Intro To Wave Physics

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/349721510
Introduction to Wave Physics
Book · March 2021
CITATIONS READS
0 1,124
1 author:
Anthony L Gerig
Viterbo University
25 PUBLICATIONS 171 CITATIONS
SEE PROFILE
All content following this page was uploaded by Anthony L Gerig on 06 March 2021.
The user has requested enhancement of the downloaded file.

Introduction to Wave Physics
Anthony L. Gerig, Ph.D.

ii
Copyright © 2021 Anthony L. Gerig

All rights reserved.
Dedication
To Julie, Clayton, Gretchen and my students
iii
iv DEDICATION
Contents
Dedication iii
Preface ix
1 One-dimensional Waves 1
1.1 Wave Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Harmonic Waves . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Complex Representation . . . . . . . . . . . . . . . . . . . . . 7
1.4 Interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 One-dimensional Wave Equation . . . . . . . . . . . . . . . . . 13
1.5.1 Wave Equation for a String . . . . . . . . . . . . . . . 14
1.5.2 Separation of Variables . . . . . . . . . . . . . . . . . . 15
1.6 Fourier Transforms . . . . . . . . . . . . . . . . . . . . . . . . 18
1.7 Energy Transmission . . . . . . . . . . . . . . . . . . . . . . . 21
1.8 Reflection and Transmission . . . . . . . . . . . . . . . . . . . 22
1.9 Finite Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.9.1 Fixed Strings . . . . . . . . . . . . . . . . . . . . . . . 26
1.9.2 Free Strings . . . . . . . . . . . . . . . . . . . . . . . . 30
1.9.3 Joined Strings . . . . . . . . . . . . . . . . . . . . . . . 33
1.10 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.11 Vector Waves and Polarization . . . . . . . . . . . . . . . . . . 37
2 Three-dimensional Wave Functions 41

2.1 Three-dimensional Wave Equations . . . . . . . . . . . . . . . 41
2.2 Separable Solutions . . . . . . . . . . . . . . . . . . . . . . . . 44
2.2.1 Rectangular Wave Functions . . . . . . . . . . . . . . . 45
2.2.2 Spherical Wave Functions . . . . . . . . . . . . . . . . 49
2.2.3 Cylindrical Wave Functions . . . . . . . . . . . . . . . 60
v
vi CONTENTS
2.3 Energy Transmission . . . . . . . . . . . . . . . . . . . . . . . 65

2.3.1 Acoustic (Scalar) Intensity . . . . . . . . . . . . . . . . 65
2.3.2 Electromagnetic (Vector) Intensity . . . . . . . . . . . 69
3 Interference 73
3.1 Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2 Spherical Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Application: Linear Arrays . . . . . . . . . . . . . . . . . . . . 80
4 Reflection and Transmission 87

4.1 Plane Waves at Normal Incidence . . . . . . . . . . . . . . . . 88
4.1.1 Acoustic Waves . . . . . . . . . . . . . . . . . . . . . . 88
4.1.2 Electromagnetic Waves . . . . . . . . . . . . . . . . . . 90
4.2 Plane Waves at Oblique Incidence . . . . . . . . . . . . . . . . 92
4.2.1 Acoustic Waves . . . . . . . . . . . . . . . . . . . . . . 92
4.2.2 Electromagnetic Waves . . . . . . . . . . . . . . . . . . 97
4.3 Non-Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . 102
5 Cavities and Waveguides 103

5.1 Rectangular Cavity . . . . . . . . . . . . . . . . . . . . . . . . 104
5.1.1 Acoustic . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.1.2 Electromagnetic . . . . . . . . . . . . . . . . . . . . . . 108
5.2 Cylindrical Cavity . . . . . . . . . . . . . . . . . . . . . . . . 113
5.3 Spherical Cavity . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.4 Rectangular Waveguide . . . . . . . . . . . . . . . . . . . . . . 122
5.4.1 Acoustic . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.4.2 Electromagnetic . . . . . . . . . . . . . . . . . . . . . . 128
5.5 Cylindrical Waveguide . . . . . . . . . . . . . . . . . . . . . . 132
6 Radiation 135
6.1 Boundary Condition Method . . . . . . . . . . . . . . . . . . . 135
6.1.1 Pulsating Sphere . . . . . . . . . . . . . . . . . . . . . 136
6.1.2 Vibrating Sphere . . . . . . . . . . . . . . . . . . . . . 138
6.1.3 Sources Near Boundaries . . . . . . . . . . . . . . . . . 140
6.2 Inhomogeneous Wave Equation Method . . . . . . . . . . . . . 140
6.2.1 Radiation by Point Sources . . . . . . . . . . . . . . . 143
6.2.2 Radiation by Extended Sources . . . . . . . . . . . . . 149
CONTENTS vii
7 Diffraction 159
7.1 Kirchoff Diffraction Theory . . . . . . . . . . . . . . . . . . . 159
7.2 Rectangular Aperture . . . . . . . . . . . . . . . . . . . . . . . 165
7.3 Circular Aperture . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.4 Multiple Apertures . . . . . . . . . . . . . . . . . . . . . . . . 172
8 Scattering 177
8.1 Scattering Terminology . . . . . . . . . . . . . . . . . . . . . . 177
8.2 Boundary Condition Method . . . . . . . . . . . . . . . . . . . 180
8.3 Inhomogeneous Medium Method . . . . . . . . . . . . . . . . 187
8.4 Kirchoff Method . . . . . . . . . . . . . . . . . . . . . . . . . . 193
viii CONTENTS
Preface
This book developed out of a one-semester course on Waves and Optics that I
taught junior and senior physics majors at Viterbo University. I had planned
to adopt a text for the course that would prepare students both for further
work in the areas of acoustics, electromagnetics and optics, and for the rig-
ors of a senior-level course in quantum mechanics. However, I had difficulty
finding an undergraduate-level textbook that addressed all three topics of
radiation, diffraction and scattering, and that consistently covered material
in three dimensions. I therefore chose not to require a textbook and instead
pulled material from various sources to produce course notes that were dis-
tributed to the class. Former students who were doing graduate research
or were employed in the areas described above often mentioned that they
were continuing to use the notes in the course of their work. Consequently, I
decided to to make the notes more usable by converting them into this text.
The book is designed to proceed from simple to complex, using one-
dimensional waves to introduce major concepts in the first chapter, and ex-
tending each of those concepts to three dimensions in subsequent chapters
for rectangular, cylindrical and spherical coordinates. Both scalar and vec-
tor waves are addressed, using acoustic waves as the primary example of
the former and electromagnetic/optical waves as the primary example of the
latter.
Because three-dimensional waves are covered extensively, readers should
be familiar with multivariable calculus and have some exposure to curvilinear
coordinates. Courses in differential equations, mid-level mechanics and mid-
level electricity/magnetism are helpful, but not required.
ix
x PREFACE
Chapter 1
One-dimensional Waves
Waves are interesting, but also vital, natural phenomena. They are com-
monly the physical mechanism that is responsible whenever energy or infor-
mation is transmitted from one location to another. Without them, there
would be no light or sound, and many modern technologies would be nonex-
istent, including radio communication, microwave ovens, medical ultrasound,
X-ray imaging, radar, sonar, fiber-optics, etc.
Fundamentally, a wave is a variation, or a change, in a physical quantity
that propagates through space over time. For example, when a stone strikes
the surface of a body of water, it creates movement, or a displacement, in the
water at that location. That displacement, however, isn’t confined to a single
location. The moving water exerts a force on the surrounding water, which
causes it to become displaced. That water exerts a force on the neighboring
water, which causes it to become displaced and so on. In this manner, the
displacement propagates outward from the location of the strike in the form
of expanding circular ripples on the surface of the water.
There are many physical quantities that will propagate when varied.
Waves can be classified according to whether the propagating quantity is
a scalar, vector or tensor. Examples of propagating scalar quantities include
fluid pressure, which is experienced as sound, and the quantum mechani-
cal wave function for a spin-less particle. Examples of propagating vector
quantities include electromagnetic fields, which are experienced as light, and
material displacements like that described above for water. An example of a
propagating tensor quantity is the metric tensor associated with gravity.
Waves can be further categorized according to whether the propaga-
tion occurs in one, two or three spatial dimensions. An example of a one-
1
2 CHAPTER 1. ONE-DIMENSIONAL WAVES
dimensional wave is a displacement wave propagating along a plucked string.

A good example of a two-dimensional wave is the one described above: a dis-
placement wave propagating across the surface of a body of water. Examples
of three-dimensional waves include an electromagnetic/optical wave propa-
gating outward from an antenna, or a pressure/acoustic wave propagating
outward from a speaker.
The primary purpose of this book is to introduce the reader to the funda-
mental aspects of wave behavior and the mathematics necessary to describe
that behavior. It focuses on three-dimensional waves due to their importance
and prevalence. However, the first chapter introduces many of the founda-
tional concepts using one-dimensional waves, with displacement waves on a
string as the primary example, because they ease the reader into the required
mathematics. The remaining chapters extend the concepts to three dimen-
sions, using acoustic waves as the primary example of a three-dimensional
scalar wave and electromagnetic waves as the primary example of a three-
dimensional vector wave.
1.1 Wave Functions

A wave is represented mathematically by a wave function, which is a math-
ematical function that yields the value of the propagating quantity at each
location in space for any given time. It can be determined intuitively for most
one-dimensional waves. Consider, for example, a Gaussian pulse propagating
on a string with speed v as illustrated in Figure 1.1. The figure plots the
displacement of the string ψ as a function of position x, which is called the
waveform, both initially and at a later time t. (Note that although displace-
ment is a vector quantity ψ that can point in any direction perpendicular
to the string, here it is constrained to the plane of the page to simplify the
mathematics and can be represented by a scalar ψ.) For a Gaussian pulse,
the initial waveform is given by the function:
x2
ψ(x) = Ae− 2σ2
where σ is a measure of the pulse width. The shape of the waveform at t is

identical, but shifted in the positive x direction a distance vt. The function
for this waveform can therefore be obtained by subtracting vt from x in the
original function:
1.1. WAVE FUNCTIONS 3
vt
ψ(x)
Figure 1.1: A pulse propagating on a string in the positive x direction with

speed v.
(x−vt)2
ψ(x, t) = Ae− 2σ 2
Because the value of t is arbitrary, this function yields the displacement of

the string at each location in space for any time and is therefore the wave
function for the pulse. For a pulse traveling in the opposite direction, the
wave function is obtained by adding vt rather than subtracting it. The
process is identical for any other wave on the string, so the wave function
can be written generally as:
ψ(x, t) = f (x ± vt) (1.1)

where the function f specifies the shape of the initial waveform.
The propagation speed for most waves is determined by the properties

of the medium, or material, through which they propagate. For displace-
ment waves, the propagation speed is higher when medium elements respond
quickly to the movement of neighboring elements, which occurs when they are
lighter and the forces exerted between them are stronger. The propagation
speed for a displacement wave on a string is given by:
s
T
v= (1.2)
µ
where T is the tension in the string and µ is the linear mass density of the
string. The lower the mass density and the higher the tension, the faster the
propagation.
1.2 Harmonic Waves

When a propagating quantity oscillates harmonically, it produces a specific
type of wave called a harmonic wave. Imagine, for example, that the dis-
placement of a string at the origin is given by:
2π
ψ(t) = A cos t = A cos 2πf t
T
where A is the amplitude, or maximum displacement, of the string, T is the
period, or the time it takes to complete one cycle, and f is the frequency,
which is the inverse of the period and a measure of the number of cycles
completed per unit time. The top pane of Figure 1.2 illustrates the motion.
Because the motion propagates, the displacement at every other location
along the propagation direction is also harmonic, but delayed by a time equal
to the distance from the origin divided by the propagation speed. Assuming
propagation in the positive x direction, the displacement at the location x is
therefore given by:
x
ψ(x, t) = A cos 2πf (t − )
v
2πf
ψ(x, t) = A cos (vt − x)
v
2πf
ψ(x, t) = A cos (x − vt)
v
1.2. HARMONIC WAVES 5
T
ψ(t)
λ
ψ(x)
Figure 1.2: Harmonic motion of a string element (top) and the initial wave-
form produced by the motion (bottom).
which is the wave function for the harmonic wave.

Inserting t = 0 into the wave function to get the initial waveform yields:
2πf
ψ(x) = A cos x
v
which is plotted in the bottom pane of Figure 1.2. The distance between
identical displacements on the waveform is called the wavelength λ, and is
equal to the distance the displacement at a given location propagates away
in the time it takes the string at that location to complete one cycle and
return to the same displacement:
v
λ = vT = or v = f λ (1.3)
f
Notice that because the wavelength and period are proportional, the wave-
length and frequency are inversely related quantities, meaning that if one
increases, the other decreases by the same factor.
Substituting the wavelength into the wave function and accounting for
the possibility of propagation in the negative x direction:
2π
ψ(x, t) = A cos (x ± vt)
λ
Using Equation 1.3, this can be rewritten as:
2π v
ψ(x, t) = A cos ( x ± 2π t)
λ λ
2π
ψ(x, t) = A cos ( x ± 2πf t)
λ
ψ(x, t) = A cos (kx ± ωt)
where k = 2π/λ is called the wavenumber and ω = 2πf the angular frequency
of the harmonic wave. The two can be related using Equation 1.3:
ω = kv (1.4)
Because the cosine function is even, A cos (kx + ωt) = A cos (−kx − ωt) and
the wave equation can be recast as:
ψ(x, t) = A cos (±kx − ωt)
where the positive sign indicates propagation in the positive x direction and
the negative sign propagation in the negative x direction (the reason for
switching the location of the ± will become clear later). Finally, the wave
function above assumes that the displacement of the string at the origin is
initially at a maximum (ψ(0, 0) = A) as illustrated in the top pane of Figure
1.2. If this is not the case, a phase angle φ (often referred to simply as the
phase) is required in the argument of the wave function to shift the curve to
a different initial displacement (ψ(0, 0) = A cos(φ)). Including it yields the
wave function for a harmonic wave in its most general form:
ψ(x, t) = A cos (±kx − ωt + φ) (1.5)

1.3. COMPLEX REPRESENTATION 7
1.3 Complex Representation

An imaginary number is the square root of any negative number and is
written in the following manner:
√ √ √
−n = n −1 = bi
where
√ √
i = −1 and b = n
A complex number is a number with real and imaginary parts, and is ex-
pressed as:
a + bi
where
Re(a + bi) = a and Im(a + bi) = b
Taking the complex conjugate of a complex number is defined as reversing the

sign on its imaginary part by flipping the sign in front of i, and is represented
by a superscript asterisk:
(a + bi)∗ = a − bi
Complex numbers can be plotted in two dimensions as shown in Figure 1.3,
where the real part is the horizontal coordinate and the imaginary part the
vertical coordinate. As illustrated in the figure, the complex number can
also be represented as a vector called a phasor with magnitude A and angle
θ given by:
√
A= a2 + b2 and θ = tan−1 b/a (1.6)
or
a = A cos θ and b = A sin θ (1.7)
Euler’s identity expresses a complex exponential as a complex number,

and can be derived by Taylor expanding the exponential:
Figure 1.3: Plot of a complex number in two dimensions.
(iθ)0 (iθ)1 (iθ)2 (iθ)3

eiθ = + + + ...
0! 1! 2! 3!
1 1
= (1 − θ2 + ...) + i(θ − θ3 + ...)
2 6
= cos θ + i sin θ (1.8)
Making use of Euler’s identity, any complex number can be written in terms
of its phasor magnitude and angle using a complex exponential:
a + bi = A cos θ + iA sin θ = Aeiθ (1.9)

Using this expression, any harmonic wave function can be written as the real
part of a complex number:
1.4. INTERFERENCE 9
Re(Aei(±kx−ωt+φ) ) = A cos (±kx − ωt + φ) = ψ(x, t) (1.10)

which is referred to as its complex representation. Equivalently, it is the
horizontal component of a phasor as in Figure 1.3 with θ = ±kx−ωt+φ. For
simplicity, it is common practice to drop the real operator when working with
the complex representation as long as the mathematics are not affected and it
is understood that the imaginary part is extraneous. The phase angle is also
commonly combined with the amplitude to produce a complex amplitude:
ψ(x, t) = Aei(±kx−ωt+φ) = Aeiφ ei(±kx−ωt) = Ãei(±kx−ωt)

The benefit of using the complex/phasor representation for harmonic
wave functions is that mathematical operations are generally easier to per-
form on an exponential/vector than a cosine function. In particular, deriva-
tives and integrals can be applied simply by adding or removing factors of
the exponent. Adding and subtracting harmonic wave functions is also much
easier using phasors, as will be demonstrated in the next section.
1.4 Interference
When multiple waves encounter one another at the same location, they in-
teract to produce a combined wave in a phenomenon defined as interference.
The Superposition Principle dictates that the combined wave is simply the
sum of the interfering waves.
As an example, consider the interference between two identical harmonic
waves with different phase angles whose wave functions are given by:
ψ1 = Aei(kx−ωt) and ψ2 = Aei(kx−ωt+φ)

The combined wave function is therefore:
ψ = ψ1 + ψ2
= Aei(kx−ωt) + Aeiφ ei(kx−ωt)
= (A + Aeiφ )ei(kx−ωt)
where the sum of the complex amplitudes for the interfering waves yields
the complex amplitude for the combined wave, and can be found by adding
their phasors as illustrated in Figure 1.4. Because the three phasors form an
isosceles triangle, the phase angle for the combined wave is given by:
φc = φ/2 (1.11)
The amplitude for the combined wave is determined by applying Equation 1.6
to the real (A+A cos φ) and imaginary (A sin φ) components of the resultant,
and using the trigonometric identity cos x = 2 cos2 (x/2) − 1 to simplify the
result:
p
Ac = (A + A cos φ)2 + (A sin φ)2
p
= 2A2 + 2A2 cos φ
p
= 2A2 + 2A2 [2 cos2 (φ/2) − 1]
= 2A cos(φ/2) (1.12)
Notice that when the phase difference between the interfering waves φ is
zero or an even multiple of π, all three phasors in Figure 1.4 align, meaning
that their corresponding wave functions also align as illustrated in Part (a)
of Figure 1.5. The interfering waves are therefore said to be in phase. The
amplitude of the combined wave in this case is simply the sum of the ampli-
tudes of the interfering waves (Ac = 2A), which is also its maximum value.
Because the interfering waves reinforce one another, this type of interference
is called constructive. When the phase difference is an odd multiple of π,
the phasors for the interfering waves and their corresponding wave functions
in Part (b) of Figure 1.5 are anti-aligned. The amplitude of the combined
wave is therefore the difference between the amplitudes of the interfering
waves (Ac = 0). Because the interfering waves cancel each other, this type
of interference is called destructive. Between these two extremes, there is
no alignment between the phasors or their corresponding wave functions as
illustrated in Part (c) of Figure 1.5, and the interfering waves are said to
be out of phase. The amplitude of the combined wave is also between the
extremes and the interference is therefore called incomplete.
As another example, consider the interference between two identical har-
monic waves traveling in opposite directions whose wave functions are given
by:
ψ1 = Aei(kx−ωt) and ψ2 = Aei(−kx−ωt)

1.4. INTERFERENCE 11
Figure 1.4: Addition of phasors representing the complex amplitudes of inter-

fering harmonic waves, where the resultant represents the complex amplitude
of the combined wave.
ψ = ψ1 + ψ2
= Aeikx e−iωt + Ae−ikx e−iωt
= A(eikx + e−ikx )e−iωt
Applying Euler’s identity:
= A[cos kx + i sin kx + cos (−kx) + i sin (−kx)]e−iωt

= 2A cos (kx)e−iωt
ψ(x)
(a) x
ψ(x)
(b) x
ψ(x)
(c) x
Figure 1.5: Interfering (dashed and dotted) and combined (solid) harmonic
wave functions at t = 0 for (a) constructive interference, (b) destructive
interference, and (c) incomplete interference when φ = π/4.
Taking the real part:
ψ = 2A cos (kx) cos ωt
Notice that every point on the string now has the same time dependence
(cos ωt), meaning that they all oscillate in unison. There no delay between
the movement at one location and the corresponding movement at another
location farther down the string, implying that the displacement no longer
propagates. This type of wave is therefore referred to as a standing wave. In
addition, the amplitude of the oscillation is no longer uniform, but varies with
location on the string according to the absolute value of the leading term in
the wave function |2A cos kx|. Whenever kx = nπ or x = n(λ/2) where n is
1.5. ONE-DIMENSIONAL WAVE EQUATION 13
an integer, the amplitude takes on its maximum value of 2A. These locations
occur every half-wavelength, and are called anti-nodes. Whenever kx =
(n+1/2)π or x = (n+1/2)(λ/2), the amplitude is equal to zero and the string
is motionless. These locations also occur every half-wavelength between the
anti-nodes, and are called nodes. Figure 1.6 illustrates the motion of the
string, highlighting node and anti-node locations.
t =0
(1/8)T
N ode Antinode (3/8)T
(1/2)T
ψ(x)
Figure 1.6: A standing wave on a string generated by interfering harmonic

waves traveling in opposite directions. The waveform is shown at several
different times to illustrate the motion.
1.5 One-dimensional Wave Equation

Not all varying physical quantities propagate. The only ones that produce
waves are those whose dynamics are governed by the wave equation, which
in one-dimension is given by:
∂ 2ψ 1 ∂ 2ψ
= (1.13)
∂x2 v 2 ∂t2
where v is the propagation speed.
1.5.1 Wave Equation for a String

Consider string displacement as an example. Figure 1.7 shows a represen-
tative string element of length dx under tension T with mass density µ.
The vertical displacement of the string ψ is governed by the component of
Newton’s second law in that direction:
Figure 1.7: A representative string element of length dx under tension T

with mass density µ.
X
Fψ = maψ
From the figure:
∂ 2ψ
T sin θx+dx − T sin θx = (µdx) 2
∂t
sin θx+dx − sin θx µ ∂ 2ψ
=
dx T ∂t2
Applying the small angle approximation sin θ ≈ tan θ:
tan θx+dx − tan θx µ ∂ 2ψ

=
dx T ∂t2
∂ψ
Using tan θ = slope of string = ∂x
:
∂ψ ∂ψ
∂x x+dx
− µ ∂ 2ψ
∂x x
= 2
dx
T ∂t 2
∂ ∂ψ µ∂ ψ
=
∂x ∂x T ∂t2
∂ 2ψ 1 ∂ 2ψ
=
∂x2 T /µ ∂t2
which is
pthe wave equation, where the propagation speed of the displacement
is v = T /µ as given in Equation 1.2.
1.5.2 Separation of Variables

The general solution to the wave equation (Equation 1.1) was found earlier
using an intuitive approach. Its validity can be demonstrated by plugging it
into the wave equation:
1 ∂ 2ψ 1 ∂
= [±vf 0 (x ± vt)]
v 2 ∂t2 v 2 ∂t
1
= 2 v 2 f 00 (x ± vt)
v
∂2
= f (x ± vt)
∂x2
∂ 2ψ
=
∂x2
However, it is much more difficult to apply the same approach in multiple
spatial dimensions. A more methodical technique called separation of vari-
ables is required. By assuming that the solutions to the wave equation can
be separated into a product of functions of a single, independent variable,
the equation can be separated into multiple, one-dimensional equations that
are easier to solve.
The method is best illustrated by initially applying it to the one-dimen-
sional wave equation. The first step is to assume that the solutions can be
written as a product of functions of a single variable:
ψ(x, t) = X(x)T (t)

Inserting this function into the wave equation yields:
d2 X(x) 1 d2 T (t)
T (t) = X(x)
dx2 v2 dt2
2 2 2
v d X(x) 1 d T (t)
2
=
X(x) dx T (t) dt2
Because x and t are independent variables, the above equation can only hold
true for all values of x and t if both sides are equal to the same constant.
Otherwise, changing the value of x while holding t constant, for example,
would change the left-hand side without affecting the right-hand side and
the equality would no longer be valid. This constant is referred to as the
separation constant, and can be represented by any combination of other
constants. For reasons that will become clear later, −ω 2 is chosen as the
separation constant in this case:
v 2 d2 X(x) 2 1 d2 T (t)
= −ω =
X(x) dx2 T (t) dt2
The above can now be written as two separate equations, one a function of
t alone and the other a function of x alone:
d2 X(x) ω 2 d2 T (t)
+ X(x) = 0 and + ω 2 T (t) = 0
dx2 v2 dt2
The two equations are identical and equivalent to the equation of motion
for a simple harmonic oscillator, which has both complex and real solutions.
The complex solutions are given by:
˜ −ikx
T (t) = ãeiωt + b̃e−iωt and X(x) = c̃eikx + de
where ã, b̃, c̃ and d˜ are complex integration constants and k = ω/v. Multi-
plying the solutions produces the full wave function:
˜ i(−kx−ωt)
ψ(x, t) = X(x)T (t) = b̃c̃ei(kx−ωt) + b̃de
˜ i(−kx+ωt) (1.14)
+ ãc̃ei(kx+ωt) + ãde
The final two terms are redundant and can be dropped by setting ã = 0 be-
cause their real parts are identical to those for the first two terms. Combining
the complex integration constants and writing them as complex exponentials
yields:
ψ(x, t) = A1 ei(kx−ωt+φ1 ) + A2 ei(−kx−ωt+φ2 )

These solutions are the propagating harmonic wave functions of Equation
1.10, and will be referred to as the travelling wave solutions.
The real solutions for X(x) and T (t) are given by:
T (t) = a cos ωt + b sin ωt and X(x) = c cos kx + d sin kx

where a, b, c and d are integration constants. Multiplying the solutions pro-
duces the full wave function:
ψ(x, t) = A1 cos kx cos ωt + A2 cos kx sin ωt

+ A3 sin kx cos ωt + A4 sin kx sin ωt (1.15)
where the integration constants have been combined. These solutions are the
wave functions for harmonic standing waves, and will therefore be referred
to as the standing wave solutions. Because standing waves are produced
by interfering harmonic waves traveling in opposite directions, the standing
wave solutions are linear combinations of the traveling wave solutions and
not independent.
1.6 Fourier Transforms
It may initially seem that the solutions produced by the separation of vari-
ables method are not exhaustive. After all, there are many wave functions
that take the form of Equation 1.1 that are not harmonic. However, har-
monic waves with different frequencies, amplitudes and phases can interfere
to produce other types. For the moment, consider harmonic waves traveling
in the positive x direction exclusively. According to the Superposition Prin-
ciple, the wave function for a set of interfering harmonic waves spanning a
continuous and infinite range of frequencies/wavenumbers is given by:
Z ∞
ψ(x, t) = Re F̃ (k)ei(kx−ωt) dk
0
where the integral represents a continuous sum over frequency/wavenumber

and F (k) is an amplitude density (unit of amplitude per unit wavenumber)
to account for the multiplication by dk.
Extending the lower limit on the integral to negative infinity allows the
frequencies/wavenumbers to take on negative values. The real parts of the
corresponding wave functions are identical to those for their positive coun-
terparts (cos (kx − ωt) = cos (−kx + ωt)), so including them is redundant.
However, they play a useful role if the amplitude density for −k is required
to be the complex conjugate of that for k, F̃ (−k) = F̃ (k)∗ , because the
imaginary parts of their wave functions cancel:
1.6. FOURIER TRANSFORMS 19
F̃ (k)ei(kx−ωt) + F̃ (−k)ei(−kx+ωt)
= F̃ (k)ei(kx−ωt) + F̃ (k)∗ ei(−kx+ωt)
= F (k)eiφ ei(kx−ωt) + F (k)e−iφ ei(−kx+ωt)
= F (k)ei(kx−ωt+φ) + F (k)ei(−kx+ωt−φ)
= F (k) cos (kx − ωt + φ) + iF (k) sin (kx − ωt + φ)+
F (k) cos (−kx + ωt − φ) + iF (k) sin (−kx + ωt − φ)
= 2F (k) cos (kx − ωt + φ)
Including the negative wavenumbers with this condition on their amplitude
densities therefore simplifies the mathematics by allowing the real operator
to be dropped from the integral:
Z ∞
ψ(x, t) = F̃ (k)ei(kx−ωt) dk
−∞
Next, assume that the harmonic waves can interfere to produce a general
traveling wave of the form f (x − vt) with initial waveform f (x). To be true,
there must be a function F̃ (k) such that:
Z ∞
f (x) = ψ(x, 0) = F̃ (k)eikx dk (1.16)
−∞
It can be found by applying what is commonly referred to as Fourier’s Trick.
Both sides of the equation are multiplied by a second complex exponential,
and then integrated over space:
Z ∞ ZZ ∞
−ik0 x 0
f (x)e dx = F̃ (k)eikx e−ik x dkdx
−∞
Z ∞−∞ Z ∞
0
= F̃ (k) ei(k−k )x dxdk
−∞ −∞
Integrating a complex exponential yields a delta function multiplied by 2π,

so:
Z ∞
= F̃ (k)2πδ(k − k 0 )dk
−∞
= 2π F̃ (k 0 )
Dropping the prime from k 0 yields:

Z ∞
1
F̃ (k) = f (x)e−ikx dx (1.17)
2π −∞
which is called the Fourier transform of f (x). Its inverse, Equation 1.16, is
called the inverse Fourier transform of F̃ (k) or the Fourier integral of f (x).
Notice that if f (x) is a real function:
Z ∞
1
F̃ (−k) = f (x)e−i(−k)x dx
2π −∞
Z ∞
1
= f (x)eikx dx
2π −∞
Z ∞
1 ∗
f (x)e−ikx dx

