Sei sulla pagina 1di 1142

VECTOR AREA

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 1.61.
This is a brief mathematical interlude since well need the results here
when we discuss magnetic dipoles.
The vector area of a surface is the integral of the differential area vector
over the surface. That is

(1) a da
S
Remember that da is normal to the surface at each point.
As an example, we can calculate a for a hemisphere of radius R. In
spherical coordinates |da| = R2 sin d d . Since the vector area definition
is a vector equation, its easiest to split it up into 3 equations in rectangular
coordinates. The normal to the hemisphere is in the r direction at every
point, and

(2) r = sin cos x + sin sin y + cos z


For the x component, we have
/2 2
2
(3) ax = R cos sin3 d d
0 0
(4) = 0

since the integral over gives zero. Similarly, ay = 0. For az we get


/2 2
2
(5) az = R cos sin d d
0 0
(6) = R2
Thus the vector area of a hemisphere has the same magnitude as the circle
that is its base.
For a full sphere, we just extend the upper limit on to in the above
integrals, and we find that all three components are zero. In fact, this is a
1
VECTOR AREA 2

special case of a more general theorem, which is that a = 0 for any closed
surface. To see this, we can apply the divergence theorem in the following
way.
Suppose we have a vector field v = cT , where c is a constant vector and
T is some scalar field. Then if S is a closed surface and V is the volume it
encloses:

(7) v da = vd 3 r
S V
(8) c T da = (cT ) d 3 r
S
V
(9) = [c T + T c] d 3 r
V

(10) = c T d 3 r
V

where the last equality follows because c is a constant. Since c is arbitrary,


the two integrals must be equal:

(11) T da = T d 3 r
S V
If we take T = 1 then the LHS is just a and the RHS is zero since T is a
constant. Thus the vector area of any closed surface is zero. This is because
although the actual surface area is non-zero, the vector components always
cancel each other out as we integrate over the surface.
Consider now a closed surface and divide it into two parts by cutting it
along some closed curve L. Since the total vector area is zero, we must
have aupper = alower . Now the shape of the surface is arbitrary, so for
some curve L, we can keep the lower surface constant while varying the
shape of the upper surface. That is, alower is held constant while the upper
surface varies. However, the equality of the two areas must always hold, so
any surface enclosed by L must have the same vector area. This explains
why the vector area of a hemisphere is just the area of the circle that defines
its base: these are just two different surfaces sharing a boundary curve.
We can apply this to get a different formula for a in terms of the boundary
curve. Suppose we draw a cone with its vertex at the origin and with its base
being the curve L. (The base need not lie in a plane, since L doesnt have to
be flat. This wont affect the argument.) The vector area of this cone must
be the same as any other surface that shares L. Now if we divide up L into
line increments dl then we divide up the cone into a sequence of triangles
VECTOR AREA 3

with sides r (the vector from the vertex of the cone (the origin) to dl), dl
and r + dl. The area of this triangle is half the area of the parallelogram
with two adjacent sides r and dl, and that in turn is the magnitude r dl.
Therefore

1
(12) a= r dl
2 L
Finally, we can derive a similar result to the divergence one above using
Stokess theorem. Again using a vector field v = cT we have

(13) v dl = [ (cT )] da

L
S

(14) c T dl = [T c c T ] da
L
S

(15) = (c T ) da
S

(16) = c (T da)
S
The last line uses the triple vector product identity A (B C) = B
(C A). Equating integrals gives

(17) T dl = (T da)
L S
If we let T = c r for a constant vector c, we get

(18) (c r) dl = (c r) da
L
S

(19) = [c ( r) + (c )r] da
S

(20) = c da
S
(21) = c a = a c
The second line omits the derivatives of c which are all zero. To get the
third line, we use r = 0 and (c ) r = c (both of which can be proved
by direct calculation).
P INGBACKS
Pingback: Magnetic dipole
ELECTROSTATICS - LINEAR CHARGE DISTRIBUTIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 2.1.4, Problems 2.3 - 2.5.
When faced with a continuous distribution of charge, we can work out
the electric field as a function of position by using integration instead of
summation. In general, we have


1 (r0 ) r r0 3 0
(1) E(r) = d r
40 |r r0 |2 |r r0 |

Here r0 is the position of volume element d 3 r0 and (r0 ) is the charge den-
sity at that point. There are three types of problems that occur commonly
with continuous charge distributions: linear, surface and volume charges.
Well do a few examples using linear charges here to see how this works in
practice. In many problems in electrostatics, its advisable to make use of
any symmetries that the configuration has.
Example 1
Suppose we have a line segment extending along the x axis from L to
L. This line segment contains a constant linear charge density (measured
in Coulombs/metre). What is the electric field at a point z on the z axis?
We can split the problem in two by solving for the x and z components of
E separately. Because the z axis divides the linear charge precisely in two,
we can use symmetry to conclude that there is no net x component in the
field. To work out the z component, we note that the contributions from +x
and x are equal.
In the formula above, r r0 is the vector from a point on the linear charge
to the point z so we get

p
(2) |r r0 | = x2 + z2

rr0

2 2
The unit vector |rr 0 | has a z component of z/ x + z and the charge

density is = so we get for the magnitude of E in the z direction:


1
ELECTROSTATICS - LINEAR CHARGE DISTRIBUTIONS 2

L
1 2 z
(3) Ez = dx
40 0 (x2 + z2 )3/2
1 2 L
(4) =
40 z L2 + z2

where you can either work out the integral by hand or look it up or use
software like Maple.
Note that for z  L, we get

1 2 L
(5) Ez
40 z2

which is equivalent to the field of a point charge q = 2 L at a distance z.


For L  z we get

1 2
(6) Ez
40 z

which is the formula for the field due to an infinitely long line of charge.
Example 2
A slight variant on this problem is to remove one half of the line segment,
so the linear charge now extends from x = 0 to x = L, with the test point still
on the z axis. The z component of the field will now be half that calculated
above:

1 L
(7) Ez =
40 z L2 + z2
However, since the problem is no longer symmetric about the origin, Ex
rr0
is no longer zero. The x component of |rr0 | is x/ x2 + z2 (the negative
sign arises because the vector points from r0 to r, and since all r0 locations
are on the +x axis, and r is on the z axis, the x component of r r0 is always
negative) so we get
L
1 x
(8) Ex = dx
40 0 (x2 + z2 )3/2

z L2 + z2
(9) =
40 z L2 + z2
For z  L:
ELECTROSTATICS - LINEAR CHARGE DISTRIBUTIONS 3

1 L
(10) Ez
40 z2
(11) Ex 0
Example 3
We now have a square loop (like a wire bent into a square) lying in the xy
plane with sides parallel to the axes and centred at the origin, and we want
to find the field at some point z on the z axis. By symmetry, the field will be
entirely along the z direction, and the contributions from all four sides will
be equal. If we consider the qedge where y = a/2, then the distance from
2
a point on this edge to z is x2 + a4 + z2 . Using the same reasoning as in
example
q 1, the z component of the unit vector connecting this point with z
2
is z/ x2 + a4 + z2 so we get

a/2
1 4 z
(12) Ez = 3/2 dx
40 a/2

2 a2 2
x + 4 +z
1 8 az
(13) =  
40 2a2 + 4z2 z2 + a2
4

As a check, when z  a, we get

1 4 a
(14) Ez
40 z2

which is the equivalent of the field due to a point charge q = 4 a.


Example 4
Now consider a circular loop of radius r in the xy plane, centred at the
origin. By symmetry, the field is entirely in the z direction. A line segment
on the circle has length rd , where is the angle in the xy plane. The
contribution from all line segments is the same. The z component of the
unit vector is z/ z2 + r2 so the field is

2
1 rz
(15) Ez = d
40 (z + r2 )3/2
2
0
1 2r z
(16) =
40 (z + r2 )3/2
2
ELECTROSTATICS - LINEAR CHARGE DISTRIBUTIONS 4

P INGBACKS
Pingback: Electrostatics - surface charges
Pingback: Electric potential from charges - Examples 1
Pingback: Electric field & potential - more examples
Pingback: Force on a dielectric sphere due to a charged wire
Pingback: Electric and magnetic forces in two charged wires
Pingback: Electron speed in a copper wire
Pingback: Currents and relativity
Pingback: Coaxial wave guides: TEM mode
Pingback: Fields due to a moving linear charge
Pingback: Gravitoelectric and gravitomagnetic acceleration for a moving
wire
ELECTROSTATICS - SURFACE CHARGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.6 - 2.8.
Now a few examples of surface charge.
Example 1
First, we consider a circular disk of radius R with surface charge density
lying in the xy plane and centred at the origin. Find the electric field at a
point on the z axis.
To solve this we can make use of the solution to the circular loop. In this
case were considering a circular ring of circumference 2r and thickness
dr, so the amount of charge in the ring is 2r dr and from the earlier
solution, the field due to this ring is

1 2r z
(1) Ez = dr
40 (z + r2 )3/2
2

To get the total field from the disk, we integrate over r:

R
1 r
(2) E= 2 z 2 2 3/2
dr
40 0 (z + r )
 
2 2
z +R z
1
(3) = 2
40 z2 + R2
To get the limiting behaviours we can Taylor-expand the result. For z 
R, we expand about R = 0 and find the leading non-zero term is in R2 :

1 1
(4) E 2 2 R2
40 2z
1 R 2
(5) =
40 z2
This is correct since the total charge on the disk is R2 so the field is that
due to a point charge of that amount.
If we let R we get
1
ELECTROSTATICS - SURFACE CHARGES 2

1
(6) E 2
40

(7) =
20

This is the field due to an infinite plane of charge. Note that the field is
independent of z so is the same no matter how far away from the plane we
are.
Example 2
We have a spherical shell of charge with radius R and surface density ,
centred at the origin. Again, we seek the field at a point on the z axis.
Using spherical coordinates, a point on the sphere has coordinates (R, , )
where is the angle from the positive z axis, and is the azimuthal angle.
We can use the cosine law to write the distance between a point on the
sphere and the field point:

p
(8) |r r0 | = z2 + R2 2zR cos
By symmetry, the field will again be in the z direction, so we need the z
component of r r0 . To get this, we need the angle between r r0 and the
z axis. To get this, project the point on the sphere onto the z axis; this gives
a point with z coordinate R cos . The remaining distance along the z axis
to the field point is therefore z R cos , but this distance is the projection
of r r0 onto the z axis. The cosine of the angle between r r0 and the z
axis is this projection divided by |r r0 |, so we get

z R cos
(9) cos =
|r r0 |
z R cos
(10) =
z2 + R2 2zR cos
Now for a given value of , we have a ring of charge with radius R sin
and thickness Rd at z distance |r r0 | from the field point, so we can
integrate over to get the total field.


(z R cos )(2R sin )(Rd )
(11) E= 3/2
40 0 (z2 + R2 2zR cos )
This integral can be done using Maple, but there are two possibilities.
First, if z > R so the field point is outside the sphere, we get
ELECTROSTATICS - SURFACE CHARGES 3

1 4R2
(12) E=
40 z2
Since 4R2 is the total charge on the sphere, we see that the sphere
behaves like a point charge for all field points outside it.
Second, if z < R so we are inside the sphere, we get

(13) E =0
So anywhere inside a spherical shell with a uniform charge distribution,
we feel no field at all.
Example 3
The result of the last example can be used to find the field due to a sphere
that contains a uniform volume charge density . Since each spherical shell
within the sphere behaves as a point charge to all points outside the shell,
the field at a point outside the sphere (z > R) is just

1 4R3
(14) E=
40 3z2
At a point inside the sphere, all shells outside the field point contribute
nothing, so we get, for z < R:

1 4z3
(15) E =
40 3z2
z
(16) =
30

The field thus increases linearly within the sphere and then falls off as an
inverse square outside.
P INGBACKS
Pingback: Electric potential from charges - Examples 1
Pingback: Electrostatic boundary conditions - examples
Pingback: Electrostatic pressure
Pingback: Field of a polarized object - examples
Pingback: Force between two sheets of current
GAUSSS LAW - EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.11 - 2.18.
Gausss law in electrostatics relates the integral over a closed surface
of the electric field to the integral over the enclosed volume of the charge
density. That is

1
(1) E da = (r)d 3 r
0

where it is important to note that the integral on the left is over the enclosing
surface (often called the Gaussian surface), while that on the right is over
the volume enclosed by that surface.
In certain rather specialized situations, Gausss law allows the electric
field to be found quite simply, without having to do sometimes horrendous
integrals. The situations rely on the geometry of the charge distribution
having some kind of symmetry. Here well give a few examples of how
Gausss law can be used in this way.
Example 1. We have a spherical shell with radius R and constant surface
charge density . By taking a spherical Gaussian surface inside the shell,
we see that E = 0 inside the shell, since there is no enclosed charge here.
Outside the shell, we can take a spherical Gaussian surface with a radius
r > R. Outside the shell E is radially symmetric. The magnitude can be
found by integrating E da over the Gaussian surface:

q
(2) 4r2 E =
0
4R2
(3) =
0
2
R
(4) E =
0 r2
Thus outside the shell, the charge behaves as a point charge at the centre
of the sphere.
Example 2. Now we take a sphere of radius R that has a uniform volume
charge density . For r < R
1
GAUSSS LAW - EXAMPLES 2


q
(5) E da =
0
4r3
(6) 4r2 E =
30
r
(7) E =
30

Outside the sphere, the sphere behaves as a point charge of magnitude


4R3 /3 so

R3
(8) E=
30 r2

Example 3. For an infinitely long charged wire of linear charge density


we can choose a cylindrical Gaussian surface of length L and radius s
centred on the wire. By symmetry the field points radially away from the
wire and the end caps contribute nothing. The enclosed charge is then q =
L and the integral over the cylindrical surface gives


L
(9) E da =
0
L
(10) 2sLE =
0

(11) E =
2s0

Example 4. A sphere of radius R carries a volume charge density = kr


where r is the distance from the centre and k is a constant. Inside the sphere,
the enclosed charge as a function of r is

r
(12) q(r) = 4 (kr0 )(r02 )dr0
0
4
(13) = kr

Therefore, using Gausss law, we get


GAUSSS LAW - EXAMPLES 3


kr4
(14) E da =
0
2 kr4
(15) 4r E =
0
kr2
(16) E =
40
Outside the sphere, the enclosed charge is kR4 so

kR4
(17) E =
40 r2
kR4
(18) =
40 r2
Example 5. A hollow spherical shell contains charge density = k/r2
for a r b. In the region r < a, E = 0 since again there is no enclosed
charge. In the region a r b we first calculate the enclosed charge.
r
k 02 0
(19) q(r) = 4 02
(r )dr
a r
(20) = 4k(r a)
Gausss law then says

q
(21) E da =
0
4k(r a)
(22) 4r2 E =
0
k(r a)
(23) E =
r 2 0
For r > b we get

k(b a)
(24) E=
r 2 0
Example 6. A coaxial cable has a cylindrical inner core of radius a with
uniform volume charge density , and an outer cylindrical shell of radius b
with a surface charge density that is of opposite sign to the charge on the
core. The surface charge density is such that the cable is electrically neutral.
Inside the inner cylinder, we can use the result of example 3. The field will
GAUSSS LAW - EXAMPLES 4

point radially outward from the cylinders axis. Since the volume charge
density is , the linear charge density for that portion of the cylinder inside
radius s is s2 , so the field is

s2
(25) E =
2s0
s
(26) =
20
Between the inner cylinder and the outer shell, the linear charge density
is a2 so the field becomes

a2
(27) E =
2s0
a2
(28) =
2s0
Outside the outer shell, the total enclosed charge is zero since the cable
is neutral, so E = 0.
Example 7. An infinite plane slab has thickness 2d, and carries a uniform
volume charge density . If the y axis is perpendicular to the plane and the
plane y = 0 is the centre plane of the slab, we can choose a Gaussian surface
that is a cylinder of radius a with axis perpendicular to the slab and thickness
2y. The enclosed charge is a2 (2y), and by symmetry E points away from
the slab on both sides and only contributes on the ends of the cylinder, so

q
(29) E da =
0
2a2 y
(30) E(2a2 ) =
0
y
(31) E =
0
Outside the slab

d
(32) E=
0
That is, the electric field is constant no matter how far from the slab we
are.
Example 8. We have two spheres, each of radius R, one of which has
volume charge density + and the other of which has density . The
vector from the centre of the positive sphere to the centre of the negative
GAUSSS LAW - EXAMPLES 5

sphere is d. The two spheres have a region of overlap and we want the
electric field within this region.
We might be tempted to say that since, in the region of overlap, any vol-
ume contains zero net charge (since the densities are equal and opposite),
there is zero field within this region. However, the problem with this argu-
ment is that when working out the surface integral of the field, there is no
obvious symmetry we can invoke. Thus although it is correct to say that any
integral E da over a closed surface entirely within the region of overlap
is zero, this doesnt automatically translate to the field being zero as it did
in earlier examples.
The problem does, however, have a simple solution. If we look back
at example 2, we see that the electric field inside a uniformly positively
charged sphere is (restoring the vector notation)


(33) Er = r
30

where r is the vector from the centre of the sphere to the point in question.
Now suppose that s is the vector from the centre of the negative sphere to
the same point. Because the charge is negative, we get


(34) Es = s
30

so the total field is, using superposition

(35) E = Er + Es

(36) = (r s)
30

(37) = d
30

where d is the vector joining the two centres. Thus the field is constant in
the region of overlap, although it is not zero.
This is a bit of a trick question, since it relies on the field being directly
proportional to the radius vector. For other geometries, no such simple
solution exists.
P INGBACKS
Pingback: Electric potential - examples
Pingback: Electrostatic boundary conditions - examples
Pingback: Work and energy - continuous charge
GAUSSS LAW - EXAMPLES 6

Pingback: Electrostatics - conductors


Pingback: Electric field & potential - more examples
Pingback: Potential of two charged wires
Pingback: Laplaces equation - average values of solutions
Pingback: Polarizability of hydrogen
Pingback: Polarizability of hydrogen - quantum version
Pingback: Polarizability - linear charge distribution
Pingback: Field of a polarized object
Pingback: Ohms law, conductivity and resistivity
Pingback: Negative heat capacity in gravitational systems; estimating the
Suns temperatur
CURL & POTENTIAL IN ELECTROSTATICS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 2.20.
An important consequence of the curl of the electric field being zero in
electrostatics (that is, in the absence of any moving charges) comes from
Stokess theorem, which for the electric field is


(1) ( E) da = E dl
S L

where the integral on the left is taken over some open surface S and that on
the right is over the boundary curve L of that surface. Since the curl is zero,
this means that the line integral of the electric field around any closed path
is zero. This means that the line integral of E dl along a path connecting
two points a and b is independent of the path. This follows since if the
integral did depend on the path, we could choose one path from a to b and
then a different path (where the integral had a different absolute value) on
the return path b to a. In that case the integral over the combined path,
which is a closed curve, would not be zero, which isnt allowed.
In fact the path integral is another way of defining the electric potential
function V : the potential difference between two points is the negative of
the line integral of E dl along any path connecting those two points:

b
(2) V (b) V (a) = E dl
a

Although it is only the potential difference which has any physical signif-
icance, it is traditional to define the potential function so that it has a definite
value at each point. To do this, we need to specify a reference point p at
which V (p) = 0 by definition, and then calculate all other values of the po-
tential by integrating from that reference point to the point in question. For
many applications, the reference point is chosen as infinity, because for any
localized distribution of charge, its effect will fall off to zero as we get in-
finitely far from it. However, this isnt cast in stone, and in some problems,
other reference points are more convenient.
1
CURL & POTENTIAL IN ELECTROSTATICS 2

The curl condition E = 0 means that we cant specify any old vector
field as a description of an electric field in space. For example, if we tried
to specify E as
 
(3) E = c xyi + 2yzj + 3xzk

we can prove this isnt a valid expression for an electric field by calculating
its curl. We have

i j k


(4) E x

y z


cxy 2cyz 3cxz

We need to work out only the x component to discover that it is 2cyi 6=


0, which means the curl cannot be zero so this is not a valid electric field.
However, the field given by

E = c y2 i + (2xy + z2 )j + 2yzk
 
(5)

is a valid electric field since E = 0 as can be verified by direct calcula-


tion.
We can work out the potential function for this field, but we need to
specify a reference point. Although this is a valid electric field, it is a bizarre
one since it gets larger the further from the origin we are. Since the field
is zero at the origin, this seems a sensible place to set our reference point
for the potential. We can use 2 to work out the potential, but we need a
path of integration. The integral is independent of the path so it doesnt
matter which path we choose. Since the field is specified in rectangular
coordinates, a sensible path from the origin out to a point (x, y, z) seems to
be as follows. Start at the origin and move along the x axis to the point
(x, 0, 0). Then from this point, keep x and z constant and follow the line
from (x, 0, 0) to (x, y, 0). Finally, keep x and y constant and move along the
line from (x, y, 0) to (x, y, z).
Along the first line segment, E dl = cy2 dx, but since y = 0 along our
(x,0,0)
chosen path, the integral (0,0,0) cy2 dx0 contributes zero.
Along the second line segment E dl = c(2xy + z2 )dy = 2xydy since z = 0
(x,y,0)
along this segment. We get (x,0,0) 2cxy0 dy0 = cxy2 .
(x,y,z)
Finally along the third line segment E dl = 2cyzdz and (x,y,0) 2cyz0 dz0 =
cyz2 .
CURL & POTENTIAL IN ELECTROSTATICS 3

The total potential is the sum of these three contributions, so we get

(6) V (x, y, z) = (0 + cxy2 + cyz2 )


(7) = cxy2 cyz2
The answer can be verified directly by calculating E = V .
P INGBACKS
Pingback: Electrostatic boundary conditions
Pingback: Electric potential - examples
Pingback: Work and energy - point charges
Pingback: Potential of two charged wires
Pingback: Deviation from Coulombs law
ELECTRIC POTENTIAL - EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.21 - 2.24.
Weve seen that the electric field can be expressed as the gradient of a
potential function

(1) E = V

and that the potential can be calculated from a line integral of the field
r
(2) V (r) = E dl
a

where a is some arbitrarily chosen reference point. The actual value of


a doesnt matter, since it always disappears when calculating the field, or
when calculating a potential difference. However, if we want to define a
potential function, we do need to specify this point.
Calculating V from the field is something of an artificial exercise, since
usually it is the field we want, not the potential. However, its useful to run
through a few examples to see how the potential can be calculated in this
way.
Example 1. We have a uniformly charged solid sphere with radius R
and total charge q. We can find the potential at any point inside or outside
the sphere, since we worked out the field as Example 2 in an earlier post,
although there we used the charge density rather than the total charge q.
For r < R:

r
(3) E=
30

where is the charge density. In terms of q this is


for r < R:

r
(4) E=
30
1
ELECTRIC POTENTIAL - EXAMPLES 2

where is the charge density. In terms of q this is

4R3
(5) q =
3
3q
(6) =
4R3
So

q r
(7) E=
40 R3
Outside the sphere

R3
(8) E =
30 r2
q
(9) =
40 r2
Using infinity as the reference point, the potential outside the sphere is
r
q 1
(10) V = dr0
40 (r0 )2
q 1
(11) =
40 r
Inside, we have
r
q 1 q r0 0
(12) V= 3
dr
40 R R 40 R
 
q 1 1 2 2

(13) = + R r
40 R 2R3
r2
 
q
(14) = 3 2
80 R R
Example 2. For an infinitely long charged wire of linear charge density
we can use the field calculated in Example 3 on the earlier post. The
electric field is, for a distance s from the wire


(15) E=
2s0

The reference point for the potential calculation in this case is a bit tricky.
ELECTRIC POTENTIAL - EXAMPLES 3

Suppose we choose some arbitrary distance from the wire, which well rep-
resent as s = a. Then
s
1 0
(16) V = ds
20 a s0
s
(17) = ln
20 a
Choosing either a = or a = 0 doesnt work, since the logarithm term
blows up at both points. However, since the reference point is arbitrary, we
might just as well leave it at some finite, non-zero value of a.
Since the potential depends only on the distance s, the gradient in cylin-
drical coordinates contains only a radial term, and we get

(18) E = V
 
a1
(19) = s
20 s a

(20) = s
2s0
Note that a cancels out, so its precise value doesnt matter.
Example 3. A hollow spherical shell contains charge density = k/r2
for a r b. We worked out the field in Example 5 in the earlier post. The
electric field is E = 0 for r < a; E = k(ra)
r2 0
for a < r < b and E = k(ba)
r 2 0
for
r > b. Using infinity as the reference point, we can get the potential at the
centre of the sphere.
b a
k(b a) k(r a)
(21) V = dr dr
r2 0 b r2 0
(There is no integral for r < a since E = 0 there.) The integrals are all
fairly simple, and we get

k b
(22) V= ln
0 a
We can get the potential in the region a < r < b by evaluating the integrals

b
k(b a) 0 r
k(r0 a) 0
(23) V = dr dr
r02 0 b r02 0
 
k a b
(24) = 1 + ln
0 r r
ELECTRIC POTENTIAL - EXAMPLES 4

Finally, the potential outside the sphere (r > b) is


r
k(b a) 0
(25) V = dr
r02 0
k ba
(26) =
0 r
Note that the potentials are all continuous at the various boundaries (al-
ways a good idea to check this!).
Example 4. A coaxial cable has a cylindrical inner core of radius a with
uniform volume charge density , and an outer cylindrical shell of radius b
with a surface charge density that is of opposite sign to the charge on the
core. The surface charge density is such that the cable is electrically neutral.
We worked out the field in Example 6 in the earlier post. The field inside
s
the inner cylinder is E = 2 0
and between the inner cylinder and the outer
2
a
cylinder it is E = 2s0
where s is the distance from the axis. The potential
difference between the axis and the outer cylinder is then

b
(27) V (b) V (0) = E dl
0
a b 2
s a
(28) = ds ds
0 20 a 2s0
a2
 
b
(29) = 2 ln + 1
40 a
Note that in calculating a potential difference, we dont need to use a
reference point for the potential.

C OMMENTS
October 26, 2017 at 7:15 am
tomas says: I have a simple question about the Electrodynamics Spher-
ical potential example 1. If the path integral goes from infinity to r, why
doesnt the dl point inwards? It would appear an extra minus which will
make it all wrong. But arent the Electric Field and the dl pointing in opos-
site directions?
Thanks.
==================
I think its just convention that if you choose your reference point to be
at infinity, then dl always points away from the origin. You can always fix
the sign of V if youre calculating it from E dl, since you know that you
ELECTRIC POTENTIAL - EXAMPLES 5

must have E = V . Thus if E points outwards (away from the origin),


then V has to decrease as you move away from the origin.
P INGBACKS
Pingback: Electric potential from charges - Examples 1
Pingback: Electric potential from charges - Examples 2
Pingback: Laplaces equation in cylindrical coordinates
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 1

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.25-2.26.
Weve seen that the electric potential function can be defined in terms of
the charge distribution:

1 1
(1) V (r) (r0 )d 3 r0
40 |r r0 |
This is the more usual problem in calculating the potential, as opposed
to the case we considered earlier where the potential is calculated from the
electric field. Usually, the problem we wish to solve is: given a charge
distribution, find the electric field. Although this can be solved by doing
a vector integral directly, often finding the potential first and then using
E = V to find the field is easier.
Here well consider a few examples of calculating the field from the
charge distribution.
Example 1. We have two point charges, each of charge +q a distance d
apart, so that they lie on the x axis at locations x = d/2. Find the potential
at a point on the z axis. In this case, we can use the discrete version of the
potential formula, which can be obtained from 1 by using delta functions
for the point charges, or else we can use a simple sum formula.

1 2 qi
(2) V=
40 i=1 |r ri |
2q
(3) = p
40 z2 + d 2 /4
The field from this distribution along the z axis points in the z direction
by symmetry, and is

(4) E = V
2qz
(5) = z
40 (z2 + d 2 /4)3/2
1
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 1 2

This answer agrees with that calculated directly. Note that we cannot
calculate the field off the z axis from this potential formula, since at other
locations the symmetry does not apply. We would need to calculate a more
general formula for the potential first.
This is illustrated more clearly if we change one of the charges to q.
In that case, the above formula gives V = 0 on the z axis. We cant infer
the electric field from this however, since there is no longer any symmetry
along the z axis. The field is, of course, not zero; rather it points in the x
direction.
Example 2. We have a linear charge extending from x = L to x = +L
and wish to find the potential along the z axis. The potential due to an
element of the linear charge of length dx is ( /40 z2 + x2 )dx so the
total potential is

L
2 dx
(6) V=
40 0 z2 + x 2
2 h p i
(7) = ln(L + z2 + L2 ) ln z
40

Since the configuration is again symmetric about the z axis, we can cal-
culate the field

(8) E = V

2 z 1
(9) =  +
40 z2 + L2 L + z2 + L2 z
2

2 2
L z +L +L2
(10) =  
40 z z2 + L2 L + z2 + L2

2 L
(11) =
40 z z + L2
2

which agrees with Example 1 earlier.


Example 3. A flat circular disk of radius R lies in the xy plane with its
centre at the origin. Find the potential on the z axis.
The potential due
to a ring of charge of thickness dr is (2 r/40 z + r2 )dr so the total
2
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 1 3

potential from the disk is


R
r
(12) V= dr
20 0 z2 + r 2
hp 2 i
(13) = z + R2 |z|
20
Due to the symmetry of the situation, the potential must be the samefor
z, which is why weve used |z| in the last line. This term comes from z2
so we need to decide which root to take in each case.
The distribution is again symmetric about the z axis, so the field is

(14) E = V
 
|z|
(15) = 1 z
20 z2 + R2
which agrees with the Example 1 in an earlier post.
Example 4. Given a right circular cone with base radius R and height
h = R, we can find the potential difference between the centre of the base
and the tip of the cone. Let the cones tip be at z = 0 and let the cone open
upwards (like an ice-cream cone), so the centre of its base is at z = R. To
calculate the potential, we can slice the cone horizontally into thin slabs,
each of radius x at height z. Since the slant of the cone is constant, we have
z = x for each of these slabs. The area of the rim of each slab can be found
as follows.
Since the cones surface is slanted, its not correct to say that the surface
area of the rim ofpeach slab is just 2xdz.p Using Pythagoras, the actual sur-
2 2 2
face area is 2x (dx) + (dz) = 2x 1 + (dz/dx) dx = 2 2xdx since
in this specialized case, z = x for all points on the cone.
If the surface charge
density is , then the amount of charge on the sur-
face of a slab is 2 2x dx. For a general point zV on the z axis, the poten-
tial of this ring of charge at height z is


2 xdx
(16) dV = p
20 x + (zV z)2
2

2 xdx
(17) = p
20 x2 + (zV x)2

where the last line uses z = x for points on the cones surface. Thus the total
potential for the cone is
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 1 4

R
2 xdx
(18) V= p
20 0 x2 + (zV x)2
This integral can be evaluated using software, but its not pretty. How-
ever, for the special cases of the centre of the base of the cone and the tip of
the cone, the answer is a bit less messy. For the tip of the cone zV = 0 and
we get

R
2 dx
(19) V (0) =
20 0 2
R
(20) =
20
For the centre of the base, zV = R and we get

R
2 xdx
(21) V (R) = p
20 0 x2 + (R x)2
!
2 2R 2+1
(22) = ln
20 4 21
!
R 2+1
(23) = ln
40 21

The potential difference is then

" ! #
R 1 2+1
(24) V (R) V (0) = ln 1
20 2 21
R 1 
  
2
(25) = ln ( 2 + 1) 1
20 2
R    
(26) = ln 2+1 1
20

In the second
line, we multiplied the argument of the logarithm top and
bottom by 2 + 1.
In order to calculate the field, we would need to evaluate the potential for
general values of zV and then calculate the gradient.
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 1 5

P INGBACKS
Pingback: Electric potential from charges - Examples 2
Pingback: Laplaces equation - charged line segment
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 2

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.27 - 2.28.
Here are a couple more examples of calculating the potential from the
charge distribution.
Example 1. Given a uniformly charged solid sphere of radius R and total
charge q, find the potential from the formula

1 1
(1) V (r) = (r0 )d 3 r0
40 |r r0 |
In this case, the charge density is constant, and can be expressed in terms
of q by

3q
(2) =
4R3
Using the cosine law, we have
p
(3) |r r0 | = r2 + r02 2rr0 cos

so in spherical coordinates, we get

2 R
3q r02 sin d dr0 d
(4) V (r) =
16 2 0 R3 0 0 0 r2 + r02 2rr0 cos
r2
 
q
(5) = 3 2
80 R R

which agrees with the result in Example 1 in an earlier post.


Example 2. Given a uniformly charged solid cylinder of length L, radius
R and charge density with its axis along the z axis and centre at the origin,
find the potential at location zV on the z axis, where were assuming that
zV > L/2.
We can use the result of Example 3 in an earlier post. The potential of a
charged disk is
1
ELECTRIC POTENTIAL FROM CHARGES - EXAMPLES 2 2

hp 2 i
(6) V= z + R2 z
20

where z is the distance above the centre of the disk. In the case of the
cylinder, we can slice the cylinder into a number of disks, each of thickness
dz. For a slice at height z the distance from the centre of the slice to location
zV is then zV z, so the potential of the entire cylinder will be
L/2 q 

(7) V= (zV z)2 + R2 zV + z dz
20 L/2
This is an unpleasant integral, but can be done with software. Maple
gives (after condensing a few of the terms):

(8)
  q  
20 L zV L zV
q
V= 2
(2zV + L) + (2R)2 + 2
+ (2zV L) + (2R)2
8 4 8 4
(9)
"p #
R2 (2zV L)2 + (2R)2 2zV + L
+ ln p zV L
2 (2zV + L)2 + (2R)2 2zV L
Since the system has symmetry about the z axis, we can calculate the
electric field along the axis by taking the gradient. After using Maple and
doing some simplification we get (remember that we take the derivative
w.r.t. zV in the gradient):

(10) E = V
 
1 1
q q

(11) = L (2zV + L)2 + (2R)2 + (2zV L)2 + (2R)2 z
20 2 2
P INGBACKS
Pingback: Work and energy - continuous charge
Pingback: Laplaces equation - average values of solutions
Pingback: Laplaces equation - charged disk
ELECTROSTATIC BOUNDARY CONDITIONS - EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 2.30.
Weve seen that the electric field has a discontinuity of /0 when we
cross a surface charge of density , but that the potential is continuous.
Here we offer some examples that verify this condition.
Example 1. From Example 1 in a previous post, we know that the electric
field for an infinite plane is


(1) E= n
20

where is the surface charge density and n is a unit vector normal to the
plane, pointing away from the plane on both sides. The difference in field
as we cross the plane is therefore /0 .
Example 2. If we have two infinite, parallel planes of charge, with one
having a charge density of + and the other of , then by the principle of
superposition, the field will be zero everywhere except between the planes,
where it will be /0 pointing from the positive plane to the negative plane.
Crossing either plane thus gives rise to a discontinuity in the field of /0 .
Example 3. For a spherical shell of charge, from Example 1 in a previous
post, we know that the field is E = 0 inside the shell and points radially
outward with a magnitude of

R2
(2) E=
0 r2

where R is the radius of the sphere and r is the distance from the centre. At
the boundary, r = R and E = /0 so the discontinuity across the shell is
again /0 .
Example 4. Consider an infinitely long, hollow, cylindral tube with sur-
face charge density . By symmetry, the field points radially outwards, so
we can use Gausss law to find it. If the radius of the tube is R and we
consider a length L then if we consider a Gaussian cylinder enclosing this
section, the total enclosed charge is 2RL . If the Gaussian cylinder has
radius r its area (minus the end caps, which dont contribute) is 2rL so
1
ELECTROSTATIC BOUNDARY CONDITIONS - EXAMPLES 2


q
(3) E da =
0
2RL
(4) 2rLE =
0
R
(5) E =
r0
At the boundary r = R so the field outside is /0 . Inside, since there
is no enclosed charge, E = 0 so again the discontinuity at the boundary is
/0 .
Example 5. We can work out the potential due to a spherical shell of
charge from the formula

1 1
(6) V (r) = 0
(r0 )dr0
40 |r r |
If we consider a sphere of radius R, and determine the potential at a point
z on the z axis, then

p
(7) |r r0 | = z2 + R2 2zr cos

where is the angle between the z axis and the vector r0 . The integral we
need to do in this case is


2
R2 sin d
(8) V (z) =
40 0 z2 + R2 2zr cos
The value of this integral depends on the relative sizes of z and R. If z < R
so we are inside the sphere, then

R
(9) V (z < R) =
0
If z > R so we are outside the sphere, then

R2
(10) V (z > R) =
z0
At the boundary, z = R and V is R /0 on both sides, so it is continuous.
The gradient is zero inside the shell, while outside it is
ELECTROSTATIC BOUNDARY CONDITIONS - EXAMPLES 3

R2
(11) V =
z2 0

so at the boundary z = R and the gradient is /0 . Thus we get the discon-


tinuity of the field as before.
P INGBACKS
Pingback: Electric field & potential - more examples
WORK AND ENERGY - POINT CHARGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 2.4, Problem 2.31.
Since charges exert forces on each other through their electric fields, it
will require the expenditure of energy, or work, to assemble any configura-
tion of charges. Here well have a look at how much energy is required to
assemble, and thus how much energy is stored, in a collection of discrete
charges.
The force on a charge q due to an electric field E is qE. From elementary
physics, we know that the work done when an object is moved against a
force is the negative (since were opposing the force) of (force) times (dis-
tance). In general, if the force varies as a function of position, we get

(1) W = F dl

where the integral is taken over the path through which the object is moved.
The minus sign is an indication that we are opposing the force F; if we work
instead with the force that we must exert to move the object, then the minus
sign is omitted.
For the electric field, then, we get

b
(2) W = F dl
a
b
(3) = q E dl
a
(4) = q(V (b) V (a))

To get the last line, weve used the fact that, in electrostatics, the line inte-
gral of the electric field is independent of the path; it depends only on the
endpoints a and b. Weve seen earlier that the line integral of the field is the
negative of the potential difference between the two endpoints.
So, in other words, the potential difference between two points is the
work per unit charge required to move a charge between those two points.
1
WORK AND ENERGY - POINT CHARGES 2

If weve set the reference point for the potential at infinity (that is, V = 0 at
infinity), then the work required to bring in a charge from infinity to a point
r is

(5) W = qV (r)
We can apply this formula to find out how much energy is required to
assemble a collection of point charges. To place a single charge q1 at a
location r1 takes no work, since there are no fields to work against. Bringing
in a second charge q2 requires working against the field due to q1 . The
1
potential due to q1 is V (r) = 4 0
q1 /|r r1 |, so if we want to place q2 at
position r2 the work required is

1 q1
(6) W2 = q2
40 |r2 r1 |
Before we go any further, its worth noting that this formula gives rise to
a bit of a problem. What if we want to assemble a point charge itself? That
is, suppose we want to build up a point charge of a certain size by bringing
together other point charges and, in effect, gluing them together. This seems
to be a valid procedure, since after all, if a charge is truly a mathematical
point, we should be able to pile as many of these point charges on top of
each other as we like without increasing the volume (that is, zero) occupied
by the sum of all the charges.
However, if we try that, the above formula says this will require an infi-
nite amount of work (since r2 = r1 ). This is, in fact, a recognized problem
in electrodynamics, and the problems dont go away even in the quantum
mechanical theory. In fact, we cant even get out of the problem by saying
that there is no such thing as a point charge, since a lot of physicists think
that the electron might actually be a point charge (at least its diameter, if its
non-zero, is so small that nobody has actually measured it yet).
With that caution in mind, lets ignore the problem and carry on. If we
assume that the existence of point charges is possible (without taxing our
minds as to how they are built), we can continue to add more point charges
to our distribution. Adding a third charge q3 at location r3 requires work

(7) W3 = q3V1,2 (r3 )


 
1 q1 q2
(8) = q3 +
40 |r3 r1 | |r3 r2 |
The total work required to assemble all three charges is then
WORK AND ENERGY - POINT CHARGES 3

(9) W1,2,3 = W1 +W2 +W3


 
1 q1 1 q1 q2
(10) = 0+ q2 + q3 +
40 |r2 r1 | 40 |r3 r1 | |r3 r2 |
The general pattern should be fairly obvious by now. To assemble n
charges, the total work is

1 n n qj
(11) W= q i
40 i=1 j=i+1 |ri r j |

If we extend the second sum to cover 1 through n (excluding j = i), we


get

1 1 n n qj
(12) W= qi
2 40 i=1 j=1, j6=i |ri r j |
1 n
(13) = qiV (ri )
2 i=1

where V (ri ) is the potential due to all the charges in the collection except
qi .
As an example, suppose we have arranged two charges of q at the ends
of a diagonal in a square, and a charge of +q on one of the other two corners
of the square. How much work is required to bring in another charge of +q
from infinity and place it at the remaining corner of the square?
To work this out, we need to find the potential at this corner due to the
existing three charges. If the length of each side of the square is a, then we
get

(14) W4 = qV1,2,3 (r4 )


 
1 2q q
(15) = q +
40 a 2a
1 q 2 
1

(16) = 2
40 a 2
Once all four charges have been assembled, the total energy stored in the
collection is
WORK AND ENERGY - POINT CHARGES 4

1 n
(17) W= qiV (ri)
2 i=1
1 2q2
 
1 1
(18) = 2+ 2
2 40 a 2 2
q 2 
1

(19) = 2
20 a 2
h second line,i the potential at the location of one of the q charges
In the
is 40 aq + qa q2a . Multiplying this by q gives the first two terms in
1

the square brackets. Similar logic for one of the +q charges gives the last
two terms in the brackets. The factor of 2 in the numerator arises from the
fact that there are two each of q and +q charges.
P INGBACKS
Pingback: Stimulated emission of radiation: lasers
Pingback: Work and energy - continuous charge
Pingback: Method of images
Pingback: Method of images - point charge and sphere
Pingback: Method of images: two conducting planes
Pingback: Energy in a dielectric
WORK AND ENERGY - CONTINUOUS CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.32, 2.33.
Weve seen that, for discrete point charges, the work required to assemble
a collection of n charges qi is

1 n
(1) W= qiV (ri)
2 i=1

where V (ri ) is the potential at the location of charge qi due to all the other
charges (excluding qi ).
For a continuous charge density we can write this as an integral over
the volume containing the charges:


1
(2) W= (r)V (r)d 3 r
2
Note, however, that there is a subtle distinction between the discrete and
continuous formulas. In the discrete formula, the potential term in the sum
excludes the charge qi , but in the integral form, the potential is the complete
potential due to the entire charge distribution. In the continuous case, we
dont talk about point charges (unless we write the density as a sum of
delta functions), so in that sense the continuous formula is more accurate.
However, if point charges such as electrons do truly exist, then we cant
either build them or take them apart, so it seems fair enough to exclude
the energies involved in doing so. However, as weve seen in the post on
discrete charges, the energy associated with a point charge is actually infi-
nite, so its dodgy to just ignore it. This problem plagues both classical and
quantum electrodynamics, but most books just ignore it.
The integral formula can be expressed in terms of the electric field by
using a bit of vector calculus. We know from Gausss law for electrostatics
that

(3) = 0 E
1
WORK AND ENERGY - CONTINUOUS CHARGE 2

so we can write the work as



0
(4) W= ( E)V (r)d 3 r
2
A theorem from vector calculus says

(5) (V E) = V E + E V
(6) V E = (V E) E V
(7) = (V E) + E 2
In the last line, we used the relation E = V .
Therefore, the work becomes

0
(V E) + E 2 d 3 r
 
(8) W=
2
We can now use the divergence theorem on the first term, and convert it
from a volume integral to a surface integral if we select some surface that
encloses all the charge (were assuming that were dealing with realistic
systems so that all the charge is at a finite distance, and not with things like
infinite planes of charge). That is, we can say

0 0
(9) W= V E da + E 2d3r
2 2
Since all we require is that the surface encloses all the charges, we can
let the surface tend to infinity. In that case, since the charges are all at
finite distances, and we are taking the potential to be zero at infinity, and
the electric field falls off as 1/r2 , the surface integral will go to zero at
infinity. The volume integral is always positive (since were integrating the
square of the field, which is always positive), so what happens is that as we
include more volume, the surface integral decreases and the volume integral
increases in such a way as to keep the total work constant. That is, we get

0
(10) W= E 2d3r
2

where the integral now covers all space. Note the distinction between 2 and
10: in the first case, we need to integrate only over that volume where 6= 0;
in the second case we need to integrate over all space, since in general the
WORK AND ENERGY - CONTINUOUS CHARGE 3

electric field is always non-zero over any finite distance, even for a localized
charge distribution.
As an example, we can work out the energy stored in a uniformly charged
solid sphere of radius R and charge q. Well do it four different ways to show
how each of the above methods works.
Example 1. We can use 2. We found the potential of the sphere earlier
(Example 1 in this post). The charge density in terms of the total charge
is

3q
(11) =
4R3

and the potential inside the sphere (all we need here, since we need to inte-
grate only over that volume where 6= 0) is

r2
 
q
(12) V= 3 2
80 R R

Therefore, we get


1
(13) W= V d 3 r
2
R 2 
3q2 r2

(14) = 3 2 r2 sin d d dr
32 2 0 R4 0 0 0 R
3 q2
(15) =
20 0 R

Example 2. Same problem, but now we use 10. We worked out the field
earlier (Example 2 in this post). In this case, we will need the field for both
inside and outside the sphere, since we must integrate over all space. We
have, for inside:

r
(16) Ein =
30
rq
(17) =
40 R3

and for outside:


WORK AND ENERGY - CONTINUOUS CHARGE 4

R3
(18) Eout =
30 r2
q
(19) =
40 r2
The energy is then

(20)
 R 2 2 
0 2 2 2 2
W= Ein r sin d d dr + Eout r sin d d dr
2 0 0 0 R 0 0
(21)
3 q2
=
20 0 R
Example 3. This time we use the formula 9, so the integral is split be-
tween a volume integral and a surface integral. If we use a surface of radius
a > R, then the volume component can be worked out using the same inte-
grals as in Example 2, but changing the limit on the second integral.

(22)
 R 2 a 2 
0 2 2 2 2
Wvol = Ein r sin d d dr + Eout r sin d d dr
2 0 0 0 R 0 0
(23)
q2
= (6a 5R)
400 aR

Note that as a , this integral tends to the total energy as worked out in
3 q2
the previous two examples: Wvol 20 0 R .
The surface integral uses the potential and field at a distance r = a. This
time we need the potential outside the sphere, which is

q 1
(24) Vout =
40 a
The field at r = a is

q
(25) Eout =
40 a2
Both of these are constants over the bounding sphere, so we get
WORK AND ENERGY - CONTINUOUS CHARGE 5

2
0 q2
(26) Wsur f = a2 sin d d
2 16 2 a3 0 0
q2
(27) =
80 a

The total energy is

(28) W = Wvol +Wsur f


3 q2
(29) =
20 0 R

Example 4. Finally, we can build up the solid sphere by adding suc-


cessive layers of charge of thickness dr. We have, when the sphere has
intermediate radius r:

4 3
(30) q = r
3
(31) dq = 4r2 dr

This amount of charge is brought from infinity and added to a spherical


volume of charge q and radius r. We know from above that the potential of
such a sphere at its outer boundary is

q
(32) V =
40 r
r2
(33) =
30

so the amount of work needed to add this charge to the sphere is

(34) dW = V dq
4 2 4
(35) = r dr
30

We can then integrate this to find the total work:


WORK AND ENERGY - CONTINUOUS CHARGE 6


4 2 R 4
(36) W = r dr
30 0
4 2 5
(37) = R
150
3 q2
(38) =
20 0 R
P INGBACKS
Pingback: Conservation of energy in a capacitor
Pingback: Energy in a dielectric
Pingback: Energy in a magnetic field
Pingback: Poyntings theorem
ELECTROSTATICS - CONDUCTORS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 2.5, Problems 2.35 - 2.36.
A conductor is a substance containing charge that is free to move under
the influence of an electric field. A perfect conductor contains an unlim-
ited amount of free charge. Many metals come quite close to being perfect
conductors.
The microscopic processes that occur within a conductor are the sub-
ject of more advanced physics, but for now, we can visualize a conductor
as a substance in which one or more electrons per atom are free to move
throughout the volume of the conductor. Thus the charge that moves is
usually negative charge (since electrons are negative), although we often
talk about both positive and negative charge moving within a conductor.
The motion of positive charge is accomplished by the motion of negative
charge in the opposite direction.
Conductors have several properties that at first glance seem quite strange.
Property 1. E = 0 everywhere inside a conductor. The simple explana-
tion for this is that, if E 6= 0 somewhere inside a conductor, then any charge
present there would move in response to the field, and rearrange itself until
the field was zero. We can see how this works by considering what happens
if a conductor is placed in an external electric field. The field will act on
the charge inside the conductor, and the charge will move. Eventually this
charge will hit a boundary of the conductor and be forced to stop. This mo-
tion of the charge will generate its own electric field, which will continue
to adjust itself until the net field (external + induced) inside the conductor
is zero, and the charge stops moving.
Property 2. The electric field at the outside surface of a conductor is
normal to the surface. This follows from a similar argument as Property
1. If there were any component of E parallel to the surface, it would cause
charge to move along the surface until such a component was cancelled out.
Property 3. The charge density = 0 everywhere inside a conductor.
This follows from Property 1 and Gausss law. Since the field is zero every-
where inside, then E = /0 = 0 everywhere inside as well. Note that
this does not imply that the surface charge density is zero on a conductor.
As we saw in Property 1, free charge will be forced to the surface by an ex-
ternal field, so charge tends to pile up there. Property 2 states that any force
1
ELECTROSTATICS - CONDUCTORS 2

felt by charge at the surface of a conductor is normal (and outward) at the


surface, so this surface charge gets stuck there by the balance of the outward
electric field and atomic forces holding the charge inside the conductor.
Property 4. The potential is constant everywhere within a conductor.
This follows from the fact that E = 0 and E = V .
An external electric field, such as that from a point charge or other charge
distribution placed nearby, will induce a surface charge distribution on a
conductor. For example, if we place a positive point charge +q near a solid
metal sphere, the side of the sphere nearest the charge will accumulate some
negative charge, with a corresponding build up of positive charge on the far
side of the sphere. Thus the point charge and the sphere will attract each
other.
There are various mathematical techniques for calculating the induced
charge distributions in conductors, but well leave most of these till later
posts. Here well consider the special case of a cavity inside a solid con-
ductor.
Suppose we have a solid conductor (the precise shape doesnt matter at
this stage) and it contains a cavity (again, the precise shape doesnt matter
- all were requiring here is that the cavity is totally surrounded by the con-
ductor, so there are no holes leading to the outside world). If a point charge
+q is placed inside the cavity, what happens?
This point charge generates a field that must be cancelled by charge re-
distributing itself inside the conductor. Since the point charge is positive,
it will attract a negative charge distribution on the inner wall of the cavity,
and this charge will distribute itself in such a way that the electric field in
the conductor surrounding the cavity is zero. By applying Gausss law to
a volume enclosing the entire cavity (that is, the wall of the cavity and the
point charge within it) we see that the amount of negative charge that is at-
tracted to the cavitys inner surface must be precisely q. Why? Because
the electric field on the surface of the Gaussian volume is zero (its all inside
the conductor), so the total enclosed charge must also be zero. (Remember
Gausss law says that E da = Q/0 where Q is the total charge enclosed
by the surface.)
Since a total of q has been attracted to the inner surface, an amount
of charge +q must therefore distribute itself over the outer surface of the
conductor in order to maintain electric neutrality. The combination of the
fields from these three sources (point charge in the cavity, induced nega-
tive charge on the inner surface, and positive charge on the outer surface)
conspire to provide a zero field inside the conductor.
Now lets get a bit more specific and require the conductor to be a solid
metal sphere with a cavity (the cavity can still be any shape) inside it. How
does the charge on the outer surface distribute itself?
ELECTROSTATICS - CONDUCTORS 3

To answer this question fully requires that we prove that the configuration
giving rise to the condition that E = 0 inside the conductor is unique, but
that will have to wait till another post. What we can do is show that there
is at least one configuration that does satisfy this requirement (and once we
know that the configuration is unique, we know we have the only answer).
The argument goes like this: we suppose that the negative charge that
gets induced on the inner surface is capable of neutralizing the effect of the
point charge without any help from the positive charge on the outer surface.
Is it reasonable to say that? Yes, because we can visualize a situation where
the sphere is so large (light years in diameter if you like) that any charge on
the outer surface is so far away it has negligible effect on the area around
the cavity. If the inner surface and point charge fields cancel, the positive
charge will distribute itself over the outer surface uniformly, since we know
that the field inside a spherical shell is zero (see Example 1 in this post). In
other words, we are claiming that there are two separate charge distributions
(one consisting of the point charge and the cavity wall; the other from the
outer surface) and that the field inside the conductor is separately zero from
each of these two distributions.
We therefore reach the surprising conclusion that, for a conducting sphere,
the outer charge distribution does not depend on the shape or location of the
cavity; the sphere will have charge +q (equal to the point charge inside the
cavity) distributed uniformly over its surface.
Well now look at a couple of examples of how we can apply this.
Example 1. A conducting sphere of radius R carries a net charge q. This
sphere is surrounded by a concentric conducting shell with an inner radius
a and outer radius b. The shell has no net charge.
First, the inner sphere must have a surface charge density of R = q/4R2
(total charge divided by surface area of the sphere). The charge must be
entirely on the surface of the sphere, and is distributed uniformly due to
symmetry.
This situation is an instance of a sphere (the concentric shell) containing
a cavity, except that the charge inside the cavity is the inner sphere rather
than a point charge. However, outside the inner sphere, the field behaves
like a point charge of size +q, so there is a negative charge of q induced
on the inner surface of the shell. Because of the symmetry, we can now say
that this charge is distributed uniformly over the inner surface, so the charge
density is a = q/4a2 . The charge density on the outer surface is then
b = q/4b2 .
If the potential is zero at infinity, we can find it at each point in the system
by integrating.
ELECTROSTATICS - CONDUCTORS 4

b
(1) V (b) = E dl

b
q
(2) = 2
dr
40 r
1 q
(3) =
40 b

Inside the shell, E = 0 so the potential remains constant until we reach the
inner surface. Then we go from r = a to r = R:

R
q
(4) V (R) = V (b) 2
dr
a 40 r
 
q 1 1 1
(5) = +
40 b R a
The potential then remains constant inside the inner sphere.
If we then ground the outer surface of the shell so that its potential is
V (b) = 0, we simply lose the 1/b term in the second answer:
 
q 1 1
(6) Vgrounded (R) =
40 R a
Example 2. We have a solid metal sphere of radius R which contains two
non-overlapping spherical cavities, one of radius a and the other of radius b.
Inside the first cavity is a point charge qa and in the other is a charge qb . By
the principle of superposition, we can say that the induced surface charge
density in cavity a is a = qa /4a2 and that in cavity b is b = qb /4b2 .
The induced charge on the outer surface of the sphere is thus qa + qb so the
surface density is R = (qa + qb )/4R2 .
The field outside the sphere is therefore just

qa + qb
(7) Eout =
40 r2

where r is measured from the centre of the sphere. E points radially out-
ward. In other words, the external field is equivalent to that from a combined
point charge of qa + qb .
The field within each cavity is due entirely to the point charge within that
cavity, since the field from the charge in the other cavity is sheilded by the
intervening conductor. Thus within cavity a
ELECTROSTATICS - CONDUCTORS 5

qa
(8) Ea =
40 ra2

where ra is measured from the centre of cavity a. Same for cavity b (replace
subscript a by b).
Since each point charge is situated within spherical shells of induced
charge (one on the inner surface of the cavity and one on the outer sur-
face of the sphere) the field felt from both these shells is zero, so the force
is zero.
If we now placed a third charge outside the sphere, what would happen?
The induced charges on the inner surfaces of the cavities wouldnt change.
The charge distribution on the surface of the sphere would change, but it
would do so in such a way as to retain a zero field inside the conductor.
Thus there is still no force on either of qa or qb .
P INGBACKS
Pingback: Electrostatic pressure
Pingback: Capacitance
Pingback: Method of images
Pingback: Method of images - point charge and sphere
Pingback: Greens reciprocity theorem
Pingback: Dielectric examples
Pingback: Dipole between two angled conducting planes
Pingback: Decay time for free charge in a conductor
Pingback: Wave guides: derivation of the wave equation
Pingback: Radiation from a point charge near a conducting plane
ELECTROSTATIC PRESSURE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 2.5.3, Problems 2.37 - 2.38.
If an object carries a surface charge distribution, the electric field (due to
the charge distribution itself or any other external field) will exert a force on
the surface charge. If this charge is constrained to lie on the surface of the
object (for example, if the object is a conductor, so that all excess charge
lies on the surface), then the electric field will create a pressure (which is
force per unit area) by its action on this surface charge.
The question is: what is the force on a surface charge distribution? This
is a somewhat tricky question, since weve seen that the component of the
electric field that is normal to the surface is discontinuous, with a difference
of /0 from one side of the surface to the other (where is the surface
charge density).
To answer this question, we can consider some arbitrary surface (not nec-
essarily a conductor) which has surface charge distributed over it. We can
also consider a small patch on this surface and examine the fields acting on
it. Weve seen (Example 1 here) that, for an infinite plane of charge, the
field is /20 on each side of the plane, pointing away from the plane (for
positive charge) on both sides. Now if were considering a small patch of a
surface, thats clearly not an infinite plane, but if we also consider the field
just above (or below) this patch, then we can say that were considering the
field at a distance from the patch that is very small compared to the dimen-
sions of the patch. In such a case we expect the field to become arbitrarily
close to that for an infinite plane.
For example, if we consider the case of a circular disk of charge which
we solved earlier (Example 1 here), the field at location z on the axis of the
disk is

 
2
z +R z 2
1
(1) E= 2
40 z2 + R2
 
z2 + R2 z

(2) =
20 z2 + R2
1
ELECTROSTATIC PRESSURE 2

In the limit z  R (where R is the diameter of the disk), we see that this ex-
pression tends to /20 . Similar limits for other geometries always produce
the same result.
So we can split the electric field in the area of the patch into two contri-
butions. The first is due to the patch itself, and is /20 pointing normal to
the surface on both sides, and some other field E1 which could be anything,
depending on the geometry of the surface and other fields in the vicinity.
That is, we can say that the fields above and below the patch are


(3) Eabove = E1 + n
20

(4) Ebelow = E1 n
20
We can solve these two equations to find the other field

1
(5) E1 = (Eabove + Ebelow )
2
That is, if we know the field on either side of the patch, we can find
the field acting on the patch, and it turns out to be just the average of the
field on either side of the patch (that is, on either side of the discontinuity).
Furthermore, we can say that E1 is the only field that acts on the patch,
since a charges field doesnt act on the charge itself. (OK, this argument is
a bit dodgy, since were not considering a point charge, but rather a small
patch, so that technically, yes, the field produced by one part of the patch
does act on other parts of the same patch. However, were juggling two
limits here: in the first place were assuming that the size of the patch is
small enough that we can consider its field to be almost that due to a point
charge so that the field doesnt act back on the patch itself. In the second
place, were considering that the distance above the patch is small relative
to the size of the patch so that we say the field due to the patch is /20 . So
were essentially nesting one infinitesimal inside another.)
The argument so far is valid for any surface charge. In the special case
of a conductor, we know the fields above and below the surface. Below
the surface (that is, inside the conductor) we know that Ebelow = 0, and just
above the surface we know that Eabove = 0 n (the field at the surface of a
conductor is always normal to the surface). In this case, the field that acts
on the surface charge of a conductor is


(6) E1 = n
20
ELECTROSTATIC PRESSURE 3

and the force per unit area is then just the field times the surface charge
density:

2
(7) F= n
20
Since force per unit area is pressure, we can define the electrostatic pres-
sure that the charge exerts on the surface of a conductor as the magnitude
of the force per unit area

2
(8) P =
20
0 2
(9) = E
2 above

where Eabove is the electric field on the outer surface of the conductor:


(10) Eabove =
0
A couple of examples of this pressure:
Example 1. Two large metal plates, each of area A are held a distance d
apart. If there is a charge Q on each plate, then the field due to each plate
is E = /20 , with = Q/A, pointing away from the plate on each side.
Between the plates, E = 0, while outside the plates, E = /0 pointing
away from the plates. The electrostatic pressure is therefore

0 2
(11) P = E
2
0 2
(12) =
2 02
Q2
(13) =
20 A2
Note that the answer does not depend on d, provided that the linear di-
mensions of the plates are much larger than their separation, so that we can
regard them as infinite planes.
Example 2. A spherical conductor of radius R carries a total charge Q.
We can find the force of repulsion between two hemispheres of the sphere.
From symmetry, the electrostatic pressure is 20 E 2 in a radial direction. To
find the total force between two hemispheres, we can integrate the z com-
ponent of the force per unit area (that is, the pressure) over one hemisphere
ELECTROSTATIC PRESSURE 4

to find the total force on it. The electric field is zero inside the sphere, and
Q
4 R2
just outside the sphere. Thus
0


0 2 Q2 /2
(14) F = 2R cos sin d
2 (40 R2 )2 0
Q2
(15) =
320 R2
P INGBACKS
Pingback: Conservation of energy in a capacitor
Pingback: Method of images
Pingback: Method of images - point charge and sphere
Pingback: Force on a dielectric
CAPACITANCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec. 2.5.4 & Problem 2.39
The electric potential is defined as


1 1
(1) V (r) (r0 )dr0
40 |r r0 |
One of the properties of a conductor is that its surface is an equipotential;
that is, the potential is the same everywhere on the surface of (and inside) a
conductor. Although the calculation of the potential from the above formula
for some geometric shape of conductor could be very difficult, we can see
that if we change the charge density by a constant factor everywhere, the
potential will change by the same factor. That is, the potential is effectively
proportional to the total charge.
The absolute value of the potential depends on the reference location we
use; as weve seen before, its traditional to choose V = 0 at infinity. How-
ever, the potential difference between two points does not depend on this
reference location (since it cancels out when taking the difference). Since
the potential itself is proportional to the total charge, so too is the potential
difference. Thus if we arrange two conductors in some configuration, then
there is a definite potential difference which is the same from any point on
one conductor to any point on the other conductor.
As such, it makes sense to define a quantity called the capacitance which
is the ratio of the charge on the conductors to the potential difference be-
tween them:

Q
(2) C
V

where Q is the amount of positive charge on one conductor (and Q is the


amount of negative charge on the other conductor), and V is the potential
difference between the two conductors, taken as (positive) minus (negative),
so V is always positive. Thus C is always a positive quantity, and in SI units
its unit is the farad, or coulomb per volt.
1
CAPACITANCE 2

The name capacitance can be thought of as the capacity of a system of


two conductors for holding charge. The larger the capacitance, the more
charge is required to produce a given potential difference. To calculate the
capacitance we need to find an expression for the potential in terms of the
amount of charge stored on the conductors.
Example 1. The parallel plate capacitor. If we have two parallel flat
plates, each of area A and a distance d apart, what is the capacitance of
the system? If the area is large compared to the separation, then we can
approximate the field between the plates by /0 , where is the surface
charge density: = Q/A. Since the field is therefore constant, the potential
is found from an integral.


(3) V = E dl
Q
(4) = d
A0

The capacitance is then

Q
(5) C =
V
A0
(6) =
d

Thus the capacitance increases if we increase the area of the plates, or if


we decrease the distance between them.
Example 2. If we now have two concentric conducting spheres with radii
a and b (a < b), and place a charge +Q on the inner sphere, we note that
the field between the spheres is due entirely to the inner sphere, and is

1 Q
(7) E= r
40 r2

where a < r < b. The potential difference between the spheres is then found
from
CAPACITANCE 3


(8) V = E dl
b
Q dr
(9) =
40 a r2
 
Q 1 1
(10) =
40 a b
Q ba
(11) =
40 ab
The capacitance is

Q
(12) C =
V
ab
(13) = 40
ba
Note the rather curious fact that this calculation doesnt depend at all on
how much charge is on the outer sphere, although it does depend on where
the outer sphere is (that is, on b). In fact, if the outer sphere goes off to
infinity, the capacitance tends to

C 40 a
Example 3. We have two concentric cylindrical conducting shells or radii
a and b. Find the capacitance per unit length. First, we can find the field
due to the inner cylinder. By symmetry, the field points radially outwards,
so we can use Gausss law to find it. If the radius of the cylinder is a and we
consider a length L then if we consider a Gaussian cylinder enclosing this
section, the total enclosed charge is 2aL . If the Gaussian cylinder has
radius r its area (minus the end caps, which dont contribute) is 2rL so


q
(14) E da =
0
2aL
(15) 2rLE =
0
a
(16) E =
r0
The potential difference from R = a to R = b is therefore
CAPACITANCE 4


a b 1
(17) V = dr
0 a r
a b
(18) = ln
0 a

In a unit length, the charge is Q = 2a , so

Q b
(19) V= ln
20 a

and

Q
(20) C =
V
20
(21) =
ln(b/a)
P INGBACKS
Pingback: Coaxial cable with dielectric
Pingback: Force on a dielectric
Pingback: Transmission lines
Pingback: Momentum in a capacitor
Pingback: Transforming the electric field in relativity
CONSERVATION OF ENERGY IN A CAPACITOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 2.40.
The electrostatic pressure (force per unit area) on the surface of a con-
ductor is

0 2
(1) P= E
2

where E is the electric field just outside the conductor (E = 0 inside a con-
ductor).
Suppose we have a flat parallel plate capacitor with plate area A carrying
an amount of charge so that the field between the plates is E. Each plate
experiences a pressure as above, so the total force on each plate is

0 2
(2) F = PA = E A
2
If the plates move an infinitesimal distance towards each other under
the influence of this force, the work done is

(3) W = PA
0 2
(4) = AE
2
The energy density in an electric field is

0 2
(5) u= E
2

so if the plates contract by a distance then the amount of energy lost is

0 2
(6) U = E A
2

so the energy from the electric field is used to move the plates closer to-
gether.
1
CONSERVATION OF ENERGY IN A CAPACITOR 2

We can invert the argument by starting with 6 and arguing that because
of conservation of energy, the energy lost in the field must provide the work
done in moving the plates, from which we can get the force on the plates as

U 0
(7) F= = E 2A
2

and the pressure as

F 0
(8) P= = E2
A 2
ELECTRIC FIELD & POTENTIAL - MORE EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 2.41, 2.42, 2.43, 2.44.
Here are a few more examples of the calculation of electric field and
potential.
Example 1. Given a square sheet of charge with side length a and surface
charge density , find the electric field at height z above the centre of the
sheet.
From Example 3 in this post, the field at distance z above the centre of a
square loop of charge of side length s and linear charge density is

1 32s z
(1) E=
40 2s2 + 4z2 (s2 + 4z2 )

If we think of this square loop as an element of the square sheet, where


the thickness of the loop is ds/2 (so that the overall square has side length
s + ds) then = ds/2 and the field of the sheet is

a
16 z s
(2) E= ds
40 0 2s + 4z2 (s2 + 4z2 )
2

The mathematical software (Maple, in this case) needs a bit of help with
this integral, so we can use the substitution u = s2 ; du = 2sds to get

a2
8 z du
(3) E=
40 0 2u + 4z2 (u + 4z2 )
" ! #
1 2a2 + 4z2
(4) = 8 arctan 2
40 2z
2
q

(5) = arctan 1 + a2 /2z2
0 20

The field points vertically upwards, by symmetry.


Example 2. If the electric field is given by
1
ELECTRIC FIELD & POTENTIAL - MORE EXAMPLES 2

1 
(6) Ar + B sin cos
r

(A and B are constants), find the charge density.


In this case, we can use the differential form of Gausss law: E =
/0 . In spherical coordinates, the divergence for a vector field with its
component equal to zero is


(7) = E
0
1 2  1 E
(8) = 2 r Er +
r r r sin
A B sin
(9) = 2
r r2
0
(10) = 2 (A B sin )
r
Example 3. In a uniformly charged solid sphere, find the force of repul-
sion between two hemispheres.
To make the problem definite, well consider the north and south
hemispheres, so the problem becomes one of finding the vertical force be-
tween these two hemispheres.
From Example 2 in a previous post, the electric field inside a uniformly
charged sphere with density is

r
(11) E=
30
The vertical force on a volume element is therefore
 
3 r
(12) dFz = d r cos
30
So the total repulsive force on a hemisphere is


2 2 R /2 3
(13) Fz = r cos sin d dr
30 0 0
2 R4
(14) =
30 4
Using
ELECTRIC FIELD & POTENTIAL - MORE EXAMPLES 3

3Q
(15) =
4R3

we get

1 3Q2
(16) Fz =
40 16R2
Note that we cant just find the repulsive force between two hemispheres
of a spherical shell and then integrate over shells of sizes from zero up to
the radius of the sphere because this doesnt take into account the force
between hemispheres of different sizes.
Problem 4. A northern hemispherical shell of radius R has a uniform
surface charge density of . Find the potential difference between the north
pole and the centre of the base of the hemisphere.
For an inverted hemispherical bowl, we can use the method of Example
5 in this previous post, except the limits on the integral are now 0 to /2.
We therefore get
 
2R p 2 2
q
(17) 40V (z) = R + z (R z)2
z
At the north pole, z = R so

(18) 40V (R) = 2 2 R
At the centre, z = 0 so we need to take a limit. We can rewrite the formula
as

2R2
q 
(19) 40V (z) = 2 2
1 + z /R 1 + z/R
z
The lowest order term in the expansion of the square root is

1 z2
q
(20) 1 + z2 /R2 = 1+
2 R2
so the limit as z 0 is

2R2 z
(21) 40V (0) = lim
z0 z R
(22) = 2 R
ELECTRIC FIELD & POTENTIAL - MORE EXAMPLES 4

and the potential difference is

2  
(23) V (R) V (0) = R 21
40
R  
(24) = 21
20
DIRAC DELTA FUNCTION IN THREE DIMENSIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec. 1.5.3, Problem 2.46.
Shankar, R. (1994), Principles of Quantum Mechanics, Plenum Press.
Chapter 12, Exercise 12.6.4.
[If some equations are too small to read easily, use your browsers mag-
nifying option (Ctrl + on Chrome, probably something similar on other
browsers).]
Weve seen that we can define a curious function called the Dirac delta
function in one dimension. Here we examine how this can be extended to
three dimensions, and how this extension is relevant to electrostatics.
The easiest way to define a three-dimensional delta function is just to
take the product of three one-dimensional functions:

(1) 3 (r) (x) (y) (z)


The integral of this function over any volume containing the origin is
again 1, and the integral of any function of r is a simple extension of the
one-dimensional case:

(2) f (r)3 (r a)d 3 r = f (a)

In electrostatics, there is one situation where the delta function is needed


to explain an apparent inconsistency involving the divergence theorem. If
we have a point charge q at the origin, the electric field of that charge is

1 q
(3) E= r
40 r2
According to the divergence theorem, the surface integral of the field is
equal to the volume integral of the divergence of that field:

(4) E da = Ed 3 r
V

where the integral on the left is over some closed surface, and that on the
1
DIRAC DELTA FUNCTION IN THREE DIMENSIONS 2

right is over the volume enclosed by the surface. In electrostatics, the in-
tegral on the right evaluates to the total charge contained in the volume
divided by 0

q
(5) Ed 3 r =
V 0
Now for the catch. If we calculate E (in spherical coordinates) for
the point charge, we get, since only the radial component of the field is
non-zero:

1 2 
(6) E = r Er
r2 r
q 1 (1)
(7) =
40 r2 r
At this stage, we might be tempted to say that the derivative is zero (since
the derivative of any constant is zero), but the problem is that at r = 0 we
also have a zero in the denominator, so we have the indeterminate fraction
of zero-over-zero. Thus although it is true that E = 0everywhere except
the origin, we know from the divergence theorem that V Ed 3 r = q0 so
we must have
 
1
(8) 2 r d 3 r = 4
V r
and
 
1
(9) 2 r = 0 if r 6= 0
r
These two conditions can be satisfied if
 
1
(10) 2 r = 43 (r)
r
Furthermore, since 1r = r12 r, we have

1
(11) 2 = 43 (r)
r
Example. Suppose we have some distribution of charge that gives a po-
tential function
DIRAC DELTA FUNCTION IN THREE DIMENSIONS 3

e r
(12) V (r) = A
r
We can find the field by taking the gradient

(13) E = V
e r
(14) = A 2 (1 + r) r
r
We can now find the charge distribution by taking the divergence, re-
membering what weve discussed above. Applying the divergence formula
in spherical coordinates directly gives

!
2 e r
(15) = 0 A
r

but this formula is valid only for r 6= 0. To get the full charge distribution
we need to incorporate the delta function. Using the product rule for the
divergence ( ( f A) = f A + A f ):

 
r r
(16) E = 2 Ae (1 + r)
r
 
r 1 r  
(17) = Ae (1 + r) 2 r + 2 Ae r (1 + r)
r r
!
2 e r
(18) = Ae r (1 + r) (43 (r)) A
r
!
2 e r
(19) = 4A3 (r) A
r
" #
2 e r
(20) = A0 43 (r)
r

In the fourth line, we used the fact that f (r)3 (r) = f (0)3 (r), since the
delta function is zero everywhere except at r = 0.
From this, we can find the total net charge by integrating :
DIRAC DELTA FUNCTION IN THREE DIMENSIONS 4


(21) Q = d 3 r
V

3 2 e r 3
(22) = A0 43 (r)d r A0 d r
V V r
r
e
(23) = 4A0 4A0 2 r2 dr
0 r
2
 
(24) = 4A0 1 2

(25) = 0
That is, the delta function contributes a point charge of +4A0 at the
origin, and the second term contributes a continuous charge distribution
smeared out over all space that sums up to 4A0 .
POTENTIAL OF TWO CHARGED WIRES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 2.47.
Two infinite wires lie in the x-y plane, parallel to the x axis. One carries
a charge density of + and lies at location y = +a while the other carries
a charge density of and lies at location y = a. Find the potential at a
location (x, y, z) in rectangular coordinates.
The field due to an infinite wire can be found using Gausss law in cylin-
drical coordinates. For the wire carrying charge density we have, using
a Gaussian cylinder of unit length centred on the wire (see Example 3 in
this post)


(1) 2r E =
0

where r is the cylindrical distance from the wire, and for the wire with
charge density +


(2) 2r+ E+ =
0
The potentials from the two wires add (according to the superposition
principle), so we get

(3) V = E dl
s
dr
(4) =
20 a r
s
(5) = ln
20 a

where the limits on the integral arise from taking the origin as the zero
point for potential. The distance s is the cylindrical distance from the wire
at which we want the potential.
Similarly, the potential for the positive wire is
1
POTENTIAL OF TWO CHARGED WIRES 2

s+
(6) V+ = ln
20 a

so the total potential is

h s s+ i
(7) V= ln ln
20 a a
s
(8) = ln
20 s+
In terms of (x, y, z) we have
q
(9) s = (y + a)2 + z2
q
(10) s+ = (y a)2 + z2

so we get
p
(y + a)2 + z2
(11) V= ln p
20 (y a)2 + z2
(y + a)2 + z2
(12) = ln
40 (y a)2 + z2
We can find the equipotential surfaces, that is, the surfaces where V = K
for some constant K. First, note that if y = 0 then V = 0, so the xz plane
is the equipotential surface for V = 0. To find the other surfaces, we can
consider V = K 6= 0 so we have

40 K (y + a)2 + z2
(13) = ln
(y a)2 + z2
(y + a)2 + z2
(14) e40 K/ =
(y a)2 + z2
We can now define

(15) A e40 K/

so we get
POTENTIAL OF TWO CHARGED WIRES 3

A (y a)2 + z2 = (y + a)2 + z2

(16)
A y2 2ay + a2 + z2 = y2 + 2ay + a2 + z2

(17)
(18)
(A 1)y2 (A + 1)2ay + (A 1)a2 + (A 1)z2 = 0
A+1
(19) y2 2ay + a2 + z2 = 0
A1
A+1 2 2 A+1 2
   
(20) y a +z = a a2
A1 A1
 2 2 
2 A + 2A + 1 A + 2A 1
(21) =a
(A 1)2
4a2 A
(22) =
(A 1)2

The fourth line is obtained by dividing through by (A1), which requires


A 6= 1, or K 6= 0. We saw above that K = 0 is a special case, giving the xz
plane as a surface.
This equation has the form of a circle inqthe yz plane, with centre at y =
A+1 4a2 A
A1 a and z = 0, and with a radius of R = (A1)2
. Thus the equipotential
surfaces are circular cylinders, with axes given by the lines running through
the centres of the circles. Note that since A was defined as an exponential,
it is always positive, so the argument of the square root in the equation for
the radius is always positive.
We can revert back to expressions containing K to see the relation be-
tween the surfaces and the potentials.
The radius becomes

2ae20 K/
(23) R=
e40 K/ 1
2a
(24) = 2 K/
e 0 e20 K/
a
(25) =
sinh (20 K/ )

The axis is (with z = 0):


POTENTIAL OF TWO CHARGED WIRES 4

e40 K/ + 1
(26) y= a
e40 K/ 1
e20 K/ + e20 K/
(27) = 2 K/ a
e 0 e20 K/
cosh (20 K/ )
(28) = a
sinh (20 K/ )
a
(29) =
tanh (20 K/ )
P INGBACKS
Pingback: Potential of two copper pipes
Pingback: Two charged wires and a cylinder
VACUUM DIODE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 2.48.
A vacuum diode is an electronic component consisting of two parallel
plates called the cathode, maintained at potential zero, and the anode, at
potential V0 . If the cathode is heated, the excited atoms within the plate emit
electrons which are accelerated across the gap between the plates. After
a short while, a steady state is reached in which a constant current I is
maintained across the gap. If we assume that the area A of the plates is much
greater than the square of the distance d between them, we can neglect edge
effects and assume that all quantities are functions only of x, the location
between the plates (with x = 0 at the cathode and x = d at the anode). The
relevant functions are the potential V (x), the charge density between the
plates (x), the speed of the electrons v(x) and the current I, which is a
constant independent of x once a steady state has been reached. In addition,
we can assume that once this steady state has been reached, the electron
density at the cathode is such that the electric field is zero there (although
its gradient is not).
From Gausss law, we get Poissons equation between the plates


(1) E =
0
(2) E = V

(3) 2V =
0

Since these functions depend only on x, we get

d 2V
(4) =
dx2 0

Since the force on an electron is qE (remember q is negative for an elec-


tron), we can find the speed of an electron from the work done on it, which
is translated into kinetic energy.
1
VACUUM DIODE 2

x
1 2 dV 0
(5) mv = q 0
dx
2 0 dx
(6) = qV (x)
r
2qV (x)
(7) v(x) =
m
Next, we can find the current I in terms of and v. The current is the
rate of flow of charge. Consider a thin slice (thickness dx) of the space
between the plates. The volume of this slice is Adx so the amount of charge
contained in this slice is Adx. If the charge in this slice takes time dt to
pass a given point, then the current is

dx
(8) I = A
dt
(9) = A(x)v(x)

so the charge density can be written as

I
(10) (x) =
Av(x)
r
I m
(11) =
A 2qV (x)
Note that since the charge density arises from electrons, it is negative, so
the current is also negative (current is usually defined as the flow of positive
charge, so when we have a flow of electrons, the current is actually in the
opposite direction to the flow of electrons).
Returning to Poissons equation, we can now write a differential equation
for the potential.

d 2V
(12) =
dx2 0
r
I m
(13) = V 1/2
A0 2q
(14) V 1/2V 00 = C
r
I m
(15) C
A0 2q
Again, note that C is positive and real, since both I and q are negative.
We can solve this by guessing a solution of form
VACUUM DIODE 3

(16) V = Bxk

where B and k are to be determined. Note that this form satisfies the bound-
ary conditions V (0) = V 0 (0) = 0, which we require from above (the poten-
tial at the cathode is zero, and at steady state, the electric field (= V 0 (0))
is also zero). We just have to hope that it also provides a solution to the
differential equation.
From this we get

(17) V 0 = kBxk1
(18) V 00 = k(k 1)Bxk2
(19) B1/2 xk/2 k(k 1)Bxk2 = C
(20) B3/2 k(k 1)x3k/22 = C
Since the RHS is a constant, the LHS must have the same value for all
values of x, so the exponent of x must be zero. That is,

3k
(21) 2 = 0
2
4
(22) k =
3

This allows us to find B:

4 3/2
(23) B = C
9
 2/3
9
(24) B = C
4
So we get

 2/3
9
(25) V (x) = C x4/3
4
m 2/3 4/3
 r 
9I
(26) = x
4A0 2q
1/3
81I 2 m

(27) = x4/3
32A2 02 q
VACUUM DIODE 4

In particular, the potential at the anode is

1/3
81I 2 m

(28) V0 = V (d) = d 4/3
32A2 02 q
 x 4/3
(29) V (x) = V0
d
Note that this isnt linear, as it would be if there were no charge between
the plates.
From this, we can write the current in terms of the anodes potential
 1/2
4A0 2q 3/2
(30) I= 2 V0
9d m

(weve taken the negative root, since I is negative). This relation between
current and potential difference in a diode is known as the Child-Langmuir
law.
The charge density
r
I m
(31) (x) =
A 2qV (x)
40V0
(32) = 4/3 2/3
9d x
Finally, the speed of the electrons is
r
2qV (x)
(33) v(x) =
m
r
2qV0  x 2/3
(34) =
m d
DEVIATION FROM COULOMBS LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 2.49.
An interesting alternative (and fictional) world scenario for electrostatics
arises if we alter the spatial dependence of Coulombs law so that the force
between two charges is

|r r0 | |rr0 |/
 
1 q1 q2
r r0

(1) F= 0 3
1+ e
40 |r r |

where is some (very large) length. We can see that if this formula
reduces to the normal Coulomb force law. If we assume that the principle
of superposition still holds, we can derive a few analogs to the electrostatic
equations we all know and love.
First, the electric field due to a point charge q at position r0 is

|r r0 | |rr0 |/
 
1 q 0

(2) E(r) = 1 + e r r
40 |r r0 |3

from which we can get the field due to a charge distribution



(r0 ) |r r0 | |rr0 |/
 
1 0
 3 0
(3) E(r) = 1 + e r r d r
40 V |r r0 |3
Although this looks ugly, we notice that a potential function exists for
this field by using the following argument. First we consider a point charge
at the origin, for which

1 q r  r/
(4) E(r) = 1 + e r
40 r2
Since the field depends only on r and is radial in direction, we can write,
in spherical coordinates

1 q r  r/
(5) E dl = 1 + e dr
40 r2
1
DEVIATION FROM COULOMBS LAW 2

Therefore, the line integral of the field between two points a and b is

!
b
q b
er/ er/
(6) E dl = + dr
a 40 a r2 r
!
q era / erb /
(7) =
40 ra rb

where ra is the distance from the origin to point a (similarly for rb ). Clearly
if the path is closed, then ra = rb and the integral is zero, so by Stokess
theorem, E = 0, which is the same result as in normal electrostat-
ics. Also, the integral above must be path independent, since otherwise we
could choose a closed loop where the integrals along two branches of the
loop were different, and thus the integral over the whole loop was non-zero,
contradicting what weve just said. This means that, just as with normal
electrostatics, we can define a potential relative to some reference point x
as

r
(8) V (r) E dl
x

If we pick infinity as the reference point, then we get, for a point charge
at the origin:

q er/
(9) V (r) =
40 r

Also for a point charge at the origin, we can integrate the field over a
sphere of radius R centred at the origin:


4R2 q
 
R R/
(10) E da = 1+ e
R 40 R2
 
q R R/
(11) = 1+ e
0

This isnt quite as nice as Gausss law because of the extra R-dependent
factors, but if we work out the volume integral of the potential over the same
sphere:
DEVIATION FROM COULOMBS LAW 3

R r/
3 q e
(12) Vd r = r2 dr
40 0 r
q 2
  
R/ R
(13) = 1e 1+
0
From these two results, we see that

1 q
(14) E da + 2 V d3r =
R 0

Remember that the first integral is a surface integral and the second inte-
gral is a volume integral. This is a sort-of Gausss law for the fictional
electrostatics.
In fact, this argument applies to any shape of enclosing volume. Suppose
we take an infinitesimal patch on the sphere, defined by angle increments d
and d . Then for that patch we have the increment in the surface integral:
   
1 q R R/ 2
(15) E da = 1+ e R sin d d
40 R2
 
q R R/
(16) = 1+ e sin d d
40
For the volume integral extending from the centre of the sphere up to the
patch we need integrate over r only to get the increment in this integral

  "
R r/
#
1 1 q e
(17) V d3r = 2 r2 dr sin d d
2 40 0 r
  
q R/ R
(18) = 1e 1+ sin d d
40
The sum of these two terms comes out to
   
1 3 q
(19) E da + V d r = sin d d
2 40
That is, the dependence on the radius of the sphere cancels out for each
incremental patch. This means we can consider any shape composed of in-
cremental patches, where the radius of each patch can be different, which
amounts to saying that we can consider an enclosing surface of any shape
DEVIATION FROM COULOMBS LAW 4

we like, and the same result holds. Further, by the principle of superposi-
tion, we can combine the results from any collection of charges so that we
can write

1 Q
(20) E da + 2 V d3r =
R 0

where the first integral is over any enclosing surface and the second inte-
gral is over the enclosed volume, and Q is the total charge enclosed by the
surface.
Although this result appears reasonably pretty, it does depend on the
specific nature of the field equation (that is, the pair of extra factors that
have been inserted). If we considered a different deviation from the classic
Coulombs law, these results wouldnt follow. But then, theres no evidence
that there is any deviation from Coulombs law anyway, so the whole exer-
cise is just for fun.
LAPLACES EQUATION - AVERAGE VALUES OF SOLUTIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 3.1, Problems 2.50, 3.1.
Weve seen that the electric field obeys Gausss law, which in differential
form is


(1) E =
0
The field can also be written as the gradient of a potential function, so we
get

(2) E = V

(3) 2V =
0
The last equation is a partial differential equation (PDE) known as Pois-
sons equation, and its solution gives the potential for a given charge distri-
bution.
In regions where there is no charge, = 0 and Poissons equation be-
comes Laplaces equation:

(4) 2V = 0
One of the key points in the solution of any PDE is the specification of
boundary conditions. Without boundary conditions, the problem is incom-
pletely specified, and some surprising results can occur. For example, if we
specify an electric field by the conditions (where a is a constant):

(5) Ex = ax
(6) Ey = Ez = 0

then we can verify by direct calculation that E = 0 as required in elec-


trostatics, so it looks like this is an acceptable field. However if we calculate
we get
1
LAPLACES EQUATION - AVERAGE VALUES OF SOLUTIONS 2

(7) E = a
(8) = a0

That is, the charge density is constant over all space. We get a paradoxical
situation where the charge is uniform and yet the electric field has a spe-
cific direction. Clearly wed get similar results if we specified a field by
equations like Ex = Ez = 0; Ey = by or Ex = Ey = 0; Ez = cz. Poissons
equation for the potential becomes, in this case

(9) 2V = a

There are one obvious solution to this:

a
(10) V = x2
2

which returns the original electric field when the gradient is taken, but again
seems paradoxical since the potential for a uniform charge distribution has
a dependence on x.
The problem is that although the div and curl equations are necessary
for an electric field, they are not sufficient to determine the problem were
trying to solve. We need to specify the region of space in which the field
has the given form, since the equations a given result in an infinite field as
x . Weve solved field and potential problems earlier in which there is
a uniform charge density, but it is restricted to a certain volume, such as a
sphere.
But getting back to Laplaces equation, there are a couple of important
properties possessed by all solutions to the equation. Proving these things
rigorously takes a bit of heavy mathematics, but we can get a feel for these
properties by some relatively simple calculations. The first property is that
if we have a solution of Laplaces equation in three dimensions, then if we
consider the value of the solution at a given point r, then if we calculate the
average value of the solution over any sphere centred at r, the value of this
average is equal to the value of the solution at r. That is if we integrate V
over the surface of a sphere of radius R centred at r, we must have

1
(11) V da = V (r)
4R2 R
LAPLACES EQUATION - AVERAGE VALUES OF SOLUTIONS 3

This is the main result which requires a bit of heavy-duty math to prove in
general, but once we have established this fact, the second property of solu-
tions to Laplaces equation follows quite easily: all extreme values (maxima
and minima) of a solution must occur on the boundaries of the region under
consideration. This is fairly obvious, since if an extremum occurred at some
interior point r0 then we could find some sphere around this point where the
values of V are all greater than (for a minimum) or less than (for a maxi-
mum) the value at r0 , which isnt allowed according to the first property.
We can demonstrate this first property for the special case of the potential
due to a point charge. For regions not containing the point charge, Laplaces
equation is satisfied (since there is no charge there), and weve seen earlier
that the potential due to a point charge q at location r0 is

1 q
(12) V (r) =
40 |r r0 |
Thus we know that V (r) is a solution to Laplaces equation everywhere
away from the point charge. We can therefore try to show the averaging
property for this particular solution.
To make things definite, suppose the charge is located at position z on the
z axis, and we calculate the average value of V over a sphere of radius R
centred at the origin. Well make z > R to ensure that the charge is outside
the sphere, so Laplaces equation is valid everywhere within and on the
surface of the sphere.
From the cosine law, a point on the surface of the sphere will satisfy
p
(13) |r r0 | = z2 + R2 2Rz cos

where is the angle between the z axis and the vector pointing to the point
on the sphere. Using spherical coordinates, we can calculate the average of
V:
2
1 1 q R2 sin
(14) V da = d d
4R2 R 4R2 40 0 0 z2 + R2 2Rz cos
 q 
1 q 1
q
2 2
(15) = (z + R) (z R)
8 0 zR
1 q
(16) =
40 z
It isqimportant to note that in going from the second to the third line, we
used (z R)2 = + (z R), since z > R.
LAPLACES EQUATION - AVERAGE VALUES OF SOLUTIONS 4

The final result is just the potential due to the charge at the origin, that is,
at the centre of the sphere, so in this case, the average of the potential over
the sphere is equal to its value at the centre.
This result can be generalized by using the superposition principle to
show that the average, over a sphere in a region where Laplaces equation
is satisfied, of a potential due to any distribution of charge is equal to the
potential at the centre of the sphere. Thus weve established the averag-
ing property of Laplaces equation for electrostatic potential, but this isnt,
of course, a general demonstration that all solutions of Laplaces equation
satisfy the property.
We can note that the averaging formula can also be generalized to include
charge that is inside the sphere. In the above derivation, if we consider a
q the sphere, then z < R and we need to take the other root
charge that is inside
in the last step: (z R)2 = (z R). This gives the average potential as

1 q
(17) V da =
4R2 R 40 R

That is, the radius R has replaced z in the denominator. Using the superpo-
sition principle again, we get a general formula for the average potential

Qin
(18) V = Vout +
40 R

where Vout is the average of the potential due to all charge outside the sphere
(or, what is the same thing, the potential at the centre of the sphere due to
all charge outside the sphere), and Qin is the total charge enclosed by the
sphere.
P INGBACKS
Pingback: Laplace & Poisson equations - uniqueness of solutions
METHOD OF IMAGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 3.2, Problem 3.6.
Weve seen that, in electrostatics, Laplaces equation 2V = 0 governs
the potential in those regions where there is no charge. Laplaces equation
is a special case of Poissons equation:


(1) 2V =
0

where V is the potential and is the volume charge density. Weve also
seen that, given a particular geometry of bounding surfaces and boundary
conditions specified on those surfaces, the solution of Laplaces equation is
unique. This allows a clever trick to be used in some situations. The trick is
known as the method of images. There are two standard problems that are
usually given in textbooks to illustrate the method of images; well have a
look at the first one here.
The defining problem for the method of images is the point charge +q
and the infinite, grounded conducting plane. To make things definite, well
suppose the conducting plane occupies the xy plane, and the point charge is
at location z = d on the z axis. In this case, the boundary condition is that
V = 0 on the xy plane (a grounded conductor is always assumed to be at zero
potential). The problem here is that, since the plane is a conductor, charge is
free to move around on its surface in response to the electric field from the
point charge. It seems clear that this charge will have radial symmetry about
the z axis, but beyond that, its hard to tell precisely what the distribution
will be.
The trick is to notice that if we replace the conducting plane by another
point charge q at location z = d. With just two point charges, we can
write down the potential right away. In rectangular coordinates, we get
" #
q 1 1
(2) V= p p
40 x2 + y2 + (z d)2 x2 + y2 + (z + d)2
The clever bit is to notice that in the xy plane (at z = 0), V = 0. Thus
we have a potential that satisfies Laplaces equation (at least at every point
1
METHOD OF IMAGES 2

except where the two point charges are), and also satisfies the boundary
condition that V = 0 on the xy plane. Since we know that solutions to
Laplaces equation are unique, this must also be a solution of the point
charge-conducting plane problem, at least in the half space z > 0.
From the potential, we can also find the electric field from E = V .
We can also find the surface charge density on the plane. At the surface
of a conductor, we know that the charge density is given by the normal
derivative of the potential at the surface:

V
(3) = 0
n
In this case, the normal direction is in the z direction, so we get

V
(4) = 0
z z=0
qd
(5) = 3/2
2 (x2 + y2 + d 2 )
We can also calculate the total induced charge qi by integrating this over
the entire plane. This is easier to do in polar coordinates, where r2 = x2 + y2
and the increment of area is r dr d . We integrate to get

2qd rdr
(6) qi = 3/2
2 0 (r2 + d 2 )
(7) = q
That is, the total induced charge is equal to the image charge.
Since the force is calculated from the electric field, and the field is cal-
culated from the potential, the force between the point charge and the plane
must be equal to the force between the point charge and image charge. That
is

q2
(8) F =
40 (4d 2 )
The work done to set up the configuration isnt quite as straightforward,
since we cant calculate it directly from the two point charges. We can,
however, work it out from the integral

(9) W= F dl
METHOD OF IMAGES 3

We can choose a path from infinity on the z axis up to the location of the
point charge at z = d. The force used in calculating the work done is the
negative of the force between the charge and plane, since were opposing
the force, so we get

d
q2 dz
(10) W = 2
40 (4z )
1 q2
(11) =
40 4d

Notice this is half what we would get for the two point charges on their
1 q2
own, where W = 4 0 2d
. This isnt the same problem as that of finding
the work done in assembling two point charges, since as we bring in one
point charge, the location of the image charge changes to preserve the zero
potential in the conducting plane. Another way of looking at it is that al-
though we do work on the point charge in bringing it in from infinity, the
induced charge in the plane is distributed by moving charge around a lo-
cation where the potential is constant. Since the potential is constant, the
field within the plane is zero, so no work is actually done to rearrange the
induced charge. This does seem to be a bit of a fudge to me, though, since if
there were really no transverse fields operating within the conducting plane,
the charges within the plane wouldnt move at all, so I suspect there is some
work done in the process that the theory at this level simply ignores. Com-
ments welcome.
By using the superposition principle, we can apply the method of images
to any number of point charges above the conducting plane. For example,
suppose we have a grounded conducting plane in the xy plane, and a charge
of 2q is placed on the z axis at z = d, and a second charge of +q is placed at
z = 3d. We can use the method of images combined with the superposition
principle to find the force on the charge of +q.
The +q charge has an image of q at z = 3d and the 2q charge has
an image of +2q at z = d.
Since the method of images allows us to find the potential from the image
charges, the electric field (which is the negative gradient of the potential)
must be the same in the image and original problems. Since the force is
calculated from the field, it too is the same in the two problems.
Therefore, the force on +q is
METHOD OF IMAGES 4

q2
 
2 2 1
(12) F= 2+ z
40 4d 16d 2 36d 2
q2
 
1 1 1
(13) = + z
40 d 2 2 8 36
1 29q2
(14) = z
40 72d 2
P INGBACKS
Pingback: Method of images - point charge and sphere
Pingback: Method of images: two conducting planes
Pingback: Potential of two copper pipes
Pingback: Method of images - moving charge and plane
Pingback: Method of images: point charge and two planes
Pingback: Point charge embedded in dielectric plane
Pingback: Floating a magnet above a superconductor
Pingback: Radiation from a point charge near a conducting plane
Pingback: Laplaces equation - separation of variables
Pingback: Direct product of vector spaces: 2-dim examples
METHOD OF IMAGES - POINT CHARGE AND SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.7.
A special case in which the method of images works is that of a point
charge and a grounded, conducting sphere. Well take the sphere with its
centre at the origin and give it a radius R. The point charge q is on the z axis
at location z = a, where a > R (so its outside the sphere). Now we know
that the potential everywhere on the sphere is zero (since its a grounded
conductor), so as the point charge is brought in, charge moves around on
the conductor to keep the potential at that value. The problem is to find the
potential everywhere outside the sphere.
Someone at some point noticed that if you replace the sphere by an image
charge of q0 = Rq/a at a location of z = R2 /a (since a > R this puts the
image charge inside the sphere), then the potential on the sphere is still
zero. This is most easily seen if we write the potential due to the two point
charges in spherical coordinates. At a location r the potential due to the
original charge is

1 q
(1) Vq =
40 r + a 2ar cos
2 2

where weve used the cosine rule to get the distance from q to the point r.
The angle is, as usual, the angle between the z axis and the vector r.
By the same argument, the potential due to q0 is

1 q0
(2) Vq0 = q
40 2
r2 + (R2 /a) 2r (R2 /a) cos
1 q
(3) = q
40
(ar/R)2 + R2 2ra cos
If we look at the surface of the sphere, then r = R and we get Vq = Vq0
so the total potential on the surface of the sphere is zero. As you might
expect, this is a very special case, and its highly unusual to find problems
in which the method of images works this well.
1
METHOD OF IMAGES - POINT CHARGE AND SPHERE 2

However, having found the image charge, the net potential is just the sum
of the two above, so we get

q 1 1
(4) V (r) = q
40 r2 + a2 2ar cos 2 2
(ar/R) + R 2ra cos
From here, we can work out the induced surface charge on the sphere.
For a conductor, we have for the surface charge density :

V
(5) = 0
n

That is, we need the derivative of the potential normal to the surface. In this
case, the normal direction is in the direction of increasing r, so we just take
the derivative with respect to r. This gives

(6)

q r a cos ra2 /R2 a cos
=

4 (r2 + a2 2ra cos )
3/2  3/2
2 2
(ar/R) + R 2ra cos
r=R
(7)
q a2 R2
=
4 R (R2 + a2 2Ra cos )3/2
The induced charge is negative, since q itself is positive. We can find the
total induced charge on the sphere by integrating over the surface area.


a2 R2 R2 sin


2q
(8) qi = 3/2
d
4 0 R (R2 + a2 2Ra cos )
qR
(9) =
a
That is, the total induced charge is equal to the point image charge.
Finally, we can work out the energy of the configuration. The easiest way
to do that is to recognize that the electric field and thus the force outside
the sphere are the same in both configurations of the problem (the original
one with the grounded sphere, and the image version with the two point
charges). For a point charge at location zq (= a) and image charge at location
z = R2 /a the force between the charges is
METHOD OF IMAGES - POINT CHARGE AND SPHERE 3

qq0 1
(10) F=
40 zq R2 /zq 2


q2 R
(11) =
40 zq zq R2 /zq 2


q2 R zq
(12) =
40 z2 R2 2

q
The energy is the work done in bringing the charge q from infinity up to
location zq = a. Thus we need to integrate F dl over a path between these
two points. The easiest path to take is along the z axis. We get (again, we
take the negative of the force between the charges since were opposing this
force in doing the work):
a
q2 R zq
(13) W= 2 dzq
40 z2q R2
q2 R
(14) =
80 (a2 R2 )
P INGBACKS
Pingback: Method of images - sphere in uniform field
METHOD OF IMAGES: TWO CONDUCTING PLANES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Sec 3.2, Problem 3.10.
Another example of the method of images is the problem of a point
charge +q located at point (x, y) = (a, b) in the first quadrant (at z = 0),
between two conducting planes that cover the xz and yz planes, thus these
two conducting planes meet at right angles.
Following the procedure for the simpler problem of an image charge next
to a single conducting plane, we can first place images of q at locations
(x, y) = (a, b) and (a, b). If we stopped there, the potential would be

(1)
q 1 1 1
V= q q q
40 2 2 2 2 2 2 2 2 2
(x a) + (y b) + z (x + a) + (y b) + z (x a) + (y + b) + z
This clearly isnt zero on either of the conducting planes, so we need to
add another image. If we try an image of +q at (x, y) = (a, b) then the
potential is

(2)
q 1 1 1
V= q q q +q
40
(x a)2 + (y b)2 + z2 (x + a)2 + (y b)2 + z2 (x a)2 + (y + b)2 + z2 (
Now V = 0 on both the planes x = 0 and y = 0 as can be seen by direct
substitution.
The force on the original charge can be found from the force from the 3
images:

!
q2 1 1 1
(3) F= 2 x 2 y + 3/2
(ax + by)
40 4a 4b 4 (a2 + b2 )
The third term is the force between the actual charge
 and the image 
at
(x, y) = (a, b). The magnitude of this force is q2 / 40 4a2 + 4b2
and weve resolved this along the two coordinate axes.
1
METHOD OF IMAGES: TWO CONDUCTING PLANES 2

The work required to bring in q from infinity can be found from the gen-
eral formula for work applied to the images:

1 n
(4) W= qiV (ri)
2 i=1

where V (ri ) is the potential due to all the charges in the collection except
qi . However, this work assumes that were looking at all space, whereas the
conducting planes cut the space under consideration down to a quarter of
all space, so we need to divide the result by 4.
Plugging in the values we get
 
q 1 1 1
(5) V (a, b) = +
40 2a 2b 2 a2 + b2
 
q 1 1 1
(6) V (a, b) = +
40 2a 2b 2 a2 + b2
 
q 1 1 1
(7) V (a, b) = +
40 2a 2b 2 a2 + b2
 
q 1 1 1
(8) V (a, b) = +
40 2a 2b 2 a2 + b2
Thus

1 1 q2
 
1 1 1
(9) W= 4 +
4 2 40 2a 2b 2 a2 + b2
q 2 
1 1 1

(10) = +
80 2a 2b 2 a2 + b2
This technique can actually be applied to a configuration where we have
two conducting planes meeting at other angles. Since the image charges
have to cancel in pairs, we can derive a formula for the case where we can
divide up space into an even number n of sectors. Weve just seen how it
works for n = 4, but for general even n the argument would go like this.
Suppose the sector bounded by the planes has one plane at y = 0 (that is,
the xz plane) and the other at an angle of 2/n. Lets place the test charge
at a location (x, y) = d (cos , sin ) where is the angle the radius vector
to the charge makes with the plane y = 0 and d is the distance of the charge
from the origin. We can then place the first image by drawing a line from the
charge perpendicular to the plane at angle 2/n and extending it an equal
distance on the other side. This image will have charge q and be at an
angle of 2 2 4
n = n . The next image is found by drawing a line
METHOD OF IMAGES: TWO CONDUCTING PLANES 3

from the first image perpendicular to the plane at angle 4/n and extending
it an equal distance on the other side. This charge is +q and is at angle
4
n + . We continue in this fashion until we arrive at a charge of q at
angle 2 , which is the image of the original charge in the y = 0 plane.
This will give a total of n charges (one is the test charge and the other n 1
are the images). The potential is then

(11)
n
2 1
40 1
V= q
q 4m
2 4m
2
m=0 x d cos n + + y d sin n + + z2
n
2
1
(12) q 2 2
4m 4m
m=1 x d cos n + y d sin n + z2
POTENTIAL OF TWO COPPER PIPES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 3.11.
Time to fill in a gap that I had overlooked. This problem is given in
Griffiths as belonging to the method of images, but its not really an image
problem. It goes like this: We have two infinite copper pipes, each of radius
R. The axis of one pipe is on the line x = d, (y = 0) and the other is on
x = +d. The potential of the one on the left is held at V0 and the one on
the right at +V0 . We are to find the potential everywhere.
The solution doesnt really involve images, in that no part of space is
outside the realm in which we want the answer. What it does do is exploit
the uniqueness of solutions to Laplaces equation. If we look back at the
problem of two charged wires, we found that if we had a linear charge
density of on a wire at x = a and a density of + at x = +a, the
potential is

(x + a)2 + z2
(1) V= ln
40 (x a)2 + z2
We also found that the equipotential surfaces (where V is a constant) are
cylinders with axes at

a
(2) x=
tanh (20V / )

and radii of

a
(3) r=
|sinh (20V / )|
Therefore, the copper pipe problem is equivalent to a double-wire prob-
lem if we can find a pair of wires that have the equipotential surfaces speci-
fied above. That is, we must have an equipotential of V = +V0 for a cylinder
with axis x = +d and radius r = R, and an equipotential of V = V0 for a
cylinder with axis x = d and radius r = R. This means we need to find a
and satisfying these conditions. So
1
POTENTIAL OF TWO COPPER PIPES 2

a
(4) d=
tanh (20V0 / )
a
(5) R=
sinh (20V0 / )
From which we get

d sinh (20V0 / )
(6) = = cosh (20V0 / )
R tanh (20V0 / )
20V0
(7) =
cosh1 Rd
(8) a = R sinh (20V0 / )
q
(9) = R cosh2 (20V0 / ) 1
p
(10) = d 2 R2
The equipotential V = V0 gives us the cylinder with axis at x = a.
With these substitutions, the potential is given by 1.
P INGBACKS
Pingback: Copper pipes in a conducting medium
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 1

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 3.12 - 3.13.
Here are a few examples of calculating the Fourier coefficients for some
special cases.
Example 1. Consider the infinite slot problem with the boundary at x = 0
consisting of a conducting strip with a constant potential of V0 . In this case
we get

2V0 a ny
(1) cn = sin dy
a 0 a
2V0
(2) = (1 cos n)
n
The coefficients are thus zero for even n and 4V0 / for odd n:
(
0 n even
(3) cn = 4V0
n n odd
The potential is thus

4V0
enx/a ny
(4) V (x, y) = sin
n=1,3,5,... n a

Example 2. Now suppose the boundary at x = 0 consists of two conduct-


ing strips, insulated from each other and from the infinite sheets. The first
strip, from y = 0 to y = a/2 has a constant potential V0 while the other strip,
from y = a/2 to y = a is held at potential V0 .
Here, the coefficients cn are given by
" #
a/2 a
2V0 ny ny
(5) cn = sin dy sin dy
a 0 a a/2 a
2V0 h n i
(6) = 1 2 cos + cos n
n 2
1
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 1 2

If n is odd, this comes out to zero. If n is even, there are two cases. First,
if n = 2, 6, 10, . . . the term in brackets is 4. If n = 4, 8, 12, . . . the term in
brackets is zero. Thus we get

0
n odd
8V0
(7) cn = n n = 2, 6, 10 . . .

0 n = 4, 8, 12 . . .
Thus the potential is

8V0
enx/a ny
(8) V (x, y) = sin
n=2,6,10... n a
8V0 e(4n+2)x/a (4n + 2)y
(9) = sin
n=0 4n + 2 a

where in the last line weve changed the index of summation since the non-
zero terms in the first sum are just those with n = 4m + 2 starting at m = 0.
Example 3. The infinite slot with the strip at x = 0 held at potential V0
has the solution

4V0
enx/a ny
(10) V (x, y) = sin
n=1,3,5,... n a
For a conductor, the surface charge density can be found from the deriv-
ative taken normal to the surface:

V
(11) = 0
n x=0
In this case, the normal to the surface is the x direction, so we get

4V0  n  enx/a ny
(12) = 0 a sin
n=1,3,5,... n a
x=0

4V0 ny
(13) = 0 sin
a n=1,3,5,... a
This looks fine except for the problem that the series doesnt converge.
Consider y = a/2. The series is then a sum of an alternating sequence of
+1 and 1. The original series for the potential does converge at x = 0 due
to the n in the denominator. Not sure what the solution to this is.
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 2

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 3, Post 14.
Another example of the solution of Laplaces equation in a two-dimensional
problem.
We have an infinite rectangular pipe extending to infinity in both direc-
tions, lying parallel to the z axis. The four sides of the pipe are as follows.
At y = 0 and y = a the potential is held at V = 0. At x = 0 the potential
is also V = 0, but at x = b it is some arbitrary function of y: V = V0 (y).
We can use separation of variables and Fourier series to find the potential
everywhere inside the pipe.
The general solution from separation of variables gives us

 
(1) V (x, y) = Aekx + Bekx (C sin ky + D cos ky)

for constants A, B,C, D. In this case we could choose to swap x and y in


the solution, since neither x nor y goes to infinity so theres no requirement
for either term to vanish at infinity. However, with the given boundary
conditions, the current choice makes things easier (though feel free to try
 round if you like; that is, try a solution of form V (x, y) =
it the other way
Aeky + Beky (C sin kx + D cos kx)and see how far you get).
The boundary conditions are



0 y=0

0 y=a
(2) V=

0 x=0

V0 (y) x=b

The first condition gives

(3) D=0

The second gives


1
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 2 2

n
(4) k=
a

for n = 1, 2, 3, . . .
The third gives

(5) A+B = 0
(6) A = B
We therefore get, for a particular choice of n:
  n
nx/a nx/a
(7) Vn (x, y) = AC e e sin y
a
nx ny
(8) = 2AC sinh sin
a a
nx ny
(9) cn sinh sin
a a

where in the last line weve merged the constant 2AC into the single constant
cn .
As usual, we can now form the general solution as a series of Vn terms:


nx ny
(10) V (x, y) = cn sinh sin
n=1 a a
The coefficients cn can be found from the fourth boundary condition
above, by multiplying both sides by sin my
a and integrating.

a a
my nb my ny
(11) V0 (y) sin dy = cn sinh sin sin dy
0 a n=1 a 0 a a
a mb
(12) = cm sinh
2 a
Reverting to using n as the subscript on the coefficients, we get
a
2 ny
(13) cn = V0 (y) sin dy
a sinh nb
a 0 a
We cant go any further without specifying V0 (y).
In the special case where V0 (y) = V0 =constant, we can work out the
integral on the right and get
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 2 3

(
0 n = 2, 4, 6, . . .
(14) cn = 4V0
n = 1, 3, 5, . . .
n sinh nb
a

In this case, the general solution is

ny
4V0 sinh nx
a sin a
(15) V (x, y) =
n=1,3,5... n sinh nb
a

P INGBACKS
Pingback: Laplaces equation - Fourier series example 4
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 3 -
THREE DIMENSIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.15.
The examples of solving Laplaces equation weve seen so far have all
been essentially two-dimensional, so its time to see a fully three dimen-
sional problem.
We have a cube of side length a whose faces are conducting plates. The
cube is placed with one corner at the origin, and the other corner at (x, y, z) =
(a, a, a), with its edges parallel to the coordinate axes. The face at z = a is
held at a constant potential of V0 (it is insulated from the other faces). All
other faces are grounded so their potential is V = 0. We want to find the
potential inside the cube.
We use the standard separation of variables technique we described ear-
lier. We assume that

(1) V (x, y, z) = X(x)Y (y)Z(z)

Plugging this into Laplaces equation and dividing through by V , we get


three ordinary differential equations:

1 d2X
(2) = C1
X dx2
1 d 2Y
(3) = C2
Y dy2
1 d2Z
(4) = C3
Z dz2

with the condition on the constants

(5) C1 +C2 +C3 = 0

The boundary conditions are


1
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 3 - THREE DIMENSIONS 2



0 x=0

0 x=a





0 y=0
(6) V=

0 y=a

0 z=0





V
0 z=a

The boundary conditions involving x and y suggest that a sine solution is


appropriate here, so we can try setting C1 = k2 < 0, C2 = l 2 < 0 and C3 =
k2 + l 2 > 0. Again, these choices are really inspired guesswork, and you
should feel free to try changing the signs of the constants and discovering
that the solution doesnt work out as nicely.
With these choices, we get, in a similar way to solving the two-dimensional
problem:

(7) X(x) = A sin kx + B cos kx


(8) Y (y) = C sin ly + D cos ly

k2 +l 2 z k2 +l 2 z
(9) Z(z) = Ee + Fe
The first boundary condition implies B = 0, the third implies D = 0 and
the fifth implies E = F. The second condition implies k = n/a and the
fourth implies l = m/a for some positive integers n and m. Putting all this
together, we get

nx
(10) X(x) = An sin
a
my
(11) Y (y) = Cm sin
 a 
n2 +m2 z/a n2 +m2 z/a
(12) Z(z) = Enm e e
!
n2 + m2 z
(13) = 2Enm sinh
a

where weve added subscripts to the constants to indicate that the constants
could be different for each choice of n and m.
This solution on its own clearly doesnt satisfy the sixth boundary con-
dition for z = a, but as usual, we can get around this by creating a Fourier
series from these individual solutions. The most general solution therefore
has the form
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 3 - THREE DIMENSIONS 3

!

nx my n2 + m2 z
(14) V (x, y, z) = cnm sin sin sinh
m=1 n=1 a a a

where as usual we have merged the constants together: cnm 2AnCm Enm .
We can now apply the sixth boundary condition to obtain a formula
for the coefficients cnm . We set z = a and then multiply both sides by
0 0
sin n ax sin m ay and integrate over x and y. Since V = V0 at z = a, we get

(15)
a a
n0 x m0 y p  a nx n0 x
a
my
V0 sin dx sin dy = cnm sinh 2 2
n +m sin sin dx sin s
0 a 0 a m=1 n=1 0 a a 0 a
The left side is

(
a
n0 x a
m0 y 0 n0 or m0 even
(16) V0 sin dx sin dy = 4a2V0
0 a 0 a
n0 m0 2
n0 and m0 odd
The terms on the right side are zero unless n = n0 and m = m0 . The sum
reduces to a single term, which is

p  a2
(17) n02 + m02
cn0 m0 sinh
4
Dropping the primes to clean up the notation, we therefore get

(
0 n or m even
(18) cnm = 16V
0 n and m odd
nm 2 sinh( n2 +m2 )

The overall solution is therefore


 2 2 
nx my n +m z
16V0 sin a sin a sinh a
(19) V (x, y, z) = 2  
n=1,3,5,... m=1,3,5,... nm sinh n2 + m2

Since this solution satisfies the boundary condition at z = a we get the


rather curious formula:

1 nx my 2
(20) sin sin =
n=1,3,5,... m=1,3,5,... nm a a 16
LAPLACES EQUATION - FOURIER SERIES EXAMPLES 3 - THREE DIMENSIONS 4

which is apparently true for all values of x and y such that 0 < x < a and
0 < y < a.
LAPLACES EQUATION IN SPHERICAL COORDINATES:
EXAMPLES 1

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 3.17, 3.18.
Example 1. A simple example of Laplaces equation in spherical coor-
dinates is that of a spherical shell of radius R with a constant potential V0
over its surface. We want to find the potential inside and outside the sphere.
The general solution in spherical coordinates was found in the last post:

 
l Bl
(1) V (r, ) = Al r + l+1 Pl (cos )
l=0 r
Inside the sphere, all Bl are zero to prevent an infinity at the origin, so we
get


(2) V (r, ) = Al rl Pl (cos )
l=0

At the spherical boundary, r = R so we get


(3) V0 = Al Rl Pl (cos )
l=0

The only Legendre polynomial that doesnt depend on is P0 = 1, so it


is only the l = 0 term that is non-zero, and we get A0 = V0 , so inside the
sphere, V = V0 everywhere.
Outside the sphere, all Al = 0 to prevent the potential increasing to infin-
ity for large r. Again, the potential must satisfy the boundary condition at
r = R, and we get


l B
(4) V0 = Rl+1 Pl (cos )
l=0

As above, we discard all terms except for l = 0, which gives B0 = RV0 ,


and
1
LAPLACES EQUATION IN SPHERICAL COORDINATES: EXAMPLES 1 2

RV0
(5) V=
r
Example 2. We saw that the coefficients Al and Bl can be found by
working out integrals, but in some special cases, it is easier to match up
terms in the series on both sides of the equation. This happens if we can
express V as a series of cosines (admittedly, this doesnt happen very often,
but they are popular student exercises).
For example, suppose we have a spherical shell of radius R on which the
potential is V ( ) = k cos 3 . Using some trig identities, we can convert the
cosine term.

(6) cos 3 = cos(2 + )


(7) = cos 2 cos sin 2 sin
(8) = (2 cos2 1) cos 2 sin2 cos
(9) = 2 cos3 cos 2(1 cos2 ) cos
(10) = 4 cos3 3 cos

We can now apply this to the general solution 1. Inside the sphere, the
Bl terms are all zero to prevent an infinity at the origin, so we get at the
boundary:


(11) V (R, ) = Al Rl Pl (cos )
l=0

k 4 cos3 3 cos = Al Rl Pl (cos )

(12)
l=0

Since the only terms appearing in the potential are of degree 1 and 3,
only P1 and P3 appear in the series on the right. From tables of Legendre
polynomials, we have

(13) P1 (cos ) = cos


5 3
(14) P3 (cos ) = cos3 cos
2 2
Matching up terms for the 3rd degree term, we get
LAPLACES EQUATION IN SPHERICAL COORDINATES: EXAMPLES 1 3

5
(15) 4k = A3 R3
2
8k
(16) A3 =
5R3
With this value of A3 , the l = 3 term in the series contributes a term
12
5 k cos , so combining this with the l = 1 term and equating this to the
degree 1 term on the LHS, we get

12
(17) 3k = A1 R k
5
3k
(18) A1 =
5R
The potential inside the sphere is thus:

r3
 
k r
(19) Vin (r, ) = 3 P1 (cos ) + 8 3 P3 (cos )
5 R R
Outside the sphere, we can use the technique, except this time it is the Al
terms that are all zero to avoid an infinite potential for large r. We get, at
the boundary:


l B
(20) V (R, ) = Rl+1 Pl (cos )
l=0

l B
k 4 cos3 3 cos =

(21) Rl+1 Pl (cos )
l=0
For the 3rd degree term:

5B3
(22) 4k =
2R4
8kR4
(23) B3 =
5
The l = 3 term contributes a term of 12
5 k cos as before, so combining
this with the l = 1 term, we get

B1 12
(24) 3k = k
R2 5
3kR 2
(25) B1 =
5
LAPLACES EQUATION IN SPHERICAL COORDINATES: EXAMPLES 1 4

The outside potential is

3R2 8R4
 
k
(26) Vout (r, ) = 2 P1 (cos ) + 4 P3 (cos )
5 r r
P INGBACKS
Pingback: Laplaces equation in spherical coordinates: surface charge
LAPLACES EQUATION IN SPHERICAL COORDINATES:
SURFACE CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 3.19, 3.18.
We found the general solution of Laplaces equation in spherical coordi-
nates to be:

 
l Bl
(1) V (r, ) = Al r + l+1 Pl (cos )
l=0 r

If the potential is specified as V0 ( ) on the surface of a sphere, and there


are no charges inside or outside the sphere, we can use the relations between
the potential and the surface charge density to find that charge density on
the sphere. Since the electric field is discontinuous across a charge layer,
we have

above below
(2) E E =
0

and the field is related to the potential by

(3) E = V

we can use the radial symmetry of the problem to deduce that, at the surface
of the sphere where r = R,


V V
(4) =
r out r in
0

For this problem, we know that inside the sphere, Bl = 0 and outside the
sphere, Al = 0. So we have
1
LAPLACES EQUATION IN SPHERICAL COORDINATES: SURFACE CHARGE 2


V
(5) = lAl rl1Pl (cos )
r in l=0

V Bl
(6) = (l + 1) l+2 Pl (cos )
r
out l=0 r
At the boundary, we must have


Bl l1 ( )
(7) (l + 1) Rl+2
Pl (cos ) + lAl R Pl (cos ) =
0
l=0 l=0
However, since the potential is also continuous across a charge layer, we
must have


l B
(8) V0 ( ) = Al Rl Pl (cos ) = Rl+1 Pl (cos )
l=0 l=0
Since the Legendre polynomials are orthogonal, these two series must be
equal term by term, so we get

Bl
(9) Al Rl =
Rl+1
Multiplying 8 both sides by Pm (cos ) sin and integrating from 0 to ,
we get, using the orthogonality of the polynomials

2
(10) V0 ( )Pm (cos ) sin d = Am Rm
0 2m + 1
(11) Cm
Putting this back into 7 we get

 
( ) 2l + 1 2l + 1
(12) = (l + 1) + l Cl Pl (cos )
0 l=0 2R 2R
1
(13) = (2l + 1)2 Cl Pl (cos )
2R l=0
We can apply this result to find the charge distribution for the case where
V0 = k cos 3 . By using trig identities, we can expand the cosine and express
V0 in terms of the Pl :

k
(14) V0 ( ) = (3P1 (cos ) + 8P3 (cos ))
5
LAPLACES EQUATION IN SPHERICAL COORDINATES: SURFACE CHARGE 3

From here, we see that has contributions only from l = 1 and l = 3,


and we get from the above formula for the Cl coefficients:

3k
(15) C1 =
15
8k
(16) C3 =
35  
k0 9 56
(17) ( ) = P1 (cos ) + P3 (cos )
R 5 5
P INGBACKS
Pingback: Laplaces equation - charged sphere
Pingback: Laplaces equation - cylindrical shell
Pingback: Laplaces equation: conducting sphere and shell of charge
LAPLACES EQUATION - CHARGED DISK

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.21
In Example 3 in an earlier post, we found that the potential of a uniformly
charged disk of radius R can be worked out for points on the axis of the disk
and is

p 2 
(1) V (z) = z + R2 |z|
20

In spherical coordinates, a point with coordinate z on the axis has coor-


dinates (r, 0) if z > 0 or (r, ) if z < 0. We can therefore write

p 2 2

(2) V (r, 0) = V (r, ) = r +R r
20

This formula is valid only for points on the z axis, but we can use it, to-
gether with the general solution of Laplaces equation in terms of Legendre
polynomials, to get an approximation for the potential off the z axis.
We consider first the case of r > R. In this case the most general solution
of Laplaces equation in spherical coordinates is:


B
(3) V (r, ) = rl+1l Pl (cos )
l=0

For = 0 we get, since Pl (0) = 1 for all l:


B
(4) V (r, 0) = rl+1l
l=0

To find the coefficients Bl we need to expand 2. We can expand it in


powers of (R/r):
1
LAPLACES EQUATION - CHARGED DISK 2

s
r R2
(5) V (r, 0) = 1 + 2 1
20 r

R2 R4
 
(6) = +...
20 2r 8r3
Comparing with 4, we get

R2
(7) B0 =
40
(8) B1 = 0
R4
(9) B2 =
160

so the general formula for the potential off the axis is

R2 R4
 
(10) V (r, ) = P0 (cos ) 3 P2 (cos ) + . . .
20 2r 8r
Since the odd polynomial terms all vanish, and the even Legendre poly-
nomials are even functions, this expansion is also valid for the region z < 0,
where /2 < .
For the case r < R, we have


(11) V (r, ) = Al rl Pl (cos )
l=0
When = 0, we get


(12) V (r, 0) = Al r l
l=0
We can now expand 2 in powers of r/R, and we get
r !
r2
(13) V (r, 0) = R 1+ 2 r
20 R
r2 r4
 

(14) = Rr+ +...
20 2R 8R3
Comparing this to 12, we get
LAPLACES EQUATION - CHARGED DISK 3

R
(15) A0 =
20

(16) A1 =
20

(17) A2 =
40 R
(18) A3 = 0
The off-axis expansion thus begins:

r2
 

(19) V (r, ) = RP0 (cos ) rP1 (cos ) + P2 (cos ) + . . .
20 2R
In this case, since there is a P1 term, the same expansion isnt valid for
= . However, because the odd polynomials are odd functions, for =
we have


(20) V (r, ) = (1)l Al rl
l=0
The only change we need to make in the expansion is thus to change the
sign of A1 so we get

r2
 

(21) V (r, ) = RP0 (cos ) + rP1 (cos ) + P2 (cos ) + . . .
20 2R
All higher odd polynomials have Al = 0 so this is the only change that is
required in the entire expansion.
P INGBACKS
Pingback: Laplaces equation - charged line segment
Pingback: Field of a polarized object - examples
LAPLACES EQUATION - CHARGED SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.22.
Weve seen one example of the relation between surface charge and po-
tential using the series solution to Laplaces equation. We consider a spher-
ical shell of radius R and wish to find the potential from the surface charge
on this shell. The general solution to Laplaces equation is

 
l Bl
(1) V (r, ) = Al r + l+1 Pl (cos )
l=0 r
Inside the sphere, to prevent an infinity at the origin, Bl = 0, while out-
side, to prevent infinities in the other direction, Al = 0. In the earlier post,
we found that:

( ) 1
(2)
0
= (2l + 1)2 Cl Pl (cos )
2R l=0
2
(3) Cl Al Rl
2l + 1
Simplifying, we get


( )
(4) = (2l + 1)Al Rl1 Pl (cos )
0 l=0
From here, we can use the orthogonality of the Legendre polynomials to
get an expression for Al . Multiply both sides by Pm (cos ) sin and inte-
grate from 0 to :

(5)

1
l1
( )Pm (cos ) sin d = (2l + 1)Al R Pm (cos )Pl (cos ) sin d
0 0 l=0 0

2
(6) = (2l + 1)Al Rl1 2m + 1 lm
l=0
m1
(7) = 2R Am
1
LAPLACES EQUATION - CHARGED SPHERE 2

So, reverting back to l as the index:



1
(8) Al = ( )Pl (cos ) sin d
2Rl1 0 0
As an example, consider a spherical shell with a uniform charge density
of +0 in the region 0 < /2 and 0 from /2 < . We can use
8 to work out the Al explicitly. First, note that sin is an even function over
the interval [0, ] (that is, sin(/2 + x) = sin(/2 x)). Over the same in-
terval, Pl (cos ) is even if l is even and odd if l is odd. The specified charge
distribution is odd. (Another way is to convert the integral by the substitu-
tion x = cos and then the functions are even or odd in the traditional way,
with respect to the origin x = 0 as the mid-point.)
Therefore, only odd l terms will have a non-zero integral. For these
terms, we need to work out

0
(9) Al = Il
2Rl1 0
/2
(10) Il Pl (cos ) sin d Pl (cos ) sin d
0 /2
We can look up tables of Legendre polynomials and work out these inte-
grals manually (they all involve powers of the cosine multiplied by a single
sine, so the integrals are fairly straightforward, but tedious). Or we can use
software to do the integrals, and we get

(11) I0 = I2 = I4 = I6 = . . . = 0
(12) I1 = 1
1
(13) I3 =
4
1
(14) I5 =
8
The potential, up to l = 6, is therefore, for r < R:

r3 r5
 
0
(15) V (r, ) = rP1 (cos ) 2 P3 (cos ) + 4 P5 (cos ) + . . .
20 4R 8R
For r > R, we found in the earlier post that Bl = Al R2l+1 from the re-
quirement that the potential be continuous at the boundary. Therefore

0 l+2
(16) Bl = R Il
20
LAPLACES EQUATION - CHARGED SPHERE 3

so

0 R 3 R5 R7
 
(17) V (r, ) = P1 (cos ) 4 P3 (cos ) + 6 P5 (cos ) + . . .
20 r2 4r 8r
LAPLACES EQUATION IN CYLINDRICAL COORDINATES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.23.
We can use the separation of variables technique to solve Laplaces equa-
tion in cylindrical coordinates, in the special case where the potential does
not depend on the axial coordinate z. In general, Laplaces equation in
cylindrical coordinates is

1 2V 2V
 
1 V
(1) r + 2 + =0
r r r r 2 z2
We let V (r, ) = R(r)( ) and then multiply through by r2 and divide
through by V :

1 2
 
r R
(2) r + =0
R r r 2
Since each term depends only on a separate independent variable, each
term must be a constant. Well consider first the special case where this
constant is zero. In this case, we get from the first term:

(3) r2 R00 (r) + rR0 (r) = 0


This has the solution

(4) R(r) = A ln r + B

as can be verified by direct substitution. From the second term

(5) ( ) = C + D
However, we need to remember that is an angle, and is restricted to
[0, 2], so the C term doesnt behave properly, in that its not periodic. So
we need to take C = 0, giving = constant as the only solution in this case.
1
LAPLACES EQUATION IN CYLINDRICAL COORDINATES 2

If we now consider the more general case, then the angular equation is,
taking the constant as k2 as in previous applications of separation of vari-
ables:

(6) 00 ( ) = k2

which has the general solution

(7) ( ) = C sin k + D cos k


The radial equation now becomes

(8) r2 R00 (r) + rR0 (r) k2 R(r) = 0


This has the general solution


(9) R= an r n
n=1
Substituting into the ODE, we get


an n(n 1) + an n an k2 rn = 0
 
(10)
n=1
From the uniqueness of power series, the coefficient of each power of r
must be zero, from which we get

(11) an (n2 k2 ) = 0
Thus either n = k or an = 0. This means that k must be an integer
[though see comment from Artur Gower below], and for a given choice of
k = n, the solution is

(12) Rn (r) = an rn
The general solution is the linear combination of all the particular solu-
tions, so we get (redefining the constants):

(13)
1
n
V (r, ) = A ln r +B+ r (An sin n + Bn cos n )+ rn (Cn sin n + Dn cos n )
n=1 n=
LAPLACES EQUATION IN CYLINDRICAL COORDINATES 3

For example, in the case of an infinite wire, V is independent of so we


get

(14) V (r) = A ln r + B
As weve seen earlier (Example 2 in this post), to nail this down any more
requires that we specify a reference point (or surface) where V = 0, and for
this case it doesnt really matter as long as we avoid r = 0 and r = .
P INGBACKS
Pingback: Laplaces equation - infinite pipe
Pingback: Laplaces equation - cylindrical shell
Pingback: Laplaces equation - cylindrical shell with opposing charges
Pingback: Dielectric cylinder in uniform electric field
Pingback: Electric potential from a steady current
Pingback: Electric field outside an infinite wire
LAPLACES EQUATION - INFINITE PIPE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.24.
Heres an example of using the solution to Laplaces equation in cylindri-
cal coordinates. We have an infinite cylindrical conducting pipe of radius R
lying with its axis on the z axis. A constant electric field E0 pointing in the
+x direction covers all space. Find the potential outside the pipe.
Since the pipe is a conductor, the potential is constant everywhere on it.
In this case, the potential will not tend to zero at infinity, since there is an
electric field everywhere, so we might as well choose the zero of V to be on
the pipe. Far from the pipe, the potential must be

(1) V f ar = E0 x +C

We can choose C = 0 which means we are choosing V = 0 on the yz


plane. The boundary conditions are therefore, in cylindrical coordinates:

(
0 r=R
(2) V=
E0 r cos r

The solution of Laplaces equation in cylindrical coordinates is

(3)
1
V (r, ) = A ln r +B+ rn (An sin n + Bn cos n )+ rn (Cn sin n + Dn cos n )
n=1 n=

Since the sine is odd and the cosine is even, we can rewrite this as

(4)  
 
n Cn n Dn
V (r, ) = A ln r + B + An r n sin n + Bn r + n cos n
n=1 r n=1 r

For r , the terms with rn in the denominator can be neglected, and we


get
1
LAPLACES EQUATION - INFINITE PIPE 2

(5) An = 0
(6) B1 = E0
(7) Bn = 0 (n 6= 1)
(8) A=B=0
For r = R, we have

Cn
(9) An Rn =
Rn
Dn
(10) Bn Rn = n
R
Combining these two sets of results, we get

(11) An = Cn = 0 (all n)
(12) B1 = E0
(13) D1 = E0 R2
(14) Bn = Dn = 0 (n 6= 1)
Thus the final formula for the potential is

R2
 
(15) V (r, ) = E0 r + cos
r
The induced charge density on the pipe can be found from the normal
derivative of the potential at the surface of the conductor. In this case, the
normal to the pipe is radially outwards, so we get

V
(16) = 0
r r=R
(17) = 20 E0 cos
P INGBACKS
Pingback: Laplaces equation - cylindrical shell
LAPLACES EQUATION - CYLINDRICAL SHELL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.25.
As another example of applying the solution to Laplaces equation in
cylindrical coordinates, we consider the following problem. We are given a
cylindrical non-conducting shell or radius R carrying a charge density of

(1) ( ) = k sin 5
We wish to find the potential outside and inside the cylinder.
We start with the general solution:

(2)  
 
n Cn n Dn
V (r, ) = A ln r + B + An r n sin n + Bn r + n cos n
n=1 r n=1 r
Since we are given the surface charge, we can use a similar procedure
to that used in solving problems in spherical coordinates. We can start by
finding the general form of the potential in the two regions. Outside the
shell, to keep V finite, we eliminate the ln r and rn terms. (In fairness, it has
to be pointed out that the potential of an infinite charged wire does go as
ln r so in that case it seems acceptable for V to be infinite at large distances.
Because the solution to Laplaces equation with a given set of boundary
conditions is unique, though, the fact that we can find a solution in this
problem in which V remains finite at all locations must mean that it is the
actual solution.)
Thus, outside the shell, we have

 
Dn Cn
(3) Vout = Bout + n
cos n n sin n
n=1 r r
Inside the shell, we eliminate the terms in ln r and 1/rn to prevent an
infinity at r = 0. We have


(4) Vin = Bin + [An rn sin n + Bn rn cos n ]
n=1
1
LAPLACES EQUATION - CYLINDRICAL SHELL 2

Since the potential is continuous over a surface charge, we must have


Vout (R) = Vin (R), so we get

(5)  

Dn Cn
Bout + n
cos n n sin n = Bin + [An Rn sin n + Bn Rn cos n ]
n=1 R R n=1

Equating coefficients of the sine and cosine, we get

(6) Bout = Bin


(7) Cn = An R2n
(8) Dn = Bn R2n
The outward derivative of the potential is discontinuous across a surface
charge, and we have


V V
(9) =
r out r in
0
Plugging in the formulas for Vout and Vin , we get

(10)
nR2n An nR2n Bn
  
n1 n1 k
n+1
nR An sin n + n+1
nR Bn cos n = sin 5
n=1 R n=1 R 0
Equating coefficients of sine and cosine, we see that the only non-zero
term on the left must be the sin 5 term, so we have

k
(11) 10A5 R4 =
0
k
(12) A5 =
100 R4
k 6
(13) C5 = R
100
(14) An = Cn = 0 (n 6= 5)
(15) Bn = Dn = 0 (all n)
We are left with the constants Bout = Bin and might as well take them to
be zero, since an arbitrary constant doesnt affect the potential. We therefore
get
LAPLACES EQUATION - CYLINDRICAL SHELL 3

kR6
(16) Vout (r, ) = sin 5
100 r5
kr5
(17) Vin (r, ) = sin 5
100 R4
P INGBACKS
Pingback: Laplaces equation - cylindrical shell with opposing charges
Pingback: Field of a polarized cylinder
MULTIPOLE EXPANSION IN ELECTROSTATICS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 3.4.1 & Problem 3.26
For a given charge distribution, we can write down a multipole expansion,
which gives the potential as a series in powers of 1/r, where r is the distance
from the origin to the observation point.
We know that the potential in general is

1 d 3 r0
(1) V (r) = (r0 )
40 |r r0 |
In the integral, r0 is the position of charge element (r0 )d 3 r0 . From the
law of cosines
p
(2) |r r0 | = r2 + r02 2rr0 cos 0

where 0 is the angle between r and r0 . We can rewrite this as

1 1
(3) 0
=
|r r | r + r 2rr0 cos 0
2 02
1 1
(4) = q
r 1 + r 2 r0 cos 0
02
r2 r
From the theory of Legendre polynomials, it is known that the last factor
in this expression is a generating function for the polynomials. That is, if
we write the square root as an power series, we get

 0 n
1 r 0
(5) q = Pn (cos )
02 0
1 + rr2 2 rr cos 0 n=0 r
 0 n
The coefficient of rr in the series is the Legendre polynomial Pn (cos 0 ).
This can be verified for the first few terms by calculating the Taylor series
expansion of the square root term about r0 /r = 0. This is tedious to do by
hand, but using Maple, we get, defining s r0 /r:
1
MULTIPOLE EXPANSION IN ELECTROSTATICS 2

(6)    
1 0 2 3 2 0 1 3 5 3 0 3 0
= 1+s cos +s cos +s cos cos +. . .
1 + s2 2s cos 0 2 2 2 2

It is important to note that the angle 0 is equivalent to the angle in


spherical coordinates only if the observation point r lies on the z axis, since
that is the only configuration where the angle between the observation vec-
tor and a charge element corresponds to the spherical coordinate angle .
(A more general multipole expansion uses spherical harmonics rather than
just Legendre polynomials, but thats a topic for a more advanced post.)
With this restriction, we can substitute the series expansion back into ??
to get

 0 n

1 0 r 0
(7) V (r) = (r ) Pn (cos ) d 3 r0
40 r n=0 r

1 1
(8) = (r0 )Pn (cos 0 )r0n d 3 r0
40 n=0 rn+1

The first few terms in this series have special names. The n = 0 term is


1 Q
(9) (r0 )d 3 r0 =
40 r 40 r

where Q is the total charge. This is called the monopole term, and shows
that to a first approximation, the potential of any charge distribution is just
the potential of a point charge with the same total charge.
The next term in the series is


1 0 0 0 1 3 0
(10) r P1 (cos )(r )d r = r0 cos 0 (r0 )d 3 r0
40 r2 40 r2
This is called the dipole term.
For n = 2, we get the quadrupole term

(11)  
1 02 0 0 3 0 1 02 3 2 0 1
r P2 (cos )(r )d r = r cos (r0 )d 3 r0
40 r3 40 r3 2 2
Finally, for n = 3 we get the octopole term
MULTIPOLE EXPANSION IN ELECTROSTATICS 3

(12)  
1 03 0 0 3 0 1 03 5 3
4
r P3 (cos )(r )d r = 4
r cos cos (r0 )d 3 r0
3 0 0
40 r 40 r 2 2
As an example, consider a solid sphere with a charge density

R
(13) (r0 ) = k
(R 2r0 ) sin 0
r02
We can use the integrals above to find the first non-zero term in the series,
and thus get an approximation for the potential. Note that we can do this
only for points on the z axis.
By direct calculation, we have for the monopole term:
R 2
1 0 3 0 1 R
(r )d r =
(14) k 02 (R 2r0 ) sin 0 r02 sin 0 d 0 d 0 dr0
40 r 40 r 0 0 0 r
(15) = 0

since the integral over r0 gives zero. Thus the monopole term vanishes, as it
always does if the total charge is zero.
For the dipole term, we get
R 2
1 0 0 0 1 3 0 R
r cos (r )d r =
(16) k (R 2r0 )r03 sin 0 cos 0 sin 0 d 0 d 0 dr0
40 r2 40 r2 0 0 0 r02
(17) = 0
This time, the integral over 0 gives zero, since the term cos 0 sin2 0 is
odd relative to the interval [0, ].
For the quadrupole term
  R 2
1 02 3 2 0 1 0 3 0 1 R
r cos (r )d r = k 02 (R 2r0 ) sin 0
40 r3 2 2 3
40 r 0 0 0 r
 
3 1
(18) r04 cos2 0 sin 0 d 0 d 0 dr0
2 2
 2 5
1 kR
(19) =
40 r3 48
kR5
(20) =
1920 r3
The octopole term comes out to zero, since the terms in 0 are again odd
relative to the interval [0, ].
MULTIPOLE EXPANSION IN ELECTROSTATICS 4

P INGBACKS
Pingback: Dipoles
Pingback: Multipole expansion of linear charge
Pingback: Quadrupole moment
Pingback: Magnetic dipole
DIPOLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 3.4.2 & Problem 3.27-3.29
Weve seen that the potential of a charge distribution can be written as a
multipole expansion in terms of Legendre polynomials:

1 1
(1) V (r) = rn+1 (r0 )Pn (cos 0 )r0n d 3 r0
40 n=0

where r is the vector from the origin to the observation point, r0 is the vector
from the origin to the point with charge density (r0 ) and 0 is the angle
between r and r0 .
The n = 0 term in this expansion is the monopole term, and the potential
for that reduces to the equivalent potential for a point charge.

1 Q
(2) V0 (r) = (r0 )d 3 r0 =
40 r 40 r
The n = 1 term is the dipole term, and dominates if the total charge is
zero (that is, if there are equal amounts of positive and negative charge).

1
(3) V1 (r) = (r0 )P1 (cos 0 )r0 d 3 r0
40 r2

1
(4) = r0 (r0 ) cos 0 d 3 r0
40 r2
Since 0 is the angle between r and r0 , we have r r0 = r0 cos 0 , where r
is the unit vector in the r direction. Since this unit vector is fixed for a given
r, we can take it outside the integral and get

1
(5) V1 (r) = r r0 (r0 )d 3 r0
40 r2
The integral is now a property of the charge distribution alone, without
any reference to r (although it does depend on the choice of origin). The
integral is a vector and depends on r0 , which varies as you move around
within the charge distribution. In practice, if the integrand doesnt possess
1
DIPOLES 2

any symmetry that makes it easy to integrate, the only practical way of
doing such an integral is to convert to rectangular coordinates, since the unit
vectors in the rectangular system are constants and can be taken outside the
integral.
The integral is called the dipole moment of the charge distribution, and is
usually given the symbol p:

(6) p r0 (r0 )d 3 r0
1
(7) V1 (r) = r p
40 r2
For n point charges, the dipole moment can be written as

n
(8) p = qi r0i
i=1
Example 1. We have four point charges as follows. Charge 3q placed at
(x, y, z) = (0, 0, a), q at (0, 0, a), 2q at (0, a, 0) and 2q at (0, a, 0). We
can use 8 to work out the dipole moment.

(9) p = qa(3z z 2y + 2y)


(10) = 2qaz
From here, we can get the dipole potential

1
(11) V1 (r) = p r
40 r2
In spherical coordinates, since p points along the z axis, we have p r =
2qaz r = 2qa cos , and

2qa cos
(12) V1 (r) =
40 r2
Example 2. We have a spherical shell of radius R centred at the origin
with a surface charge density of = k cos for a constant k (in spherical
coordinates), and we want to find the dipole moment.
We can use the symmetry of the situation to note that if we converted the
integral to rectangular coordinates, then since the charge density depends
only on , a charge element at location (x, y, z) will have an equal element
at (x, y, z), so the components of the dipole moment in the x and y direc-
tions will be zero. The component of r0 in the z direction is R cos , so the
moment is
DIPOLES 3

2
cos 0 (R cos 0 )(R2 sin 0 )d 0 d 0

(13) p = kz
0 0
4
(14) = kR3 z
3
The dipole component in the potential is therefore

1
(15) V1 (r) = p r
40 r2
kR3
(16) = cos
30 r2
Referring back to 1 and using the fact that = k cos = kP1 (cos ), we
can use the fact that the Legendre polynomials are orthogonal to conclude
that this must also be the only non-zero term in the multipole expansion.
Example 3. The classic dipole consists of two equal and opposite charges
a distance d apart. To make it easier to analyze in spherical coordinates, we
place the dipole with both charges on the z axis, and measure the vector r
from the midpoint of the distance between them. Thus we have a charge
+q at z = d/2 and q at z = d/2. We can then define r+ and r to be the
vectors from the two charges to the observation point. With the geometry
as given, the angle is the angle between the z axis and r, so its the usual
spherical angle . It is also the angle between r and the vector to +q, so
that is the angle between r and the vector to q.
Using the law of cosines for the sides of a triangle, we get
r
d2 d
(17) r+ = r2 + 2r cos
s 4 2
 2
d d
(18) = r 1 cos +
r 2r
r
d2 d
(19) r = r2 + 2r cos ( )
s 4 2
 2
d d
(20) = r 1 + cos +
r 2r
The potential at point r is therefore
 
q 1 1
(21) V (r) =
40 r+ r
DIPOLES 4

By substituting for r+ and r , we can then use a Taylor expansion on the


variable d/r (which will be small if the charges are close together and we
move far away from them) to get the multipole expansion. We get, for the
first few terms:

(22) V0 = 0
qd cos
(23) V1 =
40 r2
qd P1 ( )
(24) =
40 r2
(25) V2 = 0
qd 3 5 cos3 3 cos
(26) V3 =
40 8r4
qd 3 P3 (cos )
(27) =
40 4r4
(28) V4 = 0
qd 5 63 cos5 ( ) 70 cos3 + 15 cos
(29) V5 =
40 128r6
qd 5 P5 (cos )
(30)
40 16r6
Since r/r+ is the generating function for the Legendre polynomials, and
since cos( ) = cos , then r/r is the generating function for Le-
gendre polynomials with each term in the sum multiplied by (1)n . There-
fore, since V as given in 21 takes the difference between these two gener-
ating functions, all the terms involving even polynomials will cancel out,
leaving only the odd terms, as weve seen in the first few explicit terms
above.
For large enough r, the dipole term dominates, and we can say that for
large distances, the potential of a dipole of two point charges goes as 1/r2 .

