The University of Queensland

Department of Physics
2005

Lecture notes of the undergraduate course

PHYS3050/7051
ELECTROMAGNETIC THEORY III

Lecturer: Dr. Zbigniew Ficek

Physics Annexe(6): Rm 436
Ph: 3365 2331
email: cek@physics.uq.edu.au

Consultation Hours: Wednesday 2pm 4pm

1

Preface
This lecture notes covers the principal elements of classical electromagnetic
theory embodying Maxwells equations with applications mainly to situations
where electric charge can be treated as a continuous uid. The intention is
to introduce students to the background of classical eld theory and the
applications of the electromagnetic theory to solid state physics, classical
optics, radiation theory and telecommunication.
The goal of this course is to provide a compact logical exposition of the
fundamentals of the electromagnetic theory and the applications to various
areas of physics and engineering. The treatment is quantitative throughout
and an attempt has been made to imbue students with a sound understanding
of the Maxwells equations and with the ability to apply them to modern
problems in physics.
The organization of the lectures is fairly standard and includes vector
analysis, electrostatic, magnetostatic, mathematical techniques in the solution of the Maxwells equations and the Laplaces equation, time varying
elds and applications of the solution of the Maxwells equations. The material on vector analysis gives greater emphasis to the relationship between
elds and their sources.

Contents
1 The Classical Theory of the Electromagnetic Field
9
1.1 Elementary Aspects of Electromagnetism . . . . . . . . . . . . 9
1.2 Macroscopic Charges and Currents . . . . . . . . . . . . . . . 11
2 Mathematical Description of Vector Fields
2.1 Gradient of a Scalar Function . . . . . . . . .
2.2 Divergence Function . . . . . . . . . . . . . .
2.3 Gauss Divergence Theorem . . . . . . . . . .
2.4 The Continuity Equation for Electric Current
2.5 Curl (Rotation) Function . . . . . . . . . . . .
2.6 Stokes Theorem . . . . . . . . . . . . . . . .
2.7 Successive Application of
. . . . . . . . . .
2.8 Electromagnetic Field Equations and Electric
Potential . . . . . . . . . . . . . . . . . . . . .
2.8.1 Maxwells Equations . . . . . . . . . .
2.8.2 Electric Potential . . . . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

14
14
15
16
17
19
20
20

. . . . . . . . . 21
. . . . . . . . . 21
. . . . . . . . . 22

3 The Experimental Basis of the Development of Electromagnetic Theory

3.1 Coulombs Law Force between Static Charges . . . . . . . .
3.2 Derived Result Gauss Law . . . . . . . . . . . . . . . . . . .
3.3 Biot-Savart Law . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Current Element and Charge Element . . . . . . . . . . . . . .
3.5 The Lorentz Force . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Amperes Circuit Law . . . . . . . . . . . . . . . . . . . . . . .
3.7 Faradays Law of Electromagnetic Induction . . . . . . . . . .

27
27
27
34
37
38
38
42

4 Dierential Equations for the EM Field and Maxwells Theory

4.1 Dierential Equations for the EM Field . . . . . . . . . . . . .
4.1.1 Divergence of E . . . . . . . . . . . . . . . . . . . . . .
4.1.2 Curl of E . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.3 Divergence of B . . . . . . . . . . . . . . . . . . . . . .
4.1.4 Curl of B . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Maxwells Equations and Prediction of Electromagnetic Waves

45
45
46
46
47
47
51

4.3

4.2.1 The Wave Equation for EM Waves in Vacuum

4.2.2 Plane Wave Solution to the Wave Equation .
4.2.3 Harmonic Waves . . . . . . . . . . . . . . . .
The Transverse Nature of Plane Waves in Vacuum . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

5 EM Theory and Einsteins Special Theory of Relativity

5.1 Lorentz Transformation Equations for Space and Time . . .
5.2 Force Transformation Equations . . . . . . . . . . . . . . . .
5.2.1 The Force between Two Charges Moving with Constant Velocities . . . . . . . . . . . . . . . . . . . . .
5.3 Electric and Magnetic Field Lines of a Moving Charge . . .
5.3.1 Electric Field Lines . . . . . . . . . . . . . . . . . . .
5.3.2 Magnetic Field Lines . . . . . . . . . . . . . . . . . .
5.4 Field equations for an elementary point charge moving with
uniform velocity . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.1 Electromagnetic Induction . . . . . . . . . . . . . . .
6 Energy and Momentum in the Electromagnetic Field
6.1 The Energy Conservation Theorem Poyntings Theorem
6.1.1 Ohmic Heating . . . . . . . . . . . . . . . . . . . .
6.1.2 Electrostatic Field Energy Density . . . . . . . . .
6.1.3 Magnetostatic Field Energy Density . . . . . . . . .
6.1.4 No Fluxes from Static Fields . . . . . . . . . . . . .
6.2 Phase Relationships in Harmonic Waves . . . . . . . . . .
6.3 Momentum Flux . . . . . . . . . . . . . . . . . . . . . . .
6.4 Electromagnetic Energy Flow . . . . . . . . . . . . . . . .
6.4.1 Energy Flow into a Resistive Wire . . . . . . . . .
6.4.2 Energy Flow out of Battery . . . . . . . . . . . . .
6.4.3 Propagation of Energy along a Wire . . . . . . . .
7 General Solution of the Maxwells Equations
7.1 Diculty of the Direct Solution of Maxwells Equations
Time Varying Sources . . . . . . . . . . . . . . . . . .
7.2 Scalar and Vector Potentials . . . . . . . . . . . . . . .
7.2.1 Lorentz Gauge . . . . . . . . . . . . . . . . . .
7.2.2 Coulomb Gauge . . . . . . . . . . . . . . . . . .
7.3 Solution of the Inhomogeneous Wave Equations . . . .
4

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.

51
52
54
55

59
. 59
. 61
.
.
.
.

62
66
66
67

. 68
. 74
.
.
.
.
.
.
.
.
.
.
.

78
78
80
81
81
83
84
84
85
85
87
88
90

with
. . .
. . .
. . .
. . .
. . .

.
.
.
.
.

90
92
94
95
97

7.4

Alternative Solution: Green Functions Method . . . . . . . . . 100

8 Solution of Laplaces Equation and Boundary Value Problem104

8.1 Uniqueness of the Solution of Laplaces Equation . . . . . . . 105
8.1.1 Dirichlet theorem . . . . . . . . . . . . . . . . . . . . . 105
8.2 Solutions of Laplaces Equation . . . . . . . . . . . . . . . . . 106
8.2.1 Method of Separation of Variables . . . . . . . . . . . . 107
8.2.2 Solution of the Laplace Equation in Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
9 Electromagnetic Antennas: Hertzian Dipole
9.1 Field of an Element of Alternating Current . . . . . . . . . .
9.2 Power Radiated from the Current Element . . . . . . . . . .
9.3 Gain of the Dipole Antenna . . . . . . . . . . . . . . . . . .
10 Electromagnetic Theory of Polarizable
Materials
10.1 Potential and Electric Field of a Single Dipole . . . . . . . .
10.2 Polarization Vector . . . . . . . . . . . . . . . . . . . . . . .
10.3 Maxwells Equation for E in a Dielectric . . . . . . . . .
10.4 Dense Dielectrics: The Clausius-Mossotti Relation . . . . . .
10.5 Time Dependent Fields and the Complex Dielectric Susceptibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.6 The Complex Susceptibility and Permittivity . . . . . . . . .
10.7 Added Note - The Loss Tangent . . . . . . . . . . . . . . . .
11 Magnetic Fields in Magnetizable
Materials
11.1 Magnetic Polarization Currents . . . . . . . . . . . . . . .
11.2 The Magnetic Intensity Vector H . . . . . . . . . . . . . .
11.3 Linear Isotropic Magnetic Materials . . . . . . . . . . . . .
11.4 The Magnetization of Iron . . . . . . . . . . . . . . . . . .
11.5 Time Dependent Magnetic Fields and Energy Loss . . . .
11.6 The Ferromagnet . . . . . . . . . . . . . . . . . . . . . . .
11.7 Maxwells Equations in Dielectric and Magnetic Materials

.
.
.
.
.
.
.

125
. 126
. 133
. 135

.
.
.
.

138
138
140
141
146

. 147
. 151
. 154

.
.
.
.
.
.
.

156
156
162
164
167
168
170
173

12 Poyntings Theorem Revisited

175
12.1 Poynting Vector in Terms of E and H . . . . . . . . . . . . . 175
12.2 Poynting Vector for Complex Sinusoidal Fields . . . . . . . . . 177
13 Plane Wave Propagation in Dielectric and Magnetic Media 181
13.1 Wave Refraction and Skin Eect . . . . . . . . . . . . . . . . . 185
14 Transitions Across Boundaries for Electromagnetic Fields
14.1 Applications in dielectrics . . . . . . . . . . . . . . . . . . .
14.1.1 Normal Component of B . . . . . . . . . . . . . . . .
14.1.2 Normal Component of H . . . . . . . . . . . . . . . .
14.1.3 Normal Component of E . . . . . . . . . . . . . . . .
14.1.4 Tangential Component of E . . . . . . . . . . . . . .
14.1.5 Tangential Component of H . . . . . . . . . . . . . .
14.1.6 Tangential component of B . . . . . . . . . . . . . .

.
.
.
.
.
.
.

15 Reection and Transmission of Waves

Across a Boundary
15.1 Representation of Plane Waves in Dierent Directions
15.1.1 Representation of B in terms of E . . . . . . .
15.2 Directions of Reected and Transmitted Waves . . .
15.3 Snells Law of Refraction and Angle of Reection . .

199
. 199
. 200
. 201
. 206

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

191
191
191
193
193
193
196
196

16 Fresnels Equations
208
16.1 Ei normal to plane of incidence . . . . . . . . . . . . . . . . . 209
16.2 Ei in the plane of incidence . . . . . . . . . . . . . . . . . . . 211
16.3 Fresnel Equations for dielectric media . . . . . . . . . . . . . . 212
17 Applications of the Boundary Conditions and the
Equations
17.1 Applications in dielectrics . . . . . . . . . . . . . . .
17.1.1 Polarization by reection . . . . . . . . . . . .
17.1.2 Total internal reection . . . . . . . . . . . . .
17.2 Transmission and Reection at a Conducting
Surface . . . . . . . . . . . . . . . . . . . . . . . . . .
17.2.1 Field vectors at normal incidence . . . . . . .

Fresnel
214
. . . . . 214
. . . . . 214
. . . . . 216
. . . . . 217
. . . . . 219

17.3 Wave Propagation in a Conducting Rectangular Pipe (The

Rectangular Waveguide) . . . . . . . . . . . . . . . . . . . .
17.3.1 Transverse Electric (TE) Modes . . . . . . . . . . . .
17.3.2 Boundary Conditions . . . . . . . . . . . . . . . . . .
17.3.3 TE Modes in a Lossless Waveguide . . . . . . . . . .
17.3.4 Phase and Group Velocities for TE and TM Modes .
18 Relativistic Transformation of the Electromagnetic Field
18.1 The Principle of Relativity . . . . . . . . . . . . . . . . . . .
18.2 Transformation of Electric and Magnetic Field Components
18.3 Transformation Rules in Terms of Parallel and Normal Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.3.1 Rules for Parallel Components . . . . . . . . . . . . .
18.3.2 Rules for Normal Components . . . . . . . . . . . . .
18.4 Transformation of the Components of a Plane EM Wave . .
18.5 Doppler Eect . . . . . . . . . . . . . . . . . . . . . . . . . .
18.6 Transformation of Energy of a Plane EM Wave . . . . . . .

.
.
.
.
.

221
222
225
226
230

233
. 233
. 239
.
.
.
.
.
.

240
241
241
243
244
246

Literature
1. J.D. Jackson, Classical Electrodynamics, 3rd ed. Wiley 1999.
2. R.K. Wangsness, Electromagnetic Fields, 2nd ed. Wiley 1986.
3. R. Plonsey and R.E. Collin, Principles and Applications of Electromagnetic Fields, McGraw Hill 1961.
4. B.I. Bleaney and B. Bleaney, Electricity and Magnetism, 3rd ed. Oxford
U.P. 1983.
5. J.A. Stratton, Electromagnetic Theory, McGraw Hill, 1941.
6. W.K.H. Panofsky and M. Phillips, Classical Electricity and Magnetism,
2nd ed. Addison-Wesley 1962.
7. R.P. Feynman, R.B. Leighton, and M. Sands, The Feynaman Lectures
on Physics, Vol. 2, Addison Wesley, 1964.

The Classical Theory of the Electromagnetic Field

Classical theory of the electromagnetic eld or Classical Electrodynamics,

formulated by Maxwell more than 100 years ago, is now a well established
theory. In this context classical means non-quantum, but we would
like to point out that the basic equations of electromagnetism, the Maxwells
equations, hold equally in quantum and classical eld theory.
Electromagnetic interactions are ONE of FOUR fundamental types:
Type of interaction
Strong interaction (nuclear)
Electromagnetic
Weak interaction (e.g. decay)
Gravitation

Relative Strength
1
102
1012
1040

At our level of discussion there is no relation between these four types of

interaction (i.e. they cannot be considered as dierent manifestations of a
single FORCE).
Classical EM Theory is particularly interesting because of:
Its historical role in the development of Physics in the late 19th and
early 20th centuries.
Numerous applications on our scale of existence:
Chemical bonding is due to EM forces.
Communications via EM Waves (radio, TV, telephony, computers).
Generation and transmission of electrical power.
Knowledge of the rest of the universe from reception of EM waves.

1.1

Elementary Aspects of Electromagnetism

EM forces are due to ELECTRIC CHARGE which is NOT in turn

explicable in terms of anything else.
9

 Charges are of two kinds called positive and negative. In the static
limit like charges repel and unlike attract.
Charges are quantized in units of e

1.6 1019 [Coulombs].

In the static limit the inverse square (Coulomb) law of force holds:
Fq 2 =

1 q2 q1
r,

40 r 2

q1 q2 ,

that the charge q1 acts on the charge q2 with the force Fq2 . The parameter 0
determines the property of the medium and is called the electric permittivity.
The Coulomb law holds only for charges whose the spatial dimensions are small compared with the distance separating them.
The Coulomb law can be tested to great accuracy indirectly by showing
that no charge rests on the inside of a statically charged hollow conductor1 .
If the exponent in the Coulombs law
1
rn

is not n = 2 but n = 2(1 ), the potential inside the hollow conductor

would be large. Since the potential inside the hollow conductor found by the
experiments was less in magnitude than a small detectable potential, then
109 , the level of sensitivity of the detector.
Electric charge is conserved (algebraically)
q = constant
whole universe

In electromagnetic theory none of these things are explained in terms

of anything else.
1

S.J. Plimpton and W.E. Lawton, Phys. Rev. 50, 1066 (1936).

10

 In the NON-STATIC case (moving charges) the force is no longer given

by Coulombs Law. In general is given by the Lorentz equation:
Fq2 = q2 (E + v B) ,
where E and B are the electric and magnetic elds due to q1 .
The elds E and B will depend on the frame of reference of the observer
(F must follow the required relativistic transformation law).
However, we do not think that q depends on the frame of reference.
i.e. q does not depend on its velocity with respect to an observer.
This is because in ordinary matter electrons move much faster than ions,
their speeds depend on temperature, and electric elds are not observed to
arise from changes in temperature.
Why there must be a B and how E and B are computed for arbitrary
motion of charges is the substance of electromagnetic theory.
THE BASIC IDEA OF ELECTROMAGNETISM IS:

CHARGE 1 =

ELECT ROM AGN ET IC

F IELD

F ORCE ON

= CHARGE 2

Important conclusion
Fields are generated by charges - NOT by other elds.

1.2

Macroscopic Charges and Currents

We know that electric charge is quantized. The electron is a point charge on

the smallest scale measurable. We may then speak, on a subatomic scale, of
a microscopic theory of electromagnetism. On a subatomic scale there must
11

be very strong and rapidly varying electric and magnetic elds on spatial
scales 108 m and temporal scales 1010 s.
When we measure the elds around a macroscopic circuit, clearly we are
not looking at these elds. We are measuring elds on distance scales
108
10
m and time scales
10
s. The microscopic elds sum to (almost) zero. In
the macroscopic context it is convenient and justiable to regard the charge
as a continuous uid.
Charge density
When we encounter a large number of point charges in a nite volume, it is
convenient to describe the source in terms of a charge density, dened as
= lim

q
,
V

where q is the algebraic sum of the charge in the volume V .

The limit is not to zero but to a V
atomic scale size, which is still
very small on the laboratory scale.
If the charge density is represented by a continuous function , the total
charge Q in a volume V is given by
Q=

dV .

Current density
For many purposes it is necessary to introduce the idea of current density.

vdt
IA

^
n
v

12

Let IA = q/t is a current through the area A. Then the current density
is dened by
J = lim

I
q
A v t
n = lim

n = lim

nJ = v ,

A
t A
t A

where v = v .
n
The limit is taken in the same sense as for .
Total current through a surface area
If the current density J is known at every point of an arbitrary surface,
we can obtain the total current through the area.

^
n

dA

The current through A is:

I = J cos()A = J n A .

Then the current through the total area A is the sum of the contributions
from all elements of the area:
IA =
A

J n dA =

J dA ,

where dA is a vector representing the element dA of the surface A. In vector analysis it is common to represent a surface by a vector whose length
corresponds to the magnitude of the surface area and whose the direction is
specied by the unit vector n normal to the surface.

In summary, when EM Theory is formulated in terms of and J as the

sources, we speak of a MACROSCOPIC THEORY.
13

Mathematical Description of Vector Fields

The study of electromagnetic theory requires considerable knowledge of vector analysis. In this lecture, we will introduce vector operations we will need
for our study of electromagnetic theory. As we shall see, based on this lecture, it is possible to considerably simplify the formulation of electromagnetic
theory.

2.1

Gradient of a Scalar Function

Let us suppose that represents a scalar eld and that is a single valued,
continuous, and dierentiable function of position.
The gradient of the scalar function is dened as:
grad =

n,

where n is a unit vector in the direction the rate has its maximum value. In

s
other words, gradient tells us in which direction the change in is maximal.

^
n
ds

^
i
dx

+d

14

For some other direction dX:

d =

dX =

n dX =

cos dX
s
s

In rectangular (cartesian) coordinates:

=
i
ni =

cos
s
s

= lim
=
= lim
(s/ cos )
x
x

( )x =

Hence
=

k
i+
j+
x
y
z

in cartesian coordinates.
The gradient is analogous to multiplication of a vector by a scalar. The
result, of course, is a vector. We do not usually take a gradient of a vector,
the result would be a tensor.
Example
Consider a scalar function = xyz. Gradient of = xyz is
y
z
x
i + xz
j + xy
k
x
y
z

= yz + xz + xy k .
i
j

= yz

2.2

Divergence Function

The divergence is the scalar which results from operation of

upon a vector F in a fashion analogous to the dot product of two vectors. In Cartesian
coordinates:
div F =

F =

Fx Fy Fz
+
+
x
y
z

Where there is a positive divergence, there is a source of a vector eld.

When F = 0 everywhere, the eld F is called solenoidal.
15

Example
Consider a vector F = x
i.
div F =

x
x
x
(i i) +
(j i) +
(k i) = 1 .
x
y
z

Consider now a vector F = y In this case

i.
div F =

y
y
y
(i i) +
(j i) +
(k i) = 0 .
x
y
z

Thus, divergence of a given eld is dierent from zero only if the eld amplitude changes in the direction of the eld. So the divergence is related to
how the eld changes as you move in the direction of the eld.

2.3

Gauss Divergence Theorem

Gauss law, or sometimes called as the divergence theorem, is stated as

F dV =

volume integral

F n dS

closed surface integral

F
^
n
V

dS

Remember that n is the unit outward normal over S.

16

Denition of Flux
The FLUX of F through a surface S, not necessary a closed S, is dened as:

F dS =

Associated with this is the pictorial representation of elds by lines of

force, the direction of F given by the tangent to a line and the strength of F
given by the line density per unit area.
Note that from the divergence theorem:
F = lim

dV 0

1
dV

F ndS = lim

dV 0

dV

i.e. the divergence of a vector eld is the emanating ux per unit volume.

S
F

A non-zero F is then implies a source (if positive) or a sink (if negative). And if F = 0 there is no source or sink the eld lines have no
beginnings or ends.

2.4

The Continuity Equation for Electric Current

Suppose we have some substance of density in a volume V enclosed by a

surface S, as shown in Fig. 1.
17

Let v = macroscopic velocity of substance .

Let q = rate of production of substance per unit volume .
Let l = rate of annihilation of substance per unit volume .

S
^
n
dV
V
v
Figure 1:
Then, the rate of increase of total substance in the volume V
=Rate of production - rate of annihilation - rate of transport out through S.
We can express this as

dV =
V

q dV

l dV

v n dS .

Using the Gauss divergence theorem, we can write

v n dS =

(v) dV .

Thus

dV =
t

q dV

l dV
18

(v) dV

Since this relation holds for arbitrary V :

=ql
t

(v) .

For a conserved quantity such as electric charge: q = l = 0, and then

v = J .
Hence

+
t

J =0 ,

which is well known as the continuity equation.

When stationary currents are involved, then /t = 0. In this case
J = 0, that is for stationary currents the current density is solenoidal.

2.5

Curl (Rotation) Function

Curl (or rotation) is a vector which results from the operation of operator
upon a vector in a fashion analogous the cross product of two vectors
curl F

F =
i

F F F
+j
+k
,
x
y
z

or
F =
=

Fx Fz
Fy Fx
Fz Fy

i+

j+

k
y
z
z
x
x
y
k
i
j

.
x
y
z
Fx Fy Fz

Curl is nonzero when the eld increases (or decreases) in a dierent direction that the eld pointed. If the eld is pointed in the same direction as
that in which is increased, the curl is zero. So the curl is related to how the
eld changes as you move across the eld.
When

F = 0 everywhere, the eld F is called irrotational.

19

2.6

Stokes Theorem

Stokes theorem may be stated in the form

F dl =

complete loop

F ndS

surface not closed

F
dl
^
n

dS

curl F
However
F n = lim

dS0

1
dS

F dl .

This gives an intuitive meaning to any component of F in terms of

the line integral around a small element of surface.
F is a measure of
the vorticity of the eld.

2.7

Successive Application of

We can introduce scalar and vector products in which the operator

appears more than once. For example, since the gradient of an arbitrary scalar
function is a vector, we can take the divergence of the gradient

V =

V V V
2V
2V
2V
i+
j+
k =
+
+
.
x
y
z
x2
y 2
z 2
20

The same result is obtained if we take

properties
2

The operator

as a new operator

with

2
2
2
+ 2+ 2 .
x2 y
z

is called the Laplacian and is a scalar.

The Laplacian may also be applied to a vector, with the result

2

F =

2F
2F
2F
+
+
.
x2
y 2
z 2

It is also possible to form the curl of the gradient, which is identically zero

V =0.

The divergence of the curl of a vector is also identically zero

F =0 .

The proof of the above properties is left to the students.

2.8

Electromagnetic Field Equations and Electric

Potential

We will now illustrate some properties of the successive application of ,

which will allow us to introduce the concept of vector and scalar potentials
to the electromagnetic eld theory.
2.8.1

Maxwells Equations

Consider coupled dierential equations for vector elds E and B, the Maxwells
equations:

,
(1)
E =
0
B = 0 ,
(2)
E =

B
,
t

B = 0 J +
21

1 E
,
c2 t

(3)
(4)

By means of divergence theorem and Stokess theorem, they can be easily

transferred into an integral form

E ndS =

B ndS = 0 ,

q
,
0

E dl =

B ndS = 0

B ndS ,

J ndS +

1
c2 t

E ndS .

According to the Helmholtz Theorem, an arbitrary vector F can always be

written as
1
4
V
= Fl + Ft .

F =
Thus, specication of
mine F .

F and

1
F
dV +
r
4

F
dV
r

F is necessary and sucient to deter-

Hence, we need four equations of this type to determine E and B.

In the following lectures, we will discuss how Maxwell arrived at these equations. We will also discuss the eect of Einsteins special theory of relativity
on how we think about electromagnetic elds and Maxwells equations. Much
of our discussion will be about how to solve Maxwells equations.

2.8.2

Electric Potential

The solutions of the electromagnetic eld equations are not in general simple and straightforward. Often their solution is aided by the use of potentials.
A potential is a quantity from which a vector eld can be derived by some
process of dierentiation.
22

Examples
(1) Let
F = 0. Since

will have an integral of the form:

0 then the equation

F =

F = 0

and is called a scalar potential. Thus, the eld may be derived from the
gradient of the scalar potential function .
In the electrostatic eld
E = 0. Hence, E = . 2 Then the
work done per unit charge q moving from point A to point B is:
W
q

=
B

=
A

E dl =

r dl =

dl

A
B

d = B A = AB .

dl
A
2

The minus sign is inserted in the denition of to agree with the denition of as
a potential ENERGY. The negative sign can also be understood physically from the fact
that E is in the direction that a positive charge moves, hence in the direction of decreasing
potential.

23

Note that if

E = 0 then:
E dl =

E ndS = 0 .

Hence
(AB)1 = (BA)2 = (AB)2 .

1 (AB)
B
A

2 (BA)
The work done is independent of the path chosen. Thus a eld F with
F = 0 is a conservative eld of force.
(2) Let
F = 0. Since
have an integral of the form:

F =

A 0 the equation

F = 0 will

and A is called a vector potential.

Note:
It turns out that
B = 0 always in electromagnetism so there will
always be a vector eld A such that B = A and such an A is referred
to as THE vector potential in electromagnetism (though there may be other
24

electromagnetic eld functions with zero divergence).

Note also that just writing A = B does not completely specify A even
if B is known everywhere. One needs to specify A as well to completely
determine A. Equivalently we can say that dening A = B still leaves
us free to dene A.
On the other hand E = 0 in general in electromagnetism so it is not
in general possible to write E = .
Exercise in class: Vector analysis
For a given scalar function and vectors A, B, C, indicate successive steps
you would follow in the calculation of the following expressions:
(i)

A ;

ABC ;

B;

B ;

AB ;

AB

B ;
;

ABC .

(ii) Which of the expressions in (i) are vectors?

(iii) Which of the four statements is correct:
1.

A = (

2.

A =

3.

A =

4.

A =

25

)A ,

( A) ,
(

A) ,

A) .

A
A

;
BC ;

Weekend exercises
(a) For elds of the form r nr , (r = 0), nd for which values of n the

divergence is zero.

(b) For elds of the cylindrical form n , ( = 0), nd for which values
of n the curl is zero.
(c) If the potential satises the equation (Laplace equation)
show that is both solenoidal and irrotational.

26

= 0,

The Experimental Basis of the Development of Electromagnetic Theory

In this lecture, we will present the basic properties of the electrostatic eld
in vacuum. The Coulomb law for the force between two point charges is the
experimental basis for the development of electromagnetic theory.

3.1

Coulombs Law Force between Static Charges

In 1785, Coulomb investigated the nature of the force between charged bodies, and the results of his experiments can be formulated mathematically in
what is known as Coulombs law
F2 =

1 q1 q2
r.

40 r 2

We can write Coulombs law as

F2 = q 2 E 1 .
Hence
E1 =

1 q1
r.

