UNITEXT for Physics

Luca Salasnich

Quantum Physics
of Light and
Matter
A Modern Introduction to Photons,
Atoms and Many-Body Systems
UNITEXT for Physics

Series editors
Michele Cini, Roma, Italy
A. Ferrari, Torino, Italy
Stefano Forte, Milano, Italy
Inguscio Massimo, Firenze, Italy
G. Montagna, Pavia, Italy
Oreste Nicrosini, Pavia, Italy
Luca Peliti, Napoli, Italy
Alberto Rotondi, Pavia, Italy

For further volumes:

http://www.springer.com/series/13351
Luca Salasnich

Quantum Physics of Light

and Matter
A Modern Introduction to Photons,
Atoms and Many-Body Systems

123
Luca Salasnich
Fisica e Astronomia ‘‘Galileo Galilei’’
Università di Padova
Padova
Italy

ISSN 2198-7882 ISSN 2198-7890 (electronic)

ISBN 978-3-319-05178-9 ISBN 978-3-319-05179-6 (eBook)
DOI 10.1007/978-3-319-05179-6
Springer Cham Heidelberg New York Dordrecht London

Library of Congress Control Number: 2014935224

 Springer International Publishing Switzerland 2014

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publisher’s location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Preface

This book contains lecture notes prepared for the one-semester course ‘‘Structure
of Matter’’ belonging to the Master of Science in Physics at the University of
Padova. The course gives an introduction to the field quantization (second quan-
tization) of light and matter with applications to atomic physics.
Chapter 1 briefly reviews the origins of special relativity and quantum
mechanics and the basic notions of quantum information theory and quantum
statistical mechanics. Chapter 2 is devoted to the second quantization of the elec-
tromagnetic field, while Chap. 3 shows the consequences of the light field quan-
tization in the description of electromagnetic transitions. In Chap. 4, it is analyzed
the spin of the electron, and in particular its derivation from the Dirac equation,
while Chap. 5 investigates the effects of external electric and magnetic fields on the
atomic spectra (Stark and Zeeman effects). Chapter 6 describes the properties of
systems composed by many interacting identical particles. It is also discussed the
Fermi degeneracy and the Bose–Einstein condensation introducing the Har-
tree–Fock variational method, the density functional theory, and the
Born–Oppenheimer approximation. Finally, in Chap. 7, it is explained the second
quantization of the nonrelativistic matter field, i.e., the Schrödinger field, which
gives a powerful tool for the investigation of finite-temperature many-body prob-
lems and also atomic quantum optics. Moreover, in this last chapter, fermionic Fock
states and coherent states are presented and the Hamiltonians of Jaynes–Cummings
and Bose–Hubbard are introduced and investigated. Three appendices on the Dirac
delta function, the Fourier transform, and the Laplace transform complete the book.
It is important to stress that at the end of each chapter there are solved problems
which help the students to put into practice the things they learned.

Padova, January 2014 Luca Salasnich

v
Contents

1 The Origins of Modern Physics . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Special Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Axioms of Quantum Mechanics . . . . . . . . . . . . . . . . . . 8
1.2.2 Quantum Information . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Quantum Statistical Mechanics . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.1 Microcanonical Ensemble . . . . . . . . . . . . . . . . . . . . . . 13
1.3.2 Canonical Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.3 Grand Canonical Ensemble . . . . . . . . . . . . . . . . . . . . . 15
1.4 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2 Second Quantization of Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.1 Electromagnetic Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.1 First Quantization of Light . . . . . . . . . . . . . . . . . . . . . . 23
2.1.2 Electromagnetic Potentials and Coulomb Gauge . . . . . . . 24
2.2 Second Quantization of Light . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.1 Fock Versus Coherent States for the Light Field. . . . . . . 30
2.2.2 Linear and Angular Momentum
of the Radiation Field . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.3 Zero-Point Energy and the Casimir Effect . . . . . . . . . . . 34
2.3 Quantum Radiation Field at Finite Temperature . . . . . . . . . . . . 36
2.4 Phase Operators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3 Electromagnetic Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1 Classical Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Quantum Electrodynamics in the Dipole Approximation . . . . . . 53
3.2.1 Spontaneous Emission . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.2 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2.3 Stimulated Emission . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3 Selection Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4 Einstein Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

vii
viii Contents

3.4.1Rate Equations for Two-Level

and Three-Level Systems . . . . . . . . . . . . . . . . . . . . . . . 64
3.5 Life-Time and Natural Line-Width . . . . . . . . . . . . . . . . . . . . . 66
3.5.1 Collisional Broadening . . . . . . . . . . . . . . . . . . . . . . . . 67
3.5.2 Doppler Broadening . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.6 Minimal Coupling and Center of Mass. . . . . . . . . . . . . . . . . . . 69
3.7 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4 The Spin of the Electron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4.1 The Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.2 The Pauli Equation and the Spin . . . . . . . . . . . . . . . . . . . . . . . 85
4.3 Dirac Equation with a Central Potential . . . . . . . . . . . . . . . . . . 87
4.3.1 Relativistic Hydrogen Atom and Fine Splitting. . . . . . . . 88
4.3.2 Relativistic Corrections to the Schrödinger
Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . ....... 89
4.4 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 91
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 97

5 Energy Splitting and Shift Due to External Fields . . . . . . . . . . . . . 99

5.1 Stark Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.2 Zeeman Effect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2.1 Strong-Field Zeeman Effect . . . . . . . . . . . . . . . . . . . . . 102
5.2.2 Weak-Field Zeeman Effect. . . . . . . . . . . . . . . . . . . . . . 103
5.3 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

6 Many-Body Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.1 Identical Quantum Particles . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.2 Non-interacting Identical Particles . . . . . . . . . . . . . . . . . . . . . . 117
6.2.1 Uniform Gas of Non-interacting Fermions . . . . . . . . . . . 119
6.2.2 Atomic Shell Structure and the Periodic
Table of the Elements . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.3 Interacting Identical Particles . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.3.1 Variational Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.3.2 Hartree for Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.3.3 Hartree-Fock for Fermions . . . . . . . . . . . . . . . . . . . . . . 126
6.3.4 Mean-Field Approximation . . . . . . . . . . . . . . . . . . . . . 129
6.4 Density Functional Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.5 Molecules and the Born-Oppenheimer Approximation . . . . . . . . 137
6.6 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Contents ix

7 Second Quantization of Matter. . . . . . . . . . . . . . . . . . . . . . . . . . . 145

7.1 Schrödinger Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.2 Second Quantization of the Schrödinger Field. . . . . . . . . . . . . . 147
7.2.1 Bosonic and Fermionic Matter Field . . . . . . . . . . . . . . . 150
7.3 Connection Between First and Second Quantization . . . . . . . . . 153
7.4 Coherent States for Bosonic and Fermionic Matter Fields . . . . . 156
7.5 Quantum Matter Field at Finite Temperature . . . . . . . . . . . . . . 160
7.6 Matter-Radiation Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . 162
7.6.1 Cavity Quantum Electrodynamics . . . . . . . . . . . . . . . . . 163
7.7 Bosons in a Double-Well Potential . . . . . . . . . . . . . . . . . . . . . 167
7.7.1 Analytical Results with N ¼ 1 and N ¼ 2 . . . . . . . . . . . 170
7.8 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

Appendix A: Dirac Delta Function. . . . . . . . . . . . . . . . . . . . . . . . . . . 179

Appendix B: Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

Appendix C: Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Chapter 1
The Origins of Modern Physics

In this chapter we review the main results of early modern physics (which we suppose
the reader learned in previous introductory courses), namely the special relativity of
Einstein and the old quantum mechanics due to Planck, Bohr, Schrödinger, and
others. The chapter gives also a brief description of the relevant axioms of quan-
tum mechanics, historically introduced by Dirac and Von Neumann, and elementary
notions of quantum information. The chapter ends with elements of quantum statis-
tical mechanics.

1.1 Special Relativity

In 1887 Albert Michelson and Edward Morley made a break-thought experiment of

optical interferometry showing that the speed of light in the vacuum is

c = 3 × 108 m/s, (1.1)

independently on the relative motion of the observer (here we have reported an

approximated value of c which is correct within three digits). Two years later, Henry
Poincaré suggested that the speed of light is the maximum possible value for any
kind of velocity. On the basis of previous ideas of George Francis FitzGerald, in
1904 Hendrik Lorentz found that the Maxwell equations of electromagnetism are
invariant with respect to this kind of space-time transformations

x − vt
x∧ =  (1.2)
1 − vc2
2

y∧ = y (1.3)
z∧ = z (1.4)

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 1

DOI: 10.1007/978-3-319-05179-6_1, © Springer International Publishing Switzerland 2014
2 1 The Origins of Modern Physics

t − vx/c2
t∧ =  , (1.5)
1 − vc2
2

which are called Lorentz (or Lorentz-FitzGerald) transformations. This research

activity on light and invariant transformations was summarized in 1905 by Albert
Einstein, who decided to adopt two striking postulates:
(i) the law of physics are the same for all inertial frames;
(ii) the speed of light in the vacuum is the same in all inertial frames.
From these two postulates Einstein deduced that the laws of physics are invariant with
respect to Lorentz transformations but the law of Newtonian mechanics (which are
not) must be modified. In this way Einstein developed a new mechanics, the special
relativistic mechanics, which reduces to the Newtonian mechanics when the involved
velocity v is much smaller than the speed of light c. One of the amazing results of
relativistic kinematics is the length contraction: the length L of a rod measured by
an observer which moves at velocity v with respect to the rod is given by

v2
L = L0 1 − , (1.6)
c2

where L0 is the proper length of the rod. Another astonishing result is the time
dilatation: the time interval T of a clock measured by an observer which moves at
velocity v with respect to the clock is given by

T0
T= , (1.7)
v2
1− c2

where T0 is the proper time interval of the clock.

We conclude this section by observing that, according to the relativistic mechanics
of Einstein, the energy E of a particle of rest mass m and linear momentum p = |p|
is given by 
E= p2 c2 + m2 c4 . (1.8)

If the particle has zero linear momentum p, i.e. p = 0, then

E = mc2 , (1.9)

which is the rest energy of the particle. Instead, if the linear momentum p is finite
but the condition pc/(mc2 )  1 holds one can expand the square root finding

p2
E = mc2 + + O(p4 ), (1.10)
2m
1.1 Special Relativity 3

which shows that the energy E is approximated by the sum of two contributions: the
rest energy mc2 and the familiar non-relativistic kinetic energy p2 /(2m). In the case
of a particle with zero rest mass, i.e. m = 0, the energy is given by

E = pc, (1.11)

which is indeed the energy of light photons. We stress that all the predictions of special
relativistic mechanics have been confirmed by experiments. Also many predictions
of general relativity, which generalizes the special relativity taking into account the
theory of gravitation, have been verified experimentally and also used in applications
(for instance, some global positioning system (GPS) devices include general relativity
corrections).

1.2 Quantum Mechanics

Historically the beginning of quantum mechanics is set in the year 1900 when Max
Planck found that the only way to explain the experimental results of the electro-
magnetic spectrum emitted by hot solid bodies is to assume that the energy E of
the radiation with frequency ν emitted from the walls of the body is quantized
according to
E = hν n, (1.12)

where n = 0, 1, 2, . . . is an integer quantum number. With the help of statistical

mechanics Planck derived the following expression

8π 2 2 hν
ρ(ν) = ν βhν (1.13)
c3 e −1

for the electromagnetic energy density per unit of frequency ρ(ν) emitted by the hot
body at the temperature T , where β = 1/(kB T ) with kB = 1.38 × 10−23 J/K the
Boltzmann constant, c = 3 × 108 m/s the speed of light in the vacuum. This formula,
known as Planck law of the black-body radiation, is in very good agreement with
experimental data. Note that a hot solid body can be indeed approximated by the
so-called black body, that is an idealized physical body that absorbs all incident
electromagnetic radiation and it is also the best possible emitter of thermal radiation.
The constant h derived by Plank from the interpolation of experimental data of
radiation reads
h = 6.63 × 10−34 J s. (1.14)

This parameter is called Planck constant. Notice that often one uses instead the
reduced Planck constant
h
= = 1.06 × 10−34 J s. (1.15)
2π
4 1 The Origins of Modern Physics

Few years later the black-body formulation of Planck, in 1905, Albert Einstein
suggested that not only the electromagnetic radiation is emitted by hot bodies in
a quantized form, as found by Planck, but that indeed electromagnetic radiation is
always composed of light quanta, called photons, with discrete energy

 = hν. (1.16)

Einstein used the concept of photon to explain the photoelectric effect and predicted
that the kinetic energy of an electron emitted by the surface of a metal after being
irradiated is given by
1 2
mv = hν − W , (1.17)
2
where W is work function of the metal (i.e. the minimum energy to extract the electron
from the surface of the metal), m is the mass of the electron, and v is the emission
velocity of the electron. This prediction clearly implies that the minimal radiation
frequency to extract electrons from a metal is νmin = W /. Subsequent experiments
confirmed the Einstein’s formula and gave a complementary measure of the Planck
constant h.
In 1913 Niels Bohr was able to explain the discrete frequencies of electromagnetic
emission of hydrogen atom under the hypothesis that the energy of the electron
orbiting around the nucleus is quantized according to the formula

me4 1 1
En = − = −13.6 eV 2 , (1.18)
2ε20 h2 n2 n

where n = 1, 2, 3, . . . is the principal integer quantum number, e is the electric charge

of the electron and ε0 the dielectric constant of the vacuum. This expression shows
that the quantum states of the system are characterized by the quantum number n and
the ground-state (n = 1) has the energy −13.6 eV, which is the ionization energy
of the hydrogen. According to the theory of Bohr, the electromagnetic radiation is
emitted or absorbed when one electron has a transition from one energy level En to
another Em . In addition, the frequency ν of the radiation is related to the energies of
the two states involved in the transition by

hν = En − Em . (1.19)

Thus any electromagnetic transition between two quantum states implies the emission
or the absorption of one photon with an energy hν equal to the energy difference of
the involved states.
In 1922 Arthur Compton noted that the diffusion of X rays with electrons
(Compton effect) is a process of scattering between photons of X rays and
electrons. The photon of X rays has the familiar energy
1.2 Quantum Mechanics 5

c
 = hν = h , (1.20)
λ
with λ wavelength of the photon, because the speed of light is given by c = λν. By
using the connection relativistic between energy  and linear momentum p = |p| for
a particle with zero rest mass, Compton found

 = pc, (1.21)

where the linear momentum of the photon can be then written as

h
p= . (1.22)
λ
By applying the conservation of energy and linear momentum to the scattering
process Compton got the following expression

h
λ∧ − λ = (1 − cos (θ)) (1.23)
mc

for the wavelength λ∧ of the diffused photon, with θ the angle between the incoming
photon and the outcoming one. This formula is in full agreement with experimental
results.
Inspired by the wave-particle behavior exhibited by the light, in 1924 Louis de
Broglie suggested that also the matter, and in particular the electron, has wave-like
properties. He postulated that the relationship

h
λ= (1.24)
p

applies not only to photons but also to material particles. In general, p is the linear
momentum (particle property) and λ the wavelength (wave property) of the quantum
entity, that is usually called quantum particle.
Two years later the proposal of de Broglie, in 1926, Erwin Schrödinger went to
the extremes of the wave-particle duality and introduced the following wave equation
for one electron under the action of an external potential U(r)
 
∂ 2
i ψ(r, t) = − ∇ 2 + U(r) ψ(r, t), (1.25)
∂t 2m

∂ 2 ∂ 2 ∂ 2
where ∇ 2 = ∂x 2 + ∂y2 + ∂z2 is the Laplacian operator. This equation is now known
as the Schrödinger equation. In the case of the hydrogen atom, where

e2
U(r) = − , (1.26)
4πε0 r 2
6 1 The Origins of Modern Physics

Schrödinger showed that setting

ψ(r, t) = Rn (r) e−iEn t/ (1.27)

one finds  
2 e2
− ∇2 − Rn (r) = En Rn (r) (1.28)
2m 4πε0 r 2

the stationary Schrödinger equation for the radial eigenfunction Rn (r) with
eigenvalue En given exactly by the Bohr formula (1.18). Soon after, it was under-
stood that the stationary Schrödinger equation of the hydrogen atom satisfies a more
general eigenvalue problem, namely
 
2 e2
− ∇2 − ψnlml (r) = En ψnlml (r), (1.29)
2m 4πε0 r 2

where ψnlml (r) is a generic eigenfunction of the problem, which depends on three
quantum numbers: the principal quantum number n = 1, 2, 3, . . ., the angular quan-
tum number l = 0, 1, 2, . . . , n − 1, and third-component angular quantum number
ml = −l, −l + 1, . . . , l − 1, l. The generic eigenfunction of the electron in the
hydrogen atom in spherical coordinates is given by

ψnlml (r, θ, φ) = Rnl (r) Ylml (θ, φ),

where Rnl (r) is the radial wavefunction while Ylml (θ, φ) is the angular wavefunction.
Notice that in Eq. (1.28) we have Rn (r) = Rn0 (r). It is important to stress that initially
Schrödinger thought that ψ(r) was a matter wave, such that |ψ(r, t)|2 gives the
local density of electrons in the position r at time t. It was Max Born that correctly
interpreted ψ(r, t) as a probability field, where |ψ(r, t)|2 is the local probability
density of finding one electron in the position r at time t, with the normalization
condition 
d 3 r |ψ(r, t)|2 = 1. (1.30)

In the case of N particles the probabilistic interpretation of Born becomes cru-

cial, with (r1 , r2 , . . . , rN , t) the many-body wavefunction of the system, such that
|(r1 , r2 , . . . , rN , t)|2 is the probability of finding at time t one particle in the posi-
tion r1 , another particle in the position r2 , and so on.
In the same year of the discovery of the Schrödinger equation Max Born, Pasqual
Jordan and Werner Heisenberg introduced the matrix mechanics. According to this
theory the position r and the linear momentum p of an elementary particle are not
vectors composed of numbers but instead vector composed of matrices (operators)
which satisfies strange commutation rules, i.e.

r̂ = (x̂, ŷ, ẑ), p̂ = (p̂x , p̂y , p̂z ), (1.31)

1.2 Quantum Mechanics 7

such that
[r̂, p̂] = i, (1.32)

where the hat symbol is introduced to denote operators and [Â, B̂] = ÂB̂ − B̂Â is
the commutator of generic operators  and B̂. By using their theory Born, Jordan
and Heisenberg were able to obtain the energy spectrum of the hydrogen atom
and also to calculate the transition probabilities between two energy levels. Soon
after, Schrödinger realized that the matrix mechanics is equivalent to his wave-like
formulation: introducing the quantization rules

r̂ = r, p̂ = −i∇, (1.33)

for any function f (r) one finds immediately the commutation rule
 
r̂ · p̂ − p̂ · r̂ f (r) = (−ir · ∇ + i∇ · r) f (r) = if (r). (1.34)

Moreover, starting from classical Hamiltonian

p2
H= + U(r), (1.35)
2m
the quantization rules give immediately the quantum Hamiltonian operator

2 2
H=− ∇ + U(r) (1.36)
2m
from which one can write the time-dependent Schrödinger equation as

∂
i ψ(r, t) = Ĥψ(r, t). (1.37)
∂t
In 1927 the wave-like behavior of electrons was eventually demonstrated by
Clinton Davisson and Lester Germer, who observed the diffraction of a beam of
electrons across a solid crystal. The diffracted beam shows intensity maxima when
the following relationship is satisfied

2 d sin(φ) = n λ, (1.38)

where n is an integer number, λ is the de Broglie wavelength of electrons, d is the

separation distance of crystal planes and φ is the angle between the incident beam of
electrons and the surface of the solid crystal. The condition for a maximum diffracted
beam observed in this experiment corresponds indeed to the constructive interference
of waves, which is well know in optics (Bragg condition).
8 1 The Origins of Modern Physics

1.2.1 Axioms of Quantum Mechanics

The axiomatic formulation of quantum mechanics was set up by Paul Maurice Dirac
in 1930 and John von Neumann in 1932. The basic axioms are the following:
Axiom 1. The state of a quantum system is described by a unitary vector |ψ→ belonging
to a separable complex Hilbert space.
Axiom 2. Any observable of a quantum system is described by a self-adjoint linear
operator F̂ acting on the Hilbert space of state vectors.
Axiom 3. The possible measurable values of an observable F̂ are its eigenvalues f,
such that
F̂|f → = f |f →

with |f → the corresponding eigenstate. Note that the observable F̂ admits the spectral
resolution
F̂ = |f →f ∀f |,
f

which is quite useful in applications, as well as the spectral resolution of the identity

Î = |f →∀f |.
f

Axiom 4. The probability p of finding the state |ψ→ in the state |f → is given by

p = |∀f |ψ→|2 ,

where the complex probability amplitude ∀f |ψ→ denotes the scalar product of the
two vectors. This probability p is also the probability of measuring the value f of
the observable F̂ when the system in the quantum state |ψ→. Notice that both |ψ→
and |f → must be normalized to one. Often it is useful to introduce the expectation
value (mean value or average value) of an observable F̂ with respect to a state |ψ→
as ∀ψ|F̂|ψ→. Moreover, from this Axiom 4 it follows that the wavefunction ψ(r) can
be interpreted as
ψ(r) = ∀r|ψ→,

that is the probability amplitude of finding the state |ψ→ in the position state |r→.
Axiom 5. The time evolution of states and observables of a quantum system with
Hamiltonian Ĥ is determined by the unitary operator

Û(t) = exp (−iĤt/),

such that |ψ(t)→ = Û(t)|ψ→ is the time-evolved state |ψ→ at time t and F̂(t) =
Û −1 (t)F̂ Û(t) is the time-evolved observable F̂ at time t. From this axiom one finds
immediately the Schrödinger equation
1.2 Quantum Mechanics 9

∂
i |ψ(t)→ = Ĥ|ψ(t)→
∂t

for the state |ψ(t)→, and the Heisenberg equation

∂
i F̂ = [F̂(t), Ĥ]
∂t

for the observable F̂(t).

1.2.2 Quantum Information

The qubit, or quantum bit, is the quantum analogue of a classical bit. It is the unit
of quantrum information, namely a two-level quantum system. There are many two-
level system which can be used as a physical realization of the qubit. For instance:
horizontal and veritical polarizations of light, ground and excited states of atoms or
molecules or nuclei, left and right wells of a double-well potential, up and down
spins of a particle.
The two basis states of the qubit are usually denoted as |0→ and |1→. They can be
written as



1 0
|0→ = , |1→ = , (1.39)
0 1

by using a vector representation, where clearly

∀0| = (1, 0), ∀1| = (0, 1), (1.40)

and also


1
∀0|0→ = (1, 0) = 1, (1.41)
0


0
∀0|1→ = (1, 0) = 0, (1.42)
1


1
∀1|0→ = (0, 1) = 0, (1.43)
0


0
∀1|1→ = (0, 1) = 1, (1.44)
1

while



1 10
|0→∀0| = (1, 0) = , (1.45)
0 00
10 1 The Origins of Modern Physics



1 01
|0→∀1| = (0, 1) = , (1.46)
0 00



0 00
|1→∀0| = (1, 0) = , (1.47)
1 10



0 00
|1→∀1| = (0, 1) = . (1.48)
1 01

Of course, instead of |0→ and |1→ one can choose other basis states. For instance:



1 1 1 1 1 1
|+→ = √ (|0→ + |1→) = √ , |−→ = √ (|0→ − |1→) = √ ,
2 2 1 2 2 −1
(1.49)

but also



1 1 1 1 1 1
|i→ = √ (|0→ + i|1→) = √ , | − i→ = √ (|0→ − i|1→) = √ .
2 2 i 2 2 −i
(1.50)

A pure qubit |ψ→ is a linear superposition (superposition state) of the basis states
|0→ and |1→, i.e.
|ψ→ = α |0→ + β |1→, (1.51)

where α and β are the probability amplitudes, usually complex numbers, such that

|α|2 + |β|2 = 1. (1.52)

A quantum gate (or quantum logic gate) is a unitary operator Û acting on qubits.
Among the quantum gates acting on a single qubit there are: the identity gate


10
Î = , (1.53)
01

the Hadamard gate


1 1 1
Ĥ = √ (1.54)
2 1 −1

the not gate (also called Pauli X gate)


01
X̂ = = σ̂1 , (1.55)
10

the Pauli Y gate


0 −i
Ŷ = = σ̂2 , (1.56)
i 0
1.2 Quantum Mechanics 11

and the Pauli Z gate


1 0
Ẑ = = σ̂3 . (1.57)
0 −1

It is strightforward to deduce the properties of the quantum gates. For example, one
easily finds
Ĥ|0→ = |+→, Ĥ|1→ = |−→, (1.58)

but also
Ĥ|+→ = |0→, Ĥ|−→ = |1→. (1.59)

A N-qubit |N →, also called a quantum register, is a quantum state characterized by

|N → = cα1 ...αN |α1 → ⊗ · · · ⊗ |αN →, (1.60)
α1 ...αN

where |αi → is a single qubit (with αi = 0, 1), and

|cα1 ...αN |2 = 1. (1.61)
α1 ...αN

In practice, the N-qubit describes N two-state configurations. The more general 2-

qubit |2 → is consequently given by

|2 → = c00 |00→ + c01 |01→ + c10 |10→ + c11 |11→, (1.62)

where we use |00→ = |0→⊗|0→, |01→ = |0→⊗|1→, |10→ = |1→⊗|0→, and |11→ = |1→⊗|1→
to simplify the notation.
For the basis states of 2-qubits we one introduce the following vector
representation
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0 0
⎜0⎟ ⎜1⎟ ⎜0⎟ ⎜0⎟
|00→ = ⎜
⎝0⎠,
⎟ |01→ = ⎜
⎝0⎠,
⎟ |10→ = ⎜
⎝1⎠,
⎟ |11→ = ⎜
⎝0⎠.
⎟ (1.63)
0 0 0 1

Moreover, quantum gates can be introduced also for 2-qubits. Among the quantum
gates acting on a 2-qubit there are: the identity gate
⎛ ⎞
1 0 0 0
⎜0 1 0 0⎟
Î = ⎜
⎝0
⎟ (1.64)
0 1 0⎠
0 0 0 1

and the CNOT (also called controlled-NOT) gate

12 1 The Origins of Modern Physics
⎛ ⎞
1 0 0 0
⎜0 1 0 0⎟
ˆ =⎜
CNOT ⎟. (1.65)
⎝0 0 0 1⎠
0 0 1 0

Notice that the action of the CNOT gate on a state |αβ→ = |α→|β→ is such that the
“target” qubit |β→ changes only if the “control” qubit |α→ is |1→.
The 2-qubit |2 → is called separable if it can be written as the tensor product of
two generic single qubits |ψA → and |ψB →, i.e.

|2 → = |ψA → ⊗ |ψB →. (1.66)

If this is not possible, the state |2 → is called entangled. Examples of separable
states are
|00→ = |0→ ⊗ |0→, (1.67)

or
|01→ = |0→ ⊗ |1→, (1.68)

but also
1 1
√ (|01→ + |11→) = √ (|0→ + |1→) ⊗ |1→. (1.69)
2 2

Examples of entangled states are instead

1
√ (|01→ ± |10→) . (1.70)
2

but also
1
√ (|00→ ± |11→) , (1.71)
2

which are called Bell states.

Remarkably, a CNOT gate acting on a separable state can produce an entagled
state, and viceversa. For instance:

ˆ √1 (|00→ + |10→) = √1 (|00→ + |11→) ,

CNOT (1.72)
2 2

while
ˆ √1 (|00→ + |11→) = √1 (|00→ + |10→) .
CNOT (1.73)
2 2
1.3 Quantum Statistical Mechanics 13

1.3 Quantum Statistical Mechanics

Statistical mechanics aims to describe macroscopic properties of complex systems

starting from their microscopic components by using statistical averages. Statistical
mechanics must reproduce the general results of Thermodynamics both at equilib-
rium and out of equilibrium. Clearly, the problem is strongly simplified if the system
is at thermal equilibrium. Here we discuss only quantum systems at thermal equilib-
rium and consider a many-body quantum system of identical particles characterized
by the Hamiltonian Ĥ such that

(N) (N) (N)

Ĥ|Ei → = Ei |Ei →, (1.74)

where |Ei(N) → are the eigenstates of Ĥ for a fixed number N of identical particles and
EiN are the corresponding eigenenergies.

1.3.1 Microcanonical Ensemble

In the microcanonical ensemble the quantum many-body system in a volume V has
a fixed number N of particles and also a fixed energy E. In this case the Hamiltonian
Ĥ admits the spectral decomposition
(N) (N) (N)
Ĥ = Ei |Ei →∀Ei |, (1.75)
i

and one defines the microcanonical density operator as

ρ̂ = δ(E − Ĥ), (1.76)

where δ(x) is the Dirac delta function. This microcanonical density operator ρ̂ has
the spectral decomposition
(N) (N) (N)
ρ̂ = δ(E − Ei )|Ei →∀Ei |. (1.77)
i

The key quantity in the microcanonical ensemble is the density of states (or micro-
canonical volume) W given by

W = Tr[ρ̂] = Tr[δ(E − Ĥ)], (1.78)

namely (N)
W= δ(E − Ei ). (1.79)
i
14 1 The Origins of Modern Physics

The ensemble average of an observable described by the self-adjunct operator  is

defined as

Tr[Â ρ̂] 1 (N) (N)

∀Â→ = = Aii δ(E − Ei ), (1.80)
Tr[ρ̂] W
i

(N) (N) (N)

where Aii = ∀Ei |Â|Ei →. The connection with Equilibrium Thermodynamics is
given by the formula

S = kB ln (W ), (1.81)

which introduces the entropy S as a function of energy E, volume V and number N

of particles. Note that Eq. (1.81) was descovered by Ludwig Boltzmann in 1872 and
kB = 1.38 × 10−23 J/K is the Boltzmann constant. From the entropy S(E, V, N) the
absolute temperature T , the pressure P and the chemical potential μ are obtained as




1 ∂S ∂S ∂S
= , P=T , μ = −T , (1.82)
T ∂E V,N ∂V E,N ∂N E,V

which are familiar relationships of Equilibrium Thermodynamics such that

1 P μ
dS = dE + dV − dN. (1.83)
T T T
In fact, in the microcanonical ensemble the independent thermodynamic variables
are E, N and V , while T , P and μ are dependent thermodynamic variables.

1.3.2 Canonical Ensemble

In the canonical ensemble the quantum system in a volume V has a fixed number N
of particles and a fixed temperature T . In this case one defines the canonical density
operator as
ρ̂ = e−β Ĥ , (1.84)

with β = 1/(kB T ). Here ρ̂ has the spectral decomposition

(N)
ρ̂ = e−βEi |Ei(N) →∀Ei(N) |. (1.85)
i

The key quantity in the canonical ensemble is the canonical partition function (or
canonical volume) Z N given by
1.3 Quantum Statistical Mechanics 15

ZN = Tr[ρ̂] = Tr[e−β Ĥ ], (1.86)

namely
(N)
ZN = e−βEi . (1.87)
i

The ensemble average of an observable  is defined as

Tr[Â ρ̂] 1 (N) −βE (N)

∀Â→ = = Aii e i , (1.88)
Tr[ρ̂] ZN
i

(N) (N) (N)

where Aii = ∀Ei |Â|Ei →. Notice that the definition of canonical-ensemble aver-
age is the same of the microcanonical-ensemble average but the density of state ρ̂ is
different in the two ensembles. The connection with Equilibrium Thermodynamics
is given by the formula

ZN = e−βF , (1.89)

which introduces the Helmholtz free energy F as a function of temperature T , volume

V and number N of particles. From the Helholtz free energy F(T , V, N) the entropy
S, the pressure P and the chemical potential μ are obtained as




∂F ∂F ∂F
S=− , P=− , μ=− , (1.90)
∂T V,N ∂V T ,N ∂N T ,V

which are familiar relationships of Equilibrium Thermodynamics such that

dF = −SdT − PdV − μdN. (1.91)

In fact, in the canonical ensemble the independent thermodynamic variables are T , N

and V , while S, P and μ are dependent thermodynamic variables.

1.3.3 Grand Canonical Ensemble

In the grand canonical ensemble the quantum system in a volume V has a fixed
temperature T and a fixed chemical potential μ. In this case the Hamiltonian Ĥ has
the spectral decomposition
∞
(N) (N) (N)
Ĥ = Ei |Ei →∀Ei |, (1.92)
N=0 i
16 1 The Origins of Modern Physics

which is a generalization of Eq. (1.75), and one introduces the total number operator
N̂ such that
(N) (N)
N̂|Ei → = N|Ei →, (1.93)

and consequently N̂ has the spectral decomposition

∞
(N) (N)
N̂ = N|Ei →∀Ei |. (1.94)
N=0 i

For the grand canonical ensemble one defines the grand canonical density operator as

ρ̂ = e−β(Ĥ−μN̂) , (1.95)

with β = 1/(kB T ) and μ the chemical potential. Here ρ̂ has the spectral decomposi-
tion
∞
∞

(N) (N) (N) (N) (N) (N)
ρ̂ = e−β(Ei −μN)
|Ei →∀Ei |= zN e−βEi |Ei →∀Ei |, (1.96)
N=0 i N=0 i

where z = eβμ is the fugacity. The key quantity in the grand canonical ensemble is
the grand canonical partition function (or grand canonical volume) calZ given by

Z = Tr[ρ̂] = Tr[e−β(Ĥ−μN̂) ], (1.97)

namely
∞
∞

(N)
−β(Ei −μN)
Z= e = zN ZN . (1.98)
N=0 i N=0

The ensemble average of an observable  is defined as

∞
Tr[Â ρ̂] 1 (N) −βE (N)
∀Â→ = = Aii e i , (1.99)
Tr[ρ̂] Z
N=0 i

where A(N) (N) (N)

ii = ∀Ei |Â|Ei →. Notice that the definition of grand canonical-ensemble
average is the same of both microcanonical-ensemble average and canonical-enseble
average but the density of state ρ̂ is different in the three ensembles. The connection
with Equilibrium Thermodynamics is given by the formula

Z = e−β , (1.100)
1.3 Quantum Statistical Mechanics 17

which introduces the grand potential  as a function of temperature T , volume V

and chemical potential μ. From the grand potential (T , V, μ) the entropy S, the
pressure P and the average number N̄ of particles are obtained as




∂ ∂ ∂
S=− , P=− , N̄ = − , (1.101)
∂T V,μ ∂V T ,μ ∂μ V,T

which are familiar relationships of Equilibrium Thermodynamics such that

d = −SdT − PdV − μdN. (1.102)

In fact, in the grand canonical ensemble the independent thermodynamic variables

are T , μ and V , while S, P and N̄ are dependent thermodynamic variables.
To conclude this section we observe that in the grand canonical ensemble, instead
(N)
of working with eigenstates |Ei → of Ĥ at fixed number N of particles, one can work
with multi-mode Fock states

|n0 n1 n2 . . . n∞ → = |n0 → ⊗ |n1 → ⊗ |n2 → ⊗ · · · ⊗ |n∞ →, (1.103)

where |nα → is the single-mode Fock state which describes nα particles in the single-
mode state |α→ with α = 0, 1, 2, . . .. The trace Tr which appears in Eq. (1.97) is
indeed independent on the basis representation. We analyze this issue in Chaps. 2
and 7.

1.4 Solved Problems

Problem 1.1
The electromagnetic radiation of the black body has the following energy density
per unit of frequency
8π 2 hν
ρ(ν) = 3 ν 2 βhν ,
c e −1

where c is the speed of light in vacuum, h is the Planck constant, and β = 1/(kB T )
with T absolute temperature and kB Boltzmann constant. Determine the correspond-
ing energy density per unit of wavelength.
Solution
The energy density ρ(ν) is such that
 ∞
E= ρ(ν) dν
0
18 1 The Origins of Modern Physics

represents the energy of the radiation per unit of volume, i.e. the energy density. The
linear frequency ν is related to the wavelength λ by the expression

λν = c.

In practice, ν = cλ−1 , from which

dν = −cλ−2 dλ.

By changing variable the energy density becomes

 ∞ 8π 2 hc
E= βhc/λ
dλ,
0 λ e
5 −1

and consequently the energy density per unit of wavelength reads

8π 2 hc
ρ(λ) = ,
λ5 eβhc/λ − 1

such that  ∞
E= ρ(λ) dλ.
0

Problem 1.2
Calculate the number of photon emitted in 4 s by a lamp of 10 W which radiates
1% of its energy as monochromatic light with wavelength 6, 000 × 10−10 m (orange
light).
Solution
The energy of one photon of wavelength λ and linear frequency ν is given by
c
 = hν = h ,
λ
where
h = 6.63 × 10−34 J s

is the Planck constant. In our problem one gets

c 3 × 108 m/s
=h = 6.62 × 10−34 J s = 3.3 × 10−19 J.
λ 6 × 103 × 10−10 m

During the period t = 4 s the energy of the lamp with power P = 10 W is

E = P t = 10 J s × 4 s = 40 J.
1.4 Solved Problems 19

The radiation energy is instead

1 40
Erad = E × 1 % = E × = J = 0.4 J.
100 100
The number of emitted photons is then

Erad 0.4 J
N= = = 1.2 × 1018 .
 3.3 × 10−19 J

Problem 1.3
On a photoelectric cell it arrives a beam of light with wavelength 6, 500 × 10−10
m and energy 106 erg per second [1 erg = 10−7 J]. This energy is entirely used to
produce photoelectrons. Calculate the intensity of the electric current which flows
in the electric circuit connected to the photoelectric cell.
Solution
The wavelength can be written as

λ = 6.5 × 103 × 10−10 m = 6.5 × 10−7 m.

The energy of the beam of light can be written as

E = 106 erg = 106 × 10−7 J = 10−1 J.

The time interval is

t = 1 s.

The energy of a single photon reads

c 3 × 108
=h = 6.6 × 10−34 J = 3 × 10−19 J,
λ 6.5 × 10−7

where h is the Planck constant and c the speed of light in the vacuum.
The number of photons is thus given by

E 10−1 J
N= = = 3.3 × 1017 .
 3 × 10−19 J

If one photon produces one electron with electric charge e = −1.6 × 10−19 C, the
intensity of electric current is easily obtained:

|e|N 1.6 × 10−19 C × 3.3 × 1017

I= = = 5.5 × 10−2 A.
t 1s
20 1 The Origins of Modern Physics

To conclude we observe that with wavelength 6,500 ×10−10 m the photoelectric

effect is possible only if the work function of the sample is reduced, for instance by
using an external electric field.
Problem 1.4
Prove that the simple 2-qubit state

|2 → = c00 |01→ + c10 |10→

is separable if and only if c01 = 0 or c10 = 0.

Solution
Let us consider two generic 1-qubits, given by

|ψA → = αA |0→ + βA |1→, |ψB → = αB |0→ + βB |1→.

Then
|ψA →|ψB → = αA αB |00→ + αA βB |01→ + βA αB |10→ + βA βB |11→.

If the 2-qubit state |2 → is separable, it can be written as

|2 → = |ψA → ⊗ |ψB →,

namely

c00 |01→ + c10 |10→ = αA αB |00→ + αA βB |01→ + βA αB |10→ + βA βB |11→.

It follows
αA αB = βA βB = 0, αA βB = c01 , βA αB = c10 .

These conditions are all satisfied only if c01 = 0 or c10 = 0, or both.

Further Reading

For special relativity:

W. Rindler, Introduction to Special Relativity (Oxford Univ. Press, Oxford, 1991)
For quantum mechanics:
C. Cohen-Tannoudji, B. Dui, F. Laloe, Quantum Mechanics (Wiley, New York, 1991)
P.A.M. Dirac, The Principles of Quantum Mechanics (Oxford University Press,
Oxford, 1982)
For quantum information:
M. Le Bellac, A Short Introduction to Quantum Information and Quantum Compu-
tation (Cambridge University Press, Cambridge, 2006)
For quantum statistical mechanics:
K. Huang, Statistical Mechanics (Wiley, New York, 1987)
Chapter 2
Second Quantization of Light

In this chapter we discuss the quantization of electromagnetic waves, which we also

denoted as light (visible or invisible to human eyes). After reviewing classical and
quantum properties of the light in the vacuum, we discuss the so-called second quan-
tization of the light field showing that this electromagnetic field can be expressed as
a infinite sum of harmonic oscillators. These oscillators, which describe the possible
frequencies of the radiation field, are quantized by introducing creation and annihi-
lation operators acting on the Fock space of number representation. We analyze the
Fock states of the radiation field and compare them with the coherent states. Finally,
we consider two enlightening applications: the Casimir effect and the radiation field
at finite temperature.

