EE141 Text

Kurt Edmund Oughstun
Electric & Magnetic Field Theory

August 25, 2015
Springer
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 The Electric Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1 Coulombs Law and the Electric Field Intensity . . . . . . . . . . . . 7
2.1.1 Charge Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Gauss Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 The Electrostatic Scalar Potential and Work . . . . . . . . . . . . . . . 12
2.3.1 Poissons Equation for the Scalar Potential . . . . . . . . . . 14
2.4 The Concept of an Ideal Conductor . . . . . . . . . . . . . . . . . . . . . . . 18
2.5 The Electric Dipole, Quadrupole, and Multipoles . . . . . . . . . . . 21
2.5.1 The Static Electric Dipole . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5.2 The Linear Electric Quadrupole . . . . . . . . . . . . . . . . . . . . 24
2.5.3 Static Electric Multipoles . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 The Electrostatic Field Produced by an Arbitrary Static
Charge Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 The Concept of a Perfect Dielectric . . . . . . . . . . . . . . . . . . . . . . . 29
2.8 Electrostatic Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.8.1 Capacitance and Energy of Multi-Conductor Systems . 37
2.8.2 Thomsons Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.9 Forces on a Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3 Electric Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.1 Conservation of Charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Electric Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.1 Resistance and Conductance . . . . . . . . . . . . . . . . . . . . . . . 56
3.3 Joules Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
v
vi Contents
4 The Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.1 The Biot and Savart Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Amperes Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3 The Differential Relations of Magnetostatics . . . . . . . . . . . . . . . 62
4.4 The Vector Potential for the Magnetostatic Field . . . . . . . . . . . 63
4.5 Magnetic Field of a Spatially Localized Current Distribution . 64
4.5.1 A Scalar Potential for the Magnetic Field in
Source-Free Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.6 The Concept of a Perfect Magnetic Material . . . . . . . . . . . . . . . 68
4.7 Faradays Law of Electromagnetic Induction . . . . . . . . . . . . . . . 72
4.7.1 Self-Inductance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.7.2 Mutual Inductance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.8 Magnetic Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5 Transmission Line Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

5.1 Distributed Resistance and Internal Inductance . . . . . . . . . . . . 81
5.2 Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.2.1 Transmission Line Equations . . . . . . . . . . . . . . . . . . . . . . . 89
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
A The Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

A.1 The One-Dimensional Dirac Delta Function . . . . . . . . . . . . . . . . 97
A.2 The Dirac Delta Function in Higher Dimensions . . . . . . . . . . . . 104
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
B Helmholtz Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
C Greens Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
D Boundary Value Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

D.1 Boundary Value Problems in Rectangular Coordinates . . . . . . 118
D.2 Boundary Value Problems in Two-Dimensional Angular
Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
D.3 Boundary Value Problems in Cylindrical Coordinates:
Cylinder Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
D.4 Boundary Value Problems in Spherical Coordinates:
Legendre Polynomials and Spherical Harmonics . . . . . . . . . . . . 126
D.4.1 Legendres Equation and the Legendre Polynomials . . . 127
D.4.2 Spherical Coordinate Boundary Value Problems with
Azimuthal Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Contents vii
D.4.3 Associated Legendre Functions and the Spherical

Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
D.4.4 Boundary Value Problems in Spherical Coordinates . . . 133
D.4.5 The Addition Theorem for the Spherical Harmonics . . 134
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Contents 1
Chapter 1
Introduction
Classical electromagnetic field theory has its origins in the historically sepa-
rate theories of electric and magnetic fields. In order to maintain close connec-
tion with experimental observation, the fundamental field equations derived
from Coulombs law, the Biot-Savart law, and Amperes law [1] are typi-
cally expressed in terms of an appropriate spatial integration over regions
containing electric charge, both static and moving with a uniform velocity
(steady electric current), as well as from Faradays law [2] for accelerating
charge (time-varying electric current). The divergence and Stokes theorems
from vector analysis are then used to express these relations in differential
form, which is more amenable to mathematical analysis. With the inclusion
of a displacement current in Amperes law in order to satisfy conservation
of charge, Maxwell [3, 4] united electric and magnetic field theory, forming
electromagnetic field theory.
Maxwells equations are described by the set of four vector differential
relations
D(r, t) = (r, t), (1.1)

B(r, t) = 0, (1.2)
E(r, t) = B(r, t)/t, (1.3)
H(r, t) = J(r, t) + D(r, t)/t, (1.4)
in MKSA units. The first two equations are mathematical statements of

Gauss law, the third equation expresses Faradays law, and the fourth equa-
tion expresses Amperes law. These four first-order differential relations are
completed by the appropriate constitutive relations (or material equations)
relating the induced electric displacement field vector D(r, t), the magnetic
intensity field vector H(r, t), and the conduction current density Jc (r, t),
referred to as the induction field, to both the electric field intensity vector
E(r, t) and the magnetic induction field vector B(r, t), these latter two field
vectors being referred to as the primitive fields.
3
4 1 Introduction
The divergence of Amperes law (1.4) followed by substitution from Gauss

law (1.1) yields the equation of continuity
J(r, t) + (r, t)/t, (1.5)
which is the differential expression of the conservation of charge. This re-

sult depends upon the inclusion of the displacement current D(r, t)/t in
Amperes law, as was originally done by Maxwell [3, 4]. Finally, Maxwells
equation (1.1)(1.4) are connected to physical measurement through the
Lorentz force relation [5]
F(r, t) = (r, t)E(r, t) + J(r, t) B(r, t), (1.6)
from which it is seen that an electric field accelerates charge whereas a mag-
netic field changes its direction.
In free space (vacuum), the appropriate constitutive relations are simply
D(r, t) = 0 E(r, t) and H(r, t) = 1
0 B(r, t) where 0 = 8.854 10
12
F/m is
7
the dielectric permittivity and 0 = 4 10 W/Am the magnetic perme-
ability of free space. In source-free regions of space, the two curl relations of
Maxwells equations then become
E(r, t) = 0 H(r, t)/t, (1.7)

H(r, t) = 0 E(r, t)/t, (1.8)
with E(r, t) = H(r, t) = 0. The curl of the first equation (1.7) with
substitution from the second equation (1.8) and use of the vector identity
F = ( F) 2 F then yields the vector wave equation for the
electric field
1 2 E(r, t)
2 E(r, t) 2 = 0, (1.9)
c t2
and similarly, the curl of (1.8) with substitution from (1.7) yields the vector
wave equation for the magnetic field
1 2 H(r, t)
2 H(r, t) = 0, (1.10)
c2 t2

where c = 1/ 0 0 is the speed of light in vacuum. Taken together, these
results unified electric and magnetic fields into a single electromagnetic field
and established the fact that light is an electromagnetic wave.
The electromagnetic field equations uncouple into separate, independent
equations for the electric and magnetic field vectors only when the field is
static (i.e., when the time derivative of the field quantities identically van-
ish). In that idealized case, the electrostatic field is described by the pair of
relations D(r) = (r), E(r) = 0 and the steady-state magnetic field is
described by the separate pair of relations B(r) = 0 and H(r) = J(r).
References 5
As stated by Einstein regarding Maxwells theory of electromagnetism in

his 1936 article on Physics and Reality [6],
the electric field theory of Faraday and Maxwell represents probably the most
profound transformation which has been experienced by the foundations of
physics since Newtons time. . . The existence of the field manifests itself, indeed,
only when electrically charged bodies are introduced into it. The differential
equations of Maxwell connect the spacial and temporal differential coefficients
of the electric and magnetic fields. The electric masses are nothing more than
places of non-disappearing divergency of the electric field. Light waves appear
as undulatory electromagnetic field processes in space.
In regard to the contribution by Lorentz to this theory, Einsten [6] went on

to state that his theory was built on the following fundamental hypothesis:
Everywhere (including the interior of ponderable bodies) the seat of the field
is the empty space. The participation of matter in electromagnetic phenomena
has its origin only in the fact that the elementary particles of matter carry
unalterable electric charges, and, on this account are subject on the one hand to
the actions of ponderomotive forces and on the other hand possess the property
of generating a field. The elementary particles obey Newtons law of motion for
the material point.
The word ponderomotive is an adjective describing the tendency to produce

movement of a ponderable body. By a ponderable body is meant one having
nonzero mass. Finally, as a critique of the resultant Maxwell-Lorentz theory,
Einstein [6] pointed out that
This is the basis on which H. A. Lorentz obtained his synthesis of Newtons me-
chanics and Maxwells field theory. The weakness of this theory lies in the fact
that it tried to determine the phenomena by a combination of partial differen-
tial equations (Maxwells field equations for empty space) and total differential
equations (equations of motion of points), which procedure was obviously un-
natural. The unsatisfactory part of the theory showed up externally by the
necessity of assuming finite dimensions for the particles in order to prevent the
electromagnetic field existing at their surfaces from becoming infinitely great.
The theory failed moreover to give any explanation concerning the tremendous
forces which hold the electric charges on the individual particles. H. A. Lorentz
accepted these weaknesses of his theory, which were well known to him, in order
to explain the phenomena correctly at least as regards their general lines.
Einstein then concluded that [6]

What appears certain to me. . . is that, in the foundations of any consistent field
theory, there shall not be, in addition to the concept of field, any concept con-
cerning particles. The whole theory must be based solely on partial differential
equations and their singularity-free solutions.
References
1. A. M. Ampere, Memoir on the mutual action of two electric currents, Annales

de Chimie et Physique, vol. 15, pp. 5976, 1820.
6 1 Introduction
2. M. Faraday, Experimental Researches in Electricity. London: Bernard Quaritch,

1855.
3. J. C. Maxwell, A dynamical theory of the electromagnetic field, Phil. Trans.
Roy. Soc. (London), vol. 155, pp. 450521, 1865.
4. J. C. Maxwell, A Treatise on Electricity and Magnetism. Oxford: Oxford Univer-
sity Press, 1873.
5. H. A. Lorentz, Uber die Beziehungzwischen der Fortpflanzungsgeschwindigkeit
des Lichtes der Korperdichte, Ann. Phys., vol. 9, pp. 641665, 1880.
6. A. Einstein, Out of My Later Years. New York: Philosophical Library, 1950. pp.
7677.
Chapter 2
The Electric Field
The beginning is the most important part of the work. Plato
The electric field produced by a specified static distribution of charges is

developed here based solely upon Coulombs law. An historical development
of this theory is given by Elliott [1].
2.1 Coulombs Law and the Electric Field Intensity
The mathematical description of the electric field begins with Coulombs law1
for the force exerted by one point charge on another This so-called electric
charge is an inherent, inseparable physical property of certain fundamental
(elementary) particles (e.g., electrons, protons, and positrons) that are the
closest physical approximation to an ideal point charge. Let q1 be a stationary
point charge with position vector r1 relative to a fixed origin O, and let q2
be a separate, distinct point charge with position vector r2 6= r1 relative to
the same origin O. Coulombs law then states that the force F21 exerted by
q2 on q1 is given by
q1 q2
F21 = K 2 r21 (2.1)
r
where r |r1 r2 | is the distance between the two point charges and where
r21 (r1 r2 )/|r1 r2 | = (r1 r2 )/r is the unit vector directed from q2
to q1 . The force is repulsive if q1 and q2 are of the same sign and attractive
if they are of the opposite sign, accounted for by the direction of the unit
vector r21 and the sign of the product q1 q2 in Eq. (2.1). Reciprocity requires
that an equal but oppositely directed force F12 is exerted by q1 on q2 , so
1
Charles Augustin de Coulomb (17361806) demonstrated the inverse square law
of electric force in 1785 using a torsion balance. His results were preceded by the
experimental observations of Benjamin Franklin in 1755, Joseph Priestley in 1767,
John Robison in 1769, and Henry Cavendish in 1773. In an experiment performed at
the Worcester Polytechnic Institute in 1936, Plimpton and Lawton showed that this
dependency deviated from the inverse square law by less than 2 parts in 1 billion.
7
8 2 The Electric Field
that F12 = F21 , a direct consequence of Eq. (2.1) since r12 = r21 . In
the rationalized MKSA (meter, kilogram, second, ampere) system, the unit
of force is the newton (N ), the unit of charge is the coulomb (C), and the
constant appearing in Coulombs law is given by
1
K= 8.988 109 N m2 /C 2 .
40
Here 0 8.85421012F/m is the permittivity of free space, where the Farad

(F C/V ) is the unit of capacitance and the Volt (V J/C) is a measure
of work per unit charge. Coulombs law in MKSA units then becomes
1 q1 q2
F21 = r21 (N ) (2.2)
40 r2
As written, Coulombs law directly applies to any pair of point charges that
are situated in vacuum and are stationary with respect to each other. It
also applies in material media if F21 is taken as the direct microscopic force
between the two charges q1 and q2 , irrespective of the other forces arising
from all of the other charges in the surrounding material.
The Coulombic force satisfies the principle of superposition. The electro-
static force exerted on a stationary point charge q1 at r1 by a system of
stationary point charges qk at rk , k 6= 1, is consequently given by the vector
sum or linear superposition of all the Coulombic forces exerted on q1 as
X q1 X qk
F(r1 ) = Fk1 = 2 rk1 (2.3)
40 rk1
k6=1 k6=1
with rk1 r1 rk , where rk1 r1k /r1k and rk1 = |r1k |.

The electric field intensity E(r) = E(x, y, z) at a point r = 1x x+ 1y y + 1z z
in space is defined as the limiting force per unit charge exerted on a test charge
q at that point as the magnitude of the test charge goes to zero, viz.
F(r)
E(r) lim (V /m) (2.4)
q0 q
The limit q 0 is introduced in order that the test charge does not influence
the charge sources producing the electric field. The electric field is thus de-
fined in such a way that it is independent of the presence of the test charge.
Notice that this abstraction to a field concept (introduced by Michael Fara-
day) eliminates the mechanist requirement of action-at-a-distance that is
embodied in Coulombs law.
From Eqs. (2.2) and (2.4), the electric field intensity at a fixed point r due
to a single point charge qk situated at rk is
1 qk
E(r) = R, (2.5)
40 R2
2.1 Coulombs Law and the Electric Field Intensity 9
where R = rrk denotes the vector from the source point Pk at rk to the field
point P at r with magnitude R = |R|, and where R R/R is the unit vector
along that direction. As a consequence of the principle of superposition, the
electric field intensity at a fixed point r due to a system of stationary, discrete
point charges qj located at the points rj , j = 1, 2, . . . , n, is then given by the
vector sum
n
1 X qj
E(r) = Rj (2.6)
40 j=1 Rj2
where Rj = r rj denotes the vector from the source point Pj at rj to the

field point P at r with magnitude Rj and direction Rj Rj /Rj .
2.1.1 Charge Density
The microscopic charge density (r) is a scalar field whose value at any point
r in space is defined by the limiting ratio
q
(r) lim (C/m3 ) (2.7)
V 0 V
where q is the net charge in the volume element V . From a microscopic per-
spective, the charge density (r) is zero everywhere except in those regions
occupied by fundamental charged particles. From a macroscopic perspective,
the abrupt spatial variations in the microscopic charge density (r), which
are on the scale of interparticle distances, are removed through an appro-
priate spatial averaging procedure over spatial regions that are small on a
macroscopic scale but whose linear dimensions are large in comparison with
the particle spacing; a detailed description of this spatial-averaging procedure
is presented in 4.1. The result is the macroscopic charge density
(r) = hh(r)ii (2.8)
The electric field that is determined from such a macroscopic charge density
is correspondingly a spatially-averaged field and, as such, is just what would
be obtained through an appropriate laboratory measurement. Notice that
the microscopic charge density can be obtained from the macroscopic density
through the use of the Dirac delta function (see Appendix A).
With the introduction of the charge density, the vector summation ap-
pearing in Eq. (2.6) may be replaced (in the appropriate limit) by a volume
integration over the entire region of space containing the source charge dis-
tribution. Because q(r) = (r)V is the elemental charge contained in the
volume element V at the point r, Eq. (2.6) becomes

1 X (r ) 1 (r )
ZZZ
E(r) = lim 2
RV = Rd3 r , (2.9)
V 0 40 R 40 R2
where R = r r is the vector directed from the source point r to the

field point r with magnitude R and direction R, and where d3 r = dx dy dz
denotes the volume element at the source point r = (x , y , z ). Notice that
the integral appearing in Eq. (2.9) is convergent for r = r.
2.2 Gauss Law
Consider a point charge q at a fixed point in space together with a simply-

connected closed surface S. Let r denote the variable distance from q to a
Fig. 2.1 A point charge

q inside a closed surface q
S producing an electric v
d n
field E(P ) at a point P P
on S a distance r away.
da
The differential element
E(P)
of surface area da at P
has unit outward normal S
vector n at an angle to
E, subtending the solid
angle d at q.
point P on the surface S with associated unit vector r directed along the line
from q to P , and let da denote the differential element of surface area at that
point with outwardly directed unit normal vector n, as illustrated in Fig. 2.1.
The flux of E passing through the directed element of area da = nda of S is
then given by
1 r n
E nda = q da (2.10)
40 r2
The quantity (r n/r2 )da is the differential element of solid angle d sub-
tended by da at the position of the point charge. With this identification,
the flux of E passing through the directed element of area da = nda of S
becomes E nda = (q/40 )d. The total flux of E passing through S in the
outward direction His then given by integrating H this expression over the entire
surface S, where S d = 4 if q S and S d = 0 if q / S, resulting in
Gauss law for a single point charge
(
1 q, if q S
I
E nda = (2.11)
S 0 0, if q
/S
2.2 Gauss Law 11
Notice that Gauss law does not provide a result if q is situated on S because
the direction of the unit vector r is then not uniquely defined. For a system
of discrete point charges qj , the principle of superposition applies and Gauss
law becomes
1 X
I
E nda = qj , (2.12)
S 0 j
where the summation extends over only those charges that are inside the
region enclosed by the surface S. If the charge system is described by the
charge density (r), one finally obtains the integral form of Gauss law
1
I ZZZ
E nda = (r)d3 r (2.13)
S 0 V
where V is the volume enclosed by the surface S.

This derivation of Gauss law, as given in Eq. (2.11), depends only upon
the following two mathematical properties of the electrostatic field embodied
in Coulombs law: (1) the inverse square law for the force between point
charges and (2) the central nature of the force. Remove either one of these two
properties and Gauss law no longer follows. The generalized form of Gauss
law in Eq. (2.13) relies upon the additional property of linear superposition.
For a system of discrete point charges qj located at r = rj , the charge
density is given by X
(r) = qj (r rj ),
j
where (r rj ) (x xj )(y yj )(z zj ) is the three-dimensional Dirac

delta function (see Appendix A). With this substitution in Eq. (2.13), the
integral form of Gauss law becomes
1 X 1 X
I ZZZ
E nda = qj (r rj )d3 r = qj
S 0 j V 0 j
which recaptures the microscopic form (2.12) of Gauss law for a system of
discrete point charges with locations rj V. H
With application of the divergence theorem S E nda = V ( E)d3 r,
RRR
the integral form of Gauss law given in Eq. (2.13) becomes

ZZZ
( E /0 )d3 r = 0.
V
Because this expression holds for any region V, the integrand itself must then
vanish, so that
(r)
E(r) = (2.14)
0
which is the differential form of Gauss law .

This single vector differential relation is not sufficient to completely deter-
mine the electric field vector E(r) for a given charge density (r). Helmholtz
theorem (see Appendix B) states that a vector field can be specified almost
completely (up to the gradient of an arbitrary scalar field) if both its di-
vergence and curl are specified everywhere. The required curl relation for
the electrostatic field follows from the integral form (2.9) of Coulombs law,
expressed here as
1 r r 3
ZZZ
E(r) = (r ) d r, (2.15)
40 |r r |3
1 rr

the integration extending over all space. Because |rr | = |rr |3 where
the gradient operator operates only on the unprimed coordinates, then
1 (r ) 3
ZZZ
E(r) = d r. (2.16)
40 |r r |
Because the curl of the gradient of any well-behaved scalar function identi-
cally vanishes, Eq. (2.16) shows that
E(r) = 0 (2.17)
which is Faradays law for the electrostatic field.
2.3 The Electrostatic Scalar Potential and Work
The mathematical form of Coulombs law in Eq. (2.16) suggests that a scalar
potential for the electric field vector be defined as
E(r) = V (r) (2.18)
where the minus sign is introduced by convention, and
1 (r ) 3
ZZZ
V (r) = d r (V ) (2.19)
40 |r r |
the integration extending over the entire region of space containing the charge
distribution under consideration. Notice that V (r) is not uniquely determined
by Eq. (2.19) as one may add to it any quantity that has a zero gradient
without changing E(r). In addition, note that the principle of superposition
applies to the electrostatic potential V (r) just as it does to the electrostatic
field vector E(r).
2.3 The Electrostatic Scalar Potential and Work 13
The convergence properties of the integral appearing in Eq. (2.19) are

essential for the proper formulation of a given electrostatic problem. To that
end, consider the behavior of the scalar potential inside a differential element
of space containing charge. Let the volume element d3 r be a spherical shell
of thickness dr and radius r centered at the point P . Then d3 r = 4r2 dr and
the element of charge in this spherical shell produces a potential
4r2 dr
dV = = rdr,
40 r 0
which remains finite as r 0. The scalar potential for the electrostatic field
then converges for sufficiently well-behaved charge density functions (r).
Consider now determining the work done in transporting a test charge q
from a point A to another point B through an externally produced electro-
static field E(r). The electric force acting on the test charge q at any point in
the field is given by Coulombs law as F(r) = qE(r), so that the work done
in moving the test charge q slowly2 from A to B is given by the path integral
Z B Z B
W = F d = q E d, (2.20)
A A
where the minus sign indicates that this is the work done on the test charge
against the action of the field. With Eq. (2.18) this expression becomes
Z B Z B
W =q V d = q dV = q(VB VA ), (2.21)
A A
which shows that the quantity V (r) can be interpreted as the potential energy
of the charge q in the electrostatic field. The negative sign in Eq. (2.19) is
then seen to indicate that E(r) points in the direction of decreasing potential,
and hence, decreasing potential energy.
The relations appearing in Eqs. (2.20)(2.21) show that the path integral
of the electrostatic field vector E(r) between any two points is independent
of the path and is the negative of the potential difference between them, viz.
Z B
VB VA = E d (2.22)
A
Notice that the electrostatic field only determines the difference between the
electrostatic potential at the two points. If the path is closed, then
I
E d = 0 (2.23)
2
Sufficiently slow such that there are negligible accelerations resulting in negligible
energy loss due to electromagnetic radiation so that the process is reversible.
H RR
With the application of Stokes theorem C E d = S ( E) nda to this
result one immediately obtains E = 0, which is just Eq. (2.17). The
electrostatic field E(r) is then seen to be an irrotational vector field and is
therefore conservative.
For an open system, it is convenient to choose the potential at infinity to
be zero. The electrostatic potential at a point P is then given by
Z P
V (P ) = E d (2.24)

which is referred to as the absolute potential .

The set of points in space that are at a constant potential defines an
equipotential surface. Because V (r) = constant on an equipotential, then
V V V
dV = dx + dy + dz = 0. (2.25)
x y z
Because E = V , the electric field is everywhere perpendicular to the
equipotential surfaces. Lines of force are then defined such that they are
everywhere perpendicular to the equipotentials with E(r) tangent to them.
2.3.1 Poissons Equation for the Scalar Potential
Substitution of Eq. (2.18) into the differential form (2.14) of Gauss Law
yields Poissons equation 3
(r)
2 V (r) = (2.26)
0
Because 2 |r r |1 = 4(r r ), the integral solution of Poissons

equation is given by Coulombs law as [cf. Eq. (2.19)]
1 (r ) 3
ZZZ
V (r) = r r d r ,
(2.27)
40
the integration extending over the entire region containing charge. In charge-
free regions of space the electrostatic potential satisfies Laplaces equation
2 V (r) = 0, (2.28)
which states that the electrostatic potential can never have an extremum in
a charge-free region.
3
Simeon Denis Poisson (17811840) extended [2] Laplaces equation [3] in 1813 to
include regions occupied by charge.
The continuity of the charge density (r) in Poissons equation is not suffi-
cient to ensure the existence of the second partial derivatives of the potential
V (r). This is provided by the following definition due to Holder [4].
Definition 2.1 (Holder Condition). A function f (Q) of the coordinates

of a point Q is said to satisfy a Holder condition at the point P iff there exists
three positive constants A, , and r0 such that
|f (Q) f (P )| Ar Q r r0 , (2.29)
where r is the distance between the points P and Q.
If a region R exists in which f (Q) satisfies a Holder condition at every point

P for a fixed set of values for A, , and r0 , then the function f (Q) is said to
satisfy a Holder condition uniformly, or a uniform Holder condition, in R.
The following theorem due to Holder [4] then holds.4
Theorem 2.1. Let V (r) be the potential of a source distribution with piece-
wise continuous density (r) in a regular region R. Then at any interior point
P0 R at which satisfies a Holder condition, the derivatives of V (r) exist
and satisfy Poissons equation.
2.3.1.1 Boundary Value Problems & Uniqueness
The general boundary value problem for the electrostatic potential V (r) corre-
sponding to a given charge distribution (r) in a particular region of space R
amounts to determining a solution to Poissons equation (2.26) [or Laplaces
equation (2.28) if = 0] that satisfies the given boundary conditions specified
on the system of surfaces S enclosing the region R. The boundary conditions
that lead to a unique solution of Poissons equation are:
Dirichlet Boundary Conditions: The specification of V (r) on a closed sur-
face S forms a Dirichlet problem for the region R enclosed by S.
Neumann Boundary Conditions: The specification of the normal deriva-
tive V /n on a closed surface S forms a Neumann problem for the region
R enclosed by S.
Mixed Boundary Conditions: The specification of Dirichlet boundary con-
ditions on a portion S1 of S and Neumann boundary conditions on the
remainder S2 of S, where S1 S2 = , forms a mixed problem for the region
R enclosed by S = S1 S2 .
The proof of uniqueness with these boundary conditions proceeds in the
usual manner. Suppose that there are two solutions V1 (r) and V2 (r), each
satisfying Poissons equation in the region R and both satisfying the same
boundary condtions on S. Then the difference function V (r) V1 (r) V2 (r),
4
The proof may be found in Kellogg [5].
r R, satisfies Laplaces equation in R; that is 2 V (r) = 0 r R and

either V (r) = 0 on S (Dirichlet boundary conditions), or V (r)/n = 0 on S
(Neumann boundary conditions), or V (r) = 0 on S1 and V (r)/n = 0 on S2
with S = S1 S2 and S1 S2 = RRR (mixed boundary conditions).H Application
2 2
3
of Greens first integral identity R + d r = S n d r to
V (r) then gives
V 2
ZZZ I
2
3
V V + V V d r = V d r,
R S n
where V /n = V n. As 2 V = 0 in R and either V = 0 or V /n = 0

on S, then ZZZ
V 2 d3 r = 0,

R
which then implies that V = 0 everywhere in the region R. Consequently,
V is a constant in R. It then follows that:
for Dirichlet boundary conditions, V = 0 on S so that the constant must
be zero and V1 (r) = V2 (r) in R and the solution is unique;
for Neumann boundary conditions, V /n = 0 on S and, apart from an
unimportant arbitrary additive constant, the solution is unique;
for mixed boundary conditions, V = 0 on part of S so that the constant
must be zero and V1 (r) = V2 (r) in R and the solution is unique.
A solution when both V and V /n are specified arbitrarily on a closed
boundary surface S, known as Cauchy boundary conditions, does not always
exist because there are separate unique solutions for the Dirichlet and Neu-
mann boundary value problems and these will not, in general, be consistent.
The formal solution of Poissons equation using Greens functions is pre-
sented in Appendix C. This representation naturally leads to the powerful
method of images approach for solving field problems in the presence of con-
ducting surfaces. For completeness, the solution of Laplaces equation in spe-
cific separable coordinate systems by the separation of variables method is
also presented in Appendix C.5
2.3.1.2 Average Electrostatic Potential Over a Sphere
Consider a spherical surface S of radius a > 0 carrying a uniform surface

charge density s with total charge Q = 4a2 s , as illustrated in Fig. 2.2.
Because s is spherically symmetric, the electrostatic field vector is radially
directed from the center O of the sphere and is a function of the radial
distance R alone, so that E(r) = 1R E(R). By Gauss law, the electric field
intensity at an observation point P a distance R > a from the center O of the
5
A thorough resource on the solution of boundary value problems for both electric
and magnetic fields may be found in the classic text by Smythe [6].
P
r
Fig. 2.2 Spherical sur-
face S with origin O and
a R
radius a. The exterior
point P is at a distance
R > a from O, and the O
variable distance from the
surface S to P is denoted S
by r.
sphere is given by E(R) = 4Q0 R2 and the absolute potential is V (R) = 4Q0 R .
H s
The potential is also given by Coulombs law as V (R) = S 4 0r
da, where
s = Q/4a2 . Upon equating these two expressions, one obtains
Q Q da
I
= 2
, (2.30)
40 R 4a S 40 r
which then results in the geometrical identity
1 1 da
I
= , R>a (2.31)
R 4a2 S r
Theorem 2.2. The average value of 1r over a spherical surface S, where r is

the distance from a point on the surface S to an exterior point P , is equal to
1
R , where R is the distance from the center O of the sphere S to P .
If the surface charge density s = Q/4a2 is removed and a point charge Q

is placed at the exterior point P , the potential at the center O of the sphere
is then given by the left-hand side of Eq. (2.30). The right-hand side of this
equation is then just the average potential on the spherical surface. Hence:
Theorem 2.3 (Mean Value Theorem). The average potential over any
spherical surface is equal to the potential at the center of the sphere if there
are no charges inside the sphere.
Corollary 2.1. It is impossible to have a potential maximum or minimum

in a charge-free region.
Consider again a spherical surface S of radius a carrying a uniform surface

charge density s = Q/4a2 with total charge Q, as illustrated in Fig. 2.3.
Application of Gauss law to any concentric spherical surface S of radius
R < a shows that, at any point P inside S, the electrostatic field intensity
is zero because there isnt any enclosed charge. The electrostatic potential
V (R) at any point P interior to S must then be equal to the potential at the
surface, so that
r
Fig. 2.3 Spherical sur-
P
face S with origin O and a
R
radius a. The interior O
point P is at a distance
R < a from O, and the
variable distance from the S
surface S to P is denoted
by r.
Q s Q da
I I
V (R) = = da = , (2.32)
40 a S 40 r 4a2 S 40 r
which results in the geometrical identity
1 1 da
I
= , R<a (2.33)
a 4a2 S r
If the surface charge density s = Q/4a2 is removed and a point charge

