Light-Matter Interaction

Physics and Engineering at the Nano Scale

John Weiner
IFSC-CePOF-INOF, Universidade de São Paulo, São Paulo, São Carlos, Brazil

Frederico Nunes
Universidade Federal de Pernambuco, Recife and IFSC-CePOF-INOF, Universidade de São
Paulo, São Paulo, São Carlos, Brazil

1
To Annick and Marize for patience and encouragement.
Preface

Light-matter interaction is pervasive throughout the disciplines of optical and atomic

physics, condensed matter physics, and electrical engineering with frequency and
length scales extending over many orders of magnitude. Deep earth and sea com-
munications use frequencies of a few tens of Hz and X-ray imaging requires sources
oscillating at hundreds of petaHz (1015 s−1 ). Length scales range from thousands of
kilometers to a few hundred picometers. Although we cannot pretend to offer an ex-
haustive treatise on this vast subject over all these frequency and length scales, we do
propose to provide advanced undergraduates, graduate students and researchers from
diverse disciplines the principal tools required to understand and contribute to rapidly
advancing developments in light-matter interaction centered at optical frequencies and
length scales around a hundred nanometers.
After a historical synopsis of the principal ideas leading to our present understand-
ing of light and matter in Chapter 1, we enter the heart of the subject with a review
of classical electrodynamics in Chapter 2. The intent is to reacquaint the reader with
electric and magnetic force fields and their interactions with ponderable media through
Maxwell’s equations and the Lorentz force law. We emphasize here macroscopic quan-
tities of permittivity and permeability and, through the constitutive relations, po-
larization and magnetization fields. Space-propagating and surface-propagating wave
solutions to Maxwell’s equations are fundamental to understanding energy and mo-
mentum transport around and through nanoscale structured materials. The chapter
ends with a development of plane wave propagation in homogenous media and at di-
electric and metallic surfaces. This discussion lays the ground work for the comparison
between the waves and transmission lines in Chapter 4.
In addition to the main chapter we have prepared three “Complements” which
cover some key issues more thoroughly. These complements either extend the expo-
sition to a deeper level or furnish important details and digressions that may be of
specific interest to some readers but not to all. By perusing the chapters and comple-
ments the material may be studied at various levels and from various angles. The first
Complement, Energy Flow in Polarizable Matter, covers the time evolution of energy
flux when electromagnetic waves propagate through media with electric polarization.
We point out analogies between the behavior of classical fields in bulk matter with
the energy dynamics of reactive and dissipative circuits. The second Complement,
Macroscopic Polarization from Microscopic Polarizability, shows how the macroscopic
electric, polarization, and displacement fields can be related to microscopic atomic and
molecular properties. The Clausius-Mossotti equation that expresses the dielectric con-
stant of a material (a macroscopic property) in terms of the microscopic polarizability
of the constituent atoms or molecules is developed at the end of this complement.
The Classical Charge Oscillator as Dipole Antenna constitutes Complement 3. In it
viii Preface

we show how a “real” antenna can be built up from an array of oscillating charges
and how an array of macro-antennas can be used to concentrate the spatial direction
of emission or reception.
After these three Complements to Chapter 2 we devote Chapter 3 to a fairly
extensive discussion of surface waves at the interface between dielectrics and metals,
because they play such an important role in “plasmonic” structures and devices. In fact
this propagation can be expressed in terms of circuit and waveguide theory, familiar to
electrical engineers; and Chapter 4, Transmission Lines, Waveguides, and Equivalent
Circuits, establishes the correspondence between classical electromagnetics and circuit
properties such as capacitance, inductance, and impedance. Furthermore we illustrate
how waveguide modal analysis and impedance matching can be used to guide the
design of nanoscale optical devices.
On the atomic scale (a few hundred picometers) and at interaction energies less
than or comparable to the chemical bond, light-matter interaction can be very well un-
derstood through a semiclassical approach in which the light field is treated classically
and the atom quantally. We therefore retain the classical electrodynamics treatment
in Chapter 5, Radiation in Classical and Quantal Systems, while presenting a very
simple quantum atomic structure with dipole transitions among atomic and molecu-
lar internal states. We take a very down-to-earth wave mechanical approach to the
quantum description in order to bring out the analogies between classical light waves,
quantum matter waves, classical dipole radiation and atomic radiative emission.
In addition to the Complements we have included a number of Appendices that
provide some supplementary discussion of the analytical tools used to develop the
physics and engineering of light-matter interaction. Appendix A is a brief discussion of
systems of units in electricity and magnetism. Although the Système International (SI)
has now been almost universally adopted, it is still worthwhile to understand how this
system is related to others; what quantities and units can be chosen for “convenience”
and what are the universal constraints that all systems must respect. Appendix B
is a brief review of vector calculus that readers have probably already seen, but who
might find a little refresher discussion useful. Appendix C discusses how the important
differential operations of vector calculus can be recast in different coordinate systems.
Although the Cartesian system is usually the most familiar, spherical and cylindrical
coordinates are practically indispensable for frequently encountered problems. Much
of the book deals with harmonically oscillating fields, and Appendix D is succinct
review of the quite useful phasor representation of these fields. Finally Appendices E,
F, G present the properties of the special functions, Laguerre, Legendre, and Hermite
that are so commonly encountered in electrodynamics and quantum mechanics. These
Appendices are an integral part of the book, not just some “boiler plate” nailed on at
the end. Readers are strongly encouraged to pay as much attention to them as they
do to Chapters and Complements.
Most of the material in this book is not new. Excellent texts and treatises on
classical electrodynamics, physical optics, circuit theory, waveguide and transmission
line engineering, atomic physics and spectroscopy are readily available. The real aim of
this book is take the useful elements from these disciplines and to organize them into a
course of study applicable to light-matter interaction at the nanoscale. To the extent,
 Preface ix

for example that waveguide mode analysis and sound design practice in microwave
propagation informs the nature of light transmission around and through fabricated
nanostructures, it is relevant to the purposes of this book. Rugged, reliable laser sources
in the optical and near-infrared regime together with modern fabrication technologies
at the nanoscale have opened a new area of light-matter interaction to be explored.
This exploration is far from complete, but the present book is intended to serve as
point of entry and useful account of some of the principal features of this new terrain.
Acknowledgements
This work has benefited from discussions, suggestions and remarks of many people over
the past several years. In particular Ben-Hur Borges, Nader Engheta, Henri Lezec, Ma-
sud Mansuripur, and Euclydes Marega, Jr., have contributed invaluable insight. JW
was also the grateful recipient of a visiting fellowship at the Center for Nanoscale Sci-
ence and Technology (CNST) of the National Institute of Standards and Technology
(NIST) from 2009-20011. The CNST provided a uniquely stimulating human and ma-
terial infrastructure for studying fundamental questions of light-matter interaction at
the nanoscale. During the year 2011-2012 JW has also benefited from the hospitality of
the Instituto de Fı́sica de São Carlos (IFSC) of the Universidade de São Paulo (USP),
and within the institute the Centro de Pesquisas em Óptica e Fotônica (CePOF). Fi-
nancial support for this year was provided by a generous Visiting Researcher grant
from the Fondação de Amparo à Pesquisa do Estado de São Paulo (FAPESP). JW
would also particularly like to thank Vanderlei Bagnato for the years of collaborative
work and fruitful interactions over the past quarter century, studying atomic optical
cooling in general and “cold collisions” in particular. Research in light-matter interac-
tion at the atomic level has contributed significantly to a deeper appreciation of these
phenomena at the nanoscale. Finally JW thanks Euclydes Marega, Jr. for his generous
and warm hospitality during a year in which he shared much of his time, his office
space, and know-how, allowing the timely completion of this project. FN acknowledges
financial support from FAPESP during 2011-2012 that permitted him to work closely
with Ben-Hur Borges and JW at the São Carlos campus of USP. Scientific discussions
with Professor Borges together with his able technical assistance on computer pro-
gramming contributed greatly to the work presented here. FN also acknowledges the
Universidade Federal de Pernambuco (UFPE) for a year sabbatical that made his stay
in São Carlos possible and the hospitality of CePOF and the IFSC.
Contents

1 Historical Synopsis of Light-Matter Interaction 1

1.1 Light and Matter in Antiquity 1
1.2 Light and Matter in the European Renaissance 2
1.3 The Revolution Accelerates 3
1.4 One Scientific Revolution Spawns Another 4
1.5 Further Reading 6
2 Elements of Classical Field Theory 7
2.1 Introduction 7
2.2 Relations among Classical Field Quantities 7
2.3 Classical Fields in Matter 9
2.4 Maxwell’s Equations 9
2.5 Static Fields, Potentials, and Energy 11
2.6 Three Illustrative Applications 14
2.7 Dynamic Fields and Potentials 24
2.8 Dipole Radiation 27
2.9 Light Propagation in Dielectric and Conducting Media 31
2.10 Plane Electromagnetic Waves 34
2.11 Exercises 57
2.12 Further Reading 58
Complements 60
Complement A Energy Flow in Polarizable Matter 61
A.1 Poynting’s Theorem in Polarizable Material 61
A.2 Harmonically Driven Polarization Field 62
A.3 Drude-Lorentz Dispersion 63
Complement B Macroscopic Polarization from Microscopic Po-
larizability 70
B.1 Introduction 70
B.2 Electric Field inside a Material 70
B.3 Polarization and Polarizability 72
Complement C The Classical Charged Oscillator and the Dipole
Antenna 74
C.1 The Proto-antenna 74
C.2 Real Antennas 76
3 Surface Waves 81
3.1 Introduction 81
xii Contents

3.2 History of Electromagnetic Surface Waves 81

3.3 Plasmon Surface Waves at Optical Frequencies 82
3.4 Plasmon Surface Wave Dispersion 92
3.5 Energy Flux and Density at the Boundary 99
3.6 Plasmon Surface Waves and Waveguides 103
3.7 Surface Waves at a Dielectric Interface 106
3.8 Exercises 113
3.9 Further Reading 114
4 Transmission Lines, Waveguides, and Equivalent Circuits 115
4.1 Introduction 115
4.2 Elements of Conventional Circuit Theory 115
4.3 Transmission Lines 120
4.4 Special Termination Cases 128
4.5 Waveguides 131
4.6 Rectangular Waveguides 136
4.7 Cylindrical Waveguides 140
4.8 Networks of Transmission Lines and Waveguides 146
4.9 Nanostructures and Equivalent Circuits 155
4.10 Exercises 165
4.11 Further Reading 166
5 Radiation in Classical and Quantal Atoms 167
5.1 Introduction 167
5.2 Dipole Emission of an Atomic Electron 167
5.3 Radiative Damping and Electron Scattering 169
5.4 The Schrödinger equation for the hydrogen atom 170
5.5 State Energy and Angular Momentum 179
5.6 Real Orbitals 182
5.7 Interaction of Light with the Hydrogen Atom 183
5.8 The Fourth Quantum Number: Intrinsic Spin 194
5.9 Other simple quantum dipolar systems 194
5.10 Exercises 200
5.11 Further Reading 201
Complements 202
Complement D Classical Blackbody Radiation 203
D.1 Field Modes in a Cavity 203
D.2 Planck mode distribution 205
D.3 The Einstein A and B coefficients 207
Appendices 209
Appendix A Systems of Units in Electromagnetism 210
A.1 General Discussion of Units and Dimensions 210
A.2 Coulomb’s Law 211
A.3 Ampère’s Law 212
 Contents xiii

Appendix B Review of Vector Calculus 215

B.1 Vectors 215
B.2 Axioms of vector addition and scaler multiplication 217
B.3 Vector Multiplication 217
B.4 Vector Fields 221
B.5 Integral Theorems for Vector Fields 225
Appendix C Gradient, divergence and curl in cylindrical and
polar coordinates 227
C.1 The gradient in curvilinear coordinates 229
C.2 The divergence in curvilinear coordinates 230
C.3 The curl in curvilinear coordinates 231
C.4 Expressions for grad, div, curl in cylindrical and polar coordinates 231
Appendix D Properties of Phasors 235
D.1 Introduction 235
D.2 Application of Phasors to Circuit Analysis 236
Appendix E Properties of the Laguerre Functions 239
E.1 Generating function and recursion relations 239
E.2 Orthogonality and Normalization 240
E.3 Associated Laguerre Polynomials 240
Appendix F Properties of the Legendre functions 242
F.1 Generating Function 242
F.2 Recurrence Relations 243
F.3 Parity 244
F.4 Orthogonality and Normalization 244
Appendix G Properties of the Hermite polynomials 246
G.1 Generating Function and Recurrence Relations 246
G.2 Orthogonality and Normalization 247
1
Historical Synopsis of Light-Matter
Interaction

The phrase “light-matter interaction” covers a vast realm of physical phenomena from
classical to quantum electrodynamics, from black holes and neutron stars to to meso-
scopic plasmonics and nanophotonics to subatomic quantum objects. The term “inter-
action” implies that light and matter are distinct entities that influence one another
through some intermediate agent. The history of scientific inquiry from the earliest
times to the present day might be neatly summarized into three questions: what is
the nature of light itself, of matter itself, and of the interaction agent. We now know
from Einstein’s celebrated equation E = mc2 that light (E) and matter (m) are fun-
damentally manifestations of the same “thing,” related by a universal proportionality
constant, the square of the speed of light in vacuum (c2 ). Nevertheless under ambient
physical conditions normally found on earth, the distinction between light and matter
makes sense; their interaction meaningful and worth studying.

1.1 Light and Matter in Antiquity

In the 5th century BC Leucippus, a Greek philosopher from Miletus (now in Turkey),
founded the school of atomism in which the universe is composed of immutable, in-
destructible, indivisible atoms and the space through which they move, the void. His
best student was Democritus (460-370 BC) who elaborated the atomistic construct of
the universe, attributing natural phenomena to the motion of atoms and the diversity
of material objects to their shapes and interlocking structures. The most extensive ac-
count of the Leucippus-Democritus atomic theory appears in an extended epic poem,
De rerum natura (The nature of things) by Lucretius, a Roman, who lived much later
(99-55 BC). A contemporary of Democritus, the Greek philosopher Empedocles (490-
430 BC), proposed that the cosmos was composed of four elements: fire, air, water,
and earth. Like the atomist school, these elements were immutable and the diversity of
nature arose from their combinations. The dynamics of the combinations are effected
by two forces, repulsive and attractive, called love and strife, respectively. Empedocles
is also credited with proposing the first theory of light. His idea was that light particles
stream out of the eyes and contact material objects. Euclid (∼ 300 BC) assumed this
flux moved in straight lines and used the idea to explain some optical phenomena in
Euclid’s Optics, a very influential early treatise on optics. Euclid’s Optics in turn influ-
enced Claudius Ptolemy (90-168 AD), a Roman citizen living in Egypt, whose writing
on geocentric astronomy were considered definitive until the European Renaissance.
2 Historical Synopsis

1.2 Light and Matter in the European Renaissance

By the beginning of the 17th century, the certitude of received ideas was crumbling.
Earth as center of the universe and Europe as center of the earth was cast into doubt.
The Americas had been discovered by European explorers between 1492 and 1504, the
earth had been circumnavigated by 1522, and the Ptolemaic geocentric astronomical
system, had been effectively overthrown by the Copernican heliocentric revolution of
1543. The invention of the telescope in 1608 enabled Galileo to show that Jupiter’s
moons revolved around that planet, not the earth. Into this fluid situation stepped
René Descartes (1596-1650) with a new world view. Descartes proposed that the uni-
verse consisted only of matter and motion. Forces could only be propagated among
massive bodies by actual contact, and therefore the apparent space between celestial
bodies, the “void” of Democritus, was actually filled with a kind of very fine-grained
material medium or plenum. Light emission, reflection, refraction, and absorption were
all explained in terms of material flux. The notion of force “fields” and action at a
distance had no place in the Cartesian system of the universe. Descartes”s interpreta-
tion of refraction, however, was severely challenged by Pierre de Fermat (1601-1665)
who explained the deviation of light rays on the basis of the principle of least time.
Applying this principle, Fermat derived that the sines of the incident and refracted
angles are in constant ratio, essentially the equivalent of what we now commonly term
“Snell’s law”. Descartes also derived this law, but his interpretation of light as particle
flux required greater velocity in the denser medium whereas the principle of least time
imposed a slower velocity. Fermat’s principle is in accord with the modern expression
for the velocity of light, v = c/n where n, the index of refraction, is unity in free
space and greater than unity in material media. The next significant observation was
light “diffraction,” a term coined by Grimaldi (1618-1663) to describe the appearance
of light beyond the geometrical shadow boundary defined by the supposed rectilinear
motion of light-particle flux. Diffraction was also observed by Robert Hooke (1635-
1703) who conjectured that light was due to rapid vibratory motion of the very small
particles of which ordinary matter is composed. Furthermore Hooke had the brilliant
insight that light (still considered as a kind of matter flux) propagated outward from
the center of each tiny vibrating center in circular figures and that light “rays” were
trajectories at right angles to these circular figures. This view of light propagation
laid the foundation for the construction of wave fronts with which Hooke was able
to explain refraction. He also tried to interpret colors in terms of refraction, but his
color theory was challenged by Isaac Newton (1642-1727) who correctly interpreted
color as an intrinsic property of light and not a distortion of it due to refraction. Al-
though Hooke took the first steps toward a wave theory of light, it was Christiaan
Huygens (1629-1695) who put it on a firmer foundation by expressing refraction in
terms of the principle that each element of a wave-front may be regarded as the center
of a secondary disturbance giving rise to spherical waves. The wavefront at any later
time is the envelope of all such secondary wavelets. Later this principle was refined
and extended by Augustin-Jean Fresnel (1788-1827) to establish modern wave optics
based on the principle of Huygens-Fresnel. It successfully explains light intensity mod-
ulations due to diffraction. Throughout antiquity and into the 17th century, however,
light and matter were not considered to be intrinsically different. Light was simply
 The Revolution Accelerates 3

a manifestation of matter, either in linear flux, or in vibratory motion. Huygens also

discussed the phenomenon of double diffraction in Iceland crystal and interpreted it
in terms of the propagation of ordinary and “extraordinary” waves within the crystal.
These studies were later considered by Newton and led to the discovery of light polar-
ization. Newton believed that longitudinal, compressive waves could never account for
polarization and his arguments essentially laid to rest the wave interpretation of light
until it was revived by Thomas Young (1773-1829) who demonstrated interference in
the celebrated Young’s double slit experiment.
Meanwhile the atom theory of matter was being transformed from an antique spec-
ulative philosophical proposition to a working scientific hypothesis, refined, enlarged,
and tested by quantitative experiment. In 1643 Evangelista Torricelli (1608-1647) es-
tablished that the “air” of everyday experience exerted pressure and was therefore
a tangible, if rarified, material composed of particles in constant motion. In Eng-
land Newton published the Philosophiae Naturalis Principia Mathematica in 1687,
laying out his laws of motion and theory of gravitation supported by precise, quanti-
tative celestial mechanics. Although Newton did not like to admit it, his gravitational
force law did not require an intervening plenum between bodies with mass. Contrary
to Descartes, action-at-a-distance was tacitly if not overtly admitted. Robert Boyle
(1627-1691) established one of the fundamental gas laws, that the pressure was in-
versely proportional to volume at fixed temperature and helped the science of chem-
istry emerge from obscurantist alchemy by publishing a truly scientific treatise, The
Sceptical Chymist. His distinction between heterogeneous mixtures and homogenous
compounds laid the ground work for the truly revolutionary advances of the 18th
century.

1.3 The Revolution Accelerates

At the beginning of the 18th century the nature of light was still very much in doubt.
The notion, commonly attributed to Newton, that light consisted of beams of particles,
or “corpuscles,” traveling in straight lines through homogenous media, held sway. But
from 1801 to 1803 Thomas Young carried out well-designed diffraction experiments
that put the wave theory of light back in the race. Although Young’s experiments
(especially the double-slit experiment published in 1807) dealt a body blow to the
corpuscular theory, light waves were still considered to be vibratory motion in the
longitudinal direction, along the direction of propagation. As such they could not
account for polarization. In 1821 the French engineer Augustin-Jean Fresnel (1788-
1827) showed that polarization was consistent with the wave picture if the periodic
vibration was transverse to the direction of propagation. This finding removed the
principal remaining objection to the wave model of light. The dramatic Arago spot
demonstration, in which a bright central spot can be observed in the geometric shadow
of an opaque circular screen, laid to rest the corpuscular model forever (or so it seemed
at the time). At the same time that Young and Fresnel were correctly characterizing
the wave nature of light, Michael Faraday (1791-1867) was carrying out experiments in
electricity and magnetism. Faraday pictured magnetic influences acting on bodies not
in direct contact as lines of force. These lines, originating and terminating in closed
loops, were the beginnings of force fields acting on bodies through space with no actual
4 Historical Synopsis

physical contact. Together with Newton’s theory of gravitation these ideas elevated
action-at-a-distance to serious consideration. The next step was the great unification
of electric, magnetic, and optical phenomena effected by James Clerk Maxwell (1831-
1879). In 1865 Maxwell published A Dynamical Theory of the Electromagnetic Field in
which he set forth the proposition that light was in fact a transverse electromagnetic
wave. In modern form the four Maxwell equations and the Lorentz force law constitute
a unified classical theory of electricity, magnetism, and light.
While Young, Fresnel, and Maxwell were putting light, electricity, and magnetism
on a firm foundation, understanding of ponderable matter and chemical interaction was
also advancing rapidly. Building on Boyle’s experiments with gases, Joseph Priestly
(1733-1804) conducted extensive experiments on “airs” and identified nitric (NO) and
nitrous (N2 O) oxides and oxygen (O2 ), which he called “dephlogisticated air”. This
term refers to “phlogiston,” a substance that was thought to be contained within
matter and expelled during combustion. Although now discredited by modern un-
derstanding of oxidation, Priestly used the principle of phlogiston to rationalize his
observations. In France Joseph Louis Gay-Lussac (1778-1850) annunciated another
physical gas law, complementary to Boyle’s law. Gay-Lussac found that for a given
quantity of gas, the volume was directly proportional to the temperature at constant
pressure. Back in England John Dalton (1766-1844) determined that elements com-
bine in simple number ratios and began to determine atomic masses. Dalton also
determined that the pressure of a mixture of gases was equal to the sum of the pres-
sures of the individual constituents. But the major unifying advance in understanding
chemical reactivity was carried out by Antoine Lavoisier (1743-1794), considered the
father of modern chemistry. Lavoisier was the first to clearly annunciate that between
reactants and products of a chemical reaction mass was conserved. The recognition
of mass conservation as a fundamental principle distinguished light, now recognized
as an electromagnetic field wave, from ponderable matter. Finally Dimitri Mendeleev
(1834-1907) unified the growing list of elements into a rational order with the periodic
table in 1869. As the nineteenth century drew to a close, there were grounds for great
satisfaction. Although the inner structure of the atomic elements was not known, their
chemical identity and relationships to each other were firmly established. The periodic
table not only rationalized known elements but predicted those not yet discovered.
Oliver Lodge could write in 1888 “The whole subject of electrical radiation seems to
be working itself out splendidly.” There was only the somewhat perplexing null result
of the Michelson-Morley experiment in 1887 that failed to measure the velocity of
the “luminiferous aether” through which light was supposed to travel, and the fail-
ure of the equipartition of energy principle to account for the spectral distribution of
black-body radiation. But these were considered minor blemishes on a near perfect
masterwork of human understanding.

1.4 One Scientific Revolution Spawns Another

The null result of the Michelson-Morley experiment and its subsequent refinements
was finally explained by the theory of special relativity proposed in 1905 by Albert
Einstein (1879-1955). Special relativity showed that the speed of light in vacuum was
not just a property of the electromagnetic wave but a universal constant independent
 One Scientific Revolution Spawns Another 5

Fig. 1.1 Major developments in the history of light and matter.

of the inertial reference frame in which it is measured. It follows that any experiment
attempting to determine the luminiferous aether wind with respect to the earth’s mo-
tion must yield a null result. Furthermore, in a subsequent paper published later in the
same year, Einstein showed that special relativity implied the equivalence of matter
and energy through the famous formula, m = E/c2 . The “new philosophy” originat-
ing with Descartes and culminating in the achievements of Maxwell and Mendeleev
appeared to establish a fundamental distinction between light and matter. Within
a few decades of their master works, however, special relativity announced a funda-
mental equivalence between them. Furthermore, in a subsequent paper of the same
year, Einstein proposed the quantization of radiation that reestablished some of the
old Newtonian corpuscular properties of light. Energy quanta in the form of E = hν,
where ν is the frequency of light and h is Planck’s constant, also removed the “ultravio-
let catastrophe” from the Rayleigh-Jeans formulation of black-body radiation. Finally
6 Historical Synopsis

the quantum mechanics established by M. Born, E. Schrödinger, W. Heisenberg, and

P. A. M. Dirac in the first decades of the 20th century explained the periodic table of
the chemical elements in terms of the inner structure of atoms. Quantum mechanics
in its nonrelativistic form is quite adequate for an understanding of matter outside
the atomic nucleus. Together with Dirac’s initial formulation of a quantized version
of electromagnetism, the second revolution in physics was almost complete. In 1916
Einstein was able to generalize his theory of relativity and apply it to gravity. The
result was a new geometric conception of “space-time” relevant to cosmological energy
and length scales. Of the four known classes of forces, gravity, electromagnetism, the
strong and the weak nuclear forces, only two have been unified into a consistent the-
oretical framework: electromagnetism and the weak force. To date a quantum theory
of gravity does not exist and current efforts to unify the force classes into a “theory
of everything” using string theory or M-theory continue. However, for the purposes of
this book classical electrodynamics, essentially as formulated by the Maxwell-Lorentz
force laws, together with nonrelativistic quantum mechanics is perfectly adequate to
describe light-matter interactions.

1.5 Further Reading

1. E. T. Whittaker, A history of the theories of aether and electricity: from the age
of Descartes to the close of the nineteenth century, BiblioLIfe, Charleston, South
Carolina.
2. F. Wilczek, The lightness of being, Basic Books, New York, (2008).
3. S. Hawking and L. Mlodinow, The Grand Design, Bantam Books, New York,
(2010).
4. J. Gribbin, The Scientists, Random House, New York, (2004).
5. D. C. Lindberg, The Beginnings of Western Science, University of Chicago Press,
Chicago, (2007)
6. C. A. Pickover, Archimedes to Hawking, Oxford University Press, Oxford, (2008).
2
Elements of Classical Field Theory

2.1 Introduction
Before we can discuss the interaction of light and matter we have to establish a working
vocabulary for the vector fields of classical electrodynamics and the equations that
relate them. These equations, Maxwell’s equations, together with the Lorentz force
law serve as the postulates of that theory. A class of solutions to these equations
describes the propagation of plane electromagnetic waves, and we will explore this
propagation in vacuum, in dielectrics, and in good conductors. In complements to the
main chapter we will study polarization fields in bulk matter, the relation between the
macroscopic polarization field and the microscopic atomic polarizability, and finally
develop further the central role played by the radiating dipole in the construction of
antennas and antenna arrays. Later in a complement to Chapter 5 we will also describe
how standing waves are established in a cavity and how to calculate the energy density
by counting modes. The energy density is then applied to the problem of classical black-
body radiation theory, and we will describe how it leads to a fundamental contradiction
between the predictions of classical theory and experimental measurement.

2.2 Relations among Classical Field Quantities

Since virtually all students now learn electricity and magnetism with the SI (Système
International) or rationalized MKS (meters, kilograms, seconds) family of units, we
adopt it here. A discussion of units and fundamental quantities in electromagnetism is
presented in Complement A. There we will see that although RMKS is now the most
common choice, other units systems have their advantages and that in any case all
expressions can be readily transposed from one system to another.
The present choice of RMKS means that we write Coulomb’s force law between
two electric point charges q, q 0 separated by a distance r as
 0
1 qq
F= r (2.1)
4π0 r3
and Ampère’s force law (force per unit length) of magnetic induction between two
infinitely long wires carrying electric currents I, I 0 , separated by a distance r as
µ0 II 0
 
d |F|
= (2.2)
dl 2π r
where 0 and µ0 are called the permittivity of free space and the permeability of free
space, respectively. In this units system the permeability of free space is defined as
8 Elements Classical Field Theory

µ0
≡ 10−7 (2.3)
4π
and the numerical value of the permittivity of free space is fixed by the condition that
1
= c2 (2.4)
0 µ0
where c is the speed of light in vacuum. Therefore we must have
1
= 10−7 c2 (2.5)
4π0
The time-independent vector fields issuing from these force expressions are the electric
field, E, and the magnetic induction field, B. The E-field is the Coulomb force per
unit charge,
1 q
E= r (2.6)
4π0 r3
and the magnetic induction B-field is given by the Biot-Savart law,
I×r dl × r
Z Z
µ0 µ0
B= dl = I (2.7)
4π r3 4π r3
where I is the current running in a wire and dl is an element of the wire length. The
Lorentz force law succinctly summarizes the effect of the E- and B-fields on a charged
particle moving with velocity v,
F = q(E + v × B) (2.8)
In addition to these two fields, the displacement field D and magnetic field H are
needed to describe the modification of force fields in dielectric or conductive matter.
In linear, isotropic materials these two additional fields are linked to E and B by the
“constitutive relations”
D = E (2.9)
1
H= B (2.10)
µ
where  and µ are the permittivity and permeability of the material. These material
parameters are related to those of free space by
 = 0 r (2.11)
µ = µ0 µr (2.12)
where r and µr are the relative (and unitless) permittivity and permeability respec-
tively. The relative permittivity r is often called the dielectric constant, but the rel-
ative permeability rarely finds application and does not have a common alternative
name. Unfortunately many authors use  for the dielectric constant so there is danger
of confusion between permittivity (with MKS units of C 2 /N · m2 ) and the unitless
dielectric constant. However one can always discern from the context what is meant
by the symbol . There is also some difference of opinion among authors as to the most
appropriate nomenclature for the B-field and H-field. Here we follow conventional us-
age and call the B-field the “magnetic induction field” and the H-field the “magnetic
field”.
 Classical Fields in Matter 9

2.3 Classical Fields in Matter

In free space E, B, D, H are related by

D = 0 E (2.13)
B = µ0 H (2.14)

When these force fields act on a material medium, however, the D and H fields take
on added terms. Matter consists of positively charged core nuclei surrounded by dis-
tributions of negatively charged electrons. If the core nuclei are arranged according to
some symmetric spatial extension the material is crystalline; if not the material may
be a glassy solid, a liquid or a gas. The electric charge distribution may be bound to
the nuclei or delocalized throughout the crystal structure. In any case if E and B are
present, the the Lorentz forces acting on the electric charge distribution will clearly
modify it. These modifications can be characterized by the introduction of two new
fields, the polarization P and the magnetization M. In the presence of matter the
displacement and magnetic fields now are expressed as

D = 0 E + P (2.15)
B = µ0 H + M (2.16)

For E-and B-fields that are not too strong the P and M fields themselves are propor-
tional to E and H.

P = 0 χe E (2.17)
M = µ0 χm H (2.18)

where χe , χm are electric and magnetic susceptibility. Equations 2.15, 2.16 then become

D = 0 (1 + χe )E (2.19)
B = µ0 (1 + χm )H (2.20)

and from Eqs. 2.9, 2.10 we see that the relative permittivity and permeability can be
written in terms of the corresponding susceptibilities as

r = 1 + χ e (2.21)
µr = 1 + χm (2.22)

The relative permittivity and permeability are unitless but may be complex, the imag-
inary parts reflecting absorptive loss.

r = 0 + i00 = 1 + χ0e + iχ00e (2.23)

0 00
µr = µ + iµ = 1 + χ0m + iχ00m (2.24)

2.4 Maxwell’s Equations

In the foregoing we have presented the fields of classical electrodynamics, D, E, B, H,
and the fields induced in materials, P and M which represent the response of matter
10 Elements Classical Field Theory

to electric and magnetic forces. The equations governing the spatial and temporal
behavior of these fields are Maxwell’s equations,

∇ · D = ρfree (2.25)
∇·B=0 (2.26)
∂B
∇×E=− (2.27)
∂t
∂D
∇ × H = Jfree + (2.28)
dt
The first two equations make statements about field sources. The first states that the
source of the displacement field D is the “free” electric charge density ρfree . In fact
the charge density is composed of two terms: the free charge density and the “bound”
charge, ρbound .
ρ = ρfree + ρbound (2.29)
The bound charge density is defined as the negative divergence of the polarization
field,
ρbound = −∇ · P (2.30)
so that from Eqs. 2.25, 2.30 we see that
ρ
∇·E= (2.31)
0
The second source equation, Eq. 2.26, states that the magnetic induction field B does
not originate from a magnetic charge density, ρm . In fact Equation 2.26 implies that
magnetic source “charges” do not exist; and magnetic monopoles have never been
found in nature (although Dirac’s quantum electrodynamics implies that that must
exist somewhere or have existed at some time). The second pair of equations, termed
Faraday’s Law and the Maxwell-Ampère Law respectively, describe the spatial and
temporal behavior of the fields. Equation 2.28, Ampère’s Law, introduces another field,
the charge current density Jfree . Just as Eq. 2.29 expresses the total charge density
as the sum of the free and bound charge densities so the total current density J is
composed of the sum of free and bound current densities.

J = Jfree + Jbound (2.32)

The bound charge density is defined by two contributions: one from the polarization
field and the other from the magnetization field.
∂P 1
Jbound = + ∇×M (2.33)
∂t µ0
and consequently, by Eqs. 2.15, 2.16, 2.28, 2.33,
∂D
Jfree = ∇ × H − (2.34)
∂t
The term ∂D/∂t is called the displacement current.
 Static Fields, Potentials, and Energy 11

2.5 Static Fields, Potentials, and Energy

2.5.1 Electric Field Energy
The E-field resulting from a single point charge, Eq. 2.6, can be written
1 q
E= r̂ (2.35)
4π0 r2
where r is the distance from the charge and the radial unit vector r̂ = r/r. Written in
this way the familiar 1/r2 radial fall-off of the field is made explicit. Any vector field
is specified by the divergence and the curl of the field, and in the case of Eq. 2.35 the
divergence is
1
∇ · E = ρ(r) (2.36)
0
where ρ(r) is the charge density at the position r.
It can be easily shown by the application of Stokes’ theorem (see Appendix B for
a discussion of the differential and integral operations common in vector fields) to
Eq. 2.35 that the curl of E is always null.
∇×E=0 (2.37)
By the principal of vector field superposition, this property is true not just for the
E-field of a single point charge but for any spatial distribution of point charges. The
E-field is said to be irrotational and therefore can be set equal to the gradient of a
scaler function, say,
E = −∇V (2.38)
This characterization of the E-field as the gradient of a scaler potential V is justified
by the vector calculus identity that the curl of the gradient of any scalar function is
null.
∇ × E = −∇ × ∇V = 0 (2.39)
The negative sign on the right side of Eq. 2.38 is a convention. A units analysis shows
that V has units of energy per charge, and the work done to bring n charges from
infinity to a given spatial configuration is
n
1X
Eelec = qi V (ri ) (2.40)
2 i=1
The factor of 1/2 on the right side of Eq. 2.40 avoids double counting. This expression
can be generalized to a smooth charge density distribution ρ(r),
Z
1
Eelec = ρV dτ (2.41)
2
and taking into account Eqs. 2.36, 2.38, the energy required to assemble the charge
distribution can be expressed in terms of the resulting E-field,
Z
0
Eelec = E 2 dτ (2.42)
2
where the integration is over all space. Equation 2.42 can simply be interpreted as the
energy of the E-field.
12 Elements Classical Field Theory

2.5.2 Magnetic Field Energy

We know from Eq. 2.26 that the divergence of the B-field is null.

∇·B=0 (2.43)

but we find from the Biot-Savart law describing the magnetic field issuing from current
flowing in a wire, Eq. 2.7, that the curl of the B-field is given by

∇ × B = µ0 J (2.44)

where J is the distribution of the current densities giving rise to B. Just as we intro-
duced the scaler potential V (r) to characterize the curl-less E-field, so we can introduce
a vector potential A to characterize the divergence-less B-field.

B=∇×A (2.45)

and setting the divergence of the vector potential to zero, we find the standard ex-
pression for A is
J(r0 ) 0
Z
µ0
A(r) = dτ (2.46)
4π r
where r = |r − r0 |, the distance between the point r where A is evaluated and r0 the
position of the current density J. Just as it is possible to write the energy of some
distribution of electric charge as an integral over the product of the charge density
distribution ρ(r) and the scalar potential V (r) (Eq. 2.41), we can write the magnetic
energy as the integral over scaler product of the current density distribution J(r) and
the vector potential A(r) Z
1
Emag = J · A dτ (2.47)
2
By recognizing that J is proportional to the curl of B from Eq. 2.45, substituting in
Eq. 2.47 and integrating by parts, we can write the magnetic energy entirely in terms
of the B-field in an expression analogous to Eq. 2.42
Z
1
Emag = B 2 dτ (2.48)
2µ0

where the integral is taken over all space.

2.5.3 Poynting’s Theorem

From Eqs. 2.42, 2.48 it appears plausible that the energy of an electromagnetic field is
the sum of the energy of the constituent parts.
Z  
1 1 2
Eem = 0 E 2 + B dτ (2.49)
2 µ0

and this result can in fact be obtained directly from one of the basic postulates of
classical electrodynamics, the Lorentz force law,
 Static Fields, Potentials, and Energy 13

F = q(E + v × B) (2.50)

The force F acting on charged particles set in motion with velocity v does work on
the particles at the rate
dEmech
= qE · v (2.51)
dt
where Emech is the mechanical energy of the particle system. Of course the B-field,
always acting
R at right angles to the direction of motion, can do no work on the particles.
But q = ρdτ and v = Jfree /ρ so
Z
dEmech
= E · Jfree dτ (2.52)
dt

Now we can use the Ampère-Maxwell law, Eq. 2.28, to eliminate Jfree in Eq. 2.52. The
dot product in the integrand becomes

∂D
E · Jfree = E · (∇ × H) − E · (2.53)
∂t

The next step is to invoke a vector-field identity and the Faraday law, Eq. 2.27, to
rewrite the first term on the right of Eq. 2.53. The vector-field identity is

∇ · (E × H) = H · (∇ × E) − E · (∇ × H) (2.54)

and Faraday’s law is

∂B
∇×E=−
∂t
so
∂B
E · (∇ × H) = −H · − ∇ · (E × H) (2.55)
∂t
and
∂B ∂D
E · Jfree = −H · −E· − ∇ · (E × H) (2.56)
∂t ∂t
Now we substitute the E · Jfree expression back into the integrand in Eq. 2.52, and use
Stokes’ theorem to convert the volume integral of the ∇ · (E × H) term to a surface
integral with integrand E × H. The resulting expression for the time rate of work done
by the electromagnetic field on the charged particles is
Z   I
dEmech ∂B ∂D
=− H· +E· dτ − (E × H) · da (2.57)
dt ∂t ∂t σ

where the integral of the second term on the right is taken over the surface σ bounding
the volume integral in the first term. The two terms in the volume integral are identified
with the electric and magnetic parts of the electromagnetic field energy. In a free-space
14 Elements Classical Field Theory

volume or in any medium with negligible polarization the decrease in the field energy,
as work is being done on the system of particles, can be written,
Z  
dEem d 1 1 2
− =− B + 0 E 2 dτ (2.58)
dt dt 2 µ0

confirming that the total electromagnetic field energy can be identified as the sum
of the separate static magnetic and electric terms as in Eq. 2.49. We can now write
Eq. 2.57 as an energy conservation expression,
I
dEmech dEem
+ = − E × H · da (2.59)
dt dt
Where the term on the right is interpreted as an integral over the energy flux (energy
per unit area per unit time) flowing across the bounding surface. This cross product
expression is termed the Poynting vector, S.

S = (E × H) (2.60)

Notice that energy increasing within the volume is equal to a negatively signed energy
flux into the volume across the boundary surface. The negative sign on the Poynting
vector is simply a consequence of the sign convention on the vector surface differential
da, positive pointing outward.

2.6 Three Illustrative Applications

Gauss’s law plays a pivotal role in many of the applications we will study in the
following chapters. We write out in detail here the solutions to three problems the
results of which will be useful in subsequent topics. The integral form of Gauss’s law
is also developed at the end of this section.
Example 2.1 Estimate the electron oscillation frequency in a hydrogen atom, using
the differential form of Gauss’s law.

Solution 2.1
To solve this problem, let us consider the following model for the hydrogen atom. We
assume that the electron is represented by a cloud of charge density ρ, symmetrically
distributed in a sphere of radius R centered at the origin x = 0 where a positively
charged proton is located.
e e
ρ= = (2.61)
V 4/3πR2
The force of attraction between the proton and the electron charge distribution is
given by Coulomb’s law.
e2
F=− r̂ (2.62)
4π0 r3
Under the action of a constant external E-field, the cloud will be displaced from its
equilibrium position. Let us call x the displacement distance between charge centers at
 Three Illustrative Applications 15

Fig. 2.1 Displacement of the electron charge cloud under the influence of a constant external
E-field.

which a new balance of forces will occur. Figure 2.1 illustrates this displacement. We
seek to calculate the restoring force on the electron charge density using Gauss’s law.
This restoring force will be exerted by the charge fraction δq located at the surface
of the sphere with radius x. The fraction of charge outside the volume Vx = 4/3πx3
produces no force on the proton. Taking ρ as the electron charge density, we have
 x 3
δq = (ρ)(Vx ) = −e . (2.63)
R
The magnitude of the effective Coulomb force between the proton and the displaced
electron charge cloud is now

eδq e2 x3
|F| = 2
=− = −Kx (2.64)
4π0 x 4π0 x2 R3
which shows that small displacements result in a linear restoring force with the force
constant K.
e2
K= (2.65)
4π0 R3
Remembering that the frequency of oscillation is given by
r s
K e2
ω= = (2.66)
m 4π0 R3 me

and taking R = 0.529−10 m as the radius of the hydrogen atom in the ground state,
and me = 9.109 × 10−31 Kg, the rest mass of the electron, we calculate the charge
frequency of oscillation ωe = 4.13 × 1016 s−1 .
Example 2.2 Calculate the E-field of a three-dimensional static electric dipole from
the potential.
16 Elements Classical Field Theory

Solution 2.2

In Section 2.8 we studied the oscillating dipole and emphasized the time-dependence
of the field and radiation. Here we study the static dipole and illustrate the usefulness
of calculating the E-field by first writing down the scalar potential, then taking its
divergence.
A dipole is a distribution of two equal but opposite charges, separated by a distance
a as shown in Fig. 2.2. The dipole is a vector which by convention points from the

Fig. 2.2 Electric dipole in a system of polar and cartesian coordinates.

negative to the positive charge. In Fig. 2.2 the dipole p is oriented along z, p = qaˆ
z .
The calculation of the E-field is carried out through the electric potential using E =
−∇V (x, yz). Since the dipole is oriented along the z-axis and has cylindrical symmetry
around this axis, the dipole field is invariant with rotation about the azimuthal angle
ϕ. From Fig. 2.2, at point r we have for the potential,
 
q 1 1
V (x, y, z) = − (2.67)
4π0 r1 r2

Since we know from the law of cosines,

 Three Illustrative Applications 17

 a 2  a 2
r12 = r2 + − a r cos θ and r22 = r2 + + a r cos θ (2.68)
2 2
Eq. 2.67 takes the form
 
q  1 1
V (r, θ) = 1/2 −  (2.69)

4π0 
 1/2 
a 2 a 2



r2 + 2 − ar cos θ r2 + 2 + ar cos θ

Taking advantage of the dipole cylindrical symmetry, we calculate the electric field by
applying the gradient operation to the potential in polar coordinates,

E(r, θ) = Er ˆr + Eθ ˆθ (2.70)

with
∂V 1 ∂V
Er = − and Eθ = − (2.71)
∂r r ∂θ
Carrying out the derivative operations on Eq. 2.69 we find
 
a a
q  r− 2 cos θ r+ 2 cos θ
Er = 3/2 −  (2.72)

4π0  2
 3/2 
a 2 a 2



r + 2 − ar cos θ r2 + 2 + ar cos θ

and
 
a a
q  2 sin θ 2 sin θ
Eθ = 3/2 +  (2.73)

4π0 
 3/2 
a 2 a 2



r2 + 2 − ar cos θ r2 + 2 + ar cos θ

In many practical problems the dipole potential and field are of interest when r >> a,
the region of space termed the far field. Under these conditions we can write
  a 2 n   a 2  a  n
2 2n
r + ± a r cos θ = r 1+ ± cos θ
2 2r r
   
2n a 2 a
'r 1+n ± cos θ (2.74)
2r r
and therefore write the dipole potential Eq. 2.67 in the far field as
     
q 1 1  a 2  a  1  a 2  a 
V (r, θ) = 1− − cos θ − 1 − + cos θ
4π0 r 2 2r r 2 2r r
(2.75)
which, after simplification, yields
qa cos θ
V (r, θ) ' (2.76)
4π0 r2
and the far-field dipole E-field
18 Elements Classical Field Theory

Fig. 2.3 Relation between cartesian and polar coordinates for dipole oriented along z axis

2qa cos θ qa sin θ

E(r, θ) = ˆr + ˆθ (2.77)
4π0 r3 4π0 r3

We can write the dipole potential and E-field, Eqs. 2.76, 2.77, in vector form as

p · ˆr
V (r, θ) = (2.78)
4π0 r2

and
(2p · ˆr )ˆ
r − (p · ˆθ )ˆ
θ
E(r, θ) = 3
(2.79)
4π0 r
where ˆr , ˆθ are unit vectors in polar coordinates. As discussed at some length in
Complement C, the dipole vector E-field can be expressed either in polar or cartesian
coordinates. The unit vectors of the E-field in polar coordinates are ˆr , ˆθ , ˆϕ , and
the dipole p in Fig. 2.3 can be written in terms of the polar components as

p = (p · ˆr ) ˆr + (p · ˆθ ) ˆθ + (p · ˆϕ ) ˆϕ (2.80)

The relations between the unit vectors in polar coordinates and cartesian coordinates
are given by

ˆr = ˆx sin θ cos ϕ + ˆy sin θ sin ϕ + ˆz cos θ (2.81)
ˆθ = ˆx cos θ cos ϕ + ˆy cos θ sin ϕ − ˆz sin θ (2.82)
ˆϕ = −ˆ
x sin ϕ + ˆy cos ϕ (2.83)
 Three Illustrative Applications 19

Fig. 2.4 Dielectric sphere immersed in a uniform electric field oriented along the z direction.

By multiplying the first of these relations by cos θ and the second by − sin θ we can
easily show that
r − sin θˆ
cos θˆ θ = ˆz (2.84)
In Fig. 2.3 we see that in an r, θ plane, carrying out the dot products in Eq. 2.80 with
the help of Eqs. 2.81, 2.82, 2.83 results in

r − sin θˆ
p = p(cos θˆ θ ) (2.85)

and therefore

(p · ˆθ )ˆ
θ = −p sin θˆ
θ = p − (p cos θ)ˆ
r = p − (p · ˆr )ˆ
r (2.86)

Substituting this last expression in Eq. 2.77 results in

(3p · ˆr )ˆ
r − p
E(r, θ) = (2.87)
4π0 r3

Example 2.3 Analyze the polarization of a dielectric sphere of radius R immersed in

a uniform electric field with permittivity .

Solution 2.3
Consider a dielectric sphere placed in a uniform electric field E as shown in Fig. 2.4
We analyze the problem in polar coordinates. The potential along the z direction is
given by
V = −E · ẑ = −E0 r cos θ (2.88)
Our first goal is to determine the potential within and external to the sphere. To find
the potential we solve Laplace’s equation,

∇2 V = 0 (2.89)
20 Elements Classical Field Theory

in both regions and the join the solutions at the boundary. We show in Complement
C, Eq. C.38 how to express the Laplacian operator in polar coordinates. An equivalent,
slightly different form is
1 ∂2 ∂2V
 
1 ∂ ∂V 1
∇ · ∇V = (rV ) + · sin θ + (2.90)
r ∂r2 r2 sin θ ∂θ ∂θ r2 sin2 θ ∂ϕ2
The problem of the dielectric sphere subject to a constant electric field along the z axis
is obviously cylindrically symmetric. Before attacking this problem directly we examine
the form of the solutions to Laplace’s equation for spherical symmetry. Assuming for
the moment a spherically symmetric potential V , we write V as a product of a radial
function R(r)/r, a polar angular function Θ(θ), and an azimuthal angular function
Φ(ϕ).
R(r)
V (r, θ) = Θ(θ) Φ(ϕ) (2.91)
r
Now we substitute this form back into Eq. 2.90 and arrange all terms with r and θ
dependence on the left side and all terms with ϕ dependence on the right side. After
some algebra the result is
r2 sin2 θ ∂ 2 R(r) ∂ 2 Θ(θ) 1 ∂2Φ
 
sin θ ∂Θ(θ)
+ cos θ + sin θ = − (2.92)
R(r) ∂r2 Θ(θ) ∂θ ∂Θ2 (θ) Φ(ϕ) ∂ϕ2
Since r, θ, ϕ are independent variables the only way that Eq. 2.92 can be true is for
each side to be equal to a constant, called the “separation constant”. We can choose
the form of this constant to be anything we want, so we choose it to be m2 . From the
right side of Eq. 2.92 we then have
1 ∂2Φ
= −m2 (2.93)
Φ(ϕ) ∂ϕ2
and we can see by simple inspection and can verify by substitution that the solutions
are
Φ(ϕ) = e−imϕ (2.94)
In order for Φ to be a single-valued function in ϕ (modulo 2π), m must be an integer,
m = 0, ±1, ±2, . . .. After division by sin2 θ the left side of Eq. 2.92, also equal to m2 ,
can be rearranged to
r2 ∂ 2 R ∂ 2 Θ(θ) m2
 
1 ∂Θ(θ)
+ cos θ + sin θ − =0 (2.95)
R(r) ∂r2 sin θΘ(θ) ∂θ ∂θ2 sin2 θ
The first term in Eq. 2.95 depends only on r and the next two only on θ. Therefore
they can also be separated and set equal to a separation constant, the form of which
we choose to be l(l + 1).
r2 d2 R
= l(l + 1) (2.96)
R(r) dr2
d2 Θ(θ) m2
   
1 dΘ(θ)
− cos θ + sin θ − = l(l + 1) (2.97)
(sin θ)Θ(θ) dθ dθ2 sin2 θ
 Three Illustrative Applications 21

Table 2.1 Legendre Polynomials

Index l Function Polynomial

0 P0 (x) 1
1 P1 (x) x

2 P2 (x) 1/2 3x2 − 1

3 P3 (x) 1/2 5x3 − 3x

4 P4 (x) 1/8 35x4 − 30x2 + 3

Equation 2.97 can be made more concise by recognizing that the term in square brack-
ets is an expanded differential of a product of functions. Multiplying both sides by Θ
we have
m2
 
1 d dΘ
sin θ − Θ = −l(l + 1)Θ (2.98)
sin θ dθ dθ sin2 θ
The radial and angular differential equations can finally be rearranged to
d2 R l(l + 1)
− =0 (2.99)
dr2 r2
d2 Θ m2

dΘ
+ cot θ + l(l + 1) − Θ=0 (2.100)
dθ 2 dθ sin2 θ
We have already observed that m must be zero or an integer in order to maintain Φ as
a single-valued function. In this field-matching problem of a dielectric sphere immersed
in a constant electric field, oriented along z, the E-field will show no ϕ dependence,
and we can set m = 0. The resulting expression is called Legendre’s equation.
d2 Θ dΘ
+ cot θ + [l(l + 1)] Θ = 0 (2.101)
dθ2 dθ
Often the independent variable of the Legendre equation is taken to be x = cos θ
rather than θ itself. Then Eq. 2.101 takes the form
 
d
dΘ
1 − x2 + l(l + 1)Θ = 0 (2.102)
dx dx
The physically admissible solutions to Eq. 2.102 are a family of polynomials in cos θ
called the Legendre polynomials. They are labeled by the index l which must only
assume the positive integer values l = 0, 1, 2, . . .. The first few Legendre polynomials
are listed in Table 2.1 The Rodrigues’ formula is a concise expression of any desired
member l of the Legendre polynomials
1 dl
l
Pl (x) = l l
x2 − 1 (2.103)
2 l! dx
The solution to the separated radial equation, Eq. 2.99, is
R(r) = Arl + Br−(l−1) (2.104)
where A and B are as yet undetermined constants. That Eq. 2.104 really is a solution
can be verified by direct substitution.
22 Elements Classical Field Theory

Now we are in a position to write the general solution, Eq. 2.91, to Maxwell’s
equation governing the response of a dielectric sphere immersed in a constant electric
field.
∞ h
V (r, θ) X i
= U (r, θ, ϕ) = Arl + Br−(l+1) Pl (cos θ)e−imϕ (2.105)
r
l=0

Because the problem has cylindrical symmetry and the solution is invariant about the
z axis of the applied E-field, the m = 0 solution of Φ(ϕ) = e−imϕ is the appropriate
choice, and Φ(ϕ) = 1. Furthermore, we must fit the solution to the physical boundary
conditions at the origin, at infinity and at the interface on the surface of the sphere.
Since the potential must remain finite as r → 0, it is clear that we should choose B = 0
for r inside r0 , the radius of the sphere. In this region
∞
X
Uin (r, θ) = Al rl Pl (cos θ) (2.106)
l=0

Outside the sphere, in the limit r → ∞ we see from Eq. 2.88 that the asymptotic
solution must take on the form −E0 cos θ. In order to fit this result, A1 = −E0 , with
all other Al = 0. The solution outside the sphere must therefore have the form
∞
X
Uout (r, θ) = −E0 r cos θ + Bl r−(l+1) Pl (cos θ) (2.107)
l=0

In order to determine the remaining parameters Al and Bl we apply continuity con-

ditions at the sphere surface. The solutions must match there, and in order to match
smoothly the first derivatives must match has well. Therefore we have at r = r0
∞
X ∞
X
Al r0l Pl (cos θ) = −E0 r0 cos θ + Bl r−(l+1) Pl (cos θ) (2.108)
l=0 l=0

for the first continuity condition. For the second condition we remember that
∂U (r, θ)
Er = − (2.109)
∂θ
and that components of the D-field normal to the boundary must be continuous. Then
we have for the second continuity condition
∞ ∞
" #
(l−1) −(l+1)
X X
−in Al lr0 Pl (cos θ) = −out −E0 r0 cos θ + Bl r0 Pl (cos θ) (2.110)
l=0 l=0

where in , out are the permittivities inside and outside the sphere. Next we will elim-
inate all terms with summations in Eqs. 2.108, 2.110 except one by using the orthogo-
nality and normalization properties of the Legendre polynomials.
Z +1
2
Pl (cos θ)Pm (cos θ) d(cos θ) = δ(ml) (2.111)
−1 2m +1
 Three Illustrative Applications 23

The result is
   
2 2 −(m+1) 2
Am r0m = − E0 r0 + Bm r0 (2.112)
2m + 1 3 2m + 1

and
     
−(m+1) 2 out m + 1 2 out 1
Bm r0 1+ = E0 r0 1 − · (2.113)
2m + 1 in m 3 in m
or, h i

2

2 1 − out 1
·m
−(m+1) in
Bm r0 = E0 r0 h
i (2.114)
2m + 1 3 1 + out m+1in m

Substituting Eq. 2.114 into the right side of Eq. 2.112 results in the following expression
involving Am and r0 .
h i
out 1

2

2 2 1 − in · m
Am r0m = − E0 r0 + E0 r0 h
i (2.115)
2m + 1 3 3 1+ out m+1
in m

The continuity condition should hold for any choice of r0 , the radius of the sphere,
and we see that if we choose m = 1, then the expression for Am will be independent
of r0 . Then we find  
3out
A1 = −E0 (2.116)
in+2out
and from Eq. 2.114  
in − out
B1 = r03 E0 (2.117)
in + 2out
Now we can substitute A1 and B1 back into Eqs. 2.107, 2.106 to obtain the expressions
for the scalar potential U (r, θ) outside and inside the dielectric sphere.

r3
 
in − out
U (r, θ)out = −E0 r cos θ + E0 02 cos θ outside potential (2.118)
r in + 2out
 
3out
U (r, θ)in = −E0 r cos θ inside potential (2.119)
in + 2out
From Eqs. 2.118, 2.119 we can easily calculate the E-field outside and inside the sphere
which was the initial goal we set out to reach.
 
