Guillaume Gervais
2007
Contents
1
1.1
1.1.1
Coulomb Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.2
1.1.3
Gauss Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.4
1.1.5
1.1.6
Capacity
1.2
1.3
1.4
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.2
1.2.3
1.2.4
Electric Displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Magnetostatics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.1
BiotSavart Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.2
Vector Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.3
Amperes Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.4
Magnetic Dipole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.5
1.3.6
Electrodynamics
2.1
25
Faraday Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.1
Electromotive Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.2
Faraday Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2
2.1.3
2.2
Inductance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2.1
2.3
2.4
General Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Maxwells Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.1
2.3.2
2.3.3
Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Electromagnetic Potentials
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4.1
2.4.2
Gauge Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4.3
2.4.4
Electromagnetic Waves
3.1
3.2
3.3
44
3.1.2
Sinusoidal Waves in 1D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.1.3
3.1.4
Polarization of Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.5
3.2.2
3.2.3
3.2.4
3.2.5
3.4
3.3.1
Normal Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.2
Oblique Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.3
Brewster angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.3.4
3.4.2
3.4.3
3.4.4
3.4.5
Propagation in a Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.4.6
Guided Waves
4.1
4.2
4.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Rectangular Waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.1
TE Modes (Ez = 0, Bz 6= 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.2
TM Modes (Ez 6= 0, Bz = 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.3.3
4.5
4.3
82
Electromagnetic Cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5.1
What is an EM cavity? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5.2
Electromagnetic Radiation
97
4
5.1
5.2
5.3
5.1.2
Retarded Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.1.3
LienartWiechert Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.1.4
5.2.2
5.2.3
5.3.2
Coulomb Law
F12
F21
q1
r12
q2
F~12 =
q1 q2
r
2 12
4o r12
with
r12
= ~r1 ~r2
~r1 ~r2
r12 =
r12
1.1.2
~
F~ = q E
qi
ri
(ri)
ri
rri
rri
We call the variable ~x0 the field point variable, whereas ~x is the source point variable.
1 X
qi
3 (~x ~x0i )
4o
i ~
~0
x xi
1
4o
(~x0 )
~0
3 (~x ~x0 )dx
~x ~x0
.
~ d3 x dV dxdydz is the differential volume element.
Here dx
7
1.1.3
Gauss Law
Gaussian Surface
~ =
~ ds
E
where
i Qi
X Qi
i
o
~ =
~ ds
E
~
~ E
~ dx
then
I
s
~ =
~ ds
E
Z
V
~ = 1
~ E
~ dx
o
~
dx
~ E
~ = (~x)
o
2 =
o
Z
=
2. Electric Force is Conservative: Since the electric field derives from a gradient of a scalar
function, we must have that
~ E
~ = ~0
i.e. the curl of any gradient of a scalar function must be identically zero, or
I
~ ~l =
Ed
~ E)d~
~ s=0
(
3. Important Identity
1.1.4
1
~x ~x0
1
(~x ~x0 )
o
1X
1 1 X qi qj
qi (x~i ) =
2
2 40
i
i,j ~
xi x~j
9
0 .
o
d~x(~x)(~x) =
2
d~x2
~ ()
~
~
~ as
We can rewrite using the identity 2 =
o
U=
2
~ 2 + o
d~x (E)
2
~
d~s E
Here the second term, the surface integral, as obtained using Gauss divergence theorem and does
not contribute since at infinity, the electric field is falling off as r12 , the potential as 1r and the
surface grows only as r2 . Far from the source, the resulting 1r behaviour dominates and the integral
does not contribute.
2
~
Finally, if we define an Energy Density as E = o /2E
then we can write the potential energy as
Z
Ed~x =
U=
1.1.5
2
o ~
E d~x
2
Imagine a closed surface S in which charges are flowing in and out through it:
S
q
q
q
10
i
j
k
~ s
Jd~
~ s=
Jd~
~ J)d~
~ x=
(
t
V
~
(~x)dx
~ J~ + = 0
1.1.6
Capacity
Because of the principle of superposition, the potential of a conductor is proportional to the charges
in the system. If we have N charges, Qi , located on N conductors, we can define the capacity as
Qi =
Cij j
where Cij is termed the coefficient of capacity for i = j, and the coefficient of electrostatic
influence for i 6= j.
11
Q = C
U=
1X
1X
Qi i =
Cij i j
2
2
i
i,j
1
U = C2
2
~ 0 (x~0 )x~0
dx
p~(~x) =
V
12
1.2.1
(~x) =
(2 ) (~x)
l=0
l
(2l )
1
(~x) =
4o rl+1
where Pl (t) are the Legendre Polynomial of order l, and where we have defined r ~x and r0 x~0 .
Note that the above relation results from the following mathematical expansion
X r0l
1
=
Pl (t)
rl+1
~x x~0
l=0
Monopole case, for l = 0:
(1)
1
(~x) =
4o r
d~x(x~0 ) =
q
4o r
1
(~x) =
4o r3
(2) (~x) =
1
p~ ~x
4o r3
1
p~ ~x
4o r3
~ 0 (x~0 )x~0
dx
p~(~x) =
V
+q
q
a
1.2.2
~ p~
U = E
~
= p~ E
~ dip
~ (2) (~x)
E
1
~
Edip =
3(~p ~x)~x p~
4o r3
14
1.2.3
We call the average polarization P~ (~x) = p~(~x)N where p~(~x) is the polarization of each molecule
inside the dialectric material. The applied electric field will create an overall charge in the material
that is bound to it. We define this bound charge as
~ P~
0 =
where the we use to denote the bound charge, as oppose to which describes the free charge
inside the material.