Q placed at the interior point P , then the final expression in Eq. (2.33) is
seen to be the average potential taken over the spherical surface S. This then
establishes the following theorem:
Theorem 2.4. The average electrostatic potential taken over any spherical
Q
surface is equal to 4 0a
, where a > 0 is the radius of the sphere and Q is the
total enclosed charge, provided that there is no charge outside the sphere.
2.4 The Concept of an Ideal Conductor
In order to complete the formal mathematical structure of the electrostatic

field, it is necessary to introduce the concept of an ideal conductor.
Definition 2.2 (Ideal Conductor). An ideal conductor is defined as a
medium inside and on the surface of which electric charge freely flows un-
der the influence of an externally applied electric field.
Such an idealized medium is necessary not only from a mathematical point
of view as it provides a convenient approach by which idealized boundary
conditions may be introduced into the theory, but also from a physical point
of view since an ideal conductor may be viewed as a material system whose
electrical properties can be approached in some physically realizable manner.
Because the electrostatic field is itself an idealization, it is entirely compatible
with the concept of an ideal (or perfect) conductor. As an immediate con-
sequence of this definition, the electrostatic field inside an ideal conductor
2.4 The Concept of an Ideal Conductor 19
must be zero as any nonzero internal field would induce a current flow. The
propagation of such a current either in or on an ideal conductor would nec-
essarily involve a dissipation of energy in the field and hence cannot occur in
any static state of the field. This result them leads to the following equivalent
definition of an ideal conductor.
Definition 2.3 (Ideal Conductor). An ideal conductor may be defined as
a medium that is incapable of sustaining an electrostatic field in its interior.
Application of Gauss law [see Eq. (2.15)] to the interior region V of an
ideal conductor results in the immediate conclusion that
(r) = 0 E(r) = 0; r V. (2.34)
Hence, any net static charge on an ideal conductor must reside on its surface.
When a conductor is charged, the charges arrange themselves in such a man-
ner that the fields they produce in the interior region of the conductor body
are mutually balanced and the net interior electric field is zero. If an ideal con-
ductor is placed in an external electrostatic field, the charges flow temporarily
within it (this has now temporarily become a non-static arrangement) so as
to set up a surface charge distribution which produces an additional electric
field that, when added to the initial external electrostatic field, results in a
zero field inside the conductor once static conditions are reestablished. The
electrostatic field external to the conductor body is altered in this process as
one electrostatic arrangement is replaced by another.
Fig. 2.4 A perfectly

conducting body with a
cavity. The dotted surface
S, to which Gauss law
S
is applied to determine
the enclosed charge, lies
entirely within the body
and encloses the cavity.
As a practical application of the properties of an ideal conductor, consider a

conducting body containing a cavity that is placed in an externally generated
electrostatic field. Application of Gauss law to any surface S that encloses the
cavity and is completely contained
H within the conducting body, as depicted
in Fig. 2.4, yields Qenc = 0 S E nda = 0 for the charge enclosed, since the
electrostatic field vanishes on S. If charges are completely contained within
the cavity, then the induced surface charge residing on the inner cavity surface
exactly cancels their net charge. Hence, if no charge is enclosed within the
cavity then there is no surface charge induced on the inner cavity surface.
The following two special cases are then realized:
1. No externally applied electrostatic field: A charge q inside the cavity then

induces an equal but opposite surface charge on the inside surface of the
conductor. In order to preserve the net charge neutrality of the conductor
body itself, a net surface charge q must then appear on the outside surface
of the conductor body which, in turn, produces an external E-field.
2. An externally applied electrostatic field: Let there now be no charge en-
closed within the cavity. Configure the Gaussian surface S such that it
is entirely within the conductor body and is situated an infinitesimal dis-
tance from the outer surface of the conductor body. Because Qenc = 0,
the entire net charge induced on the conductor must be external to S and
hence, resides on the outside surface of the conductor body. Thus, when a
hollow conductor with zero enclosed charge is placed in an externally gen-
erated electrostatic field, a surface charge is induced only on the outside
surface of the conductor body.
Consequently, a charge situated in the region exterior to a hollow conductor
does not produce an electrostatic field in the interior cavity of the conductor
body, whereas a charge situated in the interior cavity of the conductor body
produces an electrostatic field in the region external to the conductor. The
first of these two results forms the basis of electrostatic shielding by which
the enclosed region may be shielded from external fields.
The boundary conditions imposed on the external electrostatic field vector
E(r) at the surface of an ideal conductor follow directly from the fact that
E identically vanishes inside the conductor body. Application of Gauss and
Faradays laws [see Eqs. (2.14) and (2.17)] to any point on the conductor
surface yields
n E(r) = s (r)/0 , r S, (2.35)

n E(r) = 0, r S, (2.36)
respectively, where n is the unit outward normal vector to the conductor

surface S and where s (r) denotes the surface charge density on S. The second
relation states that the tangential component of the external electrostatic field
must vanish on the conductor surface if there is to be no induced current flow
along the surface. Hence, the electrostatic field vector must be everywhere
normal to the conductor surface S so that, from Eq. (2.35),
s (r)
E(r) = n , r S. (2.37)
0
Furthermore, because E(r) = V (r), this result shows that the electro-
static scalar potential of the field must be constant on the surface of an ideal
conductor; that is, the surface of an ideal conductor is an equipotential sur-
face of the electrostatic field. The distribution of charge over the conductor
surface is then given by s = 0 n E = 0 nV , so that
2.5 The Electric Dipole, Quadrupole, and Multipoles 21
V (r)
s (r) = 0 , r S, (2.38)
n
the derivative of the potential being taken along the outward normal to the
surface. The total charge on the conductor is then given by
V
I
Qs = 0 da. (2.39)
S n
Consider now the scalar potential V (r) that is established in some region
D of space by a system of charged conductors and (or) other external sources.
The potential then satisfies Laplaces equation in D. Assume that this poten-
tial has a maximum value at some point P D that is not on the boundary of
some subregion where there is no field. The point P may then be surrounded
by a closed surface that is contained entirely within the region D and on which
the normal derivative satisfies the inequality V /n < 0; if P is on the bound-
ary of a subregion where there is no field, then 2 V = 0 in the region about
that point and the field domain may accordingly be extended (in a mathe-
matical sense) across theH boundary without altering the field. Integration over
the surface then gives V n da < 0. However, application of the divergence
theorem followed by Laplaces equation yields V
H RRR 2 3
n da = V d r = 0,
and one has obtained a contradiction. Hence, in any charge-free region of
space the electrostatic potential V (r) can assume maximum and minimum
values only at the boundaries of regions where there is a field. This then
shows that a test charge q that is introduced into an electrostatic field cannot
be in static equilibrium because there is no point where its potential energy
qV would have a minimum. This result may then be generalized as:
Theorem 2.5 (Earnshaws Theorem). A charged body placed in an elec-

trostatic field cannot be maintained in stable equilibrium under the influence
of the electrostatic forces alone.
2.5 The Electric Dipole, Quadrupole, and Multipoles
The properties of the electrostatic field produced by simple charge config-

urations leads to important results concerning the dielectric properties of
material media and so are now considered in some detail.
2.5.1 The Static Electric Dipole
The electric dipole consists of a positive and negative charge of equal mag-
nitude Q separated by a distance s, as depicted in Fig. 2.5. With the z-axis
P(r,,)
Fig. 2.5 Static electric

rb
dipole formed by two
point charges of equal
r
but opposite charge Q +Q
v ra
1r
separated by the fixed s/2
distance s. The origin of O
coordinates O is taken s/2
-Q
at the midpoint of the y
line joining the two point
charges and the field x
observation point P is
located at the spherical
polar coordinates (r, , ).
along the dipole axis through the two point charges and the origin O at the
midpoint between them, the electrostatic potential at any point P (r, , ) is
given by the superposition of the potential due to each charge alone as

Q 1 1
V (r, , ) = . (2.40)
40 rb ra
Application of the law of cosines to the pair of triangles QOP in Fig.

2
2.5 gives r = r2 + (s/2)2 rs cos with ra = r+ and rb = r , so that
2 2 2 2 3
r 1 s s 3 s s
r = 1 2 4r 2 r cos + 8 4r 2 r cos 15 s s
48 4r 2 r cos + .
Upon arranging terms in ascending powers of s/r, there results
r s s2 15s3
= 1 P1 (cos ) + 2 P2 (cos ) P3 (cos ) + , (2.41)
r 2r 4r 8r3
where P0 (cos ) = 1, P1 (cos ) = cos , P2 (cos ) = 21 (3 cos2 1), P3 (cos ) =

1 3
2 (5 cos 3 cos ), etc., are the Legendre polynomials (see Appendix C). The
electrostatic dipole potential is then given by
15s2 P3 (cos )

Qs
V (r, , ) = P1 (cos ) 1 + + , (2.42)
40 r2 4r2 P1 (cos )
which falls off as 1/r2 , the potential due to a point charge falling off as 1/r.
The dipole moment of the charge pair is defined as
p Qs = Qs1s (2.43)
with magnitude p = Qs, where the unit vector 1s is directed from the negative
to the positive charge along the dipole axis. With this definition, the first-
term approximation of the expression (2.42) for the absolute electrostatic

dipole potential becomes
p 1r
V (r) , r 3 s3 , (2.44)
40 r2
the correction factor being given by the bracketed quantity in Eq. (2.42).
An ideal point dipole is obtained in the limit as s 0 with fixed dipole
moment p and so is given by Eq. (2.44) without approximation. Equipoten-
tial surfaces V (r, ) = constant for an ideal point dipole are indicated by the
dotted curves in Fig. 2.6, with darker shading indicating increasing poten-
tial magnitude. Notice that the absolute potential is positive (negative) in
the half-space that the dipole moment p is directed into (out of), the poten-
tial vanishing in the equatorial plane = /2. In addition, notice that the
Axis of
Symmetry
(Dipole Axis)
Fig. 2.6 Electrostatic

field lines of force (dashed
curves) and equipoten-
V>0
tials (dotted curves) for dr Er
an ideal point dipole. The r
ds
full three-dimensional rd
V=0
field structure is obtained E
E(r,)
by rotating the figure

about the dipole axis.
V<0
The darkness of shad-
ing between neighboring
equipotentials indicates
increasing potential mag-
nitude.
dipole potential identically vanishes in the equatorial plane ( = /2) since

ra = rbh for all s 0. Finally, ifor a non-ideal dipole (s > 0), the correction
2 P3 (cos )
factor 1 + 15s4r 2 P1 (cos ) + appearing in Eq. (2.42) begins to deviate sig-
nificantly from unity when r decreases below 2s. For smaller values of r, a
series expansion of 1/r in powers of r/s is appropriate.
The electrostatic field vector for a point dipole is obtained from the neg-
ative gradient of either Eq. (2.44) or Eq. (2.42) with the result
3(p r)r r2 p p
E(r) = 5
= 3
1r 2 cos + 1 sin , (2.45)
40 r 40 r
the second form of the result applying when the point dipole p = p1z is
located at the origin as in Fig. 2.5. As indicated in Fig. 2.6, the lines of force
dr Er 2 cos dr 2 cos d
for this field are specified by rd = E = sin so that r = sin =
2d(sin ) 2
sin ,with solution r = A sin . Electrostatic field lines are indicated by
the dashed curves in Fig. 2.6, the electrostatic dipole field vector being in the
direction from higher to lower potential, as indicated by the arrows in the
figure.
2.5.2 The Linear Electric Quadrupole
The linear electric quadrupole consists of an inverted pair of electric dipoles of

equal dipole moment magnitude p = Qs with centers displaced by the dipole
separation s along their common axis, as illustrated in Fig. 2.7. With the
z
Fig. 2.7 Linear electric
quadrupole formed by two P(r,,)
colinear dipoles of equal
but opposite moment p
with centers displaced by rb

the dipole separation s
r
along their common axis.
ra
The origin of coordinates +Q
v
1r
O is taken at the midpoint s
of the line joining the -2Q
s
two dipoles and the field
+Q
observation point P is
located at the spherical
polar coordinates (r, , ).
z-axis taken along the linear quadrupole axis with origin O at the midpoint
of the arrangement, the potential at any point P (r, , ) is given by the
superposition of the potential due to each charge alone as

Q 1 1 2 Q r r
V (r, , ) = + = + 2 . (2.46)
40 ra rb r 40 r ra rb
From the expansion given in Eq. (2.41) [with 2s replaced by s] one has that
r s s2 r s s2
ra 1 r P1 (cos ) + r 2 P2 (cos ) and rb 1 + r P1 (cos ) + r 2 P2 (cos ),
provided that (s/r)3 1. With these substitutions, the expression (2.46) for
the absolute electrostatic potential of a linear quadrupole becomes
2Qs2
V (r, , ) P2 (cos ), r 3 s3 , (2.47)
40 r3
which varies inversely as the cube of the radial distance r > 0 from the center
of the quadrupole charge structure. Equipotential surfaces and electrostatic
field lines for the ideal linear quadrupole are illustrated in Fig. 2.8.
Axis of
Symmetry
(Quadrupole Axis)
V>0
Fig. 2.8 Electrostatic V=0

field lines of force (dashed
curves) and equipoten-
tials (dotted curves)
for an ideal axial point V<0
quadrupole. The full

three-dimensional field
structure is obtained by
rotating the figure about
V=0
the quadrupole axis. The
darkness of shading be-
tween neighboring equipo- V>0
tentials indicates increas-
ing potential magnitude.
2.5.3 Static Electric Multipoles
The field properties of static electric multipoles can be understood through

the following geometric construction sequence (with |sj | s):
The monopole field is a single point charge Q with potential
V0 Q/r.
The dipole field is obtained by displacing a monopole through a distance s1

and replacing it by one of equal but opposite sign, resulting in the potential
V1 Qs/r2 .
The quadrupole field is obtained by displacing a dipole through a distance s2

and replacing it by one of equal but opposite sign, resulting in the potential
V2 Qs2 /r3 .
The octupole field is obtained by displacing a quadrupole through a distance

s3 and replacing it by one of equal but opposite sign, resulting in the potential
V3 Qs3 /r4 .
For a 2 -multipole, with denoting the number of independent displacements

s1 , s2 , , s required to specify the charge arrangement, the potential is
V Qs /r+1 , (2.48)
and the associated electric field then varies as
E Qs /r+2 . (2.49)
For a linear point multipole, the angular dependence is described by the

corresponding Legendre polynomial P (cos ) given in Eq. (C.70) of Appendix
C. The electrostatic potential due to an ideal monopole is then given by
Q
V0 (r, , ) = P0 (cos ) (2.50)
40 r
with P0 () = 1, that due to an ideal (point) dipole [cf. Eq. (2.44)]
Qs
V1 (r, , ) = P1 (cos ) (2.51)
40 r2
with P1 () = , that due to an ideal linear (point) quadrupole [cf. Eq. (2.47)]
Qs2
V2 (r, , ) = P2 (cos ) (2.52)
40 r3
with P2 () = 12 (3 2 1), that due to an ideal (point) octupole
Qs3
V3 (r, , ) = P3 (cos ), (2.53)
40 r4
with P3 () = 12 (5 3 3), and so-on.
2.6 The Electrostatic Field Produced by an Arbitrary

Static Charge Distribution
Consider now obtaining an expansion of the electrostatic field produced by

an arbitrary (but fixed) charge distribution with density (r ) occupying a
region in free-space and extending to a maximum distance rmax from the
origin O of a fixed coordinate system in terms of its multipole moments
about that point, as illustrated in Fig. 2.9. It is assumed that O is positioned
either within or is in close proximity to the charged region . The absolute
electrostatic potential V (r) = V (x, y, z) is then given by
1 (x , y , z ) 3
ZZZ
V (x, y, z) = d r (2.54)
40 r
2.6 The Electrostatic Field Produced by an Arbitrary Static Charge Distribution 27
Fig. 2.9 An arbitrary P(x,y,z)

r
(but fixed) charge distri-
P(x,y,z)
bution (r ) occupying a
finite region in free-
r
space. The unit vector r

1r is directed from the
origin O (assumed to be V
located either within or O 1r

in close proximity to it) to
the field point P (x, y, z)
a distance r away, where y
x
r > rmax .
1/2
where r = (x x )2 + (y y )2 + (z z )2

is the distance from the
source point at P (x , y , z ) to the field point at P (x, y, z) a distance r from
O. Because the source point P is near to the origin O and provided that the
field point P is far removed from O such that r > rmax , the quantity 1/r
may be expanded in a Taylor series about the origin as
1 1 1 1 2 1

r = r O +(x x +y y +z z ) r O + 2! (x x +y y +z z ) r O + ,
(2.55)
2 2 2 2
2 2 2
where (x x + y y + z z ) = x2 x 2 + y y 2 + z z 2 + 2x y x y +
2 2
2x z x z + 2y z y z . The first term appearing in the Taylor series expan-
sion (2.55) is given by 1/r |O = 1/r |(x ,y ,z )=(0,0,0) = 1/r. For the second
term, one has that (/x )(1/r )|O = (x x )/r3 |O = /r2 , and similarly
that (/y )(1/r )|O = m/r2 and (/z )(1/r )|O = n/r2 , where x/r,
m y/r, and n z/r are the cosines of the angles between the position
vector r = 1r r and the x, y, and z-axes, respectively. For the third term, one
has that ( 2 /x2 )(1/r )|O = (32 )/r3 , ( 2 /y 2 )(1/r )|O = (3m2 )/r3 ,
( 2 /z 2 )(1/r )|O = (3n2 )/r3 , ( 2 /x y )(1/r )|O = ( 2 /y x )(1/r )|O
= 3m/r3 , ( 2 /x z )(1/r )|O = ( 2 /z x )(1/r )|O = 3n/r3 , and
( 2 /y z )(1/r )|O = ( 2 /z y )(1/r )|O = 3mn/r3 . With these substi-
tutions in Eq. (2.55), the expression (2.54) for the electrostatic potential at
the field point P (r) = P (x, y, z) may be expressed as
V (x, y, z) = V0 (x, y, z) + V1 (x, y, z) + V2 (x, y, z) + , (2.56)
where each term Vj (x, y, z), j = 0, 1, 2, 3, . . . , corresponds to the related term

in the Taylor series expansion of 1/r .
The zeroth-order term in this expansion,
1 Q
ZZZ
V0 (x, y, z) = (x , y , z )d3 r = , (2.57)
40 r 4 0r
where Q is the net charge in , is called the monopole term because it is

the potential one would have at P if the entire charge distribution were
concentrated at O. This term is zero in the multipole expansion (2.56) only if
the net charge Q is zero. If Q is non-vanishing, then V0 (x, y, z) is the dominant
term in the multipole expansion as r because it decreases only as r1 .
Notice that the value of Q is independent of the origin O.
The first-order term in the multipole expansion (2.56) is the dipole term
p 1r
V1 (x, y, z) = (2.58)
40 r2
with 1r 1x + 1y m + 1z n denoting the unit vector along the radial line

from O to the field point P (r) = P (x, y, z), as indicated in Fig. 2.9, where
ZZZ
p r (r )d3 r (2.59)

is the dipole moment of the charge distribution in taken with respect to

O. Hence, the first-order term V1 (r) describes the potential at the field point
P (r) due to an effective dipole at the origin O with dipole moment p.
The dipole moment of an extended charge distribution may also be defined
as
p = Qr , (2.60)
where Q is the total net charge in the distribution, and where
RRR 3
r (r )d r 1
ZZZ
r = RRR )d3 r
= r (r )d3 r (2.61)

(r Q
is a vector extending from the origin O to the charge centroid of the extended
charge distribution. Notice that if Q = 0, then r and p, as given by
Eq. (2.60), is indeterminate; however, Eq. (2.59) always determines the dipole
moment unambiguously and is to be used in that singular case. Notice that
in the Q = 0 case, the dipole moment is independent of the choice of origin.
Finally, if Q 6= 0, then the dipole moment of the charge distribution can
always be made to vanish by choosing the origin O at the centroid of the
charge distribution which is determined by setting r = 0.
The second-order term in the expansion (2.56) is the quadrupole term

1
V3 (r) = 3mQxy + 3nQxz + 3mnQyz
40 r3

1 1 1
+ (32 1)Qxx + (3m2 1)Qyy + (3n2 1)Qzz . (2.62)
2 2 2
The scalar quantities Q define the quadrupole moment tensor Q = (Q ),

, = x, y, z, of the charge distribution, with
2.7 The Concept of a Perfect Dielectric 29
ZZZ
Q = Q (r )d3 r = Q . (2.63)

This set of expressions for the quadrupole term is simplified considerably

if the charge distribution possesses certain symmetries. For example, if the
charge distribution in possesses cylindrical symmetry about the z-axis,
then Qxy = Qyz = Qxz = 0 and Qxx = Qyy . It is then convenient to define
a single scalar quadrupole moment Q of the charge distribution as
ZZZ
Q 2(Qzz Qxx ) = (3z 2 r2 )(r )d3 r , (2.64)

where r2 = x2 + y 2 + z 2 . Under these conditions, Eq. (2.63) becomes
Q 3n2 1 Q 3 cos2 1
V3 (r) = 3
= , (2.65)
40 4r 40 4r3
where n z/r = cos with denoting the angle between the positive z-axis
and the line segment of length r extending from the origin O to the field point
P = P (r, ). The multipole expansion of the electrostatic potential due to
a cylindrically-symmetric charge distribution (r ) is then given by [cf. Eqs.
(2.50)(2.53)]
Q p Q
V (r, ) = + P1 (cos ) + P2 (cos ) + (2.66)
40 r 40 r2 80 r3
in terms of the Legendre polynomials P (cos ). The first nonvanishing term
in this multipole expansion then dominates the behavior of V (r, ) as r .
2.7 The Concept of a Perfect Dielectric
In a perfect dielectric medium, all of the charged particles are bound either
in atomic or molecular configurations. When an external electric field is ap-
plied, the positive and negative charges bound in each molecule are displaced
in opposite directions and the molecular charge density of each molecule
is accordingly distorted. Positive charge is displaced in the direction of E
and negative charge in the opposite direction so that the induced molecular
dipole moment is in the same direction as E; each molecule then produces
an average electric field that is in a direction opposite to E. After static con-
ditions are re-established, the multipole moments of each molecule will differ
from their zero field (unperturbed) values. For a simple dielectric, the domi-
nant multipole that is induced is the dipole, all higher-order multipoles being
negligible by comparison. The dielectric is then said to be polarized by the
external electric field with its molecules possessing induced dipole moments.
In a nonpolar dielectric, the molecules have zero permanent dipole moments
so that, in the absence of an applied electric field, the dipole (and higher-
order multipole) moments are all zero. In a polar dielectric, the molecules
possess a nonzero permanent dipole moment such that, in the absence of an
applied field, the molecular dipole moments are randomly oriented so that
the spatially-averaged dipole (and higher-order multipole) moments are again
all zero. Application of an external electric field in such a simple dielectric
then produces a macroscopic electric polarization density P(r) given by the
spatial average of the microscopic dipole moment as
X
P(r) = Nj hpj (r)i, (2.67)
j
where pj (r) is the microscopic dipole moment of the jth-type of molecule

comprising the dielectric, hpj (r)i is the spatial average of this microscopic
dipole moment taken over a macroscopically small but microscopically large
region centered at the point r, and where Nj is the average number density of
j-type molecules in that region. Because the net charge in a perfect dielectric
is zero, this macroscopic dipole moment density P(r) is independent of the
choice of origin [see the discussion following Eq. (2.61)].
The connection between microscopic and macroscopic field quantities is
obtained through the same spatial-averaging process. The macroscopic elec-
tric field vector E(r) for the electrostatic field is defined as the spatial average
of the microscopic electric field vector e(r) as
E(r) he(r)i. (2.68)
There is then no distinction between microscopic and macroscopic fields in

vacuum. When the spatial-averaging procedure is applied to Faradays law
(2.17) for the microscopic electrostatic field, the same equation results, viz.
E(r) = 0. (2.69)
This result then implies that the macroscopic field may likewise be expressed
in terms of a macroscopic scalar potential V (r) as
E(r) = V (r). (2.70)
The spatial average of Gauss law (2.14) for the microscopic electrostatic field
yields
1
E(r) = h(r)i, (2.71)
0
so that the divergence of the macroscopic electrostatic field vector is deter-
mined by the spatial average of the microscopic charge density in the dielec-
tric. The proper description of this quantity deserves careful attention (see
Ch. 3 of Lorrain and Corson [7] and Ch. 13 of Kittel [8] as well as Vol. I of
Bottchers two volume treatise on electric polarization [9]).
Consider first determining the electrostatic field produced by a dielectric
body with macroscopic dipole moment density P(r) at a field point exterior
to the body in vacuum. From Eq. (2.44), the potential dV (r) produced at the
exterior point r due to the dipole P(r )d3 r at the interior point r is given
by (40 )dV (r) = (P(r ) 1r /r2 )d3 r = [P(r ) (1/r)]d3 r , where 1r is the
unit vector directed from the source point r to the field point r, separated
by the (nonvanishing) distance r = [(x x )2 + (y y )2 + (z z )2 ]1/2 , and
where operates on the primed (source) coordinates. The total potential
at the exterior point r in vacuum is then given by

1 1
ZZZ

V (r) = P d3 r
40 V r
P 3
Z Z Z
1 P
ZZZ
3
= d r d r
40 V r V r
P 3
I
1 P n 2
ZZZ
= d r d r ,
40 S r V r
where S is the boundary surface to the dielectric region V with outward unit
normal vector n. Comparison of this result with the form (2.9) of Coulombs
law shows that this external field is produced by both a surface polarization
charge density sb (r) P(r) n, r S, and a volume polarization charge
density b (r) P(r), r V, so that
sb (r ) 2 b (r ) 3
I
1
ZZZ
V (r) = d r + d r .
40 S r V r
Consider next an arbitrary but fixed point inside the dielectric body
V such that a sphere of radius R centered at can be constructed with
surface S lying entirely within V, thereby dividing the dielectric into two
regions, the region V inside S and the region V outside S but inside S.
The radius R is chosen sufficiently small such that the enclosed volume V
is macroscopically small, in which case the macroscopic electric field vector
E(r), dipole moment density P(r), and polarization charge density b =
P do not vary appreciably in V . The potential at due to the dipole
distribution in the exterior region V is given by the preceding result with
the surface integral taken over both S and S and the volume integral taken
over V , with electric field intensity
"I #
1 sb (r ) 2 sb (r ) 2 b (r ) 3
I ZZZ

E () = r d r + r d r + r 2 d r ,
40 S r2 S r2 V r
where r is the unit vector from the dipole source to the field point and r is
the distance between these two points.
The average electric field intensity E () due to the near dipoles enclosed
by S is given by the summation over j of the product of the molecular dipole
field hpj i/(40 R3 ), where hpj i denotes the average dipole moment of the
j-type molecule in S , times the number density Nj of j-type molecules times
the volume (4/3)R3 enclosed by S , so that