40 r 2

The electric eld is an example of a vector eld. In principle, we can always calculate an electric eld using Coulombs law. However, there is an
alternative way we can nd the electric eld. In particular, the eld may
be represented by means of the ux concept. The total ux of E from a
point charge q1 may be readily calculated by integrating E dS over a surface
enclosing q1 .

3.2

Derived Result Gauss Law

Consider a macroscopic charge q1 closed by a surface S. We will show that

the ux of the electric eld produced by the charge q1 is proportional to the
charge q1 , and is independent of the shape of the surface closing the charge.
27

S2
S1
^
r

r0
^
n

Consider rst the ux through a spherical surface:

S1 =
S1

E ndS =

q1
=
2
40 r0

S1

dS =

q1
r ndS

2
40 r0
q1
q1
2
4r0 =
.
2
40 r0
0

Next, consider the ux through an arbitrary surface

S2 =

dS2 =
S2

S2

q1
r ndS ,

40 r 2

where S2 is an arbitrary surface enclosing q1 .

But from the inverse square law and some geometry:
q1
r ndS

40 r 2
q1 dS cos
q1
=
d ,
=
2
40
r
40

dS2 =

where d is the solid angle subtended by dS at q1 .

The element of solid angle d is independent of where we cut the bundle
of electric lines of force.
Thus
q1
.
dS2 = dS1 and
dS2 = dS1 =
0
28

Furthermore for some arbitrary number of charges q1 , q2 , q3 , , qn within

a surface S:

S
q1
q2 q3
q4

d =

(E1 + E2 + E3 + ) n dS

= 1 + 2 + 3 +
1
q1 q2 q3
+ +
+ =
=
0 0 0
0
where

q,

q is the algebraic sum of all charges within the surface S.

If q is outside S, the surface integral vanishes since the total solid angle
subtended at q by the surface is zero. Thus
d1 + d2 = 0 .
Proof:
q r n1 dS1

2
40
r1
dS1
= 2 = d .
r1

d1 =
r n1 dS1

2
r1
Similarly
d2 =

q
d = d1 .
40
29

^2
n
dS 1

dS 2

^
n

^
r

Since
r n2 dS2

= +d ,
2
r2
we nd that d1 + d2 = 0 and integrating over all S:
=

d = 0 .
S

The physical interpretation of this result is that eld lines originating from
an external charge and entering the surface S must also leave this surface.
In summary, the Gauss law says that the total electric ux through a closed
surface S is:
=

q
,
0

where q =

charges INSIDE S .

Using the denition of the ux, we often write the Gauss law as
S

E dS =

q
.
0

The power of the Gauss law lies in the fact that we are free to apply
it to any closed surface whose shape can be chosen arbitrary such that the
evaluation of the surface integral becomes a simple straightforward task. The
Gauss law is particularly useful in simplifying the calculation of the electric
30

eld produced by certain symmetrical charge distributions. We illustrate this

in the following examples.
Example of an application of the Gauss Law
An innitely long line is positively and uniformly charged with a constant
linear charge density l . Use (a) Coulomb law, (b) Gauss Law to nd the
electric eld about the line.
(a) If we have to calculate the eld due to a static macroscopic distribution of charge using the Coulomb law, we divide the macroscopic charge into
innitesimal (point) charges dq which produce an electric eld dE. The eld
dE is given by the Coulomb eld
dE =

1 dq
r.

40 r 2

Then the total electric eld is found by vector addition

E=

dE .

Since we are adding vectors, a caution must be employed. We use the following procedure, which is general and can be employed to any system:
1. Write the expression for the electric eld dE produced by the innitesimal (point) charge dq.
2. Resolve this vector into components dEx ,dEy , and dEz .
3. Calculate each component of E by integration, e.g. Ex =

dEx .

4. Find the resultant E from its components

E = Ex + Ey + Ez k .
i
j
Return now to our example of the charged innitely long line.
Take a small element dl of the line containing a point charge dq. Electric
31

dE

r
h

dl

eld produced by the point charge at A distance r from dl is given by the

Coulomb eld

dE =

1 dq
r.

40 r 2

We see from the gure that

h
,
sin

r=

l = h cot .

Hence
dl =

h
d ,
sin2

dq = l dl ,

and then
sin2
l
h
2
d cos + sin
i
j
40
h2
sin
l
=
cos + sin d ,
i
j
40 h

dE =

where we have decomposed the unit vector r into two (x, y) components

r = cos + sin .

i
j
32

Integrating over from = 0 to = , as the line is innite, (at x =

, = 0 and at x = , = ), we obtain
l
( sin 0 + sin ) + (cos 0 cos )
i
j
40 h
l
j.
=
20 h

E =

Thus, the electric eld produced by the charged line depends inversely on
the distance from the line and points in the direction perpendicular to the
line.
(b) Let us now calculate the eld by the direct application of the Gausss
law.

E
dS

dS

dS
A

The electric eld near the uniformly charged line must be radially directed
because of the symmetry of the problem. The eld must have cylindrical
symmetry because the problem is unchanged by rotating the line about its
axis. The eld must also be independent of position along the line because
the distance to either end is innite. This is the ideal situation for the
application of Gausss Law. We can apply a cylinder surface of radius h and
length L centered about the line of charge, see the Figure above.
According to the Gausss law
S

E dS =

q
,
0

where q = l L is the charge closed by the cylinder surface.

33

The ux through the cylinder surface splits into three uxes

S

E dS =

E dS A +

E dS B +

E dS C .

Since E dS A , E dS C , E dS B , and the magnitude of E is constant

along the surface B, the ux through the cylinder reduces to
S

E dS =

EdSB = 2hLE ,

i.e. a nonzero ux exists only through the surface B, and since the cylinder
is symmetrically positioned about the line of charge, the magnitude of E is
constant over the surface B.
Then, according to the Gausss law
2hLE =

l L
,
0

which gives
E=

l
.
20 h

Note how simple are the calculations of the electric eld using the Gausss
law. However, we were able to solve this problem because we knew the direction of the eld at any point around the line.

3.3

Biot-Savart Law

In 1819, H.C. Oersted showed the magnetic eects of electric current.

In 1827, Ampere showed that quantitatively the magnetic forces in macroscopic circuits can be accounted for by what has come to be known as the
Biot-Savart Law.
Experiment

34

I2

I1

Two long parallel wires. Force between the wires:

F =

0 I 1 I 2
,
4 d

where 0 = 4 107 [H/m] in SI units, is the permeability of the vacuum.

(This denes the SI unit of current, the Ampere.)
If I1 I2 then the force F is attractive.
If I1 anti I2 then F is repulsive.
If one wire is rotated through 90 then F = 0.
All such observations are explained by the law between current elements
being:
d F2 =

0 I 2 d
4

(I1 d
r2

r)

We can write the force as

d F2 = I 2 d

dB ,

where
dB =

0 I 1 d 1 r
,
2
4
r
35

which is known as the Biot-Savart law for magnetic eld produced by the
current element I1 d 1 .
The Biot-Savart law allows to compute magnetic eld produced by an
arbitrary current distribution Id
B=

0 I
4

d r

.
r2

(5)

The method requires integration of small current elements.

We can simplify the calculations of B by using the following procedure.
If we replace r /r 2 by (1/r), the integrand becomes

(1/r) .

However, using a vector identity that

1
d
=
r
r
since

1
= d
r

d d

1
r

d = 0, we can write
B=

0 I
4

d
.
r

(6)

Since we can write

B=
as

A ,

B = 0 always, we can rst calculate A:

A=

0 I
4

d
r

(7)

and then using (6), we will nd B.

The integral for A is easier to calculate than the original expression (5)
for B. Since the curl operation is readily performed, we may use (7) as an
intermediate step for nding B in a simpler way.
The vector A is called a vector potential, and will see later in the course
many useful applications of A in electromagnetic theory.
36

3.4

Current Element and Charge Element

A charge dq moving with a velocity v is equivalent to an element of current Id :

dq = Idt = I

d
,
v

where dt = the time for all the charge in d to pass out of the volume.

I
v
dl
Hence
dqv = Id .
The Biot-Savart Law can then be written as
d F2 =

0
(dq1 v1 r)

dq2 v2
,
2
4
r

dF2 = dq2 v2 dB1 ,

with
dB1 =

v1 r

0
dq1 2 ,
4
r

which shows that magnetic eld is produced by a moving electric charge.

37

3.5

The Lorentz Force

The Lorentz force is obtained putting Coulombs Law and the Biot-Savart
Law together: FE = q E and FM = qv B:
FEM = FE + FM = q(E + v B) .
Thus, a motion of electric charges is modied by both the electric and magnetic forces. If the charge is stationary, the force depends only on E, if it
moves, there is an additional force proportional to v.

3.6

Amperes Circuit Law

Amperes circuit law is a useful relation between currents and magnetic elds.
This law allows us to calculate magnetic eld produced by some currents in
a very eective way.

dl

The Amperes law says that for an arbitrary closed path around a current
carrying conductor, the component of magnetic eld tangent to the path is
proportional to the net current passing through the surface bounded by the
path
B d = 0 I
closed loop

I = total current through loop

38

Proof:
Consider a long wire of radius a carrying current I. Let P is a point on
the integration path, see Figure 2.
The magnetic eld at P is:

P
^
r

dl

-dl

ds

Figure 2: The source circuit and the integration path to prove the Ampere
law.

B=

0 I
4

ds ()
r
.
r2

Moving P by d is equivalent to moving the current circuit by d .

The solid angle subtended by d , ds at P is:
(d ds) r
d ds r

ds
r
=
=d
.
2
2
2
r
r
r
(The element of area normal to r is d s r )

39

Thus due to the path element d , the change in solid angle subtended
at P by the circuit is:
d = d

ds ()
r
.
r2

Hence
d = d

4
0 I

dB =
s

4
Bd ,
0 I

where integration is around the circuit s giving the magnetic eld B at some
point P as shown in the diagram.
Now integrating round the closed path:
Bd =

0
I
4

d .

If P moves round a closed path (returning to its original position but not
circulating through the current loop:
d = 0 .
But if P circulates through the loop:
d = 4 ,
and then
B d =

0
I 4 = 0 I .
4

We conclude that the line integral of the magnetic eld round a closed
loop path is equal to 0 I where I is the current passing through the loop.
The Amperes law can be applied in highly symmetric situations to nd
the magnetic eld more easily than by computing with the Biot-Savart law.
40

In either case, the result is the same. In case that lack the proper symmetry,
Amperes law is not easily applied.
Example of an application of the Amperes Law
An innitely long wire caries a constant current I. Use (a) Biot-Savart
law, (b) Amperes law to nd the magnetic eld about the wire.

r
h

dl

dl

(a) The element of magnetic eld dB due to an element I d at distance

r = r is found from the Biot-Savart formula:
r
dB =

0
Id 2
4
r

where we note that all dB are in the same direction normal to the direction
of the current. So we see from this symmetry that the eld lines are circles
concentric with the current. Furthermore, along any such circular path the
eld is constant in magnitude.
Let us calculate the magnitude of the magnetic eld. Since

d r = dl sin ,

r=

h
,
sin

we have
dl =

h
d
sin2
41

l = h cot ,

and then
dB =

h
0 I sin2
0 I

2
sin d .
sin d =
2
4 h
4h
sin

Integrating the above equation over the length of the wire, we obtain
0 I
0 I (cos 0 cos )

sin d =
4h 0
4h
0 I
.
=
2h

B =

(b) Let us now calculate the eld using the Amperes law.
Since the eld lines are circles concentric with the current, and along
any such circular path the eld is constant in magnitude, this is the ideal
situation for the application of Amperes law:
B d = 0 I 2hB = 0 I B =

0 I
.
2h

Note how simple are the calculations of the magnetic eld using the Amperes law.

3.7

Faradays Law of Electromagnetic Induction

Faraday discovered electromagnetic induction by changing magnetic eld. If

we consider a closed stationary circuit located in a varying magnetic eld,
the induced electromotive force around this circuit is equal to the negative
time rate of change of the magnetic ux through the circuit
E =

d
,
dt

where is the total magnetic ux through the circuit.

From the denition of the ux
=
S

B dS ,
42

dl

and from that emf E = work done per unit charge, we have
E
=
q

(E + v B) d .

In a stationary circuit v = 0 (and anyway v B d since v

v B d = 0.
Hence
E=

d ). Thus

E d

and nally
E d =

B dS ,

or
E d =

B
dS .
t

This is the Faradays law written in the integral form.

Using the Stokess theorem, we can rewrite the Faradays law in the differential form
E =
43

B
.
t

The Faradays law tells us that time-varying magnetic elds give rise to
electric elds. This shows that the elds are related to each other, and we
then must speak of electromagnetic elds, rather than separate electric
and magnetic elds.

Questions:
1. Prove that the total electric ux through a closed surface S is proportional to the total charge inside the surface.
2. Prove the Amperes circuit law.
3. Derive the integral form of the Faradays law and then transform it
into the dierential form
E =

B
.
t

Weekend exercises
(a) Find the pressure (force per unit area) between two innite and opposite charged parallel planes of charge density .
(b) Describe the magnetic eld associated with E = x
j.

44

Dierential Equations for the EM Field and

Maxwells Theory

We know now that electromagnetic forces are carried by electromagnetic

elds that propagate at speed c 3 108 ms1 . Because of the nite propagation speed we are forced to assign energy & momentum to the elds i.e.
we must think of them as real physical entities as against mere mathematical
conveniences (as is the case for static elds). An electromagnetic system then
qualies as static only if all the charges have been at rest longer than the
time taken to traverse the system at speed c.
In the 1830s Michael Faraday carried out experiments to measure a nite
electromagnetic propagation speed. He was unsuccessful due to lack of time
resolution in his apparatus. Faraday would have had no reason to think that
the electromagnetic speed was the same as the speed of light (then known).
In the 1860s, James Clerk Maxwell, seeking to advance Faradays ideas about
electromagnetic elds, by a brilliant process of intuition worked out how to
generalize certain dierential equations deduced from static experiments. He
produced a set of eld equations known by his name today. Maxwell also
had a theory i.e. a set of qualitative ideas underpinning his equations. The
theory (unlike the equations) has not stood the test of time.

4.1

Dierential Equations for the EM Field

Let us take as the source of the electromagnetic eld a continuous charge

and current distribution ( and J). Then we will produce macroscopic eld
equations
q=

dV ,

I=

J n dS .

We nd dierential vector equations from the integral forms of observational

results discussed above.

45

4.1.1

Divergence of E
E dS =

q
=
0

dV .
0

Applying Gauss Theorem to Coulombs Law:

dV .
0

E dV =

Since this must hold for arbitrary V no matter how small:

E =

4.1.2

.
0

Curl of E

From Faradays ux cutting rule we had:

Ed =

B
n dS .

Applying Stokess Theorem:

E n dS =

B
n dS .

For arbitrary S and n implies

E =

46

B
.
t

4.1.3

Divergence of B

From Biot-Savart law, we had:

B=

0
Id 2 .
4
r

Applying the rule

(A B) = B

AA

B ,

we obtain
B =
However,

0 I
4

r2

d d

r2

d = 0, since d is a constant vector, and

0 for any n ,
rn

Hence
B =0 .

4.1.4

Curl of B

From Amperes Circuit Law:

B d = 0 I = 0

J n dS .

Applying Stokess Theorem, the relation

B n dS = 0

47

J n dS

dl

B
I

holds for arbitrary n. Thus

B = 0 J .
However, Maxwell realized that unlike the previous three dierential
equations, this one could not be generally true.
To see this, take its divergence and remember that F 0 for any
vector function F :

B = 0 = 0

J .

Thus
J 0 !!!
We have already seen that conservation of electric charge requires
J =
Thus

.
t

J 0 implies that

0,
t

i.e. we could never charge or discharge a capacitor.

Maxwell guessed at the right value of
48

B as follows:

Since
J +

= 0 and = 0
t

(from the rst Maxwell equation for

Hence
J +

0
t

E)

= 0 or

( J + 0

E
)=0
t

If then we (after Maxwell) write:

B = 0 (J + 0

E
),
t

instead of
B = 0 J ,
we obtain

B 0

( J + 0

E
)=0,
t

which is in accord with conservation and motion of charge.

Note from above that the term that Maxwell added
0

E
,
t

has the dimensions of current density. Maxwell called it the displacement

current density.
Maxwell had a theory underpinning his equations in which the displacement
current was a real physical current - due to polarization of the electromagnetic ether. This theory has not survived. Nevertheless the above term is
still referred to as the displacement current.
We write the fourth Maxwell equation then as:
B = 0 J + 0 0
49

E
.
t

The dierential form of Maxwells equations is easier to interpret physically and is also useful in deriving the boundary conditions that the eld
vectors must satisfy.
The Maxwells equations are self-consistent and no experimental evidence
for requiring any further modications has been found.

Exercise in class: Fields within a capacitor

A plane parallel capacitor is being charged with a current I. Show that
the displacement current between the plates of the capacitor is equal to the
conduction current I in the external charging circuit. Remember that the
displacement current density is, by denition, 0 E so the displacement curt
rent through a surface S is:
ID =

0
S

E
n dS .

Assume that the external wires are perfect conductors so that E is zero in
them. Assume the space between the plates is a perfect insulator so no conduction current ows within the capacitor.
Can you see any curious consequence in this case if the displacement
current is assumed to be a real physical current (ow of charges)?

50

4.2

Maxwells Equations and Prediction of Electromagnetic Waves

The basic equations to study electromagnetic theory are Maxwells Equations, which govern the behavior of the time-varying electromagnetic eld
I.
II.
III.
IV.

,
0
B =0 ,
B
,
E =
t
E =

B = 0 J + 0 0

E
.
t

The rst equation is just Gausss law, the second tells us about the nonexistence of magnetic charges, the third equation is Faradays law, and the
nal equation is Amperes law.
Maxwells immediate triumph was to predict the existence of electromagnetic waves and their propagation speed. The calculated speed came (within
experimental error) to be equal to the measured speed of light. This prediction obviously led to the conclusion that light was electromagnetic in nature.
Thus arose a synthesis of electromagnetism and optics.

4.2.1

The Wave Equation for EM Waves in Vacuum

In a vacuum there are no sources i.e. = 0 and J = 0.

Hence, the Maxwells equations III and IV reduce to the following:
E =

B
,
t

B = = 0 0

(8)
E
.
t

(9)

We eliminate E or B between equations (8) and (9) to obtain dierential

equations for E or B alone, using where required E = 0 and B = 0.
51


Method: Think of and t as linear (dierential) operators. By analogy
with methods of solving linear algebraic equations, applying into (8) and
/t into (9), we obtain

B
,
t
2E
B = 0 0 2 .
t

= 0 0

2E
,
t2

E = 0 0

2E
.
t2

t
Hence

E
Because

E = 0 in the vacuum, we nally obtain

E=

1 2E
,
c2 t2

c2 =

1
.
0 0

(10)

where

The parameter c has the dimensions of velocity and is numerically equal

to 3 108 ms1 .
Equation (10) is the standard form of a three-dimensional vector wave
equation. The eld B satises the same equation.
4.2.2

Plane Wave Solution to the Wave Equation

The wave equation in a vacuum is

2

X=

1 2X
c2 t2

for X E, B.
52

Look for plane wave solutions propagating in the z direction.

In this case, /x = /y 0, and then
2

2
2
2
2
+ 2+ 2 = 2 .
x2 y
z
z

The dierential equations for E and B both have the same form:
1 2X
2X
= 2 2 .
z 2
c t
Such an equation has solutions of the form X = f (z ct) where f is an
arbitrary function.
This solution represents a signal propagating with speed c as can be seen
from the following discussion.
Let X0 = f (z0 ct0 ) i.e. X at t = t0 and z = z0 .
Now examine X at time t later and distance z further along in z. Since
a harmonic wave does not change in vacuum, we have
X1 = f (z0 + z c(t0 + t))
= f (z0 ct0 ) = X0
when

z = ct ,

i.e. the signal propagates a distance z = ct in time t i.e. it propagates

with speed c.
Proof of solution:
Let f represent f (z ct).
f
(z ct)
f
f
=
=
=f ,
z
(z ct)
z
(z ct)

where means dierentiation wrt z ct.

Similarly
2f
z 2
f
t

f
f
(z ct)
=
=f
z
(z ct)
z
f
(z ct)
=
= cf
(z ct)
t
=

53

Similarly
2f
= (c)(c)f .
t2
Consequently
1 2f
2f
= 2 2 .
z 2
c t
Thus
c2 =

1
1
c=
,
0 0
0 0

which with the numerical values of the parameters

0

8.85 1012 Fm1

0 = 4 107 Hm1

gives
c

3 108 ms1 .

What determines the form of the function f ?

Boundary conditions do.
Source (charges & current), propagating EM in a vacuum, detectors (other
charges)

4.2.3

Harmonic Waves

In vacuum, we chose a plane wave representation for the propagating EM

wave. For a harmonic wave
c = f =

,
k

= 2f,

k=

2
,

where f = frequency (Hz), = wavelength (m),

= radian frequency (radians s1 ), and k = propagation constant (m1 ).
f (z ct) = f (z

t) = f1 (t kz) .
k
54

A plane wave is represented by

E = E0 ei(tkz) .
Note that some textbooks on electromagnetism, for engineers in particular,
use j instead of i for the imaginary number. We will use i throughout this
lecture notes.

4.3

The Transverse Nature of Plane Waves in Vacuum

We now investigate the relations between the amplitudes and phases of the
electric and magnetic elds in a plane harmonic wave. While it is true that
the magnetic eld satises the same wave equation as the electric eld, it is
not independent of the latter, since one must satisfy the Maxwell equations III
and IV.
Since B = 0 always in electromagnetism, and
Bx By Bz
+
+
x
y
z
Bz
= 0+0+
z

B =

for a plane wave propagating in the z direction, we have

Bz
=0.
z
However, for a plane wave
Bz
= ikBz .
z
Hence, the rhs must be zero, which means that either k = 0 (zero frequency)
or Bz = 0 (transverse wave).
For a propagating wave k = 0, so the wave is transverse in B.
In a vacuum E = 0 and then by the same argument we conclude that
the plane wave is also transverse in E. In other cases in electromagnetism
(e.g. for a plasma in a magnetic eld) a plane wave may not be purely transverse in E.
55

In addition: E B for a plane EM wave in a vacuum.

Proof:
Consider a harmonic wave propagating parallel to the z axis.
In this case the eld components are ei(tkz)
We will expand Maxwell equation III
B
t
remembering that /t i

= i(Bx + By + Bz k)
i
j
z
Ez
E =

in Cartesian coordinates

i
j

Ex Ey

(11)

For electromagnetic plane waves propagating along the z axis in a vacuum:

=
= 0,
= ik .
Ez = Bz = 0,
x
y
z
Hence, Eq. (11) reduces to

i
j
k
i
j)
0 0 ik = i(Bx + By .
Ex Ey 0
Comparing the left and right-hand sides:
x component
y component

k
ikEy = iBx Bx = Ey

k
ikEx = iBy By = Ex

Consider a scalar product

E B = ( x + y ) ( x + y )
iE
jE
iB
jB
k
k
= ( x + y ) ( Ey + Ex )
iE
jE
i
j

k
k
= Ex Ey + Ex Ey = 0 ,

56

which means that E B.

Note that
E

= =c.
B
k
In electromagnetic theory when E and B are related, their ratio is always a
velocity characteristic of the problem in hand.
This is as far as Maxwell took the subject. It was for others like Heinrich
Hertz 1884 to show how to solve Maxwells equations with source terms , J
included (i.e. the generation of electromagnetic waves). We will consider
this later.
You should be aware that we have not derived Maxwells equations from
the static limits like Coulombs Law and the Biot-Savart Law. The solutions
to Maxwells equations include the static limits as special cases but many
more. Maxwells equations have the status of postulates suggested by experimental results.
In summary, we have the following important results for related electric
and magnetic elds propagating in vacuum:
1. The electric and magnetic elds propagate in a form of plane waves,
so-called electromagnetic (EM) waves.
2. The plane EM wave is transverse in E and B.
3. The electric and magnetic elds are perpendicular to each other.
4. The ratio E/B is constant and equal to the velocity of the wave, that
is equal to the speed of light in vacuum.

57

Questions:
(1) Starting from the Maxwells equations derive the continuity equation,
i.e. show that conservation of charge is built into the Maxwells equations.
(2) Using the Maxwells equations show that B satises the same wave
equation as E.
(3) Show, using the proof of solution of the wave equation, that f (z + ct)
represents a signal propagating in the negative z direction with speed c.

Weekend exercises
(a) Demonstrate that the Coulomb eld E =
point charge, follows from the Maxwells equations.

58

Q
r

40 r 2

for a stationary

EM Theory and Einsteins Special Theory

of Relativity

Special relativity (1905) grew out of Einsteins meditation on electromagnetic theory and the properties of space and time. Historically, the insights
of Einsteins theory follow after electromagnetism. Logically however, special
relativity contains more general statements about nature than electromagnetism. Electromagnetic eld theory is just one of a possible set of eld
theories that are compatible with the Einstein theory of space and time. It
is evident that relativistic eects are important if we were to calculate the
eld of a charge moving with a speed comparable to that of light. What is
not so obvious is that special relativity oers insights in to aspects of electromagnetic theory even in the case of the low speed charges we consider in
this course.
Two such aspects are:
(1) The unity of the electromagnetic eld i.e. the eld is a single entity
with 6 components (represented by two vectors E and B, each with three
components).
(2) Understanding the nature of causal relationships in electromagnetic
theory, e.g.
E =

B
.
t

Does this mean that a time changing B causes a spatially changing E?

5.1

Lorentz Transformation Equations for Space and

Time

The principle of relativity

1. The laws of physics are the same in all inertial reference frames.
59

2. The speed of light in vacuum is independent of the uniform motion of

the observer or source.
The constancy of the velocity of light, independent of the motion of the
source, gives rise to the relations between space and time coordinates in different inertial reference frames known as Lorentz transformations.
Consider a stationary reference frame S and a inertial frame S moving
with a velocity u parallel to the x axis.
y
y

S
x
x

Let x, y, z are coordinates in the S frame and x , y , z are the coordinates

in the S frame which is moving with a constant velocity u parallel to the x
axis with respect to S.
The time and space coordinates in S are related to those in S by the 1D
Lorentz transformations:
x
y
z
t

= (x ut)
= y
= z
ux
= t 2
c

60

where
u2
= 1 2
c

1
2

is the Lorentz factor.

The above transformation corresponds to a situation of u parallel to the x
axis. Later in the course, we will consider the general case of the velocity u
of the frame S in an arbitrary direction.

5.2

Force Transformation Equations

We will demonstrate how one might infer the law of Biot-Savart from application of special relativity to Coulomb law.

Consider a particle which is moving with velocity v = vx + vy + vz k in

i
j
the S frame and is acted on by a force with components Fx , Fy and Fz . Then
in the S frame:
Fx = F x

Fz

Fy

uvy
uvz
Fz ,
uvx Fy 2
c2 )
c (1 uv2x )
c

c2 (1

(1 v2 ) 2
Fy
c
,
=
uvx Fy =
1 c2
(1 uv2x )
c

(1 v2 ) 2
Fz
c
=
.
uvx Fz =
1 c2
(1 uv2x )
c

Suppose Fx , Fy , Fz represents the velocity independent Coulomb force.

Then in the S frame (source of the eld now moving) the force is no longer
velocity independent. In electromagnetic theory we say that there is now a
magnetic force and we dene a magnetic eld B that determines the magnetic force.
We now present a detailed calculation that illustrates how the form of
the Maxwells equations is determined by nature obeying Einsteins special
theory of relativity.