=
2π −∞
Z ∞ ∗
1 −ikx
= f (x)e dx
2π −∞
= F̃ (k)∗
so the condition required to include the negative wavenumbers in the Fourier
integral is guaranteed. Inserting the time dependence into the initial wave-
form to get the wave function:
Z ∞ Z ∞
ik(x−vt)
f (x − vt) = F̃ (k)e dk = F̃ (k)ei(kx−ωt) dk (1.18)
−∞ −∞
For a wave that propagates in the negative x direction, the sign is simply
reversed:
Z ∞ Z ∞
i(kx+ωt)
f (x + vt) = F̃ (k)e dk = F̃ (k)ei(−kx−ωt) dk (1.19)
−∞ −∞
Consider the specific example of a Gaussian pulse. F̃ (k) is determined

by evaluating Equation 1.17:
Z ∞
1 2 /2σ 2

F̃ (k) = Ae−x e−ikx dx
2π −∞
1.7. ENERGY TRANSMISSION 21
which can be found in a table of integrals:
A√ 2 2
= 2πσ 2 e−k σ /2
2π
Aσ 2 2
= √ e−k σ /2
2π
If the pulse is travelling in the positive x direction, its wave function is given
by:
Z ∞
−
(x−vt)2 Aσ 2 2
ψ(x, t) = Ae 2σ 2
= √ e−k σ /2 ei(kx−ωt) dk
−∞ 2π
Notice that the Fourier transform of the Gaussian pulse F̃ (k) is also Gaus-
sian, but its width is determined by the value of 1/σ rather than σ. The
width of the Fourier transform, also called the bandwidth of the pulse, is
therefore inversely related to the width of the pulse. The broader the pulse,
the narrower the range of interfering harmonic waves required to produce
it. The reverse is also true. This trade-off is universal regardless of pulse
shape and is responsible for a wide variety of physical phenomena, including
Heisenberg’s Uncertainty Principle in quantum mechanics.
Because any traveling wave can be viewed as a combination of interfering
harmonic waves, the separation of variables solutions are exhaustive. For
this reason, the remainder of the book will focus on the mechanics of har-
monic waves. The behavior of any other wave can be determined simply by
adding the results for its constituent harmonic waves (multiplying by F̃ (k)
and integrating).
1.7 Energy Transmission

Physical quantities generally carry energy when they propagate. For a string,
it is the transfer of mechanical energy from one element to the next that
causes displacement to propagate. Calculating the energy transfer is usually
a straightforward process. Consider, for example, a harmonic wave propa-
gating along a string in the positive x direction, where the wave function is
given by Equation 1.5. The mechanical power transferred to a representative
string element, like the one shown in Figure 1.7, by the neighboring element
to the left is given by:
P =F ·v
where v is its velocity and F is the force exerted by the neighboring element.
Because the motion of the element is transverse:
P = Ft v
where Ft is the transverse component of the force, which in this case is the
vertical component of the tension in the string. From the figure:
∂ψ
P = −T sin θx
∂t
As noted in the derivation of the wave equation, sin θx ≈ ∂ψ/∂x:
∂ψ ∂ψ
= −T
∂x ∂t
= −T [−kA sin (kx − ωt + φ)][ωA sin (kx − ωt + φ)]
= T kωA2 sin2 (kx − ωt + φ)
substituting T = µv 2 from Equation 1.2 and k = ω/v:
P = µvω 2 A2 sin2 (kx − ωt + φ)
Because the average of sine squared over one period is equal to one half, the
average transmitted power is:
1
Pav = µvω 2 A2
2
Although this result is particular to a string, it is generally the case for any
harmonic wave that the energy transmission is a function of the propagation
speed and the squared amplitude.
1.8 Reflection and Transmission

To this point, waves that travel through a uniform medium have been con-
sidered exclusively. Finding wave functions for media whose properties vary
1.8. REFLECTION AND TRANSMISSION 23
continuously can be difficult and is beyond the purview of this book. How-
ever, cases where waves encounter boundaries between uniform media with
different properties will be addressed. When a propagating wave strikes such
a boundary, some of its energy is generally reflected and some is transmitted.
For boundaries with simple shapes, wave functions for the the reflected and
transmitted waves can be calculated by enforcing certain physical laws at
those boundaries, referred to as boundary conditions.
Figure 1.8: Reflection/transmission of a harmonic wave by a boundary at

x = 0 between strings with different mass densities.
Consider, as an example, a harmonic wave propagating on a semi-infinite

string that encounters a boundary with a second semi-infinite string pos-
sessing a different mass density as illustrated in Figure 1.8. The incident,
reflected and transmitted wave functions are:
ψi (x, t) = Ãi ei(ki x−ωt) ψr (x, t) = Ãr ei(−ki x−ωt)

ψt (x, t) = Ãt ei(kt x−ωt)
(1.20)
where ki = ω/vi and kt = ω/vt . The complex amplitude of the incoming wave
is arbitrary. The other two are determined by imposing boundary conditions
on the waves. The first condition enforces continuity at the boundary, mean-
ing that the conjoined strings must stay connected. Because the boundary
is located at x = 0, this condition requires that:
ψ+ (0, t) = ψ− (0, t)
ψt (0, t) = ψi (0, t) + ψr (0, t)
Ãt = Ãi + Ãr
where the incident and reflected waves are added because they interfere to
produce a combined wave on the same string. The second condition enforces
Newton’s third law at the boundary, meaning that the forces exerted by
the strings on one another must be equal. Imposing this condition on the
transverse component of the tension requires:
T sin θ+ = T sin θ−
As noted in the derivation of the wave equation, sin θ ≈ tan θ = ∂ψ/∂x:
∂ψ ∂ψ
T (0, t) = T (0, t)
∂x + ∂x −
T ikt Ãt = T iki Ãi − T iki Ãr
kt Ãt = ki Ãi − ki Ãr
The two equations produced by imposing the boundary conditions can

be used to solve for the complex amplitudes of the reflected and transmitted
waves in terms of the complex amplitude of the incident wave. Substituting
the first equation into the second:
1.8. REFLECTION AND TRANSMISSION 25
kt (Ãi + Ãr ) = ki Ãi − ki Ãr

Ãi (ki − kt ) = Ãr (ki + kt )
Ãr ki − kt
=
Ãi ki + kt
Substituting this result back into the first equation:
ki − kt
Ãt = Ãi + Ãi
ki + kt

ki + kt ki − kt
= Ãi +
ki + kt ki + kt
Ãt 2ki
=
Ãi ki + kt
The impedance for a string is defined as:

p
Impedance = Z = µT (1.21)
p
Using this definition, a substitution can be made for k = ω/v = ω/ T /µ =
Zω/T into the results above for Ãr and Ãt such that they can be written in
a form that is dependent upon the properties of the media alone:
Ãr Zi − Zt
R= = (1.22)
Ãi Zi + Zt
Ãt 2Zi
T = = (1.23)
Ãi Zi + Zt
where R is referred to as the reflection coefficient and T the transmission

coefficient for the boundary.
Notice that for transmission into a very heavy string (µt µi ), R → −1
and T → 0. The incident wave is almost completely reflected and flipped
(corresponding to a 180-degree phase change), and the transmitted wave is
almost nonexistent. In the extreme limit as µt → ∞, the incident string is
unable to move at the boundary and is referred to as fixed. In the opposite
direction (µt decreases), the reflection coefficient decreases and the transmis-
sion coefficient increases. When µt = µi , R = 0 and T = 1, implying that the
transmission is complete and there is no reflection. As the transmitted string
becomes lighter than the incident string, reflection and transmission both in-
crease and the reflected wave no longer undergoes a phase change. When
the transmitted string becomes very light (µt µi ), R → 1 and T → 2.
The incident wave is almost completely reflected and the transmitted wave,
although present, carries almost no energy due to the string’s lack of mass.
In the extreme limit as µt → 0, the incident string is effectively unattached
at the boundary and therefore referred to as free.
1.9 Finite Media

The previous section examined the behavior of waves in semi-infinite media
with a single boundary. This section will address finite media with two
boundaries, each of which could be fixed, free or joined to another medium.
The wave functions for such a medium can be found by applying appropriate
boundary conditions at both ends.
1.9.1 Fixed Strings

As an initial example, consider a string fixed at both ends. In this case,
there is no wave incident upon one of the boundaries that sets the string
in motion. Rather, the string is ’plucked’ to generate an initial wave that
continuously reflects off of the boundaries and interferes with itself. For an
initial harmonic wave, the reflected waves traveling in opposite directions
interfere to produce standing waves. Consequently, boundary conditions are
applied to the standing wave solutions to the wave equation (Equation 1.15)
rather than the traveling wave solutions.
Assume that the boundaries for the string are located at x = 0 and x = L.
Because the force at a fixed boundary is effectively infinite, the boundary con-
dition enforcing Newton’s third law is not applied. The continuity condition
requires ψ = 0 at both boundaries given that the ends of the string do not
move. Applying this condition to Equation 1.15 at x = 0:
1.9. FINITE MEDIA 27
0 = ψ(0, t) = A1 cos 0 cos ωt + A2 cos 0 sin ωt

+ A3 sin 0 cos ωt + A4 sin 0 sin ωt
= A1 cos ωt + A2 sin ωt
This can only be true for all times if A1 = A2 = 0, leaving:
ψ(x, t) = A3 sin kx cos ωt + A4 sin kx sin ωt (1.24)

Applying the continuity condition for the second boundary at x = L:
0 = ψ(L, t) = sin kL(A3 cos ωt + A4 sin ωt)

which can only be true for all times if:
nπ
sin kL = 0 ⇒ kL = nπ ⇒ kn = (1.25)
L
where n is an integer, meaning that the wavenumber has a discrete set of
possible values. The fixed string wave functions are therefore given by:
ψ(x, t) = sin kn x(A3 cos ωt + A4 sin ωt) (1.26)

where each of the standing waves in the discrete set is referred to as a normal
mode.
Because the allowed wavenumbers are discrete, the corresponding wave-
lengths and frequencies are also discrete:
2π 2L
λn = = (1.27)
kn n
and
2πv πv
ωn = 2πfn = = n (1.28)
λn L
where the set of allowed frequencies is commonly referred to as a harmonic
series. Rearranging the top equation, L = (λn /2)n, makes it clear that for
the nth normal mode, exactly n half wavelengths span the string as illustrated
in Figure 1.9. Fundamentally, the set of normal modes is the subset of all
harmonic standing waves that possess nodes at x = 0 and x = L where the
string is required to be motionless.
ψ(x)
n=1
n=2
n=3
n=4
Figure 1.9: The first four normal modes at t − 0 for a string fixed at both
ends.
Of course, multiple normal modes can exist on the same fixed string and
interfere with one another. The combined wave function is given by the sum:
X
ψ(x, t) = sin kn x(A3n cos ωn t + A4n sin ωn t) (1.29)
n
where the individual amplitudes, A3n and A4n , are determined by the initial
state of the string when it is set in motion. Consider, for example, a string
that is plucked such that its initial shape is given by the function f (x) and
its initial velocity by v(x). Applying the first condition:
X
f (x) = ψ(x, 0) = sin kn x(A3n cos 0 + A4n sin 0)
n
X
= A3n sin kn x
n
The values for A3 n can be found by applying a variant of Fourier’s trick. Each
side of the equation is multiplied by a second sine function and integrated
over the length of the string:
Z L X Z L
f (x) sin km xdx = A3n sin kn x sin km xdx
0 n 0
The result of the second integral is 0 unless m = n, in which case it is L/2:

Z L
L
f (x) sin km xdx = A3m
0 2
Recasting the subscript:
Z L
2
A3n = f (x) sin kn xdx (1.30)
L 0
Applying the second condition:
∂ψ X
v(x) = (x, 0) = ωn sin kn x(−A3n sin ωn 0 + A4n cos ωn 0)
∂t n
X
= A4n ωn sin kn x
n
Using Fourier’s trick yields:

Z L
2
A4n = v(x) sin kn xdx (1.31)
Lωn 0
Because the functions f (x) and v(x) are completely general, any wave that is
initiated on the string can be viewed as a combination of interfering harmonic
standing waves.
1.9.2 Free Strings

Next, consider the case where one end of the string segment is fixed and
the other end is free. If the fixed end is located at x = 0, the harmonic
wave functions for the string are given by Equation 1.24 after the continuity
condition has been enforced. Because the other end is free, the continuity
condition does not apply there, but the transverse component of the force
must be zero. Applying this condition at x = L:
∂ψ
0= (L, t) = k cos kL(A3 cos ωt + A4 sin ωt)
∂x
which can only be true for all times if:

1 1 π
cos kL = 0 ⇒ kL = n + π ⇒ kn = n + (1.32)
2 2 L
where n is an integer. The final fixed-free wave functions are therefore given
by Equation 1.26 where A3 and A4 are determined by the initial state of the
string.
The discretized wavelengths and frequencies for the fixed-free string are
given by:
2π 2L
λn = = (1.33)
kn n + 1/2
and

2πv πv 1
ωn = 2πfn = = n+ (1.34)
λn L 2
Solving the top equation for L = (λn /2)n + λn /4 indicates that n half wave-
lengths plus a quarter wavelength span the string for the nth normal mode as
illustrated in Figure 1.10. The set of normal modes for the fixed-free string is
therefore the subset of harmonic standing waves that possess nodes at x = 0
and antinodes at x = L.
For a string segment that is free at both ends, the transverse component
of the force must be zero at both boundaries. Applying this condition to
Equation 1.15 at x = 0 yields:
1.9. FINITE MEDIA
ψ(x) 31
n=1
n=2
n=3
n=4
Figure 1.10: The first four normal modes at t = 0 for a string fixed at one
end and free at the other.
∂ψ
0= (0, t) = −A1 k sin 0 cos ωt − A2 k sin 0 sin ωt
∂x
+ A3 k cos 0 cos ωt + A4 k cos 0 sin ωt
= A3 k cos ωt + A4 k sin ωt
A3 and A4 must therefore be equal to zero, leaving:
ψ(x, t) = cos kx(A1 cos ωt + A2 sin ωt) (1.35)

Applying the same boundary condition at x = L:
∂ψ
0= (L, t) = −k sin kL(A1 cos ωt + A2 sin ωt)
∂x
which follows only if:

π
sin kL = 0 ⇒ kL = nπ ⇒ kn = n (1.36)
L
where n is an integer. The final wave functions are therefore:
ψ(x, t) = cos kn x(A1 cos ωt + A2 sin ωt) (1.37)

where A1 and A2 are determined by the initial state of the string segment.
Because the condition on the wavenumber is identical to that for a fixed
string, the permitted wavelengths and frequencies are the same. However, the
corresponding normal modes are shifted a quarter wavelength, as illustrated
in Figure 1.11, such that antinodes are located at x = 0 and x = L where
the string is free.
n =1
n =2
n =3
n =4
ψ(x)
Figure 1.11: The first four normal modes at t = 0 for a string free at both
ends.
1.9.3 Joined Strings

Finally, consider the case where the string segment is joined to semi-infinite
strings at either end, with a traveling harmonic wave incident from the neg-
ative x direction as illustrated in Figure 1.12. The incident, reflected and
transmitted wave functions are identical to those for a single boundary (see
Equation 1.20). The incident and reflected wave functions within the segment
are given by:
ψis (x, t) = Ãis ei(ks x−ωt) ψrs (x, t) = Ãrs ei(−ks x−ωt)
Figure 1.12: Reflection/transmission of a harmonic wave by a string segment.
Applying the continuity condition at both boundaries yields:
Ãi + Ãr = Ãis + Ãrs and Ãis eiks L + Ãrs e−iks L = Ãt eikt L
Applying Newton’s third law at both boundaries produces:
Ãi ki − Ãr ki = Ãis ks − Ãrs ks and Ãis ks eiks L − Ãrs ks e−iks L = Ãt kt eikt L
Substituting for the wavenumber in terms of the impedance (k = Zω/T ),

these become:
Ãi Zi − Ãr Zi = Ãis Zs − Ãrs Zs and Ãis Zs eiks L − Ãrs Zs e−iks L = Ãt Zt eikt L
The four boundary condition equations can be used to solve for the un-
known complex amplitudes in terms of the incident amplitude. The reflection
and transmission coefficients are given by:
Ãr (1 − Zt /Zi ) cos ks L − i(Zt /Zs − Zs /Zi ) sin ks L

R= =
Ãi (1 + Zt /Zi ) cos ks L − i(Zt /Zs + Zs /Zi ) sin ks L
Ãt 2Zi Zs
T = =
Ãi (Zs Zt + Zi Zs ) cos ks L − i(Zi Zt + Zs2 ) sin ks L
The amplitudes for the waves within the string segment are given by:
Ãis (Zi + Zs ) + (Zs − Zi )R

=
Ãi 2Zs
Ãrs (Zs − Zi ) + (Zi + Zs )R
=
Ãi 2Zs
Because all four equations are complex, there are typically phase differences
between the waves.
1.10 Radiation
Waves are produced by sources, such as a stone striking the surface of a
body of water, an oscillator vibrating a string, or a finger plucking a string
segment. The generation of a wave by a source is defined as radiation. The
corresponding wave function can be determined as long as the behavior of
the source is well-characterized. If it acts at a single point in time, the initial
1.10. RADIATION 35
conditions it produces in the medium can be used to calculate the wave

function as illustrated in the previous section for a fixed string. However, if
it acts continuously, a different approach is required.
One option is to treat the location of the source as a boundary for the
medium, and impose a boundary condition consistent with the behavior of
the source on the general solutions to the wave equation. For example,
consider the case of a harmonic oscillator attached to an infinite string. If
the oscillator is attached at the origin of the coordinate system, the infinite
string is effectively split into two semi-infinite strings, each with a boundary
located at x = 0. The condition on each boundary is that its motion must
match that of the source:
ψ(0, t) = Ãe−iωt
This condition is imposed on the general solutions to the wave equation,
which in this case are the traveling wave solutions:
ψ(x, t) = Ã+ ei(kx−ωt) + Ã− ei(−kx−ωt)

Sources only produce waves that propagate outward from their location, so
Ã− = 0 for the semi-infinite string on the positive side of the origin and
Ã+ = 0 for the semi-infinite string on the negative side, leaving:
ψ+ (x, t) = Ã+ ei(kx−ωt) and ψ− (x, t) = Ã− ei(−kx−ωt)

where ψ+ (x, t) and ψ− (x, t) are the wave functions for the strings on the
positive and negative sides of the origin respectively. Applying the boundary
condition to both yields:
Ãe−iωt = ψ± (0, t) = Ã± e−iωt

Ã = Ã±
Consequently, the source produces an outward propagating harmonic wave

on each side with complex amplitude Ã. The two wave functions can be
combined into a single expression using an absolute value operator:
ψ(x, t) = Ãei(k|x|−ωt) (1.38)

A second option is to adapt the wave equation to account for the presence
of the source and solve it. Referring back to Figure 1.7, Newton’s second law
dictated that the equation of motion for a string element was given by:
∂ 2ψ
T sin θx+dx − T sin θx = (µdx)
∂t2
If there is an additional force acting on the element to produce radiation, the
equation of motion becomes:
∂ 2ψ
T sin θx+dx − T sin θx + F (x, t)dx = (µdx)
∂t2
where F (x, t) is the force per unit length (force density) applied to the el-
ement. Following the same steps used to derive the wave equation, this
becomes:
∂ 2ψ 1 ∂ 2ψ −F (x, t)
2
− 2 2
= (1.39)
∂x v ∂t T
Specifically for the case of a harmonic oscillator attached to the string at the
origin:
∂ 2ψ 1 ∂ 2ψ −F̃
2
− 2 2 = δ(x)e−iωt (1.40)
∂x v ∂t T
where F̃ is the complex amplitude of the force applied by the oscillator
and the delta function specifies that it is located at the origin. From here,
various techniques can be used to solve the equation, but Equation 1.38 can
be verified as the solution.
Non-harmonic sources are more complicated, but can be handled by tak-
ing advantage of the Principle of Superposition. Just as any waveform can
be written as a sum of harmonic waveforms as in Equation 1.16, the time
dependence of any source f (t) can be represented as a sum of the time de-
pendencies for harmonic sources:
Z ∞
f (t) = F̃ (ω)e−iωt dω (1.41)
−∞
where F̃ (ω) is the Fourier transform of f (t):

Z ∞
1
F̃ (ω) = f (t)eiωt dt (1.42)
2π −∞
1.11. VECTOR WAVES AND POLARIZATION 37
A non-harmonic source can therefore be treated as a combination of harmonic

sources whose complex amplitudes are determined by its Fourier transform.
The Principle of Superposition then dictates that the wave function produced
by the combination is the sum/integral of the wave functions generated by
the individual harmonic constituents. For a string driven by a non-harmonic
source located at the origin, for example, the wave function produced would
be given by:
Z ∞
ψ(x, t) = F̃ (ω)ei(k|x|−ωt) dω (1.43)
−∞
1.11 Vector Waves and Polarization

This final section of the chapter addresses the behavior of one-dimensional
vector waves, which have been ignored until now to simplify the introduction
of new concepts. Beginning with the wave equation, the version for a vector
quantity is identical to that for a scalar quantity:
∂ 2ψ 1 ∂ 2ψ
= (1.44)
∂x2 v 2 ∂t2
Writing ψ in terms of its rectangular components yields:
∂2 1 ∂2
(ψx x̂ + ψy ŷ + ψz ẑ) = (ψx x̂ + ψy ŷ + ψz ẑ)
∂x2 v 2 ∂t2
∂ 2 ψx ∂ 2 ψy ∂ 2 ψz 1 ∂ 2 ψx 1 ∂ 2 ψy 1 ∂ 2 ψz
x̂ + ŷ + ẑ = x̂ + ŷ + ẑ
∂x2 ∂x2 ∂x2 v 2 ∂t2 v 2 ∂t2 v 2 ∂t2
which can be broken up into three separate equations, one for each unit
vector:
∂ 2 ψn 1 ∂ 2 ψn
= n = x, y, z
∂x2 v 2 ∂t2
Each of these is a scalar wave equation that is independent of the other
two, meaning that the methods of the previous sections can be employed to
find the wave functions for each of the components independently, and then
recombined to form a vector afterward.
Vector waves can possess all three components, but it is not a requirement.
A transverse wave, for example, is a wave for which the propagating vector
quantity is perpendicular to the propagation direction. Such waves can have z

and/or y components, but no x component. A displacement wave on a string
is a perfect example. A longitudinal wave, on the other hand, is a wave for
which the propagating vector is parallel to the propagation direction. These
waves only possess an x component. A common example is a displacement
wave on a coiled spring, where the individual coils move move forward and
backward along the length of the spring creating regions of compression and
rarefaction as illustrated in Figure 1.13.
Figure 1.13: A snapshot of a longitudinal wave propagating along coiled

spring.
To this point in the chapter, string displacement has been restricted to

a single plane, making it a scalar quantity. Without that constraint, the
displacement becomes a vector quantity with both z and y components. For
harmonic waves traveling in the positive x direction, the independent wave
1.11. VECTOR WAVES AND POLARIZATION 39
functions for the two components are given by:
ψy = Ay ei(kx−ωt+φy ) and ψz = Az ei(kx−ωt+φz )

How the two components sum to produce a vector wave for different combi-
nations of amplitude and phase is referred to as the wave’s polarization.
The simplest possible case is when the phases are equal. The combined
wave function becomes:
ψ(x, t) = (Ay ŷ + Az ẑ)ei(kx−ωt+φ)

Consider the motion of the string at the specific location x = 0. Taking the
real part of the equation above leaves:
ψ(x, t) = (Ay ŷ + Az ẑ) cos (ωt − φ)

This is simply a vector at angle tan−1p (Az /Ay ) with respect to the y-axis
whose length oscillates between A = A2y + A2z and −A. The motion is
therefore identical to that of a string constrained to a plane, but tilted at an
angle. Because the movement occurs along a single line, it is referred to as
linear polarization.
The motion becomes more complicated if the phases are not equal. If
they differ by ninety degrees but the amplitudes are equal, the combined
wave function becomes:
ψ(x, t) = Aei(kx−ωt+φ) ŷ + Aei(kx−ωt+φ±π/2) ẑ

Taking the real part at x = 0 leaves:
ψ(x, t) = A cos (ωt − φ)ŷ + A cos (ωt − φ ∓ π/2)ẑ

ψ(x, t) = A cos (ωt − φ)ŷ ± A sin (ωt − φ)ẑ
This vector has a constant length A and traces out a circle in the yz plane
with angular velocity ±ω as illustrated in Figure 1.14. For this reason, the
motion is referred to as circular polarization. As x increases, the string
motion continues to be circular but is increasingly out of phase with the
motion at x = 0. Consequently, the movement of the entire string looks like
a rotating corkscrew.
If the phases differ by ninety degrees but the amplitudes are unequal, the
shape traced by the vector goes from being a circle to an ellipse whose major
Figure 1.14: The motion of a string at x = 0 for a circularly polarized wave.
and minor axes are determined by the values of Ax and Ay . This type of
motion is therefore referred to as elliptical polarization. If the phases differ
by an angle other than ninety degrees, the ellipse is tilted.
Chapter 2
Three-dimensional Wave
Functions
The initial step in extending the concepts introduced in the first chapter to
three dimensions is to find the harmonic solutions to the three-dimensional
wave equation by applying the separation of variables technique. The bulk of
this chapter is devoted to deriving those solutions in the rectangular, spher-
ical and cylindrical coordinate systems. The remainder focuses on energy
transmission by acoustic waves as the primary example of three-dimensional
scalar waves, and electromagnetic waves as the primary example of a three-
dimensional vector waves.
2.1 Three-dimensional Wave Equations

The three-dimensional wave equation is an intuitive extension of the one-
dimensional equation. The scalar form in rectangular coordinates is given
by:
∂ 2ψ ∂ 2ψ ∂ 2ψ 1 ∂ 2ψ
∇2 ψ(x, y, z, t) = + + = (2.1)
∂x2 ∂y 2 ∂z 2 v 2 ∂t2
where ∇2 is referred to as the Laplacian operator and is the divergence of
the gradient operator:
∂2 ∂2 ∂2
∇2 = ∇ · ∇ = + +
∂x2 ∂y 2 ∂z 2
41
42 CHAPTER 2. THREE-DIMENSIONAL WAVE FUNCTIONS
where:
∂ ∂ ∂
∇= x̂ + ŷ + ẑ
∂x ∂y ∂z
For an acoustic wave, ψ represents the pressure in a material medium. If the
compressibility of the medium is κ, defined as the inverse of the bulk modulus
B, and the mass density is ρ, then:
r s
1 B
v= = (2.2)
κρ ρ
The three-dimensional vector wave equation is identical:
1 ∂ 2ψ
∇2 ψ = (2.3)
v 2 ∂t2
For an electromagnetic wave:
1
v=√ (2.4)
µ
where is the permittivity and µ the permeability of the medium through
which the wave propagates. The electric (E) and magnetic (B) field vectors
that comprise the wave separately satisfy the wave equation. However, the
wave equations for the fields are less restrictive than Maxwell’s equations
from which they are derived:
∇ · E = 0 Gauss’s Law (sourceless) (2.5)

∇·B =0 (2.6)
∂B
∇×E =− Faraday’s Law (2.7)
∂t
∂E
∇ × B = µ Ampere’s Law (sourceless) (2.8)
∂t
The solutions must therefore meet additional criteria to be valid. In partic-
ular, Equations 2.5 and 2.6 specify that both fields must be divergenceless,
while Equations 2.7 and 2.8 indicate that they are not independent, i.e. the
solution to one of the fields determines the solution to the other.
2.1. THREE-DIMENSIONAL WAVE EQUATIONS 43
Imposing these additional conditions can be difficult, so in some instances

it is easier to introduce and solve for potentials from which the vector quan-
tities can be obtained. Because the magnetic field is divergenceless, it can be
written as the curl of another vector field which is referred to as the vector
potential :
B =∇×A (2.9)
Plugging the vector potential into Faraday’s law yields:
∂
∇ × E = − (∇ × A)
∂t
∂A
∇× E+ =0
∂t
Because the curl of the term in parentheses is zero, it can be written as the
gradient of a scalar field, which is named the scalar potential :
∂A
−∇V = E +
∂t
∂A
E = −∇V − (2.10)
∂t
Plugging Equations 2.10 and 2.9 into Gauss’s and Faraday’s laws:
∂
∇ · E = −∇2 V − (∇ · A) = 0
∂t
and
∂V ∂ 2A
∇ × B = ∇ × ∇ × A = −µ∇ − µ 2
∂t ∂t
or using the vector identity ∇ × ∇ × A = ∇(∇ · A) − ∇2 A:
∂V ∂ 2A
−∇2 A + ∇(∇ · A) = −µ∇ − µ 2
∂t ∂t
2

∂ A ∂V
∇2 A − µ 2 − ∇ ∇ · A + µ =0
∂t ∂t
Solving these two equations for the potentials and plugging the results into
Equations 2.10 and 2.9 produces the same results for E and B as solving
Maxwell’s equations directly. They are not simple equations in their current
form, but because the electric and magnetic fields are calculated by taking
derivatives of the potentials, there is some freedom in setting the values of
V and A without effecting the results for E and B. This freedom is referred
to as gauge freedom, and picking a particular form for V and A is referred
to as setting the gauge.
The Lorentz Gauge sets V and A such that:
∂V
∇ · A = −µ (2.11)
∂t
Plugging this additional condition into the two equations above yields:
∂ 2V ∂ 2A
∇2 V = µ and ∇ 2
A = µ (2.12)
∂t2 ∂t2
which are clearly wave equations. Consequently, finding appropriate wave
functions for the scalar and vector potentials and plugging the results into
Equations 2.10 and 2.9 is an equivalent alternative to solving the wave equa-
tion for the electric and magnetic fields directly and imposing any additional
conditions required by Maxwell’s equations.
2.2 Separable Solutions

Solutions to the three-dimensional wave equation can be found by employing
the separation of variables technique introduced in the first chapter. The first
step is to separate the time dependence of the solutions from their spatial
dependence. Beginning with scalar wave functions and assuming that they
can be written as a product of time and space dependent functions:
ψ(r, t) = ψ(r)T (t)

Plugging this expression into the wave equation:
1 d2 T (t)
T (t)∇2 ψ(r) = ψ(r)
v2 dt2
Dividing through by ψ(r, t):
2.2. SEPARABLE SOLUTIONS 45
v2 1 d2 T (t)
∇2 ψ(r) =
ψ(r) T (t) dt2
Choosing −ω 2 as the separation constant:
d2 T (t) 2 2 ω2
+ ω T (t) = 0 and ∇ ψ(r) + 2 ψ(r) = 0
dt2 v
The solutions to the first equation, which were introduced in the first chapter,
are:
T (t) = a cos ωt + b sin ωt Standing Wave Solutions

or
T (t) = ãeiωt + b̃e−iωt Traveling Wave Solutions
As in one dimension, ã is generally set to zero to avoid redundant terms in

the traveling wave functions. Substituting k = ω/v, the spatial equation
becomes:
∇2 ψ(r) + k 2 ψ(r) = 0 (2.13)

which is called the Helmholtz equation.
2.2.1 Rectangular Wave Functions

The next step in the process is to assume ψ(r) can be written as a prod-
uct of functions of a single spatial variable, which vary depending upon the
coordinate system being used. Beginning with the rectangular coordinate
system:
ψ(r) = X(x)Y (y)Z(z)