2.1 Electromagnetic Waves

The light is an electromagnetic field characterized by the coexisting presence of an

electric field E(r, t) and a magnetic field B(r, t). From the equations of James Clerk
Maxwell in vacuum and in the absence of sources, given by

∇ · E = 0, (2.1)
∇ · B = 0, (2.2)
νB
∇∧E = − , (2.3)
νt
νE
∇ ∧ B = ρ0 μ0 , (2.4)
νt
one finds that the coupled electric and magnetic fields satisfy the wave equations
(Fig. 2.1)

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 21

DOI: 10.1007/978-3-319-05179-6_2, © Springer International Publishing Switzerland 2014
22 2 Second Quantization of Light

Fig. 2.1 Plot of a sinusoidal electromagnetic wave moving along the x axis

 
1 ν2
−∇ 2
E = 0, (2.5)
c2 νt 2
 
1 ν2
− ∇ 2 B = 0, (2.6)
c2 νt 2

where
1
c= ∇ = 3 × 108 m/s (2.7)
ρ0 μ0

is the speed of light in the vacuum. Note that the dielectric constant (electric per-
mittivity) ρ0 and the magnetic constant (magnetic permeability) μ0 are respectively
ρ0 = 8.85 × 10−12 C2 /(N m2 ) and μ0 = 4π × 10−7 V · s/(A · m). Equations (2.5)
and (2.6), which are fully confirmed by experiments, admit monochromatic complex
plane wave solutions

E(r, t) = E0 ei(k·r−βt) , (2.8)

B(r, t) = B0 e i(k·r−βt)
, (2.9)

where k is the wavevector and β the angular frequency, such that

β = c k, (2.10)

is the dispersion relation, with k = |k| is the wavenumber. From Maxwell’s equations
one finds that the vectors E and B are mutually orthogonal and such that

E = cB, (2.11)

where E = |E| and B = |B|. In addition they are transverse fields, i.e. orthogonal to
the wavevector k, which gives the direction of propagation of the wave. Notice that
Eq. (2.11) holds for monocromatic plane waves but not in general. For completeness,
let us remind that the wavelength λ is given by

2π
λ= , (2.12)
k
2.1 Electromagnetic Waves 23

and that the linear frequency ε and the angular frequency β = 2πε are related to the
wavelength λ and to the wavenumber k by the formulas
β
λε = = c. (2.13)
k

2.1.1 First Quantization of Light

At the beginning of quantum mechanics Satyendra Nath Bose and Albert Einstein
suggested that the light can be described as a gas of photons. A single photon of a
monochromatic wave has the energy

λ = hε = β, (2.14)

where h = 6.63×10−34 J· s is the Planck constant and  = h/(2π) = 1.05×10−34 J· s

is the reduced Planck constant. The linear momentum of the photon is given by the
de Broglie relations
h
p = n = k, (2.15)
λ
where n is a unit vector in the direction of k. Clearly, the energy of the photon can
be written also as
λ = p c, (2.16)

which is the energy one obtains for a relativistic particle of energy


λ= m 2 c4 + p 2 c2 , (2.17)

setting to zero the rest mass, i.e. m = 0. The total energy H of monochromatic wave
is given by 
H= β n s , (2.18)
s

where n s is the number of photons with angular frequency β and polarization s in the
monochromatic electromagnetic wave. Note that in general there are two possible
polarizations: s = 1, 2, corresponding to two linearly independent orthogonal unit
vectors ε1 and ε2 in the plane perpendicular to the wavevector k.
A generic electromagnetic field is the superposition of many monochromatic
electromagnetic waves. Calling βk the angular frequency of the monochromatic
wave with wavenumber k, the total energy H of the electromagnetic field is

H= βk n ks , (2.19)
k s
24 2 Second Quantization of Light

where n ks is the number of photons with wavevector k and polarization s. The

results derived here for the electromagnetic field, in particular Eq. (2.19), are called
semiclassical, or first-quantization results, because they do not take into account the
so-called “quantum fluctuations of vacuum”, i.e. the following remarkable experi-
mental fact: photons can emerge from the vacuum of the electromagnetic field. To
justify this property of the electromagnetic field one must perform the so-called
second quantization of the field.

2.1.2 Electromagnetic Potentials and Coulomb Gauge

In full generality the electric field E(r, t) and the magnetic field B(r, t) can be
expressed in terms of a scalar potential θ(r, t) and a vector potential A(r, t) as
follows
νA
E = −∇θ − , (2.20)
νt
B = ∇ ∧ A. (2.21)

Actually these equations do not determine the electromagnetic potentials uniquely,

since for an arbitrary scalar function (r, t) the so-called “gauge transformation”

ν
θ → θ∀ = θ + , (2.22)
νt
A → A∀ = A − ∇, (2.23)

leaves the fields E and B unaltered. There is thus an infinite number of different
electromagnetic potentials that correspond to a given configuration of measurable
fields. We use this remarkable property to choose a gauge transformation such that

∇ · A = 0. (2.24)

This condition defines the Coulomb (or radiation) gauge, and the vector field A is
called transverse field. For a complex monochromatic plane wave

A(r, t) = A0 ei(k·r−βt) (2.25)

the Coulomb gauge (2.24) gives

k · A = 0, (2.26)

i.e. A is perpendicular (transverse) to the wavevector k. In the vacuum and without

sources, from the first Maxwell equation (2.1) and Eq. (2.20) one immediately finds
2.1 Electromagnetic Waves 25

ν
∇2θ + (∇ · A) = 0, (2.27)
νt
and under the Coulomb gauge (2.24) one gets

∇ 2 θ = 0. (2.28)

Imposing that the scalar potential is zero at infinity, this Laplace’s equation has the
unique solution
θ(r, t) = 0, (2.29)

and consequently

νA
E=− , (2.30)
νt
B = ∇ ∧ A. (2.31)

Thus, in the Coulomb gauge one needs only the electromagnetic vector potential
A(r, t) to obtain the electromagnetic field if there are no charges and no currents.
Notice that here E and B are transverse fields like A, which satisfy Eqs. (2.5) and (2.6).
The electromagnetic field described by these equations is often called the radiation
field, and also the vector potential satisfies the wave equation
 
1 ν2
− ∇2 A = 0. (2.32)
c2 νt 2

We now expand the vector potential A(r, t) as a Fourier series of monochromatic

plane waves. The vector potential is a real vector field, i.e. A = A√ and consequently
we write
  eik·r e−ik·r

√
A(r, t) = Aks (t) ∇ + Aks (t) ∇ εks , (2.33)
k s V V

where Aks (t) and A√ks (t) are the dimensional complex conjugate coefficients of the
∇
expansion, the complex plane waves eik·r / V normalized in a volume V are the
basis functions of the expansion, and εk1 and εk2 are two mutually orthogonal real
unit vectors of polarization which are also orthogonal to k (transverse polariza-
tion vectors).
Taking into account Eqs. (2.30) and (2.31) we get

  eik·r e−ik·r

E(r, t) = − Ȧks (t) ∇ + Ȧ√ks (t) ∇ εks , (2.34)
k s V V
  eik·r e−ik·r

B(r, t) = Aks (t) ∇ − A√ks (t) ∇ ik ∧ εks , (2.35)
k s V V
26 2 Second Quantization of Light

where both vector fields are explicitly real fields. Moreover, inserting Eq. (2.33)
into Eq. (2.32) we recover familiar differential equations of decoupled harmonic
oscillators:
Äks (t) + βk2 Aks (t) = 0, (2.36)

with βk = ck, which have the complex solutions

Aks (t) = Aks (0) e−iβk t . (2.37)

These are the complex amplitudes of the infinite harmonic normal modes of the
radiation field.
A familiar result of electromagnetism is that the classical energy of the electro-
magnetic field in vacuum is given by
  
ρ0 1
H= d 3r E(r, t)2 + B(r, t)2 , (2.38)
2 2μ0

namely  ρ
νA(r, t)  1
2 
0 2
H= d 3r + ∇ ∧ A(r, t) (2.39)
2 νt 2μ0

by using the Maxwell equations in the Coulomb gauge (2.30) and (2.31). Inserting
into this expression Eq. (2.33) or Eqs. (2.34) and (2.35) into Eq. (2.38) we find


H= λ0 βk2 A√ks Aks + Aks A√ks . (2.40)
k s

√ (t)
It is now convenient to introduce adimensional complex coefficients aks (t) and aks
√
related to the dimensional complex coefficients Aks (t) and Aks (t) by


Aks (t) = aks (t). (2.41)
2ρ0 βk


A√ks (t) = a √ (t). (2.42)
2ρ0 βk ks

In this way the energy H reads

  βk

√ √
H= aks aks + aks aks . (2.43)
s
2
k

This energy is actually independent on time: the time dependence of the complex
√ (t) and a (t) cancels due to Eq. (2.37). Instead of using the complex
amplitudes aks ks
√ (t) and a (t) one can introduce the real variables
amplitudes aks ks
2.1 Electromagnetic Waves 27

2 1
√

qks (t) = aks (t) + aks (t) (2.44)
βk 2
 1

√
pks (t) = 2βk aks (t) − aks (t) (2.45)
2i
such that the energy of the radiation field reads
 
 2
pk,s1 2 2
H= + βk qks . (2.46)
s
2 2
k

This energy resembles that of infinitely many harmonic oscillators with unitary mass
and frequency βk . It is written in terms of an infinite set of real harmonic oscillators:
two oscillators (due to polarization) for each mode of wavevector k and angular
frequency βk .

2.2 Second Quantization of Light

In 1927 Paul Dirac performed the quantization of the classical Hamiltonian (2.46)
by promoting the real coordinates qks and the real momenta pks to operators:

qks → q̂ks , (2.47)

pks → p̂ks , (2.48)

satisfying the commutation relations

[q̂ks , p̂k∀ s ∀ ] = i ∂k,k∀ ∂s,s ∀ , (2.49)

where [ Â, B̂] = Â B̂ − B̂ Â. The quantum Hamiltonian is thus given by

 
 2
p̂k,s
1 2 2
Ĥ = + βk q̂ks . (2.50)
s
2 2
k

The formal difference between Eqs. (2.46) and (2.50) is simply the presence of the
“hat symbol” in the canonical variables.
Following a standard approach for the canonical quantization of the Harmonic oscil-
lator, we introduce annihilation and creation operators
 
βk i
âks = q̂ks + p̂ks , (2.51)
2 βk
  
+ βk i
âks = q̂ks − p̂ks , (2.52)
2 βk
28 2 Second Quantization of Light

which satisfy the commutation relations

[âks , âk+∀ s ∀ ] = ∂k,k∀ ∂s,s ∀ , (2.53)

and the quantum Hamiltonian (2.50) becomes

  
+ 1
Ĥ = βk âks âks + . (2.54)
s
2
k

Obviously this quantum Hamiltonian can be directly obtained from the classical one,
√ to operators:
given by Eq. (2.43), by promoting the complex amplitudes aks and aks

aks → âks , (2.55)

√ +
aks → âks , (2.56)

satisfying the commutation relations (2.53).

+
The operators âks and âks act in the Fock space F, i.e. the infinite dimensional
Hilbert space of “number representation” introduced in 1932 by Vladimir Fock. A
generic state of this Fock space F is given by

| . . . n ks . . . n k∀ s ∀ . . . n k∀∀ s ∀∀ . . . ⊗, (2.57)

meaning that there are n ks photons with wavevector k and polarization s, n k∀ s ∀ pho-
tons with wavevector k∀ and polarization s ∀ , n k∀∀ s ∀∀ photons with wavevector k∀∀ and
polarization s ∀∀ , et cetera. The Fock space F is given by
∈

F = H0 ∞ H1 ∞ H2 ∞ H3 ∞ · · · ∞ H∈ = Hn , (2.58)
n=0

where
Hn = H ∓ H ∓ · · · ∓ H = H∓n (2.59)

is the Hilbert space of n identical photons, which is n times the tensor product ∓ of
the single-photon Hilbert space H = H∓1 . Thus, F is the infinite direct sum ∞ of
increasing n-photon Hilbert states Hn , and we can formally write
∈

F= H∓n . (2.60)
n=0

Notice that in the definition of the Fock space F one must include the space H0 =
H∓0 , which is the Hilbert space of 0 photons, containing only the vacuum state

|0⊗ = | . . . 0 . . . 0 . . . 0 . . . ⊗, (2.61)
2.2 Second Quantization of Light 29

+
and its dilatations ψ|0⊗ with ψ a generic complex number. The operators âks and âks
are called annihilation and creation operators because they respectively destroy and
create one photon with wavevector k and polarization s, namely
∇
âks | . . . n ks . . . ⊗ = n ks | . . . n ks − 1 . . . ⊗, (2.62)
+

âks | . . . n ks . . . ⊗ = n ks + 1 | . . . n ks + 1 . . . ⊗. (2.63)

Note that these properties follow directly from the commutation relations (2.53).
Consequently, for the vacuum state |0⊗ one finds

âks |0⊗ = 0 F , (2.64)

+
âks |0⊗ = |1ks ⊗ = |ks⊗, (2.65)

where 0 F is the zero of the Fock space (usually indicated with 0), and |ks⊗ is clearly
the state of one photon with wavevector k and polarization s, such that

eik·r
↑r|ks⊗ = ∇ εks . (2.66)
V

From Eqs. (2.62) and (2.63) it follows immediately that

+
N̂ks = âks âks (2.67)

is the number operator which counts the number of photons in the single-particle
state |ks⊗, i.e.
N̂ks | . . . n ks . . . ⊗ = n ks | . . . n ks . . . ⊗. (2.68)

Notice that the quantum Hamiltonian of the light can be written as

  
1
Ĥ = βk N̂ks + , (2.69)
s
2
k

and this expression is very similar, but not equal, to the semiclassical formula (2.19).
The differences are that N̂ks is a quantum number operator and that the energy E vac
of the the vacuum state |0⊗ is not zero but is instead given by
 1
E vac = βk . (2.70)
s
2
k

A quantum harmonic oscillator of frequency βk has a finite minimal energy βk ,

which is called zero-point energy. In the case of the quantum electromagnetic field
there is an infinite number of harmonic oscillators and the total zero-point energy,
30 2 Second Quantization of Light

given by Eq. (2.70), is clearly infinite. The infinite constant E vac is usually eliminated
by simply shifting to zero the energy associated to the vacuum state |0⊗.
The quantum electric and magnetic fields can be obtained from the classical
expressions, Eqs. (2.34) and (2.35), taking into account the quantization of the clas-
sical complex amplitudes aks and aks √ and their time-dependence, given by Eq. (2.37)
and its complex conjugate. In this way we obtain

 βk  + −(ik·r−βk t)

Ê(r, t) = i âks ei(k·r−βk t) − âks e εks , (2.71)
s
2ρ0 V
k

    k
+ −i(k·r−βk t)
B̂(r, t) = âks ei(k·r−βk t) − âks e i ∧ εks . (2.72)
s
2ρ0 βk V |k|
k

It is important to stress that the results of our canonical quantization of the radi-
ation field suggest a remarkable philosophical idea: there is a unique quantum elec-
tromagnetic field in the universe and all the photons we see are the massless particles
associated to it. In fact, the quantization of the electromagnetic field is the first step
towards the so-called quantum field theory or second quantization of fields, where all
particles in the universe are associated to few quantum fields and their corresponding
creation and annihilation operators.

2.2.1 Fock Versus Coherent States for the Light Field

Let us now consider for simplicity a linearly polarized monochromatic wave of the
radiation field. For instance, let us suppose that the direction of polarization is given
by the vector ε. From Eq. (2.71) one finds immediately that the quantum electric
field can be then written in a simplified notation as

β  
Ê(r, t) = i â ei(k·r−βt) − â + e−i(k·r−βt) ε (2.73)
2ρ0 V

where β = βk = c|k|. Notice that, to simplify the notation, we have removed the
subscripts in the annihilation and creation operators â and â + . If there are exactly n
photons in this polarized monochromatic wave the Fock state of the system is given
by
1
+ n
|n⊗ = ∇ â |0⊗. (2.74)
n!

It is then straightforward to show, by using Eqs. (2.62) and (2.63), that

↑n|Ê(r, t)|n⊗ = 0, (2.75)

2.2 Second Quantization of Light 31

for all values of the photon number n, no matter how large. This result holds for
all modes, which means then that the expectation value of the electric field in any
many-photon Fock state is zero. On the other hand, the expectation value of Ê(r, t)2
is given by  
β 1
↑n| Ê(r, t) |n⊗ =
2
n+ . (2.76)
ρ0 V 2

For this case, as for the energy, the expectation value is nonvanishing even when
n = 0, with the result that for the total field (2.71), consisting of an infinite number
of modes, the expectation value of Ê(r, t)2 is infinite for all many-photon states,
including the vacuum state. Obviously a similar reasoning applies for the magnetic
field (2.72) and, as discussed in the previous section, the zero-point constant is usually
removed.
One must remember that the somehow strange result of Eq. (2.75) is due to the
fact that the expectation value is performed with the Fock state |n⊗, which means
that the number of photons is fixed because

N̂ |n⊗ = n|n⊗. (2.77)

Nevertheless, usually the number of photons in the radiation field is not fixed, in
other words the system is not in a pure Fock state. For example, the radiation field of
a well-stabilized laser device operating in a single mode is described by a coherent
state |φ⊗, such that
â|φ⊗ = φ|φ⊗, (2.78)

with
↑φ|φ⊗ = 1. (2.79)

The coherent state |φ⊗, introduced in 1963 by Roy Glauber, is thus the eigenstate of
the annihilation operator â with complex eigenvalue φ = |φ|eiα . |φ⊗ does not have
a fixed number of photons, i.e. it is not an eigenstate of the number operator N̂ , and
it is not difficult to show that |φ⊗ can be expanded in terms of number (Fock) states
|n⊗ as follows
∈
−|φ|2 /2 φn
|φ⊗ = e ∇ |n⊗. (2.80)
n=0
n!

From Eq. (2.78) one immediately finds

N̄ = ↑φ| N̂ |φ⊗ = |φ|2 , (2.81)

and it is natural to set 

φ= N̄ eiα , (2.82)
32 2 Second Quantization of Light

where N̄ is the average number of photons in the coherent state, while α is the phase
of the coherent state. For the sake of completeness, we observe that

↑φ| N̂ 2 |φ⊗ = |φ|2 + |φ|4 = N̄ + N̄ 2 (2.83)

and consequently
↑φ| N̂ 2 |φ⊗ − ↑φ| N̂ |φ⊗2 = N̄ , (2.84)

while
↑n| N̂ 2 |n⊗ = n 2 (2.85)

and consequently
↑n| N̂ 2 |n⊗ − ↑n| N̂ |n⊗2 = 0. (2.86)

The expectation value of the electric field Ê(r, t) of the linearly polarized mono-
chromatic wave, Eq. (2.73), in the coherent state |φ⊗ reads

2 N̄ β
↑φ|Ê(r, t)|φ⊗ = − sin(k · r − βt + α) ε, (2.87)
ρ0 V

while the expectation value of Ê(r, t)2 is given by

2 N̄ β
↑φ| Ê(r, t)2 |φ⊗ = sin2 (k · r − βt + α). (2.88)
ρ0 V

These results suggest that the coherent state is indeed a useful tool to investigate the
correspondence between quantum field theory and classical field theory. Indeed, in
1965 at the University of Milan Fortunato Tito Arecchi experimentally verified that a
single-mode laser is in a coherent state with a definite but unknown phase. Thus, the
coherent state gives photocount statistics that are in accord with laser experiments
and has coherence properties similar to those of a classical field, which are useful
for explaining interference effects.
To conclude this subsection we observe that we started the quantization of the light
field by expanding the vector potential A(r, t) as a Fourier series of monochromatic
plane waves. But this is not the unique choice. Indeed, one can expand the vec-
tor potential A(r, t) by using any orthonormal set of wavefunctions (spatio-temporal
modes) which satisfy the Maxwell equations. After quantization, the expansion coef-
ficients of the single mode play the role of creation and annihilation operators of
that mode.
2.2 Second Quantization of Light 33

2.2.2 Linear and Angular Momentum of the Radiation Field

In addition to the total energy there are other interesting conserved quantities which
characterize the classical radiation field. They are the total linear momentum

P= d 3 r ρ0 E(r, t) ∧ B(r, t) (2.89)

and the total angular momentum


J= d 3 r ρ0 r ∧ (E(r, t) ∧ B(r, t)). (2.90)

By using the canonical quantization one easily finds that the quantum total linear
momentum operator is given by

  
1
P̂ = k N̂ks + , (2.91)
s
2
k

where k is the linear momentum of a photon of wavevector k.

The quantization of the total angular momentum J is a more intricate problem.
In fact, the linearly polarized states |ks⊗ (with s = 1, 2) are eigenstates of Ĥ and
P̂, but they are not eigenstates of Ĵ, nor of Jˆ2 , and nor of Jˆz . The eigenstates of
Jˆ2 and Jˆz do not have a fixed wavevector k but only a fixed wavenumber k = |k|.
These states, which can be indicated as |k jm j s⊗, are associated to vector spherical
harmonics Y j,m j (α, θ), with α and θ the spherical angles of the wavevector k. These
states |k jm j s⊗ are not eigenstates of P̂ but they are eigenstates of Ĥ , Jˆ2 and Jˆz ,
namely

Ĥ |k jm j s⊗ = ck|k jm j s⊗ (2.92)
Jˆ2 |k jm j ⊗ = j ( j + 1)2 |k jm j s⊗ (2.93)
Jˆz |k jm j s⊗ = m j |k jm j s⊗ (2.94)

with m j = − j, − j + 1, . . . , j − 1, j and j = 1, 2, 3, . . .. In problems where the

charge distribution which emits the electromagnetic radiation has spherical symmetry
it is indeed more useful to expand the radiation field in vector spherical harmonics
than in plane waves. Clearly, one can write

|ks⊗ = c jm j (α, θ)|k jm j s⊗, (2.95)
jm j
34 2 Second Quantization of Light

where the coefficients c jm j (α, θ) of this expansion give the amplitude probability of
finding a photon of fixed wavenumber k = |k| and orbital quantum numbers j and
m j with spherical angles α and θ.

2.2.3 Zero-Point Energy and the Casimir Effect

There are situations in which the zero-point energy E vac of the electromagnetic field
oscillators give rise to a remarkable quantum phenomenon: the Casimir effect.
The zero-point energy (2.70) of the electromagnetic field in a region of volume
V can be written as
 
 1 d 3k  2
E vac = c k x2 + k 2y + k z2 = V c k x + k 2y + k z2 (2.96)
s
2 (2π)3
k

by using the dispersion relation βk = ck = c k 2y + k 2y + k z2 . In particular, consid-
ering a region with the shape of a parallelepiped of length L along both x and y and
length a along z, the volume V is given by V = L 2 a and the vacuum energy E vac
in the region is
  
+∈ L dk x +∈ L dk y +∈ a dk z  2
E vac = c k x + k 2y + k z2
−∈ 2π −∈ 2π −∈ 2π

 
c 2 ∈ ∈ n2 π2 
= L dk↓ k↓ dn k↓2 + , (2.97)
2π 0 0 a2

where the second expression is obtained setting k↓ = k x2 + k 2y and n = (a/π)k z
(Fig. 2.2).
Let us now consider the presence of two perfect metallic plates with the shape
of a square of length L having parallel faces lying in the (x, y) plane at distance a.
Along the z axis the stationary standing waves of the electromagnetic field vanishes
on the metal plates and the k z component of the wavevector k is no more a continuum
variabile but it is quantized via
π
kz = n , (2.98)
a
where now n = 0, 1, 2, . . . is an integer number, and not a real number as in Eq.
(2.97). In this case the zero-point energy in the volume V = L 2 a between the two
plates reads
 ∈ k n2 π2 
∈

∀ c ↓
E vac = L2 dk↓ k↓ + k↓2 + 2 , (2.99)
2π 0 2 a
n=1
2.2 Second Quantization of Light 35

Fig. 2.2 Graphical represen-

tation of the parallel plates in
the Casimir effect

where for k z = 0 the state |(k z , k y , 0)1⊗ with polarization σk1 parallel to the z axis
(and orthogonal to the metallic plates) is bound, while the state |(k z , k y , 0)2⊗ with
polarization σk2 orthogonal to the z axis (and parallel to the metallic plates) is not
bound and it does not contribute to the discrete summation.
The difference between E vac∀ and E vac divided by L 2 gives the net energy per
unit surface area E, namely
 ∈
c π 3  1 
∀ ∈

E vac − E vac
E= = A(0) + A(n) − dn A(n) , (2.100)
L2 2π a 2 0n=1

where we have defined

 +∈  1 
A(n) = dσ σ σ 2 + n 2 = (n 2 + ∈)3/2 − n 3 , (2.101)
0 3

with σ = (a/π)k↓ . Notice that A(n) is clearly divergent but Eq. (2.100) is not
divergent due to the cancellation of divergences with opposite sign. In fact, by using
the Euler-MacLaurin formula for the difference of infinite series and integrals

 ∈  ∈
1 1 dA 1 d3 A
A(0) + A(n) − dn A(n) = − (0) + (0)
2 0 6 · 2! dn 30 · 4! dn 3
n=1
1 d5 A
− (0) + · · · (2.102)
42 · 6! dn 5
and Eq. (2.101) from which

dA d3 A d5 A
(0) = 0, (0) = −2, (0) = 0 (2.103)
dn dn 3 dn 5
36 2 Second Quantization of Light

and all higher derivatives of A(n) are zero, one eventually obtains

π 2 c
E =− . (2.104)
720 a 3
From this energy difference E one deduces that there is an attractive force per unit
area F between the two plates, given by

dE π 2 c
F =− =− . (2.105)
da 240 a 4
Numerically this result, predicted in 1948 by Hendrik Casimir during his research
activity at the Philips Physics Laboratory in Eindhoven, is very small

1.30 × 10−27 N m2
F =− . (2.106)
a4
Nevertheless, it has been experimentally verified by Steven Lamoreaux in 1997 at
the University of Washington and by Giacomo Bressi, Gianni Carugno, Roberto
Onofrio, and Giuseppe Ruoso in 2002 at the University of Padua.

2.3 Quantum Radiation Field at Finite Temperature

Let us consider the quantum radiation field in thermal equilibrium with a bath at the
temperature T . The relevant quantity to calculate all thermodynamical properties of
the system is the grand-canonical partition function Z, given by

Z = T r [e−δ( Ĥ −μ N̂ ) ] (2.107)

where δ = 1/(k B T ) with k B = 1.38 × 10−23 J/K the Boltzmann constant,


Ĥ = βk N̂ks , (2.108)
k s

is the quantum Hamiltonian without the zero-point energy,


N̂ = N̂ks (2.109)
k s

is the total number operator, and μ is the chemical potential, fixed by the conservation
of the particle number. For photons μ = 0 and consequently the number of photons
is not fixed. This implies that
2.3 Quantum Radiation Field at Finite Temperature 37

Z= ↑ . . . n ks . . . |e−δ Ĥ | . . . n ks . . . ⊗
{n ks }
 
βk N̂ks
= ↑ . . . n ks . . . |e−δ ks | . . . n ks . . . ⊗
{n ks }
   
βk n ks
= e−δ ks = e−δ βk n ks
{n ks } {n ks } ks
  ∈
= e−δ βk n ks = e−δ βk n
ks n ks ks n=0
 1
= . (2.110)
ks
1 − e−δ βk

Quantum statistical mechanics dictates that the thermal average of any operator  is
obtained as
1
↑ Â⊗T = T r [ Â e−δ( Ĥ −μ N̂ ) ]. (2.111)
Z

In our case the calculations are simplified because μ = 0. Let us suppose that  = Ĥ ,
it is then quite easy to show that

1 ν  ν
↑ Ĥ ⊗T = T r [ Ĥ e−δ Ĥ ] = − ln T r [e−δ Ĥ ] = − ln(Z). (2.112)
Z νδ νδ

By using Eq. (2.110) we immediately obtain

 
ln(Z) = ln 1 − e−δ βk , (2.113)
k s

and finally from Eq. (2.112) we get

 βk 
↑ Ĥ ⊗T = = βk ↑ N̂ks ⊗T . (2.114)
k s
eδ βk −1 k s

In the continuum limit, where

 
d 3k
→V , (2.115)
(2π)3
k

with V the volume, and taking into account that βk = ck, one can write the energy
density E = ↑ Ĥ ⊗T /V as
  ∈
d 3k ck c k3
E =2 
= 2 dk , (2.116)
(2π) e
3 δc k −1 π 0 eδck − 1
38 2 Second Quantization of Light

where the factor 2 is due to the two possible polarizations (s = 1, 2). By using
β = ck instead of k as integration variable one gets
 ∈  ∈
 β3
E= 2 3 dβ δ β = dβ ρ(β), (2.117)
π c 0 e −1 0

where
 β3
ρ(β) = (2.118)
π 2 c3 eδ β − 1

is the energy density per frequency, i.e. the familiar formula of the black-body radi-
ation, obtained for the first time in 1900 by Max Planck. The previous integral can
be explicitly calculated and it gives

π 2 k 4B 4
E= T , (2.119)
15c3 3
which is nothing else than the Stefan-Boltzmann law. In an similar way one deter-
mines the average number density of photons:
 ∈
↑ N̂ ⊗T 1 β2 2σ(3)k 3B 3
n= = 2 3 dβ = T . (2.120)
V π c 0 eδ β − 1 π 2 c 3 3

where σ(3)  1.202. Notice that both energy density E and number density n of
photons go to zero as the temperature T goes to zero. To conclude this section, we
stress that these results are obtained at thermal equilibrium and under the condition of
a vanishing chemical potential, meaning that the number of photons is not conserved
when the temperature is varied.

2.4 Phase Operators

We have seen that the ladder operators â and â + of a single mode of the electromag-
netic field satisfy the fundamental relations
∇
â|n⊗ = n |n − 1⊗, (2.121)
∇
â + |n⊗ = n + 1 |n + 1⊗, (2.122)

where |n⊗ is the Fock state, eigenstate of the number operator N̂ = â + â and describ-
ing n photons in the mode. Ramarkably, these operator can be expessed in terms of
the Fock states as follows
2.4 Phase Operators 39

∈ ∇
 ∇ ∇
â = n + 1|n⊗↑n + 1| = |0⊗↑1| + 2|1⊗↑2| + 3|2⊗↑3| + · · · , (2.123)
n=0
∈ ∇ ∇ ∇
â + = n + 1|n + 1⊗↑n| = |1⊗↑0| + 2|2⊗↑1| + 3|3⊗↑2| + · · · . (2.124)
n=0

It is in fact straightforward to verify that the expressions (2.123) and (2.124) imply
Eqs. (2.121) and (2.122).
We now introduce the phase operators

fˆ = â N̂ −1/2 , fˆ+ = N̂ −1/2 â + , (2.125)

which can be expressed as

∈

fˆ = |n⊗↑n + 1| = |0⊗↑1| + |1⊗↑2| + |2⊗↑3| + · · · , (2.126)
n=0
∈
fˆ+ = |n + 1⊗↑n| = |1⊗↑0| + |2⊗↑1| + |3⊗↑2| + · · · , (2.127)
n=0

and satisfy the very nice formulas

fˆ|n⊗ = |n − 1⊗, (2.128)

ˆ+
f |n⊗ = |n + 1⊗, (2.129)

showing that the phase operators act as lowering and raising operators of Fock states
without the complication of coefficients in front of the obtained states.
The shifting property of the phase operators on the Fock states |n⊗ resembles that
of the unitary operator ei p̂/ acting on the position state |x⊗ of a 1D particle, with p̂
the 1D linear momentum which is canonically conjugated to the 1D position operator
x̂. In particular, one has

ei p̂/|x⊗ = |x − 1⊗, (2.130)

e−i p̂/|x⊗ = |x + 1⊗. (2.131)

Due to these formal analogies, in 1927 Fritz London and Paul Dirac independently
suggested that the phase operator fˆ can be written as

ˆ
fˆ = ei  , (2.132)

where  ˆ is the angle operator related to the number operator N̂ = â + â. But, contrary
to e i p̂/ , the operator fˆ is not unitary because

fˆ fˆ+ = 1̂, fˆ+ fˆ = 1̂ − |0⊗↑0|. (2.133)

40 2 Second Quantization of Light

ˆ is not Hermitian and

This implies that the angle operator 

ˆ+
fˆ+ = e−i  . (2.134)

In addition, a coherent state |φ⊗ is not an eigenstate of the phase operator fˆ. However,
one easily finds
∈
2  |φ|2n
↑φ| fˆ|φ⊗ = φe−|φ| ∇ , (2.135)
n=0
n!(n + 1)!

∈ n ex
and observing that for x  1 one gets n=0
∇ x
n!(n+1)!
 ∇
x
this implies

φ
↑φ| fˆ|φ⊗  = eiα , (2.136)
|φ|
∇
for N̄ = |φ|2  1, with φ = N̄ eiα . Thus, for a large average number N̄ of photons
the expectation value of the phase operator fˆ on the coherent state |φ⊗ gives the phase
factor eiα of the complex eigenvalue φ of the coherent state.

2.5 Solved Problems

Problem 2.1
Show that the eigenvalues n of the operator N̂ = â + â are non negative.
Solution
The eigenvalue equation of N̂ reads

â + â|n⊗ = n|n⊗.

We can then write

↑n|â + â|n⊗ = n↑n|n⊗ = n,

because the eigenstate |n⊗ is normalized to one. On the other hand, we have also

↑n|â + â|n⊗ = (â|n⊗)+ (â|n⊗) = |(â|n⊗)|2

Consequently, we get
n = |(â|n⊗)|2 ≥ 0.

We stress that we have not used the commutations relations of â and â + . Thus we
have actually proved that any operator N̂ given by the factorization of a generic
operator â with its self-adjunct â + has a non negative spectrum.
2.5 Solved Problems 41

Problem 2.2
Consider the operator N̂ = â + â, where â and â + satisfy the commutation rule
â â + − â + â = 1. Show that if |n⊗ is an eigenstate of N̂ with eigenvalue n then a|n⊗
is an eigenstate of N̂ with eigenvalue n − 1 and a + |n⊗ is an eigenstate of N̂ with
eigenvalue n + 1.
Solution
We have
N̂ â|n⊗ = (â + â)â|n⊗.

The commutation relation between â and â + can be written as

â + â = â â + − 1.

This implies that

N̂ â|n⊗ = (â â + − 1)â|n⊗ = â(â + â − 1)|n⊗ = â( N̂ − 1)|n⊗ = (n − 1)â|n⊗.

Finally, we obtain

N̂ â + |n⊗ = (â + â)â + |n⊗ = â + (â â + )|n⊗ = â + (â + â + 1)|n⊗ = â + ( N̂ + 1)|n⊗

= (n + 1)â + |n⊗.

Problem 2.3
Taking into account the results of the two previous problems, show that the spectrum
of the number operator N̂ = â + â, where â and â + satisfy the commutation rule
â â + − â + â = 1, is the set of integer numbers.
Solution
We have seen that N̂ has a non negative spectrum. This means that N̂ possesses a
lowest eigenvalue n 0 with |n 0 ⊗ its eigenstate. This eigenstate |n 0 ⊗ is such that

â|n 0 ⊗ = 0.

In fact, on the basis of the results of the previous problem, â|n 0 ⊗ should be eigenstate
of N̂ with eivenvalue n 0 −1 but this is not possible because n 0 is the lowest eigenvalue
of N̂ . Consequently the state â|n 0 ⊗ is not a good Fock state and we set it equal to 0.
In addition, due to the fact that

N̂ |n 0 ⊗ = n 0 |n 0 ⊗


= â + â|n 0 ⊗ = â + â|n 0 ⊗ = â + (0) = 0

we find that n 0 = 0. Thus, the state |0⊗, called vacuum state, is the eigenstate of N̂
with eigenvalue 0, i.e.
N̂ |0⊗ = 0|0⊗ = 0,
42 2 Second Quantization of Light

but also
â|0⊗ = 0.

Due to this equation, it follows that the eigenstates of N̂ are only those generated by
applying m times the operator â + on the vacuum state |0⊗, namely

1
|m⊗ = ∇ (â + )m |0⊗,
m!

where the factorial is due to the normalization. Finally, we notice that it has been
shown in the previous problem that the state |m⊗ has integer eigenvalue m.
Problem 2.4
Consider the following quantum Hamiltonian of the one-dimensional harmonic oscil-
lator
p̂ 2 1
Ĥ = + mβ 2 x̂ 2 .
2m 2
By using the properties of the annihilation operator
  
mβ i
â = x̂ + p̂ ,
2 mβ

determine the eigenfunction of the ground state of the system.

Solution
Let us observe the following property of the annihilation operator
  
mβ  ν
↑x|â = x+ ↑x|.
2 mβ νx

In addition, from
â|0⊗ = 0,

we find   
mβ  ν
↑x|â|0⊗ = x+ ↑x|0⊗ = 0.
2 mβ νx

Introducing the characteristic harmonic length



lH = ,
mβ

the adimensional coordinate x

x̄ = ,
lH
2.5 Solved Problems 43

and the adimensional eigenfunction

τ̄n (x̄) = l H τn (x̄l H ),

we obtain  
ν
x̄ + τ̄0 (x̄) = 0.
ν x̄

The solution of this first order differential equation is found by separation of variables:

d τ̄0
x̄ d x̄ = − ,
τ̄0

from which  
1 −x̄ 2
τ̄0 (x̄) = exp ,
π 1/4 2

having imposed the normalization


d x̄ |τ̄0 (x̄)|2 = 1.

Problem 2.5
Consider the quantum Hamiltonian of the two-dimensional harmonic oscillator

p̂12 + p̂22 1
Ĥ = + mβ 2 (x̂12 + x̂22 ).
2m 2
By using the properties of the creation operators
  
mβ i
âk+ = x̂k − p̂k , k = 1, 2
2 mβ

determine the eigenfunctions of the quantum Hamiltonian with eigenvalue 6β.

Solution
The eigenvalues of the two-dimensional harmonic oscillator are given by

E n 1 n 2 = β(n 1 + n 2 + 1),

where n 1 , n 2 = 0, 1, 2, 3, . . . are the quantum numbers.

44 2 Second Quantization of Light

The eigenstates |n 1 n 2 ⊗ corresponding to the eigenvalue 6β are:

|50⊗, |41⊗, |32⊗, |23⊗, |14⊗, |05⊗.

Thus the eigenfunctions to be determined are

θ50 (x1 , x2 ), θ41 (x1 , x2 ), θ32 (x1 , x2 ), θ23 (x1 , x2 ), θ14 (x1 , x2 ), θ05 (x1 , x2 ).

In general, the eigenfunctions θn 1 n 2 (x1 , x2 ) = ↑x1 x2 |n 1 n 2 ⊗ can be factorized as

follows
θn 1 n 2 (x1 , x2 ) = τn 1 (x1 ) τn 2 (x2 )

where τn j (x j ) = ↑x j |n j ⊗, j = 1, 2. It is now sufficient to calculate the following

eigenfunctions of the one-dimensional harmonic oscillator:

τ0 (x), τ1 (x), τ2 (x), τ3 (x), τ4 (x), τ5 (x).

We observe that the creation operator of the one-dimensional harmonic oscillator

satisfies this property
  
+ mβ  ν
↑x|â = x− ↑x|,
2 mβ νx

and consequently
     
+ mβ  ν mβ  ν
↑x|â |n⊗ = x− ↑x|n⊗ = x− τn (x).
2 mβ νx 2 mβ νx

Reminding that ∇
â + |n⊗ = n + 1|n + 1⊗,

we get
1
↑x|n + 1⊗ = ∇ ↑x|a + |n⊗,
n+1

and the iterative formula

  
1 mβ  ν
τn+1 (x) = ∇ x− τn (x).
n+1 2 mβ νx

Now we introduce the characteristic harmonic length



lH = ,
mβ
2.5 Solved Problems 45

the adimensional coordinate x

x̄ = ,
lH

and the adimensional wavefunction

τ̄n (x̄) = l H τn (x̄l H ),

finding  n
ν 1
τ̄n (x̄) = ∇ x̄ − τ̄0 (x̄).
n
2 n! ν x̄

The function τ̄0 (x̄) of the ground state is given by (see Exercise 1.4)
 
1 −x̄ 2
τ̄0 (x̄) = exp .
π 1/4 2

Let us now calculate the effect of the operator

 n
ν
x̄ − .
ν x̄

For n = 1:      
ν −x̄ 2 −x̄ 2
x̄ − exp = 2 x̄ exp .
ν x̄ 2 2

For n = 2:  2    
ν −x̄ 2 −x̄ 2
x̄ − exp = (2 x̄ 2 − 1) exp .
ν x̄ 2 2

For n = 3:
 3    
ν −x̄ 2 −x̄ 2
x̄ − exp = (8x̄ − 12 x̄) exp
3
.
ν x̄ 2 2

For n = 4:
 4    
ν −x̄ 2 −x̄ 2
x̄ − exp = (16x̄ 4 − 48x̄ 2 + 12) exp .
ν x̄ 2 2

For n = 5:
 5    
ν −x̄ 2 −x̄ 2
x̄ − exp = (32 x̄ 5 − 160 x̄ 3 + 120 x̄) exp .
ν x̄ 2 2
46 2 Second Quantization of Light

Finally, the eigenfunctions of H with eigenvalue 6β are linear combinations of

functions θn 1 n 2 (x1 , x2 ), namely

(x1 , x2 ) = cn 1 n 2 θn 1 n 2 (x1 , x2 ) ∂n 1 +n 2 ,5 ,
n1 n2

where the coefficients cn 1 n 2 are such that


|cn 1 n 2 |2 ∂n 1 +n 2 ,5 = 1.
n1 n2

Problem 2.6
Show that the coherent state |φ⊗, defined by the equation

â|φ⊗ = φ|φ⊗,

can be written as
∈
 φn
e−|φ|
2 /2
|φ⊗ = ∇ |n⊗.
n=0
n!