P INGBACKS
Pingback: Stimulated emission of radiation: lasers
Pingback: Average field over a sphere
Pingback: Electric field of a dipole
Pingback: Quadrupole moment
Pingback: Polarizability of hydrogen
Pingback: Polarizability - linear charge distribution
Pingback: Dipole force and energy
DIPOLES 5

Pingback: Field of a polarized object


Pingback: Dipole in dielectric sphere
Pingback: Volume current density and dipole moment
Pingback: http://physicspages.com/pdf/Griffiths EM/Griffiths Problems
09.23.pdf
Pingback: Frequency dependence of electric permittivity
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
DIPOLE-DIPOLE INTERACTIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.1.3; Problems 3.33, 4.5, 4.6.
Weve seen that the electric field due to a dipole (aligned so its centre is
at the origin and p points along the +z axis) is (in spherical coordinates):

p  
(1) E= 2 cos r + sin
40 r3

We can express this in a coordinate-free form as follows. The angle is


the angle between p and the unit vector r in the radial direction, so p r =
p cos . The unit vector makes an angle of + /2 with p, so p =
p cos( + /2) = p sin . Therefore

(2) p = (p r)r + (p )
(3) = p cos r p sin

So

p  
(4) E= 3
2 cos r + sin
40 r
1
(5) = [2(p r)r p + (p r)r]
40 r3
1
(6) = [3(p r)r p]
40 r3

For a physical dipole where charges +q and q are separated by a vector


distance d, the net force in a uniform field E is zero, since the force on the
positive charge cancels that on the negative one. However, there is a torque
acting about the centre point. This torque N is
1
DIPOLE-DIPOLE INTERACTIONS 2

(7) N = r+ F+ + r F
 
1 1
(8) = d (qE) + d (qE)
2 2
(9) = qd E
(10) = pE
Although this derivation isnt technically valid for an ideal dipole, since
the separation is zero there, we can work out a couple of problems assuming
its true.
Example 1. We have two perfect dipoles p1 and p2 separated by a dis-
tance r. Their alignment is such that p1 is perpendicular to the line separat-
ing them (pointing upwards) and p2 is parallel to the line separating them
(pointing away from p1 ). From the formula above, the field E12 due to p1
at p2 is

1
(11) E12 = [3(p1 r12 )r12 p1 ]
40 r3
1
(12) = p1
40 r3

since p1 r12 . The torque on p2 is therefore

1
(13) N2 = p2 p1
40 r3
1
(14) = p1 p2
40 r3
The field E21 due to p2 on p1 is

1
(15) E21 = [3(p2 r21 )r21 p2 ]
40 r3
1
(16) = [3p2 r21 p2 ]
40 r3
2
(17) = p2
40 r3

since p2 is anti-parallel to r21 . The torque on p1 is therefore

2
(18) N1 = p1 p2
40 r3
DIPOLE-DIPOLE INTERACTIONS 3

This torque is in the same direction as the first one, and twice the magni-
tude. This appears to violate Newtons third law, in that the torque exerted
by one dipole on the other is not equal and opposite to the inverse situation.
The problem is because the two torques are calculated relative to different
points, so were not comparing like with like. See here for a resolution of
this problem.
Example 2. A perfect dipole is situated a distance z above an infinite,
grounded, conducting plane. The dipole moment p makes an angle with
the perpendicular axis to the plane. What is the torque on the dipole?
Infinite grounded conducting planes make us think of the method of im-
ages. To see what the image of a dipole in a plane is, suppose its a physi-
cal dipole with a finite separation between the two charges. If the positive
charge is further away from the plane at a distance d1 , then in the image
this charge will be a distance d1 and be negative. Similarly, the negative
charge will have a positive image in the plane. The result is that dipoles
image has the same z-component but opposite components in the x and y
directions. That is, the angle between the dipole and its image is 2 .
The field at the dipole from its image p0 , using the formula above and
taking r in the +z direction is

1  0
3(p r)r p0

(19) E= 3
40 r
1  0

(20) = 3p cos r p
40 r3

The last line uses the fact that the magnitudes of the dipole and its image
are the same.
The torque is then

(21) N = pE
1 
3p cos p r p p0

(22) = 3
40 r
p2
(23) = (3 cos sin sin 2 ) x
40 r3
p2
 
1
(24) = sin 2 x
40 r3 2

Were assuming the coordinate system is aligned so that the x axis is


perpendicular to p and r.
DIPOLE-DIPOLE INTERACTIONS 4

Thus the torque is zero when = 0 or = /2, so the dipole feels no


torque when its axis is either perpendicular or parallel to the plane. How-
ever, the position parallel to the plane is an unstable equilibrium, since the
torque always tends to rotate the dipole so that it is perpendicular to the
plane.
P INGBACKS
Pingback: Dipole force and energy
Pingback: Dipole-dipole forces and torques
Pingback: Momentum in a magnetized and polarized sphere
METHOD OF IMAGES - MOVING CHARGE AND PLANE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.34.
As an example of the method of images, consider a point charge +q of
mass m that starts at rest a distance d from an infinite conducting plane.
How long will it take for the charge to hit the plane?
Weve seen that this setup can be replaced by an equal and opposite
charge q a distance d on the other side of the plane. The force on the
charge is therefore

q2
(1) F =
40 (4d 2 )
If we let r be the distance of the charge from the plane, and let it be a
function of time t, and use Newtons law we get (a dot above a variable
indicates a time derivative; a double dot is the second derivative):

q2
(2) mr =
40 (4r2 )
C
(3) r = 2
r
q2
(4) C
160 m
We therefore need to solve the differential equation

C
(5) r =
r2

subject to the initial conditions

(6) r(0) = d
(7) r(0) = 0
Solving the equation requires a trick, in that we multiply both sides by r:
1
METHOD OF IMAGES - MOVING CHARGE AND PLANE 2

r
(8) rr = C
r2
We need to integrate this from r = d to r = 0. We can integrate the LHS
by parts to get (using the initial condition r(0) = 0):

t t
0
t
(9) rrdt = r2 0 rrdt
0 0
t
1
(10) rrdt = r2 (t)
0 2

On the RHS, we get, also by integrating by parts and using r(0) = d:

t   t
r 1 1 r
(11) C 2
dt = C 2C 2
dt
0 r r(t) d 0 r
t  
r 1 1
(12) C 2
dt = C
0 r r(t) d

Combining the two sides, we get

 
1 2 1 1
(13) r (t) = C
2 r(t) d
1 1/2
 
1
(14) r(t) = 2C
r(t) d

r
d r
(15) = 2C
rd
Weve taken the negative square root on line 2 since r is decreasing with
time. Note also that this equation satisfies the initial condition r(0) = 0,
since r(0) = d.
We can separate the variables and get

0r t 0
rd
(16) dr = 2C dt
d d r 0

The integral on the left can be done with software, and we get
METHOD OF IMAGES - MOVING CHARGE AND PLANE 3

1
(17) d 3/2 = 2Ct
2
d 3/2
(18) t=
2 2C

d 3/2 4 0 m
(19) =
2 2q

(d)3/2 20 m
(20) =
q
METHOD OF IMAGES: POINT CHARGE AND TWO PLANES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.35.
Given two parallel, grounded, infinite conducting planes a distance a
apart, we place a charge +q between the plates, a distance x from one of
them. What is the force on the charge?
At first glance, this might seem easy, since you might think all we need
to do is create an image charge behind each of the two planes. However, we
need to maintain the potential on both planes at V = 0, and introducing an
image charge behind one plane messes up the potential on the other plane.
The only solution is to introduce an infinite sequence of images, where
each image counters the one introduced just before it. If we imagine the
two planes to be oriented vertically with our point of view looking at them
edge on. Then with the charge at position x, we place the first image of q
at distance x on the other side of the left plane (that is, a distance of 2x from
the charge). This image is a distance x + a from the right plane, so we need
to place an image of +q a distance x + a behind the right plane. This image
is a distance (a x) + x + a = 2a from the charge. This latest image is now
a distance of x + 2a from the left plane, so we need another image of q
a distance of x + 2a behind the left plane (a distance of 2x + 2a from the
charge). The process continues like this, and repeats for the right plane, so
we get a series of images as follows:
Images of q behind the left plane (to the left of the charge)
Distance from plane Distance to left of charge
x 2x
2a + x 2a + 2x
4a + x 4a + 2x
Images of +q behind the left plane:
Distance from plane Distance to left of charge
2a x 2a
4a x 4a
6a x 6a
Images of q behind the right plane (to the right of the charge)
1
METHOD OF IMAGES: POINT CHARGE AND TWO PLANES 2

Distance from plane Distance to right of charge


ax 2a 2x
3a x 4a 2x
5a x 6a 2x
Images of +q behind the right plane.
Distance from plane Distance to right of charge
a+x 2a
3a + x 4a
5a + x 6a
To work out the force, it is the distances of the images from the charge
that we must use. We will count a force as negative if it pulls the charge to
the left, and positive if it pulls it to the right. The forces from the positive
images cancel out, and from the negative images we get:

(1)
q2
 
1 1 1 1 1
F= ...+ + +...
40 (2x)2 (2a + 2x)2 (4a + 2x)2 (2a 2x)2 (4a 2x)2
I dont think this series has any closed form representation, so thats about
as simple as we can get the final answer.
We can test it with a couple of special cases. If a , then one of the
plates is taken off to infinity and were left with the simple case of a point
charge a distance x from a single conducting plane. In that case, all the
terms in the sum tend to zero except the first one, and we get

q2
(2) F
40 (2x)2
In the case where x = a/2 so the charge is exactly halfway between the
plates, the force should be zero by symmetry. Plugging this value into the
force series above, we get

q2
 
1 1 1 1 1
(3) F(a/2) = ...+ + +...
40 (a)2 (3a)2 (5a)2 (a)2 (3a)2
We see that the terms all cancel in pairs, so the result appears to be cor-
rect.
TWO CHARGED WIRES AND A CYLINDER

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.36.
Another example of the method of images. We are given two infinite
charged wires, one with linear charge density + and the other with .
These wires lie in the x y plane along the lines y = a. Between them is
an infinite conducting cylinder of radius R < a, with its axis on the x axis.
The cylinder carries no net charge. We are to find the potential everywhere
outside the cylinder.
The solution to this problem relies on an earlier example in which we
worked out the potential due to two charged wires on their own. We saw
there that the equipotential surfaces for two charged wires were circular
cylinders, so we should be able to set up an image configuration in which the
surface of the cylinder can be replaced by image wires inside the cylinder.
In the charged wires problem, we had two wires at y = a and found that
the potential could be written as

(y + a)2 + z2
(1) V= ln
40 (y a)2 + z2
We also found that for a constant potential K, if we define the value
A e40 K/ we could get expressions for the radius R and axis y0 of the
equipotential cylinders:

A+1
(2) y0 = a
A1
s
4a2 A
(3) R =
(A 1)2
A useful relation between these two is

(4) y20 R2 = a2
To apply the method of images and the above solution, we need to create
one image for each wire. Consider first the wire on the left, at y = a.
We want to create an image of this wire at y = b, where this location
1
TWO CHARGED WIRES AND A CYLINDER 2

is inside the cylinder. We can map this image problem onto the two-wire
problem by noting that the midpoint between the wire and its image is at
y = b (a b)/2. If we define y1 = y + b + (a b)/2, then the potential
due to this wire-image configuration is

(y1 + ab 2
2 ) +z
2
(5) V1 = ln
40 (y1 ab 2
2 ) +z
2

(y + a)2 + z2
(6) = ln
40 (y + b)2 + z2
On the other side, the wire at y = +a has an image at y = +b. Defining
y2 = y b (a b)/2, we get for the potential of this pair:

(y2 + ab 2
2 ) +z
2
(7) V2 = ln
40 (y2 ab 2
2 ) +z
2

(y b)2 + z2
(8) = ln
40 (y a)2 + z2
The only remaining problem is to find b. We can use 4 for this. Consider
the first wire-image pair. The distance y0 is the distance from the midpoint
between the wire and image to the axis of the cylinder, which is y0 = ab2 +
a+b
b = 2 . The quantity a in this relation is half the distance between the wire
and its image, so in this case must be replaced by ab2 . We therefore have

 2  2
a+b 2 ab
(9) R =
2 2
R2
(10) b =
a
Since potentials obey the superposition principle, we can just add the two
potentials from the two image-wire pairs to get the total potential:

(11) V = V1 +V2
(y + a)2 + z2 (y b)2 + z2
 

(12) = ln + ln
40 (y + b)2 + z2 (y a)2 + z2
(y + a)2 + z2 (y b)2 + z2
 

(13) = ln
40 (y + b)2 + z2 (y a)2 + z2
(y + a)2 + z2 (y R2 /a)2 + z2
 

(14) = ln
40 (y + R2 /a)2 + z2 (y a)2 + z2
TWO CHARGED WIRES AND A CYLINDER 3

To convert this to cylindrical coordinates, we can use y2 + z2 = r2 and


y = r cos :

(15)
r2 + a2 + 2ar cos r2 2R2 r cos /a + R4 /a2
 

V (r, ) = ln 2
40 r + 2R2 r cos /a + R4 /a2 r2 + a2 2ar cos
(16)
r2 + a2 + 2ar cos a2 r2 /R2 2ar cos + R2
 

= ln 2 2 2
40 a r /R + 2ar cos + R2 r2 + a2 2ar cos
Note that if r = R, then the argument of the logarithm becomes 1, so
V = 0 on the surface of the cylinder. This happens even though the potential
of each wire-image pair separately is not zero at r = R. In fact, for the first
pair, we have

a2
(17) V1 (R, ) = ln 2
40 R

and for the second pair, we get

R2
(18) V2 (R, ) = ln 2
40 a
Also, if = /2 or 3/2, V = 0. This corresponds to the xz plane, which
is the plane of symmetry in the problem.
LAPLACES EQUATION: CONDUCTING SPHERE AND SHELL
OF CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.37.
Another example of using the series solution to Laplaces equation to find
the potential using boundary conditions.
We have a conducting sphere of radius a maintained at a constant poten-
tial V0 . Concentric with this sphere is a larger spherical shell of radius b
with a surface charge density of ( ) = k cos . Find the potential in the
two regions (i) a r b and (ii) r > b.
The general solution in terms of Legendre polynomials is

 
l Bl
(1) V (r, ) = Al r + l+1 Pl (cos )
l=0 r
In the inner region, we must use the most general form of the solution,
since r tends neither to zero nor infinity. In the outer region, to keep the
potential finite, we dispose of the terms in rl , so we have


C
(2) Vout (r, ) = rl+1l Pl (cos )
l=0
Note that we must use a different set of coefficients since the potential
will have a different functional form outside the shell.
We can now apply the various boundary conditions. First, at r = a we
must have V = V0 so

 
l Bl
(3) Al a + al+1 Pl (cos ) = V0
l=0
From the orthogonality of the Pl , only the l = 0 term on the left is non-
zero, so we get

B0
(4) A0 + = V0
a
(5) Bl = a2l+1 Al (l > 0)
1
LAPLACES EQUATION: CONDUCTING SPHERE AND SHELL OF CHARGE 2

Next, we look at r = b. Here, the potential must be continuous, so

 
Bl C
(6) A b l
l bl+1 l + P (cos ) = rl+1l Pl (cos )
l=0 l=0
Equating coefficients of the Pl we get

Bl Cl
(7) Al +
l+1
= l+1
b b
Using the above relation between Al and Bl we get, first for l = 0:

B0 C0
(8) A0 + =
b b

and for l > 0:

a2l+1
 
l Cl
(9) Al b l+1 = l+1
b b
 
(10) Cl = Al b2l+1 a2l+1
Finally, we use the surface charge density on the shell. From our earlier
example of this type, we get

V V
(11) =
r out r in
0
 
Cl Bl k
(12) (l + 1) l+2 lAl bl1 + (l + 1) l+2 Pl (cos ) = cos
l=0 b b 0
Since P1 (cos ) = cos , we can equate coefficients on both sides to get

C0 B0
(13) + =0
b2 b2
C1 B1 k
(14) 2 3 A1 + 2 3 =
b b 0
Cl Bl
(15) (l + 1) l+2 lAl bl1 + (l + 1) l+2 = 0 l>1
b b
The first of these equations gives us

(16) B0 = C0
Substituting this into 8 gives us A0 = 0; B0 = aV0 = C0 .
LAPLACES EQUATION: CONDUCTING SPHERE AND SHELL OF CHARGE 3

The second equation can be reduced by substituting for B1 and C1 in


terms of A1 from the above relations:
 3
b a3 a3

k
(17) 2 + 1 + 2 A1 =
b3 b3 0
k
(18) A1 =
30
This gives us, from the above relations

a3 k
(19) B1 =
30
k
b3 a3

(20) C1 =
30
The third equation can be solved by taking Al = Bl = Cl = 0 for all l > 1.
Since the solution to Laplaces equation is unique, this must be the solution.
Putting all this together, we get:
 3

aV0 + k
r ar2 cos arb
r 30
(21) V (r, ) =  3 3
aV
0 k b a
r + 30 r2
cos r>b
The induced surface charge on the conductor can be found from the nor-
mal derivative at the surface r = a (the potential inside the conductor is
constant, so its derivative is zero):

V
(22) =
0 r r=a
V0 k
(23) = + cos
a 0
V0 0
(24) = k cos
a
If the sphere is grounded so that V0 = 0, then the induced charge is just
the negative of the charge on the shell. The total charge in the system is
thus the surface area of the conducting sphere times the constant term:

V0 0
(25) Q = 4a2
a
(26) = 40 aV0
For large distances, from the above formula
LAPLACES EQUATION: CONDUCTING SPHERE AND SHELL OF CHARGE 4

aV0
(27) V
r
This is consistent with the total charge, since the potential for a point
charge is

Q
(28) V =
40 r
aV0
(29) =
r
LAPLACES EQUATION - CHARGED LINE SEGMENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.38.
An example of using the general series solution to Laplaces equation.
We have a linear charge density along the z axis between z = a and
z = +a. Find the potential for any position where r > a.
We can approach this problem using the same technique as we used for
finding the potential due to a charged disk. We worked out the potential for
points in the xy plane earlier (see Example 2 in this earlier post). Using the
coordinates suitable for the current problem, we have

2 h  p
2 2
 i
(1) V (r, /2) = ln a + r + a ln r
40
" r #
a a2
(2) = ln + 1+ 2
20 r r

We can expand the log term in powers of 1/r and get

a 1 a3 3 a5
 

(3) V (r, /2) = + +...
20 r 6 r3 40 r5

This series must match the general form of Laplaces equation for r > a.
This consists entirely of the terms in inverse powers of r, since we require
the potential to be finite for large r. That is, we must have


B
(4) V (r, ) = rl+1l Pl (cos )
l=0

By comparing the series, we see that terms in odd l must vanish, so Bl = 0


if l is odd. The second series must match the earlier one for the special case
of = /2, that is when cos = 0. Looking up tables of the Legendre
polynomials, we find for the first few:
1
LAPLACES EQUATION - CHARGED LINE SEGMENT 2

(5) P0 (0) = 1
1
(6) P2 (0) =
2
3
(7) P4 (0) =
8
Therefore, we have

B0 B2 3 B4
(8) V (r, /2) = 3 + 5 +...
r 2r 8r
Comparing the two series, we get

a
(9) B0 =
20
a3
(10) B2 =
20 3
a5
(11) B4 =
20 5
The general solution is then

a a3 a5
 

(12) V (r, ) = + P2 (cos ) + 5 P4 (cos ) + . . .
20 r 3r3 5r
a2 a4
 
a
(13) = 1 + 2 P2 (cos ) + 4 P4 (cos ) + . . .
20 r 3r 5r
We can write this in terms of the total charge in the line segment, which
is Q = 2a :

a2 a4
 
Q
(14) V (r, ) = 1 + 2 P2 (cos ) + 4 P4 (cos ) + . . .
40 r 3r 5r
LAPLACES EQUATION - CYLINDRICAL SHELL WITH
OPPOSING CHARGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.39.
As another example of applying the solution to Laplaces equation in
cylindrical coordinates, we consider the following problem. We are given
a cylindrical non-conducting shell or radius R carrying a charge density of
0 on its upper half ( < < 0) and 0 on its lower half (0 < < ).
Find the potential everywhere.
We can begin in the same manner as for the other problem involving a
cylindrical shell that we solved earlier. The solution is the same up until the
point where we introduce the surface charge.
Thus, outside the shell, we have

 
Dn Cn
(1) Vout = Bout + n
cos n n sin n
n=1 r r
Inside the shell, we have


(2) Vin = Bin + [An rn sin n + Bn rn cos n ]
n=1
Since the potential is continuous over a surface charge, we must have
Vout (R) = Vin (R), so we get

(3)  

Dn Cn
Bout + cos n sin n = B in + [An Rn sin n + Bn Rn cos n ]
n=1 Rn Rn n=1
Equating coefficients of the sine and cosine, we get

(4) Bout = Bin


(5) Cn = An R2n
(6) Dn = Bn R2n
The outward derivative of the potential is discontinuous across a surface
charge, and we have
1
LAPLACES EQUATION - CYLINDRICAL SHELL WITH OPPOSING CHARGES 2


V V
(7) =
r out r in
0
Plugging in the formulas for Vout and Vin , we get

(8) (
0
nR2n An 2n B < < 0
   
n1 nR n n1
nR An sin n + nR Bn cos n = 0 0
n=1 Rn+1 n=1 Rn+1 0< <
0

Since the surface charge is an odd function of , we can eliminate the


cosine terms, since the cosine is an even function. Therefore, we have Bn =
Dn = 0. We are free to choose the potential at infinity to be any constant, so
we might as well take it to be zero, in which case we have Bin = Bout = 0.
We are therefore left with, after simplifying the term in brackets:
(
0

n1 0 < < 0
(9) 2 nR An sin n =
n=1 00 0< <
To find the An , we can use the fact that the set of sin n functions is
orthogonal over the interval [, ]. That is


(
0 n 6= m
(10) sin m sin n d =
n=m
Therefore we can multiply both sides by sin m and integrate to get
 0 
m1 0
(11) 2mR Am = sin m d sin m d
0 0
The integrals in brackets on the right come out to

 0
(

0 m even
(12) sin m d sin m d =
0 m4 m odd
Therefore (changing the index from m to n for convenience):
(
0 n even
(13) An = 20
0 n2 Rn1
n odd

We thus get
LAPLACES EQUATION - CYLINDRICAL SHELL WITH OPPOSING CHARGES 3

(
0 n even
(14) Cn = Rn+1
20 2 n odd
0n

The final formula for the potential is


( 2
r n
0
0

n odd n2 Rn1 sin n r<R
(15) V (r, ) = 20 Rn+1
0
n odd n2 rn sin n r>R
MULTIPOLE EXPANSION OF LINEAR CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.40.
This is a rather specialized example of the multipole expansion. We are
given a linear charge distribution on the z axis between z = a and z = +a.
For each of three such distributions, find the leading term in the multipole
expansion.
We saw in the post where we derived the multipole expansion that


1 1
(1) V (r) = rn+1 (r0 )Pn (cos 0 )r0n d 3 r0
40 n=0

where is the volume charge density and Pn is a Legendre polynomial.


When using this formula, it is crucial to understand what the vectors r
and r0 , and the angle 0 refer to. The vector r is the vector from the origin
to the point at which we wish to calculate the potential. It can be in the
region where the charge density is non-zero or outside it. The integration
variable r0 is the vector from the origin to the volume element at which
we measure the charge density (r0 ). In practice, r0 extends only over that
region of space where there is charge, that is, where (r0 ) 6= 0. Finally, the
angle 0 is the angle between r and r0 . Inside the integral, 0 varies as r0
varies over the region where there is charge, and in practice we would need
to work out the relation between 0 and r0 in order to do the integral.
For the special case in this example, however, we can simplify things a bit
by considering the geometry of the situation. Since the charge lies entirely
on the z axis between z = a and z = +a, then the angle 0 is the same for
all charge lying in the interval 0 < z +a. For charge in the interval a
z < 0, the angle between r and r0 is 0 . Since cos( 0 ) = cos 0
we have for the Legendre polynomials in the region z < 0

(2) Pn (cos( 0 )) = Pn ( cos 0 )


(3) = (1)n Pn (cos 0 )

since Pn is even (odd) if n is even (odd).


1
MULTIPOLE EXPANSION OF LINEAR CHARGE 2

Now if we remember that r0 is the vector from the origin to a charge


element, then in the case where all the charge resides on the z axis, we have
r0 = |z|. For z < 0, we therefore have (r0 )n = (1)n zn . If we combine these
two results, then we see that for z < 0

(4) Pn (cos( 0 ))r0n = (1)n Pn (cos 0 )(1)n zn


(5) = Pn (cos 0 )zn
That is, due to the two negatives cancelling out, the product of these two
terms is the same over the entire interval a < z < a.
For a given value of r (that is, for a given observation point), 0 is constant
over the entire charge distribution, so we can take Pn (cos 0 ) outside the
integral, and we end up with

a
1 Pn (cos )
(6) V (r) = rn+1 zn (z)dz
40 n=0 a

where weve used the more traditional symbol to denote a linear charge
density, and weve dropped the prime off since this variable depends only
on r and not on the charge location.
Note that this formula works only for the special case where we have a
linear charge distributed along the z axis. Thats why we said that this a
very specialized situation.
Now we need only specify (z) and work out the integrals. Well con-
sider three cases, and find only the first non-zero term in each case.
(a) = k cos(z/2a). In this case the leading term is n = 0, since

a
4ka
(7) k cos(z/2a)dz =
a

so

1 P0 (cos ) 4ka
(8) V0 =
40 r
1 4ka
(9) =
40 r
In this case, the monopole term dominates.
(b) = k sin(z/a). In this case the total charge is zero, so we look at
the dipole (n = 1) term.
MULTIPOLE EXPANSION OF LINEAR CHARGE 3

a a
(10) z (z)dz = k z sin(z/a)dz
a a
2ka2
(11) =

so

1 P1 (cos ) 2ka2
(12) V1 =
40 r2
1 2ka 2
(13) = cos
40 r2
(c) = k cos(z/a). In this case, the monopole and dipole terms are both
zero, so we look at the quadrapole term.
a a
2
(14) z (z)dz = k z2 cos(z/a)dz
a a
4ka3
(15) =
2

so

1 P2 (cos ) 4ka3
(16) V2 =
40 r3 2
1 2ka3
(17) = (3 cos2 1)
40 2 r3
AVERAGE FIELD OVER A SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.41.
The average electric field inside a sphere of radius R due to a point charge
q at position r inside the sphere is the field integrated over the volume of
the sphere divided by the spheres volume:

1 3q r0 r
(1) Eav = d 3 r0
40 4R3 |r0 r|3
where the integral extends over the interior of the sphere. (Note that the
statement of the problem in Griffiths book has a typo: the unit vector in the
integral in part (a) should be script r, not r.)
Now suppose we have a uniformly charged sphere with charge density
and wish to find the field at point r due to this charge. This time the field at
r due to volume element d 3 r0 is (r r0 ) d 3 r0 /40 |r0 r|3 so the overall
field is

1 r r0
(2) E = d 3 r0
40 |r r0 |3
The two fields are the same if we set

3q
(3) =
4R3
From Gausss law, we can work out E if we work out the integrals below
for a sphere of radius r < R:

1
(4) E da = d 3 r0
0
1 3q 4r3
(5) 4r2 E =
0 4R3 3
1 qr
(6) E =
40 R3
Since E points in the radial direction due to symmetry,
1
AVERAGE FIELD OVER A SPHERE 2

1 q
(7) E = r
40 R3
The dipole moment of a single point charge q at position r is p = qr, so
the field can be written as

1 p
(8) E = Eav =
40 R3
From the superposition principle, we can extend this result so that it ap-
plies to any distribution of charge within the sphere.
If r is outside the sphere, the formula for Eav is the same as before, with
the integral still extending over the interior of the sphere. The formula for
E is also the same, but this time if we use Gausss law and integrate over
a spherical surface of radius r > R, we get

1
(9) E da = d 3 r0
0
1 3q 4R3
(10) 4r2 E =
0 4R3 3
1 q
(11) E =
40 r2
1 q
(12) E = r
40 r2
The RHS of the second line arises from the fact that all of the charged
sphere is now interior to the surface of integration, while on the LHS, we
are still integrating over a spherical surface of radius r > R. The average
field produced by a point charge is thus the same as the field produced by
this charge at the centre of the sphere. By superposition we can apply this
argument to any distribution of charge external to the sphere.
P INGBACKS
Pingback: Electric field of a dipole
Pingback: Relation between polarizability and susceptibility
Pingback: Average magnetic field within a sphere
ELECTRIC FIELD OF A DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 3.4.4 & Problem 3.42.
Weve seen that the potential of a pure dipole with dipole moment p (that
is, a system of charge where only the dipole term in the multipole expansion
is non-zero) is, in spherical coordinates:

1
(1) V = p r
40 r2
1
(2) = p cos
40 r2
By taking the gradient in spherical coordinates, we can find the electric
field of a dipole, since E = V .

V 1 V 1 V
(3) E= r
r r r sin
p  
(4) = 2 cos r + sin
40 r3
We can calculate the average field of a dipole over a sphere of radius
R but to do this we need to express the spherical unit vectors in terms of
rectangular unit vectors, since the spherical unit vectors chage with position.
We have

(5) r = sin cos x + sin sin y + cos z


(6) = cos cos x + cos sin y sin z
Substituting these into the field and collecting terms we get

(7)
p
E= [3 sin cos cos x + 3 sin cos sin y + (2 cos2 sin2 )z]
40 r3
If we integrate this over the sphere, doing the angular integrals first, we
get zero, since (remember we need to multiply by the spherical volume
element r2 sin d d dr):
1
ELECTRIC FIELD OF A DIPOLE 2

2 2
(8) cos d = sin d = (2 cos2 sin2 ) sin d = 0
0 0 0
R
However, the integral over r is infinite, since 0 (1/r)dr is infinite at the
lower limit. We are thus left with an infinity times zero situation, so the
answer is not well defined.
Weve seen that the average field of a charge distribution within a sphere
is

1 p
(9) Eav =
40 R3
The problem clearly arises at the point r = 0, since if we consider inte-
grating the expression for dipole field above from some infinitesimal sphere
r = out to r = R, we will always get zero due to the angular component.
If we accept that the expression for the average field is valid, then we can
patch things up by requiring the field to have a delta-function component at
r = 0. That is, if we say that there is a term

1
(10) E0 = p 3 (r)
30

then the average field comes out right since



3 1 p
(11) 3
E0 d 3 r =
4R 40 R3
We cant really rewrite the total field as

p  1
p 3 (r)

(12) E= 3
2 cos r + sin
40 r 30

since the first term applies only at r 6= 0. As always with delta functions,
the physical interpretation is a bit dodgy.
P INGBACKS
Pingback: Dipole-dipole interactions
Pingback: Point charge and neutral atom
Pingback: Energy in a dielectric
Pingback: Dipole-dipole forces and torques
Pingback: Magnetic dipole: average field over a sphere
GREENS RECIPROCITY THEOREM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 3.43 & 3.44.
There is an interesting theorem that relates two separate charge distri-
butions. Suppose we have a charge distribution 1 with its associated po-
tential V1 , and a completely separate charge distribution 2 with potential
V2 . These two distributions do not co-exist; they are completely separate
situations.
Now consider the electric fields E1 and E2 produced by these two distri-
butions. We can consider the following integral, taken over all space:

3
(1) E1 E2 d r = V1 E2 d 3 r

Lets consider the first term in the dot product, and use integration by
parts:

V1 E2x
(2) E2x dxdydz = V1 E2x |all x dydz + V1 dxdydz
x x
If we make the usual assumption that the potential V1 vanishes at infinity
then the integrated term is zero. Doing similar integrals for the other terms
in the dot product (integrating with respect to y and then z first for the second
and third terms respectively) gives us:

3
(3) V1 E2 d r = V1 E2 d 3 r

1
(4) = V1 2 d 3 r
0

where in the last line we used Gausss law relating the field to the charge
distribution.
We could just as well have done the same calculation interchanging the
subscripts 1 and 2, so we get

3
(5) V1 2 d r = V2 1 d 3 r
1
GREENS RECIPROCITY THEOREM 2

which is Greens reciprocity theorem.


Now for a few examples of its use.
Example 1. Suppose we have two conductors, each of which can be of
arbitrary shape and location. First, we place a charge Q on conductor 1,
which induces a potential V12 on conductor 2 (which has no net charge).
Second, we do the opposite: we place the same charge Q on conductor 2,
which induces a potential V21 on conductor 1 (which now has no net charge).
Since the potential on a conductor is always constant, we can use the
reciprocity theorem to say


3
(6) V12 2 d r = V21 1 d 3 r

The two integrals are the same, and give the total charge Q, so we can
conclude that V12 = V21 . Note that this result does not depend on the shape
or location of the conductors; its a universal result.
Example 2. We have two parallel infinite conducting planes, both of
which are grounded. The distance between the plates is d. We place a point
charge q between the plates at a distance x from plate 1 (which we take to
be the left plate). Find the total charge induced on each plate.
To apply the reciprocity theorem, we need two distinct charge distribu-
tions. For the first, we can take the system as described. For the second, we
can remove the charge q and also remove the condition that the plates are
grounded, so each plate can be at a different potential.
First, consider the distribution as given. Let the potential of the left plate
be Vl and of the right plate be Vr . Since the two plates are grounded, we
have Vl = Vr = 0. Also, since the plates are grounded, the induced charge
must cancel out the point charge so there is zero net charge in the system.
That is Ql + Qr = q.
Now consider the distribution without the point charge q. In this case we
can take the potential of the left plate to be Vl0 = 0 and of the right plate
to be Vr0 = V0 . Note that this assumes the right plate is not grounded. This
doesnt matter, since the second distribution can be anything we like. We
assume that the plates here have total charges Q0l and Q0r , although well see
we dont need these values anyway.
Since the second distribution contains no point charge, the potential varies
linearly between the two plates, so the potential at position x is Vx0 = Vl0 +
V0 x/d = V0 x/d. Now were ready to apply the reciprocity theorem. The
charge density 2 consists only of the charge on the two plates, since weve
removed q. We have, on one side:
GREENS RECIPROCITY THEOREM 3


(7) V1 2 d 3 r = Vl Q0l +Vr Q0r
(8) =0

since Vl = Vr = 0.
On the other side, we have

(9) V2 1 d 3 r = Vl0 Ql +Vx0 q +Vr0 Qr
 qx 
(10) = V0 + Qr
d

From the theorem, this must be zero, so we get

qx
(11) Qr =
d
qx
(12) Ql = q +
 d x
(13) = q 1
d
Note that the reciprocity theorem in this case allows us to calculate only
the total charge on each plate; finding the actual surface charge density is a
considerably harder problem.
Example 3. We have two concentric spherical conductors of radii a and
b > a, and a point charge q between them at a location r such that a < r < b.
Again assuming the spheres are grounded, find the total induced charge on
each sphere.
Using similar notation to the last example, we again consider the two
distributions to be the original configuration (with the charge q) and a con-
figuration without q. In the first case, since the conductors are grounded,
we have

(14) Va = Vb = 0
In the second case, we can take Va0 = V0 . In this case, since we dont have
the charge q, the system has spherical symmetry, so any charge distributed
over the spheres must be uniform, so the potential and the field are the same
as if the charge were concentrated at the centre of the spheres. This means
that the potential between the spheres has a 1/r dependence, so we can write
GREENS RECIPROCITY THEOREM 4

(15) Va0 = V0
a
(16) Vr0 = V0
r
0 a
(17) Vb = V0
b
Applying the reciprocity theorem, we get

(18) V1 2 d 3 r = Va Q0a +Vb Q0b
(19) =0
On the other side, we have

(20) V2 1 d 3 r = Va0 Qa +Vr0 q +Vb0 Qb
 a a 
(21) = V0 Qa + q + Qb
r b
Again, since the two spheres are grounded in the first configuration, we
must have Qa + Qb = q, so we get

3
 a a 
(22) V2 1 d r = V0 q Qb + q + Qb
r b
(23) =0
Solving this for Qb we get

(r a)b
(24) Qb = q
(b a)r
(25) Qa = q Qb
(b r)a
(26) = q
(b a)r
QUADRUPOLE MOMENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.45.
(a) Weve seen that the dipole term in the multipole expansion can be
written as

1 p r
(1) Vdip =
40 r2
3
1
(2) = ri pi
40 r2 i=1

where p is the dipole moment and the subscript i indicates the ith component
of the corresponding vector. In particular


(3) pi = ri0 (r0 )d 3 r0

We can use a similar derivation to define a quadrupole moment. The


quadrupole term in the multipole expansion is


1 1
(4) Vquad = r02 P2 (cos 0 )(r0 )d 3 r0
40 r3
 
1 1 02 3 2 0 1
(5) = r cos (r0 )d 3 r0
40 r3 2 2

The angle 0 is the angle between the vectors r0 and r, so r0 cos 0 = r0 r


where the hat denotes a unit vector. We can therefore write the quadrupole
term as


1 1 2 
(6) Vquad = 3 r0 r r0 r0 (r0 )d 3 r0
40 2r3
In terms of components
1
QUADRUPOLE MOMENT 2

3
(7) r0 r = ri0ri
i=1
0 0 02
(8) r r = r

so we can write

3
!
3
! !
1 1
(9) Vquad =
40 2r3
3 ri0ri r0j r j r02 (r0 )d 3 r0
i=1 j=1
3
1 1
3ri0 r0j r02 i j (r0 )d 3 r0

(10) = ri r j
40 2r3 i,
j=1

1 1 3
(11) ri r j Qi j
40 2r3 i,
j=1

where in the second line i j is the usual Kronecker delta symbol and we
used the fact that

3
(12) ri r j i j = 1
i, j=1

since r is a unit vector.


The quadrupole moment is defined by

2 
(13) Qi j 3r0i r0j ri0 i j (r0 )d 3 r0

The quadrupole moment is a second-rank tensor, which can be repre-


sented by a 3 3 symmetric matrix.
(b) As an example, consider a configuration of four point charges at the
corners of a square of side a in the xy plane, with +q at locations (a/2, a/2)
and (a/2, a/2), and q at (a/2, a/2) and (a/2, a/2). Then we have

a2
(14) Q11 = q (2 + 2 2 + 2)
4
(15) = 0
We can work out the other components in the same way and get
QUADRUPOLE MOMENT 3

3qa2 0

0 12 0 0
a2
(16) Q = q 12 0 0 = 3qa2 0 0
4 0 0 0 0 0 0
(c) How does the quadrupole moment depend on the choice of origin?
Suppose we shift the origin by an amount a. Then

2 
Q0i j 3 ri0 + ai r0j + a j ri0 + ai i j (r0 + a)d 3 r0
 
(17) =

2 
(18) = 3r0i r0j ri0 i j (r0 + a)d 3 r0 +

(19) 0 0 3 0
3ai r j (r + a)d r + 3a j ri0 (r0 + a)d 3 r0

(20) 2ai i j ri0 (r0 + a)d 3 r0 +

(21) 0 3 0
3ai a j (r + a)d r ai i j (r0 + a)d 3 r0
2

The first term in the expanded integral (line 2) is actually just Qi j , since
we are multiplying the charge at location r0 +a (which in the old coordinates
was at r0 ) by the old coordinates. By the same argument, the three terms in
lines 3 and 4 are all constants multiplied by the dipole moment, and the two
terms in line 5 are constants multiplied by the monopole moment (which is
just the total charge). Thus if both the dipole and monopole moments are
zero, then the quadrupole moment is independent of the origin.
(d) We can use the same approach to define higher order moments. For
example, the next moment in the series is the octopole moment. To define
it, we start with the octopole term in the expansion:

1 1
(22) Voct = r03 P3 (cos 0 )(r0 )d 3 r0
40 r4

1 1 03 3 0 0
(r0 )d 3 r0

(23) = r 5 cos 3 cos
40 2r4
For the first term, we get
3
(24) 5r03 cos3 0 = 5 r0 r
3
(25) =5 ri r j rk ri0 r0j rk0
i, j,k=1
For the second term, we have
QUADRUPOLE MOMENT 4

(26) 3r03 cos 0 = 3r02 r0 cos 0


(27) = 3r02 r0 r
3
(28) = 3r02 ri0 ri
i=1
3
(29) = 3r02 ri r j rk jk ri0
i, j,k=1

where in the last line we used 12 above.


Putting these two expansions back into the expression for Voct we get

3
1 1
5ri0 r0j rk0 3r02 ri0 jk (r0 )d 3 r0

(30) Voct = ri r j rk
40 2r4 i, j,k=1
We can therefore define an octopole moment as the integral term:

5ri0 r0j rk0 3r02 ri0 jk (r0 )d 3 r0

(31) Oi jk =

Note that this isnt the only way we could have defined it, since the sec-
ond term above could be written in various ways. It is probably more tra-
ditional to write it so that it is symmetric with respect to all three of its
indexes. That is, we could write it as

3
3r03 cos 0 = r02 ri r j rk jk ri0 + ik r0j + i j rk0
 
(32)
i, j,k=1

which makes the octopole moment come out to



5ri0 r0j rk0 r02 jk ri0 + ik r0j + i j rk0 (r0 )d 3 r0
 
(33) Oi jk =

The two forms do have different components, but they are equally valid
since when placed into the original sum they both give the same potential.
METHOD OF IMAGES - SPHERE IN UNIFORM FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 3.46.
The method of images can be used to find the potential of the configu-
ration where we have a grounded conducting sphere in a uniform electric
field (uniform except around the sphere, of course, where the induced sur-
face charge will distort the field).
Weve seen that in the case of a single charge q outside a sphere of radius
R, we have


q 1 1
(1) V (r) = q
40 2 2
r + a 2ar cos 2 2
(ar/R) + R 2ra cos

where the charge is assumed to be on the z axis at z = a and is the usual


spherical angle. This solution was found by introducing an image charge
q0 = Rq/a at location z = R2 /a, and we showed in the earlier post that
these two charges produce V = 0 at r = R, so the boundary condition is the
same as in the case of a grounded conducting sphere.
We can now introduce another point charge q at location z = a, and
its image q00 = Rq/a at location z = R2 /a. Since this is just the mirror
image of the above configuration, it too will maintain the potential of the
sphere at zero, so the boundary condition remains unchanged. The potential
due to these two charges is identical to that of the first two, except that the
angle is now . Since cos( ) = cos , we get
1
METHOD OF IMAGES - SPHERE IN UNIFORM FIELD 2

(2)

q 1 1
V (r) = q
40 2 2
r + a 2ar cos 2
(ar/R) + R2 2ra cos
(3)

q 1 1
+ +q
40 r2 + a2 + 2ar cos 2
(ar/R) + R2 + 2ra cos
(4)
" #
q 1 1
= p p
40 a 1 + r2 /a2 2 (r/a) cos 1 + r2 /a2 + 2 (r/a) cos
(5)
" #
q R 1 R 1
+ p + p
40 a r 1 + R /(ar) 2 (R /ra) cos r 1 + R /(ar) + 2 (R2 /ra) cos
4 2 2 4 2

The idea now is to let a so the external charges become very far
away from the sphere, which makes the electric field due to them essentially

uniform. In this approximation, we can use the expansion of 1/ 1 + x
1 x/2 to get

R3
 
q r
(6) V (r) 2 cos 2 2 cos
40 a a r a
2q

R 3 
(7) = r cos
40 a2 r2
For r , we require that the electric field is uniform, with a value of
E0 say. For large r the second term in the above expression can be ignored,
and we get

2q
(8) V (r) r cos
40 a2
2q
(9) = z
40 a2
Since the field is in the z direction, we have E0 = V / z = 42qa2 .
0
That is, this quantity must remain constant as a , so q must also become
infinitely large. This makes sense, since if the charges are infinitely far
away, they must be infinitely large to provide a finite field.
LAPLACES EQUATION - FOURIER SERIES EXAMPLE 4

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 3.47.
Here is another example of solving Laplaces equation in rectangular co-
ordinates. We have a rectangular pipe extending in the z direction with
boundaries at x = b, y = 0 and y = a. The potential on each face is held
constant and satisfies



0 y=0

V
0 y=a
(1) V=

0 x = b

0 x=b

The problem is to find V (x, y) everywhere inside the pipe. We can use
the separation of variables technique to arrive at a solution of form:

 
(2) V (x, y) = Aekx + Bekx (C sin ky + D cos ky)

Because the problem is symmetric in x we must have V (x, y) = V (x, y),


which means that A = B and we can write

(3) V (x, y) = (C sin ky + D cos ky) cosh kx

where weve absorbed the constants A and B into C and D.


The boundary condition at y = 0 means that D = 0. At first glance we
cant do much with the boundary conditions at the other three faces. Lets
leave this solution for the moment and try a different approach.
In the original derivation of the separation of variables solution, we had
the equations
1
LAPLACES EQUATION - FOURIER SERIES EXAMPLE 4 2

1 d2X
(4) = C1
X dx2
1 d 2Y
(5) = C2
Y dy2
1 d2Z
(6) = C3
Z dz2
We made the implicit assumption that the constants were non-zero, re-
quiring only that their sum is zero so as to satisfy the full Laplaces equa-
tion:

(7) C1 +C2 +C3 = 0


In this problem, though, we can use a trick to get the solution. Since the
problem doesnt depend on z, we can consider the two equations

d2X
(8) = 0
dx2
d 2Y
(9) = 0
dy2
The general solution of this pair of equations is linear in each of x and y
separately in order for the second derivatives to be zero. Thus we get

(10) X(x) = c1 x + c2
(11) Y (y) = c3 y + c4

and the overall solution is thus

(12) V = (c1 x + c2 ) (c3 y + c4 )


Applying the boundary conditions on y we get

(13) V (x, 0) = c4 (c1 x + c2 )


(14) = 0
This must be true for all x so either c4 = 0 or c1 = c2 = 0. If we choose
the latter case, then V = 0 everywhere, which isnt much use, so well take
c4 = 0.
At the other boundary, we have
LAPLACES EQUATION - FOURIER SERIES EXAMPLE 4 3

(15) V (x, a) = c3 a (c1 x + c2 )


(16) = V0
Again, this must be true for all x, so c1 = 0 and c2 c3 = V0 /a. Thus a
solution that satisfies the y boundary conditions is

V0
(17) V (x, y) = y
a
Clearly this doesnt satisfy the x boundary conditions, but we can now
return to our earlier solutions to the separation of variables problem to see
how to fix this. If we find a solution that satisfies the new boundary condi-
tions


0 y=0

0 y=a
(18) V=

Va0 y x = b
V0

ay x=b

then we can add this solution to the solution we just found. This new solu-
tion will satisfy all four boundary conditions, since the sums of the poten-
tials at each of the four boundaries come out to what was originally spec-
ified. We can now use the usual technique of building an infinite series of
solutions and using the orthogonality of the sine functions to work out the
coefficients. That is, returning to our first solution, we have, after satisfying
the boundary condition at y = 0:

(19) V (x, y) = C sin ky cosh kx


If we now require this solution to be zero at y = a, then we must have

n
(20) k=
a

for n = 1, 2, 3, . . .. We can then construct the sum of these terms and com-
bine it with the other solution we found to give


V0  ny   nx 
(21) V (x, y) = y + Cn sin cosh
a n=1 a a
To satisfy the x boundary conditions we must have
LAPLACES EQUATION - FOURIER SERIES EXAMPLE 4 4

 ny   
nb V0
(22) Cn sin cosh = y
n=1 a a a

for 0 < y < a. We now multiply both sides of this equation by sin my
a and
integrate from 0 to a, using the orthogonality of the sine, that is

a
(
my ny 0 n 6= m
(23) sin sin = a
0 a a 2 n=m
Therefore

a nb V0 a ny
(24) Cn cosh = y sin dy
2 a a 0 a
(1)n a2
 
V0
(25) =
a n
n
(1) aV0
(26) =
n
2(1)nV0
(27) Cn =
n cosh(nb/a)
The final form of the potential is

V0 2V0 (1)n sin (ny/a) cosh (nx/a)


(28) V (x, y) = y+
a n=1 n cosh(nb/a)

P INGBACKS
Pingback: Thompson-Lampard theorem
THOMPSON-LAMPARD THEOREM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 3.48a.
In the previous post, we worked out the potential inside a rectangular pipe
with three grounded sides and the fourth side held at a constant potential of
V0 . The boundary conditions are


0 y=0

V
0 y=a
(1) V=

0 x = b

0 x=b

The answer is

V0 2V0 (1)n sin (ny/a) cosh (nx/a)


(2) V (x, y) = y+
a n=1 n cosh(nb/a)
We can now consider a more specialized version of this problem in which
the pipe has a square cross-section, so that b = a/2. In this case, we have

V0 2V0 (1)n sin (ny/a) cosh (nx/a)


(3) V (x, y) = y+
a n=1 n cosh(n/2)
We can find the charge density on the face opposite the V0 face (that is,
at y = 0) by using the usual normal derivative formula:

V
(4) = 0
n
V
(5) = 0
y y=0
0V0 20V0 (1)n cosh (nx/a)
(6) = cosh(n/2)
a a n=1
We can work out from this the charge per unit length for this face in the
pipe. Since the charge density depends only on x we can integrate x from
a/2 to a/2 to get
1
THOMPSON-LAMPARD THEOREM 2

a/2
(7) = (x)dx
a/2
20V0 (1)n 2 sinh (n/2)
(8) = 0V0 n cosh(n/2)
n=1
" #
4 (1)n  n 
(9) = 0V0 1 + tanh
n=1 n 2
Although I couldnt find any tricky mathematical way of summing the
series into a closed form, the geometry of the problem is an instance of the
Thompson-Lampard theorem, which applies to any case of cross capaci-
tors; that is two pairs of opposing plates such as we have in this problem.
The theorem states that in such a case

(10) eC1 /0 + eC2 /0 = 1

where C1 and C2 are the two capacitances per unit length. In our problem,
because of the symmetry of the configuration, C1 = C2 C and the formula
can then be solved for C:

0
(11) C= ln 2

Since the capacitance is defined as C = Q/V where Q is the charge per
unit length on the plates and V is the potential difference between them, we
have

0V
(12) Q= ln 2

If we consider a unit length of the rectangular pipe and look at the pair of
plates whose potential differs by V0 we get

0V0
(13) Q= = ln 2

This is the charge on the plate with positive potential, so the charge on
the opposite plate is the negative of this.
The series above converges quite slowly since for large n tanh (n/2) 1
and the alternating harmonic series (1)n /n converges very slowly. How-
ever, we can convert the formula by expressing tanh in its exponential form:
THOMPSON-LAMPARD THEOREM 3

sinh x
(14) tanh x =
cosh x
ex ex
(15) = x
e + ex
ex + ex 2ex
(16) =
ex + ex
ex
(17) = 12 x
e + ex
1
(18) = 12
1 + e2x
So we get

(1)n (1)n
 
 n  1
(19) n tanh 2 = n 1 2 1 + en
n=1 n=1
(1) n
The series
n=1 n = ln 2 (standard calculus result), so we get

(1)n (1)n
 
 n  1
(20) n tanh 2 = ln 2 2 n 1 + en
n=1 n=1
The remaining series does converge quite fast since the denominator in-
creases exponentially, so we can add up the first few terms numerically and
compare the results. Taking the first 20 terms, we get (using software to do
the sum)

4 (1)n 4 ln 2 8 20 (1)n
 
 n  1
(21) 1 + tanh 1
n=1 n 2 n=1 n 1 + en
(22) 0.2206356003
To the same number of significant figures, we have

ln 2
(23) 0.2206356001

[Note that the answer given in problem 3.48 in Griffithss book is wrong
- the denominator should be , not 2.]
POLARIZABILITY OF HYDROGEN

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.1.2; Problem 4.1.
When a neutral atom is placed in an electric field, the positive nucleus
gets pulled in the direction of the field, and the negative electron cloud gets
pulled in the direction opposite to the field. This results in the centres of
positive and negative charge no longer being in the same place, so the atom
obtains a small dipole moment. Experimentally, it is found that for small
fields (that is, fields small compared to the field required to ionize the atom
by removing one or more electrons), this dipole moment is approximately
proportional to the applied field:

(1) p = E

where the constant is called the atomic polarizability, and has a charac-
teristic value for each type of atom.
A crude model of polarizability assumes that the electron cloud is a uni-
formly charged sphere and the nucleus is a point charge. An external field
is assumed to displace the cloud and nucleus so that the distance from the
centre of the cloud (which is assumed still to be spherical) to the nucleus is
d. We can find d by equating the force between the cloud and nucleus to
the opposite of the external force from the applied field.
Weve seen (Example 2 in here) that the field a distance d from the centre
of a uniformly charged sphere with charge density is

d
(2) E=
30
If the total charge in the cloud is q then = 3q/4a3 where a is the
radius of the cloud, so we can write the field as

qd
(3) E =
40 a3
p
(4) =
40 a3
1
POLARIZABILITY OF HYDROGEN 2

where p is the induced dipole moment. Therefore

(5) p = 40 a3 E
(6) = 40 a3

and we have a crude formula for the atomic polarizability. Experimentally,


this turns out to accurate to within a factor of 4 or so, which may not sound
that great, but at least its within an order of magnitude, and the model is
very crude. For example, for hydrogen, if we take a = 5.29 1011 m (the
Bohr radius), we get


(7) = 1.48 1031 m3
40

Experimentally, /40 for hydrogen is 6.67 1031 m3 .


As an example, suppose we have a hydrogen atom between a pair of
metal plates 1 mm apart, with a potential difference of 500 V across the
gap. In that case, the field is E = 5 105 V/m. Using the experimental
value for the polarizability, we get

(8) p = qd
(9) = E
(10) = 40 6.67 1031 5 105
(11) = 3.71 1035 C m

using 40 = 1.11 1010 in MKS units.


With the electron charge of 1.6 1019 coulombs, we get

(12) d = 2.32 1016 m


To estimate the voltage required to ionize the hydrogen atom, we can
work out the field required to make d = a. That is

qa
(13) Eion =

1.6 1019 5.29 1011
 
(14) =
(6.67 1031 ) (1.11 1010 )
(15) = 1.14 1011 V /m
POLARIZABILITY OF HYDROGEN 3

With plates 1 mm apart, we would therefore need a potential difference


of 1.14 108 volts.
P INGBACKS
Pingback: Polarizability of hydrogen - quantum version
Pingback: Polarizability - linear charge distribution
Pingback: Point charge and neutral atom
Pingback: Dielectric constant
Pingback: Relation between polarizability and susceptibility
Pingback: http://physicspages.com/pdf/Griffiths EM/Griffiths Problems
09.23.pdf
Pingback: Frequency dependence of electric permittivity
POLARIZABILITY OF HYDROGEN - QUANTUM VERSION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.2.
In the last post, we did a crude estimate of the separation between elec-
tron and proton in hydrogen due to polarization in an electric field. We as-
sumed there that the electron cloud was a uniformly charged sphere. A more
accurate representation is provided by quantum mechanics, which says that
the ground state of the electron has a charge density of

q 2r/a
(1) (r) = e
a3
We can use this to provide another estimate of the atomic polarizability.
As before, we assume that an electric field separates the electron cloud and
proton by a distance d, and that the cloud retains its spherical shape. Using
Gausss law (its a calculation similar to Example 2 here) we can work out
the electric field of a sphere with this charge density at a distance d from its
centre. Using the symmetry we can do the integrals over and to get 4
and we are left with the radial integral:
d 2
q
(2) 2
4d E = 3 r2 e2r/a sin d d dr
a 0 0 0 0
d
4q
(3) = r2 e2r/a dr
a3 0 0
This integral can be done twice by parts, or we can just use software, to
get
 3  3
a2

2 q a 2d/a a a 2
(4) d E= 3 e + d+ d
a 0 4 4 2 2
If we assume that d  a (as was the case when we assumed a uniform
charge density in the sphere), we can expand this result in a Taylor series
(again, by hand or using software) and keep only the leading non-zero term
in d. In fact, we discover that all terms up to and including d 2 cancel out,
so the lowest order term that isnt zero is d 3 /3. We therefore have, in this
approximation:
1
POLARIZABILITY OF HYDROGEN - QUANTUM VERSION 2

qd 3
(5) d2E
3a3 0
qd
(6) E
3a3 0
We can write this in the usual way using p = qd as the dipole moment,
and we have

(7) p 3a3 0 E

so in this approximation, the polarizability is

(8) = 3a3 0
With the uniformly charged sphere, the answer was = 4a3 0 so there
really isnt much difference. However, since the uniformly charged spheres
estimate was already on the low side, this supposedly more accurate form
for the electron density is actually worse.
P INGBACKS
Pingback: Polarizability - linear charge distribution
POLARIZABILITY - LINEAR CHARGE DISTRIBUTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.3.
In an earlier post, we did a crude estimate of the separation between elec-
tron and proton in hydrogen due to polarization in an electric field. In that
case, we assumed that the electron cloud was a uniformly charged sphere,
and the result was that the induced dipole moment is proportional to the
applied electric field.
Although this is also the result experimentally, it is possible to construct
artificial cases where this relation isnt true. For example, suppose that the
electron cloud is a sphere where the charge density is proportional to the
distance from the centre out to a distance R. That is

(1) (r) = kr
We can find the total charge:
R
(2) q = 4 kr r2 dr
0
4
(3) = kR
We can rewrite the charge density in terms of the total charge:

q
(4) (r) = r
R4
If the applied field separates the electron cloud and nucleus by a distance
d then we can do the same calculation as in the earlier post to get a relation
between the dipole moment and applied field. The nucleus must feel an
equal and opposite field when it moves out the distance d so using Gausss
law we get


2 4 d q
(5) 4d E = r r2 dr
0 0 R4
qd 4
(6) =
0 R4
1
POLARIZABILITY - LINEAR CHARGE DISTRIBUTION 2

The dipole moment is p = qd so we get


p
(7) p = R2 40 qE
In this case p E 1/2 .
If we have a radially symmetric charge density, we can work out the
condition on (r) so that p E. In general, from Gausss law, we have

2 4 d
(8) 4d E = (r)r2 dr
0 0
Since d is very small, we can expand (r) in the region [0, d] in a Taylor
series to get

(9) (r) = (0) + r 0 (0) + . . .


In order for p E, the leading term in the integral must contain d 3 , which
means that we must have (0) 6= 0. This is true of the uniformly charged
sphere and also the exponential distribution predicted by quantum mechan-
ics, but obviously not for the case in the current example.
POINT CHARGE AND NEUTRAL ATOM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.4.
In an earlier post, we saw that placing a neutral atom in an external field
induces a dipole moment p where for small fields the experimentally deter-
mined relation is

(1) p = E

where is the atomic polarizability.


If this external field is due to a point charge q at a distance r from the
atom then the field at the atom due to the charge is

1 q
(2) E=
40 r2
The induced dipole moment is therefore

q
(3) p=
40 r2
Weve seen that the electric field due to a dipole is (in spherical coordi-
nates):

p  
(4) E= 3
2 cos r + sin
40 r
The field from the dipole induced by the point charge is therefore

q  
(5) E= 2 cos r + sin
(40 )2 r5

where r points along the line from the atom to the point charge. The point
charge is thus located at = 0 so the force on the charge due to the dipole
is
1
POINT CHARGE AND NEUTRAL ATOM 2

2q2
(6) F=
(40 )2 r5
DIPOLE FORCE AND ENERGY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.1.3; Problems 4.7, 4.8, 4.9.
In a constant electric field, a perfect dipole feels no force since the force
on the positive charge is cancelled by that on the negative charge. However,
if the field varies, there could be a net force on the dipole. Over atomic
or molecular distances, the field would need to change rapidly in order for
there to be any appreciable force. This force is the difference between the
forces on the two charges, so we get

(1) F = qE+ qE
(2) = q (Ex x + Ey y + Ez z)
For an infinitesimal distance, as would be found in an ideal dipole, we
can write the change in field as the dot product of the gradient of each
component with the infinitesimal displacement dl:

(3) Ex = Ex dl
The above relation for the force can then be written as

(4) F = (qdl Ex ) x + (qdl Ey ) y + (qdl Ez ) z


(5) = q (dl ) E
(6) = (p ) E
To work out the energy of a dipole in an electric field, suppose we start
with two equal and opposite charges at the same location, and that the elec-
tric field is parallel to the z axis. Now we move +q to location d/2 and q
to d/2. The force on +q is qE and the work done in moving it is qE d/2.
Similarly, the work in moving q is also qE(d/2) = qEd/2. Thus the
total work is qE d = p E. If were working against the force, the potential
energy is negative so we get U = p E, which is the standard formula.
However, there are a couple of things about this that are a bit unclear.
First, we assumed that E is constant over the dimensions of the dipole. For
an ideal dipole, this is fair enough, since ideal dipoles are essentially point
objects. However, suppose we have a field that is constant over all space.
1
DIPOLE FORCE AND ENERGY 2

In that case, the force on the dipole is zero everywhere, regardless of the
orientation or position of the dipole. So we should be able to start the dipole
off in its final orientation and move it in from infinity, all with doing zero
work. This seems to imply that dipoles have zero energy in a constant field.
But, we also know that rotating a dipole requires work, since the field
exerts a torque on the dipole. Weve seen that this torque is

(7) N = pE

If we start off with p and E perpendicular, and then rotate the dipole to
some angle we have to do work against the torque, which is


(8) W = pE sin 0 d 0
/2
(9) = pE cos
(10) = pE

Thus in one case, we get the energy from separating the two charges
against the electric field, and in the second case we get it from rotating the
dipole against the field. Theres clearly something Im missing here.
Anyway, given that the energy has the form given, we can work out the
interaction energy of two dipoles. We saw earlier that the field due to a
dipole p1 can be written as

1
(11) E= [3(p1 r)r p1 ]
40 r3

The other dipole therefore has an energy within this field of

(12) U = p2 E
1
(13) = [p1 p2 3(p1 r)(p2 r)]
40 r3

As a final example, suppose we have a dipole and a point charge, with the
dipole making an angle with the line connecting it with the charge. The
force on the dipole due to the charge can be found from the formula above
for the force. We can take the line connecting the two as the x axis, so the
field due to the charge is
DIPOLE FORCE AND ENERGY 3

q
(14) Eq = r
40 r2
" #
q xx + yy + zz
(15) =
40 (x2 + y2 + z2 )3/2
The force from 6 is then
" #
q xx + yy + zz
(16) Fp = p 3/2
40 (x2 + y2 + z2 )
We get

q
px 2x2 y2 z2 + 3py xy + 3pz xz
  
(17) Fpx = 5/2
40 (x2 + y2 + z2 )
q 
3x (px x + py y + pz z) px x2 + y2 + z2

(18) = 5
40 r
q 
3xp r px r2

(19) = 5
40 r
q  2 2 2
 
(20) Fpy = 5/2
3px xy + p y 2y x z + 3p z xz
40 (x2 + y2 + z2 )
q 
3yp r py r2

(21) = 5
40 r
q  2 2 2

(22) Fpz = 5/2
3px xz + 3yzp y + p z 2z x y
40 (x2 + y2 + z2 )
q 
3zp r pz r2

(23) = 5
40 r
We can combine these to get a single vector expression for F:

q
(24) Fp = (p 3 (p r) r)
40 r3
To get the force on the charge, we could just use Newtons third law
(equal action and reaction) to say the force is the negative of that on the
dipole, but we can also observe this from the formula for the field of a
dipole. We get, for a unit vector that points in the opposite direction to that
above

1
(25) Ep = [3(p (r)) (r) p]
40 r3
DIPOLE FORCE AND ENERGY 4

so the force on the charge is just

q
(26) Fq = [3(p r)r p]
40 r3
(27) = F p
P INGBACKS
Pingback: Dipole-dipole forces and torques
Pingback: Dipole between two angled conducting planes
Pingback: Force on a dielectric sphere due to a charged wire
Pingback: Susceptibility of a polar dielectric
Pingback: Force on a magnetic dipole in a slab of current
Pingback: Energy of magnetic dipole in magnetic field
FIELD OF A POLARIZED OBJECT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.2.1; Problem 4.10.
The potential of an ideal dipole is

1
(1) V (r) = r p
40 r2

where r is the vector from the dipole to the observation point.


If we now consider an object consisting of polarized dielectric, we can
define the polarization density, or polarization per unit volume as P. The
potential due to such an object is then

1 1 (r r0 ) P(r0 ) 3 0
(2) V (r) = d r
40 |r r0 |2 |r r0 |

Note that this is a volume integral over the primed coordinates r0 , that is,
over the location of the volume element containing the polarized material.
We can transform this integral by doing a calculation of a gradient:

 
1 1
(3) 0 = x + . . .

|r r0 | x 0 1/2
h i
0 2 0 2 0 2
(x x ) + (y y ) + (z z )
(x x0 )
(4) =h i3/2 x + . . .
0 2 0 2 0 2
(x x ) + (y y ) + (z z )
1 (r r0 )
(5) =
|r r0 |2 |r r0 |

where weve omitted the terms in y and z in the first two lines since they
have the same form. We can therefore write the original integral as
 
1 0 1
(6) V (r) = P d 3 r0
40 |r r0 |
1
FIELD OF A POLARIZED OBJECT 2

A standard theorem from vector calculus says, for a function f and vector
field A:

(7) ( f A) = A f + f A

so we can transform the integral to get


 
1 0 P 3 0 1 1
(8) V (r) = 0
d r 0
0 Pd 3 r0
40 |r r | 40 |r r |
The first integral is a volume integral of a divergence so we can apply the
divergence theorem to transform this to a surface integral, so we finally get
 
1 1 0 1 1
(9) V (r) = 0
P da 0 Pd 3 r0
40 |r r | 40 |r r0 |

where the first integral is over the surface of the polarized object, and the
second integral is over its volume. That is, the potential of a polarized object
can be expressed as the sum of the potential of a surface charge density and
a volume charge density, where we have

(10) b P n
(11) b 0 P
These charge distributions are known as bound charges, which is why
weve used a subscript b. If were interested in the electric field of a dipole
distribution, we can work out these integrals and then take the negative
gradient to get E = V , or, if the problem has the right symmetry, we can
use Gausss law to work out the fields of the two charge distributions and
then add them together.
As an example, suppose we have a sphere of radius R with a polarization
density given by

(12) P(r0 ) = kr0

where k is a constant. Then

(13) b = kR
(14) b = k0 r
(15) = 3k
FIELD OF A POLARIZED OBJECT 3

where the last line can be found by expressing r in rectangular coordinates.


Note that the total charge in the shell qs = 4R2 kR is equal and opposite to
the total volume charge 43 R3 (3k).
Rather than working out the integrals above, we can use Gausss law to
find the electric field from a spherical shell and a uniformly charged sphere.
Weve already solved this problem in examples 1 and 2 here, so we can just
quote the results. Inside the sphere, the spherical shell contributes nothing,
and the volume charge gives a radial field of

kr
(16) E =
0
Outside the sphere, the shell contributes

kR3
(17) Es =
0 r 2

and the volume charge contributes

kR3
(18) Ev =
0 r 2

so the total field is Es + Ev = 0.


P INGBACKS
Pingback: Field of a polarized object - examples
Pingback: Field of a polarized cylinder
Pingback: Bound charges in a dielectric
Pingback: Electric displacement
Pingback: Electric field within a cavity inside a dielectric
Pingback: Point charge embedded in dielectric plane
Pingback: Energy in a dielectric
Pingback: Dielectric cube: bound charges
Pingback: Point charge in dielectric sphere
Pingback: Conducting sphere half-embedded in dielectric plane
Pingback: Magnetization: bound currents
Pingback: Magnetostatic formulas from electrostatic formulas
Pingback: Uniform charge, polarization & magnetization
Pingback: Magnetic dipoles versus monopoles; an experiment
Pingback: Maxwells equations in matter
FIELD OF A POLARIZED OBJECT - EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 4.11, 4.12.
We saw before that we can write the potential of a polarized object as


1 1 (r r0 ) P(r0 ) 3 0
(1) V (r) = d r
40 |r r0 |2 |r r0 |

Note that this is a volume integral over the primed coordinates r0 , that is,
over the location of the volume element containing the polarized material.
Although we can work out the field due to a uniformly polarized sphere
using the techniques in the last post, it is also possible to do this using this
integral directly. For a uniformly polarized sphere, P(r0 ) is constant over
the volume of the sphere, so we can take it outside the integral.


1 1 (r r0 ) 3 0
(2) V (r) = P d r
40 |r r0 |2 |r r0 |
The remaining integral depends only on the position r of the observation
point, and not on the polarization vector. We can therefore take r to lie on
the z axis. The angle between r and r0 is therefore the polar angle . By
symmetry, only the z component of the vector in the integral will be non-
zero, so we can work out that on its own. This is actually the same problem
we faced when calculating the electric field of a uniformly charged sphere,
which was done here. Quoting the answer from that post, we have

( 3
1 (r r0 ) 3 0 4R
3r 2 z r > R
(3) d r =
|r r0 |2 |r r0 | 4r
3 z r < R

where the plus sign is taken if r points in the +z direction and the minus
sign in the other case.
We can now insert the polarization vector so that it makes an angle with
the observation vector r and since P z = P cos we have
1
FIELD OF A POLARIZED OBJECT - EXAMPLES 2

R3
(
30 r2
P cos r>R
(4) V (r) = r
30 P cos r<R
Note that the electric field inside the sphere is uniform. Since z = r cos :

(5) E = Vin
P
(6) = z
30
As a simple example of the use of the bound charges representation of a
polarized object, we can look at a cylinder which contains a uniform polar-
ization P parallel to its axis. In this case, P = 0 everywhere inside the
cylinder and P n = 0 on the sides of the cylinder. On the ends, P n = P,
so the only bound charge is on the ends. We therefore have two charged
disks separated by the length L of the cylinder. Weve worked out an ap-
proximation for the potential of a single charged disk before, but its diffi-
cult to get much of a picture of what the field looks like from that solution.
However, if the length is much greater than the radius, we have essentially
two point charges of equal and opposite sign, which is a physical dipole.
For L much less than the radius, we have a dipole layer, so we would expect
field lines coming out of the side with positive charge and curling round to
enter on the negative side. If the radius and length are roughly equal, we
would expect field lines going directly between the disks, with lines curving
round the outside as well.
P INGBACKS
Pingback: Electric field within a cavity inside a dielectric
Pingback: Dielectric sphere in uniform electric field
Pingback: Energy in a dielectric
FIELD OF A POLARIZED CYLINDER

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.13.
As another example of the use of the bound charges representation of a
polarized object, we can look at a cylinder which contains a uniform polar-
ization P perpendicular to its axis. In this case, P = 0 everywhere inside
the cylinder. If we take the direction of P to be = 0 then P n = P cos .
We thus must find the potential of an infinite cylinder with a surface charge
of

(1) b = P cos
Weve worked out the general solution to Laplaces equation in cylindri-
cal coordinates before, so we can use the results from there. The solution
inside the cylinder is


(2) Vin = Bin + [An rn sin n + Bn rn cos n ]
n=1

while outside it is

 
Dn Cn
(3) Vout = Bout + n
cos n n sin n
n=1 r r
From the boundary condition requiring the potential to be continuous at
the surface of the cylinder we get the relations

(4) Bout = Bin


(5) Cn = An R2n
(6) Dn = Bn R2n

where R is the radius of the cylinder.


From the condition on the derivative of the potential at the boundary, we
get
1
FIELD OF A POLARIZED CYLINDER 2



n1
2nRn1Bn cos n = 0b
  
(7) 2nR An sin n +
n=1 n=1
From 1 we see that all coefficients of the sine terms are zero, as are all
coefficients of cosine terms except for n = 1. We therefore get

P
(8) B1 =
20
(9) D1 = R2 B1
PR2
(10) =
20
The potential in the two regions is thus

P
(11) Vin = r cos
20
PR2
(12) Vout = cos
2r0
From this we can get the field by taking the negative gradient. Since
r cos = x, the field inside the cylinder is

(13) Ein = Vin


P
(14) = x
20
The interior field is thus uniform, just as the field inside a uniformly
polarized sphere is uniform.
Outside the cylinder we need to take the gradient in cylindrical coor-
diantes. We get

Vout 1 Vout
(15) Eout = r
r r
R2  
(16) = 2
P cos r + P sin
2r 0
We can write this in terms of the polarization vector. If P points in the
direction = 0 then we have

(17) P = P cos r P sin


FIELD OF A POLARIZED CYLINDER 3

where the minus sign on the second term is because points in the direction
of increasing , which is clockwise from = 0. We have then

(18) P cos r + P sin = 2 (P r) r P

so the field is

R2
(19) Eout = [2 (P r) r P]
2r2 0
BOUND CHARGES IN A DIELECTRIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.14.
Weve seen that potential due to the polarization density in a dielectric
can be represented as bound charges which divide into a volume charge
distribution and a surface charge distribution. The relation between these
charge distributions and the polarization density P is

(1) b = P n
(2) b = P

where n is the unit normal to the surface of the dielectric.


Since a dielectric is neutral, the polarization results from moving around
charge within it, not from adding or removing charge, so the total charge
should remain as zero. This follows directly from the divergence theorem,
since

2
(3) P nd r = Pd 3 r
A V

where the first integral is over the surface of the dielectric and the second is
over its volume. From that relationship, we see that the total surface charge
is equal and opposite to the total volume charge, so their sum is always zero.
In the special case (such as weve considered in the last few posts) where
the polarization inside the volume is uniform, P = 0 everywhere inside
the dielectric, so the total surface charge must also be zero.
P INGBACKS
Pingback: Electric displacement
Pingback: Dielectric constant

1
ELECTRIC DISPLACEMENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.3.1 and Problem 4.15.
Weve seen that potential due to the polarization density in a dielectric
can be represented as bound charges which divide into a volume charge
distribution and a surface charge distribution. The relation between these
charge distributions and the polarization density P is

(1) b = P n
(2) b = P

where n is the unit normal to the surface of the dielectric.


These bound charges can be used to work out the electric field due to
the polarization. However in a more general situation, there could be other
charges, which well call free charges not associated with the polarization,
and these charges will also generate an electric field. In a real dielectric,
there is no real distinction between the volume and surface charge densi-
ties, since the volume charge density will not change discontinuously at the
surface. We can therefore concentrate on the volume charge density b in
the analysis that follows.
If we add a free charge density f to the bound charge density, then
the total charge density is = b + f . From Gausss law we know that
0 E = b + f = P + f . We can combine the two divergence terms
to get

(3) (0 E + P) = f
This new vector is called the electric displacement D:

(4) D 0 E + P
The units of D are those of polarization density, which is dipole moment
per unit volume. The dipole moment has units of charge times distance,
so the units of D are charge times distance over volume, or charge per unit
area.
1
ELECTRIC DISPLACEMENT 2

Another way of looking at it is in terms of a parallel plate capacitor,


initially in a vacuum. If the capacitor has flat plates that carry a charge of f
(positive on one plate, negative on the other), then the electric field between
the plates is E = f /0 , pointing from the positive plate to the negative one.
Now if we introduce a dielectric between the plates, the field will induce
bound charges in the dielectric. In particular, there will be a surface charge
b on the surfaces of the dielectric in contact with the plates.  Thus the
effective field between the plates is reduced to E = f b /0 . From
the formula above, we can write this as

(5) 0 E + b = f
(6) 0 E + P = f
Thus the displacement is the density of surface charge required to pro-
duce a given field in a capacitor filled with a dielectric. The actual value of
P will of course depend on the material used for the dielectric.
We can integrate the divergence equation and use the divergence theorem
to get

(7) Dd 3 r = Q f
V

(8) = D da
A

where the integral in the first line is taken over a volume and that in the
second line is over the surface bounding the volume.
This integral can be used to simplify the calculation of electric field in
some situations. For example, if we have a spherical dielectric shell with
inner radius a and outer radius b, and this shell has a polarization given by

k
(9) P = r
r

we could calculate the electric field by first working out the bound charges
and then using Gausss law. From above, the bound charges are
(
k
b r=b
(10) b =
ak r=a
and
ELECTRIC DISPLACEMENT 3

(11) b = P
 
1 k
(12) = 2 r2
r r r
k
(13) = 2
r
Using Gausss law, in the region r < a, there is no enclosed charge, so
Er<a = 0. Inside the shell, we have the contributions from the volume
charge and the surface charge on the inner surface of the shell:
r
4a2 k k 02 0
(14) Q = 4 r dr
a a r02
(15) = 4kr
The field is thus obtained from

(16) 4r2 0 E = Q
(17) = 4kr
k
(18) E =
r0
Outside the shell, the net enclosed charge is zero (as we saw in the last
post), so again E = 0.
Using the displacement, however, since there is no free charge anywhere,
we have

(19) D da = 0
A

in all three regions. Inside the inner surface and outside the outer surface,
we know that P = 0, so this is equivalent to

(20) E da = 0
A

and from the symmetry of the problem, we know that any field would have
to be radial, so this gives E = 0 in these two regions.
Between the surfaces of the shell, again we know that both the field and
the polarization density are radial and symmetric, so this implies that D = 0
as well, which gives
ELECTRIC DISPLACEMENT 4

P
(21) E =
0
k
(22) = r
r0
Its important to notice that in a more general situation, the equation 19
does not always imply D = 0. This is because we need both the diver-
gence and the curl of a vector field to specify it uniquely, and although the
curl of E is always zero in electrostatic problems, there is no corresponding
condition on P. In the problem above, we have to invoke the extra condi-
tions imposed by the symmetry of the problem to conclude that D = 0; we
couldnt conclude that solely from the integral 19.
P INGBACKS
Pingback: Electric field within a cavity inside a dielectric
Pingback: Dielectric constant
Pingback: Dielectric examples
Pingback: Dielectric sphere with free charge
Pingback: Dielectric cylinder in uniform electric field
Pingback: Energy of conducting sphere in a dielectric shell
Pingback: Energy in a dielectric
Pingback: Point charge in dielectric sphere
Pingback: Electric displacement: boundary conditions
Pingback: Uniqueness of potential in dielectrics
ELECTRIC FIELD WITHIN A CAVITY INSIDE A DIELECTRIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.16
We can use superposition to work out the electric field within a small
cavity that is hollowed out inside a block of dielectric. Well take the elec-
tric field within the dielectric to be E0 (which need not be constant), and the
polarization to be P (again, not necessarily constant). The displacement is
then D0 = 0 E0 + P. In the small region of interest, well take P pointing
upwards, so that the field it induces points downwards.
Now if we hollow out a small sphere (assumed to be small enough that
the field and polarization can be taken as constant within it), we can work
out the field within the hollow sphere. The trick is that the polarization
within empty space must be zero, so we can simulate the situation by su-
perimposing a sphere with equal and opposite polarization P on top of the
dielectric. Weve worked out the field within a uniformly polarized sphere
earlier, and with polarization P this field is

1
(1) Es = P
30
Thus the net field within a spherical cavity is

1
(2) E = E0 + Es = E0 + P
30

Since the net polarization within the cavity is zero, the displacement is

1 2
(3) D = 0 E0 + P = D0 P
3 3
Now suppose the cavity is shaped like a long thin needle parallel to P.
Again, we superimpose a needle with the opposite polarization. We can
work out the bound charges, and since the polarization within the needle is
constant, b = (P) = 0 and since the polarization is parallel to the axis,
b = P n = 0 on the sides of the needle. On the ends of the needle, b
is non-zero, but if the needle is long enough with small end points, this will
1
ELECTRIC FIELD WITHIN A CAVITY INSIDE A DIELECTRIC 2

contribute a very small field so we can approximate the situation by saying


that the electric field is not significantly modified within the cavity:

(4) E = E0
(5) D = 0 E0 = D0 P
Finally, we consider a thin, circular wafer shaped cavity perpendicular to
the polarization. In this case, the bound volume charge is again zero, but
the surface charge is

(6) b = P n = P

with the plus sign on the bottom of the wafer and the minus sign on top. The
field due to the bottom (positive) surface is (by Gausss law) P/20 upwards,
and the field due to the top (negative) surface is also P/20 upwards, so the
total field within the wafer is

1
(7) E = E0 + P
0
(8) D = 0 E = D0
Thus the needle leaves E unchanged and the wafer leaves D unchanged.
P INGBACKS
Pingback: Magnetic field within a cavity
Pingback: Magnetic dipoles versus monopoles; an experiment
DIELECTRIC CONSTANT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.4.1 and Problem 4.18.
Experimentally, the dipole moment of an atom is proportional to the ap-
plied electric field (for small fields). Since the polarization of a dielectric is
due to individual atoms within the dielectric being given dipole moments,
it should come as no surprise that the polarization density P of a dielectric
is also proportional to the applied field. The relation is written as

(1) P = 0 e E

where e is the electric susceptibility. Due to the presence of the 0 , e is


dimensionless.
Not all substances obey such a simple law, but for those that do, they are
called linear dielectrics.
If this condition holds, we get a simple relationship between the displace-
ment, the field and the polarization. We have

(2) D = 0 E + P
(3) = 0 (1 + e ) E
The quantity 1 + e is called the dielectric constant for a material. The
entire proportionality constant is called the permittivity of the material:

(4) = 0 (1 + e )
Since a vacuum cannot be polarized at all, e = 0 for a vacuum, so that
= 0 in that case. This is why 0 is called the permittivity of free space.
As a simple example of how these relations can be used, suppose we have
a parallel plate capacitor whose plates are separated by a distance 2a. On
one plate there is a surface charge density of and on the other there is .
Between the plates are two slabs of dielectric, each of thickness a. Slab 1
(next to the positive plate) has a dielectric constant of 2 and slab 2 has a
dielectric constant of 1.5.
1
DIELECTRIC CONSTANT 2

We can begin by finding the displacement D. Using Gausss law for


displacement, we can build a little cylindrical Gaussian pillbox of radius r
with one end in the positive plate and the other in slab 1. By symmetry D is
parallel to the axis of the cylinder so there are no contributions to Dda from
the sides of the cylinder. The end of the cylinder inside the plate will have
D = 0 (since were inside a conductor), while the other end will contribute
r2 D. The charge enclosed by the cylinder is r2 so we get

(5) D da = Q

(6) r2 D = r2
(7) D =
This result is independent of the dielectric, so it holds everywhere be-
tween the plates. If we assume the positive plate lies above the negative
one, we have D = z since the displacement vector points from positive
to negative.
From the relation above, we can find E. For the two slabs, we have

1
(8) E1 = D
0 (1 + e )1

(9) = z
20

(10) E2 = z
1.50
The polarization within the two slabs can now be found from 1.

(11) P1 = 0 (2 1)E1

(12) = z
2

(13) P2 = z
3
The potential difference between the plates is

(14) V = E1 a + E2 a
7 a
(15) =
6 0
The bound charges resulting from the polarization can be found. Since
the polarization is uniform within each slab, there is no volume bound
charge since P = 0 everywhere. The surface bound charge is found from
DIELECTRIC CONSTANT 3

(16) b = P n
Applying this at the boundary between the positive plate and slab 1, we
get b = /2. At the boundary between slabs 1 and 2, there is a bound
charge of + /2 on slab 1 and /3 on slab 2 (for a net bound charge of
+ /6 at the boundary). At the boundary between slab 2 and the negative
plate, we have + /3.
We can use these bound charges to check the values for the field found
above. Choosing a Gaussian cylinder with one end in the positive plate and
the other in slab 1, the enclosed charge is r2 ( /2) which is equal to
0 times the surface integral of the field over the cylinders end cap which is
r2 E1 . Thus E1 = /20 (pointing downwards) as before. Doing the same
calculation for slab 2 we get E2 = 2 /30 as before.
P INGBACKS
Pingback: Dielectric examples
Pingback: Dielectric sphere with free charge
Pingback: Coaxial cable with dielectric
Pingback: Dielectric cylinder in uniform electric field
Pingback: Dielectric sphere in uniform electric field
Pingback: Point charge embedded in dielectric plane
Pingback: Point charge in dielectric sphere
Pingback: Uniqueness of potential in dielectrics
Pingback: Transmission lines
Pingback: Skin depth of electromagnetic waves in conductors
Pingback: http://physicspages.com/pdf/Griffiths EM/Griffiths Problems
09.23.pdf
Pingback: Frequency dependence of electric permittivity
Pingback: Microwave shielding for perfect transmission
Pingback: Maxwells equations in matter: boundary conditions
DIELECTRIC EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.4.1 and Problem 4.19.
In a linear dielectric, we have a simple relationship between the displace-
ment, the field and the polarization. We have

(1) D = 0 E + P
(2) = 0 (1 + e ) E
(3) E
The quantity 1 + e is called the dielectric constant (sometimes called
the relative permittivity) of a material. If we use the symbol r for this, we
have

(4) r = 1 + e

(5) =
0
From the definition of displacement, we have that

(6) D = f

where f is the free charge (as opposed to bound charge) in the dielectric.
In the case of a linear dielectric, we have for the curl:

(7) D = 0 E + P
(8) = P

since E = 0 in all electrostatic problems. Is the curl of the polarization


zero in general? We might think that, for linear dielectrics, it has to be,
since P is proportional to E. However, it is the constant of proportionality
0 e that messes things up, since if we have two adjacent linear dielectrics
with different susceptibilities, the constant is different on either side of the
boundary. The argument goes something like this:
1
DIELECTRIC EXAMPLES 2

Suppose we want to calculate the line integral of P along a rectangu-


lar path where the long sides of the rectangle lie on opposite sides of the
boundary (and are parallel to the boundary) while the short sides of the rec-
tangle are perpendicular to the boundary. We can consider an extreme case
where one side of the boundary is a vacuum (so P = 0), while the other is
a dielectric with P 6= 0. In the general case where P is not normal to the
boundary, P dl 6= 0 within the dielectric, so the line integral wont be zero.
By Stokess theorem, the curl isnt zero either.
At this point, you might think: but what about the case of the electric field
at the boundary of a conductor? We know that E = 0 inside the conductor,
and if the conductor has some surface charge, E 6= 0 outside the conductor.
Cant we use the same argument to show that E 6= 0 in this case also?
The key point here is that the electric field just outside a conductor is always
normal to its surface, so in this case E dl = 0, so the line integral is zero
and the curl is zero as it has to be. (Actually, we know from the original
derivation of the E = 0 condition, using the fact that the electric field
is the gradient of a potential, that the line integral around any path is zero,
so the argument doesnt rely on our choosing a rectangle as the path. The
electric field will always be oriented in such a way that the line integral
comes out to zero.)
Returning to the polarization, we can see that if the dielectric is uniform
and homogeneous (that is, its the same dielectric everywhere there is an
electric field), then the curl will be zero, and in fact the displacement D
satisfies (almost) the same equations as that of the vacuum electric field Ev :

(9) D = f = 0 Ev
(10) D = 0 = Ev

Whenever two vector fields have the same divergence and curl, they must
be equal, so we can say that

(11) D = 0 Ev

However, we also know that

(12) D = E

where the E in this relation is actual electric field inside the linear dielectric.
Therefore
DIELECTRIC EXAMPLES 3

0
(13) E = Ev

1
(14) = Ev
r

That is, a uniform dielectric reduces the electric field by a factor equal to
the dielectric constant.
As an example, suppose we have a parallel plate capacitor with a spacing
d between the plates. Half the distance between the plates is filled by a
dielectric slab with constant r with the remaining half being a vacuum. If
the potential difference between the plates in a vacuum is V , what will the
potential difference be with the dielectric?
The half of the distance that is vacuum will be unaffected by the dielec-
tric, and accounts for half the original potential difference, or V /2. In the
dielectric the electric field is reduced by a factor r , so the potential differ-
ence within the dielectric will be V /2r . The total potential difference will
now be

 
V 1
(15) V1 = 1+
2 r
(1 + r )V
(16) =
2r

This effectively changes the capacitance, since C = Q/V . Thus the ratio
of the new capacitance to the original is

C1 V
(17) =
C V1
2
(18) =
(1 + 1/r )
2r
(19) =
1 + r

The actual location of the dielectric slab doesnt matter in this calculation,
but to make things definite, suppose its positioned midway between the
plates, so there is a gap of d/4 on each side of the slab. We can then use
similar techniques to the earlier post to find the various quantities.
The electric field in the vacuum sections is
DIELECTRIC EXAMPLES 4

V /2
(20) Ev =
d/2
V
(21) =
d
2r V1
(22) =
1 + r d
That is, its unchanged from the pure vacuum case. Within the dielectric,
we have

Ev
(23) Ed =
r
V
(24) =
r d
2 V1
(25) =
1 + r d
The surface charge on the capacitor plates must produce the vacuum field,
so

(26) = 0 Ev
V
(27) = 0
d
2r 0 V1
(28) =
1 + r d
The polarization in the vacuum is zero, and within the dielectric it is

(29) Pd = 0 (r 1) Ed
r 1 V
(30) = 0
r d
r 1 V1
(31) = 20
r + 1 d
Since the bound surface charge on the dielectric is b = P n on the upper
surface of the dielectric the polarization and normal vector point in opposite
directions so

r 1 V1
(32) b = 20
r + 1 d
DIELECTRIC EXAMPLES 5

Finally, the displacement is D = 0 E + P (since all vectors are parallel


here, we can use magnitudes), so

20 r V1
(33) Dv =
1 + r d
(34) Dd = 0 Ed + Pd
20 r V1
(35) =
1 + r d
The displacement is the same everywhere between the plates.
Now suppose the dielectric covers the entire gap between the plates, but
occupies only half the area of the plates. Because the capacitor plates are
conductors, the potential must be the same everywhere inside a given plate.
Therefore, the potential difference is the same whether we go through the
dielectric or the vacuum. This in turn means the electric field must be the
same on both sides, and since the electric field inside a dielectric is reduced
by a factor r , there must be more charge on the plate where it contacts the
dielectric. That is, on the vacuum side we have

V2
(36) Ev =
d
v
(37) =
0

while on the dielectric side we have

(38) Ed = Ev
d
(39) =
0
r v
(40) =
0
So the charge density on the plate next to the dielectric is

(41) d = r v
V2
(42) = r 0
d
This means that the capacitor holds more charge than if there were a vac-
uum between its plates, so, since both regions contribute an equal amount:
DIELECTRIC EXAMPLES 6

C2 v + d
(43) =
C 2v
1 + r
(44) =
2
The polarization within the dielectric is

(45) Pd = 0 (r 1)Ed
V2
(46) = 0 (r 1)
d
The surface charge on the dielectric is

V2
(47) b = 0 (r 1)
d
The displacement is

(48) Dv = 0 Ev
V2
(49) = 0
d
(50) Dd = 0 Ed + Pd
V2 V2
(51) = 0 + 0 (r 1)
d d
V2
(52) = 0 r
d
P INGBACKS
Pingback: Point charge embedded in dielectric plane
DIELECTRIC SPHERE WITH FREE CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.4.1 and Problem 4.20.
Suppose we have a sphere of linear dielectric (radius R) in which is em-
bedded a uniform free charge of density .
From the definition of displacement, we have that

(1) D =
The integral form of this is, using Gausss law:

(2) D da = Q

where Q is the charge enclosed by the surface over which the integral is
done. In this case, from symmetry, we get for points r < R:

4 3
(3) 4r2 D = r
3
r
(4) D =
3
Outside the sphere, we have

4 3
(5) 4r2 D = R
3
R3
(6) D =
3r2
In both regions, the relation between D and E is D = 0 r E where r is
the dielectric constant, so we have
( r
30 r 0<r<R
(7) E= R3
30 r2
r>R
Outside the sphere the dielectric constant is 1, since we are in a vacuum
there.
1
DIELECTRIC SPHERE WITH FREE CHARGE 2

From this we can find the potential at the centre of the sphere, relative to
infinity.
0
(8) V = Edr

R 0
R3 dr
(9) = 2
rdr
30 r 30 r R
2
R 2
R
(10) = +
30 60 r
R2
 
1
(11) = 1+
30 2r
COAXIAL CABLE WITH DIELECTRIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.21.
Suppose we have a coaxial cable with an inner conducting core of radius
a and an outer conducting cylinder of radius c. Part of the space between
the conductors is filled with a linear dielectric with dielectric constant r
which extends from radius b to c.
We can use an analysis similar to that which we used in working out the
capacitance of a coaxial cable earlier. If we place a surface charge density of
on the inner conductor, then using the analysis from before, the potential
in the region a < r < b is

a r
(1) V= ln
0 a
For the region b < r < c, the electric field is reduced by a factor of r so
in this region we have for the potential relative to radius b:

a r
(2) V= ln
0 r b
The total potential difference between the two conductors is then

a b a c
(3) Vtot = ln + ln
0 a 0 r b
 
a b 1 c
(4) = ln + ln
0 a r b
The capacitance per unit length of the cable is found from C = Q/V . The
charge per unit length is Q = 2a , so we get

2a
(5) C = h i
a b 1 c
0 ln a + r ln b
20
(6) =
ln ab + 1r ln bc

1
DIELECTRIC CYLINDER IN UNIFORM ELECTRIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.22.
In our original investigation into electrostatic boundary conditions in a
vacuum, we found that the electric field is discontinuous across a layer of
surface charge. The discontinuity was

above below
(1) E E =
0
We can do a similar analysis for a dielectric by using the displacement
instead of the bare electric field. We know that

(2) D da = Q f
A

where Q f is the free charge enclosed by the surface of integration. There-


fore, if we consider a dielectric with a free surface charge of f , we can use
a small cylindrical Gaussian surface with ends on either side of the surface
to say that

(3) Dabove
Dbelow
= f
Thus if there is no free surface charge on the dielectric, the perpendicular
component of D is continuous across the surface. The dielectric may be
polarized, since polarization induces only a bound surface charge, and that
has no effect on the continuity of D .
In a linear dielectric, D = E and as usual E = V , so we can convert
this into a condition on the potential on each side of the surface layer:

V1 V2
(4) 1 2 = f
n n
As an example, suppose we have a long cylinder of dielectric that is in
a uniform electric field E0 , where the field is perpendicular to the axis of
the cylinder. We can find the potential everywhere, and thus find the field
inside the cylinder.
1
DIELECTRIC CYLINDER IN UNIFORM ELECTRIC FIELD 2

To solve this problem, we set up a cylindrical coordinate system with its


axis centred on the axis of the cylinder. We can define the = 0 direction as
the direction of the electric field. In addition to the boundary condition 4, we
have the addition boundary condition that the potential must be continuous,
and we also have the asymptotic condition that for large r, the electric field
must tend to a constant. Since there is no free charge, f = 0 and the three
conditions become

V1 V2
(5) 1 = 2
n r=R n r=R
(6) V1 (R) = V2 (R)
(7) lim V2 (r) = E0 r cos
r
Here, V1 is the potential inside the cylinder and V2 is the potential out-
side. R is the radius of the cylinder. The last condition says that the poten-
tial for large r must be E0 where r cos is a coordinate measured
along the = 0 direction. It would be confusing to call this coordinate z,
since in cylindrical coordinates, z denotes the direction along the axis of the
cylinder, which weve already said is the axis of the dielectric. With this
definition, the field is E0 = (dV /d ) = E0 .
Since there is no free charge, the system satisfies Laplaces equation, so
we can make use of our solution in cylindrical coordinates. The general
solution is

(8)

1
V (r, ) = A ln r +K + rn (An sin n + Bn cos n )+ n
(Cn sin n + Dn cos n )
n=1 n=1 r
Inside the cylinder we can throw away the log term and the 1/rn terms to
keep the potential finite at r = 0. Thus inside we have


(9) V1 = K1 + rn (An sin n + Bn cos n )
n=1
Outside, we again throw away the log term. We can also throw away
all the rn terms except for n = 1, since we need the asymptotic behaviour
referred to above. Thus we get


1
(10) V2 = K2 E0 r cos + n
(Cn sin n + Dn cos n )
n=1 r
We can now apply the two other boundary conditions above. First:
DIELECTRIC CYLINDER IN UNIFORM ELECTRIC FIELD 3

(11) V1 (R) = V2 (R)


(12)

n 1
K1 + R (An sin n + Bn cos n ) = K2 E0 R cos + n
(Cn sin n + Dn cos n )
n=1 n=1 R

From the uniqueness of series, we can equate coefficients of the various


trig functions, so we get

(13) K1 = K2
Cn
(14) An Rn =
Rn
Dn
(15) Bn Rn = (n 6= 1)
Rn
D1
(16) RB1 = E0 R +
R
Since the value of the constant K1 makes no difference to the field (it
disappears when we take the derivative), we might as well take K1 = K2 = 0.
Now from the derivative condition, we have, using 1 = 0 r , where r is
the dielectric constant, and 2 = 0 , since outside the dielectric we have a
vacuum:

V1 V2
(17) 1 = 2
n r=R n r=R
(18)
" #

n
r nRn1 (An sin n + Bn cos n ) = E0 cos Rn+1
(Cn sin n + Dn cos n )
n=1 n=1

Again, equating coefficients, we get

n
(19) r nAn Rn1 = Cn
Rn+1
n
(20) r nBn Rn1 = Dn (n 6= 1)
Rn+1
D1
(21) r B1 = E0
R2
Combining these equations with those from the first boundary condition
above, we get, in each of the first two equations, the condition that either
r = 1 (impossible, since dielectric constants are always 1) or An = Cn =
DIELECTRIC CYLINDER IN UNIFORM ELECTRIC FIELD 4

0 and, for n 6= 1, Bn = Dn = 0. We are therefore left with the last equation


from each boundary condition, and we can solve these two equations to get

2E0
(22) B1 =
r + 1
r 1 2
(23) D1 = R E0
r + 1
We therefore get for the potentials

2E0
(24) V1 = r cos
r + 1
r 1 R2 E0
(25) V2 = cos E0 r cos
r + 1 r
From this we can get the field inside the cylinder

2E0
(26) V1 =
r + 1
dV1
(27) Ein =
d
2
(28) = E0
r + 1
Rather surprisingly, the field inside the cylinder is uniform.
P INGBACKS
Pingback: Dielectric sphere in uniform electric field
Pingback: Dielectric shell surrounding conducting sphere
Pingback: Dipole in dielectric sphere
DIELECTRIC SPHERE IN UNIFORM ELECTRIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.23.
We can work out the field inside a sphere of dielectric placed in a uniform
electric field in the same way we tackled the cylinder in the last post (for
those interested, this is done in Griffithss book). However, we can also use
a method of successive approximations to get the answer.
The starting point is to assume that the uniform exterior field E0 covers all
space, including the interior of the sphere. We then know that this uniform
field will cause the sphere to become polarized. Since were dealing with a
linear dielectric, this polarization is

(1) P0 = 0 e E0
However, a uniformly polarized sphere produces its own electric field,
which we need to add onto the original uniform field. We worked out the
field inside a uniformly polarized sphere earlier, so we can quote that result.
We get

1
(2) E1 = P0
30
e
(3) = E0
3
This new field produces more polarization, so we get

(4) P1 = 0 e E1
2
(5) = 0 e E0
3
Another round gives the next correction to the field:

e2
(6) E2 = 2 E0
3
Its fairly obvious that the general pattern is
1
DIELECTRIC SPHERE IN UNIFORM ELECTRIC FIELD 2

 n
e
(7) En = (1)n E0
3

so the total field is the sum of all these increments:

 n
n e
(8) E= (1) E0
n=0 3
The sum is a geometric series, so we can sum it using the standard for-
mula and we get

1
(9) E = E0
1 + e /3
3
(10) = E0
2 + r

where r = 1 + e is the dielectric constant. This agrees with the result


obtained using Legendre polynomials, as given in Griffiths.
P INGBACKS
Pingback: Magnetic field of a sphere in a uniform field
DIELECTRIC SHELL SURROUNDING CONDUCTING SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.24.
As an exercise in applying Laplaces equation to a problem with di-
electrics, suppose we have a conducting sphere of radius a surrounded by a
spherical shell of dielectric of outer radius b and susceptibility , with the
whole system in a uniform electric field E0 .
The general solution to Laplaces equation for the potential in spherical
coordinates is

 
n Bn
(1) V (r, ) = An r + n+1 Pn (cos )
n=0 r

where Pl is the degree-l Legendre polynomial.


Inside the sphere, the potential is constant (since the field is zero inside a
conductor), so we might as well take it to be zero.
In the region a < r < b, the solution is the general one given above. For
r > b, in order to avoid an infinite field for large r, we have to drop the Al
terms except for A1 since we need a potential that gives a constant field. If
we take the field to lie in the z direction, then since z = r cos = rP1 (cos )
we have for this region


Cn
(2) Vr>b = E0 rP1 + P (cos )
n+1 n
n=0 r
Note that weve used a different set of coefficients Cn here, since the
solution in this region is distinct from that for the region a < r < b. That is,
the coefficients Bn 6= Cn .
We can obtain conditions on the coefficients by equating terms of the
various Legendre polynomials in the boundary conditions. Continuity of
the potential at r = a gives the condition

Bn
(3) An an + =0
an+1
Continuity at r = b gives
1
DIELECTRIC SHELL SURROUNDING CONDUCTING SPHERE 2

B1 C1
(4) A1 b + 2
= E0 b + 2
b b
B n C n
(5) An bn + n+1 = n+1 (n 6= 1)
b b
Finally, we can use the condition at the boundary of two dielectrics to
get, since there is no free charge at the boundary:

V1 V2
(6) 1 = 2
n n
(n + 1) B n (n + 1)Cn
(7) An nbn1 n+2
= 0 (n 6= 1)
b bn+2
B1 C1
(8) A1 2 3 = 0 E0 20 3
b b
Consider first the terms for n 6= 1. We get

(9) Bn = a2n+1 An
a2n+1
 
n Cn
(10) An b n+1 = n+1
b b
Cn = An b2n+1 a2n+1

(11)
Since both Bn and Cn are proportional to An , the only way the third bound-
ary condition (at the dielectric/vacuum boundary) above can be satisfied is
if An = Bn = Cn = 0 for n 6= 1. We are therefore left with the n = 1 terms.
For these we get

(12) B1 = a3 A1
a3
 
C1
(13) A1 b 2 = E0 b + 2
b b
C1 = A1 b3 a3 + b3 E0

(14)
Plugging these into the third boundary condition gives

30 b3 E0
(15) A1 =
(b3 + 2a3 ) + 20 (b3 a3 )
30 a3 b3 E0
(16) B1 =
(b3 + 2a3 ) + 20 (b3 a3 )
In terms of the dielectric constant r = /0 we get
DIELECTRIC SHELL SURROUNDING CONDUCTING SPHERE 3

3b3 E0
(17) A1 =
r (b3 + 2a3 ) + 2 (b3 a3 )
3a3 b3 E0
(18) B1 =
r (b3 + 2a3 ) + 2 (b3 a3 )
The potential inside the dielectric shell is therefore

B1
(19) Va<r<b (r, ) = A1 r cos + cos
r2
3b3 E0 cos a3
 
(20) = r + 2
r (b3 + 2a3 ) + 2 (b3 a3 ) r
The field can be found from the gradient

(21)

E = V
(22)
V 1 V
= r
r r
(23)
3b3 E0 a3 a3
    
= 1 + 2 3 cos r + 1 + 3 sin
r (b3 + 2a3 ) + 2 (b3 a3 ) r r
This reduces to the situation of a conducting sphere in a uniform field
if we set r = 1 (effectively replacing the dielectric by a vacuum). In that
case, the potential reduces to

a3
(24) Vr =1 = E0 r cos + E0 cos
r2
The first term is just the applied field, and the second term arises from
the induced charge on the conductor.
POINT CHARGE EMBEDDED IN DIELECTRIC PLANE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.25.
The following is a problem that combines the theory of linear dielectrics
with the method of images. Suppose we divide 3-d space into two parts,
separated by the plane z = 0. The region z < 0 is filled with a dielectric
with constant r , while the region z > 0 is filled with a dielectric with a
different constant r0 . We also place a point charge q on the z axis at z = d.
This problem is similar to that of a point charge above a conducting plane
that served as the introductory example for the method of images. However,
in that case, we were able to replace the conductor by a single point charge
q at location z = d and then show that the resulting potential was valid
for z > 0. (Inside a grounded conductor, V = 0.) Here, the situation isnt
quite as simple.
The point charge will polarize both dielectrics with the result that there
will be bound surface charge where the dielectrics meet. We call this surface
charge b and b0 for z < 0 and z > 0 respectively. The surface charge is
related to the polarization by

(1) b = P n

and for a linear dielectric

(2) P = 0 e E

so at the boundary between the dielectrics, we get

(3) b = P n
(4) = 0 e Ez
(5) 0 = P0 n0
b
(6) = 0 e0 Ez0
We need to be careful with the directions of the various vectors here.
Suppose that q is positive. Then the induced polarization will be such that
1
POINT CHARGE EMBEDDED IN DIELECTRIC PLANE 2

negative charges are attracted to q, so that b will consist of negative charge


and b0 will consist of positive charge. That is, the polarization vector points
towards q in both cases.
In the calculation of P n the normal vector to the plane points upwards
but in the calculation of P0 n0 the normal vector points downwards, while
P0 still points in the same direction as P. That is the reason for the minus
sign in the result for b0 .
The electric field is discontinuous across a layer of surface charge, with
the discontinuity given by


(7) Ez(above) Ez(below) =
0

The fields due to the surface charges in this problem are therefore

(8) Ez0 = Ez(above)


b + b0
(9) =
20
(10) Ez = Ez(below)
b + b0
(11) =
20

The perpendicular component of the field due to the point charge is, at
z = 0:

1 q
(12) Ez(q) = 0
cos
4 r + d 2
2
1 qd
(13)
4 0 (r2 + d 2 )3/2

Here 0 = 0 r and reflects the fact that the electric field is reduced by the
factor of r inside a dielectric. The variable r is the distance from the origin
to a point on the z = 0 plane, as usual. The angle is that between the z
axis and a line from q to the point on the plane, so that cos = d/ d 2 + r2 .
This formula is valid on both sides of the boundary.
We now have enough information to write equations for the two surface
charges. We get
POINT CHARGE EMBEDDED IN DIELECTRIC PLANE 3

(14) b = 0 e Ez
" #
1 qd b + b0
(15) = 0 e
4 0 (r2 + d 2 )3/2 20
(16) b0 = 0 e0 Ez0
" #
+ 0
1 qd b
(17) = 0 e0 b
4 0 (r2 + d 2 )3/2 20
As a check at this stage, we observe that if we set e = e0 (that is, we
make the same dielectric fill all space, thus eliminating the boundary), and
then add these two equations together, we get

b + b0 = e b + b0

(18)
Since e 6= 0 in general, this means that b + b0 = 0 in this case. This
makes sense, since with no boundary between the dielectrics, we would
expect there to be no net surface charge.
Returning to the general case, we can solve these two simultaneous equa-
tions (by hand or using Maple), and get, using the relation 0 = 0 (1 + e0 )

e qd
(19) b = 3/2
2 (2 + e + e0 ) (r2 + d 2 )
e0 qd (1 + e )
(20) b0 = 3/2 0
2 (2 + e + 0 ) (r2 + d 2 ) (1 + e )
e
Again, we note that if e = e0 , b = b0 .
In general:

(e0 e ) qd
(21) b + b0 = 3/2
2 (2 + e + e0 ) (1 + e0 ) (r2 + d 2 )
At this stage, if we want to find the potential, we might try a direct inte-
gration, since

q b (r0 ) + b0 (r0 ) 0 0
(22) 40V (r) = + dx dy
|r d| |r r0 |

where r is the point at which we wish to find the potential, and the integra-
tion variable r0 ranges over the plane z = 0. However, I couldnt find any
way of doing this integral (Maple just got stuck) V .
POINT CHARGE EMBEDDED IN DIELECTRIC PLANE 4

Ive never liked the method of images since its largely informed guess-
work, but anyway...
If we work out the total bound surface charge we get

b + b0 rdr

(23) Q = 2
0
(e0 e ) q
(24) =
(1 + e0 ) (2 + e + e0 )
We pause here to look at a few limiting cases. If e0 = 0, so that the upper
region becomes a vacuum, we get

e
(25) Q= q
2 + e

which is the result given in Griffithss book for that case. Further, if we let
e , we get Q = q, which is the result for a point charge next to a
conducting plane. Also, if we let e0 , we get Q = 0, since in that case
weve embedded q inside a conductor, which shields it completely, so there
is no induced charge at the boundary.
Getting back to the method of images, we have to find the potential in
the two regions. For z > 0, we have a shielded charge with effective charge
q/r0 at z = d and we want to replace the surface charge with a point image
in the region z < 0. What should be the amount of this image charge? In
the case of the conducting plane, the image charge was the same amount as
the total surface charge, and at location z = d, so we can try that. In this
case, we get

1 q Q
(26) Vabove = q +q
40 0 2 2 2 2
r r + (z d) r + (z + d)
For z < 0, we still see the original charge at z = +d, and we can try
replacing the surface charge by a point charge the same distance beyond the
plane as we did when calculating the image for the z > 0 region. That is, we
place an image charge of size Q at z = +d (so it coincides with the original
point charge). In this case, the potential is

1 q/r + Q
(27) Vbelow = q
40
r2 + (z d)2
POINT CHARGE EMBEDDED IN DIELECTRIC PLANE 5

We can check that this potential satisfies the required boundary condi-
tions. First, it must be continuous at z = 0, which it obviously is:

1 q/r0 + Q
(28) Vabove (z = 0) =
40 r2 + z2
(29) = Vbelow (z = 0)
Second, we should be able to derive the discontinuity in the field above.
We have
 
Vabove d q
(30) = Q
z z=0 40 (r2 + d 2 )3/2 r0
 
Vbelow d q
(31) = +Q
z z=0 40 (r2 + d 2 )3/2 r0
Since Ez = V / z we get

Qd
(32) Ez(above) Ez(below) = 3/2
20 (r2 + d 2 )
(e0 e ) qd
(33) = 3/2
20 (r2 + d 2 ) (1 + e0 ) (2 + e + e0 )
b + b0
(34) =
0
So this checks out as well.
ENERGY OF CONDUCTING SPHERE IN A DIELECTRIC
SHELL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.4.3 & Problem 4.26.
Weve seen that the energy in a system containing dielectrics can be writ-
ten as


1
(1) W= (D E) d 3 r
2
This is the energy required to place the free charges, and includes the
energy needed to polarize the dielectric.
As a simple example of this formula, suppose we have a spherical con-
ductor of radius a that has a free charge Q on it, and we surround this con-
ductor with a spherical shell of linear dielectric that extends from r = a to
r = b. The free charge on the conductor will polarize the dielectric, resulting
in surface charges on the inner and outer surfaces of the dielectric.
We can find the displacement D from its relation to the free charge. That
is,


(2) D da = Q f
A

where Q f is the free charge (excluding the bound charge) that is contained
within the surface of integration.
In this case, Q f = Q and because the system has spherical symmetry, we
can take the surface of integration to be a sphere. Since the enclosed free
charge is Q for any surface with r > a, we get in this region:

(3) 4r2 D = Q
Q
(4) D = r
4r2
For a linear dielectric D = E = 0 (1 + e ) E. We therefore have for the
electric field:
1
ENERGY OF CONDUCTING SPHERE IN A DIELECTRIC SHELL 2

Q
(
40 (1+e )r2
r a<r<b
(5) E= Q
40 r2
r r>b
The energy is then

1
(6) W= (D E) d 3 r
2
" 2 #
b
4Q2 r2 dr r dr
(7) = 2 4
+
2 (4) 0 a (1 + e ) r b r4
Q2
   
1 1 1 1
(8) = +
80 1 + e a b b
Q 2 
1 e

(9) = +
80 (1 + e ) a b
If we remove the dielectric, this is the same as setting e = 0, so the
energy stored in the field of a charged conducting sphere is

Q2
(10) W=
80 a
ENERGY IN A DIELECTRIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.4.3 & Problem 4.27.
We now have a look at the analogous formula when we have free charges
embedded in a dielectric. To find the work required to assemble only the
free charges (excluding the bound charges), we start with the ideas devel-
oped originally for finding the energy required to add a point charge to an
existing configuration of charges. Assuming that the potential is zero at in-
finity and we bring in the charge q from there to point r, the energy required
to do this is W = qV (r).
In the continuous case, if we bring in a small amount f of free charge,
then the energy required to do this is

f (r) V (r) d 3 r
 
(1) W =

(Note that it is incorrect to start with the formula W = 21 (r)V (r) d 3 r,
since that formula was derived for a situation where we considered all the
charges (not just the free charges) already in place, so that the potential
in the formula is from the final distribution of charges. In our case, were
essentially retracing the derivation used originally for the case of adding a
point charge to an existing distribution and building up the formula from
there.)
Since D = f (see here), we can write this as

(2) W = [D]V (r) d 3 r

Using the vector calculus identity

(3) (AV ) = V A + A V

and the relation between field and potential E = V , we get



3
(4) W = [V D] d r + D Ed 3 r
1
ENERGY IN A DIELECTRIC 2

Using the divergence theorem, we can convert the first integral into a
surface integral and let the surface go to infinity, making the usual assump-
tion that the quantity V D goes to zero faster than 1/r2 , so that the surface
integral also goes to zero. We therefore end up with

(5) W = D Ed 3 r

Now for a linear dielectric, D = 0 r E, so we can write, for incremental


changes

(6) (D E) = 0 r (E E)
(7) = 0 r E 2
(8) = 20 r EE
(9) = 2E D
So if we pull the physicists trick of interchanging differentials and inte-
grals, we can write

1
(10) W = (D E) d 3 r
2
So in general, the work done to assemble free charge in a dielectric is

1
(11) W= (D E) d 3 r
2
Earlier we saw that the energy of a static charge distribution could be
written in terms of the electric field:

0
(12) W= E 2d3r
2
If we compare these two formulas they appear to be incompatible, since
if D = 0 r E, the first integral becomes

0 r
(13) W= E 2d3r
2

so that theres a factor of r in the first integral that is absent from the second.
The resolution of this dilemma is that we derived the dielectric formula
for free charge only. In the absence of a dielectric, all charge is free charge,
and r = 1 so the two formulas become the same. However, when we have
dielectric present, bringing in free charge will cause a polarization of the
ENERGY IN A DIELECTRIC 3

dielectric, which requires energy, since we are essentially pulling apart the
positive and negative bound charges in the dielectrics atoms. Thus the
energy required to place a free charge next to a dielectric is greater (by the
factor r ) than that required to place the same charge without the dielectric
being there.
As an example of the difference between the two formulas, consider
again the problem of the uniformly polarized sphere with polarization P.
We found that the field inside the sphere was uniform:

1
(14) Er<R = P
30
Outside the sphere, the potential is

R3
(15) Vr>R = P cos
30 r2
This is the same as the potential of a pure dipole with dipole moment

4
(16) p = R3 P
3
Thus the field outside the sphere is the same as that of a pure dipole, so
we get

p 
(17) Er>R = 2 cos r + sin
40 r3
Outside the sphere, the energy is the same no matter which formula we
use since there is no dielectric here. We therefore have

1
(18) Wr>R = 0 E 2 d 3 r
2
2
p2 0 r sin 2 2

(19) = 2 4 cos + sin d dr
2 (40 )2 R 0 r6
p2
(20) =
120 R3
4P2 R3
(21) =
270
Inside the sphere, since the field is constant, we have, using the total
charge formula
ENERGY IN A DIELECTRIC 4


0
(22) Wr<R = E 2d3r
2
2P2 R3
(23) =
270
The total energy calculated this way is thus

4P2 R3 2P2 R3
(24) Wtot = +
270 270
2
2P R 3
(25) =
90
Calculated using the free charge formula, we use the definition of dis-
placement as D = 0 E + P = P/3 + P = 2P/3. Then

1
(26) Wr<R = (D E) d 3 r
2
4P2 R3
(27) =
270
The total energy in this case thus comes out to Wtot = 0, which is because
there is no free charge in the problem. However, the derivation above also
relied on the dielectric being linear, which isnt the case in this problem
since we have a frozen in polarization with no external electric field. Thus
the second result doesnt really mean much anyway.
P INGBACKS
Pingback: Energy of conducting sphere in a dielectric shell
FORCE ON A DIELECTRIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Section 4.4.4 & Problem 4.28.
Weve seen that a conductor carrying some free charge experiences elec-
trostatic pressure due to the interaction of the free charges. A dielectric in
an external electric field also experiences a force from that field.
In the case of a parallel plate capacitor filled with dielectric, the field over
most of the plates area is perpendicular to the plates and thus there is no
net force on the dielectric since it is held in place between the plates and
the electric force is balanced by the mechanical force exerted by the plates.
However, at the edges of the plates, there is a fringing field which curves
outwards from the plates themselves, and this tends to pull the dielectric
further into the capacitor. If the dielectric fills the capacitor, then the forces
at all the edges balance each other and the dielectric experiences no net
force. However, if the dielectric only partially fills the space between the
plates, there is a net force tending to pull the dielectric further in.
A nice example of this is the following problem. We have two concentric
cylindrical tubes. The inner tube has radius a and the outer tube has radius b.
The tubes are placed vertically into a dish of dielectric oil with susceptibility
e and mass density , and a constant voltage of V is applied between the
two tubes. The oil will be attracted into the space between the tubes, so it
will tend to rise to a height h.
Weve already seen that the capacitance per unit length of a coaxial pair
of cylinders in a vacuum is

20
(1) C=
ln (b/a)
If the total length of the cylinder is L, and the height of the oil is h, then
the portions of the cylinder will have capacitances of

20 r
(2) Coil = h
ln (b/a)
20
(3) Cvac = (L h)
ln (b/a)
1
FORCE ON A DIELECTRIC 2

(Note that the capacitance of the region with the oil is increased by a
factor of the dielectric constant r = 1 + e .) The total capacitance is thus

20
(4) C = [r h + L h]
ln (b/a)
20
(5) = (e h + L)
ln (b/a)
Since were holding the voltage constant and the capacitance increases as
the oil rises, the amount of charge on the plates must also increase. That is,
we must transfer charge from the inner to the outer tube in order to maintain
a constant V . The amount of work required to transfer an amount of charge
dq through a potential difference V is V dq. There are thus two sources of
work done here: the energy required to make the oil rise between the plates,
and the energy required to transfer charge from one tube to the other.
If we choose instead to assume that the charge on the plates rather than
the voltage remains constant then no charge is transferred so no work is
done by the V dq term, which is now zero. In this case, the voltage would
change as the oil rises. However, since the oil would still rise, work is still
being done.
Since the energy stored in a capacitor is W = 12 CV 2 , if V is constant, the
change in energy if we move the oil a distance dh is

1 dC
(6) dW = V 2 dh
2 dh
Since the force we apply to move the oil is in opposition to the electrical
force F we get

dW
(7) F =
dh
1 2 dC
(8) = V
2 dh
However, in the constant voltage problem, we also have to move a charge
dq from one plate to the other, requiring work V dq. This adds another term
onto the force, which is, since q = CV :

dq
(9) Fq = V
dh
dC
(10) = V2
dh
This results in a total force of
FORCE ON A DIELECTRIC 3

(11) Ft = F + Fq
1 dC
(12) = V2
2 dh
Its worth looking at this in a bit more detail. The capacitance C increases
as the dielectric constant r increases, so if we keep the charge Q on the
capacitor constant as the oil rises (and thus increases the dielectric constant
over a greater range, increasing C) then since Q = CV , V must decrease. The
2
energy stored in the capacitor is W = 21 CV 2 = Q 2C therefore also decreases.
This energy goes into the work done to raise the oil.
However, if we keep V constant as the oil rises, then from Q = CV , Q
must increase as C increases. The energy stored in the capacitor is W =
1 2
2 CV and thus it increases. The extra energy is provided by an external
battery which adds charge to the capacitor in order to maintain a constant
voltage, as well as provide the energy to raise the oil. This is the extra V dq
term.
Returning to the problem of determining the height of the oil, the electri-
cal force on the oil is balanced by the gravitational force, so we get

1 2 dC
(13) V = mg
2 dh
1 2 20 e
= b2 a2 gh

(14) V
2 ln (b/a)
0 eV 2
(15) h =
g (b2 a2 ) ln (b/a)
P INGBACKS
Pingback: Conducting sphere half-embedded in dielectric plane
DIPOLE-DIPOLE FORCES AND TORQUES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.29.
Well return to the problem of the interaction of two dipoles that we con-
sidered a while back. We have two perfect dipoles p1 and p2 separated by
a distance r. Their alignment is such that p1 is perpendicular to the line
separating them (pointing upwards) and p2 is parallel to the line separating
them (pointing away from p1 ). The problem is to find the force that each
dipole exerts on the other. Weve seen that the force felt by a dipole in an
electric field is

(1) F = (p ) E
The field in each case is produced by the other dipole, and the formula
for such a field is

p  
(2) E= 2 cos r + sin
40 r3
Its important to be clear on the coordinate definitions in this formula.
This is a spherical coordinate system centred on the dipole producing the
field, with the vertical, or z, axis parallel to the dipole moment p. The angle
is the angle measured from the z axis.
First, lets consider the force on p2 due to p1 . In this case, we centre the
coordinate system on p1 . In this system, we can place p2 so that it points
along the +x axis. The force on p2 is therefore

(3) F2 = (p2 ) E1

where E1 is the field produced by p1 .


It turns out that its very difficult to work on this problem if we stay in
spherical coordinates, so well need to switch to rectangular coordinates. In
our coordinate system, p2 = p2 x, so p2 = p2 x . This means that we can
hold y and z constant in E1 . Holding z constant means that is constant at
= /2 (since were in the xy plane), and holding y constant means that
= 0 (since were restricting our attention to the x axis). In the xy plane,
1
DIPOLE-DIPOLE FORCES AND TORQUES 2

= z. Finally, r = x. Therefore, the field were interested in can be


written as

p1
(4) E1 = z
40 x3
From this, we get

3p1 p2
(5) F2 = z
40 x4
3p1 p2
(6) = z
40 r4
That is, the force on p2 is upwards.
Now for the force on p1 . This time, we centre the coordinates on p2 with
the z axis parallel to p2 . In this system, p1 is at position rz, and it points
in the x direction, so p1 = p1 x. Unfortunately, at this stage things dont
simplify as much as in the previous case. We get p1 = p1 x , which
means we want to hold y and z constant, as before, but because p1 is at rz,
varying x will cause changes in both r and (although at least we can keep
= 0 constant). There may be a simpler way of doing this, but one way
that does seem to work is to express the formula for E above entirely in
rectangular coordinates.
The standard formulas for converting the unit vectors are:

(7) r = sin cos x + sin sin y + cos z


(8) = cos cos x + cos sin y sin z
If we set = 0 and substitute these formulas into the formula for E2 , we
get

p2
(9) E2 = [2 cos (sin x + cos z) + sin (cos x sin z)]
40 r3
p2  2
 
(10) = 3 sin cos x + 3 cos 1 z
40 r3
In the case = 0, we have

x
(11) sin =
r
z
(12) cos =
r

so:
DIPOLE-DIPOLE FORCES AND TORQUES 3

p2  2 2
 
(13) E2 = 3xzx + 3z r z
40 r5
p
Using the product rule and r = x2 + y2 + z2 , we get

(14)
5p1 p2 x  p1 p2
3xzx + 3z2 r2 z
 
F1 = p1 E2 = 7
[3zx 2xz]
x 40 r 40 r5
3p1 p2  2
5x r2 zx + 5z2 r2 xz
  
(15) = 7
40 r
We need to evaluate this at x = y = 0; z = r, so we get

3p1 p2
(16) F1 = x
40 r4
With the orientation of our coordinate system, the +x direction is down-
wards, so F1 = F2 , which is consistent with Newtons third law of equal
and opposite forces.
We can now work out the torques on the two dipoles relative to each other.
In our original analysis of this problem we found that the torques exerted
by one dipole on the other were not symmetric, with the torque on p1 being
twice that on p2 . However, each torque was calculated relative to the centre
of the respective dipole, so we werent comparing like with like. What we
really need to do is compare the torques on the dipoles with respect to the
same point. For example, since we worked out the torque on p1 relative to
its centre as

2
(17) N1 = p1 p2
40 r3

we need to work out the torque on p2 , also relative to the centre of p1 .


A quick reminder of how to transform torques from one axis to another.
Suppose that the dipole is slightly non-ideal, in that there is a finite separa-
tion between the two charges. Call the vector from the centre of the dipole
(the midpoint between the charges) to one of the charges ra and the vector
from the centre to the other charge rb . If charge a has a force Fa acting on
it, and charge b has force Fb , then the torque on the dipole is

(18) Ncentre = ra Fa + rb Fb
DIPOLE-DIPOLE FORCES AND TORQUES 4

Now suppose we want the torque due to these same two forces, but about
a different point O. Call the vector from O to the midpoint of the dipole s.
Then the vectors from O to charges a and b are

(19) sa = s + ra
(20) sb = s + rb
The torque relative to O is therefore

(21) NO = sa Fa + sb Fb
(22) = s (Fa + Fb ) + ra Fa + rb Fb
(23) = s (Fa + Fb ) + Ncentre
That is, the torque is the sum of the original torque and a new torque
consisting of the vector product of s and the total force acting on the dipole.
In terms of our ideal dipoles, the torque on p2 relative to p1 is then

(24) N21 = r F2 + N2
In our earlier analysis, we found that

1
(25) N2 = p1 p2
40 r3
Using the coordinate system centred on p1 ,

p1 p2
(26) N2 = y
40 r3
3p1 p2
(27) r F2 = y
40 r3
2p1 p2
(28) N21 = y
40 r3
We can see that N21 = N1 so that when we measure the torques acting
on the two dipoles with respect to the same point, they are in fact equal and
opposite, as we would expect.
P INGBACKS
Pingback: Dipole-dipole interactions
DIPOLE BETWEEN TWO ANGLED CONDUCTING PLANES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.30.
Suppose we have two conducting planes that make angles of + and
respectively with the xy plane, and meet along the y axis (although they are
insulated from each other). The top plane is held at a potential of +V and
the bottom plane at V .
Since the electric field is always normal to the surface of a conductor,
the field lines will bulge outwards in the +x direction and have a vertical
tangent (that is, parallel to the z axis) when they pass through the xy plane.
Since the field lines travel from positive to negative, the field lines start on
the top plane and arc towards the bottom plane.
Now suppose we place a dipole between the planes. The dipole is centred
on the xy plane and the dipole moment p points in the +z direction. What
force does this dipole feel?
This depends on whether we are talking about a physical dipole (that is,
two charges separated by a finite distance) or an ideal dipole, in which the
separation is zero. Weve seen that the force felt by an ideal dipole in an
electric field is

(1) F = (p ) E
Since p is parallel to z, F = p z E and since E is tangent to the vertical at
the location of the dipole, E/ z = 0 and the force is zero.
If were talking about a physical dipole, then the positive charge is slightly
above the xy plane where the field points slightly to the right of z, so the
force on the positive charge will be in that direction. Similarly, the negative
charge is slightly below the xy plane, where the field points slightly to the
left of z. The force on the negative charge will thus be in the opposite
direction, or slightly to the right of +z. From the symmetry of the problem,
the vertical components of force cancel and the dipole feels a net force to
the right.

1
DIELECTRIC CUBE: BOUND CHARGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.31.
Another example of the bound charge due to polarization in a dielectric.
Suppose we have a dielectric cube of side length a centred at the origin,
with a frozen-in polarization of P = kr, for some constant k. The bound
charges induced by this polarization are

(1) b P n
(2) b 0 P

The volume bound charge density is

(3) b = k r
(4) = 3k

which is found by expressing r in rectangular coordinates.


Since the charge density is constant, the total volume bound charge is

(5) Qv = 3ka3

For the surface charge density, by symmetry all 6 faces will be the same,
so we can look at the upward facing side of the cube. Here

(6) b = kr n
a
(7) = k
2

since the dot product is just the projection of r onto the z axis which for the
upper face, is always a/2. Again, the surface charge density is a constant,
so the charge on one face is k 2a a2 and the total charge on the surface of the
cube is
1
DIELECTRIC CUBE: BOUND CHARGES 2

a
(8) Qs = 6k a2
2
(9) = 3ka3
The total bound charge is zero, as required.
POINT CHARGE IN DIELECTRIC SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.32.
Yet another example of the bound charge due to polarization in a dielec-
tric. Suppose we have a point charge at the centre of a sphere of linear
dielectric (susceptibility e ) of radius R. We can begin the solution by cal-
culating the displacement, since the system has spherical symmetry. We
know that

(1) Dd 3 r = Q f
V

(2) = D da
A

where Q f is the free charge enclosed by the surface of integration. Here we


can take the surface to be a sphere of radius r centred at the point charge,
which is the only free charge in the problem. Therefore

(3) D da = 4r2 D(r)
A
(4) = q
q
(5) D (r) = r
4r2
For a linear dielectric we have

(6) D = 0 (1 + e ) E

so we can get the field


( q r
40 (1+e ) r2 0<r<R
(7) E= q r
40 r2 r>R
The polarization, again because were dealing with a linear dielectric, is
1
POINT CHARGE IN DIELECTRIC SPHERE 2

(8) P = 0 e E
e q r
(9) =
4 (1 + e ) r2
The bound charges induced by this polarization are

(10) b P n
(11) b P
At the surface of the sphere, n = r and we get

e q
(12) b =
4 (1 + e ) R2
(13) Qs = 4R2 b
e q
(14) =
(1 + e )
For the bound charge, we make use of the formula for the three-dimensional
delta function:
 
r
(15) 2 = 43 (r)
r
The bound charge is therefore

(16) b = P
e q
(17) = 3 (r)
(1 + e )
The total volume bound charge is the integral of this over the sphere,
which is just

e q
(18) Qv = 3 (r)d 3 r
V (1 + e )
e q
(19) =
(1 + e )
All the volume bound charge is concentrated at the centre, and the total
bound charge is Qv + Qs = 0 as required.
ELECTRIC DISPLACEMENT: BOUNDARY CONDITIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.33.
We can work out analogs of the boundary conditions on the electric field
in the case of the displacement vector D. In general

(1) D = 0 E + P

where P is the polarization density.


Weve already seen that

(2) D = f

where f is the free charge density. To apply this to a boundary, suppose we


have a boundary between two dielectric materials, with a surface free charge
density f at this boundary. Using the same argument as in the case of E,
we can apply Gausss law over a small box that straddles the boundary and
whose surfaces on each side of the boundary are parallel to the boundary
layer itself. Well find that

(3) Dabove
Dbelow
= f
For the parallel component of D, we look at the line integral around a
rectangle that straddles the boundary, just as we did with E. We know
from Stokess theorem that since E = 0 in all electrostatic problems,
the parallel component of E is continuous across the boundary. However,
in general P 6= 0, so we cant conclude that the parallel component of
D is continuous, and we get

(4) Dabove
k Dbelow
k = Pkabove Pkbelow
As an example, we can work out how much the electric field bends when
we cross a boundary between two linear dielectrics, assuming there is no
free charge at the boundary. Without free charge, we have from 3:
1
ELECTRIC DISPLACEMENT: BOUNDARY CONDITIONS 2

(5) Dabove
= Dbelow

above below
(6) 0 a E = 0 b E

where a,b is the dielectric constant on either side of the boundary.


For the parallel component, we get from 4 (remember were dealing with
a linear dielectric):

(7) 0 a Ekabove 0 b Ekbelow = 0 a Ekabove 0 b Ekbelow


(8) (1 + a ) Ekabove (1 + b ) Ekbelow = a Ekabove b Ekbelow
(9) Ekabove = Ekbelow
Thus the second condition just reproduces the original condition on the
continuity of the parallel component of the field, so it doesnt really tell us
anything new.
We can express the components of the field in terms of the angle between
E and the normal to the boundary on either side, which well call a and b .
On the above side, we have

above
(10) E = E above cos a
(11) Ekabove = E above sin a

with similar conditions on the below side, so we can rewrite the equations
above as

above below
(12) a E = b E
(13) a E above cos a = b E below cos b
(14) E above sin a = E below sin b
Dividing the third equation by the second, we get

tan a a
(15) =
tan b b
A larger dielectric constant means a larger tangent and thus a larger angle
with the normal to the surface, so the field tends to spread out when entering
a dielectric with a higher constant.
ELECTRIC DISPLACEMENT: BOUNDARY CONDITIONS 3

P INGBACKS
Pingback: Boundary between magnetic materials
Pingback: Maxwells equations in matter: boundary conditions
DIPOLE IN DIELECTRIC SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.34.
Another example of solving a boundary value problem in a system with
linear dielectrics. Suppose we have an ideal dipole p at the centre of a
sphere of dielectric (dielectric constant r ) and radius R. What is the poten-
tial at any point (inside or outside the sphere)?
We can use a similar approach to that of the problem of a dielectric cylin-
der in an electric field. We first specify the boundary conditions. At the
surface of the sphere, the potential must be continuous, so we have

(1) Vin (R) = Vout (R)


As we saw in the cylinder problem, the condition on the normal derivative
of the potential is, since on the outside of the sphere, = 0 :

Vin Vout
(2) = 0
r r=R r r=R
Since there is no external field, we must have V 0 as r . Finally, as
r 0, the potential must behave like that of an ideal dipole, so, assuming
that p points in the +z direction:

p cos
(3) lim Vin (r) =
r0 4r2
Note that were using rather than 0 in this formula, since the dipole
is inside the dielectric and the potential is reduced by a factor of r , where
= 0 r .
With these conditions, we must solve Laplaces equation in spherical co-
ordinates, so we can quote the general form of the solution:

 
l Bl
(4) V (r, ) = Al r + l+1 Pl (cos )
l=0 r
Inside the sphere, we have
1
DIPOLE IN DIELECTRIC SPHERE 2


p cos
(5) Vin = 2
+ Al rl Pl (cos )
4r l=0

Outside, we have


B
(6) Vout = rl+1l Pl (cos )
l=0

Applying the continuity condition 1 and equating coefficients of the Pl ,


we get

p B1
(7) 2
+ A1 R =
4R R2
Bl
(8) Al Rl = (l 6= 1)
Rl+1
We can therefore express the Bl in terms of Al :

p
(9) B1 = + A1 R3
4
(10) Bl = R2l+1 Al (l 6= 1)

Using the condition on the derivatives 2 and equating coefficients as be-


fore, we get

p 2B1
(11) + A 1 = 0
2R3 R3
(l + 1) Bl
(12) Al lRl1 = 0 (l 6= 1)
Rl+2
Substituting for Bl into the second equation gives

(13) Al lR2l+1 = 0 (l + 1) Al R2l+1

The only way this can be satisfied is if Al = 0 for l 6= 1, so we get Al =


Bl = 0 for l 6= 1.
For the l = 1 case, we can solve the two equations in A1 and B1 to get
(using = 0 r ):
DIPOLE IN DIELECTRIC SPHERE 3

 
p 0
(14) A1 =
2R3 + 20
 
2p r 1
(15) =
4R3 r + 2
 
p r 1
(16) B1 = 1+2
4 r + 2
 
p 3
(17) =
40 r + 2
The potential is
 
p cos 2pr cos r 1
(18) Vin (r) = +
4r2 4R3 r + 2
r 3 r 1
  
p cos
(19) = 1+2 3
4r2 R r + 2
 
p cos 3
(20) Vout (r) =
40 r2 r + 2
UNIQUENESS OF POTENTIAL IN DIELECTRICS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.35.
When we considered electric fields in a vacuum, we found that if we
specify the charge distribution in some region V , and also specify either
the potential or its normal derivative on the boundary of that region, then
the potential inside V is unique. Here, well show that uniqueness of the
potential also applies to the case where V contains some linear dielectric.
The assumptions we make are:
(1) The potential V is specified on all boundaries of V .
(2) The free charge distribution f is specified everywhere within V .
(3) The distribution of dielectric within V is fixed, and all dielectric
constants are specified.
The proof follows a similar line of reasoning to that used in the electric field
case. As before well suppose that there are two distinct potentials V1 and
V2 that satisfy the conditions. Well also define the displacements due to
these potentials as D1 and D2 . Now we consider the difference between
these two solutions, so we have V3 V1 V2 and D3 D1 D2 . Now we
look at this volume integral and convert it to a surface integral in the usual
way:

3
(1) (V3 D3 ) d r = V3 D3 da
V A
By assumption, on the surface A, V3 = V1 V2 = 0 since the potential is
specified everywhere on the boundary. Therefore:

(2) (V3 D3 ) d 3 r = 0
V
Now we expand the integrand using a standard theorem from vector cal-
culus:

(3) (V3 D3 ) = D3 V3 +V3 D3


From the formula for the divergence of the displacement:
1
UNIQUENESS OF POTENTIAL IN DIELECTRICS 2

(4) D3 = (D1 D2 )
(5) = f f
(6) = 0
This follows, since the free charge distribution was fixed by assumption.
Therefore we have

(7) (V3 D3 ) = D3 V3
For a linear dielectric, D = E and in general, E = V , so

(8) (V3 D3 ) = D3 V3
(9) = E32
(10) = |E1 E2 |2
Thus the volume integral becomes

(11) 3
(V3 D3 ) d r = |E1 E2 |2 d 3 r
V V
(12) = 0
(Note that we cant take outside the integral since in general it varies
over the volume, depending on what dielectrics are present.)
Now the integrand is non-negative everywhere, since the permittivity
0 so the only way the integral can be zero is if E1 = E2 everywhere inside
V . This means that V1 V2 = k for some constant k, but since V1 = V2 on
the boundary, we must have k = 0 and V1 = V2 everywhere.
P INGBACKS
Pingback: Conducting sphere half-embedded in dielectric plane
CONDUCTING SPHERE HALF-EMBEDDED IN DIELECTRIC
PLANE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.36.
An interesting special case of a dielectric problem is one in which a con-
ducting sphere is partially embedded in an infinite half-space of dielectric.
That is, we have a metal sphere of radius R with its centre at the origin, and
the space z < 0 is completely filled with a linear dielectric with suscepti-
bility e , so that the lower hemisphere of the conductor is embedded in the
dielectric with the top hemisphere in vacuum.
If we hold the sphere at a potential of V0 then in the absence of the di-
electric, the potential would be
(
V0 r<R
(1) V (r) =
V0 Rr r>R
From this we get the field (which is zero inside the sphere of course):

(2) E = V
V0 R
(3) = r (r > R)
r2
For a linear dielectric, the polarization is (for z < 0):

(4) P = 0 e E
V0 R
(5) = 0 e 2 r
r
The volume bound charge induced by this is

(6) b = P
(7) = 0
(since r/r2 = 43 (r) and we are considering only points with r >

R so we avoid the singularity at the origin).
The surface bound charge is b = P n and is zero on the plane z = 0
since P n. That is, since P is parallel to r and r is horizontal across the
1
CONDUCTING SPHERE HALF-EMBEDDED IN DIELECTRIC PLANE 2

plane z = 0, all the polarization is parallel to the plane so there is no bound


surface charge.
On the lower hemisphere, we consider the surface of the dielectric which
has a surface normal pointing towards the origin, which means that n = r
so on this surface

(8) b = P n
V0
(9) = 0 e
R
In order for the sphere to be maintained at a constant potential, the surface
charge density on the lower hemisphere would have to increase to compen-
sate for this induced bound charge on the dielectric. This is the same argu-
ment as that used when we increase the capacitance of a capacitor by putting
dielectric between its plates. Because the dielectric cancels out some of the
field between the plates, we need to put more charge on the plates in order
to maintain the same potential difference between the plates. In this case,
were trying to maintain the same potential difference between the sphere
and infinity (where V = 0), so we need to increase the charge on the lower
hemisphere to maintain it at V0 .
In the absence of dielectric, we can work out the surface charge density
on the sphere from the field, since across a surface layer of charge


(10) E =
0
V0
Since E = 0 inside and E = R r outside, we have for the upper hemi-
sphere

0V0
(11) =
R
In the lower hemisphere, the net surface density must be the same, so

0V0 V0
(12) = + 0 e
R R
V0
(13) =
R

where = 0 (1 + e ) = 0 r . That is, the net charge density is the same as if


the dielectric was not present at all. Thus by the uniqueness of the potential
in systems containing dielectrics, the potential for the half-embedded sphere
is the same as that for a sphere without any dielectric.
CONDUCTING SPHERE HALF-EMBEDDED IN DIELECTRIC PLANE 3

This argument hinges on the fact that the boundary of the dielectric lies
on a surface that is parallel to the field (the plane z = 0 here). Thus we could
use the same argument for a sphere embedded in any dielectric in which the
surface of the dielectric lay entirely parallel to the field, for example a cone
centred at the origin. It would not work for, say, a sphere embedded in a
dielectric plane that didnt split the sphere into two equal hemispheres.
FORCE ON A DIELECTRIC SPHERE DUE TO A CHARGED
WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.37.
The force on an ideal dipole in an electric field is

(1) F = (p ) E
In a dielectric with polarization density P (r), we can integrate this to get
the total force on the dielectric due to an external field:

(2) F= (P ) Ed 3 r
V

where V is the volume occupied by the dielectric.


In the case of a sphere of linear dielectric of radius R situated a distance
z from an infinite wire carrying charge density , we can use this formula
to work out the force on the sphere. Well assume that the sphere is very
small, so that R  z.
The field due to the wire along a line perpendicular to the wire and pass-
ing through the centre of the sphere (which were taking as the z axis) is

1 2
(3) E= z
40 z

Assuming that the sphere is small enough, we can approximate this situation
by assuming that the field is constant over the sphere. In that case, we can
use the formula for the field inside a dielectric sphere in a uniform field:

3
(4) Ein = E
2 + r
3 1 2
(5) = z
2 + r 40 z
Since the dielectric is linear
1
FORCE ON A DIELECTRIC SPHERE DUE TO A CHARGED WIRE 2

(6) P = 0 e Ein
6 e
(7) = z
4 (2 + r ) z
The integrand above is then

3 2 e
(8) (P ) E = z
4 2 0 (2 + r ) z3
By assuming this is constant over the sphere, we get the force as

4R3 3 2 e
(9) F z
3 4 2 0 (2 + r ) z3
e 2 R3
(10) = z
0 (3 + e ) z3
The force is attractive as youd expect since the positive charge on the
wire would induce a negative surface charge on the side of the sphere facing
the wire.
RELATION BETWEEN POLARIZABILITY AND
SUSCEPTIBILITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.38.
The polarization density of a linear dielectric can be written as

(1) P = 0 e E

where E is the macroscopic field applied to the dielectric. From a micro-


scopic point of view, the dipole moment of an atom is

(2) p = Elocal

where Elocal is the external field at the location of the atom. The differ-
ence between E and Elocal is that Elocal excludes the field produced by the
atomic dipole. We can use this distinction to work out the relation between
, the polarizability (a microscopic quantity), and e , the susceptibility (a
macroscopic quantity).
The average field for a charge distribution within a sphere of radius R
can be written in terms of its dipole moment, so if we assume that an atom
occupies a sphere with this radius, we have for the local field due to the
atom:

1
(3) Eatom = p
40 R3
The total field at the location of the atom is then

(4) E = Elocal + Eatom


1
(5) = Elocal p
40 R3
 

(6) = 1 Elocal
40 R3
1
RELATION BETWEEN POLARIZABILITY AND SUSCEPTIBILITY 2

We can write this in terms of the number density of atoms. If one atom
occupies a sphere of radius R then the number of atoms per unit volume is

3
(7) N=
4R3

so
 
N
(8) E = 1 Elocal
30
From this we can get the relation between and e , since the polarization
density is

(9) P = Np
(10) = NElocal
(11) 0 e E = NElocal
 
N
(12) 0 e 1 Elocal = NElocal
30
N/0
(13) e =
1 N/30
In terms of the dielectric constant r = 1 + e , so

N/0
(14) r 1 =
1 N/30
N N
(15) r 1 (r 1) =
30 0
30 r 1
(16) =
N r + 2
This last equation is known as the Clausius-Mossotti formula, and can
be used to work out the polarizability if the dielectric constant and number
density are known. The dielectric constant can usually be measured fairly
easily, and at least for gases, the number density can be found from the ideal
gas law.
P INGBACKS
Pingback: Clausius-Mossotti formula - examples
Pingback: Susceptibility of a polar dielectric
CLAUSIUS-MOSSOTTI FORMULA - EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.39.
The Clausius-Mossotti formula gives the atomic polarizability in terms
of the dielectric constant for a linear dielectric. We can check the formula
for a few gases. The formula is

30 r 1
(1) =
N r + 2
Griffiths gives the dielectric constants for some gases at a pressure of 1
atm and temperature of 20C = 293K. We need the other two values to plug
into the formula. We have:

(2) 0 = 8.85418782 1012 Farads/m


We can use the ideal gas law PV = nRT to find the number of molecules
in a cubic metre. Using R = 8.205746 105 m3 atm K 1 mol 1 and V =
1m3 , we get, using Avogadros number 6.0221415 1023

PV
(3) n =
RT
1
(4) =
293 8.205746 105
(5) = 41.592 moles/m3
(6) N = 2.505 1025 molecules/m3
Using these values, we get, using numbers from Griffithss book:
Gas C M Measured
He 0.2065 0.205
H 0.794 0.667
Ne 0.413 0.396
Ar 1.652 1.64
(Values are for /40 in units of 1030 m3 .) The values arent a brilliant
match, but theyre not bad considering its a fairly crude calculation.

