Fundamentals of Electromagnetic Waves

Thomas E. Roth
Elmore Family School of Electrical and Computer Engineering
Purdue University

Spring 2023
ii
Preface

This set of lecture notes was prepared for a junior level course on the fundamentals of
electromagnetic waves that is offered each semester in the Elmore Family School of Electrical
and Computer Engineering at Purdue University. The class was originally created by Prof.
Kevin Webb at Purdue. The class serves as a follow-up to a first course on electromagnetic
physics that primarily covers electrostatics and magnetostatics, with a little time left at the
end of the semester to begin discussing the very basics of plane waves. The second course on
electromagnetic waves covers a wide range of topics, beginning from time-varying Maxwell’s
equations and then developing in depth the fundamentals of plane waves. After that, we
cover in detail transmission lines before discussing the field theory treatment of waveguides
and resonators. We finally wrap up the course by discussing the basics of antenna theory. In
addition to these lecture notes, we use David K. Cheng’s Field and Wave Electromagnetics
(2nd edition) as a required textbook for the course. In preparing these lecture notes, I drew
from Cheng’s textbook, David Pozar’s Microwave Engineering (4th edition), Constantine
Balanis’s Antenna Theory: Analysis and Design (3rd edition), and Jian-Ming Jin’s Theory
and Computation of Electromagnetic Fields (2nd edition). Due to the well-established nature
of the material covered, I make no attempt to exhaustively include citations throughout these
lecture notes.
Looking through the table of contents of these lecture notes, one will quickly see that our
ordering of topics does not follow a typical progression found in most textbooks. Although
it leads to an unusually long and broad first chapter, we follow this order in the class to
balance the content presented to students with the goal of keeping the class more engaging.
In particular, we intersperse fundamental electromagnetic topics (e.g., boundary conditions,
Poynting’s theorem, material models etc.) with discussions of plane waves or other more
applied topics that lend themselves to introducing how electromagnetics plays a role in a
wide range of practical technologies. This also allows us to cover topics in a sequence such
that each weekly homework assignment can have more meaningful problems that build the
desired mathematical and problem-solving skills we wish to impart to our students. We have
found that most students really enjoy this sequence and stay more engaged throughout the
semester as a result.

iii
iv
Contents

1 Fundamentals of Electromagnetics and Plane Waves 1

1.1 Maxwell’s Equations and Electromagnetic Quantities . . . . . . . . . . . . . 1
1.2 Vector Calculus Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Integral Form of Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Free-Space Wave Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Maxwell’s Equations in the Frequency Domain . . . . . . . . . . . . . . . . . 17
1.5.1 Frequency Domain Wave Equation . . . . . . . . . . . . . . . . . . . 22
1.6 Introduction to Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.7 Boundary Conditions for Maxwell’s Equations . . . . . . . . . . . . . . . . . 27
1.7.1 Curl Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.7.2 Divergence Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.7.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.8 Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.8.1 Polarization Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.8.2 Magnetization Intensity . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.9 Plane Waves in Lossy Media . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.9.1 Useful Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.10 Poynting’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.10.1 Time Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.10.2 Frequency Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.10.3 Power Flow of a Plane Wave . . . . . . . . . . . . . . . . . . . . . . . 46
1.11 Radiation from Current Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.12 Plane Wave Scattering from a PEC Region (Normal Incidence) . . . . . . . . 53
1.13 Lorentz Force and Material Response . . . . . . . . . . . . . . . . . . . . . . 57
1.13.1 Plasma Material Response . . . . . . . . . . . . . . . . . . . . . . . . 57
1.13.2 Other Material Response Models . . . . . . . . . . . . . . . . . . . . 60
1.14 Polarization of Electromagnetic Waves . . . . . . . . . . . . . . . . . . . . . 60
1.14.1 Introduction to Polarization . . . . . . . . . . . . . . . . . . . . . . . 60
1.14.2 Applications of Polarized Waves . . . . . . . . . . . . . . . . . . . . . 65
1.15 Doppler Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
1.15.1 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
1.15.2 Doppler Radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
1.15.3 Introduction to Synthetic Aperture Radar . . . . . . . . . . . . . . . 72
1.16 Dispersion and Group Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 73

v
CONTENTS

1.17 Plane Waves Propagating in Arbitrary Directions . . . . . . . . . . . . . . . 77

1.18 Dispersion Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
1.18.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
1.18.2 Anisotropic Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
1.19 Plane Wave Scattering from a PEC Sheet (Oblique Incidence) . . . . . . . . 85
1.19.1 Geometry Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
1.19.2 Perpendicular Polarization Case . . . . . . . . . . . . . . . . . . . . . 89
1.19.3 Parallel Polarization Case . . . . . . . . . . . . . . . . . . . . . . . . 91
1.20 Normal Incidence at a Homogeneous Half Space . . . . . . . . . . . . . . . . 93
1.21 Normal Incidence at Multiple Planar Interfaces . . . . . . . . . . . . . . . . 97
1.22 Oblique Incidence at a Homogeneous Half Space . . . . . . . . . . . . . . . . 99
1.22.1 Perpendicular Polarization . . . . . . . . . . . . . . . . . . . . . . . . 100
1.22.2 Parallel Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
1.22.3 Comparison of Parallel and Perpendicular Polarizations . . . . . . . . 104
1.22.4 Special Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
1.23 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

2 Transmission Lines 115

2.1 Introduction to Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . 115
2.2 Telegrapher’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
2.2.1 Lossless Line Wave Equation . . . . . . . . . . . . . . . . . . . . . . . 118
2.2.2 Lossy Line Wave Equation . . . . . . . . . . . . . . . . . . . . . . . . 120
2.2.3 Computing R, L, G, and C . . . . . . . . . . . . . . . . . . . . . . . . 120
2.3 Terminated Lossless Transmission Lines . . . . . . . . . . . . . . . . . . . . . 123
2.3.1 Derivation of Reflection Coefficient . . . . . . . . . . . . . . . . . . . 123
2.3.2 Other Characterizations of Reflections . . . . . . . . . . . . . . . . . 125
2.3.3 Input Impedance and Impedance Transformation . . . . . . . . . . . 127
2.3.4 Special Cases of Lossless Terminated Lines . . . . . . . . . . . . . . . 129
2.4 Quarter-Wave Transformer Matching . . . . . . . . . . . . . . . . . . . . . . 131
2.5 Generator and Load Mismatches . . . . . . . . . . . . . . . . . . . . . . . . . 134
2.6 Power Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.7 Transmission Line Ports and Impedances . . . . . . . . . . . . . . . . . . . . 141
2.7.1 Voltage and Current – A Closer Look . . . . . . . . . . . . . . . . . . 141
2.7.2 One-Port Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
2.7.3 Summary of Impedances . . . . . . . . . . . . . . . . . . . . . . . . . 145
2.8 Introduction to Port Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 146
2.8.1 Impedance Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 147
2.8.2 Admittance Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.8.3 Scattering Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.8.4 Scattering Parameters of Common Components . . . . . . . . . . . . 153
2.9 Introduction to the Smith Chart . . . . . . . . . . . . . . . . . . . . . . . . . 158
2.10 Single-Stub Impedance Matching . . . . . . . . . . . . . . . . . . . . . . . . 161
2.10.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
2.11 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

vi
 CONTENTS

3 Waveguides and Resonators 171

3.1 TEM, TE, and TM Waveguide Modes . . . . . . . . . . . . . . . . . . . . . 171
3.1.1 General Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
3.1.2 TEM Waves, Ez = 0, Hz = 0 . . . . . . . . . . . . . . . . . . . . . . 173
3.1.3 TE Waves, Ez = 0, Hz ̸= 0 . . . . . . . . . . . . . . . . . . . . . . . . 174
3.1.4 TM Waves, Ez ̸= 0, Hz = 0 . . . . . . . . . . . . . . . . . . . . . . . 175
3.2 Rectangular Waveguide Modes . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.2.1 TE Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.2.2 TM Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
3.2.3 Examples of Waveguide Modes . . . . . . . . . . . . . . . . . . . . . 183
3.2.4 Practical Waveguide Operation . . . . . . . . . . . . . . . . . . . . . 183
3.3 Introduction to Resonators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
3.3.1 Lumped Element Resonators: A Short Review . . . . . . . . . . . . . 185
3.3.2 Transmission Line Resonators . . . . . . . . . . . . . . . . . . . . . . 186
3.3.3 Rectangular Waveguide Resonators . . . . . . . . . . . . . . . . . . . 188
3.4 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

4 Antennas 193
4.1 Introduction to Antenna Concepts . . . . . . . . . . . . . . . . . . . . . . . . 193
4.2 Electromagnetic Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
4.3 Potentials Produced by Known Source Distributions . . . . . . . . . . . . . . 199
4.4 Fields of an Infinitesimal Dipole . . . . . . . . . . . . . . . . . . . . . . . . . 203
4.4.1 Field Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.5 Fundamental Parameters of Antennas . . . . . . . . . . . . . . . . . . . . . . 206
4.6 Friis Transmission Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
4.7 Introduction to Antenna Arrays . . . . . . . . . . . . . . . . . . . . . . . . . 212
4.8 Practice Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

Bibliography 217

vii
CONTENTS

viii
Chapter 1

Fundamentals of Electromagnetics
and Plane Waves

1.1 Maxwell’s Equations and Electromagnetic Quanti-

ties
Maxwell’s equations (completed in 1865) describe the fundamental properties and interplay
between electricity and magnetism. A number of the principles behind these equations
were originally introduced by other researchers. Maxwell took these experimental results
and theoretical concepts and developed a comprehensive and unifying summary of these
disparate concepts.
Maxwell’s equations represent one of the greatest triumphs ever achieved in physics. They
are for almost all intents and purposes, perfect. In 2012, their theoretical accuracy had been
experimentally validated to one part in a trillion. This level of accuracy is equivalent to
measuring the distance from the Earth to the Moon and being correct to within the width
of a single human hair [1]. Many other disciplines in physics can only dream of having
foundational equations that are this accurate or well-behaved. For instance, in the field of
fluid dynamics, the Navier-Stokes equations are of practical interest for many applications
ranging from modeling weather to complex aerodynamics. Even though they have had great
success in practical applications, the proof of existence and smoothness of solutions to the
Navier-Stokes equations are still considered one of the most important open problems in
mathematics. Such is the importance that the Clay Mathematics Institute has offered a
$1M USD bounty for anyone who can solve it!
Further, many areas of physics can only develop equations with various approximations
in place – i.e., they are only valid for certain situations and uses, making their application
difficult for the many “boundary cases” that inevitably occur in practical engineering analysis
and design. In the field of electromagnetics we rarely have to be concerned about this, we
can always fall back on the validity of Maxwell’s equations to know that we have firm footing
to investigate further concepts, develop numerical solutions, etc. We are truly privileged and
indebted to the great work of Maxwell and others who were so successful in developing the
theory of electromagnetism!
With this in mind, we turn our attention to Maxwell’s equations.

1
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Maxwell’s Equations
Ampere′ s Law : ∇ × H = ∂t D + J (1.1)
Faraday′ s Law : ∇ × E = −∂t B (1.2)
Gauss′ Law of Electricity : ∇·D=ρ (1.3)
Gauss′ Law of Magnetism : ∇·B=0 (1.4)

These equations relate the following electromagnetic quantities to each other.

ˆ E(r, t): electric field intensity [V/m]

ˆ H(r, t): magnetic field intensity [A/m]

ˆ D(r, t): electric flux density [C/m2 ]

ˆ B(r, t): magnetic flux density [T = Wb/m2 ]

One of the fundamental contributions of Maxwell was generalizing these equations to suggest
that a time-varying electric flux density can produce a magnetic field in the same way that
a time-varying magnetic flux density was known to produce an electric field. As we will
see, this relationship is what allows for electromagnetic waves to propagate, and thus allows
for many kinds of technologies to exist (e.g., long-distance wireless communication, radar,
etc.). In addition to electric and magnetic fields/fluxes being able to produce one another,
there are also two additional sources in Maxwell’s equations that can produce electric and
magnetic fields.
ˆ J: electric current density [A/m2 ]

ˆ ρ: electric charge density [C/m3 ]

As one would expect, these quantities are not independent of each other. This fact is
codified in the current continuity equation. The current continuity equation is a statement
of the fact that charges are conserved. In differential form, this is

∇ · J = −∂t ρ. (1.5)

Intuitively, we can think of the divergence operation as helping us to measure the current
density that is entering/exiting any point in space (more on this later). This must always be
balanced out by changes in the amount of charge density at a particular location to ensure
that charge is conserved.
Continuity equations are usually related to conservation laws, which play a very impor-
tant role in many areas of physics. They usually take a similar kind of form to the current
continuity equation. For instance, conservation laws play a very central role in the theories
of special and general relativity. In those settings, these concepts often come about when
considering stress-energy (also called energy-momentum) tensors which are useful quantities
to help us “track” how energy flows/transfers through a system.
Before moving on, it is good to take a closer look at the units of E, H, D, and B.
From this, we see that the electric and magnetic fields each have a unit of 1/m, while

2
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

the electric and magnetic fluxes each have a unit of 1/m2 . As we become more and more
acquainted with these quantities, we can often have a tendency to treat the fields and fluxes
as interchangeable. However, it is always good to remember that there are some fundamental
(and important) differences, as suggested by the units. In particular, we should always think
of E and H as quantities that are “built” to be integrated along a line/curve, while D and
B are “built” to be integrated along surfaces.
Even very talented researchers have fallen into the trap of treating the fields and fluxes
as interchangeable time and time again! There are many instances throughout the history of
computational electromagnetics (a field devoted to numerically solving Maxwell’s equations
with computers) where the researchers forgot this core principle. These early numerical
methods would produce “spurious” or erroneous solutions that were sometimes difficult to
properly diagnose. However, in many cases, the issue was that E and H weren’t being
treated as quantities “built” to be integrated along a curve, while D and B weren’t being
treated as quantities “built” to be integrated over a surface. Once these quantities were
treated appropriately, all these spurious numerical issues vanished!
Now, if we take a closer look at Maxwell’s equations, we can see that at this point we
have two vector equations and two scalar equations. This is equivalent to having 8 equations
that we could try and solve (each vector equation can actually be considered to be “worth” 3
scalar equations). However, we need to compute 4 vector quantities (E, H, D, and B), which
implies that we need 12 equations to fully specify Maxwell’s equations. So what is missing?
The constitutive relations are our answer. These specify additional interrelationships between
E, H, D, and B in the presence of simple forms of matter.
For simple materials (we will be more specific on what this means later), D can be related
to E, while B can be related to H. The particular relationships are

D = ϵE, (1.6)
B = µH, (1.7)

where ϵ is the permittivity of the material (with units of [F/m]) and µ is the permeability of the
material (with units [H/m]). Recall our discussion on units and the distinction in character
between the fields and fluxes (in terms of what they should be integrated over). We see that
even though these constitutive relationships appear very simple, they are fundamentally
changing the character of the fields into fluxes (and, in some sense, vice-versa). Looking at
the units we also see that ϵ will augment/contribute to capacitive effects, while µ will do the
same for inductive effects.
In free space (i.e., a vacuum), the permittivity and permeability are denoted as ϵ0 ≈
8.854 × 10−12 [F/m] and µ0 = 4π × 10−7 [H/m], respectively (Note: the subscript 0 is gen-
erally used in electromagnetics to denote that the quantity being considered corresponds
to its free space value). Since these are rather inconvenient numbers that are difficult to
remember, we typically normalize permittivities and permeabilities of different materials by
the corresponding free space value. These normalized quantities are called the relative per-
mittivity and relative permeability, and are denoted as ϵr and µr , respectively. This leads to
much simpler values to remember, like that Teflon has a relative permittivity of about 2.1,
as opposed to Teflon has a permittivity of 2.1 × 8.854 × 10−12 (whatever that ugly number
is). Before continuing, it is worth mentioning that many different terminologies are used for

3
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

these quantities in practice. For instance, it is very common to refer to the relative permit-
tivity (ϵr ) as the dielectric constant of a material. We prefer to not use this terminology,
since, as we will learn later, the relative permittivity is never truly a “constant”.
The constitutive relationships written in (1.6) and (1.7) are only valid for special kinds
of simple (albeit extremely prevalent) types of materials. In particular, this is for isotropic
materials. This means that the permittivity/permeability are the same regardless of which
direction they are measured in. There are special kinds of materials, called anisotropic
materials, where this is not the case. Simple examples of where anisotropic materials are used
include in polarized sunglasses and liquid crystal displays. However, there are more exotic
cases, such as for metamaterials, which are synthetic materials designed to have advantageous
electromagnetic properties not typically available in nature. For now, we won’t worry about
the consequences of anisotropic materials, but we will revisit this briefly at a later point in
the semester.
It is good to take stock at this point to see if we have enough equations to fully specify our
electromagnetic system. If we add everything up (including constitutive relations), we now
appear to have 4 vector equations and 2 scalar equations, making for a total of 14 equations
with 12 unknowns to solve. Did we go too far, have we now over-specified everything? It
turns out that the issue is in us naively treating the two Gauss’ laws as being independent
from the other Maxwell’s equations. We can actually show that the two Gauss’ laws can be
derived from Ampere’s and Faraday’s laws (see the practice problems). So why then do we
write them as separate equations? The answer is that these equations are only derivable for
time-varying systems. Since we want our Maxwell’s equations to also describe electrostatic
and magnetostatic systems, it is best to keep Gauss’ laws in our definition of what we consider
to be “Maxwell’s equations”. However, this does illustrate the important point that we do
not always need to try and independently solve all of Maxwell’s equations. Sometimes, we
can get by with just considering the constitutive relations and a few other key parts of
Maxwell’s equations.
Another kind of constitutive relationship that exists is Ohm’s law. You are no doubt
very familiar with Ohm’s law from your circuit theory classes. However, as we will learn
throughout this course, circuit theory at a fundamental level is a (very useful) simplification
of Maxwell’s equations. In the full electromagnetic picture, Ohm’s law is

J = σE (1.8)

In this equation, σ is the conductivity of the material. Typically, a unit of [S/m] is used
for the conductivity. Again note how the units taking part in this constitutive relation is
fundamentally changing the character of the electric field.

1.2 Vector Calculus Review

Vector calculus plays a very important role in the study of Maxwell’s equations. Because
of this, we will want to develop intuition for many of the operations we will use repeatedly
throughout the semester. Many of these concepts can feel abstract while first learning
them in a mathematics class, but studying them in the context of an applied theory like
electromagnetism can help bring the material to life. With that in mind, let’s review some

4
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.1: Illustration of the gradient. The colors denote the value of the scalar field and
the arrows denote the vector field produced by taking the gradient of the underlying scalar
field (image from Wikipedia [2]).

of the most important pieces of vector calculus. To help build intuition, we will only focus
on the definitions of these quantities appropriate for Cartesian coordinates. It is essential
to remember that these formulas cannot be naively applied to other coordinate systems; in
general, that will not work!
We will start with the most basic derivative in vector calculus, the gradient. The gradient
takes a differentiable scalar function and turns it into a vector field. For scalar function f (r),
the gradient operation is denoted as ∇f (r). The components of the vector field ∇f (r) (in
Cartesian coordinates) are
 
∂x f (r)
∇f (r) = ∂y f (r) (1.9)
∂z f (r)

The direction of the vector field ∇f is pointed toward where f is increasing or changing the
fastest. Hence, it helps us understand the “flow” of a scalar function. By taking the dot
product of ∇f with a unit vector, we can measure how quickly the scalar field is changing
along the direction specified by the unit vector. This is illustrated in Fig. 1.1.
To go along with the gradient, we have the gradient theorem. This tells us what happens
when we try and take the line integral of the gradient. You should recall from the definition
of a derivative that it involves dividing by the infinitesimal distance that we are computing
the change in a function over. Hence, we can think of it as carrying with it a change in units
of a function by a factor of 1/m. If you recall our discussion about the character of E and
H (i.e., they should be integrated along lines), we see that the gradient is taking a general
function and preparing it to be integrated along a line/curve. This is captured in a very
special way in the gradient theorem, which says
ˆ
∇f (r) · dℓ = f (r2 ) − f (r1 ) (1.10)
C

5
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.2: Illustration of the gradient theorem using a finite difference approximation to
the derivative.

where r1 (r2 ) are the beginning (ending) points of the curve C. This tells us that if we
integrate the gradient of a function, all we really need to consider is what happens at the
boundary of our integration range (in this case, the endpoints of the line). See Fig. 1.2 for
more details on how to interpret this result using finite difference approximations. This can
be very important in electromagnetism, and is certainly prevalent in electrostatics.
As we will see shortly, for every derivative we will frequently use in vector calculus there is
a matching integration theorem that tells us the only thing that we need to monitor is what
happens at the boundary of the integration range. This has many important consequences
throughout electromagnetism.
Next, we will consider the curl of a vector field. The curl of a vector field helps measure
the “rotation/circulation” of the vector field. It takes a vector field f and defines a new vector
field ∇ × f with magnitude equaling the maximum circulation of f and direction denoting
the axis of this maximum circulation. In Cartesian coordinates, the curl is
 
x̂ ŷ ẑ ∂y fz (r) − ∂z fy (r)
∇ × f (r) = ∂x ∂y ∂z = ∂z fx (r) − ∂x fz (r) (1.11)
fx (r) fy (r) fz (r) ∂x fy (r) − ∂y fx (r)
We see that each component (e.g., x̂-component) involves derivatives “measuring” the rate of
change around a closed loop in the plane orthogonal to the component (e.g., the (y, z)-plane).
This is more clearly seen by expanding the derivatives in a finite difference approximation,
as shown in Fig. 1.3.
To go along with the curl, we have the integration theorem known as Stokes’ theorem.
Stokes’ theorem is used to integrate the curl of a vector field over a surface S. It tells us
that this result only depends on what is happening on the boundary of S. Mathematically,
this is given as
¨ ˛
∇ × f (r) · dS = f · dℓ, (1.12)
S C

6
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.3: Illustration of one component of the curl using a finite difference approximation
for the derivatives. The portion of the vector field that aligns with the blue arrows will
accumulate into a positive value of the curl.

Figure 1.4: Illustration of Stokes’ theorem. The circulations from nearby points cancel with
one another, except over the boundary where there are not “matching” points to allow for
the cancellation to occur (image from [3]).

where C forms the boundary of S and ℓ defines the integration direction around C. The
intuitive picture is that at all adjacent interior points the derivatives involve terms with equal
magnitudes but opposite directions, making them cancel each other out. As a result, the
only location where a result can actually accumulate to a non-zero value is on the boundary
of the integration region where there are no adjacent points to cancel with. A rough picture
of this effect can be seen in Fig. 1.4.
Next up, the divergence of a vector. The divergence of a vector field helps measure the
“outward flux” of the vector field. It takes a vector field f and defines a new scalar field ∇ · f
with magnitude equaling the net flux exiting/entering the point. Positive magnitudes denote
flux moving out of the point of measurement. In Cartesian coordinates, the divergence is

∇ · f (r) = ∂x fx + ∂y fy + ∂z fz . (1.13)

7
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.5: Illustration of the divergence in two-dimensions using a finite difference approx-
imation. The portion of the vector field that aligns with the blue arrows will accumulate
into a positive value of the divergence.

An illustration of the divergence shown in Fig. 1.5 using finite differences helps illustrate
why this computes the “outward flux”.
As with the last two derivatives, there is a matching integration theorem. This is typically
referred to as the divergence theorem, but is also sometimes referred to as Gauss’ theorem
or Ostrogradsky’s theorem. The divergence theorem is used to integrate the divergence of
a vector field (typically over a volume). It tells us that this result only depends on what is
happening on the boundary of the integration region, for similar reasons to what was seen
with Stokes’ theorem. For integrating over a volume, the divergence theorem is given as
˚ ‹
∇ · f (r)dV = f · n̂dS, (1.14)
V S
where S forms the boundary of V and n̂ is the outward pointing unit normal vector. An
illustration of the derivatives calculated at interior points leading to cancellation in the
integration is shown in Fig. 1.6.
The trend should hopefully be clear now. Each kind of derivative we take in vector
calculus gives us a new quantity that is “built” to be integrated over some new kind of
region; be it a curve, a surface, or a volume. For instance, when we originally discussed E
and H we mentioned that they were built to be integrated along curves. However, they are
related to D and B in Faraday’s and Ampere’s laws, which we said should always be thought
of as being integrated over surfaces. How do we reconcile this? The answer is that the curl
operations in both of these equations change the characters of E and H so that they are
ready to be integrated over surfaces.
Before concluding our discussion on vector calculus, there are two last identities that are
important to remember. These are that


∇ × ∇f = 0, (1.15)

8
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.6: Illustration of the divergence theorem (image from [3]).

∇ · ∇ × f = 0. (1.16)

In words, we see that the gradient vector field always has no “circulation” and that the
circulation vector field always has no net “outward flux”. We can think of these derivatives
as measuring “orthogonal” rates of change, i.e., they are in some sense independent measure-
ments/specifications of how a quantity is changing. This is a very fundamental concept that
is typically encapsulated in what is known as the Helmholtz decomposition theorem. This
theorem is central to the use of electromagnetic potential functions that will be discussed
later in the course when we consider analyzing simple antennas.

1.3 Integral Form of Maxwell’s Equations

Armed with these vector calculus concepts, we can now use them to determine the integral
form of Maxwell’s equations. This form of Maxwell’s equations is useful for deriving certain
results such as the boundary conditions needed to solve the differential form of Maxwell’s
equations. As we will see, the integral form of Maxwell’s equations are also useful in deriving
circuit theory from electromagnetic theory.
To begin finding the integral form of Maxwell’s equations, let’s start with Ampere’s law.
From our previous discussions, we see that all the quantities in this equation are “built” to
be integrated over a surface. So let’s do exactly this! Integrating Ampere’s law over surface
S gives us
¨ ¨



∇ × H · n̂ dS = ∂t D + J · n̂ dS. (1.17)
S S

Looking at the left-hand side, we see that we can simplify this by using Stokes’ theorem.
The result is
˛ ¨


H · dℓ = ∂t D + J · n̂ dS. (1.18)
C S

We see that the magnetic field circulating a surface depends on the total current flowing
through that surface (where ∂t D is typically referred to as the displacement current).

9
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Next, let’s consider Faraday’s law. The same treatment with Stokes’ theorem can be
applied here to give us
˛ ¨
E · dℓ = − ∂t B · n̂ dS. (1.19)
C S

This tells us that the electric field around a loop will counteract changes in the magnetic
flux intersecting that loop. This effect can be leveraged to make many different kinds of
electromechanical machines.
Now let’s turn our attention to Gauss’ laws of electricity and magnetism. In both cases,
we are taking the divergence of the fluxes. Hence, these equations are giving us quantities
that should be integrated over a volume. So let’s do just that! This gives us
˚ ˚
∇ · D dV = ρ dV (1.20)

for Gauss’ law of electricity. Looking at the left-hand side, we see that we can simplify this
using the divergence theorem. This gives
‹ ˚
D · n̂ dS = ρ dV. (1.21)
S

In words, we see that the total flux exiting some volume depends on the total amount
of charge contained within that volume. Following an identical process for Gauss’ law of
magnetism gives us
‹
B · n̂ dS = 0. (1.22)
S

This equation tells us two (related) things. First, because there are no magnetic charges the
flux leaving a closed surface will always be exactly balanced out by an equal amount of flux
entering the surface. Stated another way, this tells us that B always forms closed loops.
In summary, the integral form of Maxwell’s equations are the following.

Maxwell’s Equations (Integral Form)

˛ ¨
′


Ampere s Law : H · dℓ = ∂t D + J · n̂ dS (1.23)
˛C ¨
S

Faraday′ s Law : E · dℓ = − ∂t B · n̂ dS (1.24)

‹C ˚
S

Gauss′ Law of Electricity : D · n̂ dS = ρ dV (1.25)

‹S
Gauss′ Law of Magnetism : B · n̂ dS = 0 (1.26)
S

We can also follow similar steps for the current continuity equation. By integrating this
over a volume and using the divergence theorem, we get
‹ ˚
J · n̂ dS = − ∂t ρ dV. (1.27)
S

10
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.7: Illustration of Kirchoff’s current law. For simple circuits, the integration of the
current densities reduces to a summation of the total current flowing on the wires, which
makes the current continuity equation reduce to Kirchoff’s current law.

We can relate this to circuit theory by integrating around a “circuit node” where there are
only perfect wires connected to the node. In this case, no charge can build up at the node,
so we have that the sum of all the currents entering/exiting the node will be equal to 0. This
is Kirchoff ’s current law, which is illustrated in Fig. 1.7.
Similarly, Kirchoff ’s voltage law can be inferred from Maxwell’s equations. This is done
by applying Faraday’s law around the loop of a circuit that doesn’t have a time-varying
magnetic flux density intersecting it. This is illustrated in Fig. 1.8. Although this leads to a
simple result, it should be recognized that it is an idealization of Faraday’s law by assuming
there is no time-varying magnetic flux intersecting the loop. In reality, many electrical
components can produce time-varying magnetic fluxes (intentionally or unintentionally) that
will intersect “loops” in practical circuits. This induces unexpected voltages that can cause
a circuit to function incorrectly. Determining the sources of this kind of interference and
mitigating the issues is typically referred to as the field of electromagnetic compatibility
(EMC) or electromagnetic interference (EMI).
Next, we can understand the I-V relationships for inductors and capacitors through
Maxwell’s equations. This is done by applying Faraday’s law around a closed loop that
contains an inductor. The total flux going through the inductor is related to the inductance
by definition. In particular, we have that L = Φtot /i. We can compute the total flux passing
through the inductor by integrating the B that passes through the surface of the inductor
as
¨
Φtot = B · n̂dS. (1.28)
S

This result can then be substituted into Faraday’s law to see the expected I-V relationship.
See Fig. 1.9 for more details.
Finally, we can understand the I-V relationship for a capacitor in a similar way. By
applying Ampere’s law at various points of the circuit, we can see that the displacement

11
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.8: Illustration of Kirchoff’s voltage law. For simple circuits with ideal wires, the
only voltage changes occur through circuit components. In this simplified setting, Faraday’s
law becomes Kirchoff’s voltage law when it is assumed there is no time-varying magnetic
flux intersecting the circuit loop.

Figure 1.9: Illustration of the I-V relationship for an inductor.

current flowing through the capacitor must equal the current flowing through the circuit.
From this we can infer the expected I-V relationship for a capacitor. This is most easily done
for the simple case of a parallel plate capacitor. In this case, we know that the capacitance will
approximately be C = ϵA/d, where A is the area of the capacitor plates and d is the distance
between the two plates. If we ignore fringing fields, we know from electrostatic theory that
the magnitude of the electric field in between the two plates should be approximately constant
and equal to |E| = vc /d, where vc is the voltage across the capacitor. Using the constitutive
relationship of D = ϵE and that the current flowing through the middle of the capacitor
must be equal to
¨
i= ∂t D · n̂dS (1.29)
S

due to Ampere’s law, we can determine that if S covers the entire area of the parallel plates
(and D is constant over the area) that
 
ϵvc
i = ∂t A = C∂t vc . (1.30)
d
See Fig. 1.10 for more details.

12
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.10: Illustration of the I-V relationship for a capacitor.

1.4 Free-Space Wave Equation

We can now go about deriving the wave equation for the electric and magnetic fields in free
space. Originally, Maxwell was able to determine the wave equation because he had added
the displacement current term to Ampere’s law. Without this modification, propagating
electromagnetic waves would not be able to exist.
To begin deriving the wave equation, we will rewrite Ampere’s and Faraday’s laws using
the constitutive relations. This will allow us to combine the two coupled 1st-order partial
differential equations (PDEs) to arrive at 2nd-order PDE. It is this 2nd-order PDE that
clearly shows evidence of wave propagation. With this in mind, Ampere’s law becomes
∇ × H = ∂t D
(1.31)
= ∂t (ϵ0 E)

and Faraday’s law becomes

∇ × E = −∂t B
(1.32)
= −∂t (µ0 H).

We want to combine these two equations into a single equation. To do this, we need to make
it so that we can substitute one of the equations into the other one. This will be possible by
taking the curl of (1.32), which gives

∇ × ∇ × E = −µ0 ∇ × ∂t H. (1.33)

Note that because µ0 does not vary as a function of position, we were able to bring it
“outside” of the curl operation. Next, we will swap the order of the temporal and spatial
derivatives on H. This is safe to do in this case because our “time” and “space” variables are
independent. However, this “swapping” of the order of second-order (or higher) derivatives

13
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

is generally acceptable under fairly easy requirements of the function being differentiated.
Doing this, we have


∇ × ∇ × E = −µ0 ∂t ∇ × H . (1.34)

We can now substitute the righ-hand side of (1.31) into this equation for ∇ × H. This gives

∇ × ∇ × E = −µ0 ϵ0 ∂t2 E. (1.35)

Moving everything to one side of the equation gives us the vector wave equation for the
electric field. This is

∇ × ∇ × E + µ0 ϵ0 ∂t2 E = 0. (1.36)

We can follow a very similar derivation to get the vector wave equation for the magnetic
field as well. This is

∇ × ∇ × H + µ0 ϵ0 ∂t2 H = 0. (1.37)

Since we are in a source-free region, we can use some more vector calculus identities to
rewrite these equations into another form. In particular, we note that

∇ × ∇ × f = ∇ ∇ · f − ∇2 f ,


(1.38)

where ∇2 = ∇ · ∇ (i.e., the divergence of the gradient of the vector) is the vector Laplacian
operator. Using this in (1.36), we have

∇ ∇ · E − ∇2 E + µ0 ϵ0 ∂t2 E = 0.


(1.39)

Has this helped us out at all? Well, not yet. However, if we note that Gauss’ law of electricity
in a source-free region of free space is

ϵ0 ∇ · E = 0, (1.40)

we can further simplify our wave equation. Using (1.40) in (1.39), we have

∇2 E − µ0 ϵ0 ∂t2 E = 0. (1.41)

This is the vector Helmholtz wave equation. It is traditional for wave equations to define
the square of the propagation speed as the inverse of the constants multiplying the time
derivative term. Here, this gives us the speed of light, denoted as
1
c0 = √ ≈ 3 × 108 [m/s]. (1.42)
µ0 ϵ0

In our current digital age, there are very few numbers that I will recommend for you to mem-
orize. However,this is certainly one of them. Another good one to note is that c0 ≈ 1 [ft/ns].
We will examine this equation in greater detail shortly, and will spend a significant amount of

14
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

time understanding the properties of solutions to this equation for different electromagnetic
systems.
So how do we go about interpreting the wave equation (1.41)? The first item to tackle
is understanding what the Laplacian operator does when it is applied to a vector quantity.
However, before we jump to that, let’s first recall how it works with a scalar function. In
Cartesian coordinates, the Laplacian operator applied to a scalar function takes the following
form:

∇2 f (r) = ∇ · ∇f (r) .
 
(1.43)

That is, we first take the gradient of f and then take the divergence of the result. This
evaluates to

∇2 f (r) = ∂x2 f (r) + ∂y2 f (r) + ∂z2 f (r). (1.44)

If we recall our interpretations of the gradient and divergence operations, we see that in
words the Laplacian operator is computing the net outward flux of the gradient flow of f . It
is also a higher order derivative. These occur very commonly in physics to describe things
like diffusion. Further, if you recall the Taylor series of a function, we can associate with the
second derivative as being useful in establishing a “higher order” understanding of how the
function f is changing around the point we are evaluating our derivatives at. Understanding
quantities through their “curvature” is a hugely fundamental concept that occurs in many
areas of modern physics.
Now that we have a better feel for what the Laplacian means when applied to a scalar
function, we can turn our attention to using it on a vector function. However, we are
immediately hit with needing to compute the gradient of a vector function. Is this allowed?
If so, what does it even mean and what does the result look like? It turns out that this is
allowed, and we go about computing it by taking the gradient of each of the components
of the vector function. Considering this, the output of this operation is 3 vector quantities
that are all related to each other. We need a good way to handle the bookkeeping of this
quantity to not make mistakes or make our notation too cumbersome. Traditionally, this
has been done using dyads. In the same way that we sometimes represent vector functions
using a column vector notation, dyads can be represented as a matrix of values. Hence, the
gradient of a vector can be represented as
 
∂x fx (r) ∂x fy (r) ∂x fz (r)
∇f (r) = ∂y fx (r) ∂y fy (r) ∂y fz (r) . (1.45)
∂z fx (r) ∂z fy (r) ∂z fz (r)
Successfully manipulating dyads and performing calculus on them requires learning a number
of new rules. This approach has fallen out of favor and is becoming a bit of a dying art in
modern times for a number of reasons. One main reason is that better methods were needed
to handle generalizations of dyads (called tensors) that are important in studying various
areas of modern physics, such as general relativity. You will occasionally come across dyads
still being used in some limited research fields, and occasionally they will still be used in
some graduate level electromagnetics courses. However, we favor leaving them as a dying
art, so we will not dwell on them in detail.

15
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Instead, we will just note the following simple form of the vector Laplacian that occurs in
Cartesian coordinates (and only Cartesian coordinates! ). In this case, the Laplacian operator
applied to a vector function takes the following simple form:
 2 
∇ fx (r)
∇2 f (r) = ∇2 fy (r) . (1.46)
∇2 fz (r)
From this, we see that the vector Laplacian operator simply is the scalar Laplacian operator
applied to each component of the vector. However, it cannot be emphasized enough that
this simple form only applies in Cartesian coordinates!
Now that we understand the vector Laplacian operator better, let’s revisit the vector
Helmholtz wave equation. This was
1 2
∇2 E − ∂ E = 0. (1.47)
c20 t
Considering the complexity of using the vector Laplacian operator, this form of the wave
equation is typically only useful for understanding simple wave solutions. So let’s consider
one of these now. In particular, let’s assume that we can represent E(r, t) as
E(r, t) = x̂Ex (z, t). (1.48)
This electric field only varies as a function of z and t, and has only a single vector direction
(we say that is it polarized along the x-direction). In this case, the vector Helmholtz equation
reduces down to the much simpler scalar Helmholtz equation. That is,
1 2
∇2 Ex (z, t) − ∂ Ex (z, t) = 0. (1.49)
c20 t
Considering that Ex only varies with z in space, we can further simplify this to be
1 2
∂z2 Ex (z, t) − ∂ Ex (z, t) = 0. (1.50)
c20 t
What will a solution to this need to look like? Well, we can first of all obviously say that
it will need to be twice-differentiable for the derivatives to be able to be correctly applied.
Next, we see that whatever the spatial derivative does to Ex is going to need to be very
carefully balanced out by what the temporal derivative does to Ex . More formally, we can
likely expect that the z and t variables will enter into Ex in the same functional form.
Considering these points, let’s assume for now that Ex (z, t) actually takes the form of
Ex (a1 t±a2 z), where a1 and a2 are some constants that we want to determine. Let’s substitute
this proposed solution form into (1.50) and see what we get. At first, we have
1 2
∂z2 Ex (a1 t ± a2 z) = ∂ Ex (a1 t ± a2 z), (1.51)
c20 t
which isn’t immediately useful. However, we can use a simple trick with the chain rule to
recognize that
′ d  ′
∂z Ex (a1 t ± a2 z) = Ex (a1 t ± a2 z) a1 t ± a2 z = ±a2 Ex (a1 t ± a2 z), (1.52)
dz

16
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

where the ′ symbol is being used to denote a derivative with respect to the entire argument
of a function, which is a1 t ± a2 z in this case. Similarly, we see that
′′
∂z2 Ex (a1 t ± a2 z) = a22 Ex (a1 t ± a2 z). (1.53)

Likewise, we see that

′′
∂t2 Ex (a1 t ± a2 z) = a21 Ex (a1 t ± a2 z). (1.54)

If we substitute (1.53) and (1.54) into (1.51), we now have

′′ a21 ′′
a22 Ex (a1 t ± a2 z) = E (a1 t ± a2 z). (1.55)
c20 x

We can cancel out the identical terms to see that this functional form of a solution will work
so long as

a21
c20 = . (1.56)
a22

It is customary to choose a1 = 1 and a2 = 1/c0 , so that our functional form of a solution to

the wave equation becomes
 
z
E(r, t) = Ex t ± . (1.57)
c0

This solution exhibits a simple form of wave propagation. If we advance t to t + ∆t, we

see that the same point of the function Ex can be reached by correspondingly adjusting our
z location to z ∓ ∆z if ∆z = c0 ∆t. We say that in time ∆t the wave has moved/advanced
by a distance of c0 ∆t. This is why c0 is referred to as the speed of light. Further, we see that
Ex (t − z/c0 ) can be associated with waves traveling in the +z-direction and that Ex (t + z/c0 )
can be associated with waves traveling in the −z-direction.

1.5 Maxwell’s Equations in the Frequency Domain

In general, solving Maxwell’s equations in the time domain is a very difficult undertaking.
It is a good rule of thumb that each time we increase the number of dimensions in a partial
differential equation we are trying to solve analytically the problem becomes exponentially
harder. Due to this, it is very common that we will try and find useful ways to simplify
the problems we are studying to make them more tractable. An excellent example of this is
provided by time-harmonic analysis. You should already be familiar with using the Fourier
transform to simplify the analysis of problems in your AC circuit theory and linear systems
courses. In your AC circuits courses it was typically used to convert a differential equation
into an algebraic one that could be solved much more readily. Similarly, your linear systems
courses should have taught you how we can build significant intuition into how linear, time-
invariant systems behave by “probing” them with sinusoidal functions. In electromagnetics,

17
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

we can use a combination of these concepts to greatly simplify the study of many practical
problems.
To begin, let’s recall some basic details about the Fourier transform. The Fourier trans-
form provides us with a way to express a time domain signal in terms of an “infinite sum-
mation” of sinusoidal functions (and vice-versa). If you recall your linear algebra theory,
the Fourier transform can be predominantly viewed as a kind of basis transformation. The
particular form of the Fourier transform that we will use in this course is
ˆ ∞
f (ω) = F{f (t)} = f (t)e−jωt dt. (1.58)
−∞

We can interpret this integration as an inner product of our function f (t) with a particular
Fourier harmonic, namely, ejωt . Hence, we are measuring how much of f (t) “overlaps” with
the particular Fourier harmonic of interest.
We can invert our Fourier transform using a similar transform. In particular, this will be
ˆ ∞
−1 1
f (t) = F {f (ω)} = f (ω)ejωt dω, (1.59)
2π −∞
which can be interpreted as an “infinite summation” of all the different Fourier harmonics
weighted by f (ω) to properly reproduce f (t).
One of the most important properties of the Fourier transform for simplifying differential
equations is how a derivative of a Fourier transform variable changes. To see this, let’s take
the time derivative of f (t) as expressed through the inverse Fourier transform written above.
This is
ˆ
d 1 d ∞
f (t) = f (ω)ejωt dω
dt 2π dt −∞
ˆ ∞ (1.60)
1
= jωf (ω)ejωt dω.
2π −∞
Hence, we see that

d
⇐⇒ jω (1.61)
dt
where the ⇐⇒ denotes how these quantities transform under the action of the Fourier and
inverse Fourier transforms.
Let’s see how we can use this to simplify Maxwell’s equations. We will start by assuming
that we are dealing with a linear, time-invariant system so that our Fourier representations
of functions are appropriate. Now, let’s rewrite all of the quantities in Ampere’s law using
the representation provided by the inverse Fourier transform. This gives us
∇ × H(r, t) = ∂t D(r, t) + J(r, t)
 ˆ ∞   ˆ ∞  ˆ ∞
1 jωt 1 jωt 1
∇× H(r, ω)e dω = ∂t D(r, ω)e dω + J(r, ω)ejωt dω
2π 2π −∞ 2π −∞ (1.62)
ˆ ∞ −∞ ˆ ∞ ˆ ∞
1 1 1
∇ × H(r, ω)ejωt dω = jωD(r, ω)ejωt dω + J(r, ω)ejωt dω
2π −∞ 2π −∞ 2π −∞

18
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Since our system is linear and time-invariant, we can analyze this system at each frequency
component individually and then add up all the results later to recover the full time domain
response. Due to this, we don’t need to “worry” about the integrations and common terms
in this equation. A more exact justification for this is possible by “projecting” our equation
onto a particular Fourier harmonic of interest. We can do this by taking the inner product
of (1.62) with a Fourier harmonic, e.g., exp [jω ′ t]. For notational simplicity, we will only
focus on what happens to the left-hand side of this equation (the same process will work for
all of the terms in the overall equation). Now, taking the inner product gives us
ˆ ∞  ˆ ∞ 
−jω ′ t 1 jωt
e ∇ × H(r, ω)e dω dt. (1.63)
−∞ 2π −∞

Swapping the order of integrations then gives us

ˆ ∞ ˆ ∞ 
1 j(ω−ω ′ )t
∇ × H(r, ω) e dt dω. (1.64)
2π −∞ −∞

The inner integration will only be non-zero if ω = ω ′ . More generally, we can recognize this
inner integration as being proportional to the Dirac delta function δ(ω − ω ′ ). Hence, we have
that (1.64) simplifies to
ˆ ∞
∇ × H(r, ω)δ(ω − ω ′ )dω. (1.65)
−∞

This final integration can be evaluated easily using the sifting property of the delta function.
This finally gives us the simplified result that
ˆ ∞  ˆ ∞ 
−jω ′ t 1
e ∇ × H(r, ω)e dω dt = ∇ × H(r, ω ′ ).
jωt
(1.66)
−∞ 2π −∞

This same process can be applied to the rest of the terms in Ampere’s law given in (1.62). By
finally swapping the arbitrary notations of ω ′ to ω at the end to keep a consistent notation,
we have

∇ × H(r, ω) = jωD(r, ω) + J(r, ω). (1.67)

We can follow suit with the rest of Maxwell’s equations. The end result is summarized
below.

Maxwell’s Equations: Frequency Domain

Ampere′ s Law : ∇ × H = jωD + J (1.68)

Faraday′ s Law : ∇ × E = −jωB (1.69)
Gauss′ Law of Electricity : ∇·D=ρ (1.70)
Gauss′ Law of Magnetism : ∇·B=0 (1.71)

19
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Similarly, the current continuity equation becomes

∇ · J = −jωρ. (1.72)

In electrical engineering, we typically use phasors to represent our time-harmonic quan-

tities. These are completely specified if we establish their magnitude, frequency, and initial
phase. We do this so that we can represent our phasors using complex-valued exponential
functions (for which algebra is very easy) as opposed to the more cumbersome trigonometric
functions. For a cosine reference at a fixed position in space, an example of this is


Ex (z = 0, t) = E0 cos ωt + φ
(1.73)
 
jφ jωt
= Re E0 e e .

We typically call the underlined portion the phasor. It has a magnitude of E0 , a phase of φ,
and a frequency of ω. We can see that the two lines in (1.73) are equivalent by using Euler’s
formula,

ejx = cos(x) + j sin(x), (1.74)

to write the exponential functions as trigonometric functions. We will use the phasor repre-
sentation shown in (1.73) in most cases throughout this course. As a result of this, it is good
to remember that we can recover the instantaneous expression of a phasor solution by mul-
tiplying it by exp(jωt) and taking the real part. You will need to do this frequently in this
course, and can also be useful in practice; e.g., for generating time-harmonic visualizations
of simulations produced by various computational electromagnetics software.
Before moving on, it will be good to review a few additional mathematical preliminaries.
Two identities that we will find particularly useful throughout this course are

2 cos(x) = ejx + e−jx (1.75)

and

2j sin(x) = ejx − e−jx . (1.76)

We will often use these identities to simplify solutions to problems involving reflected fields
and to compute the power flow in an electromagnetic system. We will also often need to take
the complex conjugate of expressions involving many imaginary numbers. It is important to
remember that to successfully perform this operation we simply flip the sign anywhere that
we see the imaginary number. For example, we have that
 ∗
5−jx
je = −je5+jx . (1.77)

Now, let’s put some of these frequency domain and phasor concepts together to look at
a particularly useful form of the solution to the wave equation. This is

Ex (z, t) = E0 cos(ωt − k0 z), (1.78)

20
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.11: Illustration of the period and wavelength of a sinusoidal wave.

where k0 = 2π/λ0 = ω/c0 is the free-space wavenumber of the wave, and λ0 = c0 /f is the
wavelength in free space (Note: this formula for the wavelength is only valid in a limited
number of circumstances, and should be used with caution throughout this course!). You can
see that this is in fact a solution to the wave equation by substituting it back into (1.50).
Considering the units of the wavenumber, we see that it helps us understand how much
the phase will change by moving backward/forward some physical distance along the wave.
Similarly, if we hold at a fixed spatial location, ω = 2πf = 2π/T helps us understand how
much the phase will change as time elapses. Both of these quantities are illustrated in Fig.
1.11.
We can also use this expression to introduce the concept of phase velocity. This is the
velocity that a particular position of the wave propagates at. We can find this by setting
the argument of the cosine function equal to a constant and seeing what result we get. For
simplicity, we will choose the constant to be 0. For this situation, we have

ωt − k0 z = 0. (1.79)

We can now rearrange this equation so that we will have units that look like a velocity. This
gives
ω z
= . (1.80)
k0 t
We will call the left-hand side of this equation the phase velocity of the wave, which will be
denoted as vp . If we expand out the definitions of ω and k0 we can find that
ω ω
= = c0 . (1.81)
k0 ω/c0
That is, the phase velocity matches the speed of light. So what was the point in defining
the phase velocity as a concept if it is just equal to the speed of light? It turns out that
the phase velocity will only match the speed of light for a few special cases of propagating

21
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

waves. Somewhat surprisingly, we will even find that for some kinds of propagating waves
the phase velocity can be greater than the speed of light! The reason for this lies in the
fact that the wavenumber does not always have such a simple form as ω/c0 . We will discuss
this in much more detail when we reach other portions of the course, such as the sections on
waveguides.