3out
Ein (r, θ) = E0 êz (2.120)
in + 2out

in − out r03

 
Eout (r, θ) = E0 ẑ + E0 (2 cos θ êr + sin θ êθ ) (2.121)
in + 2out r3
24 Elements Classical Field Theory

2.6.1 Integral Form of Gauss’s Law

Maxwell’s equations are usually summarized in their differential form, but they can be
cast in an integral form, often more useful for practical calculation. The integral form
of Eq. 2.31 is found by applying Gauss’s theorem of vector calculus (see Appendix B),
Z Z
∇ · K dτ = K · dσ (2.122)
V S
This theorem states that the integral of the divergence of a vector field K over a
volume is equal to the integral of the field itself over a surface enclosing the volume.
Applying this theorem to the differential form of Gauss’s law we have
Z Z Z
∇ · D dτ = 0 ∇ · E dτ = q = 0 E · dσ (2.123)
V V S
R
where q is the total charge enclosed by the volume, q = V ρ dτ . Figure 2.5 illustrates
the physical significance of the integral form of Gauss’s law.

Fig. 2.5 Charges q1 , q2 , q3 , . . . are point sources on which E-field lines originate. The integral
form of Gauss’s law states that the surface integral of E over the volume enclosing these
charges is equal to the total charge divided by the permittivity.

2.7 Dynamic Fields and Potentials

2.7.1 Maxwell’s Equations Revisited
In section 2.4 we wrote down Maxwell’s equations

ρfree ∂B
∇·E= ∇×E=− Faraday’s Law
0 ∂t
∂D
∇·B=0 ∇×H= +J Maxwell-Ampère Law
∂t
 Dynamic Fields and Potentials 25

In the case of static fields the E-field is irrotational, ∇ × E = 0, and the B-field
is “divergenceless”, ∇ · B = 0. Therefore the E-field can be specified by the gradient
of a scalar function, the scalar potential V and the B-field is determined by the curl
of a vector function, the vector potential A.

E = −∇V B=∇×A (2.124)

In the case of fields changing in time the E-field is no longer irrotational because
∂B ∂A
=∇×
∂t ∂t
and substituting into Faraday’s Law we have
∂A
∇ × E = −∇ ×
∂t
or  
∂A
∇× E+ =0 (2.125)
∂t
The quantity E + ∂A/∂t is itself irrotational and can therefore be written as the
gradient of a scaler function.
∂A
E+ = −∇V
∂t
or
∂A
E = −∇V − (2.126)
∂t
Then from Maxwell’s E-field divergence equation
 
∂A ρfree
∇ · E = ∇ · −∇V − =
∂t 0
or
∂ ρfree
∇2 V + (∇ · A) = − (2.127)
∂t 0
Substituting B = ∇ × A into the left side of the Maxwell-Ampère Law and taking the
time derivative of Eq. 2.126 we find

∂2A
 
∂V
∇ × (∇ × A) = µ0 J − µ0 0 ∇ − µ0 0 2
∂t ∂t

and using the vector field identity ∇ × (∇ × A) = ∇(∇ · A) − ∇2 A we can simplify to

∂2A
   
∂V
∇2 A − µ0 0 2 − ∇ ∇ · A + µ0 0 = −µ0 J (2.128)
∂t ∂t

Equations 2.127 and 2.128 are just a reformulation of the Maxwell relations in terms
of the time-dependent potentials rather than fields.
26 Elements Classical Field Theory

2.7.2 Lorentz Gauge

The vector potential A can be adjusted to a certain extent since the B-field only
depends on the curl of A, not A itself. In particular, since ∇ × (∇f ) = 0, where f
is some arbitrary scaler function, we can add the gradient of a scaler function to the
vector potential without changing its curl. The choice of what scaler function to use
is called the choice of gauge, and one particularly useful choice of scaler function is
called the Lorentz gauge. It is defined by
∂V
f = ∇ · A = −µ0 0 (2.129)
∂t
Substituting Eq. 2.129 into Eq. 2.128 we see that it simplifies to
∂2A
∇2 · A − µ0 0 = −µ0 J (2.130)
∂t2
and Eq. 2.127 becomes
∂2V ρfree
∇2 V − µ0 02
=− (2.131)
∂t 0
The strategy then, given the source terms J and ρ, is to solve the two inhomogeneous
wave equations, Eqs. 2.130, 2.131, for A and V . Then use Eqs. 2.124, 2.126 to find the
fields B and E.

2.7.3 Retarded Potentials

The points in space where the potentials are evaluated may be very far from the
location of the sources. Therefore as the sources J(r0 , t), ρ(r0 , t) vary with time, the
changes in the potentials at V (r, t), A(r, t) will only take effect at a later time t,
governed by the distance from the source position to the points in space where the
potentials are evaluated and the speed of light. The time that the changes actually
take place at the sources is called the “retarded time” tr , and the time at which these
changes take effect at the potential points is just denoted by t. The relation between
the two is tr = t − `/c, where ` is the distance from the source point to the field point,
` = |r − r0 |. The spatial relations between coordinate origin, source position and field
position is shown in Fig. 2.6. The formal solutions for distributed charge and current
density sources, ρ(r0 , tr ), J(r0 , tr ) are
ρ(r0 , tr ) 0
Z
1
V (r, t) = dτ (2.132)
4π0 `
and
J(r0 , tr ) 0
Z
µ0
A(r, t) = dτ (2.133)
4π `
Using these formal solutions for V and A we can obtain formal solutions to the fields
from Eqs. 2.126 and 2.124.
Z " 0 #
1 ρ(r , tr ) ˆ ρ̇(r0 , tr ) ˆ J̇(r0 , tr )
E(r, t) = `+ `− dτ 0 (2.134)
4π0 `2 c` c2 `
 Dipole Radiation 27

Fig. 2.6 Relations between coordinate origin O, position r0 of differential volume dτ 0 in

distributed source, and distance ` = |r − r0 | between differential source volume dτ 0 , located
at point S, and field point P .

and Z " #
µ0 J(r0 , tr ) J̇(r0 , tr ) ˆ 0
B(r, t) = + × `dτ (2.135)
4π `2 c`

In Eqs. 2.134, 2.135 J̇ denotes the time derivative of the source current density, and
thus we write Z
∂A µ0 J̇ 0
= dτ (2.136)
∂t 4π `
which is used in the third term on the right side of Eq. 2.134.

2.8 Dipole Radiation

In this section we use the results for the time-dependent retarded potentials developed
in section 2.7 to calculate the E-field and B-field solutions for a harmonically oscillating
dipole source. We take the dipole as two charges +q and −q placed at the ends of
very short conducting wire of length a. The charges oscillate along the wire with a
frequency ω such that the time dependence of the charge at any point along the wire
is q(t) = q0 cos(ωt). The electric dipole, aligned along the z axis, is then given by

p(t) = q(t)aẑ = q0 a cos(ωt)ẑ = p0 cos(ωt)ẑ (2.137)

The current i associated with this dipole is just

dq
i(t) = ẑ = −q0 ω sin ωtẑ (2.138)
dt
28 Elements Classical Field Theory

Fig. 2.7 Electric dipole oscillator aligned along z axis. The maximum length of the dipole
is a and the coordinates of field point p, with respect to the dipole midpoint, are indicated
in spherical coordinates r, θ, φ.

Substituting the current into Eq. 2.133 the vector potential for the harmonic oscillator
is
µ0 a/2 −q0 ω sin ω t − rc ẑ
Z 

A(r, t) = dz (2.139)
4π a/2 r

But since the length a of the oscillator is short compared to the wavelength we can
simply replace the integral with the integrand multiplied by a.
µ0 p0 ω h  r i
A(r, t) ' − sin ω t − ẑ (2.140)
4πr c

We can write down the retarded scaler potential for some point r at time t, V (r, t) by
using Eq. 2.132 and Fig. 2.7.
 
1 q0 cos [ω(t − r1 /c)] q0 cos [ω(t − r2 /c)]
V (r, t) = − (2.141)
4π0 r1 r2

The denominators r1 , r2 can be written in terms of r and the angle θ as

  a 2 1/2
2
r1,2 = r ∓ ra cos θ + (2.142)
2

Now if the point where the field is to be evaluated is sufficiently far from the dipole
such that|r| >> a and the length of the dipole is sufficiently short such that the
 Dipole Radiation 29

wavelength of light λ = ω/c >> a then it can be easily shown by expanding 1/r1,2 in
a Taylor series that
 
p0 cos θ ω h  r i 1 h  r i
V (r, θ, t) ' − sin ω t − + 2 cos ω t − (2.143)
4π0 rc c r c

In order to calculate the time-varying E-field we need to know A(r, t) as well as

V (r, θ, t) because
∂A
E = −∇V − (2.144)
∂t
Using Eq. 2.140 for the vector potential of a dipole aligned along the z axis,

∂A µ0 p0 ω 2 h  r i
=− cos ω t − ẑ (2.145)
∂t 4πr c

expressed in spherical coordinates as

∂A µ0 p0 ω 2 h  r i
=− cos ω t − (cos θr̂ − sin θθ̂) (2.146)
∂t 4πr c

Substituting Eqs. 2.143 and 2.146 into 2.144 and writing the gradient operator ∇ in
spherical coordinates (see Appendix C),

∂ 1 ∂V
∇= r̂ + θ̂
∂r r ∂θ

we find the radial component of the E-field

 
2p0 cos θ 1 h  r i ω h  r i
Er = cos ω t − − 2 sin ω t − r̂ (2.147)
4π0 r3 c r c c

and the angular component

(
p0 sin θ 1 h  r i ω h  r i
Eθ = cos ω t − − sin ω t −
4π0 r3 c r2 c c
)
ω2 h  r i
− 2 cos ω t − θ̂ (2.148)
rc c

Then, using B = ∇ × A we find for the oscillator B-field

r i ω 2
 
µ0 p0 sin θ ω h  h  r i
Bϕ = − sin ω t − + cos ω t − ϕ̂ (2.149)
4π r2 c rc c
30 Elements Classical Field Theory

2.8.1 Far Field

In the “far field” where r >> λ the only surviving E-field term is
p0 sin θ ω 2 h  r i
Eθ = − cos ω t − θ̂ (2.150)
4π0 rc2 c
and the surviving B-field terms is
µ0 p0 sin θ ω 2 h  r i
Bϕ = − cos ω t − ϕ̂ (2.151)
4π rc c
which can be rewritten, using µ0 0 = 1/c2 , as
p0 sin θ ω 2 h  r i
Bϕ = − cos ω t − ϕ̂ (2.152)
4π0 rc3 c
Comparing Eqs. 2.150 and 2.152 we see that the field amplitudes are related simply
by a factor of c,
1
Bϕ = Eθ (2.153)
c
that the Eθ and Bϕ components are in phase, and that they are orthogonal to each
other and to the radial propagation direction r̂. The components Eθ and Bϕ in the
far field as r → ∞ become the plane waves discussed in section 2.10.
The Poynting vector describes the energy flux emitted into the far-field region by
the oscillating dipole.
µ0 p20 ω 4 sin2 θ
 
1 2
h  r i
S= (E × B) = cos ω t − r̂ (2.154)
µ0 16π 2 c r2 c
The optical-cycle averaged energy flux hSi (energy per unit area per unit time) is
µ0 p20 ω 4 sin2 θ
 
hSi = (2.155)
32π 2 c r2
and the total power W (energy per unit time) emitted by the oscillator is calculated
by integrating the energy flux over all space
µ0 p20 ω 4 sin2 θ 2 µ0 p20 ω 4
Z Z
W = hSi · da = 2 2
r sin θdθdϕ = (2.156)
32π c r 12πc
2.8.2 Quasi-stationary Regime
In the regime where r . λ or equivalently where ωr/c . 2π we can drop the retardation
term in the time arguments of Eqs. 2.147, 2.148, 2.149 and rewrite them as
2p0 cos θ h  ωr  i
Er = cos ωt − sin ωt r̂ (2.157)
4π0 r3 c
  ωr 2 
p0 sin θ  ωr 
Eθ = cos ωt − sin ωt − cos ωt θ̂ (2.158)
4π0 r3 c c
   ωr 2 
p0 sin θ ωr
Bϕ = − sin ωt + cos ωt ϕ̂ (2.159)
4π0 cr3 c c
 Light Propagation in Dielectric and Conducting Media 31

We see that in the quasi-stationary regime the surviving E-field and B-field terms from
an oscillator source (the first terms on the right in Eqs. 2.157–2.159) are in quadra-
ture, not in phase as in far-field radiation. This regime corresponds to conventional
engineering circuit response where the product of frequencies ω (radio and microwave)
and circuit size r are well within the ωr/c . 2π criterion.
Note that when the quasi-stationary regime goes to the truly stationary, ω → 0,
Eqs. 2.157–2.159 go to their electrostatic dipole limits,

2p0 cos θ p0 sin θ

Er = Eθ = Bϕ = 0
4π0 r3 4π0 r3

2.9 Light Propagation in Dielectric and Conducting Media

So far we have assumed that light propagates either through a vacuum or through a
gas so dilute that we need consider only the isolated field–particle interaction. Here we
study the propagation of light through a continuous dielectric (nonconducting) medium
and near the surface of a good conductor. Interaction of light with such media permits
us to reexamine the important quantities of polarization, magnetization, susceptibility,
index of refraction, extinction coefficient, and absorption coefficient. We shall see later
how the polarization field can be usefully regarded as a density of transition dipoles
induced in the dielectric by the oscillating light field. Formally, as the definitions
Eqs. 2.17, 2.18 show, the magnetization M of materials is on an equal footing with
the polarization P, but in practice electric polarization is encountered much more
commonly. We will restrict the present discussion therefore to the response of materials
to the incident electric field component of a plane wave. We begin by recalling the
definition of material polarization P with respect to an applied electric field E as

P = 0 χ e E (2.160)

where χe is the linear electric susceptibility, an intrinsic property of the medium re-
sponding to the light field. Recall the relation between the electric field E, the po-
larization P, and the displacement field D in a material medium. In the rationalized
MKS system of units the relation is

D = 0 E + P (2.161)

Furthermore, for isotropic materials, in all systems of units, the so-called constitutive
relation between the displacement field D and the imposed electric field E is written

D = E

with  referred to as the permittivity of the material. Therefore

D = 0 (1 + χ)E

and
 = 0 (1 + χ)
32 Elements Classical Field Theory

The susceptibility χ is often a strong function of frequency ω around resonances and

can be spatially anisotropic. It is a complex quantity having a real, dispersive compo-
nent χ0 and an imaginary absorptive component χ00 :
χ = χ0 + iχ00
Real and imaginary parts of the permittivity are related to the susceptibility by
 = 0 (1 + χ0 ) + i0 χ00 (2.162)
with
0 = 0 (1 + χ0 ) (2.163)
00 00
 = 0 χ (2.164)
A number of familiar expressions in free space become modified in a dielectric medium.
For example
 2
kc
=1 ; free space
ω
 2
kc
= 1 + χ ; dielectric
ω
In a dielectric medium kc/ω becomes a complex quantity that is conventionally ex-
pressed as
kc
= η + iκ
ω
where η is the refractive index and κ is the extinction coefficient of the dielectric
medium. The relations between the refractive index, the extinction coefficient, and
the two components of the susceptibility are
η 2 − κ2 = 1 + χ0
2ηκ = χ00
Note that in a transparent lossless dielectric medium
0
η 2 − κ2 = 1 + χ0 =
0
But if the medium has an absorptive component
00
2ηκ =
0
The unitless terms 0 /0 and 00 /0 are called the real and imaginary parts of the
relative permittivity or the dielectric constant. They are denoted as
0
0r = (2.165)
0
00
00r = (2.166)
0
The subscript r emphasizes relative permittivity. Unfortunately in the scientific liter-
ature the dielectric constant is often just written  = 0 + i00 with the same notation
 Light Propagation in Dielectric and Conducting Media 33

as the permittivity, and one must decipher from the context whether the permittivity
(units of C 2 /J · m) or relative permittivity–dielectric constant (unitless) is intended.
For example it is clear from expressions involving the refractive index such as
√
n = η + iκ = 

or a propagation parameter √
k = k0 n = k 
that  is the dielectric constant whereas in an expression involving a constitutive
relation such as
D = E
the  factor denotes a material permittivity.
In a dielectric medium the traveling wave solutions of Maxwell’s equation become,
ηz
E = E0 ei(kz−ωt) −→ E0 e[iω( )−ω κc z]
c −t

the relation between magnetic and electric field amplitudes is

√ √
B0 = 0 µ0 E0 −→ B0 = 0 µ0 (η + iκ) E0

and the period-averaged field energy density is

1 2 1 2
ρ̄ω = 0 |E| −→ ρ̄ω = 0 η 2 |E| (2.167)
2 2
Now the light-beam intensity in a dielectric medium is attenuated exponentially by
absorption:
 
1 2 1 2 c 1 ωκ
I¯ω = 0 c |E| −→ I¯ω = 0 η 2 |E| = 0 cηE02 e−2 c z = I¯0 e−Kz (2.168)
2 2 η 2
where
1
I¯0 = 0 cηE02 (2.169)
2
is the intensity at the point where the light beam enters the medium, and
ωκ ω
K=2 = χ00 (2.170)
c ηc
is termed the absorption coefficient. Note that the energy flux I¯ω in the dielectric
medium is still the product of the energy density
1 2
ρ̄ω = 0 η 2 |E|
2
and the speed of propagation c/η. Note also that, although light propagating through
a dielectric maintains the same frequency as in vacuum, the wavelength contracts as
c/η
λ=
ν
34 Elements Classical Field Theory

2.10 Plane Electromagnetic Waves

2.10.1 Phasor Form of the Vector Fields
Up to now we have tacitly assumed that the vector force fields of electricity and
magnetism are real. However, for most of the following chapters we shall be studying
electromagnetic waves of various sorts, propagating through dielectric and conductive
media. The sources of these waves will be posited to be dipoles harmonically oscillating
at the angular frequency ω = 2πν and removed spatially to negative infinity. Therefore
the time dependence of the fields will always be set to a factor of e−iωt , and the spatial
oscillation will always have an associated phase, eiϕ . The general form of the vector
fields can then be expressed as
A(r, t) = A(r)eiϕ e−iωt (2.171)
If the spatial dependence of the wave is sinusoidal as well, then the spatial phase ϕ
can be written
ϕ=k·r+δ (2.172)
where k is called the propagation vector, the magnitude of which is related to the
wavelength λ by
2π
k= (2.173)
λ
and δ is just the reference phase at the wave origin, usually set to zero. The direction
of the propagation vector is given by
k = kx x̂ + ky ŷ + kz ẑ (2.174)
The form1 are always expressed as of the general vector field, Eq. 2.171, is then
A(r, t) = A(r)ei(kx x+ky y+kz z) e−iωt (2.175)
The vector form of the amplitude A(r) reflects the fact that in most instances the am-
plitudes components Ax , Ay , Az are not equal, and the amplitude is polarized linearly,
circularly, or radially. In the subsequent discussion we will assume the simplest case
of linear polarization with electromagnetic vector fields aligned along some cartesian
axis. The direction of polarization is described by a polarization vector ê(r) of unit
length so that we can write
A(x, y, z, t) = ê(r) · A(r)ei(kx x+ky y+kz z) e−iωt (2.176)
| {z }
phasor

and the complex spatial part of the field is called a phasor. Unless otherwise explicitly
stated otherwise we will assume from here forward that the usual fields of electromag-
netism, E, B, D, H, P, M, J, S are in the phasor form.
1 The reader should be mindful that physics literature and engineering literature use different
conventions to describe time-harmonic phasor fields. The form A = A0 ei(k·r−ωt) is the physics
convention. The form A = A0 e−j(β·r−ωt) is the engineering convention. An important consequence
of this choice is that complex quantities ã must be expressed as ã = a+iα with the physics convention
and ã = a−jα with the engineering convention. Thus a complex propagation parameter in the physics
convention would be written k̃ = k + iκ while engineers would write β̃ = α − jγ. Most of the circuit,
transmission line and waveguide literature is written by engineers.
 Plane Electromagnetic Waves 35

2.10.2 Decoupling Maxwell’s Curl Equations

We saw in section 2.8.1 that the far-field spherical waves morphed asymptotically into
local plane waves. We can also simply consider that the dipole source is removed to
−∞ and posit the plane wave form as possible solutions to Maxwell’s equations. The
two curl equations, Eqs. 2.27, 2.28, are at the heart of classical electrodynamics. They
are two coupled first order differential equations, but they can be uncoupled to form
two second order differential equations that admit propagating electromagnetic wave
solutions. The standard way to carry out the uncoupling is to apply the curl operation
to Eqs. 2.27, 2.28, invoke a vector calculus identity, and posit source-free conditions.
The curl operations yields
 
2 ∂B
∇ × (∇ × E) = ∇ (∇ · E) − ∇ E = ∇ × − (2.177)
∂t
∂ ∂2E
= − (∇ × B) = −µ0 0 2 (2.178)
∂t ∂t
In source-free regions ∇ · E = 0 so
∂2E 1 ∂2E
∇2 E = µ0 0 = (2.179)
∂t2 c2 ∂t2
Similarly
 
2 ∂E
∇ × (∇ × B) = ∇(∇ · B) − ∇ B = ∇ × µ0 0 (2.180)
∂t
2
∂ ∂ B
= µ0 0 (∇ × E) = −µ0 0 2 (2.181)
∂t ∂t
According to Eq. 2.26, the divergence of the magnetic induction field is always null,
∇ · B = 0, so
∇2 B 1 ∇2 B
∇2 B = µ0 0 2 = 2 (2.182)
∂t c ∂t2
Equations 2.179, 2.182 clearly have solutions of the form of plane waves propagating
with velocity c. We shall generally be concerned with light propagating in free space,
or in dielectric and conductive materials. We start by describing the properties of
the simplest form of light propagation, the plane electromagnetic wave composed of
two fields, electric and magnetic, oscillating at a single frequency ω. Figure 2.8 shows
the essential features of the plane light wave. The electric field E constituent of the
electromagnetic field, propagating in free space as a plane wave in direction k, is given
by
E = êE E˜0 ei(k·r−ωt (2.183)
In particular, with propagation in the z direction, the wave has the form
E = êE E˜0 ei(kz z−ωt) (2.184)
The complex field amplitude is E˜0 , the polarization direction êE , and the frequency
ω. The frequency, wavelength λ and velocity v of the constant phase front are related
by
36 Elements Classical Field Theory

Fig. 2.8 Plane electromagnetic wave propagating in free space. The E-field is aligned along
x, the B-field along y, and the wave vector k is oriented along z.

2π
ω= v = kv (2.185)
λ
In free space v is the speed of light c and in other homogeneous materials is given by
c
v=
n
where n is the index of refraction of the material. The index of refraction in free space
is unity and in matter is always greater. In materials with negligible absorption, n is a
real number greater than unity, but in absorbing, “lossy” materials n is complex with
a positive imaginary term. As light of a single frequency propagates through materials
of different indices of refraction, ω remains constant, but the phase velocity changes as
c/n and the wavelength shortens by λ/n. It is important not to confuse the “angular”
frequency ω with the frequency of optical cycles ν. The relation is always

ω = 2πν

The magnetic induction field B, the other constituent of the plane electromagnetic
wave propagating in free space in the z direction, is given by

B = êB B̃0 ei(kz z−ωt (2.186)

with complex amplitude B̃0 and polarization direction êB . The reason for denoting
the field amplitudes complex is to express a possible phase relation between them. In
fact the divergence and curl relations in Maxwell’s equations insure that the E- and
B-fields are orthogonal to the direction of propagation and that the amplitudes are in
 Plane Electromagnetic Waves 37

phase. The E and B fields combine to form the electromagnetic (E-M) field. They are
mutually orthogonal and orthogonal to the direction of propagation.

êE · êB = 0 and êE × êB = êk

where êk is the unit vector of the propagation parameter k. The relative amplitude
between magnetic induction and electric fields in vacuum is given by,

1 √
B0 = E0 = 0 µ0 E0 (2.187)
c

and for a traveling wave through free space and homogeneous matter, the time oscil-
lation of the E and B fields is always in phase.

2.10.3 Plane Wave Field Relations from Maxwell’s Equations

For plane waves with E-field and H-field in the x − y plane, propagating along z in a
dielectric medium with permittivity  and permeability µ and oscillating at frequency
ω we have, from the curl relations of Faraday’s law and the Maxwell-Ampère law,
Eqs. 2.27, 2.28

∂Ex ∂Hy
= −µ = iωµHy (2.188)
∂z ∂t
∂Ey ∂Hx
=µ = −iωµHx (2.189)
∂z ∂t

and

∂Hy ∂Ex
= − = iωEx (2.190)
∂z ∂t
∂Hx ∂Ey
= = −iωEy (2.191)
∂z ∂t

Equations 2.188 and 2.190, relating Ex and Hy , are coupled as are Eqs. 2.189 and
2.191, relating Ey and Hx . The first of these pairs can be decoupled by differentiating
Eq. 2.188 by z and Eq. 2.190 by t. The order of differentiation is interchangeable which
results in
∂ 2 Ex ∂ 2 Ex
− µ =0 (2.192)
∂z 2 ∂t2
Taking into account the harmonic time variation at frequency ω, we have

∂ 2 Ex
+ µω 2 Ex2 = 0 (2.193)
∂z 2

The product of the material parameters µ = 1/v 2 , where v is the phase velocity
of the plane wave in the medium. The plane wave propagation parameter k = ω/v
38 Elements Classical Field Theory

so Eq. 2.193 can be written finally as the one-dimensional wave equation for the Ex
component of the plane wave.

∂ 2 Ex
+ k 2 Ex2 = 0 (2.194)
∂z 2

Clearly analogous equations can be written for all the field components in the relations
Eqs. 2.188–2.191. Furthermore, substitution of plane wave E-field solution, Eq. 2.184,
into the second-order wave equation, 2.179, results in

∇2 E + k 2 E = 0 (2.195)
√
with k = ω/c = ω µ0 0 when the waves are propagating in free space. Equation 2.195
is clearly valid for each of the Cartesian components and can be written simply in
terms of the wave amplitude,
∇2 E + k 2 E = 0 (2.196)

as such it is called the Helmholtz scalar wave equation. The analogous equation is
obviously valid for the plane-wave H-field, and the relation between the E-field and
H-field plane-wave amplitudes is specified by Eq. 2.189.
r
E µ0
B= and therefore E= H (2.197)
c 0
p
The proportionality factor between the amplitudes, µ0 /0 , has units of impedance
and is variously called the impedance of free space, or intrinsic impedance, or wave
impedance and is denoted by Z0 .
r
µ0
Z0 = = 376.73 Ohms (2.198)
0

In lossless media µ and  are real numbers, but in lossy dielectrics or conductive
materials such as metals or semiconductors, the permittivity becomes complex. We
examine the interaction of plane waves and lossy or conductive media in the next
section.

2.10.4 Plane Waves in a Lossy, Conductive Medium (low frequency

limit)
In this section we examine the propagation of plane waves in a lossy conductor. The
constitutive relations for permeability, permittivity, and conductivity, µ, , σ, respec-
tively characterize the medium. As usual we do not consider “magnetic” materials so
the permeability remains µ0 even inside the conductor but the permittivity becomes
complex,  = 0 (0 + i00 ). Here we assume that σ is real which is perfectly adequate for
RF and microwave frequencies. Later in Section we shall see that at optical frequencies
the conductivity must also be generalized to a complex quantity.
 Plane Electromagnetic Waves 39

We start again with Faraday’s law and the Maxwell-Ampère law including both
displacement and charge current in a medium characterized by µ, , σ. Assume further
that Ohm’s law obtains so that J = σE.
∂B ∂H
∇×E=− = −µ
∂t ∂t
∂D ∂E
∇×H= +J= + σE
∂t ∂t
Let us posit plane solutions with the E-field aligned along x, B-field aligned along y,
and propagating along z.
Ex = E0 ei(kz z−ωt)
(2.199)
By = B0 ei(kz z−ωt)
The single frequency, harmonic time dependence results in
∇ × E = iµωH (2.200)
∇ × H = (−iω + σ)E (2.201)
Take the curl of both sides of Eq. 2.200 and apply the standard “curl-curl” identity
∇ × ∇ × E = iµω∇ × H (2.202)
2
∇(∇ · E) − ∇ E = iµω∇ × H (2.203)
2
−∇ E = iµω∇ × H (2.204)
Because, as usual, we assume a source-free environment where the plane wave is prop-
agating, ∇ · E evaluates to zero in Eq. 2.203. Now substituting Eq. 2.201 into the right
side of Eq. 2.204 results in
 σ
∇2 E + µω 2 1 + i E=0 (2.205)
ω
This is clearly a “Helmholtz-like” wave equation, and since we have posited the E-field
along x,
∂ 2 Ex  σ
2
+ µω 2 1 + i Ex = 0 (2.206)
∂z ω
and the H-field along y
∂ 2 Hy  σ
2
+ µω 2 1 + i Hy = 0 (2.207)
∂z ω
The k of Eq. 2.196, there assumed real, generalizes to a complex expression
√  σ 1/2
k = µω 1 + i (2.208)
ω
For nonmagnetic lossy materials we can write µr = 1 an  = 0 r = 0 (0 + 00 ). Then
  1/2
2π 0 00 σ
k=  +i  + (2.209)
λ0 ω0
40 Elements Classical Field Theory

Plane Waves in a Lossy Dielectric (but poor Conductor). If σ/ω0 << 00 then the
material is considered a poor conductor or good insulator and the propagation param-
eter becomes
2π 0
k→ ( + i00 )1/2 (2.210)
λ0
A plane wave propagating through this medium reflects dissipative loss due to 00 .
0 00 1/2
Ex = E0 ei[2π/λ0 ( +i ) z−ωt]
(2.211)
In many cases |00 /0 | << 1 in which case we can write
2π √ 0 00
 
k'  1+i 0 (2.212)
λ0 2
and Eq. 2.211 can be written
00 √
−(2π/λ0 ) √ z i[(2π/λ0 ) 0 z−ωt]
Ex = E0 e 2 0 e (2.213)
The first exponential factor with the real argument damps the amplitude E0 as the
wave propagates along z. The second exponential factor√ represents the usual harmonic
wave oscillation through a medium characterized 0 . The two factors evidently rep-
resent a damped travelling plane wave. As 00 → 0 the damping vanishes and the
oscillatory term represents√a plane wave propagating through a lossless dielectric with
real refractive index η = 0 . The wave impedance in a lossy medium is again deter-
mined from Faraday’s law. We find in this case
r
µ0
Ex = Hy (2.214)
0 (0 + i00 )
and r
µ0
Z= (2.215)
0 (0 + i00 )
In the limit of a lossless, nonmagnetic dielectric, 00 → 0,
r
µ0 1
Z= = Z0 √ (2.216)
 0
Plane Waves in a Good Conductor (low frequency limit). As in Section 2.10.4 here
we continue to treat σ as a real quantity. We shall see in Section 2.10.7 that although
this assumption is adequate for radio and microwave frequencies, it breaks down at
optical frequencies.
When σ/ω0 >> 00 we have a good conductor. The generalized propagation pa-
rameter, Eq. 2.209 becomes, as 00 ω0 /σ → 0,
 1/2
2π 0 σ
k→  +i (2.217)
λ0 ω0
The good conductor (often a metal such as silver or gold) is then characterized by
the real part of the permittivity and the conductivity of the metal at frequency ω.
 Plane Electromagnetic Waves 41

If the dissipative part 00 of the permittivity is retained, then the imaginary terms in
Eq. 2.209 can be grouped together and labeled as a “total” 00tot .

σ
00tot = 00 + (2.218)
ω0

and an “equivalent conductivity” including both terms of the imaginary part of the
permittivity can be defined as
σequiv = ω0 00tot (2.219)

Returning to Eq. 2.273, we see that if σ/ω0 >> 0 then a plane wave entering the
conductor will by damped by the conductivity term.
1/2
2π √

σ
k' i (2.220)
λ0 ω0
√ √
Using the identity i = (1 + i)/ 2,
 1/2
2π σ
k' (1 + i) (2.221)
λ0 2ω0

The imaginary term damps the plane wave propagating into the conductor. The depth
of penetration, the skin depth δ is defined as inverse of this imaginary term
 1/2
λ0 2ω0
δ= (2.222)
2π σ

As a specific example take silver metal as a good conductor and λ0 = 514.5 nm, the
“green line” of an Argon-ion laser. At this wavelength, 0 ' −15 and 00 ' 0.3. The
conductivity of silver is σ ' 6.3 × 107 S/m. The ratio σ/ω0 ' 2 × 103 much greater
than the real or imaginary parts of the dielectric constant. The expression for the skin
depth is therefore valid and for this example δ ' 2.6 × 10−9 m. At optical frequencies
an incident plane only penetrates a few nanometers below the surface of a good metal.
The impedance is again determined from Faraday’s law, specifying the relation
between the E- and H-fields in the conductor.
r  −1 r
µ0 (1 + i) ω0
Ex = √ Hy (2.223)
0 2 σ

and
r
ω0
Z = Z0 e−i(π/2) (2.224)
σ

Note that the E-field lags the H-field in the conductor by a phase angle of 45 degrees.
42 Elements Classical Field Theory

2.10.5 Energy Density and Flux in Harmonic Phasor Fields

The discussion of energy density and power in Section 2.5 must now be extended to
account for the complex form of harmonic phasor fields. The expressions for the E-field
energy and B-field energy (Eqs. 2.42, 2.48) in the static case and in vacuum are
Z
1
Eelec = E · D dτ
2
Z
1
Emag = H · B dτ
2
If the fields are time-harmonic phasors than the optical-cycle-averaged expressions for
the electric and magnetic field energies in vacuum are
Z
0
Eelec = E · E∗ dτ (2.225)
4
Z
µ0
Emag = H · H∗ dτ (2.226)
4
From the relation between the amplitudes of the electric and magnetic fields
1
|H| = |E| (2.227)
µ0 c
It is clear that in vacuum
Emag = Eelec (2.228)
the magnetic field energy of the wave is equal to the electric field energy. Starting with
an expression analogous to Eq. 2.52 we can develop a power balance relation between
work done on moving charges and the energy flow from plane-wave phasor fields. In
this case we will include charges flowing in conductors Jc as well as freely moving
charged particles Jfree and write the total current as

Jtot = Jfree + Jc (2.229)

and the time-rate of doing work on or by these charges by the electromagnetic field is
is
dE
= E · Jtot (2.230)
dt
Now we use the Maxwell-Ampère law and Faraday’s law in their forms appropriate
for time-harmonic phasors. We generalize the permittivity to include the possibility
of lossy dielectrics and conductors and write  = 0 (0 + i00 ), but do not consider the
much less frequent case of magnetic materials. Therefore we leave the permeability as
µ0 and write
∂E∗
∇ × H∗ = ∗ + J∗ tot = iω∗ E∗ + J∗ tot (2.231)
∂t
∂H
∇ × E = −µ0 = iωµ0 H (2.232)
∂t
 Plane Electromagnetic Waves 43

Now we “dot-multiply” the conjugate Faraday’s law by E and the Maxwell-Ampère

law by H∗ and invoke the vector field identity

∇ · (E × H∗ ) = (∇ × E) · H∗ − (∇ × H∗ ) · E (2.233)

Substituting the result of the dot-multiplications into Eq. 2.233 we find

∇ · (E × H∗ ) = iω µ0 |H|2 − 0 0 |E|2 − ω0 00 |E|2 − J∗free · E − σ|E|2

(2.234)

Then applying Stokes’ theorem to the term on the left, rearranging the terms on the
right, and dividing all terms by 2,
Z I Z Z
1 ∗ 1 ∗ 1 00 2 1
− Jfree · E dV = (E × H ) · da + ω0  |E| dV + σ|E|2 dV
2 2 S 2 2
Z 
iω 0 2 2


+ 0  |E| − µ0 |H| dV (2.235)
2

This power balance equation requires careful interpretation. The term on the left can
be considered a source term from which energy is emitted. The negative sign denotes
power flowing away from the emitter through the surface S of the enclosing volume V .
The first term on the right is immediately recognizable as the cycle-averaged power
passing out of the volume enclosed by surface S. This term is the phasor version of
Eqs. 2.57, 2.59; and we can identify the integrand with the Poynting vector

S = E × H∗ (2.236)

The next two terms indicate dissipative loss: absorptive loss from 00 and conductive
loss from σ due to material present within the enclosed volume. The last term describes
energy “stored” in the material. Taking into account Eq. 2.227, we can write this last
term as Z  Z 
dEstored iω 0 2 1 dP
= χ 0 |E| dV = − E· dV (2.237)
dt 2 2 dt
and we see that this term represents power flow into the polarizability of the material.
Since polarizability is a density of dipoles, we can think of this term as that part of the
total emitted energy that is stored in the dipole oscillators of the material rather than
the dipole oscillators of the field. The negative sign indicates that the stored energy
and field energy oscillate out of phase.

2.10.6 Reflection and Transmission of Plane Waves at an Interface

Reflection and transmission of light at surfaces is encountered frequently in the study
of light-matter interaction. We discuss here a few simple cases in which a plane wave,
incident from a half-space of vacuum (0 , µ0 , σ = 0), impinges at the surface of a
material characterized by , µ, σ. The cases are organized according to the nature (real
or complex) and magnitude of the permittivity and conductivity. As before we consider
the material nonmagnetic and keep the permeability as µ0 .
44 Elements Classical Field Theory

General Case. We posit a plane propagating from left to right along z, with E-field
linearly polarized along x and H-field along y as shown in Fig. 2.8. The interface is in
the x − y plane and situated at z = 0. The half-space to the left (z < 0) is vacuum,
and the half-space to the right (z > 0) is characterized by , µ0 , σ. The incident E-
and H-phasor fields are given by

Ei = E0 eik0 z x̂ (2.238)
E0 ik0 z
Hi = H0 eik0 z ŷ = e ŷ (2.239)
Z0
p
where Z0 is the impedance of free space, Z0 = µ0 /0 . At the interface a fraction of
the wave is reflected and a fraction is transmitted. We write the reflected wave as2

Er = −RE0 e−ik0 z x̂ (2.240)

E0
Hr = R e−ik0 z ŷ (2.241)
Z0
where R is the reflection coefficient. The transmitted wave is given by

Et = T E0 eikz x̂ (2.242)
E0 ikz
Ht = T e ŷ (2.243)
Z
In the material to the right, the propagation parameter k is given by Eq. 2.209 and
rewritten here for convenience,
  1/2
2π 0 σ
k=  + i 00 +
λ0 ω0
At z = 0 the total parallel field amplitudes on either side of the boundary must be
continuous across it. Therefore

Ei + Er = Et (2.244)
Ei − REi = T Ei (2.245)
Hi + RHi = T Hi (2.246)
Ei Ei Ei
+R =T (2.247)
Z0 Z0 Z
From Eqs. 2.245, 2.247 we have

1−R=T (2.248)
1 R T
+ = (2.249)
Z0 Z0 Z
2 The choice of phase for the reflected wave is a matter of convention. The reflected waves could
also be written Er = RE0 e−ik0 z x̂ and Hr = −R(E0 /Z0 )e−k0 z ŷ. The first convention is often used
in physical optics while the second is common in electrical engineering. The second choice results in
a change of sign for the reflection coefficient with the consequence that R = (Z − Z0 )/(Z0 + Z) and
1 = T − R.
 Plane Electromagnetic Waves 45

From these two relations we find

1=R+T (2.250)
Z0 − Z
R= (2.251)
Z0 + Z
2Z
T = (2.252)
Z0 + Z
Lossless Dielectric. For a lossless dielectric 00 = 0 and σ = 0 so the propagation
vector becomes k = (2π/λ0 ) 0 . The Poynting vector on the left side of the boundary
is

Sz<0 = (Ei + Er ) × (H∗i + H∗r ) (2.253)

 
ik0 x −ik0 x

E0 −ik0 z E0 ik0 z
= E0 e − RE0 e × e +R e ẑ
Z0 Z0
E2 
= 0 1 − R2 − i2R sin (2k0 z) ẑ


(2.254)
Z0
At z = 0 the energy flux from the left is

E02
1 − R2 ẑ


S− = (2.255)
Z0
and since R < 1, the net energy flux propagates from left to right. In the region z < 0
the oscillatory term in Eq. 2.254 is due to the standing wave set up by reflection of
the plane wave at the boundary. Notice that the phase of the standing wave is in
quadrature with respect to the incident wave. The Poynting vector on the right side is

Sz>0 = (Et ) × (H∗t ) ẑ (2.256)

2
2 E0
=T ẑ (2.257)
Z
The energy flux across the boundary should be continuous, and by using the expres-
sions for R and T in terms of the impedances, Eqs. 2.248, 2.249, the continuity can be
verified.

Lossy Dielectric. In the case of a lossy dielectric the propagation vector becomes
2π 0 1/2
k= ( + i00 ) (2.258)
λ0
2π
= (η + iκ) (2.259)
λ0
and the wave impedance becomes complex.
r r
µ0 µ0 Z0
Z= = = (2.260)
 0 (0 + i00 ) η + iκ
46 Elements Classical Field Theory

Therefore, from Eqs. 4.60, 2.252 the reflection and transmission coefficients become
complex, and the parallel field component continuity conditions at the boundary now
give
2
T = (2.261)
(η + 1) + iκ
(η − 1) + iκ
R= (2.262)
(η + 1) + iκ

Figure 2.9 shows a plane wave incident from the left on a slightly absorbing slab of
SiO. The shortened wave length and slight loss of amplitude due to absorption can be
discerned within the material. The Poynting vector in the z < 0 region is
2
|E0 |  2

Sz<0 = 1 + R∗ ei2k0 z − Re−i2k0 z − |R| ẑ (2.263)
Z0
and in the z > 0 region
2 2
2 |E0 | −2κz |E0 | 4(η − iκ) −2κz
Sz>0 = |T | e ẑ = · e ẑ (2.264)
Z∗ Z0 (η + 1)2 + κ2

At the boundary (z = 0) the flux from the z < 0 region is

2
− |E0 |
|E0 |2 4(η − iκ)
Sz=0 = 1 + R∗ − R − |R|2 = · 2 (2.265)
Z0 Z0 (η + 1) + κ2

and from the z > 0 region

2
+ |E0 | 4(η − iκ)
Sz=0 = · (2.266)
Z0 (η + 1)2 + κ2

So we see that the complex Poynting vector is conserved across the boundary. When
the Poynting vector is complex, we interpret the energy flux or the power flow as the
cycle-averaged real part of the Poynting vector. The net energy flux traversing the
boundary from the z < 0 region is therefore
2
1 |E0 | 2η
P− = Re[S − ] = · (2.267)
2 Z0 (η + 1)2 + κ2

and the energy flux penetrating the lossy dielectric on the z > 0 side of the boundary
is
2
1 |E0 | 2η
P + = Re[S + ] = · e−2κz (2.268)
2 Z0 (η + 1)2 + κ2
Note that the incident energy flux on the boundary from the left is
2
E0∗ −ik0 z
 
1 |E0 |
Pi = Re E0 eik0 z · e = (2.269)
2 Z0 2Z0
 Plane Electromagnetic Waves 47

Fig. 2.9 Top panel shows the real part of the incident E-field Ei , reflected E-field Er and
total E-field ET in air to the left of the air-dielectric interface. The wave length of the
propagating wave in air is λ0 = 620 nm. Within the dielectric the transmitted E-field Et is
shown. The dielectric is SiO with index of refraction n = η + iκ = 1.969 + i0.01175. Bottom
panel shows the real part of the incident H-field Hi . reflected H-field Hr and total H-field HT
in air to the left of the air-dielectric interface. Within the dielectric the transmitted H-field
Ht is shown. Note the shortening of the wave length and the slight decrease in amplitude as
the wave penetrates into the lossy dielectric.

and the reflected flux is

2
R∗ E0∗ ik0 z
 
1 |E0 |
Pr = Re −RE0 e−ik0 z · e = −|R|2 (2.270)
2 Z0 2Z0
The sum of the incident and reflected power
48 Elements Classical Field Theory

2  |E |2
|E0 |  2 0 2η
PT = Pi + Pr = 1 − |R| = · (2.271)
2Z0 Z0 (η + 1)2 + κ2

accords with the net power flow calculated from the superposed incident and reflected
fields (Eqs. 2.263, 2.267).

Good Conductor (Low Frequency Regime). A plane wave incident on a perfect con-
ductor and on a good conductor are shown in Figs. 2.10, 2.11. The difference between
“perfect” and “good” is essentially the magnitude of the conductivity, σ. When σ → ∞
a conductor approaches the perfect conductor idealization. In the case of a good con-
ductor
 1/2
2π 0 σ
k=  +i (2.272)
λ0 ω0
As we saw in the case of silver metal, for good conductors σ/ (ω0 ) is on the order of
thousands while 0 is on the order of unity. We can then to a quite good approximation
write
 1/2
2π σ
k' i (2.273)
λ0 ω0
and recalling from Eq. 2.222 that the skin depth is defined as
 1/2
λ0 2ω0
δ=
2π σ

we can express k as
√
2i 1+i
k' =± (2.274)
δ δ
The sign of k is chosen so that the wave amplitude decays exponentially as it propagates
into the good conductor in the z > 0 half space. Wave impedance in the conductor is
r r   
µ0 µ0 −1/2  ω0 1/2 k0 δ
Z= = ·i = Z0 · (1 − i) (2.275)
 0 σ 2

The reflection and transmission coefficients are

2
2 − (k0 δ) 2k0 δ
R= 2 +i 2 (2.276)
2 + 2k0 δ + (k0 δ) 2 + 2k0 δ + (k0 δ)
' 1 − ik0 δ (2.277)
h i
2
2 (k0 δ) + k0 δ 2k0 δ
T = 2 −i 2 (2.278)
2 + 2k0 δ + (k0 δ) 2 + 2k0 δ + (k0 δ)
' (1 − i)k0 δ (2.279)

At the boundary (z = 0) the flux from the z < 0 region is

 Plane Electromagnetic Waves 49

Fig. 2.10 Top panel shows the real part of the incident E-field Ei , reflected E-field Er
and total E-field ET in air to the left of the air-perfect metal interface. The wave length
of the propagating wave in air is λ0 = 632 nm. Within the perfect conductor (σ → ∞) no
electromagnetic field penetrates. Bottom panel shows the real part of the incident H-field
Hi . reflected H-field Hr and total H-field HT in air to the left of the air-dielectric interface.
Within the perfect conductor the transmitted H-field Ht is null.

2
− |E0 |
· 1 + R∗ − R − |R|2


Sz=0 =
Z0
2
|E0 | 4k0 δ
= · 2 (1 − i) (2.280)
Z0 2 + 2k0 δ + (k0 δ)

and the flux penetrating into the conductor (z > 0) is

50 Elements Classical Field Theory

Fig. 2.11 Top panel shows the real part of the incident E-field Ei , reflected E-field Er
and total E-field ET in air to the left of the air-dielectric interface. The wave length of the
propagating wave in air is λ0 = 632 nm. Within the dielectric the transmitted E-field Et is
shown. Bottom panel shows the real part of the incident H-field Hi , reflected H-field Hr and
total H-field HT in air to the left of the air-dielectric interface. Within the dielectric the
transmitted H-field Ht is shown.
2 2
|E0 | |T | −(2/δ)z
Sz>0 = e ẑ
Z∗
2
|E0 | 4k0 δ −(2/δ)z
= · 2 (1 − i)e ẑ (2.281)
Z0 2 + 2k0 δ + (k0 δ)

At the z = 0 boundary the flux from the z > 0 region is

 Plane Electromagnetic Waves 51

2 2
+ |E0 | |T |
Sz=0 =
Z∗
2
|E0 | 4k0 δ
= · 2 (1 − i) (2.282)
Z0 2 + 2k0 δ + (k0 δ)
− +
and it is clear that the complex Poynting vectors Sz=0 = Sz=0 . Because the skin depth
is only a few nanometers we see that the energy flux penetrates significantly only a
few tens of nanometers. As in the case of the lossy dielectric the power penetrating
into the metal is given by
1  + 
P+ = Re Sz=0
2
2
|E0 | 2k0 δ
= · 2 (2.283)
Z0 2 + 2k0 δ + (k0 δ)
(2.284)

And the power entering the metal from the z < 0 side is
1  − 
P− = Re Sz=0 (2.285)
2
2
|E0 | 2k0 δ
= · 2 (2.286)
Z0 2 + 2k0 δ + (k0 δ)
(2.287)

So we see that P − = P + and the power across the boundary is conserved. Furthermore
we can confirm that the sum of the incident power Pi and reflected power Pr is equal
to the power transmitted across the boundary.

1 |E0 |2
1 − |R|2


Pi + Pr = (2.288)
2 Z0
|E0 |2 2k0 δ
= · 2 (2.289)
Z0 2 + k0 δ + (k0 δ)
(2.290)

Now from the earlier example in Section 2.10.4 for an air-silver metal interface, the
value for k0 = 1.22 × 107 m−1 , the skin depth δ = 2.6 × 10−9 m, and the product
k0 δ = 3.2 × 10−2 . Therefore
|E0 |2
P− = P+ ' · k0 δ (2.291)
Z0
The cycle-averaged incident power is

|E0 |2
Pi = (2.292)
2Z0
52 Elements Classical Field Theory

and therefore the fraction of the incident power entering the metal is

P+
= 2k0 δ = 6.4 × 10−2 (2.293)
Pi

About 6% of the incident power is transmitted to the metal.

The reflected power is given by

|E0 |2
Pr = −|R|2 (2.294)
2Z0

and the fraction of the incident power reflected is

Pr
' (1 − k0 δ) = 96.8 × 10−2 (2.295)
Pi

We see that about 97% of the incident power is reflected.

Current Density and Joule Heating. We can easily calculate the magnitude and phase
of the current density induced in the real metal by using Ohm’s law and the field
continuity conditions at the interface. From Ohm’s law we have

J = σEt x̂ (2.296)

and from the continuity condition

Ei (1 − R) = Ei T = Et

For a high-conductivity material we have

r r
1 ω0 ω0
Z = Z0 · √ = Z0 = Z0 e−iπ/4 (2.297)
m iσ σ

where the expression for the conductor permittivity m is obtained from Eq. 2.273.
Since |Z| << |Z0 | the transmission coefficient T can be written
r
2Z 2Z ω0
T = ' = 2 e−iπ/4 (2.298)
Z0 + Z Z0 σ

So the transmitted and incident E-field amplitudes are related by

r
−iπ/4 ω0
Et = 2 e E0 (2.299)
σ

We see that the transmitted E-field exhibits a phase lag with respect to the incident
E-field by π/4, and Eq. 2.296 shows that the current density induced in the metal also
lags the incident E-field by π/4.
 Plane Electromagnetic Waves 53

Both the current density J and the transmitted E-field are rapidly attenuated expo-
nentially as they penetrate the conductor. From Eq. 2.274 we can write the propagation
parameter in the conductor in terms of the skin depth.
√
J(z) = 2 σω0 E0 ei(z/δ−π/4) e−z/δ x̂ (2.300)
r
ω0
Et (z) = 2 E0 ei(z/δ−π/4) e−z/δ x̂ (2.301)
σ

The power density (W m−3 ) associated with “Joule heating” in the conductor is given
by
dPJoule
= Et · J∗ (2.302)
dV
In order to compare this Joule heating power to P + or P − we need first to integrate
over the penetration depth in the z direction, then integrate over a cross sectional
area of one square meter in the x − y plane in order to calculate the volume power
dissipated in the conductor. This power dissipation corresponds to the third term on
the right in the power-balance, Eq. 2.235.
Z 1 Z 1 Z ∞ 
1
PJoule = Re Et · J∗ dxdydz
2 0 0 0
2 δ
= 2 ω0 |E0 |
2
2
|E0 |
= · k0 δ (2.303)
Z0
and we see that this result agrees with our earlier power calculation, Eq. 2.291, based
on the Poynting vector energy flux entering the conductor normal to its surface.