61

5.2.1

The Force between Two Charges Moving with Constant Velocities

Invariance of electric charge

In ordinary matter, electrons move with much greater speeds than protons
yet there is no associated electric eld. This implies that electric charge is
independent of velocity unless electromagnetic laws modied in some complicated way (see discussion by King in Physical Review Letters 5, 562 (1960)).
Consider charges q1 and q2 moving with velocities u and v in an inertial
frame S. No loss of generality occurs if v is taken parallel to the x axis.

u
q1

u
r

q2 (v,0,0)

q1

q2 (0,0,0)

Now consider a frame S moving with velocity v along x axis, that q2 is

stationary in S . Assume that at time t = 0, the frames S and S overlap.
From the principle of relativity, in the S frame Coulombs law holds. The
force on q1 seen in S is therefore
F =

1 q1 q2
r .
40 r 3

We shall transform this expression to nd the force observed in the S

frame in which the source of the eld (q2 ) is moving.
We shall see that what we normally call the MAGNETIC FIELD arises
as a natural consequence of relativistic invariance with no extra assumptions.
62

The x component of the force is

Fx =

1 q1 q2
x
40 r 3

and similarly for y and z.

The Lorentz transformations at t = 0 are
x
y
z
t

=
=
=
=

x
y
z
vx
2
c

To do the transformations, we also need transformation from r to r. Clearly,

there is an axial symmetry here and we should involve the angle .
r 2 = x 2 + y 2 + z 2 = 2 x2 + y 2 + z 2
y2 + z2
v2
2
2
2
2
= x +
= x + 1 2
y2 + z2
2

c
2
v
= 2 x2 + y 2 + z 2 2 y 2 + z 2
c
2
v2
v 2 2
2 2
2
2
= r 2 r sin = r 1 2 sin2 ,
c
c
where sin =
Hence

(y 2 + z 2 )/r.
1
2

v2
r = r 1 2 sin2
c

Substituting into Fx , we obtain

Fx =

1
40

q1 q2 x

3 r3 1
= q1 gx ,
63

v2
c2

sin2

3
2

where
1
40

g=

q2
2 r3 1

v2
c2

3
2

sin2

Similarly
Fy =

q1 gy

Fz =

q1 gz

The force transformations are

a) x component
Fx = F x

vuy
c2 1

Fy

vux
c2

vuz
c2 1

vux
c2

Fz

Thus
q1 gx = Fx

vuy
c2

vux
c2

Fy

vuz
c2

vux
c2

Fz

b) y component
Fy =

Fy
1

vux
c2

vux
vux
Fy = 1 2
c2
c
vux
= q1 gy 1 2
c

Fy = 1
Fy

c) z component
Fz = q1 gz 1

vux
c2

Substituting for Fy,z in x equation

Fx = q1 gx + q1 gy

vuy
vuz
+ q1 gz 2 .
2
c
c
64

q1 gy

Note: Here is the germ of the magnetic eld. The last two terms are typical
second order relativistic terms v 2 /c2 . In a nonrelativistic calculation we
would have Fx = Fx .
We now combine results a), b) and c) into a single vector equation for
the force F . First, note that
vux = u v .
Next, we can write the x component as
vux
vux
+ q1 gx 2
2
c
c
vuz
vuy
+ q1 gy 2 + q1 gz 2
c
c
v
vux
= q1 g 1 2 x + q1 g 2 (u r)
c
c

Fx = q1 gx 1

and with the y and z components

vux
y
c2
vux
= q1 g 1 2 z
c

Fy = q 1 g 1
Fz

these three components combine into

v
vu
r + q1 g 2 (u r)
2
c
c
q1 g
= q1 gr + 2 [v (u r) r (u v)]
c
q1 g
F = q1 gr + 2 u (v r)
c

F = q1 g 1

We can write this equation in the form of the Lorentz force

F = q1 E + u B
where
E = gr =

1
40

q2
2 r3 1
65

v2
c2

sin2

3
2

and
B=

v gr
vE
=
.
2
c
c2

Note that as v 0, 1, and then

E

1 q2
r.
40 r 3

Moreover, the ration of magnitudes of magnetic to electric term in the force

equation is uv/c2 , i.e. magnetic forces are second order relativistic eects.

5.3

5.3.1

Electric and Magnetic Field Lines of a Moving

Charge
Electric Field Lines

Let = v/c. Then, we can write the electric eld as

E=

1
40

q(1 2 )
r3

sin

3
2

r.

For a given , the electric eld E still varies as 1/r 2 , but the eld lines are
crowded in the direction perpendicular to v.
In the forward direction = 0 and then
E=

1 q(1 2 )
< Es ,
40
r2

where Es is the static electric eld (at v = 0).

In the perpendicular direction, = /2, and then
E=

1 q
1 2
2
40 r

1
2

> Es .

The electric eld lines radiate from the present position of the charge.

66

static field
v

5.3.2

Magnetic Field Lines

From the relation between electric and magnetic elds, we nd

vE
c2
1
=
40

B =

q(1 2 ) v r
c2 r 3 1 2 sin2

3
2

In spherical polar coordinates

B = B r r + B + B = B

since B v r, and r , , are unit vectors.

The magnetic eld lines form concentric rings about v, and there is symmetry about the plane = /2.
In the non-relativistic case of v
1, 0, and then
B=

1 q vr
,
40 c2 r 3

which is the Biot-Savart law.

Applied to a continuous line current I:
B=

1 I dl r
.
40 c2 r 3
67

E
v

The constant 1/(0 c2 ) is normally written 0 - the magnetic permeability of

free space.

5.4

Field equations for an elementary point charge moving with uniform velocity

(1) Equation for the total electric ux

Consider the total electric ux
E =

E n dS .

We will use the axial symmetry and break sphere up into rings lying between
and + d.

68

Since E

n, the ux dE is

dE = E n dS = EdS .

However
dS = 2(r sin ) rd = 2r 2 sin d .
Hence
1
40

dE =

q(1 2 )
r 2 1 2 sin2

q(1 2 )
20

3
2

2r 2 sin d

sin d

1 2 sin2

3
2

Thus
E =

q(1 2 )
dE =
20

=0

sin d
1 2 sin2

3
2

To calculate the integral, put cos = x, so that sin d = dx, and then
dx

I =
=
where a =

(1

2 x2 )3/2

dx

1
3

(a2

+ x2 )3/2

1
3

dx
1 2
2

+ x2

1 2 /. Performing the integration, we obtain

I=

x
1
.
a2 3 (a2 + x2 )1/2

Thus, including the limits of the integral

1

I =

dx

(1 2 + 2 x2 )3/2
1
1
1
1
2
= 2 3 2
+ 2 3 2
=
a a +1 a a +1
1 2
1

69

3/2

and then
E =

q(1 2 )
2
q
=
.
2
20
1
0

Hence, the electric eld produced by a moving charge satises the Gausss law.
(2) Magnetic ux through a closed surface
Consider the magnetic ux through a closed surface
M =

B n dS .

If we choose a sphere centered on q to calculate the integral, we nd that B

is perpendicular to n everywhere, and then M = 0.

Similarly, B = 0 always. It is easy to see that

1
vE
c2
1
= 2 E vv
c

B =

since

E =0

v = 0 as v is constant, and v is perpendicular to

E.

(3) Spatial E derivative related to temporal B derivative.

We shall show that for a point charge under uniform velocity
E =

B
.
t

In spherical polar coordinates

r

(sin E ) E

r sin

1 Er (rE )

r sin
r

(rE ) Er
+
.

r
r

E =

70

Since
E = E = 0 ,
then
E =

1 Er

with
1
40

Er =

q(1 2 )
r 2 1 2 sin2
K

=
1

sin

3
2

3
2

Next
Er

3
K 1 2 sin2
2
3K 2 sin 2
=
5 .
2 1 2 sin2 2
=

5
2

2 2 sin cos

Hence
E =

1 3q(1 2 )
40
2r 3

2 sin 2
1

sin

5
2

To calculate B/t, we use the theorem of partial derivatives. If

y = f (x, t)
then from the maximum change of y
dy =

y
y
dx +
dt = 0 ,
x
t

we obtain
y x
y
=
.
t
x t
71

Thus
B
B
= v
.
t
x
Alternatively, to see this physically, remember that the eld pattern
moves with constant velocity v. Let a stationary observer measure the change
in the eld B in a time interval dt. This change is the same as he would
observed at a xed time by moving a distance dx = vdt, i.e.
dB

in dt dB

in dx = vdt .

Hence
B
B
=
,
x
vt
and then
B
B
= v
.
t
x
Now
B=

1
40

q(1 2 ) v sin
c2 r 2

sin

3
2

and
B B r
B
=
=

x
x
r x
Since
sin =

y2 + z2
a
=
r
r

we can write B as
B =

Ka
r3 1

2 a2
r2

3/2

72

Ka
(r 2 2 a2 )3/2

where
1 q(1 2 )v
K=
40
c2

and a =

y2 + z2

Next
1/2

r
(x2 + y 2 + z 2 )
=
x
x
1/2
1 2
x
=
x + y2 + z2
2x = = cos
2
r
3
B
= Ka
r 2 2 a2
r
2
3Ka r
=
(r 2 2 a2 )5/2

5/2

2r

Hence
B
1 3q(1 2 ) vr 2 sin cos
=
5
x
40
c2 r 5 1 2 sin2 2
=

1
40

3q(1 2 ) v sin 2
2c2 r 3 1 2 sin2

5
2

and then
B
B
1
3q(1 2 ) v 2 sin 2
= v
=
5
t
x
40 2 3
2c r 1 2 sin2 2
=

Comparing with

1 3q(1 2 ) 2 sin 2
5
40
3 1 2 sin2 2
2r

E, we see that
E =
73

B
.
t

5.4.1

Electromagnetic Induction

It has been known since about 1831 when Faraday rst waved a magnet
near an electric circuit and played with transformers that when the magnetic
ux through a circuit changes, an electromotive force (emf) E appears in it.
Faraday gave the rule
E=

M
.
t

The nature of this phenomenon is, however often misinterpreted.

Consider a point charge moving near a closed circuit as shown in the
gure.

(1)
(2)

Because of the dependence of E, the electric eld on side (1) is larger

than that on side (2). Thus, there is a net driving force round the circuit.
Calculate the resulting electromotive force in the circuit, which is equal
to the work done on a charge in the circuit
E = Wq =

F dl
=
q

(E + u B) dl ,

where u is the velocity of the charge q.

Now since u B is perpendicular to both u and B, it is also perpendicular
to dl, and then
u B dl = 0 .
74

Hence
E =

E dl =

E n dS =

E dS

B
M
dS =
.
B dS =
t
t
t
However, it is obvious in this example that the changing magnetic ux is not
the CAUSE of the emf. The changing magnetic eld and the electric eld
have a common CAUSE through the charge q.
We can conclude that: Electric and magnetic elds do not produce
each other - they are both due to electric charges.
It is often thought in the textbooks, however that e.g. in a transformer
the changing M produces E. It happens because the ux cutting rule is an
extremely powerful one for calculating the integrated electric eld of electric
currents.
The Faradays rule E = M /t, which arises from E = B/t
should not be thought of as a casual relationship. What it means is that if
a charge moving with a constant velocity produces a time varying magnetic
eld then that charge also produces a spatially varying electric eld.
=

(4) Relation between spatial variation of B and temporal variation of E.

Since
vE
,
c2
= E
v (v

B =
vE

)E +v

E E(

and v is constant, we obtain

v =0 ,

v=0,

and then
1
vE
c2
1
= 2 (v ) E + v
c

B =

75

v)

Now
(v

)E =

vx

= v

E
+ vy
+ vz
x
y
z

E
E
=
x
t

as
vx = v

vy = vz = 0 and

= v
t
x

and then

1
B = 2 v
c

For points in free space

B =

E
E+
.
t
1 E
.
c2 t

In summary
Maxwells equations for a point charge moving with uniform velocity are
E = 0
B = 0

B
t
1 E
B = 2
c t
E =

These equations arise from the necessity for the correct relativistic transformations between frames in uniform relative motion. If the postulates of
relativity are correct and Coulombs law gives the eld of a stationary charge,
these equations follow, and the force on a charge is
F =q E +vB
76

Questions
(1) Show that magnetic and electric elds of a charge moving with a
constant velocity v are related by
B=

vE
c2

(2) Show that the magnetic eld lines produced by a charge moving with
a constant velocity v form concentric rings about v.
(3) Show that the electric eld produced by a moving charge satises the
Gausss law.
(4) Explain the statement: Electric and magnetic elds do not produce
each other - they are both due to electric charges.

77

Energy and Momentum in the Electromagnetic Field

Since the electromagnetic eld propagates at nite speed it must be assigned

energy and momentum or else we must say that energy and momentum are
not conserved in the nite time of propagation.

SOURCE

t =r/c

DETECTOR

The eld takes a time t = r/c to travel from the source charge to the
detector (another charge).

6.1

The Energy Conservation Theorem Poyntings

Theorem

Energy may be transported through space by means of electromagnetic waves.

We expect energy ow in the direction of propagation of the wave: E B.

E
PLANE WAVE

EXB
B
We will show that the power ow across an element of area dS is given
by c2 0 E B dS. To derive this relation consider a volume V bounded by
78

a closed surface S, and containing a charged material.

Consider the expression
(E B) .

(12)

If we employ the vector identity

(E B) = B

E E

and use the Maxwell equations III and IV:

E =

B
t

and

B = 0 J + 0 0

E
,
t

we can write Eq. (12) as

(E B) = B

E
B
0 E J 0 0 E
,
t
t

or
1
0

(E B) = E J

1
1 B2
0 E 2 +
2
2 0

On lhs we put 1/0 = 0 c2 and integrate the equation over some closed
surface S enclosing a volume V . Then, we obtain
0 c2

(E B) dV =

E J dV

1
1 B2
0 E 2 +
dV . (13)
2
2 0

Now we apply Gauss theorem to the lhs of the above equation, and nd

0 c2 (E B) n dS

E J dV

Energy ux

Rate of doing work by eld on the current

1 B2
1
2
0 E +
dV
2
2 0

Field energy.

This interpretation of the terms in this equation is suggested by some

special cases:
79

6.1.1

Ohmic Heating

From the circuit theory, we know that in a resistive medium Ohms Law is:
V = IR and R = R /A, where R is the resistivity of the medium.

Thus, I = V /R = (V A)/(R ). Since E = V / (E =

I/A =

V ), we get

1
E.
R

Dening the conductivity = 1/R, Ohms Law becomes:

I/A = J = E
The rate of conversion of electromagnetic eld energy in to heat is then:
H = IV =

V2
E2 2
= E 2 A = E 2 V ,
=
R
RA

where V = A is the volume of the resistive medium.

Hence
H
= E 2 = E J = E J .
V
Thus, E J is the rate of heating per unit volume in this case.

80

6.1.2

Electrostatic Field Energy Density

It is well known that the work required to charge a capacitor of a capacitance

1
C to a voltage V is W = 2 CV 2 . According to the eld theory of electromagnetism this work done corresponds to conversion from some other form
of energy into electrostatic eld energy
0 A
1
,
E = W = CV 2 and C =
2
d
and the electric eld in the capacitor is given by V = Ed. Thus:
E=

1 0 A 2 2 1
E d = 0 E 2 V ,
2 d
2

where V = Ad is the volume of the capacitor.

Hence
1
E
= 0 E 2 .
V
2
6.1.3

Magnetostatic Field Energy Density

It is well known that the work required to energize an inductor of L [henrys]

1
to a current I is E = 2 LI 2 and that the magnetic eld within a long solenoid
of self-inductance L = 0 n2 A is B = 0 nI. According to the eld theory
of electromagnetism this work done corresponds to transformation of some
other form of energy to magnetic eld energy.
Thus, for a solenoid of length and area of cross section A we have energy:
1
1
1
1
E = W = LI 2 = 0 n2 A I 2 = (0 nI)2 A
2
2
2
0
2
1B
=
V
2 0
or
E
1 B2
=
.
V
2 0
81

All our observations are in accord with supposing that the expressions
for energy density found in electrostatic and magnetostatic situations are
applicable in general.
The expression
N = 0 c2 (E B)
is referred to as the Poynting vector.
It represents the energy ux in the electromagnetic eld, i.e. the energy
ow per unit area (measured normal to the ow) per unit time.
The units of N are thus watts/square meter.
The energy ow equation can be converted into the form of a dierential
continuity equation or energy conservation law. From Eq. (13), we have
U
+
t

N = J E ,

where
1 B2
1
U = 0 E 2 +
2
2 0
is the energy density of the EM eld.
The physical meaning of the dierential continuity equation is that the
time rate of change of electromagnetic energy within a certain volume, plus
the energy owing out through the boundary surface of the volume is equal
to the negative of the total work done by the elds on the source inside the
volume. Thus. J E is a conversion of electromagnetic energy into heat
energy.
Example
An Earth satellite transmits a power P = 5 W. What are the typical E
and B in the radiation eld at a distance of 1000 km from the transmitter?
The mean ux (Poynting vector magnitude) at a distance r meters is:
N=

P
5
=
= 4 1013 [Wm2 ] .
2
4r
41012
82

Since
EB

and B =

E
,
c

N = 0 cE 2

4 1013 = 8.85 1012 (3 108 ) E 2

Hence
E

1.2 104 Vm1 .

Thus, an antenna consisting of a straight piece of wire 10 cm long would have

a voltage V
E
1.23 104 0.1 or
12 V induced in it if it were
placed parallel to E.
Then
B = E/c = 1.2 104 /(3 108 )

4 1013 [T] .

It is a very weak eld. For, example magnetic eld produced by a hair

dryer is of order 103 T. Magnetic eld produced by a large laboratory electromagnet is about 5 T.

6.1.4

No Fluxes from Static Fields

Consider a source of electrostatic eld E and magnetostatic eld B. If it is

arranged so that E B it is sometimes asked whether we should expect to
see an energy ux of 0 c2 E B.

SOURCE OF E
STATIC E

SOURCE OF
STATIC B

The question is easily answered in the negative by noting that the electrostatic eld is E, 0 and the magnetostatic eld is 0, B. Each of these has
83

E B = 0. The E, B obtained by mentally combining the electrostatic E

from one source and the magnetostatic B does not constitute an electromagnetic eld.
Electromagnetic elds do not interact with each other
The elds interfere (as the term is used in physical optics). Beyond the
interference region the two plane waves are the same as they would have been
had there been no interference. If there are charges in the region of interference then this is a dierent matter. The charges may absorb and re-emit
radiation and so permanently modify the radiation elds.

6.2

Phase Relationships in Harmonic Waves

Only the in-phase components of E and B contribute to net energy ow

averaged over a whole cycle of the radiation.
If E = E0 cos(t) and B = B0 cos(t) then:
i
j

N = 0 c2 E B = 0 c2 E0 B0 cos2 (t) k .

Since, cos2 (t) = 1 , we obtain

2

1
N = 0 c2 E0 B0 = 0 c2 Erms Brms ,
2

where Erms = E0 / 2 and Brms = B0 / 2.

If however E = E0 cos(t) and B = B0 sin(t) then:
i
j

N = 0 c2 E B = 0 c2 E0 B0 cos(t) sin(t) k .
Since, cos(t) sin(t) = 0, we have N = 0.

6.3

Momentum Flux

To obtain an expression for the momentum carried by the electromagnetic

eld, we may employ the relativistic energy-momentum relationship
E 2 = p 2 c2 + m 2 c4 .
0
84

Since for electromagnetic radiation m0 = 0, we obtain

p=

E
.
c

Thus the momentum ux of the electromagnetic eld is:

M=

6.4

0 c2 E B
= 0 cE B .
c

Electromagnetic Energy Flow

We now consider two simple examples illustrating applications of the Poynting vector to circuit theory to show how the eld theory provides an alternative way of viewing some circuit problems.
6.4.1

Energy Flow into a Resistive Wire

Consider a wire (resistor) of length , carrying a current I. Let V is a potential dierence applied along a wire.
Circuit theory calculation

I
l

According to circuit theory the power dissipated in the wire is P = V I,

where I is the current owing through the wire. Thus, according to the circuit theory, energy ows along the wire.
Field theory calculation
Lets look at the same problem from the point of view of the eld theory.
According to electromagnetic eld theory the ow of energy should be described by the Poynting vector N = 0 c2 E B.
85

^
z

^
r
Calculate the electric and magnetic eld produced by the current. The
electric eld propagates along the wire, and is given by
E=

V =

z.

From the Amperes line integral theorem, we nd the magnetic eld around
the wire
B=

0 I
.
2a

Hence
N = 0 c2 E B = 0 c2

V 0 I
r

2a

since z = r .

Since 0 0 c2 = 1, we nally obtain
N=

VI
r.

2 a

Thus, the eld theory predicts that energy ows into the wire from the air
not along the wire. The energy is in the elds, the wire provides boundary
conditions and guides the elds.
If the energy is in the elds, it means that the electromagnetic energy
goes out of a battery into the air, and then goes into the wire from the air.
This is exactly the case we will show in the following example.

86

The total rate at which eld energy ows in to the surface is given by

N dS =

N r dS =

= N

dSside =

N dS
VI
2 a = V I ,
2 a

(N dS on the ends of the cylinder), which is in agreement with the result

of the circuit theory. This result demonstrates quantitatively that the power
which heats the resistor enters through the sides not through the wires.
6.4.2

Energy Flow out of Battery

In the above example, we have shown that the energy enters the resistor from
the air. Then, a question arises: If the energy enters the resistor from the
air, how does the energy get out to the air from a source of energy (battery)?
Consider a battery, which provides energy to the resistor.

+
E

Inside the battery J and E are in opposite directions. The magnetic eld
circulates around the battery, so we see that the Poynting vector N points
out into the air, not along the wire.

87

6.4.3

Propagation of Energy along a Wire

Above we showed that eld energy ows in to a wire so that it can be dissipated as heat. It is well known that electromagnetic energy can be transmitted along a wire with very little loss. This is a dierent situation. If we

/2
+

Ex B
B

Ex B

consider the case of a perfect conductor we nd that there will be a current

wave along the wire with surface charges induced producing a radial electric
eld. Then E B is parallel to the wire and the eld description of energy
transmission is that it is transmitted in the space around the wire. In the
space around the wire E and B are in phase and N is always in the same
direction. Within a perfect conductor we will show that E and B are /2
out of phase so the mean N averaged over a cycle is 0. So there is no net
energy transmission within the perfectly conducting wire.

Exercise in class: Energizing of a capacitor

Consider the energizing of a plane parallel capacitor with circular plates.
Show that circuit and eld calculations agree as to the rate of energizing of
the capacitor, i.e. Pc = Pf where:
Pc = V I = rate of doing work (by current I and voltage V between the
plates) in charging the capacitor according to circuit theory.

88

Pf =

0 c2 E B dS = rate of energy ow into the surface of the ca-

pacitor according to eld theory.

From which direction does the energy ow in to the capacitor according

to eld theory?
We will keep the calculation simple assuming the plates of the capacitor
to be uniformly charged. Under what conditions is this assumption likely to
be true?

Questions:
(1) Using the Maxwells equations derive the continuity equation for the
Poynting vector.
(2) Show, using the eld theory calculation, that the power dissipated
along a resistive wire is P = V I, the same predicted by the circuit theory.

89

General Solution of the Maxwells Equations

The electromagnetic elds of charges in uniform motion are eectively bound

to the charges. The elds of accelerated charges on the other hand can propagate as electromagnetic (EM) waves at the speed c and can have a life of
their own (until absorbed by some other charges).
Consider a simple model showing why only accelerating charges can emit
electromagnetic waves.
Charges in uniform motion produce a constant magnetic eld, i.e. when
v = const then B = const. EM wave is produced when B = const. From the
Faraday induction law
E(t) =

d
= const ,
dt

where = B(t) S is the magnetic ux.

B = const when v = const (acceleration).

7.1

Diculty of the Direct Solution of Maxwells Equations with Time Varying Sources

The existence of an electromagnetic eld implies the presence of currents and

charges. If the currents and charges are known, we can nd the elds solving
the Maxwells equations.
Consider the Maxwells equations
I.
II.

E = /0

B =0

B
t

III.

E =

IV.

B = 0 J +
90

1
E.
c2 t

In general case, E(r, t) and B(r, t) depend on (r, t), the charge and current
densities also depend on (r, t).
First, we will try to separate the Maxwells equations into an equation for E
alone or B alone.
Assuming in the usual way that space and time operators commute, we act

with c12 t on III and on IV, and obtain

1
c2 t
1
c2 t

E +

1 2
B = 0
c2 t2
B) = 0

J .

Eliminating E by subtraction
(

B) +

1 2
B = 0
c2 t2

and using the vector identity for double product, and II

(

B) =

B+

B) =

B,

we obtain
2

1 2
B = 0
c2 t2

J .

(14)

Similarly, elimination of B gives

2

1 2
E=
+ 0 J .
2 t2
c
0
t

(15)

Equations (14) and (15) are in the form of coupled wave equations known
as inhomogeneous Helmholtz equations. We see that the current density J
enters into these equations in a relatively complicated way, and for this reason these equations and are not readily soluble in general.
In the absence of currents and charges, J = 0, = 0, and then the above
equations describe a free EM eld, and can be solved separately.
91

Solution of the wave equations in the absence of currents and charges is given
in a form of plane transverse waves
Uk ei(k tkr) ,

U=

(16)

where U (E, B), k = k /c, and Uk is the amplitude of the E or B wave

propagating in the k direction.

The general solution of the wave equations, in the presence of space and
time varying currents and charges, is more readily attained via the electromagnetic potentials.

7.2

Scalar and Vector Potentials

Generally, we do not nd elds E and B directly by integration of Eqs. (14)

and (15). We rather rst compute scalar and vector potentials from which
the elds may be found.
We will illustrate here the advantage of working with the scalar and vector potentials.
Scalar potential is a quantity from which a eld can be derived by a process of dierentiation, e.g. in electrostatics
E= ,
where is the electrostatic potential.
Introduce the vector potential A dened such that the Maxwells equation II remains unchanged. Since B = 0, we can always write
B=

A .

Substitute this relation to the Maxwells equation III, and since

where is an arbitrary scalar function, we nd
E=

A
t
92

= 0,
(17)

In the static limit of A/t = 0, the scalar function reduces to the

familiar electrostatic potential.
The above equation shows that the electric eld depends on the specic
choice of the potentials.
We can dene new potentials without changing E and B
A =A+

.
t

(18)

Proof:

A
t

=
A +
t

= A =E .
t

=
=

A =

A+

A=B .

( )

as required.
The transformation (18) is called a gauge transformation, and the invariance of the elds under such transformations is called gauge invariance.
The denition B =
A does not completely dene A despite the fact
that B is completely dened. The vector potential A is arbitrary to the extent that the gradient of some scalar function can be added. Thus, innite
set of possible potentials corresponds to an innite set of possible vector
potentials. However, the Maxwells equations should be independent of the
specic choice of potentials.

93

7.2.1

Lorentz Gauge

Recall the Helmholtz Theorem which says that any vector eld can be
written as a sum two terms
F =

1
4

F
1
dV +
r
4

F
dV
r

= Fl + Ft ,
where Fl is called the longitudinal part of the eld and has
Fl = 0,
while Ft is called the transverse part and has Ft = 0.
We see that F and F together determine F but neither do alone.
Thus, if we dene
A, we complete the denition of A. This is called
choosing the gauge of the potential. The above is an excellent illustration
of the power of the Helmholtz theorem. This theorem enables us to recognize basic common properties of vector elds independent of their individual
physical properties.
Derive dierential equations for A and .
From the Maxwells equation I and (17), we nd
E =

The electric eld E will satisfy the Maxwells equation I when

= /0 .