When the polarization, P~ (~x) is discontinuous, such as at an interface between a dialectric and
vacuum, a surface charge distribution is created
Z
~ =
~ P~ dx
d~s P~ = P~ ~nA
0s = P~ n
1.2.4
Electric Displacement
We shall define the total charge tot inside the material as the one that includes both the contribution from free electrons, and the bound charge:
tot = + 0
where here is the free charge and 0 is the bounded charge.
Note that from Gauss law it is the total charge that must be taken into account, therefore:
1
~ E
~ = 1 (~
~ P~ )
+ ~0 ) = (
o
o
15
so that
P~
~
~
=
E+
o
o
~ as
We define the electric displacement, D
~ o E
~ + P~
D
~ D
~ =
1.3 Magnetostatics
In electrostatics, we were concerned by the study of stationnary charges, i.e. with constant
electric fields, and in analogy magnetostatics refers to the study of steady currents, thereby
generating constant magnetic fields.
1.3.1
BiotSavart Law
In analogy to the Coulomb force, two constant currents flowing in wires will generate a force known
as the BiotSavart law. For two parallel conductors, each carrying a current, I, we have
dF~12 =
such that
I
o I2 d~l2 r12
~
B(x~1 ) =
2
4 c
r12
I
~ x1 )
F~12 = I1 d~l1 B(~
c
We note that for a point charge in motion with a constant velocity ~v , the BiotSavart law recovers
the form
~
F~mag = q~v B
known as the Lorentz magnetic force.
1.3.2
Vector Potential
~ x) to
We can generalize the BiotSavart law for an arbitrary current configuration, J(~
~ x) = o
B(~
4
Z
V
~0
~ ~0
~ 0 J(x ) (~x x )
dx
~x x~0 3
In magnetostatics, we impose that the charge are constant in time, so that /t = 0, and hence
~ J~ = 0, so we can write
~ =
~ A
~
B
~ x ) = o
A(~
4
~ ~0
~ 0 J(x )
dx
~x x~0 
~ B
~ =0
;
i.e. unlike in electrostatics, there is no single magnetic monopole.
1.3.3
Amperes Law
~ d~l = o
B
J~ d~s =
o Ii
f~(~r) d~l =
~ f~(~r)) d~s
(
~ B)
~ = o J~
(
.
18
1.3.4
Magnetic Dipole
~ in a
Similar to the electrostatic scalar potential, we can expand the magnetic vector potential, A,
l)
(2
multipole series, A (~x)
Z
o X 1
~ 0 J(
~
~ x~0 )Pl (cos)
A(~x) =
dx
4
rl+1 V
l=0
~ ~x
~ (2) (~x) = o m
A
4 r3
where we have defined r ~x and the magnetic dipole moment m,
~ as
Z
m
~ =
V
~ =1
~ (~x)dx
M
2
~
~ x)dx
~x J(~
.
We can show that the component of the magnetic field due to the dipolar term in the vector
potential is
o 1
B (~x) =
3(m
~ ~x) m
~
4 r3
(2)
19
Z
V
~ x J)
~ =I
dx(~
2
with
~x d~l = I Area
J~ d~s
I=
s
and so
m
~ = I Area
Z
m
~ = I d~s
s
X qi
1X
Li
qi (~xi ~vi ) =
2
2m
i
where Li is the angular momentum of charge i, and m is the mass of charge i, and thefore
m
~ =
q ~
L
2m
.
It is easy to show that the force and potential energy of a magnetic dipole in a constant magnetic
~ is given by
field B
~ m
~
F~ = (
~ B)
and
~
U = m
~ B
20
~
=m
~ B
.
1.3.5
Electric fields applied to a material will give rise to a not overall polarization, and likewise magnetic
~ , i.e. local magnetic dipolar moments will be
fields applied in matter will create magnetization, M
~
induced by the external field. The vector potential A in matter will thus contain two contributions:
o
A(~x) =
4

Z
V
~ ~0
~ (x~0 ) (~x x~0 )
J(x )
M
+
~x x~0 
~x x~0 3
{z
} 
{z
}
~0
dx
F ree
Bound
~0
~ (x~0 )
dx~0 M
1
~x x~0 
Z
=
V
~
~
~ M (~x)
dx
~x x~0 
~ M
~
J~ 0 =
and, similarly as in electrotstatics, at interfaces there will be a superficial density of bound current
~0 = M
~ n
K
Amperes law in matter should therefore be rewritten to include the sum of all currents as
~ B
~ = o J~ +
~ M
~
21
~ by the
and similarly as for the electric displacement, we define the magnetic field in a medium, H,
equation
~ =B
~ o M
~
o H
~ is the sum of the applied field, B,
~ plus the effect of the magnetization induced in the medium,
i.e. H
~
M . Amperes law in matter is therefore rewritten as
~ H
~ = J~
~
~
~ = B M
H
o
22
[See also: Cambridge Scientific Minds, eds. Harman and Mitton, ch 1 by Stephen Pumfrey, 2002]
23
~ D
~
~ E
~
~ B
~
~ H
~
=
=
=
=
0
0
J~
o
~ E
~ = 0
~ B
~ = 0
~ B
~ = o J~
~ E
~ =
24
2 Electrodynamics
2.1 Faraday Law
2.1.1
Electromotive Force
E =
~ d~l
E
which has the units of a potential, or volts. When charges are stationary /t = 0, as in electrostatics and magnetostatics, the circulation of the electric field over a closed loop is
E =
~ d~l =
E
~ E)
~ d~s = 0
(
~ E
~ = 0 when E
~ =
~ (Maxwells equations for static electromagnetic fields). In
since
electrodynamics, charges and current are not steady nor stationary and consequently /t 6= 0,
i.e. the charge and current are now time dependent and so will be the electric and magnetic fields.