2.10.7 Plane Waves in a Lossy, Conductive Medium (high frequency

limit)
In Sections 2.10.4 and 2.10.6 we developed the physics of a plane electromagnetic wave
interacting at the surface of and within a good conductor (most often encountered as
a real metal). This development assumed that the conductivity σ is real. We show here
that, although this assumption is adequate for radio and microwave frequencies, the
conductivity must be considered complex in the optical range of the electromagnetic
spectrum.
We start with the wave equation for the E-field component of an electromagnetic
wave propagating in a medium, Eq. 2.206, that we rewrite here for convenience.
∂ 2 Ex 2
 σ
+ µω 1 + i Ex = 0 (2.304)
∂z 2 ω
and assume, as usual, a plane wave solution as in Eq. 2.199

Ex = E0 ei(kz z−ωt) (2.305)

Substituting 2.305 into 4.30 results in

54 Elements Classical Field Theory

kz2 = µω 2 + iµσω (2.306)

As before we assume nonmagnetic materials (µ = µ0 ) and consider the permittivity

to be complex,  = 0 (0 + i00 ). But now we consider the conductivity to be complex
as well, σ = σ 0 + iσ 00 . Now we rewrite kz as km to emphasize wave propagation inside
the metallic conductor; and recognize that it is complex as well.
0 00
km = km + ikm (2.307)

Substituting the complex quantities into Eq. 2.306 and gathering the real and imagi-
nary parts,
02 00 2
km − km = µ0 0 0 ω 2 − µ0 σ 00 ω
σ 00 σ 00
   
= µ0 0 ω 2 0 − = k02 0 − = α2 (2.308)
0 ω ω0
0 00
2km km = µ0 0 00 ω 2 + µ0 σ 0 ω
σ0 σ0
   
= µ0 0 ω 2 00 + = k02 00 + = β2 (2.309)
0 ω ω0

The two coupled equations, 2.308, 2.309 can be solved for the real and imaginary parts
of km .
r
α
q
0 4
km = ± √ 1 ± 1 + (β/α) (2.310)
2
r
α
q
00 4
km = ± √ −1 ± 1 + (β/α) (2.311)
2

We also know that the propagation parameter within the metal km is related to the
free-space propagation parameter k0 through the complex index of refraction n in the
metal.  2
km
= n2 = (η + iκ)2 (2.312)
k0
But n is also related to the real and complex “effective” dielectric constants in the
metal by
n2 = 0eff + i00eff (2.313)
The relations between η, κ and 0eff , 00eff are
r r
0eff q
η= 1 ± 1 + (00eff /0eff ) (2.314)
2
r r
0
eff q
κ= 1 ± −1 + (00eff /0eff ) (2.315)
2
But
 Plane Electromagnetic Waves 55

0
km = k0 η (2.316)
00
km = k0 κ (2.317)
Then comparing Eqs. 2.316, 2.317 to Eqs. 2.310, 2.311 we see that
σ 00
0eff = 0 − (2.318)
ω0
σ0
00eff 00
= + (2.319)
ω0
The two terms 0 , 00 are the real and imaginary parts of the metal dielectric constant
excluding the highly polarizable conduction electrons. They represent the response of
the much more tightly bound valence electrons to the transmitted plane wave. The
terms involving σ 0 , σ 00 represent the frequency-dependent response of the conduction
electrons to the transmitted wave. In high-conductivity materials these latter two
terms dominate the response. Now real metals are dispersive which means that 0eff and
00eff depend on the frequency. A simple dispersion relation can be found if the metal
conductor is modeled as a harmonically oscillating free-electron gas. The harmonic
motion is driven at the incident wave frequency ω and damped at a rate Γ, inserted into
the model phenomenologically and corresponding to acoustic and radiative dissipation.

Damped Harmonic Oscillator Model for Conduction Current. We posit that the equa-
tion of motion of the conduction electrons is governed by harmonic acceleration driven
by a plane-wave electromagnetic field propagating in the metal and polarized along
the x direction.
d2 x dx
me 2 + me Γ = eEt = eT E0 eikm z−ωt (2.320)
dt dt
where me , e are the electron mass and charge, Γ a phenomenological damping constant,
and eEt = eT E0 ei(km z−ωt) the driving force on the conduction current. Dropping the
common e−iωt factor we have for the solutions of position, velocity and acceleration
1
x=− eT E0 (2.321)
me (ω 2+ iΓω)
dx ω
=i eT E0 (2.322)
dt me (ω 2 + iΓω)
d2 x ω2
2
= 2
eT E0 (2.323)
dt me (ω + iΓω)
Now the amplitude of the conduction current density is given by
dx ω
Γω − iω 2 eT E0


Jc = eNe = eNe (2.324)
dt me
where the electron density in the metal conduction band Ne is related to the resonance
frequency of the oscillating electrons ωp by
e 2 Ne
ωp2 = (2.325)
me 0
56 Elements Classical Field Theory

The resonant frequency is called the “bulk plasmon frequency”. The conduction cur-
rent Jc can now be written as

ωp2
Jc = 0 T E 0 (2.326)
Γ − iω
But from the standard constitutive relation we have
1
Jc x̂ = σEt x̂ = σ0 T E0 x̂ (2.327)
0
Comparing Eqs. 2.327 and 2.326 and separating real and imaginary parts, we see that

σ Γ ω
= 2  + i  (2.328)
0 Γ ω2 Γ2 ω2
ω2 + p ωp2 ωp2 + ωp2

Now from Eq. 2.325 we can estimate the bulk plasmon frequency and compare it to ω
and Γ. A high-conductivity metal such as silver exhibits a conduction electron density
Ne ' 6 × 1028 m−3 and therefore ωp ' 1.3 × 1016 s−1 . The damping rate Γ due to
radiative loss and acoustic coupling is typically ∼ 1014 s−1 , and ω in the visible and
near IR is ∼ 1015 s−1 . Therefore we can write

σ ωp2 ωp2
'Γ 2 +i (2.329)
0 ω ω

Thus at relatively low frequencies (radio and microwave) the real term dominates, and
the conductivity is said to be “ohmic”. As the driving frequency into the optical range,
however, the imaginary term dominates. At ω ∼ 1015 s−1 the real term is only about
5% of the imaginary term. We can, from Eqs. 2.329, 2.318, 2.319, write the frequency
dependence of the “effective” dielectric constants in the metal as
!
0 0
ωp2
eff =  − 2 (2.330)
ω
!
00 00
Γωp2
eff =  + 3 (2.331)
ω

The valence electrons in a metal are very tightly bound with a large energy gap between
the valence ground and excited states. Therefore we can to good approximation write
0 = 1 and 00 = 0. The effective, frequency-dependent dielectric constants then simplify
to
!
2
ωp
0eff ' 1 − 2 (2.332)
ω
!
00
Γωp2
eff ' (2.333)
ω3
 Exercises 57

Fig. 2.12 Plane wave propagating from left to right, reflected at the interface.

Returning to Eq. 2.329 we see that at high frequency

ωp2
σ → iσ 00 = i 0 (2.334)
ω
and the E-field in the conductor becomes
r
ω0
Et = T E0 → 2e−iπ/2 E0 (2.335)
σ 00
The E-field in the conductor now lags the incident driving field by π/2 and the con-
duction current, from Eq. 2.327, becomes
ωp2 ωp2 p
Jc → i 0 Et = i 0 T E0 = 2 σ 00 ω0 E0 = 2ωp 0 E0 (2.336)
ω ω
We see that, in the optical frequency regime, the conduction current induced in the
metal is in phase with the driving field, proportional to the incident amplitude, and
independent of the driving field frequency.

2.11 Exercises
1. A plane wave propagates in free space with an E-field and B-field given by
Eqs. 2.184 and 2.186, calculate the Poynting vector S associated with this wave.
2. As shown in Fig. 2.12 a plane wave of wavelength 620 nm is incident from the
left on a lossless, nonmagnetic dielectric slab with dielectric constant 0 = 2.25.
Calculate the reflection and transmission coefficients R, T , the reflected power
and the transmitted power, assuming and E-field amplitude of 1 V/m. Assume
the dielectric slab to be infinitely thick.
3. Now reconsider the dielectric slab in Fig. 2.12 as a slightly lossy dielectric with an
index of refraction corresponding to SiO at λ0 = 620 nm. At this wavelength n =
η + iκ = 1.969 + i0.01175. Calculate the reflection and transmission coefficients,
the reflected power and the transmitted power for this case.
58 Elements Classical Field Theory

4. Silicon nitride, Si3 N4 , is a very low-loss dielectric material commonly used in

micro- and nanofabrication. The index of refraction of silicon nitride in the visible
region of the spectrum is 2.05. Calculate the ratio of the impedance of a plane
wave propagating in Si3 N4 to that of a plane wave propagating in free space.
5. The bulk plasma frequency of silver (Ag) metal is, ωp = 1.3 × 1016 s−1 . A plane
wave impinging on a silver surface penetrates to a skin depth of about 10 nm.
Calculate the current density Jc induced in the Ag near the surface.
6. Consider a hydrogen atom subject to a constant, uniform electric field. Suppose
the hydrogen atom consists of a positive point charge with a proton mass, sitting
at the origin. A negatively charged uniform electron charge density cloud rep-
resents the electron. The volume integral of the electron charge density is equal
to the proton charge. Because of the external E-field, the electron charge cloud
is displaced from the proton. Use Gauss’s law to calculate the restoring force
between the charge centers. Show that this restoring force is equivalent to that
of a harmonically vibrating spring and calculate the “natural” hydrogen atom
frequency.
7. With the electron in the ground state of the hydrogen atom the total energy
of the system is known to be −13.6 eV.3 The total energy is the sum of the
kinetic energy of the system in the ground state and the potential energy. If the
kinetic energy of the electron is 13.6 eV, what is the potential energy (due to the
Coulomb attractive force between the electron and proton)? Use your answer and
the Planck relation E = ~ω to calculate the corresponding frequency ν. Compare
this frequency to the harmonic oscillator frequency you calculated in Exercise 6.
Remember the angular frequency and cyclic frequency are related by ω = 2πν.
8. Redo Exercise 6 above supposing that the applied E-field is harmonically oscillat-
ing with an angular frequency ω. (a) Estimate the average value of the oscillating
dipole established between the positive and negative charge centers of the atom.
(b) Calculate the rate of energy (in Watts) emitted by the radiating dipole.
9. The ionosphere can be considered charged-particle plasma gas. Calculate the
plasma resonance frequency ωp , using the Drude model. Assume that the density
of free electrons is between 104 and 106 cm−3 . Remember that, according to the
dispersion diagram for the charged-particle-plasma-gas model, waves can propa-
gate into the plasma above ωp but are reflected below it. In radio transmission,
above what frequency are we broadcasting to the solar system (and beyond)?

2.12 Further Reading

1. D. J. Griffiths, Introduction to Electrodynamics, 3rd edition, Pearson Addison
Wesley, Prentice Hall, 1999.
2. J.-P. Pérez, R. Carles, R. Fleckinger, Électromagnétism Fondements et applica-
tions, 3ème édition, Masson, 1997.
3. H. A. Haus, Waves and Fields in Optoelectronics, Prentice-Hall, 1984.
4. D. M. Pozar, Microwave Engineering, 3rd edition, John Wiley & Sons, 2005.
3 The unit of energy “eV” is the electron volt and is the energy acquired by an electron after
passing through a potential difference of one volt. On eV is equivalent to 1.602 × 10−19 J.
 Further Reading 59

5. M. Mansuripur, Field, Force, Energy and Momentum in Classical Electrodynam-

ics, Bentham e-Books, 2011.
6. F. Nunes, T. Vasconcelos, M. Bezerra, J. Weiner, Electromagnetic energy density
in dispersive and dissipative media Journal of the Optical Society of America B,
vol 28, pp 1544-1552 (2011).
Complements

60
A
Energy Flow in Polarizable Matter

A.1 Poynting’s Theorem in Polarizable Material

Recall from Chapter 2, Eq. 2.59 that the energy flow across the surface of a spatial
volume is I
dEmech dEem
+ = − E × H · da (A.1)
dt dt
and that the integrand on the right side is identified with the energy flux crossing the
surface,
S=E×H (A.2)
with S called the Poynting vector. The two terms on the left are the rate of change of
“mechanical” energy stored in any material included in the bounded volume and the
rate of change of “field” energy associated with the E- and B-fields within the same
volume. Let us deconstruct some of the characteristics of the Poynting vector to bring
out the explicit time-dependent terms. Firstly

∇ · S = ∇ · (E × H) (A.3)
= −E · (∇ × H) + H · (∇ × E) (A.4)

From Maxwell-Faraday law and the Maxwell-Ampère law, Eqs. 2.27, 2.28, we have
∂B ∂H
∇×E=− = −µ0 (A.5)
∂t ∂t
and
∂D
∇×H= + Jfree (A.6)
∂t
so that
   
∂D ∂H
∇·S=−E· + Jfree + H · −µ0 (A.7)
∂t ∂t
 
∂D ∂H
= − E · Jfree + E · + µ0 H · (A.8)
∂t ∂t

Now specify the case where there no free currents (Jfree = 0), but material within the
enclosed volume is polarizable. Then the displacement field is

D = 0 E + P
62 Polarizable Matter

and
 
∂D ∂H
−∇ · S = E · + µ0 H · (A.9)
∂t ∂t
 
1 ∂ 1 ∂ ∂P
= 0 · (E · E) + µ0 (H · H) + E ·
2 ∂t 2 ∂t ∂t
 
∂ 1 ∂P
0 E 2 + µ0 H 2 + E ·


= (A.10)
∂t 2 ∂t

The term in brackets on the right side of Eq. A.10 is the field energy and the second
term is the time rate of change of the “mechanical energy” of the material. We see that
the flow of energy into or out of any material included within the volume is controlled
by the time dependence of the polarization field in that material.

A.2 Harmonically Driven Polarization Field

The polarization can be considered a density of harmonic dipole oscillators charac-
terized by a “natural” frequency ω0 and a damping frequency γ. If a harmonically
oscillating external electric field E is applied to this dipole density, the equation of
motion of the response will be

∂2P ∂P
+γ + ω02 P = ωd2 0 E (A.11)
∂t2 ∂t
where ωd2 is a “strength of external field coupling” parameter and E is the external
driving field. The coupling parameter ωd clearly is the frequency of the driving field
E. After a few lines of tedious but straight-forward algebra we find
stored
z }| {
 dissapative
z }| {
 2  2
∂P 1 1 ∂   ∂P 2 2 
 γ ∂P
E· = + ω0 P  + (A.12)
∂t 0 ωd2 2 ∂t  ∂t | {z }  0 ωd2 ∂t

| {z } potential
kinetic

Remembering that for a polarizable but nonmagnetic material the bound current den-
sity is given by,
∂P
Jbound = (A.13)
∂t
and
dEmech
= E · Jbound (A.14)
dt
we see immediately that Eq. A.12 represents the flow of mechanical energy into and
out of the ensemble of electric dipoles constituting the material of the system. Three
terms constitute this mechanical energy: the kinetic and potential energy stored in the
dipoles and the dissipated mechanical energy associated with the damping rate γ as
denoted in Eq. A.12.
 Drude-Lorentz Dispersion 63

A.3 Drude-Lorentz Dispersion

Equation A.12 describes the rate of change of mechanical energy in terms of the
time dependence of the polarization field, but the response of matter to an external
driving field varies with the frequency. If the response to the driving field is linear, the
polarization field is proportional to the applied electric field.
P = 0 χE (A.15)
where χ is the linear susceptibility of the material. The susceptibility is frequency
dependent and can be complex, χ(ω) = χ0 (ω)+iχ00 (ω). The relation characterizing the
material frequency response is called the material dispersion and is usually expressed
as the frequency dependence of the complex dielectric constant (ω) = 0 (ω) + i00 (ω).
For linear materials the relation between susceptibility and dielectric constant is given
by
χ(ω) = (ω) − 1 (A.16)
A simple dispersion relation for metals is the Drude-Lorentz model that expresses the
frequency dependence of the dielectric constant.
ωp2
(ω) = 1 + (A.17)
ω02 − ω 2 − iγω
In Eq. A.17 ω is the driving frequency, ω0 is the resonance frequency, γ is the irrecov-
erable rate of energy dissipation to “heat”, and ωp is the bulk plasmon frequency of
the electrons in the conduction band of the metal. Separating the real and imaginary
parts of ,
"
# " #
ωp2 ω02 − ω 2 γωp2 ω
(ω) = 1 + 2 2
+i 2 2
(A.18)
(ω02 − ω 2 ) + (γω) (ω02 − ω 2 ) + (γω)
= 0 + i00
Now from the constitutive relation between the displacement field and the electric field
and the definition of the displacement field we can write
D = 0 E (A.19)
= 0 E + P (A.20)
So we can write the polarization field in terms of the dielectric constant and the electric
field as
P = ( − 1)0 E (A.21)
Therefore we can express the factor in square brackets in the stored energy part of
Eq. A.12 " #
 2 2
∂P ∂E
+ ω02 P2 = ( − 1)2 20 + ω02 E2 (A.22)
∂t ∂t
The right hand side of Eq A.22 is complex due to the (−1)2 factor. In order to develop
expressions for the stored and dissipated energies it is therefore convenient to express
64 Polarizable Matter

the fields in complex representation and calculate the absolute value of the energy
terms. In complex notation the E-field and its time derivative are written
∂E
E = E0 e−iωt = −iωE
and
∂t
and the absolute value of the stored energy factor is
 2 " 2 #
∂P 2 2 2 2 ∂E 2 2
+ ω0 P = ( − 1) 0 + ω0 E
∂t ∂t
 2
∂P 2
h 2
i
+ ω02 |P| =20 ω02 + ω 2 (0 − 1)2 + (00 ) E02 (A.23)
∂t
Now from the dispersion relation Eq. A.17 we can also write

2
0 2 00 2
ωp4 ω02 − ω 2 γ 2 ωp4 ω 2
( − 1) + ( ) = h i2 + h i2 (A.24)
2 2
(ω02 − ω 2 ) + (γω)2 (ω02 − ω 2 ) + (γω)2
ωp4
= 2 (A.25)
(ω02 − ω 2 ) + (γω)2
Therefore, from Eqs. A.22, A.23, A.25 we have the absolute value of the stored energy
from Eq. A.12
" 2 # "
#
1 ∂P 2 2 0 ωp2 ω02 + ω 2
+ ω0 |P| = E02 (A.26)
20 ωp2 ∂t 2 (ω02 − ω 2 )2 + (γω)2
By “stored energy” we mean that part of the energy density that flows from the fields
into the dipoles, is stored there and can flow back to the fields.
Making analogous substitutions we have for the dissipative part of dEmech /dt,
2
γωp2 ω 2

γ ∂|P|
= 2 0 E02
0 ωp2 ∂t 2 2
(ω0 − ω ) + (γω)2

= 00 0 ωE02 (A.27)

This dissipative loss is irreversible and cannot be recovered by the fields.
Returning to the absolute value of the stored energy, the factor in square brackets
on the right side of Eq. A.26 can be written as the sum of two fractions,


ωp2 (ω 2 + ω02 ) ωp2 ω02 − ω 2 2ω 2 ωp2
2 = 2 + 2
(ω02 − ω 2 ) + (γω)2 (ω02 − ω 2 ) + (γω)2 (ω02 − ω 2 ) + (γω)2
and from Eq. A.18 we can identify the first term as
2ω 00
0 − 1 and the second as 
γ
Therefore " 2 #  
1 ∂|P| 2 0 2ω 00
+ ω02 |P| = (0 − 1) +  E02 (A.28)
20 ωp2 ∂t 2 γ
 Drude-Lorentz Dispersion 65

A.3.1 Stored Energy Density

From Eqs. A.10, A.12, and Stokes’ theorem we can write the energy conservation the-
orem as
Z Z    2 ! 
∂ 1 2 2 1 ∂P 2 2
− S · da = 0 E + µ0 H + + ω0 P dV +
A V ∂t 2 0 ωp2 ∂t
Z  2
γ ∂P
dV (A.29)
V 0 ωp2 ∂t
Then taking the absolute value of the right side of Eq. A.29 and substituting Eq. A.27
and Eq.A.28 into it, we have
Z Z     
∂ 1 2ω 00
− S · da = 0 0 +  E 2 +µ0 H 2 dV +
A V ∂t 2 γ
Z
0 00 ωE 2 dV (A.30)
V

and writing the magnetic field energy in terms of the electric field energy
0 0 p 0 2
H2 = ||E 2 =  + 00 2 E 2 (A.31)
µ0 µ0
the stored energy density Us within the curly brackets on the right side of Eq. A.30
can be written   p 
0 0 2ω 00 2 2
Us =  + 0
 +  + 00 E2 (A.32)
2 γ
and averaging Us over an optical cycle
  p 
0 2ω 00
hUs i = 0 +  + 0 2 + 00 2 E02 (A.33)
4 γ
In the limit of a lossless material, 00 , γ → 0,
1 0 2
hUs i = 0  E 0 (A.34)
2
as expected.
Finally, we can express this conservative energy density in terms of the real and
imaginary parts of the index of refraction, η and κ, respectively.
0 = η 2 − κ2


(A.35)
00 = 2ηκ (A.36)
So that  
2 2ωηκ
U s = 0 η + E2 (A.37)
γ
and averaging Us over an optical cycle,
 
0 2 2ωηκ
hUs i = η + E02 (A.38)
2 γ
66 Polarizable Matter

A.3.2 Dissipated Energy Density

The rate of energy density dissipation ∂Ud /∂t is given by the integrand of the second
term on the right side of Eq. A.30
 2
∂Ud γ ∂P
= (A.39)
∂t 0 ωp2 ∂t

Remembering that P = 0 ( − 1)E we have

h i
2 2
P2 = 20 (0 − 1) + (00 ) E 2 (A.40)

As usual, we take E to be a harmonically oscillating field, E = E0 e−iωt . Then

q
2 2
P = 0 (0 − 1) + (00 ) E (A.41)

and
∂P 2 ∂E
q
2
= 0 (0 − 1) + (00 ) (A.42)
∂t ∂t
So substituting A.42 into A.39 we have

∂Ud γ0 ω 2 h 0 2 2
i
= 2
( − 1) + (00 ) E 2 (A.43)
∂t ωp

where E0 is the electric field amplitude. This expression can be further simplified by
noting that
2 2 ωp4
(0 − 1) + (00 ) = 2
(ω02 − ω 2 ) + (γω)2

and therefore, taking into account A.27,

∂Ud
= 0 00 ωE 2 (A.44)
∂t

Averaging over an optical cycle,

 
∂Ud 1 00
= 0  ωE02 (A.45)
∂t 2

and the average energy dissipated in one optical cycle (T = 2π/ω) is

hUd i = π0 00 E02 (A.46)

 Drude-Lorentz Dispersion 67

A.3.3 Time Dependence of Stored and Dissipated Energy Density

Representing the driving E-field as a real quantity, E = E0 cos ωt, we see from Eq. A.37
that the stored energy is
 
2 2ωηκ
Us = 0 η + E02 cos2 ωt (A.47)
γ

and is therefore in phase with the driving field energy flux, Sz = Ex Hy cos2 ωt. There-
fore the rate of change of stored energy is in quadrature
 2
dUs 2 2ωηκ
= −2ω0 η + E02 cos ωt sin ωt (A.48)
dt γ
In contrast from Eq. A.44 the rate of change of the dissipated energy density is in
phase with the driving field energy flux
dUd
= 0 00 ωE02 cos2 ωt (A.49)
dt
while the dissipated energy itself accumulates over the optical cycles
 
ωt sin 2ωt
Ud = 0 00 E02 + (A.50)
2 4
Figure A.1 shows the time dependence of the various constituents of the energy density
in a polarizable medium.

A.3.4 Frequency Dependence of Stored and Dissipated Energy

Returning to the expression for the stored energy in terms of the real and imaginary
parts of the dielectric constants, Eq. A.32, and taking the limit of negligible loss,

Us ' 0 0 E 2 (A.51)

The dispersion relation of the Drude-Lorentz model for 0 is

0
ωp2 ω02 − ω 2
 (ω) = 2 2
+1 (A.52)
(ω02 − ω 2 ) + (γω)
and therefore the frequency dependence of the stored energy in lossless media is
"
#
ωp2 ω02 − ω 2
Us (ω) ' 0 2 2
+ 1 E2 (A.53)
(ω02 − ω 2 ) + (γω)

The stored energy exhibits the same “dispersive” frequency dependence as a driven
harmonic oscillator, in phase with the driving field below resonance, ω0 − ω > 0 and
π out of phase at frequencies above resonance, ω0 − ω < 0. In the simplest case of
a metal modeled with a single resonance at the bulk plasmon frequency, ω0 = ωp ,
the stored energy responds in phase at all frequencies “to the red” of ωp . For most
68 Polarizable Matter

Fig. A.1 Top panel: Red (full) curve shows time dependence of stored energy (normalized
to 0 E02 ), Eq. A.47, and blue (dashed) curve shows time dependence of the rate of change
of stored energy, Eq. A.48. Bottom panel: Red (full) curve shows accumulation of dissipated
energy with time, Eq. A.50 and blue (dashed) curve shows rate of change of dissipated energy
(dissipated power), Eq. A.49. These graphs are plotted with the following parameters taken
for silver metal: 0 = −9.108, 00 = 0.753, ω = 3.768 × 1015 s−1 ,(λ = 500 nm), γ = 1.0 × 1014 .

common noble metals ωp is in the blue or near ultraviolet regions of the spectrum.
Therefore, for incident driving fields in the visible or near-infrared, the usual case in
plasmonic studies, the stored energy is in phase with the incident energy flux.
The dispersion relation for the material loss term is
γωωp2
00 (ω) = 2 2
(A.54)
(w02 − ω 2 ) + (γω)
and, using Eq. ?? the dissipative energy density is peaked at the resonance frequency
ω0 , decreasing symmetrically as the driving frequency is tuned above or below reso-
 Drude-Lorentz Dispersion 69

nance. " #
γωωp2
 
ωt sin 2ωt
Ud = 0 2 2
E02 + (A.55)
(w02 − ω 2 ) + (γω) 2 4
The dispersive and absorptive frequency profiles are simply those of a driven, damped
harmonic oscillator. This behavior is consistent with the polarization of the material
modeled as a density of dipoles.
B
Macroscopic Polarization from
Microscopic Polarizability

B.1 Introduction
In Chapter 2 the polarization vector field was introduced in Eqs. 2.15, 2.17. The polar-
ization field P adds vectorially to the applied electric field (multiplied by the permittiv-
ity) 0 E to produce the total displacement field D, and in many cases the polarization
field itself is proportional to the applied electric field. The proportionality constant,
the susceptibility χ, is a property of the material; and the dielectric constant of the
material is simply related to this susceptibility by  = 1 + χ. As the term suggests, the
susceptibility is a measure of the response of the material to an external applied field.
Since matter is actually composed of atoms and molecules, the macroscopic response
must somehow be related to the distortion of local electric fields around the constituent
microscopic particles. The goal of this complement is to develop that relation.

B.2 Electric Field inside a Material

At the truly microscopic distance scale, comparable to the distance between atoms in
condensed matter, the electric field due to electronic and nuclear charge densities is
extremely complicated and constantly changing in time due to the orbital motion of
electron charge densities around nuclei and the thermal motion of the nuclei them-
selves. The calculation of this field is hopeless, but in fact it is not the field we really
seek anyway. What we really want is the field at a fixed point inside the material due
to two influences: the external E-field applied to the bulk material and the E-field
internal to the material arising from its polarization. Figure B.1 is a sketch of the var-
ious elements of the problem. The figure shows a slab of material with susceptibility
χ, dielectric constant , placed between two conducting plates across which an electric
field Eext is applied. The field at a point internal to the material can be considered the
sum of two superposed fields. The source of the first field is the charge density induced
at the surface of the material by the applied external field. This surface charge density
produces a uniform polarization field P opposite in direction to the applied E-field.
The E-field due to this polarization field is Eint = −P/0 . The source of the second
field are the microscopic induced dipoles themselves. As we have already discussed
above, at a spatial resolution of an individual dipole, the E-field is neither practically
calculable nor useful. We seek instead the field at a specific point due to the average
influence of many induced dipoles. To this end we construct a sphere whose volume
is large with respect to the volume of an individual dipole but negligible compared to
 Electric Field inside a Material 71

Fig. B.1 External electric field Eext is applied to two external conducting parallel plates.
The material slab with dielectric constant  and susceptibility χ exhibits a polarization field
−P the source of which is the surface charge density σP induced at the slab-plate interface.
The volume of the sphere inside the slab is large with respect to the size of individual dipoles
comprising the material but small compared to the bulk volume of the slab itself.

the bulk volume. Using Gauss’s law, we can compute this E-field EP at the center of
the sphere due to the bound charge density around the surface. The net field at the
center of the sphere, experienced by a molecule at that site, is then Eext − P/0 + EP .

The displacement field D = 0 Eext must be continuous across the slab boundary,
and therefore the E-field at the point of interest due to the external applied field is

D P
= Eext + (B.1)
0 0

The surface charge density at the two slab interfaces σPslab is the source of a polariza-
tion field at all points within the slab equal to −P/0 . Now we invoke Gauss’s law to
calculate the E-field EP at center of the sphere from the polarization charge density
σPsphere at the sphere surface. This bound charge density is determined from the po-
larization component normal to the sphere, and therefore the distribution of charge
density around the surface goes as σPsphere (θ) = |P| cos θ where θ is the angle between
the z axis and the surface normal as shown in Fig. B.1. The E-field at the sphere center
is then
72 Polarization and Polarizability

σPsphere cos θ
Z
1
EP = dS (B.2)
4π0 S r02
where r0 is the sphere radius. The surface differential is dS = r02 sin θ dθ dϕ, and
integrating over the azimuthal angle ϕ we have

dS = 2πr02 sin θ dθ (B.3)

Putting σPsphere (θ) and dS into Eq. B.2 we have

π
|P|2πr02
Z
EP = sin θ cos2 θ dθ (B.4)
4π0 r02 0
|P|
= (B.5)
30
and
P
EP = (B.6)
30
The effective E-field at the center of the sphere, experienced by a material molecule
placed at that position, is therefore
D P
Eeff = − + EP (B.7)
0 0
P P P
= Eext + − + (B.8)
0 0 30
P
Eeff = Eext + (B.9)
30
Note that the effective E-field does not depend on the radius of the sphere. Any sphere
will yield the same result as long as the spherical volume is not comparable to an
individual dipole element. The only question that remains is the possible contribution
of the E-fields of individual dipole elements inside the sphere. Since the material was
initially assumed isotropic with no net polarization in the absence of an external field,
we can safely assume that there is no E-field contribution to the dipole elements
themselves.

B.3 Polarization and Polarizability

Now we want to use Eeff to obtain the macroscopic polarization field from the micro-
scopic dipole moment induced in each molecule. The connection from the micro- to
the macro-world is through the polarizability, a property of matter at the molecular
level. The polarizability α is defined through the relation,

p = α0 Eeff (B.10)

where p is the induced dipole moment of an individual molecule. Since the macroscopic
polarization P can be interpreted as a dipole density, we can write
 Polarization and Polarizability 73

 
P
P = N p = N α0 Eeff = N α0 E + (B.11)
30

where N is the number of molecules per unit volume. Remembering that for linear
response
P = 0 ( − 1)E (B.12)
We can set B.11 equal to B.12 and find
−1 Nα N0 ρm α
= = (B.13)
+2 3 3M
where in the last term on the right N0 is Avogadro’s constant, ρm is the density and M
is the molecular mass. This expression is called the Clausius-Mossotti relation. It tells
us how we can calculate the dielectric constant , a macroscopic property, from the
microscopic polarizability, α. As such it furnishes a key bridge between microscopic
and macroscopic physics of matter.
C
The Classical Charged Oscillator and
the Dipole Antenna

C.1 The Proto-antenna

The classical charge dipole, p0 = qa, oscillating at frequency ω, the electromagnetic
fields of which we considered in Section 2.8, emits radiation of wavelength λ = 2πc/ω
and can therefore be considered a prototypical or microscopic antenna radiative source.
We recall Eqs. 2.150, 2.152, the far-field solutions to Maxwell’s equations for a small
dipole, a << λ.

p0 sin θ ω 2 h  r i
Eθ = − cos ω t − θ̂ (C.1)
4π0 rc2 c
µ0 p0 sin θ ω 2 h  r i
Bϕ = − cos ω t − ϕ̂ (C.2)
4π rc c
and taking account of µ0 = 1/0 c2 we note that the electric and magnetic field ampli-
tudes are related by, B = E/c. We can associate a characteristic “current” ic of the
dipole with length a oscillating at frequency ω,
ωp0
ic = (C.3)
a
and a charactersitic electric field Ec
 1/2
µ0 ic a sin θ
Ec = (C.4)
0 2λr
where λ is the wavelength of light at frequency ω. We recall from Eq. 2.155 that the
cycle-averaged energy flux emitted by the oscillator is given by

µ0 p20 ω 4 sin2 θ
 
hSi = (C.5)
32π 2 c r2
or in terms of ic and Ec ,
 1/2  2
1 µ0 ic a sin θ
hSi = (C.6)
8 0 λr
Figure C.1 shows the distribution of radiated power density from a point dipole. The
 The Proto-antenna 75

Fig. C.1 Angular distribution of power density (power radiated through a unit area) from
a radiating point dipole

cycle-averaged total power radiated over all space is

 1/2   2
1 ω 4 p20 2π µ0 a 2 ic
hPi = = (C.7)
4π0 3c3 3 0 λ 2

where we have used the fact that ω 4 p20 = ω 4 p20 /2. Now from a units analysis it is
easy to show that
 1/2
µ0
' 377 ohms (C.8)
0
and the cycle-averaged power emitted by the oscillator can be set equal to the average
power required to drive the oscillator
1
hPi = Rc i2c (C.9)
2
where Rc is the characteristic oscillator impedance to radiation,
 1/2
2π µ0
Rc =
3 0
=790 ohms

and for a given dipole length and wavelength (a << λ) the radiative impedance Rr is
 1/2  
2π µ0 a 2
Rr = (C.10)
3 0 λ
76 Dipole Antenna

Fig. C.2 Geometric disposition for determining the E-field of a line antenna of length R at
point M by integrating contributions from point dipoles aligned along z.

Finally the directional gain of any antenna, G(θ), is defined as the ratio of power
emitted in a given direction to the total radiated power. In the case of the dipole
proto-antenna, from Eqs. C.5, C.7,

hSi 3
G(θ) = 4πr2 = sin2 θ (C.11)
hPi 2

C.2 Real Antennas

We consider a real, practical antenna consisting of charge oscillating along a straight
conductor of finite length, comparable to the wavelength of the emission. We suppose
that the electronmagnetic field resulting from such an antenna can be analyzed as
a linear combination of proto-antennas arranged along the length of the conductor.
Figure C.2 shows the antenna aligned along the z axis. At point M the field is due
to the linear superposition of contributions from the proto-antennas arranged along z.
One of these point dipoles is shown at position P along z. The E-field at M for one
proto-antenna is
 1/2
µ0 sin θ
i(z)dz e[−iω(t− c )] θ̂
PM
E= (C.12)
0 2λP M

We seek to write the distance P M in terms of r and then integrate the resulting
expression for E over the length of the antenna. Invoking the law of cosines we can
write
 Real Antennas 77

 2 !1/2
2 2

1/2 OP 2OP cos θ
P M = r + OP − 2r OP cos θ =r 1+ − (C.13)
r r

At distances far from the antenna where r >> OP we have P M → r. Then substitut-
ing this result into C.12 and using k = ω/c we have
 1/2 "Z #
+R/2
µ0 sin θ
dz e−[ω(t− c )] θ̂
r
ikz z
E(M ) ' ic (z)e (C.14)
0 2λr −R/2

where  
2π
kz = k cos θ =
λ
C.2.1 The Half-wave Antenna
We posit that the antenna length is λ/2 and that a proto-dipole oscillator source,
located at the midpoint, is driving current to the antenna extremeties. A standing
wave is set up along the antenna such that
 
2πz
ic (z) = ic cos = ic cos (kz) (C.15)
λ
Then the integral in Eq. C.14 becomes
Z λ/4  
2πz
Ic (k) = ic cos eik cos θz dz (C.16)
−λ/4 λ
and the resulting E-field at P is
 1/2    θ


µ0 ic cos π cos 2
h  r i
Ehw (P ) = − cos ω t − θ̂ (C.17)
0 2πr sin θ c
The cycle-averaged energy flux for the half-wave dipole antenna, hSihw , is

ic cos2 π cos θ2
 2 

1
hSihw = (C.18)
8π 2 0 c r2 sin2 θ
and the total cycle-averaged power of the half-wave antenna is

1 2 π cos2 π cos θ2
Z 

hPihw = i dθ (C.19)
4π0 c c 0 sin θ
hPihw =36.5i2c
1
hPihw = i2c Rr
2
and therfore the radiative impedance of the half-wave antenna is Rr = 73 ohms. A
typical FM radio emission frequency is 100 MHz, and from the above calculation we
see why a typical FM radio antenna consists of a 75 ohm cable about 1.5 meters in
length.
78 Dipole Antenna

Fig. C.3 Plot of the radiation distribution, normalized to the maximum at θ = π/2, from a
linear half-wave antenna and single point-dipole. Angles 78◦ and 90◦ correspoind to half-power
points for half-wave antenna and point dipole, respectively.

C.2.2 Array of Half-wave Antennas

The directionality of the emission for a single half-wave antenna is given by Eq. C.18
and Fig. C.3 shows that it is only slightly more peaked along the θ = π/2 direction
than the proto-dipole oscillator. The emission can be made more directional by im-
plementing an array of half-wave dipole antennas along the x axis as indicated in
Fig. C.4. The E-field of the array at any point is essentially the linear superposition of
the E-fields of the individual elements, each element offset by a phase shift related to
the array pitch d along x. Suppose we have N = 2n + 1 array elements symetrically
arranged along ±x with n elements on each side of the origin. The the field at point
P is simply,
m=n
X
Earray (P ) = Ehw (P ) e−imϕ
m=−n

m=(N −1)
X
= Ehw (P )einϕ e−imϕ
m=0

1 − e−iN ϕ
= Ehw (P )einϕ
1 − e−iϕ
sin(N ϕ/2)
= Ehw (P ) (C.20)
sin(ϕ/2)

where ϕ is the azimuthal phase shift associated with the displacement a along x.
Figure C.5 shows the angular distribution of the E-field of a five-element linear array
as shown in Fig. C.4.
 Real Antennas 79

Fig. C.4 Five λ/2 antennas arranged in a linear array along the x axis, each element of the
array separated by a distance d.

2π
ϕ = kx a = sin θ cos ϕ a
λ

Figure C.5 was determined from an oscillation frequency of ν = 1 × 109 Hz and a

Fig. C.5 Angular E-field distribution of a half-wave antenna array with five elements.

wavelength λ = 0.3 m. The separation between the elements of the array is d = λ/2 =
0.15 m. We see that the maximum of the principal lobes of intensity occur for θ = π/2
80 Dipole Antenna

Fig. C.6 Angular distribution of radiated power from the five-element array shown in
Fig. C.4. Note that the power directed along x with half-power points of 10.4◦ compared
to a single half-wave antenna with half-power points at 78◦ .

but we see small adjacent lobes, reminiscent of a diffraction pattern. This result is
hardly surprising since Eq. C.20 is the same as for linear diffraction grating of N
elements. The corresponding radiated power distribution of the five-element array is
shown in Fig. C.6. The concentration of the radiated power along x is clearly evident.
The angular aperture of the half-power points in the five-element array is only 10.4◦
compared to 78◦ for the single-element antenna.
3
Surface Waves

3.1 Introduction
We treat here an important wave phenomenon, the surface wave. It occurs in many
physical systems both mechanical and electromagnetic. Mechanical surface waves can
exist at the interface between two media with different densities such as the oceans
and earth. Ordinary sea waves and tsunamis are surface waves. Seismic events can pro-
duce both longitudinal and shear waves at the earth’s surface, resulting in earthquakes.
Electromagnetic surface waves can occur at the interface between dielectric and con-
ductive media or between two dielectrics satisfying appropriate boundary conditions.
Air-salt water or glass-metal interfaces support electromagnetic surface waves. Here
we focus on electromagnetic waves at the interface between common dielectrics and
noble metals such as air or glass and silver or gold. We will characterize these waves by
their distinctive properties of dispersion, spatial field distribution, and polarization.
These properties will be used to analyze useful phenomena such as wave guiding within
subwavelength structures and spatial light localization below the diffraction limit.

3.2 History of Electromagnetic Surface Waves

The impetus for the study of electromagnetic surface waves began with the devel-
opment of radio communication near the end of the nineteenth century. Ordinary
Maxwellian “space” waves cannot directly propagate beyond the horizon, although
Marconi had been able to demonstrate long-range radio communication including
transatlantic messaging between 1899 and 1901. A possible explanation for this suc-
cessful over-the-horizon transmission was the excitation and propagation of electro-
magnetic waves at the air-earth or air-sea interface that could be “guided” over the
curvature of the planet. It was therefore necessary to find solutions to Maxwell’s equa-
tions that could explain the observations. Stimulated by Arnold Sommerfeld’s analysis
of electromagnetic wave propagation along a single conducting wire in 1899, his stu-
dent J. Zenneck presented a paper on planar surface electromagnetic waves in 1907.
Zenneck found that Maxwell’s equations had a solution corresponding to a wave cou-
pled to a flat surface interface at the boundary between a dielectric (air) and a medium
with finite conductivity (salt water). The configuration used by Zenneck is illustrated
in Fig. 3.1. Soon after Zenneck’s article appeared Sommerfeld again published an in-
fluential paper in 1909 in which he analyzed the propagation of radio waves along the
surface of the earth. He considered a vertical oscillating dipole (radio antenna) per-
pendicularly positioned to the interface between two media: one dielectric, the other
conducting. Sommerfeld studied secondary waves produced by the discontinuity at the
82 Surface Waves

Fig. 3.1 Schematic arrangement for Zenneck wave generation.

interface between air and land, considered as a weakly conductive material. His anal-
ysis showed that the Hertz vector Π, that describes the electromagnetic field in each
medium is given by the sum of two fields, P and Q corresponding to two waves. The
wave Q represents a space propagating wave and P a surface pguided wave. For long
distances R from the source, Q varies as (1/R), where R = (r2 + z 2√ ), (see Fig. 3.1
for a definition of the coordinates) while P varies asymptotically as 1/ r and is con-
fined to the Earth’s surface. Over the course of the next two decades a controversy
arose over the existence of the so-called Zenneck waves because Sommerfeld’s original
analysis contained a sign error, and it was found that it was difficult to excite the
surface-guided wave with practical antenna sources or to detect and distinguish the
true surface wave from the spatially propagating wave very close to the surface. In
any event, as far as long-distance wireless communication was concerned, the whole
question became irrelevant when it was discovered that reflection from the ionosphere
was the principal agent responsible for over-the-horizon radio transmission.

3.3 Plasmon Surface Waves at Optical Frequencies

Recent interest in electromagnetic surface waves stems from their potential appli-
cations in nanoscale opto-electronic devices. In contrast to the Sommerfeld-Zenneck
analyses emphasizing radio frequencies, modern studies at optical frequencies usually
take Raether’s influential treatise, Surface Plasmons on Smooth and rough Surfaces
and on Gratings as their point of departure. The modern terminology for the Zennick
wave at the interface between a metal and a dielectric is the surface plasmon or the
surface plasmon-polariton (spp). At optical frequencies the surface plasmon exhibits
strong spatial localization at the dielectric-conductor interface and can be simply and
efficiently excited by appropriately polarized light incident on surfaces decorated with
features (slits, holes, grooves, etc.) at the subwavelength scale (tens to hundreds of
nanometers). Here, we consider the interface between two media, one dielectric and
 Plasmon Surface Waves at Optical Frequencies 83

Fig. 3.2 Configuration of field components for TE and TM polarized plane waves incident
on a dielectric-metal interface.

the other a metallic conductor. Each material is characterized by a relative permittiv-

ity, r : dr for the dielectric and m
r for the metal. In order to lighten notation we drop
the subscript r and write the relative permittivities (dielectric constants) as d and m
for the dielectric and metal, respectively. The metal also exhibits a high conductivity,
σ and obeys “Ohm’s law” E = σJ where E is the electric field applied to the metal and
J is the current density induced in it. We consider the dielectric medium as essentially
transparent (lossless) so that d = 0d where 0d is real. In the case of metals absorptive
loss in the optical region of the electromagnetic spectrum is significant, the dielectric
constant is complex, and m = 0m + i00m , where the imaginary term represents ab-
sorptive loss. Figure 3.2 shows two orthogonal linear polarization orientations of E-M
fields incident on the metal surface.1 The plane of incidence is defined by the incident
and reflected plane-wave propagation vectors k, and in Fig. 3.2 this plane is aligned
with the x − z plane. With H perpendicular to the plane of incidence (right panel,
Fig. 3.2) the polarization is called “transverse-magnetic” (TM). TM polarization re-
stricts the incident wave to field components Ex , Ez , Hy . With E perpendicular to
the same plane (left panel, Fig. 3.2) the polarization is “transverse-electric” (TE). The
field components of a TE polarized wave are Hx , Hz , Ey . We will consider continuity
conditions at the surface for each case.
We start with the two Maxwell curl equations, the Faraday law and the Ampère-
Maxwell law, Eqs. 2.27, 2.28, and posit two solutions: E- and H-fields characterized
by the propagation vector k = (2π/λ)k̂, where λ is the wavelength, and oscillating
harmonically in time with frequency ω.

E(r, t) = E0 (r)ei(k·r−ωt (3.1)

i(k·r−ωt
H(r, t) = H0 e (3.2)

1 Although in Chapter 2 the direction of propagation is always along z, in this chapter we must
consider both the incident propagating wave and the surface wave. The incident space-propagating
wave is taken to be incident along z, while the surface wave propagates along x.
84 Surface Waves

Application of the curl operation leads to the following relations,

∂Ey
= (iωµ0 )Hx (3.3)
∂z
∂Hx ∂Hz TE polarization
− = −i(ω0 )Ey (3.4)
∂z ∂x
∂Ey
= −(iωµ0 )Hz (3.5)
∂x
and
∂Hy
= (iω0 )Ex (3.6)
∂z
∂Ex ∂Ez TM polarization
− = (iωµ0 )Hy (3.7)
∂z ∂x
∂Hy
= −(iω0 )Ez (3.8)
∂x

The incident and reflected waves propagate in the dielectric medium characterized
by the dielectric constant d . The propagation vector in free space is denoted k0 and
within the dielectric kd = k0 n where n is the index of refraction. The dielectric con-
stant and the index of refraction of any medium are related by  = n2 . Since the
incident k = ki is in the x − z plane we can write

ki2 = kx2 + kz2 (3.9)

= d k02 (3.10)

Similarly on the metal side of the interface for the transmitted propagation vector k m
we have

kt2 = kx2 + kz2 (3.11)

= m k02 (3.12)

3.3.1 Surface Waves with TE Polarization

In TE polarization we find the wave solution for Ey by applying the posited plane-
wave expression, Eq. 3.1, to the relations Eqs. 3.3, 3.4, 3.5. The result (in the dielectric
medium) is
d d
Eyd (x, z) = E0y
d i[(kx x+kz z)−ωt]
e (3.13)
From this E-field component the H-field components on the dielectric side of the
interface are readily obtained.

i ∂Eyd kd
Hxd = − = z Eyd (3.14)
ωµ0 ∂z ωµ0
d
i ∂Ey kd
Hzd = = − x Eyd (3.15)
ωµ0 ∂x ωµ0
 Plasmon Surface Waves at Optical Frequencies 85

The same analysis applies to the metal side of the interface with the result
m m
Eym (x, z) = E0y
m i[(kx x+kz
e z)−ωt]
(3.16)

and
i ∂Eym km
Hxm = − = z Eym (3.17)
ωµ0 ∂z ωµ0
m
i ∂Ey km
Hzm = = − x Eym (3.18)
ωµ0 ∂x ωµ0

At the interface, z = 0, continuity conditions for Hx and Ey require that

Hxd (z = 0) = Hxm (z = 0) (3.19)

Eyd (z = 0) = Eym (z = 0) (3.20)

and therefore at the boundary

kzd = kzm (3.21)

√ √
d k0 = m k0 (3.22)

But this last equation can only be valid for k0 = 0 since the dielectric constants on the
two sides of the boundary can never be equal. Therefore there can be no wave solution
at the boundary for TE polarization.

3.3.2 Surface Waves with TM Polarization

With TM polarization we have three field components, Hy , Ex , Ez . We find the wave
solution for Hy on the dielectric side of the interface

Hyd (x, z) = Hy0

d i[(kx x+kz z)−ωt]
e (3.23)

Then using Eqs. 3.6, 3.8, we find expressions for Exd and Ezd .

kzd
Exd = H d (x, z) (3.24)
ω0 d y
kd
Ezd = − x Hyd (x, z) (3.25)
ω0 d

On the metal side of the interface we have

kzm
Exm = H m (x, z) (3.26)
ω0 m y
km
Ezm = − x Hym (x, z) (3.27)
ω0 m

Continuity conditions at the interface (z = 0) are

86 Surface Waves

Exd (x, 0) = Exm (x, 0) (3.28)

Hyd (x, 0) = Hym (x, 0) (3.29)
d Ezd (x, 0) = m Ezm (x, 0) (3.30)
Then substituting Eqs. 3.24, 3.25 and 3.26, 3.27 into the continuity conditions we find
that at the boundary
kxd = kxm = kx (3.31)
kzd kzm
= (3.32)
d m
The wave vector relations on the dielectric and metal sides of the interface, Eqs. 3.10, 3.12
allow us to write
(kzm )2 m k02 − (kxm )2
= (3.33)
(kzd )2 d k02 − (kxd )2
Then, using Eqs. 3.31, 3.32, we can eliminate the kz components and obtain an expres-
sion for kxs , the magnitude of the surface wave vector propagating along the boundary,
in terms of the free-space wave vector k0 and the dielectric constants on each side of
the interface. r
d m
kxs = k0 (3.34)
d + m
The z components of the surface wave, on the dielectric and metallic sides of the
boundary, can be easily obtained from
 

2 2 m
kzs,d = d k02 − (kxs ) k02 = 1 − d k02 (3.35)
m + d
 
2 2 d
(kzs,m ) = m k02 − (kxs ) k02 = 1 − m k02 (3.36)
m + d

d
kzs,d = ± √ k0 (3.37)
d + m
m
kzs,m = ±√ k0 (3.38)
m + d
Note that if the real part of the dielectric constant in the metal is negative and if
d + m < 0, then kzm , kzd will be imaginary. The Ez component of the surface wave on
both sides of the interface will then be evanescent.
d
kzs,d → ±iκds = ±i √ k0 (3.39)
d + m
m
kzs,m → ±iκm
s = ±i √ k0 (3.40)
d + m
The choice of sign (±) ensures that the wave amplitude decreases exponentially with
increasing distance (±z) from the interface. With the choice of coordinate axes as
shown in Fig. 3.2,
 Plasmon Surface Waves at Optical Frequencies 87

kzs,d → −iκds z<0 (3.41)

kzs,m → iκm
s z>0 (3.42)
and the components of the surface wave projecting onto the dielectric side take on the
following forms,
d s
Hys,d (x, z, t) = H0y eκs z ei(kx x−ωt) (3.43)
iκds d s
Exs,d (x, z, t) = − H0y eκs z ei(kx x−ωt) (3.44)
ω0 d
ks d s
Ezs,d (x, z, t) = − x H0y eκs z ei(kx x−ωt) (3.45)
ω0 d
Similar expressions obtain on the metallic side with the appropriate change of sign for
iκm
s (Eq. 3.42). Figure 3.3 shows how the electric and displacement fields of surface
wave decrease exponentially on both dielectric-metal interface. In addition to the ex-
ponential decrease in the ±z direction the plots in Fig. 3.3 also show that the absolute
magnitude of the dielectric constant on the metal side is typically much greater than
on the dielectric side. Therefore field penetration into the metal is limited to the skin
depth. For gold or silver in the optical interval (350 ≤ λ0 ≤ 850 nm) of the electro-
magnetic frequency spectrum significant field penetration is of the order 20 − 30 nm.
Furthermore the E-field is discontinuous at the boundary, due to the presence of sur-
face charge density, while the displacement field (D = 0 E) takes into account the
sign change in the dielectric constant across the boundary and is therefore continu-
ous (although the derivative of the D-field with respect to the boundary normal is
not continuous). Table 3.1 summarizes the various components of the electromagnetic
wave in the dielectric, the metal, and at the interface for TM polarization.
Equation 3.34 shows that kxs is complex since m is complex. The three components
of the surface wave Exs , Ezs , Hys propagate in the x direction at the interface. Taking
Exs as an example, s
Exs (x) = E0x
s ikx
e x = E0xs ik0 ns x
e (3.46)
and the argument of the exponential will have a real and imaginary term since kxs =
k0 ns is complex. The factor ns = ηs +iκs is the complex surface index of the refraction
at the interface. The imaginary term represents the exponential loss in amplitude
due to dissipation in the metal as the wave propagates along the surface. From the
definition of the surface index of refraction, ns = kxs /k0 , the complex dielectric constant
of the metal, m = 0m + i00m and Eq. 3.34 we can determine the corresponding complex
index of refraction.
r
m d
ns = (3.47)
m + d
00
0 d 1 + i 0 m

n2s = 0 m · m
00 (3.48)
m + d 1 + i 0 m
 +d m

For the two metals commonly used in plasmonic devices, silver and gold, |00m /0m | <<
1 over the optical range of interest, and this characteristic can be used to simplify
88 Surface Waves

Fig. 3.3 Vertical (z) components of the surface wave electric field (E-field) and displacement
field (D-field). Note that Ez is discontinuous at the surface while Dz is continuous. The units
of the E-field and D-field are arbitrary.
 Plasmon Surface Waves at Optical Frequencies 89

Fig. 3.4 Real part of the dielectric constant 0 vs. wavelength for silver from the Johnson
and Christy measurements (see citation 8 in the reading list at the end of the chapter).

Eq. 3.48. Figure 3.4 shows a plot the real and imaginary terms of the dielectric constant
of silver metal as a function of wavelength over the optical range; and Fig. 3.5 plots
the ratio of |00m /0m | << 1 for the same silver data. Separating Eq. 3.48 in real and
imaginary terms while dropping fractional terms quadratic in 00m , we find

0m d 00m d
 
2
ns = 0 · 1+i 0 0 (3.49)
m + d m (m + d )
and from the definition of the complex index of refraction

n2s = η 2 − κ2s + i2ηκs (3.50)

Setting the real and imaginary terms in Eqs. 3.49, 3.50 equal,
0m d
ηs2 − κ2s = (3.51)
0m + d
0m 00m d
2ηs κs = 2 (3.52)
0m (0m + d )
Equations 3.51, 3.52 can be decoupled and solved for ηs and κs , the two components
of the surface complex index of refraction. The result is
90 Surface Waves

Fig. 3.5 Imaginary part of the dielectric constant 00 vs. wavelength for silver from the
Johnson and Christy measurements (reference 8 in the reading list).
s
0m d
ηs = 0
(3.53)
m + d
3/2
1 00m 0m d

κs = (3.54)
2 (0m )2 0m + d

These expressions can be simplified further if the dielectric is vacuum or air. In that
case d can be set equal to unity and
s
0m
ηs = (3.55)
0m + 1
1 00m
κs = p · (3.56)
2 0m (0m + 1)3/2

As a typical example of a surface wave at a metal-dielectric interface, we take λ0 =

632 nm, d = 1 (air), m = −15.625 + i(1.04059) (silver) and calculate ηs , κs . From
Eqs. 3.55, 3.56 and the definition of k0 we find
 Plasmon Surface Waves at Optical Frequencies 91

Fig. 3.6 Ratio 00 /0 vs. wavelength for silver from the Johnson and Christy measurements
(reference 8 in the reading list).