1
SUSCEPTIBILITY OF A POLAR DIELECTRIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 4.40.
The Clausius-Mossotti formula gives the atomic polarizability in terms
of the dielectric constant for a linear dielectric. This formula assumes that
the entire polarization is caused by an external electric field, and that the
atoms of the dielectric possess no intrinsic polarization. Many substances,
such as water, do possess an intrinsic polarization, so we need a different
formula for them.
The energy of a dipole in an external electric field is

(1) u = p E
From statistical mechanics (which I hope to get to eventually), we know
that for a material in thermal equilibrium at temperature T , the probability
of a molecule having energy u is proportional to eu/kT , where k is the
Boltzmann constant. The average energy of molecules in such a system is
then
u/kT
ue du
(2) hui = u/kT
e du

where both integrals are taken over all possible energies in the system.
In the case of a dipole in a field, the energy can range from pE to +pE,
so these are the limits on the integrals. If we write the energy in terms of
the angle between p and E, we have u = pE cos and

(3) du = pE sin d

Making this substitution, the limits on the integral become 0 to and we


have

p2 E 2 e pE cos /kT cos sin d
(4) hui = 0
pE 0 e pE cos /kT sin d
1
SUSCEPTIBILITY OF A POLAR DIELECTRIC 2

The bottom integral is easily done straight away, and the top integral can
be done by parts (integrate e pE cos /kT sin and differentiate cos ), so we
get
   
1+e2pE/kT pE + 1 e2pE/kT kT
(5) hui =
e2pE/kT 1
Multiplying top and bottom by epE/kT and converting to hyperbolic
functions, we get
 
pE
(6) hui = kT pE coth
kT
If there are N molecules per unit volume, then the polarization density is

(7) P = N hpi
The only component of p that contributes to the energy is the component
that is parallel to E. Suppose we look only at dipoles that make an angle
with E, so that the energy of such a dipole is u = pE cos . Now consider
the projection of one of these dipoles into the plane perpendicular to E. If
we take E to be along the z axis, this projection lies in the xy plane, has a
magnitude pE sin and makes an angle with the x axis. Since has no
effect on the energy, all values of are equally likely for each particular
value of . As a result, we would expect hpi to be parallel to E, since
all perpendicular components average out to zero. Therefore, since the net
polarization density is N hpi, it is parallel to E, so we have

(8) N hui = N hp Ei
(9) = hP Ei
(10) = PE
N hui
(11) P =
E  
pE NkT
(12) = N p coth
kT E
   
pE kT
(13) = N p coth
kT pE
 
P pE kT
(14) = coth
Np kT pE
This is known as the Langevin formula.
SUSCEPTIBILITY OF A POLAR DIELECTRIC 3

A plot of P/N p against pE/kT looks like this:

The curve is sigmoidal (S-shaped) and tends to 1 for large fields and/or
low temperatures (the sign being determined by the sign of p), indicating
that polarization tends to become complete in these cases as you might ex-
pect.
In the region where pE/kT is small, we can approximate the RHS of the
formula by expanding it in a series. The series for coth x can be looked up
(or you can work it out from the series for sinh x and cosh x) and is

1 x x3
(15) coth x = + +...
x 3 45
From this we get, to first order in pE/kT :

 
pE
(16) P ' Np
3kT
NE p2
(17) =
3kT
In this region, the material behaves like a linear dielectric (since P E),
so we can derive an expression for the susceptibility:
SUSCEPTIBILITY OF A POLAR DIELECTRIC 4

(18) P = 0 e E
N p2
(19) e =
30 kT
For water, we can work out estimates for the dielectric constant. We need

(20) 0 = 8.85418782 1012 Farads m1


(21) k = 1.3806503 1023 m2 kg s2 K1
(22) pH2 O = 6.1 1030 C m
For water at 20C = 293 K, we need the molecular density N. Waters
mass is 18.0153 g/mole and 1 m3 of water weighs 1000 kg, so contains
55508 moles. To get the molecular density, we multiply by Avogadros
number 6.0221415 1023 , so we have

(23) N = 3.343 1028


(24) r = e + 1
(25) = 12.58
The measured value is 80.4, so this isnt a particularly accurate estimate.
For water vapour at 100C = 373 K and 1 atm pressure we can use the
ideal gas law PV = nRT to find the number of molecules in a cubic metre.
Using R = 8.205746 105 m3 atm K 1 mol 1 and V = 1m3 , we get

PV
(26) n =
RT
1
(27) =
373 8.205746 105
(28) = 32.672 moles/m3
(29) N = 1.968 1025 molecules/m3
(30) r = 1.00535
This time, the measured value is 1.00587, so the answer isnt too far off.
MAGNETOSTATICS - THE LORENTZ FORCE LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.1.
Up to now, weve looked only at problems in electrostatics, that is, situa-
tions where we have a stationary distribution of charge.
The other half of electromagnetism is, of course, magnetism, so well
start by looking at magnetostatics, which is the study of static magnetic
fields (possibly mixed in with static electric fields).
The starting point in the study of magnetism is the Lorentz force law,
which states that a charge q moving at velocity v through a magnetic field
B feels a force of

(1) F = qv B
This law is not derived; it is merely an expression of what is observed in
experiments.
Magnetism is an unusual force in several ways. First, it acts on a charge
only if the charge is moving relative to the field. Second, it produces a
force perpendicular both to the field and the direction of motion. Third, as
a consequence of the second point, a magnetic force cannot do any work.
This is because work is defined as the integral of F dl, in which only the
component of force in the direction of motion appears. Since the magnetic
force is always perpendicular to the direction of motion, F dl = 0 always,
so no work is done. The magnetic force can therefore change only the
direction of motion, and not its speed.
As a simple example of this, suppose we send a charged particle into a re-
gion with a constant magnetic field. The particles velocity is perpendicular
to the field. The particle will feel a constant force of magnitude qvB per-
pendicular to its direction of motion, so that this force acts as a centripetal
force, and the particle moves in a circle. Equating these two forces, we get

mv2
(2) qvB =
r
This is known as the cyclotron formula, since it describes the principle
used in a cyclotron, where charged particles are made to travel in circles by
shooting them between the poles of a large electromagnet.
1
MAGNETOSTATICS - THE LORENTZ FORCE LAW 2

Using this formula, the radius of the circle is

mv
(3) r=
qB
The momentum of the particle can be expressed in terms of the radius,
charge and field:

(4) p = mv
(5) = qrB
Suppose the field points along the +y direction and the particle starts off
moving in the +x direction. After moving in the field a horizontal distance
a, we find the particle is deflected in the z direction by a vertical distance
d. By applying the right hand rule for cross products, we find that the
particle must be positively charged. We can also work out its momentum
(and thus its mass, assuming we measure its speed) if we can find the radius
of the circular path.
If we draw a triangle with one vertex at the centre of the circle, another
where the particle has reached its displacement of d below the xy plane, and
then draw a horizontal line from the second vertex to intersect the z axis, we
get a right angled triangle with sides of r, r d and a. From Pythagoras, we
have

(6) r2 = (r d)2 + a2
(7) 0 = 2rd + d 2 + a2
a2 + d 2
(8) r =
2d
Therefore, the momentum is

(9) p = qrB
qB 2
a + d2

(10) =
2d
P INGBACKS
Pingback: Cycloid motion in crossed electric and magnetic fields
Pingback: Currents in magnetic fields
Pingback: Divergence of magnetic field: magnetic monopoles
Pingback: Currents and relativity
Pingback: Hall effect
Pingback: Radially symmetric magnetic field
MAGNETOSTATICS - THE LORENTZ FORCE LAW 3

Pingback: Force between current loops: Newtons third law


Pingback: Torque on a magnetic dipole
Pingback: Changing a particles speed in a cyclotron
Pingback: Accelerating an atomic electron with a changing magnetic
field
Pingback: Poyntings theorem
Pingback: Electromagnetic field tensor: conservation of mass
Pingback: Electromagnetic field tensor: change in kinetic energy
Pingback: Electromagnetic field tensor: justification
Pingback: Gravitomagnetic acceleration is perpendicular to velocity
Pingback: Gravitoelectric and gravitomagnetic densities
Pingback: Gravitoelectric and gravitomagnetic acceleration for a moving
wire
CYCLOID MOTION IN CROSSED ELECTRIC AND MAGNETIC
FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.2.
One of the classic examples of how mixing electric and magnetic fields
can produce unusual results is the case of cycloid motion. The idea is that
we have a constant magnetic field pointing in the x direction, and a constant
electric field perpendicular to the magnetic field. We might as well align
the electric field with the z axis.
If we place a particle with charge q at the origin and either give it an
initial velocity perpendicular to B or start it at rest, then neither the electric
nor the magnetic field will exert a force in the x direction, so the motion is
restricted entirely to the yz plane.
Combining the Lorentz force law F = qvB and electrostatics with New-
tons law F = ma we get, using a dot to denote a derivative w.r.t. time:

(1) my = qBz
(2) mz = qE qBy
At this point, we can note that its quite easy to generalize this system
so that E has an arbitrary direction perpendicular to B. However, theres
not much point in doing so, since this amounts merely to a rotation of the
coordinates about the x axis. Giving E a component along x, however, does
introduce significant complexity into the system, since we then have a force
component along x, so we need another differential equation in the above
set.
We can solve the system of 2 ODEs above by using software (or by hand,
if youre familiar with methods of solving systems like this), and we get

c3 c2 E
(3) y (t) = cos t + sin t + t + c4
B
c3 c2 E
(4) z (t) = sin t + cos t + + c1
B

where qB m and the ci s are constants of integration. Note that has


units of inverse time, as we can see from the Lorentz force law above. qB
1
CYCLOID MOTION IN CROSSED ELECTRIC AND MAGNETIC FIELDS 2

has units of force/velocity, which has units of mass time1 . (For those
reading Griffithss book, the constants in our solution can be redefined to
produce Griffithss equation 5.6, since E, B, q, m and are all constants
here.)
The four constants can be determined by specifying the initial position
and velocity of the particle. If it starts at rest at the origin (that is, x (0) =
y (0) = x (0) = y (0) = 0), we can apply these conditions to the solution
(again, using software to ease the mathematics), and get

E E
(5) y (t) = sin t + t
B B
E E
(6) z (t) = cos t +
B B
These two equations give the trajectory in parametric form, so we can
vary t to get a plot of z versus y (weve taken all the constants to be 1
E = B = q = m = 1 for convenience):

E
Taking an initial velocity of B along the y axis (y (0) = EB ), we get

E
(7) y (t) = t
B
(8) z (t) = 0
CYCLOID MOTION IN CROSSED ELECTRIC AND MAGNETIC FIELDS 3

In this case the particle moves at constant speed along the y axis. Why?
The electric force is qE in the +z direction (assuming q is positive), and the
initial magnetic force is qvB = q EB B = qE in the z direction (use the right
hand rule in the cross product). Thus the two forces cancel each other, so
there is no net acceleration and the initial velocity remains unchanged.
E
If y (0) = 2B , we get

E E
(9) y (t) = sin t + t
2B B
E E
(10) z (t) = cos t +
2B 2B
This gives a similar trajectory to the first example, but with a reduced
amplitude:

Finally, if we start the particle off travelling at an angle of /4 w.r.t. the


y axis, so that y (0) = z (0) = EB , we get

E E E
(11) y (t) = cos t + t +
B B B
E
(12) z (t) = sin t
B
CYCLOID MOTION IN CROSSED ELECTRIC AND MAGNETIC FIELDS 4

This gives a trajectory identical in shape to the first case, but shifted down
so that its midpoint is now on the y axis:

P INGBACKS
Pingback: Electron charge to mass ratio
ELECTRON CHARGE TO MASS RATIO

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.3.
A classic experiment that is often done in undergraduate physics labs is
the use of crossed electric and magnetic fields for determining the charge
to mass ratio of the electron. The experimental setup involves electric and
magnetic fields that are mutually perpendicular, and then firing a beam of
electrons into these fields with a velocity that is perpendicular to both fields.
If the fields are adjusted so that the electrons dont deflect at all, then
from the last post, we know that v = EB .
If the electric field is removed at this point, the electron beam will be
deflected into a circular path. The magnetic field provides the centripetal
force, so

mv2
(1) = qvB
r
If we measure the radius of the circle, then we get

q v
(2) =
m rB
E
(3) =
rB2
I can remember doing this experiment when I was an undergraduate back
in the 1970s. They must have been on a fairly restricted budget since the
method we used for measuring the radius of the circle was to take a piece of
wooden dowelling and fiddle with the magnetic field until the electron beam
matched the dowelling. We had to judge this by eye, so the resulting value
for mq wasnt all that accurate. I always did prefer theoretical to experimental
physics anyway.

1
CURRENTS IN MAGNETIC FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.4.
Because an electric current involves moving charge, it feels a force when
a magnetic field is applied to it. Currents also produce magnetic fields, but
thats a topic for a future post. Here were concerned with calculating the
effect on a current of a magnetic field.
The essence of the argument is to apply the Lorentz force law to a col-
lection of moving charges. In its most general form, if we have a charge
distribution (r) with a velocity field v (r) immersed in a magnetic field
B (r) then the total force on the charge due to the magnetic field is

(1) F= v Bd 3 r

In this general case, all factors in the integrand depend on position. This
formula is often written as

(2) F= J Bd 3 r

where J v is the volume current density, and represents the charge per
unit area per unit time flowing past a given point.
We can make various simplifying assumptions to get some common spe-
cial cases. A surface current is a current that flows over the surface of some
volume, so the analogous formula is

(3) F = v B da

(4) = K Bda

In this case, is the surface charge density and K v is the charge per
unit width per unit time flowing past a given point on the surface.
Finally, for a linear current (such as that in an infinitesimally thin wire),
we get
1
CURRENTS IN MAGNETIC FIELDS 2


(5) F = v B dl

(6) = I Bdl

Here, is the linear charge density and I v is simply the charge per
unit time flowing past a point in the wire, and is the current commonly
used in describing electric circuits. Technically, I is a vector, since it has a
direction, but in most circuit calculations, the directional aspect of current
is ignored, since its constrained to flow wherever the circuit takes it.
As an example of calculating the magnetic force on a linear current, sup-
pose we have a square loop of wire of side length a lying in the yz plane
with its centre at the origin. It carries a constant current I, flowing counter-
clockwise when viewed down the x axis (that is, looking towards negative
x).
If we apply a constant magnetic field B = Bx, then the net force on the
loop is zero. This follows since the force on the edge y = 2a is

(7) Fy=a/2 = a (I z) (Bx)


(8) = aIBy

That is, the right-hand side of the square experiences a force pointing to the
right, since the current flows upwards and the magnetic field is pointing to-
wards you (assuming youre standing on the positive x axis looking towards
the origin). By a similar calculation, the force on the left-hand side of the
square is aIBy, since the current has reversed direction and the magnetic
field is the same. Similarly, the top and bottom edges cancel out, giving a
net force of zero.
Suppose we now impose a varying magnetic field of the form

(9) B = kzx

Notice that the field reverses direction as we cross the xy plane. To find
the force, we can start with the top and bottom edges z = a2 , since the field
is constant along each of these edges. On the top edge z = 2a , B = ka 2 x and
I = I y so
CURRENTS IN MAGNETIC FIELDS 3

 
ka
(10) Fz=a/2 = a I y x
2
ka2 I
(11) = z
2
On the bottom edge, z = a2 , B = ka
2 x and I = I y so

ka2 I
(12) Fz=a/2 = z
2
On the left and right sides, the forces cancel, since for each value of z
there is a segment of the loop with current +I z on the right and a corre-
sponding segment with current I z on the left. The magnetic field is the
same for these two segments so the net force is zero.
The total force is then

(13) F = Fz=a/2 + Fz=a/2


(14) = ka2 I z
P INGBACKS
Pingback: Surface and volume current in a wire
Pingback: Surface and volume current density 2
Pingback: Volume current density and dipole moment
Pingback: Biot-Savart law - current loops
Pingback: Biot-Savart law: force on other currents
Pingback: Current loop in a magnetic field
Pingback: Motional emf
SURFACE AND VOLUME CURRENT IN A WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.5.
A couple of examples of surface and volume current.
If we have a wire of radius a which carries a current I, what is the surface
current K if the current in the wire is spread uniformly over the surface of
the wire?
Since K is defined as the charge per unit width per unit time, we need the
width over which the surface current is spread. This is just the circumfer-
ence of the wire, so

I
(1) K=
2a
Now suppose the current is distributed in the volume of the wire such
that volume current density is inversely proportional to the distance from
the axis. That is,

A
(2) J=
r

for some constant A. We need the total flow across the cross-sectional area
of the wire to be the total current I, so we must have, using cylindrical
coordinates:
a 2
(3) I = Jd rdr
0 0
(4) = 2Aa
I
(5) A =
2a
Thus the volume current density is

I
(6) J=
2ar

1
SURFACE AND VOLUME CURRENT DENSITY 2

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.6.
A couple more examples of surface and volume current.
Given a disk with a uniform surface charge density rotating at angular
speed , its linear speed at a point r from the centre is v = r, so its surface
current is

(1) K = r
Now consider a uniformly charged sphere with density spinning about
its axis with angular speed . If we centre a spherical coordinate system at
the centre of the sphere and align the axis of rotation with the z axis, then
a point at location (r, , ) has linear speed r sin , so the volume current
density is

(2) J = r sin

1
VOLUME CURRENT DENSITY AND DIPOLE MOMENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.7.
There is a relation between the volume current density and the dipole
moment. Suppose we have a collection of charges and currents contained
within a finite volume. Consider the integral

(1) (xJ) d 3 r
V
Looking at the integrand, we have

(2) (xJ) = x J + x J
(3) = x J + x J
(4) = Jx + x J
The current density represents the rate of flow of charge across a surface,
so the total rate of change of charge must be equal to the net flow across the
surface. That is

d
(5) J da = d 3 r
A dt V

where the minus sign indicates that a net flow of charge out of the volume
(indicated by the integral on the LHS being positive) means a decrease in
the amount of charge inside the volume.
By the divergence theorem, the integral on the LHS can be converted into
a volume integral, so we get

(6) J da = Jd 3 r
A V

3
(7) = d r
V t

where weve taken the time derivative inside the integral, and converted it
to a partial derivative since depends on space as well.
1
VOLUME CURRENT DENSITY AND DIPOLE MOMENT 2

This relation has to be true for any volume, so we can equate the inte-
grands to get


(8) J =
t
Returning to the first equation, we can now say:

(9) (xJ) = Jx + x J

(10) = Jx x
t
Integrating the LHS and using the divergence theorem in the opposite
direction, we get

3
(11) (xJ) d r = xJ da
V A

Now if we take the bounding surface A to be outside all the charges and
currents, then the RHS is zero, so we get


3 3
(12) Jx d r = x d r
V V t
Clearly we can write similar equations for the y and z components and
then multiply each equation by the corresponding unit vector and add all
three up to get


3 3
(13) Jd r = r d r
V V t
The integral on the RHS is


3 dp
(14) r d r=
V t dt

where p is the total dipole moment of the charge in the volume. We there-
fore get the relation between current density and dipole moment:

dp
(15) Jd 3 r =
V dt
VOLUME CURRENT DENSITY AND DIPOLE MOMENT 3

P INGBACKS
Pingback: Ampres law for steady currents
Pingback: Momentum in electromagnetic fields
Pingback: Decay time for free charge in a conductor
BIOT-SAVART LAW - CURRENT LOOPS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.8.
As mentioned earlier, besides feeling a force from an external magnetic
field, an electric current also produces its own magnetic field. The exper-
imentally determined rule for calculating this generated magnetic field is
known as the Biot-Savart law. For a steady current (one that doesnt vary
with time) in a wire, this law can be written as


0 I (r r0 )
(1) B (r) = dl 0
4 |r r0 |3

where r0 is a location on the wire and r is the point at which you want to
determine the magnetic field. The constant 0 is known as the permeability
of free space, and is the magnetic analogue to 0 in electrostatics. Its value
is

(2) 0 = 1.25663706 106 m kg s2 Amp2

Again, this law isnt derived from anything more fundamental; its a
generalization of experiment, although the value of 0 is fixed at exactly
4 107 m kg s2 Amp2 .
As an example, suppose we want to find the field generated by a steady
current travelling round a square loop of side length 2R, at the centre of
the square. We can do this by finding the field generated by a single wire
segment first.
Because of the cross product in the integrand, the field will be perpendic-
ular to the plane of the square, so if we call this direction the z axis, and set
the edge of the square at x = R, we can write the integral as (setting r = 0
since were interested in the field at the origin):
1
BIOT-SAVART LAW - CURRENT LOOPS 2


I0 |r r0 | sin
(3) B1 (r) = z dl 0
4 |r r0 |3

I0 R R
(4) = z dy
4 R (R2 + y2 )3/2

2I0
(5) = z
4R

where is the angle between I and r r0 , so that sin = R/ |r r0 |.


By symmetry, the contribution from all 4 sides is equal, so we get for the
total field

2I0
(6) B = 4B1 = z
R
Now suppose we have a current loop consisting of a regular polygon with
n sides. In this case, each side subtends an angle of 2/n, so if we align
one side parallel to the y axis at x = R, this side will extend from an angle
of /n to +/n, and will have a length of 2R tan n . Now the integral for
a single side is

R tan n
I0 R
(7) B1 (0) = z 3/2
dy
4 (R2 + y2 )
R tan n

I0 2R tan n
(8) = z q
4 2
R R2 + R tan n
I0
(9) = z sin
2R n
Each of the n sides will still contribute an equal amount, so the total field
is

nI0
(10) B = nB1 = z sin
2R n
As n , this formula should give us the field due to a circular loop.
In this limit, we can approximate the sine by the first term in its Taylor
expansion sin n n so we get

I0
(11) lim B = z
n 2R
BIOT-SAVART LAW - CURRENT LOOPS 3

P INGBACKS
Pingback: Biot-Savart law: a couple more examples
Pingback: Biot-Savart law: force on other currents
Pingback: Electric and magnetic forces in two charged wires
Pingback: Ampres law for steady currents
Pingback: Ampres law: slab of current
Pingback: Solenoid with arbitrary cross-section
Pingback: Magnetic vector potential: div, curl and Laplacian
Pingback: Magnetic dipole moment of a current loop
Pingback: Currents and relativity
Pingback: Solenoid field from Biot-Savart law
Pingback: Magnetic field of a semi-circular current loop
Pingback: Magnetic field of current loop - off axis field
Pingback: Force between current loops: Newtons third law
Pingback: Average magnetic field within a sphere
Pingback: Force on a magnetic dipole
Pingback: Toroidal solenoid with a gap
Pingback: Faradays law and the Biot-Savart law
Pingback: Mutual inductance
Pingback: Maxwells equations with varying charge but constant current
Pingback: Jefimenkos equation for time-dependent magnetic field
BIOT-SAVART LAW: A COUPLE MORE EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.9.
Here are a couple more examples of calculating the magnetic field due to
a steady current in a wire.
For the first configuration (bear with me as my picture drawing skills
arent up to much so Ill describe this in words), we have a loop defined as
follows.
Starting at x = a, y = 0, draw a circular arc of radius a through an angle
of /2 up to x = 0, y = a. Then draw a line out to x = 0, y = b, then another
circular arc of radius b back down to x = b, y = 0, then finally join this up
with a line to x = a, y = 0 to complete the circuit. Find the field at the centre
of the circles defining the arcs.
We apply the Biot-Savart law

0 I (r r0 )
(1) B (r) = dl 0
4 |r r0 |3

where r0 is a location on the wire and r is the point at which you want to
determine the magnetic field. Along the two radial segments I is parallel to
r r0 so the cross product is zero, thus the only contributions come from
the two arcs.
From the earlier post, we saw that the field due to a complete circular
loop of radius R is, where the current flows in a counter-clockwise direction
as seen from above

I0
(2) B = z
2R
Each arc is 14 of a circle, and the current flows in the opposite direction
in the outer arc, so the net field is

I0 I0
(3) B = z z
8a  8b 
I0 1 1
(4) = z
8 a b
1
BIOT-SAVART LAW: A COUPLE MORE EXAMPLES 2

Now consider a semi-circular arc of radius R extending from = + 2 to


= 32 . Join each end of this arc to a horizontal wire extending to infinity
in the x direction (so we have infinite wires along the lines y = R, starting
at the y axis and extending out to positive infinity in the x direction). The
current flows in along the bottom wire and out along the top wire. We want
the field at the centre of circle defining the semi-circular arc.
The field due to the semi-circle is (negative since the current is flowing
clockwise)

I0
(5) Bs = z
4R
For the two straight segments, we can use the same approach as in the
earlier post. For the bottom wire, we get

I0 R
(6) Bb = z 3/2
dx
4 0 (R2 + x2 )
I0
(7) = z
4R
By symmetry, the top wire contributes the same amount, so the total field
is

(8) B = Bs + 2Bb
 
I0 2
(9) = z 1+
4R
BIOT-SAVART LAW: FORCE ON OTHER CURRENTS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.10.
Here are a couple more examples of calculating the magnetic field due to
a steady current in a wire.
Suppose we have an infinite straight wire on the x axis carrying a current
I in the +x direction. A square loop of side length a (also carrying current
I flowing clockwise around the square as viewed from above) is placed in
the xy plane so that the side nearest the wire is parallel to the wire and along
the line y = s. What is the force on the loop due to the wire?
First, we need to work out the magnetic field caused by the current in the
wire. We can apply the result we used earlier in calculating the magnetic
field due to a wire segment. In this case, we get, for a wire a perpendicular
distance d from the observation point r:


I0 |r r0 | sin 0
(1) B1 (r) = z dl
4 |r r0 |3

I0 d
(2) = z dx
4 (d 2 + x2 )3/2
I0
(3) = z
2d

The magnetic field points in the +z direction above the x axis (that is,
where y > 0) and in the z direction on the other side of the x axis. Since
the loop is on the lower side of the x axis and the current flows clockwise
as seen from above, I is in the +x direction on the side of the loop nearest
the wire and in the x direction on the side farthest from the wire. From
the symmetry of the problem the forces on the two sides of the square that
are perpendicular to the wire cancel, so we need to work out only the forces
on the two edges parallel to the wire. We get, using the form of the Lorentz
force law for linear currents
1
BIOT-SAVART LAW: FORCE ON OTHER CURRENTS 2


(4) F = I Bdl

aI 2 0 1
 
1
(5) = y
2 s s+a
Now suppose we have an equilateral triangle (side length a) with its base
aligned with the wire, and along the line y = s.. As before, the current
flows clockwise around the triangle.
The force on the edge nearest the wire is the same as for the square,
namely

aI 2 0
(6) F1 = y
2s
To calculate the force on one of the other edges, we can take the coor-
dinates of the vertex on this edge nearest the wire to be (0, s). Then the
equation of the line of which that edge is a part is


(7) y = x tan s
3
(8) = 3x s
The line increment along this edge is
p
(9) dl = dx2 + dy2
(10) = 2dx
The integral we need to do to figure out the force on this one edge is

I 2 0 dl
(11) F2 =
2 d

where d is the perpendicular distance from a point on the edge of the triangle
to the infinite wire. In our coordinates, d = |y| and over the extent of this
edge x varies from 0 to a2 so we get

a
I 2 0 2dx
2
(12) F2 =
2 0 3x + s
!
I 2 0 3a
(13) = ln 1 +
3 2s
BIOT-SAVART LAW: FORCE ON OTHER CURRENTS 3

This is the total force on the edge, which from the cross product right
hand rule is perpendicular to the edge and pointing towards the interior of
the triangle. The force on the third edge will, by symmetry, point in a
direction that is the mirror image of F2 (reflected about the line x = a/2).
Thus the components of these two forces parallel to the wire cancel, and the
net force from these two edges is twice the perpendicular component of F2 .
Since F2 is perpendicular to the edge of the triangle, which in turn makes
an angle of /3 with the x axis, F2 makes an angle of /6 with the x axis
(draw a little diagram if this isnt clear), so its perpendicular component is
F2 sin 6 = F22 and the total perpendicular component from the two edges of
the triangle is 2 F22 = F2 . This component is directed in the +y direction
while the force on the bottom edge is in the y direction, so the net force is
" ! #
I 2 0 3a aI 2 0
(14) F = ln 1 + y
3 2s 2s
" ! #
I 2 0 2 3a a
(15) = ln 1 + y
2 3 2s s

P INGBACKS
Pingback: Magnetic field of a solenoid
Pingback: Ampres law for steady currents
Pingback: Electron speed in a copper wire
Pingback: Magnetic vector potential of an infinite wire
Pingback: Currents and relativity
Pingback: Magnetic vector potential from magnetic field
Pingback: Emf near an infinite wire
MAGNETIC FIELD OF A SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.11.
A solenoid is a helical coil of wire wound round an insulating cylinder.
We can find the magnetic field due to a solenoid carrying a steady current
I as follows. First, we work out the field due to a single circular loop of
radius a as measured at a point P on the axis (which well take to be the z
axis) of the loop. The diagram below shows a side view of the solenoid.

In the diagram, well take the circular loop to be the first loop in the
solenoid (closest to P), whose radius subtends an angle 1 at P. The distance
from P along the axis to the centre of the circle is z. (Well refer to the other
parts of the diagram later.)
By symmetry, the components of the field in the x and y directions cancel,
so we need calculate only the z component. The Biot-Savart law in this case
is

I0 dl (r r0 )
(1) B (r) = z
4 |r r0 |3
The line element dl is tangent to the circle, and the vector r r0 is always
perpendicular to it. Look at the point on the loop where the tail of the1
vector meets the current loop, and assume the current is coming out of the
page at this point. Then applying the right-hand rule, we see that the field
vector lies in the plane of the figure, and points diagonally to the upper right,
making an angle of 2 1 with the z axis.  To project out the z component
a
of the field, we multiply by cos 2 1 = sin 1 = |rr0 | . This angle is a
1
MAGNETIC FIELD OF A SOLENOID 2

constant, as is |r r0 |, and the integration extends round the circumference


of the circle, so we get

I0 a 2a
(2) Bz = 0
4 |r r | |r r0 |2
I0 3
(3) = sin 1
2a
Now suppose the solenoid has n turns per unit length, and that this num-
ber is large enough that we can approximate the solenoid by a circular sur-
face current density of In per unit length, and thus use integration to de-
termine the overall field. To do the integral, we need to work out relation
between a change in and a change in z, as shown in the diagram. An
increase of dz means a decrease in angle of d as shown. To work out the
relation:

a
(4) sin =
a2 + z2
a
(5) sin ( d ) = q
a2 + (z + dz)2
We can expand the second equation in a Taylor series and retain only first
terms in the differentials, and we get

a az
(6) sin cos d = dz
a2 + z2 (a2 + z2 )3/2
Replacing the terms on the right by trig functions, we get

a az 1
(7) 3/2
dz = sin cos sin2 dz
2
a +z2
(a2 + z2 ) a
Thus we get

a
(8) dz = d
sin2
The current in an infinitesimal slice of the solenoid is therefore

Ina
(9) Indz = d
sin2
We can now find the total field by doing the integral
MAGNETIC FIELD OF A SOLENOID 3

1
0 Ina
(10) Bz = sin3 1 2 d
2 4a sin
n0 I
(11) = (cos 2 cos 1 )
2
This might look like its independent of the radius a, but this dependence
is included in the angles.
For an infinite solenoid, 1 and 2 0 and we get

(12) B = n0 I
P INGBACKS
Pingback: Solenoid with arbitrary cross-section
Pingback: Magnet falling through a metal pipe
Pingback: Faradays law, Ampres law and the quasistatic approxima-
tion
Pingback: Self-inductance
Pingback: Energy in a magnetic field
Pingback: Angular momentum in electromagnetic fields
Pingback: Energy transfer in a solenoid
Pingback: Angular momentum conservation: example with a solenoid
Pingback: Momentum of a point charge outside a solenoid
Pingback: Magnetic systems in thermodynamics
ELECTRIC AND MAGNETIC FORCES IN TWO CHARGED
WIRES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.12.
The electric field a distance d from an infinite line of charge (linear charge
density ) is

1 2
(1) E=
40 d
For positive charge, this field points radially outwards from the wire.
Now if we move this line of charge at speed v, thus creating a current
I = v, it will produce a magnetic field

I0
(2) B =
2d
v0
(3) =
2d
The field will circle the line in a direction determined by the right-hand
rule.
Now suppose we place an identical line of charge, moving at the same
speed, a distance d from the first one. Since both lines carry the same sign
of charge, the electric force between them is repulsive. From the Biot-Savart
law and the right hand rule, we can see that the magnetic force is attractive.
Can we adjust v so that the electric and magnetic forces balance each other?
For a unit length of wire, the electric force is

1 2 2
(4) FE =
40 d

and the magnetic force is

( v)2 0
(5) FB =
2d
Setting these two equal to each other, we get
1
ELECTRIC AND MAGNETIC FORCES IN TWO CHARGED WIRES 2

1
(6) v=
0 0
1
(7) =p
(8.85418782 10 ) (1.25663706 106 )
12

(8) = 2.997924 108 m s1


This happens to be the speed of light, so its not possible to get the charge
moving fast enough for the magnetic force to balance the electric force.
Its not a coincidence that 10 0 = c; in fact this quantity comes out of
Maxwells equations as the speed of electromagnetic waves, and is what led
Einstein to postulate that c is a universal constant and thus create his theory
of special relativity. Well get to this eventually.
P INGBACKS
Pingback: Force between two sheets of current
AMPRES LAW FOR STEADY CURRENTS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.13.
The Biot-Savart law is the mathematical statement of what is observed
from experiments on steady currents and their production of magnetic fields.
It is, if you like, the magnetic analogue to Coulombs law in electrostatics.
Just as we derived Gausss law from Coulombs law and the principle of
superposition, so we can derive a law from the Biot-Savart law and the
magnetic principle of superposition. This law is known as Ampres law.
Starting with the most general form of the Biot-Savart law, which treats
currents throughout a volume, we have

0 J (r0 ) (r r0 )
(1) B (r) = d 3 r0
4 V |r r0 |3

where the volume V is assumed to be the finite volume enclosing all currents
under consideration.
We can take the curl of this expression if we remember that the derivatives
within the curl are with respect to the components of r, not r0 . That is

" #
0 J (r0 ) (r r0 )
(2) r B (r) = r d 3 r0
4 V |r r0 |3

where weve added a suffix r to to emphasize that it applies only to r


components.
An identity from vector calculus comes in handy here:

(3) (A B) = (B ) A (A ) B + A ( B) B ( A)
If we set

A = J r0

(4)
(r r0 )
(5) B =
|r r0 |3
1
AMPRES LAW FOR STEADY CURRENTS 2

then all terms involving the derivative of J are zero since derivatives are
being taken with respect only to unprimed coordinates. Therefore we get

(6) " # !
J (r0 ) (r r0 )   (r r0 ) (r r0 )
= J r0 0

r + J r
|r r0 |3 |r r0 |3 |r r0 |3
The divergence in the second term can be written in terms of the 3-d
delta function. Again, remember that were taking the derivative only with
respect to r components, so the presence of the r0 serves merely to shift the
origin, so we have

(r r0 )
= 43 r r0

(7)
|r r0 |3
To handle the first term, we can try to convert it from a volume integral
to a surface integral and then do the usual trick of letting the surface go to
infinity, and use the assumption that the volume containing the currents is
finite to set the integral to zero. To do this, we can split this term into its 3
components and treat each one separately. So for the x component, we have

  x x0
(8) J r0
|r r0 |3
We need another vector calculus identity, and this time we can use

(9) A f = f A ( f A)
There is one slight snag, however. In order to use the divergence theorem
to convert a volume integral of a divergence into a surface integral of the
normal component of a vector field, the divergence needs to be with respect
to the integration variable, which the operator here isnt. However, look-
ing at the term 8, we see that the primed and unprimed coordinates occur
in an anti-symmetric fashion (wherever an unprimed variable occurs, it is
paired with a primed variable with the opposite sign), so a derivative with
respect to an unprimed variable is just the negative of the derivative with
respect to a primed variable. That is

 x x0  x x0
J r0 r 0
 
(10) = J r r 0
|r r0 |3 |r r0 |3
AMPRES LAW FOR STEADY CURRENTS 3

We can now convert the term into the difference of two divergences with
respect to the integration variable:
!
0
 x x0 x x0 0 x x0
J r0
  
(11) J r r0 = r 0 J r r0
|r r0 |3 |r r0 |3 0
|r r |3

At this point, we need to make an observation about the divergence of the


current density. We saw that, due to conservation of charge:


(12) J =
t
If were dealing with steady currents, then the amount of charge within
the volume doesnt change, and

(13) J = 0
Making this assumption, the volume integral now becomes

!
 x x0 3 0 x x0
J r0 r0 J r0 d 3 r0
 
(14) d r = r0
V |r r0 |3 V 0
|r r |3

Now finally we can convert to a surface integral and use the fact that
J = 0 outside V . Thus


 x x0 3 0
J r0 r0

(15) d r =0
V |r r0 |3

for the special case of steady currents. In the more general case, 13 wont
be true and we have to use 12 instead, which means the first term in the
integral wont be zero in general. But more on that later.
Finally, we can look at the last term in 6 and use its delta function equiv-
alent to get

0)
!
0 (r r
J r0 d 3 r0

(16) r B (r) =
4 V |r r0 |3

= 0 J r0 3 r r0 d 3 r0
 
(17)
V
(18) = 0 J (r)
AMPRES LAW FOR STEADY CURRENTS 4

This is Ampres law in differential form, which relates a steady current


density to the curl of the magnetic field it produces. In integral form, using
Stokess theorem, we get


(19) B dl = 0 J da

That is, the integral of B around a closed loop is equal to the total current
passing through that loop (times 0 ).
We can use this form of Ampres law to work out the field configurations
in some simple cases in much the same way we used Gausss law to work
out E for simple charge distributions.
As an example, suppose we have a cylindrical wire of radius a carrying a
current I. If the current is uniformly distributed over the surface of the wire,
what is the magnetic field inside and outside the wire?
Inside the wire we can choose a circular loop of radius r < a centred on
the axis of the wire. From symmetry, B dl is a constant all around this loop,
and since the enclosed current is zero (its all on the surface of the wire),
we get

(20) 2rB = 0
(21) B = 0

Thus there is no field inside the wire.


Outside the wire, we can again choose a circular loop of radius r > a.
This time, the enclosed current is I so we get

(22) 2rB = 0 I
0 I
(23) B =
2r

so the field decreases proportionally to 1/r, as we saw in an earlier calcula-


tion using the Biot-Savart law.
Now suppose the current is distributed within the wire such that J = kr
for some constant k. First, we find k. Since the total current is I, we must
have
AMPRES LAW FOR STEADY CURRENTS 5

a
(24) I = 2 kr2 dr
0
2ka3
(25) =
3
3I
(26) k =
2a3
Inside the wire, at a radius s the enclosed current is
s
3I
(27) I (s) = 2 r2 dr
2a3 0
Is3
(28) =
a3
Thus the field is

(29) 2sB (s) = 0 I (s)


0 Is2
(30) B (s) =
2a3
Outside the wire, B is the same as before since all the current is enclosed
within the loop.

P INGBACKS
Pingback: Ampres law: slab of current
Pingback: Force between two sheets of current
Pingback: Solenoid with arbitrary cross-section
Pingback: Ampres law: surface of integration
Pingback: Divergence of magnetic field: magnetic monopoles
Pingback: Magnetic vector potential
Pingback: Magnetic vector potential of an infinite wire
Pingback: Magnetic scalar potential
Pingback: Faraday field and magnetic vector potential
Pingback: Magnetization: bound currents
Pingback: Faradays law, Ampres law and the quasistatic approxima-
tion
Pingback: Coaxial cable with varying current
Pingback: Faradays law and the Biot-Savart law
Pingback: Energy in a magnetic field
Pingback: Maxwells correction to Ampres law
Pingback: Faraday field and magnetic vector potential
AMPRES LAW FOR STEADY CURRENTS 6

Pingback: Maxwells equations with varying charge but constant current


Pingback: Coaxial wave guides: TEM mode
Pingback: Fields due to a moving linear charge
Pingback: Gravitoelectric and gravitomagnetic acceleration for a moving
wire
AMPRES LAW: SLAB OF CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.14.
Heres another simple example of Ampres law. We have an infinite slab
parallel to the xy plane, extending from z = a to z = +a, and carrying a
volume current density J in the +x direction. What is the magnetic field
both inside and outside the slab?
First, we need to find the direction of the field. From the Biot-Savart law,
we know the field must be perpendicular to the current, so there can be no
field in the x direction. Now if we consider a thin line of current parallel
to the x axis, we know that the field loops around the line so that it forms a
counterclockwise cylinder when viewed down the x axis. By the symmetry
of the setup, the z component of the field in the xz plane due to a line of
current extending along the line z = b, y = c (for some constants b and c) is
cancelled by the z component from the line of current along z = b, y = c,
so the net z component of the field in the xz plane must be zero. Since
the slab is infinite in extent in the xy plane, we can apply this symmetry
argument to all planes parallel to the xz plane, so there is no z component of
the field anywhere. Thus the field is entirely in the y direction.
To find its magnitude, well consider first the region outside the slab. The
field is in the y direction above the slab and in the +y direction below.
Take a rectangular loop cutting through the slab, with one pair of sides hav-
ing length 1, parallel to the yz plane, centred on the xy plane, and extending
to z = d where d > a. Using Ampres law we get


(1) B dl = 0 J da
(2) 2B = 0 2aJ
(3) B = 0 aJ

Inside the slab, we can observe that in the xy plane, B = 0 since any
contribution to the y component of B from an element at z = b is cancelled
by the contribution from z = b. Also from symmetry arguments, the field
at z = d is By where By is the y component to be determined. Using
another loop like the previous one except this time with d < a, we get
1
AMPRES LAW: SLAB OF CURRENT 2


(4) 2 By = 0 2dJ

(5) By = 0 dJ
In terms of z, this gives

(6) B = 0 Jzy

for a z a.
P INGBACKS
Pingback: Force between two sheets of current
Pingback: Magnetic vector potential: sheet of current
Pingback: Force on a magnetic dipole in a slab of current
COAXIAL SOLENOIDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.15.
This problem is a simple extension of the result we got for the infinite
solenoid. We have two coaxial solenoids along the z axis. The inner one,
with radius a, has n1 turns per unit length and carries current I clockwise
when viewed down the z axis. The outer one, with radius b, has n2 turns per
unit length and carries current I counterclockwise.
To find the field everywhere, we can use the principle of superposition,
together with the facts that (i) the field outside a solenoid is zero and (ii) the
field inside is B = 0 nI.
The field inside both solenoids is therefore

(1) Binner = 0 I (n1 n2 ) z

where the direction is determined by the right-hand rule.


Between them, only the outer solenoids field is non-zero, so we get

(2) Bmiddle = 0 n2 I z
Outside both solenoids, the field is zero.

1
FORCE BETWEEN TWO SHEETS OF CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.16.
As an example of Ampres law applied to surface currents, suppose we
have two parallel, infinite sheets of charge parallel to the xy plane, with the
upper sheet having a charge density of + and the lower sheet having .
If the sheets are both moving with speed v in the +x direction, what is the
magnetic field everywhere?
This problem is similar to that of the slab of current that we treated earlier.
Consider first the upper sheet. By the same arguments as in the earlier case,
the field above the sheet is in the y direction, and below the sheet is in
the +y direction. The magnitude of the field is obtained from Ampres
law, using a surface current of K = vx. Using an integration loop with a
side-length of 1 unit, we get

(1) Bu dl = 0 K dn
(2) 2Bu = 0 v
0 v
(3) Bu = y
2

with the plus sign for points below the sheet, and the minus sign for points
above. In the first line, dn represents a normal to a line segment in the
cross section of one of the sheets of charge, and we integrate over this cross
section of length 1 unit. dn is essentially xd` where d` is the length of the
infinitesimal line segment. This line segment is not the same as dl in the
integral of magnetic field, since that integral is a line integral around a loop.
For the lower sheet, since the charge density is negative, the field is re-
versed, so we get

0 v
(4) Bl = y
2
Thus above the top sheet and below the bottom sheet, the net field is zero.
Between the sheets (if youll pardon the phrase), the two fields add, and we
get
1
FORCE BETWEEN TWO SHEETS OF CURRENT 2

(5) B = 0 vy
The force per unit area on the top sheet can be found from the Lorentz
force law. For a unit area, we have

(6) Fmag = vx Bl
0 ( v)2
(7) = z
2
That is, the force pushes the sheets apart.
The electrical force is attractive, since the two sheets carry opposite charges,
and the electric field due to an infinite plane of charge is


(8) E= z
20
The force per unit area is therefore

(9) Felec = E
2
(10) = z
20
If we try to balance the electric and magnetic forces, we must have

0 ( v)2 2
(11) =
2 20
1
(12) v =
0 0
(13) = c
(that is, the speed of light). Not surprisingly, this is the same result that
we got for two line charges.
P INGBACKS
Pingback: Transmission lines
SOLENOID WITH ARBITRARY CROSS-SECTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.17.
We calculated the magnetic field of a solenoid with a circular cross-
section and found that, for an infinite solenoid

(1) B = 0 nI

where n is the number of turns per unit length and I is the current flowing
through the wire wrapped around the cylinder.
We can actually find the magnetic field of an infinite solenoid of any
cross-section, provided that cross-section is constant over the entire length.
First, we use the Biot-Savart law to find the direction of the field. Well
choose a rectangular coordinate system such that the axis of the solenoid
is parallel to the y axis. This means that the cross-section of the solenoid
is parallel to the xz plane, so the current has no y component (well, OK, it
does have a very slight y component since the wire is wound as a helix and
current does flow down the length of the solenoid, but we can assume here
that weve got an idealized solenoid where the current flows in closed loops
around the axis). That is

(2) I = Ix x + Iz z

Now pick a point on the solenoid at position r01 = x0 x+y0 y+z0 z. If we put
the observation point r in the xz plane, then it has coordinates r = xx + zz.
Then the Biot-Savart law says that the contribution to the magnetic field
from a line element dl 0 at point r0 is

0 0
 0
(3) dB = I r r dl
4 |r r0 |3

The cross product is what interests us, and we get for the current element
under consideration
1
SOLENOID WITH ARBITRARY CROSS-SECTION 2

I r r01 = [Ix x + Iz z] x x0 x y0 y + z z0 z
    
(4)
= y0 Iz x Ix z z0 Iz x x0 y + y0 Ix z
  
(5)
Because of the symmetry of the solenoid, there is a point on the solenoid
at position r02 = x0 x y0 y + z0 z which has the same current and (since r has
no y component) the same value of |r r0 |. Thus the only difference is that
the sign of y0 has changed, so we get

I r r02 = y0 Iz x Ix z z0 Iz x x0 y y0 Ix z
   
(6)
When we add the contributions from these symmetric elements, only the
y component is left, so we conclude that the field must be entirely in the y
direction.
The magnitude of the field could be worked out from the Biot-Savart law
if we knew the shape of the cross-section, but using Ampres law, we can
choose a rectangular loop of unit length parallel to the yz plane. If this loop
has both edges outside the solenoid, with the nearer edge in the plane z = a
and the farther edge at z = b, then there is no enclosed current so since the
contribution from the two edges parallel to the z axis cancel (the path along
one of these edges is equal and opposite to the path along the other), we get

(7) B dl = B (b) B (a)
(8) = 0
That is, the field outside the solenoid is constant. At this point, the as-
sumption is made that the field must go to zero at infinite distance (some-
thing that doesnt sit quite right, since we are dealing with an infinite solenoid
after all, but anyway) therefore be zero everywhere outside the solenoid.
Inside the solenoid, we could set up another loop that doesnt cross the
current and conclude again that the field inside is also constant. To get its
magnitude, we use a loop that does cross the current, with one edge inside
and one outside. The enclosed current for a unit length is nI and the only
edge that contributes to the line integral is the edge inside the solenoid so
we have

(9) B dl = Bin
(10) = 0 nI
Thus the magnetic field inside an infinite solenoid is independent of the
cross-sectional shape.
SOLENOID WITH ARBITRARY CROSS-SECTION 3

Using a similar technique (example 5.10 in Griffiths), it is possible to


show that the magnetic field inside a torus-shaped solenoid (again, the cross-
section doesnt matter) is

0 NI
(11) Btorus =
2s

where s is the radial distance from the axis of the torus, if this places the
observation point inside the torus (the field is again zero outside). In this
case, the field inside is not constant, but falls off linearly with distance from
the axis. For a torus with a radius large compared with its cross-sectional
dimensions, we can take s to be essentially the radius of the torus, so that
N/2s = n and the formula reduces to that for a linear solenoid.
P INGBACKS
Pingback: Coaxial solenoids
Pingback: Magnetostatic boundary conditions: a couple of examples
Pingback: Solenoid field from Biot-Savart law
Pingback: Magnetic dipole field of a finite solenoid
Pingback: Magnetization: bound currents
Pingback: Toroidal solenoid with a gap
Pingback: Faradays law, Ampres law and the quasistatic approxima-
tion
AMPRES LAW: SURFACE OF INTEGRATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.18.
Ampres law in integral form states that


(1) B dl = 0 J da

That is, if we choose a closed loop and integrate B dl around this loop,
the result is the same as integrating 0 J da over the area enclosed by this
loop. However, a closed loop can be the boundary of an infinite number of
surfaces, so since the law doesnt state which surface over which to inte-
grate, does it make any difference which surface we choose?
To analyze this problem, think about what J da represents. It is the
component of current travelling perpendicular to the surface at a given point
on the surface. To make things concrete, suppose we take the loop to be a
circle, and consider two surfaces bounded by this circle. The first is a flat
disk, and the second is a hemisphere.
If the currents are steady (not changing with time), then from conserva-
tion of charge, any current that flows through the disk can do one of two
things. First, it can continue on and flow through the hemisphere as well or
second, it can turn around and flow back out through the disk. It cant enter
the region between the disk and the hemisphere and then cycle endlessly
within that volume or just pile up somewhere, since if it did, charge would
continually be flowing into the volume with no charge flowing out, so the
amount of charge within the volume would be increasing, which means that
the currents within the volume are not constant, thus violating our assump-
tion of steady currents.
Now if current flowing through the disk continues on to flow through the
hemisphere as well, then the amount of current flowing through the two
surfaces is the same. On the other hand, if some current that flows through
the disk turns around and flows out again without reaching the hemisphere,
then that current makes a net contribution of zero to the amount of current
flowing through the disk, so again the net amount of current flowing through
the two surfaces is the same. Thus the net total amount of current flowing
through any surface is the same.
1
AMPRES LAW: SURFACE OF INTEGRATION 2

Note in particular that this argument requires the assumption of steady


currents, which is expressed mathematically by requiring J = 0, which
is to say that there are no sources or sinks for current. This assumption was
in fact used in the derivation of Ampres law.
P INGBACKS
Pingback: Ampres law and the steady current assumption
ELECTRON SPEED IN A COPPER WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.19.
Its interesting to work out some actual numbers for electric and mag-
netic properties of a real material, such as copper. The density of copper
is 8.96 g cm3 and its atomic weight is 63.546 so the number of atoms per
cubic cm is

8.96
(1) nCu = 6.022142 1023
63.546
(2) = 8.49123 1022
Assuming 2 free electrons per atom the density of free charges is there-
fore 1.69825 1023 cm3 .
Suppose we have a copper wire 1 mm in diameter carrying a current
of 1 Amp. One amp is one coulomb per second, so that means we have
6.2415 1018 electrons per second passing a given point in the wire. Given
the charge density, we can work out the linear charge density in the wire. A
centimetre length of the wire has a volume of (0.05)2 = 7.854103 cm3 ,
so the linear charge density is

(3) = 1.69825 1023 7.854 103


(4) = 1.3338 1021 cm1
Since we require 6.2415 1018 electrons per second, the speed at which
charge moves down the wire is

6.2415 1018
(5) v =
1.3338 1021
(6) = 4.6795 103 cm s1
In other words, it takes about 3.5 minutes for an electron to move 1 cm,
which is extremely slow. This seems paradoxical; after all, when you flip
a light switch it doesnt take several hours for the light to come on, as it
would if we had to wait for electrons to trickle all the way from the switch
to the light bulb.
1
ELECTRON SPEED IN A COPPER WIRE 2

The answer to this paradox is that its not the electrons themselves that
carry the energy (or the signal, in the case of, say, a computers network
connection or a telephone line), its the electromagnetic wave that travels
down the wire. In a vacuum, these waves travel at the speed of light, while
in unshielded copper, the wave travels at up to 97% of the speed of light.
For more details, see this Wikipedia article.
If we place 2 such wires parallel to each other at a distance of 1 cm, we
can work out the force between them. Since they are electrically neutral,
there is no electric force, but there is a magnetic force. The magnetic field
due to one wire is

I0
(7) B=
2d
For a current of 1 amp and distance of 0.01 m, the field is (using MKS
units)

(1) 4 107
(8) B =
2 (0.01)
(9) = 2 105 T

For a unit length of the other wire (1 metre), the force is then

(10) F = IB
(11) = 2 105 N m1

If we had a linear charge density as above consisting of electrons only


(no positive charges), then the electric field from one wire is (using MKS
units)

2
(12) E=
40 d
1.3338 1023 1.6022 1019
(13) =
2 (8.854 1012 ) (0.01)
(14) = 3.8414 1016 V m1

The force on a metre of the other line charge is then


ELECTRON SPEED IN A COPPER WIRE 3

(15) F = E
 
(16) = 1.3338 1023 1.6022 1019 3.8414 1016
(17) = 8.21 1020 N m1
This is around 4 1025 times greater than the magnetic force, so clearly
the electric force is a lot stronger.
AMPRES LAW AND THE STEADY CURRENT ASSUMPTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.20.
This is a bit of a footnote to the post on the surface used to evaluate the in-
tegral in Ampres law. It was pointed out there that one of the assumptions
made in deriving Ampres law was that currents were steady, in particu-
lar that J = 0. This assumption makes the differential form of the law
consistent. This form states that

(1) B = 0 J
An identity from vector calculus states that, for any vector field,
( A) = 0. This is true in the steady-current case, but breaks down in
the more general case where J = /t, that is, where the current is
responsible for redistributing the charge distribution.
There is a related problem with the electric field equation E = 0.
Although this doesnt pose any problems mathematically, since the iden-
tity ( E) = 0 is still true. However, the curl-free field equation was
derived from Coulombs law, which in turn gave rise to the fact that in elec-
trostatics, the field can be written as the gradient of a potential, and the
vector identity = 0 then gives us the zero curl. Once charges start
to move, they generate magnetic fields, and as well see (eventually), vary-
ing magnetic fields can affect the electric field and give rise to a non-zero
curl.

1
DIVERGENCE OF MAGNETIC FIELD: MAGNETIC
MONOPOLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.21.
In electrostatics, Gausss law states that


(1) E =
0

that is, that free charge (represented by the charge density ) acts as a source
or sink for the electric field. Is there an analogous law in magnetostatics?
To find out, we can start with the Biot-Savart law which gives the mag-
netic field B in terms of the currents in a volume


0 J (r0 ) (r r0 )
(2) B (r) = d 3 r0
4 V |r r0 |3
What happens if we take the divergence of this formula? The divergence
operator here takes derivatives with respect to the unprimed coordinates
only, since the primed coordinates are the variables of integration and dont
appear in the final result for B. We get

" #
0 J (r0 ) (r r0 )
(3) B = d 3 r0
4 V |r r0 |3
We can use an identity from vector calculus to expand the integrand:

(4) (A C) = C ( A) A ( C)
We can make the correspondences:

A = J r0

(5)
r r0
(6) C =
|r r0 |3
1
DIVERGENCE OF MAGNETIC FIELD: MAGNETIC MONOPOLES 2

Now J (r0 ) = 0 since J depends only on primed coordinates. For the


other vector, consider the x component:

(7)
! ! !
r r0 z z0 y y0
=
|r r0 |3 x
y |r r0 |3 z|r r0 |3
3  0
 0
 0
 0

(8) = 2 y y z z 2 z z y y
2 |r r0 |5
(9) =0

where weve used

q
r r0 = (x x0 )2 + (y y0 )2 + (z z0 )2

(10)
The other two components are also zero, so
!
r r0
(11) =0
|r r0 |3

and in general

(12) B = 0
That is, there are no magnetic sources or sinks, so there is no such thing
as magnetic charge or, as they are more commonly known, magnetic
monopoles.
Unlike some of the other laws that weve covered, the divergence-free
nature of the magnetic field is true not only in magnetostatics, but in more
general situations where both the electric and magnetic fields vary with
time.
Although magnetic monopoles are excluded from the classical theory of
electromagnetism, more modern theories such as string theory do predict
their existence. So far, no magnetic monopoles have ever been found.
If magnetic monopoles did exist, then we could make a few changes to
the electromagnetic laws weve discussed so far. Basically, these adjust-
ments are creating terms for the magnetic charge analogous to the electric
charge. So we would get a divergence for B:

(13) B = 0 m
DIVERGENCE OF MAGNETIC FIELD: MAGNETIC MONOPOLES 3

where 0 is a constant analogous to 0 and m is the magnetic charge den-


sity.
Since B depends on the flow of electric charge, we might expect that
E would now depend on the flow of magnetic charge, so me would get
something like

(14) E = 0 Jm

where 0 is another constant and Jm is magnetic current caused by flowing


magnetic monopoles.
The Lorentz force law gives the force resulting from electric charge in-
teracting with a magnetic field, so we might expect an analogous force law
resulting from the interaction of a magnetic charge with an electric field,
along the lines of

(15) Fm = qm v E
At this point, we need to examine the units. The electric field was orig-
inally defined as the force per unit of electriccharge, so the units of the RHS
here are (magnetic charge) length time1 mass length time2 electric charge1

If the unit of magnetic charge is the same as that of electric charge, then
the units of the RHS work out to forcevelocity so this doesnt work. We
could therefore define the units of magnetic charge as (electric charge) /(velocity).
Presumably, a magnetic charge would also be affected by a magnetic field
in the same way an electric charge is affected by an electric field, so the total
force on a magnetic monopole would be something like

(16) F = qm B + qm v E
However, with the definition of the unit of magnetic charge as given
above, the units dont work out correctly for the first term (since qe v B
has the units of force, qm B has units of force/velocity2 ). We would there-
fore need to multiply the first term by a quantity that has units of velocity2 .
If we define some constant velocity as v0 then we might get something like

(17) F = v20 qm B + qm v E
If we want v0 to be a universal constant, the fact that it is a velocity
suggests that it might be the speed of light c.
MAGNETIC VECTOR POTENTIAL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.22.
The Biot-Savart law gives the magnetic field B in terms of the currents in
a volume:

0 J (r0 ) (r r0 )
(1) B (r) = d 3 r0
4 V |r r0 |3
By straightforward calculation, we can show that

r r0 1
(2) =
|r r0 |3 |r r0 |
Plugging this into the integral, we get

0 1
J r0 d 3 r0

(3) B (r) =
4 V |r r0 |
The gradient operator operates on the unprimed coordinates only, so we
can rewrite the integrand as

1 J (r0 )
J r0

(4) =
|r r0 | |r r0 |
To see that this works, consider the general equation

(5) J = (J)

where on both sides, the operator operates only on and not on J. Look-
ing at each side separately and isolating the x component, we get

(6) [J ]x = Jy z + Jz y
(7) [ (J)]x = y (Jz ) z (Jy )
(8) = Jz y Jy z
1
MAGNETIC VECTOR POTENTIAL 2

where in the last line we can pull the components of J outside the deriva-
tives since it depends only on the primed coordinates. The same derivation
obviously works for the y and z components as well. Therefore we can
rewrite the Biot-Savart law as


0 J (r0 ) 3 0
(9) B (r) = d r
4 V |r r0 |
The vector quantity A (r) is defined by


0 J (r0 ) 3 0
(10) A (r) = d r
4 V |r r0 |

and is known as the magnetic vector potential. This is a magnetic analog of


the electrostatic condition E = , as we can write

(11) B = A
Being able to write B as a curl makes its divergence zero automatically.
However, weve defined A only by specifying what its curl is, and since
we need both the curl and the divergence to determine a vector field uniquely,
A is not uniquely specified by its derivation. Since all we require is that
B = A, we can add any vector field to A that has a zero curl. Since
the curl of any gradient is zero, we can write the most general form of the
vector potential as


0 J (r0 ) 3 0
(12) A (r) = d r + (r)
4 V |r r0 |

where is any scalar function of position. Transforming the vector poten-


tial in this way is known as a gauge transformation. A common choice is
to choose so that A = 0, which can be done by requiring


2 0 J (r0 ) 3 0
(13) = d r
4 V |r r0 |
The quantity on the RHS is a scalar function of r, so this is an instance of
Poissons equation. For steady currents, we can actually work out the RHS.
First, note that
MAGNETIC VECTOR POTENTIAL 3

J (r0 )
  
0 1

(14) = J r
|r r0 | |r r0 |
 
0
 0 1
(15) = J r
|r r0 |

where in the second line, we are now taking the derivative with respect to
the primed coordinates. We now get

J (r0 ) 3 0
 
0 1 0
d 3 r0

(16) d r = J r
V |r r0 | V
0
|r r |
0
J (r0 ) 3 0

J
(17) = + d r
|r r0 | V |r r |
0

(18) = 0+0
In the second line, weve integrated by parts. The integrated term is eval-
uated at infinity and is zero assuming that all currents are contained within
a finite volume and the second term is zero if currents are steady, since
0 J (r0 ) = 0. Thus 2 = 0 everywhere. This is an instance of Laplaces
equation, but since it applies over all space, if we require to be finite at
infinity, the only solution is = constant, which in turn implies = 0 so
we can in fact just write

0 J (r0 ) 3 0
(19) A (r) = d r
4 V |r r0 |
With this choice of gauge, we can get another relation for A by using the
vector identity

(20) B = A = ( A) 2 A
Applying A = 0 and quoting Ampres law, we get

(21) 2 A = 0 J
This is another instance of Poissons equation, this time with a separate
equation for each of the three components.
Finding the vector potential involves working out similar integrals to
those for finding B from the Biot-Savart law. As an example, suppose we
have a wire segment extending from z1 to z2 on the z axis and carrying a
steady current I. The form of 19 for a linear current is
MAGNETIC VECTOR POTENTIAL 4


0 I dl0
(22) A (r) =
4 |r r0 |

We have

q
r r0 = x2 + y2 + (z z0 )2

(23)

Doing the integral we get


0 I z2
dz0
(24) A= q
4 z1 x2 + y2 + (z z0 )2
q
x 2 + y2 + (z z )2 + z z
0 I 2 2
(25) = ln q z
4 2
x2 + y2 + (z z1 ) + z1 z

To check that this is correct, we can calculate

(26)

B = A
(27)

y y x x
=     x +     y
z2 z z1 z z1 z z2 z
r22 r2 + 1
2
r 1 r1 + 1 2
r 1 r1 + 1 2
r 2 r2 + 1

where

q
(28) ri x2 + y2 + (z zi )2

To compare this with eqn 5.35 in Griffiths, we need to consider a partic-


ular field point r, so suppose we look at r = [0, s, 0]. Then we can define the
angles i to be the angle between a line from r to zi and the xy plane (these
are the same angles Griffiths uses in his example 5.5).
MAGNETIC VECTOR POTENTIAL 5

From the diagram, we see that

zi z
(29) = sin i
ri
s
(30) = cos i
ri
For the point r we get
 
0 I s s
(31) B= x
4 r22 (sin 2 + 1) r12 (sin 1 + 1)
cos2 2 cos2 1
 
0 I
(32) = x
4 s (sin 2 + 1) s (sin 1 + 1)
0 I 1 sin2 2 1 sin2 1
 
(33) = x
4 s (sin 2 + 1) s (sin 1 + 1)
 
0 I (1 sin 2 ) (1 + sin 2 ) (1 sin 1 ) (1 + sin 1 )
(34) = x
4 s (sin 2 + 1) s (sin 1 + 1)
0 I
(35) = (sin 1 sin 2 ) x
4s
(The sign is opposite to Griffiths eqn 5.35 because of the orientation of
the axes. My x axis points into the page, so the magnetic field points out of
the page, in agreement with Griffiths.)

P INGBACKS
Pingback: Magnetic vector potential and current
Pingback: Magnetic vector potential of constant field
Pingback: Magnetic vector potential of an infinite wire
Pingback: Magnetic vector potential: div, curl and Laplacian
Pingback: Magnetic scalar potential
MAGNETIC VECTOR POTENTIAL 6

Pingback: Magnetic field of rotating sphere of charge


Pingback: Divergenceless vector field as a curl
Pingback: Magnetic dipole
Pingback: Faraday field and magnetic vector potential
Pingback: Magnetic vector potential from magnetic field
Pingback: Magnetic vector potential as the curl of another function
Pingback: Average magnetic field within a sphere
Pingback: Magnetization: bound currents
Pingback: Mutual inductance
Pingback: Energy in a magnetic field
Pingback: Faraday field and magnetic vector potential
Pingback: Coulomb and Lorenz gauges
Pingback: Electromagnetic field tensor: four-potential
MAGNETIC VECTOR POTENTIAL AND CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.23.
The relation between magnetic vector potential A and current density J
is given by

(1) 2 A = 0 J

This equation makes it fairly easy to find the current density produced
by a given vector potential, but we do need to be careful if were using a
coordinate system other than the rectangular one.
In rectangular coordinates, this equation is essentially 3 Poisson equa-
tions, one for each component. Because the rectangular unit vectors are
constant over all space, the Laplacian operator doesnt affect them. When
we use a different coordinate system, such as cylindrical or spherical, the
unit vectors do vary from place to place so the Laplacian does have an ef-
fect on them, and we cant just separate A into its components and apply
the equation to each one.
There are two ways we could approach this problem: one is to convert A
to rectangular coordinates, and the other is to use the vector identity

(2) 2 A = ( A) ( A)

since the two vector derivatives on the RHS are well defined for all coordi-
nate systems.
As a simple example of this, suppose we have a vector potential in cylin-
drical coordinates:

(3) A = k

where k is a constant. The formulas for the divergence, gradient and curl
are:
1
MAGNETIC VECTOR POTENTIAL AND CURRENT 2

1 1 A Az
(4) A = (rAr ) + +
r r r z
(5)
    "  #
1 Az A Ar Az 1 rA Ar
A = r + + z
r z z r r r
f 1f f
(6) f = r + + z
r r z
Applying these formulas we get

(7) A = 0
k
(8) A = z
r
k
(9) ( A) = 2
r
Therefore

k
(10) J=
0 r2
This is a current that rotates around the z axis and decreases with distance
from the axis. Wed obviously need to avoid r = 0 so the current remains
finite.
MAGNETIC VECTOR POTENTIAL OF CONSTANT FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.24.
The magnetic vector potential A for a uniform (that is, constant) magnetic
field B is

1
(1) A = rB
2
To check this, we calculate its divergence and curl. For the divergence,
using a vector calculus identity

1
(2) A = [B ( r) r ( B)]
2
The last term is zero since B is a constant, and r = 0 as can be
checked by direct calculation, so A = 0 which is one condition required
of A.
For the curl, we get (omitting terms involving a derivative of B, which
are all zero):

1
(3) A = (r B)
2
1
(4) [(B ) r B ( r)]
2
1
(5) = [B 3B]
2
(6) = B
To get the third line from the second, heres a sample calculation of the x
component of the first term:

(7) (B ) xx = (Bx x + By y + Bz z ) xx
(8) = Bx x
The y and z components work the same way. Weve also used r = 3
as can be checked by direct calculation.
1
MAGNETIC VECTOR POTENTIAL OF CONSTANT FIELD 2

Thus the div and curl of A give the correct values. Its not unique, since
we can add any vector field to A so long as its div and curl are both zero.
This is true of any constant field, so in general

1
(9) A = rB+C
2

where C is constant over all space.


P INGBACKS
Pingback: Magnetic vector potential from magnetic field
Pingback: Magnetic vector potential as the curl of another function
MAGNETIC VECTOR POTENTIAL OF AN INFINITE WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.25.
The magnetic vector potential A can be evaluated from


0 J (r0 ) 3 0
(1) A (r) = d r
4 V |r r0 |

provided that all currents are contained within a finite volume. If the cur-
rents extend to infinity we have to use a different method.
One example is that of an infinite wire carrying a steady current I. We can
work out the potential by applying Stokess theorem. From the definition of
A, we know that A = B, so if we define a closed loop then we have


(2) A dl = B da
A

where the integral on the LHS is a line integral around the loop and the
integral on the right is over the area A enclosed by the loop. For a flat loop,
the direction of integration on the LHS is such that the right hand rule gives
a vector pointing in the same direction as da on the RHS.
For an infinite wire along the z axis, the magnetic field, using cylindrical
coordinates, is

I0
(3) B=
2r

where r is the distance from the axis of the wire.


We choose a rectangular loop with one edge a distance a from the wire
and the opposite edge a distance b from the axis (with length of the parallel
sides taken as 1), we can integrate B over this area:
1
MAGNETIC VECTOR POTENTIAL OF AN INFINITE WIRE 2

b
I0
(4) B da = dr
A a 2r
I0
(5) = (ln b ln a)
2
By symmetry, A has the same magnitude for all points at a given distance
r from the axis, and since the 2 edges perpendicular to the wire are traversed
in opposite directions, their contributions to the line integral cancel, and we
get

(6) A dl = A(a) A(b)

Thus it seems a good candidate is

I0
(7) A= ln rz
2
We can check its div and curl, and we find (using standard formulas for
div and curl in cylindrical coordinates):

(8) A = 0
I0
(9) A =
2r
Now suppose we look inside the wire, taking the wires radius to be R.
At a distance r from the axis, only current within that radius contributes to
the field, so we get from Ampres law using a circular loop within the wire

r2
(10) 2rB = 0 2 I
R
0 rI
(11) B =
2R2
Following the same procedure as above, only this time choosing a rect-
angular loop entirely within the wire, we get

b
0 I
(12) A (a) A (b) = rdr
2R2 a
0 I
b2 a2

(13) = 2
4R
Thus it seems a candidate is
MAGNETIC VECTOR POTENTIAL OF AN INFINITE WIRE 3

0 I 2
(14) A (r) = r z
4R2
Again, we get A = 0 and A = B, so it looks like were done.
However, there is one slight snag here: A isnt continuous at r = R. Al-
though in practice this might not make much difference since it is only the
derivatives of A that have physical meaning (and weve seen that both the
div and curl are correct), its nice to tidy up the solution.
The key is that when we did the line integrals above, we ended up only
with expressions for the difference between A at two distances from the
wire. The potential itself could have a non-zero constant of integration
which cancels out in the difference (and also wouldnt contribute to the
div or curl). Thus wed like to choose a couple of constants so that

I0 I0 2

(15) (ln R +C1 ) = R +C 2
2 4R2
The simplest choice is to make both sides zero, so

(16) C1 = ln R
(17) C2 = R2
Our final solution is then
(
I
2
0
ln Rr rR
(18) A= 0 I 2 2

4R2 r R rR

P INGBACKS
Pingback: Magnetic vector potential: sheet of current
MAGNETIC VECTOR POTENTIAL: SHEET OF CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.26.
Another example of calculating the magnetic vector potential in a case
where the current extends to infinity. We consider a uniform sheet of cur-
rent in the xy plane, carrying surface current density K x. Using the same
argument as in the case of a slab of current, the magnetic field due to this
current is

(
0 K y z>0
(1) B = 0 2
2 K y z<0

Note that the field is independent of the distance from the sheet of current.
We can work out the potential by applying Stokess theorem. From the
definition of A, we know that A = B, so if we define a closed loop then
we have


(2) A dl = B da
A

where the integral on the LHS is a line integral around the loop and the
integral on the right is over the area A enclosed by the loop. For a flat loop,
the direction of integration on the LHS is such that the right hand rule gives
a vector pointing in the same direction as da on the RHS.
We need to be careful in defining directions for the area and line integrals
to get the signs right. Well take as our area A a rectangle above the xy
plane and parallel to the xz plane. The sides of the rectangle parallel to the
xy plane are of length 1, with the lower side at a distance a and the upper
side at a distance b from this plane. The normal to the area points in the y
direction Since B is uniform everywhere above the plane, we get


0
(3) B da = K (b a)
A 2
1
MAGNETIC VECTOR POTENTIAL: SHEET OF CURRENT 2

For the line integral, the path around the rectangle is clockwise when
looking in the +y direction, and by symmetry, the integral of A dl along
the vertical sides cancels out, so

(4) A dl = Ax (b) Ax (a)

Comparing these two, a reasonable candidate is

0
(5) A= Kzx
2
We can check this by finding the div and curl, as usual:

(6) A = 0
0
(7) A = K y
2
Below the sheet of current, the sign is reversed so we have

0
(8) A= Kzx
2
MAGNETIC VECTOR POTENTIAL: DIV, CURL AND
LAPLACIAN

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.27.
Weve seen how the magnetic vector potential is derived from the fact
that the magnetic field can be expressed as the curl of a vector field. The
fact that A gives the Biot-Savart equation for the magnetic field can be
obtained by just reversing the derivation of A.
We can check a couple of other derivatives from the definition of A,
which is


0 J (r0 ) 3 0
(1) A (r) = d r
4 V |r r0 |
First, the divergence A. Remember that the derivative is taken only
with respect to the unprimed coordinates, so the divergence works out to


0 1
J r0 d 3 r0

(2) A = 0
4 V |r r |
If we look at the derivative with respect to x term, we have

(3)
   

0 1 3 0 0 1
d 3 r0

Jx r d r = Jx r
V x |r r0 | V x 0 |r r0 |

0
 1 1
Jx r0 d 3 r0

(4) = Jx r 0
+ 0 0
|r r | A
V |r r | x

1
In the first line, we used the fact that the derivative of |rr0 | with respect
0
to x is the negative of the derivative with respect to x . We then integrated
by parts. The integrated term is evaluated over the boundary surface A. We
can take this surface to lie outside the region in which currents exist, so
Jx = 0 there. We will get similar results for the y and z terms in the original
integral, so if we combine them, we get
1
MAGNETIC VECTOR POTENTIAL: DIV, CURL AND LAPLACIAN 2


0 1 1
d 3 r0 = 0 0
  3 0
(5) J r 0
J r d r
V |r r | V |r r0 |
For steady currents, 0 J (r0 )
= 0, so in the case of localized, steady
currents, we have A = 0, which is what we assumed in our derivation of
A, so this is consistent.
We also had a version of Poissons equation for the potential:

(6) 2 A = 0 J
Taking the x component, we have

0 1
2
d 3 r0
Jx r0 2

(7) Ax (r) = 0
4 |r r |

V

= 0 Jx r0 3 r r0 d 3 r0
 
(8)
V
(9) = 0 Jx (r)

where weve used a result for the delta function.


Similar relations for y and z verify Poissons equation.
MAGNETIC SCALAR POTENTIAL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.28.
Since the electric field can be expressed as the gradient of a scalar poten-
tial (E = V ), its natural to wonder if the magnetic field could be written
as the gradient of a scalar potential, as well as the curl of a vector potential.
That is, we would like to try

(1) B = U

for some function U. This has one obvious flaw, in that Ampres law states
that the curl of B is non-zero in the presence of currents. That is

(2) B (r) = 0 J (r)

This is a problem, because the curl of a gradient is always zero (this


is an identity from vector calculus). Still, we might try to define a scalar
potential in current-free areas. If we try this, we can see one consequence
by considering the magnetic field due to the current I in an infinite wire. In
cylindrical coordinates, this is

0 I
(3) B=
2r

If we write this as U, then using the form for the gradient in cylindrical
coordinates there will be only a term:

1 U
(4) U =
r

Now if we apply Ampres law to a circular path surrounding the wire


1
MAGNETIC SCALAR POTENTIAL 2


(5) B dl = U dl
2  
1 U 
(6) = r d
0 r
2
U
(7) = d
0
(8) = U (0) U (2)
(9) = 0 I
That is, the difference U (2) U (0) 6= 0 even though = 0 and = 2
represent the same point. The potential must therefore be a multiple-valued
quantity, so its not of much use in physical situations.
P INGBACKS
Pingback: Magnetic scalar potential: dipole and rotating sphere
MAGNETIC FIELD OF ROTATING SPHERE OF CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.29.
Since any moving charge generates a magnetic field, one way of produc-
ing a novel current is to take a uniform sphere of charge and set it spinning
on its axis. To work out the field produced by such a sphere, we can start
with the field generated by a spinning spherical shell of charge. The deriva-
tion of this field is surprisingly tricky, and is given by Griffiths as example
5.11. The results for the vector potential are
(
1
3 0 R r sin rR
(1) A (r, , ) = 0 R4 sin
3 r2
rR
Here, R is the radius of the shell, is the surface charge density and is
the angular velocity, where the spheres axis is taken to be the z axis.
From this we can calculate the magnetic field B = A using the stan-
dard formula for the curl in spherical coordinates:
( 2
0 R

3 cos r sin rR
(2) B= 0 R4 1 
3 r 3 2 cos r + sin rR
Notice that although A is continuous at r = R, B is not. Well look at
this in more detail in a future post. Its also worth noting that since z =
cos r sin , the field inside the shell points along the rotation axis and
is uniform.
To find the field due to a solid spinning sphere of charge with charge
density , we can integrate a series of spherical shells. In doing this, we
need to be very careful in interpreting the various symbols for the radius.
For a thin shell of thickness dr, the charge per unit area on this shell is
dr, so this will replace in the equations above. In these equations, R is
the radius of the shell, and r is the observation radius. We need the overall
radius of the solid sphere, which well define as R0 . Thus the integration
variable will be R, since it is the radius of the shells that we need to vary.
Then for a value of r inside the sphere, we will get a contribution to the
field from those shells with a radius R < r by using the second equation
1
MAGNETIC FIELD OF ROTATING SPHERE OF CHARGE 2

above, and for R > r by using the first equation. Note that we do not inte-
grate over either r or , since these two coordinates define the observation
point. We get for the field due to shells interior to the observation radius:
r
0 1
R4 dR

(3) Bi = 3
2 cos r + sin
3 r 0
0 r2 
(4) = 2 cos r + sin
3 5
From shells outside the observation radius:

20  R0
(5) Bo = cos r sin RdR
3 r
0 2
R0 r2 cos r sin
 
(6) =
3
The total field is the sum so
    
0 3 2 6 2
(7) B= R20 r cos r + 2
r R0 sin
3 5 5
3Q
If we know the total charge Q in the sphere, then = 4R30
and

3 r2
   2  
0 Q 6r
(8) B= 1 cos r + 1 sin
4R0 5 R20 5 R20
3 r2 6 r2
 
0 Q
(9) = z cos r + sin
4R0 5 R20 5 R20
Although the last version mixes spherical and rectangular coordinates, it
shows that there is a uniform field in the z direction with a varying field
superimposed on top of it.
P INGBACKS
Pingback: Magnetostatic boundary conditions: a couple of examples
Pingback: Magnetic force of attraction between hemispheres
Pingback: Magnetic fields of spinning disk and sphere
Pingback: Magnetic scalar potential: dipole and rotating sphere
Pingback: Magnetic potential and field of a rotating sphere of charge
DIVERGENCELESS VECTOR FIELD AS A CURL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.30.
This is more mathematics than physics, but it relates to the magnetic
vector potential so here we go. The vector potential A is defined so that B =
A and B = 0. Wed like to prove that, in general, any divergenceless
vector field F can be written as the curl of another vector field A. The curl
F = A has three components:

(1) y Az z Ay = Fx
(2) x Az z Ax = Fy
(3) x Ay y Ax = Fz
Thus in principle, we have 3 coupled PDEs to solve, but we can get a
solution by starting with the assumption that Ax = 0 (this isnt the only
solution, of course, but it does give a solution). Then

(4) x Az = Fy

Az = Fy x0 , y, z dx0 + G (y, z)

(5)
(6) x Ay = Fz

Fz x0 , y, z dx0 + H (y, z)

(7) Ay =

where G and H are functions of integration; since the integrals are with
respect to x, the constants of integration can be functions of y and z.
We can now plug these into the first component of the curl above:

(8)
z Fz x , y, z dx0 + y G (y, z)
0
y Fz x0 , y, z dx0 z H (y, z)
 
y Az z Ay =

(9) = ( F + x0 Fx0 ) dx0 + y G (y, z) z H (y, z)
1
DIVERGENCELESS VECTOR FIELD AS A CURL 2

If we now want the result at a particular point (x, y, z), we can introduce
limits on the integral, and also use the requirement that F = 0:

x
(10) Fx (x, y, z) = x0 Fx0 dx0 + y G (y, z) z H (y, z)
0
(11) = Fx (x, y, z) Fx (0, y, z) + y G (y, z) z H (y, z)
(12) 0 = Fx (0, y, z) + y G (y, z) z H (y, z)

At this point we can proceed in various ways, since the functions G and
H have between them only this one condition. One option is to choose

y
Fx 0, y0 , z dy0

(13) G (y, z) =
0
(14) H (y, z) = 0

With this option, we get

y x
0 0
Fy x0 , y, z dx0
 
(15) Az = Fx 0, y , z dy
0 x 0

Fz x0 , y, z dx0

(16) Ay =
0
(17) Ax = 0

We could equally as well have chosen

(18) G (y, z) = 0
z
Fx 0, y, z0 dz0

(19) H (y, z) =
0

or some combination of the two.


Using the first option, we can check the curl.
DIVERGENCELESS VECTOR FIELD AS A CURL 3

(20)
 y x   x 
Fx 0, y , z dy0
0 0 0
Fz x , y, z dx0
0
  
( A)x = y Fy x , y, z dx z
0 0 0
(21)
x
= Fx (0, y, z) + ( F + x0 Fx0 ) dx0
0
(22)
= Fx (0, y, z) + Fx (x, y, z) Fx (0, y, z)
(23)
= Fx (x, y, z)
(24)
 y x 
Fx 0, y , z dy0
0 0
  0
( A)y = x Fy x , y, z dx + z (0)
0 0
(25)
= Fy (x, y, z)
(26)
x
Fz x0 , y, z dx0 y (0)

( A)z = x
0
(27)
= Fz (x, y, z)

For the divergence

(28) 
x y x 
Fz x , y, z dx0 +z
0 0 0
Fy x , y, z dx0
0
  
A = x (0)+y Fx 0, y , z dy
0 0 0

This isnt zero in general.


For a specific example, consider

(29) F = yx + zy + xz

Then
DIVERGENCELESS VECTOR FIELD AS A CURL 4

(30) Ax = 0
x
Fz x0 , y, z dx0

(31) Ay =
0 x
(32) = x0 dx0
0
x 2
(33) =
2 y x
Fx 0, y , z dy0
0
Fy x0 , y, z dx0
 
(34) Az =
0 y x 0
0 0
(35) = y dy zdx0
0 0
y2
(36) = xz
2
By direct calculation

(37) ( A)x = y Az z Ay
(38) = y = Fx
(39) ( A)y = z Ax x Az
(40) = z = Fy
(41) ( A)z = x Ay y Ax
(42) = x = Fz
MAGNETOSTATIC BOUNDARY CONDITIONS: A COUPLE OF
EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.31.
Here are a few examples of the magnetic boundary conditions. First con-
sider again the infinite solenoid, which we treated as a surface current travel-
ling around a cylinder. We found that the magnetic field outside the solenoid
was zero, and inside it was uniform with a value of B = 0 nI where n is the
number of turns of wire per unit length and I is the current. Translated into
surface current density, we have nI = K, so B = 0 K. Thus the discontinuity
at the surface is indeed 0 K as our previous analysis showed.
Now consider the rotating spherical shell of charge. We found that the
vector potential was
(
1
3 0 R r sin rR
(1) A (r, , ) = 0 R4 sin
3 r2
rR
At the surface r = R and its clear that A is continuous there. For the
normal derivative, we have
(
1
3 0 R sin rR
(2) r A = 4
20 R3 sinr3 rR
The difference as we cross the shell is (taking the outer minus the inner
values):

(3) r (A) = 0 R sin


The linear velocity of a patch of area at angle is R sin , so the surface
current at that point is K = R sin . Thus r (A) = 0 K as required.

1
MAGNETOSTATIC BOUNDARY CONDITIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.32.
A while back, we worked out the behaviour of the electric field and po-
tential at a layer of surface charge. We can do a similar analysis for the
magnetic field and its behaviour as we cross a surface current.
Suppose we have a surface with a surface current K. At a local point,
lets say this current flows in the +x direction. Consider a small patch of
area and build a little pillbox that straddles the surface at this area. If we
make the thickness of the box infinitesimally thin, then a surface integral is
essentially over the two sides of the box on either side of the current. Since
B = 0, Gausss law says that

3
(1) Bd r = B da = 0
V A
Now B da = B da so the second integral says that
 
(2) B da = Babove
Bbelow
A=0
A

Therefore, the normal component of B is continuous across a surface cur-


rent.
For the parallel component, consider first the component of B parallel to
the surface but perpendicular to the current. We can define a little loop that
straddles the surface, where the area enclosed by the loop is perpendicular
to the current. If we make the vertical sides of the loop infinitesimal and the
horizontal sides of length 1, then we can use Stokess theorem to say

(3) B dl = Babove
k Bbelow
k = 0 I = 0 K

Finally, if we take a loop perpendicular to the


surface but parallel to the
current, then the loop encloses zero current so B dl = 0 and this compo-
nent of the field is continuous.
Thus the only component of B that has a discontinuity is the one parallel
to the surface but perpendicular to the current. That is, the discontinuity is
1
MAGNETOSTATIC BOUNDARY CONDITIONS 2

perpendicular both to the normal n to the surface and the current K, so the
difference must be expressible as the cross product of these two vectors:

(4) Babove Bbelow = 0 K n


To get the direction of the discontinuity, suppose we look in the +x direc-
tion (that is, in the direction of the current). Then the field above the current
points to the right (that is, in the y direction) and below it points to the left
(+y direction). Thus the difference Babove Bbelow points to the right. This
works out properly if K points in the +x direction and n points in the +z
direction.
As for the vector potential, assuming A = 0 means that A is contin-
uous (using the same argument as above). Since B = A, we can again
use Stokess theorem to integrate A around a loop straddling the surface:


(5) A dl = B da

where the second integral is over the area enclosed by the loop. Unlike the
integral Bdl above, though, the flux of B enclosed by the loop diminishes
to zero as we make the loop thinner and thinner. There is no infinitesimally
thin sheet of magnetic field that is always enclosed by the loop as there
was in the case of enclosed current. Thus in the limit, A dl = 0 for
any orientation of the loop across the surface, and all components of A are
continuous across a surface current.
The derivative of A does have a discontinuity however. To see this, sup-
pose we set up a coordinate system with z the normal to the area patch and
y the direction of the current K. Then

(6) B = Babove Bbelow = 0 K n = 0 K x


Writing out B = (A) we have

(7) Bx = y (Az ) z (Ay ) = 0 K


(8) By = z (Ax ) x (Az ) = 0
(9) Bz = x (Ay ) y (Ax ) = 0
Since A is continuous across the surface, the derivatives in directions
parallel to the surface (that is, x and y) will be the same on both sides, so x
and y derivatives of A will all be zero. This gives us
MAGNETOSTATIC BOUNDARY CONDITIONS 3

(10) Bx = z (Ay ) = 0 K
(11) By = z (Ax ) = 0
To get the third derivative we use A = 0, which gives

(12) z (Az ) = x (Ax ) y (Ay ) = 0


Therefore, the y component of the normal derivative of A has a disconti-
nuity:

(13) z (A) = 0 K y
P INGBACKS
Pingback: Magnetostatic boundary conditions: a couple of examples
Pingback: Magnetostatic boundary conditions; Laplaces equation
Pingback: Boundary between magnetic materials
Pingback: Surface current induced by a magnet floating over a supercon-
ductor
MAGNETIC DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.33.
In much the same way as the electric potential, we can write the mag-
netic potential as a multipole expansion. Because the magnetic potential
is a vector, the expansion is a series of vector integrals rather than volume
integrals.
The most general form of the vector potential is

0 J (r0 ) 3 0
(1) A (r) = d r
4 V |r r0 |
For a steady line current I, this becomes a line integral around the current
loop:

0 I 1
(2) A (r) = dl0
4 |r r0 |
We can expand the integrand in terms of Legendre polynomials in the
same way as for the electric potential:

1 1
(3) 0
=
|r r | r + r 2rr0 cos 0
2 02
1 1
(4) = q
r 1 + r02 2 r0 cos 0
r2 r
 0 n
1 0 r
(5) = Pn (cos )
r n=0 r
The potential becomes

0 I 1
(6) A (r) = rn+1 Pn (cos 0 )r0n dl0
4 n=0
The first three terms in this sum are called the monopole, dipole and
quadrupole terms. The monopole term is
1
MAGNETIC DIPOLE 2


0 I
(7) A0 = dl0 = 0
4r

since integration of the vector line element around any closed loop is zero.
For the dipole:

0 I
(8) A1 = r0 cos 0 dl0
4r2
Since 0 is the angle between r0 and r, we can write this as

0 I
(9) A1 = r0 rdl0
4r2
Since r is a constant as far as the integral is concerned, we can write the
integral in terms of the vector area a enclosed by the loop.

0 I
(10) A1 = a r
4r2
The quantity

(11) m Ia

is defined as the magnetic dipole moment of the current loop, so the dipole
potential is

0
(12) A1 = m r
4r2
The magnetic field due to a dipole is

(13) B1 = A1
 
0 r
(14) = m 2
4 r
We can use a vector identity to expand the curl, and use the fact that m is
a constant for a given current loop, so its derivatives are all zero. We get
   
0 r r
(15) B1 = m 2 (m ) 2
4 r r
We can now use
MAGNETIC DIPOLE 3

1
(16) r = (xx + yy + zz)
r
p
(17) r = x2 + y2 + z2
Therefore

(18)
r x y z
2
= x 3 + y 3 + z 3
r r r r

3 2x 2 1
 
3 2y2 1
 
3 2z2 1

(19) = 5 + 3 + 5 + 3 + 5 + 3
2 r r 2 r r 2 r r
(20) =0
(21)
3 2x2 1
  
r 3 2xy 3 2xz
(m ) 2 = mx 5 + 3 x y z + my [. . .] + mz [. . .]
r 2 r r 2 r5 2 r5
xmx mx ymy my zmz mz
(22) = 3 5 r + 3 x 3 5 r + 3 y 3 5 r + 3 z
r r r r r r
1
(23) = 3 [3 (m r) r + m]
r
Putting this all together, we get

0
(24) B1 = [3 (m r) r m]
4r3

P INGBACKS
Pingback: Magnetic dipole moment of a current loop
Pingback: Magnetic dipole moment of spinning disk
Pingback: Magnetic dipole moment of spinning spherical shell
Pingback: Magnetic dipole moment of a current loop
Pingback: Magnetic dipole in a constant field
Pingback: Magnetic potential and field of a rotating sphere of charge
Pingback: Magnetic dipole moment for volume current
Pingback: Magnetic dipole field of a finite solenoid
Pingback: Torque on a magnetic dipole
Pingback: Magnetization: bound currents
Pingback: Energy of magnetic dipole in magnetic field
Pingback: Magnetic dipole embedded in sphere
Pingback: Mutual inductance
Pingback: Mutual inductance between two magnetic dipoles
MAGNETIC DIPOLE 4

Pingback: Surface current induced by a magnet floating over a supercon-


ductor
Pingback: Floating a magnet above a superconductor
Pingback: Momentum in a magnetized and polarized sphere
Pingback: Fields of an oscillating magnetic dipole
Pingback: Radiation from the magnetic dipoles of Earth and pulsars
Pingback: Magnetic cooling
MAGNETIC DIPOLE MOMENT OF A CURRENT LOOP

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.34.
Heres a simple example of the magnetic dipole. Weve got a circular
loop of current I with radius R lying in the xy plane and centered at the
origin. The dipole moment is

(1) m = Ia

where a is the vector area of the loop, which is just a = R2 z. Thus

(2) m = IR2 z
Using the dipole field as an approximation for points far from the loop,
we have

0
(3) B1 = [3 (m r) r m]
4r3
0 IR2
(4) = (3 cos r z)
4r3
0 IR2 
(5) = 2 cos r + sin
4r3
For points on the z axis this becomes

0 IR2
(6) B1 = z
2z3
The exact formula for points on the z axis (given as example 5.6 in Grif-
fiths) is

0 IR2
(7) B= 3/2
z
2 (R2 + z2 )
For z  R, this gives the same result as the dipole field.
1
MAGNETIC DIPOLE MOMENT OF A CURRENT LOOP 2

P INGBACKS
Pingback: Magnetic dipole moment of spinning disk
Pingback: Magnetic dipole moment of spinning spherical shell
Pingback: Magnetic scalar potential: dipole and rotating sphere
Pingback: Gyromagnetic ratio
Pingback: Magnetic dipole: average field over a sphere
Pingback: Magnetic dipole field of a finite solenoid
Pingback: Torque on a magnetic dipole
MAGNETIC DIPOLE MOMENT OF SPINNING DISK

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.35.
Heres another example of the magnetic dipole. We have a circular disk
with surface charge density that is rotating at an angular speed . We
can use the result for the magnetic dipole moment of a circular loop and
integrate this to get the dipole moment for the disk.
The moment for the loop is

(1) m = Ir2 z
The linear speed of a loop of radius r is r so the current at that radius is
dI (r) = r dr. The total dipole moment is then
R
(2) m = z r3 dr
0
R4
(3) = z
4

1
MAGNETIC DIPOLE MOMENT OF SPINNING SPHERICAL
SHELL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.36.
Now for the magnetic dipole moment of a spinning spherical shell of
charge. We can divide the sphere up into a number of circular loops. Each
loop corresponds to a particular value of and has radius R sin . The
current of each loop is dI = R sin Rd and the magnetic moment of
the loop is

(1) dm = z (R sin )2 R2 sin d


The total dipole moment is then

4
(2) m = z R sin3 d
0
4
(3) = R4 z
3
For r > R, the potential of the sphere is (Griffiths example 5.11):

0 R4 sin
(4) A=
3 r2
The dipole approximation for the potential is

0
(5) A = m r
4r2
0 R4 sin
(6) =
3 r2

which follows since is the angle between z and r so z r = sin . Thus


outside the sphere, the dipole potential is the exact potential.
P INGBACKS
Pingback: Magnetic potential and field of a rotating sphere of charge
Pingback: Magnetic dipole embedded in sphere
1
MAGNETIC DIPOLE MOMENT OF A CURRENT LOOP

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.37.
Earlier we worked out the magnetic field at the centre of a square loop
(side length w) of current which lies in the xy plane with its centre at the
origin. We can generalize this a bit by working out the field at any point on
the z axis. Start with the Biot-Savart law for linear currents:


0 I dl0 (r r0 )
(1) B (r) =
4 |r r0 |3

Now consider the edge that extends from y = w/2 to y = +w/2 at x =


+w/2. Along this edge we have

(2) r r0 = x0 x y0 y + zz
w
(3) = x y0 y + zz
2
(4) dl = dy0 y
0

0 0
  w  0
(5) dl r r = zx + z dy
2
The integral is then


0 I  w  w/2 dy0
(6) B1 = zx + z 3/2
4 2 
w/2 w2
4 + y02 + z2
0 I  w  8w
(7) = zx + z
4 2 2w2 + 4z2 (w2 + 4z2 )

By symmetry, the opposite edge at x = w/2 will contribute the same


z component and the opposite x component. Similarly, the two edges at
y = w/2 will contribute two more z components with the y components
cancelling. Thus the total field is
1
MAGNETIC DIPOLE MOMENT OF A CURRENT LOOP 2

0 I w 8w
(8) B = 4 z
4 2 2w2 + 4z2 (w2 + 4z2 )
4w2 0 I
(9) = z
2w2 + 4z2 (w2 + 4z2 )
As a check, when z = 0 this reduces to

2 20 I
(10) B (0) = z
w
This matches the earlier post (where w = 2R).
The dipole moment of the square is m = Ia = w2 z, so the dipole compo-
nent of the field is

0
(11) B1 = [3 (m r) r m]
4z3
0 I
3w2 w2 z

(12) = 3
4z
0 w2 I
(13) = z
2z3
If we take z  w in the exact formula, it reduces to

4w2 0 I
(14) B z
4z2 (4z2 )
0 w2 I
(15) = z
2z3
Thus the dipole term is a good approximation for large distances.
CURRENTS AND RELATIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.38.
Using the Biot-Savart law and the Lorentz force law, we can see that if we
have two parallel wires carrying the same current in the same direction, they
will feel a magnetic force of attraction. Using the same logic, we can view
the current within a wire (with a non-zero radius) as a number of parallel
currents, so it would seem that the current within a wire should be drawn
together towards the centre of the wire. The reason this doesnt happen, of
course, is because there are also the stationary positive charges within the
wire, so if the negative electrons carrying the current clustered towards the
centre, the electrical attraction of the positive ions they leave behind will
tend to pull them back towards the outside of the wire.
We can make this quantitative by looking at the case of a wire in which
the positive charges have a uniform density + and are stationary, and the
negative charges have a density and move relative to the positive ions
with a speed v.
The magnetic field due to the current in an infinite wire is

0 I
(1) B=
2d

where d is the distance from the centre of the wire, and I is the current
interior to this distance (current in the cylindrical shell outside d doesnt
produce any field within the shell). This current is given by I = d 2 v, so

0
(2) B= d v
2
The magnetic force on a unit test charge moving with speed v along the
wire is then given by the Lorentz force law and is

0
(3) FB = (1) vB = d v2
2
The electric field due to an infinite line charge is
1
CURRENTS AND RELATIVITY 2


(4) E=
20 d

where is the net charge density. Here, this means = d 2 (+ + )


(where is taken to be negative), so the electric force on a unit charge a
distance d from the centre is

d (+ + )
(5) FE = (1) E =
20
Equating the two forces gives

(6) FB = FE
0 d (+ + )
(7) d v2 =
2 20
+
(8) =
1 v2 0 0
+
(9) =
1 v2 /c2

where c = 1/ 0 0 is the speed of light.
Thus it appears that the negative charge density is larger than the positive
density, which appears to contradict the supposed neutrality ofp the wire.
The key to resolving this is to notice the telltale term = 1/ 1 v2 /c2 ,
that is

(10) = 2 +
This factor turns up all over the place in special relativity, so lets try
analyzing the situation from the two frames of reference: the positive ions
and the moving electrons. In the ion frame (the rest frame for the wire),
we observe (backed by experimental data) that the wire is neutral, so in that
frame, we must have the density of the positive charges equal and opposite
to the density of negative charges. If we let 0 be the positive charge density,
then in the rest frame

(11) + = 0 =
However, we are actually specifying the densities of the two charge types
in two different inertial frames. The positive charges are at rest, but the neg-
ative charges are moving with speed v. If these two densities are specified
CURRENTS AND RELATIVITY 3

using a unit length in the rest frame, how will they appear in the electron
frame?
The unit length of positive charge is at rest in the ion frame, so when
we transform to the electron frame, it will undergo Lorentz contraction and
appear have a length of 1/. Since the length is smaller, but still contains
the same number of charges, the density is higher, so we get

0
(12) + = + = 0
Now the unit length for negative charge as measured by the ions is mov-
ing along at a speed v relative to the ions, so it is already Lorentz contracted.
If we shift to the electron frame where this length is at rest, it will appear to
expand by a factor of , which means the density of negative charge appears
to decrease, giving

0
(13) = / = 0 /
Thus in the electron frame, we get

0 0
(14) + = 2
(The reason the positive and negative charges are reversed from what they
were in the first derivation above is that traditional electromagnetic theory
assumes that current is caused by motion of positive charge, whereas in the
relativistic derivation, we assumed that negative charge is moving.)
Thus, according to relativity, a current-carrying wire actually does ac-
quire a net charge density. It turns out that this is true, but a full analysis
of how electric fields are calculated in relativity would take us too far afield
here.
P INGBACKS
Pingback: A flawed theory of gravity
HALL EFFECT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.39.
The Hall effect occurs when a current-carrying substance is placed in a
magnetic field that is perpendicular to the direction of the current. Suppose
we have a wire with a rectangular cross-section that carries current in the
+y direction. A magnetic field pointing in the +x direction is applied to
the wire. From the Lorentz force law, a moving charge in the wire feels a
magnetic force qv B, so it will be deflected in the z direction, where the
sign of the deflection depends on the sign of the charge and the direction of
motion. If the charges are positive and flowing in the +y direction, they are
deflected in the z direction.
As a result, a charge imbalance is created inside the wire resulting in an
electric field in the z direction. Equilibrium is established when the electric
and magnetic forces balance, and this happens when qE = qv B. For
positive charges, this means that E = vB, so if the wire has a thickness t in
the z direction, the potential difference across the wire is Et = vBt.
If the charges are negative, then to produce the same current as above
they would have to be moving in the y direction. Since the direction of
motion and the sign of the charges are both opposite to the first case, the
negative charges will still be deflected downwards, so the direction of the
induced electric field will be reversed. Thus by measuring the sign of the
potential difference we can tell whether the charge carriers are positive or
negative.

1
CURRENT LOOP IN A MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.40.
The generalization of the Lorentz force law to the case of a current-
carrying wire in a magnetic field is

(1) F=I dl B

where the integral is taken along the path of the wire. In the case of a
constant field, this becomes
 
(2) F=I dl B

That is, the field comes outside the integral and the integral is just the vector

sum of segments along the path, so the result of the integral is w = dl
which is the vector pointing from the start to the end of the path. For a
closed loop entirely within a constant field, w = 0 and there is no force. For
a loop (of any shape; it doesnt even have to be flat) that is partially within
a constant field (that is, one continuous portion of the loop is within the
field, while the remainder of the loop lies outside the field), w is the vector
pointing from the point on the wire where the wire enters the field to the
point where it leaves it. If the field is perpendicular to w, then the force is
perpendicular both to the field and to w. The magnitude of the force in this
case is F = IwB.
P INGBACKS
Pingback: Dropping loops through magnetic fields

1
RADIALLY SYMMETRIC MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.41.
This is a rather contrived example of the use of the Lorentz force law. We
have a circular disk of radius R with a magnetic field applied perpendicular
to the disk. The field depends only on the distance r from the centre of the
disk, and its distribution satisfies B da = 0. That is, the field varies in
direction and magnitude in such a way that the total flux through the disk is
zero. Since the field is radially symmetric, this means

R
(1) B da = 2 B(r)rdr = 0
0

Now suppose we start off a particle with charge q at the centre of the
disk moving with a velocity v. Assuming that v is such that the particle
eventually reaches the edge of the disk, at what angle to the edge will it
leave?
We can approach this problem by considering the particles angular mo-
mentum, since if the particle leaves the disk along a line parallel to the
radius vector pointing to its point of departure, the angular momentum is
zero.
Since the particle starts at the origin, its initial angular momentum is zero,
so the total change in angular momentum as the particle travels to the edge
must be zero in order for it leave radially. The rate of change of angular
momentum is the torque N, so we have, with the total time being T :

T
(2) L = Ndt
0
T
(3) = r Fdt
0
T
(4) = q r (v B) dt
0
T
(5) = q [v (r B) B (r v)] dt
0
1
RADIALLY SYMMETRIC MAGNETIC FIELD 2

The motion occurs entirely in the plane of the disk, so r B for the
entire path, so the first term in the integrand is zero. In the second term,
r v = rr v = r dr
dt , since r v is the component of the velocity along the
radial direction. Therefore
T
dr
(6) L = q Br dt
0 dt
R
(7) = qz B (r) rdr
0
(8) = 0
MAGNETIC FORCE OF ATTRACTION BETWEEN
HEMISPHERES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.42.
Another example of the Lorentz force law. This time well find the mag-
netic force between the north and south hemispheres of a rotating shell of
charge. Were faced with a bit of a dilemma here, since the magnetic field
is discontinuous across the shell:
( 2  20 R
0 R
3 cos r sin = 3 z r < R
(1) B= 0 R4 1 
3 r 3 2 cos r + sin r>R

where R is the radius of the sphere, is the angular speed and is the
surface charge density.
The current of an infinitesimal strip is

(2) I = (R sin ) (Rd )


The usual procedure in cases like this is to take the average of the field
on either side of the current layer. In this case, we can simplify things a
little bit by noticing that the field inside the sphere is uniform and in the z
direction. The force on a surface element at angles and has the direction
of I B, so it lies parallel to the xy plane and points radially outward (that
is, there is no component in the z direction). For each surface element at a
given there is an equivalent element at angle , so the forces on these
two elements cancel. Thus the interior field exerts no net force in the z
direction. The total force will therefore be half that exerted by the exterior
field.
For the exterior field, we have

/2 2
(3) Fz = [I B]z (R sin ) d d
0 0
/2
2
0 R4 2 2 sin sin2 r +2 sin cos

(4) = z
d
3 0
1
MAGNETIC FORCE OF ATTRACTION BETWEEN HEMISPHERES 2

The extra factor R sin in the first line comes from calculating the path
length of the strip at angle . To get the z component of the vector in the
integrand, we take the dot product with z = cos r sin and we get
/2
2 4 2 2
(5) Fz = 0 R 3 sin3 cos d
3 0
2
(6) = 0 R4 2 2
4
To get the total force, we divide this by 2 as described above, and get, for
the northern hemisphere:


(7) F = 0 R4 2 2 z
4
That is, the net force is attractive, since it points downward.
P INGBACKS
Pingback: Maxwell stress tensor
MAGNETIC MONOPOLE: FORCE ON A CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.43.
Well have a look at a bit of fantasy physics, in the sense that it deals with
something (magnetic monopoles) that are believed not to exist. Suppose
they did exist, and that a magnetic charge qm produced a magnetic field
according to

0 qm
(1) B= r
4r2

where r is the vector position from the magnetic monopole. Then a par-
ticle of mass m with an ordinary charge q moving with velocity v would
experience a force:

(2) F = qv B
dv 0 qqm
(3) = v r
dt 4mr2
Thus the acceleration is perpendicular to v which means that the magni-
tude |v| is a constant. The argument is the same as that used for a centripetal
force:

dv2 d (v v)
(4) =
dt dt
dv
(5) = 2v
dt
(6) = 0
Thus v2 and hence v is a constant.
We now introduce a rather cryptic quantity

0 qqm
(7) Q mr v r
4
This too is a constant, as we can see by taking its time derivative:
1
MAGNETIC MONOPOLE: FORCE ON A CHARGE 2

dQ dv 0 qqm d r
(8) = mv v + mr
dt dt 4 dt
The first term is zero, and we can substitute from above for dv dt in the
second. The third term is a bit special though. The derivative of r is not
zero, since the direction of the radius vector changes as the particle moves.
We can work it out as follows:

d r d r
(9) =
dt dt r
v 1 dr
(10) = r
r r2 dt
v 1 dr
(11) = r
r r dt
The trick here is to notice that dr
dt is not in general equal to v. This is
dr
because dt measures the rate of change of distance from the origin, so if the
particle moves, for example, in a circular orbit, dr
dt = 0 while v 6= 0. So what
is it? It measures the rate of change of radial position, in other words, the
component of velocity parallel to r, which is given by dr dt = v r. Plugging
all this in, we get
 
dQ 0 qqm 1 1
(12) = r (v r) (v (v r) r)
dt 4 r2 r
0 qqm
(13) = [r (v r) (v (v r) r)]
4r
0 qqm
(14) = [v (r r) r (v r) (v (v r) r)]
4r
(15) =0

where we used the vector identity for a triple product in line 3 to expand the
first term.
Since Q is constant, we can introduce spherical coordinates with the z
axis along the Q direction. By taking dot products of Q with each of the
unit vectors, we can derive some useful properties.

(16) Q = m (r v)

(17) = mrv r
(18) = mrv
MAGNETIC MONOPOLE: FORCE ON A CHARGE 3

where weve used another vector identity in line 2.


However, since Q is along the vertical axis and is always horizontal,
Q = 0. Therefore v = 0 as well. To see what this means, we can
write out v as the derivative of the linear path of the particle. In spherical
coordinates, a line element is

(19) dl = drr + rd + r sin d


The velocity is

dl dr d d
(20) v= = r + r + r sin
dt dt dt dt
Therefore

d
(21) v = r =0
dt

which means that is a constant.


Next, consider Q r = |Q| cos (since Q is along the z axis). From the
definition of Q we get

0 qqm
(22) |Q| cos =
4
0 qqm
(23) |Q| =
4 cos
Notice that this formula requires the sign of cos to be opposite to that
of qqm , since |Q| > 0. This means that Q may point in the z direction so
that this formula works out with the right sign. We dont actually use the
direction of Q (that is, whether it points up or down; we do, of course, use
the fact that it is parallel to the z axis) in any of the calculations, so this isnt
particularly important, but its worth noting.
Finally, we consider

(24) Q = mr (r v)

(25) = mrv r
(26) = mrv
Using 20 we get
MAGNETIC MONOPOLE: FORCE ON A CHARGE 4

d
(27) mrv = mr2 sin
dt
Since the angle between the z axis and is + 2 , we have
(28) Q = |Q| sin
0 qqm sin
(29) =
4 cos
Since sin 0 for 0 , the same sign relation between cos and
qqm holds here as well.
Therefore we have

d 0 qqm sin
(30) mr2 sin =
dt 4 cos
d 0 qqm
(31) =
dt 4mr2 cos
k
(32) = 2
r
0 qqm
(33) k
4m cos
We can eliminate the time by calculating (since d
dt = 0)

(34) v2 = r2 sin2 2 + r2
2  2
2 2 k dr
(35) = r sin 4 +
r d
2  2 2
2 2 k dr k
(36) = r sin 4 +
r d r4
r
dr v2 r4
(37) = r2 sin2
d k2
Since v and are constants, this is a differential equation in r and only.
We can integrate it to get
!
1 k sin
(38) arctan p = + 0
sin v2 r2 k2 sin2
k sin
(39) r=
v sin ([ + 0 ] sin )
SOLENOID FIELD FROM BIOT-SAVART LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.44.
Weve used Ampres law to find the field inside and outside a solenoid.
The fundamental formula for finding the magnetic field due to a current is
the Biot-Savart law, so it should be possible to work out the solenoid field
from that as well. Since we treat the current in a solenoid as cylindrical
surface current, the form of the Biot-Savart law to use is

0 K (r r0 )
(1) B= da0
4 |r r0 |3
where K is the surface current density. For a solenoid with n turns per unit
length carrying a current I, K = nI.
Although the natural coordinates to use are cylindrical, I find it easier to
set the problem up in rectangular coordinates and then convert to cylindrical
later on. To define the problem, put the axis of the solenoid on the z axis
and place the observation point r on the x axis. The source point r0 lies on
the solenoid. If the radius of the solenoid is R, then

(2) r0 = xx + yy + zz
(3) x2 + y2 = R2
(4) r = rx
(5) r r0 = (r x) x yy zz
Since K points around the circumference of the solenoid,

(6) K = nI
(7) = nI ( sin x + cos y)
0

(8) K r r = nI [z cos x z sin y + (y sin (r x) cos ) z]
From the symmetry of the setup, if we replace z by z in the source point
r0 , K remains unchanged, but the x and y components of K(r r0 ) change
sign. Thus these components cancel out and the net field must lie in the z
direction, so we can restrict our attention to that from now on.
1
SOLENOID FIELD FROM BIOT-SAVART LAW 2

We can now make the conversion to cylindrical coordinates using

(9) y = R sin
(10) x = R cos
(11) da0 = Rd dz
We get
2 R sin2 + cos2 r cos

0 nI
(12) B= z i3/2 Rdzd
4 0
h
2
(r x) + y2 + z2
2
0 nI R2 rR cos
(13) = z 3/2
dzd
4 0 [R2 + r 2 2rR cos + z2 ]
At this point its important to do the integrals in the right order. Attempt-
ing to do the integral first leads to a mess containing elliptic functions. If
we do the z integral first, we get
2
20 nI R2 rR cos
(14) B= z d
4 0 R2 + r2 2rR cos
Using software, this integral comes out to
(
0 nI z r < R
(15) B=
0 r>R
This reproduces, after a lot of effort, the result obtained from Ampres
law.
MAGNETIC FIELD OF A SEMI-CIRCULAR CURRENT LOOP

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.45.
Another example of finding the magnetic field due to a current using the
Biot-Savart law.
The system this time consists of a semi-circular wire of radius R in the xy
plane, extending from = to = 0 and carrying current I in a counter-
clockwise direction. The problem is to find the magnetic field on a point on
the upper semi-circle of radius R extending from = 0 to = . (There is
no wire or current on this upper semi-circle; were just interested in finding
the field there.) Let the angle of the observation point be = .
The Biot-Savart law for a line current is


0 I dl0 (r r0 )
(1) B=
4 |r r0 |3

Since dl0 lies on the semi-circular wire

(2) dl0 = Rd ( sin x + cos y)

For a given observation point r and source point r0


1
MAGNETIC FIELD OF A SEMI-CIRCULAR CURRENT LOOP 2

(3) r = xx + yy
(4) = R cos x + R sin y
(5) r0 = x0 x + y0 y
(6) = R cos x + R sin y
r r0 = x x0 x + y y0 y
 
(7)
(8) = R (cos cos ) x + R (sin sin ) y
q
r r0 = R (cos cos )2 + (sin sin )2

(9)
p
(10) = R 2 (1 sin sin cos cos )
p
(11) = 2R 1 cos ( )
(12)
dl0 r r0 = R2 d [sin (sin sin ) + cos (cos cos )] z


(13) = R2 d [1 sin sin cos cos ] z


(14) = R2 d [1 cos ( )] z
We can now write out the integral:


0 IR2 [1 cos ( )] d
(15) B = 3/2
z
42 R 3
(1 cos ( ))3/2

0 I d
(16) = 3/2 z p
2 4R 1 cos ( )
Using software, this integral comes out to

1 cos2 12 ( ) tanh1 cos 12 ( )


   
0 I
(17) B = 3/2 z 2 q
2 4R 1 cos2 12 ( ) sin 12 ( )
   


  
0 I 1 1
(18) z 2 tanh cos ( )
23/2 4R 2
" 1 #
0 I 1 + cos 2 ( )
(19) = z ln
1 cos 12 ( )
 
16R

where weve used the identity tanh1 x = 12 ln 1x


1+x

.
Up to now, Ive purposely left off the limits of integration. If we use the
limits as stated in the question above, we get
MAGNETIC FIELD OF A SEMI-CIRCULAR CURRENT LOOP 3

"  # 0
1 + cos 12 ( )

0 I
(20) B= z ln
1 cos 12 ( )
 
16R

"  1 #

0 I 1 + cos 2 1 cos 2 ( + )
(21) = z ln
1 cos 2 1 + cos 12 ( + )
  
16R
" #
0 I tan2 14 ( + )
(22) = z ln 
16R tan2 4
" #
0 I tan 1 ( + )
4
(23) = z ln

8R
tan 4

where weve used the trig identities tan 2x = 1cos x sin x


sin x = 1+cos x .
This matches the answer given in the question in Griffiths (where his
is my ). However, if we choose the limits to be = to = 2, we get
" #
1 cos 2 1 cos 12 ( )
 
0 I
(24) B = z ln
1 + cos 2 1 + cos 12 ( )
  
16R
     
0 I 1
(25) = z ln tan
tan ( )
8R 4 4
Since tan x 2 = tan1 x this comes out to


" #
0 I tan 1 ( + )
4
(26) B= z ln

8R

tan 4
That is, by merely changing the limits of integration by adding 2, weve
reversed the direction of B. This clearly doesnt make sense, since the inte-
grand 1 is identical under this change, so it seems that we need
1cos( )
to take the absolute value of the answer. We can see how this errant sign
crept in by looking back at the original integral solution above. There, we
cancelled off the cos and sin terms that multiplied the arctanh, but the square
root term could be positive or negative (something which the software didnt
tell us), so presumably we need to select the right sign in order for the direc-
tion of the field to come out right. The correct direction can be determined
from the right hand rule and for the direction of current assumed, B must
point in the +z direction.
HELMHOLTZ COIL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.46.
The magnetic field due to a circular current loop of radius R and current
I a distance z above the centre of the loop is (Griffiths example 5.6):

0 IR2 1
(1) B=
2 [R2 + z2 ]3/2
If we take two such current loops a distance d apart and set z = 0 to be
halfway between them, then the net field on the axis is, by the principle of
superposition:

0 IR2 1 1
(2) B= +

2
h
2
i3/2 h 2
i3/2
R2 + z d2 R2 + z + d2
 

The derivative of B is

B 0 IR2 3 2zd 3 2z+d
(3) = 

z 2 2 2
5/2 2 
2
5/2
R2 + z 21 d R2 + z + 12 d
 

The derivative is always zero at z = 0.


Now the second derivative is

2 2 B 15 (2 z d)2 3
(4) 2 2
= 7/2
 2 5/2
0 IR z 4 
2 1
 2

2 1
R + z 2 d R + z 2 d
15 (2 z + d)2 3
+ 7/2
 2 5/2
4 
2 1
 2

2 1
R + z+ 2 d R + z+ 2 d

At z = 0 this becomes
1
HELMHOLTZ COIL 2

!
2 B 0 IR2 15 d2 6
(5) =
z2 2 2 R2 + 1 d 2 7/2
R2 + 1 d 2 5/2

4 4
We can make this derivative zero as well if we choose

(6) d=R
Under this condition, the field at the centre is

8I0
(7) B (0) =
5 5R
Such a pair of circular loops is known as a Helmholtz coil and is used to
produce a localized region of a relatively constant magnetic field. In fact,
the third derivative also turns out to be zero at z = 0, so the first non-zero
derivative is the fourth derivative which (if youre interested) comes out to

4B 27648 5 0 I
(8) (z = 0) =
z4 3125 R5
MAGNETIC FIELDS OF SPINNING DISK AND SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.47.
Another application of the magnetic field due to a circular current loop
of radius R and current I a distance z above the centre of the loop (as given
by Griffiths example 5.6):

0 IR2 1
(1) B=
2 [R2 + z2 ]3/2

We start with a spinning disk with surface charge density . We can treat
this as a collection of concentric current loops, with the current at radius r
given by

(2) dI (r) = rdr

where is the angular velocity. The field of the spinning disk is then

R
0 r3
(3) B (z) = 3/2
dr
2 (r2 + z2 )
0

0 2z2 + R2
 
(4) = 2z
2 R2 + z2

Now we can work out the field due to a spinning solid sphere of charge
density as a collection of spinning disks. We fix our observation point
z, and place the origin at the centre of the sphere. A given disk within the
sphere has a radius given by R sin for a particular value of . The distance
of this disk from the observation point is z R cos and the charge density
in the disk is Rd . It might seem that the way to proceed is to replace
Rd R R sin and z z R cos in the formula for the spinning
disk and then integrate over from 0 to . That is, we try:
1
MAGNETIC FIELDS OF SPINNING DISK AND SPHERE 2

(5)


2 2
0 R 2 (z R cos ) + (R sin )
B (z) = q 2 (z R cos ) d
2 0 2 2
(R sin ) + (z R cos )

Even after multiplying out the various squares and trying to simplify this
expression, plugging it into software gives a complex sum of elliptic func-
tions. It turns out that a better approach is to use an integral over the z
coordinate of the disks. Let be the z coordinate of a given disk. Then
the distancepfrom the disk to the observation point is z , the radius of
the disk is R2 2 and the charge density is d . Now the appropriate
substitutions
p into the formula for the disks magnetic field are z z ,
R R 2 and d , and the integral is over between R and
2
R:



R 2
0 2 (z ) + R2 2
(6) B (z) = q 2 (z ) d
2 R 2
R2 2 + (z )
" #
0 R 2z2 + R2 4z + 2
(7) = p 2 (z ) d
2 R R2 + z2 2z
2 0 R5
(8) =
15 z3
Software was used for the integral, but the first term in the integrand can
be split into 3 (one for each term in the numerator) and either integrated
directly or by parts, so its not too hard even by hand. Note that weve
assumed that z > R. If we take 0 z R then we get contributions to the
field from two different regions: 0 z and z R. The field B1 from
the inner region can be obtained from 7 by setting R = z and integrating from
z to z

z
" #
0 3z2 4z + 2
(9) B1 = p 2 (z ) d
2 z 2z2 2z
2
(10) = 0 z2
15
Each spherical shell of radius in the outer region contributes a uniform
field of magnitude (as worked out in Griffithss example 5.11):
MAGNETIC FIELDS OF SPINNING DISK AND SPHERE 3

2
(11) Bshell = 0
3

ere is the surface charge density on the shell. A shell from our solid
sphere of charge has a charge density of = d so the field due to the
outer region is
R
2
(12) B2 = 0 d
3 z
1
0 R2 z2

(13) =
3
The total field for z < R is then

(14) Bz<R = B1 + B2
2 1
0 z2 + 0 R2 z2

(15) =
15 3
 2
z2

R
(16) = 0
3 5
This agrees with our earlier calculation of the field inside a rotating sphere
    
0 2 3 2 6 2 2
(17) B= R0 r cos r + r R0 sin
3 5 5
To get the field along the z axis, we set = 0 and r = z, so we get the
same formula.
MAGNETIC FIELD OF CURRENT LOOP - OFF AXIS FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.48.
The magnetic field due to a circular current loop of radius R and current
I a distance z above the centre of the loop is (as given by Griffiths example
5.6):

0 IR2 1
(1) B=
2 [R2 + z2 ]3/2

The derivation of this was made easier by restricting our attention to


points on the z axis. In principle, its easy enough to find the field at any
other point, but the integrals prove somewhat problematical. Suppose we
take our observation point r to be in the yz plane, so that

(2) r = yy + zz

The loop lies in the xy plane, so

(3) r0 = x0 x + y0 y
(4) = R cos 0 x + R sin 0 y
0
R cos 0 x + y R sin 0 y + zz

(5) rr =
q
r r0 = (R cos 0 )2 + (y R sin 0 )2 + z2

(6)
q
(7) = R2 + r2 2yR sin 0

The line element is


1
MAGNETIC FIELD OF CURRENT LOOP - OFF AXIS FIELD 2

(8)
dl0 = sin 0 x + cos 0 y Rd 0
 

(9)
n h  i o
0 0 0 0 0 0 0 2
z Rd 0
 
dl r r = z cos x + z sin y + R sin y sin + R cos
(10)
= z cos 0 x + z sin 0 y + R y sin 0 z Rd 0
  

The field is then, using the Biot-Savart formula:



0 I 2
dl0 (r r0 )
(11) B=
4 0 |r r0 |3

0 IR 2
[z cos 0 x + z sin 0 y + (R y sin 0 ) z] d 0
(12) = 3/2
4 0 (R2 + r2 2yR sin 0 )
The x component integrates easily to
2
0 Iz 1
(13) Bx =

p
4y R2 + r2 2yR sin 0 0

(14) = 0
The fact that Bx = 0 isnt terribly surprising, since the symmetry of the
problem makes that fairly obvious without doing any calculations.
The other two integrals are non-trivial, however. Plugging them into soft-
ware we get a blizzard of elliptic functions with no obvious way of simpli-
fying the result. We can check that these integrals reduce to the correct
form when the observation point is on the z axis (that is, when y = 0). In
that case we get
2
0 IR z
(15) By = sin 0 d 0 = 0
4 (R2 + z2 )3/2 0
2
0 I R2 0 0 IR2 1
(16) Bz = 3/2
d =
4 (R2 + z2 ) 0 2 [R2 + z2 ]3/2
FORCE BETWEEN CURRENT LOOPS: NEWTONS THIRD
LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.49.
We can write the force on one current loop due to a second current loop
in a symmetric form by combining the Biot-Savart law and Lorentz force
law. The force on loop 1 due to loop 2 is

(1) F1 = I1 dl1 B2

The magnetic field produced by loop 2 at a point r1 on loop 1 is



0 I2 dl2 (r1 r2 )
(2) B2 =
4 |r1 r2 |3
If we define r r1 r2 , we can substitute this field into the force integral
to get

0 I1 I2 dl1 [dl2 r]
(3) F1 =
4 r3

0 I1 I2 dl2 (dl1 r) r (dl1 dl2 )
(4) =
4 r3

0 I1 I2 dl2 (dl1 r) r (dl1 dl2 )
(5) =
4 r2

where weve used a vector identity in line 2.


Consider now the term dl1 r. The dl1 is an increment along loop 1 while
r2 is held constant. Thus dl1 is the change in r as we move a bit along loop
1. The dot product dl1 r is the component of this change that is parallel to
r; that is, it is the change in the magnitude (as opposed to the direction) of
r. We can therefore write

(6) dl1 r = dr
The first term therefore becomes
1
FORCE BETWEEN CURRENT LOOPS: NEWTONS THIRD LAW 2


0 I1 I2 dr
(7) dl2
4 r2

where weve taken dl2 outside the first integral since the point on loop 2 is
held constant in the inner integral. The inner integral is zero, since we can
split the closed loop into 2 parts, one from point a to point b and the other
from b back to a by carrying along in the same direction around the loop.
The first integral is
b
dr 1 1
(8) 2
=
a r r (a) r (b)
That is, it depends only on the value of r at the endpoints and not on the
path taken between them. The integral on the reverse path is therefore just
the negative of the first integral, so the integral around the closed loop is
zero. We therefore get for the force

0 I1 I2 r (dl1 dl2 )
(9) F1 =
4 r2
If we interchange the loops to find the force F2 on loop 2 due to loop 1,
everything is the same except that r becomes r, so F2 = F1 which is an
example of Newtons third law (equal and opposite forces).
MAGNETIC VECTOR POTENTIAL FROM MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.51.
The magnetic vector potential can be used to find the field through the
relation B = A, but there is no common formula for calculating the
potential from the field. The potential is usually calculated from the current
density via the formula

0 J (r0 ) 3 0
(1) A (r) = d r
4 V |r r0 |
One attempt at a formula for deriving A from B is
r
(2) A (r) = B dl
O

where the integral is a line integral over some path from the origin (which
must be defined independently) to the observation point r. For a constant
field, this gives

(3) A = Br
(4) = r B
This differs from the correct value by a factor of 12 , but there is a more
serious problem in that the line integral isnt always independent of the
path. To see this, consider the case of an infinite wire with a current I. The
magnetic field circles the wire so in cylindrical coordinates

0 I
(5) B=
2r
If we integrate B around a closed square loop of side 1 that is perpendic-
ular to the field, the contributions from the two edges perpendicular to the
wire will cancel so we get
 
0 I 1 1
(6) B dl =
2 a a+1
1
MAGNETIC VECTOR POTENTIAL FROM MAGNETIC FIELD 2

where a is the distance of the near edge from the wire. This is clearly not
zero, which it would have to be for the integral to be path independent.
Another attempt at a formula is

1
(7) A (r) = r B ( r) d
0

(This formula is stated in Griffithss problem.)


For a constant field, this gives

1
(8) A = r B d
0
1
(9) = rB
2

which is correct.
In the case of the infinite wire, we have

0 I
(10) B ( r) =
2 ( s)

where s is the perpendicular distance from the wire to the point r. Therefore

1 1
0 I
(11) B ( r) d = d
0 2s 0
0 I
(12) =
2s
In cylindrical coordinates, the position vector r = ss + zz so

0 I
(13) A = (ss + zz)
2s
0 I
(14) = (zs sz)
2s
In this case, A = 0 and A = B as can be checked by direct calcula-
tion, so this is a valid potential. However, in general, a potential calculated
this way does not satisfy A = 0. We have
MAGNETIC VECTOR POTENTIAL FROM MAGNETIC FIELD 3

 1 
(15) A = r B ( r) d
0
1 1
(16) = r B ( r) d B ( r) rd
0 0
1
(17) = r 2 0 J ( r) d 0
0

where weve used Ampres law and the chain rule in the first term, and
the fact that r = 0 in the second. Thus the only way this can be zero
for all observation points r is for the integral to be zero, which effectively
means that there can be no current. So this formula for the potential is not
in general divergence-free.
MAGNETIC SCALAR POTENTIAL: DIPOLE AND ROTATING
SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.52.
We can try finding a magnetic scalar potential for a few examples. First,
lets try a pure magnetic dipole. The field of a dipole is

0 m 
(1) B= 2 cos r + sin
4r3

where m is the dipole moment. We can find the scalar potential by solving
B = U for U. In spherical coordinates

U 1 U 1 U
(2) U =
r + +
r r r sin
Equating components we find that

U 0 m
(3) = 2 cos
r 4r3
0 m
(4) U = cos + f ( , )
4r2
Checking this with the term, we find that

U 0 m
(5) = sin
4r2

so f ( , ) = 0 and

0 m
(6) U= cos
4r2
As another example, we revisit the rotating spherical shell, which had the
field
( 2
0 R

3 cos r sin r<R
(7) B= 0 R4 1 
3 r 3 2 cos r + sin r>R
1
MAGNETIC SCALAR POTENTIAL: DIPOLE AND ROTATING SPHERE 2

The r > R region is identical to the dipole, so in that region:

0 R4
(8) U= cos
3r2
For r < R, we can follow the same procedure as above and find that

20 R
(9) U = r cos
3
Finally, we can try to find a scalar potential for the solid rotating sphere.
The field inside the sphere is

3 r2
   2  
0 Q 6r
(10) B= 1 cos r + 1 sin
4R0 5 R20 5 R20
Using the r component as above, we find

1 r3
 
0 Q
(11) U = r cos + f ( , )
4R0 5 R20

where f ( , ) is a constant (relative to r) of integration. From this, the


component of the gradient must be

1 r2
 
1 U 0 Q 1f
(12) = 2
1 sin +
r 4R0 5 R0 r
Since f cannot depend on r, there is no way to make this agree with the
term in B above.
MAGNETIC VECTOR POTENTIAL AS THE CURL OF
ANOTHER FUNCTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.53.
The fact that B is divergenceless allows us to express it as the curl of a
vector potential: B = A. By assuming A = 0 as well, can express it
in turn as the curl of another function: A = W. Although this is rarely
useful, we can try to work out W in a few cases.
First, we can express W in terms of B by noting

(1) B = A
(2) = ( W)
(3) = ( W) 2 W
If we assume that W is divergenceless as well, then we get

(4) 2 W = B
This is the same form as the equation for the vector potential in terms of
the current density:

(5) 2 A = 0 J
This could be written in integral form as

0 J (r0 ) 3 0
(6) A (r) = d r
4 V |r r0 |
Therefore we can write

1 B (r0 ) 3 0
(7) W (r) = d r
4 V |r r0 |
Its important to note that this equation relies on B 0 at infinite dis-
tance, just as the equation for A relied on the current density being finite in
extent.
1
MAGNETIC VECTOR POTENTIAL AS THE CURL OF ANOTHER FUNCTION 2

We can now find W in a couple of special cases. First, consider the case
of a constant field. We might be tempted to try calculating W by simply
taking B outside the integral above, but this integral formula doesnt apply
in this case since B extends to infinity. We can, however, use the result we
got earlier for the vector potential for a constant field.

1
(8) A = rB
2
We need to solve the equation

1
(9) W = rB
2
One way of approaching this is to split the vector equation into its com-
ponents. For the x component we have

Wz Wy 1
(10) = (yBz zBy )
y z 2
We can try a solution of the form

Wz 1
(11) = yBz
y 2
1
(12) Wz = y2 Bz + f (x, z)
4
Wy 1
(13) = zBy
z 2
1
(14) Wy = z2 By + g (x, y)
4

where f and g are functions of integration.


Working out the z component, we get

1
(15) Wy = x2 By + h (y, z)
4
Comparing the two equations for Wy we get

1
Wy = By x2 + z2

(16)
4
The other two components can be worked out the same way and we get
MAGNETIC VECTOR POTENTIAL AS THE CURL OF ANOTHER FUNCTION 3

1
Bx y2 + z2 x + By x2 + z2 y + Bz x2 + y2 z
   
(17) W=
4
By direct calculation we can check that W = 0 so this is a valid solu-
tion. (There may be some fancy way of expressing this entirely in terms of
vectors and their products, but if so, it eluded me.) The solution is unlikely
to be unique.
For a second example, we can return to the infinite solenoid. Griffiths
shows in his Example 5.12 that the vector potential inside a solenoid with n
turns per unit length and of radius R is

(
0 nIr
2 r<R
(18) A= 0 nIR2
2r r>R
Again, we seek components of W that satisfy the equation A = W.
Outside the solenoid, we have, using the equations for the curl in cylindrical
coordinates:

1 Wz W
(19) = 0
r z
Wr Wz 0 nIR2
(20) =
" z r# 2r

1 rW Wr
(21) = 0
r r

We can try the solution

0 nIR2 z
(22) Wr =
2r
(23) W = 0
(24) Wz = 0
This satisfies all three curl equations, and also satisfies

1 (rWr )
(25) W = +0+0
r r
(26) = 0
Inside the solenoid, we have
MAGNETIC VECTOR POTENTIAL AS THE CURL OF ANOTHER FUNCTION 4

1 Wz W
(27) = 0
r z
Wr Wz 0 nIr
(28) =
" z r# 2

1 rW Wr
(29) = 0
r r
If we try a similar solution, we have

0 nIrz
(30) Wr =
2
(31) W = 0
(32) Wz = 0
This satisfies all of the curl equations but unfortunately the divergence
isnt zero:

1 (rWr )
(33) W = +0+0
r r
(34) = 0 nIz
We can fix this by adding in a z component, so we get

0 nIrz
(35) Wr =
2
(36) W = 0
0 nIz2
(37) Wz =
2

This doesnt affect the curl, and makes the divergence zero. Again, this
solution is unlikely to be unique.
MAGNETIC FIELD: UNIQUENESS CONDITIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.54.
Weve seen that solutions to Laplaces and Poissons equations are unique,
so its natural to ask if a similar condition exists in the case of the magnetic
field and its vector potential. That is, suppose we specify the field B or the
potential A on the boundary of some volume, and also specify the current
density J within that volume. Does this determine the field uniquely within
the volume?
First, we need an identity which can be derived from the divergence the-
orem and a vector calculus identity. For some vector fields U and V, we
have

(1) U ( V) da = [U ( V)] d 3 r
A
V
(2) = [( U) ( V) U ( ( V))] d 3 r
V

Following the same logic as in the electrostatic case, we suppose that


there are two different solutions B1 and B2 (with corresponding potentials
A1 and A2 ) and consider the difference B3 B2 B1 . Because the curl
operator is linear, we have

(3) A3 = (A2 A1 )
(4) = B2 B1
(5) = B3
Setting U = V = A3 in the identity above, we get on the LHS

(6) U ( V) da = (A3 B3 ) da
A A

Note that if we specify either A or B on the boundary surface, then either


A3 = 0 or B3 = 0 since there can be either a unique potential or a unique
field on the boundary. Thus this surface integral is always zero.
1
MAGNETIC FIELD: UNIQUENESS CONDITIONS 2

Its important to note here that specifying A on the boundary only does
not specify B on the boundary as well, since we need to know A in three
dimensions in order to calculate its curl.
On the RHS, we get

(7) 0= [B3 B3 A3 J3 ] d 3 r
V
Since weve specified the current inside the volume J3 = J2 J1 = 0 so
we get

(8) B23 d 3 r = 0
V
Since the integral of a positive definite quantity over the volume is zero,
it must be zero everywhere, so B3 = 0 and B2 = B1 .
MAGNETIC DIPOLE IN A CONSTANT FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.55.
A pure magnetic dipole has a curious property when it is placed in a
uniform magnetic field: there is a sphere surrounding the dipole which is not
crossed by any magnetic field lines. To see this, we start with the equation
for the field of a dipole with dipole moment m:

0
(1) Bdip = [3 (m r) r m]
4r3
Well align the dipole so that it points along the negative z axis: m =
m0 z and apply the magnetic field also along the z axis: Bext = B0 z. Then
the total field is

(2) B = Bdip + Bext


0
(3) = [3m0 cos r + m0 z] + B0 z
4r3
If there is a sphere which is not crossed by field lines, then B must be
tangent to the sphere everywhere on that sphere. In other words, B is per-
pendicular to the surface normal vector, which for a sphere is just r. So we
look for a solution of B r = 0. We get

0
(4) B r = [3m0 cos + m0 cos ] + B0 cos = 0
4r3
m0 0
(5) = B0
2r3
m0 0 1/3
 
(6) r=
2B0
Presumably the field lines inside this sphere resemble those of the pure
dipole; that is, they emerge outwards from the north pole of the dipole and
loop round to converge on the south pole. Outside the sphere the field lines
far from the sphere are straight lines, while closer to the sphere they diverge
around the sphere, just like water flowing around an obstacle in a stream.

1
GYROMAGNETIC RATIO

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.56.
The gyromagnetic ratio is the ratio of an objects magnetic dipole mo-
ment to its angular momentum. The numerical value of depends on the
units being used (that is, its not a dimensionless quantity).
For a circular wire loop of radius r carrying charge Q and having mass
M we can calculate as follows. Suppose the loop is rotating with angular
Q Q
speed . Then the current in the loop is I = 2r r = 2 . The magnetic
dipole moment is therefore

Q Q
(1) m = Ia = r2 = r2
2 2
2
The angular momentum is L = mr , so the gyromagnetic ratio of a
circular wire loop is

m Q
(2) = =
L 2M
The ratio is independent of r and , which might seem a bit bizarre since
we can calculate even when the loop isnt spinning, and therefore there is
no current or angular momentum. In that case, should be interpreted as a
limit, since otherwise it would involve dividing zero by zero.
Since the ratio is independent of r, it applies to any circular loop so we
can apply this to any solid of revolution, such as a sphere or cylinder. All
such objects have the same formula for .
We can attempt to find for an elementary particle such as an electron,
although the resulting answer isnt correct (as you might expect when ap-
plying classical physics to quantum objects). The electron has spin 21 which
means its angular momentum is 2h . We then get

2me e
(3) e = =
h 2Me

where me is the dipole moment of the electron, Me is its mass and e is the
elementary charge. Plugging in the numbers we get (in SI units):
1
GYROMAGNETIC RATIO 2

1.602 1019
(4) e = = 8.792 1010 C kg1
2 9.11 1031
From this we can get the magnetic dipole moment of the electron

h
(5) me = e = 4.636 1024 Amp m2
2
Experimentally, the value is me = 9.285 1024 Amp m2 which is al-
most exactly twice the calculated value. When the moment is calculated
using quantum electrodynamics, the agreement is pretty much exact.
P INGBACKS
Pingback: Magnetic potential and field of a rotating sphere of charge
Pingback: Magnetic moment in oscillating magnetic field
Pingback: Wigner-Eckart Theorem - adding orbital and spin angular mo-
menta
AVERAGE MAGNETIC FIELD WITHIN A SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.57.
We worked out the average electric field at the centre of a sphere earlier,
so can we do something similar for the average magnetic field? Suppose we
are given a steady current density within the sphere.
The starting point is the definition of the average field:

3
(1) Bav = Bd 3 r
4R3 V
We can write this in terms of the vector potential and then in terms of the
current density:

3 3 3
(2) Bd r = Ad 3 r
4R2 4R 3
V V

3
(3) = A da
4R3 A

30 J (r)
(4) = dad 3 r
16 R V A |r r0 |
2 3

In the second line we used the vector identity V Ad 3 r = A A da
and in line 3 we used the definition of the vector potential.
We are now faced with a surface integral and a volume integral. We can
do the surface integral first, but we need to be clear about which coordinates
are which. If we do the surface integral first, then we are choosing a partic-
ular volume element d 3 r and holding it constant while we integrate over the
surface. That is, the observation point r points to the volume element and
the source point r0 points to the surface element. This is why weve explic-
itly stated the dependence of J (r) since it depends on the volume element,
and not the surface element.
The only term containing r0 is therefore the |rr 1
0 | factor, so the integral
1
we need to do is A |rr0 | da. Since r is fixed, we can point it along the z
axis, making the angle between r and r0 . Also, since r0 always points to a
surface element, its magnitude is always R. We therefore have
1
AVERAGE MAGNETIC FIELD WITHIN A SPHERE 2

1 1
(5) 0
=
|r r | 2 2
r + R 2rR cos
The surface element da always points radially outward since its on the
surface of a sphere, so by symmetry the integral over the x and y components
of da will cancel out and were left with only the z component, which is
cos |da| = R2 sin cos d d . The integral we must do is thus
2
1 sin cos d d
(6) 0
da = R2 z
A |r r | 0 0 r2 + R2 2rR cos
Using software, we find that

(7)
p
1 r2 + R2 2 rR cos ( ) r2 + R2 + rR cos ( )

sin cos d
=
r2 + R2 2rR cos 3 r2
Evaluating the limits requires specifying whether R is larger or smaller
than r. If were interested in currents within the sphere, then R > r > 0 and
we have
2
2 sin cos d d 4
(8) R z = rz
0 0 r2 + R2 2rR cos 3
Plugging this back into the average field formula above we get

0
(9) Bav = J rd 3 r
4R3 V
We can write this in terms of the magnetic dipole moment as follows:

20
(10) Bav = m
4R3
If we look at currents outside the sphere, then R < r when we evaluate
the limits on the integral above, giving
2
2 sin cos d d 2R3
(11) R z = 2 z
0 0 r2 + R2 2rR cos 3r
Plugging this into the average field formula we get

0 Jr 3
(12) Bav = d r
4 V r3
AVERAGE MAGNETIC FIELD WITHIN A SPHERE 3

Remember that r points from the centre to the volume element, so it is


the negative of the vector rc from the volume element to the centre. In this
form we get

0 J rc 3
(13) Bav = d r
4 V rc3
This is just the volume form of the Biot-Savart law for calculating the
field at the centre of the sphere thus for currents outside the sphere, the
average of their field over the sphere is equal to the field produced at the
centre of the sphere.
P INGBACKS
Pingback: Magnetic potential and field of a rotating sphere of charge
Pingback: Magnetic dipole: average field over a sphere
Pingback: Magnetization: microscopic versus macroscopic
Pingback: Magnetostatic formulas from electrostatic formulas
MAGNETIC POTENTIAL AND FIELD OF A ROTATING
SPHERE OF CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.58.
We return to the rotating solid sphere of charge. If the sphere has radius
R, contains total charge Q and is rotating with angular speed we can find
its magnetic moment from the gyromagnetic ratio. The angular momentum
of a sphere is L = 25 MR2 , so

m Q
(1) = =
L 2M
LQ
(2) m =
2M
1 2
(3) = QR
5
We can align the sphere so its magnetic moment is along the z axis. From
this we can find the average field within the sphere:

20
(4) Bav = m
4R3
0 2Q
(5) = z
4 5R
We worked out the exact field earlier and found

3 r2
   2  
0 Q 6r
(6) B= 1 cos r + 1 sin
4R 5 R2 5 R2
3 r2 6 r2
 
0 Q
(7) = z cos r + sin
4R 5 R2 5 R2
We can calculate the average field within the sphere from this formula by
finding

3
(8) Bav = Bd 3 r
4R3 V
1
MAGNETIC POTENTIAL AND FIELD OF A ROTATING SPHERE OF CHARGE 2

The average field has three contributions. The z contribution is constant,


so is just

0 Q
(9) Bav;z = z
4R
The r contribution, by symmetry, has a non-zero component only in the
z direction, and since the angle between r and z is , this contribution will
2
0 Q 3 r 2
be Br = 4R 5 R2 cos z. If we take the average of this, we have


3
(10) Bav;r = Br d 3 r
4R3 V
R 2
0 Q 3 3 2 2
 2 
(11) = z r cos r sin d d dr
4R3 5 4R3 0 0 0
3 0 Q
(12) = z
25 4R

For the contribution, again the non-zero component is in thez direction.