1.5.1 Frequency Domain Wave Equation

Now that we know how to convert Maxwell’s equations to the frequency domain, let’s look
at how our source-free wave equations change. As we will see, the derivation changes in only
a minor way compared to the time domain one. Considering this, let’s make things a bit
more interesting by deriving the wave equation for a homogeneous medium instead of free
space. This means that all of space is completely filled with a material characterized by a
general permittivity and permeability, denoted as ϵ and µ, respectively.
With this in mind, we again start with Ampere’s law. In the frequency domain and after
using the constitutive relations, this is

∇ × H = jωϵE. (1.82)

Similarly, Faraday’s law is

∇ × E = −jωµH. (1.83)

To combine these two equations into a single one, we will take the curl of (1.83). This gives

∇ × ∇ × E = −jωµ∇ × H. (1.84)

We can now substitute the right-hand side of (1.82) into this equation for ∇ × H. This gives

∇ × ∇ × E = µϵω 2 E. (1.85)

Moving everything to one side of the equation gives us the vector wave equation for the
electric field in the frequency domain. This is

∇ × ∇ × E − µϵω 2 E = 0. (1.86)
√
If we define the speed of light (in this material) to be c = 1/ µϵ, then we can rewrite this
to be
 2
ω
∇×∇×E− E = 0. (1.87)
c

Again, we can define the wavenumber (in this material) to be k = ω/c so that we arrive at

∇ × ∇ × E − k 2 E = 0. (1.88)

Before moving on, it is useful to note that for simple materials with µ, ϵ > 0 the speed of
light will be reduced in the material. As a result, the wavenumber becomes larger. That

22
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

is, the phase varies more quickly as a function of position in the material than it would in
free space. Hence, the wavelength is also smaller. Although most materials have µ, ϵ > 0,
it is possible for either or both of µ and ϵ to be less than 0. An example of a natural
material that can have ϵ < 0 is a plasma. More recently, there has been significant research
interest in designing synthetic materials that behave as if they have negative µ and/or ϵ.
These man-made materials are known as metamaterials, and they can be engineered to have
many exciting properties that are not possible with some of the simpler materials that are
commonly encountered in the natural world.
Now, returning to the derivation of the wave equation, since we are still in a source-free
homogeneous region we can use the same vector calculus identities to rewrite these equations
into the Helmholtz wave equation form we are familiar with. In particular, we note that

∇ × ∇ × f = ∇ ∇ · f − ∇2 f .


(1.89)

Using this in (1.88), we have

∇ ∇ · E − ∇2 E − k 2 E = 0.


(1.90)

Noting that Gauss’ law of electricity in a source-free region of a homogeneous medium is

ϵ∇ · E = 0, (1.91)

we can further simplify the wave equation. Using (1.91) in (1.90), we have

∇2 E + k 2 E = 0. (1.92)

This is the vector Helmholtz wave equation in the frequency domain.

We can now turn our attention to coming up with solutions to this equation in our
homogeneous medium. As with the time domain, this form of the wave equation is mostly
useful when we consider a relatively simple case so we can avoid the difficulty of dealing with
dyadic functions. So, let’s go ahead and again assume that we have an electric field polarized
along the x-direction that is only varying in the z-direction. That is, E = x̂Ex (z). Note
that in this expression we have dropped the explicit reference to the fact that this electric
field should be defined at a particular frequency. This is done to save space in our equations,
and is done very often when it is understood by context that a particular time convention is
being used to suppress the frequency variables.
So what should the functional form of Ex (z) be to solve (1.92)? To start out, let’s
substitute our expression for E into (1.92) to get our simpler equation. Recall that because
we are working in Cartesian coordinates and our electric field is polarized along a single axis
and only varies along a single dimension, the vector Laplacian simplifies greatly. We end up
with our equation being

d2
2
Ex (z) + k 2 Ex (z) = 0. (1.93)
dz
The first thing that we should notice is that whatever functional form we choose, it is going
to need to “return” back to itself in some manner after we have taken two derivatives.

23
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

The options that should jump to mind for this kind of behavior should be cos(kz), sin(kz),
and exp(±jkz). As we will see throughout the semester, we can certainly use any of these
functions as our “baseline” solution to the wave equation. However, we can often simplify
the algebra in our overall analysis of a problem if we make a good choice as to which general
solution to use. For “open” problems that don’t have any boundaries that we need to worry
about the behavior of our solution at, it will generally be the case that the exponential
functions are good choices of general solutions.
Considering this, let’s write our solution as

Ex (z) = E + e−jkz + E − ejkz (1.94)

for the time being, where E + and E − are complex-valued constants denoting the amplitude
and phase of the two solutions. The superscripts on these constants are suggestively written
to point out that the first function on the right-hand side of (1.94) corresponds to a wave
propagating in the +z-direction, while the second function on the right-hand side of (1.94)
corresponds to a wave propagating in the −z-direction. How can we tell that this is the
case? The easiest way will be to transform our phasor solutions back to the time domain.
Recall that we go about doing this by multiplying these solutions by exp(jωt) (our time
convention) and taking the real part of the result. From this, we get
     
jωt + −jkz jωt − jkz jωt
Re Ex (z)e = Re E e e + Re E e e
(1.95)
+ + − −
= |E | cos(ωt − kz + ∠E ) + |E | cos(ωt + kz + ∠E ).

In this equation, we have used the symbol ∠E ± to represent the phase of the complex-valued
constant E ± . Hence, we see that the first function on the right-hand side of (1.95) does
represent a wave propagating in the +z-direction, and likewise for the negative propagating
wave.

1.6 Introduction to Plane Waves

The wave solution that we studied in the previous section is an example of a very special
kind of wave that is often used in studying electromagnetic systems. In particular, this kind
of wave is called a plane wave. This name is in reference to the fact that if we were to freeze
time we would see that the isophase surfaces (the surface formed by all points having the
same phase) of this wave form planes in 3D space (see Fig. 1.12). Further, we can call
this wave a uniform plane wave because all of the points on our isophase surface have the
same magnitude. This kind of wave, although very simple, is extremely important in many
practical applications. The reason for this is that if we move a sufficiently far distance away
from the source of a propagating wave we can almost always consider the wave reaching us
to effectively be a uniform plane wave (see Fig. 1.13). This greatly simplifies the analysis
we need to perform when designing various systems, such as antennas, as we will see later
in this course.
Before moving on, a quick note on terminology. If someone refers to something simply as
a plane wave, it is a good rule of thumb to assume they actually mean a uniform plane wave.

24
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.12: Example of the wavefronts of a plane wave. Note that the wavefronts are
orthogonal to the direction of propagation (image from Wikipedia [4]).

Figure 1.13: Radiated fields far away from their source quickly begin to look locally flat; i.e.,
like plane waves.

The “uniform” part is very commonly omitted since it is generally safe to assume that what
we are discussing are uniform plane waves.
Let’s now take some time to look at the properties of plane waves in a little more detail.
To do this, we will first go back to using our vector wave equation in the form of

∇ × ∇ × E(r) − k 2 E(r) = 0. (1.96)

Let’s see what happens when we apply the curl operation to the simple plane wave solution
we have been considering, i.e., E(r) = x̂ exp(±jkz). This is

  x̂ ŷ ẑ
±jkz
∇ × x̂e = ∂x ∂y ∂z = ±jk ŷe±jkz . (1.97)
e±jkz 0 0

If we take the curl again, we arrive at

  x̂ ŷ ẑ
±jkz
∇× ± jk ŷe = ∂x ∂y ∂z = k 2 x̂e±jkz , (1.98)
±jkz
0 ±jke 0

25
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.14: Illustration of how the propagation direction and polarization directions of E
and H form a right-handed coordinate system for a plane wave. We can take cross-products
between the unit vectors of the different components of this coordinate system to find the
direction of the remaining component (e.g., Ê × Ĥ = k̂ or k̂ × Ê = Ĥ).

which was the expected result from our previous study. So why did we do this exercise?
The reason is that when we evaluate the derivatives in this way, we can notice an important
“shortcut” when dealing with plane waves. To see this, we note that the result of the first
curl operation can be viewed as
 
±jkz
∇ × x̂e = −jk(∓ẑ) × x̂e±jkz , (1.99)

while the second curl operation can be viewed as

 
±jkz
= (−jk)2 (∓ẑ) × (∓ẑ) × x̂ e±jkz .


∇ × ± jk ŷe (1.100)

The trend is that the final direction of the curl operation applied to a plane wave simplifies
down to taking the cross product between the direction of propagation and the polarization
direction of the plane wave. This is multiplied by −jk and the original plane wave function
exp(±jkz) to get the complete result of the curl operation. We will revisit this concept in a
more general setting when we learn more about plane waves later in this course. For now, the
important takeaway is that due to this relationship, a plane wave can only be polarized in a
direction orthogonal to its propagation direction. This is why we never considered solutions
of the form ẑ exp(±jkz). Simply put, this is not a valid solution to the free-space wave
equation.
This same kind of “shortcut” can also be used to help us quickly determine what the
magnetic field associated with our plane wave will look like if we have already determined
the behavior of the electric field. In particular, we can look at Faraday’s law

∇ × E = −jωµH, (1.101)

26
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

to see that
1 jk 1
H=− ∇×E= (∓ẑ) × x̂e±jkz = ∓ŷ e±jkz , (1.102)
jωµ jωµ η
p
where η = µ/ϵ is the intrinsic impedance of the medium. In free space, the intrinsic
impedance is η0 = 120π ≈ 377Ω. We see that the magnetic field is also pointing in a direction
orthogonal to the propagation direction. Further, it is orthogonal to the direction the electric
field is pointing in. Overall, we get a “right-handed” coordinate system relationship where
we see that the cross product between the polarization direction of the electric and magnetic
fields (in that order) gives us the propagation direction of the plane wave. This is illustrated
in Fig. 1.14.
A wave that has both it’s electric and magnetic fields orthogonal to the propagation
direction is called a transverse electromagnetic wave. A plane wave is a typical example
of this, but there are other special kinds of transmission lines that also can support the
existence/propagation of transverse electromagnetic waves. We will learn about these more
in the transmission lines portion of the course.

1.7 Boundary Conditions for Maxwell’s Equations

In many ways, the integral and differential forms of Maxwell’s equations can be viewed as
being equivalent. However, in reality, the integral forms are a more “complete” or “funda-
mental” representation of Maxwell’s equations. The reason for this is that if we only have the
differential forms of Maxwell’s equations we find some significant difficulty in determining
how the fields and fluxes should behave at boundaries between regions of space with different
electromagnetic properties (e.g., free space meeting a dielectric material, free space meeting
a perfect conductor, etc.). The key to overcoming these difficulties is by specifying bound-
ary conditions for the different quantities involved in Maxwell’s equations. These boundary
conditions tell us how the fields and fluxes should behave at any kind of boundary/interface
between materials, allowing us to move forward with solving Maxwell’s equations. As we
will see shortly, these boundary conditions are actually “embedded” in the integral form
of Maxwell’s equations. That is, we can derive the boundary conditions from the integral
form of Maxwell’s equations. The fact that this information is already contained in the in-
tegral form of Maxwell’s equations is why we can think of them as being a more “complete”
representation of Maxwell’s equations than the differential form.

1.7.1 Curl Equations

To begin deriving boundary conditions, let’s consider the situation illustrated in Fig. 1.15.
This setup consists of an interface between two materials that have distinct electromagnetic
properties, denoted by ϵ1 ̸= ϵ2 and µ1 ̸= µ2 . We will use Faraday’s law with this picture to
determine how the electric field should behave at this kind of interface. To do this, we will
integrate over a small rectangular surface centered at the interface (as shown in Fig. 1.15)
while the height of the rectangle becomes infinitesimally small. For this situation, Faraday’s

27
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.15: Illustration of regions to derive boundary conditions from Maxwell’s curl equa-
tions.

law becomes
˛ ¨
lim E · dℓ = −∂t lim B · n̂s dS. (1.103)
∆h→0 C ∆h→0 S

Due to the limit, the terms aligned with n̂ will go to 0, and as a result we will not need
to consider them. Likewise, the surface integral will evaluate to 0 in the limit because the
surface area goes to 0. Hence, we simplify Faraday’s law to be


∆ℓt̂ · E2 − E1 = 0. (1.104)

This is starting to look like a usable boundary condition. However, both ∆ℓ and t̂ (a unit
vector tangential to the material interface) are not uniquely determined quantities. We
would prefer to phrase our boundary condition using only uniquely specified quantities, so
it would be good for us to work to rewrite (1.104).
Since ∆ℓ is applied in equal proportion to both E1 and E2 , we can simply “divide” it
out of this equation. We can’t simply remove t̂ in a similar way because then we would lose
the essential information that this boundary condition only applies to the components of E1
and E2 that are tangential to our interface.
There are various ways to mathematically specify that the boundary condition should
only apply to the tangential components of E1 and E2 . We will follow a somewhat round-
about derivation for a particular specification that is commonly used in electromagnetics.
Although this is roundabout for deriving a boundary condition from Faraday’s law, it is
more useful in considering Ampere’s law (which we will do shortly).
In particular, if we are clever we can see that t̂ = n̂ × (t̂ × n̂). This allows us to use a few

28
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

vector identities to rewrite the left-hand side of (1.104) as

t̂ · E2 − E1 = E2 − E1 · t̂
 


= E2 − E1 · n̂ × (t̂ × n̂)
(1.105)
 


= −(t̂ × n̂) · n̂ × E2 − E1 .

In moving from line two to line three, we used the vector identity that

a · (b × c) = b · (c × a) = c · (a × b). (1.106)

This is a very useful identity to memorize. It is easy to memorize it by recognizing that the
result stays the same under “cyclic permutations” of the vectors in the identity.
Returning to (1.105), we see that we now have a useful result. We can omit the (t̂ × n̂)
part safely from our boundary condition because the n̂× that is being applied to E1 and E2
has the desired result of projecting the field vectors into the plane of the interface between
the two regions. Further, the unit normal vector is uniquely determined by the geometry of
the interface. Hence, our final boundary condition from Faraday’s law becomes


n̂ × E1 − E2 = 0. (1.107)

In words, this boundary condition tells us that the tangential components of the electric field
are continuous across a boundary.
We can of course use our constitutive relations to relate this boundary condition to the
electric fluxes as well. This tells us that

n̂ × ϵ−1 −1
1 D1 = n̂ × ϵ2 D2 . (1.108)

That is, the tangential components of the electric flux is discontinuous across a boundary if
the two materials have different permittivities. This is again another helpful reminder to us
that the electric fields and fluxes are fundamentally different quantities.
We can now turn our attention to Ampere’s law. We will consider the situation shown
in Fig. 1.16, and will explain some of the changes from Fig. 1.15 shortly. For this situation,
Ampere’s law becomes
˛ ¨


lim H · dℓ = lim ∂t D + J · n̂s dS. (1.109)
∆h→0 C ∆h→0 S

Using similar arguments to what we did for Faraday’s law, we can reason that the normal
components of H and the surface integral of ∂t D aren’t going to be important in this limit.
It is tempting to also use this logic to suggest that the surface integral of J is also equal
to 0 in this limit. However, there are many situations where it is convenient to consider
the concept of a surface current density. This is a form of current that can flow in two
dimensions, but is confined to an infinitesimally thin sheet. Although surface currents are
always an approximation to an actual current density that can physically exist, they can be
very useful in simplifying the mathematical analysis of common problems. For instance, this

29
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.16: Illustration of regions to derive boundary conditions from Ampere’s law when
a surface current density is present.

kind of surface current can be used when analyzing conductors with very high conductivity
(which is the case in many practical scenarios), and can also be a useful mathematical tool
in modeling certain kinds of antennas. In general, a surface current has units of [A/m]. We
will often denote a surface current as Js in this set of notes. However, it is very common to
simply use the symbol J for any kind of current density when the type of current density
(e.g., volume or surface currents) can be inferred from the context.
Now, we can evaluate (1.109) assuming that there is a surface current density present.
The integral evaluates to


∆ℓt̂ · H2 − H1 = ∆ℓ(t̂ × n̂) · Js . (1.110)

Note that we have used n̂s = (t̂ × n̂) on the right-hand side to act as the unit normal vector
to the surface we integrated over in Fig. 1.16. Dropping common terms and recalling the
vector identities we used to simplify the boundary condition from Faraday’s law, we arrive
at
 


(t̂ × n̂) · n̂ × H1 − H2 = (t̂ × n̂) · Js . (1.111)

We can again omit the common vector terms (t̂ × n̂) since they no longer contribute any
meaningful information to the boundary condition. As a result, the final boundary condition
becomes


n̂ × H1 − H2 = Js . (1.112)

This tells us that the tangential components of the magnetic field can be discontinuous at
an interface between two regions only if there is a surface current density flowing at the
interface. As alluded to previously, this will occur at the interface with a perfect electric

30
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.17: Illustration of regions to derive boundary conditions from Maxwell’s divergence
equations.

conductor (an approximation that assumes the conductivity σ ≈ ∞). However, for many
cases there will not be a Js and we can usually conclude that in similarity to the electric
field, the tangential components of the magnetic field will be continuous across an interface.
We can also use the constitutive relations to write this boundary condition in terms of the
magnetic flux. The interested reader can do this themselves and see that a very similar
conclusion to that from Faraday’s law can be found here as well.

1.7.2 Divergence Equations

We have so far determined what happens to the tangential components of the fields and
fluxes at interfaces between two distinct regions. Although this is good progress, we are
not finished yet. In particular, we need to also determine what happens to the normal
components of the electric and magnetic fields/fluxes at an interface. We can do this by
considering the other two Maxwell’s equations. The illustration of our integration regions
that will be appropriate for these equations are shown in Fig. 1.17.
Let’s start with Gauss’ law of magnetism. For this situation, we will perform our inte-
grations with the limit of ∆h going to 0. This gives us
‹
lim B · dS = 0. (1.113)
∆h→0 S
We can see that we have 3 distinct surfaces to consider in Fig. 1.17. The top and bottom
disks of our cylinder will be able to retain non-zero values, but the limit of ∆h going to 0
will cause the integration around the rest of the cylinder to go to zero due to the shrinking
surface area. Hence, our result becomes


∆S n̂ · B1 − B2 = 0. (1.114)
We can remove the ∆S for our unique description of the boundary condition. As a result,
we end up with


n̂ · B1 − B2 = 0. (1.115)

31
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

This tells us that the normal component of the magnetic flux density is continuous across
a boundary. We can use the constitutive relations to easily see that this means that the
normal component of the magnetic field will not be continuous across the material interface
(if the materials have different permeabilities).
We can follow a similar process now for Gauss’ law of electricity. This becomes
‹ ˚
lim D · dS = lim ρdV. (1.116)
∆h→0 S ∆t→0

The left-hand side will evaluate in the same way as for Gauss’ law of magnetism. We may
be tempted to suggest that the right-hand side is equal to 0. However, in the same way that
we needed to consider surface current densities for Ampere’s law, we also need to consider
surface charge densities. This is essentially the same concept as a surface current density,
but is now applied to charge densities. In particular, a surface charge density, denoted as
ρs with units [C/m2 ], is a set of charges that are confined to an infinitesimally thin region
along the boundary between the materials.
Combining these concepts, we have that


n̂ · D1 − D2 = ρs . (1.117)

This tells us that the normal component of the electric flux density can be discontinuous
across a material interface if there is a surface charge density there. Similarly, we can see that
if there is no surface charge density then the normal component of the electric flux density
will be continuous across the interface. Finally, if the regions have different permittivities,
we can conclude that the normal component of the electric field will be discontinuous across
the interface even if there is no surface charge density there.

1.7.3 Summary
The boundary conditions are summarized below.

Boundary Conditions


n̂ × E1 − E2 = 0 (1.118)


n̂ × H1 − H2 = Js (1.119)


n̂ · D1 − D2 = ρs (1.120)


n̂ · B1 − B2 = 0 (1.121)

It should be emphasized that in this class, Js and ρs will only be non-zero at the interface
between a homogeneous medium and a perfect conductor or will be explicitly specified as
part of the problem statement (e.g., in analyzing a simple “antenna” in a few lectures).

1.8 Material Properties

Up to this point, we have been fairly unconcerned about whether we are talking about free
space or a homogeneous medium. We saw a few lectures ago that the only real difference was

32
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

that we adjusted our permittivity and permeability to account for the fact that we were in a
material as opposed to being in free space. However, from our other physics courses we know
that real materials are made up of a huge collection of atoms/molecules arranged in very
special fashions. This raises the question of whether or not our simple picture of accounting
for the effects of a material with a modified permittivity and permeability is appropriate.
As we will see, this is reasonable for many situations that we are interested in for practical
applications. However, there are situations where a more detailed description of the features
of a material are needed.
One of the driving concepts behind this simplification can be captured in the name macro-
scopic electromagnetics. This is a framework for studying electromagnetic systems where we
can make simplifications such as treating materials through a shift in the permittivity and
permeability. The reason this is acceptable is that the length scale that the atomic/molecular
physics operates over is generally incredibly small compared to the wavelengths of electro-
magnetic fields. As a result, we can take a “macroscopic” viewpoint where we don’t need
to consider what is happening to each of the individual atoms/molecules. If we were to try
and consider how each individual atom/molecule is behaving, we would be taking a “micro-
scopic” viewpoint of the physics. For certain situations, we need to attempt to use some
kind of in-between picture of the physics. This is called the “mesoscopic” regime. Each of
these different viewpoints have important applications in practical engineering. However, in
this course, we will mainly focus on a macroscopic viewpoint.
With this in mind, we will now consider how materials respond to electric and magnetic
fields to better rationalize the macroscopic electromagnetism viewpoint that we have been
using somewhat freely up to this point. We will start by considering the polarization of a
dielectric material followed by the magnetization of a magnetic material.

1.8.1 Polarization Intensity

We will first consider what happens to the electric charges in a medium where an electric
field is applied, and correspondingly, how these changes to the electric charges affect the
electromagnetic fields and fluxes. To simplify this, we will consider the material to be made
up of generic atoms/molecules that are composed of a positively charged nucleus surrounded
by negatively charged electrons. Further, we assume that the number of positive and negative
charges over the range of the entire material are perfectly balanced; i.e., the material is
neutrally charged. We will also assume that the strength of electric fields we are considering
are not powerful enough to “rip” the electrons away from the nucleus (i.e., we do not ionize
the material ).
In the absence of an electric field, we can assume that all of the atoms/molecules are
essentially randomly oriented so that even if they are polar molecules there is no net polarity
to the material (see Fig. 1.18). However, this changes if the material is placed in a uniform
electric field. In this situation, the electric field can exert a force on the charges that will
“stretch” the nuclei and electrons away from their equilibrium positions and cause them to
be oriented in a regular arrangement (see Fig. 1.18). As a result, we can view each of these
atoms/molecules as being a small electric dipole. We can consider these tiny dipoles through

33
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.18: (Left) In the absence of an applied electric field, the “dipoles” that make up
the material are oriented randomly. (Right) When an external electric field is applied, the
“dipoles” are stretched and aligned into a regular pattern.

their dipole moments, defined as

p = qℓ, (1.122)

where q is the charge of the dipole and ℓ is a vector pointing from the center of mass of the
negative charges to the center of mass of the positive charges.
We can describe the overall effect of these many tiny dipoles through the polarization
vector/intensity, defined as
X pi
P = lim . (1.123)
∆v→0
i
∆v

How can we add this into our description of Maxwell’s equations? The key is recognizing
that the divergence of P is going to be able to be considered to be a new kind of charge
density (look at the units). In particular, we have a bound charge density (because it is
bound to the material we are considering) that is given as

ρb = −∇ · P. (1.124)

The negative sign is needed because the divergence computes an outward flux, while we are
actually more concerned with what is happening “inside” the region we are considering.
Now that we have associated the polarization intensity with a charge density, we can
rewrite Gauss’ law of electricity within a material as

∇ · (ϵ0 E) = ρ + ρb
(1.125)
= ρ − ∇ · P.

We can move the polarization intensity term to the left-hand side and consolidate terms
under the divergence symbol to be


∇ · ϵ0 E + P = ρ. (1.126)

34
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Comparing this to what we originally stated Gauss’ law to be, we can see that the effect of
the material can be elegantly incorporated into Maxwell’s equations if we say that

D = ϵ0 E + P. (1.127)

In many cases, working directly with the polarization intensity is inconvenient. For-
tunately, it is often the case that the polarization intensity that occurs in a material can
be easily related back to the applied electric field. In particular, we typically have in the
frequency domain that

P = ϵ0 χe E, (1.128)

where χe is the electric susceptibility. Each material will have a different χe that characterizes
the unique properties of that particular material. We will consider how to derive χe from a
simple model appropriate for many types of materials a few lectures from now. As mentioned
previously, this type of relationship is commonly adequate. Important cases where it is not
acceptable include for nonlinear materials, which require a more sophisticated description.
If we are performing our analysis in the frequency domain, the electric susceptibility will
typically be a complex number. In this situation, it is represented as

χe = χ′e − jχ′′e , (1.129)

where χ′e is the real part and χ′′e is the imaginary part of the susceptibility. The imaginary
part is necessary to describe lossy materials, i.e., materials which absorb some of the energy
from the electric field. In all physical materials, there will be some loss. However, there
are plenty of materials for which the loss is small enough that we can ignore it for quick
calculations. We can typically associate this loss of energy with a kind of frictional force
that occurs due to collisions between nearby atoms/molecules inside the material as they
are being stretched and “flipped” (for time-varying fields). It turns out that water molecules
have very high loss at microwave frequencies. This is why your microwave oven works so
well for heating up food, since many foods that we eat have a high amount of water content.
Moving on, we can substitute our new representation for the polarization intensity into
(1.127) to get

D = ϵ0 E + ϵ0 χe E
(1.130)
= ϵ0 (1 + χe )E.

By defining the relative permittivity as

ϵr = ϵ′r − jϵ′′r = (1 + χe ), (1.131)

we arrive back at our simple constitutive relationship of

D = ϵ0 ϵr E = ϵE. (1.132)

Our interpretation of the imaginary part of the material property corresponding to loss also
applies to the relative permittivity. Note that within the context of a lossy dielectric, it is
common to refer to ϵ′r as the dielectric constant of the material.

35
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.19: Illustration of the origin of the name “loss tangent”.

Now, one common way to specify the loss associated with a material is through its loss
tangent. This is typically written as tan(δ) = ϵ′′ /ϵ′ = ϵ′′r /ϵ′r , and is illustrated in Fig. 1.19.
This quantity is particularly useful for determining whether a material corresponds to a low-
loss dielectric or not. Generally, we want tan(δ) ≪ 1 to classify a material as being low-loss.
When this is the case, certain approximate formulas can be used to analyze the physics of
the situation, as we will see shortly.
As promised, we see from (1.132) that so long as this macroscopic electromagnetics
perspective works we can indeed account for a material medium by simply changing the
permittivity we use in our equations. This type of perspective can break down in a few
different ways. One manner is when the frequencies of our electromagnetic waves become
very high. In these situations, the small wavelengths of the waves can more directly interact
with the atoms/molecules (including their internal structure). At this point, we typically
need to resort to some kind of quantum mechanical description of the atoms/molecules to
begin to understand phenomena like spontaneous and stimulated emission. Another situation
where our macroscopic viewpoint can break down is if the material region we are considering
becomes extremely thin or extremely small. Again, we start to run into quantum mechanical
effects becoming more pronounced in these extremely small volumes. We will not study these
cases in this course, however, we will mention that understanding and engineering these
kinds of fundamental quantum interactions is a very active and vibrant research area at
the heart of achieving things like quantum computation. In most physical implementations
being pursued to date, electromagnetic devices/physics play a crucial role in controlling and
interacting with the quantum systems being used as quantum bits (qubits).

1.8.2 Magnetization Intensity

The effect that a magnetic material has on the applied magnetic field can be captured in a
similar manner to the polarization intensity. In this situation, we consider the material to
have many tiny magnetic dipole moments that can be viewed as being due to the orbiting
of electrons, as well as their spin orientation in certain cases. These details are beyond the

36
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.20: (Left) Illustration of a single magnetic dipole moment arising from a small
current loop. (Right) Computing the circulation of the magnetization intensity is related to
how much the individual current loops of the magnetic dipoles align in a particular direction.

scope of this course, so we will simply say that the material has many tiny dipole moments
that are due to small current loops flowing inside the material. The magnetic dipole moment
for these currents can be given by m = r × i, where r points from the origin of the current
loop to the loop itself and i is the vector characterizing the direction of current flow, as
shown in Fig. 1.20.
We will define the magnetization intensity/vector as
X mi
M = lim , (1.133)
∆v→0
i
∆v

where mi is the magnetic dipole moment of the ith current loop. We will associate with this
magnetization a new volume current density given by
Jm = ∇ × M. (1.134)
We can see that this leads to a sensible volume current density by considering how the
circulation of the magnetization intensity relates to the small current loops that make up
the magnetic dipoles, as shown in Fig. 1.20.
Now, this volume current density can be incorporated into Ampere’s law as
∇ × µ−1
0 B = jωD + J + Jm
(1.135)
= jωD + J + ∇ × M.
We can move the magnetization to the left-hand side and consolidate terms underneath the
curl operation to have
∇ × µ−1


0 B − M = jωD + J. (1.136)
Comparing this to our original form of Maxwell’s equations we can recognize that we should
adjust the definition of H to be
H = µ−1
0 B − M. (1.137)
Alternatively, we can write this as


B = µ0 H + M . (1.138)

37
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Similar to the case of electrical response, we can define a magnetic susceptibility to relate
the magnetization more directly to the magnetic field. In particular, we have in the frequency
domain that

M = χm H, (1.139)

where

χm = χ′m − jχ′′m . (1.140)

Similarly, we can define a relative permeability as

µr = µ′r − jµ′′r = 1 + χm . (1.141)

This allows us to write

B = µ0 µr H = µH, (1.142)

as we had been previously doing in this course. A loss tangent can also be written for µr
similar to what was done with ϵr . This is not as commonly done for magnetic materials,
since most magnetic materials are significantly lossy at microwave frequencies. As a result,
tracking the dependence on other physical quantities in terms of a magnetic loss tangent
isn’t as useful as it is with an electric loss tangent.
Now, why were we careful to point out at various times in this derivation (and for the
polarization) that we are working in the frequency domain? The reason is that for any
real material the permittivity and permeability will be functions of frequency. This can be
viewed as a consequence of inertia, i.e., the various atoms/molecules/currents that make up
the dipole moments can’t respond immediately to changes in the applied fields. This delay
in motion will cause the response of the materials to change as a function of frequency. From
your linear systems courses you should recall that if we have two quantities multiplied in the
frequency domain that corresponds to a convolution in the time domain. Hence, all of our
constitutive relations in the presence of a material become convolutions in the time domain
if the material properties vary as a function of frequency. There are certainly times where
we can approximate the frequency variation of the material as being unimportant so that
the convolutions can be safely ignored. However, in general, this is not always going to be
the case. This is another reason why performing analysis in the frequency domain can be
significantly simpler than working directly in the time domain.

1.9 Plane Waves in Lossy Media

Previously, we discussed how materials could be incorporated into Maxwell’s equations and
briefly commented on how they were mathematically described when they were lossy. We will
now see how a lossy material impacts our solutions to the wave equation in a homogeneous
medium. To be general, we will assume that the material has dielectric, magnetic, and
conductive losses. Since we are looking for plane wave solutions, we will also consider a
source-free region of space. Despite this, there can still be induced current flowing inside of
the material due to the conductive losses. We will denote this conduction current as Jc .

38
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Now, for this situation, Ampere’s law becomes

∇ × H = jωD + Jc
(1.143)
= jωϵE + σE,

where we have used the constitutive relation and Ohm’s law to make all the terms on
the right-hand side proportional to the electric field. Following our standard approach for
deriving the wave equation, we next look at Faraday’s law, which for this situation is

∇ × E = −jωµH. (1.144)

Remember that both ϵ and µ are complex-valued quantities to account for the dielectric and
magnetic losses, respectively.
Now, we take the curl of Faraday’s law and substitute in from Ampere’s law to arrive at


∇ × ∇ × E + jωµ σ + jωϵ E = 0. (1.145)

If we define
q

γ= jωµ σ + jωϵ , (1.146)

then we can write this wave equation as

∇ × ∇ × E + γ 2 E = 0. (1.147)

We typically call γ the propagation constant, since it characterizes many details about how
the wave will propagate in a particular medium.
It should be noted that in many cases, it is common to “lump” the dielectric and conduc-
tive losses together into a single complex-valued ϵ. We can see how to do this by returning
to Ampere’s law and writing it as

∇ × H = jωϵE + Jc
= jω(ϵ′ − jϵ′′ )E + σE
   (1.148)
′ ′′ σ
= jω ϵ − j ϵ + E.
ω

If we consider ϵ′′ + σ/ω to be the imaginary part of the permittivity, we find that our
expressions for γ will reduce to
√
γ = jω µϵ, (1.149)

where both µ and ϵ can be complex-valued. If a conductivity is not explicitly given or referred
to in a problem, you should typically use (1.149) in your calculations rather than (1.146).
Now, returning to (1.147), we can use our vector identity in a source-free region of space
to rewrite this wave equation as a vector Helmholtz wave equation. This is

∇2 E − γ 2 E = 0. (1.150)

39
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

From similarity to our previous equations, we see that an identical form of plane wave
solution is going to be appropriate here. That is, if E = x̂Ex (z), we have that
Ex (z) = E + e−γz + E − eγz . (1.151)
It is generally useful to break γ into its real and imaginary parts as
γ = α + jβ, (1.152)
where α is the called the attenuation constant and β is called the phase constant. The reasons
for this should hopefully be clear. Since α is real, it will correspond to exponential decay,
while the jβ term will correspond to a varying phase as a function of position. It is important
to note that because α could accidentally portray exponential growth, we must be careful
with our signs and specify for which regions of a problem our mathematical representations
are valid.
The SI units for α are neper per meter (Np/m) and the SI units for β are radians
per meter. A neper is a dimensionless quantity that is defined so that if α = 1 [Np/m] a
unit wave amplitude will have decreased to a magnitude of 1/e after traveling a distance
of 1 meter. We often work with decibel units in electromagnetic engineering. In these
cases, the α is often reported in terms of dB/m. We can convert between units by noting
that 1 [Np/m] = 20 log10 e = 8.686 [dB/m]. You should never plug an attenuation constant
expressed in dB/m into an exponential function when computing aspects of a solution, this
will lead to the wrong answer!
It should also be noted that in a lossy medium we will need to use β to compute quantities
like the phase velocity or wavelength of our plane waves. This is because the speed of light
√
naively defined as c = 1/ µϵ will become a complex number and lose its useful meaning.
Instead, we will need to compute the phase velocity as
ω
vp = (1.153)
β
and the wavelength as
2π
λ= . (1.154)
β
These expressions will generally work in most situations, and so are much safer to use than
something like λ = c/f , which only holds in lossless and unbounded homogeneous regions of
space (which is a very specific situation!).
Along with the modified propagation constant, we also need to modify our intrinsic
impedance. For this kind of lossy material, this will become
s
jωµ
η= , (1.155)
σ + jωϵ

or simply
r
µ
η= (1.156)
ϵ

40
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

if any conductive losses have been “lumped” into the definition of ϵ. Again, if a conductivity
is not explicitly given or referred to in a problem, you should typically use (1.156) in your
calculations rather than (1.155).
Before moving on, let’s look at how to find the instantaneous expression for a plane wave
in a lossy medium. To do this, we will remember that we can simply multiply our phasors
by exp[jωt] and then take the real part. Doing this for the wave that propagates in the
+z-direction from (1.151), we get
 
+ −γz + −γz jωt
E e =⇒ Re E e e
 
+ −αz −jβz jωt
=Re E e e e
(1.157)
 
=e−αz Re E + e−jβz ejωt

=|E + |e−αz cos ωt − βz + ∠E + .

It is also important to remember that when computing the electric or magnetic fields from
one another in a lossy media that these operations should be done with phasors and then
converted into instantaneous expressions from there. The issue is that you cannot simply
divide by something like η in the instantaneous expressions, since in the lossy media η is a
complex number which cannot be used directly in the time domain.

1.9.1 Useful Approximations

We can now specialize our new formulas to a few cases of particular interest. It is important
to note that these expressions are only approximations valid for limited ranges of material
parameters. It is always acceptable to simply use the full formulas, rather than approximate
ones in working with lossy dielectrics.
The first special case of interest will be for a good dielectric, which is characterized by
ωϵ ≫ σ. That is, the conductive losses are very small. Approximate formulas appropriate to
this regime can be found in Section 8.3-1 of your textbook. One approximate formula that
is of particular interest is that for γ. This is
 ′′
 ′′ 2 
p 1 ϵ 1 ϵ
γ = α + jβ ≈ jω µϵ′ 1 − j ′ + . (1.158)
2ϵ 8 ϵ′

This formula highlights how the loss tangent (recall this is given as ϵ′′ /ϵ′ ) becomes a mean-
ingful parameter in an approximate expansion of the propagation characteristics for good
dielectrics.
Another useful case to consider is that of a good conductor. Approximate formulas ap-
propriate to this regime can be found in Section 8.3-2 of your textbook. For this material,
we will have that σ ≫ ωϵ. We can then simplify our propagation constant to be
p
γ ≈ jωµσ. (1.159)

41
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Similarly, the intrinsic impedance becomes

p
η≈ jωµ/σ. (1.160)
For a good conductor, it is useful to consider the concept of skin depth. The skin depth
is defined as the distance into a material where the field decays to 1/e of its original value.
For a good conductor, we have that
r
ωµσ
α= . (1.161)
2
Hence, our skin depth (denoted as δ), will be
r
1 2
δ= = . (1.162)
α ωµσ
Since we are dealing with exponential decay, it usually only takes a handful of skin depths
for a field to effectively decay to 0. At that point, the field will not be able to “push” charges
around anymore to cause currents to flow. Hence, currents will typically only flow within a
small number of skin depths of the surface of a good conductor.
To get a better feel for this, let’s see how thick the skin depth is for a commonly used
material in circuit board and antenna design. At the frequency of WiFi signals (2.4 GHz),
the skin depth is approximately 1.33 µm for copper cladding. Many circuit board materials
use very thin thicknesses of copper to define their traces (which can also be used to make
antennas, e.g., in your phone). However, they still are typically on the order of 30 – 40 µm
thick, so we can conclude that for the most part all of the current is effectively flowing on
the surfaces of the conductors. This is part of the reason why surface current densities are
such a useful concept. It is much easier to solve problems analytically (and numerically) if
we consider only an infinitesimal layer of current as opposed to rigorously accounting for the
decay profile of the current in the conductor. However, there are situations where this is no
longer adequate. For instance, when the sizes of the conductors start becoming comparable
to the skin depth. This can become important for the design of modern computer chips
because of the huge amount of miniaturization that has happened over the years to continue
improving chip performance.

1.10 Poynting’s Theorem

The flow of energy and power throughout a system is important in many areas of electrical
engineering, including in electromagnetics. You are no doubt familiar with the concept of
the conservation of energy. In the area of electromagnetics, the conservation of energy is
captured in Poynting’s theorem. We will now take some time to consider what this looks
like in the time and frequency domains.

1.10.1 Time Domain

To begin, we will write out Faraday’s and Ampere’s laws for a medium characterized by µ,
ϵ, and σ. In addition to this, we will also assume that there is an impressed current present,

42
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.21: Illustration of the region of space considered for deriving Poynting’s theorem.

as depicted in Fig. 1.21. An impressed current can be viewed as a source of electromagnetic

fields that is “forced” onto the system. The important aspect of this is that we don’t
allow the electromagnetic fields to have any meaningful “back action” on the current density
itself. That is, the current density does not change due to the presence of the produced
electromagnetic fields. This is, in general, not a completely accurate view of many physical
phenomena in their entirety. However, it does provide us with a very useful mathematical
tool to go about solving for the accurate description of the physics. This is much like the
ideal voltage or current sources that you used in circuit analysis. For instance, no real current
source can drive an exact current value through a circuit regardless of the composition of the
circuit. However, this is still a very useful model that allows us to neglect the complicated
composition of the source for the many situations where the practical source is operated in
a way where it behaves close to ideal.
Now, our equations are

∇ × E = −µ∂t H (1.163)

∇ × H = ϵ∂t E + σE + Ji . (1.164)

We are interested in looking at energy and power. To start getting to quantities with the
correct kinds of units for this to be the case, we can take the dot product of (1.163) with H.
We can see this is a reasonable idea by remembering that the magnetic energy stored in an
inductor is given by 21 LI02 . Performing the dot product with (1.163) gives

H · ∇ × E = −µH · ∂t H, (1.165)

from which we can see that the right-hand side is starting to have units related to an energy
density. To make this more explicit, we can rewrite the right-hand side by noting that
 
1 1
µH · ∂t H = µ∂t (H · H) = ∂t µH · H . (1.166)
2 2

43
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We define

1
wm = µH · H (1.167)
2

as the magnetic energy density. This has units of H/m · (A/m)2 = J/m3 , so it is in fact a
volumetric energy density. Using this energy density, we can write (1.165) as

H · ∇ × E = −∂t wm . (1.168)

We can follow a similar process for (1.164) by taking the dot product of it with E. This
gives us

E · ∇ × H = ϵE · ∂t E + σE · E + E · Ji . (1.169)

We can use a similar trick with the first term on the right-hand side to define an electric
energy density as

1
we = ϵE · E. (1.170)
2

Using this, we can rewrite (1.169) as

E · ∇ × H = ∂t we + σE · E + E · Ji . (1.171)

We would like to combine (1.168) and (1.171) into a single equation to describe energy
and power. To do this, we will subtract (1.171) from (1.168) to get

H · ∇ × E − E · ∇ × H = −∂t wm − ∂t we − σE · E − E · Ji . (1.172)

We can simplify the left-hand side by noting the vector calculus identity

∇ · (a × b) = b · ∇ × a − a · ∇ × b. (1.173)

Using this gives us,

∇ · (E × H) = −∂t wm − ∂t we − σE · E − E · Ji . (1.174)

We can integrate this equation over a volume to get a description of the total power and
energy contained within the volume. If we use the divergence theorem on the left-hand side
of this volume-integrated equation we get
‹ ˚  
n̂ · (E × H)dS = − ∂t wm + ∂t we + σE · E + E · Ji dV, (1.175)
S V

where V is some finite volume we are integrating over and S is the boundary surface of V .
By noting that the left-hand side has units of V/m · A/m = W/m2 , we can see that the
left-hand side can be viewed as measuring the total amount of power flowing into/out of the

44
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

surface S that encloses V . We see that this is equal to the change in energy density in V
plus a few other terms.
Let’s take a moment to consider the other terms. If we look at the third set of terms
on the right-hand side, we see that this has units of S/m · (V/m)2 = W/m3 . Hence, we can
associate this with the amount of power that is being dissipated in our conductive medium.
The units on our last term on the right-hand side also gives us W/m3 . However, from our
original discussion of this as an impressed current we can conclude that this is power being
supplied to our system.
We can now rearrange terms to arrive at Poynting’s theorem, which characterizes the
balance of electromagnetic energy and power. In particular, we have that
‹ ˚   ˚
n̂ · S dS + ∂t wm + ∂t we + σE · E dV = − E · Ji dV, (1.176)
S V V

where S(r, t) = E(r, t) × H(r, t) is the instantaneous Poynting vector. The remaining terms
on the left-hand side account for the magnetic and electric energy densities and the power
dissipated in the medium. The right hand side is the power supplied to the medium by the
current Ji (recall that negative powers are by convention supplied powers). Hence, Poynting’s
theorem is a statement of the conservation of energy in a simple electromagnetic medium.

1.10.2 Frequency Domain

We can now construct Poynting’s theorem in the frequency domain. It should be no surprise
by now that much of this derivation is going to follow closely from the time domain derivation.
Hence, we will begin with Faraday’s and Ampere’s laws as

∇ × E = −µjωH (1.177)

∇ × H = ϵjωE + σE + Ji . (1.178)

To go about combining these equations, we will still need to take the dot products with
the electric and magnetic fields. However, because we are now dealing with complex-valued
quantities in the frequency domain we will need to be a little more careful.
The decision on which way to take complex conjugates comes down to convention. By
convention, we will define the average Poynting vector as

1 
Savg (r, ω) = Re E(r, ω) × H∗ (r, ω) . (1.179)
2
To get the left-hand side of our combined equations to give us the desired average Poynting
vector, we will need to take the dot product of (1.177) with H∗ . Similarly, we will take the
dot product of E with the complex conjugate of (1.178). This gives us

H∗ · ∇ × E = −µjω|H|2 (1.180)

E · ∇ × H∗ = −ϵ∗ jω|E|2 + σ|E|2 + E · J∗i . (1.181)

45
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We can subtract (1.181) from (1.180) to get

H∗ · ∇ × E − E · ∇ × H∗ = −µjω|H|2 + ϵ∗ jω|E|2 − σ|E|2 − E · J∗i (1.182)

We can now use the same vector calculus identity and the divergence theorem to get
‹ ˚   ˚
∗ ∗
2 2 2
E · J∗i dV. (1.183)


n̂ · (E × H )dS + jω µ|H| − ϵ |E| + σ|E| dV = −
S V V

This can be viewed as being Poynting’s theorem in the frequency domain. We can arrive at
the time average power relationships by taking 1/2 Re{(1.183)}. The main point of interest
here is to note that the only way that ϵ and µ will impact the real power (i.e., dissipation) is
if ϵ and µ have imaginary parts. If you recall from our early discussions on these quantities,
we mentioned that the imaginary parts were associated with loss. We now have another way
to see this through Poynting’s theorem.

1.10.3 Power Flow of a Plane Wave

We will now take a closer look at how the Poynting vector relates to power flow by considering
the case of a plane wave in free space. To do this, we will return back to our standard example
of a plane wave. This is

E = x̂E0 cos(ωt − k0 z) (1.184)

in the time domain and

E = x̂E0 e−jk0 z (1.185)

in the frequency domain. To evaluate the Poynting vector for our plane wave, we will also
need the expression for the magnetic field. We can get this from Faraday’s law, which says
that

∇ × E = −jωµ0 H. (1.186)

From this, we can determine that

E0 −jk0 z
H = ŷ e (1.187)
η0
in the frequency domain and
E0
H = ŷ cos(ωt − k0 z) (1.188)
η0
in the time domain.
Let’s start by considering the time-average Poynting vector in the frequency domain.
This will be
  ∗ 
1  ∗ 1 E0 −jk0 z jk0 z
Savg = Re E × H = x̂ × ŷ Re E0 e e , (1.189)
2 2 η0

46
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.22: Illustration of the Poynting vector following the right-hand rule.

which simplifies down to

|E0 |2
Savg = ẑ . (1.190)
2η0

This basic structure is a good one to memorize. The power flow density for a plane wave has
a direction given by the cross product of the directions that the electric and magnetic fields
point in (in that order to get the correct propagation direction). The magnitude is given by
the remainder of the equation. This is easy to remember by recalling using Ohm’s law to
express powers in your circuits class. The expression in (1.190) is very similar to P = V 2 /R
that you can recall from circuits classes. If instead of using E0 we had used the amplitude
of the magnetic field we would then have

η0 |H0 |2
Savg = ẑ , (1.191)
2

which should remind you of P = I 2 R.

Although we showed this with a plane wave, it is a general property of the Poynting
vector that it points in the direction of power flow. As a result, once we know the electric
and magnetic fields we can also determine the power flow throughout a system. From the
definition of the Poynting vector, we can also see that the direction of power flow follows the
right-hand rule, as illustrated in Fig. 1.22.
Let’s now turn our attention to looking at how the Poynting vector evaluates in the time
domain. Using (1.184) and (1.188), we get

E02
S(r, t) = E(r, t) × H(r, t) = ẑ cos2 (ωt − k0 z). (1.192)
η0

47
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.23: Illustration of the instantaneous Poynting vector at z = 0.