Table 3.1 Field Components for TM Polarization

Region Ex (x, z, t) Ez (x, z, t) Hy (x, z, t)

kzd kd d d
dielectric ω0 d H( x, z, t) − ω0xd Hy (x, z, t) Hy0 ei(kx x+kz z−ωt)

kzm km m m
metal ω0 m Hy (x, z, t) − ω0xm Hy (x, z, t) Hy0 ei(kx x+kz z−ωt)

iκd ks d s
d side − ω0sd Hy (x, z, t) − ω0xd Hy (x, z, t) H0y eκs z ei(kx x−ωt)
interface
iκm ks m s
m side ω0 m Hy (x, z, t)
s
− ω0xm Hy (x, z, t) H0y e−κs z ei(kx x−ωt)

ηs = 1.0336
κs = 2.3534 × 10−3
k0 = 2π/λ0 = 9.942 × 106 m−1

Taking the “effective length” xeff to be the distance over which the amplitude of the
surface waves falls to 1/e of the initial value, we find xeff = 42.74 µm. Figure 3.7
shows the behavior of the surface wave Ex -field amplitude at the interface (z = 0),
propagating in the x direction. Figure 3.8 shows the variation of Ex , Ez near the
metal-dielectric interface.
92 Surface Waves

Fig. 3.7 The surface wave Ex -field amplitude at the interface (z = 0), propagating in the
x direction. The distance along which the amplitude decreases to 1/e of the initial value at
x = 0 is called the “effective” length.

3.4 Plasmon Surface Wave Dispersion

3.4.1 The Free-electron Plasma
The dielectric constant of materials in general depends on the frequency of the elec-
tromagnetic fields with which they interact. The equation describing this dependence
is called the dispersion relation, and the free-electron plasma model leads to a very
simple dispersion relation for high-conductivity metals. The force on a free electron
subject to an applied E-field in one dimension is

d2 x
me = −eEx (3.57)
dt2
and if the E-field is time-harmonic, E = E0 e−iωt , we have

−ω 2 me x = −eEx (3.58)

The polarization field P of the free-electron plasma, considered as a volume density

of one-dimensional dipoles px = −ex, is

e2
P = −Ne E (3.59)
me ω 2
with Ne the electron concentration of the plasma. From the definition of the displace-
ment field D and the constitutive relation between D and E we have
 Plasmon Surface Wave Dispersion 93

Fig. 3.8 Instantaneous distribution of the electric field Ex and Ez in the x and z directions,
respectively. The signal variation of the field Ez along x is accompanied by the variation of
electric charge (+,-) corresponding to the condition of Ez field discontinuity. The represen-
tation of surface charge density along z is exaggerated for clarity.

D = 0 E + P (3.60)
D = 0 E (3.61)

and therefore the dielectric constant of the free-electron plasma is

P Ne e2
(ω) = 1 + =1− (3.62)
0 E 0 me ω 2
The characteristic plasma frequency is defined by

Ne e 2
ωp2 = (3.63)
0 m e
So finally we have the lossless dielectric constant

ωp2
(ω) = 1 − (3.64)
ω2
3.4.2 The Drude Model of Metals
The free-electron plasma model describes harmonic electron motion in the limit of
a negligible restoring force acting on the electron and no “collisional” or dissipative
losses. The Drude model does not introduce a finite restoring force, but does add
dissipation to the equation of motion for the electrons in the plasma gas (when a
94 Surface Waves

linear restoring force is introduced into the conduction electron equation of motion,
the theory is called the Drude-Lorentz model of metals). The equation of motion
for the simple Drude model of the conduction electrons can still be described by a
one-dimensional driven harmonic oscillator expression, where the driver issues from
an electromagnetic field incident at the dielectric-metal boundary. If we assume the
incident radiation is polarized along x, then the Drude model equation of motion can
be written
d2 x dx
me 2 + me Γ = −eE(x, t) = −eE0m e−iωt (3.65)
dt dt
where me is the electron mass, Γ is a phenomenological damping constant, and eE0m e−iωt
the driving force of an E-M field in the metal with amplitude E0m and frequency ω.
Again dropping the common oscillatory phase term, the solutions for the position,
velocity and and acceleration are
1
x= eE0m (3.66)
me (ω 2 + iΓω)
dx iω
=− eE0m (3.67)
dt me (ω 2 + iΓω)
2
d x ω2
=− eE0m (3.68)
dt2 me (ω 2 + iΓω)

Again we write the polarization as a density of dipoles, Px = −Ne ex, and now the
dielectric constant is complex

Ne e 2 ωp2
(ω) = 1 − = 1 − (3.69)
0 me (ω 2 + iΓω) ω 2 + iΓω

Separating the real and imaginary terms

ω 2 − ωp2 + Γ2 ωp2 Γ
(ω) = + i (3.70)
ω 2 + Γ2 ω(ω 2 + Γ2 )

If Γ << ω as is the usual case at optical frequencies,

(ω) = 0 (ω) + i00 (ω)

ωp2 ωp2 Γ
(ω) ' 1 − + i (3.71)
ω2 ω3
Drude Model Dispersion Curves. The dispersive response to electromagnetic excita-
tion is characterized by the frequency dependence of the dielectric constant of a mate-
rial, (ω). Many common optical dielectrics (air, glass, fuzed quartz) exhibit dielectric
constants with near-zero imaginary terms (negligible absorption) and real terms that
vary little with frequency over the the visible range. We have just seen how metallic
response can be described by simple physical models such as the free-plasma model
or the Drude model. The behavior of a material is usually summarized by the “dis-
persion curve” that plots the energy or frequency of the exciting radiation against the
 Plasmon Surface Wave Dispersion 95

propagation constant k of the light in the material. There is a simple chain of relations
between the propagation constant k(ω) and the dielectric constant (ω). First

2π λ0
k= and λ= (3.72)
λ n
where λ0 is the free-space wavelength and n the index of refraction. Then the material
index of refraction can be complex, n = η + iκ and is related to the dielectric constant
 by
p
n(ω) = (ω) (3.73)
p
0 00
η(ω) + iκ(ω) =  (ω) + i (ω) (3.74)

and therefore

η 2 − κ2 = 0 (ω) (3.75)
00
2ηκ =  (ω) (3.76)

Decoupling η and κ results in a quadratic equation in η

1 00 2
η 4 − η 2 0 (ω) − [ (ω)] = 0 (3.77)
4
The solution to Eq. 3.77 can be obtained directly from the elementary quadratic for-
mula. The real part of the index of diffraction η can then be substituted into Eq. 3.76
to obtain the imaginary part κ. Thus η, κ and 0 , 00 can be readily interconverted. Fig-
ure 3.9 shows the dielectric constant terms and the refractive index terms, calculated
from Eq. 3.71 as a function of ω/ωp where ωp is the plasma frequency for silver metal,
ωp = 1.35 × 1016 s−1 and the relaxation rate Γ is taken to be Γ = 1 × 1014 s−1 .
The dispersion curve for a Drude metal can be obtained by starting with the
expressions for the real and imaginary parts of the dielectric constant, Eq. 3.71 and
rescaling the frequency variable to y = ω/ωp .

1
0m = 1 − 2 (3.78)
y
 
1 Γ
00m = 3 (3.79)
y ωp

and taking the propagation constant of the surface wave along the dielectric-metal
interface as ks = k0 ns . The factor ns is the surface of index of refraction, the real and
imaginary parts of which we write from Eqs. 3.53, 3.54.
" 1/2 3/2 #
d 0m 00m d 00m

ks = k0 ns = k0 +i 0 2 (3.80)
d + 0m 2(m ) d + 0m
ks = ks0 + iks00 (3.81)

Thus
96 Surface Waves

Fig. 3.9 Real (panel a) and imaginary (panel b) terms of the index of refraction and the
dielectric constant calculated from the Drude model (Eq. 3.71) vs. frequency ω normalized to
ωp , the plasma frequency for silver. The relaxation rate Γ is taken to be Γ = 1014 s−1 and
ωp = 1.35 × 1016 s−1 .

Fig. 3.10 Dispersion curve for a silver-glass interface. The real dielectric constant for glass
is taken to be 0 = 2.25. The loss term in the metal is not included.
 Plasmon Surface Wave Dispersion 97

1/2
d 0m

ks0 = k0 (3.82)
d + 0m
3/2
00m d 00m

00
ks = (3.83)
2(0m )2 d + 0m

At first neglecting the imaginary term, we define a new “reduced” variable

ks0 c k0
x= = s (3.84)
ωp kp

and write Eq. 3.82

1/2
d (y 2 − 1)

x=y (3.85)
y 2 (d + 1) − 1
Now Eq. 3.85 results in a quadratic equation in y the solution of which is
v q
u
u  + ( + 1)x2 ± ( + ( + 1)x2 )2 − 4 x2
t d d d d d
y= (3.86)
2d

The solution constitutes the dispersion curve (in “reduced” units) of the frequency vs.
the propagation parameter for the surface wave. The reduced variable x is defined in
terms of the real part of the propagation parameter ks0 and therefor this expression de-
scribes only the real part of the dispersion curve. The solution consists of two branches,
plotted in Fig. 3.10. The lower branch describes the surface-wave dispersion from the
long wavelength limit (y → 0; x → 0) to an asymptote (y → ysp = ωsp /ωp ; x → ∞).
This frequency asymptote at the short wavelength limit corresponds to the surface
plasma resonance where the surface charge density oscillates collectively over the en-
tire surface and the phase velocity of the surface wave dω/dks slows to near zero. An
expression for this surface plasma asymptotic frequency can be obtained from Eq. 3.86.
ωp
ω → ωsp = √ (3.87)
1 + d

The surface plasma resonance is often called the surface plasmon resonance and bares
a simple relation to ωp , the bulk plasmon resonance derived from the free-electron
plasma model. Between these two frequencies there is no charge density surface wave
or bulk wave propagation. This “forbidden zone” is sometimes termed a stopband. At
frequencies above ωp propagation once again occurs on the surface and in the bulk.
This frequency limit corresponds to the point where 0m → 0. Below ωp , 0m < 0 and
above ωp , 0m > 0. Also shown in Fig. 3.10 is the linear “light line”, the dispersion
curve of light freely propagating in the dielectric medium, ωll = k0 c/η.
Inclusion of the imaginary part of the surface parameter ks in the definition of the
reduced variable x in Eq. 3.84 and subsequent solution of Eqs. 3.85, 3.86 modifies the
dispersion curve and opens propagation in the stopband. The real and imaginary parts
of the full dispersion curve are shown in Fig. 3.11. Note that in panel (a) of this figure
98 Surface Waves

Fig. 3.11 Panel (a): dispersion curve for the real part of the surface propagation parameter
ks0 . Panel (b): Dispersion curve for the imaginary part of the surface propagation parame-
ter ks00 . Both propagation parameters have been normalized to kp = ωp /c. Inclusion of the
imaginary term results in finite propagation in the stopband (compare with Fig. 3.10).

the phase velocity

ω
vphase = (3.88)
ks0
reverses sign in the stopband region. This property is a necessary characteristic of
“negative index” metamaterials. Panel (b) shows that, unfortunately for many poten-
tial applications, this region is also one of high absorption.

Conduction Current in a Drude Metal. The conduction current in the metal is given
by
dx ω
Jc = eNe = eNe eE0m (3.89)
dt me (Γω − iω 2 )
and with the electron density related to the bulk plasma frequency, Eq. 3.63, we have
ωp2
Jc = 0 E0m (3.90)
(Γ − iω)
From the standard constitutive relations
σ
Jc = σE = 0 E (3.91)
0
and so, comparing Eqs. 3.90, 3.91, the expression for the complex conductivity can be
expressed as
σ Γ ω
= 2  + i  (3.92)
0 Γ
2 +
ω 2
2
Γ2
2 +
ω2
2
ωp ωp ωp ωp

We can compare the damping rate Γ to the plasma frequency ωp if we can estimate
the latter, and from Eq. 3.63 we see that this requires knowledge of the conduction
 Energy Flux and Density at the Boundary 99

Fig. 3.12 Indication of the components of the Poynting vector of the dielectric-metal inter-
face in which a wave propagates SPP.

electron density. A typical “good” metal such as silver exhibits an electron density
Ne ' 6 × 1028 m−3 . and therefore ωp ' 1.5 × 1016 s−1 . The damping due to radiative
or collisional loss is typically ∼ 1014 s−1 , and therefore Γ << ωp . Therefore in the
optical regime we can write
σ Γωp2 ωp2
' 2 +i (3.93)
0 ω ω
This last expression for the conductivity shows that at relatively low frequencies (RF,
microwave, radio wave) the real term dominates, and the conductivity is in phase
with the driving field and “ohmic”. As the frequency increases into the optical range,
however, the electrons can no longer follow the driving field in phase, and the imaginary
term dominates. At ω ' 1015 s−1 , typical of the visible optical regime, the real term is
only about 5 per cent of the imaginary term. Figure 3.10 plots the dispersion relation
of the Drude model, Eq. 3.71

3.5 Energy Flux and Density at the Boundary

3.5.1 Energy Flux Normal to the Boundary
As indicated in Eq. 2.60, the Poynting vector is defined as

S=E×H (3.94)

and the energy flux of an electromagnetic wave averaged over an optical cycle is given
by
1
S = Re[E × H∗ ] (3.95)
2
Figure 3.12 shows the x and z components of the energy flow on the two sides of the
surface between a dielectric and a metal. Writing out these components explicitly from
100 Surface Waves

Table 3.1, we have at normal incidence the transmitted energy flux impinging at the
surface from the dielectric into the metal along the z direction.
 d 
1   1 kz
Szd = Re Exd Hy∗ = Re 2
H0y (3.96)
2 2 d 0 ω
 m 
m 1  m ∗ 1 kz 2
Sz = Re Ex Hy = Re H0y (3.97)
2 2 m 0 ω
From the continuity relation at the surface, Eq. 3.33, we see immediately that the en-
ergy flux is continuous across the boundary in the z direction. On the dielectric side
the permittivity 0 d and the propagation constant kzd are real (for a lossless mate-
rial), and therefore the expression for the energy flux is also real. On the metal side
the dielectric constant is complex, m = 0m + i00m , as is kzm = k0 nm = k0 (ηm + iκm ).
The metal index of refraction nm is related to the metal dielectric constant through
√
nm = m , and therefore the expression for the product of Exm and Hy∗ in Eq. 3.97
exhibits real and imaginary terms. The real part of Eq. 3.97 represents the transmit-
ted, exponentially decreasing energy flux into the metal. In order to obtain explicit
z
expressions for nm and km we need to examine the propagation of electromagnetic
waves inside real metals.

3.5.2 Electromagnetics in Metals

We start by writing the curl equations of Maxwell in terms of E and B, the electric
field and magnetic induction field.
∂B
∇×E=− Faraday Law (3.98)
∂t
∂E
∇ × B = µm m + µm Jc Ampère-Maxwell Law (3.99)
∂t
where µm and m are the permeability and permittivity of the metal, and Jc the
free charge current. For a high-conductivity metal such as silver or gold Jc is directly
proportional to E with the conductivity σ the proportionality constant, Jc = σE.
Equations 3.98, 3.99 can be decoupled by first applying the curl operation to the Fara-
day equation,
 
∂B ∂(∇ × B)
∇ × (∇ × E) = ∇(∇ · E) − ∇2 E = ∇ × =− (3.100)
∂t ∂t
∂2E ∂E
= −µm m 2 − µm σ (3.101)
∂t ∂t
which together with ∇ · E = 0 results in a wave equation for E in the metal
∂2E ∂E
∇2 E = µm m + µm σ (3.102)
∂t2 ∂t
Similarly applying the curl operation to Eq. 3.99 produces the wave equation for B
∂2B ∂B
∇2 B = µm m + µm σ (3.103)
∂t2 ∂t
 Energy Flux and Density at the Boundary 101

Plane-wave solutions to these equations for waves propagating in the z direction can
be written
m
Em = E0m ei(kz z−ωt)
(3.104)
i(kzm z−ωt)
Bm = B0m e (3.105)

The amplitudes E0m , B0m and the propagation vector kzm are generally complex quan-
tities. Substituting these solutions back into the uncoupled wave equations 3.102, 3.103,
results in an expression for the propagation parameter in terms of the permeability,
complex permittivity, and complex conductivity of the metal,
2
(kzm ) = µm m ω 2 + iµm σω (3.106)

with
kzm = kr + iki m = r + ii σ = σr + iσi (3.107)
Equating real and imaginary terms after substitution into Eq. 3.106 yields

kr2 − ki2 = µ0 r ω 2 − µ0 σi ω (3.108)

2
2kr ki = µ0 i ω + µ0 σr ω (3.109)

where we have set µm = µ0 . Now writing the complex permittivity in terms of the
dielectric constants, r = 0 1 and i = 0 2 we have
   
σi σi
kr2 − ki2 = µ0 0 ω 2 1 − = k02 1 − (3.110)
0 ω 0 ω
   
σr σr
2kr ki = µ0 0 ω 2 2 + = k02 2 + (3.111)
0 ω 0 ω

where µ0 0 = 1/c2 and k0 = ω/c have been used. Setting

 
2 σi
k 0 1 − = β2 (3.112)
0 ω
 
2 σr
k0 2 + = γ2 (3.113)
0 ω

we separate Eqs 3.110, 3.111

v s
u  4
β u γ
kr = ± √ 1± 1+ (3.114)
t
2 β
v s
u  4
β u γ
ki = ± √ −1 ± 1 + (3.115)
t
2 β

But the complex propagation parameter km can also be expressed in terms of the
complex index of refraction, nm .
102 Surface Waves

km = k0 nm = k0 (ηm + iκm ) (3.116)

and nm is related to the metal dielectric constant m through
√ p
nm = m = 0m + i00m (3.117)
Again equating real and imaginary terms
v s
r u  00 4
0
m u 
ηm = 1± 1+ m (3.118)
t
2 0m
v s
r u  00 4
0
m u 
κm = −1 ± 1 + m (3.119)
t
2 0m
and consequently
v s
r u  00 4
0m u 
kr = k0 ηm = k0 1± 1+ m (3.120)
t
2 0m
v s
r u  00 4
0
m tu 
ki = k0 κm = k0 −1 ± 1 + m (3.121)
2 0m
Comparing Eqs. 3.120, 3.121 with Eqs. 3.114, 3.115 shows that
 
0 σi
m = 1 − (3.122)
0 ω
 
00 σr
m = 2 + (3.123)
0 ω
Equations 3.122, 3.123 separate the metal dielectric constant into two parts: first 1
and 2 represent the dispersive and absorptive response of the metal excluding the con-
duction electrons and second σi,r /0 ω, represent the conducting electrons contribution
to the dielectric constant. Substituting for σi,r /0 ω from Eq. 3.93, we have
! !
0
ωp2 00
Γωp2
m = 1 − m = 2 + (3.124)
0 ω 2 0 ω 3
If we consider the metal response, excluding the conduction electrons, to behave essen-
tially as a lossless dielectric material, we can posit 1 and 2 as 1 and 0, respectively.
This assumption is equivalent to the Drude model in the high frequency limit, Eq. 3.71.
3.5.3 Energy Flux along the Boundary
The energy flux of the surface wave propagating along the x direction.
1 1 kxs
Sxs,d = − Ezs,d (Hyd )∗ = H2 (3.125)
2 2 d 0 ω 0y
1 1 kxs
Sxs,m = − Ezs,m (Hym )∗ = H2 (3.126)
2 2 m 0 ω 0y
 Plasmon Surface Waves and Waveguides 103

Fig. 3.13 Design of a planar waveguide formed by three zones of different permittivity 1 , 3 ,
the “cladding” or shell can be considered metallic and 2 , the “core,” a dielectric. This type
of structure is sometimes termed a MIM waveguide (metal-insulator-metal). The complemen-
tary structure, IMI waveguides, are also possible. Propagating modes can be symmetric or
antisymmetric with respect to the core centerline.

3.6 Plasmon Surface Waves and Waveguides

3.6.1 Introduction
The plasmon surface wave at a metal-dielectric interface can be considered a guided
wave since it follows the surface. In this sense the interface itself acts as a one-
dimensional (1D) waveguide. We will take up a systematic discussion of transmission
lines and waveguides in Chapter 4. Here we show that plasmon-like electromagnetic
wave can propagate in the presence of two parallel interfaces separated by a subwave-
length gap and obtain the properties of this “gap” plasmon propagating mode. Figure
3.13 shows a slab waveguide with two exterior metallic cladding layers and one interior
dielectric layer. The lowest-order propagating mode of this guide resembles a surface-
wave, “surface plasmon polariton” (spp), bounded by the two outer metal claddings
instead of the usual single surface. Sometimes this propagating mode is called a “gap
plasmon”. Other modes, symmetric (cosine-like) or antisymmetric (sine-like) with re-
spect to the core centerline are possible as the width of the core increases. If the two
metal cladding layers are different (e.g. gold and silver), the guided wave will not be
perfectly symmetric or antisymmetric since the dispersion curves of different metals
vary somewhat. The allowed modes can be determined by matching the fields in the
transverse z direction at the cladding-core boundaries. The matching condition results
in a transcendental equation that can be solved numerically for the propagation pa-
104 Surface Waves

rameter q as a function of the core width a. Here we carry out in some detail the
matching operation for TM modes as an illustrative example.

3.6.2 Field Matching at Core-Cladding Boundaries

TM Polarization. For TM (transverse magnetic) polarization we have three field
components, Hy , Ex , Ez , in each of the three zones. From the coordinate setup in
Fig. 3.13 we see that Hy and Ex are parallel to the boundaries and therefor must be
continuous across them. We assume a harmonic force-field wave F propagating in the
x direction.
F (r, t) = F (y, z)ei(kx x−ωt) (3.127)
and use the two curl equations of Maxwell to find relations among the relevant field
components. We further assume that the materials are nonmagnetic with relative
permeability µ = 1.
∂Hy
= iω0 Ex (3.128)
∂z
∂Ex
− ikEz = iωµ0 Hy (3.129)
∂z
ikHy = −iω0 Ez (3.130)

Now we write down the real field components in the transverse z direction in each
of the three zones indicated in Fig. 3.13, while expressing the propagation in the x
direction as eikx .

For z < 0
Hy = Beikx epz (3.131)
 
−iB ikx pz −ip
Ex = pe e = Hy (3.132)
ω3 0 ω3 0
 
−Bk ikx pz −k
Ez = e e = Hy (3.133)
ω3 0 ω3 0
For 0<z<a
Hy = Aeikx cos(qz + ϕ) (3.134)
i
Ex = qAeikx sin(qz + ϕ) (3.135)
ω2 0
 
−k ikx −k
Ez = A cos(qz + ϕ) = Hy (3.136)
ω2 0 ω2 0
For z > a
Hy = C ikx e−rz (3.137)
 
iC ir
Ex = reikx e−rz = Hy (3.138)
ω1 0 ω1 0
 
−Ck ikx −rz −k
Ez = e e = Hy (3.139)
ω1 0 ω1 0
 Plasmon Surface Waves and Waveguides 105

Next we write the continuity conditions at each boundary for Hy and Ex .

For z = 0
B =A cos ϕ (3.140)
   
−ip iq
B= A sin ϕ (3.141)
ω3 0 ω2 0
For z = a
A cos(qa + ϕ) = Ce−ra (3.142)
iq ir −ra
A sin(qa + ϕ) = C e (3.143)
ω2 0 ω1 0
Rearranging and dividing Eq. 3.141 by 3.140 we have
2 p
tan ϕ = − (3.144)
3 q
and proceeding similarly with Eqs. 3.143 3.142 we have
tan qa + tan ϕ 2 r
tan (qa + ϕ) = = (3.145)
1 − tan qa tan ϕ 1 q
Substituting the right side of Eq. 3.144 into Eq. 3.145 results in the transcendental
equation the solutions to which define the propagating modes in the waveguide.
2 r 2 p
1 q + 3 q
tan qa = 22 pr
(3.146)
1− 1 3 q 2

The propagation parameters, p, q, r are also related through

∂Ex
− ikEz = iωµ0 Hy (3.147)
∂z
Substituting the relevant field components into this expression shows that
k 2 − p2 = 3 k02 (3.148)
2 2
q +k = 2 k02 (3.149)
2 2
−r + k = 1 k02 (3.150)
where k0 = ω/c is the propagation parameter of the free-space propagating wave
with frequency ω. Rearranging these expressions, we can define an “effective” index
of refraction that characterizes the propagating mode in the core.
q q
p = k 2 − 3 k02 = k0 n2eff − n23 (3.151)
p q
q = 2 k0 − k 2 = k0 n22 − n2eff (3.152)
q q
r = k 2 − 1 k02 = k0 n2eff − n21 (3.153)
√ √ √
where n1 = 1 , n 2 = 2 , n 3 = 3 , and neff = k/k0 .
106 Surface Waves

TE Polarization. For TE (transverse electric) polarization the relevant field compo-

nents are Ey , Hx , Hz , and carrying through the matching procedure at z = 0, z = a
we find the transcendental equation,
r p
q + q
tan qa = pr (3.154)
1− q2

As qa → 0 we find a limiting expression in TM polarization, Eq. 3.146, for the channel

width a,
1 p + 3 r
a=− (3.155)
2 pr

which is physically allowed since 1 , 3 are negative. Therefore even as a → 0 a prop-

agating mode always exists. In the case of TE polarization the limiting expression
is
p+r
a=− (3.156)
pr

and since the channel width cannot be negative, this result indicates that TE modes
are not supported in the deep subwavelength limit.

3.7 Surface Waves at a Dielectric Interface

3.7.1 Introduction
In the previous sections of this chapter we have focused almost entirely on electromag-
netic surface waves at dielectric-metal interfaces, and we have developed the physics
of charge density waves as the source of these surface plasmon-polariton fields. We
have also observed that the interface between two media can also be considered a 1-D
waveguide. In the present section we will again use boundary matching, as we did
in Section 3.6.2, to find wave solutions confined to the line defined at the interface
between two dielectric slabs, one of which is also in contact with a ground plane. We
will also find the 1-D waveguide modes of these surface wave for both TM and TE
polarization.
The geometry is shown in Fig. 3.14. We will consider the top dielectric, unbounded
in the positive x direction, as air or vacuum; and the dielectric slab with thickness
t might be glass or silicon. This arrangement can be realized practically on ridge
waveguides and is similar to stripline or microstrip geometries popular in integrated
circuit design. We study TM and TE modes separately and assume surface wave
propagation in the +z direction with propagation parameter β.

3.7.2 TM Modes
TM polarization specifies the existence of Ex , Ez , Hy polarized amplitudes. We seek
solutions to the scaler wave equation in the region above the 0 /d interface and in the
region within the d dielectric slab. We match these solutions at the interface.
 Surface Waves at a Dielectric Interface 107

Fig. 3.14 Surface waves propagate along z at the interface between vacuum 0 and material
dielectric d . The ground plane truncates wave propagation in −x direction and sets field
amplitudes to zero at the d -ground plane interface.

Above the interface the free-space propagation parameter k0 in the x − z plane is

related to kx and β by k02 = kx2 + β 2 . and the 1-D wave equation in the space above
the interface is given by  2 
∂ 2 2
+ k0 − β Ez (x) = 0 (3.157)
∂x2
and within the dielectric slab
∂2
 
2 2
+ k d − β Ez (x) = 0 (3.158)
∂x2

where kd2 = d k02 and Ez = Ez (x)eiβz . We assume that there is no variation in the
fields along the y direction. Within the dielectric slab the general solution, in terms of
linear combinations of complex traveling waves, is

Ez (x) = Aeikc x ± Be−ikc x (3.159)

where
kc2 = kd2 − β 2 = d k02 − β 2 (3.160)
Above the interface the E-field of the surface wave must decrease exponentially,
and the form of the solution must be

Eza (x) = Ce−κx (3.161)

where
κ2 = β 2 − k02 (3.162)
and Eza (x)is the phasor amplitude in the air above the dielectric slab. We anticipate
that the propagation velocity of the surface wave will be somewhat slower than a wave
of the same frequency propagating in air since a significant faction of the surface wave
amplitude will be immersed in the dielectric slab. Therefore we expect β > k0 and
Eq. 3.162 is written so that κ is real and can be be chosen with positive sign. This
108 Surface Waves

choice assures that Eq. 3.161 represents an exponentially decaying amplitude in the
+x direction.
At x = 0 the boundary condition requires Ezd (0) = 0. Therefore B = A and

Ezd (x) = ±A eikc x − e−ikc x = ±A2i sin kc x

(3.163)

Where Ezd (x) is the amplitude for the surface-wave phasor Ez (x) = Ez (x)eiβz . At x = t
the amplitudes for the E-field amplitudes on both sides of the interface have to match.

±A2i sin kc t = Ce−κt (3.164)

The H-field amplitudes have to match as well, and we obtain the H-fields from the
E-fields by using the relations obtained in Chapter 4, Section 4.5.1. In the dielectric
slab we have
i ∂Ez
Hyd (x) = ω0 d
kc2 ∂x
i
= ± ω0 d A2i cos kc x (3.165)
kc
and on the air side
i
Hya (x) = − ω0 Ce−κx (3.166)
κ
At the interface H-field amplitudes must match. Since Hya (x) has a negative amplitude,
we must choose the negative amplitude solution for Hyd (x).

i i
− ω0 d A2i cos kc t = − ω0 Ce−κt (3.167)
kc κ
Now divide Eq. 3.164 by Eq. 3.167, the two matching equations at the interface. The
result is
kc tan kc t = d κ (3.168)
which provides a relation between the wave vector kc , the thickness of the dielectric
slab t, the damping constant κ, for a given slab material with dielectric constant d .
Equations 3.159, 3.160 specify that kc is the wave vector component along x of that
part of the surface wave within the dielectric slab. We can write κ in terms of kc and
k0 by using the fact that the propagation parameter β along z must be the same for
the two parts of the surface wave: one part in the dielectric slab and the other in the
air. We can therefore eliminate β from Eqs. 3.160, 3.162 resulting in the expression

kc2 + κ2 = k02 (d − 1) (3.169)

In principle we could eliminate κ between Eqs. 3.168, 3.169 and obtain an expression
for kc as a function of three parameters: the free-space wave vector k0 , the material
dielectric constant d and the slab thickness t. However, Eq. 3.168 is a transcendental
equation with no simple analytic solution for kc . The best we can do is find solutions
for kc numerically and plot them to gain some insight into the properties of the TM
modes.
 Surface Waves at a Dielectric Interface 109

Fig. 3.15 The expressions y1 , y2 plotted as a function of kc t. The function y2 is plotted for
a range of thicknesses t from 100 nm to 1000 nm. Dashed curves show four representative
solutions where the thickness t permits modes TM0 and TM1 . Note that TM0 has no cutoff
thickness but TM1 cuts off at kc t = π. Intersection of the curves y1 , y2 > 0 constitute solutions
to the transcendental equation 3.170.

One way to plot the solutions is to divide both sides of Eq. 3.168 by kc and rearrange
it as
d κ
tan(kc t) = (3.170)
kc
and consider kc as an independent variable y. We can then plot both sides of the
equation, and identify solutions to the transcendental equation as the points of inter-
section.

y1 = tan(kc t) (3.171)
 1/2
d κt d k02 (d − 1) − kc2 t
y2 = = (3.172)
(kc t) (kc t)
Figure 3.15 shows schematically the appearance of the two families of curves y1 , y2
when plotted against kc t. Clearly y1 is just a plot of the tangent function with branches
starting at y1 = 0 for kc t = 0, π, 2π . . . It is also clear that only positive values of y1
can provide legitimate solutions to Eq. 3.168 since the numerator of y2 consists of three
factors all of which are intrinsically positive. For fixed d and k0 , y2 , as a function of
110 Surface Waves

(kc t), is a family of curves parametrically increasing with the slab thickness t. From
Eq. 3.172 we see that y2√→ ∞ as kc t → 0 and that y2 crosses 0, becoming pure
imaginary, when kc = k0 d − 1. Intersections of y2 with y1 within the positive half
of the first branch (0 ≤ kc t ≤ π/2) constitute solutions to the transcendental equation
3.168 belonging to the lowest TM mode, TM0 . We see immediately that the TM0 mode
has no cutoff for kc as kc t → 0. For fixed fixed kc , t and as k0 increases, y2 will begin
to intersect y1 both in the first and in the second positive branch (π ≤ kc t ≤ 3π/2).
These intersections in the second branch constitute √ TM1 solutions to 3.168. Higher
order TM modes begin to propagate as the kc = k0 d − 1 upper limit increases with
k0 . From Fig. 3.15 it clear that the cutoff points for TM1 , TM2 , . . . occur at kc t = nπ
with n = 1, 2, 3, . . . The general expression for the cutoff frequency (in Hz units) is
given by
kc
νc = √ (3.173)
2π µ
√
where 1/ µ is the local, “effective” speed of light. For the TM modes of this surface
wave, therefore, we have
nc
νc = √ n = 0, 1, 2, . . . (3.174)
2t d − 1

where d − 1 represents an “effective” dielectric constant for the TM modes in the 1-D
surface wave.

3.7.3 TM Field Solutions

Once kc has been found for a fixed set of parameters k0 , t, d , the TM fields Ez , Ex , Hy
can be obtained from Eqs. 3.163, 3.164 and the relations in Section 4.5.1.

Ez (x, z) = ±A0 sin(kc x)eiβz 0≤x≤t (3.175)

0 −κ(x−t) iβz
Ez (x, z) = ±A sin(kc t)e e x≥t (3.176)

Where the constant A0 replaces the constant factors in 3.163, A0 = A2i.

For Ex we find

iβ 0
Ex (x, z) = ± A cos(kc x)eiβz 0≤x≤t (3.177)
kc
iβ
Ex (x, z) = ∓ A0 sin(kc t)e−κ(x−t) eiβz x≥t (3.178)
κ

and for Hy

iω0 d 0
Hy (x, z) = ± A cos(kc x)eiβz 0≤x≤t (3.179)
kc
iω0 0
Hy (x, z) = ∓ A sin(kc t)e−κ(x−t) eiβz x≥t (3.180)
κ
 Surface Waves at a Dielectric Interface 111

3.7.4 TE Modes
In contrast to the case of plasmon surface waves, the 1-D waveguide between two
dielectrics can support TE waves as well as TM waves. The fields specified by TE
polarization are Hz , Hx , Ey and the scaler wave equation above the dielectric slab is
written  2 
∂ x 2
− κ Hz = 0 (3.181)
∂x2
and, since the dielectric slab is nonmagnetic, within the slab the wave equation is
 2 
∂ x 2 2
+ k0 − β Hz = 0 (3.182)
∂x2
The form of the solutions within the slab is

Hzd (x) = Aeikc x ± B −ikc x (3.183)

with kc2 = k02 − β 2 . On the air side above the slab we again posit an exponentially
decreasing amplitude.
Hza (x) = Ce−κx (3.184)
Again at x = 0 the E-field parallel to the ground plane must vanish. Obtaining the
E-field from the H-field using the relations in Section 4.5.1, we have
i ∂Hz
Eyd (x) = − 2
ωµ0
kc ∂x
ωµ0
=∓ A2i sin kc x (3.185)
kc
and in the air space above the dielectric slab
i
Eya (x) = ωµ0 Ce−κx (3.186)
κ
The expression for Eyd in Eq. 3.185 implies that the general solution for Hzd (x) in
Eq. 3.183 specializes to
Hzd (x) = ±A2 cos kc x (3.187)
and that therefore the H-field is maximum at x = 0, the ground-plane boundary.
At first thought this condition might seem perplexing since no field should exist in
the ground plane. What the boundary condition really means is that an infinitesimal
distance above the boundary the H-field amplitude is ±2A and an infinitesimal distance
below the boundary the H-field is null.
At the boundary x = t, Hz and Ey must match.

±A2 cos kc t = Ce−κt (3.188)

ωµ0 i
∓ A2i sin kc t = ωµ0 Ce−κt (3.189)
kc κ
Dividing 3.188 by 3.189 and rearranging results in
112 Surface Waves

Fig. 3.16 The expressions y1 , y2 plotted as a function of kc t. The function y2 is plotted

for a range of thicknesses t from 100 nm to 1000 nm. The first four dashed curves indicate
representative values of y2 (kc t) that do not admit solutions. Cutoff for the first allowed TE
mode (TE1 ) occurs at kc t = π/2. Intersection of the curves in the y domain where y1 , y2 > 0
constitute solutions to the transcendental equation 3.190.

κ
− cot kc t = (3.190)
kc
Again we set
y1 = − cot kc t (3.191)
s
κt k02 (d − 1)
y2 = = −1 (3.192)
kc t kc2
and plot y1 , y2 with kc t the independent variable. Solutions to 3.190 are at the positive
intersections of y1 , y2 . The general form of the solutions is shown in Fig. 3.16. In the first
zone 0 ≤ kc t ≤ π/2 no TE modes are possible because all the y1 solutions are negative
and therefore unphysical. The first TE mode, TE1 appears in the zone π/2 ≤ kc t ≤ π.
From Figs. 3.15, 3.16 we see that the energy ordering of the first three modes is TM0 ,
TE1 , TM1 . Subsequent TE modes first appear at kc t = 3π/2, 5π/2, 7π/2 . . . The cutoff
frequencies are then
nc
νc = √ n = 1, 3, 5, . . . (3.193)
4t d − 1
 Exercises 113

3.7.5 TE Field Solutions

Once kc has been found for a fixed set of parameters k0 , t, d , the TE fields Hz , Hx , Ey
can be obtained from Eqs. 3.184 – 3.187 and the relations in Section 4.5.1.
Hz (x, z) = ±2A cos(kc x)eiβz 0≤x≤t (3.194)
Hz (x, z) = ±2A cos(kc t)e−κ(x−t) eiβz x≥t (3.195)
For Hx (x, z)
β
Hx (x, z) = ∓ i2A sin(kc x)eiβz 0≤x≤t (3.196)
kc
iβ
Hx (x, z) = ± 2A cos(kc t)e−κ(x−t) eiβz x≥t (3.197)
κ
and for Ey (x, z)
iωµ0
Ey (x, z) = ∓ 2A sin(kc x)eiβz 0≤x≤t (3.198)
kc
iωµ0
Ey (x, z) = ± 2A cos(kc t)e−κ(x−t) eiβz x≥t (3.199)
κ
As we can see from Figs, 3.15, 3.16, the energy ordering of the first three modes is
TM0 , TE1 , TM1 .

3.8 Exercises
1. It is a well-known fact that aluminum (Al), soon after exposure to air, exhibits a
transparent aluminum oxide layer (Al2 O3 ). This oxide layer is about 100 nm thick
and has an index of refraction nAl2 O3 = 1.77. What would be the effective index
of refraction of a surface plasmon polariton wave propagating at the metal/metal
oxide interface if the incident light has a wavelength of 400 nm? The index of
refraction of Al at = 400 nm is n = 0.381 + i4.883.
2. A surface plasmon polariton (spp) wave propagates at the interface between air
and a gold (Au) layer. Calculate the phase and group velocity if the incident
wavelength is 780 nm. In the common case of good conductors where |0m |  00 ,
the group velocity is, to good approximation,
c
vg =
nspp
(s  0 3/2   0  )−1
d 0m

ω d m 1 dm 1 dd
'c + · · 02 + 2 (3.200)
d + 0m 2 d 0m  m dω d dω
and from the Drude model the complex dielectric constant for Au can be expressed
as
 ω 2
p
0m (ω) = 1 − (3.201)
ω
!
ωp2
00m = (3.202)
ω3 τ
114 Surface Waves

The bulk plasmon resonance frequency is denoted by ωp , and τ is the characteristic

damping time in the metal. For Au the values are ωp = 1.37 × 1016 s−1 and
τ = 2.47 × 10−14 s.
3. Calculate the spatial dependence of the electric and magnetic field amplitudes
of an spp wave propagating at the interface between a silver (Ag) layer and a
glass substrate. The incident wave is 780 nm. Use the Drude model of metals to
calculate the permittivity of Ag at the given wavelength. The dielectric constant
of glass is taken to be  = 2.25.
4. With reference to the preceding problem, calculate the spp energy flux along the
propagation direction (x) at two points: x = 0 and x = 100 nm.
5. An spp wave propagates along the x direction at an air-gold interface. The free-
space wavelength of the incident wave exciting the spp is 632 nm, and the origin of
the spp wave is at x = 0. Calculate the spatial extent of the spp amplitude along
the z direction, normal to the interface, at x = 0. Determine the 1/e points for
the amplitude and intensity on the air side and on the metal side of the interface.
6. Suppose a plane wave (free-space wavelength λ = λ0 ) is normally incident on a
dielectric-metal interface. Consider the energy density at the interface between
the two media: a dielectric with permittivity 0 d and a metal with permittivity
0 m . Write the relation between the electric and magnetic energy densities on the
dielectric side and the metal side of the interface. Assume the dielectric is lossless
but m = 0m + i00m .
7. In Section 3.7 we developed the properties of surface waves propagating at the
interface between two dielectrics, one of which was grounded. Follow the same
procedure of field matching at the boundaries to find the surface waves at an
air-dielectric interface where the dielectric is in contact with a perfect electrical
conductor (PEC). Find the transcendental equations for TM polarization, analo-
gous to Eq. 3.168.

3.9 Further Reading

1. A. Sommerfeld, Partial Differential Equation in Physics, Lectures on Theoretical
Physics Vol. VI, Chap. VI, Problems of Radio;(Academic Press, New York, 1964).
2. A. Sommerfeld, Ann. Phys. Chem. 67, 233 (1899).
3. J. Zenneck, Ann. Phys. 23, 846 (1907).
4. H. Raether, Surface plasmons on Smooth and Rough Surfaces and on Gratings
(Springer, Berlin, 1988).
5. M. Born and E. Wolf, Principles of Optics, 6th edition, Pergamon Press, 1993.
6. T-I. Jeon and D. Grischkowsky, THz Zenneck surface wave (THz surface plasmon)
propagation on a metal sheet, Applied Physics Letters, vol. 88, p. 061113 (2006).
7. D. M. Pozar, Microwave Engineering, 3rd edition, John Wiley & Sons, 2005.
8. P. B. Johnson and R. W. Christy, Optical Constants of the Noble Metals, Phys.
Rev. B 6 4370-4379 (1972).
4
Transmission Lines, Waveguides, and
Equivalent Circuits

4.1 Introduction
In chapter 2 we summarized the elements of electromagnetics in terms of electric and
magnetic vector force fields related by Maxwell’s equations. We saw in section 2.4 that
the two time-varying equations, Faraday’s Law, Eq. 2.27 and the Maxwell-Ampère
Law, Eq. 2.28, give rise to electromagnetic wave solutions that propagate through a
medium or, as we saw in chapter 3, at the interface between media. These solutions
are generally distributed through out a spatial extent large with respect to the char-
acteristic wavelength.
In contrast the conventional radio and microwave circuit theory of electrical en-
gineering considers time-varying electromagnetic phenomena from the standpoint of
various combinations of lumped elements, (capacitors, inductors, resistors) localized in
space and linked by interconnects of negligible impedance. Furthermore, the spatial
extent of these circuit elements, together with their sources of voltage and current,
oscillating from tens of kilohertz to hundreds of megahertz, is usually much smaller
than the characteristic wavelength. The result is that all the elements of the circuits
are subject to the same time dependence without retardation. This time behavior
is termed quasistatic, and the physical arrangement of the circuit can be said to be
subwavelength. The subwavelength scaling similarity between dielectric and metallic
nanostructures interacting with optical driving fields and electrical circuits driven by
conventional voltage and current oscillators suggests that circuit analysis might find
useful application in the design of functional plasmonic and photonic devices. The
main goal of this chapter therefore is to explore this possibility.

4.2 Elements of Conventional Circuit Theory

Circuit theory is essentially the application of Maxwell’s equations to problems com-
monly encountered in electrical engineering. The conditions that validate circuit the-
ory are: (1) that electrical effects happens instantaneously throughout the circuit—the
quasistatic approximation, (2) that the net charge on every component is null, and
(3) that magnetic coupling between or among lumped components of the circuit are
negligible. A typical circuit comprising a source, resistor, capacitor and inductor are
shown in Fig. 4.1.
116 Waveguides and Equivalent Circuits

Fig. 4.1 Schematic of an electrical circuit showing lumped elements of resistor, inductor,
capacitor in series with a voltage source. Also shown is the direction of current relative to
the polarity of the voltage source.

4.2.1 Kirchhoff ’s Rules

The starting point in circuit analysis are the two rules annunciated by Gustav Kirchhoff
in 1845. The rules are: (1) The scalar sum of the electric potential differences Vi around
any closed circuit loop is null. X
Vi = 0 (4.1)
i

(2) The scalar sum of charge currents Ii flowing out of a junction node is null.
X
Ii = 0 (4.2)
i

Kirchhoff’s Voltage Rule. We consider first the voltage rule which is the more relevant
of the two for our purposes. Equation 4.1 derives from the differential form of Faraday’s
Law. By applying Stokes’ Theorem to Eq.2.27 we have
Z I Z
∂B
(∇ × E) · dS = E · dl = − · dS (4.3)
S S ∂t

where E, B are the usual electric and magnetic induction fields, S is the outward
surface normal through which the curl of E protrudes, and l is the line around the
boundary of S, the positive direction of which is taken by using the right hand rule.
With respect to Fig. 4.1 we take the line integral around the circuit and define the
voltage difference between two reference points a, b as
Z b
Vba = − E · dl (4.4)
a

Then using Eq. 4.3 we have

 Elements of Conventional Circuit Theory 117

Z b Z d Z f Z h Z Z
∂B ∂
− E · dl − E · dl − E·− E · dl = · dS = B · dS (4.5)
a c e g S ∂t ∂t S

or Z
∂
V0 (t) + Vdc + Vf e + Vhg = B · dS (4.6)
∂t S

where V0 (t) is the voltage source connected to terminals a, b, Vdc is the resistive voltage
drop, Vf e and Vhg the inductive and capacitive voltage drops, respectively. In order
to identify Eq. 4.6 with the Kirchhoff rule we have to make two assumptions: (1) that
there is no changing magnetic flux traversing the plane of the circuit loop. In simple
circumstances where we only consider R, L, C lumped circuits, this condition is not
hard to realize. Even if there is a changing magnetic field present, it can usually be
incorporated into the inductive voltage term to make an “effective” voltage drop. (2)
Voltage drops along the wire connecting the lumped elements in the loop are negligible.

Resistance. We know from Ohm’s law the current density J through a conductor is
proportional to the E-field along the conductor, J = σE, where σ is the conductivity.
Then the voltage drop across the resistor is
Z d Z d
J
Vdc = − E · dl = − · dl (4.7)
c c σ
R
We also know that J·dl can be written in terms of the total current I flowing through
the conductor and the resistivity ρ, the inverse of the conductivity
Z d
I
Vdc = − ρ dl = −IR (4.8)
c A

where the resistance R is defined as

Z d
ρ
R= dl (4.9)
c A

and A is the cross sectional area of the conductor. The familiar expression Eq. 4.8 is
valid at low frequencies where the current flows through the conductor uniformly over
A. At high frequencies the current in not uniform across A, passing only near the
surface and within the skin depth.

Induction. Here we apply Faraday’s law to the local inductive element labeled L in
Fig. 4.1 (not the whole circuit loop). The integral has to be over a closed path linking
the terminals e, f . The first branch of the loop goes through the current-carrying coil
from e to f . The second branch is the return path from f to e along an arbitrary path
in the space outside the coil. This second branch contributes nothing to the voltage
drop across the coil.
I Z f Z e 
E · dl = E · dl + E · dl = 0 (4.10)
e f
118 Waveguides and Equivalent Circuits

Then Z f Z
∂
Vf e = E · dl = − B · dS (4.11)
e ∂t S
With the definition of the inductance L as the integral of the magnetic induction B
over the surface surrounding the inductive element, per unit current I passing through
the coil of the inductor, we have
R
B · dS
L= (4.12)
I
Therefore
∂(LI) ∂I
Vf e = − = −L (4.13)
∂t ∂t
Capacitance. We assume that the capacitive element can be represented by a par-
allel plate capacitor. From elementary electrostatics we know that the definition of
capacitance is the total charge Q accumulated on one plate divided by the voltage Vhg
across the plate.
Q
C= (4.14)
V
But Z
Q = I dt (4.15)

and thus the voltage drop across the capacitor in terms of the capacitance C and the
current looping through the circuit is
R
I dt
Vhg = −Vgh = − (4.16)
C
Putting all three lumped circuit elements together with the positive voltage source
Vba = Vs and using the Kirchhoff voltage rule, we have
Z
dI 1
Vs − IR − L − I dt = 0 (4.17)
dt C
Kirchhoff’s Current Rule. Kirchhoff’s current rule, that the charge currents Ii flowing
into and out of a circuit junction sum to zero, is illustrated in Fig. 4.2. The basis for
Eq. 4.2 is essentially the charge conservation relation that derives from the Ampère-
Maxwell equation
∂D
∇×H=J+ (4.18)
∂t
Taking the divergence of both sides of this equation and remembering that div·curl = 0
we find
∂∇ · D
∇·J+ =0
 ∂t 
∂D
∇· J+ =0 (4.19)
∂t
 Elements of Conventional Circuit Theory 119

Fig. 4.2 Circuit junction point from which currents enter and leave. The Kirchhoff current
rule states that the sum of the currents into and out of the junction must be zero.

Then using Stokes’ theorem we can write

Z   I  
∂D ∂D
∇· J+ dA = J+ · dl = 0 (4.20)
A ∂t c ∂t
Equation 4.20 says that the sum of the conduction current density J and displacement
current density ∂D/∂t around a closed loop c defining a surface A is null. From a
circuit point of view, the closed loop can be considered any loop around a junction
point into and out of which currents flow along highly spatially localized wires. The
integral over the conduction current can be considered a sum over the current Ii of
wire i by writing I X
J · δ(li − l) dl = Ii (4.21)
c i
For harmonically time varying circuits the second term on the right in Eq. 4.20 can be
written I I
∂D
· dl = −iωD · dl ' 0 (4.22)
c ∂t c
since at radio and microwave frequencies the displacement current is negligible com-
pared to the conduction current. At optical frequencies the displacement current might
120 Waveguides and Equivalent Circuits

Fig. 4.3 Lumped circuit element representation of a transmission line repeating segment
with length ∆z. The transmission line extends along the z direction

become significant and in this regime Kirchhoff’s current rule has to be checked on a
case-by-case basis. We shall assume as a working assumption that Kirchhoff’s current
rule is valid and write I  
∂D X
J+ · dl ' Ii ' 0 (4.23)
c ∂t i

4.3 Transmission Lines

4.3.1 Lumped-Element Circuit Analysis of a Transmission Line
In conventional circuit analysis, lumped circuit elements such as resistors, inductors,
and capacitors are assumed to be dimensionally “subwavelength,” i.e. much smaller
than the characteristic wavelength of the driving field. In contrast a transmission line
is a circuit entity transporting voltage and current waves over distances much greater
than a wavelength. However, Kirchhoff’s two rules can be used to analyze a trans-
mission line as a succession of lumped circuit elements. A two-wire transmission line
extending along the z direction can be represented by the diagram in Fig. 4.3 that
shows one segment of a repeating circuit laid out along ∆z. The lumped elements
R, L, C represent the resistance, inductance and capacitance per unit length, respec-
tively. The fourth quantity, G, represents the “shunt conductance” per unit length due
to dielectric absorption between the two conductors. The series resistance in the two
conductors is due to the finite (but very high) conductivity of the metal while shunt
conductance is due to the finite (but very low) conductivity of the insulating dielectric.
These four quantities divide into two groups: R and G are dissipative and represent
loss, while L and C are reactive and represent stored energy.
We can use the Kirchhoff rules to analyze this circuit segment. Proceeding around
the circuit and applying the voltage rule (Eq. 4.1) we have
∂I(z, t)
V (z, t) − R∆zI(z, t) − L∆z − V (z + ∆z) = 0 (4.24)
∂t
 Transmission Lines 121

while the current rule (Eq. 4.2) gives

∂V (z + ∆z)
I(z, t) − G∆zV (z + ∆z, t) − C∆z − I(z + ∆z, t) = 0 (4.25)
∂t
Now divide both equations by ∆z and take the limit as ∆z → 0. The result is
∂V ∂I(z, t)
= −RI(z, t) − L (4.26)
∂z ∂t
∂I ∂V (z, t)
= −GV (z, t) − C (4.27)
∂z ∂t
As usual we assume harmonic time variation so the partial derivative terms with
respect to t simplify to
dV
= − [R − iωL] I(z) (4.28)
dz
dI
= − [G − iωC] V (z) (4.29)
dz
These two equations can be uncoupled to yield
d2 V (z)
− β̃ 2 V (z) = 0 (4.30)
d z2
d2 I(z)
− β̃ 2 I(z) = 0 (4.31)
d z2
with
β̃ 2 = (R − iωL)(G − iωC) (4.32)
p
β̃ = ± (R − iωL)(G − iωC) (4.33)
and
β̃ ≡ iβ − γ (4.34)
The solutions are very reminiscent of the plane wave solutions of Chapter 2, Section
2.10. In the positive z space the voltage and current solutions look like damped travel-
ing waves propagating in the forward z direction. The voltage wave forward traveling
takes the form
V = V0+ eβ̃z = V0+ ei(β+iγ)z (4.35)
where β is the propagation parameter1 and γ the dissipation term. In the negative z
space we have backwards traveling waves.
V = V0− e−β̃z = V0− e−i(β+iγ)z (4.36)
The general solution is some linear combination of forward and backward traveling
waves
V = V0+ eβ̃z + V0− e−β̃z (4.37)
Usually the amplitude of the backward propagating wave V0− will not be equal to the
amplitude of the forward wave V0+ , and if the backward wave is reflected (the usual
1 Common usage in optics and electromagnetics is to denote k for the propagation parameter. In
transmission line theory, developed by and for electrical engineering, the term β is more common.
122 Waveguides and Equivalent Circuits

case) the forward and backward amplitudes will be related by a reflection coefficient.
From Eq. 4.28 we can write I(z) in terms of V0+ and V0− .

dV /dz iβ − γ  +
I(z) = I0+ + I0− = − V0 − V0−

=− (4.38)
R − iωL R − iωL
In analogy to Eq. 2.198 we define an impedance Z0 as the ratio of the forward voltage
amplitude to the forward current amplitude.

