Waves
Waves
Waves
Richard Fitzpatrick
Professor of Physics
The University of Texas at Austin
Contents
1 Introduction 5
4 Coupled Oscillations 43
4.1 Two Spring-Coupled Masses . . . . . . . . . . . . . . . . . . 43
4.2 Two Coupled LC Circuits . . . . . . . . . . . . . . . . . . . . 48
4.3 Three Spring Coupled Masses . . . . . . . . . . . . . . . . . 51
4.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2 OSCILLATIONS AND WAVES
1 Introduction
Vibrations and Waves, A.P. French (W.W. Norton & Co., New York NY,
1971).
The Physics of Vibrations and Waves, 5th Edition, H.J. Pain (John Wiley &
Sons, Chichester UK, 1999).
Simple Harmonic Oscillation 7
Here, k > 0 is the so-called force constant of the spring. The negative sign
indicates that f(x) is indeed a restoring force (i.e., if the displacement is
positive then the force is negative, and vice versa). Note that the magnitude
of the restoring force is directly proportional to the displacement of the mass
from its equilibrium position (i.e., |f| ∝ x). Of course, Hooke’s law only
holds for relatively small spring extensions. Hence, the displacement of the
mass cannot be made too large. Incidentally, the motion of this particular
dynamical system is representative of the motion of a wide variety of me-
chanical systems when they are slightly disturbed from a stable equilibrium
state (see Section 2.4).
Newton’s second law of motion gives following time evolution equation
for the system:
m ẍ = −k x, (2.2)
x=0
Figure 2.1: Mass on a spring
where a > 0, ω > 0, and φ are constants. We can demonstrate that Equa-
tion (2.3) is indeed a solution of Equation (2.2) by direct substitution. Plug-
ging the right-hand side of (2.3) into Equation (2.2), and recalling from
standard calculus that d(cos θ)/dθ = − sin θ and d(sin θ)/dθ = cos θ, so
that ẋ = −ω a sin(ω t − φ) and ẍ = −ω2 a cos(ω t − φ), where use has
been made of the chain rule, we obtain
2π
T= . (2.6)
ω
This result is easily obtained from Equation (2.3) by noting that cos θ is a
periodic function of θ with period 2π: i.e., cos(θ + 2π) ≡ cos θ. It follows
that the motion repeats every time ω t increases by 2π: i.e., every time t
Simple Harmonic Oscillation 9
increases by 2π/ω. The frequency of the motion (i.e., the number of oscilla-
tions completed per second) is
1 ω
f= = . (2.7)
T 2π
It can be seen that ω is the motion’s angular frequency; i.e., the frequency
f converted into radians per second. Of course, f is measured in Hertz—
otherwise known as cycles per second. Finally, the phase angle, φ, determines
the times at which the oscillation attains its maximum displacement, x = a.
In fact, since the maxima of cos θ occur at θ = n 2π, where n is an arbitrary
integer, the times of maximum displacement are
φ
tmax = T n + . (2.8)
2π
Clearly, varying the phase angle simply shifts the pattern of oscillation back-
ward and forward in time. See Figure 2.3.
Table 2.1 lists the displacement, velocity, and acceleration of the mass
at various different phases of the simple harmonic oscillation cycle. The in-
formation contained in this table can easily be derived from Equation (2.3).
Note that all of the non-zero values shown in this table represent either the
10 OSCILLATIONS AND WAVES
Figure 2.3: Simple harmonic oscillation. The solid, short-dashed, and long
dashed-curves correspond to φ = 0, +π/4, and −π/4, respectively.
ωt − φ 0 π/2 π 3π/2
x +a 0 −a 0
ẋ 0 −ω a 0 +ω a
ẍ −ω2 a 0 +ω2 a 0
Here, use has been made of the trigonometric identities cos(−θ) ≡ cos θ
and sin(−θ) ≡ − sin θ. Hence, we deduce that
q
a= x02 + (v0/ω)2, (2.11)
and
v0
−1
φ = tan , (2.12)
ω x0
since sin2 θ + cos2 θ ≡ 1 and tan θ ≡ sin θ/ cos θ.
The kinetic energy of the system, which is the same as the kinetic energy
of the mass, is written
1 1
K= m ẋ2 = m a2 ω2 sin2(ω t − φ). (2.13)
2 2
The potential energy of the system, which is the same as the potential energy
of the spring, takes the form
1 1
U= k x2 = k a2 cos2(ω t − φ). (2.14)
2 2
Hence, the total energy is
1 1
E=K+U= k a2 = m ω2 a2, (2.15)
2 2
since m ω2 = k and sin2 θ + cos2 θ ≡ 1. Note that the total energy is a con-
stant of the motion. Moreover, the energy is proportional to the amplitude
squared of the oscillation. It is clear, from the above expressions, that the
simple harmonic oscillation of a mass on a spring is characterized by a con-
tinuous backward and forward flow of energy between kinetic and potential
components. The kinetic energy attains its maximum value, and the poten-
tial energy its minimum value, when the displacement is zero (i.e., when
12 OSCILLATIONS AND WAVES
x = 0). Likewise, the potential energy attains its maximum value, and the
kinetic energy its minimum value, when the displacement is maximal (i.e.,
when x = ±a). Note that the minimum value of K is zero, since the system
is instantaneously at rest when the displacement is maximal.
where a > 0 and φ are constants. Moreover, the above equation describes a
type of oscillation characterized by a constant amplitude, a, and a constant
angular frequency, ω. The phase angle, φ, determines the times at which
the oscillation attains its maximum value. Finally, the frequency of the os-
cillation (in Hertz) is f = ω/2π, and the period is T = 2π/ω. Note that
the frequency and period of the oscillation are both determined by the con-
stant ω, which appears in the simple harmonic oscillator equation, whereas
the amplitude, a, and phase angle, φ, are both determined by the initial
conditions—see Equations (2.9)–(2.12). In fact, a and φ are the two con-
stants of integration of the second-order ordinary differential equation (2.16).
Recall, from standard differential equation theory, that the most general
solution of an nth-order ordinary differential equation (i.e., an equation in-
volving a single independent variable and a single dependent variable in
which the highest derivative of the dependent with respect to the indepen-
dent variable is nth-order, and the lowest zeroth-order) involves n arbitrary
constants of integration. (Essentially, this is because we have to integrate
the equation n times with respect to the independent variable in order to
reduce it to zeroth-order, and so obtain the solution, and each integration
introduces an arbitrary constant: e.g., the integral of ṡ = a, where a is a
known constant, is s = a t + b, where b is an arbitrary constant.)
Multiplying Equation (2.16) by ṡ, we obtain
ṡ s̈ + ω2 ṡ s = 0. (2.18)
Simple Harmonic Oscillation 13
Note that the simple harmonic oscillation (2.17) can also be written in
the form
s(t) = A cos(ω t) + B sin(ω t), (2.24)
where A = a cos φ and B = a sin φ. Here, we have employed the trigono-
metric identity cos(x − y) ≡ cos x cos y + sin x sin y. Alternatively, (2.17)
can be written
s(t) = a sin(ω t − φ ′ ), (2.25)
where φ ′ = φ − π/2, and use has been made of the trigonometric identity
cos θ ≡ sin(π/2 + θ). Clearly, there are many different ways of representing
14 OSCILLATIONS AND WAVES
then s1(t) + s2(t) is also a solution. This can be verified by adding the pre-
vious two equations, and making use of the fact that d2s1/dt2 + d2s2/dt2 =
d2(s1 + s2)/dt2. Furthermore, it is easily demonstrated that any linear com-
bination of s1 and s2, such as a s1 + b s2, where a and b are constants, is
a solution. It is very helpful to know this fact. For instance, the solution
to the simple harmonic oscillator equation (2.16) with the initial conditions
s(0) = 1 and ṡ(0) = 0 is easily shown to be
Likewise, the solution with the initial conditions s(0) = 0 and ṡ(0) = 1 is
clearly
s2(t) = ω−1 sin(ω t). (2.29)
Thus, since the solutions to the simple harmonic oscillator equation are su-
perposable, the solution with the initial conditions s(0) = s0 and ṡ(0) = ṡ0
is s(t) = s0 s1(t) + ṡ0 s2(t), or
ṡ0
s(t) = s0 cos(ω t) + sin(ω t). (2.30)
ω
2.3 LC Circuit
L İ + Q/C = 0. (2.31)
Ï + ω2 I = 0, (2.32)
where
1
ω= √ . (2.33)
LC
Comparison with Equation (2.16) reveals that (2.32) is a simple harmonic
oscillator equation with the associated angular oscillation frequency ω. We
conclude that the current in an LC circuit executes simple harmonic oscilla-
tions of the form
I(t) = I0 cos(ω t − φ), (2.34)
where I0 > 0 and φ are constants. Now, according to Equation (2.31), the
potential difference, V = Q/C, across the capacitor is minus that across the
inductor, so that V = −L İ, giving
s s
L L
V(t) = I0 sin(ω t − φ) = I0 cos(ω t − φ − π/2). (2.35)
C C
Here, use has been made of the trigonometric identity sin θ ≡ cos(θ − π/2).
It follows that the voltage in an LC circuit oscillates at the same frequency
as the current, but with a phase shift of π/2. In other words, the voltage is
maximal when the current is zero, and vice versa. The amplitude p of the volt-
age oscillation is that of the current oscillation multiplied by L/C. Thus,
we can also write s
L
V(t) = I(t − ω−1 π/2). (2.36)
C
16 OSCILLATIONS AND WAVES
L.
C
Figure 2.4: An LC circuit.
is one in which all of the circuit energy resides in the capacitor. The initial
conditions are V(0) = −L İ(0) = V0 and I(0) = 0. It is easily demonstrated
that the current evolves in time as
V0
I(t) = − p sin(ω t). (2.39)
L/C
fixed support
pivot point
l
θ
T
mg
where use has been made of Equations (2.41) and (2.43), as well as the
trigonometric identities cos(2 θ) ≡ cos2 θ−sin2 θ and sin(2 θ) ≡ 2 sin θ cos θ.
Hence, the average power absorption during a cycle of the oscillation,
Z
1 T
hPi = P(t) dt, (2.45)
T 0
is zero, since it is easily demonstrated that hcos(2 ω t)i = hsin(2 ω t)i = 0. In
other words, any energy which the capacitor absorbs from the circuit during
one part of the oscillation cycle is returned to the circuit without loss during
another. The same goes for the inductor.
I θ̈ = τ, (2.46)
Simple Harmonic Oscillation 19
where I is the moment of inertia of the mass, and τ the torque acting about
the suspension point. For the case in hand, given that the mass is essentially
a point particle, and is situated a distance l from the axis of rotation (i.e.,
the suspension point), it is easily seen that I = m l2.
The two forces acting on the mass are the downward gravitational force,
m g, where g is the acceleration due to gravity, and the tension, T , in the
string. Note, however, that the tension makes no contribution to the torque,
since its line of action clearly passes through the suspension point. From
elementary trigonometry, the line of action of the gravitational force passes
a perpendicular distance l sin θ from the suspension point. Hence, the mag-
nitude of the gravitational torque is m g l sin θ. Moreover, the gravitational
torque is a restoring torque: i.e., if the mass is displaced slightly from its
equilibrium position (i.e., θ = 0) then the gravitational torque clearly acts
to push the mass back towards that position. Thus, we can write
τ = −m g l sin θ. (2.47)
l θ̈ + g sin θ = 0. (2.48)
Note that, unlike all of the other time evolution equations which we have ex-
amined so far in this chapter, the above equation is nonlinear [since sin(a θ) 6=
a sin θ], which means that it is generally very difficult to solve.
Suppose, however, that the system does not stray very far from its equi-
librium position (θ = 0). If this is the case then we can expand sin θ in a
Taylor series about θ = 0. We obtain
θ3 θ5
sin θ = θ − + + O(θ7). (2.49)
6 120
Clearly, if |θ| is sufficiently small then the series is dominated by its first term,
and we can write sin θ ≃ θ. This is known as the small angle approximation.
Making use of this approximation, the equation of motion (2.48) simplifies
to
θ̈ + ω2 θ ≃ 0, (2.50)
where r
g
ω= . (2.51)
l
20 OSCILLATIONS AND WAVES
where θ0 > 0 and φ are constants. We conclude that the pendulum swings
back and forth at a fixed frequency, ω, which depends on l and g, but is
independent of the amplitude, θ0, of the motion. Of course, this result only
holds as long as the small angle approximation remains valid. It turns out
that sin θ ≃ θ is a good approximation provided |θ| <∼ 6◦ . Hence, the period
of a simple pendulum is only amplitude independent when the amplitude
of the motion is less than about 6◦ .
2.5 Exercises
f = −k x − m ν ẋ, (3.1)
where m > 0 is the mass, k > 0 the spring force constant, and ν > 0 a
constant (with the dimensions of angular frequency) which parameterizes
the strength of the damping. The time evolution equation of the system thus
becomes [cf., Equation (2.2)]
ẍ + ν ẋ + ω02 x = 0, (3.2)
p
where ω0 = k/m is the undamped oscillation frequency [cf., Equation
(2.5)]. We shall refer to the above as the damped harmonic oscillator equa-
tion.
Let us search for a solution to Equation (3.2) of the form
where a > 0, γ > 0, ω1 > 0, and φ are all constants. By analogy with the
discussion in Section 2.1, we can interpret the above solution as a periodic
oscillation, of fixed angular frequency ω1 and phase angle φ, whose ampli-
tude decays exponentially in time as a(t) = a exp(−γ t). So, (3.3) certainly
24 OSCILLATIONS AND WAVES
Now, the only way in which the above equation can be satisfied at all times
is if the coefficients of exp(−γ t) cos(ω1 t − φ) and exp(−γ t) sin(ω1 t − φ)
separately equate to zero, so that
γ = ν/2, (3.9)
ω1 = (ω02 − ν2/4)1/2. (3.10)
assuming that ν < 2 ω0 (since ω12 = ω02 − ν2/4 clearly cannot be negative).
We conclude that the effect of a relatively small amount of damping, pa-
rameterized by the damping constant ν, on a system which exhibits simple
harmonic oscillation about a stable equilibrium state is to reduce the angular
frequency of the oscillation from its undamped value ω0 to (ω02 − ν2/4)1/2,
and to cause the amplitude of the oscillation to decay exponentially in time
at the rate ν/2. This modified type of oscillation, which we shall refer to as
damped harmonic oscillation, is illustrated in Figure 3.1. [Here, T0 = 2π/ω0,
ν T0 = 0.5, and φ = 0. The solid line shows x(t)/a, whereas the dashed
lines show ±a(t)/a.] Incidentally, if the damping is sufficiently large that
ν ≥ 2 ω0, which we shall assume is not the case, then the system does not
Damped and Driven Harmonic Oscillation 25
oscillate at all, and any motion simply decays away exponentially in time
(see Exercise 3).
Note that, although the angular frequency, ω1, and decay rate, ν/2,
of the damped harmonic oscillation specified in Equation (3.11) are deter-
mined by the constants appearing in the damped harmonic oscillator equa-
tion, (3.2), the initial amplitude, a, and the phase angle, φ, of the oscillation
are determined by the initial conditions. In fact, if x(0) = x0 and ẋ(0) = v0
then it follows from Equation (3.11) that
x0 = a cos φ, (3.12)
ν
v0 = − a cos φ + ω1 a sin φ, (3.13)
2
giving
" #1/2
(v0 + ν x0/2)2
a = x02 + , (3.14)
ω12
v0 + ν x0/2
−1
φ = tan . (3.15)
ω1 x0
Note, further, that the damped harmonic oscillator equation is a linear dif-
ferential equation: i.e., if x(t) is a solution then so is a x(t), where a is an
26 OSCILLATIONS AND WAVES
arbitrary constant. It follows that the solutions of this equation are superpos-
able, so that if x1(t) and x2(t) are two solutions corresponding to different
initial conditions then a x1(t) + b x2(t) is a third solution, where a and b are
arbitrary constants.
Multiplying the damped harmonic oscillator equation (3.2) by ẋ, we ob-
tain
ẋ ẍ + ν ẋ 2 + ω02 ẋ x = 0, (3.16)
which can be rearranged to give
dE
= −m ν ẋ 2, (3.17)
dt
where
1 1
E= m ẋ 2 + k x2 (3.18)
2 2
is the total energy of the system: i.e., the sum of the kinetic and potential
energies. Clearly, since the right-hand side of (3.17) cannot be positive, and
is only zero when the system is stationary, the total energy is not a conserved
quantity, but instead decays monotonically in time due to the presence of
damping. Now, the net rate at which the force (3.1) does work on the mass
is
P = f ẋ = −k ẋ x − m ν ẋ2. (3.19)
Note that the spring force (i.e., the first term on the right-hand side) does
negative work on the mass (i.e., it reduces the system kinetic energy) when
ẋ and x are of the same sign, and does positive work when they are of the
opposite sign. On average, the spring force does no net work on the mass
during an oscillation cycle. The damping force, on the other hand, (i.e., the
second term on the right-hand side) always does negative work on the mass,
and, therefore, always acts to reduce the system kinetic energy.
The energy loss rate of a weakly damped (i.e., ν ≪ 2 ω0) harmonic oscillator
is conveniently characterized in terms of a parameter, Qf, which is known as
the quality factor. This quantity is defined to be 2π times the energy stored
in the oscillator, divided by the energy lost in a single oscillation period. If
the oscillator is weakly damped then the energy lost per period is relatively
small, and Qf is therefore much larger than unity. Roughly speaking, Qf is
the number of oscillations that the oscillator typically completes, after being
set in motion, before its amplitude decays to a negligible value. For instance,
Damped and Driven Harmonic Oscillation 27
the quality factor for the damped oscillation shown in Figure 3.1 is 12.6. Let
us find an expression for Qf.
As we have seen, the motion of a weakly damped harmonic oscillator is
specified by [see Equation (3.11)]
x = a e−νt/2 cos(ω1 t − φ), (3.20)
It follows that
a ν −νt/2
ẋ = − e cos(ω1 t − φ) − a ω1 e−νt/2 sin(ω1 t − φ). (3.21)
2
Thus, making use of Equation (3.17), the energy lost during a single oscil-
lation period is
Z (2π+φ)/ω1
dE
∆E = − dt (3.22)
φ/ω1 dt
Z (2π+φ)/ω1 2
−νt ν
2
= mνa e cos(ω1 t − φ) + ω1 sin(ω1 t − φ) dt.
φ/ω1 2
In the weakly damped limit, ν ≪ 2 ω0, the exponential factor is approxi-
mately unity in the interval t = φ/ω1 to (2π + φ)/ω1, so that
Z !
m ν a2 2π ν2 2 2 2
∆E ≃ cos θ + ν ω1 cos θ sin θ + ω1 sin θ dθ, (3.23)
ω1 0 4
where θ = ω1 t − φ. Thus,
π m ν a2 2 ν
∆E ≃ (ν /4 + ω12) = π m ω02 a2 , (3.24)
ω1 ω1
since, as is easily demonstrated,
Z 2π Z 2π
2
cos θ dθ = sin2 θ dθ = π, (3.25)
0 0
Z 2π
cos θ sin θ dθ = 0. (3.26)
0
Now, the energy stored in the oscillator (at t = 0) is [cf., Equation (2.15)]
1
E= m ω02 a2. (3.27)
2
Hence, we obtain
E ω1 ω0
Qf = 2π = ≃ . (3.28)
∆E ν ν
28 OSCILLATIONS AND WAVES
L .
I R
C
Figure 3.2: An LCR circuit.
where
1 2 1 Q2
E= LI + . (3.36)
2 2 C
Clearly, E is the circuit energy: i.e., the sum of the energies stored in the
inductor and the capacitor. Moreover, according to Equation (3.35), the cir-
cuit energy decays in time due to the power R I2 dissipated via Joule heating
in the resistor. Note that the dissipated power is always positive: i.e., the
circuit never gains energy from the resistor.
Finally, a comparison of Equations (3.28), (3.31), and (3.32) reveals
that the quality factor of an LCR circuit is
p
L/C
Qf = . (3.37)
R
Piston x
X=0 x=0
which one end of the spring is attached to the mass, and the other to a mov-
ing piston. See Figure 3.3. Let x(t) be the horizontal displacement of the
mass, and X(t) the horizontal displacement of the piston. The extension of
the spring is thus x(t) − X(t), assuming that the spring is unstretched when
x = X = 0. Thus, the horizontal force acting on the mass can be written [cf.,
Equation (3.1)]
f = −k (x − X) − m ν ẋ. (3.38)
The equation of motion of the system then becomes [cf., Equation (3.2)]
ẍ + ν ẋ + ω02 x = ω02 X, (3.39)
where ν > 0 is the damping constant, and ω0 > 0 the undamped oscillation
frequency. Suppose, finally, that the piston executes simple harmonic oscil-
lation of angular frequency ω > 0 and amplitude X0 > 0, so that the time
evolution equation of the system takes the form
ẍ + ν ẋ + ω02 x = ω02 X0 cos(ω t). (3.40)
We shall refer to the above as the driven damped harmonic oscillator equa-
tion.
Now, we would generally expect the periodically driven oscillator shown
in Figure 3.3 to eventually settle down to a steady (i.e., constant amplitude)
pattern of oscillation, with the same frequency as the piston, in which the
frictional energy loss per cycle is exactly matched by the work done by the
piston per cycle (see Exercise 7). This suggests that we should search for a
solution to Equation (3.40) of the form
x(t) = x0 cos(ω t − ϕ). (3.41)
Damped and Driven Harmonic Oscillation 31
Here, x0 > 0 is the amplitude of the driven oscillation, whereas ϕ is the phase
lag of this oscillation (with respect to the phase of the piston oscillation).
Now, since
Now, the only way in which the above equation can be satisfied at all times
is if the coefficients of cos(ω t) and sin(ω t) separately equate to zero. In
other words,
ω02 X0
x0 = 1/2 , (3.48)
(ω02 − ω2)2 + ν2 ω2
!
νω
ϕ = tan−1 2
. (3.49)
ω0 − ω2
Let us investigate the dependence of the amplitude, x0, and phase lag,
ϕ, of the driven oscillation on the driving frequency, ω. This is most easily
done graphically. Figure 3.4 shows x0/X0 and ϕ plotted as functions of ω
for various different values of ν/ω0. In fact, ν/ω0 = 1/Qf = 1, 1/2, 1/4,
1/8, and 1/16 correspond to the solid, dotted, short-dashed, long-dashed,
and dot-dashed curves, respectively. It can be seen that as the amount of
damping in the system is decreased the amplitude of the response becomes
progressively more peaked at the natural frequency of oscillation of the sys-
tem, ω0. This effect is known as resonance, and ω0 is termed the resonant
32 OSCILLATIONS AND WAVES
x0(ω = ω0) ω0
= = Qf. (3.50)
X0 ν
In other words, when the driving frequency matches the resonant frequency
the ratio of the amplitude of the driven oscillation to that of the piston os-
cillation is the quality factor, Qf. Hence, Qf can be regarded as the resonant
amplification factor of the oscillator. Equations (3.48) and (3.49) imply that,
for a weakly damped oscillator (i.e., ν ≪ ω0) which is close to resonance
[i.e., |ω − ω0| ∼ ν ≪ ω0],
x0(ω) ν
≃ sin ϕ ≃ . (3.51)
x0(ω = ω0) [4 (ω0 − ω)2 + ν2]1/2
π/4 and 3π/4, respectively. Furthermore, the fractional width of the peak is
∆ω ν 1
= = . (3.52)
ω0 ω0 Qf
We conclude that the height and width of the resonance peak of a weakly
damped (Qf ≫ 1) harmonic oscillator scale as Qf and Q−1 f , respectively.
Thus, the area under the resonance peak stays approximately constant as
Qf varies.
Now, the force exerted on the system by the piston is
1
hPi = k X0 x0 ω sin ϕ, (3.55)
2
since hcos(ω t) sin(ω t)i = 0 and hcos2(ω t)i = 1/2. Of course, given that
the amplitude of the driven oscillation neither grows nor decays, the average
power absorption from the piston during an oscillation cycle must be equal
to the average power dissipation due to friction (see Exercise 7). Making
use of Equations (3.50) and (3.51), the mean power absorption when the
driving frequency is close to the resonant frequency is
" #
1 ν2
hPi ≃ ω0 k X02 Qf . (3.56)
2 4 (ω0 − ω)2 + ν2
L C R
I
V
Figure 3.5: A driven LCR circuit.
ω V0
Ï + ν İ + ω02 I = cos(ω t), (3.59)
L
√
where ω0 = 1/ L C and ν = R/L. Comparison with Equation (3.40) reveals
that this is a driven damped harmonic oscillator equation. It follows, by
comparison with the analysis contained in the previous section, that the
current driven in the circuit by the oscillating emf is of the form
where
ω V0/L
I0 = 1/2 , (3.61)
(ω02 − ω2)2 + ν2 ω2
!
−1 νω
ϕ = tan 2
. (3.62)
ω0 − ω2
Recall that the steady-state (i.e., constant amplitude) solution to this equa-
tion which we found earlier takes the form
where
ω02 X0
x0 = 1/2 , (3.69)
(ω02 − ω2)2 + ν2 ω2
!
