Physics 704 Spring 2013

1 Fundamentals
1.1 Overview
The objective of this course is: to determine

1 and

1 elds in various physical
systems and the forces and/or torques resulting from them.
The domain of applicability is the classical regime in which the eld picture
is appropriate. In QM the photon is the particle that mediates the E-M force.
The eld picture is appropriate when interactions involve lots of photons. This
regime has widespread applicability. Mostly well be working with continuous
charge distributions in which the discrete nature of electrons is ignored.
Idealizations well make include:
1. The photon has zero mass: the experimental limit is :
~
< 10
51
g. (See
link on web site for details.) The mass of the photon is linked to the form of
the point charge potential ,r and the corresponding inverse square force
law. (Most experiments test for the inverse square law rather than trying to
measure mass directly.) If the photon had mass, then the potential would be
of the form

1
r
exp

:
~
c
/
r

Think about what the force law looks like in this case, and how it diers from
a pure inverse-square law.
2. Perfect conductors exist.
3. An extended surface can be maintained at a constant potential, and two
dierent surfaces can be maintained at the same potential ("common ground").
This idealization is closely related to #2.
4. Surface charge and current layers have zero thickness. The actual thick-
ness is of order a few atomic diameters, so this idealization is equivalent to
size of system few atomic diameters
1.2 Units:
The rst half of Jacksons book uses SI units, the second half uses Gaussian.
Read the appendices on units! SI is an engineering system and is theoretically
corrupt. In the Gaussian unit system, 1 and 1 are dimensionally equivalent; in
SI they are not. This creates diculties when 1 and 1 are dierent components
of the same tensor! (See Chapter 11.) In the Gaussian system all units are
derived from the three fundamental units of mass (gm), length (cm), and time
(s). For example:
1
1 =

2
r
2
= r

1
and thus the esu (electro-static unit of charge or stat-coulomb) is
1 esu = cm
r
gm cm
s
2
= cm
3/2
gm
1/2
s
1
(1)
In SI a pseudo-fundamental unit of current is also included. (This is the
corruption.) In cgs-Gaussian units there is no need for constants such as -
0
and j
0
.
The fundamental relations of electromagnetic theory are Maxwells equa-
tions. In SI units, we have
Gauss law:

1 =
j
-
0

1 = j
}

1 = 0
Faradays law

1 =
0

1
0t
Amperes law

1 = j
0

, +j
0
-
0
0

1
0t

H =

,
}
+
0

1
0t
plus the Lorentz force law

1 =

1 +

1

We will spend the entire semester learning how to solve these 5 equations, and
understand the solutions.
Charge conservation
0j
0t
+

, = 0
is a useful additional relation, but it is not independent of Maxwells equations.
To see this, start with Gauss law
0j
0t
=
0
0t

-
0

-
0
0
0t

1

t and r are independent variables, -
0
is constant
=

1
j
0

,
!
Amperes law
=

, divergence of a curl is zero

You can also derive this equation from the principle of charge conservation, as
in Lea Problem 1.6.
2
1.3 Nature of the mathematical problem
Maxwells equations are linear, 2nd order, elliptic PDEs. Classes of problems
well be interested in include:
1. Given sources (j,

,) nd the elds.
2. Given boundary conditions on the bounding surface of a nite volume,
nd the elds in the volume. (Boundary conditions show the eect of charges
and currents outside the volume.)
3. Given the elds, nd their eect on j,

,.
These are well-posed problems with unique solutions. Diculties can arise
because the elds due to sources act back on the sources, changing the position
and velocity of charges and thus changing the sources that produce the elds.
For the most part we will not get involved with these complications. (But see
plasma physics course (Phys 712) , and also J chapter 16).
Linearity is nice. It gives us the principle of superposition. We can compute
the elds due to individual sources and simply add the results. The technique
of using Greens functions depends on the linearity of Maxwells equations.
Structure of the equations.
The equations may be divided into two groups in two ways:
1. Divergence equations and curl equations. This is useful when nding
boundary conditions.
2. Equations with sources and equations without sources. This is actually
more fundamental See Ch 11.
The lack of symmetry is due to non-existence of magnetic monopoles.
1.4 Fields in media.
In a material medium, an applied

1 polarizes the material, reducing the net

1
in the interior. It is useful to dene a new eld

1, the electric displacement,

1 = -
0

1 +

1
where

1 is the polarization. Since

1 is generally proportional to the net

1,
(in which case the material is said to be linear and isotropic)

1 = -
0

1
where is the susceptibilty, then

1 = -
0
(1 +)

1 = -

1 (2)
Then

1 = -
0

1 +

1
= j j
b
= j
}
3
where
j
b
=

1
is the bound charge density and j
}
is the free charge density. See Ch 4 for
more details. (Well get to this in April.) This description is most useful in the
so-called LIH materials (linear, isotropic, homogeneous) in which - is a constant
scalar. However, - is always a function of the frequency of the applied elds,
and strictly equation (2) is true in the Fourier transform domain.

1(.) = - (.)

1(.) (3)
and this leads to an interesting relation between the time-dependent elds.
(See Lea Ch 7 P 12 and Jackson Ch 7.) In static situations we use
- = lim
.0
- (.)
For non-isotropic materials - is a tensor, and the relation is
1
I
= -
I
1

indicating that one component of

1 depends on all components of

1.
Similarly, atoms have magnetic moments which can be modied and/or
aligned by applied elds. (see eg Ch 29 in Lea and Burke and also J p 191-
194) In these cases we have a magnetization

' =
n

H and

1 = j
0

H +

'

= j

H
In situations where this description is useful (LIH) we nd that j,j
0
' 1. The
interesting cases with j,j
0
substantially dierent from one are not LIH, and
the non-linearity gives rise to some very interesting phenomena (hysteresis etc).
We wont get into these much in this course.
2 Boundary conditions
Here we investigate the relations between the eld components on either side of a
boundary between dierent materials. We start with the divergence equations,
and integrate over a pillbox placed across the boundary. The dimensions of the
pillbox are chosen as follows:
' d
2
, d 1 = length scale over which any property changes
/ d
4
Then, from Gausss law, we have
Z
pillbox

1 d\ =
Z
pillbox
j
}
d\ =
Z
j
}
dd/
The free charge is usually conned to the surface (remember: this means within
a few atomic diameters of the surface), so dening the surface charge density as
o
}
=
R
j
}
d/, we have
I

1 : d =
Z
d
Z
j
}
d/ =
Z
o
}
d
With d 1, o
}
is constant over the area, and we may approximate
I

1 : d = o
}
With / d, the sides of the pillbox do not contribute to the integral on the
LHS. Then

1
1

1
2
i
: = o
}
Thus h

1
1

1
2
i
: = o
}
(4)
For the electric eld h

1
1

1
2
i
: =
o
-
0
(5)
and since

1 satises the equation

1 = 0, then
h

1
1

1
2
i
: = 0 (6)
Always draw the pillbox to avoid sign errors. Note that in this derivation,
: points from region 2 into region 1.
5
Next we look at the curl equations and integrate around a rectangle that
crosses the boundary. We choose the dimensions such that the rectangle mea-
sures n / where, as before, / n 1. We start with Amperes law.
Z

d =
Z

,
}


d+
Z
0

1
0t


d
Here

is perpendicular to the plane of the rectangle, according to the RH rule.
Then using Stokes theorem, and for n 1,
I

H d

/ = n
Z

,
}


d/ +n
Z
0

1
0t


d/
Then as we let / 0, the short sides of the rectangle do not contribute to the
line integral, and
h

H
1

+

H
2

t
i
n = n

1
}

where

1
}
=
Z

,
}
d/
is the free surface current density. Note that as / 0, we trap all the sur-
face current inside our rectangle. (Eectively,

, is a delta function and thus
right at the surface.) However, the time derivative of

1 remains nite
everywhere, and so the 2nd integral on the RHS always goes to zero as / 0. .
We may rewrite the LHS as follows

H
2

H
1

=
h

H
2

H
1

:
i

=

1
}

Now we may orient our surface arbitrarily in the sense that

can be any vector
in the plane perpendicular to :. For example, if : = ., we can choose

= r
or

= j. Thus we get all components of the equation

H
2

H
1

: =

1
}
= :

H
1

H
2

(7)
or, in terms of

1,
:

1
1

1
2

= j
0

1 (8)
6
Similarly for

1
:

1
1

1
2

= 0 (9)
The tangential component of

1 is continuous across any surface.
3 Derivation of equations from experimental re-
sults
3.1 Coulombs Law
Physics is an experimental science, so Maxwells equations have their roots in
experiment. Cavendish (in England) and Coulomb (in France, 1785) indepen-
dently determined the electostatic force law. The force exerted on one point
charge by another is

1
on 2 due to 1

1

2
r
r
2
The constant of proportionality depends on the unit system. In Gaussian units
it is 1 (because this equation denes the charge unit, see eq. 1). In SI it is
/ = 9 10
9
N m
2
,C
2
=
1
4-
0
We dene the electric eld by placing a point test charge at a point 1 and
measuring the force on it. Then

1(1) = lim
j0

Thus the eld due to the point charge

1
at the position of
2
is

1(1) =
/
1
r
2
r
7
Next we invoke the observed linearity of the forces to obtain the principle of
superposition for

1

1(r) =
X
all charges
/

I
r
2
I
r
I
= /
X
I

I
r r
I
|r r
I
|
3
and then for a continuous distribution with d = j (r
0
) d\
0
,

1(r) = /
Z
j (r
0
)
r r
0
|r r
0
|
3
d
3
r
0
(10)
The standard convention is to use the primed coordinates for the source and
the unprimed coordinates for the eld point.
3.2 Gauss Law
Consider the electric eld at a point 1 on a surface o that surrounds a point
charge . Let do be an element of the surface with outward normal :. Then

1 : do = /

r
2
r : do
Now the dot product r : do projects the area onto the plane perpendicular to
r. Thus, by denition, this area is r
2
d, and the ux is
8
I

1 : do = /
I
d = 4/ =

-
0
If is outside o, then there is an area element on the "near" side whose contri-
bution exactly cancels the element on the far side, and the result is zero. Now
we can use the principle of superposition to sum the contributions from all the
charges inside, to obtain Gauss law:
I

1 : do =
all charge inside
-
0
=
R
\
j d\
-
0
The nal step is to use the divergence theorem on the LHS
Z
\

1 d\ =
Z
\
j
-
0
d\
Since this result must be true for any and every volume \, then

1 =
j
-
0
which is our rst Maxwell equation.
9
3.3 Faradays Law
In 1831 Faraday published his experimental result:
Emf
d
dt
(magnetic ux through circuit)
or, in symbol form
I
c

1d

/ =
d
dt
Z
S

1 : d (11)
where o is a surface spanning the curve C. Note that the time derivative can
give a non-zero result for any of the following:
1.

1 is a function of t
2. the angle between

1 and : is a function of t
3. the area of the surface is a function of t.
Items (2) and (3) give rise to "motional EMF".
Comments:
1. The electric eld is the eld that would make current ow in the circuit
if the circuit were a conducting wire. That means,

1 is measured in the
rest frame of the circuit element d

/.
2. The constant of proportionality (here unity) depends on the unit system
chosen. In Gaussian units it is 1,c.
3. The minus sign expresses energy conservation. It makes sense here be-
cause the direction of : is related to the direction of d

| through the usual

right hand rule. It is sometimes given a separate name as Lenzs Law.
First lets choose our curve C to be stationary. Then the rate of change of
ux is due entirely to the time rate of change of

1 itself, and the derivative may
be moved inside the integral, where it becomes 0,0t. (All spatial information
disappears in the integral over o). We may also convert the LHS to a surface
integral using Stokes theorem:
Z
S

: d =
Z
S
0
0t

1 : d (12)
Now this relation must hold true for any stationary curve C and any corre-
sponding spanning surface o, and so we may conclude:

1 =
0
0t

1
which is Faradays law.
But there is more to learn from this experiment. What happens when we
allow C to have a uniform velocity ? There is an additional contribution to the
10
change in ux as the curve moves to a region with a diering value of

1. Using
a Taylor series expansion, we have:
1(r +cr) = 1(r +ct) =

1(r) +

1ct +
Then
1(r +cr)

1(r) = c

1 =

1ct
to rst order in ct. Thus the change in ux due to motion of the curve is:
c
n
=
Z
S
c

1 : d =
Z
S
h

1
i
: dct
and hence, in the limit ct 0,
d
dt

n
=
Z
S
h

1
i
: d
due to motion of the curve. (This is sometimes called "motional emf".) Adding
the contribution from 0

1,0t that we have already calculated, we have

d
dt

total
=
Z
S
0
0t

1 : d+
Z
S
h

1
i
: d
and applying Faradays law:
I
c

1
0
d

/ =
Z
S
0
0t

1 : d+
Z
S
h

1
i
: d

Remember: the electric eld

1
0
is measured in the rest frame of d

/, i.e. the
frame moving with velocity with respect to the lab, and

1 is measured in the
lab frame.
Now we want to convert the last term on the right to a line integral, so we
use a result from the cover of Jackson:

=

1

1
But here we have chosen to be uniform, so all its spatial derivatives are zero,
and from the second Maxwell equation,

1 = 0, so

1
Thus
I
c

1
0
d

/
0
=
Z
S
0
0t

1 : d+
Z
S
h

1
i
: d

=
Z
S
0
0t

1 : d
I
c

/ (13)
11
Comparing equation (13) with equations (11) and (12) we may replace the
surface integral of
J
J|

1 : with a line integral involving

1 in the lab frame
1
:
I
c

1
0
d

/
0
=
I
c

1 d

/ +
I
c

/
where again the result is true for any curve C moving at uniform velocity .
Thus we obtain the (non-relativistic) transformation law:

1
0
=

1 +

1 (14)
The result is consistent with the non-relativisitic Lorentz force law in this unit
system. Thus we have established that the constant of proportionality in Fara-
days law is linked to the transformation properties of the electric eld. (See
Chapter 11 for relativistic corrections to this result.)
3.4 Amperes Law
We start with the Biot-Savart Law (discovered in 1820 by Jean-Baptiste Biot
and Felix Savart). The magnetic eld element d

1 produced by current 1 in a
wire segment d

| is
d

1 = /
n
1d

/
r
r
2
As in Coulombs law, r points from the source to the eld point. More generally,
for a current density

,,

1(r) =
Z
d

1 =
j
0
4
Z

, (r
0
)
(r r
0
)
|r r
0
|
3
d
3
r
0
(15)
Note the following useful relation:

1
|r r
0
|
=

1
q
(r r
0
)
2
+ (j j
0
)
2
+ (. .
0
)
2
=
r r
0
|r r
0
|
3
r
j j
0
|r r
0
|
3
j
. .
0
|r r
0
|
3
.

1
|r r
0
|
=
r r
0
|r r
0
|
3
=

0 1
|r r
0
|
(16)
So we may write

1(r) =
j
0
4
Z

, (r
0
)

1
|r r
0
|
d
3
r
0
From the front cover

(ca) =

c a +c

a = a

c +c

a
1
Strictly, we must apply Faradays law to a curve
0
at rest in the lab that instantaneously
coincides with the moving curve
12
where here a =

, (r
0
) and c = 1, |r r
0
| , but since

, (r
0
) is not a function of
r, then

a = 0 and so (note the sign)

1 =
j
0
4

Z

, (r
0
)
|r r
0
|
d
3
r
0
(17)
We may write this relation as

1 =

with

(r) =
j
0
4
Z

, (r
0
)
|r r
0
|
d
3
r
0
(18)
so we immediately have

1 = 0,
our third Maxwell equation.
To obtain Amperes law, we take the curl of (17):

1 =
j
0
4

Z

, (r
0
)
|r r
0
|
d
3
r
0
!
Expand the double derivative:

1 =
j
0
4
"

Z

, (r
0
)
|r r
0
|
d
3
r
0
!

2
Z

, (r
0
)
|r r
0
|
d
3
r
0
#
Move the dierential operator inside the integral, where it operates only on the
1, |r r
0
| factor. In the rst integral we also use the results (16) and

(ca) =

c a +c

a (19)
I leave it to you to verify that the last term also simplies as follows:

1 =
j
0
4

Z

, (r
0
)

1
|r r
0
|
d
3
r
0

Z

, (r
0
)

2 1
|r r
0
|
d
3
r
0

=
j
0
4

Z

, (r
0
)

0 1
|r r
0
|
d
3
r
0

+ 4
Z

, (r
0
) c (r r
0
) d
3
r
0

where we have used Lea eqn 6.26. In the rst term, use relation (19) again,
and in the second use the sifting property:

1 =
j
0
4
(

Z
"

, (r
0
)
|r r
0
|

1
|r r
0
|

, (r
0
)
#
d
3
r
0
!
+ 4

, (r)
)
Now use the divergence theorem and charge conservation.

1 =
j
0
4

"

Z
S

, (r
0
)
|r r
0
|
:
0
d
0
+
Z
1
|r r
0
|

0j (r
0
, t)
0t

d
3
r
0
#
+j
0

, (r)
13
If the current distribution is conned to a nite region of space, the surface
integral is zero. (Beware innite wires!)

1 =
0
0t
j
0
4

Z
j (r
0
)
|r r
0
|
d
3
r
0

+j
0

, (r) (20)
and evaluating the gradient and comparing with equation (10) we nally have
Amperes law.

1 =
0
0t
j
0
4
Z
j (r
0
) (r r
0
)
|r r
0
|
3
d
3
r
0
+j
0

, (r)
= j
0
-
0
0

1
0t
+j
0

, (r)
Strictly these derivations are valid when the elds change slowly, although the
results are always valid. Well investigate further in Chapters 6 and 7.
4 Potentials and gauge choice
4.1 Static case
We start with Faradays law, which in the static case reduces to

1 = 0
and thus we may write

1 =

(21)
The minus sign is convential. Comparing with equation (20) we may write the
solution for :
(r) =
1
4-
0
Z
\
j (r
0
)
|r r
0
|
d
3
r
0
(22)
provided that the integral exists.
We may also evaluate as follows:
Z
2
1

1 d

| =
Z
2
1

| =
Z
2
1
d =
1

2
(23)
The integral (22) does not exist if j = `c (r) c (j) corresponding to an innite
line charge along the .axis. Then the integral becomes
(r) =
Z
+

`
q
r
2
+j
2
+ (. .
0
)
2
d.
0
14
Let .
0
. =
p
r
2
+j
2
tan0. Then
(r) =
Z
+t/2
t/2
`
p
r
2
+j
2

1 + tan
2
0
sec
2
0d0
=
`
p
r
2
+j
2
Z
+t/2
t/2
sec 0d0
=
`
p
r
2
+j
2
ln(sec 0 + tan0)|
t/2
t/2
and now we are in trouble because the logarithm is undened at both limits.
The problem arises because our charge distribution extends to innity. In cases
like this there is usually enough symmetry that we can proceed by nding

1
using the integral form of Gauss law, and integrating

1 to get the potential
(eqn 23). For the line charge,

1
=
Z
1
2

1 d

/ =
Z
1
2
`
2-
0
r
dr
=
`
2-
0
ln

r
1
r
2

We may not choose the reference point at innity because there is charge there.
So if the reference point where = 0 is at a distance r
0
from the line, then
(r) =
`
2-
0
ln

r
0
r

(24)
Alternatively, using the expression (21) for

1 in Gauss law, we have the
dierential equation (Poissons equation) for :

2
(r) =
j (r)
-
0
(25)
Example: dipole potential.
The standard model for an ideal dipole is two equal but opposite charges
separated by a small displacement

/.
15
The dipole moment is
j = lim
j,|0

/
We begin with equation (22) with the charge density j (r) = c (r)+c

(dipole at the origin). Then

(r) =
1
4-
0
Z
c (r
0
) +c

r
0

|r r
0
|
d
3
r
0
=
1
4-
0

|r|
+

(r) =

4-
0
|r|

1 +
1
q
1 2

|r
|r|
2
+
l
2
|r|
2

=

4-
0
|r|
"
1 + 1 +

/ r
|r|
2
+
#

j r
4-
0
|r|
3
as / 0 (26)
or, placing the dipole at a point r
0
, we get

dipole
(r) =
(r r
0
) j
4-
0
|r r
0
|
3
=
1
4-
0
j

1
r r
0

(27)
Note that this potential decreases as one over the square of the distance from
the source.
16
4.1.1 Magnetic scalar potential
If

, (r) = 0, then Amperes law gives us

1 = 0, so we may also write

1 =

n
where

n
(r) = 0 (28)
Thus the magnetic scalar potential satises Laplaces equation. This equation
is easier to solve than Poissons equation, but we have lost all the information
about the sources! The result is still useful anywhere that

, = 0 (ie outside
the source) but we have to use tricks to nd the potential. (See problem 5.1
for example.)
A simple example: Current in a long straight wire. Because of the
azimuthal symmetry and translational symmetry parallel to the wire, we can
easily nd the eld using Amperes Law:

1 =
j
0
2
1
j

c
where j is the radial coordinate in a cylindrical coordinate system. This suggests
that we can use the scalar potential

n
=
j
0
2
1c
This is not ideal, however, because
n
is not single-valued. It has the countably
innite number of values
n
=

0
2t
1 (c + 2:) at any point with azimuthal
angle c. We wont get into much trouble, though, because all these values give
the same eld

1.
4.1.2 Magnetic vector potential
When

, 6= 0, we start with

1 = 0
which tells us we can express

1 as the curl of a vector eld

(see, eg, equation
17). Then Amperes law reduces to

2

= j
0

, (29)
We may add the gradient of any scalar function c to

without changing the
value of

1. We may use this freedom to set

= 0 (30)
17
This is the Coulomb Gauge. (Suppose

0
6= 0. We let

=

0
+

c. Then

= 0 =

0
+
2
c and we can solve this equation to nd a suitable c.)
Then the equation for the potential

is

2

= j
0

, (31)
If we work in Cartesian components, each component of this equation has the
same form as the equation for , and so the solution also has the same form,
as indeed we have already found (equation 18). Do not attempt to use this
result in other than Cartesian components! While true as a vector relation, the
integration must be handled with extreme care.
Again the solution (18) is valid only when the integral exists. We cannot
compute the vector potential due to a current in a long straight wire with this
equation. But for a long wire, it is easy to nd

1 using Amperes law. If we
need

rather than just

1, we have to work from the dierential equation (31),
which takes the form:

2

= j
0
1 . c (r) c (j)
So, in Cartesian components,

r
=
2

= 0
and

:
= j
0
1c (r) c (j) (32)
The gauge condition is

=
0
r
0r
+
0

0j
+
0
:
0.
= 0
Since the symmetry of the situation (translational symmetry along the direction
parallel to the wire and rotational symmetry about the wire) leads us to expect

to be independent of ., (
J.

J:
= 0), and
r
=

, we should be able to nd a
solution with
r
=

= 0. Then equation (32) has the solution:

:
=
j
0
2
1 ln
j
j
0
where j =
p
r
2
+j
2
. (Recall that
2
ln(j,j
0
) = 2c (j) . Lea Chapter 6 eqn.
6.27) The reference point where

= 0 is at distance j
0
from the line.
We can check this result by calculating

1 :

1 =

=

c

0
:
0j

=

c

j
0
2j
1

which agrees with the result from Ampres law.

18
4.2 Time dependence
When the elds are time dependent,



1 is no longer zero. But



1 is
always zero, and so it makes sense to start with the vector potential

. Then
Faradays law becomes:

1 =
0
0t

0
0t

or

1 +
0

0t
!
= 0
So now we may write

1 +
0

0t
=

leading to

1 =

0t
(33)
The rst term represents the eld due to charges, or the Coulomb eld, while
the second represents the induced eld. To see how this decomposition arises
from Maxwells equations, we put these potentials into Amperes law:

1 =

= j
0

, +j
0
-
0
0

1
0t

2

= j
0

, +j
0
-
0
0
0t

0t
!
Rearranging, and reordering the partial derivatives of , we nd

+j
0
-
0

0
0t
+j
0
-
0
0
2
0t
2

2

= j
0

+
1
c
2
0
0t

+
1
c
2
0
2
0t
2

2

= j
0

, (34)
Here the convenient gauge choice is

+
1
c
2
0
0t
= 0 (35)
the Lorentz Gauge. Then the equation for

becomes
1
c
2
0
2
0t
2

2

= j
0

, (36)
Thus

satises a wave equation with wave speed c and source j
0

,.
Next we look at Gauss law, using (33) for

1:

0t
!
=
j
-
0
19
Expand out the divergence, and use the gauge condition (35) to eliminate

2

0
0t

1
c
2
0
0t

=
j
-
0

2
+
1
c
2
0
2
0t
2
=
j
-
0
(37)
So here, too, satises a wave equation, with wave speed c, and this time the
source is j,-
0
.
In the special case that the elds are time independent, the Coulomb and
Lorentz Gauges are the same. Well have more to say about these gauges later.
Even with the gauge condition imposed, there is still some freedom in our
potentials. Suppose we have a vector potential

and a scalar potential . We
want to impose the Lorentz gauge condition, so we look for a vector

0
where

0
=

+

c and
0
where
0
= +. We want the same values of the elds,
so:

1
0
=

0
=

+

c

=

1
and

1
0
=

0
0t
=

( +)
0

+

c

0t
=

0t

0

c
0t
=

1

+
0c
0t

=

1
Thus
=
0c
0t
+C
where C is a constant. Now imposing the Lorentz gauge condition,

0
+
1
c
2
0
0t

0
= 0

+

c

+
1
c
2
0
0t
(+) = 0
If

and satisfy the Lorentz gauge condition,

0
and
0
will also if we choose
c and C to satisfy the equation:

2
c +
1
c
2
0
0t
=
2
c +
1
c
2
0
0t

0c
0t
+C

= 0
Then c satises the source-free wave equation

2
c
1
c
2
0
2
0t
2
c = 0
20
5 Magnetic scalar potential and permanent mag-
nets
The relation between

1 and

H is

1 = j
0

H +

'

Then, since

1 = 0,

H =

' (38)
Thus

' acts like a magnetic charge density in producing

H. In the
time-independent case,

H =

,
free
So if

,
free
= 0, we may write

H as the gradient of a potential,

H =

n
.
Comparing equations (25) and (38), we may immediately write the solution:

n
=
1
4
Z
\

' (r
0
)
|r r
0
|
d
3
r
0
(39)
Then we may integrate by parts using relation (19):

n
(r) =
1
4
Z
\

"

' (r
0
)
|r r
0
|
#
d
3
r
0
+
1
4
Z
\

' (r
0
)

0 1
|r r
0
|
d
3
r
0
=
1
4
Z
S

' (r
0
) :
0
|r r
0
|
d
2
r
0

1
4
Z
\

' (r
0
)

1
|r r
0
|
d
3
r
0
(40)
= 0
1
4

Z
\

' (r
0
)
|r r
0
|
d
3
r
0
Then at a large distance from the magnet, |r r
0
| '
1
:
, and

n
(r) '

1
4r

Z
\

' (r
0
) d
3
r
0
=
: r
4r
3
(41)
This is a dipole potential (cf equation 26) , with magnetic dipole moment : =
R
\

' (r
0
) d
3
r
0
.
When the magnetization is uniform within a volume \,



' (and hence
the magnetic charge density) is zero except at the edge of the volume where

'
comes abruptly to zero, and



' . Then if we put a pillbox over the
surface, we nd
21
Z
\ pillbox
j
1
d\ =
Z
\ pillbox

' d\ =
Z
S pillbox

' :
pillbox
d =
Z
S cookie
o
1
d
where the pillbox has cut a cookie from the surface of the volume. Notice
that :
pillbox
is outward from the pillbox, but on the face inside the magnet, thus
:
pillbox
= :, where : is the outward normal from the magnetized volume. Thus
the magnetic surface charge density is
o
1
=

' : (42)
In such cases the potential is

n
(r) =
1
4
Z
S

' (r
0
) :
|r r
0
|
d
2
r
0
(43)
where the integral is over all parts of the surface where

' : is not zero.
22