V0+
Z0 = (4.39)
I0+

Substituting from 4.38 and using 4.32 4.34 we have

r
R − iωL R − iωL
Z0 = − = (4.40)
iβ − γ G − iωC

where we have used the negative root of β̃ 2 from 4.33 in the denominator of 4.40.
The transmission line current can now be written in terms of the voltage and the
characteristic impedance.

1 h + β̃z i
I(z) = V0 e + V0− e−β̃z (4.41)
Z0

In most practical transmission lines R and G are very small, and in the limit of a
lossless transmission line we find r
L
Z0 → (4.42)
C
4.3.2 Lossless Plane-Parallel Transmission Line
If we start with a lossless transmission line as the point of departure, we can calculate
the inductance per unit length, a function of the transmission line geometry, directly
from Faraday’s law. Figure 4.4 shows the magnetic flux lines between two parallel
plates carrying a constant current I in opposite directions. The separation between
the plates is given by s and the width of the plates is w. The ratio of w/s is sufficiently
great that the magnetic flux density B is mostly concentrated between the plates and
can be considered constant. The magnetic field between the plates can be related
to the current flowing in one of the plate conductors through the integral form of
the Maxwell-Ampère law, Eq. 2.28. Using Stokes’ theorem as we did in Eq. 4.3 for
Faraday’s law, we write
Z I Z
(∇ × H) · dS = H · dl = J · dS (4.43)
S

Applying this relation to the cross sectional area ABCD of Fig. 4.4 we have

H0 w = I (4.44)
 Transmission Lines 123

Fig. 4.4 Idealized parallel-plate transmission line with no resistive losses in the current-car-
rying plates and no shunt conductance between the plates.
R
where I = J · dS, the integral of the current density over ABCD. Now R we substitute
the definition of inductance L from Eq. 4.12 into Eq. 4.44, taking the B · dS over the
surface shown in Fig. 4.4. The inductance per unit length along z is then
dL s
= L̄ = µ0 (4.45)
dz w
Voltage and Current along a Transmission Line. Equation 4.45 expresses the constant
distributed inductance along the planar wave guide of Fig. The voltage gradient along
the guide is then obtained from Eq. 4.13
∂V ∂I
= −L̄ (4.46)
∂z ∂t
and the current gradient from the distributed capacitance, obtained from Eq. 4.17.
∂I ∂V
= −C̄ (4.47)
∂z ∂t
4.3.3 Correspondence to Plane Waves
Transmission Line Voltage and Current Waves. Equations 4.46, 4.47 bear a close re-
semblance to Eqs. 2.188–2.191. Just as we found for Eqs. 4.30, 4.31 differential “Helmholtz-
like” equations for V and I separately can be obtained for the lossless transmission
124 Waveguides and Equivalent Circuits

line following the same procedure that produced Eqs. 2.193, 2.194 for plane waves.
Differentiation of 4.46 by z and 4.47 by t results in

∂2V ∂2V
2
− LC 2 = 0 (4.48)
∂z ∂t
and eliminating V in favor of an equation for I by reversing the differentiation variables
for each equation results in
∂2I ∂2I
− LC =0 (4.49)
∂z 2 ∂t2
Just as in the plane-wave case we associate LC with 1/v 2 where v is the phase velocity
of the V − I transmission wave running in the two-slab transmission line of Fig. 4.4.
Furthermore if we assume time-harmonic phasor solutions to Eqs. 4.48, 4.49 we can
write the solution to the voltage equation as

V = V0 ei(βz−ωt) (4.50)

Then substituting this solution into Eq. 4.48 yields

∂2V
+ LCω 2 V = 0 (4.51)
∂z 2
We see therefore that the phase velocity v and propagation parameter β for the V − I
excitation of the transmission line is
1 ω √
v=√ and β = = LCω (4.52)
LC v

Given the voltage solution, Eq. 4.50, we can find the current by substituting 4.50 into
4.46 and integrating with respect to time.
∂I 1 ∂V 1
=− = − iβV (4.53)
∂t L ∂t L
then Z r
1 1β C
I = − iβ V dt = V = ·V (4.54)
L Lω L
The ratio of the voltage to current amplitudes has units of resistance and is identified
with the characteristic impedance of the transmission line.
r
V L
Z0 = = (4.55)
I C
which is, as expected, in agreement with Eq. 4.42.

Reflection on a Lossless Terminated Line. So far we have considered the character-

istic impedance Z0 of the transmission line itself which is determined essentially by
the geometric parameters of the line — the separation and symmetry between the con-
ductors and the nature of the insulating dielectric between them. If now we insert a
 Transmission Lines 125

load impedance ZL across the line conductors, the “impedance mismatch” will provoke
some fraction of the incident wave to reflect at the load position. the transmission line
is said to be “terminated” by ZL . Suppose we put the load at z = 0. We know that in
general V and I on the otherwise lossless line are related by
V (z) = V0+ eiβz + V0− e−iβz (4.56)
1  + iβz
V0 e − V0− e−iβz

I(z) = (4.57)
Z0
At z = 0 we must have
V (0) V + + V0−
ZL = = 0+ Z0 (4.58)
I(0) V0 − V0−
so the amplitude of the reflected wave in terms of the line and load impedances is
 
ZL − Z0
V0− = V0+ (4.59)
ZL + Z0
Now, in analogy to Eqs. 2.240, 2.241 for plane wave reflection, we define2 a reflection
coefficient Γ as
V− Z0 − ZL
Γ ≡ − 0+ = (4.60)
V0 Z0 + ZL
The voltage and current on the transmission line are then given by
V (z) = V0+ eiβz − Γe−iβz


(4.61)

I(z) = I0+ eiβz + Γe−iβz

(4.62)
We see that if the load is “impedance matched” to the characteristic impedance of the
transmission line, then reflected waves are suppressed.
Note that ZL is defined as the ratio of voltage to current amplitude at z = 0. Since
the superposition of incident and reflected waves set up a standing wave on the line,
we can expect the ratio V (z)/I(z) to vary with z. Therefore the impedance along the
line varies with z when the load is mismatched. If we seek the impedance Z a distance
−d from the load , then from Eqs. 4.39, 4.61 4.62 we have
e−iβd − Γeiβd
Zline = Z0 ·
e−iβd + Γeiβd
(1 − Γ) − i(1 + Γ) tan βd
= Z0 ·
(1 + Γ) − i(1 − Γ) tan βd
ZL − iZ0 tan βd
= Z0 · (4.63)
Z0 − iZL tan βd
Figure 4.5 diagrams the transmission line terminated in the load ZL and the impedance
Zline a distance −d to the left of the terminating load.
2 We choose here the phase of the reflected voltage to be consistent with the convention used in Sec-
tion 2.10.6 for reflected electromagnetic plane waves. The electrical engineering literature commonly
uses Γ = V0− /V0+ .
126 Waveguides and Equivalent Circuits

Fig. 4.5 Transmission line impedance Zline (−d) with the line terminated by a load ZL 6= Z0 .
The impedance mismatch sets up a reflection on the line and a standing wave due to the
superposition of V + (z) and V − (z) waves.

Power Propagation on a Lossless Transmission Line. The cycle averaged power on a

transmission line is given by the real part of the product of the voltage and current.
Using Eqs. 4.61, 4.62
1
PT = Re [V (z)I ∗ (z)] (4.64)
2
2
1 V0+ h
2
i
= Re 1 + Γ∗ ei2βz − Γe−i2βz − |Γ| (4.65)
2 Z0
2
1 V0+ h 2
i
= 1 − |Γ| (4.66)
2 Z0
This power expression is reminiscent of the power calculated from the Poynting vector
of a propagating plane wave, Eqs. 2.267, 2.268, 2.271. Clearly if the line is impedance
matched so that there is no reflection, Γ = 0, then all the power propagates down the
line. If Γ 6= 0 then some power is reflected back along the incident direction, setting
up a V − I standing wave along the line.

Transmission and Reflection at a Line Junction. Suppose we have two transmission

lines, each with a characteristic impedance Z1 , Z2 respectively, and they are joined at
z = 0 with the Z1 line along z < 0 and the Z2 line along z > 0. Assume further the
second line with impedance Z2 is infinitely long (or terminated with an impedance
matching load) so that there are no reflections at the end of it. At the junction the
reflection coefficient is
Z1 − Z2
Γ= (4.67)
Z1 + Z2
so that the voltage along the z < 0 section is
 Transmission Lines 127

Fig. 4.6 Two transmission lines joined at z = 0. The line to the left (z < 0) is characterized
by line impedance Z1 , and the line to the right (z > 0) by Z2 . Reflection coefficient Γ and
transmission coefficient T determine the amplitude of the reflected and transmitted waves,
respectively.

V1 (z) = V0+ eiβ1 z − Γe−iβ1 z

z<0 (4.68)

The transmission junction is illustrated in Fig. 4.6. On the z > 0 side, the voltage
travels only in the +z direction and the transmitted amplitude is some fraction T of
V1+ . The transmitted wave is then represented by

V2 (z) = TV1+ eiβ2 z z>0 (4.69)

At the junction itself the two voltage must be equal

V1 (z = 0) = V2 (z = 0) (4.70)
V1+ (1
− Γ) = TV1+ (4.71)
2Z2
=T (4.72)
Z1 + Z2
So we find that in terms of the characteristic impedances in the two lines the reflection
and transmission coefficients are analogous to reflection and transmission of plane
waves at a material surface.
Z1 − Z2
Γ= (4.73)
Z1 + Z2
2Z2
T= (4.74)
Z1 + Z2
4.3.4 Equivalence of Plane Waves and Transmission Lines
The correspondence between electromagnetic plane wave propagation (Section 2.10,
Fig. 2.8) through media characterized by µ,  and voltage/current transport along a
lossless transmission line (Section 4.3.2) characterized by L, C is indicated in Table 4.1
below. Analogous Helmholtz-like wave equations govern propagation, reflection, and
transmission in both cases.
128 Waveguides and Equivalent Circuits

Table 4.1 Equivalent quantities between plane waves and transmission lines

Plane Wave Transmission Line

Ex (z) V (z)
Hy (z) I(z)
√ √
k = ω µ β = ω LC
q
Z = µ L
p
Z= C
R Γ
T T

4.4 Special Termination Cases

In this section we discuss some special cases for ZL in the lossless terminated trans-
mission line, Fig. 4.5 in Section 4.3.3.

4.4.1 Shorted Transmission Line

If we set ZL = 0 then the reflection coefficient, from Eq. 4.60, becomes unity, Γ = 1,
and from Eqs. 4.61, 4.62 the transmission voltage and current become

V (z) = V0+ eiβz − e−iβz = i2V0+ sin βz

(4.75)
+ iβz −iβz +


I(z) = I0 e + e = 2I0 cos βz (4.76)

Note that at the load point (z = 0) the voltage is null and the current is maximum as
would be expected from a short circuit. Note also that since the product V (z)I ∗ (z) is
pure imaginary no power is delivered to the load. The impedance along the line, from
Eq. 4.63 is
Zline = −iZ0 tan βd (4.77)

4.4.2 Open Circuit Transmission Line

Suppose ZL = ∞ so that in Fig. 4.5 the load is completely removed. In this case, from
Eq. 4.60, Γ = −1, and there is 100 % reflection at the load. The voltage and current
expressions become

V (z) = V0+ eiβz + e−iβz = 2V0+ cos βz

(4.78)
+ iβz −iβz +


I(z) = I0 e − e = i2I0 sin βz (4.79)

As expected the voltage is maximum at the load point and the current is null. The
expression for the impedance along the line, from Eq. 4.63, is

Zline = iZ0 cot βd (4.80)

and again no power is transmitted to the load.

 Special Termination Cases 129

Fig. 4.7 Diagram of how a segment of transmission line with impedance Z1 and length λ/4
can be used to match an incoming transmission line Z0 with a load ZL .

4.4.3 Line Impedance at d = −λ/2

At half-wave points along the line, Eq. 4.63 shows that

Zline = ZL (4.81)

At these points the impedance is independent of the transmission line characteristics.

4.4.4 Line Impedance at d = −λ/4

At the quarter-wave points we have, from Eq. 4.63,
ZL − iZ1 tan βd
Zline = Z1 (4.82)
Z1 − iZL tan βd
Z12
lim Zline = (4.83)
βd→π/2 ZL

This quarter-wave transformer can be used as a λ/4 length of line, with impedance
Z1 to match an input transmission line of impedance Z0 to a given load ZL . Figure
4.7 shows how this matching can be accomplished. In order to suppress reflection at
the Z0 /Z1 junction, Γ must be set equal to zero there. Therefore,

Zline = Z0 at − d = λ/4 (4.84)

and p
Z1 = Z0 ZL (4.85)
The line impedance at the quarter-wave point, Eq. 4.84, is called a quarter-wave trans-
former because the load impedance is “transformed” by the square of the characteristic
impedance of the transmission line.
130 Waveguides and Equivalent Circuits

4.4.5 Slightly Lossy Transmission Lines

An ideal, lossless transmission line is characterized by L and C, themselves calculated
from Maxwell’s equations and the geometry of the line. However, a real transmission
line always has some resistance in the conducting elements and some leakage current in
the dielectric separating them. As diagramed in Fig. 4.3, these departures from ideality
are characterized by R in the conductors and G, the “shunt conductance” through the
insulating dielectric material. The complex propagation parameter is given by Eq. 4.32,
β̃ 2 = (R − iωL)(G − iωC) (4.86)
or p
β̃ = ± (R − iωL)(G − iωC) (4.87)
Assuming that R << ωL and G << ωC we can write 4.87 as
√
r r
R G
β̃ = ±iω LC 1 + i · 1+i (4.88)
ωL ωC
Expanding the square-root expression as the first two terms in a Taylor’s series,
√
   
i R i G
β̃ = ±iω LC 1 + · 1+ (4.89)
2 ωL 2 ωC
√
 
i R i G RG
= ±iω LC 1 + + − (4.90)
2 ωL 2 ωC (ωLC)2
Since R/ωL and G/ωC are already small compared to unity, the last term on the right
can be dropped, and we write
√
  
i G R
β̃ ' ±iω LC 1 + + (4.91)
2 ωC ωL
r !
√
r
1 L C
' ±iω LC ∓ G +R (4.92)
2 C L

We then identify the real propagation parameter β

√
β = ω LC (4.93)
and the dissipation term γ
r r !
1 L C
γ= G +R (4.94)
2 C L
 
1 R
= GZ0 + (4.95)
2 Z0
The signs of β and γ have been chosen so that a wave propagating in the positive z
direction has a positive argument, eiβz , and the real exponential term γ dissipates the
amplitude to zero as z → ∞.

eiβ̃z = ei(β+iγ)z = e(iβ−γ)z (4.96)

 Waveguides 131

4.5 Waveguides
Up to now we have studied plane waves in Chapter 2 and transmission lines in the
present chapter. Waves propagating within or on guiding structures, which essentially
define the transverse field boundary conditions, are related to unbounded plane waves
and even more closely related to transmission lines. But guided waves also exhibit
unique characteristics that require extension of our familiar ideas of voltage, current,
capacitance and inductance. Waveguiding at microwave frequencies was developed for
radar during WWII, but now integrated optical circuits at the nanoscale use waveg-
uides to transport optical signals onto and around integrated microchips. We discuss
in this section the elements of waveguide theory and the relation guided waves bear
to plane electromagnetic waves and to transmission lines.
The prototypical transmission line, consisting of two parallel slab conductors sep-
arated by a dielectric, is shown in Fig. 4.4. In a sense this structure can be considered
a wave guide since the E-field terminates on the two conducting slabs separated by a
constant distance in the y − z plane, although the H-field is unbounded along y. It is
clear from the front and side views, that the E- and H-field lines are confined to the
x−y plane with no field components along the direction of propagation, z. Waves with
this property are called transverse electromagnetic (TEM) waves. An ideal plane wave
propagating in space also exhibits phase fronts confined to the plane orthogonal to
the propagation direction, so they are TEM waves as well. Note that if we added two
conducting slabs in the x − z plane so as to enclose a volume with a single conductor
of rectangular x − y cross-section, no TEM field could exist since the four conducting
sides would define an interior space of constant voltage (and hence a null E-field).
However, waveguides with rectangular, cylindrical, and even arbitrarily shaped cross-
section do exist, but in order to understand them we have to extend our ideas beyond
the familiar TEM wave. Waves propagating within structures of a single conductor
are classified into two types: transverse electric (TE) waves with an H-field component
along the propagation direction z (but E-field only in the transverse x − y plane), and
transverse magnetic (TM) waves with an E-field component along the propagation
direction (but only with transverse H-field components).
4.5.1 Field Solutions for TM and TE Waves
We assume field solutions to the Helmholtz equation that will have some spatial varia-
tion in the x − y plane, subject to the boundary conditions of a known perfect electric
conductor (PEC) guide structure, and propagate along the z direction with eiβz . For
the time being we will take the propagation parameter to be real (lossless medium).
The general form of the phasor fields are then
E(x, y, z) = [E(x, y) + Ez (x, y)ẑ] eiβz (4.97)
iβz
H(x, y, z) = [H(x, y) + Hz (x, y)ẑ] e (4.98)
Once again we write the two curl equations of Maxwell with harmonic time dependence,
no current sources in nonmagnetic material (µ = mu0 ).
∇ × E = iωµ0 H
∇ × H = −iωE
132 Waveguides and Equivalent Circuits

and write out explicitly the various components of H and E. For the H-field compo-
nents we have

∂Ez
− iβEy = iωµ0 Hx (4.99)
∂y
∂Ez
iβEx − = iωµ0 Hy (4.100)
∂x
∂Ey ∂Ex
− = iωµ0 Hz (4.101)
∂x ∂y

and for the E-field components

∂Hz
− iβHy = −iωEx (4.102)
∂y
∂Hz
iβHx − = −iωEy (4.103)
∂x
∂Hy ∂Hx
− = −iωEz (4.104)
∂x ∂y

We can use Eqs. 4.99, 4.103 to write Hx in terms of derivatives of Ez and Hz in the
transverse x − y plane. Similarly we can use Eqs. 4.100, 4.102 to also write Hy in terms
of transverse derivatives of Ez , Hz . Proceeding in the same way for Ex and Ey we have
finally four equations,

 
i ∂Ez ∂Hz
Hx = − ω − β (4.105)
(k 2 − β 2 ) ∂y ∂x
 
i ∂Ez ∂Hz
Hy = 2 ω +β (4.106)
(k − β 2 ) ∂x ∂y
 
i ∂Hz ∂Ez
Ex = 2 ωµ 0 + β (4.107)
(k − β 2 ) ∂y ∂x
 
i ∂Hz ∂Ez
Ey = − 2 ωµ0 −β (4.108)
(k − β 2 ) ∂x ∂y

where k, as always, is related to the frequency ω and velocity of light through the
√ √

medium (v = 1/ µ0 ) by the expression k = ω µ0 . The factor k 2 − β 2 can be
interpreted as the square of a propagation parameter in the x − y plane since β is
always along the z axis. This transverse propagation parameter is called the “cutoff”
parameter kc .

kc2 = k 2 − β 2 (4.109)

TE Waves. In the case of TE waves Ez = 0 and Eqs. 4.105–4.108 become

 Waveguides 133

i ∂Hz
Hx = ·β (4.110)
kc2 ∂x
i ∂Hz
Hy = ·β (4.111)
kc2 ∂y
i ∂Hz
Ex = 2
· ωµ0 (4.112)
kc ∂y
i ∂Hz
Ey = − 2
· ωµ0 (4.113)
kc ∂x

In order to find the transverse fields we have to use the Helmholtz equation to find
the permitted values of Hz . We write the expression

∂2 ∂2 ∂2
 
2 2 2


∇ +k Hz = + 2 + 2 + k Hz = 0 (4.114)
∂x2 ∂y ∂z

But since
∂ 2 Hz
= −β 2 Hz (4.115)
∂z 2
the Helmholtz equation reduces to a differential equation for the transverse Hz field
gradients
 2
∂2

∂ 2
+ + kc Hz = 0 (4.116)
∂x2 ∂y 2
Allowed solutions will be a function of the transverse boundary conditions.

TM Waves. For TM waves we follow the same procedure except we set Hz = 0 and
write Eqs. 4.105–4.108 as

i ∂Ez
Hx = − · ω (4.117)
kc2 ∂y
i ∂Ez
Hy = · ω (4.118)
kc2 ∂x
i ∂Ez
Ex = ·β (4.119)
kc2 ∂x
i ∂Ez
Ey = 2
·β (4.120)
kc ∂y

The Helmholtz equation for Ez reduces to

∂2 ∂2
 
2
+ 2 + kc Ez = 0 (4.121)
∂x2 ∂y

and allowed solutions will again be a function of the transverse boundary conditions.
134 Waveguides and Equivalent Circuits

4.5.2 Parallel Plate Waveguide

Here we apply the formal development of the last section to a simple but very useful
example, the parallel plate waveguide. This structure is essentially the same as the
parallel plate transmission line studied in Section 4.3.2, but here we will see that the
transmission line TEM waves (with no field components in the direction of propaga-
tion) can be be complemented with TE modes and TM modes. The two conducting
plates of the waveguide are separated by s between the plates (along the x direction)
and have a width w (along the y direction) that is much greater than the separation
s. We can therefore assume that the solutions we seek will not be functions of y.
TE Modes. In order to find the allowed TE modes we start with the reduced Helmholtz
equation for the TE case, Eq. 4.116, and write the solutions for Hz . By inspection we
write the general solution in terms of the real sin and cos functions and refer to the
coordinate system of Fig. 4.4. The amplitudes A, B are determined by the boundary
conditions.
Hz (x, y) = A sin kc x + B cos kc x (4.122)
Since we posit that the plates are perfect conductors (and the intervening dielectric
lossless) we have for boundary conditions
Ey (0, y) = 0 and Ey (s, y) = 0 (4.123)
Then from Eq. 4.113
i ∂Hz i ∂Hz iβz
Ey (x, y) = − · ωµ0 = − 2 · ωµ0 e (4.124)
kc2 ∂x kc ∂x
iωµ0
=− [A cos kc x − B sin kc x] eiβz (4.125)
kc
The boundary conditions, Eq. 4.123, require that A = 0 and puts a condition on the
argument of the sin term, that at x = s the E-field must be null so
nπ
kc s = nπ n = 1, 2, 3 . . . or kc = n = 1, 2, 3 . . . (4.126)
s
The boundary conditions impose that only discrete values of kc are permitted (de-
pending on the plate separation s), and the expression for the propagation parameter
along the z direction becomes
r  nπ 2
p
β = ± k 2 − kc2 or β = ± k 2 − (4.127)
s
The solutions for Hz (x, y) are
nπ
Hz (x, y) = Bn cos x (4.128)
s
or  nπ 
Hz (x, y) = Bn cos x eiβz (4.129)
s
Note that when k > kc , β is positive and the wave propagates along z. When k = kc ,
then β = 0 and the wave is stationary with zero phase velocity. If k falls below kc ,
 Waveguides 135

then β becomes pure imaginary. The wave no longer propagates in the waveguide
but decays exponentially to 1/e of its initial amplitude at a characteristic distance of
x = 1/β. The propagating wave becomes an evanescent wave. This threshold behavior
for propagation of mode n within the waveguide is the reason that kc is called the
“cutoff” parameter. Below cutoff, the wave does not propagate. The cutoff parameter
kc is related in the usual way to wavelength and frequency.
2π
kc = and ωc = kc v (4.130)
λc
where v is the propagation velocity in the gap dielectric of the waveguide. If the
dielectric is air or vacuum, then is the velocity of light, c. For a given waveguide
“mode” n the cutoff frequency must be greater than kcn v.
Once we have the solutions for Hz (x, y) we can get the solutions for all the trans-
verse field components from Eqs. 4.110–4.113. Since there is no transverse dependence
in the y direction, Eqs. 4.111, 4.112 show that Hy = Ex = 0 and

β  nπ 
Hx = −i Bn sin x eiβz (4.131)
kc s
ωµ0  nπ 
Ey = i Bn sin x eiβz (4.132)
kc s
Note the difference in sign between Hx and Ey . This sign difference is consistent with
the direction of energy propagation, determined by the Poynting vector, along the
positive z direction, S = E × H. The wave impedance is given by the ratio of the
E-field amplitude to the H-field amplitude,
|Ey | ωµ0 k
ZTE = = = Z0 (4.133)
|Hx | β β
For the parallel plate waveguide the TE modes consist of a “triplet” of components,
Hz , Hx , Ey and the modes are labeled TEn , n = 1, 2, 3 . . .

TM Modes. To find the field components for the TM modes we follow a parallel
procedure to the TE case. The first step is to find the Ez solutions from the reduced
Helmholtz equation, Eq. 4.121, subject to the boundary conditions on the parallel plate
waveguide at x = 0 and x = s. Once again the general solution is

Ez (x, y) = A sin kc x + B cos kc x (4.134)

with boundary conditions that Ez (0, y) = Ez (s, y) = 0. Therefore B = 0 and sin kc x = 0

when x = 0 or when x = s. The solutions to Eq. 4.121 that respect the boundary
conditions are therefore
 nπ 
Ez = An sin x n = 0, 1, 2, 3 . . . (4.135)
s
and  nπ 
Ez = An sin xeiβz (4.136)
s
136 Waveguides and Equivalent Circuits

with r
p  nπ 2
β = ± k 2 − kc2 = ± k2 − (4.137)
s
From Eqs. 4.117–4.120 we find the transverse field components.
ω  nπ 
Hy = i An cos x eiβz (4.138)
kc s
β  nπ 
Ex = i An cos x eiβz (4.139)
kc s
The impedance for the TM waves is
Ex β β
ZTM = = = Z0 (4.140)
Hy ω k

and the TM modes consist of a triplet of three field components: Ez , Ex , Hy . Note

that for the lowest n = 0 mode, there is no cut off frequency, Ez = 0, and the Hy , Ex
components have constant amplitudes. In fact the TM0 mode is identical to the TEM0
mode, represented in Fig. 4.4.

4.6 Rectangular Waveguides

The parallel plate waveguide is essentially a one-dimensional problem since the trans-
verse spatial variation of the TE and TM modes depend only on x. Here we examine
the more realistic case of a rectangular wave guide with finite dimensions in both x and
y. We assume again boundary conditions on a perfect electrical conductor (PEC) and
apply those E-field conditions at the waveguide edges along x and y. By convention
we denote the long and short sides of the rectangular cross section as l, h, respectively
and align l along x and h along y so that the lower left corner of the waveguide cross
section is at the coordinate origin. The guide is filled with some nonmagnetic, loss-
less dielectric material characterized by permeability µ0 and permittivity . Figure 4.8
shows the layout for the rectangular waveguide.

4.6.1 TE Modes
The transverse electric (TE) modes are defined as those waves that propagate in the
guide with E-fields only transverse to the z (propagation) direction. Therefore we seek
Hz solutions to the transverse Helmholtz equation.
 2
∂2

∂ 2
+ + kc Hz (x, y) = 0 (4.141)
∂x2 ∂y 2

The solutions will be functions of both coordinates, but we can expect them to be
uncoupled. Therefore we can posit solutions of the form Hz (x, y) = X(x)Y (y). Sub-
stitution into Eq. 4.141 and division by X(x)Y (y) results in

1 ∂2X 1 ∂2Y
+ = −kc2 (4.142)
X ∂x2 Y ∂y 2
 Rectangular Waveguides 137

Fig. 4.8 Rectangular waveguide with sides x = 2a and y = 2b. Inside the conducting
boundaries the waveguide is filled with a lossless dielectric characterized by µ0 , .

Since this result must be true for all values of x, y, the terms on the left must themselves
be equal to constants that we write

1 ∂2X 1 ∂2Y
= −kx2 = −ky2 (4.143)
X ∂x2 Y ∂y 2

and
kx2 + ky2 = kc2 (4.144)
The solution for Hz (x, y) is a product of the general solution in x and the general
solution in y.

Hz (x, y) = (A cos kx x + B sin kx x) · (C cos ky y + D sin ky y) (4.145)

The boundary conditions at x, y = 0 and x = a, y = b are

Ex (x, 0), Ex (x, b) = 0 and Ey (0, y), Ey (a, y) = 0 (4.146)

The transverse E-fields are obtained from the longitudinal H-field Hz by using Eqs. 4.112, 4.113
ωµ0
Ex = i ky (A cos kx x + B sin kx x) · (−C sin ky y + D cos ky y) (4.147)
kc2
ωµ0
Ey = i 2 kx (−A sin kx x + B cos kx x) · (C cos ky y + D sin ky y) (4.148)
kc

The condition on Ex (x, y = 0) implies that D = 0 and the condition on Ex (x, y = b)

implies that ky = nπ/b with n = 0, 1, 2 . . . Similarly the condition on Ey (x = 0, y)
138 Waveguides and Equivalent Circuits

requires that B = 0 and the condition on Ey (x = a, y) requires kx = mπ/a with

m = 0, 1, 2 . . . Therefore the general solution for Hz (x, y), Eq. 4.145, becomes
mπx nπy
Hz = Amn cos cos (4.149)
a b
and
mπx nπy iβz
Hz (x, y, z) = Amn cos cos e (4.150)
a b
The arbitrary amplitude constant Amn is a product of the constants A, C in Eq. 4.145.
Now we get the transverse fields using Eqs. 4.110–4.113.

βmπ mπx nπy iβz

Hx = −i Amn sin cos e (4.151)
kc2 a a b
βnπ mπx nπy iβz
Hy = −i 2
Amn cos sin e (4.152)
kc b a b
ωµ0 nπ mπx nπy iβz
Ex = −i 2 Amn cos sin e (4.153)
kc b a b
ωµ0 mπ mπx nπy iβz
Ey = i Amn sin cos e (4.154)
kc2 a a b

As always we get the cutoff propagation parameters and frequencies from

q r  mπ 2  nπ 2
p

β = k − kc = k − kx + ky = k 2 −
2 2 2 2 2 − (4.155)
a b
The propagation parameter along z becomes real when
r
mπ 2  nπ 2
k > kc = − (4.156)
a b
The TE wave impedance is denoted by ZTE and is independent of the mode labeling
n, m.
Ex k
ZTE = = Z0 (4.157)
Hy β
Because we have posited that a > b, Eq. 4.156 shows that the TE mode with the lowest
cutoff frequency is TEmn = TE10 .

4.6.2 TM Modes
The transverse magnetic modes are defined as those modes with H-fields only in the
transverse (x − y) plane. Therefore the field component in the z direction must be an
E-field. The waves propagate along z and Ez is written as the product Ez (x, y)eiβz .
Therefore we seek Ez solutions to the same Helmholtz wave equation as for the TE
modes.  2
∂2 ∂2

∂ 2
+ 2 + 2 + kc Ez = 0 (4.158)
∂x2 ∂y ∂z
 Rectangular Waveguides 139

The general solution for Ez has the same form as that for Hz , Eq. 4.145, and subject
to boundary conditions,

Ez (x, 0), Ez (x, b) = 0 and Ez (0, y), Ez (a, y) = 0 (4.159)

The expression for Ez is therefore

mπx nπy iβz
Ez (x, y, z) = Ez (x, y)eiβz = Bmn sin sin e (4.160)
a b
with m, n = 1, 2, 3 . . . As with the TE modes the cutoff propagation parameter is given
by
kc2 = kx2 + ky2 (4.161)
and the boundary conditions dictate that
mπ nπ
kx = and ky = (4.162)
a b
The transverse field components are obtained from Ez by using Eqs. 4.117–4.120.
ωnπ mπx nπy iβz
Hx = −i Bmn sin cos e (4.163)
bkc2 a b
ωmπ mπx nπy iβz
Hy = i 2
Bmn cos sin e (4.164)
akc a b
βmπ mπx nπy iβz
Ex = i 2
Bmn cos sin e (4.165)
akc a b
βnπ mπx nπy iβz
Ey = i Bmn sin cos e (4.166)
bkc2 a b
The propagation parameter is given, as usual, by
r  mπ 2  nπ 2
p
β = ± k − kc = ± k 2 −
2 2 − (4.167)
a b
We can see from Eqs. 4.163–4.166 that TM01 or TM10 vanish (in fact all TM modes
with a zero index vanish) so the lowest TM mode is TM11 with a cutoff propagation
parameter given by r 
π 2  π 2
kc = + (4.168)
a b
If one of the sides becomes much longer than the other (say, a >> b), then
π
kc → (4.169)
b
and the cutoff kc becomes equal the parallel plate waveguide case (Section 4.5). Finally
the TM mode impedance is given by
Ex β
ZTM = = Z0 (4.170)
Hy k
140 Waveguides and Equivalent Circuits

Fig. 4.9 Schematic of cylindrical waveguide. The cylinder surface is assumed to be a perfect
conductor. Cylindrical coordinates are r, ϕ, z as shown. The radius of the cylinder is a.

4.7 Cylindrical Waveguides

The cylindrical geometry is important not only for waveguides and transmission lines
in the conventional microwave domain but also for nano-scale holes and hole arrays in
the visible and near infrared regions. A sensible point of departure for understanding
light transmission through these cylindrically symmetric structures is an analysis of
the field modes they can support. The procedure is really no different than what we
have done for the rectangular wave guides. We choose a coordinate system in which
the solutions to the Helmholtz equation will separate, and we will be able to write the
solutions to Ez and Hz as a product of three factors R(r)Φ(ϕ)eiβz where R(r) is the
radial field dependence and Φ(ϕ) is the azimuthal angular dependence around the z
axis. A schematic of the guide is shown in Fig. 4.9

4.7.1 TE Modes
As always, TE modes are defined by the presence of electric fields only in the trans-
verse plane. The field function along the z axis must be an H-field, and we write the
Helmholtz equation
∇ 2 Hz + k 2 Hz = 0 (4.171)
Then we write the ∇2 or Laplacian operator in cylindrical coordinates and factor
Hz (r, ϕ, z) = Hz (r, ϕ)eiβz . We take the expression for the Laplacian operator from
Eq. C.33 of Appendix C.
 2
1 ∂2

∂ 1 ∂ 2
+ + + kc Hz (r, ϕ) = 0 (4.172)
∂r2 r ∂r r2 ∂ϕ2
 Cylindrical Waveguides 141

Now we take the posited product solution

Hz (r, ϕ) = R(r)Φ(ϕ) (4.173)

and substitute it into 4.172. After multiplication by r2 we have

r2 d2 R r dR 1 d2 Φ
2
+ + r2 kc2 = − (4.174)
R dr R dr Φ dϕ2
The left side is a function only of r and the right side only of ϕ. In order for the
equation to be valid for the entire range of r, ϕ the two sides must be equal to a
2
constant. We set the separation constant equal to lϕ and write

1 d2 Φ 2
− = lϕ (4.175)
Φ dϕ2
or rearranging
d2 Φ 2
+ lϕ Φ=0 (4.176)
dϕ2
From the left side of Eq. 4.174 we have

d2 R dR
r2 + r2 kc2 − lϕ
2


+r R=0 (4.177)
dr2 dr
Note that lϕ is unitless. From inspection we can write down a solution to Eq. 4.176.

Φ = eilϕ ϕ (4.178)

Now in order to represent a physical entity Φ must be a single-valued function, meaning

that Φ[lϕ (ϕ + 2π)] must have the same value as Φ(lϕ ϕ) or

Φ[(lϕ (ϕ + 2π)] = eilϕ ϕ eilϕ 2π = Φ(lϕ ϕ) (4.179)

Therefore the separation constant lϕ must be an integer, lϕ = n = 1, 2, 3 . . . The

general solution to Eq. 4.176 can also be written in terms of real sin and cos functions.

Φ(ϕ) = A sin nϕ + B cos nϕ (4.180)

The radial equation becomes

d2 R dR
r2 + r2 kc2 − n2 R = 0


2
+r (4.181)
dr dr
This expression is one of the forms of Bessel’s differential equation. There exists a
family of functions that are solutions to this equation. In fact there are two independent
families, Bessel functions of the first kind, Jn (x), and of the second kind, Nn (x). The
functions of the second kind, the Neumann functions, tend to −∞ as x → 0 so they
are not physically acceptable solutions. The Bessel functions of the first kind are finite
at the origin and exhibit physically acceptable behavior. The first few Bessel function
142 Waveguides and Equivalent Circuits

Fig. 4.10 First three Bessel functions Jn (x)

Jn (x) are plotted in Fig. 4.10. Putting together the angular and radial solutions to the
reduced Helmholtz equation, we have

Hz (r, ϕ) = (A sin nϕ + B cos nϕ) Jn (kc r) (4.182)

The next step is to apply boundary conditions at r = 0 and r = a, where a is the radius
of the hole. But in order to do so we must write the E-field and H-field components
in the r, ϕ directions in terms of the z-axis components. These components are found
from writing down the six Maxwell’s curl equations in cylindrical coordinates and
reducing them to four relations of r, ϕ in terms of z as was done for Eqs. 4.105–4.108.
The result is
 
1 ∂Ez ωµ0 ∂Hz
Er = i 2 β + (4.183)
kc ∂r r ∂ϕ
 
1 β ∂Ez ∂Hz
Eϕ = i 2 − ωµ0 (4.184)
kc r ∂ϕ ∂r
 
1 ω ∂Ez ∂Hz
Hr = −i 2 −β (4.185)
kc r ∂ϕ ∂r
 
i ∂Ez β ∂Hz
Hϕ = i 2 ω + (4.186)
kc ∂r r ∂ϕ

On the perfectly conducting wall of the cylinder the E-field tangent to the wall must
vanish. Therefore Eϕ (r = a, ϕ) = 0. From Eq. ?? we have
ωµ0
Eϕ (r, ϕ, z) = −i (A sin nϕ + B cos nϕ) Jn0 (kc r)eiβz (4.187)
kc
 Cylindrical Waveguides 143

Fig. 4.11 Derivatives of the first three Bessel functions, showing the positions of the first
few zero-crossings.

Table 4.2 Some values of (kc a)nm for TE in a cylindrical waveguide

n (kc a)n1 (kc a)n2 (kc a)n3

0 3.832 7.016 10.174
1 1.841 5.331 8.536
2 3.054 6.706 9.970

where Jn0 (kc r) is the radial derivative of the Bessel function. Applying the boundary
condition at r = a
ωµ0
Eϕ (a, ϕ, z) = −i (A sin nϕ + B cos nϕ) Jn0 (kc a)eiβz = 0 (4.188)
kc

Since the r dependence is only in the Bessel term, we must have

Jn0 (kc a) = 0 (4.189)

But this requirement fixes kc to the zeros of Jn0 (kc a). Each function Jn0 , labeled by
n will have a series of zeros where Jn0 [(kc a)nm ] corresponds to the mth zero of the
nth Bessel function derivative. The arguments of the zeros (or roots) of Jn0 are not
simple multiples of π as in the case of sin and cos functions. They have to be found
numerically and 4.2 provides values for the first few arguments (kc a)nm corresponding
to the zeros of the Jn0 functions. The first three Bessel function derivatives are plotted
in Fig. 4.11. Evidently we obtain kc for the nmth mode from
144 Waveguides and Equivalent Circuits

kc anm
kcnm = (4.190)
a
We obtain the propagation parameter along z for the nmth mode in the usual way
q
βnm = k 2 − kc2nm (4.191)

and with βnm in hand we construct Hz (r, ϕ, z) = Hz (r, ϕ)eiβnm z . Then from Eqs. 4.183–
4.186 we calculate the transverse fields for the nmth mode. The notation nm has been
omitted from kcnm and βnm so as not encumber the expressions more than necessary.
ωµ0 n
Er = i 2 (A cos nϕ − B sin nϕ) Jn (kc r)eiβz (4.192)
kc r
ωµ0
Eϕ = −i (A sin nϕ + B cos nϕ) Jn0 (kc r)eiβz (4.193)
kc
β
Hr = i (A sin nϕ + B cos nϕ) Jn0 (kc r)eiβz (4.194)
kc
βn
Hϕ = i 2 (A cos nϕ − B sin nϕ) Jn (kc r)eiβz (4.195)
kc r
In fact the A and B amplitude factors in the angular part of the transverse field expres-
sions are not independent. We can easily see this from Eq. 4.178. The “normalization
integral” of the angular solution in complex form is
Z 2π
Φ∗ Φ dϕ = 2π = const. (4.196)
0
and writing the normalization of the angular solution in terms of real sin and cos
functions Z 2π
2 1 2
A + B 2 = const.


(A sin nϕ + B cos nϕ) dϕ = (4.197)
0 2
or in other words the only restriction on the amplitudes of the sin and cos terms is that
the sum of the squares of A and B must always be equal to some constant. Therefore
we can set A = 0 and let B 2 = const. Then the transverse fields can be written in
somewhat simpler form
ωµ0 n
Er = −i 2 (B sin nϕ) Jn (kc r)eiβz (4.198)
kc r
ωµ0
Eϕ = −i (B cos nϕ) Jn0 (kc r)eiβz (4.199)
kc
β
Hr = i (B cos nϕ) Jn0 (kc r)eiβz (4.200)
kc
βn
Hϕ = −i 2 (B sin nϕ) Jn (kc r)eiβz (4.201)
kc r
The TE mode impedance is given by
Er k
ZTE = = Z0 (4.202)
Hϕ β
 Cylindrical Waveguides 145

Table 4.3 Some values of (kc a)nm for TM in a cylindrical waveguide

n (kc a)n1 (kc a)n2 (kc a)n3

0 2.405 5.520 8.654
1 3.832 7.016 10.174
2 5.135 8.417 11.620

4.7.2 TM Modes
In the case of TM modes, there is no component Hz along the axis of symmetry, and
we seek solutions to the Helmholtz equation for Ez (r, ϕ, z) = Ez (r, ϕ)eiβz . As with the
TE modes we have a reduced Helmholtz equation for the transverse coordinates.
 2
1 ∂2

∂ 1 ∂ 2
+ + + k c Ez (r, ϕ) = 0 (4.203)
∂r2 r ∂r r2 ∂ϕ2
This expression is the same as Eq. 4.172 so the general solutions are the same.

Ez (r, ϕ) = (A sin nϕ + B cos nϕ) Jn (kc r) (4.204)

But in this case the boundary conditions on the E-field can be applied directly the
solutions for the Helmholtz equation.

Ez (r = a, ϕ) = 0 (4.205)

Therefore from Eq. 4.204 we have

Jn (kc a) = 0 (4.206)

and we get the cutoff parameter for the TMnm mode from the zeros of Jn Bessel
function. Table 4.3 shows some values for the roots of Jn . Once again the cutoff
parameters are obtained from
kc anm
kcnm = (4.207)
a
The propagation parameter along z is
p
βnm = k 2 − kc2 (4.208)

The transverse fields from Eqs. 4.183–4.186 and setting A = 0 are

β
Er = i (B cos nϕ) Jn0 (kc r)eiβz (4.209)
kc
βn
Eϕ = −i 2 (B sin nϕ) Jn (kc r)eiβz (4.210)
kc r
ωn
Hr = i 2 (B sin nϕ) Jn (kc r)eiβz (4.211)
kc r
ω
Hϕ = i (B cos nϕ) Jn0 (kc r)eiβz (4.212)
kc
146 Waveguides and Equivalent Circuits

Finally the mode impedance is given by

Er β
ZTM = = Z0 (4.213)
Hϕ k

4.8 Networks of Transmission Lines and Waveguides

Earlier in this chapter we discussed the properties of lumped circuit elements and
how these properties, such as capacitance, inductance and resistance, can be adapted
to distributed structures like transmission lines and waveguides. The purpose of the
present section is to show how transmission lines and waveguides themselves can be
grouped together into a network to achieve some predetermined design result with-
out having to solve Maxwell’s equations directly for elaborate source and boundary
conditions. The ultimate goal is then to adopt these ideas to nanoscale structures in
order to build networks of 2-D surface or gap plasmon waveguides or 1D interfacial
waveguides. The discussion here will be somewhat brief since an extended discussion
of networked-transmission-line theory is really a digression from the subject of light-
matter interaction. The idea is to introduce key concepts and notation so that readers
not trained in electrical engineering can grasp the significance of these analysis tools.

4.8.1 Impedance and Admittance Matrices

Figure 4.12 shows a generalized waveguide network where we see propagation into
and away from some arrangement of couplers, dividers, amplifiers, etc. that constitute
the network. Often our interest in focused on the inputs and outputs at the network
“ports”, labeled in the figure as ports S1 − S4 . The ports are assumed to be single-
mode waveguides, and we can in principle know the transverse wave properties and
the propagation parameter of the active input and output ports. We can write the
voltage and current passing the nth port entrance-exit plane.

Vn = Vn+ + Vn−
In = In+ − In−

and relating the voltages and currents by the impedance Z for each port of the four-
port network of Fig. 4.12,
     
V1 Z11 Z12 Z13 Z14 I1
 V2   Z21 Z22 Z23 Z24   I2 
 = ·  (4.214)
 V3   Z31 Z32 Z33 Z34   I3 
V4 Z41 Z42 Z43 Z44 I4

or in short-hand matrix notation

[V ] = [Z] · [I] (4.215)

The inverse of the impedance is called the admittance Y , and the inverse of the
impedance matrix [Z] is the admittance matrix [Y ].

[Y ] = [Z]−1 (4.216)
 Networks of Transmission Lines and Waveguides 147

Fig. 4.12 A schematic waveguide network with two ports S1 , S2 showing inputs and outputs
and two ports S3 , S4 showing only outputs. The outputs at S1 , S2 can be thought of as
reflected waves.

or written out explicitly for the four-port case Eq. 4.216 becomes
     
I1 Y11 Y12 Y13 Y14 V1
 I2   Y21 Y22 Y23 Y24   V2 
 I3   Y31 Y32 Y33 Y34  ·  V3 
 =    (4.217)
I4 Y41 Y42 Y43 Y44 V4
The matrix equation 4.215 means that for each element Zij we have
Vi
Zij = (4.218)
Ij
where all Ik with k 6= j have been set to zero. Equation 4.218 means that we can
determine the impedance Zij if we inject current I at port j and measure the resulting
voltage V at port i with all the other ports closed or “open-circuited” in engineering
parlance. For the admittance matrix we have the analogous situation.
Ii
Yij = (4.219)
Vj
148 Waveguides and Equivalent Circuits

which means that in order to determine the admittance matrix element Yij , we apply
a voltage V to port j and measure the current I at port i with all other ports dis-
connected from the network (open-circuited). It is important to remember that the
network can be lossy in which case the matrix elements Zij or Yij are complex. It can
be shown that lossless networks have all Zij and Yij pure imaginary and that recipro-
cal networks are represented by [Z] and [Y ] matrices that are symmetric with respect
to interchange of indices. That is, for example, Zij = Zji . Physically the reciprocal
property means that if a current injected at port m produces a voltage at port n,
then injecting the same current in port n will produce the same voltage at port m.
Networks of linear, passive components such as resistors, capacitors, and inductors are
reciprocal. Networks with active nonlinear components with gain, such as amplifiers,
or not reciprocal.

4.8.2 Illustrative Example: two-port voltage divider

We illustrate these ideas with a simple two-port voltage divider network depicted in
Fig. 4.13. The first step is to shut off port S2 and inject current I1 at port S1 . Then

V1
Z11 = = ZA + ZC (4.220)
I1
because no current runs through ZB . The “transfer impedance” Z12 is obtained by
shutting off port S1 and injecting current into port S2 . No current flows through ZA
so all the injected current I2 flows through ZB and ZC . Then V2 = I2 (ZB + ZC ).
The voltage V1 appearing at the port S1 is just the current I2 flowing through the
“pull-up” resistor ZC . Therefore Z12 is ZC . It is easy to verify that this network is
reciprocal and that therefore Z22 = Z11 and Z12 = Z21 .

4.8.3 The Scattering Matrix

For high-frequency networks in the microwave and optical regime, it is difficult to
characterize voltage and current points because the transverse amplitudes across wave-
fronts need not be constant. Therefore, instead of using impedance and admittance,
the scattering matrix (or S-matrix) is often employed instead. An S-matrix relates volt-
age waves incident and reflected at the various network ports. Incident and reflected
power from traveling and standing waves lend themselves to easier measurement in
high-frequency networks. Consider port 1 in Fig. 4.12. The S-matrix relating the inci-
dent voltage wave amplitude V1+ to the reflected wave amplitude V1− is given by
 −    +
V1 S11 S12 S13 S14 V1
−
 V   S21 S22 S23 S24   V + 
 2−  =   ·  2+  (4.221)
 V3   S31 S32 S33 S34   V3 
V4− S41 S42 S43 S44 V4+

The usual rules of matrix multiplication obtain here. Thus, for a given network, the
voltage amplitude of a reflected wave at port i is related to the voltage amplitude of
an incident wave at port j by
Vi− = Sij Vj+ (4.222)
 Networks of Transmission Lines and Waveguides 149

Fig. 4.13 (A) Two-port impedance network. (B) Two-port impedance network implementing
a voltage divider. Note that the directions of I1 , I2 point into the ports towards the network.

with the implicit assumption that all other ports are quiescent with voltages at the
port inputs set to zero. The diagonal matrix elements express the ratio of reflected
amplitude to incident amplitude at a give port, assuming all other ports are quiescent
(i.e. matched impedances at all other ports). Thus Sii is a reflection coefficient at
port i. Since the S-matrix is essentially intended for waveguide networks, the ports
are assumed to have some characteristic impedance (for example, Z0 = 50Ω). The Sij
matrix element expresses the transmission coefficient from port j to port i.
150 Waveguides and Equivalent Circuits

4.8.4 Relation of S-matrix to Z-matrix

We know that the voltage and current at port n is the net result of transmission and
reflection.
Vn = Vn+ + Vn− (4.223)
In = In+ − In− (4.224)
Now suppose that the characteristic impedance at port n is unity, Z0n = 1. Then we
can also write
In = In+ − In− = Vn+ − Vn− (4.225)
and the voltage, current, and impedance matrices are related by
[V ] = [Z] · [I] = [Z] · [V + ] − [V − ] = [V + ] + [V − ]

Rearranging by grouping matrices [V + ] and [V − ] we have

[V + ] − [V + ][Z] = −[V − ] − [Z][V − ] = − [V − ] + [Z][V − ]


(4.226)
We can matrix-factor this last matrix equation if we define a unit matrix [U ] with only
elements of unity on the diagonal and zero elements elsewhere. Then we have
[V + ] ([U ] − [Z]) = −[V − ] ([U ] + [Z]) (4.227)
or
[V + ] ([Z] − [U ]) = [V − ] ([U ] + [Z])
[V − ] [Z] − [U ]
= (4.228)
[V + ] [Z] + [U ]
Now division by a matrix really means multiplication by the matrix inverse so
[V − ] −1
= ([Z] − [U ]) · ([Z] + [U ]) (4.229)
[V + ]
and from the definition of the S-matrix, Eq. 4.221, we can identify the S-matrix with
the right side of Eq. 4.229
[V − ] −1
[S] = = ([Z] − [U ]) · ([Z] + [U ]) (4.230)
[V + ]
4.8.5 The ABCD Matrix
Although the preceding sections have assumed that a network has n ports, in most
practical cases circuits are designed with many two-port networks in series (in “cas-
cade” in engineering parlance). A given two-port network is described by the matrix
equation      
V1 AB V
= · 2 (4.231)
I1 CD I2
In equation form
V1 = AV2 + BI2 (4.232)
I1 = CV2 + DI2 (4.233)
The diagram in Fig. 4.14 show how the ABCD matrix functions. If we have two net-
 Networks of Transmission Lines and Waveguides 151

Fig. 4.14 Schematic ABCD network. Note that the positive current I2 is pointing away from
the port entrance. This ABCD-matrix convention differs from the S-matrix where all positive
quantities point into the port.

Fig. 4.15 Two-port ABCD network: one impedance element.

works in series the overall result is obtained by matrix multiplication.

       
V1 A1 B 1 A2 B 2 V
= · · 3 (4.234)
I1 C1 D1 C2 D2 I3

The ABCD matrices can be used to construct elaborate networks from a few simple
elements. For example, suppose we have have an elementary two-port network con-
sisting of one impedance element as shown in Fig. 4.15 By applying Ohm’s law we
can identify the ABCD matrix elements. By inspection it is easy to determine that
A = 1, B = Z, C = 0, D = 1. Another example is a two-port network with a single
admittance as shown in Fig. 4.16 Again by Ohm’s law and remembering the Y = 1/Z,
a simple inspection of the coupled equations reveals that A = 1, B = 0, C = Y, D = 1.
Figure 4.17 shows an impedance T-network. The ABCD matrix for this network
can be obtained by a more systematic approach. First consider the A element and
Eq. 4.232. We see that if no current flows at the output (open circuit with I2 = 0),
then A = V1 /V2 . But in that case from Ohm’s law V1 = I1 (Z1 + Z3 ) and V2 = I1 Z1 .
Therefore A = (Z1 + Z3 ) /Z1 . To determine the B element we set V2 = 0 (output short
circuit). Then B = V1 /I2 . The result of applying Ohm’s law (or the Kirchhoff voltage
and current rules) is that B = Z1 + Z2 + Z1 Z2 /Z3 . Similarly we consider Eq. 4.233
152 Waveguides and Equivalent Circuits

Fig. 4.16 Two-port ABCD network: one admittance element.

Fig. 4.17 Two-port ABCD impedance T-network.

and set successively I2 , V2 = 0 to find the matrix elements C and D. The result is
C = 1/Z3 and D = 1 + Z2 /Z3 .
Another common network is the admittance network shown in Fig. 4.18. This
arrangement is often called the Pi-network. By methodically “open-circuiting” and
“shorting” the output terminals the elements of the ABCD matrix can be determined.

Two more useful elementary functions useful for constructing more elaborate net-
works are the transmission line segment and the inductive transformer. Figures 4.19
and 4.20 sketch these two circuit functions, and the ABCD matrix elements for all
these elementary circuit functions are entered in Table 4.4.

4.8.6 Reciprocal and Lossless Networks

Most common network elements such as capacitors, inductors, and resistors are linear,
and passive in the sense that they do not contribute to voltage gain. The matrix
elements representing these networks are in general complex quantities. Therefore in
general, a Z, Y, S, or ABCD matrix will contain 2N 2 independent elements where N
 Networks of Transmission Lines and Waveguides 153

Fig. 4.18 Two-port ABCD admittance Pi-network.