Plugging this expression into the Helmholtz equation yields:
d2 X d2 Y d2 Z
YZ + XZ + XY + k 2 XY Z = 0
dx2 dy 2 dz 2
Dividing through by ψ(r):
1 d2 X 1 d2 Y 1 d2 Z
+ + = −k 2
X dx2 Y dy 2 Z dz 2
Because each of the terms is dependent upon a single spatial variable, the
equation can only hold true for all values of x, y and z if each term is
a constant. Choosing −kn2 as the separation constant for the nth spatial
variable:
d2 X d2 Y d2 Z
2
+ kx2 X = 0 2
+ ky2 Y = 0 2
+ kz2 Z = 0
dx dy dz
where:
kx2 + ky2 + kz2 = k 2 (2.14)

The solutions are given by:
X(x) = c cos kx x + d sin kx x

Y (y) = f cos ky y + g sin ky y Standing Wave Solutions
Z(z) = h cos kz z + l sin kz z
or
˜ −ikx x
X(x) = c̃eikx x + de
Y (y) = f˜eiky y + g̃e−iky y Traveling Wave Solutions
Z(z) = h̃eikz z + ˜le−ikz z
where, as in one dimension, the standing wave solutions are linear combina-
tions of the traveling wave solutions and therefore not independent.
Consider the traveling wave solutions for a simple case where kx = ky = 0.
The wave function becomes:
ψ(r, t) = XY ZT = Ãei(±kx x−ωt)

which, as in one dimension, represents a harmonic wave propagating along
the x axis with wavenumber kx = k = ω/v. In this particular case, the wave
function does not vary in the y or z directions. Consequently, for any plane
perpendicular to the x axis, ψ oscillates in unison at all points on the plane.
Such a surface is called a wavefront, and this type of wave is referred to as a
plane wave because its wavefronts are planar.
For the more general case:
ψ(r, t) = Ãei(±kx x±ky y±kz z−ωt)

which can be simplified by introducing what is referred to as the wavevector :
k = ±kx x̂ ± ky ŷ ± kz ẑ (2.15)
where
q ω
|k| = kx2 + ky2 + kz2 = k = (2.16)
v
from Equation 2.14. Inserting the wavevector into the wave function yields:
ψ(r, t) = Ãei(k·r−ωt) = Ãei(krk −ωt) (2.17)

where rk is the projection of r onto k, or the length/component of r along
the wavevector direction (see Figure 2.1). This wave function therefore repre-
sents a harmonic wave propagating in the direction of increasing values of rk
along the wavevector. Given that all points on a plane perpendicular to the
wavevector have the same value of rk and therefore oscillate in unison, the
wavefronts for these waves are also planar and perpendicular to the direction
of propagation.
As in one dimension, any possible three-dimensional wave function can be
written as a sum of harmonic plane wave functions using a three-dimensional
Fourier transform. Any three-dimensional wave can therefore be viewed as a
combination of interfering harmonic plane waves.
Also as in one dimension, the vector wave equation can be split into three
independent scalar wave equations, one for each vector component. When
solved using the separation of variables technique in rectangular coordinates,
the resulting harmonic plane waves can have different wavevectors for each
component. However, most physical vector waves must satisfy additional
conditions that require them to be identical. Under that assumption, the
traveling wave solutions are given by:
ψ(r, t) = Ãx ei(k·r−ωt) x̂ + Ãy ei(k·r−ωt) ŷ + Ãz ei(k·r−ωt) ẑ = Ãei(k·r−ωt) (2.18)

If there are no restrictions on the amplitudes or phases of the components of
Ã, the waves can be longitudinal, transverse (with linear, circular or elliptical
polarization), or a combination of the two.
Figure 2.1: Visualization of a plane wave propagating in the direction of the

wavevector.
For electromagnetic fields, additional conditions are placed on the solu-

tions by Maxwell’s equations. Applying Gauss’ law to the electric field:
0 = ∇ · E = ∇ · Ãei(k·r−ωt)
= iÃx kx ei(k·r−ωt) + iÃy ky ei(k·r−ωt) + iÃz kz ei(k·r−ωt)
= Ãx kx + Ãy ky + Ãz kz
= Ã · k
This result implies that the electric field must be perpendicular to the direc-
tion of propagation, meaning that the wave must be transverse. Because the
magnetic field is also divergenceless, it must also be a transverse wave.
For all harmonic waves, where the time dependence is given by e−iωt ,
Faraday’s law and Ampere’s law become:
ω
∇ × E = iωB and ∇ × B = −i (2.19)
v2
Applying Faraday’s law:
i i
B = − ∇ × E = − ∇ × Ãei(k·r−ωt)
ω ω
i
= − [(iky Ez − ikz Ey )x̂ + (ikz Ex − ikx Ez )ŷ + (ikx Ey − iky Ex )ẑ]
ω
i
= − (ik × E)
ω
k
= k̂ × E
ω
1
= k̂ × E (2.20)
v
This result requires the electric and magnetic fields to be perpendicular to
one another, and their amplitudes to differ by a multiplicative constant equal
to the propagation speed (see Figure 2.2). Applying Ampere’s law yields the
same result.
2.2.2 Spherical Wave Functions

Unlike in one dimension, there are alternative coordinate systems in three
dimensions that also possess separable solutions to the wave equation. One
example is the spherical coordinate system, which is illustrated in Figure 2.3
along with the transformations to/from the rectangular coordinate system.
These transformations, along with the chain rule, can be applied to rewrite
the Laplacian operator in terms of spherical coordinates:
1 ∂2

21 ∂ 2 ∂ 1 ∂ ∂
∇ = 2 r + sin θ + (2.21)
r ∂r ∂r sin θ ∂θ ∂θ sin2 θ ∂φ2
Inserting this result into the scalar wave equation produces:
1 ∂ 2ψ 1 ∂ 2ψ

1 ∂ 2 ∂ψ 1 ∂ ∂ψ
r + sin θ + =
r2 ∂r ∂r sin θ ∂θ ∂θ sin2 θ ∂φ2 v 2 ∂t2
Figure 2.2: Illustration of the relationship between the electric and magnetic
fields for an electromagnetic plane wave.
which can be solved using separation of variables.

The first step is to separate the time dependence, which yields the Helm-
holtz equation written in terms of spherical coordinates:
1 ∂ 2 ψ(r)

1 ∂ 2 ∂ψ(r) 1 ∂ ∂ψ(r)
r + sin θ + + k 2 ψ(r) = 0
r2 ∂r ∂r sin θ ∂θ ∂θ sin2 θ ∂φ2
The next step is to assume a separable solution of the form:
ψ(r) = R(r)Θ(θ)Φ(φ)
Plugging this expression into the Helmholtz equation and dividing by ψ pro-
duces:
Figure 2.3: The spherical coordinate system, consisting of a radial coordinate

r, a zenith angle θ, and an azimuthal angle φ. The corresponding unit vectors
are also shown.
d2 Φ

1 1 d 2 dR 1 d dΘ 1
r + sin θ + + k2 = 0
r2 R dr dr Θ sin θ dθ dθ Φ sin2 θ dφ2
Multiplying through by r2 sin2 θ and isolating the φ-dependent term:
1 d2 Φ

2 1 d 2 dR 1 d dΘ 2 2
sin θ r + sin θ +k r =−
R dr dr Θ sin θ dθ dθ Φ dφ2
Choosing m2 as the separation constant:

d2 Φ
+ m2 Φ = 0 (2.22)
dφ2
and

2 1 d 2 dR 1 d dΘ
sin θ r + sin θ + k r = m2
2 2
R dr dr Θ sin θ dθ dθ
Dividing by sin2 θ and isolating the r- and φ-dependent terms:
m2

1 d 2 dR 2 2 1 d dΘ
r +k r = − sin θ
R dr dr sin2 θ Θ sin θ dθ dθ
Choosing l(l + 1) as the separation constant:
m2

1 d dΘ
sin θ + l(l + 1) − Θ=0 (2.23)
sin θ dθ dθ sin2 θ
and

1 d dR
r2 + k 2 r2 = l(l + 1)
R dr dr
The second equation can be simplified:

d 2 dR
r + [k 2 r2 − l(l + 1)]R = 0
dr dr
dR d2 R
2r + r2 2 + [k 2 r2 − l(l + 1)]R = 0
dr dr
d2 R 2 dR

2 l(l + 1)
+ + k − R=0 (2.24)
dr2 r dr r2
The solutions to Equation 2.22 are typical:
Φ(φ) = c cos mφ + d sin mφ Standing Wave Solutions

or
˜ −imφ Traveling Wave Solutions
Φ(φ) = c̃eimφ + de
However, m must be an integer for Φ to be a single-valued function, i.e. Φ

must have the same value at φ and φ + 2π. The solutions to Equation 2.23
are a series of special functions called Associated Legendre Polynomials:
|m|
Θ(θ) = Pl (cos θ)
where l must be an integer for Θ not to explode at 0 and π, and |m| must be
less than or equal to l for Θ to be non-zero. The first few are given in Table
2.1 and plotted in Figure 2.4.
P00 (cos θ) = 1
P10 (cos θ) = cos θ
√
P11 (cos θ) = 1 − cos2 θ = sin θ
P20 (cos θ) = 3 cos2 θ − 1
√
P21 (cos θ) = cos θ 1 − cos2 θ = cos θ sin θ
P22 (cos θ) = 1 − cos2 θ = sin2 θ
P30 (cos θ) = 5 cos3 θ − 3 cos θ
√
P31 (cos θ) = (5 cos2 θ − 1) 1 − cos2 θ = (5 cos2 θ − 1) sin θ
P32 (cos θ) = cos θ(1 − cos2 θ) = cos θ sin2 θ
P33 (cos θ) = (1 − cos2 θ)3/2 = sin3 θ
Table 2.1: Associated Legendre Polynomials to l = 3.
When multiplied together, the solutions for Φ and Θ provide the full angular
dependence of the separable wave functions and are referred to as Spherical
Harmonics:
|m|
Ylm = CΘ(θ)Φ(φ) = CPl (cos θ)e±imφ
where C is a constant that normalizes the integral of |Ylm |2 over all angles.
The first few are listed in Table 2.2.
Finally, Equation 2.24 has both standing wave and traveling wave solu-
tions. The standing wave solutions are a combination of special functions
called spherical Bessel functions:
1
P1m(cos θ)
0
−1
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
2
P2m(cos θ)
−1
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
2
P3m(cos θ)
−2
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
cos θ
Figure 2.4: Associated Legendre Polynomials from l = 1 to 3. Solid lines are

for m = 0, dashed lines for m = 1, dotted lines for m = 2, and dashed-dotted
lines for m = 3.
R(r) = f jl (kr) + gnl (kr) Standing Wave Solutions
where the jl are spherical Bessel functions of the first kind, and the nl are
spherical Bessel functions of the second kind or spherical Neumann functions.
Both are combinations of decaying sinusoids. The first few are provided in
Table 2.3 and plotted in Figure 2.5. Notice that the spherical Neumann
functions explode at the origin and are therefore not viable solutions at that
location. The traveling wave solutions are a combination of spherical Hankel
functions:
r
1
Y00 =
4π
r
3
Y10 = cos θ
4π
r
3
Y1±1 = ∓ sin θe±iφ
8π
r
0 5
Y2 = (3 cos2 θ − 1)
16π
r
15
Y2±1 = ∓ cos θ sin θe±iφ
8π
r
15
Y2±2 = sin2 θe±2iφ
32π
r
0 7
Y3 = (5 cos3 θ − 3 cos θ)
16π
r
±1 21
Y3 = ∓ (5 cos2 θ − 1) sin θe±iφ
64π
r
105
Y3±2 = cos θ sin2 θe±2iφ
32π
r
35
Y3±3 = ∓ sin3 θe±3iφ
64π
Table 2.2: Spherical Harmonics to l = 3.
(1) (2)
R(r) = f˜hl (kr) + g̃hl (kr) Traveling Wave Solutions
where
(1) (2)
hl (kr) = jl (kr) + inl (kr) and hl (kr) = jl (kr) − inl (kr)
(1) (2)
and the hl are spherical Hankel functions of the first kind and the hl are
spherical Hankel functions of the second kind. Because the spherical Bessel
functions are combinations of sinusoids, the spherical Hankel functions are
combinations of complex exponentials. For example:
sin kr
j0 (kr) =
kr
cos kr
n0 (kr) = −
kr
sin kr cos kr
j1 (kr) = −
(kr)2 kr
cos kr sin kr
n1 (kr) = − −
(kr)2 kr

3 sin kr cos kr
j2 (kr) = 2
−1 −3
(kr) kr (kr)2

3 cos kr sin kr
n2 (kr) = − 2
−1 −3
(kr) kr (kr)2

15 6 sin kr 15 cos kr
j3 (kr) = − − − 1
(kr)3 kr kr (kr)2 kr

15 6 cos kr 15 sin kr
n3 (kr) = − 3
− − 2
−1
(kr) kr kr (kr) kr
Table 2.3: Spherical Bessel functions to l = 3.
(1) sin kr cos kr eikr

h0 = −i = −i
kr kr kr
−ikr
(2) sin kr cos kr e
h0 = +i =i
kr kr kr
The full solutions to the wave equation are given by the product of the
terms above along with the solutions to the time-dependent equation:
ψ(r, θ, φ, t) = R(r)Θ(θ)Φ(φ)T (t)

Like the rectangular solutions, the spherical solutions can be combined by
integrating over k and summing over l and m to produce any legitimate
three-dimensional wave function.
Consider the traveling wave solutions for l = m = 0:
±ikr
e ˜ e−iωt = Ã ei(±kr−ωt)

ψ(r, t) = ∓ (1)(c̃ + d)
kr r
1.5
l =0
1 l =1
l =2
jl (kr)
0.5 l =3
−0.5
0 5 10 15
0.5
0
nl (kr)
−0.5
−1
−1.5
0 5 10 15
kr
Figure 2.5: Spherical Bessel functions to l = 3.
where all of the multiplicative constants have been absorbed into Ã. This
function represents a harmonic wave traveling in the direction of increasing/
decreasing values of r, i.e. propagating away from/towards the origin. Be-
cause all points on a spherical shell surrounding the origin have the same
value of r and therefore oscillate in unison, the wavefronts are spherical.
Consequently, waves of this type are referred to as spherical waves. As for
plane waves, the wavefronts are perpendicular to the direction of propagation
(see Figure 2.6). Unlike plane waves, the amplitude decreases with distance
from the origin by a factor of r.
If the standing wave solutions are used for Φ, the wave functions for l > 0
take the form:
(1) (2) |m|

ψ(r, t) = [f˜hl (kr) + g̃hl (kr)]Pl (cos θ)(c cos mφ + d sin mφ) e−iωt

Figure 2.6: Visualization of a spherical wave propagating away from the

origin.
Because the spherical Hankel functions are combinations of complex expo-

nentials, these wave functions all contain terms of the form ei(±kr−ωt) and are
therefore spherical waves. At large distances from the origin (kr 1 ⇒ r
λ), the higher order 1/(kr) terms in the spherical Hankel functions drop out
such that:
(1) (−i)l+1 ikr (2) il+1 −ikr

hl≈ e hl ≈ e (2.25)
kr kr
Absorbing the angular dependence and complex constants into a single term,
the wave functions become:
Ã(θ, φ) i(±kr−ωt)
ψ(r, t) = e (2.26)
r
where the only difference from the case where l = 0 is that the amplitude
possesses angular dependence.
Because the rectangular components of a vector wave independently solve
the wave equation, the separable solution in spherical coordinates for each is
the same as for a scalar. Assuming that the wavenumbers are identical, the
traveling wave solutions take the form of Equation 2.26 at large distances
from the origin with different amplitudes for each component. Pulling the
common terms leaves an expression for a simple spherical wave:
1
ψ(r, t) = [A˜x (θ, φ)x̂ + Ãy (θ, φ)ŷ + Ãz (θ, φ)ẑ] ei(±kr−ωt)
r
= e (2.27)
r
For electromagnetic waves, Maxwell’s equations require the electric and
magnetic fields to be divergenceless. Rigorously enforcing this additional
condition on the general spherical solutions is complicated and beyond the
level of this book. However, applying Guass’ law in spherical coordinates to
Equation 2.27 specifically:
" #
1 ∂ Ã r (θ, φ)
0=∇·E = 2 r2 ei(±kr−ωt)
r ∂r r
" # " #
1 ∂ Ãθ (θ, φ) i(±kr−ωt) 1 ∂ Ãφ (θ, φ) i(±kr−ωt)
+ sin θ e + e
r sin θ ∂θ r r sin θ ∂φ r
The last two terms are proportional to r−2 and can be dropped for large
values of r. Differentiating the first term produces:
Ãr (θ, φ) ±ikr

0 = e−iωt ± ikre±ikr

2
e
r
Dropping the first term, which is proportional to r−2 , yields:
Ãr (θ, φ) i(±kr−ωt)

0 = ±ik e
r
This equation holds only if Ãr = 0, i.e. the electric field has no radial com-
ponent. The electric field must therefore be perpendicular to the direction of
propagation. Because the magnetic field is also divergenceless, it must also be

perpendicular to the propagation direction. Consequently, electromagnetic
spherical waves must be transverse like plane waves.
The electric and magnetic fields can be related by applying Equation 2.19.
Using the product rule ∇ × (ba) = ∇b × a + b∇ × a:
−i
B= ∇×E
ω ±ikr
e±ikr

−i −iωt e
= e ∇ × Ã(θ, φ) + ∇ × Ã(θ, φ)
ω r r
The curl is proportional to r−1 , which means that the term on the right is
proportional to r−2 and can be dropped. Calculating the gradient on the left
yields:
−i −iωt ∂ e±ikr

B= e r̂ × Ã(θ, φ)
ω ∂r r
e±ikr e±ikr

−i −iωt
= e ±ik − 2 r̂ × Ã(θ, φ)
ω r r
Dropping the second term, which is proportional to r−2 leaves:
k ei(±kr−ωt)
B = ± r̂ × Ã(θ, φ)
ω r
k
= ± r̂ × E
ω
1
= k̂ × E
v
which is identical to the relationship between the electric and magnetic fields
for plane waves, and requires them to be perpendicular to one another.
2.2.3 Cylindrical Wave Functions

The cylindrical coordinate system is the other primary example of an alterna-
tive three-dimensional coordinate system. Figure 2.7 displays the coordinates
Figure 2.7: The cylindrical coordinate system, consisting of a radial coordi-

nate r, an azimuthal angle φ, and an elevation z. The corresponding unit
vectors are also shown.
along with transformations to/from the rectangular system. These transfor-

mations, along with the chain rule, can be applied to rewrite the Laplacian
operator in terms of cylindrical coordinates:
∂2 1 ∂ 1 ∂2 ∂2
∇2 = + + + (2.28)
∂r2 r ∂r r2 ∂φ2 ∂z 2
Inserting this result into the scalar wave equation yields:
∂ 2 ψ 1 ∂ψ 1 ∂ 2ψ ∂ 2ψ 1 ∂ 2ψ
+ + + =
∂r2 r ∂r r2 ∂φ2 ∂z 2 v 2 ∂t2
which can be solved using separation of variables.
Separating the time dependence first yields the Helmholtz equation in

terms of cylindrical coordinates:
∂ 2 ψ(r) 1 ∂ψ(r) 1 ∂ 2 ψ(r) ∂ 2 ψ(r)

+ + + + k 2 ψ(r) = 0
∂r2 r ∂r r2 ∂φ2 ∂z 2
Inserting a separable solution of the form:
ψ(r) = R(r)Φ(φ)Z(z)
dividing by ψ, and isolating the z-dependent term produces:
1 d2 R 1 dR 1 d2 Φ 2 1 d2 Z
+ + + k = −
R dr2 rR dr r2 Φ dφ2 Z dz 2
Choosing kz2 as the separation constant:
d2 Z
+ kz2 Z = 0 (2.29)
dz 2
and
2
1dR 1 dR 1 d2 Φ
+ + + k 2 − kz2 = 0
R dr2 rR dr r2 Φ dφ2
The bottom equation can be simplified by introducing a wavevector:
k = kr r̂ + kz ẑ (2.30)
where
p ω
|k| = kr2 + kz2 = k = (2.31)
v
such that kr2 = k 2 − kz2 and:
1 d2 R 1 dR 1 d2 Φ
+ + + kr2 = 0
R dr2 rR dr r2 Φ dφ2
Multiplying through by r2 and isolating the φ-dependent term:
r2 d2 R r dR 2 2 1 d2 Φ
+ + r kr = −
R dr2 R dr Φ dφ2
Choosing the separation constant m2 yields:
d2 Φ
+ m2 Φ = 0 (2.32)
dφ2
and
2
m2

d R 1 dR 2
+ + kr − 2 R = 0 (2.33)
dr2 r dr r
The solutions to Equation 2.29 are identical to those for the rectangular
z coordinate:
Z(z) = h cos kz z + l sin kz z Standing Wave Solutions

or
Z(z) = h̃eikz z + ˜le−ikz z Traveling Wave Solutions
The solutions to Equation 2.32 are identical to those for the spherical φ
coordinate:
Φ(φ) = c cos mφ + d sin mφ Standing Wave Solutions

or
˜ −imφ Traveling Wave Solutions
Φ(φ) = c̃eimφ + de
where m must be an integer.
Equation 2.33 is comparable to Equation 2.24 and has similar solutions.
The standing wave solutions are called Bessel functions:
R(r) = f Jm (kr r) + gNm (kr r) Standing Wave Solutions

where the Jm are Bessel functions of the first kind and the Nm are Bessel
functions of the second kind or Neumann functions. Both are expressed as
an infinite series of polynomials. The first few of each type are plotted in
Figure 2.8. Notice that the functions look like decaying sinusoids, and at
large distaces from the origin (kr r 1 ⇒ r λ):
r
2 1 π
Jm (kr r) ≈ cos kr r − m +
πkr r 2 2
r (2.34)
2 1 π
Nm (kr r) ≈ sin kr r − m +
πkr r 2 2
Like their spherical counterparts, the Neumann functions explode at the

origin and are not viable solutions at that location. The traveling wave
solutions are a combination of Hankel functions:
1.5
m =0
1 m =1
=2
J m(kr r)
m
0.5 m =3
−0.5
0 5 10 15
0.5
0
Nm (kr r)
−0.5
−1
−1.5
0 5 10 15
kr r
Figure 2.8: Bessel functions to m = 3.
R(r) = f˜Hm
(1) (2)
(kr r) + g̃Hm (kr r) Traveling Wave Solutions
where
(1) (2)
Hm (kr r) = Jm (kr r) + iNm (kr r) and Hm (kr r) = Jm (kr r) − iNm (kr r)
(1) (2)
and the Hm are Hankel functions of the first kind and the Hm are Han-
kel functions of the second kind. Substituting Equation 2.34 for the Bessel
functions in their definitions indicates that they are decaying complex expo-
nentials at large distances from the origin:
r
(1) 2 i[kr r−(m+ 12 )( π2 )]
Hm (kr r) ≈ e
πkr r
r (2.35)
(2) 2 −i[kr r−(m+ 12 )( π2 )]
Hm (kr r) ≈ e
πkr r
The full solutions to the wave equation are given by the product of the
terms above along with the solutions to the time-dependent equation:
ψ(r, φ, z, t) = R(r)Φ(φ)Z(z)T (t)

Like the rectangular and spherical solutions, the cylindrical solutions can
be combined by integrating over k and summing over m to produce any
legitimate three-dimensional wave function. As for the spherical solutions,
the separable solution in cylindrical coordinates for each of the rectangular
components of a vector wave is the same as for a scalar.
2.3 Energy Transmission

In three dimensions, the energy transmitted by a wave is spread over a surface
perpendicular to the direction of propagation. Consequently, the fundamen-
tal measure of energy transmission is instantaneous intensity, or the power
transmitted per unit area. It is a vector quantity that points in the direction
of energy flow/propagation and is represented by S. Average intensity, or
simply intensity, is the time-averaged power transmitted per unit area. It
is represented by I and is calculated for harmonic waves by averaging the
instantaneous intensity over one period:
1 T
Z
I= Sdt (2.36)
T 0
Just as the equation derived for power transmission in the first chapter was
particular to a string, the calculation of instantaneous intensity is specific to
the type of wave being considered.
2.3.1 Acoustic (Scalar) Intensity

Acoustic waves are propagating pressure variations within a fluid caused by
the displacement of particles that comprise the fluid. When the particles
move, they collide with neighboring particles causing them to move, which
allows the motion to propagate through the fluid longitudinally like on a
coiled spring. In locations where the particle density is high (a compression)
the pressure is higher than normal, whereas in locations where the particle
density is low (a rarefaction) the pressure is lower than normal (see Figure
2.9). The relationship between particle displacement, velocity and pressure
is given by Euler’s equation, which is an expression of Newton’s second law
for materials:
∂ 2ψ ∂u
ρ 2
=ρ = −∇p (2.37)
∂t ∂t
where ρ is fluid density, ψ is particle displacement, u is particle velocity and
p is the pressure.
Figure 2.9: Illustration of particle motion and the resulting density/pressure

variations for a harmonic acoustic wave.
Just as the mechanical power transferred by an object can be calculated

by taking the dot product between its velocity and the force it applies, the
instantaneous intensity for an acoustic wave at a given location is given by
the product of particle velocity and pressure at that location:
S = pu (2.38)
where the energy propagation and particle velocity directions are identical
because the wave is longitudinal. Euler’s equation can be used to substitute
a simple expression for the particle velocity in terms of the pressure:
Z
1
u=− ∇pdt
ρ
For harmonic plane waves in particular:
Z
1 h i
u=− ∇ Ãei(k·r−ωt) dt
ρ
Z
1 h i
=− ik Ãei(k·r−ωt) dt
ρ
1 k h i(k·r−ωt) i
= Ãe
ρω
1 h i
= k̂ Ãei(k·r−ωt)
ρv
p
= k̂
ρv
Plugging this result into the equation for instantaneous intensity yields:
p2
S= k̂ (2.39)
ρv
Note that when physical quantities are multiplied/squared, as pressure is
above, they must be real. Otherwise, the imaginary parts of complex rep-
resentations become real when multiplied and contribute to the real part of
the product. Consequently:
1 i(k·r−ωt+φ) 2 A2
S= Re Ae k̂ = cos2 (k · r − ωt + φ)k̂ (2.40)
ρv ρv
and the intensity is given by:

T
A2 A2
Z
1
I= cos2 (k · r − ωt + φ)k̂dt = k̂ (2.41)
T 0 ρv 2ρv
which is very similar to the expression for the average power transmitted by
a harmonic wave on a string (see Section 1.7). The modulus of a complex
number is defined as:
p √ √
|Ã| = ÃÃ∗ = Aeiφ Ae−iφ = A2 = A (2.42)
which is equal to its phasor magnitude, just as the modulus of a vector:
q
|A| = A2x + A2y + A2z = A
is equal to its magnitude. Consequently, the intensity can also be written in

terms of the complex amplitude as:
|Ã|2
I= k̂ (2.43)
2ρv
For harmonic spherical waves at large distances from the origin:
Z " #
1 Ã(θ, φ) i(±kr−ωt)
u=− ∇ e dt
ρ r
Applying the gradient operator in terms of spherical coordinates yields:
" # " #
Ã(θ, φ) i(±kr−ωt) ∂ Ã(θ, φ) i(±kr−ωt)
∇ e = r̂ e
r ∂r r
" # " #
1 ∂ Ã(θ, φ) i(±kr−ωt) 1 ∂ Ã(θ, φ) i(±kr−ωt)
+ θ̂ e + φ̂ e
r ∂θ r r sin θ ∂φ r
" #
Ã(θ, φ) i(±kr−ωt) Ã(θ, φ) i(±kr−ωt)
= r̂ − e ± ik e
r2 r
1 ∂ Ã(θ, φ) i(±kr−ωt) 1 ∂ Ã(θ, φ) i(±kr−ωt)
+ θ̂ e + φ̂ e
r2 ∂θ r2 sin θ ∂φ
The terms of order r−2 can be dropped at large distances from the origin
leaving:
1 ±k Ã(θ, φ) i(±kr−ωt) p
u= e r̂ = ± r̂
ρ ω r ρv
The instantaneous intensity is therefore given by:
p2 A2 (θ, φ)
S=± r̂ = ± cos2 (±kr − ωt + φ)r̂ (2.44)
ρv ρvr2
and the intensity by:
A2 (θ, φ) |Ã(θ, φ)|2

I=± r̂ = ± r̂ (2.45)
2ρvr2 2ρvr2
Notice that the intensity is almost identical to that for plane waves, but drops
off as 1/r2 . Physically, this is a consequence of energy conservation. The
total radiated power for a spherical wave can be calculated by integrating its
intensity over the surface of a spherical shell at any distance from the origin
r. Energy conservation requires the result to be the same regardless of the
value of r, but this follows only if the intensity drops off (1/r2 ) at the same
rate that the surface area of the spherical shell increases (r2 ).
2.3.2 Electromagnetic (Vector) Intensity

The instantaneous intensity for electromagnetic waves is given by the Poynt-
ing Vector :
1
S= E×B (2.46)
µ
For harmonic plane and spherical waves, B = (1/v)k̂ × E such that:
1 1
S= E × (k̂ × E)
µ v
Because the waves are transverse, the cross products yield a simple result:
1
S= |E|2 k̂ (2.47)
µv
The electric field for a plane wave is given by:
E = Ãei(±k·r−ωt)
Without loss of generality, the z-axis can be chosen to point along the direc-
tion of propagation such that:
E = Ax ei(kz−ωt+φx ) x̂ + Ay ei(kz−ωt+φy ) ŷ
Using the real part to calculate E because multiplication is required, the
Poynting Vector becomes:
1 2
S= [A cos2 (kz − ωt + φx ) + A2y cos2 (kz − ωt + φy )]ẑ (2.48)
µv x
The intensity is therefore given by:
1 |A|2
I= (A2x + A2y )ẑ = k̂ (2.49)
2µv 2µv
Making use of the definition of the modulus of a complex vector:
p
|Ã| = Ã · Ã∗
q
= Ax eiφx Ax e−iφx + Ay eiφy Ay e−iφy + Az eiφz Az e−iφz
q
= A2x + A2y + A2z = |A| (2.50)
it can also be written in terms of the complex vector amplitude:
|Ã|2
I= k̂ (2.51)
2µv
The electric field for a spherical wave is given by:
E= e
r
Aθ i(±kr−ωt+φθ ) Aφ i(±kr−ωt+φφ )
= e θ̂ + e φ̂
r r
Inserting the real part into the expression for the Poynting Vector:
1 A2θ A2φ