For simplicity, we will focus on the instantaneous Poynting vector at z = 0. This gives us
E02
S(r, t) = ẑ cos2 (ωt). (1.193)
η0
We can now use some trigonometric identities to see how this relates to the time-average
Poynting vector we had in (1.190). In particular, we can recall that
1 1
cos2 (x) = + cos(2x). (1.194)
2 2
We can use this to write our instantaneous Poynting vector as
E02

S(r, t) = ẑ 1 + cos(2ωt) . (1.195)
2η0
We see that we get a DC component (the first term in the parentheses) and another com-
ponent that oscillates at twice the frequency of our plane wave (the second term in the
parentheses), which is illustrated in Fig. 1.23.
We can now compute the time average of the instantaneous Poynting vector. This should
equal the time-average Poynting vector result that we arrived at from our frequency domain
analysis. To see this, we note that the time average is computed as
ˆ
1 T
⟨S(0, t)⟩ = S(0, t)dt, (1.196)
T 0
where T = 2π/ω is a single period of the waveform. Evaluating this gives us

E02
⟨S(0, t)⟩ = Savg = ẑ (1.197)
2η0

because the integral of the cosine function is 0. As expected, our frequency and time domain
results do match.

48
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.24: Illustration of a current sheet located in the (x, y)-plane.

1.11 Radiation from Current Sheets

We will now get our first taste of how to go about solving Maxwell’s equations for a situation
where we can’t consider all of space to be a simple, source-free region of space. We will
consider a relatively straightforward example to begin learning the necessary strategies that
will be used repeatedly throughout the rest of this course. In particular, we will consider
the radiation produced by an infinitely large planar current sheet.
The basic setup of this problem is shown in Fig. 1.24. In it, we have a current sheet that
is covering the entire (x, y)-plane. The mathematical definition of the current is given by
(
x̂J0 cos(ωt), z = 0,
Js (r, t) = (1.198)
0, elsewhere.
As with many things, we could also write this current density in another equivalent form.
For instance, we could use the Dirac delta function to write this current density as
Js (r, t) = x̂J0 cos(ωt)δ(z). (1.199)
Regardless of how we write it, the concept should hopefully be clear. We have a current
density that covers the (x, y)-plane and points in the +x-direction with a fixed amplitude
(J0 ) and frequency (ω).
Now, we will go about determining what the electromagnetic fields should be equal to
in all regions of space. The basic strategy that we will reuse frequently in this course is the
following:
1. Identify the distinct regions of the problem and the boundaries that exist between the
regions.
2. Propose a suitable general solution to Maxwell’s equations in the distinct regions of the
problem. These general solutions will involve a number of unknown constants that we
will need to solve for. For instance, the amplitude, phase, or frequency of the solution.
3. Determine the specific values of the unknown constants by making use of the bound-
ary conditions we know must be satisfied at various boundaries between the different
regions of the problem.

49
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Let’s now run through these steps for the current problem.
For Step 1, we can see that we have two distinct regions – free space for z < 0 and free
space for z > 0. The boundary between these regions is at the (x, y)-plane and is filled by
the current sheet. We will identify the region where z < 0 as Region 1 and the region where
z > 0 as Region 2. For our derivation of boundary conditions, we assumed that n̂ points
into Region 1. Hence, for this problem we will have n̂ = −ẑ for any boundary conditions we
use.
Throughout this course we will learn a number of different general wave solutions to
Maxwell’s equations that we can consider to be part of our “catalog” for answering the
questions posed in Step 2. Each solution will have different characteristics and be well-
suited for satisfying different boundary conditions. At this point, we only have one wave
solution that we are familiar with – a plane wave. Fortunately, this wave solution is exactly
what we will need for this problem (funny how that works out...).
We can rationalize that a plane wave is a good solution because the current sheet fills the
entire (x, y)-plane and has a constant amplitude and phase over this entire region of space.
Considering the boundary condition for the magnetic field, i.e.,

n̂ × (H1 − H2 ) = Js , (1.200)

we see that the magnetic field is going to have a spatial and temporal distribution similar
to the current sheet. Hence, we have a good feeling that using a plane wave as a general
solution will be a good choice.
Before writing out our proposed form of solution, the last item we will address is the
expected propagation direction for the two regions. Since the current sheet is a source of
the electromagnetic waves, it is reasonable to expect that the waves are going to propa-
gate away from the source. If this weren’t the case, antennas wouldn’t work very well for
communications or radar systems, so this seems like a reasonable thing to expect of our
solution.
With this in mind, our proposed solutions in the two regions will be
(
H1 (r, t) = x̂Hx1 cos(ω1 t + k1 z + φx1 ) + ŷHy1 cos(ω1 t + k1 z + φy1 ), z < 0,
H(r, t) =
H2 (r, t) = x̂Hx2 cos(ω2 t − k2 z + φx2 ) + ŷHy2 cos(ω2 t − k2 z + φy2 ), z > 0.
(1.201)

We now need to go about the process of solving for the unknown constants to complete our
solution. For reference, we need to solve for Hx1 , Hy1 , Hx2 , Hy2 , φx1 , φy1 , φx2 , φy2 , k1 , k2 ,
ω1 and ω2 . This may look like a lot, but many of these constants will be quite easy to solve
to the point we won’t even really need to worry about them in future problems, we will just
know what they should be.
This brings us to Step 3 of the process. We are going to solve for these constants by
ensuring they give us the correct behavior at the boundary between the two regions. Let’s
start with the boundary condition for the magnetic field that we already wrote out earlier.
This gives us

−ẑ × (H1 − H2 ) = Js . (1.202)

50
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We can expand these terms at z = 0 to get

− ŷHx1 cos(ω1 t + φx1 ) + x̂Hy1 cos(ω1 t + φy1 )

− (−ŷHx2 cos(ω2 t + φx2 ) + x̂Hy2 cos(ω2 t + φy2 )) = x̂J0 cos(ωt). (1.203)

This looks like a bit of a mess, so let’s simplify it down. First, we can immediately see
that because the current source is only directed in the x-direction we can safely say that
Hx1 = Hx2 = 0. Since this removes these terms from our general solution, we also no longer
need to worry about solving for φx1 or φx2 . Next, we can see that for this equation to be
satisfied we are going to need to have the arguments of all the remaining cosine functions
equal one another for all t. From this, we can conclude that the frequencies of our two
waves match that of the source. That is, ω1 = ω2 = ω. Further, we must also have that
φy1 = φy2 = 0. Recalling that our wavenumbers are linked to the material properties in
the regions (which are identical for this case) and the frequency of the wave, we can also
conclude that k1 = k2 = k0 = ω/c0 . Now that we have that out of the way, the equation can
be simplified down to be

Hy1 − Hy2 = J0 . (1.204)

At this point, you can probably guess that this is going to work out to be Hy1 = −Hy2 = J0 /2.
However, let’s be more careful about this and make sure that we can compute this without
“guessing”.
To do this, we will make use of another boundary condition. In particular, we will use the
boundary condition for the electric field. Hence, we will need to compute what the electric
field is considering our proposed form of solution for the magnetic field. We can do this
either by using Ampere’s law or the tricks we learned previously about manipulating plane
waves. Following the first approach, we have that computing the curl of our plane wave gives
 
∇ × ŷ cos(ωt ± k0 z) = k0 (∓ẑ) × ŷ cos′ (ωt ± k0 z), (1.205)

where the ′ denotes a derivative with respect to the entire argument of the function. We can
then use Ampere’s law to see that
k0

E1 (r, t) = −x̂ Hy1 cos(ωt + k0 z), z < 0,


ωϵ0
E= (1.206)
k
E2 (r, t) = +x̂ 0 Hy2 cos(ωt − k0 z), z > 0.


ωϵ0
We can simplify these expressions by recalling that k0 /(ωϵ0 ) = η0 . This gives us
(
E1 (r, t) = −x̂η0 Hy1 cos(ωt + k0 z), z < 0,
E= (1.207)
E2 (r, t) = +x̂η0 Hy2 cos(ωt − k0 z), z > 0.

If we had instead used our “tricks” to manipulate plane waves we would first have determined
the appropriate polarization directions. For E1 × H1 to give us the correct propagation

51
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

direction of −ẑ, we would see that we would need E1 to be polarized in the −x̂-direction. To
adjust the amplitude of the wave, we would recall that we can use the intrinsic impedance in
a manner similar to Ohm’s law to convert amplitudes. In this case, we would recognize that
V = ZI, and so we need to multiply our magnetic field amplitude by η0 to get the amplitude
for E1 . Combining these steps, we see that we arrive at the same result as explicitly using
Ampere’s law. We can of course also follow this process to find E2 , but we will leave that as
a practice exercise for the interested reader.
Now that we have our electric fields, we can expand the boundary condition on the
electric field to get another equation. In particular, the boundary condition

−ẑ × (E1 − E2 ) = 0 (1.208)

tells us that

Hy1 + Hy2 = 0. (1.209)

At this point, we can solve (1.204) and (1.209) together to give us the result that we guessed
earlier; i.e., Hy1 = −Hy2 = J0 /2.
We now summarize our final results. We have that the magnetic field is
J0

H1 (r, t) = ŷ cos(ωt + k0 z),
 z < 0,
H(r, t) = 2 (1.210)
H (r, t) = −ŷ J0 cos(ωt − k z), z > 0,

2 0
2
and the electric field is
J0

E1 (r, t) = −x̂ cos(ωt + k0 z),
 z < 0,
E(r, t) = 2 (1.211)
J
E (r, t) = −x̂ 0 cos(ωt − k z),

2 0 z > 0.
2
These are also shown in Fig. 1.25.
We can briefly look over these results to see if they make sense intuitively. The first thing
we see is that the waves have equal amplitudes and opposite propagation directions. This
matches our expectations, since we don’t have any reason to believe that the current sheet
should produce a stronger wave in one direction compared to the other. Next, we see that the
polarization of the electric field is the same on either side of the current sheet. Again, this
matches our expectations since the boundary condition told us that the tangential component
of the electric field should be continuous. To achieve the different propagation directions,
we see that the tangential component of the magnetic field is discontinuous. Again, this is
expected since this is exactly what the boundary condition tells us to expect.
This exercise of “checking your solution” is a good habit to get in. It is likely to serve
you well in your academic and professional career, especially if you ever find yourself running
numerical analysis programs. It is easy for these tools to accidentally be set up incorrectly.
Despite that, they will still typically run and produce some kind of solution. Doing a “sanity
check” that the solution makes sense is always a good idea to make sure that things are
working properly.

52
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.25: Illustration of the solution to the current sheet problem.

1.12 Plane Wave Scattering from a PEC Region (Nor-

mal Incidence)
We will now use the strategies we learned in the previous section to analyze the scattering
of a plane wave from a region containing a perfect electric conductor (PEC). To simplify the
analysis, we will only consider the “normal incidence” case. This case is when the isophase
surfaces of the plane wave are in the same plane as the outer surface of the PEC region.
Equivalently, we can say that there is no angle formed between the propagation direction of
the plane wave and the unit normal vector of the PEC region (“no angle” here can mean 0◦
or 180◦ ). We will consider the more general case, referred to as oblique incidence, later in
the course.
The problem we will solve in this section is illustrated in Fig. 1.26. In it, we have a
PEC region that begins in the z = 0 plane and extends into z > 0. We also have a plane
wave incident on the PEC region from the z < 0 portion of space. When we say that it
is “incident”, we mean that it is part of the specification of the problem. That is, we are
not concerned with how it was produced, all we care about is what must happen “next” to
ensure that Maxwell’s equations are satisfied in all regions of space given that this particular
specified field already exists. From an intuitive perspective, we have a wave approaching
an obstacle so we expect that some portion of the wave is likely to be reflected from the
boundary. In this section, we will show how we can use Maxwell’s equations to prove that
this is indeed the case (and also provide a more detailed description of how it specifically
reflects from the boundary).
For our current situation, the electric field of our incident plane wave is given by

Einc (z, t) = x̂E + cos(ωt − k0 z). (1.212)

We will analyze this problem in the frequency domain to show how our process from the
previous section looks in the frequency domain. Considering this, our incident plane wave

53
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.26: Illustration of a PEC region that begins in the z = 0 plane and continues into
the z > 0 region of space. There is an incident field approaching the PEC region from the
z < 0 region.

becomes
Einc (z) = x̂E + e−jk0 z . (1.213)
Let’s now apply the three-step process from the previous section. In Step 1, we need to
identify the distinct regions of the problem, the boundaries between them, and any sources
or existing fields within each region. In this problem, we have Region 1 with z < 0 and
Region 2 with z > 0. Region 1 is made up of free space and has the incident field Einc , while
Region 2 is filled with PEC. We have a boundary made of PEC at the z = 0 plane.
Step 2 can be approached in a couple ways. We will use the simplest way here, which
is to note that because we already have a plane wave that is incident on our boundary in
Region 1 we will go ahead and assume that we should also consider plane waves in the other
regions. Included in this, is noting that since we already have a plane wave propagating
in the +z-direction in Region 1 we will need to introduce another plane wave propagating
in the −z-direction. We will also assume that we have a plane wave propagating in the
+z-direction in Region 2. This would correspond to the transmitted field. Considering this,
we have the total electric field being given as
(
E1 (r, t) = x̂E + e−jk0 z + (x̂Ex− + ŷEy− )ejk0 z , z < 0,
E= + + −jkz
(1.214)
E2 (r, t) = (x̂Ex2 + ŷEy2 )e , z > 0.
Let’s dissect this for a moment. First of all, we have included two polarizations for the plane
waves with unknown constants because we don’t necessarily know at this point whether only
having x-polarized plane waves will be enough to solve the problem. We also see that we
have k in the second line instead of k0 . This is because this field is in a different medium
than free space. Although we haven’t specified an ϵ and µ for a PEC region, the important
thing to note here is that this k would still be calculated at the same frequency that we are
performing our analysis at since we have a linear, time-invariant system.

54
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We now turn our attention to Step 3, which covers solving for the different constants of
the problem using the information we have at the various boundaries of the problem. We
start out by writing the boundary condition down at z = 0, which is
n̂ × (E1 − E2 ) = 0. (1.215)
We can simplify this by noting that a PEC material is characterized by having “σ = ∞”.
Since J can’t take on an infinite value, Ohm’s law tells us that E must be equal to 0 in a
PEC region. Hence, we have that E2 = 0. Due to this, even if the PEC region only extends a
short distance into the z > 0 the physics of the scenario being considered will be unchanged.
As a result, it is common to just illustrate PEC “sheets” in drawings, like in Fig. 1.26. Now,
the boundary condition given in (1.215) then tells us that the total tangential electric field
in Region 1 must be equal to 0 at z = 0. This gives us
 
+ − −
−ẑ × x̂E + (x̂Ex + ŷEy ) = 0, (1.216)

which simplifies to
−ŷE + − ŷEx− + x̂Ey− = 0. (1.217)
The first thing we note from (1.217) is that Ey− must be 0 because there is no other term
with the same polarization left in the equation. The remaining part of the equation tells us
that Ex− = −E + .
When dealing with systems where reflections can occur, we generally will define a reflec-
tion coefficient. The exact definition can depend somewhat on the situation of interest, but
in general a reflection coefficient is defined as the ratio between the complex-valued ampli-
tudes of the reflected and incident fields referenced to a particular location in space. Where
this reference point makes sense to be taken at depends on the specifics of a problem. For
this problem, we will reference our reflection coefficient to z = 0. The reflection coefficient
is typically denoted with the symbol Γ. For this problem, we have
E − (z = 0)
Γ(z = 0) = , (1.218)
E + (z = 0)
which in general will be a complex-valued number. For passive systems (i.e., systems that
don’t provide some kind of amplification) the magnitude of the reflection coefficient will be
less than or equal to 1. That is, |Γ| ≤ 1. By convention, the reflection coefficient is always
specified in terms of the ratio of electric fields for this kind of problem. If you calculate
the reflection coefficient in terms of the magnetic field, you will get the wrong result! The
reflection coefficient is often a very important figure of merit of many different microwave
systems. We generally want to minimize the reflection coefficient in the design of transmission
line and antenna systems. We will learn much more about this during those portions of the
course.
Before finishing our discussion of this problem, it is useful to take one last look at the
final answer that we came up with. In particular, we have that
(
E1 (r, t) = x̂E + e−jk0 z − ejk0 z , z < 0,


E= (1.219)
E2 (r, t) = 0, z > 0.

55
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.27: Illustration of a standing wave produced by the reflected field.

The portion inside the parentheses in the first line should look familiar to you. In particular,
we can use Euler’s identities to rewrite this as a sine function. We end up with
(
E1 (r, t) = −x̂j2E + sin(k0 z), z < 0,
E= (1.220)
E2 (r, t) = 0, z > 0.
The form of the solution in the top line is often referred to as a standing wave. There are
many reasons for this name, and we will look into this more later in the course. The intuitive
reason is that this solution no longer shows us the same kind of wave propagation effect we
had seen previously. It has fixed points in space where the amplitude of the wave is always 0.
These “nulls” do not move in space as a function of time, and neither does any other part of
the wave. Hence, it looks like it is “standing” in space. An illustration of the standing wave
pattern is shown in Fig 1.27. This is an example of wave interference, which is a broadly
useful resource. As you will see, much of engineering with waves involves carefully interfering
them to achieve a desired outcome.
Before concluding, it is worth looking at why we used the boundary condition for the
tangential components of E rather than the tangential components of H. The reason lies
in that at a PEC boundary the tangential magnetic field is discontinuous. In particular, it
goes from having a non-zero finite value on one side of the boundary to immediately having
a value of zero in the PEC region. As we discussed previously, this kind of discontinuity
can only occur if there is a surface current flowing along the PEC boundary. Since we don’t
know what value the surface current has going into the problem, we can’t use
n̂ × (H1 − H2 ) = Js (1.221)
to solve for H1 (noting that H2 = 0 in the PEC region). However, now that we have solved
the problem, we can easily use
n̂ × H1 = Js (1.222)
to compute the surface current from the solved for fields.

56
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

1.13 Lorentz Force and Material Response

We have already discussed informally in this class that the electric field can apply a force to a
charged particle, and consequently, put it into motion. In addition to this, the magnetic flux
may also influence the motion of a charged particle. These effects are captured in the Lorentz
force, which describes how the electromagnetic fields influence the motion of a point charge.
In this course, we are not often concerned with looking at the motion of individual electrons
(this is something physicists tend to be more interested in, and can also be important in
the design of certain kinds of traveling wave tube amplifiers that are needed in high-power
communication systems). However, the Lorentz force provides a useful way to determine
the material response of certain systems, such as plasmas. Surprisingly, we will see shortly
that even using relatively simple equations can often provide a reasonable description of the
physics. These simple models have been leveraged to great effect to analyze many different
physical systems.
Overall, the complete Lorentz force is given by
F = qE + qv × B, (1.223)
where F is the total force applied to the point charge that has velocity v and charge q. We
have previously discussed how charges and currents can produce electromagnetic fields, now
we see how these fields can influence charges.

1.13.1 Plasma Material Response

We now turn our attention to using the Lorentz force to determine the material response of a
plasma medium. A plasma can often be viewed as an ionized gas with a number of “species”
of ions present. Typically, the electrons are substantially lighter than the positively-charged
ions that also make up the plasma. As a result, the electrons are generally the predominant
component of the plasma that the electromagnetic field will interact with. Hence, we will
develop an equation of motion for the electrons in the plasma medium and see how this can
be used to derive an effective material response/description of the plasma.
Since we are dealing with a classical, non-relativistic system, our equation of motion for
each electron will simply be Newton’s law. We typically write this for a plasma as
mr̈ + mγ ṙ + κr = qE. (1.224)
The first term in (1.224) is the acceleration of the particle, the second term is a fric-
tion/collision term that accounts for loss in the system, and the third term is a restoring
force similar to Hooke’s law. For simplicity, we have omitted the magnetic flux density. This
can often be acceptable when the velocity of the charge is small and the magnetic field is
relatively weak (it is often weaker than the electric field since |B| = µ|H| = µ|E|/η for a
plane wave). These assumptions are generally good approximations for a propagating plane
wave. However, in more general applications where magnetic fields can be applied from other
sources than the wave itself the more complete description of the full Lorentz force can be
needed.
Depending on which terms in (1.224) are kept, we can arrive at a number of different
models for material response that are appropriate for different situations. We will start

57
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

our analysis by considering all of the terms and then looking at a particular special case of
interest from there. The easiest way to analyze this system is to consider a time harmonic
case. Then we can apply a Fourier transform to arrive at
q
−ω 2 r + jωγr + ω02 r = E, (1.225)
m
p
where ω0 = κ/m. We can solve for r to get

1 q
r= E. (1.226)
ω02 2
− ω + jωγ m

Now, recall from our discussions about the polarization intensity of a material that we can
treat the displacements of the charged particles as small dipole moments given by q r. If we
sum the effects of all of these little dipoles up we get the overall polarization intensity, from
which we can identify the electric susceptibility of the material. In particular, if we have N
electrons per unit volume we will have that

1 N q2
P = N qr = E. (1.227)
ω02 − ω 2 + jωγ m

Recalling that we also have

P = ϵ0 χe E, (1.228)

we see that the electric susceptibility is going to be given by

N q2 1
χe = 2
. (1.229)
ϵ0 m ω0 − ω 2 + jωγ

We will now consider the particular case where the collision frequency γ and restoring
force can be neglected. This can occur, for example, when the density of the plasma is very
low (this case is sometimes referred to as a cold collisionless plasma). Under this approx-
imation (since it is always an approximation), our electric susceptibility can be simplified
greatly to be

N q2 ωp2
χe = − = − 2, (1.230)
ϵ0 mω 2 ω

where
s
N q2
ωp = (1.231)
mϵ0

is called the plasma frequency. The plasma frequency corresponds to the natural oscillation
frequency of the plasma medium.

58
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

(a) (b)

Figure 1.28: (a) Illustration of an over-the-horizon radar system interrogating a location

that is not visible to a line-of-sight radar wave propagation. (b) The low operating frequen-
cies of over-the-horizon radars require massive antennas to operate effectively (images from
Wikipedia [5]).

We get different kinds of material responses depending on the value of the electromagnetic
field’s frequency relative to the plasma frequency. To see this easily, we recall that the relative
permittivity of the medium can be described by

ωp2
ϵr = 1 + χe = 1 − , (1.232)
ω2
where ω is the frequency of the electromagnetic wave. Correspondingly, our propagation
constant for a plane wave in this plasma medium would be
r
√ ωp2
γ = jω µ0 ϵ0 1 − 2 . (1.233)
ω
If ω < ωp , the propagation constant becomes a purely real number, and as a result the plane
wave will be quickly attenuated as it travels into the plasma medium. The name for this
kind of behavior is an evanescent wave. Although we won’t look at this in detail right now,
this case also often results in a large reflection of the plane wave off of the “boundary” of
the plasma medium since it isn’t able to propagate into the medium. This can occur off
of the ionosphere for particularly low frequency waves (generally less than 40 MHz), and is
leveraged to build over-the-horizon radar systems. These systems bounce radar waves off of
the ionosphere so that they can search for targets of interest that would otherwise not be
visible for a standard line-of-sight radar operation (see Fig. 1.28).
On the other hand, if we have that ω ≫ ωp , the relative permittivity is approximately 1
and the plasma has almost no effect on the propagating wave. Since satellite communication
signals are operated well above the plasma frequency, they can typically propagate through
the ionosphere without any issue. However, the ionosphere is a complicated system with
the plasma density always slowly varying depending on the time of day, solar wind effects,
etc. As a result, the plasma density can still impact the propagation of electromagnetic
waves in important ways. For instance, global navigation systems (e.g., GPS) are only able
to achieve the level of accuracy that we have grown accustomed to on a day-to-day basis

59
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

by constantly calculating updates to internal error models to approximately correct for the
ionospheric propagation effects. The correction terms for these error models are calculated
at well-surveyed locations (often cell phone network towers), which then can distribute these
corrections to nearby users connected to the network. Without these corrections, the GPS
positioning accuracy for standard users can diminish by multiple meters (sometimes much
more in extreme cases).

1.13.2 Other Material Response Models

As mentioned previously, depending on which terms of (1.224) we keep in our description
of a material we can arrive at various other material responses. One common model is the
Drude model for metals. This case assumes that the collision force dominates, and leads to
a description of a conduction current and conductivity of a material. Other models include
that of a Lorentz oscillator, which corresponds to keeping the acceleration and restoring force
terms; as well as the Debye relaxation model, which keeps the friction/collision and restoring
force terms. These models are useful for describing the particular dispersion profiles (more
on this shortly) of various kinds of materials.

1.14 Polarization of Electromagnetic Waves

Up to this point in the course, we have occasionally mentioned in passing how a particular
plane wave is polarized. Since we were only dealing with rather simple descriptions of plane
waves, we did not need to be very detailed in our descriptions of what this means exactly.
However, the polarization of electromagnetic waves brings with it incredibly rich and useful
physics that can be leveraged to great effect in many application areas. We will now turn
our attention to providing a more complete description of electromagnetic wave polarization,
followed by a brief look at how this important phenomenon can be utilized. Following that,
we will finish introducing the essential topics related to the polarization of electromagnetic
waves.

1.14.1 Introduction to Polarization

The polarization of a plane wave is defined by the locus of points that the tip of the electric
field intensity vector traces out as a function of time with space fixed. In particular, the locus
of points is traced out in a plane orthogonal to the propagation direction. Let’s take a closer
look at our standard plane wave example to see how this definition works out.
In particular, we have

E(r) = x̂E0 e−jkz . (1.234)

We want to understand the behavior in time, so we will convert our phasor back into the
time domain by multiplying by exp jωt and taking the real part. This gives us

E(r, t) = x̂|E0 | cos(ωt − kz + φ), (1.235)

60
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.29: Illustration of linear polarization for a plane wave propagating in the +z-
direction.

where φ is any phase factor that may have been embedded in the complex-valued E0 . For
now, we will assume that φ = 0 to keep things simple. Further, we will look at the polar-
ization in the plane z = 0. It is conventional to look at how the polarization varies from the
perspective of an observer that the plane wave is traveling toward. Since our current plane
wave is propagating in the +z-direction, we will be looking at the z = 0 plane from above
when we plot the electric field vector as a function of time.
An illustration of how the tip of the electric field vector will move around in the z = 0
plane as a function of time is shown in Fig. 1.29. As expected from the cosine function in
(1.235), the tip of the electric field vector simply oscillates back and forth between x = −|E0 |
and x = |E0 |. Since the shape in the z = 0 plane is a line, we refer to this kind of polarization
as linear polarization. We would refer to this plane wave as being linearly polarized.
We can of course also have a plane wave polarized along the y-axis as well. This function
would look like

E(r, t) = ŷ|E0 | cos(ωt − kz + φ). (1.236)

The only change to our illustration in Fig. 1.29 would be that the line would now be varying
along the y-axis instead of the x-axis.
What happens if we combine two linearly polarized plane waves? Is the resulting polar-
ization still linear? The answer depends on the relative phase between the two plane waves.
For instance, if we have a relative phase of 0, our plane wave can be represented as

E(z = 0, t) = x̂|E0 | cos(ωt) + ŷ|E0 | cos(ωt), (1.237)

in the z = 0 plane. If we draw this out in the z = 0 plane we get the result shown in Fig.
1.30. We see that the result is still a line, so this combination of plane waves is still linearly
polarized. If we were to change the magnitude of the amplitudes between the two plane
waves, the angle the line would make with the x-axis would change. In the end, we find that
if there is no relative phase difference between the two plane waves with orthogonal spatial

61
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.30: Illustration of a combination of two plane waves with orthogonal linear polar-
izations and no relative phase difference propagating in the +z-direction.

polarizations the resulting total plane wave (i.e., the summation of the two components) will
be linearly polarized.
However, if we have a non-zero relative phase difference between our plane waves the
story can be different. For instance, if we have that our y-polarized plane wave has a leading
phase of 90 degrees our expression for the electric field becomes

E(z = 0, t) = x̂|E0 | cos(ωt) + ŷ|E0 | cos(ωt + π/2)

(1.238)
= x̂|E0 | cos(ωt) − ŷ|E0 | sin(ωt).

If we trace out the tip of the electric field vector for this case, we get the result shown in
Fig. 1.31. We see that the angle the field vector makes with respect to the x-axis is now
changing as a function of time. Further, because the magnitudes of the two plane waves
are equal to each other we end up with the tip of the electric field vector tracing out a
circular shape. Hence, we refer to this as being a circular polarization. However, this is not
a full specification of our polarization. We also must relate the sense of the rotation. For
this particular case, we see that it is following a clockwise rotation. Alternatively, we can
also call it a left-hand circular polarization because when the fingers of the left hand curl
to follow the sense of the rotation the thumb points in the direction of propagation. It is
important to remember that this sense of polarization must be specified from the perspective
of an observer that the wave is traveling toward.
If we instead have a lagging phase of 90 degrees, the sense of our rotation will change.
To see this, we note that the expression for our electric field now becomes

E(z = 0, t) = x̂|E0 | cos(ωt) + ŷ|E0 | cos(ωt − π/2)

(1.239)
= x̂|E0 | cos(ωt) + ŷ|E0 | sin(ωt).

If we trace out the tip of the electric field vector for this case, we get the result shown in
Fig. 1.32. However, this is not a full specification of our polarization. We see that the sense

62
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.31: Illustration of a combination of two plane waves with orthogonal linear polar-
izations and a 90◦ relative phase difference propagating in the +z-direction. The y-polarized
wave has a leading phase difference so we get a clockwise/left-hand sense to our rotation.

of the rotation is following a counterclockwise rotation. Alternatively, we can also call it a

right-hand circular polarization because when the fingers of the right hand curl to follow the
sense of the rotation the thumb points in the direction of propagation. An illustration of
how right- and left-hand circularly polarized waves propagating through space look is shown
in Fig. 1.33. As a final note, we emphasize that for a circularly polarized wave to exist,
we must have two components that have equal magnitude and are orthogonal in time and
space.
In general, it takes a very special kind of source to produce only a single linear polarization
or to produce a pure form of circular polarization. More generally, the non-ideal nature of
realistic systems means we will have some perturbation from these desired results. This can
come in the form of cross polarization (unintended production of an orthogonal polarization),
or by corrupting the polarization that was produced. This leads us to a very general type
of polarization, known as elliptical polarization. Elliptical polarization will occur if we have
two spatially orthogonal components that do not meet all of the conditions for being either
linearly or circularly polarized. An illustration of an elliptically polarized wave is shown in
Fig. 1.34.
There are a few important parameters used to characterize an elliptically polarized wave.
First, the amplitudes of the semi-minor and semi-major axes. These will often be reported
together in the axial ratio of the wave, which is equal to the ratio between the minor and
major axis of the polarization ellipse. Due to the difficulty in generating pure circularly
polarized waves, practical applications will generally specify an axial ratio bandwidth as being
the frequencies for which the axial ratio is less than 3 dB. Second, the angle of inclination (ψ
in Fig. 1.34) is defined, which gives the angle between the semi-major axis and the x-axis.
Finally, the sense of rotation also needs to be specified in the same way that it was for a
circularly polarized wave.
It is often a useful skill to be able to quickly determine the polarization of a wave by

63
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.32: Illustration of a combination of two plane waves with orthogonal linear polar-
izations and a 90◦ relative phase difference propagating in the +z-direction. The y-polarized
wave has a lagging phase difference so we get a counterclockwise/right-hand sense to our
rotation.

Figure 1.33: Illustration of propagating right- and left-hand circularly polarized waves (im-
ages from [6]).

looking at its phasor representation. We will look at a few cases here to get the general idea.
 
ˆ E(z) = x̂|E1 | + ŷ|E2 | e−jkz , |E1 | =
̸ |E2 | =⇒ Linear Polarization

64
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.34: Illustration of a general elliptical polarization (image from Wikipedia [7]).

 
ˆ E(z) = x̂|E1 | + ŷ|E2 | e−jkz , |E1 | = |E2 | =⇒ Linear Polarization
 
ˆ E(z) = x̂|E1 | + ŷj|E2 | e−jkz , |E1 | = |E2 | =⇒ Left-Hand Circular Polarization
 
ˆ E(z) = x̂|E1 |− ŷj|E2 | e−jkz , |E1 | = |E2 | =⇒ Right-Hand Circular Polarization
 
ˆ E(z) = x̂|E1 | + ŷj|E2 | e−jkz , |E1 | =
̸ |E2 | =⇒ Left-Hand Elliptical Polarization
 
ˆ E(z) = x̂|E1 |− ŷj|E2 | e−jkz , |E1 | =
̸ |E2 | =⇒ Right-Hand Elliptical Polarization

Although it is good to be able to quickly identify these, you can always convert them back to
a time domain representation and plot a few points to identify the polarization of the wave
you are given.
Before completing our discussion of polarization, it is useful to note that we can always
decompose a generally polarized wave into some set of orthogonal components. In general,
we can think of two orthogonal polarizations being able to be used as a basis to describe
other polarizations. For instance, in the above exercise we very clearly see that we can
always decompose these waves into two orthogonal linear polarization components. It is also
common to decompose waves into orthogonal circular polarizations, where right-hand and
left-hand circularly polarized waves form our basis. Other options can be used as well, but
these are the most often used in practice.

1.14.2 Applications of Polarized Waves

We are often most interested in polarization in the context of antennas, since these are
typically the devices that we will use to generate and receive plane waves. A quick way to

65
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.35: A crossed dipole antenna (image from Wikipedia [8]).

see how polarization impacts antenna design is by recalling from our discussion on Poynting’s
theorem that the power going into/out of the source function J was modeled as E·J∗ . Hence,
the electric field needs to have a polarization that matches the direction of the current flow
to be able to extract any power from the currents.
This has a few practical consequences. First, if we can make an antenna that predomi-
nantly has all of its current only able to flow along a single direction it will only be able to
transmit or receive the components of plane waves polarized along the direction of the cur-
rent flow. A good example of this kind of antenna is a wire antenna – like a dipole antenna.
If we take two dipoles and orient them orthogonally to each other we get a crossed dipole
antenna, as seen in Fig. 1.35. This allows us to break the signals we are receiving into their
two orthogonal constituents.
Why might we be interested in doing this? One example is in the area of wireless
communications. Our ability to transmit information simultaneously is dependent on the
number of sufficiently independent channels we have access to. A classic example would be a
frequency division multiplexing scheme where channels can be made up of various dedicated
portions of the electromagnetic spectrum. However, because our two orthogonal antenna
channels are independent of each other, we can effectively double our number of channels by
making use of each polarization. This is sometimes referred to as polarization diversity, and
is often an important component of making practical multiple-input multiple-output (MIMO)
communication systems that are integral parts of things like 5G networks.
In communication systems with particularly challenging isolation requirements, polariza-
tion diversity can be a vital tool in lowering the coupling between communication channels.
An excellent example is in the area of satellite communications. The huge distance between
a satellite and the earth results in massive propagation losses. For instance, in S-band (say
2 GHz), the loss to a typical low earth orbit is approximately 155 dB while the loss to a
geostationary orbit is approximately 185 dB. As a result, the receivers on a satellite must be
exceptionally sensitive. Likewise, the transmitter power on the satellite must also be large
so that users on the Earth (who typically have much cheaper receivers, e.g., in their phone)
can receive the signals. Locating a sensitive receiver next to a high power transmitter is

66
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.36: Example of a traditional “spotlight mode” SAR image of some buildings, park-
ing lots, and New Mexico terrain. Courtesy of Sandia National Laboratories, Radar ISR.

a challenging engineering problem that can typically require multiple levels of isolation to
meet system requirements (e.g., 100 dB of isolation). In these systems, polarization diversity
between the transmit and receive channels can be an important component of the overall
isolation strategy.

Polarization diversity can also be a useful tool in areas like remote sensing. A great
example is in synthetic aperture radar (SAR) systems. These systems coherently process
many different radar pulses together to form images of an area being investigated (this
requires a combination of careful timing and processing of phase shifts related to the Doppler
effect). A typical processed SAR image can require careful training to be able to properly
interpret since they aren’t as intuitive as optical images that we are used to looking at. An
example of a traditional SAR image is shown in Fig. 1.36. However, if you instead use a
polarimetric SAR sensor that transmits and receives the SAR signals from two orthogonal
polarizations you can make much richer images. This is possible because the manner in
which the polarization of a wave changes after interacting with an object can be indicative
of certain properties about the target being observed. This kind of information can be added
into a SAR image as “colors” to help the viewer better interpret what different parts of the
image may correspond to. For instance, a colored polarimetric SAR image is shown in Fig.
1.37. We can see that this image has much richer content and more clearly differentiates
between different kind of terrain (e.g., trees vs. buildings, grass vs. road, etc.)

67
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.37: Example of a polarimetric SAR image of some buildings, parking lots, and New
Mexico terrain. The coloring is done through processing of the polarization information
received by the radar. Courtesy of Sandia National Laboratories, Radar ISR.

1.15 Doppler Effect

The Doppler effect is a general wave phenomena that occurs when there is relative motion
in the direction of wave propagation between a transmitter and receiver of a wave. You are
familiar with this effect in everyday life when an emergency response vehicle with its sirens
on drives past you. As the vehicle comes toward you, the sirens have a higher pitch and as
the car moves past you the siren begins to have a lower pitch as it drives away. The general
idea is that because the transmitter or receiver is moving relative to the wave propagation,
the wave becomes stretched or compressed, effectively changing the frequency of the overall
wave. This is illustrated in Fig. 1.38.
With this in mind, we will first look at the derivation of the Doppler effect that is
applicable for most applications in Section 1.15.1. Following this, we will look at a special
case of how it is used in Doppler radars in Section 1.15.2.

68
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.38: A simplified illustration of the Doppler effect. Note that the blueshift and
redshift terminology is common in the physics community, and is in reference to the extremes
of the visible light spectrum. Relativistic Doppler effects have important consequences in
general relativity and our ability to interpret details about the universe from measurements
of received electromagnetic energy on Earth.

(a) (b)

Figure 1.39: Geometry for the derivation of the Doppler effect: (a) initial time and (b) after
some additional time has passed.

1.15.1 Derivation
To derive the Doppler effect, we will assume that we have a moving transmitter and a
stationary receiver. The transmitter is initially located a distance R1 away from the receiver.
The transmitter is moving with speed v at a relative angle of θ to the line connecting the
initial locations of the transmitter and receiver. This is illustrated in Fig. 1.39(a). If we
consider the portion of the wave that is emitted at the initial time t = 0, we can see that it
will arrive at the receiver at time
R1
t1 = , (1.240)
c
where c is the speed of the light.
If we now consider the case at a later time, t = ∆t, we see that the transmitter will be
at a new position (see Fig. 1.39(b)). The wave emitted from this position will reach the
receiver at time
R2
t2 = ∆t + , (1.241)
c
where R2 is the distance between the transmitter and receiver at t = ∆t. Of course, the
distance R2 can be related back to R1 and the parameters describing the motion of the

69
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

transmitter. This requires a bit of trigonometry, but the overall result is that
q
R2 = R12 − 2R1 (v∆t) cos(θ) + (v∆t)2 . (1.242)

If we substitute this into (1.241), we get a bit of an ugly expression that is

1
q
t2 = ∆t + R12 − 2R1 (v∆t) cos(θ) + (v∆t)2 . (1.243)
c
However, we can simplify the final result into a nicer form if we make a useful approximation
at this point. In particular, for most normal operating conditions we will have that v ≪ c,
since it is very difficult to move at a speed close to the speed of light. As a result, if we only
keep terms that are “first order” in the parameter v/c we can simplify this to be
s
R1 2(v∆t)
t2 = ∆t + 1− cos(θ) (1.244)
c R1

Since the second term

√ in the square root is still a small number, we can use a small argument
approximation to 1 + x to get
 
R1 (v∆t)
t2 = ∆t + 1− cos(θ)
c R1
(1.245)
R1 (v∆t)
= ∆t + − cos(θ).
c c
At this point, we can determine what the difference in receive times look like at the
receiver. This gives

∆t′ = t2 − t1
(1.246)
 
v
= ∆t 1 − cos(θ) .
c

If we assume that ∆t was a full period of the transmitters harmonic waveform we can arrive
at the perceived receive frequency to get

1 f
f′ = ′
= v . (1.247)
∆t 1 − cos(θ)
c
For standard cases where v/c ≪ 1, this can be further simplified to give us that
 
′ v
f ≈ f 1 + cos(θ) . (1.248)
c

This gives us the expected result that if the transmitter is coming toward the receiver (θ = 0)
the perceived frequency is higher, while if the transmitter is moving away from the receiver
(θ = 180) the perceived frequency is lower.

70
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

1.15.2 Doppler Radar

In this section, we will look at a simplified schematic of a Doppler radar to get a better
understanding of how a real device can use the effects derived in the previous section. The
goal of a Doppler radar is to use the Doppler shift of a moving target to determine how
fast it is moving. Hence, we are hoping to “isolate” the amount of frequency shift that has
occurred. We will denote the Doppler shift as fd in this section.
A simplified schematic of how we can isolate the Doppler shift is shown in Fig. 1.40.
In interpreting this schematic it is important to note that we are dealing with waves. As
a result, we can use various components to “direct” the waves through our system in a
desirable way so that is has a particular “flow” through the schematic. Hence, it is quite
different from a traditional circuit schematic where you would consider voltages and currents
instantaneously interacting between every component of the circuit.

Figure 1.40: A simplified schematic of a continuous wave Doppler radar.

With this in mind, we begin to work our way through the schematic. The first component
is the local oscillator (LO), which is our source for the radar. It produces a sinusoidal
waveform with frequency f . The LO is connected to a directional coupler, which is a special
kind of device that allows us to split our wave into two components. One component has a
much higher power than the other component by design. Our higher power signal will be
what we transmit at our target, while the lower power component will be “saved” in our
system to use as a reference signal later on.
We will now follow the path of our higher power signal that we will transmit at the
target. It next goes to a device called a circulator. This is a kind of non-reciprocal device
that provides different behavior depending on which direction a wave approaches it from.
As the symbol suggests, it takes a signal entering one port and “directs” it to exit from the
next port in a clockwise fashion. Nothing goes out of the port on the counter-clockwise port.

71
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

The utility of a circulator is that it allows us to separate the signals that the antenna will
transmit and receive.
After the transmit signal flows out of the circulator, it reaches the antenna and is trans-
mitted toward the target. The transmitted wave will reflect off of the target and pick up a
Doppler shift in its frequency based off of the speed of the target. This reflected signal prop-
agates back to the antenna and is received. The received signal propagates back toward the
circulator. It is directed through the circulator in a clockwise fashion and exits the device.
This received signal is then fed into a mixer. At the other port of the mixer we add in
our reference signal that we had coming out of the directional coupler. The mixer performs
a multiplication operation between the two input signals. The output leaves the mixer from
a third port. If we expand the multiplication of our two input signals we get
 
1


cos(f ) cos(f ± fd ) = cos f − (f ± fd ) + cos f + (f ± fd )
2
 
1


= cos ∓ fd + cos 2f ± fd (1.249)
2
 
1


= cos fd + cos 2f ± fd .
2

We see that we get one sinusoid oscillating at the Doppler shift frequency and another
sinusoid oscillating at a much higher frequency. It is generally much easier to process lower
frequency signals (e.g, digitizing them), so we will want to isolate this part only. This can be
readily accomplished by using a low pass filter. We can then feed this signal into a variety
of different processing electronics to make our final detection of the Doppler shift.

1.15.3 Introduction to Synthetic Aperture Radar

As we discussed previously, a synthetic aperture radar (SAR) is a type of imaging radar. It is
able to form images due to a combination of clever processing techniques involving Doppler
shifts and pulsed radar waveforms. The general idea is shown in Fig. 1.41.
A SAR system generally involves a moving platform that is attempting to image a sta-
tionary scene. The relative motion between the transmitter and the various points in the
scene being imaged leads to a Doppler shift on the reflected signals. Since the transmitter
antenna illuminates a relatively large imaging scene, a single transmit pulse will interact
with many points in the scene. The reflected signal at each point of the scene will have some
amount of Doppler shift when it is received due to the relative motion of the transmitter
(which also acts as the receiver). However, this Doppler shift is not unique, since each point
in the scene at a fixed angle from the transmitter will pick up the same amount of relative
motion. Hence, we cannot use a Doppler shift alone to try and discriminate between the
energy reflected from various points in the illuminated scene.
To provide more unique information, we transmit our waveform in pulses where the
waveform is periodically turned off. This allows us to now also time when signals are received.
This alone does not provide us with enough unique information to locate where a pulse came
from because each point that is the same distance away from the transmitter will produce
the same time delay. However, by combining the information from the time delay and the

72
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.41: Illustration of a single SAR pulse geometry.

Doppler shift we can begin to have unique enough information to determine where a signal
came from.
Although this provides useful information, it often is not able to provide a sufficiently
high enough resolution in the image to be particularly useful. The key to SAR is by observing
the same scene over multiple transmit pulses as the transmitter moves around the extent of
the scene. By coherently processing the results from all of these different transmitter pulse
locations, we can provide a much higher resolution image. This is what leads to the exquisite
kinds of images that we looked at previously in the course.
This kind of “multi-look” imaging approach to improve resolution is a very common
technique. For instance, a computed tomography (CT) scan does a similar approach with
X-rays. The sensor is rotated around the person to take cross-sectional images at various
points. The results of all the scans can be combined using various reconstruction algorithms
to form a complete image of the object being analyzed.

1.16 Dispersion and Group Velocity

When we originally discussed plane waves propagating in lossless media, we introduced the
concept of phase velocity. This was the velocity at which a constant phase point on our wave
moved. We saw that for an unbounded and lossless medium, the phase velocity could be
expressed as
ω
vp = , (1.250)
β

where β was the phase constant of the wave. In reality, this formula for the phase velocity
holds in more general cases so long as we can find the phase constant. However, for the

73
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.42: Example of a pulse distorting as it propagates through a dispersive medium.

√
simple case of an unbounded and lossless medium, we had that β = ω µϵ. Since this β is a
linear function of ω, we saw that the phase velocity became a constant number as a function
of frequency in this situation. Further, the phase velocity was equal to the speed of light.
Previously, we have seen that many materials will necessarily have some kind of frequency
dependence in their permittivity (or permeability). As a result, the phase constant will no
longer have a pure linear dependence on frequency in most materials. Later in the course,
we will also discuss the case where a geometry causes the phase constant to not linearly
depend on frequency. Regardless of the source of the augmented frequency dependence, the
end result is that waves with different frequencies will have constant phase points propagate
with different phase velocities. This effect is known as dispersion, because it causes the
shape of a signal composed of multiple frequency components (which is essentially every
practical transmitted signal) to distort/disperse as it propagates (see Fig. 1.42). As a result,
a medium that has a frequency-dependence in its material properties is known as a dispersive
medium.
Let’s now take a closer look at what happens to a signal as it propagates through a
dispersive medium. To start, we will assume that we have a relatively narrowband signal.
This is common in wireless communication systems because the amount of data that needs to
be transmitted typically occupies a small bandwidth compared to the carrier frequency that
the information is transmitted on. Let’s consider the simplest case of a wave packet composed
of two slightly different frequencies equally offset from a particular center frequency (denoted
by ω0 ).
For this situation, our electric field can be expressed as
   
E(z, t) = E0 cos [ω0 + ∆ω]t − [β0 + ∆β]z + E0 cos [ω0 − ∆ω]t − [β0 − ∆β]z , (1.251)

where ∆ω ≪ ω0 and ∆β is the corresponding change in β as β : β(ω0 ) → β(ω0 ± ∆ω). Note

that we are implicitly assuming that we can treat β as being a linear function locally to ω0
when we assume that ∆β is appropriate for both the positive and negative frequency shift.
Hence, we can see that we are using a local Taylor series representation of β in our discussion
here. We restricted our discussion to narrowband signals so that this approximation is
reasonable.
Now, we can simplify our expression for E(z, t) by using standard trigonometric identities.
These give us
E(z, t) = 2E0 cos(∆ωt − ∆βz) cos(ω0 t − β0 z). (1.252)

74
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.43: Illustration of phase and group velocities for a simple wave packet. The envelope
is shown with dashed lines, while the compete signal is shown with a solid line.

Since by assumption ∆ω ≪ ω0 , we see that our overall wave packet can be decomposed into
two parts. The second cosine function has a fast oscillation at the center frequency of the
wave. This response is modulated by a slowly-varying function with frequency dependent
on ∆ω. This is usually referred to as an envelope function. A simple illustration of this
situation is shown in Fig. 1.43.
We now see that the wave inside the envelope propagates with a phase velocity of ω0 /β0 .
We can also look at the velocity of the envelope by setting its argument equal to a constant
function. Assuming this is 0 for simplicity, we have that

∆ωt − ∆βz = 0, (1.253)

from which we can define the group velocity as

∆ω
vg ≈ . (1.254)
∆β

We can take a limit to convert these ∆’s into a derivative. We typically write this as
 −1
dβ
vg = . (1.255)
dω

We write this in this somewhat odd form because we generally are able to compute β as
a function of ω, rather than the other way around. We see that this result matches the
intuition that we had about a Taylor series representation from earlier. In particular, we
are using a first order (linear) Taylor series result to approximate the local behavior of β to
estimate the velocity that the overall wave packet moves at. Although this is usually how
vg is written, it will also sometimes be written as dω/dβ as well.