This time, the angle between and z is + 2 and cos + 2 = sin ,
2
0 Q 6 r 2
so the contribution is B = 4R 5 R2 sin z The average is


3
(13) Bav; = B d 3 r
4R3 V
R 2
0 Q 6 3 2 2
 2 
(14) = z r sin r sin d d dr
4R3 5 4R3 0 0 0
12 0 Q
(15) = z
25 4R
The total average field is then

(16) Bav = Bav;z + Bav;r + Bav;


 
3 12 0 Q
(17) = 1 z
25 25 4R
2 0 Q
(18) = z
5 4R
This agrees with the calculation above using the magnetic moment.
The dipole term in the vector potential is
MAGNETIC POTENTIAL AND FIELD OF A ROTATING SPHERE OF CHARGE 3

0
(19) A1 = m r
4r2
0 R2 Q
(20) = sin
4r2 5

so this is what wed expect the potential to be for large distances from the
sphere. However, weve seen that for a spherical shell, the potential outside
the sphere is exactly equal to that of a perfect dipole. We might expect,
therefore, that for a solid sphere, the dipole potential is also the exact po-
tential. The potential for the shell of radius r0 is

0 r04 sin
(21) A=
3 r2
The surface charge density for an infinitesimal shell of thickness dr0
within the solid sphere is

3Q
(22) = dr0
4R3

so the total potential is (remember were holding the observation point con-
stant, so r and are both constants):
R
0 3Q sin
(23) A = r04 dr0
3 4R3 r2 0
2
0 R Q
(24) = sin
4r2 5
Thus the exact potential is indeed equal to the dipole potential.
P INGBACKS
Pingback: Magnetic dipole: average field over a sphere
MAGNETIC DIPOLE: AVERAGE FIELD OVER A SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.59.
The magnetic field of a perfect dipole pointing the z direction is

0 m 
(1) Bdip = 2 cos r + sin
4r3

We can average this over a sphere using the same logic as we did in
the case of the solid rotating sphere. The average is, by symmetry, in the
z direction, and the r component contributes cos z and the component
contributes sin z, so we get

R 2
0 m 1 2 2
 2 
(2) Bav = z cos sin r sin d d dr
4 0 0 0 r3
R 2
0 m 1
cos2 sin sin3 d d dr

(3) = z
4 0 0 0 r

If we do the integral over first, we find that this integral works out to
zero. Clearly this doesnt agree with the result we obtained earlier for the
average magnetic field inside a sphere due to currents within the sphere:

20
(4) Bav = m
4R3

The problem is actually the same one that we encountered when dealing
with the electric dipole. The integral over r blows up at r = 0 (since ln r
is infinite there); If we integrate over any lower limit > 0 out to R, the
problem goes away and the integral is a well-behaved zero. The solution to
our current problem is the same as in the electric dipole case: introduce a
delta function component to the field. That is we add in a term K 3 (r) for
some constant K in order to make the average field come out to what we
want. We must have
1
MAGNETIC DIPOLE: AVERAGE FIELD OVER A SPHERE 2


3
(5) Bav = 3
K 3 (r) d 3 r
4R V
20 m
(6) =
4R3
2
(7) K = 0 m
3
As in the electric case, its not really quite correct to write

0 m  2
(8) Bdip = 3
2 cos r + sin + 0 m 3 (r) z
4r 3
0 2
(9) = [3 (m r) r m] + 0 m 3 (r) z
4r3 3

since the first term doesnt apply at r = 0 but its a convenient notation as
long as its interpreted with caution.
MAGNETIC DIPOLE MOMENT FOR VOLUME CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.60.
The expansion of the magnetic vector potential was done originally for
line currents, but a more general expression is possible for volume currents.
We start with the relation between the potential and current:

0 J (r0 ) 3 0
(1) A (r) = d r
4 V |r r0 |
We can expand the integrand in terms of Legendre polynomials in the
same way as for the electric potential:

1 1
(2) 0
=
|r r | r + r 2rr0 cos 0
2 02
1 1
(3) = q
r 1 + r 2 r0 cos 0
02
r2 r
 0 n
1 0 r
(4) = Pn (cos )
r n=0 r
Substituting this into the potential, we get

0 1
Pn (cos 0 )r0n J r0 d 3 r0

(5) A= n+1
4 n=0 r V
 
0 1 0
 3 0 1 0 0 0
 3 0
(6) = J r d r + 2 r cos J r d r + . . .
4 r V r V
The first integral can be rewritten by observing

(7) (xJ) = x J + x J
(8) = x J + x J
(9) = Jx + x J
For steady currents, J = 0, so for the x component of the integral we
have
1
MAGNETIC DIPOLE MOMENT FOR VOLUME CURRENT 2


3 0
(10) Jx d r = (xJ) d 3 r0
V
V
(11) = xJ da0
A

If the current is confined to a finite volume, we can take the surface of


integration outside this volume, making the surface integral zero. A sim-
ilar argument applies to the other two components of J, so in general the
monopole term in the expansion is zero.
The dipole term is a bit trickier. We can start with a similar vector identity
to the above, but this time with two coordinates. With x and y we get:

(12) (xyJ) = (xy) J + xy J


(13) = xy J + yx J + xy J
(14) = xy J + yx J + xy J
(15) = xJy + yJx + xy J

Again, taking J = 0 and converting to a surface integral we get


3
(16) (xyJ) d r = (xJy + yJx ) d 3 r
V
V
(17) = xyJ da
A
(18) = 0

3
(19) xJy d r = yJx d 3 r
V v

Choosing the two coordinates to be the same gives


(20) xJx d 3 r = 0
V

Returning to the dipole integral, we have for the x component (using


r0 cos 0
= rr r0 ):
MAGNETIC DIPOLE MOMENT FOR VOLUME CURRENT 3


1 r
(21) r cos Jx d r = 3 r0 Jx d 3 r0
0 0 3 0
r2 r
V
V
1
xx0 + yy0 + zz0 Jx d 3 r0

(22) = 3
r V

1  0
y y Jx x0 Jy + z z0 Jx x0 Jz d 3 r0
 
(23) = 3
2r V
h
1 i
y J r0 z z J r0 y d 3 r0

(24) = 3
2r V
 
1 0 3 0
(25) = 3 r Jr d r
2r V x
The other two components work out the same way. We can make the
dipole term equivalent to the formula for line currents:

0
(26) A1 = m r
4r2

if we define the magnetic dipole moment here as



1
(27) m = J r0 d 3 r0
2 V

1
(28) = r0 Jd 3 r0
2 V
P INGBACKS
Pingback: Average magnetic field within a sphere
MAGNETIC DIPOLE FIELD OF A FINITE SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 5.61.
For our last post on magnetostatics, well consider a finite solenoid of
radius R and length L, with a surface charge density rotating at angular
speed . We know that the field outside an infinite solenoid is zero, but
what about a finite solenoid? For points far from the axis, we can use a
dipole approximation.
We align the axis along the z axis, and consider it to be a stack of indi-
vidual current loops, each with its own dipole moment. The moment of a
current loop is

(1) m = IR2 z

In terms of the parameters of the problem, each loop has a thickness of


dz and thus carries a current of I = Rdz. The contribution of the loop at
coordinate z is therefore

(2) dm = R3 dzz

The field of the dipole from this current loop is

0
(3) dBdip = [3 (dm r) r dm]
4r3

We now need to consider carefully what is meant by r. Well take the


observation point to be a distance s along a line perpendicular to the axis
and intersecting the axis at its midpoint. Well define s to be on the x axis.
Then for a given current loop at coordinate z, the vector r points from the
centre of this loop to s. Therefore
1
MAGNETIC DIPOLE FIELD OF A FINITE SOLENOID 2

p
(4) r = s2 + z2
z
(5) cos =
r
s
(6) sin =
r
(7) dm r = R3 cos dz
z
(8) = R3 dz
r

where is, as usual, the angle between r and z. Therefore

(9) r = cos z + sin x


z s
(10) = z + x
r r
The total dipole field of the solenoid is therefore

(11)
L
" ! #
0 2 3z2 1 sz
Bdip = R3 5/2
3/2
z + 5/2
x dz
4 L2 (s2 + z2 ) (s2 + z2 ) (s2 + z2 )
(12)
0 R3 L
=   3/2 z + 0x
2 L 2
4 s + 2
(13)
0 R3 L
=   3/2 z
2 L 2
4 s + 2
1
Note that as L , the field does tend to zero as L2
which is the correct
value for an infinite solenoid.
TORQUE ON A MAGNETIC DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.1.
As a prelude to looking at magnetic fields in matter, we need to work
out the torque and force on a magnetic dipole, since atoms are composed
of such dipoles in the form of electrons (which have both orbital and spin
magnetic moments).
Consider a small square current loop carrying current I and having side
length a. Place the loop initially in the xy plane with its centre at the origin
and sides aligned with the coordinate axes, and then apply a constant mag-
netic field B in the z direction. Now if we rotate the loop about the x axis
by an angle , what are the net force and torque due to the field B?
Using the Lorentz force law, the force on each side of the square is F =
aI B and, by using the right hand rule, we see that the forces on the two
sides parallel to the y axis point in the x directions and thus cancel, and
the forces on the sides parallel to the x axis point in the y directions and
also cancel, so that the net force on the loop is zero (although the forces all
tend to stretch the loop by pulling it outwards on all four sides).
The torque on the sides parallel to the y axis is zero, but on the other
two sides, the torques do not cancel. To see this, recall that the definition
of the torque is N = r F. For a segment on one of the sides, we can
split r = rk + r where rk is the component of r that is parallel to the side
of the loop and r is perpendicular to the side. On a given edge, each
contribution of rk F from a small segment on one side of the centre of
this edge is cancelled by an equal and opposite contribution from a small
segment on the other side of the centre, so the net torque is the sum of
r F from all the segments on the edge, where F = I Bdx is the force
on the small segment. Now |r | = a2 so the torque on one edge is N =
a/2 a a2
a/2 2 IB sin dx = 2 IB sin . If I on this edge points in the x direction,
the direction of N is also x. There will be an equal torque from the opposite
edge, since both r and I point in opposite directions on the opposite edge,
so the total torque is

a2
(1) N=2 IB sin x = a2 IB sin x = m B
2
1
TORQUE ON A MAGNETIC DIPOLE 2

where m = Ia is the magnetic moment of the loop.


As a simple example, suppose we have a small circular current loop of
radius b lying in the xy plane at the origin, and a square current loop with
side length a on the y axis at a distance r from the origin. Both loops carry
current I.
If we take r  a, b, then we can approximate the circular loop as a pure
dipole with a magnetic moment of

(2) m = Ib2 z
The magnetic field due to a dipole is

0
(3) B= [3 (m r) r m]
4r3
In this case, r = y so

0 2 Ib2 0
(4) B= Ib z = z
4r3 4r3
The torque on the square loop is then

0 I 2 a2 b2
(5) N= m z
4r3

where m is a unit vector in the direction of the squares magnetic moment.


If the square is free to rotate, it will line itself up so that m = z.
P INGBACKS
Pingback: Torque on a magnetic moment; general current distribution
Pingback: Diamagnetism
Pingback: Floating a magnet above a superconductor
TORQUE ON A MAGNETIC MOMENT; GENERAL CURRENT
DISTRIBUTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.2.
Jackson, J. D. (1999) Classical Electrodynamics, 3rd Edition; Wiley -
Sections 5.6 and 5.7.
The torque on a square current loop in a constant, uniform magnetic field
is given by

(1) N = mB
It turns out that this formula applies to any current distribution, although
the proof is a bit tricky. Since the magnetic moment for a collection of line
currents is defined as

1
(2) m = Ia = I r dl
2
the natural generalization of magnetic moment to a general volume current
density is

1
(3) m= r Jd 3 r
2
The Lorentz force law for a volume current is

(4) F= J Bd 3 r

so the torque is

(5) N= r (J B) d 3 r

We can convert this using a vector identity:



3
(6) N= (r B) Jd r (r J) Bd 3 r
1
TORQUE ON A MAGNETIC MOMENT; GENERAL CURRENT DISTRIBUTION 2

To proceed further, we need another vector identity, which states that for
scalar fields f and g and localized vector field J, we have


(7) [ f J g + gJ f + f g J] d 3 r = 0

This follows, since if we integrate the middle term by parts, we get


(8) gJ f = f g (Jx + Jy + Jz ) f (gJ) d 3 r

(9) = 0 [ f J g + f g J] d 3 r

where the integrated term is zero if we assume that J goes to zero at infinity
(that is, its localized).
Now if we consider steady currents, then J = 0 so the identity reduces
to


(10) [ f J g + gJ f ] d 3 r = 0
p
If we take f = g = r = x2 + y2 + z2 , then r = r/r and


r
(11) 2 rJ d 3 r = 2 r Jd 3 r = 0
r

so the second integral in the torque 6 is zero (remember B is constant, so it


comes outside the integral).
For the first term, we use the identity with f = x, g = y:


(12) (xJy + yJx ) d 3 r = 0

Choosing the other coordinates in turn, we have in general for i and j


equal to any combination of x, y and z:


ri J j + r j Ji d 3 r = 0

(13)

Now look at the ith component of the first term of 6:


TORQUE ON A MAGNETIC MOMENT; GENERAL CURRENT DISTRIBUTION 3


(14) Ni = (r B) Ji d 3 r

(15) = B j r j Ji d 3 r
j

1
r j Ji ri J j d 3 r

(16) = Bj
2j

where we used 13 to get the last line.


If we take i = x for example, we get

1 1
(17) Nx = By (yJx xJy ) d r + Bz (zJx xJz ) d 3 r
3
2 2

1 h i
(18) = (r J)z By + (r J)y Bz d 3 r
2
  
1 3
(19) = r Jd r B
2 x
The other two components work out similarly, so we get
 
1 3
(20) N = r Jd r B
2
(21) = mB
I suspect this isnt the way Griffiths meant this problem to be solved,
since he deals only with line currents, but this is a more general proof any-
way.
FORCE BETWEEN TWO MAGNETIC DIPOLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.3.
A general formula for the force on a magnetic dipole due to a varying
magnetic field is

(1) F = (m B)
A special case of this occurs when we have a circular current loop sus-
pended in a radially symmetric field (such as that produced at the end of a
solenoid, or along the axis of another dipole). In that case, imagine that at
the location of the circular loop, the field spreads outwards like the petals
of a flower, so that it has a vertical component (parallel to the axis of sym-
metry) and a horizontal, radial component, perpendicular to the axis. The
force due to the field on the current is

(2) F=I d` B

integrated around the circle. The contributions from the vertical component
of B cancel out, but the horizontal component of magnitude B cos (where
the angle between the field and the horizontal is ) adds up around the circle
to give a net force (downwards) of

(3) F = 2RIB cos

where R is the radius of the circular loop.


We can apply both these formulas to the case of two magnetic dipoles m1
and m2 oriented along the z axis (so their axes are parallel) and separated
by a distance r. First, we use the gradient formula at the top. We start with
the dipole field of m1 in spherical coordinates:

0 m1 
(4) B= 3
2 cos r + sin
4r
Since m2 = m2 z, we get
1
FORCE BETWEEN TWO MAGNETIC DIPOLES 2

0 m1 m2
2 cos2 sin2

(5) m2 B = 3
4r
Taking the gradient in spherical coordinates, we get

(6) F = (m2 B)
0 m1 m2 
3 2 cos2 sin2 r 6 sin cos
 
(7) = 4
4r
In the case of an infinitesimal dipole on the z axis, = 0 and we get

30 m1 m2
(8) F = r
2r4
30 m1 m2
(9) = z
2r4
That is, the dipole m2 feels an attractive force towards m1 .
Doing the same calculation with the seemingly simpler formula 3 is ac-
tually more complicated, mainly because were dealing with an ideal, and
thus infinitesimal, dipole rather than a finite circular current loop. We can
start by working out the field due to m1 at the location of a finite loop with
radius R. Here, well redefine r as the distance from m1 to the centre of the
loop, so the field on the loop itself is then

0 m1 
(10) B= 3
2 cos r + sin
4

where
p
(11) r 2 + R2
r
(12) cos =

R
(13) sin =

We need the horizontal component of B at the loop, which we can get by
considering the r and components separately. The horizontal component
of Br is

0 m1   
(14) Brh = 2 cos cos
4 3 2
0 m1
(15) = (2 cos sin )
4 3
FORCE BETWEEN TWO MAGNETIC DIPOLES 3

Similarly, the horizontal component of B is

0 m1
(16) B h = (sin cos )
4 3
The total horizontal component of the field is the sum of these two, and
replaces the B cos factor in 3:

30 m1
(17) Bh = sin cos
4 3
Now we need to express the 2RI factor in terms of m2 . The magnetic
moment of the loop is m2 = R2 I, so

2m2
(18) 2RI =
R

and the force on the loop is then

30 m1 m2
(19) F= sin cos
2 3 R
To get the formula for an infinitesimal loop, we need to take the limit as
R 0. In this limit r, cos 1 and sin R/r so we get

30 m1 m2
(20) lim F =
R0 2r4

which agrees (in magnitude) with the formula obtained above by the gradi-
ent method.
P INGBACKS
Pingback: Magnetic toy
Pingback: Surface current induced by a magnet floating over a supercon-
ductor
FORCE ON A MAGNETIC DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.4.
Although the net force on a local, steady current distribution in a constant,
uniform magnetic field is zero, this is not true if the field varies in space.
One way of showing this (although it constrains the coordinate axes rather
artificially) goes like this.
Suppose we have an infinitesimal square loop carrying a current I. The
loop has side length and lies in the yz plane with one corner at the origin.
The magnetic field can vary over the loop, but we assume the variation is
small enough that a Taylor expansion of the field provides enough accuracy
in the calculations. Starting at the origin, we can get a Taylor expansion
along the y axis from y = 0 to y = :

B (0, y, 0) = B (0, 0, 0) + yy B (0, 0, 0) + O 2



(1)

where the notation y B (0, 0, 0) means B/ y evaluated at the origin. Sim-


ilarly, the field along the segment lying on the z axis is

B (0, 0, z) = B (0, 0, 0) + zz B (0, 0, 0) + O 2



(2)

On the top edge (where z = ), we can use the value of the field on the
bottom edge as a starting point for the Taylor expansion, so we have

B (0, y, ) = B (0, y, 0) + z B (0, y, 0) + O 2



(3)
= B (0, 0, 0) + yy B (0, 0, 0) + z B (0, 0, 0) + O 2

(4)

In the last line, weve approximated z B (0, y, 0) by z B (0, 0, 0) since


thats the first term in its Taylor expansion. Higher terms (involving the
second and higher derivatives of B are dropped since they contribute terms
of order 2 .
Similarly, the field on the right edge is approximated by
1
FORCE ON A MAGNETIC DIPOLE 2

B (0, , z) = B (0, 0, z) + y B (0, 0, z) + O 2



(5)
= B (0, 0, 0) + zz B (0, 0, 0) + y B (0, 0, 0) + O 2

(6)

With these expressions for B, we can use the Biot-Savart law to integrate
around the square and get the net force. The general formula is


(7) F=I dl B

Starting with the lower edge, we have dl = dyy (assuming the current is
flowing in the +y direction in this edge). The force on this edge is


(8) Fz=0 = I dy (Bz x Bx z)

where the subscripts x and z indicate which component of B were using.


Plugging in the above approximations, we get


(9) Fz=0 = I dy (Bz + yy Bz ) x I dy (Bx + yy Bx ) z
0 0

where all magnetic field terms are evaluated at the origin, and are therefore
constants. Doing the integrals, we get

2 2
   
(10) Fz=0 = I Bz + y Bz x I Bx + y Bx z
2 2
Along the right edge, dl = dzz and we have


(11) Fy= = I dz (By x + Bx y)

2
 
2
(12) = I By z By y By x+
2

2 
2
(13) I Bx + z Bx + y Bx y
2
Along the top, dl = dyy and
FORCE ON A MAGNETIC DIPOLE 3


(14) Fz= = I dy (Bz x + Bx z)

2
 
2
(15) = I Bz y Bz z Bz x+
2

2 
2
(16) I Bx + y Bx + z Bx z
2
Finally, along the left edge, dl = dzz and

(17) Fy=0 = I dz (By x Bx y)

2 2
   
(18) = I By + z By x I Bx + z Bx y
2 2
Adding up the four components of the force we find

(19) F = I 2 [ (y By + z Bz ) x + y Bx y + z Bx z]
Since B = 0, the first term is x Bx x so we get

(20) F = I 2 [x Bx x + y Bx y + z Bx z]
(21) = I 2 Bx
Because of the way weve oriented the axes, the magnetic moment of the
current loop is

(22) m = I 2 x

so the force can be written as

(23) F = (m B)
Although this derivation used a special case in the way the coordinates
were aligned, the formula is actually true in general.
P INGBACKS
Pingback: Force between two magnetic dipoles
Pingback: Force on a magnetic dipole in a slab of current
Pingback: Force on a magnetic dipole - a better derivation
Pingback: Floating a magnet above a superconductor
FORCE ON A MAGNETIC DIPOLE IN A SLAB OF CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.5.
A general formula for the force on a magnetic dipole due to a varying
magnetic field is

(1) F = (m B)
We can apply this formula to a case weve considered before: a slab
of uniform current. The slab is parallel to the yz plane and extends from
x = a to x = +a. The current density is J = J0 z. Using our earlier result,
the magnetic field inside the slab is

(2) B = 0 J0 xy

where the direction is determined by the right-hand rule.


For a dipole m = m0 x at the origin, the force is zero, since m B = 0. For
m = m0 y, m B = 0 m0 J0 x, so the force is

(3) F = 0 m0 J0 x
Note that even though m is a constant, we cant take it outside the
operator. For example, with m = m0 x, we would get

(4) (m ) B = 0 m0 J0 y 6= 0

and for m = m0 y we would get

(5) (m ) B = 0 6= 0 m0 J0 x
In the electrostatic analog, F = (p ) E = (p E). If we expand the
second term using the vector product rule, we get

(6) (p E) = p ( E) + E ( p) + (p ) E + (E ) p
1
FORCE ON A MAGNETIC DIPOLE IN A SLAB OF CURRENT 2

If p is a constant, the second and fourth terms are zero, and because
E = 0 in electrostatics, the first term is zero, which proves the relation.
In the magnetic case, B = 0 J 6= 0 so

(7) (m B) = 0 m J + (m ) B
We can see that by adding the 0 m J term to the (m ) B terms calcu-
lated above, we do indeed get the correct values as calculated by 1.
MAGNETIZATION: BOUND CURRENTS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problems 6.7, 6.8, 6.9.
By analogy with the polarization in electrostatics, we can define the mag-
netization M, which is the magnetic dipole moment per unit volume. We
can get a formula for the vector potential (and hence the magnetic field,
although there are often easier ways of finding the field) of an object con-
taining a given magnetization (which is a vector field). Starting with the
vector potential of an ideal dipole at the origin:

0
(1) A= m r
4r2

we can write this more generally as the potential when the dipole is at posi-
tion r0 :

0
m r r0

(2) A=
4 |r r0 |3
Then if M = M (r0 ) we can get the potential due to a distribution of mag-
netic dipoles as

0 1
M r0 r r0 d 3 r0
 
(3) A=
4 |r r0 |3
For pretty well any configuration, this integral is difficult or impossible
to calculate analytically, but we can transform it into a different form, in a
similar way to that used in the electrostatic case for polarization. First, we
use the formula

1 r r0
(4) 0 =
|r r0 | |r r0 |3

so the potential becomes



0 1
M r0 0 d 3 r0

(5) A=
4 |r r0 |
1
MAGNETIZATION: BOUND CURRENTS 2

Now we can use a vector product rule:

(6) ( f V) = f ( V) V f
With f = 1/ |r r0 | and V = M we get
 
0 1 3 0 0 M
(7) A= Md r d 3 r0
4 |r r0 | 4 |r r0 |
The first integral looks like the potential of a volume current density

(8) Jb M
 
0 Jb 3 0 0 M
(9) A = d r d 3 r0
4 |r r0 | 4 |r r0 |
The second integral can be transformed into a surface integral by using
the divergence theorem. For a general vector field V and a constant vector
field c we have, using a vector identity in the first line:

3 0 3 0
(10) (V c) d r = V ( c) d r c ( V) d 3 r0

(11) = c ( V) d 3 r0

(12) = c ( V) d 3 r0

(13) = (V c) da0

(14) = c V da0

Thus

3 0
(15) ( V) d r = V da0

so

0 Jb 3 0 0 M
(16) A= 0
d r + 0
da0
4 |r r | 4 |r r |
If we now define a surface current

(17) Kb M n
MAGNETIZATION: BOUND CURRENTS 3

where n is the unit normal to the surface, we get


0 Jb 0 Kb
(18) A= 0
d 3 r0 + da0
4 |r r | 4 |r r0 |
That is, we can replace the volume magnetization by a volume bound
current Jb and a surface bound current Kb and use them to calculate the
potential and the field.
In situations where there is some symmetry, we can use Ampres law
to calculate the field from these bound currents, as this usually proves a lot
easier than trying to do the integrals.
Here are a few examples.

Example 1. First, suppose we have an infinite circular cylinder containing


a uniform magnetization M parallel to its axis. Since M is constant, Jb = 0
and

(19) Kb = M

That is, the surface current flows in a circle around the outside of the cylin-
der. This is essentially the same as an infinite solenoid, so we know that the
field outside the cylinder is zero, and inside, we have B = 0 nI where n is
the number of turns per unit length and I is the current, so nI = Kb = M.
The direction of the field is given by the right hand rule, which means its
pointing in the same direction as M, thus inside:

(20) B = 0 M

Example 2. An infinite circular cylinder of radius R has magnetization M =


kr2 for a constant k. The bound currents are

(21) Kb = M n = kR2 z
1
r kr2 z

(22) Jb = M =
r r
(23) = 3krz

Inside the cylinder, using Ampres law the field due to the volume cur-
rent (there is no contribution from the surface current) is
MAGNETIZATION: BOUND CURRENTS 4

r 2
(24) 2rBJ = 2 0 (3k) r0 dr0
0
(25) BJ = 0 kr2
Outside, the field due to the volume current is obtained from Ampres
law by integrating around a circle of radius r. The volume current stops at
a radius R < r, so we have

R3
(26) BJ = 0 k
r
and the field due to the surface current is

(27) 2rBK = 0 (2R) kR2


R3
(28) BK = 0 k
r
In the first line, the total surface current is obtained by multiplying Kb by
the circumference of the cylinder, which is 2R.
Thus the total field is

(29) B = BJ + BK = 0
Note that the total current (volume + surface) is zero, since they are equal
in magnitude but opposite in direction. Thus by Ampres law, the field
outside the cylinder must be zero since the enclosed current is zero.
Example 3. Finally, for a cylinder with finite length and a constant M par-
allel to its axis, JB = 0 and KB = M as before, but KB = 0 on the ends
of the cylinder. The magnetic field lines come out of the north pole end of
the cylinder and loop around to go back into the south pole end. For a long
narrow cylinder, were back to the infinite solenoid, while for a short wide
cylinder we have essentially a planar current loop.

P INGBACKS
Pingback: Magnetization: microscopic versus macroscopic
Pingback: Auxiliary magnetic field H
Pingback: Magnetic field within a cavity
Pingback: Magnetostatic formulas from electrostatic formulas
Pingback: Uniform charge, polarization & magnetization
Pingback: Magnetic dipole embedded in sphere
Pingback: Magnetic dipoles versus monopoles; an experiment
MAGNETIZATION: BOUND CURRENTS 5

Pingback: Maxwells equations in matter


TOROIDAL SOLENOID WITH A GAP

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.10.
As another example of how we can calculate the magnetic field due to the
magnetization of an object, suppose we start with an iron bar with square
cross-section of side length a and length L, and having a constant magneti-
zation M parallel to the length of the bar. We now bend the bar into a circle
so that the end points almost touch, but are separated by a gap of width w.
Our task is to find the magnetic field at the centre of the gap.
The first thing to recognize is that by bending the bar into a circle, the
magnetization is no longer constant in magnitude within the bar, since the
outer edge of the circle will be stretched relative to the inside edge. The
new magnetization will be

L
(1) M 0 (r) = M
2r
We can verify this by finding the magnitude of the total dipole moment
in the bar:

L/2+a
(2) 2raM 0 (r) dr = a2 LM
L/2

so the total magnetic moment is the same as in the original unbent bar. Since
the magnetization now points along the circle, its vector form is

L
(3) M0 = M
2r
From the formula for the curl in cylindrical coordinates, we then get for
the bound volume current

(4) Jb = M0 = 0
The surface current has the values on the four sides of the square:
1
TOROIDAL SOLENOID WITH A GAP 2



Mz inside edge
LM

r top edge
(5) Kb = 2r ML

(L+2a) z outside edge

LM
2r r bottom edge
If we assume that L  a so that r L/2 for the whole width of the rod,
we can approximate Kb as a current density of magnitude M on all four
sides, so we have essentially a torus-shaped solenoid with a square cross
section. Griffiths works out the field inside a toroidal solenoid of arbitrary
cross-section as Example 5.10, using methods similar to those we used for
a linear solenoid, with the result (translating to the quantities used in this
post):

0 ML
(6) Btorus =
2 (L/2)
(7) = 0 M
This is the field inside a complete torus. To handle the gap, we can ap-
proximate it by a square current loop with current opposing that in the torus.
Weve worked out the field at the centre of such a loop before, so again
translating this into the quantities used here, we have:

2 2Mw0
(8) Bloop =
a
The net field at the centre of the gap is the sum of these two:
!
2 2w
(9) B = 0 M 1
a
MAGNETIZATION: MICROSCOPIC VERSUS MACROSCOPIC

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.11.
An object containing a magnetization distribution gives rise to a vector
potential according to:


0 1 0
 0
 3 0
(1) A= M r r r d r
4 |r r0 |3
This formula treats the magnetization as a continuous vector field, whereas
in reality, it is due to many discrete current loops created by spinning and
orbiting electrons and atomic nuclei. Even if the macroscopic currents are
constant, the microscopic configuration of the currents still changes rapidly,
so the microscopic magnetization is a rapidly varying function. Thus the
formula really relies on M being an average magnetization over some re-
gion around the observation point.
However, there appears to be something of a contradiction here, since
this formula was derived from the formula for a single, ideal dipole

0
m r r0

(2) A= 3
0
4 |r r |
In other words, we applied a microscopic formula to derive a macro-
scopic one. We can reconcile the two situations by the following argument.
We define the macroscopic magnetic field at a point r within the object
as the sum of the average field Bout due to all currents outside a sphere of
radius R centred at r and the field Bin produced by the currents within the
same sphere:

(3) B = Bout + Bin

For the exterior currents, we can assume that these currents are far enough
away that only their average effect is felt at the observation point, so we can
use the formula above to find the potential of these exterior currents within
the sphere:
1
MAGNETIZATION: MICROSCOPIC VERSUS MACROSCOPIC 2


0 1
M r0 rin r0 d 3 r0
 
(4) Aout (rin ) =
4 out |rin r0 |3

where the integral is taken over points r0 outside the sphere, and rin rep-
resents a point inside the sphere. So in principle, we need to average this
formula over all points rin inside the sphere. However, we showed earlier
that the average field within the sphere due to any current distribution out-
side a sphere is equal to the exact field produced at the centre of the sphere.
Thus the average of the field produced by these external currents is just

0 1 0
 0
 3 0
(5) Aav out = M r r r d r
4 out |r r0 |3

where r now points to the centre of the sphere. The magnetic field can then
be derived from this by the usual formula Bout = Aav out . Thus the field
due to the external currents is given correctly by using 1 to calculate the
field.
For the currents within the sphere, we also showed that the average mag-
netic field within the sphere for these currents is

0
(6) Bav = m
2R3

where m is the total magnetic moment of the currents within the sphere.
Now were assuming that although R is large compared to the scale of mi-
croscopic fluctuations, it is small enough that the magnetization M is essen-
tially constant over the sphere, so that

4
(7) m = R3 M
3

so the average field due to interior currents is

2
(8) Bin = 0 M
3
Griffiths shows in his example 6.1 that the internal magnetic field due to a
uniformly magnetized sphere also happens to be B = 32 0 M, and that result
was derived ultimately from the formula 1. In other words, the average field
at a point r as derived from the magnetization formula is the same as that
derived by considering the microscopic currents, so the extension of the
MAGNETIZATION: MICROSCOPIC VERSUS MACROSCOPIC 3

formula for ideal dipoles to a volume magnetization actually does give the
correct macroscopic result.
AUXILIARY MAGNETIC FIELD H

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.12.
Jackson, J. D. (1999) Classical Electrodynamics, 3rd Edition; Wiley -
Sections 5.8, 5.9.
An object containing a magnetization distribution can be modelled by
replacing the magnetization by bound volume and surface currents Jb and
Kb . If we add in some extra free current J f not due to the magnetization
(for example, by plugging the object into the electric mains), then the total
current at a point inside the object is

(1) J = Jb + J f
Since Jb = M by definition, we can write Ampres law as

1
(2) B = J
0
(3) = Jf +M
 
1
(4) BM = Jf
0
The quantity in parentheses is given the symbol H:

1
(5) H BM
0
H is sometimes called the auxiliary magnetic field or sometimes just the
magnetic field, with B being called the magnetic flux density. This allows a
variant form of Ampres law:

(6) H = Jf

(7) H d` = I f

where I f is the free current enclosed by the path of integration in the second
line.
1
AUXILIARY MAGNETIC FIELD H 2

One thing that is a bit worrying is that the bound surface current Kb
seems to have vanished in this derivation. Griffiths makes no mention of
this, but Jackson gets round the problem by saying that the surface integral
0 M 0
from which Kb was defined, namely 4 |rr0 | da is zero by assuming
that the magnetization M is well-behaved and localized, and we can take the
surface at infinity where M = 0. He then states later that in some idealized
problems, it is convenient to assume that M is discontinuous at the boundary
between two objects (for example, between a magnetized object and the
surrounding air), and in that case, the surface current term must be added
in. However, in any real physical situation, discontinuities never occur so
the surface term doesnt appear and the definition of H above is valid.
In any case, we can use this definition of H to calculate B more easily
in some idealized situations. For example, if we have an infinitely long
cylinder of radius R with a fixed M = krz, we can find B by two methods.
First, we use the bound current approach.

(8) Jb = M
(9) = k
(10) Kb = M (R) n
(11) = kR
Note that the total bound current is zero, since the total volume current is

R
(12) Jb dr = kR
0
Both bound currents effectively produce solenoids, so the field outside
the cylinder is zero. Inside, we have field due to the surface current, which
is

(13) BK = 0 kRz

and we must add to that the field due to those parts of the cylinder with
a radius greater than the the radius r of interest. The total current outside
radius r is

R
(14) Jb dr = k (R r)
r
so the total field is
AUXILIARY MAGNETIC FIELD H 3

(15) B = 0 kRz 0 k (R r) z = 0 krz = 0 M


Using H, we can take a loop of integration of radius r centred on the z
axis. Since there is no free current, we get

(16) H d` = 0

and from the symmetry of the problem we can conclude that H = 0.


Alternatively, we can work from the curl equation which gives H =
0. This on its own isnt enough to conclude that H = 0, but we can also
calculate the divergence

1
(17) H = BM
0
Since B = 0 always, and by direct calculation we can show that
M = 0 in this case, we have both the curl and divergence of H as zero, so H
must be zero. From that, we can conclude immediately from the definition
of H that B = 0 M.
P INGBACKS
Pingback: Magnetic field within a cavity
Pingback: Magnetostatic boundary conditions; Laplaces equation
Pingback: Magnetic field within a current-carrying wire
Pingback: Magnetic systems in thermodynamics
MAGNETIC FIELD WITHIN A CAVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.13.
We can do a similar analysis to that for cavities in a dielectricto calculate
the magnetic fields in cavities within magnetized material. We start with
a magnetic material within which the magnetic field is B0 and the mag-
netization is M (neither of which need be constant). The auxiliary fieldis
H0 = 10 B0 M.
If we hollow out a spherical cavity we can simulate this by superimposing
a sphere with opposite magnetization onto the material. Griffiths shows in
his example 6.1 that the internal magnetic field due to a uniformly magne-
tized sphere with magnetization M is Bs = 23 0 M so the net field inside
the cavity is

2
(1) B = B0 0 M
3
Since the magnetization within the cavity is zero, the auxiliary field is

1 1 2 1
(2) H= B = B0 M = H0 + M
0 0 3 3
For a needle-shaped cavity with magnetization M, a bound surface cur-
rentis induced which is

(3) Kb = M n = M
This is effectively a solenoid, so its field within the needle is Bs = 0 M
and the net field is

(4) B = B0 0 M
1
(5) H = B0 M = H0
0
Finally, for a thin circular wafer of radius R and thickness t, the surface
current is the same as for the needle, but the setup can be approximated
1
MAGNETIC FIELD WITHIN A CAVITY 2

by a circular current loop, the field of which is derived by Griffiths in his


example 5.6. For a point in the centre of the loop, this gives

0 I
(6) B=
2R

where I is the current in the loop. In our case, I = Mt, so the field is

0t
(7) Bw = M
2R
If t  R (that is, the wafer is very thin), then Bw 0 so

(8) B = B0
1
(9) H = B0 = H0 + M
0
Thus B is unchanged for the wafer and H is unchanged for the needle.
P INGBACKS
Pingback: Magnetic dipoles versus monopoles; an experiment
MAGNETOSTATIC BOUNDARY CONDITIONS; LAPLACES
EQUATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.15.
From the magnetostatic boundary conditions on the magnetic field B we
can work out the boundary conditions on the auxiliary field H. First, we
need the divergence of H:

1
(1) H = B M = M
0
Using a similar argument to that for the electric field, we can see that H
and M have equal and opposite discontinuites at a boundary, so

 
(2) Habove Hbelow = M
above below
M

Using the same argument as for magnetostatic boundary conditions, the


boundary condition on the component of H parallel to the boundary is

(3) Habove
k Hbelow
k = K f n

where K f is the free surface current density on the boundary.


The quantity Hk is written as a vector since it lies in the tangent plane to
the surface and has a direction given by the cross product K f n.
In the special case where there is no free current anywhere, then

(4) H = Jf = 0

which means that the vector field H can be written as the gradient of a scalar
field W :

(5) H = W
1
MAGNETOSTATIC BOUNDARY CONDITIONS; LAPLACES EQUATION 2

As an example of using this, we can derive the field due to a uniformly


magnetized sphere. The magnetization is M = Mz within the sphere and
zero outside. Since M is constant everywhere except at the boundary, then

(6) H = 2W = 0

everywhere except at the boundary. This is Laplaces equation and in spher-


ical coordinates, the general solution is

 
l Bl
(7) W (r, ) = Al r + l+1 Pl (cos )
l=0 r

We can follow the usual procedure for finding the coefficients by impos-
ing boundary conditions on W . Just as in the electrostatic case, the potential
must be continuous at the boundary, and must be finite everywhere. This
means that inside the sphere, we must have Bl = 0 and outside the sphere
Al = 0 so that


(8) Win = Al rl Pl (cos )
l=0

B
(9) Wout = rl+1l Pl (cos )
l=0

At the boundary we must have Win = Wout and equating coefficients of


each Legendre polynomial we get

(10) Bl = Al R2l+1

We can now consider the derivative of W in the r direction, which gives


the normal component of H as W / r. Using 2 we have


W W above below
(11) = M M
r out r in
(12) = Mr z
(13) = M cos

Taking the derivatives of the series above, we get for l = 1 after using 10
MAGNETOSTATIC BOUNDARY CONDITIONS; LAPLACES EQUATION 3

l +1
(14) l+2
Bl = lAl Rl1 M
R
M
(15) A1 =
3
M 3
(16) B1 = R
3
For l 6= 1 we have

l +1
(17) l+2
Bl = lAl Rl1
R
(18) (l + 1) Rl1 Al = lAl Rl1
(19) (2l + 1) Al = 0
Since this must be true for all l 6= 1 we must have Al = Bl = 0 for these
cases. Thus

M
(20) Win = r cos
3
M R3
(21) Wout = cos
3 r2
Taking the negative gradient, we get, since r cos = z

M
(22) Hin = z
3
MR3 
(23) Hout = 3
2 cos r + sin
3r
From this we can get the field B = 0 (H + M) (remember M = 0 outside
the sphere):

20 M
(24) Bin = z
3
0 MR3 
(25) Bout = 2 cos r + sin
3r3
P INGBACKS
Pingback: Boundary between magnetic materials
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.16.
The magnetic analog to the electric dielectric constant is the magnetic
permeability. In a linear magnetic material, the magnetization is directly
proportional to the auxiliary field, so we have

(0.1) M = m H

where m is the magnetic susceptibility. Given this proportionality, the field


is also directly proportional to H:

(0.2) B = 0 (H + M) = 0 (1 + m ) H H

where is the magnetic permeability. In a vacuum, the magnetization is


always zero, so m = 0 in this case, and the permeability becomes = 0 ,
which is known as the permeability of free space.
The permeability is a property of the material so different substances
have different permeabilities. In general, H can be found without knowing
anything about the material, but in order for B to be found, the permeability
must be known.
As a simple example, suppose we have two infinitely long concentric
tubes (hollow cylinders), with the inner tube having radius a and carrying
current I z and the outer tube having radius b and current I z.
By symmetry, we know that B, H and M are all circumferential (that is,
proportional to ) so we can use Ampres law:


(0.3) H d` = I f

where I f is the free current enclosed by the path of integration. Choosing a


circular path of radius r between the two cylinders we get

(0.4) 2rH = I
I
(0.5) H =
2r
(0.6) B = 0 (1 + m ) H
I
(0.7) = 0 (1 + m )
2r
1
2

We can also find the magnetization and resulting bound currents. We


have

Im
(0.8) M = m H =
2r
(0.9) Jb = M = 0
Im
(0.10) Kbi = M n = z
2a

where Kbi in the last line is the bound surface current density on the inner
cylinder, where the normal vector points inwards since were in the region
between the cylinders, so n = r. Having determined the bound currents,
we can work out B from these currents (and the free currents) without wor-
rying about the magnetization. Since the field due to a wire carrying a total
current It is (from Ampres law):

0 It
(0.11) B=
2r

and the total bound current flowing on the inner cylinder is 2aKbi = Im
the field due to the combined free and bound currents here is

0 (1 + m ) I
(0.12) B=
2r

in agreement with the earlier calculation. Note that the outer cylinder has
no effect on the field between the cylinders since the path of integration
excludes this current. If we go outside the outer cylinder, the net free current
is zero so we get

(0.13) 2rH = 0
(0.14) H = 0

so B = 0 here as well. The bound current on the outer cylinder is

Im
(0.15) Kbo = M n = z
2b

since here the normal to the surface points outwards, so n = +r. The total
bound current on the outer cylinder is thus 2Kbo = Im so the total bound
current from both cylinders is zero, making the total current zero.
3

P INGBACKS
Pingback: Diamagnetism
Pingback: Boundary between magnetic materials
Pingback: Transmission lines
Pingback: Maxwells equations in matter: boundary conditions
Pingback: Magnetic systems in thermodynamics
MAGNETIC FIELD WITHIN A CURRENT-CARRYING WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.17.
Another example of using the auxiliary field to find the magnetic field.
This time, we have a long wire of radius a carrying a uniform current I.
The wire is made of a linear metal so that the magnetization is directly
proportional to H.

(1) M = m H

where m is the magnetic susceptibility. Given this proportionality, the field


is also directly proportional to H:

(2) B = 0 (H + M) = 0 (1 + m ) H H

where is the magnetic permeability.


From the symmetry of the setup, H is circumferential, so we can take a
circular integration path to get


(3) H d` = I f

r2
(4) 2rH = I
a2
rI
(5) H =
2a2
From this we get the field for r < a

0 (1 + m ) rI
(6) B=
2a2
Outside the wire, m = 0 and the enclosed free current is just I, so for
r>a
1
MAGNETIC FIELD WITHIN A CURRENT-CARRYING WIRE 2

0 I
(7) B=
2r
The magnetization is

m rI
(8) M=
2a2

so the bound currents are

(9) Jb = M
m I
(10) = z
a2
(11) Kb = M n
m I
(12) = z
2a
The total bound current is

(13) Ib = a2 Jb + 2aKb = 0
MAGNETIC FIELD OF A SPHERE IN A UNIFORM FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.18.
If we place a linear magnetic sphere in a uniform magnetic field B0 =
B0 z, we can find the magnetic field inside the sphere using the method of
successive approximations we used for a dielectric sphere in an electric
field.
We start by assuming that the only field present is the uniform field,
which then induces a magnetization within the sphere:

m
(1) M0 = B0

We then find the field produced by this magnetization (which is done by
Griffiths in example 6.1):

2 20 m
(2) B1 = 0 M0 = B0
3 3
This field produces another bit of magnetization:

m 20 m2
(3) M1 = B1 = B0
3 2

which in turn produces a bit more field:

 2
2 20 m
(4) B2 = 0 M1 = B0
3 3
The process repeats so that for the nth iteration we get

 n
20 m
(5) Bn = B0
3
The total field is then the sum of all the individual contributions:
1
MAGNETIC FIELD OF A SPHERE IN A UNIFORM FIELD 2

 n
20 m
(6) B = B0
n=0 3
1
(7) = B0
1 23
0 m

1
(8) = B0 2m
1 3(1+ m)
3 + 3m
(9) = B0
3 + m
Unfortunately, the same method doesnt work for finding the field outside
the sphere, since that field is not constant in space.
P INGBACKS
Pingback: Magnetostatic formulas from electrostatic formulas
DIAMAGNETISM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.19.
If a magnetic dipole is placed in an external field, the field exerts a torque
on the dipole according to

(1) N = mB
In a magnetic material, this torque tends to line up the electron spins thus
producing an overall magnetization in the same direction as the field. This
effect is called paramagnetism. Because of the Pauli exclusion principle,
the electrons in an atom pair up, with the spins of the two electrons within
the pair in opposite directions, causing the overall magnetization to can-
cel out if each atom has an even number of electrons. Paramagnetism is
therefore strongest in atoms with odd numbers of electrons.
There is a much weaker effect known as diamagnetism, in which the mag-
netization is opposite to the applied field. This effect occurs in all atoms
but in atoms with odd numbers of electrons, it is masked by the stronger
paramagnetism, so it is mainly in even-numbered atoms that it appears. Al-
though diamagnetism is ultimately a quantum phenomenon, a rough idea of
its cause can be found from a purely classical argument.
Rather than looking at the spin of the electron, we look at its orbital
motion. Classically, an electron orbits the nucleus in a circular orbit of
radius R at a speed v, giving rise to a steady current I = ev/2R, with e
being the electrons charge. (Technically, a single moving charge doesnt
constitute a steady current, but if the electron orbits fast enough, its a fair
approximation.) The dipole moment of a current in a circular loop is

1
(2) m = R2 I z = evRz
2

where weve taken the plane of the orbit as perpendicular to the z axis.
We want to figure out what changes in the orbit when a field is applied.
Without a field, the electron is kept in its orbit by an electrical force produc-
ing the centripetal force needed. If the electron is orbiting a nucleus with Z
protons, then we get
1
DIAMAGNETISM 2

Ze2 v2
(3) = m e
40 R2 R

where me is the mass of the electron. (This ignores effects such as the
shielding of the nucleus from outer electrons by inner electrons and so on,
but since its a classical argument for a quantum phenomenon, we cant get
too picky anyway.)
Now if a field B in the z direction is applied, then the electron feels an
additional force ev B which will add to the centripetal force (that is, the
magnetic force is towards the centre of the orbit) if the electron is moving
so as to give the dipole moment stated above. (If the electron is moving in
the opposite direction around its orbit, the magnetic force is outwards.)
This extra centripetal force means that the radius of the orbit and/or the
speed of the electron must change. If we assume that the radius stays the
same, then the speed must increase, so that, for the new speed v:

Ze2 v2
(4) + evB = me
40 R2 R
Subtracting the first equation from the second, we get

me 2  me
(5) evB = v v2 = (v + v) (v v)
R R
If we now assume that the change in speed is small so that v + v 2v we
get

eBR
(6) v = >0
2me
The change in dipole moment is then

1 e2 R2 B
(7) m = evRz = z
2 4me
That is, the change in dipole moment (which is what causes the mag-
netization) is opposite to the applied field, which is what is observed in
diamagnetism.
If we started with the electron moving in the opposite direction, then the
sign of m is opposite, as is the direction of the magnetic force. If we still
assume the radius remains the same, then we must have v < v or v < 0.
Thus we would then get
DIAMAGNETISM 3

1 e2 R2 B
(8) m = evRz = z
2 4me
That is, the change in magnetization is the same as before, so it is still
opposite to the applied field.
To get an idea of the numbers involved, we need a value for R. Measured
values for elements in the middle of the periodic table are around 1.35
1010 m so using that value, we can calculate the magnetization:

m 3e2
(9) M = 4 3
= B
3 R
16me R
(10) = 12.46B
m
From this we can get the susceptibility, since M = m H = 0 B, so

(11) m = 12.460 = 1.57 105


Typical values as given in Griffiths Table 6.1 are in this general area,
which is surprising considering it was a classical derivation.
ENERGY OF MAGNETIC DIPOLE IN MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.21.
Using similar techniques to those for finding the energy of an electric
dipole in an electric field, we can work out the energy of a magnetic dipole
in a magnetic field. The torque exerted by a field on a dipole is

(1) N = mB
If we bring in a dipole from infinity in such a way that no work is done
(by moving the dipole along a path that is always perpendicular to the force
exerted by the field) then when we arrive at the desired location, we must
rotate the dipole into its final position, and this requires work to be done
against the torque. If we start with m B, then the work done is

(2) U = mB sin d = mB cos = m B
/2
The choice of = /2 as the zero point for the energy appears to be
arbitrary, although it does make the situation more symmetric. The maxi-
mum energy of mB occurs when m is anti-parallel to B and the minimum
energy of mB when m and B are parallel. If we had chosen a different
zero angle, the difference in energy between two angles would still be the
same, and that is all that matters physically.
The interaction energy of two dipoles can be written down from the for-
mula for the field due to a dipole:

0
(3) B1 = [3 (m1 r) r m1 ]
4r3
We then get

(4) U12 = m2 B1
0
(5) = [m1 m2 3 (m1 r) (m2 r)]
4r3
Here r is the unit vector pointing from m1 to m2 , so if we define i to be
the angle between r and mi then
1
ENERGY OF MAGNETIC DIPOLE IN MAGNETIC FIELD 2

0 m1 m2
(6) U12 = (cos (1 2 ) 3 cos 1 cos 2 )
4r3
The stable configuration is when this energy is a minimum, so consider-
ing the trig functions alone, we have

U12
(7) = sin (1 2 ) + 3 sin 1 cos 2 = 0
1
U12
(8) = sin (1 2 ) + 3 cos 1 sin 2 = 0
2
Using the shorthand notation s12 sin (1 2 ), s1 sin 1 , c1 cos 1 ,
etc, we have

(9) s12 = 3s1 c2


(10) s1 c2 + c1 s2 = 0
(11) tan 1 = tan 2
Since 0 1 , 2 , this means that either tan 1 = tan 2 = 0 or 1 =

2 + x; 2 = 2 x. Either way we have s1 = s2 s and c1 = c2 c, so

(12) s12 = 3sc


(13) s1 c2 c1 s2 = 3sc
(14) sc cs = 3sc
(15) 2sc = 3sc
The only way the last equation can be true is if s = 0 and/or c = 0, which
means the choices are 1,2 = 0, 2 , . We can try the various possibilities,
but its fairly obvious anyway that the stable configuration is when the two
moments are parallel, that is, 1 = 2 = 0 (or 1 = 2 = ). If we line up
a sequence of compass needles, they will align themselves so that the north
end of one points to the south end of the next, and so forth.
P INGBACKS
Pingback: Mutual inductance between two magnetic dipoles
FORCE ON A MAGNETIC DIPOLE - A BETTER DERIVATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.22.
Weve seen a rather crude derivation of the force on a magnetic dipole in
a varying magnetic field, which gives the result

(1) F = (m B)
In that derivation, we considered a particular geometry of dipole (a square)
and constrained the orientation of the dipole within the coordinate system.
A more general approach is given here.
If B (r) is a general vector field, we can write its value near the location
r0 of the dipole as a 3-d Taylor expansion to first order:

(2) B (r) = B (r0 ) + [(r r0 ) ] B0

where the subscript 0 on B0 in the second term indicates that the derivatives
of B are evaluated at r0 .
Putting this into the Lorentz force law, we have (given the current I pro-
ducing the dipole)

(3) F = I d` B (r)

(4) = I d` B (r0 ) + I d` [(r r0 ) ] B0

The first integral and r0 term in the second integral come out to zero,
since we are integrating a constant around a closed loop. Thus we are left
with

(5) F=I d` (r ) B0

We can write the cross product in rectangular coordinates using the Levi-
Civita symbol i jk = +1 for a cyclic permutation of 1,2,3, 1 for an anti-
cyclic permutation and zero if any two indices are equal. In general
1
FORCE ON A MAGNETIC DIPOLE - A BETTER DERIVATION 2

3
(6) (A B)i = i jk A j Bk
j,k=1

The dot product can be written as

3
(7) AB = Al Bl
l=1
Using these two sums, we have for a component of the force:

(8) Fi = I i jk rl d` j l Bk0
j,k,l

The integral is of a form that weve encountered when discussing the


vector area a of a curve. For a constant vector c:

(9) c rd` = a c

In components, this is

(10) rl cl d` j = jmn am cn
l m,n

Using c = Bk0 , we can plug this back into the force equation to get

(11) Fi = I i jk jmn am (Bk0 )n


j,k,m,n

We can now use an identity for a sum over the Levi-Civita symbols:

(12) i jk n jm = inkm imkn


j

Because of the cyclic property of the n jm , we have n jm = jmn , so we


now get for the force:

(13) Fi = I (in km im kn ) am (Bk0 )n


k,m,n
(14) = I ak (Bk0 )i I ai (Bk0 )k
k k
For a fixed dipole, the moment is a constant and is
FORCE ON A MAGNETIC DIPOLE - A BETTER DERIVATION 3

(15) m = Ia

so the first term is

(16) I ak (Bk0 )i = mk (Bk0)i


k k
(17) = (mk Bk0)i
k
(18) = (m B0 )i
In the second term, the sum comes out to

(19) I ai (Bk0 )k = Iai (Bk0 )k


k k
Bk0
(20) = Iai
k xk
(21) = Iai B0
(22) = 0

since B = 0 in general.
Thus we get the previous result

(23) F = (m B)
MAGNETOSTATIC FORMULAS FROM ELECTROSTATIC
FORMULAS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.23.
You might have noticed that there are many parallels between the equa-
tions of electrostatics and magnetostatics. We can summarize these with the
relations between the core quantities in each case.
For electrostatics, in the absence of free charge (although bound charges
due to polarization effects may still be present):

(1) D = 0, E = 0, 0 E = D P
For magnetostatics, in the absence of free currents (although again, bound
currents due to magnetization may be present):

(2) B = 0, H = 0, 0 H = B 0 M
Thus if we have the equations for problems in electrostatics, we should
be able to generate the corresponding results for magnetostatics by applying
the substitutions:

D B, E H, P 0 M, 0 0
Here are 3 examples of how this works.
First, the electric field inside a uniformly polarized sphere is given in
Griffithss example 4.2:

1
(3) E= P
30
Making the substitutions, we get the magnetic field inside a uniformly
magnetized sphere:

1
(4) H = M
3
2
(5) B = 0 (H + M) = 0 M
3
1
MAGNETOSTATIC FORMULAS FROM ELECTROSTATIC FORMULAS 2

which agrees with Griffithss example 6.1.


Second, the electric field inside a sphere of linear dielectric placed in a
uniform electric field E0 is given in Griffithss example 4.7:

3 3
(6) E= E0 = E0
r + 2 3 + e

where r = 1 + e is the dielectric constant.


To transform this equation we need to know how to transform e . From
its definition, we have

(7) P = 0 e E
Making the substitutions on this equation, we have

(8) 0 M = 0 XH

where X is the transformation of e . From the relation between M and H in


a linear material, we know that M = m H, so X = m , and thus the formula
for the field inside a sphere of linear magnetic material placed in a uniform
field is

3
(9) H= H0
3 + m
To get this in terms of B, we note that H0 = B0 /0 since it is the back-
ground field outside the sphere, where M = 0. We also have

(10) B = 0 (H + M) = 0 (1 + m ) H

so

1 3 1
(11) B = B0
0 (1 + m ) 3 + m 0
3 + 3m
(12) B = B0
3 + m
This agrees with the result we found earlier by another method.
Finally, we have the average field within a sphere due to charge dis-
tributed arbitrarily within that sphere:
MAGNETOSTATIC FORMULAS FROM ELECTROSTATIC FORMULAS 3

1 p
(13) Eav =
40 R3

where p is the total dipole moment of the sphere.


Since we know how to transform the polarization density P, and p =
4 3
3 R P where P is the average polarization density, p should transform the
same way, so

4 3 4
(14) R P R3 0 M = 0 m
3 3

where m is the total magnetic dipole moment. Therefore the overall trans-
formation is

1
(15) Hav = m
4R3
(16) Bav = 0 (Hav + M)
 
1 3
(17) = 0 m+ m
4R3 4R3
20
(18) = m
4R3
Again, this agrees with the result we got earlier via another method.
UNIFORM CHARGE, POLARIZATION & MAGNETIZATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.24.
If you cast your mind back to the original definition of the electric field,
we have

1  r r0 3 0
(1) E(r) = r0 d r
40 |r r0 |3

where the integral extends over all space. If is constant, it comes out of
the integral:

r r0 3 0
(2) E(r) = 0 3
d r
40 |r r | 40
The potential of a polarized object is

1 (r r0 ) P(r0 )
(3) V (r) = d 3 r0
40 |r r0 |3
Again, if the polarization is constant, we get

1 (r r0 ) 1
(4) V (r) = P d 3 r0 = P
40 |r r0 |3 40
Finally, for a magnetized object, the vector potential is

0 1
M r0 r r0 d 3 r0
 
(5) A=
4 |r r0 |3
For constant magnetization, we get

0 r r0 0
(6) A= M d 3 r0 = M
4 |r r0 |3 4
Therefore, if we can work out the electric field for a given shape where
the charge density is constant, we can get the potentials of a uniformly
polarized or magnetized object of the same shape by simple substitution.
1
UNIFORM CHARGE, POLARIZATION & MAGNETIZATION 2

For example, for a charged sphere of radius R, we can find the electric
field easily using Gausss law, and we get
(
30 rr r<R
(7) E (r) = R3
30 r2
r r>R
The potential for a uniformly polarized sphere is found by replacing by
P and taking the dot product, assuming P points along the z axis:
(
r r cos
30 P r = 30 P r<R
(8) V (r) = R 3 R3 cos
30 r 2 P r = 30 r2
P r>R
The vector potential of a uniformly magnetized sphere is then found by
replacing /0 by 0 M and taking the cross product:
( 0 r sin
0r
3 M r = 3 M r<R
(9) A (r) = 0 R3 0 R3 sin
3r2
M r = 3r2 M r>R
MAGNETIC TOY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.25.
A magnetic toy consists of some torus-shaped magnets placed on a cen-
tral pole. (Admittedly, I couldnt find any mention of such a toy, possibly
because there isnt a lot you can do with it, so it wouldnt be much fun
to play with.) If we treat each torus as an ideal dipole, we can find the
distances between successive magnets by equating the magnetic forces on
each torus with the gravitational force.
Lets start with just two magnets, each with a dipole moment m and mass
md . The first magnet goes at the bottom of the pole and the second magnet
is placed on top of it. Weve already worked out the force between two
dipoles, and weve seen that if the dipole moments are parallel, the force
is attractive, so if we want the top magnet to float over the bottom one, we
need to align the dipole moments so they are anti-parallel. In this case, the
force on the top magnet is

30 m2
 
(1) F= md g z
2z4
Setting this to zero and solving for the height z we get

1/4
30 m2

(2) z=
2md g
If we now add a third magnet on top of the first two, with its moment
parallel to the bottom magnet, then at equilibrium the forces on the middle
and top magnets are
" ! #
30 m2 1 1
(3) Ft = md g z = 0
2 z42 (z1 + z2 )4
30 m2
   
1 1
(4) Fm = 4 + 4 md g z = 0
2 z2 z1

where z2 is the distance between the middle and top magnets and z1 between
1
MAGNETIC TOY 2

the middle and bottom magnets. Since both forces are zero, we can equate
them to get

1 1 1 1
(5) = +
z42 (z1 + z2 )4 z42 z41
1 1
(6) 1 4
= 1
( + 1) 4

where z1 /z2 . We can solve this equation numerically to get

(7) = 0.8501149795 . . .
That is, the lower two magnets are slightly closer together than the upper
two.
BOUNDARY BETWEEN MAGNETIC MATERIALS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.26.
Just as electric field lines bend when crossing a boundary between two
different dielectrics, so magnetic field lines bend when crossing a boundary
between different permeabilities.
From the boundary conditions on the field B and on the auxiliary field H
we have

(1) Babove
= Bbelow

(2) Habove
k Hbelow
k = K f n
If there is no free current at the boundary, then K f = 0 and Habove
k =
Hbelow
k . From the definition of permeability, we also have, for linear mate-
rials:

(3) B = H
If i is the angle between the normal to the boundary and the field lines
Bi in material i, then the conditions above come out to

(4) B1 cos 1 = B2 cos 2


B1 sin 1 B2 sin 2
(5) =
1 2
Dividing these two equations, we get

tan 1 1
(6) =
tan 2 2
P INGBACKS
Pingback: Floating a magnet above a superconductor

1
MAGNETIC DIPOLE EMBEDDED IN SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.27.
Well now consider the problem of finding the magnetic field due to an
ideal dipole embedded at the centre of a sphere of linear magnetic material
with permeability .
This is a surprisingly tricky problem, and I must confess that I had to
seek a bit of help via everyones friend: Google. The starting point is to
try to figure out the bound currents and calculate the field from them. For
a linear material, the magnetization is proportional to the auxiliary field
(M = m H), so the bound volume current is

(1) Jb = M = (m H) = m H = m J f

where J f is the free volume current.


At first glance, it might seem that there is no free current in this problem,
so that Jb = 0 as well. However, the dipole embedded in the sphere must
be caused by a current. (Recall that for a planar loop carrying current I, the
dipole moment is m = IA, where A is the area enclosed by the loop, and an
ideal dipole is just a limiting case as the size of the loop goes to zero, with
the current becoming infinite in such a way as to preserve the magnitude
of the magnetic moment.) Since the dipole moment is proportional to the
current, we can see that the free dipole m gives rise to a bound dipole
mb = m m, also at the centre of the sphere. The effective dipole moment at
the centre is therefore the sum of the free and bound dipoles, giving


(2) me = m + mb = (1 + m ) m = m
0
Since the embedded dipole is the only free current in the problem, we
know that Jb = 0 everywhere except at r = 0, so we can work out the field
due to the bound volume current at this point.
Weve already worked out the field due to a dipole, so the field due to this
effective dipole is
1
MAGNETIC DIPOLE EMBEDDED IN SPHERE 2

0
(3) B1 = [3 (me r) r me ]
4r3

(4) = [3 (m r) r m]
4r3
Thats the easy bit. The really tricky bit comes when we try to work out
the surface bound current Kb . As Griffiths gives the answer in the question,
we can see that the field due to the surface current is supposed to be a
constant times m. So lets say that the total field is

(5) B = B1 + m
What we need in order to find Kb = M n = M r is the magnetization
M. Since the material is linear, we know that M = m H = m B so we get

m m
(6) M= [3 (m r) r m] + m
4r3
Now at the surface of the sphere, r = R so we get

(7) Kb = M r
m m
(8) = 3
m r + m r
4R
 
1
(9) = m m sin
4R3

where the last line follows because were taking m = mz and is the polar
angle in spherical coordinates.
At this point, we have to notice that the form of Kb is that of a spinning
shell with a constant surface charge density . This is because in that case,
Kb = v, where v is the velocity of a point on the sphere, and for a sphere
rotating at constant angular velocity this is v = r = R sin . Grif-
fiths shows in his example 5.11 that the field inside the sphere due to this
surface current is

2
(10) B = 0 R z
3
Comparing the two equations, we can make the substitution
 
1
(11) R = m m
4R3
MAGNETIC DIPOLE EMBEDDED IN SPHERE 3

so the field due to Kb is

 
2 1
(12) B2 = 0 m mz
3 4R3
 
2 1
(13) = 0 m m
3 4R3

In order for this to match up with our original assumption 5, we must


have

 
2 1
(14) = 0 m
3 4R3
 1
2 0 m 2 0 m
(15) = 1
3 4R3 3


We can now use m = 0 1 to get

   1
2 ( 0 ) 2 0
(16) = 1 1
3 4R3 3
1 20 1
 
2 ( 0 )
(17) =
3 4R3 3 3
2 (0 )
(18) =
4R3 (20 + )

Putting it all together, we get

 
3 2 (0 )
(19) B= 3
(m r) r m + m
4 r 4R3 (20 + )

which is the answer given in Griffithss question.


Outside the sphere, the field is the sum of that due to the effective dipole
at the centre and the surface bound charge. Since a spinning spherical shell
behaves as an exact dipole when seen from outside, the dipole moment due
to the surface charge is, using the substitution 11
MAGNETIC DIPOLE EMBEDDED IN SPHERE 4

 
4 3 1
(20) ms = R m m
3 4R3
  
1 2 (0 )
(21) = 1 1 m
3 20 + 0
(0 )
(22) = m
0 (20 + )
The total dipole moment seen from outside the sphere is then

(23) mout = ms + me
 
0
(24) = m +1
0 20 +
3
(25) = m
20 +
The field seen outside the sphere is therefore

0 3
(26) Bout = 3
[3 (m r) r m]
4r 20 +
Now thats what I call a hard problem for this level of textbook. Id
advise Griffiths to mark it with the ! (more difficult than normal) symbol in
the next edition :-)
MAGNETIC DIPOLES VERSUS MONOPOLES; AN
EXPERIMENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007) Introduction to Electrodynamics,
3rd Edition; Prentice Hall - Problem 6.28.
An interesting application of the calculation of the magnetic and elec-
tric fields within a cavity is the following experiment. We start with a long
cylinder of magnetized material with magnetization M = Mz. This magne-
tization produces a bound surface currentKb = M which in turn produces
a magnetic field inside the cylinder:

(1) Bm = 0 Mz
Suppose now that magnetism is caused by magnetic monopoles instead
of dipoles produced by microscopic current loops. In that case, the magne-
tization would induce a bound surface magnetic charge density of c =
M n = M on the ends of the cylinder, but no surface charge on the sides. If
the ends of the cylinder are small compared to its length, the magnetic field
produced by this charge on the ends can be approximated by a point charge,
so it will be

0 c A
(2) Bc = z
4 r2

where A is the area of the end of the cylinder and r is the distance from the
end to the observation point. The minus sign appears because the magneti-
zation causes positive charges to accumulate on the top end of the cylinder
so that the magnetic field points downwards, from positive to negative. If
r2  A, then Bc 0.
Since the fields produced in the two cases are quite different, it should be
possible to conduct an experiment to determine which field actually occurs.
We can try this by hollowing out a small cavity within the cylinder and
inserting a probe to measure the field. We can use our earlier calculations
to find the expected field within the cavity in each case.
First, if the field is caused by dipoles, then using B0 = Bm for a spherical
cavity we get
1
MAGNETIC DIPOLES VERSUS MONOPOLES; AN EXPERIMENT 2

2 1
(3) B = Bm 0 M = 0 M
3 3
If the field is caused by monopoles, then using B0 = Bc = 0 we get for a
spherical cavity (using the substitutions E B, P M and 0 1/0 ):

1 1
(4) B = Bc + 0 M = 0 M
3 3
Thus for a spherical cavity, the two fields are the same. However, it would
seem that if we made the cylinder shorter, then Bc would be significantly
non-zero, so there should be a detectable difference.
For the other two shapes of cavity, we have:
Needle: B = 0 for both cases.
Wafer: B = 0 M for both cases.
OHMS LAW, CONDUCTIVITY AND RESISTIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.1.
Up to now, our study of electromagnetism has concentrated on static
charge distributions (electrostatics) or steady current distributions (giving
rise to magnetostatics). Now we start looking in more detail at what hap-
pens when currents flow in arbitrary ways.
To begin, we can state the relation between electromagnetic force and
current density:

(1) J = (E + v B)

where is the conductivity of the material in which the current is flowing.


This is an empirical relation, based on observation, rather than a theoret-
ically derived result. Note that here is not surface charge density; it is
the standard notation for conductivity. Its inverse is = 1/ , called the
resistivity, and again should not be confused with volume charge density.
Typically, the velocity v of the charges is so small that the magnetic force
term can be neglected.
As an example, suppose we have two concentric conducting spheres of
radii a and b with b > a, held at a potential difference of V . The area
between the spheres is filled with a conducting material with conductivity
. Our problem is to determine the current that flows between the spheres.
The electric field between the spheres is due entirely to the inner sphere,
which well assume has a surface charge density of s. Then the field be-
tween the spheres is

4a2 s r a2 s r
(2) E= =
40 r2 0 r 2
Ignoring the magnetic term, the current density is

a2 s r
(3) J = E =
0 r 2
1
OHMS LAW, CONDUCTIVITY AND RESISTIVITY 2

and the total current between the spheres is


2 2
a2 s r sin 4a2 s
(4) I= J da = d d =
0 0 0 r2 0

where the integral is done over a sphere of radius r such that a < r < b.
We still need to get rid of the explicit reference to the surface charge
density s, which we can do by imposing the condition that the potential
difference between the spheres is V . We have
b
sa2 b dr sa2 1 1
 
sa (b a)
(5) V = E d` = 2
= =
a 0 a r 0 a b 0 b
Thus the surface charge density is

0 bV
(6) s=
a (b a)

and the current is

4ab
(7) I= V
ba
Note that the current is proportional to the potential difference, which is
often true in resistive materials and is known as Ohms law, which is more
usually written as

(8) V = IR

where R is the resistance of the material. In this case

ba
(9) R=
4ab
For large b, I 4aV and R 1/4a . This is presumably because
there is much more conducting material at a large radius, so it contributes
less to the total resistance between the spheres.
A setup used to measure the conductivity of sea water is to place two con-
ducting spheres each of radius a a large distance apart (say, at opposite ends
of a ship) in the water and pass current between them by holding one sphere
at potential zero and the other at potential V . (Im not 100% convinced of
this argument, but here goes arrangement as two instances of the concentric
OHMS LAW, CONDUCTIVITY AND RESISTIVITY 3

sphere setup in this problem. The sphere at potential V can be viewed as


concentric spheres with the one at a at potential V and a distant outer sphere
at potential V /2. The other sphere (at potential 0) can be viewed as an inner
sphere at potential 0 and an outer distant one at potential V /2. The currents
within both pair of spheres must be equal (since there is a constant overall
current passing from V to 0), so if we look at the first sphere, the current is
(taking b ):

V
(10) I = 4a = 2aV
2

which must also be the current passing into the second sphere. Thus the
conductivity of the sea water can be measured as

I
(11) =
2aV
P INGBACKS
Pingback: Charging and discharging a capacitor: applying Ohms law
Pingback: Charging and discharging a capacitor: general case
Pingback: Ohms law with variable conductivity
Pingback: Electromotive force
Pingback: Motional emf
Pingback: Maxwells equations in matter
Pingback: Current between two plates with a spherical perturbation
Pingback: Resistance in a conical capacitor
Pingback: Electric potential from a steady current
Pingback: Alfvens theorem
Pingback: Decay time for free charge in a conductor
Pingback: Skin depth of electromagnetic waves in conductors
Pingback: Reflection at a conducting surface: the physics of mirrors
CHARGING AND DISCHARGING A CAPACITOR: APPLYING
OHMS LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.2.
As a simple example of Ohms law, suppose we have a capacitor C with
an initial charge of Q0 connected in a circuit with a resistor R. The volt-
age across the capacitor is V (t) = Q (t) /C which must balance the voltage
across the resistor, so that

Q (t)
(1) = I (t) R = QR
C
The differential equation has the solution

(2) Q (t) = Q0 et/RC


(3) = CV0 et/RC

where V0 is the initial voltage across the capacitor. The current is then

V0 t/RC
(4) I (t) = e
R

where the minus sign indicates that the direction of the current is such as to
discharge the capacitor.
The initial energy stored in the capacitor is WC = 12 CV02 and the total
energy dissipated as heat in the resistor is

(5) WR = I 2 (t) Rdt
0

V02 2t/RC
(6) = e dt
R 0
1
(7) = CV02
2
Thus the two energies balance out.
1
CHARGING AND DISCHARGING A CAPACITOR: APPLYING OHMS LAW 2

Suppose we now start with an uncharged capacitor and connect it and the
resistor in series with a battery that is held at constant potential V0 . In this
case, the sum of the voltages across C and R must add up to V0 , so we have

Q (t)
(8) V0 = + QR
C
The solution of this differential equation with initial condition Q (0) = 0
is
 
(9) Q (t) = CV0 1 et/RC
V0 t/RC
(10) I (t) = Q (t) = e
R
This time, the total energy delivered by the battery is the sum of the en-
ergy stored in the capacitor and the energy dissipated as heat in the resistor.
Because the voltage is constant, the power generated is I (t)V0 and the en-
ergy from the battery is

(11) WB = I (t)V0 dt
0

V02
(12) = et/RC dt
R 0
(13) = CV02
Thus the battery delivers twice the energy that ends up stored in the ca-
pacitor. That it is independent of R isnt terribly surprising; with a larger
resistance it just takes longer to charge the capacitor, but the total energy
used to do it is the same.
P INGBACKS
Pingback: Charging and discharging a capacitor: general case
CHARGING AND DISCHARGING A CAPACITOR: GENERAL
CASE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.3.
The problem of charging and discharging a capacitor can be generalized
to two chunks of conducting material of arbitrary shape embedded within a
medium with conductivity . First, we can work out the resistance between
the two conductors. Suppose at some instant in time the potential difference
between the two conductors is V , and the current flowing between them is
I. This current is the surface integral of the current density over some area
that encloses one of the conductors; lets make it the conductor with a net
positive charge Q on its surface. That is


(1) I= J da

By our definition of conductivity, were taking J = E so


Q
(2) I= E da =
0

using Gausss law. The capacitance of the system is given by C = Q/V and
Ohms law says that V = IR so

V Q Q
(3) I = = =
R CR 0
0
(4) R =
C

Given the resistance between the conductors, the problem of finding the
charge, current and potential as functions of time reduces to that in the
previous post. So we have
1
CHARGING AND DISCHARGING A CAPACITOR: GENERAL CASE 2

(5) Q (t) = CV0 et/RC


(6) = CV0 et/0
CV0 t/0
(7) I (t) = Q (t) = e
0
RCV0 t/0
(8) V (t) = I (t) R = e = V0 et/0
0
Charging the capacitor with a battery of fixed potential V0 gives the same
results as in the previous post with RC = 0 / :
 
t/0
(9) Q (t) = CV0 1 e
V0 t/0 CV0 t/0
(10) I (t) = Q (t) = e = e
R 0
Q (t)  
(11) V (t) = = V0 1 et/0
C
The voltage across the resistor is VR (t) = V0 V (t) = V0 et/0 = I (t) R.
P INGBACKS
Pingback: Momentum in a capacitor
Pingback: Angular momentum in electromagnetic fields
OHMS LAW WITH VARIABLE CONDUCTIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.4.
Heres an example of using Ohms law to calculate the resistance of a
system with variable conductivity. We have a coaxial cylinder of length L
with an inner cylinder of radius a and an outer cylinder of radius b. The
conductivity is (r) = k/r where k is a constant. If the two cylinders are
held at a constant potential difference, then the current passing between
them is also constant I. We can get I by integrating over a cylindrical surface
between the two cylinders:

(1) I = (r) E da

By symmetry (ignoring end effects, or else considering a length L within


a much longer cylinder), E is radial and can depend at most on r, so if we
choose an integration cylinder of a fixed radius r then

k
(2) I = 2rEL = 2kLE
r
I
(3) E =
2kL
The field is therefore constant between the cylinders, which means the
potential difference between the cylinders is just

I
(4) V = E (b a) = (b a)
2kL

giving a resistance of (using Ohms law V = IR):

ba
(5) R=
2kL

1
ELECTROMOTIVE FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.5.
The electromotive force E (abbreviated emf) is the voltage produced by
a source of electric field, such as a battery or, as well see in a few posts, a
time-varying magnetic field. Although it is called a force it actually has
the dimensions of energy as it is defined as

(1) E = f d`

where f is the total force per unit charge and the integral is taken around
a closed loop (typically an electric circuit). The force can be produced
by a variety of things, such as chemical reactions (as in most household
batteries), solar power, etc. Any real battery has an internal resistance which
detracts from the effective voltage the battery can deliver. If this resistance
is r, then the net emf is E Ir where as usual I is the current flowing through
the circuit.
Suppose a battery with a constant emf and internal resistance is placed
in a circuit with a variable resistor R. The maximum power that can be
delivered by varying R can be found as follows. The voltage across R is
E Ir = IR (from Ohms law), and the power delivered to R is P = I 2 R, so

E
(2) I =
r+R
E 2R
(3) P =
(r + R)2
" #
dP 1 2R
(4) = E2 =0
dR (r + R)2 (r + R)3
2R
(5) 1 =
r+R
(6) R = r
Thus the maximum power is delivered when the resistance in the circuit
is equal to the batterys internal resistance.
1
ELECTROMOTIVE FORCE 2

P INGBACKS
Pingback: Perpetual motion in an electric circuit?
Pingback: Motional emf
Pingback: Emf in a spinning spherical shell
Pingback: Energy transfer in a solenoid
PERPETUAL MOTION IN AN ELECTRIC CIRCUIT?

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.6.
Suppose we try to build a perpetual motion machine as follows. We take
a parallel plate capacitor and charge it up so that the field between the plates
is E. Well take the plates to lie parallel to the xy plane, so that E points in
the z direction. Now we take a rectangular loop of wire and align it so that
it lies in the xz plane with one side (of length h) of the rectangle between
the plates, and the other end outside the plates. Thus between the plates, E
points along one edge of the rectangle and is perpendicular to the two edges
that lead into and out of the capacitor.
A naive calculation now tells us that the electromotive force in the loop
should be

E = E d` = Eh

since the field E is the force per unit charge, and E is parallel to the side of
the rectangle inside the plates, is perpendicular to the two sides leading into
and out of the plates, and the field is zero on the fourth edge since it lies
outside the plates. If the wire has a resistance R, this emf would therefore
give rise to a current I = E /R = Eh/R. Since the capacitor never discharges,
this current appears to be conjured out of nothing and would flow forever.
Clearly this is impossible, since it would result in a steady dissipation of
energy as heat from the resistor, which violates conservation of energy. The
resolution of the paradox lies in the fact that the field between the plates
is not exactly perpendicular to the plates, especially near the edges, and
there is also a fringe field that extends outside the plates. To see that this
gives an emf that is in the right direction to counter that generated by the
side of length h inside the plates, consider a point just outside the plates
on the x axis. If the bottom plate is positive and the top negative, then the
field at this point bends outwards from the lower plate and curves round
to bend back inwards towards the top plate. If the path of integration is
clockwise looking down on the xz plane, then E d` is negative on both the
lower and upper edges (since the angle between E and d` lies between 2
and in both cases), while E d` is positive on the vertical side of length
1
PERPETUAL MOTION IN AN ELECTRIC CIRCUIT? 2

h. Obviously wed need to do a complete analysis to actually prove that


these two contributions to the emf cancel
out (and thats not easy), but we
know they must since the integral E d` = 0 for all closed loops in an
electrostatic system (a consequence of Stokess law and the fact that
E = 0 in electrostatics).
P INGBACKS
Pingback: Faradays law
MOTIONAL EMF

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.7.
The most common type of electromotive force in practice is so-called mo-
tional emf, since the emf is generated by moving a wire through a magnetic
field. This is the principle used by all electric power stations: some exter-
nal force (hydro power to turn the generators, steam produced by burning
some fuel such as coal or gas, or by nuclear reactions, turns the generators,
wind power in wind turbines, and so on) is used to move the wires within a
magnetic field, and electric current is produced.
The simplest example of motional emf is a rectangular loop of wire with
one end inside a magnetic field and the other outside the field. If we pull
the loop at velocity v so that the edge inside the field moves perpendicular
to the field, then the Lorentz force law says that the force per unit charge is
fmag = v B and if we integrate this along the edge (of length `, say) we get

(1) E = fmag d` = `Bv

Only one edge contributes to the integral, since the upper and lower edges
are perpendicular to the force and the outside vertical edge is outside the
field. This motional emf then generates a current according to Ohms law:
I = E /R = `Bv/R, where R is the resistance in the wire.
However, once a current in the vertical direction starts flowing, there is
a vertical component of velocity for the charges, which adds a component
to the magnetic force. From the Lorentz force law, the force on the vertical
segment of wire carrying a current I is


`2 B2 v
(2) F =I |d` B| = I`B =
R

where the direction of F is given by the right hand rule and the direction of
the current, and will oppose the force pulling the wire through the magnetic
field. Note that F is the total force on the wire due to the magnetic field
acting on the overall current, while fmag is only the force per unit charge, so
1
MOTIONAL EMF 2

we cant combine the two to get the total force on the wire unless we know
the charge density within the wire.
Now suppose we have a loop composed of two parallel metal rails a dis-
tance ` apart connected at one end by a fixed wire containing a resistor R.
A metal bar (perpendicular to the rails) is free to slide along the rails to
form the other end of the loop. The bar starts off sliding at a velocity v0 . A
uniform magnetic field B is applied over the entire loop, facing downwards.
The motional emf generated by the bars motion is then

(3) E = `Bv0

at the start. However, there will be a horizontal force opposing the bars
motion as given above. When the bar has slowed to a speed v the force is

`2 B2 v dv
(4) F = =m
R dt

where m is the mass of the bar, and weve introduced a minus sign to indi-
cate that the bar is slowing down, so dv/dt < 0. The solution of this is

2 B2 /mR
(5) v (t) = v0 et`
Since the initial kinetic energy of the bar is 21 mv20 this energy should be
transferred as heat to the resistor by the time the bar comes to rest. We have

(6) W = I 2 Rdt
0

` B2 2
2
(7) = v (t) dt
R 0

`2 B2 v20 2t`2 B2 /mR
(8) = e dt
R 0
1 2
(9) = mv
2 0
P INGBACKS
Pingback: Emf near an infinite wire
Pingback: Magnetic flux
Pingback: Balancing magnetic force with gravity
Pingback: Dropping loops through magnetic fields
Pingback: Superconducting loop in a magnetic field
EMF NEAR AN INFINITE WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.8.
Heres a simple illustration of calculating motional emf from the change
in magnetic flux. An infinite straight wire lies along the x axis and carries a
current I in the +x direction. A square loop of wire of side length a lies in
the first quadrant of the xy plane with its sides parallel to the x and y axes,
and with the nearest edge a distance s from the infinite wire.
The magnetic field from the infinite wire is, at a distance r from the wire:

I0
(1) B=
2r

that is, it circles the wire in a clockwise direction looking towards the +x
direction. The magnetic flux is then

0 I s+a adr
(2) =
2 s r
0 Ia s + a
(3) = ln
2 s
If the loop is now moved away from the wire at a constant speed v, then
the emf is

d
(4) E =
dt
0 Ia s  a  ds
(5) = 2
2 s + a s dt
2
0 Ia v
(6) =
2s (s + a)
To work out the direction of the current, we note that from the right hand
rule, the magnetic force is in the +x direction on both the edge at distance
s and the edge at distance s + a but since the closer edge experiences the
stronger field, the net current is in the +x direction on the closer edge so the
current flows counterclockwise as seen from above the xy plane.
1
EMF NEAR AN INFINITE WIRE 2

If the loop is moved parallel to the wire (that is, keeping s constant),
there is a force along the two edges perpendicular to the wire, but because
the magnetic field at a given distance from the wire is always the same,
there is no net current and the generated emf is zero.
MAGNETIC FLUX

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.9.
When we introduced motional emf, we saw that for a simple rectangular
loop, the emf generated by pulling the loop through a magnetic field with a
speed v, assuming that one side (of length `) was inside the field region and
moved perpendicular to its length, was

(1) E = `Bv

If the distance from the edge of the rectangle generating the emf to the
edge of the field region is x, then as we pull the loop out of the field, the
distance x inside the field region decreases with speed v. That is, dx/dt =
v. Since ` remains constant, the area of the loop inside the field is A = `x
and its rate of change is dA/dt = ` dxdt = `v. We can therefore write the
emf as

dA
(2) E = B
dt
At this point, we can define the magnetic flux as the integral of the field
over the area enclosed by the loop:


(3) B da

In this particular case, with B constant, we have

d
(4) E =
dt
In fact, this is a general relation that applies for any shape of loop and
any magnetic field, even one that varies over space. The proof is given in
Griffiths, so we wont repeat it here, as it doesnt really add anything to our
understanding of the physics.
1
MAGNETIC FLUX 2

Note that we didnt specify the surface of integration in 3. In fact, the


integral is independent of the surface as we can see by the following ar-
gument. Suppose we consider a closed surface (such as a sphere, but any
closed surface will do). Then by Gausss theorem

(5) B da = Bd 3 r = 0

since B = 0 always. Now suppose we divide this closed surface into two
separate surfaces (labelled surface 1 and surface 2) by drawing some closed
curve around the closed surface. Then we must have

(6) B da = B da
1 2
Now we can keep the curve fixed, and also keep one of the
surfaces (say,
surface 2) fixed, while varying surface 1. Its clear that 1 B da has to
remain the same no matter what we define surface 1 to be, provided it has
the same boundary curve. Thus the independence of the integral on the
bounding curve is a consequence of B being divergenceless.
P INGBACKS
Pingback: Emf near an infinite wire
Pingback: Alternating current
Pingback: Faradays law
ALTERNATING CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.10.
The alternating current common in households can be generated by ro-
tating a loop of wire in a constant magnetic field. Suppose we have a square
loop of side length a rotating at a constant angular velocity within a con-
stant magnetic field B which is perpendicular to . Then the magnetic flux
is given by

(t) = B da = Ba2 cos t
The emf is then
d
E = = Ba2 sin t
dt
The emf thus alternates between positive and negative with period 2/.
A power station uses some external force to rotate the wire and provides the
magnetic field, with the result that electricity is generated.

1
BALANCING MAGNETIC FORCE WITH GRAVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.11.
Heres an illustration of how a magnetic force can balance gravity, even
for a non-magnetic substance. We have a square loop of aluminum (or
aluminium, if you must) with a constant cross-sectional area and side
length a. It is dropped from rest between the poles of a strong magnet with
a field strength of B = 1 Tesla, so that the field is horizontal, and the top
edge of the square is within the field, while the bottom edge is outside the
field. As weve seen in a similar problem, once the square has a non-zero
downwards velocity v, a magnetic force arises that opposes the fall (so it
acts upwards), and that force has magnitude

a2 B2 v
(1) Fmag =
R

where R is the resistance of the aluminum loop. The net force on the loop
is then

dv a2 B2 v
(2) m = mg
dt R

where m is the mass of the loop and g is the acceleration of gravity.


The system reaches terminal velocity when dv/dt = 0, so

mgR
(3) vterm =
a2 B2
To reduce this to physically measurable quantities we first note that the
resistance of a wire of uniform cross section is given by Griffiths in his
example 7.1 and is

L 4a
(4) R= =

1
BALANCING MAGNETIC FORCE WITH GRAVITY 2

where is the resistivity of the wire and L is its length. The mass of the
wire is

(5) m = Ld = 4ad

where d is the mass density of the wire. Therefore

16gd
(6) vterm =
B2
For aluminum, the values are

(7) = 2.82 108 ohm m


(8) d = 2700 kg m3
(9) g = 9.8 m s2

Plugging in the numbers gives

(10) vterm = 0.0119 m s1

By the way, to check that the units are correct its worth noting that

(11) 1 ohm = 1 kg m2 s3 Amp2


(12) 1 Tesla = 1 kg s2 Amp1

Plugging these into the formula for vterm does indeed give units of veloc-
ity.
We can work out the time required to achieve 90% of terminal velocity
by solving the differential equation 2. The solution turns out to be

mgR  ta2 B2 /Rm



(13) v (t) = 1 e
a2 B2 
(14) = vterm 1 egt/vterm

The 90% time is reached when


BALANCING MAGNETIC FORCE WITH GRAVITY 3

v (t90 )
(15) = 0.9
vterm
(16) egt90 /vterm = 0.1
vterm
(17) t90 = ln 0.1
g
(18) = 2.8 103 s
Thus when a square loop of a conductor (since most conductors have
similarly small resistivities) is dropped in a strong magnetic field, it will
drift down at a constant speed almost immediately.
If the loop were cut, no current could flow, so there would be no magnetic
force to oppose gravity, and the loop would fall at the normal rate.
P INGBACKS
Pingback: Dropping loops through magnetic fields
FARADAYS LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.12.
The relation between emf and change of magnetic flux turns out to be
a special case of a more general law discovered by Michael Faraday and
called Faradays law. It turns out that any change of magnetic flux through
a loop, whatever the cause, results in an emf being generated around the
loop. This can be caused by moving the loop relative to a fixed magnet,
moving the magnet relative to a fixed loop, or keeping both loop and magnet
fixed and varying the field strength.
Actually, it may seem surprising that the first two cases are treated sep-
arately; surely the relative motion of loop and magnet means they are both
the same? Special relativity would, of course, confirm this, and it was partly
this aspect of electromagnetism that inspired Einstein to think about rela-
tive motion. However, in the 19th century, motion was always considered
relative to a fixed reference frame so the two cases were quite different. In
particular, if the loop is considered fixed and the magnet moves, then the
charges in the loop are at rest, so should not feel any magnetic force from
the Lorentz force law. The fact that an emf is generated in this case as well
led Faraday to postulate that a changing magnetic field produces an electric
field. Faradays law states that the emf generated in a loop is minus the rate
of change of magnetic flux through the loop. That is

d
(1) E = E d` =
dt

Since = B da then if the area enclosed by the loop stays the same,
we have

d B
(2) E d` = = da
dt t
From Stokess theorem, we have

B
(3) E d` = ( E) da = da
t
1
FARADAYS LAW 2

so in differential form, we have

B
(4) E =
t
This is the generalization of the electrostatic condition E = 0 which
applied in the absence of magnetic fields.
As a simple example, suppose we have solenoid in which the current in
the wire wrapped around it varies with time according to I (t) = I0 cos t,
so that the magnetic field inside the solenoid is

(5) B (t) = zB0 cos t


A circular loop of wire with resistance R and radius a/2 is placed inside
the solenoid so that its axis is coincident with the solenoids axis. The flux
through the loop is then

a2 B0
(6) (t) = cos t
4

so the emf generated in the loop is

a2 B0
(7) E = sin t
4

and the current is

E a2 B0
(8) I= = sin t
R 4R
P INGBACKS
Pingback: Faraday field and magnetic vector potential
Pingback: Lenzs law
Pingback: Magnet falling through a metal pipe
Pingback: Faradays law, Ampres law and the quasistatic approxima-
tion
Pingback: Coaxial cable with varying current
Pingback: Faradays law: wire loop encircling a solenoid
Pingback: Faradays law: cutting a current-carrying wire
Pingback: Faradays law and the Biot-Savart law
Pingback: Mutual inductance
FARADAYS LAW 3

Pingback: Faraday field and magnetic vector potential


Pingback: Changing a particles speed in a cyclotron
Pingback: Maxwells equations with varying charge but constant current
Pingback: Alfvens theorem
Pingback: Poyntings theorem
Pingback: Momentum in a capacitor
Pingback: Angular momentum in a magnetized sphere
Pingback: Angular momentum conservation: example with a solenoid
Pingback: Rectangular wave guides: transverse electric waves
Pingback: Magnetic systems in thermodynamics
LENZS LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.13.
Faradays law states that the emf generated in a circuit is determined by
the rate of change of magnetic flux through that circuit, according to

d
(1) E = E d` =
dt

Since = B da then if the area enclosed by the loop stays the same,
we have

d B
(2) E d` = = da
dt t
Although the direction of the induced emf can be determined by the
Lorentz force law in cases where the loop moves relative to the magnetic
field, this doesnt work if both loop and magnet are at rest and only the field
strength changes. A simple criterion known as Lenzs law can be used in
all cases to determine the direction of the emf. Lenzs law states that the
induced emf always tends to oppose the flux change that produces it.
As a simple example, suppose we have a square loop of side length a in
the first quadrant of the xy plane, with one corner at the origin and two sides
lying along the x and y axes. A magnetic field B = ky3t 2 z (where k is a
positive constant) is applied to the loop. The flux through the loop is then
a
2 1
(3) = kt a y3 dy = kt 2 a5
0 4

and the emf is

d 1
(4) E = = kta5
dt 2
Since the magnetic field monotonically increases with time, the flux is
also increasing and is positive (where the positive direction is taken to be the
+z direction). By Lenzs law, the emf must oppose this flux so the current
1
LENZS LAW 2

must generate a field that points downward, which means the current flows
clockwise around the loop as seen from above.
P INGBACKS
Pingback: Magnet falling through a metal pipe
Pingback: Coaxial cable with varying current
Pingback: Faradays law: wire loop encircling a solenoid
Pingback: A bizarre example of Faradays law
Pingback: Angular momentum conservation: example with a solenoid
MAGNET FALLING THROUGH A METAL PIPE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.14.
You might think that if you dropped a magnet down a pipe made of some
non-magnetic metal such as aluminum or copper, that the magnet would just
fall straight through at the same rate as if it were dropped in air. However, if
we think about the situation using Faradays law and Lenzs law we can get
an idea of what really happens. Faradays law states that the emf generated
in a circuit is determined by the rate of change of magnetic flux through that
circuit, according to

d
E =
dt

and Lenzs law states that the direction of the induced current is such that
its magnetic field opposes the change in flux.
Consider a thin cross-sectional slice through the pipe; this essentially
constitutes a circular loop of wire. Before the magnet is dropped into the
pipe, the magnetic flux through all such slices is essentially zero. As the
magnet falls through the pipe, the flux in a given slice increases as the lead-
ing edge of the magnet approaches, so a current is induced in the slice. The
magnetic field due to this current attempts to oppose the flux change, so it
produces a flux opposing that from the magnet. In terms of the magnetic
field lines, if the field emanating from the magnet points downwards at the
leading edge (it also spreads out, but the overall effect is a downward point-
ing field), the induced current produces field lines pointing upwards. Thus
the two fields repel each other, slowing the magnets descent.
As the mid-section of the magnet passes a slice, the flux remains essen-
tially constant, since the field inside the magnet is pretty well constant (as
in the interior of a solenoid), so no current is induced in slices as the mid-
section of the magnet passes.
When the trailing end passes, the flux decreases, so the induced current
attempts to counter this by providing a field that increases the flux. The field
due to the magnet points into the trailing edge, so still points downwards.
Since this flux is decreasing, the induced current produces a field in the
same direction as that from the magnet, so the direction of the current is
1
MAGNET FALLING THROUGH A METAL PIPE 2

opposite to that at the leading edge. This time, the two fields attract, so the
force again reduces the speed of the magnet as it falls. Thus the two induced
currents at either end of the magnet both slow the magnet down.
FARADAYS LAW, AMPRES LAW AND THE QUASISTATIC
APPROXIMATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.15.
Faradays lawin differential form is

B
(1) E =
t
There is a similarity to Ampres law, which says

(2) B = 0 J
Considering only electric fields generated from changing magnetic fields
(and not those generated by free charges), we then have E = 0, since
there is no free charge. For magnetic fields, B = 0 always. Once we
specify both the curl and divergence of a vector field, the field is deter-
mined uniquely (up to a constant), so Faradays law is formally equivalent
to Ampres law, except that curl is determined by tB instead of 0 J. In
particular, we can use the right hand rule to determine the direction of E
if we know tB and we can use Amprian loops to calculate E in those
problems where the symmetry makes the calculation easier.
There is one important difference between Faradays and Ampres law,
however. Ampres law assumes that the currents J are steady, that is, that
there is no dependence on time. Faradays law clearly does depend on time,
in that an electric field is generated only if the magnetic field is changing.
In differential form, this doesnt pose a problem, since both sides of the
equation refer to a single point (x, y, z). However, in its original integral
form:

B
(3) E d` = da
t

the t variable is assumed to be the same at all points in the integrals. If we


choose some enormous loop for the integral on the left, then any change in
B, even one in some small, remote corner of the area enclosed by the loop,
is implicitly assumed to affect E instantaneously around the entire loop.
1
FARADAYS LAW, AMPRES LAW AND THE QUASISTATIC APPROXIMATION 2

This is a problem inherent in all non-relativistic physics. In Newtons


gravitational theory, for example, no provision is made for any travel time
from one mass to the other; if the sun were to suddenly lose half its mass,
say, the effect would be felt at the Earth immediately. In reality, of course,
nothing can travel faster than the speed of light, so changes in one part of a
system will not be felt at other parts until some signal informing these re-
mote parts of the change has reached them. In the case of electromagnetism,
the signal speed happens to be exactly that of light, so when we apply Fara-
days law in integral form, we really need to take this into account.
In practice, when were dealing with finite electrical circuits or situations
within an Earth-bound laboratory, the distances are usually so short that we
can make the approximation that the travel time is zero. This is known as
the quasistatic approximation.
As a simple example, suppose we have an infinite solenoid with n turns
per unit length and of radius a, carrying a time-dependent current I (t). The
fact that the solenoid is infinite means that the quasistatic approximation
could well break down for large distances, but well do the calculation any-
way and see what we get.
Inside the solenoid, the field is

(4) B = n0 I (t) z

so

B
(5) = n0 I (t) z
t
Using the analogy between Ampres and Faradays laws, since the direc-
tion of tB is z, we can apply the right-hand rule to find that the induced
electric field is circumferential and in the direction (that is, clockwise,
looking towards z). The magnitude of E is obtained by integrating the
field around a circle of radius r < a:

B
(6) E d` = 2rE = da = r2 n0 I (t)
t

1
(7) E = rn0 I (t)
2
Outside the solenoid, B is always zero, so there is no contribution to E
here. A circular integration path at a distance r > a still contains the flux
inside the solenoid, so
FARADAYS LAW, AMPRES LAW AND THE QUASISTATIC APPROXIMATION 3


B
(8) E d` = 2rE = da = a2 n0 I (t)
t

1 a2
(9) E= n0 I (t)
2 r
P INGBACKS
Pingback: Coaxial cable with varying current
Pingback: Faradays law: wire loop encircling a solenoid
Pingback: Faradays law: cutting a current-carrying wire
Pingback: Faradays law and the Biot-Savart law
Pingback: Induction between a wire and a torus
Pingback: Angular momentum in electromagnetic fields
Pingback: Jefimenkos equation for time-dependent magnetic field
COAXIAL CABLE WITH VARYING CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.16.
Heres another (mis)application of Faradays law, using the quasistatic
approximation to calculate the fields involved. This time we have a coaxial
cable in which the current flows in the +z direction on the inner wire and
back in the z direction on the coaxial cylinder. The current varies with
time according to

(1) I (t) = I0 cos t


The magnetic field due to the inner wire is given by

0 I0
(2) Bw (r,t) = cos t
2r
The magnetic field due to the coaxial cylinder is just the negative of this
outside the cylinder, and zero inside:

(
0 I0
2r cos t r>a
(3) Bc (r,t) =
0 r<a

Therefore the net magnetic field outside the cylinder is always zero, and
the field inside is given by Bw alone. The induced electric field is parallel
to the z axis, as can be seen by a similar argument to that given for the
magnetic field of a solenoid. First, E cannot have a radial component, since
if we reversed the directions of the currents, then B and hence E reverse
their directions. But reversing the currents is equivalent to turning the wire
through 180 degrees, and that shouldnt affect the radial component of the
field, so Er = 0.
E also cannot have a component. Since B is entirely in the direc-
tion, so is B/t, so if we choose a path of integration that is a circle at
constant r, and a corresponding surface of integration that is perpendicular
to the z direction (that is, to the cable), then da is always perpendicular to
B/t so the integral on the right is zero:
1
COAXIAL CABLE WITH VARYING CURRENT 2


B
(4) E d` = da
t

By symmetry, E must be constant around the circle so we get E d` =
2rE = 0, so E = 0. Therefore, E must have a z component only.
To figure out the actual value of E, we can use the same argument as
in calculating the solenoids magnetic field. We choose a square loop with
side length ` that lies in the yz plane. First, we take the loop to lie entirely
outside the coaxial cable, where B = 0 always. This means that (taking the
near edge to be a distance s from the cables axis):

(5) E d` = ` (E (s + `) E (s)) = 0

If we assume that E 0 at infinity, then E = 0 everywhere outside the


cable.
To find E inside the cable, we can take the loop of integration to have its
inner edge at distance s from the axis, where 0 < s < a, and the outer edge
outside the cylinder, where E = 0. Then

(6) E d` = `E (s)

B
(7) = da
t
a
` 0 I0 dr
(8) = sin t
2 s r
` 0 I0 a
(9) = sin t ln
2 s
To get the direction of E we use Lenzs law and the right hand rule. If
the current flows in the +z direction and is increasing, then the magnetic
field is in the + direction and increasing, so the flux through the loop
is increasing. The induced electric field will oppose this increase, which
means that E inside the cylinder must point in the z direction. Conversely,
if I is decreasing, E points in the +z direction. Since I starts off at its
maximum value at t = 0, the current is decreasing initially, so the initial
value of E must point in the +z direction, thus:
(
0 I0
z 2 sin t ln as 0<s<a
(10) E (s,t) =
0 s>a
COAXIAL CABLE WITH VARYING CURRENT 3

As always, we need to issue a caution that this analysis is deeply flawed,


since were applying the quasistatic approximation to a cable of infinite
length, and were assuming that the current can change everywhere along
the length of the cable at the same rate, which doesnt happen in real life.
P INGBACKS
Pingback: Displacement current is very small
FARADAYS LAW: WIRE LOOP ENCIRCLING A SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.17.
Heres another example of Faradays law, using the quasistatic approxi-
mation involving a solenoid. A solenoid of radius a with its axis along the
z axis has n turns per unit length carrying current in the + direction. The
current I is increasing at a constant rate, such that I = k. A loop of wire with
resistance R encircles the solenoid and we want to find the induced current
IR in the wire.
By Faradays law, the emf around the loop depends only on change in
flux enclosed by the loop (note that it doesnt depend on the shape of the
loop). Thus we have

d
(1) E = = a2 0 nk
dt
E a2 0 nk
(2) IR = =
R R
Since Lenzs law tells us the induced current opposes the change in flux,
the current in the wire flows in the direction.
Now suppose the current in the solenoid is held constant, and the solenoid
is removed from the loop and reinserted in the opposite direction. How
much charge flows through the resistor in total during this process? Charge
is the integral of current over time, so it might appear that we would need to
know the rate at which the solenoid is moved out and replaced, but in fact
we dont. Consider the stage where the solenoid is removed from the loop.
The charge through the resistor due to this stage is

1 d 1
(3) Q= IR (t) dt = =
R dt R

where is the change in flux before and after the solenoid is removed.
That is, the details of how fast the solenoid is moved dont matter; all that
matters is the start and end values of the flux. But the start value of the flux
is s = a2 0 nI and the end value is e = 0, so the magnitude of charge
moved is
1
FARADAYS LAW: WIRE LOOP ENCIRCLING A SOLENOID 2

a2 0 nI
(4) Q=
R
The process of reinserting the solenoid the other way round will move
the same amount of charge, so the total charge through the resistor is

2a2 0 nI
(5) QT =
R
P INGBACKS
Pingback: Faradays law: cutting a current-carrying wire
FARADAYS LAW: CUTTING A CURRENT-CARRYING WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.18.
And yet another example of Faradays law, using the quasistatic approx-
imation in a situation similar to the last problem. This time we have an
infinite wire carrying a current I in the +z direction and a square wire loop
of side length a in the xz plane, with its nearest edge parallel to and at a
distance s from the wire. The wire is suddenly cut so that the current drops
to zero. What can we say about the current and charge flowing through the
loop?
The magnetic field produced by the wire is

0 I (t)
(1) B (r,t) =
2r

so the field points in the +y direction through the loop. If we take the area
element to also point in the +y direction, then the flux is
s+a
0 aI (t) dr 0 aI (t) s + a
(2) (t) = B da = = ln
2 s r 2 s

where I (t) is the current as a function of time; the current will drop sharply
to zero when the wire is cut, but of course it wont be a step function in real
life.
When the wire is cut, the flux will suddenly decrease to zero, so the in-
duced current will oppose this, and the current flows in a counterclockwise
direction (that is, the current in the side of the square closest to the wire will
be in the +z direction). The actual magnitude of the current depends on the
resistance R in the wire and how fast the current drops to zero, so we cant
specify those with the information given.
The total charge that moves through the loop is, as in the previous exam-
ple, determined from the flux difference before and after the cut.

1 0 aI s + a
(3) |Q| = = ln
R 2R s

1
FARADAYS LAW AND THE BIOT-SAVART LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.19.
As an example of the analogy between Faradays law and Ampres law,
suppose we have a toroidal solenoid with inner radius a, width w and height
h, where both w and h are much less than a. The solenoid has N turns in
total, and carries a current I that is increasing at the constant rate of I = k.
The magnetic field inside the solenoid is given by Griffiths in his example
5.10 and is

0 NI
(1) B=
2r

where r is the radial distance. In our case we can approximate r by the


constant a since the torus is very thin compared to its radius, so we get

0 NI
(2) B
2a
0 NIhw
(3)
2a
With this approximation, the change in flux is

d 0 Nkhw
(4) =
dt 2a
Since the magnetostatic case and the Faraday case are mathematically
equivalent (provided there is no free charge), we can use the Biot-Savart
law to calculate the electric field generated by a steady change in magnetic
flux if we replace J by 10 tB , or, in the case of a linear current in the
magnetostatic case, we replace I by 10 t .
For example, we can find the electric field at a point on the axis of the
torus by using the Biot-Savart law:

1 d`0 (r r0 )
(5) E=
4 t |r r0 |3
1
FARADAYS LAW AND THE BIOT-SAVART LAW 2

Griffiths works this out in his example 5.6, and the derivation goes like
this: the line segment d`0 is always in the xy plane and is tangent to the torus.
The vector r r0 points from a point on the torus to a point on the z axis, so
it is always perpendicular to d`0 . The cross product d`0 (r r0 ) makes a
constant angle with the z axis, and is also the angle between r r0 and
the xy plane. By symmetry, any x and y components of the cross product
will cancel out in the integration, so were left with only the z component,
which is |r r0 | cos d`.
Furthermore, the magnitude
of r r0 is always the
same, and is |r r0 | = a2 + z2 , and cos = a/ a2 + z2 , so

d`0 (r r0 ) a2 + z2 2a2
(6) = 2a cos z = z
|r r0 |3 (a2 + z2 )
3/2
(a2 + z2 )
3/2

and the electric field is then

a2 0 Nkhw a
(7) E= 3/2
z = z
t 2 (a2 + z2 ) 4 (a2 + z2 )3/2

P INGBACKS
Pingback: Mutual inductance
MUTUAL INDUCTANCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.20.
One consequence of Faradays law is that if we have two separate circuits
in close proximity to each other, a varying current in one circuit will induce
an emf in the other circuit, and vice versa. This is the principle used in
the electrical transformers commonly found on power cords for household
electronic devices. This phenomenon is known as induction. Well begin
by looking at the magnetic flux through one circuit by a steady current in
the other.
The Biot-Savart law says that the magnetic field produced by a steady
linear current I1 is

0 I1 d`1 (r r1 )
(1) B1 (r) =
4 |r r1 |3

where the integral is taken around circuit number 1. For any given circuit,
this integral will probably be very difficult to work out, but the important
thing is that, for a steady current, B1 is proportional to that current. This
means that the flux through the other circuit due to this field is also propor-
tional to I1 , since the flux through circuit number 2 due to B1 is

(2) 21 = B1 da2

where the integral is over the area enclosed by circuit 2.


From the Biot-Savart law and the flux equation we can see that apart from
the proportionality to I1 , everything else (well, apart from the constant 0 )
that goes into the formula for flux depends only on the geometry of the two
loops. This everything else is called the mutual inductance M21 , that is

(3) 21 = M21 I1

with
1
MUTUAL INDUCTANCE 2

" #
0 d`1 (r r1 )
(4) M21 = da2
4 |r r1 |3
The curious thing is that the mutual inductance of circuit 1 relative to
circuit 2 is the same, that is

(5) M12 = M21


This isnt exactly obvious from the double integral above, but we can see
it as follows. We can write 2 in terms of the magnetic vector potential as
follows:

(6) 21 = ( A1 ) da2

Using Stokess theorem, this becomes a line integral:



(7) ( A1 ) da2 = A1 d`2

From our discussion of the vector potential, we have the formula:



0 I1 d`1
(8) A1 (r2 ) =
4 |r2 r1 |

so

0 I1 d`1 d`2
(9) 21 =
4 |r2 r1 |
From this formula, its obvious that if the same current I1 flows in either
circuit, the flux through the other circuit is the same, since the integrand is
identical under interchange of the subscripts 1 and 2.
As one example that this is true, suppose we have a circular loop of radius
b in the xy plane with its centre at the origin, and a second loop of radius
a  b parallel to the first with its centre on the z axis a distance z above
the first loop. If a current I flows through the larger loop, what is the flux
through the smaller one?
The magnetic field at a point on the z axis above a current loop is given by
Griffiths in his example 5.6 (for a similar derivation involving the electric
field produced by a toroidal solenoid, see here), and is
MUTUAL INDUCTANCE 3

0 I b2
(10) B1 =
2 (b2 + z2 )3/2

Since a  b, we can take B1 to be approximately constant over the


smaller loop, and so the flux through it is

0 I a2 b2
(11) 21 = B1 a2 =
2 (b2 + z2 )3/2

Now suppose the same current flows through the smaller loop. What is
the flux through the larger one? This time we cant take the field at the
larger loop to be constant, but since the smaller loop is really small, we can
approximate it by a magnetic dipole, of dipole moment m = Ia2 z. The
formula for the field due to a dipole is

0
(12) B2 = [3 (m r) r m]
4r3
To get the flux, we need to integrate this over the area enclosed by the
larger loop. The distance r is the distance from the upper dipole to a point
in the area enclosed by the lower loop. If this point is at radial distance s
from the centre of the loop, and the two loops are a distance z apart, then

p
(13) r= s2 + z2

We can also define as the angle between z and the radial vector r. Then

z
(14) m r = Ia2 z r = Ia2 cos = Ia2
s2 + z2
Therefore the field is

0 a2 I
 
3z
(15) B2 = 3/2
r z
4 (s2 + z2 ) s2 + z2

Now if we integrate this over the area of circuit b, only the component of
r in the z direction will contribute, as the other components cancel out due
to symmetry. Thus well get another factor of z r so
MUTUAL INDUCTANCE 4

"
b b #
0 a2 I 2 2sds 2sds
(16) 12 = B2 da = 3z 5/2
3/2
4 0 (s2 + z2 ) 0 (s2 + z2 )
b
2 2
b
0 a I z 1
(17) = + 2 2

2 3/2
(s2 + z2 ) 0 s +z 0
b
2 2 2
0 a I z + s + z 2
(18) = 3/2
2

(s2 + z2 ) 0

0 I a2 b2
(19) = = 21
2 (b2 + z2 )3/2
Thus the two fluxes are indeed the same. The mutual inductance is there-
fore the same both ways, and we can drop the subscripts and just refer to it
as M:

0 a2 b2
(20) M=
2 (b2 + z2 )3/2

P INGBACKS
Pingback: Mutual inductance between two loops
Pingback: Self-inductance
Pingback: Mutual inductance between two magnetic dipoles
Pingback: Mutual inductance: calculation from the integral
MUTUAL INDUCTANCE BETWEEN TWO LOOPS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.21.
Heres a simple example of how the equality of mutual inductance be-
tween two circuits can make problem solving a bit easier. Suppose we have
a small, square loop of side length a situated midway between two very
long wires a distance 3a apart. The two long wires are connected at both
ends to form a circuit, but well assume that these ends are far enough away
that they dont influence what happens in the small square.
If the small square has a current that is increasing at a constant rate of
I = k > 0, what emf is induced in the large loop?
Trying to work out the field generated by the small loop and calculate
from that the change in flux in the large loop would be very difficult, so we
can invert the problem by supposing that the current flows in the large loop
and from that obtain the mutual inductance M.
The magnetic field produced by each of the long wires is

0 I (t)
(1) B (r,t) =
2r

where the direction is obtained by using the right-hand rule and the direc-
tion of current, and r is the distance from the wire. Since the little square is
midway between the two wires, each wire contributes the same flux, and in
the same direction (since the current in each of the two wires is opposite to
that in the other wire), so since the little square spans a distance between a
and 2a from each wire:
2a
0 I (t) adr 0 aI (t)
(2) (t) = 2 = ln 2
2 a r
The mutual inductance is then

0 a
(3) M= = ln 2
I

and the induced emf is


1
MUTUAL INDUCTANCE BETWEEN TWO LOOPS 2

d dI 0 ak
(4) E = = M = ln 2
dt dt
This is also the emf induced in the large loop by a changing current in the
small square. If the current in the small square is flowing clockwise, then
the induced current in the large loop opposes the increase of current in the
small square, so it will be counterclockwise.
P INGBACKS
Pingback: Self-inductance of a long rectangle
SELF-INDUCTANCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.22.
In addition to mutual inductance between two different circuits, a sin-
gle circuit also has self-inductance, or just inductance. The flux through a
circuit is

(1) = B da

and for a steady current I, this is proportional to I:

(2) = LI

where L is the inductance of the circuit. The emf induced in a circuit by


changing the current in that circuit is then

d
(3) E = = LI
dt
The inductance (and mutual inductance) are measured in a unit called
the henry, named after Joseph Henry, who is generally acknowledged to
be a co-discoverer of inductance around the same time as Michael Faraday
(Faraday, of course, has been immortalized in the farad, the unit of capaci-
tance).
As a simple example, lets work out the inductance of a long solenoid
with n turns per unit length. The magnetic field of an infinite solenoid is

(4) B = 0 nI

so the flux through each turn of the solenoid is, if the solenoids radius is R

(5) 1 = R2 0 nI
The inductance per turn is thus
1
SELF-INDUCTANCE 2

(6) L1 = R2 0 n

and the inductance per unit length is

(7) L = R2 0 n2
P INGBACKS
Pingback: Self-inductance of a long rectangle
Pingback: Induction between a wire and a torus
Pingback: LC circuit - the oscillator
Pingback: Energy in a magnetic field
Pingback: LR circuit
Pingback: Superconducting loop in a magnetic field
Pingback: Transformers
Pingback: Transmission lines
SELF-INDUCTANCE OF A LONG RECTANGLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.23.
Well return to the earlier problem in which a small loop was enclosed
between a larger circuit consisting essentially of two very long wires a dis-
tance d apart, connected at the far ends. This time, well remove the small
loop and try to find the self-inductance of the large loop. The magnetic field
produced by each of the long wires is

0 I
(1) B =
2r

where the direction is obtained by using the right-hand rule and the di-
rection of current, and
r is the distance from the wire. Since the flux from
each wire is w = B da = LI it would seem that all we need to do is a
simple integral to get the inductance. The problem, however, arises in the
limits of the integral. The total flux is twice that due to each wire, so for a
unit length, we get

0 I d dr
(2) =2
2 0 r
The integral gives a logarithm, which is infinite at the lower limit of 0. In
reality, of course, no wire is infinitely thin, so if we give the wire a radius
of , then we get
d
0 I dr 0 I d
(3) =2 = ln
2 r

so the inductance per unit length is

0 d
(4) L= ln

1
INDUCTION BETWEEN A WIRE AND A TORUS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.24.
An interesting inductance problem consists of an infinite straight wire
carrying an alternating current I (t) = I0 cos t. This wire lies along the
axis of a toroidal solenoid of rectangular cross-section, with inner radius a,
outer radius b and height h, and with a total of N turns. The solenoid is
connected in series with a resistor R. What emf is induced in the solenoid,
and thus what current flows through the resistor?
The magnetic field due to this current is (using the quasistatic approxi-
mation, in which the current changes everywhere along the wire at exactly
the same rate):

0 I
(1) B =
2r
The flux per turn in the solenoid is then

h0 I b dr h0 I b
(2) 1 = = ln
2 a r 2 a
The total flux through the torus is then

h0 IN b
(3) = ln
2 a
The emf induced in the solenoid is

d h0 IN b h0 N b
(4) E = = ln = I0 sin t ln
dt 2 a 2 a
The current through the resistor is then

E h0 N b
(5) Ir = = I0 sin t ln
R 2R a
Putting in some actual numbers, we set a = 1 cm, b = 2 cm, h = .01 m,
N = 1000, I0 = 0.5 A, = 60 Hz = 2 60 s1 and 0 = 4 107 N
A2 we get
1
INDUCTION BETWEEN A WIRE AND A TORUS 2

(6) E = 2.613 104 sin t Volts


If the resistor is R = 500 the current through it is

2.613 104 sin t


(7) Ir = = 5.23 107 sin t Amp
500
In other words, the voltage and current involved are very small.
The changing current within the solenoid generates a back emf because of
the solenoids self-inductance. In Griffithss example 7.11, the inductance
of a toroidal solenoid is given as

h0 N 2 b
(8) L= ln = 1.386 103 henries
2 a
The flux through the solenoid due to the induced current is

(9) b = LIr = 7.25 1010 sin t

so the back emf is

db
(10) Eb = = 2.73 107 cos t V
dt
The ratio of back emf to forward emf is then

Eb
(11) = 1.046 103
E
Thus the back emf is only a small fraction of the original induced emf.
(For those keeping track of such things, this is post #666.)
P INGBACKS
Pingback: Detecting magnetic monopoles
LC CIRCUIT - THE OSCILLATOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.25.
Another inductance problem is that of the oscillator circuit consisting of
a capacitor and an inductor, with a switch in the circuit that can be opened
or closed. The capacitor is charged initially to a potential V0 and then the
switch is closed. What happens?
Its important to be clear about the directions (and thus the signs) of the
charge and current in this problem. To be definite, suppose we draw the
circuit as a square with the inductor on the left edge and the capacitor on
the right. The capacitor starts off with a charge +Q on its lower plate, and
we define the positive direction of current to be clockwise. When the switch
is closed, the current starts to flow from the positive to the negative plate of
the capacitor, so the initial current is clockwise (positive) and increasing.
However, the charge on the capacitor is decreasing, so Q < 0. Since the rate
of change of charge is (in magnitude) equal to the current, we must have

(1) I = Q
The back emf provided by the inductor must oppose this current, so since
I > 0 initially, we have

(2) EL = LI = +LQ

and the total emf in the circuit is the sum of the contributions from the
inductor and capacitor, or

Q
(3) E = EC + EL = + LQ
C
Since there is no resistor in the circuit, the total emf around the circuit
must be zero, so we get the differential equation for Q:

Q
(4) + LQ = 0
C
1
LC CIRCUIT - THE OSCILLATOR 2

Note that we should always get the same ODE no matter what directions
we specify for the current. For example, if we had defined positive current
as in the counterclockwise direction, and started off with the capacitor in the
same configuration as before, then initially wed have I = +Q. However,
this time I (0) < 0 since the current is becoming more negative (its still
increasing in magnitude but in the negative direction), so this time EL = +LI
and we get the same ODE as before.
The ODE can be written in the more familiar form

1
(5) Q = Q 2 Q
LC

where = 1/ LC. This has the general solution

(6) Q (t) = A sin t + B cos t


(7) I (t) = Q = A cos t + B sin t

with initial conditions

(8) Q (0) = CV0


(9) Q (0) = I (0) = 0

from which we get A = 0 and B = CV0 . Thus the current is

r
C t
(10) I (t) = V0 sin
L LC

Thus the current oscillates with angular frequency 1/ LC.
If we add a resistor R in series with L and C, then the combined emf of L
and C produces a voltage of IR = QR and the ODE now becomes

Q
(11) + LQ = QR
C
Q
(12) LQ + RQ + = 0
C
We can get the solution to this ODE by proposing Q = Aet and substitut-
ing. We get, after cancelling off the factors of Aet , the quadratic equation
LC CIRCUIT - THE OSCILLATOR 3

1
(13) L 2 + R + = 0
C q
R R2 4L
C
(14) =
2L !
r
R 4L
(15) = 1 1 2
2L RC
p
If R > 2 L/C then is real and bothproots are negative, so the current
decays exponentially to zero. If R < 2 L/C, the roots are complex, so
there will be an exponential decay factor and an oscillatory factor, so the
current is a damped oscillation that decays to zero.
ENERGY IN A MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.26.
Weve seen that the energy stored in an electric field is

0
(1) WE = E 2d3r
2

where the integral is over all space. Here well look at the derivation of a
similar formula for the magnetic field.
The magnetic flux through an inductor carrying current I is

(2) = LI

where L is the inductance of the circuit. The emf induced in a circuit by


changing the current in that circuit is then

d
(3) E = = LI
dt
The power generated by a current is the voltage multiplied by the current,
and this power is the rate at which work is done, so

dWB
(4) = E I = LI I
dt
The sign in this equation reflects the fact that increasing the current through
the inductor induces an emf E that opposes the increase, so we need to do
work against this back emf to generate the current. This work should be
positive for an increasing current (that is, for I > 0), so weve defined the
signs to make this true.
If we increase the current from zero to some final value I then by inte-
grating this equation we get the total work done:

1
(5) WB = LI 2
2
1
ENERGY IN A MAGNETIC FIELD 2

This formula shows that the total energy delivered to the inductor de-
pends only on the final current, and not the route by which we get there.
Thus the current could increase for a while, then decrease for a bit, and then
increase again, but as long as it ends up at the final value of I the same work
is required.
To convert this expression to one containing the magnetic field requires
a bit of juggling with vector calculus, so here we go. First, we can write LI
in terms of B:

(6) LI = = B da

In terms of the vector potential, this becomes, using Stokess theorem:



(7) LI = ( A) da

(8) = A d`

where weve converted to a line integral around the circuit bordering the
area. Therefore, the work is

1 1
(9) WB = I A d` = (A I) d`
2 2
where weve made the current a vector in place of the directed line ele-
ment. If the current occupies a volume rather than a linear circuit, we can
write the generalization of this as

1
(10) WB = (A J) d 3 r
2
If these are steady currents (that is, weve increased the current up to a
certain value and then held it there), we can apply Ampres law in the form
B = 0 J to get

1
(11) WB = A ( B) d 3 r
20
Now we can use a vector calculus identity:

(12) (A B) = B ( A) A ( B)
ENERGY IN A MAGNETIC FIELD 3

to write

1 3 1
(13) WB = B ( A) d r (A B) d 3 r
20 20

1 2 3 1
(14) = B d r (A B) d 3 r
20 20
We can use the divergence theorem to convert the second integral to a
surface integral, and then perform the usual trick of letting the surface go to
infinity. We get

1 2 3 1
(15) WB = B d r (A B) da
20 20
If the currents are all localized, then both A and B tend to zero at infinity,
so we can ignore this final integral and get

1
(16) WB = B2 d 3 r
20
This is the energy stored in a (localized) magnetic field produced by
steady currents.
As an example, well consider the standard case of the infinite solenoid
(I know, I know, we derived this formula for finite current distributions,
so you can think of this problem as a very long solenoid rather than an
infinite one), with n turns per unit length carrying current I. Well work out
the energy per unit length of a section far from the ends.
If the solenoid has radius R, its inductance per unit length is

(17) L = R2 0 n2

so from 5 we have

1
(18) WB = R2 0 n2 I 2
2
We can also work this out from 9 using the vector potential of the solenoid
given by Griffiths in his example 5.12:

0 nIr
(19) A=
2
ENERGY IN A MAGNETIC FIELD 4

for r < R.
In a solenoid, I is in the direction so the total current in unit length is
nI , the line integral goes around the solenoid at r = R and we get


1 1 0 nIR 1
(20) (A I) d` = 2RnI = R2 0 n2 I 2
2 2 2 2
Next, we can use 16. The field inside the solenoid is a constant

(21) B = n0 I

and the volume per unit length is R2 so

1 1
(22) WB = R2 (n0 I)2 = R2 0 n2 I 2
20 2
Finally, we can use 15. Instead of taking the integration volume to be all
space, we can use any volume that completely encloses the current, so we
can use a cylindrical tube of inner radius r = a < R to r = b > R. Outside
the solenoid, B = 0 so we need look only at the region a r R. The first
integral is


1 1
B2 d 3 r = R2 a2 0 n2 I 2

(23)
20 arR 2
To do the surface integral, we first work out the direction of A B. A
is in the direction and B is in the z direction, so A B is in the radial
direction, pointing outwards. On the inner surface of the tube, da points
radially inwards, so (A B) da < 0, and the integral is


1 1 0 nIa 1
(24) (A B) da = 2a n0 I = a2 0 n2 I 2
20 r=a 20 2 2
Adding the two contributions, we get

1 1 1
WB = R2 a2 0 n2 I 2 + a2 0 n2 I 2 = R2 0 n2 I 2

(25)
2 2 2
Thus all four methods give the same answer.
ENERGY IN A MAGNETIC FIELD 5

P INGBACKS
Pingback: Energy in a toroidal solenoid
Pingback: Inductance of a coaxial cable
Pingback: LR circuit
Pingback: Poyntings theorem
Pingback: Energy flow in time-dependent fields
Pingback: Magnetic systems in thermodynamics
ENERGY IN A TOROIDAL SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.27.
Heres another simple example of calculating the energy in the magnetic
field. This time well look at a toroidal solenoid with N turns carrying a
current I. The field is given in Griffithss example 5.10:

0 NI
(1) B=
2r
This formula applies to a torus with an arbitrary cross section, but well
look at a rectangular cross section here. The inner radius is a, outer radius
b and height h. The energy is

1
(2) WB = B2 d 3 r
20
 2 b
1 0 NI 2rh
(3) = dr
20 2 a r2
0 N 2I2h
b
(4) = ln
4 a
1 2
Since WB = 2 LI we can use this formula to find the inductance of the
torus.

0 N 2 h b
(5) L= ln
2 a

which agrees with Griffithss equation 7.27 in example 7.11.

1
INDUCTANCE OF A COAXIAL CABLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.28.
Heres another simple example of calculating the energy in the magnetic
field and using this to calculate the inductance. We have a coaxial cable
with an inner cable of radius R in which the current I flows uniformly, and
an outer sheath at radius R in which the current makes its return journey.
The magnetic field within the inner cable is given by Ampres law if we
take a loop of radius r centred on the axis. The enclosed current is then

r2 2r
(1) Ienc = I 2
= B
R 0
0 rI
(2) B =
2R2
Outside the cable, B = 0, so the energy per unit length is
R
1
(3) WB = B2 2rdr
20 0
 2 R
2 0 I
(4) = r3 dr
20 2R2 0
0 I 2
(5) =
16
1 2
From WB = 2 LI we get the inductance per unit length:

0
(6) L=
8
In this case, the inductance is independent of R.

1
LR CIRCUIT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.29.
Suppose we have a circuit consisting of a battery delivering a voltage
E0 , an inductor L and a resistor R, all in series. If the circuit has been
connected long enough to stabilize, suppose we then throw a switch that
bypasses the battery and isolates L and R in their own loop. The inductor
will then produce some back emf that opposes the drop in current, in the
amount E = LI. Since this is now the only source of emf in the circuit, it
must equal the voltage drop across the resistor, so

(1) LI = IR
R
(2) I = I
L
E0 Rt/L
(3) I (t) = I0 eRt/L = e
R

where I0 is the initial current, which is I0 = E0 /R.


The total energy delivered to the resistor by the inductor is

1
(4) W =R 2
I (t) dt e2Rt/L dt = LI02
= I02 R
0 0 2
From the energy formula for an inductor, this is equal to the initial energy
in the inductors magnetic field.
 2
1 E0
(5) W= L
2 R

1
MUTUAL INDUCTANCE BETWEEN TWO MAGNETIC
DIPOLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.30.
This problem relates mutual inductance to the interaction energy of two
magnetic dipoles. Suppose we have two tiny loops of wire with area vectors
a1 and a2 . The loops become dipoles if they carry a current I so that their
magnetic dipole moments are mi = Iai (i = 1, 2). We can work out the
mutual inductance between these two loops if we assume that the areas are
small enough that the field produced by one dipole is constant over the area
of the other dipole. The magnetic field of dipole 2 as felt by dipole 1 is

0
(1) B21 = 3
[3 (m2 r21 ) r21 m2 ]
4r12
0 I
(2) = 3
[3 (a2 r21 ) r21 a2 ]
4r12
where r21 is the unit vector pointing from dipole 2 to dipole 1. The mutual
inductance can be worked out if we know the flux through dipole 1. With
our approximation of constant field over dipole 1, we get

(3) 21 = B21 a1
0 I
(4) = 3
[3 (a2 r21 ) (r21 a1 ) a2 a1 ]
4r12
The mutual inductance is

21 0
(5) M21 = = 3
[3 (a2 r21 ) (r21 a1 ) a2 a1 ]
I 4r12
Conversely, if we run a current I through dipole 2 and look at the flux
through dipole 1, we get

(6) 12 = B12 a2
0 I
(7) = 3
[3 (a1 r12 ) (r12 a2 ) a1 a2 ]
4r12
1
MUTUAL INDUCTANCE BETWEEN TWO MAGNETIC DIPOLES 2

Since r12 = r21 , we see that 12 = 21 and thus M12 = M21 as required.
If we now start off with a current I1 flowing in dipole 1 but no current in
dipole 2, then switch on a current I2 in dipole 2, the emf induced in dipole
1 is given by the change in flux:

d21
(8) E21 =
dt
dI2
(9) = M21
dt
0 dI2
(10) = 3 dt
[3 (a2 r21 ) (r21 a1 ) a2 a1 ]
4r12
To get the total work required if we want to maintain I1 in dipole 1, we
need to consider how this induced emf affects I1 . If we didnt attempt to
maintain I1 , then I1 would change in the direction prescribed by the sign
of E21 . So if we want to maintain I1 , we have to oppose E21 , and thus the
rate at which the work is done is dW21 /dt = E21 I1 , where the minus sign
indicates were opposing E21 .

dW21
(11) = E21 I1
dt
0 I1 dI2
(12) = 3 dt
[3 (a2 r21 ) (r21 a1 ) a2 a1 ]
4r12
Since were keeping I1 and all the geometrical terms constant, we can
just integrate this expression to get the total work done:
t
0 I1 dI2
(13) W21 = 3
[3 (a 2 r21 ) (r21 a 1 ) a 2 a 1 ] dt
4r12 0 dt
0 I1 I2
(14) = 3
[3 (a2 r21 ) (r21 a1 ) a2 a1 ]
4r12
0
(15) = 3
[3 (m2 r21 ) (r21 m1 ) m2 m1 ]
4r12
This last expression is exactly equal to the interaction energy between
the two dipoles except that its opposite in sign. Im not really sure why
this should be (assuming this answer is correct), though the two cases are
physically different. In this case, we start with one dipole at full strength
given by m1 and then turn the other one on, so the work doesnt involve
any movement of the dipoles; we can move dipole 2 into position before
we turn the current on, so there is no interaction with the first dipole until
its in place. In the previous case, both dipoles were switched on and then
MUTUAL INDUCTANCE BETWEEN TWO MAGNETIC DIPOLES 3

moved into position, so the motion of the dipoles causes the change in flux
(and thus the back emf), rather than any change in the current (and hence in
the dipoles strengths). Not sure.
MAXWELLS CORRECTION TO AMPRES LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.31.
We run into problems if we try to extend Ampres law to a non-static
situation. Ampres law in differential form is

(1) B = 0 J
Since the divergence of a curl is always zero, this law implies that

(2) ( B) = 0 = 0 J
For steady currents, the RHS is indeed true, but in general


(3) J =
t

where is the charge density. Thus for time-varying currents, Ampres


law is not true. Maxwell fixed this problem by making a postulate that is, in
a way, the complement of Faradays postulate that a changing electric field
produces a magnetic field; Maxwell proposed that a changing electric field
induces a magnetic field. In particular, he proposed that

E
(4) B = 0 J + 0 0
t
This fixes the divergence problem, since we now get

( E)
(5) ( B) = 0 = 0 J + 0 0
t

but from Gausss law, E = /0 , so the RHS is indeed now zero.


Since 0 tE has the dimensions of current density, Maxwell called it the
displacement current:

E
(6) Jd 0
t
1
MAXWELLS CORRECTION TO AMPRES LAW 2

As a simple example, suppose we have a circular solid wire of radius a


that has a narrow gap of width w  a cut into it. If a steady current I is
passed along the wire, the two surfaces on either side of the gap act as a
parallel plate capacitor, as charge builds up on these surfaces. As a result,
the electric field between the plates changes as the charge builds up, so we
should get an induced magnetic field. Neglecting fringe effects at the edges
of the plates, the electric field between the plates is E = /0 where is
the surface charge density. What were really interested in isnt E itself, but
its derivative, which we can write as

E 1 d
(7) =
t 0 dt
1 dQ
(8) =
a2 0 dt
I
(9) =
a2 0
Taking a circular path of radius r between the plates, there is no current
here so J = 0 and we have

(10) B d` = 2rB

E
(11) = 0 0 da
t
I
(12) = 0 0 r2 2
a 0
r 2
(13) = 0 I 2
a
Ir
(14) B = 0
2a2
By symmetry and the right hand rule, if we take the z direction to be from
the positive to the negative plate, then

Ir
(15) B = 0
2a2
P INGBACKS
Pingback: Displacement current in a capacitor
Pingback: Displacement current is very small
Pingback: Maxwells equations in matter
Pingback: Maxwells equations with varying charge but constant current
MAXWELLS CORRECTION TO AMPRES LAW 3

Pingback: Displacement current in an infinite wire


Pingback: Poyntings theorem
Pingback: Poyntings theorem and conservation of energy
DISPLACEMENT CURRENT IN A CAPACITOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.32.
Heres a slight variant on the previous problem in which we considered
a thick wire of radius a with a small gap of width w  a. A steady current
I flows along the wire with the result that the gap acts like a parallel plate
capacitor.
This time, well replace the thick wire with a very thin wire and replace
the gap with two circular plates, again with radius a and separated by a
gap of width w  a, so we have a more realistic capacitor. We still have a
steady current I, and well assume that charge piles up on each plate in such
a way that the surface charge density is constant in space (obviously is
increasing with time).
The electric field between the plates is the same as in the previous prob-
lem, so


(1) E= z
0
Now is increasing as current flows onto the plates, so

E 1
(2) =
t 0 t
1 dQ
(3) =
a2 0 dt
I
(4) =
a2 0

and the field as a function of time is

It
(5) E= z
a2 0
The displacement current is then
1
DISPLACEMENT CURRENT IN A CAPACITOR 2

(6) Jd = 0 E
0 dQ
(7) = z
a2 0 dt
I
(8) = z
a2
To find the induced magnetic field between the plates we can take as
the path of integration a circle of radius r halfway between the plates, and
the calculation is the same as in the previous post if we take the area of
integration to be the flat circular area filling the circle. Thus we get

r
(9) B = 0 I
2a2
However, in applying Stokess law, we are allowed to choose as the area
bounded by the curve any area we like, not necessarily a planar one. So we
can choose an open-ended cylinder with the open end bounded by the circle
between the plates, the sides of the cylinder extending back through one of
the plates to a point behind the plate, and closed off with a circular cap that
cuts through the wire leading up to the plate. Since only part of the sides of
the cylinder project into the gap between the plates, and the normal to this
area is perpendicular to Jd , there is no contribution to the area integral from
the displacement current.
Behind the plate, we can apply the original form of Ampres law since
the currents here are all steady. There is a current I crossing into the surface
where the wire cuts the end cap, but if r < a, current is also flowing out of
the cylinder as it spreads out over the plate. Thus the net current enclosed
by the surface is

a2 r 2

r2
(10) Ienc = I I = I
a2 a2
The line integral of the magnetic field is the same as before (since the
circle around which the integral is taken hasnt changed), so

(11) 2rB = 0 Ienc


r
(12) B = 0 I
2a2

so we get the same answer whatever surface we choose.


DISPLACEMENT CURRENT IS VERY SMALL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.33.
Well consider again the problem of induced magnetic and electric fields
in a coaxial cable with alternating current. A thin wire carries current I (t) =
I0 cos t and the current returns down the coaxial cylinder of radius a. We
saw earlier that the induced electric field is
(
0 I0
z 2 sin t ln as 0<s<a
(1) E (s,t) =
0 s>a
The displacement current density inside the cylinder is then

E 2 0 I0 0 a
(2) Jd = 0= z cos t ln
t 2 s
The total displacement current is then
a
2 0 I0 0 a
(3) Id = cos t 2s ln ds
2 0 s
 2 a
2 s a s2
(4) = 0 I0 0 cos t ln +
2 s 4 0
2 0 I0 0 a2
(5) = cos t
4
The ratio of the two currents is

Id 2 0 0 a2
(6) =
I 4
We can convert this to a numerical value if we note that 0 0 = 1/c2 =
2
1/ 3 108 so

Id
(7) = 2.78 1018 2 a2
I
If a = 1 mm, then to get Id up to 1% of I, we would need a frequency of
1
DISPLACEMENT CURRENT IS VERY SMALL 2

(0.01)
(8) 2 = 2
(2.78 1018 ) (103 )
(9) = 6 1010 s1

or in cycles per second

(10) = 9.55 109 Hz


Ordinary household current is in the range 50 - 60 Hz, but 1010 Hz is in
the microwave region of the EM spectrum.
MAXWELLS EQUATIONS AND AN EXPANDING SHELL OF
CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.34.
We can now summarize Maxwells equations


(1) E =
0
(2) B = 0
B
(3) E =
t
E
(4) B = 0 J + 0 0
t
A novel example of electric and magnetic fields that satisfy these equa-
tions is

1 q
(5) E = H (vt r) r
40 r2
(6) B = 0

where H (x) is the step function


(
1 x>0
(7) H (x)
0 x<0
(Ive used H instead of Griffithss for the step function to avoid confu-
sion with the spherical angle .)
Fairly obviously, B = B = 0 and, since E has only a radial com-
ponent which depends only on r, E = 0 (if you look up the form of the
curl in spherical coordinates, the only derivatives of Er are with respect to
and which are both zero).
You might think that E is fairly easy to calculate using the definition of
the divergence in spherical coordinates, but we must remember that some-
thing bizarre happens at r = 0; we must use the formula for the 3-d delta
function:
1
MAXWELLS EQUATIONS AND AN EXPANDING SHELL OF CHARGE 2

 
r
(8) 2 = 4 3 (r)
r

and the product rule for the divergence of the product of a scalar field f and
a vector field A:

(9) ( f A) = A f + f A
Here we have

(10) f = H (vt r)
1 q
(11) A = r
40 r2
Using the derivative of the step function dH (x) /dx = (x) we have

dH (vt r)
(12) = (vt r)
dr

so
 
1 q q r
(13) E = 2
r (H (vt r)) H (vt r) 2
40 r 40 r
1 q q
(14) = 2
r ( (vt r) r) H (vt r) 3 (r)
40 r 0
The second term is non-zero only at r = 0 so, assuming vt > 0 we can
omit the step function since it is always 1 at r = 0. We therefore have

q q
(15) E = 2
(vt r) 3 (r)
4r 0 0
The charge density is therefore

(16) = 0 E
q
(17) = (vt r) q 3 (r)
4r2
This corresponds to a point charge q at the origin and an expanding
(with speed v) spherical shell of charge of total amount +q. The electric
field inside the shell is due entirely to the point charge at the origin, and the
field outside the shell is zero since the total enclosed charge is zero.
The current density is obtained from B = 0, so:
MAXWELLS EQUATIONS AND AN EXPANDING SHELL OF CHARGE 3

E
(18) 0 J + 0 0 = 0
t
E
(19) J = 0
t
qv
(20) = (vt r) r
4r2
The expanding shell provides the current density; the stationary charge at
the origin doesnt contribute.
P INGBACKS
Pingback: Electromagnetic waves in vacuum
Pingback: Maxwells equations in terms of potentials
Pingback: Maxwells equations in cylindrical coordinates
MAGNETIC MONOPOLE FORCE LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.35.
If magnetic monopoles did exist, we could consider the behaviour of
magnetic charges qm . We could postulate a Coulombs law for them:

0 qm1 qm2 (r2 r1 )


(1) F=
4 |r2 r1 |3
Griffiths sets the problem of working out the force law for a magnetic
charge passing through electric and magnetic fields. This strikes me as
rather difficult, since the Lorentz force law for electric charges is essentially
a postulate of the theory (confirmed by experiment); it was never derived.
The simplest equivalent magnetic force law would seem to be something
like

(2) F = qm B + qm v E
We do need to check the units, however. First, what are the units of a
magnetic charge qm ? The units of 0 are m kg s2 A2 so the Coulomb
force law gives us

(3) kg m s2 = m kg s2 A2 [qm ]2 m2

so the units of qm must be A m = Coulomb m s1 , that is, charge times


velocity. How does this fit with our proposed Lorentz-like force law? The
units of B are kg m s2 A1 m1 , so qm B has the units of force, so that term
seems OK.
For the second term, we know that electric charge times electric field
give force, so that means that qm v E has the units of force times velocity
squared, so we need to divide by some constant velocity squared. Since we
know that 0 0 = 1/c2 this seems a logical choice, so we can propose

qm
(4) F = qm B + 0 0 qm v E = qm B + vE
c2
In fact, the commonly accepted form is
1
MAGNETIC MONOPOLE FORCE LAW 2

qm
(5) F = qm B vE
c2
I could not find a derivation of this result; rather it seems that this form
is adopted to make this force law invariant under a duality transformation.
P INGBACKS
Pingback: Duality transformation for magnetic and electric charge
DETECTING MAGNETIC MONOPOLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.36.
Blas Cabrera (1982), Physical Review Letters, 48, 1378.
Although magnetic monopoles have never been found (at least, repro-
ducibly), one experiment that tried to find them was that of Cabrera in
1982. The experiment was a fairly simple setup, consisting of a super-
conducting (and therefore zero-resistance) wire loop with a magnet aligned
so that monopoles, should they exist, could pass through the loop. Assum-
ing that a single magnetic charge qm emits a magnetic field that obeys a
Coulomb-like law, that is

0 qm
(1) B= r
4 r2

we can work out the magnetic flux through the loop as a single monopole
falls through it.
If the speed of the monopole is v and it falls along the axis of the loop
then, assuming the loop has radius b and we take asthe area of integration
the flat circle within the loop, we need to work out B da to get the flux.
Suppose the monopole is a distance d from the centre of the loop. Then the
distance from the monopole to a point on the disk with radius s is d 2 + s2
so the field strength at that point is

0 qm
(2) B=
4 d 2 + s2
The term B da isolates the component of B that is perpendicular to the
disk, which is B cos where is the angle between the axis and a line from
the monopole to a point on the disk at radius s. We get

d
(3) cos =
d + s2
2

so the flux from the monopole is


1
DETECTING MAGNETIC MONOPOLES 2


(4) = B da

0 qm b 2sd
(5) = ds
4 0 (d 2 + s2 )3/2
b
0 qm d
(6) =
2 d 2 + s2 0
 
0 qm d
(7) = 1
2 d 2 + b2
If we take t = 0 to be the time when the monopole crosses the plane of
the disk and take the surface normal at the disk to point in the direction of
the monopoles velocity, then d = vt for t < 0 and d = +vt for t > 0, and
B da > 0 for t < 0 and B da < 0 for t > 0. That is
 
0 qm vt
2 1+ 2 2 t <0


(vt) +b
(8) (t) =  
q
02 m 1 vt2 2 t >0


(vt) +b

To go further, we need to modify Maxwells equations to include mag-


netic charge. The relevant one is Faradays law which needs an extra term:

B
(9) E = 0 Jm
t

where Jm is the magnetic current density. This is the analog to the B


equation which involves electric current density. Applying Stokess theo-
rem, we can integrate the LHS around the loop and the RHS over the disk
enclosed:


B
(10) E d` = 0 Jm da da
t
d
(11) E = 0 Imenc
dt

where E is the induced back emf and Imenc is the magnetic current flowing
through the loop. This emf can be written in terms of the self-inductance L
of the loop:
DETECTING MAGNETIC MONOPOLES 3

dI (t)
(12) E = L
dt
What are we to make of Imenc considering we have only a single mono-
pole to make up the current? We can write it as a delta function:

(13) Imenc = qm (t)

That is, there is a current consisting of a single charge across the disk only
at time t = 0. Since the delta function is the derivative of the step-function
H, we can integrate Faradays law to get

(14) LI (t) = 0 qm H (t) + (t)


So we get
 
q vt
2L 1 + 2 2
0 m
t <0


(15) I (t) = (vt) +b 
q 0 qm
0L m 2L 1 vt2 t >0


(vt) +b2
The t > 0 term comes out to be the same as the t < 0 term, so we get in
general

0 qm vt
(16) I (t) = 1+ q
2L 2 2
(vt) + b
For t , I 0, while for t , I 0 qm /L. This means that, when
the monopole is infinitely far away and approaching, there is no induced
current. The current increases as the monopole gets closer, but since the
superconducting loop has no resistance, the current that is built up due to
the back emf never dissipates and tends to a constant non-zero value as the
monopole recedes into the distance.
MAXWELLS EQUATIONS IN MATTER

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.37.
Inside matter, weve seen that the polarization and magnetization give
rise to bound charges and bound currents. The earlier results applied in the
electrostatic and magnetostatic cases, respectively, so wed like to general-
ize these to get the corresponding equations in electrodynamics.
For a medium with polarization P and magnetization M, the static cases
are

(1) P = b
(2) P n = b
(3) M = Jb
(4) M n = Kb

where b and b are the bound volume and surface charge densities and Jb
and Kb are the bound volume and surface currents. These equations assume
that nothing changes with time.
Now suppose that the polarization does change with time, so that P =
P (t). Considering a small cylindrical volume element with an axis parallel
to P, there is no bound surface charge on the sides of the cylinder, and on
one end b = P and on the other b = P. If P increases by P over time t
then the surface charge density also increases (or decreases, at the negative
end, but both ends increase in magnitude) by P. This change in charge
density must be due to a current density J p parallel to P of amount

b P
(5) Jp = n =
t t
This current density is called the polarization current. It is a current that
is not present at all in the static case, so its a new current in addition to the
bound current Jb and free current J f .
If the magnetization changes with time, this will change the bound cur-
rents Jb and Kb but it doesnt introduce any new currents into the system,
so the only change we need to make in a dynamic system is the addition of
1
MAXWELLS EQUATIONS IN MATTER 2

the polarization current J p . The charge density is thus still composed of


two contributions:

(6) = f + b = f P

but the current density now has an additional term:

P
(7) J = J f + Jb + J p = J f + M +
t

Gausss law in matter can thus be written

1 
(8) E = = f P
0 0
(9) (0 E + P) = f
(10) D = f

where D = 0 E + P.
where is the permittivity of the medium and is constant everywhere in the
medium.
Ampres law with Maxwells correction must take into account the po-
larization current, so we have

E
(11) B = 0 J + 0 0
 t 
P E
(12) = 0 J f + M + + 0 0
t t
 
1
(13) B M = J f + (0 E + P)
0 t
D
(14) H = Jf +
t

The other two Maxwell equations dont change, since they dont refer
explicitly to charge or current. Thus we get Maxwells equations in matter:
MAXWELLS EQUATIONS IN MATTER 3

(15) D = f
(16) B = 0
B
(17) E =
t
D
(18) H = Jf +
t
For linear, homogeneous materials, the polarization and magnetization
depend linearly on E and B, respectively, so

(19) D = E
1
(20) H = B

where

(21) = 0 (1 + e )
(22) = 0 (1 + m )

are the permittivity and permeability of the material. Maxwells original


displacement current, 0 E/t, becomes D/t.