Solution
The coherent state |φ⊗ can be expanded as
∈

|φ⊗ = cn |n⊗.
n=0

Then we have
∈
 ∈
 ∈
 ∈

∇ ∇
â|φ⊗ = â cn |n⊗ = cn â|n⊗ = cn n|n − 1⊗ = cn+1 n + 1|n⊗
n=0 n=0 n=1 n=0
∈ ∈
φ|φ⊗ = φ cn |n⊗ = φ cn |n⊗.
n=0 n=0

Since, by definition
â|φ⊗ = φ|φ⊗,

it follows that ∇
cn+1 n + 1 = φ cn

from which φ
cn+1 = ∇ cn ,
n+1
2.5 Solved Problems 47

namely
φ φ2 φ3
c1 = ∇ c0 , c2 = ∇ c0 , c3 = ∇ c0 , · · ·
1 2! 3!

and in general
φn
cn = ∇ c0 .
n!

Summarizing
∈
 φn
|φ⊗ = c0 ∇ |n⊗,
n=0
n!

where the parameter c0 is fixed by the normalization

∈
 φ2n
= |c0 |2 e|φ| ,
2
1 = ↑φ|φ⊗ = |c0 |2
n!
n=0

and consequently
c0 = e−|φ|
2 /2
,

up to a constant phase factor.

Problem 2.7
Calculate the probability of finding the Fock state |n⊗ in the vacuum state |0⊗.
Solution
The probability p is given by

1 if n = 0
p = |↑0|n⊗|2 = ∂0,n = .
0 if n ⊥= 0

Problem 2.8
Calculate the probability of finding the coherent state |φ⊗ in the vacuum state |0⊗.
Solution
The probability is given by
p = |↑0|φ⊗|2 .

Because
∈
 ∈
φn φn 1
e−|φ|
2 /2
e−|φ| /2 ∇ ↑0|n⊗ = e−|φ| /2 ∇ = e−|φ| /2 ,
2 2 2
↑0|φ⊗ = ↑0| ∇ |n⊗ =
n=0
n! n=0
n! 0!

we get
p = e−|φ| .
2
48 2 Second Quantization of Light

Problem 2.9
Calculate the probability of finding the coherent state |φ⊗ in the Fock state |n⊗.
Solution
The probability is given by
p = |↑n|φ⊗|2 .

Because
∈
 ∈
φm φm φn
e−|φ|
2 /2
e−|φ| /2 ∇ ↑n|m⊗ = e−|φ| /2 ∇ ,
2 2
↑n|φ⊗ = ↑0| ∇ |m⊗ =
m=0
m! m=0
m! n!

we get
φ2n
p = e−|φ|
2
,
n!
which is the familiar Poisson distribution.
Problem 2.10
Calculate the probability of finding the coherent state |φ⊗ in the coherent state |δ⊗.
Solution
The probability is given by
p = |↑δ|φ⊗|2 .

Because
∈
 ∈  ∈
φm (δ √ )n φm
e−|φ|
2 /2
e−|δ| /2 e−|φ| /2 ∇ ∇ ↑n|m⊗
2 2
↑δ|φ⊗ = ↑δ| ∇ |m⊗ =
m=0
m! n=0 m=0
n! m!
∈
 (δ √ φ)n √φ 2 +|δ|2 −2φδ √ )/2
= e−(|φ|
2 +|δ|2 )/2
= e−(|φ|
2 +|δ|2 )/2
eδ = e−(|φ| ,
n!
n=0

we get
p = e−|φ−δ| .
2

This means that two generic coherent states |φ⊗ and |δ⊗ are never orthogonal to each
other.

Further Reading

For the second quantization of the electromagnetic field:

F. Mandl, G. Shaw, Quantum Field Theory, Chap.1, Sects. 1.1 and 1.2 (Wiley, New
York, 1984)
2.5 Solved Problems 49

M.O. Scully, M.S. Zubairy, Quantum Optics, Chap. 1, Sects. 1.1 and 1.2 (Cambridge
University Press, Cambridge, 1997)
U. Leonhardt, Measuring the Quantum State of Light, Chap. 2, Sects. 2.1, 2.2, and
2.3 (Cambridge University Press, Cambridge, 1997)
F.T. Arecchi, Measurement of the statistical distribution of Gaussian and laser
sources. Phys. Rev. Lett. 15, 912 (1965)
For the Casimir effect:
R.W. Robinett, Quantum Mechanics: Classical Results, Modern Systems, and Visu-
alized Examples, Chap. 19, Sect. 19.9 (Oxford University Press, Oxford, 2006)
S.K. Lamoreaux, Demonstration of the Casimir force in the 0.6 to 6 m range. Phys.
Rev. Lett. 78, 5 (1997)
G. Bressi, G. Carugno, R. Onofrio, G. Ruoso, Measurement of the Casimir force
between parallel metallic surfaces. Phys. Rev. Lett. 88, 041804 (2002)
For the quantum radiation field at finite temperature:
K. Huang, Statistical Mechanics, Chap. 12, Sect. 12.1 (Wiley, New York, 1987)
Chapter 3
Electromagnetic Transitions

In this chapter we investigate the crucial role of the quantum electromagnetic field
on the spontaneous and stimulated transitions between two atomic quantum states.
After reviewing some properties of classical electrodynamics, we analyze the quan-
tum electrodynamics within the dipole approximation. We calculate the rate of spon-
taneous emission, absorption, and stimulated emission and connect them with the
transition coefficients introduced by Einstein. Finally, we discuss the life time of an
atomic state and the line width of an electromagnetic transition.

3.1 Classical Electrodynamics

Let us consider a classical system composed of N particles with masses mi , electric

charges qi , positions ri and linear momenta pi , under the presence of an electromag-
netic field. The Hamiltonian of free matter is given by


N
p2i
Hfree = , (3.1)
2mi
i=1

where p2i /(2mi ) = p2i /(2mi ) is the kinetic energy of ith particle. The presence of the
electromagnetic field is modelled by the Hamiltonian

H = Hshift + Hrad , (3.2)

where

N
(pi − qi Ai )2
Hshift = + qi νi , (3.3)
2mi
i=1

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 51

DOI: 10.1007/978-3-319-05179-6_3, © Springer International Publishing Switzerland 2014
52 3 Electromagnetic Transitions

with Ai = A(ri , t), νi = ν(ri , t), where ν(r, t) and A(r, t) are the electromagnetic
potentials and   
ρ0 2 1 2
Hrad = d r 3
E (r, t) + B (r, t) , (3.4)
2 T 2μ0

is the radiation Hamiltonian. We observe that by using the Hamilton equations

πHshift
ṙi = (3.5)
πpi
πHshift
ṗi = − (3.6)
πri

on the shift Hamiltonian (3.3) it is straightforward to derive Newton equations with

the Lorentz force acting on the ith particle
 
πAi
mi r̈i = qi −∧i νi − + vi ∧ (∇ i ∧ Ai ) = qi (Ei + vi ∧ Bi ) (3.7)
πt

where vi = ṙi and, as always, the electric field E is obtained as

πA
E = −∇ν − , (3.8)
πt
while the magnetic field B can be written as

B = ∇ ∧ A. (3.9)

In Eq. (3.4) it appears the transverse electric field ET which, in the Coulomb gauge
∇ · A = 0, is related to the total electric field E by the following decomposition into
longitudinal EL and transverse ET fields

E = EL + ET , (3.10)

such that
πA
EL = −∇ν, ET = − . (3.11)
πt
Remarkably, the longitudinal electric field EL gives rise to the instantaneous
electrostatic interaction between the charges. Indeed, Eq. (3.3) can be rewritten as


N
(pi − qi Ai )2
Hshift = + HC , (3.12)
2mi
i=1
3.1 Classical Electrodynamics 53

where

N
1 
N
qi qj
HC = qi νi = , (3.13)
2 4β0 |ri − rj |
i=1 i,j=1

is the Hamiltonian of the Coulomb interaction. The Hamiltonian Hshift can be further
decomposed as follows
Hshift = Hmatt + HI , (3.14)

where
Hmatt = Hfree + HC (3.15)

is the Hamiltonian of the self-interacting matter and

 

N
qi q2
HI = − Ai · pi + i A2i (3.16)
m 2mi
i=1

is the interaction Hamiltonian between matter and radiation. Combining the above
results, we obtain the complete Hamiltonian

H = Hshift + Hrad = Hmatt + Hrad + HI . (3.17)

Notice, however, that this Hamiltonian does not take into account the possible spin
of particles.

3.2 Quantum Electrodynamics in the Dipole Approximation

The quantization of electrodynamics is obtained promoting the classical Hamiltonian

of the system to a quantum Hamiltonian. For simplicity we consider the hydrogen
atom with Hamiltonian
p̂2 e2
Ĥmatt = − , (3.18)
2m 4β0 |r|

where p̂ = −i∇ is the linear momentum operator of the electron in the state |p∇, in
the presence of the radiation field with Hamiltonian

+
Ĥrad = εk âks âks , (3.19)
k s

where âks and âks are the annihilation and creation operators of the photon in the
state |ks∇. Here −e is the electric charge of the electron, e = 1.60 × 10−19 C is the
electric charge of the proton, and m = me mp /(me + mp ) → me is the reduced mass
54 3 Electromagnetic Transitions

of the electron-proton system, with me = 9.11 × 10−31 kg the electron mass and
mp = 1.67 × 10−27 kg the proton mass.
We take into account the matter-radiation interaction by using the so-called dipolar
Hamiltonian e
ĤD = Â(0, t) · p̂. (3.20)
m

With respect to the complete interaction Hamiltonian ĤI , given in our case by
Eq. (3.16) with N = 1, we neglect the term e2 Â2 /(2m) of Eq. (3.16) which rep-
resents only a tiny perturbation in the atomic system. In addition, we also neglect the
spatial variations in the vector potential operator

 

+ −i(k·r−εk t)
Â(r, t) = âks ei(k·r−εk t) + âks e εks . (3.21)
s
20 εk V
k

This dipolar approximation, which corresponds to

1
eik·r = 1 + ik · r + (ik · r)2 + · · · → 1, (3.22)
2
1
e−ik·r = 1 − ik · r + (ik · r)2 + · · · → 1, (3.23)
2
is reliable if k · r ∀ 1, namely if the electromagnetic radiation has a wavelength
λ = 2β/|k| very large compared to the linear dimension R of the atom. Indeed,
the approximation is fully justified in atomic physics where λ → 10−7 m and R →
10−10 m. Notice, however, that for the θ electromagnetic transitions of atomic nuclei
the dipolar approximation is not good and it is more convenient to expand the vector
potential operator Â(r, t) into vector spherical harmonics, i.e. photons of definite
angular momentum.
The total Hamiltonian of our system is then given by

Ĥ = Ĥ0 + ĤD , (3.24)

where
Ĥ0 = Ĥmatt + Ĥrad (3.25)

is the unperturbed Hamiltonian, whose eigenstates are of the form

|a∇| . . . nks . . . ∇ = |a∇ √ | . . . nks . . . ∇ (3.26)

 with eigenvalue Ea and | . . . nks . . . ∇ it the

where |a∇ is the eigenstate of Ĥmatt
eigenstate of Ĥrad with eigenvalue ks εk nks , i.e.
3.2 Quantum Electrodynamics in the Dipole Approximation 55

Ĥ0 |a∇| . . . nkr . . . ∇ = Ĥmatt + Ĥrad |a∇| . . . nks . . . ∇
 

= Ea + εk nks |a∇| . . . nks . . . ∇. (3.27)
ks

The time-dependent perturbing Hamiltonian is instead given by


e  

+ iεk t
ĤD (t) = âks e−iεk t + âks e εks · p̂. (3.28)
m s
20 εk V
k

Fermi golden rule: Given the initial |I∇ and final |F∇ eigenstates of the unperturbed
Hamiltonian Ĥ0 under the presence to the time-dependent perturbing Hamiltonian
ĤD , the probability per unit time of the transition from |I∇ to |F∇ is given by

2β
WIF = |⊗F|ĤD (0)|I∇|2 ∂(EI − EF ), (3.29)

with the constraint of energy conservation.
This is the so-called Fermi golden rule, derived in 1926 by Paul Dirac on the basis
of the first order time-dependent perturbation theory, and named “golden rule” few
years later by Enrico Fermi.

3.2.1 Spontaneous Emission

Let us now apply the Fermi golden rule to the very interesting case of the hydrogen
atom in the state |b∇ and the radiation field in the vacuum state |0∇. We are thus
supposing that the initial state is

|I∇ = |b∇|0∇. (3.30)

Notice that, because we are considering the hydrogen atom, one has

Ĥmatt |b∇ = Eb |b∇, (3.31)

where
mc2 ψ2 13.6 eV
Eb = − →− (3.32)
2nb2 nb2

is the well-known quantization formula of the nonrelativistic hydrogen atom with

quantum number nb = 1, 2, 3, . . . and
56 3 Electromagnetic Transitions

e2 1
ψ= → (3.33)
(4β0 )c 137

the fine-structure constant. In addition we suppose that the final state is

|F∇ = |a∇|ks∇, (3.34)

+
i.e. the final atomic state is |a∇ and the final photon state is |ks∇ = |1ks ∇ = âks |0∇.
Obviously, we have
Ĥmatt |a∇ = Ea |a∇, (3.35)

where
mc2 ψ2 13.6 eV
Ea = − 2
→− (3.36)
2na na2

with na = 1, 2, 3, . . .. In this process, where Eb > Ea , there is the spontaneous

production of a photon from the vacuum: a phenomenon strictly related to the quan-
tization of the electromagnetic field.
From Eqs. (3.28) and (3.29) one finds
 
spont 2β  e 2 
Wba,ks = |εks · ⊗a|p̂|b∇|2 ∂(Eb − Ea − εk ), (3.37)
 m 20 εk V

because
âk∞ s∞ |I∇ = âk∞ s∞ |b∇|0∇ = |b∇âk∞ s∞ |0∇ = 0, (3.38)

while
âk+∞ s∞ |I∇ = âk+∞ s∞ |b∇|0∇ = |b∇âk+∞ s∞ |0∇ = |b∇|k∞ s∞ ∇, (3.39)

and consequently

⊗F|p̂ âk∞ s∞ |I∇ = 0, ⊗F|p̂ âk+∞ s∞ |I∇ = ⊗a|p̂|b∇ ∂k∞ ,k ∂s∞ ,s . (3.40)

+
Remember that p̂ acts on a different Hilbert space with respect to âks and âks . By
using the equation of the linear momentum operator p̂ of the electron

p̂ 1
= [r, Ĥmatt ], (3.41)
m i
we get

1 m
⊗a|p̂|b∇ = ⊗a|m [r, Ĥmatt ]|b∇ = ⊗a|rĤmatt − Ĥmatt r|b∇
i i
m
= (Eb − Ea )⊗b|r|a∇ = −imεba ⊗a|r|b∇, (3.42)
i
3.2 Quantum Electrodynamics in the Dipole Approximation 57

where εba = (Eb − Ea )/, and consequently

spont βεba 2
Wba,ks = |εks · ⊗a|e r|b∇|2 ∂(εba − εk ). (3.43)
V 0 εk

The delta function is eliminated by integrating over the final photon states

 
spont spont d 3 k  spont
Wba = Wba,ks =V Wba,ks
s
(2β)3
k s=1,2
  
V spont
= dkk 2 d Wba,ks , (3.44)
8β 3
s=1,2

where d is the differential solid angle. Because εk1 , εk2 and n = k/k form a
orthonormal system of vectors, setting rab = ⊗a|r|b∇ one finds

|rab |2 = |εk1 · rab |2 + |εk2 · rab |2 + |n · rab |2 = |εks · rab |2 + |rab |2 cos2 (φ),
s=1,2
(3.45)
where φ is the angle between rba and n. It follows immediately

|εks · rab |2 = |rab |2 (1 − cos2 (φ)) = |rab |2 sin2 (φ), (3.46)
s=1,2

namely 
|εks · ⊗a|r|b∇|2 = |⊗a|r|b∇|2 sin2 (φ). (3.47)
s=1,2

In addition, in spherical coordinates one can choose d = sin (φ)dφdν, with φ ∈

[0, β] the zenith angle of colatitude and ν ∈ [0, 2β] the azimuth angle of longitude,
and then   2β  β
8β
d sin2 (φ) = dν dφ sin3 (φ) = . (3.48)
0 0 3

In this way from Eq. (3.44) we finally obtain

spont εba
3
Wba = |⊗a|d|b∇|2 , (3.49)
3β0 c3

where the d = −e r is the classical electric dipole momentum of the hydrogen atom,
i.e. the dipole of the electron-proton system where r is the position of the electron of
charge −e < 0 with respect to the proton of charge e > 0, and ⊗a|d|b∇ = −⊗a|e r|b∇
is the so-called dipole transition element.
58 3 Electromagnetic Transitions

3.2.2 Absorption

We now consider the excitation from the atomic state |a∇ to the atomic state |b∇ due
to the absorption of one photon. Thus we suppose that the initial state is

|I∇ = |a∇|nks ∇, (3.50)

while the final state is

|F∇ = |b∇|nks − 1∇, (3.51)

where Ea < Eb . From Eqs. (3.28) and (3.29) one finds

 
absorp 2β  e 2 
Wab,ks = nks |εks · ⊗b|p̂|a∇|2 ∂(Ea + εk − Eb ), (3.52)
 m 20 εk V

because
∓
⊗F|p̂ âk∞ s∞ |I∇ = nks ⊗b|p̂|a∇ ∂k∞ ,k ∂s∞ ,s , ⊗F|p̂ âk+∞ s∞ |I∇ = 0. (3.53)

Note that with respect to Eq. (3.37) in Eq. (3.52) there is also the multiplicative term
nks . We can follow the procedure of the previous section to get

absorp βεba 2
Wab,ks = nks |εks · ⊗b|e r|a∇|2 ∂(εba − εk ). (3.54)
V 0 εk

Again the delta function can be eliminated by integrating over the final photon states
but here one must choose the functional dependence of nks . We simply set

nks = n(εk ), (3.55)

and after integration over k and s, from Eq. (3.54) we get

absorp εba
3
spont
Wab = |⊗b|d|a∇|2 n(εba ) = Wba n(εba ). (3.56)
3β0 c3

Note that without the integration over the solid angle, which gives a factor 1/3,
one obtains the absorption probability of one photon with a specific direction. For
a thermal distribution of photons, with α(ε) the energy density per unit of angular
frequency specified by the thermal-equilibrium Planck formula

ε 3 1
α(ε) = n(ε), n(ε) = ε/(k T ) , (3.57)
β c
2 3 e B −1

where kB is the Boltzmann constant and T the absolute temperature, we can

also write
3.2 Quantum Electrodynamics in the Dipole Approximation 59

spont β c
1 2 3
absorp
Wab = |⊗b|d|a∇|2 α(ε ba ) = W α(εba ). (3.58)
30 2 ba
εba
3

3.2.3 Stimulated Emission

In conclusion we consider the stimulated emission of a photon from the atomic state
|b∇ to the atomic state |a∇. Thus we suppose that the initial state is

|I∇ = |b∇|nks ∇, (3.59)

while the final state is

|F∇ = |a∇|nks + 1∇, (3.60)

where Eb > Ea . From Eqs. (3.28) and (3.29) one finds

 
2β  e 2 
stimul
Wba,ks = (nks + 1) |εks · ⊗a|p̂|b∇|2 ∂(Eb − Ea − εk ),
 m 20 εk V
(3.61)
because
∓
âks |I∇ = âks |b∇|nks ∇ = |b∇âks |nks ∇ = nks |b∇|nks − 1∇, (3.62)

while
+ + +

âks |I∇ = âks |b∇|nks ∇ = |b∇âks |nks ∇ = nks + 1 |b∇|nks + 1∇, (3.63)

and consequently

+

⊗F|p̂ âks |I∇ = 0, ⊗F|p̂ âks |I∇ = nks + 1 ⊗a|p̂|b∇. (3.64)

Note that with respect to Eq. (3.52) in Eq. (3.61) there is the factor nks + 1 instead
of nks . It is straightforward to follow the procedure of the previous section to obtain

εba
3
absorp spont
stimul
Wba = |⊗a|d|b∇|2 (n(εba ) + 1) = Wab + Wba , (3.65)
3β0 c3
stimul of stimulated emission reduces to the spon-
which shows that the probability Wba
spont
taneous one Wba when n(εba ) = 0. It is then useful to introduce

spont
stimul
W̃ba = Wba
stimul
− Wba (3.66)

which is the effective stimulated emission, i.e. the stimulated emission without the
contribution due to the spontaneous emission. Clearly, for a very large number of
60 3 Electromagnetic Transitions

Fig. 3.1 Scheme of the 3 mechanisms of radiative transition

photons (n(εba ) ↑ 1) one gets W̃ba

stimul → W stimul . Moreover, for a thermal distribu-
ba
tion of photons, with the energy density per unit of angular frequency α(ε), we can
also write
spont β c
2 3
absorp
stimul
W̃ba = Wab = Wba α(εba ). (3.67)
εba
3

Remarkably, the probability of stimulated emission is different from zero only if the
emitted photon is in the same single-particle state |ks∇ of the stimulating ones (apart
when the stimulating light is the vacuum |0∇). In the stimulated emission the emitted
photon is said to be “coherent” with the stimulating ones, having the same frequency
and the same direction (Fig. 3.1).

3.3 Selection Rules

It is clear that, within the dipolar approximation, no electromagnetic transition, either

spontaneous or stimulated, will occur between the atomic states |a∇ and |b∇ unless
at least one component of the dipole transition matrix element ⊗b|d|a∇ is nonzero. It
is possible to show that the matrix elements are zero for certain pairs of states. If so,
the transition is not allowed (at least in the electric dipole approximation), and the
results can be summarized in terms of simple selection rules governing the allowed
changes in quantum numbers in transitions.
Since the electric dipole d = −er changes sign under parity (r ↓ −r), the matrix
element ⊗b|d|a∇ is zero if the states |a∇ and |b∇ have the same parity. Therefore
3.3 Selection Rules 61

the parity of the wavefunction must change in an electric dipole transition.

This means that if σa (r) and σb (r) are the eigenfunctions of the states |a∇ and |b∇
and, for instance, both eigenfunctions are even, i.e.

σa (−r) = σa (r) and σb (−r) = σb (r), (3.68)

then the dipole matrix element is such that


⊗b|d|a∇ = −e d 3 r σb∗ (r) r σa (r) = 0. (3.69)

 
In fact, under the trasformation r ↓ −r one finds d 3 (−r) = d 3 r while ⊗b|d|a∇ =
−⊗b|d|a∇, and then it follows ⊗b|d|a∇ = 0.
Let us recall that in our calculations a generic eigenstate |a∇ = |nlm∇ of the matter
Hamiltonian Ĥmatt is such that

σa (r) = σnlm (r) = ⊗r|a∇ = ⊗r|nlm∇ = Rnl (r)Ylm (φ, ν), (3.70)

with Ynm (φ, ν) the spherical harmonic function. The spherical harmonics satisfy the
orthonormalization condition
 β  2β
dφ sin (φ) dν Yl∗∞ m∞ (φ, ν) Ylm (φ, ν) = ∂l,l∞ ∂m,m∞ , (3.71)
0 0

∗ (φ, ν) = (−1)m Y
where Yl,m l,−m (φ, ν). Moreover under parity transformation
one finds

σnlm (−r) = Rnl (r)Ylm (β − φ, ν + β) = (−1)l Rnl (r)Ylm (φ, ν) = (−1)l σnlm (r),
(3.72)
thus the radial part Rnl (r) of the wavefunction is unchanged and the parity of the
state is fully determined from the angular part.
The generic dipole matrix element is given by
 ∞  β  2β
∞ ∞ ∞
⊗n l m |d|nlm∇ = −e dr r Rn∞ l∞ (r)Rnl (r)
3
dφ sin (φ) dν Yl∗∞ m∞ (φ, ν)
0 0 0
× (cos (ν) sin (φ), sin (ν) sin (φ), cos (φ)) Ylm (φ, ν), (3.73)

with r = r (cos (ν) sin (φ), sin (ν) sin (φ), cos (φ)). Setting l = l − l∞ and m =
m − m∞ , by using the properties of the spherical harmonics it is possible to prove that
the dipole matrix element ⊗n∞ l ∞ m∞ |d|nlm∇ is different from zero if only if l = ±1 and
m = 0, ±1 (see, for instance, Problem 3.2). Thus, in the dipole approximation,
the orbital angular momentum l of the atomic state and its third component m =
−l, −l + 1, ..., l − 1, l must satisfy the selection rules
62 3 Electromagnetic Transitions

l = ±1 m = 0, ±1. (3.74)

This means that in the electric dipole transitions the photon carries off (or brings in)
one unit of angular momentum. It is important to stress that the above selection rules
are obtained within the dipole approximation. This means that they can be violated
by rare electromagnetic transitions involving higher multipolarities.
Concluding this section we observe that in the general case of N particles with
positions ri and electric charges qi , i = 1, 2, ..., N, the electric dipole momentum is
defined
N
d= qi ri . (3.75)
i=1

Moreover, due to Eq. (3.42), the dipolar interaction Hamiltonian (3.20) can be effec-
tively written as
ĤD = −d · Ê, (3.76)

which is the familiar expression of the coupling between the electric dipole d and
the electric field Ê.

3.4 Einstein Coefficients

In 1919 Albert Einstein observed that, given an ensemble of N atoms in two possible
atomic states |a∇ and |b∇, with Na (t) the number of atoms in the state |a∇ at time t
and Nb (t) the number of atoms in the state |b∇ at time t, it must be

N = Na (t) + Nb (t) (3.77)

and consequently
dNa dNb
=− . (3.78)
dt dt
Einstein suggested that, if the atoms are exposed to an electromagnetic radiation of
energy density per unit of frequency α(ε), the rate of change of atoms in the state
|a∇ must be
dNa
= Aba Nb + Bba α(εba ) Nb − Bab α(εba ) Na . (3.79)
dt
where the parameters Aba , Bba , and Bab are known as Einstein coefficients. The first
two terms in this equation describe the increase of the number of atoms in |a∇ due
to spontaneous and stimulated transitions from |b∇, while the third term takes into
account the decrease of the number of atoms in |a∇ due to absorption with consequent
transition to |b∇.
3.4 Einstein Coefficients 63

Einstein was able to determine the coefficients Aba , Bba and Bab by supposing that
the two rates in Eqs. (3.78) and (3.79) must be equal to zero at thermal equilibrium, i.e.

dNa dNb
=− = 0, (3.80)
dt dt
In this way Einstein found
 
Nb Nb
Aba = α(εba ) Bab − Bba . (3.81)
Na Na

Because the relative population of the atomic states |a∇ and |b∇ is given by a
Boltzmann factor
Nb e−δEb
= −δE = e−δ(Eb −Ea ) = e−δ εba , (3.82)
Na e a

Einstein got
Aba
α(εba ) = . (3.83)
Bab eδ εba − Bba

At thermal equilibrium we know that

εba
3
1
α(εba ) = . (3.84)
β c3 eδ εba − 1
2

It follows that
εba3
Aba = Bba , Bab = Bba . (3.85)
β 2 c3
Notice that in this way Einstein obtained the coefficient Aba of spontaneous decay
by simply calculating the coefficient of stimulated decay Bba .
Actually, by using the results we have obtained in the previous section, it is
clear that

spont εba
3
Aba = Wba = |⊗a|d|b∇|2 , (3.86)
3β0 c3
stimul
W̃ba β 2 c3
Bba = = A ,
3 ba
(3.87)
α(εba ) εba
absorp
Wab β 2 c3
Bab = = A .
3 ba
(3.88)
α(εba ) εba

It is important to stress that the laser device, invented in 1957 by Charles Townes
and Arthur Schawlow at Bell Labs, is based on a generalization of Eqs. (3.78) and
(3.79). In the laser device one has population inversion, which is achieved with an
64 3 Electromagnetic Transitions

out-of-equilibrium pumping mechanism strongly dependent on the specific charac-

teristics of the laser device.

3.4.1 Rate Equations for Two-Level and Three-Level Systems

Let us analyze under which conditions one can get population inversion. We have
seen that for a two-level system under the action of an electromagnetic pump with
energy density α(εba ) the rate equations of Einstein are given by

dNa
= Aba Nb + R (Nb − Na ), (3.89)
dt
dNb
= −Aba Nb − R (Nb − Na ), (3.90)
dt

where R = Bba α(εba ) is the pumping rate. We now suppose that the system is in
a steady state, but not at thermal equilibrium, i.e. the energy density α(εba ) of the
electromagnetic pump is not given by the Planck distribution (3.84). At the steady
state we have
dNa dNb
= = 0, (3.91)
dt dt
from which, taking into account that N = Na + Nb , we obtain

Aba + R
Na = N , (3.92)
Aba + 2R
R
Nb = N , (3.93)
Aba + 2R

and the population imbalance ζ = (Na − Nb )/N reads

Aba
ζ= . (3.94)
Aba + 2R

Eq. (3.94) shows that in the absence of electromagnetic pump (R = 0) one has
ζ = 1, which simply means that all two-level atoms are in the lower state |a∇.
Instead, if the electromagnetic pump is active (R ≥= 0) the population imbalance
ζ decreases by increasing the pumping rate R, but ζ cannot change sign because
ζ ↓ 0 for R ↓ +∞. Thus, we conclude that the population inversion (ζ < 0), and
consequently laser light, is impossible in a strictly two-level system.
Let us now consider an ensemble of N atoms in three possible atomic states |a∇,
|b∇ and |c∇, with Na (t) the number of atoms in the state |a∇ at time t, Nb (t) the number
of atoms in the state |b∇ at time t, and Nc (t) the number of atoms in the state |b∇ at
time t. In this ensemble of three-level atoms it must be
3.4 Einstein Coefficients 65

N = Na (t) + Nb (t) + Nc (t). (3.95)

Moreover, we suppose that Ea < Eb < Ec , with Ea the energy of the state |a∇, Eb
the energy of the state |b∇, and Ec the energy of the state |c∇. In the presence of an
electromagnetic pump of energy density α(εca ), with εca = (Ec − Ea )/, between
the state |a∇ and |c∇ the rate equations of the three-level system are given by

dNa
= Aba Nb + Aca Nc + R (Nc − Na ) + Bba αba (Nb − Na ), (3.96)
dt
dNb
= −Aba Nb + Acb Nc − Bba αba (Nb − Na ) + Bcb αcb (Nc − Nb ), (3.97)
dt
dNc
= −Acb Nc − Aca Nc − R (Nc − Na ) − Bcb αcb (Nc − Nb ), (3.98)
dt

where R = Bca α(εca ) is the pumping rate and the energy densities αba and αcb
are induced by the electromagnetic transitions. These equations are simplified under
the following physical assumptions (Ruby laser): (i) there is no electromagnetic
transition from |c∇ and |b∇, namely αcb = 0; (ii) the time decay of the spontaneus
transition from |c∇ to |a∇ is very long, namely Aca → 0; (iii) the time decay of the
spontaneus transition from |c∇ to |b∇ is very short and consequently Nc ∀ Na . In this
way the rate equations become

dNa
= Aba Nb − R Na + S (Nb − Na ), (3.99)
dt
dNb
= −Aba Nb + Acb Nc − S (Nb − Na ), (3.100)
dt
dNc
= −Acb Nc + R Na , (3.101)
dt
where S = Bba αba is the stimulated rate. At stationarity one has

dNa dNb dNc

= = = 0, (3.102)
dt dt dt
from which we get

Aba + S
Na = N , (3.103)
Aba + 2S + R
R+S
Nb = N , (3.104)
Aba + 2S + R
R
Nc = Na , (3.105)
Acb

and the population imbalance ζ = (Na − Nb )/N reads

66 3 Electromagnetic Transitions

Aba − R
ζ= . (3.106)
Aba + 2S + R

This equation clearly shows that for R > Aba one gets ζ < 0, thus one obtains the
desired population imbalance.

3.5 Life-Time and Natural Line-Width

spont
We have seen that the Einstein coefficient Aba = Wba gives the transition proba-
bility per unit of time from the atomic state |b∇ to the atomic state |a∇. This means
that, according to Einstein, in the absence of an external electromagnetic radiation
one has
dNb
= −Aba Nb (3.107)
dt
with the unique solution
Nb (t) = Nb (0) e−Aba t . (3.108)

It is then quite natural to consider 1/Aba as the characteristic time of this spontaneous
transition.
More generally, the life-time τb of an atomic state |b∇ is defined as the reciprocal
of the total spontaneous transition probability per unit time to all possible final atomic
states |a∇, namely
1
τb =  . (3.109)
a Aba

Clearly, if |b∇ is the ground-state then Aba = 0 and τb = ∞.

On the basis of the time-energy indetermination principle of Werner Heisenberg,
in the radiation energy spectrum the natural line-width N due to the transition from
the state |b∇ to the state |a∇ can be defined as
 
1 1
N =  + . (3.110)
τb τa

Indeed it is possible to prove that in this transition the intensity of the emitted elec-
tromagnetic radiation follows the Lorentzian peak

I0 N2 /4
I() = , (3.111)
( − Eba )2 + N2 /4

where  = ε is the energy of the emitted photon and Eba = Eb − Ea is the energy
difference of the two atomic states. The Lorentzian peak is centered on  = Eba and
N is its full width at half-maximum.
3.5 Life-Time and Natural Line-Width 67

It is important to observe that the effective line-width  measured in the exper-

iments is usually larger than N because the radiating atoms move and collide. In
fact, one can write
 = N + C + D , (3.112)

where in addition to the natural width N there are the so-called collisional broad-
ening width C and the Doppler broadening width D .

3.5.1 Collisional Broadening

The collisional broadening is due to the collision among the radiating atoms of a gas.
The collision reduces the effective life-time of an atomic state and the collisional
width can be then written as

C = , (3.113)
τcol

where τcol is the collisional time, i.e. the average time between two collision of atoms
in the gas. According to the results of statistical mechanics, τcol is given by

1
τcol = , (3.114)
nσvmp

where n is the number density of atoms, σ is the cross-section, and vmp is the most
probable speed of the particles in the gas.
In the quantum theory of scattering one usually assumes that an incoming particle,
described by a plane wave along the z-axis, scatters with a target, modelled by an
external potential V (r), such that the outcoming asymptotic wavefunction of the
particle reads
eikr
σ(r) = eik·r + fk (φ) , (3.115)
r
where fk (φ) is the scattering amplitude. The differential cross section is given by
dσ/d = |fk (φ)|2 , which is the scattering probability per unit time of the particle
with mass m, wavevector k and energy 2 k 2 /(2m). If one considers only s-wave
scattering the differential cross section does not depend on the angle φ, and the total
scattering cross section is just

4β
σ(k) = 4β|fk |2 = sin2 (∂0 (k)), (3.116)
k2

where ∂0 (k) is the so-called s-wave phase shift of the cross-section. The s-wave
scattering length as , which characterizes the effective size of the target, is then defined
as the following low-energy limit
68 3 Electromagnetic Transitions

1
as = − lim tan (∂0 (k)). (3.117)
k↓0 k

In this way one clearly finds

lim σ(k) = 4βas2 . (3.118)
k↓0

Thus, for a dilute and cold gas of atoms the cross-section σ of low-energy atoms
can be indeed approximated as
σ = 4βas2 , (3.119)

where the value of as can be indeed determined by solving the reduced single-
particle Schrödinger equation with the actual inter-atomic potential V (r). Moreover,
by considering the Maxwell-Boltzmann distribution of speeds in an ideal gas, the
most probable speed vmp is given by

2kB T
vmp = , (3.120)
m

where T is the absolute temperature, kB is the Boltzmann constant and m is the mass
of each particle.

3.5.2 Doppler Broadening

The Doppler broadening is due to the Doppler effect caused by the distribution of
velocities of atoms. For non-relativistic velocities (vx ∀ c) the Doppler shift in
frequency is  vx
ε = ε0 1 − , (3.121)
c
where ε is the observed angular frequency, ε0 is the rest angular frequency, vx is
the component of the atom speed along the axis between the observer and the atom
and c is the speed of light. The Maxwell-Boltzmann distribution f (vx ) of speeds
vx = −c(ε − ε0 )/ε0 at temperature T , given by
 1/2
m
e−mvx /(2kB T ) dvx ,
2
f (vx ) dvx = (3.122)
2βkB T

becomes
 1/2
m c
e−mc
2 (ε−ε
0)
2 /(2ε 2 k T )
f (ε) dε = 0 B dε (3.123)
2βkB T ε0
3.5 Life-Time and Natural Line-Width 69

in terms of the angular frequency ε. This is the distribution of frequencies seen by

the observer, and the full width at half-maximum of the Gaussian is taken as Doppler
width, namely 
2 ln (2)kB T
D = ε0 . (3.124)
mc2

3.6 Minimal Coupling and Center of Mass

Up to now the interaction between atom and electromagnetic radiation has been
investigated by taking into account only the electronic degrees of freedom of the
atoms. Here we consider also the effect of the atomic nucleus and for simplicity we
study the hydrogen atom, whose classical complete Hamiltonian reads

p2p p2e e2
Hmatt = + − , (3.125)
2mp 2me 4βρ0 |rp − re |

where rp and re are the positions of proton and electron, and pp and pe are their
corresponding linear momenta, with mp the proton mass and me the electron mass.
By imposing the minimal coupling to the electromagnetic vector potential A(r),
in the Coulomb gauge we get

(pp − eA(rp ))2 (pe + eA(re ))2 e2

Hshift = + − . (3.126)
2mp 2me 4βρ0 |rp − re |

We now introduce center of mass position and relative position vectors:

mp rp + me re
rcm = , (3.127)
M
rrel = re − rp , (3.128)

where M = mp + me is the total mass and μ = mp me /M is the reduced mass, and

also the center of mass momentum and relative momentum vectors:

pcm = pp + pe , (3.129)
mp me
prel = pe − pp . (3.130)
M M
First of all we notice that

p2cm p2 e2
Hmatt = + rel − , (3.131)
2M 2μ 4βρ0 |rrel |
70 3 Electromagnetic Transitions

moreover
Hshift = Hmatt + HI , (3.132)

where
e me e mp
HI = − pcm · A(rcm − rrel ) + pcm · A(rcm + rrel )
M M M M
e me e mp
+ prel · A(rcm − rrel ) + prel · A(rcm + rrel )
mp M me M
e2 me e2 mp
+ A(rcm − rrel )2 + A(rcm + rrel )2 . (3.133)
2mp M 2me M

In this case, the dipole approximation means

me mp
A(rcm − rrel ) → A(rcm + rrel ) → A(rcm ) (3.134)
M M
and the interaction Hamiltonian becomes

e e2
HI = prel · A(rcm ) + A(rcm )2 . (3.135)
μ 2μ

We stress that, within the dipole approximation, in the Hamiltonian HI the depen-
dence on pcm disappears and that the only dependence on the center of mass is
included in A(rcm ). Indeed, this Hamiltonian is precisely the dipolar Hamiltonian of
Eq. (3.20) by using the center of mass as origin of the coordinate system.

3.7 Solved Problems

Problem 3.1
108 sodium atoms are excited in the first excited state of sodium by absorption of
light. Knowing that the excitation energy is 2.125 eV and the life time is 16 ns,
calculate the maximum of the emitting power.
Solution
The total absorbed energy is given by

E = N,

where N is the number of atoms (and also the number of absorbed photons) while 
is the transition energy

 = 2.125 eV = 2.125 × 1.6 × 10−19 J = 3.4 × 10−19 J.

3.7 Solved Problems 71

The absorbed energy is then

E = 108 × 3.4 × 10−19 J = 3.4 × 10−11 J.

The absorbed energy is equal to the energy emitted by spontaneous de-excitation.

The emitting power P(t) decays exponentially with time t as

P(t) = P0 e−t/τ ,

where P0 is the maximum of the emitting power and τ = 16 ns is the life time of the
excited state. It must be  ∞
E= P(t) dt = P0 τ ,
0

from which we get

E 3.4 × 10−11 J J
P0 = = = 2.1 × 10−3 = 2.1 mW.
τ 16 × 10−9 s s

Problem 3.2
Derive the selection rule m = 0 of the matrix element ⊗n∞ l ∞ m∞ |ẑ|nlm∇ of the
hydrogen atoms by using the formula [L̂z , ẑ] = 0.
Solution
Remember that for any eigenstate |nlm∇ of the electron in the hydrogen atom one has

L̂z |nlm∇ = m |nlm∇.

Moreover
[L̂z , ẑ] = L̂z ẑ − ẑL̂z .

We observe that

⊗n∞ l ∞ m∞ |[L̂z , ẑ]|nlm∇ = ⊗n∞ l ∞ m∞ |L̂z ẑ − ẑL̂z |nlm∇

= ⊗n∞ l ∞ m∞ |L̂z ẑ|nlm∇ − ⊗n∞ l ∞ m∞ |ẑL̂z |nlm∇
= m∞  ⊗n∞ l ∞ m∞ |ẑ|nlm∇ − ⊗n∞ l ∞ m∞ |ẑ|nlm∇ m
= (m∞ − m) ⊗n∞ l ∞ m∞ |ẑ|nlm∇.

Because [L̂z , ẑ] = 0 it follows that ⊗n∞ l ∞ m∞ |[L̂z , ẑ]|nlm∇ = 0 and also

(m∞ − m) ⊗n∞ l ∞ m∞ |ẑ|nlm∇ = 0.

72 3 Electromagnetic Transitions

From the last expression we get

⊗n∞ l ∞ m∞ |ẑ|nlm∇ = 0

if m∞ ≥= m. This is exactly the selection rule m = 0. We can then write

⊗n∞ l ∞ m∞ |ẑ|nlm∇ = ⊗n∞ l ∞ m|ẑ|nlm∇ ∂m∞ ,m .