P
4 3 j Nj hpj (r)i P
E () = R = ,
3 40 R3 30
P
where P = j Nj hpj (r)i, as defined in Eq. (2.67).
The integral taken over the spherical surface S centered at may be
evaluated in the following manner. Construct an axis in the direction 1P
parallel to the dipole moment density P and let be the angle between 1P
and the line extending from a point on the surface S to the point , so that
sb = P n = P cos , where n is the unit normal to the surface S directed
inwards toward the center . Then
Z
sb (r ) 2 4
I
r 2
d r = 2P cos2 sin d = P,
S r 0 3
since r = 1 cos for this geometry. This term then cancels the near dipole
contribution E () = P/30 given above.
The total electrostatic field intensity produced at an interior point of a
polarized dielectric is then given by
sb (r ) 2 b (r ) 3
I
1
ZZZ
E(r) = r d r + r d r ,
40 S r2 V r2
where the integration over the region V has been extended to the entire
volume V of the dielectric body by adding the contribution from the spherical
region V which vanishes as it describes the electrostatic field at the center
of a uniform spherical charge distribution.
Taken together, these results show that the electrostatic potential and field
both inside and outside a simple dielectric with free charge are given by
s (r ) + sb (r ) 2 (r ) + b (r ) 3
I
1
ZZZ
V (r) = d r + d r ,
40 S r V r
s (r ) + sb (r ) 2 (r ) + b (r ) 3
I
1
ZZZ
E(r) = r d r + r d r ,
40 S r2 V r2
where s is the free surface charge density on S and is the free volume
charge density in V.
For a simple dielectric, the spatial average of the microscopic charge den-
sity [see Eq. (2.71)] is thus seen to be given by
h(r)i = (r) P(r), (2.72)
where denotes the macroscopic charge density in the dielectric and P is the
macroscopic polarization density defined in Eq. (2.67). The presence of the
divergence of P(r) in this spatial average of the microscopic charge density
accounts for any spatial nonuniformity in this vector field and is referred to
as the polarization or bound charge density b , where
b (r) P(r) (2.73)
for all points r in the interior of the dielectric body. In addition, associ-
ated with this macroscopic dipole moment density is the surface polarization
charge density
sb (r) P(r) n (2.74)
for all points on the surface of the dielectric body with outward unit normal
vector n. With this substitution, the spatial average of Gauss law becomes
D(r) = (r), (2.75)
where
D(r) 0 E(r) + P(r) (2.76)
is the electric displacement vector (in C/m2 ) for a simple dielectric.
Because
1 1
E(r) = D(r) P(r), (2.77)
0 0
the macroscopic electric field intensity E(r) inside a simple dielectric is then
seen to be given by the sum of two vector fields: the field D(r)/0 associated
with the spatially-averaged molecular charge density (typically zero) plus any
free charge f that is externally supplied, where (D(r)/0 ) = (r)/0 ; and
the field P(r)/0 associated with the bound or polarization charge of the
dielectric, where (P(r)/0 ) = b (r)/0 . The field lines of the electric
displacement vector D then begin and end only on externally supplied (free)
charge as well as on the spatially-averaged molecular charge when this latter
quantity is nonzero, whereas the lines of force of the macroscopic electric field
vector E begin and end on either free or bound (polarization) charge.
For a simple dielectric, the macroscopic electric polarization density P(r)
is linearly related to and in the same direction as the macroscopic electric
field intensity E(r) at that point, so that
P(r) = 0 e E(r), (2.78)
where e is the electric susceptibility of the simple dielectric. Notice that e

is dimensionless and is real-valued in the static case. Taken together, Eqs.
(2.76) and (2.78) give
D(r) = 0 (1 + e )E(r)
= 0 r E(r) = E(r), (2.79)
where is the dielectric permittivity of the medium, given by

0 (1 + e ), (2.80)
and where r = /0 is the relative permittivity of the dielectric. Notice that

is real-valued in the static case.
If the dielectric material is spatially inhomogeneous so that e = e (r)
and = (r) vary with position, then substitution of Eq. (2.71) into Gauss
law (2.75) results in the expression (r)2 V (r) + (r) V (r) = (r),
which may be rewritten as
(r)
2 V (r) + ln ((r)) V (r) =

. (2.81)
(r)
If the dielectric is spatially homogeneous so that is independent of position

within the material, then the above expression reduces to Poissons equation
(r)
2 V (r) = , (2.82)

which further reduces to Laplaces equation in charge-free regions of space.
Boundary conditions on the electrostatic field vectors across an interface
S separating two simple dielectrics with permittivities 1 and 2 may be
obtained by direct application of the integral form of Gauss law to a simple
closed surface with identical faces on opposite sides of S and Faradays law
to a simple closed circuit with identical segments on opposite sides of S with
the results

n D2 (r) D1 (r) = s (r) , r S, (2.83)

n E2 (r) E1 (r) = 0 , r S, (2.84)
where n is the unit normal to the surface at the point r, directed from medium
1 into medium 2.
2.8 Electrostatic Energy
Whenever two charges qa and qb are brought within a distance Rab of each
other, work is expended against the Coulombic force [Eq. (2.2)] in consum-
mating the process. Once the charges are in place, the persistence of this force
makes the energy stored in the electrostatic field potentially available when-
ever demanded. If it is assumed that the charges are moved slowly enough
into place (i.e. reversibly), then their kinetic energies may be neglected and
any loss due to electromagnetic radiation effects, significant if rapid charge
accelerations occur, may then be neglected.
Consider then the energy stored in a fixed configuration of n charges, given
by the reversible work required to assemble the static charge configuration.
Assume that all n charges q1 , q2 , . . . , qn are initially located at infinity in their
zero potential state. Upon bringing just q1 from infinity to its final position
P1 , no work is expended because no other charges are present. The work done
2.8 Electrostatic Energy 35
(1) (2)
in bringing q2 from infinity to P2 is then given by U2 = q2 V2 = q1 V1 ,
(1)
where V2 = q1 /(40 R12 ) denotes the potential at P2 due to the charge q1
(2)
at P1 , and where V1 = q2 /(40 R21 ) denotes the potential at P1 due to the
charge q2 at P2 . The work done in bringing a third charge q3 in from infinity
(1) (2) (3) (3)
to P3 is then given by U3 = q3 V3 + q3 V3 = q1 V1 + q2 V2 , and so on for
the remaining charges q4 , q5 , . . . , qn , taking note of the symmetry relation
(j) (k)
qk Vk = qj Vj (2.85)
(j)
where Vk denotes the electrostatic potential at Pk due to the charge qj at
Pj . The total energy Ue = U1 + U2 + + Un stored in the assembled charge
configuration can then be written in two different ways: First by adding the
(1) (1) (2)
first forms of the above equations, giving Ue = q2 V2 + q3 V3 + q3 V3 +
(1) (2) (3) (1) (2) (n1)
q4 V4 +q4 V4 +q4 V4 + +qn Vn +qn Vn + +qn VN , or by adding the
(2) (3) (3)
second forms of the above equations, giving Ue = q1 V1 + q1 V1 + q2 V2 +
(4) (4) (4) (4) (n) (n)
q1 V1 + q2 V2 + q3 V3 + + q1 V1 + q2 V2 + + qn1 VN . The average
of these two expressions then yields the result
n
1X
Ue = qk Vk (J) (2.86)
2
k=1
for the potential energy of the assembled charge configuration, where qk de-
notes the charge of the k th particle located at the fixed point Pk , and where
Vk denotes the absolute potential at Pk due to all of the charges in the config-
uration except qk . This result has rather general applicability provided that
the potential Vk is properly determined. Notice further that this expression
does not include the self-energy of the individual charges, this being defined
as the energy that would be liberated if each charge was allowed to expand
to an infinite volume. Because of this, Eq. (2.86) identically vanishes for a
single point charge, as required.
For a continuous volume charge density (r), the expression given in Eq.
(2.86) for the electrostatic potential energy generalizes to
1
ZZZ
Ue = (r)V (r)d3 r (2.87)
2 V
with analogous expressions for surface S (r) and line (r) charge densities.
Notice that this generalized expression for the electrostatic potential energy
includes the self-energies of the charges. For example, for a uniform spherical
charge distribution = 3Q/(4r03 ) for r r0 with total charge Q and radius
r0 , the absolute electrostatic potential inside the sphere is found to be given
by
Q Q
r02 r2 +

V (r) = 3 ; r r0 .
80 r0 40 r0
From Eq. (2.87), the self-energy of this spherical charge distribution is then
3Q r0 3Q2
Z
Q 2 2
Q 2
Use = 3 r r + r dr = ,
2r0 0 80 r03 0 40 r0 200 r0
so that Use as r0 0 with fixed Q 6= 0, showing that it requires

infinite energy to construct an ideal point charge. On the other hand, Use =
(/0 )r02 0 as r0 0 with fixed charge density ; the generalized expression
(2.87) for the electrostatic potential energy should then be viewed from this
latter point of view.
Consider finally deriving an expression for the electrostatic energy in terms
of the field quantities alone. From Poissons equation for a spatially homoge-
neous simple dielectric [see Eq. (2.82)], the charge density may be expressed
as (r) = 2 V (r) at every point in the field which, when substituted in
Eq. (2.87), yields

ZZZ
Ue = V (r)2 V (r)d3 r, (2.88)
2 V
where V is any volume containing all of the charges
3 in Hthe system.2 From
2
RRR
Greens first integral identity
VRRR
+ d r = S nd r with
2
)2 d3 r = S V V nd2 r,
H
(r) = (r) = V (r), one obtains V V V +(V
so that I

ZZZ
2 2 3
Ue = V V nd r (V ) d r (2.89)
2 S V
Because V can be any volume that contains all of the charges in the system,
the surface S may be chosen at an arbitrarily large distance from the charge
distribution. Furthermore, because V (r) falls off at least as fast as 1/r as
r , then V (r) falls off at least as fast as 1/r2 as r , and because
the surface area of S increases as r2 in that limit, then the surface integral
appearing in Eq. (2.89) decreases at least as fast as 1/r as r and can
be made arbitrarily small by choosing S sufficiently distant from the source
charge distribution. Because V (r) = E(r), the electrostatic energy is then
given by
1
ZZZ ZZZ
Ue = D(r) E(r)d3 r = |E(r)|2 d3 r (2.90)
2 V 2 V
where the volume V must only be large enough to include all regions where
the electrostatic field E(r) produced by the charge distribution is nonzero; if
this is not satisfied, then the electrostatic energy is given by Eq. (2.89). Notice
that this expression includes the self-energies of all the charges in the system.
The integrand in Eq. (2.90) is defined as the electrostatic energy density
1
ue (r) D(r) E(r) (J/m3 ) (2.91)
2
which is associated with the field energy at each point in space.
2.8.1 Capacitance and Energy of Multi-Conductor

Systems
The total energy of the electrostatic field produced by a system of n charged

ideal conductors embedded in a spatially homogeneous simple dielectric
medium is given by Eq. (2.90) with the integration taken over the entire
region V external to the conductors. Because E = V in V, this expression
may be rewritten as

ZZZ
Ue = E V d3 r
2
Z Z Z

ZZZ
= (V E)d3 r V Ed3 r
2

ZZZ
= (V E)d3 r,
2
the second integral vanishing because E = 0 throughout V. Application
of the divergence theorem transforms the remaining integral into a sum of
surface integrals taken over each surface Sj of the entire system of conduc-
tors which spatially bound the electrostatic field plus an additional integral
over an infinitely remote surface. This latter integral vanishes if the sys-
tem of charged conductors is situated within a finite region of space as the
electrostatic field produced by them then diminishes with sufficient rapidity
(E R2 and V R1 ) as R . With the subscript j denoting the j th
conductor and Vj the constant value of the electrostatic potential on that
conductor, the above expression becomes
n
X
I
Ue = Vj E nda,
2 j=1 Sj
where n here denotes the unit outward normal to the conductor surface,
directed from the conductor body into the field region (notice that this is
opposite to the convention used in the divergence theorem,
H as reflected in the
change of sign in the above equation). Because Qj = Sj E nda, one finally
obtains
n
1X
Ue = Qj Vj (2.92)
2 j=1
which is analogous to the expression (2.86) for the electrostatic potential

energy of a system of point charges.
The charges Qj and potentials Vj of the conductor bodies cannot both
be arbitrarily prescribed and consequently must be related in some fashion.
Because the field equations are linear and homogeneous, it is expected that
these relations should also be linear. In order to prove this6 , consider a fixed
geometrical arrangement of n conductor bodies. First, assume that all of
the conductors are uncharged except for the k th conductor which carries the
nonzero charge Qk . Let the corresponding solution of Laplaces equation in
the region V exterior to the conductor bodies be denoted by V (k) (r) and let
(k)
Vj , j = 1, 2, . . . , n, denote the constant potential value on the j (th) conduc-
tor. Now let the charge on the k th conductor be changed to Qk , where is
a constant. The function V (k) (r) then satisfies Laplaces equation in V. The
electrostatic potential and all of its derivatives are thus everywhere changed
by the factor . In addition, because s = V /n, it then follows that the
charge density on each conductor surface Sj is multiplied by the same factor
, in which case the charge on the k th conductor body becomes Qk while the
other conductors remain uncharged. Because a solution of Laplaces equation
that satisfies a particular set of Dirichlet boundary conditions is unique (see
2.1.5.1), the potential V (k) (r) is then the correct solution of this modified
system, establishing the linearity of the relationship between the charges and
potentials of the conductor bodies. More importantly, however, is the con-
clusion that the potential of each conductor body is directly proportional to
the charge Qk of the k th conductor. Hence
(k)
Vj = pjk Qk , j = 1, 2, . . . , n,
where the coefficients pjk depend only upon the geometry of the conductor
system. By superposition, the electrostatic potential on the j th conductor
Pn (k)
when all of the conductors are charged is given by Vj = k=1 Vj , so that
n
X
Vj = pjk Qk (2.93)
k=1
for j = 1, 2, . . . , n, which is the desired linear relationship between the charges

and potentials on the conductor bodies of the n conductor system. The co-
efficients pjk , which depend only upon the geometry of the multi-conductor
system, are called coefficients of potential, where pjk is the potential of the
j th conductor per unit charge on the k th conductor.
Substitution of this result into Eq. (2.92) for the potential energy gives
6
The proof presented here follows that given in 3.12 and 6.5 of Reitz and Milford
[10].
n n
1 XX
Ue = pjk Qj Qk (2.94)
2 j=1
k=1
Hence, the electrostatic potential energy of a system of charged conductors is

a quadratic function of the charges on the conductors comprising the system.
Theorem 2.6. The coefficients of potential pjk for a multi-conductor system
satisfy the three fundamental properties:
pjk = pkj ,
pjk 0, (2.95)
pjj pjk , k.
Proof. (1). The first property follows from Eq. (2.94). The total differential
of Ue is given by

Ue Ue Ue
dUe = dQ1 + dQ2 + + dQn .
Q1 Q2 Qn
If the only charge in the n-conductor system that is changed is Qj , then

n
Ue 1X
dUe = dQj = (pjk + pkj )Qj dQj .
Qj 2
k=1
The work expended (and hence the change in potential energy) transporting
an element of charge dQj from a zero potential reservoir to the j th conductor
is given by
Xn
dUe = Vj dQj = pjk Qk dQj .
k=1
Comparison
Pn of this expressionPfor dUe with the preceding expression then
n
gives 12 k=1 (pjk + pkj )Qk = k=1 pjk Qk . Because this result must be valid
for all possible values of Qk , one must then have that 12 (pjk + pkj ) = pjk , and
hence pjk = pkj , which proves the first property.
(2). The second property that the coefficients of potential pjk are non-negative
follows from the fact that the potential due to a net positive charge is positive.
(3). In order to prove the third property, let the j th conductor carry a positive
charge Qj while all the remaining conductors in the n conductor system
remain uncharged. Because the k th conductor (k 6= j) is uncharged, then the
net number of electrostatic lines of force leaving this conductor body is zero.
Two distinct cases must then be separately considered:
Case I: There are no lines of force either leaving or impinging upon the
k th conductor, in which case it is in an equipotential region, and hence,
is shielded by another conductor. It is then either contained within the
body of the charged j th conductor, in which case its potential is Vj so that
pjk = pjj , or it is located inside some other conductor 6= j in which case

pjk = pj and attention is then shifted to the th conductor.
Case II: The lines of electric flux leaving the k th conductor are balanced by
the lines of flux impinging upon it. Because the origin of this electric flux
is the charge Qj on the j th conductor, it is then possible to trace a given
flux line that is impinging on the k th conductor back (possibly via other
conductors) to the j th conductor. This means that the j th conductor is at
a higher potential than the k th conductor, so that Vj > Vk when Qj > 0.
From Eq. (2.93), this then implies that pjj > pjk , to which an equality
sign must be added to cover Case I.
This proves the third property, completing the proof of the theorem.

The relation given in Eq. (2.93) is a set of n linear equations expressing

the potentials Vj on each of the n conductors in terms of the charges Qj
residing on them. This system of equations may be inverted to yield a set
of equations giving the charge values Qj on the conductors in terms of their
potentials Vj with the result
n
X
Qj = cjk Vk (2.96)
j=1
for j = 1, 2, . . . , n. The coefficients cjj are the coefficients of capacitance (or

capacity coefficients) whereas the coefficients cjk with j 6= k are the elec-
trostatic induction coefficients. The coefficients of capacitance and induction
form a matrix C = (cjk ), with the coefficients of capacitance forming the
diagonal and the coefficients of inductance the off-diagonal elements, that is
the inverse of the matrix P = (pjk ) of the coefficients of potential pjk , so that
C = P 1 . (2.97)
As a consequence of theorem 2.6, the following equivalent theorem holds.

Theorem 2.7. The coefficients of capacitance and induction cjk for a multi-
conductor system satisfy the three fundamental properties:
cjk = ckj ,
cjj > 0, (2.98)
cjk 0, j 6= k.
Substitution of Eq. (2.96) into Eq. (2.92) for the electrostatic potential
energy yields
n n
1 XX
Ue = cjk Vj Vk (2.99)
2 j=1
k=1
The electrostatic potential energy of a system of charged conductors is thus

seen to be a quadratic function of either the charges or the potentials on the
various conductors comprising the system.
From Eq. (2.96), the capacitance of a conductor is seen to be given by
the total charge on the conductor when it is maintained at unit potential,
all other conductors in the system being held at zero potential. In that case
Qj = cjj Vj and
Qj
cjj = . (2.100)
Vj
A pair of isolated conductors which can store equal and opposite charges
Q, independently of whether other conductors in the system are charged,
form what is known as a capacitor. This independence from the presence of
other charges in the system implies that one of the conductor bodies shields
the other, as illustrated in Fig. 2.10, so that the electrostatic potential con-
tributed to each conductor of the pair by any external charge is the same; if
this isnt the case, then one must consider the entire multi-capacitor system
comprised of the n conductors. If two ideal conductors S1 and S2 form such
a capacitor, application of Eq. (2.93) to that arrangement yields
V1 = p11 (Q) + p12 Q + VE ,

V2 = p21 (Q) + p22 Q + VE ,
when +Q is on S2 and Q is on S1 and where VE denotes the common

potential due to the presence of any external charges. The potential difference
V = V2 V1 between the two conductors is then given by
V = (p11 + p22 2p12 )Q, (2.101)
where the potential reference is taken on the negatively charged conductor in

order to make V non-negative. The difference in electrostatic potential be-
tween the two conductor bodies of a capacitor is then seen to be proportional
to the stored charge Q.7 This result may then be written as V = Q/C,
where
Q
C = (p11 + p22 2p12 )1 (F ) (2.102)
V
is the capacitance of the capacitor. The capacitance is then seen to be the
charge stored per unit of potential difference. The unit of capacitance in
MKSA units is the farad (F ) where F C/V .
+ +

+
Fig. 2.10 A capacitor +
formed by a pair of con- + -- -

- - +
ductors separated by a + -
-
simple dielectric with - +

- -
permittivity with one +
- S1 +
-
conductor (S1 ) being -
+
shielded by the other (S2 ) S2 +
+
from other charges in the
n conductor system.
Consider now a simple capacitor consisting of two (perfectly) conduct-

ing bodies separated in space by a simple dielectric with permittivity and
brought to a static charge state with net charge +Q on body 1 and Q on
body 2. Such a capacitor then possesses the following properties:
Free charges Q reside on the conductor surfaces, resulting in a surface
charge density sj (r) on each body (j = 1, 2 respectively) such that
I
Q = sj (r)d2 r. (2.103)
Sj
Gauss law shows that the E-field lines originate normally from the surface
S1 of the positively charged body and terminate normally on the surface
S2 of the negatively charged body, with
I
D(r) nd2 r = Q (2.104)
S
for any surface S enclosing either S1 (+Q) or S2 (Q).

As a consequence of the perpendicularity of E at each conductor surface,
they are both equipotential surfaces, where V (r) = V1 when r S1 and
V (r) = V2 when r S2 .
A single-valued potential difference V = V1 V2 then exists between the
two conducting bodies, given by
7
By convention, the absolute value of the charge on one of the two conductor bodies
is referred to as the charge on the capacitor.
Z S1 Z S2 Z S1
V = E(r) d + E(r) d = E(r) d. (2.105)
S2
In order to make V positive, the potential reference S2 is taken on the

negatively charged conductor body.
From Eqs. (2.102) and (2.104)(2.105), the capacitance of this simple linear
capacitor is then given by
D(r) nd2 r
H
Q
C= = SR1S1 (2.106)
V E(r) d S2
The voltage difference due to a charge Q on a linear capacitor with capac-

itance C is given by V = Q/C. The amount of work required to trans-
fer an additional incremental charge dQ from S2 to S1 is then given by
dUe = V dQ = C1 QdQ. Beginning with an uncharged capacitor and trans-
fering charge until Q is reached on S1 , the total work expended is given by
Q
1 Q2 1
Z
Ue = QdQ = = C(V )2 , (2.107)
C 0 2C 2
in agreement with Eq. (2.99) for the potential energy of an isolated conductor.
The capacitance of a two-conductor system is then seen to be given in terms
of its electrostatic energy Ue by either of the two equations
2Ue 1
ZZZ
C= = D(r) E(r)d3 r, (2.108)
(V )2 (V )2 V
or
Q2 Q2
C= = RRR , (2.109)
2Ue V
D(r) E(r)d3 r
where the integration domain V contains all of the charged bodies comprising
the multi-conductor capacitor.
The mutual capacitance of this two conductor system can also be expressed
in terms of the capacity and induction coefficients cjk . From Eq. (2.94) the
electrostatic potential energy of this two conductor capacitor is given by
Ue = 21 Q2 (p11 + p22 2p12 ). Because P = C 1 , then
1
p11 p12 c11 c12 1 c22 c12
= = ,
p12 p22 c12 c22 c11 c22 c212 c12 c11
so that
1 2 1
Ue = Q (c11 + c22 + 2c12 ).
2 c11 c22 c212
Comparison of this expression with that given in Eq. (2.107) then yields
c11 c22 c212

C= (2.110)
c11 + c22 + 2c12
for the mutual capacitance of the two conductor system.
Consider now determining the change in the electrostatic energy of a sys-
tem of conductors that is caused by an infinitesimal change in either their
charges or their potentials, as treated by Landau and Lifshitz [11]. Beginning
with Eq. (2.90), the variation of the total electrostatic energy of a system of
n charged ideal conductors is given by
1
ZZZ ZZZ
Ue = (E E)d3 r = E Ed3 r. (2.111)
2
Upon setting E = V and using the fact that E = 0 in the dielectric
so that E = 0, the above result becomes
ZZZ n
X I
Ue = (V E)d3 r = Vj E nd2 r
j=1 Sj
after application of the divergence theorem, where n denotes the outward

unit normal vector to the conductor surface Sj . The variation
H in charge on
the j th conductor is obtained from Eq. (2.39) as Qj = Sj E nd2 r, so
that
Xn
Ue = Vj Qj . (2.112)
j=1
This expression then gives the change in electrostatic energy due to a change
in the charges on the conductors. This is simply the work required to bring
a set of n infinitesimal charges Qj from infinity to the various conductor
bodies in the system in the presence of the (fixed) potential Vj .
The expression (2.112)
RRR for the change in electrostatic energy can also be
E (V )d3 r = (EV )d3 r. Because in-
RRR
written as Ue =
finitesimal changes in the potentials, just like the potentials themselves, are
constant over the surface of each conductor, this expression becomes
n
X I
Ue = Vj E nd2 r
j=1 Sj
after application of the divergence theorem. Each surface integral here is

recognized as the charge Qj = Sj E nd2 r on the corresponding conductor,
H
so that
Xn
Ue = Qj Vj , (2.113)
j=1
which expresses the change in electrostatic energy in terms of the change in

the potentials of the conductor bodies.
The relations given in Eqs. (2.112) and (2.113) show that, by differentiating
the electrostatic energy Ue of a system of charged conductors with respect to
the charges on the conductors, the potentials on the individual conductors
are obtained as
Ue
Vj = , (2.114)
Qj
whereas the derivatives of Ue with respect to the potentials gives the charges
on the conductors as
Ue
Qj = . (2.115)
Vj
The symmetry relation cjk = ckj for the coefficients of capacitance given in
theorem 2.7 then follows from the fact that 2 Ue /Vj Vk = 2 Ue /Vk Vj .
The remaining properties in that theorem follow from the positive-definiteness
of the quadratic form given in Eq. (2.99).
2.8.2 Thomsons Theorem
The energy of the electrostatic field of a system of charged ideal conductor

bodies possesses an extremum property that may be obtained in the follow-
ing manner. Suppose that the charge distribution on the conductor bodies
undergoes an infinitesimal change in such a manner that the total charge on
each conductor remains unchanged; charge must then be allowed to penetrate
into the conductor body. The resultant change in electrostatic energy begins
with Eq. (2.90) which must now be extended over all space, including the
regions occupied by the conductor bodies. One then has that
ZZZ ZZZ
Ue = E Dd3 r = V Dd3 r
Z Z Z ZZZ
= (V D)d3 r V (D)d3 r .
Application of the divergence theorem to the first integral in the above ex-
pression results in a surface integral at infinity and consequently vanishes.
For the second integral, Gauss law (2.75) gives (D) = , so that
ZZZ
Ue = V d3 r,
where describes the infinitesimal change in the charge distribution on the

conductors. Because is zero in the region surrounding the conductors, this
remaining integration is taken only over the conductor bodies, and since V
is constant over each conductor, one obtains
X ZZZ
Ue = Vj d3 r = 0,
j Vj
the remaining integral over the volume Vj of each conductor body being equal
to zero since its total charge remains unaltered.
The energy of the actual electrostatic field is thus an extremum. In order
to determine whether it is a maximum or a minimum, let V , E represent any
other possible electrostatic field. Because the total charge on the conductors
is fixed, one then has that Sj (E E ) nd2 r = 0 for each conductor body
H
j in the system. If Ue and Ue denote the respective electrostatic energies of

the two fields, then their difference is given by
Z Z Z

ZZZ
Ue Ue = (E E) (E E)d3 r + 2 E (E E)d3 r .
2
Because E = 0 and (E E) = 0, then the second integral on the

right-hand side of the above equation vanishes (see problem 2.5), so that

ZZZ
E E2 d3 r,

Ue Ue =
2
which is a non-negative quantity. Hence, the energy Ue is a minimum, for if
E differs from E in any region of space, the resultant potential energy Ue
will be greater than Ue . This then proves the following useful theorem8 .
Theorem 2.8 (Thomsons Theorem). The energy of the actual electro-
static field produced by a system of charged conductors is a minimum relative
to the energies of the fields which could be produced by any other distribution
of the charges on or in the conductors.
As a corollary to Thomsons theorem, one has that:
Corollary 2.2. The introduction of an uncharged conductor into the field
produced by a system of charged conductors (or a fixed set of charges) reduces
the total energy of the electrostatic field.
As a consequence, an uncharged conductor that is in a region remote from
a given system of charged conductor bodies is attracted towards the system.
Earnshaws theorem (see Theorem 2.5 in 2.1.6) also applies, which states
that a charged body placed in an electrostatic field cannot be maintained in
stable equilibrium under the influence of the electrostatic forces alone.
8
It is unclear how this theorem came to be attributed to W. Thomson (Lord Lelvin)
as no reference is made to him in connection with the statement of this theorem in
the historical treatise on The Mathematical Theory of Electricity & Magnetism by
Sir James Jeans [12]. This theorem is not to be confused with the theorem due to J.
J. Thomson who is referenced in a footnote on page 154 of the 1954 Dover reprint
of Maxwells Treatise on Electricity & Magnetism [13] where it is stated (without
proof) that the normal derivative of the magnitude of the electric field intensity at
a point on the surface of a conductor (or any equipotential surface) is related to the
mean curvature ( R1 + R1 ) of the surface at that point, where R1 and R2 are the
1 2
principal radii of curvature at that point, as defined in differential geometry. A proof
has been given by Pappas [14].
2.9 Forces on a Conductor 47
2.9 Forces on a Conductor
The electrostatic force that acts between charged bodies can be obtained
through a consideration of the change in the total electrostatic energy of the
system under a small virtual displacement. As an illustration, consider deter-
mining the force per unit area acting on the surface S of an ideal conductor
carrying a surface charge density s (r). An element of surface charge s (r)da
at a point r S experiences an electrostatic force due to the electrostatic
field of all the other charges in the system. Because the E-field is normal
to the surface of a perfect conductor, this force is perpendicular to the con-
ductor surface S, and because the element of charge s (r)da is bound to the
conductor by internal forces, the force acting on it is directly transferred to
the body of the conductor. In the immediate neighborhood of the conductor
surface the electrostatic energy density is given by [see Eq. (2.91)]
2 1
ue (r) = E(r) = 2s (r). (2.116)
2 2
An infinitesimally small virtual displacement of an elemental area a of
the conductor surface will then result in a decrease in the electrostatic energy
by an amount given by the product of the energy density ue (r) at that point
1 2
and the excluded volume a, so that Ue = 2 s (r)a. Because the
magnitude of the force F(r) is given by Ue /, this result means that there
is an outward force per unit area equal in magnitude to
dF (r) 1
= 2s (r) = ue (r) (2.117)
da 2
at the surface of the conductor.
By Gauss law, the total E-field flux emerging from the surface element da
with surface charge density s (r) is given by 1 s (r)da, half of it directed into
the body and half directed out of the body of the conductor. The electrostatic
field intensity due to the local surface charge density alone is then comprised
1
of two parts directed along n with equal magnitude 2 s (r). Because the
electric field intensity at an exterior point infinitesimally close to the surface
of the conductor is given by E(r) = ns (r)/ [cf. Eq. (2.38)] it is then seen
that the element of surface charge s (r)da at that point produces exactly
half of the total field external to that point. The electric field intensity acting
on the element of surface charge s (r)da due to all of the other charges in
the system is then given by E0 (r) = ns (r)/2. The force dF(r) acting on
the element of surface charge s (r)da, and consequently acting on the surface
element da of the conductor itself, is then given by dF = (s da)E0 , so that
dF(r) 1
= 2s (r)n = ue (r)n, (2.118)
da 2
in agreement with the expression given in Eq. (2.117), where n is the outward
unit normal vector to the conductor surface at the point r S. Hence, the
electrostatic force on a conductor tends to pull the conductor into the field;
that is, an electrostatic field exerts a negative pressure on a conductor with
magnitude equal to the energy density in the field at that point. The total force
acting on a conductor body is then obtained by integrating the expression
for the force per unit area given in in Eq. (2.118) over the entire surface S of
the conductor body as
1 1
I I
2
F= (r)nda = D(r) E(r)nda. (2.119)
2 S s 2 S
2.9 Problems 49
Problems
2.1. Determine the general solution for the vector X to each of the equations
AX= k & A X = K,
where the scalar k and vectors A and K are known constants. In addition,
determine the solution when the vector X satisfies both of these equations.
2.2. Prove that the specification of both the divergence and curl of a well-
behaved vector field F(r) throughout a finite, regular region V of space, in
addition to the specification of the tangential component n F of the vector
field over the closed boundary surface S of V results in a unique determination
of F(r) in V .
2.3. Derive the symmetry properties for the Greens function GD (r , r ) with
Dirichlet boundary conditions given in Eq. (C.7) and for the modified Greens
function GN (r , r ) with Neumann boundary conditions given in Eqs. (C.10)
(C.11) of Appendix C.
2.4. Prove Greens reciprocation theorem: If (r) is the electrostatic potential
due to a volume charge density (r) within a region R and a surface charge
density s (r) on the surface S bounding the region R, and if (r) is the
electrostatic potential due to another volume charge density (r) in R and
surface charge density s (r) on S, then
ZZZ I ZZZ I
(r) (r)d3 r + s (r) (r)d2 r = (r)(r)d3 r + s (r)(r)d2 r.
R S R S
2.5. Use Greens reciprocation theorem to derive Eq. (C.16) in Appendix C.
2.6. Prove the following theorem due to Stratton [15]: The integral over all
space of the scalar product of an irrotational vector field F(r) and a solenoidal
vector field G(r) is zero, viz.
ZZZ
F(r) G(r) = 0,
provided that F(r) and G(r) and their first partial derivatives are continuous
everywhere except on a finite number of closed, regular surfaces Sj across
which their discontinuities satisfy

n F+ (r) F (r) = 0, r Sj ,

n G+ (r) G (r) = 0, r Sj ,
and provided that their limiting behavior at infinity satisfies
lim r|F(r)| = 0,
r
lim r|G(r)| = 0,
r
where r = |r|.
2.7. Prove the corollary 2.2 to Thomsons theorem.