−1 νω
ϕ = tan 2
. (3.70)
ω0 − ω2
ẍ + ν ẋ + ω02 x = 0, (3.71)
to (3.68) then the result will still be a solution to Equation (3.67). Now,
from Section 3.1, the most general solution to the above equation can be
written
xtr(t) = A e−νt/2 cos(ω1 t) + B e−νt/2 sin(ω1 t), (3.72)
where ω1 = (ω02 − ν2/4)1/2, and A and B are arbitrary constants. [In terms
of the standard solution (3.11), A = a cos φ and B = a sin φ.] Thus, a
more general solution to (3.67) is
In fact, since the above solution contains two arbitrary constants, we can be
sure that it is the most general solution. Of course, the constants A and B are
Damped and Driven Harmonic Oscillation 37
determined by the initial conditions. It is, thus, clear that the most general
solution to the driven damped harmonic oscillator equation (3.67) consists
of two parts. First, the solution (3.68), which oscillates at the driving fre-
quency ω with a constant amplitude, and which is independent of the initial
conditions. Second, the solution (3.72), which oscillates at the natural fre-
quency ω1 with an amplitude which decays exponentially in time, and which
depends on the initial conditions. The former is termed the time asymptotic
solution, since if we wait long enough then it becomes dominant. The latter
is called the transient solution, since if we wait long enough then it decays
away.
Suppose, for the sake of argument, that the system is initially in its equi-
librium state: i.e., x(0) = ẋ(0) = 0. It follows from (3.73) that
since Qf = ω0/ν. Thus, the driven response oscillates at the resonant fre-
quency, ω0, since both the time asymptotic and transient solutions oscillate
at this frequency. However,
the amplitude
of the oscillation grows monotoni-
cally as a(t) = X0 Qf 1 − e −νt/2 , and so takes a time of order ν−1 to attain
its final value X0 Qf, which is, of course, larger that the driving amplitude
by the resonant amplification factor (or quality factor), Qf. This behavior
is illustrated in Figure 3.6. [Here, T0 = 2π/ω0 and Qf = ω0/ν = 16. The
solid curve shows x(t)/X0 and the dashed curves show ±a(t)/X0.]
Consider, now, the situation in which there is no damping, so that ν = 0.
In this case, Equation (3.81) yields
ω0
x(t) = X0 sin[(ω0 − ω) t/2] sin[(ω0 + ω) t/2], (3.83)
ω0 − ω
where use has been made of the trigonometry identity cos a − cos b ≡
−2 sin[(a + b)/2] sin[(a − b)/2]. It can be seen that the driven response
Damped and Driven Harmonic Oscillation 39
oscillates relatively rapidly at the “sum frequency” (ω0 + ω)/2 with an am-
plitude a(t) = X0 [ω0/(ω0 − ω)] sin[(ω0 − ω)/t] which modulates relatively
slowly at the “difference frequency” (ω0 − ω)/2. (Recall, that we are assum-
ing that ω is close to ω0.) This behavior is illustrated in Figure 3.7. [Here,
T0 = 2π/ω0 and ω0 − ω = ω0/16. The solid curve shows x(t)/X0 and the
dashed curves show ±a(t)/X0.] The amplitude modulations shown in Fig-
ure 3.7 are called beats, and are produced whenever two sinusoidal oscilla-
tions of similar amplitude, and slightly different frequency, are superposed.
In this case, the two oscillations are the time asymptotic solution, which
oscillates at the driving frequency, ω, and the transient solution, which os-
cillates at the resonant frequency, ω0. The beats modulate at the difference
frequency, (ω0 − ω)/2. In the limit ω → ω0, Equation (3.83) yields
X0
x(t) = ω0 t sin(ω0 t), (3.84)
2
since sin x ≃ x when |x| ≪ 1. Thus, the resonant response of a driven un-
damped oscillator is an oscillation at the resonant frequency whose ampli-
tude, a(t) = (X0/2) ω0 t, increases linearly in time. In this case, the period
of the beats has effectively become infinite.
Finally, Figure 3.8 illustrates the non-resonant response of a driven dam-
ped harmonic oscillator, obtained from Equation (3.81). [Here, T0 = 2π/ω0,
40 OSCILLATIONS AND WAVES
3.7 Exercises
1. Show that the ratio of two successive maxima in the displacement of a damped
harmonic oscillator is constant.
2. If the amplitude of a damped harmonic oscillator decreases to 1/e of its initial
value after n ≫ 1 periods show that the ratio of the period of oscillation to
the period of the oscillation with no damping is
1/2
1 1
1+ ≃1+ .
4 π2 n2 8 π2 n2
3. Many oscillatory systems are subject to damping effects which are not exactly
analogous to the frictional damping considered in Section 3.1. Nevertheless,
such systems typically exhibit an exponential decrease in their average stored
Damped and Driven Harmonic Oscillation 41
I0 cos(ω t)
L R C
.
(c) Demonstrate that this expression yields (3.56) in the limit that the driv-
ing frequency is close to the resonant frequency.
8. A generator of emf V(t) = V0 cos(ω t) is connected in series with a resistance
R, an inductor L, and a capacitor C. Let I(t) be the current flowing in the
circuit, and Q(t) the charge on the capacitor. Suppose that I = Q = 0 at
t = 0. Find I(t) and Q(t) for t > 0.
9. The equation m ẍ + k x = F0 sin(ω t) governs the motion of an undamped
harmonic oscillator driven by a sinusoidal force of angular frequency ω.
Show that the steady-state solution is
F0 sin(ω t)
x= ,
m (ω02 − ω2 )
p
where ω0 = k/m. Sketch the behavior of x versus t for ω < ω0 and
ω > ω0 . Demonstrate that if x = ẋ = 0 at t = 0 then the general solution is
F0 ω
x= sin(ω t) − sin(ω0 t) .
m (ω02 − ω2 ) ω0
F0
x≃ [sin(ω0 t) − ω0 t cos(ω0 t)] .
2 ω02
4 Coupled Oscillations
Here, we have made use of the fact that a mass attached to the left end
of a spring of extension x and spring constant k experiences a horizontal
force +k x, whereas a mass attached to the right end of the same spring
experiences an equal and opposite force −k x.
Equations (4.1)–(4.2) can be rewritten in the form
x1 x2
k k k
m m
x1 = 0 x2 = 0
p
where ω0 = k/m. Let us search for a solution in which the two masses
oscillate in phase at the same angular frequency, ω. In other words,
^1 cos(ω t − φ),
x1(t) = x (4.5)
^2 cos(ω t − φ),
x2(t) = x (4.6)
where x
^1, x
^2, and φ are constants. Equations (4.3) and (4.4) yield
− ω2 x
^1 cos(ω t − φ) = −2 ω02 x
^1 + ω02 x
^2 cos(ω t − φ), (4.7)
−ω2 x
^2 cos(ω t − φ) = ω02 x
^1 − 2 ω02 x
^2 cos(ω t − φ), (4.8)
or
^ 2 − 2) x
(ω ^1 + x
^2 = 0, (4.9)
^ 2 − 2) x
^1 + (ω
x ^2 = 0, (4.10)
where ω ^ = ω/ω0. Note that by searching for a solution of the form (4.5)–
(4.6) we have effectively converted the system of two coupled linear differ-
ential equations (4.3)–(4.4) into the much simpler system of two coupled
linear algebraic equations (4.9)–(4.10). The latter equations have the trivial
solutions x
^1 = x^2 = 0, but also yield
^1
x 1
=− 2 ^ 2 − 2).
= −(ω (4.11)
^2
x (ω
^ − 2)
^ 2 − 2) (ω
(ω ^ 2 − 2) − 1 = 0. (4.12)
^ 2−2
! ! !
ω 1 ^1
x 0
2
= , (4.13)
1 ^ −2
ω ^2
x 0
then it is clear that the criterion (4.12) can also be obtained by setting the
determinant of the associated 2 × 2 matrix to zero.
Equation (4.12) can be rewritten
^ 4− 4ω
ω ^ 2 + 3 = (ω
^ 2 − 1) (ω
^ 2 − 3) = 0. (4.14)
Coupled Oscillations 45
It follows that √
^ = 1 or
ω 3. (4.15)
Here, we have neglected
√ the two negative frequency roots of (4.14)—i.e.,
ω^ = −1 and ω ^ = − 3—since a negative frequency oscillation is equivalent
to an oscillation with an equal and opposite positive frequency, and an equal
and opposite phase: i.e., cos(ω t − φ) ≡ cos(−ω t + φ). It is thus apparent
that the dynamical system pictured in Figure
√ 4.1 has two unique frequencies
of oscillation: i.e., ω = ω0 and ω = 3 ω0. These are called the normal
frequencies of the system. Since the system possesses two degrees of freedom
(i.e., two independent coordinates are needed to specify its instantaneous
configuration) it is not entirely surprising that it possesses two normal fre-
quencies. In fact, it is a general rule that a dynamical system with N degrees
of freedom possesses N normal frequencies.
The patterns of motion associated with the two normal frequencies can
easily be deduced from Equation (4.11). Thus, for ω = ω0 (i.e., ω ^ = 1), we
get x ^2, so that
^1 = x
^ 1 cos(ω0 t − φ1),
x1(t) = η (4.16)
^ 1 cos(ω0 t − φ1),
x2(t) = η (4.17)
Note that we can be sure that this represents the most general solution to
Equations (4.1) and (4.2) because it contains four arbitrary constants: i.e.,
^ 1, φ1, η
η ^2, and φ2. (In general, we expect the solution of a second-order
ordinary differential equation to contain two arbitrary constants. It, thus,
follows that the solution of a system of two coupled ordinary differential
equations should contain four arbitrary constants.) Of course, these con-
stants are determined by the initial conditions.
For instance, suppose that x1 = a, ẋ1 = 0, x2 = 0, and ẋ2 = 0 at t = 0. It
follows, from (4.20) and (4.21), that
^1 cos φ1 + η
a = η ^2 cos φ2, (4.22)
√
^1 sin φ1 + 3 η
0 = η ^ 2 sin φ2, (4.23)
^1 cos φ1 − η
0 = η ^2 cos φ2, (4.24)
√
^1 sin φ1 − 3 η
0 = η ^ 2 sin φ2, (4.25)
It follows from (4.20) and (4.21) that, in the presence of both normal
modes,
^1 cos(ω0 t − φ1),
η1(t) = η (4.30)
√
^2 cos( 3 ω0 t − φ2).
η2(t) = η (4.31)
Consider the LC circuit pictured in Figure 4.3. Let I1(t), I2(t), and I3(t) be
the currents flowing in the three legs of the circuit, which meet at junctions
A and B. According to Kichhoff’s first circuital law, the net current flowing
into each junction is zero. It follows that I3 = −(I1 +I2). Hence, this is a two
degree of freedom system whose instantaneous configuration is specified by
the two independent variables I1(t) and I2(t). It follows that there are two
independent normal modes of oscillation. Now, the potential differences
across the left, middle, and right legs of the circuit are Q1/C + L İ1, Q3/C ′ ,
and Q2/C + L İ2, respectively, where Q̇1 = I1, Q̇2 = I2, and Q3 = −(Q1 +
Q2). However, since the three legs are connected in parallel, the potential
differences must all be equal, so that
Q1/C + L İ1 = Q3/C ′ = −(Q1 + Q2)/C ′ , (4.34)
′ ′
Q2/C + L İ2 = Q3/C = −(Q1 + Q2)/C . (4.35)
Differentiating with respect to t, and dividing by L, we obtain the coupled
time evolution equations of the system:
Ï1 + ω02 (1 + α) I1 + ω02 α I2 = 0, (4.36)
Ï2 +ω02 (1+ α) I2 + ω02 α I1
= 0, (4.37)
√
where ω0 = 1/ L C and α = C/C ′ .
It is fairly easy to guess that the normal coordinates of the system are
η1 = (I1 + I2)/2, (4.38)
η2 = (I1 − I2)/2. (4.39)
Forming the sum and difference of Equations (4.36) and (4.37), we obtain
the evolution equations for the two independent normal modes of oscilla-
tion:
η̈1 + ω02 (1 + 2 α) η1 = 0, (4.40)
η̈2 + ω02 η2 = 0. (4.41)
Coupled Oscillations 49
L . A L .
I1 I3 I2
Q1 C Q3 C′ Q2 C
B
Figure 4.3: Two degree of freedom LC circuit.
Substitution into the time evolution equations (4.36) and (4.37) yields the
matrix equation
^ 2 − (1 + α) ^I1
! ! !
ω −α 0
= , (4.50)
−α 2
^ − (1 + α)
ω ^I2 0
where ω^ = ω/ω0. The normal frequencies are determined by setting the
determinant of the matrix to zero. This gives
h i2
^ 2 − (1 + α)
ω − α2 = 0, (4.51)
or
^ 4 − 2 (1 + α) ω
ω ^ 2−1
^ 2+ 1 + 2α = ω ^ 2 − [1 + 2 α] = 0.
ω (4.52)
^1 cos(ω t − φ),
x1(t) = x (4.60)
^2 cos(ω t − φ),
x2(t) = x (4.61)
^3 cos(ω t − φ).
x3(t) = x (4.62)
^ 2−2
ω 1 0 ^1
x 0
1 ω 2
^ −2 1 ^2 = 0 , (4.63)
x
0 1 ^ 2−2
ω ^3
x 0
52 OSCILLATIONS AND WAVES
or h √ ih 2 √ i
^ 2 − 2) ω
(ω ^ 2−2− 2 ω ^ − 2 + 2 = 0. (4.65)
√ √ √ √
Thus, ^ = 2 (1 − 1/ 2)1/2, 2, and 2 (1 +
the normal frequencies are ω
√ 1/2
1/ 2) . According to the first and third rows of Equation (4.63),
^1 : x
x ^2 : x ^ 2 : 1,
^3 :: 1 : 2 − ω (4.66)
^1 : x
x ^2 : x
^3 :: −1 : 0 : 1 (4.67)
when ω ^ 2 = 2. Note that we can only determine the relative ratios of x ^1,
^2, and x
x ^3, rather than the absolute values of these quantities. In other
words, only the direction of the vector x ^ = (^
x1, x ^3) is well-defined. [This
^2, x
follows because the most general solution, (4.71), is undetermined to an
arbitrary multiplicative constant. That is, if x(t) = (x1(t), x2(t), x3(t)) is a
solution to the dynamical equations (4.57)–(4.59) then so is a x(t), where
a is an arbitrary constant. This, in turn, follows because the dynamical
equations are linear.] Let us arbitrarily set the magnitude of x ^ to unity. It
follows
√ that
√ the normal mode associated with the normal frequency ω ^1 =
2 (1 − 1/ 2)1/2 is
1 1 1
^1 =
x , √ , . (4.68)
2 2 2
√
Likewise, the normal mode associated with the normal frequency ω ^2 = 2
is
1 1
^2 = − √ , 0, √ .
x (4.69)
2 2
√
Finally,
√ 1/2 the normal mode associated with the normal frequency ω ^ 3 = 2(1+
1/ 2) is
1 1 1
^3 =
x , −√ , . (4.70)
2 2 2
Let x = (x1, x2, x2). It follows that the most general solution to the problem
is
^1 + η2(t) x
x(t) = η1(t) x ^2 + η3(t) x
^3, (4.71)
Coupled Oscillations 53
where
^1 cos(ω
η1(t) = η ^ 1 t − φ1), (4.72)
^2 cos(ω
η2(t) = η ^ 2 t − φ2), (4.73)
^3 cos(ω
η3(t) = η ^ 3 t − φ3). (4.74)
Here, η
^1,2,3 and φ1,2,3 are constants. Equation (4.71) yields
√
x1 1/2 1/ 2 1/2 η1
√ √
x2 =
−1/ 2 0 1/ 2 η2 (4.75)
√
x3 1/2 −1/ 2 1/2 η3
The above equation can easily be inverted by noting that the matrix is uni-
tary: i.e., its transpose is equal to its inverse. Thus, we obtain
√
η1 1/2 −1/ 2 1/2 x1
√ √
η2 =
1/ 2 0 −1/ 2 x2 (4.76)
√
η3 1/2 1/ 2 1/2 x3
This equation determines the three normal coordinates, η1, η2, η3, in terms
of the three conventional coordinates, x1, x2, x3. Note that, in general, the
normal coordinates are undetermined to arbitrary multiplicative constants.
4.4 Exercises
4. Consider two simple pendula with the same length, l, but different bob
masses, m1 and m2 . Suppose that the pendula are connected by a spring
of spring constant k. Let the spring be unextended when the two bobs are in
their equilibrium positions. Demonstrate that the equations of motion of the
system (for small amplitude oscillations) are
g
m1 θ̈1 = −m1 θ1 + k (θ2 − θ1 ),
l
g
m2 θ̈2 = −m2 θ2 + k (θ2 − θ1 ),
l
where θ1 and θ2 are the angular displacements of the respective pendula
from their equilibrium positions. Show that the normal coordinates are η1 =
(m1 θ1 + m2 θ2 )/(m1 + m2 ) and η2 = θ1 − θ2 . Find the normal frequencies
and normal modes. Find a superposition of the two modes such that at t = 0
the two pendula are stationary, with θ1 = θ0 , and θ2 = 0.
5. Find the normal frequencies and√normal modes of the coupled LC circuit
shown below in terms of ω0 = 1/ L C and α = L ′ /L.
C C
I1 I2
L L′ L
.
.
Transverse Standing Waves 55
bead
y wall m string
T
x
a a a
equilibrium position of string
Figure 5.1: A beaded string.
56 OSCILLATIONS AND WAVES
θi+1
θi
y
T
yi−1 yi yi+1
x
xi−1 xi xi+1
oscillations in phase with one another. See Figure 5.4. Moreover, the os-
cillations have an amplitude A sin(k xi) which varies sinusoidally along the
length of the string (i.e., in the x-direction). The pattern of oscillations is
thus periodic in space. The spatial repetition period, which is usually known
as the wavelength, is λ = 2π/k. [This follows from (5.9) because sin θ is a
periodic function with period 2π: i.e., sin(θ + 2π) ≡ sin θ.] The constant
k, which determines the wavelength, is usually referred to as the wavenum-
ber. Thus, a small wavenumber corresponds to a long wavelength, and vice
versa. The type of solution specified in (5.9) is generally known as a standing
wave. It is a wave because it is periodic in both space and time. (An oscilla-
tion is periodic in time only.) It is a standing wave, rather than a traveling
wave, because the points of maximum and minimum amplitude oscillation
are stationary (in x). See Figure 5.4.
Substituting (5.9) into (5.8), we obtain
which yields
− ω2 sin(k xi) = ω02 (sin[k (xi − a)] − 2 sin(k xi) + sin[k (xi + a)]) . (5.11)
or
ω2 = 4 ω02 sin2(k a/2), (5.13)
where use has been made of the trigonometric identity 1−cos θ ≡ 2 sin2(θ/2).
Note that an expression, such as (5.13), which determines the angular fre-
quency, ω, of a wave in terms of its wavenumber, k, is generally known as a
dispersion relation.
Now, the solution (5.9) is only physical provided y0 = yN+1 = 0: i.e.,
provided that the two ends of the string remain stationary. The first con-
straint is automatically satisfied, since x0 = 0 [see (5.1)]. The second con-
straint implies that
where the integer n is known as the mode number. Clearly, a small mode
number translates to a small wavenumber, and, hence, to a long wavelength,
and vice versa. We conclude that the possible wavenumbers, k, of the nor-
mal modes of the system are quantized such that they are integer multiples
of π/[(N + 1) a]. Thus, the nth normal mode is associated with the charac-
teristic pattern of bead displacements
ni
yn,i(t) = An sin π cos(ωn t − φn), (5.16)
N+1
where
n π
ωn = 2 ω0 sin . (5.17)
N+1 2
Here, the integer i = 1, N indexes the beads, whereas the mode number n
indexes the normal modes. Furthermore, An and φn are arbitrary constants
determined by the initial conditions. Of course, An is the peak amplitude of
the nth normal mode, whereas φn is the associated phase angle.
So, how many unique normal modes does the system possess? At first
sight, it might seem that there are an infinite number of normal modes,
corresponding to the infinite number of possible values that the integer n
can take. However, this is not the case. For instance, if n = 0 or n = N + 1
then all of the yn,i are zero. Clearly, these cases are not real normal modes.
Moreover, it is easily demonstrated that
Figure 5.3: Normal modes of a beaded string with eight equally spaced beads.
Transverse Standing Waves 61
Figure 5.4: Time evolution of the n = 2 normal mode of a beaded string with
eight equally spaced beads.
Figure 5.5: Normal frequencies of a beaded string with eight equally spaced
beads.
is known as a node. The n = 1 normal mode has two nodes (counting the
stationary points at each end of the string as nodes), the n = 2 mode has
three nodes, the n = 2 mode four nodes, etc. In fact, the existence of nodes
is one of the distinguishing features of a standing wave.
Figure 5.5 shows the normal frequencies of an N = 8 beaded string
plotted as a function of the normalized wavenumber. Recall that, for an
N = 8 system, the relationship between the normalized wavenumber, k a,
and the mode number, n, is k a = (n/9) π. It can be seen that the an-
gular frequency increases as the wavenumber increases, which implies that
shorter wavelength modes have higher oscillation frequencies. Note that the
dependence of the angular frequency on the normalized wavenumber, k a,
is approximately linear when k a ≪ 1. Indeed, it can be seen from Equa-
tion (5.17) that if k a ≪ 1 then the small angle approximation sin θ ≃ θ
yields a linear dispersion relation of the form
n
ωn ≃ (k a) ω0 = π ω0. (5.22)
N+1
We, thus, conclude that those normal modes of a uniformly beaded string
whose wavelengths greatly exceed the bead spacing (i.e., modes with k a ≪
1) have approximately linear dispersion relations in which their angular fre-
Transverse Standing Waves 63
∂y(x, t) 1 ∂2y(x, t)
y(x + δx, t) = y(x, t) + δx + (δx)2 + O(δx)3. (5.24)
∂x 2 ∂x2
Likewise,
∂y(x, t) 1 ∂2y(x, t)
y(x − δx, t) = y(x, t) − δx + (δx)2 + O(δx)3. (5.25)
∂x 2 ∂x2
It follows that
y(x − δx, t) − 2 y(x, t) + y(x + δx, t) ∂2y(x, t)
= + O(δx). (5.26)
(δx)2 ∂x2
∂2y 2
2∂ y
= v , (5.27)
∂t2 ∂x2
where s
T
v= (5.28)
ρ
is a quantity having the dimensions of velocity. Equation (5.27), which is
the transverse equation of motion of the string, takes the form of a very
famous partial differential equation known as the wave equation. The quan-
tity v turns out to the the propagation velocity of transverse waves along the
string. See Section 7.1.
By analogy with Equation (5.9), let us search for a solution of the wave
equation of the form
y(0, t) = 0, (5.31)
y(l, t) = 0. (5.32)
It can be seen that the solution (5.29) automatically satisfies the first bound-
ary condition. However, the second boundary condition is only satisfied
when sin(k l) = 0, which immediately implies that
π
k=n , (5.33)
l
where the mode number, n, is an integer. We, thus, conclude that the
possible normal modes of a taut string, of length l and fixed ends, have
wavenumbers which are quantized such that they are integer multiples of
π/l. Moreover, it is clear that this quantization is a direct consequence of
the imposition of the physical boundary conditions at the two ends of the
string.
It follows, from the above analysis, that the nth normal mode of the
string is associated with the pattern of motion
x
yn(x, t) = An sin n π cos(ωn t − φn), (5.34)
l
where
πv
ωn = n . (5.35)
l
Here, An and φn are constants which are determined by the initial con-
ditions. See Section 5.3. So, how many unique normal modes are there?
Well, the choice n = 0 yields y0(x, t) = 0 for all x and t, so this is not a real
normal mode. Moreover,
provided that A−n = −An and φ−n = −φn. We, thus, conclude that modes
with negative mode numbers give rise to the same patterns of motion as
modes with corresponding positive mode numbers. However, the modes
with positive mode numbers each correspond to unique patterns of motion
which oscillate at unique frequencies. It follows that the string possesses
an infinite number of normal modes, corresponding to the mode numbers
n = 1, 2, 3, etc. Recall that we are dealing with an infinite degree of freedom
system, which we would expect to possess an infinite number of unique
normal modes. The fact that we have actually found an infinite number
of such modes suggests that we have found all of the normal modes of the
system.
Figure 5.6 illustrates the spatial variation of the first eight normal modes
of a uniform string with fixed ends. The modes are all shown at the instances
in time at which they attain their maximum amplitudes: i.e., at ωn t − φn =
0. It can be seen that the modes are all smoothly varying sine waves. The
low mode number (i.e., long wavelength) modes are actually quite similar
in form to the corresponding normal modes of a uniformly beaded string.
See Figure 5.3. However, the high mode number modes are substantially
different. We conclude that the normal modes of a beaded string are similar
to those of a uniform string, with the same length and mass per unit length,
provided that the wavelength of the mode is much larger than the spacing
between the beads.