Problem 3.3
A gas of hydrogen atoms is prepared in the state |n = 2, l = 1, m = 0∇. Calculate
the electric dipole moment associated to the transition to the ground-state |n = 1,
l = 0, m = 0∇.
Solution
The generic eigenfunction of the electron in the hydrogen atom in spherical coordi-
nates is given by
σnlm (r, φ, ν) = Rnl (r) Ylm (φ, ν),

where Rnl (r) is the radial wavefunction, which depends on the quantum numbers n
(principal) ed l (angular momentum), while Ylm (φ, ν) is the angular wavefunction,
which depends on the quantum numbers l (angular momentum) and m (projection
of angular momentum).
The wavefunction of the the electron in the ground-state is

σ100 (r) = R10 (r)Y00 (φ, ν),

where
2
R10 (r) = 3/2
e−r/r0
r0
1
Y00 (φ, ν) = ∓ ,
4β

with r0 the Bohr radius. The wavefunction of the excited state is instead

σ210 (r) = R21 (r)Y10 (φ, ν),

where
1 r
R21 (r) = ∓ e−r/(2r0 )
3(2r0 )3/2 r0

3
Y10 (φ, ν) = cos φ.
4β
3.7 Solved Problems 73

The vector of the dipole transition is given by


∗
rT = ⊗σ100 |r|σ210 ∇ = d 3 r σ100 (r) r σ210 (r).

Remembering that in spherical coordinates

r = (r cos ν sin φ, r sin ν sin φ, r cos φ )

with ν ∈ [0, 2β] and φ ∈ [0, β], and moreover

d 3 r = dr r 2 dν dφ sin φ,

the 3 components of the transition vector rT = (xT , yT , zT ) read

∓  ∞  2β  β
3
xT = drr 2 R10 (r)rR21 (r) dν cos ν dφ sin2 φ cos φ,
4β 0 0 0
∓  ∞  2β  β
3
yT = drr 2 R10 (r)rR21 (r) dν sin ν dφ sin2 φ cos φ,
4β 0 0 0
∓  ∞  2β  β
3
zT = drr 2 R10 (r)rR21 (r) dν dφ sin φ cos2 φ.
4β 0 0 0

One finds immediately that

xT = yT = 0,

because  
2β 2β
dν cos ν = dν sin ν = 0,
0 0

while  
β 1 2
dφ sin φ cos2 φ = d(cos φ) cos2 φ = .
0 −1 3

As a consequence, we have
 ∞  ∞
1 2 5 r0 2 8 r0
zT = ∓ 4 r 4 e−3r/(2r0 ) dr = ∓ η 4 e−η dη = ∓ ,
3 2r0 0 36 2 0 35 2

because  ∞
η n e−η dη = (n + 1) = n!.
0
74 3 Electromagnetic Transitions

Finally, the modulus of the electric dipole is

28
|d| = e|rT | = e zT = ∓ e r0 ,
35 2

where e is the electric charge of the electron. Performing the numerical calculation
we find

|d| = 1.6 × 10−19 C × 0.53 × 10−10 m × 0.74 = 6.3 × 10−30 C m.

Problem 3.4
Calculate the rates of spontaneous emission W spont to the ground-state in a gas of
hydrogen atoms which are initially:
(a) in the state |n = 3, l = 2, m = 1∇;
(b) in the state |n = 3, l = 1, m = 1∇.

Solution
The ground-state corresponds to |n = 1, l = 0, m = 0∇.

(a) In this case the transition probability is zero, within the dipolar approximation,
because l = 2 and the permitted transitions are only with l = ±1 and
m = 0, ±1.
(b) In this case the calculations must be done explicitly. The wavefunction of the
electron in the ground-state is

1 1
σ100 (r) = ∓ 3/2 e−r/r0 ,
βr
0

while the wavefunction of the excited state is

 
2 r r
σ311 (r) = 3 ∓ 5/2 1 − e−r/(3r0 ) eiν sin φ,
3 βr 6r0
0

where r0 is the Bohr radius. The vector of the dipole transition is given by

∗
rT = ⊗σ100 |r|σ311 ∇ = d 3 r σ100 (r) r σ311 (r).

Remembering that in spherical coordinates

r = (r cos ν sin φ , r sin ν sin φ , r cos φ ) ,

and also
d 3 r = dr r 2 dν dφ sin φ ,
3.7 Solved Problems 75

the 3 components of the transition vector rT = (xT , yT , zT ) read

 ∞    2β  β
2 r −4r/(3r0 )
xT = − 3 dr r 1− 4
e dν cos ν e iν
dφ sin3 φ,
3 β r04 0 6r0 0 0
 ∞    2β  β
2 r −4r/(3r0 )
yT = − 3 dr r 4
1 − e dν sin ν e iν
dφ sin3 φ,
3 β r04 0 6r0 0 0
 ∞    2β  β
2 r −4r/(3r0 )
zT = − 3 dr r 4
1− e dν e iν
dφ cos φ sin2 φ.
3 β r04 0 6r0 0 0

One find immediately that

zT = 0,

because  β
dφ cos φ sin2 φ = 0,
0

while  β 22
dφ sin3 φ = .
0 3

In addition, we get
 2β  2β
dν cos ν eiν = dν cos2 ν = β,
0 0

and similarly
 2β  2β
dν sin ν e iν
=i dν sin2 ν = i β.
0 0

Instead, for the radial part

 ∞  
r 37
dr r 1−4
e−4r/(3r0 ) = 10 r05 ,
0 6r0 2

and consequently
   2
2 37 5 2 33
xT = − 3 r β = − r0
3 β r04 210 0 3 27
   
2 37 5 22 33
yT = − 3 r 0 i β = −i r0 .
3 β r04 210 3 27
76 3 Electromagnetic Transitions

The squared modulus of the transition vector is given by

36 2
|rT |2 = |xT |2 + |yT |2 = 2 r = 0.088 r02 = 2.48 × 10−22 m2 ,
214 0

where r0 = 5.29 × 10−11 m.

The rate of spontaneous transition is given by the formula

4
W spont = ψ εT3 |rT |2 ,
3c2

where c = 3 × 108 m/s, ψ = e2 /(4β0 c) → 1/137,

 
E3 − E1 13.60 eV 1 13.60 × 1.60 × 10−19 J 8
εT = = 1− 2 =
  3 1.05 × 10−34 J s/rad 9
= 1.84 × 1016 rad/s.

Finally, the rate of spontaneous transition reads

W spont = 1.67 × 108 rad/s.

Problem 3.5
A gas of hydrogen atoms is irradiated from all directions by light with spectral density
 
(ε − ε0 )2
α(ε) = α0 exp − ,
1,000 ε02

where ε0 is the frequency of the ψ-Lymann (1s–2p) transition.

(a) Calculate the value of α0 which produces stimulated emission 2p↓1s with the
rate W stimul = 12.5 × 108 s−1 .
(b) Determine the effective electric field Erms of the incoming light, defined as
 ∞
1
0 Erms
2
= α(ε) dε.
2 0

Solution

(a) From the Einstein relations we know that the rate of stimulated emission W stimul
can be written as β
W stimul = |d|2 α(ε0 ),
30 2

where d is the electric dipole momentum. In this transition we are considering

3.7 Solved Problems 77

d = −e⊗2p|r|1s∇,

where e is the electric charge of the electron. This dipole has been previously calcu-
lated and its modulus is equal to

|d| = 6.3 × 10−30 C m.

Because in this case α(ε0 ) = α0 , we obtain

30 2 W stimul
α0 = .
β |d|2

Finally, we get
α0 = 2.95 × 10−12 J s/m3 .

(b) The integral to be calculated is

   
∞ ∞ (ε − ε0 )2
E= α(ε) dε = α0 exp − dε.
0 0 1,000 ε02

With the position

ε − ε0
t=∓
1,000 ε0

the integral becomes

  ∞  ∓  ∞ 
E= 1,000α0 ε0 exp −t 2
dt → 10 10 α0 ε0 exp −t 2 dt
1
− ∓1,000 0
∓
∓ β ∓
= 10 10 α0 ε0 = 5 10β α0 ε0 .
2
The frequency ε0 is
 
13.6 eV 1 1
ε0 = − − 2 = 1.55 × 1016 s−1 ,
 2 2 1

and consequently
E = 1.28 × 106 J/m3 .

The effective electric field reads


2E
Erms = = 5.38 × 108 V/m,
0

where 0 = 8.85 × 10−12 J/(m V2 ) is the dielectric constant in the vacuum.

78 3 Electromagnetic Transitions

Problem 3.6
Calculate both natural and Doppler width of the ψ-Lymann line in a hydrogen gas at
the temperature 1,000 K, knowing that the life time of the excited state is 0.16×10−8 s
and the wavelength of the transition is 1214 × 10−10 m.
Solution
The ψ-Lymann line is associated to the 1s ↓ 2p transition of the hydrogen atom.
The natural width between the states |i∇ and |j∇ is given by
 
1 1 1
νN = +
2β τi τj

where τi is the life time of the state |i∇ and τj is the life time of the state |j∇. In our
problem τ1s = ∞, because |1s∇ is the ground-state, while τ2p = 0.16 × 10−8 s for
the excited state |2p∇. It follows that
   
1 1 1 1 1
νN = + = 0+ s−1 = 9.9 × 107 Hz.
2β τ1s τ1p 6.28 0.16 × 10−8

The Doppler width depends instead on the temperature T = 1,000 K and on the
frequency ν of the transition according to the formula
 
 kB T ν kB T
νD = ν 8 ln(2) 2
= 8 ln(2) ,
mH c c mH

where kB = 1.3 × 10−23 J/K is the Boltzmann constant, mH = 1.6 × 10−27 kg is

the mass of an hydrogen atom, while c = 3 × 108 m/s is the speed of light in the
vacuum. We know that
ν 1 1
= = = 8.22 × 106 m−1 ,
c λ 1.216 × 10−7 m

and then we get


1.3 × 10−23 × 103
νD = 8.22 × 106 × 2.35 × Hz = 5.6 × 1010 Hz.
1.6 × 10−27

Problem 3.7
Calculate the pressure (collisional) width of the ψ-Lymann line for a gas of hydrogen
atoms with density 1012 atoms/m3 and collisional cross-section 10−19 m2 at the
temperature 103 K.
3.7 Solved Problems 79

Solution
The collisional width is given by

1
νC = ,
2βτcol

where τcol is the collision time. This time depends on the cross-section σ = 10−19 m2
and on the gas density n = 106 m−3 according to the formula

1
τcol = ,
n σ vmp
∓
where vmp = 2kB T /mH is the velocity corresponding the the maximum of the
Maxwell-Boltzmann distribution. We have then

n σ 2kB T
νC = .
2β mH

Because T = 103 K, kB = 1.3 × 10−23 J/K and mH = 1.6 × 10−27 kg, we

finally obtain
νC = 6.4 × 10−5 Hz.

Problem 3.8
Helium atoms in a gas absorb light in the transition a ↓ b, where |a∇ and |b∇ are
two excited states. Knowing that the wavelength of the transition is 501.7 nm and
that the life times are τb = 1.4 ns and τa = 1 ms, calculate: (a) the natural width; (b)
the Doppler width. The gas is at the temperature 1,000 K.
Solution
The natural width of the spectral line between the states |a∇ e |b∇ is given by
 
1 1 1
νN = + .
2β τa τb

Consequently we obtain
 
1 1 1
νN = + Hz = 1.14 × 108 Hz.
6.28 10−3 1.4 × 10−9

The Doppler width is instead


ν kB T
νD = 8 ln(2) ,
c mHe
80 3 Electromagnetic Transitions

where T = 103 K is the temperature of the gas, kB = 1.3 × 10−23 J/K is the
Boltzmann constant, mHe = 4 × 1.6 × 10−27 kg is the mass of a helium atom, while
c = 3 × 108 m/s is the velocity of light in the vacuum. We know that

ν 1 1
= = = 1.99 × 106 m−1 ,
c λ 501.7 × 10−9 m

and then we get

νD = 6.77 × 109 Hz.

Further Reading

For classical and quantum electrodynamics and radiative transitions:

F. Mandl, G. Shaw, Quantum Field Theory, Chap. 1, Sects. 1.3 and 1.4 (Wiley, New
York, 1984).
For life time and line widths:
B.H. Bransden, C.J. Joachain: Physics of Atoms and Molecules, Chap. 4, Sects. 4.6
and 4.7 (Prentice Hall, Upper Saddle River, 2003).
Chapter 4
The Spin of the Electron

In this chapter we explain the origin of the intrinsic angular momentum, also known
as the spin, of the electron on the basis of the Dirac equation. After introducing
the Dirac equation for a relativistic massive and charged particle coupled to the
electromagnetic field, we study its non relativistic limit deriving the Pauli equation.
This Pauli equation, which is nothing else than the Schrödinger equation with an
additional term containing the spin operator, predicts very accurately the magnetic
moment of the electron. Finally, we discuss the relativistic hydrogen atom and the
fine-structure corrections to the non relativistic one.

4.1 The Dirac Equation

The classical energy of a nonrelativistic free particle is given by

p2
E= , (4.1)
2m
where p is the linear momentum and m the mass of the particle. The Schrödinger
equation of the corresponding quantum particle with wavefunction ν(r, t) is easily
obtained by imposing the quantization prescription

ρ
E ∧ i , p ∧ −i∇. (4.2)
ρt
In this way one gets the time-dependent Schrödinger equation of the free particle,
namely
ρ 2 2
i ν(r, t) = − ∇ ν(r, t), (4.3)
ρt 2m

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 81

DOI: 10.1007/978-3-319-05179-6_4, © Springer International Publishing Switzerland 2014
82 4 The Spin of the Electron

obtained for the first time in 1926 by Erwin Schrödinger. The classical energy of a
relativistic free particle is instead given by

E= p2 c2 + m 2 c4 , (4.4)

where c is the speed of light in the vacuum. By applying directly the quantization
prescription (4.2) one finds

ρ 
i ν(r, t) = −2 c2 ∇ 2 + m 2 c4 ν(r, t). (4.5)
ρt
This equation is quite suggestive but the square-root operator on the right side is
a very difficult athematical object. For this reason in 1927 Oskar Klein and Walter
Gordon suggested to start with

E 2 = p2 c2 + m 2 c4 (4.6)

and then to apply the quantization prescription (4.2). In this way one obtains

ρ2  
− 2 ν(r, t) = − ∇
2 2 2
c + m 2 4
c ν(r, t) (4.7)
ρt 2
the so-called Klein-Gordon equation, which can be re-written in the following form
 
1 ρ2 m 2 c2
− ∇2 + 2 ν(r, t) = 0, (4.8)
c ρt
2 2 

i.e. a generalization of Maxwell’s wave equation for massive particles. This equation
has two problems: (i) it admits solutions with negative energy; (ii) the space integral
over the entire space of the non negative probability density π(r, t) = |ν(r, t)|2 ∇ 0
is generally not time-independent, namely

d
π(r, t) d 3 r →= 0. (4.9)
dt R3

Nowadays we know that to solve completely these two problems it is necessary to pro-
mote ν(r, t) to a quantum field operator. Within this second-quantization (quantum
field theory) approach the Klein-Gordon equation is now used to describe relativistic
particles with spin zero, like the pions or the Higgs boson.
In 1928 Paul Dirac proposed a different approach to the quantization of the rel-
ativistic particle. To solve the problem of Eq. (4.9) he considered a wave equation
with only first derivatives with respect to time and space and introduced the classical
energy
E = c α̂ · p + β̂ mc2 , (4.10)
4.1 The Dirac Equation 83

such that squaring it one recovers Eq. (4.6). This condition is fulfilled only if α̂ =
(α̂1 , α̂2 , α̂3 ) and β̂ satisfy the following algebra of matrices

α̂12 = α̂22 = α̂32 = β̂ 2 = Iˆ, (4.11)

α̂i α̂ j + α̂ j α̂i = 0̂, i →= j (4.12)
α̂i β̂ + β̂ α̂i = 0̂, ∀i (4.13)

where 1̂ is the identity matrix and 0̂ is the zero matrix. The smallest dimension in
which the so-called Dirac matrices α̂i and β̂ can be realized is four. In particular, one
can write
   
0̂2 ε̂i Iˆ2 0̂2
α̂i = , β̂ = , (4.14)
ε̂i 0̂2 0̂2 − Iˆ2

where Iˆ2 is the 2 × 2 identity matrix, 0̂2 is the 2 × 2 zero matrix, and
     
01 0 −i 1 0
ε̂1 = ε̂2 = ε̂3 = (4.15)
10 i 0 0 −1

are the Pauli matrices. Equation (4.10) with the quantization prescription (4.2) gives

ρ  
i (r, t) = −ic α̂ · ∇ + β̂ mc2 (r, t), (4.16)
ρt

which is the Dirac equation for a free particle. Notice that the wavefunction (r, t)
has four components in the abstract space of Dirac matrices, i.e. this spinor field can
be written
 ⎛
ν1 (r, t)
 ν2 (r, t) ⎜
(r, t) =  ⎜
 ν3 (r, t) ⎝ . (4.17)
ν4 (r, t)

In explicit matrix form the Dirac equation is thus given by

 ⎛  ⎛
ν1 (r, t) ν1 (r, t)
ρ  ν2 (r, t) ⎜  ⎜
⎜ = Ĥ  ν2 (r, t) ⎜
i  (4.18)
ρt  ν3 (r, t) ⎝  ν3 (r, t) ⎝
ν4 (r, t) ν4 (r, t)

where
84 4 The Spin of the Electron
 ρ ρ ⎛
mc2 0 −i c ρz −i c( ρx − i ρρy )
 mc2 ρ
−i c( ρx + i ρρy ) ρ
i c ρz ⎜
 0 ⎜
Ĥ =  ρ ρ ρ ⎜ . (4.19)
 −i c ρz −i c( ρx − i ρ y ) −mc 2 0 ⎝
ρ ρ ρ
−i c( ρx + i ρ y ) i c ρz 0 −mc 2

It is easy to show that the Dirac equation satisfies the differential law of current
conservation. In fact, left-multiplying Eq. (4.16) by
⎞ ⎟
 + (r, t) = ν1√ (r, t), ν2√ (r, t), ν3√ (r, t), ν4√ (r, t) (4.20)

we get
ρ
i + = −ic  + α̂ · ∇ + mc2  + β̂ . (4.21)
ρt
Considering the Hermitian conjugate of the Dirac Eq. (4.16) and right-multiplying
it by (r, t) we find instead

ρ +
− i  = ic α̂ · ∇ +  + mc2  + β̂ . (4.22)
ρt
Subtracting the last two equations we obtain the continuity equation

ρ
π(r, t) + ∇ · j(r, t) = 0, (4.23)
ρt
where
⎠
4
+
π(r, t) =  (r, t)(r, t) = |νi (r, t)|2 (4.24)
i=1

is the probability density, and j(r, t) is the probability current with three components

jk (r, t) = c  + (r, t)α̂k (r, t). (4.25)

Finally, we observe that from the continuity Eq. (4.23) one finds

d
π(r, t) d 3 r = 0, (4.26)
dt R3

by using the divergence theorem and imposing a vanishing current density on the
border at infinity. Thus, contrary to the Klein-Gordon equation, the Dirac equation
does not have the probability density problem.
4.2 The Pauli Equation and the Spin 85

4.2 The Pauli Equation and the Spin

In this section we analyze the non-relativistic limit of the Dirac equation. Let us
suppose that the relativistic particle has the electric charge q. In presence of an
electromagnetic field, by using the Gauge-invariant substitution

ρ ρ
i ∧ i − q λ(r, t) (4.27)
ρt ρt
−i∇ ∧ −i∇ − qA(r, t) (4.28)

in Eq. (4.16), we obtain

ρ  ⎞ ⎟ 
i (r, t) = c α̂ · p̂ − qA(r, t) + β̂ mc2 + q λ(r, t) (r, t), (4.29)
ρt

where p̂ = −i∇, λ(r, t) is the scalar potential and A(r, t) the vector potential.
To workout the non-relativistic limit of Eq. (4.29) it is useful to set
 ⎛
ν1 (r, t)  
 ν2 (r, t) ⎜
(r, t) = e−imc t/ 
2
⎜ = e−imc2 t/ ν(r, t) , (4.30)
 θ1 (r, t) ⎝ θ(r, t)
θ2 (r, t)

where ν(r, t) and θ(r, t) are two-component spinors, for which we obtain
    
ρ ν qλ c σ̂ · (p̂ − qA) ν
i = (4.31)
ρt θ c σ̂ · (p̂ − qA) q λ − 2mc2 θ
⎞ ⎟
where σ̂ = ε̂1 , ε̂2 , ε̂3 . Remarkably, only in the lower equation of the system it
appears the mass term mc2 , which is dominant in the non-relativistic limit. Indeed,
ρ
under the approximation i ρt − q λ + 2mc2 θ ⊗ 2mc2 θ, the previous equations
become
 ρν    
i ρt qλ c σ̂ · (p̂ − qA) ν
= , (4.32)
0 c σ̂ · (p̂ − qA) −2mc2 θ

from which
σ̂ · (p̂ − qA)
θ= ν. (4.33)
2mc
Inserting this expression in the upper equation of the system (4.32) we find
⎡⎣ ⎤2 ⎦
ρ σ̂ · (p̂ − qA)
i ν = + q λ ν. (4.34)
ρt 2m
86 4 The Spin of the Electron

From the identity

⎣ ⎞ ⎟⎤2
σ̂ · p̂ − qA = (p̂ − qA)2 − i q (p̂ ∞ A) · σ̂ (4.35)

where p̂ = −i∇, and using the relation B = ∇ ∞ A which introduces the magnetic
field we finally get
 
ρ (−i∇ − qA(r, t))2 q
i ν(r, t) = − B(r, t) · Ŝ + q λ(r, t) ν(r, t),
ρt 2m m
(4.36)
that is the so-called Pauli equation with


Ŝ = σ̂. (4.37)
2
the spin operator. This equation was introduced in 1927 (a year before the Dirac
equation) by Wolfgang Pauli as an extension of the Schrödinger equation with the
phenomenological inclusion of the spin operator. If the magnetic field B is constant,
the vector potential can be written as

1
A= B∞r (4.38)
2
and then
⎞ ⎟2
p̂ − qA = p̂2 − 2qA · p̂ + q 2 A2 = p̂2 − qB · L̂ + q 2 (B ∞ r)2 , (4.39)

with L̂ = r ∞ p̂ the orbital angular momentum operator. Thus, the Pauli equation for
a particle of charge q in a constant magnetic field reads
 2 2   q2 
ρ  ∇ q
i ν(r, t) = − − B · L̂ + 2Ŝ + (B ∞ r)2 + q λ(r, t) ν(r, t).
ρt 2m 2m 2m
(4.40)

In conclusion, we have shown that the spin Ŝ naturally emerges from the Dirac equa-
tion. Moreover, the Dirac equation predicts very accurately the magnetic moment μ S
of the electron (q = −e, m = m e ) which appears in the spin energy E s = −μ̂ S · B
where μB
μ S = −ge Ŝ (4.41)


with gyromagnetic ratio ge = 2 and Bohr magneton μ B = e/(2m) ⊗ 5.79 · 10−5

eV/T.
4.3 Dirac Equation with a Central Potential 87

4.3 Dirac Equation with a Central Potential

We now consider the stationary Dirac equation with the confining spherically-
symmetric potential V (r ) = V (|r|), namely
 
−ic α̂ · ∇ + β̂ mc2 + V (r ) (r) = E (r). (4.42)

This equation is easily derived from Eq. (4.29) setting A = 0, qλ = V (r ), and

(r, t) = e−i Et/ (r). (4.43)

The relativistic Hamiltonian

Ĥ = −ic α̂ · ∇ + β̂ mc2 + V (r ) (4.44)

commutes with the total angular momentum operator


Ĵ = L̂ + Ŝ = r ∞ p̂ + σ̂ (4.45)
2
because the external potential is spherically symmetric. In fact, one can show that

[ Ĥ , L̂] = −ic α̂ ∞ p̂ = −[ Ĥ , Ŝ]. (4.46)

Consequently one has

[ Ĥ , Ĵ] = 0, (4.47)

and also
[ Ĥ , Jˆ2 ] = 0, [ Jˆ2 , Jˆx ] = [ Jˆ2 , Jˆy ] = [ Jˆ2 , Jˆz ] = 0, (4.48)

where the three components Jˆx , Jˆy , Jˆz of the total angular momentum Ĵ =
( Jˆx , Jˆy , Jˆz ) satisfy the familiar commutation relations

[ Jˆi , Jˆj ] = i ∂i jk Jˆk (4.49)

with ⎧
⎨ 1 if (i, j, k) is (x, y, z) or (z, x, y) or (y, z, x)
∂i jk = −1 if (i, j, k) is (x, z, y) or (z, y, x) or (y, x, z) (4.50)
⎩
0 if i = j or i = k or j = k

the Levi-Civita symbol (also called Ricci-Curbastro symbol). Note that these com-
mutation relations can be symbolically synthesized as

Ĵ ∞ Ĵ = i Ĵ. (4.51)
88 4 The Spin of the Electron

Indicating the states which are simultaneous eigenstates of Ĥ , Jˆ2 and Jˆz as |n jm j ∈,
one has

Ĥ |n jm j ∈ = E n j |n jm j ∈, (4.52)
Jˆ2 |n jm j ∈ = 2 j ( j + 1) |n jm j ∈, (4.53)
Jˆz |n jm j ∈ = m j |n jm j ∈, (4.54)

where j is the quantum number of the total angular momentum and m j = − j, − j +1,
− j + 2, ..., j − 2, j − 1, j the quantum number of the third component of the total
angular momentum.
In conclusion, we have found that, contrary to the total angular momentum Ĵ, the
orbital angular momentum L̂ and the spin Ŝ are not constants of motion of a particle
in a central potential.

4.3.1 Relativistic Hydrogen Atom and Fine Splitting

Let us consider now the electron of the hydrogen atom. We set q = −e, m = m e and

e2 e2
V (r ) = − =− . (4.55)
4ψφ0 |r| 4ψφ0 r

Then the eigenvalues E n j of Ĥ are given by

mc2
En j =  − mc2 , (4.56)
α2
1+   2
n− j− 2 + ( j+ 21 )2 −α2
1

with α = e2 /(4ψ∂0 c) ⊗ 1/137 the fine-structure constant. We do not prove this
remarkable quantization formula, obtained independently in 1928 by Charles Galton
Darwin and Walter Gordon, but we stress that expanding it in powers of the fine-
structure constant α to order α4 one gets
 ⎡ ⎦
α2 1 3
E n j = E n(0) 1+ − , (4.57)
n j+ 1
2
4n

where
1 α2 13.6 eV
E n(0) = − mc2 2 = − (4.58)
2 n n2
4.3 Dirac Equation with a Central Potential 89

Fig. 4.1 Fine splitting for the

hydrogen atom. On the left
there are the non-relativistic
energy levels obtained by
solving the Schrödinger equa-
tion with a Coulomb potential
(Coulomb). On the right there
are the energy levels obtained
by taking into account rel-
ativistic corrections (Fine
Structure)

is the familiar Bohr quantization formula of the non relativistic hydrogen atom. The
term which corrects the Bohr formula, given by
⎡ ⎦
α2 1 3
E = E n(0) − , (4.59)
n j+ 1
2
4n

is called fine splitting correction. This term removes the non relativistic degeneracy
of energy levels, but not completely: double-degenerate levels remain with the same
quantum numbers n and j but different orbital quantum number l = j ±1/2 (Fig 4.1).
We have seen that, strictly speaking, in the relativistic hydrogen atom nor the
orbital angular momentum L̂ nor the spin Ŝ are constants of motion. As a consequence
l, m l , s and m s are not good quantum numbers. Nevertheless, in practice, due to the
smallness of fine-splitting corrections, one often assumes without problems that both
L̂ and Ŝ are approximately constants of motion.

4.3.2 Relativistic Corrections to the Schrödinger Hamiltonian

It is important to stress that the relativistic Hamiltonian Ĥ of the Dirac equation

in a spherically-symmetric potential, given by Eq. (4.44), can be expressed as the
familiar non relativistic Schrödinger Hamiltonian

2 2
Ĥ0 = − ∇ + V (r ) (4.60)
2m
plus an infinite sum of relativistic quantum corrections. To this aim one can start from
the Dirac equation, written in terms of bi-spinors, i.e. Eq. (4.31) with A(r, t) = 0
and qλ(r, t) = V (r ), which gives
90 4 The Spin of the Electron
    
ν̃ V (r ) c σ̂ · p̂ ν̃
E = (4.61)
θ̃ c σ̂ · p̂ V (r ) − 2mc2 θ̃

setting ν(r, t) = ν̃(r) e−i Et/ and θ(r, t) = θ̃(r) e−i Et/. The lower equation of
this system can be written as

c σ̂ · p̂
θ̃ = ν̃. (4.62)
E − V (r ) + 2mc2

This is an exact equation. If E − V (r ) ∓ 2mc2 the equation becomes

σ̂ · p̂
θ̃ = ν̃, (4.63)
2mc

which is exactly the stationary version of Eq. (4.33) with A(r, t) = 0. We can
do something better by expanding Eq. (4.62) with respect to the small term (E −
V (r ))/2mc2 obtaining
 
σ̂ · p̂ E − V (r )
θ̃ = 1− + ... ν̃. (4.64)
2mc 2mc2

Inserting this expression in the upper equation of the system and neglecting the higher
order terms symbolized by the three dots, and after some tedious calculations, one
finds
E ν̃ = Ĥ ν̃, (4.65)

where
Ĥ = Ĥ0 + Ĥ1 + Ĥ2 + Ĥ3 , (4.66)

with

4
Ĥ1 = − ∇ 4, (4.67)
8m 3 c2
1 1 d V (r )
Ĥ2 = L · S, (4.68)
2m 2 c2 r dr
2
Ĥ3 = ∇ 2 V (r ), (4.69)
8m 2 c2

with Ĥ1 the relativistic correction to the electron kinetic energy, Ĥ2 the spin-orbit
correction, and Ĥ3 the Darwin correction.
If the external potential V(r) is that of the hydrogen atom, i.e. V (r ) = −e2 /
(4ψφ0 |r|), one finds immediately that H3 = (2 e2 )/(8m 2 c2 φ0 )α(r) because
∇ 2 (1/|r|) = −4ψα(r). In addition, by applying the first-order perturbation theory
to Ĥ with Ĥ0 unperturbed Hamiltonian, one gets exactly Eq. (4.57) of fine-structure
4.3 Dirac Equation with a Central Potential 91

correction. Physically one can say that the relativistic fine structure is due to the
coupling between the spin Ŝ and the orbital angular momentum L̂ of the electron.
Moreover, we observe that

1  ˆ2 
L̂ · Ŝ = J − L̂ 2 − Ŝ 2 (4.70)
2
because
Jˆ2 = (L̂ + Ŝ)2 = L̂ 2 + Ŝ 2 + 2 L̂ · Ŝ, (4.71)

and this means that the Hamiltonian Ĥ of Eq. (4.66) commutes with L̂ 2 and Ŝ 2 but
Ĥ does not commute with L̂ z and Ŝz .
Actually, also the nucleus (the proton in the case of the hydrogen atom) has its
spin Î which couples to electronic spin to produce the so-called hyperfine structure.
However, typically, hyperfine structure has energy shifts orders of magnitude smaller
than the fine structure.

4.4 Solved Problems

Problem 4.1
One electron is set in a uniform magnetic field B = (0, 0, B0 ). Calculate the ex-
pectation value of the spin Ŝ along the x axis if at t = 0 the spin is along the z
axis.
Solution
The Hamiltonian operator of the spin is

Ĥ = −μ̂ S · B,

where μ̂ is the magnetic dipole moment of the electron, given by

μB 1
μ S = −ge Ŝ = − ge μ B σ̂,
 2
with ge = 2.002319 ⊗ 2 the gyromagnetic ratio of the electron, μ B = e/(2m) =
9.27 · 10−24 J/T the Bohr magneton and σ̂ = (ε̂1 , ε̂2 , ε̂3 ) the vector of Pauli
matrices:
 
01
ε̂1 =
10
 
0 −i
ε̂2 =
i 0
 
1 0
ε̂3 = .
0 −1
92 4 The Spin of the Electron

The Hamiltonian operator can be written as

 
1 1 1 0
Ĥ = σ0 ε̂3 = σ0 ,
2 2 0 −1

where σ0 = ge μ B B0 / is the Larmor frequency of the system, with  = h/(2ψ) the

reduced Planck constant. The initial state of the system is
 
1
|ν(0)∈ = = | ↑ ∈,
0

while  
0
= | ↓∈
1

is the state along the third component of spin. The state at time t is then given by

|ν(t)∈ = e−i Ĥ t/|ν(0)∈ = e−iσ0 ε̂3 t/2 | ↑ ∈ = e−iσ0 t/2 | ↑ ∈,

because
ε̂3 | ↑ ∈ = | ↑ ∈

and
e−iσ0 ε̂3 t/2 | ↑ ∈ = e−iσ0 t/2 | ↑ ∈.

The expectation value at time t of the spin component along the x axis is then

  
 Ŝx (t)∈ = ν(t) |ε̂1 |ν(t)∈ = ↑ |eiσ0 t/2 ε̂1 e−iσ0 t/2 | ↑ ∈ = ↑ |ε̂1 | ↑ ∈.
2 2 2
Observing the
ε̂1 | ↑ ∈ = | ↓ ∈,

and also
↑ | ↓ ∈ = 0,

we conclude that
 Ŝx (t)∈ = 0.

This means that if initially the spin is in the same direction of the magnetic field it
remains in that direction forever: it is a stationary state with a time-dependence only
in the phase. Instead, the components of spin which are orthogonal to the magnetic
field have always zero expectation value.
4.4 Solved Problems 93

Problem 4.2
One electron is set in a uniform magnetic field B = (0, 0, B0 ). Calculate the ex-
pectation value of the spin Ŝ along the x axis if at t = 0 the spin is along the x
axis.
Solution
The Hamiltonian operator of spin is given by

1
Ĥ = σ0 ε̂3
2
where σ0 = ge μ B B0 / is the Larmor frequency of the system, with ge gyromagnetic
factor and μ B Bohr magneton. The initial state of the system is

1
|ν(0)∈ = √ (| ↑ ∈ + | ↓ ∈) ,
2

because
1 ⎞ ⎟ 1
ε̂1 |ν(0)∈ = √ ε̂1 | ↑ ∈ + ε1 | ↓ ∈ = √ (| ↓ ∈ + | ↑ ∈) = |ν(0)∈.
2 2

The state at time t is then

1
|ν(t)∈ = e−i Ĥ t/|ν(0)∈ = e−iσ0 ε̂3 t/2 √ (| ↑ ∈ + | ↓ ∈)
2
1  −iσ0 t/2 
=√ e | ↑ ∈ + eiσ0 t/2 | ↓ ∈ ,
2

because

ε̂3 | ↑ ∈ = | ↑ ∈
ε̂3 | ↓ ∈ = −| ↓ ∈

and

e−iσ0 ε̂3 t/2 | ↑ ∈ = e−iσ0 t/2 | ↑ ∈

e−iσ0 ε̂3 t/2 | ↓ ∈ = eiσ0 t/2 | ↓ ∈.

The expectation value at time t of the spin component along x axis reads
⎞ ⎟ ⎞
 Ŝx (t)∈ = ν(t) 2 |ε̂1 |ν(t)∈ = 4 ↑ |eiσ0 t/2 + ↓ |e−iσ0 t/2 ε̂1 e−iσ0 t/2 | ↑ ∈
⎟ ⎞ ⎟
+ eiσ0 t/2 | ↓ ∈ = 4 eiσ0 t ↑ | ↑ ∈ + e−iσ0 t ↓ | ↓ ∈ = 2 cos (σ0 t).
94 4 The Spin of the Electron

Problem 4.3
One electron with spin Ŝ, initially along z axis, is under the action of a time-dependent
magnetic field B = B1 (cos (σt)ex + sin (σt)e y ). Calculate the spin state at time t.
Solution
The Hamiltonian operator of spin is given by

1 ⎞ ⎟
Ĥ (t) = σ1 cos (σt) ε̂1 + sin (σt) ε̂2 ,
2
where σ1 = ge μ B B1 , with ge gyromagnetic factor of the electron and μ B Bohr
magneton. In matrix form the Hamiltonian operator reads
 
 0 σ1 e−iσt
Ĥ (t) = ,
2 σ1 eiσt 0

while the spin initial state is

 
1
|ν(0)∈ = | ↑ ∈ = .
0

The spin state at time t can be written instead as

   
1 0
|ν(t)∈ = a(t)| ↑ ∈ + b(t)| ↓ ∈ = a(t) + b(t) ,
0 1

where a(t) and b(t) are time-dependent functions to be determined with initial con-
ditions
a(0) = 1, b(0) = 0.

The time-dependent Schrödinger equation

ρ
i |ν(t)∈ = H (t)|ν(t)∈,
ρt
can be written as
    
a(t)  0 σ1 e−iσt a(t)
i = .
b(t) 2 σ1 eiσt 0 b(t)

Thus the complex functions a(t) and b(t) satisfy the following system of first-order
ordinary differential equations:

2 i ȧ(t) = σ1 e−iσt b(t),

2 i ḃ(t) = σ1 eiσt a(t).
4.4 Solved Problems 95

The time-dependent phase factors can be removed introducing the new variables c(t)
and d(t) such that

a(t) = eiσt/2 c(t),

d(t) = e−iσt/2 d(t).

The differential system then becomes

2 i ċ(t) = σ c(t) + σ1 d(t),

2 i ḋ(t) = −σ d(t) + σ1 c(t).

This system can be solved by using the Laplace transform

 ≥
F(s) = f (t) e−st dt
0

Indeed, by applying the Laplace transform, the system becomes

2 i (s C(s) − c(0)) = σ C(s) + σ1 D(s).

2 i (s D(s) − d(0)) = −σ D(s) + σ1 C(s).

Taking into account the initial conditions c(0) = 1 and d(0) = 0, one can get D(s)
from the second equation, namely
σ1
D(s) = C(s),
2is + σ

and insert it into the first one. In this way

s σ σR
C(s) = − +i ,
σ 2R +s 2 2σ R σ R + s 2
2


where σ R = 21 σ12 + σ 2 is the so-called Rabi frequency. By applying the Laplace
anti-transform we immediately obtain
σ
c(t) = − cos (σ R t) + i sin2 (σ R t).
2σ R

Coming back to the initial variables a(t) and b(t) we eventually find

σ2 σ12 σ12
|a(t)|2 = + cos2 (σ R t), |b(t)|2 = sin2 (σ R t).
4σ 2R 4σ 2R 4σ 2R
96 4 The Spin of the Electron

which clearly satisfy the ralation |a(t)|2 + |b(t)|2 = 1. Notice that for σ = 0 it
follows σ R = σ1 /2 from which

|a(t)|2 = cos2 (σ R t), |b(t)|2 = sin2 (σ R t).

In this case there is complete spin-flip during the dynamics.

Problem 4.4
Derive the non relativistic formula of Bohr for the spectrum of the hydrogen atom
from the relativistic expression

mc2
En j =  − mc2
α2
1+   2
n− j− 21 + ( j+ 21 )2 −α2

which is obtained from the Dirac equation, expanding it in powers of α ⊗ 1/137 at

order α2 .
Solution
Setting x = α2 the relativistic spectrum cha be written as

E(x) = mc2 f (x) − mc2 ,

where  −1/2
x
f (x) = 1 + √ ,
(A + B − x)2

with A = n − j − 1/2 e B = ( j + 1/2)2 . Let us now expand f (x) in MacLaurin

series at first order:
f (x) = f (0) + f ⊥ (0)x,

where
f (0) = 1
f ⊥ (0) = − 21 1
√ = − 21 n12 .
(A+ B)2

We get at first order in x

1 x 1 x
E(x) = mc2 − mc2 2 − mc2 = − mc2 2 ,
2 n 2 n
namely
1 α2
E n = − mc2 2 ,
2 n
which is exactly the non relativistic Borh formula.
4.4 Solved Problems 97

Problem 4.5
Calculate the fine splitting for the state |3 p∈ of the hydrogen atom.
Solution
From the relativistic formula of the energy levels of the hydrogen atom

mc2
En j =  − mc2
α2
1+   2
n− j− 21 + ( j+ 21 )2 −α2

expanding to order α4 we get

 ⎡ ⎦
α2 1 3
En j = E n(0) 1+ −
n j+ 1
2
4n

where
1 α2 13.6 eV
E n(0) = − mc2 2 = − .
2 n n2
The integer number j is the quantum number of the total angular momentum J =
L + S, where j = 1/2 if l = 0 and j = l − 1/2, l + 1/2 if l →= 0. The state |3 p∈
means n = 3 and l = 1, consequently j = 1/2 or j = 3/2. The hyperfine correction
for j = 1/2 reads
 
(0) α (0) α
2 1 2 13.6 eV 1
E 3, 1 = E 3 1− = E3 =− = −2.01 · 10−5 eV.
2 3 4 4 9 1372 · 4

The hyperfine correction for j = 3/2 is instead

 
(0) α (0) α
2 1 1 2 13.6 eV 1
E 3, 3 = E3 − = E3 =− = −6.71 · 10−6 eV.
2 3 2 2 12 9 1372 · 12

Further Reading

For the Dirac equation and the Pauli equation:

J.D. Bjorken and S.D. Drell, Relativistic Quantum Mechanics, chapt. 1, sections 1.1,
1.2, 1.3, 1.4 (McGraw-Hill, New York, 1964).
For the relativistic hydrogen atom and the fine splitting:
V.B. Berestetskii, E.M. Lifshitz, L.P. Pitaevskii, Relativistic Quantum Theory, vol.
4 of Course of Theoretical Physics, chapt. 4, sections 34, 35, 36 (Pergamon Press,
Oxford, 1971).
B.H. Bransden and C.J. Joachain: Physics of Atoms and Molecules, chapt. 5, sections
5.1 (Prentice Hall, Upper Saddle River, 2003).
Chapter 5
Energy Splitting and Shift Due
to External Fields

In this chapter we study the Stark effect and the Zeeman effect on atoms, and in
particular on the Hydrogen atom. In the Stark effect one finds that an external constant
electric field induces splitting of degenerate energy levels and energy shift of the
ground-state. Similarly, in the Zeeman effect an external constant magnetic field
induces splitting of degenerate energy levels, but the splitting properties strongly
depend on the intensity of the external magnetic field.