2.8. Show that r 8
= (r).
r3 3
2.9. Show that r M n M
(M ) = 3 3 3 n.
r3 r r
2.10. Beginning with the divergence theorem ( S F nd2 r = V Fd3 r),
H R
where the closed surface S forms the copmplete boundary of the region V
with outward unit normal vector n, show that, for any sufficiently continuous
vector field G = G(r) and scalar field f = f (r):
I Z
f (r)nd2 r = (f (r)) d3 r, (2.120)
S V
I Z
n G(r)d2 r = ( G(r))d3 r. (2.121)
S V
2.11. Beginning with Stokes theorem ( C F dr = S ( F) nd2 r), where

H R
the closed contour C forms the complete boundary of the surface S with unit
normal vector n taken in the positive direction, show that, for any sufficiently
continuous vector field G = G(r) and scalar field f = f (r):
I Z
dr G(r) = (n ) G(r)d2 r, (2.122)
C S
I Z
f (r)dr = (n f (r))d2 r. (2.123)
C S
References
1. R. S. Elliott, Electromagnetics: History, Theory, and Applications. Piscataway,

NJ: IEEE, 1993.
2. S. D. Poisson, Remarks on an equation which occurs in the theory of attractions
of spheroids, Bull. de la Soc. Philomathique, vol. 3, pp. 388392, 1813.
3. P. S. Laplace, Theory of attractions of spheroids, Mem. de lAcademie Royale,
pp. 113196, 1785.
4. O. Holder, Beitrage zur Potentialtheorie. PhD thesis, Polytechnikum (Stuttgart),
1882.
5. O. D. Kellogg, Foundations of Potential Theory. Berlin: Springer, 1929.
6. W. R. Smythe, Static and Dynamic Electricity. New York: Hemisphere Publish-
ing, third, revised printing ed., 1989.
7. P. Lorrain and D. R. Corson, Electromagnetic Fields and Waves. New York: W.
H. Freeman and Co., second ed., 1970.
8. C. Kittel, Introduction to Solid State Physics. New York: John Wiley & Sons,
fourth ed., 1971. Ch. 13.
9. C. J. F. Bottcher, Theory of Electric Polarization: Volume I. Dielectrics in Static
Fields. Amsterdam: Elsevier, second ed., 1973.
References 51
10. J. R. Reitz and F. J. Milford, Foundations of Electromagnetic Theory. Reading,

MA: Addison-Wesley, second ed., 1967.
11. L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media. Oxford:
Pergamon, 1960. Ch. IX.
12. J. Jeans, The Mathematical Theory of Electricity and Magnetism. Cambridge:
Cambridge University Press, 1908.
13. J. C. Maxwell, A Treatise on Electricity and Magnetism. Oxford: Oxford Uni-
versity Press, 1873.
14. R. C. Pappas, Differential-geometric solution of a problem in electrostatics,
SIAM Review, vol. 28, no. 2, pp. 225227, 1986.
15. J. A. Stratton, Electromagnetic Theory. New York: McGraw-Hill, 1941.
Chapter 3
Electric Current
Thunder is good. Thunder is impressive. But its the lightning that does the
work. Samuel Clemens
The transport of electric charge constitutes electric current and the physical
process by which this occurs is either conductive or convective. Conduction
current is due to the relative drift motion of charge carriers through a ma-
terial medium, whereas convection current is due to the mass transport of
the charged medium itself such as occurs in atmospheric electricity as well
as in the motion of charged particles in vacuum (e.g., electron flow in a vac-
uum tube). In electric conduction, the entire material body typically remains
electrostatically neutral, whereas in electric convection the system is not elec-
trostatically neutral so that its electrostatic charge must be accounted for.
In a metal, conduction current is entirely due to valence electrons, the
more massive positive ions held fixed in the crystal lattice. In an electrolyte,
conduction current is comprised of both positive and negative ions traveling
in opposite directions, thereby contributing to current flow in the same di-
rection. In a gas discharge, conduction current is comprised of both electrons
and positive ions, but because the positive ions are much heavier and thus
less mobile than the electrons, the current is primarily due to the electrons.
In a semiconductor, conduction current is carried by both valence electrons
and so-called positively-charged holes in the materials electronic structure.
Electrons that are excited into the conduction band of the semiconductor
material leave unoccupied states, or holes, in the valence band, both of which
contribute to the electrical conductivity. Although the holes dont actually
move, an electron from a neighboring atom can move to fill the hole, leaving
behind a hole at the original position, and in this way the holes appear to
move, behaving as if they were positively charged particles.
By convention, the directional sense of electric current is taken in the
direction that positive charge moves. Because positive charge carriers move
in the direction of an externally applied electric field and negative charge
carriers in the opposite direction (from Coulombs law F = qE), the net
current is in the direction of the applied field.
53
54 3 Electric Current
3.1 Conservation of Charge
Consider a conducting medium that is comprised of a single type of charge

carrier with charge q and number density N (the number of carriers per
unit volume). The spatial average of their velocities under the application of
an external electric field results in a drift velocity v, their random thermal
motions averaging to zero. In the time interval t each carrier moves the
average distance vt so that the total charge Q crossing the directed area
nda in this time interval is given by qN v ndat, resulting in the current
dI Q/t = N qv nda. For more than one type of charge carrier involved,
" #
X
dI = Ni qi vi nda, (3.1)
i
the summation extending over different carrier types. The quantity appearing
in the brackets of this expression is then identified as the current density
X
J(r) Ni qi vi (A/m2 ) (3.2)
i
With this definition, Eq. (2.120) becomes dI = J nda and the total current
flowing through a regular surface S is given by
Z
I= J nda (3.3)
S
in the sense specified by the unit normal vector n to S.

For a stationary regular region V with simple closed surface S, the electric
current entering V is given by
I Z
I = J nda = Jd3 r (3.4)
S V
after application of the divergence theorem, where n is the outward unit

normal to S. Because I = dQ/dt is the rate at which charge is transported
into the region V, then
d (r, t) 3
Z Z
I= (r, t)d3 r = d r (3.5)
dt V V t
after application of Leibnitzs theorem, where (r, t) describes the macro-

scopic charge density in V. Taken together, Eqs. (2.123) and (2.124) express
the fundamental physical law of conservation of charge, which then requires
that Z
(r, t)
J(r) + d3 r = 0.
V t
3.2 Electric Conductivity 55
Because this result is valid for arbitrary regular regions V, the integrand itself
must vanish everywhere, so that
(r, t)
J(r) + =0 (3.6)
t
which is the equation of continuity relating the charge and current densi-
ties. Under steady-state conditions, the time derivative of the charge density
vanishes so that J(r) = 0 and the conduction current density is solenoidal.
3.2 Electric Conductivity
The action of an applied electrostatic field E(r) in a conducting medium is

to accelerate both electrons and holes (if present) through the Coulombic
force F(r) = qE(r). Due to collisions with the molecules of the material, this
motion is reduced to an average drift velocity which is proportional to the
applied field strength E. The average electron drift velocity is then given by
ve = e E, where e is the electron mobility (in m2 /V s), and the average
hole drift velocity is given by vh = +h E, where h is the hole mobility.
With Ne denoting the free electron and Nh the free hole number densities,
the electron and hole charge densities are then given by e = Ne qe and
h = +Nh qe , respectively, where qe 1.6 1019 C is the magnitude of the
electronic charge. The conduction current density is then given by
Jc (r) = Je (r) + Jh (r) = e ve + h vh

= e e + h h E(r). (3.7)
The static electric conductivity 0 is then defined through the point form of
Ohms law
Jc (r) = 0 E(r), (3.8)
where
0 = e e + h h = Ne e + Nh h qe (3.9)
1
in mhos/m (/m where A/V ). For a good conductor, Nh h Ne e
and 0 e e = Ne e qe . Two special limiting cases are to be noted: (1) for
a perfect dielectric, 0 = 0 in which case Jc = 0 regardless of the strength of
E, and (2) for a perfect conductor, 0 = so that E = Jc /0 = 0 regardless
of the strength of the conduction current density Jc .
Boundary conditions on the conduction current density Jc across an in-
terface S separating two conductors with static conductivities 1 and 2 are
directly obtained from the boundary conditions on the electric field intensity
1
The conductivity is also expressed in Siemens/m (S/m) after Ernst Werner von
Siemens (18161892).
given in Eqs. (2.83)(2.84) through the static form of Ohms law Jc = 0 E.

For the normal boundary conditions, Eq. (2.83) [D1n D2n = s ] directly
yields
1 2
J1n J2n = s . (3.10)
1 2
However, because the conduction current density is solenoidal when steady-
state conditions areHsatisfied, a straightforward application of the integral
form of Gauss law Jc nda = 0 to a simple closed surface with identical
faces on opposite sides of the interface surface S yields
J1n J2n = 0. (3.11)
Combination of this result with that given in Eq. (2.129) results in the re-
quirement
1 2
s = Jcn (3.12)
1 2
relating the surface charge density to the normal component of the conduction
current density at the interface, where Jcn = J1n = J2n . For the tangential
boundary conditions, Eq. (2.84) [E1t = E2t ] gives
J1t J2t
= . (3.13)
1 2
3.2.1 Resistance and Conductance
Consider a length of a conductor with static conductivity 0 and with a

uniform cross-section surface S. Application of a voltage difference
Z S2
V = V1 V2 = E d. (3.14)
S1
between the cross-sectional surfaces S1 and S2 of the conductor results in an

electrostatic field E and a conduction current density Jc = 0 E. The total
current I flowing through that section of the conductor is then given by
ZZ ZZ
I= Jc ds = 0 E ds. (3.15)
S S
The resistance of the conductor is then defined as

R
V C E d
R = RR () (3.16)
S 0 E ds
I
and the conductance is

3.3 Joules Law 57
RR
1 0 E ds
G = SR ( = S) (3.17)
R C E d
For example, for a uniform wire of cross-sectional area A and length , V = E

and I = 0 EA so that R = VI = 0A .
From Eqs. (2.106) and (2.135) as applied to a pair of conductors embedded
in a material with uniform static dielectric permittivity s and conductivity
H RS
0 , the capacitance and resistance are given by C = s S1 E ds/ S21 E d
RS H
and R = S21 E d/0 S1 E ds, respectively, whose product is then
s
RC = (3.18)
0
3.3 Joules Law
Consider a differential volume element V containing the electron charge

qe = e V and hole charge qh = h V. Under the action of an applied
electrostatic field E, these charges experience the forces Fe = qe E = e EV
and Fh = qh E = h EV, respectively. The work expended in moving qe the
differential distance e and qh the differential distance h is then given by
W = Fe e + Fh h with associated power expended P W/t
given by P = Fe e /t+Fh h /t. With ue e /dt identified as the
electron drift velocity and uh h /dt identified as the hole drift velocity,
expended power becomes P = Fe ue +Fh uh so that
this expression for the
P = e ue + h uh E V = Jc E V, where Jc denotes the conduction
current density [see Eq. (2.126)]. The total dissipated power in a volume V is
then given by Joules Law
ZZZ
P = Jc (r) E(r) d3 r (W ) (3.19)
V
With the point form of Ohms Law given in Eq.(2.129), Joules law becomes
ZZZ
P = 0 (r)|E(r)|2 d3 r. (3.20)
V
As a simple example, for a uniform wire with cross-sectional area A and

length and with constant static conductivity 0 , Joules law yields P =
RR 2
R
A 0 |E|d r 0 |E| d, where the first integral yields the current I = 0 |E|A
flowing through the wire and the second integral the voltage drop V = |E|
across the length of the wire, their product then yielding the familiar result
P = IV = I 2 R.
Problems
3.1. Consider a temporary imbalance of charge existing in a region inside

of a conductor with dielectric permittivity and conductivity such that a
net charge density 0 exists at time t = 0. Derive the first-order differential
equation for this charge density (t) and determine its solution for all t 0
in terms of its time constant /.
3.2. Compare the time constants for copper, silver, gold and platinum.
3.3. Assuming a uniform current distribution, determine the resistance per

unit length of a conductor with uniform cross-section with cross-sectional
area A and conductivity .
3.4. Determine the resistance per unit length of the dielectric insulation in
a coaxial cable with inner conductor radius a and outer conductor radius b,
dielectric permittivity and conductivity .
References

Chapter 4
The Magnetic Field
Ampere was the Newton of electricity. J. C. Maxwell
The magnetic field produced by steady-state current flow is now considered

based upon the Biot and Savart law [Biot and Savart (1820); Ampere (1820
1825)]. An historical development of this theory is given by Elliott [1].
4.1 The Biot and Savart Law
If d is a differential element of length that is oriented along the direction of

flow of a filament of current I and if r is the position vector from the current
element Id to a fixed observation point P , as illustrated in Fig. 2.11, then
the differential element of magnetic flux density dB produced at P is given
by the Biot and Savart law
0 d r
dB = I (4.1)
4 r3
in MKSA units, where the magnetic flux density is in webers per square meter
(also known as the tesla T W b/m2 = N/C), where 0 = 4 107 W b/(A
m) is the magnetic permeability of free space, the weber (W b = V s = J/A)
denoting the unit of magnetic flux.
As an illustration, consider determining the magnetic induction field B of
an infinitely long straight filament of current I. With the z-axis of a polar
cylindrical coordinate system (R, , z) chosen along the filament in the di-
rection of positive current flow, d r = (1z d) (1z z + 1R R) = 1 Rd, so
that B = 1 B with magnitude
Z
0 R 0 I
B = I d = . (4.2)
4 (R2 + 2 )3/2 2R
This is the experimental result first established by Biot and Savart.
59
60 4 The Magnetic Field
Fig. 4.1 Geometric re-

lationship between a dif-
ferential element Id of
current filament and the
differential element dB of dl dB
magnetic flux density it r
produces at the oriented
distance r from the dif- P
ferential element of path
length d, as expressed by
the Biot and Savart law.
4.2 Amperes Law
Amperes experiments were concerned with the force that one current carry-
ing wire experiences in the presence of another. The differential element of
force dF experienced by a current element I1 d1 due to its interaction with
the magnetic induction field B is given by the Lorentz force relation
dF = I1 d1 B. (4.3)
The total magnetic force on a curved filament of current I extending from

P1 to P2 in a uniform magnetic field is then given by Fm = I B where
RP
= P12 d is the distance vector from P1 to P2 ; if the filament forms a closed
loop, then the total magnetic force vanishes.
If the magnetic field B(r12 ) is due to a closed circuit C2 carrying current
I2 , as illustrated in Fig. 2.12, then the total force F12 that the closed circuit
C1 with current I1 experiences is obtained from Eqs. (2.140) and (2.142) as
C1
dl2
Fig. 4.2 Geometric re- C2
lationship between a dif-
ferential element I1 d1 r12
of current I1 flowing in
the circuit C1 and the dl1
differential element I2 d2 I1
of current I2 flowing in
the circuit C2 separated I2
by the vector distance r12
for Amperes law.
4.2 Amperes Law 61
0 d1 (d2 r12 )
I I
F12 = I1 I2 3 . (4.4)
4 C1 C2 r12
Because d1 (d3
r12
2 r12 )
= dr13r12 d2 dr13d2 r12 , where the first term on the
12 12
right is a perfect differential in 1 and so integrates to zero over the closed
contour C1 , one obtains Amperes law for the force between two current loops
0 r12
I I
F12 = I1 I2 3 d1 d2 (4.5)
4 C1 C2 r12
which exhibits the inherent symmetry of the problem. Just as in Coulombs

law, this expression of Amperes law embodies the mechanistic concept of
action-at-a-distance that is absent in the field description.
As an elementary application of Amperes law, consider two infinitely long,
parallel, straight wires a distance d apart, carrying currents I1 and I2 . Each
of these wires then experiences a force per unit length that is directed per-
pendicularly towards the other wire with magnitude
Z
0 d 0 I1 I2
f= I1 I2 3/2
d = .
4 (2 + d2 ) 2 d
This force is attractive if the two currents flow in the same direction and
repulsive if they are oppositely directed.
As another example of the magnetic force, consider comparing the electric
and magnetic forces between two moving point charges: charge q with velocity
v and charge q1 with velocity v1 . From the Lorentz force relation, the mag-
netic force on q is given by Fm = qvB where B = (0 /4)(q1 /r2 )v1 (r/r)
is the magnetic field produced by q1 . Because 0 = 1/(0 c2 ), this magnetic
force may be written as
v1 E
B= ,
c c
where E = (1/40 )(q1 /r2 )(r/r) is the electric field produced by q1 with force
Fe = qE exerted on q. In addition, the magnetic force exerted on q, given by
1 qq1 v v1 r

Fm = 4 0 r
2 c c r , may be written as
v v1
Fm = Fe .
c c
This then shows that Fm Fe for non-relativistic velocities v c, so that
the magnetic force can usually be neglected in comparison to the electric force
except in exceptional circumstances. One such exceptional circumstance is
that of a conduction current where both positive and negative charges are
present with equal densities. The macroscopic electric field is then zero while
the magnetic field produced by the moving charges is not.
Finally, note that two measurements of magnetic force Fm (r) = qv B(r)
at a fixed point r are sufficient to uniquely determine the magnetic induction
field vector B(r) at that point (see Problem 2.12).
4.3 The Differential Relations of Magnetostatics
The fundamental differential relation for the magnetic induction field given
in Eq. (2.140) can be expressed in terms of the current density J(r) as
0 (r r ) 3
ZZZ
B(r) = J(r ) d r, (4.6)
4 |r r |3
the integration taken over the region containing the current density. Because
(r r )/|r r |3 = |r r |1 , this expression may be rewritten as
0 1
ZZZ
B(r) = J(r ) d3 r
4 |r r |
0 J(r ) 3
ZZZ
= d r, (4.7)
4 |r r |
so that
B(r) = 0 (4.8)
and the B-field is solenoidal. This is the differential form of Gauss law for
the magnetic induction field.
For the curl of the magnetic induction filed vector, Eq. (2.145) yields
0 J(r ) 3
ZZZ
B(r) = d r
4 |r r |
ZZZ
0 1 1
ZZZ
3 2 3
= J(r ) d r J(r ) d r .
4 |r r | |r r |
Because (|r r |1 ) = (|r r |1 ) and 2 (|r r |1 ) = 4(r r ),

the above result becomes
0 1
ZZZ
B(r) = J(r ) d3 r + 0 J(r).
4 |r r |
Integration by parts with the surface term at infinity vanishing results in
0 J(r ) 3
ZZZ
B(r) = 0 J(r) + d r. (4.9)
4 |r r |
Because J(r) = 0 for steady-state magnetic phenomena [see the discussion

following Eq. (2.125)], one obtains the differential form of Amperes law
B(r) = 0 J(r) (4.10)
Integration of Eq. (2.147) over a simply-connected region of space bounded

by a closed surface S yields, after application of the divergence theorem with
n denoting the outward unit normal vector to the surface S,
4.4 The Vector Potential for the Magnetostatic Field 63
I
B nd2 r = 0 (4.11)
S
which is the integral form of Gauss law for the magnetic induction field.
Application of Stokes theorem to the integral of the normal component of
the curl of B given by Eq. (2.149) over an open surface So that is bounded
by a simple closed contour C yields the integral form of Amperes law
I
B d = 0 I (4.12)
C
J nd2 r is the current passing through the closed contour C.

R
where I = So
4.4 The Vector Potential for the Magnetostatic Field
Because B = 0 everywhere, then the magnetic induction field vector B(r)

may always be expressed as the curl of a vector potential A(r) as
B(r) = A(r). (4.13)
From Eq. (2.146) it is seen that the general form of the vector potential is
0 J(r ) 3
ZZZ
A(r) = d r + (r). (4.14)
4 |r r |
The added gradient of an arbitrary scalar function (r) shows that the vector
potential can be transformed as A(r) A (r) = A(r) + (r) without
effecting the magnetic field vector B(r). This is the gauge transformation for
the magnetostatic vector potential.1
Substitution of Eq. (2.152) into the differential form (2.149) of Amperes
law results in
( A) 2 A = 0 J.
The freedom of choice implied by the gauge transformation for the vector
potential allows one to set
A = 0, (4.15)
so that A(r) is solenoidal. This choice is known as the Coulomb gauge. The
vector potential in the Coulomb gauge then satisfies Poissons equation
2 A(r) = 0 J(r) (4.16)
1
An analogous gauge transformation for the scalar potential of the electrostatic field
did not arise because it is just a trivial constant potential.
The choice of gauge given in Eq. (2.154) implies that the gauge function
satisfies Laplaces equation 2 = 0 once one is in the Coulomb gauge, in
which case is a constant provided that there are no sources at infinity. The
expression (2.153) for the vector potential then becomes
0 J(r ) 3
ZZZ
A(r) = d r (W b/m) (4.17)
4 |r r |
Comparison of this expression with that given in Eq. (2.27) for the elec-
trostatic scalar potential reveals their similarity in mathematical form, the
principal difference being the scalar and vector forms of the two potentials.
4.5 Magnetic Field of a Spatially Localized Current

Distribution
Consider now the magnetic field structure due to a general steady-state cur-
rent distribution described by the current density vector J(r) that is localized
in a small region of space with vector potential given by Eq. (2.156). By
choosing the origin of coordinates O in close proximity to the localized current
distribution region , as depicted in Fig. 2.9, the denominator in the inte-
grand of Eq. (2.156) may then be expanded in inverse powers of the distance r
to the field point P as |rr |1 = 1/r+(rr )/r3 +[3(rr )2 r2 r2 ]/2r5 + .
The vector potential then has the expansion
0 0
ZZZ ZZZ
A(r) = J(r )d3 r + 3
J(r )(r r )d3 r
4r 4r
0
ZZZ
J(r ) 3(r r ) r2 r2 d3 r + , (4.18)
2

+ 5
8r
where the integration may be extended to all of space because J vanishes

outside of .
By the divergence theorem, if each of the functions f (r), g(r), and J(r) is
continuous throughout all space, then
ZZZ I
f (r )g(r )J(r ) d3 r = f (r )g(r )J(r ) nd2 r 0

as the surface recedes to infinity, so that [because J(r ) = 0]

ZZZ

f (r )J(r ) g(r ) + g(r )J(r ) f (r ) d3 r = 0.