From the Maxwells equation IV and B =

(

A) = 0 J +

A, we have

A
2 t
c
t

which can be written as

A+

A) = 0 J
94

1 2
1
A 2
2 t2
c
c

.
t

Hence
2

1 2
A 2 2 A = 0 J +
c t

A+

c2 t

We have already shown that

(
t

A)

= /0 .

(19)

The freedom of choosing A and means that we can choose a set of

potentials to satisfy the condition
A+

1
=0.
c2 t

This is called the Lorentz gauge and denes

times called the Lorentz equation.

A. This equation is some-

Under the Lorentz gauge, the Maxwell equations reduce to two uncoupled
wave equations
2

1 2
A = 0 J ,
c2 t2

and
2

7.2.2

1 2
= /0 .
c2 t2

Coulomb Gauge

Another useful gauge of the potentials is the Coulomb gauge or transverse

gauge
A=0 .
The origin of the name Coulomb gauge is in equation (19) that under the
condition A = 0 reduces to the Poisson equation
2

= /0 ,
95

that determines the Coulomb potential due to the charge density .

Before we proceed further, we will show that the solution of the Poisson
equation is of the form
(r) =

dV .
r

1
40

It can be proved in the following way.

From the Coulombs law
E=

1
40

r
dV .
r2

and using the relation

1
r

= 2 ,
r
r

we can write the electric eld as

E=

1
40

dV = ,
r

where
=

1
40

dV .
r

(20)

Since the electric eld satises the Maxwells equation I, we nd

E =

,
0

as required.
Now, we can nd the wave equation for A under the Coulomb gauge,
Under the Coulomb gauge, the vector potential satises the inhomogeneous wave equation
2

1 2
1
A = 0 J + 2
2 t2
c
c
96

.
t

The term involving the scalar potential is called longitudinal as it has

vanishing . This suggests that it may cancel the longitudinal part of the
current density J . According to the Helmholtz Theorem, the current density
can be written as
J =

1
4

J
1
dV +
r
4

J
dV
r

= Jl + Jt .
Using the continuity equation

+
t

J = 0

and the solution of the Poisson equation, we nd the longitudinal part of the
current density
Jl

1
=
4
=

1
4 t

1
J
dV =
r
4

t
dV
r

dV = 0
.
r
t

Then
0 Jl = 0 0

1
= 2
t
c

.
t

Hence, the inhomogeneous term in the wave equation for A can be expressed entirely in terms of the transverse current.
2

1 2
A = 0 Jt .
c2 t2

This is the origin of the name transverse gauge.

7.3

Solution of the Inhomogeneous Wave Equations

We have shown that the Maxwell equations can be reduced to two independent wave equations for the potentials A and . In fact, we have four scalar
97

equations for (Ax , Ay , Az , ). Each of these equations has the same form.
We will illustrate the solution on one of the equations, :
2

1 2
= /0 .
c2 t2

A general solution may be found by considering two limiting cases:

(a)

Electrostatic limit /t 0

In this limit the wave equation for reduces to the Poisson equation
whose the solution is
(r) =

1
40

(r )
dV .
r

(b) Source free limit = 0

In this case, the wave equation reduces to the homogeneous equation
2

1 2
2 2 =0 .
c t

This equation has a spherically symmetric solution of the form

(r, t) =

f (t r/c)
,
r

where f (t r/c) is an arbitrary function of the retarded time t r/c. The

retardation r/c is equal to the time needed for the electromagnetic wave to
pass the distance from the source to a given point in space.
Proof:
If there are no boundary surfaces, the potential can depend only on r,
and must in fact be spherically symmetric. Thus, in spherical coordinates
only the radial part of the Laplacian will contribute to the wave equation
2

r2
2 r
r
r
98

Hence, we have
f
f tr
1
=
= f ,
r
tr r
c
where tr = t r/c and f = f /tr .
Thus
2

1
f
f
+ 2
r2
2 r
r
cr r
1 rf
= 2
+f
r r c
r
1
1
1 f
+
f + f
= 2
r c
c
c
c
1
f ,
=
rc2

where f = 2 f /t2 .
r
Since /t = /tr , we have
1 2
1
= 2 f ,
2 t2
c
cr
as required.
We can construct a general solution of the wave equation by noting that
it must represent a spherical wave outside the source and reduce to the
appropriate static limit. This solution is
(r, t) =

1
40

(t r/c)
dV ,
r

where r is the distance coordinate from the source (from the charge dV ) at
the time when the potential wave left it. This exhibits the causal behavior
associated with the wave disturbance. The argument of shows that an
eect observed at the point r at time t is caused by the action of the source
a distant r away at an earlier or retarded time t = t r/c. The time r/c is
the time of propagation of the disturbance from the source to the point r.
Thus, the Maxwells equations satisfy the causality principle.

99

7.4

Alternative Solution: Green Functions Method

The wave equations all have the basic structure

2

(r, t)

1 2 (r, t)
= 4f (r, t) ,
c2 t2

where f (r, t) is a known (source distribution) function.

To solve this equation, we will introduce the Green function of the equation and solve it as an inhomogeneous Helmholtz equation.
Suppose that (r, t) and f (r, t) have the Fourier integrals
(r, ) =
f (r, ) =

(r, t)eit dt ,
f (r, t)eit dt .

When we insert it into the wave equation, we nd that the Fourier transform (r, ) satises the inhomogeneous Helmholtz wave equation
2

+ k 2 (r, ) = 4f (r, ) ,

where k = /c is the wave number.

The advantage of working in Fourier components is to remove the derivative over time.
Green function
For a unit point source the potential satises the Poisson equation
21

= 4(r) .

The function 1/r = G(r) is called a Green function of the above dierential equation.
In analogy, we can dene the Green function of the wave equation
2

1 2
G(r, t t0 ) = 4(r)(t t0 ) .
c2 t2
100

The Fourier transform gives

2

+ k 2 Gk = 4(r)eit0 .

where Gk is the Fourier transform of the Green function G(r, t t0 ), which

we are trying to nd.
If there are no boundary surfaces, the Green function depends only on r,
and then the Laplacian operator in spherical coordinates depends only on r
giving
1 d2
rGk eit0 + k 2 Gk eit0 = 4(r) .
r dr 2
Everywhere except r = 0, the function rGk eit0 satises the homogeneous equation
d2
rGk eit0 + k 2 (rGk eit0 ) = 0 ,
2
dr
whose the solution is
rGk eit0 = Aeikr + Beikr .
Thus, the general solution for the Green function is
Gk =

eikr eit0
.
r

Using the inverse Fourier transform, we nd

G(r, ) =

1
2

eikr i
e
d ,
r

where = t t0 .
The integral
1
2
is the delta function (

ei(

r/c)

r/c). Thus
1
G(r, ) = (
r
101

r/c) .

The Green function is a casual response function, and has the same property
as the scalar potential of a point source.
In summary
The general (retarded) solutions of the Maxwells equations are
B=

A ,

E =

A
,
t

(21)

with
1
(t r/c)
dV ,
40
r
1
J(t r/c)
dV .
A(r, t) =
2
40 c
r
(r, t) =

(22)
(23)

In practice, we calculate the scalar and vector potentials from Eqs. (22)
and (23), and then nd the electric and magnetic elds from Eqs. (21).

102

Questions
(1) Find a dierential equation for the vector potential A in the Lorentz
gauge.
(2) Explain, why the Coulomb gauge is often called Transverse gauge.
(3) Prove that the homogeneous wave equation
2

1 2
2 2 =0
c t

has the solution of the form

(r, t) =

f (t r/c)
r

where f (t r/c) is an arbitrary function of the retarded time t r/c.

103

Solution of Laplaces Equation and Boundary Value Problem

In the previous lecture, we have shown that electric and magnetic elds
are readily calculated with a help of the scalar and vector potentials. In
this lecture, we will illustrate applications of the scalar potential to physical
problems involving bounded elds.
There is a class of problems in electromagnetism in which a eld can be
derived without involvement of the complete set of the Maxwells equations
as the gradient of a scalar potential which satises Laplaces equation
2

=0.

The condition for this to happen is that

F =0

and

F =0 .

Since, F = 0 is the condition for F =

function, then F = 0 means that
( ) =

(24)

, where is an arbitrary scalar

=0.

Thus, the scalar potential contains all the necessary information to completely specify the eld of the properties (24).
Examples
1. Electrostatic problems involving Laplaces equation

Since in general E = /0 and E = t B we see that the requirement for Laplaces equation to be relevant is that = 0 and /t = 0, i.e. a
source-free region and static conditions. Of course there must be a source of
charges somewhere or there would be no eld anywhere. The typical situation where solution of Laplaces equation is relevant in electrostatic is where
we have source-free non-conducting regions between statically charged conductors.

104

2. Magnetostatic problems involving Laplaces equation

Since in general B = 0 and B = 0 J + c12 t E, we see that the

requirement for Laplaces equation to be relevant is that J = 0 and /t = 0,
i.e. a source-free region and static conditions. Again, there must be currents
somewhere or there would be no elds anywhere. The typical situation is
to be calculating the magnetic eld in the non-conducting region between
constant currents.

8.1

Uniqueness of the Solution of Laplaces Equation

As we will see below, general solutions of Laplaces equation are in terms

of some constants which are usually found from boundary conditions for a
given problem. A question arises: What boundary conditions are appropriate for the Laplace equation to ensure that a unique and
well-behaved (physically reasonable) solution will exist inside the
bounded region? Our experience leads to believe that specication of the
potential on a closed surface denes a unique potential problem. This is
called Dirichlet theorem or Dirichlet boundary conditions.

8.1.1

Dirichlet theorem

Consider a volume V completely bounded by a closed surface S. Within S

there is a potential satisfying 2 = 0. The Dirichlet theorem says that
the value of is uniquely determined by the variation of over S.
Proof:
Suppose, to the contrary, that there exist two solutions 1 and 2 satisfying the same boundary condition, i.e. 2 1 = 0 and 2 2 = 0 within S,
but 1 = 2 on S.

105

Let U = 1 2 is the dierence between the solutions. Since 1 and 2

are known to be solutions of the Laplace equation, then from the linearity
of the 2 operator 2 U = 0, i.e. U is also a solution of the Laplace equation.
We will prove that U = 0 inside the volume. To show this, we introduce
a vector F = U U . Then using the property
F =

U U =U

( U) +

and the Gauss Divergence Theorem, we get

F dV =

=
=

U dV +

U dV +

U dV

( U )2 dV

U U dS .

Now RHS= 0 because U = 0 over S.

Also, the integral
V

U dV = 0 ,

because 1 and 2 both satisfy the Laplace equation throughout V .

Hence
V

( U )2 dV = 0 .

Since the integral from a positive function is always positive, U must be

zero for the integral to be zero. Thus U = 0 and consequently, inside V , U
is constant. Since U = 0 on S, so that inside V , 1 = 2 .

8.2

Solutions of Laplaces Equation

There are dierent methods of solving the Laplace equation

106

 Method of Images
Green functions method
Variational method
Method of lattices
Numerical Monte-Carlo simulations method
Method of separation of variables
Solution in spherical coordinates
We will illustrate last two methods which can be applied to a large class
of problems in electromagnetism. The other methods can be applied to specic problems. For these methods it is necessary that the boundaries over
which the potential is specied coincide with the constant bounding surfaces.

8.2.1

Method of Separation of Variables

In cartesian coordinates the Laplace equation for the scalar potential can be
written as
2 2 2
+
+ 2 =0.
x2
y 2
z

(25)

Since x, y, z are independent variables, the solution of the Laplace equation

is of the form
(x, y, z) = X(x)Y (y)Z(z) .
Substituting this into the Laplace equation and dividing both sides by XY Z,
we obtain
1 d2 Y
1 d2 Z
1 d2 X
+
+
=0.
X dx2
Y dy 2
Z dz 2

107

This equation can be separated into three independent equations. To show

this, we write this equation as
1 d2 Y
1 d2 Z
1 d2 X
=

.
X dx2
Y dy 2
Z dz 2
Both sides of the above equation depend on dierent (independent) variables,
thus are equal to a constant, say 2 :
1 d2 X
X dx2

1 d2 Z
1 d2 Y

Y dy 2
Z dz 2

= 2
= 2 .

The second equation can be written as

1 d2 Z
1 d2 Y
= 2
.
Y dy 2
Z dz 2
Again, both sides depend on dierent variables, thus are equal to a constant,
say 2 :
1 d2 Y
Y dy 2
1 d2 Z
2
Z dz 2

= 2
= 2 .

Hence, after the separation of the variables, we get three independent ordinary dierential equations
1 d2 X
+ 2 = 0 ,
X dx2
1 d2 Y
+ 2 = 0 ,
Y dy 2
1 d2 Z
(2 + 2 ) = 0 .
Z dz 2
The solutions of these equations depend on whether 2 and 2 are positive
or negative. If we choose 2 and 2 to be positive, the solutions of the
108

dierential equations are

Ak eix + Bk eix

X(x) =
k

Cl eiy + Dl eiy

Y (y) =
l

Ep e

Z(z) =

2 + 2 z

+ F p e

2 + 2 z

The solutions can also be written in the form

[Ak sin(x) + Bk cos(x)]

X(x) =
k

[Cl sin(y) + Dl cos(y)]

Y (y) =
l

Z(z) =

Ep sinh

2 + 2 z + Fp cosh

2 + 2 z

The above solutions are in a general form, where the constants

, , Ak , Bk , Cl , Dl , Ep and Fp can be found from specic boundary conditions.
Consider two examples:
1. We have a solution with known boundary conditions, nd the problem.
2. We have a problem with specic boundary conditions, nd the solution.
Example 1.
Consider the following two-dimensional solution of the Laplace equation
(x, z) = X(x)Z(z) = V0 sin(x) sinh(z)
with the lower boundary min = 0 and the upper boundary max = V0 .
In what circumstance is the above the solution?
Consider (x, z) in some limits. X = 0 for x = 0 or x = , i.e. x = /.
The solution thus satises the boundary conditions along the vertical
lines for = /b.
109

V=0

V
0

x=0
z=0

0.282b z=0
x=b

Since sinh(z) = 0 for z = 0, the boundary condition along the lower boundary is satised.
For the solution to satisfy the upper boundary condition, the shape of
the upper boundary must be such that
V0 sin(x) sinh(z) = V0
for all points x, z on the line, i.e.
sin
Since sin

x
b

Hence, sinh

x
z
sinh
b
b

=1.

0 at the edges, it is greatest (= 1) at the center x = b/2.

z
b

must be equal to one at x = b/2. This happens when

z
= arc sinh(1) 0.885
b
0.885b
= 0.282b
z =

We usually have reverse problems to the above that we have a set of

electrodes which constitute equipotential lines or surfaces, and need to nd
the appropriate solution of the Laplace equation. This is illustrated in the
next example, where we will try to nd potential inside a rectangular box
whose three sides have potential equal to zero, and the remaining side has a
potential V0 .
110

Example 2.

x=b

V=0

x=0
z=0

V0

z=a

This two-dimensional problem has a general solution

(x, z) =

[An sin(x) + Bn cos(x)]

n

[En sinh(z) + Fn cosh(z)] .

The boundary condition = 0 at x = 0 can be satised by Bn = 0.
The boundary condition = 0 at z = 0 can be satised by Fn = 0.
To have = 0 at x = b we must have b = n.
Hence the solution reduces to

(x, z) =

Kn sin
n=1

nz
nx
sinh
b
b

where Kn = An En .
To nd Kn we apply the remaining boundary condition = V0 at z = a

(x, a) = V0 =

Kn sin
n=1

nx
na
sinh
b
b

This is a Fourier series in x and in the usual way we use the orthogonality
properties of sine functions to calculate Kn .
2
0

sin(m) sin(n) d =
111

0 for m = n
for m = n

2
0
2
0

cos(m) cos(n) d =

0 for m = n
for m = n

sin(m) cos(n) d = 0 for all m and n

Multiplying both sides by sin(mx/b) and integrating over x = 0 b,

we get
b
0

V0 sin

mx
dx =
b

na
b
n=1
b
nx
mx
sin
dx .
sin

b
b
0
Kn sinh

All integrals on the RHS are equal to zero except for m = n

V0 cos

nx
b

b
0

b
na
= Kn sinh
n
b

b
0

sin2

which we can write as

V0 b
na
[1 cos(n)] = Kn sinh
n
b
b 1
2nx

1 cos
b
0 2
The cos

2nx
b

nx
dx ,
b

dx .

integrates to zero over the range 0 b, giving

V0 b
na b
[1 cos(n)] = Kn sinh
n
b
2

and nally
Kn =

2V0 1 cos(n)
.
n sinh na
b

If n is an even number then cos(n) = 1 and Kn = 0.

If n is an odd number then cos(n) = 1 and 1 cos(n) = 2.
Hence
1
4V0
Kn =
for odd n
n sinh na
b

112

and then
(x, z) =
odd n

8.2.2

4V0 sinh
n sinh

nz
b
na
b

sin

nx
b

Solution of the Laplace Equation in Spherical Coordinates

In this lecture, we continue the discussion of boundary-value problems and

will illustrate solution of the Laplace equation for general problems of spherical symmetry.
In spherical coordinates (r, , ), the Laplace equation can be written as
2

r2
2 r
r
r

1
2
=0.
sin
+ 2 2
+ 2
r sin

r sin 2

Multiplying by r 2 , the Laplace equation can be written as a sum of two

separate parts

r2
r
r

1 2
sin
+
=0.
sin

sin2 2

The rst part depends only on r, whereas the second part depends only
on , . Thus, the solution is of the separable form
= R(r)Y (, ) .
Hence, substituting = R(r)Y (, ) and dividing by R(r)Y (, ), we
obtain
dR
1 d
r2
R dr
dr

1
Y

1
Y
sin
sin

113

1 2Y
sin2 2

Both sides of the above equation depend on dierent variables, thus must
be equal to the same constant, say :
d
dR
r2
+ R = 0
dr
dr
1 2Y
+
Y = 0 .
sin2 2

Y
1
sin
sin

Thus, the Laplace equation splits into two independent dierential equations.
We will call them (A) and (B).
(A) Consider the equation for Y .
Multiplying both sides by sin2 , we get
sin

sin

sin2 Y +

2Y
=0.
2

This equation contains two separate parts, one dependent only on and the
other dependent only on . Therefore, the solution will be of the form
Y (, ) = X()() .
Hence, we get
dX
d
1
sin
sin
X
d
d

sin2 =

1 d2
.
d2

As before, both sides must be equal to a constant, say m2 :

dX
d
1
sin
sin
X
d
d
2
1d
= m2
d2

sin2 = m2

(A1) First, we will solve the equation for , which we can write as
d2
= m2 .
d2
114

It is the familiar dierential equation for a harmonic motion. The solution

of this equation is of simple exponent form:
() = A exp(im) ,
where A is a constant.
To determine the constant m note that in rotation, and + 2 correspond to the same position in space: () = ( + 2), which is satised
when
exp(im) = exp[im( + 2)] .
This leads to
exp(i2m) = 1 ,
that is satised when m = 0, 1, 2, . . ..
Hence, the constant m2 is not an arbitrary number, is an integer.
(A2) Now, we will nd X().
Using the solution for , the dierential equation for X() can be written as
1 d
dX
sin
sin d
d

m2
X=0.
sin2

Introducing a new variable z = cos , we can rewrite this equation as

1 z2
or
d
dz

d2 X
m2
dX
X=0
+
2z
dz 2
dz
1 z2

1 z2

dX
m2
X=0.
+
dz
1 z2

The above equation is known as the generalized Legendry dierential equation, and its solutions are the associated Legendry polynomials. For m = 0,
the equation is called the ordinary Legendry dierential equation whose solution is given by the Legendry polynomials.
115

Lets look into the solution procedure of the above equation. This will
allow us to nd and X().
We assume that the whole range of z (cos ), including the north and south
poles (z = 1), is in the region of interest. The desired solution should be
single valued, nite, and continuous on the interval 1 z 1 in order to
represent a physical potential.
The dierential equation for X has poles at z = 1. In order to nd the
solution of this equation, we rst check what solution could be continuous
near the poles.
Lets check a possible solution near z = 1. Substituting x = 1 z, then
dx = dz and in terms of x the equation takes a form
d
dX
m2
x (2 x)
+
X =0.
dx
dx
x(2 x)
We look for a solution in the trial form of power series in x
X(x) = x

a n xn .

n=0

Substituting this into the dierential equation for X, we get

2s2 a0 xs1 + (s + 1)(2sa1 sa0 + 2a1 )xs + . . .
m2
+
(a0 + a1 x + . . .)xs = 0 .
x(2 x)
Near x 0, we can replace x(2 x) by 2x, and obtain
2s2 a0

m2
a0 xs1 + (. . .)xs . . . = 0 .
2

This equation is satised for all x only if the coecients at xs , xs1 , . . . are
zero. From this, we nd that
1
s = |m| .
2
1
We take only s = + 1 |m| as for s = 2 |m| the solution for X(x) at x = 0
2
would go to innity. We require the solution to be nite at any point x.

116

Thus, the solution that is continuous near x = 0 is of the form

1

X(x) = x 2 |m|

a n xn

n=0

or in terms of z

X(x) = (1 z) 2 |m|

an z n .

n=0

Using the same procedure, we can show that near the pole z = 1, the
continuous solution is

X(x) = (1 + z) 2 |m|

an z n .
n=0

Hence, we will try to nd the solution in the form

X(x) = 1 z 2

1
|m|
2

bn z n .

n=0

Substituting this equation into the dierential equation for X(z) and collecting all terms at the same powers of z n , we obtain
n

{(n + 1)(n + 2)bn+2 n(n 1)bn

2(|m| + 1)nbn ( + |m| + m2 )bn z n = 0 .

Hence, we get a recurrence relation for the coecients bn
bn+2 =

(n + |m|)(n + |m| + 1) +
bn .
(n + 1)(n + 2)

We have two separate solutions for even and odd n. For b0 = 0, we put
b1 = 0, and the solution is given in terms of even n. For b0 = 0, we put
b1 = 0, and the solution is given in terms of odd n.
We cannot accept both the even and odd solutions at the same time, because
in this case the solution X(z) would not be a single valued function.
For example, for b0 = 0, we have = |m| m2 , but for b1 = 0, we have
= 2 3|m| m2 . If we would accept both of the solutions at the same
time, the potential would have two dierent values.
117

We check now whether the series is converting when n which would

ensure that the potential is nite.
Since bn+2 > bn , the series diverges for z = 1. Therefore, in order to get
the potential nite everywhere in the space, we have to terminate the series
at some n = n0 . In other words, we assume that bn0 +1 = bn0 +2 = . . . = 0.
The series terminating at n = n0 indicates that
(n0 + |m|)(n0 + |m| + 1) + = 0 .
Introducing
l = n0 + |m| ,
we see that l |m|, and
= l(l + 1) ,

l = 0, 1, 2, . . .

Thus, the solution for X(z) is

Xlm (z) = 1 z 2

1
|m|
2

l|m|

bn z n ,
n

where the sum is over even n when l |m| is an even number, and over odd
n when l |m| is an odd number.
First few solutions
0
X00 (z) = b0 = b0 P0 (z)
0
X10 (z) = b1 z = b1 P1 (z)

1
X11 (z) = b0 1 z 2 = b0 P1 (z) ,

0
0
1
where P0 (z) = 1, P1 (z) = z, P1 (z) = 1 z 2 , . . . are the associate Legendry
polynomials of the order l.
Useful examples [in terms of (z = cos )]:
0
P0 (cos ) = 1
0
P1 (cos ) = cos
1
P1 (cos ) = sin

118

1
[3 cos(2) + 1]
4
3
1
P2 (cos ) =
sin(2)
2
3
2
P2 (cos ) =
[1 cos(2)]
2
0
P2 (cos ) =

Orthogonality of the Legendry polynomials is an important property:

1
1

n
Plm (cos )Pk (cos ) d(cos ) = 0

unless m = n and l = k, for which

1
1

[Plm (cos )]2 d(cos ) =

2
(l + m)!
2l + 1 (l m)!

With the above notation, the solution for Y (, ) is of the form

Alm Plm (cos )eim

Y (, ) =
l

(B) What left is to nd the radial part R(r).

With = l(l + 1), the dierential equation for R is of the form
dR
d
r2
l(l + 1)R = 0 .
dr
dr
Dividing by r and introducing a new function U (r) = rR(r), we obtain
d2 U
l(l + 1)

U =0.
2
dr
r2
Lets rst check the asymptotic solution for r
1. In this case we can ignore
the second term in the dierential equation, and nd that the asymptotic
equation has a solution U (r
1) = Cr, where C is a constant.

119

Following this asymptotic behavior, we will try the general solution of a

form
U (r) = r s .
Substituting this into the dierential equation, we obtain
[s(s 1) l(l + 1)] r s2 = 0 .
This equation is satised for any r when
s = (l + 1)

or

s = l .

Thus, the general solution is of the form

U (r) = C1 r l+1 + C2 r l ,
or
R(r) = C1 r l + C2 r (l+1) .
Thus, solution of the Laplace equation in spherical polar coordinates is of
the form:
C1l r l + C2l r (l+1)

(r, , ) =
l

Alm Plm (cos )eim .

The solution can be written as
C1l r l + C2l r (l+1)

(r, , ) =
l

[alm cos(m) + blm sin(m)] Plm (cos )} .

Example 1: Potential inside an uniformly charged sphere

Consider an example of boundary-value problem with azimuthal symmetry:
A conducting sphere of a radius a in an uniform electric eld.

120

The conducting sphere is an equipotential volume (else there would be

electric elds driving currents till it became so). Take the zero potential on
the sphere.
The boundary conditions to be satised are:
1. The potential on the surface of the sphere V (a, , ) = 0.
2. The potential at innity is the uniform eld potential (no eect of the
sphere at innity), so = Er cos at innity.
Since the applied potential is independent of the angle phi, the induced
potential will also be independent of . Thus, we can set m = 0 in the
general solution and get
C1l r l + C2l r (l+1) Pl0 (cos ) .

(r, ) =
l

The boundary condition at innity is satised for all constants C1l = 0 except
for l = 1 (remember P1 = cos ).
(r, ) = C11 rP1 (cos ) +
l

C2l Pl (cos )
.
r l+1

As r , the potential (r, ) C11 r cos = Er cos .

Therefore C11 = E.
121

The other boundary condition to be satised is (a, ) = 0 on the surface

of the sphere.
(a, ) = 0 = EaP1 (cos ) +

C2l Pl (cos )
.
al+1

We can determine the coecients C2l using the orthogonality properties of

the Legendry polynomials. Multiplying the above equation by Pk (cos ) and
integrating over cos , we obtain
1

0 = aE

P1 (cos )Pk (cos ) d(cos )

C2l
al+1

+
l

1
1

Pl (cos )Pk (cos ) d(cos )

If k = 1:
The rst term vanishes by orthogonality of the Legendry polynomials.
All terms in the summation vanish except that for l = k. Thus
0 =

C2k
ak+1

1
1

Pk (cos )Pk (cos ) d(cos )

Since the integral is nonzero, then C2k = 0 for k = 1.