5.1 Stark Effect

Let us consider the Hydrogen atom under the action of a constant electric field E.
We write the constant electric field as

E = E uz = (0, 0, E), (5.1)

choosing the z axis in the same direction of E. The Hamiltonian operator of the
system is then given by

Ĥ = Ĥ0 + ĤI , (5.2)

where
p̂2 e2
Ĥ0 = − (5.3)
2m 4νρ0 r

is the non-relativistic Hamiltonian of the electron in the hydrogen atom (with m = me

the electron mass), while

ĤI = −e π(r) = eE · r = −d · E = eEz (5.4)

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 99

DOI: 10.1007/978-3-319-05179-6_5, © Springer International Publishing Switzerland 2014
100 5 Energy Splitting and Shift Due to External Fields

is the Hamiltonian of the interaction due to the electric scalar potential π(r) = −E · r
such that E = −∇π, with −e electric charge of the electron and d = −er the electric
dipole.
Let |nlml ∧ be the eigenstates of Ĥ0 , such that

Ĥ0 |nlml ∧ = En(0) |nlml ∧ (5.5)

with
13.6
En(0) = − eV, (5.6)
n2
the Bohr spectrum of the hydrogen atom, and moreover

L̂ 2 |nlml ∧ = 2 l(l + 1) |nlml ∧, L̂z |nlml ∧ =  ml |nlml ∧. (5.7)

At the first order of degenerate perturbation theory the energy spectrum is given by

En = En(0) + En(1) , (5.8)

(1)
where En is one of the eigenvalues of the submatrix ĤIn , whose elements are

al(n)    
 m ,lm = ∇nl ml |ĤI |nlml ∧ = eE∇nl ml |z|nlml ∧. (5.9)
l l

Thus, in general, there is a linear splitting of degenerate energy levels due to the
external electric field E. This effect is named after Johannes Stark, who discovered it
in 1913. Actually, it was discovered independently in the same year also by Antonino
Lo Surdo.
It is important to stress that the ground-state |1s∧ = |n = 1, l = 0, ml = 0∧ of the
hydrogen atom is not degenerate and for it ∇1s|z|1s∧ = 0. It follows that there is no
linear Stark effect for the ground-state. Thus, we need the second order perturbation
theory, namely

E1 = E1(0) + E1(1) + E1(2) , (5.10)

where
(1)
E1 = ∇100|eEz|100∧ = 0, (5.11)
 |∇nlml |eEz|100∧|2 ∀
(2) |∇n10|z|100∧|2
E1 = (0) (0)
= e 2 2
E (0) (0)
, (5.12)
nlml →=100 E1 − En n=2 E1 − En

where the last equality is due to the dipole selection rules. This formula shows that
the electric field produces a quadratic shift in the energy of the ground state. This
phenomenon is known as the quadratic Stark effect.
5.1 Stark Effect 101

The polarizability βp of an atom is defined in terms of the energy-shift E1 of

the atomic ground state energy E1 induced by an external electric field E as follows:

1
E1 = − βp E 2 . (5.13)
2
Hence, for the hydrogen atom we can write
∀
 |∇n10|z|100∧|2 9 e2 r02
βp = −2e2 (0) (0)
=− , (5.14)
E1 − En 4 E (0)
n=2 1

where the last equality is demonstrated in Problem 5.2, with r0 the Bohr radius.

5.2 Zeeman Effect

Let us consider the hydrogen atom under the action of a constant magnetic field B.
According to the Pauli equation, the Hamiltonian operator of the system is given by
 2
p̂ + eA e2
Ĥ = − − µ̂S · B (5.15)
2m 4νρ0 r

where e
µ̂S = − Ŝ (5.16)
m

is the spin dipole magnetic moment, with Ŝ the spin of the electron, and A is the
vector potential, such that B = ∇ √ A. Because the magnetic field B is constant, the
vector potential can be written as
1
A= B √ r, (5.17)
2
and then
 2
p̂ + eA = p̂2 + 2ep̂ · A + e2 A2 = p̂2 + 2eB · L̂ + e2 (B √ r)2 , (5.18)

with L̂ = r √ p̂. In this way the Hamiltonian can be expressed as

Ĥ = Ĥ0 + ĤI , (5.19)

where
p̂2 e2
Ĥ0 = − (5.20)
2m 4νρ0 r

is the non-relativistic Hamiltonian of the electron in the hydrogen atom, while

102 5 Energy Splitting and Shift Due to External Fields

e2
ĤI = −µ̂ · B + (B √ r)2 (5.21)
8m
is the Hamiltonian of the magnetic interaction, with
e
µ̂ = µ̂L + µ̂S = − (L̂ + 2̂S) (5.22)
2m
the total dipole magnetic moment of the electron. We now write the constant magnetic
field as

B = B uz = (0, 0, B), (5.23)

choosing the z axis in the same direction of B. In this way the interaction Hamiltonian
becomes

eB   e2 B2
ĤI = L̂z + 2Ŝz + (x 2 + y2 ). (5.24)
2m 8m
The first term, called paramagnetic term, grows linearly with the magnetic field B
while the second one, the diamagnetic term, grows quadratically. The paramagnetic
term is of the order of μB B, where μB = e/(2m) = 9.3 × 10−24 J/T = 5.29 ×
10−5 eV/T is the Bohr magneton. Because the unperturbed energy of Ĥ0 is of the
order of 10 eV, the paramagnetic term can be considered a small perturbation.

5.2.1 Strong-Field Zeeman Effect

Usually the diamagnetic term is much smaller than the paramagnetic one, and
becomes observable only for B of the order of 106 /n4 T, i.e. mainly in the astrophys-
ical context. Thus in laboratory the diamagnetic term is usually negligible (apart for
very large values of the principal quantum number n) and the effective interaction
Hamiltonian reads
eB  
HI = L̂z + 2Ŝz . (5.25)
2m
Thus (5.20) is the unperturbed Hamiltonian and (5.25) the perturbing Hamiltonian. It
is clear that this total Hamiltonian is diagonal with respect to the eigenstates |nlml ms ∧
and one obtains immediately the following energy spectrum

En,ml ,ms = En(0) + μB B (ml + 2ms ) , (5.26)

(0)
where En is the unperturbed Bohr eigenspectrum and μB = e/(2m) is the Bohr
magneton, with m the mass of the electron. Equation (5.26) describes the high-field
5.2 Zeeman Effect 103

Zeeman effect, first observed in 1896 by Pieter Zeeman. The field B does not remove
the degeneracy in l but it does remove the degeneracy in ml and ms . The selection
rules for dipolar transitions require ms = 0 and ml = 0, ±1. Thus the spectral
line corresponding to a transition n ⊗ n is split into 3 components, called Lorentz
triplet (Fig. 5.1).

5.2.2 Weak-Field Zeeman Effect

In the hydrogen atom the strong-field Zeeman effect is observable if the magnetic field
B is between about 1/n3 T and about 106 /n4 T, with n the principal quantum number.
In fact, as previously explained, for B larger than about 106 /n4 T the diamagnetic
term is no more negligible. Instead, for B smaller than about 1/n3 T the splitting due
to the magnetic field B becomes comparable with the splitting due to relativistic fine-
structure corrections. Thus, to study the effect of a weak field B, i.e. the weak-field
Zeeman effect, the unperturbed non-relativistic Hamiltonian Ĥ0 given by the Eq.
(5.20) is no more reliable. One must use instead the exact relativistic Hamiltonian
or, at least, the non-relativistic one with relativistic corrections, namely

Ĥ0 = Ĥ0,0 + Ĥ0,1 + Ĥ0,2 + Ĥ0,3 , (5.27)

where

2 2 e2
Ĥ0,0 = − ∞ − (5.28)
2m 4νρ0 r
 4
Ĥ0,1 = − 3 2 ∞ 4, (5.29)
8m c
1 1 dV (r)
Ĥ0,2 = L · S, (5.30)
2m2 c2 r dr
2
Ĥ0,3 = ∞ 2 V (r), (5.31)
8m2 c2

with Ĥ0,0 the non relativistic Hamiltonian, Ĥ0,1 the relativistic correction to the elec-
tron kinetic energy, Ĥ0,2 the spin-orbit correction, and Ĥ0,3 the Darwin correction.
In any case, ml and ms are no more good quantum numbers because L̂z and Ŝz do not
commute with the new Ĥ0 .
The Hamiltonian (5.27) commutes instead with L̂ 2 , Ŝ 2 , Ĵ 2 and Ĵz . Consequently,
for this Hamiltonian the good quantum numbers are n, l, s, j and mj . Applying again
the first-order perturbation theory, where now (5.27) is the unperturbed Hamiltonian
and (5.25) is the perturbing Hamiltonian, one obtains the following energy spectrum

(0) (1)
En,l,j,mj = En,j + En,l,s,j,mj , (5.32)

(0)
where En,j is the unperturbed relativistic spectrum and
104 5 Energy Splitting and Shift Due to External Fields

Fig. 5.1 Zeeman effect for the hydrogen atom. Due to a (strong) magnetic field the energy of the
degenerate excited state |2p∧ = |n = 2, l = 1∧ is split into 3 levels and the energy of the degenerate
excited state |3d∧ = |n = 3, l = 2∧ is split into 5 levels. But the spectral lines are always three
(Lorentz triplet)

(1) eB
En,l,s,j,mj = ∇n, l, s, j, mj |L̂z + 2Ŝz |n, l, s, j, mj ∧ (5.33)
2m
is the first-order correction, which is indeed not very easy to calculate. But we can
do it. First we note that

∇n, l, s, j, mj |L̂z + 2Ŝz |n, l, s, j, mj ∧ = ∇n, l, s, j, mj |Ĵz + Ŝz |n, l, s, j, mj ∧

= mj + ∇n, l, s, j, mj |Ŝz |n, l, s, j, mj ∧.
(5.34)

Then, on the basis of the Wigner-Eckart theorem,1 we have

1 The Wigner-Eckart theorem states that for any vector operator V̂ = (V̂1 , V̂2 , V̂3 ) such that
[Ĵi , V̂j ] = iρijk V̂k holds the identity
5.2 Zeeman Effect 105

2 j(j+1)∇n, l, s, j, mj |Ŝz |n, l, s, j, mj ∧ = ∇n, l, s, j, mj |(Ŝ·Ĵ) Ĵz |n, l, s, j, mj ∧, (5.35)

from which we obtain

2 j(j + 1)∇n, l, s, j, mj |Ŝz |n, l, s, j, mj ∧

= mj ∇n, l, s, j, mj |Ŝ · Ĵ|n, l, s, j, mj ∧
1 
= mj ∇n, l, s, j, mj | Ĵ 2 + Ŝ 2 − L̂ 2 |n, l, s, j, mj ∧
2
1 2
= mj  (j(j + 1) + s(s + 1) − l(l + 1)) . (5.36)
2
In conclusion, for a weak magnetic field B the first-order correction is given by

(1)
En,l,s,j,mj = μB B gl,s,j mj , (5.37)

where μB = e/(2m) is the Bohr magneton, and

j(j + 1) + s(s + 1) − l(l + 1)

gl,s,j = 1 + (5.38)
2j(j + 1)

is the so-called Landé factor. Clearly, in the case of the electron s = 1/2 and the
Landé factor becomes
j(j + 1) − l(l + 1) + 3/4
gl,j = 1 + . (5.39)
2j(j + 1)

Strictly speaking, the energy splitting described by Eq. (5.37) is fully reliable only
for a weak magnetic field B in the range 0 T ∈ B ∓ 1/n3 T, with n the principal
quantum number. In fact, if the magnetic field B approaches 1/n3 T one observes
a complex pattern of splitting, which moves by increasing B from the splitting
described by Eq. (5.37) towards the splitting described by Eq. (5.26). This tran-
sition, observed in 1913 by Friedrich Paschen and Ernst Back, is now called the
Paschen-Back effect.

(Footnote 1 continued)

∇n, l, s, j, mj |V̂|n, l, s, j, mj ∧ = ∇n, l, s, j, mj |(V̂ · Ĵ) Ĵ|n, l, s, j, mj ∧.

In our case V̂ = Ŝ and we have considered the z component only.
106 5 Energy Splitting and Shift Due to External Fields

5.3 Solved Problems

Problem 5.1
Determine the Stark splitting for the state |2p∧ of the hydrogen atom in a electric
field of 107 V/m.
Solution
We write the constant electric field as E = E uz , choosing the z axis in the same
direction of E. The Hamiltonian operator is given by

Ĥ = Ĥ0 + Ĥ1 ,

where
p̂2 e2
Ĥ0 = −
2m 4νρ0 r

is the Hamiltonian of the unperturbed system, while

Ĥ1 = e E · r = eEz

is the Hamiltonian of the perturbation due to the electric field, with e electric charge
of the electron. Let |nlml ∧ be the eigenstate of Ĥ0 , such that

Ĥ0 |nlml ∧ = En(0) |nlml ∧

with
13.6
En(0) = − eV,
n2
the Bohr spectrum of the hydrogen atom, and moreover

L̂ 2 |nlml ∧ =  l(l + 1) |nlml ∧, L̂z |nlml ∧ =  ml |nlml ∧.

At the first order of perturbation theory the energetic spectrum is given by

En = En(0) + En(1) ,

(1)
where En is one of the eigenvalues of the submatrix Ĥ1n , whose elements are

(n)
al m ,lm = ∇nl  ml |Ĥ1 |nlml ∧ = eE ∇nl ml |z|nlml ∧.
l l

If n = 2 then l = 0, 1 and ml = −l, −l + 1, ..., l − 1, l, i.e. ml = 0 if l = 0 and

(2)
ml = −1, 0, 1 if l = 1. It follows that the submatrix Ĥ1 is a 4 × 4 matrix, given by
5.3 Solved Problems 107
 (2) (2) (2) (2) 
a00,00 a00,10 a00,11 a00,1−1
 a(2) (2) (2) (2) 
(2)  a10,10 a10,11 a10,1−1 
Ĥ1 =  10,00
(2) (2) (2) (2) .
 a11,00 a11,10 a11,11 a11,1−1 ⎛
(2) (2) (2) (2)
a1−1,00 a1−1,10 a1−1,11 a1−1,1−1

Note that

∇nl ml |z|nlml ∧ = ∇nl ml |z|nlml ∧ δm ,ml δl+l ,1 ,

l

i.e. the selection rules l = ±1 and ml = 0 hold. It follows that many elements
of the submatrix Ĥ1(2) are zero:
 (2) 
0 a00,10 0 0
(2)  a(2) 0
Ĥ1 = 
 0
10,00 0 0 ,
0 0 0⎛
0 0 0 0

(2) (2) (2)

and also a00,10 = a10,00 . One finds immediately that 2 eigenvalues of Ĥ1 are zero.
To determine the other 2 eigenvalues it is sufficient to study the 2 × 2 given by
⎜ ⎝
(2)
(2) 0 a00,10
Ĥ1 = (2) ,
a00,10 0

whose eigenvalues are

(2)
ε1,2 = ±a00,10 = ±e E ∇200|z|210∧.

We must now calculate ∇200|z|210∧. By using the completeness relation

⎞
1= d 3 r|r∧∇r|,

this matrix element can be written as

⎞ ⎞
∇200|z|210∧ = ∇200| d r|r∧∇r|z|210∧ = d 3 r∇200|r∧z∇r|210∧
3

⎞
↑ ↑
= d 3 rλ200 (r)zλ210 (r).

For the hydrogen atom one has

108 5 Energy Splitting and Shift Due to External Fields

1 r −r/(2r0 )
λ200 (r) = ↓ 3/2
(1 − )e ,
2 2νr0 2r0
1 r −r/(2r0 )
λ210 (r) = ↓ 3/2 r
e cos θ,
4 2νr 0
0

where r0 = 0.53 × 10−10 m is the Bohr radius. Remembering that in spherical

coordinates z = r cos θ, we get

⎞ ⎞∀ ⎞2ν ⎞ν
↑ ↑ ↑
3
d r λ200 (r)zλ210 (r) = drr 2
dπ dθ sin θ λ200 (r) r cos θ λ210 (r)
0 0 0
⎞∀ ⎞2ν ⎞1
1 −r/r0 r
= dr e r (1 −
4
) dπ d(cos θ) cos2 θ
16νr04 2r0
0 0 −1
⎞∀
1 r
= dr e−r/r0 r 4 (1 − )
12r04 2r0
0
1
= (−36r05 ) = −3r0 ,
12r04

namely
∇200|z|210∧ = −3r0 .

(1)
Finally, the possible values of the perturbing energy E2 read
⎟
⎠ 3eEr0
(1)
E2 = 0 .
⎡
−3eEr0

(0) (0)
It follows that the unperturbed energy E2 splits into 3 levels, given by E2 = E2 +
E2(1) , but the central one of them coincides with the unperturbed level. In conclusion,
the Stark splitting is

V
E2 = ±3eEr0 = ±3e × 107 × 0.53 × 10−10 m = ±1.6 × 10−3 eV.
m
Problem 5.2
Calculate the Stark shift of the state |1s∧ of the hydrogen atom in a electric field of
107 V/m.
Solution
We write the constant electric field as E = E ez , choosing the z axis in the same
direction of E. The Hamiltonian operator is given by
5.3 Solved Problems 109

Ĥ = Ĥ0 + Ĥ1 ,

where
p̂2 e2
Ĥ0 = −
2m 4νρ0 r

is the Hamiltonian of the unperturbed system, while

Ĥ1 = e E · r = eEz

is the Hamiltonian of the perturbation due to the electric field, with e electric charge
of the electron. At first order of perturbation theory the energy of |1s∧ = |n = 1, l =
0, ml = 0∧ is given by
(0) (1)
E1 = E1 + E1 ,

where
(0)
E1 = −13.6 eV,

(1)
while E1 is
(1)
E1 = ∇100|Ĥ1 |100∧ = e E ∇100|z|100∧ = 0.

This matrix element is zero due to the selection rules l = ±1 e ml = 0. Thus,
we need the second order perturbation theory, namely

(0) (1) (2)

E1 = E1 + E1 + E1 ,

where

 ∀

|∇nlml |eEz|100∧|2 |∇n10|eEz|100∧|2
E1(2) = = ,
nlml →=100 E1(0) − En(0) n=2 E1(0) − En(0)

by using again the selection rules. Remembering that

(0)
E1
En(0) = ,
n2
we can write
∀
(2) e2 E 2  n2
E1 = (0)
|∇n10|z|100∧|2 .
E1 (n2 − 1)
n=2

In addition, knowing that in spherical coordinates z = r cos θ, the matrix element

∇n10|z|100∧ reads
110 5 Energy Splitting and Shift Due to External Fields
⎞
↑
∇n10|z|100∧ = d 3 rλn1o (r)zλ100 (r)

⎞∀ ⎞2ν ⎞
= drr Rn1 (r)R10 (r)
3
dπ dθ sin θ cos θY10 (θ, π)Y00 (θ, π).
0 0

Because ⎣
1 3
Y00 (θ, π) = ↓ Y10 (θ, π) = cos θ,
4ν 4ν

we get
⎞∀
1 1
∇n10|z|100∧ = ↓ drr 3 Rn1 (r)R10 (r) = ↓ r0 f (n),
3 3
0

where r0 = 0.53 × 10−10 m is the Bohr radius and f (n) is a function of the principal
quantum number n. It is possible to show that
∀
 n2 27
f (n)2 = .
n2 − 1 8
n=2

From this exact relation we obtain

(2) e2 E 2 r02 27 9 e2 E 2 r02

E1 = (0) 8
= .
3E 8 E (0)
1 1

Inserting the numerical values we finally get

(2)
E1 = −2.3 × 10−15 eV.

Problem 5.3
Determine the Zeeman splitting for the states |1s∧ e |2p∧ of the hydrogen atom in a
magnetic field of 10 T.
Solution
In presence of a magnetic field with intensity B between 1/n3 and 106 /n4 T, one has
the strong-field Zeeman effect and the energy spectrum is given by

En,m,ms = En + μB B(ml + 2ms ),

where
13.6 eV
En = −
n2
5.3 Solved Problems 111

is the Bohr spectrum, ml is the quantum number of the third component of the orbital
angular momentum L and ms is the quantum number of the third component of the
spin S. In addition μB = e/(2m) = 9.3 × 10−24 J/T is the Bohr magneton. In our
problem we are in the regime of validity of the strong-field Zeeman effect and

J 1019
μB B = 9.3 × 10−24 × 10 T = 9.3 × 10−23 J = 9.3 × 10−23 × eV
T 1.6
= 5.8 × 10−4 eV.

For the state |1s∧ = |n = 1, l = 0, ml = 0∧ one has s = 1

2 and consequently
ms = − 21 , 21 . It follows that

E1,0,− 1 = −μB B = −5.8 × 10−4 eV

2

E1,0, 1 = μB B = 5.8 × 10−4 eV.

2

For the state |2p∧ = |n = 2, l = 1, ml = −1, 0, 1∧ one has s = 1

2 and then
ms = − 21 , 21 . It follows that

E2,−1,− 1 = −2μB B = −11.6 × 10−4 eV

2

E2,0,− 1 = −μB B = −5.8 × 10−4 eV

2

E2,1,− 1 = E2,−1, 1 = 0
2 2

E2,0, 1 = μB B = 5.8 × 10−4 eV

2

E2,1, 1 = 2μB B = 11.6 × 10−4 eV.

2

Problem 5.4
Calculate the Zeeman splitting for the state |2p∧ of the hydrogen atom in a magnetic
field of 5 Gauss.
Solution
The magnetic field B = 5 Gauss = 5 × 10−4 T is quite weak (the magnetic field
of the Earth is about 0.5 Gauss). With a field B between 0 and 10−2 T, the energy
spectrum is well described by

En,l,j,mj = En + μB B gl,j mj ,

where μB = e/(2m) = 9.3 × 10−24 J/T is the Bohr magneton, mj is the quantum
number of the third component of the total angular momentum J = L + S, and gl,j
is the Landé factor, given by

j(j + 1) − l(l + 1) + 3/4

gl,j = 1 + .
2j(j + 1)
112 5 Energy Splitting and Shift Due to External Fields

In this problem the magnetic field is very weak, and one has the weak-field Zeeman
effect. For the state |2p∧ = |n = 2, l = 1, ml = −1, 0, 1∧ one has s = 21 and then
ms = − 21 , 21 . In addition, j = l ± 21 = 21 , 23 from which mj = − 23 , − 21 , 21 , 23 . In the
case j = 1/2 the Landé factor reads

3/4 − 2 + 3/4 2
g1, 1 = 1 + = 1 − 1/3 = .
2 3/2 3

In the case j = 3/2 the Landè factor is instead given by

15/4 − 2 + 3/4 1 4
g1, 3 = 1 + =1+ = .
2 15/2 3 3

In our problem

J
μB B = 9.3 × 10−24 × 5 × 10−4 T = 4.6 × 10−27 J
T
1019
= 4.6 × 10−27 × eV = 2.9 × 10−8 eV.
1.6

We can now calculate the Zeeman splittings. For j = 1

2 one has mj = − 21 , 21 and
then
1
E1, 1 ,− 1 = − μB B = −0.97 × 10−8 eV
2 2 3
1
E1, 1 , 1 = μB B = 0.97 × 10−8 eV.
2 2 3

For j = 3
2 one has instead mj = − 23 , − 21 , 21 , 23 and then

E1, 3 ,− 3 = −2μB B = −5.8 × 10−8 eV

2 2
2
E1, 3 ,− 1 = − μB B = −1.93 × 10−8 eV
2 2 3
2
E1, 3 , 1 = μB B = 1.93 × 10−8 eV
2 2 3
E1, 3 , 3 = 2μB B = 5.8 × 10−8 eV.
2 2
5.3 Solved Problems 113

Further Reading

For the Stark effect:

R.W. Robinett, Quantum Mechanics: Classical Results, Modern Systems, and Visu-
alized Examples, Chap. 19, Sect. 19.6 (Oxford University Press, Oxford, 2006)
For the Zeeman effect:
B.H. Bransden, C.J. Joachain, Physics of Atoms and Molecules, Chap. 6, Sects. 6.1
and 6.2 (Prentice Hall, Upper Saddle River, 2003)
Chapter 6
Many-Body Systems

In this chapter we want to analyze atoms with many electrons and, more generally,
systems with many interacting identical particles. We first consider the general prop-
erties of many identical particles with their bosonic or fermionic many-body wave-
functions, and the connection between spin and statistics which explains the Pauli
principle and the main features of the periodic table of elements. We then discuss
the Hartree-Fock approximation and also the density functional theory, historically
introduced by Thomas and Fermi to model the electronic cloud of atoms and recently
applied also to atomic Bose-Einstein condensates. Finally, we illustrate the Born-
Oppenheimer approximation, which is very useful to treat molecules composed by
many electrons in interaction with several nuclei.

6.1 Identical Quantum Particles

First of all, we introduce the generalized coordinate x = (r, ν) of a particle which

takes into account the spatial coordinate r but also the intrinsic spin ν pertaining to
the particle. For instance, a spin 1/2 particle has ν = −1/2, 1/2 =∧, ↑. By using
the Dirac notation the corresponding single-particle state is

|x∇ = |r ν∇. (6.1)

We now consider N identical particles; for instance particles with the same mass and
electric charge. The many-body wavefunction of the system is given by

(x1 , x2 , . . . , x N ) = (r1 , ν1 , r2 , ν2 , . . . , r N , ν N ) , (6.2)

According to quantum mechanics identical particles are indistinguishible. As a con-

sequence, it must be

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 115
DOI: 10.1007/978-3-319-05179-6_6, © Springer International Publishing Switzerland 2014
116 6 Many-Body Systems

|(x1 , x2 , . . . , xi , . . . , x j , . . . , x N )|2 = |(x1 , x2 , . . . , x j , . . . , xi , . . . , x N )|2 ,

(6.3)

which means that the probability of finding the particles must be independent on the
exchange of two generalized coordinates xi and x j . Obviously, for 2 particles this
implies that
|(x1 , x2 )|2 = |(x2 , x1 )|2 . (6.4)

Experiments suggests that there are only two kind of identical particles which satisfy
Eq. (6.3): bosons and fermions. For N identical bosons one has

(x1 , x2 , . . . , xi , . . . , x j , . . . , x N ) = (x1 , x2 , . . . , x j , . . . , xi , . . . , x N ) , (6.5)

i.e. the many-body wavefunction is symmetric with respect to the exchange of two
coordinates xi and x j . Note that for 2 identical bosonic particles this implies

(x1 , x2 ) = (x2 , x1 ). (6.6)

For N identical fermions one has instead

(x1 , x2 , . . . , xi , . . . , x j , . . . , x N ) = −(x1 , x2 , . . . , x j , . . . , xi , . . . , x N ) ,
(6.7)

i.e. the many-body wavefunction is anti-symmetric with respect to the exchange of

two coordinates xi and x j . Note that for 2 identical fermionic particles this implies

(x1 , x2 ) = −(x2 , x1 ). (6.8)

An immediate consequence of the anti-symmetry of the fermionic many-body wave-

function is the Pauli Principle: if xi = x j then the many-body wavefunction is zero.
In other words: the probability of finding two fermionic particles with the same
generalized coordinates is zero.
A remarkable experimental fact, which is often called spin-statistics theorem
because can be deduced from other postulates of relativistic quantum field theory, is
the following: identical particles with integer spin are bosons while identical particles
with semi-integer spin are fermions. For instance, photons are bosons with spin 1
while electrons are fermions with spin 1/2. Notice that for a composed particle it
is the total spin which determines the statistics. For example, the total spin (sum of
nuclear and electronic spins) of 4 He atom is 0 and consequently this atom is a boson,
while the total spin of 3 He atom is 1/2 and consequently this atom is a fermion.
6.2 Non-interacting Identical Particles 117

6.2 Non-interacting Identical Particles

The quantum Hamiltonian of N identical non-interacting particles is given by


N
Ĥ0 = ĥ(xi ) , (6.9)
i=1

where ĥ(x) is the single-particle Hamiltonian. Usually the single-particle Hamil-

tonian is given by
2 2
ĥ(x) = − → + U (r) , (6.10)
2m

with U (r) the external confining potential. In general the single-particle Hamiltonian
ĥ satisfies the eigenvalue equation

ĥ(x) ρn (x) = πn ρn (x) , (6.11)

where πn are the single-particle eigenenergies and ρn (x) the single-particle eigen-
functions, with n = 1, 2, . . ..
The many-body wavefunction (x1 , x2 , . . . , x N ) of the system can be written in
terms of the single-particle wavefunctions ρn (x) but one must take into account the
spin-statistics of the identical particles. For N bosons the simplest many-body wave
function reads

(x1 , x2 , . . . , x N ) = ρ1 (x1 ) ρ1 (x2 ) . . . ρ1 (x N ) , (6.12)

which corresponds to the configuration where all the particles are in the lowest-
energy single-particle state ρ1 (x). This is indeed a pure Bose-Einstein condensate.
Note that for 2 bosons the previous expression becomes

(x1 , x2 ) = ρ1 (x1 )ρ1 (x2 ). (6.13)

Obviously there are infinite configuration which satisfy the bosonic symmetry of the
many-body wavefunction. For example, with 2 bosons one can have

(x1 , x2 ) = ρ4 (x1 )ρ4 (x2 ) , (6.14)

which means that the two bosons are both in the fourth eigenstate; another example
is
1
(x1 , x2 ) = ∀ (ρ1 (x1 )ρ2 (x2 ) + ρ1 (x2 )ρ2 (x1 )) , (6.15)
2
118 6 Many-Body Systems

Fig. 6.1 Ground-state of a system of identical non-interacting bosons (a) and fermions (b) is a
harmonic trap

∀
where the factor 1/ 2 has been included to maintain the same normalization of the
many-body wavefunction, and in this case the bosons are in the first two available
single-particles eigenstates.
For N fermions the simplest many-body wave function is instead very different,
and it is given by
 
ρ1 (x1 ) ρ1 (x2 ) . . . ρ1 (x N )
1   ρ2 (x1 ) ρ2 (x2 ) . . . ρ2 (x N ) 

(x1 , x2 , . . . , x N ) = ∀  (6.16)
N! ... ... ... ... 
ρ N (x1 ) ρ N (x2 ) . . . ρ N (x N )

that is the so-called Slater determinant of the N × N matrix obtained with the N
lowest-energy single particle wavefunctions βn (x), with n = 1, 2, . . . , N , calculated
in the N possible generalized coordinates xi , with i = 1, 2, . . . , N . Note that for 2
fermions the previous expression becomes

1
(x1 , x2 ) = ∀ (ρ1 (x1 )ρ2 (x2 ) − ρ1 (x2 )ρ2 (x1 )) . (6.17)
2

We stress that for non-interacting identical particles the Hamiltonian (6.9) is sep-
arable and the total energy associated to the bosonic many-body wavefunction (6.12)
is simply
E = N π1 , (6.18)

while for the fermionic many-body wavefunction (6.16) the total energy (in the
absence of degenerate single-particle energy levels) reads

E = π1 + π2 + · · · + π N , (6.19)
6.2 Non-interacting Identical Particles 119

which is surely higher than the bosonic one. The highest occupied single-particle
energy level is called Fermi energy, and it indicated as π F ; in our case it is obviously
πF = πN .
Notice the above definition of Fermi energy π F can be considered acceptable
in many contexts and is a simple and effective definition, especially for the zero-
temperature Fermi gas. Unfortunately it is imprecise in some contexts, e.g. finite
systems with a nonzero gap between the last full and first empty state. A better
definition, such as the limit for T √ 0 of the chemical potential μ (see Sect. 7.5)
would lead to some intermediate value between the energy of the highest occupied
state and that of the lowest empty state.

6.2.1 Uniform Gas of Non-interacting Fermions

A quite important physical system is the uniform gas of non-interacting fermions. It

is indeed a good starting point for the description of all the real systems which have
a finite interaction between fermions.
The non-interacting uniform Fermi gas is obtained setting to zero the confining
potential, i.e.
U (r) = 0 , (6.20)

and imposing periodicity conditions on the single-particle wavefunctions, which are

plane waves with a spinor
1
ρ(x) = ∀ eik·r χν , (6.21)
V

where χν is the spinor for spin-up and spin-down along a chosen z asis:

1 0
χ↑ = , χ∧ = . (6.22)
0 1

At the boundaries of a cube having volume V and side L one has

eik x (x+L) = eik x x , eik y (y+L) = eik y y , eikz (z+L) = eikz z . (6.23)

It follows that the linear momentum k can only take on the values

2ε 2ε 2ε
kx = nx , ky = n y , kz = nz , (6.24)
L L L
where n x , n y , n z are integer quantum numbers. The single-particle energies are given
by
2 k 2 2 4ε 2 2
πk = = (n + n 2y + n 2z ). (6.25)
2m 2m L 2 x
120 6 Many-Body Systems

In the thermodynamic limit L √ ⊗, the allowed values are closely spaced and one
can use the continuum approximation
 
√ dn x dn y dn z , (6.26)
n x ,n y ,n z

which implies
  
L3 d 3k
√ d k=V
3
. (6.27)
(2ε)3 (2ε)3
k

The total number N of fermionic particles is given by


N=  (π F − πk ) , (6.28)
ν k

where the Heaviside step function (x), such that (x) = 0 for x < 0 and (x) = 1
for x > 0, takes into account the fact that fermions are occupied only up to the
Fermi energy π F , which is determined by fixing N . Notice that at finite temperature
T the total number N of ideal fermions is instead obtained from the Fermi-Dirac
distribution, namely
 1
N= λ(π −μ)
, (6.29)
ν
e k +1
k

where λ = 1/(k B T ), with k B the Boltzmann constant, and μ is the chemical potential
of the system. In the limit λ √ +⊗, i.e. for T √ 0, the Fermi-Dirac distribution
becomes the Heaviside step function and μ is identified as the Fermi energy π F .
In the continuum limit and choosing spin 1/2 fermions one finds

d 3k 2 k 2
N= V  πF − , (6.30)
(2ε)3 2m
ν=↑,∧

from which one gets (the sum of spins gives simply a factor 2) the uniform density

3/2
N 1 2mπ F
θ= = . (6.31)
V 3ε 2 2

The formula can be inverted giving the Fermi energy π F as a function of the density
θ, namely
2 ⎛ 2 ⎜2/3
πF = 3ε θ . (6.32)
2m
In many applications the Fermi energy π F is written as
6.2 Non-interacting Identical Particles 121

2 k 2F
πF = , (6.33)
2m
where k F is the so-called Fermi wave-number, given by
⎛ ⎜1/3
k F = 3ε 2 θ . (6.34)

The total energy E of the uniform and non-interacting Fermi system is given by

E= πk  (π F − πk ) , (6.35)
ν k

and using again the continuum limit with spin 1/2 fermions it becomes

d 3 k 2 k 2 2 k 2
E= V  π F − , (6.36)
(2ε)3 2m 2m
ν=↑,∧

from which one gets the energy density

E 3 3 2 ⎛ 2 ⎜2/3 5/3
E= = θ πF = 3ε θ (6.37)
V 5 5 2m
in terms of the Fermi energy π F and the uniform density θ.

6.2.2 Atomic Shell Structure and the Periodic

Table of the Elements

The non-relativistic quantum Hamiltonian of Z identical non-interacting electrons

in the neutral atom is given by


Z
Ĥ0 = ĥ(ri ), (6.38)
i=1

where ĥ(r) is the single-particle Hamiltonian given by

2 2
ĥ(r) = − → + U (r), (6.39)
2m
with
Z e2
U (r) = − (6.40)
4ε∂0 |r|
122 6 Many-Body Systems

Table 6.1 Lightest atoms and their ground-state energy E on the basis of single-particle energies
Z Atom Symbol E
1 hydrogen H π1 (1)
2 helium He 2π1 (2)
3 lihium Li 2π1 (3) + π2 (3)
4 berylium Be 2π1 (4) + 2π2 (4)
5 boron B 2π1 (5) + 3π2 (5)
6 carbon C 2π1 (6) + 4π2 (6)
7 nitrogen N 2π1 (7) + 5π2 (7)
8 oxygen O 2π1 (8) + 6π2 (8)
9 fluorine F 2π1 (9) + 7π2 (9)
10 neon Ne 2π1 (10) + 8π2 (10)
11 sodium Na 2π1 (11) + 8π2 (11) + π3 (11)
12 magnesium Mg 2π1 (12) + 8π2 (12) + 2π3 (12)
13 aluminium Al 2π1 (13) + 8π2 (13) + 3π3 (13)
14 silicon Si 2π1 (14) + 8π2 (14) + 4π3 (14)

the confining potential due to the attractive Coulomb interaction between the single
electron and the atomic nucleus of positive charge Z e, with e > 0.
Because the confining potential U (r) is spherically symmetric, i.e. U (r) = U (|r|),
the single-particle Hamiltonian ĥ satisfies the eigenvalue equation

ĥ(r) ρnlm l m s (r) = πn (Z ) ρnlm l m s (r), (6.41)

where
Z2
πn (Z ) = −13.6 eV (6.42)
n2
are the Bohr single-particle eigenenergies of the hydrogen-like atom, and
ρnlm l m s (r) = Rnl (r )Ylm l (ψ, ρ) are the single-particle eigenfunctions, which de-
pends on the principal quantum numbers n = 1, 2, . . ., the angular quantum num-
ber l = 0, 1, . . . , n − 1, the third-component angular quantum number m l =
−l, −l +1, . . . , l −1, l, and the third-component spin quantum number m s = − 21 , 21 .
Notice that here
ρnlm l m s (r) = ρnlm l (r, ν) (6.43)

with ν =↑ for m s = 21 and ν =∧ for m s = − 21 (Table 6.1).

Due to the Pauli principle the ground-state energy E of this system of Z electrons
strongly depends of the degeneracy of single-particle energy levels. In the Table we
report the ground-state energy E of the lightest atoms on the basis of their single-
particle energy levels πn (Z ).
The degeneracy of the single-particle energy level πn (Z ) is clearly independent
on Z and given by
6.2 Non-interacting Identical Particles 123


n−1
deg(πn (Z )) = 2(2l + 1) = 2n 2 , (6.44)
l=0

which is the maximum number of electrons with the same principal quantum number
n. The set of states with the same principal quantum number is called theoretical shell.
The number of electrons in each theoretical shell are: 2, 8, 18, 32, 52. One expects
that the more stable atoms are characterized by fully occupied theoretical shells.
Actually, the experimental data, namely the periodic table of elements due to Dmitri
Mendeleev, suggest that the true number of electrons in each experimental shell are
instead: 2, 8, 8, 18, 18, 32, because the noble atoms are characterized the following
atomic numbers:

2, 2 + 8 = 10, 2 + 8 + 8 = 18, 2 + 8 + 8 + 18 = 36, 2 + 8 + 8 + 18 + 18 = 54,

2 + 8 + 8 + 18 + 18 + 32 = 86 ,

corresponding to Helium (Z = 2), Neon (Z = 10), Argon (Z = 18), Krypton

(Z = 36), Xenon (Z = 54), and Radon (Z = 86). The experimental sequence is
clearly similar but not equal to the theoretical one, due to repetitions of 8 and 18.
It is important to stress that the theoretical sequence is obtained under the very
crude assumption of non-interacting electrons. To improve the agreement between
theory and experiment one must include the interaction between the electrons.

6.3 Interacting Identical Particles

The quantum Hamiltonian of N identical interacting particles is given by


N
1 
N
Ĥ = ĥ(xi ) + V (xi , x j ) = Ĥ0 + Ĥ I , (6.45)
2
i=1 i, j=1
i∞= j

where ĥ is the single-particle Hamiltonian and V (xi , x j ) is the inter-particle potential

of the mutual interaction. In general, due to the inter-particle potential, the Hamil-
tonian (6.45) is not separable and the many-body wavefunctions given by Eqs. (6.12)
and (6.16) are not exact eigenfunctions of Ĥ .

6.3.1 Variational Principle

Many approaches to the determination of the ground-state of an interacting many-

body system are based on the so-called variational principle, which is actually a
theorem.
124 6 Many-Body Systems

Theorem 1 For any normalized many-body state |∇, i.e. such that ∈|∇ = 1,
which belongs to the Hilbert space on which acts the Hamiltonian Ĥ , one finds

∈| Ĥ |∇ ∓ E gs , (6.46)

where E gs is the ground-state energy of the system and the equality holds only if |∇ =
|gs ∇ with |gs ∇ ground-state of the system, i.e. such that Ĥ |gs ∇ = E gs |gs ∇.