Figure 5.7 illustrates the temporal variation of the n = 4 normal mode
of a uniform string. The mode is shown at ω4 t − φ4 = 0, π/8, π/4, 3π/8,
π/2, 5π/8, 3π/2, 7π/8 and π. It can be seen that all points on the string
attain their maximal transverse displacements, and pass through zero dis-
placement, simultaneously. Note that the n = 4 mode possesses five nodes,
at which the string remains stationary. Two of these are located at the ends
of the string, and three in the middle. In fact, it is clear from Equation (5.34)
that the nodes correspond to points at which sin[n (x/l) π] = 0. Hence, the
nodes are located at
j
xn,j = l, (5.38)
n
where j is an integer lying in the range 0 to n. Here, n indexes the normal
mode, and j the node. Thus, the j = 0 node lies at the left end of the
string, the j = 1 node is the next node to the right, etc. It is apparent, from
the above formula, that the nth normal mode has n + 1 nodes which are
uniformly spaced a distance l/n apart.
Finally, Figure 5.8 shows the first eight normal frequencies of a uniform
string with fixed ends, plotted as a function of the mode number. It can
Transverse Standing Waves 67
Figure 5.8: Normal frequencies of the first eight normal modes of a uniform
string.
Transverse Standing Waves 69
be seen that the angular frequency of oscillation increases linearly with the
mode number. Recall that the low mode number (i.e., long wavelength)
normal modes of a beaded string also exhibit a linear relationship between
normal frequency and mode number of the form [see Equation (5.22)]
1/2
nπ nπ T
ωn = ω0 = . (5.39)
N+1 N+1 ma
which is identical to Equation (5.35). We, thus, conclude that the normal
frequencies of a uniformly beaded string are similar to those of a uniform
string, with the same length and mass per unit length, as long as the wave-
length of the associated normal mode is much larger than the spacing be-
tween the beads.
(5.41)
where use has been made of (5.34) and (5.35). Note that this expression is
obviously a solution of (5.27), and also automatically satisfies the boundary
conditions (5.31) and (5.32). As we have already mentioned, the constants
An and φn are determined by the initial conditions. Let us see how this
comes about in more detail.
Suppose that the initial displacement of the string at t = 0 is
It follows that
sin(k π) 1 k=0
= , (5.49)
kπ 0 k 6= 0
Transverse Standing Waves 71
where k is any integer. This result can be combined with Equation (5.46),
recalling that n and n ′ are both positive integers, to give
Z
2 l x x
′
sin n π sin n π dx = δn,n′ . (5.50)
l 0 l l
Here, the quantity
1 n = n′
δn,n′ = , (5.51)
0 n 6= n ′
where n and n ′ are integers, is known as the Kronecker delta function.
Let us multiply Equation (5.44) by (2/l) sin(n π x/l) and integrate over
x from 0 to l. We obtain
Z
2 l x
y0(x) sin n π dx
l 0 l
Z
X 2 π x x
′
= An′ cos φn′ sin n π sin n π dx
l 0 l l
n′ =1,∞
X
= An′ cos φn′ δn,n′ = An cos φn, (5.52)
n′ =1,∞
where use has been made of Equations (5.50) and (5.51). Similarly, Equa-
tion (5.45) yields
Z
2 l x
v0(x) sin n π dx = n π An sin φn. (5.53)
v 0 l
Thus, defining the integrals
Zl
2 x
Cn = y0(x) sin n π dx, (5.54)
l 0 l
Zl
2 x
Sn = v0(x) sin n π dx, (5.55)
nπv 0 l
for n = 1, ∞, we obtain
Cn = An cos φn, (5.56)
Sn = An sin φn, (5.57)
and, hence,
An = (Cn2 + Sn2)1/2, (5.58)
φn = tan−1(Sn/Cn). (5.59)
72 OSCILLATIONS AND WAVES
Thus, the constants An and φn, appearing in the general expression (5.41)
for the time evolution of a uniform string with fixed ends, are ultimately
determined by integrals over the string’s initial displacement and velocity
which are of the form (5.54) and (5.55).
As an example, suppose that the string is initially at rest, so that
v0(x) = 0, (5.60)
This triangular pattern is zero at both ends of the string, rising linearly to
a peak value of A halfway along the string, and is designed to mimic the
initial displacement of a guitar string which is plucked at its mid-point. See
Figure 5.10. A comparison of Equations (5.55) and (5.60) reveals that, in
this particular example, all of the Sn coefficients are zero. Hence, from
(5.58) and (5.59), An = Cn and φn = 0 for all n. Thus, making use
of Equations (5.41), (5.54), and (5.61), the time evolution of the string is
governed by
X
x
t
y(x, t) = An sin n π cos n 2π , (5.62)
l τ
n=1,∞
Figure 5.9 shows the ratio An/A1 for the first ten overtone harmonics
of a uniform guitar string plucked at its midpoint: i.e., the ratio An/A1 for
odd-n modes with n > 1, calculated from Equation (5.65). It can be seen
that the amplitudes of the overtone harmonics are all small compared to the
amplitude of the fundamental. Moreover, the amplitudes decrease rapidly
in magnitude with increasing mode number, n.
In principle, we must include all of the normal modes in the sum on the
right-hand side of Equation (5.62). In practice, given that the amplitudes
of the normal modes decrease rapidly in magnitude as n increases, we can
truncate the sum, by neglecting high-n normal modes, without introducing
significant error into our calculation. Figure 5.11 illustrates the effect of
such a truncation. In fact, this figure shows the reconstruction of y0(x), ob-
tained by setting t = 0 in Equation (5.62), made with various different num-
bers of normal modes. The long-dashed line shows a reconstruction made
with only the largest amplitude normal mode, the short dashed-line shows a
reconstruction made with the four largest amplitude normal modes, and the
solid line shows a reconstruction made with the sixteen largest amplitude
normal modes. It can be seen that sixteen normal modes is sufficient to very
accurately reconstruct the triangular initial displacement pattern. Indeed, a
Transverse Standing Waves 75
Figure 5.11: Time evolution of a uniform guitar string plucked at its mid-point.
at x = l/2. At this instant, the string is again found in its initial position.
The pattern of motion then repeats itself ad infinitum. The period of the
oscillation is the time required for a kink to propagate two string lengths,
which is τ = 2 l/v. This, of course, is also the oscillation period of the n = 1
normal mode.
5.4 Exercises
where n π
ωn = 2 ω0 sin ,
N 2
ω0 is as defined in Section 5.1, An and φn are constants, the integer i = 1, N
indexes the beads, and the mode number n indexes the modes. How many
unique normal modes does the system possess, and what are their mode
numbers? Show that the lowest frequency mode has an infinite wavelength
and zero frequency. Explain this peculiar result. Plot the normal modes
and normal frequencies of an N = 8 beaded string in a similar fashion to
Figures 5.3 and 5.5.
I1 I2 IN −1 IN
. . . .
L L L L
C C C C
3. The above figure shows the left and right extremities of a linear LC network
consisting of N identical inductors of inductance L, and N + 1 identical ca-
pacitors of capacitance C. Let the instantaneous current flowing through the
Transverse Standing Waves 77
Zl
2 x
Sn = v0 (x) cos n π dx.
l 0 l
Suppose that the string is initially in its equilibrium state. At t = 0 it is struck
by a hammer in such a manner as to impart an impulsive velocity u0 to a
small segment of length a < l centered on the mid-point. Find an expression
for the subsequent motion of the string. Plot the motion as a function of time
in a similar fashion to Figure 5.11, assuming that a = l/10.
7. The linear LC circuit considered in Exercise 3 can be thought of as a discrete
model of a uniform lossless transmission line: e.g., a co-axial cable. In this
interpretation, Ii (t) represents I(xi , t), where xi = i δx. Moreover, C = C δx,
and L = L δx, where C and L are the capacitance per unit length and the
inductance per unit length of the line, respectively. Show that, in the limit
δx → 0, the evolution equation for the coupled currents given in Exercise 3
reduces to the wave equation
∂2 I ∂2 I
2
= v2 2 ,
∂t ∂x
√
where I = I(x, t), x measures distance along the line, and v = 1/ L C.
If Vi (t) is the potential difference (measured from the top to the bottom)
across the i + 1th capacitor (from the left) in the circuit shown in Exercise 3,
and V(x, t) is the corresponding voltage in the transmission line, show that
the discrete circuit equations relating the Ii (t) and Vi (t) reduce to
∂V 1 ∂I
= − ,
∂t C ∂x
∂I 1 ∂V
= − ,
∂t L ∂x
in the transmission line limit. Hence, demonstrate that the voltage in a trans-
mission line satisfies the wave equation
∂2 V ∂2 V
2
= v2 .
∂t ∂x2
8. Consider a uniform string of length l, tension T , and mass per unit length ρ
which is stretched between two immovable walls. Show that the total energy
of the string, which is the sum of its kinetic and potential energies, is
Z " 2 2 #
1 l ∂y ∂y
E= ρ +T dx,
2 0 ∂t ∂x
p
where v = T/ρ. Demonstrate that
X
E= En ,
n=1,∞
where
1
En = m ωn2 An2
4
is the energy of the nth normal mode. Here, m = ρ l is the mass of the string,
and ωn = n π v/l the angular frequency of the nth normal mode.
80 OSCILLATIONS AND WAVES
Longitudinal Standing Waves 81
ψi−1 ψi ψi+1
K K
m m m
xi−1 xi xi+1
p
where ω0 = K/m. Since there is nothing special about the ith mass,
the above equation is assumed to hold for all N masses: i.e., for i = 1, N.
Note that Equation (6.2), which governs the longitudinal oscillations of a lin-
ear array of spring coupled masses, is analogous in form to Equation (5.8),
which governs the transverse oscillations of a beaded string. This observa-
tion suggests that longitudinal and transverse waves in discrete dynamical
systems (i.e., systems with a finite number of degrees of freedom) can be
described using the same mathematical equations.
We can interpret the quantities ψ0 and ψN+1, which appear in the equa-
tions of motion for ψ1 and ψN, respectively, as the longitudinal displace-
ments of the left and right extremities of springs which are attached to the
outermost masses in such a manner as to form the left and right boundaries
of the array. The respective equilibrium positions of these extremities are
x0 = 0 and xN+1 = (N + 1) a. Now, the end displacements, ψ0 and ψN+1,
must be prescribed, otherwise Equations (6.2) do not constitute a complete
set of equations: i.e., there are more unknowns than equations. The partic-
ular choice of ψ0 and ψN+1 depends on the nature of the physical boundary
conditions at the two ends of the array. Suppose that the left extremity of
the leftmost spring is anchored in an immovable wall. This implies that
ψ0 = 0: i.e., the left extremity of the spring cannot move. Suppose, on the
other hand, that the left extremity of the leftmost spring is not attached to
anything. In this case, there is no reason for the spring to become extended,
which implies that ψ0 = ψ1. In other words, if the left end of the array is
fixed (i.e., attached to an immovable object) then ψ0 = 0, and if the left end
is free (i.e., not attached to anything) then ψ0 = ψ1. Likewise, if the right
end of the array is fixed then ψN+1 = 0, and if the right end is free then
ψN+1 = ψN.
Suppose, for the sake of argument, that the left end of the array is free,
and the right end is fixed. It follows that ψ0 = ψ1, and ψN+1 = 0. Let us
search for normal modes of the general form
where A > 0, k > 0, ω > 0, and φ are constants. Note that the above
expression automatically satisfies the boundary condition ψ0 = ψ1. This
follows because x0 = 0 and x1 = a, and, consequently, cos[k (x0 − a/2)] =
cos(−k a/2) = cos(k a/2) = cos[k (x1 − a/2)]. The other boundary condi-
tion, ψN+1 = 0, is satisfied provided
The dynamical system pictured in Figure 6.1 can be used to model the ef-
fect of a planar sound wave (i.e., a longitudinal oscillation in position which
is periodic in space in one dimension) on a crystal lattice. In this applica-
tion, the masses represent parallel planes of atoms, the springs represent
the interatomic forces acting between these planes, and the longitudinal
oscillations represent the sound wave. Of course, a macroscopic crystal con-
tains a great many atomic planes, so we would expect N to be very large.
Note, however, from Equations (6.5) and (6.8), that, no matter how large
N becomes, k a cannot exceed π (since n cannot exceed N), and ωn cannot
exceed 2 ω0. In other words, there is a minimum wavelength that a sound
wave in a crystal lattice can have, which turns out to be twice the inter-
atomic spacing, and a corresponding maximum oscillation frequency. For
waves whose wavelengths are much greater than the interatomic spacing
(i.e., k a ≪ 1), the dispersion relation (6.6) reduces to
ω ≃ kc (6.9)
p
where c = ω0 a = K/m a is a constant which has the dimensions of
velocity. It seems plausible that (6.9) is the dispersion relation for sound
waves in a continuous elastic medium. Let us investigate such waves.
86 OSCILLATIONS AND WAVES
∂ψ(x, t)
ǫ(x, t) = . (6.11)
∂x
Of course, it is assumed that the strain is small: i.e., |ǫ| ≪ 1. Stress, σ(x, t),
in an elastic rod is defined as the elastic force per unit cross-sectional area. In
a conventional elastic material, the relationship between stress and strain
(for small strains) takes the simple form
σ = Y ǫ. (6.12)
assuming that the strain is positive (i.e., the force acts to lengthen the sec-
tion). Likewise, the force acting on the right boundary of the section is
A Y ǫ(x + δx/2), and is directed in the positive x-direction, assuming that
the strain is positive (i.e., the force again acts to lengthen the section). Fi-
nally, the mean longitudinal (i.e., x-directed) acceleration of the section is
∂2ψ(x, t)/∂t2. Hence, the section’s longitudinal equation of motion becomes
∂2ψ(x, t)
ρ A δx = A Y [ǫ(x + δx/2, t) − ǫ(x − δx/2)] . (6.13)
∂t2
In the limit δx → 0, this expression reduces to
∂2ψ(x, t) ∂ǫ(x, t)
ρ 2
=Y , (6.14)
∂t ∂x
or
∂2ψ 2
2∂ ψ
= c , (6.15)
∂t2 ∂x2
p
where c = Y/ρ is a constant having the dimensions of velocity, which
turns out to be the speed of sound in the rod (see Section 7.1), and use has
been made of Equation (6.11). Of course, (6.15) is a wave equation. As
such, it has the same form as Equation (5.27), which governs the motion
of transverse waves on a uniform string. This suggests that longitudinal
and transverse waves in continuous dynamical systems (i.e., systems with
an infinite number of degrees of freedom) can be described using the same
mathematical equations.
In order to solve (6.15), we need to specify boundary conditions at the
two ends of the rod. Suppose that the left end of the rod is fixed: i.e., it
is clamped in place so that it cannot move. This implies that ψ(0, t) = 0.
Suppose, on the other hand, that the left end of the rod is free: i.e., it is not
attached to anything. This implies that σ(0, t) = 0, since there is nothing
that the end can exert a force (or a stress) on, and vice versa. It follows from
(6.11) and (6.12) that ∂ψ(0, t)/∂x = 0. Likewise, if the right end of the rod
is fixed then ψ(l, t) = 0, and if the right end is free then ∂ψ(l, t)/∂x = 0.
Suppose, for the sake of argument, that the left end of the rod is free,
and the right end is fixed. It follows that ∂ψ(0, t)/∂x = 0, and ψ(l, t) = 0.
Let us search for normal modes of the form
where A > 0, k > 0, ω > 0, and φ are constants. Note that the above
expression automatically satisfies the boundary condition ∂ψ(0, t)/∂x = 0.
88 OSCILLATIONS AND WAVES
cos(k l) = 0, (6.17)
which yields
k l = (n − 1/2) π, (6.18)
where n is an integer. As usual, the imposition of the boundary conditions
leads to a quantization of the possible mode wavenumbers. Substitution of
(6.16) into the equation of motion (6.15) yields the normal mode dispersion
relation s
Y
ω = kc = k . (6.19)
ρ
Note that this dispersion relation is consistent with the previously derived
dispersion relation (6.9), since m = ρ A a and K = A Y/a. Here, a is the
interatomic spacing, m the mass of a section of the rod containing a sin-
gle plane of atoms, and K the effective force constant between neighboring
atomic planes.
It follows, from the above analysis, that the nth longitudinal normal
mode of an elastic rod, of length l, whose left end is free, and whose right
end is fixed, is associated with the characteristic displacement pattern
x
ψn(x, t) = An cos (n − 1/2) π cos(ωn t − φn), (6.20)
l
where
πc
ωn = (n − 1/2) . (6.21)
l
Here, An and φn are constants which are determined by the initial condi-
tions. It is easily demonstrated that only those normal modes whose mode
numbers are positive integers yield unique displacement patterns: i.e., n > 0.
Equation (6.20) describes a standing wave whose nodes (i.e., points at which
ψ = 0 for all t) are evenly spaced a distance l/(n − 1/2) apart. Of course,
the boundary condition ψ(l, t) = 0 ensures that the right end of the rod
is always coincident with a node. On the other hand, the boundary con-
dition ∂ψ(0, t)/∂x = 0 ensures that the left hand of the rod is always co-
incident with a point of maximum amplitude oscillation [i.e., a point at
which cos(k x) = ±1]. Such a point is known as an anti-node. It is easily
demonstrated that the anti-nodes associated with a given normal mode lie
halfway between the corresponding nodes. Note, from (6.21), that the nor-
mal mode oscillation frequencies depend linearly on mode number. Finally,
it is easily demonstrated that, in the long wavelength limit k a ≪ 1, the
Longitudinal Standing Waves 89
πc X x
ψ̇(x, 0) = (n ′ − 1/2) An′ sin φn′ cos (n ′ − 1/2) π . (6.24)
l ′ l
n =1,∞
Thus, multiplying (6.23) by (2/l) cos[(n − 1/2) π x/l], and then integrating
over x from 0 to l, we obtain
Zl
2 x
Cn ≡ ψ(x, 0) cos (n − 1/2) π dx = An cos φn, (6.26)
l 0 l
where use has been made of (6.25) and (5.51). Likewise, (6.24) gives
Zl
2 x
Sn ≡ ψ̇(x, 0) cos (n − 1/2) π dx = An sin φn. (6.27)
c (n − 1/2) π 0 l
and
X
x
t
ψ(x, t) = An cos (n − 1/2) π sin (n − 1/2) π , (6.29)
l τ
n=1,∞
where τ = l/c. Figure 6.4 shows the time evolution of the normalized rod
^ t) = (c/V0 a) ψ(x, t), calculated from the above equa-
displacement, ψ(x,
tions using the first 100 normal modes (i.e., n = 1, 100), and choosing
a/l = 0.1. The top-left, top-right, middle-left, middle-right, bottom-left,
and bottom-right panels correspond to t/τ = 0.01, 0.02, 0.04, 0.08, 0.16,
0.32, 0.64, and 1.28, respectively. It can be seen that the hammer blow gen-
erates a displacement wave that initially develops at the free end of the rod
(x/l = 0), which is the end that is struck, propagates along the rod at the
velocity c, and reflects off the fixed end (x/l = 1) at time t/τ = 1 with no
phase shift.
where p is the pressure, V the volume, and γ the ratio of specific heats (i.e.,
the ratio of the gas’s specific heat at constant pressure to its specific heat at
constant volume). This ratio is approximately 1.4 for ordinary air.
Consider a sound wave in a column of gas of cross-sectional area A. Let
x measure distance along the column. Suppose that the wave generates an
x-directed displacement of the column, ψ(x, t). Consider a small section of
the column lying between x − δx/2 and x + δx/2. The change in volume of
92 OSCILLATIONS AND WAVES
the section is δV = A [ψ(x + δx/2, t) − ψ(x − δx/2, t)]. Hence, the relative
change in volume, which is assumed to be small, is
δV(x, t) ∂ψ(x, t)
= . (6.32)
V ∂x
The pressure perturbation δp(x, t) associated with the volume perturbation
δV(x, t) follows from (6.30), which yields
or
(1 + δp/p) (1 + δV/V)γ ≃ 1 + δp/p + γ δV/V = 1, (6.34)
giving
δV ∂ψ
δp = −γ p = −γ p , (6.35)
V ∂x
where use has been made of (6.32).
Consider a section of the gas column lying between x − δx/2 and x +
δx/2. The mass of this section is ρ A δx. The x-directed force acting on
its left boundary is A [p + δp(x − δx/2, t)], whereas the x-directed force
acting on its right boundary is −A [p + δp(x + δx/2, t)]. Finally, the average
longitudinal (i.e., x-directed) acceleration of the section is ∂2ψ(x, t)/∂t2.
Thus, the section’s longitudinal equation of motion is written
∂2ψ(x, t)
ρ A δx = −A [δp(x + δx/2, t) − δp(x − δx/2, t)] . (6.36)
∂t2
In the limit δx → 0, this equation reduces to
∂2ψ(x, t) ∂δp(x, t)
ρ 2
=− . (6.37)
∂t ∂x
Finally, (6.35) yields
∂2ψ 2
2∂ ψ
= c , (6.38)
∂t2 ∂x2
p
where c = γ p/ρ is a constant with the dimensions of velocity, which turns
out to be the sound speed in the gas (see Section 7.1).
Longitudinal Standing Waves 93
f = f1 f = 3 f1 f = 5 f1
Figure 6.5: First three normal modes of an organ pipe.
where A > 0, k > 0, ω > 0, and φ are constants. This expression auto-
matically satisfies the boundary condition ψ(0, t) = 0. The other boundary
condition is satisfied provided
cos(k l) = 0, (6.40)
which yields
k l = (n − 1/2) π, (6.41)
where the mode number n is a positive integer. Equations (6.38) and (6.39)
yield the dispersion relation
ω = k c. (6.42)
94 OSCILLATIONS AND WAVES
with y(t + τ) = y(t) for all t. This waveform rises linearly from an initial
value −A at t = 0 to a final value +A at t = τ, discontinuously jumps
back to its initial value, and then repeats ad infinitum. According to Equa-
tions (6.51) and (6.52), the Fourier harmonics of the waveform are
Zτ Z 2π
2 A
Cn = A (2 t/τ − 1) cos(n ω t) dt = 2 (θ − π) cos(n θ) dθ,
τ 0 π 0
(6.54)
Zτ Z 2π
2 A
Sn = A (2 t/τ − 1) sin(n ω t) dt = 2 (θ − π) sin(n θ) dθ,
τ 0 π 0
(6.55)
Cn = 0, (6.56)
2A
Sn = − . (6.57)
nπ
Hence, the Fourier reconstruction of the waveform is written
2 A X sin(n 2π t/τ)
y(t) = − . (6.58)
π n
n=1,∞
Given that the Fourier coefficients fall off like 1/n, as n increases, it seems
plausible that the above series can be truncated, after a finite number of
terms, without unduly affecting the reconstructed waveform. Figure 6.6
shows the result of truncating the series after 4, 8, 16, and 32 terms (these
cases correspond the top-left, top-right, bottom-left, and bottom-right pan-
els, respectively). It can be seen that the reconstruction becomes increas-
ingly accurate as the number of terms retained in the series increases. The
annoying oscillations in the reconstructed waveform at t = 0, τ, and 2τ are
known as Gibbs phenomena, and are the inevitable consequence of trying
to represent a discontinuous waveform as a Fourier series. In fact, it can be
demonstrated mathematically that, no matter how many terms are retained
in the series, the Gibbs phenomena never entirely go away.
We can slightly generalize the Fourier series (6.47) by including an n = 0
term. In other words,
X
Cn′ cos(n ′ ω t) + Sn′ sin(n ′ ω t) ,
y(t) = C0 + (6.59)
n′ =1,∞
98 OSCILLATIONS AND WAVES
and that Equations (6.51) and (6.52) still hold for n > 0.
Longitudinal Standing Waves 99
where y(t + τ) = y(t) for all t. This waveform rises linearly from zero at
t = 0, reaches a peak value A at t = τ/2, falls linearly, becomes zero again
at t = τ, and repeats ad infinitum. Moreover, the waveform clearly has a
non-zero average. It is easily demonstrated, from Equations (6.51), (6.52),
(6.62), and (6.63), that
A
C0 = , (6.64)
2
and
sin2(n π/2)
Cn = −A (6.65)
(n π/2)2
for n > 1, with Sn = 0 for n > 1. Note that only the odd-n Fourier har-
monics are non-zero. Figure 6.7 shows a Fourier reconstruction of the “tent”
waveform using the first 1, 2, 4, and 8 terms (in addition to the C0 term) in
the Fourier series (these cases correspond to the top-left, top-right, bottom-
left, and bottom-right panels, respectively). It can be seen that the recon-
struction becomes increasingly accurate as the number of terms in the series
increases. Moreover, in this example, there is no sign of Gibbs phenomena,
since the tent waveform is completely continuous.
Now, in our first example—i.e., the sawtooth waveform—all of the Sn
Fourier coefficients are zero, whereas in our second example—i.e., the tent
waveform—all of the Cn coefficients are zero. It is easily demonstrated that
this occurs because the sawtooth waveform is odd in t—i.e., y(−t) = −y(t)
for all t—whereas the tent waveform is even—i.e., y(−t) = y(t) for all t.