Fig. 4.19 Two-port ABCD network element: transmission line segment of length z.

is the number of ports in the network. For components with linear, passive response,
however, the matrices are always symmetric so that Xij = Xji . All the network matrix
representations considered here are symmetric. Furthermore it can also be shown that
if the network is lossless, containing only reactive elements that store power but do
not dissipate it, the network elements will be pure imaginary quantities. Coupling
networks that contain capacitance and induction but no significant resistance fall into
this category.

4.8.7 Comparison of Impedance Matrix and ABCD Matrix

We can specialize the impedance matrix, discussed in Section 4.8.1, to the two-port
network,
154 Waveguides and Equivalent Circuits

Fig. 4.20 Two-port ABCD network element: inductive coupler.

Table 4.4 ABCD elements for frequently used network functions

Network A B C D
Single Z 1 Z 0 1
Single Y 1 0 Y 1
Z1 Z1 Z2 1 Z2
T-Impedance 1+ Z3 Z1 + Z2 + Z3 Z3 1+ Z3
Y2 1 Y1 Y 2 Y1
Pi-Admittance 1+ Y3 Y3 Y1 + Y2 + Y3 1+ Y3

Ideal Transmission Line cos βz −iZ0 sin βz − Zi0 sin βz cos βz

1
Real Transmission Line cosh βz Z0 sinh βz Z0 sinh βz cosh βz
1
Inductive Coupler N 0 0 N

     
V1 Z11 Z12 I1
= · (4.235)
V2 Z21 Z22 −I2
Note that the sign of I2 has been reversed. The reason is that we seek to compare
the ABCD network scheme with the Z scheme. But the impedance network shown in
Fig. 4.13 assumes that positive current I2 points into port 2, whereas in the ABCD
network I2 points outward from the port. Written out in equation form we have

V1 = Z11 I1 − Z12 I2 (4.236)

V2 = Z21 I1 − Z22 I2 (4.237)

The set of Z-coupled equations 4.236, 4.237 must be compared to the set of ABCD-
coupled equations, 4.232, 4.233. Suppose we take I2 output current to zero (open-circuit
port 2). Then from Eq. 4.232 A = V1 /V2 and from Eqs. 4.236, 4.237 Z11 = V1 /I1 and
Z21 = V2 /I1 . From these relations we find that the A element of the ABCD matrix
is related to the impedance matrix elements by A = Z11 /Z21 . If we short-circuit port
2 (V2 = 0) we find B = Z11 (Z22 /Z21 ) − Z12 . Proceeding similarly for C, D we find
C = 1/Z21 and D = Z22 /Z21 .
 Nanostructures and Equivalent Circuits 155

Table 4.5 Equivalence between ABCD and Z-matrix elements in two-port networks

ABCD Matrix Z-Matrix T-Network

Z11 Z1
A Z21 1+ Z3
Z11 Z22 −Z12 Z21 Z1 Z2
B Z21 Z1 + Z2 + Z3
1 1
C Z21 Z3
Z22 Z2
D Z21 1+ Z3

Fig. 4.21 T-Network equivalent circuit showing Z1 , Z2 , Z3 impedances in terms of Z-matrix

elements

The equivalence between the ABCD matrix elements, the impedance matrix ele-
ments, and the impedance-equivalent circuit for the T-network is summarized in Table
4.5. The T-network, implemented with impedance elements, is shown in Fig. 4.21.

4.9 Nanostructures and Equivalent Circuits

4.9.1 Nanosphere Driven by a Harmonic Electric Field
We discuss how the response of nanoscale objects to light in the visible-near IR range
can be considered as an “equivalent circuit”. Our discussion follows the treatment of
N. Engheta and his research group (Ref. 3 listed at the end of the chapter). We begin
by revisiting the sphere subject to an electric field that we considered in Chapter 2,
Section 2.6. Let us consider a homogenous sphere of nanoscale dimension, radius r0
immersed in a harmonically varying E-field, E = E0 e−iωt . We posit the permittivity
of the material as  without specifying, for the present, whether the sphere is dielectric
or conductive. Since the sphere is subwavelength in size, we can use the quasistatic
approximation for the E-fields in the space within and outside the sphere. We have
already obtained these fields in Eqs. 2.120 2.121, but we rewrite them here for conve-
nience. The permittivity of the sphere itself is labeled in and outside the sphere the
permittivity is out .
156 Waveguides and Equivalent Circuits

Fig. 4.22 Polarized sphere subject to an external E-field Ez = E0 e−iωt . E-field lines are
sketched in the vicinity of the sphere, and displacement field (D-field) lines are shown within
the sphere and looping outside the sphere in a dipole pattern. Note that the D-field lines
normal to the surface are continuous through the interface.
 
3out
Ein (r, θ) = E0 êz (4.238)
in + 2out
in − out r03
 
Eout (r, θ) = E0 êz + E0 (2 cos θ êr + sin θ êθ ) (4.239)
in + 2out r3

Figure 4.22 shows the shape and direction of the applied E-field lines (along the z
direction) in the vicinity of the sphere and the displacement field inside the sphere,
Din = in Ein , and outside the sphere, Dout = out Eout . At the surface of the sphere,
matching conditions impose that the normal component of the displacement field be
continuous across the surface.
   
in − out
in Ein · n̂ = out E0 êz · n̂ + E0 (2 cos θêr · n̂ + sin θêθ · n̂)
in + 2out
     
3out in − out
in E0 êz · n̂ = out E0 êz · n̂ + E0 2êz · n̂ (4.240)
in + 2out in + 2out

The applied displacement field, parallel to z, inside the sphere is given by

Dinappl = in Ein = in E0 êz (4.241)

The net or residual displacement field within the sphere is the difference between the
total displacement field (left side of Eq. 4.240) and the applied displacement field.
 
out − in
Dnet = in E0 êz (4.242)
in + 2out
 Nanostructures and Equivalent Circuits 157

The component normal to the sphere surface of this net displacement field is
 
out − in
Dnet · n̂ = in E0 êz · n̂ (4.243)
in + 2out
Now we subtract Eq. 4.243 from both sides of Eq. 4.240 and rearrange the terms to
put the applied fields inside and outside the sphere on the left and everything else on
the right.
   
in − out in − out
(in − out ) E0 êz · n̂ = in E0 êz · n̂ + out E0 2êz · n̂ (4.244)
in + 2out in + 2out
The displacement current is given by the time derivative of the displacement field.
Taking the time derivative of Eq. 4.244 and integrating over the surface of the half-
sphere with positive charge in Fig. 4.22 yields
 
2 in − out
−iω (in − out ) E0 · πr0 = −iωin E0 · πr02
in + 2out
 
in − out
− iωout E0 · 2πr02 (4.245)
in + 2out
The integration over the surface is carried out by noting that the êz · n̂ factor in
Eq. 4.244 is cos θ and that therefore the integration over an infinitesimal of half the
sphere surface dS is
Z Z π/2
2 2
r0 êz · n̂ dS = r0 cos θ sin θ dθ dϕ
0
 π/2
1
= r02 2π sin2 θ = πr02
2 0
The term on the left side of Eq. 4.245 is a bound current source arising from the
polarization of the sphere due to the external applied electric field. The first term
on the right is the bound current passing through the sphere surface and the second
term is the external “dipole current” looping back to the source (Fig. 4.22). The three
terms of Eq. 4.245 express the Kirchhoff current law (Section 4.2.1), that all currents
entering and leaving a node in a circuit must sum to zero.
Dielectric Sphere. We suppose for the moment that in represents an (almost) lossless
dielectric. In that case the permittivity of the sphere is complex but with a negligible
imaginary term and in > 0 . The circuit of the displacement current can then be
represented as in Fig. 4.23. Now from the electric field at the surface of the sphere we
the potential difference across the sphere. The potential is given by
Z
Vsphere = − Esurf · dl
σ
Z 0  
out − in
= −2 E0 cos θ r0 sin θ dθ
−π/2 in + 2out
 
out − in
= r0 E0 (4.246)
in + 2out
158 Waveguides and Equivalent Circuits

Fig. 4.23 Equivalent circuit for a subwavelength dielectric sphere subject to an applied oscil-
lating electric field. The displacement currents into and out of the round node at the branch
obey Kirchhoff’s current rule. The dielectric sphere is represented by capacitor in parallel
with a very low-loss resistor, and the dipole loop displacement current is also represented by
a capacitor.

Now from Vsphere and Isphere we can identify the characteristic impedance of the sphere
as
Vsphere
Zsphere =
Isphere
i
= (4.247)
ωin πr0
and the impedance of the dipolar field is
Vsphere
Zdipole =
Idipole
i
= (4.248)
ωout 2πr0
From the phasor form of the impedance (Section D.2 in Appendix D) we have
Z = R + i (XL + XC ) (4.249)
XL = −ωL inductive reactance (4.250)
1
XC = capacitive reactance (4.251)
ωC
Comparing 4.251 with 4.247 and 4.248 we see that
Csphere = in πr0 (4.252)
Cdipole = out 2πr0 (4.253)
 Nanostructures and Equivalent Circuits 159

Fig. 4.24 Equivalent circuit of a subwavelength metallic sphere subject to an external os-
cillating electric field. The source displacement current drives the sphere and Kirchhoff’s
current rule applies to the branch point node. Current through the sphere exhibits inductive
reactance in parallel with a nonnegligible resistive loss. The dipole branch exhibits capacitive
reactance as in the case of the dielectric sphere.

Metallic Sphere. In the case of a metallic sphere the real part of the permittivity can
be strongly negative and the resistive loss nonnegligible. In that case the equivalent
circuit can be construed as shown in Fig. 4.24. The sphere impedance becomes

metal i
Zsphere = (4.254)
ωRe[in ]πr0

and comparing 4.254 with 4.250 we see that

i
Lmetal
sphere = − (4.255)
ω 2 Re[ in ]πr0

Thus we see that the nanosphere can act as a capacitive element or an inductive
element by choosing the material property: insulating or conducting.

4.9.2 Equivalent Circuit for Plasmon Surface Waves

In this section we reconsider surface waves at the interface between a dielectric and
metal. We apply what we have learned about planar transmission lines, their rela-
tion to plane waves (Section 4.3.4), and the equivalent circuits that can represent a
transmission line. The basic idea is to treat a plane wave incident normally on the
interface as a planar transmission line and the interface itself as a junction between
two transmission lines: one in the dielectric and the other in the metal.

Transmission Line Equivalent Circuit. A plane wave impinging on a dielectric-metal

interface can be modeled as a real transmission line. Figure 4.25 shows in the left
160 Waveguides and Equivalent Circuits

panel the plane wave incident normal to the interface and the right panel the equiv-
alent impedance T-network representing a transmission line. The expressions for the
impedances are (see reference 4 at the end of the chapter),
 
βl
Zs = Zc tanh (4.256)
2
Zc
Zp = (4.257)
sinh(βl)
where
γc
Zc = (4.258)
iωc
and
γc2 = β 2 − c k02 = −kc2 (4.259)
where β is the complex propagation constant,l the length of the transmission line,
Zc the characteristic impedance, and c the characteristic dielectric constant of the
medium. Using these expressions for the series impedance Zs and parallel impedance
Zp , and applying the rules for finding the ABCD matrix elements of the T-circuit we
developed in Section 4.8.5, we find that for long transmission lines (l >> 1/β).

A = cosh βl (4.260)
B = Z0 sinh βl (4.261)
1
C= sinh βl (4.262)
Z0
D = cosh βl (4.263)

in accordance with the expressions listed in Table 4.4. Now as the length of the trans-
mission line tends to infinity, Z → 0; and in the limit of very long lines we have the
situation shown in Fig. 4.26. In fact the plane wave propagates through the dielectric
and the metal, although the propagation length in the metal penetrates only as far as
the skin depth. Therefore we can characterize the light incident on the interface as two
transmission lines connected at the interface with two characteristic impedances, Z0d
and Z0m for the dielectric and metal respectively. Figure 4.27 shows the correspondence
between plane wave propagation in the dielectric and metal and the equivalent circuit
of the interface. Now the transmission line impedances can be written in a particularly
useful form if we consider Maxwell’s equations in a Laplace-transformed space. The
Laplace transform is an integral operator of the form
Z ∞
L̂ = e−γz dz (4.264)
0

and transformation of field A(x, z) is expressed by,

Z ∞
Ã(γ0 ) = L̂A(x, z) = A(x, z)e−γz dz (4.265)
0

with
 Nanostructures and Equivalent Circuits 161

Fig. 4.25 Comparison between plane wave and equivalent impedance t-network. Panel(a)
Plane wave incident on a dielectric-metal interface. (b) Equivalent circuit for a transmission
line representing the plane wave.

γ 2 = −kz2 (4.266)
where γ may be complex to take into account transmission line losses and kz the
propagation parameter in the z direction. Details of this transformation and its con-
sequences are discussed in a recent publication (reference 6 at the end of the chapter).
Here we simply state one of the key results for the Ampère-Maxwell law in Laplace
space for incident waves with TM polarization.
γ H̃y = iωẼx (4.267)
where H̃y and Ẽx are the Laplace-transformed y and x components of the H- and E-
field, respectively, ω and  have their usual meanings of frequency and permittivity. We
identify the characteristic impedance of the transmission line with the corresponding
impedance of the Laplace-transformed wave.
Ẽx γ
Z0 = = (4.268)
H̃y iω
162 Waveguides and Equivalent Circuits

Fig. 4.26 The limiting form of the equivalent circuit of a long transmission line.

Any surface waves traveling in the x direction must have the same propagation pa-
rameter β in the dielectric and metal materials. With k0 the wave propagation vector
in free space we have from k-vector addition on each side of the interface
q
γd = β 2 − d k02 = ikzd (4.269)
q
γm = β 2 − m k02 = ikzm (4.270)

The picture we have now is of two transmission lines along z that join at the dielectric-
metal interface. The characteristic impedances for the dielectric and metal transmis-
sion lines are
γd
Z0d = (4.271)
iωd
γm
Z0m = (4.272)
iωm
Transmission Line Transverse Resonance. When a transmission line is populated by
a wave that is just at the cutoff condition, kc , a standing wave is present in the
line, and no net energy propagates along the line. A transmission line in this state
is said to be in “resonance”. The two transmission lines, along z in our problem, are
transverse to the interface x. In order for stable surface waves to propagate along
the interface, they must be subject to the same restriction that no energy propagates
along z. Therefore a necessary condition for stable, propagating surface waves along x
is that our transmission lines along z fulfill the resonance condition. Since no net wave
propagates along z, the net propagation parameter along z must be null. Therefore,
according to Eqs. 4.271, 4.272, a necessary condition for transverse resonance in our
two joined transmission lines is

Z0d = Z0m for all z (4.273)

 Nanostructures and Equivalent Circuits 163

Fig. 4.27 Correspondence between plane waves propagating in dielectric and metal, panel
(a), and two equivalent circuit transmission lines joined at the dielectric-metal interface.

Therefore
γd γm
= (4.274)
iωd iωm
or
kzd km
= z (4.275)
d m
This “impedance matching” condition at the interface, arising from the transmission
line at resonance, is equivalent to the field matching condition that we found in Section
3.3, Eq. 3.32. Thus we see that a transmission line point of view gives us a supplemen-
tary insight into the physical conditions required for stable surface waves.

4.9.3 Equivalent Circuit of a Dielectric Slit in a Metal Layer

In this section we consider the equivalent circuite of a dielectric slit milled in a metal
layer. We suppose that both the thickness of the metal layer and the slit width are
164 Waveguides and Equivalent Circuits

Fig. 4.28 Schematic of a subwavelength slit of width w milled into a subwavelength metal
layer. The dielectric in the slit is the same as the dielectric above and below the metal surface.
In the usual case the metal is the same on both sides of the slit so Z1 = Z2 . The equivalent
circuits of the two dielectric-metal interfaces and the T-circuit of the slit itself are shown in
the impedance equivalent circuit.

subwavelength. Figure 4.28 shows a schematic of the physical slit and the correspond-
ing impedance equivalent circuit. The equivalent circuit consists of the T-circuit that
we considered earlier in Section 4.9.2 for the slit and two symmetrical voltage sources
and impedances that represent voltage and current runnning in the metal near the
surfaces. We posit that a plane wave propagating along z is incident on the top sur-
face and sets up a standing wave there. We saw in Section 2.10.6 that the magnetic
field component of the wave is adjacent to the surface and penetrates to the skin depth
within a real metal. This time-harmonically oscillating magnetic field, via Faraday’s
law, induces currents and voltages within the skin depth on the top surface and on
the slit walls. A detailed discussion of the currents circulating in the skin depth near
the slit is given in Reference 7 at the end of the chapter.
The first step in the analysis is to assume that the circuit is symmetrical with the
metal wall impedances Z1 = Z2 = Zm . Then we consider the circuit with only one
source, say the one on the left associated with Z1 in Fig. 4.28 and effectively “short-
circuit” the other source on the right. The resulting circuit consists of a single voltage
source and a series-parallel network of impedances that can be combined into a single
impedance ZT . The expression for ZT is
 
(Zm + Zsd )Zpd + (Zm + Zsd )(Zm + Zsd + Zpd )
ZT = (4.276)
Zm + Zsd + Zpd
Now we impose the “resonant transmission line” condition to insure that no power is
propagated perpendicular to the slit walls (along the x direction). As we saw in Section
 Exercises 165

3.6 power is propagated only along z with exponential fall-off of the fields penetrating
the metal walls of the slit. This condition is that ZT = 0, which results in the following
impedance expression
2 2 2
(Zsd + Zpd ) + (2Zm ) (Zsd + Zpd ) + Zm − Zm + Zpd =0 (4.277)

Adapting the definitions Eqs. 4.256 - 4.259, we have

γd W
Zsd = Zd tanh (4.278)
2
Zd
Zpd = (4.279)
sinh γd W
where
γd
Zd = ± (4.280)
iωd
and
γd2 = β 2 − d k02 = −kd2 (4.281)
with W the width of the slit. Substituting these definitions into 4.277 we arrive, after
some algebra, at
2Zd Zm 2 (Zm /Zd )
tanh γd W = ± 2 2
=± 2 (4.282)
Zd + Zm 1 + (Zm /Zd )
This expression is the transcendental equation that specifies the allowed modes in the
slit that propagate in z and are stable against radiation along x. Now recalling the
identity
2 tanh (x/2)
tanh x = 2 (4.283)
1 + [tanh (x/2)]
and comparing with 4.282 we can identify
γd W Zm
tanh =± (4.284)
2 Zd
From the definition of the hyperbolic tangent in terms of exponential functions this
last expression can be further simplified to
 
γd W Z d + Z m
e = ±1 (4.285)
Zd − Zm

This is the final expression for the slit transcendental function in terms of the equivalent
circuit quantities. It is consistent with the transcendental equations found for the slit
in Chapter 3 obtained from a wave treatment.

4.10 Exercises
1. Suppose two lossless transmission lines with different characteristic impedances
are coupled. If Z1 = 50 Ω and Z2 = 75 Ω, calculate the fractonal power reflected
and transmitted at the interface between the two lines.
166 Waveguides and Equivalent Circuits

2. For a tranmission line of length λ/4 calculate the impedance of the line if it is (a)
shorted, (b) open-circuited.
3. For a rectangular waveguide calculate the cutoff wavevector kc for TM modes
if the guide is square with side s = 800 nm and if the guide is very thin with
s1 = 800 nm and s2 = 100 nm. Assume the guide is excited by a plane wave
of λ = 500 nm impinging normally on the front fae of the guide. Calculate the
characteristic impedance of this transmission line.
4. Consider the interface between silver and glass. Assume that 1 mW power is cou-
pled to the interface and converted to spp waves. From what you learned in
Chapter 3 calculate the stable surface wave propagation parameter assuming the
exciting field wavelength is 832 nm. Use the following dielectric constants: glass
0 = 2.40, silver  = −32.8 + i0.46.
5. Using the same data as in Exercise 4 calculate the impedance Zp of the equivalent
T-circuit and the charactersitic impedance Z0 of the corresponding transmission
line.
6. Calculate the TE and TM modes that can propagate in a planar metal-dielectric-
metal (MDM) waveguide with water between the two metal walls. The distance
between the two metal slabs is 0.25 µm and the free-space wavelength is 1.55 µm.
The dielectric constant of water at λ0 = 1.55µm can be considered real and equal
to 1.77.
7. Calculate (a) the capcitance of a glass nanosphere with radius equal to 50 nm and
(b) the capcitive reactance of the sphere at an excitation wavelength of 832 nm.
The dielctric constant of glass is 2.25.
8. Suppose we have a sphere with the same dimensions as in Exercise 7 but fabricated
from silver metal. Using the dielectric constant data of Exercise 4, calculate (a)
the inductance of the sphere and (b) the inductive reactance.

4.11 Further Reading

1. D. M. Pozar, Microwave Engineering,3rd edition, John Wiley & Sons, Hoboken,
NJ, 2005.
2. S. Ramo, J. R. Whinnery, and T. Van Duzer, Fields and Waves in Communication
Electronics, 3rd edition, John Wiley & Sons, New York, 1994.
3. N. Engheta, A. Saladrino, and A. Alù, Circuit elements at optical frequencies:
nano-inductors, nano-capacitors, and nano-resistors Phys. Rev. Lett. 95, 095504
(2005).
4. C. D. Papageorgiou and J. D. Kanellopoulos, Equivalent circuits in Fourier space
for the study of electromagnetic field, J. Phys. A: Math. Gen. 15, 2569-2580 (1982)
5. S. E. Kocabaş, G. Veronis, D. A. B. Miller, and S. Fan, Transmission Line and
Equivalent Circuit Models for Plasmonic Waveguide Components, IEEE J. Sel.
Topics Quant. Elec. 14, 1462-1472 (2008)
6. F. Nunes and J. Weiner, Equivalent Circuits and Nanoplasmonics, IEEE T. Nano,
8, 298-302 (2009).
7. J. Weiner, The electromagnetics of light transmission through subwavelength slits
in metallic films, Opt. Express 19, 16139-16153 (2011).
5
Radiation in Classical and Quantal
Atoms

5.1 Introduction
The hydrogen atom is the simplest atomic structure interacting with light in the optical
regime. We know from the earliest days of atomic spectroscopy that the hydrogen atom
emits and absorbs dipole radiation. But the emitted and radiated light only appears
at certain frequencies; the hydrogen atom does not exhibit a continuous spectrum as
the classical dipole antenna. The reason is that the atom is a quantum object, not a
classical dipole. Nevertheless a “semiclassical” picture of the light-atom interaction is
possible. In this picture the electron, with dynamics defined by the quantal nature of
atomic internal states, still acts as a receiving antenna or a source of classical (usually
dipole) radiation. This hybrid picture of quantal matter and classical electrodynamics
captures a great deal of the atom-optical physics. Nonrelativistic quantum mechanics
and the classical radiation field combine to explain almost all the features of the
hydrogen atom spectrum. We will use these tools to develop an explanation for the
quantization of the energy levels of the atom and the “selection rules” for dipole
transitions between them.

5.2 Dipole Emission of an Atomic Electron

In Section 2.8 we considered a classical electric dipole aligned along the z axis and
wrote the dipole as
p(t) = q0 a cos ωtẑ
If we consider an atom as an electron bound to a massive proton, subject to a radial
restoring force centered at the proton position, then we can write down a similar dipole
expression,
p(t) = −er(t) p0 = −er0 (5.1)
where −e is the electron charge, r is the radial distance from the atom center, and r0 is
the maximum radial distance. The electron is pictured as vibrating radially through the
proton position and in this simple model exhibits no angular momentum from orbital
motion about the charge center. Using the result for power emitted by a classical
dipole, Eq. 2.156 we can write the optical-cycle-averaged power emitted by the atom
as
1 ω04 p20 e2 ω04 r02
hW i = 3
= (5.2)
4π0 3c 4π0 3c3
168 Radiation in Classical and Quantal Atoms

For harmonic oscillation the maximum acceleration is a0 = −ω02 r0 and

e2 a20
hW i = (5.3)
4π0 3c3
Writing the cycle-averaged acceleration as ha · ai = a2 = (1/2)a20 we have

e2 2 a 2
hW i = (5.4)
4π0 3c3
and we see that the average power emitted by the oscillating electron is proportional to
the average square of the electron acceleration. A charged particle must be accelerating
to emit radiation; an electron moving at constant velocity, not subject to external
forces, does not emit.
The energy (cycle-averaged) emitted by the oscillator must come from mechanical
(kinetic + potential) energy, Emech .
1
me ω02 r02
Emech = (5.5)
2
where me is the electron mass. The average power emitted by the oscillator can be
equated to the diminution of mechanical energy with time.
Emech
hW i = − (5.6)
τ
From Eqs. 5.2, 5.5, and 5.6 we see that that
 2 −1
e 3me c3
τ= (5.7)
4π0 2ω02
The “lifetime” of the oscillator τ can be written a little more simply by introducing
the classical radius of the electron,
e2 1
rc = ' 2.8 × 10−15 m (5.8)
4π0 me c2
and
3c
τ= (5.9)
2rc ω02
In differential form Eq. 5.6 is written
dEmech
hW i = −
dt
so
dEmech dt
=− (5.10)
Emech τ
and
Emech (t) = Emech (0)e−t/τ (5.11)
We see that the internal energy of the atom, modeled as a classical electron oscillating
around a positive charge center, decreases exponentially with time as the energy is
radiated away.
 Radiative Damping and Electron Scattering 169

A specific example provides some feeling for the relevant time and energy scales.
Consider a sodium atom, structured with 10 electrons in a “closed shell” and one
electron outside. This exposed or “active” electron is subject to an effective positive
charge at the nucleus roughly equal to that of of one proton. The potential and electric
field experienced by the active electron is radial and the simple dipole oscillator model
might therefore be appropriate. Emission from the first excited state to the ground
state is concentrated in a narrow band of wavelengths centered at 590 nm, the well-
known “D line” of atomic spectroscopy. Considering this emission as emanating from
a dipole oscillator with resonant frequency ω0 ,
2πc
ω0 = = 3.2 × 1015 s−1
λ
and
3c
τ= ' 16 × 10−9 s
2rc ω02
This result is actually quite close to the measured lifetime of the first excited state
of the sodium atom and lends credence to the harmonic oscillator model. In reality,
however, the atom must be treated quantum mechanically because there is no lower
limit to the classical radiative emission, and the classical electron would fall into the
nucleus as its mechanical energy diminishes to zero. Furthermore, the first excited
state with a quantum mechanically “allowed” transition possesses one quantum of
orbital angular momentum, and this angular momentum does not appear in the simple
oscillator model.

5.3 Radiative Damping and Electron Scattering

Nevertheless the damped harmonic oscillator model provides some insight into the
physics of atomic emission and absorption of radiation and of atomic scattering. We
can write down the equation of motion of the classical, radiatively damped, externally
driven harmonic oscillator, moving along the z axis, as

d2 z 1 dz eEz −iωt
+ + ω02 z = − e (5.12)
dt2 τ dt me
where as before ω0 is the characteristic frequency of the oscillator, e, me are the electron
charge and mass, respectively, τ the classical oscillator lifetime, and Ez is the amplitude
of the external driving field oscillating at frequency ω. The solution to this equation
of motion is
−eEz e−iωt
z= (5.13)
me [(ω02 − ω 2 ) − iω/τ ]
From this result we can write down the frequency-dependent dipole moment,

−eEz e−iωt
p = −ez ẑ = ẑ (5.14)
me [(ω02 − ω 2 ) − iω/τ ]

From Eq. 5.2 we have for the average dissipated power

170 Radiation in Classical and Quantal Atoms

e4 Ez2 ω 4
hW i = h
2
i (5.15)
12π0 m2e (ω02 − ω 2 ) + (ω/τ )2

Using the classical electron radius, Eq. 5.8, we can write this expression for the dissi-
pated power as the product of an energy flux and a scattering cross section,
( " #)
0 Ez2 8πrc2 ω4
 
hW i = c· 2 2
(5.16)
2 3 (ω02 − ω 2 ) + (ω/τ )
| {z } | {z }
energy flux cross section

Where the scattering cross section is defined as

" #
8πrc2 ω4

σ(ω) = 2 2
(5.17)
3 (ω02 − ω 2 ) + (ω/τ )

5.4 The Schrödinger equation for the hydrogen atom

The basic expression for the Schrödinger equation of any atomic system is described
by an operator equation,
∂Ψn (r, t)
ĤΨn (r, t) = i~ (5.18)
∂t

where Ψn (r, t) represents the state n of the atom, and Ĥ is the Hamiltonian operator,
Ĥ = i~∂/∂t. According to the postulates of quantum mechanics, a characteristic or
“eigen” state of a system must be represented by a wavefunction that is single-valued,
normalized, and orthogonal to all the other eigenstates of the system. An arbitrary
system state may be expressed as a linear combination of these eigenstates that form
a “complete set” in a linear, orthogonal vector space. If the state is “stationary” or
time-independent, as it is for the isolated atom, we may write the state function as a
product of a spatial term ψ(r) and a time-varying phase, exp(−iωt)

Ψ(r, t) = ψ(r)e−iωt (5.19)

The frequency of the phase is related to the energy En of the stationary state by
the Planck relation, En = ~ω. The spatial, time-independent part can factor and be
separated by substitution of Eq. 5.19 into Eq. 5.18 with the result

Ĥ [ψ(r) − En ] = 0 (5.20)

where ψ(r) is the spatial wave function that is the solution to the time-independent
Schrödinger equation, 5.20. Now the Hamiltonian operator Ĥ is constructed from the
Bohr correspondence principle that permits us to identify certain mathematical op-
erators with their dynamical variable counterparts in the Hamiltonian formulation of
 The Schrödinger equation for the hydrogen atom 171

classical mechanics. The classical Hamiltonian H expresses the energy of a system.

For example,
H =T +V (5.21)
where T is the kinetic energy and V the potential energy. The correspondence principle
is a rule that indicates the equivalence of these classical functions to the operators
representing them in quantum mechanics. For example,
H → Ĥ T → T̂ V → V̂ (5.22)
The kinetic and potential energies themselves are functions of position and momen-
tum which correspond to the operators r̂ and p̂, respectively. The “operator” for the
position is the position itself
r̂ → r (5.23)
and the operator for the momentum is
p̂ → −i~∇ (5.24)
Therefore the operator expression for the kinetic energy of a particle with mass m
becomes
p2 ~2 2
T = → T̂ = − ∇ (5.25)
2m 2m
and if the potential energy is a function only of position, then
V (r) → V̂ (r̂) (5.26)
From Eqs. 5.25, 5.26 we can construct the Hamiltonian operator and the Schrödinger
operator equation for the energy of the system,
~2 2
h i  
Ĥ − En ψ = 0 → − ∇ + V (r) − En ψ = 0 (5.27)
2m
The physically allowed solutions to the time-independent Schrödinger equation, Eq. 5.27
constitute the “states” of the physical system described by the energy terms of the
Hamiltonian. In the case of a hydrogen atom, the system is quite simple, consisting
of a single proton and an single electron bound together by the electrostatic Coulomb
potential. Strictly speaking this situation is a two-particle problem, and the kinetic
energy term in Eq. 5.27 should be the sum of two terms for the kinetic energies of the
proton and electron. However, the rest mass of the proton is greater than that of the
electron by a factor of nearly 2000, and it is convenient to consider the “reduced mass”
µ of the equivalent single-particle problem.
me mp
µ= ' me (5.28)
me + mp
and the center-of-mass (cm) coordinate,
mp rp + me re
rcm = ' rp (5.29)
mp + me
So to a very good approximation we can consider a model system where the reduced
mass can be considered essentially that of the electron and the center-of-mass co-
ordinate that of the proton. By placing the origin of the coordinate system at the
172 Radiation in Classical and Quantal Atoms

center-of-mass position we simplify the problem to a single electron moving about a

stationary positive charge centered at the origin. For all practical purposes therefore
we can write the time-independent Schrödinger equation for the hydrogen atom as
~2 2 1 e2
− ∇ ψ(x, y, z) − ψ(x, y, z) − En ψ(x, y, z) = 0 (5.30)
2me 4π0 (x2 + y 2 + z 2 )1/2
where the Laplacian operator ∇2 in Cartesian coordinates is
∂2 ∂2 ∂2
∇2 = + + (5.31)
∂x2 ∂y 2 ∂z 2
and the Coulomb potential is
1 e2
V =− (5.32)
4π0 (x2 + y 2 + z 2 )1/2
Since the Coulomb potential term exhibits spherical symmetry, Eq. 5.30 can be solved
far more easily in spherical coordinates, r, θ, φ where
x = r sin θ cos ϕ (5.33)
y = r sin θ sin ϕ (5.34)
z = r cos θ (5.35)
2 1/2
r = x2 + y 2 + z


(5.36)
and the Laplacian operator ∇2 is expressed as
∂2
   
1 ∂ ∂ 1 1 ∂ ∂
∇2 = 2 r2 + 2 2 + sin θ (5.37)
r ∂r ∂r r sin θ ∂ϕ2 r2 sin θ ∂θ ∂θ
The Schrödinger equation in spherical coordinates takes the form
~2 2 1 e2
− ∇ ψ(r, θ, ϕ) − ψ(r, θ, ϕ) − En ψ(r, θ, ϕ) = 0 (5.38)
2me 4π0 r
The relation between Cartesian and spherical coordinates and the representation of the
Laplacian operator in spherical coordinates is discussed in Appendix C. In spherical
coordinates the wave function solution to Eq 5.30 factors into three functions of each
coordinate,
ψ(r, θ, ϕ) = R(r)P (θ)Φ(ϕ) (5.39)
Substituting this form back into Eq. 5.38 results in the separation into three differential
equations in r, θ, ϕ, respectively.
 
1 d 2 dRnl 2µ Rnl
2
r + 2 [E − V (r)] Rnl = l(l + 1) 2 (5.40)
r dr dr ~ r
 
1 d dPl Pl l(l + 1)
− sin θ + 2 = Pl (5.41)
sin θ dθ dθ sin θ m2l
d2 Φ
= −m2l Φ (5.42)
dϕ2
2
The expressions l(l + 1) and − (ml ) are constants of separation.
 The Schrödinger equation for the hydrogen atom 173

5.4.1 Radial Equation

The independent variable of the radial equation 5.40 is r, the distance of the electron
from the origin; and the physically admissible solutions are labeled by the quantum
number integers n, l. By “physically admissible” we mean those solutions such that
Rnl → 0 as r → ∞ and Rnl remains finite as r → 0. There are in fact a family of
admissible solutions each member of which is identified by a unique n, l label. The
n quantum number is called the principal quantum number and is directly identified
with the total energy of the atom.

µe4 me e4
En = 2 '− 2 (5.43)
(4π0 ) 2~2 n2 (4π0 ) 2~2 n2
where n can take on the integer values n = l, l + 1, l + 2, . . . and l can be an integer
l = 0, 1, 2, . . .. The radial solutions take the form of three factors: a normalization
factor, a polynomial, and an exponential. In fact Eq. 5.40 can be rearranged into a
form that bears a striking resemblance to the differential equation whose solutions are
the “modified Laguerre functions”. The rearranged form is obtained by multiplying
Eq. 5.40 by −2µ/~2 and transposing two terms.

~2 1 d e2 ~2 l(l + 1)
 
2 dRnl
− 2
r − Rnl + Rnl = En Rnl (5.44)
2µ r dr dr 4π0 r 2µ r2
The differential equation for the modified Laguerre functions and its solutions, Eqs. E.18, E.19,
are discussed in Appendix E. Now if we make the following substitutions, Eq. 5.44 can
be cast into the form of Eq. E.19.
r
8µE
ρ = αr and α= − 2 (5.45)
~
where it is understood that E < 0 since we only consider the allowed bound-state
energies of the hydrogen atom, and the reference zero of energy is at the ionization
limit r → ∞. The quantity α has units of m−1 so ρ is unitless. We make an additional
substitution to consolidate many constants,

µe2
λ= (5.46)
2π0 α~2
In order for the λ substitution to be consistent with the physically admissible values
of the energy En , λ must be restricted to integers, λ = 1, 2, 3, . . .. and defining χ(ρ) =
R(ρ/α), we can recast Eq. 5.44 as,
   
1 d 2 dχ(ρ) λ 1 l(l + 1)
ρ + − − χ(ρ) = 0 (5.47)
ρ2 dρ dρ ρ 4 ρ2
Using one last substitution trick,

1 d 2 dχ 1 d2
ρ = (ρχ) (5.48)
ρ2 dρ dρ ρ dρ2
174 Radiation in Classical and Quantal Atoms

we have finally
d2 (ρχ)
 
1 λ l(l + 1)
+ − + − ρχ = 0 (5.49)
dρ2 4 ρ ρ2
Now we compare this last result to Eq. E.19 rewritten here for convenience,

d2 Φkn (x) 1 2n + k + 1 k 2 − 1
 
+ − + − Φkn (x) = 0
dx2 4 2x 4x2

We see that Eq. 5.49 become equivalent to Eq. E.19 if

k−1
ρ↔x l↔ λ→n+l+1 (5.50)
2
Then the solutions to Eq. 5.49 are

ρχ(ρ) = e−ρ/2 ρl+1 L2l+1

λ−l−1 (ρ) (5.51)

equivalent to the solutions for Eq. E.19

Φkn (x) = e−x/2 x(k+1)/2 Lkn (x) (5.52)

where Lkn (x) are the associated Laguerre polynomials. Thus Eq. 5.51 expresses the
solutions to Eq. 5.49 that in turn is just the the hydrogen radial equation, Eq. 5.40,
in a different guise. Now we recover from the scaling parameters α and ρ terms of
physically meaningful quantities. From the expression for the allowed bound energies
En of the hydrogen atom, Eq. 5.43, we can write α and ρ in terms of n the principal
quantum number.
2 2
α= and ρ= r (5.53)
na0 na0
where a0 = 4π0 ~2 /me e2 , the well-known atomic radius of the Bohr model. Finally
we can write out the solutions to the hydrogen radial equation, Eq. 5.40,
−αr/2
Rnl (r) = (αr)l L2l+1
n−l−1 (αr)e (5.54)

In quantum mechanics the solutions to the Schrödinger equation play an analogous

role to force fields in the Maxwell equations. Although the wave function is not a
vector function, it is in general complex. Just as in electrodynamics the force field
F itself is not observed but the intensity, F F ∗ , guaranteed to be a real quantity, is
the detected signal, so in quantum mechanics it is the probability density, ψψ ∗ that
connects theory to experiment. Note that EE ∗ and HH ∗ are directly proportional
to the energy density of an electromagnetic field. It is therefore plausible to assign
ψψ ∗ the role of probability density. When ψ is a function only of space and is the
solution to a single-particle Schrödinger equation, then ψ(r)ψ ∗ (r) is interpreted as
the probability density of finding the particle at position r. Evidently then,
Z
ψ(r)ψ ∗ (r) dr = 1 (5.55)
all space
 The Schrödinger equation for the hydrogen atom 175

Fig. 5.1 Normalized radial probability density as a function of the radial coordinate (in
units of a0 ) for three radial wave functions.

and when this condition is fulfilled; the wavefunction is said to be “normalized”. In

order to be consistent with this probability density interpretation,
p the functions Rnl (r)
must be prefixed by a constant normalizing factor NRnl such that
Z ∞
1 ∗
Rnl (r)Rnl (r)r2 dr = 1 (5.56)
NRnl 0
It can be shown from the normalization condition of the associated Laguerre functions,
Eq. E.16, that
 na 3 2n(n + l)!
0
NRnl = (5.57)
2 (n − l − 1)!
so that finally the normalized radial solutions to the hydrogen atom Schrödinger are
" 3 #1/2  l
2 (n − l − 1)! 2 −(2/na0 )r/2
Rnl (r) = r L2l+1
n−l−1 e (5.58)
na0 2n(n + l)! na0

Figure 5.1 shows the normalized radial probability density functions

∗
P (r) = Rnl Rnl r2 (5.59)
for the hydrogen atom ground state n = 1, l = 0 and two excited states. Table 5.1 lists
the radial functions, Rnl used to plot the probability densities of Fig. 5.1.

5.4.2 Angular Equations

Functions of the azimuthal angle ϕ. The solutions to Eq. 5.41, are a family of func-
tions of the polar angle θ and are usually denoted Pl (cos θ) with cos θ as the inde-
pendent variable. This family is called the associated Legendre functions, and their
176 Radiation in Classical and Quantal Atoms

Table 5.1 Hydrogen radial wave functions Rnl , n=1 − 3;l=0

n l Rnl
3/2

1 0 1
a0 2e−r/a0
 3/2 h i
1 r −r/2a0
2 0 2a0 2 − a0 e
 3/2     2 
3 0 1
3a0 2 1 − 2 a0 + 27 ar0
r 2
e−r/3a0

properties are discussed in Complement F. The solutions to Eq. 5.42, the functions of
the azimuthal angle ϕ, can be obtained by inspection. It is easy to verify that
Φml (ϕ) = eiml ϕ (5.60)
According to the postulates of quantum mechanics, the solution to a differential equa-
tion must be single-valued in the independent variable in order for the function to
suitably represent a physical state. Therefore, as ϕ cycles through integer multiples of
2π, increasing from 0 → 2π, 2π → 4π, . . . the value of the function Φml should remain
single-valued. Thus Φml (ϕ) must have the same value as Φml (ϕ0 ) if ϕ0 = ϕ+n2π where
n = 1, 2, 3, . . .. In order to assure this behavior, the value of ml must be an integer,
ml = ±1, ±2, ±3, . . .. Furthermore, Φml is subject to a normalization requirement
analogous to that of Eq. 5.56.
1 2π ∗
Z
Φml (ϕ)Φml (ϕ) dϕ = 1 (5.61)
N 0
The normalized azimuthal wavefunction is therefore
1
Φml (ϕ) = √ ei(±ml )ϕ (5.62)
2π
Functions of the polar angle θ . The solution to Eq. 5.41 constitutes a family of
functions, Plml (cos θ) called the associated Legendre functions. The properties of the
Legendre functions and the Legendre polynomials, the relevant functions when ml = 0,
are discussed in some detail in Complement F. These functions describe the polar-angle
dependence of the hydrogen atom wave function, ψ(r, θ, ϕ) = Rnl Plml (cos θ)Φml (ϕ).
Just as Rnl and Φml must be normalized, so must the functions Plml (cos θ). The
normalization requirement specifies that
1 π ∗ml
Z
P (cos θ)Plml (cos θ) sin θ dθ = 1 (5.63)
N 0 l
with N expressed by
2 (l + ml )!
N= · (5.64)
2l + 1 (l − ml )!
The family of functions that are the product of Pl (cos θ) and Φ(ϕ) is called the spherical
harmonics, Ylml (θ, ϕ). The normalized spherical harmonics are
 The Schrödinger equation for the hydrogen atom 177

ml
Table 5.2 Angular Functions, Spherical Harmonics Yl

n l ml Ylml
q
1 1
1 0 0 2 π
q
2 1 −1 1
2
3
2π sin θe−ϕ
q
1 3
2 1 0 2 2π cos θ
q
2 1 1 − 21 3
2π sin θeϕ
q
3 2 −2 1
4
15
2π sin2 θe−2ϕ
q
3 2 −1 1
2
15
2π sin θ cos θe−ϕ
q
1 5 2


3 2 0 4 π cos θ − 1
q
3 2 1 − 12 15
2π sin θ cos θeiϕ
q
3 2 2 1
4
15
2π sin2 θei2ϕ

s
2l + 1 (l − ml )! ml
Ylml (θ, ϕ) = (−1)ml · P (cos θ)eiml ϕ (5.65)
4π (l + ml )! l

and Z π Z 2π
Yl∗ml (θ, ϕ)Ylml (θ, ϕ) sin θ dθ dϕ = 1 (5.66)
0 0
In order to maintain physically admissible, well-behaved functions there are certain
restrictions on the function labels, n, l, ml . We summarize them here without proof.

n = 1, 2, 3, . . . (5.67)
l = 0, 1, 2, . . . n − 1 (5.68)
ml = −l, −l + 1, −l + 2, . . . 0 . . . , l − 2, l − 1, l (5.69)

Table 5.2 lists the spherical harmonic functions in spherical coordinates up to n = 3.

Putting the spherical harmonic angular wavefunctions together with the radial
wavefunctions gives us finally the full three-dimensional wavefunctions that specify
that quantum state of the hydrogen atom

ψ(r, θ, ϕ)nlml = Rnl (r)Ylml (θ, ϕ) (5.70)

∗
The expression ψnlml
(r, θϕ)ψnlml (r, θ, ϕ) is the probability density of finding the elec-
tron at r. . .

r = rr̂ polar coordinates (5.71)

r = xx̂ + y ŷ + z ẑ Cartesian coordinates (5.72)
178 Radiation in Classical and Quantal Atoms

ml
Table 5.3 Hydrogen wave functions Rnl Yl , n=1 − 3

n l ml Rnl Ylml
 3/2
1 0 0 √1
π
1
a0 e−r/a0 1
 3/2 h  i
2 0 0 √1
4 2π
1
a0 2− r
a0 e−r/2a0 1
 3/2  
2 1 0 √1
4 2π
1
a0
r
a0 e−r/2a0 cos θ
 3/2  
2 1 ±1 8 π
1
√ 1
a0
r
a0 e−r/2a0 sin θe±iϕ
 3/2     2 
3 0 0 1
√
81 3π
1
a0 27 − 18 ar0 + 2 ar0 e−r/3a0 1
q  3/2   
1 2 1 −r/3a0
3 1 0 81 π a0 6 − ar0 r
a0 e cos θ
 3/2   
−r/3a0
3 1 ±1 1
√
81 π a0
1
6 − r
a0
r
a0 e sin θe±iϕ
 3/2  2 2
3 2 0 √1
81 6π a0
1 r
a0 e−r/3a0 3 cos2 θ − 1
 3/2  2 2
3 2 ±1 1
√
81 π a0
1 r
a0 e−r/3a0 sin θ cos θe±iϕ
 3/2  2 2
3 2 ±2 1√
162 π a0
1 r
a0 e−r/3a0 sin2 θe±i2ϕ

The probability dP of finding the electron in a volume interval between V and V + dV

is
∗
dP = ψnlml
(r, θϕ)ψnlml (r, θ, ϕ) dV (5.73)
∗
dP = ψnlml
(r, θϕ)ψnlml (r, θ, ϕ) r2 sin θ dθ dϕ dr (5.74)
and Z Z ∞ Z π Z 2π
∗
dP = ψnlml
ψnlml r2 sin θ dθ dϕ dr = 1 (5.75)
0 0 0
all space

Table 5.3 lists the normalized ψnlml solutions to the hydrogen atom Schrödinger equa-
tion for n = 1 − 3.
5.4.3 Orthogonality
If ψn,l,ml represents an eigenstate of the hydrogen atom with an eigen energy and
angular momentum, then according to the postulates of quantum mechanics it must
be not only single-valued and normalized, but also orthogonal to all other states with
different energies or angular momenta. More succinctly,
Z
∗
ψn,l,ml
ψn0 ,l0 ,m0l dτ = δnn0 δll0 δml m0l (5.76)
all space
 State Energy and Angular Momentum 179

where the δ function takes the value of unity when n, l, ml are equal to their primed
counterparts but is null otherwise. Wavefunctions that exhibit this property are said
to be “orthonormal” because they are orthogonal and normalized. In polar coordinates
the wavefunctions of the hydrogen atom are products of radial and angular functions

ψn,l,ml (r, θ, ϕ) = Rnl (r)Ylml (θ, ϕ) = Rnl (r)Plml (cos θ)Φml (ϕ) (5.77)

and therefore these radial and angular functions must be orthonormal as well.

5.5 State Energy and Angular Momentum

Now that we have the functional form of the radial and angular solutions to the
Schrödinger equation for the hydrogen atom, we can examine their properties in more
detail.

5.5.1 Eigen energies

Considering the time-independent Schrödinger equation as an operator equation, Eq. 5.20,
we see that an eigen state, when operated upon by the energy operator Ĥ, returns the
energy eigen value of that state,

Ĥψn,l,ml = En ψn,l,ml (5.78)

The eigen energies En are labeled by the principal quantum number n and are charac-
terized by the analytical expression given in Eq. 5.43. The constants in this expression
can be collected together to facilitate numerical calculation in convenient units.

2.18 × 10−18 13.6 1.097 × 105

En = − J=− eV = − cm−1 (5.79)
n2 n2 n2

The allowed values of n are n = 1, 2, 3, . . .. The lowest lying (ground) state of the
atom corresponds to n = 1, and E1 = −13.6 eV. The first excited state corresponds to
n = 2, and E2 = −3.40 eV, and so forth. As n → ∞, En → 0, and the zero-reference
energy corresponds to the ionization limit where the electron is no longer bound by
the Coulomb potential. Figure 5.2 shows a diagram of the hydrogen atom energy levels
bound within the Coulomb potential.

5.5.2 Angular Momentum

In addition to energy, the states of the hydrogen atom may exhibit orbital angular
momentum. In the classical Bohr model, the electron rotated around the proton in
various orbital configurations reminiscent of the planets around a star. The uncertainty
principle no longer permits us to make such a literal interpretation, but the Schrödinger
equation does show us how to calculate the orbital angular momentum of the hydrogen
atom states. We can obtain the quantum mechanical operator corresponding to the
classical angular momentum by appeal to the “correspondence principle” discussed
180 Radiation in Classical and Quantal Atoms

Fig. 5.2 Coulomb potential to which the bound electron in the hydrogen atom is subject.
The lowest four bound states (n = 1 − 4) are shown.

in Section 5.4. From classical mechanics we have for the angular momentum of the
electron L from the position r and the linear momentum p

L=r×p (5.80)

and the three Cartesian components

Lx = ypz − zpy (5.81)

Ly = zpx − xpz (5.82)
Lz = xpy − ypx (5.83)

Then from the correspondence principle,

r → r̂ (5.84)
p → −i~∇ (5.85)

The operator equivalents of three angular momentum components are therefore

 State Energy and Angular Momentum 181

 
∂ ∂
L̂x = −i~ y −z (5.86)
∂z ∂y
 
∂ ∂
L̂y = −i~ z −x (5.87)
∂x ∂z
 
∂ ∂
L̂z = −i~ x −y (5.88)
∂y ∂x
and the operator corresponding to the square of the total orbital angular momentum
is  
Lˆ2 = Lˆ2x + Lˆ2y + Lˆ2z (5.89)

In polar coordinates, as discussed in Complement C,we have

   
ˆ ∂ 1 ∂
Lx = i~ sin ϕ + cos ϕ (5.90)
∂θ tan θ ∂ϕ
   
ˆ ∂ 1 ∂
Ly = i~ − cos ϕ + sin ϕ (5.91)
∂θ tan θ ∂ϕ
∂
Lˆz = −i~ (5.92)
∂ϕ
and Eq. 5.89 transforms to

1 ∂2
   
1 ∂ ∂
Lˆ2 = −~2 sin θ + (5.93)
sin θ ∂θ ∂θ sin2 θ ∂ϕ2
From Eqs. 5.41, 5.42 and the definition of the spherical harmonic functions, Eq. 5.65
we see that

Lˆ2 Pl0 (θ) = ~2 l(l + 1)Pl0 (θ) (5.94)

Lˆz Yl,ml (θ, ϕ) = ~ml Yl,ml (θ, ϕ) (5.95)

The value of the orbital angular momentum for a given eigen state of the hydrogen
atom is therefore p
L = ~ l(l + 1) (5.96)
and the Lz component is given by

Lz = ~ml (5.97)

Thus we see that the ground state exhibits no orbital angular momentum, with n = 1,
l = 0 and L, ml = 0. The first excited state, n = 2, can have values of l = 0, 1. Therefore
√
the n = 2 level exhibits two states of orbital angular momentum, L = 0 and L = ~ 2.
The level with l = 1 possesses three possible values of Lz : ~, 0, −~. It is already clear
from these results that Schrödinger equation yields a physical interpretation of the
internal atomic structure quite different from that of the Bohr orbits. First, since the
ground state has no angular momentum, we cannot imagine the electron as orbiting
the nucleus. The electron is restricted to radial motion. Second, even when the orbital
182 Radiation in Classical and Quantal Atoms

Table 5.4 Real orbitals from linear combinations of spherical harmonics

n l real orbital linear combination Cartesian

coordinates
q q
1
Y1−1 − Y11 3 x


2 1 px 2 4π · r
q q
i 12 Y1−1 + Y11 3 y


2 1 py 4π · r
q
3 z
2 1 pz Y10 4π · r
q q 2 2
15 x −y
1
Y2−2 + Y22 1


3 2 dx2 −y2 2 4 π · r2
q q
i 12 Y2−2 − Y22 1 15 xy


3 2 dxy 2 π · r2
q q
1 −1 1 15 xz


3 2 dxz 2 Y2 − Y21 2 π · r2
q q
i 12 Y2−1 + Y21 1 15 yz


3 2 dyz 2 π · r2
q 2 2 2
1 5 2z −x −y
3 2 dz 2 Y20 4 π · r2

state does have nonzero angular momentum, the projection L on the z axis can only
take on discrete values. In the case of l = 1 only three Lz values are allowed. In a
classical orbit, there would be no restriction on Lz from −L to L. A third nonclassical
feature of the angular momentum is the peculiar status of the two other components,
Lx , Ly . A consequence of the uncertainty principle is that the position and momentum
of a particle cannot be known (measured) simultaneously. If all three components of
the angular momentum were measurable and the magnitude of L was known as well,
then the position and momentum of the electron in the atom would be determined
simultaneously in violation of this fundamental postulate of the quantum theory.

5.6 Real Orbitals

It is sometimes useful to consider linear combinations of the spherical harmonics, which
of course are themselves solutions to the angular part of the Schrödinger equation. In
particular linear combinations can be chosen such that resulting solutions are real.
These real one-electron functions are termed “orbitals” and they play an important
role in molecular structure. These real orbitals give spatial directionality to the angular
wavefunctions and can help to interpret the geometry of chemical bonding when atoms
combine to form molecules. Linear combinations of l = 1 solutions to the Schrödinger
equation are called p-orbitals. They exhibit elecron probability density localized along
the three orthogonal x, y, z coordinate axes, as shown in the first row of Fig. 5.3.
Combinations of the five l = 2 solutions result in d-orbitals. The electron probability
density of three of these orbitals are indicated in the second row of Fig. 5.3. Table 5.4
lists the linear combinations of spherical harmonics for l = 1 and l = 2. The labels for
the real orbitals arise from the functional form of these combinations when expressed
in Cartesian coordinates as listed in the right-most column. Higly symmetric molecu-
 Interaction of Light with the Hydrogen Atom 183

lar structures can be explained on the basis of the linear combination of orbitals. For
example the methane CH4 exhibits a spatial structure with the carbon atom at the
center and the four hydrogen atom at the vertices of an equilateral tetrahedron. This
structure is not explained by the p-orbitals of the central carbon atom alone but is
due to a further orbital mixing. The l = 0 s-orbital combines with the three p-orbitals
to produce the tetrahedral structure. This further mixing of s- and p-orbitals is called
hybridization because it involves linear combinations of solutions to the Schrödinger
equation with different angular momenta. As such these “hybridized” orbitals are not
pure quantum angular momentum states, but they nevertheless appear in nature be-
cause the hybridization lowers the energy of the molecule. Highly symmetric molecules
like methane CH4 or sulfur hexafluoride SF6 do not exhibit dipole absorption or emis-
sion between rotational or vibrational states because they have no permanent dipole
moment. However, a simple substitution of a chlorine atom for one of the hydrogen
atoms in methane, for example, yields chloromathane CH3 Cl which does have a per-
manent dipole moment. Rotation and vibration of the molecular structure results in
oscillation of the dipole which in turn permits rotation-vibration transitions through
a dipolar interaction with incident light. We will examine in more detail this type
of molecular spectra for simple diatomic molecules when we discuss the rigid-rotor,
harmonic oscillator model of molecular spectra in Section 5.9.