Example. Suppose we have a large parallel plate capacitor embedded in


sea water and maintain an AC voltage across the plates, so that

(23) V (t) = V0 cos (2t)

In terms of the conductivity of the sea water, the current density due to
conduction is J = E and for a capacitor with plates separated by a distance
d, E (t) = V (t) /d, so

V0
(24) J= cos (2t)
d
The displacement current is
MAXWELLS EQUATIONS IN MATTER 4

D
(25) Jd =
t
E
(26) =
t
V0
(27) = 2 sin (2t)
d
The ratio of the amplitudes is therefore

J 1
(28) = =
Jd 2 2

where is the resistivity of sea water.


Plugging in the values given by Griffiths, we have = 0.23 m, =
810 = 7.17 1010 m3 kg1 s4 A2 , = 4 108 s1 and

J
(29) = 2.41
Jd
P INGBACKS
Pingback: Electromagnetic waves in matter: normal reflection and trans-
mission
Pingback: Decay time for free charge in a conductor
Pingback: Maxwells equations in matter: boundary conditions
CURRENT BETWEEN TWO PLATES WITH A SPHERICAL
PERTURBATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.38.
Another example of calculating the current flowing between two conduc-
tors separated by a weakly conducting medium. We have two large parallel
plates separated by a distance d, with the lower plate at potential 0 and the
upper plate at potential V = V0 . A metal hemisphere of radius a  d rests
on the lower plate so its potential is also zero. The region between the plates
is filled with a substance of conductivity . We want the current flowing to
the hemisphere.
The relation between current and electric field is

(1) J = E

so we need the field around the hemisphere. This is worked out (well, the
potential is anyway) for a complete sphere by using a series solution of
Laplaces equation in Griffithss example 3.8, so we can quote the result
here:

a3
 
(2) V = E0 r 2 cos
r

Here E0 is the uniform electric field surrounding the hemisphere and


since the hemisphere is small relative to the separation of the plates, E0 =
V0 /d. We can use E = V to get the field in spherical coordinates:

a3 a3
   
V0 V0
(3) E= 1 + 2 3 cos r + 1 + 3 sin
d r d r

Since were concerned only with the field between the plates, and the
plates themselves provide the uniform background field, we can also apply
this formula to the hemisphere.
1
CURRENT BETWEEN TWO PLATES WITH A SPHERICAL PERTURBATION 2

To get the current flowing onto the hemisphere, we need to integrate this
over the surface of the hemisphere. Since the normal to the surface is par-
allel to r only the Er component contributes and we get, taking r = a since
were integrating over the surface of the hemisphere:

(4) I = E da

V0 /2
(5) = 3 cos (2a sin ) (ad )
d 0
3 a2V0
(6) =
d

where in the second line, the spherical surface element is the horizontal
circumference of the sphere at angle (which is 2a sin ) multiplied by
the arc length on the surface of an increment in (which is ad ).
COPPER PIPES IN A CONDUCTING MEDIUM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.39.
Here well revisit the problem of two parallel copper pipes, except this
time well embed the pipes in a weakly conducting medium with conduc-
tivity and find the current that flows between them per unit length. The
geometry of the problem is the same as before. We have two infinite copper
pipes, each of radius a. The axis of one pipe is on the line x = d, (y = 0)
and the other is on x = +d. The potential of the one on the left is held at
V0 and the one on the right at +V0 . We found the potential of this system
is

(x + b)2 + z2
(1) V= ln
40 (x b)2 + z2

where

20V0
(2) =
cosh1 da
p
(3) b = d 2 a2
The dual pipe problem is transformed into a problem involving two par-
allel wires lying along the lines x = b and carrying linear charge densities
. We can use Gausss law to integrate the electric field over a cylinder
of unit length around the wire with charge density to get the current
flowing into the wire:

2V0 2V0
(4) I= E da = = 1 d
=  q  
0 cosh a d d 2
ln a + a 1

Gausss law says that the surface integral of E is q/0 , where q is the charge
enclosed by the surface. The minus sign indicates that the field points to-
wards the wire, and thus in the opposite direction to the surface normal so
the current flows towards the wire.
1
COPPER PIPES IN A CONDUCTING MEDIUM 2

Incidentally, if we wanted to follow the same procedure as in the previous


problem (and not use the parallel wire equivalent problem), we need to find
the electric field and then integrate it over a surface enclosing one of the
pipes. This is a lot harder, but heres how it goes. Its easier to use a
cylindrical coordinate system with its axis equal to the axis of one of the
pipes, so lets choose the z axis to be along the line x = d. We can then
transform coordinates to x + d so the potential becomes

( + b d)2 + z2
(5) V= ln
40 ( b d)2 + z2
In terms of the cylindrical coordinates, we have = r cos , so

r2 + 2 (b d) r cos + (b d)2
(6) V= ln
40 r2 2 (b + d) r cos + (b + d)2
Now to find the electric field we take the negative gradient, but since
were integrating over the surface of the pipe, the integrand E da will in-
volve only the radial component Er = V / r. As this involves the loga-
rithm of a quotient of two polynomials, the derivative isnt exactly simple
and at this stage its easier to turn the problem over to Maple. The deriva-
tive is too complex to write down here, but we can then integrate this over
a cylinder of radius a (so that r = a and the integral is over from 0 to 2),
and Maple confirms we get the same answer as above.
RESISTANCE IN A CONICAL CAPACITOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.40.
Heres another example of calculating the resistance. We have a truncated
cone with the radius of the shorter end cap being a and of the longer being b,
and the length between the two ends being L. The b end is held at constant
potential V0 and the a end is at V = 0. What is the current between the ends
if the cone is filled with a material with resistivity ?
If we make the (incorrect) assumption that the electric field between the
ends is constant and parallel to the axis, we can derive the resistance as
follows. Consider a thin slice of thickness dz through the cone. The voltage
drop across this slice is dV = V0 dz/L, so the current flowing through the
slice is

dV 1 2
(1) I= r
dz

where r is the radius of the cone at point z. From Ohms law, the resistance
dR of the slice can be found:

(2) dV = IdR
dz
(3) dR =
r2
We need r as a function of z which we can get from similar triangles.
Suppose we extend the cone at the a end so that its vertex is restored, and
call the distance from the vertex to the a circles centre z0 . Then we get

a r b
(4) = =
z0 z + z0 z0 + L

where z is the distance along the axis from the a end towards to b end (so
z = 0 at a and z = L at b).
We know everything in these equations except r and z0 , so we can solve
to get them:
1
RESISTANCE IN A CONICAL CAPACITOR 2

aL
(5) z0 =
 a
b
z
(6) r = a 1+
z0
 
z (b a)
(7) = a 1+
aL
We can now get the total resistance of the truncated cone as

L
dz
(8) R = 2
0 r
L
z (b a) 2


(9) = 1+ dz
a2 0 aL
L
(10) =
ab
The problem with this derivation is that E is not constant along the cones
length, since the field spreads out from the short end to the long end.
An alternative situation that does work out properly is obtained by replac-
ing the flat ends of the cone with circular sections from a sphere centred at
the cones vertex. In this case the a end is a concave surface of radius a and
the b end is a convex surface of radius b. We can take L to be the distance
between the centres of the circles at either end of the cone. The parame-
ters r and z0 are therefore the same as before. This time the infinitesimal
slices are concentric with the spherical cap and have a thickness dy (where
y is now the radial coordinate, since weve used r and R for other things
already). The area of a shell of spherical radius y is

0
2
(11) A = 2y sin d = 2y2 (1 cos 0 )
0

where 0 is the angle subtended by the radius of the shell. (Note that this
formula becomes 4y2 for 0 = so it correctly gives the surface area of a
sphere).
The angle 0 is the same for all shells and from the lower end of the cone
we have

z0
(12) cos 0 = q
z20 + a2
RESISTANCE IN A CONICAL CAPACITOR 3

The resistance of a spherical shell is then

dy
(13) dR =
A

so we need to sum up all the shells. We can do this two ways: by integrating
over y or over z. First, we integrate over y. To do this, we need the end
points, which are the spherical radii for each end of the cone. These are

q
(14) ya = z20 + a2
q
(15) yb = (z0 + L)2 + b2

The resistance is then

yb
dy
(16) R =
2 (1 cos 0 ) ya y2
(b a)2
(17) = q 
2
2ab L2 + (b a) L

If we do the integral over z we need to transform the differential. For a


general circle within the cone, we have

q
(18) y = (z0 + z)2 + r2
s  2
2 a
(19) = (z0 + z) + (z0 + z)
z0
q
(20) = (z0 + z) 1 + a2 /z20
q
(21) dy = 1 + a2 /z20 dz
1
q
(22) = L2 + (b a)2 dz
L

The resistance integral is then


RESISTANCE IN A CONICAL CAPACITOR 4

L
1
q
2 2 dz
(23) R = L + (b a) i2
2 (1 cos 0 ) L
h
0 (z0 + z)2 + a
z0 (z0 + z)
(b a)2
(24) = q 
2 2
2ab L + (b a) L

after doing all the substitutions.


ELECTRIC POTENTIAL FROM A STEADY CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.41.
Mark A. Heald (1984), American Journal of Physics, 52, 522.
Ohms law is derived from the assumption that in a conducting medium
J = E. If the currents are steady, then the charge density is independent
of time, so J = d
dt = 0, which implies that (if is constant throughout
the material) E = 2V = 0, and we can use the various methods we
derived earlier to solve Laplaces equation and determine the potential. One
example of this is given here.
We have a cylindrical pipe of conductor of radius a, with a very thin slot
cut down the side. A battery is attached to either side of the slot so that one
edge of the slot is maintained at potential +V0 /2 and the other at V0 /2.
This results in a uniform, steady current flowing around the circumference
of the pipe. Since the conductivity of the pipe is uniform, the potential at an
angle on the surface of the pipe is

V0
(1) V (a, ) =
2

for < < +. Our job is to find the potential both inside and outside
the cylinder.
We can approach this problem using the series solution to Laplaces equa-
tion in cylindrical coordinates that we worked out earlier:

(2)

1
V (r, ) = A ln r +B+ rn (An sin n + Bn cos n )+ n
(Cn sin n + Dn cos n )
n=1 n=1 r

with the boundary condition given above.


Inside the cylinder, to keep V finite we must throw away the log term and
the series in inverse powers of r, so inside we have


(3) Vr<a = Bi + rn (An sin n + Bn cos n )
n=1
1
ELECTRIC POTENTIAL FROM A STEADY CURRENT 2

Similarly, outside we get


1
(4) Vr>a = Bo + n
(Cn sin n + Dn cos n )
n=1 r
At the boundary, we must have a continuous potential satisfying the
boundary condition, so

(5)

V0 1
Bi + an (An sin n + Bn cos n ) = = Bo + n (Cn sin n + Dn cos n )
n=1 2 n=1 a
At this point, we would like to be able to equate the coefficients of the sin
and cos terms, but the boundary condition depends on and not explicitly
on sines or cosines. However, one of the nice things about Fourier series
(which these series are, really) is that we can express any function in terms
of them, so we can try to find a series such that


V0
(6) = Fn sin n
2 n=1
We can use the usual procedure for finding Fn : multiply both sides by
sin m and integrate:

V0
(7) sin m d = Fn sin m sin n d
2 n=1

Since the sine function is orthogonal on the interval of integration, we get



V0
(8) sin m d = Fm sin2 m d
2
Working out the integrals, we get

(1)m+1 V0
(9) = Fm
m

so

V0 V0 (1)n+1
(10) = n sin n
2 n=1
Now we can equate coefficients, so we get for r < a
ELECTRIC POTENTIAL FROM A STEADY CURRENT 3

(11) Bi = 0
(12) Bn = 0
(1)n+1
(13) An =
nan
V0  r n sin n
(14) Vr<a =
n=1 a n
Similar reasoning on the outside gives

(15) Bo = 0
(16) Dn = 0
(1)n+1 an
(17) Cn =
n
V0  a n sin n
(18) Vr>a =
n=1 r n
To get the answer given in Griffiths, we need to sum the series. Maple is
able to do this explicitly, and we get
 a 
V0 r sin
(19) Vr>a = arctan
1 + ar cos
 
V0 a sin
(20) = arctan
r + a cos
The potential inside the cylinder is the same except with a and r swapped,
so we have:
 
V0 r sin
(21) Vr<a = arctan
a + r cos
We can check that these formulas do indeed satisfy the boundary con-
dition using a couple of trig identities. Using double angle formulas we
have


(22) sin = 2 sin cos
2 2
2
(23) cos = 2 cos 1
2

so at r = a, we have
ELECTRIC POTENTIAL FROM A STEADY CURRENT 4


V0 2a sin 2 cos 2
(24) V (a, ) = arctan  
2
a 1 + 2 cos 2 1

V0 sin 2
(25) = arctan
cos 2
V0
(26) =
2
The surface charge density on the cylinder can be found from the radial
component of the electric field at the surface, which is

V
(27) Er =
r r=a
V0 sin
(28) =
2a (1 + cos )

with the plus sign for the inner surface and the minus sign for the outer
surface.
The surface charge density is 0 Er , or

0V0 sin
(29) =
2a (1 + cos )
(The answer given in Griffithss question is clearly wrong; the charge
density does not depend on r (what Griffiths calls s), and this answer agrees
with that given in Healds paper.)
SURFACE CURRENT INDUCED BY A MAGNET FLOATING
OVER A SUPERCONDUCTOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.43.
Heres another example of the effect of levitating a magnet above a super-
conductor. In the last post, we showed that the equilibrium position of the
magnetic dipole occurs when it aligns itself horizontally above the planar
superconductor. In this post, well look at what happens if the dipole is con-
strained to point normal to the plane. That is, we start with a dipole m which
points in the +z direction and sits at a height h above the superconducting
plane.
As we saw in the previous post, the system can be modelled using the
method of images, so the image dipole in this case is m at position z = h.
Assuming the dipole has mass M, the height h can be found by equating the
magnetic and gravitational forces. Weve already found the formula for the
force between two dipoles whose axes are parallel; to get the force when
the axes are anti-parallel, we just reverse the sign, giving

30 m2
(1) F= z
2r4

where r is the distance between the dipoles, so at equilibrium, r = 2h. The


height is thus found from

30 m2
(2) = Mg
2 (2h)4
1/4
30 m2

1
(3) h =
2 2Mg
This is slightly higher (by a factor of 21/4 ) than in the horizontal dipole
case.
The above result was derived by using the fact that the normal compo-
nent of B is continuous across boundaries. The tangential component of
B, however, is not continuous; it depends on the surface current density K.
The relation is
1
SURFACE CURRENT INDUCED BY A MAGNET FLOATING OVER A SUPERCONDUCTOR
2

(4) Babove Bbelow = 0 K n

In our case, Bbelow = 0 due to the Meissner effect (exclusion of mag-


netic fields from a superconductor), so if we can find Babove , we can get the
surface current density. We start with the formula for the field due to the
dipole:

0
(5) B= [3 (m r) r m]
4r3
Since the dipole points along the +z direction, the system has cylindrical
symmetry so consider a cross section lying in the xz plane, and choose a
point on the x axis (which lies on the boundary with the superconductor).
The unit vector r points from the dipole to the point x and makes an angle
with the z axis. Since m = mz, m r = m cos .
First, we show that there is no component of B in the y direction. The
component By = B y, but since r lies in the xz plane r y = 0. Also, since
m = mz, m y = 0, so By = 0.
Now to get Bx = B x. The angle between r and x is 2 , so r x =
cos 2 = sin , and m x = 0, so

0
(6) Bx = (3m cos sin )
4r3
By the way, the image solution does not, of course, give the correct an-
swer for the field just inside the boundary (where the field is zero), but thats
ok, since image solutions are valid only outside conductors.
To express the cos and sin in terms of the distances, consider the right
triangle with vertices at the origin, the dipole and the point x. The vertical
side has length
h, the horizontal side has length x and the hypotenuse has
length r = h2 + x2 . The angle betweensides h and r is so the angle
between sides x and r is + 2 = 2 . Therefore

h  
(7) = sin = cos
r 2
x 
(8) = cos = sin
r 2
Putting it all together, we get
SURFACE CURRENT INDUCED BY A MAGNET FLOATING OVER A SUPERCONDUCTOR
3

30 mhx
(9) Bx =
4r5
30 mhx
(10) = 5/2
4 (h2 + x2 )
By symmetry, the x component from the image dipole will be the same,
so the total field is

30 mhx
(11) Bx;tot = 5/2
2 (h2 + x2 )
The minus sign indicates that the field points towards the origin, so the
net field on the xy plane is radial, pointing inwards. That is, in cylindrical
coordinates

30 mhr
(12) B= 5/2
r
2 (h2 + r2 )
where r now represents the cylindrical radius measured from the z axis.
The current density is therefore

3mhr
(13) K= 5/2

2 (h2 + r2 )
where here the minus sign indicates that the current flows in a clockwise
direction when viewed from above.
It may seem odd that a static configuration of a magnet floating over a
superconductor can give rise to a current, but remember that the magnet
must be brought in to its final position and due to the Meissner effect, a
surface current appears to negate the field inside the superconductor. Since
the superconductor has zero resistance, once the current is established, it
remains forever.
FLOATING A MAGNET ABOVE A SUPERCONDUCTOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.44.
A superconductor has a number of curious properties, one of which is
the Meissner effect, which is the total exclusion of magnetic fields from
the interior of the superconductor. Because the normal component of B
is continuous across a boundary, B = 0 on both sides of the boundary
between the superconductor and the region above it.
Suppose we place magnet (which well idealize as an ideal magnetic
dipole) above the superconductor, which is assumed to fill the half-space
below the xy plane. We can apply the method of images to work out the
field above the superconductor, and the resultant force on the dipole. We
need an image dipole below the xy plane which, together with the real
dipole above, gives a zero normal component of B at the surface.
Let the z axis extend along a line that goes through the upper dipole and is
normal to the xy plane. Suppose the upper dipole m1 is placed at a distance
h above the plane and makes an angle with the z axis, and we place an
image dipole m2 at a distance h below the plane, so that the angle between
m2 and the z axis is (that is, m2 is an exact mirror image of m1 , with
|m1 | = |m2 | m). The fields due to the dipoles are

0
(1) Bi = [3 (mi ri ) ri mi ]
4r3

where i = 1, 2. Now suppose we want the field at a point on the x axis, in


the xy plane. The distance r from both dipoles to this point is the same. The
normal (that is, z) component of the field is

(2) Biz = Bi z
0
(3) = [3 (mi ri ) (ri z) mi z]
4r3
Because the two dipoles are mirror images, their z components are equal
and opposite, but so are the z components of r1 and r2 . The result of this is
that angle between m1 and r1 is the same as the angle between m2 and r2
so
1
FLOATING A MAGNET ABOVE A SUPERCONDUCTOR 2

(4) m1 r1 = m2 r2
(5) r1 z = r2 z
(6) m1 z = m2 z
Therefore, B1z = B2z so the total normal field is zero everywhere on
the surface, no matter what the orientation of the upper dipole (provided its
image is a true mirror image). Therefore, the condition on the normal field
tells us nothing about the equilibrium orientation of the upper dipole. For
that, we need the torque on a dipole in a field, which is

(7) N = mB
The torque on the upper dipole due to the image is thus

(8) N = m1 B
0
(9) = [3 (m2 r2 ) (m1 r2 ) m1 m2 ]
4r3
Since both dipoles lie on the z axis, r2 points in the +z direction. The
relevant angles are: between m1 and r2 the angle is ; between m2 and r2
the angle is and between m1 and m2 the angle is = 2 .
To make things definite (and without loss of generality), we can take the two
dipoles to lie in the xz plane, with positive x components, so that all cross
products will lie along y. We get for the various products of vectors:

(10) m2 r2 = m cos ( ) = m cos


(11) m1 r2 = m sin y
(12) m1 m2 = m2 sin ( 2 ) y = m2 sin 2 y = 2m2 sin cos y
Putting it all together, we get

0  2
3m sin cos y 2m2 sin cos y

(13) N = 3
4r
0 2
(14) = m sin cos y
4r3
0 2
(15) = m sin 2 y
8r3
The torque is therefore zero at = 0, 2 , , that is, when the upper dipole
points straight up, straight down, or is horizontal (with any azimuthal an-
gle). To see which of these angles is stable, note that if 0 < < 2 , the
torque is in the +y direction, which tends to rotate the dipole towards the
FLOATING A MAGNET ABOVE A SUPERCONDUCTOR 3

horizontal. Similarly, if 2 < < , the torque is in the y direction, which


again rotates the dipole back towards the horizontal. Thus = 2 is the
stable position, and the magnet will orient itself so that it is horizontal.
What force does the magnet feel? The formula for the force on a mag-
netic dipole is

(16) F = (m B)
With both dipoles horizontal, we have for the field due to the image:

0
(17) B= [3 (m2 r2 ) r2 m2 ]
4r3
This time, m2 is fixed while r2 varies so to avoid confusion, well define
the angle between r2 and the +z axis to be . Since m2 is horizontal, the
angle between m2 and r2 is 2 so

0 h   i
(18) B = 3m cos r2 m2
4r3 2
0
(19) = [3m sin r2 m2 ]
4r3
Since m1 and m2 are parallel (they are actually the same vector), we then
get

0  2 2 2

(20) m1 B = 3m sin m
4r3
0 m2  2

(21) = 3 sin 1
4r3
We can now revert to the usual spherical coordinates (that is, well replace
by 2 , where the subscript 2 reminds us that the centre of the spherical
coordinate system is the image dipole) to get

(22) F = (m1 B)
0 m2 
 
2

(23) = 3 sin 2 1
4r3
30 m2  2
 60 m2
(24) = 3 sin 2 1 r2 + sin 2 cos 2 2
4r4 4r4
Were interested in the force when the dipole is at 2 = 0, where we get

30 m2
(25) F (2 = 0) = r2
4r4
FLOATING A MAGNET ABOVE A SUPERCONDUCTOR 4

Since this points in the +z direction, the force is repulsive, meaning that
the magnet floats above the superconductor. To get the height, we can give
the magnet a mass of M and equate the magnetic force with the gravitational
force:

30 m2
(26) = Mg
4r4
1/4
30 m2

(27) r =
4Mg
Since r is the distance from the image dipole, the distance above the xy
plane is half that, or

1/4
30 m2

1
(28) h=
2 4Mg
P INGBACKS
Pingback: Surface current induced by a magnet floating over a supercon-
ductor
EMF IN A SPINNING SPHERICAL SHELL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.45.
Heres a simple example of calculating the emf directly from its definition
as the work done on a unit charge over a path.


(1) E = f d`

Usually, the integral is taken around a closed loop such as an electric


circuit, but we can use the same definition to calculate the emf between two
points on a surface. Suppose we have a superconducting spherical shell of
radius a that is spinning on its axis (taken to be the z axis) with angular
speed . We impose a constant, uniform magnetic field B = Bz. If we look
at an infinitesimal horizontal slice through the sphere taken at a polar angle
of , we can think of this slice as a circular loop carrying a current. If the
sphere is spinning in the + direction (counterclockwise seen from above),
then the magnetic force per unit charge is

(2) f = vB
(3) = aB sin z

The vector z points horizontally outward. If we now want to find the


emf induced by the field between the north pole (at = 0) and the equator
(at = 2 ), we can do the integral above over a path along a line of longitude
between those two angles. Along this path

(4) d` = a (d )

The angle between z and is also so


1
EMF IN A SPINNING SPHERICAL SHELL 2


(5) E = f d`
/2
2

(6) = a B sin z
0
/2
(7) = a2 B sin cos d
0
/2
1 2
(8) = a B sin 2 d
2 0
1 2
(9) = a B
2
Note that this is the same emf as that produced in a flat, circular disk of
radius a rotating at angular speed (Griffiths, example 7.4) so the sphere
problem effectively reduces to the disk problem by projecting the sphere
onto its equatorial plane.
DROPPING LOOPS THROUGH MAGNETIC FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.46.
Heres a questionable variant on the problem of a metal loop falling
through a magnetic field. We begin, as before, with a square loop of side
length a, except this time we rotate it by /4 so that it makes a diamond
shape. In the diagram, the magnetic field points out of the page at all points
above the horizontal line, and the loop moves downward under gravity, so
the force per unit charge f points to the left, as shown in the figure:

If we now drop it from rest, what happens? As before, we first calcu-


late the emf around the square, and take the path of integration as counter-
clockwise. For the top left side of the square, which is entirely within the
magnetic field, we have


vBa
(1) Euple f t = f d` =
2

By symmetry, we get the same contribution for the upper right, so the
emf for the top two edges is
1
DROPPING LOOPS THROUGH MAGNETIC FIELDS 2


(2) Etop = 2vBa
The bottom two edges are only partially within the field, so we need to
find the distance d. The ratio of d to a full side a is the same as the ratio of
the height of the centre of the square y to half a diagonal. That is:

d y
(3) =
a a/ 2

(4) d = 2y
The emf from the bottom two edges opposes that from the top two edges,
so

vBd
(5) Ebottom = 2 = 2vBy
2
The total emf is thus
 
(6) E = vB 2a 2y

If the resistance of the loop is R, this emf generates a current according


to Ohms law:

E vB  
(7) I=
= 2a 2y
R R
Weve seen that the force on a current loop that is partially enclosed in a
magnetic field is

(8) F = IwB

where w is the length of the chord at the edge of the field. Here, w is as
shown in the diagram, and by similar triangles

a y
w 2
(9) =
2a a
2

(10) w = 2a 2y
Thus the force on the loop is
DROPPING LOOPS THROUGH MAGNETIC FIELDS 3

vB2  2
(11) F= 2a 2y
R
If we follow the same argument here as we did for the square loop with
its edges parallel to the edge of the field, we then say

(12) mv = mg F
vB2  2
(13) = mg 2a 2y
R
For the terminal velocity, we set v = 0, so we get

mgR
(14) vterm2 =  2
B2 2a 2y
The original terminal velocity was

mgR
(15) vterm1 =
a2 B2

so the ratio is

vterm1  y 2
(16) = 22
vterm2 a

which agrees with the answer in Griffiths. However, this derivation doesnt
really make any sense, since the velocity v = y, so if v = 0, this implies
that v is a constant (presumably not zero). In that case, y is continually
decreasing. However, 14 implies that if v = vterm2 = constant, y must also
be a constant, giving a contradiction.
I think to treat this properly, we need to write the equation of motion
entirely in terms of y, giving

B2  2
(17) my = mg + y 2a 2y
R

with initial conditions y (0) = a/ 2 and y (0) = 0. Not surprisingly, this
non-linear ODE cannot be solved analytically, but using Maple, we can get
a numerical solution, which is shown for m = 1, g = 9.8, B = 1, a = 1 and
R = 0.01. The plot shows y (t) (in red) and y (t) (in green).
DROPPING LOOPS THROUGH MAGNETIC FIELDS 4


This shows the portion where y decreases from a/ 2 to zero (that is, as
the lower half of the diamond falls across the boundary of the field).The
spike in v around t = 5 is actually a smooth peak as can be seen with a more
detailed graph. Also, the sharp corner in v just after t = 0 is also smooth.
The latter looks a bit like the velocity rising rapidly and then levelling off
and in fact, plugging the numbers into 14, we see that v at this point does
actually level off (roughly) at a value close to
vterm2 . The spike in v from the
solution of the ODE occurs around y = a/ 2, at which point vterm2 ,
which is clearly nonsense, so the calculations here give only an approxima-
tion to the true behaviour.
Using similar approximate reasoning, we can work out what happens if
we drop a circular loop of radius a in the same field. This time well let y
be the vertical distance between the circles centre and the edge of the field,
and describe a point on the circle by an angle , taking = 0 to be the point
at the top of the circle.
DROPPING LOOPS THROUGH MAGNETIC FIELDS 5

The field covers the circle from = y to = +y , where y = 2 + ,


and = arcsin ay . The force f still points to the left and v points down,
but now d` is tangent to the circle, pointing counterclockwise. The angle
between f and d` is , so the emf is

y
(18) E = vB cos (a d )
y
y
(19) = 2 vB cos (a d )
0
(20) = 2avB sin y
 y
(21) = 2avB sin + arcsin
2 y
a
(22) = 2avB cos arcsin
r a
 y 2
(23) = 2avB 1
a
From this we get the current:

E
r  y 2
2avB
(24) I= = 1
R R a
To get the magnetic force on the loop, we need the chord w, but this is
r  y 2
w
(25) = a cos = a 1
2 a
DROPPING LOOPS THROUGH MAGNETIC FIELDS 6

The force is therefore

(26) F = IwB
4a2 vB2
  y 2 
(27) = 1
R a
Again, taking v = 0 we get

4a2 vB2
  y 2 
(28) 1 = mg
R a
mgR
  y 2 1
(29) v = 1
4a2 B2 a
The time taken for the loop to fall completely out of the field area is found
from v = dy/dt
a 
4a2 B2  y 2 
(30) t = 1 dy
mgR a a
16a3 B2
(31) =
3mgR
FARADAY FIELD AND MAGNETIC VECTOR POTENTIAL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 5.50a, 7.47.
Weve seen that in the magnetostatic case, the Biot-Savart law gives the
magnetic field produced by steady currents:

0 J (r0 ) (r r0 )
(1) B (r) = 3
d 3 r0
4 V |r r0 |
This law can be used to derive Ampres law which in differential form
is

(2) B = 0 J
The Biot-Savart law can be seen as the solution of Ampres law, com-
bined with the divergence equation B = 0.
In the case where the magnetic field is changing with time, Faradays law
gives the electric field produced as a result of the changing magnetic field:

B
(3) E =
t
If there is no free charge, then E = 0, so this pair of equations for the
electric field are formally equivalent to those above for the magnetic field.
We can therefore write down an analogous solution by substituting tB for
0 J in the Biot-Savart law to get

1 B (r0 ) (r r0 )
(4) E= 3
d 3 r0
4 tV |r r0 |
We can extend the same argument to the magnetic vector potential A. We
have B = A and we can choose A = 0, so again by analogy with the
Biot-Savart law we can write

1 B (r0 ) (r r0 )
(5) A= d 3 r0
4 V |r r0 |3
Comparing the last two equations, we get
1
FARADAY FIELD AND MAGNETIC VECTOR POTENTIAL 2

A
(6) E=
t
Taking the curl of both sides gives us back Faradays law: E =
A
t = tB . Remember that this is true only for those electric fields
produced as a result of changing magnetic fields; it does not apply to elec-
tric fields produced by free charge.
Suppose now we have a spherical shell of radius R with a uniform surface
charge density which spins at a variable rate (t). Griffiths works out
the vector potential of this sphere in his example 5.11 and gets
(
0 R
3 r sin rR
(7) A= 0 R4 sin
3 r2
rR
Inside the sphere there is no Coulomb field, while outside, this field is

Q
(8) Ecoulomb = r
40 r2
4R2
(9) = r
40 r2
R2
(10) = r
0 r2
Using 6 for the Faraday field, we get the total field:
( R
0 3 r sin rR
(11) E= R2 4

0 r2
r 0 R3 sinr2 rR
CHANGING A PARTICLES SPEED IN A CYCLOTRON

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.48.
We can now revisit the problem of a charged particle in a cyclotron field.
Suppose we start with a charged particle of mass m and charge q at rest a
distance R from the axis of the cyclotron, and we want to increase the speed
of the particle in its circular orbit while keeping it at the same radius (such
a device is called a betatron). We can do this by varying the magnetic field
B (t) such that the cyclotron relation is always satisfied:

mv2
(1) qvB =
R
mv
(2) B =
qR
Taking the time derivative, we get

mv
(3) B =
qR
The changing magnetic field induces a circumferential electric field E,
and since this field is parallel to the particles direction of motion it will act
as the force that accelerates the particle. From Newtons law, F = mv = qE,
so

E
(4) B =
R
If we assume the cyclotron has cylindrical symmetry, then we can inte-
grate this equation along the particles orbit, along which both fields are
constant in magnitude. That is

E
(5) Bc d` = d`
R

(6) Ed` = R Bc d`

(7) = 2R2 Bc
1
CHANGING A PARTICLES SPEED IN A CYCLOTRON 2

This equation applies on the circumference of the orbit only (not in the
interior of the orbit), so weve added a suffix c to Bc to emphasize this point.
From Faradays law in integral form, we have also that the integral of the
electric field is given by the change in flux:

d
(8) Ed` =
dt

(9) = B da

where now we are integrating over all points within the orbit.
If we start with the particle at rest in zero field, then we have

(10) 2
2R Bc = B da = R2 B

where B is the average field across the orbit.


A word about the signs here. Suppose the magnetic field points in the z
direction (as shown in Fig. 7.52 in Griffiths), and we take the area vector
to point in the +z direction. Further, if the magnetic field is increasing in
magnitude, then B points towards z as well. Thus B da < 0, giving the
sign shown.
We can integrate both sides to some time t to get

1
(11) Bc (t) = B (t)
2
Thus we can speed up (or slow down, by decreasing the field) the particle
by keeping the average field equal to twice the field at the radius of the orbit.
ACCELERATING AN ATOMIC ELECTRON WITH A
CHANGING MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.49.
This is another example of the cyclotron field, though this time were
dealing with an electron orbiting an atomic nucleus. The treatment here is
purely classical, so its certainly not a valid model of how a magnetic field
interacts with an atom (for that we need quantum mechanics). However,
lets have a look and see where the argument takes us.
Suppose an electron of charge q orbits a nucleus of charge Q, and ini-
tially there are no external magnetic or electric fields. Equating centripetal
and Coulomb forces, we get

mv2 Qq
(1) =
r 40 r2

Now suppose we turn on a magnetic field perpendicular to the plane of


the orbit, in the +z direction. In time dt the field increases by an amount
dB, so that the force equation now becomes (assuming r doesnt change,
which well justify below):

mv21 Qq
(2) = qv1 dB
r 40 r2

where v1 is the new velocity of the electron. The minus sign occurs because
were taking the direction of the orbit to be the + direction (clockwise as
viewed from above). This results in a force pointing radially outwards (for
positive q), so it detracts from the attractive Coulomb force. Griffiths works
out the difference v1 v in his section 6.1.3, but we wont need the actual
value here; rather well just say that v changes by a small amount so that
v1 v = dv. To first order:
1
ACCELERATING AN ATOMIC ELECTRON WITH A CHANGING MAGNETIC FIELD 2

mv21 mv2 mv
(3) = + 2 dv
r r r
Qq
(4) = qv dB
40 r2
mv2
(5) = qv dB
r
The change in kinetic energy due to the change in v is therefore

1 2 1 2
(6) dT = mv mv
2 1 2
1
(7) = qrv dB
2
Since magnetic forces always act perpendicular to the direction of mo-
tion, they never do any work, so this change in kinetic energy must come
from somewhere else. Since a changing magnetic field gives rise to an elec-
tric field by Faradays law, and electric forces can do work, that must be the
source of the change in energy. If we can show that this induced electric
field gives exactly this energy, then weve justified the assumption that r
doesnt change. Faradays law in integral form says

d
(8) E d` =
dt
Integrating this over the circle of the orbit, we get

dB
(9) 2rE = r2
dt
The electric field causes a force on the electron:

qr dB
(10) FE = qE =
2 dt
Since E acts in the + direction, the force is parallel to the direction of
motion. Over a time dt, this force will do work equal to FE ds, where ds
is the distance travelled in that time. Since the velocity changes from v to
v + dv and to first order, this change is linear, the average velocity over that
time is 12 (v + v + dv) = v + 12 dv. The distance, again to first order, is then
 
1
(11) ds = v + dv dt = v dt
2
ACCELERATING AN ATOMIC ELECTRON WITH A CHANGING MAGNETIC FIELD 3

so the work done, which is equal to the change in kinetic energy, is

(12) dT = FE ds
qr dB
(13) = (v dt)
2 dt
qvr
(14) = dB
2
This is exactly the change in kinetic energy we worked out earlier, so the
electric force provides the required work, provided that r doesnt change.
A BIZARRE EXAMPLE OF FARADAYS LAW

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.50.
Romer, Robert. H. (1982), Am. J. Phys. 50, 1089.
This is a rather bizarre illustration of Faradays law. Suppose we set up
the circuit as shown in the figure:

The circle marked S is a solenoid perpendicular to the plane of the screen,


and V1 and V2 are two ideal voltmeters (in the sense that they have infinite
resistance and thus draw no current). The current through the solenoid is
varied at a constant rate such that the magnetic flux in the interior of the
solenoid varies according to (t) = t, where is a constant. According
to Faradays law, the changing flux induces an electric field such that


(1) E d` =

for any path enclosing the solenoid. If the magnetic field points into the
screen, then by the right hand rule, the current through the solenoid runs
clockwise around the solenoid. By Lenzs law, the induced electric field
opposes the flux change, so it runs counterclockwise around the loop con-
taining the two resistors, and the current is given by

(2) = I (R1 + R2 )

The voltage drop across R1 is therefore


1
A BIZARRE EXAMPLE OF FARADAYS LAW 2

(3) V1 = IR1
R2
(4) =
R1 + R2
That is, the voltage drops by this amount when we travel up the branch of
the circuit containing R1 . Similarly, if we travel down the branch containing
R2 , the voltage drop is

R2
(5) V2 =
R1 + R2

where the minus sign is because both voltmeters are measuring the voltage
for a path travelling upwards.
This result seems bizarre because both meters are essentially connected
to the same points in the circuit. In his paper, Romer does an experiment in
which he actually measures the two voltages and shows that they do actually
behave this way. The key point seems to be that what we are measuring is
the actual integral E d` along a specific path, so if we put both voltmeters
on the same side of the solenoid, then they would read the same.
SUPERCONDUCTING LOOP IN A MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.51.
When we introduced the idea of motional emf, we used an example of
a circuit containing a resistance R. What happens if the circuit contains no
resistance (as with a superconductor)? In this case, we need to consider
the self-inductance of the circuit. As a simple example, suppose we have
a rectangular loop of wire with dimensions of h and ` (height and length).
The loop is placed in a magnetic field that is perpendicular to the plane of
the loop, and which covers a distance x < ` along the length. If the loop
is free to move parallel to its length, what motion does it follow, assuming
that it is given an initial velocity v0 ?
The motion of the loop within the field causes a current to flow due to
the Lorentz force qv B. This force is perependicular to the wire along the
top and bottom edges of the loop, so it is only the end of the loop within
the field that experiences the force, since here the Lorentz force is parallel
to the wire. Suppose the current is I. This current flows up the end of the
loop, which means that an additional component of the Lorentz force acts.
If the field points into the page and the current is flowing upwards, then this
force is to the left. If the speed of the charges along the wire is u, the total
force on the end of the loop is huB, where is the linear charge density.
However, the current is the amount of charge that flows past a point in unit
time, so the total charge in that end of the loop is I = u so the force is IhB.
Now, the self-inductance of a loop relates the magnetic flux through the
loop to the current:

d
(1) = LI
dt
The rate of change of flux is

d
(2) = hBx
dt

so
1
SUPERCONDUCTING LOOP IN A MAGNETIC FIELD 2

Bh
(3) I = x
L
Bh
(4) I = (x x0 )
L

where x0 is the position at time t = 0. Thus I < 0 (upwards) as the loops


edge nears the edge of the field, and I > 0 (downwards) when the edge is far
from the edge of the field. If we define the positive direction as the direction
of increasing x, this means that positive is to the left here. This means that
a negative current gives a positive force, so

(5) F = IhB
B2 h2
(6) mx = (x x0 )
L
This has the general solution

(7) x (t) = A sin (t + 0 ) + x0

with the frequency given by

Bh
(8) =
mL
MUTUAL INDUCTANCE: CALCULATION FROM THE
INTEGRAL

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 7.52.
The general formula (known as the Neumann formula) for mutual induc-
tance is

0 d`1 d`2
(1) M=
4 |r2 r1 |
Although this formula isnt used much for direct calculation of M, we
can revisit our initial mutual inductance problem, in which we have two
parallel, circular loops sharing a common axis and separated by a distance
z. The upper loop has radius a and the lower loop b. The integral is easiest
if we use cylindrical coordinates. Both the line increments are in the
direction with the upper increment having magnitude d`1 = a da and the
lower one d`2 = b db . The angle between the two increments is a b
and both angles are integrated over the range [0 . . . 2]. Therefore

(2) d`1 d`2 = ab cos (a b ) da db


For the denominator, we have

h i1/2
(3) |r2 r1 | = (a cos a b cos b )2 + (a sin a b sin b )2 + z2
1/2
= z2 + a2 + b2 2ab cos (a b )

(4)
If we define

ab
(5)
z2 + a2 + b2

we get
r
1 1
(6) = p
|r2 r1 | ab 1 2 cos (a b )
1
MUTUAL INDUCTANCE: CALCULATION FROM THE INTEGRAL 2

We need to do the integral


p
0 ab 2 2 cos (a b ) da db
(7) M= p
4 0 0 1 2 cos (a b )
This has no closed form solution, but we can expand the denominator in
a Taylor series to get (defining c cos (a b )):

1 1
(8) p =p
1 2 cos (a b ) 1 2 c
3 5
(9) = 1 + c + ( c)2 + ( c)3 + . . .
2 2
The integral is then
p 
ab 2 2

0 2 3 2 3 5 3 4
(10) M = c + c + c + c + . . . da db
4 0 0 2 2
p  
0 ab 15
(11) = 0 + 2 2 + 0 + 3 2 + . . .
4 4
p  
0 ab 15 2
(12) = 1+ +...
2 8
If the upper loop is much smaller than the lower one and the distance
between them (a  b and a  z), then to first order in a, we get
ab/ b2 + z2 , 2 0 and

0 a2 b2
(13) M
2 (b2 + z2 )3/2

which agrees with the earlier result.


TRANSFORMERS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 7.53, 7.54.
An electrical transformer (as opposed to Optimus Prime) is a device that
uses mutual induction to transform one voltage and current to another with
the same average power. To see how this works in a simplified model, sup-
pose we have two wires that are wound around the same cylinder, with wire
1 having N1 turns and wire 2 N2 turns. If wire 1 is driven by an alternating
voltage, then (assuming no resistance in wire 1), the back emf must balance
the driving voltage, and is determined by

d1
(1) E1 =
dt

where 1 is the total magnetic flux through wire 1. If the flux per turn is
then 1 = N1 and

d
(2) E1 = N1
dt
If the windings of the two wires are arranged so that all loops of both
wires contain the same flux, then the induced emf in wire 2 is

d2 d
(3) E2 = = N2
dt dt

so

E2 N2
(4) =
E1 N1
That is, the voltage is stepped up or down, depending on the ratio of the
number of turns in each coil. In order to conserve energy, the power P = E I
must remain constant, so an increase in voltage means a decrease in current,
and vice versa. To see this, suppose the driving voltage in wire 1 is

(5) Vin = V1 cos t


1
TRANSFORMERS 2

and that the self inductances of the two coils are L1 and L2 , and their mutual
inductance is M. Then the back emf in wire 1 must match Vin , and it results
from the change in the current I1 in wire 1 (via the self inductance L1 ) and
from the change in I2 (via the mutual inductance M), so we get

(6) E1 = Vin = V1 cos t = L1 I1 + M I2


If wire 2 is connected to a resistor R, then the voltage drop across the
resistor must equal the back emf in wire 2 (as there is no driving voltage),
so

(7) E2 = Vout = I2 R = L2 I2 + M I1
To solve these two equations, we need a relation between M, L1 and L2 .
The self inductance per unit length of a solenoid of radius r and n turns per
unit length is

(8) L = r2 0 n2

so for our two intermeshed solenoids (assumed to be the same length `):

r2 2
(9) L1,2 = 0 N1,2
`
Using the same argument, the mutual inductance between two solenoids
is found as follows. The field due to solenoid 1 is B1 = 0 N`1 I1 . This pro-
duces a flux per turn through solenoid 2 of t = r2 0 N`1 I1 , for a total flux
through 2 of = r2 0 N`1 N2 I1 . The mutual inductance is therefore

N1
(10) M = r2 0 N2
`
Comparing the two, we see that

(11) M 2 = L1 L2
Substituting this into 6 and 7 above, we get


(12) V1 cos t = L1 I1 + L1 L2 I2

(13) I2 R = L2 I2 + L1 L2 I1
TRANSFORMERS 3
p
Mutliplying the first equation by L2 /L1 and subtracting the second, we
find
r
L2
(14) V1 cos t = I2 R
L1
r
L2 V1
(15) I2 (t) = cos t
L1 R
The current in wire 1 is then
r
V1 L2
(16) I1 = cos t I2
L1 L1
V1 L2 V1
(17) I1 (t) = sin t + cos t
L1 L1 R

assuming the constant of integration in the last line is zero (which means
that I1 has no DC component).
The voltage in wire 2 is thus
r
L2
(18) Vout = I2 R = V1 cos t
L1
The ratio of the two voltages is
r
Vout L2 N2
(19) = =
Vin L1 N1

using 9.
Finally, we can work out the power in each wire.

(20) P1 = I1Vin
V12 L2 V12
(21) = sin t cos t + cos2 t
L1 L1 R
(22) P2 = I2Vout
L2 V12
(23) = cos2 t
L1 R
If we integrate these two powers over a full cycle (for t = 0 . . . 2/) the
first term in P1 integrates to zero and the last term in P1 equals P2 so the
average power in each wire is the same.
TRANSFORMERS 4

2/
L2 V12 L2 V12
(24) hP1 i = hP2 i = cos2 t dt =
L1 R 2 0 2L1 R
MAXWELLS EQUATIONS WITH VARYING CHARGE BUT
CONSTANT CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 7.55.
An interesting situation in electrodynamics is one where the current den-
sity J is constant in time, but the charge density isnt. Clearly this means
that charge is piling up somewhere, which could happen when charging a
capacitor. However, the rate at which changes is constrained in this situ-
ation, as we can show using Maxwells equations. First, from Gausss law:

1
(1) E =
0
Then from Faradays law

(2) ( B) = 0 J + 0 0 E
Since the divergence of a curl is always zero, we get

1
(3) E = J
0
(4) = J
Since J is independent of time, we can integrate this directly to get

(5) (r,t) = J (r)t + (r, 0)

so the charge density is a linear function of the time.


In this situation, the Biot-Savart law still holds, as we can see by using
Maxwells equations. The law is

0 J (r0 ) (r r0 )
(6) B (r) = d 3 r0
4 |r r0 |3
First, we can check the extended version of Ampres law:
1
MAXWELLS EQUATIONS WITH VARYING CHARGE BUT CONSTANT CURRENT 2

E
(7) B = 0 J + 0 0
t
Taking the curl of the Biot-Savart law, we get


0 J (r0 ) (r r0 )
(8) r B (r) = r d 3 r0
4 |r r0 |3

where the subscript r reminds us that the curl is with respect to the unprimed
coordinates.
Using a product rule we get

(9) " # !
J (r0 ) (r r0 )   (r r0 ) (r r0 )
= J r0 0

r + J r
|r r0 |3 |r r0 |3 |r r0 |3

The divergence in the second term on the RHS can be written in terms
of the 3-d delta function. Again, remember that were taking the derivative
only with respect to r components, so the presence of the r0 serves merely
to shift the origin, so we have

(r r0 )
= 43 r r0

(10)
|r r0 |3
We can work out the integral of the first term on the RHS using the same
technique that was applied to the original, magnetostatic version of Am-
pres law. If you refer back to that derivation, we transformed the x com-
ponent of the integrand to the form

(11) !
 x x0 x x0 x x0
J r0 r0 0
J r0
  
= r 0 J r + r0
|r r0 |3 |r r0 |3 0
|r r |3

where the derivatives are now with respect to the primed coordinates. The
second term on the RHS can be converted to a surface integral using the
divergence theorem and, if the currents are localized, works out to zero. In
the magnetostatic case r0 J (r0 ) = 0, but here that isnt true, as we saw
above. However, if we return to three dimensions are look again at the
integral, we get
MAXWELLS EQUATIONS WITH VARYING CHARGE BUT CONSTANT CURRENT 3


r r0 0 3 0 r r0
d 3 r0

(12) r0 J r d r =
|r r0 |3 |r r0 |3

r r0 3 0
(13) = d r
t |r r0 |3
However, the integral is essentially the definition of the electric field:

1 r r0
(14) E= d 3 r0
40 |r r0 |3
Combining this with 10, we get

(15) ( B) = 0 J + 0 0 E

which is just Ampres law again.


P INGBACKS
Pingback: Jefimenkos equation for time-dependent electric field
DISPLACEMENT CURRENT IN AN INFINITE WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 7.56.
Terry, William K. (1982), Am. J. Phys. 50, 742.
This is an illustration of the role of Maxwells displacement current in
Ampres law

(1) B = 0 J + 0 Jd

with the displacement current defined as

E
(2) Jd 0
t
Consider the case of an infinite straight wire carrying a current I. This is
a magnetostatic case, and as such we can calculate the magnetic field from
Ampres law with E/t = 0 and get the usual formula

0 I
(3) B=
2r
In this case the displacement current is zero. Suppose, however, we in-
troduce a small gap of width in the current. (This is an idealized situation
in which we picture a moving line of charge where the charge density is
everywhere except in the gap, where there is no charge.) We can simulate
this situation by considering a tiny segment of wire of length carrying the
opposite charge density travelling at a speed v, where I = v. Superim-
posing this wire segment onto the infinite wire gives a small gap travelling
along at speed v. Because of the small gap in the wire, E/t is no longer
zero as this segment moves along. Well place the wire along the z axis and
calculate the field in the xy plane. From the integral form of Ampres law
we have, for an integration path around a circle of radius R in the xy plane:


2
(4) RB = J da + E da
0 t
1
DISPLACEMENT CURRENT IN AN INFINITE WIRE 2

Consider a time when the gap straddles the origin. The net current across
the surface is then J = 0, so any contribution to the magnetic field must
come from the changing electric field. The infinite wire produces a constant,
radial electric field so contributes nothing to E/t, so we can look at the
field due to the short segment of wire.
For a segment of wire of length dz at position z, the z component of field
(all were interested in, since its E da we have to calculate) at a point on a
circle of radius s in the xy plane is

1 dz z
(5) dEz = 2 2

40 (s + z ) s2 + z2
Weve used z in the numerator to get the direction of the field right.
If dz is in the region z < 0 then Ez is negative (points towards z) since
the charge is negative. If we take da to point in the +z direction, we want
E da < 0 in this case. Similarly, if z > 0, Ez should be positive, since the
field points towards +z. Therefore

1 z dz
(6) dEz =
40 (s2 + z2 )3/2
The total field at a point in the xy plane is thus (assuming that the leading
edge of the segment crosses z = 0 at t = 0):
vt
z dz
(7) Ez = 3/2
40 vt (s2 + z2 )
 1/2  1/2 
2 2 2 2
(8) = s + (vt ) s + (vt)
40
The total flux through the circle from the segment is then

(9) E = E da

2 R  2 2
1/2 
2 2
1/2 
(10) = s + (vt ) s + (vt) s ds
40 0
q q  R
2 2 2 2

(11) = s + (vt ) s + (vt)
20 0
At this point were faced with a bit of a conundrum, for when evaluating
the s = 0 limit, we need to choose whether to take the positive or negative
square root. If the segment straddles the origin and we take the positive root
in both cases, then we get
DISPLACEMENT CURRENT IN AN INFINITE WIRE 3

q q 
2 2 2 2
(12) E = R + (vt ) R + (vt) ( vt) + vt
20
q q 
2 2 2 2
(13) = R + (vt ) R + (vt) + 2vt
20
This result looks a bit dubious since the flux doesnt go to zero when
0, which you would intuitively think it must. The catch appears to
be that if were requiring the segment to straddle the origin, then the inte-
grand has a singularity when s = vt = 0. Indeed, in the original paper by
Terry, he doesnt derive the result this way; rather he begins with a point
charge moving along the z axis with the result that the flux contains a step
function when the charge passes through the circle. This in turn gives the
displacement current a delta function at the origin.
In any event, the displacement current here is

dE
(14) Id = 0
dt
v2 t t
(15) = v+ q q
2 2 2
R2 + (vt ) R2 + (vt)
As 0, Id v = I, so in theory, the displacement current provides
the magnetic field in the gap.
I have to confess that I dont find this derivation particularly convincing,
due to the fudging of the limits of integration and the sweeping under the
carpet of singularities. Terrys original treatment is much clearer.
ELECTRIC FIELD OUTSIDE AN INFINITE WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 7.57.
The electric field inside a resistor with a constant cross-section is given
in Griffithss Example 7.1 as

I
(1) E= z
A

where in this case is the resistivity (not the charge density!), I is the
current and A is the cross-sectional area of the resistor. If we consider the
simple case where the cross section is circular and make the resistor very
long (so we can call it a wire with radius a), then

I
(2) E= z
a2
Surprisingly, the problem of finding the field outside the wire is not com-
pletely solved. If we assume that the current returns through a supercon-
ducting coaxial cylinder of radius b around the wire, and that the magnetic
field is constant in time, so that E = 0 and thus E = V , then we can
apply Laplaces equation in cylindrical coordinates to find the potential. In
order to solve it, we need boundary conditions. Since the coaxial cylinder is
a perfect conductor, V = 0 there. On the surface of the inner wire, we have

Iz
(3) V (a, z) =
a2
The problem is, what is V in between the two cylinders? The solution we
derived earlier was for the case where V was independent of z:

(4)
1
n
V (r, ) = A ln r +B+ r (An sin n + Bn cos n )+ rn (Cn sin n + Dn cos n )
n=1 n=
1
ELECTRIC FIELD OUTSIDE AN INFINITE WIRE 2

However, in the special case where V (r, z) = z f (r), this solution also
works for finding f (since using separation of variables on all three cylin-
drical variables separates the z from the rest of V ). Since there is no
dependence, both sums disappear and we are left with

(5) V (r = b) = 0 = A ln b + B
(6) B = A ln b
On the wire:

Iz
(7) V (a) =
a2
(8) = A ln a + B
a
(9) = A ln
b
Iz
(10) A = 2
a ln (a/b)
So

Iz ln (r/b)
(11) V (r, z) =
a2 ln (a/b)
The field is then

(12) E = V
I hz r i
(13) = r + ln z
a2 ln (a/b) r b
The surface charge density on the wire is found from the difference in
radial components of the field. Inside the wire there is no radial component,
so

I0 z
(14) = 0 Er = 3
a ln (a/b)
As Griffiths remarks, the results for the field and surface charge are pe-
culiar (I would say wrong) since they depend on z, which shouldnt be
the case for an infinite wire with no variation along its length. It is surely
impossible since the results depend on where we set the origin, which is
completely arbitrary.
TRANSMISSION LINES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 7.58.
In this problem, weve got a transmission line consisting of two thin rib-
bons (each of width w) separated by an insulating vacuum space of thick-
ness h  w. Weve looked at a similar problem before, so well start from
there. If we assume that the magnetic field is confined to the region between
the ribbons (in reality there will be some fringe effects, but if the ribbons
are very close together, these will be small), then the field per unit length is

(1) B = 0 v

where is the surface charge density and v is the speed of the current I, so
that I = vw and

0 I
(2) B=
w
The energy in a magnetic field is

1 2 1
(3) WB = LI = B2 d 3 r
2 20
The volume between a unit length of ribbons is hw and we can take the
field to be constant in this volume, so we can get the inductance per unit
length:

0 I 2
 
1 2 1
(4) LI = hw
2 20 w
h
(5) L = 0
w
The capacitance of a unit length can be found by treating the cable as a
parallel plate capacitor:

0 w
(6) C=
h
1
TRANSMISSION LINES 2

The product is thus:

1
(7) LC = 0 0 =
c2

where c is the speed of light.


If the ribbons are separated by a material with permittivity and perme-
ability then we get (see Griffiths, Example 4.6):

0 w w
(8) C= =
0 h h
For the inductance, we
can start from the definition of inductance, which
is = LI. The flux is B da and within matter, B = Bvac /0 so L must
scale the same way:

h h
(9) L= 0 =
0 w w

and the product is

(10) LC =

If the velocity is taken as v = 1/ LC we get

1
(11) v=

P INGBACKS
Pingback: Poyntings theorem
ALFVENS THEOREM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 7.59.
Starting from Ohms law in the form

(1) J = (E + v B)

we can derive Alfvens theorem, which states that in a perfectly conducting


fluid, the magnetic field lines get frozen, which means that if we take some
closed loop within the fluid, the magnetic flux through that loop remains
constant as the loop gets carried along by the fluid.
Since the fluid is a perfect conductor, the conductance = so if J is
finite, we must have

(2) E = v B
Applying Faradays law we get

B
(3) E = = (v B)
t
Now consider a loop P enclosing a surface S at time t. Over a time
interval dt, S moves a distance v dt with the fluid, and ends up as a surface
S 0 surrounded by loop P 0 . The surface between S and S 0 forms a ribbon
R. The change in flux through the loop is then

(4) d = B (t + dt) da B (t) da
S0 S
The three surfaces S (the minus means we are reversing the surface
normal so it points outward from the combined volume), S 0 and R taken
together form a closed surface, and if we integrate B (t + dt) over this closed
surface, we can use the divergence theorem to say:

(5) B da = Bd 3 r = 0
S +S 0 +R V
1
ALFVENS THEOREM 2

since B = 0 always. Therefore



(6) B (t + dt) da + B (t + dt) da = B (t + dt) da
S0 R S
Plugging this back into 4 we get

(7) d = B (t + dt) da B (t) da B (t + dt) da
S S R

B
(8) = dt da B (t + dt) da
S t R
Keep in mind that the ribbon R is a differential area, so the differentials
in this equation still balance out.
The differential area is a parallelogram bounded by a line element d`
around the loop P and the vector vdt, so

(9) da = d` vdt
The integral over da is thus equivalent to adding up these area elements
as we move around the loop P. That is

(10) B (t + dt) da = B (t + dt) (d` vdt)
R P
To first order in dt, this integrand is

B (t + dt) (d` vdt) = B (t) (d` vdt) + O dt 2



(11)

so we can write it as

(12) B (t + dt) da = dt B (t) (d` v)
R P
Using the vector triple product rule (we can cyclically permute the three
vectors), we get

(13) dt B (t) (d` v) = dt d` (v B (t))
P P
Since were now integrating v B around the boundary of surface S , we
can use Stokess theorem:
ALFVENS THEOREM 3


(14) dt d` (v B (t)) = dt (v B (t)) da
P S
We therefore get finally:

B
(15) d = dt da dt (v B (t)) da
S t S
 
d B
(16) = da (v B (t)) da
dt S t
Using 3 we get finally

d
(17) =0
dt

so the flux through the loop doesnt change.


DUALITY TRANSFORMATION FOR MAGNETIC AND
ELECTRIC CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 7, Post 60.
Continuing our excursion in the fantasy world where magnetic monopoles
exist, we can explore the duality transformations, which describe a rotation
in an E-B space, as follows:

(1) E0 = E cos + cB sin


(2) cB0 = cB cos E sin
(3) cq0e = cqe cos + qm sin
(4) q0m = qm cos cqe sin


where is an arbitrary angle and c = 1/ 0 0 . Currents and charge den-
sities transform the same way as the corresponding point charges.
Its a straightforward, though tedious, exercise to verify that Maxwells
equations (including magnetic charge) are invariant under these transforma-
tions. The two equations that are different are:

(5) B = 0 m
B
(6) E = 0 Jm
t
A few sample calculations should show how it goes. Well do one involv-
ing a divergence:

(7) c B0 = c B cos E sin


e
(8) = c0 m cos sin
0
(9) = c0 (m cos ce sin )
(10) = c0 m0
And another involving a curl:
1
DUALITY TRANSFORMATION FOR MAGNETIC AND ELECTRIC CHARGE 2

(11) E0 = E cos + c B sin


   
B E
(12) = 0 Jm cos + c 0 Je + 0 0 sin
t t
B 1 E
(13) = 0 (Jm cos cJe sin ) cos + sin
t c t
B0
(14) = 0 J0m
t
The other two equations transform similarly.
If we take the force law for electric and magnetic charges to be
 
1
(15) F = qe (E + v B) + qm B 2 v E
c

then the transformed law is


 
0 1
= q0e 0 0
+ q0m
0 0

(16) F E +vB B 2vE
c
 qm 
(17) = qe cos + sin
 c  
1
(18) E cos + cB sin + v B cos E sin +
c
(19) (qm cos cqe sin )
 
1 1
(20) B cos E sin 2 v (E cos + cB sin )
c c
 
1
(21) = qe (E + v B) + qm B 2 v E
c
(22) =F
To get the penultimate line it is just a matter of multiplying out the first
expression and using cos2 + sin2 = 1, and cancelling off terms.
P INGBACKS
Pingback: Magnetic monopole force law
Pingback: Radiation from a magnetic dipole composed of monopoles
Pingback: Fields of a moving magnetic monopole
POYNTINGS THEOREM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.1.
If we want to calculate the work done by electromagnetic forces on a
collection of charges and currents, we can start with the Lorentz force law

(1) F = qE + qv B
In order for work to be done, the force has to act over a displacement
of the charge, so in time dt the charge moves vdt. As the magnetic force
always acts perpendicular to the velocity, it does no work so all the work
comes from the electric field:

(2) dW = F vdt = qE vdt


If we now consider a continuous distribution of charge with density
then we can find the work done on a volume element d 3 r by replacing q with
d 3 r. Also, the current density J is the charge density times the velocity,
J = v, so we get

d 3 r E vdt

(3) dW =
(4) = E Jd 3 rdt
The rate at which work is done over a volume V is therefore

dW
(5) = E Jd 3 r
dt V

We can write this entirely in terms of E and B by using one of Maxwells


equations

E
(6) B = 0 J + 0 0
t
Thus
1
POYNTINGS THEOREM 2

1 E
(7) EJ = E ( B) 0 E
0 t
An identity from vector calculus says

(8) (E B) = B ( E) E ( B)

so

1 1
(9) E ( B) = [B ( E) (E B)]
0 0
Using another of Maxwells equations (Faradays law)

B
(10) E =
t

we get
 
1 1 B
(11) E ( B) = B + (E B)
0 0 t
For any vector field A we have

2
(12) A = (A A) = 2A A
t t t

so

1 1 B2
 
1
(13) E ( B) = + (E B)
0 0 2 t
Combining this with 12 for A = E and 7 we get

1 B2 0 E 2 1
(14) EJ = (E B)
20 t 2 t 0
The total rate of work done in volume V is therefore
 
dW 1 1 2 2 3 1
(15) = B + 0 E d r (E B) d 3 r
dt 2 t V 0 0 V
Using the divergence theorem on the last integral, we get
POYNTINGS THEOREM 3

 
dW 1 1 2 2 3 1
(16) = B + 0 E d r (E B) da
dt 2 t V 0 0 S

where the integral is now over the surface S enclosing the volume V .
Weve seen earlier that the energy stored in electric and magnetic fields
is

0
(17) WE = E 2d3r
2 V

1
(18) WB = B2 d 3 r
20 V

so the first integral in 16 is the negative of the rate of change of the en-
ergy stored in the two fields. That is, as the fields do work on the charges
they expend their energy (so it decreases, giving a negative derivative) but
the work done on the charges is positive. The second integral is the rate at
which energy flows across the surface S . This result is known as Poynt-
ings theorem, named after John Henry Poynting (1852-1914), an English
physicist.
The vector field

1
(19) S EB
0

is called the Poynting vector, and represents the rate per unit area at which
energy crosses a surface.
Example 1. We have a coaxial cable with inner radius a and outer radius
b that carries a current I down the inner cylinder and I back on the outer
cylinder. There is a potential difference V between the inner and outer cylin-
ders.
This is a magnetostatic situation, so we can apply Ampres law to find
B. The magnetic field is circumferential so at a radius r between the two
cylinders

(20) 2rB = 0 I
0 I
(21) B =
2r
We can find the electric field from Gausss law. If the linear charge den-
sity on the inner cylinder is then by symmetry the electric field points
POYNTINGS THEOREM 4

radially outward from the axis of the cable. If we take a cylindrical Gauss-
ian surface of radius r and length 1, then


(22) 2rE =
0

(23) E =
20 r
In order for the potential difference between the cylinders to be V , we
must have
b
(24) V = Edr
a
b
dr
(25) =
20 a r
b
(26) = ln
20 a
20V
(27) =
ln (b/a)
V
(28) E = r
r ln (b/a)
The Poynting vector is therefore

1
(29) S = EB
0
IV
(30) = 2
z
2r ln (b/a)
The total rate at which power flows along the cable is then
b
2IV rdr
(31) S da =
2 ln (b/a) a r2
(32) = IV
This is a special case of the general formula known as the Joule heating
law, which is that the rate at which energy flows (the power) is P = IV .
Example 2. Well revisit the case of the transmission line consisting of two
thin ribbons (each of width w) separated by an insulating vacuum space of
thickness h  w, with each ribbon carrying current I (in opposite directions)
and held at a potential difference V . Because the separation of the two
ribbons is much less than their width, we can approximate the situation by
POYNTINGS THEOREM 5

considering each ribbon to be a sheet of current. This gives a magnetic field


from each ribbon perpendicular to the cable. Taking the cable to lie along
the x axis, and the xy plane to be the plane containing the ribbons,

0 I
(33) Br =y
2w
The fields from the two ribbons will add in the space between them so
the total field is

0 I
(34) B = 2Br = y
w
The electric field is

V
(35) E= z
h

so

1
(36) S = EB
0
IV
(37) = x
hw

and the total rate of energy flow is this times the area between the ribbons:

IV
(38) S da = hw = IV
hw
The minus sign is there because of the direction we chose for the x axis;
clearly because the cable is symmetric, it just means that the power flows
along the cable at a rate of IV , as wed expect.

P INGBACKS
Pingback: Poyntings theorem and conservation of energy
Pingback: Maxwell stress tensor
Pingback: Momentum in electromagnetic fields
Pingback: Energy transfer in a solenoid
Pingback: The electron as electromagnetic energy and angular momen-
tum
Pingback: Electromagnetic waves: energy, momentum and light pressure
Pingback: Average of product of two waves
POYNTINGS THEOREM 6

Pingback: Electromagnetic waves in conductors: energy density and in-


tensity
Pingback: Wave guide: energy flows at the group velocity
Pingback: Transmission coefficient for a wave passing through 3 media
Pingback: Energy flow in time-dependent fields
Pingback: Power due to point charge moving with constant velocity
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
Pingback: Gausss law for a relativistic point charge
POYNTINGS THEOREM AND CONSERVATION OF ENERGY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.2.
Weve seen that Poyntings theorem allows us to calculate the rate at
which energy changes within a given volume due to the action of electro-
magnetic fields. The theorem says
 
dW 1 1 2 2 3 1
(1) = B + 0 E d r (E B) da
dt 2 t V 0 0 S
The first term gives the rate at which energy stored within the fields
changes within the volume V and the second term gives the rate at which
energy flows across the surface S enclosing V .
As an example, we can revisit the following problem. We have a solid
wire with a circular cross-section of radius a that has a narrow gap of width
w  a cut into it. If a steady current I is passed along the wire, the two
surfaces on either side of the gap act as a parallel plate capacitor, as charge
builds up on these surfaces. As a result, the electric field between the plates
changes as the charge builds up, so we should get an induced magnetic field.
We saw before that the electric field within the gap is

E I
(2) =
t a2 0
I
(3) E (r,t) = t z
a2 0

assuming that E (0) = 0, and the z axis is parallel to the wire, pointing from
the positive side of the gap to the negative. The electric field is independent
of the distance r from the axis.
The magnetic field is given by

Ir
(4) B (r,t) = 0
2a2
E
Because t is independent of time, so is B, although it does depend on
r.
The energy density uem in the fields is then
1
POYNTINGS THEOREM AND CONSERVATION OF ENERGY 2

 
1 1 2
(5) uem = B + 0 E 2
2 0
0 I 2 r2 I 2t 2
 
1
(6) = +
2 4 2 a4 2 a4 0
I2 0 r2 t 2
 
(7) = +
2 2 a4 4 0
The Poynting vector is

1
(8) S = EB
0
I 2 rt
(9) = 2 4 r
2 a 0
Note that S points inwards, indicating that energy is flowing into the gap
in the wire, causing the increase in the energy density.
In general, if there is any charge in the volume V , then the electric field
will do work on the charge if the charge moves around (magnetic fields
do no work). This work will change the mechanical energy umech of the
charges (that is, their potential or kinetic energy). The total energy flowing
into the volume must therefore add up to the sum of umech and uem ; that is,
the energy flowing into a volume is stored either in the energy of the fields
or in the mechanical energy of the charge. Since the Poynting vector S gives
the rate at which energy flows across a surface, the integral of S over the
surface bounding V must equal the rate of change of umech + uem , that is


d
(10) S da = (umech + uem ) d 3 r
S dt V

where the minus sign on the LHS is there because we want the energy flow-
ing into the volume and da is an outward pointing normal to the surface.
Using the divergence theorem to convert the LHS to a volume integral, we
get the differential form:


(11) S = (umech + uem )
t
In our present example, there is no charge in the gap, so umech = 0 so we
should have
POYNTINGS THEOREM AND CONSERVATION OF ENERGY 3

uem
(12) S =
t
We can check this:

1
(13) S = (rSr )
r r
1 2I 2 rt
(14) =
r 2 2 a4 0
I 2t
(15) =
2 a4 0
uem
(16) =
t

so it checks out.
The total energy in the gap as a function of time can be found by inte-
grating 7 over the volume of the gap:

I2 0 r2 t 2

(17) Uem (t) = + d3r
2 2 a4 V 4 0
2 a
2w0 I 3 a2 wI 2t 2
(18) = r dr +
8 2 a4 0 2 2 a4 0
0 wI 2 wI 2t 2
(19) = +
16 2a2 0
Uem wI 2t
(20) =
t a2 0

I 2 at
(21) S da = (2aw)
S 2 2 a4 0
wI 2t
(22) =
a2 0
Uem
(23) =
t
Thus again it checks out.
P INGBACKS
Pingback: Electromagnetic stress-energy tensor
MAXWELL STRESS TENSOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.3.
Here well look at a purely classical, non-relativistic form of the tensor
in electromagnetism. In doing so, well look only at the spatial components
of the tensor, so it becomes a 3 3 matrix.
The derivation starts with a calculation of the total force due to electro-
magnetic fields on the charges and currents within some volume V . From
the Lorentz force law, we have

(1) F = (E + v B) d 3 r
V
(2) = (E + J B) d 3 r
V
We can think of the integrand as a force density, or force per unit volume
f:

(3) f E + J B
We can express this entirely in terms of fields by using Maxwells equa-
tions:

(4) = 0 E
1 E
(5) J = B 0
0 t
So we get
 
1 E
(6) f = (0 E) E + B 0 B
0 t
We now need to do a bit of vector calculus gymnastics. From the product
rule

E B
(7) (E B) = B+E
t t t
1
MAXWELL STRESS TENSOR 2

and from Faradays law

B
(8) = E
t

Combining these two we get

E B
(9) B = (E B) E
t t t

(10) = (E B) + E ( E)
t
We can insert this into 6 and while were at it, we can add on a term
1
0 ( B) B. This is always zero because B = 0, but it gives the equation
a symmetry that will be useful in a minute. We get for the force density:

1 1 E
(11) f = 0 ( E) E + ( B) B + ( B) B 0 B
0 0 t
1 1
(12) = 0 ( E) E + ( B) B + ( B) B
0 0

(13) 0 (E B) 0 E ( E)
t
Now another identity from vector calculus says

(14) (A B) = A ( B) + B ( A) + (A ) B + (B ) A
If A = B = E, we get

E 2 = 2E ( E) + 2 (E ) E

(15)

so

1
E 2 (E ) E

(16) E ( E) =
2
1
B2 (B ) B

(17) B ( B) =
2
Putting this into 12 we get
MAXWELL STRESS TENSOR 3

1
(18) f = 0 ( E) E + ( B) B 0 (E B)
0 t
1
B ( B) 0 E ( E)
0
1
(19) = 0 ( E) E + ( B) B 0 (E B)
0 t
 
1 1 1
0 E 2 + B2 + 0 (E ) E + (B ) B
2 0 0
1
(20) = 0 [( E) E + (E ) E] + [( B) B + (B ) B]
0
 
1 1
0 E 2 + B2 0 (E B)
2 0 t

It might not seem that were making any progress, since the equations
just get longer with each alteration. However, we can now introduce the

Maxwell stress tensor T which is a 3 3 matrix with components defined
by

   
1 2 1 1 2
(21) Ti j 0 Ei E j i j E + Bi B j i j B
2 0 2

Note that the tensor is symmetric: Ti j = T ji . If we define the scalar prod-


uct of the tensor with an ordinary vector to be another vector:

h i
(22) a T = ai Ti j
j i

where the subscript j indicates the jth component of the resulting vector,
then the divergence is
MAXWELL STRESS TENSOR 4

h i
(23) T = iTi j
j i
 
 1 2
(24) = 0 (i Ei ) E j + Ei i E j i j i E +
i 2
 
1  1 2
(i Bi ) B j + Bi i B j i j i B
0 i 2
 
1 2
(25) = 0 ( E) E j + (E ) E j j E +
2
 
1 1 2
( B) B j + (B ) B j j B
0 2


Comparing this with 20, we see that we can write f in terms of T and
the Poynting vector as

S
(26) f = T 0 0
t
The total force on the volume is then

(27) F = fd 3 r
V  

S 3
(28) = T 0 0 d r
V t
From the formula 23 for the divergence, we can see that the vector re-
sulting from the divergence has as its components the divergences of each

column of T . Therefore we can apply the divergence theorem to the first
term in the integrand to get




(29) F= T da 0 0 Sd 3 r
S t V

where S is any surface that encloses only the charges and currents within
V.
Example. We can revisit the problem of finding the magnetic force between
the two halves of a spherical shell of surface charge density rotating with
angular velocity = z. In our earlier solution we used the Biot-Savart
law and integrated over each differential ring in the rotating sphere. Using
the stress tensor, we can integrate over any volume that encloses the upper
MAXWELL STRESS TENSOR 5

half of the sphere, so we can choose the half space consisting of all space
above the xy plane (were assuming that the centre of the sphere is at the
origin, so the xy plane contains the spheres equator). Since the distribution
of charges and currents is finite, all fields will go to zero at infinity, so we
need to integrate only over the xy plane.
We saw earlier that the magnetic field inside and outside the sphere is

( 2  20 R
0 R
3 cos r sin = 3 z r < R
(30) B= 0 R4 1 
3 r 3 2 cos r + sin r>R

In the xy plane, = /2 so the field is

( 2
0 R
3 z r<R
(31) B= 4
0 R3 r13 z r>R

Since were interested only in the magnetic field, we can ignore E here,
although there is a repulsive force between the two hemispheres due to the
electric field as well. Also, as the currents are steady, S/t = 0.
From the symmetryhof the problem, the force is in the z direction, so we
i
need to work out only T da . We get Txz = Tyz = 0 because Bx = By = 0
z
on the xy plane, so were left with just Tzz :

(
2 2 2 2
1 2 9 0 R r<R
(32) Tzz = B = 2 2 R8
20 z 1
18 0 r6 r>R

The total force is then (the minus sign is because Tzz > 0 and da points
towards z):



(33) F= T da
S
 R 
2 2 2 2 2 2 2 8 r dr
(34) = z 0 R 2 r dr + 0 R
9 0 18 R r6
 
2
(35) = z 0 2 2 R4 + 0 2 2 R4
9 36

(36) = 0 2 2 R4 z
4
This agrees with the result we got earlier using the Biot-Savart law.
MAXWELL STRESS TENSOR 6

P INGBACKS
Pingback: Maxwell stress tensor: force between two charges
Pingback: Momentum in electromagnetic fields
Pingback: Maxwell stress tensor and electromagnetic waves
MAXWELL STRESS TENSOR: FORCE BETWEEN TWO
CHARGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.4.
Heres another, albeit impractical, example of using the Maxwell stress
tensor to calculate the force on a charge. We have two charges q of the same
sign, with one at z = +a and the other at z = a. We can integrate the stress
tensor over some volume containing, say, the lower charge to find the force
on it. (Obviously, we could just write down the force as F = q 2 z
40 (2a)
but we want to see if we can get the same answer from the stress tensor.)
Well use a spherical coordinate system centred on the lower charge, and
integrate the tensor over the xy plane (recall that we can do the integral over
any volume enclosing the charge, and since the fields go to zero at infinity,
we can use the lower half space as our volume, with the xy plane as its
bounding surface). Then a point on the xy plane has rectangular coordinates
[r sin cos , r sin sin , r cos ]. Since the electric field radiates out from
each of the charges with a magnitude

q
(1) E=
40 r2

where r is the distance from the charge, the rectangular components of E1


from the lower charge are

q
(2) E1x = sin cos
40 r2
q
(3) E1y = sin sin
40 r2
q
(4) E1z = cos
40 r2

By symmetry, the x and y components of the upper charges field are the
same as those from the lower charge, and the z component is equal and
opposite, so the total field in the xy plane is
1
MAXWELL STRESS TENSOR: FORCE BETWEEN TWO CHARGES 2

2q
(5) Ex = sin cos
40 r2
2q
(6) Ey = sin sin
40 r2
(7) Ez = 0

Also from symmetry, the net force is in the z direction, as is the normal
to the surface over which were integrating, so we need only the component
Tzz .

0 2
Ez Ex2 Ey2

(8) Tzz =
2
 2
0 q
(9) = sin
2 20 r2

Since well be integrating over the xy plane, we can use the 2-d polar
coordinates s (for distance from the origin in the xy plane) and the azimuthal
angle . We then have

p
(10) r =a2 + s2
s s
(11) sin = =
r a2 + s2
a a
(12) cos = =
r a + s2
2

q2 s2
(13) Tzz = 2
8 0 (a2 + s2 )3

The force is then given by




(14) F = T da
S

q2 z
s2 (s ds)
(15) = 2 (2) 3
8 0 0 (a2 + s2 )

The integral can be done with software or by using the substitution u =


a2 + s2 ,
du = 2s ds
MAXWELL STRESS TENSOR: FORCE BETWEEN TWO CHARGES 3



s3 ds
u a2
(16) 3
= du
0 (a2 + s2 ) a2 2u3
1 a2
(17) =
2a2 4a4
1
(18) =
4a2
Thus the force is

q2
(19) F= z
40 (2a)2
which is a roundabout way of getting Coulombs law.
If the charges are opposite, then the x and y components of the field cancel
and the z component adds, so we get for the force on the lower charge

(20) Ex = 0
(21) Ey = 0
2q
(22) Ez = cos
40 r2
qa
(23) = 3/2
20 (a2 + s2 )
q2 a2 1
(24) Tzz =
8 0 (a2 + s2 )3
2

q2 a2 z s ds
(25) F = 2
(2) 3
8 0 0 (a2 + s2 )
q2
(26) = + z
40 (2a)2
This also agrees with Coulombs law; this time the force is upwards so
the charges are attracted to each other.
MOMENTUM IN ELECTROMAGNETIC FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.5.
Weve seen that the force on a collection of charges in a volume V en-
closed within a surface S due to electromagnetic fields is given in terms of
the Maxwell stress tensor:



(1) F= T da 0 0 Sd 3 r
S t V


where T is the stress tensor and S is the Poynting vector. It turns out
that electromagnetic fields carry both momentum and angular momentum.
To see this, we can write the above equation in terms of the mechanical
momentum pmech (that is, the momentum of the particles present, rather
than the fields):

dpmech

(2) = T da 0 0 Sd 3 r
dt S t V
Looking back at Poyntings theorem:
 
dW 1 1 2 2 3
(3) = B + 0 E d r S da
dt 2 t V 0 S

we saw that there, we had a volume integral representing the rate at which
energy in the fields was changing and a surface integral representing the
rate at which energy was flowing into (or out of) the volume. If we make
an analogy with 2, we can interpret the volume integral as the rate at which
momentum in the fields is changing and the surface integral as the rate at
which momentum flows into the volume. That is

(4) pem 0 0 Sd 3 r
V

is the momentum of the fields themselves and


1
MOMENTUM IN ELECTROMAGNETIC FIELDS 2

(5) pem = 0 0 S

is the momentum density of the fields.


The other integral can be converted into a volume integral using the di-
vergence theorem:



(6) T da = T d3r
S V

If we define the mechanical momentum density pmech then we can write 2


in differential form


(7) (pem + pmech ) = T
t
We can look at this equation as a set of three equations, one for each
component of the momentum. That is,

3

(8) (pem + pmech )i = j Ti j
t j=1
Compare this with the continuity equation relating charge density and
current density J:


(9) = J
t
This says that if there is a divergence in J, that is, current is flowing out
of some infinitesimal volume, then the charge density decreases in that vol-
ume. Equation 7 says something similar by relating the total momentum

density pem + pmech to the divergence of the stress tensor T . In partic-
ular, the rate at which a particular component i of momentum changes is

the divergence of the ith column of T (the minus sign occurs because a


positive divergence of T leads to an increase in momentum density, so it

must be the negative of T that gives the rate at which momentum flows out
of a volume), so Ti j is the rate at which momentum component i flows in
the j direction. This is the same interpretation as the spatial coordinates of
the stress-energy tensor in relativity.
In electrostatics, S = 10 E B = 0 because B = 0, so electric fields on
their own store no momentum. However, the stress tensor is not zero in
electrostatics, so momentum still does flow in an electric field.
MOMENTUM IN ELECTROMAGNETIC FIELDS 3

Example. For example, consider an infinite flat plate capacitor with surface
charge density + on one plate and on the other. From Gausss law,
the electric field from each plate is


(10) E=
20

pointing away from the plate on both sides. Outside the capacitor, the fields
cancel, but between the plates they add, giving a net field of


(11) E=
0
Taking the z axis to be perpendicular to the plates, we have for the stress
tensor between the plates

Ez2

0 0

0
(12) T = 0 Ez2 0
2
0 0 Ez2

2 1 0 0

(13) = 0 1 0
20 0 0 1
We can get the force on a plate from 1. Since da is in the z direction,
only Tzz contributes (and S = 0 so the second integral is zero), so if we
integrate over a unit area of the plate



(14) Fupper = T da
S
(15) = Tzz z
2
(16) = z
20
The minus sign occurs because da points in the z direction and Tzz is
positive. On the lower plate, da points in the +z direction and Tzz is the
2
same, so Flower = + 2 0
z. The force on each plate is towards the other plate,
as wed expect since they are oppositely charged.
Interpreting the stress tensor as momentum flux density, we have that
Tzz is the momentum flux in the z direction; that is, its the amount of
momentum per unit area per unit time that crosses any plane parallel to
and between the plates. When this momentum is absorbed by a plate, the
plate experiences a force. The momentum flux density flowing in the +z
MOMENTUM IN ELECTROMAGNETIC FIELDS 4

2
direction is therefore Tzz = 2 0
and this is the momentum delivered to
the plate per unit area per unit time, which is the force per unit area, so we
get the same answer as in 16.
To get the force on the lower plate, we can use the fact that a momentum
flux density of Tzz in the +z direction is the same as a density of +Tzz in
the z direction, so the momentum delivered to the lower plate per unit area
2 2
per unit time is +Tzz = + 2 0
giving F lower = + 20
z as before.
Incidentally, it might seem that equations 1 and 7 are inconsistent in this

case where T is constant, since it would appear that T = 0 which

would imply, by the divergence theorem, that S T da = 0 as well. The

catch is that T isnt really constant over the surface surrounding a unit area
on one of the plates, since it has non-zero components between the plates,


but is zero everywhere outside the plates. Thus T must drop to zero as we
cross the plate. In an idealized problem where the plate is infinitesimally


thin, we could use a step function to represent T , which means that T
would involve delta functions.
P INGBACKS
Pingback: Angular momentum in electromagnetic fields
Pingback: Angular momentum in a magnetized sphere
Pingback: Momentum in a magnetized and polarized sphere
Pingback: The electron as electromagnetic energy and angular momen-
tum
Pingback: Angular momentum of a charge and magnetic monopole
Pingback: Momentum of a point charge outside a solenoid
Pingback: Electromagnetic waves: energy, momentum and light pressure
Pingback: Maxwell stress tensor and electromagnetic waves
Pingback: Force and momentum with two moving point charges
MOMENTUM IN A CAPACITOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.6.
The momentum density of an electromagnetic field is given by

(1) pem = 0 0 S = 0 E B
Rather bizarrely, this applies even in cases where the electric and mag-
netic fields do not change with time, so a system such as a coaxial cable
carrying a steady current, with a potential difference between its inner and
outer cylinders, contains momentum in the fields. In fact, there is a balanc-
ing momentum due to a relativistic effect involving magnetic dipoles which
balances the momentum in the fields so that a static configuration actually
does have zero momentum, but that takes us further afield than we want to
go at the moment.
As an example of conservation of the field momentum, we can look at an
idealized parallel plate capacitor, with plate area A and plate separation d.
Its capacitance is given by

A
(2) C = 0
d
If we align the plates parallel to the xy plane and charge the capacitor,
then the electric field between the plates is E = E0 z assuming that the lower
plate is positive and the upper one negative. If the potential difference be-
tween the plates is V0 then E0 = V0 /d.
We now apply a constant magnetic field B = B0 x along the +x direction,
so that the momentum density between the plates (neglecting edge effects)
is

(3) pem = 0 E0 B0 y

and the total momentem in the fields between the plates is just the density
multiplied by the volume:

(4) pem = 0 E0 B0 Ad y
1
MOMENTUM IN A CAPACITOR 2

Now suppose we connect a resistor R between the plates along the z axis.
The capacitor will discharge, with the potential difference given by

(5) V (t) = V0 et/RC


The current in the resistor is in the +z direction of magnitude

V0 t/RC
(6) I (t) =
e
R
Because we now have moving charges (the current) in a magnetic field,
the resistor feels a force

(7) F = qv B = dI (t) By
The total impulse I felt by the resistor as the capacitor discharges (the im-
pulse is the force integrated over time, and gives the change in momentum
a force imparts to an object) is

(8) I = Fdt
0

dV0 B
(9) = y et/RC dt
R 0
(10) = CdV0 By
A
(11) = 0 dE0 dBy
d
(12) = 0 E0 B0 Ad y
(13) = pem
Thus the momentum in the fields gets transferred to the resistor after the
capacitor discharges.
We can also eliminate the momentum in the field by reducing the mag-
netic field to zero and leaving the capacitor charged. By Faradays law, the
changing magnetic field will induce an electric field (in addition to the field
due to the charge on the plates) by the formula

B
(14) E =
t

or in integral form

B
(15) E d` = da
t
MOMENTUM IN A CAPACITOR 3

We can choose the path of integration on the LHS to be a rectangle par-


allel to the yz plane that is a cross-section of the area between the plates. If
the width of the plates is ` then this rectangle has width ` and height d. If
we reduce B uniformly, then at any particular time tB is a constant over the
rectangle and the RHS can be evaluated so we get

B
(16) E d` = `d
t
As for E, remember that were considering only the field that is induced
due to changing B. E cannot have a z component since, if it did, we could
invert this z component by turning the system upside-down. However, since
B is uniform and points in the x direction, turning it upside-down would
leave it unchanged, so the induced E cannot change either, so Ez = 0. We
must also have Ex = 0 since any path in a plane parallel to the x axis contains
no magnetic flux (since its parallel to B; the same argument also shows that
Ez = 0). Thus the induced field must be entirely in the y direction.
If we neglect edge effects in the capacitor, then by symmetryE is constant
over any plane parallel to the plates, so the integral on the LHS becomes
(taking the path of integration as counterclockwise when viewed down the
x axis)

(17) E d` = `Ey (z = d) + `Ey (z = 0)
B
(18) E (0) E (d) = d
t
This field exerts a force in the y direction on the plates. If the surface
charge density on the lower plate is + and on the upper plate is then
the net force on the capacitor is

(19) F = A [E (0) E (d)] y


B
(20) = A d y
t
The impulse is then

B
(21) I = Ad y dt
0 t
(22) = AdB0 y

B
(The minus sign is cancelled out because B is decreasing so t < 0.) The
MOMENTUM IN A CAPACITOR 4

surface charge density on a capacitor plate is related to the vertical electric


field by


(23) E0 =
0

so the impulse is

(24) I = 0 E0 AdB0 y

which is the same as 12. Thus the fields momentum is transferred to the
capacitor as we turn off the magnetic field.
ANGULAR MOMENTUM IN ELECTROMAGNETIC FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.7.
The momentum density of an electromagnetic field is given by

(1) pem = 0 0 S = 0 E B
If we have linear momentum, then we automatically have angular mo-
mentum with respect to some origin by using the classical definition of an-
gular momentum L = r p. We can define the angular momentum density
of an electromagnetic field by

(2) Lem r pem = 0 r (E B)


Just as with linear momentum, even static fields can have angular mo-
mentum. As an example, suppose we have a long solenoid with n turns per
unit length carrying current I0 and a radius R, with its axis along the z axis.
The magnetic field inside the solenoid is

(3) B0 = 0 nI z

The field is zero outside the solenoid.


Now suppose we add two other cylinders (not solenoids), both coaxial
with the solenoid. One cylinder has radius a < R (so it lies inside the
solenoid) and carries surface charge +Q; the other cylinder has radius b > R
(outside the solenoid) and carries charge Q. Both cylinders have length
`. From Gausss law, the electric field between these two cylinders is, for
a<r<b

Q r
(4) E0 =
20 ` r

That is, the field points radially outward from the axis. The electric field is
zero for r < a and r > b. (Were neglecting end effects, so were assuming
that `  b > a.)
1
ANGULAR MOMENTUM IN ELECTROMAGNETIC FIELDS 2

The linear momentum density is non-zero in the region a < r < R (where
both fields are non-zero) and we have

0 nIQ
(5) pem =
2` r

so the angular momentum density is

(6) Lem = r pem


0 nIQ
(7) = z
2`
Conveniently, this is constant so the total angular momentum is just the
density times the volume of the cylindrical tube in the region a < r < R

 0 nIQ
(8) Lem = ` R2 a2 z
2`
 0 nIQ
(9) = R2 a2 z
2
Now suppose we (quasistatically) discharge the two cylinders by con-
necting a resistor R between them. Wed like to show that the angular
momentum gets transferred from the fields to the physical devices in the
problem. The two cylinders are effectively a capacitor with some capaci-
tance C, so we know that the current in the resistor will decay exponentially

V0 t/RC
(10) I (t) = e
R

where V0 is the potential difference between the cylinders at t = 0. The


force dF on a segment of the resistor of length dr is

(11) dF = I (t) drr B0


(12) = I (t) drB0

so the torque on this segment is

(13) dN = r dF
(14) = I (t) B0 rdrz
The total torque on the resistor at time t is
ANGULAR MOMENTUM IN ELECTROMAGNETIC FIELDS 3

R
(15) N (t) = I (t) B0 z r dr
a
1
= I (t) B0 R2 a2 z

(16)
2
The angular impulse is the integral of torque over time, so we get

(17) I = N (t) dt
0

1 2 2

(18) = B0 R a z I (t) dt
2
0
1 V0 t/RC
B0 R2 a2 z

(19) = e dt
2 0 R
1
B0 R2 a2 CV0 z

(20) =
2
1
0 nI R2 a2 Qz

(21) =
2

where we used the relation between capacitance, charge and voltage Q =


CV . We see that this agrees with 9, so all the angular momentum in the
fields is transferred to the resistor as the electric field is reduced to zero.
P INGBACKS
Pingback: Angular momentum in a magnetized sphere
Pingback: The electron as electromagnetic energy and angular momen-
tum
Pingback: Angular momentum of a charge and magnetic monopole
Pingback: Angular momentum conservation: example with a solenoid
Pingback: Momentum of a point charge outside a solenoid
ANGULAR MOMENTUM IN A MAGNETIZED SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.8.
Heres another slightly more complex example of the conservation of
angular momentum in electromagnetic fields. We have a solid iron sphere
of radius R with a uniform magnetization Mz and carrying a surface charge
Q. We want to show that the angular momentem in the electric and magnetic
fields is conserved as we switch off either the magnetic or electric field.
First, we need to calculate the angular momentum in the fields before
anything is switched off.
The momentum density of an electromagnetic field is given by

(1) pem = 0 0 S = 0 E B

and the angular momentum density is

(2) Lem r pem = 0 r (E B)


Griffiths shows in his Example 6.1 that the magnetic field of a uniformly
magnetized sphere is
(
2 2

3 0 Mz = 3 0 M cos r sin inside
(3) B= 0 MR3 
3r3
2 cos r + sin outside
The outside formula is just the field of a perfect magnetic dipole.
The electric field is
(
0 inside
(4) E= Q
40 r2
r outside
The linear momentum density is then

0 MQR3
(5) pem = 0 E B = sin
12r5
1
ANGULAR MOMENTUM IN A MAGNETIZED SPHERE 2

so the angular momentum density is

0 MQR3
(6) Lem = r pem = sin
12r4
The total angular momentum is thus
2
0 MQR3 sin 2
(7) Lem = r sin d d dr
12 R 0 0 r4
As the unit vector is not constant in direction, we need to express it in
rectangular coordinates:

(8) = cos cos x + cos sin y sin z


The integral of cos and sin over the interval [0, 2] gives zero, so only
the z component survives, giving

0 MQR3 dr
(9) Lem = z (2) 2
sin3 d
12 R r 0
2
(10) = 0 MQR2 z
9
Now suppose we slowly switch off the magnetic field, keeping Q con-
stant. This will generate an electric field according to Faradays law. Con-
sider a slice of the sphere parallel to the xy plane, that is, at some value of
. The electric field is circular around the axis of the sphere and satisfies
the equation

d B
(11) E d` = = da
dt t
The radius of the circle is R sin and if we switch off the magnetic field
in a uniform way, then tB is constant over this area, so we get

B
(12) 2RE sin = (R sin )2
t
1 B
(13) E = R sin
2 t

The minus sign in the last line cancels the fact that tB < 0 and gives an
electric field that opposes the decrease in magnetic field.
ANGULAR MOMENTUM IN A MAGNETIZED SPHERE 3

This field exerts a force on the strip of charge in the slice. The surface
charge density is

Q
(14) =
4R2

so the amount of charge in the slice is

Q 1
(15) dq = 2
(2R sin ) (Rd ) = Q sin d
4R 2
The force on the slice is

1 B
(16) dF = Edq = RQ sin2 d
4 t

and the torque is

(17)
1 B 1 B
dN = r dF = (R sin ) RQ sin2 d z = R2 Q sin3 d z
4 t 4 t
The total torque on the sphere is

(18) N = dN ( )
0

1 2 B
(19) = R Q z sin3 d
4 t 0
1 B
(20) = QR2 z
3 t
The impulse is obtained by integrating this over time:
 
1 B 1 2 2 2
(21) I = QR2 z dt = QR z 0 M = 0 MQR2 z
3 0 t 3 3 9

which agrees with 10, showing that the angular momentum is transferred to
the sphere.
Now lets keep the magnetization constant and decrease the electric field
by draining off the spheres surface charge. We arrange to do this by con-
necting a resistor R between the north pole and ground, but we do this in a
way that the surface charge density remains uniform over the sphere as the
total charge decreases. Doing this creates a surface current that experiences
ANGULAR MOMENTUM IN A MAGNETIZED SPHERE 4

a force from the magnetic field. Since were draining the charge off at the
north pole, the surface current everywhere on the sphere will head towards
= 0, that is, it flows in the direction. To work out the current at each
point on the sphere, consider the same horizontal slice that we used above.
The current passing through this slice must be the sum of the current pass-
ing through the slice just below it (that is, at +d ) and the charge draining
away from the slice itself. That is

d
(22) I ( ) = I ( + d ) + (2R sin ) (Rd )
dt
dI d
(23) = 2R2 sin
d dt
d
(24) I ( ) = 2R2 cos +
dt

where is a constant of integration. Since the charge is draining off through


the north pole, wed like the current at the south pole to be zero, so we
choose

d
(25) =2R2
dt
We can treat the sphere/ground system as a capacitor C, so the charge
density is given by

(26) (t) = (0) et/RC


Q t/RC
(27) = e
4R2

so

Q t/RC
(28) I ( ) = e (1 + cos )
2RC
This is the total current flowing across the surface of a slice at angle .
The qv term in the force law is the current times the distance over which
it flows, which in this case is Rd . We can now apply the Lorentz force
law F = qv B to work out the force on the sphere due to the current in-
teracting with the magnetic field. Since the current is parallel to , only
the r component of B will survive the cross product. Note that although the
component of B is not continuous across the surface of the sphere, the r
component is continuous, so it doesnt matter whether we use the inside or
outside formula. Thus we get
ANGULAR MOMENTUM IN A MAGNETIZED SPHERE 5

   
Q t/RC 20 M 
(29) dF = e (1 + cos ) (Rd ) cos
2RC 3
0 MQR t/RC
cos + cos2

(30) = e
3RC
The torque on a slice is

(31) dN = r dF
0 MQR2 t/RC
z cos + cos2 sin

(32) = e
3RC
The total torque is thus

0 MQR2 t/RC
cos + cos2 sin d

(33) N = e z
3RC 0
20 MQR2 t/RC
(34) = e z
9RC
Finally we integrate the torque over time to get the impulse:

(35) I = Ndt
0

20 MQR2
(36) = z et/RC dt
9RC 0
2
(37) = 0 MQR2 z
9
This again agrees with 10 so all is well.
ENERGY TRANSFER IN A SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.9.
Heres a simple example of conservation of energy in an electromagnetic
system. We have an infinite solenoid of radius a carrying n turns per unit
length and current Is . The magnetic field is zero outside the solenoid and
inside we have

(1) Bs = 0 nIs z
Now suppose we put a circular wire loop of radius b  a and resistance
R around the solenoid. If we now decrease the current in the solenoid, the
changing magnetic flux will induce a circumferential electric field around
the solenoid, which will in turn create a current in the wire. To find the
current Ir in the resistor we can use the fact that the electric field creates an
electromotive force (emf) around the wire:

d
(2) E =
dt
dIs
(3) = a2 0 n
dt
The current is

E a2 0 n dIs
(4) Ir = =
R R dt
To find the direction of the current, remember that it in turn generates
a magnetic field that opposes the reduction in the solenoids field, so the
current must flow in the + direction (counterclockwise as viewed from
above).
The rate at which energy is dissipated by the resistor is the power, which
is Ir2 R. This energy must come from the solenoid via the Poynting vector.
We can calculate the Poynting vector just outside the solenoid as follows.
First, we need E and B outside the solenoid. The electric field is produced
by the changing magnetic field in the solenoid, and by Faradays law we
have
1
ENERGY TRANSFER IN A SOLENOID 2


B
(5) E d` = da
t
Taking a circular path of radius a we get

dIs 2
(6) 2aE = 0 n a
dt
0 na dIs
(7) E =
2 dt
Ir R
(8) =
2a
Remember that dI s
dt < 0 so E points in the + direction.
There is no magnetic field due to the solenoid outside the solenoid itself,
but the current in the resistor generates a magnetic field due to the Biot-
Savart law. Griffiths works out the magnetic field on the z axis due to a
circular loop in his Example 5.6 and since were taking the radius b of the
loop to be much greater than the radius a of the solenoid, we can use this
formula as a good approximation. We have

0 Ir b2
(9) Br = z
2 (b2 + z2 )3/2
The Poynting vector is then

1
(10) S = EB
0
I2R b2
(11) = r r
4a (b2 + z2 )3/2
We can integrate the magnitude of the vector over the surface of the
solenoid to find the rate at which energy is radiating away from the solenoid.
We get

(12) P = S da

I2R
b2
(13) = (2a) r 3/2
dz
4a (b2 + z2 )
(14) = Ir2 R
Thus the power in the resistor is indeed coming from the solenoid.
MOMENTUM IN A MAGNETIZED AND POLARIZED SPHERE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.10.
Heres another example of finding momentum in electromagnetic fields.
Consider a non-conducting sphere of radius R with a uniform magnetization
M and a uniform polarization P, which need not be in the same direction as
M. What is the total momentum stored in the fields?
The momentum density is

(1) pem = 0 E B
Its easiest to work with the fields in their coordinate free forms. The
field of a pure magnetic dipole at the origin is

0
(2) B1 = [3 (m r) r m]
4r3

and for a pure electric dipole we have

1
(3) E= [3(p r)r p]
40 r3
Griffiths shows in his Example 6.1 that the magnetic field of a uniformly
magnetized sphere is

(
2
(4) B= 3 0 M inside
0
4r3
[3 (m r) r m] outside

where the total magnetic dipole moment of the sphere is

4
(5) m = R3 M
3

so
1
MOMENTUM IN A MAGNETIZED AND POLARIZED SPHERE 2

(
2
3 0 M inside
(6) B= 0 R3  1

r 3 (M r) r 3 M outside
Griffiths also shows in his Example 4.2 that a uniformly polarized sphere
has electric field
(
310 P inside
(7) E= 1
4 r3
[3 (p r) r p] outside
0

where the total electric dipole moment is

4
(8) p = R3 P
3

so
(
310 P inside
(9) E = R3 
(P r) r 31 P outside

r3 0

To get the momentum density we use 1


(
29 0 P M inside
(10) pem = 0 R6 
(P r) r 31 P (M r) r 13 M outside
  
r6
The momentum density is a constant inside the sphere. The outside den-
sity requires a bit of vector juggling to get it into a useful form. We start
with the identity

(11) A (B C) = (A C) B (A B) C
Putting A = r, B = P and C = M, we get

(12) r (P M) = (r M) P (r P) M
Taking the cross product again:

(13) r (r (P M)) = (r M) (r P) (r P) (r M)
(14) = [(r M) (P r) + (r P) (r M)]
Expanding the RHS of 10 and using r r = 0 we get
MOMENTUM IN A MAGNETIZED AND POLARIZED SPHERE 3

(15)
   
1 1 1 1
(P r) r P (M r) r M = [(r M) (P r) + (r P) (r M)] + P M
3 3 3 9
1 1
(16) = r (r (P M)) + P M
3 9
So
(
2 0 P M inside
(17) pem = 0 9R6  1

3r6
r (r (P M)) + 3 P M outside
Without loss of generality we can choose the z axis to lie along P M.
In that case, the spherical angle is the angle between r and P M, so we
have

(18) r (P M) = |P M| sin
(19) r (r (P M)) = |P M| sin
To get the total momentum in the field, we need to integrate 17 over all
space, so we get (using the symbol P for total momentum)

(20) (
43 R3 29 0 P M = 27
8
0 R 3 P M inside
P = 0 R6 2 r2  1

3 R 0 0 r6 |P M| sin + 3 P M sin d d dr outside
To do the integral, we need to express in rectangular coordinates

(21) = cos cos x + cos sin y sin z


The integral will kill off the x and y components, so we are left with

2 0 R6 
dr 3 1
(22) Pout = 4
|P M| sin z + P M sin d
3 R r 0 3

2 0 R6  
dr 3 1
(23) = PM 4
sin + sin d
3 R r 0 3
2 0 R 3 
4 2

(24) = PM +
9 3 3
4
(25) = 0 R 3 P M
27
The total momentum is thus
MOMENTUM IN A MAGNETIZED AND POLARIZED SPHERE 4

8 4
(26) P = 0 R3 P M 0 R3 P M
27 27
4 3
(27) = 0 R P M
9
4
(28) = 0 R 3 M P
9
THE ELECTRON AS ELECTROMAGNETIC ENERGY AND
ANGULAR MOMENTUM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.11.
Early in the twentieth century, a classical model of the electron was pro-
posed in which the electron was taken to be a spherical shell of charge of
radius R spinning with angular speed , and that its entire mass (strictly,
mass-energy) was composed of the energy stored in its electromagnetic
fields, and its angular momentum was that of the electromagentic fields
as well.
First, we work out the energy, which is
 
1 1 2
(1) W= B + 0 E d 3 r
2
2 V 0
The electric field of a spherical shell is
(
0 r<R
(2) E= q
40 r2
r r>R
Griffiths works out the magnetic vector potential of a spinning spherical
shell of charge in his Example 5.11:
(
1
3 0 R r sin rR
(3) A= 1 4 sin
3 0 R r2 rR

from which we can get B from


(
2
3 0 R z rR
(4) B = A = 1 0 R4
 
3 r 3 2 cos r + sin rR
The surface charge density = q/4R2 so we get
(
0 q
6R z rR
(5) B= 0 qR2  
12r3
2 cos r + sin rR
1
THE ELECTRON AS ELECTROMAGNETIC ENERGY AND ANGULAR MOMENTUM 2

The energy is

(6)
2
0 qR2 4 cos2 + sin2 2

1 4 3  0 q 2 1
W= R + (2) r sin d dr+
20 3 6R 20 12 R 0 r6
 2
0 q 1 2
(2) 6
r sin d dr
2 40 R 0 r
(7)

0 2 2 0 q2 R4 2 3 cos2 + 1 q2 sin
= q R + 4
sin d dr + d dr
54 144 R 0 r 160 R 0 r4
(8)
0 q2 R 2 0 q2 R 2 q2
= + +
54 108 80 R
(9)
30 q2 R 2 q2
= +
108 80 R
To calculate the angular momentum, we first need the linear momentum.
The momentum density is

(10) pem = 0 E B
Only the region outside the sphere contributes (since E = 0 inside), so
we get

0 q2 R2 sin
(11) pem =
48 2 r5
(12) Lem = r pem
0 q2 R2 sin
(13) =
48 2 r4
The total angular momentum is
2
0 q2 R2 sin 2
(14) Lem = r sin d d dr
48 2 R 0 0 r4
To do the integral, we need to express in rectangular coordinates

(15) = cos cos x + cos sin y sin z


The integral will kill off the x and y components, so we are left with
THE ELECTRON AS ELECTROMAGNETIC ENERGY AND ANGULAR MOMENTUM 3

3
0 q2 R2 sin
(16) Lem = z d dr
24 R 0 r2
0 q2 R
(17) = z
18
The idea now is to equate the energy with relativistic rest mass and the
angular momentum with the quantum mechanical spin, so we get from 9
and 17

30 q2 R 2 q2
(18) + = mc2
108 80 R
0 q2 R h
(19) =
18 2
Solving for R and we get

1 18 h2 2 0 + q4 0
(20) R =
8 mc2 0 0 q2
h 2 mc2 0
(21) = 72
18 h2 2 0 + q4 0
We can plug in the values of the various constants (all in SI units):

(22) h = 1.0546 1034


(23) m = 9.109 1031
(24) 0 = 8.854 1012
(25) 0 = 1.2566 106
(26) c = 3.0 108

and we get

(27) R = 2.978 1011 m


(28) = 3.105 1021 s1
This gives an equatorial speed of

(29) R = 9.246 1010 m s1


THE ELECTRON AS ELECTROMAGNETIC ENERGY AND ANGULAR MOMENTUM 4

which is more than 300 times the speed of light, so this model of the electron
doesnt work.
ANGULAR MOMENTUM OF A CHARGE AND MAGNETIC
MONOPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.12.
Returning to the fantasy land where magnetic monopoles exist, suppose
we have a single electric charge e and a single magnetic monopole (mag-
netic charge) b. Then the fields due to these charges are

e
(1) E = r
40 r2
0 b
(2) B = rb
4rb2

where weve placed e at the origin. The magnetic charge b is not at the ori-
gin, so rb is the vector from bs location to the point where were measuring
the fields. Wed like to find the total angular momentum contained in these
two fields.
Well place b on the z axis a distance d from e, so its location is d z. Then
the vectors d z, r and rb form a triangle with the spherical angle lying
between d z and r, so it is the angle opposite the side rb . Therefore

(3) rb = r d z

(4) = rr d cos r sin
(5) = (r d cos ) r + d sin

By the cosine rule for triangles

(6) rb2 = d 2 + r2 2dr cos

so we get
1
ANGULAR MOMENTUM OF A CHARGE AND MAGNETIC MONOPOLE 2

e
(7) E = r
40 r2
0 b  
(8) B = 3/2
(r d cos ) r + d sin
4 (d 2 + r2 2dr cos )
The momentum density is

(9) pem = 0 E B
0 ebd sin
(10) = 2

16 r2 (d 2 + r2 2dr cos )3/2

and the angular momentum density is

(11) Lem = r pem


0 ebd sin
(12) =
16 2 r (d 2 + r2 2dr cos )3/2
To find the total angular momentum, we integrate over all space

(13) 2
0 ebd sin
L= r2 sin d drd
16 2 0 0 0 r (d 2 + r2 2dr cos )
3/2

As usual, we must first express in rectangular coordinates

(14) = cos cos x + cos sin y sin z


The integral will kill off the x and y components, so we are left with


0 ebd r sin3
(15) L= z (2) drd
16 2 0 0 (d 2 + r2 2dr cos )
3/2

This is a rather unpleasant integral that can be done using Maple, with
the result


r sin3 2
(16) 3/2
drd =
0 0 (d 2 + r2 2dr cos ) d

so the angular momentum becomes


ANGULAR MOMENTUM OF A CHARGE AND MAGNETIC MONOPOLE 3

0 eb
(17) L= z
4
I could leave it at that, but for once I decided to solve the integral by hand,
just to see how its done (since the difficulty of the integral would seem to
be the main reason Griffiths has this marked as a hard problem). It seems
easiest to do the r integral first, so we can rewrite the integral as


3 r
(18) d sin 3/2
dr
0 0 (d 2 + r2 2dr cos )

If we had the derivative of the expression in parentheses in the denom-


inator present in the numerator, the integral would be easy. So we can try
something like this (Ill call cos c and sin s to save a bit of writing;
is a constant in the r integral anyway so it shouldnt cause any confusion):

(19)

r 1 2r 2dc + 2dc
3/2
dr = 3/2
dr
0 (d 2 + r2 2drc) 2 (d 2 + r2 2drc)
0

1 1
(20) = + dc 3/2
dr
d 2 + r2 2drc 0
0 (d 2 + r 2 2drc)

1 1
(21) = + dc 3/2
dr
d 0 (d 2 + r 2 2drc)

We can now complete the square in the remaining integral to get


1 1
(22) 3/2
dr = 3/2 dr
(d 2 + r2 2drc)

0 0 2 2 2 2
(r dc) + (d d c )

1
(23) =  3/2 dr
0 2 2 2
(r dc) + d s

3/2
The last integral is of the form dx/ a2 + x2 which can be looked
up in tables, but just to be complete, lets work that one out too. We can use
a trigonometric substitution
ANGULAR MOMENTUM OF A CHARGE AND MAGNETIC MONOPOLE 4

(24) r dc = ds tan u
(25) dr = ds sec2 udu
(r dc)2 + d 2 s2 = d 2 s2 1 + tan2 u

(26)
(27) = d 2 s2 sec2 u
Putting all this into 23 we get (well leave the limits off for now):

1 ds sec2 udu
(28)  3/2 dr = d 3 s3 sec3 u
(r dc)2 + d 2 s2

1
(29) = 2 2 cos u du
d s
1
(30) = 2 2 sin u
d s
To convert back to r, we use
r
1
(31) cos u =
sec2 u
r
1
(32) =
1 + tan2 u
v
1
u
(33) =
u
t 2
1 + (rdc)
2
d s 2
p
(34) sin u = 1 cos2 u
r dc
(35) = q
d 2 s2 + (r dc)2

so the integral comes out to (restoring the limits):





1 1 r dc
(36)  3/2 dr = d 2 s2 q
2
2

0 2 2
d s + (r dc)
(r dc) + d 2 s2 0
1
(37) = (1 + c)
d 2 s2
Since the integrand is positive over its entire range, we must take the +
sign in the answer. We can now plug this back into 21 to get
ANGULAR MOMENTUM OF A CHARGE AND MAGNETIC MONOPOLE 5


r 1 dc
(38) 3/2
+ 2 2 (1 + c)
dr =
0 (d 2 + r 2 2drc) d d s
 
1 c (1 + c)
(39) = 1+
d 1 c2
 
1 c (1 + c)
(40) = 1+
d (1 c) (1 + c)
1
(41) =
(1 cos ) d
Plugging this into 16, we now have the remaining integral over :

(42)


r sin3 1
sin3
drd = d
0 0 (d 2 + r2 2dr cos )
3/2 d 0 1 cos

1 sin 1 cos2

(43) = d
d 0 1 cos

1 sin (1 cos ) (1 + cos )
(44) = d
d 0 1 cos

1
(45) = (sin + sin cos ) d
d 0
2
(46) =
d
Q.E.D.
Now you see why I use Maple to work out integrals!
ANGULAR MOMENTUM CONSERVATION: EXAMPLE WITH A
SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.13.
Well revisit the earlier problem of the two charged cylinders and the
solenoid. To reiterate, we have a long solenoid with n turns per unit length
carrying current I0 and a radius R, with its axis along the z axis. The mag-
netic field inside the solenoid is

(1) B0 = 0 nI z

The field is zero outside the solenoid.


We add two other cylinders (not solenoids), both coaxial with the solenoid.
One cylinder has radius a < R (so it lies inside the solenoid) and carries sur-
face charge +Q; the other cylinder has radius b > R (outside the solenoid)
and carries charge Q. Both cylinders have length `. From Gausss law,
the electric field between these two cylinders is, for a < r < b

Q r
(2) E0 =
20 ` r

That is, the field points radially outward from the axis. The electric field is
zero for r < a and r > b. (Were neglecting end effects, so were assuming
that `  b > a.)
In our earlier solution, we worked out the angular momentum contained
in the fields and showed that it is equal to the mechanical angular mo-
mentum transferred to the two cylinders if the current in the solenoid is
slowly reduced. However, there is another effect that we neglected: when
the charged cylinders start to rotate, they generate a changing magnetic field
inside them which in turn creates a circumferential electric field in the space
between the cylinders. When the final rotation speeds of the two cylinders
are reached (that is, when the current through the solenoid has been reduced
to zero), the cylinders continue rotating, thus generating a static magnetic
field that interacts with the electric field to produce an extra amount of an-
gular momentum in the fields. Well consider this static magnetic field first.
1
ANGULAR MOMENTUM CONSERVATION: EXAMPLE WITH A SOLENOID 2

The rotating cylinders are effectively solenoids themselves. The surface


charge density on the two cylinders is
(
Q
2a` inner cylinder
(3) a,b = Q
2b` outer cylinder
The surface current densities are
(
Q(aa )
2a` inner cylinder
(4) Ka,b = b)
Q(b
2b` outer cylinder

so the magnetic field due to each cylinder is


(
0 Qa
(5) Ba,b = 2` z inner cylinder
0 Qb
2` z outer cylinder
To get the directions of B we see from Griffithss Example 8.4 that the
angular momentum of the inner cylinder is in the +z direction and of the
outer cylinder is in the z direction. Since the outer cylinder has negative
charge, however, its magnetic field points in the same direction as that of
the inner cylinder. Another way of putting this is to note that the angular
velocities are

(6) a = a z
(7) b = b z

so the minus sign for b cancels the minus sign for the charge on the outer
cylinder.
The linear momentum density is non-zero only in the region a r b
and is

(8) pem = 0 E B
0 Q2 b
(9) = 2 2
4 ` r
The angular momentum density is

(10) Lem = r pem


0 Q2 b
(11) = z
4 2 `2
ANGULAR MOMENTUM CONSERVATION: EXAMPLE WITH A SOLENOID 3

which is constant, so the total angular momentum is

0 Q2 b 
b2 a2 ` z
 
(12) L = 2 2
4 `
0 Q2 b 2
b a2 z

(13) =
4`
Now we can look at whats happening as the cylinders are spinning up to
their final speeds. As they speed up, the magnetic field due to each cylinder
is changing according to

(
0 Qa
Ba,b 2` z inner cylinder
(14) = 0 Qb
2` z outer cylinder
t

where the dot indicates a time derivative. According to Faradays law, this
changing magnetic field induces a circumferential electric field:


Ba,b
(15) E d` = da
t

This electric field will exert a torque on each cylinder, whose integral over
time will give the angular momentum transferred to the cylinders. First we
need to calculate the field at each cylinder. We choose a circular path of
integration at the surface of each cylinder. Remember that the field of the
inner cylinder is non-zero only for r < a, which the the field for the outer
cylinder covers the entire region r < b.

1 0 Q (a + b )
(16) Ea = a2
2a 2`
1 0 Q a2 a + b2 b

(17) Eb =
2b 2`
To confirm the direction of E, recall Lenzs law, which states that the
induced field opposes the change that produced it. Since the magnetic field
is increasing in the +z direction, the induced electric field must oppose this
increase so it must be in the direction.
The torque on the cylinders is then
ANGULAR MOMENTUM CONSERVATION: EXAMPLE WITH A SOLENOID 4

(18) Na = r Fa
(19) = r QEa
0 Q2 (a + b )
= a2

(20) r
4`
0 Q2 (a + b )
(21) = a2 z
4`
(22) Nb = r QEb
0 Q2 a2 a + b2 b

(23) = z
4`
Adding them together we get

(24) N = Na + Nb
0 Q2 b2 a2 b

(25) = z
4`
Integrating over the time it takes to reach the final speed b we get

0 Q2 b2 a2 b

(26) L= z
4`

which is equal and opposite to 13. Thus the total angular momentum intro-
duced into the system by magnetic and electric fields induced by the rotating
cylinders is zero, showing that angular momentum is conserved.
MOMENTUM OF A POINT CHARGE OUTSIDE A SOLENOID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.14.
Another example of calculating momentum and angular momentum in
electromagnetic fields. We have an infinite solenoid along the z axis, of
radius R with n turns per unit length and carrying current I. At position
ax there is a point charge q. We want to find the momentum and angular
momentum of the resulting fields.
The first step in this problem is deciding which coordinate system to
use. The solenoid makes us think of cylindrical coordinates, while the point
charge suggests spherical. However, these two systems dont mesh well, so
we can try the fallback of using rectangular coordinates.
The field of the solenoid is zero outside and uniform inside, where it is

(1) B = 0 nI z for x2 + y2 < R2


If the point charge were located at the origin, its field would be

q xx + yy + zz
(2) E=
40 (x2 + y2 + z2 )3/2
If we shift the charge to ax then we get

q (x a) x + yy + zz
(3) E= 3/2
40 
(x a)2 + y2 + z2

The momentum density is

(4) pem = 0 E B
0 nIq
(5) =  3/2 [ (x a) y + yx]
2
4 (x a) + y2 + z2

To get the total momentum we need to integrate this over the interior of
the solenoid. It turns out to be easiest to do this by integrating first over z
1
MOMENTUM OF A POINT CHARGE OUTSIDE A SOLENOID 2

and then converting to polar coordinates for the remaining two integrations.
Integrating over z we get (using Maple or tables)

(6)

0 nIq dz 0 nIq 2
[ (x a) y + yx] = [ (x a) y + yx]
4 3/2 4 (x a)2 + y2
 

(x a)2 + y2 + z2
0 nIq (a x) y + yx
(7) =
2 (x a)2 + y2

If we integrate the x component over y we have


R2 x2
yx
(8) =0
R2 x2 (x a)2 + y2

because the integrand is an odd function of y and the integral is over a


symmetric interval. This leaves us with the y component, and it is here that
we turn to polar coordinates. Using x = r cos and y = r sin we have

R 2 R 2
(a x) y (a r cos ) ry
(9) 2
r d dr = d dr
0 0 (x a) + y2 0 0 r2 2ar cos + a2
R
2r
(10) = y dr
0 a
R2
(11) = y
a

where we used Maple to do the integral. The total momentum is thus

0 nIq R2 0 nIqR2
(12) Pem = y = y
2 a 2a

The angular momentum density is


MOMENTUM OF A POINT CHARGE OUTSIDE A SOLENOID 3

(13)
Lem = r pem
0 nIq
(14) =  3/2 [xx + yy + zz] [ (x a) y + yx]
2 2 2
4 (x a) + y + z
0 nIq  2
 
(15) =  3/2 z (x a) x + yzy + x (a x) y z
4 (x a)2 + y2 + z2
Integrating the x and y components over z gives zero because the inte-
grand is an odd function of z integrated over a symmetric interval. Thus we
are left with the z component which we can again integrate first over z and
then over the other two coordinates using polar coordinates. We have
R 2 x (a x) y2 r d dr

0 nIq
(16) Lem = z dz  3/2
4 0 0 2 2 2
(x a) + y + z
R 2 2 x (a x) y2

0 nIq
(17) = z r d dr
4 0 0 (x a)2 + y2
R r cos (a r cos ) r2 sin2 r
2

0 nIq
(18) = z d dr
2 0 0 r2 2ar cos + a2
R 2
ar cos r2 r

0 nIq
(19) = z d dr
2 0 0 r2 2ar cos + a2
R
2r2 2r2

0 nIq
(20) = z r dr
2 0 a2 r2 a2 r2
(21) =0

where again we used Maple to do the integral.


WAVE EQUATION: DERIVATION AND EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.1.
As a prelude to the study of electromagnetic waves, well have a look at
a derivation of the wave equation.
Waves can appear in any form of matter, as well as in electromagnetic
fields, so well look at the easiest case for a derivation. Suppose we have
a string under a tension T . Initially, well stretch out the string along the z
axis so that its perfectly straight. Now suppose we pull the string slightly
to one side (in the x direction, say) so that the string now follows a slightly
curved path. Suppose that at position z the tangent to the string makes an
angle with the z axis and at a slightly different position z + dz the angle
is + d . The tension can be resolved into an x component (perpendicular
to the original orientation of the string) and a z component (parallel to the
original orientation). We get, at position z:

(1) Tx (z) = T sin


(2) Tz (z) = T cos

and at position z + dz:

(3) Tx (z + dz) = T sin ( + d )


(4) Tz (z + dz) = T cos ( + d )
For small angles, the sine and tangent are equal to first order:

(5) sin tan

and the tangent is the slope of tangent to the string. Since the string was
pulled back in the x direction, x is the displacement, so

x
(6) tan =
z
1
WAVE EQUATION: DERIVATION AND EXAMPLES 2

and the difference in the x component of the tension between the two points
on the string is

(7) Tx (z + dz) Tx (z) T tan ( + d ) T tan


 
x x
(8) = T
z z
z+dz z
The second derivative is defined as

2x
 
1 x x
(9) 2
= lim
z dz0 dz z z+dz z z

so the difference in Tx is, in the limit dz 0

2x
(10) dTx = T dz
z2
However, from Newtons law F = ma, the net force on the string segment
is also equal to the mass of that string segment times its acceleration. If the
mass per unit length is , then

2x
(11) dTx = dz
t 2
Equating these last two formulas gives us the wave equation

2x 2x
(12) =
z2 T t 2
The coefficient /T has the units of mass divided by force, which is
velocity squared, so we can define a velocity
s
T
(13) v

and write the wave equation as

2x 1 2x
(14) =
z2 v2 t 2
Although we havent shown it here, it turns out that v is the speed of
propagation of the wave.
WAVE EQUATION: DERIVATION AND EXAMPLES 3

It turns out that any function of form

(15) x = f (z vt)

is a solution of the wave equation as can readily be seen by direct differen-


tiation.
Example 1. With

2
(16) f = Aeb(zvt)

we have

2x b(zvt)2 2 2 2

(17) = 2bAe 2bv t 4bzvt + 2bz 1
z2
2x 2 b(zvt)2 2 2 2
 2
2 x
(18) = 2bAv e 2bv t 4bzvt + 2bz 1 = v
t 2 z2

so 14 is satisfied.
Example 2. With

(19) f = A sin (b (z vt))

we have

2x
(20) = Ab2 sin (b (z vt))
z2
2x 2 2
2
2 x
(21) = Ab v sin (b (z vt)) = v
t 2 z2
Example 3. With

A
(22) f=
b (z vt)2 + 1

we have
WAVE EQUATION: DERIVATION AND EXAMPLES 4

2x 8Ab2 (z vt)2 2Ab


(23) 2
=  3  2
z
b (z vt)2 + 1 b (z vt)2 + 1
2x 8Ab2 v2 (z vt)2 2Abv2 2
2 x
(24) =
3  2 = v
t 2 z2

b (z vt)2 + 1 b (z vt)2 + 1

Functions that dont have the form 15 dont satisfy the wave equation.
Example 4. With

2 +vt
(25) f = Aeb(bz )

we have

2x 2
= 2Ab2 eb(bz +vt ) 2b2 z2 1

(26) 2
z
2x 2 2 b(bz2 +vt ) 2 x
2
(27) = Ab v e 6
= v
t 2 z2
Example 5. With

(28) f = A sin (bz) cos (bvt)3

we have

(29)
2x
2
= Ab2 sin (bz) cos (bvt)3
z
(30)
2x 3 3

3 3 3
 2
2 x
= 3A sin (bz) b v t 3 (bvt) cos (bvt) + 2 sin (bvt) 6
= v
t 2 z2
P INGBACKS
Pingback: Wave equation: wave speed and standing waves
Pingback: Wave equation: sinusoidal waves and complex notation
Pingback: Wave equation: solution by separation of variables
Pingback: Waves: boundary conditions
Pingback: Waves: boundary condition with a massive knot
WAVE EQUATION: DERIVATION AND EXAMPLES 5

Pingback: Waves in a viscous fluid


Pingback: Electromagnetic waves in vacuum
Pingback: Wave solution of the weak-field Einstein equation
WAVE EQUATION: WAVE SPEED AND STANDING WAVES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.2.
Any function of form f = f (z vt) (for a constant v) is a solution of the
wave equation

2 f 1 2 f
(1) =
z2 v2 t 2
Once we specify f , we can plot it as a function of z for a fixed value of t,
say t1 . This shows a snapshot of the wave at a particular time. If we then plot
f as a function of z at some later time t2 , the wave has the same shape but is
shifted along the z axis. A point at which f has the value f1 f (z1 vt1 )
will have moved to a position z2 where z2 vt2 = z1 vt1 . In other words,
that particular point on the wave moves from z1 to z2 in a time t2 t1 , and
must be moving with speed

z2 z1 v (t2 t1 )
(2) = =v
t2 t1 t2 t1
Thus v is the speed of the wave. If z2 > z1 , then v > 0 and the wave
moves towards increasing z. If z2 < z1 , then v < 0 and the wave moves
towards decreasing z.
Since the wave equation is linear, any linear combination of solutions
is also a solution, so we can add together waves with equal and opposite
speeds. For example, if we have two sine waves with equal and opposite
speeds, we can add them together to get

(z,t) = sin (k (z vt)) + sin (k (z + vt)) = sin kz cos kvt cos kz sin kvt +
f(3)
sin kz cos kvt + cos kz sin kvt
(4) = 2 sin kz cos kvt

where k is a constant.
The latter form is not a function of form f (z vt) but as its a sum of
two functions of this type, it is still a solution of the wave equation. In fact,
it represents a standing wave, that is, a wave which doesnt travel along the
1
WAVE EQUATION: WAVE SPEED AND STANDING WAVES 2

z axis, but rather oscillates in place. We can see this since for all points
zn = n/k where n is an integer, f (zn ,t) = 0 for all times t. For all times
tn = (2n + 1) /2kv, f (z,tn ) = 0 for all positions, so the wave becomes a
flat line. In between these times, the wave oscillates between its maximum
and minimum heights of 2 sin kz.
P INGBACKS
Pingback: Wave equation: sinusoidal waves and complex notation
WAVE EQUATION: SINUSOIDAL WAVES AND COMPLEX
NOTATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.3.
Any function of form f = f (z vt) (for a constant v) is a solution of the
wave equation

2 f 1 2 f
(1) =
z2 v2 t 2
Probably the most commonly used solution is the sine wave, which can
be written as either a sine or a cosine, either of which gives the most general
sinusoidal solution. For example, we can write

(2) f (z,t) = A cos (k (z vt) + )

where k, v and are constants. The wavelength is the distance (at constant
time) between two successive peaks, which occurs when z advances far
enough to increase the argument of the cosine by 2. Therefore

(3) k = 2
2
(4) =
k
The parameter k is called the wave number.
As we saw earlier, v is the speed of the wave. The parameter is the
phase of the wave, and allows for waves that dont have their maximum at
z vt = 0. Suppose that at t = 0 the maximum of the wave that is closest
to z = 0 is at z = zm . Then the argument of the cosine must be zero at that
point, so

(5) kzm =

(6) zm =
k
1
WAVE EQUATION: SINUSOIDAL WAVES AND COMPLEX NOTATION 2

Thus a positive phase ( > 0) means that the peak of the wave lags behind
z = 0, while a negative phase ( < 0) means the peak leads z = 0.
Its more common to write the wave in terms of k and the angular fre-
quency, defined as

2v
(7) kv =

so we get

(8) f (z,t) = A cos (kz t + )


It turns out that its often more convenient to represent a sinusoidal wave
as the real part of a complex exponential. That is
 
(9) f (z,t) = Aei(kzt+ )
We can then define a complex wave function

(10) f (z,t) Aei(kzt)


(11) A Aei
Physical waves are, of course, always real functions (well, except in
quantum mechanics, where complex wave functions are the norm, but even
there, any physical interpretation of the quantum wave function requires
extracting a real value from the complex function).
Working with complex exponentials is, in most cases, easier than work-
ing with sines and cosines, although the formulas arent quite trivial. For
example, if we want to add two waves f1 and f2 then we get

(12) f3 = f1 + f2
= f1 + f2

(13)
If the two waves to be added have the same k and , then we need to find
the new amplitude A3 and phase 3 in terms of the amplitudes and phases
of the constituent waves. That is

(14) f3 = A3 ei(kzt)
A3 ei3 ei(kzt) = A1 + A2 ei(kzt)

(15)
(16) A3 ei3 = A1 ei1 + A2 ei2
WAVE EQUATION: SINUSOIDAL WAVES AND COMPLEX NOTATION 3

Taking the square modulus of both sides, we get


 
(17) A23 = A21 + A22 + A1 A2 ei(1 2 ) + ei(1 2 )
(18) = A21 + A22 + 2A1 A2 cos (1 2 )
q
(19) A3 = A21 + A22 + 2A1 A2 cos (1 2 )
To get the phase, we can use

(20) A3 = A3 (cos 3 + i sin 3 )


A3
(21) tan 3 =
A3

A1 + A2
(22) = 
A1 + A2
A1 sin 1 + A2 sin 2
(23) =
A1 cos 1 + A2 cos 2
 
A1 sin 1 + A2 sin 2
(24) 3 = arctan
A1 cos 1 + A2 cos 2
Keep in mind that these formulas work only for adding two waves with
the same wave number k and frequency .
P INGBACKS
Pingback: Waves: boundary condition with a massive knot
Pingback: Waves: polarization
Pingback: Electromagnetic waves in vacuum
WAVE EQUATION: SOLUTION BY SEPARATION OF
VARIABLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.4.
We can use separation of variables to solve the wave equation

2 f 1 2 f
(1) =
z2 v2 t 2
As usual, we propose a solution of form

(2) f0 (z,t) = Z (z) T (t)


Substituting into the wave equation and dividing through by ZT we get

1 d2Z 1 d2T
(3) =
Z dz2 v2 T dt 2
Since the LHS depends only on z and the RHS only on t, both sides must
be equal to a constant, which we can call k2 . Thus

1 d2Z
(4) = k2
Z dz2
1 d2T
(5) = k2
v2 T dt 2
The general solutions are

(6) Z (z) = Aeikz + Beikz


(7) T (t) = Ceikvt + Deikvt
(8) = Ceit + Deit

where kv. Therefore


1
WAVE EQUATION: SOLUTION BY SEPARATION OF VARIABLES 2

 
f0 (z,t) = Aeikz + Beikz Ceit + Deit

(9)
(10) = ADei(kzt) + BCei(kzt) + ACei(kz+t) + BDei(kz+t)
The most general solution is the weighted integral of this quantity over
all values of k, that is

(11) h i
f (z,t) = c (k) ADei(kzt) + BCei(kzt) + ACei(kz+t) + BDei(kz+t) dk
0
If we allow k (and therefore also ) to take on negative and positive
values, we can expand the integral to and combine terms 1 and 2, and
terms 3 and 4:
 
(12) f (z,t) = A1 (k) ei(kzt) + A2 (k) ei(kz+t) dk

Technically, this is as far as we can go if we want the full complex solu-
tion, but in reality we are interested only in the real part. The real part of
the first exponential is the same as the real part of the second exponential,
while the imaginary parts are equal and opposite, so from a physical point
of view, we can write the general solution as

(13) f (z,t) = A (k) ei(kzt) dk

where weve added tildes to indicate that this is a physical (rather than a
proper mathematical) solution, and that we should look only at the real part
of f to get the actual equation of the wave. (We could equally well have used
the second exponential in our physical solution, but the first exponential is
more traditional.)
P INGBACKS
Pingback: Fourier transform of superposition of plane waves
WAVES: BOUNDARY CONDITIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.5.
For a one dimensional wave (on a string, say) suppose we now place a
boundary at the point z = 0. For a string, this could be a point at which one
string is joined with another string of a different mass per unit length. If we
send a wave down the string from large negative z, when this wave reaches
z = 0, there will be a reflected wave that returns towards negative z and a
transmitted wave that proceeds beyond z = 0 towards positive z. We can
get some idea of the nature of these reflected and transmitted waves if we
impose some boundary conditions at the point z = 0.
First, we require the wave function f to be continuous, for the simple
reason that there is no break in the string at z = 0. The second boundary
condition requires that the derivative f / z is also continuous. The reason
for this is a bit more subtle. Assuming there is no point mass (such as a
knot) at the joining position, if the tangent to the string at that point were
not continuous, then the second derivative would be infinite, meaning that
there would be an infinite force at that point.
To see how the incident, reflected and transmitted waves are related, sup-
pose we have an incident wave I(z v1t), a reflected wave R (z + v1t) and a
transmitted wave T (x v2t), where I and T are moving to the right, with I
defined for z < 0 and T for z > 0, and R moving to the left for z < 0.
The continuity of the wave function at z = 0 gives us

(1) I (v1t) + R (v1t) = T (v2t)

The continuity of the derivative, if applied directly, just gives the same
equation with each function replaced by its derivative, so doesnt help much:


I R T
(2) + =
z z=0 z z=0 z z=0+

However, if we consider the original definition of a derivative as a limit,


we can make some progress. Consider first the derivative of the incident
wave just below z = 0, at time t = 0:
1
WAVES: BOUNDARY CONDITIONS 2


I I (0) I (z)
(3) = lim
z z=0 z0
z
The wave amplitude at the point (z,t) = (z, 0) will be at z = 0 after it
travels the distance z, which takes a time t = z/v1 .
By a similar argument, the derivative of the reflected wave is

R R (0) R (z)
(4) = lim
z z=0 z0
z
This time, the wave amplitude at (z, 0) was at z = 0 at time t = z/v1
since this wave is travelling to the left. Finally, for the transmitted wave

T T (z) T (0)
(5) = lim
z z=0+ z0
z

since this wave is defined for z > 0. The wave amplitude at (z, 0) was at
z = 0 at time t = z/v2 since the transmitted wave is travelling to the right
with speed v2 .
We can now use the continuity condition 1 to eliminate either R or T
from the limits. Start by eliminating R by evaluating everything at time
t = z/v1 :

(6)     
R (0) R (z) 1 v2
lim = lim T (0) I (0) T z I (z)
z0 z z0 z v1
Now we can insert this into the continuity equation for derivatives 2, and
well leave off the limit and 1/z to simplify the notation:

(7)
   
v2
I (0) I (z) + T (0) I (0) T z I (z) = T (z) T (0)
v1
(8)
   
v2
(I (0) I (z)) + (I (z) I (0)) = T z T (0) + (T (z) T (0))
v1
Restoring the limit and 1/z we get
 
I v2 T
(9) 2 = +1
z v1 z
WAVES: BOUNDARY CONDITIONS 3

This condition is strictly true only at z = 0, but it must be true for all
times. We can convert the derivatives into time derivatives by noting that
since I = I (z v1t) we have

I 1 I
(10) =
z v1 t
Similarly for R and T :

T 1 T
(11) =
z v2 t
R 1 R
(12) =
z v1 t
Returning to 9 we get
 
2 I 1 v2 T
(13) = +1
v1 t v2 v1 t
T 2v2 I
(14) =
t v1 + v2 t
At z = 0, we can integrate with respect to time to get

2v2
(15) T (v2t) = I (v1t) + KT
v1 + v2
where KT is a constant of integration. Although I, T and R are functions
of both z and t, they are all actually functions of only one variable, since z
and t must always occur in the combination zv1,2t. Thus what 15 is saying
is that, if we have the incident wave in the form I (u), then the transmitted
wave has the form
 
v2 2v2
(16) T u = I (u) + KT
v1 v1 + v2

or, conversely
 
2v2 v2
(17) T (u) = I u + KT
v1 + v2 v1
 
For T (u), u = z v2t for z 0 (and all times t), while for I vv12 u , u =
z v1t for z 0. If we pick a particular numerical value for u, say 42, then
we can write
WAVES: BOUNDARY CONDITIONS 4

 
2v2 v2
(18) T (zT v2tT ) = I (zI v1tI ) + KT
v1 + v2 v1
 
2v2 v2
(19) T (42) = I 42 + KT
v1 + v2 v1

and this equation is valid for all values of z and t such that

(20) zT v2tT = 42 (z 0)
(21) zI v1tI = 42 (z 0)
That is, the values of zT and zI need not be equal, and neither must tT = tI .
All that matters is that zT v2tT = zI v1tI .
For the reflected wave, we eliminate T using 1 at time t = z/v2 :

(22)      
T (z) T (0) 1 v1 v1
lim = lim I z + R z I (0) R (0)
z0 z z0 z v2 v2
Substitute into 2, again without the limit and 1/z to simplify the nota-
tion:

(23)
   
v1 v1
I (0) I (z) + R (0) R (z) = I z + R z I (0) R (0)
v2 v2
(24)
      
v1 v1
(I (0) I (z)) I z I (0) = R (0) R z (R (0) R (z))
v2 v2
Restoring the limit and 1/z we get
   
v1 I v1 R
(25) 1 = 1+
v2 z v2 z
   
1 v1 I 1 v1 R
(26) 1 = 1+
v1 v2 t v1 v2 t
v2 v1
(27) R (+v1t) = I (v1t)
v1 + v2
v2 v1
(28) R (u) = I (u)
v1 + v2
Again, for R, u = z + v1t with z 0 and for I, u = z v1t with z 0.
These results apply to any wave shape, not just to sinusoidal waves.
WAVES: BOUNDARY CONDITIONS 5

P INGBACKS
Pingback: Waves: boundary condition with a massive knot
Pingback: Waves in a viscous fluid
WAVES: BOUNDARY CONDITION WITH A MASSIVE KNOT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.6.
In our earlier analysis of the boundary conditions for a one dimensional
wave, we looked at the case of one string of mass per unit length 1 joined at
z = 0 to another string with mass density 2 , and we assumed that the knot
joining the two strings was massless. This led to the boundary conditions
being that both the wave function and its first derivative are continuous at
z = 0.
Suppose we now take the knot to have a non-zero mass m. In that case,
the derivative no longer needs to be continuous at the knot. We can analyze
the situation the same way as we derived the original wave equation. Since
the tensions on either side of the knot are equal in magnitude but can now
point in different directions, we have the net transverse force on the knot
 
f f
(1) Ft = T
z 0+ z 0

where T is the tension in the strings.


This force is equal to the knots mass times its transverse acceleration, so

2 f
 
f f
(2) T =m 2
z 0+ z 0 t 0
Now suppose we have an incident sinusoidal wave

(3) I (z,t) = AI ei(k1 zt)

using complex notation, so AI is a complex amplitude with real magnitude


aI and phase 1 :

(4) AI = aI ei1
(Ill drop the tildes that Griffiths uses in his book, since well be dealing
with complex notation pretty well exclusively until we extract the physical
meaning at the end.)
1
WAVES: BOUNDARY CONDITION WITH A MASSIVE KNOT 2

When this wave hits the knot, part of it is reflected and part is transmitted,
so the overall wave function is
(
AI ei(k1 zt) + AR ei(k1 zt) z<0
(5) f (z,t) =
AT ei(k2 zt) z>0
The frequency is the same on both sides of z = 0, but the wave numbers
k1 and k2 (and thus the wave speeds vi = /ki ) are different. The continuity
condition gives us

(6) AI + AR = AT

and the derivative condition 2 gives us

m 2
(7) ik2 AT ik1 (AI AR ) = AT
T
These two equations can be solved to give AR and AT in terms of AI .
Example. Suppose the string for z < 0 has a non-zero mass 1 so the waves
speed v1 and wave number k1 are both finite and non zero. However, the
string for z > 0 is taken to be very light (effectively massless). In this case,
the speed is (as we saw in the derivation of the wave equation)
s
T
(8) v2 = =
2

so


(9) k2 = =0
v2
Then from 7 and 6 we have

T k1
(10) i (AI AR ) = AT = AI + AR
m 2
With the definition

T k1
(11)
m 2

we can solve this to get


WAVES: BOUNDARY CONDITION WITH A MASSIVE KNOT 3

i 1
(12) AR = AI
1 + i
2i
(13) AT = AI
1 + i
Taking magnitudes we get the real amplitudes of the reflected and trans-
mitted waves:

(14) aR = aI
2
(15) aT = aI
1 + 2
To get the phases, we can find the ratio of the imaginary and real parts in
12 and 13 to get the tangents of the phases. This is a bit messy so its easiest
to use Maple to do the algebra. We get

2 cos 1 + 2 1 sin 1

(16) tan R =
( 2 1) cos 1 2 sin 1
2 + 2 1 tan 1

(17) =
( 2 1) 2 tan 1
2
2 1
+ tan 1
(18) =
1 22 1 tan 1

We can simplify this equation by using the formula for the tangent of the
sum of two angles:

tan u + tan v
(19) tan (u + v) =
1 tan u tan v
We get

 
2
(20) tan R = tan arctan 2 + 1
1
2
(21) R = arctan 2 + 1
1
For the transmitted wave, we have
WAVES: BOUNDARY CONDITION WITH A MASSIVE KNOT 4

cos 1 + sin 1
(22) tan T =
sin 1 + cos 1
1
+ tan 1
(23) =
1 1 tan 1
 
1
(24) = tan arctan + 1

1
(25) T = arctan + 1

It might seem odd that the reflected wave has the same amplitude as
the incident wave, yet the transmitted wave still has a non-zero amplitude.
Doesnt this violate conservation of energy, since the kinetic energy of the
string due to the reflected wave is equal to that of the incident wave, so how
can there be any energy left over for a transmitted wave? The answer is that
since the transmitted wave is in a massless string, it carries no energy, so
energy can still be conserved.
WAVES IN A VISCOUS FLUID

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.7.
Another example of a transverse wave is that of a string embedded in
a viscous fluid which adds a drag force on the string. The drag force is
proportional to the strings transverse speed, so we need to insert a term

f
(1) Fdrag = z
t

where is a constant determined by the viscosity of the fluid, into our


derivation of the original wave equation. This gives us a modified wave
equation:

2 f f 2 f
(2) T =
z2 t t 2

where T is the tension in the string and is the mass per unit length of the
string.
If we assume that strings frequency is constant, then we can write

(3) f (z,t) = eit F (z)

and the wave equation reduces to an ODE:

(4) T F 00 + iF = 2 F
i + 2
(5) F 00 = F 2F
T
The general solution is

(6) F (z) = Aez + Bez


However, is a complex number, so F contains both oscillatory and
exponential parts.
1
WAVES IN A VISCOUS FLUID 2

Using Maple, we can find the :


"r q r q #
1
(7) = 2 2 + 2 2 i 2 2 + 2 + 2
2T
To keep F (z) finite for large z (were assuming the string extends from
z = 0 to z = +), we must have A = 0 and therefore

(8) ( "r q r q #)
z
F (z) = AT exp 2 2 + 2 2 i 2 2 + 2 + 2
2T

where AT is the (complex) amplitude of the wave.