75
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Let’s now consider the particular case of the simple plasma medium we considered pre-
viously (i.e., no restoring force or loss). For this situation, we have that
s  2
ω ωp
β= 1− . (1.256)
c0 ω

We can easily find that the phase velocity is given by

ω c0
vp = =s  2 . (1.257)
β ωp
1−
ω

As discussed previously, we will have wave propagation when ω > ωp . We see that this
corresponds to a phase velocity that is larger than the speed of light. Is this an issue? It
turns out that it isn’t, mostly because the phase velocity is not the velocity that information
travels at. When a phase point approaches the “edge” of the wave packet envelope it isn’t
able to move outside of or past the envelope. Hence, that phase point has reached the end
of the packet and no longer is of interest to be tracked.
For a narrowband signal, “information” will propagate at the group velocity. For this
situation, we get that
s  2
ωp
vg = c0 1 − , (1.258)
ω

which is less than c0 for our case of ω > ωp . Hence, we have no “issues” with one of the
central principles of special relativity (that information can’t propagate faster than the speed
of light).
There are many “tools” that we can use to help visualize and understand the behavior
of dispersive media. One of the common ones is a dispersion diagram. These can come in
different forms, but generally involve plotting either ω as a function of β or vice-versa. An
example of a dispersion diagram (also called an ω-β diagram for this case) for our plasma
case is shown in Fig. 1.44.
Dispersion diagrams are particularly useful when we consider more complicated propaga-
tion environments. In these situations, it is common that only a few very specially structured
“modes” will be valid solutions to Maxwell’s equations. We can plot a trace on a single dis-
persion diagram for each mode we are considering to understand how energy will be able to
travel through our medium. This kind of tool is very prevalent in the analysis of different
kinds of periodic structures, such as metamaterials, photonic crystals, and electromagnetic
bandgap structures.
It has become such an important topic that researchers have begun discussing dispersion
engineering. This is a field aimed at designing different materials to have desirable dispersion
characteristics. For instance, one may wish to design a material that compensates for the
dispersion of another natural material so that when a signal propagates through both of
them the end result is a wave packet that has significantly reduced distortion (see Fig. 1.45).
Other areas of interest are making bandgaps where no signal is able to propagate through the

76
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.44: Example of a dispersion diagram for a simple plasma medium. The slope of the
line running from the origin to the current ω point gives the phase velocity. Similarly, the
slope of the tangent line to the curve gives the local group velocity.

Figure 1.45: Example of dispersion compensation to correct the properties of a pulse. (a)
Input pulse, (b) pulse after propagating through a long dispersive medium, and (c) pulse
after applying dispersion compensation (images from [9]).

medium. This can be useful for lowering the interactions between nearby electromagnetic
devices, such as antennas. This is a technique that has garnered reasonable research interest
for designing MIMO (multiple-input, multiple-output) antenna systems, for example (see
Fig. 1.46).

1.17 Plane Waves Propagating in Arbitrary Directions

Up to this point, we have typically used relatively simple descriptions of plane waves. In
particular, we always considered them to be propagating along a particular axis of our
coordinate system so that we could treat our wave equation as a one-dimensional problem.
However, there are many practical cases where it is easiest to approach them by considering

77
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.46: Example of an electromagnetic bandgap (EBG) structure to lower the electro-
magnetic interference between nearby antennas. (a) Geometry of the antenna array with
EBG, (b) surface currents without and (c) with the EBG (images from [10]).

a plane wave that is propagating in an arbitrary direction, potentially having a component

propagating along all of our coordinate axes. A good example of when this is useful is in
considering the oblique scattering of a plane wave off of a planar interface (we will consider
this problem in upcoming lectures).
We will now consider how to approach this more general problem. This will also give us
our first look at one of the very common approaches at solving complicated partial differential
equations – the separation of variables strategy. To begin, let’s assume that we are trying to
solve the vector Helmholtz wave equation in free space. We then have that

∇2 E(x, y, z) + k02 E(x, y, z) = 0. (1.259)

As mentioned earlier, we are going to assume that E can have a spatial dependence along
all 3 of our coordinate axes. In order to make the solution of the problem more tractable,
we are going to see if we can find a solution that has the form

Ψ(x, y, z) = X(x)Y (y)Z(z), (1.260)

where Ψ could be any of Ex , Ey , or Ez . That is, we are looking for a solution that “cleanly”
factorizes into a function that only depends on x, another that only depends on y, and yet
another that only depends on z. As we will see shortly, searching for a solution of this form
leads to a greatly simplified differential equation. This type of solution approach is what is
referred to as using the separation of variables.
Let’s now plug (1.260) into our wave equation. Recalling that we are in Cartesian coor-
dinates and that Ψ is a scalar function, we see that we get

∂xx XY Z + ∂yy XY Z + ∂zz XY Z + k02 XY Z = 0.

(1.261)

We can simplify our derivatives by noting that they each will only apply to one of the
functions in each of the parentheses. Hence, we get

Y Z∂xx X + XZ∂yy Y + XY ∂zz Z + k02 XY Z = 0. (1.262)

78
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Next, we divide our equation by XY Z to arrive at the desired form of equation for the
separation of variables strategy to be useful. This is

1 d2 1 d2 1 d2
X + Y + Z + k02 = 0. (1.263)
X dx2 Y dy 2 Z dz 2
Now, the key result of the separation of variables strategy is that we may now argue that
each of the terms on the left-hand side of our equation must be equal to a constant. The
reason for this is that k02 is a constant and each of the remaining terms all vary independently
as functions along different coordinate axes. Hence, there is no way for them to “cancel each
other out” in order for the overall summation of terms to equal a constant at every single
point in space.
Considering this argument, we can set each of the terms on the left-hand side (with the
exception of k0 ) of our equation to be equal to different separation constants. For instance,
we have that
1 d2
2
X = −kx2 , (1.264)
X dx
1 d2
Y = −ky2 , (1.265)
Y dy 2
1 d2
Z = −kz2 . (1.266)
Z dz 2
Let’s look at the first equation for simplicity and then note that each of the other equations
will behave in the same manner due to the identical form. Now, we can rearrange our terms
to get

d2
X + kx2 X = 0. (1.267)
dx2
This is exactly the kind of equation that we encountered when we were considering our
“simple” plane waves! As a result, we see that a valid solution for X will be a plane wave
with “wavenumber” being equal to kx . More explicitly, we have that

X(x) = X0 e±jkx x , (1.268)

where X0 is some constant-valued amplitude factor. Since all of the separated equations will
behave in the same way, we get as an overall solution that

Ψ(x, y, z) = E0 e±jkx x e±jky y e±jkz z , (1.269)

where E0 is again some constant-valued amplitude factor that has absorbed all the constants
from the individual solutions. A common shorthand way to write this function out is as

Ψ(x, y, z) = E0 ejk·r , (1.270)

where k = x̂kx + ŷky + ẑkz and r = x̂x + ŷy + ẑz. We often refer to k as the wave vector. It
points in the direction of propagation.

79
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

To determine the magnitude of the wave vector, it is important to note that kx , ky , and
kz are not independent of each other. In particular, to be a valid solution we will need to go
back to our original equation in (1.263) before we separated them all out. Plugging in our
separation constants, we see that

kx2 + ky2 + kz2 = k02 , (1.271)

where it is important to remember that k0 still has the same definition as we have used
previously, i.e., k0 = ω/c0 . From this, we can readily see that k0 will be the correct magnitude
for our wave vector. We often refer to this kind of equation as a dispersion relation. It tells
us what are valid combinations of separation constants to still solve Maxwell’s equations.
Depending on the situation, we can plot aspects of this kind of equation on a dispersion
diagram. Hence, the dispersion diagram is plotting out values that correspond to valid
solutions to Maxwell’s equations.

1.18 Dispersion Relations

We will see repeatedly throughout this course that determining dispersion relations is a very
common and fundamental part of solving Maxwell’s equations for a variety of situations. To
see how it is useful, we will first consider a simple form of Maxwell’s equations when dealing
with plane wave solutions in Section 1.18.1. Following this, we will see how this approach
provides us with a powerful set of tools to determine how waves will propagate in more
complicated types of material media in Section 1.18.2.

1.18.1 Basics
In Practice Problem 8 from Section 1.23, you will prove that if E and H are of the form

E(r) = E0 ejk·r , H(r) = H0 ejk·r , (1.272)

then Maxwell’s equations in a source-free region reduce to

k × E = ωB, (1.273)

k × H = −ωD, (1.274)

k · D = 0, (1.275)

k · B = 0. (1.276)

We can use this form of Maxwell’s equations to derive a “wave equation” for plane waves
that only involves vector algebra as opposed to vector calculus. As we will see, this will allow
us to quickly derive dispersion relations to determine our solutions to Maxwell’s equations.
The real value of this approach is that it will allow us to analyze fairly complicated situations,

80
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

e.g., those involving permittivity and permeability tensors characteristic of complex kinds of
media, with a relatively straightforward process.
Let’s go about doing this process now. Following a similar process to deriving an actual
wave equation, we will start by taking the cross product of “Faraday’s law” with the wave
vector. This gives us

k × k × E = ωµk × H. (1.277)

We can now use “Ampere’s law” to substitute for k × H on the right-hand side. This gives
us

k × k × E = −ω 2 µϵE = −k 2 E. (1.278)

We can move everything to one side to get a “wave equation”. This is

k × k × E + k 2 E = 0. (1.279)

We can simplify this by using the vector identity that

a × (b × c) = (a · c)b − (a · b)c. (1.280)

This allows us to write

k × k × E = (k · E)k − (k · k)E. (1.281)

From our simplified form of “Maxwell’s equations” used earlier we conclude that k · E = 0
from “Gauss’ law”. Hence, we can write our “wave equation” as

(k · k − k 2 )E = 0. (1.282)

This equation will only be satisfied if we have

k · k = k2, (1.283)

which matches the dispersion relation we had found earlier since

k · k = kx2 + ky2 + kz2 . (1.284)

1.18.2 Anisotropic Media

We will now use the techniques from the previous section to determine the dispersion relation
appropriate to a more complicated case. In particular, we will consider the types of plane
wave solutions that can exist in a uniaxial medium. This is a special kind of anisotropic
medium that has a different permittivity only along a single axis. Mathematically, we repre-
sent this type of medium as a dyad/tensor – which we can write down as a matrix of values
in this case. For a uniaxial dielectric medium, we have
 
ϵ1 0 0
ϵ =  0 ϵ1 0  , µ = µ0 . (1.285)
0 0 ϵz

81
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We can use the same steps that we followed in the homework problem to see that
Maxwell’s equations can be simplified to

k × E = ωµH, (1.286)

k × H = −ωϵ · E, (1.287)

k · D = k · [ϵ · E] = 0, (1.288)

k · B = µk · H = 0, (1.289)

for this particular situation. Following the process of the previous section, we want to now
derive a “wave equation” that we will be able to find a dispersion relation from. However,
we must be careful in the steps that we follow to not forget that ϵ is a matrix. As a result,
ϵ can’t be “factored” out of expressions in simple ways; we must respect the rules of matrix
algebra.
To begin, we will start by taking the cross product of “Faraday’s law” with the wave
vector. This gives us

k × k × E = ωµk × H. (1.290)

We may now use “Ampere’s law” to substitute for k × H. This gives us

k × k × E = −ω 2 µϵ · E. (1.291)

In the previous section, we used

k × k × E = (k · E)k − (k · k)E (1.292)

to simplify our “wave equation”. Can we do that this time? The answer is no, because our
“Gauss’ law” no longer tells us that k · E = 0, instead it says that k · [ϵ · E] = 0. Hence, this
vector identity won’t be able to help us at this point. Instead, we are going to need to just
evaluate the vectors in (1.291) and use some concepts from linear algebra to help us find our
dispersion relation.
With this in mind, let’s look at a special case to simplify our algebra a bit. In particular,
we will assume that wave propagates in the x-direction, which correspondingly means that
E is confined to the (y, z)-plane. In this case, we have that

k × k × E = −ŷkx2 Ey − ẑkx2 Ez . (1.293)

We can then write out (1.291) in a full matrix notation to get some ideas for how to find the
dispersion relation. This makes (1.291) become
     
0 0 0 0 ϵ1 0 0 0
0 −kx2 2 
0  Ey = −ω µ 0 ϵ1 0
  Ey  , (1.294)
2
0 0 −kx Ez 0 0 ϵz Ez

82
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

where we have explicitly written that we are considering the case where Ex = 0. Now, we
have the same vector on each side of our equation, so we can simplify things by combining
everything into a single matrix. This gives
 2    
ω µϵ1 0 0 0 0
2 2
 0 ω µϵ1 − kx 0  Ey = 0 .
  (1.295)
2 2
0 0 ω µϵz − kx Ez 0

Now, if you recall from linear algebra, the only way for this matrix equation to have a non-
trivial solution is if the determinant of the matrix is equal to 0. Since we also have that
Ex = 0 due to the assumed propagation direction, we are really interested in making sure
that the lower 2 × 2 block matrix has a determinant of 0. This gives us

(ω 2 µϵ1 − kx2 )(ω 2 µϵz − kx2 ) = 0, (1.296)

which can of course be satisfied if one of the expressions in the parentheses are equal to 0.
This means that we have two solutions, or in this case, two dispersion relations. These can
be summarized as

1. The ordinary wave, with dispersion relation

ko2 = ω 2 µϵ1 (1.297)

and Ey ̸= 0, Ez = 0. (Note: the subscript here is an “o” for ordinary, not a “0”.)

2. The extraordinary wave, with dispersion relation

ke2 = ω 2 µϵz (1.298)

and Ez ̸= 0, Ey = 0.

We see that the propagation characteristics (phase velocity, attenuation constant, etc.) de-
pend on the polarization of the plane wave. This polarization-dependent propagation effect
can be utilized to make a number of useful devices.
For example, if we have a material for which ϵz is significantly lossy and ϵ1 is approx-
imately lossless, we can preferentially attenuate one polarization. This effect is used in
making polarized sunglasses to reduce the glare produced by light reflecting off of surfaces.
Later, we will learn more about how the reflection coefficient changes for an electromagnetic
wave depending on its polarization with respect to the orientation of the reflecting surface.
For now, we will just note that at large incidence angles the horizontally polarized light
reflects strongly while the vertically polarized light reflects to a much weaker degree. As a
result, polarized sunglasses are designed to block the horizontally polarized light that can
be particularly “glaring/bright”.
Another type of device that can be made with a uniaxial medium is a quarter-wave
plate. These are typically made with materials that have both a relatively lossless ϵ1 and ϵz .
However, the two permittivities are different enough that the waves associated with the two
polarizations will accumulate a different phase shift relative to each other. This effect can

83
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.47: Illustration of a sequence of uniaxial materials being used to generate circularly
polarized light. First, unpolarized light is passed through a linear polarizer so that the angle
of the polarization is known. By keeping this angle at 45◦ with respect to the axis of the
quarter-wave plate circular polarization is generated at the output of the quarter-wave plate
(image from Wikipedia [11]).

be used to take a linearly polarized wave and convert it into a circularly polarized wave (see
Fig. 1.47).
One final effect we will discuss is Faraday rotation. This can occur in gyrotropic media,
such as magnetized plasmas. Depending on the frequency of a wave, the ionosphere can act
as a magnetized plasma due to the influence of the Earth’s (approximately) static magnetic
field. If we were to account for the presence of the Earth’s magnetic field in our analysis of
a plasma medium previously (i.e., taking into account the v × B term in the Lorentz force)
we would end up with a permittivity tensor that looks like
 
ϵx jg 0
ϵ = −jg ϵy 0  . (1.299)
0 0 ϵz
Fully analyzing the propagation of waves for this kind of permittivity is beyond the scope
of this course, however, the main effect of interest is that the polarization of the wave
“rotates” as it propagates through the medium (see Fig. 1.48). For instance, if a linearly
polarized wave enters the gyrotropic medium with an electric field polarized along the x-axis
it will begin to rotate so that further into the medium the polarization will now be at some
angle with respect to the x-axis. This can be an important effect for certain components of
satellite communication systems, and is one reason why these systems tend to use at least
one circularly polarized antenna in the overall communication link (since the unpredictable
rotation of the waves could make the communication systems cross-polarized and no longer
able to communicate with one another). More broadly, it is common to include at least one
circularly polarized antenna in a communication link regardless of if it will propagate through
the ionosphere to ensure that the signal reception is always possible without inconvenient
“drop outs” due to the two antennas being oriented incorrectly with respect to each other.

84
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.48: Illustration of Faraday rotation. The polarization angle of the electric field
progressively changes as the wave propagates through the gyrotropic medium (image from
Wikipedia [12]).

The effect of Faraday rotation can also be used in designing various non-reciprocal mi-
crowave components. For instance, one common way to design circulators is by using ferrite
materials to cause Faraday rotation on the waves propagating through the device. Waves
that propagate through the device in different directions (e.g., clockwise and counterclock-
wise) can be carefully controlled so that they destructively interfere at one port of the device
and constructively add at the desired output port. Another common non-reciprocal com-
ponent built using ferrite materials that leverage the Faraday rotation effect are isolators.
These allow wave propagation through the device in one direction only, and are useful to
protect sensitive components from any reflections that may be produced at other locations
in a system. This can be important for protecting sensitive nonlinear components, such as
oscillators or amplifiers, and to help prevent positive feedback loops from happening unin-
tentionally in a system.

1.19 Plane Wave Scattering from a PEC Sheet (Oblique

Incidence)
In a previous lecture, we considered how a plane wave scattered from a planar PEC region
when the plane wave was normally incident on the planar boundary. We now turn our
attention to how to characterize the scattering that occurs when the plane wave has an
oblique incidence angle. This means that the unit normal vector of the planar boundary is
no longer perfectly aligned with the direction of propagation of the incident plane wave.

85
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.49: Illustration of the perpendicular (left) and parallel (right) polarizations. Other
commonly used names are also written next to the illustrations.

1.19.1 Geometry Basics

For this oblique incidence case, we will have two general kinds of situations depending on the
orientation of the plane wave’s polarization with respect to the planar boundary. There are
a number of different terminologies that are commonly used to differentiate between these
two cases. We will refer to these as perpendicular and parallel polarizations in this class, but
we will also mention the other names as we discuss the two cases in more detail.
To see why these are the relevant names, we must first understand the concept of the
plane of incidence for a planar scattering problem. This is defined as the plane that contains
the planar surface’s unit vector and the unit vector describing the direction of propagation
of the plane wave. The naming convention for our two polarizations mentioned above are
in reference to the orientation of the electric field and the plane of incidence. Hence, our
perpendicular polarization case occurs when the direction of the electric field is perpendicular
to the plane of incidence. Similarly, our parallel polarization case occurs when the direction
of the electric field is in the plane of incidence. Incident wave definitions for these two
cases are illustrated in Fig. 1.49. We will often denote a perpendicular polarization with the
subscript ⊥ and a parallel polarization with the subscript ∥. As we will see later, the different
polarizations will result in very different scattering behavior for the two polarizations.
Let’s now turn our attention to writing out the general expressions for the electric and
magnetic fields of perpendicular and parallel polarized plane waves. Having a quick reference
of these will be quite useful when we go about enforcing boundary conditions to solve for
the reflected and transmitted fields for specific scattering problems.

86
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.50: Illustration of the perpendicular polarization that we will derive expressions of
E and H from.

Perpendicular Polarization
We will start with the perpendicular polarization case. This is also referred to as the trans-
verse electric (TE) wave or E-wave case as well. The name transverse electric originates
from the fact that the electric field in this case is transverse to the plane of incidence (i.e.,
it is not contained in it). Let’s assume that we have the fields given in Fig. 1.50 to make
things concrete and determine the electric and magnetic fields.
For the perpendicular polarization case, we see that we can give a very simple definition
for the electric field. In particular, we have that
E⊥ = ŷE⊥ e−jk·r . (1.300)
We will now go about writing the magnetic field in terms of the electric field using what we
know about plane waves. First, we know that the magnetic field must be orthogonal to the
electric field and the direction of propagation. Further, we know that the cross product of
the electric and magnetic fields (in that order) must give us the direction of propagation.
Hence, we can draw out H⊥ as shown in Fig. 1.50. We also know that the the magnitude
of H⊥ will need to be E⊥ /η for a plane wave. We now want to write out the mathematical
expression for H⊥ . With a little geometry (illustrated in Fig. 1.50) we get that
 
E⊥ E⊥
H⊥ = − x̂ cos(θ) + ẑ sin(θ) e−jk·r . (1.301)
η η
It can be interesting to compare the wave impedances for the different polarizations,
which can be roughly thought of as the effective impedance for the plane wave components
propagating along the different coordinate axes. Since we have two components of the
magnetic field, we can define a wave impedance for each component. In particular, we define
Ey η
Zz = = , (1.302)
−Hx cos(θ)

87
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.51: Illustration of the parallel polarization that we will derive expressions of E and
H from.

Ey η
Zx = = , (1.303)
Hz sin(θ)
where the sign conventions have been selected to give us positive impedances. The signs can
be determined by how the signs on the components of E and H need to be to lead to a plane
wave propagating in the correct direction – i.e., positive z or positive x for this example.

Parallel Polarization
Let’s now take a look at the parallel polarization case. This is also referred to as the
transverse magnetic (TM) wave or H-wave case as well. For this situation we will assume
that we have the fields given in Fig. 1.51 to make things concrete and determine the electric
and magnetic fields.
For the parallel polarization case, we see that we can give a very simple definition for the
magnetic field. In particular, we have that
H∥ = ŷH∥ e−jk·r . (1.304)
We will now go about writing the electric field in terms of the magnetic field using what we
know about plane waves. First, we know that the electric field must be orthogonal to the
magnetic field and the direction of propagation. Further, we know that the cross product
of the electric and magnetic fields (in that order) must give us the direction of propagation.
Hence, we can draw out E∥ as shown in Fig. 1.51. We also know that the the magnitude
of E∥ will need to be ηH∥ for a plane wave. We now want to write out the mathematical
expression for E∥ . With a little geometry (illustrated in Fig. 1.51) we get that
 
E∥ = x̂ηH∥ cos(θ) − ẑηH∥ sin(θ) e−jk·r . (1.305)

The wave impedances for this polarization will be

Ex
Zz = = η cos(θ), (1.306)
Hy

88
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.52: Problem setup for the perpendicular polarization oblique incidence scattering
from a PEC sheet.

−Ez
Zx = = η sin(θ), (1.307)
Hy
where the sign conventions have been selected to give us positive impedances. Comparing
these to the wave impedances given in (1.302) and (1.303), we see that the dependence on
the angle has changed. As we will see when analyzing the scattering for these two kinds of
plane waves, this difference leads to fundamentally different scattering characteristics for the
two polarizations.
With the necessary geometric basics determined, we can now turn our attention to solving
for the reflected fields produced when a plane wave is obliquely incident on a planar PEC
sheet. We will start out by analyzing this for the perpendicular polarization case before then
considering the parallel polarization.

1.19.2 Perpendicular Polarization Case

The problem setup for the perpendicular polarization case is shown in Fig. 1.52. We will
now go through our three step process for solving these kinds of problems to determine the
reflected fields.
In Step 1, we identify the different regions of the problem, the boundaries between them,
and any “excitations” for the problem (incident fields in this case). For the scenario shown
in Fig. 1.52, we have
ˆ Region 1: z < 0, homogeneous medium with incident field Einc
⊥ ,

ˆ Region 2: z > 0, PEC,

89
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

ˆ Boundary 1: PEC and the homogeneous medium interface at z = 0. The unit normal
for boundary conditions is −ẑ.

Before moving on to Step 2, we first give our explicit expressions for the incident electric
and magnetic fields. These are
inc −jkinc ·r
Einc
⊥ = ŷE⊥ e = ŷE⊥inc e−jk(x sin θinc +z cos θinc ) (1.308)

E⊥inc
Hinc − x̂ cos θinc + ẑ sin θinc e−jk(x sin θinc +z cos θinc )


⊥ = (1.309)
η
Now, for Step 2 we need to propose an appropriate general solution in the different regions
of the problem. From our experience solving for the normal incidence case of scattering from
a PEC sheet, we already know that the fields will need to be 0 in Region 2 due to the infinite
conductivity. In Region 1, we expect that we will have a reflected field that will propagate
away from the interface. We also have no reason to expect that the PEC sheet will produce
new components of polarization directions, so we will assume that it will have a similar
mathematical expression and see if that gets us a workable solution. Our particular trial
solution will be based on the illustration in Fig. 1.52. For this field, we have
inc −jkref ·r
Eref
⊥ = ŷΓ⊥ E⊥ e = ŷΓ⊥ E⊥inc e−jk(x sin θref −z cos θref ) , (1.310)

E⊥inc
Href x̂ cos θref + ẑ sin θref e−jk(x sin θref −z cos θref ) .


⊥ = Γ⊥ (1.311)
η
We now move on to Step 3, where we use the boundary conditions of our problem to
solve for the unknown constants in our trial solutions. For this case, we need to solve for Γ⊥
and θr . From our previous experience, we know that the boundary condition on the electric
field is probably a good one to use for this situation. This tells us that the total tangential
electric field at the PEC boundary (z = 0) needs to be 0. Mathematically, we express this
condition as

−ẑ × (Einc ref

⊥ (z = 0) + E⊥ (z = 0)) = 0. (1.312)

From this, we get

E⊥inc e−jkx sin θinc + Γ⊥ E⊥inc e−jkx sin θref = 0. (1.313)

For this equation to have a possible solution for all x, we see that we will need the remaining
parts of the exponential functions to be equal to one another. This is known as a phase
matching condition, which gives us the additional equation

sin θinc = sin θref . (1.314)

From this, we immediately see that θinc = θref , that is the angle of reflection matches the
angle of incidence. This is also often referred to as Snell’s law of reflection.

90
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Using Snell’s law, we can simplify (1.313) to now be

E⊥inc + Γ⊥ E⊥inc = 0. (1.315)

From this equation, we see that Γ⊥ = −1.

We can use this solution to now write out the total field in Region 1. This gives us
inc −jkz cos θinc
E1 = Einc ref
⊥ + E⊥ = ŷE⊥ (e − ejkz cos θinc )e−jkx sin θinc , (1.316)

which we can simplify to be

E1 = −ŷ2jE⊥inc sin(kz cos θinc )e−jkx sin θinc . (1.317)

We can dissect this solution to see that it looks like we have a standing wave that is formed
in the z-direction and a kind of plane wave propagating in the x-direction with a phase
velocity of
ω c
vp = = . (1.318)
k sin θinc sin θinc
However, this is no longer a uniform plane wave propagating in the x-direction. This is
because the amplitude of the overall wave changes as a function of z.

1.19.3 Parallel Polarization Case

Let’s now look at the parallel polarization case. The problem setup for this scenario is
shown in Fig. 1.53. We will now go through our three step process for solving these kinds
of problems to determine the reflected fields.
As usual, in Step 1, we identify the different regions of the problem, the boundaries
between them, and any “excitations” for the problem (incident fields in this case). For the
scenario shown in Fig. 1.53, we have

ˆ Region 1: z < 0, homogeneous medium with incident field Hinc

∥ ,

ˆ Region 2: z > 0, PEC,

ˆ Boundary 1: PEC and the homogeneous medium interface at z = 0. The unit normal
for boundary conditions is −ẑ.

Before moving on to Step 2, we first give our explicit expressions for the incident electric
and magnetic fields. These are
inc −jkinc ·r
Hinc
∥ = ŷH∥ e = ŷH∥inc e−jk(x sin θinc +z cos θinc ) (1.319)

Einc inc
x̂ cos θinc − ẑ sin θinc e−jk(x sin θinc +z cos θinc )


∥ = ηH∥ (1.320)

Now, for Step 2 we need to propose an appropriate general solution in the different regions
of the problem. From our experience solving for the normal incidence case of scattering from

91
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.53: Problem setup for the parallel polarization oblique incidence scattering from a
PEC sheet.

a PEC sheet, we already know that the fields will need to be 0 in Region 2 due to the infinite
conductivity. In Region 1, we expect that we will have a reflected field that will propagate
away from the interface. We also have no reason to expect that the PEC sheet will produce
new components of polarization directions, so we will assume that it will have a similar
mathematical expression and see if that gets us a workable solution. Our particular trial
solution will be based on the illustration in Fig. 1.53. For this field, we have
inc −jkref ·r
Href
∥ = −ŷΓ∥ H∥ e = −ŷΓ∥ H∥inc e−jk(x sin θref −z cos θref ) (1.321)

Eref inc
x̂ cos θref + ẑ sin θref e−jk(x sin θref −z cos θref )


∥ = ηΓ∥ H∥ (1.322)

We now move on to Step 3, where we use the boundary conditions of our problem to
solve for the unknown constants in our trial solutions. For this case, we need to solve for Γ∥
and θref . From our previous experience, we know that the boundary condition on the electric
field is probably a good one to use for this situation. This tells us that the total tangential
electric field at the PEC boundary (z = 0) needs to be 0. Mathematically, we express this
condition as

−ẑ × (Einc ref

∥ (z = 0) + E∥ (z = 0)) = 0. (1.323)

From this, we get

ηH∥inc cos θinc e−jkx sin θinc + ηΓ∥ H∥inc cos θref e−jkx sin θref = 0. (1.324)

92
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We see that this equation can only be solved if we have the same phase matching condition
as we did for the perpendicular polarization case. In particular, this gives us

sin θinc = sin θref , (1.325)

from which we see that θinc = θref . Hence, we still get the same Snell’s law of reflection for
the parallel and perpendicular cases.
Using Snell’s law, we can simplify (1.324) to now be

1 + Γ∥ = 0, (1.326)

so that we again have Γ∥ = −1.

We can use this solution to now write out the total field in Region 1. This is easiest to
interpret by looking at the magnetic field, which gives us
inc −jkz cos θinc
H1 = Hinc ref
∥ + H∥ = ŷH∥ (e + ejkz cos θinc )e−jkx sin θinc , (1.327)

which we can simplify to be

H1 = ŷ2H∥inc cos(kz cos θinc )e−jkx sin θinc . (1.328)

We can dissect this solution to see that it looks like we have a standing wave that is formed
in the z-direction and a kind of plane wave propagating in the x-direction with a phase
velocity of
ω c
vp = = . (1.329)
k sin θinc sin θinc
However, this is no longer a uniform plane wave propagating in the x-direction. This is
because the amplitude of the overall wave changes as a function of z.

1.20 Normal Incidence at a Homogeneous Half Space

We will now consider what happens when a plane wave is normally incident on a half space
made of a homogeneous material. What this means is that at some location, say z =
0, we abruptly transition from one material (e.g., free space) to a homogeneous medium
characterized by a constant ϵ and µ for all locations with z > 0. We are interested in
determining what the electric and magnetic fields will need to be at all locations in space
for Maxwell’s equations to be satisfied. To do this, let’s go ahead and apply our three step
process for solving these kinds of problems. We will consider the specific case shown in Fig.
1.54.
In Step 1, we identify the different regions of the problem, the boundaries between regions,
and the specified fields (if any). For this problem, we have the following setup.

ˆ Region 1: z < 0, free space (µ0 , ϵ0 ), Einc (z) = x̂E inc e−jk0 z

ˆ Region 2: z > 0, homogeneous medium with µ, ϵ

93
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.54: Illustration of the problem for scattering from a homogeneous half space. Inci-
dent field is part of the specification of the problem, the reflected and transmitted fields are
postulated solutions that we need to solve for to satisfy Maxwell’s equations.

ˆ Boundary 1: z = 0, between free space and the homogeneous medium, the unit normal
for boundary conditions is −ẑ

We now move on to Step 2 where we propose reasonable trial solutions for the different
regions of the problem. From our previous experience, we expect that we may need to have
a reflected field in Region 1. Since Region 2 is not some kind of perfect conductor, we have
a feeling that we will also now need to consider the possibility of some amount of the plane
wave being able to transmit into Region 2 and continue propagating. Considering this, we
will write out our reflected and transmitted field trial solutions as

Eref (z) = x̂E ref ejk0 z , (1.330)

Etr (z) = x̂E tr e−jkz . (1.331)

Let’s now move to Step 3 where we enforce our boundary conditions to determine the
different constants we need to finish solving our problem. To begin, we will use the boundary
condition on the electric field. This was

n̂ × (E1 − E2 ) = 0. (1.332)

From this equation, we are able to deduce that at z = 0 we will need

E inc + E ref = E tr . (1.333)

We only know E inc in this equation, so we are going to need to determine another equation
to solve this problem.

94
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Which equation should we use? The best approach is to use the boundary condition
on the tangential components of the magnetic field. For this situation, there is no surface
current density so we will have that

n̂ × (H1 − H2 ) = 0. (1.334)

Obviously, we will need to evaluate our magnetic fields for each of the different plane waves
we are working with in order to use this equation. We can quickly find that

E inc −jk0 z
Hinc (z) = ŷ e , (1.335)
η0

E ref jk0 z
Href (z) = −ŷ e , (1.336)
η0

E tr −jkz
Htr (z) = ŷ e . (1.337)
η
We can plug these into our boundary condition to get

E inc E ref E tr
− = . (1.338)
η0 η0 η

We can now go about solving (1.333) and (1.338) to find what E ref and E tr should be.
We end up finding that
η − η0 inc
E ref = E , (1.339)
η + η0

2η
E tr = E inc . (1.340)
η + η0
We can now define a reflection coefficient as

E ref η − η0
Γ= = (1.341)
E inc η + η0
and a transmission coefficient as
E tr 2η
T = inc
= (1.342)
E η + η0
(note that your textbook uses the symbol τ for this).
This reflection coefficient is the same as what we have encountered before. It is a dimen-
sionless quantity that ranges from −1 to 1, and in general, can be a complex-valued number.
It is defined as the ratio of the reflected and incident fields, and as a result is a function
of position. If someone doesn’t specify where the reflection coefficient is referenced to for a

95
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

problem like this they typically will have it referenced to the interface between the regions,
which is what we did here.
We also now have a new concept of a transmission coefficient. This plays a similar role to
the reflection coefficient, but helps us understand how much of the wave is transmitted into
the new medium. To understand what values the transmission coefficient can range over, we
can go back to (1.333) to see that at z = 0 we have

1 + Γ = T. (1.343)

Hence, we see that 0 < T < 2.

This raises an interesting question. If, for instance, we have a Γ of 0.9, we get a T of
1.9. How is this possible? How are we getting a “stronger” plane wave in medium 2 than
what we had in our incident field? The answer can be put somewhat simply as there is no
law of conservation of voltages. We only need to have power conserved, but the electric field
(which is like a voltage) is not a conserved quantity. To see what happens with the power
of the waves, let’s take a look at the time-average Poynting vectors at either side of z = 0.
In Region 1, we have that

E1 (z) = Einc (z) + Eref (z) = x̂E inc (e−jk0 z + Γejk0 z )

(1.344)
 
inc −jk0 z j2k0 z
= x̂E e 1 + Γe

E inc −jk0 z
H1 (z) = Hinc (z) + Href (z) = ŷ (e − Γejk0 z )
η0
(1.345)
E inc −jk0 z
 
j2k0 z
= ŷ e 1 − Γe .
η0

The time-average Poynting vector is then

1
S1,avg = Re{E1 × H∗1 }
2
|E inc |2
= ẑ Re{(1 + Γej2k0 z )(1 − Γ∗ e−j2k0 z )}
2η0
|E inc |2
= ẑ Re{(1 − |Γ|2 ) + |Γ|(ej(2k0 z+∠Γ) − e−j(2k0 z+∠Γ) )} (1.346)
2η0
|E inc |2
= ẑ Re{(1 − |Γ|2 ) + j|Γ|2 sin(2kz + ∠Γ)}
2η0
|E inc |2
= ẑ (1 − |Γ|2 ).
2η0

In Region 2, we have only one plane wave so the math is much simpler. This gives us that

|E inc |2 2
S2,avg = ẑ |T | . (1.347)
2η

96
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We can equate our time-average Poynting vectors at the boundary to see that power will
only be conserved if
η0
1 − |Γ|2 = |T |2 . (1.348)
η
We can plug in our expressions for Γ and T to see that this equation does hold with our
definitions as is. Hence, there is no issue with the fact that 1 + Γ = T . Power is still
conserved, it is simply a situation that voltages and electric fields are not!
We can also take a closer look at our total electric field in Region 1 to get a better
understanding of what is occurring. In particular, we can be clever and add and subtract
Γ exp(−jk0 z) to write it as
 
inc −jk0 z jk0 z −jk0 z


E1 (z) = x̂E (1 + Γ)e +Γ e −e
  (1.349)
inc −jk0 z
= x̂E Te + Γj2 sin(k0 z) .

From this, we see that we can interpret the overall field as having a component that is a
propagating plane wave and another part that is a standing wave. The overall field will then
show interference between these two different kinds of waves.

1.21 Normal Incidence at Multiple Planar Interfaces

We will now consider the case where we have a plane wave incident on a “slab” of finite
thickness. As we will see, this problem has many similarities with what we just considered.
However, because we now have multiple interfaces between regions of differing constitutive
parameters we will have a bit more algebra to do in order to solve the overall problem.
However, the overall solution process will not need to change much at all from the previous
scenario. The specific illustration of the problem we will consider is shown in Fig. 1.55.
Let’s now begin our three step process to solving this kind of problem. We begin with
Step 1, where we note the different regions of the problem, interfaces, and any excitations.
ˆ Region 1: z < 0, homogeneous medium with ϵ1 , µ1 , incident field of Einc
ˆ Region 2: 0 < z < d, homogeneous medium with ϵ2 , µ2
ˆ Region 3: z > d, homogeneous medium with ϵ3 , µ3
ˆ Boundary 1: z = 0, between Region 1 and Region 2, unit normal is −ẑ
ˆ Boundary 2: z = d, between Region 2 and Region 3, unit normal is −ẑ
We now move to Step 2 where we specify appropriate general solutions in each region of
the problem. As with the previous case, we expect we will need a reflected plane wave in
Region 1. In Region 2, we expect to have a transmitted field as with the previous problem.
However, because there is another interface at z = d there is also the possibility of a reflected
field in Region 2. Hence, we will need two possible solutions in this region, namely, forward-
and backward-propagating plane waves. Now, in Region 3 we should only have a transmitted
field that is propagating in the +z direction. Summarizing, we have

97
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.55: Illustration of the problem for scattering from multiple planar interfaces. The
incident field is part of the specification of the problem, the reflected and transmitted fields
are postulated solutions that we need to solve for to satisfy Maxwell’s equations.

ˆ Region 1

– Einc = x̂E inc e−jk1 z

E inc −jk1 z
– Hinc = ŷ e
η1
– Eref = x̂ΓE inc ejk1 z
E inc jk1 z
– Href = −ŷΓ e
η1
ˆ Region 2

– E+ + −jk2 z
2 = x̂E2 e
E2+ −jk2 z
– H+
2 = ŷ e
η2
– E− − jk2 z
2 = x̂E2 e
E2− jk2 z
– H−
2 = −ŷ e
η2
ˆ Region 3

– Etr = x̂T E inc e−jk3 (z−d)

T E inc −jk3 (z−d)
– Htr = ŷ e
η3
We can now move to Step 3 to solve for the unknown quantities by applying our boundary
conditions at the interfaces. We see that we have 4 quantities to solve for (Γ, T , E2+ , E2− ),

98
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

so we will need 4 equations. We can get these by ensuring the continuity of the tangential
electric and magnetic fields at the two boundaries of the problem. Let’s start at Boundary
2 and work our way back. At Boundary 2, we have that

E2+ e−jk2 d + E2− ejk2 d = T E inc , (1.350)

E2+ −jk2 d E2− jk2 d T E inc

e − e = . (1.351)
η2 η2 η3

Similarly, at Boundary 1 we have that

E inc (1 + Γ) = E2+ + E2− , (1.352)

E inc E2+ E2−

(1 − Γ) = − . (1.353)
η1 η2 η2

We can solve (1.350) to (1.353) for all of the unknown constants. However, the algebra be-
comes rather tedious if one attempts to do it directly. In some situations, certain strategies
can be used to quickly determine certain results without requiring solving the full prob-
lem. For instance, if one is only concerned with the overall reflection produced by the slab
geometry, one can use some tricks from analyzing transmission lines to determine that

Γ12 + Γ23 e−j2k2 d

Γ= , (1.354)
1 + Γ12 Γ23 e−j2k2 d

where

ηj − ηi
Γij = (1.355)
ηj + ηi

is an intrinsic reflection coefficient. This is the reflection coefficient that would exist between
these two media if they were both semi-infinite in extent. The expression in (1.354) appears
quite frequently when analyzing systems with multiple interfaces, and is a good one to
remember.

1.22 Oblique Incidence at a Homogeneous Half Space

We will now consider what happens when a plane wave approaches a homogeneous half space
with oblique incidence. For this situation, we will have different behavior depending on the
polarization of the incident field. Hence, we will need to consider the perpendicular and
parallel polarization cases separately.

99
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.56: Illustration of the problem for perpendicular polarization oblique incidence
scattering from a homogeneous half space. The incident field is part of the specification of
the problem, the reflected and transmitted fields are postulated solutions that we need to
solve for to satisfy Maxwell’s equations.

1.22.1 Perpendicular Polarization

We will start with the perpendicular polarization case, which is illustrated in Fig. 1.56. We
will now go through the three step process of solving this problem. In Step 1, we note the
different regions of the problem, interfaces, and excitations. For this problem, we have the
following setup.

ˆ Region 1: z < 0, homogeneous medium with ϵ1 , µ1 , and

Einc (r) = ŷEi e−jk1 (x sin θinc +z cos θinc ) (1.356)

Ei
Hinc (r) = (−x̂ cos θinc + ẑ sin θinc )e−jk1 (x sin θinc +z cos θinc ) (1.357)
η1

ˆ Region 2: z > 0, homogeneous medium with ϵ2 , µ2

ˆ Boundary 1: z = 0, between free space and the homogeneous medium, the unit normal
for boundary conditions is −ẑ

We now move on to Step 2 where we propose reasonable trial solutions for the different
regions of the problem. From our previous experience, we expect that we may need to
have a reflected field in Region 1. Similarly, we now know we need to also consider the
possibility of some amount of the plane wave being able to transmit into Region 2 and

100
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

continue propagating. Considering this, we will write out our reflected and transmitted field
trial solutions as

Eref (r) = ŷEr ejk1 (z cos θref −x sin θref ) , (1.358)

Er
Href (r) = (x̂ cos θref + ẑ sin θref )ejk1 (z cos θref −x sin θref ) , (1.359)
η1

Etr (r) = ŷEt e−jk2 (x sin θtr +z cos θtr ) , (1.360)

Et
Htr (r) = (−x̂ cos θtr + ẑ sin θtr )e−jk2 (x sin θtr +z cos θtr ) . (1.361)
η2

We can plug (1.356), (1.358), and (1.360) into our boundary condition for the tangential
components of the electric field at z = 0 to get

Ei e−jk1 x sin θinc + Er e−jk1 x sin θref = Et e−jk2 x sin θtr . (1.362)

The only way for this equation to have a possible solution is for the arguments of all of the
exponential functions to match, i.e., our phase matching condition. Hence, we have that

k1 sin θinc = k1 sin θref = k2 sin θtr . (1.363)

We have already seen the first equality before. This was Snell’s law of reflection. The next
equality is referred to as Snell’s law of refraction. It tells us that the angle of transmission
will depend on the angle of incidence and the ratio of material properties between the two
media. More explicitly, we see that
r 
µ1 ϵ1
θtr = arcsin sin θref . (1.364)
µ2 ϵ2

After applying our phase matching condition, we can simplify (1.362) to be

Ei + Er = Et . (1.365)

We need another equation to solve for the two unknowns. We can get this from the continuity
of the tangential magnetic field. This will give us
1 Et
(Ei − Er ) cos θinc = cos θtr . (1.366)
η1 η2

We can solve our two equations to determine reflection and transmission coefficients of

Er (η2 / cos θtr ) − (η1 / cos θinc )

Γ⊥ |z=0 = = , (1.367)
Ei (η2 / cos θtr ) + (η1 / cos θinc )

101
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.57: Illustration of the problem for parallel polarization oblique incidence scattering
from a homogeneous half space. The incident field is part of the specification of the problem,
the reflected and transmitted fields are postulated solutions that we need to solve for to
satisfy Maxwell’s equations.

Et 2(η2 / cos θtr )

T⊥ |z=0 = = . (1.368)
Ei (η2 / cos θtr ) + (η1 / cos θinc )
We also still have that
1 + Γ⊥ = T⊥ . (1.369)
Hence, we see that our expressions look very close to the normal incidence case. We just
need to scale the impedance of the different regions by 1/ cos θ where θ is the angle of the
wave’s propagation in that medium. You should recall that we saw this same behavior when
defining the wave impedance for perpendicular polarized plane waves in a previous lecture.
We will now turn our attention to solving the parallel polarization case. Afterward, we
will compare the two situations and also look at a couple special cases of particular interest.

1.22.2 Parallel Polarization

The parallel polarization case is illustrated in Fig. 1.57.
We will now go through the three step process of solving this problem. In Step 1, we
note the different regions of the problem, interfaces, and excitations. For this problem, we
have the following setup.
ˆ Region 1: z < 0, homogeneous medium with ϵ1 , µ1 , and
Einc (r) = Ei (x̂ cos θinc − ẑ sin θinc )e−jk1 (x sin θinc +z cos θinc ) (1.370)

Ei −jk1 (x sin θinc +z cos θinc )

Hinc (r) = ŷe (1.371)
η1

102
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

ˆ Region 2: z > 0, homogeneous medium with ϵ2 , µ2

ˆ Boundary 1: z = 0, between free space and the homogeneous medium, the unit normal
for boundary conditions is −ẑ
We now move on to Step 2 where we propose reasonable trial solutions for the different
regions of the problem. From our previous experience, we expect that we may need to
have a reflected field in Region 1. Similarly, we now know we need to also consider the
possibility of some amount of the plane wave being able to transmit into Region 2 and
continue propagating. Considering this, we will write out our reflected and transmitted field
trial solutions as

Eref (r) = Er (x̂ cos θref + ẑ sin θref )e−jk1 (x sin θref −z cos θref ) , (1.372)

Er −jk1 (x sin θref −z cos θref )

Href (r) = − ŷe , (1.373)
η1

Etr (r) = Et (x̂ cos θtr − ẑ sin θtr )e−jk2 (x sin θtr +z cos θtr ) , (1.374)

Et −jk2 (x sin θtr +z cos θtr )

Htr (r) = ŷe . (1.375)
η2
We can plug (1.370), (1.372), and (1.374) into our boundary condition for the tangential
component of the electric field at z = 0 to get

Ei cos θinc e−jk1 x sin θinc + Er cos θref e−jk1 x sin θref = Et cos θtr e−jk2 x sin θtr . (1.376)

The only way for this equation to have a possible solution is for the arguments of all of the
exponential functions to match, i.e., our phase matching condition. Hence, we have that

k1 sin θinc = k1 sin θref = k2 sin θtr , (1.377)

which is exactly what we saw for the perpendicular polarization case.

After applying our phase matching condition, we can simplify (1.376) to be

(Ei + Er ) cos θinc = Et cos θtr . (1.378)

We need another equation to solve for the two unknowns. We can get this from the continuity
of the tangential magnetic field. This will give us
1 Et
(Ei − Er ) = . (1.379)
η1 η2
We can solve our two equations to determine reflection and transmission coefficients of
Er η2 cos θtr − η1 cos θinc
Γ∥ |z=0 = = , (1.380)
Ei η2 cos θtr + η1 cos θinc

103
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Et 2η2 cos θinc

T∥ |z=0 = = . (1.381)
Ei η2 cos θtr + η1 cos θinc

We also have that

1 + Γ∥ = T∥ (cos θtr / cos θinc ). (1.382)

We see that our expressions are still somewhat similar to the normal incidence case, but
to a lesser degree than the perpendicular polarization expressions were. As a result, it is
vital that we remember that for oblique incidence cases on a homogeneous half space that
our parallel and perpendicular polarizations exhibit different behavior, and so, have different
formulas that must be used in evaluating certain properties.