Hence, if the electric field is allowed to vary in time, the circulation over a closed path will now
gives
~ E
~ 6= 0
i.e. a net potential difference will appear, and the electric field is not irrotational anymore.
2.1.2
Faraday Law
~ d~s =
B
~ d~l
A
~ =
~ A
~ and we have used Stokes theorem. Faraday Law, which is empirical in nature
since B
and dates from 1830, states that:
25
~ d~l E = B
E
t
~ d~l =
E
t
C
~ =
~ ds
B
~
~ E
~ = B
26
~ E)
~ d~s
(
[For more, see: Michael Faraday and the Royal Institution, John Meurig Thomas, 1991]
27
2.2 Inductance
The Faraday law states that an electric field will be induced by a timedependent magnetic field,
or flux, with a loop C. Such a loop, or circuit will be characterized by an inductance L which
will depend on the geometry of the loop.
2.2.1
General Case
Consider a loop C1 in which a current I1 is flowing. In such a loop, the magnetic field generated is
proportional to the current
I ~
o
dl ~r
I1
4
r2
~ 1 I1
B
~1
B
This magnetic field will extend beyond the loop, and could influence other loops located in the
vicinity of loop 1. Consider a second loop C2 adjacent to C1 . On that loop, the magnetic flux going
through the loop (generated by C1 ) is
Z
2 =
~ d~s2 I1
B
so it will be directly proportional to the current I1 . In general we can write the following linear
relation between fluxes on a loop i created by currents j as
(i)
B =
Mij Ij
C2
C1
B2
B1
I1
28
Lets consider the vector ~xi which is a coordinates of a point on a loop Ci . The vector potential at
~ xi ), due to a current in loop Cj is given by
position ~xi , A(~
~ xi ) = Ij o
A(~
4
dlj
~xi ~xj 
Cj
~ d~s =
B
~ d~l
A
o
Mij =
4
I I
d~li d~lj
xi xj 
B = LI
E
= L
dI
dt
dI
L d(I 2 )
P ower = E I V I = L
I=
dt
2 dt
so the total work stored into the circuit, or inductor, will be given by
Z
W =
0
L d(I 2 )
LI 2
dt =
2 dt
2
W =
LI 2
2
In a static system /t = 0, and Faraday law showed us that a timedependent magnetic field will
induce an electric field inside a circuit, or loop. In electrodynamics
6= 0
t
and so the continuity equation, which is always valid
~ J~ + = 0
t
describes that the changes in charges must be compensated by changes in current flow. Lets
consider Gauss Law
~ E
~ = /o
~ J~ + o (
~ E)
~ = 0.
t
30
We also know from Amperes Law that the magnetic field is related to the current density J by
~ B
~ = o J~
and by taking the divergence on both side, we now find the situation where
~ (
~ B)
~ = o (
~ J)
~ 6= 0.
But the divergence of a curl must always be zero. So we are now force to modify Ampere law and
make the following addition:
o J~ o J~ + o o
~
E
t
and
~
~ B
~ = o J~ + o o E .
Now we can easily verify that the divergence of the curl will indeed be zero,
~ B)
~ = o
~ J~ + o o (
~ E
~
(
t
~ ~
~
~
= o J + o ( E) .
t

{z
}
=0
This modification to Amperes law lead to an important consequence on the magnetic field, namely:
A changing electric field induces a magnetic field.
The term
31
~
E
J~D o
t
2.3.2
o
~
~ E
~ = B
t
~
B = 0
~ E
~ =
~
~ B
~ = o J~ + o o E
t
and similarly, we can express them in matter, recalling the definity for the electric field displacement
~ and magnetic field in matter H:
~
D
~ = o E
~ + P~
D
~ P~ = 0
~
~ = B M
~
H
o
~ M
~ = J~0
where 0 is the bound charge density, is the free charge density, such that the total charge is
tot = + 0
, and similarly is J~0 is the bound current density J~ the free current density such that the total
current is
J~tot = J~ + J~0
.
Maxwells equations in a medium becomes
32
~ D
~ =
~
~ E
~ = B
t
~
~
B = 0
~
~ H
~ = J~ + D
For the special case of a medium that is linear and isotropic, i.e that will respond linearly to fields,
and which response will not depend on orientation of the fields, we can write the polarization and
magnetization vector as:
~
P~ = o e E
~ = E
~
D
~ = m H
~
M
~
~ = B
H
where e is the electric susceptibility and m is the magnetic susceptibility. For such
medium, we can then express the the permeativity of the medium by
= o (1 + e )
and its magnetic permeability as
= o (1 + m )
2.3.3
Boundary conditions
33
1 1
2 2
~1 D
~ 2) n
(D
= s
(1)
~1 E
~ 2) n
(E
= s
(2)
~ = (D
~ D
~ dr
~1 D
~ 2) n
A = s A
~ =
~ dl
E
t
C
~ d~s
B
~ d~s
B
lim
0 t
~ d~s 0
B
we obtain the boundary condition for the parallel component of the electric field
~1 E
~ 2) n
(E
=0
~ B
~ = 0 that at the interface between two media of different magnetic
Similarly, we can show from
properties, we have that
~1 B
~ 2) n
(B
=0
(3)
~ B
~ = 0 (no magnetic charge!)