In fact, it is a general rule that waveforms which are even in t only have
cosines in their Fourier series, whereas waveforms which are odd only have
sines. Of course, waveforms which are neither even nor odd in t have both
cosines and sines in their Fourier series.
Fourier series arise quite naturally in the theory of standing waves, since
the normal modes of oscillation of any uniform continuous system possess-
ing linear equations of motion (e.g., a uniform string, an elastic solid, an
ideal gas) take the form of spatial cosine and sine waves whose wavelengths
are rational fractions of one another. Thus, the instantaneous spatial wave-
form of such a system can always be represented as a linear superposition of
cosine and sine waves: i.e., a Fourier series in space, rather than in time. In
fact, we can easily appreciate that the process of determining the amplitudes
100 OSCILLATIONS AND WAVES
and phases of the normal modes of oscillation from the initial conditions is
essentially equivalent to Fourier analyzing the initial conditions in space—
see Sections 5.3 and 6.2.
6.5 Exercises
1. Estimate the highest possible frequency (in Hertz), and the smallest possible
wavelength, of a sound wave in aluminium, due to the discrete atomic struc-
ture of this material. The mass density, Young’s modulus, and atomic weight
of aluminium are 2.7 × 103 kg m−3 , 6 × 1010 N m−2 , and 27, respectively.
2. Consider a linear array of N identical simple pendulums of mass m and
length l which are suspended from equal height points that are evenly spaced
a distance a apart. Suppose that each pendulum bob is attached to its two
immediate neighbors by means of light springs of unstretched length a and
spring constant K. The figure shows a small part of such an array. Let xi = i a
be the equilibrium position of the ith bob, for i = 1, N, and let ψi (t) be its
horizontal displacement. It is assumed that |ψi |/a ≪ 1 for all i. Demonstrate
that the equation of motion of the ith pendulum bob is
g K
ψ̈i = − ψi + (ψi−1 − 2 ψi + ψi+1 ).
l m
Consider a general normal mode of the form
ψi (t) = [A sin(k xi ) + B cos(k xi )] cos(ω t − φ).
Show that the associated dispersion relation is
g 4K
ω2 = + sin2 (k a/2).
l m
Suppose that the first and last pendulums in the array are attached to immov-
able walls, located a horizontal distance a away, by means of light springs of
xi x xi+1
K m
ψi ψi+1
a
Longitudinal Standing Waves 101
unstretched length a and spring constant K. Find the normal modes of the
system. Suppose, on the other hand, that the first and last pendulums are
not attached to anything on their outer sides. Find the normal modes of the
system.
3. Find the system of coupled inductors and capacitors which is analogous to
the system of coupled pendulums considered in the previous exercise, in
the sense that the time evolution equation for the current flowing through
the ith inductor has the same form as the equation of motion of the ith
pendulum. Consider both types of boundary condition discussed above. Find
the dispersion relation.
4. Consider a periodic waveform y(t) of period τ, where y(t + τ) = y(t) for all
t, which is represented as a Fourier series:
X
y(t) = C0 + [Cn cos(n ω t) + Sn sin(n ω t)] ,
n>1
where
where y(t + T ) = y(t) for all t, and τ < T , has the Fourier representation
τ 2A X sin(n π τ/T )
y(t) = A + (−1)n cos(n 2π t/T )
T π n
n=1,∞
Demonstrate that if τ ≪ T then the most significant terms in the above series
have frequencies (in Hertz) which range from the fundamental frequency
1/T to a frequency of order 1/τ.
Traveling Waves 103
7 Traveling Waves
We saw earlier, in Sections 5.2, 6.2, and 6.3, that a small amplitude trans-
verse wave on a uniform string, and a small amplitude longitudinal wave in
an elastic solid or an ideal gas, are all governed by the wave equation, which
(in one dimension) takes the general form
∂2ψ 2
2∂ ψ
= c , (7.1)
∂t2 ∂x2
where ψ(x, t) represents the wave disturbance, and c > 0 is a constant, with
the dimensions of velocity, which is a property of the particular medium
that supports the wave. Up to now, we have only considered media of finite
length: e.g., media which extend from x = 0 to x = l. Generally speaking,
we have encountered two distinct types of physical constraint which hold at
the boundaries of such media. Firstly, if a given boundary is fixed then the
wave displacement is constrained to be zero there: e.g., if the left bound-
ary is fixed then ψ(0, t) = 0. Secondly, if a given boundary is free then the
spatial derivative of the displacement (which usually corresponds to some
sort of force) is constrained to be zero there: e.g., if the right boundary is
free then ∂ψ(l, t)/∂x = 0. It follows that a fixed boundary corresponds to a
node—i.e., a point at which the amplitude of the wave disturbance is always
zero—whereas a free boundary corresponds to an anti-node—i.e., a point
at which the amplitude of the wave disturbance is always locally maximal.
Consequently, the nodes and the anti-nodes of a wave, of definite wave-
length, supported in a medium of finite length with stationary boundaries,
which can be either fixed or free, are constrained to be stationary. The only
simple solution of the wave equation (7.1) which has stationary nodes and
anti-nodes is a standing wave of the general form
ψ(x, t) = [A cos(k x) + B sin(k x)] cos(ω t − φ). (7.2)
The associated nodes are located at the values of x that satisfy
A cos(k x) + B sin(k x) = 0, (7.3)
which implies that they are indeed stationary, and also evenly spaced a dis-
tance λ/2 apart, where λ = 2π/k is the wavelength. Moreover, the anti-
nodes are situated halfway between the nodes. For example, suppose that
104 OSCILLATIONS AND WAVES
both boundaries of the medium are fixed boundaries. It follows that the
points x = 0 and x = l must each correspond to a node. This is only possible
if the length of the medium, l, is a half-integer number of wavelengths: i.e.,
l = n λ/2, where n is a positive integer. We conclude that, in this case, the
possible wavenumbers of standing wave solutions to the wave equation are
quantized such that
k l = n π. (7.4)
Moreover, the same is true if both boundaries are free boundaries. Fi-
nally, if one boundary is free, and the other fixed, then the quantization
of wavenumbers takes the slightly different form
k l = (n − 1/2) π. (7.5)
Now, those standing wave solutions that satisfy the appropriate quantiza-
tion criterion are known as the normal modes of the system. Moreover,
substitution of (7.2) into the wave equation (7.1) yields the standing wave
dispersion relation
ω2 = k2 c2. (7.6)
Thus, the fact that the normal mode wavenumbers are quantized immedi-
ately implies that the associated oscillation frequencies are also quantized.
Finally, since the wave equation is linear, the most general solution which
satisfies the boundary conditions is a linear superposition of all of the nor-
mal modes. Such a solution has the appropriate node or anti-node at each
of the boundaries, but does not necessarily have any stationary nodes or
anti-nodes in the interior of the medium.
phase angle φ. Note, in particular, that ψ(x+λ, t) = ψ(x, t) and ψ(x, t+T ) =
ψ(x, t) for all x and t: i.e., the wave is periodic in space with period λ, and
periodic in time with period T . Now, a wave maximum corresponds to a
point at which cos(k x − ω t − φ) = 1. It follows, from the well known
properties of the cosine function, that the various wave maxima are located
at
k x − ω t − φ = n 2π, (7.8)
where n is an integer. Thus, differentiating the above expression with re-
spect to t, and rearranging, the equation of motion of a particular maximum
becomes
dx ω
= . (7.9)
dt k
We conclude that the wave maximum in question propagates along the x-
axis at the velocity
ω
vp = . (7.10)
k
It is easily demonstrated that the other wave maxima, as well as the wave
minima and the wave zeros, also propagate along the x-axis at the same
velocity. In fact, the whole wave pattern propagates in the positive x-
direction without changing shape. The characteristic propagation velocity
vp is known as the phase velocity of the wave, since it is the velocity with
which points of constant phase in the wave disturbance (i.e., points which
satisfy k x − ω t − φ = constant) move. For obvious reasons, the type of
wave solution given in (7.7) is called a traveling wave.
Substitution of (7.7) into the wave equation (7.1) yields the familiar
dispersion relation
ω2 = k2 c2. (7.11)
We immediately conclude that the traveling wave solution (7.7) satisfies the
wave equation provided
ω
vp = =c: (7.12)
k
i.e., provided that the phase velocity of the wave takes the fixed value c.
In other words, the constant c2, which appears in the wave equation (7.1),
can be interpreted as the square of the velocity with which traveling waves
propagate through the medium in question. It follows, from the discussion
in Sections 5.2, 6.2, and 6.3, that transverse waves propagate along
p strings
of tension T and mass per unit length ρ at the phase velocity T/ρ, that
longitudinal sound waves propagate through elastic
p media of Young’s mod-
ulus Y and mass density ρ at the phase velocity Y/ρ, and that sound waves
106 OSCILLATIONS AND WAVES
Table 7.1: Calculated versus measured sound velocities in various solid materi-
als. [From Vibrations and Waves, A.P. French (W.W. Norton & Co., New York
NY, 1971).]
However, since traveling waves in infinite media are not subject to boundary
conditions, it follows that there is no restriction on the possible wavenum-
bers, or wavelengths, of such waves. Hence, any traveling wave solution
whose wavenumber, k, and angular frequency, ω, are related according to
ω = kc (7.15)
[which is just the square root of the dispersion relation (7.11)] is a valid
solution of the wave equation. Another way of putting this is that any trav-
eling wave solution whose wavelength, λ = 2π/k, and frequency, f = ω/2π,
are related according to
c = fλ (7.16)
is a valid solution of the wave equation. We, thus, conclude that high fre-
quency traveling waves propagating through a given medium possess short
wavelengths, and vice versa.
Consider the alternative wave solution
dx ω
=− . (7.19)
dt k
Clearly, (7.17) represents a traveling wave that propagates in the minus x-
direction at the phase velocity vp = ω/k. Moreover, substitution of (7.17)
into the wave equation (7.1) again yields the dispersion relation (7.15),
which implies that vp = c. It follows that traveling wave solutions to the
wave equation (7.1) can propagate in either the positive or the negative
x-direction, as long as they always move at the fixed speed c.
But, what is the relationship between traveling wave and standing wave so-
lutions to the wave equation (7.1) in an infinite medium? To help answer
108 OSCILLATIONS AND WAVES
Since the wave equation (7.1) is linear, it follows that the above superpo-
sition is a solution provided the two component waves are also solutions:
i.e., provided that ω = k c, which we shall assume to be the case. How-
ever, making use of the trigonometric identity cos a + cos b ≡ 2 cos[(a +
b)/2] cos[(a − b)/2], the above expression can also be written
ψ(x, t) = (A1 + A2) cos(k x) cos(ω t) + (A1 − A2) sin(k x) sin(ω t). (7.23)
∂2y 2
2∂ y
= v , (7.24)
∂t2 ∂x2
p
where v = T/ρ is the phase velocity of traveling waves on the string.
Consider a section of the string lying between x = x1 and x = x2. The
kinetic energy of this section is
Z x2
1 ∂y 2
K= ρ dx, (7.25)
x1 2 ∂t
since ∂y/∂t is the string’s transverse velocity. The potential energy is simply
the work done in stretching the section, which is T ∆s, where ∆s is the
difference between the section’s stretched and unstretched lengths. Here,
it is assumed that the tension remains approximately constant when the
section is slowly stretched. Now, an element of length of the string is
" 2#1/2
dy
2 2 1/2
ds = (dx + dy ) = 1+ dx. (7.26)
dx
Hence,
Z x2 " #1/2 Z x2 2
dy 2 1 ∂y
∆s = 1+ − 1 dx ≃ dx, (7.27)
x1 dx x1 2 ∂x
∂y ∂2y ∂y ∂2y
ρ = T , (7.30)
∂t ∂t2 ∂t ∂x2
since v2 = T/ρ. This expression yields
or
∂E ∂I
+ = 0, (7.33)
∂t ∂x
where " 2 2#
1 ∂y ∂y
E(x, t) = ρ +T (7.34)
2 ∂t ∂x
is the energy per unit length of the string, and
∂y ∂y
I(x, t) = −T . (7.35)
∂t ∂x
Finally, integrating (7.33) in x from x1 to x2, we obtain
Z
d x2
E dx + I(x2, t) − I(x1, t) = 0, (7.36)
dt x1
or
dE
= I(x1, t) − I(x2, t). (7.37)
dt
Here, E(t) is the energy stored in the section of the string lying between
x = x1 and x = x2 [see Equation (7.29)]. Now, if we interpret I(x, t) as the
instantaneous energy flux (i.e., rate of energy flow) in the positive-x direc-
tion, at position x and time t, then the above equation can be recognized
as a simple declaration of energy conservation. Basically, the equation states
that the rate of increase in the energy stored in the section of the string lying
between x = x1 and x = x2, which is dE/dt, is equal to the difference be-
tween the rate at which energy flows into the left end of the section, which is
I(x1, t), and the rate at which it flows out of the right end, which is I(x2, t).
Traveling Waves 111
Note that the string must conserve energy, since it lacks any mechanism for
energy dissipation. The same is true of the other wave media discussed in
this section.
Consider a wave propagating in the positive x-direction of the form
According to Equation (7.35), the energy flux associated with this wave is
where p
Z= ρY (7.51)
is the impedance of the solid. The units of Z are pressure over velocity, so, in
this case, the impedance measures the typical pressure in the solid required
to produce a unit longitudinal velocity. Analogous arguments to the above
reveal that the impedance of an ideal gas of density ρ, pressure p, and ratio
of specific heats γ, is (see Section 6.3)
√
Z = ρ γ p. (7.52)
√
velocity is v = 1/ L C. Multiplying (7.53) by C V, (7.54) by L I, and then
adding the two resulting expressions, we obtain the energy conservation
equation
∂E ∂I
+ = 0, (7.56)
∂t ∂x
where
1 1
E = L I2 + C V 2 (7.57)
2 2
is the electromagnetic energy per unit length of the line, and
I = IV (7.58)
is the electromagnetic energy flux along the line (i.e., the energy per unit
time which passes a given point) in the positive x-direction. Consider a sig-
nal propagating along the line, in the positive x-direction, whose associated
current takes the form
where
V0 = I0 Z. (7.61)
Here, s
L
Z= (7.62)
C
is the characteristic impedance of the line, and is measured in Ohms. It
follows that the mean energy flux associated with the signal is written
1 1 1 V02
hIi = hI Vi = I0 V0 = Z I02 = . (7.63)
2 2 2 Z
Likewise, for a signal propagating along the line in the negative x-direction,
1 1 1 V02
hIi = − I0 V0 = − Z I02 = − . (7.66)
2 2 2 Z
Traveling Waves 115
Consider two uniform semi-infinite strings which run along the x-axis, and
are tied together at x = 0. Let the first string be of density per unit length
ρ1, and occupy the region x < 0, and let the second string be of density per
unit length ρ2, and occupy the region x > 0. Now, the tensions in the two
strings must be equal, otherwise the string interface would not be in force
balance in the x-direction. So, let T be the common tension. Suppose that a
transverse wave of angular frequency ω is launched from a wave source at
x = −∞, and propagates towards the interface. Assuming that ρ1 6= ρ2, we
116 OSCILLATIONS AND WAVES
would expect the wave incident on the interface to be partially reflected, and
partially transmitted. Of course, the frequencies of the incident, reflected,
and transmitted waves are all the same, since this property of the waves
is ultimately determined by the oscillation frequency of the wave source.
Hence, in the region x < 0, the wave displacement takes the form
or
The only way in which the above equation can be satisfied for all values of
t is if φi = −φr = φt. This being the case, the common cos(ω t + φi) factor
cancels out, and we are left with
Ai + Ar = At. (7.76)
Secondly, since the two strings lack an energy dissipation mechanism, the
energy flux into the interface must match that out of the interface. In other
words,
1 2 1
ω Z1 (Ai2 − Ar2) = ω 2 Z2 At2, (7.77)
2 2
Traveling Waves 117
√ √
where Z1 = ρ1 T and Z2 = ρ2 T are the impedances of the first and
second strings, respectively. The above expression reduces to
where use has been made of the trigonometric identity cos a − cos b ≡
2 sin[(a + b)/2] sin[(b − a)/2]. We conclude that the incident and reflected
waves interfere in such a manner as to produce a standing wave with a node
at the fixed boundary.
Suppose that the density
√ per unit length of the second string, ρ2, tends
to zero, so that Z2 = ρ2 T → 0. It follows from (7.80) and (7.81) that
Ar = Ai and At = 2 Ai. Likewise, (7.82) and (7.83) yield R = 1 and T = 0.
Hence, the interface between the two strings oscillates at twice the ampli-
tude of the incident wave (i.e., the interface is a point of maximal amplitude
oscillation), and there is no transmitted energy. In other words, the second
string acts exactly like a free boundary. It follows that when a transverse
wave on a string is incident on a free boundary then it is perfectly reflected
with no phase shift: i.e., Ar = Ai. Thus, the resultant wave displacement
on the string becomes
where use has been made of the trigonometric identity cos a + cos b ≡
2 cos[(a + b)/2] cos[(a − b)/2]. We conclude that the incident and reflected
waves interfere in such a manner as to produce a standing wave with an
anti-node at the free boundary.
Suppose that two strings of mass per unit length ρ1 and ρ2 are separated
by a short section of string of mass per unit length ρ3. Let all three strings
have the common tension T . Suppose that the first and second strings oc-
cupy the regions x < 0 and x > a, respectively. Thus, the middle string
occupies the region 0 ≤ x ≤ a. Moreover, the interface between the first
and middle strings is at x = 0, and the interface between the middle and
second strings is at x = a. Suppose that a wave of angular frequency ω is
launched from a wave source at x = −∞, and propagates towards the two
interfaces. We would expect this wave to be partially reflected and partially
transmitted at the first interface (x = 0), and the resulting transmitted wave
to then be partially reflected and partially transmitted at the second inter-
face (x = a). Thus, we can write the wave displacement in the region x < 0
as
y(x, t) = Ai cos(k1 x − ω t) + Ar cos(k1 x + ω t), (7.87)
where Ai is the amplitude of the
p incident wave, Ar is the amplitude of the
reflected wave, and k1 = ω/ T/ρ1. Here, the phase angles of the two
waves have been chosen so as to facilitate the matching process at x = 0.
Traveling Waves 119
where A+ and A− are the amplitudes of the rightp and left moving waves
on the middle string, respectively, and k3 = ω/ T/ρ3. Continuity of the
transverse displacement at x = 0 yields
Ai + Ar = A+ + A−, (7.90)
where a common factor cos(ω t) has cancelled out. Continuity of the energy
flux at x = 0 gives
Suppose that the length of the middle string is one quarter of a wavelength:
i.e., k3 a = π/2. Furthermore, let φt = k2 a + π/2. It follows that cos(k3 a −
ω t) = sin(ω t), cos(k3 a + ω t) = − sin(ω t), and cos(k2 a − ω t − φt) =
sin(ω t). Thus, canceling out a common factor sin(ω t), the above expres-
sion yields
A+ − A− = At. (7.94)
Continuity of the energy flux at x = a gives
2 2
Z3 (A+ − A− ) = Z2 At2. (7.95)
Now, suppose that the impedance of the middle string is √the geometric mean
of the impedances of the two outer strings: i.e., Z3 = Z1 Z2. In this case,
it is clear, from the above two equations, that R = 0 and T = 1. In other
words, there is no reflection of the incident wave, and all of the incident
energy ends up being transmitted across the middle string from the leftmost
to the rightmost string. Thus, if we wish to transmit transverse wave energy
from a string of impedance Z1 to a string of impedance Z2 (where Z2 6= Z1)
in the most efficient manner possible—i.e, with no reflection of the incident
energy flux—then we can do this by connecting the two strings via a short
section of string
√ whose length is one quarter of a wavelength, and whose
impedance is Z1 Z2. This procedure is known as impedance matching.
It should be reasonably clear that the above analysis of the reflection and
transmission of transverse waves at a boundary between two strings is also
applicable to the reflection and transmission of other types of wave incident
on a boundary between two media of differing impedances. For example,
consider a transmission line, such as a co-axial cable. Suppose that the line
occupies the region x < 0, and is terminated (at x = 0) by a load resistor of
resistance RL. Such a resistor might represent a radio antenna (which acts
just like a resistor in an electrical circuit, except that the dissipated energy
Traveling Waves 121
is radiated, rather than being converted into heat energy). Suppose that a
signal of angular frequency ω is sent down the line from a wave source at
x = −∞. The current and voltage on the line can be written
Z − RL
Ir = Ii. (7.106)
Z + RL
Hence, the coefficient of reflection, which is the ratio of the power reflected
by the load to the power sent down the line, is
2 2
Ir Z − RL
R= = . (7.107)
Ii Z + RL
4 Z RL
T =1−R= . (7.108)
(Z + RL)2
It can be seen, by comparison with Equations (7.82) and (7.83), that the
load terminating the line acts just like another transmission line of imped-
ance RL. Moreover, it is clear that power can only be efficiently sent down a
transmission line, and transferred to a terminating load, when the impedan-
ce of the line matches the effective impedance of the load (which, in this
case, is the same as the resistance of the load). In other words, when Z = RL
there is no reflection of the signal sent down the line (i.e., R = 0), and all
of the signal energy is therefore absorbed by the load (i.e., T = 1). As
an example, a half-wave antenna (i.e., an antenna whose length is half the
wavelength of the emitted radiation) has a characteristic impedance of 73 Ω.
Hence, a transmission line used to feed energy into such an antenna should
122 OSCILLATIONS AND WAVES
It can be seen that the above equations are just like the corresponding vac-
uum equations, (7.109) and (7.110), except that ǫ0 has been replaced by
ǫ ǫ0. It immediately follows that the phase velocity of an electromagnetic
wave propagating through a dielectric medium is
1 c
v= √ = , (7.129)
ǫ ǫ0 µ0 n
√
where c = 1/ ǫ0 µ0 is the velocity of light in vacuum, and the quantity
√
n= ǫ (7.130)
where Ei is the amplitude of (the electric component of) the incident wave,
Er the amplitude of the reflected wave, k1 = n1 ω/c, and Z1 = Z0/n1. The
wave electric and magnetic fields in the region z > 0 take the form
Ei + Er = Et, (7.136)
π radian phase shift of the reflected wave, relative to the incident wave, on
reflection from a boundary with a medium of greater refractive index. Con-
versely, there is no phase shift on reflection from a boundary with a medium
of lesser refractive index.
Note that Equations (7.138)–(7.141) are analogous to Equations (7.80)–
(7.83), with refractive index playing the role of impedance. This suggests,
by analogy with earlier analysis, that we can prevent reflection of an elec-
tromagnetic wave normally incident at a boundary between two transparent
dielectric media of different refractive indices by separating the media by a
thin transparent layer whose thickness is one quarter of a wavelength, and
whose refractive index is the geometric mean of the refractive indices of
the two media. This is the physical principle behind the non-reflective lens
coatings used in high-quality optical instruments.
7.8 Exercises
and voltage oscillate π/2 radians out of phase everywhere along the line.
Demonstrate that there is zero net flux of electromagnetic energy along the
line.
4. Suppose that the transmission line in the previous exercise is short circuited,
such that the line is effectively terminated by a negligible resistance. Find the
relationship between Ir and Ii . Show that the current and voltage oscillate
π/2 radians out of phase everywhere along the line. Demonstrate that there
is zero net flux of electromagnetic energy along the line.
5. Suppose that the transmission line of Exercise 3 is terminated by an inductor
of inductance L, such that
∂I(0, t)
V(0, t) = L .
∂t
Find the relationship between Ir and Ii . Obtain expressions for the cur-
rent, I(x, t), and the voltage, V(x, t), along the line (which only involve Ii ).
Demonstrate that the incident and the reflected wave both have zero net
associated energy flux.
6. Suppose that the transmission line of Exercise 3 is terminated by a capacitor
of capacitance C. Find the relationship between Ir and Ii . Obtain expressions
for the current, I(x, t), and the voltage, V(x, t), along the line (which only
involve Ii ). Demonstrate that the incident and the reflected wave both have
zero net associated energy flux.
7. A lossy transmission line has a resistance per unit length R, in addition to
an inductance per unit length L, and a capacitance per unit length C. The
resistance can be considered to be in series with the inductance. Demonstrate
that the Telegrapher’s equations generalize to
∂V 1 ∂I
= − ,
∂t C ∂x
∂I R 1 ∂V
= − I− ,
∂t L L ∂x
where I(x, t) and V(x, t) are the voltage and current along the line. Derive
an energy conservation equation of the form
∂E ∂I
+ = −R I2 ,
∂t ∂x
where E is the energy per unit length along the line, and I the energy flux.
Give expressions for E and I. What does the right-hand side of the above
equation represent? Show that the current obeys the wave-diffusion equation
∂2 I R ∂I 1 ∂2 I
2
+ = .