1.22.3 Comparison of Parallel and Perpendicular Polarizations

We have just seen that the reflection and transmission coefficients for the perpendicular and
parallel polarization cases are different from each other. It is natural to then ask, are there
any practical consequences of this?
There are of course many practical consequences of this, but one that you have probably
experienced is in the use of polarized sunglasses. These are specifically designed to filter
out “horizontally” polarized light. That is, light which is in parallel to the surface of the
earth. The reason for this is that much of the sunlight that is getting reflected off of surfaces
outside will be from “horizontal” surfaces (like the ground). If we assume that the sunlight
itself isn’t preferentially polarized (which turns out to be the case), we can assume that we
will have approximately an equal amount of light approaching the ground in parallel and
perpendicular polarizations. Thinking about the geometry of the problem, we see that the
“horizontal” polarization will match the perpendicular polarization case, while a “vertical”
polarization would correspond to the parallel polarization case.
We can write down the two reflection coefficients again to compare them. That is,

(η2 / cos θtr ) − (η1 / cos θinc )

Γ⊥ = , (1.383)
(η2 / cos θtr ) + (η1 / cos θinc )

η2 cos θtr − η1 cos θinc

Γ∥ = . (1.384)
η2 cos θtr + η1 cos θinc

With the exception of θinc = 0, we can quickly see that as θinc gets larger Γ⊥ will tend to get
larger while Γ∥ will get smaller. To see this more clearly, we have √ plotted the magnitude of
the reflection coefficients for the case where η1 = η0 and η2 = η0 / 10 in Fig. 1.58. From this,
we see that at relatively large incidence angles the perpendicular reflection coefficient is much
larger than the parallel reflection coefficient. Hence, by blocking the horizontal polarization
(that corresponds to the perpendicular polarization) a pair of polarized sunglasses can block
much of the glare that would come about due to reflected sunlight.

104
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.58: Comparison of parallel and perpendicular polarization reflection coefficients.

1.22.4 Special Angles

Moving on, there are two kinds of angles that are typically considered to be “special”. These
are the critical angle and the Brewster angle. The critical angle is the angle where the
reflection coefficient is equal to 1 (i.e., perfect reflection) and the Brewster angle is the exact
opposite – i.e., the angle where no reflection occurs. Let’s take a look at what situations can
allow for this to happen.

Critical Angle
We will begin with the critical angle. This typically occurs when we have a wave in a
“denser” medium attempting to propagate into a “less dense” medium (density is a term
sometimes used to refer to the material properties, a denser medium has a higher constitutive
parameter). For many practical cases of interest, we won’t have magnetic materials present
so that we can assume that ϵ1 > ϵ2 . The quickest way to find the critical angle is from Snell’s
law of refraction. We will have total reflection when θtr = π/2, since in this case no wave
actually propagates into the less dense medium. Hence, we want to find

k1 sin θc = k2 sin(π/2) = k2 . (1.385)

We can write this as

r 
ϵ2
θc = arcsin . (1.386)
ϵ1

We will continue to have a complete reflection for any situation where the incident angle is
larger than the critical angle. Although there is complete reflection, we do have some fields
“leak” into medium 2. These are in the form of evanescent waves that do not propagate,
but instead exponentially attenuate. It should be noted that because there are still fields in

105
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Figure 1.59: Illustration of total internal reflection within an optical fiber (illustration shows
a cross section of a cylindrical fiber).

this “evanescent region” they can interact with objects near to the boundary between the
propagating and evanescent regions. This is a general kind of wave phenomena, and is at
the core of why tunneling of waves is possible. For instance, a number of semiconductor
electronics are able to operate due to electrons tunneling through potential barriers. In the
region of the tunneling barrier the electron wave is evanescent, but as long as the potential
barrier is sufficiently thin it is possible for the electron to exit out the other side and begin
propagating again.
A practical electromagnetic device that uses the total internal reflection effect provided
by the critical angle is an optical fiber. The core of the fiber is made of a denser material
than the cladding that surrounds the core. Light can then be input into the fiber. Any light
that is propagating with an angle larger than the critical angle of the overall fiber geometry
will be able to propagate down the fiber. You can picture a ray bouncing back and forth
from the different walls of the fiber core, as shown in Fig. 1.59.

Brewster Angle
Next, we will discuss the Brewster angle for perpendicular and parallel polarizations. This is
the angle at which no reflection occurs. (Note: it is important to recognize that just because
there is no reflection, it does not necessarily mean that the transmission coefficient is equal
to 1, e.g., see (1.382)!) Looking at the numerators of both of the reflection coefficients,
we see that this seems like it should potentially be possible if we can find an appropriate
combination of η1 , η2 , and θinc for the two terms to cancel each other. Denoting the Brewster
angle for perpendicular polarization as θB⊥ , we want to find when
η2 η1
= . (1.387)
cos θtr cos θB⊥
We can use Snell’s law of refraction to write cos θtr as
r
p
2 µ1 ϵ1
cos θtr = 1 − sin θtr = 1 − sin2 θinc . (1.388)
µ2 ϵ2

106
 CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

We can plug this into (1.387) to eventually find that

1 − µ1 ϵ2 /(µ2 ϵ1 )
sin2 θB⊥ = . (1.389)
1 − (µ1 /µ2 )2
The first thing we can notice is that if µ1 = µ2 then the denominator will become infinite and
we won’t have a Brewster angle. Many practical materials are dielectrics, whose magnetic
properties match that of free space (i.e., µr = 1). Hence, for any interface between two
dielectric regions there will be no Brewster angle for the perpendicular polarization. There
are certain combinations of material properties where the Brewster angle can exist for a
perpendicular polarization, however, these rarely are found in natural materials. As a result,
it is extremely rare to encounter a Brewster angle for a perpendicular polarization in practical
situations.
Although this is the case for perpendicular polarization, the story is rather different for
a parallel polarization. The same set of steps can be used to find that the Brewster angle
for a parallel polarization is
1 − µ2 ϵ1 /(µ1 ϵ2 )
sin2 θB∥ = . (1.390)
1 − (ϵ1 /ϵ2 )2
It is extremely common to be able to have an interface between two regions with different
permittivities, so the denominator is rarely going to have a situation where it would be-
come infinite. As a result, it is generally possible to find a Brewster angle for the parallel
polarization between two dielectric materials. For instance, this occurred in Fig. 1.58 at
72.5◦ .

1.23 Practice Problems

1. (a) Show that Gauss’ laws of electricity and magnetism can be derived from the other
equations in Maxwell’s equations.
Hint: You will need to use some of the vector calculus identities discussed in the
lecture notes to complete this question.
(b) Explain why your derivation only works for time-varying systems.
2. (a) Prove that


∇ · f A = f ∇ · A + A · ∇f (1.391)
for differentiable scalar f and vector A. You may use Cartesian coordinates for
simplicity.
(b) Provide an intuitive explanation for what each term on the right-hand side of this
identity means.
3. (a) Prove that



∇ × f A = f ∇ × A + ∇f × A (1.392)
for differentiable scalar f and vector A. You may use Cartesian coordinates for
simplicity.

107
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

(b) Provide an intuitive explanation for what each term on the right-hand side of this
identity means.

4. Verify in Cartesian coordinates that

∇ × ∇f = 0 (1.393)

and


∇ · ∇ × A = 0, (1.394)

5. A medium where the permittivity and permeability may vary as a function of position
is called an inhomogeneous medium.

(a) Derive a time domain wave equation for the electric field in an inhomogeneous
medium where both ϵ and µ are functions of position.
(b) Derive a time domain wave equation for the magnetic field in an inhomogeneous
medium where both ϵ and µ are functions of position.

Hint: start from Maxwell’s equation written in terms of E, H, D, and B and be careful
to not factor ϵ and µ out of any spatial derivatives. Instead, think about if there are
simple ways you can rearrange the equations prior to taking the spatial derivatives to
simplify the rest of the derivation. Alternatively, follow the procedure given in class
and use appropriate vector calculus identities (you can find the appropriate one in
HW1) to find a form where Faraday’s and Ampere’s laws can be substituted into one
another. These two approaches will yield different looking wave equations, but both
are equivalent and valid.

6. The electric field of a uniform plane wave in free space is given by

E(x, t) = ẑ 34 cos(4.8π × 109 t − kx) [mV/m]. (1.395)

(a) What direction is this plane wave traveling in?

(b) Find the frequency f , the wavenumber k, and the wavelength λ of this wave.
Note: do not forget to label the units of your answer. These reminders will not
be made again.
(c) Sketch Ez (x, t) as a function of t at x = 0 and x = λ/4.
(d) Sketch Ez (x, t) as a function of x at t = 0 and t = π/ω.

7. The magnetic field of a Ku-band satellite broadcast signal propagating in free space is
given by

H(z, t) = x̂ H0 cos(ωt − 261.8z) [mA/m]. (1.396)

(a) Find f and λ for this wave.

(b) Find E(z, t).

108
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

Point of Interest: Ku-band is the name of a portion of the electromagnetic spectrum

that contains the frequency of this broadcast signal. The exact range of frequencies
contained in Ku-band varies depending on whose definition you use (e.g., commercial,
government, military, NATO, etc.), but the rough size of the band corresponds to
the frequencies that can be correctly transmitted down a standard-sized rectangular
waveguide (in particular, a WR-62 waveguide). We will learn more about why a
rectangular waveguide can only correctly transmit a portion of the electromagnetic
spectrum in the waveguides section of the course.

8. For a time harmonic plane wave propagating in a homogeneous medium, both E and
H vary with the exponential “phase factor” exp[−jk · R], where k = x̂kx + ŷky + ẑkz
and R = x̂x + ŷy + ẑz. Assuming that E = Ê0 exp[−jk · R] and H = Ĥ0 exp[−jk · R],
show by direct substitution that Maxwell’s equations in a source-free region reduce to

k × E = ωµH, (1.397)

k × H = −ωϵE, (1.398)

k · E = 0, (1.399)

k · H = 0. (1.400)

9. Although there is no physical evidence that magnetic current densities exist, the effects
of certain real source distributions can be conveniently modeled by treating them as
“effective” magnetic current densities. This is the case, for example, in the analysis of
aperture antennas or various kinds of coupled waveguides. If we denote the effective
magnetic current density as M, Maxwell’s equations become

∇ × H = ∂t D + J, (1.401)

∇ × E = −∂t B − M, (1.402)

∇ · D = ρ, (1.403)

∇ · B = 0. (1.404)

Note: do not confuse M with the magnetization intensity of a material that we dis-
cussed previously; instead, think of it like a counterpart to J.

(a) Derive (1.404) from (1.402). This will require you to specify a particular condition
on a derivative of M as a “postulate” of our modified theory. What does this
postulate say about the spatial variation of M (recall our discussions on the
spatial variation of B)?

109
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

(b) Derive the integral forms of the modified Maxwell’s equations given in (1.401) to
(1.404).
(c) Using your answer from 1(b), derive the boundary condition that is applicable to
(1.402) assuming that there is a surface magnetic current density present at the
boundary between the two regions.

10. A 5 GHz y-polarized electric field is propagating in the +z-direction through a low-loss
dielectric medium which has ϵ′r = 3.5. Answer the following questions given that at
the same instant in time at z = 0 [m] the phasor of the electric field is 10∠40◦ [V/m]
and that at z = 0.1 [m] the phasor of the electric field is 8.22∠−3.34◦ [V/m].
(a) What is the attenuation constant in dB/m?
(b) What is the phase constant in rad/m? Compare this to the wavenumber that
would occur in a lossless dielectric medium with ϵr = 3.5.
Hint: If your two answers are significantly different you have done something
wrong.
(c) Write down an expression for the instantaneous magnetic field that is valid for all
z ≥ 0.
11. A 5 GHz z-polarized plane wave is propagating in the −y-direction in a lossy dielectric
medium with ϵr = 3 − j2.5. If the electric field phasor at y = 0 is 4∠30◦ V/m, write
down the instantaneous expression for the magnetic field.
12. A 10 GHz electric field for a plane wave in a lossy dielectric medium is given by
E = x̂Ex + j ẑEz eαy+jβy .

If the relative permittivity of the medium is 10 − j6, find an expression for the instan-
taneous magnetic field. Make sure to compute explicit numeric values for all quantities
except the arbitrary complex-valued amplitudes Ex and Ez .
13. In this problem, we will again consider a form of Maxwell’s equations that includes an
effective magnetic current density, M. Considering this, Maxwell’s equations are given
by (1.401) to (1.404). Note: you do not need to consider conduction currents for this
problem.
(a) Derive Poynting’s theorem in the time domain for the modified Maxwell’s equa-
tions given in (1.401) to (1.404).
(b) Repeat this derivation in the frequency domain.
14. Assume that we have a surface magnetic current density in free space located at x = 0
that covers the entire (y, z)-plane. Our expression for this current sheet is
(
ẑM0 cos(ωt − φ0 ), x = 0,
Ms (r) = (1.405)
0, elsewhere.
Using the boundary conditions appropriate to the modified Maxwell’s equations given
in (1.401) to (1.404), determine what E and H are equal to in all regions of space.

110
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

15. Previously, we discussed the concept of a perfect electric conductor (PEC). We saw that
the boundary condition between free space and a PEC region was that the tangential
component of the electric field was equal to 0. In similarity to this, we can also define
the concept of a perfect magnetic conductor (PMC). The boundary condition that is
appropriate between free space and a PMC region is that the tangential component of
the magnetic field must be equal to 0. Further, the perfect conductivity means that
all fields will equal 0 inside the PMC region.

(a) Now, assume that we have a PMC filling the entire region of space for y ≥ 0. If
we have a plane wave incident on this PMC region described by
Hinc (y) = ẑH + e−jk0 y , (1.406)
determine the expressions for the reflected E and H that must be produced to
ensure that Maxwell’s equations are satisfied in all regions of space.
(b) Determine the reflection coefficient for this scenario at y = 0.
Note: the reflection coefficient by convention is always specified in terms of the
ratio of the electric field.
(c) Compare this reflection coefficient to the one we determined previously for the
reflected fields produced at the interface with a PEC region.

Point of Interest: Neither PECs or PMCs exist in the real world. However, a PEC is a
reasonable approximation for many good conductors at microwave frequencies, and so
it is used frequently in practical engineering design. In contrast to this, there are no
“good magnetic conductors” that exist in the real world, so PMCs have traditionally
just been useful as a mathematical concept for simplifying the analysis of certain
advanced problems. Recently, clever antenna engineers have been able to design a type
of metasurface known as an artificial magnetic conductor (AMC). These are surfaces
engineered to approximately replicate some of the properties of a PMC over a particular
band of frequencies. These have found use in designing low-profile antennas and in
improving the bandwidth and radiation patterns of certain types of antennas.
16. Determine the polarization for each of the following plane waves. At a minimum,
please show your work by specifying the logic behind your answer (i.e., don’t simply
write linear, right-hand circular, etc.; provide some kind of justification for the answer).
Don’t forget to specify the sense of the polarization if it is circular or elliptical, and
pay attention to the propagation direction of the wave.
(a) E(z) = j x̂|E1 | + ŷ|E2 | e−jkz , |E1 | = |E2 |




(b) E(z) = x̂|E1 | − j ŷ|E2 | ejkz , |E1 | =
̸ |E2 |

jkz
(c) E(z) = x̂|E1 | − ŷ|E2 | e , |E1 | = ̸ |E2 |
jπ/3 −jkz


(d) E(z) = x̂|E1 | + ŷ|E2 |e e , |E1 | = |E2 |
17. Prove that an arbitrarily-polarized plane wave can be decomposed into a superposition
of right-hand and left-hand circularly polarized waves. Stated another way, show that
right-hand and left-hand circularly polarized waves form a basis for plane waves.

111
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

18. Prove that the instantaneous Poynting vector of a circularly polarized plane wave in a
lossless homogeneous medium is constant as a function of time and space.

19. A plane wave in medium 1 (ϵ1 , µ1 ) is incident normally onto a lossless dielectric slab
(ϵ2 , µ2 ) that is backed by a PEC plane. If the dielectric slab is of thickness d, determine
the reflected electric field in medium 1, the total electric field in the dielectric slab,
and the time-average Poynting vector everywhere.
Hint: If you denote Ei0 as the amplitude of the incident field and Er as the complex-
valued amplitude of the reflected field you should find that
Er jη2 tan(β2 d) − η1
= .
Ei0 jη2 tan(β2 d) + η1
To make it easiest to arrive at this expression, the phase factors in the exponentials
of your proposed solutions inside medium 2 should be ±jβ2 (z − d); i.e., their phase
reference is set to the surface of the PEC located at z = d. Along these lines, if E2+
is the complex-valued amplitude of the field propagating in the +z-direction inside
medium 2, you should find that

E2+ η2
= .
Ei0 η1 cos(β2 d) + jη2 sin(β2 d)
Also note that you may not use these relationships as given in your solution, you must
derive them yourself.

20. A 10 GHz parallel polarization plane wave with a peak amplitude of 10 mV/m is
propagating in free space and is incident on a lossless dielectric half space at an angle
of 40◦ . If the transmitted wave propagates in the dielectric medium at an angle of 20◦
with respect to the z-axis, determine the following.

(a) The relative permittivity of the dielectric medium.

(b) Expressions for the incident and reflected electric fields in free space.
(c) An expression for the total electric field in the dielectric medium.

Note: make sure to fully evaluate the expressions in (b) and (c); i.e., don’t leave
symbols like Γ∥ in your expressions if they can be numerically evaluated.

21. A 10 GHz circularly polarized plane wave given by


Einc (y, z, t) = |E0 | x̂ cos(2π × 1010 t + 66.28y − 198.83z)

10
− (ŷ 0.9487 + ẑ 0.3162) sin(2π × 10 t + 66.28y − 198.83z)

is propagating in free space and is incident at the Brewster angle on a lossless dielectric
half space located at y = 0. If the relative permittivity of the dielectric half space is
ϵr = 9, answer the following questions.

112
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

(a) What is the sense of rotation for the incident plane wave?
(b) What is the polarization of the reflected wave?
(c) Find an expression for the transmitted wave.
(d) What is the polarization of the transmitted wave?

22. A 14 GHz plane wave given by

Einc (x, y, t) = −E0 (x̂ 0.7660 + ŷ 0.6428)ej(342.62x−408.31y)

is propagating in a lossless dielectric and is incident on another lossless dielectric half

space that begins at x = 0. If the relative permittivity of the dielectric half space the
wave will be transmitted into is ϵr2 = 16.5, answer the following questions.

(a) What is the angle of incidence for this wave?

(b) What is the relative permittivity of the dielectric medium the incident wave is
propagating in?
(c) What is the transmission angle for this problem?
(d) Find an expression for the reflected magnetic field.
(e) Find an instantaneous expression for the transmitted electric field.

23. A 5 GHz plane wave propagating in free space is given by

Einc = ŷ5e−j(85.840x−60.106z) [V/m].

This wave is incident on a dielectric half space that begins at z = 0. If the relative
permittivity of the dielectric half space is ϵr = 20, answer the following questions.

(a) What is the angle of incidence of this wave (measured relative to the z-axis)?
(b) What will the angle of transmission be?
(c) Find an expression for the incident magnetic field.
(d) Find an instantaneous expression for the reflected magnetic field.
(e) Find an expression for the transmitted electric field.

113
CHAPTER 1. FUNDAMENTALS OF ELECTROMAGNETICS AND PLANE WAVES

114
Chapter 2

Transmission Lines

2.1 Introduction to Transmission Lines

Although the most accurate solution to a problem will always come from performing a com-
plete field analysis of the system, this is often impractical at many stages of a design. In
certain cases, we can apply some significant approximations to simplify Maxwell’s equations
down to a “0-dimensional” kind of circuit theory. However, there are many practical applica-
tions where this kind of circuit theory is no longer sufficient. In particular, this occurs when
the sizes of the circuit begin to approach the size of the wavelength of the electromagnetic
fields existing in the system. Transmission line theory provides us with a set of tools to
bridge the gap between circuit theory and a full field analysis of systems. As a result, it is
significantly simpler than performing a complete field analysis, but still allows us to model
the wave behavior of voltages and currents that are propagating through our circuitry. This
forms a kind of one-dimensional wave theory that is extremely useful in analyzing and de-
signing a wide range of practical systems. This transmission line theory is an indispensable
tool in the fields of radio frequency (RF) and microwave engineering.
Before we begin our discussions on transmission line theory, it is useful to look at a few
examples of the different kinds of geometries used to guide waves between circuit components.
In Fig. 2.1, we show some common kinds of non-planar transmission lines. Similarly, Fig. 2.2
shows some of the most common kinds of planar transmission line structures. Transmission
line theory will provide us with a way to analyze this diverse range of geometries using a
unified theoretical framework.

2.2 Telegrapher’s Equations

We now turn our attention to seeing how wave behavior can begin to manifest itself in a
transmission line. This can typically be done in two ways. One approach is to begin with
Maxwell’s equations and simplify them for a specific geometry. Alternatively, one can start
with a simplified model and more quickly derive the wave behavior. We take the second
approach here, and will see aspects of the first approach later in the course.
To derive general equations to describe transmission line behavior, we must introduce the
concept of a distributed parameter or distributed element network. The idea is that because

115
CHAPTER 2. TRANSMISSION LINES

Figure 2.1: Examples of non-planar transmission lines.

Figure 2.2: Examples of planar transmission lines. Dark boxes are conductors and lighter
boxes are dielectric substrates.

the physical dimensions of the transmission line are of a relevant size to the wavelength
of the fields, the voltages and currents that flow along the transmission line can change in
magnitude and phase in appreciable ways before reaching other components the transmission
line connects. Due to this, we need a more general description for how voltages and currents
change as they move along the transmission line.
The strategy to model this behavior is to initially focus on an infinitesimal portion of
a transmission line. We would typically draw this small section of transmission line in the

116
 CHAPTER 2. TRANSMISSION LINES

Figure 2.3: Schematic of an infinitesimal portion of a continuous transmission line.

Figure 2.4: Lumped element model of an infinitesimal portion of a continuous transmission

line.

manner shown in Fig. 2.3. Now, because the portion of the transmission line is very small
compared to the wavelength we can assume that we may describe the small changes that will
occur in the voltages and currents using lumped elements. That is, a standard circuit theory
expression of various components. A general lumped element model for the infinitesimal
portion of transmission line is shown in Fig. 2.4.
This circuit is composed of various per unit length quantities that characterize the be-
havior of the transmission line. These are:

ˆ R = series resistance per unit length, for all conductors present, in Ω/m ,

ˆ L = series inductance per unit length, for all conductors present, in H/m ,

ˆ G = shunt conductance per unit length, in S/m ,

ˆ C = shunt capacitance per unit length, in F/m.

As we will see later, these quantities can be computed for a specific geometry from a field-
based analysis of a cross-section of the transmission line using Maxwell’s equations. Briefly,
R accounts for the loss of power due to the finite conductivity of the conductors, while G
accounts for loss in the material potentially filling the transmission line geometry. Meanwhile,
L represents the total self-inductance of the geometry and C is a capacitance due to the close
proximity of the conductive surfaces forming the transmission line.

117
CHAPTER 2. TRANSMISSION LINES

Since we are only considering a very small portion of the overall transmission line, we
can use the basic techniques of circuit theory to determine equations for the voltages and
currents in Fig. 2.4. We begin with Kirchhoff’s voltage law (KVL) to get

V (z, t) − R∆zI(z, t) − L∆z∂t I(z, t) − V (z + ∆z, t) = 0. (2.1)

We can rearrange terms and divide by ∆z to get

V (z + ∆z, t) − V (z, t)
= −RI(z, t) − L∂t I(z, t). (2.2)
∆z
We recognize the left-hand side as being a finite difference approximation to a derivative.
Hence, if we take the limit of ∆z → 0, we will have that

∂z V (z, t) = −RI(z, t) − L∂t I(z, t). (2.3)

Similarly, we can use Kirchhoff’s current law (KCL) to get

I(z, t) − G∆zV (z + ∆z, t) − C∆z∂t V (z + ∆z, t) − I(z + ∆z, t) = 0. (2.4)

We rearrange terms to get

I(z + ∆z, t) − I(z, t)
= −GV (z + ∆z, t) − C∂t V (z + ∆z, t), (2.5)
∆z
which becomes

∂z I(z, t) = −GV (z, t) − C∂t V (z, t) (2.6)

after taking the limit of ∆z → 0.

In summary, the telegrapher’s equations are “Maxwell’s equations” for a simple 1D trans-
mission line system. By this, we mean that they are the governing partial differential equa-
tions for the voltage and current that exist on the transmission line. The telegrapher’s
equations in the time domain are

KVL : ∂z V (z, t) = −RI(z, t) − L∂t I(z, t), (2.7)

KCL : ∂z I(z, t) = −GV (z, t) − C∂t V (z, t). (2.8)

2.2.1 Lossless Line Wave Equation

We can combine these to derive a wave equation. This is easiest if we begin by assuming a
lossless case where R = G = 0. Then, we have

∂z V (z, t) = −L∂t I(z, t) (2.9)

∂z I(z, t) = −C∂t V (z, t). (2.10)

We take a derivative with respect to z on (2.9) to get

∂z2 V (z, t) = −L∂t ∂z I(z, t). (2.11)

118
 CHAPTER 2. TRANSMISSION LINES

We then use (2.10) to substitute for ∂z I(z, t). This gives us

∂z2 V (z, t) = LC∂t2 V (z, t). (2.12)

We can rearrange to finally have

∂z2 V (z, t) − LC∂t2 V (z, t) = 0. (2.13)

At this point, we can easily recognize this as a one-dimensional wave equation.

To proceed, let’s now convert to the frequency domain to have

∂z2 V (z) + β 2 V (z) = 0, (2.14)

where the phase constant is

√
β = ω LC, (2.15)

the wavelength is
2π
λ= , (2.16)
β
and the phase velocity can be seen to be
ω 1
vp = =√ . (2.17)
β LC
Since L and C depend on the geometry of the transmission line, we will find that all of these
parameters characterizing the propagation of waves along a transmission line are also very
much a function of its geometry.
We know that solutions of (2.14) will be of the form

V (z) = V0+ e−jβz + V0− ejβz . (2.18)

We can compute what I(z) will be by using

∂z V (z) = −jωLI(z). (2.19)

Evaluating the derivative gives us that

β + −jβz β − jβz
I(z) = V0 e − V e . (2.20)
ωL ωL 0
We can simplify this expression by defining a characteristic impedance. This is the ratio
between the voltage and current (with appropriate signs adopted to achieve a positive result).
For the lossless case, the characteristic impedance is defined as
r
L
Z0 = . (2.21)
C

119
CHAPTER 2. TRANSMISSION LINES

We can use this to see that (2.20) is

V0+ −jβz V0− jβz

I(z) = e − e . (2.22)
Z0 Z0
It should be noted that when using the characteristic impedance to quickly find a voltage or
current from the other, it is vital to remember the need to flip the sign if the wave if propa-
gating in the negative direction. This is similar to the sign changes that happened between
electric and magnetic fields when plane waves were propagating in opposite directions. In
the plane wave case, we had the built-in check on our work to avoid sign errors by making
sure that E×H pointed in the correct propagation direction. We no longer have this built-in
check for transmission line problems, and so we must be very careful to not forget to account
for this when we are working through a problem!

2.2.2 Lossy Line Wave Equation

We will now consider the full case of lossy transmission lines. For this situation, we have
R, G ̸= 0. In the frequency domain, the telegrapher equations become

∂z V (z) = −(R + jωL)I(z), (2.23)

∂z I(z) = −(G + jωC)V (z). (2.24)

We can combine these to form a wave equation, which is

∂z2 V (z) − γ 2 V (z) = 0, (2.25)

where
p
γ = α + jβ = (R + jωL)(G + jωC). (2.26)

For this situation, we will have that λ = 2π/β, where β is calculated from (2.26). The
characteristic impedance must also be updated for lossy lines. It now becomes
s
R + jωL
Z0 = . (2.27)
G + jωC

Considering this definition in comparison to the lossless case of (2.21), we see that if the
characteristic impedance is complex-valued we have a lossy line. Further, we see that it is the
imaginary part of the characteristic impedance that accounts for the loss in the transmission
line.

2.2.3 Computing R, L, G, and C

In order for the transmission line theory to be useful, we are going to need some way to
calculate all of the different per unit length transmission line parameters that we were using
in the previous section. We now turn our attention to doing this.

120
 CHAPTER 2. TRANSMISSION LINES

Figure 2.5: Illustration of a cross section of a transmission line where the per unit length
transmission line parameters can be computed from.

We assume that we already know the description of the fields that exist within a cross
section of the transmission line geometry (e.g., that illustrated in Fig. 2.5). Although we
won’t concern ourselves with this now, we can compute these fields by first solving Maxwell’s
equations for the transmission line geometry. We will revisit this later in the course.
We assume that we have a voltage V0 e±jγz and current I0 e±γz defined on the transmission
line. We will define the different per unit length parameters so that the definitions of power
and energy densities will give the same results between Poynting’s theorem and what we
know from circuit theory. Starting with L, we note that the time-averaged magnetic energy
density stored in the cross section of the transmission line will be equal to
ˆ
µ′
wm = H · H∗ dS. (2.28)
4 S
Meanwhile, the circuit theory result would be that
wm = L|I0 |2 /4. (2.29)
Hence, we can define the inductance per unit length to be
ˆ
µ′
L= 2
H · H∗ dS. (2.30)
|I0 | S
We may define the capacitance per unit length in a similar manner. We first note that the
time-average electric energy density stored in the cross section of the transmission line will
be
ˆ
ϵ′
we = E · E∗ dS. (2.31)
4 S
The circuit theory result for a capacitor would be
we = C|V0 |2 /4. (2.32)
Hence, we can define the capacitance per unit length to be
ˆ
ϵ′
C= E · E∗ dS. (2.33)
|V0 |2 S

121
CHAPTER 2. TRANSMISSION LINES

The power loss due to finite conductivity that is accounted for by the per unit length
resistance R can be calculated in a few different ways. One common approach to account for
this is to make the approximation that even though the conductor has finite conductivity,
the current on the conductor still flows as a surface current. We can then introduce a surface
resistivity to account for the loss that occurs in reality due to the finite penetration depth of
the fields and corresponding current flow in a real conductor. When we do this, we can then
compute the corresponding power loss in terms of the magnetic field that is tangential to the
cross section of the transmission line. These fields are proportional to the surface current
densities flowing along the conductors (think about the boundary condition). Hence, we can
write the power loss as
ˆ
Rs
PC = H · H∗ dℓ, (2.34)
2 ∂S
where Rs is the surface resistivity of the metal, ∂S denotes that the integral should be taken
over the boundary of the surface S, and dl is simply the one dimensional integration along
those bounding curves. The circuit theory result would have

PC = R|I0 |2 /2, (2.35)

so the resistance per unit length is

ˆ
Rs
R= H · H∗ dℓ. (2.36)
|I0 |2 ∂S

Finally, the per unit length conductance G needs to be defined. This is due to loss in the
dielectric material filling the medium. We would typically have this be
ˆ
ωϵ′′
Pd = E · E∗ dS (2.37)
2 S
from Poynting’s theorem. The circuit theory result is

Pd = G|V0 |2 /2, (2.38)

so the conductance per unit length is

ˆ
ωϵ′′
G= E · E∗ dS. (2.39)
|V0 |2 S

If magnetic materials are used that contribute magnetic loss, these formulas can be modified
as needed to account for these additional loss effects. However, this is a fairly uncommon
situation to encounter in typical practice, so this is not considered here.
Once these per unit length transmission line parameters have been computed for a ge-
ometry, we can continue with our analysis of a problem using the simplified one-dimensional
transmission line theory that we will continue to study. By breaking the analysis of a com-
plete problem into these smaller parts, we are able to greatly simplify the overall analysis,
leading to a very useful design and analysis tool that is still used today.

122
 CHAPTER 2. TRANSMISSION LINES

Figure 2.6: Example of common SMT component sizes sitting on the head of a matchstick
(image from Surface Mount Process).

2.3 Terminated Lossless Transmission Lines

One of the fundamental problems that we need to be able to analyze using transmission
line theory is what happens when a transmission line is terminated in some kind of load
impedance. For this kind of problem, we assume that the load impedance can be represented
as a lumped element. Implicit to this assumption is that the load impedance is small compared
to the wavelength of the voltage/current waves on the transmission line.
From a practical perspective, we tend to be able to use this kind of representation fairly
pervasively in microwave engineering when we use surface mount technology (SMT). These
are very small components that can be soldered completely on a single surface of a printed
circuit board (i.e., they do not have leads that poke through the board like larger components
you likely used in circuits labs). There are many standard sizes of SMT parts, but you can
readily buy ones that are only 20 mils long by 10 mils wide (see Fig. 2.6)! These often
have to be handled with tweezers and soldered under a microscope. However, even with
components that are this small, the physical nature of the devices that our waves interact
with can become important at microwave frequencies. As a result, it is often a certain kind
of “balancing act” in including enough approximate detail into our description of a lumped
element load impedance for our transmission line equations to provide sufficiently accurate
solutions for practical engineering design. We won’t be able to focus on this here, but it is
something that you will likely become quite familiar with in other advanced courses or in
your career if you work as a microwave engineer. Generally, many of these non-ideal effects
are described as being due to parasitic reactances that occur because of the physical features
of a practical component.

2.3.1 Derivation of Reflection Coefficient

We now return to considering the case of a transmission line terminated with a load impedance.
We illustrate the setup of our problem in Fig. 2.7. In it, we have our transmission line char-
acterized by Z0 and β. At z = 0, we have a load impedance ZL connecting the signal line
of the transmission line to ground. As is convention in solving transmission line problems,
we typically consider the transmission line to have some finite length ℓ that measures into
negative z values. This won’t be important for this particular problem, but will be used in

123
CHAPTER 2. TRANSMISSION LINES

Figure 2.7: Illustration of a transmission line terminated with a load impedance.

future scenarios. We further assume that we have some voltage wave propagating toward
the load impedance as part of our problem statement.
We can approach this problem using the same three step process we learned about for
solving Maxwell’s equations with plane wave solutions. In Step 1, we identify the regions of
the problem. Here, we simply have our transmission line for z < 0 and then at z = 0 we
have our lumped element ZL . As an excitation to the problem, we have an incident voltage
wave given by

V + (z) = V0+ e−jβz . (2.40)

The corresponding current is

V0+ −jβz
I + (z) = e . (2.41)
Z0
In Step 2, we propose a suitable general solution. Much like with plane waves encoun-
tering different kinds of materials, the load impedance acts as a discontinuity to our voltage
wave. Hence, we will need to have some reflected voltage wave be generated to ensure that
the boundary conditions of our problem are met. We will express this wave as

V − (z) = V0− ejβz , (2.42)

V0− jβz
I − (z) = − e . (2.43)
Z0
Note that the minus sign comes about for the current due to the spatial derivative on z in
the telegrapher’s equations. It is essential that you remember to include this minus sign, or
you are likely to get incorrect results in future problems!
We can now move to Step 3 where we use boundary conditions to solve for our unknown
quantities. For this problem, we need to solve for V0− . Now, what should our boundary
condition be? We don’t have anything clearly appropriate from the boundary conditions we
derived for fields previously, so we have to think a bit harder about this. It turns out that
our load impedance is actually our boundary condition. This is because the load impedance

124
 CHAPTER 2. TRANSMISSION LINES

by definition expresses the ratio of the voltage and current at the load impedance through
Ohm’s law. Hence, we will need to have the total voltage and current at the load impedance
give us
V (0)
ZL = , (2.44)
I(0)
where V and I are the total voltage and current on the transmission line. By recalling that

V (z) = V0+ e−jβz + V0− ejβz (2.45)

and
 
1 + −jβz − jβz
I(z) = V e − V0 e , (2.46)
Z0 0
we can fill in these expressions in (2.44) to get
V0+ + V0−
ZL = Z0 . (2.47)
V0+ − V0−
We can rearrange this to solve for V0− , giving
ZL − Z0 +
V0− = V . (2.48)
ZL + Z0 0
This formula should look rather familiar to you. In particular, the relationship between the
impedances looks almost identical to the reflection coefficient that we had when considering
a plane wave normally incident on a half space of some homogeneous medium. This com-
monality is no accident, and will happen time and time again when analyzing transmission
lines. It turns out that the analysis of transmission lines and plane waves normally inci-
dent on different regions of space are mathematically equivalent, since both are essentially
one-dimensional wave equations.
Now, considering this similarity, we often define a voltage reflection coefficient at the load
impedance as
V − (0) ZL − Z0
ΓL = = . (2.49)
V + (0) ZL + Z0
It is important to point out that by convention, the reflection coefficient is always defined
with respect to the voltages.

2.3.2 Other Characterizations of Reflections

Characterizing the reflections on a transmission line is such an important concept that a
number of different conventions have arisen over the years to describe this behavior. In
addition to the reflection coefficient, engineers will also often refer to the return loss (RL)
of a system. This is simply a logarithmic representation of the reflection coefficient as

RL = −20 log |ΓL | [dB]. (2.50)

125
CHAPTER 2. TRANSMISSION LINES

It is referred to as return loss because it helps characterize the “loss” in our overall system
due to power being reflected at a load rather than being absorbed into it. The minus sign is
in the expression to make the return loss always be a positive number. Oftentimes, engineers
will accidentally refer to something as a return loss without including this minus sign so that
they report a “negative loss”. This is unphysical for any passive system (i.e., it doesn’t have
any amplifiers/gain), and is an error that you should not make. Considering the range of
|ΓL |, we see that 0 < RL < ∞. Further, it should be noted that a larger return loss is a
better value (i.e., it represents smaller reflections). As a rule of thumb, a return loss of 10 dB
is generally considered a minimally acceptable value for “complex” parts (e.g., an antenna or
a phase shifter), while a value in the range of 20 to 40 dB is typical for simpler components
(e.g., an attenuator).
To see why the return loss characterizes the power “lost” to reflections, we can consider
the average power on our transmission line. Considering that
 
+ −jβz jβz
V (z) = V0 e + ΓL e (2.51)

and
V0+ −jβz
 
jβz
I(z) = e − ΓL e , (2.52)
Z0
the average power is
1 |V + |2
Pavg = Re{V (z)I(z)∗ } = 0 Re{1 − Γ∗L e−2jβz + ΓL e2jβz − |ΓL |2 }. (2.53)
2 2Z0
The middle two terms combine to be a purely imaginary quantity so that the average power
is
|V0+ |2
Pavg = (1 − |ΓL |2 ). (2.54)
2Z0
Hence, we see that the reflected power that lowers the average power delivered to a load is
proportional to the return loss given in (2.50).
There is yet another way that engineers will characterize the reflections on a transmission
line. This is with the voltage standing wave ratio (VSWR), which is the ratio of the maximum
and minimum voltage anywhere on the transmission line. This is also sometimes simply
referred to as the standing wave ratio (SWR). We can look at the total expression for the
voltage on our line to quickly determine what these values will be. In particular, we end up
having
Vmax 1 + |ΓL |
VSWR = = . (2.55)
Vmin 1 − |ΓL |
This quantity can range anywhere from 1 < VSWR < ∞, with larger values indicating a
worse system (i.e., larger reflections). A VSWR of 2 : 1 is roughly equivalent to a 10 dB
return loss. An illustration of the VSWR for a situation where ZL = RL < Z0 is shown in
Fig. 2.8.

126
 CHAPTER 2. TRANSMISSION LINES

Figure 2.8: Illustration of the voltage standing wave ratio.

2.3.3 Input Impedance and Impedance Transformation

Previously, we discussed how to incorporate a load on a transmission line into our transmis-
sion line analysis. We followed a very general purpose approach that was reminiscent of how
we solved certain plane wave problems earlier in the course. Although this is a solid general
purpose approach, it is rather laborious and quickly becomes impractical when we want to
analyze a complicated microwave network. We will now discuss certain useful “shortcuts”
that are invaluable in analyzing transmission line problems.
For now, we will return to considering the case of a transmission line terminated with
a load impedance. We illustrate the setup of our problem in Fig. 2.9. In it, we have our
transmission line characterized by Z0 and β. At z = 0, we have a load impedance ZL
connecting the signal line of the transmission line to ground. As is convention in solving
transmission line problems, we typically consider the transmission line to have some finite
length ℓ that measures into negative z values. We further assume that we have some voltage
wave propagating toward the load impedance as part of our problem statement.
At this point, we have seen that a terminated transmission line will often be accompanied
with some amount of reflected waves existing on the transmission line. Due to the presence
of these reflected waves, the total voltage becomes an oscillatory function of position, similar
to what we have seen previously in many plane wave and transmission line problems. This
brings us to the idea of an input impedance. You likely encountered this concept at some
point in your circuits courses. Basically, there are many practical situations where it is
useful to represent some complex network of circuit components by an effective impedance
that properly gives the relationship between the voltage and current that would exist at
the input terminals to this complex network. This helps us have a very simple problem to
consider from the viewpoint of a source/generator.
What happens with transmission lines is that the input impedance of some transmis-
sion line network will vary as a function of position due to the oscillations in the total
voltage and current as functions of position. Hence, the effect that some load impedance
will have on a source/generator will depend on how far the load is situated away from the
source. Importantly, we can characterize complex combinations of transmission lines and

127
CHAPTER 2. TRANSMISSION LINES

Figure 2.9: Illustration of a transmission line terminated with a load impedance.

load impedances as a simple lumped element input impedance. This provides us with a very
powerful strategy for solving transmission line problems. We can often remove the presence
of the transmission lines (and the corresponding wave behavior) by computing appropriate
input impedances and then simply use circuit theory to solve for the voltages and currents
that are of interest.
To see the basic way this works, we will again consider the simple case of Fig. 2.9. We
will now look at the input impedance of the system at some position −ℓ. For this, we have
V (−ℓ) V + (ejβℓ + ΓL e−jβℓ ) ejβℓ + ΓL e−jβℓ
Zin (−ℓ) = = Z0 0+ jβℓ = Z 0 . (2.56)
I(−ℓ) V0 (e − ΓL e−jβℓ ) ejβℓ − ΓL e−jβℓ
A more commonly used form can be found by replacing ΓL with
ZL − Z0
ΓL = (2.57)
ZL + Z0
and simplifying. Initially, this gives us
(ZL + Z0 )ejβℓ + (ZL − Z0 )e−jβℓ
Zin (−ℓ) = Z0 . (2.58)
(ZL + Z0 )ejβℓ − (ZL − Z0 )e−jβℓ
We can regroup terms to write this expression using trigonometric functions. This gives us
ZL cos(βℓ) + jZ0 sin(βℓ)
Zin (−ℓ) = Z0 . (2.59)
Z0 cos(βℓ) + jZL sin(βℓ)
Typically, the cosine function is factored out so that we only have to deal with one trigono-
metric function in the numerator and denominator. This gives us

ZL + jZ0 tan(βℓ)
Zin (−ℓ) = Z0 . (2.60)
Z0 + jZL tan(βℓ)

This is an incredibly useful formula. It allows us to quickly take a load impedance and
transform its effect into a new lumped element load impedance located at a more convenient
position in our transmission line network. This is often referred to as impedance transfor-
mation. Once the impedance transformation has been completed, we can treat the new load

128
 CHAPTER 2. TRANSMISSION LINES

as a lumped element in the rest of our transmission line analysis because the impedance
transformation has accounted for all of the complex wave interference effects that occur in
the transmission line section “transformed” out of the analysis. This is a significant simplifi-
cation, and allows us to rapidly solve complex problems that would be extremely tedious to
consider if we were to try and apply the full 3-step solution process we have used previously
at all sections of the transmission line system. Final Note: ℓ should be plugged into the RHS
of this equation as a positive value!
Before moving on, we briefly mention that this derivation may be repeated quite easily
for a lossy transmission line. In this case, the impedance transformation equation becomes

ZL + Z0 tanh(γℓ)
Zin (−ℓ) = Z0 . (2.61)
Z0 + ZL tanh(γℓ)

In this equation, tanh(x) is a hyperbolic tangent function.

2.3.4 Special Cases of Lossless Terminated Lines

It is useful to consider a few special cases of a terminated transmission line that occur rather
frequently in microwave engineering. The first case we will consider is what happens when a
transmission line is terminated with a short circuit. For this case, ZL = 0, so the impedance
transformation equation (2.60) becomes

Zin (−ℓ) = jZ0 tan(βℓ). (2.62)

This is purely imaginary and can take on values between +j∞ and −j∞. Hence, a short-
circuited transmission line can effectively look like any purely reactive impedance we can
imagine! This is quite useful for designing different kinds of transmission line systems,
especially matching networks. The voltage, current, and effective reactance of the short-
circuited transmission line is shown in Fig. 2.10. Looking at the variation of the voltage and
current in Fig. 2.10, we see that this short circuit boundary is similar to a PEC boundary
condition for plane waves.
The next case we will consider is what happens when a transmission line is terminated
with an open circuit. For this case, “ZL = ∞”, so the impedance transformation equation
(2.60) becomes

Zin (−ℓ) = −jZ0 cot(βℓ). (2.63)

This is also purely imaginary and can take on values between +j∞ and −j∞. Hence,
we get quite similar behavior to the short circuit case. The voltage, current, and effective
reactance of the open-circuit terminated transmission line is shown in Fig. 2.11. Looking at
the variation of the voltage and current in Fig. 2.11, we see that this open circuit boundary
is similar to a PMC boundary condition for plane waves.
In practice, it depends on the type of transmission line structure being used as to whether
it is convenient to implement a short or open circuit termination. Additionally, because the
type of reactance achieved closer to the line is opposite for open vs. short circuit terminations,

129
CHAPTER 2. TRANSMISSION LINES

Figure 2.10: Illustration of the (a) voltage, (b) current, and (c) reactance as a function of
position for a short circuit terminated line (image from [13]).

it can often be more practical to use whichever case will give you a shorter overall transmission
line in practice.
There are also some “special” lengths of transmission lines that are useful to be aware of.
The first is what happens if ℓ = λ/2 (or some multiple of λ/2). For this case, the tangent
function becomes 0 so that the input impedance becomes

Zin = ZL . (2.64)

That is, it simply looks like the load impedance.

Next, we also often want to consider quarter-wavelength sections of transmission line.
These are useful because they make the input impedance

Z02
Zin = . (2.65)
ZL

To see why this is the result, we note that for a quarter-wavelength line we will have that
βℓ = (2π/λ)(λ/4) = π/2. We can’t plug this into (2.60) directly because tan(π/2) → ∞.

130
 CHAPTER 2. TRANSMISSION LINES

Figure 2.11: Illustration of the (a) voltage, (b) current, and (c) reactance as a function of
position for a open circuit terminated line (image from [13]).

To evaluate the result, we can divide the numerator and denominator of (2.60) by tan(βℓ)
so that we get
ZL
+ jZ0
tan(βℓ)
Zin (−ℓ) = Z0 . (2.66)
Z0
+ jZL
tan(βℓ)
By now taking the limit as βℓ → π/2, we can readily see that we recover the result given
in (2.65). This type of line is significant because it can be very useful in developing simple
matching networks. It is often referred to as a quarter-wave transformer.

2.4 Quarter-Wave Transformer Matching

Let’s now consider a simple design example to see how a quarter-wave transformer can be
used in practice. For this example, consider that we have a known load impedance that

131
CHAPTER 2. TRANSMISSION LINES

Figure 2.12: Illustration of inserting a matching network between a load and transmission
line to eliminate reflections from the perspective of the connected transmission line.

Figure 2.13: Illustration of a quarter-wave transformer for matching a real load impedance.

is purely resistive (and thus is also purely real) that is different from the characteristic
impedance Z0 of the transmission lines we are using in the rest of our system. We want to
develop a simple matching network that we can insert in between the load resistance and
the transmission line in our system so that from the view of our transmission line there will
be no reflection due to the load resistance. This is schematically illustrated in Fig. 2.12.
One simple matching network we can use in this case is a quarter-wave transformer. This
solution is illustrated in Fig. 2.13. Our goal now becomes choosing an appropriate value
for the characteristic impedance of the quarter-wave transformer (Z1 ) so that the input
impedance at the output of this line matches the characteristic impedance of our system
(Z0 ).
From our previous analysis, we know that the input impedance of this quarter-wave
transformer will be

Z12
Zin = . (2.67)
RL

To find the necessary value of Z1 to achieve a perfect match (i.e., no reflections), we need to

132
 CHAPTER 2. TRANSMISSION LINES

Figure 2.14: Frequency dependence of the reflection coefficient for a quarter-wave transformer
for a load resistance of RL = 100 Ω and a system characteristic impedance of Z0 = 50 Ω
(image from [13]).