since
~ is discontinuous by K,
~ the superficial density of free current
But, the parallel component of H
~1 H
~ 2) n
~
(H
=K
~ = E,
~ B
~ = H,
~ so the boundary conditions yield:
For linear media, D
~ 1 2 E
~ 2) n
(1 E
= s
~1 E
~ 2) n
(E
= 0
~1 B
~ 2) n
(B
= 0
~
~2
B1 B
~
n
= K
1
2
35
(4)
We recall that both the electric and magnetic fields derive from potentials, in one case a scalar
~ In electromagnetostatics, when = 0 they are
potential and in the other a vector potential A.
t
related to the fields by the following relations:
~ E
~ = ~0
~ =
~
E
~ B
~ =0
~ =
~ A
~
B
and as we have seen previously, the curl of a gradient and the divergence of a curl must always be
(mathematicallly) zero.
But in electrodynamics, once we introduce timedependent sources, we are are now confronted with
the following situation:
~
~
~
~ E
~ = B = ( A)
t
t
and thus the Faraday law reads
~
~ (E
~ + A ) = ~0
t
~
~
which is NOT zero for E = . To remedy to this, we must rewrite the scalar potential, , and
~ and relate them to the fields in the following way:
vector potential, A
~
~ =
~ A
E
t
~
~
~
B =A
~ A)
~
~ E
~ = (
t
~
A
~
~
E + t = ~0
~
~
36
~
A
t
~
A
t
= ~0
since the curl of gradient scalar function is always zero. Importantly, we note that if the potential
vector A is now timedependent, which is true for timedependent magnetic field, it
will act as a potential source for the electric field.
2.4.2
Gauge Transformations
The correspondence between fields and potentials is not unique. In fact, we can always choose a
potential arbitrarily for as long as Maxwells equations are satisfied. We can therefore freely modify
~ or B.
~ This
the potentials by the addition of a function, , without ever modifying the fields E,
transformation to the potentials is called a Gauge Transformation.
~0 A
~ +
~
A
and we show that this transformation does not modify the fields in any way:
i) Scalar potential:
~0
~ 0 =
~ 0 A
E
t
~
A
~
~
~
+
= +
t
t
t
~
~ A
=
t
~
= E
ii) Vector potential:
~0 =
~0
~ A
B
~ (A
~ + )
~
=
:
~ A
~ +
~
~
=
~
= B
Gauge transformations are very important for other fields of physics such as for example in quantum
electrodynamics. This nonuniqueness of potentials imply that we can impose certain conditions
37
~ and B.
~ These are called gauges. There are two gauges that
on potentials without changing E
are particularly useful in classical electrodynamics, the socalled Lorentz and Coulomb (transverse)
gauge:
1.) Lorentz Gauge:
In the Lorentz gauge, we impose the following condition to the potentials:
~ A
~ + o o = 0
~ 0,
If the potentials do not satisfy this condition, we can always find a function, , so that 0 , or A
does satisfy it. Lets transform the potentials for this condition according to a gauge transformation
by the function :
2
0
~ 0 + o o =
~ A
~ + o o + 2 o o
~ A
t
t
t2
~ 0 provided
and so we see that we can always impose the Lorentz condition on the potential 0 and A
that the function satisfied the equation
2 o o
2
~ A
~ + o o }
=
{
t2
t
i.e. all satisfying the wave equation (left hand side) with a source (right hand side) will
be a Lorentz gauge.
2.) Coulomb Gauge:
In the Coulomb, also called transverse gauge, the following condition is imposed on the potential
~
vector A:
~ A
~=0
and we can readily show that in that case, we can always make a transformation so that the
potential vector is in the transverse gauge provided that we solve the Poisson euquation for :
38
~0 =
~ A
~ A
~ + 2
R
~ ~ x)
1
(~x) = 4
dx~0 A(~
~0
~
x x 
o
dx~0
(x~0 )
~x x~0 
2.4.3
In electrodynamics, the introduction of the timedependence for the sources, and hence for the
fields, has linked the electric and magnetic fields in a dynamical way, and through their dependence. We have seen that we can now induce an electric field by varying a magnetic field (Faraday)
and conversely, we can generate a magnetic field by varying an electric field. We shall now see that
~ can be treated very similarly within the Lorentz gauge and
both potential, scalar and vector A,
that in case, they will obey to the same dynamical wave equation.
~ E
~ =
~
~
o
~
A
t
=
o
2 +
~ A)
~
(
=
t
o
39
(*)
~
~
~
~
~
( A) = o J + o o
t
t
and making use of the mathematical identity
~
~ (
~ A)
~ = (
~
~ A)
~
~ 2A
.
Lets impose the Lorentz gauge
~
~ A)
~ + o o
~ 0
(
t
~ obeys to the following equation
an so, in that the case, the potential vector A
~ o o
2 A
~
2A
= o J~
t2
which is simply a wave equation with a source. If we go back to the equation (*), and
~
~ A)
~ = o o
~ we obtain the following equation for the
replacing the Lorentz condition (
t
scalar potential
2 o o
2
=
2
t
o
2 = 2
1 2
2
2
=
o
o
t2
v2 t2
40
where v is the phase velocity of the wave, and which will be discussed in great details in chapter
~ and :
3. We can therefore rewrite the following equations for A,
o
~ = o J~
2 A
2 =
i.e. the same wave equation but with a distinct source in each case.