We can think of the imaginary part of the exponential as the wave number
k2 and the real part as a spatial decay factor . That is
r q
1
(9) k2 2 2 + 2 + 2
2T
r q
1
(10) 2 2 + 2 2
2T
(11) = ik2
(12) F (z) = AT e z+ik2 z
Because of the negative real part in the exponent, the wave is attenuated
(its amplitude falls off with distance) with a characteristic penetration dis-
tance d of

1 2T
(13) d= =q p

2 2 + 2 2
This is the distance at which the amplitude falls to 1/e of its value at
z = 0.
Finally, we can consider the case of two strings joined by a massless knot
at z = 0, with the string on the right embedded in the viscous fluid. If the
wave is sinusoidal, then we have
(
AI ei(k1 zt) + AR ei(k1 zt) z<0
(14) f (z,t) = it
e F (z) z>0

where F (z) is given by 8.


WAVES IN A VISCOUS FLUID 3

The boundary conditions require the wave function and its derivative to
be continuous at z = 0, so we get

(15) AI + AR = AT
(16) ik1 (AI AR ) = AT = (AI + AR )

Solving for AR we get

+ ik1
(17) AR = AI
ik1
+ i (k1 k2 )
(18) = AI
i (k1 + k2 )

If AI is real (that is, the incident wave has zero phase: I = 0) then the
magnitude of the amplitude of the reflected wave is

s
2 + (k1 k2 )2
(19) |AR | = AI
2 + (k1 + k2 )2

We can convert 18 by multiplying top and bottom by the complex conju-


gate of the denominator:

( + i (k1 k2 )) ( + i (k1 + k2 ))
(20) AR = AI
2 + (k1 + k2 )2
2 + k22 k12 + 2i k1
(21) = AI
2 + (k1 + k2 )2

The phase of the reflected wave is

(AR )
(22) tan R =
(AR )
2 k1
(23) =
+ k22 k12
2

If = 0 (so there is no fluid surrounding string 2), then = 0 and


WAVES IN A VISCOUS FLUID 4

k22 k12
(24) AR = AI
(k1 + k2 )2
k1 k2
(25) = AI
k2 + k1
(26) tan R = 0

which agrees with Griffiths equation 9.29 for sinusoidal waves.


WAVES: POLARIZATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.8.
For the sinusoidal wave on a string, lets assume that the wave travels
in the +z direction and then define x and y axes in the usual way. We can
produce the wave by shaking the string in the xz plane, in which case there
is no component of the wave in the y direction, or by shaking the string in
the yz plane, in which case there is no motion in the x direction. Or we
could shake the string in some other plane intermediate between xz and xy,
as long as that plane contains the z axis.
If the x axis points upwards and the y axis points horizontally, then a wave
moving in the xz plane can be said to be vertically polarized, while a wave
moving in the yz plane is horizontally polarized. (These terms arent really
technical terms, but you get the idea.) Any transverse wave whose motion
is restricted to a plane is linearly polarized. We can represent this in vector
notation by writing

(1) f (z,t) = Aei(kzt) n

where n is a unit vector perpendicular to the z axis and parallel to the plane
of polarization. Thus for a wave moving in the xz plane, n = x and so on.
If is the angle between n and x, then in general

(2) n = cos x + sin y

and a wave linearly polarized in the n direction is given by

(3) f (z,t) = Aei(kzt) (cos x + sin y)


More generally, we can have a wave that is the vector sum of a vertically
and horizontally polarized wave:

(4) f (z,t) = Ax ei(kzt) cos x + Ay ei(kzt) sin y


1
WAVES: POLARIZATION 2

where the complex amplitudes are given by

(5) Ai = ai eii

with i = x, y. The real amplitudes ax and ay must be the same for both
components in order for the polarization to be in the n plane.
For a wave to be linearly polarized, the phases of the two component
waves must also be equal, so that x = y = . In that case, the components
of the wave at a fixed location z = z0 vary with time according to

(6) fx (z0 ,t) = aei(kz0 t+ ) cos x


(7) fy (z0 ,t) = aei(kz0 t+ ) sin y
Since it is only the real parts of these equations that are physically mean-
ingful, we see that the time dependence is the same in both cases and is
cos (kz0 t + ). Thus the horizontal and vertical components oscillate
in phase, so that the wave remains in the n plane.
Now suppose that y = /2 and x = 0 so that the two components are
out of phase. Then the real components of the wave are

(8) x = ( fx ) = a cos (kz0 t) cos


 
(9) y = ( fy ) = a cos kz0 t + sin
2
(10) = a sin (kz0 t) sin
In this case, the point on the string at position z = z0 follows a curve with
equation

x2 y2
(11) + =1
(a cos )2 (a sin )2
which is the equation of an ellipse. In this case, the wave is elliptically
polarized. If = 4 then we get a circle and the wave is circularly polarized.
To see which direction the string moves, try a couple of values of t at
z0 = 0. For t = 0, x (0, 0) = a cos , y (0, 0) = 0. Then for t = /2 we
have x (0, /2) = 0, y (0, /2) = a sin . Thus if increases counter-
clockwise as we look down the z axis towards the origin, the point is rotat-
ing counterclockwise. The shape of the string at any given instant of time is
a helix (either elliptical or circular, depending on ). We can generate this
by shaking the end of the string in an ellipse or circle.
WAVES: POLARIZATION 3

If we set y = /2 then

(12) x = ( fx ) = a cos (kz0 t) cos


 
(13) y = ( fy ) = a cos kz0 t sin
2
(14) = a sin (kz0 t) sin

and this time, a point on the string moves clockwise around the z axis.
ELECTROMAGNETIC WAVES IN VACUUM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.9.
Starting with Maxwells equations


(1) E =
0
(2) B = 0
B
(3) E =
t
E
(4) B = 0 J + 0 0
t

we can now investigate what happens if we have time-varying electric and


magnetic fields in vacuum. In that case, there is no charge or current so
= J = 0 and we get

(5) E = 0
(6) B = 0
B
(7) E =
t
E
(8) B = 0 0
t
We can transform these equations into separate equations for E and B by
taking the curl of the last two:

(9) ( E) = ( E) 2 E
B
(10) =
t
2E
(11) = 0 0 2
t
Since E = 0 in vacuum, we get
1
ELECTROMAGNETIC WAVES IN VACUUM 2

2E
(12) 2 E = 0 0
t 2
A similar calculation for B gives us

(13) ( B) = ( B) 2 B
E
(14) = 0 0
t
2
B
(15) = 0 0 2
t

so we get

2B
(16) 2 B = 0 0
t 2
The wave equation can be generalized to higher dimensions. In two di-
mensions, we can consider the force on a patch of membrane held under
tension (as in a drum), and the wave variable is the displacement of the
membrane from equilibrium. In three dimensions, we can consider the
change in some property of a 3-d substance. For example, we can think
of the change in density of a fluid such as water as a sound wave passes
through it. A proper derivation of the 3-d wave equation would take us a bit
far afield here, so well just quote the result. For a scalar field (that is, the
quantity that waves) f the 3-d wave equation is

1
(17) 2 f = t f
v2

where v is the speed of the wave through the substance.


Given the 3-d wave equation, we can see that each component of E and
B satisfies the wave equation, and that the speed of the wave is the same in
both cases, namely

1
(18) v=
0 0
Experimentally, it was found that v = c, the speed of light. This result
seems to me to be one of the most magical results in physics. It predicts
that electric and magnetic fields, once produced, sustain each other and
ELECTROMAGNETIC WAVES IN VACUUM 3

propagate as a wave. Not only that, it suggests (it doesnt really predict)
that light is itself an electromagnetic wave.
We can derive a few more properties of the electromagnetic wave by
applying Maxwells equations to solutions of the wave equations. Since
any solution of the wave equation can be expressed as a sum (or integral)
over sinusoidal functions (thats Fourier analysis), we can consider only
sinusoidal solutions from now on. Considering waves that consist of only a
single frequency that travel in the +z direction and have no dependence
on x or y, we can write the solutions as

(19) E = E0 ei(kzt)
(20) B = B0 ei(kzt)

where the tilde indicates were using complex notation, and that the physical
wave is the real part. The parameters E0 and B0 are constants under the
assumptions weve made here. Such a wave is called monochromatic (one
colour) because it contains only one frequency (and hence, for visible light,
only one colour) and plane because the wave is constant over any plane
perpendicular to the direction of propagation.
We can now apply Maxwells equations to these solutions. First, an ob-
servation about the complex notation. For the fields above, the real parts
depend on space and time through a term cos (kz t) and the imaginary
parts through a term sin (kz t). Maxwells equations involve only first
derivatives with respect to space and time, and these derivatives will convert
all cosines into sines and vice versa. Therefore, if the real part of E or B
satisfies Maxwells equations (as it does), then applying the same equations
to the imaginary parts just replaces all cosines by sines and thus the imagi-
nary parts must also be solutions. So its safe to apply Maxwells equations
to the full complex functions E and B.
In a vacuum, both E = 0 and B = 0 from which we get

(21) E = ikE0z ei(kzt) = 0


(22) B = ikB0z ei(kzt) = 0
Since this must be true for all z, we must have

(23) E0z = B0z = 0


That is, the wave has only x and y components, so it must be a transverse
wave: a wave that oscillates in a plane perpendicular to the direction of
propagation.
ELECTROMAGNETIC WAVES IN VACUUM 4

We can now apply E = tB , which gives


   
(24) E = ikE0y ei(kzt) x + ikE0x ei(kzt) y
   
i(kzt) i(kzt)
(25) = i B0x e x + i B0y e y

so

k
(26) B0x = E0y

k
(27) B0y = E0x

which can be written in vector form as

k 1
(28) B0 = z E0 = z E0
c
Therefore, not only are B and E transverse, they are also perpendicular
to each other.
Now suppose we want a monochromatic, plane wave that travels in some
arbitrary direction given by k. The value of the wave function at some point
r in 3-d space depends on the projection of r onto the direction of propaga-
tion, since for a plane wave, the wave function depends only on the distance
weve moved along this direction. This projection is r k. We can therefore
define the wave vector k = kk that points along the propagation direction
and replace the kz in the equations above by k r, so a monochromatic plane
wave travelling in direction k is then

(29) E = E0 ei(krt)
(30) B = B0 ei(krt)
For such a wave, the directions given by E0 and B0 are fixed, and its con-
ventional to take the direction of E0 as the polarization direction n. Since B
is perpendicular both to E and k its direction is given by k n so

(31) E = E0 ei(krt) n
(32) B = B0 ei(krt) k n
1
(33) = k E
c
ELECTROMAGNETIC WAVES IN VACUUM 5

Example 1. Suppose we have a monochromatic plane wave of (real) am-


plitude E0 , frequency and phase angle = 0. The wave is travelling in
the x direction and polarized in the z direction. We then have

(34) k kx
=
(35) n z=
(36) kr kx
=
(37) E0 E0=
(38) E E0 cos (kx t) z
=
E0
(39) B = cos (kx t) y
c
Example 2. Now we have the same wave, except that it is travelling in the
direction [1, 1, 1] and is polarized parallel to the xz plane. This time

k
(40) k = [1, 1, 1]
3

Since n is parallel to the xz plane and k is perpendicular to n, we have

(41) n = [nx , 0, ny ]
(42) k n = 0
1
(43) n = [1, 0, 1]
2

Thus

 
E0 k
(44) E = cos (x + y + z) t [1, 0, 1]
2 3
1
(45) B = k E
c  
E0 k
(46) = cos (x + y + z) t [1, 2, 1]
6c 3

or, since = ck
ELECTROMAGNETIC WAVES IN VACUUM 6

 
E0
(47) E = cos (x + y + z) t [1, 0, 1]
2 3c
1
(48) B = k E
c  
E0
(49) = cos (x + y + z) t [1, 2, 1]
6c 3c
P INGBACKS
Pingback: Electromagnetic waves: energy, momentum and light pressure
Pingback: Average of product of two waves
Pingback: Maxwell stress tensor and electromagnetic waves
Pingback: Electromagnetic waves in matter: normal reflection and trans-
mission
Pingback: Skin depth of electromagnetic waves in conductors
Pingback: Electromagnetic waves in conductors: phases and amplitudes
Pingback: Refraction and dispersion coefficients
Pingback: Wave guides: derivation of the wave equation
Pingback: Spherical electromagnetic wave
Pingback: Total internal reflection
Pingback: Potentials for an electromagnetic wave
Pingback: Fields and radiated power from an oscillating electric dipole
Pingback: Relativistic invariants involving electromagnetic fields
Pingback: Relativistic transformation of electromagnetic waves; the Doppler
effect
ELECTROMAGNETIC WAVES: ENERGY, MOMENTUM AND
LIGHT PRESSURE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.10.
Weve seen that Maxwells equations in a vacuum predict that an electro-
magnetic field can propagate as a wave with the speed of light. Weve also
seen that any region where both an electric and magnetic field exists results
in energy flowing through the region. The rate per unit area at which energy
crosses a surface is given by the Poynting vector

1
(1) S= EB
0
In an electromagnetic wave, E and B are perpendicular to each other,
and also perpendicular to the direction k of propagation of the wave. Thus
their cross product is parallel to k, indicating that an electromagnetic wave
carries energy.
Electromagnetic fields also carry momentum, with the momentum den-
sity given by

1
(2) p = 0 0 S = S
c2
For a monochromatic plane wave travelling in the z direction

(3) E = E0 cos (kz t + ) x


E0
(4) B = cos (kz t + ) y
c
E02
(5) S = cos2 (kz t + ) z
0 c
(6) = E02 c0 cos2 (kz t + ) z

The average of cos or sin squared over a complete cycle is 21 , so the


averages of energy and momentum density are
1
ELECTROMAGNETIC WAVES: ENERGY, MOMENTUM AND LIGHT PRESSURE 2

1 2
(7) hSi = E c0 z
2 0
1 2
(8) hpi = E 0 z
2c 0
The magnitude of hSi has the units of energy per unit area per unit time,
or power per unit area, and is known as the intensity of the radiation:

1
(9) I hSi = E02 c0
2
1 2 I
(10) hpi = E0 0 = 2
2c c
If an electromagnetic wave hits a surface, its momentum is either ab-
sorbed or reflected (or bits of both), so EM radiation actually exerts a pres-
sure on a surface. Pressure is force per unit area, which is momentum trans-
ferred per unit time per unit area. The quantity hpi is the momentum density
in the wave, so the amount of momentum that falls upon an area A in time
t is the volume of the wave that falls on the area times hpi. The wave is
travelling at speed c so this volume is Act and the momentum that falls on
the area is

(11) p = hpi Act


The average pressure is the average force per unit area, which in turn is
the momentum received by the area per unit area per unit time, so

p 1 I
(12) P= = hpi c = E02 0 =
At 2 c
This assumes a perfect absorber, so the radiation is just absorbed and not
reflected. A perfect reflector would experience a pressure twice as large,
since the momentum in the fields is not just stopped, it is reversed.
To get an idea of how big this pressure is, suppose were outside on a
sunny day at noon, when the solar radiation is at its maximum. At my
latitude (Scotland) this intensity never gets very far above 1000 Watts m2
(if youre interested in my weather data see here) but well use Griffithss
value of 1300 Watts m2 . If this sunlight strikes a perfect absorber then

1300
(13) P= = 4.33 106 N m2
3 108
For a perfect reflector, the pressure is just twice this, or 8.66106 N m2 .
ELECTROMAGNETIC WAVES: ENERGY, MOMENTUM AND LIGHT PRESSURE 3

For comparison, standard atmospheric pressure is 101325 N m2 so its


no surprise that the pressure exerted by sunlight isnt noticeable.
P INGBACKS
Pingback: Maxwell stress tensor and electromagnetic waves
Pingback: Electromagnetic waves in matter: reflection and transmission
coefficients
Pingback: Electromagnetic waves in conductors: energy density and in-
tensity
Pingback: Refraction and dispersion coefficients
Pingback: Relativistic transformation of electromagnetic waves; the Doppler
effect
Pingback: Solar sails
AVERAGE OF PRODUCT OF TWO WAVES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.11.
A common calculation that is required when analyzing any system that
varies with a sinusoidal period is a time average over one cycle. For exam-
ple, a monochromatic plane wave with amplitude A, direction k, frequency
and phase can be written as

(1) f = A cos (k r t + ) = Aei(krt)


(2) A = Aei

Now suppose we have two waves with the same direction and frequency,
but different amplitudes and phases. Then

(3) f = A cos (k r t + a )
(4) g = B cos (k r t + b )

The average of the product of these waves over a single cycle is then

2/
AB
(5) h f gi = cos (k r t + a ) cos (k r t + b ) dt
2 0

We can transform this integral by defining

(6) k r t
(7) d = dt

AB 2
(8) h f gi = cos ( + a ) cos ( + b ) d
2 0
Weve used the limits of 0 and 2 since any interval of 2 covers one
complete cycle of .
The two cosines have the same period and differ only in their phase, so
we will get the same result from the integral if we replace them by
1
AVERAGE OF PRODUCT OF TWO WAVES 2

(9)
cos ( + a ) cos ( + b ) cos cos ( + a b )
(10) = cos2 cos (a b ) cos sin sin (a b )
We now have

(11)
2 2
AB 2 AB
h f gi = cos (a b ) cos d sin (a b ) cos sin d
2 0 2 0
(12)
1
= AB cos (a b ) 0
2
(13)
1
= AB cos (a b )
2
(14)
1 1
= ( f g ) = ( f g)
2 2
Thus we can get the answer using complex notation without doing any
integrals.
This applies to vector products as well, since the components of vector
products are just products of scalar functions. For example, the time average
of the Poynting vector becomes, when the electric and magnetic fields are
written in complex notation:

1
E B

(15) hSi =
20
The electromagnetic energy density in the fields has a time average of
 
1 1
(16) huem i = 0 E E + B B
4 0
 
1 1
(17) = 0 E E + B B
4 0
 
1 1
(18) = E E + B B
40 c2
We dropped the in line 2 since the quantity in parentheses is automati-
cally real anyway, and in the last line we used
AVERAGE OF PRODUCT OF TWO WAVES 3

1
(19) 0 0 =
c2
P INGBACKS
Pingback: Wave guide: energy flows at the group velocity
MAXWELL STRESS TENSOR AND ELECTROMAGNETIC
WAVES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.12.

The Maxwell stress tensor T gives the components of momentum flux
density. That is, the component Ti j is the momentum component i per unit
area per unit time crossing a surface normal to the j direction. The formula
is

   
1 2 1 1 2
(1) Ti j 0 Ei E j i j E + Bi B j i j B
2 0 2
For a monochromatic plane wave with amplitude A, direction z, fre-
quency and phase polarized in the x direction, we have

(2) E = E0 cos (kz t + ) x


E0
(3) B = cos (kz t + ) y
c


All the off-diagonal elements of T are zero, and the diagonal elements

are, using c = 1/ 0 0 :

1 1
(4) Txx = 0 E02 cos2 (kz t + ) 2
E02 cos2 (kz t + )
2 20 c
1 1
(5) = 0 E02 cos2 (kz t + ) 0 E02 cos2 (kz t + )
2 2
(6) =0
1 1
(7) Tyy = 0 E02 cos2 (kz t + ) + E 2 cos2 (kz t + )
2 20 c2 0
(8) =0
1 1
(9) Tzz = 0 E02 cos2 (kz t + ) E 2 cos2 (kz t + )
2 20 c2 0
(10) = 0 E02 cos2 (kz t + )
1
MAXWELL STRESS TENSOR AND ELECTROMAGNETIC WAVES 2

Thus there is no momentum flux in the x or y directions, and the mo-


mentum flux density in the z direction is 0 E02 cos2 (kz t + ). Since the
average of cos2 x over a single cycle is 21 , the average momentum flux den-
sity is

1
(11) hTzz i = 0 E02
2
Since the wave moves with speed c, a unit area sweeps out a volume c in
unit time. Thus the average momentum density (that is, momentum per unit
volume) is

1 1
(12) hpi = hTzz i = 0 E02
c 2c

which agrees with our earlier result.


Also in the earlier post, we showed that rate per unit area of energy flow
is given by the Poynting vector and is

(13) S = 0 cE02 cos2 (kz t + ) z


(14) = cTzz z
The energy density is thus S/c = Tzz so the energy density is the same
as the momentum flux density.
ELECTROMAGNETIC WAVES IN MATTER: REFLECTION AND
TRANSMISSION COEFFICIENTS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.13.
Continuing with our study of electromagentic waves in matter, well
carry on with the system of an incident wave travelling in the +z direction
(so k = z) and polarized in the x direction (so n = x). Suppose the boundary
is the xy plane, with medium 1 on the left (z < 0) and medium 2 on the right
(z > 0). Then the reflected and transmitted (complex) amplitudes are

1
(1) E0R = E0
1+ I
2
(2) E0T = E0
1+ I

where E0I is the incident amplitude and

1 v1 1 n2
(3) =
2 v2 2 n1

with vi the speed of the wave in medium i and ni = c/vi the index of refrac-
tion.
The intensity of a wave in a vacuum is defined as the mean (over time) of
the magnitude of the Poynting vector:

1
(4) I hSi = E02 c0
2
If we follow through the derivation of I for a wave in matter, we see that
the only difference is that c is replaced by v and 0 by , so the intensity
becomes

1
(5) I = vE02
2
The reflection coefficient R is the ratio of reflected to incident intensity:
1
ELECTROMAGNETIC WAVES IN MATTER: REFLECTION AND TRANSMISSION COEFFICIENTS
2

1 2
2 1 v1 E0R
(6) R = 1 2
2 1 v1 E0I
1 2
 
(7) =
1+
The transmission coefficient T is the ratio of transmitted to incident in-
tensity:

1 2
2 2 v2 E0T
(8) T = 1 2
2 1 v1 E0I
42 v2
(9) =
1 v1 (1 + )2
4
(10) =
(1 + )2
where in the last line we used

1
(11) vi =
i i
1
(12) i =
i v2i
We can see that R + T = 1 which is just an expression of the conservation
of energy. The larger n2 is relative to n1 , the larger is which means that
R 1 and T 0.
The theory here is incomplete, as in practice the index of refraction de-
pends not only on the material but also on the wavelength of radiation. This
is largely a quantum phenomenon as it depends on the distances between
the atoms in the refracting medium, whereas the classical theory assumes
the medium is continuous.
P INGBACKS
Pingback: Fresnel equations for perpendicular polarization
Pingback: Reflection at a conducting surface: the physics of mirrors
Pingback: Transmission coefficient for a wave passing through 3 media
ELECTROMAGNETIC WAVES IN MATTER: NORMAL
REFLECTION AND TRANSMISSION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.14.
To see how electromagnetic waves propagate in matter, we can start with
Maxwells equations in matter:

(1) D = f
(2) B = 0
B
(3) E =
t
D
(4) H = Jf +
t
In the medium is linear, homogeneous and contains no free charge or
current, these equations reduce to

(5) E = 0
(6) B = 0
B
(7) E =
t
E
(8) B =
t
These equations are identical to those for a vacuum, except that the per-
meability 0 and permittivity 0 of free space have been replaced by their
corresponding values and in the medium. Thus electromagnetic waves
propagate the same way in a medium, so we can write them as

(9) E = E0 ei(krt) n
(10) B = B0 ei(krt) k n
1
(11) = k E
v

where
1
ELECTROMAGNETIC WAVES IN MATTER: NORMAL REFLECTION AND TRANSMISSION
2

1
(12) v=

is the speed of the wave in the medium. Since the permittivities and per-
meabilities of materials are almost always greater than those for free space,
the speed v < c in almost all cases, so light travels more slowly through
a medium than through a vacuum. The ratio of the speeds is the index of
refraction:

r
c
(13) n =
v 0 0
We can apply the boundary conditions at the interface between two media
to find out how light behaves when passing from one medium into another.

(14) 1 E1 2 E2 = 0
(15) B
1 B2 = 0
k k
(16) E1 E2 = 0
1 k 1 k
(17) B B = 0
1 1 2 2
Well start with an incident wave travelling in the +z direction (so k = z)
and polarized in the x direction (so n = x) and suppose the boundary is
the xy plane, with medium 1 on the left (z < 0) and medium 2 on the right
(z > 0). Then the incident wave is

(18) EI = E0I ei(k1 zt) x


(19) BI = B0I ei(k1 zt) z x
1
(20) = E0 ei(k1 zt) y
v1 I
Well have a reflected wave (k = z) and a transmitted wave (k = z), so

(21) ER = E0R ei(k1 zt) (cos R x + sin R y)


(22) BR = B0R ei(k1 zt) z (cos R x + sin R y)
1
(23) = E0 ei(k1 zt) (sin R x cos R y)
v1 R
ELECTROMAGNETIC WAVES IN MATTER: NORMAL REFLECTION AND TRANSMISSION
3

(24) ET = E0T ei(k2 zt) (cos T x + sin T y)


(25) BT = B0T ei(k2 zt) z (cos T x + sin T y)
1
(26) = E0T ei(k2 zt) ( sin T x + cos T y)
v2

where R and T are the angles of polarization for the reflected and trans-
mitted waves.
Since there are no components of the fields perpendicular to the boundary
at z = 0, 14 and 15 tell us nothing. Applying 16 to the x and y components,
we have

(27) E0I x + E0R (cos R x + sin R y) = E0T (cos T x + sin T y)


(28) E0I + E0R cos R = E0T cos T
(29) E0R sin R = E0T sin T
From 17 we have

1 1
(30) E0R sin R = E0 sin T
1 v1 2 v2 T
1  1
(31) E0I E0R cos R = E0 cos T
1 v1 2 v2 T
Substituting from 29 into 30 we have

1 1
(32) E0T sin T = E0 sin T
1 v1 2 v2 T
The only way this equation can be satisfied is if sin T = 0 so from 29 we
conclude that sin R = 0 as well. Thus T and R are either 0 or , so the
cosines are 1. Thus the plane of polarization is the same for the reflected
and transmitted waves as for the incident wave.
If we define

1 v1
(33)
2 v2

we can write 31 as

(34) E0I E0R cos R = E0T cos T


Adding this to 28 we get
ELECTROMAGNETIC WAVES IN MATTER: NORMAL REFLECTION AND TRANSMISSION
4

1+
(35) E0I = E0T cos T
2
2
(36) E0T = E0
1+ I
Subtracting 34 from 28 we get

1 1
(37) E0R = E0T = E0
2 1+ I
Thus the reflected and transmitted waves are either in phase or half a
cycle out of phase with the incident wave.
P INGBACKS
Pingback: Electromagnetic waves in matter: reflection and transmission
coefficients
Pingback: Three laws of geometrical optics
Pingback: Fresnel equations for perpendicular polarization
Pingback: Reflection at a conducting surface: the physics of mirrors
Pingback: Refraction and dispersion coefficients
Pingback: Transmission coefficient for a wave passing through 3 media
Pingback: Microwave shielding for perfect transmission
THREE LAWS OF GEOMETRICAL OPTICS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.15.
We got the equations for reflected and transmitted waves when an electro-
magnetic wave is incident on a boundary head-on. Using similar (although
somewhat more involved) methods, we can solve the general case where a
wave is incident on a boundary at any angle.
Suppose the wave vector of the incident wave is kI and the boundary is
located on the xy plane. The incident wave is

(1) EI = E0I ei(kI rt)


1
(2) BI = kI EI
v1

where the direction of polarization is given by the direction of E0I . At this


stage, we dont know the direction of either the reflected or transmitted
waves, so we can just write them as

(3) ER = E0R ei(kR rt)


1
(4) BR = kR ER
v1
(5) ET = E0T ei(kT rt)
1
(6) BT = kT ET
v2
The speeds vi of the waves are determined by the medium and not by the
direction of the waves and since the frequency is the same for all three
waves and = kv, we get

(7) kI v1 = kR v1
(8) kI = kR
(9) kI v1 = kT v2
v1 n2
(10) kT = kI = kI
v2 n1
1
THREE LAWS OF GEOMETRICAL OPTICS 2

where ni is the index of refraction.


This gives us the magnitudes of the wave vectors, but not the directions.
To get those, we need to apply the boundary conditions for zero surface
charge:

(11) 1 E1 2 E2 = 0
(12) B
1 B2 = 0
k k
(13) E1 E2 = 0
1 k 1 k
(14) B B = 0
1 1 2 2

where the subscript 1 refers to fields in medium 1 (z < 0) and 2 refers to


medium 2 (z > 0). That is

(15) E1 = EI + ER
(16) E2 = ET

and similarly for B. Thus all these boundary conditions have the form

(17) AI ei(kI rt) + AR ei(kR rt) = AT ei(kT rt)

for coefficients Ai that vary, depending on the boundary condition being im-
posed, but that dont depend on r or t. The space and time dependencies are
entirely within the complex exponentials. Since weve taken the boundary
to be the xy plane, z = 0 in these equations, so they become

(18) AI ei(kIx x+kIy yt ) + AR ei(kRx x+kRy yt ) = AT ei(kT x x+kTy yt )


(19) AI ei(kIx x+kIy y) + AR ei(kRx x+kRy y) = AT ei(kT x x+kTy y)
These equations must be valid over the entire xy plane, so they are valid
for x = 0 or y = 0:

(20) AI eikIx x + AR eikRx x = AT eikT x x


To see what this implies, consider the general case

(21) Aeiax + Beibx = Ceicx


THREE LAWS OF GEOMETRICAL OPTICS 3

for all x. That is, both sides are the same function of x, so they must have
the same Taylor expansion. Starting at point x and expanding to get the
function at x + x we get up to first order in x:

(22)
Aeiax (1 + iax) + Beibx (1 + ibx) + O x2 = Ceicx (1 + icx) + O x2
 

Using 21 and cancelling common factors we get

(23) aAeiax + bBeibx = cCeicx


 
(24) = c Aeiax + Beibx
(25) A (a c) eiax + B (b c) eibx = 0
This last equation must be true for all x so we must have

(26) a=b=c
Going back to 19, we therefore must have

(27) kIx = kRx = kT x


(28) kIy = kRy = kTy
If we choose axes so that kI is in the xz plane, then kR and kT must also
lie in the same plane. This gives the first law of geometrical optics:
The wave vectors of the incident, reflected and transmitted waves all lie
in the same plane, and this plane also contains the normal to the boundary.
If the angle of incidence is I (that is, the angle between kI and the normal
to the xy plane) with R and T the corresponding angles for reflection and
transmission, then from 8 we have

(29) kI sin I = kR sin R


(30) = kI sin R
(31) sin I = sin R
(32) I = R
This is the familiar condition for specular reflection: the angle of inci-
dence is equal to the angle of reflection. This is the second law of geomet-
rical optics.
Finally, for the transmitted wave, from 10 we have
THREE LAWS OF GEOMETRICAL OPTICS 4

(33) kI sin I = kT sin T


n2
(34) = kI sin T
n1
sin I n2
(35) =
sin T n1
This is Snells law, or the law of refraction, which is the third law of
geometrical optics.
We havent actually applied the boundary conditions yet, apart from 20,
but well look at that next to get an idea of the reflection and transmission
coefficients.
P INGBACKS
Pingback: Fresnel equations for perpendicular polarization
Pingback: Total internal reflection
Pingback: Lensmakers equation
FRESNEL EQUATIONS FOR PERPENDICULAR
POLARIZATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.16.
Continuing our study of electromagnetic waves incident on a surface at an
oblique angle well now use the boundary conditions to derive the reflection
and transmission coefficients. The boundary conditions are

(1) 1 E1 2 E2 = 0
(2) B
1 B2 = 0
k k
(3) E1 E2 = 0
1 k 1 k
(4) B B = 0
1 1 2 2

where the subscript 1 refers to fields in medium 1 (z < 0) and 2 refers to


medium 2 (z > 0). That is

(5) E1 = EI + ER
(6) E2 = ET

and similarly for B. As we saw last time, the space-time dependence cancels
out of the boundary conditions, we can replace all fields by their (complex)
amplitudes so we get
 
(7) 1 E0I + E0R = 2 E0T
(8) B
0I + B0R = B0T
k k k
(9) E0I + E0R = E0T
1 
k k
 1 k
(10) B0I + B0R = B
1 2 0T
There are actually two cases to consider: polarization parallel to the inci-
dent plane (that is, E in the xz plane in our example) or perpendicular to the
1
FRESNEL EQUATIONS FOR PERPENDICULAR POLARIZATION 2

incident plane (so E is polarized along the y axis). Griffiths does the parallel
case in his section 9.3.3 so well look at the perpendicular case here. (Its
important to be clear about what is perpendicular or parallel to what. In the
four boundary conditions above, the and k symbols mean perpendicular
and parallel to the boundary (that is, the xy plane) not the incident plane.
The polarization were considering is perpendicular to the incident plane.)
The incident wave travels along wave vector kI at an angle of I to the
normal to the xy plane, the reflected wave travels along kR also at an an-
gle of I , and the transmitted wave travels along kT at angle T such that,
according to Snells law

sin I n2
(11) =
sin T n1

Condition 7 tells us nothing since E is in the y direction, so has no com-


ponent perpendicular to the xy plane. To use the other conditions, we need
to work out the components of E and B. We know E is in the y direction
so thats easy. The direction of B is given by k E. Consider BI . Here kI
points towards the xy plane (from the left) at an angle I to the normal to
this plane. The cross product k E therefore lies in the xz plane and points
to the lower right at an angle 2 I to the normal, so the components of BI
are (to keep the notation simple in what follows well drop the 0 subscript
and tilde for the amplitudes):

  1
(12) BIz = BI cos I = EI sin I
2 v1
  1
(13) BIx = BI sin I = EI cos I
2 v1

BIx is negative since BI points towards negative x and positive z.


Assuming the reflected wave still has polarization in the +y direction,
the direction of BR is now kR ER and kR points away (towards the left)
from the xy plane at angle R = I so BR points to the upper right and has
components

 1 
(14) BRz = BR cos I = ER sin I
2 v1
  1
(15) BRx = BR sin I = ER cos I
2 v1
FRESNEL EQUATIONS FOR PERPENDICULAR POLARIZATION 3

Finally, the transmitted wave has direction kT which points away (to-
wards the right) from the xy plane, so BT = v12 kT ET points to the lower
right and has components

 1
(16) BTz = BT cos T = ET sin T
2 v2
  1
(17) BTx = BT sin T = ET cos T
2 v2

Were now ready to apply the boundary conditions. First, we use 8, which
applies to the z components of B so we have

1 1 1
(18) EI sin I + ER sin I = ET sin T
v1 v1 v2
v1 sin T
(19) EI + ER = ET
v2 sin I
n1 v1
(20) = ET
n2 v2
(21) = ET

Condition 9 just gives us the same relation, so we dont learn anything


new from it. Condition 10 applies to Bx only since B has no y component.

 
1 1 1 1
(22) EI cos I + ER cos I = ET cos T
1 v1 v1 2 v2
1 v1 cos T
(23) EI ER = ET
2 v2 cos I
(24) = ET

where

cos T
(25)
cos I
1 v1 1 n2
(26) =
2 v2 2 n1

Solving these two equations gives


FRESNEL EQUATIONS FOR PERPENDICULAR POLARIZATION 4

1
(27) ER = EI
1 +
2
(28) ET = EI
1 +
These are Fresnels equations for perpendicular polarization. For normal
incidence, I = T = 0, = 1 and they reduce to the equations we got in
that case. Plots of ER /EI (red) and ET /EI (blue) for n2 /n1 = 1.5 are as
shown:

The negative values for ER /EI indicate that the reflected wave is out
of phase with the incident wave. For I = 0 (normal incidence) 80% of the
amplitude is transmitted, dropping to zero when I = /2 (incident wave is
parallel to the surface).
The Fresnel equations for parallel polarization (see Griffiths) turn out to
be


(29) ER = EI
+
2
(30) ET = EI
+
FRESNEL EQUATIONS FOR PERPENDICULAR POLARIZATION 5

We can see that if = , ER = 0 and there is no reflected wave. This


occurs at Brewsters angle B , given by

12
(31) sin2 B =
(n1 /n2 )2 2
For perpendicular polarization, there is no reflection if we can find an
angle B such that = 1/ .

1 sin2 T
(32) 2 =
1 sin2 I
1 (n1 /n2 )2 sin2 I
(33) =
1 sin2 I
1
(34) =
2
1 1/ 2
(35) sin2 B =
(n1 /n2 )2 1/ 2
2 1
(36) =
2 (n1 /n2 )2 1
In practice, the permeabilities of media are approximately equal, so that
1 2 and n2 /n1 from 26. In this case, the expression for sin2 B
blows up so there is no solution, and thus no Brewster angle for perpen-
dicular polarization (unless = 1 which occurs only if n1 = n2 so there is
effectively no boundary).
The reflection and transmission coefficients are

1 2
2 1 v1 ER
(37) R = 1 2
2 1 v1 EI
(1 )2
(38) =
(1 + )2
1 2
2 2 v2 ET
(39) T = 1 2
2 1 v1 EI
4
(40) =
(1 + )2

and it can be seen that R + T = 1.


FRESNEL EQUATIONS FOR PERPENDICULAR POLARIZATION 6

P INGBACKS
Pingback: Optical properties of diamond
Pingback: Total internal reflection
OPTICAL PROPERTIES OF DIAMOND

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.17.
Weve seen how to derive the properties of reflected and transmitted
waves in the case where the wave is polarized perpendicular to the plane
of incidence. The derivation in the case of parallel polarization is very sim-
ilar and is given in Griffiths 9.3.3. Here well have a look at some of these
properties at an interface between air and diamond.
The Fresnel equations for parallel polarization, giving the reflected and
transmitted amplitudes in terms of the incident amplitude, turn out to be


(1) ER = EI
+
2
(2) ET = EI
+

where the angle of incidence is I , the angle of transmission is T and

cos T
(3)
cos I
1 n2
(4)
2 n1

Taking 1 = 2 = 0 and using diamonds index of refraction n2 = 2.42,


we can draw plots of ER /EI and ET /EI (red for reflected and blue for trans-
mitted):
1
OPTICAL PROPERTIES OF DIAMOND 2

At normal incidence I = T = 0 and

ER
(5) = 0.415
EI
ET
(6) = 0.585
EI
The negative value for the reflected amplitude indicates that the wave is
out of phase with the incident wave.
The reflected and transmitted amplitudes are equal where the curves cross,
which occurs at an angle obtained from solving ER = ET :

(7) R=T = 1.362 rad = 78.06


We can see that if = , ER = 0 and there is no reflected wave. This
occurs at Brewsters angle B , given by

12
(8) sin2 B =
(n1 /n2 )2 2
For the air-diamond interface, we get
OPTICAL PROPERTIES OF DIAMOND 3

(9) B = 1.179 rad = 67.55


DECAY TIME FOR FREE CHARGE IN A CONDUCTOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.18a.
When we looked at conductors in electrostatics, we saw that any free
charge in a conductor must reside on the surface. In electrodynamics, how-
ever, we can place some free charge inside a conductor and then watch it
(figuratively) move to the surface. How long does this migration take?
We start with Maxwells equations within matter:

(1) D = f
(2) B = 0
B
(3) E =
t
D
(4) H = Jf +
t
For linear media, D = E and H = B/, so these equations become

1
(5) E = f

(6) B = 0
B
(7) E =
t
E
(8) B = J f +
t
Within a conductor with conductivity (not surface charge density!)
Ohms law can be written as

(9) Jf = E

so Maxwells equations become


1
DECAY TIME FOR FREE CHARGE IN A CONDUCTOR 2

1
(10) E = f

(11) B = 0
B
(12) E =
t
E
(13) B = E +
t
The conservation of charge (continuity condition) says

f
(14) Jf =
t

so from 9 and 10 we get

f
(15) = E = f
t
For a fixed location within the conductor we can integrate this with re-
spect to time to get

(16) f (t) = f (0) et/


That is, the free charge density decays exponentially with a characteristic
time = / .
Example. For glass, the conductivity is around 1011 S m1 (S stands for
Siemens, which is the SI unit of conductance, where 1 S = 1 kg1 m2 s3 A2 )
and the permittivity is 4.70 = 4.16 1011 m3 kg1 s4 A2 so the charge
would move to the surface in a time of about

(17) = 4.16 s
Some forms of glass have much smaller conductivities, so the migration
time would be correspondingly larger.
P INGBACKS
Pingback: Skin depth of electromagnetic waves in conductors
SKIN DEPTH OF ELECTROMAGNETIC WAVES IN
CONDUCTORS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.18b-c.
Weve seen that any free charge within a conductor migrates to the sur-
face with a characteristic time that depends on the conductors conductance
and permittivity. Once that transient effect subsides, we can take the free
charge density to be zero: f = 0. This doesnt mean that the free current
density J f is zero, though. We can still have a current in an electrically
neutral conductor caused by electrons moving relative to stationary atomic
nuclei.
In linear media, Maxwells equations are

1
(1) E = f

(2) B = 0
B
(3) E =
t
E
(4) B = J f +
t
Within a conductor with conductivity (not surface charge density!)
Ohms law can be written as

(5) Jf = E

so with f = 0, Maxwells equations become

(6) E = 0
(7) B = 0
B
(8) E =
t
E
(9) B = E +
t
1
SKIN DEPTH OF ELECTROMAGNETIC WAVES IN CONDUCTORS 2

We can take the curl of the last two equations in the same way as when
we derived the wave equation in a vacuum.

(10) ( E) = ( E) 2 E
B
(11) =
t
2
E E
(12) = 2
t t

Since E = 0 we get

2E E
(13) 2 E = 2
+
t t

A similar calculation for B gives

2B B
(14) 2 B = 2
+
t t

We thus get the wave equation modified by the addition of an extra first-
derivative term. Conveniently, these equations have a similar solution to the
ordinary wave equation. For a plane wave travelling in the z direction

(15) E (z,t) = E0 ei(kzt )


(16) B (z,t) = B ei(kzt )
0

Substituting 15 into 13 we get, after cancelling terms

(17) k2 = 2 + i

The fact that the wave vector k is complex means that the resulting wave
has both an oscillatory and an exponentially decaying factor. Finding the
square root of this using Maple gives
SKIN DEPTH OF ELECTROMAGNETIC WAVES IN CONDUCTORS 3

(18)
r q r q
1 1
k = 2 4 2 2 + 2 2 2 + 2 2 + i 2 4 2 2 + 2 2 2 2 2
2 2
(19)

sr sr
 2  2
= 1+ +1+i 1+ 1
2 2
(20)
k + i
The wave is then given by

(21) E (z,t) = E0 ez ei(kzt)


(22) B (z,t) = B0 ez ei(kzt)
If an electromagnetic wave hits a conductor (starting in air or vacuum,
say), the wave will attenuate as it penetrates the conductor with a character-
istic distance, called the skin depth d of

" r !#1/2
1 1  2
(23) d= = 1+ 1
2
The skin depth depends not only on the conductivity and permittivity, but
also on the frequency of the incident radiation.
Example 1. The skin depth for good conductors such as metals is very
small for a wide range of frequencies. For silver, the conductivity is =
6.30 107 S m1 . To get the permittivity of silver, we can return to its
definition:

(24) = 0 (1 + e )

where the electric susceptibility e is defined by the amount of polarization


in the material that is produced by an applied electric field:

(25) P = 0 e E
For a perfect conductor, no polarization is produced by any amount of
electric field, so e = 0 and = 0 . Thus its a reasonable approximation for
SKIN DEPTH OF ELECTROMAGNETIC WAVES IN CONDUCTORS 4

a good conductor like silver to take 0 . Most materials also have a per-
meability 0 . At a microwave frequency of 1010 Hz = 2 1010 s1


(26) = 1.13 108
0

so to a good approximation
s s
1 2 2
(27) d= = = 6.34 107 m
0 0

The real part of k gives the oscillatory part of the wave, so the wavelength
is

2
(28) =
k

and the wave speed is


(29) v = = =
2 k
The index of refraction is the ratio of c to v as usual:

c ck
(30) n= =
v
Example 2. For copper, = 5.96 107 S m1 . For radio waves with a
frequency of 1 MHz = 2 106 s1 we have


(31) = 1.07 1012
0
r
0
(32) k
2
(33) = 1.533 104 m1
2
(34) = = 4.1 104 m
k
The wave speed is


(35) v= = 410 m s1
k
This is obviously very slow for electromagnetic radiation.
SKIN DEPTH OF ELECTROMAGNETIC WAVES IN CONDUCTORS 5

In air or vacuum, v = c = 3 108 m s1 and = v/ = 300 m which is


closer to what wed expect for radio waves.
P INGBACKS
Pingback: Skin depth of water and metals
Pingback: Electromagnetic waves in conductors: phases and amplitudes
Pingback: Reflection at a conducting surface: the physics of mirrors
SKIN DEPTH OF WATER AND METALS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.19a-b.
Electromagnetic waves in a conductor (where there is free current but no
free charge) can be written as

(1) E (z,t) = E0 ei(kzt )


(2) B (z,t) = B ei(kzt )
0

where the wave vector is complex:

(3)

sr sr
 2  2
k = 1+ +1+i 1+ 1 k + i
2 2
For a poor conductor, the conductivity is small, so for large enough
frequencies  and we can approximate by

r
1  2
(4) 1+ 1
2 2
r

(5) =
2
Since the imaginary part of k governs the attenuation of the wave as it
penetrates the material, the skin depth for a poor conductor is
r
1 2
(6) d= =

For pure (deionized) water = 5.5 106 S m1 and = 80.10 (at
20 C) (we can take 0 ) so the skin depth of water is

(7) d = 8635 m
Because the skin depth is so large, water is transparent.
1
SKIN DEPTH OF WATER AND METALS 2

For a good conductor,  and we can approximate


r r

(8) = k
2 2

so the skin depth is


s
2 1 2
(9) d= =
k

where is the wavelength within the material. For a typical metal,


107 S m1 and 0 so the skin depth at visible frequencies 1015 s1
is

(10) d 1.26 108 m


With a skin depth this small, even a thin film of metal is effectively im-
pervious to any penetration by visible light.
P INGBACKS
Pingback: Reflection at a conducting surface: the physics of mirrors
ELECTROMAGNETIC WAVES IN CONDUCTORS: PHASES
AND AMPLITUDES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.19c.
Electromagnetic waves in a conductor(where there is free current but no
free charge) can be written as

(1) E (z,t) = E0 ei(kzt )


(2) B (z,t) = B ei(kzt )
0

where the wave vector is complex:

(3)

sr sr
 2  2
k = 1+ +1+i 1+ 1 k + i
2 2

so

(4) E (z,t) = E0 ez ei(kzt)


(5) B (z,t) = B0 ez ei(kzt)
By applying Maxwells equations in a conductorwe can get a few more
properties of these waves. The equations are

(6) E = 0
(7) B = 0
B
(8) E =
t
E
(9) B = E +
t
Using the same techniques as in analyzing waves in vacuum. Both E =
0 and B = 0 from which we get
1
ELECTROMAGNETIC WAVES IN CONDUCTORS: PHASES AND AMPLITUDES 2

(10) E = (ik ) E0z ei(kzt) = 0


(11) B = (ik ) B0z ei(kzt) = 0
Since this must be true for all z, we must have

(12) E0z = B0z = 0


That is, the wave has only x and y components, so it must be a transverse
wave: a wave that oscillates in a plane perpendicular to the direction of
propagation. If we orient the axes so that E is polarized in the x direction
then

(13) E (z,t) = E0 ez ei(kzt) x


Applying 8 to this gives

(14) E = i (k + i) E0 ez ei(kzt) y
(15) = ikE0 ez ei(kzt) y
B
(16) =
t
(17) = i B0 ez ei(kzt)
k
(18) B (z,t) = E0 ez ei(kzt) y

As in vacuum, E and B are perpendicular and transverse to the direction
of propagation. Unlike in the vacuum, however, the two components of the
wave may not be in phase, due to the presence of the complex variable k in
the equation for B. If we write k in modulus-phase form we have

(19) k = Kei

where
p
(20) K = k2 + 2
s r
 2
(21) = 1+


(22) = arctan
k
ELECTROMAGNETIC WAVES IN CONDUCTORS: PHASES AND AMPLITUDES 3

Then the complex amplitudes of the two components can be written as

(23) E0 = E0 eiE
K
(24) B0 = E0 ei(E + )

and the ratio of the real amplitudes is


s r
B0 K  2
(25) = = 1+
E0
Example. For a good conductor,  so from 3 k so from 22 the
phase difference between B and E is /4. The ratio of amplitudes is
r
B0
(26) =
E0
For a typical good conductor 107 S m1 and 0 and at visible
frequencies 1015 s1 so

B0
(27) = 1.12 107 s m1
E0
P INGBACKS
Pingback: Electromagnetic waves in conductors: energy density and in-
tensity
ELECTROMAGNETIC WAVES IN CONDUCTORS: ENERGY
DENSITY AND INTENSITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.20.
We can write the electromagnetic wave inside a conductor as (if we orient
the axes so that E is polarized in the x direction)

(1) E (z,t) = E0 ez ei(kzt) x


(2) = E0 ez ei(kzt+E ) x
k
(3) B (z,t) = E0 ez ei(kzt) y

s r  2
(4) = 1+ E0 ez ei(kzt+E + ) y

where

(5)

sr sr
 2  2
k = 1+ + 1+i 1+ 1 k +i Kei
2 2
The actual fields are the real parts of these equations, so

(6) E (z,t) = E0 ez cos (kz t + E ) x


s r
 2
(7) B (z,t) = 1 + E0 ez cos (kz t + E + ) y

The energy density in the wave is
 
1 2 1 2
(8) u= E + B
2
Taking the time average (over one cycle) of this we have (since the aver-
age of cos2 t over one cycle = 2/ is 21 ):
1
ELECTROMAGNETIC WAVES IN CONDUCTORS: ENERGY DENSITY AND INTENSITY2

r !
E 2 e2z  2
(9) u= 0 + 1+
4
For a good conductor,  so

E02 e2z  
(10) u +
4
E02 e2z
(11)
4
From 10, we see that the magnetic contribution ( /) is much larger
than the electric contribution () for a good conductor.
We can express this in terms of the wave vector k by using 5 for a good
conductor.
r

(12) k
2
r

(13) =
2
2k 2
(14) =

E02 e2z k2
(15) u
2 2
The intensity is the energy crossing a unit area in unit time, which is the
energy density times the volume crossing a unit area per unit time, which is

(16) I = uv

where v is the speed of the wave, which is /k so

E02 e2z k
(17) I=
2
REFLECTION AT A CONDUCTING SURFACE: THE PHYSICS
OF MIRRORS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.21.
We can analyze reflection of an electromagnetic wave at a nonconductor-
conductor interface in a similar way to that used for a nonconductor-nonconductor
interface. Well look only at the case of normal incidence here.
As before, we start with the boundary conditions in linear media derived
from Maxwells equations:

(1) 1 E1 2 E2 = f
(2) B
1 B2 = 0
k k
(3) E1 E2 = 0
1 k 1 k
(4) B B = K f n
1 1 2 2
Well take medium 1 as the nonconductor (air, say) and medium 2 as the
conductor. Were allowing for the presence of free surface charge density
f and free current density K f at the boundary.
If were dealing with a conductor that obeys Ohms law, the volume
current density is proportional to the electric field

(5) Jf = E

where here is the conductivity, not a charge density. Recall that J f is the
amount of current flowing through a unit area in the conductor. If we had a
surface current density K f , this current flows along the boundary as a sheet
of moving charge with infinitesimal thickness, so that the cross-sectional
area occupied by K f is essentially zero, making the volume charge density
infinite. For a finite conductivity it would take an infinite electric field to
produce this surface current, so we can safely assume that K f = 0 in what
follows.
The incident and reflected waves are both in medium 1, so if we polarize
the wave in the x direction, we have for the incident wave:
1
REFLECTION AT A CONDUCTING SURFACE: THE PHYSICS OF MIRRORS 2

(6) EI = E0I ei(k1 zt) x


1
(7) BI = E0 ei(k1 zt) y
v1 I

where v1 is the speed of the wave in medium 1.


The reflected wave is travelling in the z direction and has equations

(8) ER = E0R ei(k1 zt) x


1
(9) BR = E0R ei(k1 zt) y
v1
The transmitted wave is inside the conductor, so its equations can be
written as

(10) ET (z,t) = E0T ei(k2 zt ) x


k2
(11) BT (z,t) = E0 ei(k2 zt ) y
T
where the wave vector is complex:

(12)

sr sr
 2  2
k = 1+ +1+i 1+ 1 k + i
2 2
We can now apply the boundary conditions. Equation 1 tells us that f =
0 since there is no perpendicular component of E (remember the wave is
transverse). Equation 2 tells us nothing (0 = 0). From 3, assuming that the
boundary is at z = 0, we get, since all components of E are in the x direction:

(13) E0I + E0R = E0T


Finally, from 4 we get, since all components of B are in the y direction
and K f = 0:

1  k2
(14) E0I E0R = E0
1 v1 2 T

which we can rewrite as


REFLECTION AT A CONDUCTING SURFACE: THE PHYSICS OF MIRRORS 3

1 v1
(15) E0I E0R = k2 E0T
2
(16) E0T
Adding 13 and 16 we get

1 
(17) E0I = 1 + E0T
2
2
(18) E0T = E0I
1 +
Subtracting 13 and 16 we get

1 
(19) E0R = 1 E0T
2
1
(20) = E0I
1 +
These are deceptively simple equations, since everything with a tilde on
it is a complex number. To get the actual amplitudes and phases we need to
extract the real and imaginary parts.
Example. To put some numbers into these equations, lets consider an air-
silver interface. For a good conductor such as silver,  and in 12
r

(21) k
2
In air, v1 c and we can take 1 = 2 = 0 so from 16
r
0
(22) c (1 + i)
2
For silver = 6 107 S m1 and at an optical frequency of = 4
1015 s1 we get

(23) = 29.1 (1 + i)
1 28.1 29.1i 30.1 29.1i
(24) =
1 + 30.1 + 29.1i 30.1 29.1i
1
(25) = (1693 58.2i)
1753
REFLECTION AT A CONDUCTING SURFACE: THE PHYSICS OF MIRRORS 4

To get the reflection coefficient we can write the complex amplitudes in


modulus-phase form as

(26) E0R = E0R eiR


(27) E0I = E0I eiI
The intensity of a wave is the average over one cycle of the magnitude
of the Poynting vector, so the fact that the incident and reflected waves may
have different phases doesnt matter (since they have the same frequency).
This means that

1 2
2 c0 E0R
(28) R = 1 2
2 c0 E0I
|E0R |2
(29)
|E0I |2
16932 + 58.22
(30) =
17532
(31) = 0.93
Silver reflects 93% of the incident light, so it makes a good mirror.
PHASE VELOCITY AND GROUP VELOCITY IN A WAVE
PACKET

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.22.
A monochromatic wave has velocity called the phase velocity given by


(1) vp =
k

where is the frequency and k = 2/ is the wave number. However,


if we have a compound wave that is composed of individual waves with a
range of frequencies, each individual wave has a velocity given by 1, but
the amplitudes of the waves add up to produce a wave packet which has a
velocity all its own. This velocity is called the group velocity and is usually
different from the individual phase velocities of the waves that make up the
packet.
This effect arises from the fact that typically the frequency is a function
of the wave number: = (k). Suppose we have a wave packet made up
of a range of individual waves. We can write this as

(2) (x,t) = A (k) ei(kxt) dk

where A (k) is a function giving the contribution to the packet of waves with
wave number k. Now suppose that most of these waves have values of k
that are close to some value k0 . In that case, we can expand (k) and keep
only the first order term:

(3) (k0 + k) = 0 + k00

where

d
(4) 00
dk k0
Plugging this into 2 we get
1
PHASE VELOCITY AND GROUP VELOCITY IN A WAVE PACKET 2


0
(5) (x,t) = e i(k0 x0 t)
A (k) ei(kxk0t ) d (k)

We can see that the wave packet is composed of a monochromatic wave


represented by the exponential outside the integral modulated by the inte-
gral factor. The speed of the monochromatic wave is just the phase velocity
of the main wave:

0
(6) vp =
k0
The velocity of the modulation is now a constant given by

k 00

d
(7) vg = =
k dk k0
Its this latter velocity that is the group velocity. This derivation relies
on the waves making up the packet all having wave numbers (and hence
wavelengths) lying close to each other.
Example 1. A wave travelling on the surface of water has a phase velocity
that is proportional to the square root of its wavelength (provided is less
than the depth of the water). That is,


(8) vp = =A
kr
2
(9) = A
k

(10) (k) = A 2k
d
(11) vg =
dk r
1 2
(12) = A
2 k
1
(13) = vp
2
Thus the phase velocity of deep water waves is twice the group velocity.
Example 2. Weve seen that a free particle can be represented in quantum
mechanics by a superposition of waves, each of which has the form

(14) (x,t) = Aei(pxEt)/h


PHASE VELOCITY AND GROUP VELOCITY IN A WAVE PACKET 3

where p is the momentum and E is the energy, given by

p2
(15) E=
2m
In terms of frequency and wave number, we have

p
(16) k =
h
p2
(17) =
2hm
hk2
(18) =
2m
d hk p
(19) vg = = =
dk m m
p vg
(20) vp = k = =
2m 2
So in this case, the phase velocity is half the group velocity. Classically
a particles velocity is given by v = p/m so it is the quantum group velocity
that corresponds to classical velocity.
P INGBACKS
Pingback: Group velocity of electromagnetic waves in a dispersive medium
FREQUENCY DEPENDENCE OF ELECTRIC PERMITTIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.23a.
Experimentally, it is known that the permittivity of a material when an
electromagnetic wave passes through it depends on the frequency of the
wave. To develop a (relatively crude) theory of how this comes about, its
worth recalling the definition of permittivity , which arises from the ability
of an external electric field E to polarize a dielectric, producing a polariza-
tion density P:

(1) P = 0 e E
e is the electric susceptibility, and the permittivity is defined in terms of
it by

(2) = 0 (1 + e )
Therefore, if we want to discover the dependence of on frequency, we
might start by trying to find a relation between P and E, where E arises
from the electromagnetic wave passing through the material. The idea is
to look at a typical electron bound to one of the atoms in the dielectric and
work out the dipole moment of this atom in terms of the applied field in the
wave.
The electron (with charge q) is subject to several forces. First, there
is the force from the wave. The waves electric component has the form
(assuming its polarized in the x direction):

(3) E = E0 ei(kzt) x
For a fixed point, say z = 0, the field oscillates in place so the force on
the electron is the real part of the field times the charge:

(4) FE = qE0 cos t x


Second, the electron experiences a binding force with the nucleus. A
simple model that we used earlier took the electron to be a sphere of uniform
charge density, of radius a (the Bohr radius in hydrogen, which is 5.29
1
FREQUENCY DEPENDENCE OF ELECTRIC PERMITTIVITY 2

1011 m) centred on the nucleus. In this case, when the electron is displaced
a distance x from equilibrium, the binding force is

Zq2
(5) Fb = xx
40 a3

where Z is the atomic number (number of protons in the nucleus) and the
minus sign is because Fb pulls the electron back towards equilibrium. This
force is a harmonic oscillator force, since

(6) Fb = kb xx
Zq2
(7) kb
40 a3
The harmonic oscillator force has a natural frequency of
r
kb
(8) 0 =
m

so we can write the binding force as

(9) Fb = m02 xx
Example 1. We can work out this natural frequency for a hydrogen-like
atom with Z = 1. We get

2
1.6 1019
(10) kb = 3
4 (8.85 1012 ) (5.29 1011 )
(11) = 1.55 103 kg s2
r
kb
(12) 0 =
m
s
1.55 103
(13) =
9.1 1031
(14) = 4.13 1016 s1
0
(15) 0 = = 6.58 1015 s1
2
This frequency is in the near ultraviolet, just beyond the violet end of the
visible spectrum.
FREQUENCY DEPENDENCE OF ELECTRIC PERMITTIVITY 3

Finally, there will be a damping force because, once the wave is turned
off, we expect the electron to eventually return to its equilibrium position.
This can happen by radiating away energy or from interactions with other
fields in the material. The simplest damping force is proportional to, and
opposite in direction to, the velocity, so we can let

(16) Fd = mxx

where is the damping constant.


Since all the forces act in the x direction, we can drop the vector notation
and apply Ftotal = mx to get

(17) mx = mx m02 x + qE0 cos t


At this point, we can do the usual trick of introducing complex numbers
by defining x to be the real part of a complex variable x. This equation then
becomes

q
(18) x + x + 02 x = E0 eit
m
The solution is

(19) x (t) = x0 eit

since if we substitute this into the ODE we get

q
(20) 2 x0 eit i x0 eit + 02 x0 eit = E0 eit
m
qE0
(21) x0 = 2

m 0 2 i
The dipole moment p of the atom is the charge times the separation,
which is given by x, so it is the real part of

(22) p = qx0
q2 E0
(23) = 2
 eit
2
m 0 i
q2
(24) p =  E
m 02 2 i
FREQUENCY DEPENDENCE OF ELECTRIC PERMITTIVITY 4

We have now achieved our objective (for a single electron) since we have
the dipole moment in terms of the applied electric field. However, be-
cause of the damping term, the real part of p is not directly proportional
to E = E0 cos t so the medium isnt linear. We still need to calculate the
polarization density P to get the permittivity. If there are f j electrons with
natural frequency 0 = j and damping constant = j per atom (or mol-
ecule) and N atoms per unit volume, then

(25) P = N f j p j
j
Nq2 fj
(26) = E  
m j 2 2
i
j j

Generalizing 1 so that the susceptibility is complex, we have

P = 0 e E
Nq2 fj
e =  
0 m j 2 2 i
j j

This gives a complex dielectric constant and permittivity:

(27) r = 1 + e
Nq2 fj
(28) = 1+  
0 m j 2 2 i
j j

(29) = 0 r
Nq2 fj
(30) = 0 +  
m j 2 2 i
j j

P INGBACKS
Pingback: Refraction and dispersion coefficients
Pingback: Radiation damping
REFRACTION AND DISPERSION COEFFICIENTS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.23b.
In a dispersive medium, the permittivity depends on the frequency of
electromagnetic radiation.

Nq2 fj
(1) = 0 + 2
m j j 2 i j

where there are f j electrons per atom with natural frequency j and damp-
ing factor j , and there are N atoms per unit volume. Because is complex,
the medium isnt linear in the sense that the polarization is directly propor-
tional to the applied field, but if we take both the polarization P and field E
to be complex, then the medium is linear in the sense that

(2) P = 0 e E
With this assumption, we can substitute the complex permittivity for
the ordinary real permittivity in Maxwells equations and follow through
the same steps to get the wave equation, which now becomes

2 E
(3) 2 E =
t 2
Just as before, we can get plane wave solutions of the form

(4) E (z,t) = E0 ei(kzt )

where k is a complex wave vector

p
(5) k =
The actual real and imaginary parts of k are complicated expressions
since is a sum of complex numbers, but we can use the shortcut nota-
tion
1
REFRACTION AND DISPERSION COEFFICIENTS 2

(6) k = k + i

giving (assuming E is polarized in the x direction):

(7) E (z,t) = E0 ez ei(kzt) x


2
The intensity of the radiation is proportional to E so the intensity falls
off according to e2z as we penetrate the medium. The absorption coeffi-
cient is defined as

(8) 2

and gives a measure of the reciprocal of the distance at which the intensity
is attenuated.
We can write the complex permittivity in 1 as

(9)
" #
Nq2 fj 2j 2 + i j
= 0 +
m 2 2 2 2
j j i j j + i j
(10)
 
Nq2 f j 2j 2 Nq2 f j j
= 0 + 2 + i 2
m j m j
 
2 2
2j 2 + j 2j 2 + j


If we stay away from the resonant frequencies, where j , the sum


terms are quite small so we can approximate them in 5 by the first order term
in a Taylor expansion. If we also take 0 as is true of most materials,

and use c = 1/ 0 0 , we get

s

(11) k =
0 c2

Using 1 + x 1 + 12 x for small x, we get
REFRACTION AND DISPERSION COEFFICIENTS 3

(12)  
2
fj j 2
Nq2 Nq2 f j j
k 1 + +i

c
20 m j
 2
2

c

20 m  2
2

2j 2 + j 2j 2 + j
 
j

From 7 the speed of the wave is


(13) v=
k

so the index of refraction is

c
(14) n =
v  
Nq2 f j 2j 2
(15) 1+
20 m
 2 2
j 2j 2 + j

and the absorption coefficient is

(16) = 2
Nq2 2 f j j
(17) 2
c0 m j
 2
2
j 2 + j
If we stay away from resonances, the damping term becomes insignifi-
cant so the index of refraction is approximately

Nq2 fj
(18) n 1+ 2
20 m j j 2
If the frequency of the wave is significantly less than all the resonant
1
frequencies j we can further approximate this using 1x 1 + x for small
x:

Nq2 fj
(19) n 1+  
20 m j 2 1 2 / 2
j j

Nq2 fj 2 Nq
2 fj
(20) 1+ 2
+
20 m j j 20 m j 4j
REFRACTION AND DISPERSION COEFFICIENTS 4

In a vacuum, c = = /2 so

Nq2 f j 4 2 c2 Nq2 fj
(21) n 1+ +
20 m j 2j 2 20 m j 4j
 
B
(22) = 1+A 1+ 2

where

Nq2 fj
(23) A =
20 m j 2j
4 2 c2 Nq2 fj
(24) B =
A 20 m j 4j
Eqn 22 is known as the Cauchy formula, although Cauchy had many
equations named after him (particularly in the area of complex variable the-
ory), so the name is easily confused with other formulas. The parameter A is
the coefficient of refraction and B is the coefficient of dispersion. The more
usual form of Cauchys equation seems to be n = 1 + A + B2 (I couldnt find
any sources that gave the formula in the form used by Griffiths).
Example. Applying this model to hydrogen at 0 C and atmospheric pres-
sure (that is, standard temperature and pressure, or STP), the number of
electrons per molecule of H2 is f j = 2. In the previous post, we found
that the resonant frequency is 0 = 4.13 1016 s1 . At STP, an ideal gas
occupies 22.414 m3 kmol1 , so the number density is

6.02 1023 (1000)



(25) N= = 2.69 1025
22.414
The parameters are

Nq2 2
(26) A= = 5 105
20 m 02
4 2 c2 Nq2 2 4 2 c2
(27) B= 4
= 2
= 2.08 1015 m2
A 20 m 0 0
The experimental values quoted by Griffiths are

(28) A = 1.36 104


(29) B = 7.7 1015
REFRACTION AND DISPERSION COEFFICIENTS 5

so the values from the model are at least around the right order of magni-
tude.
P INGBACKS
Pingback: Resonances in a dispersive medium
Pingback: Group velocity of electromagnetic waves in a dispersive medium
RESONANCES IN A DISPERSIVE MEDIUM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.24.
In a dispersive medium, the index of refraction is
 
Nq2 f j 2j 2
(1) n 1+
20 m
 2 2
j 2j 2 + j

and the absorption coefficient is

Nq2 2 f j j
(2) 2
c0 m j
 2
2
j 2 + j

We can look at the behaviour of these coefficients near one of the reso-
nances, that is, when j for some j. To simplify things, well assume
that there is only one term in the sum (that is, only one resonance). In
practice, when were near one resonance, the other resonances dont affect
things much unless they are very close together.
In this case, well take the one natural frequency to be 0 and the associ-
ated damping coefficient to be 0 , and define


(3) x
0
Then we get, after dividing top and bottom by 02 :

Nq2 f0 1 x2
(4) n 1+
20 m 2 (1 x2 )2 + 2 x2
0 0
Nq2 f 0 0 x2
(5)
m0 c 2 (1 x2 )2 + 2 x2
0 0
The index of refraction n rises to a peak when is just before 0 , then
dips sharply, reaching a minimum just after 0 , after which it rises slowly
1
RESONANCES IN A DISPERSIVE MEDIUM 2

again. We can find the maximum and minimum by setting the derivative to
zero and solving for x. Using Maple to simplify the result, we get
 2 
x x 4 2x2 + 1 0
dn Nq2 f
0 2
(6) =   2  0  =0
dx 0 m 2 x4 0 2 x2 + 1
0 20

The roots are


r r
0 0
(7) x = 0, 1 + , 1
0 0
The negative and q
zero roots arent of interest, soqwe see that n reaches
its maximum at x = 1 0 and minimum at x = 1 + 00 . If 0  0 ,
0

these approximate to

1
(8) n max 0 0
2
1
(9) n min 0 + 0
2

so the width of the anomalous region is 0 .


From 5, we see that the absorption reaches a maximum when x = 1 (this
can be checked by calculating the derivative) and has a value of

Nq2 f0
(10) max =
m0 c0
Substituting the positive roots from 7 into 5 and dividing by max we find

n max 0 0 1
(11) =
max 20 0 2
n min 0 + 0 1
(12) =
max 20 + 0 2
Thus for small , the index of refraction reaches its maximum and mini-
mum values roughly where the absorption is half its maximum.
P INGBACKS
Pingback: Radiation damping
GROUP VELOCITY OF ELECTROMAGNETIC WAVES IN A
DISPERSIVE MEDIUM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.25.
In a dispersive medium, the permittivity is a complex quantity given by

Nq2 fj
(1) = 0 + 2
m j j 2 i j

and the wave vector is also complex:

p
(2) k =
If the sum term in
1 is small compared to 0 , we can approximate it in
the square root using 1 + x 1 + 12 x.

" #1/2
Nq2 fj
(3) k = 0 1 +
m0 j 2j 2 i j
!
Nq2 fj
(4) 1+
c 2m0 2 2
j j i j


where weve taken 0 and used 0 0 = 1c .
The solution to the wave equation is

(5) E (z,t) = E0 ei(kzt )


If the damping factors j are all small and were away from any resonance
frequencies ( j ), then we can ignore the imaginary term in the denominator
and get a real wave vector:
!
Nq2 fj
(6) k= 1+ 2
c 2m0 j j 2
1
GROUP VELOCITY OF ELECTROMAGNETIC WAVES IN A DISPERSIVE MEDIUM 2

d
We can get the group velocity dk of a wave packet by implicit differen-
tiation:

d 1 Nq2 fj 2 f j
(7) 1= + + 

2 2 2
dk c 2m0 c j j j 2j 2
" !#1
d Nq2 fj 2 2
(8) = c 1+ 1+ 2
dk 2m0 2
j j
2 j 2
" !#1
Nq2 fj 2j 2 2 2
(9) = c 1+ +
2m0 2
j j
2 2j 2 2j 2
  1
2
fj j + 2
Nq2
(10) = c 1 +


2m0 j

2
2
j 2

Since everything inside the square brackets is positive, we see that d


dk < c
for all frequencies. On the other hand, from 6 we see that the phase velocity
is
!1
Nq2 fj
(11) = c 1+
k 2m0 j 2j 2

which can actually exceed c for some frequencies near the resonant frequen-
cies, since the denominator in the sum can be negative there. Griffiths states
that ordinarily, energy (and thus information) carried by the wave travels at
the group velocity which is always less that c. However, it is possible for
the group velocity to exceed c in some cases.
P INGBACKS
Pingback: Wave guide: energy flows at the group velocity
WAVE GUIDES: DERIVATION OF THE WAVE EQUATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.26.
A wave guide is a hollow tube which allows electromagnetic waves to
travel down it. Wave guides are usually made of conductors, so well as-
sume that they are made of a perfect conductor so that E = 0 everywhere
inside the conducting boundary.
The boundary conditions implied by Maxwells equations are

(1) 1 E1 2 E2 = f
(2) B
1 B2 = 0
k k
(3) E1 E2 = 0
1 k 1 k
(4) B B = K f n
1 1 2 2
In particular, 3 tells us that the parallel component of E is zero at the
boundary of the wave guide. As usual, well take the z axis to be parallel to
wave guides axis, so the waves have the form

(5) E = E0 (x, y) ei(kzt)


(6) B = B0 (x, y) ei(kzt)
Maxwells equations inside the guide (assumed to be a vacuum) are

(7) E = 0
(8) B = 0
B
(9) E =
t
1 E
(10) B = 2
c t
Previously, we applied the divergence equations to show that the waves
were transverse (no z component), but that relied on the waves being un-
bounded plane waves, which is not the case here. It turns out that waves in
1
WAVE GUIDES: DERIVATION OF THE WAVE EQUATION 2

a wave guide are not transverse in general, in that at least one of E and B
must have a longitudinal component. We therefore write

(11) E0 (x, y) = Ex x + Ey y + Ez z
(12) B0 (x, y) = Bx x + By y + Bz z

where all the components on the RHS depend on x and y, and may be com-
plex functions. Putting these together with 5 and 6 into 9, we get (remem-
bering that the components do not depend on z):

(13) E = (y Ez ikEy ) x + (x Ez + ikEx ) y + (x Ey y Ex ) z


B
(14) =
t
(15) = iBx x + iBy y + iBz z
Equating components, we get

(16) y Ez ikEy = iBx


(17) x Ez + ikEx = iBy
(18) x Ey y Ex = iBz
We can apply exactly the same procedure to 10 to get the analogous equa-
tions


(19) y Bz ikBy = i Ex
c2

(20) x Bz + ikBx = i 2 Ey
c

(21) x By y Bx = i 2 Ez
c
We can solve these 6 equations to get the x and y components in terms of
the z components. For example, multiplying 17 through by k and 19 through
by and adding, we get

2
(22) kx Ez + ik2 Ex i Ex = y Bz
c2
1
(23) Ex = (kx Ez + y Bz )
i (k2 2 /c2 )
i
(24) = (kx Ez + y Bz )
2 /c2 k2
WAVE GUIDES: DERIVATION OF THE WAVE EQUATION 3

Similarly, we can get the other 3 equations. Multiply 16 by k and 20 by


and subtract to get

i
(25) Ey = (ky Ez x Bz )
2 /c2 k2
Multiply 16 by /c2 and 20 by k and subtract to get

i  
(26) Bx = k B
x z E
y z
2 /c2 k2 c2
Multiply 17 by /c2 and 19 by k and add to get

i  
(27) By = k y B z + x Ez
2 /c2 k2 c2
To get the wave equations we can apply 7 and 8:

i
(28) E = [(kxx Ez + yx Bz ) + (kyy Ez xy Bz )] + ikEz
2 /c2 k2
i
(29) = [kxx Ez + kyy Ez ] + ikEz
2 /c2 k2
(30) =0
The wave equation for Ez is thus

xx + yy + 2 /c2 k2 Ez = 0
 
(31)
Exactly the same procedure applied to B = 0 gives

xx + yy + 2 /c2 k2 Bz = 0
 
(32)
Solving these two equations subject to the boundary conditions will give
us all 3 components of each field.
P INGBACKS
Pingback: Rectangular wave guides: transverse electric waves
Pingback: Wave guide: energy flows at the group velocity
Pingback: Rectangular resonant cavity
RECTANGULAR WAVE GUIDES: TRANSVERSE ELECTRIC
WAVES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.27.
The wave equations for an electromagnetic wave in a wave guide are

xx + yy + 2 /c2 k2 Ez = 0
 
(1)
xx + yy + 2 /c2 k2 Bz = 0
 
(2)
If Ez = 0 we have a transverse electric or TE wave, so we need to solve
only the Bz equation. For a rectangular wave guide with dimensions of a in
the x direction and b in the y direction, we can use separation of variables
to get a solution (the technique is mathematically the same as that used in
solving the infinite square well in quantum mechanics). That is, we assume
that

(3) Bz (x, y) = X (x)Y (y)

and substitute this into 2, then divide through by XY :

X 00 Y 00 2
(4) + + 2 k2 = 0
X Y c
Because this equation must hold for all values of x and y, the two deriva-
tive terms must separately be constant, so we have

(5) X 00 = kx2 X
(6) Y 00 = ky2Y

for some constants kx and ky , which satisfy

2
(7) kx2 ky2 + k2 = 0
c2
1
RECTANGULAR WAVE GUIDES: TRANSVERSE ELECTRIC WAVES 2

The general solution to these ODEs can be written as either trigonometric


functions or complex exponentials. If we choose trig functions, we have

(8) X (x) = A sin kx x + B cos kx x


(9) Y (y) = C sin ky y + D cos ky y
The boundary conditions require that at x = 0 and x = a

(10) B
1 B2 = 0
The component of B perpendicular to the walls of the guide in the x
direction is Bx , and since B = 0 inside the conducting wall, this condition
requires Bx = 0. In our derivation of the wave equation for a wave guide,
we found that

i  
(11) Bx = 2 2 kx Bz 2 y Ez
/c k2 c

and since Ez = 0, we must also have x Bz = 0, which in turn requires that

(12) X 0 (0) = X 0 (a) = 0


From 8, this means that A = 0 and

(13) kx B sin kx a = 0

so

m
(14) kx =
a

for m equal to a non-negative integer.


Exactly the same analysis on Y gives us

n
(15) ky =
b

so the separation of variables solution gives us

mx ny
(16) Bz (x, y) = B0 cos cos
a b
This is known as the TEmn mode. From 7 we get the wave number
RECTANGULAR WAVE GUIDES: TRANSVERSE ELECTRIC WAVES 3

s
2
 2
n2

2
m
(17) k= +
c2 a2 b2
In fact, at least one of m or n must be non-zero, as we can see by the
following argument. Suppose m = n = 0. Then k = c . We need the results
we got in the previous post:

(18) y Ez ikEy = iBx


(19) x Ez + ikEx = iBy
(20) x Ey y Ex = iBz


(21) y Bz ikBy = i Ex
c2

(22) x Bz + ikBx = i 2 Ey
c

(23) x By y Bx = i 2 Ez
c

With k = c and Ez = 0 we get from 21 and 19


(24) y Bz i By = i 2 Ex
c c

(25) i 2 Ex = i By
c c
Adding these equations gives us

(26) y B z = 0
Similarly, from 22 and 18 we get

(27) x Bz = 0
Therefore, Bz is constant in both the x and y directions, so it is a constant
overall. To find what constant this is, we can use Faradays law in integral
form:

d B
(28) E d` = = da
dt t
The area of integration on the RHS that well choose is a cross-section of
the wave guide, so the path of integration on the LHS is around the rectan-
gular boundary of the guide. We know that
RECTANGULAR WAVE GUIDES: TRANSVERSE ELECTRIC WAVES 4

(29) B = B0 ei(kzt)

so

B
(30) = iB0 ei(kzt)
t
In this case da points in the z direction, so we get

i(kzt)
(31) E d` = ie Bz da

(32) = iabei(kzt) Bz

since Bz is constant over the area of integration.


Since weve chosen the path of integration to be the boundary of the wave
guide, we can use the boundary condition

k k
(33) E1 E2 = 0
The field inside the conducting wall of the wave guide is zero, so the
parallel field at the boundary must also be zero and, since E d` = E k d`, the
line integral must also be zero, so we must have

(34) Bz = 0
That is, if m = n = 0, both the electric and magnetic fields must be trans-
verse (known as TEM mode). Griffiths shows in section 9.5 of his book
that in this case, for a hollow wave guide of any cross section, the electric
field must actually be zero everywhere, which means no wave can propa-
gate down the guide. Thus the T E00 mode cannot exist in a hollow wave
guide.
P INGBACKS
Pingback: Rectangular wave guides: TE modes and the cutoff frequency
Pingback: Wave guide: energy flows at the group velocity
Pingback: Rectangular wave guide: transverse magnetic modes
Pingback: Coaxial wave guides: TEM mode
RECTANGULAR WAVE GUIDES: TE MODES AND THE
CUTOFF FREQUENCY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.28.
The possible TE modes in a rectangular wave guide are given by

mx ny
(1) Bz (x, y) = B0 cos cos
a b

where m and n are integers and a and b are the dimensions of the rectangle,
with a b. These values are related to the wave number by

s
2
 2 2

m n
(2) k= 2 +
c2 a2 b2

so any modes where

m2 n2
 
2 2 2
(3) <c +
a2 b2

are not allowed, as they would give an imaginary value of k. The frequency

r
m2 n2
(4) mn = c +
a2 b2

is therefore the cutoff frequency for the mode TEmn .


As an example, suppose we had a wave guide with a = 2.28 cm, b =
1.01 cm and a frequency of = 1.70 1010 Hz. Then the condition for a
real k is
1
RECTANGULAR WAVE GUIDES: TE MODES AND THE CUTOFF FREQUENCY 2

2 m2 n2
(5) +
2 c2 a2 b2
4 2 m2 n2
(6) +
c2 a2 b2
 2
n2

4 4 m
(7) 1.284 10 10 +
5.2 1.02
(8) 1.02m2 + 5.2n2 6.813
Since at least one of m and n must be non-zero, the allowed modes are
TE10 , TE20 , TE01 and TE11 (where the first subscript is m and the second is
n).
If we wish to allow only one mode, this would have to be TE10 (since
that gives the lowest value on the LHS). The angular frequency range that
would do this is between 10 and 20 , that is

c 2c
(9) 10 = < = 20
a a
For the values above

(10) 10 = 4.134 1010 s1


(11) 20 = 8.268 1010 s1
The corresponding frequencies and wavelengths are

10
(12) 10 = = 6.58 109 Hz
2
(13) v20 = 1.32 1010 Hz
c
(14) 10 = = 0.0456 m
10
(15) 20 = 0.0228 m
WAVE GUIDE: ENERGY FLOWS AT THE GROUP VELOCITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.29.
Its possible for the phase velocity of an electromagnetic wave to exceed
c in some situations. In the case of a TE wave in a rectangular wave guide
its possible to show that the energy in the wave actually moves at the group
velocity rather than the phase velocity, and that the group velocity is always
less than c. The Poynting vector gives the rate per unit time and per unit
area at which energy flows, and its time average is given by

1
E B

(1) hSi =
20
The surface integral of hSi over the cross section of the wave guide thus
gives the energy flux along the guide. Since the normal da to the cross
section is along the z direction, we need only the z component of hSi:

1
Ex By Ey Bx

(2) hSz i =
20
Well also need the time average of the energy density
 
1 1
(3) huem i = E E + B B
40 c2
The z component of magnetic field for a TE mode is

mx ny
(4) Bz (x, y) = B0 cos cos
a b

where m and n are integers and a and b are the dimensions of the rectangle,
with a b. These values are related to the wave number by

1
q
(5) k= 2 mn
2
c

where
1
WAVE GUIDE: ENERGY FLOWS AT THE GROUP VELOCITY 2

r
m2 n2
(6) mn = c +
a2 b2

is the cutoff frequency for the mode TEmn .


The components of E and B in the TE mode can be obtained from, where
weve set Ez = 0:

i
(7) Ex = y Bz
2 /c2 k2
i
(8) Ey = 2 2 x Bz
/c k2
i
(9) Bx = kx Bz
/c2 k2
2

i
(10) By = ky Bz
/c2 k2
2

Substituting 4 into these equations, we get

nB0 mx ny
(11) Ex = i cos sin
b a b
mB0 mx ny
(12) Ey = i sin cos
a a b
mkB0 mx ny
(13) Bx = i sin cos
a a b
nkB0 mx ny
(14) By = i cos sin
b a b

where

2
(15) k2
c2

We now get
WAVE GUIDE: ENERGY FLOWS AT THE GROUP VELOCITY 3

(16)
2 kB20 n2 m2 2 mx
 
2 mx 2 ny 2 ny
hSz i = cos sin + 2 sin cos
2 2 0 b2 a b a a b
(17)
2 B20 n2 m2 2 mx
  2 
2 mx 2 ny 2 ny 2
huem i = 2 cos sin + 2 sin cos +k
0 b2 a b a a b c2
B20 mx ny
+ cos2 cos2
40 a b
We can now integrate these two quantities over the rectangle 0 x a,
0 y b. This is straightforward but tedious so we can get Maple to do it.
We get

(18)
b
2 k n2 a2 + m2 b2 B20
a

e f low hSz i dx dy =
0 0 8ab 2 0
abk 2
(19) = 2 2 mn B20
8c 0
(20)
b
2 n2 a2 + m2 b2 B20 2
a
 
abB20

2
evol huem i dx dy = + k +
0 0 16ab 2 0 c2 160
abB20 mn 2 2
   
(21) = + k2 + 1
160 2 c2 c2

using 6.
The first integral gives the rate at which energy is flowing along the z
direction, while the second integral gives the total energy in a volume of
unit length with the cross sectional area of the wave guide. The speed vE of
the energy flow is the rate at which energy flows past a certain point (given
by the first integral) divided by the amount of energy in a unit length (the
second integral). That is

e f low
(22) vE =
evol
2  2 2  1
2kmn mn 2
(23) = +k +1
c2 2 2 c2 c2
It is actually easier to work out 1/vE first. We get
WAVE GUIDE: ENERGY FLOWS AT THE GROUP VELOCITY 4

2
2
1 2 +k c2 2
(24) = c + 2
vE 2k 2kmn
We can eliminate k using 5:

2
(25) = k2
c2
mn2
(26) =
c2
1 2 2 mn
2 2
mn
(27) = p + p
vE 2c 2 mn 2 2c 2 mn
2

(28) = p
2
c mn 2

Therefore, the speed of the energy flow is


r
2
mn
(29) vE = c 1
2
This is less than c since mn is the minimum frequency at which the mode
TEmn will propagate in the wave guide, so mn .
From 5, the group velocity is

d 1
(30) vg = = = vE
dk dk/d

so we see that the energy propagates at the group velocity.


RECTANGULAR WAVE GUIDE: TRANSVERSE MAGNETIC
MODES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.30.
We can work out the theory of a transverse magnetic (TM) wave in a
rectangular wave guide of dimensions a in the x direction and b in the y
direction, in the same way as for the TE wave. In a TM wave the component
Bz parallel to the axis of the wave guide is zero, so we have only a single
wave equation to solve.

2
 
2
(1) xx + yy + 2 k Ez = 0
c
As this is identical to the equation for TE waves, we have the same solu-
tion:

(2) Ez = X (x)Y (y)


(3) X (x) = A sin kx x + B cos kx x
(4) Y (y) = C sin ky y + D cos ky y
The boundary conditions are different here, however, as we require

k k
(5) E1 E2 = 0
If the wave guide is a perfect conductor, then E = 0 inside it, so Ek = 0 at
all boundaries of the guide. In particular, Ez = 0 for x = 0, a and y = 0, b.
This means B = D = 0, so

(6) Ez (x, y) = E0 sin kx x sin ky y

and

m
(7) kx =
a
n
(8) ky =
b
1
RECTANGULAR WAVE GUIDE: TRANSVERSE MAGNETIC MODES 2

Because Ez uses sines rather than cosines, the integers m and n must both
be non-zero in order for the wave to exist at all, so the lowest TM mode is
TM11 .
The wave number has the same form as in the TE case:
s
2
 2
n2

2
m
(9) k = +
c2 a2 b2
1
q
(10) 2 mn
2
c s
 2
n2

m
(11) mn = c +
a2 b2
The phase and group velocities are the same as for TE waves


(12) vp =
k r
d 2
(13) vg = = c 1 mn
dk 2
The ratio of lowest cutoff frequencies is
q
TM
11 c a12 + b12 1p 2
(14) TE
= = a + b2
10 c/a b
COAXIAL WAVE GUIDES: TEM MODE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.31.
Although purely transverse electromagnetic (TEM) waves cant exist within
a hollow wave guide, it is possible to have TEM waves in a coaxial trans-
mission line. To see this, we start with Maxwells equations in the form

(1) y Ez ikEy = iBx


(2) x Ez + ikEx = iBy
(3) x Ey y Ex = iBz


(4) y Bz ikBy = i Ex
c2

(5) x Bz + ikBx = i 2 Ey
c

(6) x By y Bx = i 2 Ez
c
By setting Bz = Ez = 0 we see from 1 and 5 that

k
(7) Bx = Ey

k 2
(8) Ey = 2 Ey
c

(9) k =
c
Substituting this into 1 and 2 we get

(10) Ey = cBx
(11) Ex = cBy

from which we see that

(12) E B = [cBy , cBx , 0] [Bx , By , 0] = 0


1
COAXIAL WAVE GUIDES: TEM MODE 2

so the electric and magnetic fields are perpendicular to each other.


From 3, 6 and the other two Maxwell equations for vacuum: E = 0
and B = 0 we get

(13) x Ey y Ex = 0
(14) x By y Bx = 0
(15) x Ex + y Ey = 0
(16) x Bx + y By = 0
Since the components of E and B dont depend on z [remember the z de-
pendence is contained in the complex exponential in the form E (x, y, z,t) =
E0 (x, y) ei(kzt) and B (x, y, z,t) = B0 (x, y) ei(kzt) ] the first equation is
equivalent to saying E0 = 0 and the second to B0 = 0. Together
with E = 0 and B = 0, these are Maxwells equations for static fields
[ tE = tB = 0] in empty space (no free charge or current). In a cylindrical
coaxial cable with an inner cylinder of radius a and an outer cylinder of
radius b, the magnetic field is, for a < r < b

0 I
(17) B0 =
2r

where I is the steady current in the inner cylinder. The electric field due to
an infinite line of charge is


(18) E0 = r
20 r

where is the linear charge density. These are formal solutions for the case
of cylindrical symmetry; the important thing is that the fields have the forms

A
(19) B0 =
cr
A
(20) E0 = r
r

for some constant A.


Plugging these into the full formulas for E and B and taking the real part,
we get
COAXIAL WAVE GUIDES: TEM MODE 3

A cos (kz t)
(21) E = r
r
A cos (kz t)
(22) B =
cr
These equations satisfy Maxwells equations:

1
(23) E = (rEr ) = 0
r r
1 B
(24) B = =0
r
Er
(25) E =
z
A sin (kz t)
(26) = k
r
A sin (kz t)
(27) =
c r
B
(28) =
t
B 1 
(29) B = r + rB z
z r r
A sin (kz t)
(30) = k r
cr
A sin (kz t)
(31) = r
c2 r
1 E
(32) =
c2 t
The boundary conditions are

(33) B
1 B2 = 0
k k
(34) E1 E2 = 0

and since B is circumferential, its component normal to the cylinders is


zero, and since E is radial, its parallel component is zero, so the boundary
conditions are satisfied.
By comparison with 17 and 18 we can write the analogs for the full fields
COAXIAL WAVE GUIDES: TEM MODE 4

0 I (z,t)
(35) B=
2r

(z,t)
(36) E= r
20 r
From this we can read off the current and charge density on the inner
cylinder.

2A cos (kz t)
(37) I (z,t) =
0 c
(38) (z,t) = 20 A cos (kz t)
FOURIER TRANSFORM OF SUPERPOSITION OF PLANE
WAVES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.32.
The general solution to the wave equation can be written as a superposi-
tion of plane waves:


(1) f (z,t) = A (k) ei(kzt) dk

where A (k) is the (complex) amplitude of waves with wave number k, that
is, its the contribution to the overall wave f of waves with a given wave
number. At t = 0, this is


(2) f (z, 0) = A (k) eikz dk

and its time derivative at t = 0 is


(3) f (z, 0) = i (k) A (k) eikz dk

Note that we cant take outside the integral as it is, in general, a function
of k: = (k).
According to Plancherels theorem, the Fourier transform and its inverse
of a function (k) are


(4) (z) = (k) eikz dk


1
(5) (k) = (z) eikz dz
2
Therefore, the inverse transforms of 2 and 3 are
1
FOURIER TRANSFORM OF SUPERPOSITION OF PLANE WAVES 2


1
(6) A (k) = f (z, 0) eikz dz
2

1
(7) i (k) A (k) = f (z, 0) eikz dz
2

i
(8) f (z, 0) eikz dz = f (z, 0) eikz dz

In this case, we can put inside the integral in the last line, since it
doesnt depend on z.
Now let

(9) f = f + ig

where f and g are real functions. Then from 8, the real and imaginary parts
must be equal on each side, so

ikz i
f + ig eikz dz

(10) ( f + ig) e dz =

 
i 1
(11) = f g eikz dz

g
(12) f =

f
(13) g =

[We can ignore the e ikz in these calculations since the same factor ap-
pears in both integrals, so in order for the real and imaginary parts of the
two integrands to be equal, the factor multiplying eikz must have equal real
and imaginary parts on both sides.]
Substituting this back into 6 we get
 
1 i
(14) A (k) = f (z, 0) + f (z, 0) eikz dz
2
SPHERICAL ELECTROMAGNETIC WAVE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 9, Post 33.
One form of spherical wave has an electric component given by
 
sin sin (kr t)
(1) E (r, ,t) = A cos (kr t)
r kr

where A is a constant. We can derive the corresponding magnetic field from


Maxwells equations in vacuum:

(2) E = 0
(3) B = 0
B
(4) E =
t
E
(5) B = 0 0
t
The calculations are straightforward but can get messy, so well use Maple
to do the derivatives. We get

1  1 
(6) E = sin E r rE
r sin  r r 
2A cos sin (kr t)
(7) = cos (kr t) r+
r2 kr
  
A sin 1 cos (kr t)
k 2 sin (kr t) +
r kr r
Integrating this with respect to t we get
 
2A cos cos (kr t)
(8) B= sin (kr t) + r+
r2 kr
  
A sin 1 2
(9) kr cos (kr t) + r sin (kr t)
r3 k
1
SPHERICAL ELECTROMAGNETIC WAVE 2

We can verify that E and B satisfy the other 3 Maxwell equations by


direct calculation.

1 
(10) E = E = 0
r sin
1 2  1
(11) B = 2 r Br + (sin B )
r r  r sin 
2A cos 1 sin (kr t)
(12) = k 2 cos (kr t) +
r2 kr r
   
2A cos 1 1 2 sin (kr t)
kr cos (kr t) +
r2 r2 k r
(13) =0
 
1 Br
(14) B = (rB )
r r
  
2A sin 1
(15) = kr cos (kr t) + sin (kr t) +
r3 kr
   
A sin 1 2
kr cos (kr t) kr 1 sin (kr t) +
r3 k
 
2A sin cos (kr t)
sin (kr t) +
r3 kr
sin
(16) = A 2 [kr sin (kr t) + cos (kr t)]
r c

where in the last line we used /k = c and collected terms. From 1 we get

 
E sin cos (kr t)
(17) = A sin (kr t) +
t r kr
sin
(18) = A 2 [krc sin (kr t) + c cos (kr t)]
r
2
(19) = c B

so the final Maxwell equation is satisfied.


The Poynting vector is, after simplifying
SPHERICAL ELECTROMAGNETIC WAVE 3

(20)
1
S= EB
0
A2 sin2
  
1 2 2 3 3 2
= r r k sin [2 (kr t)] kr cos [2 (kr t)] + k r cos (kr t) +
0 kr5 2
(21)
A2 sin 2 1 2 2
 

k r 1 sin [2 (kr t)] + kr cos [2 (kr t)]
0 kr5 2
The intensity, or time average of the Poynting vector is

2/
(22) I = hSi = Sdt
2 0
A2 k sin2
(23) = r
20 r2
The energy flows radially outwards and falls off as r2 .
The total power radiated is the integral of I da over a sphere:

(24) P = I da

A2 k r 2 3
(25) = sin d
0 0 r 2
4A2 k
(26) =
30
4A2
(27) =
30 c
P INGBACKS
Pingback: Fields of an oscillating magnetic dipole
TRANSMISSION COEFFICIENT FOR A WAVE PASSING
THROUGH 3 MEDIA

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.34.
We can extend the analysis of reflection and transmission of waves at a
boundary by considering the case of an electromagnetic wave starting out
in medium 1 (with wave speed v1 , wave number k1 and index of refraction
n1 = c/v1 ), then passing at normal incidence to medium 2 at z = d and
then to medium 3 at z = 0. (Weve changed the origin from that stated in
Griffithss problem to make the analysis a bit easier, as well see). Wed
like to find the transmission coefficient between mediums 1 and 3, that is,
wed like to see how much of the waves energy gets transmitted all the way
through the middle medium. Well assume the mediums are all homoge-
neous and linear, and that = 0 in all of them.
The analysis is much the same as the earlier method, but a bit more com-
plicated. We have a wave E1R travelling in towards the right in medium
1. At the boundary with medium 2, it gives rise to a reflected wave E1L
travelling to the left in medium 1 and a trasmitted wave E2R travelling to
the right in medium 2. When this wave hits the boundary with medium 3,
there is a reflected wave E2L travelling to the left and a transmitted wave
E3R travelling to the right. There are, of course, corresponding magnetic
waves B1L and so on. We can then apply the boundary conditions to work
out the amplitudes. [Actually, the wave reflected back to the left from the 2-
3 boundary will hit the 1-2 boundary and be reflected and transmitted there
too, so that there is, in principle, an infinite number of reflected and trans-
mitted waves resulting from the wave bouncing back and forth between the
two boundaries. However, we can subsume all the left-moving waves into
E1L and E2L and all the right moving waves into E1R , E2R and E3R . The
important thing is that these waves must satisfy the boundary conditions.]
We can take the electric component to be polarized along the x direction,
so that the magnetic component is then along the y direction. The waves are
1
TRANSMISSION COEFFICIENT FOR A WAVE PASSING THROUGH 3 MEDIA 2

(1) E1L = E1L ei(k1 zt) x


(2) E1R = E1R ei(k1 zt) x
(3) E2L = E2L ei(k2 zt) x
(4) E2R = E2R ei(k2 zt) x
(5) E3R = E3R ei(k3 zt) x
1
(6) B1L = E1L ei(k1 zt) y
v1
1
(7) B1R = E1R ei(k1 zt) y
v1
1
(8) B2L = E2L ei(k2 zt) y
v2
1
(9) B2R = E2R ei(k2 zt) y
v2
1
(10) B3R = E3R ei(k3 zt) y
v3

The negative signs for the left-moving magnetic waves are to keep the
Poynting vector pointing to the left. The coefficients E1L and so on are
actually complex numbers, but weve dropped the tilde (and the subscript 0
that Griffiths uses) to make the notation simpler.
Because the medium 2-3 boundary consists of an incident right-moving
wave, a reflected left-moving wave and a transmitted wave, it is identical
to the case we treated earlier, provided we take z = 0 at this point (which
weve done). We can therefore write down the results:

v3 v2
(11) E2L = E2R
v2 + v3
2v3
(12) E3R = E2R
v2 + v3

Now for the medium 1-2 boundary at z = d. From the boundary condi-
k k k k
tion E1 = E2 and (since = 0 everywhere) B1 = B2 we get

(13) E2R eik2 d + E2L eik2 d = E1R eik1 d + E1L eik1 d


1   1 
(14) E2R eik2 d E2L eik2 d = E1R eik1 d E1L eik1 d
v2 v1
TRANSMISSION COEFFICIENT FOR A WAVE PASSING THROUGH 3 MEDIA 3

These 4 equations are linear in the E coefficients so its straightforward


(although tedious) to solve them. Its easiest to let Maple handle this part,
and we get (since were interested only in expressing E3R in terms of E1R ):

(15)
eik1 d h ik2 d i
v2 v3 + v1 v2 v22 v1 v3 + eik2 d v2 v3 + v1 v2 + v22 + v1 v3

E1R = E3R e
4v2 v3
(16)
eik1 d 
2 cos (k2 d) (v2 v3 + v1 v2 ) 2i sin (k2 d) v22 + v1 v3

= E3R
4v2 v3
The intensity of a wave is

|E|2
(17) I=
2v

and the transmission coefficient is

I3R
(18) T=
I1R

so we get

(19)
2 (k d) (v v + v v )2 + 4 sin2 (k d) v2 + v v 2

v3 4 cos 2 2 3 1 2 2 1 3
T 1 = 2
v1 16v22 v23
(20)
2 +v v 2
" #
(v2 v3 + v1 v2 )2

1 v 1 3
1 sin2 (k2 d) + 2 2 sin2 (k2 d)

=
4v1 v3 v22 v2
(21)
2 + v v 2 (v v + v v )2
"  #
1 v 1 3 2 3 1 2
= (v3 + v1 )2 + sin2 (k2 d) 2
4v1 v3 v22
(22)
" #
v21 v22 v23 v22

1
= (v3 + v1 )2 + sin2 (k2 d)
4v1 v3 v22
We can express this in terms of the indexes of refraction by noting that
since ni = c/vi and the power of the vs is the same in numerator and denom-
inator so the factors of c cancel out and we can replace vi with 1/ni . We can
TRANSMISSION COEFFICIENT FOR A WAVE PASSING THROUGH 3 MEDIA 4

then multiply the first term top and bottom by n21 n23 and the second term top
and bottom by n21 n23 n42 to get
" #
2 n2 n2 n2

1 n
(23) T 1 = (n1 + n3 )2 + sin2 (k2 d) 1 2 3 2
4n1 n3 n22
Finally we note that the wave speed in medium 2 is v2 = /k2 = c/n2 so
k2 = n2 /c and we get
"  2 #
2 n2 n2



1 n 2 d n n
(24) T 1 = (n1 + n3 )2 + sin2 1 2 3 2
4n1 n3 c n22

C OMMENTS
From: Paul Landini
Time: October 22, 2017 at 7:30 pm
Comment: In problem 9.34 of Griffiths Electrodynamics book, how
do you arrive at equation 0.19 based from 0.16? I do not see where the
e^(ik_1d) disappears at, nor do I see where the imaginary terms go in the
expansion of the squared terms.
Thanks! Paul
=============
Equation 19 is the ratio of two intensities, which are defined by equation
17, which takes the square modulus of E. The square modulus of a complex
ix
2 2
number a + ib is a + b , which is why the i disappears. Also e = 1 for

any real x.
P INGBACKS
Pingback: Microwave shielding for perfect transmission
Pingback: Transmission coefficients from water through glass into air
MICROWAVE SHIELDING FOR PERFECT TRANSMISSION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.35.
Heres a simple example of the transmission coefficient for waves passing
from medium 1 through medium 2 into medium 3. Suppose we want to
protect a microwave antenna from the weather by enclosing it in a plastic
shield, where the dielectric constant of the plastic is 2.5. If the antenna
radiates at 10 GHz, what is the best (that is, smallest) thickness of plastic?
The transmission coefficient is given by
"  2 #
2 n2 n2



1 n 2 d n n
(1) T 1 = (n1 + n3 )2 + sin2 1 2 3 2
4n1 n3 c n22

where n j is the index of refraction of medium j, given by


r

(2) n=
0 0
If = 0 then


r
2
(3) n2 = = r2 = 2.5
0

where r2 is the dielectric constant of the plastic.


We can see that if the sine is zero and n1 = n3 (which were assuming
here, since both mediums 1 and 3 are air with n = 1), then T = 1 and we
get perfect transmission. The smallest thickness d is such that

n2 d
(4) =
c
2d 2.5
(5) =
c
c
(6) d =
2 2.5
(7) = 9.49 103 m
1
MICROWAVE SHIELDING FOR PERFECT TRANSMISSION 2

A plastic shield of about 9.5 mm thickness would allow perfect transmis-


sion at that frequency.
TRANSMISSION COEFFICIENTS FROM WATER THROUGH
GLASS INTO AIR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.36.
Heres another simple example of the transmission coefficient for waves
passing from medium 1 through medium 2 into medium 3. Suppose we
have an aquarium filled with water (index of refraction n1 = 43 ). Light from
the aquarium passes (at normal incidence) through a sheet of glass (n2 = 32 )
and into air (n3 = 1).
The transmission coefficient is given by
"  2 #
n1 n22 n23 n22
 
1 1 2 2 n2 d
(1) T = (n1 + n3 ) + sin
4n1 n3 c n22

The minimum value of T occurs when the sine is 1:


" #
n21 n22 n23 n22

1 1
(2) Tmin = (n1 + n3 )2 +
4n1 n3 n22
(3) Tmin = 0.935
The maximum is when the sine is zero:

4n1 n3
(4) Tmax =
(n1 + n3 )2
(5) = 0.980
It doesnt matter much what the frequency of the light is or how thick the
glass is; most of the light makes it through in any case. Since 1 is symmetric
in n1 and n3 a fish inside the aquarium can see out as easily as we can see
in.

1
TOTAL INTERNAL REFLECTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.37.
When we looked at the behaviour of waves passing from one medium to
another at an angle, one of the consequences was Snells law of refraction
which says

sin I n2
(1) =
sin T n1

where I is the angle of incidence (angle between the wave vector and the
normal to the surface) in medium 1 with index of refraction n1 , and T is
the angle of the refracted wave in medium 2. If n1 > n2 , that is, the wave
is incident from a medium (such as water) with a higher index of refraction
than medium 2 (such as air), then we can reach a critical incident angle c
where the refracted angle is /2, so that the refracted wave moves parallel
to the interface. This happens when

n2
(2) sin c =
n1
If I > c , no wave is transmitted through the interface and the entire
wave is reflected back into medium 1. This is known as total internal re-
flection.
To see what happens in this case, we can follow through the same deriva-
tion as before, except we allow the wave vector kT of the transmitted wave
to be complex. That is, for a given T we say that (assuming that the inci-
dent and transmitted plane is the xz plane):

(3) kT = kT (sin T x + cos T z)


The wave vector has, as usual, the magnitude of

n2
(4) kT = =
v2 c
Now suppose I > c . In that case,
1
TOTAL INTERNAL REFLECTION 2

n1 n1 n2
(5) sin T = sin I > =1
n2 n2 n1
q
(6) cos T = 1 sin2 T
q
(7) = i sin2 T 1
s
n21 2
(8) =i sin I 1
n22
s !
n1 n 2
(9) kT = kT sin I x + i 1
2
sin2 I 1z
n2 n2
s !
n2 n1 n21 2
(10) = sin I x + i sin I 1z
c n2 n22
(11) = kx + i z

where

n1
(12) k sin I
cq

(13) n21 sin2 I n22
c
That is, we venture into the realm of complex variables, and T can no
longer be interpreted as a geometric angle. However, lets proceed with the
analysis and see what happens.
First, we look at the electric field, which has the general form

(14) ET (r,t) = E0T ei(kT rt)


Plugging in 11, we get

(15) ET (r,t) = E0T ez ei(kxt)


That is, the transmitted wave propagates in the x direction (parallel to the
interface) and is attenuated in the z direction.
How much of the wave is reflected in this case? For a wave with polar-
ization parallel to the incident plane (that is, E has only an x component),
we found earlier That the reflected amplitude is
TOTAL INTERNAL REFLECTION 3


(16) ER = EI
+

where

cos T
(17)
cos I
1 v1 1 n2
(18) =
2 v2 2 n1
In this case, is purely imaginary and is real, so the reflection coeffi-
cient is

2
ER
(19) R = = =1
EI +

since

(20) | |2 = ||2 + 2 = | + |2

For perpendicular polarization, we have

1
(21) ER = EI
1 +

and again, since 1 is real and is purely imaginary

1 2

(22) R = =1
1 +
Thus the reflection is indeed total for both polarizations.
Still with perpendicular polarization, the electric field is entirely in the y
direction so

(23) ET = E0 ez ei(kxt) y
n2
The magnetic field is given by (using 11 and kT = c kT and v2 = c/n2 )
TOTAL INTERNAL REFLECTION 4

1
(24) BT = kT ET
v2
1 c
(25) = E0 ez ei(kxt) (kz i x)
v2 n2
E0 z i(kxt)
(26) = e e (kz i x)

Taking the real parts to get the actual fields, we have

(27) ET = E0 ez cos (kx t) y


E0 z
(28) BT = e (k cos (kx t) z + sin (kx t) x)

To check that these fields satisfy Maxwells equations requires grinding
away at some derivatives, which we can do in Maple. The divergences are
fairly easy and we get

(29) ET = BT = 0
The curls give

(30) ET = E0 ez (k sin (kx t) z + cos (kx t) x)


BT
(31) =
t
E0 z
sin (kx t) y k2 2

(32) BT = e

n2
(33) = 22 E0 ez sin (kx t) y
c
1 ET
(34) = 2
v2 t
Thus all four of Maxwells equations are satisfied.
The Poynting vector is

1
(35) S= ET BT

E 2 e2z
= 0 k cos2 (kx t) x sin (kx t) cos (kx t) z

(36)

TOTAL INTERNAL REFLECTION 5

Integrating this over one cycle (t = 0 to 2/) gives zero for the z com-
ponent, thus no energy is transmitted perpendicular to the interface, and all
the energy flows in the x direction, parallel to the interface.
RECTANGULAR RESONANT CAVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 9.38.
A wave guide consisting of a completely enclosed volume is known as a
resonant cavity. The simplest resonant cavity is created by taking a rectan-
gular wave guide and closing off the ends to form a rectangular box with
dimensions of a in the x direction, b in the y direction and d in the z direc-
tion. To find the fields, we cant just assume that the wave propagates in
the z direction with the same z dependence for all three components of each
field, that is, we cant take the waves to be

(1) E = E0 (x, y) ei(kzt)


(2) B = B0 (x, y) ei(kzt)
This time, there has to be an explicit dependence on z in E0 and B0 , so
we take the waves to be

(3) E = E0 (x, y, z) eit


(4) B = B0 (x, y, z) eit
We can apply the two curl Maxwell equations to get

B
(5) E =
t
(6) = i B
1 E
(7) B = 2
c t

(8) = i 2 E
c
As the curl affects only the spatial coordinates, we can cancel off eit
from both sides of these equations to get

(9) E0 = i B0

(10) B0 = i 2 E0
c
1
RECTANGULAR RESONANT CAVITY 2

Taking the curl of the first of these equations, we get

= E0 2 E0
 
(11) E0
(12) = 2 E0
(13) = i B0
2
(14) = E0
c2
Therefore we get

2
(15) 2 E0 = E0
c2

giving the same differential equation for each component. We can use sepa-
ration of variables to solve them. For the x component, let Ex = X (x)Y (y) Z (z).
Then

2
(16) X 00Y Z + XY 00 Z + XY Z 00 = XY Z
c2
X 00 Y 00 Z 00 2
(17) + + = 2
X Y Z c
As usual, each term on the LHS must be equal to a constant, and the
solution of the resulting 3 equations is a sum of a sine and a cosine, so we
have

(18) X (x) = A sin kxx x + H cos kxx x


(19) Y (y) = C sin kxy y + D cos kxy y
(20) Z (z) = F sin kxz z + G cos kxz z

with similar equations for the other components of E. The double subscript
such as kxy means that this k belongs to Ex in the function Y (y).
The boundary conditions are

(21) Ek = 0
(22) B = 0

at all boundaries, so we can use this to constrain the ki s. Ex must be zero at


y = 0, b and z = 0, d which means
RECTANGULAR RESONANT CAVITY 3

(23) D = G=0
n
(24) kxy =
b
`
(25) kxz =
d
We cant, at this stage, put any constraints on kxx since it doesnt figure in
any of the boundary conditions. Putting it all together, and condensing the
constants into X (x), we get

n `
(26) Ex = (A sin kxx x + H cos kxx x) sin
y sin z
b d
For Ey we get the same solutions 18, 19 and 20 as above. Ey must be zero
at x = 0, a and z = 0, d so we get

(27) H = G=0
m
(28) kyx =
a
`
(29) kyz =
d
This gives

m `
(30) Ey = (C sin kyy y + D cos kyy y) sin
x sin z
a d
Finally, for Ez the boundary conditions require it to be zero at x = 0, a
and y = 0, b so we get

(31) H = D=0
m
(32) kzx =
a
n
(33) kzy =
b
This gives

m n
(34) Ez = (F sin kzz z + G cos kzz z) sin
x sin y
a b
Now we can invoke Gausss law in vacuum, which states that E = 0.
This gives
RECTANGULAR RESONANT CAVITY 4

n `
(35) E = kxx (A cos kxx x H sin kxx x) sin y sin z +
b d
m `
kyy (C cos kyy y D sin kyy y) sin x sin z +
a d
m n
kzz (F cos kzz z G sin kzz z) sin x sin y
a b
(36) = 0
This equation must be true for all values of (x, y, z) so if we choose x = 0
we get A = 0, or if y = 0 then C = 0, or if z = 0 then F = 0, so we have

n `
(37) E = kxx H sin kxx x sin y sin z +
b d
m `
kyy D sin kyy y sin x sin z +
a d
m n
kzz G sin kzz z sin x sin y
a b
(38) = 0
We should now be able to conclude that arguments of the sine functions
for each variable are equal, that is, that kxx = m a and so on. I havent
been able to find a proof of this, although it seems to have something to do
with Fourier analysis. The argument is along the lines of: the only way an
expansion of sines can be zero for all points is if they are all the same sine
and they add up to zero identically. Given that, we get

 
m n ` m n `
(39) H + D + G sin x sin y sin z = 0
a b d a b d
m n `
(40) H + D+ G = 0
a b d
We can therefore simplify the notation by defining

m
(41) kx =
a
n
(42) ky =
b
`
(43) kz =
d
So the electric field is
RECTANGULAR RESONANT CAVITY 5

(44) E = Heit cos kx x sin ky y sin kz zx + Deit sin kx x cos ky y sin kz zy+
Geit sin kx x sin ky y cos kz zz
The magnetic field can be found from the Maxwell equation

B
(45) E =
t
(46) = (Gky Dkz ) eit sin kx x cos ky y cos kz zx +
(Hkz Gkx ) eit cos kx x sin ky y cos kz zy +
(Dkx Hky ) eit sin kx x cos ky y cos kz zz
Integrating with respect to time gives

i
(47) B = (Gky Dkz ) eit sin kx x cos ky y cos kz zx

i
(Hkz Gkx ) eit cos kx x sin ky y cos kz zy

i
(Dkx Hky ) eit sin kx x cos ky y cos kz zz

In all cases, 17 requires that

2
(48) kx2 + ky2 + kz2 =
c2

so the resonant frequency for mode mn` is


s
 m 2  n 2  ` 2
(49) mn` = c + +
a b d
s
 m 2  n 2  ` 2
(50) = c + +
a b d
MAXWELLS EQUATIONS IN TERMS OF POTENTIALS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.1.
Maxwells equations are


(1) E =
0
(2) B = 0
B
(3) E =
t
E
(4) B = 0 J + 0 0
t
The are completely general, in that they describe the interactions between
electric and magnetic fields and the charges and currents that generate them,
in both the static (fixed charge and current densities) and dynamic (time-
varying) cases. In the static case, the two time derivatives are zero, which
means that E = 0. Since any vector field whose curl is zero can be
represented as the gradient of a scalar field, we could write E = V where
V is the potential. The presence of the term tB , however, means that we
cant, in general, write E as the gradient of a scalar field in the dynamic
case.
Because B = 0 we can write B as the curl of a vector field

(5) B = A
This is still true in the dynamic case. Substituting this into 3 we get

A
(6) E =
  t
A
(7) E+ = 0
t
The LHS is the curl of a vector field that combines the electric field and
the magnetic vector potential, and since it is zero even in the dynamic case,
we can write this vector field as the gradient of a scalar field V . [We should
be able to use the same symbol V for the potential in the dynamic case,
1
MAXWELLS EQUATIONS IN TERMS OF POTENTIALS 2

since if tA = 0 we regain the static equations E = 0 and E = V .]


That is, we can write

A
(8) E+ = V
t
A
(9) E = V
t
Equations 5 and 9 effectively replace 2 and 3. Using 5 and 9 we can
eliminate E and B from 1 and 4:


(10) E = 2V ( A)
t

(11) 2V + ( A) =
t 0
(12) B = ( A)
E
(13) = 0 J + 0 0
t
V 2A
(14) ( A) = 0 J 0 0 0 0 2
t t
We can rearrange the last equation using the identity

(15) ( A) = ( A) 2 A

so we get

V 2A
(16) ( A) 2 A = 0 J 0 0 0 0 2
t t
(17)
2A
 
2 V
A 0 0 2 A + 0 0 = 0 J
t t
To summarize:


(18) 2V + ( A) =
t 0

2A
 
2 V
(19) A 0 0 2 A + 0 0 = 0 J
t t
MAXWELLS EQUATIONS IN TERMS OF POTENTIALS 3

These two equations comprise 4 equations (one from 18 and one for each
vector component in 19) for four functions (V and A), and their solution
allows us to calculate both E and B by means of 9 and 5, so they form a
complete replacement for the original set of 4 Maxwell equations that we
started with. This reduces the number of equations to be solved from 6 (for
the 3 components of each of E and B) to 4.
We can write these equations in a more compact form using the dAlembertian
operator

2
(20) 2 2 0 0
t 2

and defining the function

V
(21) L A + 0 0
t
We get

2 2 2V
(22) V = V 0 0 2
t
L
(23) = 2V + ( A)
t t
L
(24) 2V + = 2V + ( A)
t t

(25) =
0

using 18 in the last line. Also,

2A
(26) 2 A = 2 A 0 0
t 2
(27) 2 A L = 0 J

using 19. In summary

L
(28) 2V + =
t 0

(29) 2 A L = 0 J
MAXWELLS EQUATIONS IN TERMS OF POTENTIALS 4

P INGBACKS
Pingback: Energy flow in time-dependent fields
Pingback: Potentials for a point charge
Pingback: Potentials for an electromagnetic wave
Pingback: Jefimenkos equation for time-dependent electric field
Pingback: Fields of a moving point charge
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
ENERGY FLOW IN TIME-DEPENDENT FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.2.
Suppose we have time-dependent electric and magnetic potentials given
by

(1) V =0

(
0 k 2
(2) A= 4c (ct |x|) z for |x| < ct
0 for |x| > ct

where k is a constant and c is the speed of light.


We can get the fields by using our earlier formulas

A
(3) E = V
t
(4) B = A
We get, for |x| < ct

0 k
(5) E = (ct |x|)
2
0 k
(6) B = (ct |x|)2 y
4c x
0 k
(7) = (ct |x|) y
2c

where in the last equation the + applies to x > 0 and the to x < 0. For
|x| > ct, all fields are zero.
Griffiths shows in his example 10.1 that these fields satisfy Maxwells
equations and arise from a surface current flowing in the x = 0 plane given
by

(8) K = kt z
1
ENERGY FLOW IN TIME-DEPENDENT FIELDS 2

for t 0. That is, the current starts at t = 0 and its effects travel outwards
in the x direction at the speed c. What well do here is calculate the energy
flow due to this current. Well consider a rectangular box of width w in the
y direction and length ` in the z direction extending from x = d to x = d + h.
At time t1 = d/c the energy is zero inside the box since the fields from
the current have only just reached the lower face of the box at x = d. At
time t2 = (d + h) /c, the fields have reached the outer face of the box so the
entire box is filled with fields. At this time the energy within the box can be
calculated from the energy density

(9)
 
1 2 3 1 2 3
U= 0 E d r + B d r
2 0
(10)
"   2 d+h #
0 k 2 d+h

w` 1 0 k
= 0 (d + h x)2 dx + (d + h x)2 dx
2 2 d 0 2c d
(11)

w`0 k2 d+h
= (d + h x)2 dx
4c2 d
(12)

w`0 k2 h 2
= v dv
4c2 0
(13)
w`0 k2 h3
=
12c2

where in the second line we used 0 0 = 1/c2 and in the third line we used
the substitution v = d + h x in the integral.
Now we can calculate the flow of energy into the box as a function of
time using the Poynting vector which gives the rate per unit area at which
energy crosses a surface. We have for the surface x = d

1
(14) S = EB
0
1 02 k2
(15) = (ct d)2
0 4c
The total rate at which energy flows into the box via this face is thus
ENERGY FLOW IN TIME-DEPENDENT FIELDS 3

dU w`0 k2
(16) = (ct d)2
dt 4c

and the total energy that flows into the box between times t1 = d/c and
t2 = (d + h) /c is
(d+h)/c
w`0 k2
(17) U = (ct d)2 dt
4c d/c
w`0 k 2 h
(18) = 2
v2 dv
4c 0
2
w`0 k h 3
(19) =
12c2

which of course agrees with our earlier calculation of the energy in the box.
[We used the substitution v = ct d in the first line.]
P INGBACKS
Pingback: Coulomb and Lorenz gauges
POTENTIALS FOR A POINT CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.3.
As a simple (though unusual) example of specifying a system through
electric and magnetic potentials suppose we have

(1) V (r,t) = 0
1 qt
(2) A (r,t) = r
40 r2
These potentials give rise to the fields

(3) B = A = 0
A 1 q
(4) E = V = r
t 40 r2
The expression for E is just that of a point charge q at the origin, while
a zero magnetic field indicates that there is no current. If we want to be
pedantic, we can also get these results from the potentials. For the charge
density, we had


(5) 2V + ( A) =
t 0
Using the formula
 
1
(6) 2 r = 43 (r)
r

we get

qt
(7) A = 3 (r)
0
(8) = q3 (r)
For the current, we can use the equation
1
POTENTIALS FOR A POINT CHARGE 2

V 2A
(9) ( A) = 0 J 0 0 0 0 2
t t

and we see that all the terms not involving J are zero, so J = 0 as well. Thus
this is a bizarre way of writing potentials for a point charge. This illustrates
that the potentials giving rise to a particular charge and current distribution
are not unique.
P INGBACKS
Pingback: Gauge transformations in electrodynamics
Pingback: Coulomb and Lorenz gauges
POTENTIALS FOR AN ELECTROMAGNETIC WAVE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.4.
We can define an electromagnetic wave in terms of electric and magnetic
potentials as follows. Using rectangular coordinates, let

(1) V = 0
(2) A = A0 sin (kx t) y
These potentials give rise to the fields

(3) B = A = A0 k cos (kx t) z


A
(4) E = V = A0 cos (kx t) y
t
We can check that these fields satisfy Maxwells equations in vacuum.
First,

(5) E = B = 0
For the curls, we have

B
(6) E = A0 k sin (kx t) z =
t
(7) B = A0 k2 sin (kx t) y

The second equation should be equal to 0 0 tE = 1 E


c2 t
so

1 E
(8) A0 k2 sin (kx t) y =
c2 t
1
(9) = 2 A0 2 sin (kx t) y
c

which can be true only if


1
POTENTIALS FOR AN ELECTROMAGNETIC WAVE 2

2
(10) k2 =
c2

which is the usual relation between wave number k and angular frequency
.
P INGBACKS
Pingback: Coulomb and Lorenz gauges
GAUGE TRANSFORMATIONS IN ELECTRODYNAMICS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.5.
The potentials V and A that give rise to a particular configuration of fields
E and B are not unique. For example, for a point charge at the origin, the
usual potentials would be

q
(1) V =
40 r
(2) A = 0

but as weve seen, the potentials

(3) V (r,t) = 0
1 qt
(4) A (r,t) = r
40 r2

give rise to the same fields. The question arises: what is the most general
set of potentials that give a particular configuration of fields? That is, what
changes can we make to V and A without changing E and B?
The fields can be calculated from the potentials via the equations

(5) B = A
A
(6) E = V
t
We cant multiply V and A by anything and leave the fields unchanged,
but we might be able to add something. Suppose we add a vector function
(r,t) to A and a scalar function (r,t) to V :

(7) A0 = A +
(8) V0 = V +
To get the same magnetic field, we must have
1
GAUGE TRANSFORMATIONS IN ELECTRODYNAMICS 2

(9) B = A
(10) = A0
(11) = (A + )
(12) = 0
Since its curl is zero, we can write as the gradient of a scalar field:

(13) (r,t) = 0 (r,t)


We also have to get the same electric field E so

A
(14) E = V
t
(A + )
(15) = (V + )
t

(16) + = 0
t
Combining these results, we get

0
 
(17) + =0
t
The term in parentheses must not depend on position (since its gradient
is identically zero everywhere), but it might depend on time. Define this
term to be k (t); then

0
(18) = k (t)
t
If we define
t
0
k t 0 dt 0

(19)
0

then


(20) =
t
0
Since and differ only by a function of time (and not of space), their
gradients are equal, so we can use in place of 0 in 13 without changing
GAUGE TRANSFORMATIONS IN ELECTRODYNAMICS 3

. Therefore we can change the potentials as follows, without changing the


fields they produce:

(21) A0 = A +

(22) V0 = V
t

where = (r,t) is an arbitrary scalar field. Such a change to the potentials


is called a gauge transformation.
Example. Earlier, we saw the unusual potentials 3 and 4 for a point charge
at the origin. We can transform them using the gauge function

qt
(23) =
40 r
We get

qt
(24) = r
40 r2
q
(25) =
t 40 r

so the new potentials are

q
(26) V =
40 r
(27) A = 0

which is the more usual set of potentials 1 and 2 for a point charge.
P INGBACKS
Pingback: Coulomb and Lorenz gauges
Pingback: Relativistic electromagnetic potentials
COULOMB AND LORENZ GAUGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.6.
Two common gauges in electrodynamics are the Coulomb gauge and the
Lorenz gauge. [Note that Griffiths calls the latter the Lorentz gauge, al-
though the gauge was introduced by the Danish physicist Ludvig Lorenz.
My thanks to Gerrit Jan van Dijk for pointing this out.] Each gauge amounts
to specifying a value for A. The Coulomb gauge sets A = 0, and is the
gauge we used when introducing the magnetic vector potential. To see that
its always possible to transform from an arbitrary gauge to the Coulomb
gauge, we need to find a function such that the transformation

(1) A = A0 +

(2) V = V0
t

gives A = 0. To do this, we must have

(3) A = 0 = A0 + 2
(4) 2 = A0

The last line is just Poissons equation and for any reasonable original
potential A0 it is possible to solve it. In the Coulomb gauge, the potential
forms of Maxwells equations:


(5) 2V + ( A) =
t 0
2
A

V

(6) 2 A 0 0 2 A + 0 0 = 0 J
t t

reduce to
1
COULOMB AND LORENZ GAUGES 2


(7) 2V =
0
2A
 
2 V
(8) A 0 0 2 = 0 J + 0 0
t t
The scalar potential V is therefore also a solution of Poissons equation,
and once we have found it, we can, in principle, solve the second equation
(which is a wave equation with a complicated driving term on the RHS) for
the vector potential A.
The Lorenz gauge sets

V
(9) A = 0 0
t

which transforms 5 and 6 to

2V
(10) 2V 0 0 2
=
t 0
2
A
(11) 2 A 0 0 2 = 0 J
t
In terms of the dAlembertian operator

2 2 2
(12)  0 0 2
t

we can write these two equations as


(13) 2V =
0
2
(14)  A = 0 J

so that both potentials now become solutions of the wave equation with a
driving term, but now V and A are decoupled.
Example 1. For the potentials

(15) V =0
COULOMB AND LORENZ GAUGES 3

(
0 k 2
(16) A= 4c (ct |x|) z for |x| < ct
0 for |x| > ct

we have

(17) A = 0

and because V = 0,

V
(18) A = 0 0
t
These potentials satisfy the conditions for both the Coulomb and Lorenz
gauges.

Example 2. For the potentials

(19) V (r,t) = 0
1 qt
(20) A (r,t) = r
40 r2

we have

qt
(21) A = 3 (r)
0

so it uses neither the Coulomb nor Lorenz gauge.

Example 3. For the potentials

(22) V = 0
(23) A = A0 sin (kx t) y

we again have A = 0 so it satisfies both the Coulomb and Lorenz gauges.


COULOMB AND LORENZ GAUGES 4

P INGBACKS
Pingback: Lorenz gauge is always possible
Pingback: Retarded potentials
Pingback: Relations among charge, current, potentials and fields
Pingback: Point charge moving at constant velocity satisfies Lorenz gauge
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
Pingback: Relativistic electromagnetic potentials
LORENZ GAUGE IS ALWAYS POSSIBLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.7.
The Lorenz gauge is defined by setting

V
(1) A = 0 0
t
but is it always possible to do this? We can show that it is using a similar
technique to that for the Coulomb gauge. We want a function which
we can use to transform some arbitrary potentials A0 and V so that that A
satisfies 1 as follows:

(2) A = A0 +

(3) V = V0
t
Taking the divergence of both sides, we get

(4) A = A0 + 2
V
(5) = 0 0
t
V 0 2
(6) = 0 0 + 0 0 2
t t
Combining the first and last equations we get

2 V 0
(7) 2 0 0 = 0 0 A0
t 2 t
V 0
(8) 2 = 0 0 A0
t
That is, is the solution of the wave equation with a driving term, which
we can, in principle, always solve (although it may not be easy!). Therefore
we can always find the function to convert an arbitrary pair of potentials
A0 and V 0 to the Lorenz gauge.
1
LORENZ GAUGE IS ALWAYS POSSIBLE 2

In general (not necessarily in the Lorenz gauge), we could always set


V = 0 by choosing


(9) = V0
t
t
V 0 r,t 0 dt 0

(10) =
0
We cant always choose A = 0 however, since B = A and if the
magnetic field is non-zero, then A cant be zero everywhere.
RETARDED POTENTIALS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Section 10.2.1. Problem 10.8.
In electrostatics and magnetostatics, charge distributions and currents
were all constant in time. When they vary, we need to take into account
the finite speed of light in calculating potentials and fields. If we want the
fields at some point P then, if the charge or current changes at some point
Q a distance d from P, an observer at P wont know about the change until
the signal from Q reaches him, which in vacuum takes a time d/c. To take
account of this, the potentials at position r and time t in a dynamic system
are taken to be

1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d

where

d
(3) tr t
c

and

d r r0

(4)
q
(5) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(6) d =
d
That is, each potential is the sum over all locations where there is charge
or current, and each location r0 is sampled at the time tr in the past which
is the time a light signal would have left r0 to arrive at r at time t. These
potentials are called retarded potentials, since they depend on the situation
at various times in the past to get the fields at the present time.
1
RETARDED POTENTIALS 2

Griffiths shows in his section 10.2.1 that these potentials (well V anyway;
the argument for A is similar) satisfy the wave equations in the Lorenz
gauge


(7) 2V =
0
2
(8)  A = 0 J
We also need to show that the potentials satisfy the Lorenz gauge condi-
tion

V
(9) A = 0 0
t
Starting from 2 we need to find

J (r0 ,tr ) 3 0
 
0
(10) A = d r
4 d
Note that the is a derivative with respect to r (the observers position)
and not r0 (the source positions and variable of integration), and that both tr
and d depend on both r and r0 . We begin by writing
   
J 1 1
(11) = J+J
d d d
We can also use the derivative with respect to r0 :
   
J0 1 0 0 1
(12) = J+J
d d d
We have (you can work this out by using 5 if you dont believe me):
   
1 d 0 1
(13) = 2 =
d d d
Therefore
   
0 J 1 0 1
(14) = JJ
d d d
   
1 1 0 J
(15) J = J 0
d d d
Inserting this into 11 we get
RETARDED POTENTIALS 3

   
J 1 1 0 0 J
(16) = J+ J
d d d d
Now we need to work out the two divergences J and J0 . To do
this, we need to remember that J = J (r0 ,tr ), so it depends on r only via tr
but it depends on r0 both explicitly through its first argument and implicitly
through tr . Using the chain rule, we get for the contribution from x

Jx tr
(17) ( J)x =
tr x
1 d
(18) = Jx
c x
1
(19) = Jx (d)x
c

where the dot over the J is a derivative with respect to t, which is the same
as a derivative with respect to tr since tr = t d/c and d doesnt depend on
time.
The other two coordinates give similar results and we get

1
(20) J = J d
c
For the other divergence, things are a bit trickier since J depends explic-
itly on r0 . Here we used the extended chain rule

g (x, f (x)) g g f
(21) = +
x x f x
Therefore

1
(22) 0 J = 0r0 J J 0 d
c

where weve used the specialized notation 0r0 to indicate the divergence
with respect to the explicit r0 dependence in J. From Maxwells fourth
equation

E
(23) 0r0 B = 0 J + 0 0
t
RETARDED POTENTIALS 4

where the fields and currents depend on r0 , we can take the explicit diver-
gence to get

0 0 E
0r0 0r0 B = 0 0r0 J + 0 0 r

(24)
t
The divergence of a curl is always zero, so we get

0r0 E
(25) 0r0 J = 0
t
(26) =

where weve used Maxwells first equation


(27) 0r0 E =
0
Plugging this into 22 we get

1
(28) 0 J = J 0 d
c
1
(29) = + J d
c

since from 5

(30) 0 d = d
Putting 20 and 29 into 16 we get
     
J 1 1 1 0 J
(31) = J d + J d
d d c c d
 
J
(32) = 0
d d
Finally, from 2 we have

  
0 0 J
(33) A (r,t) = d 3 r0
4 d d
RETARDED POTENTIALS 5

Using the divergence theorem, the second term can be converted to a


surface integral at infinity where (presumably) the current J is zero, so this
term vanishes. Using 1 we then get

0 3 0
(34) A (r,t) = d r
4 d
V
(35) = 0 0
t

which is the Lorenz gauge condition, as required.


P INGBACKS
Pingback: Retarded potentials in an infinite straight wire
Pingback: Retarded potential of a wire loop
Pingback: Jefimenkos equation for time-dependent electric field
Pingback: Jefimenkos equation for time-dependent magnetic field
Pingback: Linard-Wiechert potentials for a moving point charge
Pingback: Retarded potentials for a sinusoidal current loop
Pingback: Relations among charge, current, potentials and fields
Pingback: Point charge moving at constant velocity satisfies Lorenz gauge
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
Pingback: Fields of an oscillating magnetic dipole
Pingback: Electric dipole radiation from an arbitrary source
Pingback: Radiation from a current loop with time-varying current
RETARDED POTENTIALS IN AN INFINITE STRAIGHT WIRE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.9.
Here are a couple of examples of calculating the retarded potential. In
both examples, we have an infinite straight wire that carries a current. At
time t = 0 a current I (t) is switched on, and we want to find the electric and
magnetic fields after that time. Well assume the wire is electrically neutral
so there is no free charge density and the scalar potential is thus V = 0.
Example 1. The current is linearly increasing:

(1) I (t) = kt
For a linear current the vector potential is (assuming the current flows in
the z direction):


0 I (r0 ,tr )
(2) A (r,t) = z dz
4 d

where

d
(3) tr t
c

and

p
(4) d = r2 + z2

where were using cylindrical coordinates, so r is the perpendicular distance


from the wire and z is the distance along the wire.
q d < ct will
At time t and distance r, only those parts of the wire where
contribute to the potential, so the integrals limits are z = (ct)2 r2 and
we get
1
RETARDED POTENTIALS IN AN INFINITE STRAIGHT WIRE 2

(ct)2 r2
0 k t r2 + z2 /c
(5) A = z dz
4 (ct)2 r2 r2 + z2
(ct)2 r2  
0 k ct
(6) = z 2 1 dz
4c 0 r2 + z2
q
2 2
0 k ct + (ct) r q
(7) = z ct ln (ct)2 r2
2c r

From this we can find E and B using the equations

A
(8) E = V
t
(9) B = A
The derivatives are messy and best done with Maple. The results are
q
2
0 k ct + (ct) r2
(10) E = ln z
2 r
q 
2 2 2 2
0 k (ct) r ct + (ct) r
(11) B =  q 
2 2
2rc ct + (ct) r
q
0 k (ct)2 r2
(12) =
2rc
Example 2. This time the current is an impulse at t = 0 given by

(13) I (t) = q0 (t)


The argument above still applies; only the integrand is different. We get

0 I (r0 ,tr )
(14) A (r,t) = z dz
4 d

(ct)2 r2 t r2 + z2 /c

0 q0
(15) = z dz
2 0 r 2 + z2
We can transform the integral using the substitution
RETARDED POTENTIALS IN AN INFINITE STRAIGHT WIRE 3


r2 + z2
(16) u = t
p c
(17) r2 + z2 = ct cu
q
(18) z = (ct cu)2 r2
c (ct cu)
(19) dz = q du
2 2
(ct cu) r
The limits transform as

r
(20) z=0 u=t
q c
(21) z= (ct)2 r2 u = 0
The integral becomes
t cr
0 q0 c (ct cu) (u)
(22) A (r,t) = z q du
2 0 2 2
(ct cu) (ct cu) r
c q
(23) = z q 0 0
2 (ct)2 r2
The corresponding fields are

0 q0 c3t
(24) E = 3/2
z
2 (c2t 2 r2 )
0 q0 cr
(25) B = 3/2

2 (c2t 2 r2 )
RETARDED POTENTIAL OF A WIRE LOOP

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.10.
One more example of calculating the retarded potential. We have a loop
of wire in the following shape. It extends along the x axis from b to a,
then in a semicircular loop of radius a clockwise around to x = +a, then
along the x axis from +a to +b, then in a semicircular loop of radius b back
to x = b. A linearly increasing current

(1) I (t) = kt

flows through the loop in the direction given above. Assuming the wire is
electrically neutral, V = 0 so our job is to find A.
Calculating A in general is a complex task, so well look only at the value
of A at the origin. Consider first the inner loop of radius a. All points on
this loop are at the same distance a from the origin, so the retarded time is
the same for all points on the loop. Since the current goes clockwise around
the semicircle, the contribution to A is

0 0 k t ac

(2) Aa = a d
4 a

0  a 
(3) = k t ( sin x + cos y) d
4 c 0
0  a 
(4) = k t x
2 c
We get a similar expression for the loop around the outer semicircle ex-
cept this time the current flows counterclockwise so the sign is reversed:
 
0 b
(5) Ab = k t x
2 c
Adding these two together we get

0 b a
(6) Aab = k x
2 c
1
RETARDED POTENTIAL OF A WIRE LOOP 2

The contributions from each of the two horizontal segments are equal, so
for these two segments we have
b
0 t xc
(7) Ax = 2 kx dx
4 a x
 
0 b ba
(8) = kx t ln
2 a c
The total potential is then

(9) A (0,t) = Aab + Ax


0 b
(10) = kxt ln
2 a
Because we have the potential at only a single point in space, we cant
calculate any of its derivatives, so we cant calculate B = A. However
we can calculate E:

A
(11) E =
t
0 b
(12) = kx ln
2 a
The electric field is constant in time at the origin. An electrically neu-
tral wire can produce an electric field since the changing current induces a
changing magnetic field which in turn produces an electric field.
JEFIMENKOS EQUATION FOR TIME-DEPENDENT ELECTRIC
FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.11.
Using the retarded potentials, we can find a time-dependent expression
for the electric field E. The potentials are

1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d

where

d
(3) tr t
c

and

d r r0

(4)
q
(5) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(6) d =
d
The expression for E is given by

A
(7) E = V
t
The derivatives are complicated by the fact that the integrands in 1 and
2 depend on r both via tr and d. We get (note that indicates derivatives
with respect to components of r only, not r0 ):
1
JEFIMENKOS EQUATION FOR TIME-DEPENDENT ELECTRIC FIELD 2


(r0 ,tr ) 3 0

1
(8) V = d r
40 d
 
1 1 1
(9) = + d 3 r0
40 d d
Using the chain rule


(10) = tr
tr
1
(11) = d
c t

(12) = d
c

since tr = t because of 3. By direct calculation we have

(13) d = d
 
1 d
(14) = 2
d d
Plugging everything into 9 we get
 
1
(15) V (r,t) = + dd 3 r0
40 d 2 cd
The second term in 7 is just


A 0 J (r0 ,tr ) 3 0
(16) = d r
t 4 d

so the time-dependent field is (using 0 0 = 1/c2 ):


J (r0 ,tr ) 3 0

1 3 0 0
(17) E (r,t) = + dd r d r
40 d 2 cd 4 d

(r0 ,tr ) (r0 ,tr ) J (r0 ,tr ) 3 0

1
(18) = d + d d r
40 d2 cd c2 d
This is Jefimenkos equation for the electric field. In the static case, all
time derivatives are zero and there is no dependence on tr so we get
JEFIMENKOS EQUATION FOR TIME-DEPENDENT ELECTRIC FIELD 3


1 (r0 ) 3 0
(19) E (r) = dd r
40 d2

which is just Coulombs law from electrostatics.