Figure 2.15: Frequency dependence of the reflection coefficient for Chebyshev multisection
matching transformers with varying numbers of sections for a load resistance of RL = 100 Ω
and a system characteristic impedance of Z0 = 50 Ω (image from [13]).

set Zin = Z0 . Doing this, we can solve for Z1 to find that

p
Z1 = Z0 RL . (2.68)

This solution is very straightforward, and comes with some drawbacks. One of the pri-
mary drawbacks is that the “perfectly matched” condition will only occur when the quarter-
wave transformer is a quarter-wave long (or an odd multiple of this) – i.e., at a single
frequency. Hence, for practical systems that must operate over a band of frequencies, this
simple solution may not be sufficient since the performance can reduce quite quickly as a
function of frequency. This effect is illustrated in Fig. 2.14, where the frequency dependence
of the reflection coefficient is shown for the case of RL = 100 Ω and Z0 = 50 Ω.
One simple way to overcome the bandwidth limitation is to utilize a more sophisti-
cated matching network design. Many examples exist, but they can often require the use

133
CHAPTER 2. TRANSMISSION LINES

Figure 2.16: Illustration of a typical transmission line problem involving a generator and
load that are not matched to the transmission line connecting the two.

of additional sections of transmission lines with different characteristic impedances (called

multisection matching transformers). The example of the bandwidth improvement possible
with different Chebyshev multisection matching transformers is shown in Fig. 2.15. Al-
though better results are possible, this comes at the cost at a more complex design and often
requires the use of more “real estate” (i.e., it takes up more area in the physical design).

2.5 Generator and Load Mismatches

In solving practical transmission line problems, it is often necessary to consider a source or
generator being connected to a transmission line. The simplest example is shown in Fig.
2.16. In this problem, we have some known source with a source impedance ZS connected to
a transmission line of length ℓ that is terminated in a load impedance ZL . In general, ZS may
not exactly equal Z0 , so we will have the possibility of the voltage wave that was originally
reflected at ZL also be reflected at ZS . This leads to an “infinite” number of reflections that
can make evaluating certain aspects of the problem somewhat involved. We already saw this
same kind of case when we considered a plane wave normally incident on a finite thickness
slab of a dielectric material. At that time, we formulated our problem but didn’t go through
the process of solving it due to the tedious nature. Here, we use what we have just learned
about transmission lines to see how we can very quickly extract the important information
from the problem without going through the full tedious process.
To begin, we note that our general solution of the total voltage wave in the transmission
line will still need to consist of a combination of a forward and backward traveling waves.
We know from the boundary condition at the load that we can express the total voltage
wave as
V (z) = V + (e−jβz + ΓL ejβz ). (2.69)
The catch here is that we don’t know what V + is because it depends on the voltage source
and the reflections that can be produced at z = −ℓ. To solve for V + we can use a boundary
condition at z = −ℓ. The easiest way to do this is to transform our load impedance back a
distance ℓ so that our entire circuit can be represented as a simple lumped element circuit.
This gives us the illustration shown in Fig. 2.17.

134
 CHAPTER 2. TRANSMISSION LINES

Figure 2.17: Illustration of a typical transmission line problem after transforming the
impedance back along the length of the transmission line.

Figure 2.18: Example circuit to look at the effects of generator and load mismatches.

We can easily see from Fig. 2.17 that Vin can be found from a simple voltage divider
circuit. This gives us
Zin
Vin = VS . (2.70)
ZS + Zin
We must have that V (z) equals Vin at z = −ℓ. Hence, we see that
Vin = V + (ejβℓ + ΓL e−jβℓ ). (2.71)
Equating (2.70) and (2.71), we can solve for V + . We get that
Zin VS
V+ = . (2.72)
ZS + Zin e + ΓL e−jβℓ
jβℓ

It is helpful to see this full process in action, so we now conclude with an example
calculation for the circuit shown in Fig. 2.18. Our goal is to determine expressions for the
voltage and current at any point in the transmission line. To begin, we need to evaluate the
different parts of (2.69) that will be needed to determine V + later. We will first evaluate the
load reflection coefficient, which for this circuit is
ZL − Z0 25 − 50 1
ΓL = = =− . (2.73)
ZL + Z0 25 + 50 3
Next, to determine V + we will need to use the boundary condition at z = −λ/4. Hence, we
will need to know what βℓ is equal to for this position to evaluate (2.69). We can quickly
find that
2π λ π
βℓ ℓ=λ/4 = × = , (2.74)
λ 4 2

135
CHAPTER 2. TRANSMISSION LINES

Figure 2.19: Example circuit after using an impedance transformation to simplify.

where we have used the fact that β = 2π/λ to simplify the evaluation. Plugging this into
(2.69) and simplifying gives us
 
+ jπ/2 1 −jπ/2
V (z = −λ/4) = V e − e
3
 
j (2.75)
=V+ j+
3
4
= j V +.
3
Now, to determine V + , we need to find what the total voltage V (z = −λ/4) should be
equal to. We can do this by reducing our transmission line circuit into an equivalent lumped
element circuit (i.e., all the wave portions have already been accounted for elsewhere). This
is done by finding the input impedance of the circuit from the position ℓ = −λ/4. We use
the impedance transformation formula to find this. We write this in a slightly different form
of
ZL
+ jZ0
tan(βℓ)
Zin (ℓ) = Z0 (2.76)
Z0
+ jZL
tan(βℓ)

to simplify the analysis. The reason that this helps is because when we plug in for βℓ → π/2
we have that tan(βℓ) → ∞. With the rearranged form of (2.76) we can quickly evaluate this
limit to find that
Z02
Zin = , (2.77)
ZL
which is the result that we have already discussed for a quarter-wave transformer. Plugging
in our numbers for Z0 and ZL we find that Zin = (50)2 /25 = 100 Ω. Our simplified circuit
after this impedance transformation is now shown in Fig. 2.19. This simple voltage divider
tells us that Vin = 2/3.
Considering this, we can now find V + by recognizing that Vin = V (z = −λ/4). Using
our result from (2.75) and that Vin = 2/3 we find that
4 2 j
j V+ = =⇒ V + = − . (2.78)
3 3 2

136
 CHAPTER 2. TRANSMISSION LINES

Figure 2.20: Circuit diagram for a generator and load connected by a transmission line.

We can now use this in (2.69) to find that

 
j −jβz 1 jβz
V (z) = − e − e . (2.79)
2 3

We can also evaluate the current quickly by remembering that we simply need to divide the
total voltage by the characteristic impedance of the line and switch the sign on the portion
of the wave propagating in the −z-direction. Considering this, we have that
 
j −jβz 1 jβz
I(z) = − e + e . (2.80)
2(50) 3

2.6 Power Transfer

We now begin to consider general expressions for how much power is transferred to a load
of a transmission line network. Our presentation will follow that given in [13]. Typically,
we are interested in maximizing the power that we deliver to a load. For instance, we can
view an antenna as a load to a transmission line system. The more power we deliver to
the antenna the more power the antenna can correspondingly radiate to achieve different
applications (e.g., communication or radar systems).
To analyze the relevant scenario, we will again consider the case where both the generator
and load impedance may not be matched to the transmission line connecting the two devices.
This is illustrated in Fig. 2.20. We won’t go through the detailed derivation again to solve
this problem. However, recall that the process that we went through started by transforming
the load impedance back along the length of transmission line. Once we did this, we had a
simple circuit shown in Fig. 2.21 that we could solve using standard circuit theory equations
since the wave aspects of the problem had been accounted for already.
After solving for all the necessary unknown variables, we found that

Zin
Vin = VS , (2.81)
ZS + Zin

137
CHAPTER 2. TRANSMISSION LINES

Figure 2.21: Circuit diagram for a generator and load connected by a transmission line after
the load has been transformed back along the length of the transmission line.

where the input impedance was given by

ZL + jZ0 tan(βℓ)
Zin = Z0 . (2.82)
Z0 + jZL tan(βℓ)
For the purposes of this section, it will be much easier for us to just consider Zin in a general
sense and not worry about its explicit expression in terms of this formula. It is simply
important to remember that whatever we determine about Zin can be eventually related
back to ZL following some appropriate math.
We now want to determine how much power is transferred to the load under different
operating conditions that commonly occur. This analysis can be simplified by noting that
because we have a lossless transmission line (since Z0 is assumed to be real for this problem)
all of the real power delivered to Zin in our simplified circuit schematic will have to be
absorbed in the load. Hence, we have that the power delivered to the load is
1 ∗ 1 −1
PL = Re{Vin Iin } = |Vin |2 Re{Zin }. (2.83)
2 2
We can use our expression for Vin from (2.81) to get
2
1 Zin −1
PL = |VS |2 Re{Zin }. (2.84)
2 Zin + ZS
If we now express ZS = RS + jXS and Zin = Rin + jXin , we can eventually simplify our
above expression to be
1 Rin
PL = |VS |2 . (2.85)
2 (Rin + RS ) + (Xin + XS )2
2

We will now consider three different scenarios. First, when the load impedance is matched
to the transmission line impedance. Second, when the generator is matched to the loaded
line (i.e., matched to Zin ). Finally, what is known as conjugate matching.
Load Matched to the Transmission Line
Mathematically, having the load matched to the transmission line means that no reflections
will be produced at the interface between the load and the transmission line. Recalling that
the reflection coefficient at the load is
ZL − Z0
ΓL = , (2.86)
ZL + Z0

138
 CHAPTER 2. TRANSMISSION LINES

we see that for ΓL = 0 we will need ZL = Z0 . We can also see from our impedance
transformation formula (2.82) that Zin = Z0 for this case as well. Hence, the power delivered
to the load given in (2.85) becomes
1 Z0
PL = |VS |2 . (2.87)
2 (Z0 + RS )2 + XS2
Generator Matched to the Loaded Line
For the generator to be matched to the loaded line means we should have Zin = ZS . From a
practical perspective, if we want to achieve this we can typically do so by making appropriate
selections of βℓ and Z0 . That is, by changing properties of the transmission line connecting
the generator and load.
For this situation, we can consider there to be no overall reflection coefficient from the
perspective of the generator because
Zin − ZS
ΓS = = 0. (2.88)
Zin + ZS
However, this does not mean that there are no reflections in the overall system! In particular,
it is likely that we will have a standing wave exist on the transmission line since Z0 and ZL will
in general not be equal to each other for this case. We are simply adjusting the parameters
of the system so that this standing wave “looks” like a perfectly matched line from the view
of the generator.
Now, the power transferred to the load becomes
1 RS
PL = |VS |2 . (2.89)
2 4(RS + XS2 )
2

If we compare this to (2.87), we can see that the power transferred to the load for the
current scenario may actually be less than the case where only the load was matched to the
transmission line. Hence, to determine under what conditions we will transfer a maximum
amount of power to a load will require us to be more clever in our approach. We will consider
this next.
Conjugate Matching
To determine the maximum power that can be transferred to a load, we will assume that
the generator impedance is fixed. This is often reasonable for many practical applications
since our sources will typically be a commercial-of-the-shelf (COTS) part that we don’t have
much design control over. Depending on your application, the load impedance may very well
be fixed as well. However, because we will often have some freedom over how we connect
the generator and load to one another, we can still consider Zin to be something that is
reasonably under our control as designers.
With this in mind, we will determine how to maximize power transfer to the load by
differentiating (2.85) with respect to the real and imaginary parts of Zin and setting the
results equal to 0. Beginning with the real part, we have that
 
∂PL 1 2 1 −2Rin (Rin + RS )
= |VS | + . (2.90)
∂Rin 2 (Rin + RS )2 + (Xin + XS )2 [(Rin + RS )2 + (Xin + XS )2 ]2

139
CHAPTER 2. TRANSMISSION LINES

If we set this equal to 0 and do some algebra, we can eventually find that this will require

RS2 − Rin
2
+ (Xin + XS )2 = 0. (2.91)

We can similarly differentiate with respect to the imaginary part of the input impedance to
get

∂PL 1 −2Rin (Xin + XS )

= |VS |2 . (2.92)
∂Xin 2 [(Rin + RS )2 + (Xin + XS )2 ]2

If we set this equal to 0, we will need to have

Rin (Xin + XS ) = 0. (2.93)

We now need to solve (2.91) and (2.93) together to get

Xin = −XS , Rin = RS . (2.94)

We can write this compactly as

Zin = ZS∗ , (2.95)

which is why this is referred to as conjugate matching. For this case, the power delivered to
the load becomes
1 1
PL = |VS |2 , (2.96)
2 4RS
which is larger than the previous cases we considered.
From an intuitive perspective, this case can be seen to potentially have many multiple
reflections occurring since the load is not matched to the transmission line and the input
impedance is also not matched to the generator. However, we arrive at a larger overall power
transfer than the previous two cases we considered because we have found a case where the
multiple reflections can add in phase constructively. The end result is a larger amount of
power going to the load.
It should also be mentioned that often times it is desirable to make the reactive part of
the generator impedance as low as possible. If this happens, we end up with an intuitive
result that maximum power transfer will happen if the input impedance of the loaded line
is matched to the generator impedance, i.e., Zin = ZS = RS . Many types of loads may have
some amount of reactance to them (an antenna is an excellent example where this is the
case), which can make having Zin = RS over a large bandwidth somewhat difficult if we only
rely on the impedance transformation formula to modify Zin . In these cases, it is often useful
to design an impedance matching circuit that sits in between the final load impedance and
the transmission line that connects the load to the generator. This circuit often serves the
purpose of eliminating the reactive part of the load impedance and leaving the real part of
the impedance as something that can be reasonably connected to the transmission line or
generator without causing many reflections. We will learn a bit about a simple impedance
matching circuit later in this course.

140
 CHAPTER 2. TRANSMISSION LINES

Figure 2.22: Simplified circuit diagram of a Doppler radar.

2.7 Transmission Line Ports and Impedances

Designing practical microwave systems often involves connecting a number of components
together to achieve a desired goal. For instance, when we discussed the basic idea of how
a Doppler radar worked we needed to use a number of different components to generate
our desired signals, route them through our network, and eventually process them. As a
refresher, we include again the simplified schematic of the Doppler radar in Fig. 2.22.
Looking at this as a block diagram, we see that the different blocks/components often
have a different number of “inputs” and “outputs” depending on the purpose of the device.
In a microwave engineering context, we will refer to each of the locations where we can
connect to a device as a terminal or port. To make analysis of large systems tractable, we
often only want to consider how a particular device behaves at its ports (i.e., we don’t want
to worry about finding the voltage and current at every point within the device). This is very
much like specifying “input” and “output” relationships of a block diagram like you would
see in a signal processing or computer engineering course. The complication here is that
because we are dealing with waves that can reflect at various points in a practical device, we
must consider every port to simultaneously be an “input” and “output” port! As a result, we
will need to use appropriate analysis tools to understand how to design microwave systems.
We now begin to learn about how to do this by introducing a few common ways systems are
described in terms of ports and what are referred to as port parameters. This overall theory
is typically referred to as microwave network analysis.

2.7.1 Voltage and Current – A Closer Look

To begin, we need to take a harder look at some of the concepts we have been using in
describing transmission lines up to this point. In particular, we have been somewhat vague

141
CHAPTER 2. TRANSMISSION LINES

about what exactly a voltage and current wave look like and how they should be explicitly
defined for a given transmission line. The unfortunate reason for this is that with the
exception of a particular set of special cases, the definitions of voltage and current for a
transmission line are somewhat arbitrary, i.e., there is no unique definition that works for
every situation of interest.
We won’t dwell on this complication in detail in this course. We will simply mention that
a relatively standard way of dealing with this ambiguity is to follow a few guiding principles
that often work well for practical situations. First, voltage and current are usually defined
to be proportional to the transverse electric and magnetic fields of the transmission line,
respectively. Further, we will want to define the voltage and current in a manner so that
their product gives us the correct power flow on the transmission line for the particular
set of fields that exist (i.e., we want our circuit power flow to match the magnitude of the
Poynting vector). Finally, we want the ratio of the voltage to the current to be equal to the
characteristic impedance of the transmission line. Unfortunately, because the voltage and
current are ambiguous, the characteristic impedance of a general transmission line is also
somewhat arbitrary. Typically, we either choose the characteristic impedance to be equal to
the wave impedance of the transmission line fields or to be equal to 1.
Now, for simple kinds of transmission lines known as transverse electromagnetic (TEM)
lines, we can actually uniquely define voltages and currents. These kinds of lines always must
have at least two distinct conductors that can be held at different potentials, as illustrated
in Fig 2.23.
Now, for this kind of system the transverse electric field is electrostatic in nature. As a
result, we can define a unique voltage as
ˆ −
V = E · dℓ. (2.97)
+

The “uniqueness” is in the sense that the voltage does not depend on the integration path
taken because the electric field can be expressed as the gradient of some scalar potential
function. To determine the total current of the transmission line, we can use Ampere’s law
to get
˛
I= H · dℓ. (2.98)
C

The integration path is usually taken to be any closed path that encompasses the entire
“positive” conductor. We can then determine the characteristic impedance of the line as
V
Z0 = . (2.99)
I
If we have also determined the propagation constant of the line, we can proceed with using
the circuit theory of transmission lines that we have considered previously.

2.7.2 One-Port Network

Let’s now consider an arbitrary one-port network to understand how we can use concepts of
impedance to make a connection between the full electromagnetic field analysis problem and

142
 CHAPTER 2. TRANSMISSION LINES

Figure 2.23: Illustration of a two conductor transmission line that can support a TEM mode.
The solid (dashed) lines are the electric (magnetic) field lines.

Figure 2.24: Illustration of an arbitrary one-port network from the perspective of Poynting’s
theorem.

a simpler transmission line network one. We will consider the problem illustrated in Fig.
2.24. We assume that the only way that power and energy can enter/exit the arbitrary one-
port network is through a terminal plane. Attached to the terminal plane is a transmission
line whose cross section exactly matches the configuration of conductors and materials at
the terminal plane of the one-port network.
Considering this, the complex power delivered or extracted from the one-port network is

143
CHAPTER 2. TRANSMISSION LINES

given by
¨
1
P = E × H∗ · dS = Pℓ + 2jω(Wm − We ), (2.100)
2 S

where Pℓ is the power dissipated in the one-port network and Wm (We ) is the total magnetic
(electric) energy density in the one-port network. Now, let’s assume that we can express the
electric and magnetic fields over the terminal plane as

Et (x, y, z) = V (z)e(x, y) (2.101)

Ht (x, y, z) = I(z)h(x, y) (2.102)

and that the transverse fields e and h are normalized such that
¨
e × h∗ · dS = 1. (2.103)
S

Then we can use these expressions to write the complex power at the terminal as
¨
1 1
P = V (z)I ∗ (z)e × h∗ · dS = V (z)I ∗ (z), (2.104)
2 S 2
where z would be set at the terminal plane.
We can use this relationship to consider a definition of an overall input impedance for
the arbitrary one-port network. In particular, we will have that
V
Zin = Rin + jXin = . (2.105)
I
We can work this into a form that relates to the power quantities that we had in (2.104) by
multiplying (2.105) by I ∗ /I ∗ . This gives us
V I∗ 2P
Zin = 2
= 2. (2.106)
|I| |I|

We can now substitute for P from our original statement of Poynting’s theorem in (2.100)
to get
2Pℓ + 4jω(Wm − We )
Zin = . (2.107)
|I|2
This provides us with a very important result. We have taken a potentially very complicated
one-port device and simplified it down to an effective lumped element impedance from the
perspective of our transmission line. We can incorporate this lumped element impedance
into our transmission line analysis in the same way as we had been doing previously for
what we were treating as some kind of arbitrary impedance that was just “there” as part of
the problem. This gives us a way to start using our transmission line theory for practical
problems involving complex microwave devices.

144
 CHAPTER 2. TRANSMISSION LINES

Now, from (2.107) we can see that the impedance of our network will only have losses if
it has a non-zero real part. We also see that if Wm > We then the network will be inductive
and if We > Wm the network will look like a capacitive load. Although we won’t go into
detail on this, it should be noted that the effective resistance and reactance of the one-port
network will generally be a function of frequency (sometimes a strong one at that).
It is very important to remember what an impedance represents in order to properly in-
terpret it. For instance, we have just seen that when we characterize an arbitrary microwave
network in terms of a lumped impedance it will only have a real part if the microwave net-
work contains some amount of loss. To make things more complicated, this is the exact
opposite behavior to what we had when discussing the characteristic impedance of a trans-
mission line. For transmission lines, the characteristic impedance for a lossless line had to
be purely real. We must be careful to keep these things straight. To help with this, we
summarize in the next section some details about the different kinds of impedances that we
have encountered throughout the course. With time you will become adept at recalling these
different properties, but starting out it is essential to make sure you are careful and don’t
accidentally make a mistake due to the many different kinds of impedance concepts we use
in electromagnetic engineering.

2.7.3 Summary of Impedances

1. Intrinsic impedance of a medium
r
µ
ˆ Formula: η = .
ϵ
ˆ Purpose: this impedance characterizes how a material medium will affect the
relationship between the electric and magnetic fields in an unbounded medium.
It is only a function of the material properties of a material.
ˆ Real/Imaginary/Complex: purely real in a lossless medium, becomes a complex
number in a lossy medium.

2. Wave impedance
Etan
ˆ Formula: Zw = .
Htan
ˆ Purpose: this impedance is the ratio of the tangential components of the electric
and magnetic fields. It depends on the type of wave that exists in a medium,
and as a result can be dependent on the type of waveguide supporting the wave.
Generally, it will also be a function of frequency. We saw how this could be useful
when studying the reflection and transmission coefficients for oblique incidence
plane waves on the boundary between homogeneous materials. It is also useful
in defining the characteristic impedance of complicated waveguides that don’t
support TEM waves.
ˆ Real/Imaginary/Complex: purely real for a propagating wave in a lossless medium.
The wave impedance often has η somewhere in its definition, so it will inherit the
complex-valued nature of the intrinsic impedance in a lossy medium.

145
CHAPTER 2. TRANSMISSION LINES

3. Characteristic impedance

V+
ˆ Formula: Z0 = .
I+
ˆ Purpose: this impedance characterizes the ratio of voltage and current of a trav-
eling wave on a transmission line. It depends on the geometry of the transmission
line, the materials that may exist around the transmission line, as well as the kind
of wave supported by the transmission line.
ˆ Real/Imaginary/Complex: purely real for a lossless transmission line, becomes a
complex number for a lossy line.

4. Terminal/Port/Network impedance

ˆ Formula: Z (explicit formula depends on properties of the system it is character-

izing).
ˆ Purpose: this impedance characterizes the ratio of the total voltage and current
at the terminal or port of an arbitrary transmission line or microwave network.
It depends on all of the properties of the system it is characterizing. The input
impedance of a terminated transmission line can be viewed as a simple example of
this kind of impedance. This kind of impedance is useful for building equivalent
circuit models of complex devices.
ˆ Real/Imaginary/Complex: purely imaginary for a lossless system, becomes a real
or complex number if the system is lossy.

2.8 Introduction to Port Parameters

Now that we have an idea of what a port is, we can begin to look at how systems with many
ports are typically described. We will begin by discussing the impedance and admittance
matrix approach, followed by the very important scattering matrix. Often, people will refer
to these quantities as being port parameters. Due to this terminology, one will often refer
to the impedance parameters or the scattering parameters, for instance, of a system when
discussing the properties of different microwave components.
For all of these parameters, the general setup that we are considering is shown in Fig.
2.25. In it, we assume that we have some arbitrary electromagnetic system that eventually
transitions into N -ports at which we can imagine connecting a long section of a constant
cross-section transmission line. At some location along the transmission line, we consider
that we have a reference or terminal plane. We will define all of our voltages and currents
with respect to these reference planes.
We should note before moving on that a single physical “port” of a device may be able
to support multiple kinds of waves. As a result, we will have to introduce additional effective
ports to our description of our arbitrary system. In general, each impedance/admittance/scattering
parameter will only account for one kind of wave at each physical port a device may have.

146
 CHAPTER 2. TRANSMISSION LINES

Figure 2.25: Diagram of an arbitrary N-port microwave network.

2.8.1 Impedance Parameters

The impedance parameters of an N-port network relate the total voltage and currents that
exist at each terminal plane of a device to each other. At the nth terminal plane, we have
that the total voltage and current will be given by

Vn = Vn+ + Vn− , (2.108)

In = In+ − In− , (2.109)

where the right hand sides of each of these equations are written in terms of the forward-
and backward-propagating voltages and currents that make up the total waves on the trans-
mission line.
The impedance parameters can then be used to relate these voltages and currents to
each other. A matrix notation is typically best for this, so we write the impedance matrix
as [Z] and column vectors of voltages and currents as {V } and {I}. Within this notation,
our relationship between voltages and currents is compactly written as

{V } = [Z]{I}. (2.110)

This can be expanded out explicitly as

    

 V1  Z11 Z12 . . . Z1N   I1 
 V2 
   Z21 Z22 . . . Z2N   I2 
 
= . . (2.111)
 
.. .. .. . .
 .. ..   .. 



 . 
 . .  
  
 
VN ZN 1 . . . . . . ZN N IN
 

147
CHAPTER 2. TRANSMISSION LINES

Figure 2.26: Circuit diagram of a two-port T-network.

How do we go about calculating or measuring one of these impedance parameters in

practice? We know that we need to be careful about this due to the wave characteristics
that are going to exist in our device. To figure out the correct procedure, we can simply look
at the structure of (2.111). In particular, we see that

Vi
Zij = . (2.112)
Ij Ik =0, ∀k̸=j

In words, we can find Zij by driving port j with a current Ij and then measuring the total
voltage at port i. For this to work properly, we need to make sure that we leave all other
ports of the device open-circuited so that no current can exist at the other ports. Due to the
complexity in defining voltages for some transmission lines, the measurement of impedance
parameters is not extremely common. However, they can be computed from other commonly
performed measurements and there are certain kinds of numerical methods that can be used
to calculate them directly. As a result, the impedance parameters remain an important
diagnostic tool for understanding the properties of a particular microwave network.
Of particular interest are the Zii entries of the impedance matrix. These are the input
impedances of each port when the other ports are open-circuited. This information can
be particularly useful when trying to design matching circuits for a particular microwave
network. As we will touch on later in the course, designing a matching circuit often in-
volves “canceling” out certain undesirable impedance behaviors as a function of frequency
(e.g., some excess reactance, similar to what we had for conjugate matching). Hence, un-
derstanding the impedance parameters of a device can be a fundamental step in this design
process.

Impedance Parameter Derivation Example

Let’s look at how to derive the impedance parameters for a two-port T-network. This system
is shown in Fig. 2.26. T-networks find a number of uses in microwave engineering. They
can be used in simplifying complex circuits, in providing a reasonable approximation for the
behavior of many simple two-port devices, as well as in forming a useful configuration for
implementing certain kinds of attenuators.
Let’s start by deriving the input impedances, Z11 and Z22 . Beginning with Z11 , we will
need to determine the ratio of V1 and I1 when port 2 is left open-circuited. We see that for
this situation the total impedance seen by port 1 will simply be the series combination of

148
 CHAPTER 2. TRANSMISSION LINES

ZA and ZC . Hence, we have that

V1
Z11 = = ZA + ZC . (2.113)
I1 I2 =0

A similar process may be followed for Z22 to quickly yield

V2
Z22 = = ZB + ZC . (2.114)
I2 I1 =0

Now, for the transfer impedance Z12 we need to measure the open-circuit voltage at port 1
when port 2 is driven by a current I2 . Since port 1 is open-circuited, no current will flow
through ZA , so the port 1 voltage will simply equal the voltage drop over ZC . This can be
given by a voltage divider to be
ZC
V1 = V2 . (2.115)
ZB + ZC
The trouble here is that we don’t know V2 because it is the voltage drop over an ideal
current source. This dilemma can be quickly solved by noting that if we plug (2.115) into
the definition for Z12 for V1 we will get
V1 ZC V2
Z12 = = . (2.116)
I2 I1 =0 ZB + ZC I2

We can recognize V2 /I2 as being equivalent to Z22 . Hence, if we plug (2.114) into the above
expression for V2 /I2 we will get
V1
Z12 = = ZC . (2.117)
I2 I1 =0

Alternatively, we could have also noted that since the ideal current source will cause current
I2 to flow through the series combination of ZB and ZC (since port 1 is open-circuited) the
actual current in ZC will be I2 . Hence, the voltage drop will be given by V = ZC I2 , which
is equal to V1 due to the open circuit at port 1. As a result, V1 /I2 = ZC . Now, by the
symmetry of the system, we can quickly recognize that Z21 = ZC as well. We could verify
this easily following the same process as what we used to derive the expression for Z12 given
in (2.117).
Consolidating our results together, we have that
 
ZA + ZC ZC
[Z] = . (2.118)
ZC ZB + ZC
This is a symmetric matrix, which is a characteristic that indicates that this system is
reciprocal. That is, the transfer characteristics of the device do not depend on which direction
we apply a signal from. Most practical microwave components are reciprocal. However, some
very important non-reciprocal devices play key roles in building communication and radar
systems. For instance, the circulator that we saw in the simplified schematic of a Doppler
radar was a kind of non-reciprocal component.

149
CHAPTER 2. TRANSMISSION LINES

2.8.2 Admittance Parameters

For certain types of systems/circuits, it can be more convenient to work with a set of ad-
mittance parameters. These are very similar to the impedance parameters, and are in some
sense their “inverse” representation. In particular, we have that
{I} = [Y ]{V }, (2.119)
where [Y ] is the admittance matrix. The individual elements of the admittance matrix can
be computed as
Ii
Yij = . (2.120)
Vj Vk =0, ∀k̸=j

In words, we can find Yij by driving port j with a voltage Vj and then measuring the total
current at port i. For this to work properly, we need to make sure that we leave all other
ports of the device short-circuited so that no voltage can exist at the other ports. From the
similarity of the matrix equations relating voltages and currents in terms of [Z] and [Y ], we
can quickly see that [Y ] = [Z]−1 .

2.8.3 Scattering Parameters

Although the impedance and admittance parameters can be a useful tool, they are not always
convenient to measure, calculate, or use. One of the main difficulties comes from the fact
that working with the total voltage and current at a port is not typically how we want to
think of a device’s operation. Generally, it is far more intuitive to think of the device in terms
of a set of input and output relationships. The scattering parameters, often abbreviated as
S-parameters, provide us with such a tool.

Introduction
In particular, the S-parameters relate an incident set of voltage waves to the voltage waves
leaving each port of an N-port microwave network. In matrix form, known as the scattering
matrix, we have
{V − } = [S]{V + }, (2.121)
which can be expanded out to be
 −    +
 V1  S11 S12 . . . S1N  V1 
V2−  . . . S2N  V2+ 
   
   S21 S22  
 
= . . (2.122)

.. .. .. ..  ..

 . 
  .. . . .   . 

 −
  
 + 
VN SN 1 . . . . . . SN N VN
A specific S-parameter can be calculated as

Vi−
Sij = . (2.123)
Vj+
Vk+ =0, ∀k̸=j

150
 CHAPTER 2. TRANSMISSION LINES

Figure 2.27: Example of a 3 dB attenuator matched to 50 Ω.

In words, we need to measure the voltage wave leaving port i when there is only an incident
wave coming in at port j. For all the incident waves to be zero except at port j we need to
terminate each port of the device not actively being used with a matched load. A matched
load perfectly absorbs all of the energy incident on it from the port so that there are no
reflections that can become an unintended V + at a particular port.
Considering these definitions, we can interpret the different S-parameters in the following
manner. Each Sii is the reflection coefficient at a particular port when all other ports have
been terminated with matched loads. Similarly, Sij is a transmission coefficient from port
j to port i when all other ports have been terminated with matched loads. It is important
to remember the ordering of the indices for this relationship, the second index is the source
and the first index is the receiver.

Derivation Example
Let’s now consider an example of how to derive the S-parameters for a particularly simple
device. In particular, we will consider a kind of 3 dB attenuator shown in Fig. 2.27. This
device is designed to dissipate half of the power transmitted through it (hence, 3 dB).
Although it may initially seem odd for us to want to simply remove power from our system,
there are many practical scenarios where it is desirable to use attenuators. A few examples are
providing protection to sensitive circuits that can be damaged if too much power is incident
on them, equalizing the powers coming from different sources, and improving the impedance
match of certain devices. Further, it is often much easier to generate a tunable/variable
attenuator rather than a source that has a tunable output power, so variable attenuators
can also be useful in generating a desired variable output power level.
To begin, let’s find S11 . By definition, this will be the reflection coefficient at port 1
when port 2 is terminated with a matched load. To determine this, we need to know what
impedance this device is matched to. Due to the symmetry of the device, we can comfortably
conclude that the characteristic impedance that port 1 and port 2 should be matched to will
be equivalent. Considering this, we can conclude that Zin,1 = Z0 , where Zin,1 is the input
impedance at port 1 when port 2 is terminated in Z0 . Considering this, we can expand Zin,1
as
141.8(8.56 + Z0 )
8.56 + . (2.124)
141.8 + (8.56 + Z0 )

151
CHAPTER 2. TRANSMISSION LINES

Figure 2.28: Computing S21 for the 3 dB attenuator example.

Setting this equal to Z0 , and then solving for Z0 , gives us Z0 = 50 Ω. This is actually a
standard characteristic impedance that almost all practical microwave engineering circuits
operate at. Almost all measurement instruments and COTS devices are matched to 50 Ω.
Now, since we know that Zin,1 = Z0 = 50 Ω we can return to calculating S11 . To do this,
we need to compute the reflection coefficient at port 1 when we attach a transmission line
with a 50 Ω characteristic impedance to it and have port 2 terminated in Z0 as well. This
tells us that
Zin,1 − Z0
S11 = = 0. (2.125)
Zin,1 + Z0

Due to the symmetry of the device, we see that S22 = 0 as well.

Let’s now compute S21 , while also noting that due to the symmetry S21 = S12 . By
definition, to calculate S21 we need to inject an incident voltage wave V1+ at port 1 and
measure the output voltage wave V2− at port 2. We must do this when V2+ = 0, which
requires that we imagine connecting an infinite length transmission line of impedance Z0
at port 2. We are then trying to compute the transmission coefficient into this impedance.
Now, from the perspective of port 1 there is no difference between having an infinite length
transmission line of Z0 and a lumped impedance of Z0 at port 2. Hence, we really just need
to compute the voltage over a lumped element impedance of Z0 at port 2 to determine S21 .
Our corresponding circuit is shown in Fig. 2.28.
We begin finding the voltage over the Z0 impedance in Fig. 2.28 by first finding the
voltage over the parallel combination of 141.8 Ω and Z0 + 8.56 = 58.56 Ω. This parallel
combination is 41.44 Ω, and by a simple voltage divider we can find the voltage over it as
41.44
Veq = V1 = 0.8288V1 . (2.126)
41.44 + 8.56

We then find V2− = V2 as another voltage divider of Veq over the 8.56 Ω and Z0 impedances
in Fig. 2.28. This gives us that
50 1
V2− = Veq = 0.707V1+ ≈ √ V1+ . (2.127)
50 + 8.56 2
√
Hence, we see that S21 ≈ 1/ 2. If we recognize that the power transmitted through the
device is proportional to |S21 |2 , we see that half of the input power will emerge at port 2. In

152
 CHAPTER 2. TRANSMISSION LINES

dB, we have the output power is proportional to 20 log10 (|S21 |) = −3 dB. Since there are no
reflections at port 1, the remaining power is dissipated in the resistors in the device.
This computation of S21 should hopefully help you understand why it is so convenient for
us to try and make sure that all of our devices are matched to a single common impedance
reference. If this were not the case, we would have significantly more reflected waves occurring
in our devices and it would be far more difficult to compute the S-parameters. In reality, no
device is matched perfectly and so all of these reflections will still occur and it is often very
important for us to computationally account for them for many practical designs. However, if
we can keep the reflections low we can typically come up with a reasonable intuition for how
a system will operate without requiring these fully detailed analyses. When the full analysis
is needed, there are various analytical tools to solve these problems, as well as powerful
numerical methods available to us. We won’t learn about these more general methods in
this class, but if you take a course on microwave engineering you will have an opportunity
to learn more about these different analysis methods.

Measuring S-Parameters
Accurately measuring the S-parameters of a microwave device is often a very important task
in the development of a design. From the definition of the S-parameters, we see that this kind
of measurement is certainly going to require a much more sophisticated piece of technology
than what we are used to working with in typical circuits labs. The particular instrument
that is most often used to measure S-parameters now is known as a vector network analyzer
(VNA). These are truly amazing instruments that automate a large amount of the work
involved in measuring S-parameters. A basic block diagram of how these instruments work
is included in Fig. 2.29.
One of the key components in the VNA are its directional couplers. These are the
components that are directly connected to the device under test in Fig. 2.29. The special
features of these devices are that they allow us to separately sample the forward and backward
traveling waves that go through the directional coupler, which is obviously a vital step in
measuring the S-parameters of a device.
Before moving on, it should be noted that VNAs are finely tuned instruments and that
measuring the S-parameters accurately requires significantly more care than a standard cir-
cuit measurement, even when using something as powerful as a VNA. We won’t cover the
practical aspects of measurements in this course, but we will mention that it is vital to reg-
ularly calibrate the VNA (each time you do a new set of measurements, essentially) so that
it can utilize its sophisticated error models to provide highly accurate results.

2.8.4 Scattering Parameters of Common Components

We will now briefly discuss a number of common passive components (i.e., they do not provide
gain) that are frequently used in the design of microwave systems. To keep our conversation
brief, we will only be able to report the ideal S-parameters for these components rather than
actually deriving them. The derivation of many of these S-parameters requires the use of
analytical techniques that are outside of the scope of this course, but would often be covered
in a course dedicated to the study of microwave engineering. We will also briefly discuss

153
CHAPTER 2. TRANSMISSION LINES

Figure 2.29: Simplified block diagram of a vector network analyzer (image from [13]).

a few of the uses of the different components that we will introduce. We will organize our
discussion in terms of the number of ports that the devices have.

2-Port Devices
We will begin our discussion with 2-port devices. As a result, these are all described by 2x2
scattering matrices.
Attenuators – We have already discussed the basic concept of an attenuator as an ex-
ample of deriving the S-parameters for a simple circuit. Depending on the circumstances, a
wide range of different attenuation values may be needed in the design and testing of various
microwave systems. As a result, it is possible to buy many different attenuators that have
fairly standard values (e.g., 3 dB, 6 dB, 10 dB, 20 dB, etc.) – these are known as fixed
attenuators. More complicated devices also exist that can provide a variable amount of at-
tenuation based off of external settings – these are known as variable attenuators. These can
sometimes be digitally controlled, but in older technology they often had to be mechanically
controlled (e.g., rotating a part of the component).
The ideal scattering parameters for an attenuator are
 
0 |α|
[S] = , (2.128)
|α| 0

where |α| < 1 and sets the desired value of attenuation. In reality, the S11 and S22 values
will not be perfectly matched (i.e., will not equal 0). However, because the attenuator is

154
 CHAPTER 2. TRANSMISSION LINES

already providing loss to the system, it is often possible for them to be able to achieve very
low values of |S11 | and |S22 | (e.g., less than −30 dB).
Phase Shifters – Another very useful component that is typically a 2-port device (but
doesn’t necessarily have to be) is a phase shifter. The goal of a phase shifter is to apply a
well-characterized shift in the phase of the signal as it propagates through the device. This
can be a very important feature in a number of different microwave systems. An excellent
example that we will discuss briefly toward the end of the course is an antenna array. The
overall radiation pattern of an antenna array can be controlled by modifying the amplitude
and phase of the signals reaching each antenna element within an array. Phase shifters
can be a versatile technology to help control the phase part of this engineering problem.
Furthermore, it is common to use variable phase shifters that can have the amount of phase
shift applied to the signal entering the device changed in situ. Digital phase shifters that
provide a relatively constant phase shift over a band of frequencies with 6 bits of resolution
over the full 360◦ range are widely available. However, these components are typically rather
lossy (e.g., about 1 to 1.5 dB of loss per bit of resolution), so they have to be used carefully
in system design.
The ideal scattering parameters for a phase shifter are
 
0 |α|ejφ
[S] = , (2.129)
|α|ejφ 0

where φ sets the desired value of phase shift and |α| describes the (typically unavoidable) loss
that occurs in the device. In reality, the S11 and S22 values will not be perfectly matched,
and it is much harder to achieve a good match with a wideband constant-phase phase shifter
(i.e., it provides the same phase shift over a band of frequencies). As a result, typical values
of |S11 | or |S22 | may be closer to −10 to −15 dB.
Isolators – Isolators are a kind of non-reciprocal 2-port device that only allows propa-
gation through the device in one direction. This is often useful in protecting certain sensitive
components from reflections that may be produced in other nearby components – e.g., pre-
venting the output of an oscillator or amplifier from perturbing the bias point of these devices.
Traditionally, isolators were built using magnetic materials to provide the non-reciprocity.
More recent designs have found other clever engineering approaches to avoid the typically
high loss and heavy weight associated with using magnetic materials (e.g., using time-varying
or nonlinear systems).
The ideal S-parameters of an isolator that can only allow transmission from port 1 to
port 2 would be
 
0 0
[S] = . (2.130)
1 0

In reality, the isolator may not be perfectly matched, will result in some loss so that |S21 | < 1,
and will not be able to have a perfect amount of isolation (so that |S12 | ̸= 0). Depending
on the type of isolator being used, the isolation may be somewhere in the 20 to 30 dB range
(i.e., |S12 | will be around −20 to −30 dB).
Filters – Filters are often 2-port devices that are designed to provide a particular
frequency response in S21 (and likewise S12 since most filters will be reciprocal) while also

155
CHAPTER 2. TRANSMISSION LINES

being matched well. However, because the frequency response will be seen as a function of
frequency, looking at the scattering matrix at a single frequency isn’t particularly helpful.
However, ideally, the loss in the filter will be small so that in the passband of the filter the
|S21 | ≈ 1, while in the stopband the |S21 | ≈ 0.

3-Port Devices
We will now turn our discussion to 3-port devices, which are all described by 3x3 scattering
matrices.
Power Dividers/Combiners – The first component we discuss will be power dividers.
These devices are designed to take a single input and divide it into output signals with equal
phase but amplitudes set by the power division ratio of the device. For instance, one of the
most common power dividers would be an equal one that splits the input signal into two
equal amplitude outputs. For some applications, such as phased array antennas, it can be
desirable to utilize an uneven amplitude division to help achieve a desired array weighting
function. Since power dividers are reciprocal devices, they can also be used equally well as
power combiners, where they are now viewed as having multiple inputs that are combined
(according to the amplitude ratio of the device) to a single output. It is also possible to
design power dividers/combiners that have more than 3-ports, however, in their most basic
form they will be 3-port devices.
As an example, the ideal S-parameters of an equal division power divider would be
 √ √ 
0√ 1/ 2 1/ 2
[S] = −j 1/√2 0 0 . (2.131)
1/ 2 0 0
This kind of power divider is known as a Wilkinson power divider. A detailed analysis can
show that it is impossible for a power divider to be matched, reciprocal, and lossless at the
same time. The Wilkinson power divider is matched and reciprocal, and thus incorporates
some loss in the system (typically through a carefully located resistance). This loss is also
used in maintaining isolation between the two “output ports” (ports 2 and 3). This is why
|S23 | = |S32 | = 0 in the given S-parameters.
Circulators – We discussed the basic idea of a circulator when we discussed the basic
functionality of a Doppler radar. We saw that a circulator was a kind of non-reciprocal
device that could be useful in separating signals propagating in opposite directions; e.g.,
the signal traveling to/from an antenna to separate the transmit and receive signals. Like
isolators, circulators were traditionally designed using magnetic materials that were heavy,
bulky, and potentially more lossy than desirable. As a result, designing compact and efficient
circulators is still and important component design task (e.g., this is still a pain point in
designing quantum computers that operate at microwave frequencies).
The ideal S-parameters of a circulator would be
 
0 0 1
[S] = 1 0 0 . (2.132)
0 1 0
It can also be possible to design circulators with more ports. These generalize the above
S-parameters in a simple fashion.

156
 CHAPTER 2. TRANSMISSION LINES

4-Port Devices

We will now turn our discussion to 4-port devices, which are all described by 4x4 scattering
matrices.
Directional Couplers – The first 4-port device we will discuss is a directional coupler.
We discussed these in the contexts of Doppler radar and vector network analyzers, where
in each instance the directional coupler was used to route a small portion of a signal into
another direction to be used elsewhere in the device (typically to be compared against another
signal). The output port that has the main signal is referred to as the through port, while
the port with the smaller signal is referred to as the coupled port. The amount of power that
is routed to the coupled port is given in terms of the coupling factor. The fourth port of a
directional coupler is known as the isolated port, and ideally should have no power exiting
it. In reality, it is arbitrary which port we label as the “input port”, and so any port can
become the through port, the coupled port, or the isolated port depending on how the device
is operated.
The ideal S-parameters of a directional coupler would be
 
0 τ κ 0
τ 0 0 κ
[S] = 
κ
, (2.133)
0 0 τ
0 κ τ 0

where τ is the transmission coefficient (to the through port) and κ is the coupling coefficient
(to the coupled port). These are, in general, complex numbers with magnitudes less than 1.
In reality, the isolation will not be perfect, so additional terms in the S-matrix will not be 0.
Hybrid Couplers – A hybrid coupler is a special kind of directional coupler that equally
divides the power between the through and coupled ports. It also provides a 90◦ relative
phase shift between the signals exiting the through and coupled ports. As a result, this kind
of coupler can be very useful in connecting to two ports of a dual linearly polarized antenna
to generate circular polarization. It is also useful in I-Q mixers that require decomposing
signals into their “real” and “imaginary” components.
The ideal S-parameters of a hybrid coupler would be
 
0 1 j 0
1 1 0 0 j
[S] = √  . (2.134)
2 j
 0 0 1
0 j 1 0

Just like with the regular directional couplers, the isolation will not be perfect so that
additional terms in the S-matrix will be non-zero. Also, a completely lossless system is
not possible, so there will be some amount of insertion loss lowering the amplitudes of the
signals exiting the through and coupled ports. However, these losses can be made small and
be balanced so that the desired equal amplitude relationship is still achieved for the signals
exiting the through and coupled ports.

157
CHAPTER 2. TRANSMISSION LINES

2.9 Introduction to the Smith Chart

The Smith chart was created by P. H. Smith in the 1940’s at Bell Labs. It is one of the
most useful graphical tools for building intuition about many transmission line problems,
especially impedance matching problems. Most VNAs and commercial software that is used
to analyze microwave engineering designs have a built-in capability for plotting data on a
Smith chart.
The Smith chart can often be a daunting tool to learn how to use, so we will introduce
it in stages as opposed to jumping directly into looking at the entire chart at once. We will
simply mention that there is a reason one of the most popularly used Smith charts available
online has the subtitle “Black Magic Design” on it!
Now, the key to understanding the Smith chart is to first recognize that it is based on
a polar plot of the reflection coefficient on the complex plane. For a passive device (which
is what the standard Smith chart is applicable to), we know that |Γ| ≤ 1. As a result, the
Smith chart is completely contained in the unit circle of the complex plane. A particular
reflection coefficient is then plotted as Γ = |Γ|ejθ , where θ is measured from the positive real
axis in a counterclockwise direction. If we have a matched system, then |Γ| = 0, and we will
be at the center of the Smith chart. The goal of impedance matching problems is to get to
the “center of the Smith chart”, i.e., make |Γ| = 0.
The key realization that Smith had was that for transmission line problems we can
make a one-to-one correspondence between a particular reflection coefficient and normalized
impedances or normalized admittances. To see this, we recall that

ZL − Z0
Γ= . (2.135)
ZL + Z0
To use the Smith chart, we will work with normalized quantities, where the normalization
is often taken to be the characteristic impedance of the transmission line. We will write
normalized quantities as lowercase letters, which is a common convention. Hence, we have
that
zL − 1
Γ= , (2.136)
zL + 1

where zL = ZL /Z0 . The Smith chart provides us with a very quick way to change back
and forth between these representations to quickly solve problems without needing to use
any complex formulas. This is why it is so valuable for gaining intuition about solving
transmission line problems.
To see how this is done, we first solve for zL in (2.136) to get that

1+Γ
zL = . (2.137)
1−Γ
We can write this in terms of real and imaginary parts as

(1 + Γr ) + jΓi
rL + jxL = . (2.138)
(1 − Γr ) − jΓi

158
 CHAPTER 2. TRANSMISSION LINES

We can separate this into real and imaginary parts to get

1 − Γ2r − Γ2i
rL = (2.139)
(1 − Γr )2 + Γ2i

2Γi
xL = . (2.140)
(1 − Γr )2 + Γ2i

Both of these equations can be rearranged to get equations for two families of circles in the
(Γr , Γi )-plane. Doing this gives us,
 2  2
rL 1
Γr − + Γ2i = (2.141)
1 + rL 1 + rL

 2  2
2 1 1
(Γr − 1) + Γi − = . (2.142)
xL xL

Let’s start by considering the family of rL -circles given by (2.141). The first thing to
recognize is that all of these circles have their centers on the real axis. Next, we see that
as rL increases, the circles will get smaller and will have their centers move progressively
toward the (1, i0) point on the complex plane. This point corresponds to a real impedance
of ∞, which we can recognize as being equivalent to an open circuit. At the opposite side of
the Smith Chart (i.e., the leftmost side), we have the (−1, i0) point, which corresponds to a
short circuit. Finally, we can note that all of the rL -circles pass through the “open circuit”
point at the rightmost side of the Smith chart. A plot with a few of these circles drawn on
it is shown in Fig. 2.30.
Next, we consider the family of xL -circles given by (2.142). These circles have centers on
the Γr = 1 line at the rightmost edge of the Smith chart. All circles with xL > 0 have points
above the Γr axis and those with xL < 0 have points below the Γr axis. Hence, inductive
reactances are always above the real line and capacitive reactances are always below the real
line. Similar to the rL -circles, as xL increases the circles become smaller and move closer to
the real line, eventually merging with the open circuit point on the Smith chart. Further,
we can note that all of the xL -circles pass through the “open circuit” point as well. Some of
these circles are also shown in Fig. 2.30.
Not drawn on most Smith charts are constant |Γ| circles. These are circles that are
centered on the origin of the complex plane. Every point on one of these circles has a radius
that is equal to the magnitude of the reflection coefficient, i.e., |Γ|. Further, whatever angle a
particular normalized impedance point makes with the positive real axis is also equal to the
angle of the reflection coefficient. Although these circles don’t come standard on the Smith
chart, we often draw them ourselves as part of solving transmission line problems. To do
this accurately by hand requires the use of a drawing compass. These circles are very useful
in determining the input impedance of a particular loaded transmission line some length
forward or backward along a transmission line.