2.4.4
We now consider moving charges in free space that would be subjected to both an electric and a
~ field with Amperes Law, with that
magnetic field. We are interested in the sum of the resulting E
~
of the B field with the Faraday law. We therefore consider the quantity:
~
~ (
~ E)
~ = B
~ B
B
t
(1)
~
~ (
~ B)
~ = o E
~ J~ + o o E
~ E
E
t
(2)
and
~
~ E)
~ E(
~
~ B)
~ = o E
~ J~
B(
1
o
~2
1 B
2 t
~2
o o E
2
t
2
~
1
B
2
~
~ E)
~ E(
~
~ B)
~
~
~ J~
B(
= 2 t o + o E
E
it gives
41
~2
1 B
1 ~
2
~ J~
~
~
~
(E B) =
+ o E E
o
2 t o
and finally
~2
1 B
1 ~
2
~
~
~
~ B).
~
EJ =
+ o E
(E
2 t o
o
We already recognize on the right hand side the magnetic and electric energy density given by 12 B
o
and 2o E 2 , respectively. We are interested in the total energy stored over the whole volume, so we
can integrate both side of the equation over the whole volume, or sphere, taken at infinity
Z
~ J)d~
~ x = 1
(E
2 t
V
Z ~2
Z
B
1
2
~
~ (E
~ B)d~
~ x.
+ o E d~x
o
o V
V
~ J)d~
~ x =
(E
Ki
t
~
qi~vi E
vi F~i
i.e. it is given by the timederivative of the kinetic energy of the charged particle, or the power.
Making use of that, and of Gauss divergence theorem, we can rewrite the volume integrals as
Z ~2
I
K
1
B
1
2
~
~ B)
~ d~s
=
+ o E d~x
(E
t
t 2
o
o S
which is known as the Poyntings WorkEnergy Theorem. The first term of the equation
describes the total energy stored in the field while the second term is the energy flux that is carried
out. The Poynting theorem states that the
Work done on the charges by the electromagnetic force is equal to the decrease in energy stored in
the field less the energy that flowed out through the surface
42
~ 1E
~ B
~
S
o
~2
1 B
2
~
E=
+ o E
2 o
.
Making use of these notations, the Poynting theorem can be rewritten in a more compact differential form
~ J~ = E +
~ S
~
E
t
The Poynting theorem therefore defines a rate equation for the energy stored inside the fields must
be compensated by the energy flux carried away at a sphere taken at infinity.
It is also usual to define the following quantity for the electromagnetic field
1. Density of momentum for the EM field:
~ = o o S
~
P
i.e. the Poynting vector represents the momentum carried away by the electromagnetic fields
and similarly, we can define
2. Angular momentum density:
~ = ~r P
~ = o (~r (E
~ B))
~
43
3 Electromagnetic Waves
3.1 Waves in Vacuum (General Case)
3.1.1
A wave is simply a disturbance of a continuous medium that propagates with a fixed shape and
a constant velocity.
Let be a function, = (x, t) which depends on position, x, and time, t. In general, one can
show that will be a function which describes a wave propagation if it is a solution of the classical
wave equation:
1 2
2
=
x2
v 2 t2
1
=
u1
x v t
1
=
+
u2
x v t
and so we can write the wave equation as
2
=0
u1 u2
= 1 (u1 ) + 2 (u2 )
or, writing in x, t coordinates
44
This solution, known as the DAlambert Solution is the sum of right traveling and left traveling
waves, which is the most general solution of the wave equation.
3.1.2
Sinusoidal Waves in 1D
= Acos[k(xvt) + ]
where:
k wave vector = 2
v propagating speed
A amplitude
phase
y
v
A
Travelling Wave
45
and
2
= A(vk)2 cos[k(x vt) + ] = (vk)2
t2
Therefore we have
2
= k 2
x2
and
1 2
= k 2
v 2 t2
and so
2
1 2
=
x2
v 2 t2
which is the wave equation.
We call k the wave number (in 1D), or ~k the wave vector (in 2D), and we have
k=
since, when x x + 2/k, the cosine has executed a complete cycle, ie.
2
2
x+
, vt = Acos k x +
vt + = (x, vt)
k
k
Similarly, for a full cycle, we define the period as
T =
2
kv
So that
46
kv
v
1
=
=
T
2
= 2 = kv
We can therefore rewrite the wave equation in terms of wave vector, k, and angular frequency, ,
as
(x, t) = Acos(kx t + )
The inversion of k k transforms the wave from a right travelling to a left travelling wave and
viceversa.
Complex Notation:
Eulers Formula states
ei = cos + isin
and so we can write the sinusoidal solution as
3.1.3
v2
v1
v3
k1
k2
Wave Boundaries/Interfaces
48
T (x, t) = AT e
i(k2 xt)
Note that the frequency, , will be the same for all the waves. Therefore, since = kv, kv must
be constant and so
k1 v1 = k2 v2
giving the result
k2
v1
=
k1
v2
(0 , t) = (0+ , t)
(1)
=
x o+
x o
(2)
(i)
k1 (AI AR ) = k2 AT
or
k2
(AI AR ) = Ar
k1
Subtracting (i)(ii) gives
k2
2AR = AT 1
k1
and adding (i)+(ii) gives
k2
2AI = AT 1 +
k1
giving
AT =
2k1
AI
k1 + k2
and
2AR =
2k1
k1 k2
AI
(k1 + k2 )
k1
AR =
k1 k2
AI
k1 + k2
50
(ii)
v2 v1
AI
v2 + v1
2v2
AI
=
v2 + v1
AR =
AR
Which gives the amplitudes of the reflected and transmitted waves as a function of the time incident
wave front.