(4.19)
With f = 1 and g(r ) = rj , this general result gives

4.5 Magnetic Field of a Spatially Localized Current Distribution 65
ZZZ
J(r )d3 r = 0, (4.20)
so that the first term (the monopole term) in the expansion (2.157) of the
vector potential vanishes.
For the second term in this expansion, let f (r ) = ri and g(r ) = rj , in
RRR
ri Ji (r ) + rj Ji (r ) d3 r = 0, so that

which case Eq. (2.158) becomes
ZZZ 3
X ZZZ
r r Ji (r )d3 r = rj rj Ji (r )d3 r
j=1
3
1X
ZZZ
ri Jj (r ) rj Ji (r ) d3 r

= rj
2 j=1
3 3
1 XX
ZZZ
= ijk rj (r J(r ))k d3 r ,
2 j=1
k=1
where ijk is the permutation symbol defined to be 0 if two or more indices

are equal, to be +1 if ijk is a permutation of the indices 123 (or xyz) and to
be 1 if it is not. The final result is then given by
1
ZZZ ZZZ
3
r J(r ) d3 r .

r r J(r )d r = r (4.21)
2
The magnetic moment density or magnetization of the localized current dis-
tribution is defined as (in A/m)
1
M(r) r J(r), (4.22)
2
and its integral is defined as the magnetic moment (in A m2 )
1
ZZZ
m r J(r )d3 r . (4.23)
2
Notice that, if the current is confined to a plane Hloop with arbitrary shape,
then its magnetic moment is given by m = 12 I C r d, and since da =
1
2 r d is the oriented differential element of area swept out by d, then the
magnitude of the magnetic moment is given by the current times the area of
the loop, regardless of the shape of the circuit C.
For the third term in the expansion given in Eq. (2.157), one has that
1
ZZZ
J(r ) 3(r r )2 r2 r2 d3 r

r 5

3 3
1 X X xi xj
ZZZ
= (3xi xj ij r2 )J(r )d3 r .
2 i=1 j=1 r5
The quadrupole moment tensor of the localized current distribution is then

defined as (in A m3 )
ZZZ
Qij (3xi xj ij r2 )J(r )d3 r . (4.24)

With these results, the expansion (2.157) of the vector potential produced
by a spatially localized current distribution becomes

3 3
0 m r 1 X X xi xj
A(r) = + Q ij r5 + , (4.25)
4 r3 2 i=1 j=1

the sum of the first two terms of which is O(r3 ), the sum of the remaining
terms being O(r4 ) as r .
The first nonvanishing term in the expansion given in Eq. (2.164) is the
magnetic dipole vector potential
0 m r
A(r) = . (4.26)
4 r3
The associated magnetic induction vector B(r) = A(r) is then given by
0 r
B(r) = m 3
4 h r
0 r r i
= m 3 (m ) 3 .
4 r r
Because rr3 = 8 r m nm

3 (r) and (m ) r 3 = r 3 3 r 3 n, where n r/r is
a unit vector in the direction of the position vector r (see problems 2.72.8),
then the above expression for the magnetic induction field vector produced
by a magnetic dipole with moment m becomes
0 (3n m)n m
B(r) = , (4.27)
4 r3
where the delta function term is not included because r is bounded away
from zero. The magnetic dipole field then has a component along the dipole
axis as well as along the direction of observation from the origin O. With the
dipole axis taken along the polar axis of a spherical coordinate system r, ,
with origin O at the center of a circular current loop, the above expression
becomes
0 m
B(r) = 3
1r 2 cos + 1 sin , (4.28)
4r
where m = |m| is the magnitude of the magnetic dipole moment. Comparison
of this result with that given in Eq. (2.45) for the electric dipole field with
dipole axis oriented along the polar axis reveals that they have the same
spatial behavior in the far-zone r a, where a is a measure of the spatial
extent of the dipole source (e.g., the linear separation of the charges for the
electric dipole and the radius of the current loop for the magnetic dipole).
4.5 Magnetic Field of a Spatially Localized Current Distribution 67
Finally, notice that the magnetic dipole field given in Eq. (2.167) with r
bounded away from zero can be expressed as the negative gradient of a scalar
function as
B(r) = m (r) (4.29)
with
0 m r
m (r) = (4.30)
4 r3
viewed as a scalar potential for the magnetic dipole field.
4.5.1 A Scalar Potential for the Magnetic Field in

Source-Free Regions
In general, a scalar potential for the magnetostatic field can be introduced

wherever the current density vanishes. Amperes law (2.149) becomes
B(r) = 0 = B(r) = m (r). (4.31)
With Gauss law (2.147), this scalar potential then satisfies Laplaces equation
2 m (r) = 0 (4.32)
everywhere that J(r) = 0. The expression in Eq. (2.170) may be integrated

to yield an expression for this scalar potential as
Z r
m (r) = B() d (4.33)
Pref
where Pref is some fixed reference point. It is important that the path of
integration taken in Eq. (2.172) not pass through any region where J 6= 0.
This scalar potential is then ideally suited for obtaining the multipole ex-
pansion of a localized current distribution [2] with the reference point Pref
chosen at infinity and the origin O chosen near to the localized current dis-
tribution. As before, let J vanish outside a sphere of radius rmin centered at
O. With substitution from Eq. (2.145), the scalar potential for the magnetic
field is then given by (with = 1r d = 1r d)
1r J(r ) (1r r )
Z r
0
ZZZ
3
m (r) = d r d .
4 1r r 3

Because 1r J (1r r r ) = 1r (r J), the above result becomes

r
0 d
ZZZ Z
d3 r 1r r J(r )

m (r) =
4 1r r 3

!
r
0 d 1
Z ZZZ
d3 r r J(r )

= ,
4 1r r
which may be integrated by parts over the primed variables to yield
0 r d
ZZZ Z
m (r) = d3 r r J(r ) . (4.34)
4 r r
1
The expansion of the Greens function |r r |1 given in Eq. (C.101) of

Appendix C may then be substituted in the integrand of the above expression
with r< = r and r> = and the integral over evaluated term by term with
the result
X
X 1 Mm
m (r) = 0 Ym (, ), (4.35)

2 + 1 r+1
=1 m =
where
1
Z

3
Mm Ym ( , ) r J(r ) r d r. (4.36)
+1
Notice that the = 0 term is absent in the expansion given in Eq. (2.174) as
it results in a nonphysical magnetic monopole contribution.
4.6 The Concept of a Perfect Magnetic Material
Magnetic material properties have their origin in the motion of bound atomic
electrons together with their intrinsic quantum mechanical spin. In most ma-
terials, the electrons orbit the positively charged nucleus within specific en-
ergy bands. These energy bands are subdivided into orbitals which are typ-
ically evenly filled with electrons that pair together such that the paired or-
bits are orientationally alike but oppositely directed. In that case, the paired
magnetic dipoles are oppositely directed and consequently cancel each other,
resulting in a material exhibiting negligible magnetic properties. In those few
materials that do exhibit macroscopic magnetic properties, the inner electron
orbitals are not evenly filled so that those electrons are not completely paired.
The resultant interaction between adjacent microscopic magnetic dipoles re-
sults in a coupling that tends to align the magnetic moments of the electrons
involved, resulting in a material that exhibits magnetic properties. If the
magnetic moments tend to align themselves under the influence of an exter-
nal magnetic field such that they tend to weaken the field, then the material
is said to be diamagnetic. If they tend to align themselves with the external
4.6 The Concept of a Perfect Magnetic Material 69
magnetic field such that they tend to strengthen the field, then the material
is said to be paramagnetic.
In a ferromagnetic material, the microscopic (atomic or molecular) mag-
netic moments are highly aligned in the absence of an applied magnetic field.
Because it is energetically favored, a ferromagnetic material is broken up into
randomly oriented ferromagnetic domains where each domain is fully magne-
tized. The macroscopic magnetization properties of ferromagnetic materials
is then a result of changes in the domain structure due to the application
of a magnetic field, either through domain wall movement or through rota-
tion of the domain magnetization, depending upon the applied magnetic field
strength. Nearly reversible domain wall motion is the dominant effect in a
weak applied field, but as the applied field strength increases, this process
becomes increasingly irreversible. Finally, at a sufficiently large applied mag-
netic field strength the process becomes dominated by domain rotation. Once
the process becomes irreversible, the substance remains magnetized when the
external magnetic field is removed.
Electron spin alignment also plays a role in determining the magnetization
properties of a material. If the spin alignment of neighboring atoms is parallel,
then the material is ferromagnetic. If the favored spin alignment results in
a net zero macroscopic magnetic moment, then the material is said to be
antiferromagnetic. Finally, if the spin structure is comprised of both spin-up
and spin-down components that results in a nonzero macroscopic magnetic
moment, then the material is said to be ferrimagnetic.
In a perfect magnetic material, all of the electronic motion is confined to the
atomic structure in the form of atomic currents and their moments are ran-
domly oriented in the absence of an external applied magnetic field. They are
then either paramagnetic or diamagnetic materials. For a simple magnetic or
magnetizable material, the dominant multipole moment of the atomic current
distribution is the magnetic dipole, all higher-order moments being negligi-
ble by comparison. Let mj denote the microscopic magnetic dipole moment
of the jth-type of atom comprising the material. The macroscopic magnetic
moment density or magnetization is then given by the spatial average of the
microscopic magnetic dipole moment per unit volume as
X
M(r) = Nj hmj (r)i, (4.37)
j
where hmj (r)i is the spatial average of this atomic dipole moment taken
over a macroscopically small but microscopically large region centered at the
point r, and where Nj is the average number density of j-type atoms in that
microscopic region.
The macroscopic magnetic induction field vector B(r) for the steady-state
magnetic field is defined as the spatial average of the microscopic magnetic
induction field vector b(r) as
B(r) hb(r)i. (4.38)
There is then no distinction between the microscopic and macroscopic field

quantities in vacuum. When the spatial-averaging procedure is applied to
Gauss law (2.147) for the microscopic magnetic field, the same equation
B(r) = 0 (4.39)
is found to hold for the macroscopic field vector. This result then implies
that the macroscopic magnetic field may likewise be expressed in terms of a
macroscopic vector potential A(r) as
B(r) = A(r). (4.40)
The spatial average of Amperes law (2.149) for the microscopic magnetic
field yields
B = 0 hj(r)i, (4.41)
so that the curl of the macroscopic magnetic induction vector is determined
by the spatial average of the microscopic current density in the magnetic
material.
Just as for the spatial average of the microscopic charge density in a per-
fect dielectric, the spatial average of the microscopic current density in a per-
fect magnetic material deserves careful attention. Fortunately, the method
of analysis for the present problem closely parallels that for a perfect dielec-
tric, as given in 2.1.9. In particular, the vector potential A(r) and magnetic
induction field B(r) at points either interior or exterior to (but not on the
surface of) the body of a perfect magnetic material are found to be given by
(see Ch. 9 of Lorrain and Corson [3])
M(r ) n 2 M(r ) 3
I
0
ZZZ
A(r) = d r + d r ,
4 S r V r
"I #
M(r ) n R 2 M(r ) R 3

0
ZZZ
B(r) = d r + d r ,
4 S r V r
where R = (r r )/|r r | and r = |r r |. Notice that a free surface current

density may be added to the numerator of the integrand in the above surface
integral contributions and that a free volume current density may be added
to the numerator of the integrand in the above volume integral contributions
in order to account for any externally supplied current sources.
For a simple magnetic material, the spatial average of the microscopic
current density is thus seen to be given by
hj(r)i = J(r) + M(r), (4.42)

4.6 The Concept of a Perfect Magnetic Material 71
where J denotes the macroscopic current density in the magnetic material

and where M is the macroscopic magnetization defined in Eq. (2.176). The
magnetization current density Jm is then defined as the curl of the magneti-
zation as
Jm (r) M(r) (4.43)
with associated magnetization surface current density Jsm (r) M(r) n.
With substitution from Eq. (2.181), the spatial average of Amperes law given
in Eq. (2.180) becomes
H(r) = J(r), (4.44)
where
1
H(r) B(r) M(r) (4.45)
0
is the magnetic intensity vector (in A/m) for a simple magnetic material.
Because
B(r) = 0 H(r) + 0 M(r), (4.46)
the macroscopic magnetic induction field B(r) inside a simple magnetic or
magnetizable material is seen to be given by the sum of two vector fields: the
field 0 H(r) associated with the spatially averaged atomic current density
plus any externally supplied current, where (0 H(r)) = 0 J(r), and
the field 0 M(r) associated with the magnetization current density of the
magnetizable medium, where (0 M(r)) = 0 JM (r).
For a simple magnetizable material, the macroscopic magnetization M(r)
is linearly related to and in the same direction as the macroscopic magnetic
intensity vector H(r) at that point, so that
M(r) m H(r), (4.47)
where m is the magnetic susceptibility. Substitution of this result in Eq.

(2.185) then gives
B(r) = 0 (1 + m )H(r). (4.48)
If m is positive, then the magnetizable material is said to be paramagnetic
and the magnetic induction field is strengthened by its interaction with the
material. If m is negative, then the magnetizable material is said to be
diamagnetic and the magnetic induction field is weakened by its interaction
with the material. In general, the magnitude of m is quite small for both
paramagnetic and diamagnetic materials, so that2
|m | 1. (4.49)
2
Measured static values for several typical paramagnetic materials are given by
m 2.3 105 (Al), m 1.2 105 (Mg), m 7.06 105 (Ti), and m
6.8 105 (W). Measured static values for several typical diamagnetic materials
are m 1.66 105 (Bi), m 0.98 105 (Cu), m 3.2 105 (Hg),
m 2.6 105 (Ag), and m 0.24 105 (Na).
The magnetic permeability of the magnetizable medium is then defined as
0 (1 + m ), (4.50)
and the relative permeability is then given by r = /0 , where is real-

valued in the static case. The relation (2.187) between the macroscopic mag-
netic induction and intensity vectors then becomes
B(r) = H(r) (4.51)
for simple (perfect) magnetic materials. This simple linear relationship does
not apply for either ferromagnetic, antiferromagnetic, or ferrimagnetic ma-
terials as they exhibit hysteresis, a rate-independent branching nonlinearity
relating the B- and H-fields. [4]
Boundary conditions on the magnetostatic field vectors across an interface
S separating two simple magnetic materials with permeabilities 1 and 2
may be obtained by direct application of the integral form of Gauss law to a
simple closed surface with identical faces on opposite sides of S and Amperes
law to a simple closed circuit with identical segments on opposite sides of S
with the results

n B2 (r) B1 (r) = 0 , r S, (4.52)

n H2 (r) H1 (r) = Js (r) , r S, (4.53)
where n is the unit normal to the surface at the point r, directed from medium
1 into medium 2. Notice that, unless one of the materials is a superconductor
(a material with 0 so that E = 0 in its interior and which also
completely excludes magnetic flux in its interior so that B = 0), the surface
current density Js (r) = 0.
4.7 Faradays Law of Electromagnetic Induction
Consider a simple closed circuit C enclosing an open surface S in a magnetic

field B(r). The magnetic flux m linking the circuit C is then defined as
ZZ
m B nd2 r (4.54)
S
where n is the unit normal vector to S in the chosen positive direction. Fara-
days law of electromagnetic induction (1831) then states that the induced
electromotive force (emf ) I
E E d (4.55)
C
4.7 Faradays Law of Electromagnetic Induction 73
in the positive sense about the closed circuit C (as determined by the choice
of n) is related to the change in magnetic flux through the circuit by
dm
E= (4.56)
dt
This electromotive force then acts as the driving force for the current induced
in the closed circuit C, and Faradays law states that this emf results from
any change in the magnetic flux linking that circuit. Notice that the mathe-
matical expression (2.195) of Faradays law is a statement of an independent
experimental law and is entirely independent of the manner in which the
magnetic flux linkage changes. Nevertheless, due care must be taken in defin-
ing the respective reference frames that contain the circuit and the magnetic
flux source, as is the case in the so-called Faradays paradox.3
The negative sign appearing in Faradays law (2.195) indicates that the
induced emf opposes the change that is producing it. This then leads to Lenzs
law which states that a change in magnetic flux induces an emf that produces
a current whose magnetic field opposes the original change in magnetic flux.
4.7.1 Self-Inductance
According to the Biot-Savart law (2.142), the magnetic flux linking an isolated
circuit is dependent on both the geometry of the circuit and is linearly related
to the current in the circuit. For a stationary circuit with fixed shape, the
only change in its magnetic flux linkage must then result from changes in
the current, so that dm /dt = (dm /dI)(dI/dt). Because m is directly
3
Faradays paradox results from an experimental arrangement for Faradays law that
consists of a conducting cylindrical disc and a coaxial cylindrical permanent magnet
that are separated by a small distance along their common axis. The magnets B-
field is then directed along this axis. In order to measure the effects of this field, an
electric circuit is formed through a pair of sliding contacts situated at the center and
the rim of the conducting disc. When the disc is rotated with angular frequency
and the magnet is held fixed, the magnetic Lorentz force Fm = qv B produces
a radial current that is outwardly directed in the conducting disc, resulting in a
measured direct current in the circuit. In this arrangement the device is known as a
Faraday generator or Faraday disc. However, if the disc is held fixed while the magnet
is rotated about its axis, no current is measured. In addition, if both the disc and
magnet are rotated together, a current is measured. This experiment is described as
a paradox because the results appear to violate Faradays law as the magnetic flux
linkage with the conducting disk appears to be the same regardless of which element
is rotating so that dm /dt = 0 and there should never be an induced emf or current
flow. Alternatively, from the viewpoint of the magnetic lines of flux, the induced emf
is proportional to the rate at which the magnetic flux lines are cut. If the flux lines
originate in the magnet, then they are stationary in the frame of reference of the
magnet so that rotating the disc relative to the magnet, either by rotating the disc or
the magnet, should then produce an emf, but rotating both of them together should
not. The resolution of this apparent paradox is obtained simply by noting that the
motion of the magnet is entirely irrelevant in Faradays law.
proportional to the current I, then dm /dI is independent of I. This quantity

is defined as the inductance L, where
dm
L (H) (4.57)
dI
With this identification, Faradays law (2.195) becomes
dI
E = L , (4.58)
dt
where the unit of inductance in the MKSA system is the henry (H V s/A).
Notice that if a ferromagnetic material is used in the inductor, then the
relation between m and I is nonlinear and dm /dI is no longer independent
of I. In that case dm /dI is called the incremental inductance and the ratio
m /I the inductance.
4.7.2 Mutual Inductance
Consider a system of n isolated (nonintersecting) closed circuits Cj , j =

1, 2, . . . , n, each with current Ij . The magnetic flux mi linking the ith circuit
is then given by the sum of the individual flux linkages produced by the
currents in each of the n circuits, so that
n
X
mi = mij , (4.59)
j=1
where mij is the magnetic flux linking the circuit Ci due to the current Ij
flowing in the circuit Cj . The induced emf in the closed circuit Ci is then
given by Faradays law as
n
dmi X dmij
Ei = = . (4.60)
dt j=1
dt
If each circuit Cj is rigid and stationary in the laboratory reference frame, then
the only change in each magnetic flux linkage mij is that which is caused by
changes in the currents Ij , so that dmij /dt = (dmij /dIj )(dIj /dt), where
dmij /dIj is independent of the current Ij in the linear case. This quantity
then defines the mutual inductance
dmij
Mij , i 6= j (H) (4.61)
dIj
between circuit i and circuit j. Notice that this expression reduces to the
self-inductance of the ith circuit when i = j, where Li = Mii .
4.8 Magnetic Energy 75
The mutual inductance for a pair of stationary, rigid circuits in a medium

with linear magnetic properties is given by M21 = dm21 /dI1 = m21 /I1 .
The magnetic flux m21 is obtained from the Biot-Savart law (2.140) and Eq.
(2.193) as
d1 (r2 r1 )
Z Z I

m21 = I1 ndr22 . (4.62)
4 S2 C1 |r2 r1 |3
Because C1 d|r1 (r2 r1 )

= 2 C1 |r2d
H H
r1 | , the mutual inductance M21 linking
1
3
2 r1 |
circuit 2 to the current flowing in circuit 1 is given by
I
d1
ZZ
M21 = 2 ndr22 .
4 S2 C1 |r2 r1 |
Finally, application of Stokes theorem to the surface integral in this expres-

sion results in Neumanns formula for the mutual inducatnce
d1 d2
I I
M21 = (4.63)
4 C2 C1 |r2 r1 |
The inherent symmetry of this expression then shows that, in general
Mij = Mji . (4.64)
Notice that, in principle, Neumanns formula is applicable to determining the

self-inductance of a circuit. However, due care must be given to the logarith-
mic singularity at r2 = r1 in the resultant contour integral. This singular
behavior may be resolved by including the thickness of the current-carrying
circuit, a quantity which, in most cases, may be conveniently neglected in
determining the mutual inductance.
4.8 Magnetic Energy
Because the magnetic force dFm = dqvB acting on an element of charge dq

is perpendicular to its velocity v, then unlike an electric field, a magnetic field
does no work on the charges moving through it. Consequently, in order to
determine the change in energy of a system of moving charges (i.e., currents)
when a magnetic field is applied, one must first consider the electric field that
is induced by the magnetic field variation (through Faradays law) and then
determine the work that it does on the currents producing the magnetic field.
The current I produced in a circuit C with resistance R when an external
source of emf E0 is applied is given by E0 + E = IR, where E is the induced
emf. The work done by the external emf E0 in moving the increment of charge
dq = Idt through the circuit C is then given by E0 dq = E0 Idt so that
E0 dq = EIdt + I 2 Rdt
= Idm + I 2 Rdt,
where dm = Edt by Faradays law. The I 2 Rdt term accounts for the
irreversible Joule heat loss in the circuit while the Idm term describes the
work done in opposition to the induced emf in the circuit. This (externally
supplied) incremental work
dU = Idm (4.65)
is positive when the change in magnetic flux dm linking the circuit is in
the same direction as the magnetic flux produced by the current I. If the
circuit is linear (i.e., non-hysteretic) and not changing (i.e., it is rigid and
fixed in space), then dU = dUm describes the change in magnetic energy of
the circuit.
Consider now determining the magnetic energy that is required to establish
the currents Ij in a system of n interacting circuits Cj , j = 1, 2, . . . , n. From
Eq. (2.204), the incremental electrical work done against the induced emf Ej
in each circuit is given by
n
X
dU = Ij dj . (4.66)
j=1
When the changes in magnetic flux dj are produced by current changes dIj
in the n circuits, then
n n
X djk X
dj = dIk = Mjk dIk , (4.67)
dIk
k=1 k=1
where Mjk is the mutual inductance between circuits j and k. As before,

if each circuit Cj is rigid and stationary, then dU is equal to the change in
magnetic energy dUm of the system of interacting circuits; if not, then Eq.
(2.205) describes both the change in the mechanical and magnetic energy of
the system. In the former (mechanically static) case, the magnetic energy is
given by
Xn X n Z Ij
Um = Mjk Ij dIk . (4.68)
j=1 k=1 0
Because the magnetic energy of the final state is independent of the manner
in which each current Ij is established in each corresponding circuit Cj of the
system, consider the simplest case when they are all brought to their final
values in concert such that at each instant they are all at the same fraction
of their final values so that Ij = Ij and dIk = Ik d where increases
from = 0 to = 1. Then
4.8 Magnetic Energy 77
n n
1 XX
Um = Mjk Ij Ik (J) (4.69)
2 j=1
k=1
Pn
Because j = k=1 Mjk Ik , this expression may also be written as
n
1X
Um = j Ij (J) (4.70)
2 j=1
which should be compared with the expression (2.86) for the electrostatic
energy of an assembled charge configuration.
For the special case of two coupled circuits, the magnetic energy is given
by
1 1
Um = L1 I12 + M I1 I2 + L2 I22 ,
2 2
where M M12 = M21 . Note that the quantity M I1 I2 may be either positive
or negative, whereas the magnetic energy Um must be non-negative. With
I1 /I2 , this expression becomes Um = 12 I22 L1 2 + 2M + L2 0 which

has an extremum at = M/L1 where dUm /d = I22 (L1 + M ) = 0; because

d2 Um /d 2 = L1 I22 0, this extremum is a minimum. This minimum value of
Um must be non-negative so that M 2 /L1 2M 2 /L1 +L2 0, or L1 L2 M12 2
.
In general
2
Lj Lk Mjk . (4.71)
In order to extend this result to include the magnetic energy of a volume
distribution of current described by the current density J(r) in a region R,
notice first that any single current loop can be treated as a system of contigu-
ous filamentary currents Ij flowing along closed circuits Cj , j = 1, 2, . . . , N
that have magnetic flux linkage
ZZ I
j = B nda = A dj ,
Sj Cj
where Sj is the open surface bounded by the circuit Cj and where A is

the local magnetic vector potential. Substitution of this expression into Eq.
(2.209) for the magnetic energy results in
N
1X
I
Um = Ij A dj .
2 j=1 Cj
Because Ij dj = Jaj dj = Jd3 r , then as N , the above expression

becomes
1
Z
Um = J(r) A(r)d3 r. (4.72)
2 R
With substitution from Amperes law [J(r) = H(r)] for a steady-state

magnetic field, the vector identity (A H) = H ( A) A ( H)
yields J A = H ( A) (A H) so that with B = A and
application of the divergence theorem, Eq. (2.211) becomes
1 1
Z I
3
Um = H(r) B(r) d r (A(r) H(r)) nd2 r, (4.73)
2 R 2 S
where the closed surface S bounds the region R. Because the current density
J(r) vanishes outside of the finite region R, the region of integration may
be extended to infinity without changing the result. Because H(r) falls off
at least as fast as 1/r3 and A(r) falls off at least as fast as 1/r2 as r
[see Eqs. (2.164)(2.166)], where r is the radial distance from a fixed origin
O situated near the current source distribution to a point on the surface S
as S is expanded to infinity, and since the surface area varies as r2 , then the
integrand in the surface integral appearing in Eq. (2.212) falls off at least as
fast as 1/r3 and so vanishes in this limit. Consequently, the total magnetic
energy is given by
1
ZZZ
Um = H(r) B(r) d3 r (4.74)
2
the integration extending over the region of space containing the current
density J(r). The integrand appearing in Eq. (2.213) is defined as the mag-
netostatic energy density
1
um (r) H(r) B(r) (J/m3 ) (4.75)
2
which is associated with the field energy at each point in space.

References 79
Problems
4.1. Show that two measurements F1 and F2 of the magnetic force Fm =

qv B at a fixed point are sufficient to uniquely determine the magnetic
induction field vector B at that observation point as

1 F1 v1 (F2 v2 ) v1
B= + v1 , (4.76)
q v12 v12 v22
provided that v1 and v2 are mutually orthogonal velocities.
4.2. Calculate the = 0 term in the multipole expansion for the magnetic
scalar potential given in Eq. (2.174). Show that this monopole term is zero.
References
1. R. S. Elliott, Electromagnetics: History, Theory, and Applications. Piscataway,

NJ: IEEE, 1993.
2. J. B. Bronzan, The magnetic scalar potential, Am. J. Phys., vol. 39, no. 11,
pp. 13571359, 1971.
3. P. Lorrain and D. R. Corson, Electromagnetic Fields and Waves. New York: W.
H. Freeman and Co., second ed., 1970.
4. E. D. Torre, Magnetic Hysteresis. Piscataway, NJ: IEEE Press, 1999.
5. R. A. Chipman, Theory and Problems of Transmission Lines. New York: McGraw-
Hill, 1968.
Chapter 5
Transmission Line Analysis
The history of science shows that theories are perishable. Nikola Tesla
5.1 Distributed Resistance and Internal Inductance of

Homogeneous, Isotropic Cylindrical Conductors
A problem of considerable practical importance involving both electric and

magnetic field effects concerns the distributed resistance and internal induc-
tance of transmission line conductors. For the elementary case of a solid
homogeneous, isotropic wire with circular cross-section of radius a and static
conductivity 0 , the resistance per unit length at zero frequency is given by
R0 = (0 a2 )1 [see the example following Eq. (2.136)] . The inductance per
unit length is another matter, as described below.
As described by Chipman [1], the distributed inductance of a conductor
(either isolated or not) is comprised of two parts: that due to internal flux-
current linkages, designated as Li (), and that due to external flux-current
linkages, designated as Lx (), where = 2f is the angular oscillation fre-
quency of the current source.
Consider a uniform, solid conductor with magnetic permeability 0
and circular cross-section of radius a carrying a dc (f = 0) current I. The
differential amount of current flowing in an infinitesimal annular tube of
radius r a and radial width dr is then given by (I/a2 )(2r dr ). With
the definition in Eq. (2.196) that inductance is the flux linking a circuit
per unit current flowing in the circuit, this cylindrical differential element
constitutes a fraction (2r dr )/(a2 ) of the conductor as a circuit element.
Because the magnetic flux lines due to each cylindrical current element in the
wire are circles concentric with the conductor, this fractional circuit element
is linked by the magnetic flux inside the conductor between r and a, where
81
82 5 Transmission Line Analysis
2

I r
B(r) = (5.1)
2r a2
with r r a. The contribution to the distributed dc internal inductance

Li (0) per unit length due to the fractional circuit consisting of the cylindrical
tube with thickness dr at radius r is given by
2r dr a I r2
Z
dLi (0) = dr
a2 I r 2r a2
= a2 r2 r dr

(5.2)
in H/m. The total internal dc inductance is then obtained by integrating this

expression over the radial extent of the wire as
Z a

Li (0) = dLi (0) = , (5.3)
0 8
and is thus independent of the conductor radius.

From Eq. (2.216), the incremental internal dc inductance per unit annular
cross-sectional area is given by dLi (0)/(2r dr ) = (/2) a2 r2 so that
the distributed internal inductance decreases monotonically as r increases
from the center r = 0 to the periphery r = a of the cylindrical conductor.
This result then implies that, at a fixed frequency, the distributed internal
reactance XL = iL also decreases as r increases from 0 to a. If an ac
voltage is applied across the ends of a section of the cylindrical conductor,
less current per unit area will flow in the high reactance region near the center
of the conductor than in the low reactance region nearer to the periphery
of the conductor. Unlike the uniform current density obtained in the zero
frequency limit, the current density is now nonuniform and this effect becomes
increasingly pronounced as the frequency increases. At a sufficiently high
frequency the current flow will be essentially confined to a very thin layer
at the conductor surface, a phenomenon known as the skin effect. A more
detailed analysis of this effect is now given for the important case of cylindrical
symmetry about the z-axis wherein both the electric field and current density
are functions of the radius r alone.
For a time-harmonic field with angular frequency n = 2fo, the phasor
form of the conduction current density Jz (r, t) = Jz (r)eit is given by
Ohms law as [cf. Eq. (2.127)]
Jz (r) = ()Ez (r), (5.4)
where Ez (r) is the complex phasor of the electric field intensity along the
cylinder axis of the conductor. The phasor form of the azimuthally-directed
magnetic induction field at a radial distance rRfrom the conductor axis pro-
= r Jz (r )2r dr flowing along
duced by the total conduction current I(r) 0
the z-direction within that cylinder of radius r is obtained from Eq. (2.141)
5.1 Distributed Resistance and Internal Inductance 83
as

I(r)
Z r
B (r) = = Jz (r )dr . (5.5)
2r 0
The phasor electric Ez (r) and magnetic B (r) fields are related through the
phasor form of Faradays law [cf. Eqs. (2-193)(2.195)] as
I ZZ
E d = i B nda. (5.6)
C S
Application of this law to a rectangular contour situated in the rz-plane of

the conductor at a distance r from the conductor axis with length z and
radial width r then gives
(" # ) Z r
Ez (r)
Ez (r) + z Ez (r)z r = i zr Jz (r )dr .
r r 0
With Eq. (2.218) one then obtains the integro-differential equation
dJz (r) r
Z
r = i() Jz (r )dr (5.7)
dr 0
for the phasor conduction current density in the conductor. Differentiation

of this expression with respect to r then gives the second-order differential
equation
d2 Jz (r) 1 dJz (r)
+ i()Jz (r) = 0, (5.8)
dr2 r dr
which is a modified form of Bessels equation of order zero [see Eq. (C.41) of
Appendix C], with solution

Jz (r) = a1 J0 2i (r/()) + a2 Y0 2i (r/()) , (5.9)
where a1 and a2 are constants of integration, and where

s
2
() (5.10)
()
is the skin depth of the conductor at angular frequency . Because the

argument of the Bessel functions appearing in Eq. (2.223) is complex-valued,
one introduces the real and imaginary parts through the Kelvin functions
n o
ber() J0 i , (5.11)
n o
bei() J0 i , (5.12)
n o
ker() Y0 i , (5.13)
n o
kei() Y0 i , (5.14)
where is real-valued. With these substitutions, Eq. (2.223) becomes

h i h i
Jz (r) = a1 ber 2 r/ + ibei 2 r/ + a2 ker 2 r/ + ikei 2 r/ .