159
CHAPTER 2. TRANSMISSION LINES

Figure 2.30: Development of the Smith chart. A few constant resistance circles are shown
in solid lines and a few constant reactance circles are shown in dashed lines. Point labeled
by P is zL = 1.7 + j0.6, which corresponds to a Γ = 0.33 ∠ 28.07◦ (image from [14]).

To see why this is the case, we need to recall that the input impedance of a loaded
transmission line at location −ℓ from the load is

V (−ℓ) ejβℓ + ΓL e−jβℓ 1 + ΓL e−2jβℓ

Zin = = Z0 jβℓ = Z0 . (2.143)
I(−ℓ) e − ΓL e−jβℓ 1 − ΓL e−2jβℓ

For a normalized input impedance, we will have

1 + ΓL e−2jβℓ
zin = . (2.144)
1 − ΓL e−2jβℓ

We can consider the term ΓL e−2jβℓ to simply be a new reflection coefficient referenced to a
different position along the line (in particular, the position is −ℓ). If we write ΓL = |ΓL | ∠θΓL ,
we can recognize that our new reflection coefficient Γ(−ℓ) will have the same magnitude as
ΓL . Since only the phase of Γ(−ℓ) changes, we can find the corresponding zin on the Smith
chart that would be associated with a reflection coefficient Γ(−ℓ) by simply rotating in a
clockwise direction along a constant |Γ| circle.
To facilitate this operation, the Smith chart has a set of “scales” measured in electrical
length along the perimeter of the Smith chart (by electrical length, we mean a physical
distance normalized by the operating wavelength). You determine the point to rotate to by
first drawing a radial line from the origin and through the point zL out to the perimeter of
the Smith chart. You read the point on the scale you are at, move along the scale marked
“wavelengths toward generator” an appropriate amount, and then draw a new radial line

160
 CHAPTER 2. TRANSMISSION LINES

through the point on the perimeter you are at to the origin of the Smith chart. Where this
line intersects the constant |Γ| circle you had been on originally marks your new zin !
At this point, we can now appreciate a full Smith chart, which is shown in Fig. 2.31.
There are many other tricks that can be performed with the Smith chart to rapidly determine
other quantities of interest, like the standing wave ratio or return loss of a loaded line. We
won’t go into detail on these methods in this course, however, you can learn about some of
them from your textbook and other online references if you are interested. Instead, we will
be more interested in having an intuitive understanding of the Smith chart and how it can
be “traversed” in the course of practical design problems like impedance matching.

2.10 Single-Stub Impedance Matching

To get an idea of how the Smith chart can be used in practical design scenarios, we will look
at an example of what is known as a single-stub matching circuit. The goal of this circuit
is to take a generic load impedance ZL and match it to a lossless transmission line so that
looking into the single-stub matching circuit there are no reflections (i.e., Zin = Z0 ). The
configuration of a single shunt-stub matching circuit is shown in Fig. 2.32.
Due to the stub being in parallel to the transmission line with the load impedance, it is
easiest to think of the operation of this matching circuit in terms of admittances. Considering
this, the basic idea of the circuit is to transform the admittance YL = 1/ZL back along a
length of transmission line until it has an admittance with its real part equal to Y0 = 1/Z0
and its susceptance equal to some arbitrary number B. A “stub” is then connected in
parallel to this part of the circuit, and consists of either an open- or short-circuited length of
transmission line. We learned previously that by choosing an appropriate length of line, this
stub will be able to achieve an input impedance that is equal to any possible susceptance.
Hence, we simply need to choose a susceptance that will cancel out the susceptance B of
the transformed load admittance due to the parallel combination of the transformed load
admittance and the stub. Since admittances add in parallel, we can determine that the
needed susceptance of the stub will simply be −B.
It is certainly possible to determine analytical expressions for what ℓ and d (see Fig. 2.32
for definitions) need to be to achieve the desired match. However, this is fairly unintuitive and
rather tedious. Solving these problems on a Smith chart is quite easy, and can provide better
intuition as to how certain circuit components can be used to transform a load impedance
around the Smith chart to a new desired point.
As mentioned earlier, actually solving this kind of problem is most easily done in terms
of admittances (since the parallel combination of the stub and loaded line becomes a simple
addition). One can follow some simple steps to use the Smith chart (which is written in terms
of normalized impedances) to be used in terms of normalized admittances instead. We won’t
go through the details here, but to determine the corresponding normalized admittance
(denoted as yL ) from a particular normalized impedance one only needs to draw a constant
|Γ| circle through the zL point. The yL point is then the point on the constant |Γ| circle that
is diametrically opposite to the zL point.

161
CHAPTER 2. TRANSMISSION LINES

Figure 2.31: Full Smith chart that can be used in solving transmission line problems.

162
 CHAPTER 2. TRANSMISSION LINES

Figure 2.32: Circuit schematic of a single-stub matching circuit. The “stub” is the portion
coming off of the main transmission line sections in parallel to ZL and the section of line
with length d.

2.10.1 Example
Let’s now work out an example. The filled in Smith chart is shown in Fig. 2.33. We will list
out the steps in order here. The problem that we are tasked with is matching a normalized
load impedance of zL = 0.55 − j0.3.

1. Determine d:

(a) Mark the normalized load impedance on the Smith chart. This is the point labeled
zL in Fig. 2.33.
(b) Draw the constant |Γ| circle that intersects zL . This is the smaller blue circle
shown in Fig. 2.33.
(c) We want to work with admittances, since this will make the parallel combination
of the loaded line and stub easier to calculate. To find the normalized load
admittance, we “reflect” our zL point through the origin to the opposite side of
the constant |Γ| circle. The result is marked with the point yL in Fig. 2.33.
(d) Later on we will need to know how long d should be. To find this we will need to
know where yL lies on the “wavelength toward generator” scale on the perimeter
of the Smith chart. We have extended the line through the origin, zL , and yL far
enough so that we can read the desired number off the scale as 0.188λ.
(e) We now need to determine the points on our constant |Γ| circle that will intersect
the “normalized conductance equals 1” circle. This is the larger blue circle in Fig.
2.33. We see that our constant |Γ| circle intersects this new circle at two points,
marked as S1 and S2 . We can read these points off of the Smith chart to see that
they are S1 = 1 − j0.75 and S2 = 1 + j0.75.

163
CHAPTER 2. TRANSMISSION LINES

(f) We can now “rotate” our normalized admittance around the constant |Γ| circle
by extending the length of line that connects it to the point where we add in
the stub (i.e., by changing d). This will rotate us in a clockwise direction, so
we see that we are going to reach the point S1 far sooner than we can reach S2 .
Correspondingly, the length of line to get to S1 will be smaller, which is usually
the more practical design from an engineering perspective. We have extended a
line through the origin and S1 out to the edge of the Smith chart to read off what
this point is on the “wavelength toward generator” scale. The result we find is
0.347λ.
(g) We can determine d as the difference between the two spots we marked on the
“wavelength toward generator” scale. That is, d = 0.347λ − 0.188λ = 0.159λ.

2. Determine ℓ

(a) We can determine ℓ on the Smith chart as well, however, this leads to a rather
crowded figure so we have opted to do this using formulas instead. Since we
are only dealing with a single open- or short-circuited stub, the formulas are
quite simple. What we need to do is find the length of line that gives us a
normalized susceptance that cancels the susceptance at the point S1 . Recalling
that S1 = 1 − j0.75, we see that we need a susceptance of b = +j0.75.
(b) Now, we need to choose whether to use a short- or open-circuited stub. For
many transmission line geometries, it may only be possible to do a good job
implementing either a short circuit or an open circuit. However, some planar
transmission lines can allow either short or open circuits given the right conditions.
All things equal, it is usually preferable to choose whichever solution will lead to
a smaller stub length. For this problem, we can see that the j0.75 circle will
intersect the |Γ| = 1 circle closer to the leftmost side of the Smith chart, so we
will want to use an open circuit stub. To see this, remember that the open circuit
impedance point is the rightmost spot on the Smith chart, but as an admittance
this gets reflected through the origin which makes the leftmost spot the open
circuit admittance point.
(c) We can now use an impedance transformation formulas we encountered earlier to
compute ℓ. For an open-circuited line we had that the normalized input impedance
would be

zin = −j cot(βℓ). (2.145)

In terms of admittances, this will become

yin = j tan(βℓ). (2.146)

We can solve for ℓ to get yin = j0.75. This will tell us that ℓ = 0.1024λ.
Note: we could also have read this distance off of the Smith chart quite easily.
Try and do this yourself to test your understanding.

164
 CHAPTER 2. TRANSMISSION LINES

(d) We have graphically marked what this shunt stub does in Fig. 2.33 in green.
In particular, as we add a compensating susceptance we rotate in a clockwise
direction along the “normalized conductance equals 1” circle toward the origin.
The reason we rotate along this circle is because we are modifying the susceptance
of our system without changing the conductance. As a result, we stay on the same
constant conductance circle as we had started on.

Figure 2.33: Smith chart solution of a single-stub matching circuit design.

165
CHAPTER 2. TRANSMISSION LINES

2.11 Practice Problems

1. A general TEM wave propagating in the +z-direction will have Ez = Hz = 0 by
definition. Considering this, find the relations between Ex (x, y), Ey (x, y), Hx (x, y), and
Hy (x, y) by using the source-free Maxwell’s equations and noting that Ez = Hz = 0.
You will further want to note that all field components have an exponential phase
factor of exp[−jβz].
2. Beginning from the scalar Helmholtz wave equation, show that a general TEM wave
propagating in the z-direction will have all four scalar field components satisfy the
two-dimensional Laplace’s equation.
3. Assume that the fields within a coaxial cable are given by
V0 ρ̂
E= e−γz , (2.147)
ρ ln(b/a)

I0 ϕ̂ −γz
H= e , (2.148)
2πρ
where a is the radius of the inner conductor and b is the radius to the beginning of
the outer conductor. Inside the region a < ρ < b, the transmission line is filled with
a material characterized by ϵ = ϵ′ − jϵ′′ and µ = µ′ . The conductors in this problem
can also be characterized with a surface resistivity of Rs . Considering this, answer the
following questions:
(a) Find the per unit length inductance,
(b) Find the per unit length capacitance,
(c) Find the per unit length resistance,
(d) Find the per unit length conductance,
(e) Find the characteristic impedance of this line assuming that the losses can be
neglected.
4. For a purely reactive load impedance and a purely real characteristic impedance, show
that the magnitude of the reflection coefficient is always one.
5. A lossless transmission line with characteristic impedance Z0 = 50 Ω and length
ℓ = 0.7λ is terminated with a load impedance ZL = 45 + j30 Ω. Find the follow-
ing quantities.
(a) The reflection coefficient at the load.
(b) The standing wave ratio on the line.
(c) The input impedance of the system.
(d) If the end of this transmission line (i.e., the part that is 0.7λ away from the
load) were connected to another transmission line with characteristic impedance
of 100 Ω, what would be the resulting reflection coefficient at the connection point?

166
 CHAPTER 2. TRANSMISSION LINES

6. Beginning with the formula for the propagation constant in a lossy transmission line,
show that if
R G
= (2.149)
L C
the group velocity on the transmission line will equal the phase velocity.
Point of Interest: This means that there will be no dispersion on the line. This kind
of line is referred to as a distortionless line, and often requires special engineering of
the transmission line to achieve this behavior if it is needed.

7. For the circuit shown in Fig. 2.34, calculate the time-average power propagating for-
ward in the finite-length transmission line (Pi ), the time-average power propagating
backward in the finite-length transmission line (Pr ), and the time-average power prop-
agating in the forward direction in the infinitely-long transmission line (Pt ). Note that
Z0 is the characteristic impedance of the half-wavelength section of transmission line
and Z1 is the characteristic impedance of the infinitely-long transmission line.

Figure 2.34

Hint: think about a simple way to represent an infinitely-long section of transmission

line from the perspective of the rest of the circuit that the infinitely-long line is attached
to.

8. Derive the admittance parameters for the π-network shown in Fig. 2.35.

Figure 2.35: Example of a two-port π-network.

167
CHAPTER 2. TRANSMISSION LINES

Note: Pay attention to the direction convention for the currents when evaluating the
admittance parameters. It is easy to accidentally be off by a minus sign if you aren’t
careful. You will not receive full credit for the problem if you do an incorrect procedure
but get the correct answer due to errors canceling out.

9. Determine the Y11 and Y22 elements of the admittance matrix for the transmission line
circuit shown in Fig. 2.36.

Figure 2.36

10. Assuming that the characteristic impedance for every port of an arbitrary n-port
network is equal to 1, show that the scattering matrix can be computed from the
impedance matrix as

[S] = ([Z] + [I])−1 ([Z] − [I]) (2.150)

where [I] is the n-dimensional identity matrix.

11. Use the circuit shown in Fig. 2.37 to answer the following questions. For this problem,
all of the transmission lines have a characteristic impedance of 50 Ω and β = 200π.

Figure 2.37

(a) What is the input impedance of the circuit at the location marked by Zin ?

168
 CHAPTER 2. TRANSMISSION LINES

Note: This input impedance must account for the combination of both of the
transmission line segments shown (i.e., the one that is d long and the one that is
ℓ long).
(b) From (a), you have found that this circuit is not matched to the 50 Ω transmission
line it is connected to. You have been tasked with changing the length of one of
the sections of transmission line (either the 1.8388 mm line or the 1.655 mm line,
but not both) so that this circuit functions correctly as a single stub matching
circuit. Determine which line length should be adjusted and find the correct
length it should be.
Hint: Review the lecture material on how a single stub matching circuit works.
Determine a test you can make to see if the line of length d is accomplishing
its purpose for a properly functioning single stub matching circuit. If it is not
working than d must be fixed. If it is working then ℓ must be fixed.

169
CHAPTER 2. TRANSMISSION LINES

170
Chapter 3

Waveguides and Resonators

3.1 TEM, TE, and TM Waveguide Modes

In the last portion of the course, we discussed how a one-dimensional transmission line
theory could be used to solve many practical problems in microwave engineering. We also
saw that in order to use this simplified theory we would need to be able to derive certain
characteristic values about our transmission line, such as its propagation constant and char-
acteristic impedance. At the time, we simply assumed that we already knew these values
so that we could focus on learning the fundamentals of transmission line theory. We now
revisit this problem by performing a more complete field analysis of common transmission
line systems. From this field analysis, we will be able to calculate the properties needed
to apply transmission line theory for an “arbitrary” geometry. Of course, we will only be
able to perform a full analytical analysis of fairly simple geometries. Numerical methods or
judicious approximations are typically necessary to handle more complex transmission line
geometries.
To begin this process, we will consider different kinds of solutions to Maxwell’s equations
in a general waveguide geometry. In particular, we will consider the properties of three
different kinds of solutions; namely, transverse electromagnetic (TEM), transverse electric
(TE), and transverse magnetic (TM) modes. In this section, we will focus on deriving
the general equations that these different kinds of waves will need to obey for a particular
geometry. We will then consider solving these equations for a rectangular waveguide later.

3.1.1 General Equations

To determine the general equations that will be useful for analyzing waveguide geometries,
we will assume that we have a geometry that is oriented along the z-axis, is infinitely long,
and has a constant cross sectional shape over the entirety of the waveguide. Under this
assumption, we can assume that we will have a simple propagating wave characteristic for
the z-dependence of the electric and magnetic fields contained in the waveguide. Due to
the orientation of the geometry, it makes since to break our electric and magnetic fields into
their transverse (i.e., in the cross section of the waveguide) and longitudinal (i.e., along the

171
CHAPTER 3. WAVEGUIDES AND RESONATORS

length of the waveguide) components. Considering this, we can write our fields as
E(x, y, z) = [Et (x, y) + ẑEz (x, y)]e−jβz , (3.1)

H(x, y, z) = [Ht (x, y) + ẑHz (x, y)]e−jβz , (3.2)

where Et and Ht contain the transverse components (i.e., x- and y-components) of the
electric and magnetic fields, respectively.
For a general waveguide analysis, we will be considering the electromagnetic fields that
exist in the source-free region contained inside of a particular waveguide geometry. Hence, we
can use the source-free form of Maxwell’s equations to derive the Helmholtz wave equation
for E and H following our standard process. This gives us for the electric field
∇2 E + k 2 E = 0, (3.3)
√
where k = ω µϵ is, as usual, the wavenumber. If we write
∇2 = ∇2t + ∂z2 , (3.4)
we can simplify our wave equation given the known z-dependence of our field given in (3.1).
In particular, we will get that
∇2t E + (k 2 − β 2 )E = 0. (3.5)
This looks similar to some of the expressions we saw when discussing the separation of
variables in the context of arbitrarily-directed plane waves. This is no accident, and it is
advantageous to define a new kind of wavenumber suggested by (3.5) as
kc2 = k 2 − β 2 . (3.6)
We refer to kc as the cutoff wavenumber. The reason for this terminology can be seen by
rearranging this into the form of a dispersion relation as
β 2 = k 2 − kc2 . (3.7)
We see that if k 2 < kc2 , our propagation constant will become imaginary. Recalling the as-
sumed z-dependence of our fields in (3.1), we see that this will correspond to an exponentially
attenuating set of fields as a function of z. Hence, we will not have wave propagation unless
our frequency is such that our wavenumber is above the cutoff wavenumber. For the kinds
of geometries we have been considering in class up to this point, we haven’t typically had to
worry about this notion because it turned out that the cutoff wavenumber was 0. However,
we will consider shortly cases where the cutoff wavenumber is greater than 0, which causes
these particular waveguide geometries to act as a kind of high pass filter (i.e., only signals
above a certain frequency will be able to be transmitted through them).
Now, to see a useful strategy for actually solving practical waveguide problems it can be
more advantageous to return to the source-free form of Maxwell’s equations. If we evaluate
Faraday’s law in terms of Cartesian coordinates for (3.1) and (3.2), we can get
∂y Ez + jβEy = −jωµHx , (3.8)
−jβEx − ∂x Ez = −jωµHy , (3.9)
∂x Ey − ∂y Ex = −jωµHz . (3.10)

172
 CHAPTER 3. WAVEGUIDES AND RESONATORS

A similar set of equations can be derived from Ampere’s law. It turns out that all of these
equations can be rearranged to express the transverse components of E and H in terms of
the longitudinal components of E and H. In particular, we get that

j
Hx = (ωϵ∂y Ez − β∂x Hz ), (3.11)
kc2

j
Hy = − (ωϵ∂x Ez + β∂y Hz ), (3.12)
kc2

j
Ex = − (β∂x Ez + ωµ∂y Hz ), (3.13)
kc2

j
Ey = (−β∂y Ez + ωµ∂x Hz ). (3.14)
kc2

The important result is that if we can solve for Ez and Hz we can calculate the rest of the
components without returning to solving the wave equation from scratch. This turns out
to be very useful for waveguide problems. It also suggests a useful way to decompose the
different wave solutions that may exist into different “families” based on their characteristics
in terms of Ez and Hz . In particular, we will consider TEM modes characterized by Ez =
Hz = 0, TE modes characterized by Ez = 0, Hz ̸= 0, and TM modes characterized by
Ez ̸= 0, Hz = 0. We will now consider some general properties about each of these different
kinds of waves before explicitly finding these kinds of solutions for a rectangular waveguide.

3.1.2 TEM Waves, Ez = 0, Hz = 0

The first thing to note about these waves is that if we try to use (3.11) to (3.14) we won’t get
very far because everything will look like a trivial solution. However, we can actually notice
from these equations an important detail about kc for TEM waves. In particular, there were
no undue assumptions in the derivation of (3.11) to (3.14), so these can be considered to be
valid equations. As a result, we see that the only way that all of our transverse components
of E and H won’t be trivial will be if (3.11) to (3.14) produce some kind of indeterminate
result. This will only be possible if kc = 0, which is a property of TEM waves (i.e., they
have no cutoff frequency).
Although this is a useful conclusion, it doesn’t help us to determine the spatial structure
of the transverse field components. To do this, we will need to return to the Helmholtz wave
equation and simplify it for E = Et e−jβz (since Ez = 0). You actually have already done this
as a homework problem, where you saw that both Et and Ht will need to satisfy Laplace’s
equation in the transverse cross section of the waveguide geometry for a TEM wave. That
is,

∇2t Et = 0, (3.15)

173
CHAPTER 3. WAVEGUIDES AND RESONATORS

∇2t Ht = 0. (3.16)

We won’t consider in detail solving these equations for waveguide geometries in this class.
The important concept to note is that because these equations are the equations that govern
static fields, we can conclude that the spatial structure of the transverse fields of a TEM
wave are the same as the static fields that can exist between the conductors. Hence, the
concepts you learned in your earlier electromagnetics courses can be valuable for intuitively
picturing what the fields will look like for a particular TEM wave.
A few other points of interest for TEM waves are the following.
1. Due to the static field behavior, TEM waves can only exist if we have multiple con-
ductors in our geometry (like in a coaxial cable or a parallel plate waveguide).
√
2. Since kc = 0, we have that β = k = ω µϵ. This tells us that the phase velocity of
our TEM waves will be constant. As a result, the only dispersion that will exist for
our TEM waves is due to the dispersion inherent in the material filling the waveguide
itself.

3. The wave impedance is equal to the intrinsic impedance of the materials filling the
transmission line, that is
Ex Ey
ZTEM = =− = η. (3.17)
Hy Hx

4. The transverse fields can be related to each other as

1
Ht = ẑ × Et . (3.18)
ZTEM

3.1.3 TE Waves, Ez = 0, Hz ̸= 0
Let’s now consider TE waves. We will see that the process for solving for TE waves is
somewhat different from solving for TEM waves. However, the approach for TE waves will
also apply to TM waves (and other more complicated kinds of waves) so that it can be viewed
as a more general process, while solving for TEM waves is somewhat of a special case.
Now, from (3.11) to (3.14) with Ez = 0 we can see that if we are able to calculate Hz
we can find all the remaining transverse field components from this single scalar component.
Hence, it will serve us well to find the equation for Hz . Following an identical process to what
gave us the wave equation only involving transverse derivatives in (3.5) for the magnetic field
will give us

∇2t H + kc2 H = 0, (3.19)

where the cutoff wavenumber is still given by kc2 = k 2 − β 2 . In Cartesian coordinates, it is

quite easy for us to simplify this to only consider the Hz components. This will give us the
equation

∇2t Hz + kc2 Hz = (∂x2 + ∂y2 )Hz + kc2 Hz = 0. (3.20)

174
 CHAPTER 3. WAVEGUIDES AND RESONATORS

We must solve this equation subject to the boundary conditions of a particular waveguide
geometry. For most geometries we can handle by hand, we will be able to use the separation
of variables approach to solve this equation.
Despite not being able to solve this equation yet, we can still conclude some useful
properties of TE waves. These are summarized briefly here.
1. TE waves can exist in single conductor systems, such as a hollow metal “pipe”.
2. kc > 0 for all TE waves, so the waveguide geometries that support TE waves can often
be looked at as being like a high pass filter
3. The wave impedance for a TE wave is
Ex Ey kη
ZTE = =− = . (3.21)
Hy Hx β
p
Since β = k 2 − kc2 we see that ZTE will be a function of frequency.
4. The transverse fields can be related to each other as
1
Ht = ẑ × Et . (3.22)
ZTE
5. The guided wavelength for a TE wave is given by λg = 2π/β. Hence, we have that
2π
λg = p . (3.23)
k 2 − kc2
From this, we can see that the guided wavelength will in general not be equal to the
“free-space” or “homogeneous medium” wavelength of λ = c/f . This is important to
remember when dealing with non-TEM waves.

3.1.4 TM Waves, Ez ̸= 0, Hz = 0
For TM waves, we will be able to use a very similar process to what we did for TE waves.
However, we now have from (3.11) to (3.14) with Hz = 0 that we will need to calculate Ez
to quickly find the remaining transverse field components from this single scalar component.
Hence, it will serve us well to find the equation for Ez . Here, we will have
∇2t E + kc2 E = 0, (3.24)
where the cutoff wavenumber is still given by kc2 = k 2 − β 2 . In Cartesian coordinates, it is
quite easy for us to simplify this to only consider the Ez components. This will give us the
equation
∇2t Ez + kc2 Ez = (∂x2 + ∂y2 )Ez + kc2 Ez = 0. (3.25)
We must solve this equation subject to the boundary conditions of a particular waveguide
geometry. For most geometries we can handle by hand, we will be able to use the separation
of variables approach to solve this equation.
Despite not being able to solve this equation yet, we can still conclude some useful
properties of TE waves. These are summarized briefly here.

175
CHAPTER 3. WAVEGUIDES AND RESONATORS

1. TM waves can exist in single conductor systems, such as a hollow metal “pipe”.
2. kc > 0 for all TM waves, so the waveguide geometries that support TM waves can
often be looked at as being like a high pass filter
3. The wave impedance for a TM wave is
Ex Ey βη
ZTM = =− = . (3.26)
Hy Hx k
p
Since β = k 2 − kc2 we see that ZTM will be a function of frequency. We can also see
that ZTM ̸= ZTE ̸= ZTEM , in general.
4. The transverse fields can be related to each other as
1
Ht = ẑ × Et . (3.27)
ZTM

5. The guided wavelength for a TM wave is given by λg = 2π/β. Hence, we have that
2π
λg = p . (3.28)
k 2 − kc2
From this, we can see that the guided wavelength will in general not be equal to the
“free-space” or “homogeneous medium” wavelength of λ = c/f . This is important to
remember when dealing with non-TEM waves.

3.2 Rectangular Waveguide Modes

We will now consider how to solve for the waves that can exist for the rectangular waveguide
geometry shown in Fig. 3.1. From this figure, we see that this waveguide is a single conductor
system, and as a result will not be able to support TEM waves. In our solution process,
we will find that there are actually many different TE and TM waves that can exist in this
geometry. To differentiate between all of the different valid solutions, we will refer to each
of them as modes of the waveguide. Each mode will have its own unique set of propagation
characteristics (e.g., guided wavelength, propagation constant, cutoff wavenumber, wave
impedance), which can make the analysis of systems involving these kinds of waveguide
geometries much more complicated than simple plane wave situations we considered earlier
in the course. Although things are more complex here, we often build practical systems that
limit the number of waveguide modes that need to be simultaneously considered so that the
conceptual analysis of these systems is still intuitive and possible in practice.

3.2.1 TE Modes
Recall that for TE modes, we have that Ez = 0, Hz ̸= 0. Hence, we need to solve the
equation
(∂x2 + ∂y2 )Hz + kc2 Hz = 0, (3.29)

176
 CHAPTER 3. WAVEGUIDES AND RESONATORS

Figure 3.1: Geometry of a rectangular waveguide made from a perfect electric conductor.

under the assumption that

Hz (x, y, z) = Hz (x, y)e−jβz . (3.30)

Considering the structure of this partial differential equation, we see that if we follow a
separation of variables approach we will be able to simplify the solution of this equation. In
particular, we will assume that

Hz (x, y, z) = X(x)Y (y)e−jβz . (3.31)

Substituting this into (3.29) will, after some rearranging, give us

1 d2 1 d2
2
X+ 2
Y + kc2 = 0. (3.32)
X dx Y dy
By the standard separation of variables argument, we can conclude that each term on the
LHS of this equation will need to individually equal a constant for this equation to be
satisfied. Hence, we can define separation constants kx and ky to get that

d2
2
X + kx2 X = 0, (3.33)
dx

d2
Y + ky2 Y = 0. (3.34)
dy 2
We will also have the modified dispersion relation from (3.32) that

kc2 = kx2 + ky2 . (3.35)

We can now follow our standard three step process for solving Maxwell’s equations. We
have already completed Step 1 in Fig. 3.1. To be more explicit, we can identify Region 1 as
being inside the waveguide that is filled with a material characterized by µ and ϵ. The walls

177
CHAPTER 3. WAVEGUIDES AND RESONATORS

of the waveguide are made from a perfect electric conductor (PEC). We have walls at x = 0
and x = a that range continuously from 0 ≤ y ≤ b. Similarly, we have another two walls at
y = 0 and y = b which range continuously over the range 0 ≤ x ≤ a. Region 2 would be the
rest of space, which can be assumed to be fully filled with PEC material.
Now, for Step 2, we need to propose a general solution in each region of our problem. We
already know that there will be no fields in Region 2 because it is filled with PEC material,
so we can ignore this region for the rest of the problem. In Region 1, we can look at the
two separated 1D wave equations in (3.33) and (3.34) and see that our standard set of wave
solutions will be applicable here for X(x) and Y (y). However, because we are working with
a finite-sized region it will now be more advantageous to use trigonometric functions as
our wave solutions as opposed to exponential functions (this makes enforcing the boundary
conditions easier algebraically). Hence, we will have as our general wave solution that



Hz (x, y) = X(x)Y (y) = Xc cos(kx x) + Xs sin(kx x) Yc cos(ky y) + Ys sin(ky y) . (3.36)

We can now begin Step 3 to solve for the unknown constants based on the boundary
conditions of our problem. For a PEC boundary, we would typically use the fact that the
tangential component of the electric field will need to vanish as our boundary condition.
Since we have our proposed solution in terms of Hz , one solution approach would be to use
Hz to compute Ex and Ey and then apply the boundary condition to these field components.
This is facilitated by using (3.13) and (3.14) and by noting that for a TE wave Ez = 0. This
then gives us that
j
Ex = − ωµ∂y Hz , (3.37)
kc2

j
Ey = ωµ∂x Hz . (3.38)
kc2

Alternatively, we can use (3.37) and (3.38) to recast the boundary condition on the tan-
gential components of E into an equivalent boundary condition on H. For a PEC boundary,
this will tell us that the normal derivative of Hz at the PEC boundary should be 0 (i.e., at
a surface with n̂ = x̂, we would take ∂x Hz = 0 as our boundary condition). Hence, we can
determine that the four boundary conditions we should enforce to solve our current problem
are

∂x Hz = 0 (i.e., Ey = 0) at x = 0, (3.39)

∂x Hz = 0 (i.e., Ey = 0) at x = a, (3.40)

∂y Hz = 0 (i.e., Ex = 0) at y = 0, (3.41)

∂y Hz = 0 (i.e., Ex = 0) at y = b. (3.42)

178
 CHAPTER 3. WAVEGUIDES AND RESONATORS

We can now plug (3.36) into these boundary conditions. Beginning with (3.39), we see
that we will get
 
− Xc kx sin(kx 0) + Xs kx cos(kx 0) Y (y) = Xs kx Y (y) = 0. (3.43)

This tells us that Xs = 0, since we can’t in general have kx = 0 or Y (y) = 0 as a completely

general solution. Similarly, we can use (3.40) to get that

−Xc kx sin(kx a)Y (y) = 0 (3.44)

after taking into account that Xs = 0. We can’t simply say that Xc = 0 since this would
completely remove the X(x) dependence in our separation of variables expression. Hence,
we will need to have that

sin(kx a) = 0, (3.45)

from which we can determine that

mπ
kx = , m ∈ Z. (3.46)
a

This tells us that there are an infinite number of values of m for which this equation will be
able to be satisfied. This is our first “taste” of why there will be many different possible TE
modes that can exist in a particular waveguide geometry.
Continuing, we move on to enforcing our boundary conditions in (3.41) and (3.42). Start-
ing with (3.41), we get that
 
− Yc ky sin(ky 0) + Ys ky cos(ky 0) X(x) = Ys ky X(x) = 0. (3.47)

By similar logic to what we saw earlier, we can conclude that Ys = 0. Now, we can use
(3.42) to get that

sin(ky b)X(x) = 0, (3.48)

which tells us that

nπ
ky = , n ∈ Z. (3.49)
b

This tells us we can also have an infinite number of values of n for which our equation may
be able to be solved.
Putting these results together, we get that
   
mπ nπ
Hz (x, y) = H0 cos x cos y , m, n ∈ Z. (3.50)
a b

179
CHAPTER 3. WAVEGUIDES AND RESONATORS

We can compute the remaining field components of Hx , Hy , Ex , and Ey using (3.11) to

(3.14) and noting that Ez = 0. As a result, by finding (3.50) we have completed our solution
of this problem. Before moving on, it is important to consider the integers m and n more
closely. For simplicity of notation, we have written in (3.50) that these may be any integer.
However, the values of m and n do loosely depend on each other. In particular, we need
to make sure that the combination of m and n don’t cause Hz to either be equal to 0 or a
constant. We can’t have these cases because they either invalidate our assumption for this
kind of wave that Hz ̸= 0 or represent a trivial case that is not of interest. Considering this,
we see that the only constraint on m and n with respect to each other is that they cannot
be simultaneously equal to 0. However, they can be 0 individually and still give us a valid
solution.
We can now look at the values of some important propagation characteristics for these
TE modes. We will begin with the propagation constant, which becomes
s  2  2
p q mπ nπ
β = k 2 − kc2 = k 2 − (kx2 + ky2 ) = k 2 − − . (3.51)
a b

Considering this, we can ask ourselves the question of what will be the first TE mode that
can propagate in our particular waveguide (i.e., which mode will have the lowest kc ). To
help organize ourselves, we will keep track of modes by labeling them as TEmn . Since a > b,
we will find that the cutoff frequency for the TE10 mode will be the first mode that can
propagate in our waveguide. We will typically refer to the first mode that can propagate in
a waveguide as being the dominant mode. After analyzing the TM mode case next, we will
find that the TE10 mode is in fact the dominant mode of a rectangular waveguide (i.e., all
TM modes have higher cutoff frequencies).

3.2.2 TM Modes
Recall that for TM modes, we have that Hz = 0, Ez ̸= 0. Hence, we need to solve the
equation

(∂x2 + ∂y2 )Ez + kc2 Ez = 0, (3.52)

under the assumption that

Ez (x, y, z) = Ez (x, y)e−jβz . (3.53)

Once we have solved for Ez , we can use this to compute all other non-zero field components
in the waveguide by recalling that we had previously derived from Maxwell’s equations that
j
Hx = (ωϵ∂y Ez − β∂x Hz ), (3.54)
kc2

j
Hy = − (ωϵ∂x Ez + β∂y Hz ), (3.55)
kc2

180
 CHAPTER 3. WAVEGUIDES AND RESONATORS

j
Ex = − (β∂x Ez + ωµ∂y Hz ), (3.56)
kc2

j
Ey = (−β∂y Ez + ωµ∂x Hz ). (3.57)
kc2

Further, by considering the structure of (3.52), we see that if we follow a separation of

variables approach similar to what we did for TE modes we will be able to simplify the
solution of this equation in the same way. This will give us the same modified dispersion
relation

kc2 = kx2 + ky2 , (3.58)

and the different separation equations will also be the same (i.e., 1-D wave equations).
We can now follow our standard three step process for solving Maxwell’s equations. We
have the same geometry as for TE modes so Step 1 doesn’t change. Similarly, we have the
same separated equations (although, the boundary conditions do change), so Step 2 doesn’t
need to really change either. Hence, we will have as our general wave solution that



Ez (x, y) = X(x)Y (y) = Xc cos(kx x) + Xs sin(kx x) Yc cos(ky y) + Ys sin(ky y) . (3.59)

We can now begin Step 3 to solve for the unknown constants based on the boundary
conditions of our problem. For a PEC boundary, we can use the fact that the tangential
component of the electric field will need to vanish as our boundary condition. Hence, we
have that

Ez = 0 at x = 0 and x = a, (3.60)

Ez = 0 at y = 0 and y = b. (3.61)

We can now plug (3.59) into these boundary conditions. Beginning with (3.60) at x = 0, we
see that we will get
 
Xc cos(kx 0) + Xs sin(kx 0) Y (y) = 0 =⇒ Xc = 0. (3.62)

Similarly, we can use (3.60) at x = a to get that

Xs sin(kx a)Y (y) = 0 (3.63)

after taking into account that Xc = 0. We can’t simply say that Xs = 0 since this would
completely remove the X(x) dependence in our separation of variables expression. Hence,
we will need to have that

sin(kx a) = 0, (3.64)

181
CHAPTER 3. WAVEGUIDES AND RESONATORS

from which we can determine that

mπ
kx = , m ∈ Z. (3.65)
a

This tells us that there are an infinite number of values of m for which this equation will be
able to be satisfied, just as we had for TE modes.
Continuing, we move on to enforcing our boundary conditions in (3.61). Starting with
(3.61) at y = 0, we get that
 
Yc cos(ky 0) + Ys sin(ky 0) X(x) = 0 =⇒ Yc = 0. (3.66)

Now, we can use (3.61) at y = b to get that

Ys sin(ky b)X(x) = 0, (3.67)

which tells us that

nπ
ky = , n ∈ Z. (3.68)
b

This tells us we can also have an infinite number of values of n for which our equation may
be able to be solved.
Putting these results together, we get that
   
mπ nπ
Ez (x, y) = E0 sin x sin y , m, n ∈ (Z − ∅). (3.69)
a b

All of the transverse field components can be easily computed from this expression for Ez
using (3.54) to (3.57) and noting that Hz = 0. These transverse fields can be found in your
book in (10-134) to (10-137) [14]. Now, we have labeled that m and n can be any integers
that aren’t zero (this is what we mean by m, n ∈ (Z − ∅)). We can’t allow m or n to equal
0 because this would cause the entire Ez = 0, which violates the assumption of a TM mode.
Hence, the first TM mode that can propagate in a rectangular waveguide is the TM11 mode.
We can now look at the values of some important propagation characteristics for these
TM modes. We will begin with the propagation constant, which becomes
s  2  2
p q mπ nπ
β = k − kc = k 2 − (kx2 + ky2 ) =
2 2 k2 − − , (3.70)
a b

which matches the result we had for TE modes. However, because our first TM mode is the
TM11 , we see that the dominant mode of the overall waveguide will be the TE10 mode since
that will have the lowest cutoff frequency.

182
 CHAPTER 3. WAVEGUIDES AND RESONATORS

Figure 3.2: Examples of the transverse electric field for low-order modes in a rectangular
waveguide (images from [15]).

3.2.3 Examples of Waveguide Modes

To help better understand the field structure of rectangular waveguide modes, it is useful
to plot the fields in various cross sections of the waveguide. The transverse electric field
for a number of low-order modes are shown in Fig. 3.2. Other cross sections can be found
online and in your book. The figures of interest in your book are Figs. 10-11 to 10-13.
In general, each mode number represents the number of half-sinusoidal variations that the
field solution will take over a particular dimension of the waveguide. For example, the TE10
mode means that there will be a single half-sinusoidal variation along the x-dimension of
the waveguide and no variation along the y-dimension. Similarly, TE21 would have a full
sinusoidal variation (i.e., one period) along the x-dimension of the waveguide and a single
half-sinusoidal variation along the y-dimension.

3.2.4 Practical Waveguide Operation

Before moving on, it is useful to comment on a few remaining details with regard to how
rectangular waveguides are used in practice. The first point of note is that there are many
cases that can occur when two different modes will have the same cutoff frequency (e.g., the
TE11 and TM11 modes as a trivial example). These are referred to as degenerate modes.
Dealing with these can sometimes be inconvenient, so it is typical to design waveguides
that minimize the number of degenerate modes. One common way to do this is to make
practical waveguides have cross-sectional dimensions of a ≈ 2b, but avoiding the case of
a = 2b. However, for certain applications, it is actually desirable to have two degenerate
orthogonal polarized modes (for instance, feeding certain circularly polarized antennas). In
this case, one might use a square waveguide that has many degenerate modes. However, this
is typically the exception, rather than the rule.

183
CHAPTER 3. WAVEGUIDES AND RESONATORS

Figure 3.3: Dispersion diagram for a WR-90 rectangular waveguide illustrating the typical
operating bandwidth.

Another important point about the use of waveguides is to define their typical operating
band. This is illustrated in Fig. 3.3. We see that the beginning of the operating band is
generally taken to be about 1.25 times the cutoff frequency of the waveguide. This is done to
avoid the highly nonlinear portion of the dispersion curve near the cutoff frequency because
this region has significant dispersion that can lead to distortion of practical signals that
encompass a band of frequencies. The upper limit for the operating band is usually set to be
slightly lower than the cutoff frequency of the second propagating mode in the waveguide.
This is done to ensure that only a single mode can propagate in the waveguide, which makes
the analysis and operation of a waveguide much smoother in practical situations.
As an example of a standard waveguide, we can consider the WR-90 waveguide that is
designed to correspond to the X-band of frequencies (8.2 – 12.4 GHz). The “WR” stands for
“waveguide rectangular” and the “90” denotes that the “a” dimension of the waveguide is
0.9” wide. It is a straightforward exercise to compute that the cutoff frequency of the TE10
mode is approximately 6.557 GHz for this waveguide. We can see that 8.2/6.557 = 1.25, as
mentioned earlier.

3.3 Introduction to Resonators

Microwave resonators are a valuable tool in many practical applications. For instance, they
can form an essential part of many kinds of filters, oscillators, and amplifiers. At many
microwave frequencies, it is not achievable to form resonators using lumped element compo-
nents. As a result, alternative approaches using distributed resonators (e.g., formed by some
configuration of transmission line structure) are often needed.

184
 CHAPTER 3. WAVEGUIDES AND RESONATORS

(a) (b)

Figure 3.4: Examples of lumped element resonant circuits. (a) Series case and (b) parallel
case.

3.3.1 Lumped Element Resonators: A Short Review

Before discussing distributed resonators, it will be useful to recall some basic properties
of lumped element resonators. The simplest examples of resonant circuits are series and
parallel RLC networks. These each support a single resonance frequency, which occurs when
the reactance or susceptance of the inductor and capacitor completely cancel each other
out. If we consider the series resonant circuit in Fig. 3.4(a), we can see that our resonance
condition will require

1
jω0 L + = 0. (3.71)
jω0 C

We can solve this to find that our resonance frequency will be given by

1
ω0 = √ . (3.72)
LC

If we repeat these steps for the parallel resonant circuit we arrive at an identical result. For
both cases, we typically define the bandwidth of a resonant circuit to√occur at the “half-
power” point, or when the input impedance has changed by a factor of 2 from its value at
resonance.
Another important quantity that is used to characterize resonators is its Q (also called
quality factor ). This is defined as

average energy stored Wm + We

Q=ω =ω . (3.73)
energy loss / second Ploss

This helps us understand how long a resonator will continue to resonate after the excitation
has been removed. In general, the total Q of a resonator will depend on what other circuitry
is connected to it (e.g., a source or a load). When we only consider the “intrinsic” Q of a
resonator, we refer to it as the unloaded Q. If we take into account the loading due to other

185
CHAPTER 3. WAVEGUIDES AND RESONATORS

circuits, we refer to the new Q as the loaded Q. For the series resonance case we have that
the unloaded Q, denoted as Q0 , is
r
1 L ω0 L 1
Q0 = = = . (3.74)
R C R ω0 RC
Similarly, the unloaded Q of the parallel resonator is
r
C R
Q0 = R = = ω0 RC. (3.75)
L ω0 L
More generally, the total Q of a system can be found as
1 1 X 1
= + (3.76)
Q Q0 i
Qi

where the Qi characterize the quality factors of additional components, such as extra load
components. For simple resonators, we can relate the total Q to the bandwidth of the
resonator as BW = 1/Q.
Although these various definitions are usually introduced in the context of analyzing
simple lumped element circuits, we use the same terminology to analyze more complicated
microwave resonators. Of course, the evaluation of the different quantities involved in the
definitions will become more complex for this case. However, much of the intuition about
simpler circuits can be useful in understanding the more complicated cases of microwave
resonators. Depending on how the input impedance of a microwave resonator varies as a
function of frequency, certain distributed resonators will also be referred to as supporting
“series” or “parallel” resonances in analogy to the lumped element case. We won’t use this
terminology here, but you may encounter it in your future career.

3.3.2 Transmission Line Resonators

To achieve a resonance with a transmission line, we typically must terminate both ends of the
transmission line with a particular boundary condition (if part of the transmission line were
to extend to infinity we would just end up with the cases we have considered previously that
weren’t resonators). We can typically think of our resonant solutions as being a standing
wave solution that “fits” into our particular geometry perfectly. To see this, let’s consider
a simple example of a lossless transmission line of length ℓ that is terminated with a short
circuit at both ends (shown in Fig. 3.5).
Now, we want to determine the solutions to the wave equation that we got from the
telegrapher’s equations; i.e.,

d2
V (z) + β 2 V (z) = 0, (3.77)
dz 2
√
where we had that β = ω LC. Since our geometry has a finite length, we know that it
will often be quickest to determine a valid solution if we write our general solution in terms
of sine and cosine as opposed to exponential functions (which are traveling wave solutions).

186
 CHAPTER 3. WAVEGUIDES AND RESONATORS

Figure 3.5: Example of a short-circuited transmission line resonator. The boundary condi-
tions on the line will only be able to be satisfied for particular values of ℓ considering the
fixed transmission line parameters Z0 and β.

Considering this, our proposed general solution for the voltage on the transmission line can
be
V (z) = Vs sin(βz) + Vc cos(βz). (3.78)
We now need to enforce our boundary conditions at z = 0 and z = ℓ. Since we have a
short circuit at both of these locations, we will need V (z) to vanish at these positions on the
line. From this, we can readily find out that Vc = 0 and that
mπ
β= , m ∈ Z. (3.79)
ℓ
√
However, as mentioned previously, we also know that β = ω LC = 2π/λg . Considering
this, we can rearrange our equation to see that we will only have a valid solution when
λg
ℓ=m , m ∈ Z. (3.80)
2
That is, our transmission line will only be able to support a valid solution when the length
of the line is exactly a multiple of half a guided wavelength. This exercise can be repeated
for a transmission line that is terminated in open-circuits at both ends as well, and the same
result will be achieved. The unique aspects of how these resonances differ can be tied back
to notions of “series” and “parallel” resonances, but we will not delve into this in this course.
Consolidating these results, we have that valid solutions of the voltage will be
Vm (z) = V0,m sin(βm z), (3.81)
with βm = mπ/ℓ. If we wish to find the current that goes with these voltage solutions, we
can use the telegrapher’s equations. In particular, we can recall that
∂z V = −jωLI (3.82)
to find that
V0,m
Im (z) = − βm cos(βm z). (3.83)
jωL

187
CHAPTER 3. WAVEGUIDES AND RESONATORS
√
Recalling that βm will still need to equal ω LC at the particular resonant frequency,
p we
can simplify this expression into terms of the characteristic impedance Z0 = L/C. In
particular, we get that

V0,m
Im (z) = − cos(βm z). (3.84)
jZ0

Now, it should be understood that finding the resonant solutions of a system in this way
has a few deficiencies to it. First, what we have seen is that the resonance only exists at a
discrete set of frequencies when the transmission line is a multiple of half a guided wavelength.
In reality, the loss of a line will cause the resonance to have a finite bandwidth. Further,
we haven’t accounted for any loading effects that may exist due to nearby circuitry. These
kinds of effects can usually be accounted for in different ways as perturbations to the results
we have just found. Overall, the idealization of a lossless system is often mathematically
convenient for quickly finding general ideas about resonators, such as how long they should
be to achieve resonance at a particular frequency and that they will have equally-spaced
resonance frequencies. However, a more careful consideration is needed to successfully design
a properly functioning resonator for a particular application, and is something that can be
learned about in more advanced microwave engineering courses.

3.3.3 Rectangular Waveguide Resonators

We won’t consider these in depth in lecture, but you will solve some problems related to these
in your homework. However, it should hopefully be clear to you that if we take a rectangular
waveguide of finite length and terminate both ends of it with a perfectly conducting wall
(to end up with a hollow metal box) we will find a similar result to what we just found for
a transmission line resonator. In this case, we can still support the same set of transverse
modes, but the finite size along the longitudinal direction now forces us to find standing
wave solutions along this axis as well. The end result is that we can only support a non-
trivial solution at discrete frequencies. We typically number the resulting modes as TEmnp
and TMmnp modes, where the index p is associated with the longitudinal dimension of the
waveguide. The first mode that exists will typically be the TE101 mode, since the b dimension
of the waveguide cavity is generally taken to be the smallest dimension by convention.