∂t L ∂t L C ∂x2
Traveling Waves 129
where λ is the wavelength of the incident light in air. Assume that n = 1.5,
and that this value remains approximately constant for light whose wave-
lengths lie in the visible band. Suppose that λ0 = 550 nm, which corresponds
to green light. It follows that R = 0 for green light. What is R for blue light of
130 OSCILLATIONS AND WAVES
wavelength λ = 450 nm, and for red light of wavelength 650 nm? Comment
on how effective the coating is at suppressing unwanted reflection of visible
light incident on the lens.
12. A glass lens is coated with a non-reflective coating whose thickness is one
quarter of a wavelength (in the coating) of light whose frequency is f0 .
Demonstrate that the coating also suppresses reflection from light whose fre-
quency is 3 f0 , 5 f0 , etc., assuming that the refractive index of the coating and
the glass is frequency independent.
13. An plane electromagnetic wave, polarized in the x-direction, and propagating
in the z-direction though a conducting medium of conductivity σ is governed
by
∂Ex σ 1 ∂Hy
= − Ex − ,
∂t ǫ0 ǫ0 ∂z
∂Hy 1 ∂Ex
= − ,
∂t µ0 ∂z
where Ex (z, t) and Hy (z, t) are the electric and magnetic components of the
wave. Derive an energy conservation equation of the form
∂E ∂I
+ = −σ Ex2 ,
∂t ∂x
where E is the electromagnetic energy per unit volume, and I the electro-
magnetic energy flux. Give expressions for E and I. What does the right-
hand side of the above equation represent? Demonstrate that Ex obeys the
wave-diffusion equation
∂2 E x σ ∂Ex 2
2 ∂ Ex
+ = c ,
∂t2 ǫ0 ∂t ∂z2
√
where c = 1/ ǫ0 µ0 . In the high frequency, low conductivity limit ω ≫
σ/ǫ0 , show that the above equation has the approximate solution
8 Wave Pulses
for all x. Recall, from Section 6.4, that we can represent such a function as
a Fourier series: i.e.,
X
F(x) = [Cn cos(n δk x) + Sn sin(n δk x)] , (8.2)
n=1,∞
where
2π
δk =. (8.3)
L
[Note that we have neglected the n = 0 term in (8.2), for the sake of con-
venience.] The expression (8.2) automatically satisfies the periodicity con-
straint (8.1), since cos(θ + n 2π) = cos θ and sin(θ + n 2π) = sin θ for all θ
and n (with the proviso that n is integer). The so-called Fourier coefficients,
Cn and Sn, appearing in (8.2), can be determined from the function F(x) by
means of the following easily demonstrated results (see Exercise 1):
Z L/2
2
cos(n δk x) cos(n ′ δk x) dx = δn,n′ , (8.4)
L −L/2
Z L/2
2
sin(n δk x) sin(n ′ δk x) dx = δn,n′ , (8.5)
L −L/2
Z L/2
2
cos(n δk x) sin(n ′ δk x) dx = 0, (8.6)
L −L/2
Incidentally, it is clear, from the above equations, that C(−k) = C(k) and
S(−k) = −S(k). The Fourier space (i.e., k-space) functions C(k) and S(k)
are known as the cosine Fourier transform and the sine Fourier transform of
the real space (i.e., x-space) function F(x), respectively. Furthermore, since
we already know that any periodic function can be represented as a Fourier
series, it seems plausible that any aperiodic function can be represented as
a Fourier transform. This is indeed the case. Note that Equations (8.10)–
(8.12) effectively enable us to represent a general function as a linear super-
position of sine and cosine functions. Let us now consider some examples.
Wave Pulses 133
Now, given that cos(−k x) = cos(k x) and sin(−k x) = − sin(k x), it is ap-
parent, from (8.11) and (8.12), that if F(x) is even in x, so that F(−x) =
F(x), then S(k) = 0, and if F(x) is odd in x, so that F(−x) = −F(x), then
C(k) = 0. Hence, since the square-wave function (8.13) is clearly even in
x, its sine Fourier transform is automatically zero. On the other hand, its
cosine Fourier transform takes the form
Z
1 l/2 l sin(k l/2)
C(k) = cos(k x) dx = . (8.14)
2π −l/2 2π k l/2
Figure 8.1 shows the function F(x), together with its associated cosine trans-
form, C(k).
As a second example, consider the Gaussian function
!
x2
F(x) = exp − . (8.15)
2 σx2
zero. On the other hand, the cosine Fourier transform is easily shown to be
(see Exercise 2) !
1 k2
C(k) = exp − , (8.16)
(2π σk2)1/2 2 σk2
where
1
σk = . (8.17)
σx
Note that this function is a Gaussian in Fourier space of characteristic width
σk = 1/σx. In fact, the Gaussian is the only mathematical function which
is its own Fourier transform. Now, the original function F(x) can be recon-
structed from its Fourier transform using
Z∞
F(x) = C(k) cos(k x) dk. (8.18)
−∞
p √
the full width at half maximum of the function is ∆x = 2 π/2 σx = 2π σx.
Likewise, the full width√ at half maximum of the Fourier space Gaussian
function C(k) is ∆k = 2 π σk. Thus,
∆x ∆k = 2π, (8.19)
In other words, a Gaussian function in real space, of unit height and char-
acteristic width σx, has a cosine Fourier transform which is a Gaussian in
Fourier space, of characteristic width σk = 1/σx, whose integral over all
k-space is unity.
Consider what happens to the above mentioned real space Gaussian, and
its Fourier transform, in the limit σx → ∞, or, equivalently, σk → 0. There
is no difficulty in seeing, from Equation (8.15), that
F(x) → 1. (8.21)
In other words, the real space Gaussian morphs into a function which takes
the constant value unity everywhere. The Fourier transform is more prob-
lematic. In the limit σk → 0, Equation (8.16) yields a k-space function
which is zero everywhere apart from k = 0 (since the function is negli-
gible for |k| >∼ σk), where it is infinite [since the function takes the value
(2π σk)−1/2 at k = 0]. Moreover, according to Equation (8.20), the integral
of the function over all k remains unity. Thus, the Fourier transform of the
uniform function F(x) = 1 is a sort of integrable “spike” located at k = 0.
This strange function is known as the Dirac delta function, and is denoted
δ(k). Thus, one definition of a delta function is
!
1 k2
δ(k) ≡ lim exp − . (8.22)
σk →0 (2π σ 2)1/2 2 σk2
k
Z∞
1
cos(k x) sin(k ′ x) dx = 0. (8.31)
2π −∞
∂2ψ 2
2∂ ψ
= v , (8.32)
∂t2 ∂x2
where ψ(x, t) is the wave amplitude, and v the characteristic phase velocity.
We have seen a number of particular solutions of this equation. For instance,
ω = kv : (8.34)
i.e., provided that the wave propagates with the fixed phase velocity v. We
can also write the solution (8.33) as
Now, the wave equation (8.32) is linear. This suggests that its most
general solution can be written as a linear superposition of all of its valid
wavelike solutions. In the absence of specific boundary conditions, there is
no restriction on the possible wavenumbers of such solutions. Thus, it is
plausible that the most general solution of (8.32) can be written
Z∞
ψ(x, t) = C+(k) cos[k (x − v t)] dk
−∞
Z∞
+ S+(k) sin[k (x − v t)] dk
−∞
Z∞
+ C−(k) cos[k (x + v t)] dk
−∞
Z∞
+ S−(k) sin[k (x + v t)] dk : (8.37)
−∞
It follows that we can indeed uniquely determine the functions C+(k), C−(k),
S+(k), and S−(k), appearing in expression (8.37), for any ψ(x, 0) and ψ̇(x, 0).
This proves that (8.37) is the most general solution of the wave equation
(8.32).
140 OSCILLATIONS AND WAVES
where
Z∞
F(x) = [C+(k) cos(k x) + S+(k) sin(k x)] dk, (8.46)
−∞
Z∞
G(x) = [C−(k) cos(k x) + S−(k) sin(k x)] dk. (8.47)
−∞
What does the expression (8.45) signify? Well, F(x − v t) represents a wave
disturbance of arbitrary shape which propagates in the positive x-direction,
at the fixed speed v, without changing shape. This should be clear, since a
point with a given amplitude on the wave, F(x − v t) = c, has an equation
of motion x − v t = F−1(c) = constant, and thus propagates in the positive
x-direction at the velocity v. Moreover, since all points on the wave propa-
gate in the same direction at the same velocity it follows that the wave does
not change shape as it moves. Of course, G(x + v t) represents a wave dis-
turbance of arbitrary shape which propagates in the negative x-direction, at
the fixed speed v, without changing shape. Thus, we conclude that the most
general solution to the wave equation (8.32) is a superposition of two wave
disturbances of arbitrary shapes which propagate in opposite directions, at
the fixed speed v, without changing shape. Such solutions are generally
termed wave pulses. So, what is the relationship between a general wave
pulse and the sinusoidal traveling wave solutions to the wave equation that
we found previously. Well, as is clear from Equations (8.46) and (8.47),
a wave pulse can be thought of as a superposition of sinusoidal traveling
waves propagating in the same direction as the pulse. Moreover, the ampli-
tude of cosine waves of wavenumber k in this superposition is simply equal
to the cosine Fourier transform of the pulse shape evaluated at wavenumber
k. Likewise, the amplitude of sine waves of wavenumber k in the superpo-
sition is equal to the sine Fourier transform of the pulse shape evaluated at
wavenumber k.
For instance, suppose that we have a triangular wave pulse of the form
(see Figure 8.3)
1 − 2 |x|/l |x| ≤ l/2
F(x) = . (8.48)
0 |x| > l/2
Wave Pulses 141
The sine Fourier transform of this pulse shape is zero by symmetry. However,
the cosine Fourier transform is (see Exercise 7)
Z∞
1 l sin2(k l/4)
C(k) = F(x) cos(k x) dx = . (8.49)
2π −∞ 4π (k l/4)2
The functions F(x) and C(k) are shown in Figure 8.3. It follows that the
right-propagating triangular wave pulse
1 − 2 |x − v t|/l |x − v t| ≤ l/2
ψ(x, t) = (8.50)
0 |x − v t| > l/2
becomes Z∞
1 sin2(k l/4)
ψ(x, t) = cos[k (x + v t)] l dk. (8.53)
4π −∞ (k l/4)2
142 OSCILLATIONS AND WAVES
8.3 Bandwidth
where C(ω) and S(ω) are the temporal cosine and sine Fourier transforms
of the waveform, respectively. By analogy with Equations (8.10)–(8.12), we
can invert the above expression to give
Z
1 ∞
C(ω) = F(t) cos(ω t) dt, (8.55)
2π −∞
Z
1 ∞
S(ω) = F(t) sin(ω t) dt. (8.56)
2π −∞
These equations make it clear that C(−ω) = C(ω), and S(−ω) = −S(ω).
Moreover, it is apparent that if F(t) is an even function of t then S(ω) = 0,
but if it is an odd function then C(ω) = 0.
The current in the antenna of an amplitude-modulated (AM) radio trans-
mitter is driven by a voltage signal which oscillates sinusoidally at a fre-
quency, ω0, which is known as the carrier frequency. In fact, in commercial
(medium wave) AM radio, each station is assigned a single carrier frequency
which lies somewhere between about 500 kHz and 1600 kHz. However,
the voltage signal fed to the antenna does not have a constant amplitude.
Rather, it has a modulated amplitude which can be expressed, somewhat
schematically, as a Fourier series:
X
A(t) = A0 + An cos(ωn t − φn), (8.57)
n>0
which, with the help of some standard trigonometric identities, can also be
written
1X
ψ(t) = A0 cos(ω0 t) + An cos[(ω0 + ωn) t − φn)]
2
n>0
1X
+ An cos[(ω0 − ωn) t + φn)]
2
n>0
1X
= A0 cos(ω0 t) + An cos φn cos[(ω0 + ωn) t]
2
n>0
1X
+ An sin φn sin[(ω0 + ωn) t]
2
n>0
1X
+ An cos φn cos[(ω0 − ωn) t]
2
n>0
1X
− An sin φn sin[(ω0 − ωn) t]. (8.59)
2
n>0
We can calculate the cosine and sine Fourier transforms of the signal,
Z
1 ∞
C(ω) = ψ(t) cos(ω t) dt, (8.60)
2π −∞
Z
1 ∞
S(ω) = ψ(t) sin(ω t) dt, (8.61)
2π −∞
Z∞
1 1
sin(ω t) sin(ω ′ t) dt = δ(ω − ω ′ ) − δ(ω + ω ′ ) ,
2π −∞ 2
(8.63)
Z∞
1
cos(ω t) sin(ω ′ t) dt = 0. (8.64)
2π −∞
Here, we have only shown the positive frequency components of C(ω) and
S(ω), since we know that C(−ω) = C(ω) and S(−ω) = −S(ω).
The AM frequency spectrum specified in Equations (8.65) and (8.66)
is shown, somewhat schematically, in Figure 8.4. The spectrum consists
of a series of delta function spikes. The largest spike corresponds to the
carrier frequency, ω0. However, this spike carries no information. Indeed,
the signal information is carried in so-called sidebands which are equally
spaced on either side of the carrier frequency. The upper sidebands corre-
spond to the frequencies ω0 + ωn, whereas the lower sidebands correspond
to the frequencies ω0 − ωn. Thus, in order for an AM radio signal to carry
all of the information present in audible sound, for which the appropriate
modulation frequencies, ωn, range from about 0 Hz to about 20 kHz, the
signal would have to consist of a superposition of sinusoidal oscillations
with frequencies which range from the carrier frequency minus 20 kHz to
the carrier frequency plus 20 kHz. In other words, the signal would have to
occupy a range of frequencies from ω0 − ωN to ω0 + ωN, where ωN is the
largest modulation frequency. This is an important result. An AM radio sig-
nal which only consists of a single frequency, such as the carrier frequency,
transmits no information. Only a signal which occupies a finite range of fre-
quencies, centered on the carrier frequency, is capable of transmitting useful
information. The difference between the highest and the lowest frequency
components of an AM radio signal, which is twice the maximum modula-
tion frequency, is called the bandwidth of the signal. Thus, to transmit all
Wave Pulses 145
carrier frequency
C(ω)
ω→
S(ω)
ω→
Of course, we must have ω0 σt ≫ 1: i.e., the period of the carrier wave must
be much less than the duration of the bit. Figure 8.5 illustrates a digital bit
calculated for ω0 σt = 20.
The sine Fourier transform of the signal (8.67) is zero by symmetry.
However, its cosine Fourier transform takes the form
Z∞ !
1 t2
C(k) = exp − cos(ω0 t) cos(ω t) dt, (8.68)
2π −∞ 2 σt2
Wave Pulses 147
Z∞ !
1 t2
= exp − {cos[(ω − ω0) t] + cos[(ω + ω0) t]} dt.
4π −∞ 2 σt2
where
1
σω = . (8.70)
σt
In other words, the Fourier transform of the signal takes the form of a Gaus-
sian in ω-space, which is centered on the carrier frequency, ω0, and is of
characteristic width σω = 1/σt. Thus, the bandwidth of the signal is of or-
der σω. Note that the shorter the signal duration the higher the bandwidth.
This is a general
√ rule. A signal of full width at half maximum temporal du-
ration ∆t = 2π σt generally √ has a Fourier transform of full width at half
maximum bandwidth ∆ω = 2π σω, so that
∆ω ∆t ∼ 2π. (8.71)
of the screen, that includes many horizontal lines, has a flicker rate of 60
Hz.) Thus, the rate at which the instructions “turn on” and “turn off” must
be sent to the electron beam is 30 × 2.5 × 105 or 8 × 106 times a second.
The transmitted TV signal must therefore have about 107 on-off instruction
blips per second. If temporal overlap is to be avoided, each blip can be no
longer than ∆t ∼ 10−7 seconds in duration. Thus, the required bandwidth
is ∆f ∼ 1/∆t ∼ 107 Hz = 10 MHz. The carrier wave frequencies used for
conventional broadcast TV lie in the so-called VHF band, and range from
about 55 to 210 MHz. Our previous discussion of AM radio might lead us
to think that the 10 MHz bandwidth represents the combined extents of an
upper and a lower sideband of modulation frequencies. In practice, the car-
rier wave and one of the sidebands are suppressed: i.e., they are filtered
out, and never applied to the antenna. However, they are regenerated in
the receiver from the information contained in the single sideband which
is broadcast. This technique, which is called single sideband transmission,
halves the bandwidth requirement to about 5 MHz. Thus, between 55 and
210 MHz there is room for about 30 TV channels, each using a 5 MHz band-
width. (Actually, there are far fewer TV channels than this in the VHF band,
because part of this band is reserved for FM radio, air traffic control, air
navigation beacons, marine communications, etc.)
8.4 Exercises
1. Verify Equations (8.4)–(8.6). Derive Equations (8.7) and (8.8) from Equa-
tion (8.2) and Equations (8.4)–(8.6).
2. Suppose that
x2
F(x) = exp − .
2 σx2
Demonstrate that
Z∞
k2
1 ikx 1
F̄(k) ≡ F(x) e dx = q exp − ,
2π −∞ 2 π σk2 2 σk2
where i is the square-root of minus one, and σk = 1/σx . [Hint: You will
need to complete the square of the exponent of e, transform the variable of
R∞ 2
integration, and then make use of the standard result that −∞ e−y dy =
√
π.] Hence, show from de Moive’s theorem, exp( i k x) ≡ cos x + i sin x, that
Z
1 ∞ k2
1
C(k) ≡ F(x) cos(k x) dx = q exp − ,
2π −∞ 2 π σk2 2 σk2
Wave Pulses 149
Z∞
1
S(k) ≡ F(x) sin(k x) dx = 0.
2π −∞
3. Demonstrate that
Z∞
k2
1
exp − dk = 1.
2 σk2
q
−∞ 2 π σk2
9. Suppose that F(t) is zero, except in the interval from t = −∆t/2 to t = ∆t/2.
Suppose that in this interval F(t) makes exactly one sinusoidal oscillation at
the angular frequency ω0 = 2π/∆t, starting and ending with the value zero.
Find the above defined Fourier transforms C(ω) and S(ω).
10. Demonstrate that
Z∞ Z∞
F2 (t) dt = 2π [C2 (ω) + S2 (ω)] dω,
−∞ −∞
where the relation between F(t), C(ω), and S(ω) is defined above. This
result is known as Parseval’s theorem.
11. Suppose that F(t) and G(t) are both even functions of t with the cosine trans-
forms F̄(ω) and Ḡ(ω), so that
Z∞
F(t) = F̄(ω) cos(ω t) dω,
−∞
Z∞
G(t) = Ḡ(ω) cos(ω t) dω.
−∞
Let H(t) = F(t) G(t), and let H̄(ω) be the cosine transform of this even
function, so that Z∞
H(t) = H̄(ω) cos(ω t) dω.
−∞
150 OSCILLATIONS AND WAVES
Demonstrate that
Z∞
1
H̄(ω) = F̄(ω ′ ) Ḡ(ω ′ + ω) + Ḡ(ω ′ − ω) dω ′ .
2 −∞
This result is known as the convolution theorem, since the above type of in-
tegral is known as a convolution integral. Suppose that F(t) = cos(ω0 t).
Show that
1
H̄(ω) = Ḡ(ω − ω0 ) + Ḡ(ω + ω0 ) .
2
Dispersive Waves 151
9 Dispersive Waves
ω = ω(k). (9.2)
ω = k v, (9.3)
which is clearly the equation of a wave pulse that propagates in the positive
x-direction, at the fixed speed v, without changing shape (see Chapter 8).
The above analysis would seem to suggest that arbitrarily shaped wave
pulses generally propagate at the same speed as sinusoidal waves, and do so
without dispersing or, otherwise, changing shape. In fact, these statements
are only true of pulses made up of superpositions of sinusoidal waves with
linear dispersion relations. There are, however, many types of sinusoidal
152 OSCILLATIONS AND WAVES
wave whose dispersion relations are nonlinear. For instance, the disper-
sion relation of sinusoidal electromagnetic waves propagating through an
unmagnetized plasma is (see Section 9.2)
q
ω= k2 c2 + ωp2, (9.6)
where g is the acceleration due to gravity, T the surface tension, and ρ the
mass density. Sinusoidal waves which satisfy nonlinear dispersion relations,
such as (9.6) and (9.7), are known as dispersive waves, as opposed to waves
which satisfy linear dispersion relations, such as (9.3), which are known as
non-dispersive waves. As we saw above, a wave pulse made up of a linear
superposition of non-dispersive sinusoidal waves, all traveling in the same
direction, propagates at the common phase velocity of these waves, without
changing shape. But, how does a wave pulse made up of a linear superposi-
tion of dispersive sinusoidal waves evolve in time?
Suppose that
!
1 (k − k0)2
C(k) = q exp − : (9.8)
2π σk2 2 σk2
It follows that
Z∞ !
sin(k0 x − ω0 t) (k − k0)2
− exp − sin[(k − k0) (x − vg t)] dk,
2 σk2
q
2π σk2 −∞
(9.13)
where use has been made of a standard trigonometric identity. The inte-
gral involving sin[(k − k0) (x − vg t)] is zero, by symmetry. Moreover, an
examination of Equations (8.15)–(8.18) reveals that
Z∞ ! !
1 k2 x2
exp − cos(k x) dk = exp − , (9.14)
2 σk2 2 σx2
q
2π σk2 −∞
velocity (since, if ω = k v then ω/k = dω/dk = v). However, for the case
of dispersive waves, the two velocities are, in general, different.
Equation (9.15) indicates that, as the wave pulse propagates, its en-
velope remains the same shape. Actually, this result is misleading, and is
only obtained because of the neglect of second-order terms in the expan-
sion (9.11). If we keep more terms in this expansion then we can show
that the wave pulse does actually change shape as it propagates. How-
ever, this demonstration is most readily effected by means of the following
simple argument. The pulse extends in Fourier space from k0 − ∆k/2 to
k0 + ∆k/2, where ∆k ∼ σk. Thus, part of the pulse propagates at the velocity
vg(k0 − ∆k/2), and part at the velocity vg(k0 + ∆k/2). Consequently, the
pulse spreads out as it propagates, since some parts of it move faster than
others. Roughly speaking, the spatial extent of the pulse in real space grows
as
dvg(k0)
∆x ∼ (∆x)0 + [vg(k0 + ∆k/2) − vg(k0 − ∆k/2)] t ∼ (∆x)0 + ∆k t,
dk
(9.17)
where (∆x)0 ∼ σx = σ−1
k is the extent of the pulse at t = 0. Hence, from
(9.12),
d2ω(k0) t
∆x ∼ (∆x)0 + . (9.18)
dk2 (∆x)0
We, thus, conclude that the spatial extent of the pulse grows linearly in
time, at a rate proportional to the second derivative of the dispersion rela-
tion with respect to k (evaluated at the pulse’s central wavenumber). This
effect is known as pulse dispersion. In summary, a wave pulse made up of a
linear superposition of dispersive sinusoidal waves, with a range of different
wavenumbers, propagates at the group velocity, and also gradually disperses
as time progresses.
d2x
m = q Ex, (9.20)
dt2
where x measures its wave-induced displacement in the x-direction. The
above equation can easily be solved to give
q E0
x=− cos(k z − ω t). (9.21)
m ω2
Thus, the wave causes the particle to execute sympathetic simple harmonic
oscillations, in the x-direction, with an amplitude which is directly propor-
tional to its charge, and inversely proportional to its mass.
Suppose that the wave is actually propagating through an unmagne-
tized electrically neutral plasma consisting of free electrons, of mass me and
charge −e, and free ions, of mass mi and charge +e. Since the plasma is
assumed to be electrically neutral, each species must have the same equi-
librium number density, ne. Now, given that the electrons are much less
massive than the ions (i.e., me ≪ mi), but have the same charge (mod-
ulo a sign), it follows from (9.21) that the wave-induced oscillations of the
electrons are of much higher amplitude than those of the ions. In fact,
to a first approximation, we can say that the electrons oscillate whilst the
ions remain stationary. Assuming that the electrons and ions are evenly
distributed throughout the plasma, the wave-induced displacement of an
individual electron generates an effective electric dipole moment in the x-
direction of the form px = −e x (the other component of the dipole is a
stationary ion of charge +e located at x = 0). Hence, the x-directed electric
dipole moment per unit volume is
Px = ne px = −ne e x. (9.22)
Given that all of the electrons oscillate according to Equation (9.21) (with
q = −e and m = me), we obtain
ne e2 E0
Px = − cos(k z − ω t). (9.23)
me ω2
156 OSCILLATIONS AND WAVES
where Z is the effective impedance of the plasma, and Px in the form (9.23),
Equations (9.24) and (9.25) can easily be shown to yield the nonlinear dis-
persion relation (see Exercise 1)
ω2 = k2 c2 + ωp2, (9.27)
√
where c = 1/ ǫ0 µ0 is the velocity of light in vacuum, and the so-called
plasma frequency,
!1/2
ne e2
ωp = , (9.28)
ǫ0 me
is the characteristic frequency of collective electron oscillations in the plasma.
Equations (9.24) and (9.25) also yield
Z0
Z= , (9.29)
n
p
where Z0 = µ0/ǫ0 is the impedance of free space, and
!1/2
kc ωp2
n= = 1− 2 (9.30)
ω ω
the effective refractive index of the plasma. We, thus, conclude that sinu-
soidal electromagnetic waves propagating through a plasma have a nonlin-
ear dispersion relation. Moreover, it is clear that this nonlinearity arises
because the effective refractive index of the plasma is frequency dependent.
The expression (9.30) for the refractive index of a plasma has some
rather unusual properties. For wave frequencies lying above the plasma
Dispersive Waves 157
frequency (i.e., ω > ωp), it yields a real refractive index which is less than
unity. On the other hand, for wave frequencies lying below the plasma
frequency (i.e., ω < ωp), it yields an imaginary refractive index. Neither
of these results makes much sense. The former result is problematic be-
cause if the refractive index is less than unity then the wave phase velocity,
vp = ω/k = c/n, becomes superluminal (i.e., vp > c), and superluminal
velocities are generally thought to be unphysical. The latter result is prob-
lematic because an imaginary refractive index implies an imaginary phase
velocity, which seems utterly meaningless. Let us investigate further.
Consider, first of all, the high frequency limit, ω > ωp. According to
(9.30), a sinusoidal electromagnetic wave of angular frequency ω > ωp
propagates through the plasma at the superluminal phase velocity
ω c c
vp = = = . (9.31)
k n (1 − ωp /ω2)1/2
2
vg = n c = (1 − ωp2/ω2)1/2 c. (9.34)
158 OSCILLATIONS AND WAVES
Note that the group velocity is subluminal (i.e., vg < c). Hence, as long as
we accept that high frequency electromagnetic waves transmit information
through a plasma at the group velocity, rather than the phase velocity, then
there is no problem with causality. Incidentally, it should be clear, from
this discussion, that the phase velocity of dispersive waves has very little
physical significance. It is the group velocity which matters. For instance,
according to Equations (7.120), (9.29), (9.30), and (9.34), the mean flux of
electromagnetic energy in the z-direction due to a high frequency sinusoidal
wave propagating through a plasma is given by
1 1
hIi = ǫ0 E02 n c = ǫ0 E02 vg, (9.35)
2 2
p √
since Z0 = µ0/ǫ0 and c = 1/ ǫ0 µ0. Thus, if the group velocity is zero, as
is the case when ω = ωp, then there is zero energy flux associated with the
wave.
The fact that the energy flux and the group velocity of a sinusoidal wave
propagating through a plasma both go to zero when ω = ωp suggests that
the wave ceases to propagate at all in the low frequency limit, ω < ωp. This
observation leads us to search for spatially decaying standing wave solutions
to (9.24) and (9.25) of the form,
ne e2 E0 −kz
Px = − e cos(ω t). (9.38)
me ω2
Substitution into Equations (9.24) and (9.25) reveals that (9.36) and (9.37)
are indeed the correct solutions when ω < ωp (see Exercise 2), and also
yields q
kc = ωp2 − ω2, (9.39)
as well as
ω
Z = Z0 = Z0 (ωp2/ω2 − 1)−1/2. (9.40)
kc
Furthermore, the mean z-directed electromagnetic energy flux becomes
hIi = hEx Hyi = E02 Z−1 e−2kz hcos(ω t) sin(ω t)i = 0. (9.41)
Dispersive Waves 159
where Et is the amplitude of the decaying wave which penetrates into the
plasma, φt is the phase of this wave with respect to the incident wave, and
!1/2
ωp2
α= −1 . (9.46)
ω2
These two equations, which must be satisfied at all times, can be solved to
give (see Exercise 3)
Er = Ei, (9.49)
2α
tan φr = , (9.50)
1 − α2
2 Ei
Et = , (9.51)
(1 + α2)1/2
tan φt = α. (9.52)
is unity, which implies that all of the incident wave energy is reflected by
the plasma, and there is no energy absorption. The relative phase of the
reflected wave varies from 0 (when ω = ωp) to π (when ω ≪ ωp) radians.
The outer regions of the Earth’s atmosphere consist of a tenuous gas
which is partially ionized by ultraviolet and X-ray radiation from the Sun,
as well as by cosmic rays incident from outer space. This region, which is
known as the ionosphere, acts like a plasma as far as its interaction with
radio waves is concerned. The ionosphere consists of many layers. The two
most important, as far as radio wave propagation is concerned, are the E
layer, which lies at an altitude of about 90 to 120 km above the Earth’s
surface, and the F layer, which lies at an altitude of about 120 to 400 km.
The plasma frequency in the F layer is generally larger than that in the E
layer, because of the greater density of free electrons in the former (recall
√
that ωp ∝ ne). The free electron number density in the E layer drops
steeply after sunset, due to the lack of solar ionization combined with the
gradual recombination of free electrons and ions. Consequently, the plasma
frequency in the E layer also drops steeply after sunset. Recombination in
the F layer occurs at a much slower rate, so there is nothing like as great
a reduction in the plasma frequency of this layer at night. Very High Fre-
quency (VHF) radio signals (i.e., signals with frequencies greater than 30
MHz), which include FM radio and TV signals, have frequencies well in ex-
cess of the plasma frequencies of both the E and the F layers, and thus pass
straight through the ionosphere. Short Wave (SW) radio signals (i.e., signals
with frequencies in the range 3 to 30 MHz) have frequencies in excess of the
plasma frequency of the E layer, but not of the F layer. Hence, SW signals
Dispersive Waves 161
VHF
night
MW SW F
Earth E
day
pass through the E layer, but are reflected by the F layer. Finally, Medium
Wave (MW) radio signals (i.e., signals with frequencies in the range 0.5 to 3
MHz) have frequencies which lie below the plasma frequency of the F layer,
and also lie below the plasma frequency of the E layer during daytime, but
not during nighttime. Thus, MW signals are reflected by the E layer during
the day, but pass through the E layer, and are reflected by the F layer, during
the night.
The reflection and transmission of the various different types of radio
wave by the ionosphere is shown schematically in Figure 9.1. This diagram
explains many of the features of radio reception. For instance, due to the
curvature of the Earth’s surface, VHF reception is only possible when the
receiving antenna is in the line of sight of the transmitting antenna, and
is consequently fairly local in nature. MW reception is possible over much
larger distances, because the signal is reflected by the ionosphere back to-
wards the Earth’s surface. Moreover, long range MW reception improves at
night, since the signal is reflected at a higher altitude. Finally, SW radio re-
ception is possible over very large distances, because the signal is reflected
at extremely high altitudes.
A so-called Ohmic conductor is a medium that satisfies Ohm’s law, which can
be written in the form
j = σ E, (9.54)
162 OSCILLATIONS AND WAVES
where j is the current density (i.e., the current per unit area), E the electric
field-strength, and σ a constant known as the conductivity of the medium in
question. (Of course, the current generally flows in the same direction as
the electric field.) The z-directed propagation of an electromagnetic wave,
polarized in the x-direction, through an Ohmic conductor of conductivity σ
is governed by
∂Ex σ 1 ∂Hy
+ Ex = − , (9.55)
∂t ǫ0 ǫ0 ∂z
∂Hy 1 ∂Ex
= − , (9.56)
∂t µ0 ∂z
∂Ex 1∂2Ex
= , (9.57)
∂t µ0 σ ∂z2
∂Hy 1 ∂Ex
= . (9.58)
∂t µ0 ∂z
where
1/2
2
d= , (9.61)
µ0 σ ω
and
1/2 1/2
ω µ0 ω ǫ0
Z= = Z0. (9.62)
σ σ
Equations (9.59) and (9.60) indicate that the amplitude of an electromag-
netic wave propagating through a conductor decays exponentially on a char-
acteristic lengthscale, d, which is known as the skin-depth. Consequently, an
electromagnetic wave cannot penetrate more than a few skin-depths into a
conducting medium. Note that the skin-depth is smaller at higher frequen-
cies. This implies that high frequency waves penetrate a shorter distance
into a conductor than low frequency waves.
Dispersive Waves 163
hIi = hEx Hyi = |Ex|2 Z−1 hcos(ω t − z/d) cos(ω t − z/d − π/4)i
|E |2
= √x , (9.64)
8Z
where Et is the amplitude of the decaying wave which penetrates into the
conductor, φt is the phase of this wave with respect to the incident wave,
and
Z ǫ0 ω 1/2
α= = ≪ 1. (9.67)
Z0 σ
The appropriate matching conditions are the continuity of Ex and Hy at
z = 0: i.e.,
Equations (9.68) and (9.69), which must be satisfied at all times, can be
solved, in the limit α ≪ 1, to give (see Exercise 6)
√
Er ≃ −(1 − 2 α) Ei, (9.70)
√
φr ≃ − 2 α, (9.71)
Et ≃ 2 α Ei, (9.72)
φt ≃ π/4. (9.73)
If the water is to remain incompressible then the inflow and outflow rates
must match: i.e.,
vx(x, z, t) dy dz+vz(x, z, t) dx dy = vx(x+dx, z, t) dy dz+vz(x, z+dz, t) dx dy,
(9.75)
or
vx(x + dx, z, t) − vx(x, z, t) vz(x, z + dz, t) − vz(x, z, t)
+ dx dy dz = 0.
dx dz
(9.76)
Hence, the incompressibility constraint reduces to
∂vx ∂vz
+ = 0. (9.77)
∂x ∂z
Consider the equation of motion of a small volume element of water ly-
ing between x and x + dx, y and y + dy, and z and z + dz. The mass of
this element is ρ dx dy dz, where ρ is the uniform mass density of water.
Suppose that p(x, z, t) is the pressure in the water, which is assumed to be
isotropic. The net horizontal force on the element is p(x, z, t) dy dz − p(x +
dx, z, t) dy dz (since force is pressure times area, and the external pressure
forces acting on the element act inward across its surface). Hence, the ele-
ment’s horizontal equation of motion is
∂vx(x, z, t) p(x + dx, z, t) − p(x, z, t)
ρ dx dy dz =− dx dy dz, (9.78)
∂t dx
which reduces to
∂vx ∂p
ρ =− . (9.79)
∂t ∂x
The vertical equation of motion is similar, except that the element is subject
to a downward acceleration, g, due to gravity. Hence, we obtain
∂vz ∂p
ρ =− − ρ g. (9.80)
∂t ∂z
Now, we can write
p = p0 − ρ g z + p1, (9.81)
where p0 is atmospheric pressure (i.e., the pressure at the surface of the
water), and p1 is the pressure perturbation due to the wave. Of course, in
the absence of the wave, the water pressure at a depth h below the surface
is p0 + ρ g h. Substitution into (9.79) and (9.80) yields
∂vx ∂p1
ρ = − , (9.82)
∂t ∂x
∂vz ∂p1
ρ = − . (9.83)
∂t ∂z
Dispersive Waves 167
It follows that
∂2vx ∂2vz ∂2p1 ∂2p1
ρ −ρ =− + = 0, (9.84)
∂z ∂t ∂x ∂t ∂z ∂x ∂x ∂z
which implies that
∂ ∂vx ∂vz
ρ − = 0, (9.85)
∂t ∂z ∂x
or
∂vx ∂vz
− = 0. (9.86)
∂z ∂x
(Actually, the above quantity could be non-zero and constant in time, but
this is not consistent with an oscillating wave-like solution.)
Equation (9.86) is automatically satisfied if
∂φ
vx = , (9.87)
∂x
∂φ
vz = . (9.88)
∂z
Equation (9.77) then gives
∂2φ ∂2φ
+ 2 = 0. (9.89)
∂x2 ∂z
Finally, Equations (9.82) and (9.83) yield
∂φ
p1 = −ρ . (9.90)
∂t
As we have just seen, surface waves in water are governed by Equation
(9.89), which is known as Laplace’s equation. We next need to derive the
physical constraints which must be satisfied by the solution to this equation
at the water’s upper and lower boundaries. Now, the water is bounded from
below by a solid surface located at z = −d. Assuming that the water always
remains in contact with this surface, the appropriate physical constraint at
the lower boundary is vz(x, −d, t) = 0 (i.e., there is no vertical motion of
the water at the lower boundary), or
∂φ
= 0. (9.91)
∂z z=−d
The physical constraint at the water’s upper boundary is a little more compli-
cated, since this boundary is a free surface. Let ζ(x, t) represent the vertical
displacement of the water’s surface. It follows that
∂ζ ∂φ
= vz|z=0 = . (9.92)
∂t ∂z z=0
168 OSCILLATIONS AND WAVES
Now, the physical constraint at the surface is that the water pressure be
equal to atmospheric pressure, since there cannot be a pressure discontinu-
ity across a free surface. Thus, it follows from (9.81) that
Hence, the problem boils down to solving Laplace’s equation, (9.89), subject
to the physical constraints (9.91) and (9.94).
Let us search for a propagating wave-like solution of (9.89) of the form
d2F
− k2 F = 0, (9.96)
dz2
whose independent solutions are exp(+k z) and exp(−k z). Hence, the most
general wavelike solution to Laplace’s equation takes the form
This boundary condition can be combined with the solution (9.98), in the
deep water limit k d ≫ 1, to give the modified deep water dispersion rela-
tion s
T
ω = g k + k3. (9.107)
ρ
Hence, the phase velocity of the waves takes the form
s
ω g T
vp = = + k, (9.108)
k k ρ
170 OSCILLATIONS AND WAVES
and the ratio of the group velocity to the phase velocity can be shown to be
" #
vg 1 1 + 3 T k2/(ρ g)
= . (9.109)
vp 2 1 + T k2/(ρ g)
√
Thus, the phase velocity attains a minimum value of 2 (g T/ρ)1/4 ∼ 0.2 m s−1
when k = k0 ≡ (ρ g/T )1/2, which corresponds to λ ∼ 2 cm. The group veloc-
ity equals the phase velocity at this wavelength. For long wavelength waves
(i.e., k ≪ k0), gravity dominates surface tension, the phase velocity scales
as k−1/2, and the group velocity is half the phase velocity. On the other
hand, for short wavelength waves (i.e., k ≫ k0), surface tension dominates
gravity, the phase velocity scales as k1/2, and the group velocity is 3/2 times
the phase velocity. The fact that the phase velocity and the group veloc-
ity both attain minimum values when λ ∼ 2 cm means that when a wave
disturbance containing a wide spectrum of wavelengths, such as might be
generated by throwing a rock into the water, travels across the surface of
a lake, and reaches the shore, the short and long wavelength components
of the disturbance generally arrive before the components of intermediate
wavelength.
9.5 Exercises
where ωp2 (z) = ne (z) e2 /(ǫ0 me ), and ωp2 (z0 ) = ω2 . Show that if ne ∝ zp
then h ∝ ω2/p .
9. A uniform rope of mass per unit length ρ and length L hangs vertically. De-
termine the tension T in the rope as a function of height from the bottom of
the rope. Show that the time required for pa transverse wave pulse to travel
from the bottom to the top of the rope is 2 L/g.
10. The aluminium foil used in cooking has an electrical conductivity σ = 3.5 ×
107 (Ω m)−1 , and a typical thickness δ = 2 × 10−4 m. Show that such foil can
be used to shield a region from electromagnetic waves of a given frequency,
provided that the skin-depth of the waves in the foil is less than about a
third of its thickness. Since skin-depth increases as frequency decreases, it
follows that the foil can only shield waves whose frequency exceeds a critical
value. Estimate this critical frequency (in Hertz). What is the corresponding
wavelength?
11. A sinusoidal surface wave travels from deep water toward the shore. Does its
wavelength increase, decrease, or stay the same, as it approaches the shore?
Explain.
12. Demonstrate that the dispersion relation (9.107) for surface water waves
generalizes to
T
ω2 = g k + k3 tanh(k d)
ρ
in water of arbitrary depth.
172 OSCILLATIONS AND WAVES
10 Multi-Dimensional Waves
k x − ω t = j 2π, (10.2)
x = d, (10.3)
n · r = d, (10.4)
n
r
plane
d
origin
where the constant vector k = (kx, ky, kz) = k n is called the wavevector.
The wave represented above is conventionally termed a three-dimensional
plane wave. It is three-dimensional because its wavefunction, ψ(x, y, z, t),
depends on all three Cartesian coordinates. Moreover, it is a plane wave
because the wave maxima are located at
k · r − ω t = j 2π, (10.6)
or
n · r = j λ + v t, (10.7)
where λ = 2π/k, and v = ω/k. Note that the wavenumber, k, is the mag-
nitude of the wavevector, k: i.e., k ≡ |k|. It follows, by comparison with
Equation (10.4), that the wave maxima consist of a series of parallel planes,
normal to the wavevector, which are equally spaced a distance λ apart, and
propagate in the k-direction at the fixed speed v. See Figure 10.2. Hence,
the direction of the wavevector specifies the wave propagation direction,
whereas its magnitude determines the wavenumber, k, and, thus, the wave-
length, λ = 2π/k. Actually, the most general expression for the wavefunc-
tion of a plane wave is ψ = ψ0 cos(φ + k · r − ω t), where φ is a constant
phase angle. As is easily appreciated, the inclusion of a non-zero phase angle
in the wavefunction merely shifts all the wave maxima a distance −(φ/2π) λ
in the k-direction. In the following, whenever possible, φ is set to zero, for
the sake of simplicity.
Multi-Dimensional Waves 175
∂2ψ 2
2∂ ψ
= v . (10.8)
∂t2 ∂x2
Likewise, the three-dimensional plane wave solution (10.5) satisfies the
three-dimensional wave equation (see Exercise 1),
!
∂2ψ 2 ∂2 ∂2 ∂2
= v + + ψ. (10.9)
∂t2 ∂x2 ∂y2 ∂z2
Note that both of these equations are linear, and, thus, have superposable
solutions.
kr z=0
kt
x
θr θt
z x=0
θi ki
n1 n2
The first of these relations states that the angle of incidence, θi, is equal to
the angle of reflection, θr. Of course, this is the familiar law of reflection.
Moreover, the second relation corresponds to the equally familiar law of
refraction, otherwise known as Snell’s law.
Incidentally, the fact that a plane wave propagates through a uniform
medium with a constant wavevector, and, therefore, a constant direction of
propagation, is equivalent to the well known law of rectilinear propagation,
which states that light propagates through a uniform medium in a straight-
line.
It is clear, from the above discussion, that the laws of geometric optics
(i.e., the law of rectilinear propagation, the law of reflection, and the law of
refraction) are fully consistent with the wave properties of light, despite the
fact that they do not appear to explicitly depend on these properties.
10.4 Waveguides
As we saw in Section 7.5, transmission lines (e.g., ethernet cables) are used
to carry high frequency electromagnetic signals over distances which are
long compared to the signal wavelength, λ = c/f, where c is the veloc-
ity of light and f the signal frequency (in Hertz). Unfortunately, conven-
tional transmission lines are subject to radiative losses (since the lines ef-
fectively act as antennas) which increase as the fourth power of the signal
frequency. Above a certain critical frequency, which typically lies in the mi-
crowave band, the radiative losses become intolerably large. Under these
circumstances, the transmission line must be replaced by a device known as
a waveguide. A waveguide is basically a long hollow metal box within which
electromagnetic signals propagate. Moreover, if the walls of the box are
much thicker than the skin-depth (see Section 9.3) in the wall material then
the signal is essentially isolated from the outside world, and the radiative
losses are consequently negligible.
Consider an evacuated waveguide of rectangular cross-section which
runs along the z-direction, and is enclosed by perfectly conducting (i.e.,
infinite conductivity) metal walls located at x = 0, x = a, y = 0, and y = b.
Suppose that an electromagnetic wave propagates along the waveguide in
the z-direction. For the sake of simplicity, let there be no y-variation of
the wave electric or magnetic fields. Now, the wave propagation inside the
waveguide is governed by the two-dimensional wave equation [cf., Equa-
Multi-Dimensional Waves 179
tion (10.9)] !
∂2ψ 2 ∂2 ∂2
= c + ψ, (10.20)
∂t2 ∂x2 ∂z2
where ψ(x, z, t) represents the electric component of the wave, which is as-
sumed to be everywhere parallel to the y-axis, and c is the velocity of light
in vaccum. The appropriate boundary conditions are
ψ(0, z, t) = 0, (10.21)
ψ(a, z, t) = 0, (10.22)
since the electric field inside a perfect conductor is zero (otherwise, an infi-
nite current would flow), and, according to standard electromagnetic theory,
there cannot be a tangential discontinuity in the electric field at a conduc-
tor/vacuum boundary. (There can, however, be a normal discontinuity. This
allows ψ to be non-zero at y = 0 and y = b.)
Let us search for a separable solution of (10.20) of the form
where k represents the z-component of the wavevector (rather than its mag-
nitude), and is the effective wavenumber for propagation along the waveg-
uide. The above solution automatically satisfies the boundary condition
(10.21). The second boundary condition (10.22) is satisfied provided
π
kx = j , (10.24)
a
where j is a positive integer. Suppose that j takes its smallest possible value
1. (Of course, j cannot be zero, since, in this case, ψ = 0 everywhere.)
Substitution of expression (10.23) into the wave equation (10.20) yields
the dispersion relation
ω2 = k2 c2 + ω02, (10.25)
where
πc
ω0 = . (10.26)
a
Note that this dispersion relation is analogous in form to the dispersion
relation (9.27) for an electromagnetic wave propagating through a plasma,
with the cut-off frequency, ω0, playing the role of the plasma frequency, ωp.
The cut-off frequency is so-called because for ω < ω0 the wavenumber is
imaginary (i.e., k2 < 0), which implies that the wave does not propagate
along the waveguide, but, instead, decays exponentially with increasing z.
180 OSCILLATIONS AND WAVES
On the other hand, for wave frequencies above the cut-off frequency the
phase velocity,
ω c
vp = =q , (10.27)
k 1 − ω 2/ω20
is superluminal. This is not a problem, however, since the group velocity,
dω q
vg = = c 1 − ω02/ω2, (10.28)
dk
which is the true signal velocity, remains subluminal. (Recall, from Sec-
tion 9.2, that a high frequency electromagnetic wave propagating through a
plasma exhibits similar behavior.) Not surprisingly, the signal velocity goes
to zero as ω → ω0, since the signal ceases to propagate at all when ω = ω0.
It turns out that waveguides support many distinct modes of propaga-
tion. The type of mode discussed above is termed a TE (for transverse
electric-field) mode, since the electric field is transverse to the direction
of propagation. There are many different sorts of TE mode, corresponding,
for instance, to different choices of the mode number, j. However, the j = 1
mode has the lowest cut-off frequency. There are also TM (for transverse
magnetic-field) modes, and TEM (for transverse electric- and magnetic-
field) modes. TM modes also only propagate when the wave frequency
exceeds a cut-off frequency. On the other hand, TEM modes (which are the
same type of mode as that supported by a conventional transmission line)
propagate at all frequencies. Note, however, that TEM modes are only pos-
sible when the waveguide possesses an internal conductor running along its
length.
v v
x
v
Figure 10.4: A cylindrical wave.
10.6 Exercises
1. Show that the one-dimensional plane wave (10.1) is a solution of the one-
dimensional wave equation (10.8) provided that
ω = k v.
182 OSCILLATIONS AND WAVES
ω = |k| v.
∂2 ψ
2
2 ∂ ψ 1 ∂ψ
= v + .
∂t2 ∂ρ2 ρ ∂ρ
Show that
ψ0
ψ(ρ, t) ≃ cos(φ + k ρ − ω t)
ρ1/2
is an approximate solution of this equation in the limit k ρ ≫ 1, where v =
ω/k.
4. Demonstrate
p that for a spherically symmetric wavefunction ψ(r, t), where
r = x2 + y2 + z2 , the three-dimensional wave equation (10.9) can be re-
written
∂2 ψ
2
2 ∂ ψ 2 ∂ψ
= v + .
∂t2 ∂r2 r ∂r
Show that
ψ0
ψ(r, t) = cos(φ + k r − ω t)
r
is a solution of this equation, where v = ω/k. Explain why the wave am-
plitude attenuates as r−1 . What sort of wave source would be most likely to
generate the above type of wave solution?
Wave Optics 183
11 Wave Optics
11.1 Introduction
Here, ψ represents the electric component of the wave, ψ0 > 0 the wave
amplitude, φ the phase angle, k > 0 the wavenumber, ω = k c the angular
frequency, and c the velocity of light in vacuum. Let the wave be normally
184 OSCILLATIONS AND WAVES
opaque screen
cylindrical
projection screen
ρ1
symmetry plane
R
incident d/2 ρ2
plane θ y=0
wave d/2
diffracted
cylindrical
wave y
x=0
(i.e., toward the region x < 0). (It is actually possible to demonstrate that
this is, in fact, the case using advanced electromagnetic theory, but such a
demonstration lies well beyond the scope of this course.) As discussed in
Section 10.5, we would expect a uniform line source to emit a cylindrical
wave. It follows that each slit emits a half-cylindrical light wave in the for-
ward direction. See Figure 11.1. Moreover, these waves are emitted with
equal amplitude and phase, since the incident plane wave has the same am-
plitude (i.e., ψ0) and phase (i.e., φ − ω t) at both of the slits, and the slits
are identical. Finally, we expect the cylindrical waves emitted by the two
slits to interfere with one another (see Section 7.3), and to, thus, generate
a characteristic interference pattern on the cylindrical projection screen. Let
us now determine the nature of this pattern.
Consider the wave amplitude at a point on the projection screen which
lies an angular distance θ from the plane y = 0. See Figure 11.1. The
wavefunction at this particular point is written
!
cos(φ + k ρ1 − ω t) 1
ψ(θ, t) ∝ 1/2
+O 3/2
ρ1 k ρ1
!
cos(φ + k ρ2 − ω t) 1
+ 1/2
+O 3/2
, (11.2)
ρ2 k ρ2
assuming that k ρ1, k ρ2 ≫ 1. In other words, the overall wavefunction
in the region x > 0 is the superposition of cylindrical wavefunctions [see
Equation (10.30)] of equal amplitude (i.e., ρ−1/2) and phase (i.e., φ + k ρ −
ω t) emanating from each slit. Here, ρ1 and ρ2 are the distances which the
cylindrical waves emitted by the first and second slits (located at y = d/2
and y = −d/2, respectively) have travelled by the time they reach the point
on the projection screen under discussion.
Standard trigonometry (i.e., the law of cosines) reveals that
!1/2 " !#
d 1 d2 1d d2
ρ1 = R 1 − sin θ + =R 1− sin θ + O 2 . (11.3)
R 4 R2 2R R
Likewise, " !#
1d d2
ρ2 = R 1 + sin θ + O 2 . (11.4)
2R R
Hence, expression (11.2) yields
ψ(θ, t) ∝ cos(φ + k ρ1 − ω t) + cos(φ + k ρ2 − ω t)
1 d
+O +O , (11.5)
kR R
186 OSCILLATIONS AND WAVES
Figure 11.2: Two-slit far-field interference pattern calculated for d/λ = 5 with
normal incidence and narrow slits.
1 1
ψ(θ, t) ∝ cos φ + k (ρ1 + ρ2) − ω t cos k (ρ1 − ρ2)
2 2
1 d
+O +O , (11.6)
kR R
or
" !# " !#
k d2 1 k d2
ψ(θ, t) ∝ cos φ + k R − ω t + O cos − k d sin θ + O
R 2 R
1 d
+O +O . (11.7)
kR R
1
ψ(θ, t) ∝ cos(φ + k R − ω t) cos k d sin θ . (11.9)
2
Wave Optics 187
Figure 11.3: Two-slit far-field interference pattern calculated for d/λ = 1 with
normal incidence and narrow slits.
Now, the orderings (11.8), which can also be written in the form,
d2
R ≫ d, λ, , (11.10)
λ
are satisfied provided that the projection screen is located sufficiently far
away from the slits. Consequently, the type of interference described in this
section is known as far-field interference. The characteristic features of far-
field interference are that the amplitudes of the cylindrical waves emitted
by the two slits are approximately equal to one another when they reach
a given point on the projection screen (i.e., |ρ1 − ρ2|/ρ1 ≪ 1), whereas
the phases are, in general, significantly different (i.e., k |ρ1 − ρ2| >∼ π). In
other words, the interference pattern generated on the projection screen is
entirely due to the phase difference between the cylindrical waves emitted by
the two slits when they reach the screen. This phase difference is produced
by the slight difference in path length between the slits and a given point on
the projection screen. (Recall, that the two waves are in phase when they
are emitted by the slits.)
The mean energy flux, or intensity, of the light striking the projection
screen at angular position θ is
Figure 11.4: Two-slit far-field interference pattern calculated for d/λ = 0.1
with normal incidence and narrow slits.
1
2
∝ cos k d sin θ , (11.11)
2
where h· · ·i denotes an average over a wave period. (The above expression
follows from the standard result I = E2/Z0, for an electromagnetic wave,
where E is the electric component of the wave, and Z0 the impedance of
free space. See Section 7.7. Recall, also, that ψ ∝ E.) Here, we have made
use of the easily established result hcos2(φ + k R − ω t)i = 1/2. Note that,
given the very high oscillation frequency of a light wave (i.e., f ∼ 1014 Hz),
it is the intensity of light, rather than the rapidly oscillating amplitude of
its electric component, which is typically detected experimentally (e.g., by a
photographic film, or photo-multiplier tube). Hence, for the case of two-slit
far-field interference, assuming normal incidence and narrow slits, the in-
tensity of the characteristic interference pattern appearing on the projection
screen is specified by
d
2
I(θ) ∝ cos π sin θ . (11.12)
λ
Figure 11.2 shows the intensity of the typical two-slit far-field interfer-
ence pattern produced when the slit spacing, d, greatly exceeds the wave-
length, λ, of the light. It can be seen that the pattern consists of multiple
bright and dark fringes. A bright fringe is generated whenever the cylindri-
cal waves emitted by the two slits interfere constructively at given point on
Wave Optics 189
the projection screen. This occurs if the path lengths between the two slits
and the point in question differ by an integer number of wavelengths: i.e.,
ρ2 − ρ1 = d sin θ = j λ, (11.13)
where j is an integer, since this ensures that the phases of the two waves
differ by an integer multiple of 2π, and, hence, that the effective phase dif-
ference is zero. Likewise, a dark fringe is generated whenever the cylindrical
waves emitted by the two slits interfere destructively at a given point on the
projection screen. This occurs if the path lengths between the two slits and
the point in question differ by a half-integer number of wavelengths: i.e.,
since this ensures that the effective phase difference between the two waves
is π. We conclude that the innermost (i.e., low j, small θ) bright fringes are
approximately equally-spaced, with a characteristic angular width ∆θ ≃ λ/d.
This result, which follows from Equation (11.13), and the small angle ap-
proximation sin θ ≃ θ, can be used experimentally to determine the wave-
length of a monochromatic light source (see Exercise 11.4).
Figure 11.3 shows the intensity of the interference pattern generated
when the slit spacing is equal to the wavelength of the light. It can be seen
that the width of the central (i.e., j = 0, θ = 0) bright fringe has expanded to
such an extent that the fringe occupies almost half of the projection screen,
leaving room for just two dark fringes on either side of it.
Finally, Figure 11.4 shows the intensity of the interference pattern gen-
erated when the slit spacing is much less than the wavelength of the light. It
can be seen that the width of the central bright fringe has expanded to such
an extent that the band occupies the whole projection screen, and there are
no dark fringes. Indeed, I(θ) becomes constant in the limit that d/λ ≪ 1,
in which case the interference pattern entirely disappears.
It is clear, from Figures 11.2–11.3, that the two-slit far-field interference
apparatus shown in Figure 11.1 only generates an interesting interference
patten when the slit spacing, d, is greater than the wavelength, λ, of the
light.
Suppose, now, that the plane wave which illuminates the interference
apparatus is not normally incident on the slits, but instead propagates at an
angle θ0 to the x-axis, as shown in Figure 11.5. In this case, the incident
wavefunction (11.1) becomes
opaque screen
incident
plane
wave
projection screen
interference
pattern
j = 0 bright fringe
d/2
θ0 y=0
d/2 R
x=0
Thus, the phase of the light incident on the first slit (located at x = 0, y =
d/2) is φ + (1/2) k d sin θ0 − ω t, whereas the phase of the light incident on
the second slit (located at x = 0, y = −d/2) is φ − (1/2) k d sin θ0 − ω t. As-
suming that the cylindrical waves emitted by each slit have the same phase
(at the slits) as the plane wave which illuminates them, Equation (11.2)
generalizes to
cos(φ1 + k ρ1 − ω t) cos(φ2 + k ρ2 − ω t)
ψ(θ, t) ∝ 1/2
+ 1/2
, (11.16)
ρ1 ρ2
where φ1 = φ + (1/2) k d sin θ0 and φ2 = φ − (1/2) k d sin θ0. Hence,
making use of the far-field orderings (11.10), and a standard trigonometric
identity, we obtain
1 1 1
ψ(θ, t) ∝ cos (φ1 + φ2) + k R − ω t cos (φ1 − φ2) − k d sin θ
2 2 2
1
∝ cos(φ + k R − ω t) cos k d (sin θ − sin θ0) . (11.17)
2
For the sake of simplicity, let us concentrate on the limit d ≫ λ in which
the innermost (i.e., low j) interference fringes are located at small θ. (Note
Wave Optics 191
d
2
I(θ) ∝ cos π (θ − θ0) . (11.19)
λ
λ
θ = θ0 + j , (11.20)
d
11.3 Coherence
opaque screen
incoherent
B light sources projection screen
D/2
θ0/2 d/2
y=0
L θ0/2 d/2 R
D/2
y
A
x
x=0
timescales much greater than τ, since the phase angle, φ of the plane wave
which illuminates the two slits is subject to the same changes. Nevertheless,
the relative phase angle, φ1 − φ2 = k d sin θ0, between the two cylindrical
waves remains constant. Moreover, as is clear from Equation (11.17), the
interference pattern appearing on the projection screen is generated by the
phase difference (1/2) (φ1 − φ2) − (1/2) k d sin θ between the two cylindrical
waves at a given point on the screen, and this phase difference only depends
on the relative phase angle. Indeed, the intensity of the interference pattern
is I(θ) ∝ cos2[(1/2) (φ1 − φ2) − (1/2) k d sin θ]. Hence, the fact that the
relative phase angle, φ1 − φ2, between the two cylindrical waves emitted
by the slits remains constant on timescales much longer than the charac-
teristic coherence time, τ, of the light source implies that the interference
pattern generated in a conventional two-slit interference apparatus is gener-
ally unaffected by the temporal incoherence of the source. Strictly speaking,
however, the preceding conclusion is only accurate when the spatial extent
of the light source is negligible. Let us now broaden our discussion to take
spatially extended light sources into account.
Up until now, we have assumed that our two-slit interference apparatus
is illuminated by a single plane wave, such as might be generated by a line
source located at infinity. Let us now consider a more realistic situation in
194 OSCILLATIONS AND WAVES
which the light source is located a finite distance from the slits, and also has
a finite spatial extent. Figure 11.6 shows the simplest possible case. Here,
the slits are illuminated by two identical line sources, A and B, which are
a distance D apart, and a perpendicular distance L from the opaque screen
containing the slits. Assuming that L ≫ D, d, the light incident on the slits
from source A is effectively a plane wave whose direction of propagation
subtends an angle θ0/2 ≃ D/2 L with the x-axis. Likewise, the light incident
on the slits from source B is a plane wave whose direction of propagation
subtends an angle −θ0/2 with the x-axis. Moreover, the net interference
pattern (i.e., wavefunction) appearing on the projection screen is the linear
superposition of the patterns generated by each source taken individually
(since light propagation is ultimately governed by a linear wave equation
with superposable solutions—see Section 10.2.). Let us determine whether
these patterns reinforce one another, or interfere with one another.
The light emitted by source A has a phase angle, φA, which is constant
on timescales much less than the characteristic coherence time of the source,
τ, but is subject to abrupt random changes on timescale much longer than τ.
Likewise, the light emitted by source B has a phase angle, φB, which is con-
stant on timescales much less than τ, and varies significantly on timescales
much greater than τ. Furthermore, there is, in general, no correlation be-
tween φA and φB. In other words, our composite light source, consisting of
the two line sources A and B, is both temporally and spatially incoherent on
timescales much longer than τ.
Again working in the limit d ≫ λ, with θ, θ0 ≪ 1, Equation (11.18)
yields the following expression for the wavefunction on the projection screen:
1
ψ(θ, t) ∝ cos(φA + k R − ω t) cos k d (θ − θ0/2)
2
1
+ cos(φB + k R − ω t) cos k d (θ + θ0/2) . (11.22)
2
1
2 2 2
I(θ) ∝ hψ i ∝ h cos (φA + k R − ω t)i cos k d (θ − θ0/2)
2
+2 hcos(φA + k R − ω t) cos(φB + k R − ω t)i
1 1
× cos k d (θ − θ0/2) cos k d (θ + θ0/2)
2 2
2 1
2
+h cos (φB + k R − ω t)i cos k d (θ + θ0/2) .
2
Wave Optics 195
(11.23)
1 1
2
I(θ) ∝ cos k d (θ − θ0/2) + cos2 k d (θ + θ0/2)
2 2
d d
= 1 + cos 2π θ cos π θ0 , (11.24)
λ λ
d d
I(θ) ∝ 1 + cos 2π θ sinc π θ0 , (11.25)
λ λ
where sinc(x) ≡ sin x/x. Now, we can conveniently parameterize the visibil-
ity of the interference pattern, appearing on the projection screen, in terms
of the quantity
Imax − Imin
V= , (11.26)
Imax + Imin
where the maximum and minimum values of the intensity are taken with
respect to variation in θ (rather than θ0). Of course, V = 1 corresponds to a
196 OSCILLATIONS AND WAVES
Moreover, when this expression is averaged over θ0, in order to generate the
interference pattern produced by a uniform coherent light source of angular
extent θ0, we obtain
d d
2
I(θ) ∝ 2 cos π θ 1 + sinc π θ0 . (11.31)
λ λ
Thus, making use of the far-field orderings (11.10), where d now repre-
sents the typical spacing between neighboring slits, and assuming normally
incident collimated light, Equation (11.5) generalizes to
X
ψ(θ, t) ∝ cos(φ + k R − ω t − k yn sin θ), (11.33)
n=1,N
yn = [n − (N + 1)/2] d (11.37)
l λ
sin θ = , (11.41)
Nd
200 OSCILLATIONS AND WAVES
tan θj
δθj = . (11.43)
Nj
∆λ
∆θj = tan θj . (11.44)
λ
Clearly, in order for the spectroscope to resolve the incident light into its two
constituent wavelengths, at the jth spectral order, the angular spacing, ∆θj,
between the jth-order maxima associated with these two wavelengths must
be greater than the angular widths, δθj, of the maxima themselves. If this
is the case then the overall jth-order maximum will consist of two closely
spaced maxima, or “spectral lines” (centered at θj and θj + ∆θj). On the
other hand, if this is not the case then the two maxima will merge to form
a single maximum, and it will consequently not be possible to tell that the
incident light consists of a mixture of two different wavelengths. Thus, the
condition for the spectroscope to be able to resolve the spectral lines at the
202 OSCILLATIONS AND WAVES
∆λ 1
> . (11.45)
λ Nj
Here, for the sake of simplicity, we have assumed that P the arrangement of
slits is symmetric with respect to the plane y = 0, so that n=1,N sin(α yn) =
0 for any α. Now, using the well-known properties of the delta function [see
Equation (8.26)], expression (11.47) can also be written
ψ(θ, t) ∝ cos(φ + k R − ω t) F̄(θ), (11.48)
where Z∞
F̄(θ) = F(y) cos[k (sin θ − sin θ0) y] dy. (11.49)
−∞
In the following, we shall assume that Equation (11.48) is a general result,
and is valid even when the slits in the opaque screen are of finite width (i.e.,
δ >∼ λ). This assumption is equivalent to the assumption that each unblocked
section of the screen emits a cylindrical wave in the forward direction that
is in phase with the plane wave which illuminates it from behind. The latter
assumption is known as Huygen’s principle. (Huygen’s principle can actu-
ally be justified using advanced electromagnetic theory, but such a proof lies
well beyond the scope of this course.) Note that the interference/diffraction
function, F̄(θ), is just the Fourier transform of the aperture function, F(y).
This is an extremely powerful result. It implies that we can work out the
far-field interference/diffraction pattern associated with any arrangement
of parallel slits, of arbitrary width, by simply Fourier transforming the asso-
ciated aperture function. Of course, once we have calculated the interfer-
ence/diffraction function, F̄(θ), the intensity of the interference/diffraction
pattern appearing on the projection screen is readily obtained from
h i2
I(θ) ∝ F̄(θ) . (11.50)
Suppose that the opaque screen contains a single slit of finite width. In fact,
let the slit in question be of width δ, and extend from y = −δ/2 to y = δ/2.
Thus, the associated aperture function is
1/δ |y| ≤ δ/2
F(y) = . (11.51)
0 |y| > δ/2
It follows from (11.49) that
Z
1 δ/2 δ
F̄(θ) = cos[k (sin θ − sin θ0) y] dy = sinc π (sin θ − sin θ0) ,
δ −δ/2 λ
(11.52)
204 OSCILLATIONS AND WAVES
Figure 11.9: Single-slit far-field diffraction pattern calculated for δ/λ = 20.
where sinc(x) ≡ sin(x)/x. Finally, assuming, for the sake of simplicity, that
θ, θ0 ≪ 1, which is most likely to be the case when δ ≫ λ, the diffraction
pattern appearing on the projection screen is specified by
δ
2
I(θ) ∝ sinc π (θ − θ0) . (11.53)
λ
Note, from L’Hopital’s rule, that sinc(0) = limx→0 sin x/x = limx→ 0 cos x/1 =
1. Furthermore, it is easily demonstrated that the zeros of the sinc(x) func-
tion occur at x = j π, where j is a non-zero integer.
Figure 11.9 shows a typical single-slit diffraction pattern calculated for
a case in which the slit width greatly exceeds the wavelength of the light.
Note that the pattern consists of a dominant central maximum, flanked by
subsidiary maxima of negligible amplitude. The situation is shown schemat-
ically in Figure 11.10. When the incident plane wave, whose direction of
motion subtends an angle θ0 with the x-axis, passes through the slit it is
effectively transformed into a divergent beam of light (where the beam cor-
responds to the central peak in Figure 11.9) that is centered on θ = θ0. The
angle of divergence of the beam, which is obtained from the first zero of the
single-slit diffraction function (11.53), is
λ
δθ = : (11.54)
δ
i.e., the beam effectively extends from θ0 − δθ to θ0 + δθ. Thus, if the slit
width, δ, is very much greater than the wavelength, λ, of the light then the
Wave Optics 205
opaque screen
incident
plane
wave
diffracted
beam
δθ
δ θ0 y=0
x=0
beam divergence is negligible, and the beam is, thus, governed by the laws
of geometric optics (according to which there is no beam divergence). On
the other hand, if the slit width is comparable with the wavelength of the
light then the beam divergence is significant, and the behavior of the beam
is, consequently, very different than that predicted by the laws of geometric
optics.
The diffraction of light is a very important physical phenomenon, since
it sets a limit on the angular resolution of optical instruments. For instance,
consider a telescope whose objective lens is of diameter D. When a plane
wave from a distant light source of negligible angular extent (e.g., a star)
enter the lens it is diffracted, and forms a divergent beam of angular width
λ/D. Thus, instead of being a point, the resulting image of the star is a disk
of finite angular width λ/D. Such a disk is known as an Airy disk. Suppose
that two stars are an angular distance ∆θ apart in the sky. As we have just
seen, when viewed through the telescope, each star appears as a disk of
angular extent δθ = λ/D. Clearly, if ∆θ > δθ then the two stars appear as
separate disks. On the other hand, if ∆θ < δθ then the two disks merge to
form a single disk, and it becomes impossible to tell that there are, in fact,
two stars. It follows that the maximum angular resolution of a telescope
206 OSCILLATIONS AND WAVES
λ
δθ ≃ . (11.55)
D
This result is usually called the Rayleigh criterion. Note that the angular
resolution of the telescope increases (i.e., δθ decreases) as the diameter of
its objective lens increases.
1 X
F(y) = F2(y − yn), (11.56)
N
n=1,N
where
yn = [n − (N + 1)/2] d, (11.57)
and
1/δ |y| ≤ δ/2
F2(y) = . (11.58)
0 |y| > δ/2
Of course, F2(y) is the aperture function for a single slit, of finite width δ,
which is centered on θ = 0—see Equation (11.51).
Assuming normal incidence (i.e., θ0 = 0), the interference/diffraction
function, which is the Fourier transform of the aperture function, takes the
form [see Equation (11.49)]
Z∞
F̄(θ) = F(y) cos(k sin θ y) dy. (11.59)
−∞
Hence,
Z
1 X ∞
F̄(θ) = F2(y − yn) cos(k sin θ y) dy
N
n=1,N −∞
" Z∞
1 X
= cos(k sin θ yn) F2(y ′ ) cos(k sin θ y ′ ) dy ′
N −∞
n=1,N
Wave Optics 207
Z∞ #
′ ′ ′
− sin(k sin θ yn) F2(y ) sin(k sin θ y ) dy (11.60)
−∞
X Z∞
1
= cos(k sin θ yn) F2(y ′ ) cos(k sin θ y ′ ) dy ′ ,
N −∞
n=1,N
R∞
where y ′ = y−yn. Here, we have made use of −∞ F2(y ′ ) sin(α y ′ ) dy ′ = 0,
for any α, which follows because F2(y ′ ) is even in y ′ , whereas sin(α y ′ ) is
odd. We have also employed the standard identity cos(x−y) ≡ cos x cos y−
sin x sin y. The above expression reduces to
where
1 X
F1(y) = δ(y − yn), (11.63)
N
n=1,N
where j is an integer. The latter function (see Figure 11.9, which shows
[F̄2(θ − θ0)]2) is of order unity for |θ| <∼ sin−1(λ/δ), and much less than
unity for |θ| >∼ sin−1(λ/δ). It follows that the intensity of the interfer-
ence/diffraction pattern associated with N identical equally-spaced parallel
slits of finite width, which is given by
h i2 h i2 h i2
I(θ) ∝ F̄1(θ) F̄2(θ) ∝ F̄1(θ) F̄1(θ) , (11.66)
is similar to that for N identical equally-spaced parallel slits of negligible
width, [F̄1(θ)]2, except that the heights of the various maxima in the pattern
are modulated by [F̄2(θ)]2. Hence, those maxima lying in the angular range
|θ| < sin−1(λ/δ) are of similar height, whereas those lying in the range |θ| >
sin−1(λ/δ) are of negligible height. This is illustrated in Figure 11.11, which
shows the multi-slit interference/diffraction pattern calculated for N = 10,
d/λ = 20, and δ/λ = 2. As expected, the maxima lying in the angular range
|θ| < sin−1(0.5) = π/6 have relatively large amplitudes, whereas those lying
in the range |θ| > π/6 have negligibly small amplitudes.
11.8 Exercises
1 − zN
S= .
1−z
demonstrate that X
C= cos(α yn ),
n=1,N
where
yn = [n − (N + 1)/2] d,
evaluates to
sin(N α d/2)
C= .
sin(α d/2)
2. An interference experiment employs two narrow parallel slits of separation
0.25 mm, and monochromatic light of wavelength λ = 500 nm. Estimate the
minimum distance that the projection screen must be placed behind the slits
in order to obtain a far-field interference pattern.
3. A double-slit of slit separation 0.5 mm is illuminated at normal incidence by
a parallel beam from a helium-neon laser that emits monochromatic light
of wavelength 632.8 nm. A projection screen is located 5 m behind the slit.
What is the separation of the central interference fringes on the screen?
4. Consider a double-slit interference experiment in which the slit spacing is
0.1 mm, and the projection screen is located 50 cm behind the slits. Assuming
monochromatic illumination at normal incidence, if the observed separation
between neighboring interference maxima at the center of the projection
screen is 2.5 mm, what is the wavelength of the light illuminating the slits?
5. What is the mean length of the classical wavetrain (wave packet) correspond-
ing to the light emitted by an atom whose excited state has a mean lifetime
τ ∼ 10−8 s? In an ordinary gas-discharge source, the excited atomic states do
not decay freely, but instead have an effective lifetime τ ∼ 10−9 s, due to col-
lisions and Doppler effects. What is the length of the corresponding classical
wavetrain?
210 OSCILLATIONS AND WAVES
12 Wave Mechanics
12.1 Introduction
According to classical physics (i.e., physics prior to the 20th century), par-
ticles and waves are two completely distinct classes of physical entity that
possess markedly different properties. 1) Particles are discrete: i.e., they
cannot be arbitrarily divided. In other words, it makes sense to talk about
one electron, or two electrons, but not about a third of an electron. Waves,
on the other hand, are continuous: i.e., they can be arbitrarily divided. In
other words, given a wave whose amplitude has a certain value, it makes
sense to talk about a similar wave whose amplitude is one third, or any
other fraction whatsoever, of this value. 2) Particles are highly localized in
space. For example, atomic nuclei have very small radii of order 10−15 m,
whilst electrons act like point particles: i.e., they have no discernible spatial
extent. Waves, on the other hand, are non-localized in space. In fact, a wave
is defined to be a disturbance that is periodic in space, with some finite peri-
odicity length: i.e., wavelength. Hence, it is fairly meaningless to talk about
a disturbance being a wave unless it extends over a region of space that is
at least a few wavelengths in dimension.