(5.15)
This represents the general solution for the current density for a hol-
low tubular conductor with r bounded away from zero. However, because
lim0 ker() = , the coefficient a2 must be zero for a solid tubular con-
ductor, so that the solution becomes
h i
Jz (r) = a1 ber 2 r/ + ibei 2 r/ ,

(5.16)
for 0 r a.
Because the current has penetrated into the conductor due to the external
electric field at the conductors surface (r = a), the quantity of interest here is
the current density Jz (r) relative to the current density Jz (a) at the surface
of the conductor, given by

Jz (r)

ber 2 r/ + ibei 2 r/
= . (5.17)
Jz (a)

ber 2 a/ + ibei 2 a/
The radial dependence of both the real and imaginary parts of this expression
for the relative current density is depicted in Figs. 2.13 and 2.14 for angular
frequency values at which a/ = 5 and a/ = 10, respectively. Notice that as
the skin depth decreases (i.e. as a/ increases), the current density becomes
increasingly concentrated near the surface of the conductor. In particular,
since (0) = 0 , where 0 denotes the static conductivity, then and
a/ 0 as 0 so that the current density Jz (r) becomes uniformly
distributed (i.e. independent of r) throughout the cross-section of the wire
under static conditions.
The distributed internal impedance Zi of a solid cylindrical conductor is
related to the distributed resistance R and distributed internal inductance Li
of the conductor by Zi = R + iLi . By definition, this complex impedance
is given by the ratio of the longitudinal phasor potential difference per unit
length of the conductor at the conductors surface, obtained from Eq. (2.218)
as Ez (a) = Jz (a)/(), to the total phasor current Iz in the conductor, so
that
5.1 Distributed Resistance and Internal Inductance 85
0.8
0.6
Jz(r)/Jz(a)
0.4
~
Fig. 5.1 Radial depen-
dence of the real (solid ~
0.2
curve) and imaginary

0
(dashed curve) parts of
the relative current den-
sity Jz (r)/Jz (a) in a solid
-0.2
cylindrical conductor of -0.4

radius a at a frequency 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/a
for which a/ = 5.
0.8
0.6
Jz(r)/Jz(a)
0.4
~
Fig. 5.2 Radial depen-

~
dence of the real (solid 0.2
curve) and imaginary

0
(dashed curve) parts of
the relative current den-
sity Jz (r)/Jz (a) in a solid
-0.2
cylindrical conductor of -0.4

radius a at a frequency 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/a
for which a/ = 10.
Jz (a)
Zi () = R() + iLi () = . (5.18)
()Iz
It now remains to determine an expression for the total phasor current Iz in
the cylindrical conductor. Because of the cylindrical symmetry of the con-
ductor geometry, E(r) = 1z Ez (r) and H(r) = 1 H (r). From the integral
form of Amperes law [see Eq. (2.151)], Iz is given by the contour integral of
H (r) around the periphery (r = a) of the cylindrical conductor, with the
result
Iz = 2aH (a). (5.19)

From the phasor form of Faradays law, given by E = i B, with B = H,
one obtains
1 Ez (r) 1 Jz (r)
H (r) = = (5.20)
i r i r
after substitution from Eq. (2.218). With substitution from Eq. (2.230),

2

a1 ber 2 r/ + ibei 2 r/ , for

this expression becomes H (r) = i
0 r a, where the prime denotes differentiation with respect to the ar-
gument. With this result, Eq. (2.233) for the total phasor current becomes
3/2
Iz = 2ia
a1 ber 2 a/ + ibei 2 a/ . With a1 given by Eq. (2.230)

evaluated at r = a, one finally obtains

"

#
3/2 ber 2 a/ + ibei 2 a/
2 a
Iz = Jz (a) , (5.21)
i

ber 2a/ + ibei 2a/
so that the expression given in Eq. (2.232) for the distributed internal
impedance becomes
" #
Rs () ber 2a/ + ibei 2a/
Zi () = R() + iLi () = i , (5.22)
2a ber 2 a/ + ibei 2 a/
where
1
r
Rs () = (5.23)
()() 2()
is the surface resistivity (also known as the skin resistivity), in /m2 .
5.2 Transmission Lines
A transmission line may be defined as a two-port network connecting a gen-

erator circuit at the transmitter to a load circuit at the receiver, as depicted
in Fig. 2.15. Analysis of the interrelated voltage and current behavior along
the transmission line then relies upon the following set of assumptions [1]:
Assumption 1: The uniform transmission line is comprised of two straight
parallel conductors, beginning at the source terminal AA and ending at
the load terminal BB .
Comment (a): The analysis is applicable to transmission line systems
with more than two two parallel conductors provided that they are
interconnected in a manner that presents only two terminals at the
source AA and load BB ends.
5.2 Transmission Lines 87

Thevenin-
Equivalent
RLC Source
Network A i(z,t) B
Terminal
Load
v(z,t)
RLC
Network
A i(z,t) B
Source
Voltage
Vg(t)
(a)
Zg
~
A I(z) B
~
V(z) ZL
A ~ B
I(z)
~
Vg
(b)
Fig. 5.3 Generalized transmission line circuit with associated source (connected at
terminal AA ) and load (connected at terminal BB ) networks in (a) the time domain
and (b) the frequency domain.
Comment (b): By a uniform transmission line it is meant that the ma-

terial properties [mass density m , electric conductivity (), dielec-
tric permittivity (), magnetic permeability ()] of both the conduc-
tors and surrounding medium, the line geometry (including bends and
twists1 ) as well as the cross-sectional conductor geometry are indepen-
dent of the distance z along the entire length of the line.
Assumption 2: The electric current only flows in a direction parallel to the
axial line of the conductors, taken here to be along the z-axis.
Assumption 3: At any instant of time t and position z along the transmis-
sion line, the instantaneous total currents i(z, t) in the two conductors are
1
The perturbative effects of bends and twists are negligible provided that the rate
of the bend or twist in the line geometry is less than approximately one degree in a
line length that is comparable with the line separation.
equal in magnitude and flowing in opposite directions, as depicted in Fig.

2.15(a).
Assumption 4: At any instant of time t and position z along the transmis-
sion line, there is a unique value of the potential difference v(z, t) between
the conductors, as depicted in Fig. 2.15(a).
Assumption 5: The electrical behavior of the transmission line is completely
described by the following set of four uniformly distributed electric circuit
coefficients:
Resistance R per unit length (/m) and inductance L per unit length
(H/m) uniformly distributed in series along the entire length of the
transmission line.
Capacitance C per unit length (F/m) and leakage conductance G per
unit length (S/m) uniformly distributed as shunt circuit elements along
the entire length of the transmission line.
The values of each of these distributed circuit coefficients at any given
frequency are, by assumption 1(b), determined only by the geometry,
dimensions, and material properties of the line conductors and the sur-
rounding medium, are independent of the line voltage and current as well
as independent of time. As a consequence, the transmission line, as con-
sidered here, is a linear passive network. Finally, for transverse electromag-
netic (TEM) transmission lines2 , these line parameters satisfy the relations
[see Eq. (2.137)]
LC = , (5.24)
G
= , (5.25)
C
where is the magnetic permeability, the dielectric permittivity, and
the conductivity of the surrounding medium. For example, for a coaxial line
with inner radius a and outer radius b containing an insulating material
with material parameters , , , the inductance per unit length is given by
L = (/2) ln (b/a), the shunt conductance per unit length is given by G =
2/ ln (b/a), and the capacitance per unit length is C = 2/ ln (b/a).
2
Electromagnetic analysis of uniform two-conductor transmission lines shows that
there are three fundamental mode types that may propagate: transverse electric (TE)
modes in which the electric field vector is orthogonal to the z-axis and the magnetic
field vector has a non-vanishing component along the z-direction; transverse mag-
netic (TM) modes in which the magnetic field vector is orthogonal to the z-axis and
the electric field vector has a non-vanishing component along the z-direction; and
transverse electromagnetic (TEM) modes in which both the electric and magnetic
field vectors are orthogonal to the z-direction. Both the TE and TM modes have a
cutoff frequency below which they cannot propagate. This cutoff frequency is in the
microwave frequency range ( 0.5GHz 40GHz) for typical transmission lines.
However, the TEM mode has no nonzero cutoff frequency.
5.2.1 Transmission Line Equations
Consider the equivalent circuit of a differential length z of a uniform two-

conductor transmission line depicted in Fig. 2.16. Application of Kirchhoff s
Fig. 5.4 Equivalent cir-

z
cuit of a differential length
z of a two-conductor i(z,t)
Lz Rz
i(z+z,t)
transmission line with re- N N+1
sistance R per unit length

and inductance L per v(z,t) Gz Cz v(z+z,t)
unit length in series, and
with capacitance C per
unit length and leakage i(z,t) i(z+z,t)
conductance G per unit z z+z
length in parallel.
voltage law (KVL) to this circuit results in
I(z, t)
V (z, t) RzI(z, t) Lz V (z + z, t) = 0,
t
which may be rearranged in the more suggestive form

V (z + z, t) V (z, t) I(z, t)
= RI(z, t) + L .
z t
In the limit as z 0, this expression becomes
V (z, t) I(z, t)
= RI(z, t) + L (5.26)
z t
which is the first of a pair of transmission line equations. The second equation
of this pair is obtained from application of Kirchhoff s current law (KCL) to
node N + 1 in Fig. 2.16, with the result
V (z + z, t)
I(z, t) GzV (z + z, t) Cz I(z + z, t) = 0,
t
which may be rearranged in the more suggestive form

I(z + z, t) I(z, t) V (z + z, t)
= GV (z + z, t) + C .
z t
In the limit as z 0, this expression becomes
I(z, t) V (z, t)
= GV (z, t) + C (5.27)
z t
Taken together, Eqs. (2.240) and (2.241) comprise the time-domain form of
the transmission line equations or telegraphers equations.
5.2.1.1 Time-Harmonic Wave Solutions
It is more convenient to work with the phasor form of these equations ob-
tained through the time-harmonic assumption
n o n o
V (z, t) = V (z)et
& I(z, t) = I(z)e t
, (5.28)
where = (by convention) and where = 2f is the angular frequency of

the line voltage and current. With these substitutions, Eqs. (2.240)(2.241)
become
dV (z)
= (R + L)I(z),
dz

dI(z)
= (G + C)V (z).
dz
The frequency-dependence of the transmission line parameters R R(),
L L(), G G(), and C C() may now be introduced in these
frequency-domain equations with the understanding that their correspond-
ing terms in the time-domain form of the transmission line equations are
convolution operations. The frequency-domain form of the transmission line
equations or the phasor form of the telegraphers equations are then given by
dV (z)

= R() + L() I(z) (5.29)
dz

dI(z)
= G() + C() V (z) (5.30)
dz
An equation for the phasor voltage alone may be obtained by differentiat-
ing Eq. (2.243) with respect to z and substituting Eq. (2.244) into the result
as
d2 V (z)
dI(z)
2
= R() + L()
dz dz

= R() + L() G() + C() V (z).
Similarly, an equation for the phasor current alone may be obtained by dif-
ferentiating Eq. (2.244) with respect to z and substituting Eq. (2.243) into
the result as

d2 I(z) dV (z)
2
= G() + C()
dz dz

= G() + C() R() + L() I(z).
This pair of phasor wave equations may be rewritten in a more compact form
as
d2 V (z)
2 ()V (z) = 0 (5.31)
dz 2

d2 I(z) =0
2 ()I(z) (5.32)
dz 2
These are the Helmholtz equations for wave propagation on a dispersive trans-
mission line. For either equation
2 () = R() + L() G() + C()

1 L 1 C
= R2 + L2 2 e tan ( R ) G2 + C 2 2 e tan ( G )
p p
1 L 1 C
= (R2 + L2 2 ) (G2 + C 2 2 )e[tan ( R )+tan ( G )] ,
p
where the explicit frequency dependence of the line parameters is now left
understood. The magnitude of () = |()|e () is then given by
1/4
R 2 + L2 2 G2 + C 2 2

|()| = , (5.33)
and its phase is

1 L C
() arg () = tan1 + tan1 . (5.34)
2 R G
The complex propagation factor () may then be expressed as
() = () + (), (5.35)
where
() {()} = |()| sin ( ()) (5.36)
is the propagation factor in radians per meter (rad/m), and
() {()} = |()| cos ( ()) (5.37)
is the attenuation factor in nepers3 per meter (Np /m). Notice that
() = () (5.38)
for real .
Traveling wave solutions to the phasor wave equations (2.245)(2.246) are
3
By definition, a neper is a base e logarithmic measure of the ratio between two
voltage or current magnitudes.
V (z) = V0+ e()z + V0 e()z , (5.39)

= I + e()z + I e()z ,
I(z) (5.40)
0 0
where
e()z et = e()z e(()zt)
describes attenuated wave propagation in the positive z-direction, and
e()z et = e()z e(()z+t)
describes attenuated wave propagation in the negative z-direction.

Substitution of the traveling wave solution given in Eq. (2.253) into the
phasor wave equation given in Eq. (2.243) and solving for the phasor current
yields
=
V0+ ez + V0 ez .

I(z)
R + L
Comparison of this result with the expression given in Eq. (2.254) for the
phasor current shows that
V0+ V0
= = Z0 , (5.41)
I0+ I0
independent of the position z along the line, where

1/2
R() + L() R() + L()
Z0 () = (5.42)
() G() + C()
is the characteristic impedance of the line. The traveling wave solution (2.254)
for the phasor current may now be expressed as
= 1 + ()z
I(z) V0 e + V0 e()z . (5.43)
Z0 ()
In general, let V0+ = |V0+ |e+ at the generator end and let V0 = |V0 |e
at the load end of the transmission line be given. Then the instantaneous
voltage on the line is given by
n o
V (z, t) = V (z)et
n o
= V0+ e()z + V0 e()z et

+ ()z t()z++ ()z t+()z+
= |V0 |e e + |V0 |e e
so that
V (z, t) = |V0+ |e()z cos t ()z + +

+|V0 |e()z cos t + ()z + ,

(5.44)
where the first term on the right describes a wave traveling in the +z-direction
and the second term a wave traveling in the z-direction. A similar expression
holds for the instantaneous current on the line. Both time-harmonic waves
propagate with the phase velocity

vp () = = . (5.45)
() |()| sin
5.2.1.2 Temporal Pulse Solutions
By linear superposition, a voltage pulse traveling in the positive z-direction

along a dispersive transmission line is given by the Fourier integral represen-
tation Z
1
V (z, t) = V0 ()e()z et d, (5.46)
2
where Z
V0 () = V0 (t)et dt (5.47)

is the initial voltage pulse spectrum with V0 (t) V (0, t). From Eq. (2.257),
the related current pulse is then given by

1 V0 () ()z t
Z
I(z, t) = e e d, (5.48)
2 Z0 ()
R
where I0 (t) = I(0, t) = (1/2) (V0 ()/Z0 ())et d describes the initial
current pulse with initial pulse spectrum I0 () = V0 ()/Z0 (). Because both
V0 (t) and I0 (t) are real-valued [as are V (z, t) and I(z, t)], it then follows from
Eq. (2.261) that the following symmetry relations hold:
V0 () = V0 (),
I0 () = I0 (), (5.49)
Z0 () = Z0 (),
for real . The real parts of each of these spectra are then even functions of
real and their imaginary parts are odd-functions of real .
For either an amplitude-modulated (AM) or pulse-modulated (PM) volt-
age pulse, the initial voltage pulse is given by
V0 (t) = u(t) sin (c t), (5.50)

where u(t) is the envelope function and c is the fixed angular frequency of
the carrier wave. The Fourier spectrum of this pulse type is then given by
1
Z
u(t) ec t ec t et dt

V0 () =
2
Z Z
1 (c )t (+c )t
= u(t)e dt u(t)e dt
2
1
= u( c ) u( + c ) , (5.51)
2
R
where u() = u(t)et dt is the spectrum of the initial pulse envelope
function which, because u(t) is real-valued, satisfies the symmetry relation
u () = u(), (5.52)
for real . As a consequence of both this symmetry relation and the symmetry
relation (2.252) satisfied by (), Eq. (2.260) may be written as
Z Z
1 1 ()z t ()z t
V (z, t) = u( c)e e d u ( c)e e d
2 2
( Z
1 1
= u( c )e()z et d
2 2
Z )
1 ()z t
+ u( c )e e d ,
2
so that Z
1 1
V (z, t) = u( c )e()z et d , (5.53)
2
for z 0. In a similar manner, the associated current pulse is given by
Z
1 1 u( c ) ()z t
I(z, t) = e e d , (5.54)
2 Z0 ( c )
for z 0.
For an amplitude modulated (AM) signal, the envelope function u(t) is
strictly slowly-varying with respect to the period Tc = 2/c of the carrier
wave. In that case, the spectrum u( c ) is strongly peaked about the
carrier frequency c . The attenuation and propagation factors may then be
approximated as
() (c ), (5.55)
() (c ) + (c )( c ), (5.56)
for | c |/c 1, where () ()/. If |u( c )| is negligible

outside of this narrow frequency interval, the Fourier integral representation
(2.267) of the AM voltage signal may be approximated as

1 (c )z
V (z, t) e e((c )zc t)
2
Z

u( c )e( (c )zt)(c )) d ,

with the result
V (z, t) u ( (c )z t) e(c )z sin (c )z c t ,

(5.57)
for z 0. Within this linear dispersion approximation, the AM signal en-

velope propagates undistorted, but attenuated, at the group velocity vg (c ),
where
1 1
vg () = , (5.58)
() ()/
while the carrier wave propagates at the phase velocity vp (c ) = c /(c ) [cf.
Eq. (2.259)]. Notice that similar results do not apply for the pulse-modulated
(PM) case as the approximations used in deriving Eq. (2.272) no longer
strictly apply.
Problems
5.1. Prove that the complex propagation factor () defined in Eqs. (2.247)
(2.249) satisfies the symmetry relation given in Eq. (2.252), given that the
real quantities R = R(), L = L(), G = G(), and C = C() satisfy similar
symmetry relations. Based upon this expression, what symmetry relations do
() {()} and () {()} satisfy for real ?
5.2. Derive Eq. (2.268) for the propagation of a current pulse.
5.3. Derive Eq. (2.271) describing the evolution of an amplitude-modulated

voltage signal along a dispersive line.
5.4. Beginning with Eq. (2.268), derive an equivalent expression describing

the evolution of an amplitude-modulated current signal along a dispersive
line.
References
1. R. A. Chipman, Theory and Problems of Transmission Lines. New York: McGraw-

Hill, 1968.
Appendix A
The Dirac Delta Function
A.1 The One-Dimensional Dirac Delta Function
The Dirac delta function [1] in one-dimensional space may be defined by the
pair of equations
(x) = 0; x 6= 0, (A.1)
Z
(x) dx = 1. (A.2)

It is clear from this definition that (x) is not a function in the ordinary
mathematical sense, because if a function is zero everywhere except at a single
point and the integral of this function over its entire domain of definition
exists, then the value of this integral is necessarily also equal to zero. Because
of this, it is more appropriate to regard (x) as a functional quantity with
a certain well-defined symbolic meaning. For example, one can consider a
sequence of functions (x, ) that, with increasing values of the parameter ,
differ appreciably from zero only over a decreasing x-interval about the origin
and which are such that Z
(x, ) dx = 1 (A.3)

for all values of . Although it may be tempting to try to interpret the Dirac
delta function as the limit of such a sequence of well-defined functions (x, )
as , it must be recognized that this limit need not exist for all values
of the independent variable x. However, the limit
Z
lim (x, ) dx = 1 (A.4)

must exist. As a consequence, one may interpret any operation that involves
the delta function (x) as implying that this operation is to be performed
with a function (x, ) of a suitable sequence and that the limit as is
97
98 A The Dirac Delta Function
to be taken at the conclusion of the calculation. The particular choice of the

sequence of functions (x, ) is immaterial, provided that their oscillations
(if any) near the origin x = 0 are not too violent [2]. Each of the following
functions forms a sequence with respect to the parameter that satisfies the
required properties.
2 2
(x, ) = e x ,

(x, ) = rect1/ (x),

(x, ) = sinc(x),

(r, ) = circ1/ (r),

(r, ) = J1 (2r),
r
where rect1/ (x) /2 when |x| < 1/ and is zero otherwise, circ1/ (r)
2 / when r < 1/ and is zero otherwise, and sinc(x) sin (x)/x when x 6= 0
and is equal to its limiting value of unity when x = 0, where the last two of
the above set of functions are appropriate for polar coordinates.
Let f (x) be a continuous and sufficiently well-behaved function of x
(, ) and consider the value of the definite integral
Z Z
f (x)(x a)dx = lim f (x)(x a, )dx.

When the parameter is large, the value of the integral appearing on the
right-hand side of this equation depends essentially on the behavior of f (x)
in the immediate neighborhood of the point x = a alone, and the error that
results from the replacement of f (x) by f (a) may be made as small as desired
by taking sufficiently large. Hence
Z Z
lim f (x)(x a, )dx = f (a) lim (x a, )dx,

so that Z
f (x)(x a)dx = f (a). (A.5)

This result is referred to as the sifting property of the delta function. That is,
the process of multiplying a continuous function by (x a) and integrating
over all values of the variable x is equivalent to the process of evaluating the
function at the point x = a. Notice that, for this result to hold, the domain of
integration need not be extended over all x (, ); it is only necessary
that the domain of integration contain the point x = a in its interior, so that
Z a+2
f (x)(x a)dx = f (a), (A.6)
a1
A.1 The One-Dimensional Dirac Delta Function 99
where 1 > 0, 2 > 0. It is then seen that f (x) need only be continuous at
the point x = a.
The above results may be written symbolically as
f (x)(x a) = f (a)(x a), (A.7)
the meaning of such a statement being that the two sides yield the same
result when integrated over any domain containing the point x = a. For the
special case when f (x) = xk with k > 0 and a = 0, Eq. (A.7) yields
xk (x) = 0, k > 0. (A.8)
Theorem A.1. Similarity Relationship (Scaling Law). For all a 6= 0

1
(ax) = (x). (A.9)
|a|
Proof. In order to prove this relationship one need only compare the integrals
of f (x)(ax) and f (x)(x)/|a| for any sufficiently well-behaved continuous
function f (x). For the first integral one has (for any a 6= 0)
Z
1 1
Z
f (x)(ax)dx = f (y/a)(y)dy = f (0),
a |a|
where the upper or lower sign choice is taken accordingly as a > 0 or a < 0,
respectively, and for the second integral one obtains
Z
1 1
f (x) (x)dx = f (0).
|a| |a|
Comparison of these two results then shows that (ax) = (x)/|a|, as was to
be proved.

For the special case a = 1, Eq. (A.9) yields
(x) = (x), (A.10)
so that the delta function is an even function of its argument.

Theorem A.2. Composite Function Theorem. If y = f (x) is any con-
tinuous function of x with simple zeroes at the points xi [i.e., y = 0 at x = xi
and f (xi ) 6= 0] and no other zeroes, then
X 1
(f (x)) = (x xi ). (A.11)
i
|f (xi )|
Proof. In order to prove this theorem, let g(x) be any sufficiently well-
behaved continuous function and let {xi } denote the set of points at which
y = 0. Under the change of variable x = f 1 (y) one has that

1
Z Z
g f 1 (y) (y)

g(x)(f (x))dx = dy
R |f (f 1 (y)) |
X 1
g f 1 (0)

=
xi
|f (f 1 (0)) |
X 1
= g(xi ) ,
i
|f (xi |
where R denotes the range of f (x). In addition,

Z ( ) Z
X 1 X 1
g(x) (x xi ) dx = g(x)(x xi )d
i
|f (xi )| i
|f (xi )|
X 1
= g(xi ) .
i
|f (xi |
Comparison of these two expressions then proves the theorem.

As an example, consider the function f (x) = x2 a2 which has simple zeroes

at x = a. Then |f (a)| = 2|a| so that, for a 6= 0,
1
(x2 a2 ) = ((x a) + (x + a)) .
2|a|
An additional relationship of interest that employs the Dirac delta function

is Z
( x)(x )dx = ( ), (A.12)

which is seen to be an extension of the sifting property to the delta function

itself. This equation then implies that if both sides are multiplied by a con-
tinuous function of either or and the result integrated over all values of
either or , respectively, an identity is obtained. That is, because
Z
f ()( )d = f (),

and
Z Z
f () ( x)(x )dx d

Z Z
= f ()( x)d (x )dx

Z
= f (x)(x )dx = f ()

then the expression in Eq. (A.12) follows. In a similar manner, because

Z Z
g() ( x)(x )dx d

Z Z
= g()(x )d ( x)dx

Z
= g(x)( x)dx = g()

and
Z
g()( )d = g(),

then the expression in Eq. (A.12) is again obtained.

Consider next what interpretation may be given to the derivatives of the
delta function. This is accomplished through R use of the function sequence
(x, ). Consider then the ordinary integral f (x) (x, )dx which may be
evaluated by application of the method of integration by parts with u = f (x)
and dv = (x, )dx, so that
Z Z
f (x) (x, )dx = f ()(, ) f ()(, ) f (x)(x, )dx.

Upon proceeding to the limit as , the first two terms appearing on the
right-hand side of this equation both vanish because
lim (, ) = 0, (A.13)

with the result Z

f (x) (x)dx = f (0). (A.14)

Upon repeating this procedure n times for the nth-order derivative of the
delta function, one obtains the general result
Z
f (x) (n) (x)dx = (1)n f (n) (0). (A.15)

As a special case of Eq. (B.14), let f (x) = x so that

Z Z
x (x)dx = 1 = (x)dx,

and one then has the equivalence
x (x) = (x). (A.16)

Because (x) is an even function and x is an odd function, it then follows

that (x) is an odd function of its argument; that is
(x) = (x). (A.17)
The generalization of Eq. (A.16) may be directly obtained from Eq. (B.15)
by letting f (x) = xn . In that case, f (n) (x) = n! and this relation gives
Z Z
n n n
x (x)dx = (1) n! = (1) n! (x)dx,

and one then has the general equivalence
xn (n) (x) = (1)n n!(x). (A.18)
This final relationship shows that the even-order derivatives of the delta
function are even functions and the odd-order derivatives are odd functions
of the argument.
It is often convenient to express the Dirac delta function in terms of the
Heaviside unit step function U (x) that is defined by the relations U (x) = 0
when x < 0, U (x) = 1 when x > 0. Consider the behavior of the derivative of
U (x). If, as before, a superscript prime denotes differentiation with respect
to the argument, one obtains formally upon integration by parts (with the
limits x1 < 0 and x2 > 0),
Z x2 Z x2
x2
f (x)U (x)dx = [f (x)U (x)]x 1
f (x)U (x)dx
x1 x1
Z x2

= f (x2 ) f (x)dx
0
= f (x2 ) [f (x2 ) f (0)]
= f (0),
where f (x) is any continuous function. Upon setting x = y a, and f (x) =

f (y a) = F (y), and then proceeding to the limits as x1 and
x2 +, the above result becomes
Z
F (y)U (y a)dy = F (a),

and the derivative U (x) is seen to satisfy the sifting property given in Eq.
(A.5). In particular, with F (y) = 1 and a = 0, this expression becomes
Z
U (y)dy = 1,

and U (x) also satisfies the property given in Eq. (A.2) which serves to par-
tially define the delta function. Moreover, U (x) = 0 for all x 6= 0 and property
(A.1) is also satisfied. Hence, one may identify the derivative of the unit step
function with the delta function, so that
dU (x)
(x) = . (A.19)
dx
In addition, it is seen that
Z x
U (x) = ()d, (A.20)

which follows from Eqs. (A.6) and (A.19).