5.7 Interaction of Light with the Hydrogen Atom

5.7.1 Semiclassical Absorption and Emission
We will first consider a two-level system is order to keep the notation simple and the
ideas clear. Once we have a expressions for the rates of absorption and emission we
will apply the result to the transistions between the hydrogen atom ground state and
the first excited state.
We start by returning to the time-dependent Schrödinger equation, Eq. 5.18, rewrit-
ten here for convenience.
∂Ψn (r, t)
ĤΨn (r, t) = i~ (5.98)
∂t
The Hamiltonian operator now consists of two terms: the first term is the Hamiltonian
of the atom itself which we have developed earlier in the chapter, Ĥatom . To this atomic
Hamiltonian we add a time-dependent term, V̂ (t), which will be closely related to a
driving electromagnetic field, with frequency ω0 . The total Hamiltonian is the sum of
these two terms. Now we write the stationary state n of the atom as

Ψn (r, t) = e−iEn t/~ ψn (r) = e−iωn t ψn (r) (5.99)

where we have used

En = ~ωn (5.100)

the Planck relation between quantized energy and frequency (see Complement D at the
end of this chapter). For the case of the hydrogen atom, the label n can be considered
184 Radiation in Classical and Quantal Atoms

Fig. 5.3 Plots of the p-orbitals and three of the d-oribitals showing their spatial directions
and orientations.

the principal quantum number n discussed in Section 5.4.1. The time-independent

Schrödinger equation becomes
Ĥatom ψn (r) − En ψn (r) = 0 (5.101)
For the moment we consider the hydrogen atom as a two-level system and write
Ĥatom ψ1 = E1 ψ1 = ~ω1 ψ1
Ĥatom ψ2 = E2 ψ1 = ~ω1 ψ2

with
~ω0 = ~ (ω2 − ω1 )) = E2 − E1
We now add the V̂ (t) term, the interaction energy of the atom with the driving field
with frequency close to ω0 .
 Interaction of Light with the Hydrogen Atom 185

Ĥ = Ĥatom + V̂ (t) (5.102)

With the field turned on the state of the system is no longer stationary and becomes
a coupled, time-dependent linear combination of the two stationary states.

Ψ(r, t) = C1 (t)ψ1 e−iω1 t + C2 (t)ψ2 e−iω2 t (5.103)

with C1 , C2 the coupling coeffieicients. We require Ψ(r, t) to be normalized.

Z
2 2 2
|Ψ(r, t)| dV = |C1 (t)| + |C2 (t)| = 1

Now, if we substitute the time-dependent solution, Eq. 5.103, back into the time-
dependent Schrödinger equation, Eq. 5.98, multiply from the left with the complex
conjugate, ψ1∗ eiωt , and integrate over all space, we obtain
Z Z
dC1
C1 ψ1∗ V̂ ψ1 dr + C2 e−iω0 t ψ1∗ V̂ ψ2 dr = i~ (5.104)
dt

To simplify notation we write the integral “matrix elements” ψ1∗ V̂ ψ1 dr and ψ1∗ V̂ ψ2 dr
R R

as V11 and V12 . so we have

dC1
C1 V11 + C2 e−iω0 t V12 = i~ (5.105)
dt
and similarly for C2 we get
dC2
C1 eiω0 t V21 + C2 V22 = i~ (5.106)
dt
These two equations define the time evolution of the atom-plus-field system as the atom
undergoes a stimulated change between states 1 and 2. We need to solve this pair of
equations for C1 (t) and C2 (t). The squares of the absolute values of these coefficients
can be intperpreted as the probability of finding the atom in either stationary state 1
2 2
or 2, as a function of time. Therefore we are really more interested in |C1 (t)| , |C2 (t)|
than the coefficients themselves.

Atom-field coupling operator. We now turn our attention to the details of V̂ (t). This
term is an interaction energy operator, and from the classical theory of electrostatics
we know that the leading interaction V between an electric field and neutral matter
is the dipole term.
V = −p · E (5.107)
where p is the dipole moment and E is the electric field of, say, a propagating elec-
tromagnetic wave through a polarizable material. The classical dipole itself is defined
as X X
p= pi = qi ri (5.108)
i i
where qi is the charge at position ri with respect to the origin of the coordinate sysem
(see Section ??). In the case of a classical atom with one electron orbiting a massive
proton, the instantaneous classical dipole can be taken as
186 Radiation in Classical and Quantal Atoms

p = −er (5.109)
where e is the charge on the electron and r is the radial position of the electron with
respect to the origin, centered at the proton. If an external E-field is now applied to
the classical atom, the interaction energy is given by
V = −p · E = e r · E (5.110)
Passing to the quantal atom, we use the correspondence principle to write the r and
p operators.
r → r̂ (5.111)
p → p̂ (5.112)
but we do not quantize the E-field. The rate of atomic absorption and emission we will
obtain is based on this semiclassical approximation in which the atom is quantized
but the field remains classical. We write the dipole interaction operator as
V̂ = −p̂(r) · E (5.113)
= e r̂ · E (5.114)

where r̂ = r is the electron radial coordinate centered at the proton. The operator V̂
exhibits odd parity with respect to this electron coordinate so that the matrix elements
V11 and V22 in Eqs. 5.105, 5.106 will vanish. Only atomic states of opposite parity can
be coupled by the dipole interaction. For the E-field we write the real form
E = êE0 cos(ωt − kz)
were ê is the polarization unit vector, E0 the field amplitude, ω the frquency, and k the
propagation parameter. We have the chosen the field to be propagating along z. The
interaction takes place over the spatial extent of the atom, ∼ a0 , which is only about
0.5 nm. For electromagnetic fields ranging in wavelength from the near ultraviolet to
the naear infrared, the wavelength is hundreds of nanometers. Therefore, the kz term in
the field argument is negligible and the E-field can be considered effectively constant
over atomic dimensions. The absolute interaction matrix element can therefore be
written
|V12 (t)| = |V12 | cos ωt
Z 
∗
= ψ1 (r)erψ2 (r) dV E0 cos ωt

= p12 E0 cos ωt (5.115)

Since the dipole-field interaction is an energy, we can also write it as
|V12 (t)| = ~Ω0 cos ωt
with the Rabi frequency given by
p12 E0
Ω0 = (5.116)
~
 Interaction of Light with the Hydrogen Atom 187

5.7.2 Selection rules for hydrogen atom transitions

In classical electromagnetic theory, an oscillating charged dipole can radiate a contin-
uous spectrum. We found in Chapter 2, Eq. 2.156 that the power emitted by a classical
dipole varies as the fourth power of the oscillation frequency ω, on which there is no re-
striction. In contrast, the absorption and emission of radiation by atoms only occurs at
certain well-defined frequencies, giving rise to the characteristic atomic line spectrum.
The rules that govern these allowed transition frequencies are called “selection” rules.
The quantization of the emission and absorpion of radiation is one of the principal
differences between classical and quantal light-matter interaction. We can understand
the origin of the selection rules by studying the quantal dipole transistion moment
p12 .
Z
∗
p12 = ψnlm l
(r, θ, ϕ)erψn0 l0 m0l (r, θ, ϕ)r2 sin θ dθ dϕ dr (5.117)

where the primes on the quantum number labels indicate a final state different from
the intial state. This dipole matrix element can be simplified by factoring the wave-
functions into their r, θ, ϕ constitutents and resolving the vector dipole moment into
its x, y, z components.
Z
∗
p12 = e rRnl (r)Rn0 l0 r2 drPl∗ (cos θ)Pl0 (cos θ)e−iml ϕ eiml0 ϕ sin θ dθ dϕ (5.118)

pz = r cos θ (5.119)
px = r sin θ cos ϕ (5.120)
py = r sin θ sin ϕ (5.121)

First, considering pz , we see that this component is independent of ϕ. Therefore the

dipole transition matrix element will have a factor
Z 2π
pz ∼ ei(ml −ml0 )ϕ dϕ (5.122)
0

Since the integral over the period of any oscillatory exponential function with imagi-
nary argument is zero, the only way this factor can be nonzero is if the argument itself
is zero. Therefore we have the first seclection rule

ml − ml0 = ∆ml = 0 (5.123)

Next, still considering pz , we see that that we will have another factor in transition
moment integral
Z π
pz ∼ Pl∗ (cos θ)Pl0 (cos θ) cos θ sin θ dθ (5.124)
0

Now, as we show in Apppendix F, the Legendre functions Pl (cos θ) have definite parity.
Functions labeled with even l are unchanged (even) with respect to the parity operation
and functions labeled with odd l exhibit odd parity (change sign). The parity operation
188 Radiation in Classical and Quantal Atoms

is sign reversal of the function variable. In the case of the Legendre functions the
variable is cos θ. The parity operation is

cos θ → cos(π − θ) = − cos θ (5.125)

and the parity property of the Legendre functions is that

Pl [cos(π − θ)] → Pl (cos θ) l = 0, 2, 4, . . .

pl [cos(π − θ)] → −Pl (cos θ) l = 1, 3, 5, . . .

The integrand of the integral in Eq. 5.124 must be even with respect to θ over the
limits of integration. Since the factor cos θ sin θ exhibits odd parity, the product Pl Pl0
must also be odd so as to render the overall integrand even. Therefore we see that
∆l = 0 is a “forbidden” transition and ∆l = ±1 is allowed, or at least not forbidden,
by the integrand symmetry. Turning now to px , py , we will have factors
Z 2π
px ∼ e−iml ϕ cos ϕeiml0 ϕ dϕ (5.126)
0
Z 2π
py ∼ e−iml ϕ sin ϕeiml0 ϕ dϕ (5.127)
0

and since cos ϕ and sin ϕ are linear combinations of eiϕ and e−iϕ , it is clear that in
order to avoid a null integral another selection rule for these two components of the
transition moment is
∆ml = ±1 (5.128)
More generally we have to consider the overall parity operation that transforms the
dipole vector er to −er. In polar coordinates this parity inversion is accomplished by

r → −r (5.129)
θ →π−θ (5.130)
ϕ→ϕ±π (5.131)

We see from Eq. 5.118 that the radial part of the integrand has odd parity and therefore
the product of the Legendre functions must show odd parity as well. Therefore l must
differ from l0 by ±1 or we have the selection rule

∆l = ±1 (5.132)

Summarizing, the selection rules for allowed dipole transitions in the hydrogen atom
are

∆l = ±1 (5.133)
∆ml = ±1, 0 (5.134)

The various allowed atomic transitions are often summarized in a schematic chart
called a Grotrian diagram. Figure 5.4 shows a Grotrian diagram for some of the
 Interaction of Light with the Hydrogen Atom 189

Fig. 5.4 Grotrian diagram for nl energy levels of the hydrogen atom. The arrows indicate
some of the allowed transitions. Even with the restrictive selection rule ∆l = ±1, the com-
plicated cascading transition from high-lying energy states to lower-lying states is evident.

transtions in the hydrogen atom. The notation for the levels derives from atomic
spectroscopy. The prinicipal quantum number n is indicated as n = 1, 2, 3, 4, . . . but
the l quantum number is assigned a letter s, p, d, f, . . .. The letters indicate progres-
sions of spectroscopic lines that have some common property. The quantum number
l = 1 is labeled s for “sharp” lines. The l = 2 levels are labeled p for “principal” lines,
l = 2 → d for “diffuse” lines.

Calculation of atomic absorption and emission rates. Now let us go back to the
coupled equations 5.105, 5.106 and rewrite them in terms of the Rabi frequency and
ω0 , the frequency difference between the ground and excited states.
dC1
Ω0 cos ωt e−iω0 t C2 = i (5.135)
dt
dC2
Ω∗0 cos ωt eiω0 t C1 = i (5.136)
dt
We take the initial conditions (at t = 0) to be unit probability of finding the two-state
system in state 1 and null probability for finding it in the excited state 2. At later
2
times the probability for finding the system in state 2 is P = |C2 (t)| . At early times
the rate of increase of this probability is
190 Radiation in Classical and Quantal Atoms

2
dP |C2 (t)|
= (5.137)
dt t
so our task is to calculate C2 (t) at times just later than t = 0. Applying the inital
conditions, it is not difficult to show that
Ω∗0 1 − ei(ω0 +ω)t 1 − ei(ω0 −ω)t
 
C2 (t) = + (5.138)
2 ω0 + ω ω0 − ω
If the frequency ω of the driving wave approaches the two-state difference frequency
ω0 it is quite clear that the first term in Eq. 5.138 will be negligible compared to the
second. Near resonance we can therefore write to good approxiamtion
Ω∗0 1 − ei(ω0 −ω)t
 
C2 (t) ' (5.139)
2 ω0 − ω
We have therefore
2 2 sin2 [(ω0 − ω)(t/2)]
|C2 (t)| = |Ω0 | 2 (5.140)
(ω0 − ω)
and when ω → ω0 , around resonance, we obtain a limiting expression
2 1 2
|C2 (t)| = |Ω0 | t2 (5.141)
4
Thus with ω near ω0 and at early times under weak-field conditions, the probability
rate for finding the system in the excited states increases linearly with time.
dP 1 2
= |Ω0 | t (5.142)
dt 4
5.7.3 Finite spectral width of absorption
The incident driving field always has some spectral width. The source might be, for
example, a broad band arc lamp or the output from a monochromator coupled to
a synchrotron. Therefore we need to write the field eneergy as an integral over the
spectral energ density of the excitation source in the neighborhood of the transition
frequency.
Z ω0 + 21 ∆ω
1 2
0 E0 = ρω dω (5.143)
2 ω0 − 12 ∆ω
where ρω is the spectral energy density of the excitation source with a line width
of ∆ω. Note that the MKS units of spectral energy density are Jm−3 s. Now from
Eq. 5.140 we can write
2
sin2 [(ω0 − ω)t/2]

2 p12 E0
|C2 (t)| = 2 (5.144)
~ (ω0 − ω)
and substituting from Eq. 5.143 we have
ω0 + 12 ∆ω
2p212
Z
2
|C2 (t)| = ρω dω (5.145)
0 ~2 ω0 − 21 ∆ω

For conventional broad-band excitation sources we can safely assume that the spectral
density is constant over the line width of the atomic transition and replace ρ(ω) in the
 Interaction of Light with the Hydrogen Atom 191

integrand with ρ(ω0 ) outside it. Assuming a continuous wave (cw) excitation source,
the integral can be evaluated
ω0 + 12 ∆ω
sin2 [(ω0 − ω)t/2]
Z
πt
2 dω = (5.146)
ω0 − 21 ∆ω (ω0 − ω) 2

So finally the expression for the time rate of increase of the probability of finding
the system in state 2 (i.e. the radiation absorption rate), integrated over the spectral
width of the source, is
2
dP |C2 (t)| πp212
= = ρ(ω0 ) (5.147)
dt t 0 ~2
5.7.4 Comparison to Einstein B coefficient
Now from Complement D we know that the Einstein A and B coefficients represent
spontaneous emission and stimulated absorption under conditions of equilibrium be-
tween a radiation field and a collection of material absorber-emitters. These coefficients
were introduced into phenomenological rate equations and originally had no underlying
interpretation of just how the light was absorbed or emitted. However, by recoginizing
that the Einstein absorption rate, B12ρω , is equivalent to the semiclassical dP/dt we
can write the B coefficient in terms of the quantum mechanical dipole matrix element
p12 . More succinctly
dP πp212
= ρ(ω0 ) = B12 ρ(ω0 ) (5.148)
dt 0 ~2
or
πp212
B12 = (5.149)
0 ~2
The p12 that appears in Eq. 5.148 has a specific orientation with respect to the po-
larization of the exciting field. Equation 5.115 shows the angle θ of p12 with respect
to the E-field of the exciting light is zero. However the Einstein B coefficient assumes
isotropic radiation, so in order to really make the comparison of B and p12 meaningful
we have to average the right side of Eq. 5.149 over all angles. The average value of the
square of the dipole moment matrix element is given by
1 2
p212 = p212 cos2 θ = p (5.150)
3 12
So, finally we have
πp212
B12 = (5.151)
30 ~2
We have restricted this development to the consideration of an artificial atom, the
two-level system without any substructure or energy-level degeneracy. No real atom
corresponds to this simple scheme. Even the simplest atom, hydrogen, with one elec-
tron and one proton, exhibits orbital angular momentum degeneracy in the excited
p level. However it is a simple matter to take degeneracy into account in writing the
expressions for B12 , B21 and A21 , the stimulated absorption, stimulated emission, and
192 Radiation in Classical and Quantal Atoms

spontaneous emission coefficients, respectively. The stimulated absorption and emis-

sion coefficients are related to the degeneracies g1 , g2 of states 1, 2 respectively by
g1 B12 = g2 B21 (5.152)
The spontaneous emission rate A21 is independent of the upper state degeneracy but
is related to B21 by
~ω 3
A21 = B21 2 03 (5.153)
π c
Therefore we have
g1 g1 π 2 c3
B21 = B12 = (5.154)
g2 g2 ~ω03
and, using Eq. 5.151
g1 ω03 p212
A21 = (5.155)
g2 3π0 ~c3
Thus if we can calclulate or determine experimentally the transition dipole moment
of a quantal atom we can obtain the spontaneous emission rate and the stimulated
emission and absorption rate coeffficients.

5.7.5 Interpretation of Dipole Radiation in Quantal Atoms

We saw in Chapter 2, Section 2.8, Eq. 2.137 that the leading source of classical elec-
tromagnetic radiation is the harmonically oscillating charge dipole,
p(t) = p0 cos(ωt)ẑ = q0 a cos(ωt)ẑ
where q0 is a point charge oscillating over a length a aligned along ẑ with frequency ω.
In classical electrodynamics an oscillating dipole is a tangible, physical entity with ob-
servable properties. In quantum mechanics the dipole, like all observables, is cast as an
abstract mathematical opertor, p̂(t) = qr̂(t). How then does one connect the classical
picture with the quantal formulation? The answer is through the dipole “matrix ele-
ment” or expression for the “average value” of the dipole operator. The conventional
form for the matrix element is to sandwich the operator between the wavefunction
representing the state of the dipole (on the right) and its complex consugate (on the
left). The “sandwich” is then integrated over all space.
Z
hp̂i = Ψ∗ (r, t)p̂Ψ(r, t) dV (5.156)

The rationale for this expression comes from the interpretation of the product of a
wavefunctions and its complex conjugate as a probability density P.
P = Ψ∗ (r, t)Ψ(r, t)
Assuming the wavefunction is normalized,
Z Z
P dV = Ψ∗ (r, t)Ψ(r, t) dV = 1

In classical statistics the average value of any quantity is calculated by integrating

the probability distribution of values over the space of all possible values. A familiar
 Interaction of Light with the Hydrogen Atom 193

example is the average speed v of a gas molecule subjet to the Maxwell-Boltzmann

distribution at fixed temperature T .
Z  3/2 Z
m 1 2
hvi = P(v)v dv = 4πv 2 e− 2 mv /(kmb T )
· v dv
2πkmb T

where kmb is the Maxwell-Boltzmann constant. Similarly in quantum mechanics

Z
hp̂(t)i = Ψ∗ (r, t)Ψ(r, t)p̂(t) dV (5.157)

but because p̂ is an operator, it is conventionaly placed to the left of Ψ so that it

can “operate” on it, left to right. Now, the state of the atom, when subject to a
harmonic driving field, is not one of the stationary states but a time-dependent linear
combination of them. In the two-level atom we have

Ψ(r, t) = C1 (t)ψ1 (r)e−iω1 t + C2 (t)ψ2 (r)e−iω2 t (5.158)

with mixing coefficients C1 , C2 subject to the normalization condition

2 2
|C1 | + |C2 | = 1 (5.159)

The differential probability of finding the dipole p(r) = −er between r an r + dr is

then
P p dV = Ψ∗ Ψp dV = −eΨ∗ Ψr dV (5.160)

and the average value of the dipole moment is the integral of this differential probability
over all space.
Z
 ∗ ∗
C1 ψ1 (r)eiω1 t + C2∗ ψ2 eiω2 t (−er)·

hp(t)i = −e hr(t)i =

C1 ψ1 (r)e−iω1 t + C2 ψ2 (r)e−iω2 t dV
 
Z
= 2 cos(ω0 t) C1 C2∗ ψ1∗ (r)(−er)ψ2 (r) dV

= 2C12 p0 cos(ω0 t) (5.161)

with ω0 = ω2 − ω1 and C12 = C1∗ C2 = C2∗ C1 and

Z
p0 = ψ1∗ (−er)ψ2 dV (5.162)

We see that the dipole matrix element oscillates in time at frequency ω0 = ω2 − ω1

and with an amplitude controlled by the coefficients C1 , C2 . When the two stationary
states are strongly coupled to form the tine-dependent oscillating states the dipole
transition is strong.
194 Radiation in Classical and Quantal Atoms

5.8 The Fourth Quantum Number: Intrinsic Spin

Up to this point we have considered only the quantum numbers n, l, ml . We have
used these indices to identify properties of the stationary states of the hydrogen atom.
The number n indexes the energy of the state, the number l the magnitude of its
orbital angular momentum, and ml the projection of the angular momentum on the
axis of quantization; specifying, within the limits of the uncertainty principle, the di-
rection in space to which the orbital angular momentum orients. In addition to these
three, a fourth quantum number ms exists. This quantum lables the orientation of
the intrinsic spin angular momentum S of the electron. This spin angular momentum
has no classical analog. It does not correspond to the classical magnetic moment of
a spinning electron and it does not correspond to the magnetic field of an electron
circulating in a Bohr orbit. It arises naturally from the theory of quantum electrody-
namics (QED) which is, unfortunately, beyond
p the scope of this book. Every electron
carries a quantity of intrinsic spin S = ~ s(s + 1) where the spin quantum number
s = 1/2. Therefore
√ the magnitude of spin angular momentum intrinsic to every ele-
cron is S = ( 3/2)~. Analogous to the orbital angular momentum, the spin direction
is given by the projection of S on the quantization axis. For the electron spin there
are only two possible values, ms = ±1/2, and the spin state is labeled by this quan-
tum number. A complete specification of a stationary state of the hydtrogen atom
is therefore given by four quantum numbers: n, l, ml , ms . The quantum charge-dipole
operator does not couple states of different intrinsic spin, and so the quantum number
ms remains unchanged during electric dipole transitions in the hydrogne atom.1 Our
last selction rule is therefore ∆ms = 0.
In fact stationary states of the hydrogen atom with l 6= 0 do exhibit a magnetic
moment that can be interpreted as arising from the electron orbital motion. The
orbiting electron is a current, and Ampère’s law tells us that a magetic field is always
produced at right angles to this loop current. This orbital magnetic is not the intrinsic
spin, but in multielectron atoms the two magnetic dipoles, orbital and spin, can couple.
The results of this coupliing can be observed in the atomic transition spectra, but these
considerations are the proper domain of atomic structure and spectroscopy and will
not be considered further here.

5.9 Other simple quantum dipolar systems

5.9.1 Rigid rotor model of a diatomic molecule
A heteronuclear diatomic molecule has an asymmetric charge density localized near
the two nuclei comprising the molecule. This charge asymmetry produces a net charge
separated by internuclear distance of the molecule, and therefore a permanent dipole.
In the rigid rotator model, we ignore the molecular vibration and consider only the
molecular rotation in space. The projection of a rotating dipole on any axis contained in
1 In multielectron atoms the orbital and spin angular momenta may couple. When this “spin-orbit
coupling” becomes significant, the individual angular momenta L and S are no longer conserved; and
only the total angular momentum J = L + S specifies the state of the system. Further discussion of
angular momentum coupling in multielectron systems is, however, beyond the scope of what we can
discuss here.
 Other simple quantum dipolar systems 195

Fig. 5.5 Rigid rotor model of a heteronuclear diatomic molecule. Internuclear separation
r is considered constant. Mass m2 > m1 . Permanent electric dipole is aligned along r and
rotates with the molecule.

the plane of rotation reveals an effective dipole oscillation and therefore the possibility
of absorption and emission of diople radiation. Note that this dipolar oscillation is
due to nuclear motion; not the electron orbital motion. The nuclei of molecules move
much more slowly than the electrons in the charge density binding the nuclei togther.
Because of this great difference between effective nuclear and elecron velocities, their
motions are, in many cases, effectively decoupled. The energy of the molecule can then
be calculated as the sum of the rotational, vibrational, and electronic energy; and we
can write down a Schrödinger equation for each type of motion, solve them separately
and write the total wave function as a product of three. Thus

Emolecule ' Erot + Evib + Eelec (5.163)

ψmolecule ' ψrot ψvib ψelec (5.164)

This motional decoupling is called the Born-Oppenheimer approximation.

Figure 5.5 shows the setup for modeling the rotational motion. The classical rota-
tional energy is usually expressed in terms of the angular frequency ω and the moment
of inertia, I.
1
Erot = Iω 2 (5.165)
2
The moment of inertia is given by

I = m1 r12 + m2 r22 (5.166)

where
m1 m2
r1 = r and r2 = r (5.167)
m1 + m2 m1 + m2
196 Radiation in Classical and Quantal Atoms

Now substitute 5.167 into 5.166. The result is

m1 m2 2
I= r (5.168)
m1 + m2
and identify the reduced mass µ as
m1 m2
µ= (5.169)
m1 + m2
We now have reduced the problem to one mass and one coordinate
I = µr2 (5.170)
and the energy of rotation can be written in terms of the rigid rotor angular momen-
tum, J.
J = Iω (5.171)
and
J2
Erot = (5.172)
2I
The quantum rigid rotor problem is setup usig the correspondencee principle between
classical dynamical variables and their quantum operator counterparts. From Eq. 5.172
we draw the following correspondences,
Erot → Ĥrot (5.173)
J → Jˆ = −~2 ∇2 (5.174)
r → r̂ = r (5.175)
Then we write the time-independent Schrödinger equation
h i
Ĥrot − Erot ψrot = 0 (5.176)
 
2 2I
∇ + 2 ψrot = 0 (5.177)
~
The solutions to 5.177 are analgous to the orbital angular momentum of the electron
in the hydrogen atom problem. There is a family of solutions
m
ψrot = PJ (cos θ)eimJ ϕ = YJ j (θ, ϕ) (5.178)
m
where PJ (cos θ) are again the associated Legendre functions and YJ j are the spherical
harmnics. The eigne energies are
~2 J(J + 1)
Erot = ; J = 0, 1, 2, . . . (5.179)
2I
where J are the rigid-rotor angular momentum quantum number (analgous to l for
the hydrogen atom). The selection rules on rigid rotor transitions are similar to the
selection rules in hydrogen.
∆mJ = 0, ±1; ∆J = ±1 (5.180)
Figure 5.6 shows a diagram of the energy levels and pure rotational transitions in the
rigid rotor model.
 Other simple quantum dipolar systems 197

Fig. 5.6 Energy level diagram of the rigid rotor eigen states.

In the realm of molecular spectroscopy the energy levels are often classified ac-
cording to terms, by which is meant that the transition energy is expressed in units of
reciprocal length. Thus Eq. 5.179 is expressed as

Erot (J) h
F (J) = = J(J + 1) = BJ(J + 1) (5.181)
hc 8π 2 cI

where h, c have their usual meanings and B is called the rotational constant. It charac-
terizes the rotational energy in units of reciprocal length. Traditionally in spectroscopy
the conventional unit is cm−1 , called a “wavenumber”. Pure rotational transitions
typically have rotational constants of some tens of cm−1 . For example the rotational
constant of hydrogne iodide, HI, is about 6.5 cm−1 and that of H2 is about 61 cm−1 .

5.9.2 Harmonic oscillator model of a diatomic molecule

For heteronuclear diatomic molecules, a permanent dipole moment exists, oriented
along the internuclear axis. Harmonic molecular vibrations along this axis will result
in a dipole oscillation at the same frequency, and therefore quantized dipole emission
and absorption can take place. Figure 5.7 shows the harmonic oscillator model of the
heteronuclear diatomic moleucle. The energy of a harmonic oscillator is the sum of its
kinetic and potential energies. From elementary treatments we know that the kinetic
energy is p2 /2µ, where p is the linear momentum of the reduced mass µ, and the
potential energy V (r) is given by
198 Radiation in Classical and Quantal Atoms

Fig. 5.7 Harmonic oscillator model of molecular vibration along the internuclear axis r. If
m1 6= m2 then a charge dipole exists along the axis, the amplitude of which will vary with
the oscillation of r.

1 2
V (r) = kr (5.182)
2
where k is the force constant given by

k = µω 2 (5.183)

with ω the angular frequency of oscillation. Once again we use the correspondence
principle to form the Hamiltonian operator.
∂
p → p̂ = −i~ (5.184)
∂r
r → r̂ = r (5.185)
2
p̂ 1
Evib → Ĥvib = + kr̂2 (5.186)
2µ 2
The harmonic oscillator Schrödinger equation then becomes
h i
Ĥvib − Evib ψvib = 0 (5.187)

The eigne energies take on the form

 
1
Evib = ~ω v + ; v = 0, 1, 2, . . . (5.188)
2
√
and the eigen functions are closely related to the Hermite polynomials Hv ( αr) dis-
cussed in Appendix G. The physically admissable solutions to Eq. 5.187 are
 Other simple quantum dipolar systems 199

Fig. 5.8 Energy level diagram showing allowed energies of the quantum harmonic oscillator.
Note that the ground state energy E0 = 1/2~ω.
√ 2
ψvib = Nnorm Hv ( αr)e−µω/~x (5.189)
√
where Nnorm = 1/ 2v π 1/2 v! is a nomalization factor and α = µω/~ is a scaling factor
for the coordinate r. Figure 5.8 shows the equally spaced ladder of energies Evib = ~ω
starting from the ground state with zero-point energy 1/2~ω.
In spectroscopy the energy of vibrational transitions are expressed as vibrational
terms analogous to the spectroscopy of pure rotations. The vibrational terms are
expressed as  
Evib 1
G(v) = =ω v+ ; v = 0, 1, 2, . . . (5.190)
hc 2
where ω means the vibrational energy, usually reported in units of cm−1 . This vibra-
tional ω should not be confused with the angular frequency of some wave or transition.
Unfortunately spectroscopic notation became entrenched long before the modern as-
sociation of ω with a unit of angular frequency. The vibrational transitions of diatomic
molecules exhibit much greater energy than rotational transitions. For example the vi-
brational termG(v) in HI is 2309 cm−1 , factor of about 350 greater than the rotational
term.
5.9.3 Electronic states of molecules
In addition the rotational and vibrational transitions, arising from the motion of per-
manent dipoles, molecules can undergo electronic transitions, somewhat analogous to
200 Radiation in Classical and Quantal Atoms

Table 5.5 Molecular transitions, term energies, and frequencies

Transition Type Term Energy (cm−1 ) Frequency Range

Rotation 3-30 EHF Radio
Vibration 1000 - 3000 THz-IR
Electronic 10 000 - 30 000 Vis-UV

the transitions we have studied in the hydrogen atom. Diatomic molecules have lower
symmetry (cylindrical) than atoms (spherical), and therfore the great simplification of
variable separation is not so easy to carry out. Nevertheless one can characterize molec-
ular orbitals and assign symmetry quantum number to electronic states. The proper
study of molecular structure and spectroscopy is outside the scope of this book, but
suffice it to state that molecular electronic transitions do occur and their energies are
again much greater than pure vibrational or rotational transitions. Continuing with
the example of HI, the first electronic transition between molecular bound states is the
X → B transition. Here X indicates a molecular ground electronic state and B labels
a bound excited electronic state. The electronic spectroscopic term for this transition
is about 5.6 × 103 cm−1 , about 25 times greater than the vibrational term. Hierarchy
of Transitions Rotation, vibration, and electronic transitions follow an energy hierar-
chy from relatively low to high. Table 5.5 summarizes the typical energies for these
dipole transitions. Rotational spectroscopy can be studied with microwave technology,
but pure vibrational transitions (at least for “typical” diatomic molecules) fall in an
awkward part of the electromagnetic spectrum. Radiaton sources and detectors in the
THz (1012 Hz and far infrared (IR) regimes are not, at present, a well-developed and
readily available technology. Electronic spectroscopy, in contrast, can be studied with
a vast array of sources, optical dispersion instrumentation, and detectors. Of course
when a molecule absorbs or emits light in the visible-ultraviolet range it undergoes a
“rovibronic” transition in which all three types of internal molecular motion change.
Figure 5.9 shows schematically how these rovibronic transitions couple lower states
to higher states. The rotation-vibration motion gives rise to underlying manifolds of
transition lines, riding on top of the electronic transition. The amplitude and spacing
of the transition lines within the manifold provides information on vibrational and
rotational motion.

5.10 Exercises
1. Calculate the average value of the electron radius for the hydrogen atom in its
ground state, R10 . Use the expression from Table 5.1.
2. For the same state, calculate the maximum probability for finding the electron at
some distance from the proton.
3. Explain why the average value for the electron radius and the most probable value
for the electron radius are not equal.
4. Given the ground state and first excited states of the hydrogen as
1 1
ψ100 = √ 3/2 e−r/a0
πa
0
 Further Reading 201

Fig. 5.9 Schematic of rovibronic transitions where radiation connects a ground-state mani-
fold of vibration-rotation states to an excited-state manifold. Separations between states are
not to scale.

and
1 1
ψ210 = √ re−r/2a0 cos θ
π (2a0 )5/2
show that
215/2 ea0
p12 =
35
where e is the charge on the electron and a0 is the Bohr radius.
5. The energy level spacing for this first transition in hydrogen is called the Ly-
man α transition. The wavelength at line center is λ = 121.7 nm. Calculate the
spontaneous emission rate A21 (s−1 ) for this transition.

5.11 Further Reading

1. J.-P. Pérez, R. Carles, and R. Fleckinger, Électromagnétism, Fondements et ap-
plications, 3ème édition, Masson, 1997.
2. C. Cohen-Tannoudji, B. Diu, F. Laloë, Quantum Mechanics, Wiley, 1977.
3. G. B. Arfken and H. J. Weber, Mathematical Methods for Physicists, 6th edition,
Elsevier Academic Press, 2005.
4. R. Loudon, The Quantum Theory of Light, 3rd edition, Oxford Press, 2003.
5. A. Corney, Atomic and Laser Spectroscopy, Clarendon Press, Oxford, 1977.
6. J. Weiner, P.-T. Ho, Light-Matter Interaction: Fundamentals and Applications,
Wiley-Interscience, 2003.
7. G. Herzberg, Molecular Spectra and Molecular Structure; I Spectra of Diatomic
Moleucles, Van Nostrand Reinhold, 1950.
Complements

202
D
Classical Blackbody Radiation

D.1 Field Modes in a Cavity

Now that we have established the essential vocabulary of radiation fields and the equa-
tions of motion governing them, we can begin our discussion of light–matter interaction
by applying our new language to a straight-forward problem from the classical theory
of radiation. What we seek to do is calculate the energy density inside a bounded
conducting volume. We will then use this result to describe the interaction of the light
with a collection of two-level atoms inside the cavity.
The basic physical idea is to consider that the electrons inside the conducting
volume boundary oscillate as a result of thermal motion and, through dipole radiation,
set up electromagnetic standing waves inside the cavity. Because the cavity walls are
conducting, the electric field E must be zero there. Our task is twofold: first to count
the number of standing waves that satisfy this boundary condition as a function of
frequency; second, to assign an energy to each wave, and thereby determine the spectral
distribution of energy density in the cavity. After decoupling the electric and magnetic
fields in Eqs. 2.27, 2.28 we arrive at an expression that describes the equation of motion
of an E-field wave,
1 ∂2E
∇2 E = 2 2
c ∂t
propagating in a charge-free space containing no E-field sources,

∇·E=0

Harmonic waves standing-wave solutions factor into oscillatory temporal and spatial
terms. Now, respecting the boundary conditions for a three-dimensional box with sides
of length L, we have for the components of E

Ex (x, t) = E0x e−iωt cos(kx x) sin(ky y) sin(kz z)

Ey (y, t) = E0y e−iωt sin(kx x) cos(ky y) sin(kz z) (D.1)
−iωt
Ez (z, t) = E0z e sin(kx x) sin(ky y) cos(kz z)

where again k is the wave vector of the light, with amplitude

2π
|k| = (D.2)
λ
and components
204 Blackbody Radiation

Fig. D.1 Mode points in k space. Left panel shows one-half the volume surrounding each
point. Right panel shows one-eighth the volume of spherical shell in this k space.

πn
kx = n = 0, 1, 2, ...
L
and similarly for ky , kz . Notice that the cosine and sine factors for the Ex field
component show that the transverse field amplitudes Ey , Ez have nodes at 0 and L as
they should and similarly for Ey and Ez . In order to calculate the mode density, we
begin by constructing a three-dimensional orthogonal lattice of points in k space as
π
shown in Fig. D.1. The separation between points along the kx , ky , kz axes is L , and
the volume associated with each point is therefore
 π 3
V =
L
Now the volume of a spherical shell of radius |k| and thickness dk in this space is
4πk 2 dk. However, the periodic boundary conditions on k restrict kx , ky , kz to positive
values, so the effective shell volume lies only in the positive octant of the sphere. The
number of points is therefore just this volume divided by the volume per point:
1 2


8 4πk dk 1 3 k 2 dk
Number of k points in spherical shell = = L
π 3 2 π2


L

Remembering that there are two independent polarization directions per k point, we
find that the number of radiation modes between k and dk is,

k 2 dk
Number of modes in spherical shell = L3 (D.3)
π2
and the spatial density of modes in the spherical shell is
 Planck mode distribution 205

Number of modes in shell k 2 dk

= dρ(k) =
L3 π2
We can express the spectral mode density, mode density per unit k, as

dρ(k) k2
= ρk = 2
dk π
and therefore the mode number as
k2
ρk dk = dk
π2
with ρk as the mode density in k - space. The expression for the mode density can be
converted to frequency space, using the relations
2π 2πν ω
k= = =
λ c c
and
dν c
=
dk 2π
Clearly
ρν dν = ρk dk
and therefore
8πν 2 dν
ρν dν =
c3
The density of oscillator modes in the cavity increases as the square of the frequency.
Now the average energy per mode of a collection of oscillators in thermal equilibrium,
according to the principal of equipartition of energy, is equal to kB T , where kB is the
Boltzmann constant. We conclude therefore that the energy density in the cavity is

8πν 2 kB T dν
ρRJ
E (ν)dν = (D.4)
c3
which is known as the Rayleigh–Jeans law of blackbody radiation; and, as Fig. D.2
shows, leads to the unphysical conclusion that energy storage in the cavity increases
as the square of the frequency without limit. This result is sometimes called the “ul-
traviolet catastrophe” since the energy density increases without limit as oscillator
frequency increases toward the ultraviolet region of the spectrum. We achieved this
result by multiplying the number of modes in the cavity by the average energy per
mode. Since there is nothing wrong with our mode counting, the problem must be in
the use of the equipartition principle to assign energy to the oscillators.

D.2 Planck mode distribution

We can get around this problem by first considering the mode excitation probability
distribution of a collection of oscillators in thermal equilibrium at temperature T .
206 Blackbody Radiation

Fig. D.2 Left panel: Rayleigh–Jeans blackbody energy density distribution as a function
of frequency, showing the rapid divergence as frequencies tend toward the ultraviolet (the
ultraviolet catastrophe). Right panel: Planck blackbody energy density distribution showing
correct high-frequency behavior.

This probability distribution Pi comes from statistical mechanics and can be written
∞
in terms of the Boltzmann factor e−i /kB T and the partition function q = e−i /kB T :
P
i=0

e−i /kB T
Pi =
q
Now, Planck suggested that instead of assigning the average energy kB T to every oscil-
lator, this energy could be assigned in discrete amounts, proportional to the frequency,
such that  
1
i = n i + hν
2
where ni = 0, 1, 2, 3 . . . and the constant of proportionality h = 6.626 × 10−34 J·s. We
then have

ni
e−hν/2kB T e−ni hν/kB T e−hν/kB T
Pi = ∞ = P
∞
ni (D.5)
e−hν/2kB T e−ni hν/kB T e−hν/kB T
P
ni =0 ni =0
  ni  
= e−hν/kB T 1 − e−hν/kB T (D.6)
∞
ni
e−hν/kB T = 1/ 1 − e−hν/kB T . The average
P

where we have recognized that
ni =0
energy per mode then becomes
∞
X
ε̄ = Pi  i =
i=0
∞   ni 
X  hν
e−hν/kB T 1 − e−hν/kB T (ni )hν = (D.7)
ni =0
ehν/kB T −1
 The Einstein A and B coefficients 207

and we obtain the Planck energy density in the cavity by substituting ε̄ from Eq. D.7
for kB T in Eq. D.4
8πh 3 1
ρP l
E (ν)dν = 3
ν hν/k T dν (D.8)
c e B −1
This result, plotted in Fig. D.2, is much more satisfactory than the Rayleigh–Jeans
result since the energy density has a bounded upper limit and the distribution agrees
with experiment.

D.3 The Einstein A and B coefficients

Let us consider a two-level atom or collection of atoms inside the conducting cavity.
We have N1 atoms in the lower level E1 and N2 atoms in the upper level E2 . Light
interacts with these atoms through resonant stimulated absorption and emission, E2 −
E1 = ~ω0 , the rates of which, B12 ρω , B21 ρω , are proportional to the spectral energy
density ρω of the cavity modes. Atoms populated in the upper level can also emit
light “spontaneously” at a rate A21 that depends only on the density of cavity modes
(i.e., the volume of the cavity). This phenomenological description of light absorption
and emission can be described by rate equations first written down by Einstein. These
rate equations were meant to interpret measurements in which the spectral width of
the radiation sources was broad compared to a typical atomic absorption line width
and the source spectral flux I¯ω (W/m2 ·Hz) was weak compared to the saturation
intensity of a resonant atomic transition. Although modern laser sources are, according
to these criteria, both narrow and intense, the spontaneous rate coefficient A21 and the
stimulated absorption coefficient B12 are still often used in the spectroscopic literature
to characterize light–matter interaction in atoms and molecules. These Einstein rate
equations describe the energy flow between the atoms in the cavity and the field modes
of the cavity, assuming of course, that total energy is conserved:

dN1 dN2
=− = −N1 B12 ρω + N2 B21 ρω + N2 A21 (D.9)
dt dt
dN1
At thermal equilibrium we have a steady-state condition dt = − dN
dt = 0 with
2

ρω = ρth
ω so that
A
ρth
ω =   21
N1
N2 B12 − B21

and the Boltzmann distribution controlling the distribution of the number of atoms
in the lower and upper levels,
N1 g1
= e−(E1 −E2 )/kT
N2 g2

where g1 , g2 are the degeneracies of the lower and upper states, respectively. So
A
21
A21 B21
ρth
ω =   =  (D.10)
g1 ~ω0 /kT g1 ~ω0 /kT B12
g2 e B12 − B21 g2 e B21 − 1
208 Blackbody Radiation

But this result has to be consistent with the Planck distribution, Eq. D.8:
8πh 3 1
ρP l
E (ν) dν = ν dν (D.11)
c3 0 ehν0 /kB T − 1
~ 1
ρP l 3
E (ω) dω = 2 3 ω0 ~ω0 /kB T dω (D.12)
π c e −1
Therefore, comparing these last two expressions with Eq. D.10, we must have
g1 B12
=1 (D.13)
g2 B21
and
A21 8πh
= 3 ν03
B21 c
or
A21 ~ω 3
= 2 03
B21 π c
These last two equations show that if we know one of the three rate coefficients, we
can always determine the other two.
It is worthwhile to compare the spontaneous emission rate A21 to the stimulated
emission rate B21
A21
= e~ω0 /kT − 1
B21 ρth
ω

which shows that for ~ω0 much greater than kT (visible, UV, X - ray), the spontaneous
emission rate dominates; but for regions of the spectrum much less than kT (far IR,
microwaves, radio waves), the stimulated emission process is much more important. It
is also worth mentioning that even when stimulated emission dominates, spontaneous
emission is always present.
Appendices

209
A
Systems of Units in
Electromagnetism

A.1 General Discussion of Units and Dimensions

We conventionally decide to take the three mechanical units of mass m, length l, and
time t, as three fundamental, basic units. However, this choice is just for convenience.
The number of units need not be three and they need not be these three. For example
we could increase the number of units by defining the kinetic energy as
 2
l
E =k·m· = kv 2 (A.1)
t
where k is a constant of proportionality with some dimensions, say, the ratio of charge
to mass of an electron, (e/me ). We could define
e
k=2 (A.2)
me
Then kinetic energy would be expressed by
 
e
E= v2 (A.3)
me

and would have units of l2 /qt2 , where q is some unit of electrical charge. The number of
basic units is now four: mass, length, time, and electric charge. We could also decrease
the number of units by setting some constants equal to unity and without dimensions.
For example setting me = e = ~ = 1 is such a choice, sometimes called atomic units,
and convenient in atomic and molecular quantum mechanical expressions.

Here is another example: The volume can be defined as function of the length unit,

V = kl3 (A.4)

where k is an arbitrary constant, usually set equal to unity. But suppose we had some
problem dealing only with spherical volumes,
4
V = k πr3 (A.5)
3
3
We could define k to be 4π which would make the volume read
 Coulomb’s Law 211

V = r3 (A.6)
in some system of ’rationalized’ units.

Now electromagnetic (E-M) theory differs from classical mechanics in that a fun-
damental constant, ”the speed of light”, c appears explicitly. This constant has di-
mensions, l/t, and if these units are changed, the appearance of E-M expression will
change as well. We will see how the speed of light makes it’s appearance by comparing
electrostatic and magnetostatic forces.

A.2 Coulomb’s Law

Suppose we start with Coulomb’s law,where ”force” has been defined in terms of mass,
length, and time but the constant of proportionality, K1 , or the units of electrical
charge q have not yet been fixed.
q q 
1 2
F = K1 · (A.7)
r2
We now choose to set K1 = 1 which then determines the dimensionality of the charge
to be l3/2 m1/2 /t. This choice of proportionality constant in Coulomb’s law is used in
the esu (electrostatic units) system of units. The usual units of mass, length, and time
in this system are the gram, centimeter, and second. The unit of charge is called the
’statcoulomb’

We can use Coulomb’s law to write the electric field as ’force per unit charge’,
F q2
=E= 2 (A.8)
q1 r
Or more properly in vector form
Z 
r̂ r̂
E=q → ρ(x)d3 x (A.9)
r3 V r3
where r̂ is the unit vector pointing in the E direction, and ρ(x) is the charge density.
Now from Gauss’s law we know
I Z
E · n da = 4π ρ(x)d3 x (A.10)
S V

and from the divergence theorem

I Z
E · n da = ∇ · E d3 x (A.11)
S V

from which we get the differential form of Gauss’s law,

∇ · E = 4πρ (A.12)
1
But instead of choosing K1 = 1 in Eq. A.7 we could have chosen K1 = 4π which would
make the 4π factor in Eqs. A.10, A.12 disappear. Such a choice of scaled units might
212 Systems of Units

be called ’rationalized’ units.

Up to now we have only considered E-M fields in a vacuum, but when we generalize
to dielectric media the electric vector field E, becomes the displacement vector D and
is expressed as
D = 0 E + κP (A.13)
The vector field P is called the ’polarization’ and represents the density of dipole
moments in the material. The constants 0 and κ are proportionality constants. These
constants are not the same in esu and MKS units. In the esu system the displacement
vector field is
D = E + 4πP (A.14)
In this system 0 is chosen to be a pure number with unit value. This choice implies
that D, E, and P all have the same dimensions.

In the MKS system

D = 0 E + P (A.15)
As usual, the 4π factor is eliminated, but the factor 0 appears explicitly. Note from
Eq. A.7 and the choice of K1 = 1/ (4π0 ) that this factor in the MKS system has
dimensions of q 2 t2 /ml3 and units of ”Coulomb squared per Joule meter” (C 2 /J · m)
so the electric field vector E does not have the same dimensions as the displacement
vector D and the polarization vector P. To be consistent with the displacement field
in polarizable materials, the displacement field in free space is just written as
D0 = 0 E0 (A.16)
and P is zero since we do not consider free space to be polarizable (at least in classical
electrodynamics). Now going back to the differential form of Gauss’s law (Eq. A.12)
and inserting 0 on both sides, we have the generalization to the displacement field
form that gives us
∇ · D0 = 4π0 ρ (A.17)
But again, instead of choosing K1 = 1 as we did in Eq. A.7 we could have chosen
1
K1 = 4π 0
which would result in the disappearance of messy constants in the final
expression for Gauss’s law.
1
∇ · D0 = ρ and ∇ · E0 = ρ (A.18)
0
This choice for K1 forms the basis for the ’rationalized MKS’ system of units.

A.3 Ampère’s Law

Currents I1 , I2 moving in two parallel wires separated by some known distance d also
give rise to a force. The force equation, analogous to the Coulomb force in electrostat-
ics, is
dF I1 I2
= 2K2 (A.19)
dl d
where K2 is a proportionality constant whose value depends on the choice of units.
Note that here the force is ’per unit length’ of the infinitely long wires. Now from
 Ampère’s Law 213

Table A.1 Common systems of units in electromagnetism

Name units K1 K2 Remarks

2
Electrostatic cgs 1 1/c charge unit statcoulomb
Electromagnetic cgs c2 1 current unit abampere
Heaviside-Lorentz cgs 1/4π 1/(4πc2 ) rationalized emu
Gaussian cgs 1 1/c2 convenient formally
SI m,kg,s 1/(4π0 ) µ0 /4π convenient numerically

Eqs. A.7 and A.19 we can do a dimensional analysis of the ratio, K1 /K2 . The result
is that the ratio of the two constants has units of velocity squared (l/t)2 It turns
out from measurement that for equal electrostatic and magnetostatic forces the value
of this ratio is equivalent to the speed of light in vacuum squared, c2 . So we have
K1 /K2 = c2 .

The magnetic induction field B is defined as the ’force per unit current’ analogous
to the ’force per unit charge’ that defines the electric field E.
I
B = 2K2 (A.20)
d
However, Eq. A.20 is not quite the whole story. A defning relation between E and B
would imply that B is just proportional to the ’force per unit current’ so that strictly
speaking we should write
I
B = 2K2 K3 (A.21)
d
and then make some convenience argument about the choice of K3 . In order to really
set K3 correctly we need the equation that relates E to B. Such an equation exists. It
is Faraday’s law of induction.
∂B
∇ · E = −K3 (A.22)
∂t

Notice that the Faraday relates a spatial derivitive of E to a time derivitive of B.

Suppose, just to keep things simple, that E and B are the field components of a plane
wave. Then we have immediately the scalar relation
ω
ikE = K3 ωB or E = K3 B (A.23)
k
The factor ω/k has units of l/t or velocity. We have therefore two easy choices for K3 .
If we set K3 = 1, then B = (1/v) E; and in free space B = (1/c) E. We could also set
K3 = 1/c in which case B = E.
We see that in the esu system of units the choice of K1 =1 means that K2 = c2 .
In the rationalized MKS system K1 = 1/(4π0 ). The constant K2 has to be chosen so
that the ratio is c2 . The choice is the following:
1 µ0
K1 = = 10−7 c2 and K2 = = 10−7 (A.24)
4π0 4π
214 Systems of Units

Table A.2 Electromagnetic symbols and units

Quantity Symbol Units

Electric charge q C
Electric charge density ρ Cm−3
Electric current density J Am−2
Electric current I A
Electric conductivity σ Sm−1
Electric permittivity  F m−1
Electric suseptibility χe unitless
Electric field E V m−1
Displacement field D Cm−2
Magnetic permeability µ Hm−1
Magnetic susceptibility χm unitless
Magnetic field H Am−1
Magnetic induction field B V sm−2

where the ’permeability of free space’ µ0 has been introduced. The permeability serves
the same function for magnetic fields that permittivity serves for electric fields. The
difference between the Electrostatic system and the Gaussian system is in the the
choice of K3 . In the Electrostatic system of units K3 = 1 while in the Gaussian
system K3 = 1/c. The Heaviside-Lorentz system is similar to the Gaussian except
it is “rationalized” thereby eliminating factors of 4π that often appear in the field
equations of the Gaussian, Electrostatic, and Electromagnetic systems.
The magnetic fields in matter, analogous to Eqs. A.14 and A.15, are
1
H = c2 B − 4πM (esu) and H= B − M (MKS) (A.25)
µ0
The vector field H is, strictly speaking, the magnetic field while B should always
referred to as the magnetic induction field. Obviously the magnetic field H plays an
analogous role to the displacement field D in the macroscopic equations of electrostat-
ics. The vector field M is called the ’magnetization’ and plays a role analogous to P,
the polarization. Note that neither in the esu system nor in the MKS system does B
have the same units as H.

