degenerate, namely
γ 0 ψ p (x) = eiα ψ(x).

Therefore one can easily verify that

P ψ̄(x)ψ(x)P −1 = ψ̄ p (x)ψ p (x) = e−iα ψ † (x)γ 0 eiα ψ(x) = ψ̄(x)ψ(x).

As for the transformation property of ψ̄(x)γ 5 ψ(x), one finds that

35:17.
5.7 Gamma matrix algebra and PCT in Dirac spinor system 183

P ψ̄(x)γ 5 ψ(x)P −1 = ψ̄ p (x)γ 5 ψ p (x) = −ψ̄(x)γ 5 ψ(x).

It is because of these peculiar properties given above that we call ψ̄(x)ψ(x)
a scalar and ψ̄(x)γ 5 ψ(x) a pseudoscalar.
We shall summarize the results of the transformation properties for
the rest of three bilinear spinor combinations as follows:
P(ψ̄(x)γ µ ψ(x))P −1 = ψ̄ p (x)γ µ ψ p (x)

= (ψ̄(x)γ 0 ψ(x), −ψ̄(x)γψ(x)),

P(ψ̄(x)γ 5 γ µ ψ(x))P −1 = ψ̄ p (x)γ 5 γ µ ψ p (x)

= (ψ̄(x)γ 5 γ 0 ψ(x), +ψ̄(x)γ 5 γψ(x)),

P(ψ̄(x)σ µν ψ(x))P −1 = ψ̄ p (x)σ µν ψ p (x)

= (−ψ̄(x)σ 0i ψ(x), +ψ̄(x)σ ij ψ(x)).

As one can also prove, it is of course much more complicated than in
the case of parity that under the Lorentz transformation, the bilinear
spinor combinations in class 1 or in class 3 behave as a Lorentz scalar,
while class 2 or class 4 behave as a Lorentz vector and class 5 behave
as a second rank antisymmetric tensor. We shall leave the proof to
students of inquiring minds.

The charge conjugation property can only be analyzed in the system

with the electromagnetic interaction. Let us express the Dirac equation
for a particle with charge q as follows
   
µ iq
iγ ∂µ − Aµ − m0 c ψ(x) = 0. (5.122)
~c

Denoting the charge conjugation operator by C, and defining ψ c (x) =

Cψ(x), then the last equation can be cast into

35:17.
184 Chapter 5

   
iq
iγ ∂µ + Aµ − m0 c ψ c (x) = 0,
µ
(5.123)
~c

where CqC −1 = −q is employed in obtaining the second term.

Taking the complex conjugate of Eq. (5.123), the equation becomes

   
iq
iγ µ∗ ∂µ − Aµ + m0 c ψ c∗ (x) = 0,
~c

which can be converted back into the Dirac equation

   
iq
µ
iγ ∂µ − Aµ − m0 c Cψ c∗ (x) = 0,
~c

where the solution is denoted by Cψ c∗ (x) if the following condition is

met, namely,

Cγ µ∗ C−1 = −γ µ ,

Cψ c∗ (x) = eiβ ψ(x),

with eiβ standing for a phase factor and C a 4 × 4 matrix.

The solution of matrix C is strictly representation dependent. If
we adopt γ µ in Weyl representation, we have the solution C = γ 5 γ 2 ,
because the Dirac matrices are of real matrix elements except γ 2 . The
charge conjugate spinor takes the following expression

ψ C (x) = e−iβ γ 5 γ 2 ψ ∗ (x). (5.124)

As for the time reversal transformation of the following Dirac equa-

tion, one considers

35:17.
5.8 Exercises 185

n m0 c o −1
T iγ µ ∂µ − T T ψ(x) = 0,
~
n
m0 c o t
or −iγ 0 ∂0 + iγ · ∂ − ψ (x) = 0, (5.125)
~

where one takes the time t into −t in the equation, T ψ(x) = ψ t (x), i.e.
T.R.
T : ∂0 7−→ ∂00 = −∂0 .

Equation (5.125) can easily be converted back into the Dirac equa-
tion of the following expression
n
m0 c o t
iγ 0 ∂0 + iγ · ∂ − Tψ (x) = 0,
~

by multiplying a matrix T, if the following conditions are met, i.e.

Tγ 0 T−1 = −γ 0 ,

TγT−1 = γ.

We shall leave it to the readers to verify that T = γ 0 γ 5 , and ψ t (x) =

e−iη γ 0 γ 5 ψ(x) with η being a phase factor.

5.8 Exercises

Ex 5.8.1
Show that a space-like vector is orthogonal to a time-like vector.
(Hint: if x0 > |~x|, then y 0 < |~y | for x · y = 0.)

Ex 5.8.2
Show that the sum of two time-like vectors in the same light cone is
also a time-like vector.
(Hint: make use of the Schwarz inequality.)

35:17.
186 Chapter 5

Ex 5.8.3

Show that two consecutive Lorentz boost transformations in the

same direction with velocities β1 and β2 are equivalent to a boost with
velocity β given by

β1 + β2
β= .
1 + β1 β2

Ex 5.8.4

2
( 1 ,0)
2 ∗ 0 (0, 1 )∗
Verify that Dmm 0 = γ(m)D−m−m0 γ (m ) can be cast in the matrix

form as follows

! !
( 12 ,0) 0 γ( 12 ) 1 0 γ ∗ (− 21 )
D = D(0, 2 )∗ .
1
γ(− 2 ) 0 γ ∗ ( 12 ) 0

Ex 5.8.5

Verify that

! !
0 1 0 −1
σi∗ −1 = σi∗ = −σi .
−1 0 1 0

Ex 5.8.6

Let ψ and ϕ̇ be the left-handed-spinor and co-right-handed spinor

respectively. Prove then that ϕ̇ψ is an invariant scalar under Lorentz
transformation.

35:17.
5.8 Exercises 187

Ex 5.8.7
Show that the Lorentz transformation matrix element Λµν can be
expressed as follows (Hint: by Eqs. (5.77) and (5.82).)
1 n 1 1
o 1 n 1 1
o
Λµν = Tr σcµ D( 2 ,0) σν D( 2 ,0)† = Tr σ µ D(0, 2 ) σcν D(0, 2 )† .
2 2

Ex 5.8.8
Show that Dγ µ D −1 = Λµν γ ν where

1
! !
D(0, 2 ) 0 0 σcν
D= ( 12 ,0)
and γ ν = .
0 D σν 0

(Hint: since X 0 = DXD −1 by definition.)

Ex 5.8.9
Define a new Dirac spinor as
iq
ψd0 (x) = e ~c α(x) ψd (x).
Show that the Dirac equation for a charge particle with EM interac-
tion is invariant under such gauge transformation. (Hint: with a new
vector potential A0µ (x) = Aµ (x) − ∂µ α(x).)

35:17.
This page intentionally left blank

35:17.
Bibliography

Dirac, P. A. M., The Principles of Quantum Mechanics, 4th Edition, (Oxford

University Press, London, 1958).
Feynman, R. P. and A. P. Hibbs, Quantum Mechanics and Path Integrals,
(McGraw-Hill, Inc. 1965).
Gürsey, Feza, Editor; Group Theoretical Concepts and Methods in Elementary
Particle Physics, (Gordon and Breach, Science Publishers, Inc., New York,
1964).
Jordan, Thomas F., Linear Operators for Quantum Mechanics, 2nd Edition,
(John Wiley & Sons, 1982).
Messiah, Albert, Quantum Mechanics, Vol. I & Vol. II, (North-Holland Pub-
lishing Co., Amsterdam; Interscience Publishers Inc., New York, 1962).
Riesz, F. and B. Sz-Nagy, Functional Analysis, translated by L. F. Boron, (F.
Ungar Publishing Co., New York, 1955).
Tung, Wu-Ki, Group Theory in Physics, (World Scientific Publishing Co Pte
Ltd., Singapore, 1985).
von Neumann, John, Mathematical Foundations of Quantum Mechanics, (Prince-
ton University Press, Princeton, New Jersey, 1955).
Wigner, Eugene P., Group Theory, (Academic Press, New York, 1959).
Wybourne, Brian G., Classical Groups for Physicists, ( John Wiley & Sons,
1974).

189
This page intentionally left blank
Index

α-string, 102 Cartan’s criteria

4-vector, 152 for semisimple group, 94
Casimir operator, 96
Abelian group, 94 Cauchy sequence, 7
active transformation, 159 charge conjugate, 110
adjoint conjugate, 11, 25, 29 charge conjugation, 110
adjoint conjugate matrix, 29 chiral transformation, 163
algebra closure relation, 33
Lie algebra, 87, 93 commutability, 16
O(3) algebra, 88, 128, 143 commutator, 17, 20
O(4) algebra, 143 compact group
SO(3) algebra, 117 compact Lie group, 91
SO(3,1) algebra, 156 compatibility, 16
SU(2) algebra, 128 compatible observables, 16
angular momentum, 54, 109, 117, completeness, 6
128 complete vector space, 8
angular momentum operator, conjugate
53, 91, 130, 135, 137 adjoint conjugate, 11, 25
orbital angular momentum , 135 charge conjugate, 110
annihilation operator, 40, 41, 44
contravariant vector, 101
anticommutator, 23
convergent sequence, 7
antilinear operator, 112
covariant vector, 101
antiparticle, 110
cyclic permutation, 97
antiunitary operator, 163
cyclotron frequency, 53
axial vector, 109
axial vector operator, 109
degeneracy, 15, 33, 98
base vector, 33, 51 Kramers degeneracy, 113
basis, 3 m-fold degenerate, 15
boost Lorentz transformation, 153 dimension, 3
bra vector, 31 of the irreducible representation,
131
canonical commutation relations, 20, of vector space, 3
41, 57 Dirac equation, 168

191
192 Index
index

Dirac matrices, 173 state function, 13, 35, 37

Dirac spinor, 172 wave function, 13, 35, 37, 39, 46,
Dirac-Pauli representation of gamma 71
matrices, 176 fundamental commutation relation,
direct product 19
of Hilbert spaces, 50
of operators, 51 g-factor of magnetic moment, 179
of spaces, 19, 50 g-metric tensor of Minkowski space,
of states, 19 152
of vectors, 51, 54 g-tensor of Lie group, 95
direct product vector, 51, 54 generator
direct sum, 172 matrix representation, 120, 133,
discrete 147, 154, 157, 161
discrete group, 107 Gram-Schmidt orthogonalization, 4
discrete symmetry, 107 group
discrete transformation, 107 Abelian group, 94
dual space, 28 continuous group, 83
dual vector space, 27 discrete group, 107
dynamical observable, 12, 14, 19 exceptional group, 105
group generator, 89
eigengenerator, 98 homogeneous Lorentz group,
eigenstate, 14 153
eigenvalue, 14 infinitesimal generator, 89
eigenvector, 14 invariant subgroup, 94, 115
electromagnetic interaction, 177 Lie group, 87, 93
principle of minimal interaction, O(3) group, 117
177 O(3,1) group, 151
vector potential, 177 O(4) group, 141
equation proper Lorentz group, 153
Dirac equation, 168 r-parameter group, 86
Klein-Gordon equation, 173, 176 rotational group, 91, 118
Schrödinger equation, 37, 64, 79 semisimple group, 94
Euclidean space, 2 semisimple Lie group, 94
Euler angles, 122 SL(2,C) group, 151
expectation value, 13, 22, 68 SO(3) group, 117
SO(3,1) group, 156
Fourier SO(4) group, 156
Fourier expansion, 58 SU(2) group, 117, 124
Fourier transform, 39 symmetry group, 83
4-vector, 152 translational group, 84, 90
function U(2) group, 124
δ-function, 33, 36 group composition rules, 84
index
Index 193

group generator, 89 Legendre polynomial, 137

group manifold, 85 Associated Legendre poly., 137
group parameter, 83 Lenz operator, 143
group parameter space, 85 Lenz vector, 143
gyromagnetic ratio, 179 Lie algebra, 87, 93
rank of, 98
Heisenberg standard form of, 98
Heisenberg picture, 68 Lie group, 93
Heisenberg’s uncertainty compact Lie group, 91
relation, 23 linear dependent, 2
helicity, 177 linear independent, 2
Hermitian Lorentz boost transformation, 153
Hermitian conjugate, 28 Lorentz group
Hermitian operator, 12–14 proper Lorentz group, 153
Hilbert space, 6, 8, 12, 32, 33 Lorentz spinor, 156, 168
homogeneity, 83
homogeneous Lorentz group, 153 m-fold degenerate, 15
homogeneous Lorentz matrix
transformation, 152 Λ matrix of Lorentz
transformation, 152
idempotent operator , 12 Dirac matrices, 173
identity orthogonal matrix, 120
identity element of a group, 84, Pauli matrices, 126, 147
86 metric tensor, 152
identity operator, 8 Minkowski
identity transformation, 87 metric tensor, 151
Jacobi identity, 94, 115 Minkowski inequality, 24
space, 151
infinitesimal generator, 89
mirror image, 107
inner product, 3
momentum operator, 34
intrinsic
intrinsic compatibility, 19 norm, 6
intrinsic compatible observables, of a vector, 31, 124
19 normal coordinate, 57
intrinsic spin, 179 null vector, 3
invariance number operator, 42
invariant subgroup, 94, 115
operator
Jacobi identity, 94 adjoint operator, 11
angular momentum operator,
ket vector, 31 53, 91, 130, 135, 137
Klein-Gordon equation, 173, 176 annihilation operator, 40, 41, 44
Kramers degeneracy, 113 antilinear operator, 112
194 Index
index

antiunitary operator, 163 proper Lorentz group, 153

axial vector operator, 109 proper time, 168
bounded operator, 8 pseudoscalar, 181
Casimir operator, 96
constant operator, 21 recursion formula, 103
continuous operator, 9 relativistic wave equation, 152, 159,
creation operator, 40, 41, 44 176
demotion operator, 44 representation
deviation operator, 22 Dirac-Pauli representation of
helicity operator, 177 gamma matrices, 176
Hermitian operator, 12–14 irreducible representation, 133,
idempotent operator, 12 135, 156
identity operator, 8 irreducible representation of
inverse operator, 10 SO(3,1), 156
Lenz operator, 143 n-representation, 46
linear operator, 8 p-representation, 33
momentum operator, 34 q-representation, 33
null operator, 8 standard representation, 176
number operator, 42 Weyl representation of gamma
projection operator, 12, 30–32, matrices, 175
71 Rodrigues formula
promotion operator, 44 for associated Legendre
time evolution operator, 64 polynomials, 141
translational operator, 34 for Legendre polynomials, 141
unit operator, 8 of Hermite polynomials, 46, 48
unitary operator, 10, 34 root vector, 101, 104
vector operator, 108
orthogonal matrix, 120 scalar product, 3
orthogonal vectors, 6 Schrödinger equation, 37, 64, 79
orthonormal, 5 Schrödinger picture, 68
Schwarz inequality, 4
particle-wave duality, 22 semisimple Lie group, 94
path integration, 76 space
physical observable, 13 C n -space, 2
postulate of quantum dynamics, 64 L2 (a, b)-space, 2, 3
postulates of quantum mechanics, 12 Rn -space, 2
∗ n
1st postulate of, 13 C -space, 28
2nd postulate of, 16 complete vector space, 8
3rd postulate of, 20 complex vector space, 2, see also
4th postulate of, 66 C n -space
principle of minimum interaction, 52 dimension of vector space, 3
propagator, 71 direct product space, 19, 50
index
Index 195

dual vector space, 28 boost Lorentz transformation,

Euclidean space, 2, see also Rn - 153
space charge conjugate, 110
Hilbert space, 6, 8, 12, 32, 33 charge conjugation, 110
orthogonal complement space, chiral transformation, 163
12 continuous transformation, 107,
subspace, 12, 17 163
vector space, 1, 27 discrete transformation, 107
spatial translation, 33, 90 homogeneous Lorentz
spherical harmonics, 135 transformation, 152
spin 1/2 particle, 178 identity transformation, 87
spinor Lorentz transformation, 151
co-left-handed spinor, 166 orthogonal transformation, 91,
co-right-handed spinor, 167 see also O(3), O(3,1), O(4)
Dirac spinor, 172 group
left-handed spinor, 159 parity transformation, 107
Lorentz spinor, 156, 168 time reversal transformation,
right handed spinor, 159 107, 110
standard representation, 176 unitary transformation, 70, 117,
state, 13 see also U(2), SU(2) group
physical state, 13 translation, 27, 84
structure constant, 93
subgroup uncertainty
invariant subgroup, 94, 115 deviation, variance, ∆p, ∆q, 22
semisimple Lie group, 94 Heisenberg’s uncertainty, 23
subspace, 12, 17 quantum uncertainty, 40, 79
symmetry, 83 uncertainty principle, 22
discrete symmetry, 107 uncertainty relation, 22
dynamical symmetry, 141 unitarity, 125, 126, 157, 158
rotational symmetry, 83 unitary
transformation, 83 unitary group, see U(2),SU(2)
translational, 83 group
unitary operator, 10, 34
temporal reflection, 112 unitary transformation, 70, 117
tensor
antisymmetric tensor, 120, 183 vector
metric tensor, 152 base vector, 33, 51
time evolution, 12, 64 basis of, 3
time reversal transformation, 107, bra vector, 31
110 contravariant vector, 101
transformation, 111 direct product vector, 51, 54
active transformation, 159 4-vector, 152
196 Index
index

ket vector, 31
Lenz vector, 143
linear dependent, 2
linear independent, 2
null vector, 3
orthogonal vectors, 6
orthonormal set of, 5
root vector, 101, 104
state vector, 13
vector space, 1, 27
vector diagram, 104

wave
relativistic wave equation, 152,
159, 176
wave function, 13, 35, 37, 39, 46,
71
wave number, 61, 63