3.1.4
Polarization of Waves
displacement
propagation
Wave Propagation
n
z = 0
and
n
= cos
x + sin
y
51
i(kzt)
i(kzt)
t) = Acose
(z,
x + Asine
y
3.1.5
The sinusoidal solution of the wave equation can be written using complex notation as
i(kxt)
= Ae
In fact, any wave can be decomposed into a linear combination of sinusoidal waves:
(x,
t) =
i(kzt)
A(k)e
dk
with = (k)
We consider the Maxwells Equations with no source an in free space (vacuum). Although no
charge sources are present, fields may exist in the form of EM waves, i.e. charge and /t can be
far away, and yet can propagate through the vacuum with constant energy into places where no
charges are present.
Maxwells equations tell us
~ E
~ =0
~ B
~ =0
~ E
~ = B~
~ B
~ =
t
~
o o tE
52
(i)
(ii)
(iii)
(iv)
:0 ~ 2 ~
~
~
~ (
~ E)
~ =
~
(
E)
E
~
B
~
=
t
~ B
~
=
t
~
2E
= o o 2
t
~
2E
2~
~
E = o o 2
t
:0
~ (
~ B)
~ =
~
~
~
(
B)
~
~ 2B
=
~ E)
~
(
= o o
t
2
~
B
= o o 2
t
giving
~ = o o
2 B
~
2B
t2
~ and B
~ fields that obey the
So, we have shown that Maxwells Equations in free space generate E
wave equation
2
~2 1
v 2 t2
53
such that
~ =0
E
~ =0
B
1
= o o
v2
1
c
o o
~ = E
~
D
~ = 1B
~
H
v=
1
c
n
o o
r =
R
o
v<c
54
3.2.2
In last section, we have seen that Maxwells equations in free space provides us with a classical
wave equation for both the electric and magnetic field. The solution of the wave equation will give
us the electromagnetic field as a threedimensional wave, propagating along a specific direction ~k,
the wavevector for the wave, which in magnitude describe the spatial extent of the wave and its
direction will provide the direction of propagation for the wave.
~ x, t) = Re{E~0 ei(~k~xt) }
E(~
~ x, t) = Re{B~0 ei(~k~xt) }
B(~
where the amplitudes E~0 and B~0 are complex (they include the phase shift previously discussed),
but as usual, the solution for the fields must be real.
These amplitudes, if we write ~k = ~kk must obey to the following relations
E~0 k = 0
~k ~
B~0 =
(k E0 )
~ E
~ = 0 and the second from
~ E
~ = B~ .
where the first relation can be readily derived from
t
These relations are important for if we fix the wave vector along the zaxis, i.e. ~k = k
z , then we
must have that E~0 z = 0 = B~0 z, or in another words, the zcomponent of the electromagnetic
field must be zero. The second condition implies that the electric field and magnetic field must
~ B
~ and ~k
be orthogonal between them, and also mutually with the wave vector ~k. Therefore, E,
forms an oriented tryad and the electromagntic field is transverse, the fields oscillate in direction
55
The second relation discussed above also implies that, since in free spave
the fields must related by
= c, the magnitude of
B~0 
1
=
~
c
E0 
which means that the magnetic part of the electromagnetic field will reduced in magnitude by an
amount of c compared to the electric field.
3.2.3
~ x, t) = Re{E
~ o ei(~k~xt) n
E(~
}
~ o ei(~k~xt) k n
~ x, t) = Re{ 1 E
}
B(~
c
E1 = ei E1 
E2 = ei E2 
choosing 1 x
, 2 y and k z then
56
Ex = E1 cos(kz t + )
Ey = E2 cos(kz t + )
k, z
2
2
k, z
2
3.2.4
According to Poyntings Theorem, the energy density and energy flux density for the electromagnetic field in general is given by
1 ~ 2
1
2
~
o E + B
E =
2
o
~ B)
~
~ = 1 (E
S
o
We are now interested in calculating the energy density and Poynting vector but in the case of
monochromatic plane waves propagating in free space. Defining the directing of the electric field
along an arbitrary axis labeled n
, in this case:
~ x, t) = Re{E
~ o ei(~k~xt) n
E(~
}
1 ~ i(~k~xt)
~
B(~x, t) = Re Eo e
kn
c
~ 2
~ o 2 = Eo 
B
c2
which will yield the following
1
B~0 2
2
~
=
o E0  +
cos2 (kx t + )
2
o
1
=
(o E~0 2 + o E~0 2 )cos2 (kx t + )
2
= o E~0 2 cos2 (kx t + ).
This energy density is therefore oscillating periodically as the wave travel in space. We are interested
in the timeaveraged energy density that will be stored in the fields, and for this we need to
calculate the time average over one period, so
since the time averaging of cos2 (x) = 12 . Similarly, the Poynting vector becomes
1 ~ 2 2
Eo  kcos (kx t + )
o c
~ o 2 cos2 (kx t + )k
= co E
~ =
S
~ = cE k
S
.
The momentum density carried by the EM field is given by
~
~ = S = Ek
P
c2
c
3.2.5
Lets consider a region inside matter, i.e. in a medium characterized by , but free of charges
and currents. In this case, the Maxwells equation are:
~ D
~ = 0
~
~ E
~ = B
t
~ B
~ = 0
~
~ H
~ = D
59
If in addition the medium is linear and isotropic, we can rewrite the Maxwells equations using
~ = E
~
D
~ = H
~
B
so that for a homogeneous medium, ie. , 6= f (~x), Maxwells equations reduce to
~ E
~ = 0
~
~ E
~ = B
t
~ B
~ = 0
~
~ B
~ = E .