A special case is that of constant current but varying charge. In that case

(20) (r,t) = (r, 0)t + (r, 0)

where

(21) (r, 0) = J (r)


In this case, the integrand of 18 is

(22)
(r0 ,tr ) (r0 ,tr ) J (r0 ,tr ) (r0 , 0) t dc + (r0 , 0) (r0 , 0)

d + d = d + d
d2 cd c2 d d2 cd
(r0 , 0)t + (r0 , 0)
(23) = d
d2
(r0 ,t)
(24) = d
d

so the field is

1 (r0 ,t) 3 0
(25) E (r,t) = dd r
40 d2
That is, Coulombs law is valid with the charge density evaluated at the
current (non-retarded) time.
P INGBACKS
Pingback: Jefimenkos equation for time-dependent magnetic field
Pingback: Relations among charge, current, potentials and fields
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
JEFIMENKOS EQUATION FOR TIME-DEPENDENT
MAGNETIC FIELD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 10.12.
Using the retarded potentials, we can find a time-dependent expression
for the magnetic field B to complement the equation we got earlier for the
electric field. The potentials are

1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d

where

d
(3) tr t
c

and

d r r0

(4)
q
(5) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(6) d =
d
The magnetic field is

(7) B = A

( J (r0 ,tr ))
 
0 0 1
d 3 r0

(8) = J r ,tr
4 d d

where weve used the identity

(9) ( f A) = f ( A) A f
1
JEFIMENKOS EQUATION FOR TIME-DEPENDENT MAGNETIC FIELD 2

We have

 
1 d
(10) = 2
d d
To get J, its easiest to look at a single component of the curl, say the
x component:

Jz Jy
(11) ( J)x =
y z
Using the chain rule

Jz Jz tr
(12) =
y tr y
1 Jz d
(13) =
c t y
1
(14) = Jz (d)y
c
Therefore

1 
(15) ( J)x = Jz (d)y Jy (d)z
c
1 
(16) = J d x
c
The other two components work out the same way so we have

1
(17) J = J d
c
1
(18) = J d
c
Putting this back into 8 we get


J (r0 ,tr ) d

0 0
 d 3 0
(19) B (r,t) = + J r ,tr 2 d r
4 cd d
This is Jefimenkos equation for the magnetic field. For steady currents,
J = 0 and the dependence on tr disappears, so were left with the Biot-Savart
law:
JEFIMENKOS EQUATION FOR TIME-DEPENDENT MAGNETIC FIELD 3


0  d
(20) B (r) = J r0 2 d 3 r0
4 d
In the special case where the current changes slowly enough that we can
approximate it by a first-order Taylor series:

(21) J (r,tr ) = J (r,t) + (tr t) J (r,t)


d
(22) = J (r,t) J (r,t)
c
J
(23) = J (r,t)
tr

we get from 19

J (r0 ,t) 1
 
0 0
 d 0
dd 3 r0

(24) B (r,t) = + 2 J r ,t J r ,t
4 cd d c

0 J (r0 ,t) d 3 0
(25) = d r
4 d2
That is, for slowly varying currents, to first order, Jefimenkos equation
gives the Biot-Savart law where all currents are evaluated at the current
time. Were essentially assuming that the travel time for signals from all
locations of the current to the observation point is zero. This was the qua-
sistatic approximation.
P INGBACKS
Pingback: Relations among charge, current, potentials and fields
Pingback: Electric dipole time-varying potentials; Lorenz gauge condi-
tion
LINARD-WIECHERT POTENTIALS FOR A MOVING POINT
CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 10.13.
We now look at the retarded potentials for a moving point charge q. The
potentials are

1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d

where

d
(3) tr t
c

and

d r r0

(4)
q
(5) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(6) d =
d
The charge density of a point charge is represented by a delta function in
space, so if the charges trajectory is given by w (t 0 ) then

r0 ,t 0 = q 3 r0 w t 0
 
(7)
To work out V , we need the charge density at the retarded time tr , which
we can write as the integral over time of the charge density multiplied by
another delta function:

0 0 0
3
t 0 tr dt 0
  
(8) r ,tr = q r w t
1
LINARD-WIECHERT POTENTIALS FOR A MOVING POINT CHARGE 2

We need to keep straight the different times were using here. The time
t is the observation time, t 0 is the integration variable and tr is the retarded
time, which is the time at which the signal that we are receiving at time t
left the moving charge, which is

|r r0 |
(9) tr = t
c
The potential can now be written as an integral over both time and space:

3 0 0
q 3 0 (r w (t ))
dt 0 t 0 tr

(10) V (r,t) = d r 0
40 |r r |
0 0
3 |r r0 |
  
q 3 0 (r w (t )) 0 0
(11) = d r dt t t
40 |r r0 | c

We can do the spatial integration which sets r0 = w (t 0 )


|r w (t 0 )|
  
40 0 1 0
(12) V (r,t) = dt t t
q |r w (t 0 )| c

The trick now is to transform the argument of the delta function so we can
do the integral. To do this, we need to work out ( f (x)) for some function
f (x). To work this out, we use the substitution

(13) u = f (x)
(14) du = f 0 (x) dx

so we get


(u)
(15) ( f (x)) dx = du
| f 0 (x)|
1
(16) = 0
| f (x (0))|

where we need to solve for x as a function of u from 13 and then find


x (u = 0).
For our problem, we have
LINARD-WIECHERT POTENTIALS FOR A MOVING POINT CHARGE 3

|r w (t 0 )|
 
0 0

(17) f t = t t
c
df 1 d
r w t0

(18) 0
= 1+ 0
dt c dt
Lets select our coordinate axes so that, at time t 0 , w = wx and dw
dt 0 =
+ cx where 0 < < 1. That is, the charge is on the negative x axis and is
moving in the +x direction with a speed c. Then we have

d 0 2

r w t 0 d r w t 0
 
(19) r w t = 2
dt 0 dt 0
d 
r w t0 r w t0
 
(20) = 0
dt
 dw
(21) = 2 r w t 0 0
dt
(22) = 2 c r w t 0 x
d 0
 c (r w (t 0 )) x
(23) r w t =
dt 0 |r w (t 0 )|
(r w (t 0 )) v
(24) =
|r w (t 0 )|

where in the last line v cx is the velocity of the charge. Therefore

df (r w (t 0 )) v
(25) = 1
dt 0 c |r w (t 0 )|
Returning to 12 we have f (t 0 ) = 0 when t 0 = tr so we can do the integral
over the delta function to get

|r w (t 0 )|
  
q 0 1 0
(26) V (r,t) = dt t t
40 |r w (t 0 )| c
q 1
(27) =  
40 |r w (t )| 1 (rw(tr ))v
r c|rw(tr )|
qc 1
(28) =
40 (c |r w (tr )| (r w (tr )) v)
The current density for a moving point charge is just

(29) J = v
LINARD-WIECHERT POTENTIALS FOR A MOVING POINT CHARGE 4

so the derivation of A from 2 follows exactly the same path and we get

0 qc v
(30) A (r,t) =
4 (c |r w (tr )| (r w (tr )) v)
v
(31) = 2 V (r,t)
c
These are the Linard-Wiechert potentials for a moving point charge.
Griffiths gives a heuristic argument as to why the extra term (r w (tr ))
v turns up in the denominator. The effect arises because of the perception of
size of a moving object. If we see a metre stick coming directly at us with
a speed v, we will perceive it to be slightly longer than it actually is, since
the light from the far end of the stick left the stick when it was further away
from us than the light from the near end. Although this argument does give
the right answer and the argument doesnt depend ultimately on the size of
the object approaching, I find the argument unsatisfying when applied to
a point object, since Id still expect that the effect should disappear in that
case. The argument above, using delta functions, is a lot more abstract than
the moving metre stick argument, but at least it shows rigorously how the
effect arises.
Example. We have a point charge q moving in a circle of radius a in the xy
plane at constant angular speed so that its position is given by

(32) w (t) = ax cos t + ay sin t


The velocity is

dw
(33) v (t) = = a x sin t + a y cos t
dt
For an observation point r = zz the retarded time is

|r w (t 0 )| z2 + a2
(34) tr = t =t
c c
This is independent of the charges position, since its always at the same
distance from a point on the z axis. Also by direct calculation

(35) (r w (tr )) v = 0

so the potentials are


LINARD-WIECHERT POTENTIALS FOR A MOVING POINT CHARGE 5

qc 1
(36) V (r,t) =
40 (c |r w (tr )| (r w (tr )) v)
q
(37) =
40 |r w (tr )|
q
(38) =
40 z2 + a2
qa
(39) A (r,t) = (x sin tr + y cos tr )
40 c2 z2 + a2
P INGBACKS
Pingback: Linard-Wiechert potentials for a charge moving with constant
velocity
Pingback: Point charge in hyperbolic motion: visible and invisible points
Pingback: Linard-Wiechert potential for a charge moving on a hyper-
bolic trajectory
Pingback: Fields of a moving point charge
Pingback: Point charge moving at constant velocity satisfies Lorenz gauge
LINARD-WIECHERT POTENTIALS FOR A CHARGE
MOVING WITH CONSTANT VELOCITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 10.14.
Griffiths shows in his example 10.3 that the Linard-Wiechert potentials
for a point charge q moving at constant velocity v that passes through the
origin at time t = 0 are

1 qc
(1) V (r,t) = q
40 2
(c2t r v) + (c2 v2 ) (r2 c2t 2 )
0 qcv
(2) A (r,t) = q
4 2
(c2t r v) + (c2 v2 ) (r2 c2t 2 )
These potentials can be expressed in a simpler form by defining the vector

(3) R r vt
We can eliminate r from 1 as follows.

(4) R v = r v v2t
(5) r v = R v + v2t
(6) R2 = r2 + v2t 2 2r vt
(7) r2 = R2 v2t 2 + 2r vt
(8) = R2 + v2t 2 + 2R vt
2 2
(9) c2t r v = c2t R v v2t
c2 v2 r2 c2t 2 = c2 v2 R2 + v2t 2 + 2R vt c2t 2
   
(10)
Adding the last two RHSs together and cancelling terms, we get

2
c2t r v + c2 v2 r2 c2t 2 = (R v)2 + c2 v2 R2
  
(11)
If is the angle between R and v, then this becomes
1
LINARD-WIECHERT POTENTIALS FOR A CHARGE MOVING WITH CONSTANT VELOCITY
2

(R v)2 + c2 v2 R2 = R2 c2 v2 1 cos2
 
(12)
v2 2
 
2 2
(13) = R c 1 2 sin
c
Inserting this back into 1 we get

1 q
(14) V (r,t) = r
40 2

R 1 vc2 sin2

Note that R and are both functions of time since they vary as the charge
moves. For non-relativistic speeds, v  c and the formula reduces to the
Coulomb potential from electrostatics:

q
(15) V=
40 R
POINT CHARGE IN HYPERBOLIC MOTION: VISIBLE AND
INVISIBLE POINTS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 10.15.
When we work out Linard-Wiechert potentialsfor a moving point charge,
weve been implicitly assuming that we can receive a signal from the charge
from only one retarded time. That is, if the charge is moving on some tra-
jectory w (t), there is only one point on that trajectory where a signal can
be sent such that it will reach us at the time of observation. This assump-
tion actually depends on special relativity with its postulate that nothing can
travel faster than light. To see how this comes about, suppose there were, in
fact, two different points P1 and P2 on the trajectory that could send signals
that both arrived at the same time. Then if the distances to these two points
are d1 and d2 , the retarded times for these points are

d1
(1) t1 = t
c
d2,
(2) t2 = t
c

or, in terms of the distances

(3) d1 d2 = c (t2 t1 )
If the particle gets closer to us between t1 and t2 then the difference d1
d2 represents the amount by which the distance to the particle has decreased.
According to 3, the speed at which the particle must move to cover this
distance is

d1 d2
(4) =c
t2 t1

That is, the average speed in the radial direction has to be c. If the particle
also has some transverse velocity on its trajectory, its total speed must be
greater than c, which isnt allowed. (If the velocity is entirely radial, then
the particle would have to be moving at exactly c, which also isnt allowed
1
POINT CHARGE IN HYPERBOLIC MOTION: VISIBLE AND INVISIBLE POINTS 2

for any particle with rest mass.) Thus there can be at most one point where
we receive a signal from a moving point charge.
It turns out that there are situations where a moving particle cannot be
seen at all. For a simple example, suppose we have a particle moving along
the x axis with a trajectory

p
(5) w (t) = b2 + c2t 2 x

where < t < +.


Its easiest to draw the trajectory on a spacetime diagram, with horizontal
axis x and vertical axis ct, as usual. In the following diagram, b = 1:

The red curve (a hyperbola) is the trajectory, so we see that the particle
approaches from the right until it reaches closest approach to the origin at
x = 1, then it moves away again. The grey lines are the asymptotes of the
hyperbola. The yellow lines represent photons emitted by the particle at
various points in its motion. As usual, they move upwards to the left and
right parallel to the lines ct = x. We can see from the diagram that no
photons can ever reach points below the line ct = x, so any observers in
this region will be unaware of the particles existence (so the potentials are
zero in this area).
POINT CHARGE IN HYPERBOLIC MOTION: VISIBLE AND INVISIBLE POINTS 3

For a stationary observer at location x, his world line is a vertical line


travelling upwards, so he will first see the particle when he crosses the line
ct = x. Once the particle becomes visible, it will remain visible forever,
since the particle is visible everywhere above the line ct = x.
P INGBACKS
Pingback: Linard-Wiechert potential for a charge moving on a hyper-
bolic trajectory
Pingback: Force of point charge in a hyperbolic trajectory on a fixed
point charge
Pingback: Radiation from a charge in hyperbolic motion
Pingback: Four-velocity of a particle in hyperbolic motion
Pingback: Self-force on a dipole in hyperbolic motion
LINARD-WIECHERT POTENTIAL FOR A CHARGE MOVING
ON A HYPERBOLIC TRAJECTORY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 10.16.
We can now work out the Linard-Wiechert potentials for a point charge
moving on a hyperbolic trajectory. Were trying to find

qc 1
(1) V (r,t) =
40 (c |r w (tr )| (r w (tr )) v)
The motion is in one dimension, given by
p
(2) w (t) = b2 + c2t 2 x
Its worth noting here that w b > 0 for all times, and the velocity is

dw c2t
(3) v (t) = = x
dt b2 + c2t 2

which is negative for t < 0 (when the particle is moving in from the right)
and positive for t > 0 (when the particle is moving back out again).
Well consider only observation points r that lie on the x axis to the right
of the particles location, so the separation is
p
(4) d (t) = r w (t) = x b2 + c2t 2 > 0
Using this notation, the potential is

qc 1
(5) V (r,t) =
40 (cd (tr ) d (tr ) v)
The retarded time is given by

(6) d (tr ) = c (t tr )
q
(7) x b2 + c2tr2 = c (t tr )
1
LINARD-WIECHERT POTENTIAL FOR A CHARGE MOVING ON A HYPERBOLIC TRAJECTORY
2

We can solve this by isolating the square root on one side and then squar-
ing both sides, to get

x2 2xct + c2t 2 b2
(8) tr =
2c (ct x)
Note that for t = 0, as x , tr which makes sense, since the
further out on the x axis we place the observer, the further back in time we
need to go to get a signal from the particle. We can substitute this back into
2 to get
s
2
1 (x2 2xct + c2t 2 + b2 )
(9) w (tr ) =
2 (ct x)2

Although the operand of the square root is a perfect square, we need to be


careful when taking the square root to ensure we get the correct sign. Since
w > 0 we can look at t = 0 as before, at which point
s
2
1 (x2 + b2 )
(10) w (tr ) = >0
2 (x)2

so we need to take the negative root of the operand to get w > 0. That is

x2 2xct + c2t 2 + b2
(11) w (tr ) =
2 (x tc)
We can now calculate

(12) d (tr ) = x w (tr )


x2 c2t 2 b2
(13) =
2 (x ct)

after simplifying.
Now we need to find v (tr ). Substituting 8 into 3

c b2 x2 + 2 xct c2t 2

(14) v (tr ) = r 2
2 (b2 x2 +2 xctc2t 2 )
2 (ct x) b + 2
4(x+ct)

Simplifying the operand of the square root, we get


LINARD-WIECHERT POTENTIAL FOR A CHARGE MOVING ON A HYPERBOLIC TRAJECTORY
3

c b2 x2 + 2 xct c2t 2

(15) v (tr ) = r 2
(x2 2xct+c2t 2 +b2 )
2 (ct x) 2
4(ctx)
Again, we need to take the correct sign when taking the square root. For
t = 0 we get

b2 x 2
(16) v (tr ) = c r 2
(x2 +b2 )
2x
4(x)2
For large x, the signal comes from the particle in the distant past when
it was moving to the left, so we should have v negative in this case. This
again requires taking the negative root, so we get

c b2 x2 + 2 xct c2t 2

(17) v (tr ) =
x2 2xct + c2t 2 + b2
Putting it all together, we get

(18)
  2 2

x2 + c2t 2 + b2 x2 + c2t 2 + b2 c b (x + ct)

1 c 1
cd (tr ) d (tr ) v =  
2 x + ct 2 (x + ct) b2 + (x + ct)2
(x + ct) x2 + c2t 2 + b2 c

(19) =
x2 2 xct + c2t 2 + b2
Weve used Maple to do some of the algebra. Therefore the potential is

1 q x2 2 xct + c2t 2 + b2

(20) V (r,t) =
40 (x + ct) (x2 + c2t 2 + b2 )
 
2 2
1 q (x ct) + b
(21) =
40 (x ct) (x2 c2t 2 b2 )
FIELDS OF A MOVING POINT CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problems 10.17.
We can use the Linard-Wiechert potentialsfor a moving point charge to
calculate the fields produced by the charge. The potentials are

qc 1
(1) V (r,t) =
40 (c |r w (tr )| (r w (tr )) v)
q v
(2) A (r,t) =
40 c (c |r w (tr )| (r w (tr )) v)

where w (tr ) is the position of the particle at the retarded time tr and v =
dw/dtr is the velocity at the same time.
Ill use the shorthand variable

(3) r r w (tr )

in what follows (since I dont know of any Latex symbol to match the script-
r used by Griffiths). Thus the potentials are written as

qc 1
(4) V (r,t) =
40 (cr r v)
q v
(5) A (r,t) =
40 c (cr r v)
Given the potentials, we can work out the fields using

A
(6) E = V
t
(7) B = A
Because of the convoluted dependence of the various quantities on the
observers location r and time t, these derivatives get quite involved. Grif-
fiths works out V in his section 10.3.2, with the result
1
FIELDS OF A MOVING POINT CHARGE 2

qc  2 2
 
(8) V = (cr r v) v c v + r a r
40 (cr r v)3
so well deal with tA here.
First, its useful to work out tr dtr /dt. The retarded time tr is defined
implicitly by the equation

(9) r = |r w (tr )| = c (t tr )
We start with

1
(10) rr = (r r)
t 2r t
1 2
r 2r w + w2

(11) =
2r t 
1 dw dw
(12) = 2r
tr + 2w
tr
2r dtr dtr
1
(13) = tr v (w r)
r
1
(14) = tr v r
r
From the RHS of 9 we get

d
(15) (t tr ) = c (1 tr )
c
dt
Combining these two equations we get

1
(16) c (1 tr ) = tr v r
r
rc
(17) tr =
rc r v
rc
(18) =
ru

where

(19) u cr v
Returning to 5 we now have
FIELDS OF A MOVING POINT CHARGE 3

4c0 A v 1 1
(20) = v 2
(cr r v)
q t t (cr r v) (cr r v) t
v 1 1 r r v
(21) = tr v 2
c vr tr
tr (cr r v) (cr r v) t t tr
The derivative

v
(22) a (tr )
tr

is the acceleration of the particle at the retarded time. The other two deriva-
tives are, from 9
r
(23) = c (1 tr )
t
r dw
(24) = tr = vtr
t dtr
Therefore

(25)
" #
40 c A a v 2 2
 c2 v
= tr c + v r a
q t (cr r v) (cr r v)2 (cr r v)2
(26)
" #
rc a v 2 2
 c2 v
= c + v r a
rc r v (cr r v) (cr r v)2 (cr r v)2
(27)
1  2 2
 2 
= rc a (cr r v) + v c v + r a c v (cr r v)
(cr r v)3
(28)
1
rca c2 v (cr r v) + rcv c2 v2 + r a
  
=
(cr r v)3
(29)
A qc h ra  rv 2 i
2
= v (cr r v) + c v +ra
t 40 (cr r v)3 c c
Combining this with 8 and inserting into 6 gives the result quoted in
Griffiths as equation 10.65:
FIELDS OF A MOVING POINT CHARGE 4

qr
c2 v2 u + r (ua)
  
(30) E (r,t) =
40 (r u)3
For reference, the magnetic field is

(31) B (r,t) = A
1
(32) = r E (r,t)
c
P INGBACKS
Pingback: Fields of a point charge moving in one dimension
Pingback: Fields of a point charge moving in a circle
Pingback: Radiation from a point charge; the Larmor formula
Pingback: Radiation reaction: the Abraham-Lorentz force
Pingback: Physical basis of the radiation reaction force
FIELDS OF A POINT CHARGE MOVING IN ONE DIMENSION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.18.
Heres a simple example of calculating the fields due to a moving point
charge. Suppose we have a charge constrained to move on the x axis in the
+x direction, so that v = vx.
The fields are

qr
c2 v2 u + r (ua)
  
(1) E (r,t) = 3
40 (r u)
1
(2) B (r,t) = r E (r,t)
c
where

(3) r r w (tr )
(4) u cr v

and w (tr ) is the particles position at the retarded time. If the observer is to
the right of the particle then

(5) r = +rx
(6) u = (c v) x
Since the motion is constrained to the x axis, any velocity and accelera-
tion must be parallel, so u a = 0 and we have

qr
c2 v2 u

(7) E =
40 (r u)3
q c2 v2

(8) = (c v) x
40 (c v)3 r2
q (c + v)
(9) = x
40 (c v) r2
Because r is parallel to E, B = 0 from 2.
1
FIELDS OF A POINT CHARGE MOVING IN ONE DIMENSION 2

If the observer is to the left of the charge, then

(10) r = rx
(11) u = (c + v) x
qr 2 2

(12) E = c v u
40 (r u)3
q c2 v2

(13) = (c + v) x
40 (c + v)3 r2
q (c v)
(14) = x
40 (c + v) r2
(15) B = 0
FIELDS DUE TO A MOVING LINEAR CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.19.
In his example 10.4, Griffiths works out the fields due to a point charge
moving with constant velocity v. They are

q 1 v2 /c2 R
(1) E (r,t) =
40 1 v2 sin2 /c2 3/2 R2


where

(2) R r vt

is the vector from the particles present (not retarded) position to the ob-
server (assuming the particle passes through the origin at t = 0) and is the
angle between R and v. We can use this formula to rederive the equation
for the electric field due to an infinite line charge with linear charge density
. From electrostatics, we know the field is given by


(3) E=
20 z

where z is the perpendicular distance from the line (wire). Lets see if we
can get the same result using the formula above.
The field due to a small segment of the wire of length dx at position is
that due to a point charge dx. For an observation point at r, the length of
R is
p
(4) R= z2 + x2

and since the velocity is parallel to the wire, we have

z
(5) sin =
z2 + x2
1
FIELDS DUE TO A MOVING LINEAR CHARGE 2

Since E is parallel to R, by symmetry the components of E parallel to the


wire will cancel out, since there will be equal and opposite contributions
from points x. The perpendicular component is E sin so the total field is

v2

sin dx s
(6) E= 1 2
40 c 1 v2 sin2 /c2
3/2 (z + x2 )
2

where the s direction is radial. We can convert this to an integral over by


noting that

xz
(7) cos d = 3/2
dx
(z2 + x2 )
!
dx z2 + x2 xz
(8) = dx
z2 + x 2 xz (z2 + x2 )
3/2

z2 + x2
(9) = cos d
xz

But

x
(10) cos =
z2 + x2
so

dx d
(11) =
z2 + x 2 z
(12)
 
v2 v2
 
sin dx s sin d
1 2 3/2 2 2
= 1 2 3/2 s
40 c 1 v2 sin2 /c2
 (z + x ) 40 z c 0 1 v2 sin2 /c2
The integral can be evaluated using Maple, and we get



sin d cos
(13) 3/2 s = s

  q
0 1 v2 sin2 /c2 v2
1 c2 v2 2
1 c2 sin

0
2
(14) = 2 s
1 vc2

so we get back the correct field


FIELDS DUE TO A MOVING LINEAR CHARGE 3


(15) E= s
20 z
The magnetic field of a point charge is given by Griffiths as

1
(16) B (r,t) = v E (r,t)
c2
Since v is a constant, the total magnetic field can be found from the same
integral as above. Its direction is given by x s = which circles the wire
in a direction given by the usual right-hand rule. Since v = I (the current),
we get

I
(17) B =
20 c2 z
0 I
(18) =
2z

which agrees with the magnetostatic formula using Ampres law.


P INGBACKS
Pingback: Power due to point charge moving with constant velocity
Pingback: Force and momentum with two moving point charges
Pingback: Gausss law for a relativistic point charge
FIELDS OF A POINT CHARGE MOVING IN A CIRCLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.20.
We can now calculate the electric and magnetic fields due to a point
charge moving in a circle of radius a at constant angular velocity . The
charges position and velocity are given by

(1) w (t) = ax cos t + ay sin t


(2) v (t) = a x sin t + a y cos t

The calculation is easier if we use cylindrical coordinates, where

(3) s (t) = x cos t + y sin t


(4) (t) = x sin t + y cos t
(5) z = z

Here, s is the radial coordinate and is the angular coordinate, and their
unit vectors are both functions of time. Using these coordinates, we have

(6) w = as
(7) v = a

The fields of a moving point charge are

qr
c2 v2 u + r (ua)
  
(8) E (r,t) =
40 (r u)3
1
(9) B (r,t) = r E (r,t)
c

where
1
FIELDS OF A POINT CHARGE MOVING IN A CIRCLE 2

(10) r r w (tr )
(11) u cr v (tr )
(12) a = v (tr )
For the fields at the centre of the circle r = 0 so we get

(13) r = w (tr )
(14) = as (tr )
(15) u = cs a
(16) a = a 2 s
(17) u a = a2 3 z
(18) r (ua) = a3 3
c2 v2 u + r (ua) = c2 v2 cs a a3 3
  
(19)
= ca2 2 c3 s c2 a

(20)
(21) r u = ac
Putting all this together, we get

q
a2 2 c2 s (tr ) ca (tr )
 
(22) E (0,t) = 2 2
40 a c
q
(as) a2 2 c2 s ca
 
(23) B (0,t) = 3 3
40 a c
q
(24) = z
40 ac2
0 q
(25) = z
4a
The electric field depends on time, but the magnetic field is constant.
We can use this result to find the magnetic field at the centre due to a
steady current caused by a line charge density of moving at angular ve-
locity . In that case each infinitesimal segment of the loop has length a d
so q = a d . The current is I = a so the field is

2
0 a d
(26) B = z
0 4a
0 I
(27) = z
2a
FIELDS OF A POINT CHARGE MOVING IN A CIRCLE 3

In his example 5.6 Griffiths gives the magnetic field along the z axis of a
uniform circular current, and his result is

0 a2
(28) B= 3/2
z
2 (a2 + z2 )
which reduces to our result at the centre of the circle where z = 0.
RETARDED POTENTIALS FOR A SINUSOIDAL CURRENT
LOOP

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.21.
Heres another example of calculating retarded potentials


1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d
We have a circular loop of radius a with a charge distribution (at t = 0)
of 0 sin 2 . The loop is then spun at angular velocity and we wish to find
the potentials at the centre of the loop.
In this case, the distance from the observation point (r = 0) to the retarded
position of the charge is always the same (d = a) so the potentials become


1
r0 ,tr d 3 r0

(3) V (r,t) =
40 d

0
J r0 ,tr d 3 r0

(4) A (r,t) =
4d
The scalar potentials integral is thus just the total charge on the loop,
which doesnt depend on time so we have

2
0 a
(5) V (r,t) = sin d
40 a 0 2
0
(6) =
0
For the vector potential we need J (r0 ,tr ). To get this, we use the follow-
ing argument. At t = 0, the charge element at angle is 0 a sin 2 . At time t,
this element has moved through angle t so its now at an angle of + t,
so the linear charge density as a function of time is
1
RETARDED POTENTIALS FOR A SINUSOIDAL CURRENT LOOP 2


(7) (t) = 0 a sin [cos ( + t) x + sin ( + t) y]
2
The linear current is therefore


(8) I ( ,t) = = 0 a sin [ sin ( + t) x + cos ( + t) y]
2
The integral of the current is

(9)
2
0
 3 0
J r ,tr d r = I ( ,tr )
0
2  

(10) = 0 a sin sin ( + tr ) x + sin cos ( + tr ) y d
0 2 2
4 h   a    a  i
(11) = 0 a sin t x cos t y
3 c c
[The integrals can be done using the trigonometric addition formulas for
cos (x + y) and sin (x + y).]
The vector potential is therefore

0 0 a h   a    a  i
(12) A (r,t) = sin t x cos t y
3 c c
RELATIONS AMONG CHARGE, CURRENT, POTENTIALS AND
FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.22.
With a complete theory of electrodynamics, its useful to summarize the
relations between the sources of the fields (charge density and current
density J), the potentials V and A and the fields E and B.
Starting with and J, we can calculate the retarded potentials

1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d
Or we can get the fields from Jefimenkos equations for the electric and
magnetic fields.

(r0 ,tr ) (r0 ,tr ) J (r0 ,tr ) 3 0

1
(3) E (r,t) = r + r d r
40 r2 cr c2 r

J (r0 ,tr ) r

0 0
 r 3 0
(4) B (r,t) = + J r ,tr 2 d r
4 cr r
Inverting the procedure, we can get the sources from the potentials, al-
though we need to know the gauge were using. In the Lorenz gauge we
have


(5) 2V =
0
2
(6)  A = 0 J
From the fields, we can use Maxwells equations


(7) E =
0
1 E
(8) B = 0 J
c2 t
1
RELATIONS AMONG CHARGE, CURRENT, POTENTIALS AND FIELDS 2

Finally, from the potentials we can get the fields

A
(9) E = V
t
(10) B = A
POINT CHARGE MOVING AT CONSTANT VELOCITY
SATISFIES LORENZ GAUGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.23.
Griffiths shows in his example 10.3 that the Linard-Wiechert potentials
for a point charge q moving at constant velocity v that passes through the
origin at time t = 0 are

1 qc
(1) V (r,t) = q
40 2
(c2t r v) + (c2 v2 ) (r2 c2t 2 )
0 qcv
(2) A (r,t) = q
4 2
(c2t r v) + (c2 v2 ) (r2 c2t 2 )
These potentials were derived from the general retarded potential formu-
las that in turn satisfy the Lorenz gauge condition

1 V
(3) A =
c2 t
It is possible to show explicitly that these two potentials satisfy the Lorenz
gauge condition, although to do so by hand is straightforward, but rather te-
dious. Its easier to let Maple handle the derivatives, and we find

qc v2t r v

(4) A =  3/2
2 2 2 2 2 2 2
40 (c t r v) + (c v ) (r c t )
qc3 v2t r v

V
(5) =  3/2
t 2 2 2 2 2 2 2
40 (c t r v) + (c v ) (r c t )

so the condition is satisfied.

1
FORCE OF POINT CHARGE IN A HYPERBOLIC
TRAJECTORY ON A FIXED POINT CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.24.
Well now return to the point charge q2 moving on a hyperbolic trajectory
along the x axis. Its position is given by
p
(1) x (t) = b2 + c2t 2
Suppose there is a second charge q1 fixed at x = 0. What force does
q1 exert on q2 at time t? If we look at the problem from q2 s perspective,
it needs to know what influence q1 has on the point x that it (q2 , that is)
currently occupies. In order for a signal to reach x at time t, it had to leave
q1 at time tr = t xc . But since q1 is fixed at the origin, the force that q1
exerts on any charge q2 at the point x is always just given by Coulombs law
without any retarded time, that is

q1 q2
(2) F2 (t) =
40 x2
q1 q2 1
(3) =
40 b + c2t 2
2

The total impulse delivered to q2 is



(4) I2 = F2 (t) dt


q1 q2 1
(5) = dt
40 b 1 + c22 t 2
2
b
q1 q2
(6) =
40 bc
To calculate the force q2 exerts on q1 we do need the retarded time, since
q2 is moving. [Incidentally, it might seem that by considering q1 as at rest
and q2 as moving, and treating the two cases differently, were violating the
principle of relativity, but were not. The reason is that q2 is not in an inertial
frame; the hyperbolic trajectory means that its velocity is never constant, so
q1 and q2 are not equivalent.]
1
FORCE OF POINT CHARGE IN A HYPERBOLIC TRAJECTORY ON A FIXED POINT CHARGE
2

The retarded time is calculated from

(7) |r w (tr )| = c (t tr )
p
Here r = 0 is the location of q1 and w (tr ) = b2 + c2tr2 x is the position
of q2 . We get
q
(8) b2 + c2tr2 = c (t tr )
t b2
(9) tr = 2
2 2c t
Note that as t 0, tr so q2 is not visible to q1 before t = 0, so
the retarded potential is zero for t < 0. The force for t > 0 is therefore (the
minus sign indicates the force is to the left if both charges are the same sign)

q1 q2 1
(10) F1 (t) =
40 2 2
2
 
b2 + c4 t cb2t
q1 q2 c2t 2
(11) =
0 4c2 b2t 2 + (c2t 2 b2 )2
q1 q2 c2t 2
(12) =
0 (c2t 2 + b2 )2
The total impulse is found by integrating F1 from t = 0 to infinity, since
there is no force for t < 0.

q1 q2 c2t 2
(13) I1 = dt
0 0 (c2t 2 + b2 )2
q1 q2
(14) =
40 bc
[The integral can be done by parts, although I used Maple.] Thus the two
impulses are equal and opposite.
POWER DUE TO POINT CHARGE MOVING WITH CONSTANT
VELOCITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 10.25.
Returning to the case of a point charge moving with constant velocity v,
the fields due to the charge are

q 1 v2 /c2 R
(1) E (r,t) =
40 1 v2 sin2 /c2 3/2 R2
1
(2) B (r,t) = 2 v E (r,t)
c

where R is a vector from the charges location at the present time to the
observer and is the angle between R and v. In what follows, well take v
to be in the +x direction.
Lets calculate the total power passing through the plane x = a when the
charge is at x = 0. The power per unit area is given by the Poynting vector

1
(3) S= EB
0

If we use cylindrical coordinates, with the x axis serving as the more


usual z axis, s being the radial distance from the x axis and the angular
coordinate, then E has components in the s and x directions, and the unit
vector R is given by

(4) R = x cos + s sin

where
1
POWER DUE TO POINT CHARGE MOVING WITH CONSTANT VELOCITY 2

s
(5) sin =
s2 + a2
a
(6) cos =
s + a2
2

(7) R2 = s2 + a2

To calculate B, only the s component will survive the cross-product since


v is parallel to x, so we get, using x s =

v q 1 v2 /c2 sin
(8) B=
c2 40 1 v2 sin2 /c2 3/2 s2 + a2


To get the power flowing through the plane x = a, we need the component
of S parallel to x. Since B is in the direction and s = x, we can get the
x component of S by multiplying B by the s component of E, which well
call Es . We have

q 1 v2 /c2 sin
(9) Es =
40 1 v2 sin2 /c2 3/2 s2 + a2


so the x component of the Poynting vector is

2 2
1 v2 /c2 sin2

v q
(10) Sx = 3 2
0 c2 40 1 v2 sin2 /c2 (s2 + a2 )
2
vq2 1 v2 /c2 sin2
(11) = 3 2
16 2 0 1 v2 sin2 /c2 (s2 + a2 )


using c2 = 1/0 0 .
We therefore need to integrate this over the plane x = a, which we can do
by expressing sin in terms of s using 5. The increment of area is s ds d
so the integral is
POWER DUE TO POINT CHARGE MOVING WITH CONSTANT VELOCITY 3

(12)
2
vq2 1 v2 /c2 2
s3
Px=a = ds d 3 3
16 2 0 0 0 (1 v2 s2 / (s2 + a2 ) c2 ) (s2 + a2 )
(13)
2
vq2 1 v2 /c2
s3 ds
= 3
80 0 (s2 (1 v2 /c2 ) + a2 )
(14)
2
vq2 1 v2 /c2 1
= 2
80 4a2 (1 v2 /c2 )
(15)
vq2
=
320 a2
I did the integral using Maple, but if you want to do it by hand, you can
use partial fractions, since

(16)
s3 sc6 sc8 a2
3
= 2
3
(s2 (1 v2 /c2 ) + a2 ) (s2 (c2 v2 ) + a2 c2 ) (c2 v2 ) (s2 (c2 v2 ) + a2 c2 ) (c2 v2 )
FORCE AND MOMENTUM WITH TWO MOVING POINT
CHARGES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 10. Problem 10.26.
Returning to the case of a point charge q2 moving with constant velocity
v, the fields due to the charge are

q2 1 v2 /c2 R
(1) E2 (r,t) =
40 1 v2 sin2 /c2 3/2 R2
1
(2) B2 (r,t) = 2 v E2 (r,t)
c

where R is a vector from the charges location at the present time to the
observer and is the angle between R and v. In what follows, well take v
to be in the +z direction and place the charge at the origin at t = 0.
We now place another charge q1 at the origin (and hold it there). The
force F12 of q1 on q2 is given by Coulombs law, since q1 is at rest and its
field is given by

q2 r
(3) E1 (r,t) =
40 r2

so the force is

q1 q2
(4) F12 = z
40 v2t 2
To get the force exerted by q2 on q1 we can use the field 1 with =
and R = vt:

(5) F21 = q1 E
q1 q2 1 v2 /c2

(6) = z
40 v2t 2
Thus Newtons third law (action = reaction) is not obeyed here. The
reason is that both charges must be experiencing other forces than the force
1
FORCE AND MOMENTUM WITH TWO MOVING POINT CHARGES 2

from the other charge. Since q1 is at rest and q2 is in uniform motion, the
net force on each charge must be zero (no acceleration), so some external
force cancels the electric force on each charge.
We can calculate the total linear momentum p stored in the fields of the
two charges (or at least the time-varying part of the momentum) from the
momentum density

(7) p = 0 E B
Since q1 is at rest, it generates no magnetic field, so the total momentum
density is

(8) p = 0 (E1 + E2 ) B2
However, if we integrate E2 B2 over all space, the result (although com-
plicated) must be independent of time since the magnetic and electric fields
due to q2 dont change relative to each other. The time varying part of the
momentum density is therefore

(9) pt = 0 E1 B2
Weve already calculated B2 as

v q2 1 v2 /c2 sin
(10) B2 = 2

c 40 1 v2 sin2 /c2 3/2 R2

At this point a diagram is helpful:


FORCE AND MOMENTUM WITH TWO MOVING POINT CHARGES 3

We can get the direction of pt by considering the two components of E1 .


The component perpendicular to the z axis points up when were above the
z axis and down when were below it. Since B2 circles around the z axis
it points out of the page above the z axis and into the page below it. Since
the components of both E1 and B2 reverse on either side of the z axis, their
cross product is the same on both sides and points in the +z direction.
The other component of E1 points in the +z direction on both sides of
the z axis so, since B2 reverses direction on the two sides of the z axis,
the cross product of the z component of E1 with B2 at a point above the z
axis is cancelled by the point on the other side of the axis. Thus the only
component of E1 that contributes to pt when it is integrated over all space
is E which is, from the diagram:

q1 sin
(11) E =
40 r2

The momentum density is therefore

(12) pt = 0 E B2 z
q1 q2 v 1 v2 /c2 sin

sin
(13) = 2 2 2 3/2 z
16 0 c r 1 v2 sin2 /c2 R2

At this point, we need to choose which coordinate system to use in or-


der to integrate this expression over all space. I originally tried cylindrical
coordinates since the problem has cylindrical symmetry, but I couldnt get
Maple to do the resulting integral. So we can resort to spherical coordinates
if we can express and R in terms of r and .
Using the cosine rule for triangles (see diagram) we have

(14) R2 = r2 + v2t 2 2rvt cos


(15) R sin = r sin
r sin
(16) sin =
R
r sin
(17) =
r + v t 2 2rvt cos
2 2

We therefore have for the total time dependent momentum


FORCE AND MOMENTUM WITH TWO MOVING POINT CHARGES 4

(18)
 2
q1 q2 v 1 v2 /c2 sin (r sin ) r2 sin d drd
p = z
16 2 0 c2 0 0 0 r2  3/2
v2 r2 sin2 3/2
1 c2 (r2 + v2t 2 2rvt cos )
(r2 +v2t 2 2rvt cos )
(19)

q1 q2 v 1 v2 /c2 r sin3 drd
= z
80 c2 0 0
 3/2
r2 + v2t 2 2rvt cos c12 v2 r2 sin2
(20)

q1 q2 v 1 v2 /c2 r sin3 drd
= z
80 c2 0 0
 
2 v2 sin2

2 2
3/2
r 1 c2 2rvt cos + v t
(21)
r=


2 2

q1 q2 v 1 v /c

sin (r cos vt)
= z  2
 r  d
2 v 
80 c 1 c2 vt 0 2

2 v2 sin 2 2
r 1 c2 2rvt cos + v t
r=0
(22)
 q 
v2 2
sin cos + 1 c2 sin
q1 q2
= z d
80 c2t
q
0 v2 2
1 c2 sin
(23)
r
q1 q2 v2 2
= z 1 2 sin cos

80 c2t c
0
(24)
0 q1 q2
= z
4t

where we used c2 = 1/0 0 in the last line.


If we take the sum of the two forces above, we get

q1 q2 0
(25) F12 + F21 = z
4t 2
dp
(26) =
dt
FORCE AND MOMENTUM WITH TWO MOVING POINT CHARGES 5

This indicates that the momentum change in the fields is equal and oppo-
site to the forces between the charges, which means that the external forces
are causing the momentum change. If there were no external forces, then
F12 + F21 = 0 and there would be no change in the momentum contained in
the fields.
ELECTRIC DIPOLE TIME-VARYING POTENTIALS; LORENZ
GAUGE CONDITION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Section 11.1.2. Problem 11.1.
The power radiated by a system of charges and currents is defined as
the amount of energy that leaves the system and is non-zero even at an
infinite distance. As the total rate at which energy is radiated by electric
and magnetic fields is given by the integral of the Poynting vector over a
surface (e.g. a sphere) enclosing the system, then the system will radiate
power if the integral over the sphere


1
(1) P= (E B) da
0

is non-zero as the spheres radius becomes infinite. For static charges and
currents, both the electric and magnetic fields fall off at least as fast as r12
so their product falls off at least as fast as r14 . Thus the integral falls off
at least as fast as r12 so will go to zero at infinity. Thus static charge and
current distributions dont radiate. We need changing charge and/or current
densities to produce radiation.
This follows from Jefimenkos equations for electric and magnetic fields


(r0 ,tr ) (r0 ,tr ) J (r0 ,tr ) 3 0

1
(2) E (r,t) = d + d d r
40 d2 cd c2 d

J (r0 ,tr ) d

0 0
 d 3 0
(3) B (r,t) = + J r ,tr 2 d r
4 cd d

where the retarded time is

d
(4) tr t
c

and
1
ELECTRIC DIPOLE TIME-VARYING POTENTIALS; LORENZ GAUGE CONDITION 2

d r r0

(5)
q
(6) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(7) d =
d
Both time-varying fields contain terms involving time derivatives and J
that depend on 1/r (via their dependence on d) so the Poynting vector will
have terms that depend on 1/r2 . This means that integrating these over the
surface of a sphere will cancel off the dependence on r so that the radiated
energy is independent of distance from the source charges and currents.
We can see this in action for a time-varying electric dipole, which we set
up as two charges connected by a wire of length l along the z axis. At the
start, the top charge is +q and the bottom charge is q. We then drive the
top charge through the wire so that it cancels out the bottom charge, then
we drive the positive charge upwards again and so on, all with an angular
frequency so that the top charge is

(8) q (t) = q0 cos (t)


To see what electric and magnetic fields this dipole produces, we need
to find the scalar potential V and the vector potential A from which we can
calculate the electric field as

A
(9) E = V
t

and the magnetic field in the usual way as

(10) B = A
We can get these potentials using the formulas for retarded potentials

1 (r0 ,tr ) 3 0
(11) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(12) A (r,t) = d r
4 d
The details are given by Griffiths in his section 11.1.2, but the essential
points to note are as follows. First, we want the fields for a perfect dipole,
which is when the separation l becomes infinitesimal. This is implemented
by taking l  r, that is, the distance between the charges is much less than
ELECTRIC DIPOLE TIME-VARYING POTENTIALS; LORENZ GAUGE CONDITION 3

the distance from the dipole to the observation point (ultimately we want
r so this is a reasonable assumption).
Second, we assume that l  c/. If the dipole gives rise to electromag-
netic waves travelling at c then the wavelength of a wave with frequency
is = 2c/ so this is equivalent to assuming that the distance between the
charges is much less than the wavelength of radiation they produce. Elec-
tromagnetic waves can have very small wavelengths (visible light is in the
region of 107 m) but since were taking l to be infinitesimal, we can again
say this is reasonable.
Using just these two assumptions, we can calculate the potentials by sav-
ing up to first order terms in small quantities, and we get

   r 
p0 cos   r  1
(13) V (r, ,t) = sin t + cos t
40 r c c r c
0 p0   r 
(14) A (r, ,t) = z sin t
4r c

where is the polar angle between the z axis (the axis of the dipole) and
the observation point r and p0 is the maximum dipole moment p0 = ql.
V can be further approximated by assuming r  c/, that is, assuming
that the observation point is much greater than the wavelength of the waves.
Locations satisfying this condition are said to lie in the radiation zone. In
this case we can drop the second term in 13 and get

p0 cos   r 
(15) V (r, ,t) = sin t
40 rc c

We can show that the potentials 13 and 14 satisfy the Lorenz gauge con-
dition

V
(16) A = 0 0
t
Since

p
(17) r= x2 + y2 + z2

we have
ELECTRIC DIPOLE TIME-VARYING POTENTIALS; LORENZ GAUGE CONDITION 4

(18)
 
0 p0 1   r 
A = sin t
4 z r c
   r 
0 p0 z   r  1  z 
(19) = 3 sin t + cos t
4 r c r rc c

From the geometry of the setup we have

z
(20) = cos
r

so we get
   r 
0 p0 cos 1   r 
(21) A = sin t + cos t
4r r c c c
From 13 we have

(22)
2
   r 
V p0 0 cos   r 
0 0 = cos t sin t
t 4r c c r c
 
0 p0 cos   r  1   r 
(23) = cos t + sin t
4r c c r c
(24) = A

Thus the gauge condition is satisfied.


P INGBACKS
Pingback: Fields and radiated power from an oscillating electric dipole
Pingback: Radiation resistance
Pingback: Radiation from a rotating dipole
Pingback: Fields of an oscillating magnetic dipole
Pingback: Radiation from a magnetic dipole composed of monopoles
FIELDS AND RADIATED POWER FROM AN OSCILLATING
ELECTRIC DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 2
The potentials for an oscillating dipole at a large distance from the dipole
are

p0 cos   r 
(1) V (r, ,t) = sin t
40 rc c
0 p0   r 
(2) A (r, ,t) = z sin t
4r c
These formulas apply in the special case where the dipole axis is the z
axis, so that the dipole moment is

(3) p = p0 cos (t) z

We can rewrite these formulas for a dipole pointing in any direction by


noting that

(4) p0 cos = p0 r

so

  r 
(5) V (r, ,t) = (p0 r) sin t
40 rc c
0   r 
(6) A (r, ,t) = p0 sin t
4r c
The fields can be calculated from the potentials using straightforward
differentiation. Griffiths shows the details in his section 11.1.2. After as-
suming that r  c (equivalent to assuming that the observation point is
much greater than the wavelength of the radiation) we get from 1 and 2:
1
FIELDS AND RADIATED POWER FROM AN OSCILLATING ELECTRIC DIPOLE 2

A
(7) E = V
t
0 p0 2 sin   r 
(8) = cos t
4 r c
(9) B = A
0 p0 2 sin   r 
(10) = cos t
4c r c
Note that E and B are perpendicular and in phase, and that E/B = c just
as with plane waves in vacuum. We can write these equations for general
dipole directions by noting that

(11) z r = sin
(12) r =
(13) (z r) r = sin
Therefore

0 2   r 
(14) E = cos t (p0 r) r
4r c
0 2   r 
(15) B = cos t (p0 r)
4rc c
The energy radiated per unit area per unit time is given by the Poynting
vector:

1
(16) S = EB
0
0 p0 2 sin
   r 2
(17) = cos t r
c 4 r c
For a general dipole direction, this is

0 2 |p0 r|
   r 2
(18) S= cos t r
c 4 r c
The intensity is the average of S over a single time cycle (that is, over a
time 2/). The average of cos2 x over a single cycle is 21 , so

2
0 2 p0 sin

(19) hSi = r
2c 4 r
FIELDS AND RADIATED POWER FROM AN OSCILLATING ELECTRIC DIPOLE 3

or in direction-independent form

2
0 2 |p0 r|

(20) hSi = r
2c 4 r
There is no radiation along the dipoles axis, and the maximum radiation
occurs perpendicular to the axis.
The average total power radiated is the surface integral of hSi over a
sphere of radius r, so we get from 19

2
0 2 p0 sin

(21) hPi = r2 sin r da
2c 4 r
4 2 2
0 p0
(22) = sin3 d d
32 2 c 0 0
0 4 p20
(23) =
12c
The result is independent of distance r from the dipole, so we see that
this power remains constant out to infinity.
P INGBACKS
Pingback: Radiation resistance
Pingback: Radiation from a rotating dipole
Pingback: Electric dipole radiation from an arbitrary source
Pingback: Electric quadrupole radiation
RADIATION RESISTANCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 3
The model of an oscillating dipole that weve been using consists of two
charges exchanging charge by passing a current along a wire of length l
joining the charges. The average total power radiated by the dipole is

0 4 p20
(1) hPi =
12c

where p0 is the maximum dipole moment

(2) p0 = q0 l
Here q0 is the maximum charge at one end of the dipole, and the charge
oscillates according to

(3) q (t) = q0 cos t


If the energy lost through radiation were to be lost instead by heat gen-
erated by the current passing through the wire joining the charges, what
would the resistance of the wire need to be? This resistance is known as the
radiation resistance. The power generated by a current I passing through a
resistor R is

(4) P = I2R

and in this case, the current is given by

dq
(5) I= = q0 sin t
dt
Were interested in the average power, so we want the average of I 2 over
a single cycle. The average of sin2 x over a cycle is 12 so we get
1
RADIATION RESISTANCE 2


2 1 2 2
(6) I = q0
2
1 2 2
(7) hPi = q0 R
2
Equating this to 1 we get

0 2 p20
(8) R =
6cq20
0 c 2 q20 l 2
(9) =
6c2 q20
0 c 2 2 2
 
(10) = l
6
2 0 c l 2
(11) =
3 2
l2
(12) = 787 2

where in line 3 we used /c = 2/ .


For a typical radio, we can take the length of a wire connecting compo-
nents to be around l = 5 cm, while radio waves typically have wavelengths
around 1 km, so the radiative resistance is around

0.05 2
 
(13) R = 787 = 2 106
103
This is much smaller than typical resistances in a radios circuits.
P INGBACKS
Pingback: Radiation resistance of an oscillating magnetic dipole
RADIATION FROM A ROTATING DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 4
We can simulate a rotating dipole by superimposing two perpendicular
oscillating dipoles. If our rotating dipole is located at the origin and rotates
about the z axis (so the axis of the dipole lies in the xy plane), then we get

(1) p = p0 (cos t x + sin t y)


Since the fields obey the superposition principle (fields from 2 sources
just add), we can work through the formulas we found earlier to get the
fields and thus the radiated power. To simplify the notation, well use the
shorthand
  r 
(2) c cos t
c
(3) c cos
(4) c cos

with analogous notation for the sines of these quantities.


The fields for a dipole pointing in an arbitrary direction are

0 2   r 
(5) E = cos t (p0 r) r
4r c
0 2   r 
(6) B = cos t (p0 r)
4rc c
Superposing the two perpendicular dipoles we get

0 p0 2
(7) E = [(c x + s y) r] r
4r
0 p0 2
(8) B = [(c x + s y) r]
4rc
To do the cross products we convert the rectangular unit vectors to spher-
ical unit vectors:
1
RADIATION FROM A ROTATING DIPOLE 2

(9) x = s c r + c c s
(10) y = s s r + c s + c
Then

(11) x r = c c s
(12) (x r) r = c c + s
(13) y r = c s + c
(14) (y r) r = c s c
Plugging everything in and collecting terms we get

0 p0 2   
(15) E= c c c s s + c s s c
4r
0 p0 2   
(16) B= c s s c + c c c s s
4rc
Defining

(17) E c c c + s s

(18) E c s s c

we can write the fields as

0 p0 2  
(19) E = E + E
4r
0 p0 2  
(20) B = E + E
4rc
In this form, its obvious that E B = 0, so E and B are perpendicular.
The Poynting vector is

1
(21) S= EB
0
2
0 p0 2  2
 
(22) = E + E2 r
c 4r
2
0 p0 2  2

2 2 
(23) = c c c + s s + c s s c r
c 4r
RADIATION FROM A ROTATING DIPOLE 3

The average energy radiated is the average of S over a single time cycle,
so its the average over the terms involving c and s . These are of two
types: terms involving c2 or s2 and the cross terms involving s c . The
average of s c over a cycle is zero and the average of c2 or s2 is 21 , so the
cross terms contribute nothing and we get

2 
p0 2
 
0 1 2 2 2
 1
2 2
(24) hSi = c c + s + c + s
c 4r 2 2
2
0 p0 2

1 + cos2

(25) =
2c 4r
The average radiated power is maximum in the z directions where =
0, and minimum (though not zero) in the xy plane, where = 2 . There is
no dependence on which is what wed expect on average since the dipole
rotates uniformly through all values of . [There is a dependence on
within each cycle, since the radiated power in a given azimuthal direction
depends on where the dipole is in its rotation.]
The total average radiated power is

2 2
p0 2 1 + cos2 2

0
(26) hPi = r sin d d
2c 4 0 0 r2
0 p20 4
(27) =
6c
This is exactly twice the power from a single oscillating dipole. Although
power doesnt ordinarily obey the superposition principle since it depends
on the product of E and B, it does here because the cross terms in 23 average
out to zero over a time cycle, since the two perpendicular dipoles are 2
out of phase. If they were exactly in phase, we would replace s by c
everywhere in the calculation, and then the cross terms wouldnt average out
to zero and the combined power would not be twice the individual power.
P INGBACKS
Pingback: Power radiated by a spinning ring of charge
FIELDS OF AN OSCILLATING MAGNETIC DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 5
We can analyze an oscillating magnetic dipole in a similar way to the
electric dipole. We begin with a small circular current loop of radius b in
the xy plane, centred at the origin. The current is driven to be alternating,
so that

(1) I (t) = I0 cos t


The magnetic dipole moment of a current loop is

(2) m = Ia

where a is the vector area of the loop, which for a planar circular loop is
just b2 z. Thus

(3) m (t) = b2 I0 cos (t) z


(4) m0 cos (t) z
If the loop is electrically neutral, the electric potential is V = 0, so we
need to calculate only A. The retarded potential is


0 I0 cos (t d/c) 0
(5) A (r,t) = d`
4 d

where the integral is taken around the loop and the retarded time is

d
(6) tr t
c

and
1
FIELDS OF AN OSCILLATING MAGNETIC DIPOLE 2

d r r0

(7)
q
(8) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(9) d =
d

where r0 is the position on the loop being integrated over.


To work out the integral, we can start by considering r to be some point in
the xz plane. The line integral can be broken down into pairs of increments
on the circle located at (x, y, 0), that is, each pair of points is symmetric
about the x axis. The increment d` at (x, y, 0) has components (dx, dy, 0)
while the increment d`+ at (x, y, 0) has components (dx, dy, 0). Thus the
x components cancel in pairs in the integral, while the y components add
in pairs, so the net result of the integral around the entire circle is a vector
pointing in the y direction. Since the circle is symmetric about the z axis,
we can generalize this result to deduce that A has a direction that is always
tangential to the circle, which means that, in spherical coordinates, it is in
the direction and has the same magnitude at all points around the circle.
Griffiths goes through the calculation of A in detail in his section 11.1.3
so I wont repeat that here, other than to note that he uses the same approx-
imations as were used with the electric dipole, namely that the radius of the
loop b is much less than the observation distance r, and that b is also much
smaller than the wavelength of the radiation, represented by the condition
b  c/. The result is

 
0 m0 sin 1
(10) A (r, ,t) = cos [ (t r/c)] sin [ (t r/c)]
4 r r c

At this stage, Griffiths invokes a further approximation by assuming that


r  c/ (observer is much further away than the wavelength of the radia-
tion). However, we can calculate the fields without making that approxima-
tion to see how much of an effect that approximation has. Since V = 0 we
have
FIELDS OF AN OSCILLATING MAGNETIC DIPOLE 3

A
(11) E =
t
0 m0 sin 2
 

(12) = cos [ (t r/c)] + sin [ (t r/c)]
4r c r
(13) B = A
 
0 m0 cos 1
(14) = cos [ (t r/c)] sin [ (t r/c)] r+
2r2 r c
1 2
  
0 m0 sin
cos [ (t r/c)] sin [ (t r/c)]
4r r2 c2 rc
These fields have the same form as those we worked out earlier for a
spherical wave in vacuum.
The Poynting vector can be worked out and simplified using Maple to
combine the trig products using double angle formulas:

(15)
1
S= EB
0
(16)
0 m20 sin2  2 2 4 3
 3 3 2
 4 3

= 2c r r cos [2 (t r/c)] + c 2 cr sin [2 (t r/c)] r r+
32 2 c3 r5
0 m20 sin (2 )  2 2 3 3 2
 
2c r cos [2 (t r/c)] + c cr sin [2 (t r/c)]
32 2 c3 r5
To get the average intensity, we integrate S over a complete cycle, that
is, for t = 0 to t = 2/ and then multiply by /2 to get the average.
Integrating over one cycle causes each of the double angle trig functions to
go through two complete cycles, so they all integrate to zero and the only
term that is left is the 4 r3 term in the radial component, so we get

0 m20 4 sin2
(17) hSi = r
32 2 c3 r2
This value is the same as that obtained by assuming r  c/ from the
start (equation 11.39 in Griffiths).

P INGBACKS
Pingback: Radiation resistance of an oscillating magnetic dipole
Pingback: Radiation from a magnetic dipole composed of monopoles
Pingback: Radiation from a current loop with time-varying current
FIELDS OF AN OSCILLATING MAGNETIC DIPOLE 4

Pingback: Radiation from a radio station


RADIATION RESISTANCE OF AN OSCILLATING MAGNETIC
DIPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 6
We can find the radiation resistance of an oscillating magnetic dipole
produced by an AC current in a circular wire loop of radius b in the same
way as for an oscillating electric dipole. The average power radiated by the
magnetic dipole is the integral of the intensity over a sphere, so we have


(1) hPi = hSi da

0 m20 4 sin2
(2) = r da
32 2 c3 r2
2
0 m20 4 sin 2
(3) = 2 r sin d
32 2 c3 0 r2
0 m20 4
(4) =
12c3

The resistance R required in the



wire
loop to generate the same power
loss through heat is given by P = I 2 R where the current is

(5) I (t) = I0 cos t

with the maximum current I0 given in terms of the maximum magnetic mo-
ment m0 :

m0
(6) I0 =
b2

1
Since the average of cos2 x over a complete cycle is 2 we get
1
RADIATION RESISTANCE OF AN OSCILLATING MAGNETIC DIPOLE 2


2 m20
(7) I R = R
2 2 b4
0 m20 4
(8) =
12c3
0 4 b4
(9) R =
6c3
80 5 c b4
(10) =
3 4
b4 5
(11) = 3.08 10 4

where in the fourth line we used /c = 2/ .


To compare this to the electric dipoles radiation resistance, which is
given in terms of the length l of the wire joining the two charges as:

l2
(12) Re = 787
2

we can take l = 2b so the lengths of wire in the two cases are the same.
Then

Re 787 (2)2 2
(13) =
Rm 3.08 105 b2
2
(14) = 0.1 2
b
Since were assuming that b  , Re  Rm .
RADIATION FROM A MAGNETIC DIPOLE COMPOSED OF
MONOPOLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 7
Weve seen that the fields produced by an oscillating magnetic dipole are

A
(1) E=
t
0 m0 sin 2
 

(2) = cos [ (t r/c)] + sin [ (t r/c)]
4r c r
(3) B = A
 
0 m0 cos 1
(4) = cos [ (t r/c)] sin [ (t r/c)] r+
2r2 r c
0 m0 sin

1 2 


cos [ (t r/c)] sin [ (t r/c)]
4r r2 c2 rc

By making the approximation that the observation distance r is much


larger than the wavelength of radiation, so that r  c/, these formulas
simplify to

0 m0 2 sin
(5) E = cos [ (t r/c)]
4cr
0 m0 2 sin
(6) B = cos [ (t r/c)]
4c2 r

Now lets return to the fantasy world where magnetic monopoles exist, so
that another way we can create a magnetic dipole is to connect two magnetic
charges by a wire and then drive charge back and forth between the ends of
the wire, in the same way that we did for the electric dipole. Earlier, weve
seen that if we include magnetic charge in Maxwells equations, the duality
transformation produces fields that still satisfy Maxwells equations:
1
RADIATION FROM A MAGNETIC DIPOLE COMPOSED OF MONOPOLES 2

(7) E0 = E cos + cB sin


(8) cB0 = cB cos E sin
(9) cq0e = cqe cos + qm sin
(10) q0m = qm cos cqe sin

where is a rotation angle in E B space. If we start with the fields gen-


erated by an oscillating electric dipole and then choose = 2 so that we
convert all electric charge into magnetic charge, we can generate the fields
that would be produced by an oscillating magnetic dipole constructed us-
ing magnetic charge as described above. The original fields for the electric
dipole are

A
(11) E = V
t
0 p0 2 sin   r 
(12) = cos t
4 r c
(13) B = A
0 p0 2 sin   r 
(14) = cos t
4c r c

The required transformations with = 2 are

(15) E0 = cB
(16) cB0 = E
(17) cq0e = qm
(18) q0m = cqe

As the electric dipole moment p0 = qe l is the product of an electric charge


qe and the length l of the wire joining the two charges, it transforms in the
same way as qe so we have, if we take the magnetic moment to be m0 = q0m l:

(19) m0 = cp0

Applying these transformations to 12 and 14 we get


RADIATION FROM A MAGNETIC DIPOLE COMPOSED OF MONOPOLES 3

0 (m0 /c) 2 sin   r 


(20) E0 = c cos t
4c r c
0 m0 2 sin
(21) = cos [ (t r/c)]
4cr
1 0 (m0 /c) 2 sin
    r 
(22) B0 = cos t
c 4 r c
0 m0 2 sin
(23) = cos [ (t r/c)]
4c2 r
These fields are the same as 5 and 6 that we got from the current loop.
Thus we cant tell whether magnetic dipole radiation is coming from a cur-
rent loop or from magnetic monopoles.
ELECTRIC DIPOLE RADIATION FROM AN ARBITRARY
SOURCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 8.
Having examined electromagnetic radiation from an oscillating electric
dipole, we can now look at radiation from an arbitrary source of moving
charges. The derivation of the results is rather long and Griffiths treats it
in detail in his section 11.1.4 so I wont go over it all again here, except to
point out the key assumptions made in the derivation.
To calculate the retarded potentials we start with

1 (r0 ,tr ) 3 0
(1) V (r,t) = d r
40 d

0 J (r0 ,tr ) 3 0
(2) A (r,t) = d r
4 d

where

d
(3) tr t
c

and

d r r0

(4)
q
(5) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(6) d =
d
The first assumption is that the overall size of the charge distribution is
much smaller than the distance to the observer, so that

0
(7) rmax r
This allows us to approximate by saving only up to first order terms in r0 .
0
The second approximation is that rmax is much less than all the terms
1
ELECTRIC DIPOLE RADIATION FROM AN ARBITRARY SOURCE 2

0 c
(8) rmax 
|(d n /dt n ) /|1/(n1)
for n 2. For an oscillating system, (t) = A cos t so

dn
(9) = (1)n n (t)
dt n

so
n 1/(n1)
1 d
(10)
dt n =

0
and this assumption is equivalent to rmax  that we made in analyzing
the oscillating dipole. In practice, it means that we keep up to first order
terms in r0 .
After making these two assumptions, we arrive at approximate formulas
for the potentials:
 
1 Q r p (t r/c) r p (t r/c)
(11) V (r,t)
= + +
40 r r2 rc
0 p (t r/c)
(12) A (r,t)
=
4 r

where p is the dipole moment



r0 r0 ,t r/c d 3 r0

(13) p=

and Q is the total charge in the system.


By making a further assumption that r itself is very large (essentially
approaching infinity, since ultimately we are interested only in radiation
that makes it to infinity), we arrive at approximate formulas for the fields:

0
(14) E (r,t)
= [r (r p)]
4r
0
(15) B (r,t)
= (r p)
4rc

where in both cases p is evaluated at the retarded time t r/c. Note that
the fields depend on the second time derivative of the dipole moment, which
ELECTRIC DIPOLE RADIATION FROM AN ARBITRARY SOURCE 3

means that no radiation is produced unless the charges are accelerating. The
only way to accelerate something is, of course, to apply a force to it so we
are doing work on the system, and this work is being converted (at least
partly) into radiation.
If we use spherical coordinates with the z axis in the direction of p then
the fields can be written as

0 p (t r/c)
(16) E (r,t)
= sin
4r
0 p (t r/c)
(17) B (r,t)
= sin
4rc
The Poynting vector is

1 0 p2 sin2
(18) S= EB
= r
0 16 2 c r2

and the total radiated power is



0 p2
(19) P= S da
=
6c
Example. We can apply these formulas to the case of the rotating dipole.
In that case, we had a dipole rotating in the xy plane, so its dipole moment
is given by

(20) p (t r/c) = p0 (cos (t r/c) x + sin (t r/c) y)


Therefore

(21) p = p0 2 (cos (t r/c) x + sin (t r/c) y)

so from 14 and 15 we get

0 p0 2
(22) E (r,t)
= [r (r (cos (t r/c) x + sin (t r/c) y))]
4r
0 p0 2
(23) B (r,t)
= (r (cos (t r/c) x + sin (t r/c) y))
4rc

which are the same equations we got earlier (after swapping the orders of
the cross products):
ELECTRIC DIPOLE RADIATION FROM AN ARBITRARY SOURCE 4

0 p0 2 h   r    r   i
(24) E = cos t x + sin t y r r
4r c c
0 p0 2 h   r    r   i
(25) B = cos t x + sin t y r
4rc c c
In this case, its not convenient to use the spherical coordinate forms for
the fields, since the direction of the dipole moment (and hence its second
derivative) is changing with time. However, since the power 19 is obtained
by integrating over all angles, it does give the same result, since

(26) p2 = p20 4
0 p20 4
(27) P =
6c

which is the same as we got earlier.


P INGBACKS
Pingback: Power radiated by a spinning ring of charge
Pingback: Radiation from a charge falling under gravity
Pingback: Electric quadrupole radiation
Pingback: Radiation from a charge on a spring
Pingback: Radiation from a point charge near a conducting plane
POWER RADIATED BY A SPINNING RING OF CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 9.
Heres a generalization of the rotating dipole problem we did earlier. This
time we have a circular ring with radius b with a linear charge distribution,
at t = 0, of

(1) = 0 sin

where is the azimuthal angle. The disk is set spinning with an angular
velocity of .
Because sin ( + ) = sin , this disk is essentially a collection of dipoles
with charges b d separated by distance 2b. Therefore, at time t, the
dipole moment is

2
(2) p (t) = 2b 0 sin [cos (t + ) x + sin (t + ) y] d
0

(3) = 2b2 0 ( sin t x + cos t y)
2
(4) p (t) = b 0 2 ( sin t x + cos t y)
2
2
(5) p2 = b2 0 2
The total power radiated is therefore

0 p2 0 02 4 b4
(6) P= S da
= =
6c 6c
Calculating the fields and Poynting vector is more complicated, as they
both change with time.

1
RADIATION FROM A CHARGE FALLING UNDER GRAVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 10.
If a charge falls under the influence of gravity, it accelerates and therefore
radiates. This means that not all of the potential energy lost as the charge
falls is converted to kinetic energy, so a charged object falls more slowly
than an uncharged one. Will this difference be noticeable?
Suppose we drop a single electron from rest at z = 0. After it has fallen
to a position z, its dipole moment is

(1) p = ezz

(The dipole moment is in the +z direction since the electrons charge is


negative and it falls to a point z < 0.) The power radiated is

0 p2
(2) P
=
6c

where

(3) p = ez
(4) = eg

so

0 e2 g2
(5) P= = 5.7 1052 J s1
6c

which is a constant.
To find how much energy is radiated as the electron falls, say, 1 cm, we
need to know how long it takes the electron to fall 1 cm. If all its lost
potential energy were converted to kinetic, then we get v = gt and d = 12 gt 2 .
Since the power is very small, its a safe bet that very little of the energy is
1
RADIATION FROM A CHARGE FALLING UNDER GRAVITY 2

radiated, so we can assume that d = 21 gt 2 and then check that our answer is
consistent. From this we get
r
2 0.01
(6) t= = 0.045 s
9.8

so the total energy radiated is

(7) Pt = 2.46 1053 J


The potential energy lost is

(8) V = mgh = 0.01mg = 8.92 1032 J

so the fraction of potential energy radiated is

Pt
(9) = 2.76 1022
V
So hardly any of the energy is radiated.
ELECTRIC QUADRUPOLE RADIATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 11.
In analyzing radiation from an arbitrary configuration of charge, we made
the assumption that the maximum dimension of the source is much smaller
than the observation distance, so that we can retain only first order terms
in r0 , the variable that is integrated over the source. In some cases, the first
order contribution is zero and in that case, we need to look at the next order.
This leads to electric quadrupole (and magnetic dipole, but well leave that
for now) radiation. A simple model that illustrates this is as follows.
Suppose we have two oscillating electric dipoles situated on the z axis,
with p+ at z = + d2 and p at z = d2 . The dipole oscillate exactly out of
phase, so that the dipole moment of the upper dipole is always the negative
of the dipole moment of the lower one. We can work out the fields of this
setup by using the same approximations we used in deriving the ordinary
oscillating dipole. First, we need to define a few terms. (Id draw a diagram,
but thats a painful process, so bear with me.)
Let the observation point r make an angle with the z axis, and let the
vector from p+ to r be r+ and the vector from p to r be r . The vectors
r make angles with the z axis.
The potential formulas for a dipole at the origin are

p0 cos   r 
(1) V (r, ,t) = sin t
40 rc c
0 p0   r 
(2) A (r, ,t) = z sin t
4r c
However, here, the dipoles are not at the origin so we need to adapt these
formulas. For p+ we must use r+ and + so we have

p0 cos +   r 
+
(3) V+ = sin t
40 r+ c c
From the law of cosines we have
r
d2 d
(4) r+ = r2 + 2r cos
4 2
1
ELECTRIC QUADRUPOLE RADIATION 2

and from the geometry of the setup

d
(5) r cos = r+ cos + +
2
Now assuming d  r we have
 
d
(6) r+ r 1 cos
2r
 
d d
(7) r cos r 1 cos cos + +
2r 2
r cos d2
(8) cos +
r d2 cos
  
1 d d
(9) r cos 1 + cos
r 2 2r
d
cos2 1

(10) cos +
2r
d
(11) = cos sin2
2r
Also,

  r    
+ r  d
(12) sin t sin t + cos
c c 2c
h  r i d h  r i
(13) sin t + cos cos t
c 2c c

to first order in d.
Plugging these into 3 we get

(14)
  
p0 d 2 d
V+ = cos sin 1 + cos
40 cr 2r 2r
 h  
r i d h  r i
sin t + cos cos t
c 2c c
(15)
 h  r i
p0 h  r i d h  r i d 2
cos sin t + sin t cos 2 + cos cos t
40 cr c 2r c 2c c
Under our approximation of d  r we can drop the middle term to get
ELECTRIC QUADRUPOLE RADIATION 3

(16) 
p0 h  r i d h  r i
2
V+ cos sin t + cos cos t
40 cr c 2c c
For p we can do the same calculation to get (note the opposite sign of
p0 since the dipole is opposite to the top one)

p0 cos   r 

(17) V = sin t
40 r+ c c
 
d
(18) r r 1 + cos
2r
d
(19) cos cos sin2
2r
  r  h  r i d h  r i

(20) sin t sin t cos cos t
c c 2c c
Putting this together, we get
 h  r i
p0 h  r i d
2
(21) V cos sin t cos cos t
40 cr c 2c c
The total potential is

(22) V = V+ +V
p0 2 d 2
h  r i
(23) = cos cos t
40 c2 r c
0 p0 d2 h  r i
2
(24) = cos cos t
4r c

using c2 = 1/0 0 .
For the vector potential, we get

0 p0   r 
+
(25) A+ = z sin t
4r+ c
     r 
0 p0 r  d cos
(26) z sin t cos t
4r c 2c c
   
0 p0 r  d cos   r 
(27) A z sin t + cos t
4r c 2c c
0 p0 2 d   r 
(28) A z cos cos t
4rc c
ELECTRIC QUADRUPOLE RADIATION 4

With the potentials, we can calculate the fields. To simplify the notation,
well use the shorthand
  r 
(29) c cos t
c
(30) c cos

and so on.

A
(31) E = V
t
2
0 d p0 2 0 p0 d 2  s c 

(32) = c c r + c s
4r2 4cr 2r2
Using the approximation r  c/ we can drop all but one term to get

0 p0 d 3
(33) E c s s
4rc
For the magnetic field, we get

(34) B = A
0 p0 d 2  s
c 
(35) = c s
rc 4c 2r
Again, using the approximation r  c/ we drop the second term to get

0 p0 d 3
(36) B c s s
4rc2
The Poynting vector is

1
(37) S = EB
0
2
0 p0 d 3

(38) (c s s )2 r
c 4rc
The intensity is the time average of S:

0 p20 d 2 6
(39) hSi = 2 3 2
(c s )2 r
32 c r

and the power is the integral of this over a sphere of radius r:


ELECTRIC QUADRUPOLE RADIATION 5


(40) hPi = S da

0 p20 d 2 6
(41) = 2 cos2 sin3 d
32 2 c3 0
0 p20 d 2 6
(42) =
60c3
RADIATION FROM A CURRENT LOOP WITH TIME-VARYING
CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 12.
Weve looked at the fields produced by a magnetic dipole that oscillates
with a regular frequency . By following the procedure in Griffithss sec-
tion 11.1.4, where he derives the fields due to an electric dipole of arbitrary
shape, we can derive the formulas for a magnetic dipole consisting of a cir-
cular current loop carrying a time-dependent current I (t) where the time
dependence is arbitrary.
We assume the current loop has radius b and lies in the xy plane with its
centre on the z axis. Since at any instant, the magnitude of the current is
the same everywhere in the loop, we can use the same argument as in the
oscillating case to deduce that for some observation point r in the xz plane,
the vector potential A points in the y direction, and thus, since A is always
tangential to the loop, its direction in general is in the direction. If the
loop is electrically neutral, the electric potential is V = 0, so we need to
calculate only A. The retarded potential is


0 I (t d/c) 0
(1) A (r,t) = d`
4 d
2
0 b I (t d/c)
(2) = cos 0 d 0
4 0 d

where 0 is the azimuthal angle around the loop so that the y component of
d`0 is b cos 0 and the retarded time is

d
(3) tr t
c

and
1
RADIATION FROM A CURRENT LOOP WITH TIME-VARYING CURRENT 2

d r r0

(4)
q
(5) = (x x0 )2 + (y y0 )2 + (z z0 )2
r r0
(6) d =
d

where r0 is the position on the loop being integrated over.


For our observation point in the xz plane, we have

(7) r = r sin x + r cos z

and for a point on the loop

(8) r0 = b cos 0 x + b sin 0 y


In what follows, well use the notation c cos , s sin , etc to
simplify the notation.
Therefore, assuming b  r (the loop is very small)
p
(9) d = r2 + b2 2r r0
 
b
(10) = r 1 s c 0
r
 
1 1 b
(11) = 1 + s c 0
d r r
We can expand the current in a Taylor series about t0 t cr :
   
d r b
(12) I t = I t + s c 0
c c c
 2
b 1 b
(13) = I (t0 ) + I (t0 ) s c 0 + I (t0 ) s c 0 + . . .
c 2! c
We are justified in dropping the last term if

 2
1 b b
(14) I (t0 ) s c 0  I (t0 ) s c 0
2! c c
c
(15) b 
I/I
RADIATION FROM A CURRENT LOOP WITH TIME-VARYING CURRENT 3

If we compare higher derivative terms with the first order term, we get
the general condition

I n1

(16) b  c n

d I/dt n

Assuming this is true, we get from 2:

2   
0 b b b
(17) A (r,t)
= I (t0 ) + I (t0 ) s c 0 1 + s c 0 c 0 d 0
4r 0 c r
0 b I (t0 ) I (t0 )
2  
(18)
= + s
4r r c

2
where to get the second line, we discarded the term in b3 and used 0 cos 0 d 0 =
2
0 and 0 cos2 0 d 0 = . If were interested only in the radiation produced
by this dipole, we can ignore any terms in the potential that are of order 2
or higher in 1r , since it is only r12 terms in the Poynting vector that will con-
tribute to radiation that escapes to infinity. Therefore, we can throw away
the first term above to get our final approximation:

0 b2
(19) A (r,t) = I (t0 ) s
4rc

We can write this in terms of the magnetic moment of the loop, which is

(20) m (t0 ) = b2 I (t0 )

so

0
A (r,t)
= m (t0 ) s
4rc

We can now calculate the fields:


RADIATION FROM A CURRENT LOOP WITH TIME-VARYING CURRENT 4

A
(21) E =
t
0 s
(22) = m (t0 )
4rc
(23) B = A
1
(24) = (rA)
r r
0 s m
(25) =
4rc r
1
where in calculating B, we ignored the term rs (s A) r since it gives a
term containing r12 .
Since m = m t cr we have


m 1
(26) = m
r c
0 s
(27) B = m
4rc2
The Poynting vector is

1
(28) S = EB
0
0 s2 m2
(29) = r
16 2 r2 c3
The power radiated is the integral of S over a large sphere of radius r:

(30) P = S da

2 0 m2
sin2 2
(31) = r sin d
16 2 c3 0 r2
0 m2
(32) =
6c3
P INGBACKS
Pingback: Radiation from the magnetic dipoles of Earth and pulsars
RADIATION FROM A POINT CHARGE; THE LARMOR
FORMULA

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 13.
Weve seen that the fields produced by a moving point charge are

qr
c2 v2 u + r (ua)
  
(1) E (r,t) = 3
40 (r u)
1
(2) B (r,t) = r E (r,t)
c

where r is the vector from the charge to the observation point r and u =
cr v.
If were interested only in the power radiated by a moving point charge,
we need keep only those terms in the fields that depend on 1r and discard
higher order terms such as r12 . This is because the power radiated depends
on the product of the fields via the Poynting vector, and any terms in the
Poynting vector of order r13 or higher will go to zero when integrated over
a large sphere. Its actually more convenient to use a coordinate system
centred on the moving charge, so that r and r are actually the same, and
r becomes the radius of the enclosing sphere. Looking at 1, we see that
there is a factor of order r12 out front, and of the two terms inside the square
brackets, only the second term contains another factor of r. Combining
these two terms means that the only term that will contribute to radiation
from the point charge is the second one, so we have

qr
(3) Erad = r (ua)
40 (r u)3
As it depends on the acceleration and results in radiation, this term is
known as either the acceleration field or radiation field. The first term in 1
also results in energy flux generated by the moving charge, but as it goes as
1/r2 it is a localized flux, and does not contribute to radiation as it drops to
zero for large r.
From here, the derivation of the formula for the total power radiated is
fairly straightforward and is given in detail by Griffiths in his section 11.2.1.
1
RADIATION FROM A POINT CHARGE; THE LARMOR FORMULA 2

The only further assumption that is made is that v = 0 at the particular


instant of time that were considering. [Recall that it is possible for the
velocity of a particle to be zero while its acceleration is non-zero, as with
a mass oscillating on a spring when it reaches the high and low points of
its trajectory.] This assumption gives the Larmor formula for the power
radiated by a point charge:

0 q2 a2
(4) P=
6c
Note that this formula has an  implicit time dependence through the ac-
r
celeration a = a (tr ) = a t c , although the other parameters are all con-
stants. Since were dealing with a point charge, there is only one retarded
time that we need to keep track of. The radiation detected at the enclosing
sphere, which is centred on the charge (so it moves as the charge moves) is
the radiation that left the charge at time t cr .
Although the Larmor formula was derived by assuming that v = 0 it is
actually a good approximation for all non-relativistic speeds.
Example. As an example, suppose we have an electron moving at a thermal
speed of v0 = 105 m s1 (so v  c) within a solid, such as a metal, and
by colliding with an atom it experiences a constant deceleration so that it
comes to rest after travelling d = 3 109 m. To find the power radiated
by the electron using the Larmor formula, we need to know how long the
deceleration takes. Since a is constant, we have

v0
(5) t=
a
The total energy radiated over this time is

(6) E = Pt
0 q2 a2 v0
(7) =
6c a
0 q2 av0
(8) =
6c
The fraction of the electrons initial kinetic energy radiated away is

2E
(9) f =
mv20
0 q2 a
(10) =
3mcv0
RADIATION FROM A POINT CHARGE; THE LARMOR FORMULA 3

To find a we use the formula

1 2 v20
(11) d = at =
2 2a
v20
(12) a =
2d
(13) = 1.67 1018 m s2
[Yes, thats an enormous deceleration!]
Plugging in the other constants in 10 we get

(14) f = 2 1010
Thus the amount of energy lost to radiation due to electronic collisions is
very small.
P INGBACKS
Pingback: Radiative decay of the Bohr hydrogen atom
Pingback: Linards generalization of the Larmor formula for an accel-
erating charge
Pingback: Radiation reaction: a few examples
Pingback: Radiation reaction: the Abraham-Lorentz force
Pingback: Fields of a moving magnetic monopole
Pingback: Radiation reaction: energy conservation with a constant exter-
nal force
Pingback: Radiation reaction with a delta-function external force
RADIATIVE DECAY OF THE BOHR HYDROGEN ATOM

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 14.
The instantaneous power radiated by an accelerating point charge is given
by the Larmor formula (valid for a charge moving at a speed v  c):

0 q2 a2
(1) P=
6c
One historic application of this formula was to Bohrs early model of
the hydrogen atom as an electron in a classical circular orbit around the
proton, with the centripetal force provided by the Coulomb attraction. Since
a particle moving in a circle is accelerating, it will radiate away energy, so its
orbit should eventually decay until the electron crashes into the proton. This
classical instability of atoms was one motivation behind the introduction of
quantum theory, but thats another story. Here, well investigate how long
it would take a Bohr hydrogen atom to decay.
First, we need to reassure ourselves that the electron is moving non-
relativistically. From equating centripetal and Coulomb forces, we have

mv2 q2
(2) =
r 40 r2
q
(3) v =
40 mr
r
0
(4) = qc
4mr

using c = 1/ 0 0 . Plugging in the numbers we get

v 5.3 108
(5) =
c r
The Bohr radius is a = 5 1011 m so at that radius v/c = 0.0075 so
were safe here. As r gets smaller, of course, v will increase but since the
dependence is on the square root, the rate of increase of v is fairly small, so
1
RADIATIVE DECAY OF THE BOHR HYDROGEN ATOM 2

that even when r = a/100, v has increased only to 0.075c. So for most of
its journey towards the proton, the electron is moving non-relativistically.
To work out how long it takes for the decay to occur, consider the energy
radiated during a time dt, which is P dt. From conservation of energy, this
must be equal to the amount of energy lost by the electron. The total energy
of the electron is its kinetic plus potential energy, so

1 2 q2
(6) E = mv
2 40 r
2 2
q c 0 q2 c2 0
(7) =
8r 4r
2 2
q c 0
(8) =
8r

Therefore, the energy lost is

q2 c2 0
(9) dE = dr
8r2

[Weve taken dE as negative, since the electron loses energy.] Putting


these results together, we get

0 q2 a2
(10) Pdt = dt
6c
q2 c2 0
(11) = dr
8r2
3c3
(12) dt = 2 2 dr
4r a

We need a to solve this, but this is just the centripetal acceleration, so

v2
(13) a =
r
q2 c2 0
(14) =
4mr2

Therefore,
RADIATIVE DECAY OF THE BOHR HYDROGEN ATOM 3

12 2 m2 2
(15) dt = r dr
q4 c02
(16) = 3.159 1020 r2 dr
T 0
20
(17) dt = 3.159 10 r2 dr
0 a
1 3
3.159 1020 5 1011

(18) =
3
(19) T = 1.31 1011 s
Thus in classical electrodynamics, the hydrogen atom is so unstable that
it would decay in a tiny fraction of a second. We should be grateful that
quantum mechanics saves the universe.
LINARDS GENERALIZATION OF THE LARMOR FORMULA
FOR AN ACCELERATING CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 15.
For an accelerating charge q that is instantaneously at rest, the power
radiated is given by the Larmor formula:

0 q2 a2
(1) P=
6c

where the acceleration a is a function of time. Griffiths shows in his section


11.2.1 that the Larmor formula is the integral of the Poynting vector over
a large sphere. Another way of looking at it is that the charge radiates an
amount of power dP into an element of solid angle d = sin d d and
the formula for this is obtained from the Poynting vector by multiplying by
r2 to make the result independent of distance from the charge.

dP 0 q2 a2 2
(2) = sin
d 16 2 c
We can see from this formula that the angle at which the maximum power
is radiated is max = 2 .
To generalize these formulas to the case where v 6= 0 requires a bit of a
slog through the mathematics, but the results are quoted by Griffiths as his
equations 11.72 and 11.73:

dP 0 q2 |r (u a)|2
(3) =
d 16 2 c2 (r u)5
" #
0 q2 6 2 |v a|2
(4) P = a
6c c2

where r is a unit vector pointing from the charge at the retarded time to the
observation point on the enclosing sphere and u = cr v. These formulas
are Linards generalization of the Larmor formula.
1
LINARDS GENERALIZATION OF THE LARMOR FORMULA FOR AN ACCELERATING CHARGE
2

For a charge with v k a (at the instant of retarded time), Griffiths shows
in his Example 11.3 that

dP 0 q2 a2 sin2
(5) =
d 16 2 c (1 cos )5

where v/c. This formula reduces to 2 when = 0.


We can find the angle max for the case 6= 0 by differentiating 5 and
setting the result to 0.
" #
d dP 0 q2 a2 sin ( ) cos ( ) (sin ( ))3
(6) = 2 5 =0
d d 16 2 c (1 cos ( ))5 (1 cos ( ))6
The solution = 0 gives the angle of minimum power, so if we take 6= 0
we can cancel off sin and then multiply through by (1 cos ( ))6 to get
 
(7) 2 (1 cos ( )) cos ( ) 5 1 (cos ( ))2 = 0

which has the solution


"p #
15 2 + 1 1
(8) max = arccos
3
In the ultrarelativistic case, we can write

(9) = 1x

where x  1 and expand in a Taylor series (using Maple to do the heavy


lifting):
q
15 (1 x)2 + 1 1
p
15 2 + 1 1
(10) =
3 3 (1 x)
x 27 2
x + O x3

(11) = 1
4 128
Since x is small, the series on the right is very close to 1, which means
that the argument of the arccos is close to 1, so max is close to 0, so we can
approximate
LINARDS GENERALIZATION OF THE LARMOR FORMULA FOR AN ACCELERATING CHARGE
3

2
max
(12) cos max 1
2
x
(13) 1
r 4
x
(14) max
2
r
1
(15) =
2
To compare the power output at the maximum angles in the two cases,
we have from 2

dP 0 q2 a2
(16) =
d rest 16 2 c

and from 5

dP 0 q2 a2 sin2 max
(17) =
d rel 16 2 c (1 cos max )5

so the ratio is

dP/drel sin2 max


(18) =
dP/drest (1 cos max )5
q
1
For max 2  1 we can approximate

1
(19) sin2 max
2
x
(20) =
2  
1
(21) 1 cos max 1 1
4
 x
(22) 1 (1 x) 1
4
5x
(23)
4
Therefore
LINARDS GENERALIZATION OF THE LARMOR FORMULA FOR AN ACCELERATING CHARGE
4

dP/drel 45 x
(24)
dP/drest 2 (5x)5
512
(25) =
3125 (1 )4
p
To express this in terms of the relativistic factor = 1/ 1 2 we can
approximate for = 1 x:

1
(26) = q
1 (1 x)2
1
(27) =
2x x2
1
(28)
2x
1
(29) = p
2 (1 )
Therefore

1
(30) 4
16 8
(1 )
dP/drel
(31) 2.62 8
dP/drest
Since gets very large for 1, the power generated by a relativistic
charge is enormously greater than that of a charge at rest.
P INGBACKS
Pingback: Synchrotron radiation
Pingback: Power radiated by radiation reaction force
Pingback: Radiation from a charge in hyperbolic motion
SYNCHROTRON RADIATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 16.
One common instance of an accelerated charge is a charge moving in a
circle. In this case the particles instantaneous velocity v is always perpen-
dicular to its instantaneous acceleration a. This is known as synchrotron
radiation, since it is the radiation given off by particles in a synchrotron
particle accelerator, where charged particles move in circular orbits between
the poles of a magnet.
We can use the Linard formulato work out the power radiated by such a
charge:

dP 0 q2 |r (u a)|2
(1) =
d 16 2 c2 (r u)5
" #
0 q2 6 2 |v a|2
(2) P = a
6c c2

At one instant of time, we can take

(3) v = vz
(4) a = ax
(5) r = s c x + s s y + c z
(6) u = cr v

where were using our usual shorthand for trig functions: s sin , c
cos and so on. We can now work out the components of 1:
1
SYNCHROTRON RADIATION 2

(7) r u = cr r vc
(8) = c vc
(9) = c (1 c )
(10) u a = cr ax vz ax

(11) = ca s s z + c y avy

x y z

(12) r (u a) = s c
s s c

0 cac av cas s
h i
(13) = x cas2 s2 c (cac av) +
ycas2 s c + z cas c c avs c


Taking the square of this last vector leads to a lengthy expression which
can be simplified by applying s2 +c2 = 1 repeatedly. We get, using = v/c:

(14)
1
|r (u a)|2 = s4 c4 + 1 + c2 2 2s2 c2 + 2s2 c2 c 2c +
a2 c2
s4 s2 c2 + s2 c2 c2 2s2 c2 c + s2 c2 2
We can simplify this as follows. The first and seventh terms combine to
give
 
(15) s4 c4 + s4 s2 c2 = s4 c2 c2 + s2 = s4 c2

Combining this with the eighth term:

s4 c2 + s2 c2 c2 = s2 c2 s2 + c2 = s2 c2

(16)
Combining this with the fourth and last terms we get

s2 c2 2s2 c2 + s2 c2 2 = 1 2 s2 c2

(17)
The second, third and sixth terms combine to give

(18) 1 + c2 2 2c = (1 c )2
Finally, the fifth and ninth terms cancel, so were left with
SYNCHROTRON RADIATION 3

1 2 2 2 2 2

(19) |r (u a)| = (1 c ) 1 s c
a2 c2
Putting everything together we get

2 2 2 2

dP 0 q2 a2 (1 c ) 1 s c
(20) =
d 16 2 c (1 c )5
To get the total power, we need to integrate this over all solid angles, so
we get

2 (1 c )2 1 2 s2 c2

0 q2 a2
(21) P= d d s
16 2 c 0 0 (1 c )5
The integral over is easy, using

2
(22) c2 d =
0

so were left with the integral over :


2 (1 c )2 1 2 s2

0 q2 a2
(23) P = d s
16c 0 (1 c )5

2 (1 c )2 1 2 1 c2
 
0 q2 a2
(24) = d s
16c 0 (1 c )5
This nasty looking integral can be done by using partial fractions, since
it is the ratio of two polynomials in c . I did the integral using Maple, but if
youre interested in doing it by hand, the partial fraction decomposition is

(25)
2 (1 c )2 1 2 1 c2 4 22 +1 2 1 2 +1
    
= +2
(1 c )5 2 ( cos ( ) 1)5 2 ( cos ( ) 1)4 ( cos ( ) 1)3
The presence of the extra sin from the solid angle element saves the day,
since it multiplies each term in the partial fraction expansion, providing the
derivative of cos on the top of each fraction. For example

sin 1
(26) d =
( cos ( ) 1) 5
4 ( cos ( ) 1)4
SYNCHROTRON RADIATION 4

with the other two terms having similar integrals.


The result of the integral is

2 (1 c )2 1 2 1 c2

 
8
(27) d s =
0 (1 c )5
3 (1 )2 (1 + )2
8
(28) = 2
3 (1 2 )
8 4
(29) =
3

so we get for the total power

0 q2 a2 8 4
(30) P =
16c 3
0 q2 a2 4
(31) =
6c
RADIATION REACTION: A FEW EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 17.
Here are a few simple examples of the radiation reaction force, calculated
using the Abraham-Lorentz formula.
Example 1. A charge q moves in a circle of radius R with constant speed
v. The reaction force is given by

0 q2
(1) Frad = a
6c
For uniform circular motion, the acceleration is always directed towards
the centre of the circle, so

v2
(2) a = r
R
v2
(3) = 2r
R
v2
(4) a = 2 v
R
0 q2 v2
(5) Frad = v
6cR2
The force we need to apply to counter the reaction force is just the nega-
tive of this, so

0 q2 v2
(6) Fe = v
6cR2
The power generated by applying this force is

0 q2 v4
(7) Pe = Fe v =
6cR2
The Larmor formula for radiated power is
1
RADIATION REACTION: A FEW EXAMPLES 2

0 q2 a2
(8) P =
6c
0 q2 v4
(9) =
6cR2
(10) = Pe
So the power we must exert to counter the reaction force is equal to the
power radiated away.
Example 2. Now suppose we have a charge on a spring which moves with
simple harmonic motion according to

(11) r (t) = Az cos t


(12) v (t) = A z sin t
(13) a (t) = A 2 z cos t
(14) a (t) = A 3 z sin t
The reaction force is now

0 q2 A 3
(15) Frad = z sin t
6c

and the force Fe we need to apply to counter it is

0 q2 A 3
(16) Fe = z sin t
6c

giving a power of

(17) Pe = Fe v
0 q2 A2 4 2
(18) = sin t
6c
The radiated power is

0 q2 a2
(19) P =
6c
0 q2 A2 4
(20) = cos2 t
6c
RADIATION REACTION: A FEW EXAMPLES 3

Thus in this case, the applied power is not equal to the radiated power at
each instant of time, but remember that the Abraham-Lorentz formula was
derived by taking the average power over a period of time after which the
system returns to its initial state. If we average these two powers over one
cycle, we get
2/
0 q2 A2 4
(21) hPe i = sin2 t
6c 2 0
0 q2 A2 4
(22) =
12c
(23) = hPi
Thus, on average, the powers are equal.
Example 3. For a charge falling in a constant gravitational field with accel-
eration g, a = 0 so the reaction force is zero. However, since the acceleration
is not zero, the charge does radiate with a power of

0 q2 g2
(24) P=
6c
In this case, there is no time interval after which the charge returns to its
initial state, so we cant average the power over any time interval and the
Abraham-Lorentz formula isnt valid.
RADIATION DAMPING

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 18.
The radiation reaction force is given by the Abraham-Lorentz formula

(1) Frad = m a

where

0 q2
(2)
6mc
The force gives rise to radiation damping. A simple example is that of a
charge on a spring with natural frequency 0 . If the charge is subject to a
driving force that forces the charge to oscillate with frequency then from
Newtons law

(3) mx = Fspring + Frad + Fdriving


...
(4) = m02 x + m x + Fdriving
Since the charge is being forced to move sinusoidally with frequency
we have

(5) x (t) = x0 cos (t + )

where x0 is the amplitude and is the phase. Thus

...
(6) x = x0 3 sin (t + )
(7) = 2 x
We can rewrite 4 as

(8) mx = m02 x m 2 x + Fdriving


(9) Fdriving = mx + m02 x + m 2 x
1
RADIATION DAMPING 2

Typically, a damping force (such as the drag force from travelling through
a viscous fluid, as given by Stokess law) is proportional to the velocity, so
we can write the radiation damping term as

(10) Frad = m 2 x = m x

where is the damping factor:

(11) = 2
Example. If we return to the case of resonances in a dispersive medium, we
can investigate the effect of radiation damping on the motion of an electron
driven by an electromagnetic wave in the optical region. We found that the
index of refraction in such a medium tends to increase with frequency, ex-
cept near certain resonance frequencies where the index of refraction drops
briefly before resuming its rise after the resonance frequency is passed. The
width of this anomalous dispersion region is approximately , pro-
vided that  0 (the resonance frequency).
For an electron,
2
4 107 1.6 1019

(12) = = 6.24 1024 s
6 (9.11 1031 ) (3 108 )
Taking the optical region to be green light, the frequency is about =
6 1014 s1 , so = 2 = 3.77 1015 s1 , so

(13) = 8.87 107 s1


This is much less than in the optical region.
Using the simple model of the hydrogen atom as an electron embedded
in a uniform sphere of charge, we found that its resonant frequency was

(14) 0 = 4.13 1016 s1

so  0 for an electron. [0 here is in the near ultraviolet.] The width of


the anomalous dispersion region in this model is therefore

(15) = 02 = 1.06 1010 s1


Compared to the resonant frequency, is very small.
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 19.
If a charged particle accelerates, it radiates away energy. This means that
if an external force is applied to a charge, not all of the energy transferred
to the charge by the force is converted to the kinetic energy of the charge;
some of the energy is radiated away in the form of electromagnetic waves.
Looked at another way, the charge accelerates less than an uncharged par-
ticle of the same mass. From Newtons law F = ma, the net force on the
charge must be less than the applied external force. In effect, the fields
surrounding the charge exert a recoil or reaction force on the charge.
Since the power radiated by a charge (in the non-relativistic case, any-
way) is given by the Larmor formula, we might expect that this formula
could be used to work out the reaction force, Frad . If this force acts on
the charge as it moves a distance r, the work done is Frad r, so the rate
at which this force does work, which is the power lost to radiation, is the
time derivative of this, or Frad v. The problem with this argument is that
the Larmor formula measures only that radiation that extends out to infinity.
The fields of a moving point charge are (with u = rc v)

qr
c2 v2 u + r (ua)
  
(1) E (r,t) = 3
40 (r u)
1
(2) B (r,t) = r E (r,t)
c

and it is only those terms that go as 1r that contribute to radiation that is mea-
sured by the Larmor formula. The other term, namely qr c2 v2 u/40 (r u)3


(and its counterpart in B), falls off as 1/r2 so contributes nothing to the in-
tegral of the Poynting vector over a large sphere, since it gets multiplied
into either another term in 1/r2 or the term in 1/r, giving terms in 1/r4 or
1/r3 . This 1/r2 term is called the velocity field (for lack of a better name;
its a bit misleading as both the velocity field and the acceleration field de-
pend on velocity, but never mind) and, although it doesnt contribute to the
radiation in the Larmor formula, it does store energy, so some of the energy
1
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 2

imparted by the force that gets the charge moving must be siphoned off to
create these velocity fields.
These velocity fields are curious beasts, however, for they contain energy
that is never actually lost to the charge. If a charge is accelerated to some
velocity, the velocity fields are constructed around the moving charge, but
if the charge is then decelerated to rest again, the velocity fields disappear
without having radiated away any energy. Where does this energy go? Grif-
fiths doesnt address this point, but it would seem to be reabsorbed by the
charge as it slows down.
In any case, if we look at a charged particle that starts off in some state,
then goes through an acceleration followed by a deceleration, and finally
ends up in the same state that it started from, what we can say is that the
velocity fields are the same at the end as they were at the start, so over this
period, the only energy that is truly lost from the particle is the energy that
is radiated away, that is, the energy measured by the Larmor formula, which
states

0 q2 a2
(3) P=
6c
Therefore, if we average the energy over such a period, we get

t2 t2
(4) Frad vdt = Pdt
t1 t1

0 q2 t2 2
(5) = a dt
6c t1
t2 t2
2
(6) a dt = v vdt
t1 t1
t2
(7) = v v|tt21 v vdt
t1

The integrated term in the last line is zero because the charge is in the
same state at t1 and t2 so its velocity and acceleration are the same at those
two times. Combining the remaining integral with the LHS in 4 we get

t2  t2
0 q2

(8) Frad vdt = v vdt
t1 6c t1
t2 
0 q2

(9) Frad v vdt = 0
t1 6c
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 3

Since v = a, one possible solution of this equation is if the quantity in paren-


theses is identically zero, or

0 q2
(10) Frad = a
6c

This is known as the Abraham-Lorentz formula for radiation reaction.


Note that it depends on the rate of change of the acceleration, and not on
the acceleration itself, so the reaction force can be non-zero even when the
acceleration is momentarily zero, that is, at a time when the particle is not
radiating.
This formula has become infamous in classical electrodynamics as it
makes predictions which are seemingly at odds with common sense. One
argument goes as follows. If Frad is the only force acting on the particle
then from Newtons law

0 q2
(11) Frad = ma = a
6c
This differential equation has the general solution

(12) a (t) = a0 et/

where

0 q2
(13)
6mc
From this it seems that the acceleration spontaneously increases with
time. However, Im not sure why this particular point is considered to be
a problem, because if (as we assumed) the charge is subject to no exter-
nal force, then surely its velocity must be constant (from Newtons law) so
the acceleration and all its derivatives are zero, so Frad = 0 as well. This
amounts to taking a0 = 0 in the solution above. Problem solved. Or not
Before we do an example indicating the problem in more detail, first
consider the general case where a charge is subject to a reaction force and a
separate external force F. In that case
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 4

(14) ma = Frad + F
0 q2
(15) = a + F
6c
F
(16) a = a +
m
For an electrically neutral particle, the reaction term disappears, meaning
that if F is discontinuous (such as a force that is switched on at a certain
time), then so too is a. However, it turns out that the addition of the a term
guarantees that the acceleration is always continuous, even if the force is
not. Suppose we integrate this equation over some small time interval from
t to t + . Then we get
t+ t+ t+
1
(17) a dt = a dt + F dt
t t m t
t+
1
(18) = (a (t + ) a (t )) + F dt
m t
Provided both a and F are finite over the time interval, both their integrals
must tend to zero as 0. In particular, F is allowed to be discontinuous
(but it cant contain any infinities, such as a delta function). In that case

(19) lim (a (t + ) a (t )) = 0
0

so that a must be continuous. [The neutral particle case corresponds to q = 0


which means = 0 from 13. In that case, the term (a (t + ) a (t ))
disappears and the argument above fails.]
Example. Now for a specific example. Suppose that the external force F is
switched on at t = 0, remains constant until t = T and is then switched off.
To determine the acceleration we need to solve 16 subject to the boundary
condition that a is continuous everywhere. We have

t/ t <0
a0 e

(20) F
a (t) = a1 et/ + m 0<t <T
a et/

t >T
2

Now were faced with a problem. It would seem that since F = 0 for
both t < 0 and t > T , there should be no acceleration in either of these
zones. In particular, there clearly shouldnt be any acceleration for t < 0
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 5

since the force hasnt been switched on yet, and theres no way the charge
could know that a force is about to be switched on. So lets take a0 = 0.
From continuity at t = 0 we therefore must have

F
(21) a1 + = 0
m
F
(22) a1 =
m
At t = T, we therefore must have

F 
(23) a2 eT / = 1 eT /
m
F  T / 
(24) a2 = e 1
m
Having a2 6= 0 means that we get a spontaneously increasing (well, actu-
ally, increasing in the negative direction since eT / 1 < 0) acceleration
for t > T , even though there is no external force at this time.
If we try to eliminate this runaway acceleration by starting with a2 = 0
and working backwards, we get

F
(25) a1 eT / + = 0
m
F
(26) a1 = eT /
m
F 
(27) a0 = 1 eT /
m
Because a0 6= 0, the charge starts to accelerate before any force has been
applied (actually, it accelerates starting at t = which is clearly absurd).
Not only that, but the rate of acceleration depends on the length of time T
for which the force will be applied and the magnitude F of the force. This
preacceleration violates causality, so something is clearly wrong.
In the case where we eliminate the runaway solution (when a2 = 0), we
can work out the acceleration and velocity (by integrating the acceleration
and requiring continuity at all times):

 
F
1 eT / et/ t <0
m 


(28) a (t) = F 1 e(tT )/ 0<t <T
m

0 t >T

RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 6

 
F
1 eT / et/ t <0
m


(29) v (t) = F t e(tT )/ + F 0<t <T
m m

FT
m t >T

For an uncharged particle, the corresponding values are


0
t <0
(30) a (t) = mF
0<t <T

0 t >T


0
t <0
(31) v (t) = mF
t 0<t <T
FT

t >T
m

Plotting these results (for F = m = = T = 1) we get (red curves for the


charged particle; blue curves for the uncharged particle):
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 7

These dual problems of preacceleration and runaway acceleration have


plagued the subject ever since the formula was first proposed. Clearly there
is something wrong somewhere, but so far nobody has come up with a sat-
isfactory solution. There is an alternative formula originally devised by
Landau and Lifshitz as an approximation to the Abraham-Lorentz formula
which states simply that

(32) Frad = Fext


That is, the reaction force depends on the rate of change of the exter-
nal force, so that if there is no external force, there is also no reaction force.
This formula suffers from neither the preacceleration nor the runaway prob-
lems, but it raises the question of how an approximation can be more ac-
curate than the formula it is approximating. Griffiths, Proctor and Schoeter
consider the problem in more detail in a paper published after Griffithss
textbook (David J. Griffiths, Thomas C. Proctor and Darrell F. Schroeter,
American Journal of Physics, 78, 391 (2010)).

P INGBACKS
Pingback: Radiation reaction: a few examples
Pingback: Radiation damping
RADIATION REACTION: THE ABRAHAM-LORENTZ FORCE 8

Pingback: Abraham-Lorentz formula - physical basis


Pingback: Radiation reaction: energy conservation with a constant exter-
nal force
Pingback: Radiation reaction with a delta-function external force
Pingback: Tunnelling through a potential barrier with the radiation reac-
tion force
Pingback: Physical basis of the radiation reaction force
Pingback: Power radiated by radiation reaction force
ABRAHAM-LORENTZ FORMULA - PHYSICAL BASIS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 20.
Weve seen that the radiation reaction force is given by the Abraham-
Lorentz formula

0 q2
(1) Frad = a
6c
This formula was derived for the special case of a particle that returns to
its initial state after some time, and is calculated by taking the average work
done over that time period.
The formula is derived from the conservation of energy, but it doesnt
really tell us much about how the force originates. As with most things
about radiation reaction, this origin isnt particularly well understood, but
it appears that the force is a form of self-force, that is, a force felt by a
charge in motion due to the field it emitted at a previous time. If you think
of a charge moving along some trajectory, then after it has moved some
distance x, it will feel the effect of the field that it emitted at a previous
time of x/c, since the fields travel at the speed of light.
In his section 11.2.3, Griffiths derives the Abraham-Lorentz formula by
considering a charge q moving along the x axis. In order to deduce the
force of a charge on itself, he splits the charge into a dumbbell whose axis
is perpendicular to the direction of motion. There is a charge of q2 at each
end of the dumbbell, and these two half-charges are separated by a distance
d, which is imagined to tend to zero as we consider the limit of a point
charge. By considering the force of the bottom charge on the top charge, he
shows that this force comes out to

0 q2 a
(2) F=
24c
(This formula applies in the special case where the charge is instanta-
eously at rest at the retarded time when the field felt at the current time was
emitted.) By symmetry, there is an equal force exerted by the top charge on
the bottom one, so the total force due to these two interactions is double the
force above, so
1
ABRAHAM-LORENTZ FORMULA - PHYSICAL BASIS 2

0 q2 a
int
(3) Frad =
12c
However, there is also a force felt by each charge from its own field that
it emitted at the retarded time. Applying the formula 1, the self-force felt
by the upper charge is

0 (q/2)2 a 0 q2 a
(4) Fsel f1 = =
6c 24c
Again, from symmetry, there is an equal self-force contribution from the
lower charge, so

0 (q/2)2 a 0 q2 a
(5) Fsel f2 = =
6c 24c
(6) Fsel f = Fsel f1 + Fsel f2
0 q2 a
(7) =
12c
int
(8) Ftotal = Frad + Fsel f
0 q2 a
(9) =
6c
This derivation isnt very convincing, since we used the Abraham-Lorentz
formula to prove itself, so our reasing is circular. Another approach is to ap-
ply the original cross-charge force to a strip of charge smeared out over the
entire length L of the dumbbell. That is, for each segment dy on the dumb-
bell, we consider the force felt by fields emitted by charge segments from
the entire strip at the retarded time. If the linear charge density is then

L
(10) dy = L = q
0

If we now consider a pair of segments on this strip, then the charge on each
segment is dy, which replaces q/2 in 3, so the reaction force on these two
segments is

0 (q/2)2 a 0 2 a
(11) dF = = dy1 dy2
3c 3c
The total reaction force on the strip is
ABRAHAM-LORENTZ FORMULA - PHYSICAL BASIS 3

L L
1 0 2 a
(12) F = dy1 dy2
2 3c 0 0
0 (L )2 a
(13) =
6c
0 q2 a
(14) =
6c

where we introduced the factor of 12 in the first line to avoid double-counting


each pair of segments.
Although this derivation does give the Abraham-Lorentz formula, its
not a general proof, because if we use different configurations of charge
(such as a strip of charge whose axis is parallel to the direction of motion,
or a sphere of charge) we dont get the Abraham-Lorentz formula out of
the derivation. This is yet another problem with this formula to add to
preacceleration and runaway acceleration.
P INGBACKS
Pingback: Physical basis of the radiation reaction force
Pingback: Self-force on a dipole in hyperbolic motion
RADIATION FROM A CHARGE ON A SPRING

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 21.
Heres another example of radiation from an accelerated charge. We hang
a spring with spring constant k from the ceiling and attach a particle of
mass m and charge q to its lower end, which has an equilibrium distance
of h above the floor. If the particle is pulled down a distance d below the
equilibrium and released at t = 0, we wish to find the intensity (average
power per unit area) that hits the floor.
To solve this problem, we can use the Poynting vector for an arbitrary
charge distribution that we found earlier:

1 0 p2 sin2
(1) S= EB
= r
0 16 2 c r2
If we take the axis of the spring as the z axis and the equilibrium position
as z = 0 and z increasing downwards, the dipole moment of the charge is

(2) p (t) = qd cos t


(3) p (t) = qd 2 cos t
In this coordinate system, the angle is the angle between the z axis and
a point on the floor on a circle of radius R centred on the z axis. If the charge
is at z = 0 then

R
(4) sin =
R + h2
2
h
(5) cos =
R + h2
2

(6) r2 = R2 + h2
Therefore the Poynting vector is

0 q2 d 2 4 R2 cos2 t
(7) S= 2
r
16 2 c (R2 + h2 )
1
RADIATION FROM A CHARGE ON A SPRING 2

This gives the power per unit area and unit time radiated along the r direc-
tion, pointing radially away from the charge. If we want the power received
per unit area on the floor, we note that the angle between the normal to r
and the floor is equal to the angle between r and the normal to the floor, the
latter of which is the z axis. Thus the angle between the normal to r and the
floor is just , so that a unit area on the plane normal to r spreads out to an
area of 1/ cos when projected onto the floor. Therefore the power per unit
time per unit area of floor is

S
(8) S f loor =
1/ cos
0 q2 d 2 4 R2 cos2 t
(9) = 2
cos
16 2 c (R2 + h2 )
0 q2 d 2 4 R2 h cos2 t
(10) = 5/2
16 2 c (R2 + h2 )

Although we derived this for z = 0, it is a good approximation for all times


provided that d  h, so the amplitude of oscillation is very small, since in
that case the angle changes very little over a cycle.
The intensity of radiation on the floor is the time average of this over one
cycle, and since the average of cos2 t over a cycle is 12 , we get

0 q2 d 2 4 R2 h
(11) hSi = 5/2
32 2 c (R2 + h2 )
The radius Rmax that receives the highest intensity of radiation is found
by setting the derivative to zero:
" #
d hSi 0 q2 d 2 4 h 2R 5R3
(12) =
dR 32 2 c (R2 + h2 )
5/2
(R2 + h2 )
7/2

(13) =0

6h
(14) Rmax =
3
If the floor is infinitely large, we can integrate 11 over all values of R to
find the total power received by the floor:

2
0 q2 d 2 4 h
R2
(15) Pf loor = RdR d
32 2 c 0 (R2 + h2 )
5/2
0
RADIATION FROM A CHARGE ON A SPRING 3

R
The integral can be done by parts (integrate 5/2 and differentiate
(R2 +h2 )
R2 ) so we get


0 q2 d 2 4 h 4
(16) Pf loor =
32 2 c 3h
0 q2 d 2 4
(17) =
24c
The total radiated power should be


0 p2
(18) P= S da
=
6c

Averaged over a cycle this comes out to (using 3)

0 p2


(19) hPi =
6c
0 12 q2 d 2 4
(20) =
6c
0 q d 2 4
2
(21) =
12c
Thus half the power gets absorbed by the floor, which is what wed expect
(the other half goes upwards and gets absorbed by the ceiling).
Finally, since the charge is radiating away energy, its total energy is de-
creasing so the amplitude of oscillation will decrease. The amount of power
lost in time dt is P dt, which is equal to the negative change in energy, dE.
We therefore have, defining the amplitude as a function A (t) of time:

1 2 1 2
(22) E = mv + kz
2 2
1
mA2 2 sin2 t + kA2 cos2 t

(23) =
2
1
mA2 2 sin2 t + m 2 A2 cos2 t

(24) =
2
1 2 2
(25) = mA
2
Therefore, the energy lost as the amplitude decreases is
RADIATION FROM A CHARGE ON A SPRING 4

dE
(26) dE = dA
dA
(27) = mA 2 dA
The power for an amplitude A is

0 q2 A2 4
(28) hP (A)i =
12c

so

0 q2 A2 4
(29) dt = mA 2 dA
12c
12cm dA
(30) dt =
0 q2 2 A
To find how long it takes for the amplitude to drop from A = d to A = d/e,
we have
T
12cm d/e dA
(31) dt =
0 0 q2 2 d A
 
12cm d
(32) T = ln d ln
0 q2 2 e
12cm
(33) T = ln e
0 q2 2
12cm
(34) T =
0 q2 2
P INGBACKS
Pingback: Radiation from a radio station
RADIATION FROM A RADIO STATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 22.
A radio stations transmitter consists of a circular current loop acting as
a magnetic dipole. The dipole is on a radio tower 200 m above the ground,
and is emitting a total power of 3.5 104 watts. We can find the position
on the ground receiving the maximum power, although unlike the previous
problem, were concerned only with the actual magnitude of the power and
not with the power per unit area of ground. The formula for intensity is

0 m20 4 sin2
(1) hSi = r
32 2 c3 r2
For a location on the ground that is a distance R from the base of the
tower,

R
(2) sin =
R2 + h2
(3) r2 = R2 + h2

so the intensity is

0 m20 4 R2
(4) hSi = r
32 2 c3 (R2 + h2 )2
The position on the ground receiving maximum intensity is determined
from

d hSi
(5) = 0
dR
4R3 2R
(6) = 3
+ 2
(R2 + h2 ) (R2 + h2 )
(7) Rmax = h

and the intensity at this location is


1
RADIATION FROM A RADIO STATION 2

0 m20 4
(8) hSimax =
128 2 c3 h2
The total power is obtained by integrating over a sphere of radius r:

(9) hPi = S da

0 m20 4 2 sin2 2
(10) = r sin d d
32 2 c3 0 0 r2
0 m20 4
(11) =
12c3
Therefore the maximum intensity can be written in terms of the power:

12 hPi
(12) hSimax =
128 h2
12 3.5 104
(13) =
128 2002
(14) = 2.6 102 watts m2
(15) = 2.6 106 watts cm2
RADIATION FROM THE MAGNETIC DIPOLES OF EARTH
AND PULSARS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 23.
Because Earths magnetic north pole is not at the same place as the orbital
north pole, a component of Earths magnetic dipole rotates as the planet
rotates each day. This gives rise to radiation being emitted. Should we
worry that Earth is losing a lot of energy in this way?
Suppose the angle (latitude difference) between true north and magnetic
north is and the magnetic dipole moment is M. Then the component of
M that rotates has magnitude M sin so we can write this component as

(1) Mrot = M sin [x cos t + y sin t]


The power radiated by a time-varying magnetic dipole is

0 m2
(2) P=
6c3

where m (t) is the magnitude of the magnetic dipole moment. In our case

1/2
(3) m = 2 M sin cos2 t + sin2 t = 2 M sin

so the power radiated is

0 4 M 2 sin2
(4) P=
6c3
The magnetic field due to a dipole moment M is

0
(5) B= [3 (M r) r M]
4r3
Taking Earths field strength at the equator to be around 0.5 Gauss =
5 105 Tesla and taking M r = 0 at the equator and the radius of Earth to
be r = 6.37106 m, we get an estimate of Earths magnetic dipole moment:
1
RADIATION FROM THE MAGNETIC DIPOLES OF EARTH AND PULSARS 2

4r3 B
(6) M = 1.3 1023 Amp m2
0
Using = 11 and = 2/ (60 60 24) = 7.27 105 s1 we get

(7) P = 4 105 watts


We neednt worry about Earth losing a significant amount of energy
through radiation from its magnetic field.
For a pulsar, however, its a different story. A typical pulsar has a ra-
dius of around 10 km = 104 m, a magnetic field strength of 108 Tesla and a
rotation period of 103 s giving = 2 103 s1 . This gives

(8) M = 1027 Amp m2


(9) P = 3.8 1036 sin2 watts
Taking an average value of 21 for sin2 we get a power of around 2
1036 watts. By comparison, the Sun produces only around 41026 wattsso
a pulsar produces the output of 5 billion suns from its magnetic dipole radi-
ation alone.
Not all pulsars generate this much power, though. The period of the crab
nebula pulsar is around 33 ms, so (assuming the radius and magnetic field
are the same as above) this gives a power of around 2 1030 watts which is
still pretty impressive.
FIELDS AND RADIATION FROM A TIME-DEPENDENT SHEET
OF CURRENT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 24.
Suppose we have an infinite plane of current with surface current density
K (t) z; that is, the current density at any instant of time is the same every-
where on the yz plane and flows in the z direction, but K can change with
time.
To find the fields generated by this surface current, we need first to find
the vector potential (were assuming that the plane is electrically neutral so
that the scalar potential V = 0). From symmetry, A can depend only on x,
the perpendicular distance from the yz plane. Suppose were at some dis-
tance x on the x axis (it doesnt matter what y and z are, but for convenience
well take y = z = 0) from the plane. Then the potential is given by

0 K(t r/c) 2
(1) A (x,t) = z d r
4 r

where the integral is taken over the yz plane and for a point on the yz plane
a distance r from the origin, we have
p
(2) r= x2 + r 2
Its easiest to do the integral using polar coordinates, since signals from
all points at a given radius r on the yz plane that arrive at the observation
point at time t left their points of origin at the same retarded time tr = t r/c.
Therefore
 2 2
2 K t x c+r
0
(3) A (x,t) = z r d dr
4 0 0 x2 + r 2

K t x2 +r2

0 c
(4) = z r dr
2 0 x2 + r 2
We can simplify the integral using the substitution (remember that x is a
constant since it is the fixed point of observation):
1
FIELDS AND RADIATION FROM A TIME-DEPENDENT SHEET OF CURRENT 2


x2 + r 2
(5) w =
c
r
(6) dw = dr
c x2 + r 2
x
(7) r=0 w=
c
(8) r= w=
So we get

0 c
(9) A (x,t) = z K (t w) dw
2 x/c
Finally, to get rid of the inconvenient lower bound on the integral, we can
use another substitution:

x
(10) u = w
c
x
(11) w= u=0
c
(12) w= u=

0 c x 
(13) A (x,t) = z K t u du
2 0 c
From here, we can find the fields:

A
(14) E (x,t) =
t

0 c x 
(15) = z K t u du
2 0 c
However, from the chain rule

K t xc u
 x  
(16) K t u =
c u

K t xc u

0 c
(17) E (x,t) = z du
2 0 u
0 c h  x i
(18) = z K () K t
2 c
The original form 1 is actually valid only for finite current distributions,
so we need to assume that at some finite past time t0 , K (t) = 0 for all t < t0 .
This means that the portion of the yz plane that contributes to A is always
FIELDS AND RADIATION FROM A TIME-DEPENDENT SHEET OF CURRENT 3

finite, since if t cr < t0 , K = 0, so r must be restricted to r < c (t t0 ).


Therefore, the field is

0 c  x 
(19) E (x,t) = K t z
2 c
The magnetic field is

(20) B (x,t) = A
Az
(21) = y
x

K t xc u

0 c
(22) = y du
2 0 x
From the chain rule

K t xc u 1 K t xc u
 
(23) =
x c u

so, again invoking the condition that K () = 0:


K t xc u

0
(24) B (x,t) = y du
2 0 u
0 h  x i
(25) = y K () K t
2 c
0  x 
(26) = K t y
2 c
[To be precise, we should note that weve been assuming x > 0, so that
it represents the distance from the sheet of current rather than a strict x
coordinate and thus the time tr = t xc is actually a retarded (earlier) time.
If we look on the other side of the sheet where x < 0 then we need to replace
K t xc u by K t + xc u to get the correct retarded time. This means
that

K t + xc u 1 K t + xc u
 
(27) =
x c u

which leads to

0  x 
(28) B (x,t) = K t + y
2 c
FIELDS AND RADIATION FROM A TIME-DEPENDENT SHEET OF CURRENT 4

for x < 0.]


Note that the electric field does not reduce to the correct value if we take
K to be a constant for all time. In that case, since the sheet of current
is electrically neutral and nothing is changing with time, E = 0. This is
because we cant apply the analysis above to states where the current is not
localized. [Curiously, though, we do get the correct answer from 18 if K
is constant, since then K () K t xc = 0. However, this logic doesnt
work with B, since for a constant surface current B = 02K y and taking K
constant in 25 gives B = 0. The moral is, we just cant apply this analysis
to infinite current distributions.]
The power radiated can be found from the Poynting vector:

1
(29) S = EB
0
0 c 2  x 
(30) = K t x
4 c
This is the energy per unit time per unit area radiated away from the sheet
of current on each side of the sheet, so the total radiated energy is twice this.
Also, the energy that leaves the surface of the plane (rather than the energy
detected at some distance x from the plane) is found when x = 0, so the total
radiated energy is

0 c 2
(31) E= K (t)
2
Example 1. We can find the fields for a couple of special cases. First we
consider
(
0 t 0
(32) K (t) =
K0 t >0
From 19 and 26 we have
(
x
0 t< c
(33) E (x,t) = 0 c x
2 K0 z t > c

0
t < xc
(34) B (x,t) = 20 K0 y t > xc , x > 0
0 K y t > x , x < 0

2 0 c
FIELDS AND RADIATION FROM A TIME-DEPENDENT SHEET OF CURRENT 5

It may seem odd that E gets switched on and remains constant for t > xc ,
since once the current is on and constant, there is no electric field generated.
However, the current gets switched on at one particular time over the entire
yz plane, so due to retardation, the observer at point x receives news of this
switching on from ever increasing circles on the yz plane as time progresses,
so after the first news arrives (from the point directly below the observer, at
a distance of x) there is a continual stream of news coming in that the current
has switched on from circles that are ever further away, which results in a
steady electric field.
It is also interesting that the magnetic field takes on the same value as that
produced by an infinite sheet of constant current (see Griffiths, Example
5.8). This is because the magnetic field is determined solely by the point on
the sheet directly below the observer. Contributions to B from other points
on the sheet cancel each other out.
Example 2. This time, the current is linearly increasing
(
0 t 0
(35) K (t) =
t t >0
From 19 and 26 we have

x
(
0 t< c
(36) E (x,t) = 0 c

|x|

x
2 t c z t > c

0
t < xc
x
B (x,t) = 02 t c y t > xc , x > 0

(37)
0 t + x y t > x , x < 0

2 c c
RADIATION FROM A POINT CHARGE NEAR A CONDUCTING
PLANE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 25.
For another example of radiation, we can return to the example of the
point charge q at a distance z from an infinite conducting plane, which we
looked at in electrostatics as an example of the method of images. We saw
then that we could replace the conducting plane by an equal and opposite
charge q at position z. This makes the xy plane (the plane of the con-
ductor) an equipotential surface with V = 0, so it satisfies the boundary
conditions for the half-space z 0.
Because the potentials of the original and image systems are the same in
the physical region (the region outside the conductor), so too are the fields,
and since the force on the charge is determined by the fields, the forces in
the two systems are also equal, so we have

1 q2
(1) F= z
40 (2z)2

where the negative sign indicates that the force is towards the plane (or
towards the image charge, in the image system).
If the charge is not restrained, this force will cause it to accelerate, so

F 1 q2
(2) a= = z
m 40 (2z)2 m
This acceleration will cause the charge to radiate, and we can calculate
the dipole radiation from the formula

0 p2
(3) P
=
6c

where p is the dipole moment of the charge distribution. The dipole moment
of the charge and its image is
1
RADIATION FROM A POINT CHARGE NEAR A CONDUCTING PLANE 2

(4) p = 2qzz

since the two charges are separated by a distance 2z. Therefore

1 q3
(5) p = 2qzz = 2qa = z
20 (2z)2 m

so the power is

!2
0 1 q3
(6) P
=
6c 20 (2z)2 m
!2
0 c2 0 q3
(7) =
6c 2 (2z)2 m
3
0 cq2

1
(8) =
4 6m2 z4
Although this is the answer given in Griffithss book (in his problem
11.25), Im not convinced its actually correct. The actual charge distribu-
tion that we calculated when studying the method of images, that is, using
the point charge q at position z and the surface charge density, given by

qz
(9) = 3/2
2 (r2 + z2 )

where r is the radial distance on the xy plane measured from the point di-
rectly beneath the point charge. If we take the origin to be the point in the
xy plane directly beneath the point charge, the dipole moment of the surface
charge comes out to zero (by symmetry; for each charge element at a point
r there is an equal charge element at point r so their dipole moments can-
cel out), while the dipole moment of the point charge is qzz, so the total
dipole moment of the point charge + surface charge is also qzz. Since the
dipole moment is lower by a factor of 2, the total power radiated is lower
by a factor of 4, and would seem to be

3
0 cq2

1
(10) P=
4 24m2 z4
RADIATION FROM A POINT CHARGE NEAR A CONDUCTING PLANE 3

If we assume that the half space z 0 is entirely occupied by the con-


ductor, there can be no free charge anywhere except on the surface of the
conductor (and of course the point charge itself). I suspect the problem may
have something to do with the fact that the method of images really applies
only in electrostatics, though Im not entirely sure. Comments welcome.
FIELDS OF A MOVING MAGNETIC MONOPOLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 26.
We can use the duality transformations to find the analog of the Larmor
formula for a point magnetic monopole. The transformations are

(1) E0 = E cos + cB sin


(2) cB0 = cB cos E sin
(3) cq0e = cqe cos + qm sin
(4) q0m = qm cos cqe sin
The fields for a moving point electric charge are

qe r
c2 v2 u + r (ua)
  
(5) E (r,t) =
40 (r u)3
1
(6) B (r,t) = r E (r,t)
c

Applying the transformations with = 2 we get

qm r
c2 v2 u + r (ua)
  
(7) cBm = 3
4c0 (r u)
qm r
c2 v2 u + r (ua)
  
(8) Bm = 3
4c2 0 (r u)
1 1
(9) Em = r cBm (r,t)
c c
(10) Em = cr Bm (r,t)
The Larmor formula for electric charge is

0 q2e a2
(11) Pe =
6c

so for magnetic charge it becomes


1
FIELDS OF A MOVING MAGNETIC MONOPOLE 2

0 q2m a2
(12) Pm =
6c3
RADIATION REACTION: ENERGY CONSERVATION WITH A
CONSTANT EXTERNAL FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 27.
Here, well return to the example of a constant force acting for a fixed
time on a charge, and explore the energy distribution taking into account
radiation reaction.
The external force F is switched on at t = 0, remains constant until t = T
and is then switched off. The resulting acceleration and velocity are, if we
take a = 0 for t > T and v = 0 at t = :

 
F T / et/
m 1 e t <0



(1) a (t) = F 1 e(tT )/ 0<t <T
m

0 t >T

 
F
1 eT / et/ t <0
m


(2) v (t) = F t e(tT )/ + F 0<t <T
m m

FT
m t >T

First, well find how much work the external force does. Since the force
acts only in the interval 0 < t < T , we have

T
(3) W = Fv dt
0
T    F 
F (tT )/
(4) = F t e + dt
0 m m
F2 T 2
   FT
2 T /
(5) = 1e +
m 2 m
The charge radiates away energy whenever a 6= 0, so it radiates for all
t < T . The Larmor formula gives the rate at which energy is radiated:
1
RADIATION REACTION: ENERGY CONSERVATION WITH A CONSTANT EXTERNAL FORCE
2

0 q2 a2
(6) P= = ma2
6c
To get the total energy radiated R, we integrate P over time:

(7)
0 T 2
F2  T /
2
2t/ F (tT )/
2
R = m 2
1 e e dt + m 2
1 e dt
m 0 m
(8)
F 2 
 2     
T / T / 2T /
= 1e + T 2 1 e + 1e
m 2 2
(9)
F 2 h  i
= + T 2 + + eT / ( + 2) + e2T /
m 2 2 2 2
(10)
F 2   
= T 1 eT /
m
Since the final velocity is FT /m, the final kinetic energy is

F 2T 2
(11) K=
2m
By comparing terms, we see that

(12) W = R+K

so that the work done by the force is divided between the energy radiated
away and the final kinetic energy, so that energy is conserved.
Given that this problem suffers from preacceleration (violating causality)
its something of a miracle that energy is actually conserved. The charge
seems to know just how much to accelerate in anticipation of the force
to be applied so that it can radiate away just the right amount of energy to
balance out the work done and the final kinetic energy. There just has to be
something wrong with the theory!
P INGBACKS
Pingback: Radiation reaction with a delta-function external force
RADIATION REACTION WITH A DELTA-FUNCTION
EXTERNAL FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 28.
Heres another example of applying an external force to a charge feeling
the radiation reaction force. In general, a charges acceleration obeys the
differential equation

F
(1) a = a +
m

where F is the external force and

0 q2
(2)
6mc

Suppose now that the force is a delta function:

(3) F = k (t)

for some constant k. In the earlier post, we showed that if F is finite ev-
erywhere, then a must be continuous everywhere. However, here F is not
finite at t = 0. As before we start by integrating 1 over a small time interval
around t = 0:


1
(4) a dt = [a () a ()] + F dt
m

Provided that a is finite everywhere, the integral on the LHS goes to zero
as 0 so were left with
1
RADIATION REACTION WITH A DELTA-FUNCTION EXTERNAL FORCE 2


1
(5) a = F dt
m

k
(6) = (t) dt
m
k
(7) =
m
k
(8) a =
m
We can repeat the calculations we did earlier to check that energy is con-
served here. Since F = 0 everywhere except t = 0, the general solution of
1 is

(
a0 et/ t <0
(9) a (t) =
a1 et/ t >0

If we eliminate the runaway acceleration for t > 0 by requiring a1 = 0


then the condition 8 requires a0 = k/m, so

(
k t/
m e t <0
(10) a (t) =
0 t >0

By requiring v = 0 at t = and that v is continuous at t = 0 we get

(
k t/
me t <0
(11) v (t) = k
m t >0

The work done by the force is



(12) W = Fv dt


(13) = k (t) v dt

(14) = kv (0)
k2
(15) =
m
The energy radiated R is given by integrating the Larmor formula
RADIATION REACTION WITH A DELTA-FUNCTION EXTERNAL FORCE 3

0 q2 a2
(16) P= = ma2
6c

so we get
0  2
k
(17) R = m e2t/ dt
m
k2
(18) =
2m
The final kinetic energy is

1 k2 k2
(19) K= m 2 =
2 m 2m

Thus

(20) W = R+K

and energy is conserved.


P INGBACKS
Pingback: Tunnelling through a potential barrier with the radiation reac-
tion force
TUNNELLING THROUGH A POTENTIAL BARRIER WITH THE
RADIATION REACTION FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 29.
Besides giving rise to runaway acceleration and violation of causality,
the radiation reaction force also predicts a classical version of tunnelling
through a potential barrier, something you might think is confined to quan-
tum mechanics. Suppose a charged particle travels in along the x axis, start-
ing at x = with some initial velocity vi . Between x = 0 and x = L there
is a finite potential barrier of height U0 .
In general, the charges acceleration obeys the differential equation

F
(1) a = a +
m

where F is the external force and

0 q2
(2)
6mc
Because the derivative of a step function is delta function and the force is
the negative gradient of the potential, the force is

(3) F = U0 ( (x) + (x L))

Weve treated radiation reaction with a delta function force before, but in
that case, the force was a delta function in time rather than space. However,
we can envision the particle travelling in until it arrives at x = 0 at which
point it feels a delta function force, then travelling along to x = L where it
feels another delta function force. Previously, to solve 1 around the time
t = 0 we integrated the equation over a small interval about t = 0:


1
(4) a dt = [a () a ()] + F dt
m
1
TUNNELLING THROUGH A POTENTIAL BARRIER WITH THE RADIATION REACTION FORCE
2

If F were a delta function in time, the integral is easily done, as (t) dt =
1. In our case, with the delta function a function of position, we can con-
sider the position as a function of time and use the chain rule:

(5) (x (t)) dt = (u) dt

where u x (t) so that du = x dt = v dt and dt = du/v. Therefore



(u)
(6) (x (t)) dt = du
v
1
(7) =
v (u = 0)

The velocity in the last line is evaluated at u = 0, which corresponds to


x = 0. If we define our origin of time at this point, then this is also the
velocity at t = 0, so weve converted the problem into the one weve already
solved.
In our earlier solution, we saw that the acceleration for a force of k (t)
has a discontinuity at t = 0, with a = k/m. Here, k = U0 /v0 , where
v0 is the velocity at t = 0, so

U0
(8) a0 = +
mv0
Similarly, at x = L we can say that the particle reaches this point at t = T
when the force is equal in magnitude but opposite in direction, so

U0
(9) aT =
mvT

where vT is the velocity at t = T . The general solution for the acceleration


is therefore

t/ t <0
a0 e

(10) a (t) = a1 et/ 0<t <T
a et/

t >T
2
If we set a2 = 0 to prevent runaway acceleration for t > T , then we can
apply 8 and 9 to determine a1 and a0 . We get at t = T

U0 T /
(11) a1 = e
mvT
TUNNELLING THROUGH A POTENTIAL BARRIER WITH THE RADIATION REACTION FORCE
3

Because a2 = 0, there is no further acceleration after t = T , so the velocity


at this time remains constant for all future times, and we can write it as v f ,
the final velocity. Therefore

U0 T /
(12) a1 = e
mv f
At t = 0 we get

U0
(13) a1 a0 =
mv0
!
U0 eT / 1
(14) a0 =
m vf v0
In summary,
 T / 
U0 e 1 t/
v f v0 e t <0

m

(15) U0 T / t/
a (t) = mv e e 0<t <T
f
0 t >T
Integrating this to get the velocity, we have
 T / 
U0 e 1 t/ + v

m
vf v0 e i t <0
U0
(16) v (t) = mv f eT / et/ + v1 0<t <T


v
f t >T
Integrating again, we get the position
 
U0 eT / 1 t/ + v t + x

m
vf v0 e i 0 t <0
(17) U
x (t) = mv0 eT / et/ + v1t + x1

f
0<t <T


v (t T ) + L t >T
f

where in the last line, weve just imposed the condition that the particle
moves at a constant velocity v f starting from x = L at t = T .
We now apply boundary conditions to find the various constants. First,
to get x0 we require x (0) = 0, which gives

U0  
(18) x0 = v f v0 eT /
mv0 v f
TUNNELLING THROUGH A POTENTIAL BARRIER WITH THE RADIATION REACTION FORCE
4

Also using t = 0, we can find x1 from the middle expression for x (t):

U0 T /
(19) x1 = e
mv f

Also from the middle expression for x (t) we can find v1 by requiring
x = L at t = T :

 
1 U0  T /
(20) v1 = L 1e
T mv f

We can now plug this into the middle expression for v (t), set t = 0 and
require the velocity to be v0 to find v0 :

 
1 U0  T /
(21) v0 = L+ + ( + T ) e
T mv f

Substituting this back into the middle expression for v (t) at t = T , when
the velocity is v f we find

U0  
T /

(22) L = vf T T 1e
mv f

Finally, we can find vi by taking the first expression for v (t), setting t = 0
and requiring the result to be equal to v0 . This gives a rather unpleasant
expression which can be simplified by collecting terms
TUNNELLING THROUGH A POTENTIAL BARRIER WITH THE RADIATION REACTION FORCE
5

(23)
   
m2 v4f U0 m 1 eT / v2f +U02 1 eT /
vi =  
mv f mv2f U0 1 eT /
(24)
 
U0 U0 eT / 1
= vf 
mv f mv2f +U0 eT / 1
(25)
     
mv 2 +U eT / 1 mv2 U eT / 1 +U eT / 1
U0 f 0 f 0 0
= vf
mv f +U0 eT / 1
2

mv f
(26)
   
mv2f +U0 eT / 1 U0 e T / 1
U0
= vf 1
mv2f +U0 eT / 1

mv f
(27)
!
U0 mv2f
= vf 1
mv2f +U0 eT / 1

mv f
(28)

U0 1
= vf 1 U0 
mv f 1+ mv2f
eT / 1

In the case where the final kinetic energy is half the barrier height, we
have mv2f = U0 and we get

vf
(29) vi =
eT /
This can be expressed in terms of the barrier length from 22:
 
(30) L = v f 1 eT /
vf
(31) vi =
1 L/v f
Wed like to find a set of values such that both vi and v f are positive, and
also the initial kinetic energy is less than U0 . Choosing L = v f /4 gives
TUNNELLING THROUGH A POTENTIAL BARRIER WITH THE RADIATION REACTION FORCE
6

4
(32) vi = vf
3
1 2 8 2
(33) mvi = mv
2 9 f
8
(34) = U0
9
Therefore the particle will actually tunnel through the barrier. However,
given the crazy predictions arising from the reaction force, Im not sure how
much I would believe this result.
PHYSICAL BASIS OF THE RADIATION REACTION FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 30a.
Weve seen that the radiation reaction force is given by the Abraham-
Lorentz formula

0 q2
(1) Frad = a
6c
Griffiths gives a derivation of this formula in his section 11.2.3 by con-
sidering a charge q split into a dumbbell of length d with its axis in the
y direction moving along the x axis. A half-charge q/2 is at each end of
the dumbbell, and the idea is that as the dumbbell moves along the x axis,
each charge feels a force due to the fields emitted at the retarded time tr
by both the charge itself and the charge at the other end of the dumbbell.
Griffithss derivation, however, is for the special case where the dumbbell
is momentarily at rest at the retarded time, which simplifies the calcula-
tions significantly. Here, well run through the derivation when the retarded
velocity is not zero.
Due to the immense amount of algebra involved in this derivation, it
makes sense to use Maple to do the calculations. However, even using
Maple, the process is far from simple, so this post will be as much a tutorial
on how to solve problems like this with Maple as it is on physics.
We start with the electric field due to a moving charge

qr
c2 v2 u + r (ua)
  
(2) E (r,t) =
40 (r u)3
where

(3) u = cr v

and r is the vector from the source charge at the retarded time to the desti-
nation charge at the current time.
If the dumbbell has moved a distance l along the x axis since the retarded
time tr , then if were considering the field felt by a charge q/2 at one end of
1
PHYSICAL BASIS OF THE RADIATION REACTION FORCE 2

the dumbbell due to the charge at the other end emitted at the retarded time,
we have
p
(4) r = l2 + d2
(5) r = l x + d y
Since all motion is along the x axis, we have

(6) r u = cr lv
(7) r a = la
cl
(8) ux = v
r
(9) r (ua) = u (r a) a (r u)
(10) = lau (cr lv) a
Well refer to the top charge as charge 1 and the bottom charge as charge
2. From symmetry, the y component of u for the field due to charge 1 felt
by charge 2 is equal and opposite to the same component for the field due
to charge 2 felt by charge 1, so when added together, they cancel, and we
need consider only the x component in what follows.
We can also ignore the magnetic forces since the magnetic field of a mov-
ing point charge is

1
(11) B (r,t) = r E (r,t)
c
Both E and r lie in the xy plane, so

(12) B = (rx Ey ry Ex ) z
For charge 2, both Ey and ry are the negative of their counterparts for
charge 1, while Ex and rx are the same for both charges, so the magnetic
field cancels out when added for the two charges.
Plugging in the above results and taking the x component we get for
charge 1 (which is q/2) [Ill use r instead of r in what follows since its
easier to type.]
  2   
q r cl c lv
(13) Ex = v + la ac r
80 c3 r lv 3 r 2 c
c
p
where 1/ 1 v2 /c2 .
PHYSICAL BASIS OF THE RADIATION REACTION FORCE 3

What we want is an expansion of Ex as a power series in d, the distance


between the charges (length of the dumbbell). In the process, we want to
eliminate l and r from the expression. To do this, we first write things in
terms of the time elapsed since the retarded time:

(14) T t tr
First, we can expand the position

1 1
(15) x (t) = x (tr ) + v (tr ) T + a (tr ) T 2 + a (tr ) T 3 + . . .
2! 3!
The distance travelled is l = x (t) x (tr ) so (dropping the dependence on
tr ; until further notice, v, a and a are all assumed to be at the retarded time):

a a
(16) l = vT + T 2 + T 3
2 6
How do we know how many terms to keep in this expansion? Ultimately,
were interested in taking the limit as d 0, so were looking for the con-
stant term in the series expansion of 13. We know the final formula involves
terms up to a so as a rule of thumb, we keep terms up to that point.
Things seem to be getting worse, in that weve introduced another pa-
rameter T , which we now need to get rid of. We note that T is the time
taken for a signal to travel from charge 1 at the retarded time to charge 2 at
the present time, so

(17) c2 T 2 = r2 = l 2 + d 2
a 2 a 3 2
 
(18) = vT + T + T + d2
2 6
va a2
 
2 2 2 3
(19) = d + v T + avT + + T4
3 4
c2 2 va a2
 
2 3
(20) 2
T = d + avT + + T4
3 4
Again, weve kept powers of T only up until the first term containing a.
Now we express T as a series in powers of d:

(21) T = A0 + A1 d + A2 d 2 + A3 d 3
By trial and error, we find that we need up to the cubic term to include
a term with a. We can now substitute this into 20 and equate powers of d
PHYSICAL BASIS OF THE RADIATION REACTION FORCE 4

on both sides. This is where its useful to bring in Maple. We can use the
command:
[code]
dist := c^2*T^2/gamma^2 = d^2+v*T^3*a+(1/3)*v*T^4*A+(1/4)*a^2*T^4;
Teq := subs(T = A__3*d^3+A__2*d^2+A__1*d+A__0, dist);
simplify(zip(`=`,
[coeffs(select(t -> degree(t, d) <= 4, lhs(expand(Teq))), d)],
[coeffs(select(t -> degree(t, d) <= 4, rhs(expand(Teq))), d)]));
As := solve(Tcoeffs, {A__0, A__1, A__2, A__3});
assign(As[1]);
[/code]
This Maple code defines an equation called dist that is equivalent o 20,
then substitutes 21 into it. The next command selects all the coefficients
from terms with degree of d less than or equal to 4 from the LHS and RHS
of the expanded equation and applies the = operator between each pair. We
then call solve to solve for the values of the Ai . The final assign assigns
the Ai s to the values found by solve. (The [1] means to take the first set of
solutions; there are actually two sets of solutions but the second set contains
negative values so we discard it.) The results are

(22) A0 = 0

(23) A1 =
c
av 4
(24) A2 =
c4
5
15a2 2 v2 + 4avc2 + 3a2 c2

(25) A3 = 7
24c
5
4avc2 + 3a2 2 c2 + 4v2

(26) = 7
24c
We can now plug these values back into 21 and then into 16 to get an
expansion for l. Keeping only terms up to the first occurrence of a we use
the Maple code
[code]
l := select(t -> degree(t, d) < 4, collect(expand(l), d));
l__2 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), coeff(l, d, 2)));
l__3 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), coeff(l, d, 3)));
l__3 := simplify(subs(c^2-v^2 = c^2/gamma^2, l__3));
[/code]
We also do a few tweaks and substitutions by hand to convert a few c2 v2
terms into terms until we get the result
PHYSICAL BASIS OF THE RADIATION REACTION FORCE 5

v 4a 5
d + 2 d2 + 15a2 2 v + 4ac2 d 3

(27) l= 5
c 2c 24c
Since weve now got l as a series in d, we can use it to find expansions of
the other factors in 13.
For r we get in Maple
[code]
r__s := series(r, d, 4);
r__1 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), coeff(r__s, d, 1)));
r__2 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), coeff(r__s, d, 2)));
r__2 := simplify(subs(c^2-v^2 = c^2/gamma^2, r__2));
r__3 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), coeff(r__s, d, 3)));
r__3 := simplify(subs(c^2-v^2 = c^2/gamma^2, r__3));
[/code]
With a few more simplify commands, we get

4 va 2 5 3 a2 c2 2 + 12 a2 2 v2 + 4 ac2 v 3

(28) r = d + 3 d + d
2c 24c6
The other terms in 13 can be worked out similarly by using Maples se-
ries command together with a few simplifys:

cl a 2 2 2 2
 2
(29) v = d + 3a v + 2 ac d
r 2c 12c4
15 a3 2 v + 4 aac2 5 d 3

c2 c2 va d 4 a2 d 2
(30) + la = 2 + + +
2 c 2c2 24c5
lv c2 v2 7 a2 c2 v2 3
 
(31) r = d + d
c c2 8 c6
lv 3 3 9 a2 3 15 a4 d
 
(32) r = 3 3/8 4 +
c d c d 32 c8
Now we can put everything together to find Ex :
[code]
clrv := simplify(series(c*l/r-v, d, 5));
clrv__1 := coeff(clrv, d, 1);
clrv__1 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), clrv__1));
clrv__2 := simplify(subs(gamma = 1/sqrt(1-v^2/c^2), coeff(clrv, d, 2)));
clrv__2 := simplify((2*A*c^2/gamma^2+3*a^2*v)/(12*c^4/gamma^4));
clrv := clrv__2*d^2+clrv__1*d;
cgla := c^2/gamma^2+l*a;
rlvc := simplify(series(r-l*v/c, d, 4));
PHYSICAL BASIS OF THE RADIATION REACTION FORCE 6

rlvc := simplify(subs(c^2-v^2 = c^2/gamma^2, rlvc));


rlvc := convert(rlvc, polynom);
n__1 := collect(expand(clrv*cgla), d);
n__2 := n__1-a*c*rlvc+n__1;
n__3 := select(t -> degree(t, d) < 4, collect(n__2, d));
d__1 := series(1/rlvc^3, d, 4);
d__1 := convert(d__1, polynom);
E := q*select(t -> degree(t, d) < 1, collect(r*n__3*d__1, d))/(8*Pi*epsil
[/code]
In the last line, we select only those terms of degree less than 1 in d, since
were interested only in terms that dont vanish as d 0. After a final
simplify we get

3 aq 1 q 4 3 a2 2 v + Ac2

(33) Ex = +
160 c2 d 48c5 0
The net force is 2 (q/2) Ex since there is a force from both q/2 charges
at the retarded time. Thus

3 aq2 1 q2 4 3 a2 2 v + Ac2

(34) Fx = +
160 c2 d 48c5 0
Somewhat embarrassingly, the first term blows up as d 0, but it is
moved over to the LHS and incorporated as part of the particles mass in a
process called renormalization (basically physics-speak for fudge). This
leaves the last term as the actual reaction force from one charge on the other.
To convert everything to the current time, we can use some more expan-
sions (where a subscript t indicates current time):

(35) v = vt + vt (tr t) = vt at T
(36) a = at at T
We take only up to first order terms in T , since the worst power of d in
Fx is d 1 , so a first order correction multiplied into a d 1 term will give an
adjustment to the constant term, which is the term of interest. We need to
transform not only the bare a and v terms in the force, but also the factor,
since it depends on v. In Maple, we get (I use A to represent a in the Maple
code):
[code]
E__0 := simplify(coeff(E, d, 0));
E__0t := subs({a = -A__t*T+a__t, v = -T*a__t+v__t}, subs(gamma = 1/sqrt(1
E__0ts := series(E__0t, d, 1);
PHYSICAL BASIS OF THE RADIATION REACTION FORCE 7

simplify(E__0ts);
E__m1 := simplify(coeff(E, d, -1));
E__m1t := subs({a = -A__t*T+a__t, v = -T*a__t+v__t}, subs(gamma = 1/sqrt(
E__m1ts := series(E__m1t, d, 2);
E__m1ts := simplify(subs(c^2-v__t^2 = c^2/gamma^2, E__m1ts));
E__t := simplify(coeff(E__0ts, d, 0)+coeff(E__m1ts, d, 1));
E__t := simplify(subs(c^2-v__t^2 = c^2/gamma^2, E__t));
[/code]
We isolate the constant (d 0 ) coefficient from 33, substitute for and trans-
form a and v and then do the same for the d 1 coefficient. The final result
for the field is the sum of the order d term from the coefficient of d 1 (since
the ds cancel out to give a constant term) plus the order d 0 term from the
coefficient of d 0 . The final result is

q 4 c2 at + 12 at 2 2 vt + 3 at c2

(37) Et =
48c5 0
q2 4 c2 at + 12 at 2 2 vt + 3 at c2

(38) Ft =
48c5 0
q2 4 3at2 2 vt
 
(39) = at +
120 c3 c2
0 q2 4 3at2 2 vt
 
(40) = at +
12c c2
When the force of each end on itself is included, this doubles the answer,
so we get

0 q2 4 3at2 2 vt
 
(41) F= at +
6c c2
P INGBACKS
Pingback: Power radiated by radiation reaction force
Pingback: Radiation from a charge in hyperbolic motion
Pingback: Self-force on a dipole in hyperbolic motion
POWER RADIATED BY RADIATION REACTION FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 30b.
The radiation reaction force is given, for a particle momentarily at rest,
by the Abraham-Lorentz formula

0 q2
(1) Frad = a
6c
The more general formula for a charge moving with an arbitrary velocity
is

0 q2 4 3a2 2 v
 
(2) Frad = a +
6c c2
The power radiated by a moving charge is given by

" #
0 q2 6 2 |v a|2
(3) P= a
6c c2

which reduces to the following formula if v and a are parallel:

0 q2 6 a2
(4) P=
6c
The Abraham-Lorentz formula was derived by assuming that the charge
is in the same state (effectively, that its velocity and acceleration are the
same) at two points in time, and then calculating the average power emitted
over that interval. That is, we assume

t2 t2
(5) Frad vdt = Pdt
t1 t1

We can check that the general formula 2 is consistent with 4 under this
assumption. Since the motion is in one dimension, we want to show that
1
POWER RADIATED BY RADIATION REACTION FORCE 2

t2 t2
0 q2 4 3a2 2 v 0 q2 6 a2
 
(6) a + v dt = dt
t1 6c c2 t1 6c
t2  t2
3a2 6 v2

4
(7) av + 2
dt = 6 a2 dt
t1 c t1
Well need the following formula to do the integrals:

d 1
(8) = p
dt 1 v2 /c2
vv
(9) = 3/2
c2 (1 v2 /c2 )
va
(10) = 3 2
c
Looking at the first term of the integral on the LHS, we can integrate by
parts.
t2 t2
 
4
t 4 4 d 4
(11) av dt = av t2 a a + v dt
t1
1
t1 dt
t2 
4av2 6

(12) = a a 4 + dt
t1 c2
t2
4v2
 
2 6 2
(13) = a + 2 dt
t1 c
t2
v2 4v2
 
2 6
(14) = a 1 2 + 2 dt
t1 c c
t2
3v2
 
(15) = a2 6 1 + 2 dt
t1 c
The integrated term in line 1 is zero because v and a are the same at both
limits by assumption.
Adding this to the second term on the LHS of 7 we get
t2  t2
3a2 6 v2 3v2 3v2
  
4 2 6
(16) av + dt = a 1 + 2 + 2 dt
t1 c2 t1 c c
t2
(17) = a2 6 dt
t1

so equation 7 is confirmed.
RADIATION FROM A CHARGE IN HYPERBOLIC MOTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 11, Post 31.
A particle moving along the x axis in hyperbolic motion has a position
given by
p
(1) x (t) = b2 + c2t 2

where b is a constant giving the closest approach to the origin. If the particle
has charge q then its radiation is given by

0 q2 6 a2
(2) P=
6c
To calculate this we need

tc2 tc2
(3) v (t) = =
b2 + c2t 2 x
c2 4
ct 2 c b2
2
(4) a (t) = 3 = 3
x x x
1
(5) = p
1 v2 /c2
1
(6) = q
2 2
1 t xc2
x
(7) =
b
The power is therefore

2
0 q2  x 6 c2 b2

(8) P =
6c b x3
0 q2 c3
(9) =
6b2
1
RADIATION FROM A CHARGE IN HYPERBOLIC MOTION 2

Thus the radiated power is constant as the charge moves along its trajec-
tory.
The radiation reaction force is given by

0 q2 4 3a2 2 v
 
(10) Frad = a +
6c c2

for which we need

3c4 b2t
(11) a =
x5
We get

2    2 
3a2 2 v c2 b2

3 x 2 tc
(12) 2
= 2
c c x3 b x
3c4 b2t
(13) =
x5
(14) = a
(15) Frad = 0
This is a rather curious case in that although the charge radiates, there is
no radiation reaction force.
INERTIAL FRAMES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 1-2.
Although weve already looked at special relativity in earlier posts, I
think its a good idea to revisit the theory by following the treatment in
Griffithss book in his Chapter 12. The reason is that Griffithss book devel-
ops the theory specifically with a view to showing how it applies to classical
electromagnetism, which is something we havent really looked at before.
Central to both classical and relativistic mechanics is the notion of an
inertial frame of reference. Griffithss definition of an inertial frame is one
of the simplest that Ive seen. Basically, an inertial frame is a frame in which
Newtons first law holds. Newtons first law states that, in the absence of an
external force, an object travels in a straight line at constant speed (or just
at constant velocity, since velocity incorporates both speed and direction of
travel).
Example 1. If S is an inertial frame, then if another frame S moves at
constant velocity v relative to S , it too is an inertial frame. To prove this,
we need to specify whether we are using Galilean or relativistic mechan-
ics. In Galilean mechanics, the velocity addition rule is that if an objects
velocity as measured in S is v1 then its velocity as measured in S is

(1) v1 = v1 v
If both v and v1 are constant, then so too is v1 , so S must be an inertial
frame.
In relativistic mechanics, the velocity addition formula is (if v and v1 are
parallel):

v1 v
(2) v1 =
1 v1 v/c2

and again, if v1 and v are constants, so too is v1 so S is an inertial frame.


[If v and v1 are not parallel, this formula applies to the components of the
velocities that are parallel. The perpendicular components are the same in
both frames, so they too must be constant.]
1
INERTIAL FRAMES 2

Conversely, if S and S are both inertial systems, then the velocities


of an object with no external forces acting on it must be constant in both
frames, which implies that the relative velocity v must be a constant, as can
be seen from either of the formulas above.

Example 2. Considering classical (Galilean) mechanics only, suppose that


in an inertial frame S we have a collision between a particle A (mass mA
and velocity uA ) and another particle B (mass mB and velocity uB ). During
the collision, some of the mass of A gets transferred to B, so that afterwards
we have particles C and D with masses mC , mD and velocities uC , uD . If
momentum is conserved, then

(3) mA uA + mB uB = mC uC + mD uD
(4) mA + mB = mC + mD

In inertial frame S which moves with velocity v relative to S , the ve-


locities of the particles transform as

(5) ui = ui v

The conservation of momentum equation is then

(6) mA (uA + v) + mB (uB + v) = mC (uC + v) + mD (uD + v)


(7) mA uA + mB uB + (mA + mB ) v = mC uC + mD uD + (mC + mD ) v
(8) mA uA + mB uB = mC uC + mD uD

where to get the last line, we assumed that mass is conserved as in 4.


An elastic collision is one in which the total kinetic energy is conserved.
[In an inelastic collision, momentum is conserved but kinetic energy is
transformed into another form of energy, such as heat. If two objects of
equal mass are travelling towards each other with the same speed and upon
collision, they stick together and come to rest, then the total momentum is
zero before and after the collision, but the kinetic energy has been reduced
to zero.] If the collision is elastic in S , then
INERTIAL FRAMES 3

(9)
1 1 1 1
mA u2A + mB u2B = mC uC2 + mD u2D
2 2 2 2
(10)
mA u2A + v2 + 2uA v + mB u2B + v2 + 2uB v = mC uC2 + v2 + 2uC v + mD u2D + v2 + 2uD v
  

(11)
mA u2A + mB u2B + (mA + mB ) v2 + 2 (mA uA + mB uB ) v = mC uC2 + mD u2D + (mC + mD ) v2 + 2 (mC uC + mD
We can cancel the middle terms on both sides using 4 and the last terms
on both sides using 8, so were left with

(12) mA u2A + mB u2B = mC uC2 + mD u2D

so the collision is elastic in S as well.


P INGBACKS
Pingback: Relativistic momentum and energy
VELOCITY ADDITION IN SPECIAL RELATIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 12.3.
The velocity addition formula in special relativity is (if va and vb are
parallel):

va + vb
(1) vr =
1 + va vb /c2
Example 1. To get a feel for how small the correction to the classical for-
mula vc = va + vb is for everyday speeds, suppose va = 5 miles/hour =
2.235 m s1 and vb = 60 miles/hour = 26.82 m s1 . Since va vb /c2 in this
case is very small, we can approximate vr by

 va vb 
(2) vr (va + vb ) 1 2
c
The percentage error in the classical formula is then

vc vr
(3) v = 100%
vc
(va + vb ) va vb /c2

(4) 100%
va + vb
va vb
(5) = 100%
c2
2.235 26.82
(6) = 2
100%
(3 108 )
(7) = 6.66 1014 %

Its not surprising that no relativistic effects are seen in the everyday
world.

Example 2. Suppose you could run at va = 0.5c (relative to the train) down
the corridor of a train travelling at vb = 0.75c. An observer on the ground
would see your speed relative to the ground as
1
VELOCITY ADDITION IN SPECIAL RELATIVITY 2

0.5 + 0.75
(8) v = c
1 + (0.5) (0.75)
(9) = 0.91c

Even though the classical sum of velocities is greater than c, the relativis-
tic formula still gives a result that is less than c.

Example 3. The formula 1 always gives a result that is less than c. To prove
this, we can simplify the notation by using velocities that are fractions of c
so that a va /c and b vb /c with the sum given by s vr /c. Then

a+b
(10) s=
1 + ab

To check if there are any extrema in the region 0 a 1, 0 b 1 we


can take the two partial derivatives and set them to zero:

s 1 (a + b) b
(11) = =0
a 1 + ab (1 + ab)2
s 1 (a + b) a
(12) = =0
b 1 + ab (1 + ab)2

The only solution within the region is a = b = 1. We can see that this
must be a maximum within the region, since along the border a = 1 or the
border b = 1 we have s = 1, along the border a = 0 we have s = b and along
the border b = 0 we have s = a so that s 1 on all borders. Thus s 1
everywhere in the region 0 a 1, 0 b 1. The surface 10 within this
region looks like this:
VELOCITY ADDITION IN SPECIAL RELATIVITY 3

The point a = b = 1 is, however, actually a saddle point, as an expanded


plot of the region 0 a 2, 0 b 2 reveals:

P INGBACKS
Pingback: Velocity addition formulas for all 3 directions
Pingback: Spectroscopic, eclipsing binary stars: mass, radius and tem-
perature
VELOCITY ADDITION: CHASING SPACE PIRATES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 4.
The velocity addition formula in special relativity is (if va and vb are
parallel):

va + vb
(1) vr =
1 + va vb /c2
Example. Suppose some space pirates are fleeing from the solar system
police in a spacecraft that is moving at 43 c (relative to the Earth). The po-
lices spaceship is travelling at only 12 c but in an attempt to stop the pirates
they fire a torpedo at them. The torpedos velocity, relative to the polices
ship, is 13 c. Using the classical velocity addition formula, the velocity of the
torpedo relative to the Earth is

1 1 5 10
(2) vc = c + c = c = c
2 3 6 12
3 9
Since this is greater than 4 c = 12 c, the torpedo will eventually overtake
the pirates ship.
However, using the relativistic formula 1, the velocity of the torpedo rel-
ative to the Earth is

1 1
2c + 3c 5 20
(3) vr = = c= c
1 + 16 7 28
This is less than 34 c = 21
28 c so the torpedo will never catch the pirates.
P INGBACKS
Pingback: Velocity addition: chasing space pirates viewed in four refer-
ence frames

1
SIMULTANEITY; SEEING VERSUS OBSERVING

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 5.
The standard thought experiment for proving that the simultaneity of two
events depends on the observer is that of a railway car of length L travelling
down a straight track at some speed v. In the centre of the car is a light
bulb which is switched on just as the bulb passes the origin of an observer
standing beside the track. To an observer inside the car, the light appears to
reach both ends of the car at the same time, namely L/2c.
The observer beside the track, however, sees the light hit the back end of
the car before it hits the front end, because to this observer, the light still
travels with speed c, and the back end of the car is moving towards the light
source while the front end is moving away from it. To this observer, the
distance the light must travel to reach the front end is

L
(1) df = + vt f
2

where t f is the time taken for the light to reach the front end. This time must
also be equal to the distance travelled divided by c, so

L
2+ vt f
(2) tf =
c
L
(3) tf =
2 (c v)
A similar calculation for the back end gives

L
(4) tb =
2 (c + v)
Since t f 6= tb if v > 0, the observer beside the track does not see the light
hitting both ends of the car at the same time.
Griffiths draws a distinction between what someone sees and what he
observes, which to my mind, is a bit confusing since seeing and observing
are essentially the same thing. He defines seeing as the unprocessed data
you receive about some phenomenon, and observing as what you conclude
1
SIMULTANEITY; SEEING VERSUS OBSERVING 2

about that phenomenon after interpreting the data. His point, however, is
valid: you need to make sure that you interpret what you see to draw the
correct conclusion. For example, if two cannons at different distances from
you are fired in succession so that the bangs they make when fired happen
to arrive at your location at the same time, it is incorrect to interpret this as
showing that the cannons were fired at the same time. Clearly you have to
take into consideration the distance each cannon is from you and divide that
by the speed of sound to determine when it was fired.
Example. Suppose there are synchronized clocks at 1 million km inter-
vals in a straight line, with the first clock sitting right beside you. The
nth clock is n 109 m away from you, so the light signal will take time
tn = n 109 /c to reach you. When the clock next to you reads 12 noon,
the time you will see on the nth clock will read tn seconds before noon,
although if you allow for the travel time, the actual time registered by that
clock when your clock reads 12 noon is also 12 noon.
For the 90th clock down the line, you will see it read

90 109
(5) t90 = = 300 s
3 108

before 12 noon, or 11.55 AM. Thus you see the time on the 90th clock as
11.55 AM but you observe it to be 12 noon.
P INGBACKS
Pingback: Time dilation: resolving the paradox
Pingback: The twin paradox analyzed using Lorentz transformations
APPARENT SPEEDS GREATER THAN THE SPEED OF LIGHT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 6.
Although no object can travel faster than light, it is possible for the ap-
parent speed of an object to be greater than c. A common example is that
of the apparent motion of a star across the sky. Suppose a star is a distance
a = L from Earth at time ta and that its velocity v is towards Earth at an
angle to the line of sight. At time tb it arrives at a point b whose distance
from Earth is

(1) b = L v cos (tb ta )

The times of arrival at Earth of the light emitted at distances a and b are

L
(2) Ta = ta +
c
1
(3) Tb = tb + [L v cos (tb ta )]
c

During this time, the star moves a distance perpendicular to the line of
sight of v sin (tb ta ), so the apparent speed as seen from Earth is

v sin (tb ta )
(4) u =
Tb Ta
v sin (tb ta )
(5) =
(tb ta ) 1 vc cos


v sin
(6) =
1 vc cos

This speed has a maximum at an angle which can be found by setting


the derivative to zero:
1
APPARENT SPEEDS GREATER THAN THE SPEED OF LIGHT 2

v cos ( ) 1 v2 (sin ( ))2 v cos ( ) 2


   
du
(7) = v cos ( ) 1 1
d c c c
(cos ( ) c v) cv
(8) =
(cos ( ))2 v2 2 c cos ( ) v + c2
(9) =0
(10)
v
cos =
c
At this angle, the apparent speed is
p
v 1 v2 /c2
(11) umax =
1 v2 /c2
(12) = v
Since as v c, umax can be much larger than c even though the
actual speed of the star is less than c. This again illustrates the importance
of correctly interpreting the raw data that we see, and of allowing for the
travel time of the light.
TIME DILATION: RESOLVING THE PARADOX

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 7.
Using our usual setup of a train moving with velocity v next to an ob-
server standing beside the track, suppose we have a light bulb on the ceiling
of the train which is turned on at t = 0. If the height of the train is h, then to
observer T on the train, the light takes a time tT = h/c to reach the floor,
since the light just travels straight down from ceiling to floor. To the ob-
server G on the ground, the light travels in a diagonal path due to the trains
motion, so if it takes time tG to reach the floor, this time is determined by

1
q
(1) tG = h2 + (vtG )2
c
Solving for tG :

h
(2) tG = p
c 1 v2 /c2
[Note that weve implicitly assumed that the height of the train is mea-
sured to be h to both observers. Well return to this point when we consider
length contraction.]
Since tT = h/c, we get the time dilation formula:

(3) tG = tT
1
(4) p
1 v2 /c2
That is tG > tT , a principle that is often stated as moving clocks run
slow. Another way of putting it is that the proper time is always the mini-
mum time measured between two events. The proper time is the time mea-
sured by an observer moving with the clock being used to make the time
measurements. In the case of the train, the proper time is tT since this is
the time interval measured by a clock moving with the train.
Example. A popular demonstration of time dilation is the effect it has on
the lifetime of unstable particles, such as the muon which has a half-life of
1
TIME DILATION: RESOLVING THE PARADOX 2

2 106 s when it is at rest. If the muon is travelling at some speed v, its


lifetime measured by proper time is still 2 106 s, since the proper time
clock is moving with the muon. To an observer in the lab, however, the
elapsed time is longer by a factor of . Suppose a muon is observed in the
lab to move 800 m from the time of its creation to its decay (and that its
proper time lifetime is equal to the half-life). In that case, its lab lifetime is

(5) tL = 2 106

and is speed is given by

800 m
(6) v =
2 106
r
v2
(7) = 4 108 1
c2
(8) v = 2.4 108 m s1
If we hadnt taken time dilation into account its velocity would be v =
800/2 106 = 4 108 m s1 which is greater than c.
Newcomers to relativity often get confused by the apparent contradiction
with time dilation. After all, if the train clock runs slow relative to the
ground clock, then the ground clock must run fast relative to the train clock.
But to an observer on the train, it is the ground clock that is moving, so it
should be running slow relative to the train clock.
This apparent paradox arises because what we are actually measuring in
the above experiment with the light beam isnt symmetric between the two
observers. Observer T on the train uses only one clock to measure both the
start and end points of the lights journey from ceiling to floor. Observer
G on the ground, however, must use two clocks, one of which is situated at
the point where the light leaves the ceiling and the second at a point down
the line where the light reaches the floor. As these two clocks are stationary
relative to each other, observer G can synchronize them so that the times
registered on the two clocks will always agree. However, as weve seen,
simultaneity is relative, so two events that appear synchronous to G will not
appear synchronous to T . Thus T disputes Gs claim that his clocks are
synchronized, which results in the difference in time measurements.
Another way of looking at it is this. Suppose we now have two trains,
one of which is stationary (the G train) and the other (the T train) which
moves with speed v relative to G. This time, both trains run the experiment
of firing a light beam from the ceiling to the floor. As weve seen above,
TIME DILATION: RESOLVING THE PARADOX 3

observer G thinks that it takes a longer time for the light beam on T to reach
the floor than T does. By the same logic, T will think it takes longer for Gs
light beam to reach the floor of G than G does.
T says that it takes time h/c for his own light beam to reach his own floor.
However, if G fires his light beam at the same time as T (in other words,
both G and T define their respective time origins by the simultaneous firing
of their light beams when the two trains are at the same location), T says
that, since Gs beam has further to travel, by the time T s beam has hit the
floor, Gs beam hasnt yet completed its journey. According to G, his own
beam takes time h/c to reach the floor, so if his own beam hasnt yet reached
the floor, Gs time must be less than h/c, showing that Gs clock runs slow
relative to T s clock. The important point is that the clock that runs slow in
both cases is the one that is fixed relative to the respective train. Thus Gs
fixed clock runs slow relative to the two clocks that T uses to measure the
travel time of Gs light beam, and T s fixed clock runs slow relative to the
two clocks that G uses to measure the travel time of T s light beam. The
fact that both experiments involve comparing one clock in one reference
frame to two clocks in the other reference frame shows that the theory is
consistent and there is no paradox.
P INGBACKS
Pingback: Time dilation: a rocket passing Earth
Pingback: Lorentz (length) contraction: a simple example
Pingback: Lorentz transformations: forward and backward
Pingback: The twin paradox analyzed using Lorentz transformations
Pingback: Four-velocity again
Pingback: The light clock and time dilation
Pingback: Length contraction and time dilation: a few examples
Pingback: Another trip to Alpha Centauri: more Lorentz transformation
examples
TIME DILATION: A ROCKET PASSING EARTH

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 8.
As an example of time dilation, suppose a rocket going at v = 35 c passes
Earth at t = 0 (Griffithss statement of the problem is that the rocket leaves
Earth at that time, but that would involve acceleration, so to be picky, Ill
just have the rocket going at a constant speed passing Earth at t = 0). One
hour later, as measured on the rocket, the rocket sends a light signal back
to Earth. According to clocks in Earths frame, at what time did the rocket
send the signal?
As always in time dilation calculations, we need to identify the frame
in which the clock measures proper time, that is, the frame in which the
clock is fixed to the moving object. In this case, the 1 hour measured by the
rockets clock is the rockets proper time, and the time in Earths frame is
measured by two clocks, one on Earth when the rocket passes by, and one
beside the rocket when it sends the signal. Therefore in this case, it is the
rockets clock that appears to run slow, so the time measured by Earth will
be longer than 1 hour.

(1) tE = tR
1 hour
(2) = p
1 v2 /c2
5
(3) = hours
4
Now we want to find the time, as measured by Earth, that the signal
arrives at Earth. Here we can do all calculations in Earths frame. Since
the signal left the rocket at t = 54 hours and the rocket is travelling at 35 c it
was at a distance of d = 54 35 c = 34 c light hours. Therefore it takes the signal
3 5 3
4 hour to arrive back on Earth, so it will arrive at t = 4 + 4 = 2 hours.
How long after the rocket left Earth did the signal arrive on Earth, accord-
ing to an observer on the rocket? This time, we are looking at two events
that both occur on Earth, so the 2 hours (as measured by Earth) that have
elapsed on Earth between the rocket passing by and the signal arriving is
the proper time, so the time measured by the rocket will be longer than this.
1
TIME DILATION: A ROCKET PASSING EARTH 2

To measure the time in the rockets frame, the rocket will have to use two
clocks, one at the location of the Earth when the rocket passes it, and the
other at the location of the Earth when the signal arrives. Therefore

(4) tR = tE
5
(5) = 2 hours
4
5
(6) = hours
2
LORENTZ (LENGTH) CONTRACTION: A SIMPLE EXAMPLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 9.
In addition to time dilation, the other main kinematic prediction of special
relativity is length contraction. The formula can be derived using a similar
light bulb in a train experiment that we used to derive time dilation. As
usual, we have a train moving at speed v along a straight track, and an
observer T on the train and a second observer G on the ground. We have a
light bulb at one end of the trains car and a mirror at the other end. If the
length of the car, as measured by T , is xT , then the time taken for the light
to make a round trip from the bulb to the mirror and back is

xT
(1) tT = 2
c
To G, however, the the outward and return trips of the light take different
times, due to the motion of the train. The outward journey takes

xG + vtGo
(2) tGo =
c

and the return journey takes

xG vtGr
(3) tGr =
c
Note that we are assuming that the length of the car may be different to
the two observers. We can solve these two equations for the times and get

xG
(4) tGo =
cv
xG
(5) tGr =
c+v
The total time measured by G is therefore
1
LORENTZ (LENGTH) CONTRACTION: A SIMPLE EXAMPLE 2

(6) tG = tGo + tGr


 
1 1
(7) = xG +
cv c+v
 
2c
(8) = xG 2
c v2
2xG 2
(9) =
c

The two times are related by the time dilation formula. To apply this
correctly, we need to note that T uses only the one clock to measure both
the departure and arrival of the light, since these two events happen at the
same place in his frame. Observer G must use two clocks, since the train
moves relative to G between the events. Thus T s time is the proper time
and must be less than Gs time, so

(10) tG = tT

Combining this with 1 and 9 we get

2xG 2 xT
(11) = 2
c c
xT
(12) xG =

Therefore the length of the car as measured by G is shorter than the length
measured by T , by the factor . This is the length (or Lorentz) contraction
effect.

Example. Suppose we have two spaceships A and B, and the rest length
LA of A is twice the rest length LB of B. If B is moving at vB = c/2 and
A is moving at a speed vA that makes A appear the same length as B to an
observer at rest, how fast is A moving?
We must have
LORENTZ (LENGTH) CONTRACTION: A SIMPLE EXAMPLE 3

LA LB
(13) =
A B
2 1
(14) =
A B
1 c2 /4 3
(15) = 1 =
B2 c2 4
v2A
 
3
(16) 4 1 2 =
c 4

13
(17) vA = c
4
P INGBACKS
Pingback: Lorentz contraction: no contraction in directions perpendicu-
lar to the motion
Pingback: Lorentz contraction in a rotating disk: Ehrenfests and Bells
spaceship
Pingback: Lorentz transformations: forward and backward
Pingback: Length contraction and time dilation: a few examples
Pingback: Another trip to Alpha Centauri: more Lorentz transformation
examples
LORENTZ CONTRACTION: NO CONTRACTION IN
DIRECTIONS PERPENDICULAR TO THE MOTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 10.
Weve seen that the length of a moving object in the direction of relative
motion is contracted by the factor . What about lengths perpendicular to
the relative motion? It turns out that these lengths are unaffected by the
relative motion.
Ive never seen a mathematical derivation of this result. Most books just
state it as though it were obvious, but Griffiths quotes a thought experiment
originally due to E. F. Taylor and J. A. Wheeler in 1966. Suppose we have
our usual moving train, with an observer T on the train and another observer
G on the ground. Next to the train is a wall, and before the train arrives, G
paints a horizontal red line on the wall at a distance of 1 m above the ground,
as measured by G. Then, as the train goes by, T leans out the window and
paints a blue line at a distance of 1 m above the ground, as measured by T .
After the train passes by, there are two lines painted on the wall. Which is
higher?
If there were some sort of contraction effect for distances perpendicular
to the motion, then G would say that T should perceive his 1 m above the
ground to be less than Gs 1 m, so the blue line should be below the red
one. However, to T , it is G that is moving so it is G that should perceive his
1 m to be less than T s 1 m, so the red line should be lower. As these two
results are contradictory, the only rational conclusion is that both T and G
measure 1 m to be exactly the same distance, and the blue and red lines are
at exactly the same height.
This argument might bother you, since it seems similar to arguments that
time dilation and length contraction (in the direction of motion) are contra-
dictory. If T s clock moves relative to Gs, then T s clock must run slow,
but to T , it is Gs clock that is moving and so should be running slow. Simi-
larly, for length contraction, each observer sees the others lengths as being
contracted. However, in both of these situations, the apparent paradox is
resolved by taking into account the discrepancy in simultaneity between the
two observers, and in both these cases, the disagreements relate to events
that occur at specific coordinates in both frames and that leave no lasting
effect.
1
LORENTZ CONTRACTION: NO CONTRACTION IN DIRECTIONS PERPENDICULAR TO THE MOTION
2

In the case of painting lines on the wall, the two lines remain on the wall
after the experiment is finished and, if there were a contraction for either
observer, the two lines would have to be at different heights on the wall. No
jiggery-pokery with simultaneity can change that. The fact that assuming
a contraction in the perpendicular direction leads to a contradiction shows
that lengths perpendicular to the motion are unchanged.
Example. Suppose a sailboat moves at speed v relative to an observer on
the shore. The sailboat has a mast of length L that is anchored near the front
of the boat and makes an angle (when the boat is at rest) of with the deck
of the boat. What angle will the observer on the shore see?
Since only the component of the mast parallel to the motion is contracted,
the shore observer will see an angle 0 that satisfies:

L sin L sin 0
(1) =
1
L cos
L cos 0
(2) tan 0 = tan
The faster the boat goes, the larger is , so the apparent angle will in-
crease towards /2. This makes sense, since the horizontal component gets
contracted while the vertical component remains the same.
P INGBACKS
Pingback: Lorentz contraction in a rotating disk: Ehrenfests and Bells
spaceship
Pingback: Lorentz transformations: forward and backward
Pingback: Velocity addition formulas for all 3 directions
Pingback: Transforming the electric field in relativity
Pingback: Lorentz transformations: derivation from symmetry
Pingback: The light clock and time dilation
LORENTZ CONTRACTION IN A ROTATING DISK:
EHRENFESTS AND BELLS SPACESHIP PARADOXES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 11.
Kassner, Klaus, Spatial geometry of the rotating disk and its non-rotating
counterpart, Am. J. Phys., 80, 772 (2012).
In special relativity, lengths are contracted in the direction of motion,
but not in perpendicular directions.A straightforward application of these
formulas can, however, lead you astray if youre not careful. A case in
point is that of a rotating disk of radius R. Points on the rim of the disk are
moving parallel to the rim, so we would think that the rim of the disk must
be contracted as seen by an observer at the centre of the disk. The radius
of the disk, however, is perpendicular to the direction of motion at each
point, so is not contracted. Thus the circumference of the disk would seem
to be reduced to 2R/ while the diameter remains at 2R, giving a ratio of
/ between circumference and radius. This violates basic geometry (and
is known as Ehrenfests paradox).
The first point to note is that a rotating disk is not an inertial frame, since
all points on the disk are experiencing centripetal acceleration towards the
centre.
What actually happens is that the material in the disk physically stretches
as it accelerates to its final rotation speed in such a way as to balance the
Lorentz contraction and retain a ratio of from circumference to diameter.
In other words, Lorentz contraction is a real, physical effect that actually
deforms the material. That this is so is most easily seen by examining Bells
spaceship paradox.
Suppose we have two spaceships, A and B, that accelerate from rest in
such a way that the distance between them is always constant, as seen by an
earthbound observer E (see diagram):
1
LORENTZ CONTRACTION IN A ROTATING DISK: EHRENFESTS AND BELLS SPACESHIP PARADOXES
2

In the diagram, the solid green and red paths are parallel (well, they are
meant to be anyway; the diagram is a bit skewed). The pink dashed line
connecting A and B represents the positions of the two spacecraft at one
specific time as measured by E. [The grey dashed line is the world line of a
light beam leaving A.]
Now lets look at things from the point of view of an observer aboard
A. As coordinate axes are shown as the pale green dashed lines, for a
particular point in the flight. At this time, according to A, ship B will be at
point C, since points A and C have the same time coordinate in As system.
Although we cant compare the lengths of the lines AB and AC directly from
the diagram since the units along the x and x0 axes have different lengths,
what we can say is that since C is further along ship Bs world line than the
point B, it has accelerated more than A so the distance between the two ships
must be larger than the distance AB. If the two ships were connected by a
string of length AB, then as the two ships accelerate, the string will break.
Observer A thinks this is because B is getting further away, but observer E
says it happens because as the string speeds up, it gets Lorentz contracted
so it becomes shorter than the distance AB and therefore breaks.
To apply this to the rotating disk, consider two points close to each other
on the rim. As the disk spins up, these two points are experiencing the same
tangential acceleration, so the situation is essentially the same as with the
LORENTZ CONTRACTION IN A ROTATING DISK: EHRENFESTS AND BELLS SPACESHIP PARADOXES
3

two accelerating spaceships. As a result, one point on the rim sees the other
point moving further away from it as long as the disk keeps accelerating.
That is, the disk actually stretches as it accelerates.
We havent actually proved that the amount of stretch is the correct amount
to balance the Lorentz contraction and keep the ratio of circumference to di-
ameter equal to , but hopefully the qualitative explanation helps a bit.
LORENTZ TRANSFORMATIONS: FORWARD AND BACKWARD

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 12.
We can combine the effects of time dilation and length contraction to de-
rive the Lorentz transformations. As usual, we align the two inertial frames
S and S (which moves to the right at speed v relative to S ) so that the x
and x axes coincide at their origins also coincide at t = t = 0. Consider a
point on the x axis (that, obviously, has coordinate x in the S system). Due
to length contraction, the distance from the origin of S to x as measured in
S is (remember that the measurement of a distance in S means that both
endpoints of the distance are measured at the same time t, and that these
two measurement events in S will be seen as occurring at different times t1
and t2 in the S system):

x
(1) d=

so, since S is moving at speed v relative to S , the location of point x as


measured by S is

(2) x = d + vt
x
(3) = + vt

(4) x = (x vt)

Obviously, we can invert the procedure by looking at the point of view of


S . In this case, everything is the same, except that S is moving to the left

with speed v, so

(5) x = (x + vt)

We can use this equation to eliminate x from 4 to get


1
LORENTZ TRANSFORMATIONS: FORWARD AND BACKWARD 2

x
(6) vt = (x vt)

 
1 x
(7) t = (x vt)
v
v2
   

(8) = x 1 2 1 + vt
v c
 xv 
(9) = t 2
c

Since directions perpendicular to the motion dont change, the full set of
Lorentz transformations is

(10) x = (x vt)
(11) y = y
(12) z = z
 xv 
(13) t = t 2
c

As mentioned above, its fairly obvious that the inverse transformations


can be obtained by replacing v by v:

(14) x = (x + vt)
(15) y = y
(16) z = z
 
xv
(17) t = t + 2
c

If you must, you can derive these from the original transformations. Start-
ing with 4 and then 13:
LORENTZ TRANSFORMATIONS: FORWARD AND BACKWARD 3

x
(18) x = + vt

vx v2
 
(19) t = t 2 2t
c c
 
t vx
(20) = 2
2
c
t vx
(21) =
c2
 
xv
(22) t = t + 2
c
Starting with 13 and then 4

t vx
(23) t = +
c2
vt v2
 
(24) x = x 2 x
c
 
x vt
(25) =
2
(26) x = (x + vt)
P INGBACKS
Pingback: Lorentz transformations and simultaneity
Pingback: Velocity addition formulas for all 3 directions
Pingback: The twin paradox analyzed using Lorentz transformations
Pingback: Invariance of scalar product under Lorentz transformations
Pingback: Compound Lorentz transformations
Pingback: Rapidity
Pingback: The invariant interval: some examples
Pingback: Spacetime diagrams: an example
Pingback: Four-velocity versus ordinary velocity: an example
Pingback: Contravariant gradient operator
Pingback: Lorentz transformations: derivation from symmetry
LORENTZ TRANSFORMATIONS AND SIMULTANEITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 13.
Heres a simple example of using the Lorentz transformations. Suppose
that two events occur simultaneously in the Earth frame, a distance 500 km
apart. An observer in a (very fast) plane travelling at 12
13 c along the line join-
ing the events passes the first event A such that their respective coordinate
origins coincide at that event. At what time does the plane observer think
event B occurs?
The Lorentz transformations are

(1) x = (x vt)
 xv 
(2)
t = t 2
c
Event A occurs at the same time t = t = 0 in both systems. For event B,
x = 500 km and t = 0, so
 
500v
(3) t = 2
c
13
(4) =
5  
13 12c 500
(5) t =
5 13 c2
1200
(6) =
c
1200 km
(7) =
3 105 km s1
(8) = 4 103 s
Thus the plane observer thinks that B occurs 4 milliseconds before A.

1
VELOCITY ADDITION FORMULAS FOR ALL 3 DIRECTIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.14.
Reference: Carroll, Bradley W. & Ostlie, Dale A. (2007), An Introduc-
tion to Modern Astrophysics, 2nd Edition; Pearson Education - Chapter 4,
Problem 4.10.
The relativistic velocity addition formula is most easily derived from the
Lorentz transformations:

(1) x = (x vt)
 xv 
(2) t = t 2
c

As usual, system S is moving to the right relative to system S with


velocity v. If an object is moving in the S system with velocity u = dx/dt,
then we can take differentials on both sides of the above equations to get

(3) d x = (dx v dt)


 v 
(4) dt = dt 2 dx
c
d x
(5) u =
dt
dx v dt
(6) =
dt cv2 dx
(dx/dt) v
(7) =
1 cv2 (dx/dt)
uv
(8) =
1 uv/c2
x v
(9) x =
1 xv/c2

By symmetry (or algebra if you dont trust the physical argument) this
relation can be inverted to give
1
VELOCITY ADDITION FORMULAS FOR ALL 3 DIRECTIONS 2

u + v
(10) u=
1 + uv/c2

We can also derive the formulas for transforming velocity components


perpendicular to the motion. Although perpendicular distances dont change
under the Lorentz transformation, perpendicular velocities do, because they
involve the ratio of perpendicular distance to time, and time does change
from one system to another. We get, with a dot indicating a time derivative
in the respective system:

d y
(11) y =
dt
dy
(12) =  
v
dt c2 dx
y
(13) =
(1 xv/c2 )
z
(14) z =
(1 xv/c2 )

Thus perpendicular velocities depend not only on the perpendicular ve-


locity in the original system, but also on the horizontal velocity.

Example. Suppose we have a spotlight mounted on the roof of our usual


train, and the spotlight is aimed to point towards the rear of the train, making
an angle of with the roof of the train. If the train moves with speed v, what
angle will a ground observer see the beam make with the trains roof?
In the trains system, the lights velocity components are

(15) x = c cos
(16) z = c sin

In the ground system, we get from 10:


VELOCITY ADDITION FORMULAS FOR ALL 3 DIRECTIONS 3

c cos + v
(17) x =
1 v (cos ) /c
c sin
(18) z =
(1 v (cos ) /c)
z
(19) tan =
x
c sin
(20) =
(c cos v)
As a check, we can work out x2 + z2 = c2 (after simplifying). When
the beam points directly up in the trains frame ( = 2 ), it points slightly
forward in the grounds frame. When cos = vc , tan = so the beam
points directly up in the grounds frame.
P INGBACKS
Pingback: Velocity addition: chasing space pirates viewed in four refer-
ence frames
Pingback: Spacetime diagrams: an example
Pingback: Four-velocity versus ordinary velocity: an example
Pingback: Relativistic momentum and energy
Pingback: Relativistic transformation of force, electric and magnetic fields
Pingback: Transforming the electric field in relativity
VELOCITY ADDITION: CHASING SPACE PIRATES VIEWED
IN FOUR REFERENCE FRAMES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 15.
As an example of using the relativistic velocity addition formula, we can
revisit the problem of the space police chasing some pirates. In the problem,
space pirates are fleeing from the solar system police in a spacecraft that is
moving at 34 c (relative to the Earth). The polices spaceship is travelling at
only 12 c but in an attempt to stop the pirates they fire a torpedo at them. The
torpedos velocity, relative to the polices ship, is 13 c.
Originally, we worked out the velocity of the torpedo relative to Earth as
5
7 c and saw that the torpedo would not catch the pirates. We can use the
velocity addition formula to work out the velocity of each component in the
problem relative to every other component and check that the pirates escape
no matter how we look at it. If an object has velocity u in system S which
is moving at speed v relative to system S , then the objects velocity u as
measured in S is

u + v
(1) u=
1 + uv/c2

The velocity of the torpedo relative to the Earth is

1 1
2c + 3c 5 20
(2) vr = = c= c
1 + 16 7 28

The speed v pp of the pirates relative to the police is found from the con-
dition that the velocity v pe of the pirates relative to Earth is

v pp + 21 c
(3) v pe =
1 + 12 c v pp /c2


Solving for v pp we get


1
VELOCITY ADDITION: CHASING SPACE PIRATES VIEWED IN FOUR REFERENCE FRAMES
2

v pe 12 c
(4) v pp =
1 12 c v pe /c2


3 1
4 c  2 c
(5) = 1 3 2
1 2c 4 c /c
2
(6) = c
5
Similarly, the velocity vt p of the torpedo relative to the pirates is

5 3
7c 4c
(7) vt p = 5
 3

1 2
7c 4 c /c
1
(8) = c
13
In summary, the following matrix gives the velocity of the object named
in the column header relative to the object in the row header:
Earth Police Pirates Torpedo Escape?
1 3 5
Earth 0 2c 4c 7c Yes
1 2 1
Police 2 c 0 5c 3c Yes
3 2 1
Pirates 4 c 5 c 0 13 c Yes
5 1 1
Torpedo 7 c 3 c 13 c 0 Yes
The matrix is antisymmetric (Ai j = A ji ) since the velocity of A relative
to B is just the negative of the velocity of B relative to A. In all cases,
vtorpedo < v pirates so the pirates always escape.
THE TWIN PARADOX ANALYZED USING LORENTZ
TRANSFORMATIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 16.
The twin paradox can be analyzed using Lorentz transformations to get
an idea of what each twin perceives at key points in the journey. Suppose
we have a brother Andy and his twin sister Betty. At t = t 0 = 0 (which
corresponds to the twins 21st birthday) and x = x0 = 0 in both systems,
Betty leaves on a spaceship travelling at 54 c heading towards a star X. Upon
arriving at X, Betty immediately transfers to another spaceship and heads
back to Earth, also at 45 c, arriving on her 39th birthday (according to her
own reckoning).
Andy sees Betty as moving in both directions, so to him, Bettys clock
runs slow. Using the time dilation formula, Andy sees a time interval of

(1) tA = tB
5
(2) = 18
3
(3) = 30 years

Thus Andy will be 21 + 30 = 51 years old when Betty arrives home.


According to Andy, Bettys trip took 15 years each way at a speed of 45 c,
so star X is at a distance of

4
(4) dX = 15c = 12c light years
5
In Andys frame, the coordinates of Bettys jump between spaceships is

(5) (x,t) = (12, 15)

In Bettys outgoing frame, we can apply Lorentz transformations to get


her coordinates
1
THE TWIN PARADOX ANALYZED USING LORENTZ TRANSFORMATIONS 2

(6) x0 = (x vt)
 
5 4
(7) = 12c 15c
3 5
(8) = 0
 xv 
(9) t0 = t 2
 c 
5 4
(10) = 15 12
3 5
(11) = 9 years

These values are consistent, since Betty doesnt move relative to her
own frame, so wed expect x0 = 0, and since her clock runs slow relative
to Andys clock, the time interval is shortened by a factor of 1/ giving
t 0 = 9 years.
Now consider the frame (indicated by a double prime) fixed to Bettys
returning spaceship. Well fix the coordinates so that t = t 0 = t 00 = 0 and
x = x0 = x00 = 0. Its velocity relative to Andy is v = 45 c, so in this frame,
the jump occurs at

(12) x00 = (x + vt)


 
5 4
(13) = 12c + 15c
3 5
(14) = 40 light years
 xv 
(15) t 00 = t + 2
 c 
5 4
(16) = 15 + 12
3 5
(17) = 41 years

Thus for Betty to adjust her clock so that it agreed with the S 00 frame,
she would have to advance it by 32 years just after making the jump. In this
frame, the coordinates of her arrival back on Earth are given by transforming
Andys coordinates of (x,t) = (0, 30), so we have
THE TWIN PARADOX ANALYZED USING LORENTZ TRANSFORMATIONS 3

(18) x00 = (x + vt)


 
5 4
(19) = 0 + 30c
3 5
(20) = 40 light years
 xv 
(21) t 00 = t + 2
c
5
(22) = (30 + 0)
3
(23) = 50 years

Note that x00 doesnt change as Betty travels home, again because this is
her own rest frame. Also, the time interval for her to get home is again 9
years in her own frame, the same as for the outbound journey.
Andys age, as viewed by Betty, depends on which spaceship she is on.
Just before making the jump she is in the S 0 frame, so according to her,
the time back on Earth at that point is given by taking t 0 = 9, x = 0 in the
transformation

 xv 
0
(24) t = t 2
c
5
(25) 9 = (t 0)
3
27
(26) t = = 5.4 years
5
That is, Betty thinks Andy is actually 3.6 years younger than she is when
she reaches the star. This is because to her, it is Andy who is moving so
his clock runs slow by a factor of 1/ relative to her 9 years. How can we
reconcile this with the fact that Andy thinks that Betty takes 15 years to
get to the star, so according to him, he is actually 6 years older than Betty
when she reaches the star? The confusion arises because of differences
in simultaneity as perceived by Andy and Betty. Andy says that the two
events (Andy on Earth at age 21 + 15 = 36 years, and Betty at the star
at age 21 + 9 = 30 years) are simultaneous, but Betty disagrees with this,
saying that the two events (Andy on Earth at age 21 + 5.4 = 26.4 years, and
Betty at the star at age 21 + 9 = 30 years) are simultaneous. In general,
two observers can agree about two events being simultaneous only if these
events happen at the same location in both systems.
Just after the jump, Betty can now calculate Andys age in the S 00 system
by using t 00 = 41 years and x = 0:
THE TWIN PARADOX ANALYZED USING LORENTZ TRANSFORMATIONS 4

 xv 
(27) t 00 = t + 2
c
5
(28) 41 = (t + 0)
3
123
(29) t = = 24.6 years
5
Thus Betty now thinks that Andy is 21 + 24.6 = 45.6 years old. Andys
age, of course, hasnt changed as Betty jumps between ships; only her per-
ception of it has changed.
In her own frame, Betty says the return trip to Earth takes 9 years, so by
the same logic as in 26, she says that 5.4 years elapse for Andy, making
him 45.6 + 5.4 = 51 years old when she arrives home. Thus both Andy and
Betty agree on Andys age.
The difference between the two twins is, of course, that Betty under-
goes deceleration and then acceleration in the reverse direction when she
jumps between ships at the star. A more realistic treatment would have
Betty gradually decelerating as she approached the star and then gradually
accelerating as she got onto the other ship for the journey home. During
this process, Betty is continuously changing reference frames resulting in
her seeing Andy age rapidly (but continuously, rather than the jump in age
from 26.4 to 45.6 that happens here) in the process. Thus the paradox is
real, in the sense that different amounts of time actually do pass for the two
twins.
P INGBACKS
Pingback: Another trip to Alpha Centauri: more Lorentz transformation
examples
INVARIANCE OF SCALAR PRODUCT UNDER LORENTZ
TRANSFORMATIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.17.
Reference: Carroll, Bradley W. & Ostlie, Dale A. (2007), An Introduc-
tion to Modern Astrophysics, 2nd Edition; Pearson Education - Chapter 4,
Problem 4.11.
Although the time and position of an event can change under Lorentz
transformations, the scalar product of any two four-vectors is an invariant
under a Lorentz transformation. That is

(1) ai bi = ai bi

where for motion along the 1-axis (x axis) the transformations are

a0 = a0 a1

(2)
a1 = a1 a0

(3)
(4) a2 = a2
(5) a3 = a3

and

(6) a0 = a0
(7) aj = aj

for j = 1, 2, 3. We can see this directly by calculation, using 2 = 1/ 1 2 :



1
INVARIANCE OF SCALAR PRODUCT UNDER LORENTZ TRANSFORMATIONS 2

(8)
ai bi = 2 a0 a1 b0 b1 + 2 a1 a0 b1 b0 + a2 b2 + a3 b3
   

(9)
= 2 a0 b0 1 + 2 + a1 b0 ( ) + a0 b1 ( ) + a1 b1 2 + 1 + a2 b2 + a3 b3
  

(10)
= a0 b0 + a1 b1 + a2 b2 + a3 b3
(11)
= ai bi
In particular if ai is the space-time four-vector of an event, or the differ-
ence between the four-vectors of two events, then

(12) ai ai = ai ai
This leads to the invariant interval between two events:

(13) s2 t 2 + x2 + y2 + z2
P INGBACKS
Pingback: The invariant interval: some examples
COMPOUND LORENTZ TRANSFORMATIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 18.
The Lorentz transformations can be written in matrix form as

0 0
0 0
(1) x =
0

0 1 0
0 0 0 1

where the 0 (ct) component is the first row and first column, followed by
the 1, 2, and 3 directions in order. This matrix is for relative motion along
the 1 axis.
The Galilean transformations can be written as a matrix as well, where
the first coordinate is just t rather than ct:

1 0 0 0
v 1 0 0
(2) =
0

0 1 0
0 0 0 1

or if we want to use the same symbols as in the Lorentz case, where the top
row of is a ct coordinate, we can write

1 0 0 0
1 0 0
(3) =
0

0 1 0
0 0 0 1
The Lorentz transformation along the 2 (y) axis is obtained by putting the
transformation terms in row and column 2:

0 0
0 1 0 0
(4) y =


0 0
0 0 0 1
1
COMPOUND LORENTZ TRANSFORMATIONS 2

If we apply a Lorentz transformation first in the x and then in the y direc-


tion (with different relative velocities), we get the compound matrix:

y 0 y y 0 x x x 0 0
0 1 0 0 x x
x 0 0
(5) y x =
y y 0

y 0 0 0 1 0
0 0 0 1 0 0 0 1

y x y x x y y 0
x x x 0 0
(6) = y y x y x x y

y 0
0 0 0 1
Note that although x and y are both symmetric, their product is not.
This means that applying the transformations in the opposite order gives a
different result.

(7) x y = Tx Ty
(8) = (y x )T

y x x x y y x 0
y x x x y x x y 0
(9) = y y

0 y 0
0 0 0 1
RAPIDITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 19.
An alternative way of writing the Lorentz transformations is to define a
quantity called the rapidity:

(1) tanh1
Using this definition, we have

1
(2) = p
12
1
(3) = p
1 tanh2
cosh
(4) = p
cosh2 sinh2
(5) = cosh

since cosh2 sinh2 = 1.


Also

(6) = cosh tanh = sinh

so

0 0
0 0
(7) =
0

0 1 0
0 0 0 1

cosh sinh 0 0
sinh cosh 0 0
(8) =
0 0 1 0
0 0 0 1
1
RAPIDITY 2

This is similar to a rotation through an angle in 3-d space, except both


the sinh terms are negative.
The velocity addition formula becomes

u+v
(9) u =
1 + uv/c2
u + v
(10) = c
1 + u v
tanh u + tanh v
(11) u =
1 + tanh u tanh v
(12) tanh u = tanh (u + v )

where in the last line weve used the formula for the tanh of a sum of two
arguments.
The rapidities therefore simply add, giving a simpler measure of rela-
tivistic velocity:

(13) u = u + v
P INGBACKS
Pingback: Four-velocity again
THE INVARIANT INTERVAL: SOME EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 20-21.
The invariant interval in special relativity is the scalar product of the in-
terval between two events with itself:

(1) s2 (x)i (x)i


Since x is the difference of two four-vectors, it too is a four-vector so
the invariance under Lorentz transformations follows from that fact.
Because the 0 term is negative and the other three terms are positive, s2
can be negative, zero or positive. This gives three possible types of pairs of
events:
(1) Timelike: If s2 < 0, then it is possible to find a frame in which
the two events occur at the same spatial point, but at different times,
2
since it is the time component x0 which is negative.
(2) Lightlike: If s2 = 0 then c2 (t)2 = x2 (if the motion is along the
x axis; the argument is similar for arbitrary directions), so the events
can be connected by a light signal.
(3) Spacelike: If s2 > 0, then it is possible to find a frame in which the
two events occur at the same time but at different places. Different
observers may disagree about which event occurs first.
Example 1. In system S , an event A happens at (ct, x, y, z) = (15, 5, 3, 0)
and B happens at (5, 10, 8, 0). The interval between them is

(2) s2 = 100 + 25 + 25 + 0 = 50 < 0

so the interval is timelike. There is no frame in which A and B occur simul-


taneously. However, there is a frame where they occur at the same point.
To find this frame, its easiest to orient the coordinates so that the x axis is
along the line joining the events, which points in the direction x + y, and to
redefine the origin so that x = x = 0 andt = t = 0 when A occurs. In S , the
events are separated by a distance of 5 2, so in the new coordinate system
(at rest relative to S ) we have
1
THE INVARIANT INTERVAL: SOME EXAMPLES 2

(3) A = (0, 0, 0, 0)
 
(4) B = 10, 5 2, 0, 0

We then need to find such that in the frame S, xB = 0, so using a


Lorentz transformation, we have
 
(5) xB = 0 = 5 2 + 10

2
(6) =
2
Therefore the velocity of S relative to our original frame S is

c
(7) v = (x + y)
2
Example 2. Now we take A = (1, 2, 0, 0) and B = (3, 5, 0, 0). The interval
is

(8) s2 = 4 + 9 = 5 > 0

so the interval is spacelike. Since both events are already on the x axis,
to find a frame in which the events occur at the same time, we change the
origin to event A, giving

(9) A = (0, 0, 0, 0)
(10) B = (2, 3, 0, 0)
We now use a Lorentz transformation on the time to find such that
tB = 0:

(11) tB = 0 = (tB xB )
tB
(12) =
xB
2
(13) =
3
Example 3. The previous example is easily generalized to the case where
A = (tA , xA , 0, 0) and B = (tB , xB , 0, 0). We redefine the origin to be at A,
giving B coordinates of B0 = (tB tA , xB xA , 0, 0). Assuming the interval
THE INVARIANT INTERVAL: SOME EXAMPLES 3

is spacelike, the velocity of the frame in which A and B are simultaneous is


found from

(14) tB = 0 = (tB tA (xB xA ))


tB tA
(15) =
xB xA
P INGBACKS
Pingback: Invariance of scalar product under Lorentz transformations
Pingback: Spacelike intervals
Pingback: Lorentz transformations and causality
SPACELIKE INTERVALS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 22.
If two events are separated by a spacelike interval, neither event can af-
fect the other, since different observers may disagree about the order of the
events. Here are examples of a couple of misconceptions that sometimes
arise about such intervals.

Example 1. If two people are sitting a couple of metres apart then they
are at rest relative to each other so they share the same inertial frame and
will agree about all time and space measurements. At a particular instant of
time, the interval separating the two people is spacelike, so it might seem
that they could not communicate with each other. However, if we draw
each persons world line on a spacetime diagram, then in their own frame,
each persons world line is a vertical line (remember that ct is plotted on the
ordinate (y axis) and x on the abscissa (x axis). If they are speaking to
each other, the sound waves have world lines that travel diagonally upwards
between the two vertical world lines representing the 2 people. If person A
is at x = 0 and B is at x = 2, then if A says something at t = 0, the sound
reaches B at t = 2/v where v is the speed of sound, so the slope of the sound
waves world line is

t 2 c
(1) c =c =
x 2v v

As c  v, the sounds world line is very nearly vertical but it does angle
from As vertical line over to Bs line. Similarly, if B says something to A,
the sounds world line travels in the x direction with the same speed, so
its slope is c/v.
The interval between the events of A saying something and B hearing it
is
1
SPACELIKE INTERVALS 2

(2) s2 = c2 t 2 + x2
4
(3) = c2 2 + 4
v
c2 v2
 
(4) = 4 2 1 2
v c
For c  v, this is (a very large) negative value, so the interval between
the two events is definitely timelike.
Example 2. Suppose that faster than light travel is possible, but that light
signals still travel at c. In that case it would be possible for an object to
travel from A to B such that the interval between the events of leaving A
and arriving at B is spacelike. Since different observers can disagree on the
order in which such events occur, it is possible for some observers to say
that the object arrived at B before it left A.
However, if the object then returned from B to A (also faster than light,
say), all observers would agree that the object arrived back at A after it
left A. This is because the interval between the two events (leaving A and
arriving back at A) is timelike (since they occur at the same place in As
frame), so they must be separated by a positive time interval in every inertial
frame.
SPACETIME DIAGRAMS: AN EXAMPLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 23.
As an example of a spacetime diagram, suppose we have our usual two
inertial frames with S at rest relative to observer A and S moving at speed
= 53 in the +x direction, with the origins of the two systems coinciding as
usual.
Wed like to plot the lines of constant t and x on a spacetime diagram.
Using Lorentz transformations we have

x x
(1) ct =

ct
(2) ct = x +

For various values of x and t, these two equations give two sets of parallel
lines. The first equation gives lines with slope 1/ and ct-intercepts of
x/ , while the second equation gives lines with slope and ct-intercepts
of ct/. We can plot a few lines from each set as shown:
1
SPACETIME DIAGRAMS: AN EXAMPLE 2

The red lines are lines of constant x with the top line corresponding to
x = 3 and the bottom line to x = +3, in steps of 1. The green lines are
lines of constant t with the bottom line corresponding to t = 3 and the top
line to t = +3, again in steps of 1.
The thick blue line represents the world line of an object that starts at
(t, x) = (2, 2) and moves to (t, x) = (3, 2). We can find its velocity in
S by taking its slope on the graph. Finding the exact values of x and t is
difficult by eyeballing a graph, so we can cheat a bit and use the Lorentz
transformations to find the corresponding values. We get for the starting
point:

(3) ct1 = (ct + x)


 
5 3
(4) = 2 2
4 5
(5) = 4
(6) x1 = (x + ct)
(7) = 4

and for the end point:


SPACETIME DIAGRAMS: AN EXAMPLE 3

(8) ct2 = (ct + x)


 
5 3
(9) = 3+ 2
4 5
21
(10) =
4
(11) x2 = (x + ct)
19
(12) =
4
The velocity is then

x
(13) v =
t
35
(14) = c
37
We can check this using the velocity addition formula. Its velocity in S
is

x 4
(15) v = = c
t 5

so

4
+ 35
5 35
(16) v= 4
 3 c = c
1+ 5 5 37

so it checks out.
FOUR-VELOCITY AGAIN

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 24.
Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.
The four-velocity (or proper velocity as Griffiths calls it) is defined using
the symbol (which is used by Moore for the flat space metric) as the
derivative of the four-position with respect to proper time:

dxi
(1) i
d
Ordinary velocity (which Griffiths calls u; Moore uses u for the four-
velocity. As I say, it gets confusing.) is defined as the derivative of the
four-position with respect to the time component of the four-position:

dxi
(2) ui = c
dx0

where the factor of c is there to cancel out the c in x0 = ct. Since proper
time and coordinate time are related by
r
u2
(3) d = 1 dt
c2

we have

(4) dx0 = c dt
c d
(5) = p
1 u2 /c2
2
where u2 = 3i=1 ui , that is, the square magnitude of the spatial ordinary
velocity. Therefore
1
FOUR-VELOCITY AGAIN 2

ui
(6) i = p
1 u2 /c2
In particular

u0
(7) 0 = p
1 u2 /c2
c
(8) = p
1 u2 /c2
Note that the components i can be larger than c. This is OK because
the four-velocity is the derivative of distance in one coordinate system with
respect to the time coordinate in another system, so were not actually cal-
culating the speed of an object as measured in one specific frame (unless
that frame is the objects own rest frame, in which case u = 0 and all com-
ponents i are safely less than c).
If we sum up the squares of the spatial components, we get

3 2 u2
(9) 2 i =
i=1 1 u2 /c2
2
(10) u2 =
1 + 2 /c2

(11) u = p
1 + 2 /c2

where bold-face symbols represent 3-d (spatial) vectors.


We can express in terms of rapidity , defined as

u
(12) tanh
c
We get, assuming the motion is in the x direction (using cosh2 sinh2 =
1 in the denominator):

c2 tanh2
(13) 2 =
1 tanh2
(14) = c2 sinh2
(15) = c sinh
FOUR-VELOCITY AGAIN 3

P INGBACKS
Pingback: Four-velocity versus ordinary velocity: an example
Pingback: Four-velocitys square is a universal constant
Pingback: Four-velocity of a particle in hyperbolic motion
Pingback: Relativistic momentum and energy
Pingback: Four-acceleration and Minkowski force
FOUR-VELOCITY VERSUS ORDINARY VELOCITY: AN
EXAMPLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 25.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.
Heres an example of calculating the ordinary and four-velocities of an
object. Suppose the object moves with ordinary velocity (as measured in
the ground frame) 25 c in the 45 direction (towards the upper right; note
that there is a typo
in Griffithss statement of the problem as he states that
the velocity is 2 5c which is, of course, greater than c). This resolves into
components:

r r
2 2
(1) u= cx + cy
5 5

The four velocity is defined as

ui
(2) i = p
1 u2 /c2

so the components are


(3) x = y = 2c

We now q introduce another frame moving in the x direction with ordi-


nary speed 25 c. The ordinary velocity of the object transforms using the
velocity addition formulas
1
FOUR-VELOCITY VERSUS ORDINARY VELOCITY: AN EXAMPLE 2

x v
(4) x =
1 xv/c2
y
(5) y =
(1 xv/c2 )
q q
so we get, using = 1/ 1 5 = 53 :
2

(6) ux = 0
p
2/5
(7) uy = p c
5/3 (1 2/5)
r
2
(8) = c
3
As i is a four-vector, it transforms using Lorentz transformations, for
which we need

c
(9) 0 = p = 5c
1 u2 /c2

so

x = x 0

(10)
!
5 2
r r
(11) = 2c 5c = 0
3 5

(12) y = y = 2c
We can check this using 2:

ux
(13) x = p
1 u2 /c2
(14) = 0
uy
(15) y = p
1 u2 /c2
p
2/3c
(16) = p
1/3

(17) = 2c
FOUR-VELOCITY VERSUS ORDINARY VELOCITY: AN EXAMPLE 3

so it checks out.
FOUR-VELOCITYS SQUARE IS A UNIVERSAL CONSTANT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 26.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.
The four velocity is defined as

ui
(1) i = p
1 u2 /c2
so the square is
2
c2 3i=1 ui
(2) i i = +
1 u2 /c2 1 u2 /c2
c2 + u2
(3) =
1 u2 /c2
(4) = c2
This is the equivalent of the result we derived earlier in Moores book,
except with c = 1 so i i = 1. It is a universal invariant for all four-
velocities.
In terms of four-momentum, this relation becomes

(5) pi pi = m2 c2
2
(6) p0 + p2 = m2 c2
2
(7) cp0 + c2 p2 = m2 c4
(8) E 2 p2 c2 = m2 c4

P INGBACKS
Pingback: Decay of a pion into a muon and a neutrino
1
FOUR-VELOCITYS SQUARE IS A UNIVERSAL CONSTANT 2

Pingback: Decay of a pion into two photons


Pingback: Relative energy in a particle collider
Pingback: Pair annihilation with an electron and a positron
FOUR-VELOCITY OF A PARTICLE IN HYPERBOLIC MOTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 27.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.]
We can now return to a particle travelling on a hyperbolic trajectory, so
its position (one dimensional, on the x axis) is

p
(1) x (t) = b2 + c2t 2

Here, x and t are the position and time as measured by an observer at


rest (so they are not proper time for the particle). We can find the particles
proper time as a function of t by using the relation between time intervals:

q
(2) d = 1 u2 /c2 dt

We have

(3) u = x
c2t
(4) =
b2 + c2t 2
s
c2t 2
(5) d = 1 2 dt
b + c2t 2
b
(6) = dt
b2 + c2t 2

Integrating both sides, we get, taking = 0 when t = 0 (the integral can


be done by software or looked up as it is a standard integral):
1
FOUR-VELOCITY OF A PARTICLE IN HYPERBOLIC MOTION 2

t
1
(7) = b q dt 0
0 b2 + c2 (t 0 )2
  
b 1 p
2 2 2
(8) = ln ct + b + c t
c b
We can write the position as a function of by starting with 1 and using
this last result:
 
b 1
(9) = ln (ct + x)
c b
p
(10) ct = x 2 b2
 p 
b 1 2 2
(11) = ln x b +x
c b
p
(12) bec/b = x 2 b2 + x
 2
(13) be c/b
x = x 2 b2
b  c/b 
(14) x = e + ec/b
2
c
(15) = b cosh
b
For the ordinary velocity, we have

c2t
(16) u =
b2 + c2t 2

x2 b2
(17) = c
qx
cosh2 cb 1
(18) = cb
b cosh cb
c
= c tanh
b
The four velocity is defined as

ui
(19) i = p
1 u2 /c2

so we have (Im assuming that in part (c) of Griffithss problem, he meant to


FOUR-VELOCITY OF A PARTICLE IN HYPERBOLIC MOTION 3

ask for i in terms of , not t, as the latter doesnt give anything particularly
informative):

c
(20) 0 = q
1 tanh2 cb
c
(21) = c cosh
b
u
(22) x = q
1 tanh2 cb
c tanh cb
(23) = q
1 tanh2 cb
c
(24) = c sinh
b
As a check, we note that

2 c 2 c
 
2
(25) i
i = c cosh sinh = c2
b b
RELATIVISTIC MOMENTUM AND ENERGY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 28.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.]
One way of defining linear momentum of an object of mass m in relativity
is to multiply the mass by the velocity, but since we have an ordinary
velocity and a four-velocity, we need to choose which one to use. It turns
out that if we want momentum to be conserved, we need to use the four-
velocity. We can see this as follows.
If we define momentum using ordinary velocity, so that

(1) p = mu

we can rework our collision problem from earlier. In an inertial frame S we


have a collision between a particle A (mass mA and velocity uA ) and another
particle B (mass mB and velocity uB ). During the collision, some of the mass
of A gets transferred to B, so that afterwards we have particles C and D with
masses mC , mD and velocities uC , uD . If momentum is conserved, then

(2) mA uA + mB uB = mC uC + mD uD
In inertial frame S which moves with velocity v relative to S , the ve-
locities of the particles transform using the formulas

x v
(3) x =
1 xv/c2
y
(4) y =
(1 xv/c2 )
z
(5) z =
(1 xv/c2 )
1
RELATIVISTIC MOMENTUM AND ENERGY 2

The LHS of 2 transforms as

(6)    
mA uAy uAz mB uBy uBz
(uAx v) x + y + z + (uBx v) x + y + z
1 uAx v/c2 1 uBx v/c2
We cant use 2 to convert this into the transform of the RHS (where A is
replaced by C and B by D) because of the different factors of 1u 1 v/c2 and
Ax
1
1uBx v/c2
multiplying each term, that is, for the x component for example,
we cant factor out a term mA uAx +mB uBx from this expression to set it equal
to mC uCx + mD uDx .
If we use the four-velocity to define momentum, however, things work
out properly. We define

(7) p = m

(remember that Griffiths uses for four-velocity). To make this a four-


vector, we define

mc
(8) p0 = m 0 = p
1 u2 /c2
Now we start with three-momentum (using the spatial components of the
four-momentum) conserved in S :

(9) mA A + mB B = mC C + mD D

To convert to S we can use Lorentz transformations, since the masses


are scalars (constants) and the s are four-vectors. The LHS becomes

mA A1 A0 x + A2 y + A3 z + mB B1 B0 x + B2 y + B3 z =
     

(10)
x mA A1 + mB B1 mA A0 + mB B0 + y mA A2 + mB B2 + z mA A3 + mB B3
   

If we now apply 9 to each spatial component separately, we see that this


last line is equal to

(11)
x mC C1 + mD D1 mA A0 + mB B0 + y mC C2 + mD D2 + z mC C3 + mD D3
   
RELATIVISTIC MOMENTUM AND ENERGY 3

In order for this last expression to be equal to the Lorentz transform of


the RHS of 9, the one remaining term containing A and B terms must equal
its corresponding term in the transform of the RHS. That is, we must have

(12) mA A0 + mB B0 = mC C0 + mD D0
This is used as the motivation to define the relativistic energy as

mc2
(13) E = cp0 = mc 0 = p
1 u2 /c2
With these definitions, we can see that four-momentum (which is 3-
momentum and energy together) are conserved.
P INGBACKS
Pingback: Four-velocitys square is a universal constant
Pingback: Relativistic kinetic energy
Pingback: Decay of a pion into a muon and a neutrino
Pingback: Collision of two identical particles
Pingback: Decay of a pion into two photons
Pingback: Pair annihilation with an electron and a positron
Pingback: Force in relativity
Pingback: Collision of a pion and a proton
Pingback: Relativistic energy revisited
RELATIVISTIC KINETIC ENERGY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 29.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.]
We can use the definition of relativistic energy to further define kinetic
energy. The relativistic total energy is

mc2
(1) E= p
1 u2 /c2

where m is the rest mass and u is the objects velocity in some inertial frame.
If u = 0, then we get the rest energy (the most famous formula in physics):

(2) E = mc2
Since all the rest of the energy is due to the objects motion, it is natural
to define this excess energy as the kinetic energy:
!
0
 
2 2 1 2
(3) K E mc = mc p 1 = mc 1
1 u2 /c2 c

where is the four-velocity.


Using this definition, we can show that if a collision is elastic in one
frame S it is also elastic in another frame S. Using the same example we
treated earlier, since K is conserved in S , we have

(4)  0
A0
   0    0 
2 2 B 2 C 2 D
mA c 1 + mB c 1 = mC c 1 + mD c 1
c c c c
Since total energy is conserved, we have
1
RELATIVISTIC KINETIC ENERGY 2

(5) mA A0 + mB B0 = mC C0 + mD D0

so cancelling these terms in 4, we find that rest mass is also conserved:

(6) mA + mB = mC + mD
Using a Lorentz transformation on the LHS of 4, we have in frame S:
   
A0 A1 1 + mB B0 B1 1
 
(7) mA
c c
Using conservation of energy 5 and rest mass 6, along with conservation
of the x component of momentum, that is:

(8) mA A1 + mB B1 = mC C1 + mD D1

we get
   
A0 A1 1 + mB B0 B1 1 = . . .
 
(9) mA
c c

mA A0 + mB B0 mA A1 + mB B1 (mA + mB ) = . . .
 
(10)
c c

mC C0 + mD D0 mC C1 + mD D1 (mC + mD )
 
(11)
c c
Thus the collision is also elastic in S.
Example. If a particles kinetic energy is n times its rest energy, then from
3

1
(12) p 1 = n
1 u2 /c2
s
1
(13) u = c 1
(n + 1)2
Thus if n = 0 (no kinetic energy), then u = 0 (the particle is at rest), and
as n , u c so the larger the kinetic energy, the closer the particles
speed gets to c.
CENTRE OF MOMENTUM FRAME

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 30.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.]
In classical mechanics, its often easier to analyze the motions of objects
in their centre of mass frame. In relativity, its more accurate to refer to the
centre of momentum frame, which is the frame in which the sums of the
spatial components of the momentum of all objects are zero. The problem
is to find this frame given the energies and momenta of a set of objects.
In some inertial frame, the total energy and momentum of such a set are
(assuming all motion is along the x axis):

(1) Etot = Ei = p0i c


(2) Ptot = pi = p1i
Using Lorentz transformations, we want a frame moving with velocity v
relative to this frame such that

(3) Ptot = 0

Therefore

Ptot = p1i p0i = 0



(4)
p1i
(5) =
p0i
pi
(6) v = c2
Ei
1
CENTRE OF MOMENTUM FRAME 2

P INGBACKS
Pingback: Pair annihilation with an electron and a positron
Pingback: Collision of a pion and a proton
Pingback: Elastic collision of two identical particles
DECAY OF A PION INTO A MUON AND A NEUTRINO

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 31.
[Griffithss approach to the relativistic four-velocity is similar to that of
Moore, although rather confusingly, he uses different notation (as well as
keeping factors of c in the equations rather than setting c = 1). To keep
the notation consistent with Griffiths, Ill use his notation here, but anyone
attempting to follow both books should beware.]
We can use the conservation of relativistic energy and momentum to ana-
lyze the interaction of elementary particles. For example, a pion at rest can
decay into a muon and a neutrino. Conservation of energy and 3-momentum
require

(1) E = m c2 = E + E
(2) p = 0 = p + p
We can use the relation

(3) E 2 p2 c2 = m2 c4

to relate energy and momentum. Assuming the neutrino is massless (it isnt
quite, but its close) we have

(4) E = cp

while for the muon


q
(5) E = c p2 + m2 c2

so
q
(6) m c = p2 + m2 c2 + p
1
DECAY OF A PION INTO A MUON AND A NEUTRINO 2

But p = p from 2 so
q
(7) m c = p2 + m2 c2 p
q
(8) p2 + m2 c2 = m c + p
m2 m2
(9) p = c
2m
m2 + m2 2
(10) E = c
2m

where the last line follows from 5.


The velocity of the muon can be found from

(11) E = p0 c
m c2
(12) = p
1 u2 /c2
m2 + m2 2 m c2
(13) c = p
2m 1 u2 /c2
v
u 4m2 m2
u = c 1
u
(14) t  
2
m + m 2

m2 m2
(15) = c
m2 + m2
P INGBACKS
Pingback: Collision of a pion and a proton
Pingback: Elastic collision of two identical particles
COLLISION OF TWO IDENTICAL PARTICLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 32.
Heres another example of using the conservation of relativistic energy
and momentum to analyze the interaction of elementary particles. A particle
of mass m with a total energy equal to twice its rest energy (as measured in
the lab frame) collides with an identical particle at rest, after which the two
particles stick together to form a single particle of rest mass M.
Working in the lab frame, we have from conservation of energy:

(1) Etot = 2mc2 + mc2 = M Mc2

where

1
(2) M = q
1 u2M /c2

From conservation of momentum:

(3) 2mum + 0 = M MuM


Since

1
(4) p =2
1 u2m /c2

we have


3
(5) um = c
2

so we can solve 1 and 3 to get


1
COLLISION OF TWO IDENTICAL PARTICLES 2


3
(6) uM = c
3
(7) M = 6m
Note that the rest mass of the new particle is greater than the sum of the
rest masses of the two original particles, indicating that some of the kinetic
energy has been converted to rest mass (or rest energy).
DECAY OF A PION INTO TWO PHOTONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 33.
Heres another example of using the conservation of relativistic energy
and momentum, this time in the case of a pion with momentum p = 34 mc
decaying into two photons. One way of doing this problem is to start in the
rest frame of the pion. In this frame, the two photons have equal energies
and equal and opposite momenta, so

1
(1) E = mc2
2
The momenta of the photons in the pions rest frame are found from

(2) E = |p | c
1
(3) p = mc
2
To transfer to the lab frame, we need to find the velocity of the pion:

mu 3
(4) p mu = mc
1 u2 /c2 4
3
(5) u = c
5
5
(6) =
4
Then the energies of the two photons can be found from a Lorentz trans-
formation:

(7) Elab = (E + p )
  
5 1 2 3 1 2
(8) = mc mc
4 2 5 2
1 2
(9) = mc , mc2
4
1
DECAY OF A PION INTO TWO PHOTONS 2

In the pions frame, the two photons have equal energies and therefore
equal wavelengths (from Plancks formula E = h), while in the lab frame
one photons energy is half that of its energy in the pions frame and the
other is twice the energy, so one photon is red-shifted and the other is blue-
shifted.
RELATIVE ENERGY IN A PARTICLE COLLIDER

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 34.
A technique used in particle accelerators such as the Large Hadron Col-
lider to increase the relative energy of the incident particle on its target is
to direct two opposing beams of particles at each other so that they collide
head on.
In classical physics, the kinetic energy of a particle is K = 21 mv2 so if a
particle with this kinetic energy is fired at another identical particle at rest,
this much energy is delivered into the collision. However, if we aim two
opposing beams of particles, each with kinetic energy K at each other, the
relative energy is increased by a factor of 4. This is because, classically, if
we transform to the rest frame of one of the particles, the velocity of the
other particle relative to the stationary one is 2v, so K increases by a factor
of 22 = 4.
In relativity, however, we can get a much larger gain in relative energy.
Suppose we fire two beams at each other so that, in the lab frame, each
particle in both beams has a total energy E. If we now transform to the rest
frame of one of the particles, the relative energy is

(1) E = (E + cp)

where p is the particles lab momentum, which we can get from the formula

(2) c2 p2 = E 2 m2 c4

Also, since

1 E
(3) =p = 2
1 2 mc

we have
1
RELATIVE ENERGY IN A PARTICLE COLLIDER 2

1 m2 c4 E 2 m2 c4
(4) 2 = 1 2 = 1 2 =
E E2
s
2
(E 2 m2 c4 ) E 2 m2 c4
(5) cp = =
E2 E
Putting it all together, we get

E 2 m2 c4
 
E
(6) E = E+
mc2 E
2E 2
(7) = 2
mc2
mc
That is, the relative energy increases as the square of the lab energy,
which can lead to enormous gains. For example, if we fire 2 opposing
beams of protons at each other at a lab energy of 30 GeV (the protons rest
mass is roughly 1 GeV), we find the relative energy is

2 302
(8) E = 1 = 1799 GeV
1
This is about 60 times the lab energy.
PAIR ANNIHILATION WITH AN ELECTRON AND A
POSITRON

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Post 35.
Heres a slightly more involved example of conservation of energy and
momentum in relativity. Suppose an electron with momentum pe hits a
positron (anti-electron) and they annihilate each other (a process known as
pair annihilation), producing two photons.
In passing, its worth noting that pair annihilation must produce at least
2 photons. If we looked at the problem in the centre of momentum frame,
the total momentum of the system must be zero both before and after the
collision. If only one photon were produced, it is impossible for it to have
zero momentum.
Before the collision, we have

q
(1) Etot = mc2 + mc2 + p2e c2

After the collision, suppose that one photon emerges at 60 = 3 radians


to the incident electrons direction. From conservation of momentum, we
have


(2) p1 sin = p2 sin
3

(3) p1 cos + p2 cos = pe
3

where is the angle atwhich photon 2 emerges. We can eliminate as


follows (using sin 3 = 23 , cos 3 = 12 ):

 2
3 2 1
p22 2 2

(4) sin + cos = p + pe p1
4 1 2
(5) p22 = p21 p1 pe + p2e
For a photon, E = cp, so the total energy after the collision is
1
PAIR ANNIHILATION WITH AN ELECTRON AND A POSITRON 2

(6) Etot = cp1 + cp2


From conservation of energy and 1, we get

(7)

c2 p22 = (Etot cp1 )2


 q 2
2 2 4 2 2
(8) = mc + m c + pe c cp1
 q 2  q 
2
(9) 2 4 2
= mc + m c + pe c2 2 mc + m c + pe c cp1 + c2 p21
2 2 4 2 2

We can now use 5:

(10)

c2 p22 = c2 p21 c2 p1 pe + c2 p2e


(11)
 q 2  q 
= mc + m2 c4 + p2e c2 2 mc + m2 c4 + p2e c2 cp1 + c2 p21
2 2

Solving for cp1 we get


 p 2
mc2 + m2 c4 + p2e c2 c2 p2e
(12) cp1 =  p 
2 mc2 + m2 c4 + p2e c2 cpe
p
2m2 c4 + 2mc2 m2 c4 + p2e c2
(13) =  p 
2 mc2 + m2 c4 + p2e c2 cpe
mc2
(14) = h  p i
1 cpe / 2 mc2 + m2 c4 + p2e c2
Thus photon 1 has an energy greater than the rest mass of the electron.
Note that if the pair annihilation occurs with both the electron and positron
at rest (pe = 0), then cp1 = cp2 = mc2 as wed expect.
FORCE IN RELATIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.36.
Force can be defined in relativity as the derivative of the spatial compo-
nents of four-momentum with respect to ordinary (non-proper) time:

dp
(1) F=
dt
Superficially, this looks the same as Newtons second law, but in fact the
formula for force is a bit more complex when written out in full. Using the
definition of momentum, we have

d
(2) F = [mu]
dt

where

1
(3) p
1 u2 /c2

We have

d 1 d (u u)
(4) = 3/2
dt 2c2 (1 u2 /c2 ) dt
3
(5) = (2u a)
2c2
ua
(6) = 3/2
c2 (1 u2 /c2 )

where a u is the acceleration.


Returning to 2, we have
1
FORCE IN RELATIVITY 2

d
(7) F = mu + mu
dt
m (u a) u ma
(8) = 3/2
+p
c2 (1 u2 /c2 ) 1 u2 /c2
 
m (u a) u
(9) = p a+ 2
1 u2 /c2 c u2
This formula reduces to the familiar F = ma in the limit of small u. How-
ever, if we wish to retain a fixed acceleration as u c, the required force
becomes infinite. Or looked at another way, if we want the force to remain
finite as u c, the acceleration must drop to zero. In other words, its im-
possible to accelerate an object with a non-zero rest mass to the speed of
light.
P INGBACKS
Pingback: Outrunning a light ray
Pingback: Four-acceleration and Minkowski force
Pingback: Minkowski force
Pingback: Relativistic electromagnetic force
Pingback: Relativistic acceleration in terms of force
OUTRUNNING A LIGHT RAY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.37.
Even though nothing can move faster than light, it is possible for an ob-
ject to arrive at any given location before a light ray, provided the object
gets a bit of a head start. Suppose we have an object that is subject to a con-
stant force in the +x direction. Weve seen that (ordinary) force in relativity
is the derivative of the spatial parts of the four-momentum with respect to
ordinary time. In one dimension for a constant force we therefore have

dp
(1) = F
dt
(2) p = Ft +C

where C is a constant of integration. If the object starts at t = 0 at rest (in


the lab frame), then C = 0, and

mu
(3) p= p = Ft
1 u2 /c2

which can be solved for the velocity u to give

F t
(4) u= q
m
1 + (Ft/mc)2
This can be integrated again to get the position (assuming x = 0 at t = 0):


F t
t 0 dt 0
(5) x (t) = q
m 0 1 + (Ft 0 /mc)2
mc2
q 
2
(6) = 1 + (Ft/mc) 1
F
We can rearrange this to get
1
OUTRUNNING A LIGHT RAY 2

 2  2
F (ct) Fx
(7) +1 =1
mc2 mc2

which is the equation of a hyperbola in the coordinates ct and x. The asymp-


totes are found by setting the RHS to zero, so we get
 
F (ct) Fx
(8) = +1
mc2 mc2
mc2
(9) ct = x
F
For the case of motion in the +x direction, we take the plus sign, so the
asymptote intersects the ct axis at ct = mc2 /F. We can plot this (for the
case where F/mc2 = 1 in inverse distance units) on a spacetime diagram to
get the red curve shown:

The green line is the asymptote, but it is also the world line of a light ray
that leaves x = 0 at ct = 1. Since it is the asymptote of the objects world
line, the object will reach any given value of x before the light ray, so if
an object is subjected to a constant force and given a head start (it starts
moving at ct = 0 and the light ray starts at ct = mc2 /F) it will always be
OUTRUNNING A LIGHT RAY 3

ahead of the light ray (although admittedly not by much for large x). This
is true no matter how small the force, although the smaller the force, the
larger the head start youll need to stay ahead of the light ray.
P INGBACKS
Pingback: Acceleration under a constant force
FOUR-ACCELERATION AND MINKOWSKI FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.38.
We can define a four-acceleration as the derivative of four-velocity with
respect to proper time:

d i
i
(1)
d

where

c
(2) 0 = p
1 u2 /c2
u
(3) = p
1 u2 /c2
Looking first at 0 , we have

d 0
(4) 0 =
d !
d 1
(5) = c p
d 1 u2 /c2
!
dt d 1
(6) = c p
d dt 1 u2 /c2
!
c d 1
(7) = p p
1 u2 /c2 dt 1 u2 /c2
The derivative in the last line we worked out earlier:
!
d 1 ua
(8) p = 3/2
dt 1 u2 /c2 c2 (1 u2 /c2 )

so
1
FOUR-ACCELERATION AND MINKOWSKI FORCE 2

ua
(9) 0 = 2
c (1 u2 /c2 )
The spatial components work out to

1 dp
(10) =
m d
1 dp
(11) = p
m 1 u /c dt
2 2

1
(12) = p F
m 1 u2 /c2
We alsoworked out the force earlier:
 
m (u a) u
(13) F= p a+ 2
1 u2 /c2 c u2

so
 
1 (u a) u
(14) = a+ 2
1 u2 /c2 c u2
p
The invariant square of i is (using 1/ 1 u2 /c2 )
" #
2 2 (u a)2 4 u2 (u a)2
(u a) 2
(15) i i = 8 + 4 a2 + +
c2 c2 c4
8 2 u2
 
4 2 2
(16) = a + 2 (u a) 1 + 2 + 2
c c
8 u2 u2
 
4 2 2
(17) = a + 2 (u a) 1 2 2 + 2
c c c
6
(18) = 4 a2 + (u a)2
c2 " #
1 2 (u a)2
(19) = 2
a + 2
(1 u2 /c2 ) c u2

This reduces to a2 to first order in u.


The scalar product of four-acceleration and four-velocity is, using 3:
FOUR-ACCELERATION AND MINKOWSKI FORCE 3

u2
(20) i i = 5 (u a) + 3 (u a) + 5 (u a) 2
c
 
u 2 
(21) = (u a) 3 1 2 1 2
c
(22) = 0
Finally, we can use the four-acceleration to write the Minkowski force,
which is the derivative of the four-momentum with respect to the proper
time:

dp
(23) K
d
(24) = m
The Minkowski force, used with four-acceleration, looks just like the
non-relativistic form of Newtons second law. If we complete the Minkowski
force by including its 0 component, we have

(25) K 0 = m 0

from which the invariant product with four-velocity follows:

(26) K i i = m i i = 0
P INGBACKS
Pingback: Electromagnetic Minkowski force
Pingback: Motion under a constant Minkowski force
MINKOWSKI FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.39.
The Minkowski force is defined as the derivative of four-momentum with
respect to proper time:

d pi
(1) Ki =
d
Its 0 component is therefore

0 d p0 1 dE dE
(2) K = = =
d c d c dt
0
We can relate K to the ordinary force as follows. We have

dp
(3) Fu = u
dt !
d u
(4) = m p u
dt 1 u2 /c2
Using
!
d 1 ua
(5) p = 3/2
dt 1 u2 /c2 c2 (1 u2 /c2 )
we get

mu a
(6) Fu = 3/2
(1 u2 /c2 )
!
d mc2
(7) = p
dt 1 u2 /c2
dE
(8) =
dt
So
1
MINKOWSKI FORCE 2


(9) K0 = F u
c
i
For the spatial part of K we have

dp dp
(10) KK =
d d
dp dp
(11) = 2
dt dt
2 2
(12) = F
Therefore
2
(13) Ki K i = K 0 + K K
 
2 1 2 2
(14) = 2 (F u) + F
c
u2 cos2
 
2
(15) = 1 F2
c2

where is the angle between u and F.


RELATIVISTIC ELECTROMAGNETIC FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.40.
Ordinary force in relativity is given by

 
m (u a) u
(1) F= p a+ 2
1 u2 /c2 c u2

Suppose we have a particle of charge q travelling with velocity u in a


region with electric and magnetic fields, so that the force is given by

(2) F = qE + qu B

What is the acceleration of the particle? From 1 we have

p
1 u2 /c2 (u a) u
(3) a = F 2
m c u2
p
q 1 u2 /c2 (u a) u
(4) = (E + u B) 2
m c u2
q (u a) u
(5) = (E + u B) 2
m c2
The trick is to disentangle the a from the u a term on the RHS. We can
do this by taking the dot product of this equation with u to get

p
1 u2 /c2
q (u a) u2
(6) ua = (E + u B) u 2
m c u2
p
q 1 u2 /c2 (u a) u2
(7) = Eu 2
m c u2
q (u a) u2
(8) = E u 2
m c2
We can now solve for u a:
1
RELATIVISTIC ELECTROMAGNETIC FORCE 2

1
u2 2

q
(9) ua = Eu 1+ 2
m c
q
(10) = 3 Eu
m
Substituting back into 5 we get
 
q 1
(11) a = E + u B 2 (E u) u
m c
 
q 1
q
(12) = 2 2
1 u /c E + u B 2 (E u) u
m c
Note that as u c, a 0 no matter how strong the fields are, so again
were reminded that we cant accelerate any massive object up to the speed
of light.
RELATIVISTIC TRANSFORMATION OF FORCE, ELECTRIC
AND MAGNETIC FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problems 12.41, 12.44.
Griffiths gives fairly complete derivations of the relativistic transforma-
tion laws for force and electromagnetic fields as we move from one inertial
frame to another in his sections 12.2.4 and 12.3.2, so I wont grind through
the whole thing again here. Ill just recap the ideas behind these derivations
to give a flavour of whats being done.
Ordinary force is defined as the derivative of the four-momentum (or at
least its spatial part) with respect to ordinary (not proper) time. Since four-
momentum is a four-vector, it transforms using Lorentz transformations, as
does ordinary time. The problem with force is that its transformation is
then the ratio of two Lorentz-transformed objects, so the transformation
equations get a bit messy. The derivation of the transformations is done in
much the same way as the derivation of the velocity addition formulas. For
a frame S moving at velocity v in the x direction relative to a stationary
frame S we get

d py
(1) Fy =
dt
d py
(2) =  
dt c dx
d py /dt
(3) =
(1 ux /c)
Fy
(4) =
(1 ux /c)

where u is the objects velocity in S . Remember that the velocity v used in


the calculation of and is the relative velocity of S and S and u is the
velocity of the object relative to S .
The calculation of the other two components is done in a similar way,
with the result
1
RELATIVISTIC TRANSFORMATION OF FORCE, ELECTRIC AND MAGNETIC FIELDS 2

Fx (u F) /c
(5) Fx =
1 ux /c
Fy
(6) Fy =
(1 ux /c)
Fz
(7) Fz =
(1 ux /c)
Example 1. We have a charge qA at rest at the origin of S and another
charge qB moving at speed v in the +x direction along the line y = d. When
qB crosses the y axis, it feels only an electric field (since qA is at rest it
generates no magnetic field in S ), so the force on it is

qA qB
(8) F= y
40 d 2
If we now switch to qB s frame S which is moving relative to S at
speed v along the x axis, we can find the force experienced by qB in this
frame using 6. In this case, u = v (the objects velocity in S is the same as
the relative velocity of S and S) so

Fy
(9) Fy =
(1 v/c)
Fy
(10) =
(1 v2 /c2 )
(11) = Fy
qA qB
(12) = y
40 d 2
That is, the force experienced by qB is greater in the frame at which it is
at rest.
In fact we can see from the transformation equations above that all com-
ponents of force perpendicular to the motion are at a maximum in an ob-
jects rest frame. In that frame u = 0 so the equations become

(13) Fx = Fx
Fy
(14) Fy =

Fz
(15) Fz =

RELATIVISTIC TRANSFORMATION OF FORCE, ELECTRIC AND MAGNETIC FIELDS 3

To transform the electric and magnetic fields we can use a similar ap-
proach to that for the electric field, in which we considered a parallel plate
capacitor with charge densities of 0 (in the capacitors rest frame S0 )
on the two plates. In S0 however, since the charge is at rest, there is no
magnetic field so we cant use that system as it stands to derive the general
transformation rules for electromagnetic fields. What we can do is introduce
two other frames S (moving at speed v0 relative to S0 ) and S (moving at
speed v relative to S ). From the velocity addition formula, the velocity of
S relative to S0 is then

v + v0
(16) v =
1 + vv0 /c2
Since we now have the velocities of both S and S relative to S0 (where
the charge is at rest, remember), we can eliminate S0 and express the fields
in S in terms of those in S . Griffiths goes through the details, with the
results

(17) Ex = Ex
(18) Ey = (Ey vBz )
(19) Ez = (Ez + vBy )
(20) Bx = Bx
 v 
(21) By = By + 2 Ez
c 
 v
(22) Bz = Bz 2 Ey
c
Notice how the electric and magnetic field components perpendicular to
the motion get tangled up with each other when we transform frames. This
shows how its possible for a test charge to experience only an electric field
in one frame, but a combination of electric and magnetic fields in another
frame (and vice versa).
Example 2. Now that we have the general transformation rules, we can
return to the special case of frames S in which the capacitor plates are at
rest and S where they are moving in the x direction at speed v. When the
plates are at rest, there is only an electric field (in the y direction, assuming
that the plates are parallel to the xz plane). From these equations, we see
that in any other frame moving in the x direction, the electric field will have
only a y component and the magnetic field will have only a z component.
[Im not certain this is the approach Griffiths wants in his problem 12.41 in
which he asks why Ez = 0 in the moving frame, but it does appear to answer
the question.]
RELATIVISTIC TRANSFORMATION OF FORCE, ELECTRIC AND MAGNETIC FIELDS 4

Example 3. We can also revisit Example 1 and calculate the force felt by
qB in its own rest frame by applying the field transformations. When qB
crosses the y axis (which well assume coincides with the y axis at the time
when qB crosses it) we had, from above,

qA
(23) Ey =
40 d 2

with Ex = Ez = 0, and B = 0. In qB s frame, we have

qA
(24) Ey = Ey =
40 d 2
v qA v
(25) Bz = 2
Ey =
c 40 d 2 c2

with all other components equal to zero. The force felt by qB is again en-
tirely electric (since its not moving in its own rest frame, it experiences no
magnetic force), and we get

qA qB
(26) Fy = qB Ey =
40 d 2

which is the same as 12.


P INGBACKS
Pingback: Relativistic transformation of electric and magnetic fields: an
example
Pingback: Relativistic invariants involving electromagnetic fields
Pingback: Relativistic transformation of electromagnetic waves; the Doppler
effect
TRANSFORMING THE ELECTRIC FIELD IN RELATIVITY

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.42.
Applying relativity to electromagnetism results in some rather curious
behaviour. For example, suppose we have a wire at rest in the lab frame.
The wire contains a linear charge density of positive charge and a density
of negative charge. Since the densities are equal and opposite, the wire
is electrically neutral. Now suppose we start the positive charge moving
with speed v in the +x direction, and the negative charge moving with equal
and opposite speed in the x direction, resulting in a total net current of
I = 2 v in the +x direction. This produces a magnetic field (using Ampres
law) at a distance r from the wire of

0 I 0 v
(1) B= =
2r r
If we now place a charge +q a distance r from the wire moving a velocity
u in the +x direction (parallel to the wire), it will experience no electric
force (since the wire is neutral) but it will experience a magnetic force of

0 v
(2) F = qu B = qu r
r
That is, it experiences a force towards the wire.
Now suppose we switch to a frame moving with velocity u in the +x di-
rection, in which q is at rest. Since the charge is at rest, it experiences no
magnetic force and, since the wire is neutral, we might think it experiences
no electric force either. This conclusion violates the first principle of rela-
tivity, namely that the laws of physics should appear the same in all inertial
frames. In the first frame (with moving q), the charge experiences a force
towards the wire, while in the second frame (charge at rest) it experiences
no force at all. What went wrong?
To solve this problem, we can apply relativity to the transformation be-
tween frames. In the first frame, with the wire at rest, the conclusions about
the electrical neutrality and currents are valid since were not moving. Al-
though the two lines of charge are both moving and are therefore Lorentz-
contracted, their speeds are equal so they are both contracted by the same
1
TRANSFORMING THE ELECTRIC FIELD IN RELATIVITY 2

factor. However, when we transform to the second frame, the velocities of


the two anti-parallel currents transform in different ways due to the velocity
addition formula. The velocities of the two lines of charge in this second
frame are (the subscript refers to the sign of the charge):

vu
(3) v =
1 uv/c2
Because the positive charge is now seen as moving more slowly, it will
not be Lorentz-contracted as much as the negative charge, so that in this
frame, + < . In other words, as viewed in the rest frame of the charge
q, the wire has a net negative charge so q feels an electric (not magnetic)
force of attraction to the wire. In his section 12.3.1, Griffiths goes through
the algebra to show that this electric force is equal to the magnetic force
felt in the first frame (after appropriately transforming the force between
the two frames).
We can use a similar argument to work out how the electric field trans-
forms between frames. This time, we start out with a large parallel plate
capacitor with its plates parallel to the xz plane and separated by a distance
d. If the upper plate has surface charge density +0 and the lower plate has
density 0 , then the electric field between the plates is, in the frame where
the capacitor is at rest:

0
(4) E0 = y
0
Now suppose we move to a frame moving at velocity v in the +x di-
rection. The capacitor plates are now contracted by a factor 1/ in the x
direction, but their size is unchanged in the z direction since this direction
is perpendicular to the motion. Therefore, the area of the plates is reduced
by a factor 1/ so the charge density increases by a factor resulting in an
electric field

0
(5) E= y = E0
0
Clearly the same logic applies if we oriented the plates parallel to the xy
plane (the electric field would then be in the z direction and would increase
by the same factor ), so in general we can say that

(6) E = E
0
That is, the electric field perpendicular to the motion increases by a factor
.
TRANSFORMING THE ELECTRIC FIELD IN RELATIVITY 3

To get the behaviour of the field parallel to the motion, we can orient
the plates parallel to the yz plane. This time the plates are perpendicular to
the motion so their size does not change; however the distance d between
the plates is contracted. Since the electric field doesnt depend on d (were
assuming that the plates are very large compared to their separation so we
can ignore edge effects) the parallel components of E are unchanged:

k
(7) Ek = E0
Notice that weve implicitly assumed that the charge is invariant under
a Lorentz transformation. The theoretical reasons for this appear to be
rather deep, although it seems that if we require Maxwells equations to
be Lorentz-invariant, then charge must also be Lorentz-invariant. In any
case, its a bit more than I want to get into here.
Example. Suppose we now tilt the capacitor plates so that they make an
angle of 4 with the x axis. A vector lying in the x y plane and in the plane
of the plates is x + y, so that the normal vector to the plates is

(8) n0 = x + y
In the plates rest frame, the field is

E0
(9) E0 = (x + y)
2
If we now shift to a frame moving at velocity v, only the y component of
E0 is affected, so the new field is

E0
(10) E = (x + y)
2
However, the motion also contracts the x dimension of the plates by a
factor , so a vector lying in the x y plane and in the plane of the plates is
now 1 x + y
making the new normal vector to the plates

(11) n = x + y
Therefore, in the moving frame, the angle between E and n is given by

nE 2
(12) cos = =
|n| |E| 1 + 2
TRANSFORMING THE ELECTRIC FIELD IN RELATIVITY 4

For > 1, cos 6= 1 so for the moving frame, 6= 0 and the field is not
perpendicular to the plates.
P INGBACKS
Pingback: Relativistic transformation of force, electric and magnetic fields
GAUSSS LAW FOR A RELATIVISTIC POINT CHARGE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.43.
In Griffithss example 12.13, he rederives the formula for the electric
field due to a moving charge, this time using relativity instead of retarded
potentials. The result, which weve examined before, is

q 1 v2 /c2 R
(1) E (r,t) =
40 1 v2 sin2 /c2 3/2 R2
q 1 R
(2) = 2
40 1 v2 sin2 /c2 3/2 R2
q 1 R
(3) = 2
40 1 2 sin2 3/2 R2

where

(4) R r vt

is the vector from the particles present (not retarded) position to the ob-
server (assuming the particle passes through the origin at t = 0) and is
the angle between R and v. We can verify that Gausss law holds for this
moving charge by integrating E da over a sphere of radius R.


q sin d
(5) E da =
20 2 0 1 v2 sin2 /c2
3/2

The integral is nasty because were missing a cos in the numerator that
would make the integral easy. Maple handles it easily enough, but for those
interested in how to derive it, I worked backward from Maples answer to
figure out how to do it. First, we can split the integrand into the sum of two
terms (Ill use v/c to simplify the notation):
1
GAUSSS LAW FOR A RELATIVISTIC POINT CHARGE 2

(6)
sin sin
3/2 = 3/2
1 2 sin2 (1 2 + 2 cos2 )
sin 1 2 + 2 cos2 2 cos2

1
(7) =
(1 2 ) (1 2 + 2 cos2 )
3/2

1 sin 2 cos2 sin


(8) =
(1 2 ) 1 2 + 2 cos2 (1 2 ) (1 2 + 2 cos2 )3/2
p

We can now integrate the second term by parts:

(9)
" #
2 cos2 sin 2 cos sin
d = (cos ) d
(1 2 ) (1 2 + 2 cos2 )
3/2 (1 2 ) (1 2 + 2 cos2 )
3/2

(10)
cos 1 sin
=
(1 2 ) 1 2 + 2 cos2 (1 2 )
p p
12 +
(11)
1 cos 1 sin
=
(1 2 ) (1 2 )
q p
1 2 sin2 1 2 + 2 co

The second term in the last line now cancels the integral of the first term
in 8, so were left with





sin d 1 cos
(12) =
(1 2 )
3/2 q
1 v2 sin2 /c2 2

0 2
1 sin
0
2
(13) = = 2 2
12
[There might be an easier way to do this, but I couldnt see any obvious
substitutions that worked.]
Plugging this back into 5 we get

q
(14) E da =
0

so Gausss law is satisfied.


GAUSSS LAW FOR A RELATIVISTIC POINT CHARGE 3

We can also calculate the Poynting vector by using the magnetic field of
a point charge in uniform motion, which Griffiths works out in his example
12.14:

1
(15) B = vE
c2
0 qv 1 v2 /c2 sin

(16) =
4 1 (v2 /c2 ) sin2 3/2 R2
0 qc sin
(17) = 3/2 R2
4 2 1 2 sin2

where the direction is that found by using the right-hand rule with the
thumb pointing in the direction of v, the particles motion. The Poynting
vector is

1
(18) S = EB
0
If we restrict ourselves to a charge moving in the +z direction, so that
v = vz and at the time when the charge passes through the origin, then R
in 3 becomes the radial coordinate in spherical coordinates and is the
azimuthal coordinate. In that case, R = so

q2 c sin
(19) S= 3
16 0 1 2 sin R4
2 4 2


P INGBACKS
Pingback: Relativistic transformation of electric and magnetic fields: an
example
RELATIVISTIC TRANSFORMATION OF ELECTRIC AND
MAGNETIC FIELDS: AN EXAMPLE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.45.
As an example of the transformation equations for electromagnetic fields,
consider the following situation. In the lab frame A, we have a charge q
moving at speed v in the +x direction, and a charge +q moving at the same
speed v but in the x direction, with q following the path y = 0 and +q
following y = +d. Their positions are such that their closest approach oc-
curs when they cross the y axis.
First, we can work out the fields and the force on +q at this point in the
lab frame. The fields produced by a moving point charge are

q 1 R
(1) E = 2
40 1 2 sin2  3/2 R2
qv 1 v2 /c2 sin

1
(2) B =
40 c2 1 (v2 /c2 ) sin2 3/2 R2

where R is the vector from the moving charge to the observer and the direc-
tion of is found from using the right-hand rule on the particles velocity
v, as usual. The angle is the angle between R and v. In our case, at the
point where the charges are at their closest approach = /2 and R = d so
we get

q
(3) E = y
40 d 2
qv
(4) B = z
40 c2 d 2

The force on +q with velocity v = vx can be found from the Lorentz


force law:
1
RELATIVISTIC TRANSFORMATION OF ELECTRIC AND MAGNETIC FIELDS: AN EXAMPLE
2

(5) F = q (E + v B)
q2 v2
 
(6) = y + 2 (x z)
40 d 2 c
2
q

v2

(7) = 1 + 2 y
40 d 2 c
Now suppose we switch to frame B in which +q is at rest. This frame is
moving with velocity vx with respect to A, so the transformation equations
are

(8) Ex = Ex
(9) Ey = (Ey + vBz )
(10) Ez = (Ez vBy )
(11) Bx = Bx
 v 
(12) By = By 2 Ez
c 
 v
(13) Bz = Bz + 2 Ey
c
[Note that this is a special case where the speed v of the frame B happens
to be the same as the speed of the charge in the original lab frame A, so we
can use the same symbol for both. In the more general case, the v in the
above 6 equations would be different from the v in equations 1 and 2.]
Only Ey and Bz are non-zero, so we get

q 2 v2
 
(14) E = 1 + 2 y
40 d 2 c
2q 2 v
(15) B = z
40 d 2 c2
[The y and y, and z and z axes are parallel so we can use the unit vectors
from frame A or B.] The force seen in frame B (where the velocity of +q is
v = 0 so there is no magnetic force) is thus

(16) F = qE
q2 2 v2
 
(17) = 1 + 2 y
40 d 2 c
RELATIVISTIC TRANSFORMATION OF ELECTRIC AND MAGNETIC FIELDS: AN EXAMPLE
3

Note that the force in frame B is larger by a factor of than the force in
frame A. As we saw earlier, an object experiences its maximum force in the
frame in which its at rest.
Finally, lets look at things in frame C where q is at rest. This can be
found from the results from frame A by transforming to a frame moving
at +v relative to A, so the transformation equations are equations 9 and 13
with v replaced by v, giving

(18) E = (Ey vBz ) y


q 2 v2
 
(19) = 1 2 y
40 d 2 c
q
(20) = y
40 d 2
 v 
(21) B = Bz 2 Ey z
c
(22) = 0
Since q is at rest in frame C, its electric field is just the Coulomb field
from electrostatics, and there is no magnetic field. The force on +q is there-
fore just the Coulomb force

q2
(23) F = y
40 d 2
RELATIVISTIC INVARIANTS INVOLVING
ELECTROMAGNETIC FIELDS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.46.
Using the transformation equations for electromagnetic fields, we can
show that there are a couple of quantities that are invariant under transfor-
mations between inertial frames.
The transformation equations are

(1) Ex = Ex
(2) Ey = (Ey + vBz )
(3) Ez = (Ez vBy )
(4) Bx = Bx
 v 
(5) By = By 2 Ez
c 
 v
(6) Bz = Bz + 2 Ey
c

First, we calculate the dot product E B and transform it:

(7)
h  v   v i
E B = Ex Bx + 2 (Ey + vBz ) By 2 Ez + (Ez vBy ) Bz + 2 Ey
c c
 
v2 
(8) = Ex Bx + 2 (Ey By + Ez Bz ) 1 2
c
(9) = Ex Bx + Ey By + Ez Bz
(10) = E B

Second, we can find another invariant:


1
RELATIVISTIC INVARIANTS INVOLVING ELECTROMAGNETIC FIELDS 2

(11)
  
2 2 2 2 2 v 2  v 2
E c B = Ex2 c2 B2x + 2 (Ey + vBz ) + (Ez vBy ) c2
By 2 Ez + Bz + 2 Ey
c c
(12)
v2
  
= Ex2 c2 B2x + 2 Ey2 + Ez2 c2 B2y c2 B2z

1 2
c
(13)
= E 2 c2 B2
These invariants put constraints on the forms given electric and magnetic
fields can have in different frames. For example, if B = 0 and E 6= 0 in
one frame, then E 2 c2 B2 > 0 in all frames, so its impossible to find a
frame in which E = 0. The first invariant also tells us that if E and B are
perpendicular in one frame (as they are in an electromagnetic wave, for
example) then they are perpendicular in all frames.
P INGBACKS
Pingback: The dual electromagnetic field tensor
RELATIVISTIC TRANSFORMATION OF ELECTROMAGNETIC
WAVES; THE DOPPLER EFFECT

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.47.
Lets apply the transformation equations for electromagnetic fields to an
electromagnetic wave. Suppose we have a plane EM wave with frequency
in the lab frame. The wave is polarized in the y direction and is travelling
in the x direction, so its fields are

(1) E = E0 ei(kxt) y
E0 i(kxt)
(2) B = e z
c

where


(3) k=
c
To see how this wave looks in a frame moving with velocity v in the x
direction, we can use the transformation equations

(4) Ey = (Ey vBz )


 v 
(5) Bz = Bz 2 Ey
c
We get
 v
(6) E = E0 1 ei(kxt) y
s c
1 v/c i(kxt)
(7) = E0 e y
1 + v/c
E0  v  i(kxt)
(8) B = 1 e z
cs c
E0 1 v/c i(kxt)
(9) e z
c 1 + v/c
1
RELATIVISTIC TRANSFORMATION OF ELECTROMAGNETIC WAVES; THE DOPPLER EFFECT
2

To get the final forms, we used

 v 1 v/c
(10) 1 = p
c 1 v2 /c2
1 v/c
(11) = p
(1 v/c) (1 + v/c)
s
1 v/c
(12) =
1 + v/c

The amplitude of the wave gets smaller as the speed of the frame in-
creases, becoming zero as v c.
To express this in the moving frames coordinates, we use the (inverse)
Lorentz transformations:

(13) x = (x + vt)
 
vx
(14) t = t + 2
c

giving

 v 
(15) kx t = k 2 x ( kv) t
c 
 v  v
(16) = 1 x 1 t
cs c s c
1 v/c 1 v/c
(17) = x t
c 1 + v/c 1 + v/c

Thus in the moving frame, the frequency of the wave is

s
1 v/c
(18) =
1 + v/c

and the wavelength is


RELATIVISTIC TRANSFORMATION OF ELECTROMAGNETIC WAVES; THE DOPPLER EFFECT
3

2
(19) =
k
2c
(20) =
s
2c 1 + v/c
(21) =
1 v/c
s
1 + v/c
(22) =
1 v/c
That is, the wavelength gets longer, approaching infinity as v c, while
the frequency gets smaller, approaching zero as v c. This is the Doppler
effect for light. If v > 0 so that we are moving in the direction of propagation
of the wave, the wavelength gets longer resulting in a red-shift. If v < 0 so
that we are moving against the direction of propagation, the wavelength gets
shorter and we have a blue-shift.
However, note that the speed of the wave is


(23) c = =c
2

so not surprisingly, the speed of the wave remains the same in the moving
frame.
The intensity of the wave in the lab frame is given by

E02
(24) I=
c0
2
In the moving frame this becomes

E02 1 v/c
(25) I = c0
2 1 + v/c

so the ratio is

I 1 v/c
(26) =
I 1 + v/c

and the intensity drops to zero as v c.


RELATIVISTIC TRANSFORMATION OF ELECTROMAGNETIC WAVES; THE DOPPLER EFFECT
4

P INGBACKS
Pingback: Barnards star: distance and velocity
ELECTROMAGNETIC FIELD TENSOR: LORENTZ
TRANSFORMATIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Moore, Thomas A., A General Relativity Workbook, Univer-
sity Science Books (2013) - Chapter 7; Problems 7.1.
Griffiths, David J. (2007), Introduction to Electrodynamics, 3rd Edition;
Pearson Education - Chapter 12, Problem 12.48.
To apply this to Griffiths problem 12.48, use

t 01 t 02 t 03

0
t 01 0 t 12 t 13
(1) ti j = 02 12

t t 0 t 23
t 03 t 13 t 23 0
in place of F i j in what follows. The results are the same.
The electromagnetic field tensor is

0 Ex Ey Ez
Ex 0 Bz By
(2) Fi j =
Ey Bz

0 Bx
Ez By Bx 0
We can use the usual tensor transformation rules to see how the electric
and magnetic fields transform under a Lorentz transformation. We get

x0i x0 j kl
(3) F 0i j = F
xk xl
j
(4) = i k l F kl

where the Lorentz transformation matrix is



0 0
0 0
(5) i k =
0

0 1 0
0 0 0 1
1
ELECTROMAGNETIC FIELD TENSOR: LORENTZ TRANSFORMATIONS 2

As we saw when discussing the inertia tensor, we can write this transfor-
mation as a matrix equation

(6) F 0 = FT

The first product is


Ex Ex Ey Bz Ez + By
Ex Ex Ey + Bz Ez By
(7) F =
Ey

Bz 0 Bx
Ez By Bx 0

The final product is

(8)

2 1 2 Ex Ey Bz

0  Ez + By
2 12 E 0 E + B Ez By
F 0 = FT = x y z


Ey + Bz Ey Bz 0 Bx
Ez By Ez + By Bx 0
p
Using = 1/ 1 2 we get


0 Ex Ey Bz Ez + By
Ex 0 Ey + Bz Ez By
(9) F 0 =


Ey + Bz Ey Bz 0 Bx
Ez By Ez + By Bx 0

From this, we see that

(10) Ex0 = Ex
(11) Ey0 = Ey Bz
(12) Ez0 = Ez + By
(13) B0x = Bx
(14) B0y = Ez + By
(15) B0z = Ey + Bz

Unlike lengths, the components of E and B in the direction of motion are


unchanged, while those perpendicular to the motion are altered.
ELECTROMAGNETIC FIELD TENSOR: LORENTZ TRANSFORMATIONS 3

P INGBACKS
Pingback: Symmetry of a rank 2 tensor is preserved under Lorentz trans-
formation
Pingback: The dual electromagnetic field tensor
Pingback: Electromagnetic field tensor: a couple of examples
Pingback: Electromagnetic Minkowski force
SYMMETRY OF A RANK 2 TENSOR IS PRESERVED UNDER
LORENTZ TRANSFORMATION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.49.
Weve seen that the electric and magnetic fields can be represented as
components of an anti-symmetric rank 2 tensor. In fact, any symmetric or
anti-symmetric tensor retains its symmetry property under Lorentz transfor-
mation. We can show this by some index juggling as usual in relativity.
First, suppose the tensor T is symmetric so that T i j = T ji . Then under
Lorentz transformation, we have

j
(1) T i j = ik l T kl
j
(2) = ik l T lk
j
(3) = l ik T lk
(4) = T ji
If T is anti-symmetric, then T i j = T ji and

j
(5) T i j = ik l T kl
j
(6) = ik l T lk
j
(7) = l ik T lk
(8) = T ji
If you want to use full matrix notation, you can write

(9) T = T T
Taking the transpose of a matrix product reverses the order of terms and
takes the transpose of each term, so

T
(10) T T = T T
(11) = T T T
1
SYMMETRY OF A RANK 2 TENSOR IS PRESERVED UNDER LORENTZ TRANSFORMATION
2

[Sorry about using a T for the matrix and a superscript T for transpose,
but hopefully you can keep them separate.]
Now for a symmetric matrix T T = T , so T T = T and for an anti-symmetric
matrix T T = T , so T T = T , showing that the symmetry property is pre-
served. In fact, this latter proof shows that the symmetry or anti-symmetry
of T is preserved no matter what matrix is.
THE DUAL ELECTROMAGNETIC FIELD TENSOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problems 12.50-51.
Here are a couple of examples involving the electromagnetic field tensor


0 Ex Ey Ez
Ex 0 Bz By
(1) Fi j =
Ey Bz

0 Bx
Ez By Bx 0

The tensor leads to the Lorentz transformations of the fields:

(2) Ex0 = Ex
(3) Ey0 = Ey Bz
(4) Ez0 = Ez + By
(5) B0x = Bx
(6) B0y = Ez + By
(7) B0z = Ey + Bz

Actually, the same transformations can be obtained by replacing E by B,


and B by E (using c = 1) in the original tensor. This gives us another
rank-2 tensor which is the dual tensor to F i j :


0 Bx By Bz
Bx 0 Ez Ey
(8) Gi j =
By Ez

0 Ex
Bz Ey Ex 0

If we lower both indices in these two tensors, we change the signs of all
elements in the first row and first column, to get
1
THE DUAL ELECTROMAGNETIC FIELD TENSOR 2


0 Ex Ey Ez
Ex 0 Bz By
(9) Fi j =
Ey Bz

0 Bx
Ez By Bx 0

0 Bx By Bz
Bx 0 Ez Ey
(10) Gi j =
By Ez 0 Ex
Bz Ey Ex 0
We can now calculate some invariants by finding

(11) F i j Fi j = 2E 2 + 2B2
(12) Gi j Gi j = 2B2 + 2E 2
(13) F i j Gi j = 2E B 2E B = 4E B
Comparing these results with those got earlier by directly calculating the
Lorentz transformation of these quantities, we see that the tensor products
give the same invariants (after restoring the factors of c).
As simple of example of calculating the elements in the tensors, suppose
we have an infinite straight wire along the z axis with linear charge density
moving at speed v. From Gausss law the electric field a distance x from
the wire is


(14) E= r
20 x

and from Ampres law the magnetic field is

0 v
(15) B=
2x

so at point (x, 0, 0) we have


(16) Ex =
20 x
(17) Ey = Ez = 0
0 v
(18) By =
2x
(19) Bx = Bz = 0
THE DUAL ELECTROMAGNETIC FIELD TENSOR 3

so the tensors are (again with c = 1)




0 20 x 0 0
0 v
0 0 2x

(20) F i j = 20 x


0 0 0 0
0 v
0 2x 0 0
0 v

0 0 2x 0
0 0 0 0
(21) Gi j = 0 v


2x 0 0 20 x


0 0 2 0x
0
ELECTROMAGNETIC FIELD TENSOR: A COUPLE OF
EXAMPLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problems 12.52-53.
Maxwells equations can be expressed quite simply in terms of the elec-
tromagnetic field tensor

0 Ex Ey Ez
Ex 0 Bz By
(1) Fi j =
Ey Bz

0 Bx
Ez By Bx 0

and its dual tensor:



0 Bx By Bz
Bx 0 Ez Ey
(2) Gi j =
By Ez

0 Ex
Bz Ey Ex 0
In his section 12.3.4, Griffiths shows that all four of Maxwells equations
can be expressed as

(3) j F i j = 0 J i
(4) j Gi j = 0

where J i is a four-vector known as the four-current with components

(5) J i = [ c, Jx , Jy , Jz ]
0
(6) = p [ c, ux , uy , uz ]
1 u2 /c2
In this formula, were looking at a small element of charge that is moving
with velocity u and has charge density 0 in its rest frame. The four-current
is just 0 times the four-velocity:
1
ELECTROMAGNETIC FIELD TENSOR: A COUPLE OF EXAMPLES 2

(7) J i = 0 i

If we take the divergence (in four dimensions) of 3 we get, since F i j =


F ji

(8) i j F i j = i j F ji = 0

which leads to

(9) i J i = 0

This is a compact way of expressing the continuity condition, since from


5

(c)
(10) i J i = +J
x0

(11) = +J = 0
t
The equation 4 can be written in terms of Fi j as follows. First, we can
write Gi j in terms of the components of Fi j by comparing the two matrices
2 and Fi j :


0 Ex Ey Ez
Ex 0 Bz By
(12) Fi j =
Ey Bz

0 Bx
Ez By Bx 0

We get


0 F23 F31 F12
F32 0 F03 F20
(13) Gi j =
F13 F30 0 F01

F21 F02 F10 0

Now we can read off the four equations contained in 4 by scanning the
rows of this matrix:
ELECTROMAGNETIC FIELD TENSOR: A COUPLE OF EXAMPLES 3

(14) j G0 j = 1 F23 + 2 F31 + 3 F12


(15) j G1 j = 0 F32 + 2 F03 + 3 F20
(16) j G2 j = 0 F13 + 1 F30 + 3 F01
(17) j G3 j = 0 F21 + 1 F02 + 2 F10
All four of these equations have the form

(18) j Gi j = a Fbc + b Fca + c Fab

where i, a, b and c are all different. Note that the indexes a, b and c are
cyclically permuted in each term on the RHS. Therefore we can write 4 as

(19) a Fbc + b Fca + c Fab = 0

where a, b and c are a subset of {0, 1, 2, 3}.


P INGBACKS
Pingback: Relativistic electromagnetic potentials
ELECTROMAGNETIC MINKOWSKI FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.54.
The Minkowski force is defined in general by

dp 1 dp 1
(1) K= =p =p F
d 1 2 dt 12

where p is the spatial part of the four-momentum. Griffiths shows in his


section 12.3.4 that the Minkowski force due to electromagnetic fields is

(2) K i = q j F i j

where the electromagnetic field tensor is (with c = 1):


0 Ex Ey Ez
Ex 0 Bz By
(3) Fi j =
Ey Bz

0 Bx
Ez By Bx 0

and the proper velocity is

dxi ui
(4) i = =p
d 12
Griffiths shows that the spatial components of K i work out to

q 1
(5) K= p (E + u B) = p F
1 2 12

so the relation 1 between K and F is correct.


To see what the time component of 2 gives us, we can just work it out by
reading off the first row of F i j :
1
ELECTROMAGNETIC MINKOWSKI FORCE 2

(6) K 0 = q j F 0 j
q
(7) = p (ux Ex + uy Ey + uz Ez )
12
q
(8) = p uE
12
1
(9) = p u FE
12
The term u FE is the rate at which the electric field does work on the
charge as it moves along (remember that because the direction of motion is
always perpendicular to the force exerted by a magnetic field, B never does
any work, so this is the total work done by the electromagnetic field).
CONTRAVARIANT GRADIENT OPERATOR

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.55.
The gradient of a scalar function is a covariant vector since it trans-
forms as

xi
(1) =
xa xi xa

We can therefore regard the gradient operator a on its own as a covariant


vector, so it should have a contravariant counterpart. In flat space, the only
change in switching from covariant to contravariant is that the time compo-
nent changes sign. Given that the Lorentz transformation for a contravariant
four-vector is

x0 = x0 x1

(2)
x1 = x1 x0

(3)
(4) x2 = x2
(5) x3 = x3

the transformations for the covariant four-vector are obtained by lowering


all indices and replacing the time components by their negatives:

(6) x0 = (x0 + x1 )
(7) x1 = (x1 + x0 )
(8) x2 = x2
(9) x3 = x3

where we multiplied the x0 equation through by 1. The corresponding


inverse transformations are obtained by replacing by :
1
CONTRAVARIANT GRADIENT OPERATOR 2

(10) x0 = (x0 x1 )
(11) x1 = (x1 x0 )
(12) x2 = x2
(13) x3 = x3
Thus the inverse covariant transformations are the same as the forward
contravariant transformations.
The contravariant gradient is i = xi so


(14) i =
xi
xk
(15) =
xk xi
xk
(16) = k
xi
The transformations for each value of i are then

(17) 0 = 0 1
(18) 1 = 1 0
(19) 2 = 2
(20) 3 = 3
Thus i transforms like a contravariant vector.
P INGBACKS
Pingback: Relativistic electromagnetic potentials
RELATIVISTIC ELECTROMAGNETIC POTENTIALS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.56.
Maxwells equations can be written in terms of the electromagnetic field
tensor, its dual and the current four-vector as

(1) j F i j = 0 J i
(2) j Gi j = 0
where

0 Ex Ey Ez
Ex 0 Bz By
(3) Fi j =
Ey Bz

0 Bx
Ez By Bx 0

0 Bx By Bz
Bx 0 Ez Ey
(4) Gi j =
By Ez

0 Ex
Bz Ey Ex 0
(5) J i = [, Jx , Jy , Jz ]
0
(6) = p [1, ux , uy , uz ]
12
It turns out that an even more compact form for Maxwells equations can
be written using the 4-vector potential
 
i V
(7) A = , Ax , Ay , Az
c
(8) = (V, Ax , Ay , Az )

where the last line uses relativistic units where c = 1.


Griffiths shows in section 12.3.5 that the field tensor can be written in
terms of the potentials as
1
RELATIVISTIC ELECTROMAGNETIC POTENTIALS 2

(9) F i j = i A j j Ai
Note that were using the contravariant gradient operator here, in order to
get the signs right on the time components. Because of the form of F i j , the
gauge invariance shows up naturally, since if we replace Ai by

(10) Ai Ai + i

where is any scalar function, F i j is unchanged, as the order in which


the partial derivatives of are taken doesnt matter, so drops out of the
equation for F i j . The Lorenz gauge condition is

V
(11) A = = 0V
t

[Notice were back to using the covariant gradient operator.] This can be
condensed to read

(12) i Ai = 0
[Incidentally, Griffithss equation 12.135 is wrong; it should read A / x =
0.]
Combining 1 with 9 gives

(13) j i A j j j Ai = 0 J i
If we use the Lorenz gauge, the first term is zero, so we get

(14) j j Ai = 0 J i
(15) 2 Ai = 0 J i

where the symbol 2 is the dAlembertian operator, defined as

(16) 2 j j
We can verify that the other Maxwell equation 2 is also satisfied by the
potential formulation by using the earlier result

(17) j Gi j = a Fbc + b Fca + c Fab


RELATIVISTIC ELECTROMAGNETIC POTENTIALS 3

where i, a, b and c are all different.


Lowering the indexes on 9 we get

(18) Fi j = i A j j Ai
Substituting 18 into 17 we get

(19) j Gi j = a b Ac a c Ab + b c Aa b a Ac + c a Ab c b Aa = 0
LORENTZ TRANSFORMATION IN TWO DIMENSIONS

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.57.
Although weve looked at the Lorentz transformations for a general 3-d
motion of one frame relative to another, well have a look here at the slightly
more specialized case of general 2-d motion.
Suppose that frame S moves relative to S with velocity

(1) v = c (cos x + sin y)

that is, its direction makes an angle with the x axis. To get the Lorentz
transformation here, we use the fact that distances perpendicular to v are un-
affected, and distances parallel to v transform using the regular 1-d Lorentz
transformation. Its therefore easiest to transform to a lab frame S 0 that is
rotated relative to S by , so that the basis vectors are

(2) x0 = cos x + sin y


(3) y0 = sin x + cos y
A point with coordinates [x0 , y0 ] in S 0 therefore has coordinates in terms
of S coordinates of

(4) x0 = x cos + y sin


(5) y0 = x sin + y cos
In this system, the Lorentz transformation is

0 0
0 0
0 =

(6) 0

0 1 0
0 0 0 1
We can write this out explicitly for each coordinate. The time coordinate
is the same both S and S 0 since the two frames are at rest relative to each
other, so
1
LORENTZ TRANSFORMATION IN TWO DIMENSIONS 2

t0 = t x0

(7)
(8) t = t x cos y sin
The x0 coordinate transforms as follows:

x0 = x0 t

(9)
(10) x cos + y sin = x cos + y sin t
And y0 transforms as

(11) y0 = y0
(12) x sin + y cos = x sin + y cos
Multiplying 10 by cos and 12 by sin and adding, we get

x = t cos + cos2 + sin2 x + ( 1) y sin cos



(13)
Multiplying 10 by sin and 12 by cos and adding, we get

y = t sin + ( 1) x sin cos + sin2 + cos2 y



(14)
Combining 8, 13 and 14 (along with z = z) into a matrix, we get

cos sin 0
cos cos2 + sin2 ( 1) sin cos 0
(15) =
sin ( 1) sin cos sin2 + cos2

0
0 0 0 1
This reduces to 6 when = 0, so that the axes of the two frames are
parallel. This 2-d transformation matrix is also symmetric.
COLLISION OF A PION AND A PROTON

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.58.
As another example of using conservation of energy and momentum to
work out the kinematics of particle collisions, suppose we fire a pion at a
proton at rest. One possible outcome of such a collision is the conversion of
the pion and proton into kappa and sigma particles, but this can only occur
if the momentum of the pion is high enough, since the rest energies of the
kappa plus sigma are greater than those of the pion plus proton. We can
find the minimum pion momentum (called the threshold momentum), as
measured in the lab, at which this reaction can occur. Its easiest to convert
to the centre of momentum frame to do the calculations and then convert
back at the end.
In the centre of momentum frame, the pion and proton head towards each
other with equal and opposite momenta, so using the usual relativistic nota-
tion, and expressing rest energy in MeV (so we can ignore the c2 factor):

(1) m = p p m p
At the threshold momentum, the pion and proton collide and produce a
K and at rest, so from conservation of energy

(2) m + p m p = mK + m
Using Griffithss approximate values for the rest energies, we have (in
MeV)

(3) m = 150
(4) mp = 900
(5) mK = 500
(6) m = 1200

so from 1 and 2
1
COLLISION OF A PION AND A PROTON 2

(7) 150 = 900 p p


(8) = 6 p p
(9) 150 + 900 p = 1700
34
(10) + 6 p =
3

From these equations, we get

 2
34 1
(11) 2 = 6 p =
3 1 2
 2
34
(12) 2 = 1 6 p
3
 2
34
(13) ( )2 = 6 p 1
3
(14) = 36 p2 p2

where we used 8 to get the last line. We can now solve the last two equations
to find p :

 2
34
(15) 36 p2 p2 = 6 p 1
3
 2
34
(16) = 1 136 p + 36 p2
3
 2
34
1 136 p + 36 p2 1 p2

(17) 0 =
3
 2
34
(18) 0 = 1 + 36 136 p
3
(19) p = 1.202
(20) p = 0.555

We can now get the values for the pion in the centre of momentum frame
from 11 and 12:
COLLISION OF A PION AND A PROTON 3

34
(21) = 6 p = 4.123
s3
 2
34
(22) = 1 6 p = 0.970
3
The speed of the proton in the centre of momentum frame is also the
speed of the centre of momentum frame relative to the lab frame, so we can
use a Lorentz transformation on the pions four-momentum to get back to
the lab frame:

p1 = p p1 + p p0

(23)
(24) = 1.202 (150 + 0.555 150 )
(25) = 1133 MeV/c
Notice that we must use p and p (that is, the values for the proton, not
the pion) in doing the Lorentz transformation, since its the speed of the
proton that determines the relative speed of the two frames.
ELASTIC COLLISION OF TWO IDENTICAL PARTICLES

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.59.
Heres another example the kinematics of particle collisions in relativity.
This time, well look at the common first year physics problem of elastic
scattering of two identical masses, with one mass m coming in along the x
axis with speed v and hitting another mass m at rest. In classical physics,
the two masses always fly off at right angles after the collision, but things
are more complicated in relativity. As usual, its easiest to convert to the
centre of momentum frame to do the calculations and then convert back at
the end.
The speed of the centre of momentum frame is (Ill set c = 1 and use the
standard symbol = v/c to make notation easier; we can reinsert the c at
the end):

pi
(1) =
Ei
m
(2) =
m + m

(3) =
1+

where the quantities on the RHS refer to the lab frame. Centre of momen-
tum (COM) quantities have a bar over them. Since mass 2 is at rest in the
lab, its velocity in the COM frame is


(4) 2 =
1+

and by conservation of momentum, the COM velocity of mass 1 must be


(5) 1 =
1+
Remark. At this point, we can check the results using the velocity addition
formula. For particle 2, its lab velocity is given by
1
ELASTIC COLLISION OF TWO IDENTICAL PARTICLES 2

2 +
(6) 2 = =0
1 + 2
For particle 1, we have

1 +
(7) 1 =
1 + 1
1
(8) = 2
1 + 1 + ( ) / (1 + )2
2

(1 + )2
(9) = 2
1 + (1 + )2 + ( )2
2 (1 + )
(10) =
1 + 2 + 2 + ( )2
2 (1 + )
(11) =
2 (1 2 ) + 2 + 2 + ( )2
2 (1 + )
(12) =
2 + 2 2
(13) =

where in the fifth line we used 2 1 2 = 1. Thus both velocities convert



correctly back to the lab frame.
Returning to our main calculation, what wed like to do is find the scat-
tering angle in the lab frame, given the angle between particle 1s initial
and final directions in the COM frame. As the total momentum in COM is
zero, we know that the two particles must fly off in opposite directions, so
we can write out the x and y components for both particles as

(14) 1x = cos
(15) 1y = sin
(16) 2x = cos
(17) 2y = sin
Using the velocity addition
q formulas, we can convert these back to the
lab frame, using 1/ 1 2 .
ELASTIC COLLISION OF TWO IDENTICAL PARTICLES 3

1x +
(18) 1x =
1 + 1x
cos +
(19) =
1 + 2 cos
1y
(20) 1y = 
1 + 1x
sin
(21) = 
1 + 2 cos
cos +
(22) 2x =
1 2 cos
sin
(23) 2y = 
1 2 cos

In the lab frame, the angle 1 that the incident particle makes with its
incoming direction is given by

1y
(24) tan 1 =
1x
sin
(25) = 
1 + cos

Similarly for particle 2

2y
(26) tan 2 =
2x
sin
(27) = 
1 cos

Since 1 > 0 and 2 < 0 (that is, in the lab frame, particle 1 scatters up-
wards from the x axis while particle 2 scatters downwards; obviously we
could interchange the roles of the two particles but to conserve y momen-
tum, the two particles must scatter on opposite sides of the x axis), the total
angle between the two particles is = 1 2 . Using the formula for the
tangent of the difference of two angles, we get
ELASTIC COLLISION OF TWO IDENTICAL PARTICLES 4

(28)
tan 1 tan 2
tan =
1 + tan 1 tan 2
(29)
" #1
sin2

sin 1 1
= + 1 2  
1 + cos 1 cos 1 + cos 1 cos
(30)
" #1
sin2

sin 2
= 1 
1 cos2 2 1 cos2
(31)
2
=
sin (1 1/ 2 )
(32)
2
=
2 sin
(33)
 
2
= arctan
2 sin
[This answer is the same form as that given in Griffithss question, except
that he seems to use lab coordinates for and . However, after checking
my solution I cant see anything wrong, so hopefully Ive got it right.]
In the classical limit, becomes very small and 1 so from 3,
also becomes very small and 1, so the argument of the arctan tends to
infinity and /2 as required.
MOTION UNDER A CONSTANT MINKOWSKI FORCE

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.60.
The Minkowski force K is the rate of change of four-momentum with
respect to proper time, and allows Newtons law to be written in its natural
form

(1) K = m

where is the proper acceleration, or second derivative of position with


respect to proper time. Here well investigate the behaviour of a particle
subject to a constant Minkowski force in one dimension.
In terms of ordinary force, we have

d p d p dt 1
(2) K== =p F
d dt d 1 u2 /c2
The ordinary momentum p is

mu
(3) p= p
1 u2 /c2
so its derivative is

dp m du mu2 du
(4) =p + 3/2
dt 1 u /c dt (1 u2 /c2 ) dt
2 2

Inserting this into 2 we get

K du u2 du
(5) dt = +
m 1 u2 /c2 (1 u2 /c2 )2
We can integrate both sides (using software, or integral tables) to get

c2
   
K c c+u 1 1
(6) t +C = ln +
m 4 cu 4 cu c+u
1
MOTION UNDER A CONSTANT MINKOWSKI FORCE 2

where C is a constant of integration. If the initial conditions are u = 0 at


t = 0, then C = 0 and we have

c2
   
K c c+u 1 1
(7) t = ln +
m 4 cu 4 cu c+u
This is an implicit equation for the speed of the particle as a function
of time. If we want the position as a function of time, we need a relation
between u and x. Returning to 2 and 3 we have
!
K d u
q
(8) 1 u2 /c2 = p
m dt 1 u2 /c2
We can use the chain rule to convert the derivative on the RHS to a de-
rivative with respect to x by multiplying both sides by dt/dx
! !
dt K dt d u d u
q
(9) 1 u2 /c2 = p = p
dx m dx dt 1 u2 /c2 dx 1 u2 /c2
Now dx/dt = u so dt/dx = 1/u and
p !
1 u2 /c2 K d u
(10) = p
u m dx 1 u2 /c2
If we call the expression in the parentheses on the RHS A, then we can
integrate with respect to x (since K/m is a constant):

u
(11) A p
1 u2 /c2
1K dA
(12) =
Am dx
K 1 2
(13) x +C = A
m 2
Again, starting from rest at the origin we have u = 0 when x = 0 so A = 0
also, and therefore C = 0, so we have
r
u 2Kx
(14) A= p =
2
1 u /c 2 m
At this point we could get a relation between x and t by solving 14 for u
in terms of x and then substituting this into 7. For reference, we get
MOTION UNDER A CONSTANT MINKOWSKI FORCE 3

r
2Kx 1
(15) u= p
m 1 + 2Kx/mc2

so substituting will give something of a mess. To get the answer given in


Griffiths requires a bit of algebra, but here is how I did it. Griffiths defines
the quantity z as

r
2Kx
(16) z
mc2
A
(17) =
c
u
(18) = p
c 1 u2 /c2
The quantities appearing in Griffithss answer are

p c
(19) 1 + z2 =
c2 u2
p u
(20) z 1 + z2 =
c (1 u2 /c2 )
We can rewrite 7 to get

   
2Kt 1 c+u c 1 1
(21) = ln +
mc 2 cu 2 cu c+u
Well deal with the logarithm first. Its argument is

c+u (c + u)2
(22) = 2
cu c (1 u2 /c2 )
2u u2 + c2
(23) = +
c (1 u2 /c2 ) c2 u2
2u c2 u2 + 2u2
(24) = +
c (1 u2 /c2 ) c2 u2
2u 2u2
(25) = +1+ 2
c (1 u2 /c2 ) c (1 u2 /c2 )
Now we also have
MOTION UNDER A CONSTANT MINKOWSKI FORCE 4

 p 2 p
(26) z+ 1+z2 = 2z2 + 2z 1 + z2 + 1
2u2 2u
(27) = + +1
c (1 u /c ) c (1 u2 /c2 )
2 2 2
c+u
(28) =
cu

Therefore
  r
1 c+u c+u
(29) ln = ln
2 cu cu
 p 
(30) = ln z + 1 + z2
For the second term in 21, we have
 
c 1 1 c 2u
(31) =
2 cu c+u 2 c (1 u2 /c2 )
2
u
(32) =
c (1 u2 /c2 )
p
(33) = z 1 + z2
Putting it all together, we have

2Kt  p  p
(34) = ln z + 1 + z2 + z 1 + z2
mc
SELF-FORCE ON A DIPOLE IN HYPERBOLIC MOTION

Link to: physicspages home page.


To leave a comment or report an error, please use the auxiliary blog.
Reference: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Chapter 12, Problem 12.61.
Here well revisit the problem of calculating the self-force of a moving
charge. In our earlier derivation, we considered a single charge q split into
two equal charges q/2 separated by a distance d and moving perpendicular
to the line joining the two charges. If the motion is in the x direction and the
line joining the charges is parallel to the z axis, the only net electric field felt
by one charge due to the field generated by the other charge at a retarded
time tr is

  2   
q r cl c lv
(1) Ex = v + la ac r
80 c r lv 3
3 r 2 c
c

where r is the distance from the other charge at the retarded time to the
current charge at the current time, l is the distance moved in the x direction
in the time t tpr and v and a are the velocity and acceleration at the retarded
time. = 1/ 1 v2 /c2 as usual. The calculating proceeded from here
by assuming that the separation distance d was small and deriving the self-
force in that limiting case.
Well now consider an electric dipole consisting of charges +q and q
separated by a distance d, but without assuming d to be small. The electric
field due to +q at the retarded time tr felt by q at the current time t is
therefore

  2   
q r cl c lv
(2) Ex = v + la ac r
40 c r lv 3
3 r 2 c
c
 2 
q 1 c l 2 cvr 2
(3) =
40 c2 3 2 + al 2 ar
r lv
c

where weve changed the 8 in the denominator to a 4 because weve re-


placed q/2 by q. To go any further, we need to make some assumptions
about the motion of the dipole, so well suppose that it is moving under
1
SELF-FORCE ON A DIPOLE IN HYPERBOLIC MOTION 2

hyperbolic motion. Griffiths wants us to consider the position of the charge


to be given by

mc2
q
(4) x (t) = 1 + (Ft/mc)2 1
F

However, to simplify the notation a bit, well consider the more general
form
q
(5) x (t) = b2 + (ct)2

where b is a constant with dimensions of length. This is equivalent to 4


except that x = b at t = 0 rather than x = 0. Since all calculations depend
only on how much the dipole moves over a time interval and not on its
absolute position, this change wont affect anything that follows.
Using this form, we can calculate the velocity and acceleration by taking
derivatives:

c2t c2t
(6) v (t) = q =
x
b2 + (ct)2
c2 c4 t 2
(7) a (t) = q  3/2
b2 + (ct)2 b2 + (ct)2
b2 c2 b2 c2
(8) =  =
3/2 x3
b2 + (ct)2
1 v2
(9) = 1
2 c2
b2
(10) = 2
x
Adding a subscript r to indicate a quantity evaluated at tr and plugging
these equations into 3, we get

40 c2 lctr 3 c2 b2 l 2 crtr r2
   
(11) Ex = r l+
q xr xr2 xr xr xr
We now need to express r, l and xr in terms of tr . From the geometry of
the setup
SELF-FORCE ON A DIPOLE IN HYPERBOLIC MOTION 3

(12) r2 = l 2 + d 2

and since a signal travels from +q at time tr to q at time t over a distance


r, we have

(13) r = c (t tr )
q
(14) l = c2 (t tr )2 d 2
Substituting these into 11 and simplifying gives
q
40 c2 (t tr )2 d 2
c2tr (t tr ) + d 2 xr
(15) Ex =  q 3
qb2 2
3 2 2
c tr c (t tr ) d xr t + xr tr
q q
c tr (t tr ) + d b + (ctr ) c2 (t tr )2 d 2
2 2 2 2
(16) =  q q 3
3 2 2 2 2 2
c tr c (t tr ) d b + (ct) (t tr )

where we used 5 to get rid of xr in the last line.


Also, l is the distance moved in time t tr , so

(17) l = x (t) x (tr )


q q
(18) = b + (ct) b2 + (ctr )2
2 2

[Note that we cant take l = v (t tr ), since the velocity is changing over


that time interval.] We can therefore find t in terms of tr by solving
q q q
(19) c (t tr ) d = b + (ct) b2 + (ctr )2
2 2 2 2 2

This turns out to be a quadratic equation, with solutions


 
1
q
2 2 2 2 2 2 2
(20) t = tr + ct d dtr (4b + d ) (c tr + b )
2cb2tr r
To decide which sign to take, we note that for very small d, the square
root term dominates since the first term is of order d 2 . Since we must have
t > tr , well need to take the + sign. We then get for l:
SELF-FORCE ON A DIPOLE IN HYPERBOLIC MOTION 4

(21)
q
l = c2 (t tr )2 d 2
(22)
s 
d
q
= 2 4c2 b2tr2 + 2c2 d 2tr2 + d 2 b2 + 2cdtr (4b2 + d 2 ) (c2tr2 + b2 )
2b
(23)
s 2
d
q q
= 2 ctr (4b2 + d 2 ) + d c2tr2 + b2
2b
(24)
 q 
d
q
2 2 2
= 2 ctr (4b + d ) + d c tr + b2 2
2b
Plugging this into 16 we get

(25)
 p q
b2 + (ctr )2
d
p
40 c2t 2 2 2
r (t tr ) + d 2b2
2 2
ctr (4b + d ) + d c tr + b 2
Ex = 3
qb2
  p  q
2
d
p
3 2 2 2 2 2 2
c tr 2b2 ctr (4b + d ) + d c tr + b b + (ct) (t tr )

(26)
p p
4b4 tr2 c2 d 2 2c2 b2tr (t tr ) d 2 b2 + ctr d (4b2 + d 2 ) c2tr2 + b2
= 3 i3
c
hp p
c2tr2 + b2 (2b2 (t tr ) d 2tr ) cdtr2 (4b2 + d 2 )

We can now substitute from 20

 
1
q
2 2 2 2 2 2 2
(27) t tr = ct d + dtr (4b + d ) (c tr + b )
2cb2tr r
The numerator in 26 is now

(28)
 q  q q
tr2 c2 d 2 c ctr d + dtr (4b + d ) (c tr + b ) d b + ctr d (4b + d ) c2tr2 + b2 = d 2 b2
2 2 2 2 2 2 2 2 2 2 2

and the denominator is


SELF-FORCE ON A DIPOLE IN HYPERBOLIC MOTION 5

(29)
q     3
1
q q
2 2 2 2 2 2 2 2 2 2 2 2 2 2
c tr + b ct d + dtr (4b + d ) (c tr + b ) d tr cdtr (4b + d ) =
ctr r
q  3
2 2
d 22 2
 2
(30) (4b + d ) c tr + b cdtr =
c
b6 d 3 2 2 3/2

(31) 4b + d
c3
Putting this back into 26 we get

!
40 4b4 d 2 b2
(32) E x =
qb2 c3 b6 d 3
(4b2 + d 2 )
3/2
c3
4
(33) = 3/2
d (4b2 + d 2 )
qb2
(34) Ex = 3/2
0 d (4b2 + d 2 )
q
(35) =  3/2
2
80 db 1 + (d/2b)

Miraculously, the dependence on both t and tr has disappeared, showing


that the field (and therefore the self-force) is constant. To get this answer
back into the form given by Griffiths, we compare 4 and 5.

p
(36) x = b2 + c2t 2
r
 ct 2
(37) = b 1+
b
mc2
q
(38) = 1 + (Ft/mc)2
F
mc2
(39) b =
F
[We can ignore the 1 in 4 since that just changes the origin of x, and all
the calculations above depend only on xr x so the 1 cancels out.] With
this value for b we have
SELF-FORCE ON A DIPOLE IN HYPERBOLIC MOTION 6

qF
(40) Ex =
2 3/2
 
80 mc2 d 1 + (Fd/2mc2 )
The force felt by q is therefore

q2 F
(41) Fx = qEx =
2 3/2
 
80 mc2 d 1 + (Fd/2mc2 )
The total self-force on the dipole is twice this, since there is an equal
force on +q due to q, so

q2 F
(42) Ftot =
2 3/2
 
2 2
40 mc d 1 + (Fd/2mc )
[However, it would seem to me that there should also be forces on each
charge due to their own fields at the retarded time. If these forces are equal
in magnitude to the cross-self-force calculated here, they should also be op-
posite in direction since the force is between the same charge at different
times, causing repulsion. Thus it would seem that the total force is zero,
which actually makes more sense than having the dipole constantly accel-
erating without any external force. What am I missing?]
In any case, if we set Ftot = F and solve for F, we get
s
2/3
2mc2 q2

(43) F = 1
d 40 mc2 d
s
2 2/3
0 q2

2mc
(44) = 1
d 4md

Potrebbero piacerti anche