For k = 1:
1
1

Plm (cos )Plm (cos ) d(cos )

2
(l + m)!
2l + 1 (l m)!

which for l = 1 and m = 0 is equal to 2/3. Thus

2 C21 2
0 = aE + 2
3
a 3
from which, we nd C21 = Ea3 . Hence
(r, ) = Er cos + Ea3
122

cos
.
r2

The rst term is just the potential of a uniform eld E. The second term
is the potential due to the induced surface charges or, equivalently, is the
potential of the induced dipole moment p = 40 Ea3
dip =

p cos
.
40 r 2

More useful exercises: Before you approach the tutorial problems on this
subject (tutorial set number 6), try to solve the following problems
1. Consider a two-dimensional region with boundaries at x = 0, b and
z = 0, a, as shown in the gure 3. The boundary conditions are

= 0
z
= 0
= V0

at

z=0,

at
at

x = 0, b ,
z=a.

Find the potential at any point inside the two-dimensional region.

X
=0

b
=0
z
0

=V
0
=0

Figure 3:
2. Consider a conducting sphere of radius R. The surface of the sphere is
kept at a potential
(R, , ) = V0 sin sin .
123

Using the above as a boundary condition, nd the potential at any

point inside the sphere.

Questions
(1) Explain, why in the two-dimensional case, we take from the general
three-dimensional solution of the Laplace equations
= X(x)Z(z) ,
or
= Y (y)Z(z) ,
but not
= X(x)Y (y) .

124

Electromagnetic Antennas: Hertzian Dipole

In this lecture we will illustrate an application of the vector potential A in

calculations of the electric and magnetic elds produced by a source system
containing time varying charges and currents.
Consider the retarded solutions of the Maxwells equations
B=

A ,

E =

A
,
t

with
(r, t) =

(t r/c)
dV ,
r

1
40

1
J(t r/c)
dV .
2
40 c
r
The above solution holds for the Lorentz gauge in which
A(r, t) =

1
.
c2 t
Assume that the charges and currents vary sinusoidally in time
A=

(r, t) = (r)eit
J(r, t) = J(r)eit .
In this case
A=

i
,
c2

which gives
c2
A .
i
Thus, the scalar potential can be eliminated from the eld equations
leaving only the dependence on A. Hence, we can express both E and B in
terms of the vector potential A alone. We have
=

E=

c2
(
i

A)

A
,
t
125

B=

A .

This result may seem rather strange at rst, since normally we should expect
to need both the scalar and vector potentials in order to completely determine
the eld. The explanation and in fact an another way of saying the same
thing is that time varying charge must satisfy the continuity equation
J =

= i ,
t

so that
=

J
.
i

Then
1
40

(t r/c)
dV
r
1
J
=
dV
40 i
r
c2
1
J

dV =
=
40 i
r
i

(r, t) =

A .

Thus, specication of J alone is sucient to completely determine all sources,

and hence a solution for A in terms of J contains all the necessary information to completely specify the time-varying eld.

9.1

Field of an Element of Alternating Current

Consider a linear element l carrying an alternating current I = I0 exp(it).

The current element may be viewed as two charges Q and Q oscillating
back and forth.
Assume that l is much smaller than the wavelength = 2c/. In this
case we can ignore the phase variation along l.
An understanding of the properties of such an antenna is of great interest since, in principle, all radiating structures can be considered as a sum of
small radiating elements. Moreover, many practical antennas working at low
126

z
r

I0

ei t

l
y

frequencies are very short compared with the wavelength.

If l

, then there are three spatial regions of interest:

The near eld (static) zone

The intermediate eld (induction) zone

The far eld (radiation) zone

We will see that the elds have dierent properties in the dierent zones.
In the near zone the elds have the character of static elds, with a strong
dependence on the properties of the source. In the far zone, the elds are
transverse to the radius vector and fall of as r 1 , typical of radiation elds.
The retarded current element
J(t r/c)dV = I(t r/c)dl = dlI0 ei(tkr) ,
where k = /c.
Thus, the vector potential is
A = l

I0 ei(tkr)
,
40 c2
r

where l = dl.
127

In the near zone, where r

(or kr
1), the exponent exp(ikr) can
be replaced by unity. In the far eld zone kr
1, the exponential oscillates
rapidly, and in this region it is sucient to approximate exp(ikr) 1ikr.
In the intermediate zone, all powers of kr must be retained.
In cartesian coordinates the vector potential has only one component,
say Az along the current element.
Referred to spherical coordinate system, A has components
Ar = Az cos =

I0 l cos ei(tkr)
,
40 c2
r

A = Az sin =

I0 l sin ei(tkr)
,
40 c2
r

(26)

A = 0 .
In order to nd the elds E and B, we have to calculate
that in spherical polar coordinates are given by
A =
+

A =

A and

A,

1 (A sin )
1 (r 2 Ar )
+
2
r
r
r sin

1 A
r sin
r

(A sin ) A

r sin

1 Ar (rA )

r sin
r

(rA ) Ar

r
r

Since A = 0 and there is no dependence of Ar and A , i.e. Ar, / = 0,

the above equations reduce to
A =

1 (A sin )
1 (r 2 Ar )
+
,
2
r
r
r sin

128


(rA ) Ar

.
r
r

A =

Hence, the magnetic eld of the current element is

B=

A = B ,

where
1 (rA ) Ar

r
r

B =

(27)

and Br = B = 0.
Thus, the magnetic eld is perpendicular to the radius vector at all distances.
Calculate the magnitude B . Substituting Eq. (26) into Eq. (27), we
obtain
B =
=

I0 l
40 c2 r

sin ei(tkr)
r

i(tkr)

cos

I0 l
ei(tkr)
ik sin ei(tkr) + sin
40 c2 r
r

Hence
B =

I0 l ik
1
+ 2 sin ei(tkr) .
2
40 c r
r

The magnetic eld is composed of two terms: the near zone term 1/r 2
and the far zone term 1/r. In the limit of 0, the near zone term
reduces to the Biot-Savart formula. The far zone term is only present for
an oscillating eld ( = 0) and therefore it is radiation eld arising from
accelerated (oscillating) charge.
Calculate electric eld of the current element:
E=

c2
(
i

A)
129

A
,
t

where
A

= i A = i Ar r + A ,

t
and
A =
=

1 (r 2 Ar )
1 (sin A )
+
2
r
r
r sin

1 r cos ei(tkr)
I0 l
40 c2 r 2

1
ei(tkr)
sin2
r sin
r
I0 l cos i(tkr)
=
e
ikrei(tkr)
40 c2
r2

2 sin cos ei(tkr)

r 2 sin
I0 l 1
ik
cos ei(tkr) .
=
+
2 r2
40 c
r

Since in spherical coordinates

=r

1 + ,
+
r
r r sin

we obtain
I0 l cos
ik
2
k 2 ik i(tkr)
2 e
2 3+
r
40 c2
r
r
r
r
2ik k 2 i(tkr)
I0 l cos 2
+ 2
e
,
=
40 c2
r3
r
r
A

I0 l 1
ik
+ 2 sin ei(tkr) ,
2 r3
40 c
r
A

130

=0.

Hence the radial part of the electric eld, Er , is

Er

c2
=
i

A
A)
r
t

2ik k 2 i(tkr)
I0 l cos 2
+ 2
=
e
40 i r 3
r
r
I0 l cos ei(tkr)
i
40 c2
r
2
2
ik i i(tkr)
I0 l cos
+ 2+

=
e
3
40 c
ikr
r
r
cr
Since k = /c, the 1/r terms cancel and then Er simplies to
2
I0 l cos 2
+ 2 ei(tkr) .
40 c
ikr 3 r
Similarly, we nd the component of E as
Er =

c2
=
i

A
A)

ik
I0 l sin ei(tkr)
I0 l sin 1
+ 2 ei(tkr) + i
40 i r 3 r
40 c2
r
I0 l sin
1
1
ik i(tkr)
=
e
+ 2+
40 c
ikr 3 r
r
=

The radial part of the electric eld contributes only to the near and intermediate zones, whereas the angular part contributes to all of the zones.
The 1/r 3 part is the Coulomb type contribution. It is similar in nature
to a static eld surrounding a small linear-current element and an electric
dipole.
Proof:
The Coulomb or static eld is for 0. In this limit the 1/r 3 contribution is
I0 l sin 1 i(tkr)
E =
e
40 c ikr 3
131

I0 l sin 1
1
1 ikr + (ikr)2 +
3
40 c ikr
2
I0 l sin 1
=
40 c r 2
=

where we have taken only the real (physical) part of the eld.
Since I0 = q/t and l/t = c, we get
E =

q sin 1
,
40 r 2

as required.
Electric and magnetic elds in near and far eld zones
Consider rst the near eld zone (r
electric elds are

). In this limit the magnetic and

I0 l 1

sin ei(tkr)
40 c2 r 2
I0 l 2

= i
cos + sin ei(tkr)
r
40 c kr 3

Bnear =
Enear

Since the magnetic eld is real and the electric eld is imaginary, the Pointing
vector involving the near-zone eld components is a pure imaginary quantity.
It does not represent any ow of energy. This imaginary quantity represents
energy that oscillates back and forth between the source and the region of
space surrounding the source.
Consider now the far zone or radiation components of the magnetic and
electric elds.
R
Erad = E ,
R
Brad = B ,

I0 l sin ik i(tkr)
e
40 c r
I0 l sin ik i(tkr)
R
B =
e
.
40 c2 r

R
E =

Note that:
132

1. The electric and magnetic elds oscillate in phase.

2. The electric and magnetic elds are orthogonal to each other.
3. The ratio

R
E
R
B

= c, the value for plane waves in free space.

4. The electric and magnetic elds are transverse to the radius vector at
all distances.
5. The Poynting vector N = c2 0 Erad Brad is a real quantity and is in
the direction of the radius vector, indicating that the energy of the eld
propagates away from the current element.
These properties show that in the far zone the eld is in a form of plane
waves.

9.2

Power Radiated from the Current Element

The power ux at any point is given by the Poynting vector

N = c 2 0 E B .
Then, the total power radiated across a sphere of radius r is
W =

N dS ,

where
dS = r 2 sin dd .
Only those partial products in E B which vary as 1/r 2 will have net radiated power. The other partial products are small as they fall o more rapidly
than 1/r 2 . Thus, the only part of the elds entering into the expression for
the radiated power is the far eld zone part (radiation component) consisting
of the terms varying as 1/r.

133

The volume I dl is energized and the energy ows but there is no net
energy loss over a cycle in the equilibrium situation.
The radiation components of E and B are in phase and average over
time is
E R R =
B

1
R
E
2

R
B

R
R
R
R
where E
and B are amplitudes of E and B .
0
0
Hence, the time averaged Poynting vector is

1 2
I 2 l2 k 2

N =
c 0 0 2 2 3 2 sin2
2
16 0 c r
=

2
I0 l2 4 2
I 2 l
sin2 = 0
32 2 0 c 2 r 2
80 c

sin2
.
r2

Thus, the total power is

W =
=

0
2
0

r 2 N sin dd

2
I0 l
80 c

sin3 d .

Integrating, we get
W = 2

2
I0 l
80 c

2
I0 l
4
=
3
30 c

We can write the total power radiated in terms of the power absorbed in an
equivalent resistance, called the radiation resistance:
1 2 l
W =
2 30 c

1 2
2
I0 = RI0 .
2

where
2 l
R=
30 c
134

is the radiation resistance.

Since 1/(0 c) =

0 /0 = 377 or 120, we obtain

R = 80

For example, if l/ 0.1, then R = 0.8 2 8 ohms.

This example shows that for a current element which is 10% of the wavelength long, the resistance is very small.
Thus, if l/
1, the radiation losses are negligible, that the radiated
power is very small.
This also explains why ordinary circuit theory often works so well although it ignores the loss of energy by radiation from AC currents.
Appreciable power would be radiated only if the current amplitude I0
were very large. A large current, on the other hand, would lead to large
amounts of power dissipation in the conductor, and hence a very low eciency.
We can conclude that current carrying systems that have linear dimensions small compared with the wavelength radiate negligible power. An efcient antenna should have dimensions comparable to or greater than the
wavelength.

9.3

Gain of the Dipole Antenna

A further property of the dipole antenna that is worthy of consideration is

the directional property of power radiated in dierent directions.
The gain or directivity function of a transmitting antenna is the ratio of
the Poynting ux to the ux due to an isotropic radiator emitting the same
total power W :
gT (, ) =
135

N (, )
,
Niso

where
Niso =

W
4r 2

is the energy ux uniform in all directions.

For the innitesimal dipole:
2
I0 l
N (, ) =
80 c

2
I0 l
W =
30 c

sin2
,
r2
.

Hence
gT (, ) = gT () =

3 2
sin .
2

sin2

Figure 4: The polar radiation pattern of the dipole antenna.

The directivity function gT (, ) denes a three-dimensional surface called
the polar radiation pattern of the antenna. The function varies as sin 2 , and
hence the radiation is most intense in the = /2 direction and zero in the
directions = 0, . The maximum gain then is 1.5 for directions dened by
= /2, in the equatorial plane of the dipole. The gain function is independent of .
We can conclude, that the directivity function gT (, ) is a measure of
how eective the antenna is in concentrating the radiated power in a given
direction.
136

Questions:
(1) Show that in spherical polar coordinates, the magnetic eld of a short
current element I l = lI0 exp(it) has only an azimuthal component of
the form
B =

I0 l ik
1

+ 2 sin ei(tkr) .
40 c2 r
r

(2) Show that in the far eld zone of a radiating short current element,
the electric and magnetic elds oscillate in phase and are orthogonal to each
other.
(3) Given the expressions for the EM eld of a Hertzian dipole, show
that the total radiated power from the dipole is
2
I0 l
W =
30 c

(4) Show that the time averaged Poynting vector of the EM eld emitted
by a short current element is maximal in the equatorial plane of the element.

137

10

Electromagnetic Theory of Polarizable

Materials

If an uncharged dielectric (insulator) is placed in an electric eld, the eld

within the dielectric will be found to be modied by charges induced by the
external eld. The charge is induced by rearrangement of bound charges
within the molecules of the dielectric. In dielectrics these charges are a set
of molecular dipoles. The dipole may exist permanently or may be induced
by the external eld. Dielectrics with permanent dipole moments are usually
electrically neutral due to random orientation, in the absence of electric elds,
of the dipole moments. An example is H2 O. We will study electromagnetic
theory of polarizable materials in terms of dipoles, which we will treat as
building blocks of dielectric materials.

10.1

Potential and Electric Field of a Single Dipole

Mathematically, it is convenient to deal with the dipole not as just a pair of

individual plus and minus charges, but as a separate object on its own rights.
Suppose that opposite charges q are separated by a distance d. We
will nd the potential at a distance r and angle under the assumption
that r
d, called the dipole potential.

+q
r1

-q

d cos
r2

A
We dene dipole moment as the product of the charge times the separation
p = qd .
138

Since r2 r1 d cos at r

d, we get for the potential at point A

q
q

40 r1 40 r2

q r2 r 1
q d cos
1 pr
.
=
=

2
40 r1 r2
40 r1 r2
40 r

(28)

We can nd electric eld of a dipole using E = . Since the potential

of the dipole depends on r and , it is convenient to work in the spherical
coordinates in which the electric eld is given by

E = Er r + E + E ,

where

1 2p cos
=
,
r
40
r3
1
1 p sin
=
=
,
r
40 r 3
1
=0.
=
r sin

Er =
E
E
Hence
E=

2 cos r + sin .

40 r 3

Figure 5 shows a sketch of the electric eld lines of an electric dipole moment.

p
E

Figure 5: The electric eld lines of a dipole moment.

139

10.2

Polarization Vector

If there are N dipoles per unit volume, the dipole moment per unit volume is:
N

P =

pi ,
i=1

and is called the polarization.

^
r

dV

The electric potential set up by an arbitrary volume distribution of electric

dipoles can be calculated by using the potential produced by a single dipole,
Eq. (28), and the above denition of P . The electric potential at an arbitrary
distance r from a volume element dV containing such dipoles is
d =

1 P r

dV ,
2
40 r

where r is the unit vector from dV toward X, and we have assumed that r

is much larger than the extent of the volume element dV .

Let r be the unit vector from X toward dV . (We want to integrate over V

with the position of X xed). In this case, we change r , and obtain

r
d =

1
P 2
40
r

1
40

P
140

dV

dV .
r2

The result can be transformed into one that has an interesting physical
interpretation.
Noting that

=
r2

1
,
r

we have
P
Applying a vector identity
P

1
=
r

A+A

P

r

, we thus obtain

P
.
r

P
1
dV +
r
40

1
.
r

=P

(A) =

Hence
1
=
40

r2

P
dV .
r

Transforming the rst term using Gauss divergence theorem, the potential
becomes
1
=
40

P n

1
dS +
r
40

P
dV .
r

(29)

On comparing of Eq. (29) with the general form of the potential, Eq. (20), we
deduce that the rst term on the rhs of the above equation is the potential of
a surface charge density P n. The second term is the potential of a volume

charge density P .

10.3

Maxwells Equation for

E in a Dielectric

In general, the electric eld in a dielectric can be found from the Maxwells
equation I:

E =
.
0
141

+ + + +
+ + + +
+ + + +
+ + + +
+ + + +
+ + + +
+ + + +
+

+
+ + +
+
+
+
+
+
+ +
+
+

<

Figure 6: Surface (left picture) and volume charges (right picture). Surface
charges exist because there are no neighboring charges at the end surfaces of
the material to cancel them out. Volume charges exist because the number
of dipoles per unit volume changes, that there is an incomplete cancellation
of charge density from adjacent dipoles.
In the dielectric it is convenient to (mentally) separate the polarization
charges from whatever other charges might be there also. The other charges
are usually referred to as the free charges or conducting charges to distinguish
them from the bound charges in the dielectric.
We can write
0

E = f + p ,

where f is the free charge density and p is the polarization charge density
throughout the volume. If we express p in terms of P (p = P ), we
have
0

E = f

P .

Hence
(0 E + P ) = f .

(30)

Now it is common practice to drop the subscript f , but one must remember
that now stands for the charge density not counting the polarization charges.

142

It is common practice to dene

D = 0 E + P ,

(31)

where D is called the dielectric displacement.

The reason for the name dielectric displacement can be easily understood if we refer to the Maxwells theory.
Take time derivative of both sides of Eq. (31):
E P
D
= 0
+
.
t
t
t
We know from the Maxwells theory that the rst term on the rhs of the
above equation represents displacement current density, and the second term
is the polarization current density. Therefore, D/t can be called a generalization of the displacement current density, and then D can be regarded as
the dielectric displacement.
Introducing the dielectric displacement, Eq. (30) becomes
D = .
We can read this equation that the source for the eld D is the free charge
density .
We turn now to a consideration of the macroscopic eects of the polarizability of dielectric materials. We will consider only ideal dielectrics.
Ideal dielectrics can be divided into following categories:
1. Homogeneous properties independent of the position.
2. Isotropic properties independent of direction.
3. Linear polarization proportional to E.
4. Stationary properties independent of time.

143

Case of simple isotropic and linear dielectrics

Ordinary dielectrics (glass, teon, plastics etc.) are linear in polarization
for elds not strong enough to cause dielectric breakdown i.e. P E.
For these materials, we can write
P = 0 E
where is called the dielectric susceptibility. Then, we can write
D = 0 E + P = 0 (1 + )E = 0 r E = E ,
where r is the relative permittivity or dielectric constant, and is the permittivity. Hence
D =

(E) =

or
E =

if is independent of position i.e. if is a permittivity of a homogeneous

dielectric.
The same result would be obtained by replacing 0 in Coulombs law by .
F =

1 q1 q2
r.

4 r 2

The ratio /0 then represents the relative shielding of q1 from q2 by the

polarization charges induced in the medium.
The theory of the molecular structure of a material will yield an estimate
of and . The dipole moment p of a molecule will be proportional to the
local electric eld E so we can dene a molecular susceptibility m such that
p = m 0 E.
Then
p i = N m 0 E ,

P =
i

144

where N is the number of molecules per unit volume.

Thus
= 0 r = 0 (1 + N m )
and is a quantity directly measurable from the measurement of capacitance
C=

A
0 A
= r
= r C0 ,
d
d

where C0 is the capacitance without the dielectric.

We can summarize, that lling a capacitor with dielectric multiplies its capacitance by r .

Exercise in class: Capacitor lled with a homogeneous dielectric

A plane parallel capacitor has charges + and per unit area on its plates.
The capacitor is lled with a homogeneous and linear dielectric of dielectric
constant r = 1 + .
Show that:
1. The electric eld within the dielectric is:
E=

.
r 0

2. The polarization charge per unit area on the surface of the dielectric
adjacent to the surface of the negatively charged plate is:
s =

.
1+

3. The capacitance of the capacitor is C = r C0 where C0 is the capacitance of the same capacitor without the dielectric (i.e. a vacuum or air
between the plates).

145

10.4

Dense Dielectrics: The Clausius-Mossotti Relation

The Lorentz theory of polarizability of dense dielectric materials distinguishes

between the mean electric eld E and the local electric eld Eloc as seen by a
typical dipole. The typical dipole is considered to be at the centre of a small
sphere that has been excavated from the dielectric. Eloc is thus the mean
eld E minus Eplug where Eplug is the eld of the spherical volume excavated.
We will show that:
Eloc = E +

P
.
30

If there are N molecules per unit volume each of polarizability the

poarization vector:
P = N Eloc
and by denition:

P
,
= N E +
30

P = (r 1)0 E .
Let E be the mean eld throughout the dielectric.
Let Eplug be the eld due to the spherical plug alone.
Let Eloc be the eld in the spherical hole.
E = Eloc + Eplug

and from above Eplug =

P
.
30

Thus
Eloc = E +

P
.
30

The argument now is that each molecule is at the centre of a small hole
and the eld acting on the molecule is thus Eloc . If is the molecular
polarizability, its induced dipole is thus:
p = Eloc
146

If there are N molecules per unit volume then:

P = N p = N Eloc

P
= N E +
.
30

By denition P = 0 E = (r 1)0 E. Substituting for P :

(r 1)0
E
30
N
r 1
(r + 2)
(r 1)0 = N 1 +
=
3
3
r 1
N
=
,
r + 2
30

(r 1)0 E = N E +

which is known as the Clausius-Mossotti relation for a dense dielectric.

We can solve the Clausius-Mossotti equation for r , and obtain
N
30
2N
= 1+
.
30

r 1 = (r + 2)
r 1

N
30

Note that
r =

2N
30
N
30

1+
1

as

N
1,
30

which is called the Clausius-Mossotti catastrope.

Removal of the plug leaves polarization charges, whose eld tends to line
up the dipole parallel to the external eld. The system is self-polarizing
(Clausius-Mossotti catastrope) if N 1.
30

10.5

Time Dependent Fields and the Complex Dielectric Susceptibility

The induced polarization charges do not produce any currents inside the
dielectric. There is no DC current in response to a DC electric eld, but if P
147

is changing with time (because E is changing with time) there will be an AC

current density:
J=

s
P
n=

.
t
t

Thus, P /t plays the role of polarization current density.

Let us consider what will happen if a dielectric is introduced into a timevarying electric eld.
With an alternating electric eld (due say to an AC voltage across a capacitor lled with a dielectric) the polarization P may lag in phase behind
the driving eld E. This means there is internal friction and heat dissipation as is discussed below. The capacitor will exhibit resistive as well as
capacitive properties.

+Q

-Q

I
E

d
Consider the work done in charging a capacitor. We start from the circuit
theory quantities and will express them in terms of the eld quantities.
The work done in charging the capacitor is
dW
dQ
= V I = Ed
,
dt
dt
where V is the voltage.
148

Since
Q = CV = CEd =

A
Ed , = AE = AD
d

we obtain
Q = A(0 E + P ) ,
and then
dW
d
d
= Ed A(0 E + P ) = E dA (0 E + P ) .
dt
dt
dt
Since dA = V is the volume of the capacitor, we can write
dW
=
dt

E
V

dP
d
(0 E) + E
dt
dt

dV ,

or
dW
=
dt

d 1
0 E 2 dV +
dt 2

E
V

dP
dV ,
dt

where we took into account a possibility that the electric eld and polarization can vary across the capacitors plates.
The rst term in the above equation is the rate of doing work building
up E eld. The second term is the rate of doing work on the dipoles by E.
Thus, the supplied energy to the capacitor is used to build up the electric
eld inside the capacitor and to polarize the dielectric. Consider separately
both terms.
First term:
If E = E0 cos(t), the rst term takes the form

E0 cos(t) 0 E0 sin(t) dV .

The element of work done per unit volume unit time is

dW
2
= 0 E0 cos(t) sin(t)
dV
149

Averaging over a cycle, we get

dW
2
= 0 E0
dV

2/

cos(t) sin(t) dt = 0 .
t=0

Work is done building up the eld in one part of the cycle but the stored
energy is given back in another part.
Second term:
If P = 0 E = 0 E0 cos(t) the same zero net energy conversion averaged over a cycle will happen with this term. If there is internal friction
there will be a phase dierence between P and E.
Write
P = 0 E0 cos(t + )
Then
P = 0 E0 cos cos(t) 0 E0 sin sin(t)
and
dP
= 0 E0 cos sin(t) 0 E0 sin cos(t)
dt
Hence, the work done per unit volume per cycle will be
dW
dV

2/

2
0 E0 cos cos(t) sin(t) dt
0
2/
2
0 E0 sin cos2 (t) dt .
0

Since the rst integral is zero, we get

dW
2
= 0 E0 sin
dV

2/

150

cos2 (t) dt .
0

The integral on the rhs of the above equation is positive and dW/dV must
be positive corresponding to energy dissipation (or the dielectric would keep
getting energy from its interior and building up the eld with it).
Thus, sin must be negative, so < < 0.

10.6

The Complex Susceptibility and Permittivity

In the usual way use complex exponentials to represent amplitude and phase.
We can write
E = E0 eit

then P = P0 ei(t)

( is positive) .

We can write the complex polarization in dierent forms

P = P0 ei eit = (P0 cos iP0 sin )eit ,
P =

P0
P0
cos i sin E0 eit ,
E0
E0

P = 0 ( i )E0 eit = 0 c E0 eit = 0 c E ,

where c = i is a complex susceptibility.
With the complex polarization, the dielectric displacement takes the form
D = 0 E + P = 0 E + 0 c E = 0 (1 + c )E ,
or
D = 0 (1 + i )E = 0 r E = c E ,
where c is a complex permittivity, and r is a complex relative permittivity
or dielectric constant
c = 0 (1 + ) i0 .
151

Example - equivalent circuit of the lossy capacitor

The imaginary part of the dielectric susceptibility corresponds to net energy dissipation i.e. in the circuit theory language the dielectric susceptibility
adds a resistive component to the capacitor. The material lling the capacitor could also have some ordinary ohmic conductivity (due to the presence
of free charges as well as bound charges in the material). Let us calculate
the magnitude of the resistance.

+Q = SA

Ip

-Q = -SA

Ic

Let I be the charging current in the external circuit, and

Ip =

Jp d A = A

dP
dt

be the polarization current in the dielectric. Let

Ic =

Jc dA = AE = A

V
d

be the conduction current in the dielectric due to its nite conductivity .