These are the exact same equations than in free space but with o , o , in the last equations.
Therefore, similarly as in free space, the Maxwells equation will obey a wave equation but with
the the speed of propagation for light in a linear and homogeneous medium given by
1
c
v =
n
r
n
o o
~k
c
n .For
o
o (1 + E )
r = 1 + E
where E is the electric susceptibility for the medium. Similarly, the energy density, E, and
~ will be given in the medium by
Poyntings Vector, S
60
~ =
S
~ 2
1
B
2
~
E +
2
1 ~
~
(E B)
3.3.1
Normal Incidence
We consider first the case where the wave vector of the incident propagating wave k~i is normal to
an interface defined by z = 0. In that case, there will be a reflected and transmitted waves all with
a propagating wave vector perpendicular to the interface.
ki
kr
kt
n1
n2
Lets consider a plane wave going from medium 1 to medium 2 with corresponding indicies of
refraction n1 and n2 . Since there is no free charges or current at the interface, the boundary
conditions will impose that all components of electric field and displacement and magnetic field are
continuous.
Boundary Conditions
~  ,
E
~ ,
B
~  are continuous
H
~ are continuous
D
61
1 Ez (0 ) = 2 Ez (0+ )
~  (0 ) = H
~  (0+ )
H
~  , Bz continuous at z = 0.
E
We consider that the incident plane wave is normal to the interface, and similarly to the transmitted
and reflected wave, i.e.
Ei0 , Er0 , Et0 are the amplitudes of incident, reflected and transmitted waves that are to be determined.
Note:
1. we have supposed an electric field with a polarization following x
2. the angular frequency must be the same for all waves, but, ~k is different since it depends on
the medium properties (n). The frequency is the same for all waves because the discontinuity
is in space, not in time.
Since (in magnitude) the magnetic fields is related to the electric field through B =
corresponding incident, reflected and transmitted magnetic fields are:
62
~k
E,
the
0
~ i = Ei ei(k1 zt) y
B
v1
0
~ r = Er ei(k1 zt) y
B
v1
0
~ t = Et ei(k2 zt) y
B
v2
~ =
where we have used defined the speed of propagation in the medium vi and the fact that B
vi =
1
=
=
i i
ni
k~i 
Er0
Et0
1
=
Ei0
1+
2
Ei0
=
1+
1 v1
1 n2
=
2 v2
2 n1
.
If we further assume 1 2 o then they reduce to
63
~k ~
kE
n1 n2
Ei0
=
n2 + n1
2n1
=
Ei0
n2 + n1
Er0
Et0
Note:
1. v2 > v1 (n2 < n1 ) Reflected waves is inphase with incident
2. v2 < v1 (n2 > n1 ) Reflected waves is out of phase with incident (Er is of opposite sign
with respect to Ei and so out of phase by ).
3. v2 = v1 (n2 = n1 ) No reflected waves, but Et = Ei .
R
Er0
Ei0
Ir
Ii
2
It
2 v2
T
=
Ii
1 v1
.
Et0
Ei0
n1 n2
n1 + n2
2
=
2
4n1 n2
(n1 + n2 )2
These coefficients, R and T, measure the fraction of the incident energy that is reflected and
transmitted at the interface, and we can verify that, owing to energy conversation, they obey the
relation
R+T =1
.
64
3.3.2
Oblique Incidence
In this section, we consider the case where the the wave vector for the incident wave is not normal
to the interface, i.e. there will be an angle i 6= 0 with respect to the normal of the interface.
kr
kt
r
i
ki
n1
n2
1
~i
ki E
v1
1
kr E~r
v1
1
~t .
kt E
v2
Important Notes:
1. We have now that the wave vector ~k makes an angle with the normal vector, n
, of the
interface.
2. The plane defined by ~ki , ~kr , and~kt is called the plane of incidence.
The boundary conditions must be satisfied everywhere at the interface (ie. at z = 0) and so this
~
~ i, E
~r, E
~ t at z = 0. Since all wave
imply that the phase factor ei(k ~xt) must be the same for E
vectors are lying in a same plane, and choosing the direction y perpendicular to the plane, and x
in the plane of incidence, the requirement that at z = 0 all phase must be equal is equivalent to
65
~ki x
= ~kr x
= ~kt x
which means that the xcomponent of the wave vector must be equal at the interface. But (k~i )x =
k~i sini , and similarly for the reflected and transmitted waves, so the condition above becomes:
ki sini = kr sinr = kt sint
.
We recall the dispersion relation for an electromagnetic wave in a medium free of charges and
current
c
(~k)
=
~
n
k
~k
=
c
n
and therefore in magnitude the incident and reflected wave vector must be equal
ki = kr
kt
ki
=
.
n2
n1
As a consequence of the boundary condition imposed at z = 0 which requires that ki sini = kr sinr = kt sint ,
we recover the Law of SnellDescartes:
i = r
and
n1 sin i = n2 sin t
A wave, in general, can be considered as the superposition of two polarizations: one linear polar~ i that is perpendicular to the plane of incidence, and one that is linear with E
~ i inside
ization with E
or parallel to the incident plane. So we can therefore write in general:
~ tot = E
~ + E
~ 
E
66
.