2 2
S=± cos (±kr − ωt + φθ ) + 2 cos (±kr − ωt + φφ ) r̂ (2.52)
µv r2 r
The intensity is therefore:
1 2 2 |A(θ, φ)|2 |Ã(θ, φ)|2

I=± (A + A )r̂ = ± r̂ = ± r̂ (2.53)
2µvr2 θ φ
2µvr2 2µvr2
Notice that the expressions for the intensities of both the plane and spherical
waves are almost identical to their acoustic counterparts.
Chapter 3
Interference
Chapter 1 considered the interference between identical, traveling harmonic

waves with different phase angles or opposite propagation directions. The
former gave rise to constructive, destructive and incomplete interference, and
the latter to standing waves. This behavior naturally extends to three di-
mensions. If two identical plane waves propagate in the same direction, they
interfere to produce another plane wave whose amplitude is determined by
their phase difference. If they propagate in opposite directions, they inter-
fere to produce standing waves with nodal and anti-nodal planar surfaces
perpendicular to the propagation direction. Likewise, if two identical spheri-
cal waves propagate in the same direction (toward or away from the origin),
they interfere to produce another spherical wave whose amplitude is deter-
mined by their phase difference. If they propagate in opposite directions,
they interfere to produce standing waves with nodal and anti-nodal spherical
surfaces.
However, there are additional types of interference in multiple dimensions
that are not possible in one dimension. This chapter addresses two of the
most significant: the interference between two identical plane waves travel-
ing in different directions, and between identical outgoing spherical waves
centered at different locations.
3.1 Plane Waves

If two identical plane waves are traveling in different directions, a coordinate
system can be introduced such that their wave vectors lie in the xy-plane
73
74 CHAPTER 3. INTERFERENCE
and are bisected by the x-axis as illustrated in Figure 3.1. If the origin of the
coordinate system is placed appropriately, the phase angles can be matched
such that the wave functions are given by:
Figure 3.1: Wave vectors for two interfering plane waves, where the coor-
dinate system is oriented such that the vectors lie in the xy-plane and are
bisected by the x-axis.
ψ1 = Ãei(kx x+ky y−ωt)

ψ2 = Ãei(kx x−ky y−ωt)

3.2. SPHERICAL WAVES 75
ψ = ψ1 + ψ2
= Ã(eiky y + e−iky y )ei(kx x−ωt)
= 2Ã cos (ky y)ei(kx x−ωt) (3.1)
which is the product of the wave functions for a plane wave traveling along
the x-axis (the direction that bisects the wave vectors) and a standing wave
along the y-axis (the orthogonal direction). The combined wave propagates
along the x-axis like a plane wave, but its amplitude varies sinusoidally in the
y direction like a standing wave. For two vector plane waves with identical
polarizations, the result is the same but with a vector amplitude:
ψ = 2Ã cos (ky y)ei(kx x−ωt)
3.2 Spherical Waves

Consider two identical, outgoing spherical waves whose centers are separated
by a distance d as illustrated in Figure 3.2. Using the geometry from the
figure, the wave functions for for the shifted spherical waves are given by:
Ã(θ1 , φ) i(k|r−r1 |−ωt)

ψ1 = e
|r − r1 |
Ã(θ2 , φ) i(k|r−r2 |−ωt)
ψ2 = e
|r − r2 |
The sum can not be simplified with the wave functions in their current form.
However, at large distances from the wave centers where r d and r
λ, the far-field approximation can be applied. This approximation asserts
that the amplitude terms of the wave functions are insensitive to the precise
locations of the wave centers at large distances, and can be approximated by
the amplitude for a wave centered on the origin:
Ã(θn , φ) Ã(θ, φ)
≈ (3.2)
|r − rn | r
The complex exponentials are more sensitive, but their distances can be
approximated by:
Figure 3.2: Geometry for two interfering spherical waves (solid lines) and the
far-field approximation (dashed lines).
|r − rn | ≈ r − r̂ · rn (3.3)
as illustrated in Figure 3.2 where |r − r1 | and r − r̂ · r1 are approximately
equal. The conversion in Figure 2.3 can be used to write r̂ in terms of
spherical coordinates:
r̂ = sin θ cos φx̂ + sin θ sin φŷ + cos θẑ (3.4)

such that:
|r − r1 | ≈ r − r̂ · r1 = r − r̂ · (d/2)ẑ = r − (d/2) cos θ

|r − r2 | ≈ r − r̂ · r2 = r − r̂ · (−d/2)ẑ = r + (d/2) cos θ
Substituting these values along with the amplitude approximation yields:
ψ = ψ1 + ψ2
Ã(θ, φ) i[kr−k(d/2) cos θ−ωt] Ã(θ, φ) i[kr+k(d/2) cos θ−ωt]
≈ e + e
r r
which is the wave function for two interfering spherical waves centered on
the origin with different phase angles. Consequently, the combined wave:
Ã(θ, φ) i(kr−ωt)
ψ ≈ e−ik(d/2) cos θ + eik(d/2) cos θ e
r
≈ e−ik(d/2) cos θ 1 + eikd cos θ ψsph

is also a spherical wave centered on the origin ψsph whose amplitude is deter-
mined by the phase difference (kd cos θ) between the waves stemming from
the difference in their propagation distances. The remaining sum can be eval-
uated using phasors as illustrated in Figure 3.3. Making use of the trigono-
metric identity cos x = 2 cos2 (x/2) − 1, the magnitude and angle of the
resultant are given by:
q p
A(θ) = [1 + cos (kd cos θ)]2 + sin2 (kd cos θ) = 2 + 2 cos (kd cos θ)
p
= 2 + 2[2 cos2 (k(d/2) cos θ) − 1] = 2 cos [k(d/2) cos θ]
φ = k(d/2) cos θ
Inserting this result yields:
ψ ≈ e−iφ [A(θ)eiφ ]ψsph

≈ A(θ)ψsph = 2 cos [k(d/2) cos θ]ψsph
If, as usual, the intensity of the spherical wave is proportional to its squared
amplitude, then:
I ≈ A2 (θ)Isph = 4 cos2 [k(d/2) cos θ]Isph (3.5)

where Isph is the intensity for ψsph . A2 (θ) is referred to as the interference
pattern and is plotted as a function of angle in Figure 3.4 for λ/d = 1/2. For
Figure 3.3: Phasor addition for two interfering spherical waves. Note that
φ = α/2 because the phasors form an isosceles triangle.
vector waves, the derivation is identical so ψsph can simply be replaced by its
vector equivalent.
The interference between the spherical waves is constructive when they
are in phase such that their phasors in Figure 3.3 align. This occurs at angles
for which the phase difference between the waves is an integer multiple of 2π:
kd cos θ = 2πn
λ
cos θ = n (3.6)
d
A(θ) is at its maximum value of 2 at these angles, indicating that the am-
plitude of the combined wave is twice the amplitude of the individual waves.
0
z
−1
−2
−3
−4
−4 4
−2 2
0 0
2 −2
4 −4
y
x
Figure 3.4: The interference pattern as a function of angle (θ and φ) for two
spherical waves where λ/d = 1/2.
The corresponding maximum value of the interference pattern is 4, and a

cone with this length appears in Figure 3.4 at each of the angles.
The interference is destructive when the waves are completely out of phase
such that their phasors anti-align. This occurs at angles for which the phase
difference between the waves is an odd multiple of π:
kd cos θ = π + 2πn

1 λ
cos θ = n + (3.7)
2 d
Given that A(θ) and the amplitude of the combined wave are equal to zero,
a nodal cone appears in Figure 3.4 at each of these angles. Notice that
Equations 3.6 and 3.7 indicate the spacing between maxima/nodal cones
increases with wavelength and decreases with distance between wave centers
d.
3.3 Application: Linear Arrays

A linear array is a line of evenly spaced wave sources or receivers. When
an array is used to generate waves, the individual sources usually produce
spherical waves. Consider an array of N such sources separated by a distance
d (see Figure 3.5). The wave function for the array is given by:
N N
X X Ã(θn , φ)
ψ= ψn = ei(k|r−rn |−ωt)
n=1 n=1
|r − rn |
Applying the far-field approximation where r−r̂·rn = r+[n−(N +1)/2]d cos θ
as illustrated in the figure for n = 2:
N
X Ã(θ, φ) N +1
ψ≈ ei[kr+k(n− 2
)d cos θ−ωt]
n=1
r
which is equivalent to the wave function for N interfering spherical waves
centered on the origin with different phase angles. The combined wave:
" N
#
X
i[k(n− N2+1 )d cos θ] Ã(θ, φ) i(kr−ωt)
ψ≈ e e
n=1
r
" N
#
−ik( N 2−1 )d cos θ
X
≈ e eikd cos θ(n−1) ψsph
n=1
is therefore also a spherical wave centered on the origin ψsph whose amplitude
is determined by the phase difference (kd cos θ) between the waves. The
remaining sum can be determined using phasors as illustrated in Figure 3.6,
however, it can also be evaluated directly as a geometric series:
N N
X
ikd cos θ(n−1)
X
n−1 xN − 1 eiN kd cos θ − 1
e = x = = ikd cos θ
n=1 n=1
x−1 e −1
which can be separated into a magnitude and angle for the resultant phasor:
3.3. APPLICATION: LINEAR ARRAYS 81
Figure 3.5: Geometry for a linear array with N sources. The case for an
odd value of N is shown, but the expression for the far-field approximation
is the same for an even value as long as the origin is located between the two
middle sources.
eiN kd cos θ − 1
A(θ)eiφ =
eikd cos θ − 1
eiN (k/2)d cos θ − e−iN (k/2)d cos θ eiN (k/2)d cos θ
= i(k/2)d cos θ
e − e−i(k/2)d cos θ ei(k/2)d cos θ
sin [N (k/2)d cos θ] ik( N −1 )d cos θ
= e 2
sin [(k/2)d cos θ]
sin [N (k/2)d cos θ] ik( N −1 )d cos θ
=N e 2
N sin [(k/2)d cos θ]
Substituting this result into the array wave function yields:
Figure 3.6: Phasor addition for a linear array.
ψ ≈ e−iφ A(θ)eiφ ψsph

sin [N (k/2)d cos θ]

≈ A(θ)ψsph = N ψsph
If the intensity is proportional to the squared amplitude:

2
2 2 sin [N (k/2)d cos θ]
I = A (θ)Isph = N Isph (3.8)
A representative interference pattern is plotted in Figure 3.7 for λ/d = 1/2
and N = 4. For vector wave sources, ψsph can simply be replaced by its
vector equivalent.
The interference is constructive when the waves are in phase such that all
of the phasors in Figure 3.6 align. This occurs at angles for which the phase
difference between the waves is an integer multiple of 2π:
kd cos θ = 2πn
λ
cos θ = n (3.9)
d
At these angles, A(θ) is at its maximum value of N , indicating that the
amplitude of the combined wave is N times the amplitude of the individual
waves, and the interference pattern is at its corresponding maximum value
of N 2 . A cone with this length (N 2 = 16) appears in Figure 3.7 at each of
the angles, where Equation 3.9 dictates that the spacing between them is
determined by wavelength and source separation d.
The interference is destructive at angles for which the tip of the last phasor
in Figure 3.6 meets the tail of the first such that A(θ) and the amplitude of
the combined wave are equal to zero. For this to be the case, the sum of the
phase differences between the waves (N α = N kd cos θ) must be equal to an
integer multiple of 2π:
N (kd cos θ) = 2πm

m
kd cos θ = 2π
N
mλ
cos θ = (3.10)
Nd
where m is an integer. Notice, however, that if m is an integer multiple of N ,
the phase difference between the waves becomes an integer multiple of 2π:
m nN
kdcosθ = 2π = 2π = 2πn
N N
meaning that the phasors align, the interference is constructive, and Equation
3.10 becomes identical to Equation 3.9. Figure 3.8 illustrates for m = 0...4
when N = 4. The interference is destructive for m = 1...3, but construc-
tive when m = 0 or 4. Broadly speaking, there are always 3 angles where
the interference is destructive between angles where the interference is con-
structive. These angles appear in the right pane of Figure 3.7 where the
Figure 3.7: The interference pattern for a linear array where λ/d = 1/2 and
N = 4. The two-dimensional plot on the right is included to make the nodes
between maxima visible.
interference pattern is equal to zero between angles where it is equal to its

maximum (N 2 = 16). More generally, m runs from 1 to N − 1 between n = 0
and n = 1, from N + 1 to 2N − 1 between n = 1 and n = 2, and so on such
that there are N − 1 nodes between maxima in the interference pattern.
The maximum corresponding to n = 0 at θ = 90 degrees is present for
all arrays and is called the main lobe. The others are referred to as side
lobes, and the number present is determined by the ratio λ/d in Equation
3.9. If it is greater than one, for example, then there is no solution for θ
at any value of n other than zero and there are no side lobes. In practice,
arrays are generally used to determine the direction of objects that reflect
the waves that they generate. Consequently, the absence of side lobes is
Figure 3.8: Phasor addition showing constructive and destructive interference

for a linear array with 4 elements.
desirable because any sizable reflections are known to have been produced
by objects located somewhere along the direction of the main lobe. The main
lobe can be scanned to search in multiple directions either by adjusting the
phase angles of the sources to make the interference constructive along angles
other than 90 degrees, or by physically rotating the array.
The narrower the main lobe, the better the array can localize the direction
of a reflector. Its size is defined as the angular separation between the nearest
nodes on either side, which correspond to m = ±1. In the right pane of
Figure 3.7, half of a main lobe is visible at 90 degrees with the m = 1 node
immediately to the left marking its edge. The angular separation between
the nodes can be calculated by differentiating Equation 3.10:
dm Nd
=− sin θ
dθ λ
λ 1
∆θ = ∆m
N d sin θ
Plugging in ∆m = 1 − (−1) = 2 and θ = 90 degrees yields:
2λ
∆θmain lobe = (3.11)
Nd
which indicates that the size of the main lobe is determined both by the ratio
λ/d and N , which is a measure of the size of the array. The larger the array,
the narrower the main lobe.
Chapter 4
Reflection and Transmission
In one dimension, the boundary between media where reflection and trans-
mission occurs is a point. In three dimensions, it is a surface. These surfaces
can have an infinite variety of shapes. Complex surfaces will be addressed
in subsequent chapters, but this one focuses on the simplest case of a pla-
nar boundary. It begins by considering the reflection and transmission of
plane waves at normal incidence to the boundary, then proceeds to oblique
incidence. Finally, the reflection and transmission of other types of traveling
waves is briefly discussed.
As in one dimension, boundary conditions must be applied at interfaces
between media to find wave functions for reflected and transmitted waves.
Because these conditions vary depending on wave type, this chapter will ad-
dress acoustic and electromagnetic reflection/transmission as representative
scalar and vector examples respectively. For acoustic waves, the pressure
must be equal across the boundary to satisfy Newton’s third law. In addi-
tion, the normal component of particle velocity must be equal to maintain the
integrity of the boundary. These conditions are expressed mathematically as:
p1 = p2 u⊥ ⊥
1 = u2 (4.1)
The boundary conditions for electromagnetic waves are found by enforcing
Maxwell’s equations and are very similar:
k k
1 E1⊥ = 2 E2⊥ E1 = E2
1 k 1 k (4.2)
B1⊥ = B2⊥ B1 = B2
µ1 µ2
87
88 CHAPTER 4. REFLECTION AND TRANSMISSION
4.1 Plane Waves at Normal Incidence

4.1.1 Acoustic Waves
Consider an acoustic plane wave incident upon a flat boundary between me-
dia as illustrated in Figure 4.1.
Figure 4.1: Geometry for determining reflection and transmission coefficients

for an acoustic plane wave normally incident on a flat boundary between
media.
The incident, reflected and transmitted wave functions are given by:
pi = Ãi ei(k1 x−ωt) pr = Ãr ei(−k1 x−ωt) pt = Ãt ei(k2 x−ωt)

where the complex amplitudes of the reflected and transmitted waves are
determined by applying boundary conditions. The first one requires (see
Equation 4.1):
4.1. PLANE WAVES AT NORMAL INCIDENCE 89
pi (0, t) + pr (0, t) = pt (0, t)

Ãi + Ãr = Ãt
Applying the second boundary condition:
uix (0, t) + urx (0, t) = utx (0, t)

For a plane wave (see Subsection 2.3.1):
p
u= k̂
ρv
Making this substitution:
Ãi Ãr Ãt

x̂ + (−x̂) = x̂
ρ1 v1 ρ1 v1 ρ2 v2
Ãi Ãr Ãt
− =
ρ1 v1 ρ1 v1 ρ2 v2
The specific acoustic impedance for a wave is defined as the ratio of its pres-
sure to particle speed. For a plane wave, it is given by:
p
Specific Acoustic Impedance = Z = = ρv (4.3)
u
Plugging it in yields:
Ãi Ãr Ãt

− =
Z1 Z1 Z2
Multiplying this equation by Z2 and subtracting the equation for the first
boundary condition:

Z2 Z2
− 1 Ãi − + 1 Ãr = 0
Z1 Z1
(Z2 − Z1 )Ãi = (Z2 + Z1 )Ãr
The reflection coefficient is therefore given by:

Ãr Z2 − Z1
R= = (4.4)
Ãi Z2 + Z1
and the transmission coefficient by:
Ãt Ãi + Ãr 2Z2

T = = =1+R= (4.5)
Ãi Ãi Z2 + Z1
Notice that for transmission from a high impedance medium, such as

water, into a low impedance medium, such as air, R → −1 and T → 0. The
incident wave is almost completely reflected with a 180-degree phase change,
while the transmitted wave is nearly absent. Because the pressure at the
boundary is nearly zero (pi + pr → Ãi + Ãr = Ãi − Ãi = 0), it is referred to
as a pressure release boundary. As Z1 decreases/Z2 increases, the reflection
coefficient approaches zero and the transmission coefficient increases. When
the two are equal, R = 0 and T = 1 indicating complete transmission.
Beyond this point, both coefficients increase and the reflection coefficient
becomes positive, indicating that the reflected wave no longer undergoes a
phase change. As Z2 becomes much larger than Z1 , R → 1 and T → 2.
The incident wave is almost completely reflected and the transmitted wave,
although present, carries very little energy due to the large impedance of the
medium (see Equation 2.43). Because the normal component of the fluid
velocity at the boundary is nearly zero (uix + urx → Ãi /ρ1 v1 − Ãr /ρ1 v1 =
Ãi /ρ1 v1 − Ãi /ρ1 v1 = 0), it is referred to as a rigid boundary.
4.1.2 Electromagnetic Waves

Figure 4.2 shows an electromagnetic plane wave normally incident upon a
planar boundary. In general, the electric and magnetic field vectors can have
components both parallel and perpendicular to the plane of incidence (the
plane that contains both the incident wavevector and the y-axis, i.e. the plane
of the page). Consider initially the parallel electric field component (Ey )
and the corresponding perpendicular magnetic field component (Bz ) given
by Equation 2.20, which are the ones shown in the figure. The incident,
reflected and transmitted wave functions for these components are:
4.1. PLANE WAVES AT NORMAL INCIDENCE 91

for an electromagnetic plane wave normally incident on a flat boundary be-
tween media.
Eiy = Ãi ei(k1 x−ωt) Ery = Ãr ei(−k1 x−ωt) Ety = Ãt ei(k2 x−ωt)
Ãi i(k1 x−ωt) Ãr Ãt i(k2 x−ωt)
Biz = e Brz = − ei(−k1 x−ωt) Btz = e
v1 v1 v2
Because there are no components of the fields perpendicular to the boundary,
the corresponding boundary conditions in Equation 4.2 can be ignored. For
the parallel component of the electric field:
Eiy (0, t) + Ery (0, t) = Ety (0, t)

Ãi + Ãr = Ãt
For the parallel component of the magnetic field:
1 1 1
Biz (0, t) + Brz (0, t) = Btz (0, t)
µ1 µ1 µ2
Ãi Ãr Ãt
− =
µ1 v1 µ1 v1 µ2 v2
With the substitution of µ for ρ, the two equations above are identical to
those obtained by applying acoustic boundary conditions. Consequently:
Ãr µ2 v2 − µ1 v1
R= = (4.6)
Ãi µ2 v 2 + µ1 v 1
and:
Ãt 2µ2 v2
T = =1+R= (4.7)
Ãi µ2 v2 + µ1 v1
The process can be repeated for the other two field components (Ez and By ),
but the results are identical.
Note that v = 0 for conductors because they can not support propagating
electromagnetic waves. Consequently, R = −1 and T = 0 for an electromag-
netic wave incident upon a conducting surface. Conducting boundaries are
therefore the electromagnetic equivalent of acoustic pressure release bound-
aries.
4.2 Plane Waves at Oblique Incidence

4.2.1 Acoustic Waves
Figure 4.3 shows an acoustic plane wave incident upon a planar boundary
at oblique incidence. The incident, reflected and transmitted wave functions
are given by:
pi = Ãi ei(k1 cos θi x+k1 sin θi y−ωt) pr = Ãr ei(krx x+kry y+krz z−ωt)
pt = Ãt ei(ktx x+kty y+ktz z−ωt)
Applying the first boundary condition:

4.2. PLANE WAVES AT OBLIQUE INCIDENCE 93

for an acoustic plane wave obliquely incident on a flat boundary between
media.
pi (0, y, z, t) + pr (0, y, z, t) = pt (0, y, z, t)

Ãi ei(k1 sin θi y−ωt) + Ãr ei(kry y+krz z−ωt) = Ãt ei(kty y+ktz z−ωt)
For this equation to be valid at all values of y and z, the exponents must be
equal. This requires the y and z components of the wavevectors be be equal.
For the z components:
ktz = krz = kiz = 0

which implies that all three wavevectors must lie in the same plane. If θr and
θt are the angles of the reflected and transmitted wavevectors with respect
to the x-axis, then the remaining components of the vectors can be written
as:
krx = −k1 cos θr kry = k1 sin θr

ktx = k2 cos θt kty = k2 sin θt
Making this substitution for the y-components of the wavevectors:
k1 sin θi = k1 sin θr = k2 sin θt

which requires:
θi = θr Law of Reflection (4.8)

and
sin θi sin θt
= Snell’s Law (4.9)
v1 v2
Snell’s law is also referred to as the law of refraction, and implies that the
transmitted wavevector is bent toward/away from the x-axis relative to the
incident wavevector when the wave passes into a slower/faster medium. With
the y and z components of the wavevectors equal, the boundary condition
equation becomes:
Ãi + Ãr = Ãt

uix (0, y, z, t) + urx (0, y, z, t) = utx (0, y, z, t)

pi (0, y, z, t) pr (0, y, z, t) pt (0, y, z, t)
cos θi − cos θr = cos θt
Z1 Z1 Z2
Ãi i(k1 sin θi y−ωt) Ãr i(k1 sin θr y−ωt) Ãt i(k2 sin θt y−ωt)
e cos θi − e cos θr = e cos θt
Z1 Z1 Z2
The law of reflection and Snell’s law guarantee that the exponents are iden-
tical, so:
Ãi Ãr Ãt

Z1 Z1 Z2
Substituting the first boundary condition equation into the second:
Ãi Ãr Ãi + Ãr

Z1 Z1 Z2
Ãi (Z2 cos θi − Z1 cos θt ) = Ãr (Z1 cos θt + Z2 cos θr )
Using the law of reflection to substitute θi for θr yields the reflection coeffi-
cient:
Ãr Z2 cos θi − Z1 cos θt

R= = (4.10)
Ãi Z2 cos θi + Z1 cos θt
where cos θt can be obtained from Snell’s law:
s
p
2 v22
cos θt = 1 − sin θt = 1 − 2 sin2 θi (4.11)
v1
The corresponding transmission coefficient is given by:
Ãt 2Z2 cos θi

T = =1+R= (4.12)
Ãi Z2 cos θi + Z1 cos θt
Both coefficients vary with the angle of the incident wave. Often, there
is an angle at which the reflection coefficient is zero and the transmission
coefficient is one. Because the wave is completely transmitted at this angle,
it is referred to as the angle of intromission. The reflection coefficient also
changes sign at this angle, indicating the phase of the reflected wave flips 180
degrees. The angle can be found by setting the reflection coefficient equal to
zero and solving for the incident angle:
Z2 cos θi − Z1 cos θt
0=
Z2 cos θi + Z1 cos θt
0 = ρ2 v2 cos θi − ρ1 v1 cos θt
s
v2
ρ2 v2 cos θi = ρ1 v1 1 − 22 sin2 θi
v1
ρ22 v22 cos2 θi = ρ21 v12 − ρ21 v22 (1 − cos2 θi )
(ρ22 v22 − ρ21 v22 ) cos2 θi = ρ21 v12 − ρ21 v22
s
v12 /v22 − 1
cos θi = (4.13)
ρ22 /ρ21 − 1
For the angle to be real, the speed and density ratios must either both be
greater than one or less than one. If they are greater than one, the speed ratio
must be smaller than the density ratio. If they are less than one, the speed
ratio must be larger than the density ratio. Consequently, not all boundaries
possess an angle of intromission.
If the propagation speed in the incident medium is less than the propa-
gation speed in the transmitted medium (v1 < v2 ), there is an incident angle
for which the transmitted angle is 90 degrees. It is referred to as the critical
angle (θc ), and can be found using Snell’s law:
sin θc sin (π/2)

=
v1 v2
v1
sin θc = (4.14)
v2
At this angle, the reflection coefficient is given by:
Z2 cos θi − Z1 cos(π/2)
R= =1
Z2 cos θi + Z1 cos(π/2)
indicating complete reflection. Beyond the critical angle, Snell’s law yields no
real solution for the transmitted angle (sin θt > 1), implying that the incident
wave continues to be completely reflected. For this reason, any wave with
an angle of incidence greater than or equal to the critical angle is said to
undergo total internal reflection.
4.2.2 Electromagnetic Waves
Figure 4.4 shows the electric field component parallel to the plane of inci-
dence and the corresponding perpendicular magnetic field component for an
electromagnetic wave obliquely incident upon a planar boundary. Because
boundary conditions must be enforced at x = 0 for all values of y and z, the
required conditions on the acoustic wavevectors (the y and z components
must be identical) also apply to the electromagnetic wave vectors. Conse-
quently, the law of reflection and Snell’s law are valid and can be used in
applying the boundary conditions of Equation 4.2. The first produces:

for an electromagnetic plane wave obliquely incident on a flat boundary be-
tween media (the electric field component parallel to the plane of incidence
and the corresponding perpendicular magnetic field component are shown).
1 E1⊥ = 1 E1⊥
1 Eipr (0, y, z, t) sin θi − 1 Erpr (0, y, z, t) sin θr = 2 Etpr (0, y, z, t) sin θt
1 Ãpr pr pr
i sin θi − 1 Ãr sin θr = 2 Ãt sin θt
pr v2
1 Ãpr pr
i sin θi − 1 Ãr sin θi = 2 Ãt sin θi
v1
v1 1 Ãpr pr pr
i − v1 1 Ãr = v2 2 Ãt
Ãpr Ãpr Ãpr

i
− r = t
µ1 v1 µ1 v1 µ2 v2
k k
E1 = E2
Ei (0, y, z, t) cos θi + Er (0, y, z, t) cos θr = Etpr (0, y, z, t) cos θt
pr pr
Ãpr pr pr
i cos θi + Ãr cos θi = Ãt cos θt
The third boundary condition is irrelevant in this case, and applying the
fourth yields:
1 k 1 k
B1 = B2
µ1 µ2
1 1 1
Eipr (0, y, z, t) − Erpr (0, y, z, t) = Etpr (0, y, z, t)
µ1 v 1 µ1 v 1 µ2 v2
pr pr pr
Ãi Ã Ã
− r = t
µ1 v1 µ1 v1 µ2 v2
which is identical to the first equation and therefore redundant. Plugging
the first equation into the second yields the reflection coefficient:

µ2 v2 pr µ2 v2 pr
Ãpr cos θi + Ãpr
r cos θi = Ã − Ã cos θt
i
µ1 v1 i µ1 v 1 r
µ1 v1 cos θi Ãpr pr pr pr
i − µ2 v2 cos θt Ãi = −µ1 v1 cos θi Ãr − µ2 v2 cos θt Ãr
Ãpr
r µ2 v2 cos θt − µ1 v1 cos θi
Rpr = pr = (4.15)
Ãi µ2 v2 cos θt + µ1 v1 cos θi
where cos θt is obtained from Snell’s law. From the first equation:
Ãpr
t µ2 v 2 2µ2 v2 cos θi
T pr = pr = (1 − Rpr ) = (4.16)
Ãi µ1 v 1 µ2 v2 cos θt + µ1 v1 cos θi
The process can be repeated to find the reflection and transmission coeffi-
cients for the perpendicular electric field/parallel magnetic field components.
Using the geometry of Figure 4.5 to apply the boundary conditions:

for an electromagnetic plane wave obliquely incident on a flat boundary be-
tween media (the electric field component perpendicular to the plane of inci-
dence and the corresponding parallel magnetic field component are shown).
0 = 0 Ãpp pp pp
i + Ãr = Ãt
Ãpp Ãpp Ãpp Ãpp Ãpp Ãpp

i
sin θi + r sin θr = t sin θt i
cos θi − r cos θr = t cos θt
v1 v1 v2 µ1 v1 µ1 v1 µ2 v2
Snell’s law and the law of reflection make the third equation equivalent to
the second and therefore redundant. Plugging the second into the fourth and
applying the law of reflection yields the reflection coefficient:
Ãpp Ãpp Ãpp Ãpp

i
cos θi − r cos θi = i cos θt + r cos θt
µ1 v1 µ1 v1 µ2 v2 µ2 v2
pp pp
µ2 v2 cos θi Ãi − µ1 v1 cos θt Ãi = µ2 v2 cos θi Ãr + µ1 v1 cos θt Ãpp
pp
r
pp Ãpp
r µ2 v2 cos θi − µ1 v1 cos θt
R = pp = (4.17)
Ãi µ2 v2 cos θi + µ1 v1 cos θt
From the second equation:
pp Ãpp
t 2µ2 v2 cos θi
T = pp = 1 + Rpp = (4.18)
Ãi µ2 v2 cos θi + µ1 v1 cos θt
Because the reflection and transmission coefficients for the different field
components are not identical, their angles of intromission are different. Rpp
and T pp match the results for an acoustic wave with the substitution of µ for
ρ, so the angle of intromission for the perpendicular electric/parallel mag-
netic field components is given by Equation 4.13 with the same substitution.
A valid solution for the angle often does not exist, however, because most ma-
terials have nearly identical permeabilities. The angle of intromission for the
parallel electric/perpendicular magnetic field components can be calculated
by setting their reflection coefficient equal to zero:
µ2 v2 cos θt − µ1 v1 cos θi
0 = Rpr =
µ2 v2 cos θt + µ1 v1 cos θi
0 = µ2 v2 cos θt − µ1 v1 cos θi
s
v2
µ1 v1 cos θi = µ2 v2 1 − 22 sin2 θi
v1
v24
µ21 v12 cos2 θi = µ22 v22 − µ22 sin2 θi
v12
µ21 v14 cos2 θi = µ22 v22 v12 − µ22 v24 (1 − cos2 θi )
µ2 v 2 v 2 − µ2 v 4
cos2 θi = 2 2 2 4 1 2 2 4 2
µ v − µ2 v2
v1 1
u 2
v2
−1
u
u v1
cos θi = u
t 2 2
v2 µ1 v1
v1
− µ2 v2
Because the permeabilities of most materials are nearly identical (µ1 ≈ µ2 ):

s
v22 v1
cos θi = 2 2
or tan θi = (4.19)
v2 + v1 v2
In optics, this angle is commonly referred to as Brewster’s Angle. Because
the parallel electric field component is entirely transmitted at this angle, any
reflected light must be linearly polarized with its electric field perpendicular
to the plane of incidence/parallel to the planar surface.
Given that Snell’s law applies equally to electromagnetic and acoustic
waves, critical angles are calculated using the same formula (Equation 4.14).
However, the corresponding reflection coefficients for the two sets of field
components have opposite sign:
µ2 v2 cos(π/2) − µ1 v1 cos θi
Rpr = = −1
µ2 v2 cos(π/2) + µ1 v1 cos θi
µ2 v2 cos θi − µ1 v1 cos(π/2)
Rpp = =1
µ2 v2 cos θi + µ1 v1 cos(π/2)
indicating a 180-degree phase change for the parallel electric/perpendicular

magnetic field components.
For waves incident upon a conducting surface (v2 = 0), Rpp , Rpr = −1
and T pp , T pr = 0 as expected. The component of the total electric field
parallel to the boundary must therefore be equal to zero at the boundary
kpr kpr
(Ei + Er → Ãpr pr pr pr
i cos θi + Ãr cos θr = Ãi cos θi − Ãi cos θi = 0 and
kpp kpp
Ei + Er → Ãpp pp pp pp
i + Ãr = Ãi − Ãi = 0). In addition, the component
of the total magnetic field perpendicular to the boundary must be equal
to zero at the boundary (Bi⊥pr + Br⊥pr → Ãpp pp
i sin θi /v1 + Ãr sin θr /v1 =
pp pp
Ãi sin θi /v1 − Ãi sin θi /v1 = 0).
4.3 Non-Plane Waves

It is much more difficult to apply boundary conditions at planar surfaces for
other types of traveling waves. The usual approach is to rewrite the wave as
a linear combination of plane waves as discussed in Chapter 2, calculate the
reflection coefficient for each, and sum the results. If, however, the boundary
is perfectly reflecting, the reflected wave is simply a mirror-image of the
incoming wave. If the boundary is rigid (R = 1), the reflected wave is in
phase with the incident wave. If it is a pressure release boundary, or in the
case of electromagnetic waves a conducting boundary (R = −1), the reflected
wave is 180 degrees out of phase with the incoming wave.
Chapter 5
Cavities and Waveguides
In three dimensions, a finite medium is bounded by surfaces. If the medium

is completely enclosed, it is called a cavity. If it is only partially enclosed
such that waves can freely propagate in one or two dimensions, it is called a
waveguide. This chapter examines the waves in both. As in one dimension,
the wave functions for finite media are determined by applying boundary
conditions to the separable solutions of the wave equation at the enclosing
surfaces. Because the conditions vary depending upon wave type, acoustic
waves will again be considered as a representative of scalar waves and elec-
tromagnetic waves as a representative of vector waves. Only the simplest
cases, where the boundaries are perfectly reflecting, will be addressed. Pres-
sure release and rigid boundaries are the two types of perfectly reflecting
boundaries for acoustic waves. As discussed in Subsection 4.1.1, the pressure
must be zero at pressure release boundaries, and the normal component of
the particle velocity must be zero at rigid boundaries:
p = 0 Pressure Release (5.1)

u⊥ = 0 Rigid (5.2)
Electromagnetic waves are completely reflected by conducting surfaces. At

conducting boundaries, the parallel component of the electric field and the
perpendicular component of the magnetic field must be zero (see Subsection
4.1.2):
Ek = 0 B⊥ = 0 (5.3)
103
104 CHAPTER 5. CAVITIES AND WAVEGUIDES
The chapter begins by finding the acoustic and electromagnetic wave

functions for a rectangular cavity. It then proceeds to the more complicated
cases of cylindrical and spherical cavities for acoustic waves. Rectangular
waveguides are then addressed for both wave types, and cylindrical waveg-
uides for acoustic waves. Because propagating harmonic waves are bounded
in one or more dimensions in each case, they interfere with their reflections
to produce a discrete set of possible standing waves along those directions
just as in one dimension.
5.1 Rectangular Cavity

5.1.1 Acoustic
Figure 5.1 shows a rectangular cavity with rigid walls. Because the cav-
ity possesses rectangular symmetry, boundary conditions are applied to the
separable solutions in rectangular coordinates. The standing wave solutions
are required for each coordinate because the waves are bounded in all three
directions:
p(x, y, z, t) = (a cos ωt + b sin ωt)(c cos kx x + d sin kx x)

(f cos ky y + g sin ky y)(h cos kz z + l sin kz z)
where:
(ω/v)2 = k 2 = kx2 + ky2 + kz2

Because the rigid boundary condition pertains to particle velocity, Euler’s
equation is required to determine the corresponding condition on the pres-
sure:
∂u
ρ = −∇p
∂t
∂ux ∂uy ∂uz ∂p ∂p ∂p
ρ x̂ + ρ ŷ + ρ ẑ = − x̂ − ŷ − ẑ
∂t ∂t ∂t ∂x ∂y ∂z
For the boundaries at x = 0 and x = Lx , ux is the normal component of
particle velocity. The rigid boundary condition therefore requires ux = 0 and
through Euler’s equation:
5.1. RECTANGULAR CAVITY 105
Figure 5.1: Geometry for a rectangular cavity.
∂p
= 0 at x = 0, Lx
∂x
Applying this condition to the standing wave solution at x = 0 yields:
0 = (a cos ωt + b sin ωt)(dkx )(f cos ky y + g sin ky y)(h cos kz z + l sin kz z)
which can only be valid at all values of y, z, and t if d = 0. The solution

therefore becomes:
p(x, y, z, t) = cos kx x(a cos ωt + b sin ωt)

where the constant c has been absorbed into a and b. Applying the boundary
condition at x = Lx :
0 = −kx sin kx Lx (a cos ωt + b sin ωt)

This can only be valid at all values of y, z, and t if:

lπ
sin kx Lx = 0 ⇒ kx Lx = lπ ⇒ kxl = (5.4)
Lx
where l is an integer.
The process and the results are identical for the boundaries at y = 0, Ly
and z = 0, Lz . The final wave function is therefore:
p(x, y, z, t) = cos kxl x cos kym y cos kzn z(a cos ωlmn t + b sin ωlmn t) (5.5)
lπ mπ nπ
kxl = kym = kzn = (5.6)
Lx Ly Lz
s
2 2 2
l m n
ωlmn = vklmn = πv + + (5.7)
Lx Ly Lz
where each combination of l, m and n represents a different normal mode and

the corresponding frequencies comprise a harmonic series. If any of the side
lengths (Lx ,Ly ,Lz ) are equal, multiple modes will share the same frequency
in a phenomenon referred to as degeneracy. Figure 5.2 shows cross-sections
of the solutions in the xy-plane for several values of l and m. The nodal
lines visible in the figure where p = 0 become planes when extended in the z
direction. There are l of these in the x direction and m in the y direction.
Because multiple normal modes can exist in the same medium at the
same time, the most general wave function for the cavity is given by:
X
p(x, y, z, t) = cos kxl x cos kym y cos kzn z(almn cos ωlmn t + blmn sin ωlmn t)
l,m,n
(5.8)
The modal amplitudes, almn and blmn , are determined by applying initial
conditions and using Fourier integrals as outlined in Chapter 1 for a fixed
1 1
l = 1,m = 1
l = 1,m = 0
0 0
1 1
−1 0.5 −1 0.5
0 1 y 0 1 y
x 2 0 x 2 0
1 1
l = 2,m = 2
l = 2,m = 1
0 0
1 1
−1 0.5 −1 0.5
0 1 y 0 1 y
x 2 0 x 2 0
Figure 5.2: Cross-sections in the xy-plane of several normal modes for a rigid
rectangular cavity with Lx = 2 and Ly = 1.
string. As in that case, any possible wave initiated in the cavity can be
written as a linear combination of normal modes in this manner.
If the rigid walls of the cavity are replaced with pressure release bound-
aries, the boundary condition at x = 0 and x = Lx becomes:
p = 0 at x = 0, Lx
Applying it to the standing wave solution yields the wave function:
p(x, y, z, t) = sin kxl x(a cos ωt + b sin ωt)

where:
lπ
kxl =
Lx
Because the boundary condition is identical at y = 0, Ly and z = 0, Lz , the
results are the same along the other directions and the normal modes are
given by:
p(x, y, z, t) = sin kxl x sin kym y sin kzn z(a cos ωlmn t + b sin ωlmn t) (5.9)
lπ mπ nπ
kxl = kym = kzn = (5.10)
Lx Ly Lz
s
2 2 2
l m n
ωlmn = πv + + (5.11)
Lx Ly Lz
The first few are illustrated in Figure 5.3, where there is one less nodal plane
along each direction than for the same mode in a rigid cavity.
5.1.2 Electromagnetic
Consider electromagnetic waves in a rectangular cavity with conducting walls
and the dimensions shown in Figure 5.1. The separable solution for each
component of the electric field is identical to that for acoustic pressure:
Es (x, y, z, t) = ãs e−iωt (cs cos kx x + ds sin kx x)

(fs cos ky y + gs sin ky y)(hs cos kz z + ls sin kz z)
s =x, y or z
where the complex form of the time dependence is being used to simplify the
mathematics. The condition that the parallel component of the electric field
vanish at each of the conducting boundaries requires:
Ex = 0 at y = 0, Ly and z = 0, Lz
Ey = 0 at x = 0, Lx and z = 0, Lz
Ez = 0 at x = 0, Lx and y = 0, Ly
1 1
l = 1,m = 1
l = 2,m = 1
0 0
1 1
−1 0.5 −1 0.5
0 1 y 0 1 y
x 2 0 x 2 0
1 1
l = 1,m = 2
l = 2,m = 2
0 0
1 1
−1 0.5 −1 0.5
0 1 y 0 1 y
x 2 0 x 2 0
Figure 5.3: Cross-sections in the xy-plane of several normal modes for a

pressure release rectangular cavity with Lx = 2 and Ly = 1.
Applying the top condition to the separable solution for Ex requires hx =

0, fx = 0, ky = mπ/Ly , and kz = nπ/Lz . Applying the remaining two
conditions results in similar restrictions on the solutions for Ey and Ez such
that:
Ex = ãx e−iωlmn t (cx cos kxl x + dx sin kxl x) sin kym y sin kzn z
Ey = ãy e−iωlmn t sin kxl x(fy cos kym y + gy sin kym y) sin kzn z
Ez = ãz e−iωlmn t sin kxl x sin kym y(hz cos kzn z + lz sin kzn z)
where:
lπ mπ nπ
kxl = kym = kzn =
Lx Ly Lz
s
2 2 2
l m n
ωlmn = πv + +
Lx Ly Lz
These solutions must also satisfy Maxwell’s equations. Forcing them to

be divergenceless:
0 = ∇ · E = ãx e−iωlmn t (−cx kxl sin kxl x + dx kxl cos kxl x) sin kym y sin kzn z
+ ãy e−iωlmn t sin kxl x(−fy kym sin kym y + gy kym cos kym y) sin kzn z
+ ãz e−iωlmn t sin kxl x sin kym y(−hz kzn sin kzn z + lz kzn cos kzn z)
For this to be valid at all values of x, y, z and t, the functions in each of the
summed terms must be identical. As a result, dx , gy and lz must be zero.
The equation then becomes:
0 = −ãx kxl e−iωlmn t sin kxl x sin kym y sin kzn z

− ãy kym e−iωlmn t sin kxl x sin kym y sin kzn z
− ãz kzn e−iωlmn t sin kxl x sin kym y sin kzn z
where the remaining constants have been absorbed into ãx , ãy , and ãz . Di-
viding by the common functions yields:
0 = ãx kxl + ãy kym + ãz kzn

which implies that one of the remaining constants is determined by the values
of the other two.
The corresponding magnetic field is obtained from Faraday’s law:
∂B
∇×E =− = iωB
∂t
−i ∂Ez ∂Ey −i ∂Ex ∂Ez −i ∂Ey ∂Ex
Bx = − By = − Bz = −
ω ∂y ∂z ω ∂z ∂x ω ∂x ∂y
Plugging in the solutions for the components of the electric field and using
−i = e−iπ/2 :
1
Bx = (ãz kym − ãy kzn )e−i(ωlmn t+π/2) sin kxl x cos kym y cos kzn z
ωlmn
1
By = (ãx kzn − ãz kxl )e−i(ωlmn t+π/2) cos kxl x sin kym y cos kzn z
ωlmn
1
Bz = (ãy kxl − ãx kym )e−i(ωlmn t+π/2) cos kxl x cos kym y sin kzn z
ωlmn
These solutions also satisfy the remaining two Maxwell’s equations as well as
the remaining boundary condition that the perpendicular component of the
magnetic field vanish at each of the conducting surfaces:
Bx = 0 at x = 0, Lx
By = 0 at y = 0, Ly
Bz = 0 at z = 0, Lz
It is common practice to divide the solutions into two independent sets.

The first are referred to as transverse electric (TE) solutions because there
is no z component of the electric field, meaning that ãz = 0 and therefore
ãy = −(kxl /kym )ãx from Gauss’ law. Making these substitutions yields:
Ex = ax cos(ωlmn t + δ) cos kxl x sin kym y sin kzn z

kxl
Ey = − ax cos(ωlmn t + δ) sin kxl x cos kym y sin kzn z (5.12)
kym
Ez = 0
where the imaginary part of the solutions has been excluded and the remain-
ing constants ax and δ are determined by initial conditions. The correspond-
ing magnetic field components are given by:
1 kzn kxl
Bx = ax cos(ωlmn t + δ + π/2) sin kxl x cos kym y cos kzn z
ωlmn kym
1
By = kzn ax cos(ωlmn t + δ + π/2) cos kxl x sin kym y cos kzn z (5.13)
ωlmn
2 2
−1 kxl + kym
Bz = ax cos(ωlmn t + δ + π/2) cos kxl x cos kym y sin kzn z
ωlmn kym
The second independent set are called transverse magnetic (TM) solutions
because there is no z component of the magnetic field, meaning that ãy kxl −
ãx kym = 0 and therefore ãy = (kym /kxl )ãx . In addition, from Gauss’ law:
2
kym
0 = ãx kxl + ãx + ãz kzn
kxl
2 2
kxl + kym
ãz = − ãx
kxl kzn
The components of the electric field are therefore given by:
Ex = ax cos(ωlmn t + δ) cos kxl x sin kym y sin kzn z

kym
Ey = ax cos(ωlmn t + δ) sin kxl x cos kym y sin kzn z (5.14)
kxl
2 2
kxl + kym
Ez = − ax cos(ωlmn t + δ) sin kxl x sin kym y cos kzn z
kxl kzn
and the components of the magnetic field by:
1 kym kxl − kym kzn

Bx = ax cos(ωlmn t + δ + π/2) sin kxl x cos kym y cos kzn z
ωlmn kxl
2 2 2
1 kxl + kym + kzn
By = ax cos(ωlmn t + δ + π/2) cos kxl x sin kym y cos kzn z
ωlmn kzn
(5.15)
Bz = 0
5.2. CYLINDRICAL CAVITY 113
5.2 Cylindrical Cavity

The wave functions for a cylindrical cavity of the type illustrated in Figure
5.4 are found by applying boundary conditions to the standing wave solutions
for cylindrical coordinates (see Subsection 2.2.3):
p(r, φ, z, t) = (a cos ωt + b sin ωt)[f Jm (kr r) + gNm (kr r)]

(c cos mφ + d sin mφ)(h cos kz z + l sin kz z)
where:
(ω/v)2 = k 2 = kr2 + kz2
Figure 5.4: Geometry for a cylindrical cavity.

For rigid boundaries, the normal component of particle velocity (ur at r = a

and uz at z = 0, L) must be zero. Euler’s equation in cylindrical coordinates
is given by:
∂ur ∂uφ ∂uz ∂p 1 ∂p ∂p
ρ r̂ + ρ φ̂ + ρ ẑ = − r̂ − φ̂ − ẑ (5.16)
∂t ∂t ∂t ∂r r ∂φ ∂z
Consequently, the corresponding boundary conditions on the pressure are:
∂p
= 0 at r = a
∂r
∂p
= 0 at z = 0, L
∂z
Because the Nuemann functions explode at r = 0, g = 0 in the standing
wave solution. Applying the first boundary condition to what remains:
0
0 = (a cos ωt + b sin ωt)kr Jm (kr a)
0
where Jm is the derivative of Jm . This can only be valid at all values of φ,
z, and t if:
0 0 0
Jm (kr a) = 0 ⇒ kr a = Uml ⇒ krml = Uml /a (5.17)
0 0
where Uml is the l’th root of Jm . Because the slope of a function is zero at
0
its extrema, the l’th root of Jm is also the l’th extremum of Jm . Figure 5.5
shows the first two for m = 0.
Applying the remaining boundary condition at z = 0, L yields the same
results as for a rigid rectangular cavity. The normal modes are therefore:
p(r, φ, z, t) = Jm (krml r)(c cos mφ + d sin mφ)

cos kzn z(a cos ωmln t + b sin ωmln t) (5.18)
0 nπ
krml = Uml /a kzn = (5.19)
s L
0
2
Uml nπ 2
ωmln = v + (5.20)
a L
5.2. CYLINDRICAL CAVITY 115
1.5
0 0
U01 U01 U02 U02
1
J0(U )
0.5
−0.5
0 1 2 3 4 5 6 7 8 9 10
U
Figure 5.5: The first few roots and extrema for J0 .
where the constants c and d simply determine the orientation of the mode
about the z-axis, and are set by initial conditions along with a and b. Figure
5.6 shows cross-sections of the solutions in the xy-plane for several values of
m and l with c = 1 and d = 0. Notice the presence of nodal circles at the
roots of the Bessel functions and nodal lines that cut through the origin at
the roots of cos mφ. When extended in the z direction, these become nodal
cylinders and planes respectively. For the mln mode, there are m nodal
planes, l nodal cylinders, and an additional n horizontal nodal planes along
the z direction.
For pressure release boundaries, the boundary conditions on the pressure

are:
1 1
l = 2,m = 0
l = 1,m = 0
0 0
1 1
−1 0 −1 0
−1 −1
0 y 0 y
x 1 −1 x 1 −1
1 1
l = 1,m = 1
l = 2,m = 1
0 0
1 1
−1 0 −1 0
−1 −1 y
0 y 0
x 1 −1 x 1 −1
Figure 5.6: Cross-sections in the xy-plane of several normal modes for a rigid
cylindrical cavity with a = 1.
p = 0 at r = a
p = 0 at z = 0, L
Applying the first boundary condition yields:
0 = (a cos ωt + b sin ωt)Jm (kr a)

which can only be valid at all values of φ, z, and t if:
Jm (kr a) = 0 ⇒ kr a = Uml ⇒ krml = Uml /a (5.21)

5.3. SPHERICAL CAVITY 117
where Uml is the l’th root of Jm . Figure 5.5 shows the first two for m = 0.
The results for the second boundary condition are the same as for a pressure
release rectangular cavity, so the normal modes are given by:
p(r, φ, z, t) = Jm (krml r)(c cos mφ + d sin mφ)

sin kzn z(a cos ωmln t + b sin ωmln t) (5.22)
nπ
krml = Uml /a kzn = (5.23)
s L
2
Uml nπ 2
ωmln = v + (5.24)
a L
The first few are illustrated in Figure 5.7. Each has m nodal planes through
the origin, l − 1 nodal cylinders, and n − 1 horizontal nodal planes along the
z direction.
5.3 Spherical Cavity

For a spherical cavity of the type shown in Figure 5.8, boundary conditions
are applied to the standing wave solutions for spherical coordinates (see Sub-
section 2.2.2):
|m|
p(r, θ, φ, t) = (a cos ωt + b sin ωt)Pl (cos θ)
[f jl (kr) + gnl (kr)](c cos mφ + d sin mφ)
where:
ω/v = k
If the boundary is rigid, ur = 0 at r = a. Euler’s equation in spherical
coordinates is given by:
∂ur ∂uθ ∂uφ ∂p 1 ∂p 1 ∂p

ρ r̂ + ρ θ̂ + ρ φ̂ = − r̂ − θ̂ − φ̂ (5.25)
∂t ∂t ∂t ∂r r ∂θ r sin θ ∂φ
The corresponding boundary condition on the pressure is therefore:

1 1
l = 2,m = 0
l = 1,m = 0
0 0
1 1
−1 0 −1 0
−1 −1
0 y 0 y
x 1 −1 x 1 −1
1 1
l = 1,m = 1
l = 2,m = 1
0 0
1 1
−1 0 −1 0
−1 −1 y
0 y 0
x 1 −1 x 1 −1
Figure 5.7: Cross-sections in the xy-plane of several normal modes for a

pressure release cylindrical cavity with a = 1.
∂p
= 0 at r = a
∂r
Because the spherical Neumann functions explode at r = 0, g = 0 in the
standing wave solution. Applying the boundary condition therefore yields:
|m|
0 = (a cos ωt + b sin ωt)Pl (cos θ)kjl0 (ka)(c cos mφ + d sin mφ)
where jl0 is the derivative of jl . This can only be valid at all values of θ, φ,
and t if:
jl0 (ka) = 0 ⇒ ka = u0ln ⇒ kln = u0ln /a (5.26)

Figure 5.8: Geometry for a spherical cavity.
where u0ln is the n’th root of the derivative of jl , which is also the n’th
extremum of jl . Figure 5.9 shows the first two for l = 0. The normal modes
for a rigid spherical cavity are therefore given by:
|m|
p(r, θ, φ, t) = jl (kln r)Pl (cos θ)(c cos mφ + d sin mφ)
(a cos ωln t + b sin ωln t) (5.27)
u0
kln = u0ln /a ωln = v ln (5.28)
a
where c and d determine the orientation of the mode about the z-axis, and
are set by initial conditions along with a and b. Figure 5.10 shows the angular
dependence of the solutions for several values of l and m with c = 1 and d = 0.
Notice the presence of nodal planes along φ at the roots of cos mφ and nodal
1.5
u01 u001 u02 u002
1
j0(u)
0.5
−0.5
0 1 2 3 4 5 6 7 8 9 10
u
Figure 5.9: The first few roots and extrema for j0 .
cones along θ at the roots of the Associated Legendre polynomials. Figure

5.11 shows the radial dependence specified by the spherical Bessel functions
for several values of l and n. Each root corresponds to the presence of a
nodal sphere at that radius.
For a pressure release boundary, applying the boundary condition p = 0
at r = a yields:
|m|
0 = (a cos ωt + b sin ωt)Pl (cos θ)jl (ka)(c cos mφ + d sin mφ)
which can only be valid at all values of θ, φ, and t if:
jl (ka) = 0 ⇒ ka = uln ⇒ kln = uln /a (5.29)

where uln is the n’th root of jl . Figure 5.9 shows the first two for l = 0. The
normal modes are therefore:
1 1
l = 1,m = 0
l = 1,m = 1
0 0
z
z
−1 −1
−1 1 −1 1
0 0 0 0
x 1 −1 y x 1 −1 y
2 0.5
l = 2,m = 0
l = 2,m = 1
0 0
z
−2 −0.5
−2 2 −0.5 0.5
0 0 0 0
x 2 −2 y x 0.5 −0.5 y
Figure 5.10: Angular dependence of several normal modes for a spherical

cavity.
|m|
p(r, θ, φ, t) = jl (kln r)Pl (cos θ)(c cos mφ + d sin mφ)
(a cos ωln t + b sin ωln t) (5.30)
uln
kln = uln /a ωln = v (5.31)
a
where the angular dependence is the same as for a rigid spherical cavity.
Figure 5.12 shows the radial dependence for several modes.
1.5
l = 1,n =1
l = 1,n =2
l = 2,n =1
1 l = 2,n =2
jl (kln r)
0.5
−0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r
Figure 5.11: Radial dependence of several normal modes for a rigid spherical
cavity with a = 1.
5.4 Rectangular Waveguide

5.4.1 Acoustic
Figure 5.13 shows a rectangular waveguide. Because waves are free to prop-
agate along the z-axis, boundary conditions are applied to the separable
solutions in rectangular coordinates comprised of traveling wave functions
along that direction and standing wave functions along the other two:
p(x, y, z, t) = Ãei(±kz z−ωt) (c cos kx x + d sin kx x)(f cos ky y + g sin ky y)

where:
(ω/v)2 = k 2 = kx2 + ky2 + kz2

5.4. RECTANGULAR WAVEGUIDE 123
1.5
l = 1,n =1
l = 1,n =2
l = 2,n =1
1 l = 2,n =2
jl (kln r)
0.5
−0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r
Figure 5.12: Radial dependence of several normal modes for a pressure release
spherical cavity with a = 1.
For rigid boundaries, applying the boundary conditions at x = 0, Lx and y =

0, Ly is identical to the process for a rigid rectangular cavity. Consequently,
the normal modes for the waveguide are given by:
p(x, y, z, t) = Ãei(±kz z−ωlm t) cos kxl x cos kym y (5.32)

lπ mπ
kxl = kym = (5.33)
Lx Ly
s
2 2
lπ mπ
ωlm = v + + kz2 (5.34)
Lx Ly
Consider the case where m = 0. Using Euler’s identity, the normal mode
for the wave propagating in the +z direction can be rewritten in the following
Figure 5.13: Geometry for a rectangular waveguide.
manner:
1 1
Ãei(kz z−ωl0 t) cos kxl x = Ãei(kz z+kxl x−ωl0 t) + Ãei(kz z−kxl x−ωl0 t)
2 2
which represents two interfering plane waves propagating at the angles:
±kxl kxl
θ = tan−1 = ± tan−1
kz kz
with respect to the z-axis as illustrated in Figure 5.14. The second wave
is the reflection of the first off of the boundary at x = Lx , and the first is
the reflection of the second off of the boundary at x = 0. Consequently,
this normal mode simply represents an individual plane wave propagating
through the waveguide at angle θ while constructively interfering with its
reflections. Each value of l corresponds to a plane wave propagating at a

different angle. Other angles are disallowed because the interference between
the plane wave and its reflections is destructive.
Figure 5.14: Wavevectors for a plane wave and its reflection propagating
through a rectangular waveguide.
For the case where m 6= 0, the expansion of the modes becomes:
1
Ãei(kz z−ωlm t) cos kxl x cos kym y = Ãei(kz z+kxl x+kym y−ωlm t)
4
1 i(kz z+kxl x−kym y−ωlm t) 1 i(kz z−kxl x+kym y−ωlm t)
+ Ãe + Ãe
4 4
1 i(kz z−kxl x−kym y−ωlm t)
+ Ãe
4
Figure 5.15 shows the wavevectors for the plane waves. The reflection for
each wavevector at one of the boundaries is one of the other three wavevec-
tors. For example, when the wavevector in the first quadrant encounters the
boundary at x = Lx , its reflection is the wavevector in the second quadrant.
Alternatively, when it encounters the boundary at y = Ly , its reflection is
the wavevector in the fourth quadrant. Each mode therefore represents a
single plane wave constructively interfering with its reflections as it propa-
gates through the waveguide. Different combinations of l and m correspond
to different possible wavevector orientations.
Figure 5.15: Wavevectors for a plane wave and its reflections propagating
through a rectangular waveguide.
As illustrated in the figure, the angle of the wavevector for mode lm with
respect to the z-axis is given by:
r 2 2
q
lπ mπ
2
kxl 2
+ kym Lx
+ Ly
θ = sin−1 = sin−1 (5.35)
k ω/v
The angle increases as the mode number increases until it reaches 90 de-
grees and the wave no longer propagates. Consequently, there is a limited
number of modes that can travel through a waveguide for a given wavenum-
ber/frequency. The frequency below which waves won’t propagate for a given
mode is called the cutoff frequency for that mode. From the above equation:
r 2 2
lπ mπ
Lx
+ Ly
sin(π/2) = 1 =
ω/v
s 2 2
lπ mπ
ωclm = v + (5.36)
Lx Ly
The propagation speed for the combined wave through the waveguide,
also called the phase speed, varies by mode:
ω ω
vplm = =q
kz 2
k 2 − kxl 2
− kym
ω
=r 2 2 (5.37)
(ω/v)2 − Llπx + mπ
Ly
v
=p (5.38)
1 − (ωclm /ω)2
Counterintuitively, its value is greater than v and increases with mode num-
ber as the propagation angle increases. However, it does not represent the
speed at which energy propagates through the waveguide. That quantity is
determined by the group velocity, which is the speed at which the constituent
reflecting plane waves travel through the channel:
kz
vglm = v cos θ = v
k
r 2 2
lπ mπ
(ω/v)2 − Lx
+ Ly
=v (5.39)
ω/v
p
=v 1 − (ωclm /ω)2 (5.40)
As expected, its value is less than v and decreases with increasing mode
number as the propagation angle increases.
For a waveguide with pressure release boundaries, the boundary condi-
tions at x = 0, Lx and y = 0, Ly change to those for a pressure release
rectangular cavity. The normal modes therefore become:
p(x, y, z, t) = Ãei(±kz z−ωlm t) sin kxl x sin kym y (5.41)

lπ mπ
kxl = kym = (5.42)
Lx Ly
s
2 2
lπ mπ
ωlm = v + + kz2 (5.43)
Lx Ly
where the cosine functions have become sine functions that introduce neg-
ative signs into the plane wave expansion, indicating that they undergo a
180 degree phase change when reflected at the pressure release boundaries.
Otherwise, the results are identical.
5.4.2 Electromagnetic
For electromagnetic waves in a rectangular waveguide, the initial wave func-
tion for each component of the electric field is the same as for acoustic pres-
sure:
Es (x, y, z, t) = Ãs ei(±kz−ωt) (cs cos kx x + ds sin kx x)(fs cos ky y + gs sin ky y)

s = x, y, or z
If the walls of the waveguide are conducting and have the dimensions shown
in Figure 5.13, the boundary condition on the electric field (E k = 0) requires:
Ex = 0 at y = 0, Ly
Ey = 0 at x = 0, Lx
Ez = 0 at x = 0, Lx and y = 0, Ly
Applying these conditions in the same manner as for a conducting rectangular

cavity yields:
Ex = Ãx ei(±kz z−ωlm t) (cx cos kxl x + dx sin kxl x) sin kym y
Ey = Ãy ei(±kz z−ωlm t) sin kxl x(fy cos kym y + gy sin kym y)
Ez = Ãz ei(±kz z−ωlm t) sin kxl x sin kym y
lπ mπ
kxl = kym =
Lx Ly
s
2 2
lπ mπ
ωlm = v + + kz2
Lx Ly
Forcing these solutions to be divergenceless:
0 = ∇ · E = Ãx ei(±kz z−ωlm t) (−cx kxl sin kxl x + dx kxl cos kxl x) sin kym y
+ Ãy ei(±kz z−ωlm t) sin kxl x(−fy kym sin kym y + gy kym cos kym y)
± ikz Ãz ei(±kz z−ωlm t) sin kxl x sin kym y
For this to be valid at all values of x, y, z and t, the functions in each of

the summed terms must be identical. Consequently, dx = gy = 0 and the
equation becomes:
0 = −Ãx kxl ei(±kz z−ωlm t) sin kxl x sin kym y