The Dirac delta function may also be introduced through the use of the
Fourier integral theorem [3], which may be written as
Z Z
f (a) = d dx f (x)ei2(xa) (A.21)

for any sufficiently well-behaved, continuous function f (x). Define the func-
tion sequence
Z
K(x a, ) ei2(xa) d

sin (2(x a))
= (A.22)
(x a)
with limit
K(x a) lim K(x a, ). (A.23)

Strictly speaking, this limit does not exist in the ordinary sense when x a 6=
0; however, the limit does exist and has the value zero when x a 6= 0 if it
is interpreted in the sense of a Cesaro limit [4]. Upon inversion of the order
of integration, Eq. (A.21) may be formally rewritten as
Z
f (a) = f (x)K(x a)dx, (A.24)

which should be interpreted as meaning that

Z
f (a) = lim f (x)K(x a, )dx. (A.25)

Thus, the function K(x a) satisfies the sifting property (A.5) of the delta
function. If one sets f (x) = 1 and a = 0 in Eq. (A.24), there results
Z
K(x)dx = 1,

and K(x) satisfies the property given in Eq. (A.2) which serves to partially
define the delta function. Because K(x) = lim K(x, ) = 0 when x 6= 0,
so that the property given in Eq. (A.1) is also satisfied, one then obtains from
Eq. (A.23) the relation Z
(x) = ei2x d. (A.26)

That is, the Dirac delta function may be regarded as the Fourier transform
of unity. The reciprocal relation follows from Eq. (A.25) upon setting f (x) =
exp(i2x) and a = 0, so that
Z
1= (x)ei2x dx, (A.27)

which also follows directly from the sifting property given in Eq. (A.5). Notice
that this relation by itself is not sufficient to imply the validity of Eq. (A.26).
A.2 The Dirac Delta Function in Higher Dimensions
The definition of the Dirac delta function may easily be extended to higher-
dimensional spaces. In particular, consider three-dimensional vector space in
which case the defining relations given in Eqs. (A.1)(A.2) become
(r) = 0; r 6= 0, (A.28)
Z
(r)d3 r = 1. (A.29)

The function
(r) (x, y, z)
(x)(y)(z), (A.30)
where r = 1x x + 1y y + 1z z is the position vector with components (x, y, z)

clearly satisfies Eqs. (A.28)(A.29) and so defines a three-dimensional Dirac
delta function. The sifting property given in Eq. (A.5) then becomes
Z
f (r)(r a)d3 r = f (a), (A.31)

and the similarity relationship or scaling law given in Eq. (A.9) now states
that
A.2 The Dirac Delta Function in Higher Dimensions 105
1
(ar) = (r), (A.32)
|a|3
where a is a scalar constant. The Fourier transform pair relationship ex-
pressed in Eqs. (A.26)(A.27) becomes
Z
1
(r) = eikr d3 k, (A.33)
(2)3
Z
1= (r)eikr d3 r, (A.34)

where k = 1x kx + 1y ky + 1z kz = 2(1x x + 1y y + 1z z ).
The generalization of the three-dimensional Dirac delta function to more
general coordinate systems requires more careful attention. Suppose that a
function (r) is given in Cartesian coordinates as
(r) = (x)(y)(z) (A.35)
and it is desired to express (r) in terms of the orthogonal curvilinear coor-

dinates (u, v, w) that are defined by
u = f1 (x, y, z),
v = f2 (x, y, z), (A.36)
w = f3 (x, y, z),
where f1 , f2 , f3 are continuous, single-valued functions of x, y, z with a unique

inverse x = f11 (u, v, w), y = f21 (u, v, w), z = f31 (u, v, w). That is, an ex-
pression for (r) is desired in terms of the coordinate variables (u, v, w) that
satisfies the relation
Z
(r r )(u, v, w)dV = (u , v , w ), (A.37)

where dV is the differential volume element in u, v, w-space and (u , v , w )

is the point corresponding to (x , y , z ) under the coordinate transformation
given in Eq. (A.36). If the point r = (x, y, z) is varied from r to r + r1 by
changing the coordinate variable u to u + u while keeping v and w fixed,
then
r
r1 = u.
u
Similarly, if the point r = (x, y, z) is varied from r to r + r2 by changing the
coordinate variable v to v + v while keeping u and w fixed, then
r
r2 = v.
v
The parallelogram with sides r1 and r2 then has area
A = |A|
= |r1 r2 |

r r
= uv. (A.38)
u v
If the point r = (x, y, z) is now varied from r to r + r3 by changing the

coordinate variable w to w + w while keeping u and v fixed, then
r
r3 = w,
w
and the volume of the parallelepiped with edges r1 , r2 , and r3 is then
given by
V = |r3 (r1 r2 )|

r r r
= uvw. (A.39)
w u v
The quantity

x, y, z (x, y, z)
J
u, v, w (u, v, w)

r r r
(A.40)
w u v
is recognized as the Jacobian of the coordinate transformation of x, y, z with

respect to u, v, w. With this result for the differential element of volume, Eq.
(A.37) becomes
Z
x, y, z
(r r )(u, v, w) J

u, v, w dudvdw = (u , v , w ), (A.41)

from which it is immediately seen that

x, y, z
(u)(v)(w) = J (x)(y)(z). (A.42)
u, v, w
Because this transformation is assumed to be single-valued, then
(u)(v)(w)
(x)(y)(z) =
x,y,z
J u,v,w

u, v, w
= J (u)(v)(w), (A.43)
x, y, z
A.2 The Dirac Delta Function in Higher Dimensions 107
where J(u, v, w/x, y, z) is the Jacobian of the inverse transformation.

Consider finally the description of a function (r) that vanishes everywhere
in three-dimensional space except on a surface S and is such that
Z Z
3
(r)(r)d r = (r)(r)d2 r, (A.44)
S
where (r) is the value of (r) on the surface S, that is, when r S. Choose
orthogonal curvilinear coordinates (u, v, w) such that w = w0 describes the
surface S for some constant w0 , in which case w is parallel to the normal
to the surface S, and is such that u and v are both perpendicular to the
normal to the surface S. The differential element of area of the surface S is
then given by Eq. (A.38). Furthermore, both r/w and (r/u) (r/v)
are normal to S so that

x, y, z r r r
J = . (A.45)
u, v, w w u v
With this result, Eq. (A.44) may be written as

Z
r r r
Z
(r)(r)
dudvdw = (r)(r)d2 r,
w u v S
which, with Eq. (A.38), may be expressed as

Z
r 2
Z
(r)(r)
d rdw = (r)(r)d2 r. (A.46)
w S
From this result it then follows that

(r)
(r) = r (w w0 ), (A.47)

w
which is the solution of Eq. (A.44). This result can be simplified somewhat
by noting that when the variable w is varied while u and v are held fixed,
then the changes in r and w are related by w = w r, so that
r
w = 1.
w
Moreover, because both r/w and w are normal to the surface S described
by w = w0 , then

r
w |w| = 1,

and, as a result, Eq. (A.47) becomes

(r) = |w| (r)(w w0 ) (A.48)
as the solution to Eq. (A.44).
References
1. P. A. M. Dirac, The Principles of Quantum Mechanics. Oxford: Oxford University

Press, 1930. 15.
2. M. J. Lighthill, Introduction to Fourier Analysis and Generalized Functions. Lon-
don, England: Cambridge University Press, 1970.
3. E. C. Titchmarsh, Introduction to the Theory of Fourier Integrals. London: Oxford
University Press, 1939. Ch. I.
4. B. van der Pol and H. Bremmer, Operational Calculus Based on the Two-Sided
Laplace Integral. London: Cambridge University Press, 1950. pp. 100104.
Appendix B
Helmholtz Theorem
Because
2 1
= 4(R) (B.1)
R
where R = r r with magnitude R = |R| and where (R) = (r r ) =
(x x )(y y )(z z ) is the three-dimensional Dirac delta function
(see Appendix A), then any sufficiently well-behaved vector function F(r) =
F(x, y, z) can be represented as
Z
F(r) = F(r )(r r ) d3 r
V

1 1
Z
2
= F(r ) d3 r
4 V R
1 F(r ) 3
Z
= 2 d r, (B.2)
4 V R
the integration extending over any region V that contains the point r. With
the identity = 2 , Eq. (B.2) may be written as
1 F(r ) 3 1 F(r ) 3
Z Z
F(r) = d r d r. (B.3)
4 V R 4 V R
Consider first the divergence term appearing in this expression. Because

the vector differential operator does not operate on the primed coordinates,
then
F(r ) 3

1 1 1
Z Z

d r = F(r ) d3 r . (B.4)
4 V R 4 V R
Moreover, the integrand appearing in this expression may be expressed as
109
110 B Helmholtz Theorem

1 1
F(r ) = F(r )
R R

F(r ) 1
= + F(r ), (B.5)
R R
where the superscript prime on the vector differential operator denotes

differentiation with respect to the primed coordinates alone. Substitution of
Eq. (B.5) into Eq. (B.4) and application of the divergence theorem to the
first term then yields
F(r ) 3 F(r ) F(r ) 3

1 1 1
Z Z Z
3
d r = d r + d r
4 V R 4 V R 4 V R
1 1 1 F(r ) 3
I Z
= F(r ) nd2 r + d r
4 S R 4 V R
= (r), (B.6)
which is the desired form of the scalar potential (r) for the vector field
F(r). Here S is the surface that encloses the regular region V and contains
the point r.
For the curl term appearing in Eq. (B.3) one has that
F(r ) 3

1 1 1
Z Z
d r = F(r ) d3 r
4 V R 4 V R

1 1
Z
= F(r ) d3 r . (B.7)
4 V R
Moreover, the integrand appearing in the final form of the integral in Eq.
(B.7) may be expressed as
F(r ) F(r )

1
F(r ) = , (B.8)
R R R
so that
F(r ) 3 F(r ) 3 F(r )

1 1 1
Z Z Z
d r = d r d3 r
4 V R 4 V R 4 V R
1 F(r ) 3 1 1
Z I
= d r + F(r ) nd2 r
4 V R 4 S R
= a(r), (B.9)
which is the desired form of the vector potential.

The relations given in Eqs. (B.3), (B.6), and (B.9) then show that
F(r) = (r) + a(r), (B.10)

References 111
where the scalar potential (r) is given by Eq. (B.6) and the vector potential
a(r) is given by Eq. (B.9). This expression may also be written as
F(r) = F (r) + Ft (r), (B.11)
where
F (r) = (r)
1 F(r ) 3 1 F(r )
Z I
= d r + nd2 r (B.12)
4 V |r r | 4
S |r r |
is the longitudinal or irrotational part of the vector field (where F (r ) =

0), and where
Ft (r) = a(r)
1 F(r ) 3
Z
=
d r
4 V |r r |
1 F(r ) 3 1 F(r )
Z I
= d r + nd2 r(B.13)
4 V |r r | 4
S |r r |
is the transverse or solenoidal part of the vector field (where F (r ) = 0).

If the surface S recedes to infinity and if the vector field F(r) is regular at
infinity, then the surface integrals appearing in the above expressions vanish
and Eqs. (B.12)(B.13) become
F (r) = (r)
1 F(r ) 3
Z
=
d r, (B.14)
4 V |r r |
Ft (r) = a(r)
1 F(r ) 3
Z
= d r. (B.15)
4 V |r r |
Taken together, the above results constitute what is known as Helmholtz

theorem [1].
Theorem B.1. Helmholtz Theorem. Let F(r) be any continuous vector
field with continuous first partial derivatives. Then F(r) can be uniquely ex-
pressed in terms of the negative gradient of a scalar potential (r) and the
curl of a vector potential a(r), as embodied in Eqs. (B.10)(B.13).
References
1. H. B. Phillips, Vector Analysis. New York: John Wiley & Sons, 1933.
Appendix C
Greens Functions
The formal solution of either Poissons or Laplaces equation in a finite region

R of space with either Dirichlet or Neumann boundary conditions specified
on the boundary surface S of R can be obtained through the use of Greens
2 2
3
d2 r,
RRR H
second integral identity R d r = S n n
otherwise known as Greens theorem, and the so-called Greens functions. This
class of functions G(r, r ) is defined by the set of solutions of the equation
2 G(r, r ) = 4(r r ), (C.1)
where
1
G(r, r ) = + F (r, r ) (C.2)
|r r |
with the function F (r, r ) satisfying Laplaces equation within the interior of
the region R, so that
2 F (r, r ) = 0; r R. (C.3)
From a physical point of view, a Greens function produces a response at an

observation point r due to a unit point source at r .
Let (r ) = V (r ) be the scalar potential and let (r ) = G(r, r ) be the
Greens function in Greens theorem, which can then be solved for V (r) as
1
ZZZ
V (r) = (r )G(r, r )d3 r
40 R
V (r )
I
1 G(r, r )
+ G(r, r ) V (r ) d2 r . (C.4)
4 S n n
The freedom of choice afforded by the definition of the Greens function given
in Eqs. (C.1)(C.3) then allows one to make the surface integral appearing
in Eq. (C.4) depend only upon the type of boundary condition imposed.
113
114 C Greens Functions
For Dirichlet boundary conditions one demands that the Greens function
G(r, r ) = GD (r, r ) identically vanishes on the boundary surface S, so that
GD (r, r ) = 0, r S (C.5)
and Eq. (C.4) becomes
1 1 GD (r, r ) 2
ZZZ I
3
V (r) = (r )GD (r, r )d r V (r ) d r (C.6)
40 R 4 S n
This Greens function satisfies the symmetry property
GD (r , r ) = GD (r , r ) (C.7)
which represents the physical interchangeability of the source and field points.
Neumann
H boundary conditions are H a little more complicated due to Gauss
theorem [ S d = 4 if P S and S d = 0 if P / S, where the solid angle
is subtended at the point P ] which, when applied to Eq. (C.1), gives
G(r, r ) 2
ZZZ I
4 = 2 G(r, r )d3 r = d r.
R S n
As a consequence of this result, the simplest allowable boundary condi-

tion that can be imposed on the normal derivative of the Greens function
G(r, r ) = GN (r, r ) with Neumann boundary conditions is
GN (r, r ) 4
= , r S, (C.8)
n S
where S denotes the total surface area of the boundary surface S. With this
substitution Eq. (C.4) becomes
1 1 V (r ) 2
ZZZ I
3
V (r) = hV iS + (r )GN (r, r )d r + GN (r )
d r
40 R 4 S n
(C.9)
where hV iS = S1 S V (r )d2 r denotes the average value of the potential over
H
the entire boundary surface. The typical Neumann boundary value problem is
the so-called exterior problem in which the region R is bounded between two
surfaces, one closed and at a finite distance from the origin of coordinates for
the problem and the other at infinity. In this case the surface area S is infinite
so that the boundary condition (C.8) on the Greens function becomes homo-
geneous and the average value hV iS of the potential over S vanishes. Because
of the inhomogeneous nature of the Neumann boundary condition expressed
in Eq. (C.8), the symmetry property under an interchange of source and field
points is not automatic except in the homogeneous case of the exterior prob-
lem. Nevertheless, a symmetric Greens function can always be constructed
C Greens Functions 115
once GN (r , r ) has been determined by defining a modified Greens function
1 GN (r , r) 2
I
GN (r , r ) GN (r , r ) + GN (r , r) d r (C.10)
4 S n
which satisfies the Neumann boundary condition (C.8) by virtue of the

fact that GN (r , r)/n is independent of its first argument. This modified
Greens function then satisfies the symmetry property
GN (r , r ) = GN (r , r ). (C.11)
With these results a physical interpretation of the function F (r, r ) ap-

pearing in Eq. (C.2) may now be given. Because F (r, r ) is a solution of
Laplaces equation in the region R, it then represents the electrostatic po-
tential produced by a system of charges that are external to that region. For
either Dirichlet or Neumann boundary conditions, the function F (r, r ) can
then be regarded as the electrostatic potential due to an external distribu-
tion of charges that are chosen to satisfy either the homogeneous boundary
conditions of zero potential or zero normal derivative of the potential on the
boundary surface S, respectively, when combined with the potential produced
by a unit point charge at the source point r appearing in the Greens function
(C.2). This interpretation then leads to the so-called method of images ap-
proach to solving specific types of boundary value problems in electrostatics
that involve point charges in the presence of conducting bodies maintained
at specified potential values.
For example, if the closed surface S bounding the region R is grounded so
that V (r ) = 0 for all r S, then Eq. (C.6) becomes
1
ZZZ
V (r) = (r )GD (r, r )d3 r . (C.12)
40 R
This expression simply represents the principle of superposition as applied

to both the distribution of point sources in R with charge density (r )
and their corresponding image charges. Each elemental point source in R
3
1 (r )d r
contributes the differential element of potential dV1 (r) = 4 0 |rr | and
each elemental image charge contributes the differential element of potential
1 1
dV2 (r) = 4 0
(r )F (r, r )d3 r , where F (r, r ) + |rr
| = 0 when r S. The
expression given in Eq. (C.12) then embodies the method of images.

If there are no charge sources in the region R so that the potential satisfies
Laplaces equation 2 V (r) = 0 for all r R, then Eq. (C.6) becomes
1
I
V (r) = V (r ) GD (r, r ) n d2 r , (C.13)
4 S
where n denotes the unit normal vector to S directed out of the region

R. Physically, the normal derivative n GD (r, r ) = GD (r, r ) n of the
116 C Greens Functions
Greens function at the surface represents the surface charge density [see Eq.
(1.57)]
GD (r, r )
G (r ) = 0 = 0 GD (r, r ) n (C.14)
n
that would be induced on the boundary surface S by a unit point charge at
r = r if the boundary represented a grounded conductor. With this identifi-
cation, Eq. (C.13) becomes
1
I
V (r) = G (r )V (r )d2 r . (C.15)
40 S
This relation gives the solution to the potential problem corresponding to

a specified potential V (r ) on the boundary S of the region R in terms of
the surface integral of this potential multiplied by the charge density induced
on the grounded boundary by a unit point charge placed at the field point
r R. This expression also follows from Greens reciprocation theorem (see
problems 2.42.5 in Chapter 2)
ZZZ I ZZZ I
3 2 3
(r) (r)d r + s (r) (r)d r = (r)(r)d r + s (r)(r)d2 r
R S R S
(C.16)
where (r) is the electrostatic potential due to a volume charge density (r)
within R and a surface charge density s (r) on the surface S bounding R,
and (r) is the electrostatic potential due to another volume charge density
(r) in R and surface charge density s (r) on S.
References
Appendix D
Boundary Value Problems
The types of potential problems that are associated with Laplaces equation
differ from each other in the type of boundary conditions that are applied
either through the geometric shape of the boundary surface or the prescribed
behavior of the potential on that boundary surface (in the form of either
Dirichlet or Neumann boundary conditions). If the boundary surface S sur-
rounds and confines the field to a finite region R of space, then it is a closed
surface; if not, then it is an open surface and the field extends to infinity in
at least one direction. In either case, the mathematical form of the solution
is found to be simplest when a coordinate system 1 , 2 , 3 matching the
boundary surface is constructed (for example, 1 = k with k a constant).
Let = (1 , 2 , 3 ) denote the solution of Laplaces equation that satisfies
the prescribed boundary conditions over the specified boundary surface S.
The family of surfaces defined by
(1 , 2 , 3 ) = 0 (D.1)
define the nodal surfaces or nodes for (r). A set of solutions of Laplaces
equation having all of their nodal surfaces either coincident with or orthog-
onal to, for example, the 1 -coordinate surfaces occurs only for a select few
boundary surfaces . In those cases, the solutions must be of the form [1]
(1 , 2 , 3 ) = (1 )(2 , 3 ), (D.2)
where (1 ) is independent of both 2 and 3 while (2 , 3 ) is independent

of 1 . The zeros of (1 ) are then responsible for the nodal surfaces of being
coincident with the 1 -coordinate surfaces while the function (2 , 3 ) defines
a family of nodal surfaces orthogonal to the 1 -coordinate surfaces.
This line of analysis can be taken one step further by considering the even
more limited set of coordinate systems for which there exists a set of solutions
(1 , 2 , 3 ) of Laplaces equation with nodal surfaces that are all coincident
the the three families of coordinate surfaces 1 = k1 , 2 = k2 , 3 = k3 , where
k1 , k2 , k3 are constants. These solutions must then be of the form
117
118 D Boundary Value Problems
(1 , 2 , 3 ) = 1 (1 )2 (2 )3 (3 ), (D.3)
where each k (k ) is a function of the k -coordinate alone for k = 1, 2, 3. Such

solutions are said to be separated and the particular coordinate system for
which this occurs is said to be a separable coordinate system for that partial
differential equation. Because of its linearity, all solutions of Laplaces equa-
tion can then be constructed from linear combinations of the members of the
family of separated solutions. Similar remarks hold for the Helmholtz equa-
tion (2 + k 2 ) = 0 for which there are eleven different separable coordinate
systems in three-dimensional space [1]: rectangular, circular cylinder, elliptic
cylinder, parabolic cylinder, spherical, conical, parabolic, prolate spheroidal,
oblate spheroidal, ellipsoidal, and paraboloidal coordinates. There are addi-
tional coordinate systems that are only partially separable in the sense that
the solution may be expressed in the slightly more complicated form
(1 , 2 , 3 ) = R(1 , 2 , 3 )1 (1 )2 (2 )3 (3 ), (D.4)
where the modulation factor R(1 , 2 , 3 ) is independent of the separation

constants involved in obtaining the separated solutions.
D.1 Boundary Value Problems in Rectangular

Coordinates
Laplaces equation in three-dimensional rectangular coordinates
2 2 2
+ 2 + 2 =0 (D.5)
x2 y z
admits a separated solution of the form (x, y, z) = X(x)Y (y)Z(z). With this
substitution, Laplaces equation (C.21) becomes, after division by = XY Z,
1 d2 X 1 d2 Y 1 d2 Z
2
+ 2
+ = 0. (D.6)
X(x) dx Y (y) dy Z(z) dz 2
This equation can be satisfied for arbitrary values of the independent coor-
dinates x, y, z only if each term is separately constant, so that
1 d2 X 1 d2 Y 1 d2 Z
= 2 , = 2 , = 2, (D.7)
X(x) dx2 Y (y) dy 2 Z(z) dz 2
where the separation constants , , and satisfy the relation
2 + 2 = 2 . (D.8)
D.2 Boundary Value Problems in Two-Dimensional Angular Regions 119
2 1/2
If both and are real-valued, then so also is = 2 +

and the
elementary solutions of the three ordinary differential equations appearing in
2 1/2
Eq. (C.23) are then eix , eiy , and eiz = ei( + ) z . The potential
2
(x, y, z) can then be constructed from the elementary product solutions

1/2
= eix eiy ei( + 2 )
2
z
(D.9)
through linear superposition. Allowed values of the separation constants

and are then determined by imposing specific boundary conditions on the
potential. In doing this, notice that more appropriate forms of the elementary
solutions (e.g., as sine and cosine functions) may first be inferred from the
imposed symmetry properties of the boundary conditions.
D.2 Boundary Value Problems in Two-Dimensional

Angular Regions
An important problem of considerable practical interest to electrostatics is

that in which two conducting surfaces come together in such a way that the
local geometry about the line of intersection can be approximated as the
intersection of two half-planes. The general solution of this problem provides
insight into the local behavior of the electrostatic potential and field, as well as
the surface charge density, in the neighborhood of sharp corners and wedges.
2
=0
Fig. D.1 The local two-

dimensional polar co- P(r,)
ordinate geometry for r
Laplaces equation in an
angular region formed
by the intersection of x
two conducting half-plane
surfaces.
Laplaces equation in two-dimensional polar coordinates (r, )

1 2

1
r + 2 = 0. (D.10)
r r r r 2
admits a separated solution of the form (r, ) = R(r) () which, when

substituted in Eq. (C.26), leads to the equation
1 d2

r d dR
r + = 0. (D.11)
R(r) dr dr () d2
This equation can be satisfied for arbitrary values of the independent coor-
dinates r and only if each term is separately constant, so that
1 d2

r d dR
r = = 2, (D.12)
R(r) dr dr () d2
where 2 is the separation constant. The elementary solutions of these equa-

tions are R(r) = ar + br , () = A cos () + B sin () when 6= 0, and
R(r) = a0 + b0 ln (r), () = A0 + B0 when = 0.
For boundary value problems in which there is no restriction on the az-
imuthal angle , the separation constant must be a positive integer or zero
in order for the potential to be single-valued. Single-valuedness also requires
that B0 = 0. The general solution is then of the form1

X
an rn sin (n + n ) + bn rn sin (n + n ) .

(r, ) = a0 + b0 ln (r) +
n=1
(D.13)
If the origin is included in the domain of (r, ) and there is no charge at
r = 0, then all of the coefficients bn are zero.
For the two-dimensional angular region depicted in Fig. C.1, the azimuthal
angle is restricted to the domain 0 with 2, in which case
the angle is not periodic in its value. The boundary condition that
(r, 0) = (r, ) = V r 0 (D.14)
then requires that b0 = B0 = 0 and that b = A = 0 in the elementary

solutions of the separated equations given in Eq. (C.28), so that
X
(r, ) = a0 + a r sin (). (D.15)
>0
Application of the boundary condition given in Eq. (C.30) then shows that
a0 = V and that = m/, m = 1, 2, 3, . . . . The general solution for the
electrostatic potential in this case is then given by
1
This is a special case of the general solution developed in C.2.3.
D.2 Boundary Value Problems in Two-Dimensional Angular Regions 121

X
(r, ) = V + am rm/ sin (m/). (D.16)
m=1
The remaining undetermined coefficients am are specified by the imposed

behavior of the potential in the region r 0 removed from the corner.
2
= 2
= 3/2
(|s|)/(0a1) = 5/4
1
=
Fig. D.2 Radial de-

pendence of the relative = 3/4
surface charge density
( 1)
|s |/0 a1 r = /2
at a corner with surfaces
held at the fixed potential = /4
V > 0 for several values of 0
0 0.5 1.0
the angle of the corner,
as depicted in Fig. 1.9.. r
In the immediate neighborhood of the corner (r = 0) the electrostatic

potential is approximately given by the first term in the above series as
(r, ) V + a1 r/ sin (/) as r 0, (D.17)
with associated electric field components

Er (r, ) = a1 r( 1) sin (/), (D.18)
r
1
E (r, ) = a1 r( 1) cos (/), (D.19)
r
as r 0. The surface charge densities on the conductors at = 0 and =
are equal and are given by

s (r) = 0 E (r, 0) 0 a1 r( 1) as r 0. (D.20)

Both

electric field components and the surface charge density then vary as
r 1 as r 0, as illustrated in Fig. C.2. For small one has a very deep
corner, the power of r becomes very large, and there is no charge accumulation
in the corner. For a flat surface ( = ), both the surface charge density and
the electrostatic field become independent of r. When > the corner
becomes a two-dimensional wedge and both the field and the surface charge
density become singular as r 0. This singularity in s is integrable so that
the total surface charge within a finite distance from the edge remains finite.
D.3 Boundary Value Problems in Cylindrical

Coordinates: Cylinder Functions
In cylindrical polar coordinates (r, , z) defined by the transformation equa-

tions x = r cos , y = r sin , z = z with r [0, ), [0, 2), and
z (, +), Laplaces equation 2 = 0 assumes the form
1 2 2

1
r + 2 + 2 = 0. (D.21)
r r r r 2 z
The special case when the potential has either longitudinal invariance alone
[ = (r, ) is independent of z] or both longitudinal and axial invariance
[ = (r) is a function of r alone] has been considered in C.2.2 and is not
pursued any further here.
For the general case where = (r, , z), Laplaces equation admits a
separated solution of the form (r, , z) = R(r)Q()Z(z). With this substi-
tution, Laplaces equation (C.37) may be written as
1 d2 R 1 dR 1 d2 Q 1 d2 Z
2
+ + 2 2
= = k 2 , (D.22)
R dr rR dr r Q d Z dz 2
where k 2 is a separation constant. The ordinary differential equation (ode)

for Z(z) has the elementary solutions Z(z) = ekz . The remaining part of
Eq. (C.38) may then be written as
r2 d2 R r dR 1 d2 Q
2
+ + k 2 r2 = = 2, (D.23)
R dr R dr Q d2
where 2 is another separation constant. The ode for Q() has the elementary
solutions Q() = ei . In order that the potential be single-valued when the
domain of interest covers the full azimuthal range from 0 to 2, the separation
constant must be an integer. However, unless some boundary condition is
imposed in the z-direction, the separation constant k is arbitrary. For the
present, it is assumed that k is real and positive.
The remaining radial part of Eq. (C.39) may be written as
D.3 Boundary Value Problems in Cylindrical Coordinates: Cylinder Functions 123
d2 R 1 dR 2

2
+ + k R = 0. (D.24)
dr2 r dr r2
Under the change of variable x = kr, this equation assumes the standard
mathematical form
d2 R 2

1 dR
+ + 1 R=0 (D.25)
dx2 x dx x2
which is Bessels equation. Solutions are the Bessel functions of the first kind
of order with series representation (obtained by the method of Frobenius)

x X (1)j x 2j
J (x) = (D.26)
2 j=0
j! (j + 1 ) 2
which converges for all real, finite values of x, the Bessel functions of the sec-
ond kind of order (also known as either Webers or the Neumann function)
J (x) cos () J (x)

Y (x) = (D.27)
sin ()
where the right-hand side of this equation becomes indeterminate and is

replaced by its limiting value Ym (x) = limm Y (x) when is an integer or
zero, and the Bessel functions of the third kind
H(j) = J (x) + (1)j iY (x) (D.28)
for j = 1, 2. It follows from lHopitals rule that when is an integer or zero,

( )

cos ()J (x) J (x)
Ym (x) = lim
m cos ()

1 J (x) J (x)
= lim (1)m ,
m
which then leads to the series representations

j 2j Xj
2 h x i X (1) x 1
Y0 (x) = ln + J0 (x) , (D.29)
2 j=1
(j!)2 2 n=1
n
h x i
Ym (x) = 2 ln + Jm (x)
2
x m m1
X (m j 1)! x 2j x m 1 X m
1

2 j=0
j! 2 2 m! n=1
n

j m+j
x m X (1)j x 2j X 2 X 1
+ , (D.30)
2 j=1
j!(m + j)! 2 n=1
n n=j+1 n
where m > 0 is an integer and limn 1 + 12 + 31 + + n1 ln (n) is

Eulers constant. Because of the symmetry relations Jm (x) = (1)m Jm (x)

and Ym (x) = (1)m Ym (x) when m is an integer, Eqs. (C.42) and (C.46)
can then be used for all positive and negative integer indices.
The linear independence of the Bessel functions Jm (x) and Ym (x) of the
first and second kind of integer order m follows from their respective limiting
behaviors as x 0. For m = 0, J0 (x) 1 while Y0 (x) as ln(x/2).
When m > 0, Jm (x) 0 as xm while Ym (x) as xm . Therefore,
the general solution of Bessels equation (C.41) when = m is an integer has
the form
R(x) = AJm (x) + BYm (x), (D.31)
where A and B are arbitrary constants. It is then seen that the Hankel func-
(1) (2)
tions H (x) and H (x) defined in Eq. (C.44) form a linearly independent
set of solutions to Bessels equation for all .
Notice that the major difference between Bessels differential equation
(C.41) and the simple harmonic equation for the sinusoidal functions cos x
and sin x is due to the term x1 dR
dx which has a significant impact on the behav-
ior of the solution as x 0. However, Bessels equation becomes similar to
the simple harmonic equation as x . In particular, the Bessel functions
of the first kind of order possess the asymptotic approximations
r
2
J (x) cos x , (D.32)
x 2 4
r
2
Y (x) sin x , (D.33)
x 2 4
as x . The transition from the small x behavior given by the first few
terms of Eqs. (C.42) and (C.43) to the large x asymptotic behavior given in
Eqs. (C.48) and (C.49), respectively, occurs in the region x .
Finally, by direct differentiation of the series representations given in Eqs.
(C.42) and (C.45)(C.46), one arrives at the set of recurrence relations
2
C1 (x) + C+1 (x) = C (x), (D.34)
x
dC (x)
C1 (x) C+1 (x) = 2 , (D.35)
dx
D.3 Boundary Value Problems in Cylindrical Coordinates: Cylinder Functions 125
dC (x)
= C1 (x) C (x), (D.36)
dx x
dC (x)
= C+1 (x) + C (x), (D.37)
dx x
(1)
where C (x) denotes any one of the cylinder functions J (x), Y (x), H (x),
(2)
H (x) of order or any linear combination of these functions with coeffi-
cients independent of both and x.
The alternate choice of separation constant in Eq. (C.38) is +k 2 which
leads to the equation d2 Z/dz 2 + k 2 Z = 0 with elementary solutions Z(z) =
eikz . The resultant separated ode for Q() remains unaltered with this
change so that the elementary solutions Q() = ei still apply with an
integer when covers the full azimuthal domain from 0 to 2. The radial
part of the equation then becomes [cf. Eq. (C.40)]
d2 R 1 dR 2

2
+ k + 2 R = 0. (D.38)
dr2 r dr r
The particular solutions of this equation are seen to be the Bessel functions
J (ikr) and Y (ikr) of an imaginary argument. Under the change of variable
x = kr this equation assumes the standard form
d2 r 2

1 dR
+ 1 + R=0 (D.39)
dx2 x dx x2
Real-valued solutions of this equation are the modified Bessel functions de-
fined as

x X x 2j
1
I (x) ei/2 J (ix) = (D.40)
2 j=0
j! (j + + 1) 2
where In (x) = In (x) with n an integer, and
I (x) I (x)
K (x) (D.41)
2 sin ()
where K (x) = K (x). The limiting forms of the modified Bessel functions
1
as x 0 are given by I (x) (+1) ( x2 ) , K0 (x) I0 (x) ln ( x2 ), and
K (x) 12 ()( x2 ) for > 0. In addition, their large argument asymptotic
approximations are given by
1
I (x) ex , (D.42)
2x
r
x
K (x) e , (D.43)
2x
as x . It is then seen that the modified Bessel functions I (r) are appro-
priate for boundary value problem solutions that remain bounded at r = 0,
while the modied Bessel functions K (r) are appropriate for solutions that
remain bounded as r .
The elementary solutions of Laplaces equation in cylindrical coordinates
are then given in separated form as
n (r, , z) = Rn (r)Q ()Zn (z), (D.44)
where the indices appearing in each of the separated solutions are related
to the separation constants introduced in obtaining this elementary solution.
Both the solution domain and the imposed boundary conditions then deter-
mine the appropriate from of each separated solution as well as the allowed
values of the indices and n. The full solution of Laplaces equation in the
specified solution domain that satisfies all of the boundary conditions is then
obtained through an appropriate superposition of these selected elementary
solutions as XX
(r, , z) = An n (r, , z), (D.45)
n
where the coefficients An are uniquely determined by the boundary condi-

tions.
D.4 Boundary Value Problems in Spherical Coordinates:

Legendre Polynomials and Spherical Harmonics
In spherical coordinates (r, , ) defined by the set of transformation equa-

tions x = r sin cos , y = r sin sin , z = r cos with r [0, ), [0, 2),
and [0, ], Laplaces equation 2 = 0 assumes the form
1 2 (r) 2

1 1
+ sin + 2 = 0. (D.46)
r r2 r2 sin r2 sin 2
This equation then admits separated solutions of the form
1
(r, , ) = U (r)P ()Q(), (D.47)
r
where the factor r1 is explicitly displayed in order to reflect the form of the
electrostatic potential for a spherical charge distribution. With this substitu-
D.4 Boundary Value Problems in Spherical Coordinates: Legendre Polynomials and Spherical Harmonics
127
tion, Laplaces equation (C.62) becomes
1 d2 U 1 d2 Q

1 d dP
r2 sin2 + sin = = m2 (D.48)
U dr2 P r2 sin d d Q d2
where m2 is the separation constant. The ode for Q() then has the elemen-
tary solutions Q() = eim . In order that Q() be single-valued when the
full azimuthal range = 0 2 is allowed, the separation constant m must
be an integer or zero. The remaining part of Eq. (C.64) may then be written
as
r2 d2 U m2

1 d dP
= sin + = ( + 1), (D.49)
U dr 2 P sin d d sin2
where ( + 1) is another separation constant. The ode for the radial part of
this separated equation then has the elementary solution
U (r) = Ar+1 + Br , (D.50)
where A and B are constants of integration.

The remaining angular part of Eq. (C.65) is then
m2

1 d dP ()
sin + ( + 1) P () = 0, (D.51)
sin d d sin2
where is as yet undetermined. With the change of variable = cos , for
which d = sin ()d so that d/d = sin ()d/d, this ode assumes the
standard form
m2

d dP
1 2 + ( + 1) P =0 (D.52)
d d 1 2
known as the generalized or associated Legendre equation, whose solutions

are the associated Legendre functions. In order that its solutions represent a
physically realizable potential, they must be single-valued, finite, and contin-
uous on the interval [1, 1].
D.4.1 Legendres Equation and the Legendre

Polynomials
Consider first the special case when the problem possesses azimuthal symme-
try so that there is no -dependence and m = 0. The generalized Legendre
equation (C.68) then simplifies to the ordinary Legendre differential equation

d dP
1 2 + ( + 1)P = 0 (D.53)
d d
Solutions of this equation are the Legendre polynomials of order with series
representation (obtained by the method of Frobenius)
[/2]
X (2 2j)!
P () = (1)j 2j (D.54)
j=0
2 j!( j)!( 2j)!
which are normalized such that P (1) = 1. Here [/2] denotes the greatest
integer value of /2, where [/2] = /2 if is even and [/2] = ( 1)/2
if is odd. Explicit expressions for the first few Legendre polynomials are
1 2 1 3

P0 () = 1, P1 () = , P2 () = 2 3 1 , P 3 () =
2 5 3 , P 4 () =
1 4 2 1 5 3

8 35 30 + 3 , P 5 () = 8 63 70 + 15 , and so-on for all higher-
order polynomials.
The series representation (C.70) of the Legendre polynomials may be
rewritten as
[/2]
1 d X !
P () =
(1)j 2(j) .
2 ! d j=0 j!( j)!
The summation limit of [/2] can now be extended to because the power
of 2(j) is less than after the [/2] term and so its th -order derivative
vanishes. With this replacement, the summation is seen to be the expansion

of the quantity 2 1 , so that
1 d
P () =
2 1 (D.55)
2 ! d
which is known as Rodrigues formula.

The Legendre polynomials P () form a complete orthogonal set of func-
tions on the interval [1, 1], satisfying the orthogonality relation
1
2
Z
P ()P ()d = (D.56)
1 2 + 1
Any sufficiently well-behaved function f () on the interval [1, 1] can

then be expanded in a Legendre series representation as

X
f () = A P (), 1 1, (D.57)
=0
with expansion coefficients

129
1
2 + 1
Z
A = f ()P ()d. (D.58)
2 1
D.4.2 Spherical Coordinate Boundary Value Problems

with Azimuthal Symmetry
For a given boundary value problem for Laplaces equation in spherical co-
ordinates which possesses azimuthal symmetry (i.e., is independent of the
azimuthal angle ), Eq. (C.64) demands that m = 0. The general solution
for the potential is then given by

X
(r, ) = A r + B r(+1) P (cos ). (D.59)
=0
The coefficients A and B are then determined by the imposed boundary

conditions on some prescribed surface.
As an example, suppose that the electrostatic potential is specified as
V () on the surface of a sphere of radius a, and it is required to determine
the potential within the spherical region bounded by that surface. If there is
no charge at the origin, then the potential must be finite there and B = 0
for all . The expansion given in Eq. (C.75) then becomes

X
(r, ) = A r P (cos ), r a. (D.60)
=0
The coefficients A are then determinedPby evaluating Eq. (C.76) on the

surface of the sphere, so that V () = =0 A a P (cos ). This is just a
Legendre series representation of the form given in Eq. (C.73) with = cos ,
so that the coefficients are given by
2 + 1
Z
A = V ()P (cos ) sin d. (D.61)
2a 0
Suppose next that the electrostatic potential is to be determined in the region

external to the sphere. Because the potential must now be finite at r = ,
it is required that A = 0 for all . The expansion given in Eq. (C.75) then
becomes
X B
(r, ) = P (cos ), r a. (D.62)
r+1
=0
As in the previous case, the coefficients B are determined

P by evaluating this
expression on the surface of the sphere as V () = =0 aB +1 P (cos ), so that


2 + 1 +1
Z
B = a V ()P (cos ) sin d. (D.63)
2 0
The Legendre series representation given in Eq. (C.75), with coefficients

A and B determined by the specified boundary conditions, is a unique
expansion of the potential (r, ). This uniqueness property then provides
a convenient means by which the solution of some electrostatic potential
problem may be obtained from a knowledge of the potential along the axis
of symmetry. Along the positive z-axis, z = r and cos = 1, so that

X B
(z) = A r + +1 , z 0, (D.64)
r
=0
while along the negative z-axis, z = r and cos = 1, so that

X B
(z) = (1) A r + +1 , z 0. (D.65)
r
=0
If the electrostatic potential (z) can be determined at any point z on the

symmetry axis, and if this potential can be expressed in a power series in
z = r of the form given above with known coefficients, then the solution for
the potential at any point in space is obtained simply by multiplying each
power of r and r(+1) by the Legendre polynomial P (cos ).
An expansion of considerable practical importance is that of the potential
(r, r ) = |r r |1 at a field point r due to a unit point charge at r . The
Legendre series representation for this potential is then obtained by first
rotating the coordinate axes so that the z-axis lies along the position vector
r to the unit point source. The potential (r, r ), which satisfies Laplaces
equation everywhere except at the point r = r , then possesses azimuthal
symmetry and can therefore be expanded in the form given in Eq. (C.75) as

1 X
B
= A r + P (cos ), r 6= r , (D.66)
|r r | r+1
=0
where is the angle between the vectors r and r . If the point r is on the
positive z-axis, then the right-hand side of Eq. (C.82) reduces to the form
given in Eq. (C.80) while the left-hand side becomes
1 1 1

=
|r r | 2 2
(r + r 2rr cos )1/2 |r r |
as 0. Let r< denote the smaller of r and r , and let r> denote the larger
of r and r . One then has the expansion
131

1 1 1 1 1 X r<
= = = .
|r r | r> r< r> 1 r< /r> r> r>
=0
For points off of the z-axis it is then only necessary to multiply each term in
this series expansion by the Legendre polynomial P (cos ), resulting in the
general Legendre series representation
X r
1 1 <
= = P (cos )
+1
(D.67)
|r r | 2 2
(r + r 2rr cos )
1/2
r
=0 >
which is a useful expansion of the Greens function (r, r ) = |r r |1 in

spherical coordinates.
D.4.3 Associated Legendre Functions and the Spherical

Harmonics
The general potential problem in spherical coordinates can possess azimuthal

dependency so that one can no longer set m = 0 in Eq. (C.68); however,
in order that the elementary solution Q() = eim of Laplaces equation
(C.62) be single-valued in this general situation, m must then be an integer.
It is then necessary to determine the solution of this associated Legendre
equation, which may be written as
2
m2

2 d P dP
(1 ) 2 2 + ( + 1) P = 0, (D.68)
d d 1 2
for arbitrary values of and arbitrary integer values of m. For its solution,
one defines
d
P () (1)m (1 2 )m/2 m u(),
d
in which case the associated Legendre equation becomes
dm 2

2 d u du
(1 ) 2 2 + ( + 1)u = 0. (D.69)
d m d d
The expression appearing inside the square brackets of this equation is pre-
cisely the ordinary Legendre differential equation [cf. Eq. (C.69)] with a
nonnegative integer, and so its solution is the Legendre polynomial P (). The
appropriate solution of the associated Legendre differential equation (C.84)
is then given by
dm
Pm () = (1)m (1 2 )m/2 P () (D.70)
d m
for positive integer values of m. If Rodrigues formula (C.71) is used to rep-

resent P (), an expression for the associated Legendre functions Pm () that
is valid for both positive and negative integer values of m is obtained as
d+m 2
Pm () = (1)m (1 2 )m/2 ( 1) (D.71)
d +m
from which it is seen that P0 () = P (). The solutions given by Eq. (C.87)
will be finite on the closed interval 1 1 provided that (1) is either
zero or a positive integer and (2) that m can only take on the values
m = , + 1, + 2, . . . , 1, 0, 1, . . . , 2, 1, .
Because the defining differential equation (C.84) depends only upon m2 and
m can only take on positive or negative integer values, it is seen that Pm ()
and Pm () are proportional. From Rodrigues formula, it is found that
( m)! m
Pm () = (1)m P () (D.72)
( + m)!
For a given fixed value of m, the associated Legendre functions Pm ()

form an orthogonal set on the closed interval 1 1, satisfying the
orthogonality relation
1
2 ( + m)!
Z
Pm m
()P ()d = (D.73)
1 2 + 1 ( m)!
Because of Eq. (C.88), this orthogonality relation holds for both positive
and negative integer values of m. In addition, it reduces to the orthogonality
relation given in Eq. (C.72) for the Legendre polynomials when m = 0.
The solution of Laplaces equation (C.62) in spherical coordinates by the
separation of variables method assumed a product of single variable func-
tions of the three coordinate variables r, , of the form given in Eq. (C.63).
It is convenient to combine the angular functions Qm () and Pm () of this
solution in such a way so as to construct a set of orthogonal functions over
the unit sphere. Such a set of functions is called the set of spherical har-
monics 2 . The exponential functions Qm () = eim form a complete set of
orthogonal functions in the index m on the angular interval 0 < 2, and
the associated Legendre functions Pm () form a complete set of orthogonal
functions in the index for each allowed value of m on the interval 0 .
Their product Pm ()Qm () thus forms a complete orthogonal set of func-
2
This terminology is also applied to the solutions of the generalized Legendre equa-
tion (C.68) alone. In that case the product functions Pm ()Qm () considered here
are referred to as tesseral harmonics.
133
tions on the surface of a unit sphere in the two indices and m. From the
orthogonality relation (C.89), this normalized set of functions is given by
s
(2 + 1)( m)! m
Ym (, ) P (cos )eim (D.74)
4( + m)!
q
2+1
where Y0 (, ) = 4 P (cos ). The spherical harmonics satisfy the sym-
metry relation
Y,m (, ) = (1)m Ym

(, ). (D.75)
the orthonormalization condition
Z 2 Z
d sin d Y m (, )Ym (, ) = m m (D.76)
0 0
and the completeness relation
X
X

Ym ( , )Ym (, ) = ( )(cos cos ) (D.77)
=0 m=
D.4.4 Boundary Value Problems in Spherical

Coordinates
Any sufficiently well-behaved function g(, ) that is defined on the unit

sphere can be expanded in terms of the spherical harmonics as
X
X
g(, ) = Am Ym (, ), (D.78)
=0 m=
where the expansion coefficients are found from the orthonormalization con-
dition (C.92) as I

Am = g(, )Ym (, )d, (D.79)
where d = sin dd denotes the differential element of solid angle, the

integration being taken over the entire surface of the unit sphere.3 From Eqs.
(C.63), (C.66), and (C.94), the general solution for a given boundary value
problem for Laplaces equation in spherical coordinates is given by
3
Notice that all terms in the spherical harmonic expansion given in Eq. (C.94) with
m 6= 0 vanish at = 0. In thatqcase, this expansion takes
q on the special limiting form
2+1
= A0 with A0 = 2+1 g(, )P (cos )d.
P H
g(0, ) = g(, ) =0 =0 4 4
X

X bm
(r, , ) = am r + +1 Ym (, ). (D.80)
r
=0 m=
If the potential is specified on a spherical surface, the expansion coefficients

am and bm can then be determined by evaluating this expansion on the
spherical surface which reduces it to the form given in Eq. (C.94). For the in-
terior problem one typically requires that bm = 0 , m so that the potential
remains finite at the origin whereas for the exterior problem one typically
requires that am = 0 , m so that the potential vanishes at r = . In
either case, the appropriate expansion coefficients are then obtained through
application of Eq. (C.95).
D.4.5 The Addition Theorem for the Spherical

Harmonics
The addition theorem for the spherical harmonics, which is of considerable

mathematical importance, is now briefly considered. To begin, consider two
position vectors r and r from the same origin O that are separated by an
angle , the first having spherical coordinates (r, , ) and the second having
coordinates (r , , ). Because
cos = cos cos + sin sin cos ( ),
it is then of interest to express the Legendre polynomial P (cos ) in terms of

independent spherical harmonics of the angles , and , . If r is consid-
ered to be fixed in space, then P (cos ) is a function of the angles , with
the angles , as parameters. It may then be expanded in a series of the
P P
form given in Eq. (C.94) as P (cos ) = =0 m= A m ( , )Y m (, ).
However, because P (cos ) is a spherical harmonic of order alone, this ex-
pansion simplifies to

X
P (cos ) = Am ( , )Ym (, ), (D.81)
m=
with expansion coefficients (use of the limiting form described in footnote 10

is helpful in obtaining the final form of this result)
4
I
Am ( , ) = P (cos )Ym

(, )d = Y ( , ). (D.82)
2 + 1 m
Combination of these two results then yields the addition theorem for the
spherical harmonics
References 135

4 X

P (cos ) = Ym ( , )Ym (, ) (D.83)
2 + 1
m=
where cos = cos cos + sin sin cos ( ). If the angle goes to zero
so that = and = , then Eq. (C.99) immediately yields the sum rule

2 + 1
|Ym (, )|2 =
X
(D.84)
4
m=
The addition theorem (C.99) for the spherical harmonics can now be used to
express the expansion (C.83) of the Greens function (r, r ) = |r r |1 into
its most general form as
X

1 X 1 r<
= 4 Y ( , )Ym (, )
+1 m
(D.85)
|r r | 2 + 1 r>
=0 m=
where r< denotes the smaller of r and r and r> denotes the larger of r
and r . This expression gives the potential at the point r due to a unit point
charge at the point r in a completely factorized form in terms of the spherical
coordinates of r and r . This factorization is useful in any integration over
a specified charge density where one set of variables are the variables of
integration and the other set is the coordinate set of the field point.
References
1. P. M. Morse and H. Feshbach, Methods of Theoretical Physics. New York:

McGraw-Hill, 1953. Vol. I.
Index
absolute potential, 14 diamagnetic, 68

action-at-a-distance, 8, 61 dielectric
addition theorem for the spherical perfect, 29
harmonics, 134 simple, 29
Amperes law, 62 dielectric permittivity
amplitude modulation (AM), 93 static , 33
antiferromagnetic, 69 dipole
associated Legendre equation, 131 magnetic, 66
associated Legendre functions Pm (), Dirac delta function (), 97
132 composite function theorem, 99
attenuation factor, 91 derivatives of, 101
extension to higher dimensions, 104
Bessel functions, 123 sifting property, 98
Bessels equation, 123 similarity relationship (scaling law),
Biot and Savart law, 59 99
boundary conditions Dirichlet boundary conditions, 15, 114
Dirichlet, 15, 114 divergence theorem, 11
electrostatic, 34, 72
mixed, 15 Earnshaws theorem, 21
Neumann, 15, 114 electric conductivity
static 0 , 55
capacitance, 41, 42 electric current, 53
coefficients of, 40 electric dipole, 21
capacitor, 41 electric displacement vector
capacity coefficients, 40 simple dielectric, 33
characteristic impedance Z0 , 92 electric field intensity, 8
charge density, 9 electric polarization density
coefficients of capacitance cjj , 40 macroscopic, 30
coefficients of potential pjk , 38 electric resistance, 56
conduction current, 53 electric susceptibility e , 33
conductor electromagnetic induction, 72
ideal, 18 electromotive force (emf), 72
conservation of charge, 54 electron mobility, 55
conservative vector field, 14 electrostatic induction coefficients cjk ,
convection current, 53 40
Coulombs law, 7 electrostatic shielding, 20
current density J, 54 equation of continuity, 55
137
138 Index
equipotential surface, 14 method of images, 115

Eulers constant , 124 microscopic charge density, 9
mixed boundary conditions, 15
Faradays law, 72 modified Bessel functions, 125
for an electrostatic field, 12 mutual inductance, 74
Faradays paradox, 73
ferrimagnetic, 69 nepers N p, 91
ferromagnetic, 69 Neumann boundary conditions, 15, 114
Fourier integral theorem, 103 Neumann function, 123
Neumanns formula, 75
gauge transformation nodal surface, 117
magnetostatic field, 63
Gauss law Ohms law, 55
electric field, 10
Greens first integral identity, 16 paramagnetic, 69, 71
Greens functions, 113 perfect dielectric, 29
Greens reciprocation theorem, 49, 116 permutation symbol ijk , 65
Greens second integral identity, 113 phase velocity, 93
Greens theorem, 113 point dipole, 23
group velocity, 95 Poissons equation, 14
polarization charge density, 33
Holder condition, 15 potential
Heaviside unit step function U (), 102 absolute, 14
Helmholtz equation, 91 principle of superposition, 8
Helmholtz theorem, 111 propagation factor, 91
hole mobility, 55 pulse modulation (PM), 93
hysteresis, 72
relative permittivity, 34
ideal conductor, 18 Rodrigues formula, 128
inductance L, 74
irrotational vector field, 14 scalar potential
electrostatic field, 12
Joules law, 57 separable coordinate system, 118
simple dielectric, 29
Kelvin functions, 83 skin depth, 83
Laplaces equation, 14 skin effect, 82
Legendre differential equation skin resistivity, 86
associated, 127, 131 solid angle, 10
ordinary, 127 spherical harmonics Ym (, ), 132
Legendre polynomials P (), 128 addition theorem, 134
lines of force, 14 sum rule, 135
Stokes theorem, 14
macroscopic charge density, 9 sum rule for the spherical harmonics,
macroscopic magnetic moment density, 135
69 surface polarization charge density sb ,
magnetic dipole vector potential, 66 31, 33
magnetic intensity vector, 71 surface resistivity Rs , 86
magnetic moment density, 65
magnetic permeability telegraphers equations, 90
free space 0 , 59 tesseral harmonics, 132
magnetic permeability , 72 Thomsons theorem, 46
magnetic susceptibility m , 71 transmission line equations, 90
magnetization, 65, 69 volume polarization charge density b ,
magnetization current density, 71 31, 33
magnetization surface current density,
71 Webers function, 123
mean value theorem, 17

EE141 Text

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

EE141 Text

Hochgeladen von

Copyright:

Verfügbare Formate

Kurt Edmund Oughstun

Electric & Magnetic Field Theory

2 The Electric Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4 The Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

5 Transmission Line Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

A The Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

B Helmholtz Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

C Greens Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

D Boundary Value Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

D.4.3 Associated Legendre Functions and the Spherical

D(r, t) = (r, t), (1.1)

in MKSA units. The first two equations are mathematical statements of

The divergence of Amperes law (1.4) followed by substitution from Gauss

J(r, t) + (r, t)/t, (1.5)

which is the differential expression of the conservation of charge. This re-

F(r, t) = (r, t)E(r, t) + J(r, t) B(r, t), (1.6)

E(r, t) = 0 H(r, t)/t, (1.7)

As stated by Einstein regarding Maxwells theory of electromagnetism in

In regard to the contribution by Lorentz to this theory, Einsten [6] went on

The word ponderomotive is an adjective describing the tendency to produce

Einstein then concluded that [6]

1. A. M. Ampere, Memoir on the mutual action of two electric currents, Annales

2. M. Faraday, Experimental Researches in Electricity. London: Bernard Quaritch,

The beginning is the most important part of the work. Plato

The electric field produced by a specified static distribution of charges is

2.1 Coulombs Law and the Electric Field Intensity

Here 0 8.85421012F/m is the permittivity of free space, where the Farad

with rk1 r1 rk , where rk1 r1k /r1k and rk1 = |r1k |.

where Rj = r rj denotes the vector from the source point Pj at rj to the

2.1.1 Charge Density

(r) = hh(r)ii (2.8)

where R = r r is the vector directed from the source point r to the

2.2 Gauss Law

Consider a point charge q at a fixed point in space together with a simply-

Fig. 2.1 A point charge

where V is the volume enclosed by the surface S.

where (r rj ) (x xj )(y yj )(z zj ) is the three-dimensional Dirac

the integral form of Gauss law given in Eq. (2.13) becomes

which is the differential form of Gauss law .

the gradient operator operates only on the unprimed coordinates, then

which is Faradays law for the electrostatic field.

2.3 The Electrostatic Scalar Potential and Work

E(r) = V (r) (2.18)

where the minus sign is introduced by convention, and

The convergence properties of the integral appearing in Eq. (2.19) are

which is referred to as the absolute potential .

2.3.1 Poissons Equation for the Scalar Potential

Because 2 |r r |1 = 4(r r ), the integral solution of Poissons

equation is given by Coulombs law as [cf. Eq. (2.19)]

Definition 2.1 (Holder Condition). A function f (Q) of the coordinates

where r is the distance between the points P and Q.

If a region R exists in which f (Q) satisfies a Holder condition at every point

2.3.1.1 Boundary Value Problems & Uniqueness

r R, satisfies Laplaces equation in R; that is 2 V (r) = 0 r R and

where V /n = V n. As 2 V = 0 in R and either V = 0 or V /n = 0

2.3.1.2 Average Electrostatic Potential Over a Sphere

Consider a spherical surface S of radius a > 0 carrying a uniform surface

which then results in the geometrical identity

Theorem 2.2. The average value of 1r over a spherical surface S, where r is

If the surface charge density s = Q/4a2 is removed and a point charge Q

Corollary 2.1. It is impossible to have a potential maximum or minimum

Consider again a spherical surface S of radius a carrying a uniform surface

which results in the geometrical identity