Note finally that the choice of K2 = µ0 /4π in the MKS system implies that
1
= c2 (A.26)
0 µ0

Many of the quantities, symbols, and units (in the rationalized MKS system) are
compiled in TableA.2.
B
Review of Vector Calculus

B.1 Vectors
A spatial vector is essentially a geometric entity comprising a magnitude (or length)
and a direction. A scalar is some entity characterized only by magnitude. Many physi-
cal quantities can be represented by scalars and vectors. As an example Fig. B.1 shows
an object of mass m moving in a curved trajectory with velocity v and subject to a
force F. The acceleration ac points toward the origin of the central force. The distance
from the object to the the force center is R. In this example m and R are scalars with
only one number (amplitude) associated with these physical quantities. The vector
quantities, v, F, ac have two associated properties: the amplitude and the direction.
The geometric entity of the vector can be represented by the coordinates of the two
end points. If the coordinate system is transformed to some other, the vector remains
unchanged but its representation changes with the coordinate values of the endpoints.
For example, the direction can be represented by two numbers along a line in one-
dimensional (1-D) space, two pairs of numbers in two-dimensional (2-D) space, and a
pair of three numbers in a three-dimensional (3-D) space. Very often the “tail” of the
vector is referred to the origin of the coordinate system. The 3-D velocity vector, for
example, in Cartesian coordinates takes on the form

v = vx x̂ + vy ŷ + vz ẑ (B.1)

Fig. B.1 Examples of vector quantities and a scalar quantities. Velocity v, force F, and
acceleration ac are vectors. The distance to the origin R and the mass m are scalers.
216 Vector Calculus

Fig. B.2 Coordinate rotation of coordinate axes x − y to coordinate axes x0 − y 0 by angle θ

anticlockwise around vector V. Projections onto x − y and x0 − y 0 coordinate axes are shown.

where the unit vectors in the x, y, z directions are x̂, ŷ, ẑ and the magnitude of the vec-
tor components in theq x, y, z directions are vx , vy , vz . The overall length or amplitude
of the vector is |v| = vx2 + vy2 + vz2 . A vector can also be expressed by a column of
numbers, each element corresponding to the component along a Cartesian coordinate
axis.  
vx
v =  vy  (B.2)
vz
A vector can be rotated within a coordinate space or alternatively the coordinate axes
themselves can be rotated (transformed so that the vector is represented in a new
coordinate space. The rotation operation can be represented by matrix. For a 2-D
counterclockwise rotation we have, for example,

v0
     
cos θ − sin θ v
0
v = x0 = · x (B.3)
vy sin θ cos θ vy

Writing out the matrix multiplication explicitly we have

vx0 = cos θ vx − sin θ vy (B.4)

vy0 = sin θ vx + cos θ vy (B.5)

and Fig. B.2 shows the coordinate rotation. This type of coordinate transformation of a
vector is called a unitary transformation because the length of the vector is preserved.
 Axioms of vector addition and scaler multiplication 217

This property is easy to see by taking the absolute value of the original vector and the
transformed vector. The length of the original vector is give by

1/2
|v| = vx2 + vy2 (B.6)

and the length of the rotated vector is

2 1/2
 
1/2
2
|v0 | = vx0 + vy0 = cos2 θ + sin2 θ vx2 + vy2


1/2
= vx2 + vy2 = |v| (B.7)

In 3-D we can represent the rotation matrix about the three Cartesian coordinate axes
as
 
1 0 0
Rx (θ) =  0 cos θ − sin θ  (B.8)
0 sin θ cos θ
 
cos θ 0 sin θ
Ry (θ) =  0 1 0  (B.9)
− sin θ 0 cos θ
 
cos θ − sin θ 0
Rz (θ) =  sin θ cos θ 0  (B.10)
0 0 1
These matrices rotate the vectors counterclockwise about each axis, assuming that the
rotation axis considered is pointing toward the reader.

B.2 Axioms of vector addition and scaler multiplication

Vectors can undergo algebraic operations such as addition and scaler multiplication.
The axioms that define these operations in linear algebra are summarized in Table B.1

B.3 Vector Multiplication

B.3.1 Scaler Product
The projection of some vector A onto its Cartesian coordinates provides a special
case of the definition of the scaler or “dot” product from which we can get a general
definition between two arbitrary vectors A and B. As shown in Fig. B.3 we project A
onto the coordinate axes x, y, z and define this projection as the “dot” product of A
and the unit vectors x̂, ŷ, ẑ.

Ax = A cos α ≡ A · x̂
Ay = A cos β ≡ A · ŷ
Az = A cos γ ≡ A · ẑ

We posit that the scaler product is associative and distributive.

218 Vector Calculus

Table B.1 Axiomatic properties of vector addition and scaler multiplication

Axiomatic property Vector expression

v1 + (v2 + v3 ) =
association
(v1 + v2 ) + v3
commutation v1 + v2 = v2 + v1
sum identity (existence of v+0=v
null vector)
inversion (existence of −v) v + (−v) = 0

scaler times vector sum s((v1 + v2 ) = sv1 + sv2

distribution scaler sum times vector (s1 + s2 )v = s1 v + s2 v
scaler times vector s1 (s2 v) = (s1 s2 )v

multiplication identity (exis- 1(v) = v

tence of s = 1)

Fig. B.3 Vector scaler product. A · B = AB cos θ.

A · (yB) = (yA) · B = yA · B (B.11)

A · (B + C) = A · B + A · B (B.12)

We use the associative property to get a generalization of the dot product between
two vectors A and B. We can write B in terms of its components and the unit vectors
along the Cartesian axes.
B = Bx x̂ + By ŷ + Bz ẑ
Then
A · B = A · (Bx x̂ + By ŷ + Bz ẑ) (B.13)
But then invoking the associative property
 Vector Multiplication 219

Fig. B.4 Distributive law of vector addition. A · (B + C) = A · B + A · C.

A · B = Bx A · x̂ + By A · ŷ + Bz A · ẑ
= B x Ax + B y Ay + B z Az

Therefore X X
A·B= B i Ai = Ai B i = B · A (B.14)
i i

and we see that the scaler product between two vectors is commutative as well as
associative. In the same way we defined Ax = A · x̂ = A cos α we could define the
projection of A onto the B direction as

AB = A cos θ ≡ A · B̂ = A · B/B

Then multiplying this last equivalency by B,

BA cos θ = A · B

Similarly we could specify the projection of B onto the A direction. Then BA =

B cos θ ≡ B · A/A. Then
AB cos θ = B · A
Once again we see that the scaler product between any two vectors is commutative and
equal to the product of the two vector “lengths” and the cosine of the angle between
them.
The distributive property, Eq. B.12, is illustrated in Fig. B.4. The figure shows
geometrically that the sum of the projections of B and C onto A is equivalent to the
projection of their vector sum B + C onto A.

B · A + C · A = (B + C) · A

B.3.2 Vector Product

The vector or “cross” product of two vectors A and B has meaning in 3-D and is
defined as
A×B=C (B.15)
The vector product of A and B is another vector C. The magnitude of C is related
to those of A and B by
220 Vector Calculus

|C| = |A| · |B| sin θ (B.16)

where θ is the angle between vectors A and B. The direction of C is governed by the
“right-hand screw rule”. The vectors A and B, separated by angle θ, form a plane. As
A rotates toward B, the positive direction of C points in the same direction as would
a right-hand threaded screw. A consequence of this convention is that

A × B = −B × A (B.17)

The cross product is anticommutative. From the definition we can directly obtain basic
relations among the Cartesian unit vectors.

x̂ × ŷ = ẑ ŷ × ẑ = x̂ ẑ × x̂ = ŷ (B.18)

and
ŷ × x̂ = −ẑ ẑ × ŷ = −x̂ x̂ × ẑ = −ŷ (B.19)
It is also clear from the definition, Eq. B.16 that

x̂ × x̂ = ŷ × ŷ = ẑ × ẑ = 0 (B.20)

Notice that these relations can be generated by a cyclic commutation of x̂, ŷ, ẑ starting
from any one of them. In analogy to the scaler product we posit that the cross product
operation is distributive and associative. Thus

A × (B + C) = A × B + A × C (B.21)
(A + B) × C = A × C + B × C (B.22)
A × (sB) = sA × B = (sA) × B (B.23)

Now let us take A × B and decompose this expression into its Cartesian components.

A × B = (Ax x̂ + Ay ŷ + Az ẑ) × (Bx x̂ + By ŷ + Bz ẑ) (B.24)

= (Ay Bz − Az By )x̂ − (Ax Bz − Az Bx )ŷ + (Ax By − Ay Bx ) ẑ
= Cx x̂ + Cy ŷ + Cz ẑ = C (B.25)

We see that the Cartesian components of C are constructed from combining products
of the Cartesian components of A and B. Note also that these combinations commute
cyclically. The easiest way to generate the components of the vector product is by
writing a determinant and expanding it. The first row of the determinant consists
of the Cartesian unit vectors, x̂, ŷ, ẑ. The Cartesian components of A composes the
second row and the Cartesian components of B, the third.

x̂ ŷ ẑ
C = Ax Ay Az (B.26)
Bx By Bz

Expansion of this determinant results in an expression identical to the second line

of Eq. B.24, obtained from the cross product definition applied to the Cartesian unit
vectors.
 Vector Fields 221

Fig. B.5 Geometrical interpretation of the triple scaler product. A · (B × C)

B.3.3 Triple Products

Because the cross product of two vectors is itself a vector we can “dot” this result are
“cross” it with yet another vector. Since the dot operation results in a scaler quantity,
this triple product is called the scaler triple product. Figure B.5 shows that the scaler
triple product can be interpreted geometrically as the volume of a parallelepiped.
Furthermore the scaler triple product is cyclically commutative.

A · (B × C) = B · (C × A) = C · (A × B) (B.27)

Because of the anticommutivity of the cross product operation

A · (B × C) = −A · (C × B)

The vector triple product can be simplified by the “BAC-CAB” rule

A × (B × C) = B(A · C) − C(A · B) (B.28)

which can be verified by direct expansion in Cartesian coordinates.

B.4 Vector Fields

The idea of a field is the assignment of some value or function to all points in a region
of geometric space. If the entity assigned is a scaler, then we have a scaler field. The
distribution of temperature throughout the solar system is an example of a scaler field.
In the theory of electromagnetism, the concept of a vector field is essential. A vector
field is an assignment of a vector to each point in some region of geometrical space.
The electric E and magnetic B force fields produced by an oscillating dipole source or
propagating in a plane wave are examples of vector force fields.
222 Vector Calculus

Associated with vector and scaler fields are three differential operations impor-
tant in physics and engineering: the gradient, the divergence, and (in 3-D) the curl
operations.

B.4.1 Gradient
If ϕ(x, y, z) is a scaler field then the gradient of ϕ is defined as
∂ϕ ∂ϕ ∂ϕ
∇ϕ(x, y, z) = x̂ + ŷ + ẑ (B.29)
∂x ∂y ∂z
Notice that the gradient of scaler field is a vector field consisting of the partial deriva-
tives of the scaler field pointing in the x, y, z directions. If we just considered the total
differential of ϕ in terms of its partial derivatives we have
∂ϕ ∂ϕ ∂ϕ
dϕ = dx + dy + dz (B.30)
∂x ∂y ∂z
In fact we can write this scaler differential as the scaler product of ∇ϕ and the vector
coordinate differential dr = dx x̂ + dy ŷ + dz ẑ.
dϕ = ∇ϕ · dr = |∇ϕ||dr| cos θ (B.31)
where the angle θ is the angle between the gradient and the coordinate vector. Clearly
in order for Eq. B.31 to be in accord with B.30, the angle θ = 0. If θ were any other
value, then the dot product in B.31 would be less than dϕ. So ∇ϕ as defined in Eq. B.29
points in the direction of maximum change in ϕ and its magnitude |∇ϕ| yields the
greatest rate of change (slope) along this direction.
The ∇ = ∂/∂x x̂ + ∂/∂y ŷ + ∂/∂z ẑ is a vector operator which acts on a scaler
field to produce a vector field. The most familiar example of the use of the gradient
in electromagnetism is the relation between an electrostatic field E and an electrical
potential V (x, y, z).
E = −∇V
The gradient operator is sometimes called “grad” or “del” and the gradient operation
on a scaler field ϕ is termed “grad ϕ” or “del ϕ”. The ∇ symbol itself is called the
“nabla”.

B.4.2 Divergence
The divergence operation is the dot product of the vector gradient operator with a
vector field. Suppose we have a vector field ϕ. The result of the divergence operation
is
 
∂ ∂ ∂
∇·ϕ= x̂ + ŷ + ẑ · [ϕx x̂ + ϕy ŷ + ϕz ẑ]
∂x ∂y ∂z
∂ϕx ∂ϕy ∂ϕz
= + + (B.32)
∂x ∂y ∂z
Notice that the divergence operation results in a scaler field because it is a “dot prod-
uct” operation between the vector operator ∇ and the vector field ϕ. The divergence
 Vector Fields 223

Fig. B.6 Panel (A) shows a vector field with very high negative divergence. Vector ampli-
tudes decrease with distance from origin. Panel (B) shows a constant vector field along x
which exhibits null divergence.

of a vector field is a measure of how much the field increases or diverges along the field
coordinates. A vector field radiating from a point in all directions has a very high di-
vergence. A vector field whose magnitude is invariant along some coordinate has a null
divergence. A positive (negative) divergence denotes an increasing (decreasing) field.
This geometrical representation of divergence is shown in Fig. B.6. The divergence of
vector field is sometimes written as div ϕ. The most familiar example of the use of the
divergence in electromagnetism is the differential form of Gauss’s law,
ρ
∇·E= (B.33)
0
where E is the electric field and ρ the electric charge density.

B.4.3 Curl
The vector product of ∇ and a vector field ϕ is another vector field χ, called the curl
of ϕ.
χ=∇×ϕ (B.34)
From the determinant rule for cross product of two vectors we can write
x̂ ŷ ẑ
∂ ∂ ∂
χ= ∂x ∂y ∂z
(B.35)
ϕx ϕy ϕz
Expanding the determinant we have explicitly
     
∂ϕz ∂ϕz ∂ϕz ∂ϕx ∂ϕy ∂ϕx
χ= − x̂ − − ŷ + − ẑ (B.36)
∂y ∂y ∂x ∂z ∂x ∂y
224 Vector Calculus

Fig. B.7 A vector field circulating anticlockwise in the x − y plane producing a curl field
along z but no divergence.

The curl of a vector field is measure of its “rotation” around an axis perpendicular
to the plane of rotation. A vortex field has a very large curl while a field emanating
radially from a point in all direction has no curl. Figure B.7 shows a field of all curl
and null divergence. The explicit expression for this vector field is
y x
V(x, y) = − p î + p ĵ (B.37)
2
x +y 2 x + y2
2

Figure B.8 shows a vector field with finite divergence but null curl along the x = y
diagonals. The expression for this vector field is
y x
V(x, y) = p î + p ĵ (B.38)
x2 + y2 x2 + y2
One can show that the divergence of the curl of vector field is null

∇ · [∇ × (∇ϕ)] = 0 (B.39)

and the curl of the gradient of a scaler field is null.

∇ × (∇ϕ) = 0 (B.40)

The most celebrated curl equation in electromagnetism is Faraday’s law.

∂B
∇×E=−
∂t
In this discussion we have described the grad, div, and curl operations in terms
of Cartesian coordinates. But many other curvilinear coordinate systems are a more
 Integral Theorems for Vector Fields 225

Fig. B.8 A vector field with a null curl at the origin and along the x = y diagonals but with
finite divergence for all x, y except at the point x = 0, y = 0 where the divergence is null.

convenient choice, depending on the symmetry of the problem. Two commonly encoun-
tered curvilinear coordinate systems in physics are cylindrical and polar coordinates.
The expressions for grad, div, and curl in these coordinates is discussed in Appendix
C.

B.5 Integral Theorems for Vector Fields

B.5.1 Integral of the Gradient
We saw in Eq. B.31 that the differential of a scalar field dϕ is equal to the dot product
of the gradient of the field ∇ϕ and the differential of coordinate length dr. If we
integrate the differential dϕ from ϕ(x0 , y0 , z0 ) to ϕ(x1 , y1 , z1 we have
Z 1 Z 1
dϕ = ϕ(x1 , y1 , z1 ) − ϕ(x0 , y0 , z0 ) = ∇ϕ · dr (B.41)
0 0

Equation B.41 says that the line integral of the vector field ∇ϕ using the path dr is
equal to the difference between the scaler field values at the two end points. Since this
difference does not depend on the path, the result of the line integral is independent
of the the path taken; and in particular
I
∇ϕ · dr = 0 (B.42)

for any closed loop since for these the ending point is identical to the starting point.
The integral over a path of the gradient of a scaler field is commonly encountered in
226 Vector Calculus

electrostatic problems where the task is to calculate the potential difference between,
say, two conductors that form an electric field between them. The potential difference
is given by Z 2 Z 2
V21 = V2 − V1 = − E · dr = − ∇V · dr (B.43)
1 1

B.5.2 Integral of the Divergence

The expression for the divergence theorem or Gauss’s theorem is
Z I
(∇ · χ) dV = χ · dS (B.44)
vol surf

where dV is the differential volume element and dS is a differential element of the

surface surrounding (bounding) the volume. Here χ is a vector field, in contrast to
ϕ of Eq. B.31, a scaler field. Note that the surface element dS is itself a differential
vector with the unit vector direction perpendicular to the surface and conventionally
pointing outward. A common application of the divergence theorem is the integral
form of Gauss’s law, discussed in Complement ?? of chapter 2, Eq. 2.123.
Z Z Z Z
∇ · D dV = 0 ∇ · E dV = ρ dV = 0 E · dS
vol vol vol surf

Where D is the displacement vector field, E is the electric field emanating from the
charge density ρ enclosed by the surface boundary S of volume V . Note that the surface
differential element is a vector differential with direction pointing outward from the
boundary and perpendicular to it.

B.5.3 Integral of the Curl

The surface integral over the curl of a vector is called Stokes’ theorem.
Z I
(∇ × χ) · dS = χ · dr (B.45)
surf

The line integral on the right is taken around the boundary of the surface integral on
the left. A convenient way to remember and visualize the geometric interpretation of
the curl integral is to think of Ampère’s law. In differential form it is

∇ × H = Jfree

where H is the magnetic field and Jfree is a free charge current density. The integral
form of this equation is, taking into account the curl theorem,
Z I Z
(∇ × H) · dS = H · dr = Jfree · dS = I (B.46)
surf surf

where I is the charge current running in a wire of cross section S. This relation can be
pictured as a loop of H-field surrounding a straight wire and generated by a current I
running through it.
C
Gradient, divergence and curl in
cylindrical and polar coordinates

Although Cartesian coordinates are the most familiar and serve many purposes, they
are not the only orthogonal coordinate system that can be used to define a set of basis
vectors. Often we can take advantage of the natural symmetry of a problem to express
vectors in some other orthogonal curvilinear coordinate system. The most common of
these are the cylindrical and polar coordinates because they are appropriate for many
practical problems.
In general we can expand a vector V in basis vectors of the Cartesian system or
some other system with basis vectors {q},

V = x̂Vx + ŷVy + ẑVz or V = q̂1 V1 + q̂2 V2 + q̂3 V3 (C.1)

However, we cannot treat the position vector with this general rule. In this special
case

(C.2)
r = xx̂ + y ŷ + z ẑ cartesian coordinates (C.3)
r = ρρ̂ + z ẑ cylindrical coordinates (C.4)
r = rr̂ spherical coordinates (C.5)

In the general case we convert from the Cartesian system to another through the
relations

x = x(q1 , q2 , q3 )
y = y(q1 , q2 , q3 )
z = z(q1 , q2 , q3 )

The complete differential in x can be written

∂x ∂y ∂y
dx = dq1 + dq2 + dq3
∂q1 ∂q2 ∂q3

and similarly for y and z. The differential of the position vector r in the Cartesian
basis is
dr = dx̂ + dŷ + dẑ
228 Curvilinear coordinates

and the square of the distance between two neighboring points can then be written

ds2 = dr · dr = dx2 + dy 2 + dz 2

But the differential of dr can also be expanded in the curvilinear coordinate system
∂r ∂r ∂r
dr = dq1 + dq2 + dq3 (C.6)
∂q1 ∂q2 ∂q3
and
X ∂r ∂r
ds2 = dr · dr = · dqi dqj (C.7)
ij
∂qi ∂qj

Remembering that we have supposed the curvilinear coordinates orthogonal,

q̂i · q̂j = δij

we write Eq. C.7

X ∂2r X 2
ds2 = dq 2 = g11 dq12 + g22 dq22 + g33 dq32 = (hi dqi )
i
∂qi2 i i

where we have identified gii = h2i . We see from Eq C.7 that hi is a scale factor such
that the differential length segment dsi can be written
∂r
dsi = hi dqi and = hi q̂i
∂qi
and from Eq. C.6 the position vector differential dr can be expanded in terms of these
scale factors in the curvilinear basis as

dr = h1 dq1 q̂1 + h2 dq2 q̂2 + h3 dq3 q̂3 (C.8)

From the line segment dsi we can easily write the surface and volume elements σ, τ
in the curvilinear system.

dσij = dsi dsj = hi hj dqi dqj i 6= j

and
dτ = ds1 ds2 ds3 = h1 h2 h3 dq1 dq2 d3
We can expand an area vector σ in the curvilinear basis set analogous to the position
vector r in Eq. C.8

dσ = ds2 ds3 dq̂1 + ds3 ds1 q̂2 + ds1 ds2 q̂3

= h2 h3 dq2 dq3 q̂1 + h3 h1 dq3 dq1 q̂2 + h1 h2 dq1 dq2 q̂3 (C.9)

For example we can take an ordinary vector quantity F and expand it in Cartesian
coordinates or in spherical coordinates.
 The gradient in curvilinear coordinates 229

F = Fx x̂ + Fy ŷ + Fz ẑ (C.10)
F = Fr r̂ + Fθ θ̂ + Fϕ ϕ̂ (C.11)

The unit vectors r̂, θ̂, ϕ̂ in terms of the Cartesian unit vectors are

r̂ = sin θ cos ϕ x̂ + sin θ sin ϕ ŷ + cos θ ẑ (C.12)

θ̂ = cos θ cos ϕ x̂ + cos θ sin ϕ ŷ − sin θ ẑ (C.13)
ϕ̂ = − sin ϕ x̂ + cos ϕ ŷ (C.14)

The differential length vector along the surface in spherical coordinates is then

dl = dr r̂ + r dθ θ̂ + r sin θ dϕ ϕ̂ (C.15)

and the differential volume element in spherical coordinates is obtained from the prod-
uct of ds1 ds2 ds3 with

ds1 = h1 dq1 = dr (C.16)

ds2 = h2 dq2 = r dθ (C.17)
ds3 = h3 dq3 = r sin θ dϕ (C.18)

The differential volume element in spherical coordinates is thus

dV = ds1 ds2 ds3 = dr rdθ r sin θdϕ = r2 sin θ dθ dϕ dr (C.19)

The expressions for dr and dσ in Eqs. C.8, C.9 give us the tools to write vector line
and surface integrals in the curvilinear coordinates
Z XZ
V · dr = Vi hi dqi (C.20)
i
Z XZ
V · dσ = Vi hj hk dqj dqk j 6= k 6= i (C.21)
i

C.1 The gradient in curvilinear coordinates

The gradient is a vector operator ∇ that operates on a scalar point function ψ. In
Cartesian coordinates we write
 
∂ ∂ ∂
∇ψ = x̂ + ŷ + ẑ ψ(x, y, z) (C.22)
∂x ∂y ∂z
An alternative integral definition is more convenient for finding the expression for the
gradient in some curvilinear coordinate system.
R
ψdσ
∇ψ = R lim R (C.23)
dτ →0 dτ
The gradient in general is defined such that it yields the maximum spatial rate of
change of the scalar function, and this maximum spatial change is independent of the
230 Curvilinear coordinates

coordinate system in which it is described. Therefore we should be able to express it

in coordinate systems other than Cartesian. Remembering that the length segments
dx, dy, dz in Cartesian coordinates go over into ds1 , ds2 , ds3 in our chosen curvilinear
system, we can write the gradient in our new coordinates as
∂ψ ∂ψ ∂ψ
∇ψ(q1 , q2 , q3 ) = q̂1 + q̂2 + q̂3 (C.24)
∂s1 ∂s2 ∂s3
and also remembering that dsi = hi dqi we have
1 ∂ψ 1 ∂ψ 1 ∂ψ
∇ψ(q1 , q2 , q3 ) = q̂1 + q̂2 + q̂3 (C.25)
h1 ∂q1 h2 ∂q2 h3 ∂q3

C.2 The divergence in curvilinear coordinates

The divergence ∇ · V of some vector field V can be expressed as a differential in
Cartesian coordinates as
 
∂ ∂ ∂
x̂ + ŷ + ẑ · (Vx x̂ + Vy ŷ + Vz ẑ) (C.26)
∂x ∂y ∂z
Similar to the gradient operation, the divergence can be generalized to curvilinear
coordinates by considering it as the result of taking the limit of a differential vector
field surface integral divided by the differential volume enclosed by the surface as the
volume approaches zero. Taking this limiting ratio operation as the definition of the
divergence we have R
δ V · dσ
∇ · V (q1 , q2 , q3 ) = R lim R
dτ →0 dτ
and the differential of the surface integral is
Z
δ V (q1 , q2 , q3 , ) · dσ =
 
∂(V1 h2 h3 ) ∂(V2 h3 h1 ) ∂(V3 h1 h2 )
+ + dq1 dq3 dq3
∂q1 ∂q2 ∂q3
R
and taking dτ in the limit as ds1 ds2 ds3 = h1 dq1 h2 dq2 h3 dq3 we have for the limiting
ratio

1 ∂ ∂
∇ · V (q1 , q2 , q3 ) = (V1 h2 h3 ) + (V2 h3 h1 )+
h1 h2 h3 ∂q1 ∂q2

∂
(V3 h1 h2 ) (C.27)
∂q3
Another useful general expression is the scalar Laplacian operator, ∇ · ∇ which we
can get from combining the grad and div operations,
    
1 ∂ h2 h3 ∂ψ ∂ h3 h1 ∂ψ
∇ · ∇ψ = + +
h1 h2 h3 ∂q1 h1 ∂q1 ∂q2 h2 ∂q2
 
∂ h1 h2 ∂ψ
(C.28)
∂q3 h3 ∂q3
 The curl in curvilinear coordinates 231

C.3 The curl in curvilinear coordinates

The familiar differential expression for the curl operation in Cartesian coordinates is
     
∂Vz ∂Vy ∂Vx ∂Vz ∂Vy ∂Vx
∇×V = − x̂ + − ŷ + − ẑ
∂y ∂z ∂z ∂x ∂x ∂y

and the easiest way to remember it is by writing the expression as determinant

x̂ ŷ ẑ
∂ ∂ ∂
∇×V = ∂x ∂y ∂z
Vx Vy Vz

As with the gradient and divergence operations, the curl can also be written as the
limiting integral operation
R
dσ × V
∇ × V = R lim R
dτ →0 dτ

It can be shown that by using this integral limiting form and applying Stokes’ theorem
the curl operation can be written in curvilinear coordinates as

q̂1 h1 q̂2 h2 q̂3 h3

1 ∂ ∂ ∂
∇×V = ∂q1 ∂q2 ∂q3 (C.29)
h1 h2 h3
h1 V1 h2 V2 h3 V3

C.4 Expressions for grad, div, curl in cylindrical and polar

coordinates
Circular cylindrical coordinates. Circular cylindrical coordinates consist of three in-
dependent variables, ρ, ϕ, z and their associated unit vectors, ρ̂, ϕ̂, ẑ. The limits on
these variables are,

0≤ρ≤∞ 0 ≤ ϕ ≤ 2π − ∞ < z < +∞

Figure C.1 shows the relation between the Cartesian coordinates x, y, z and the cylin-
drical coordinates, ρ, ϕ, z. The z coordinate is common to both systems. From Fig. C.1
it is evident that
p
ρ = (x2 + y 2 ) x = ρ cos ϕ y = ρ sin ϕ z=z

The differential volume element dV is

dV = dsρ dsϕ dsz = ρdρ ρdϕ dz

From which we identify

h1 = hρ = 1 h2 = hϕ = ρ h3 = hz = 1
232 Curvilinear coordinates

Fig. C.1 Cylindrical coordinates ρ, ϕ, z and unit vectors ρ̂, ϕ̂, ẑ and their relations to Carte-
sian coordinates and unit vectors, x, y, z, x̂, ŷ, ẑ.

Then from the general expressions for grad, div, curl, Eqs. C.25, C.27, C.29 we write
these operators in cylindrical coordinates as
∂ψ 1 ∂ψ ∂ψ
grad ∇ψ(ρ, ϕ, z) = ρ̂ + ϕ̂ + ẑ (C.30)
∂ρ ρ ∂ϕ ∂z
1 ∂(ρVρ ) 1 ∂Vϕ ∂Vz
div ∇·V = + + (C.31)
ρ ∂ρ ρ ∂ϕ ∂z
ρ̂ ρϕ̂ ẑ
1 ∂ ∂ ∂
curl ∇×V = ∂ρ ∂ϕ ∂z
(C.32)
ρ
Vρ ρVϕ Vz
We can also find the scalar Laplacian in cylindrical coordinates by applying the general
expression Eq. C.28
1 ∂2ψ ∂2ψ
 
2 1 ∂ ∂ψ
Laplacian ∇ ψ= ρ + 2 + (C.33)
ρ ∂ρ ∂ρ ρ ∂ϕ2 ∂z 2
Polar coordinates. Polar coordinates consist of three independent variables, r, θ ϕ
and their associated unit vectors, r̂, θ̂, ϕ̂. The limits on these variables are,

0≤r≤∞ 0≤θ≤π 0 ≤ ϕ ≤ 2π

Figure C.2 shows the relation between the Cartesian coordinates x, y, z and the polar
coordinates, r, θ, ϕ. From Fig. C.2 we see that
 Expressions for grad, div, curl in cylindrical and polar coordinates 233

Fig. C.2 Polar coordinates r, θ, ϕ and unit vectors r̂, θ̂, ϕ̂ and their relations to Cartesian
coordinates and unit vectors, x, y, z, x̂, ŷ, ẑ.
p
r= x2 + y 2 + z 2 x = r sin θ cos ϕ y = r sin θ cos ϕ z = r cos θ

From the differential volume dV and inspection of Fig. C.2 we can identify the scale
factors h1 , h2 , h3 .
dV = dsr dsθ dsϕ = dr rdθ r sin θdϕ

hr = 1 hθ = r hϕ = r sin θ
Then from Eqs. C.25, C.27, C.29 we write expression for the grad, div and curl opera-
tors in polar coordinates

∂ψ 1 ∂ψ 1 ∂ψ
grad ∇ψ = r̂ + θ̂ + ϕ̂ (C.34)
∂r r ∂θ r sin θ ∂ϕ

∂r2 Vr
 
1 ∂ sin θVθ ∂Vϕ
div ∇·V = sin θ + r + r (C.35)
r2 sin θ ∂r ∂θ ∂ϕ

r̂ rθ̂ r sin θϕ̂

1 ∂ ∂ ∂
curl ∇×V = ∂r ∂θ ∂ϕ (C.36)
r2 sin θ
Vr rVθ r sin θVϕ
Finally, again by applying Eq. C.28 we can obtain the Laplacian operator in polar
coordinates.
234 Curvilinear coordinates

    
1 ∂ 2 ∂ψ ∂ ∂ψ
Laplacian ∇2 ψ = sin θ r + sin θ +
r2 sin θ ∂r ∂r ∂θ ∂θ
1 ∂2ψ

(C.37)
sin θ ∂ϕ2

Or, expanding the radial terms,

∂ 2 ψ 2 ∂ψ
  
1 ∂ ∂ψ
Laplacian ∇2 ψ = + + sin θ +
∂r2 r ∂r r2 sin θ ∂θ ∂θ
1 ∂2ψ

(C.38)
sin θ ∂ϕ2
D
Properties of Phasors

D.1 Introduction
In Chapter 2, Section 2.10.1 we discussed the phasor form for electromagnetic prop-
agating fields consisting of a single-frequency harmonic time dependence factor e−iωt
and the spatially dependent amplitude modulation factor ei(k·r+ϕ) . The general form
of the field is
A(r, t) = A0 ei(k·r+ϕ) e−iωt (D.1)

and we separated the field into a phasor factor, involving only the spatial dependence,
and a harmonic time factor that was usually invariant.
In general a “phasor” is any function with a sinusoidal modulation, and in this
Appendix we examine some of the useful properties of phasors and their application
to circuit theory.

D.1.1 Phasor Addition

Since phasors can be represented as complex exponential, multiplication of two phasors
Aeiα and Beiβ is elementary.

Aeiα · Beiβ = ABei(α+β = Ceiγ (D.2)

The product of two phasors is another phasor with a phase angle equal to the algebraic
sum of the two individual arguments of the exponential factors.
The addition of two phasors is less obvious. In many cases we are interested in
real sin and cos functions so let us examine the following sum of two real phasors,
V1 (t) = A sin ωt and V2 (t) = B sin(ωt + ϕ).

V3 (t) = V1 (t) + V2 (t)

= A sin ωt + B sin(ωt + ϕ) (D.3)

On intuition we might suppose that the sum of two sin functions differing only in
amplitude and relative phase would be another sin function. We posit therefore that
the form of the sum is
V3 (t) = C sin(ωt + δ) (D.4)

where C and δ are to be determined. Expanding V3 (t) in Eqs. D.3, D.4 and setting
coefficients of sin ωt and cos ωt equal, we find
236 Phasors

Fig. D.1 Relations among two phasors A and B which differ in amplitude and phase. Phase
of A leads phase of B by ϕ. The interior angle γ is related to ϕ by γ = π − ϕ.
 
−1 B sin ϕ
δ = tan (D.5)
A + B cos ϕ
1/2
C = A2 + B 2 + 2AB cos ϕ

(D.6)

The expression for C, the amplitude of the phasor sum, is highly reminiscent of the
result we would expect for the length of a vector C that is the vector sum of A and
B. In this case the length of the vector C would be determined by the law of cosines
as indicated in Fig. D.1. The interior angle γ between the two component vectors is
closely related to ϕ, the relative angle between the two phasors. In fact it is easy to
show that
γ =π−ϕ (D.7)
and substitution in Eq. D.6 results in
1/2
C = A2 + B 2 − 2AB cos γ

(D.8)

confirming that two phasors add with amplitude equivalent to the resultant “length”
from the the sum of two vectors.

D.2 Application of Phasors to Circuit Analysis

As an illustration of the usefulness of phasors to harmonic circuit analysis, we consider
a simple RLC circuit with
Z
di(t) 1
V (t) = Ri(t) + L + i(t)dt (D.9)
dt C

where V, R, L, C have their usual meanings of voltage, resistance, inductance and ca-
pacitance and i(t) is the time-dependent current running through all the lumped circuit
 Application of Phasors to Circuit Analysis 237

elements. We posit that the current source is harmonically oscillating at frequency ω

and write i(t) as a phasor,
i(t) = i0 e−iωt (D.10)
Substituting the phasor form into Eq. D.9 we find
1
V (t) = Ri0 e−iωt + ωLi0 e−i(ωt+π/2) + i0 e−(ωt−π/2) (D.11)
ωC
= VR + VL + VC (D.12)

We see that the voltage drop across the resistor is in phase with the source, while the
inductive reactance, ωL shows a phase advance of π/2 and the capacitive reactance,
1/ωC lags in phase by π/2. The resistive voltage drop is purely dissipative while the
two reactances store energy originating at the source and return it to the circuit at
different points along the harmonic cycle. Equation D.9 can be rewritten in terms of
the impedance, Z(ω),

V (t) = Z(ω)i0 e−iωt (D.13)

 
1
Z(ω) = R + (ωL)e−i(π/2) + ei(π/2) (D.14)
ωC
 
i
= R − iωL + (D.15)
ωC
= R + iXL (ω) + iXC (ω) (D.16)

where XL and XC are the inductive and capacitive reactances in phasor form. We
can take this analysis further by considering the RLC circuit in series and in parallel.
Figure D.2 shows circuit and phasor diagrams for the two cases. In the series circuit
the property common to all lumped elements is the current i(t) so the relevant phasor
quantities are the voltages across the resistive, inductive and capacitive elements. For
the parallel circuit the common property is the voltage V (t) and the relevant phasor
diagram is expressed in terms of the individual currents passing through each circuit
element.
238 Phasors

Fig. D.2 (a) Series RLC circuit with harmonic voltage source and a common current i(t)
running through each circuit element. (b) Voltage phasor diagram showing the phase “lead”
for the capacitive voltage term, the phase “lag” for the inductive voltage term and the net
phase difference between the driving voltage and the resistive voltage drop. Panels (c) and
(d) show similar diagrams for current phasors in RLC parallel circuit.
E
Properties of the Laguerre Functions

The Laguerre polynomials and the “associated” Laguerre polynomials are solution to
Laguerre’s differential equation and are closely related to the radial solutions of the
Schrödinger equation for the hydrogen atom.

E.1 Generating function and recursion relations

The generating function for the Laguerre polynomials is given by
∞
e−xz/(1−z) X
g(x, z) = = Ln (x)z n |z| ≤ 1 (E.1)
1−z n=0

As we did with Legendre and Hermite polynomials, we find recurrence relations be-
tween adjacent members of the Laguerre polynomials by differentiating the generating
function with respect to z and x.

∂g(x, z)
(1 − z 2 ) = (1 − x − z)g(x, z) (E.2)
∂z
∂g(x, z)
(z − 1) = zg(x, z) (E.3)
∂x
Substituting Eq. E.1 into Eq. E.2, adjusting the summation indices, and equating terms
with equal powers of z n , results in the first recurrence relation
 
2n + 1 − x
Ln+1 (x) = Ln − nLn−1 (E.4)
n+1

It is often useful to have a recurrence relation involving the derivatives of the polyno-
mials, and we can obtain one by using Eqs. E.2 and E.3 to derive a third relation,

∂g(x, z) ∂g(x, z) ∂[zg(x, z)]

x =z −z (E.5)
∂x ∂z ∂z
Making the appropriate substitutions from Eq. E.1, adjusting summation indices and
equating terms with like powers of z n results in

xL0n (x) = nLn (x) − nLn−1 (x) (E.6)

which is the second recursion relation for Laguerre polynomials.

240 Laguerre Polynomials

E.2 Orthogonality and Normalization

The Laguerre polynomials themselves do not form an orthogonal set of functions, but
they can be made orthogonal by joining a factor e−x to them. Thus
Z ∞
e−x Lm (x)Ln (x) = δmn (E.7)
0

where the polynomials defined by the generating function, Eq. E.1 are already normal-
ized. As in the case of Hermite polynomials we can identify a closely related Laguerre
function ϕn (x) that has the orthonormal properties of a complete set of functions
involving the Laguerre polynomials,

ϕn (x) = e−x/2 Ln (x) (E.8)

The ordinary differential equation that ϕn (x) satisfies, however, is not the physically
relevant one with which we associate the Schrödinger radial equation for the hydrogen
atom. To find the solutions for that equation we need to examine the “associated”
Laguerre polynomials.

E.3 Associated Laguerre Polynomials

The associated Laguerre polynomials are defined in terms of the Laguerre polynomials
by,
dk
Lkn (x) = (−1)k k Ln+k (x) (E.9)
dx
and given the Rodrigues form for the Laguerre polynomials,
ex dn n −x
Ln (x) = (x e ) (E.10)
n! dxn
from which, after writing the Taylor series expansion of the exponential terms, we can
write a series representation for the “normal” Laguerre polynomials with integral n
n
X (−1)m n!xm
Ln (x) = (E.11)
m=0
(n − m)!m!m!

We can obtain a similar series representation for the associated Laguerre polynomials
by substituting Eq. E.11 into the definition, Eq. E.9
n
X (n + k)!
Lkn (x) = (−1)m xm k > −1 (E.12)
m=0
(n − m)!(k + m)!m!

The first few associated Laguerre polynomials are thus

Lk0 (x) = 1
Lk1 (x) = −x + k + 1
x2 (k + 2)(k + 1)
Lk2 (x) = − (k + 2)x +
2 2
 Associated Laguerre Polynomials 241

E.3.1 Generating Function, Recurrence Relations, and Orthonormality

The generating function for the associated Laguerre polynomials is given by
∞
e−xz/(1−z) X
= Lkn (x)z n |z| < 1 (E.13)
(1 − z)k+1 n=0

from which recurrence relations can be derived by differentiation and equating of terms
with like powers in z in the usual way. The resulting recurrence relation is
(n + 1)Lkn+1 (x) = (2n + k + 1 − x)Lkn (x) − (n + k)Lkn−1 (x) (E.14)
The associated Laguerre polynomials are solutions to the associated Laguerre ordinary
differential equation, but this equation is not “self-adjoint” or hermitian; and therefore
cannot function as a legitimate quantum mechanical operator. However it can be cast
into a form suitable for quantum mechanics, by multiplying the associated Laguerre
polynomials by a factor, e−x/2 xk/2 Lkn (x). This factor transforms the polynomials into
the associated Laguerre functions.
ϕkn (x) = e−x/2 xk/2 Lkn (x) (E.15)
These associated Laguerre functions have the desired orthonormality
Z ∞
(n + k)!
e−x xk Lkn (x)Lkm (x) dx = δmn (E.16)
0 n!
and they satisfy the “quantum mechanically correct” ordinary differential equation
d2 ϕkn (x) dϕkn (x) x 2n + k + 1 k 2
 
x + + − + − ϕkn (x) = 0 (E.17)
dx2 dx 4 2 4x
This equation is “correct” in the sense that the operator integrals corresponding to
average values of observable variables and the eigenvalues themselves are guaranteed to
be real. But in fact Eq E.17 is still not the ordinary differential equations corresponding
to the form of the radial Schrödinger equation for the hydrogen atom. The problem is
that the Schrödinger equation does not exhibit a term with a first-derivative operator.
However, this term can be eliminated by multiplying the Laguerre function by a factor
x1/2
Φkn (x) = e−x/2 x(k+1)/2 Lkn (x) (E.18)
This family of modified Laguerre functions is a solution to the ordinary differential
equation
d2 Φkn (x) 1 2n + k + 1 k 2 − 1
 
+ − + − Φkn (x) = 0 (E.19)
dx2 4 2x 4x2
and has a normalization
Z ∞
2 (n + k)!
e−x xk+1 Lkn (x) dx =

(2n + k + 1) (E.20)
0 n!
Notice that in Eq. E.19 there is no term in the first derivative of x, and therefore this
second-order differential equation and its family of modified Laguerre functions can
serve as a prototype for the radial Schrödinger equation of the one-electron atom.
F
Properties of the Legendre functions

The Legendre functions are a family of solutions to the Legendre differential equation.
This equation determines the angular behavior of many physical problems including
the scalar Helmholtz wave equation in optics, Maxwell’s equations in classical electro-
dynamics, and the Schrödinger wave equation in quantum mechanics. We discuss here
the properties of these functions common to all these physical situations.
The Legendre equation itself is

1 − x2 Pn00 (x) − 2xPn0 (x) + n(n + 1)Pn (x) = 0

(F.1)

where n = 0, 1, 2, . . .. In most physics problems x = cos θ, and the Legendre equation

takes the form
 
1 d dPn (cos θ)
sin θ + n(n + 1)Pn (cos θ) = 0 (F.2)
sin θ dθ dθ

or more explicitly, carrying out the derivative,

d2 Pn (cos θ) dPn (cos θ)

+ cot θ + n(n + 1)Pn (cos θ) = 0 (F.3)
dθ dθ
where the derivative is with respect to the polar coordinate θ rather than x. We
have already encountered this latter form in Eqs. 2.100, 2.101, 2.102 when studying the
angular solutions of Laplace’s equation.

F.1 Generating Function

The Legendre polynomials can always be obtained from a generating function
∞
X
g(t, x) = (1 − 2xt + t2 )−1/2 = Pn (x)tn , |t| < 1 (F.4)
n=0

The left side of this equation can be cast into a power series in tn such that left side
and right side, also a power series in tn can be equated term by term.

∞ [n/2] ∞
X X (2n − 2k)! X
(1 − 2xt + t2 )−1/2 = (−1)k (2x)n−2k) n
t = Pn (x)tn
n=0 k=0
22n−2k n!k!(n − k)! n=0
(F.5)
 Recurrence Relations 243

where the upper limit on the sum, [n/2] means n/2 when n is even and (n − 1)/2 when
n is odd. Therefore we can obtain an expression for each Pn (x) in terms of a sum over
the index k.
[n/2]
X (2n − 2k)!
Pn (x) = (−1)k n xn−2k (F.6)
2 k!(n − k)!(n − 2k)!
k=0
Using Eq. F.6 function we find, for example,
1 5 3
3x2 − 1



P0 (x) = 1 P1 (x) = x P2 (x) = P3 (x) = x − 3x
2 2

F.2 Recurrence Relations

Families of functions that are solutions to a differential equation often exhibit recur-
rence relations that express how a given solution Pn (x) is related in some simple way
to its neighbors, Pn+1 (x) and Pn−1 (x). These recurrence relations are often another,
and simpler way to spawn all the needed members of the family once one member
is known. In the case of the Legendre functions we can find a recurrence relation by
starting from the generating function, Eq. F.4 and differentiating it.
∞
∂g(t, x) x−t X
= 3/2
= nPn (x)tn−1 (F.7)
∂t (1 − 2xt + t2 ) n=0

Using Eq. F.4 we can write this last expression as sum of four terms, each of which is
a power series in t.
∞ ∞

X X
1 − 2xt + t2 nPn (x)tn−1 + (t − x) Pn (x)tn = 0 (F.8)
n=0 n=0

or
∞
X ∞
X ∞
X
nPn (x)tn−1 − x(2n + 1)Pn (x)tn + (n + 1)Pn (x)tn+1 = 0 (F.9)
n=0 n=0 n=0

Now we seek to rewrite the various terms so that they are all sums of a power series
in tn . We can do this by adjusting the indices n. Thus,
∞
X ∞
X
n−1
nPn (x)t is equivalent to (n + 1)Pn+1 tn (F.10)
n=0 n=0

and
∞
X ∞
X
(n + 1)Pn (x)tn+1 is equivalent to nPn−1 (x)tn (F.11)
n=0 n=0
and therefore we can write
(2n + 1)xPn (x) = (n + 1)Pn+1 (x) + nPn−1 (x) (F.12)
Equation F.12 is the recurrence relation for Legendre polynomials. It is a prescription
for expressing a given Pn (x) in terms of the adjacent polynomials in the series, Pn−1 (x)
244 Legendre Functions

and Pn+1 (x). For example, it is easy to remember the first two Legendre polynomials,
P0 (x) = 1, P1 = x. Using Eq. F.12 we find immediately P2 = 1/2(3x2 − 1). Then,
with P1 (x) and P2 (x) in hand we find P3 (x) = 1/2(5x3 − 3x). Thus,in principle any
member of the series may be found from the recurrence relation.

F.3 Parity
It appears by inspection that even-index Legendre polynomials are even and odd-index
polynomials are odd. In general the parity of the polynomials can be demonstrated by
recourse to the generating function, Eq. F.4.

g(t, x) = g(−t, −x)

∞
X X∞ ∞
X
Pn (x)tn = Pn (−x)(−t)n = Pn (−x)(−1)n tn
n=0 n=0 n=0

Equating equal powers of tn we have

Pn (x) = (−1)n Pn (−x) (F.13)

which shows that in general even-index Legendre polynomials are even and odd-index
polynomials are odd in the x → −x parity operation.

F.4 Orthogonality and Normalization

The Legendre equation, Eq. F.1, can be regarded as an operator-eigenfunction-eigenvalue
equation, Ôψ − λψ = 0, where Ô the operator is
d d
Ô = (1 − x2 ) (F.14)
dx dx
The eigenfunctions ψ are the Legendre polynomials Pn (x) and the eigenvalues λ =
n(n + 1). Viewed in this way, it can be shown that the “operator” is hermitian and
therefore the eigenfunctions are orthogonal and the eigenvalues real. Therefore the
Legendre polynomials are orthogonal
Z 1
Pn (x)Pm (x) dx = 0 m 6= n (F.15)
−1

and in fact they form a complete set, spanning the space of x from -1 to +1. The
normalization factor comes from the determination of the integral when m = n. We
start with the generating function, square it and write
" ∞
#2
X
(1 − 2tx + t2 )−1 = Pn (x)tn (F.16)
n=0

Then we integrate both sides and note that all the Pn Pm cross terms will vanish by
virtue of Eq. F.15.
 Orthogonality and Normalization 245

Z 1 ∞ Z 1
dx X 2
= [Pn (x)] dx t2n (F.17)
−1 1 − 2tx + t2 n=0 −1

The integral on the left can be readily evaluated by making a change of variable,
y = 1 + 2t + t2 and dy = −2tdx. Making the substitution in Eq. F.17, together with
the change in limits x = −1 → y = (1 + t)2 and x = 1 → y = (1 − t)2 we have
Z 1   ∞ Z 1
dx 1 1+t X 2
2
= ln = [Pn (x)] dx t2n (F.18)
−1 1 − 2tx + t t 1−t n=0 −1

Now the ln expression can be expanded in a power series

∞
t2n
 
1 1+t X
ln =2 (F.19)
t 1−t n=0
2n + 1

and therefore
∞ ∞ Z 1
X t2n X 2
2 = [Pn (x)] dx t2n (F.20)
n=0
2n + 1 n=0 −1

Finally, equating each term in the power series t2n on each side of Eq. F.20, we have
Z 1
2 2
[Pn (x)] dx = (F.21)
−1 2n + 1

which is the normalization condition.

Using the conventional expression that summarizes orthonormality, we write
Z 1
2δmn
Pm (x)Pn (x) dx = (F.22)
−1 2n + 1
G
Properties of the Hermite
polynomials

The Hermite polynomials constitute the essential part of the solutions to the Schrödinger
equation for the one-dimensional harmonic oscillator. Their properties are the start-
ing point for interpreting vibrational motion and dipole transition selection rules in
diatomic molecules. The Hermite polynomials can be defined from the “Rodrigues
representation”

dn 
Hn (x) = (−1)n exp x2 exp −x2


(G.1)
dxn
or by use of the generating function.

G.1 Generating Function and Recurrence Relations

The generating function is given by
∞
2 X tn
g(x, t) = e−t +2tx
= Hn (x) (G.2)
n=0
n!

from which we can obtain the recurrence relations similarly to the procedure followed
for the Legendre polynomials in Complement F. First we take the partial derivative of
g(x, t) with respect to t,
∞ ∞
∂g(x, t) 2 X ntn X tn−1
= 2(x − t)e−t +2tx = Hn (x) = Hn (x) (G.3)
∂t n=0
n! n=0
(n − 1)!

From which we find, after adjusting the summation indices,

∞ ∞ ∞
X tn X tn X tn
−2n Hn−1 (x) + 2x Hn (x) − Hn+1 (x) = 0 (G.4)
n=0
n! n=0
n! n=0 n!

Terms of equal powers of tn must be equal and the recursion relation is therefore

Hn+1 (x) = 2xHn (x) − 2nHn−1 (x) (G.5)

We can also get a recursion relation involving the derivative of the Hermite polynomials
by taking the partial derivative of the generating function with respect to x.
 Orthogonality and Normalization 247

Table G.1 Hermite polynomials

Index n Function Polynomial

0 H0 (x) 1
1 H1 (x) 2x
2 H2 (x) 4x2 − 2
3 H3 (x) 8x3 − 12x
4 H4 (x) 16x − 48x2 + 12
4

∞ ∞
∂g(x, t) 2 X tn X tn
= 2te−t +2tx = 2t Hn (x) = Hn0 (x) (G.6)
∂x n=0
n! n=0 n!

Then
∞ ∞ ∞
X tn+1 X tn X tn
2 Hn (x) = 2n Hn−1 (x) = Hn0 (x) (G.7)
n=0
n! n=0
n! n=0 n!
and
2nHn−1 (x) = Hn0 (x) (G.8)
We can get the first two members of the Hermite polynomials by expanding the gen-
erating function exponential in a Taylor series,
∞
n
2 X 2tx − t2
e−t +2tx = = 1 + (2tx − t2 ) + . . . (G.9)
n=0
n!

that results in H0 (x) = 1 and H1 (x) = 2x. Then from these and Eq. G.5, all the
polynomials up to arbitrary n can be calculated. The first few Hermite polynomials
are shown in Table G.1

G.2 Orthogonality and Normalization

G.2.1 Orthogonality
The Hermite polynomials themselves are not orthogonal in the sense that
Z ∞
Hn (x)Hm (x) dx 6= 0 m 6= n (G.10)
−∞

but a closely related integral is orthogonal,

Z ∞
Hn (x)Hm (x) exp −x2 dx = 0


m 6= n (G.11)
−∞

In fact we can define a new orthogonal function,

ϕn (x) = exp(−x2 /2)Hn (x) (G.12)
Now using the two recurrence relations, we can obtain a 2nd order differential equation
in the Hermite polynomials of index n in the following way. First, by substituting
Eq. G.8 into Eq. G.5 we eliminate the n − 1 index,
248 Hermite Polynomials

Hn+1 (x) = 2xHn (x) − Hn0 (x) (G.13)

Then differentiate the result
0
Hn+1 (x) = 2Hn (x) + 2xHn0 − Hn00 (x) (G.14)
Use Eq. G.8 again to eliminate the term with the n + 1 index,
0
Hn+1 (x) = 2(n + 1)Hn (x) (G.15)
Now regroup the terms in n
Hn00 (x) − 2xHn0 (x) + 2nHn (x) = 0 (G.16)
Differentiating Eq. G.12 and substituting into Eq. G.16 results in
ϕ00n (x) + (2n + 1 − x2 )ϕn (x) = 0 (G.17)
which is the equation for the one-dimensional quantum mechanical oscillator. There-
fore we see that the physically significant entities are not the Hermite polynomials per
se but the orthogonal functions, ϕn (x) = Hn (x) exp(−x2 /2).

G.2.2 Normalization
We find the normalization of the Hermite functions, Eq. G.12 by multiplying the gen-
erating function for Hm (x) with that for Hn (x), similar to the approach we take with
the squared generating function of the Legendre polynomials, Eq. F.16.
∞
X sm tn
exp(−s2 + 2sx) exp(−t2 + 2tx) = Hm (x)Hn (x) (G.18)
m,n=0
m!n!

Then multiply both sides by exp(−x2 ) and integrate over the interval from −∞ to
∞. Cross terms in the exp(−x2 )Hm (x)Hn (x) products drop out of the integral due to
orthogonality.
∞
(st)n ∞
X Z Z ∞
2 2
exp(−x ) [Hn (x)] dx = exp −(x − s − t)2 exp(2st) dx (G.19)
n=0
n!n! −∞ −∞

The first exponential term under the integral

√ on the right side is just the error function,
and the integral over it is known to be π. The remaining term, exp(2st), can be
expanded in a Taylor series and we have therefore
∞ ∞
(st)n ∞ √ X 2n (st)n
X Z
2
exp(−x2 ) [Hn (x)] dx = π (G.20)
n=0
n!n! −∞ n=0
n!

Now in the usual way we equate equal powers of (st)n that results in
Z ∞
2 √
exp(−x2 ) [Hn (x)] dx = 2n πn! (G.21)
−∞

Equation G.21 is the normalization expression for the Hermite functions.