− Ãy kym ei(±kz z−ωlm t) sin kxl x sin kym y
± iÃz kz ei(±kz z−ωlm t) sin kxl x sin kym y
where the remaining constants have been absorbed into Ãx and Ãy . Dividing
by the common functions yields:
0 = Ãx kxl + Ãy kym ∓ iÃz kz

which implies that one of the remaining constants is determined by the values
of the other two.
The corresponding magnetic field is obtained from Faraday’s law:
∂B
∇×E =− = iωB
∂t
−i ∂Ez ∂Ey −i ∂Ex ∂Ez −i ∂Ey ∂Ex
Bx = − By = − Bz = −
ω ∂y ∂z ω ∂z ∂x ω ∂x ∂y
Inserting the solution for the electric field and using −i = e−iπ/2 :
1
Bx = (Ãz kym ∓ iÃy kz )ei(±kz z−ωlm t−π/2) sin kxl x cos kym y
ωlm
1
By = (±iÃx kz − Ãz kxl )ei(±kz z−ωlm t−π/2) cos kxl x sin kym y
ωlm
1
Bz = (Ãy kxl − Ãx kym )ei(±kz z−ωlm t−π/2) cos kxl x cos kym y
ωlm
These solutions also satisfy the remaining Maxwell’s equations and the
boundary condition on the magnetic field (B ⊥ = 0):
Bx = 0 at x = 0, Lx
By = 0 at y = 0, Ly
For the TE solutions, Ãz = 0 implying that Ãy = −(kxl /kym )Ãx from
Gauss’ law. Making these substitutions yields:
Ex = Ãx ei(±kz z−ωlm t) cos kxl x sin kym y

kxl
Ey = − Ãx ei(±kz z−ωlm t) sin kxl x cos kym y (5.44)
kym
Ez = 0
The corresponding magnetic field components are given by:

±i kz kxl
Bx = Ãx ei(±kz z−ωlm t−π/2) sin kxl x cos kym y
ωlm kym
±i
By = kz Ãx ei(±kz z−ωlm t−π/2) cos kxl x sin kym y (5.45)
ωlm
2 2
−1 kxl + kym
Bz = Ãx ei(±kz z−ωlm t−π/2) cos kxl x cos kym y
ωlm kym
For the TM solutions, Ãy = (kym /kxl )Ãx . In addition, from Gauss’ law:
2
kym
0 = Ãx kxl + Ãx ∓ iÃz kz
kxl
2 2
kxl + kym
Ãz = ∓i Ãx
kxl kz
The components of the electric field are therefore given by:
Ex = Ãx ei(±kz z−ωlm t) cos kxl x sin kym y

kym
Ey = Ãx ei(±kz z−ωlm t) sin kxl x cos kym y (5.46)
kxl
2 2
kxl + kym
Ez = ∓i Ãx ei(±kz z−ωlm t) sin kxl x sin kym y
kxl kz
and the components of the magnetic field by:
2 2
∓i kxl + kym + kz2
Bx = kym Ãx ei(±kz z−ωlm t−π/2) sin kxl x cos kym y
ωlm kxl kz
2 2
±i kxl + kym + kz2
By = Ãx ei(±kz z−ωlm t−π/2) cos kxl x sin kym y (5.47)
ωlm kz
Bz = 0
Because the allowed wavevectors for electromagnetic and acoustic waves in

a rectangular waveguide are identical, the cutoff frequencies, phase speeds,
and group velocities are all the same.
5.5 Cylindrical Waveguide

Figure 5.16 shows a cylindrical waveguide. The radial boundary condition is
applied to the separable solutions for cylindrical coordinates with standing
wave functions along that direction and traveling wave functions along the z
direction:
p(r, φ, z, t) = Ãei(±kz z−ωt) Jm (kr r)(c cos mφ + d sin mφ)
where the Neumann functions have been excluded because they explode at
the origin and:
(ω/v)2 = k 2 = kr2 + kz2
Figure 5.16: Geometry for a cylindrical waveguide.

5.5. CYLINDRICAL WAVEGUIDE 133
For rigid boundaries, applying the boundary condition at r = a is identical

to the process for a rigid cylindrical cavity. The normal modes are therefore
given by:
p(r, φ, z, t) = Ãei(±kz z−ωml t) Jm (krml r)(c cos mφ + d sin mφ) (5.48)

0
krml = Uml /a (5.49)
s 2
0
Uml
ωml = v + kz2 (5.50)
a
where c and d determine the orientation of the mode about the z-axis.
Although the waves that propagate throughq the waveguide are more com-
2 2 . Consequently, the
plex than plane waves, krml is equivalent to kxl + kym
cutoff frequency is:
0
Uml
ωcml = v (5.51)
a
and the phase speed/group velocity are:
ω
vpml = r 2 (5.52)
0

Uml
(ω/v)2 − a
v
=p (5.53)
1 − (ωcml /ω)2
r 0 2
2 U
(ω/v) − aml
vgml =v (5.54)
ω/v
p
= v 1 − (ωcml /ω)2 (5.55)
For a cylindrical waveguide with a pressure release boundary, the radial

boundary condition changes to that for a pressure release cylindrical cavity.
The normal modes therefore become:
p(r, φ, z, t) = Ãei(±kz z−ωml t) Jm (krml r)(c cos mφ + d sin mφ) (5.56)

krml = Uml /a (5.57)
s 2
Uml
ωml = v + kz2 (5.58)
a
The expressions for the cutoff frequency, phase speed and group speed re-
main the same, but with the Bessel function roots (Uml ) replacing the Bessel
0
function extrema (Uml ).
Chapter 6
Radiation
Section 1.10 describes two different methods for finding the radiated wave
produced by a harmonic source in one dimension. The first is to treat the sur-
face of the source as a medium boundary, and impose boundary conditions on
separable solutions to the wave equation that are consistent with the behav-
ior of the source. The next section extends that method to three dimensions,
using pulsating (expanding and contracting) and vibrating (moving side to
side) spherical sources that produce acoustic waves as examples.
The second method is to adapt the wave equation to account for the
presence of the source and solve it. The remainder of the chapter addresses
this approach in three dimensions. It begins by discussing how the wave
equation can be solved when it includes a source term by recasting it in
integral form. It then applies the method to a series of point sources and
extended sources.
6.1 Boundary Condition Method

Applying boundary conditions to the separable solutions of the wave equa-
tion in a given coordinate system requires the surface of the boundary to
correspond to a constant spatial coordinate. In spherical coordinates, for ex-
ample, the boundary must be spherical (r =constant), conical (θ =constant)
or planar (φ =constant). Consequently, the boundary condition method for
finding radiated wave functions is limited to sources with simple, symmetrical
shapes such as spheres, cylinders and cones.
135
136 CHAPTER 6. RADIATION
6.1.1 Pulsating Sphere

As a first example, consider a harmonically pulsating sphere radiating acous-
tic waves into a fluid. Because the radiator has a spherical shape and the
waves it produces are free to propagate, boundary conditions are applied to
the traveling wave solutions in spherical coordinates:
X (1) (2)
p(r, θ, φ, t) = Ãlm e−iωt [f˜lm hl (kr) + g̃lm hl (kr)]
l,m
|m|
Pl (cos θ)(c̃lm eimφ + d˜lm e−imφ )
The radiated waves only propagate outward, so g̃lm = 0 leaving:
|m|
X (1)
p(r, θ, φ, t) = Ãlm e−iωt hl (kr)Pl (cos θ)(c̃lm eimφ + d˜lm e−imφ )
l,m
where f˜lm has been absorbed into Ãlm . Assuming that the pulsating sphere is
centered on the origin with radius a as illustrated in Figure 6.1, the velocity
of its surface is given by:
u(a, θ, φ, t) = ũ0 e−iωt r̂ (6.1)

The velocity of the surrounding fluid must be identical. This boundary condi-
tion on the fluid velocity can be applied to the pressure using Euler’s equation
in spherical coordinates:
∂ur ∂uθ ∂uφ ∂p 1 ∂p 1 ∂p
ρ r̂ + ρ θ̂ + ρ φ̂ = − r̂ − θ̂ − φ̂
∂t ∂t ∂t ∂r r ∂θ r sin θ ∂φ
Plugging the expressions for fluid velocity and pressure at the sphere’s surface
into the radial component:
X 0 (1) |m|
−iωρũ0 e−iωt r̂ = − Ãlm e−iωt khl (ka)Pl (cos θ)(c̃lm eimφ + d˜lm e−imφ )r̂
l,m
0 (1) (1)
where hl is the derivative of hl . Because the fluid velocity has no angular
dependence, this equation can only be valid if l = m = 0. Solving what
remains for the amplitude of the pressure:
6.1. BOUNDARY CONDITION METHOD 137
Figure 6.1: Geometry for a pulsating sphere.
iρvũ0
Ã00 = 0 (1)
h0 (ka)
Consequently, the wave function for the radiated pressure is given by:
iρvũ0 (1)
p(r, θ, φ, t) = 0 (1) h0 (kr)e−iωt
h0 (ka)
(1)
Using h0 (kr) = −ieikr /kr, the solution can be rewritten as:
ei(kr−ωt)
ρvũ0
p(r, θ, φ, t) = 0 (1) (6.2)
h0 (ka) kr
which is a simple, outgoing spherical wave. Substituting for the derivative of
the spherical Hankel function as well yields:
(ka)2 −ika ei(kr−ωt)

p(r, θ, φ, t) = ρvũ0 e
ka + i kr
For small spheres where ka 1, this becomes:
ei(kr−ωt)
p(r, θ, φ, t) ≈ −iρvũ0 (ka)2 (6.3)
kr
6.1.2 Vibrating Sphere

Next, consider a vibrating sphere. If the coordinate system is oriented such
that the sphere vibrates along the z-axis as illustrated in Figure 6.2, the
surface velocity of the sphere is given by:
u(a, θ, φ, t) = ũ0 e−iωt ẑ = ũ0 e−iωt cos θr̂ − ũ0 e−iωt sin θθ̂ (6.4)
where the unit vector ẑ has been rewritten in terms of its spherical compo-
nents. Assuming that the surface of the sphere is smooth, only the radial
component of its velocity causes fluid motion. The fluid velocity at the sur-
face of the sphere must therefore be:
u(a, θ, φ, t) = ũ0 cos θe−iωt r̂

Plugging this boundary condition into the radial component of Euler’s equa-
tion along with the separable solution for the pressure yields:
X 0 (1) |m|
−iωρũ0 cos θe−iωt = − Ãlm e−iωt khl (ka)Pl (cos θ)(c̃lm eimφ + d˜lm e−imφ )
l,m
For the angular dependence on the left and right to match, l = 1 and m = 0.
Solving for the amplitude of the remaining term in the sum:
iρvũ0
Ã10 = 0 (1)
h1 (ka)
The wave function for the radiated pressure is therefore:
iρvũ0 (1)
p(r, θ, φ, t) = 0 (1) h1 (kr) cos θe−iωt
h1 (ka)
Given that:
Figure 6.2: Geometry for a vibrating sphere.

(1) 1 i
h1 (kr) =− + eikr
kr (kr)2
where the second term in the sum can be dropped in the far-field where
kr 1. Substituting this simplified form into the solution yields:
iρvũ0 ei(kr−ωt)
p(r, θ, φ, t) ≈ − 0 (1) cos θ (6.5)
h1 (ka) kr
which is a spherical wave. Plugging in the derivative of the spherical Hankel
function:
(ka)3 ei(kr−ωt)

−ika
p(r, θ, φ, t) ≈ −iρvũ0 e cos θ
2i + 2(ka) − i(ka2 ) kr
For small spheres where ka 1:
1 ei(kr−ωt)
p(r, θ, φ, t) ≈ − ρvũ0 (ka)3 cos θ (6.6)
2 kr
6.1.3 Sources Near Boundaries

The boundary condition method can also be used to find the radiated wave
function for a source located near physical boundaries. However, the bound-
ary conditions for the source must be applied to separable solutions that also
satisfy the boundary conditions for the physical boundaries. For example, if
a source is located inside a rectangular waveguide, the boundary conditions
for the source must be applied to the solutions derived in Section 5.4.
6.2 Inhomogeneous Wave Equation Method

More complex source shapes are amenable to the alternative method of
adapting the wave equation to account for the presence of the source, al-
though finding solutions usually requires making approximations. In general,
the adapted scalar wave equation takes the form:
1 ∂ 2ψ
∇2 ψ − = −f (r, t) (6.7)
v 2 ∂t2
where the exact expression for f depends upon the type of wave under con-
sideration. For example, when a string is driven by a source, it represents
the ratio of the force per unit length applied by the source to the tension
in the string (see Equation 1.39). For harmonic sources, in particular, the
equation becomes:
1 ∂ 2ψ
∇2 ψ − 2 2
= −f (r)e−iωt (6.8)
v ∂t
The first step in solving it, as in Section 2.2 for the sourceless wave equation,
is to remove the time dependence by assuming a separable solution of the
form:
ψ(r, t) = ψω (r)e−iωt (6.9)

Plugging it in yields the Helmholtz equation with a source term:
6.2. INHOMOGENEOUS WAVE EQUATION METHOD 141
∇2 ψω (r) + k 2 ψω (r) = −f (r) (6.10)

The approach typically used to solve this equation is to first find the solution
for a point source with unit amplitude positioned arbitrarily at r 0 . The source
term becomes f (r) = δ(r − r 0 ) and the radiated wave function ψω (r) =
Gk (r, r 0 ) is called a Green’s function. Once the Green’s function is known,
it can be used to find the wave function for an arbitrary source term by
integrating it with the Helmholtz equation for the point source:
Z Z
f (r )[∇ Gk (r, r ) + k Gk (r, r )]dr = −f (r 0 )δ(r − r 0 )dr 0
0 2 0 2 0 0
Z Z
2 0 0 0 2 0 0 0
∇ f (r )Gk (r, r )dr + k f (r )Gk (r, r )dr = −f (r)
where this is just the Helmholtz equation for the arbitrary source with the
solution:
Z
ψω (r) = f (r 0 )Gk (r, r 0 )dr 0 (6.11)
The Helmholtz equation has many Green’s functions, but the one that is
used for a particular source must satisfy the boundary conditions that are
relevant for that source. For example, if the source is located inside a cavity
or waveguide, the Green’s function must satisfy the boundary conditions for
the walls. Such Green’s functions can often be found or constructed, but the
scope of this book is limited to sources in what’s called free space where no
physical boundaries are present. The Green’s function in this case, which is
called the free space Green’s function, must approach zero as r approaches
infinity and is given by:
0
0 eik|r−r |
Gk (r , r) = (6.12)
4π|r − r 0 |
Notice that this is simply the wave function for an isotropic, outgoing spher-
ical wave centered at r 0 . Plugging it into Equation 6.11 yields:
0
eik|r−r |
Z
0
ψω (r) = f (r ) 0
dr 0 (6.13)
4π|r − r |
which implies that any radiated wave can be viewed as a continuous collection
of interfering isotropic spherical waves whose amplitudes are determined by
the value of the source term.
For acoustics, the source term is given by:
∂
f (r, t) = [ρq(r, t)] (6.14)
∂t
where q(r, t) represents the volume flow density produced by the source,
and is the amount of fluid moved per unit time per unit volume by the
source at location r and time t. Given that the density of the fluid is ρ, the
product ρq(r, t) is simply the mass flow density produced by the source. For
a harmonic source:
∂
f (r, t) = [ρq(r)e−iωt ] = −iωρq(r)e−iωt (6.15)
∂t
such that the solution to the Helmholtz equation becomes:
0
eik|r−r |
Z
0
pω (r) = −iωρ q(r ) 0
dr 0 (6.16)
4π|r − r |
The vector wave equation with a source term takes the form:
1 ∂ 2ψ
∇2 ψ − = −f (r, t) (6.17)
v 2 ∂t2
which can be separated into three independent scalar wave equations corre-
sponding to the rectangular coordinates. Each can be solved for a harmonic
source using the Green’s function approach, and the results recombined to
produce:
0
eik|r−r |
Z
0
ψω (r) = f (r ) dr 0 (6.18)
4π|r − r 0 |
For radiated electromagnetic waves, it is generally best to calculate the
wave functions for the electric and magnetic fields from the vector potential
because the results are guaranteed to be divergenceless. The source term for
the vector potential is given by:
f (r, t) = µJ (r, t) (6.19)

where J (r, t) is the current density (electric current per unit volume) pro-
duced by the source at location r and time t. For a harmonic source:
f (r, t) = µJ (r)e−iωt (6.20)
such that:
0
eik|r−r |
Z
0
Aω (r) = µ J (r ) dr 0 (6.21)
4π|r − r 0 |
Once the integral is evaluated, the electric and magnetic fields can be cal-
culated outside the source using the definition of the vector potential and
Ampere’s law:
iv 2
B = ∇ × A and E = ∇×B
ω
6.2.1 Radiation by Point Sources

Equations 6.13 and 6.18 are called radiation integrals and can be evaluated or
approximated for many different sources. The solutions for several significant
combinations of point sources will be explored first.
Monopoles
A monopole is a single, harmonic point source with amplitude F̃ (or F̃ for a
vector source) called the monopole strength. Its source term is therefore:
f (r) = F̃ δ(r − r0 ) (6.22)
which returns the free space Green’s function when inserted into the radiation
integral:
F̃ eik|r−r0 |
ψω (r) = (6.23)
4π |r − r0 |
F̃ ei(k|r−r0 |−ωt)
ψ(r, t) = (6.24)
4π |r − r0 |
For a monopole situated at the origin, the wave function becomes:

F̃ eikr
ψω (r) = (6.25)
4π r
F̃ ei(kr−ωt)
ψ(r, t) = (6.26)
4π r
Dipoles
A dipole is defined as a pair of interfering monopoles that oscillate 180 degrees
out of phase. Its wave function is therefore the sum of two monopole wave
functions:
F̃ ei(k|r−r1 |−ωt) F̃ ei(k|r−r2 |−ωt−π)

ψ(r, t) = +
4π |r − r1 | 4π |r − r2 |
F̃ ei(k|r−r1 |−ωt) F̃ ei(k|r−r2 |−ωt)
= −
4π |r − r1 | 4π |r − r2 |
Centering the dipole on the origin as illustrated in Figure 6.3, the far-field
approximation can be applied as in Section 3.2 when r d, λ. Using a
distance vector d that points from the bottom monopole to the top monopole,
the distances in the complex exponentials become:
|r − r1 | ≈ r − r̂ · r1 = r − r̂ · d/2
|r − r2 | ≈ r − r̂ · r2 = r − r̂ · −d/2
and the amplitudes are replaced by the amplitude for a monopole centered
on the origin such that:
F̃ i(kr−kr̂·d/2−ωt) F̃ i(kr+kr̂·d/2−ωt)
ψ(r, t) ≈ e − e
4πr 4πr
F̃ ikr̂·d/2
≈− (e − e−ikr̂·d/2 )ei(kr−ωt)
4πr
Using Euler’s identity to simplify the term in parentheses:
F̃
ψ(r, t) ≈ − 2i sin(kr̂ · d/2)ei(kr−ωt) (6.27)
4πr
Figure 6.3: Geometry for a dipole source.
If the monopoles are close relative to the wavelength (kd 1), the argument
of the sine function is small and the small-angle approximation (sin α ≈ α)
can be applied:
ei(kr−ωt)
ψ(r, t) ≈ −ikr̂ · (F̃ d) (6.28)
4πr
where the product (F̃ d) is called the dipole moment.
The vector d and associated dipole moment are not required to point
along the z-axis as illustrated in Figure 6.3, but considering that case as an
example:
r̂ · (F̃ d) = sin θ cos φx̂ + sin θ sin φŷ + cos θẑ · F̃ dẑ = F̃ d cos θ
and:
ei(kr−ωt)
ψ(r, t) ≈ −ik(F̃ d) cos θ
4πr
As discovered in Chapter 3, interfering spherical waves always produce a
spherical wave in the far-field with an angular-dependent amplitude term
determined by their phase differences. The normalized interference pattern,
which is simply the squared angular dependence of the amplitude (cos2 θ), is
plotted in Figure 6.4.
0.5
0
z
−0.5
−1
−0.4 0.4
−0.2 0.2
0 0
0.2 −0.2
0.4 −0.4
y
x
Figure 6.4: Normalized interference pattern for a dipole in the far-field when
kd 1.
Quadrupoles
A quadrupole is a pair of interfering dipoles arranged either longitudinally or
laterally as illustrated in Figures 6.5 and 6.6. The quadrupole wave function
is therefore the sum of two shifted dipole wave functions. Using the geometry
in the figures and Equation 6.28 for the dipole wave functions:
− r1 · (F̃ d) ei(k|r−r1 |−ωt) −ik r d

−ik r d − r2 · (−F̃ d) ei(k|r−r2 |−ωt)
ψ(r, t) ≈ +
4π |r − r1 | 4π |r − r2 |
Figure 6.5: Geometry for a longitudinal quadrupole source.
where the moment for the second dipole includes a negative sign because
its orientation is reversed. For the longitudinal quadrupole, the far-field
distances in the complex exponentials are the same as for a dipole. Conse-
quently:
Figure 6.6: Geometry for a lateral quadrupole source.
ikr̂ · (F̃ d) ikr̂·d/2

ψ(r, t) ≈ (e − e−ikr̂·d/2 )ei(kr−ωt)
4πr
2kr̂ · (F̃ d)
≈− sin(kr̂ · d/2)ei(kr−ωt)
4πr
Because kd 1, this simplifies to:
ei(kr−ωt)
ψ(r, t) ≈ −k 2 (r̂ · d)2 F̃ (6.29)
4πr
Using a second distance vector d⊥ that points from the middle of the second
dipole to the middle of the first in Figure 6.6, the far-field distances in the
complex exponentials for the lateral quadrupole become:
|r − r1 | ≈ r − r̂ · r1 = r − r̂ · d⊥ /2
|r − r2 | ≈ r − r̂ · r2 = r − r̂ · −d⊥ /2
such that:
ikr̂ · (F̃ d) ikr̂·d⊥ /2

ψ(r, t) ≈ (e − e−ikr̂·d⊥ /2 )ei(kr−ωt)
4πr
2kr̂ · (F̃ d)
≈− sin(kr̂ · d⊥ /2)ei(kr−ωt)
4πr
Because kd 1:
ei(kr−ωt)
ψ(r, t) ≈ −k 2 (r̂ · d)(r̂ · d⊥ )F̃ (6.30)
4πr
The two distance vectors aren’t required to point along the z-axis and
x-axis as illustrated in Figures 6.5 and 6.6 as long as they are perpendicular,
but considering that case as an example:
r̂ · d = d cos θ
r̂ · d⊥ = d sin θ cos φ
such that:
ei(kr−ωt)
ψlong (r, t) ≈ −k 2 (F̃ d2 ) cos2 θ
4πr
ei(kr−ωt)
ψlat (r, t) ≈ −k 2 (F̃ d2 ) cos θ sin θ cos φ
4πr
where (F̃ d2 ) is called the quadrupole moment. The normalized interference
patterns for both are plotted in Figure 6.7.
6.2.2 Radiation by Extended Sources

Direct evaluation of the radiation integrals for extended sources is difficult
in most cases. An alternative approach is to rewrite the free space Green’s
1 1
0.5 0.5
0 0
z
z
−0.5 −0.5
−1 −1
−0.5 0.5 −1 0.5
0 0 0 0
y y
x 0.5 −0.5 x 1 −0.5
Figure 6.7: Normalized interference patterns for quadrupoles in the far-field

when kd 1.
function within the integrals as a linear combination of the spherical solutions

to the wave function, although the result is often just as difficult to evaluate.
The scalar integral, for example, becomes:
X Z
(1)
ψω (r) = ik hl (kr)Ylm (θ, φ) f (r 0 )jl (kr0 )Ylm∗ (θ0 , φ0 )dr 0 (6.31)
l,m
A simpler approach is to apply the far-field approximation to the free space

Green’s function instead, but the result is only valid at large distances relative
to the size of the source and the wavelength (r r0 , λ):
eikr
Z Z
1 0 ik(r−r̂·r 0 ) 0 0
ψω (r) ≈ f (r )e dr = f (r 0 )e−ik·r dr 0 (6.32)
4πr 4πr
where:
k = kr̂ = k sin θ cos φx̂ + k sin θ sin φŷ + k cos θẑ

Because the radiation integral represents a continuous collection of interfer-
ing spherical waves, the result is also a spherical wave where the angular
dependence of the amplitude is determined by the integral over the source
term. This integral can generally be evaluated, and solutions will be derived
for several acoustic and electromagnetic examples.
Pulsating Sphere
Consider the pulsating sphere illustrated in Figure 6.1 with surface velocity
given by Equation 6.1. Because the sphere only moves fluid at its surface,
the volume flow density it creates is given by:
q(r, t) = q(θ, φ, t)δ(r − a)

where q(θ, φ, t) is the volume flow per unit area at the surface:
Volume Flow Radial Surface Velocity × Area

q(θ, φ, t) = = = ũ0 e−iωt
Area Area
Plugging this into the radiation integral for acoustic sources (Equation 6.16)
with the far-field approximation applied yields:
eikr
Z
0
pω (r) ≈ −iωρũ0 δ(r0 − a)e−ik·r dr 0
4πr
Because the pulsating sphere possesses spherical symmetry, k can be aligned
with the z 0 -axis without loss of generality such that k·r 0 = k ẑ 0 ·(r0 sin θ0 cos φ0
x̂0 + r0 sin θ0 sin φ0 ŷ 0 + r0 cos θ0 ẑ 0 ) = kr0 cos θ0 . Making this substitution:
eikr 2π π ∞
Z Z Z
0 0
pω (r) ≈ −iωρũ0 δ(r0 − a)e−ikr cos θ (r02 sin θ0 )dr0 dθ0 dφ0
4πr 0 0 0
ikr −ika cos θ0 π
e e
≈ −iωρũ0 (2πa2 )

4πr ika

0
−ika
eikr eika
− e
≈ −iωρũ0 (2πa2 )
4πr ika
ikr
e sin ka
≈ −iωρũ0 (4πa2 )
4πr ka
For small spheres where ka 1, the small angle approximation can be

applied to the sine function producing:
eikr ka
pω (r) ≈ −iρ(vk)ũ0 a2
r ka
ikr
e
≈ −iρvũ0 (ka2 ) (6.33)
r
which is identical to the result obtained earlier using the boundary condition
method. Notice that the pulsating sphere is like a monopole in that the
amplitude of the spherical wave it produces has no angular dependence.
Baffled Piston
A baffled piston is a flat, vibrating surface surrounded by a barrier that con-
tains the fluid medium to the front side as illustrated in Figure 6.8. Because
the radiation is emitted into half the space, the source term is doubled. The
volume flow density is therefore:
q(r, t) = 2q(x, y, t)δ(z)

where:
Volume Flow Surface Velocity × Area

q(x, y, t) = = = ũ0 e−iωt
Area Area
Inserting this into the radiation integral:
Z ∞
eikr
ZZ
0 0 0
pω (r) ≈ −iωρũ0 δ(z 0 )e−i(kx x +ky y +kz z ) dz 0 dx0 dy 0
2πr Piston Surface −∞
ZZ
0 0
≈ ψsph e−i(kx x +ky y ) dx0 dy 0 (6.34)
ZZ Piston Surface
0 0
≈ ψsph e−i(k sin θ cos φx +k sin θ sin φy ) dx0 dy 0
Piston Surface
Notice that the integral is simply the two-dimensional Fourier transform of

the piston surface and can often be evaluated. Several examples involving
different piston shapes will be addressed in the next chapter.
Figure 6.8: A baffled, circular piston.
Vibrating Sphere
Consider the vibrating sphere in Figure 6.2 next with surface velocity given
by Equation 6.4. The volume flow density it creates is given by:
q(r, t) = δ(r − a)ũ0 cos θe−iωt