If S is the charge per unit area on the plates then:
S = (charge supplied by I)
(charge removed by Ip )
(charge removed by Ic ) .
Then
Q = SA =

I dt
152

Ip dt

Ic dt

and
E=

S
V
=
0
d

Hence
S = 0

V
d

and also remember that P = 0 E = 0 V , we obtain

d
Q = SA = 0

V
A=
d

Idt

Ip dt

Ic dt

or taking a derivative in time, we get

0

A dV
d dt

= I I p Ic
= I A

V
0 dV
A
d dt
d

Putting
dV
= iV
dt

and = i

and solving for I, we get

I=

0 A

(1 + i )i +
V .
d
0

Separating real and imaginary parts and putting C0 = 0 A/d (the capacitance there would be if the dielectric were lossless):
I = C0

+
0

+ iC0 (1 + ) V .

Since the capacitor transmits some charges through the internal dielectric,
in the circuit theory this system is equivalent to a parallel circuit
I=

1
+ iC
R
153

V .

C0

Comparing with the above result for current ow in the lossy capacitor we
see that the eective capacitance is C0 (1 + ) and the eective resistance is
R=

1
C0

where we remember that C0 is the capacitance in the absence of losses.

10.7

Added Note - The Loss Tangent

The properties of a dielectric material are usually specied by giving its dielectric constant K, and its loss tangent tan .
If we write:
I = iC0 (1 + ) i +

V .

The quantity in [ ] brackets can be dened as a complex relative permittivity r :

).
r = 1 + i( +
0
This is a generalization on the previous denition of complex relative
permittivity to include the eects of ohmic conductivity. Then we can dene
154

a generalized permittivity and dielectric constant = 0 r .

Now if we write:
r = K ei = K (cos i sin )
= K cos (1 i tan ) = K(1 i tan ) .
Then the standard form is
= 0 K(1 i tan ) ,
where tan is the loss tangent and is read out on some AC bridges as an
alternative to reading out the resistive property of a lossy capacitor.

Since K cos = (1 + ) and K sin = ( + 0 ) it follows that:

tan =

[1 + ]

In this equation, tan includes the eects of nite conductivity and the eects
of polarization damping force.

Questions:
(1) Show that the electrostatic potential due to a distribution of electric
dipoles of moment per unit volume P throughout a volume V enclosed by a
surface S is that of a volume charge density P together with a surface
charge density P n.

(2) Show that the polarization of a dielectric driven by a time varying

electric eld lags in phase the driving eld.
(3) Show that the phase dierence between the polarization and a timevarying electric eld results in a complex permittivity of the dielectric.
155

11

Magnetic Fields in Magnetizable

Materials

We have previously discussed how the polarization of dielectrics by an externally applied electric eld is equivalent to volume and surface distribution
of charge. Analogously, a magnetic eld can act on molecular scale current
loops to produce macroscopic eects. It was Amp`re who rst suggested that
e
the magnetism of matter was due to the cooperative eects of currents circulating in atoms (and not, as previously thought, due to a separate magnetic
charge called poles).

^
n
A

11.1

Magnetic Polarization Currents

We dene the magnetic moment of a current loop as

= IA ,
n
which is equal to the product of the area of the plane loop and the magnitude
of the circulating current. The vector direction n of the moment is perpen
dicular to the plane of the loop and along the direction set by the right-hand
rule.
In a material body we dene a macroscopic dipole moment per unit volume
i ,

M=
i

which is called magnetization.

156

Theorem
If a volume V enclosed by surface S has a magnetic dipole moment per
unit volume M (which may be a function of position), the macroscopic magnetic elds so produced are equivalent to those of:
A volume current density JV =

M.

A surface current density JS = M n.

Proof:
The theorem is proved by showing that the vector potential due to the
dipole distribution in a volume V closed by a surface S can be written in the
form
A=

1
40 c2

1
M
dV +
r
40 c2

M n

dS .
r

From the solution of the Maxwells equations (static) vector potential of

a current loop is given by
A=

1
40 c2

J
I
dV =
r
40 c2

dl
.
r

Consider a current loop of radius a, shown in Fig. 7.

In polar spherical coordinates

dl = ad = a sin d + a cos d .
i
j
However
r = (x a cos )2 + (y a sin )2 + z 2
which for a

1/2

R can be written as
r =

x2 + y 2 + z 2 2ax cos + a2 2ay sin

R2 2ax cos 2ay sin

R 1

ax cos ay sin

R2
R2
157

1/2

1/2

I dl
X
Figure 7: Current loop of radius a.
Hence
r 1 = R1 1 +

ax cos ay sin
+
R2
R2

where we have used the Taylor expansion of 1/(1 x) = 1 + x + . . ..

Thus
A =

ax cos ay sin
+
R2
R2
0
a sin d + a cos d .
i
j

I
40 c2 R

1+

Since
2
0

sin d =

2
0

cos d =

2
0

sin cos d = 0 ,

the formula for A simplies to

A =

I
40 c2 R3

2
0

a2 y sin2 + a2 x cos2 d ,
i
j
158

which can be written as

Ia2
A =
y
i
40 c2 R3

2
0

sin2 d + x
j

2
0

cos2 d .

Next, since
2
0

sin2 d =

2
0

cos2 d = ,

we obtain
A =

Ia2
y + x .
i
j
40 c2 R3

Using the relation

x y
j
i
x y z

i+ j+ k =

kR =k
R
R
R
R
R
we get

Ia2

kR =
k 2
40 c2 R2
40 c2
R
1
1
=

40 c2
R

A =

where we have used the result

1
R
= 2
R
R
If we consider a set of dipole moments and change, for a convenience, the

direction of R into R, we can write that the vector potential dA produced

by a set of dipole moments contained in a volume dV is
dA =

1
M
40 c2

1
dV
R

where M is the magnetic dipole moment per unit volume.

159

X
^
R
dV
Then, the vector potential produced by the whole dipole moments contained in the volume V is
1
40 c2

A =

1
dV
R

Using a vector identity

(A) =

A+

we have
M

1
1
=
M =
R
R

M
R

and then the vector potential is

A =

M
1
dV
R
40 c2

1
40 c2

In order to proceed further, we introduce a Theorem

M
dV =
R

M n

dS
R

This is an application of the more general theorem

F dV =
160

F n dS

M
dV
R

Proof:
Let C be a constant vector. Then
(F C) = (

F) C (

= C (

F)

C) F

We will prove the general theorem by using the divergence theorem

(F C) dV =

(F C) n dS

For an arbitrary constant vector C, and using (32), we get

C

F dV

=
=

V
V

F dV

(F C) dV =

F C n dS .

Hence
C

F dV =

C F n dS .

However
C F n =C F n

and then we obtain

C

F dV = C

F n dS

Since this is true for arbitrary C, we nally have

V

F dV =

F n dS

as required.
Thus
A =

1
40 c2

1
M
dV +
R
40 c2
161

M n

dS ,
R

(32)

or
A =

1
40 c2

JV
1
dV +
R
40 c2

JS
dS .
R

The eective (Ampere) currents associated with a dipole moment per unit
volume M are
(i) A current density JV = M throughout the volume.
(ii) A surface currents JS = M n.

M
B

Mxn

11.2

The Magnetic Intensity Vector H

When we were dealing with dielectric materials in the presence of electric

elds, it was convenient to introduce the displacement vector D in order to
eliminate the necessity of taking the electric dipole polarization P of the
material into account explicitly.
A similar procedure is used for the magnetic materials, and we will illustrate it here for both static and time-varying elds.
For a static current distribution, the Maxwells equation IV reduces to
B = 0 J .
162

In a medium where there are magnetic polarization currents as well as conduction currents we can write
J = Jc + Jm ,
J = Jc + M .
Hence
B = 0 ( Jc +

M) ,

which can be written as

( B 0 M ) = 0 Jc ,
or

B
M = Jc .
0

This shows that the vector B/0 M has as its source only the conduction
current Jc . Therefore, to eliminate the necessity of dealing directly with the
magnetization M , we can dene a new vector
H=

B
M ,
0

which is called the magnetic (eld) intensity vector.

Then the Maxwells equation IV for static elds in magnetic materials
takes the form
H = Jc .
In dealing with magnetic materials we often know Jc but not M (well not
directly anyway.) Think e.g. of an inductor lled with some magnetizable
material like iron. Then H becomes useful. It is a way of avoiding a detailed
calculation of the polarization currents. The magnetic intensity H is the
magnetic analogue of the dielectric displacement D in the electric case. We
may drop the subscript c and write
H =J ,
but we should remember that J is now not the total electric current density
everywhere.

163

11.3

Linear Isotropic Magnetic Materials

For most materials (excluding ferromagnetics) the magnetization M is proportional to the applied external eld. Hence, at any point the vectors B,
M , and H will be in the same direction and we may write
B = 0 r H = 0 (1 + m )H = H ,
where r is the relative permeability, m is the magnetic susceptibility, and
= 0 (1 + m ) is called the magnetic permeability.
Since
B = 0 (H + M ) ,
we have dened m such that
0 M = 0 m H =

m B
0 m B
=
.

1 + m

Hence, we get
H=

B
.
0 (1 + m )

Thus, if we know the material we use, we can nd H.

Example A solenoid lled with magnetizable material
In the expression B = 0 N I we should include the Ampere surface currents as well as the conduction currents in the wire. Since, B = 0 I , where
I is the total current per unit length
B = 0 (N I + M ) = 0 (N I + m H) = 0 (N I +
which can be written as
B(1

0 m
) = 0 N I ,

164

m B
),

or
B=

0 N I
.

1 0 m

Note that if m is positive then B is greater than it would have been in the
absence of the magnetizable material. Evidently in this case the macroscopic
Ampere current is in the same sense as the conduction current in the solenoid.
From the denitions
0 m
0 m
=1
1

0 (1 + m )
m
1
1
1
=
=
1 + m
1 + m
r
B = r 0 N I = N I = H .
The eect of lling the solenoid with a material of relative permeability r is
to multiply B by a factor r (assuming the current in the wire remains the
same).
Note: H is independent of the presence or absence of the magnetic material.
Consider a solenoid lled with a magnetic material. By the denition
H=

B
M .
0

When we remove the magnetic material, M = 0, and then

H=

B
0 N I
=
= NI .
0
0

With the material present

H=

B
N I
M =
m H = r N I m H .
0
0

Hence
H = (1 + m )N I m H
165

from which, we nd
H(1 + m ) = (1 + m )N I
and then
H = NI
as before.
This gives rise to the notion of H as an inducing eld and B as a resultant eld. This concept is much used in the study of magnetic properties of
materials.
The eect of magnetic materials on inductors
Because the relative permeability r multiplies the magnetic eld by r
for the same current, the self inductance of a solenoid is multiplied by r .

Let N is a number of turns/meter. Hence, N l is the number of turns,

and then the ux is
o = Lo I = nBo A = (N )(0 N I)A .
Thus
Lo = 0 N 2 A
with relative permeability r , so that B r Bo .
L = r 0 N 2 A = r L o .

166

11.4

The Magnetization of Iron

Iron is not a linear isotropic material. It is subject to saturation of the

internally produced magnetic elds because all the internal current loops are
lined up. Iron does not have a unique value of magnetic susceptibility because
of strong nonlinearities. In ferromagnetic materials the relation between B
and H is usually presented graphically in terms of hysteresis.

B
SATURATION

HYSTERESIS LOOP

SATURATION

Work done in magnetization

Think of the case of the solenoid of length and cross-section area A lled
with magnetic material. The applied voltage V is:
V = E =

d
d
= N BA ,
dt
dt

where N l is the number of turns.

The rate of doing work is:
P = V I = A NI

dB
dB
=V H
dt
dt

where V = Al is the volume of the material.

In the time dt that it takes to change B by dB, the work done is dW =
P dt = V H dB. Thus, the work done per unit volume is dW = H dB.
Hence, the work done per unit volume in one cycle of the hysteresis loop
of a non-linear material such as iron is given by the area of the loop.
167

H
IRON

FERRITE

It is sometimes useful to write

H dB = 0
= 0
=

H d(H + M )
H dH + 0

1
0 H 2 + 0
2

H dM

H dM ,

where the rst term on the rhs is the work to establish magnetic eld, and
the second terms is the work by the eld H to establish magnetization dM .

11.5

Time Dependent Magnetic Fields and Energy Loss

In a time-dependent magnetic eld the magnetization M may not stay in

phase with the driving eld H. This corresponds to internal friction and
heat dissipation.
Let
H = H0 cos(t) .
Then
M = M0 cos(t + ) ,
168

where M0 and represent the amplitude and phase of the magnetization

response to the magnetizing eld.
Thus
M = M0 cos cos(t) M0 sin sin(t) .
Now the work done in magnetization per cycle of the AC current producing
the magnetizing eld is
2/

W = 0

H dM = 0

H
t=0

dM
dt .
dt

Since
dM
= M0 cos sin(t) M0 sin cos(t) ,
dt
we have W = W1 + W2 . Consider the term W1 :
2/

W1 = 0 H0 M0 cos

cos(t) sin(t) dt .
t=0

This term is zero - it represents reversible energy conversion to and from H.

Consider now the term W2 :
2/

W2 = 0 H0 M0 sin

cos2 (t) dt .
0

This term represents work done against internal friction during magnetizing
and demagnetizing the material.
Since cos2 (t) dt is positive, sin must be negative so that work is done
on the material i.e. is negative.
The complex magnetic susceptibility etc
Let
H = H0 eit ,
169

which for a physical eld can be written as

H = H0 cos(t) = Re H0 eit .
Then
M = M0 ei(t) = M0 ei eit
or
M = (M0 cos iM0 sin )eit
=

M0
M0
cos i
sin H0 eit
H0
H0

This result can be written in terms of real and imaginary susceptibility

M = ( i )H0 eit = ( i )H .
Using this complex number notation, we nd
B = 0 (H + M ) = 0 H + 0 ( i )H
= [0 (1 + ) i0 ]H
and then
B = ( i )H = H ,
where = 0 (1 + ), = mo , and is the complex permeability.

11.6

The Ferromagnet

We have so far considered magnetic materials, diamagnetics and paramagnetics in which the magnetization is a function of the external eld, i.e.
M B. There is a class of materials, called ferromagnetics in which macroscopic magnetization exists even in the absence of the external eld.

170

Consider a homogeneous ferromagnetic material. In this case B inside the

material is due solely to the M
Js = M n

or in terms of the magnitudes

Js = M n .

LINES OF B

Js

M
LINES OF H

Using Amp`res law

e

B d = 0 I, we obtain

B = 0 M

i.e.

B = 0 M

Thus
H=

0 M
B
M =
M =0
0
0

in the region where M = 0, i.e. inside the magnet.

Outside the magnet
H=

B
B
0=
,
0
0

i.e. H is just a scaled replica of B.

171

Note:
B = 0 always and the lines of B form closed loops. Then
H =

B
M =
0

M .

Thus H has a source (eld lines start and stop) where M varies i.e. at the
ends of the magnet.
For a ferromagnet
H =

M = m ,

so we can think of m as a volume density of magnetic charge giving rise to

the H eld. It must be stressed that this equivalence is purely mathematical,
and does not prove a physical existence of magnetic charges.
Moreover, for the ferromagnet
H = Jc = 0 .
Note the similar mathematical properties of H here to those of the E eld
in electrostatics

E =
and
E =0 .
0
Historically, magnetostatics of permanently magnetized materials (ferromagnets) developed via use of the H eld and magnetic charges or poles.
One obtains a law analogous to Coulombs Law for the force between magnetic poles.
Exercise in class: Plane magnetized material
An innite plane surface divides the universe into a vacuum on one side
and a magnetic material on the other. Within the magnetic material there
exists a uniform magnetic moment per unit volume M which is parallel to
the surface.

172

(a) Show that while the direction of the magnetic induction vector B is
dierent on the two sides of the surface, its magnitude is given everywhere by:
B=

M
.
20 c2

(b) Find the magnitude and direction of the magnetic eld B everywhere
due to an innite plane parallel slab of material of thickness d which is permanently uniformly magnetized with dipole moment per unit volume M lying
parallel to the bounding surfaces.
(c) Find the magnetic intensity H everywhere.

11.7

Maxwells Equations in Dielectric and Magnetic

Materials

The Maxwell equation containing a current density term is

B = 0 J + 0 0

E
.
t

We think this is always true provided J is the total electric current density. Applying this in a region where there may be electric and magnetic
polarization eects we can write
J = Jc + JE + JM = J +

P
+
t

M .

Thus
B = 0 Jc + 0

P
+ 0
t

M + 0 0

E
,
t

M = Jc + (0 E + P ) ,
0
t
173

and nally
H =J +

D
,
t

where we must remember that J represents the conduction current only.

In summary: The Maxwells equations in materials are of the following

form
D = ,
B =
H =0 ,
H
E =
,
t
D
H = J +
,
t
but in general
H =

M .

These equations are supplemented by appropriate constitutive relations which

connect the electric eld E and the magnetic induction B with the displacement eld D and the magnetic eld H
D = E

and

B = H .

These relations carry information about the material.

Questions:
(1) Show that inside a ferromagnet H = 0.
(2) Derive the Maxwells equations for the EM elds in electric and magnetic materials.
174

12

Poyntings Theorem Revisited

We have seen in Section 6.1 how energy of the electromagnetic eld may
be transported through vacuum (empty space) by means of electromagnetic
waves. We have shown that the direction of propagation of energy is determined by the Poynting vector. In this lecture, we will reconsider the Poynting
theorem taking into account propagation of the electromagnetic eld in magnetizable materials.

12.1

Poynting Vector in Terms of E and H

In a vacuum
H=

B
B
=
= 0 c2 B .

Thus the Poynting vector is

N = 0 c2 E B = E H .
The cross product E H also turns out to be the correct expression for
the Poynting vector when magnetizable materials are involved and is the expression most commonly quoted for it.
Consider, as before in Section 6.1, a ow of the energy through a surface S.
Using the Gausss theorem and a vector identity
(A B) = B (

A) A (

B) ,

we obtain
s

(E H) dS =
H (

(E H) dV

E) dV

E(

Now substitute from the Maxwell equations

B,
t

H = J + D ,
t
E =

175

H) dV

and obtain
(E H) dS =

B
dV
t

D
dV
t

E J dV .

The lhs is the rate of ow of eld energy out of volume V .

1. First term on the rhs is the rate of work in establishing the magnetic
eld in V .
2. Second term is the rate of doing work in establishing the electric eld
in V .
3. Third term is the rate of doing work on the currents in V .
Note that the rate of doing work in establishing the magnetic eld is
H

B
dV .
t

Thus, to change the eld by an amount dB we have to do work per unit

volume
dWM = H dB
and then the total work is
WM =
=

H dB =
H

H d(0 H + 0 M )

d(0 H)
dt +
dt

0 H

dM
dt .
dt

The rst term on the rhs is work in energizing eld, and the second term is
work in aligning magnetic dipoles.
In a previous discussion we saw that H dM can involve loss of energy
from the eld to the material (c.f. magnetization of iron). A similar result
holds for the electric eld.

176

We go further with Poyntings Theorem

Put B = H, then we get
H

1
H 2
t 2

B
dV =
t

D
dV =
t

1 2
E
t 2

dV =

dV =

1 B2
dV
2
1 2
E dV .
2

Hence

(E H) dS =

1 2 1 B2
E +
2
2

dV

E J dV .

Thus, the rate of ow of eld energy out of volume V is equal to the rate of
changing energy of the EM eld plus the arte of doing work on the currents
in V .

12.2

Poynting Vector for Complex Sinusoidal Fields

It is well known that the electromagnetic eld (e.g. light) is a real physical
quantity (observable). However, in the electromagnetic theory it is advantageous to represent the real electromagnetic eld by complex sinusoidal quantities because of its mathematical simplicity. In addition, what we usually
measure is the average intensity of the eld, E E , which is a real quantity.
We usually write
E = E0 eit

and

H = H0 eit ,

where E0 and H0 are complex quantities including both amplitude and phase
information. We understand that the electric and magnetic elds are given
by the REAL PARTS of E and H.

177

The power of the complex exponential scheme lies in the fact that for operations such as summation, subtraction, integration etc., we take real parts
AFTER the operation. For example:
ReE1 + ReE2 = Re E1 + E2
Re

d E1
d E2
d
E1 + E2
+ Re
= Re
dt
dt
dt

Some care has to be taken in evaluating the Poynting vector.

The Poynting vector N = E H, but if we write complex exponential expressions for E H, we must remember that
N = ReEc ReHc = Re Ec Hc

where we use a subscript c to indicate that we are writing a complex exponential.

Proof:
We can write
Ec = E0 eit = Er + iEi (cos t + i sin t)
=

Er cos t Ei sin t + i Er sin t + Ei cos t .

Similarly
Hc = H0 eit = Hr + iHi (cos t + i sin t)
=

Hr cos t Hi sin t + i Hr sin t + Hi cos t ,

where Er , Ei , Hr and Hi are real vectors.

However
ReEc = Er cos t Ei sin t
ReHc = Hr cos t Hi sin t
178

Clearly, if we calculate Re(Ec Hc ) we get extra terms in addition to those

in the expression for ReEc ReHc , and then
ReEc ReHc = Re(Ec Hc ) ,
as required.
There is however a useful expression for the MEAN POYNTING VECTOR in terms of the complex exponential Ec and Hc
1
1

N = Re Ec Hc = Re Ec Hc
2
2

where bar over N means average over whole cycles of the sinusoidal eld.
Proof:
Since
N = ReEc ReHc
= Er cos t Ei sin t Hr cos t Hi sin t
= Er Hr cos2 t + Ei Hi sin2 t
(Er Hi + Ei Hr ) cos t sin t

and
1
T
T
0

T
0

cos2 t dt =

1
T

T
0

sin2 t dt =

cos t sin t dt = 0

we obtain
1

N = ( E r Hr + E i Hi ) .
2
On the other hand, take

Hc =

Ec =

Hr iHi eit ,
Er + iEi eit ,
179

1
2

and then

(Ec Hc ) = (Er Hr + Ei Hi ) + i(Ei Hr Er Hi ) .

Hence
1
1

Re Ec Hc = (Er Hr + Ei Hi ) , = N
2
2
as required.

In summary: The average Poynting vector N of complex exponential elds

satises the relation
1

N = ReEc ReHc = Re Ec Hc
2

180

13

Plane Wave Propagation in Dielectric and

Magnetic Media

Now, we shall examine in some detail how existing radiation is modied by

the material it passes through. We will nd that the conductivity is the most
signicant parameter.
We have seen that in a lossy dielectric the properties of the dielectric can
be described using a complex permittivity and similarly in a lossy magnetic
material use a complex permeability.
Thus, for a lossy material the Maxwell equations can be written as
E = / ,

B = 0 ,

H = 0) ,

B
H
=
,
t
t
E
H = J +
,
t
E =

where , are complex quantities that characterize the material, and J is

conduction current only, i.e. J = E.
Consider a plane wave propagating in the z direction
E = E0 ei(tkz) .
Then
H = E + iE = ( + i)E
Since for a plane wave propagating in the z direction the derivatives /x
and /y of E and H are zero, we have

i
j
k

0
0 z
Hx Hy Hz

= ( + i)E

181

From the LHS, we see that

(

H)z = 0 ( + i)Ez = 0 .

Hence Ez = 0 unless ( + i) = 0. Thus, E k.

Since Ez = 0, we have that
Also

E = 0.
E =

H
t

can be used to show that E H.

In addition, we can nd from the above equation that Hz = 0.
Since E = 0, the quantity occurs only in the equation for
and it is common to proceed as follows:

H,

E
= E + iE
t

H = i
+ E = i i
E
i

H = J +

Now
i

= i i = i +

which gives
H = i E .

Physically what has been done is to lump together the conduction current
and the lossy dielectric current term. To an external observer they are inseparable. Only using some theory of the internal structure of the dielectric,
they can be separated.
We can summarize our ndings of the propagation of the EM wave in a

182

conducting material that the electric and magnetic elds of the propagating
wave satisfy the following equations
E = 0

(33)

B = 0

(34)

E =

(35)

H
= iH
t
H = iE

(36)

where we have left the bar o .

We now proceed to solve Eqs. (33)(36). We will see that the solution leads
to a dispersive equation. We look for plane wave solutions, and will try to
nd how k behaves.
The procedure is as follows: Taking
(

of (35) and using (36), we obtain

E) = 2 E

E = 2 E

Hence

2E
= 2 E .
z 2

Since
2E
= k2E
z 2
we nally obtain a dispersion equation
k 2 = 2
This dispersion equation is not as simple as it looks. We cannot just say that
phase velocity is
vp =

=
k

183

as and are complex quantities and then k is complex.

What does an imaginary value of k mean?
We can write the complex number k as
k = i ,
where and are real.
Then
E = E0 ei(tkz) = E0 ei(tz+iz) = E0 ez ei(tz) .
Clearly from this, the phase velocity is
vp =

and = Im(k) is the attenuation coecient.

Physically the losses come from (1) Conduction currents, (2) Lossy dielectric,
and (3) Lossy magnetic material, which are lumped together in .
We now nd and .
We can write
1

k = () 2 =

i +

k =

+
+

1
2

Let
p =
q =

+
184

( i )

1
2

Then
1

k = (p iq) 2
k =

p +q

1
2

k = p2 + q 2

1
4

iarctan(q/p)

1
2

ei

1
where = 2 arctan q/p.
Hence

= Re(k) = p2 + q 2

1
4

= Im(k) = p2 + q 2

cos
1
4

sin

which in general is quite complicated.

13.1

Wave Refraction and Skin Eect

We will illustrate propagation of an EM wave in dielectrics and conducting

materials on two examples:
Example 1: The low loss dielectric and no magnetic eects
For dielectrics, is negligible and then we disregard magnetic properties.
+

= 0 =

1
0 c2

Hence

k =
i +
0 c

1
2

=
( )2 + +
0 c

185

1
2

1
2

ei

where
+

= arctan

Then

i/2
k =
e
c 0
k =

i
e
0

+/
2

Hence

=
c

+ /
cos
0
2

c 0

We can nd phase velocity and refractive index

0

=c
<c

n =
=
>1,
vp
0

vp =

for > 0 ,

which are not aected by the losses.

Thus, inside the dielectric, the EM wave will propagate with a phase velocity vp < c and will be refracted.
Similarly, we can nd
= Im(k) =

+ /

sin
0
2

Since for small , sin , we get

+ /
c 0
2
+ /

c 2 0

186

Thus, losses (absorption of the wave) are small.

We can summarize, that the theory predicts that the refractive index for
a lossless dielectric is given by
n=

nr .

Table below compares theoretical values of n with that obtained experimentally. An excellent agreement is observed, except for polar molecules (e.g.
water).

nr
Air
1.00029
Argon
1.00028
CO2 gas 1.00047
Benzene 1.49
Ethanol 5.3
NaCl
2.47
Water
9.0

n (experimental)
1.00029
1.00028
1.00045
1.48
1.36
1.54
1.33

Example 2: Good conductor

A good conductor, e.g. Cu, Ag.
(1) There are no dielectric and magnetic losses = = 0.
Consider the Maxwell equation IV:
H =J +

E
= E + iE .
t

(37)

Denition of a Good Conductor

Good conductor is when the conduction term on the rhs of the Maxwell
187

equation (37) dominates over the displacement current, i.e. when

For example, for copper at = 1 MHz = 2 106 rad

5.8 107

=
= 1.0 1012 .
0
2 106 8.85 1012
Consider general expression for k:
k =

i +
+

k = [ 0] i + 0

1
2
k = i

1
2

1
2

Now we might as well drop the dashes on , understanding that they are
real quantities, and obtain
1

2
k = i

k = 1 i

1
2

Remembering that for a good conductor

k =

1+

1
2

ei arctan
1, we get

i
e 2
k =

e 4
=

Since cos(/4) = sin(/4) = 1/ 2, we obtain

k =

1
1
i
2
2
188

1
2

1
2


(1 i)
2

= Re(k) =
= = Im(k) .
2
k =

Note a very heavy attenuation of this wave

E = E0 ez ei(tz)
Since
==

the distance for attenuation e fold (i.e. amplitude falls to a factor

1
original value in a distance ) is
z=

1
e

of its

i.e.
=

1
=

!!

where is called the skin depth in the conductor. It is the distance the wave
must propagate in order to decay by an amount e1 . This eect is sometimes
called skin eect as with an increasing the current ows in a narrower
and narrower layer, until in the limit of a true current exists only on
the surface of the conductor.