Proof The many-body Hamiltonian Ĥ satisfies the exact eigenvalue problem

Ĥ |φ ∇ = E φ |φ ∇, (6.47)

where E φ are the ordered eigenvalues, i.e. such that E 0 < E 1 < E 2 < · · · with
E 0 = E gs , and |φ ∇ the corresponding orthonormalized eigenstates, i.e. such that
∈φ |λ ∇ = αφ,λ with |0 ∇ = |gs ∇. The generic many-body state |∇ can be written
as 
|∇ = cφ |φ ∇, (6.48)
φ

where cφ are the complex coefficients of the expansion such that


|cφ |2 = 1. (6.49)
φ

Then one finds

  
∈| Ĥ |∇ = cφ↑ cλ ∈φ | Ĥ |λ ∇ = cφ↑ cλ E λ ∈φ |λ ∇ = cφ↑ cλ E λ αφ,λ
φ,λ φ,λ φ,λ
 
= |cφ | E φ ∓
2
|cφ | E 0 = E 0 = E gs .
2
(6.50)
φ φ

Obviously, the equality holds only if c0 = 1 and, consequently, all the other coeffi-
cients are zero.
In 1927 by Douglas Hartree and Vladimir Fock used the variational principle to
develop a powerful method for the study of interacting identical particles. We shall
analyze this variational method in the following subsections.

6.3.2 Hartree for Bosons

In the case of N identical interacting bosons the Hartree approximation is simply

given by
(x1 , x2 , . . . , x N ) = ρ(x1 ) ρ(x2 ) . . . ρ(x N ), (6.51)
6.3 Interacting Identical Particles 125

where the single-particle wavefunction ρ(x) is unknown and it must be determined

in a self-consistent way. Notice that, as previously stressed, this factorization implies
that all particles belong to the same single-particle state, i.e. we are supposing that
the interacting system is a pure Bose-Einstein condensate. This is a quite strong
assumption, that is however reliable in the description of ultracold and dilute gases
made of bosonic alkali-metal atoms (in 2001 Eric Cornell, Carl Weiman, and Wolfang
Ketterle got the Nobel Prize in Physics for their experiments with these quantum
gases), and which must be relaxed in the case of strongly-interacting bosonic systems
(like superfluid 4 He). In the variational spirit of the Hartree approach the unknown
wavefunction ρ(x) is determined by minimizing the expectation value of the total
Hamiltonian, given by

∈| Ĥ |∇ = d x1 d x2 . . . d x N  ↑ (x1 , x2 , . . . , x N ) Ĥ (x1 , x2 , . . . , x N ), (6.52)

with respect to ρ(x). In fact, by using Eq. (6.45) one finds immediately
 
1
∈| Ĥ |∇ = N d x ρ↑ (x)ĥ(x)ρ(x) + N (N − 1) d x d x ↓ |ρ(x)|2 V (x, x ↓ )|ρ(x ↓ )|2 ,
2
(6.53)

which is a nonlinear energy functional of the single-particle wavefunction ρ(x). It

is called single-orbital Hartree functional for bosons. In this functional the first term
is related to the single-particle hamiltonian ĥ(x) while the second term is related to
the inter-particle interaction potential V (x, x ↓ ). We minimize this functional with the
following constraint due to the normalization

d x |ρ(x)|2 = 1. (6.54)

We get immediately the so-called Hartree equation for bosons

⎝ ⎞
ĥ(x) + Um f (x) ρ(x) = π ρ(x), (6.55)

where the mean-field potential Um f (x) reads


Um f (x) = (N − 1) d x ↓ V (x, x ↓ ) |ρ(x ↓ )|2 (6.56)

and π is the Lagrange multiplier fixed by the normalization. It is important to observe

that the mean-field potential Um f (x) depends on ρ(x) and it must be obtained self-
consistently. In other words, the Hartree equation of bosons is a integro-differential
nonlinear Schrödinger equation whose nonlinear term gives the mean-field potential
of the system.
In the case of spinless bosons, where |x∇ = |r∇, given the local bosonic density
126 6 Many-Body Systems

θ(r) = N |ρ(r)|2 , (6.57)

under the assumption of a large number N of particles the Hartree variational energy
reads
 ⎟ ⎠ 
2 2 1
∈| Ĥ |∇ = N d 3 r ρ↑ (r) − → + U (r) ρ(r) + d 3 r d 3 r↓ θ(r)V (r − r↓ )θ(r↓ )
2m 2
(6.58)
while the Hartree equation becomes
⎟  ⎠
2 2 3 ↓ ↓ ↓
− → + U (r) + d r V (r − r ) θ(r ) ρ(r) = π ρ(r). (6.59)
2m

To conclude this subsection, we observe that in the case of a contact inter-particle

potential, i.e.
V (r − r↓ ) = g α(r − r↓ ), (6.60)

the previous Hartree variational energy becomes

 ⎟ ⎠ 
↑ 2 2 g
∈| Ĥ |∇ = N d r ρ (r) −
3
→ + U (r) ρ(r) + d 3 r θ(r)2 (6.61)
2m 2

and the corresponding Hartee equation reads

⎟ ⎠
2 2
− → + U (r) + g θ(r) ρ(r) = π ρ(r), (6.62)
2m

which is the so-called Gross-Pitaevskii equation, deduced in 1961 by Eugene Gross

and Lev Pitaevskii.

6.3.3 Hartree-Fock for Fermions

In the case of N identical interacting fermions, the approximation developed by

Hartree and Fock is based on the Slater determinant we have seen previously, namely
 
ρ1 (x1 ) ρ1 (x2 ) . . . ρ1 (x N )
1   ρ2 (x1 ) ρ2 (x2 ) . . . ρ2 (x N )  ,

(x1 , x2 , . . . , x N ) = ∀  (6.63)
N! ... ... ... ... 
ρ N (x1 ) ρ N (x2 ) . . . ρ N (x N )

where now the single-particle wavefunctions ρn (x) are unknown and they are de-
termined with a variational procedure. In fact, in the Hartree-Fock approach the
unknown wavefunctions ρn (x) are obtained by minimizing the expectation value of
6.3 Interacting Identical Particles 127

the total Hamiltonian, given by


∈| Ĥ |∇ = d x1 d x2 . . . d x N  ↑ (x1 , x2 , . . . , x N ) Ĥ (x1 , x2 , . . . , x N ), (6.64)

with respect to the N single-particle wavefunctions ρn (x). By using Eq. (6.45) and
after some tedious calculations one finds

N  
 1 ⎝
N
∈| Ĥ |∇ = d x ρi↑ (x)ĥ(x)ρi (x) + d x d x ↓ |ρi (x)|2 V (x, x ↓ )|ρ j (x ↓ )|2
2
i=1 i, j=1
i∞= j
 ⎞
− d x d x ↓ ρi↑ (x)ρ j (x)V (x, x ↓ )ρ↑j (x ↓ )ρi (x ↓ ) , (6.65)

which is a nonlinear energy functional of the N single-particle wavefunctions ρi (x).

In this functional the first term is related to the single-particle Hamiltonian ĥ(x) while
the second and the third terms are related to the inter-particle interaction potential
V (x, x ↓ ). The second term is called direct term of interaction and the third term is
called exchange term of interaction. We minimize this functional with the following
constraints due to the normalization

d x |ρi (x)|2 = 1 , i = 1, 2, . . . , N , (6.66)

where, in the case of spin 1/2 particles, one has

ρi (x) = ρi (r, ν) = ρ̃i (r, ν) χν (6.67)

with χν the two-component spinor, and the integration over x means

  
dx = d 3r , (6.68)
ν=↑,∧

such that
    
d x |ρi (x)|2 = d 3 r |ρi (r, ν)|2 = d 3 r |ρ̃i (r, ν)|2 , (6.69)
ν=↑,∧ ν=↑,∧

because χ↑ν χν = 1 and more generally χ↑ν χν↓ = αν,ν↓ .

After minimization of the energy functional we get the so-called Hartree-Fock
equations ⎝ ⎞
ĥ(x) + Ûm f (x) ρi (x) = πi ρi (x) (6.70)
128 6 Many-Body Systems

where πi are the Lagrange multipliers fixed by the normalization and Ûm f is a nonlocal
mean-field operator. This nonlocal operator is given by


N
ji
Ûm f (x) ρi (x) = Ud (x) ρi (x) − Ux (x)ρ j (x), (6.71)
j=1

where the direct mean-field potential Ud (x) reads

N 

Ud (x) = d x ↓ V (x, x ↓ ) |ρ j (x ↓ )|2 , (6.72)
j=1
j∞=i

ji
while the exchange mean-field potential Ux (x) is instead

d x ↓ ρ↑j (x ↓ ) V (x, x ↓ ) ρi (x ↓ ).
ji
Ux (x) = (6.73)

If one neglects the exchange term, as done by Hartree in his original derivation, the
so-called Hartree equations
⎝ ⎞
ĥ(x) + Ud (x) ρi (x) = πi ρi (x), (6.74)

are immediately derived. It is clearly much simpler to solve the Hartree equations
instead of the Hartree-Fock ones. For this reason, in many applications the latter are
often used. In the case of spin 1/2 fermions, given the local fermionic density

 
N 
θ(r) = |ρi (r, ν)|2 = θ(r, ν), (6.75)
ν=↑,∧ i=1 ν=↑,∧

under the assumption of a large number N of particles the Hartree (direct) variational
energy reads

N  
 ⎟ ⎠
2 2
ED = d 3 r ρi↑ (r, ν) − → + U (r) ρi (r, ν)
2m
i=1 ν=↑,∧

1 
+ d 3 r d 3 r↓ θ(r, ν)V (r − r↓ )θ(r↓ , ν ↓ ) (6.76)
2 ↓
ν,ν =↑,∧

and the corresponding Hartree equation becomes

6.3 Interacting Identical Particles 129
⎡ ⎤
 2  
⎣− → 2 + U (r) + d 3 r↓ V (r − r↓ ) θ(r↓ , ν ↓ )⎦ ρi (r, ν) = πi ρi (r, ν).
2m ↓ ν =↑,∧
(6.77)
The Hartree-Fock variational energy is slightly more complex because it includes
also the exchange energy, given by


1  
N
EX = − d 3 r d 3 r↓ ρi↑ (r, ν)ρi (r↓ , ν)V (r − r↓ )ρ↑j (r↓ , ν)ρ j (r, ν).
2
i, j=1 ν=↑,∧
i∞= j
(6.78)
Notice that in the exchange energy all the terms with opposite spins are zero due to the
scalar product of spinors: χ↑ν χν↓ = αν,ν↓ . The existence of this exchange energy E X is
a direct consequence of the anti-symmetry of the many-body wave function, namely a
consequence of the fermionic nature of the particles we are considering. Historically,
this term E X was obtained by Vladimir Fock to correct the first derivation of Douglas
Hartree who used a not anti-symmetrized many-body wavefunction.
To conclude this subsection, we observe that in the case of a contact inter-particle
potential, i.e.
V (r − r↓ ) = g α(r − r↓ ), (6.79)

the Hartee-Fock (direct plus exchange) variational energy reads

N  
 ⎟ ⎠
2 2
E= 3
d r ρi↑ (r, ν) − → + U (r) ρi (r, ν) (6.80)
2m
i=1 ν=↑,∧
 ⎝ ⎞
g  
N
+ d 3r |ρi (r, ν)|2 |ρ j (r, ν ↓ )|2 − |ρi (r, ν)|2 |ρ j (r, ν)|2 αν, ν ↓ .
2
i, j=1 ν,ν ↓ =↑,∧
i∞= j

It follows immediately that identical spin-polarized fermions with contact interaction

are effectively non-interacting because in this case the interaction terms of direct and
exchange energy exactly compensate to zero.

6.3.4 Mean-Field Approximation

A common strategy to solve the interacting many-body problem is the so-called

mean-field approximation, which corresponds on finding a suitable mean-field po-
tential which makes the system quasi-separable. The idea is the following. The in-
teraction Hamiltonian
1 
N
Ĥ I = V (xi , x j ) (6.81)
2
i, j=1
i∞= j
130 6 Many-Body Systems

can be formally written as

⎛ ⎜
Ĥ I = Ĥm f + Ĥ I − Ĥm f = Ĥm f + Hr es , (6.82)

where Ĥm f is an appropriate separable mean-field Hamiltonian, i.e. such that


N
Ĥm f = Um f (xi ), (6.83)
i=1

and Ĥr es is the residual Hamiltonian. In this way the total Hamiltonian reads

Ĥ = Ĥ0 + Ĥm f + Ĥr es = Ĥ ↓ + Ĥr es , (6.84)

where the effective Hamiltonian Ĥ ↓ is given by

N ⎝
 N ⎟
⎞ ⎠
↓ 2 2
Ĥ = Ĥ0 + Ĥm f = ĥ(xi ) + Um f (xi ) = − → + U (xi ) + Um f (xi ) .
2m i
i=1 i=1
(6.85)
Neglecting the residual Hamiltonian Ĥr es , the system is described by the effective
Hamiltonian Ĥ ↓ and the N -body problem is reduced to the solution of the 1-body
problem ⎧ ⎨
h(x) + Um f (x) ρn (x) = πn ρn (x). (6.86)

Obviously, the main difficulty of this approach is to determine the mean-field potential
Um f (x) which minimizes the effect of the residual interaction Ĥr es . The mean-field
potential Um f (x) can be chosen on the basis of the properties of the specific physical
system or on the basis of known experimental data.
As shown in the previous subsections, Um f (x) can be also obtained by using the
Hartree-Fock method. In the case of the Hartree-Fock method for bosons, introducing
the effective Hamiltonian
N ⎝
 ⎞
Ĥ ↓ = ĥ(xi ) + Um f (xi ) (6.87)
i=1

with Ûm f (xi ) given by Eq. (6.56), it is immediate to verify that the bosonic many-
body wavefunction (x1 , x2 , . . . , x N ) of Eq. (6.51), with the single-particle wave-
function ρ(x) satisfying Eq. (6.55), is the lowest eigenfunction of Ĥ ↓ , and such that

Ĥ ↓ (x1 , x2 , . . . , x N ) = N π (x1 , x2 , . . . , x N ). (6.88)

6.3 Interacting Identical Particles 131

On the other hand, (x1 , x2 , . . . , x N ) is not an eigenfunction of the total Hamiltonian

Ĥ , but it is an approximate variational wavefunction of the ground-state of the full
Hamiltonian Ĥ .
In a similar way, in the case of the Hartree-Fock method for fermions, introducing
again the effective Hamiltonian

N ⎝
 ⎞
↓
Ĥ = ĥ(xi ) + Ûm f (xi ) (6.89)
i=1

with Ûm f (xi ) given by Eq. (6.71), it is not difficult to verify that the fermionic
many-body wavefunction (x1 , x2 , . . . , x N ) of Eq. (6.63), with the single-particle
wavefunction ρ(x) satisfying Eq. (6.70), is the lowest eigenfunction of Ĥ ↓ , and such
that

N
↓
Ĥ (x1 , x2 , . . . , x N ) = πi (x1 , x2 , . . . , x N ). (6.90)
i=1

On the other hand, (x1 , x2 , . . . , x N ) is not an eigenfunction of the total Hamiltonian

Ĥ , but it is an approximate variational wavefunction of the ground-state of the full
Hamiltonian Ĥ .

6.4 Density Functional Theory

In 1927 Llewellyn Thomas and Enrico Fermi independently proposed a statistical ap-
proach to the electronic structure of atoms with a large number N of electrons, which
avoids tackling the solution of the many-body Schödinger equation by focusing on
the electron local total number density θ(r). The starting point is the (kinetic) energy
density of a uniform ideal gas of electrons at zero temperature in the thermodynamic
limit, given by
3 2
Eideal = (3ε 2 )2/3 θ5/3 , (6.91)
5 2m
where θ is the uniform number density of electrons. Thomas and Fermi considered
N electrons in a atom with nuclear charge Z e, such that

e2
V (r − r↓ ) = . (6.92)
4ε∂0 |r − r↓ |

the electron-electron Coulomb potential, while

Z e2
U (r) = − (6.93)
4ε∂0 |r|
132 6 Many-Body Systems

is the nucleus-electron Coulomb potential with the nucleus centered at the orgin of
the reference frame. Thomas and Fermi supposed that the ground-state energy of the
electrons can be captured by the functional

E T F [θ] = TT F [θ] + E D [θ] + d 3 r U (r) θ(r) , (6.94)

where 
3 2
TT F [θ] = d 3r (3ε 2 )2/3 θ(r)5/3 (6.95)
5 2m

is the Thomas-Fermi kinetic energy, i.e. the local approximation to the kinetic energy
of the ideal Fermi gas,

1
E D [θ] = d 3 r d 3 r↓ θ(r) V (r − r↓ ) θ(r↓ ) , (6.96)
2

is the (Hartree-like) direct energy of electron-electron interaction, and the third term
in Eq. (6.94) is the energy of the nucleus-electron interaction.
It is important to stress that within this formalism the electron-nucleus potential
U (r) could be more general, taking into account the effect of many nucleons. The
functional is minimized under the constraints θ(r) ∓ 0 and

N = d 3 r θ(r) (6.97)

to find the equilibrium condition

2
(3ε 2 )2/3 θ2/3 + U (r) + Um f (r) = μ , (6.98)
2m
where 
Um f (r) = d 3 r↓ V (r − r↓ ) θ(r↓ ) (6.99)

is the Hartree-like mean-field potential acting on the electrons and μ is the Lagrange
multiplier fixed by the normalization. The previous equation can be written as

3/2
(2m)3/2 3 ↓ ↓ ↓
θ(r) = μ − U (r) − d r V (r − r ) θ(r ) (6.100)
3ε 2 3

Equation (6.100) is an implicit integral equation for the local electronic density θ(r)
which can be solved numerically by using an iterative procedure.
Although this was an important first step, the Thomas-Fermi method is limited be-
cause the resulting kinetic energy is only approximate, and also because the method
does not include the exchange energy of the electrons due to the Pauli principle. A bet-
6.4 Density Functional Theory 133

ter theoretical description is obtained with the Thomas-Fermi-Dirac-von Weizsäcker

model, whose energy functional is given by

E T F DW [θ] = E T F [θ] + E X [θ] + E W [θ] , (6.101)

where E T F [θ] is the Thomas-Fermi energy functional of Eq. (6.94),


3 e2 3 1/3
E X [θ] = − d 3r ( ) θ(r)4/3 (6.102)
4 4ε∂0 ε

is the functional correction introduced by Paul Dirac in 1930 to take into account the
exchange energy which appears in the Hartree-Fock method, and

2 ⎩
E W [θ] = d 3r σ (→ θ(r))2 (6.103)
2m

is the functional correction introduced by Carl Friedrich von Weizsäcker in 1953 to

improve the accuracy of the kinetic energy with σ an adjustable parameter. Notice
that a simple dimensional analysis shows that the Thomas-Fermi kinetic energy
density must scale as 2 θ5/3 /(2m), while the exchange energy density must scale as
e2 θ4/3 /(4ε∂0 ).
In 1964 the density functional approach was put on a firm theoretical footing by
Pierre Hohenberg and Walter Kohn. We follow their original reasoning which does
not take into account explicitly the spin degrees. First of all they observed that, given
a many-body wavefunction (r1 , r2 , . . . , r N ), the associated one-body density θ(r)
reads


N 
θ(r) = ∈| α(r − ri )|∇ = N d 3 r2 . . . d 3 r N |(r, r2 , . . . , r N )|2 . (6.104)
i=1

Then they formulated two rigorous theorems for a system of identical particles
(bosons or fermions) described by the Hamiltonian

Ĥ = T̂ + Û + V̂ (6.105)

where

N
2 2
T̂ = − → (6.106)
2m i
i=1

is the many-body kinetic energy operator,


N
Û = U (ri ) (6.107)
i=1
134 6 Many-Body Systems

is the many-body external potential operator, and

1 
N
V̂ = V (ri , r j ) (6.108)
2
i, j=1
i∞= j

is the many-body interaction potential operator. We now state the two theorems of
Hohenberg and Kohn, which are based on the variational principle discussed in the
previous section, but first of all we observe that for any (r1 , r2 , . . . , r N ) one has

∈| Ĥ |∇ = ∈|T̂ + V̂ |∇ + ∈|Û |∇ = ∈|T̂ + V̂ |∇ + U (r) θ(r) d 3 r.
(6.109)
This formula shows explicitly that the external energy ∈|Û |∇ is a functional of
the local density θ(r). The Hohemberg-Kohn theorems ensure that also the internal
energy ∈| F̂|∇ = ∈|T̂ + V̂ |∇ is a functional of θ(r).

Theorem 2 For a system of N identical interacting particles in an external potential

U (r) the density θgs (r) of the non degenerate ground-state gs (r1 , r2 , . . . , r N ) is
uniquely determined by the external potential. In other words, there is a one-to-one
correspondence between U (r) and θgs (r).

Proof The proof of this theorem proceeds by reductio ad absurdum. Let there be
two different external potentials, U1 (r) and U2 (r), that give rise to the same ground-
state density θgs (r). Due to the variational principle, the associated Hamiltonians,
Ĥ1 = F̂ + Û1 and Ĥ2 = F̂ + Û2 have different ground-state wavefunctions,
gs,1 (r1 , r2 , . . . , r N ) and gs,2 (r1 , r2 , . . . , r N ), that each yield θgs (r). Using the
variational principle

E gs,1 < ∈gs,2 | Ĥ1 |gs,2 ∇ = ∈gs,2 | Ĥ2 |gs,2 ∇ + ∈gs,2 | Ĥ1 − Ĥ2 |gs,2 ∇

= E gs,2 + θgs (r)[U1 (r) − U2 (r)] d 3 r, (6.110)

where E gs,1 and E gs,2 are the ground-state energies of Ĥ1 and Ĥ2 respectively. An
equivalent expression for Eq. (6.110) holds when the subscripts are interchanged,
namely

E gs,2 < ∈gs,1 | Ĥ2 |gs,1 ∇ = ∈gs,1 | Ĥ1 |gs,1 ∇ + ∈gs,1 | Ĥ2 − Ĥ1 |gs,1 ∇

= E gs,1 + θgs (r)[U2 (r) − U1 (r)] d 3 r. (6.111)

Therefore adding the inequality (6.110) to the inequality (6.110) leads to the result:

E gs,1 + E gs,2 < E gs,2 + E gs,1 , (6.112)

6.4 Density Functional Theory 135

which is a contradiction.

From this theorems one deduces immediately two corollaries.

Corollary 1 The many-body ground-state wavefunction gs (r1 , r2 , . . . , r N ) is a
bijective function of θgs (r).

Proof As just shown in Theorem 1, θgs (r) determines U (r), and U (r) determines Ĥ
and therefore gs (r1 , r2 , . . . , r N ). This means that gs (r1 , r2 , . . . , r N ) is a bijective
function of θgs (r). Symbolically we can write

gs = G[θgs ]. (6.113)

Corollary 2 The internal energy ∈gs | F̂|gs ∇ of the ground state is a universal
functional F[θ] of θgs (r).

Proof E gs is a functional of gs (r1 , r2 , . . . , r N ). Symbolically we have

E gs = [gs ] . (6.114)

But we have just seen that gs (r1 , r2 , . . . , r N ) is a bijective function of θgs (r), in
particular
gs = G[θgs ] . (6.115)

Consequently E gs is a functional of θgs (r), namely

E gs = [G[θgs ]] = E[θgs ] . (6.116)

Due to the simple structure of the external energy, the functional E[θ] is clearly given
by 
E[θ] = F[θ] + d 3 r U (r) θ(r) , (6.117)

where F[θ] is a universal functional which describes the internal (kinetic and inter-
action) energy.

Theorem 3 For a system of N identical interacting particles in an external potential

U (r) the density functional

E[θ] = F[θ] + d 3 r U (r) θ(r)

is such that E[θ] ∓ E[θgs ] = E gs for any trial density θ(r), and the equality holds
only for θ(r) = θgs (r).

Proof We have seen that E gs is a functional of gs (r1 , r2 , . . . , r N ) but also a func-
tional of θgs (r), i.e.
136 6 Many-Body Systems

E gs = [gs ] = [G[θgs ]] = E[θgs ] . (6.118)

If (r1 , r2 , . . . , r N ) ∞= gs (r1 , r2 , . . . , r N ) from the variational principle we have

[] = E[θ] > E gs , (6.119)

where θ(r) is a local density different from θgs (r), due to Theorem 2. It follows
immediately
E[θ] > E[θgs ] , (6.120)

if θ(r) ∞= θgs (r).

The functional F[θ] is universal in the sense that it does not depend on the external
potential U (r) but only on the inter-particle potential V (r − r↓ ), which is the familiar
Coulomb potential in the case of electrons. Indeed, the description of electrons in
atoms, molecules and solids is based on the choice of U (r) while F[θ] is the same.
In the last forty years several approaches have been developed to find approximate
but reliable expressions for F[θ] in the case of electrons. For instance, within the
Thomas-Fermi-Dirac-von Weizsäcker model we have seen that

F[θ] = TT F [θ] + E D [θ] + E X [θ] + E W [θ] . (6.121)

Instead, for a Bose-Einstein condensate made of dilute and ultracold atoms, the
bosonic Hatree-Fock method suggests

g
F[θ] = E W [θ] + d 3 r θ(r)2 (6.122)
2

because V (r − r↓ )  g α(r − r↓ ), where g = 4ε2 as /m, with m the atomic mass

and as the inter-atomic s-wave scattering length. Note that on the basis of empirical
observations in the E W [θ] term one usually takes σ = 1/6 for electrons and σ = 1
for bosons.
Nowadays the most used density functional for electrons is the one proposed
by Walter Kohn and Lu Jeu Sham in 1965. In the Kohn-Sham density functional
approach the universal functional F[θ] is given by

F[θ] = TK S [θ] + E D [θ] + E XC [θ] (6.123)

where

N

2 2 ↑
TK S [θ] = ρi↑ (r) − → ρi (r) (6.124)
2m
i=1

is the Kohn-Sham kinetic energy, where the orbitals ρi (r) determine the local density,
namely
6.4 Density Functional Theory 137


N
θ(r) = |ρi (r)|2 , (6.125)
i=1

the direct (or Hartree-like) energy of interaction E D [θ] has the familiar form of Eq.
(6.96), and E XC [θ] is the so-called exchange-correlation energy, which simply takes
into account the missing energy with respect to the exact result. The minimization
of the Kohn-Sham density functional gives
⎟  ⎠
2 2 α E XC [θ]
− → + U (r) + d 3 r↓ V (r − r↓ ) θ(r↓ ) + ρi (r) = πi ρi (r)
2m αθ(r)
(6.126)
which are the local Kohn-Sham equations for the orbitals ρi (r), with πi the Lagrange
multipliers fixed by the normalization to one of the orbitals. Notice that the third
term on the left side of the previous equation is obtained by using

α E XC [θ] α E XC [θ] αθ(r) α E XC [θ]

↑ = ↑ = ρi (r) . (6.127)
αρi (r) αθ(r) αρi (r) αθ(r)

In many applications the (usually unknown) exchange-correlation energy is written

as
E XC [θ] = E X [θ] + E C [θ] , (6.128)

where the correlation energy E C [θ] is fitted from Monte Carlo calculations. Contrary
to the Hartree-Fock theory, in the Kohn-Sham approach the single-particle orbitals
ρi (r) are not related to a many-body wavefunction but to the local density θ(r): this
implies that the orbitals ρi (r) and energies πi of the Kohn-Sham density functional
have no deep physical meaning. Nevertheless, from the numerical point of view, the
solution of the Kohn-Sham equations is much simpler than that of the Hartree-Fock
ones, since, contrary to the effective Hartree-Fock mean-field potential, the effective
Kohn-Sham mean-field potential is local.
It is important to stress that the density functional theory is an extremely useful
approach for the description of atoms, molecules, and metals. Nowadays the success
of density functional theory not only encompasses standard bulk materials but also
complex materials such as proteins, carbon nanotubes, and nuclear matter.

6.5 Molecules and the Born-Oppenheimer Approximation

Up to now we have considered only single atoms with many-electrons. In this section
we discuss systems composed by many atoms, and in particular molecules.
It is well known that a generic molecule is made of Nn atomic nuclei with electric
charges Z φ e and masses Mφ (φ = 1, 2, . . . , Nn ) and Ne electrons with charges −e
and masses m. Neglecting the finite structure of atomic nuclei, the Hamiltonian of
138 6 Many-Body Systems

the molecule can be written as

Ĥ = Ĥn + Ĥe + Vne , (6.129)

where

Nn
2 2 1 
Nn
Z φ Z λ e2 1
Ĥn = − →φ + (6.130)
2Mφ 2 4ε∂0 |Rφ − Rλ |
φ=1 φ,λ=1
φ∞=λ

is the Hamiltonian of the atomic nuclei, with Rφ the position of the φ-th nucleus,


Ne
2 2 1 
Ne
e2 1
Ĥe = − →i + (6.131)
2m 2 4ε∂0 |ri − r j |
i=1 i, j=1
i∞= j

is the Hamiltonian of the electrons, with ri the position of the i-th electron, and

Nn ,Ne
1  Z φ e2 1
Vne = − (6.132)
2 4ε∂0 |Rφ − ri |
φ,i=1
φ∞=i

is the potential energy of the Coulomb interaction between atomic nuclei and elec-
trons.
It is clear that the computation of the ground-state energy and the many-body
wavefunction of an average-size molecule is a formidable task. For instance, the
benzene molecule (C6 H6 ) consists of 12 atomic nuclei and 42 electrons, and this
means that its many-body wavefunction has (12 + 42) × 3 = 162 variables: the
spatial coordinates of the electrons and the nuclei. The exact many-body Schrodinger
equation for the ground-state is given by

Ĥ (R, r ) = E (R, r ) , (6.133)

where (R, r ) = (R1 , . . . , R Nn , r1 , . . . , r Ne ) is the ground-state wavefunction,

with R = (R1 , . . . , R Nn ) and r = (r1 , . . . , r Ne ) multi-vectors for nuclear and elec-
tronic coordinates respectively.
In 1927 Max Born and Julius Robert Oppenheimer suggested a reliable approxi-
mation to treat this problem. Their approach is based on the separation of the fast elec-
tron dynamics from the slow motion of the nuclei. In the so-called Born-Oppenheimer
approximation the many-body wave function of the molecule is factorized as

(R, r ) = e (r ; R) n (R) , (6.134)

where n (R) is the nuclear wavefunction and e (r ; R) is the electronic wavefunc-

tion, which depends also on nuclear coordinates. Moreover, one assumes that the
6.5 Molecules and the Born-Oppenheimer Approximation 139

electronic wavefunction e (r ; R) is a solution of the following eigenvalue equation

⎛ ⎜
Ĥe + Vne e (r ; R) = E e (R)e (r ; R) . (6.135)

where the electronic eigenenergy E e (R) depends on the nuclear coordinates because
of the nucleon-electron potential Vne = Vne (R, r ). This equation describes an elec-
tronic eigenstate compatible with a given geometrical configuration R of the atomic
nuclei. Physically, it corresponds on assuming that nuclear and electronic motions
are somehow decoupled. In this way one gets
⎛ ⎜
Ĥ e (r ; R) n (R) = Ĥn + Ĥe + Vne e (r ; R) n (R)
⎛ ⎜
= Ĥn e (r ; R) n (R) + Ĥe + Vne e (r ; R) n (R)

= Ĥn e (r ; R) n (R) + E e (R)e (r ; R) n (R)

⎛ ⎜
 e (r ; R) Ĥn n (R) + E e (R) n (R) (6.136)

where the last approximate equality  is sound because the effect of the kinetic
term of the operator Ĥn on the electronic wavefunction e (r ; R) is small, being
the nuclear masses Mφ much larger than the electronic mass m. Finally, from the
previous equation and Eqs. (6.133) and (6.133) one finds
⎛ ⎜
Ĥn + E e (R) n (R) = E n (R) , (6.137)

that is the equation of the adiabatic motion of atomic nuclei in the mean-field potential
E e (R) generated by the electrons.
Equations (6.135) and (6.137) are still quite complex, and they are usually solved
within some approximate quantum many-body method, e.g. the Hartree-Fock method
or the density functional theory.
To conclude, we observe that in this section we have considered molecules, but
the Born-Oppenheimer approximation is crucial in any context where there is more
than a single atom around, which includes atomic gases, clusters, crystals, and many
other physical systems.

6.6 Solved Problems

Problem 6.1
By using the Gaussian variational method calculate the approximate energy of the
ground-state of a one-dimensional quantum particle under the action of the quartic
potential
U (x) = A x 4 .
140 6 Many-Body Systems

Solution
The stationary Schrödinger equation of the particle is given by
⎟ ⎠
2 δ 2
− + A x β(x) = π β(x) .
4
2m δx 2

This equation can be seen as the Euler-Lagrange equation obtained by minimizing

the energy functional
 ⎟ ⎠  ⎟ 2 ⎠
↑ 2 δ 2   δβ(x) 2
E= d x β (x) − + A x β(x) = d x
4
  + A x |β(x)| ,
4 2
2m δx 2 2m δx

with the normalization condition


d x|β(x)|2 = 1 .

According to the Gaussian variational method the wavefunction is supposed to be

given by
e−x /(2ν )
2 2

β(x) = 1/4 1/2 ,

ε ν
where ν is the variational parameter. Inserting this variational wavefunction in the
energy functional and integrating over x we get

2 3
E= 2
+ A ν4 .
4mν 4

Minimizing this energy with respect to the variational parameter ν, i.e. setting dE
dν =
0, we find
2
1/6

ν= .
6m A

Substituting this value of ν in the energy E we finally obtain

2/3
9 2
E= A1/3 ,
4 6m

which is the approximate energy of the ground-state. This energy is surely larger or
equal to the energy associated to the exact ground-state of the system.
Problem 6.2
Derive the Gross-Pitaevskii equation for a system of N Bose-condensed particles
with contact interaction
V (r − r↓ ) = g α(r − r↓ ) ,
6.6 Solved Problems 141

and prove that the Lagrange multiplier π of the Gross-Pitaevskii equation is the
chemical potential μ of the system.
Solution
The Hartree equation for N Bose-condensed particles is given by
⎟  ⎠
2 2
− → + U (r) + (N − 1) |β(r↓ )|2 V (r − r↓ ) β(r) = π β(r) ,
2m

where β(r) is the wavefunction normalized to one. Inserting the contact potential
we obtain ⎟ ⎠
2 2
− → + U (r) + (N − 1)|β(r)| β(r) = π β(r) .
2
2m

Under the reasonable hypotesis that N  1 we get

⎟ ⎠
2 2
− → + U (r) + N g|β(r)| β(r) = π β(r) .
2
2m

that is the Gross-Pitaevskii equation. This equation describes accurately a dilute

Bose-Einstein condensate where the true inter-particle potential can be approximated
with the contact potential. In this case the parameter g can be written as

4ε2 as
g= ,
m

where as is the s-wave scattering length of the true inter-particle potential V (r − r).
The energy functional associated to the Gross-Pitaevskii equation is given by
  ⎟ ⎠ 
2 2 1
E=N d 3 r β ↑ (r) − → + U (r) β(r) + N g|β(r)|4 .
2m 2

The chemical potential is defined as

δE
μ= .
δN
On the basis of this definition we obtain
  ⎟ ⎠ 
2 2
μ = d 3 r β ↑ (r) − → + U (r) β(r) + N g|β(r)|4 .
2m

On the other hand, if we insert β ↑ (r) on the left side of the Gross-Pitaevskii equation
and we integrate over r we get
142 6 Many-Body Systems
  ⎟ ⎠ 
↑ 2 2
d r β (r) −
3
→ + U (r) β(r) + N g|β(r)| = π ,
4
2m

because 
d 3 r|β(r)|2 = 1 .

By comparing the formulas of π and μ we conclude that π = μ. This result is also

known as Koopmans theorem. Indeed, more rigorously the chemical potential defined
as
μ = E N − E N −1 ,

but also in this case one easily finds that π = μ using the correct expression of the
energy
  ⎟ ⎠ 
2 2 1
EN = N d 3 r β ↑ (r) − → + U (r) β(r) + (N − 1)g|β(r)|4 ,
2m 2

where it appears (N − 1) instead of N in front of the interaction strength g.

Problem 6.3
Write the energy functional associated to the Gross-Pitaevskii equation as a functional
of the local density θ(r) of the Bose-Einstein condensate.
Solution
We have seen that the energy functional associated to the Gross-Pitaevskii equation
of the Bose-Einstein condensate is given by
  ⎟ ⎠ 
2 2 1
E=N d 3 r β ↑ (r) − → + U (r) β(r) + N g|β(r)|4 ,
2m 2

but it can also be written as

  2 
 1
E = N d 3r |→β(r)|2 + U (r)|β(r)|2 + N g|β(r)|4 ,
2m 2

if the wavefunction is zero at the surface of integration volume. We now introduce

the local density of the Bose-Einstein condensate as

θ(r) = N β(r)2 ,

supposing that the wavefunction is real. In this way the energy functional can be
immediately written as
  
2 ⎛ ⎩ ⎜2 1
E= d 3r → θ(r) + U (r)θ(r) + gθ(r)2 ,
2m 2
6.6 Solved Problems 143

that is a fuctional of the local density θ(r), as required.

Problem 6.4
By using the Gaussian variational method on the Gross-Pitaevskii functional calcu-
late the energy per particle of a Bose-Einstein condensate under harmonic confine-
ment, given by
1
U (r) = mω 2 (x 2 + y 2 + z 2 ) .
2
Solution
We start from the Gaussian variational wave function
1
e−(x
2 +y 2 +z 2 )/(2a 2 ν 2 )
β(r) = 3/2
H ,
ε 3/4 a H ν 3/2
∀
where a H = /(mω) is the characteristic length of the harmonic confinement
and ν is the adimensional variational parameter. Inserting this wave function in the
Gross-Pitaevskii energy functional
  ⎟ ⎠ 
2 2 1 1
E=N d 3 r β ↑ (r) − → + mω 2 (x 2 + y 2 + z 2 ) β(r) + N g|β(r)|4 ,
2m 2 2

after integration we obtain the energy E as a function of ν, namely

3 1 3 τ 1
E = N ω + ν2 + ,
4ν 2 4 2 ν3
∀
where τ = 2/ε N as /a H is the adimensional strength of the inter-particle interac-
tion with g = 4ε2 as /m and as the s-wave scattering length. The best choice for ν
is obtained by extremizing the energy function:

dE
=0,
dν
from which we obtain
ν(ν 4 − 1) = τ .

It is clear that ν grows with τ, if τ > 0. Instead, if τ < 0 there are two possible values
of ν: one value corresponds to a minimum of the energy E (meta-stable solution)
and the other corresponds to a maximum of the energy E (unstable solution). It is
straightforward to show that these two solutions exists only if τ > −4/55/4 .
144 6 Many-Body Systems

Further Reading

For the many-electron atom and the Hartree-Fock method:

B.H. Bransden, C.J. Joachain, Physics of Atoms and Molecules, Chap. 8, Sects. 8.1,
8.2 and 8.4 (Prentice Hall, Upper Saddle River, 2003)
For the density functional theory:
B.H. Bransden, C.J. Joachain, Physics of Atoms and Molecules, Chap. 8, Sects. 8.3
and 8.6 (Prentice Hall, Upper Saddle River, 2003)
E. Lipparini, Modern Many-Particle Physics: Atomic Gases, Quantum Dots and
Quantum Fluids, Chap. 4, Sects. 4.1, 4.2, and 4.3 (World Scientific, Singapore, 2003)
Chapter 7
Second Quantization of Matter

In this chapter we discuss the second quantization of the non-relativistic matter field,
that is the Schrödinger field. We show that the Schrödinger field can be expressed
as a infinite sum of harmonic oscillators. These oscillators, which describe the
possible eigenfrequencies of the matter field, are quantized by introducing creation
and annihilation operators acting on the Fock space of number representation. We
show how, depending on the commutation rule, these ladder operators can model
accurately bosonic and fermionic particles which interact among themselves or with
the quantum electromagnetic field. The second quantization (also called quantum
field theory) is the powerful tool to describe phenomena, both at zero and finite tem-
perature, where the number of particles is not conserved or it is conserved only on
the average.

7.1 Schrödinger Field

In Chap. 2 we have seen that the light field is composed of an infinite number of
quanta, the photons, as explained in 1927 by Paul Dirac. In the same year Eugene
Wigner and Pascual Jordan proposed something similar for the matter. They sug-
gested that in non-relativistic quantum mechanics the matter field is nothing else
than the single-particle Schrödinger field ν(r, t) of quantum mechanics, which sat-
isfies the time-dependent Schrödinger equation
 
ρ 2 2
i ν(r, t) = − ∧ + U(r) ν(r, t), (7.1)
ρt 2m

where U(r) is the external potential acting on the quantum particle. Setting

ν(r, t) = πβ (r)e−iβ t/ (7.2)

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 145
DOI: 10.1007/978-3-319-05179-6_7, © Springer International Publishing Switzerland 2014
146 7 Second Quantization of Matter

one finds
 
2
− ∧ 2 + U(r) πβ (r) = β πβ (r) (7.3)
2m

that is the stationary Schrödinger equation for the eigenfunction πβ (r) with
eigenvalue β where β is the label which represents a set of quantum numbers.
In this case the Schrödinger field ν(r, t) is a stationary state of the system, in fact

|ν(r, t)|2 = |πβ (r)|2 (7.4)

does not depend on time t.