Plugging this into the radiation integral yields:
eikr
Z
0
pω (r) ≈ −iωρũ0 δ(r0 − a) cos θ0 e−ik·r dr 0
4πr
Because the vibrating sphere possesses azimuthal symmetry, k can be placed
in the x0 z 0 -plane without loss of generality such that k · r 0 = (k sin θx̂0 +
k cos θẑ 0 ) · (r0 sin θ0 cos φ0 x̂0 + r0 sin θ0 sin φ0 ŷ 0 + r0 cos θ0 ẑ 0 ) = (k sin θ)(r0 sin θ0
cos φ0 ) + (k cos θ)(r0 cos θ0 ). Making this substitution:
2π π ∞
eikr
Z Z Z
pω (r) ≈ −iωρũ0 δ(r0 − a) cos θ0
4πr 0 0 0
−i(kr0 sin θ sin θ0 cos φ0 +kr0 cos θ cos θ0 ) 02
e r sin θ0 dr0 dθ0 dφ0
ikr Z 2π Z π
e
≈ −iωρũ0 cos θ0
4πr 0 0
−i(ka sin θ sin θ0 cos φ0 +ka cos θ cos θ0 ) 2
e a sin θ0 dθ0 dφ0
For small spheres where ka 1, the exponential can be replaced with the
first two terms of its Taylor expansion (ex ≈ 1 + x) to simplify the integral:
2π π
eikr
Z Z
pω (r) ≈ −iωρũ0 [a2 cos θ0 sin θ0 − ika3 sin θ sin2 θ0 cos θ0 cos φ0
4πr 0 0
− ika3 cos θ cos2 θ0 sin θ0 ]dθ0 dφ0
The first and second terms both integrate to zero. The third yields:
eikr
pω (r) ≈ −iωρũ0 [−ika3 cos θ(2π)(−1/3 cos3 θ0 )|π0 ]
4πr
1 eikr
≈ − (vk)ρũ0 ka3 cos θ(4π)
3 4πr
ikr
1 e
≈ − ρvũ0 (ka)3 cos θ (6.35)
3 kr
which nearly matches the result obtained earlier using the boundary condi-
tion method. Notice that the angular dependence of the amplitude in this
case is identical to that for a dipole aligned with the z-axis.
Linear Electromagnetic Antenna

For a linear electromagnetic antenna of length d placed in a coordinate system
where it is centered on the origin and oriented along the z-axis, the current
density is given by:
J (r, t) = δ(x)δ(y)I˜0 e−iωt ẑ for − d/2 ≤ z ≤ d/2 (6.36)

where the current carried by the antenna is assumed to vary harmonically

with amplitude I˜0 . Inserting this expression into the radiation integral for
electromagnetic sources (Equation 6.21) with the far-field approximation ap-
plied yields:
ikr Z
e 0
Aω (r) ≈ ẑµI˜0 δ(x)δ(y)e−ik·r dr 0
4πr
eikr d/2 −i(k cos θ)z0 0
Z
˜
≈ ẑµI0 e dz
4πr −d/2
0
eikr e−ikz cos θ d/2
≈ ẑµI˜0
4πr −ik cos θ −d/2

˜ eikr −2i sin(kd/2 cos θ)

≈ ẑµI0
4πr −ik cos θ
ikr
e sin(kd/2 cos θ)
≈ ẑµI˜0 2
4πr k cos θ
For a short antenna where kd 1, the small angle approximation can be
applied to the sine function such that:
ikr
e kd/2 cos θ
Aω (r) ≈ ẑµI˜0 2
4πr k cos θ
µ eikr
≈ ẑ I˜0 d (6.37)
4π r
The far-field spherical wave produced by any antenna can be written in
the form:
eikr
Aω (r) ≈ Ã(θ, φ)
r
Using the vector identity ∇ × (ba) = ∇b × a + b∇ × a:
eikr eikr
Bω = ∇ × A ω = ∇ × Ã(θ, φ) + ∇ × Ã(θ, φ)
r r
∂ eikr eikr

= r̂ × Ã(θ, φ) + ∇ × Ã(θ, φ)
∂r r r
ikr
eikr eikr

e
= ik − 2 r̂ × Ã(θ, φ) + ∇ × Ã(θ, φ)
r r r
In the far-field, any term of order 1/r2 can be ignored. The curl yields a 1/r
term, so only the first term remains:
eikr
Bω ≈ ik r̂ × Ã(θ, φ)
r
eikr
≈ ikr̂ × Ã(θ, φ)
r
≈ ikr̂ × Aω (r) (6.38)
Inserting the vector potential for the linear antenna yields:
µ ˜ eikr µ eikr
Bω ≈ ikr̂ × ẑ I0 d = −i (I˜0 kd) sin θ φ̂ (6.39)
4π r 4π r
Because the wave is spherical, the corresponding electric field is perpendicular
with magnitude Eω = vBω :
µ ˜ eikr
Eω ≈ −i v(I0 kd) sin θ θ̂ (6.40)
4π r
Multipole Expansion
In evaluating the radiation integral for a vibrating sphere, the complex ex-
ponential was expanded to make integration possible. The same can be done
for any source. Inserting the Taylor expansion:
∞
x
X 1 n
e = x
n=0
n!
into the radiation integral with the far-field approximation applied (Equation
6.32) yields:
∞
eikr X 1
Z
ψω (r) ≈ f (r 0 )(−ik · r 0 )n dr 0 (6.41)
4πr n=0 n!
where f is simply replaced by f for a vector wave. Although the sum is

infinite, the higher order terms quickly become negligible as long as the
source is small relative to the wavelength (k · r 0 1).
The first term in the expansion:
eikr
Z
ψω0 (r) = f (r 0 )dr 0
4πr
is the wave function for a monopole centered on the origin with monopole
strength:
Z
F̃ = f (r 0 )dr 0
The second term:
eikr eikr
Z Z
0 0 0
ψω1 (r) = f (r )(−ik · r )dr = (−ik)r̂ · f (r 0 )r 0 dr 0
4πr 4πr
is the wave function for a dipole with dipole moment:
Z
(F̃ d) = f (r 0 )r 0 dr 0
The third term is a sum of lateral and longitudinal quadrupole wave func-
tions, the fourth term a sum of octopole fields, and so on. Equation 6.41 is
therefore referred to as the multipole expansion of the source wave function.
Notice that in the far-field where kr 1, the spherical wave function
approximation for the spherical Hankel functions introduced in Subsection
2.2.2:
(1) (−i)l+1 ikr

hl (kr) ≈ e for kr 1
kr
can be substituted into Equation 6.31:
eikr X
Z
ψω (r) = i (−i) Yl (θ, φ) f (r 0 )jl (kr0 )Ylm∗ (θ0 , φ0 )dr 0
l+1 m
(6.42)
r l,m
producing a result that is very similar to the multipole expansion. In fact,

the angular dependence of the Spherical Harmonics and multipole terms is
nearly identical for l = n, with minor differences due to inaccuracies in the
multipole expansion associated with the far-field approximation. For this
reason, the terms in an expansion like the one above are often referred to by
their multipole equivalents, i.e. the l = 0 term is called a monopole term, the
l = 1 term a dipole term, the l = 2 term a quadropole term, etc.
Chapter 7
Diffraction
Chapter 4 addressed the interaction of plane waves with large, flat bound-
aries. This chapter describes their behavior when they encounter partial
boundaries. In particular, it explores the phenomenon known as diffraction,
which is the apparent bending of a wave when it passes around an edge of or
through an aperture in an otherwise impenetrable boundary (see Figure 7.1).
It begins by outlining the calculation of diffracted wave functions, and then
applies the method to rectangular apertures, circular apertures, and arrays
of apertures.
7.1 Kirchoff Diffraction Theory

In general, diffracted wave functions cannot be found by applying boundary
conditions to the separable solutions of the wave equation at partial bound-
aries. Instead, they are calculated using Green’s theorem, which asserts that
for any region of space with volume V 0 and surface S 0 , and any two well-
behaved functions ψ1 and ψ2 :
I I
0 0 0
(ψ1 ∇ ψ2 − ψ2 ∇ ψ1 ) · dS = (ψ1 ∇02 ψ2 − ψ2 ∇02 ψ1 )dV 0 (7.1)
S0 V0
Letting ψ2 be a solution to the scalar Helmholtz equation (ψω (r 0 )) and ψ1

be a scalar Green’s function (Gk (r, r 0 )):
159
160 CHAPTER 7. DIFFRACTION
Figure 7.1: Diffraction of plane wave through an aperture in two dimensions.
I
[Gk (r, r 0 )∇0 ψω (r 0 ) − ψω (r 0 )∇0 Gk (r, r 0 )] · dS 0
S0 I
= {Gk (r, r 0 )[−k 2 ψω (r 0 )] − ψω (r 0 )[−δ(r − r 0 ) − k 2 Gk (r, r 0 )]}dV 0
0
IV
= ψω (r 0 )δ(r − r 0 )dV 0
V0
= ψω (r)
Switching the order of the terms yields:

I
ψω (r) = [Gk (r, r 0 )∇0 ψω (r 0 ) − ψω (r 0 )∇0 Gk (r, r 0 )] · dS 0
S0
7.1. KIRCHOFF DIFFRACTION THEORY 161
which indicates that the scalar wave function within any region of space
can be found by performing an integral over the surface of that region. To
find the wave function transmitted through/around a partial boundary, the
surface of the region is chosen to conform to the back side of the boundary on
one side and expand toward infinity on the other as illustrated in Figure 7.2.
The free space Green’s function is typically chosen to evaluate the surface
integral because it and its gradient approach zero for large values of r 0 and
therefore nullify the integral on the infinite side leaving:
Figure 7.2: Geometry used for diffraction problems.
0 ik|r−r 0 |
eik|r−r | 0
Z
1 0 0 0e
ψω (r) = ∇ ψω (r ) − ψω (r )∇ · dS 0 (7.2)
4π bdry |r − r 0 | |r − r 0 |
which is called the Kirchoff integral theorem.

Notice that an exact solution to the integral requires knowledge of the

diffracted wave function on the boundary (ψω (r 0 )), which is generally un-
available. The Kirchoff approximation is therefore applied, which asserts
that the diffracted wave function and its gradient are approximately zero on
the back side of an impenetrable partial boundary, and that it is approxi-
0
mately equal to the wave function for the incident plane wave (Ãeiki ·r ) at
an edge or aperture. Making these substitutions yields:
0 0
eik|r−r | 0 iki ·r0 iki ·r0 0 eik|r−r |
Z
1
ψω (r) ≈ ∇ Ãe − Ãe ∇ · dS 0
4π edge/ap |r − r 0 | |r − r 0 |
which is valid as long as the distance between r and the edge/aperture is
large relative to the wavelength. The gradient of the incident wave function
is:
0
0
∇0 Ãeiki ·r = iki Ãeiki ·r
while the gradient of the Green’s function is a straightforward extension of
the gradient for an isotropic spherical wave centered on the origin:
ikr0 0
1 eikr 0

0e
∇ 0 = ik − 0 r̂
r r r0
⇓
0
ik|r−r | 0
ik|r −r|
ik|r0 −r|
0e 0e 1 e
∇ 0
=∇ 0 = ik − 0 r 0d
−r
|r − r | |r − r| |r − r| |r 0 − r|
ik|r−r0 |
1 e
= ik − r 0d −r
|r − r | |r − r 0 |
0
Because |r − r 0 | λ for the Kirchoff approximation to be valid, the second

term in the parentheses is much smaller than the first and can be ignored.
Inserting these results into the Kirchoff integral yields:
0 0
eik|r−r | eik|r−r | 0d
Z
1 iki ·r 0 iki ·r 0
ψω (r) ≈ (iki )Ãe − Ãe (ik) r − r · dS 0
4π |r − r 0 | |r − r 0 |
Substituting ki · dS 0 = −k cos θi dS 0 and r 0d

− r · dS 0 = cos θdS 0 from Figure
7.2:
7.1. KIRCHOFF DIFFRACTION THEORY 163
0
eik|r−r | 0

−iÃ
Z
cos θi + cos θ 0
ψω (r) ≈ eiki ·r dS (7.3)
λ edge/ap 2 |r − r 0 |
Like the radiation integrals from Chapter 6, this integral represents a sum of
outgoing spherical wave functions. Consequently, any diffracted wave can be
viewed as a series of interfering spherical waves spanning the edge/apertures
of the boundary whose amplitudes and phase angles are determined by the
incident plane wave.
Because the rectangular components of a diffracted vector wave separately
satisfy the scalar wave equation, each can be found using the Kirchoff integral
and then recombined. With the Kirchoff approximation applied, the result
is identical to the scalar Kirchoff integral with the amplitude of the incident
plane wave replaced by its vector equivalent:
0
cos θi + cos θ iki ·r0 eik|r−r | 0

−iÃ
Z
ψω (r) ≈ e dS (7.4)
λ edge/ap 2 |r − r 0 |
The Kirchoff integral can be difficult to solve analytically, so additional
approximations are usually required. If r is large relative to r0 over the range
of the integral, the far-field approximation can be applied:

−iÃ
Z
cos θi + cos θ 1 0 0
ψω (r) ≈ eiki ·r eik(r−r̂·r ) dS 0
λ 2 r edge/ap
ikr

−iÃ
Z
cos θi + cos θ e 0
ψω (r) ≈ e−i(k−ki )·r dS 0 (7.5)
λ 2 r edge/ap
where θ is now fixed at the origin, i.e. the angle of r. For vector waves:
−iÃ cos θi + cos θ eikr

Z
0
ψω (r) ≈ e−i(k−ki )·r dS 0 (7.6)
λ 2 r edge/ap
Because the Kirchoff integral represents a continuous series of interfering
spherical waves, the result is also a spherical wave. The remaining surface
integral affects the angular dependence of the amplitude, and will be evalu-
ated for several apertures in the examples that follow assuming a normally
incident plane wave where cos θi = 1 and ki · r 0 = 0. With the coordinate
system oriented such that the apertures lie in the xy-plane as illustrated in
Figure 7.3 for a rectangular aperture, the Kirchoff integral becomes:
Figure 7.3: Geometry used to calculate the diffracted wave for a rectangular
aperture.
−iÃ 1 + cos θ eikr

Z
0
ψω (r) ≈ e−ik·r dS 0
λ 2 r ap
ZZ
0 0
≈ ψsph e−i(kx x +ky y ) dx0 dy 0 = ψsph S(θ, φ) (7.7)
ap
or for vector waves:
eikr

−iÃ
ZZ
1 + cos θ 0 0
ψω (r) ≈ e−i(kx x +ky y ) dx0 dy 0 = ψsph S(θ, φ) (7.8)
λ 2 r ap
where kx = kr̂x = k sin θ cos φ and ky = kr̂y = k sin θ sin φ in terms of

spherical coordinates. Notice that the surface integral S(θ, φ) is essentially
7.2. RECTANGULAR APERTURE 165
the two-dimensional Fourier transform of the aperture, but can also be in-
terpreted as a phasor sum for the interfering spherical waves spanning the
aperture. It is also identical to the integral required for a baffled piston
(Equation 6.34), so the results in the examples that follow apply equally to
the radiation produced by baffled pistons with the same shapes.
If, as usual, the intensity of the diffracted spherical wave is proportional
to its squared amplitude, then:
I = S 2 (θ, φ)Isph (7.9)

where Isph is the intensity for ψsph or ψsph . S 2 (θ, φ) is the interference pattern
produced by the spherical waves spanning the aperture, and is referred to as
the diffraction pattern.
7.2 Rectangular Aperture

The surface integral for the rectangular aperture in Figure 7.3 is given by:
Z b/2 Z a/2 Z b/2 Z a/2

−i(kx x0 +ky y 0 ) 0 0 −iky y 0 0 0
S(θ, φ) = e dx dy = e dy e−ikx x dx0
−b/2 −a/2 −b/2 −a/2
Evaluating the second integral yields:
Z a/2 Z a/2 Z a/2

−ikx x0 0 0 0
e dx = cos kx x dx − i sin kx x0 dx0
−a/2 −a/2 −a/2
0 a/2
sin kx x cos kx x0 a/2
= +i
kx kx

−a/2 −a/2
sin(kx a/2)
=a
(kx a/2)
The first integral is identical so the full result becomes:
sin(kx a/2) sin(ky b/2)

S(θ, φ) = ab (7.10)
(kx a/2) (ky b/2)
An example of the corresponding diffraction pattern:
sin2 (kx a/2) sin2 (ky b/2)

S 2 (θ, φ) = (ab)2 (7.11)
(kx a/2)2 (ky b/2)2
is illustrated in Figure 7.4.
0.8
0.6
20
0.4
10
0.2
0
0
−20
−10 −10
0
10 −20 ky (1/m)
20
kx (1/m)
Figure 7.4: Normalized diffraction pattern for a rectangular aperture with

a = 1.5 and b = 1 meters.
Consider a cross section of the surface integral/phasor sum along the

xz-plane where kx = k sin θ and ky = 0 such that:
ka
b/2 a/2

sin sin θ
Z Z
−ik sin θx0
S(θ, 0) = e dx0 dy 0 = ab ka
2
−b/2 −a/2 2
sin θ
and:
7.2. RECTANGULAR APERTURE 167
2 ka

2 2 sin 2
sin θ
S (θ, 0) = (ab) 2 (7.12)
ka
2
sin θ
which is plotted in Figure 7.5. The complex exponential in the sum indicates
that the phasor angle for a spherical wave located at x0 on the aperture
surface is k sin θx0 . The total phase difference between the spherical waves
on one side of the aperture and those on the other is therefore given by:
α = k(a/2) sin θ − k(−a/2) sin θ = ka sin θ
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−10 −8 −6 −4 −2 0 2 4 6 8 10
(ka/2) sin θ
Figure 7.5: Normalized diffraction pattern for a rectangular aperture in the

xz-plane.
Figure 7.6 illustrates the continuous phasor sum for several different values
of α. The waves are in phase and the interference is constructive when all
of the phasors align, which only occurs when α = 0 at θ = 0 and corre-

sponds to the primary maximum in Figure 7.5. The value of the surface
integral/phasor sum at this angle is equal to the area of the aperture (ab),
which is unsurprising given that the number of interfering spherical waves
spanning its surface is determined by its size. The interference is destructive
whenever the tip of the last phasor meets the tail of the first such that the
sum is zero, and occurs at angles for which α is an integer multiple of 2π:
α = ka sin θ = 2πm
ka sin θ = 2πm
λ
sin θ = m (7.13)
a
Figure 7.6: Phasor sum for a rectangular aperture at several angles.

7.3. CIRCULAR APERTURE 169
These angles correspond to the nodes in Figure 7.5. Almost half way between
the nodes are secondary maxima, the heights of which become progressively
smaller as the phasor sum wraps upon itself.
The angular widths of the primary and secondary maxima are deter-
mined by the locations of the nodes on either side and can be calculated by
differentiating Equation 7.13:
dm a
= cos θ
dθ λ
λ 1
∆θ = ∆m
a cos θ
Plugging in θ = 0 and ∆m = 1 − (−1) = 2 for the primary maximum yields:
2λ
∆θprimary maximum = (7.14)
a
which indicates that its width, as well as that of the entire diffraction pattern,
is proportional to the wavelength and inversely proportional to the size of
the aperture.
7.3 Circular Aperture

Many acoustic and optical instruments make use of circular apertures/pistons
like the one illustrated in Figure 7.7. Switching from rectangular to cylin-
drical coordinates due to the circular symmetry of the aperture, the surface
integral becomes:
ZZ Z 2π Z a
−i(kx x0 +ky y 0 ) 0 0 0 cos φ0 +ky r0 sin φ0 ) 0
S(θ, φ) = e dx dy = e−i(kx r r dr0 dφ0
ap 0 0
Because the problem possesses azimuthal symmetry, the result must be in-
dependent of the azimuthal angle φ. Consequently, the integral can be eval-
uated for a single value of φ and applied to all other values. Choosing φ = 0,
which corresponds to the xz-plane, kx = k sin θ and ky = 0 such that:
Z 2π Z a
0 0
S(θ) = e−ik sin θr cos φ r0 dr0 dφ0
0 0
Figure 7.7: Geometry used to calculate the diffracted wave for a circular
aperture.
where the solution can be obtained from a table of integrals:
2J1 (ka sin θ)

S(θ) = πa2 (7.15)
(ka sin θ)
The diffraction pattern for the circular aperture:
4J12 (ka sin θ)

S 2 (θ) = (πa2 )2 (7.16)
(ka sin θ)2
is plotted in Figure 7.8. Notice the similarity with the pattern for a rectangu-
lar aperture. Both have a primary maximum equal to the squared aperture
area located along the central axis. Both also possess diminishing secondary
maxima located between a series of nodes. The locations of the nodes for the
7.3. CIRCULAR APERTURE 171
circular aperture are specified by the roots of the Bessel function:
ka sin θ = U1l
λ
sin θ = U1l (7.17)
2πa
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−10 −8 −6 −4 −2 0 2 4 6 8 10
ka sin θ
Figure 7.8: Normalized diffraction pattern for a circular aperture.

and can be used to determine the angular widths of the maxima. For the
primary maximum in particular:
∆θprimary maximum = 2θfirst node ≈ 2 sin θfirst node

λ λ λ λ
≈ U11 ≈ (1.22π) = 1.22 = 2.44 (7.18)
πa πa a D
where D is the diameter of the aperture. The result is slightly larger than
for a rectangular aperture of comparable size, but the dependencies upon
wavelength and aperture size are identical.
7.4 Multiple Apertures

Double apertures are identical apertures placed sided by side as illustrated
in Figure 7.9 for two rectangular apertures. Because the diffracted wave for
an individual aperture is a spherical wave in the far-field, a double aperture
simply produces two interfering spherical waves. From Section 3.2, the wave
function for two interfering spherical waves oriented along the z-axis is:
ψ = 2 cos[k(d/2) cos θ]ψsph = 2 cos(kz d/2)ψsph

If the double aperture is aligned with the x-axis as in Figure 7.9, kz simply
becomes kx . Making this substitution and inserting the expression for the
spherical wave produced by a single aperture:
ψ = 2 cos(kx d/2)S(θ, φ)single ψsph

The diffraction pattern for the double aperture is therefore:
S 2 (θ, φ)double = 4 cos2 (kx d/2)S 2 (θ, φ)single (7.19)

which is simply the product of the interference pattern for two spherical waves
and the diffraction pattern for a single aperture. For the double rectangular
aperture in particular:
2 2
2 2 2 sin (kx a/2) sin (ky b/2)
S (θ, φ)double = 4 cos (kx d/2)(ab) (7.20)
(kx a/2)2 (ky b/2)2
Consider a cross section of the pattern along the xz-plane where kx =
k sin θ and ky = 0:
2 ka

2 2 2 sin 2
sin θ
S (θ, 0)double = 4 cos (kd/2 sin θ)(ab) 2 (7.21)
ka
2
sin θ
which is plotted in Figure 7.10 for d/a = 4. Because the aperture spacing is
larger than the aperture size, the single-aperture diffraction pattern is wider
than the interference pattern and forms an envelope around it.
7.4. MULTIPLE APERTURES 173
Figure 7.9: Geometry used to calculate the diffracted wave for a double
rectangular aperture.
Extending the results for a double aperture to a linear array of apertures,

the diffraction pattern for the array must be the product of the diffraction
pattern for a single aperture and the interference pattern for a linear array of
spherical waves. The interference pattern from Section 3.3 for a linear array
oriented along the z-axis is given by:
2 2
2 sin [N (k/2)d cos θ] 2 sin (N kz d/2)
N =N
N sin [(k/2)d cos θ] N sin (kz d/2)
so the diffraction pattern for an array of apertures aligned with the x-axis is:
2
2 2 sin (N kx d/2)
S (θ, φ)linarray = N S 2 (θ, φ)single (7.22)
N sin (kx d/2)
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−10 −8 −6 −4 −2 0 2 4 6 8 10
(ka/2) sin θ
Figure 7.10: Normalized diffraction pattern for a double rectangular aperture

in the xz-plane where d/a = 4. The dotted line is the normalized diffraction
pattern for a single aperture.
For an array of rectangular apertures in particular, the diffraction pattern in

the xz-plane is:
2
sin2 ka

sin [N (k/2)d sin θ] sin θ
S 2 (θ, 0)linarray = N 2 (ab)2 2
2 (7.23)
N sin [(k/2)d sin θ] ka
sin θ
2
which is plotted in Figure 7.11 for d/a = 4 and N = 4. Again, the single-
aperture diffraction pattern forms an envelope around the interference pat-
tern.
A two-dimensional array is essentially a linear array of spherical waves
along the y-axis produced by a series of linear arrays aligned with the x-
7.4. MULTIPLE APERTURES 175
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−6 −4 −2 0 2 4 6
(ka/2) sin θ
Figure 7.11: Normalized diffraction pattern for a linear array of rectangular

apertures in the xz-plane where d/a = 4 and N = 4. The dotted line is the
normalized diffraction pattern for a single aperture.
axis. Consequently, the diffraction pattern is the product of the interference

pattern for a linear array along the y-axis with the diffraction pattern for a
linear array aligned with the x-axis:
2 2
2 sin (Ny ky dy /2) sin (Nx kx dx /2)
S (θ, φ)array = Ny2 Nx2 S 2 (θ, φ)single
Ny sin (ky dy /2) Nx sin (kx dx /2)
(7.24)
where Nx and Ny are the number of apertures in the x and y-directions
respectively, and dx and dy are the spacings between them.
Chapter 8
Scattering
Earlier chapters have explored the interaction of plane waves with bound-
aries that are large and flat relative to the wavelength. This final chapter
addresses scattering, which is the reflection, transmission and diffraction of
incident waves by the surfaces of small objects. It begins by introducing
the terminology used to describe scattering, and then explores the meth-
ods used to calculate scattered wave functions, including an exact method
that involves applying boundary conditions at the scatterer surface, and two
approximate methods. One treats the scatterer as a localized variation in
medium properties, while the other applies the Kirchoff integral to the scat-
terer surface.
8.1 Scattering Terminology

Figure 8.1 shows a conceptual diagram of an incident plane wave scattered
by a small object. Notice that the scattered wave is a spherical wave at large
distances, just as radiated and diffracted waves are spherical in the far-field.
Consequently, the scattered wave function takes the following form in the
far-field:
ei(kr−ωt) ei(kr−ωt)
ψs = Ãs (θ, φ, ω) or ψs = Ãs (θ, φ, ω) (8.1)
r r
where there is an implicit dependence of the amplitudes upon the orienta-
tion of the scatterer relative to the incident plane wave, and of the vector
amplitude upon the polarization of the incident plane wave.
177
178 CHAPTER 8. SCATTERING
Figure 8.1: Scattering of a plane wave by a small object.
The amplitude of the scattered wave is dependent upon the properties

and shape of the scatterer, but is also proportional to the amplitude of the
incident plane wave Ai , e.g. doubling the incident amplitude doubles the
scattered amplitude. Dividing it out removes this dependence and leaves a
quantity that is solely dependent upon the object and can therefore be used
to characterize the scattering it produces:
Ãs (θ, φ, ω) Ãs (θ, φ, ω)

L̃(θ, φ, ω) = or L̃(θ, φ, ω) = (8.2)
Ai |Ai |
This normalized scattering amplitude is referred to as the scattering length
simply because it possesses units of length.
Once the scattering length of an object is known, it can be multiplied
by the amplitude of any incident plane wave to determine the amplitude of
8.1. SCATTERING TERMINOLOGY 179
the resulting scattered wave. In most situations, however, it is the scattered

power that is of interest rather than the scattered amplitude. Naively, the
total scattered power would be equal to the product of the incoming plane
wave’s intensity and the amount of its area that contacts the scatterer. From
Figure 8.1, it is clear that this quantity is the physical cross section of the
scatterer σp so:
Ps = Ii σp (8.3)
Of course this simple model fails to account for diffraction of the incident
plane wave around the edges of the scatterer and transmission of the wave
through its surface. Consequently, the scattered power must determined from
the scattering length.
The first step in the process is to relate the scattering length to the
scattered intensity. Because the scattered wave is spherical, its intensity is
given by:
|Ãs (θ, φ, ω)|2 A2i |L̃(θ, φ, ω)|2 |Ai |2 |L̃(θ, φ, ω)|2

Is = C = C or Is = C
r2 r2 r2
where C is a constant of proportionality that depends on wave type. The

plane wave intensity and amplitude are related by the same constant (Ii =
CA2i ), so the equation can be rewritten:
Ii |L̃(θ, φ, ω)|2 Ii |L̃(θ, φ, ω)|2

Is = or Is =
r2 r2
The scattered intensity can then be used to calculate the scattered power.
Because the total power transmitted by any three-dimensional wave is found
by integrating its intensity over a surface perpendicular to its propagation
direction, the total power for a scattered spherical wave is determined by
integrating its intensity over a spherical shell. Using a shell with radius r
and working in spherical coordinates:
Z 2π Z π
Ps = Is r2 sin θdθdφ
0 0
Z 2π Z π
= Ii |L̃(θ, φ, ω)|2 sin θdθdφ
I0 0
= Ii |L̃(θ, φ, ω)|2 dΩ
where dΩ = sin θdθdφ is an increment of two-dimensional angle referred to

as solid angle.
Notice that the integral of the squared scattering length is related to the
incident intensity and scattered power in the same way as the physical cross
section of the scatterer. It is therefore referred to as the scattering cross
section of the scatterer and identified by the same variable:
I I
Ps 2
σ(ω) = = |L̃(θ, φ, ω)| dΩ or σ(ω) = |L̃(θ, φ, ω)|2 dΩ (8.4)
Ii
The derivative:
dσ dσ
(θ, φ, ω) = |L̃(θ, φ, ω)|2 or (θ, φ, ω) = |L̃(θ, φ, ω)|2 (8.5)
dΩ dΩ
is called the differential scattering cross section, and can be used to determine
the power scattered per unit solid angle in a particular direction specified by
θ and φ:
dP d dσ
(θ, φ, ω) = (Is σ(ω)) = Is (θ, φ, ω) (8.6)
dΩ dΩ dΩ
The value of the differential scattering cross section opposite the direction
of the incident plane wave is called the backscatter, and is of most practical
interest because the same device is often used to both generate the incident
wave and receive the scattered wave.
8.2 Boundary Condition Method