189

Questions:
(1) Derive the special form of Poyntings Theorem applicable in certain
material media
(E H) dS =

B
dV
t

D
dV
t

E J dV

Interpret the above equation in terms of energy storage and energy ow etc.
State qualitative meaning of each term in the equation.
(2) Prove that the useful formula for the mean Poynting vector for sinusoidal elds is
1

N = Re Ec Hc
2

where Ec = E0 exp(it) and Hc = H0 exp(it).

(3) Show that the amplitude of a plane wave propagating in a nonconducting material is damped with the rate which arises from the imaginary parts of the complex permittivity and permeability.
(4) Show that in a good conductor an EM wave propagates on the surface
of the conductor.

Weekend exercises
Find the skin depth for cooper at 60 MHz.

190

14
14.1

Transitions Across Boundaries for Electromagnetic Fields

Applications in dielectrics

Across boundaries between dierent material media there are sharp changes
in electrical properties , , . On a macroscopic scale the elds may have to
be regarded as varying discontinuously across such boundaries. The source of
such discontinuities will be the surface polarization charges and currents P n

and M n discussed previously.

We will use the Maxwells divergence equations I and II to investigate the transition of normal eld components, and the curl
equations III and IV to investigate the transition of tangential eld
components.

14.1.1

Normal Component of B

Consider the Maxwell equation II: B = 0.

Using the Gauss divergence theorem, the Maxwells equation can be written as

B n dS = 0 ,

where S is an arbitrary surface closing some area on the boundary plane.

^
n

(2)

^s
n

^
n

191

l
(1)

In order to evaluate the integral, we consider a thin cylinder of area S

and thickness including the boundary
(B2 n1 ) S + (B1 n2 ) S +

sides

B ns dS = 0 .

Since B is nite everywhere and we are interested in the transformation of

the eld at the boundary, 0, the integral
sides

B n dS 0 as 0 .

Hence
B2 n 1 + B1 n 2 = 0 .

However
n1 = 2 = n ,

and then
n ( B2 B1 ) = 0

or

B2 = B1 .

Thus, the normal component of B is continuous across a boundary

between two dierent dielectric materials.

M1
M2

^
n2

^
n1

n
M1 x ^1
n
M2 x ^2

Figure 8: Polarization currents at the boundary between two dierent materials.

We can understand this result intuitively by noting that the B elds of
polarization currents M n are parallel to the boundary and so do not aect

the normal component of B, see Figure 8.

192

14.1.2

Normal Component of H

Since B = H, we have
B1 = 1 H1 = 2 H2 = B2 .
Thus, when 1 = 2 , the normal component of H is not continuous across a
boundary. This result for the normal component of H is the consequence of
dierent magnetizations of the materials.
14.1.3

Normal Component of E

Since in nonconducting dielectrics D = 0, an identical argument to the

above will show that the normal component of D is continuous across
a boundary.
In the case of dielectrics we write D = E.
Then 1 E1 = 2 E2 can be a useful boundary condition.

P1

^
n2

Ep
=P1 n1 +P2 n2
^

P2
^
n1

The electric eld of dielectric surface charge P n is normal to the bound

ary. This causes the discontinuity in the E eld.
14.1.4

Tangential Component of E

We will use the Maxwells curl equations III and IV to investigate

the transition of tangential eld components.
193

E2

2
1
E1

For the tangential component of E, we will apply the Faraday induction

law (Maxwell equation) to a closed path such as in the Figure, in which the
sides perpendicular to the boundary are made innitely short compared to
the parallel sides L.

L
^
t

^o
n

(2)

a l

(1)
Consider the Maxwell equation III:
B
t
Integrating both sides of this equation over the surface a and applying the
Stokess theorem, we obtain
E =

E dl =

B
n0 da .

Hence

E 2 t2 + E 1 t1

L+

ends

E dl =

194

B
n0 da ,

(38)

where no is the unit vector normal to the surface a, and t1 and t2 are unit

vectors along the paths L on the side (1) and (2), respectively.
As dl 0 the rhs and the second term on lhs of Eq. (38) go to zero, since
E and B are nite everywhere. In this limit, the area enclosed by the path
approaches zero.

Since, t2 = t1 = t, we then have

E2 E1 t = 0 .

But E t is the component of E tangential to the surface. Since this is true

for any t, we obtain E1 = E2 , and we conclude that:

The tangential component of E is continuous across a boundary
between two dierent dielectric materials.
The continuity of a tangential component can be written in an equivalent
form as
n E2 E1 = 0 .

^
n

E
/2

^x E
n

Explanation
From the Figure above, we see that the cross product
n E = E sin(/2 ) = E cos ,

i.e. it is the tangential component of E.

195

14.1.5

Tangential Component of H

Consider the Maxwells equation IV:

H = Jc +

E
,
t

where Jc is the conduction current, i.e. not counting polarization currents.

Stokess theorem then gives
H dl =

Jc nda +

E
nda .

Both terms on the rhs go to zero as l 0 because Jc and E/t are nite.
Hence

H2 H1 t = 0 ,
or
n H2 H1 = 0 .

Thus, the tangential component of H is continuous across a boundary.

14.1.6

Tangential component of B

The tangential component of B is not in general continuous across a boundary because of the presence of the magnetic polarization surface currents
M n, which do not have a nite current density as they ow in an innitely

thin surface layer. Thus, if we examine the corresponding Maxwell equation

for B, the term in the integral involving J may stay nite as j 0.
Since
H = 0 c2 B M ,
we have
H2 H1 = 0 c 2 B 2 B 1 M 2 M 1
196


Take scalar product of both sides with t, and using the fact that

H2 H1 t = 0 ,
we obtain

B2 B1 t =

M2 M1 t
0 c2

If M1 = M2 the dierence between the two surface currents generates a

discontinuity in the B eld (see Figure below).

^
n2
M1

B1
B2
M1 x ^1
n

M2

M2x ^2
n
B2
B1

^
n1

Summary: Field components that are continuous across a boundary

1. The normal component of D.
2. The tangential component of E.
3. The normal component of B.
4. The tangential component of H.

197

Questions:
(1) Prove the following boundary conditions at a bounding surface between two dielectrics:
(a) The normal component of B is continuous across the boundary.
(b) The tangential component of E is continuous across the boundary.
(c) The tangential component of H is continuous across the boundary.

198

15

Reection and Transmission of Waves

Across a Boundary

At a boundary between two dierent media, the general boundary conditions cannot be satised by a transmitted wave alone. There has to be a
reected wave also. What happens is that the incident radiation is absorbed
by charges in the boundary and reradiated in all directions. The waves interfere destructively except in two directions along those of the reected and
transmitted waves. The directions, amplitudes and phases of the reected
and transmitted waves can be derived from the general boundary conditions
already obtained. We consider linear isotropic materials characterized by
electric and magnetic constants , , and .

15.1

Representation of Plane Waves in Dierent Directions

PHASE FRONT

^
np

^p r
n

Let np is the unit normal to the phasefront (wavefront) of a propagating

wave, and r is a position vector that is independent of np . The vector r

determines position of a wave-front as seen by an observer O.

199

Suppose that the wave propagates in the z direction. Then

E = E0 exp[i (t kz)] .
Since
z = nr ,

we obtain
E = E0 exp[i (t n r k)] .

Thus, the observer can distinguish dierent waves at the point A by dierent ns.

15.1.1

Representation of B in terms of E

For the eld of a plane wave, we have

B=

k
np E ,

(39)

where np is the unit vector in the direction of propagation of the eld.

Recall that the relation (39) arises from the Maxwell equation III:
E =

B
.
t

To show this, we expand both sides of this equation, and obtain

i
j
0 0
Ex Ey

B
Bx By
=
i
j
,
t
t
t

where we have used the fact that the wave propagates in the z direction.
Comparing the coecients standing at the same unit vectors, we nd that
the x component gives

Bx
Ey
=
z
t
200

from which, we obtain

ikEy = iBx

or

Ey
=
.
Bx
k

The y component gives:

Ex
By
=
z
t
from which, we obtain
ikEx = iBy

or

Ex
= .
By
k

Thus
2
2
E = Ex + E y

1
2

2
2
Bx + B y
k

1
2

B.
k

Since E, B, np , are mutually orthogonal, E B gives the direction of np

(Poynting result) of the propagation direction. Then np E gives the direc

tion of B. Hence
H=

B
k
=
np E .

(40)

In the next few sections, we will use the continuity conditions for E and H.
With the relation (40), we will be able to limit the analysis to the electric
eld alone, as knowing properties of E, we can nd from Eq. (40) properties
of H.

15.2

Directions of Reected and Transmitted Waves

Here, we will show that some familiar elementary results on reection and
transmission can be derived from the Maxwell equations.
The most useful results in this connection concern the continuity of tangential components of the E and H elds. However, most of the results can
be derived from the fact that the tangential component of E is continuous
across a boundary.
201

^
n
^
ni

i r
r

^
nr

^
t n t

Consider properties of an EM wave at a boundary between two materials.

Let the origin of position coordinates r be located on the boundary S.
The electric eld for incident, reected and transmitted waves are
Ei = E0 exp [i (t ni r k1 )] ,

Er = E1 exp [i (t nr r k1 )] ,

Et = E2 exp [i (t nt r k2 )] .

These equations determine the three electric elds relative to the direction
of observation r.
Why do we need the presence of the reected wave in the propagation of
an incident wave through the boundary?
The answer to this question is provided by the requirement that the tangential components of E and H must be continuous through the boundary.
Suppose that E is polarized in the plane of incidence. Then
Ei cos i Er cos r = Et cos t ,

(41)

Hi + H r = H t .

(42)

and

Since
H=

k
E,

202

the relation (42) takes the form

Ei + E r =

k2
Et ,
k1

(43)

as for a dielectric 1 = 2 = 0 .
Now we see that without Er , Eq. (41) gives
Ei =

cos t
Et ,
cos i

while Eq. (43) gives

Ei =

k2
Et .
k1

Thus, without Er we would get two dierent values for Ei or Et , which we

cannot accept as both continuity conditions Eqs. (41) and (42) must be satised at the same moment. Hence, we conclude that the continuity conditions
for E and H will be satised only if Er = 0.
What are the relative directions of ni , nr and nt ?

From the boundary condition for the tangential component of E:

n Ei + Er = n Et

i.e.
n E0 exp [i (t ni r k1 )] + E1 exp [i (t nr r k1 )]

= n E2 exp [i (t nt r k2 )]

This relation must hold over the whole surface S for all r (subject to n r = 0).

Thus, the exponential phase factors must all be the same. Otherwise if it
was true for one r it would not be true for other r s, but we have a freedom
of choosing r. Hence
k1 ni r = k 1 nr r = k 2 nt r ,

203

from which we have

ni r = n r r ,

and
ni r =

k2
nt r .

k1

These relations will help us to prove that

Incident, reected and transmitted rays are coplanar
i.e. ni , nr , nt are coplanar.

To show this, we use the relation
r = ( r)
n
n
Proof:
Since
n ( r) = ( r) ( n)r

n
n
n
n
and n r = 0 as the vector r lies on the plane S, we obtain

n ( r) = r

n
as required.
Hence
ni [ ( r)] = nr [ ( r)] .
n
n
n
n
Interchange () and () products
( i n) ( r) = ( r n) ( r) .
n

n
n

n
This must be true for all r in plane S. Thus:
ni n = n r n

204

^
ni

^
n
^
nr
^ xn
ni r ^

This implies that nr is in the plane of incidence, i.e. the plane containing n

and ni .

Similarly
k1 ni r = k 2 nt r

implies that
k1 ni n ( r) = k2 nt n ( r) .

n

n
Interchange () and () products, we nd that
k1 ( i n) ( r) = k2 ( t n) ( r)
n

n
n

n
for all r in S. Thus
k1 ni n = k 2 nt n ,

which implies that ni , n, and nt are coplanar.

The coplanar property of the waves is observed in any experiment. This

is an another example of a remarkable triumph of the Maxwells electromagnetic theory.

205

15.3

Snells Law of Refraction and Angle of Reection

Since
ni n = n r n

we have that
sin i = sin r
Thus
i = r .
The angle of incidence equals the angle of reection, another familiar law of
elementary optics.
Moreover
k1 ni n = k 2 nt n

from which we have

k1 sin i = k2 sin t
and then
k2
sin i
=
= n12 .
sin t
k1
This is the well-known law of refraction in optics, called the Snells law.
In the case where k1 and k2 are purely real (e.g. in dielectrics), the
refractive index has a simple physical interpretation
n12 =

k2
/k1
v1
=
,=
k1
/k2
v2

i.e. the refractive index in equal to the ratio of phase velocities.

206

Questions:
(1) Prove that in the reection and refraction at a bounding surface, the
direction of incident, reected and refracted waves are coplanar.
(2) Using the results of Question 1, derive the familiar laws of elementary
optics:
(a) Angle of reection equals to the angle of incidence.
(b) Snells law of refraction.
(3) Show, using the continuity conditions for E and H that both reection
and refraction takes place in the incidence of light on a boundary between
two dielectrics.

207

16

Fresnels Equations

The boundary condition on tangential E does not give sucient information

to calculate Er and Et in terms of Ei . For a given Ei there are still two
unknowns in the equation for continuity of tangential E viz Er and Et . We
need a second relation between Ei , Er and Et . This can be obtained from
the continuity of the magnetic eld.
We know that B tangential is not continuous across a boundary because
of the presence of M n surface currents in magnetized materials. To allow

for such possibilities we can use the more general condition that tangential
component of H is continuous, i.e.

n Hi + Hr = n Ht .

Since
H=

B
k
=
np E ,

where np is a unit ray vector in the direction of propagation, we have an

equation
k1
k1
n i Ei +

n r Er

1
1
k2
=n

n t Et ,

(44)

which together with

n Ei + Er = n Et ,

(45)

contain sucient information to determine Er and Et in terms of Ei .

The solution of these two simultaneous vector equations is greatly facilitated by considering separately, electric eld components parallel and normal
to the plane of incidence.
Equations (44) and (45) also provide a simple explanation of why we need
reected eld at the boundary to obtain the correct results for the eld amplitudes.

208

16.1

Ei normal to plane of incidence

In this case, the incident electric eld Ei is purely tangential to the boundary.
Since the materials are isotropic, the induced elds Er and Et will also be
tangential to the the boundary.

^
n

^
ni
E
Thus the condition

n Ei + Er = n Et

gives
E0 + E1 = E2 .
Note then that
n E0 = n E1 = n E2 = 0 .

The other equation for E0 , E1 and E2 comes from the H condition through
the relation
k1
k1
n i Ei +

n r Er

1
1
k2
n t Et

=n

and since the phase factors in E exp i(t n r k) are the same, we obtain

k1
n n i E0 + n n r E1

1
k2
=
n n t E2 .

2
209

Using the relation A (B C) = (A C)B (A B)C and the fact that

n E0 = n E1 = n E2 = 0, we obtain

k1
k2
n n i E0 + n n r E1 =


n n t E2

1
2
However
n ni = cos( i ) = cos i

n nr = cos r = cos i

n nt = cos( t ) = cos t

and then, we obtain
k1
k2
E0 cos i E1 cos i =
E2 cos t .
1
2
Since Es are all in the same direction, we might as well drop the vector
signs. Thus
E0 + E 1 = E 2

(46)

E0 cos i E1 cos i =

k 2 1
E2 cos t .
k 1 2

Eliminating E1 using (46), E1 = E2 E0 , we get

E0 cos i (E2 E0 ) cos i =

2E0 cos i = cos i +

k 2 1
E2 cos t
k 1 2

k 2 1
cos t E2
k 1 2

2k1 2 E0 cos i = (k1 2 cos i + k2 1 cos t ) E2

and then
E2 =

2k1 2 cos i
E0
k1 2 cos i + k2 1 cos t
210

(47)

Since
k1 sin i = k2 sin t
we can eliminate t
k2 cos t =

2
2
k2 k1 sin2 i

and obtain
E2 =

2k1 2 cos i
2
2
k1 2 cos i + 1 k2 k1 sin2 i

E0 .

(48)

Similarly, eliminating E2 = E0 + E1 from (47) above, we obtain

E1 =

2
2
k1 2 cos i 1 k2 k1 sin2 i
2
2
k1 2 cos i + 1 k2 k1 sin2 i

E0 .

(49)

Equations (48) and (49) are called Fresnel equations for the electric eld
amplitudes.
The corresponding H elds are not parallel to each other, but their relative magnitudes can be deducted from equations of the form
H=

16.2

k
nE

i.e.

H=

kE
.

Ei in the plane of incidence

In this case Hi is tangential to the boundary plane and then Hr and Ht are
tangential too. Thus
n Hi + Hr = n Ht

becomes
H0 + H1 = H2 .
The continuity of tangential E is given by
n E0 + E1 = n E2 .

211

It is simpler to work in terms of H in this case.

We express E in terms of H
E=

nB =

nH .

k
k

Thus
1
1
n i H0 + n r H1

k1
k1
2
=n

n t H2 .

k2

By a procedure similar to the case of Ei normal to plane of incidence, we

obtain
H1 =
H2 =

2
2
2
k2 2 cos i 1 k1 k2 k1 sin2 i
2
2
2
k2 1 cos i + 2 k1 k2 k1 sin2 i
2
2k2 2 cos i
2
2
2
k2 1 cos i + 2 k1 k2 k1 sin2 i

H0 ,

(50)

H0 ,

(51)

Equations (50) and (51) are called Fresnel equations for the magnetic
eld amplitudes.
This time the Es are not parallel, but their relative amplitudes can be
deducted from E = (/k)H.

16.3

Fresnel Equations for dielectric media

In a dielectric: conductivities 1 = 2 = 0, 1 = 2 = 0 , k = 2/ =real, and

vp =
Since k 1/ 1/vp

=
.
k
0

, the Snells law (k1 sin i = k2 sin t ) becomes

1 sin i = 2 sin t
212

and then
sin i
=
sin t

2
v1
=
= n12
1
v2

Consider two examples:

(1) E normal to the plane of incidence. In general
2k1 2 cos i
E2
=
E0
k1 2 cos i + k2 1 cos t
Since
1 = 2 = 0

and

k2
sin i
=
,
k1
sin t

we obtain
2 cos i
E2
2 cos i
=
=
k2
sin i
E0
cos i + k1 cos t
cos i + sin t cos t
2 cos i sin t
=
.
cos i sin t + sin i cos t
Hence
E2
2 cos i sin t
=
.
E0
sin (t + i )
Similarly
E1
sin(t i )
=
.
E0
sin (t + i )
(2) For E in the plane of incidence
E2
2 cos i cos t
=
,
E0
sin (t + i ) cos (i t )
tan(i t )
E1
=
.
E0
tan (t + i )
213

17

Applications of the Boundary Conditions

and the Fresnel Equations

Let us now examine some of the consequences of the Snells law. There
are two cases possible: n2 > n1 and n1 > n2 . In the rst case, an optical
wave travels from an optically rarer to optically denser medium. In the
second case, we have the inverse situation. We will consider these two cases
separately for dielectrics and conductors.

17.1

Applications in dielectrics

17.1.1

Polarization by reection

E
i

1
t

For the case of E in the plane of incidence the ratio of reected to incident
amplitude is
E1
tan(i t )
=
.
E0
tan(i + t )
If i + t = then tan(i + t ) = and consequently E1 = 0.
2
At the same time E1 will not be zero for the electric eld component
normal to the plane of incidence.
214

Thus if Ei has arbitrary polarization then Er will be plane polarized with Er

normal to the plane of incidence.
If i + t = then t = i . Thus
2
2
sin i
= n21 =
sin t

sin i
2
=
= tan i .
1
sin( i )
2

Hence, the angle of incidence for total linear polarization of the reected
wave is
2
.
i = arctan
1
It is known in the literature as the Brewsters angle.
An alternative proof that if E is in the plane of incidence, then E1 = 0
(no reected eld polarized in the plane of incidence).
We can use the continuity conditions for the tangential components of E
and H, from which we have
E0 cos E1 cos = E2 cos t ,
H0 + H 1 = H 2 .

and

(52)
(53)

Since
H=
and t =

, we get

k
E,

E0 E1 = E2 tan ,
E0 + E1 = n12 E2 ,

(54)

where n12 = k2 /k1 .

However, from the Snells law we have that
sin
sin
sin
=
= tan = n12 .
=

sin t
cos
sin

Thus, Eqs. (54) will be satised simultaneously only if E1 = 0, i.e. there is

no reected eld in the plane of incidence.

215

17.1.2

Total internal reection

Consider the Snells law

sin t =

sin i
,
n21

where n21 =

2
.
1

In the case n21 < 1 (2 < 1 ), going from an optically more dense to optically
less dense medium e.g. from water in to air, real angles t are obtained only
for sin i n21 .

z cost
x sin t

X
t

z
x

(x, z)

Z
For greater i , sin t > 1, and then the angle of refraction t becomes
imaginary. In this case, there is no real refracted wave, only a reected wave.
The Fresnel equations are still capable of giving a formal solution in this case
cos t =
= i

1 sin t =

sin2 i
1 2
n21

sin2 i
1 = i ,
n2
21

where is real positive. Put, e.g.

sin t =

sin i
=.
n21
216

Now consider the propagation of the transmitted wave in the less optically
dense medium, for which
n
Et = E2 ei(t t rk) ,

with
nt r = z cos t + x sin t = iz + x ,

and then
Et = E2 ei(t+i z x) = E2 e z ei(t x) .
Here = k and = k, i.e. there is attenuation in the z direction but no
phase propagation. Phase propagation occurs in the x direction along the
boundary.
This illustrates a general method of applying the Fresnel equations. For
only a limited range of circumstances will all the angles i , r , t be real. We
can however always apply a generalized Snells Law k2 sin t = k1 sin i to
nd sin t and cos t and proceed as above.
Note that the planes of constant phase are normal to the boundary (i.e. they
have their normals tangential to the boundary). The phase of the transmitted wave below the boundary must match the incident wave above. The
wavelength in the second medium is
=

2
2
=
=

2
2 sin i
0 n21

0 n21
,
sin i

where 0 is the wavelength of freely propagating waves in this medium.

The planes of constant amplitude in the transmitted medium are parallel to
the boundary.

17.2

Transmission and Reection at a Conducting

Surface

Propagation of EM elds in conductors (metals) is more complicated phenomenon than in dielectrics. Consider a propagation of an EM wave in a
217

=0

i
k1

k2

=0

medium composed of a dielectric and a conductor, and assume that the incident wave originates in the dielectric.
From previous work, we know that in the dielectric
2
k 1 = 1 1 2

and in the conductor

2
k2 = 2 2 2 i2 = 2 2 2 1

i
2

Moreover, from the Snells law

sin t =

k1
sin i .
k2

Assume a good conductor. Since k2 as 2

1, we see that sin t 0
independent of i . Thus, the direction of the transmitted wave is normal to
the surface independent of the angle of incidence i .
It follows that in the conductor the eld vectors E and H lie tangential to the
boundary and so the normal components of these vectors on the conductor
side of the boundary are zero.

It follows that:
Since the normal component of B (or H) is continuous across the
boundary, the normal component of H or B is zero on the dielectric
218

side also.
Thus the normal component of B of the reected wave must be equal
and opposite to that of the incident wave.
In the conductor
E
H

= Z2 =

E
H

0
k2

= Z1 .

This happens because

2
k2
i

=K 1
2
k1

as

1.

This means that the electric eld in the conductor (which is tangential
to the surface) 0.
Since E is continuous across the boundary E is zero also in the dielectric at the boundary.
In summary, we have two useful special boundary conditions at the
surface between a dielectric and a perfect conductor:
1. The tangential component of E = 0.
2. The normal component of B or H = 0.
17.2.1

Field vectors at normal incidence

We now consider the special case of normal incidence at a boundary, i.e.

when the wave propagation vector coincides with the normal to the boundary.
Figure 9 shows how the eld vectors must look in the plane of the surface
between dielectric and conductor.
Since for a good conductor

the eld components

Et = E i E r
219

0,

Hi

Hr
Er

Ei
Ht
Et

Figure 9: The eld vectors in the plane of the surface between dielectric and
conductor.
Ht = H i + H r

2Hi ,

In this case, the power reection coecient

p =

E2
E r Hr
= r
E i Hi
Ei2

1 as

One could think that under the normal incidence, there is only the incident
and transmitted wave with no reected wave. Here, we prove the necessity
of assuming the existence of the reected wave.
From the continuity of the tangential components, we have
Ei E r = E t ,
Hi + H r = H t .
However
H=

E=
0

0
0
E=n
E.
0
0

Thus, Eq. (56) can be written as

n1 E i + n 1 E r = n 2 E t ,
220

(55)
(56)

and then we get two equations for the amplitudes of the electric eld
Ei E r = E t ,
n2
Et .
Ei + E r =
n1
If Er is missing, we could not simultaneously satisfy both equations. Thus,
there always is a reected wave in the normal incidence.

17.3

Wave Propagation in a Conducting Rectangular

Pipe (The Rectangular Waveguide)

In this lecture, we discuss the propagation of bounded EM waves by considering the propagation of radiation through a waveguide where the radiation
is fully conned in the transverse plane. We will consider the case where the
bounding walls are planar and cross section is rectangular.

Y
y=b

x=a

Z
From the nature of the problem we see that we are likely to get standing
waves in the x direction due to reections between the walls x = 0 and x = a,
standing waves in the y direction due to reections between the walls y = 0
and y = b and waves propagating in the z direction.
We solve Maxwells equations subject to the good conductor boundary
conditions being satised at x = 0, x = a, y = 0 and y = b. We write the z
dependence of any eld component in the form
ez

Thus

,
z

221

where describes the propagation conditions, e.g. purely imaginary describes a wave propagating without loss.
We describe the electromagnetic eld by the vector pair E, H, and we use
Maxwells equations in the form

=0
E =

B = 0 or
H =0
B
H
E =
=
t
t
E
E
= E +
H = J +
t
t
Certain characteristic modes of propagation are found. In general a
characteristic mode is one which propagates with constant polarization.
In a rectangular waveguide, one may have TE (transverse electric) waves,
or TM (transverse magnetic) wave, but not TEM wave. If both E and B
elds are transverse, the wave would be going straight down the guide. However, such a wave would not satisfy various boundary conditions.

17.3.1

Transverse Electric (TE) Modes

It is possible to propagate a wave with the electric eld polarized in the xy

plane by lifting the restriction that both the electric and magnetic elds are
transverse.
Look for solutions to Maxwells equations with Ez = 0 that satisfy the
good conductor boundary conditions.
For the TE modes we use Maxwells equations to derive a wave
equation for Hz , the longitudinal component of H. This equation
is used to satisfy the boundary conditions.
Write out Maxwells equations in Cartesian coordinates.
From

H
E + t = 0

Ex Ey

k
+ iH = 0
0
222

x component
y component

Ey + iHx = 0
Ex + iHy = 0
Ey Ex

+ iHz = 0
x
y

z component
From

(57)
(58)
(59)

H =0
Hx Hy Hz
+
+
=0
x
y
z

Hence:

From

Hx Hy
+
Hz = 0
x
y
H ( E + E ) = 0
t

Hx Hy

k
( + j)E = 0
Hz

Hz
+ Hy ( + i)Ex = 0
y
Hz
( + i)Ey = 0
x
Hy Hx

=0
(Ez = 0)
x
y

Hx

From

(60)

(61)
(62)
(63)

E = / = 0

And using the TE condition Ez = 0

Ex Ey
+
=0
x
y

(Ez = 0)

(64)

From Eqs. (57) and (58)

Ex
Ey
i
=
=
=Z
Hy
Hx

223

(65)

where Z is called the wave impedance.

In Eq. (59), we can then express Ex and Ey in terms of Hy and Hx and show
that it becomes identical to Eq. (60).
Use Eq. (65) in Eqs. (61) and (62)
In Eq. (61) substitute for Ex :
Hz
i
+ Hy ( + i)
Hy = 0
y

Hence
Hy =
=

1
Hz
i
( + i) y
2

Hz

i( + i) y

(66)

In Eq. (62) substitute for Ey and similarly get

Hx =

Hz
i( + i) x

(67)

Using Eqs. (65), (66), and (67), we nd that Eqs. (63) and (64) are automatically satised.
Putting Eqs. (66) and (67) into Eq. (60), we obtain

2 Hz
2 i( + i) x2

2 Hz

Hz = 0
2 i( + i) y 2

Dividing by and multiplying by k 2 = 2 i( + i), we obtain

2 Hz 2 Hz
+
+ k 2 Hz = 0
x2
y 2

(68)

If we can solve Eq. (68) for Hz then we can solve Eqs. (66) and (67) for Hx
and Hy . Then we can solve Eq. (65) for Ex and Ey , and then we will know
all the eld components.

224

17.3.2

Boundary Conditions

The boundary conditions at the surface of a good conductor are

1. H normal to boundary (in xy plane)= 0.
2. E tangential to boundary (in xy plane)= 0.

Hy
Y

Ex
Ey
Hx

y=b

x=a

Z
Thus, we must have at the boundaries x = 0 and x = a, Hx = 0 and
Ey = 0.
Looking at Eq. (62), we see that this means that Hz /x must equal zero at
x = 0 and x = a.
Similarly, at the boundaries y = 0 and y = b we must have Hy = 0 and
Ex = 0.
Looking at Eq. (61), we see that this means that Hz /y must equal zero at
y = 0 and y = b.
The solution of Eq. (68), which satises these boundary conditions is of
the form
Hz = H0 cos(kx x) cos(ky y) eitz
with
kx a = m

and
225

ky b = n

mx
a

Hz = H0 cos

ny
b

cos

eitz .

Here m, n = 0, 1, 2 . . .. An m, n combination represents a possible TE mode

of propagation. The modes are designated in the form TEmn .
Then Eq. (68) gives

m
a

n
b

+ k 2 Hz = 0

For a non-trivial solution, Hz = 0, and then

k2 =

m
a

2 =

m
a

n
b

n
b

= 2 i( + i)
2

+ i( + i)

In general is complex. We can write = + i.

Then we can write
mx
ny
Hz = H0 cos
cos
ez ei(tz)
a
b
Then = 2/g , where g is the wavelength in the waveguide at frequency .
The phase velocity in the waveguide is vp = /.
The parameter is the attenuation coecient describing losses in the
waveguide. Energy loss may be due to:
Ohmic resistivity ( of the medium nite).
Dielectric losses (described by having an imaginary component)
Magnetic losses (described by having an imaginary component)
17.3.3

TE Modes in a Lossless Waveguide

Assume a lossless propagation for which we have = 0 and and both

purely real. In this case
2 = 2 +

m
a
226

n
b

or introducing the phase velocity, we obtain

2 =

2
m
+
2
v0
a

n
b

where v0 is the phase velocity of propagation of waves in an innite

(unbounded) medium of the type lling the waveguide.
2f
2

=
=
vm
f 0
0
where 0 is the innite medium wavelength at that frequency.
2 =

2
0

m
a

n
b

The parameter 2 can be negative or positive, and the nature of the

propagation changes according as:
1. Consider 2 negative. Then = ikg is purely imaginary and we have a
propagating wave
mx
ny
Hz = H0 cos
cos
ei(tkg z) ,
a
b
where kg is the guide propagation constant.
The wave then propagates with a guide wavelength g given by
2
= kg =
g

2
0

m
a

n
b

There is a maximum 0 (=c say) such that kg is real

m
1
=
2
c
2a

n
2b

Thus, there is a minimum frequency fmn such that the T Emn mode will
propagate down the waveguide
fmn = vm

m
2a

n
2b

This can be derived from the cut-o condition fmn c = vm .

Thus, the waveguide acts as a high-pass lter for any m, n mode.

227

2. Consider 2 positive. Then = say is purely real and

Hz = H0 cos

mx
a

cos

ny
b

ez eit .

There is no phase propagation but amplitude attenuation. Note there

is no energy loss mechanism available. This is an evanescent mode at
frequencies less that fmn analogous to the case of total internal reection.
Field components in the TE modes.
Since
Hz cos

mx
a

cos

ny
b

eitz .

From Eq. (66)

Hy

Hz
y

cos

mx
a

sin

ny
b

eitz .

Hz
x

sin

mx
a

cos

ny
b

eitz .

From Eq. (67)

Hx
And from Eq. (65)
Ex Hy cos

mx
a

sin

ny
b

eitz

Ey Hx sin

mx
a

cos

ny
b

eitz

Allowed values of m and n for TE modes

228

m = n = 0 is not allowed since Ex , Ey , Hx , Hy all contain sine terms so

all elds vanish for m = n = 0. All other T Emn modes are allowed.
m = n = 0 mode is never possible in a transmission line consisting of
a single closed conductor like a rectangular waveguide. It is possible in 2conductor lines e.g. the coaxial line or the twin wire transmission line.
TM (transverse magnetic) modes
Put Hz = 0 and go through the whole procedure again. Ez = 0 now.
Equations analogous to Eqs. (66), (67), and (68) appear for components of
the E vector this time. Consequently the previous discussion about modes
and their cut-o frequencies for TE modes is also true for TM modes. The
only dierence is that more modes are not allowed.
TM modes with either m = 0 OR n = 0 are NOT ALLOWED.
Special properties of the T E10 mode
If the transverse dimensions of the rectangular waveguide are dierent
(a = b) there is a nite range of frequencies over which the TE10 mode is the
only allowed mode. This means that a waveguide can be designed to allow
propagation in one mode only. We have
fmn = vm

f10 =

m
2a

n
2b

vm
vm
< f01 =
2a
2b

where we adopt the convention that a > b. In the frequency range f10 f01
the TE10 mode is the only mode allowed.

229

E in XY plane

H in XZ plane

g
2

g
2

17.3.4

Phase and Group Velocities for TE and TM Modes

The phase velocity

vp = f g =

=
kg

2
0

m
a

n
b

Thus
v0

vp =
1

m0
2a

n0
2b

where v0 = /k0 is the phase velocity in the unbounded medium and k0 =

2/0 .
We see that the phase velocity of the wave inside the waveguide is greater
than in an unbounded medium, and so may be greater than the speed of
light in vacuum.
Alternately we could write
m
a

n
b

=
230

2
c

2
= kc

where c is the innite medium wavelength at the cut-o frequency for

the m, n mode.
vp =

2
2
k0 kc

kc
k0

k0 1

v0

vp =

fc 2
f

v0
1

provided f > fc i.e. in the pass-band for that mode.

As f fc , the phase velocity vp .
If a vacuum or air lls the waveguide then v0 = c and vp > c.
The group velocity
Since vp > c typically we see that vp is not the velocity of propagation
of energy or information down the waveguide. The waveguide is a dispersive
d
medium. The group velocity vg = dkg is then dierent from the phase velocity vp = /kg .
The group velocity is the velocity of propagation of some modulation of
the wave that carries information. A single frequency harmonic wave carries
no information. It is just there. A nite bandwidth is required to carry in
formation. Consider the sum of two frequencies and + d
cos(t kz) + cos[( + d)t (k + dk)z]
d
dk
2k + dk
2 + d
= 2 cos
t z cos
t
z
2
2
2
2
dk
d
t z cos(t kz) ,
= 2 cos
2
2
where we have assumed that d
2 and dk
2k.
The velocity of the amplitude modulation is
vg =

d
2
dk
2

231

d
dk

In the rectangular waveguide

kg =

2
=
g

2
2 c

2
2
k0 kc =

vm

where k0 = /vm and kc = c /vm .

Remember vm is the innite medium phase velocity and c is the cut-o
(angular) frequency of the m, n mode. Then dierentiating
1
dkg

1 1 2
2
=
( c ) 2 2 =
2
d
vm 2
vm 2 c

Finally
vg =

2
vm 2 c
d
=
dkg

= vm

c
1

fc
1
f

= vm

< vm .

Thus
vg vp = v m

fc
1
f

vm
1

fc
f

2
= vm

In a vacuum-lled waveguide vg vp = c2 . Thus, relativity is still all right.

Questions:
(1) Show that under the Brewsters angle of incidence there is no reected
electric eld in the plane of the incidence.
(2) Show that in a vacuum-lled rectangular waveguide vp > c
and vg vp = c2 .
232

18

18.1

Relativistic Transformation of the Electromagnetic Field

The Principle of Relativity

1. The laws of physics are the same in all inertial reference frames.
2. The speed of light in vacuum is independent of the uniform motion of
the observer or source.
The constancy of the velocity of light, independent of the motion of the
source, gives rise to the relations between space and time coordinates in
dierent inertial reference frames known as Lorentz transformations.
Consider a stationary reference frame S and a inertial frame S moving
with a velocity u parallel to the x axis.
The time and space coordinates in S are related to those in S by the
Lorentz transformations
x
y
z
ct
where = (1 2 )

1/2

=
=
=
=

(x ct) ,
y,
z,
(ct x) ,

is the Lorentz factor, and = u/c.

The above transformation corresponds to a situation of u parallel to the x

axis. If the axis in S and S remain parallel, but the velocity u of the frame S
is in an arbitrary direction, the generalization of the above transformations is
(r )
ct
2

= r + ( 1)

ct

= ct r

where = u/c

233

Proof:
Decompose the vector r into two components:
parallel and normal to
r = r + r
Then, using the one dimensional Lorentz transformations, we have
r

= r ct

r = r

We can write the parallel and normal components as

r

(r )
2
= rr

r
Hence
r

= r + r
= r ct + r r
= r + ( 1)

(r )
ct
2

Similarly
ct = (ct x) = ct r
However
r

= r =

(r )
=r
2

which gives
ct

= ct r

as required.

234

The inverse Lorentz transformation is

r = r + ( 1)

(r )
+ ct
2

ct = ct + r
The principle of relativity indicates that the Maxwell equations and the
continuity equation should be invariant under the Lorentz transformation.
In the frame S:
D =
B = 0

B
t
D
H = J +
t

J =
t
E =

In the frame S :
D
B

=
= 0

= J +

B
t
D
t

where the prime variables are functions of the transformed variables

(t and r ).
First, we will illustrate the transformation of an arbitrary scalar function
and the divergence F .
235

Consider the transformation of a time derivative /(ct)

ct
x
y

=
+
+
(ct)
(ct ) (ct) x (ct) y (ct)
ct
r
z
=
+
+
z (ct)
(ct ) (ct)
(ct)
However
ct
= ,
(ct)

r
=
(ct)

which gives

(ct)
(ct )

Consider now the divergence

F =

Fx Fy Fz
+
+
x
y
z

Since
Fx ct
Fx x
Fx
=
+
x
(ct ) x
x x
Fx y
Fx z
+
+
y x
z x
and
ct
x
x
x
y
x

= x
= 1 + ( 1)
=

2
x
2

z
=0
x

we obtain
Fx
Fx
Fx
= x
+ x
x
(ct )
x
236

where
2
x = 1 + ( 1)x / 2

Similarly, for Fy and Fz , and nally we get

F = (

F )

F
(ct )

where is a 3 3 diagonal matrix

1 + ( 1) x
2

0
0

0
1 + (
0

0
2
1) y
2

0
2
1 + ( 1) z
2

Using the above transformations, we can derive transformations for the

current density J and the charge density .
In order to do it, we consider the continuity equation, that can be written as
J =

c
(ct)

Hence
(

J)

( J)

J
=

(ct )
(ct )
(c) =

c J
(ct )

Since

(c) =

(c)

we obtain
J c =

c J
(ct )

237

Thus, the continuity equation will be invariant under the Lorentz transformation if
c

= c J

= J c .

(69)

In order to understand the physical meaning of these equations, consider the

following example.
Example
Assume that in the S frame there is a stationary volume charge of density
= 0. Since is stationary, there are no currents in the S frame (J = 0).
What are the charge and current densities as seen in the S frame?
In the S frame
J =0,

=0

According to Eq. (69), in the S frame

J = c ,

Thus, there is a current in the S frame. As seen from S a given part of the
charge is length contracted in the direction of motion so the charge density
is correspondingly increased by the factor > 1. The length contracted
charge density appears from S to move in the opposite direction. We can
understand this result: The stationary charge in the S frame moves with
velocity u in the S frame.
Less obvious and more interesting is the following situation.
In the S frame
J=0,

=0.

Then, someone will see charge density = 0 in the S frame.

This is a pure relativistic eect.

238

18.2

Transformation of Electric and Magnetic Field

Components

To nd the transformation rules for electric and magnetic eld components

we will use the transformations of the time and space derivatives derived
above.
Consider two of the Maxwell equations that in the S frame are
D =
H = J +

D
t

These equations should be equivalent of two equations

D
H

=
= J +

D
t

in the S frame.
Using the transformations of the time and space derivatives, we have

H
(ct )

cD = J

(ct )

cD = c
(ct )

Substituting the transformations of J and , we nd that the D and H

vectors transform as
cD
H

= 1 cD + H

= cD + 1 H

239

where 1 is the inverse of the matrix

1
1 + ( 1) x
2

0
0

1+

1
(

0
2
1) y
2

From the Maxwell equations

0
2
1
1 + ( 1) z
2

B = 0

B
t
we nd that the E and B vectors transform as
E =

E
cB

18.3

= 1 E + cB

= E + 1 cB

Transformation Rules in Terms of Parallel and

Normal Components

Suppose that the frame S is moving with speed u in the direction parallel to the z axis. In this case, x = y = 0, z = = 0, and then the
transformations take the form

cD = cDx + cDy + cDz k + k H

i
j
H

E
cB

= ck D + Hx + Hy + Hz k
i
j

= Ex + Ey + Ez k + c k B
i
j

= k E + cBx + cBy + cBz k

i
j

It is useful to rephrase the transformation rules in terms of components

parallel and normal to u. The parallel components are the z components and
the normal components lie in the xy plane. For example

E = E + E = Ex + Ey + Ez k
i
j
and the same for D, H and B.

240

18.3.1

Rules for Parallel Components

E = Ez k = Ez k = E
and similarly
B =B ,

18.3.2

H =H ,

D =D

Rules for Normal Components

cD = cDx + cDy
i
j
= cDx + cDy + Hx Hy
i
j
j
i
+ (cDy + Hx )
= (cDx Hy ) i
j
H = Hx + Hy
i
j
= (Hx + cDy ) + (Hy cDx )
i
j
E = Ex + Ey
i
j
= (Ex cBy ) + (Ey + cBx )
i
j
cB = cBx + cBy
i
j
= (cBx + Ey ) + (cBy Ex )
i
j
In general
c D = c D + H
H = H c D
E = E + c B
cB = cB E
241

Example 1 - purely electric eld in S

Suppose that in S, E = 0 but B = 0.
Then from the transformation rules, in S :
E =E
B =0

E = E
B =

u E
c2

Thus
B = B =

uE
u E
=
2
c
c2

since u E = 0.
Thus what appears to be purely an electric eld to one observer is seen
as both an electric and a magnetic eld to a second observer moving with
respect to the rst.
Example 2 - purely magnetic eld in S
Now suppose that in S, E = 0 while B = 0.
Then using the transformation rules, in S :
B =B
E =0

B = B
E = u B

Thus
E = E = u B = u B
We see that what appears to be a purely magnetic eld for one observer
will appear to be both an electric and a magnetic eld to a relatively moving
242

observer.
This result could be used to calculate the emf in an electric dynamo from
the point of view of an observer watching the conductor move in a magnetic
eld or from the point of view of an observer moving with the conductor.

18.4

Transformation of the Components of a Plane EM

Wave

Example 1
Suppose that a plane wave propagates in vacuum along the z axis. Then
E = i(tkz) = 0
iEe
iE
i(tkz) = 0
B = jBe
jB
Hence from the transformation rules, in S moving in the same direction:
E
cB

= E0 cB0 = (E0 uB0 )

i
i
i

= cB0 E0 = (cB0 E0 )
j
j
j

Since in vacuum
cB0 = E0
we obtain
E

= 1
=

u
c

E0 =
i

u
c

cu
E0 i
c+u

and
cB

= 1
=

u
c

cB0
j

cu
cB0
j
c+u
243

u
c

E0
i

Thus, the ratio |E |/|B | = |E0 |/|B0 | is constant and independent of u.

Example 2
Suppose that the frame S moves in the direction of the electric eld, i.e.
u = u
i.

X
X

In this case
E

= E0 + cB0 k
i
u
= + k E0
i
c

and
cB = cB0
j
Thus, the magnetic eld remains unchanged but the electric eld turns towards the direction of propagation of the wave.

18.5

Doppler Eect

Consider a plane wave propagating in vacuum

E(t) = E0 ei(tkr)
244

In the moving frame S this wave will have a dierent frequency and
the wave vector k , but the phase of the wave will remain unchanged as in
invariant under the transformation
= t k r = t k r
Using the Lorentz transformations
r = r + ct

ct = ct + r
Hence
t k r = t k r

=
ct + r k r k ct
c

= t t k u +
k r

c
= ku t k r

Z
k
u

Y
X
Thus

= ku

= k

245

u
c2


If the wave propagates in vacuum, k = k, and then
c
= 1

u
cos
c

where is the angle between the direction of propagation of the wave and u.
For = 0
=

18.6

cu
c+u

Transformation of Energy of a Plane EM Wave

Consider an EM wave propagating in the k direction and an observer moving

in the z direction, as shown in the Figure 18.6.

Z
k
u

X
Let E = E Then, in the frame S :
i.
E

= E + c B
= E + cos cB
246


Since B = B + B k (the wave propagates in the plane yz), and cB = E,
j
we obtain
= E cos cB
i
= (1 cos ) E

Consider now energy of an electric eld of a plane EM wave conned in

a volume V . In the S frame
1
2
We = 0 E0 V
4
In the S frame
1
We = 0 (E0 )2 V
4
However
V = xyz
Assume that S frame moves in the direction of the z axis, and the wave
propagates in the direction k u = u cos . Then
V = xyz
where
z =

z
1

u
c

cos

Hence
We =

1
u
2
0 E0 2 1 cos
4
c

= We 1

u
cos
c

V
1

u
c

cos

It is interesting to compare the transformation of energy with the transformation of frequency

We = W e 1

= 1
247

u
cos
c

u
cos
c

We see that the energy and frequency transform in the similar way, indicating
that We . This proportionality was predicted in quantum physics as We =
h and forms backgrounds of the so called quantum electrodynamics.

Questions:
(1) Find the condition under which the continuity equation for and J
is invariant under the Lorentz transformation.

248

Appendix A
PHYS3050 Facts and Formulae
Gauss Divergence Theorem :
S

Stokes s Theorem :

F n dS =

Fd =

Numerical values in SI units:

0 = 8.85 1012
0 = 4 107
c = 3 108 [ms1 ]
For the electron: e = 1.6 1019 [C],
The Lorentz force law :

F n dS

m = 9.11 1031 [kg]

F = q(E + v B)

Coulomb s Law :

Biot Savart Law :

Gauss Law :
Amp`re s Circuital Law :
e

F =

1 q1 q2
r

40 r 2

0 I d r

2
4 r
Q
E n dS =

0
dB =

B d = 0 I

Maxwells Equations in vacuum:

E =
E =

B =0

B
t

B = 0 J +
249

F dV

1 E
c2 t

Maxwells Equations in material bodies:

D =
E =

H =0

B
t

H =J +

D
t

Poynting vector:
N = 0 c2 E B
Poyntings Theorem:
0 c2 (E B) n dS =

E J dV

1
1 B2
0 E 2 +
2
2 0

dV

Or in polarizable materials where it is convenient to dene D = 0 E + P and

B
H = 0 M :

E H n dS =

B
dV
t

D
dV
t

E J dV

A theorem on the calculation of the mean Poynting vector from complex

elds:
1

N = Re Ec Hc
2
The rate of doing work in magnetization
dB
dW
=H
dt
dt
and B = 0 (H + M ).
An arbitrary vector F can be written as:

F =

1
4

1
F
dV +
r
4

= Fl + Ft

250

F
dV
r

Fields and potentials:

A
B=
t
Dierential equation for the vector potential:
E =

1 2A
= 0 J +
c2 t2

A+

1
c2 t

In the Lorentz gauge:

1
c2 t
the dierential equations for the electromagnetic potentials are:
A=

1 2
=
2 t2
c
0

1 2A
= 0 J
c2 t2
and these have solutions of the form:
2

1
40

1
40 c2
The eld of a Hertzian dipole:
A=

Er =
E =

dV
r
J
dV
r

I0 cos 2
2
+ 2 ei(tkr)
40 c
ikr 3 r

1
I0 sin
1
ik i(tkr)
+ 2+
e
3
40 c
ikr
r
r

B =

1
I0 sin ik
+ 2 ei(tkr)
2
40 c
r
r

251

The mean energy ux from the Hertzian dipole:

I2

N= 0
80 c

sin2
r2

A series solution in 2 dimensions to Laplaces equation in Cartesian coordinates:

[Ak sin(x) + Bk cos(x)][Ck sinh(z) + Dk cosh(z)]

(x, z) =
k

A series solution in 3 dimensions to Laplaces equation in spherical polar

coordinates:
[alm cos(m) + blm sin(m)] P m (cos )

C1 r + C2 r ( +1)

(r, , ) =

Useful properties of trigonometrical functions:

sin( ) = sin cos sin cos
cos( ) = cos cos sin sin
1
(1 cos 2)
sin2 =
2
1
cos2 =
(1 + cos 2)
2

0
2
0
2
0
2
0

sin3 d =

4
3

sin(m) sin(n) d =

0 for m = n
for m = n

cos(m) cos(n) d =

0 for m = n
for m = n

sin(m) cos(n) d = 0 for all m and n

252

Properties of Legendry polynomials:

1
n
Plm (cos ) Pk (cos ) d(cos ) = 0

unless

m = n and l = k

1
1

[Plm (cos )]2 d(cos ) =

1

0
P1 = cos

P0 = 1

1
P2 =

2 (l + m)!
2l + 1 (l m)!
1
0
P2 = (3 cos(2) + 1)
4

1
P1 = sin

3
3
2
sin(2)
P2 = (1 cos(2))
2
2
Pl (1) = 1
for all l

A theorem on the electrostatic potential due to a distribution of electric

dipoles of moment per unit volume P :
=

1
40

P n

1
dS +
r
40

P
dV
r

A theorem on the vector potential due to a distribution of magnetic

dipoles of moment per unit volume M :
A=

1
40 c2

1
M
dV +
r
40 c2

M n

dS
r

A dispersion equation:
k=

i +
253

1
2

The skin depth in a good conductor:

2

General boundary conditions:

The normal component of B is continuous across a boundary.
The normal component of D is continuous across a boundary.
The tangential component of E is continuous across a boundary.
The tangential component of H is continuous across a boundary.
Special boundary conditions at the surface between a dielectric
and a perfect conductor:
The tangential component of E = 0.
The normal component of B or H = 0.
The characteristic impedance of free space:
Z0 =

E
=
H

0
0

377

120

ohms

The Fresnel equations:

Case 1: E normal to the plane of incidence.
Reection:
E1 =

2
2
k1 2 cos i 1 k2 k1 sin2 i
2
2
k1 2 cos i + 1 k2 k1 sin2 i

E0

Transmission:
E2 =

2k1 2 cos i
2
2
k1 2 cos i + 1 k2 k1 sin2 i

254

E0

Case 2: E in the plane of incidence.

Reection :
2
2
2
1 k2 cos i 2 k1 k2 k1 sin2 i
H1 =
2
2
2
1 k2 cos i + 2 k1 k2 k1 sin2 i
Transmission:

H2 =

2
21 k2 cos i
2
2
2
1 k2 cos i + 2 k1 k2 k1 sin2 i

In dielectric media the Fresnel equations become:

E normal to the plane of incidence:
sin(i t )
E1
=
E0
sin(i + t )
2 cos i sin t
E2
=
E0
sin(i + t )
E in the plane of incidence:
E1
tan(i t )
=
E0
tan(i + t )
E2
2 cos i cos t
=
E0
sin(i + t ) cos(i t )
Rectangular waveguides:

If the elds vary as ez :

For T E modes the longitudinal component of H satises:
2 Hz 2 Hz
+
+ k 2 Hz = 0
x2
y 2
where
k 2 = 2 i( + i)

Then satisfying the boundary conditions (assuming the walls are perfect conductors) requires:
2 =

m
a

n
b
255

+ i( + i)

For the lossless waveguide:

Cut-o frequency for the mn mode:
m
2a

fmn = vm

n
2b

Phase velocity:
vm

vp =

fc 2
f

1
Group velocity:

fc
1
f

vg = v m

VECTOR FORMULAS
( + ) =
(A + B) =

A+

(A + B) =
A+ B
() = +
(A) = A

(A B) = B (
(A) =

(A B)

( A)

( A)

A) A (

A+

= A BB
2
=

= 0
= 0
=
( A)

B)

A + (B

)A ( A

A (B C) = B (C A) = C (A B)

A (B C) = B(A C) C(A B)
256

)B

FORMS OF VECTOR OPERATIONS IN CYLINDRICAL

COORDINATES

A =

1 (A ) 1 A Az
+
+


z

A =

+
+z

A
1 Az

A Az
+
z

+z

(A ) A

1 2 2
1

+ 2 2 + 2

VECTOR AND DIFFERENTIAL OPERATIONS IN

SPHERICAL COORDINATES

A = (Ax sin cos + Ay sin sin + Az cos ) r

+ (Ax cos cos + Ay cos sin Az sin )

+ (Ax sin + Ay cos )

= Ar r + A + A

257

A =

A =
+

1
1

r+

r
r
r sin
1 (r 2 Ar )
1 (sin A )
1 A
+
+
2
r
r
r sin

r sin
1 Ar (rA )
r

(sin A ) A

r sin

r sin
r

(rA ) Ar

r
r

1
r2
2 r
r
r

sin
2 sin
r

258

1
2
r 2 sin2 2