In general, the Schrödinger field ν(r, t) is not a stationary state of the system
because the space dependence and time dependence cannot expressed as in Eq. (7.2),
but, as suggested by Wigner and Jordan, one can surely expand the field as

ν(r, t) = cβ (t) πβ (r) (7.5)
β

where 
d 3 r π∗β (r)πε (r) = λβε . (7.6)

In this way, from the time-dependent Schrödinger equation we get

i ċβ (t) = β cβ (t), (7.7)

whose general solution is obviously

cβ (t) = cβ (0) e−iβ t/. (7.8)

In this case the Schrödinger field ν(r, t) is not a stationary state of the system, in
fact

|ν(r, t)|2 = cβ∗ (0)cε (0) e−i(β −ε )t/ π∗β (r)πε (r) (7.9)
β,ε

depends on time t, while obviously its integral

 
d 3 r|ν(r, t)|2 = |cβ (0)|2 (7.10)
β

does not.
The constant of motion associated to the Schrödinger field ν(r, t) is the average
total energy of the system, given by
7.1 Schrödinger Field 147
  
∗ 2 2
H= d r ν (r, t) − ∧ + U(r) ν(r, t).
3
(7.11)
2m

Inserting the expansion (7.5) into the total energy we find

 β  
H= cβ∗ cβ + cβ cβ∗ . (7.12)
β
2

This energy is obviously independent on time: the time dependence of the complex
amplitudes cβ∗ (t) and cβ (t) cancels due to Eq. (7.8).
Instead of using the complex amplitudes cβ∗ (t) and cβ (t) one can introduce the
real variables

2 1  
qβ (t) = cβ (t) + cβ∗ (t) (7.13)
θβ 2
1  
pβ (t) = 2θβ cβ (t) − cβ∗ (t) (7.14)
2i
such that the energy of the matter field reads


p2 1

β
H= + θβ2 qβ2 , (7.15)
β
2 2

where β
θβ = (7.16)

is the eigenfrequency associated to the eigenenergy β .
This energy resembles that of infinitely many harmonic oscillators with unitary
mass and frequency θβ . It is written in terms of an infinite set of real harmonic
oscillators: one oscillator for each mode characterized by quantum numbers β and
frequency θβ . It is clear the analogy between the energy of the Schrödinger field and
the energy of the classical electromagnetic field.

7.2 Second Quantization of the Schrödinger Field

The canonical quantization of the classical Hamiltonian (7.15) is obtained by

promoting the real coordinates qβ and the real momenta pβ to operators:

qβ ∇ q̂β , (7.17)
pβ ∇ p̂β , (7.18)

satisfying the commutation relations

148 7 Second Quantization of Matter

[q̂β , p̂ε ] = i λβε , (7.19)

The quantum Hamiltonian is thus given by


p̂2 1 2 2

β
Ĥ = + θβ q̂β . (7.20)
β
2 2

The formal difference between Eqs. (7.15) and (7.20) is simply the presence of
the “hat symbol” in the canonical variables.
We now introduce annihilation and creation operators
⎛

θβ i
ĉβ = q̂β + p̂β , (7.21)
2 θβ
⎛

θβ i
ĉβ+ = q̂β − p̂β , (7.22)
2 θβ

which satisfy the commutation relations

[ĉβ , ĉε+ ] = λβ,ε , [ĉβ , ĉε ] = [ĉβ+ , ĉε+ ] = 0, (7.23)

and the quantum Hamiltonian (7.20) becomes



1
Ĥ = β ĉβ+ ĉβ + . (7.24)
β
2

Obviously this quantum Hamiltonian can be directly obtained from the classical one,
given by Eq. (7.12), by promoting the complex amplitudes cβ and cβ∗ to operators:

cβ ∇ ĉβ , (7.25)
cβ∗ ∇ ĉβ+ , (7.26)

satisfying the commutation relations (7.23).

The operators ĉβ and ĉβ+ act in the Fock space of the “particles” of the Schrödinger
field. A generic state of this Fock space is given by

| . . . nβ . . . nε . . . n∂ . . . →, (7.27)

meaning that there are nβ particles in the single-particle state |β→, nε particles in the
single-particle state |ε→, n∂ particles in the single-particle state |∂→, etc. The operators
ĉβ and ĉβ+ are called annihilation and creation operators because they respectively
destroy and create one particle in the single-particle state |β→, namely
7.2 Second Quantization of the Schrödinger Field 149

∀
ĉβ | . . . nβ . . . → = nβ | . . . nβ − 1 . . . →, (7.28)

+
ĉβ | . . . nβ . . . → = nβ + 1 | . . . nβ + 1 . . . →. (7.29)

Note that these properties follow directly from the commutation relations (7.23). The
vacuum state, where there are no particles, can be written as

|0→ = | . . . 0 . . . 0 . . . 0 . . . →, (7.30)

and

ĉβ |0→ = 0, (7.31)

ĉβ+ |0→ = |1β → = |β→, (7.32)

where |β→ is such that

√r|β→ = πβ (r). (7.33)

From Eqs. (7.28) and (7.29) it follows immediately that

N̂β = ĉβ+ ĉβ (7.34)

is the number operator which counts the number of particles in the single-particle
state |β→, i.e.
N̂β | . . . nβ . . . → = nβ | . . . nβ . . . →. (7.35)

Notice that the quantum Hamiltonian of the matter field can be written as

Ĥ = β N̂β , (7.36)
β

after removing the puzzling zero-point energy. This is the second-quantization

Hamiltonian of non-interacting matter.
Introducing the time-evolution unitary operator

Û(t) = e−iĤt/, (7.37)

one finds that obviously

Û + (t) ĉβ (0)Û(t) = ĉβ (0) e−iβ t/. (7.38)

In fact, one can write

d ⎜ + ⎝ d ⎜ iĤt/ ⎝
i Û (t) ĉβ (0) Û(t) = i e ĉβ (0) e−iĤt/
dt dt
150 7 Second Quantization of Matter


i i
= i Ĥ ĉβ (t) − ĉβ (t) Ĥ
 
= ĉβ (t) Ĥ − Ĥ ĉβ (t)
= [ĉβ (t), Ĥ]

= [ĉβ (t), ε ĉε+ (0) ĉε (0)]
ε

= ε [ĉβ (t), ĉε+ (t) ĉε (t)]
ε

= ε [ĉβ (t), ĉε+ (t)] ĉε (t)
ε

= ε λβ,ε ĉε (t)
ε
= β ĉβ (t). (7.39)

Thus, one obtains

d ⎜ + ⎝
i Û (t) ĉβ (0) Û(t) = β ĉβ (t), (7.40)
dt
with solution

Û + (t) ĉβ (0) Û(t) = ĉβ (0) eiβ t/. (7.41)

This is clearly the operatorial version of the classical solution (7.8), where

ĉβ (t) = Û + (t) ĉβ (0) Û(t) (7.42)

is the familiar expression of the time-evolution of the operator.

7.2.1 Bosonic and Fermionic Matter Field

The annihilation and creation operators ĉβ and ĉβ+ which satisfy the commu-
tation rules (7.23) are called bosonic operators and the corresponding quantum
field operator

ν̂(r, t) = ĉβ (t) πβ (r) (7.43)
β

is the bosonic field operator. Indeed the commutation rules (7.23) imply Eqs. (7.28)
and (7.29) and, as expected for bosons, there is no restriction on the number of
particles nβ which can occupy in the single-particle state |β→.
7.2 Second Quantization of the Schrödinger Field 151

To obtain fermionic properties it is sufficient to impose anti-commutation rules

for the operators ĉβ and ĉβ+ , i.e.

{ĉβ , ĉε+ } = λβε , {ĉβ , ĉε } = {ĉβ+ , ĉε+ } = 0, (7.44)

where {Â, B̂} = ÂB̂ + B̂Â are the anti-commutation brackets. An important
consequence of anti-commutation is that

(ĉβ+ )2 = 0. (7.45)

Notice that Eqs. (7.28) and (7.29) formally work also for fermions but with the
constraints that nβ = 0 or nβ = 1 for any single-particle state |β→. In fact, for
fermions
 2
N̂β2 = ĉβ+ ĉβ = ĉβ+ ĉβ ĉβ+ ĉβ (7.46)
 
= ĉβ+ 1 − ĉβ+ ĉβ ĉβ = ĉβ+ ĉβ (7.47)
= N̂β , (7.48)

which implies that the eigenvalues of N̂β can be only 0 and 1. For a fermionic two-
particle state we have

|1β 1ε → = ĉβ+ ĉε+ |0→ = −ĉε+ ĉβ+ |0→ = −|1ε 1β →, (7.49)

i.e. the state is anti-symmetric under interchange of particle labels, and moreover

|2β → = (ĉβ+ )2 |0→ = 0, (7.50)

expressing the Pauli principle that two particles cannot be created in the same single-
particle state. For fermions the field operator (7.43) is called fermionic field operator.
A remarkable property of the field operator ν̂ + (r, t), which works for bosons and
fermions, is the following:

ν̂ + (r, t)|0→ = |r, t→ (7.51)

that is the operator ν̂ + (r, t) creates a particle in the state |r, t→ from the vacuum state
|0→. In fact,
 
ν̂ + (r, t)|0→ = ĉβ+ (t) π∗β (r)|0→ = ĉβ+ eiβ t/√β|r→|0→
β β

iβ t/
= e |β→√β|r→ = |r, t→, (7.52)
β
152 7 Second Quantization of Matter

because
  
|r, t→ = eiĤt/|r→ = |β→√β|eiĤt/ |ε→√ε|r→ = eiβ t/|β→√β|r→ (7.53)
β ε β

⎞
with ĉβ+ = ĉβ+ (0), √β|ε→ = λβε , and β |β→√β| = 1.
It is straightforward to show that the bosonic field operator satisfies the following
equal-time commutation rules

[ν̂(r, t), ν̂ + (r⊗ , t)] = λ(r − r⊗ ), (7.54)

while for the fermionic field operator one gets

{ν̂(r, t), ν̂ + (r⊗ , t)} = λ(r − r⊗ ). (7.55)

Let us prove Eq. (7.54). By using the expansion of the field operators one finds

[ν̂(r, t), ν̂ + (r⊗ , t)] = πβ (r) π∗ε (r⊗ ) [cβ , cε+ ]
β,ε
 
= πβ (r) π∗ε (r⊗ ) λβ,ε = πβ (r) π∗β (r⊗ )
β,ε β
 
= √r|β→ √β|r⊗ → = √r| |β→ √β| r⊗ →
β β
= √r|r⊗ → = λ(r − r⊗ ). (7.56)

The proof for fermions is practically the same, with anti-commutators instead of
commutators.
Finally, we observe that the many-body quantum Hamiltonian (7.36) can be writ-
ten, both for bosons and fermions, in the elegant form
  
+ 2 2
Ĥ = d r ν̂ (r, t) − ∧ + U(r) ν̂(r, t).
3
(7.57)
2m

This quantum Hamiltonian can be directly obtained from the classical one, given by
Eq. (7.11), by promoting the complex classical field ν(r, t) and ν̂(r, t) to quantum
field operators:

ν(r, t) ∇ ν̂(r, t), (7.58)

ν ∗ (r, t) ∇ ν̂ + (r, t), (7.59)

satisfying the commutation relations (7.54) of bosons or the anti-commutation rela-

tions (7.55) of fermions.
7.3 Connection Between First and Second Quantization 153

7.3 Connection Between First and Second Quantization

In this section we shall analyze relationships between the formalism of second

quantization, where the number of particles is a priori not fixed, and the formal-
ism of first quantization, where the number of particles is instead fixed.
First we observe that, within the formalism of second quantization of field oper-
ators, the time-independent one-body density operator is defined as

ψ̂(r) = ν̂ + (r) ν̂(r), (7.60)

and it is such that 

N̂ = d 3 r ψ̂(r) (7.61)

is the total number operator. By using the expansion


ν̂(r) = ĉβ πβ (r) (7.62)
β

one finds immediately 

ψ̂(r) = ĉβ+ ĉε π∗β (r) πε (r) (7.63)
β,ε

and also  
N̂ = ĉβ+ ĉβ = N̂β , (7.64)
β β

because we consider an orthonormal basis of single-particle wavefunctions πβ (r).

A remarkable connection between second quantization and first quantization is
made explicit by the formula


N
ψ̂(r) |r1 r2 . . . rN → = λ(r − ri ) |r1 r2 . . . rN →, (7.65)
i=1

⎞
where ψ(r) = N i=1 λ(r − ri ) is the one-body density function introduced in the
previous chapter. This formula can be proved, for bosons, as follows:

ψ̂(r) |r1 r2 . . . rN → = ν̂ + (r) ν̂(r) |r1 r2 . . . rN →

= ν̂ + (r) ν̂(r) ν̂ + (r1 ) ν̂ + (r2 ) . . . ν̂ + (rN ) |0→
⎜ ⎝
= ν̂ + (r) λ(r − r1 ) + ν̂ + (r1 ) ν̂(r) ν̂ + (r2 ) . . . ν̂ + (rN ) |0→
= λ(r − r1 ) |r1 r2 . . . rN →
+ ν̂ + (r) ν̂ + (r1 ) ν̂(r) ν̂ + (r2 ) . . . ν̂ + (rN ) |0→
= λ(r − r1 ) |r1 r2 . . . rN →
154 7 Second Quantization of Matter
⎜ ⎝
+ ν̂ + (r) ν̂ + (r1 ) λ(r − r2 ) + ν̂ + (r2 ) ν̂(r) ν̂ + (r3 ) . . . ν̂ + (rN ) |0→
= λ(r − r1 ) |r1 r2 . . . rN → + λ(r − r2 ) |r1 r2 . . . rN →
+ ν̂ + (r) ν̂ + (r2 ) ν̂(r) ν̂ + (r3 ) . . . ν̂ + (rN ) |0→
= ···

N
= λ(r − ri ) |r1 r2 . . . rN →
i=1

+ ν̂ + (r) ν̂ + (r2 ) ν̂ + (r3 ) . . . ν̂ + (rN ) ν̂(r) |0→


N
= λ(r − ri ) |r1 r2 . . . rN →, (7.66)
i=1

taking into account the commutation relations of the field operators ν̂(r) and ν̂ + (r).
For fermions the proof proceedes in a very similar way.
We can further investigate the connection between second quantization and first
quantization by analyzing the Hamiltonian operator. In first quantization, the non-
relativistic quantum Hamiltonian of N interacting identical particles in the external
potential U(r) is given by

N 
 
1   1 
N N
(N) 2 2
Ĥ = − ∧i + U(ri ) + V (ri − rj ) = ĥi + Vij , (7.67)
2m 2 2
i=1 i,j=1 i=1 i,j=1
i∞=j i∞=j

where V (r − r⊗ ) is the inter-particle potential. In second quantization, the quantum

field operator can be written as

ν̂(r) = ĉβ πβ (r) (7.68)
β

where the πβ (r) = √r|β→ are the eigenfunctions of ĥ such that ĥ|β→ = β |β→, and
ĉβ and ĉβ+ are the annihilation and creation operators of the single-particle state |β→.
We now introduce the quantum many-body Hamiltonian
 
Ĥ = β ĉβ+ ĉβ + Vβε∂λ ĉβ+ ĉε+ ĉλ ĉ∂ , (7.69)
β βε∂λ

where 
Vβελ∂ = d 3 r d 3 r⊗ π∗β (r) π∗ε (r⊗ ) V (r − r⊗ ) πλ (r⊗ ) π∂ (r). (7.70)
7.3 Connection Between First and Second Quantization 155

This Hamiltonian can be also written as

  
2
Ĥ = d 3 r ν̂ + (r) − ∧ 2 + U(r) ν̂(r)
2m

1
+ d 3 r d 3 r⊗ ν̂ + (r) ν̂ + (r⊗ ) V (r − r⊗ ) ν̂(r⊗ ) ν̂(r). (7.71)
2

We are now ready to show the meaningful connection between the

second-quantization Hamiltonian Ĥ and the first-quantization Hamiltonian Ĥ (N) ,
which is given by the formula

Ĥ|r1 r2 . . . rN → = Ĥ (N) |r1 r2 . . . rN →. (7.72)

This formula can be proved following the same path of Eq. (7.66). In fact, one
finds that


N
ν̂ + (r) ĥ(r) ν̂(r) |r1 r2 . . . rN → = ĥ(ri )λ(r − ri ) |r1 r2 . . . rN → (7.73)
i=1

and also

ν̂ + (r) ν̂ + (r⊗ ) V (r, r⊗ ) ν̂(r⊗ ) ν̂(r) |r1 r2 . . . rN →


N
= V (ri , ri )λ(r − ri )λ(r⊗ − rj ) |r1 r2 . . . rN →. (7.74)
i,j=1
i∞=j

From these two expressions Eq. (7.72) follows immediately, after space
integration. Similarly to Eq. (7.65), Eq. (7.72) displays a deep connection between
second and first quantization.
Up to now in this section we have considered time-independent quantum field
operators. The time-dependent equation of motion for the field operator ν̂(r, t)
is easily obtained from the Hamiltonian (7.71) by using the familiar Heisenberg
equation
ρ
i ν̂ = [ν̂, Ĥ], (7.75)
ρt
which gives
  
ρ 2
i ν̂ = − ∧ 2 + U(r) ν̂(r) + d 3 r⊗ ν̂ + (r⊗ ) V (r − r⊗ ) ν̂(r⊗ ) ν̂(r). (7.76)
ρt 2m

This exact equation for the field operator ν̂(r, t) is second-quantized version
of the Hartree-like time-dependent nonlinear Schrödinger equation of a complex
156 7 Second Quantization of Matter

wavefunction (classical field) ν(r, t). Note that, in practice, one can formally use the
simpler formula
ρ λ Ĥ
i ν̂ = (7.77)
ρt λ ν̂ +

to determine the Heisenberg equation of motion, where


λ Ĥ λH
= . (7.78)
λ ν̂ + λν ∗ ν=ν̂,ν ∗ =ν̂ +

Obviously, the equation of motion for the time-dependent field operator ν̂ + (r, t) is
obtained in the same way.
Similarly, the time-dependent equation of motion of the field operator ĉβ (t) is
obtained from the Hamiltonian (7.70) by using the Heisenberg equation

d
i ĉβ = [ĉβ , Ĥ], (7.79)
dt
which gives
d 
i ĉβ = ĉβ + Vβε∂λ ĉε+ ĉλ ĉ∂ . (7.80)
dt
ε∂λ

Again, one can formally can the simpler formula

d ρ Ĥ
i ĉβ = + (7.81)
dt ρ ĉβ

to determine the Heisenberg equation of motion of ĉβ , where


ρ Ĥ ρH
= . (7.82)
ρ âβ+ ρaβ∗ aβ =âβ ,aβ∗ =âβ+

7.4 Coherent States for Bosonic and Fermionic Matter Fields

Nowadays it is possible to produce dilute and ultracold bosonic gases made of

alkali-metal atoms in a confining magnetic or optical potential U(r). For instance, one
million of 87 Rb atoms at the temperature of 100 nK in the lowest single-particle state.
The system can be so dilute that the inter-atomic interaction can be neglected. Under
these conditions one has a pure Bose-Einstein condensate made of non-interacting
bosons described by the Hamiltonian
7.4 Coherent States for Bosonic and Fermionic Matter Fields 157
   
+ 2 2
Ĥ = d r ν̂ (r) − ∧ + U(r) ν̂(r) =
3
β ĉβ+ ĉβ . (7.83)
2m β

If there are exactly N bosons in the lowest single-particle state of energy 0 , the
ground-state of the system is a Fock state given by

1  + N
|FN → = ∀ ĉ0 |0→ = |N, 0, 0, . . . →. (7.84)
N!

It is then straightforward to show that

√FN |ν̂(r)|FN → = 0, (7.85)

while the expectation value of ν̂ + (r)ν̂(r) is given by

√FN |ν̂ + (r)ν̂(r)|FN → = N |ν0 (r)|2 . (7.86)

These results are the exact analog of what we have seen for the light field in Chap. 2.
In the spirit of quantum optics one can suppose that the number of massive bosons
in the matter field is not fixed, in other words that the system is not in a Fock state.
We then introduce the coherent state |β0 →, such that

ĉ0 |β0 → = β0 |β0 →, (7.87)

with
√β0 |β0 → = 1. (7.88)

The coherent state |β0 → is thus the eigenstate of the annihilation operator ĉ0 with
complex eigenvalue β0 = |β0 |eiφ0 . |β0 → does not have a fixed number of bosons,
i.e. it is not an eigenstate of the number operator N̂ nor of N̂0 , and it is not difficult
to show that |β0 → can be expanded in terms of number (Fock) states |N, 0, 0, . . . →
as follows
∈
βN
|β0 → = e−|β0 | /2
2
∀ 0 |N, 0, 0, . . . →. (7.89)
N=0
N!

From Eq. (7.87) one immediately finds

N̄ = √β0 |N̂|β0 → = |β0 |2 , (7.90)

and it is natural to set

β0 = N̄ eiφ0 , (7.91)
158 7 Second Quantization of Matter

where N̄ is the average number of bosons in the coherent state, while φ0 is the phase
of the coherent state. The expectation value of the matter field ν̂(r) in the coherent
state |β0 → reads
√β0 |ν̂(r)|β0 → = N̄ eiφ0 π0 (r), (7.92)

with θ0 = 0 /, while the expectation value of ν̂ + (r)ν̂(r) is given by

√β0 |ν̂ + (r)ν̂(r)|β0 → = N̄ |π0 (r)|2 . (7.93)

Up to now we have considered coherent states only for bosonic fields: the elec-
tromagnetic field and the bosonic matter field. Are there coherent states also for the
fermionic field? The answer is indeed positive.
For simplicity, let us consider a single mode of the fermionic field described by
the anti-commuting fermionic operators ĉ and ĉ+ , such that

ĉĉ+ + ĉ+ ĉ = 1, ĉ2 = (ĉ+ )2 = 0. (7.94)

The fermionic coherent state |∂→ is defined as the eigenstate of the fermionic annihi-
lation operator ĉ, namely
ĉ|∂→ = ∂|∂→, (7.95)

where ∂ is the corresponding eigenvalue. It is immedate to verify that, for mathe-

matical consistency, this eigenvalue ∂ must satisfy the following relationships

¯ = 1, ∂ 2 = ∂¯ 2 = 0,
∂ ∂¯ + ∂∂ (7.96)

where ∂¯ is such that

√∂|ĉ+ = ∂√∂|.
¯ (7.97)

Obviously ∂ and ∂¯ are not complex numbers. They are instead Grassmann numbers,
namely elements of the Grassmann linear algebra {1, ∂, ∂,
¯ ∂∂}
¯ characterized by the
independent basis elements 1, ∂, ∂,
¯ with 1 the identity (neutral) element. The most
general function on this Grassmann algebra is given by

¯ ∂) = f11 + f12 ∂ + f21 ∂¯ + f22 ∂∂,

f (∂, ¯ (7.98)

where f11 , f12 , f21 , f22 are complex numbers. In fact, the function f (∂,
¯ ∂) does not have
higher powers of ∂, ∂¯ and ∂∂ ¯ because they are identically zero. The differentiation
with respect to the Grassmann variable ∂ is defined as

ρ
f (∂,
¯ ∂) = f12 − f22 ∂,
¯ (7.99)
ρ∂
7.4 Coherent States for Bosonic and Fermionic Matter Fields 159

where the minus sign occurs because one need to permute ∂¯ and ∂ before differen-
tiation. This is called left differentiation. Clearly, one has also

ρ
¯ ∂) = f21 + f22 ∂
f (∂, (7.100)
ρ ∂¯

and
ρ2 ρ2
f (∂,
¯ ∂) = − ¯ ∂) = −f22 .
f (∂, (7.101)
ρ ∂ρ∂
¯ ρ∂ρ ∂¯

The integration over Grassmann numbers is a bit more tricky. It is defined as equiv-
alent to differention, i.e.
   
ρ ρ ρ2 ρ2
d∂ = , d ∂¯ = , d ∂d∂
¯ = , d∂d ∂¯ = . (7.102)
ρ∂ ρ ∂¯ ρ ∂ρ∂
¯ ρ∂ρ ∂¯

These integrals are not integrals in the Lebesgue sense. They are called integrals
simply because they have some properties of Lebesgue integrals: the linearity and
the fundamental property of ordinary integrals over functions vanishing at infinity
that the integral of an exact differential form is zero. Notice that the mathematical
theory of this kind of integral (now called Berezin integral) with anticommuting
Grassmann variables was invented and developed by Felix Berezin in 1966. The
Berezin integral is mainly used for the path integral formulation of quantum field
theory. A discussion of the path integral formulation of quantum field theory, also
called functional integration, is outside the scope of this book.
To conclude this section we stress that, in full generality, the classical analog of
the bosonic field operator

ν̂(r) = πj (r) âj (7.103)
j

is the complex field 

ν(r) = πβ (r) βj (7.104)
j

such that
ν̂(r)|CSB → = ν(r)|CSB →, (7.105)

where ⎟
|CSB → = |βj → (7.106)
j

is the bosonic coherent state of the system, |βj → is the coherent state of the bosonic
operator âj , and βj is its complex eigenvalue. Instead, the pseudo-classical analog of
the fermionic field operator
160 7 Second Quantization of Matter

ν̂(r) = πj (r) ĉj (7.107)
j

is the Grassmann field 

ν(r) = πβ (r) ∂j (7.108)
j

such that
ν̂(r)|CSF → = ν(r)|CSF →, (7.109)

where ⎟
|CSF → = |∂j → (7.110)
j

is the fermionic coherent state of the system, |∂j → is the coherent state of the fermionic
operator ĉj , and ∂j is its Grassmann eigenvalue.

7.5 Quantum Matter Field at Finite Temperature

Let us consider the non-interacting matter field in thermal equilibrium with a bath at
the temperature T . The relevant quantity to calculate all thermodynamical properties
of the system is the grand-canonical partition function Z, given by

Z = Tr[e−ε(Ĥ−μN̂) ] (7.111)

where ε = 1/(kB T ) with kB the Boltzmann constant,


Ĥ = β N̂β , (7.112)
β

is the quantum Hamiltonian, 

N̂ = N̂β (7.113)
β

is total number operator, and μ is the chemical potential, fixed by the conservation
of the average particle number. This implies that

Z= √ . . . nβ . . . |e−ε(Ĥ−μN̂) | . . . nβ . . . →
{nβ }
 ⎞
= √ . . . nβ . . . |e−ε β (β −μ)N̂β | . . . nβ . . . →
{nβ }
7.5 Quantum Matter Field at Finite Temperature 161
 ⎞ ⎠ ⎠
= e−ε β (β −μ)nβ = e−ε(β −μ)nβ = e−ε(β −μ)nβ
nβ nβ β β nβ
⎠ ∈
 ⎠ 1
= e−ε(β −μ) n = for bosons (7.114)
β n=0 β
1 − e−ε(β −μ)
⎠ 1 ⎠⎜ ⎝
= e−ε(β −μ) n = 1 + e−ε(β −μ) for fermions (7.115)
β n=0 β

Quantum statistical mechanics dictates that the thermal average of any operator  is
obtained as
1
√Â→T = Tr[ e−ε(Ĥ−μN̂) ]. (7.116)
Z

Let us suppose that  = Ĥ, it is then quite easy to show that

1 ρ ⎜ ⎝ ρ
√Ĥ ⊗ →T = Tr[(Ĥ − μN̂) e−ε(Ĥ−μN̂) ] = − ln Tr[e−ε(Ĥ−μN̂ ] = − ln(Z).
Z ρε ρε
(7.117)
By using Eq. (7.114) or Eq. (7.115) we immediately obtain
 ⎜ ⎝
ln(Z) = ln 1 ∓ e−ε(β −μ) , (7.118)
β

where − is for bosons and + for fermions, and finally from Eq. (7.117) we get

√Ĥ→T = β √N̂β →T , (7.119)
β

with
 1
√N̂→T = . (7.120)
β
eε(β −μ) ∓1

Notice that the zero-temperature limit, i.e. ε ∇ ∈, for fermions gives


√Ĥ→0 = β √N̂β →0 (7.121)
β

with 
√N̂→0 =  (μ − β ) , (7.122)
β

where the chemical potential μ at zero temperature is nothing else than the Fermi
energy F , i.e. F = μ(T = 0).
162 7 Second Quantization of Matter

7.6 Matter-Radiation Interaction

In Chap. 3 we have discussed the quantum electrodynamics by using first quantization

for the matter and second quantization of the electromagnetic radiation. Here we write
down the fully second-quantized Hamiltonian. It is given by

Ĥ = Ĥmatt + Ĥrad + ĤI (7.123)

where
  
2
d 3 r ν̂ + (r, t) − ∧ 2 + U(r) ν̂(r, t), (7.124)
Ĥmatt =
2m
 ⎜ α  ρ Â(r, t) 
0 2 1  2 ⎝
Ĥrad = d 3 r + ∇ ↑ Â(r, t) , (7.125)
2 ρt 2μ0
  
+ e e2
ĤI = d r ν̂ (r, t) − Â(r, t) · (−i∇) +
3
Â(r, t) ν̂(r, t). (7.126)
2
m 2m

with ν̂(r, t) the scalar field operator of electrons in the external potential U(r), which
includes the instantaneous electrostatic Coulomb potential, and Â(r, t) the vector
field operator of the electromagnetic radiation. Notice that the coupling Hamiltonian
ĤI describes the very specific situation of the charged matter (usually the electron
gas) coupled to radiation, which is an important but by no means unique case of matter
interacting with electromagnetic radiation using second quantization methods. One
often uses similar techniques to describe e.g. neutral atomic gases as we shall show
in the last section.
Within the dipole approximation, where one neglects the quadratic term of the
vector potential and moreover one assumes that the spatial behavior of the vector
potential field changes more slowly than the matter field, the interaction Hamiltonian
becomes
 ⎡ e ⎣
ĤI ↓ ĤD = d 3 r ν̂ + (r, t) − Â(0, t) · (−i∇) ν̂(r, t). (7.127)
m

By performing the following expansion for the matter field


ν̂(r, t) = ĉβ e−iβ t/ πβ (r), (7.128)
β

 2
where the πβ (r) = √r|β→ are the eigenfunctions of ĥ = − 2m ∧ 2 + U(r), and the
following expansion for the radiation field

  ⎡ ⎣
+ −i(k·r−θk t)
Â(r, t) = âks ei(k·r−θk t) + âks e εks , (7.129)
20 θk V
ks
7.6 Matter-Radiation Interaction 163

we finally obtain

Ĥmatt = β ĉβ+ ĉβ , (7.130)
β

+
Ĥrad = θk âks âks , (7.131)
ks
 ⎜ ⎝
ĤD =  gβεks ĉε+ ĉβ âks e−iθk t + âks
+ iθk t
e ei(ε −β )t/ (7.132)
βεks

where
 
e 1
gβεks =− εks · d 3 r π∗ε (r)(−i∇)πβ (r). (7.133)
m 20 θk V

Notice that

d 3 r π∗ε (r)(−i∇)πβ (r) = √ε|p̂|β→ = i m θεβ √ε|r|β→, (7.134)

where θεβ = (ε − β )/ and consequently we can write

 ⎜ ⎝
ĤD =  gβεks ĉε+ ĉβ âks e−iθk t + âks
+ iθk t
e eiθεβ t (7.135)
βεks

with 
1
gβεks = −i θεβ εks · √ε|er|β→, (7.136)
20 θk V

and √ε|er|β→ the electric dipole moment of the electromagnetic transition between
the electronic states |ε→ and |β→.

7.6.1 Cavity Quantum Electrodynamics

The development of lasers has produced a new field of research in quantum optics:
cavity quantum electrodynamics (CQED). This is the study of the interaction between
laser light confined in a reflective cavity and atoms or other particles which are inside
the cavity.
The simplest system of CQED is a laser light interacting with a two-level atom.
This system can be described by the Hamiltonian (7.123) with Eqs. (7.130)–(7.132)
but considering only one mode of the electromagnetic field and only two modes of
the matter field, namely
164 7 Second Quantization of Matter
⎡ ⎜ ⎝
Ĥ = θ â+ â + 1 ĉ1+ ĉ1 + 2 ĉ2+ ĉ2 + g ĉ2+ ĉ1 âe−i(θ−θ21 )t + â+ ei(θ+θ21 )t
⎜ ⎝⎣
+ ĉ1+ ĉ2 âe−i(θ+θ21 )t + â+ ei(θ−θ21 )t , (7.137)

+
where for simplicity we set θ = θk , θ21 = (2 − 1 )/, â = âks , â+ = âks and
g = g12 = g21 . If the applied electromagnetic radiation is near resonance with
an atomic transition, i.e. θ ↓ θ21 or equivalently (θ − θ21 ) ↓ 0, the previous
Hamiltonian can be approximated as
⎡ ⎜ ⎝
Ĥ = θ â+ â + 1 ĉ1+ ĉ1 + 2 ĉ2+ ĉ2 + g ĉ2+ ĉ1 â + â+ ei(θ+θ21 )t
⎜ ⎝⎣
+ ĉ1+ ĉ2 â e−i(θ+θ21 )t + â+ . (7.138)

Moreover, within the so-called rotating-wave approximation, one can neglect the
remaining time-dependent terms which oscillate rapidly, obtaining
⎡ ⎣
Ĥ = θ â+ â + 1 ĉ1+ ĉ1 + 2 ĉ2+ ĉ2 + g ĉ2+ ĉ1 â + ĉ1+ ĉ2 â+ . (7.139)

This is the Jaynes-Cummings Hamiltonian, originally proposed in 1963 by Edwin

Jaynes and Fred Cummings. This model Hamiltonian can be considered the
drosophila of quantum optics: it contains enough physics to describe many phe-
nomena in CQED and atom optics.
Let us analyze some properties of the Jaynes-Cummings Hamiltonian. The
photonic Fock state |n→ is eigenstate of the bosonic number operator â+ â. The photon
can be in any state |n→ or in a superposition of them, namely
∈

|photon→ = βn |n→, (7.140)
n=0

⎞
where βn are complex coefficients such that ∈ n=0 |βn | = 1. In addition, the
2

electronic Fock state |g→ = |11 → is eigenstate of the fermionic number operator
ĉ1+ ĉ1 , while the electronic Fock state |e→ = |12 → is eigenstate of the fermionic num-
ber operator ĉ1+ ĉ1 . In this two-level atom the electron can be in the state |g→ or in the
state |e→, or in a superposition of both, namely

|electron→ = β|g→ + ε|e→, (7.141)

where β and ε are two complex coefficients such that |β|2 + |ε|2 = 1 and obviously

ĉ1+ ĉ1 + ĉ2+ ĉ2 = 1. (7.142)

7.6 Matter-Radiation Interaction 165

It is then useful to introduce the following pseudo-spin operators

Ŝ+ = ĉ2+ ĉ1 , (7.143)

Ŝ− = ĉ1+ ĉ2 , (7.144)
1 + 
Ŝz = ĉ ĉ2 − ĉ1+ ĉ1 , (7.145)
2 2
which have these remarkable properties

Ŝ+ |g→ = |e→, Ŝ+ |e→ = 0, (7.146)

Ŝ− |g→ = 0, Ŝ− |e→ = |g→, (7.147)
1
Ŝz |g→ = − |g→, Ŝz |e→ = 21 |e→. (7.148)
2
By using these pseudo-spin operators the Jaynes-Cummings Hamiltonian of Eq.
(7.139) becomes
⎜ ⎝
Ĥ = θ â+ â + θ21 Ŝz + g Ŝ+ â + Ŝ− â+ + ¯, (7.149)

where ¯ = (1 + 2 )/2 is a constant energy shift. Due to the deep analogy between
two-level electronic system and two-spin states it is quite natural to set


0
| ↓ → = |g→ = |11 → = , (7.150)
1


1
| ↑ → = |e→ = |12 → = . (7.151)
0

A generic Fock state of the system is then given by

|n, ↓ → = |n→ ≥ | ↓ → (7.152)

or
|n, ↑ → = |n→ ≥ | ↑ →. (7.153)

We rewrite the Hamiltonian (7.149) neglecting the unrelevant energy shift ¯ as follows

Ĥ = Ĥ0 + ĤI (7.154)

where

Ĥ0 = θ â+ â + θ21 Ŝz (7.155)

⎜ ⎝
ĤI = g Ŝ+ â + Ŝ− â+ . (7.156)
166 7 Second Quantization of Matter

Except for the state |0, ↓ →, any other eigenstate of Ĥ0 belongs to one of the infinite
set of two-dimensional subspaces {|n − 1, ↑ →, |n, ↓ →}. Thus, for a fixed and finite
value of n, the infinite matrix representing the Jaynes-Cummings Hamiltonian Ĥ on
the basis of eigenstates of Ĥ0 splis into an infinite set of 2 × 2 matrices

∀ 
θ n +∀θ21 g n
Ĥn = , (7.157)
g n θ n − θ21

because
∀
Ŝ+ â|n − 1, ↑ → = n − 1 Ŝ+ |n − 1, ↑ → = 0, (7.158)
∀ ∀
Ŝ+ â|n, ↓ → = n Ŝ+ |n − 1, ↓ → = n |n − 1, ↑ →, (7.159)
∀ ∀
Ŝ− â+ |n − 1, ↑ → = n Ŝ− |n, ↑ → = n |n, ↓ →, (7.160)
∀
Ŝ− â+ |n, ↓ → = n + 1 Ŝ− |n + 1, ↓ → = 0. (7.161)

The matrix Ĥn can be easily diagonalized yeldings the following eigenvalues

1 1
En(±) = (n − )θ ±  (7.162)
2 2
where ⎤
= λ 2 + 42 g2 n, (7.163)

is the energy splitting with

λ = (θ21 − θ) (7.164)

the detuning between the atomic (electronic) transition frequency θ21 and the photon
frequency θ. The corresponding eigenstates are instead given by

|un(+) → = cos (φ)|n, ↓ → − sin (φ)|n − 1, ↑ → (7.165)

|un(−) → = sin (φ)|n, ↓ → + cos (φ)|n − 1, ↑ →, (7.166)

where 
1 − λ/
φ = arctan . (7.167)
1 + λ/

(±)
The states |un →, known as “dressed states”, were introduced by Claude
Cohen-Tannoudji and Serge Haroche in 1968–1969. These two scientists got the
Nobel Prize in Physics (Cohen-Tannoudji in 1997 and Haroche in 2012) for their
studies on the manipulation of atoms with lasers.
7.7 Bosons in a Double-Well Potential 167

7.7 Bosons in a Double-Well Potential

In this last section we consider a quite interesting physical problem: interacting

bosons in a quasi-1D double-well potential. The starting point is the quantum-field-
theory Hamiltonian
  
+ 2 2
Ĥ = d r ν̂ (r) − ∧ + U(r) ν̂(r)
3
2m

1
+ d 3 r d 3 r⊗ ν̂ + (r) ν̂ + (r⊗ ) V (r − r⊗ ) ν̂(r⊗ ) ν̂(r), (7.168)
2

where the external trapping potential is given by

1
U(r) = VDW (x) + mθ⊥ (y + z2 ),
2 2
(7.169)
2
that is a generic double-well potential VDW (x) in the x axial direction and a harmonic
potential in the transverse (y, z) plane. We assume that the system of bosons, described
by the field operator ν̂(r), is dilute and approximate the inter-particle potential with
a contact Fermi pseudo-potential, namely

V (r − r⊗ ) = g λ(r − r⊗ ), (7.170)

with g the strength of the interaction. If the frequency θ⊥ is sufficiently large, the
system is quasi-1D and the bosonic field operator can be written as

e−(y +z )/(2l⊥ )
2 2 2

ν̂(r) = π̂(x) . (7.171)

σ 1/2 l⊥

We are thus supposing that in the transverse (y, z) plane the system is Bose-Einstein
condensed into the transverse single-particle ground-state, which is a Gaussian wave-
function of width 

l⊥ = , (7.172)
mθ⊥

that is the characteristic length of the harmonic confinement. Inserting Eq. (7.171)
into the Hamiltonian (7.168) and integrating over y and z variables we obtain the
effective 1D Hamiltonian
  
+ 2 d 2
Ĥ = dx π̂ (x) − + VDW (x) + θ⊥ π̂(x)
2m dx 2

g1D
+ dx π̂+ (x) π̂+ (x) π̂(x) π̂(x), (7.173)
2
168 7 Second Quantization of Matter

where g
g1D = 2
(7.174)
2σl⊥

is the effective 1D interaction strength. Notice that in the one-body part of the Hamil-
tonian it appears the single-particle ground-state energy θ⊥ of the harmonic con-
finement. We suppose that the barrier of the double-well potential VDW (x), with its
maximum located at x = 0, is quite high such that there several doublets of quasi-
degenerate single-particle energy levels. Moreover, we suppose that only the low-
est doublet (i.e. the single-particle ground-state and the single-particle first excited
state) is occupied by bosons. Under these assumptions we can write the bosonic field
operator as
π̂(x) = âL πL (x) + âR πR (x) (7.175)

that is the so-called two-mode approximation, where πL (x) and πR (x) are
single-particle wavefunctions localized respectively on the left well and on the right
well of the double-well potential. These wavefunctions (which can be taken real) are
linear combinations of the even (and postitive) wavefunction π0 (x) of the ground
state and the odd (and positive for x < 0) wavefunction π1 (x) of the first excited
state, namely

1
πL (x) = ∀ (π0 (x) + π1 (x)) , (7.176)
2
1
πR (x) = ∀ (π0 (x) − π1 (x)) . (7.177)
2

Clearly the operator âj annihilates a boson in the jth site (well) while the operator
âj+ creates a boson in the jth site (j = L, R). Inserting the two-mode approximation
of the bosonic field operator in the effective 1D Hamiltonian, we get the following
two-site Hamiltonian
UL UR
Ĥ = L N̂L +R N̂R −JLR âL+ âR −JRL âR+ âL + N̂L (N̂L −1)+ N̂R (N̂R −1), (7.178)
2 2

where N̂j = âj+ âj is the number operator of the jth site,
  
2 d 2
j = dx πj (x) − + VDW (x) + θ⊥ πj (x) (7.179)
2m dx 2

is the kinetic plus potential energy on the site j,

  
2 d 2
Jij = dx πi (x) − + VDW (x) + θ⊥ πj (x), (7.180)
2m dx 2

is the hopping energy (tunneling energy) between the site i and the site j, and
7.7 Bosons in a Double-Well Potential 169

Uj = g1D dx πj (x)4 (7.181)

is the interaction energy on the site j. The Hamiltonian (7.178) is the two-site Bose-
Hubbard Hamiltonian, named after John Hubbard introduced a similar model in 1963
to describe fermions (electrons) on a periodic lattice. If the double-well potential
VDW (x) is fully symmetric then L = R = , JLR = JRL = J, UL = UR = U, and
the Bose-Hubbard Hamiltonian becomes
⎜ ⎝   U⎡ ⎣
Ĥ =  N̂L + N̂R − J âL+ âR + âR+ âL + N̂L (N̂L − 1) + N̂R (N̂R − 1) .
2
(7.182)
The Heisenberg equation of motion of the operator âj is given by

d ρ Ĥ
i âj = [âj , Ĥ] = + , (7.183)
dt ρ âj

from which we obtain

d 1
i âj =  âj − J âi + U(N̂j − )âj , (7.184)
dt 2
where j = L, R and i = R, L. By averaging this equation with the coherent state
|βL βR → = |βL → ≥ |βR →, such that âj (t)|βj → = βj (t)|βj →, we find

d 1
i βj =  βj − Jβi + U(|βj |2 − )βj , (7.185)
dt 2
where
βj (t) = N̄j (t) eiφj (t) , (7.186)

with N̄j (t) the average number of bosons in the site j at time t and φj (t) the corre-
sponding phase angle at the same time t. Working with a fixed number of bosons, i.e.