The scattering length for an object is determined theoretically by calculating
the amplitude of the scattered spherical wave for a given incident plane wave,
and dividing the result by the amplitude of the incident plane wave. One
way of finding the scattered amplitude is to apply boundary conditions to the
incident, scattered and internal wave functions at the surface of the object.
The method is nearly identical to the process used in Chapter 4 to find
reflection and transmission coefficients for planar surfaces. Although exact,
like all methods that involve the application of boundary conditions, it is
limited to simple symmetrical shapes.
As an example, consider acoustic scattering from a fluid sphere as illus-
trated in Figure 8.2 where the wave speed and fluid density for the sphere
are vs and ρs respectively. Because the scatterer is spherical, boundary con-
ditions are applied to the wave functions written in terms of the separable
wave equation solutions for spherical coordinates. The wave function for the
incident plane wave becomes:
Figure 8.2: Geometry used to calculate the scattering length for a fluid
sphere, where the plane wave in incident along the +z-axis.
X
pi (r, θ, φ, t) = Ai ei(−kz−ωt) = Ai (−i)l (2l + 1)jl (kr)Pl0 (cos θ)e−iωt (8.7)
l
The scattered pressure propagates outward and has no azimuthal dependence

due to the symmetry of the problem, so its wave function is given by:
X (1)
ps (r, θ, φ, t) = Ãl hl (kr)Pl0 (cos θ)e−iωt
l
The inside of the sphere is a cavity, so the internal wave function, which also
has no azimuthal dependence, is given by:
X
pt (r, θ, φ, t) = B̃l jl (ks r)Pl0 (cos θ)e−iωt
l
The boundary conditions that are applied to the wave functions are iden-
tical to those used in Chapter 4 for a planar surface. The first is that the
pressures on the inside and outside of the sphere must be equal at the surface:
pi (a, θ, φ, t) + ps (a, θ, φ, t) = pt (a, θ, φ, t)

(1)
Ai (−i)l (2l + 1)jl (ka) + Ãl hl (ka) = B̃l jl (ks a)
Ai (−i)l (2l + 1)jl (ka) + Ãl [jl (ka) + inl (ka)] = B̃l jl (ks a)
The second is that the components of the particle velocity perpendicular to

the surface on the inside and outside of the sphere must be equal:
ur,i (a, θ, φ, t) + ur,s (a, θ, φ, t) = ur,t (a, θ, φ, t)
The pressures can be related to the radial components of particle velocity

using Euler’s equation written in terms of spherical components (Equation
5.25):
∂ur ∂p
ρ =−
∂t ∂r
∂p
−iωρur = −
∂r
i ∂p
ur = −
kvρ ∂r
i ∂p
ur = −
vρ ∂(kr)
Making this substitution into the boundary condition produces:
Ai Ãl (1)0 B̃l 0

(−i)l (2l + 1)jl0 (ka) + hl (ka) = j (ks a)
vρ vρ vs ρs l
Ai Ãl B̃l 0
(−i)l (2l + 1)jl0 (ka) + [jl0 (ka) + in0l (ka)] = j (ks a)
vρ vρ vs ρs l
Using the two boundary condition equations to solve for the scattered
amplitude:
(−i)l (2l + 1) vρjl0 (ks a)nl (ka) − vs ρs jl (ks a)n0l (ka)

Ãl = −Ai Cl = (8.8)
1 + iCl vρjl0 (ks a)jl (ka) − vs ρs jl (ks a)jl0 (ka)
Substituting the far-field approximation for the spherical Hankel functions:
(1) eikr
hl (kr) ≈ (−i)(l+1) for kr 1
kr
into the wave function for the scattered pressure yields a spherical wave:
ei(kr−ωt) X i (−1)l (2l + 1) 0

ps (r, θ, φ, t) = Ai Pl (cos θ) (8.9)
r l
k 1 + iCl
Dividing its amplitude by the amplitude of the incident plane wave leaves
the scattering length for the sphere:
X i (−1)l (2l + 1)
L̃(θ, φ, ω) = Pl0 (cos θ) (8.10)
l
k 1 + iCl
For small spheres relative to the wavelength (ka 1), the spherical
Neumann function and its derivative in Cl are large and become larger for
higher values of l. Consequently, only the lower-order terms in the sum
contribute significantly to the scattering length. Keeping
p only the monopole
(l = 0) and dipole (l = 1) terms and substituting v = 1/κρ:
1 κs − κ 1 2 3 3ρs − 3ρ
L̃(θ, φ, ω) ≈ Lmonopole + Ldipole = k 2 a3 − k a cos θ (8.11)
3 κ 3 2ρs + ρ
In Chapter 6, pulsating and vibrating spheres were associated with monopole
and dipole radiation respectively. In this case, unequal compressibilities give
rise to monopole scattering and unequal densities to dipole scattering, sug-
gesting that the incident plane wave causes the scatterer to pulsate when
its compressibility is different from that of the surrounding medium and to
vibrate when its density is different from that of the surrounding medium.
Notice also that the scattering length is proportional to the squared fre-
quency, implying that the cross sections are proportional to the fourth power
of the frequency. This small-scatterer behavior is called Rayleigh scattering
and is not specific to spheres or acoustic scattering.
For the extreme case of a rigid sphere, the boundary conditions applied
at its surface change. The component of particle velocity perpendicular to
its surface must be zero:
ur,i (a, θ, φ, t) + ur,s (a, θ, φ, t) = 0

Ai Ãl (1)0
(−i)l (2l + 1)jl0 (ka) + hl (ka) = 0
vρ vρ
Ai Ãl
(−i)l (2l + 1)jl0 (ka) + [jl0 (ka) + in0l (ka)] = 0
vρ vρ
l 0
(−i) (2l + 1)jl (ka) (−i)l (2l + 1)
Ãl = −Ai 0 = −A i
jl (ka) + in0l (ka) 1 + in0l (ka)/jl0 (ka)
which implies that:
n0l (ka)
Cl = Rigid Sphere (8.12)
jl0 (ka)
For the opposite case of a pressure release sphere, the pressure at its surface
must be zero:
pi (a, θ, φ, t) + ps (a, θ, φ, t) = 0
(1)
Ai (−i)l (2l + 1)jl (ka) + Ãl hl (ka) = 0
Ai (−i)l (2l + 1)jl (ka) + Ãl [jl (ka) + inl (ka)] = 0
(−i)l (2l + 1)jl (ka) (−i)l (2l + 1)
Ãl = −Ai = −Ai
jl (ka) + inl (ka) 1 + inl (ka)/jl (ka)
which means that:
nl (ka)
Cl = Pressure Release Sphere (8.13)
jl (ka)
Figure 8.3 shows the scattering length for rigid spheres of different sizes.
Notice that the monopole and dipole terms are dominant for the smaller
spheres, but that the higher-order terms become more significant as size
increases. For comparison, consider the scattering length for a sphere taking
only reflection of the incident plane wave into account. In that case, the
scattering cross section is equal to the physical cross section of the sphere:
I
2 dσ
πa = σr = sin θdθdφ
dΩ r
Because spheres reflect waves isotropically, the differential scattering cross
section does not depend upon angle and can be removed from the integral
such that:
I
2 dσ
πa = sin θdθdφ
dΩ r
dσ
πa2 = 4π
dΩ r
dσ a2
=
dΩ r 4
a
L̃r =
2
For the larger spheres, the scattering length is nearly equal to this value in
the backward direction between 0 and 90 degrees. The lobes in the forward
direction destructively interfere with the incident plane wave to produce a
90 1 90 2
120 60 120 60
0.5 1
150 30 150 30
180 0 180 0
210 330 210 330

240 300 240 300
270 270
ka = 1 ka = 2
90 4 90 4
120 60 120 60
2 2
150 30 150 30
180 0 180 0
210 330 210 330

240 300 240 300
270 270
ka = 5 ka = 10
Figure 8.3: Normalized Scattering Length (|L̃|/(a/2)) for rigid spheres.
shadow with oscillating intensity caused by diffraction of the plane wave

around the edges of the sphere. In this particular case, there is no effect due
to transmission of the incident plane wave into/out of the sphere because
rigid boundaries are impenetrable.
Figure 8.4 plots the backscatter for a rigid sphere as a function of fre-
quency. The results exhibit Rayleigh scattering at the lower frequencies and
approach pure reflection at higher frequencies. The oscillations at the higher
frequencies are caused by interference between the reflected wave and a por-
tion of the incident plane wave that is diffracted around the entire sphere.
8.3. INHOMOGENEOUS MEDIUM METHOD 187
1.4
1.2
1
Backscatter/(a2 /4)
0.8
0.6
0.4
0.2
0
0 2 4 6 8 10
ka
Figure 8.4: Normalized backscatter as a function of frequency for a rigid

sphere.
8.3 Inhomogeneous Medium Method

For more complicated scatterer shapes, approximate methods are required to
find the scattered amplitude. The first treats scatterers as localized variations
in medium properties. For acoustics, the wave equation for an inhomogenous
medium is given by:
1 ∂ 2p
∇2 p − = −[k 2 γκ (r)p(r, t) − ∇ · γρ (r)∇p(r, t)]
v 2 ∂t2
where γρ and γκ quantify the variations in medium density and compressibil-
ity introduced by the scatterers:
ρs (r) − ρ
γρ (r) = (8.14)
ρs (r)
κs (r) − κ
γκ (r) = (8.15)
κ
Assuming harmonic time dependence for the pressure p(r, t) = pω (r)e−iωt
produces the Helmholtz equation for inhomogeneous media:
∇2 pω + k 2 pω = −[k 2 γκ (r)pω (r) − ∇ · γρ (r)∇pω (r)]

which is identical to the Helmholtz equation for radiators with the source
term:
f (r) = k 2 γκ (r)pω (r) − ∇ · γρ (r)∇pω (r) (8.16)

The scattered wave can therefore be determined by inserting this term into
the radiation integral:
0
eik|r−r |
Z
pω,s (r) = [k 2 γκ (r 0 )pω (r 0 ) − ∇0 · γρ (r 0 )∇0 pω (r 0 )] dr 0
4π|r − r 0 |
Recasting the integral in a more usable form:
0
eik|r−r |
Z
0 0 0 0 0 0
pω,s (r) = 2
[k γκ (r )pω (r ) + γρ (r )∇ pω (r ) · ∇ ] dr 0
4π|r − r 0 |
and applying the far-field approximation to the free space Green’s function
yields:
eikr −ik·r0 0
Z
pω,s (r) = [k 2 γκ (r 0 )pω (r 0 ) + γρ (r 0 )∇0 pω (r 0 ) · ∇0 ] e dr
4πr
eikr
Z
0 0
= [k 2 γκ (r 0 )pω (r 0 )e−ik·r − γρ (r 0 )∇0 pω (r 0 ) · ike−ik·r ]dr 0 (8.17)
4πr
Notice that evaluating the radiation integral requires knowledge of the
total pressure pω (r), which is the sum of the incident and scattered pressures.
Because the scattered pressure is initially unknown, the total pressure is

often approximated as the incident pressure alone. This approximation is
called the Born approximation, and is reasonable when the scattered wave is
weak relative to the incident wave. Its accuracy can be improved by making
recursive approximations, i.e. inserting the result for the scattered wave back
into the radiation integral along with the incident wave and reevaluating.
For an incident plane wave, the approximation yields:
eikr
Z
0 0
pω,s (r) ≈ [k 2 γκ (r 0 )(Ai eiki ·r )e−ik·r
4πr
0 0
− γρ (r 0 )∇0 (Ai eiki ·r ) · ike−ik·r ]dr 0
eikr
Z
0 0
≈ Ai [k 2 γκ (r 0 )e−i(k−ki )·r + γρ (r 0 )(ki · k)e−i(k−ki )·r ]dr 0
4πr
eikr
Z
0
≈ Ai [k 2 γκ (r 0 ) + γρ (r 0 )(ki · k)]e−i(k−ki )·r dr 0
4πr
If the incident plane wave propagates in the -z direction such that ki · k =
k 2 (r̂ · −ẑ) = −k 2 cos θ:
k 2 eikr
Z
0
pω,s (r) ≈ Ai [γκ (r 0 ) − γρ (r 0 ) cos θ]e−iK·r dr 0
4πr
and:
k2
Z
0
L̃(θ, φ, ω) ≈ [γκ (r 0 ) − γρ (r 0 ) cos θ]e−iK·r dr 0 (8.18)
4π
where K = k − ki = k(r̂ + ẑ). For a single scatterer with a uniform interior,
the scattering length becomes:
k2
Z
0
L̃(θ, φ, ω) ≈ [γκ − γρ cos θ] e−iK·r dr 0
4π scat
k2
≈ [γκ − γρ cos θ]SF T (8.19)
4π
where the integral is over the volume of the scatterer and:
Z ZZZ
−iK·r 0 0 0 0 0
SF T = e dr = e−i(Kx x +Ky y +Kz z ) dx0 dy 0 dz 0 (8.20)
scat scat
is the three-dimensional Fourier transform of the scatterer.

For electromagnetic waves, the radiation integral for the electric field with
the far-field approximation applied is given by:
eikr
Z
0 0
Eω,s (r) = [k × γ (r 0 )Eω (r 0 ) × ke−ik·r − ωk × γµ (r 0 )Bω (r 0 )e−ik·r ]dr 0
4πr
(8.21)
where γ and γµ quantify the variations in the medium permittivity and
permeability introduced by the scatterers:
(r) −
γ (r) = (8.22)
s
µs (r) − µ
γµ (r) = (8.23)
µs (r)
Because the wave is spherical, the corresponding magnetic field is perpen-
dicular with magnitude Bω,s = Eω,s /v. Applying the Born approxima-
tion for an incident plane wave where the corresponding magnetic field is
Bω = k̂i /v × Eω :
eikr
Z
0 0
Eω,s (r) ≈ [k × γ (r 0 )(Ai eiki ·r ) × ke−ik·r
4πr
0 0
− ωk × γµ (r 0 )(k̂i /v × Ai eiki ·r )e−ik·r ]dr 0
eikr
Z
0
≈ [k × γ (r 0 )Ai × k − k × γµ (r 0 )(ki × Ai )]e−i(k−ki )·r dr 0
4πr
eikr
Z
0
≈ [k × γ (r 0 )Ai × k + ki × γµ (r 0 )Ai × k]e−i(k−ki )·r dr 0
4πr
If the incident plane wave propagates in the -z direction:
k 2 eikr
Z
0
Eω,s (r) ≈ |Ai | [r̂ × γ (r 0 )Âi × r̂ − ẑ × γµ (r 0 )Âi × r̂]e−iK·r dr 0
4πr
and:
k2
Z
0
L̃(θ, φ, ω) ≈ [r̂ × γ (r 0 )Âi × r̂ − ẑ × γµ (r 0 )Âi × r̂]e−iK·r dr 0 (8.24)
4π
For a single, uniform scatterer:
k2
L̃(θ, φ, ω) ≈ [r̂ × γ Âi × r̂ − ẑ × γµ Âi × r̂]SF T (8.25)
4π
If the scatterer is small relative to the wavelength and the origin for the
SF T is placed within the scatterer, K · r 0 ≈ 0 such that the SF T yields
the volume of the scatterer. For an acoustic scatterer, the scattering length
becomes:
k 2 Vscat
L(θ, φ, ω) ≈ [γκ − γρ cos θ] (8.26)
4π
which exhibits Rayleigh scattering as expected. For a spherical scatterer, in
particular:
k 2 a3

κs − κ ρs − ρ
L(θ, φ, ω) ≈ − cos θ (8.27)
3 κ ρs
which is almost identical to the exact result obtained by applying bound-
ary conditions (Equation 8.11). The discrepancy in the density terms is a
consequence of the Born approximation, and becomes significant when the
difference between the scatterer and medium densities is large. For an elec-
tromagnetic scatterer:
k 2 Vscat
L(θ, φ, ω) ≈ [r̂ × γ Âi × r̂ − ẑ × γµ Âi × r̂] (8.28)
4π
where the bracketed terms determine the angular dependence of the scatter-
ing. The first is dependent upon the sine of the angle between r̂ and the
incident electric field along Âi , and the second upon the sine of the angle
between r̂ and the incident magnetic field along −ẑ × Âi .
For larger scatterers, the SF T changes the frequency/angular dependence
of the scattering and can often be evaluated for simple shapes. For a sphere
with radius a centered on the origin:
Z a Z π Z 2π
0
SF T = e−iK·r r02 sin θ0 dφ0 dθ0 dr0
0 0 0
Because the problem possesses spherical symmetry, K can be aligned with

the z 0 -axis without loss of generality such that:
Z a Z π Z 2π
0 cos θ0 02
SF T = e−iKr r sin θ0 dφ0 dθ0 dr0
0 0 0
Z a Z π
0 cos θ0 02
= 2π e−iKr r sin θ0 dθ0 dr0
Z0 a 0
1 −iKr0 cos θ0 π 0 0
= 2π e |0 r dr
iK
Z0 a
1 iKr0 −iKr0

= 2π e −e r0 dr0
iK
Z0 a
4π
= r0 sin Kr0 dr0
K 0
The final integral can be evaluated using integration by parts to yield:
4π 1
SF T = (sin Kr0 − Kr0 cos Kr0 )|a0
K K 2
4πa3 sin Ka cos Ka

= −
Ka (Ka)2 Ka
j1 (Ka)
= 4πa3
Ka
A simple substitution can be made for K:
p
K= (kr̂ + k ẑ) · (kr̂ + k ẑ)
√
= k 2 + k 2 + 2k 2 cos θ
r
1 + cos θ
= 2k
2
= 2k cos(θ/2)
such that:
j1 (2ka cos(θ/2))
SF T = 4πa3 (8.29)
2ka cos(θ/2)
Figure 8.5 shows the corresponding acoustic scattering length, and compares
it with the exact result obtained by applying boundary conditions. Figure 8.6
8.4. KIRCHOFF METHOD 193
does the same for the backscatter. The agreement is good except in the for-
ward direction for larger values of ka. For scatterers with more complicated
shapes, where analytic calculation of the SF T is impossible, the multipole
expansion can be applied to the SF T as it was to the radiation integral in
Subsection 6.2.2 and the resulting series evaluated.
90 0.1 90 0.2
120 60 120 60
150 0.05 30 150 0.1 30
180 0 180 0
210 330 210 330
240 300 240 300

270 270
ka = 1 ka = 5
Figure 8.5: Normalized acoustic scattering length (|L̃|/(a/2)) for a sphere ob-
tained using the inhomogeneous medium method (dashed) and the boundary
condition method (solid). The density of the scatterer is 10 percent greater
and the compressibility 10 percent less than for the background.
8.4 Kirchoff Method

An alternative method for approximating the scattered amplitude employs
the Kirchoff integrals. Recall from Section 7.1 that for a scalar wave:
0.012
0.01
Backscatter/(a2 /4)
0.008
0.006
0.004
0.002
0
0 2 4 6 8 10
ka
Figure 8.6: Normalized acoustic backscatter for a sphere obtained using

the inhomogeneous medium method (dashed) and the boundary condition
method (solid). The density of the scatterer is 10 percent greater and the
compressibility 10 percent less than for the background.
I
ψω (r) = [Gk (r, r 0 )∇0 ψω (r 0 ) − ψω (r 0 )∇0 Gk (r, r 0 )] · dS 0
S0
where ψω (r) is a desired wave function and the surface integral is performed
over the boundary of the spatial region for which that wave function is de-
sired. For a scattered wave function, the inner boundary is typically chosen
to conform to the surface of the scatterer and the outer boundary is expanded
toward infinity. Selecting the free space Green’s function nullifies the integral
on the outer boundary leaving:
0 ik|r−r 0 |
eik|r−r | 0
Z
1 0 0 0e
ψω,s (r) = 0
∇ ψω,s (r ) − ψω,s (r )∇ 0
· dS 0 (8.30)
4π scat |r − r | |r − r |
which is the scalar Kirchoff integral, but evaluated over the surface of the
scatterer rather than a diffracting boundary. Applying the far-field approxi-
mation to the free space Green’s function yields:
eikr
Z
0
ψω,s (r) = e−ik·r [∇0 ψω,s (r 0 ) + iψω,s (r 0 )k] · dS 0 (8.31)
4πr scat
The Kirchoff integral is identical for vector waves, so the same expression
can be used to calculate scattered vector wave functions with ψω,s replaced
by ψω,s .
Notice that an exact solution for the scattered wave function requires
prior knowledge of its values on the surface of the scatterer. Because these
are generally unknown, they are typically approximated using the Kirchoff
scattering approximation, which asserts that the scattered wave function at
each point on the illuminated/front surface of the scatterer is approximately
equal to the reflected wave function that would be produced by an infinite
planar boundary parallel to the surface at that location, and that at each
point on the dark/back surface of the scatterer, it is approximately equal to
the negative of the incident plane wave function at that location such that the
two destructively interfere to produce a shadow. These assumptions produce
reasonable results as long as the surface of the scatterer is relatively flat, i.e.
its curvature is large in comparison to the wavelength. The Kirchoff approxi-
mation is therefore valid for short wavelengths/high frequencies. Because the
approximation ignores the effects of waves inside the scatterer, transmission
through the scatterer boundary must also be relatively weak.
For a scalar plane wave incident upon an arbitrary scatterer as shown in
Figure 8.7, the scattered wave function and its gradient on the illuminated
surface of the scatterer are approximated by:
0
ψω,s (r 0 ) ≈ R(θi )Ai eikr ·r
0
∇0 ψω,s (r 0 ) ≈ ikr R(θi )Ai eikr ·r
Figure 8.7: Geometry for the Kirchoff scattering approximation.
where R(θi ) is the reflection coefficient for the corresponding planar surface.
Recall from Section 4.2 that imposing boundary conditions requires kr · r 0 =
ki ·r 0 at a planar surface. Making this substitution to simplify the calculation:
0
ψω,s (r 0 ) ≈ R(θi )Ai eiki ·r
0
∇0 ψω,s (r 0 ) ≈ ikr R(θi )Ai eiki ·r
Inserting these results into the Kirchoff integral yields the contribution to
the total scattered wave function by the illuminated side of the scatterer:
eikr
Z h i
0 0 0
ψω,s,ill (r) ≈ e−ik·r ikr R(θi )Ai eiki ·r + ikR(θi )Ai eiki ·r · dS 0
4πr ill
iAi eikr
Z
0
≈ R(θi )[kr + k]e−i(k−ki )·r · dS 0
4πr ill
ikAi eikr
Z
0
≈ R(θi )[cos θi + cos θs ]e−i(k−ki )·r dS 0 (8.32)
4πr ill
The corresponding contribution to the scattering length is given by:
Z
i 0
L̃ill (θ, φ, ω) ≈ R(θi )[cos θi + cos θs ]e−i(k−ki )·r dS 0 (8.33)
2λ ill
Notice that for scattering in the forward direction where k ≈ ki , cos θs ≈

− cos θi and the surface integral vanishes. Physically, the scattering produced
by the illuminated side is due to reflection of the incident plane wave by the
scatterer surface, and very little reflection occurs in the forward direction.
The scattered wave function and its gradient on the dark surface of the
scatterer are approximated by:
0
ψω,s (r 0 ) ≈ −Ai eiki ·r
0
∇0 ψω,s (r 0 ) ≈ −iki Ai eiki ·r
Inserting these expressions into the Kirchoff integral yields the contribution
to the total scattered wave function by the dark side of the scatterer:
eikr
Z h i
0 0 0
ψω,s,drk (r) ≈ e−ik·r −iki Ai eiki ·r − ikAi eiki ·r · dS 0
4πr drk
iAi eikr
Z
0
≈− [ki + k]e−i(k−ki )·r · dS 0 (8.34)
4πr drk
For scattering in the backward direction, k ≈ −ki and the surface inte-
gral vanishes. Physically, the scattering produced by the dark side is due to
diffraction of the incident plane wave around the edges of the scatterer, which
only occurs in the forward direction where k ≈ ki . This substitution can be
made within the bracketed term, but the complex exponential is more sensi-
tive to small differences in k and ki . For an incident plane wave propagating
in the -z direction where ki = −k ẑ, the approximation k ≈ (kx x̂ + ky ŷ − k ẑ)
can be used such that:
i2Ai eikr
Z
0
ψω,s,drk (r) ≈ − e−i[(kx x̂+ky ŷ−kẑ)−(−kẑ)]·r ki · dS 0
4πr
Z drk
ik2Ai eikr 0 0
≈ e−i(kx x +ky y ) dSz0
4πr drk
On the dark side of the scatterer, dSz0 = −dx0 dy 0 where the integration limits
for x0 and y 0 correspond to the boundaries of the physical cross section of the
scatterer σp , so:
iAi eikr
Z
0 0
ψω,s,drk (r) ≈ − e−i(kx x +ky y ) dx0 dy 0 (8.35)
λ r σp
which is identical to the diffracted wave produced by an aperture with the

same shape as the physical cross section of the scatterer. When added to the
incident plane wave, the two destructively interfere to produce the diffracted
wave for a flat obstacle with the same shape as the physical cross section of
the scatterer. The corresponding contribution to the total scattering length
is given by:
Z
i 0 0
L̃drk (θ, φ, ω) ≈ − e−i(kx x +ky y ) dx0 dy 0 (8.36)
λ σp
Consider a rigid sphere as an example, where R(θi ) = 1. The illuminated

scattering length is difficult to calculate for all scattering angles, but is man-
ageable for the backward direction. In that case, θs = θi and for a plane
wave incident along the -z direction, k − ki = k ẑ − −k ẑ = 2k ẑ such that:
Z
i 0
L̃ill,back (ω) ≈ cos θi e−i2kz dS 0
λ ill
Working in spherical coordinates where θi = θ0 :

Z π/2 Z 2π
i 0
L̃ill,back (ω) ≈ cos θ0 e−i2ka cos θ a2 sin θ0 dφ0 dθ0
λ 0 0
Z π/2
2π 0
≈ ia2 cos θ0 sin θ0 e−i2ka cos θ dθ0
λ 0
Substituting x = cos θ0 yields:
Z 0
L̃ill,back (ω) ≈ −ika 2
xe−i2kax dx
1 Z 0
−1 −i2kax 0
2 1 −i2kax
≈ −ika xe + e dx
i2ka 1 1 i2ka

2 1 −i2ka 1 −i2kax 0
≈ −ika e + 2 2e
i2ka 4k a

1

2 1 −i2ka 1 1 −i2ka
≈ −ika e + 2 2 − 2 2e
i2ka 4k a 4k a
At high frequencies where the Kirchoff approximation is valid, ka 1 and

the last two terms can be dropped leaving:
a
L̃ill,back (ω) ≈ − e−i2ka (8.37)
2
Ignoring the variable phase, this result matches the scattering length cal-
culated earlier for reflection from a rigid sphere L̃r . The scattered wave
produced by the dark side of the sphere is identical to the diffracted wave
produced by a circular aperture with radius a calculated in Section 7.3. The
corresponding scattering length is given by:
iπa2 2J1 (ka sin θ)

L̃drk ≈ − (8.38)
λ (ka sin θ)
which compares favorably with the exact result at forward angles and high
values of ka as illustrated in Figure 8.8.
12
10
8
|L̃|/(a/2)
0
135 140 145 150 155 160 165 170 175 180
θ
Figure 8.8: Normalized scattering length (|L̃|/(a/2)) for a rigid sphere

obtained using the Kirchoff method (dotted) and the boundary condition
method (solid) for ka = 10.
View publication stats

Intro To Wave Physics

Uploaded by

Copyright:

Available Formats

Intro To Wave Physics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Intro To Wave Physics

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Introduction to Wave Physics

Book · March 2021

The user has requested enhancement of the downloaded file.

Anthony L. Gerig, Ph.D.

Copyright © 2021 Anthony L. Gerig

To Julie, Clayton, Gretchen and my students

2 Three-dimensional Wave Functions 41

2.3 Energy Transmission . . . . . . . . . . . . . . . . . . . . . . . 65

4 Reflection and Transmission 87

5 Cavities and Waveguides 103

dimensional wave is a displacement wave propagating along a plucked string.

1.1 Wave Functions

where σ is a measure of the pulse width. The shape of the waveform at t is

Figure 1.1: A pulse propagating on a string in the positive x direction with

Because the value of t is arbitrary, this function yields the displacement of

ψ(x, t) = f (x ± vt) (1.1)

The propagation speed for most waves is determined by the properties

1.2 Harmonic Waves

which is the wave function for the harmonic wave.

ψ(x, t) = A cos (±kx − ωt)

ψ(x, t) = A cos (±kx − ωt + φ) (1.5)

1.3 Complex Representation

Taking the complex conjugate of a complex number is defined as reversing the

Euler’s identity expresses a complex exponential as a complex number,

Figure 1.3: Plot of a complex number in two dimensions.

(iθ)0 (iθ)1 (iθ)2 (iθ)3

a + bi = A cos θ + iA sin θ = Aeiθ (1.9)

Re(Aei(±kx−ωt+φ) ) = A cos (±kx − ωt + φ) = ψ(x, t) (1.10)

ψ(x, t) = Aei(±kx−ωt+φ) = Aeiφ ei(±kx−ωt) = Ãei(±kx−ωt)

ψ1 = Aei(kx−ωt) and ψ2 = Aei(kx−ωt+φ)

ψ1 = Aei(kx−ωt) and ψ2 = Aei(−kx−ωt)

Figure 1.4: Addition of phasors representing the complex amplitudes of inter-

The combined wave function is therefore:

Applying Euler’s identity:

= A[cos kx + i sin kx + cos (−kx) + i sin (−kx)]e−iωt

Taking the real part:

ψ = 2A cos (kx) cos ωt

Figure 1.6: A standing wave on a string generated by interfering harmonic

1.5 One-dimensional Wave Equation

in one-dimension is given by:

1.5.1 Wave Equation for a String

Figure 1.7: A representative string element of length dx under tension T

From the figure:

Applying the small angle approximation sin θ ≈ tan θ:

tan θx+dx − tan θx µ ∂ 2ψ

1.5.2 Separation of Variables

ψ(x, t) = X(x)T (t)

ψ(x, t) = A1 ei(kx−ωt+φ1 ) + A2 ei(−kx−ωt+φ2 )

T (t) = a cos ωt + b sin ωt and X(x) = c cos kx + d sin kx

ψ(x, t) = A1 cos kx cos ωt + A2 cos kx sin ωt

1.6 Fourier Transforms

where the integral represents a continuous sum over frequency/wavenumber

Integrating a complex exponential yields a delta function multiplied by 2π,

Dropping the prime from k 0 yields:

Consider the specific example of a Gaussian pulse. F̃ (k) is determined

which can be found in a table of integrals:

1.7 Energy Transmission

substituting T = µv 2 from Equation 1.2 and k = ω/v: