MIT OpenCourseWare http://ocw.mit.

edu

Electromagnetic Field Theory: A Problem Solving Approach

For any use or distribution of this textbook, please cite as follows:

Markus Zahn, Electromagnetic Field Theory: A Problem Solving

Approach. (Massachusetts Institute of Technology: MIT
OpenCourseWare). http://ocw.mit.edu (accessed MM DD, YYYY).
License: Creative Commons Attribution-NonCommercial-Share Alike.

For more information about citing these materials or our Terms of

Use, visit: http://ocw.mit.edu/terms.
chapter 3

polarizationand
conduction
136 Polarizationand Conduction

The presence of matter modifies the electric field because

even though the material is usually charge neutral, the field
within the material can cause charge motion, called conduc-
tion, or small charge displacements, called polarization.
Because of the large number of atoms present, 6.02 x 1023 per
gram molecular weight (Avogadro's number), slight
imbalances in the distribution have large effects on the fields
inside and outside the materials. We must then self-
consistently solve for the electric field with its effect on charge
motion and redistribution in materials, with the charges.
resultant effect back as another source of electric field.

3-1 POLARIZATION

In many electrically insulating materials, called dielectrics,

electrons are tightly bound to the nucleus. They are not
mobile, but if an electric field is applied, the negative cloud of
electrons can be slightly displaced from the positive nucleus,
as illustrated in Figure 3-la. The material is then said to have
an electronic polarization. Orientational polarizability as in
Figure 3-1 b occurs in polar molecules that do not share their

No field
Electric field E

l \ +q
_E : F =qE
-4-
\ / d Torque = dx qE
=pxE

F = -qE
p = qd
Electronic polarization Orientation and ionic polarization

Figure 3-1 An electric dipole consists of two charges of equal magnitude but opposite
sign, separated by a small vector distance d. (a) Electronic polarization arises when the
average motion of the electron cloud about its nucleus is slightly displaced. (b) Orien-
tation polarization arises when an asymmetric polar molecule tends to line up with an
applied electric field. If the spacing d also changes, the molecule has ionic polarization.
 Polarization 137

electrons symmetrically so that the net positive and negative

charges are separated. An applied electric field then exerts a
torque on the molecule that tends to align it with the field.
The ions in a molecule can also undergo slight relative dis-
placements that gives rise to ionic polarizability.
The slightly separated charges for these cases form electric
dipoles. Dielectric materials have a distribution of such
dipoles. Even though these materials are charge neutral
because each dipole contains an equal amount of positive and
negative charges, a net charge can accumulate in a region if
there is a local imbalance of positive or negative dipole ends.
The net polarization charge in such a region is also a source
of the electric field in addition to any other free charges.

3S--1 The Electric Dipole

The simplest model of an electric dipole, shown in Figure

3-2a, has a positive and negative charge of equal magnitude q
separated by a small vector displacement d directed from the
negative to positive charge along the z axis. The electric
potential is easily found at any point P as the superposition of
potentials from each point charge alone:
q q
V= (1)
41reor+ 41reor-
The general potential and electric field distribution for any
displacement d can be easily obtained from the geometry
relating the distances r, and r- to the spherical coordinates r
and 0. By symmetry, these distances are independent of the
angle 4. However, in dielectric materials the separation
between charges are of atomic dimensions and so are very
small compared to distances of interest far from the dipole.
So, with r, and r_ much greater than the dipole spacing d, we
approximate them as
d
r+ - r -- cos
lim (2)
r~d d
r_=r+- cos 0
2
Then the potential of (1) is approximately
qd cos 0 p i, (3)
V2 (3)
4weor- - 41reor
where the vector p is called the dipole moment and is defined
as
p = qd (coul-m)
138 Polarizationand Conduction

v V.

-0

4•eor

Figure 3-2 (a) The potential at any point P due to the electric dipole is equal to the
sum of potentials of each charge alone. (b) The equi-potential (dashed) and field lines
(solid) for a point electric dipole calibrated for 4v•eolp = 100.
 Polarization 139

Because the separation of atomic charges is on the order of

1 A(10 - 10 m) with a charge magnitude equal to an integer
multiple of the electron charge (q = 1.6x 10-19 coul), it is
convenient to express dipole moments in units of debyes
defined as 1 debye=3.33x10-30 coul-m so that dipole
moments are of order p = 1.6 x 10-29 coul-m 4.8 debyes.
The electric field for the point dipole is then

3
p 3 = 3 ((p i )i - p
p * i,)i,
E= -V- [2 cos i, +sin i] (5)
4rreor 47reor

the last expressions in (3) and (5) being coordinate indepen-

dent. The potential and electric field drop off as a single
higher power in r over that of a point charge because the net
charge of the dipole is zero. As one gets far away from the
dipole, the fields due to each charge tend to cancel. The point
dipole equipotential and field lines are sketched in Figure
3-2b. The lines tangent to the electric field are

dr E, = 2 cot 0 r = ro sin 2 0 (6)

r dO Eo

where ro is the position of the field line when 0 = 7r/2. All field
lines start on the positive charge and terminate on the nega-
tive charge.
If there is more than one pair of charges, the definition of
dipole moment in (4) is generalized to a sum over all charges,

P= qiri (7)
all charges

where ri is the vector distance from an origin to the charge qi

as in Figure 3-3. When the net charge in the system is zero
(_ qj = 0), the dipole moment is independent of the choice of
origins for if we replace ri in (7) by ri +ro, where ro is the
constant vector distance between two origins:

p=Z qi(ri + ro)

0
= qijri + ro /q
_=qiri (8)
The result is unchanged from (7) as the constant ro could be
taken outside the summation.
If we have a continuous distribution of charge (7) is further
generalized to
r
p= rdq
all q
140 Polarizationand Conduction

p = fr dq
all q

Figure 3-3 The dipole moment can be defined for any distribution of charge. If the
net charge in the system is zero, the dipole moment is independent of the location of
the origin.

Then the potential and electric field far away from any
dipole distribution is given by the coordinate independent
expressions in (3) and (5) where the dipole moment p is given
by (7) and (9).

3-1-2 Polarization Charge

We enclose a large number of dipoles within a dielectric

medium with the differential-sized rectangular volume
Ax Ay Az shown in Figure 3-4a. All totally enclosed dipoles,
being charge neutral, contribute no net charge within the
volume. Only those dipoles within a distance d n of each
surface are cut by the volume and thus contribute a net
charge where n is the unit normal to the surface at each face,
as in Figure 3-4b. If the number of dipoles per unit volume is
N, it is convenient to define the number density of dipoles as
the polarization vector P:
P= Np= Nqd
The net charge enclosed near surface 1 is
dqi = (Nqd.)l. Ay Az = P.(x) Ay Az
while near the opposite surface 2
dq 2 = -(Nqdt) 1 .+a, Ay Az = -P.(x +Ax) Ay Az (12)
 Polarization 141

a
T/ S,

Figure 3-4 (a) The net charge enclosed within a differential-sized volume of dipoles
has contributions only from the dipoles that are cut by the surfaces. All totally enclosed
dipoles contribute no net charge. (b) Only those dipoles within a distance d - n of the
surface are cut by the volume.

where we assume that Ay and Az are small enough that the

polarization P is essentially constant over the surface. The
polarization can differ at surface 1 at coordinate x from that
at surface 2 at coordinate x + Ax if either the number density
142 Polarizationand Conduction

N, the charge.q,or the displacement d is a function of x. The

difference in sign between (11) and (12) is because near S 1 the
positive charge is within the volume, while near S2 negative
charge remains in the volume. Note also that only the
component of d normal to the surface contributes to the
volume of net charge.
Similarly, near the surfaces Ss and S4 the net charge
enclosed is
dq3 = (Nqd,) 1, Ax Az = P,(y) Ax Az
(13)
dq4 = -(Nqd,) 1 ,+a, Ax Az = -P,(y +Ay) Ax Az
while near the surfaces S 5 and S 6 with normal in the z direc-
tion the net charge enclosed is
dq5 = (Nqd.)I, Ax Ay = P,(z) Ax Ay
(14)
dq6 = -(Nqdd)l=+a, Ax Ay = -P,(z +Az) Ax Ay
The total charge enclosed within the volume is the sum of
(11)-(14):
dqT = dqI + dq2 + dqs + dq4 + dqs + dqa
(P.(x)-P.(x+Ax) P,(y)-P,(y+Ay)

+ P(z)-P(z +Az)) Ax Ay Az (15)

Az
As the volume shrinks to zero size, the polarization terms in
(15) define partial derivatives so that the polarization volume
charge density is

pdli = qT (aP + +P, = -V - P (16)

A.0o Ax Ay Az ax aVy az 6
Ay-.O

This volume charge is also a source of the electric field and

needs to be included in Gauss's law
V - (eoE) = pf+po = p -V • P (17)
where we subscript the free charge pf with the letter f to
distinguish it from the polarization charge p,. The total
polarization charge within a region is obtained by integrating
(16) over the volume,

q = pPidV=- V.PdV=- P dS (18)

where we used the divergence theorem to relate the polariza-

tion charge to a surface integral of the polarization vector.
 Polarization 143

3-1-3 The Displacement Field

Since we have no direct way of controlling the polarization

charge, it is convenient to cast Gauss's law only in terms of
free charge by defining a new vector D as
D=eoE+P (19)
This vector D is called the displacement field because it differs
from e0E due to the slight charge displacements in electric
dipoles. Using (19), (17) can be rewritten as
V - (EOE+P)= V - D= p1 (20)
where pf only includes the free charge and not the bound
polarization charge. By integrating both sides of (20) over a
volume and using the divergence theorem, the new integral
form of Gauss's law is

V.DdV=f D-dS= pdV (21)

In free space, the polarization P is zero so that D = soE and

(20)-(21) reduce to the free space laws used in Chapter 2.

3-1-4 Linear Dielectrics

It is now necessary to find the constitutive law relating the

polarization P to the applied electric field E. An accurate
discussion would require the use of quantum mechanics,
which is beyond the scope of this text. However, a simplified
classical model can be used to help us qualitatively under-
stand the most interesting case of a linear dielectric.

(a) Polarizability
We model the atom as a fixed positive nucleus with a sur-
rounding uniform spherical negative electron cloud, as
shown in Figure 3-5a. In the absence of an applied electric
field, the dipole moment is zero because the center of charge
for the electron cloud is coincident with the nucleus. More
formally, we can show this using (9), picking our origin at the
position of the nucleus:
0 2w Ro

p= Q(0) - f f irpor 3 sin 0 dr dO d (22)

/ ~=o
Jo =o

Since the radial unit vector i, changes direction in space, it is

necessary to use Table 1-2 to write i, in terms of the constant
Cartesian unit vectors:
i, = sin 0 cos i. +sin 0 sin 4i, +cos Oi.
"B d

144 Polarizationand Conduction

-~ Z
- n
o = L.
R03

2
Electric field applied r = a +r -2 racos
No electric field
(R>R o )
(a)
(a)

Figure 3-5 (a) A simple atomic classical model has a negative spherical electron cloud
of small radius Ro centered about a positive nucleus when no external electric field is
present. An applied electric field tends to move the positive charge in the direction of
the field and the negative charge in the opposite direction creating an electric dipole.
(b) The average electric field within a large sphere of radius R (R >>Ro) enclosing many
point dipoles is found by superposing the average fields due to each point charge.

When (23) is used in (22) the x and y components integrate to

zero when integrated over 0, while the z component is zero
when integrated over 0 so that p = 0.
An applied electric field tends to push the positive charge
in the direction of the field and the negative charge in the
opposite direction causing a slight shift d between the center
of the spherical cloud and the positive nucleus, as in Figure
3-5a. Opposing this movement is the attractive coulombic
force. Considering the center of the spherical cloud as our
origin, the self-electric field within the cloud is found from
Section 2.4.3b as
Qr
E, = 3 (24)

In equilibrium the net force F on the positive charge is zero,

F=Q(E 4- reoR)3 =0 (25)

where we evaluate (24) at r = d and EL, is the local polarizing

electric field acting on the dipole. From (25) the equilibrium
dipole spacing is

d = -- EELo (26)
Q
so that the dipole moment is written as
p = Qd = aELo, a = 4ireoR (27)
where a is called the polarizability.

__~__________li_______L
 Polarization 145

(b) The Local Electric Field

If this dipole were isolated, the local electric field would
equal the applied macroscopic field. However, a large
number density N of neighboring dipoles also contributes to
the polarizing electric field. The electric field changes dras-
tically from point to point within a small volume containing
many dipoles, being equal to the superposition of fields due
to each dipole given by (5). The macroscopic field is then the
average field over this small volume.
We calculate this average field by first finding the average
field due to a single point charge Q a distance a along the z
axis from the center of a spherical volume with radius R
much larger than the radius of the electron cloud (R >> Ro) as
in Figure 3-5b. The average field due to this charge over the
spherical volume is

1 " Q(ri,-ai
<E" 2 )r'sinOdrdOd,
r 0 Jo ,~o 42reo[a +r - 2ra
cos 0]31
(28)
where we used the relationships
-
rQp=a2 2racos , rQp = ri, -ai (29)
Using (23) in (28) again results in the x and y components
being zero when integrated over 4. Only the z component is
now nonzero:

EQ ( 2 " fR r (cos 0-a/r)sin Odrd0

3
2
1rR (4rEo) 0=0 ,o [a24+r -2ra cos0]
(30)
We introduce the change of variable from 0 to u
u = r +a2-2arcos 0, du = 2ar sin 0dO (31)
so that (30) can be integrated over u and r. Performing the u
integration first we have

2
<E>= 3s + 2 r 2 (r2drd
u drd u
8iR JEo J,-.) 4a
J-ro

3Q r r 2(r -a +u)\ (r+a) dr

2
83 0 4a
j=
=o0 U I u=(-a)

,
8?TwR
3Q
soa
__r2(lr
R
r=O
)dr
1
r-a
TJ
|r-a
(32)

We were careful to be sure to take the positive square root

in the lower limit of u. Then for r > a, the integral is zero so
146 Polarization and Conduction

that the integral limits over r range from 0 to a:

< E, >=-- -3Q I- 2r 2 dr= -Qa 33)

8rR soa' o 41reoR
To form a dipole we add a negative charge -Q, a small
distance d below the original charge. The average electric
field due to the dipole is then the superposition of (33) for
both charges:

<E> = Q -[a-(a-d)]= Qd s
4wsoR 4soR3 4rsoRS
(34)
If we have a number density N of such dipoles within the
sphere, the total number of dipoles enclosed is ,iTrR'N so that
superposition of (34) gives us the average electric field due to
all the dipoles in-terms of the polarization vector P = Np:
-wRNp P
<E > =- •- = (35)
4vsoR 3eo
The total macroscopic field is then the sum of the local field
seen by each dipole and the average resulting field due to all
the dipoles
- -P E+o E
E= <E> +E = (36)
360

so that the polarization P is related to the macroscopic electric

field from (27) as

, = Na(E+
P = Np= NaEL (37)

which can be solved for P as

Na Na/eo
P= E = ,eoE, x, N (38)
1- Na/3eo I - Na/3eo
where we introduce the electric susceptibility X, as the pro-
portionality constant between P and soE. Then, use of (38) in
(19) relates the displacement field D linearly to the electric
field:
D= eoE+P= o(1 +x,)E= Eoe,E= eE (39)

where E, = 1+x, is called the relative dielectric constant and

e = eeo is the permittivity of the dielectric, also simply called
the dielectric constant. In free space the susceptibility is zero
(X, = 0) so that e, = 1 and the permittivity is that of free space,
e = so. The last relation in (39) is usually the most convenient
to use as all the results of Chapter 2 are also correct within
 Polarization 147

linear dielectrics if we replace eo by e. Typical values of

relative permittivity are listed in Table 3-1 for various com-
mon substances. High dielectric constant materials are usually
composed of highly polar molecules.

Table 3-1 The relative permittivity for various common substances at room
temperature

E, = E/eo
Carbon Tetrachloride" 2.2
Ethanola 24
Methanol" 33
n-Hexane" 1.9
Nitrobenzene" 35
Pure Watera 80
Barium Titanateb(with 20% Strontium Titanate) >2100
Borosilicate Glass 4.0
Ruby Mica (Muscovite)b 5.4
Polyethyleneb 2.2
Polyvinyl Chlorideb 6.1
Teflonb Polytetrafluorethylene) 2.1
Plexiglas 3.4
Paraffin Waxb 2.2

"From Lange's Handbook of Chemistry, 10th ed., McGraw-Hill, New York, 1961, pp.
1234-37.
bFrom A. R. von Hippel (Ed.) DielectricMaterials and Applications, M.I.T., Cambridge,
Mass., 1966, pp. 301-370

The polarizability and local electric field were only intro-

duced so that we could relate microscopic and macroscopic
fields. For most future problems we will describe materials by
their permittivity e because this constant is most easily
measured. The polarizability is then easily found as
Na Na e - Eo
E -60 - Na/ e- (40)
1-Na/3eo 3ea e+2eo80
It then becomes simplest to work with the field vectors D and
E. The polarization can always be obtained if needed from
the definition
P = D-eoE = (e - eo)E (41)

EXAMPLE 3-1 POINT CHARGE WITHIN A DIELECTRIC SPHERE

Find all the fields and charges due to a point charge q

within a linear dielectric sphere of radius R and permittivity e
surrounded by free space, as in Figure 3-6.
 148 Polarizationand Conduction

e - co)q
2
41reR

q (e -Eo) q

+qpCo= cq

,
e4

Figure 3-6 The electric field due to a point charge within a dielectric sphere is less
than the free space field because of the partial neutralization of the point charge by the
accumulation of dipole ends of opposite charge. The total polarization charge on the
sphere remains zero as an equal magnitude but opposite sign polarization charge
appears at the spherical interface.

SOLUTION

Applying Gauss's law of (21) to a sphere of any radius r

whether inside or outside the sphere, the enclosed free
charge is always q:

sD dS=D,47rr =q q
D,=r247rr allr
1

The electric field is then discontinuous at r = R,

- 2 r<R
e 47rer
r = Dr = qq ,r>R
2
Lo 47EOr r

due to the abrupt change of permittivities. The polarization

field is

(e-e)q
2
r<R
Pr= D, - EoEr = 47rer
O, r>R

_· _II ~
 Polarization 149

The volume polarization charge pp is zero everywhere,

p,=-P r 8r (r2P,)=0,
-P=rl 0<r<R

except at r = 0 where a point polarization charge is present,

and at r = R where we have a surface polarization charge
found by using (18) for concentric Gaussian spheres of radius
r inside and outside the dielectric sphere:
r<R
P -(e-eo)q/E, rR
qJ,=-j • dS=-P,4rr2=
O, r>R
We know that for r <R this polarization charge must be a
point charge at the origin as there is no volume charge
contribution yielding a total point charge at the origin:
80
qT = qp +q = -q

This reduction of net charge is why the electric field within

the sphere is less than the free space value. The opposite
polarity ends of the dipoles are attracted to the point charge,
partially neutralizing it. The total polarization charge
enclosed by the sphere with r > R is zero as there is an
opposite polarity surface polarization charge at r = R with
density,
(e -e o)q
P 4reR 2

The total surface charge op,4rR2 = (( -eo)q/e is equal in

magnitude but opposite in sign to the polarization point
charge at the origin. The total p6larization charge always
sums to zero.

3-1-5 Spontaneous Polarization

(a) Ferro-electrics
Examining (38) we see that when Na/3eo = 1 the polariza-
tion can be nonzero even if the electric field is zero. We can
just meet this condition using the value of polarizability in
(27) for electronic polarization if the whole volume is filled
with contacting dipole spheres of the type in Figure 3-5a so
that we have one dipole for every volume of 3rRS. Then any
slight fluctuation in the local electric field increases the
polarization, which in turn increases the local field resulting
in spontaneous polarization so that all the dipoles over a
region are aligned. In a real material dipoles are not so
150 Polarizationand Conduction

densely packed. Furthermore, more realistic statistical models

including thermally induced motions have shown that most
molecules cannot meet the conditions for spontaneous
polarization.
However, some materials have sufficient additional contri-
butions to the polarizabilities due to ionic and orientational
polarization that the condition for spontaneous polarization is
met. Such materials are called ferro-electrics, even though
they are not iron compounds, because of their similarity in
behavior to iron compound ferro-magnetic materials, which
we discuss in Section 5.5.3c. Ferro-electrics are composed of
permanently polarized regions, called domains, as illustrated
in Figure 3-7a. In the absence of an electric field, these
domains are randomly distributed so that the net macroscopic
polarization field is zero. When an electric field is applied, the
dipoles tend to align with the field so that domains with a
polarization component along the field grow at the expense of
nonaligned domains. Ferro-electrics typically have very high
permittivities such as barium titanate listed in Table 3-1.
The domains do not respond directly with the electric field
as domain orientation and growth is not a reversible process
but involves losses. The polarization P is then nonlinearly
related to the electric field E by the hysteresis curve shown in
Figure 3-8. The polarization of an initially unpolarized
sample increases with electric field in a nonlinear way until
the saturation value Psat is reached when all the domains are
completely aligned with the field. A further increase in E does
not increase P as all the dipoles are completely aligned.
As the field decreases, the polarization does not retrace its
path but follows a new path as the dipoles tend to stick to their
previous positions. Even when the electric field is zero, a

- E

-· -····· ·· ·
No applied field Electric tield applied

(a) (b)

Figure 3-7 (a) In the absence of an applied electric field, a ferro-electric material
generally has randomly distributed permanently polarized domains. Over a macro-
scopic volume, the net polarization due to all the domains is zero. (b) When an electric
field is applied, domains with a polarization component in the direction of the field
grow at the expense of nonaligned domains so that a net polarization can result.

I I_
 Polarization 151

Figure 3-8 A typical ferro-electric hysteresis curve shows a saturation value P.., when
all the domains align with the field, a remanent polarization P, when the electric field is
removed, and a negative coercive electric field - E,, necessary to bring the polarization
back to zero.

remanent polarization P, remains. To bring the polarization

to zero requires a negative coercive field -Ec. Further magni-
tude increases in negative electric field continues the sym-
metric hysteresis loop until a negative saturation is reached
where all the dipoles have flipped over. If the field is now
brought to zero and continued to positive field values, the
whole hysteresis curve is traversed.

(b) Electrets
There are a class of materials called electrets that also
exhibit a permanent polarization even in the absence of an
applied electric field. Electrets are typically made using
certain waxes or plastics that are heated until they become
soft. They are placed within an electric field, tending to align
the dipoles in the same direction as the electric field, and then
allowed to harden. The dipoles are then frozen in place so
that even when the electric field is removed a permanent
polarization remains.
Other interesting polarization phenomena are:

Electrostriction-slightchange in size of a dielectric due to the

electrical force on the dipoles.

Piezo-electricity-when the electrostrictive effect is reversible

so that a mechanical strain also creates a field.

Pyro-electricity-induced polarization due to heating or

cooling.
152 Polarizationand Conduction

3-2 CONDUCTION

3-2-1 Conservation of Charge

In contrast to dielectrics, most metals have their outermost

band of electrons only weakly bound to the nucleus and are
free to move in an applied electric field. In electrolytic solu-
tions, ions of both sign are free to move. The flow of charge,
called a current, is defined as the total charge flowing through
a surface per unit time. In Figure 3-9a a single species of free
charge with density pf and velocity vi flows through a small
differential sized surface dS. The total charge flowing through
this surface in a time At depends only on the velocity
component perpendicular to the surface:
AQ = piAtvi - dS (1)
The tangential component of the velocity parallel to the sur-
face dS only results in charge flow along the surface but not
through it. The total differential current through dS is then
defined as
AQ,
dI = = pv's *dS= Jf • dS ampere (2)

All the charge in

dotted region has
I.ft V in q

Vni L1

AQ=- I; pf Atvi - dS= -#JlAt - dS

S' S
(b)

Figure 3-9 The current through a surface is defined as the number of charges per
second passing through the surface. (a) The current is proportional to the component
of charge velocity perpendicular to the surface. (b) The net change of total charge
within a volume is equal to the difference of the charge entering to that leaving in a
small time At.
 Conduction 153

where the free current density of these charges Jf is a vector

and is defined as
Ji = pfivi amp/m 2 (3)

If there is more than one type of charge carrier, the net

charge density is equal to the algebraic sum of all the charge
densities, while the net current density equals the vector sum
of the current densities due to each carrier:

pf=lpf, Jf=Xpfivi (4)

Thus, even if we have charge neutrality so that pf = 0, a net

current can flow if the charges move with different velocities.
For example, two oppositely charged carriers with densities
P I= -P2 Po moving with respective velocities v .and v 2 have

P =PI+P2= 0, J = P IVI +P 2V2 = P(VI - 2) (5)

With vI v2 a net current flows with zero net charge. This is
typical in metals where the electrons are free to flow while the
oppositely charged nuclei remain stationary.
The total current I, a scalar, flowing through a macroscopic
surface S, is then just the sum of the total differential currents
of all the charge carriers through each incremental size surface
element:

I=f * dS (6)

Now consider the charge flow through the closed volume V

with surface S shown in Figure 3-9b. A time At later, that
charge within the volume near the surface with the velocity
component outward will leave the volume, while that charge
just outside the volume with a velocity component inward will
just enter the volume. The difference in total charge is
transported by the current:

AQ= fv [pf(t + At) - p(t)] dV

=-¶ p.viAt - dS= - JAt - dS (7)

The minus sign on the right is necessary because when vi is in

the direction of dS, charge has left the volume so that the
enclosed charge decreases. Dividing (7) through by At and
taking the limit as At - 0, we use (3) to derive the integral
conservation of charge equation:
f r ao~_
SJ - dS+ dV= 0
atvO
154 Polarizationand Conduction

Using the divergence theorem, the surface integral can be

converted to a volume integral:

V Jf + pf V
. ] dV = 0 - Jf +• =0 (9)

where the differential point form is obtained since the

integral must be true for any volume so that the bracketed
term must be zero at each point. From Gauss's law (V - D =pf)
(8) and (9) can also be written as

dS= 0, V- J,+ =0 (10)

fS (J,+L
at) - at O (10)
where J, is termed the conduction current density and aD/at
is called the displacement current density.
This is the field form of Kirchoff's circuit current law that
the algebraic sum of currents at a node sum to zero. Equation
(10) equivalently tells us that the net flux of total current,
conduction plus displacement, is zero so that all the current
that enters a surface must leave it. The displacement current
does not involve any charge transport so that time-varying
current can be transmitted through space without charge
carriers. Under static conditions, the displacement current is
zero.

3-2-2 Charged Gas Conduction Models

(a) Governing Equations.

In many materials, including good conductors like metals,
ionized gases, and electrolytic solutions as well as poorer
conductors like lossy insulators and semiconductors, the
charge carriers can be classically modeled as an ideal gas
within the medium, called a plasma. We assume that we have
two carriers of equal magnitude but opposite sign :q with
respective masses m, and number densities n,. These charges
may be holes and electrons in a semiconductor, oppositely
charged ions in an electrolytic solution, or electrons and
nuclei in a metal. When an electric field is applied, the posi-
tive charges move in the direction of the field while the
negative charges move in the opposite direction. These
charges collide with the host medium at respective frequen-
cies v+ and v-, which then act as a viscous or frictional dis-
sipation opposing the motion. In addition to electrical and
frictional forces, the particles exert a force on themselves
through a pressure term due to thermal agitation that would
be present even if the particles were uncharged. For an ideal
gas the partial pressure p is
2 -
p = nkT Pascals [kg-s- -m '] (11)
 Conduction 155

where n is the number density of charges, T is the absolute

temperature, and k = 1.38 X 10-23 joule/K is called Boltz-
mann's constant.
The net pressure force on the small rectangular volume
shown in Figure 3-10 is
(p(x-Ax)-p(x). p(y)-p(y+Ay).

()(A) Ax Ay Az (12)

where we see that the pressure only exerts a net force on the
volume if it is different on each opposite surface. As the
volume shrinks to infinitesimal size, the pressure terms in (12)
define partial derivatives so that the volume force density
becomes

lim li = (LPx i ,.(+ 0y , +Lp i = -Va (13)

x-A.o Ax AAz a a az
Az -O

Then using (11)-(13), Newton's force law for each charge

carrier within the small volume is

av, 1
m. - qE - m v±v, -- V(n±kT) (14)
at n.

Ax)

ay)

Z -l. V

------30

x
Figure 3-10 Newton's force law, applied to a small rectangular volume Ax Ay Az
moving with velocity v, enclosing positive charges with number density 4. The pressure
is the force per unit area acting normally inward on each surface and only contributes
to the net force if it is different on opposite faces.
 156 Polarizationand Conduction

where the electric field E is due to the imposed field plus the
field generated by the charges, as given by Gauss's law.

(b) Drift-Diffusion Conduction

Because in many materials the collision frequencies are
typically on the order of v - 101s Hz, the inertia terms in (14)
are often negligible. In this limit we can easily solve (14) for
the velocity of each carrier as
1 /1
lim v,=- ±qE--V(nkT) (15)
avIMC.i v mVve n,

The charge and current density for each carrier are simply
given as
p *= qn±, J. = pav* =±qnav* (16)
Multiplying (15) by the charge densities then gives us the
constitutive law for each current as
J. = *qn,.vL = ±p±/,±E- D±Vp± (17)
where A, are called the particle mobilities and D. are their
diffusion coefficients

2
A* =-q[A-kg-'-s- 2 ], D.= kT [m -s-'] (18)

assuming that the system is at constant temperature. We see

that the ratio DL//A± for each carrier is the same having units
of voltage, thus called the thermil voltage:
D. S=AT
kvolts [kg-m 2 -A - -s -
3] (19)
IA* q
This equality is known as Einstein's relation.
In equilibrium when the net current of each carrier is zero,
(17) can be written in terms of the potential as (E = -V V)

J+= J_ = 0 = -p±aC±V V~:DVp± (20)

which can be rewritten as

V[± V+1ln p] = 0 (21)

The bracketed term can then only be a constant, so the charge

density is related to the potential by the Boltzmann dis-
tribution:
p±= po
*e /T (22)
where we use the Einstein relation of (19) and ±po is the
equilibrium charge density of each carrier when V= 0 and
are of equal magnitude because the system is initially neutral.

--
1 `
 Conduction 157

To find the spatial dependence of p and V we use (22) in

Poisson's equation derived in Section 2.5.6:
V2V (P++P-) Po (e-qVIAT_ qVkT 2PO _V
-)= (=e e sinhi
8 8E AT
(23)
This equation is known as the Poisson-Boltzmann equation
because the charge densities obey Boltzmann distributions.
Consider an electrode placed at x = 0 raised to the potential
Vo with respect to a zero potential at x = +0o, as in Figure
3-l1 a. Because the electrode is long, the potential only varies

V= Vo

0 000
0
000
G
E)E)oC

E Q 0E E,

V : i• 08
0 0 0
0 0 0
0
0 0
0
0

x~d
(b)

Figure 3-11 Opposite

inserted into a conduct
out its field for distanc
with respect to a zero I
of Vo. (b) Point charge
158 Polarization and Conduction

with the x coordinate so that (23) becomes

d 22 l - qV 12_ ekT
d- l2 sinh V= 0, V= , (24)
dx l kT' 2poq
where we normalize the voltage to the thermal voltage kT/q
and ldis called the Debye length.
If (24) is multiplied by dVldx, it can be written as an exact
differential:
d-x -[()2
a cos ] =0 (25)

The bracketed term must then be a constant that is evaluated

far from the electrode where the potential and electric field
x= -dV/dx are zero:
dl"
_x
22 1)] =
2 x>0
>1/2 (26)
E~ (cosh - 1) = - sinh -(26)
dx 1d 1d 2 x<0
The different signs taken with the square root are necessary
because the electric field points in opposite directions on each
side of the electrode. The potential is then implicitly found by
direct integration as
tanh (V/4) ixld >0
tanh (Vo/4) iex<0
U (27)
The Debye length thus describes the characteristic length
over which the applied potential exerts influence. In many
materials the number density of carriers is easily of the order
of no= 102 0/m 3, so that at room temperature (T 2930 K), id is
typically 10- 7 m.
Often the potentials are very small so that qV/kT<< 1. Then,
the hyperbolic terms in (27), as well as in the governing
equation of (23), can be approximated by their arguments:
2 V
V- 2= 0 (28)

This approximation is only valid when the potentials are

much less than the thermal voltage kT/q, which, at room
temperature is about 25 myv. In this limit the solution of (27)
shows that the voltage distribution decreases exponentially.
At higher values of Vo, the decay is faster, as shown in Figure
3-1 la.
If a point charge Q is inserted into the plasma medium, as
in Figure 3-11b, the potential only depends on the radial
distance r. In the small potential limit, (28) in spherical coor-
dinates is
I ( 2 aV V
2 r -r /- =0 (29)
r arr ar l

II __~I~
 Conduction 159

Realizing that this equation can be rewritten as

02 (rV)
- (rV)- =0O (30)

we have a linear constant coefficient differential equation in

the variable (rV) for which solutions are

rV= A e-r/d +A 2 e+rld (31)

Because the potential must decay and not grow far from
the charge, A2 = 0 and the solution is

V= - e- /Ie (32)
47rer

where we evaluated A by realizing that as r -*0 the potential

must approach that of an isolated point charge. Note that for
small r the potential becomes very large and the small poten-
tial approximation is violated.

(c) Ohm's Law

We have seen that the mobile charges in a system described
by the drift-diffusion equations accumulate near opposite
polarity charge and tend to shield out its effect for distances
larger than the Debye length. Because this distance is usually
so much smaller than the characteristic system dimensions,
most regions of space outside the Debye sheath are charge
neutral with equal amounts of positive and negative charge
density ±:Po.In this region, the diffusion term in (17) is negli-
gible because there are no charge density gradients. Then the
total current density is proportional to the electric field:

J = J++J- = po(v+-v_) = qno(iz+ + I_)E = oE (33)

where o, [siemans/m (m-3-kg-'-sS-A2 )] is called the Ohmic

conductivity and (33) is the point form of Ohm's law. Some-
times it is more convenient to work with the reciprocal
conductivity p,= (1/ar) (ohm-m) called the resistivity. We will
predominantly use Ohm's law to describe most media in this
text, but it is important to remember that it is often not valid
within the small Debye distances near charges. When Ohm's
law is valid, the net charge is zero, thus giving no contribution
to Gauss's law. Table 3-2 lists the Ohmic conductivities for
various materials. We see that different materials vary over
wide ranges in their ability to conduct charges.
The Ohmic conductivity of "perfect conductors" is large
and is idealized to be infinite. Since all physical currents in
(33) must remain finite, the electric field within the conductor
160 Polarizationand Conduction

is zero so that it imposes an equipotential surface:

E=O
lim J= o'E V=const (34)
a-0oI J = finite

Table 3-2 The Ohmic conductivity

for various common substances at
room temperature

uo [siemen/m]
Silver" 6.3 x 107
Copper" 5.9x 107
Gold" 4.2 x 107
Lead" 0.5 x 107
Tin" 0.9X 107
Zinc" 1.7x 10 7
Carbon" 7.3 x 10- 4
Mercuryb 1.06 X 106
Pure Waterb 4 x 10 - 6

Nitrobenzeneb 5 x 10-7
Methanolb 4 x 10- 5
Ethanol b 1.3 x 10- 7
Hexaneb <x 10- s

"From Handbook of Chemistry and Phy-

sics, 49th ed., The Chemical Rubber Co.,
1968,p. E80.
b From Lange's Handbook of Chemistry,
10th ed., McGraw-Hill, New York, 1961,
pp. 1220-21.

Throughout this text electrodes are generally assumed to

be perfectly conducting and thus are at a constant potential.
The external electric field must then be incident perpendic-
ularly to the surface.

(d) Superconductors
One notable exception to Ohm's law is for superconducting
materials at cryogenic temperatures. Then, with collisions
negligible (v,,= 0) and the absolute temperature low (T 0),
the electrical force on the charges is only balanced by their
inertia so that (14) becomes simply

= +--E (35)
Ot m*
We multiply (35) by the charge densities that we assume to be
constant so that the constitutive law relating the current
 Field Boundary Conditions 161

density to the electric field is

vs)_oJ. q n, = 2(+qn , q fn
a(qnv,) aJ, E=w. eE, W, = (36)
at at m* m*e
where wo• is called the plasma frequency for each carrier.
For electrons (q = -1.6 x 10' 19 coul, m_ = 9.1 x 10- ' kg) of
density n-- 1020 /m S within a material with the permittivity
of free space, e = o 8.854X 10-12 farad/m, the plasma
frequency is

o,p_ = n'n/me 5.6 x 10 11radian/sec

=f,_ = w,_/27r - 9 X 1010 Hz (37)

If such a material is placed between parallel plate elec-

trodes that are open circuited, the electric field and current
density J= J++J must be perpendicular to the electrodes,
which we take as the x direction. If the electrode spacing is
small compared to the width, the interelectrode fields far
from the ends must then be x directed and be only a function
of x. Then the time derivative of the charge conservation
equation in (10) is

(U,+J_)+e- =o (38)

The bracketed term is just the time derivative of the total

current density, which is zero because the electrodes are open
circuited so that using (36) in (38) yields

2E 2 2 2
t+opE = 0, wp =w + _p- (39)

which has solutions

E = A 1 sin wt +A2 cos Wot (40)

Any initial perturbation causes an oscillatory electric field at

the composite plasma frequency ,p.The charges then execute
simple harmonic motion about their equilibrium position.

3-3 FIELD BOUNDARY CONDITIONS

In many problems there is a surface of discontinuity

separating dissimilar materials, such as between a conductor
and a dielectric, or between different dielectrics. We must
determine how the fields change as we cross the interface
where the material properties change abruptly.
 162 Polarizationand Conduction

3-3-1 Tangential Component of E

We apply the line integral of the electric field around a

contour of differential size enclosing the interface between
dissimilar materials, as shown in Figure 3-12a. The long
sections a and c of length dl are tangential to the surface and
the short joining sections b and d are of zero length as the
interface is assumed to have zero thickness. Applying the line
integral of the electric field around this contour, from Section
2.5.6 we obtain

dl = 0
L E dl= (E, - E2)
where E1 , and E 2 , are the components of the electric field
tangential to the interface. We get no contribution from the
normal components of field along sections b and d because
the contour lengths are zero. The minus sign arises along c
because the electric field is in the opposite direction of the
contour traversal. We thus have that the tangential

tl
.. -- z, 1 -

dS D2

2 + + + +Gf

+ b ni(D 2 -Di)=Ui

1 D
dS

Figure 3-12 (a) Stokes' law applied to a line integral about an interface of dis-
continuity shows that the tangential component of electric field is continuous across
the boundary. (b) Gauss's law applied to a pill-box volume straddling the interface
shows that the normal component of displacement vector is discontinuous in the free
surface charge density of.

i 1
 Field Boundary Conditions 163

components of the electric field are continuous across the

interface
Elt=E 2 =>n x (E2 - El)= 0 (2)
where n is the interfacial normal shown in Figure 3-12a.
Within a perfect conductor the electric field is zero. There-
fore, from (2) we know that the tangential component of E
outside the conductor is also zero. Thus the electric field must
always terminate perpendicularly to a perfect conductor.

3-3-2 Normal Component of D

We generalize the results of Section 2.4.6 to include dielec-

tric media by again choosing a small Gaussian surface whose
upper and lower surfaces of area dS are parallel to a surface
charged interface and are joined by an infinitely thin cylin-
drical surface with zero area, as shown in Figure 3-12b. Then
only faces a and b contribute to Gauss's law:

D dS= (D 2 -D 1 .)dS= oydS (3)

dS
where the interface supports a free surface charge density ao
and D 2, and D 1n are the components of the displacement
vector on either side of the interface in the direction of the
normal n shown, pointing from region I to region 2. Reduc-
ing (3) and using more compact notation we have
D 2,-D, = of, n(D 2 - D) = of (4)

where the minus sign in front of DI arises because the normal

on the lower surface b is -n. The normal components of the
displacement vector are discontinuous if the interface has a
surface charge density. If there is no surface charge (of = 0),
the normal components of D are continuous. If each medium
has no polarization, (4) reduces to the free space results of
Section 2.4.6.
At the interface between two different lossless dielectrics,
there is usually no surface charge (of= 0), unless it was
deliberately placed, because with no conductivity there is no
current to transport charge. Then, even though the normal
component of the D field is continuous, the normal
component of the electric field is discontinuous because the
dielectric constant in each region is different.
At the interface between different conducting materials,
free surface charge may exist as the current may transport
charge to the surface discontinuity. Generally for such cases,
the surface charge density is nonzero. In particular, if one
region is a perfect conductor with zero internal electric field,
 164 Polarizationand Conduction

the surface charge density on the surface is just equal to the

normal component of D field at the conductor's surface,

oa = n D (5)

where n is the outgoing normal from the perfect conductor.

3-3-3 Point Charge Above a Dielectric Boundary

If a point charge q within a region of permittivity el is a

distance d above a planar boundary separating region I from
region II with permittivity E2, as in Figure 3-13, the tangential
component of E and in the absence of free surface charge the
normal component of D, must be continuous across the
interface. Let us try to use the method of images by placing an
image charge q' at y = -d so that the solution in region I is
due to this image charge plus the original point charge q. The
solution for the field in region II will be due to an image
charge q" at y = d, the position of the original point charge.
Note that the appropriate image charge is always outside the
region where the solution is desired. At this point we do not
know if it is possible to satisfy the boundary conditions with
these image charges, but we will try to find values of q' and q"
to do so.

q
Region I
e Y d

62
Region II )

q
Region I eq * q" 2
el 61 + E2

q,
q q(E-2 -e Region II
q C,2+ E1

(b)

Figure 3-13 (a) A point charge q above a flat dielectric boundary requires different
sets of image charges to solve for the fields in each region. (b) The field in region I is
due to the original charge and the image charge q' while the field in region II is due
only to image charge q".

I _
 Field Boundary Conditions 165

The potential in each region is

1 q (6)

E= -V VI

VII=- 4,re 1 [x2+(y-d)2+z2 ]/2+y-2 [X2+(y+d) 2 + z2]3/2 ) (7)

2 +zi]+]/
q" [xxi+(y-d)i
-V l = 2 _ [2 + (Y_d)+ Z I)
EII=

To satisfy the continuity of tangential electric field at y 0 we

have
EE -V VI = __ (87)

EEi = -V V 2

With no surface
continuous at y =charge,
0, the normal component of D must be

eIEI = e 2E, 11 > -q +q'= -q" (9)

Solving (8) and (9) for the unknown charges we find

(2- (81)
81+82
2162
continuous
at y = 0,
(10)

The force on the point charge q is due only to the field

from image charge q':
qq q2 - 1)

3-3-4 Normal Component of P and EE

By integrating the flux of polarization over the same Gaus-

sian pillbox surface, shown in Figure 3-12b, we relate the
discontinuity
surface in normal
polarization component
charge density ofduepolarization
using to the
the relations
surface polarization charge density cr, using the relations

M M
 166 Polarizationand Conduction

from Section 3.1.2:

fsP-dS=- odSP nP2 -P. = -o' n (P 2 -P 1 ) = -ap

(12)
The minus sign in front of ,presults because of the minus sign
relating the volume polarization charge density to the diver-
gence of P.
To summarize, polarization charge is the source of P, free
charge is the source of D, and the total charge is the source of
eoE. Using (4) and (12), the electric field interfacial dis-
continuity is

n* (E2 -Ei) = n [(D2 -DI)-(P 2-PI)] af+O', (13)

so Eo

For linear dielectrics it is often convenient to lump polariza-

tion effects into the permittivity e and never use the vector P,
only D and E.
For permanently polarized materials, it is usually con-
venient to replace the polarization P by the equivalent
polarization volume charge density and surface charge
density of (12) and solve for E using the coulombic super-
position integral of Section 2.3.2. In many dielectric prob-
lems, there is no volume polarization charge, but at surfaces
of discontinuity a surface polarization charge is present as
given by (12).

EXAMPLE 3-2 CYLINDER PERMANENTLY POLARIZED ALONG ITS

AXIS

A cylinder of radius a and height L is centered about the z

axis and has a uniform polarization along its axis, P = Poi., as
shown in Figure 3-14. Find the electric field E and displace-
ment vector D everywhere on its axis.

SOLUTION

With a constant polarization P, the volume polarization

charge density is zero:
p. = -V . P =0
Since P= 0 outside the cylinder, the normal component of P
is discontinuous at the upper and lower surfaces yielding
uniform surface polarization charges:

o,(z = L/2)= Po, Top(z = -L12)= -Po

I
 Field Boundary Conditions 167

op = PO

z =-L/2

Op = -Po

-L/2

eoEz D -P
.5)

---
-L/2

-.5
111
(b)
Figure 3-14 (a) The electric field due to a uniformly polarized cylinder of length L is
the same as for two disks of surface charge of opposite polarity + Po at z = L/2. (b) The
perpendicular displacement field D, is continuous across the interfaces at z = ± L/2
while the electric field E. is discontinuous.
 168 Polarizationand Conduction

The solution for a single disk of surface charge was obtained

in Section 2.3.5b. We superpose the results for the two
disks taking care to shift the axial distance appropriately by
. L/2 yielding the concise solution for the displacement field:
P0 ( (z+ L2) (z - L/2)
2 • [a +(z +L/2) 2V [a2 +(z -L/2)2 ] 2

The electric field is then

.=DJeo, IzL >L/2
=L(D,-Po)/eo IzI <L/2
These results can be examined in various limits. If the
radius a becomes very large, the electric field should
approach that of two parallel sheets of surface charge ±Po, as
in Section 2.3.4b:

lim E, =f0, Jz >L/2

lm . -Poleo, Izi <L/2
with a zero displacement field everywhere.
In the opposite limit, for large z (z >>a, z >>L) far from the
cylinder, the axial electric field dies off as the dipole field with
0=0

lim E=
E o , p=Powa9L

with effective dipole moment p given by the product of the

total polarization charge at z = L/2, (Poira ), and the length L.

3-3-5 Normal Component of J

Applying the conservation of total current equation in

Section 3.2.1 to the same Gaussian pillbox surface in Figure
3-12b results in contributions again only from the upper and
lower surfaces labeled "a" and "b":

n (J2-J + (D2 -D)) =0 (14)

where we assume that no surface currents flow along the

interface. From (4), relating the surface charge density to the
discontinuity in normal D, this boundary condition can also
be written as

n* (J9-J)+ ý=0 (15)

at
which tells us that if the current entering a surface is different
from the current leaving, charge has accumulated at the

1
 Resistance 169

interface. In the dc steady state the normal component of J is

continuous across a boundary.

3-4 RESISTANCE

3-4-1 Resistance Between Two Electrodes

Two conductors maintained at a potential difference V

within a conducting medium will each pass a total current I,
as shown in Figure 3-15. By applying the surface integral
form of charge conservation in Section 3.2.1 to a surface S'
which surrounds both electrodes but is far enough away so
that J and D are negligibly small, we see that the only nonzero
current contributions are from the terminal wires that pass
through the surface. These must sum to zero so that the

J, Ea -• far from the electrodes

r3
Ir* fJ'dS= 0
S.
-
S

\\

Figure 3-15 A voltage applied across two electrodes within an ohmic medium causes
a current to flow into one electrode and out the other. The electrodes have equal
magnitude but opposite polarity charges so that far away the fields die off as a dipole
oc(1/rs). Then, even though the surface S' is increasing as r' , the flux of current goes
to zero as 1/r.
170 Polarizationand Conduction

currents have equal magnitudes but flow in opposite direc-

tions. Similarly, applying charge conservation to a surface S
just enclosing the upper electrode shows that the current I
entering the electrode via the wire must just equal the total
current (conduction plus displacement) leaving the electrode.
This total current travels to the opposite electrode and leaves
via the connecting wire.
The dc steady-state ratio of voltage to current between the
two electrodes in Figure 3-15 is defined as the resistance:

R= ohm [kg-m 2 -S-3-A - 2]

(1)
I

For an arbitrary geometry, (1) can be expressed in terms of

the fields as

SEdl LE-dl
J dS oE - dS (2)

where S is a surface completely surrounding an electrode and

L is any path joining the two electrodes. Note that the field
line integral is taken along the line from the high to low
potential electrode so that the voltage difference V is equal to
the positive line integral. From (2), we see that the resistance
only depends on the geometry and conductivity aoand not on
the magnitude of the electric field itself. If we were to
increase the voltage by any factor, the field would also
increase by this same factor everywhere so that this factor
would cancel out in the ratio of (2). The conductivity a may
itself be a function of position.

3-4-2 Parallel Plate Resistor

Two perfectly conducting parallel plate electrodes of arbi-
trarily shaped area A and spacing I enclose a cylinder of
material with Ohmic conductivity oa, as in Figure 3-16a. The
current must flow tangential to the outer surface as the
outside medium being free space has zero conductivity so that
no current can pass through the interface. Because the
tangential component of electric field is continuous, a field
does exist in the free space region that decreases with
increasing distance from the resistor. This three-dimensional
field is difficult to calculate because it depends on three coor-
dinates.
The electric field within the resistor is much simpler to
calculate because it is perpendicular to the electrodes in the x
direction. Gauss's law with no volume charge then tells us that
 Resistance 171

x I
i I
1L
a·

0--
I Depth I
Jr =Er, =

(a) (b) (c)

Figure 3-16 Simple resistor electrode geometries. (a) Parallel plates. (b) Coaxial
cylinders. (c) Concentric spheres.

this field is constant:

dEr
v (sE) = o >-• =>E = Eo (3)
dx
However, the line integral of E between the electrodes must
be the applied voltage v:

E.dx= v Eo= v/1 (4)

The current density is then

J = oEoix = (orv/1)i (5)
so that the total current through the electrodes is

I= J• dS= (ov/)A (6)

where the surface integral is reduced to a pure product

because the constant current density is incident perpendic-
ularly on the electrodes. The resistance is then

Iv spacing
R = .(7)
I oA (conductivity) (electrode area)
Typical resistance values can vary over many orders of
magnitude. If the electrodes have an area A = 1 cm2 (10- m )
with spacing I = 1 mm (10 - 3 m) a material like copper has a
resistance R -0.17 x 10-6 ohm while carbon would have a
resistance R - 1.4 x 104 ohm. Because of this large range of
resistance values sub-units often used are micro-ohms
(1 fl= 10 - 6 f), milli-ohms (1 mfl= 10- 3 1), kilohm (1 kfl =
10 [l), and megohms (1 Mf = 106 fl), where the symbol 0 is
used to represent the unit of ohms.
172 Polarizationand Conduction

Although the field outside the resistor is difficult to find, we

do know that for distances far from the resistor the field
approaches that of a point dipole due to the oppositely
charged electrodes with charge density
rf(x =0) = -o-(x = 1)=eEo = evl1 (8)
and thus dipole moment
p = -or1 (x = O)Ali. = -eAvi, (9)
The minus sign arises because the dipole moment points
from negative to positive charge. Note that (8) is only
approximate because all of the external field lines in the free
space region must terminate on the side and back of the
electrodes giving further contributions to the surface charge
density. Generally, if the electrode spacing I is much less than
any of the electrode dimensions, this extra contribution is
very small.

3-4-3 Coaxial Resistor

Two perfectly conducting coaxial cylinders of length 1,

inner radius a, and outer radius b are maintained at a poten-
tial difference v and enclose a material with Ohmic conduc-
tivity or, as in Figure 3-16b. The electric field must then be
perpendicular to the electrodes so that with no free charge
Gauss's law requires

(rE r) = 0
S"(eE)= 0l• 4 ,E = c
(10)
rr r
where c is an integration constant found from the voltage
condition
SErdr = c In r =vcc (l)

The current density is then

or
J,= -E,= (12)
r In (bla)
with the total current at any radius r being a constant

I= fJr d4 dz = av2 (13)

=o0 o in (b/a)
so that the resistance is
v In (bla)
R (14)
I 2?'ol
 Capacitance 173

3-4-4 Spherical Resistor

We proceed in the same way for two perfectly conducting

concentric spheres at a potential difference v with inner
radius R 1 and outer radius R 2 , as in Figure 3-16c. With no
free charge, symmetry requires the electric field to be purely
radial so that Gauss's law yields
ia c
V - (eE)=0•7r(r E,)= O E,= r (15)

where c is a constant found from the voltage condition as

E,dr = - =v (16)
S R(1R, -/R 2 )
The electric field and current density are inversely pro-
portional to the square of the radius

], = oE, = 2( R (17)
"(1/RI-1/R2)
so that the current density is constant at any radius r
2* 4tro
I= 2 = Jr 2 sin 0dO do= (1/R- (18)

with resistance

v (1/RI-1/R2)
R 4 (19)
I 7rT

3-5 CAPACITANCE

3-5-1 Parallel Plate Electrodes

Parallel plate electrodes of finite size constrained to poten-

tial difference v enclose a dielectric medium with permittivity
e. The surface charge density does not distribute itself uni-
formly, as illustrated by the fringing field lines for infinitely
thin parallel plate electrodes in Figure 3-17a. Near the edges
the electric field is highly nonuniform decreasing in magni-
tude on the back side of the electrodes. Between the elec-
trodes, far from the edges the electric field is uniform, being
the same as if the electrodes were infinitely long. Fringing
field effects can be made negligible if the electrode spacing I is
much less than the depth d or width w. For more accurate
work, end effects can be made even more negligible by using a
guard ring encircling the upper electrode, as in Figure 3-17b.
The guard ring is maintained at the same potential as the
electrode, thus except for the very tiny gap, the field between

M ýýl
 174 Polarizationand Conduction

x , Area A

0 I+++++++++++++++++++++++++++++++++++++++++++++++
I .

(b)

Figure 3-17 (a) Two infinitely thin parallel plate electrodes of finite area at potential
difference v have highly nonuniform fields outside the interelectrode region. (b) A
guard ring around one electrode removes end effects so that the field between the
electrodes is uniform. The end effects now arise at the edge of the guard ring, which is
far from the region of interest.
 Capacitance 175

the electrodes is as if the end effects were very far away and
not just near the electrode edges.
We often use the phrase "neglect fringing" to mean that the
nonuniform field effects near corners and edges are negli-
gible.
With the neglect of fringing field effects near the electrode
ends, the electric field is perpendicular to the electrodes and
related to the voltage as

E. dx = v > E = v/1 (1)

The displacement vector is then proportional to the electric

field terminating on each electrode with an equal magnitude
but opposite polarity surface charge density given by

Dx = eE, = or(x = 0) = -o'f(x = 1)= ev/l (2)

The charge is positive where the voltage polarity is positive,
and vice versa, with the electric field directed from the posi-
tive to negative electrode. The magnitude of total free charge
on each electrode is
eA
qf= of(x = O)A = -v (3)

The capacitance C is defined as the magnitude of the ratio

of total free charge on either electrode to the voltage
difference between electrodes:
C If = eA
- qfA
v l
(permittivity) (electrode area) farad [A 2 4 -kg -2
farad [A -s4-kg- -m- ]
spacing
(4)
Even though the system remains neutral, mobile electrons on
the lower electrode are transported through the voltage
source to the upper electrode in order to terminate the dis-
placement field at the electrode surfaces, thus keeping the
fields zero inside the conductors. Note that no charge is
transported through free space. The charge transport
between electrodes is due to work by the voltage source and
results in energy stored in the electric field.
In SI units, typical capacitance values are very small. If the
electrodes have an area of A = 1cm 2 (10 - 4 m2) with spacing of
l= 1 mm (10 - s m), the free space capacitance is C-
0.9x 10- 2 farad. For this reason usual capacitance values are
expressed in microfarads ( f== 10-6 farad), nanofarads
(1 nf = 10- 9 farad), and picofarads (1 pf = 10-' farad).
176 Polarizationand Conduction

With a linear dielectric of permittivity e as in Figure 3-18a,

the field of (1) remains unchanged for a given voltage but the
charge on the electrodes and thus the capacitance increases
with the permittivity, as given by (3). However, if the total
free charge on each electrode were constrained, the voltage
difference would decrease by the same factor.
These results arise because of the presence of polarization
charges on the electrodes that partially cancel the free charge.
The polarization vector within the dielectric-filled parallel
plate capacitor is a constant
P. = D. - EoE. = (e - so)E. = (e - eo)vI/ (5)
so that the volume polarization charge density is zero.
However, with zero polarization in the electrodes, there is a
discontinuity in the normal component of polarization at the
electrode surfaces. The boundary condition of Section 3.3.4
results in an equal magnitude but opposite polarity surface
polarization charge density on each electrode, as illustrated in

(a) n+
t Qh t

Dipoles
® Free charge
Depth d

V
E, = r In (b/a
r2( R11 1)
R2/

Depth I q(R 1 ) = eE,(r = R)4rR = -q(R2) =

q(a) = eEr(r = a) 2"al = -q(b) = 4rev

eEr(r = b)r = R2)4R2
In(b/a) R-i R2
(b) (c)

Figure 3-18 The presence of a dielectric between the electrodes increases the capaci-
tance because for a given voltage additional free charge is needed on each electrode to
overcome the partial neutralization of the attracted opposite polarity dipole ends. (a)
Parallel plate electrodes. (b) Coaxial cylinders. (c) Concentric spheres.

___~
 Capacitance 177

Figure 3-18a:

o(x = 0)= -o(x= ) = -P. = -(e -o)v/II (6)

Note that negative polarization charge appears on the posi-
tive polarity electrode and vice versa. This is because opposite
charges attract so that the oppositely charged ends of the
dipoles line up along the electrode surface partially neu-
tralizing the free charge.

3-5-2 Capacitance for any Geometry

We have based our discussion around a parallel plate

capacitor. Similar results hold for any shape electrodes in a
dielectric medium with the capacitance defined as the magni-
tude of the ratio of total free charge on an electrode to
potential difference. The capacitance is always positive by
definition and for linear dielectrics is only a function of the
geometry and dielectric permittivity and not on the voltage
levels,

C=- = =feEDds (7)

v 1,E-dl 1, E dl
as multiplying the voltage by a constant factor also increases
the electric field by the same factor so that the ratio remains
unchanged.
The integrals in (7) are similar to those in Section 3.4.1 for
an Ohmic conductor. For the same geometry filled with a
homogenous Ohmic conductor or a linear dielectric, the
resistance-capacitance product is a constant independent of
the geometry:

RC- LE-d1 sfE -dS

e (8)
R
ofsE.dS ILE.dl o
Thus, for a given geometry, if either the resistance or capaci-
tance is known, the other quantity is known immediately from
(8). We can thus immediately write down the capacitance of
the geometries shown in Figure 3-18 assuming the medium
between electrodes is a linear dielectric with permittivity 6
using the results of Sections 3.4.2-3.4.4:
1 eA
Parallel Plate R= I C= -
o'A I
1
In (bla) 2•e8
Coaxial R C= (9)
2alrrl In (b/a)

Spherical R= R- /R2 C= 4r
41ro- (1/R- l/R2)
178 Polarizationand Conduction

3-5-3 CurrentFlow Through a Capacitor

From the definition of capacitance in (7), the current to an

electrode is
dq
i d (Cv) d dC
S (• -)=
C-+v- (10)
dt dt dt dt

where the last term only arises if the geometry or dielectric

permittivity changes with time. For most circuit applications,
the capacitance is independent of time and (10) reduces to the
usual voltage-current circuit relation.
In the capacitor of arbitrary geometry, shown in Figure
3-19, a conduction current i flows through the wires into the
upper electrode and out of the lower electrode changing the
amount of charge on each electrode, as given by (10). There is
no conduction current flowing in the dielectric between the
electrodes. As discussed in Section 3.2.1 the total current,
displacement plus conduction, is continuous. Between the
electrodes in a lossless capacitor, this current is entirely dis-
placement current. The displacement field is itself related to
the time-varying surface charge distribution on each elec-
trode as given by the boundary condition of Section 3.3.2.

3-5-4 Capacitance of Two Contacting Spheres

If the outer radius R 2 of the spherical capacitor in (9) is put
at infinity, we have the capacitance of an isolated sphere of
radius R as

C= 41reR (11)

------------ ---- V-
-------- "----
--- -

S-----------

dq_

Figure 3-19 The conduction current i that travels through the connecting wire to an
electrode in a lossless capacitor is transmitted through the dielectric medium to the
opposite electrode via displacement current. No charge carriers travel through the
lossless dielectric.
 Capacitance 179

If the surrounding medium is free space (e = e0) for R = 1 m,

we have that C - -x 10- 9 farad 111 pf.
We wish to find the self-capacitance of two such contacting
spheres raised to a potential Vo, as shown in Figure 3-20. The
capacitance is found by first finding the total charge on the
two spheres. We can use the method of images by first placing
an image charge q =Q=47reRVo at the center of each
sphere to bring each surface to potential Vo. However, each
of these charges will induce an image charge q2 in the other
sphere at distance b2 from the center,

Q R2 R
q2 = , b2= -= (12)
2 D 2

where we realize that the distance from inducing charge to

the opposite sphere center is D = 2R. This image charge does
not raise the potential of either sphere. Similarly, each of
these image charges induces another image charge qs in the
other sphere at disance bs,

qR Q R2
= D-b2 S= bs =- = R (13)
2 3' D-b 2

which will induce a further image charge q4, ad infinitum. An

infinite number of image charges will be necessary, but with
the use of difference equations we will be able to add all the
image charges to find the total charge and thus the capaci-
tance.
The nth image charge q. and its distance from the center b.
are related to the (n - 1)th images as

q_-iR R9
q - b. = (14)
D-b,-I D-bn-1

At potential Vo

qn_ R R2
qn D -n b=_Db_bý O - b, _

Figure 3-20 Two identical contacting spheres raised to a potential Vo with respect to
infinity are each described by an infinite number of image charges q. each a distance b.
from the sphere center.

= M ý ý ý
180 Polarizationand Conduction

where D= 2R. We solve the first relation for b.-i as

D-b,-I= R
(15)

b,= q R+D
qn+,
where the second relation is found by incrementing n in the
first relation by 1. Substituting (15) into the second relation of
(14) gives us a single equation in the q.'s:
qR Rq, 1 2 1
+D=-q 1+-+ 1= 0 (16)
q.+, qn-1 q.+, q. q.-I
If we define the reciprocal charges as
p.= /q,, (17)
then (16) becomes a homogeneous linear constant coefficient
difference equation
Pn+i +2pn + P.-i = 0 (18)
Just as linear constant coefficient differential equations have
exponential solutions, (18) has power law solutions of the
form
p, = AA " (19)

where the characteristic roots A, analogous to characteristic

frequencies, are found by substitution back into (18),
-
A "+ +2A" +A ' =0 A2 + 2+ = (A + 1)2 =O (20)

to yield a double root with A = -1. Because of the double root,

the superposition of both solutions is of the form

Pn = AI(-1)" +A2n(-)"n(21)

similar to the behavior found in differential equations with

double characteristic frequencies. The correctness of (21) can
be verified by direct substitution back into (18). The constants
Al and A 2 are determined from ql and q2 as
I/Q = - A - A 2
p = A=0
1 2 +A 2A 1 (22)
q2 Q Q
so that the nth image charge is
1 1 -(-I--1n
Qn=
= = i n n
Pn -- 1)n/ n

__
 Lossy Media 181

The capacitance is then given as the ratio of the total charge

on the two spheres to the voltage,

2 '-2 (--1) 2Q21 1 ,

Vo Vo n,1 n Vo
= 8reR In 2 (24)

where we recognize the infinite series to be the Taylor series

expansion of In(l+x) with x=l1. The capacitance of two
contacting spheres is thus 2 In 2 - 1.39 times the capacitance
of a single sphere given by (11).
The distance from the center to each image charge is
obtained from (23) substituted into (15) as
= ((-1)" (n + 1) (n- 1)
b. - ) +) 2R = 1- R (25)
n(-1)" n
We find the force of attraction between the spheres by
taking the sum of the forces on each image charge on one of
the spheres due to all the image charges on the other sphere.
The force on the nth image charge on one sphere due to the
mth image charge in the other sphere is
-qq., _Q2(_-)(-) " +
nm
fn, n 2 -i7 - (26)
41• [2R - b, - b,] 4eR (m + n)
where we used (23) and (25). The total force on the left
sphere is then found by summing over all values of m and n,
- " + "
0 Cl -- . (-I) nm
f= I Y . 2 +M)2
m=ln1 4reR2 m=L=L (n+m)

S 2 [ln 2 - ] (27)

where the double series can be explicitly expressed.* The

force is negative because the like charge spheres repel each
other. If Qo = 1 coul with R = 1 m, in free space this force is
f 6.6x 108 nt, which can lift a mass in the earth's gravity
field of 6.8 x 107 kg (=3 x 107 lb).

3-6 LOSSY MEDIA

Many materials are described by both a constant permit-

tivity e and constant Ohmic conductivity o. When such a
material is placed between electrodes do we have a capacitor

* See Albert D. Wheelon, Tables of Summable Series and Integrals Involving Bessel
Functions, Holden Day, (1968) pp. 55, 56.
182 Polarizationand Conduction

or a resistor? We write the governing equations of charge

conservation and Gauss's law with linear constitutive laws:

V- Jr+M= , Jf= oE+pfU (1)

V-D=p f , D=eE (2)

We have generalized Ohm's law in (1) to include convection

currents if the material moves with velocity U. In addition to
the conduction charges, any free charges riding with the
material also contribute to the current. Using (2) in (1) yields
a single partial differential equation in pf:

(V E)+V (pfU)+ = 0+> V *(pfU)+-pf= 0 (3)

Pf/e

3-6-1 Transient Charge Relaxation

Let us first assume that the medium is stationary so that

U = 0. Then the solution to (3) for any initial possibly spatially
varying charge distribution po(x, y, z, t = 0) is

pf=Po(x, y, z, t = 0) e- , 7 = e/o (4)

where 7 is the relaxation time. This solution is the continuum

version of the resistance-capacitance (RC) decay time in
circuits.
The solution of (4) tells us that at all positions within a
conductor, any initial charge density dies off exponentially
with time. It does not spread out in space. This is our
justification of not considering any net volume charge in
conducting media. If a system has no volume charge at t = 0
(Po = 0), it remains uncharged for all further time. Charge is
transported through the region by the Ohmic current, but the
net charge remains zero. Even if there happens to be an initial
volume charge distribution, for times much longer than the
relaxation time the volume charge density becomes negligibly
small. In metals, 7 is on the order of 10 - '9 sec, which is the
justification of assuming the fields are zero within an elec-
trode. Even though their large conductivity is not infinite, for
times longer than the relaxation time 7, the field solutions are
the same as if a conductor were perfectly conducting.
The question remains as to where the relaxed charge goes.
The answer is that it is carried by the conduction current to
surfaces of discontinuity where the conductivity abruptly
changes.
 Lossy Media 183

3-6-2 Uniformly Charged Sphere

A sphere of radius R 2 with constant permittivity e and

Ohmic conductivity ao is uniformly charged up to the radius
RI with charge density Po at time t = 0, as in Figure 3-21.
From R 1 to R 2 the sphere is initially uncharged so that it
remains uncharged for all time. The sphere is surrounded by
free space with permittivity e0 and zero conductivity.
From (4) we can immediately write down the volume
charge distribution for all time,

P=poe-/T, r<Rl
=,0, r>Rl(5
where 7=e/ol. The total charge on the sphere remains
constant, Q = 1wrR po, but the volume charge is transported
by the Ohmic current to the interface at r = R 2 where it
becomes a surface charge. Enclosing the system by a Gaussian
surface with r > R 2 shows that the external electric field is
time independent,

E, =
Qor,
r > R2 (6)
4·rore
Similarly, applying Gaussian surfaces for r < R and RI < r<
R 2 yields
pore-l'_ Qr e -u
4rsR•,' 0<r<R1
Se
E, = e_•,9 4w7 (7)
2,
4rer RI<r<R2

4P

+2

P/ =0+ +\
p+ + +
R+

+ + +j 60a
+ E 7

%++

- t
at =S4rR
- (I-e i7
2

Figure 3-21 An initial volume charge distribution within an Ohmic conductor decays
exponentially towards zero with relaxation time 7 = es/ and appears as a surface
charge at an interface of discontinuity. Initially uncharged regions are always un-
charged with the charge transported through by the current.
184 Polarization and Conduction

The surface charge density at r = R 2 builds up exponentially

with time:
0f(r = R 2 ) = eoE,(r = R 2 +)- eE,(r = R 2 -)

=- (1-e-)
e(8)
47rR

The charge is carried from the charged region (r < R 1) to the

surface at r = R 2 via the conduction current with the charge
density inbetween (R 1 < r < R2 ) remaining zero:

SI oe-"', 0<r<Rl

J.= E,= oQ e-
4--er - RI<r<R2 (9)

0, r>R2

Note that the total current, conduction plus displacement, is

zero everywhere:
Qro e - "'
-rE4retR' O<r<R
1

- J = Jd = = &_1
at 4 , R 1 <r<R2 (10)
41er2

0, r>R2

3-6-3 Series Lossy Capacitor

(a) Charging transient

To exemplify the difference between resistive and capaci-
tive behavior we examine the case of two different materials in
series stressed by a step voltage first turned on at t = 0, as
shown in Figure 3-22a. Since it takes time to charge up the
interface, the interfacial surface charge cannot instan-
taneously change at t = 0 so that it remains zero at t = 0+. With
no surface charge density, the displacement field is continu-
ous across the interface so that the solution at t - 0+ is the
same as for two lossless series capacitors independent of the
conductivities:
D. =,E,
e 1 = E (11)
The voltage constraint requires that

SE dx = Eia+Eb= V (12)
I*1
 Lossy Media 185

x x x
-9
a+

VW)

!
+ I
Depth d el e 2 V o 1 D 0o2D1
e 2 a+e 1 b El e2 Cl e2
t = 0+

(a)

x x

+ Depthd oo 2V

o 2 a +ab 02 01 02 01

t=(o
t=O (b)

C1 R 1 =-d R2 = 2ld

elld e2ld
Ct ra

Figure 3-22 Two different lossy dielectric materials in series between parallel plate
electrodes have permittivities and Ohmic conductivities that change abruptly across
the interface. (a) At t = 0, right after a step voltage is applied, the interface is
uncharged so that the displacement field is continuous with the solution the same as
for two lossless dielectrics in series. (b) Since the current is discontinuous across the
boundary between the materials, the interface will charge up. In the dc steady state the
current is continuous. (c) Each region is equivalent to a resistor and capacitor in
parallel.

so that the displacement field is

D.(t = 0+) = (13)

e2a +e lb
The total current from the battery is due to both conduction
and displacement currents. At t = 0, the displacement current
186 Polarizationand Conduction

is infinite (an impulse) as the displacement field instan-

taneously changes from zero to (13) to produce the surface
charge on each electrode:
of(x = O)= -of(x = a +b)= D. (14)
After the voltage has been on a long time, the fields
approach their steady-state values, as in Figure 3-22b.
Because there are no more time variations, the current
density must be continuous across the interface just the same
as for two series resistors independent of the permittivities,

J.(t - oo) = a'E1= o'2E2 = (15)

o 2 a +olb
where we again used (12). The interfacial surface charge is
now

(e ol - e 1o2)V
ory(x = a) = 62E2- EIEI = - 1 0 2 )V (16)
2 + o lb
2cr

What we have shown is that for early times the system is

purely capacitive, while for long times the system is purely
resistive. The inbetween transient interval is found by using
(12) with charge conservation applied at the interface:

n- (J-Jl+d (d (D,-DI) =0

0"2E 2 - OlE 1+- d [E 2E 2 - 1 E ]= 0 (17)

dt
With (12) to relate E 2 to El we obtain a single ordinary
differential equation in EI,
dE1 E 1 0'2 V
+ (18)
dt 7• 2a + e lb

where the relaxation time is a weighted average of relaxation

times of each material:

b+ 2a (19)
crb + o-a

Using the initial condition of (13) the solutions for the fields
are

' )
EI= 2V (1-e- + 12V e_1
a-ra+ olb Ena +e lb
(20)
E2= a-1V (1-e "')+ eV -Ue
oaa +oib e6a+elb
 Lossy Media 187

Note that as t -0oo the solutions approach those of (15). The

interfacial surface charge is
(820l1 812)
of (x = a) = e 2E2 e-El = (1-e - ' ')V (21)
o-2a +olb

which is zero at t = 0 and agrees with (16) for t oo.

The total current delivered by the voltage source is
dEl dE2
i(oE +e- - Id= '2EE+e--) Id
di 22 2 di

1 a 2+
00 1\
El 82 02 -\t1
c2 a+ lb 7T) e 2a+eib o 2 a+orlb)

+ 8((t) IdV (22)

E2 a + e lb

where the last term is the impulse current that instan-

taneously puts charge on each electrode in zero time at t = 0:

8( 0)= >f 08(t) dt = 1

o0, t=O0
To reiterate, we see that for early times the capacitances
dominate and that in the steady state the resistances dominate
with the transition time depending on the relaxation times
and geometry of each region. The equivalent circuit for the
system is shown in Figure 3-22c as a series combination of a
parallel resistor-capacitor for each region.

(b) Open Circuit

Once the system is in the dc steady state, we instantaneously
open the circuit so that the terminal current is zero. Then,
using (22) with i = 0, we see that the fields decay indepen-
dently in each region with the relaxation time of each region:

E2V -81, 6
cr2 a + oa b o1
(23)
E2 = V e 2-, 72 =
a 2 a + oab oU2
The open circuit voltage and interfacial charge then decay as
V
V = Ea +E2 b = [o 2 a e-'"'I + o b e '2]
o-2 a + rlb
(24)
of E 2E 2 - 1E = V [E20l e- 'T
2 - e 12 e- /''
a 2 a + or b
188 Polarizationand Conduction

(c) Short Circuit

If the dc steady-state system is instead short circuited, we
set V= 0 in (12) and (18),
Ela + E 2b= 0
(25)
di+
dt 7
where 7 is still given by (19). Since at t = 0 the interfacial
surface charge cannot instantaneously change, the initial
fields must obey the relation

lim(s 2 E 2 - 1IEl)= - -2 -+ E (e 2 1 )V (26)

t-o b o2a + 1orb2
to yield the solutions

E2 b (e 2 o 1 - elo 2 )bV e -/ (27)

E. = e (27)
a (sa +elb) (er2a+ crb)

The short circuit current and surface charge are then

L\•b+ 2a (c'2 a+rb) ea+b (28)

l- e
2E2 -e
= = (E 2 •2)
o 2a +orb

The impulse term in the current is due to the instantaneous

change in displacement field from the steady-state values
found from (15) to the initial values of (26).

(d) Sinusoidal Steady State

Now rather than a step voltage, we assume that the applied
voltage is sinusoidal,
v(t)= Vo cos wt (29)
and has been on a long time.
The fields in each region are still only functions of time and
not position. It is convenient to use complex notation so that
all quantities are written in the form

v(t) = Re (Vo0 e * )
El(t)=Re ($le"), E 2 (t)=Re ( 2 ) (30)

Using carets above a term to designate a complex amplitude,

the applied voltage condition of (12) requires

Pla + sb = Vo
 Lossy Media 189

while the interfacial charge conservation equation of (17)

becomes
aE
22 - 0E 1 +j( (e 2 E2 - E)1 = [02 +j062 ]E 2

-[o 1= 0
+jwes]• (32)
The solutions are
F1 F2 Vo (33)
2+
(iwsE 2) (iaEl+1) [b(al+jcee)+a(o'2+jWae 2 )]
which gives the interfacial surface charge amplitude as

°'= e 2 2• I1I = (34)

[b(a1 +jOai1)+a(a-2+jW6 2 )]

As the frequency becomes much larger than the reciprocal

relaxation times,
(0 >-, >W
,- (35)
61 82

the surface charge density goes to zero. This is because the

surface charge cannot keep pace with the high-frequency
alternations, and thus the capacitive component dominates.
Thus, in experimental work charge accumulations can be
prevented if the excitation frequencies are much faster than
the reciprocal charge relaxation times.
The total current through the electrodes is
I = (ol+ jOa)Elld = ('2+jae 2 )E 2 1d
Id(o1 +jWe )(a2 + jie2)Vo
[b(oi+jOe ) + a(o2 +jis•2)1
Vo
R 2= (36)
R2 R1
R2 C 2ja+1 R 1Cijw+1
with the last result easily obtained from the equivalent circuit
in Figure 3,22c.

3-6-4 Distributed Systems

(a) Governing Equations

In all our discussions we have assumed that the electrodes
are perfectly conducting so that they have no resistance and
the electric field terminates perpendicularly. Consider now
the parallel plate geometry shown in Figure 3-23a, where the
electrodes have a large but finite conductivity ac. The elec-
trodes are no longer equi-potential surfaces since as the cur-
rent passes along the conductor an Ohmic iR drop results.
 190 Polarizationand Conduction

OC RAs Depth d
Bi:ii~L'~i~ji~;a·.iiiiii81:'':':"-'"'.. ·:r::i:iir:i::·:·::l··i~---:a~::::
·· · ·· ·
I I::':':':':':'~*::I:1'::::::·::::~:1;12'
iltt|•::::::":`':"::::-:: ---~·--~ '"···-'· '·'·'··' · ''-:-'·'' ''·'·';rn;l::-:-: :·- :::·--::·a-:-::-::-::·r::
·· :·-::~r:·r:i::T::-::::-:_~:::::::H'::*~:a

v(t)
ThAZ
i~ti

h
R As

AS)
v(i - As) -v(z) = 2i()R As

i(s) -i(s + As) = CAs

dt
+GAsv(s)

Figure 3-23 Lossy parallel plate electrodes with finite Ohmic conductivity o, enclose
a lossy dielectric with permittivity e and conductivity o. (a) This system can be modeled
by a distributed resistor-capacitor network. (b) Kirchoff's voltage and current laws
applied to a section of length Az allow us to describe the system by partial differential
equations.

The current is also shunted through the lossy dielectric so

that less current flows at the far end of the conductor than
near the source. We can find approximate solutions by break-
ing the continuous system into many small segments of length
Az. The electrode resistance of this small section is
Az
R Az =
oad
where R= 1/(o-,ad) is just the resistance per unit length.
We have shown in the previous section that the dielectric can
be modeled as a parallel resistor-capacitor combination,
edAz 1 s
Cz =z
s GAz odAz
C is the capacitance per unit length and G is the conductance
per unit length where the conductance is the reciprocal of the
 Lossy Media 191

resistance. It is more convenient to work with the conduc-

tance because it is in parallel with the capacitance.
We apply Kirchoff's voltage and current laws for the section
of equivalent circuit shown in Figure 3-23b:
v(z -Az)- v(z) = 2i(z)R Az
(39)
dv(z)
i(z)- i(z + Az) = CAz + GAzv(z)
dt
The factor of 2 in the upper equation arises from the equal
series resistances of the upper and lower conductors. Divi-
ding through by Az and taking the limit as Az becomes
infinitesimally small yields the partial differential equations
av
-- = 2iR
az
(40)
- = C. ý + Gv
az at
Taking 8/az of the upper equation allows us to substitute in
the lower equation to eliminate i,
a2v av
--y= 2RC- + 2RGv (41)

which is called a transient diffusion equation. Equations (40)

and (41) are also valid for any geometry whose cross sectional
area remains constant over its length. The 2R represents the
series resistance per unit length of both electrodes, while C
and G are the capacitance and conductance per unit length of
the dielectric medium.

(b) Steady State

If a dc voltage Vo is applied, the steady-state voltage is

d v
-2RGv = v = Asinh /2-Gz +A2coshN z (42)
dz
where the constants are found by the boundary conditions at
z =0 and z = 1,
v(z = 0)= Vo, i(z = 1)= 0 (43)

We take the z = I end to be open circuited. Solutions are

v(z)= V0 cosh 2 (z- 1)

cosh 2A2- l
(44)
1 dv = -G sinh i2½ (z -1)
i(z) Vo
2R dz 2R cosh - l2G1
 192 Polarizationand Conduction

(c) Transient Solution

If this dc voltage is applied as a step at t = 0, it takes time for
the voltage and current to reach these steady-state dis-
tributions. Because (41) is linear, we can use superposition
and guess a solution for the voltage that is the sum of the
steady-state solution in (44) and a transient solution that dies
off with time:
Vo cosh 2/i (z -1) ,
v(z, t) = cosh RG (z) e-a (45)
cosh ,1lR I
At this point we do not know the function vi(z) or a. Substi-
tuting the assumed solution of (45) back into (41) yields the
ordinary differential equation

-+p2v=0, P 2 =2RCa -2RG (46)

which has the trigonometric solutions

^(z)= al sin pz +a 2 cos pz (47)
Since the time-independent' part of (45) already satisfies the
boundary conditions at z = 0, the transient part must be zero
there so that a 2 = 0. The transient contribution to the current
i, found from (40),
=Vo f--G sinh -2 (z -1)
S2R cosh l2RG
(48)
1 diO(z) pa (48)
i(z) =
2R dz 2R cos pz
must still be zero at z = 1, which means that p1 must be an odd
integer multiple of iT/2,
1 I 2+ G
pl=(2n+1)-~a. ((2n+l)= + , n = 0, 1,2,
2 2RC 21 C'
(49)
Since the boundary conditions allow an infinite number of
values of a, the most general solution is the superposition of
all allowed solutions:

cosh 2R (z -1) )
v(z, t) = VDo + Y A. sin(2n+ 1) e-"
cosh N/i- 1 0 21
(50)

This solution satisfies the boundary conditions but not the

initial conditions at t = 0 when the voltage is first turned on.
Before the voltage source is applied, the voltage distribution
throughout the system is zero. It must remain zero right after

1
 Lossy Media 193

being turned on otherwise the time derivative in (40) would

be infinite, which requires nonphysical infinite currents. Thus
we impose the initial condition
cosh -ý2_R (z -1) rz
v(z, t=0)=0O= Vo s + YI An sin(2n+1) -
cosh ,l . =o 21
(51)
We can solve for the amplitudes An by multiplying (51)
through by sin (2m+ 1) rz/21 and then integrating over z
from 0 to 1:

Vo2• cosh 2R-G(z - 1)sin(2m+1) dz

cosh I2RG
o 21
10 Z iTrz
+J An sin(2n +1) -sin(2m +1) - dz (52)
n=o 21 21

The first term is easily integrated by writing the hyperbolic

cosine in terms of exponentials,* while the last term integrates
to zero for all values of m not equal to n so that the ampli-
tudes are
1 rVo (2n + 1)
2
A= 2RG + [(2n + 1) /rl21]
The total solutions are then

Vo cosh N2--2(z- 1)
v(z,t)=
cosh v i;R I
7rVo (2n + 1) sin [(2n + 1) (lrz/21)]e-" -'
12 n=oo 2RG+[(2n+l) (r/21)]2

i(z, t)= I av
1
2R az
Vo 12R sinhVi (z - 1)
cosh I
l/2
2 -
SrVo (2n + 1)2 cos [(2n + 1)(wz/21)] e"-'
41R ,=o 2RG + [(2n + 1) ('rj21)] 2

* cosh a(z - 1)sin bz dz

=a+ b [a sin bz sinh a(z -l)-b cos bz cosh a(z -1)]

0 mm n
sin (2n + l)bz sin (2m+ 1)bz dz = n
11/2 m=n
194 Polarizationand Conduction

The fundamental time constant corresponds to the smallest

value of a, which is when n = 0:

1 C
To= -= 2 (55)
ao G+

For times long compared to •0 the system is approximately in

the steady state. Because of the fast exponential decrease for
times greater than zero, the infinite series in (54) can often be
approximated by the first term. These solutions are plotted in
Figure 3-24 for the special case where G = 0. Then the
voltage distribution builds up from zero to a constant value
diffusing in from the left. The current near z = 0 is initially
very large. As time increases, with G = 0, the current every-
where decreases towards a zero steady state.

3-6-5 Effects of Convection

We have seen that in a stationary medium any initial charge

density decays away to a surface of discontinuity. We now
wish to focus attention on a dc steady-state system of a
conducting medium moving at constant velocity Ui., as in
Figure 3-25. A source at x = 0 maintains a constant charge
density po. Then (3) in the dc steady state with constant

2
To =
8RC1
,

vf=, I
VO

a/I /fl
Figure 3-24 The transient voltage and current spatial distributions for various times
for the lossy line in Figure 3-23a with G = 0 for a step voltage excitation at z = 0 with
the z = I end open circuited. The diffusion effects arise because of the lossy electrodes
where the longest time constant is T0 = 8RC/2IIr'.
 Lossy Media 195

p (x = 0) = Po

P1 W ý poe x1AM =E
pIm a
, 1 PU

-T-

I
Figure 3-25 A moving conducting material with velocity Ui, tends to take charge
injected at x =0 with it. The steady-state charge density decreases exponentially from
the source.

velocity becomes
d+p
+"a
dpf, a P =0 (56)
dx EU

which has exponentially decaying solutions

pf = Po e - a
, 1= (57)

where 1. represents a characteristic spatial decay length. If

the system has cross-sectional area A, the total charge q in the
system is
q= pfA dx = polA (58)

3-6-6 The Earth and its Atmosphere as a Leaky Spherical Capacitor*

In fair weather, at the earth's surface exists a dc electric

field with approximate strength of 100 V/m directed radially
toward the earth's center. The magnitude of the electric field
decreases with height above the earth's surface because of the
nonuniform electrical conductivity oa(r) of the atmosphere
approximated as
cr(r)= ro + a(r- R )2 siemen/m (59)
where measurements have shown that
ro- 3 10-14

a .5 x 10-
20 (60)

* M. A. Uman, "The EarthandIts Atmosphere asa Leaky SphericalCapacitor,"Am. J. Phys.

V. 42, Nov. 1974, pp. 1033-1035.
196 Polarizationand Conduction

and R -6 x 106 meter is the earth's radius. The conductivity

increases with height because of cosmic radiation in the lower
atmosphere. Because of solar radiation the atmosphere acts
as a perfect conductor above 50 km.
In the dc steady state, charge conservation of Section 3-2-1
with spherical symmetry requires
18 C
VJ= (rJ,)= > J, = (r)E,= (61)
r2 8r r
where the constant of integration C is found by specifying the
surface electric field E,(R)* - 100 V/m
O(R)E,(R)R 2
J,(r) = 2 (62)

At the earth's surface the current density is then

J,(R) = o(R)E,(R) = roE,(R) -- 3 x 10-12 amp/m2 (63)
The total current directed radially inwards over the whole
earth is then
I = IJ,(R)47rR 21 - 1350 amp (64)
The electric field distribution throughout the atmosphere
is found from (62) as
J (r )
, =(R)E,(R)R2
E,(r) 2(r) (65)
o(r) r o(r)
The surface charge density on the earth's surface is

(r = R) = EoE,(R) - -8.85 x 10- 1' Coul/m 2 (66)

This negative surface charge distribution (remember: E,(r) <

0) is balanced by positive volume charge distribution
throughout the atmosphere

Eo 2 soo(R)E,(R)R 2 d 1
p,(r)=eoV - E= r--(rE,)= 2 L\(
r~r 22 r dr o(r)
S-soo(R)E,(R)R 2
(67)
r2((r)) 2a(r-R)

The potential difference between the upper atmosphere

and the earth's surface is

V= J- E,(r)dr

o(R)E(R)2r2[o[o+a(r-R)2]
 Field-dependentSpace ChargeDistributions 197

1 (R2 t)

r(R 2 + ) C10(R+'2 )2
( 1 a la
r(R)E,(R)
(68)
a(R' + 0)'

Using the parameters of (60), we see that rola<< R 2 so that

(68) approximately reduces to

aR 2 aR2
IoE,(R)n (69)
- 384,000 volts
If the earth's charge were not replenished, the current flow
would neutralize the charge at the earth's surface with a time
constant of order

7 = £0
-= 300 seconds (70)
0o

It is thought that localized stormy regions simultaneously

active all over the world serve as "batteries" to keep the earth
charged via negatively chairged lightning to ground and
corona at ground level, producing charge that moves from
ground to cloud. This thunderstorm current must be
upwards and balances the downwards fair weather current of
(64).

3.7 FIELD-DEPENDENT SPACE CHARGE DISTRIBUTIONS

A stationary Ohmic conductor with constant conductivity

was shown in Section 3-6-1 to not support a steady-state
volume charge distribution. This occurs because in our clas-
sical Ohmic model in Section 3-2-2c one species of charge
(e.g., electrons in metals) move relative to a stationary back-
ground species of charge with opposite polarity so that charge
neutrality is maintained. However, if only one species of
198 Polarizationand Conduction

charge is injected into a medium, a net steady-state volume

charge distribution can result.
Because of the electric force, this distribution of volume
charge py contributes to and also in turn depends on the
electric field. It now becomes necessary to simultaneously
satisfy the coupled electrical and mechanical equations.

3-7-1 Space Charge Limited Vacuum Tube Diode

In vacuum tube diodes, electrons with charge -e and mass

m are boiled off the heated cathode, which we take as our zero
potential reference. This process is called thermionic emis-
sion. A positive potential Vo applied to the anode at x = l
accelerates the electrons, as in Figure 3-26. Newton's law for a
particular electron is
dv dV
m = - eE = e (1)
dt dx
In the dc steady state the velocity of the electron depends only
on its position x so that
dv dv dx dv d 2 d
m-= = mymv ( )= -(e V) (2)
dt dx dt dx dx dx
V0
+II
|Ill

1ll

= JoA

-e +

2eV 1/ 2 +
V= [ m - E

Area A
J -Joix

+
Cathode Anode

0I II - x
0 1

(a) (b)
Figure 3-26 Space charge limited vacuum tube diode. (a) Thermionic injection of
electrons from the heated cathode into vacuum with zero initial velocity. The positive
anode potential attracts the electrons whose acceleration is proportional to the local
electric field. (b) Steady-state potential, electric field, and volume charge distributions.
 Field-dependentSpace Charge Distributions 199

With this last equality, we have derived the energy conser-

vation theorem

2 - (3)
ddx [mv2 -eV] = O mv eV= const

where we say that the kinetic energy 2mv2 plus the potential
energy -eV is the constant total energy. We limit ourselves
here to the simplest case where the injected charge at the
cathode starts out with zero velocity. Since the potential is also
chosen to be zero at the cathode, the constant in (3) is zero.
The velocity is then related to the electric potential as

= (2e V)I/ 1/ (4)

In the time-independent steady state the current density is

constant,
dJx
JJ=O - O=J = -Joi. (5)
dx
and is related to the charge density and velocity as
In 1/2
= 1 9
o = -PfvjpJf -JO(2e) V- 1
2
(6)

Note that the current flows from anode to cathode, and

thus is in the negative x direction. This minus sign is
incorporated in (5) and (6) so that Jo is positive. Poisson's
equation then requires that

V2V= -P dV Jo 'm1/2v-
\•eW (7)
Power law solutions to this nonlinear differential equation are
guessed of the form
V = Bx (8)
which when substituted into (7) yields

Bp(p - 1)x -2 = o (; B-1/2X-02

12 (9)/

For this assumed solution to hold for all x we require that

p 4
-2= -p = (10)
2 3
which then gives us the amplitude B as

B [ 4= /22/s (11)

I
200 Polarizationand Conduction

so that the potential is

V(x)-= 9'- 21 2Ex 4 (12)

The potential is zero at the cathode, as required, while the

anode potential Vo requires the current density to be

V(x = ) = Vo = I 1/22/ 4/3

4e \2e /2 /2
o= V;9 (13)

which is called the Langmuir-Child law.

The potential, electric field, and charge distributions are
then concisely written as

V(x) = Vo(! )

dV(x) 4 Vo (I\s
E(x) = - - ) (14)
2
dE(x) 4 Vo (x)- /s

and are plotted in Figure 3-26b. We see that the charge

density at the cathode is infinite but that the total charge
between the electrodes is finite,

q-= p(x)Adx= -- ve-A (15)

being equal in magnitude but opposite in sign to the total

surface charge on the anode:
4Vo
qA= of(x=1)A= -eE(x=1)A= + -4-A (16)
3 1

There is no surface charge on the cathode because the electric

field is zero there.
This displacement x of each electron can be found by
substituting the potential distribution of (14) into (4),

v S(2eVo
-
2
)2/ i s
( ~-5 _dx =(--22eVo , 1/2dt (17)

which integrates to

x= 7iýi) ' (18)

 Field-dependentSpace ChargeDistributions 201

The charge transit time 7 between electrodes is found by

solving (18) with x = 1:

= 3(1 (19)

For an electron (m = 9.1 x 10 - s ' kg, e = 1.6 10-' 9 coul) with

100 volts applied across 1 = 1 cm (10 - 2 m) this time is 7~
5 x 10 - 9 sec. The peak electron velocity when it reaches the
anode is v(x = 1)-6x 106 m/sec, which is approximately 50
times less than the vacuum speed of light.
Because of these fast response times vacuum tube diodes
are used in alternating voltage applications for rectification as
current only flows when the anode is positive and as
nonlinear circuit elements because of the three-halves power
law of (13) relating current and voltage.

3-7-2 Space Charge Limited Conduction in Dielectrics

Conduction properties of dielectrics are often examined by

injecting charge. In Figure 3-27, an electron beam with cur-
rent density J = -Joi, is suddenly turned on at t = 0.* In media,
the acceleration of the charge is no longer proportional to the
electric field. Rather, collisions with the medium introduce a
frictional drag so that the velocity is proportional to the elec-
tric field through the electron mobility /A:
v = -AE (20)
As the electrons penetrate the dielectric, the space charge
front is a distance s from the interface where (20) gives us

ds/dt = -tE(s) (21)

Although the charge density is nonuniformly distributed
behind the wavefront, the total charge Q within the dielectric
behind the wave front at time t is related to the current
density as
JoA = pE.A = - Q/t Q = -JoAt (22)
Gauss's law applied to the rectangular surface enclosing all
the charge within the dielectric then relates the fields at the
interface and the charge front to this charge as

E- dS = (E(s)-oE(0)]A = Q = -JoAt (23)

* See P. K. Watson, J. M. Schneider, andH. R. Till, Electrohydrodynamic Stability of Space

Charge Limited Currents In Dielectric Liquids. IL. ExperimentalStudy, Phys. Fluids 13
(1970), p. 1955.
 202 Polarization and Conduction

Electron beam
A= - 1-i

Space charge limited Surface of integration for Gauss's

condition: E(O)=0w: Es- A==-At
. eo law: fE__
[(s)-eoE(OI]A=Q=-JoAe
I
-
0 /
sltl
~+e-= E Moving space charge front
Se, p =0 E =ds
Electrode area -Es)
7Electrode area A

Sjo 1/2 t
E• j =

Figure 3-27 (a) An electron beam carrying a current -Joi, is turned on at t = 0. The
electrons travel through the dielectric with mobility gp.(b) The space charge front, at a
distance s in front of the space charge limited interface at x = 0, travels towards the
opposite electrode. (c) After the transit time t, = [2el/IJo] 1'2 the steady-state potential,
electric field, and space charge distributions.

The maximum current flows when E(O) = 0, which is called

space charge limited conduction. Then using (23) in (21)
gives us the time dependence of the space charge front:
ds iJot iLJot 2
= O s(t) = (24)
dt e 2e
Behind the front Gauss's law requires
dE~, P Jo dE. Jo
-=E E dE. (25)
dx e eAE. x dx EL-

~I
 Field-dependent Space Charge Distributions 203

while ahead of the moving space charge the charge density is

zero so that the current is carried entirely by displacement
current and the electric field is constant in space. The spatial
distribution of electric field is then obtained by integrating
(25) to
E•= -I2JOx , 0:xs(t) (26)
-%2_Jos/e, s(t)Sxli

while the charge distribution is

O -x s(t)
S=dE -eJo/(2x), (27)
Pf=e (27)
dx 0, s(t):x5l
as indicated in Figure 3-27b.
The time dependence of the voltage across the dielectric is
then

v(t) = Edx = -xojx d+ d

Jolt Aj2t3
e 6 , s(t)_l (28)
6 6E2

These transient solutions are valid until the space charge

front s, given by (24), reaches the opposite electrode with s = I
at time
€= 12-e11to (29)

Thereafter, the system is in the dc steady state with the

terminal voltage Vo related to the current density as

9 e;L V2
Jo= 8- (30)
8 1S

which is the analogous Langmuir-Child's law for collision

dominated media. The steady-state electric field and space
charge density are then concisely written as

3 Vo 2 dE 3E V0 1 (31)
2 1 dx- 4 1.
and are plotted in Figure 3-27c.
In liquids a typical ion mobility is of the order of
10-7 m 2 /(volt-sec) with a permittivity of e = 2e0
1.77Ox 10- farad/m. For a spacing of I= O-2 m with a
potential difference of Vo = 10 V the current density of (30)
is Jo 2 10-4 amp/m2 with the transit time given by (29)
rr0.133 sec. Charge transport times in collison dominated
media are much larger than in vacuum.
204 Polarizationand Conduction

3-8 ENERGY STORED IN A DIELECTRIC MEDIUM

The work needed to assemble a charge distribution is
stored as potential energy in the electric field because if the
charges are allowed to move this work can be regained as
kinetic energy or mechanical work.

3-8-1 Work Necessary to Assemble a Distribution of Point Charges

(a) Assembling the Charges

Let us compute the work necessary to bring three already
existing free charges qj, q2, and qs from infinity to any posi-
tion, as in Figure 3-28. It takes no work to bring in the first
charge as there is no electric field present. The work neces-
sary to bring in the second charge must overcome the field
due to the first charge, while the work needed to bring in the
third charge must overcome the fields due to both other
charges. Since the electric potential developed in Section
2-5-3 is defined as the work per unit charge necessary to bring
a point charge in from infinity, the total work necessary to
bring in the three charges is
q, \+ Iq__+ q2
W=
q,()+q2r +qsl + (1)
4 irerl 2 ' 4wer
15 4rrer 2sl
where the final distances between the charges are defined in
Figure 3-28 and we use the permittivity e of the medium. We
can rewrite (1) in the more convenient form
W= _[ q2 + qs +q q, + 3
2 4 4erTsJ,
4erel2 4ITerl2 4r823J

q + 4q 2 (2)
14rer3s 47rer23s

I
/
/
/
/
/
/
I
/
/
/
p.

Figure 3-28 Three already existing point charges are brought in from an infinite
distance to their final positions.
 Energy Stored in a Dielectric Medium 205

where we recognize that each bracketed term is just the

potential at the final position of each charge and includes
contributions from all the other charges, except the one
located at the position where the potential is being evaluated:

W= 2[q V1 +q2 V 2 +q 3 V s ] (3)

Extending this result for any number N of already existing

free point charges yields
1N
W= E 4qV,, (4)

The factor of - arises because the potential of a point charge

at the time it is brought in from infinity is less than the final
potential when all the charges are assembled.

(b) Binding Energy of a Crystal

One major application of (4) is in computing the largest
contribution to the binding energy of ionic crystals, such as
salt (NaCI), which is known as the Madelung electrostatic
energy. We take a simple one-dimensional model of a crystal
consisting of an infinitely long string of alternating polarity
point charges ±q a distance a apart, as in Figure 3-29. The
average work necessary to bring a positive charge as shown in
Figure 3-29 from infinity to its position on the line is obtained
from (4) as

1 2q 1 1 1 1
W -- 1+---+---+
- -- -- (5)
2 47rea 2 3 4 5 6
The extra factor of 2 in the numerator is necessary because
the string extends to infinity on each side. The infinite series
is recognized as the Taylor series expansion of the logarithm

2 3 4 5
In (l +x)=x--+---+-- (6)
2345

+q -q +q -q +q -q +q -q +q -q +q

Figure 3-29 A one-dimensional crystal with alternating polarity charges q a dis-

tance a apart.
206 Polarizationand Conduction

where x = 1 so that*

W = -4ia In 2 (7)
This work is negative because the crystal pulls on the charge
as it is brought in from infinity. This means that it would take
positive work to remove the charge as it is bound to the
crystal. A typical ion spacing is about 3 A (3 x 10-'1 m) so that
if q is a single proton (q = 1.6x 10- coul), the binding energy
is W 5.3x10-' joule. Since this number is so small it is
usually more convenient to work with units of energy per unit
electronic charge called electron volts (ev), which are obtained
by dividing W by the charge on an electron so that, in this
case, W -3.3 ev.
If the crystal was placed in a medium with higher permit-
tivity, we see from (7) that the binding energy decreases. This
is why many crystals are soluble in water, which has a relative
dielectric constant of about 80.

3-8-2 Work Necessary to Form a Continuous Charge Distribution

Not included in (4) is the self-energy of each charge itself

or, equivalently, the work necessary to assemble each point
charge. Since the potential V from a point charge q is pro-
portional to q, the self-energy is proportional q2 . However,
evaluating the self-energy of a point charge is difficult
because the potential is infinite at the point charge.
To understand the self-energy concept better it helps to
model a point charge as a small uniformly charged spherical
volume of radius R with total charge Q = srRspo. We assem-
ble the sphere of charge from spherical shells, as shown in
Figure 3-30, each of thickness dr. and incremental charge
dq. = 4rr2 drpo. As we bring in the nth shell to be placed at
radius r. the total charge already present and the potential
there are
4 q. rPoo

q3 4wrer. 3e

* Strictly speaking, this series is only conditionallyconvergentfor x = 1 and its sum depends on
the groupingof individual terms. If the series in (6) for x = 1 is rewritten as

1 1 1 1 1
1----+-----+...+
+ 1I---- .. 1 1
I.... kl
2 4 3 6 8 2k-1 4k-2 4k
then its sum is 2 In 2. [See J.Pleinesand S.Mahajan,On ConditionallyDivergentSeries and
a PointChargeBetween Two ParallelGrounded Planes,Am. J. Phys. 45 (1977) p. 868. ]
 Energy Stored in a DielectricMedium 207

4.
= Po4r dr,
Figure 3-30 A point charge is modelled as a small uniformly charged sphere. It is
assembled by bringing in spherical shells of differential sized surface charge elements
from infinity.

so that the work required to bring in the nth shell is

dW. = V. dq. = dr. (9)

The total work necessary to assemble the sphere is obtained

by adding the work needed for each shell:

W= dW. = 3 e dr = 20reR (10)

For a finite charge Q of zero radius the work becomes

infinite. However, Einstein's theory of relativity tells us that
this work necessary to assemble the charge is stored as energy
that is related to the mass as
W=mc= SQ 2 3Q 2
W=mc9= RR = , (11)
20ireR 204rEmc2
which then determines the radius of the charge. For the case
of an electron (Q = 1.6 x 10- 19 coul, m = 9.1 x 10- s1 kg) in free
space (e = E0 = 8.854 X 10-12 farad/m), this radius is

3(1.6 x 10-19)
Relectron =-12)9 X 1
20ir(8.854x 10-1)(9x
10 10'-)(3 x 10 )
- 15
-1.69x 10 m (12)
We can also obtain the result of (10) by using (4) where each
charge becomes a differential element dq, so that the sum-
mation becomes an integration over the continuous free
charge distribution:

W= f, Vdq, (13)
Ilql
q
208 Polarizationand Conduction

For the case of the uniformly charged sphere, dqr = Po dV, the
final potential within the sphere is given by the results of
Section 2-5-5b:

V=P (R-2_L (14)

2e \ 3/
Then (13) agrees with (10):
2
P R (R 3 r d2 41rpoR5 3SQ
W= 1 oVdV=41rp _ rdr
2 4e 3 ) 15e 201reR
(15)
Thus, in general, we define (13) as the energy stored in the
electric field, including the self-energy term. It differs from
(4), which only includes interaction terms between different
charges and not the infinite work necessary to assemble each
point charge. Equation (13) is valid for line, surface, and
volume charge distributions with the differential charge ele-
ments given in Section 2-3-1. Remember when using (4) and
(13) that the zero reference for the potential is assumed to be
at infinity. Adding a constant Vo to the potential will change
the energy u'nless the total charge in the system is zero

W= f (V+ Vo) dq

= Vdq, + Vo f
= Vdq1 (16)

3-8-3 Energy Density of the Electric Field

It is also convenient to express the energy W stored in a

system in terms of the electric field. We assume that we have a
volume charge distribution with density pf. Then, dqf= pydV,
where p, is related to the displacement field from Gauss's law:

fp, VdV=½
W=J V(V. D) dV (17)

Let us examine the vector expansion

V -(VD)=(D.V)V+ V(V. D)= V(V D)=V *(VD)+D'E

(18)
where E= -V V. Then (17) becomes

W= D. EdV+t V (VD)dV (19)

__
 Energy Stored in a Dielectric Medium 209

The last term on the right-hand side can be converted to a

surface integral using the divergence theorem:

SV (VD)dV = SVDdS (20)

If we let the volume V be of infinite extent so that the enclos-

ing surface S is at infinity, the charge distribution that only
extends over a finite volume looks like a point charge for
which the potential decreases as 1/r and the displacement
vector dies off as l/r 2 . Thus the term, VD at best dies off as
1/r s . Then, even though the surface area of S increases as r ,
the surface integral tends to zero as r becomes infinite as l/r.
Thus, the second volume integral in (19) approaches zero:

lim V (VD)dV= VD"dS=O (21)

This conclusion is not true if the charge distribution is of

infinite extent, since for the case of an infinitely long line or
surface charge, the potential itself becomes infinite at infinity
because the total charge on the line or surface is infinite.
However, for finite size charge distributions, which is always
the case in reality, (19) becomes

W= space D.EdV

= eE 2 dV (22)
all space

where the integration extends over all space. This result is

true even if the permittivity e is a function of position. It is
convenient to define the energy density as the positive-
definite quantity:
w = 2eE2 joule/m s [kg-m- -s- ]
(23)
where the total energy is

W
W= space wdV
= V(24) wd V (24)

Note that although (22) is numerically equal to (13), (22)

implies that electric energy exists in those regions where a
nonzero electric field exists even if no charge is present in that
region, while (13) implies that electric energy exists only
where the charge is nonzero. The answer as to where the
energy is stored-in the charge distribution or in the electric
field-is a matter of convenience since you cannot have one
without the other. Numerically both equations yield the same
answers but with contributions from different regions of
space.
210 Polarizationand Conduction

3-8-4 Energy Stored in Charged Spheres

(a) Volume Charge

We can also find the energy stored in a uniformly charged
sphere using (22) since we know the electric field in each
region from Section 2-4-3b. The energy density is then

Q' r>R
w = !E = 2r (25)
2 Q22 ,
32r2eR r<R

with total stored energy

W= wdV

2 -R 4 - Q3
r dr + (26)

which agrees with (10) and (15).

(b) Surface Charge

If the sphere is uniformly charged on its surface Q =
4nrR2 cro, the potential and electric field distributions are

4 0, r<R
V(r)= E, = (27)

L4r Q----§, 4Irer r>R

Using (22) the energy stored is

2
e Q \2 C dr Q
2 4ie 41JR r 8weR (28)

This result is equally as easy obtained using (13):

W= ooV(r = R) dS

= 'oV(r = R)4rR2 8 R (29)

The energy stored in a uniformly charged sphere is 20%

larger than the surface charged sphere for the same total
charge Q. This is because of the additional energy stored
throughout the sphere's volume. Outside the sphere (r > R)
the fields are the same as is the stored energy.
 Energy Stored in a Dielectric Medium 211

(c) Binding Energy of an Atom

In Section 3-1-4 we modeled an atom as a fixed positive
point charge nucleus Q with a surrounding uniform spheri-
cal cloud of negative charge with total charge -Q, as in
Figure 3-31. Potentials due to the positive point and negative
volume charges are found from Section 2-5-5b as

V+(r) = Q
41reor
3Q 2 Q2
3 (R
8weoRs L),
3 r<R
V_(r) =
Sr >R (30)
41reor
The binding energy of the atom is easily found by super-
position considering first the uniformly charged negative
sphere with self-energy given in (10), (15), and (26) and then
adding the energy of the positive point charge:
9
3Q2 2Q
W= +Q[V_(r = 0)] =- (31)
207reoR 407reoR

This is the work necessary to assemble the atom from

charges at infinity. Once the positive nucleus is in place, it
attracts the following negative charges so that the field does
work on the charges and the work of assembly in (31) is
negative. Equivalently, the magnitude of (31) is the work
necessary for us to disassemble the atom by overcoming the
attractive coulombic forces between the opposite polarity
charges.
When alternatively using (4) and (13), we only include the
potential of the negative volume charge at r = 0 acting on the
positive charge, while we include the total potential due to
both in evaluating the energy of the volume charge. We do

Total negative
charge - Q

2
-- - V() r--- 3Q(R _r2/3)
3
Q 4reor 8weoR

- -=- - - 2
- -_ R3
dr 4weor 4weo

Figure 3-31 An atom can be modelled as a point charge Q representing the nucleus,
surrounded by a cloud of uniformly distributed electrons with total charge - Q within
a sphere of radius R.
212 Polarizationand Conduction

not consider the infinite self-energy of the point charge that

would be included if we used (22):

W= QV.(r = 0)-2 -o [V+(r)+ V-(r)] i- dr

2 R
- 3Q 3Q r 3 r2 r
4
dr
16EoR 8eoR 2R 2R
9Q2
- (32)
40wsreoR

3-8-5 Energy Stored in a Capacitor

In a capacitor all the charge resides on the electrodes as a

surface charge. Consider two electrodes at voltage VI and V2
with respect to infinity, and thus at voltage difference V=
V 2 - V1, as shown in Figure 3-32. Each electrode carries
opposite polarity charge with magnitude Q. Then (13) can be
used to compute the total energy stored as

W=[ Vio. dSi+ V 20 2 ds 2 (33)

Since each surface is an equipotential, the voltages VI and V 2

may be taken outside the integrals. The integral then reduces
to the total charge ± Q on each electrode:

w=4VIJ oa dS 1+ V2 2 2dS
-Q Q

=(V2 V- )Q=Q QV (34)

e(r)
V= V 2 -V

S2-

W= Q-V2= cV = C
Q2/C

Figure 3-32 A capacitor stores energy in the electric field.

 Fieldsand Their Forces 213

Since in a capacitor the charge and voltage are linearly related

through the capacitance
Q = CV (35)
the energy stored in the capacitor can also be written as

2 (36)
W=QV=CV -

This energy is equivalent to (22) in terms of the electric field

and gives us an alternate method to computing the capaci-
tance if we know the electric field distribution.

EXAMPLE 3-3 CAPACITANCE OF AN ISOLATED SPHERE

A sphere of radius R carries a uniformly distributed sur-

face charge Q. What is its capacitance?

SOLUTION

The stored energy is given by (28) or (29) so that (36) gives

us the capacitance:
C = Q 2 /2 W = 41reR

3.9 FIELDS AND THEIR FORCES

3-9-1 Force Per Unit Area on a Sheet of Surface Charge

A confusion arises in applying Coulomb's law to find the
perpendicular force on a sheet of surface charge as the
normal electric field is different on each side of the sheet.
Using the over-simplified argument that half the surface
charge resides on each side of the sheet yields the correct
force

f= o-,(E +E) dS (1)

where, as shown in Figure 3-33a, El and E 2 are the electric

fields on each side of the sheet. Thus, the correct field to use
is the average electric field -(El + E2) across the sheet.
For the tangential force, the tangential components of E
are continuous across the sheet (El, = E2 =-E,) so that

, = s of(E, +E ,)
2 dS= IoyE, dS (2)
 214 Polarizationand Conduction

:0-b 8 -e

E2

El, =E 2V =EI

Eln
E2 n -- Eln e P0e
E
Elt
/E,

90x

Figure 3-33 (a) The normal component of electric field is discontinuous across the
sheet of surface charge. (b) The sheet of surface charge can be modeled as a thin layer
of volume charge. The electric field then varies linearly across the volume.

The normal fields are discontinuous across the sheet so that

the perpendicular force is

o = e(E 2 , - E.) f. = e (E2. - E1 .)(E. + E 2 ) dS

e(E. .- E .) dS

To be mathematically rigorous we can examine the field

transition through the sheet more closely by assuming the
surface charge is really a uniform volume charge distribution
po of very narrow thickness 8, as shown in Figure 3-33b. Over
the small surface element dS, the surface appears straight so
that the electric field due to the volume charge can then only
vary with the coordinate x perpendicular to the surface. Then
the point form of Gauss's law within the volume yields

d = P= E. = + const
dx e e

The constant in (4) is evaluated by the boundary conditions

on the normal components of electric field on each side of the

~I
 Fields and Their Forces 215

sheet

E.(x = O)= Ej., E.(x = 8)= E2. (5)

so that the electric field is

x
Ex= (E 2 .- E,,) +El, (6)

As the slab thickness 8 becomes very small, we approach a

sheet charge relating the surface charge density to the dis-
continuity in electric fields as

lim po8 = a- = e (E2n -El.) (7)

Po--O
8-0

Similarly the force per unit area on the slab of volume charge
is

F. = poE. dx

=' po((E 2 n-E

1 .) +E,. dx
2 8

1 ,.x
=[po(E2.-EE)--+E

Po8
= (E,,+E2.) (8)
2

In the limit of (7), the force per unit area on the sheet of
surface charge agrees with (3):

=
lim F.= -(E.+E2.) (E -El (9)
PoB=rf 2 2

3-9-2 Forces on a Polarized Medium

(a) Force Density

In a uniform electric field there is no force on a dipole
because the force on each charge is equal in magnitude but
opposite in direction, as in Figure 3-34a. However, if the
dipole moment is not aligned with the field there is an align-
ing torque given by t =p xE. The torque per unit volume T
on a polarized medium with N dipoles per unit volume is
then
T= Nt= Np xE= PxE
216 Polarizationand Conduction

qE
+ d) - E(r)l
V) E(r)
) E(r)
F-

F=F, +F_- =0
t=dxF÷ =qdxE=pxE

Unifprm field Nonuniform field

(a) (b)

Figure 3-34 (a) A torque is felt by a dipole if its moment is not aligned with the
electric field. In a uniform electric field there is no net force on a dipole because the
force on each charge is equal in magnitude but opposite in direction. (b) There is a net
force on a dipole only in a nonuniform field.

For a linear dielectric, this torque is zero because the

polarization is induced by the field so that P and E are in the
same direction.
A net force can be applied to a dipole if the electric field is
different on each end, as in Figure 3-34b:

f = -q[E(r)-E(r+d)] (11)

For point dipoles, the dipole spacing d is very small so that the
electric field at r + d can be expanded in a Taylor series as

E(r + d) - E(r) + d. - E(r)+ d, E(r) + d. a E(r)

ax ay az
= E(r) + (d - V)E(r) (12)
Then the force on a point dipole is

f = (qd V)E(r)= (p V)E(r) (13)

If we have a distribution of such dipoles with number

density N, the polarization force density is

F = Nf = (Np - V)E = (P . V)E (14)

Of course, if there is any free charge present we must also

add the coulombic force density pfE.

(b) Permanently Polarized Medium

A permanently polarized material with polarization Poi, is
free to slide between parallel plate electrodes, as is shown in
Figure 3-35.
 Fields and Their Forces 217

Y
St
I - II

I
P0 j 0
QQ
@
QQQ~r
@ I E° VQi_
•o
vEi
++
+++++ + +

Depth d
I

E+

F =q(E -- E_)

P= (e -Eo)E

++++++÷

Depth d

xo

(b)
Figure 3-35 (a) A permanently polarized electret partially inserted into a capacitor
has a force on it due to the Coulombic attraction between the dipole charges and the
surface charge on the electrodes. The net force arises in the fringing field region as the
end of the dipole further from the electrode edge feels a smaller electric field.
Depending on the voltage magnitude and polarity, the electret can be pulled in or
pushed out of the capacitor. (b) A linear dielectric is always attracted into a free space
capacitor because of the net force on dipoles in the nonuniform field. The dipoles are
now aligned with the electric field, no matter the voltage polarity.

We only know the electric field in the interelectrode region

and from Example 3-2 far away from the electrodes:
Vo Po
E,(x = Xo) = -, E,(x = -oo)= (15)
s EO
218 Polarizationand Conduction

Unfortunately, neither of these regions contribute to the

force because the electric field is uniform and (14) requires a
field gradient for a force. The force arises in the fringing
fields near the electrode edges where the field is nonuniform
and, thus, exerts less of a force on the dipole end further
from the electrode edges. At first glance it looks like we have a
difficult problem because we do not know the fields where the
force acts. However, because the electric field has zero curl,

VxE= E0 OE, (16)

y ax=
the x component of the force density can be written as

ay
F.=P,aE,
'ax
aE,
= -(PE,)- E, a (17)
ax ax
The last term in (17) is zero because P, = Po is a constant. The
total x directed force is then

f J F. dx dy dz
£o a
1= . ,o T- (P,E,) dxdydz (18)

We do the x integration first so that the y and z integrations

are simple multiplications as the fields at the limits of the x
integration are independent of y and z:
Posd
fJ = PoEsdlx- = PoVod +- 6 (19)
o
There is a force pulling the electret between the electrodes
even if the voltage were zero due to the field generated by the
surface charge on the electrodes induced by the electret. This
force is increased if the imposed electric field and polarization
are in the same direction. If the voltage polarity is reversed,
the force is negative and the electret is pushed out if the
magnitude of the voltage exceeds Pos/eo.

(c) Linearly Polarized Medium

The problem is different if the slab is polarized by the
electric field, as the polarization will then be in the direction
of the electric field and thus have x and y components in the
fringing fields near the electrode edges where the force
 Fields and Their Forces 219

arises, as in Figure 3-35b. The dipoles tend to line up as

shown with the positive ends attracted towards the negative
electrode and the negative dipole ends towards the positive
electrode. Because the farther ends of the dipoles are in a
slightly weaker field, there is a net force to the right tending
to draw the dielectric into the capacitor.
The force density of (14) is

F,= P. . +P, = (E o)E. +EYE (20)

ax ay ax ay
RrecaeP the elecptric field i curl free as given in (1) the
-- ------- -. - - , a -
force density is further simplified to
(e - o) 2)
F. -) (E2 +EY) (21)
2 ax
The total force is obtained by integrating (21) over the
volume of the dielectric:

d
fjI=0-o"of o(E-2 ax( (Ej +Ej) dxdyddz
(e - eo)sd(E +E)1 O (E- EO) V-(d
2 (E+E )=-M- 2 (22)
2 2 s
where we knew that the fields were zero at x = -co and uni-
form at x = xo:
E,(xo) = Vo/s, E,(xo)= 0 (23)
The force is now independent of voltage polarity and always
acts in the direction to pull the dielectric into the capacitor if
S>60o.

3-9-3 Forces on a Capacitor

Consider a capacitor that has one part that can move in the
x direction so that the capacitance depends on the coordinate
x"

q = C(x)v (24)

The current is obtained by differentiating the charge with

respect to time:
Sdq d dv dC(x)
- [C(x)vl = C(x)-+V d
dt dx dt dt
dv dG(x) dx
= C(x) -+ v
at ax t

I
 220 Polarizationand Conduction

Note that this relation has an extra term over the usual circuit
formula, proportional to the speed of the moveable member,
where we expanded the time derivative of the capacitance by
the chain rule of differentiation. Of course, if the geometry is
fixed and does not change with time (dx/dt = 0), then (25)
reduces to the usual circuit expression. The last term is due to
the electro-mechanical coupling.
The power delivered to a time-dependent capacitance is
d
p = vi = v- [C(x)v] (26)
dt
which can be expanded to the form
d 2 1 2 dC(x)
P= dt
d 1, 1 2 dC(x) dx
= -[ X)vd]+v (27)
dt dx dt
where the last term is again obtained using the chain rule of
differentiation. This expression can be put in the form
dW dx
p=-+f- (28)

where we identify the power p delivered to the capacitor as

going into increasing the energy storage W and mechanical
power f dx/dt in moving a part of the capacitor:
,
W=C(x)v = v dC(x) (29)
dx
Using (24), the stored energy and force can also be ex-
pressed in terms of the charge as
1 q2 1 q 2 dC(x) 1 2d[L/C(x)]
W= -- -f-
2 C(x)' 2 C2x) dX d
(30)
To illustrate the ease in using (29) or (30) to find the force,
consider again the partially inserted dielectric in Figure
3-35b. The capacitance when the dielectric extends a distance
x into the electrodes is
exd (1-x)d
C(x) -+
S
so- S (31)

so that the force on the dielectric given by (29) agrees with

(22):
dC(x). Vid
" = V
2x - "sO
dx s

__I ___I_
 Example 3-4 Force on a ParallelPlate Capacitor 221

Note that we neglected the fringing field contributions to

the capacitance in (31) even though they are the physical
origin of the force. The results agree because this extra
capacitance does not depend on the position x of the dielec-
tric when x is far from the electrode edges.
This method can only be used for linear dielectric systems
described by (24). It is not valid for the electret problem
treated in Section 3-9-2b because the electrode charge is not
linearly related to the voltage, being in part induced by the
electret.

EXAMPLE 3-4 FORCE ON A PARALLEL PLATE CAPACITOR

Two parallel, perfectly conducting electrodes of area A

and a distance x apart are shown in Figure 3-36. For each of
the following two configurations, find the force on the upper
electrode in the x direction when the system is constrained to
constant voltage Vo or constant charge Qo.

.-------- -------------

- - - - - - - - - - - - - - - - - - - - + V

SVo
11

Figure 3-36 A parallel plate capacitor (a) immersed within a dielectric fluid or with
(b) a free space region in series with a solid dielectric.
222 Polarizationand Conduction

(a) Liquid Dielectric

The electrodes are immersed within a liquid dielectric with
permittivity e, as shown in Figure 3-36a.

SOLUTION

The capacitance of the system is

C(x) = eA/x
so that the force from (29) for constant voltage is
. dC(x) 1 eA V
dx 2 x2

The force being negative means that it is in the direction

opposite to increasing x, in this case downward. The capacitor
plates attract each other because they are oppositely charged
and opposite charges attract. The force is independent of
voltage polarity and gets infinitely large as the plate spacing
approaches zero. The result is also valid for free space with
e = eo. The presence of the dielectric increases the attractive
force.
If the electrodes are constrained to a constant charge Qo
the force is then attractive but independent of x:
S2 d 1 1Q
dx -C(x) 2EA
For both these cases, the numerical value of the force is the
same because Qo and Vo are related by the capacitance, but
the functional dependence on x is different. The presence of
a dielectric now decreases the force over that of free space.

(b) Solid Dielectric

A solid dielectric with permittivity e of thickness s is inserted
between the electrodes with the remainder of space having
permittivity eo, as shown in Figure 3-36b.

SOLUTION

The total capacitance for this configuration is given by the

series combination of capacitance due to the dielectric block
and the free space region:
EcoA
C(x)=
eOS + (x -s)

_I
 ElectrostaticGenerators 223

The force on the upper electrode for constant voltage is

2 d E2%oA V
f=~~~v Cd(x)= 2[eos +e(x -s)]
If the electrode just rests on the dielectric so that x = s, the
force is
2AV
62=
2sos
This result differs from that of part (a) when x = s by the
factor e,= e/eo because in this case moving the electrode even
slightly off the dielectric leaves a free space region in between.
In part (a) no free space gap develops as the liquid dielectric
fills in the region, so that the dielectric is always in contact
with the electrode. The total force on the electrode-dielectric
interface is due to both free and polarization charge.
With the electrodes constrained to constant charge, the
force on the upper electrode is independent of position and
also independent of the permittivity of the dielectric block:
1 .d 1 1 Q_
odx C(x) 2 eoA

3-10 ELECTROSTATIC GENERATORS

3-10-1 Van de Graaff Generator

In the 1930s, reliable means of generating high voltages

were necessary to accelerate charged particles in atomic
studies. In 1931, Van de Graaff developed an electrostatic
generator where charge is sprayed onto an insulating moving
belt that transports this charge onto a conducting dome, as
illustrated in Figure 3-37a. If the dome was considered an
isolated sphere of radius R, the capacitance is given as C =
4wreoR. The transported charge acts as a current source feed-
ing this capacitance, as in Figure 3-37b, so that the dome
voltage builds up linearly with time:
dv i
=C v = t (1)
dt C

This voltage increases until the breakdown strength of the

surrounding atmosphere is reached, whereupon a spark dis-
charge occurs. In air, the electric field breakdown strength Eb
is 3 x 106 V/m. The field near the dome varies as E,= VR/r 2 ,
which is maximum at r = R, which implies a maximum voltage
of Vm = EbR. For Vma,,= 106 V, the radius of the sphere
 224 Polarizationand Conduction

Charge colle
the moving +
the do C v

Charge sprayec
onto the belt

(a)
Figure 3-37 (a) A Van de Graaff generator consists of a moving insulating belt that
transports injected charge onto a conducting dome which can thus rise to very high
voltages, easily in excess of a million volts. (b) A simple equivalent circuit consists of the
convecting charge modeled as a current source charging the capacitance of the dome.

must be R m so that the capacitance is C -37 pf. With a

charging current of one microampere, it takes it 37 sec to
reach this maximum voltage.

3-10-2 Self-Excited Electrostatic Induction Machines

In the Van de Graaff generator, an external voltage source

is necessary to deposit charge on the belt. In the late 1700s,
self-excited electrostatic induction machines were developed
that did not require any external electrical source. To under-
stand how these devices work, we modify the Van de Graaff
generator configuration, as in Figure 3-38a, by putting
conducting segments on the insulating belt. Rather than
spraying charge, we place an electrode at voltage V with
respect to the lower conducting pulley so that opposite
polarity charge is induced on the moving segments. As the
segments move off the pulley, they carry their charge with
them. So far, this device is similar to the Van de Graaff
generator using induced charge rather than sprayed charge
and is described by the same equivalent circuit where the
current source now depends on the capacitance C, between
the inducing electrode and the segmented electrodes, as in
Figure 3-38b.

_I
 ElectrostaticGenerators 225

n = no. of segm
entering dom
per second

+
C v
i = - 1Ci

Charges induced
onto a segmente
belt
q = -Ci V

+V-

(a)
Figure 3-38 A modified Van de Graaff generator as an electrostatic induction
machine. (a) Here charges are induced onto a segmented belt carrying insulated
conductors as the belt passes near an electrode at voltage V. (b) Now the current source
feeding the capacitor equivalent circuit depends on the capacitance Ci between the
electrode and the belt.

Now the early researchers cleverly placed another

induction machine nearby as in Figure 3-39a. Rather than
applying a voltage source, because one had not been invented
yet, they electrically connected the dome of each machine to
the inducer electrode of the other. The induced charge on
one machine was proportional to the voltage on the other
dome. Although no voltage is applied, any charge imbalance
on an inducer electrode due to random noise or stray charge
will induce an opposite charge on the moving segmented belt
that carries this charge to the dome of which some appears on
the other inducer electrode where the process is repeated
with opposite polarity charge. The net effect is that the charge
on the original inducer has been increased.
More quantitatively, we use the pair of equivalent circuits in
Figure 3-39b to obtain the coupled equations

- nCiv, = Cdv, nCiV2 = C (2)

dt dt
where n is the number of segments per second passing
through the dome. All voltages are referenced to the lower
pulleys that are electrically connected together. Because these
 226 Polarizationand Conduction

Figure 3-39 (a) A pair of coupled self-excited electrostatic induction machines

generate their own inducing voltage. (b) The system is described by two simple
coupled circuits.

are linear constant coefficient differential equations, the solu-

tions must be exponentials:
st,
vl = e
e71 v2 = V^
2 e'

Substituting these assumed solutions into (2) yields the

following characteristic roots:

s = :s= :+
C C

so that the general solution is

vi = A, e(Mci/c)t +A e,-( "CiC)9
v2 = -A, e(nc/c)t + A 2 e-(nc.ic)
where A and A 2 are determined from initial conditions.
The negative root of (4) represents the uninteresting
decaying solutions while the positive root has solutions that
grow unbounded with time. This is why the machine is self-
excited. Any initial voltage perturbation, no matter how
small, increases without bound until electrical breakdown is
reached. Using representative values of n = 10, Ci = 2 pf, and
C= 10 pf, we have that s = -2 so that the time constant for
voltage build-up is about one-half second.

I 1_1
 Electrostatic Generators 227

Collector - Conducting Cdllecting

brush strips brushes

Grounding Inducing
brush electrode
Front view
Inducing
electrodes
Side view

Figure 3-40 Other versions of self-excited electrostatic induction machines use (a)
rotating conducting strips (Wimshurst machine) or (b) falling water droplets (Lord
Kelvin's water dynamo). These devices are also described by the coupled equivalent
circuits in Figure 3-39b.

The early electrical scientists did not use a segmented belt

but rather conducting disks embedded in an insulating wheel
that could be turned by hand, as shown for the Wimshurst
machine in Figure 3-40a. They used the exponentially grow-
ing voltage to charge up a capacitor called a Leyden jar
(credited to scientists from Leyden, Holland), which was a
glass bottle silvered on the inside and outside to form two
electrodes with the glass as the dielectric.
An analogous water drop dynamo was invented by Lord
Kelvin (then Sir W. Thomson) in 1861, which replaced the
rotating disks by falling water drops, as in Figure 3-40b. All
these devices are described by the coupled equivalent circuits
in Figure 3-39b.

3-10-3 Self-Excited Three-Phase Alternating Voltages

In 1967, Euerle* modified Kelvin's original dynamo by

adding a third stream of water droplets so that three-phase

* W. C. Euerle,"A Novel Method of GeneratingPolyphase Power," M.S. Thesis, Massachusetts

Institute of Technology, 1967. See also J. R. Melcher, Electric Fields and Moving Media,
IEEE Trans. Education E-17 (1974), pp. 100-110, and thefilm by the same title produced
for the NationalCommittee on ElectricalEngineeringFilms by the EducationalDevelopment
Center, 39 ChapelSt., Newton, Mass. 02160.
228 Polarizationand Conduction

alternating voltages were generated. The analogous three-

phase Wimshurst machine is drawn in Figure 3-41a with
equivalent circuits in Figure 3-41 b. Proceeding as we did in (2)
and (3),

-nC i v = CdV2 vi= V se

dvT

- nv2sy = C--, V2 = V 2 se
dr
dv,
= C-,
- nCiv3
dr'

equation (6) can be rewritten as

nCi

Cs nC,JVsJ

Figure 3-41 (a) Self-excited three-phase ac Wimshurst machine. (b) The coupled
equivalent circuit is valid for any of the analogous machines discussed.

_·
 ElectrostaticGenerators 229

which reguires that the determinant of the coefficients of V 1,

V2, and Vs be zero:

(nC) 3 +(C
+(s) 3
=0 1
= (nC i 1 1)1s

i m( ,
(nCQ
\, e (7T Xr-l r= 1,2, 3 (8)

C
nCi
C

S2,3=!C'[I+-il
2C

where we realized that (- 1)1/ s has three roots in the complex

plane. The first root is an exponentially decaying solution, but
the other two are complex conjugates where the positive real
part means exponential growth with time while the imaginary
part gives the frequency of oscillation. We have a self-excited
three-phase generator as each voltage for the unstable modes
is 120" apart in phase from the others:

V2 V3 V1 nC _(+-j)=ei(/) (9)
V1 V2 V3 Cs2,•

Using our earlier typical values following (5), we see that the
oscillation frequencies are very low, f=(1/2r)Im(s) =
0.28 Hz.

3-10-4 Self-Excited Multi-frequency Generators

If we have N such generators, as in Figure 3-42, with the

last one connected to the first one, the kth equivalent circuit
yields
-nCi•V, = CsVk+l (10)

This is a linear constant coefficient difference equation.

Analogously to the exponential time solutions in (3) valid for
linear constant coefficient differential equations, solutions to
(10) are of the form
V 1 =AAk (11)

where the characteristic root A is found by substitution back

into (10) to yield

- nCiAA k= CsAA '+•A= - nCilCs

230 Polarizationand Conduction

-Wmh= dvc 1t
nCin
W wCidt
Figure 3-42 Multi-frequency, polyphase self-excited Wimshurst machine with
equivalent circuit.
Since the last generator is coupled to the first one, we must
have that
VN+I = Vi * N+' =A

>AN= 1
AA=: lIIN j2i•/N r=1,2,3,... ,N (13)
where we realize that unity has N complex roots.
The system natural frequencies are then obtained from
(12) and (13) as
nCA nCi -i2AwN (14)
CA CT
We see that for N= 2 and N= 3 we recover the results of (4)
and (8). All the roots with a positive real part of s are unstable
and the voltages spontaneously build up in time with oscil-
lation frequencies wo given by the imaginary part of s.
nCi
o0= IIm (s)l =- Isin 2wr/NI (15)
C
 ProbLnus 231

PROBLEMS

Section 3-1
1. A two-dimensional dipole is formed by two infinitely long
parallel line charges of opposite polarity ±X a small distance di,
apart.

-x_
(a) What is the potential at any coordinate (r, 46, z)?
(b) What are the potential and electric field far from the
dipole (r >> d)? What is the dipole moment per unit length?
(c) What is the equation of the field lines?
2. Find the dipole moment for each of the following charge
distributions:

2 I

jL
II t+ X
L + o L Xo d

L - o L

(a) (c) (d) (e)

(a) Two uniform colinear opposite polarity line charges

*Ao each a small distance L along the z axis.
(b) Same as (a) with the line charge distribution as
AAo(1-z/L), O<z<L
A-Ao(l+z/L), -L<z<O
(c) Two uniform opposite polarity line charges *Ao each
of length L but at right angles.
(d) Two parallel uniform opposite polarity line charges
* Ao each of length L a distance di, apart.