3.4 Practice Problems

1. A general guided wave solution can be written as

E = ET (x, y) + ẑEz (x, y) e−γz ,

 
(3.85)

H = HT (x, y) + ẑHz (x, y) e−γz ,

 
(3.86)

where a subscript T denotes the quantities involved are purely transverse to the direc-
tion of propagation. Hence, ET is a shorthand for the Ex and Ey components of the

188
 CHAPTER 3. WAVEGUIDES AND RESONATORS

electric field. Now, by substituting these expressions in Faraday and Ampere’s laws,
show that
1

ET = − 2
γ∇T Ez − ẑjωµ × ∇t Hz (3.87)
kc

1

HT = − 2
γ∇T Hz + ẑjωϵ × ∇t Ez . (3.88)
kc

To arrive at these expressions, you will need to solve for Ex in terms of Hz and Ez ,
and similar for the other transverse components Ey , Hx , and Hy . Then, combine the
equations for Ex and Ey into the vector expressions given above.

2. Calculate and list in ascending order the cutoff frequencies of a rectangular waveguide
with width a and height b for the following modes: TE01 , TE10 , TE11 , TE02 , TE20 ,
TM11 , TM12 , and TM22 . Do this for

(a) if a = 2b, and

(b) if a = b.

Compute the cutoff frequencies relative to that of the dominant mode of the waveguide.

3. An air-filled rectangular waveguide has a = 6.5 cm and b = 3.5 cm. If the operating
frequency is set to 3 GHz, calculate β, vp , λg , and the wave impedance of the dominant
mode.
For the following questions, assume that we have an air-filled rectangular waveguide
whose cross section is defined by 0 < x < 22.86 mm and 0 < y < 10.16 mm.

(a) Compute the time-average Poynting vector for the TM11 mode of this waveguide
at 16.5 GHz.
(b) Compare the wavelengths of the TE10 and TM11 modes to that of a plane wave
propagating in free space at 16.5 GHz.

4. A lossless transmission line of length 2.145 mm is terminated with a short circuit at

z = 0 mm and an open circuit at z = 2.145 mm. If the transmission line has per-
unit-length parameters of L = 0.7125 µH/m and C = 285 pF/m, answer the following
questions.

(a) Using the 3-step process to solving partial differential equations, determine the
expressions for the voltage and current solutions that can exist for this geometry.
(b) At what frequency will the first resonance occur?

5. Write a computer program to compute the input admittance at the position marked
in the circuit shown in Fig. 3.6 for frequencies ranging from 2 to 16 GHz. Bold lines
with lengths marked denote transmission line segments, while thin lines with no lengths
given are regular circuit connections. To perform the calculations, you may assume that

189
CHAPTER 3. WAVEGUIDES AND RESONATORS

β can be computed using formulas for a homogeneous dielectric medium with relative
permittivity ϵr = 6.325. Plot the real part of the input admittance to recreate Fig.
3.7. Further, provide a simple explanation for why the bandwidth of each resonance
increases as the frequency goes up.

Figure 3.6

Figure 3.7

6. Using the 3-step process to solving Maxwell’s equations and the separation of variables
approach, derive expressions for all relevant components of the electric and magnetic
fields for a general TMmnp mode in a rectangular cavity resonator. Assume all walls of
the resonator are made from perfect electric conductors and that the interior region of
the cavity is specified by 0 < x < a, 0 < y < b, and 0 < z < d. Given the field expres-
sions, what is the lowest order TMmnp mode that can be supported in this geometry?

Hint: Start with a general solution for Ez that is a generalization of the one we used for
solving for the TM modes of a waveguide. The application of the boundary conditions
at x = 0, x = a, y = 0, and y = b will be identical to what we did in the lecture
notes previously. At z = 0 and z = d, you will need to apply a boundary condition
on either Ex or Ey . So, you will need to determine a process to compute Ex and Ey

190
 CHAPTER 3. WAVEGUIDES AND RESONATORS

from Ez . Later, you will need to also compute Hx and Hy from Ez for this current
scenario. Think about how you can use the results from Problem 1 to quickly take
care of this rather than re-using Maxwell’s equations to relate Ez to all the transverse
field components (it has to do with “replacing” −γ in the expressions with something
else).

191
CHAPTER 3. WAVEGUIDES AND RESONATORS

192
Chapter 4

Antennas

4.1 Introduction to Antenna Concepts

Earlier in this class we spent a significant amount of time learning about the properties
of plane waves: how they propagate in an unbounded medium, what happens when they
encounter certain kinds of discontinuities, and in a crude sense one way that they could
be generated (i.e., from an infinite current sheet). We have also spent time learning about
transmission lines and how to engineer systems to try and maximize the amount of power to
a “load”, which we mentioned in passing could often be an antenna for many applications.
We are now at the point where we can begin to put these concepts together. We will see
how antennas are a practical method for generating plane waves, and will also consider how
we can represent an antenna from the perspective of a transmission line network problem.
Antenna engineering is a broad topic that continues to garner a significant amount of
research interest. Part of its longevity has come from the need to develop more and more
sophisticated kinds of antennas to support new wireless communication and other wireless
transmission applications. The many kinds of applications that exist have lead to a widely
divergent set of requirements for antennas to meet. As a result, many different kinds of
antennas have been invented over the year to provide unique sets of properties that can
be valuable for different applications. Despite the many differences, the basic purpose of
an antenna remains the same. That is, to take a “tightly bound” wave contained on some
kind of transmission line structure and find an efficient way to “separate” the fields from
the physical system so that they can begin to propagate away from the antenna. The basic
process of this for a small dipole antenna is shown in Fig. 4.1. In general, there are many
different kinds of structures that have been found that can efficiently produce radiated fields
with different properties. Some common types of antennas are shown in Figs. 4.2 to 4.6.

193
CHAPTER 4. ANTENNAS

Figure 4.1: Example of electric fields detaching from a small dipole antenna and beginning
to radiate away from the antenna (images from [16]).

Figure 4.2: Common examples of wire antennas (images from [16]).

194
 CHAPTER 4. ANTENNAS

Figure 4.3: Common examples of aperture antennas (images from [16]).

Figure 4.4: Common examples of patch antennas (images from [16]).

195
CHAPTER 4. ANTENNAS

Figure 4.5: Common examples of array antennas (images from [16]).

Figure 4.6: Common examples of reflector antennas (images from [16]).

196
 CHAPTER 4. ANTENNAS

4.2 Electromagnetic Potentials

In many cases, it can be difficult to attempt to directly solve for E, H, D, and B. In these
situations, it can be advantageous to define a set of electromagnetic potentials to help us go
about solving a problem. This is especially true in antenna problems. You should already be
familiar with this in solving electrostatic and magnetostatic problems. We will now consider
what this looks like in a full electromagnetic picture.
To begin, we start with Maxwell’s equations, which are

∇ × H = ∂t D + J, (4.1)

∇ × E = −∂t B, (4.2)

∇ · D = ρ, (4.3)

∇ · B = 0. (4.4)

We note that from Gauss’ law of magnetism (4.4) we can safely say that B can be given
by the curl of some vector field (since the divergence of a curl is always 0). Hence, we will
introduce a magnetic vector potential, denoted as A, that is related to B by

B = ∇ × A. (4.5)

We can move forward with this magnetic vector potential by using it to rewrite Faraday’s
law. This gives us

∇ × E = −∂t B
(4.6)
= −∂t (∇ × A).

We can consolidate terms on the left-hand side to get

∇ × E + ∂t A = 0 (4.7)

Recalling that the curl of a gradient is always equal to 0, we see that we can set E + ∂t A
equal to a gradient vector for this equation to always be satisfied. That is

E + ∂t A = −∇Φ, (4.8)

where Φ is the electric scalar potential. We can consolidate our potentials together to see
that

E = −∂t A − ∇Φ. (4.9)

Next, lets see what happens when we use these potentials in Ampere’s law. In a homo-
geneous medium, Ampere’s law is

∇ × µ−1 B = ∂t ϵE + J. (4.10)

197
CHAPTER 4. ANTENNAS

We can substitute in our potentials using (4.38) and (4.40) to get

∇ × µ−1 ∇ × A = −ϵ ∂t2 A + ∂t ∇Φ + J.


(4.11)

We can rearrange terms to get

∇ × ∇ × A + µϵ∂t2 A = −µϵ∂t ∇Φ + µJ. (4.12)

We see that the left-hand side looks like one of our familiar wave equations with a propagation
speed matching the speed of light.
What happens if we apply our vector identity to rewrite the ∇ × ∇× operator? This will
give us

∇2 A − µϵ∂t2 A = ∇ ∇ · A + µϵ∂t ∇Φ − µJ.

(4.13)

Have we accomplished anything? Not yet, but if you recall the Helmholtz decomposition
theorem, we realize that we have some freedom in specifying what ∇ · A should be equal to.
This freedom exists because we have introduced these auxiliary potential functions to help
us solve our problem, they don’t already come fully specified like our fields and fluxes did.
It is this freedom that we can exploit to help us simplify the solution of certain problems.
To move forward, we need to specify what we are going to force ∇ · A to equal. This is
called setting a gauge condition. Although electromagnetic theory was one of the first areas
where gauge conditions were widely used, it has become a very prevalent concept in many
areas of modern physics to describe different kinds of forces. For instance, gauge conditions
are a very prevalent and important part of the Standard Model of particle physics, which in
addition to electromagnetism includes theories for the weak and strong nuclear forces. These
additional forces obey a set of equations that can be viewed as a generalization of Maxwell’s
equations (albeit, this takes some fairly sophisticated mathematics to see).
For our current purposes, let’s use our gauge condition to try and simplify our wave
equation. We can do this by making the two terms involving A and Φ cancel on the right-
hand side of (4.13). That is, by setting

∇ · A = −µϵ∂t Φ. (4.14)

This gauge condition is used frequently enough that it has its own name. It is the Lorenz
gauge condition. Note that many people erroneously will call this the Lorentz gauge condition
(unfortunately, your textbook does this). This is incorrect and should not be done. Both
Lorenz and Lorentz made important contributions to electromagnetic theory, and they should
be appropriately commended for their contributions.
Now, after setting this gauge condition, our vector potential wave equation in the Lorenz
gauge becomes

∇2 A − µϵ∂t2 A = −µJ. (4.15)

What does the equation for Φ look like? We can find this out by substituting our electro-
magnetic potentials into Gauss’ law of electricity. This takes us from having

∇ · E = ρ/ϵ, (4.16)

198
 CHAPTER 4. ANTENNAS

to having


∇ · ∇Φ + ∂t A = −ρ/ϵ. (4.17)

We can use the Lorenz gauge condition to rewrite this as

∇2 Φ − µϵ∂t2 Φ = −ρ/ϵ. (4.18)

We see that we get a scalar Helmholtz equation for Φ. This is one of the advantages of the
Lorenz gauge. It allows us to arrive at two decoupled wave equations for our potentials that
immediately show that they both propagate at the speed of light. This is part of the reason
why this gauge condition is frequently utilized in the study of special relativity (where the
fact that electromagnetic waves propagate at the speed of light is of paramount importance).
Another gauge condition that is frequently used in certain areas of physics is the Coulomb
gauge. This is a gauge characterized by

∇ · A = 0. (4.19)

We can sometimes view this gauge as being an approximation to the Lorenz gauge where the
time variation is so small relative to the spatial variation that we can treat the ∂t Φ term as
being equal to 0. This kind of approximation is frequently used in studying the interaction
of an electromagnetic field with an atomic system in optics and quantum optics. It should
be noted, however, that it is possible to use this gauge without making any approximations.
Although we have freedom in how we specify our potentials, it is important to note that
regardless of our gauge condition A and Φ will always work together to produce the correct
fields and fluxes to solve a particular problem. Hence, we see that the potentials are in
some sense arbitrary. Due to this, it has long been said that the potentials are not physical
quantitites, but that they are simply a useful mathematical tool. However, the advent of
quantum theory has shown us that the electromagnetic potentials are very much physical
quantities that can impact the dynamics of other systems. The Aharonov-Bohm effect is an
excellent example of this. This measured effect shows that the trajectory of an electron is
affected by the potentials even when the electric and magnetic fields are identically zero in
the region the electron travels through.

4.3 Potentials Produced by Known Source Distribu-

tions
One of the fundamental tasks in antenna analysis and design is to determine the fields that
are produced from a particular “source configuration”, i.e., some set of materials and cur-
rents that exist on the antenna. We could in principle include the current distributions in
our derivation of a wave equation for E and H. However, this often leads to us needing to
compute the derivatives of the source distributions, which is often mathematically compli-
cated by the ways that we typically represent source distributions (e.g., as infinitely thin
sheets). In this situation, it can often be more efficient to first compute the electromagnetic

199
CHAPTER 4. ANTENNAS

potentials from the source distribution and then compute the fields from the potentials. Al-
though this may seem “roundabout”, it is in practice much easier than trying to directly
compute the fields.
We will now investigate how the electromagnetic potentials can be computed from a
known distribution of a source function, i.e., from J and ρ. To build up to this, we will first
recall some basic principles from electrostatics. We know that Coulomb’s law told us that
the electric field produced by a point charge was given by
q
E(r) = r̂ . (4.20)
4πϵ|r|2
Recognizing that E = −∇Φ, we can conclude that the potential would be
q
Φ(r) = . (4.21)
4πϵ|r|
To be more general in our description, it is best that we allow our point charge to be located
at a position other than 0. This generalization is possible by writing the electric field and
potential as
q(r′ ) r − r′
 
E(r) = , (4.22)
4πϵ|r − r′ |2 |r − r′ |

q(r′ )
Φ(r) = , (4.23)
4πϵ|r − r′ |

where r and r′ are position vectors pointing to the observation and source points, respectively.
If instead of having a single point charge we had a distribution, you will recall that our
formula for the potential generalizes to
˚
1
Φ(r) = ρ(r′ )/ϵ dV ′ , (4.24)
4π|r − r′ |

where the prime on dV ′ denotes that we are integrating with respect to the primed variables.
We can interpret this formula as being a superposition of the potentials produced by a
collection of point charges. A common terminology is to refer to the
1
(4.25)
4π|r − r′ |
portion of this integral as an integration kernel. The reason for this is because this is the
point source response that we are going to integrate against our distribution to determine the
potential at any other location. In the study of physical systems, this point source response
is often referred to as a Green’s function.
We can view our Green’s function as helping us to “invert” differential equations by
recalling that Poisson’s equation is

∇2 Φ = −ρ/ϵ. (4.26)

200
 CHAPTER 4. ANTENNAS

The “point source response” or Green’s function for this equation would be defined as

∇2 g(r, r′ ) = −δ(r − r′ ), (4.27)

where δ(r − r′ ) is the Dirac delta function. This PDE has as solution the Green’s function
1
g(r, r′ ) = . (4.28)
4π|r − r′ |

We can use this Green’s function to “invert” Poisson’s equation by multiplying it by g(r, r′ )
on both sides and integrating over the volume of the problem. The result is
˚ ˚
1 ′2 ′ ′ 1
′
∇ Φ(r )dV = − ′
ρ(r′ )/ϵ dV ′ . (4.29)
4π|r − r | 4π|r − r |

We can simplify our result by transferring the derivatives from Φ in (4.29) onto the Green’s
function using various kinds of integration by parts. You may recall that these forms of
“integration by parts” are typically referred to as Green’s identities from your other courses.
Once we have transferred the derivatives appropriately, we can replace the operations on
the Green’s function using (4.27) to simply be −δ(r − r′ ). At that point, evaluating the
remainder of the integration is trivial, and gives Coulomb’s law as
˚
1
Φ(r) = ρ(r′ )/ϵ dV ′ . (4.30)
4π|r − r′ |

Hence, we can see that in some sense, Coulomb’s law is just the inverse of Poisson’s equation.
This process will generally work for PDEs; however, it is often very difficult to determine
the Green’s function of a problem. Regardless, there are many situations in antenna theory
where a known and simple Green’s function can be used. We will consider this particular
Green’s function now.
Let’s now turn our attention back to the electromagnetic case. For this situation, we
want to solve the equation

∇2 Φ − µϵ∂t2 Φ = −ρ/ϵ. (4.31)

We can first transform to the frequency domain to simplify things. This gives us

∇2 Φ + k 2 Φ = −ρ/ϵ. (4.32)

We will want to “invert” this equation using a Green’s function that is the solution of

∇2 g(r, r′ ) + k 2 g(r, r′ ) = −δ(r − r′ ). (4.33)

Comparing this to (4.27), we see that the equations are very similar. Hence, we expect that
the solution may also be able to be a similar function. It turns out that the solution is
1 ′
g(r, r′ ) = ′
e−jk|r−r | . (4.34)
4π|r − r |

201
CHAPTER 4. ANTENNAS

This looks very much like our plane wave solution multiplied by our Green’s function for
static problems. Hence, we have a phase changing as we spherically move away from our
point source augmented by a reduction in magnitude. This particular Green’s function is
typically referred to as the free-space Green’s function (if ϵ = ϵ0 and µ = µ0 ), or, more
generally, as the homogeneous-medium Green’s function (if one of ϵ or µ does not equal the
corresponding free space value).
We can use it in a similar fashion to find that the potential produced by a charge distri-
bution is
˚
1 ′
Φ(r) = ′
e−jk|r−r | ρ(r′ )/ϵ dV ′ . (4.35)
4π|r − r |

Considering that our vector potential wave equation was

∇2 A − µϵ∂t2 A = −µJ, (4.36)

we see that the same Green’s function can be useful here as well. In particular, we have that
˚
1 ′
A(r) = ′
e−jk|r−r | µJ(r′ )dV ′ . (4.37)
4π|r − r |

Hence, if we know our distributions of J and ρ in a homogeneous medium it becomes a

somewhat “simple” exercise to compute the A and Φ that are produced (granted, this
may need the help of numerical integration routines if the distributions of the sources are
complicated).
Once we know the potentials we can compute the fields from them. For instance, we
know from the definition of the magnetic vector potential that

H = µ−1 ∇ × A. (4.38)

We can then use Ampere’s law to find the electric field at any point in space away from the
source distribution. Alternatively, we can use the potentials to compute the electric field
directly. In general, we have that

E = −jωA − ∇Φ. (4.39)

Often we only want to work with a single source distribution, so we will use our gauge
condition to write Φ in terms of A. For antenna problems, the Lorenz gauge is typically
used so that one will often see the electric field written in terms of A as
1
E = −jωA + ∇(∇ · A). (4.40)
jωµϵ

Next, we will be able to begin computing the fields produced by an infinitesimal dipole using
these concepts.

202
 CHAPTER 4. ANTENNAS

4.4 Fields of an Infinitesimal Dipole

Assume that we have an infinitesimal linear wire that is positioned symmetrically at the
origin of the coordinate system and is oriented along the z-axis. Due to the extremely small
dimension of the wire, we assume that we can represent the current along the wire as a linear
current density given by
(
ẑI0 , |z| ≤ ℓ/2,
J(r′ ) = (4.41)
0, elsewhere,

where I0 is the amplitude of the current and ℓ is the length of the infinitesimal current
density. In reality, it is not possible for the current density to have a perfectly uniform
density along the length of a short wire (it must go to zero at the endpoints). However,
this is a reasonable approximation for extremely small wires (compared to the wavelength of
radiation) and greatly simplifies the analysis, making it a useful starting point for analyzing
antennas. Overall, this kind of antenna is typically referred to as an infinitesimal dipole or
a Hertzian dipole.
Now, we need to evaluate the integration of this current density with the free-space
Green’s function to determine the magnetic vector potential produced by the current. This
will give us
ˆ ℓ/2 −jkR
e
A(x, y, z) = ẑµI0 dz ′ , (4.42)
−ℓ/2 4πR

where

R = |r − r′ | (4.43)

and a primed (un-primed) coordinate corresponds to a source (observation) location. In

general, this integral can be difficult to evaluate. To avoid the complications, it is assumed
here that because ℓ ≪ λ that the functions involving R don’t change appreciably over the
integration region. This allows us to then get
ˆ
µI0 −jkr ℓ/2 ′ µI0 ℓ −jkr
A(x, y, z) = ẑ e dz = ẑ e , (4.44)
4πr −ℓ/2 4πr

where r = |r|.
For antenna problems, it is often advantageous to express the radiated fields in a spherical
coordinate system. We can use standard coordinate system conversions to find that
µI0 ℓ −jkr
Ar = Az cos θ = e cos θ, (4.45)
4πr

µI0 ℓ −jkr
Aθ = −Az sin θ = − e sin θ, (4.46)
4πr

Aϕ = 0 × Az = 0. (4.47)

203
CHAPTER 4. ANTENNAS

The electric and magnetic fields may now be computed from the magnetic vector potential
using

1
E = −jωA + ∇(∇ · A), (4.48)
jωµϵ

H = µ−1 ∇ × A. (4.49)

This can be done rather easily, however, it is essential to remember that the forms of the
various vector differential operators are coordinate system dependent. As a result, we can’t
use the simple identities that we have used throughout the course for Cartesian coordinates.
Evaluating the various derivatives gives us

Hr = Hθ = 0 (4.50)

 
kI0 ℓ sin θ 1
Hϕ = j 1+ e−jkr (4.51)
4πr jkr

for the magnetic field, and

 
I0 ℓ cos θ 1
Er = η 2
1+ e−jkr (4.52)
2πr jkr

 
kI0 ℓ sin θ 1 1
Eθ = jη 1+ − 2
e−jkr (4.53)
4πr jkr (kr)

Eϕ = 0 (4.54)

for the electric field. Being able to evaluate the fields like this by hand for all regions in
space (away from the source region) is an extremely uncommon occurrence for an antenna
problem. Typically, we are only able to determine the fields that exist far away from the
antenna, which we will comment on more later.
Although we didn’t dwell much on evaluating the derivatives, it should hopefully now
make more sense why using the magnetic vector potential as part of our process is advanta-
geous. If we were to try and directly calculate the electric or magnetic fields, we would need
to compute the derivative of the current density given in (4.41). Although this is possible us-
ing somewhat more sophisticated forms of mathematics than is often taught in introductory
calculus classes, it does complicate the procedure. By first integrating the current density
against the free-space Green’s function, we end up with a function that is much easier to
compute derivatives of, leading to the fields given in (4.50) to (4.54).

204
 CHAPTER 4. ANTENNAS

Figure 4.7: Illustration of different field regions around an antenna. Only in the far-field
does the pattern stabilize as a function of distance away from the antenna.

4.4.1 Field Regions

Due to the complexity of the expressions for the fields produced by an antenna, it is typically
useful to define a few different regions where various terms of the overall field expressions will
“dominate” the other ones. This helps build intuition related to the operation of antennas,
and hints at some important properties of analyzing antenna systems. The boundaries
between regions are somewhat arbitrary and can need to be revisited for certain applications.
However, they are often all given in terms of the wavelength and a dimension D, which is
taken to be the diameter of a sphere that can completely contain the antenna being analyzed.
This dimension and the various field regions are illustrated in Fig. 4.7. We will now discuss
the various regions and their characteristics in more detail.
Now, the first region is often called the near-field, or sometimes the reactive near-field.
This is the region that is immediately surrounding the antenna, where terms with higher
inverse powers of r will dominate the overall field expressions. It can be shown that these
terms do not contribute to a real outward power flow (which would be radiation), but instead
contribute to imaginary power that is stored near the antenna. Due to the energy being
stored in this way, we can think of these fields as being “reactive”, much like a reactive load
influences a circuit or transmission line problem. Due to the dependence on higher-order
powers, the fields in this region of space typically rapidlypchange as a function of position.
The typical boundary for this region is given as R < 0.62 D3 /λ.
The next region is often referred to as the radiating near-field or Fresnel region. This is
an intermediate region where the fields are transitioning from being predominantly reactive
to being predominantly radiative. The overall shape of the fields as a function of radial
distance away from the antenna
p continue to change in this region. The typical boundary for
this region is given as 0.62 D /λ < R < 2D2 /λ.
3

The final region that is most often considered in antenna problems is the far-field or

205
CHAPTER 4. ANTENNAS

Fraunhofer region. In this region, the fields are purely radiative and propagate radially away
from the antenna. Further, the fields are “fully formed” in the sense that the distribution
of radiation over the θ and ϕ coordinates no longer changes as a function of r. This is an
essential property that greatly simplifies the analysis of antennas. Typically, we only care
about the fields far away from the antenna. We can evaluate these fields directly by making
certain approximations in the evaluation of the radiation integrals of the form shown in
(4.42). Further, in this region we are often far enough away from the antenna that we can
consider the fields to be plane waves locally. We can then use the tools we learned earlier in
the course for analyzing plane wave problems. The typical boundary for this region is given
as R > 2D2 /λ. You should memorize this distance. When we simplify fields to be in the
far-field, we drop all terms that have an inverse power of R larger than 1. That is, terms
that have dependence like R−2 or R−3 are assumed to be 0 in the far-field.
For the infinitesimal dipole, the far-field radiated fields would be
kI0 ℓ sin θ −jkr
Hϕ = j e (4.55)
4πr

kI0 ℓ sin θ −jkr

Eθ = jη e . (4.56)
4πr
Note that if we were to calculate the Poytning vector for these fields, we would end up with a
single component pointed in the radial direction, which matches our intuition for how fields
should radiate away from a source. We also can note that there are no radial components to
the field. This is a general property that always happens for fields in the far-field region of
an antenna. This also matches our intuition about antennas producing plane waves (which
are a special type of TEM wave) far away from the antenna.

4.5 Fundamental Parameters of Antennas

Now that we have a basic understanding of how to determine the fields of an antenna, we
can begin to understand some of the fundamental parameters that are used to describe the
performance of antennas. The first parameter we will discuss is the radiation pattern of an
antenna. This is a mathematical function that describes the spatial distribution of radiation
produced by an antenna, often as a function of the spatial coordinates θ and ϕ. Many
different kinds of radiation patterns can be plotted for a single antenna to characterize a
variety of properties. However, most often the radiation pattern is plotted in the far-field
zone of the antenna so that the spatial distribution will be independent of the distance from
the antenna (as long as it is far away from the antenna). Examples of radiation patterns
would be field and power patterns, which plot a particular polarization of the radiated wave
in either field or power units (e.g., V/m or W/m2 , respectively). Examples of a typical
pattern for a directional antenna is shown in Fig. 4.8.
We typically refer to each region of the pattern between zeros as being a particular lobe
of the radiation pattern. The primary lobe of the antenna where most power is radiated is
referred to as the mainlobe. All other lobes will typically be referred to as sidelobes. For
many applications, it is desirable to have a narrow mainlobe and as small of sidelobes as

206
 CHAPTER 4. ANTENNAS

Figure 4.8: Example polar plots of field and power patterns of a directive antenna. The
half-power beamwidth (HPBW) is also marked on both plots.

possible. This helps ensure the power predominantly goes toward the intended location and
a minimal amount “spills” into undesirable directions. As an example, many modern radar
applications require antennas with sidelobes that are approximately 30 to 40 dB smaller
(i.e., 3 to 4 orders of magnitude smaller) than the peak magnitude of the mainlobe. This
is typically a stringent requirement, which necessitates clever and precise engineering to be
achievable.
Another common characteristic of an antenna is the half-power beamwidth. This is defined
on Fig. 4.8, and is the angular separation between the two points on the mainlobe that are
half the value of the mainlobe peak power. This beamwidth is one common parameter that
is used to helps us understand how narrow the mainlobe is.
Often, we are interested in characterizing the radiated power associated with a particular
antenna. To get to the parameters of most interest for characterizing antennas, we will need
to go through a cascade of different power definitions. Our starting point is the time-average
Poynting vector, which from earlier in the course we know is defined as
1
Savg = Re{E × H∗ }. (4.57)
2
In the far-field, we only are concerned with the radiated power density which we sometimes
denote as Wrad = Savg . Since Wrad is assumed to be in the far-field, we can safely assume
that its vector direction is purely in the radial direction, i.e., r̂.
We can determine the total power radiated from the antenna by integrating the outward
directed radiated power density over a large sphere that encloses the entire antenna. This
will give us
‹
Prad = Wrad · r̂dS. (4.58)
S

207
CHAPTER 4. ANTENNAS

It is typically useful for us to have some convenient reference antenna to compare all other
antennas to in specifying certain radiation properties. The usual reference antenna is an
isotropic radiator, which is an ideal source that radiates equally in all directions. For this
type of antenna, we can determine a useful correspondence between the radiation density
and total radiated power of an antenna. We do this by noting that for W0 , the radiation
density of an isotropic source, we will have the total radiated power given by
‹ ‹
Prad = r̂W0 · r̂dS = W0 r2 sin θdθdϕ = 4πr2 W0 . (4.59)
S S

Hence, we have that

Prad
W0 = r̂W0 = r̂ (4.60)
4πr2
which shows the even spreading of the power density over an increasingly large sphere as r
increases.
We are often interested in removing the dependence on r in our far-field expressions
altogether. The form of (4.60) suggests that we can define what is known as the radiation
intensity as

U = r2 |Wrad |. (4.61)

The radiation intensity helps measure the power that is radiated from an antenna per unit
solid angle. Solid angles are measured in units of steradians, which are a generalization of
the concept of a radian (a one-dimensional relation on a circle) to a two-dimensional area on
a sphere. The exact definition isn’t important for our purposes here, the main idea is that
the far-field radiation intensity of an antenna will not depend on how far the observation
distance is from the antenna. As a result, it is a useful reference property that doesn’t have
any ambiguities related to how far away the antenna is, which would always persist if we
tried to specify things in terms of the radiation power density.
We can now define one of the most commonly used ways to characterize the performance
of an antenna. Known as the directivity, this parameter is equal to the ratio of the antenna’s
radiation intensity in a given direction normalized by the radiation intensity of an isotropic
source. Its mathematical definition is
U 4πU
D= = , (4.62)
U0 Prad

where we have also used the fact that the equivalent isotropic radiation intensity is equal to
U0 = Prad /4π. Often, the directivity pattern of the antenna will be plotted and the maximum
directivity will also be reported to characterize an antenna.
Let’s now calculate the directivity for the Hertzian dipole we considered earlier. Since
the directivity is a far-field property of the antenna, we can use as our beginning fields

kI0 ℓ sin θ −jkr

Hϕ = j e (4.63)
4πr

208
 CHAPTER 4. ANTENNAS

kI0 ℓ sin θ −jkr

Eθ = jη e . (4.64)
4πr
Looking at the definition of the directivity in (4.62), our first step will be finding the radiation
intensity. Evaluating the quantities in (4.61) will give us the radiation intensity of a Hertzian
dipole as
 2  2
2 2η kℓ 2 2η ℓ
U = r |Wrad | = |I0 | sin θ = |I0 | sin2 θ (4.65)
2 4π 8 λ

From the definition in (4.62), we see that we also need to determine the total radiated power.
We can find this by integrating the radiation intensity over all directions. That is,
ˆ 2π ˆ π  2
2π ℓ
Prad = U sin θdθdϕ = η|I0 | (4.66)
0 0 3 λ

We can now substitute the results of (4.65) and (4.66) into the directivity definition to get
that
3 2
D= sin θ. (4.67)
2

From this, we can conclude that the maximum directivity will occur at θ = π/2 and will be
equal to

3
D0 = . (4.68)
2
A plot of the Hertzian dipoles directivity pattern is shown in Fig. 4.9. From this, we can also
see that the there will be no far-field radiation along the axis of the dipole itself. This makes
intuitive sense, as we know that the electric field produced by the dipole near its endpoints
will be oriented in the same direction as the r̂ vector at these points, and so won’t be able
to produce a TEM wave that can propagate away from the antenna.
Although the directivity can often be computed using numerical techniques, it can be dif-
ficult to measure. The difficulty is that any real antenna will have a number of different loss
mechanisms associated with it: such as conductor losses, dielectric losses, input matching
losses, and so on. It can be difficult to determine all of these different contributing factors
accurately, and they invariably impact the total power received during any antenna measure-
ment. As a result, the loss will bias the results of any measurements of the directivity of an
antenna. To accommodate this unavoidable issue, we most often will measure the antenna
gain. This is simply the directivity of the antenna multiplied by its efficiency, i.e.,

G = e0 D (4.69)

where e0 is the total efficiency of the antenna and ranges from 0 to 1. The gain is a quantity
that we can readily measure to characterize an antenna, and is also the quantity of interest
for many system-level analyzes (e.g., a link budget between two communication system
antennas).

209
CHAPTER 4. ANTENNAS

Figure 4.9: Example polar plot of the directivity pattern for a Hertzian dipole.

Often, both the directivity and gain will be reported in terms of decibels. In this case,
one can convert either the gain or directivity into decibel units by taking 10 log10 (X), where
X should either be the directivity D or the gain G. Likewise, if a directivity or gain is
given in terms of decibels, its linear scale units can be easily found by inverting the earlier
relationship. Although it is typical for directivity or gain to be reported in decibels, it is
important to emphasize that all equations given in these lecture notes are only valid for the
directivity or gain when expressed on a linear scale.
Another important property of an antenna is its input impedance. This is the same
concept as we dealt with when discussing how a general electromagnetic network could
be viewed as an equivalent input impedance from the perspective of a transmission line
network. As a result, we can conclude that the real part of the antenna’s input impedance
will characterize the effective loss of the antenna from the perspective of the transmission
line. However, this loss can generally be broken into two parts, a loss due to actual power
dissipation and a “loss” due to radiation. Since the fields propagate away from the antenna,
this effectively looks like lost power to the transmission line network. To characterize this,
we will sometimes define a radiation resistance. This is a fictitious resistance value that
properly characterizes how much power is “lost” to radiation for a given voltage or current
flowing into the antenna.
We can also conclude that the imaginary part of the antenna’s input impedance will
largely characterize the reactive near-fields of the antenna. The form of this impedance will
depend on the type of antenna we are dealing with. For instance, the Hertzian dipole will
generally be expected to have a capacitive reactance. This makes sense intuitively, because
the extremely small dipole can be thought of as being similar to a slightly modified “open
circuit” termination to a transmission line. We know that a small distance away from an
open circuit will look like a capacitive reactance, which is the case here. This also clues us
in on one of the difficulties with Hertzian dipoles as practical antennas. In general, they

210
 CHAPTER 4. ANTENNAS

have a very large reactance and a very small radiation resistance. As a result, it is difficult
to have them efficiently radiate energy. This is a common difficulty for any antenna that is
electrically small. There is currently a large amount of research interest in determining ways
to more efficiently radiate power from electrically small antennas.
To address the issues with a small radiation resistance and a large reactance, we often
need to make our antennas electrically larger. A classic example is a half-wavelength dipole
antenna. By increasing the length of the arms of the dipole, it can begin to act more
like a resonator. If we operate the antenna at the right frequencies, we can achieve a higher
radiation resistance and a reactance that can be more easily addressed with realistic matching
circuits. For instance, a half-wavelength dipole has an input impedance of approximately
Zin = 73 + j42.5Ω, which can be readily matched using various techniques. Another effect
that happens when we make our antennas electrically larger is that they tend to be better
able to focus radiation into a more directive pattern. A great example of this is a reflector
antenna. To achieve a narrower antenna pattern we need to make the reflector larger so that
it can collect and focus energy from a larger “sampling area”.

4.6 Friis Transmission Equation

One of the fundamental purposes of an antenna is to transmit power wirelessly between two
points in space. The Friis transmission equation provides us with a simple tool to compute
how much power will be received by a particular antenna given a certain amount of power
that is transmitted by another antenna. We will consider a simple derivation of this equation
now that is valid if both antennas are located in each other’s far-field region. Before getting
into the details, there is an important fact about the reciprocity of antennas that we should
introduce. The basic principle is that due to the reciprocity of electromagnetic systems
(this is a general electromagnetic theorem you will learn about in advanced EM courses),
antennas will typically be reciprocal devices under many practical circumstances. One of
the main consequences of this is that we don’t need to use different antenna gain patterns
to describe the antenna in transmission or receive mode, the two patterns will be identical
due to reciprocity.
To begin, we need to determine the power density that will be produced by our trans-
mitter. By definition, we know that the power density for an isotropic source with transmit
power Pt will be
Pt
W0 = . (4.70)
4πR2
We also know that by definition, the power density of our actual transmitter will be equal
to W0 modified by the gain pattern of the transmitter. In particular, we will get that

Pt Gt (θt , ϕt )
Wt = , (4.71)
4πR2
where Gt is the gain of the transmitter.
To determine the power that is received by the receiving antenna, it is useful to introduce
another antenna parameter known as the effective area or effective aperture of the antenna.

211
CHAPTER 4. ANTENNAS

This is a fictitious area that when multiplied by an incident power density will give the power
delivered to a matched load at the antenna’s terminals. If we denote the effective area as
Ar , we will get that the power received Pr is
Pt Gt (θt , ϕt )
Pr = Ar . (4.72)
4πR2
One special property of the effective area is that there is a universal relationship between
the ratio of the effective area and gain for any antenna. The particular ratio is that

Ar λ2
= . (4.73)
Gr 4π
Considering this, we can rewrite (4.72) as
 2
λ
Pr = Pt Gt (θt , ϕt )Gr (θr , ϕr ). (4.74)
4πR
It is traditional to write this as a ratio between Pr and Pt to define the link budget of a
system. This gives us the simple form of Friis transmission equation to be
 2
Pr λ
= Gt (θt , ϕt )Gr (θr , ϕr ). (4.75)
Pt 4πR
The factor in the parentheses is often called the free-space loss factor, since it takes into
account the spherical spreading of power as it propagates in space. The gains in this equation
then serve to help correct for the fact that our antennas don’t actually transmit or receive
energy into a perfect sphere, but instead focus it in a particular direction.
Before moving on, it should be mentioned that this is one of the simplest forms of Friis
transmission equation. Many other factors can need to be taken into account to make the
formula more complete. For instance, one simple factor we neglected was the polarization of
the antennas with respect to each other. If two antennas are oriented so they transmit or
receive orthogonal polarizations the amount of power that can be sent between the two will
drop dramatically.

4.7 Introduction to Antenna Arrays

For many applications, it is desirable to have a large amount of control over the antenna
pattern. For instance, radar or communication systems often want to achieve a large gain
in a particular direction to avoid wasting power by allowing it to spill out into undesirable
directions. Likewise, for high-performance applications we typically have stringent require-
ments on the maximum level that sidelobes can reach without spoiling the properties of a
system, such as an imaging radar for example.
We have already seen that the radiated fields of the antenna will depend on the current
density that exists on the antenna as
˚
1 −jkR
A(r) = e µJ(r′ )dr′ , (4.76)
4πR

212
 CHAPTER 4. ANTENNAS

Figure 4.10: Illustration of the far-field approximation. When the observation point is
far away from the source point, the vectors r and R are approximately parallel and certain
simplifications can be made to simplify the evaluation of “radiation integrals” that are needed
to analyze an antenna.

where R = |r − r′ |. In the far-field, we are able to approximate the electric field as being
E ≈ −jωA, (4.77)
where A will only have non-zero θ- and ϕ-components due to the far-field assumption. Hence,
we have that
˚ −jkR
−jωµ e
E≈ J(r′ )dr′ . (4.78)
4π R
If we only care about finding the fields in the far-field region, we can simplify the integration
by assuming that
R ≈ r for amplitude variations, (4.79)
R ≈ r − r′ cos ψ for phase variations, (4.80)
where ψ is the angle between the vectors r and r′ . This approximation is illustrated in Fig.
4.10, where it shows that in the far-field region the difference between R and r will often be
offset by a small factor.
For amplitude values, this small value often doesn’t make a significant difference so we
can use the simplification given above. However, because the phase factor depends on the
wavelength, this small difference can lead to large changes in behavior. This is why we need
to add the additional term given above for accounting for phase variations in the far-field.
Although it is still approximate, it does a very good job while still greatly simplifying the
overall difficulty of the integrations involved. Considering these simplifications, we have that
˚
e−jkr ′
E ≈ −jωµ J(r′ )e−jkr cos ψ dr′ . (4.81)
4πr

213
CHAPTER 4. ANTENNAS

The important point is that if we expand r′ cos ψ in a particular set of coordinates that are
convenient for a given problem we can always view this integral as being a spatial Fourier
transform of the current density.
To see this, it is helpful to expand r′ cos ψ for the simple case of a planar antenna confined
in the z = 0 plane. In this case, we find that (4.81) becomes
¨
e−jkr ′ sin θ cos ϕ −jky ′ sin θ sin ϕ
E ≈ −jωµ J(x′ , y ′ )e−jkx e dx′ dy ′ . (4.82)
4πr

If we define kx = k sin θ cos ϕ and ky = k sin θ sin ϕ, we can quickly recognize that the
integration is simply a spatial Fourier transform of the current density along the x′ and y ′
axes. Each data point in the (kx , ky )-plane can then be considered to give us the complex
amplitude of the plane wave that would be radiated into a particular (θ, ϕ)-direction, which
corresponds to our far-field antenna pattern. As a result, we can use our intuition about
Fourier transforms to help determine the amplitude and phase that J(r′ ) should have to
achieve a desired radiation pattern.
Although this is a powerful concept, it is not typically very easy to implement in practice.
The issue is that J(r′ ) must satisfy Maxwell’s equations, and as a result it is a complicated
function of the geometry that the current exists on. Hence, we can’t arbitrarily control J(r′ )
over a large area, and in general it will be difficult to get it to even approximate a particular
structure that may be needed to produce a desired antenna pattern.
The way around this issue is to use antenna arrays. Antenna arrays are formed by a
combination of discrete antennas that are driven together to achieve some desirable radiation
characteristics. Often, each antenna in the array will have the same basic shape, as it
greatly simplifies the design, analysis, and manufacturing of the array, although there is not
necessarily a requirement for this to be the case. We can think of using an array as being
synonymous with approximating a continuous Fourier transform with a discrete Fourier
transform. The important practical consequence is that we can achieve a great amount of
control over the different excitations of each antenna within an antenna array by designing
a suitable transmission line network. This network is often referred to as a feed network.
Designing a feed network is a very tractable problem, whereas designing a continuous source
to have a desired current density is almost impossible. Hence, antenna arrays are incredibly
prevalent in practical applications.
The overall fields radiated by an antenna array is the vector superposition of the fields
radiated from each individual antenna. As a result, if we already know the radiation pattern
of the array element, we can quickly find the far-field pattern of the entire array (under
somewhat ideal conditions). To see this, let’s consider a simple two-element array oriented
along the z-axis. If both antennas have the same complex-valued electric field pattern in
the far-field given by F(θ, ϕ), then for the array shown in Fig. 4.11 we have that the total
radiated field will be

e−j(kR1 −∆/2) e−j(kR2 +∆/2)

Et = E1 + E2 = |A1 |F(θ, ϕ) + |A2 |F(θ, ϕ) (4.83)
4πR1 4πR2

where ∆ is the phase difference between the excitations of the two antenna elements. We

214
 CHAPTER 4. ANTENNAS

Figure 4.11: Illustration of the simple two-element array.

can make our far-field approximation by assuming that

R1 ≈ R2 ≈ r for amplitude variations (4.84)

d
R1 ≈ r − cos θ for phase variations (4.85)
2
d
R2 ≈ r + cos θ for phase variations. (4.86)
2
We can then write our total electric field as

e−jkr
 
j(kd cos θ+∆)/2 −j(kd cos θ+∆)/2
Et = F(θ, ϕ) |A1 |e + |A2 |e . (4.87)
4πr

If we were to set |A| = |A1 | = |A2 |, we could simplify our pattern further to be

e−jkr
 
1
Et = F(θ, ϕ) × 2|A| cos (kd cos θ + ∆) . (4.88)
4πr 2

We have used the × symbol to emphasize that we can look at the total pattern as being the
multiplication between the antenna pattern of the individual elements times what is often
referred to as the array factor. This concept is known as pattern multiplication, and is the
reason why the analysis of most antenna arrays involves primarily finding the array factor
of a proposed array. Before concluding, it is worth noting that we can use ∆ to bias the
array factor to have a maximum value at a particular value of θ. This allows us to steer the
mainlobe of an antenna array in a desirable direction by choosing suitable phase values at
the different antenna elements. This is referred to as a phased array, and is a very common
tool in many practical antenna applications.

215
CHAPTER 4. ANTENNAS

4.8 Practice Problems

1. Assume that two antennas operating at 4 GHz are in the far-field of each other and are
aligned such that there is no loss due to polarization mismatch. Further, assume that
the antennas are oriented so that the angles between the two antennas correspond to
both antennas maximum gain point in their respective patterns. You may also assume
the antennas operate in free space.

(a) If the antennas are 90 km apart, the transmit power is 500 W, and the gain of the
transmitter is 24 dB, what is the power density at the location of the receiver?
(b) If the received power at the receive antenna is 0.15 µW, what is the gain of the
receive antenna in dB?

2. Assume that a two-element array of infinitesimal dipoles are located symmetrically

about the z-axis and are a distance d = λ/2 apart. Further, the antenna excita-
tions are set such that they have the same magnitude and 180◦ phase difference. For
this scenario, roughly sketch the directivity pattern for the the two-element array of
infinitesimal dipoles as a polar plot in the (y, z)-plane.

3. Assuming that a two-element array of infinitesimal dipoles are located symmetrically

about the z-axis and are a distance d = λ/4 apart, answer the following questions.

(a) Assuming that the magnitude of the excitations of both dipoles are identical,
what should the relative phase shift between the two antenna excitations be so
that the far-field antenna pattern of the array produces no radiation at θ = 60◦ ?
(b) For the excitation scheme described in (a), roughly sketch the directivity pattern
for the the two-element array of infinitesimal dipoles in the (y, z)-plane.

216
Bibliography

[1] D. Styer, “Calculation of the anomalous magnetic moment of the electron,” 2012.

[2] Wikipedia contributors, “Gradient — Wikipedia, the free encyclopedia,” https://en.

wikipedia.org/w/index.php?title=Gradient&oldid=1196514741, 2024, [Online; accessed
9-February-2024].

[3] W. C. Chew, Waves and Fields in Inhomogeneous Media. IEEE Press, 1995.

[4] Wikipedia contributors, “Plane wave — Wikipedia, the free encyclopedia,” https:
//en.wikipedia.org/w/index.php?title=Plane wave&oldid=1187992025, 2023, [Online;
accessed 9-February-2024].

[5] ——, “Over-the-horizon radar — Wikipedia, the free encyclopedia,” https://

en.wikipedia.org/w/index.php?title=Over-the-horizon radar&oldid=1195102860, 2024,
[Online; accessed 9-February-2024].

[6] I. D. Olin, “Polarization characteristics of coherent waves,” IEEE Antennas and Prop-
agation Magazine, vol. 56, no. 3, pp. 90–102, 2014.

[7] Wikipedia contributors, “Polarization (waves) — Wikipedia, the free en-

cyclopedia,” https://en.wikipedia.org/w/index.php?title=Polarization (waves)&oldid=
1203084697, 2024, [Online; accessed 9-February-2024].

[8] ——, “Turnstile antenna — Wikipedia, the free encyclopedia,” https://en.wikipedia.

org/w/index.php?title=Turnstile antenna&oldid=1095640919, 2022, [Online; accessed
9-February-2024].

[9] C.-C. Chang, H. P. Sardesai, and A. M. Weiner, “Dispersion-free fiber transmission for
femtosecond pulses by use of a dispersion-compensating fiber and a programmable pulse
shaper,” Optics Letters, vol. 23, no. 4, pp. 283–285, 1998.

[10] S. Ullah, W.-H. Yeo, H. Kim, and H. Yoo, “Development of 60-GHz millimeter wave,
electromagnetic bandgap ground planes for multiple-input multiple-output antenna ap-
plications,” Scientific Reports, vol. 10, no. 1, pp. 1–12, 2020.

[11] Wikipedia contributors, “Waveplate — Wikipedia, the free encyclopedia,” https://

en.wikipedia.org/w/index.php?title=Waveplate&oldid=1196060382, 2024, [Online; ac-
cessed 9-February-2024].

217
BIBLIOGRAPHY

[12] ——, “Faraday effect — Wikipedia, the free encyclopedia,” https://en.wikipedia.

org/w/index.php?title=Faraday effect&oldid=1200966862, 2024, [Online; accessed 9-
February-2024].

[13] D. M. Pozar, Microwave Engineering. John Wiley & Sons, 2009.

[14] D. K. Cheng, Field and Wave Electromagnetics, 2nd ed. Addison-Wesley Publishing
Company, Inc., 1989.

[15] E. Jensen, “Cavity basics,” arXiv preprint arXiv:1201.3202, 2012.

[16] C. A. Balanis, Antenna Theory: Analysis and Design. John Wiley & Sons, 2005.

218