N = N̄L (t) + N̄R (t), (7.187)

and introducing population imbalance

N̄L (t) − N̄R (t)

z(t) = (7.188)
N
and phase difference
φ(t) = φR (t) − φL (t), (7.189)
170 7 Second Quantization of Matter

the time-dependent equations for βL (t) and βR (t) can be re-written as follows

dz 2J
=− 1 − z2 sin (φ), (7.190)
dt 
dφ 2J z U
= ∀ cos (φ) + z. (7.191)
dt  1 − z2 

These are the so-called Josephson equations of the Bose-Einstein condensate. Indeed,
under the condition of small population imbalance (|z|  1), small on-site interaction
energy (|U|z/  1) and small phase difference (φ  1) one finds

dz 2J
=− sin (φ), (7.192)
dt 
dφ 2J
= z, (7.193)
dt 
which are the equations introduced in 1962 by Brian Josephson to describe the
superconducting electric current (made of quasi-bosonic Cooper pairs of electrons)
between two superconductors separated by a thin insulating barrier (Nobel Prize in
1973). Remarkably, as shown in 2007 at Heidelberg by the experimental group of
Markus Oberthaler, Eqs. (7.190) and (7.191) describe accurately also the dynamics
of a Bose-Einstein condensate made of alkali-metal atoms confined in the quasi-1D
double-well potential by counter-propagating laser beams.

7.7.1 Analytical Results with N = 1 and N = 2

In this subsection we discuss the two-site Bose-Hubbard Hamiltonian
 
  U UR
Ĥ = −J âL+ âR + âR+ âL + N̂L (N̂L − 1) + N̂R (N̂R − 1) , (7.194)
2 2

in the case of N = 1 and N = 2 bosons. Note that, without loss of generality, we

have removed the constant energy N̂.
The ground-state |GS→ of the system with N bosons can be written as


N
|GS→ = ci |i, N − i→, (7.195)
i=0

where |i, N − i→ = |i→ ≥ |N − i→ is the state with i bosons in the left well and N − i
bosons in the right well and ci is its probability amplitude such that


N
|ci |2 = 1. (7.196)
i=0
7.7 Bosons in a Double-Well Potential 171

In the case of N = 1 bosons the ground-state is simply

|GS→ = c0 |0, 1→ + c1 |1, 0→, (7.197)

where
|c0 |2 + |c1 |2 = 1. (7.198)

Moreover, the on-site energy is identically zero and consequently it is immediate

to find
1
c0 = c1 = ∀ (7.199)
2

for any value of J and U.

In the case of N = 2 bosons the ground-state is instead given by

|GS→ = c0 |0, 2→ + c1 |1, 1→ + c2 |2, 0→, (7.200)

where
|c0 |2 + |c1 |2 + |c2 |2 = 1. (7.201)

By direct diagonalization of the corresponding Hamiltonian matrix, one finds

2
c0 = c2 = ⎤ (7.202)
16 + δ 2 + δ δ 2 + 16

and
δ+ δ 2 + 16
c1 = c0 ∀ , (7.203)
2 2

where
U
δ= (7.204)
J
is the effective adimensional interaction strength which controls the nature of the
ground state |GS→. Thus, the ground-state is strongly dependent on the ratio between
the on-site energy U and the tunneling (hopping) energy J. In particular, one
deduces that
⎦
⎧ |1, 1→ ⎝ for δ ∇ +∈
⎨ ⎜ ∀
|GS→ = 2
1
|0, 2→ + 2|1, 1→ + |2, 0→ for δ ∇ 0 . (7.205)
⎧
⎩ ∀1 (|0, 2→ + |2, 0→) for δ ∇ −∈
2
172 7 Second Quantization of Matter

7.8 Solved Problems

Problem 7.1
Consider the operator N̂ = fˆ + fˆ , where fˆ and fˆ + satisfy the anti-commutation rule
fˆ fˆ + + fˆ + fˆ = 1. Show that if |n→ is an eigenstate of N̂ with eigenvalue n then f + |n→
is eigenstates of N̂ with eigenvalue 1 − n.
Solution
We have
N̂ fˆ |n→ = (fˆ + fˆ )fˆ |n→.

The anti-commutation relation between fˆ and fˆ + can be written as

fˆ fˆ + = 1 − fˆ + fˆ .

This implies that

N̂ fˆ + |n→ = (fˆ + fˆ )fˆ + |n→ = fˆ + (fˆ fˆ + )|n→ = fˆ + (1 − fˆ + fˆ )|n→

= fˆ + (1 − N̂)|n→ = (1 − n)fˆ + |n→.

Notice that in Sect. 7.2 we have shown that the operator N̂ of fermions has only
eigenvalues 0 and 1. This means that fˆ + |0→ = |1→ and fˆ + |1→ = 0.
Problem 7.2
Show that the following equation of motion
 
ρ 2 2
i ν(r, t) = − ∧ + U(r) ν(r, t) + g |ν(r, t)|2 ν(r, t)
ρt 2m

of the classical Schrödinger field ν(r, t), that is the time-dependent Gross-Pitaevskii
equation, can be deduced by extremizing the action

S= dtd 3 r L,

where  
∗ ρ 2 2 1
L = ν (r, t) i + ∧ − U(r) ν(r, t) − g |ν(r, t)|4
ρt 2m 2

is the Lagrangian density of the system.

7.8 Solved Problems 173

Solution
First of all, we observe that
 
∇ · ν ∗ ∇ν = ν ∗ ∧ 2 ν + ∇ν ∗ · ∇ν.

We can then write

  
∗ 2
 ∗ 
d rν ∧ ν=
3
d r ∇ · ν ∇ν −
3
d 3 r ∇ν ∗ · ∇ν
V
 V
 V
 
= d 2 r ν ∗ ∇ν · n − d 3 r ∇ν ∗ · ∇ν,
S V

where S is the surface of the domain of volume V . In the limit of a very large V we
can suppose that the Schrödinger field and its derivatives are zero on the surface S,
and consequently  
d 3 r ν∗ ∧ 2 ν = − d 3 r ∇ν ∗ · ∇ν.
V V

This means that the Lagrangian density can be rewritten as

ρν 2 1
L = iν ∗ − |∧ν|2 − U(r)|ν|2 − g|ν|4 ,
ρt 2m 2

where |∧ν|2 = ∇ν ∗ · ∇ν. In this way the Lagrangian density is a function of the
Schrödinger field and only its first derivatives, namely

ρν
L = L(ν, ν ∗ , , ∇ν, ∇ν).
ρt

The Schrödinger field ν(r, t) = νR (r, t) + iνI (r, t) is complex, depending on the
two real fields νR (r, t) and νI (r, t) which can be varied independently, but in this
way the complex conjugate field ν ∗ (r, t) = νR (r, t) − iνI (r, t) is fully determined.
Alternatively, one can consider ν(r, t) and ν ∗ (r, t) as independent fields: this is our
choice and the action S is then a functional of ν(r, t) and ν(r, t). The first variation
of the action S gives


λS λS ∗
λS = dtd 3 r λν + λν ,
λν λν

where
λS ρL ρ ρL ρL
= − ρν
−∇· ,
λν ρν ρt ρ ρ∇ν
ρt
λS ρL ρ ρL ρL
∗
= ∗
− ∗ −∇ · .
λν ρν ρt ρ ρν ρ∇ν ∗
ρt
174 7 Second Quantization of Matter

Imposing that
λS = 0

we obtain two Euler-Lagrange equations

ρL ρ ρL ρL
− −∇· = 0,
ρν ρt ρ ρν ρ∇ν
ρt
ρL ρ ρL ρL
∗
− ∗ −∇ · = 0.
ρν ρt ρ ρν ρ∇ν ∗
ρt

By using our Lagrangian density the first Euler-Lagrange equation gives

 
ρ 2 2
i ν = − ∧ + U(r) ν + g |ν|2 ν,
ρt 2m

while the second Euler-Lagrange equation gives

 
ρ ∗ 2 2
−i ν = − ∧ + U(r) ν ∗ + g |ν|2 ν ∗ ,
ρt 2m

that is the complex conjugate of the previous one.

To conclude, we observe that if the Lagrangian density depends also on higher spa-
tial derivatives the Euler-Lagrange equations are modified accordingly. For instance,
in the presence of terms like ∧ 2 ν or ∧ 2 ν ∗ one gets

ρL ρ ρL ρL ρL
− ρν
−∇· + ∧2 = 0,
ρν ρt ρ ρ∇ν ρ∧ 2 ν
ρt
ρL ρ ρL ρL ρL
∗
− ∗ −∇ · ∗
+ ∧2 = 0.
ρν ρt ρ ρν ρ∇ν ρ∧ 2 ν ∗
ρt

Problem 7.3
Derive the Gross-Pitaevskii equation
 
ρ 2 2
i ν0 (r, t) = − ∧ + U(r) ν0 (r, t) + gN̄ |ν0 (r, t)|2 ν0 (r, t)
ρt 2m

of the classical field ν0 (r, t) from the Heisenberg equation of motion

 
ρ 2 2
i ν̂(r, t) = − ∧ + U(r) ν̂(r, t) + g ν̂ + (r, t)ν̂(r, t)ν̂(r, t)
ρt 2m

of the bosonic quantum field operator ν̂(r, t).

7.8 Solved Problems 175

Suggestion: expand the quantum field operator as


ν̂(r, t) = ĉβ πβ (r, t)
β

and use the coherent state |β0 → of ĉ0 .

Solution
First of all we notice that
ρ ρ
√β0 |i ν̂(r, t)|β0 → = i √β0 |ν̂(r, t)|β0 →
ρt ρt
ρ 
= i √β0 | ĉβ πβ (r, t)|β0 →
ρt β
ρ 
= i πβ (r, t)√β0 |ĉβ |β0 →
ρt β
ρ 
= i πβ (r, t)β0 λβ,β0
ρt β
ρ
= β0 i π0 (r, t).
ρt
Similarly
   
2 2 2 2
√β0 | − ∧ + U(r) ν̂(r, t)|β0 → = − ∧ + U(r) √β0 |ν̂(r, t)|β0 →
2m 2m
 
2 2 
= − ∧ + U(r) √β0 | ĉβ πβ (r, t)|β0 →
2m β
 
2 2 
= − ∧ + U(r) πβ (r, t)√β0 |ĉβ |β0 →
2m β
 
2 2 
= − ∧ + U(r) πβ (r, t)β0 λβ,β0
2m β
 
2 2
= β0 − ∧ + U(r) π0 (r, t).
2m

Moreover
  
√β0 |ν̂ + (r, t)ν̂(r, t)ν̂(r, t)|β0 → = √β0 | ĉβ+ π∗β (r, t) ĉε πε (r, t) ĉ∂ π∂ (r, t)|β0 →
β ε ∂

= π∗β (r, t)πε (r, t)π∂ (r, t)√β0 |ĉβ+ ĉε ĉ∂ |β0 →
βε∂
176 7 Second Quantization of Matter

= π∗β (r, t)πε (r, t)π∂ (r, t)|β0 |2 β0 λβ,0 λε,0 λ∂,0
βε∂

= |β0 |2 β0 |π0 (r, t)|2 π0 (r, t).

Finally, after setting N̄ = |β0 |2 , we conclude that

 
ρ 2
√β0 |i ν̂(r, t)|β0 → = √β0 | − ∧ 2 + U(r) ν̂(r, t)
ρt 2m
+ g ν̂ + (r, t)ν̂(r, t)ν̂(r, t)|β0 →

is exactly
 
ρ 2 2
i ν0 (r, t) = − ∧ + U(r) ν0 (r, t) + gN̄ |ν0 (r, t)|2 ν0 (r, t).
ρt 2m

Problem 7.4
Calculate the Brezin integral 
d ∂d∂
¯ e−∂A∂
¯
,

where ∂ and ∂¯ are Grassmann variables, and A is a complex number.

Solution
First of all we notice that, due to the fact that we are working with Grasmann numbers,
one finds
e−∂A∂
¯
= 1 − ∂A∂,
¯

because all the other terms of the Taylor-MacLaurin expansion are identically zero.
Consequently  
d ∂d∂
¯ e−∂A∂
¯
= d ∂d∂
¯ (1 − ∂A∂)
¯ .

Finally, on the basis of the definition of the Brezin integral where the integration is
defined as equivalent to differentiation, we get

ρ2
d ∂d∂
¯ (1 − ∂A∂)
¯ = (1 − ∂A∂)
¯ = A.
ρ ∂ρ∂
¯

Further Reading

For the second quantization of particles and non-relativistic quantum field theory:
N. Nagaosa, Quantum Field Theory in Condensed Matter Physics, Chap. 1, Sects.
1.1 and 1.2 (Springer, Berlin, 1999)
7.8 Solved Problems 177

A.L. Fetter, J.D. Walecka, Quantum Theory of Many-Particle Systems, Chap. 1,

Sects. 1.1 and 1.2 (Dover Publications, New York, 2003)
H.T.C. Stoof, K.B. Gubbels, D.B.M. Dickerscheid, Ultracold Quantum Fields, Chap.
6, Sects. 6.1, 6.2, 6.3, and 6.4 (Springer, Berlin, 2009)
A. Atland, B. Simons, Condensed Matter Field Theory, Chap. 2, Sect. 2.1 (Cambridge
University Press, Cambridge, 2006)
For the quantum properties of bosons in a double-well potential:
G. Mazzarella, L. Salasnich, A. Parola, F. Toigo, Coherence and entanglement in the
ground-state of a bosonic Josephson junction:from macroscopic Schrdinger cats to
separable Fock states. Phys. Rev. A 83, 053607 (2011)
Appendix A
Dirac Delta Function

In 1880 the self-taught electrical scientist Oliver Heaviside introduced the following
function

1 for x > 0
(x) = (A.1)
0 for x < 0

which is now called Heaviside step function. This is a discontinous function, with
a discontinuity of first kind (jump) at x = 0, which is often used in the context of
the analysis of electric signals. Moreover, it is important to stress that the Haviside
step function appears also in the context of quantum statistical physics. In fact, the
Fermi-Dirac function (or Fermi-Dirac distribution)

1
Fν (x) = , (A.2)
eν x +1

proposed in 1926 by Enrico Fermi and Paul Dirac to describe the quantum statistical
distribution of electrons in metals, where ν = 1/(k B T ) is the inverse of the absolute
temperature T (with k B the Boltzmann constant) and x = ρ − μ is the energy ρ of
the electron with respect to the chemical potential μ, becomes the function (−x)
in the limit of very small temperature T , namely

0 for x > 0
lim Fν (x) = (−x) = . (A.3)
ν∧+∞ 1 for x < 0

Inspired by the work of Heaviside, with the purpose of describing an extremely

localized charge density, in 1930 Paul Dirac investigated the following “function”

+∞ for x = 0
π(x) = (A.4)
0 for x ∇= 0

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 179
DOI: 10.1007/978-3-319-05179-6, © Springer International Publishing Switzerland 2014
180 Appendix A: Dirac Delta Function

imposing that  +∞
π(x) d x = 1. (A.5)
−∞

Unfortunately, this property of π(x) is not compatible with the definition (A.4). In
fact, from Eq. (A.4) it follows that the integral must be equal to zero. In other words,
it does not exist a function π(x) which satisfies both Eqs. (A.4) and (A.5). Dirac
suggested that a way to circumvent this problem is to interpret the integral of Eq.
(A.5) as
 +∞  +∞
π(x) d x = lim πρ (x) d x, (A.6)
−∞ ρ∧0+ −∞

where πρ (x) is a generic function of both x and ρ such that


+∞ for x = 0
lim πρ (x) = , (A.7)
ρ∧0+ 0 for x ∇= 0
 +∞
πρ (x) d x = 1. (A.8)
−∞

Thus, the Dirac delta function π(x) is a “generalized function” (but, strictly-speaking,
not a function) which satisfy Eqs. (A.4) and (A.5) with the caveat that the integral
in Eq. (A.5) must be interpreted according to Eq. (A.6) where the functions πρ (x)
satisfy Eqs. (A.7) and (A.8).
There are infinite functions πρ (x) which satisfy Eqs. (A.7) and (A.8). Among them
there is, for instance, the following Gaussian

1
πρ (x) = → e−x /ρ ,
2 2
(A.9)
ρ β

which clearly satisfies Eq. (A.7) and whose integral is equal to 1 for any value of ρ.
Another example is the function
1
πρ (x) = ρ for |x| ∀ ρ/2 , (A.10)
0 for |x| > ρ/2

which again satisfies Eq. (A.7) and whose integral is equal to 1 for any value of ρ.
In the following we shall use Eq. (A.10) to study the properties of the Dirac delta
function.
According to the approach of Dirac, the integral involving π(x) must be interpreted
as the limit of the corresponding integral involving πρ (x), namely
 +∞  +∞
π(x) f (x) d x = lim πρ (x) f (x) d x, (A.11)
−∞ ρ∧0+ −∞
Appendix A: Dirac Delta Function 181

for any function f (x). It is then easy to prove that

 +∞
π(x) f (x) d x = f (0). (A.12)
−∞

by using Eq. (A.10) and the mean value theorem. Similarly one finds
 +∞
π(x − c) f (x) d x = f (c). (A.13)
−∞

Several other properties of the Dirac delta function π(x) follow from its definition.
In particular

π(−x) = π(x), (A.14)

1
π(a x) = π(x) with a ∇= 0, (A.15)
|a|
 1
π( f (x)) = √
π(x − xi ) with f (xi ) = 0. (A.16)
| f (xi )|
i

Up to now we have considered the Dirac delta function π(x) with only one variable
x. It is not difficult to define a Dirac delta function π (D) (r) in the case of a D-
dimensional domain R D , where r = (x1 , x2 , ..., x D ) ⊗ R D is a D-dimensional
vector:

(D) +∞ for r = 0
π (r) = (A.17)
0 for r ∇= 0

and 
π (D) (r) d D r = 1. (A.18)
RD

Notice that sometimes π (D) (r) is written using the simpler notation π(r). Clearly,
also in this case one must interpret the integral of Eq. (A.18) as
 
(D)
π (r) d r = lim
D
πρ(D) (r) d D r, (A.19)
RD ρ∧0+ R D

(D)
where πρ (r) is a generic function of both r and ρ such that

+∞ for r = 0
lim πρ(D) (r) = , (A.20)
ρ∧0+ 0 for r ∇= 0

lim πρ(D) (r) d D r = 1. (A.21)
ρ∧0+
182 Appendix A: Dirac Delta Function

Several properties of π(x) remain valid also for π (D) (r). Nevertheless, some
properties of π (D) (r) depend on the space dimension D. For instance, one can prove
the remarkable formula
 1 2
(D) 2β ∞ (ln |r|)   for D = 2 ,
π (r) = (A.22)
− D(D−2)V
1
D
∞ 2 1
|r| D−2
for D ∈ 3

∂2 ∂2 ∂2
where ∞ 2 = ∂x12
+ ∂x22
+···+ ∂x 2D
and VD = β D/2 / (1 + D/2) is the volume of
a D-dimensional ipersphere of unitary radius, with (x) the Euler Gamma function.
In the case D = 3 the previous formula becomes

1 2 1
π (3) (r) = − ∞ , (A.23)
4β |r|

which can be used to transform the Gauss law of electromagnetism from its integral
form to its differential form.
Appendix B
Fourier Transform

It was known from the times of Archimedes that, in some cases, the infinite sum
of decreasing numbers can produce a finite result. But it was only in 1593 that the
mathematician Francois Viete gave the first example of a function, f (x) = 1/(1−x),
written as the infinite sum of power functions. This function is nothing else than the
geometric series, given by

 ∞
1
= xn, for |x| < 1. (B.1)
1−x
n=0

In 1714 Brook Taylor suggested that any real function f (x) which is infinitely
differentiable in x0 and sufficiently regular can be written as a series of powers, i.e.
∞

f (x) = cn (x − x0 )n , (B.2)
n=0

where the coefficients cn are given by

1 (n)
cn = f (x0 ), (B.3)
n!

with f (n) (x) the n-th derivative of the function f (x). The series (B.2) is now called
Taylor series and becomes the so-called Maclaurin series if x0 = 0. Clearly, the
geometric series (B.1) is nothing else than the Maclaurin series, where cn = 1. We
observe that it is quite easy to prove the Taylor series: it is sufficient to suppose that
Eq. (B.2) is valid and then to derive the coefficients cn by calculating the derivatives
of f (x) at x = x0 ; in this way one gets Eq. (B.3).
In 1807 Jean Baptiste Joseph Fourier, who was interested on wave propagation
and periodic phenomena, found that any sufficiently regular real function function
f (x) which is periodic, i.e. such that

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 183
DOI: 10.1007/978-3-319-05179-6, © Springer International Publishing Switzerland 2014
184 Appendix B: Fourier Transform

f (x + L) = f (x), (B.4)

where L is the periodicity, can be written as the infinite sum of sinusoidal functions,
namely
∞
  
a0  2β 2β
f (x) = + an cos n x + bn sin n x , (B.5)
2 L L
n=1

where
 L/2 
2 2β
an = f (y) cos n y dy, (B.6)
L −L/2 L
 L/2 
2 2β
bn = f (y) sin n y dy. (B.7)
L −L/2 L

It is quite easy to prove also the series (B.5), which is now called Fourier series. In fact,
it is sufficient to suppose that Eq. (B.5) is valid and then to  derive the coefficients

an and bn by multiplying both side of Eq. (B.5) by cos n 2β L x and cos n 2β L x
respectively and integrating over one period L; in this way one gets Eqs. (B.6) and
(B.7).
It is important to stress that, in general, the real variable x of the function f (x)
can represent a space coordinate but also a time coordinate. In the former case L
gives the spatial periodicity and 2β/L is the wavenumber, while in the latter case L
is the time periodicity and 2β/L the angular frequency.
Taking into account the Euler formula
 
2β 2β 2β
ein L x = cos n x + i sin n x (B.8)
L L
→
with i = −1 the imaginary unit, Fourier observed that his series (B.5) can be
re-written in the very elegant form

+∞
 2β
f (x) = f n ein L x , (B.9)
n=−∞

where 
1 L/2 2β
fn = f (y) e−in L y
dy (B.10)
L −L/2

are complex coefficients, with f 0 = a0 /2, f n = (an − ibn )/2 if n > 0 and f n =
(a−n + ib−n )/2 if n < 0, thus f n∓ = f −n .
The complex representation (B.9) suggests that the function f (x) can be periodic
but complex, i.e. such that f : R ∧ C. Moreover, one can consider the limit
L ∧ +∞ of infinite periodicity, i.e. a function which is not periodic. In this limit
Appendix B: Fourier Transform 185

Eq. (B.9) becomes the so-called Fourier integral (or Fourier anti-transform)
 +∞
1
f (x) = f˜(k) eikx dk (B.11)
2β −∞

with  ∞
f˜(k) = f (y) e−iky dy (B.12)
−∞

the Fourier transform of f (x). To prove Eqs. (B.11) and (B.12) we write Eq. (B.9)
taking into account Eq. (B.10) and we find

+∞   L/2

1 2β 2β
f (x) = f (y) e−in L y dy ein L x . (B.13)
n=−∞
L −L/2

Setting

2β 2β
kn = n and k = kn+1 − kn = (B.14)
L L
the previous expression of f (x) becomes

+∞  L/2
1  −ikn y
f (x) = f (y) e dy eikn x k. (B.15)
2β n=−∞ −L/2

In the limit L ∧ +∞ one has k ∧ dk, kn ∧ k and consequently

1
 +∞  +∞
−iky
f (x) = f (y) e dy eikx dk, (B.16)
2β −∞ −∞

which gives exactly Eqs. (B.11) and (B.12). Note, however, that one gets the same
result (B.16) if the Fourier integral and its Fourier transform are defined multiplying
them respectively with a generic constant and its inverse. Thus, we have found that
any sufficiently regular complex function f (x) of real variable x which is globally
integrable, i.e. such that
 +∞
| f (x)| d x < +∞, (B.17)
−∞

can be considered as the (infinite) superposition of complex monocromatic waves

eikx . The amplitude f˜(k) of the monocromatic wave eikx is the Fourier transform of
f (x).
186 Appendix B: Fourier Transform

f (x) F [ f (x)](k)
0 0
1 2βπ(k)
π(x) 1
(x) ik + β π(k)
1

eik0 x 2β π(k − k0 )
→
e−x /(2a ) a 2βe−a k /2
2 2 2 2

e −a|x| 2a
a 2 +k 2
2
sgn(x) ik
β
sin (k0 x) i [π(k − k0 ) − π(k + k0 )]
cos (k0 x) β [π(k − k0 ) + π(k + k0 )]

Table: Fourier transforms F[ f (x)](k) of simple functions f (x), where π(x) is the
Dirac delta function, sgn(x) is the sign function, and (x) is the Heaviside step
function.
The Fourier transform f˜(k) of a function f (x) is sometimes denoted as
F[ f (x)](k), namely
 ∞
f˜(k) = F[ f (x)](k) = f (x) e−ikx d x. (B.18)
−∞

The Fourier transform F[ f (x)](k) has many interesting properties. For instance, due
to the linearity of the integral the Fourier transform is clearly a linear map:

F[a f (x) + b g(x)](k) = a F[ f (x)](k) + b F[g(x)](k). (B.19)

Moreover, one finds immediately that

F[ f (x − a)](k) = e−ika F[ f (x)](k), (B.20)

F[e f (x)](k) = F[ f (x)](k − k0 ).
ik0 x
(B.21)
F[x f (x)](k) = i f˜√ (k), (B.22)
F[ f (n) (x)](k) = (ik)n f˜(k), (B.23)

where f (n) (x) is the n-th derivative of f (x) with respect to x.

In the Table we report the Fourier transforms F[ f (x)](k) of some elementary
functions f (x), including the Dirac delta function π(x) and the Heaviside step func-
tion (x). We insert also the sign function sgn(x) defined as: sgn(x) = 1 for x > 0
and sgn(x) = −1 for x < 0. The table of Fourier transforms clearly shows that
the Fourier transform localizes functions which is delocalized, while it delocalizes
functions which are localized. In fact, the Fourier transform of a constant is a Dirac
delta function while the Fourier transform of a Dirac delta function is a constant. In
general, it holds the following uncertainty theorem
Appendix B: Fourier Transform 187

1
x k ∈ , (B.24)
2
where   2
∞ ∞
x 2 = x 2 | f (x)|2 d x − x | f (x)|2 d x (B.25)
−∞ −∞

and   2
∞ ∞
k =
2
k | f˜(k)|2 dk −
2
k | f˜(k)|2 dk (B.26)
−∞ −∞

are the spreads of the wavepackets respectively in the space x and in the dual space
k. This theorem is nothing else than the uncertainty principle of quantum mechanics
formulated by Werner Heisenberg in 1927, where x is the position and k is the
wavenumber. Another interesting and intuitive relationship is the Parseval identity,
given by
 +∞  +∞
| f (x)| d x =
2
| f˜(k)|2 dk. (B.27)
−∞ −∞

It is important to stress that the power series, the Fourier series, and the Fourier
integral are special cases of the quite general expansion

f (x) = f ε λε (x) dε (B.28)

of a generic function f (x) in terms of a set λε (x) of basis functions spanned by the
parameter ε, which can be a discrete or a continuous variable. A large part of modern
mathematical analysis is devoted to the study of Eq. (B.28) and its generalization.
The Fourier transform is often used in electronics. In that field of research the
signal of amplitude f depends on time t, i.e. f = f (t). In this case the dual variable
of time t is the frequency θ and the fourier integral is usually written as
 +∞
1
f (t) = f˜(θ) e−iθt dk (B.29)
2β −∞

with
 ∞
f˜(θ) = F[ f (t)](θ) = f (t) eiθt dt (B.30)
−∞

the Fourier transform of f (t). Clearly, the function f (t) can be seen as the Fourier
anti-transform of f˜(θ), in symbols

f (t) = F −1 [ f˜(θ)](t) = F −1 [F[ f (t)](θ)](t), (B.31)

188 Appendix B: Fourier Transform

which obviously means that the composition F −1 ↑ F gives the identity.

More generally, if the signal f depends on the 3 spatial coordinates r = (x, y, z)
and time t, i.e. f = f (r, t), one can introduce Fourier transforms from r to k, from
t to θ, or both. In this latter case one obviously obtains

1
f (r, t) = f˜(k, θ) ei(k·x−θt) d 3 k dθ (B.32)
(2β)4 R4

with

f˜(k, θ) = F[ f (r, t)](k, θ) = f (k, t) e−i(k·r−θt) d 3 r dt. (B.33)
R4

Also in this general case the function f (r, t) can be seen as the Fourier anti-transform
of f˜(k, θ), in symbols

f (r, t) = F −1 [ f˜(k, θ)](r, t) = F −1 [F[ f (r, t)](k, θ)](r, t). (B.34)

Appendix C
Laplace Transform

The Laplace transform is an integral transformation, similar but not equal to the
Fourier transform, introduced in 1737 by Leonard Euler and independently in 1782
by Pierre-Simon de Laplace. Nowaday the Laplace transform is mainly used to solve
non-homogeneous ordinary differential equations with constant coefficients.
Given a sufficiently regular function f (t) of time t, the Laplace transform of f (t)
is the function F(s) such that
 +∞
F(s) = f (t) e−st dt, (C.1)
0

where s is a complex number. Usually the integral converges if the real part Re(s) of
the complex number s is greater than critical real number xc , which is called abscissa
of convergence and unfortunately depends on f (t). The Laplace transform F(s) of
a function f (t) is sometimes denoted as L[ f (t)](s), namely
 ∞
F(s) = L[ f (t)](s) = f (t) e−st dt. (C.2)
0

For the sake of completeness and clarity, we write also the Fourier transform f˜(θ),
denoted as F[ f (t)](θ), of the same function
 ∞
f˜(θ) = F[ f (t)](θ) = f (t) eiθt dt. (C.3)
−∞

First of all we notice that the Laplace transform depends on the behavior of f (t) for
non negative values of t, while the Fourier transform depends also on the behavior
of f (t) for negative values of t. This is however not a big problem, because we can
set f (t) = 0 for t < 0 (or equivalently we can multiply f (t) by the Heaviside step
function (t)), and then the Fourier transform becomes

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 189
DOI: 10.1007/978-3-319-05179-6, © Springer International Publishing Switzerland 2014
190 Appendix C: Laplace Transform
 ∞
f˜(θ) = F[ f (t)](θ) = f (t) eiθt dt. (C.4)
0

Moreover, it is important to observe that, comparing Eq. (C.2) with Eq. (C.4), if both
F and L of f (t) exist, we obtain

F(s) = L[ f (t)](s) = F[ f (t)](is) = f˜(is), (C.5)

or equivalently

f˜(θ) = F[ f (t)](θ) = L[ f (t)](−iθ) = F(−iθ). (C.6)

Remember that θ is a real variable while s is a complex variable. Thus we have found
that for a generic function f (t), such that f (t) = 0 for t < 0, the Laplace transform
F(s) and the Fourier transform f˜(θ) are simply related to each other.
In the following Table we report the Laplace transforms L[ f (t)](s) of some
elementary functions f (t), including the Dirac delta function π(t) and the Heaviside
step function (t), forgetting about the possible problems of regularity.

f (t) L[ f (t)](s)
0 0
1
1 s
π(t − ∂ ) e−∂ s
e−∂ s
(t − ∂ ) s
n!
tn s n+1
e−at 1
s+a
e−a|t| 2a
a 2 −s 2
sin (at) a
s 2 +a 2
cos (at) s
s 2 +a 2

Table. Laplace transforms L[ f (t)](s) of simple functions f (t), where π(t) is the
Dirac delta function and (t) is the Heaviside step function, and ∂ > 0 and n
positive integer.
We now show that writing f (t) as the Fourier anti-transform of f˜(θ) one can
deduce the formula of the Laplace anti-transform of F(s). In fact, one has
 ∞
1
f (t) = F −1
[ f˜(θ)](t) = f˜(θ) e−iθt dθ. (C.7)
2β −∞

Because f˜(θ) = F(−iθ) one finds also

 ∞
1
f (t) = F(−iθ) e−iθt dθ. (C.8)
2β −∞
Appendix C: Laplace Transform 191

Using s = −iθ as integration variable, this integral representation becomes

 +i∞
1
f (t) = F(s) est ds, (C.9)
2β i −i∞

where the integration is now a contour integral along any path ψ in the complex plane
of the variable s, which starts at s = −i∞ and ends at s = i∞. What we have found
is exactly the Laplace anti-transform of F(s), in symbols
 +i∞
−1 1
f (t) = L [F(s)](t) = F(s) est ds. (C.10)
2β i −i∞

Remember that this function is such that f (t) = 0 for t < 0. The fact that the
function f (t) is the Laplace anti-transform of F(s) can be symbolized by

f (t) = L−1 [F(s)](t) = L−1 [L[ f (t)](s)](t), (C.11)

which means that the composition L−1 ↑ L gives the identity.

The Laplace transform L[ f (t)](s) has many interesting properties. For instance,
due to the linearity of the integral the Laplace transform is clearly a linear map:

L[a f (t) + b g(t)](s) = a L[ f (t)](s) + b L[g(t)](s). (C.12)

Moreover, one finds immediately that

L[ f (t − a)(t − a)](s) = e−as L[ f (t)](s), (C.13)

L[e f (t)](s) = L[ f (t)](s − a).
at
(C.14)

For the solution of non-homogeneous ordinary differential equations with constant

coefficients, the most important property of the Laplace transform is the following

L[ f (n) (t)](s) = s n L[ f (t)](s) − s n−1 f (0) − s n−2 f (1) (0) − · · · − s f (n−2) − f (n−1) (0)
(C.15)

where f (n) (t) is the n-th derivative of f (t) with respect to t. For instance, in the
simple cases n = 1 and n = 2 one has

L[ f √ (t)](s) = s F(s) − f (0), (C.16)

√√ √
L[ f (t)](s) = s F(s) − s f (0) − f (0),
2
(C.17)

by using F(s) = L[ f (t)](s). The proof of Eq. (C.15) is straightforward performing

integration by parts. Let us prove, for instance, Eq. (C.16):
192 Appendix C: Laplace Transform
 +∞  +∞
 +∞ d  −st
L[ f √ (t)](s) = f √ (t) e−st dt = f (t) e−st 0 − f (t) e dt
0 0 dt
 +∞
= − f (0) + s f (t) e−st dt = − f (0) + s L[ f (t)](s)
0
= − f (0) + s F(s). (C.18)

We now give a simple example of the Laplace method to solve ordinary differential
equations by considering the differential problem

f √ (t) + f (t) = 1 with f (0) = 2. (C.19)

We apply the Laplace transform to both sides of the differential problem

L[ f √ (t) + f (t)](s) = L[1](s) (C.20)

obtaining
1
s F(s) − 2 + F(s) = . (C.21)
s
This is now an algebraic problem with solution

1 2 1 1 2 1 1
F(s) = + = − + = + . (C.22)
s(s + 1) s + 1 s s+1 s+1 s s+1

Finally, we apply the Laplace anti-transform




1 1 1 1
f (t) = L−1 + (t) = L−1 (t) + L−1 (t). (C.23)
s s+1 s s+1

By using our Table of Laplace transforms we find immediately the solution

f (t) = 1 + e−t . (C.24)

The Laplace transform can be used to solve also integral equations. In fact, one
finds

 t 
1
L f (y) dy (s) = F(s), (C.25)
0 s

and more generally

 t 
I0
L f (y) g(t − y) dy (s) = + F(s) G(s), (C.26)
−∞ s
Appendix C: Laplace Transform 193

where  0
I0 = f (y) g(t − y) dy. (C.27)
−∞

Notice that the integral which appears in Eq. (C.25) is, by definition, the convolution
( f ∓ g)(t) of two functions, i.e.
 t
( f ∓ g)(t) = f (y) g(t − y) dy. (C.28)
0
Bibliography

1. F.T. Arecchi, Measurement of the statistical distribution of Gaussian and laser sources. Phys.
Rev. Lett. 15, 912 (1965)
2. A. Atland, B. Simons, Condensed Matter Field Theory (Cambridge Univ. Press, Cambridge,
2006)
3. V.B. Berestetskii, E.M. Lifshitz, L.P. Pitaevskii, Relativistic Quantum Theory, vol. 4 of Course
of Theoretical Physics (Pergamon Press, Oxford, 1971)
4. J.D. Bjorken, S.D. Drell, Relativistic Quantum Mechanics (McGraw-Hill, New York, 1964)
5. B.H. Bransden, C.J. Joachain, Physics of Atoms and Molecules (Prentice Hall, Upper Saddle
River, 2003)
6. G. Bressi, G. Carugno, R. Onofrio, G. Ruoso, Measurement of the Casimir force between
parallel metallic surfaces. Phys. Rev. Lett. 88, 041804 (2002)
7. C. Cohen-Tannoudji, B. Dui, F. Laloe, Quantum Machanics (Wiley, New York, 1991)
8. P.A.M. Dirac, The Principles of Quantum Mechanics (Oxford University Press, Oxford, 1982)
9. A.L. Fetter, J.D. Walecka, Quantum Theory of Many-Particle Systems (Dover Publications,
New York, 2003)
10. K. Huang, Statistical Mechanics (Wiley, New York, 1987)
11. S.K. Lamoreaux, Demonstration of the Casimir force in the 0.6 to 6 m range. Phys. Rev. Lett.
78, 5 (1997)
12. M. Le Bellac, A Short Introduction to Quantum Information and Quantum Computation (Cam-
bridge Univ. Press, Cambridge, 2006)
13. U. Leonhardt, Measuring the Quantum State of Light (Cambridge Univ. Press, Cambridge,
1997)
14. E. Lipparini, Modern Many-Particle Physics: Atomic Gases, Quantum Dots and Quantum
Fluids (World Scientific, Singapore, 2003)
15. F. Mandl, G. Shaw, Quantum Field Theory (Wiley, New York, 1984)
16. G. Mazzarella, L. Salasnich, A. Parola, F. Toigo, Coherence and entanglement in the ground-
state of a bosonic Josephson junction:from macroscopic Schrdinger cats to separable Fock
states. Phys. Rev. A 83, 053607 (2011)
17. N. Nagaosa, Quantum Field Theory in Condensed Matter Physics (Springer, Berlin, 1999)
18. W. Rindler, Introduction to Special Relativity (Oxford Univ. Press, Oxford, 1991)
19. R.W. Robinett, Quantum Mechanics: Classical Results, Modern Systems, and Visualized Exam-
ples (Oxford Univ. Press, Oxford, 2006)
20. M.O. Scully, M.S. Zubairy, Quantum Optics (Cambridge Univ. Press, Cambridge, 1997)

L. Salasnich, Quantum Physics of Light and Matter, UNITEXT for Physics, 195
DOI: 10.1007/978-3-319-05179-6, © Springer International Publishing Switzerland 2014