The classical scenario, described above, in which particles and waves are
completely distinct from one another, had to be significantly modified in the
early decades of the 20th century. During this time period, physicists dis-
covered, much to their surprise, that, under certain circumstances, waves
act as particles, and particles act as waves. This bizarre phenomenon is
known as wave-particle duality. For instance, the photoelectric effect (see Sec-
tion 12.2) indicates that electromagnetic waves sometimes act like swarms
of massless particles called photons. Moreover, the phenomenon of elec-
tron diffraction by atomic lattices (see Section 12.3) implies that electrons
sometimes have wave-like properties. Note, however, that wave-particle
duality usually only manifests itself on atomic and sub-atomic lengthscales
(i.e., on lengthscales less than, or of order, 10−10 m—see Section 12.3). The
classical picture remains valid on significantly longer lengthscales. In other
words, on macroscopic lengthscales, waves only act like waves, particles only
act like particles, and there is no wave-particle duality. However, on mi-
croscopic lengthscales, classical mechanics, which governs the macroscopic
behavior of massive particles, and classical electrodynamics, which governs
the macroscopic behavior of electromagnetic fields—neither of which take
214 OSCILLATIONS AND WAVES
E = h̄ ω. (12.1)
h̄
0
0 W/h̄ ω
Figure 12.1: Variation of the kinetic energy K of photoelectrons with the wave
angular frequency ω.
W then the electron is emitted from the surface with the residual kinetic
energy
K = h̄ ω − W. (12.2)
Otherwise, the electron remains trapped in the potential well, and is not
emitted. Here, we are assuming that the probability of an electron absorb-
ing two or more photons is negligibly small compared to the probability
of it absorbing a single photon (as is, indeed, the case for low intensity il-
lumination). Incidentally, we can determine Planck’s constant, as well as
the work-function of the metal, by plotting the kinetic energy of the emitted
photoelectrons as a function of the wave frequency, as shown in Figure 12.1.
This plot is a straight-line whose slope is h̄, and whose intercept with the
ω axis is W/h̄. Finally, the number of emitted electrons increases with the
intensity of the light because the more intense the light the larger the flux of
photons onto the surface. Thus, Einstein’s quantum theory of light is capable
of accounting for all three of the previously mentioned observational facts
regarding the photoelectric effect. In the following, we shall assume that
the central component of Einstein’s theory—namely, Equation (12.1)—is a
general result which applies to all particles, not just photons.
216 OSCILLATIONS AND WAVES
p = h̄ k. (12.3)
z = x + i y, (12.6)
where x and y are real numbers. In fact, x is termed the real part of z, and
y the imaginary part of z. This is written mathematically as x = Re(z) and
y = Im(z). Finally, the complex conjugate of z is defined z∗ = x − i y.
Now, just as we can visualize a real number as a point on an infinite
straight-line, we can visualize a complex number as a point in an infinite
plane. The coordinates of the point in question are the real and imagi-
p z ≡ (x, y). This idea is illustrated in Fig-
nary parts of the number: i.e.,
ure 12.2. The distance, r = x2 + y2, of the representative point from the
origin is termed the modulus of the corresponding
p complex number, z. This
2 2
is written mathematically as |z| = x + y . Incidentally, it follows that
z z∗ = x2 + y2 = |z|2. The angle, θ = tan−1(y/x), that the straight-line
joining the representative point to the origin subtends with the real axis
is termed the argument of the corresponding complex number, z. This is
written mathematically as arg(z) = tan−1(y/x). It follows from standard
trigonometry that x = r cos θ, and y = r sin θ. Hence, z = r cos θ + i r sin θ.
Complex numbers are often used to represent waves, and wavefunctions.
All such representations depend ultimately on a fundamental mathematical
identity, known as de Moivre’s theorem (see Exercise 12.1), which takes the
form
e i φ ≡ cos φ + i sin φ, (12.7)
where φ is a real number. Incidentally, given that z = r cos θ + i r sin θ =
r [cos θ + i sin θ], where z is a general complex number, r = |z| its modulus,
and θ = arg(z) its argument, it follows from de Moivre’s theorem that any
complex number, z, can be written
z = r e i θ, (12.8)
where A > 0 is the wave amplitude, φ the phase angle, k the wavenumber,
and ω the angular frequency. Consider the complex wavefunction
imaginary
z
r
y
θ
x real
The main advantage of the complex representation, (12.14), over the more
straightforward real representation, (12.9), is that the former enables us to
combine the amplitude, A, and the phase angle, φ, of the wavefunction into
a single complex amplitude, ψ0.
E = K + U, (12.16)
where
p2
K= (12.17)
2m
is the particle’s kinetic energy. Hence,
E ψ = (K + U) ψ (12.18)
∂ψ E
= −i ω ψ0 e i (kx−ωt) = −i ψ, (12.19)
∂t h̄
which can be rearranged to give
∂ψ
E ψ = i h̄ . (12.20)
∂t
220 OSCILLATIONS AND WAVES
∂2ψ p2
2
= −k2 ψ0 e i (kx−ωt) = − 2 ψ, (12.21)
∂x h̄
which can be rearranged to give
p2 h̄2 ∂2ψ
ψ=− . (12.22)
2m 2 m ∂x2
Thus, combining Equations (12.18), (12.20), and (12.22), we obtain
∂ψ h̄2 ∂2ψ
i h̄ =− + U(x) ψ. (12.23)
∂t 2 m ∂x2
This equation, which is known as Schrödinger’s equation—since it was first
formulated by Erwin Schrödinder in 1926—is the fundamental equation of
wave mechanics.
Now, for a massive particle moving in free space (i.e., U = 0), the com-
plex wavefunction (12.15) is a solution of Schrödinger’s equation, (12.23),
provided that
h̄ 2
ω= k . (12.24)
2m
The above expression can be thought of as the dispersion relation (see Sec-
tion 5.1) for matter waves in free space. The associated phase velocity (see
Section 7.2) is
ω h̄ k p
vp = = = , (12.25)
k 2m 2m
where use has been made of (12.3). Note that this phase velocity is only
half the classical velocity, v = p/m, of a massive (non-relativistic) particle.
After many false starts, physicists in the early 20th century came to the
conclusion that the only self-consistent physical interpretation of a particle
wavefunction, which is consistent with experimental observations, is proba-
bilistic in nature. To be more exact, if ψ(x, t) is the complex wavefunction of
a given particle, moving in one-dimension along the x-axis, then the proba-
bility of finding the particle between x and x + dx at time t is
Here, we have made use of the easily demonstrated results (ψ∗ )∗ = ψ and
i∗ = −i, as well as the fact that U is real. Taking the difference between the
above two expressions, we obtain
!
∂ψ ∗ ∂ψ∗ h̄2 ∂2ψ ∗ ∂2ψ∗
i h̄ ψ + ψ =− ψ − ψ , (12.31)
∂t ∂t 2m ∂x2 ∂x2
Integrating in x, we get
Z∞
h̄2 ∂ψ ∗ ∂ψ∗
∞
d
i h̄ |ψ|2 dx = − ψ − ψ . (12.33)
dt −∞ 2 m ∂x ∂x −∞
we obtain Z∞
d
|ψ|2 dx = 0. (12.35)
dt −∞
sarily infinite. For such a wavefunction, |ψ(x, t)|2 dx gives the relative prob-
ability, rather than the absolute probability, of finding the particle between
x and x + dx at time t: i.e., [cf., Equation (12.26)]
then Z∞
1
f̄(k) = f(x) e−i kx dx. (12.40)
2π −∞
Here, f̄(k) is known as the Fourier transform of the function f(x). We can
use Fourier’s theorem to find the k-space function ψ̄(k) which generates any
given x-space wavefunction ψ(x) at a given time.
For instance, suppose that at t = 0 the wavefunction of our particle takes
the form " #
(x − x0) 2
ψ(x, 0) ∝ exp i k0 x − . (12.41)
4 (∆x) 2
Thus, the initial probability distribution for the particle’s x-coordinate is
" #
(x − x0) 2
|ψ(x, 0)| 2 ∝ exp − . (12.42)
2 (∆x) 2
yield the value k0, and very unlikely to yield a value which differs from k0
by more than 3 ∆k. Incidentally, as was previously mentioned in Section 8.1,
a Gaussian is the only mathematical function in x-space which has the same
form as its Fourier transform in k-space.
We have just seen that a wave packet with a Gaussian probability dis-
tribution of characteristic width ∆x in x-space [see Equation (12.42)] is
equivalent to a wave packet with a Gaussian probability distribution of char-
acteristic width ∆k in k-space [see Equation (12.49)], where
1
∆x ∆k = . (12.50)
2
This illustrates an important property of wave packets. Namely, in order
to construct a packet which is highly localized in x-space (i.e., with small
∆x) we need to combine plane waves with a very wide range of different
k-values (i.e., with large ∆k). Conversely, if we only combine plane waves
whose wavenumbers differ by a small amount (i.e., if ∆k is small) then the
resulting wave packet is highly extended in x-space (i.e., ∆x is large).
Now, according to Section 9.1, a wave packet made up of a superposition
of plane waves that is strongly peaked around some central wavenumber k0
propagates at the group velocity,
dω(k0)
vg = , (12.51)
dk
rather than the phase velocity, vp = (ω/k)k0 , assuming that all of the con-
stituent plane waves satisfy a dispersion relation of the form ω = ω(k).
Now, for the case of matter waves, the dispersion relation is (12.24). Thus,
the associated group velocity is
h̄ k0 p
vg = = , (12.52)
m m
where p = h̄ k0. Note that this velocity is identical to the classical velocity
of a (non-relativistic) massive particle. We conclude that the matter wave
dispersion relation (12.24) is perfectly consistent with classical physics, as
long as we recognize that particles must be identified with wave packets
(which propagate at the group velocity) rather than plane waves (which
propagate at the phase velocity).
In Section 9.1, it was also demonstrated that the spatial extent of a wave
packet of initial extent (∆x)0 grows, as the packet evolves in time, like
d2ω(k0) t
∆x ≃ (∆x)0 + , (12.53)
dk2 (∆x)0
Wave Mechanics 227
h̄ t
∆x ≃ (∆x)0 + . (12.54)
m (∆x)0
h̄
∆x ∆p >∼ . (12.56)
2
This is the famous Heisenberg uncertainty principle, first proposed by Werner
Heisenberg in 1927. According to this principle, it is impossible to simulta-
neously measure the position and momentum of a particle (exactly). Indeed,
a good knowledge of the particle’s position implies a poor knowledge of its
momentum, and vice versa. Note that the uncertainty principle is a direct
consequence of representing particles as waves.
It is apparent, from expression (12.54), that a particle wave packet of
initial spatial extent (∆x)0 spreads out in such a manner that its spatial
228 OSCILLATIONS AND WAVES
extent becomes
h̄ t
∆x ∼ (12.57)
m (∆x)0
at large t. It is easily demonstrated that this spreading of the wave packet
is a consequence of the uncertainty principle. Indeed, since the initial un-
certainty in the particle’s position is (∆x)0, it follows that the uncertainty
in its momentum is of order h̄/(∆x)0. This translates to an uncertainty in
velocity of ∆v = h̄/[m (∆x)0]. Thus, if we imagine that part of the wave
packet propagates at v0 + ∆v/2, and another part at v0 − ∆v/2, where v0
is the mean propagation velocity, then it is clear that the wave packet will
spread out as time progresses. Indeed, at large t, we expect the width of the
wave packet to be
h̄ t
∆x ∼ ∆v t ∼ , (12.58)
m (∆x)0
which is identical to Equation (12.57). Evidently, the spreading of a particle
wave packet, as time progresses, should be interpreted as representing an
increase in our uncertainty regarding the particle’s position, rather than an
increase in the spatial extent of the particle itself.
BEFORE
ψ→
x→
AFTER
ψ→
x0 x→
h̄2 d2ψ
− + U(x) ψ = E ψ. (12.62)
2 m dx2
Not surprisingly, the above equation is called the time independent Schröd-
inger equation.
Consider a particle trapped in a one-dimensional square potential well,
of infinite depth, which is such that
0 0≤x≤a
U(x) = . (12.63)
∞ otherwise
the well). Within the well, a particle of definite energy E has a stationary
wavefunction, ψ(x), which satisfies
h̄2 d2ψ
− = E ψ. (12.64)
2 m dx2
The boundary conditions are
This follows because ψ = 0 in the region x < 0 or x > a, and ψ(x) must be
continuous [since a discontinuous wavefunction would generate a singular
term (i.e., the term involving d2ψ/dx2) in the time independent Schrödinger
equation, (12.62), which could not be balanced, even by an infinite poten-
tial].
Let us search for solutions to (12.64) of the form
h̄2 k2
= E. (12.67)
2m
The solution (12.66) automatically satisfies the boundary condition ψ(0) =
0. The second boundary condition, ψ(a) = 0, leads to a quantization of the
wavenumber: i.e.,
π
k=n , (12.68)
a
where n = 1, 2, 3, etc. (Note that a “quantized” quantity is one which
can only take discrete values.) According to (12.67), the energy is also
quantized. In fact, E = En, where
h̄2 π2
En = n2 . (12.69)
2 m a2
Thus the allowed wavefunctions for a particle trapped in a one-dimensional
square potential well of infinite depth are
x E1
2
ψn(x, t) = An sin n π exp −i n t , (12.70)
a h̄
where n is a positive integer, and An a constant. Note that we cannot have
n = 0, since, in this case, we obtain a null wavefunction: i.e., ψ = 0, every-
where. Furthermore, if n takes a negative integer value then it generates
232 OSCILLATIONS AND WAVES
p = h̄ k, (12.76)
∂ψ h̄2 2
i h̄ =− ∇ ψ + U(r) ψ, (12.78)
∂t 2m
where the differential operator
∂2 ∂2 ∂2
∇2 ≡ + + (12.79)
∂x2 ∂y2 ∂z2
h̄2
E= (k 2 + ky2 + kz2). (12.95)
2m x
Thus, it follows from Equations (12.92)–(12.94) that the particle energy is
quantized, and that the allowed energy levels are
h̄2
Enx ,ny ,nz = (n 2 + ny2 + nz2). (12.96)
2 m a2 x
The properly normalized [see Equation (12.81)] stationary wavefunctions
corresponding to these energy levels are
3/2
2 x y z
ψnx ,ny ,nz (x, y, z) = sin nx π sin ny π
sin nz π .
a a a a
(12.97)
As is the case for a particle trapped in a one-dimensional potential well,
the lowest energy level for a particle trapped in a three-dimensional well is
not zero, but rather
E1,1,1 = 3 E1. (12.98)
Here,
h̄2
E1 = . (12.99)
2 m a2
is the ground state (i.e., the lowest energy state) energy in the one-dimension-
al case. Now, it is clear, from (12.96), that distinct permutations of nx, ny,
and nz which do not alter the value of nx2 + ny2 + nz2 also do not alter the
energy. In other words, in three dimensions it is possible for distinct wave-
functions to be associated with the same energy level. In this situation, the
energy level is said to be degenerate. The ground state energy level, 3 E1, is
non-degenerate, since the only combination of (nx, ny, nz) which gives this
energy is (1, 1, 1). However, the next highest energy level, 6 E1, is degener-
ate, since it is obtained when (nx, ny, ny) take the values (2, 1, 1), or (1, 2,
1), or (1, 1, 2). In fact, it is not difficult to see that a non-degenerate energy
level corresponds to a case where the three mode numbers (i.e., nx, ny, and
nz) all have the same value, whereas a three-fold degenerate energy level
corresponds to a case where only two of the mode numbers have the same
value, and, finally, a six-fold degenerate energy level corresponds to a case
where the mode numbers are all different.
Wave Mechanics 237
The solution inside the well (i.e., |x| ≤ a/2) which satisfies the symmetry
constraint ψ(−x) = ψ(x) is
where s
2 m (V + E)
k= , (12.107)
h̄2
and B is a constant. The appropriate matching conditions at the edges of the
well (i.e., x = ±a/2) are that ψ(x) and dψ(x)/dx both be continuous [since
a discontinuity in the wavefunction, or its first derivative, would generate a
singular term in the time independent Schrödinger equation (i.e., the term
involving d2ψ/dx2) which could not be balanced]. The matching conditions
yield
q = k tan(k a/2). (12.108)
Let y = k a/2. It follows that
E = E0 y2 − V, (12.109)
where
2 h̄2
E0 = . (12.110)
m a2
Moreover, Equation (12.108) becomes
p
λ − y2
= tan y, (12.111)
y
with
V
λ= . (12.112)
E0
√
Here, y must lie in the range 0 < y < λ, in order to ensure that E lies in
the range −V < E < 0.
Now, the psolutions of Equation (12.111) correspond to the intersection
of the curve λ − y2/y with the curve tan y. Figure 12.6 shows these two
curves plotted for a particular value of λ. In this case, the curves intersect
twice, indicating the existence of two totally symmetric bound states in the
well. Moreover, it is clear, from the figure, that as λ increases (i.e., as the
well becomes deeper) there are more and more bound states. However, it
is also apparent that there is always at least one totally symmetric bound
state, no matter how small λ becomes (i.e., no matter how shallow the well
Wave Mechanics 239
p
Figure 12.6: The curves tan y (solid) and λ − y2/y (dashed),
√ calculated for
λ = 1.5 π2. The latter curve takes the value 0 when y > λ.
becomes). In the limit λ ≫ 1 (i.e., the limit in which the well is very deep),
the solutions to Equation (12.111) asymptote to the roots of tan y = ∞.
This gives y = (2 n − 1) π/2, where n is a positive integer, or
π
k = (2 n − 1) . (12.113)
a
These solutions are equivalent to the odd-n infinite-depth potential well
solutions specified by Equation (12.68).
For the case of a totally antisymmetric bound state, similar analysis to
the above yields (see Exercise 12.3)
y
−p = tan y. (12.114)
λ − y2
The solutions of this equation
p correspond to the intersection of the curve
tan y with the curve −y/ λ − y2. Figure 12.7 shows these two curves plot-
ted for the same value of λ as that used in Figure 12.6. In this case, the
curves intersect once, indicating the existence of a single totally antisym-
metric bound state in the well. It is, again, clear, from the figure, that as λ
increases (i.e., as the well becomes deeper) there are more and more bound
states. However, it is also apparent that when λ becomes sufficiently small
240 OSCILLATIONS AND WAVES
p
Figure 12.7: The curves tan y (solid) and −y/ λ − y2 (dashed), calculated
for λ = 1.5 π2.
π
k = 2n . (12.115)
a
Pout ≃ 1 − 2 λ (12.116)
π2 1
Pout ≃ (12.117)
4 λ3/2
for a deep well (i.e., λ ≫ 1). It follows that the particle is very likely to be
found outside a shallow well, and there is a small, but finite, probability of
it being found outside a deep well. In fact, the probability of finding the
particle outside the well only goes to zero in the case of an infinitely deep
well (i.e., λ → ∞).
Consider a particle of mass m and energy E > 0 interacting with the simple
potential barrier
V for 0 ≤ x ≤ a
U(x) = , (12.118)
0 otherwise
where V > 0. In the regions to the left and to the right of the barrier, the
stationary wavefunction, ψ(x), satisfies
d2ψ
= −k2 ψ, (12.119)
dx2
where s
2mE
k= . (12.120)
h̄2
Let us adopt the following solution of the above equation to the left of
the barrier (i.e., x < 0):
d2ψ
= q2 ψ, (12.123)
dx2
where s
2 m (V − E)
q= . (12.124)
h̄2
The general solution to Equation (12.123) takes the form
Now, continuity of ψ and dψ/dx at the left edge of the barrier (i.e.,
x = 0) yields
1 + R = A + B, (12.126)
i k (1 − R) = q (A − B). (12.127)
Likewise, continuity of ψ and dψ/dx at the right edge of the barrier (i.e.,
x = a) gives
After considerable algebra (see Exercise 12.5), the above four equations
yield
4 k2 q2
|T | 2 = 1 − |R| 2 = . (12.130)
4 k2 q2 + (k2 + q2) 2 sinh2(q a)
Wave Mechanics 243
Here, sinh x ≡ (1/2) (e x − e−x). The fact that |R| 2 + |T | 2 = 1 ensures that
the probabilities of reflection and transmission sum to unity, as must be the
case, since reflection and transmission are the only possible outcomes for a
particle incident on the barrier. Note that, according to Equation (12.130),
the probability of transmission is not necessarily zero. This means that,
in wave mechanics, there is a finite probability for a particle incident on
a potential barrier, of finite width, to penetrate through the barrier, and
reach the other side, even when the barrier is sufficiently high to completely
reflect the particle according to the laws of classical mechanics. This strange
phenomenon is known as tunneling. For the case of a very high barrier, such
that V ≫ E, the tunneling probability reduces to
4 E −2a/λ
|T | 2 ≃ e , (12.131)
V
q
where λ = h̄2/2 m V is the de Broglie wavelength inside the barrier. Here,
it is assumed that a ≫ λ. Note that, even in the limit in which the barrier is
very high, there is an exponentially small, but nevertheless non-zero, tunnel-
ing probability. Tunneling plays an important role in the physics of α-decay
and electron field emission.
12.14 Exercises
x2 x3
ex = 1+x+ + + ···,
2! 3!
x3 x5 x7
sin x = x− + − + ···,
3! 5! 7!
x2 x4 x6
cos x = 1− + − + ···,
2! 4! 6!
which are valid for complex x, to prove de Moivre’s theorem,
e i θ = cos θ + i sin θ,
where θ is real.
where δ(k) is a Dirac delta function. Use this result to prove Fourier’s theo-
rem: i.e., if Z∞
f(x) = f̄(k) e i k x dk,
−∞
then Z∞
1
f̄(k) = f(x) e−i k x dx.
2π −∞
Pout cos3 y
= ,
Pin sin y (y + sin y cos y)
p
where λ − y2 = y tan y, and λ = V/E0 . Hence, demonstrate that for a
shallow well (i.e., λ ≪ 1) Pout ≃ 1 − 2 λ, whereas for a deep well (i.e., λ ≫ 1)
Pout ≃ (π2 /4)/λ3/2 .
4 k2 q2
|T | 2 = ,
4 k2 q2 + (k2 − q2 ) 2 sin2 (q a)
q q
where k = 2 m E/h̄2 and q = 2 m (E − V)/h̄2 . Demonstrate that the co-
efficient of transmission is unity (i.e., there is no reflection from the barrier)
when q a = n π, where n is positive integer.
7. A He-Ne laser emits radiation of wavelength λ = 633 nm. How many photons
are emitted per second by a laser with a power of 1 mW? What force does
such a laser exert on a body which completely absorbs its radiation?
8. The ionization energy of a hydrogen atom in its ground state is Eion = 13.6 eV.
Calculate the frequency (in Hertz), wavelength, and wavenumber of the elec-
tromagnetic radiation which will just ionize the atom.
10. Show that the de Broglie wavelength of an electron accelerated across a po-
tential difference V is given by
where A and a are positive real constants. For what potential U(x) does
ψ(x, t) satisfy Schrödinger’s equation?
13. Show that the wavefunction of a particle of mass m trapped in a one-dimensio-
nal square potential well of of width a, and infinite depth, returns to its
original form after a quantum revival time T = 4 m a2 /π h̄.
14. Show that the normalization constant for the stationary wavefunction
x y z
ψ(x, y, z) = A sin nx π sin ny π sin nz π
a b c
describing an electron trapped in a three-dimensional rectangular potential
well of dimensions a, b, c, and infinite depth, is A = (8/abc)1/2 . Here, nx ,
ny , and nz are positive integers.
15. An electron of momentum p passes through a slit of width ∆x. Its diffraction
as a wave can be regarded in terms of a change of its momentum ∆p in a
direction parallel to the plane of the slit (the total momentum remaining
constant). Show that the approximate position of the first maximum of the
diffraction pattern is in accordance with Heisenberg’s uncertainty principle.
16. The probability of a particle of mass m penetrating a distance x into a classi-
cally forbidden region is proportional to e−2 α x , where
α2 = 2 m (V − E)/h̄2 .
A Useful Information
sin(−α) = − sin α
cos(−α) = + cos α
tan(−α) = − tan α
sin2 α + cos2 α = 1
sin(α ± β) = sin α cos β ± cos α sin β
cos(α ± β) = cos α cos β ∓ sin α sin β
α+β α−β
sin α + sin β = 2 sin cos
2 2
α+β α−β
sin α − sin β = 2 cos sin
2 2
α+β α−β
cos α + cos β = 2 cos cos
2 2
α+β α−β
cos α − cos β = −2 sin sin
2 2
1
sin α sin β = [cos(α − β) − cos(α + β)]
2
1
cos α cos β = [cos(α − β) + cos(α + β)]
2
1
sin α cos β = [sin(α − β) + sin(α + β)]
2
248 OSCILLATIONS AND WAVES
A.3 Calculus
d n
x = n xn−1
dx
d
sin x = cos x
dx
d
cos x = − sin x
dx
d x
e = ex
dx
d 1
ln x =
dx x
d df dg
(f g) = g+ f
dx dx dx
d f df 1 dg f
= −
dx g dx g dx g2
Z Z
dg df
f dx = fg − g dx
dx dx
n (n − 1) 2
(1 + x)n = 1 + n x + x + ···
2!
x2 x3
ex = 1 + x + + + ···
2! 3!
x2 x3 x4
ln(1 + x) = x − + − + ···
2 3 4
Useful Information 249
x3 x5 x7
sin x = x − + − + ···
3! 5! 7!
x2 x4 x6
cos x = 1 − + − + ···
2! 4! 6!
x3 2 5
tan x = x + + x + ···
3 15
x2 x3
f(a + x) = f(a) + f ′ (a) x + f ′′ (a) + f ′′′ (a) + ···
2! 3!