~ i, E
~r, E
~ t for both cases, i.e with the incident electric field having a
Our goal is now to solve for E
polarization that is perpendicular and parallel to the plane of incidence.
i) Perpendicular Polarization
For the perpendiculat polarization, we consider the electric field to be polarized perpendicularly
with respect to the plane of incidence.
kr
Hr
kt
Er
r
i
Et
Ht
Ei
n1
Hi
ki
n2
(1)
1
1
(Ei0 Er0 ) cos i =
E 0 cos r
1 v1
2 v2 t
(2)
~  :
Continuity of H
and defining
1 v1
1 n2
=
2 v2
2 n1
then from equations (1) and (2) we get the Fresnel Equations for perpendicular polarization
cos i cos t 0
E
cos i + cos t i
2 cos i
=
E0
cos i + cos t i
Er0 =
Et0
67
sin(t i ) 0
E
sin(t + i ) i
2 cos i sin t 0
E
'
sin(t + i ) i
Er0 '
Et0
.
(Note that the explicit algebra for Er0 , Et0 is left as an exercise)
ii) Parallel Polarization
In the case of parallel polarization, the electric field of the electromagnetic wave is lying in the
plane of incidence.
kr
Er
kt
Hr
r
i
Ht
Et
Hi
ki
Ei
n1
n2
(2)
and again, using the same definition for as above we can combine these two equations to get the
Fresnel equations for parallel polarization:
68
cos i 1 cos t 0
E
cos i + 1 cos t i
2 cos i
Ei0
=
1
cos t + cos i
Er0 =
Et0
tan(i t ) 0
E
tan(i + t ) i
2 cos i sin t
E0
'
sin(i + t ) cos(t i ) i
Er0 '
Et0
3.3.3
Brewster angles
tan(i t ) 0
E
tan(i + t ) i
In that case, we note that if i + t = /2 then the reflected wave amplitude goes to zero. The
incident angle which satisfies this condition is termed the Brewster Angle, or angle of total
polarization, p
tan p =
=
=
=
sin p
cos p
sin p
sin(/2 p )
n2 sint
n1 sin(/2 p )
n2
n1
tan p =
69
n2
n1
3.3.4
Now consider a situation were n1 > n2 , i.e. as in the case where light passes from glass to air.
n1 sin i = n2 sin t
There is a critical angle, i c where the refracted wave does not seem to exist, i.e. for sin t =
sin /2 = 1:
c = sin
n2
n1
n1 sin i = n2 sin t
from these two equations we get
cos2 t = 1
sin2 i
sin2 i
=
1
(n2 /n1 )2
sin2 c
70
s
1
cost =
sin2 i
sin2 c
sin2 i
sin2 c
~t = E
~ 0 ekt z ei(kt x sin t t)
E
t
.
An incident wave crossing the interface at an angle beyond the critical angle i > c will have its
amplitude decay along the interface over a characteristic length defined by
1
.
kt
We say that there is dispersion when the index of refraction (or ) of the medium is a function
of frequency. In general, n is always a function of (i.e. n = n()), but this dependency can be
more, or less, pronounced.
In a region of the EM spectrum where dispersion is pronounced, a wave packet formed by a
superposition of monochromatic plane waves of nearequal frequencies will have its shape distorted
in time. Because phase velocity is a function of , v = v(), the packet will be dispersed. Similarly,
a wave is more or less refracted following n, thus depending on .
Consider a wave packet, (x, t), which is a continuous superposition of plane waves with amplitudes
~ or B.
~
A(k).
We assume plan wave, and is noted here to denote either E
Z
(x, t) =
dk
~
A(k)ei(k~xt)
2
Wave Dispersion
dk
A(k)eik
2
= A()
(x, t) =
= (x vt, 0)
d
dk
k0
Z
(x, t) =
dk
A(k)eik ei(o vg ko )t
2
= A()ei(o vg ko )t
= (x vg t, 0)ei(o vg ko )t
72
i.e., (x, t) is the product of a wave propagating with speed vg (1st term) and an oscillating function
in time (2nd term). Note that vg , the group velocity, i.e. the velocity of the wave packet as a
whole.
vg
v
Wave Packet
=
n
k
and we define the phase velocity v and group velocity vg for a dispersive medium as:
v =
vg =
3.4.2
k
d
dk
We are now interested in the deriving a microscopic model for the dialectric constant in materials
which shall provide us insights as to whether or not the medium is dispersive and attenuating.
The simplest model is know as the Drude model and it gives us a general form for the frequency
dependence of the dialectric constant in a medium. The model considers the atom as a harmonic
oscillator or as an ensemble of oscillators, i.e. the electron cloud is harmonically bound to
the nucleus with a characteristic frequency .
We want to study the response of such an oscillator upon the application of an oscillating electric
field with frequency .
~ =E
~ o ei(~k~xt)
E
73
electron
cloud
Nucleus
Drude Model
~xo =
~o
e
E
m o2 2 i
74
p~o =
~o
e2
E
.
m o2 2 i
If we now suppose a medium containing several elements with relative importance or fraction f ,
each of which with oscillators frequency and damping per unit volume, the polarization,
P~ = p~ N (~r) is simply given by
P~ =
N f e2 /m
~
E
2 2 i
~ we
and since for a linear and isotropic medium the electric susceptibility is given by P~ = 0 e E,
obtain the frequencydependent susceptibility
e () =
N f e2 /m0
.
2 2 i
We note that the electric susceptibility is now complex, and therefore so will be in that model for
the dialectric constant r = 0 = (1 + e )
r = 1 +
f
N e2 X
.
2
m0 2 i
Note:
~ = E,
~
1. We call the permittivity of the material. In a linear medium, we must have that D
whereas r is called the relative permittivity or dialectric constant of the material.
2. The tilde, i.e r indicates that the dialectric constant is now complex, i.e.
Re(
) + iIm(
)
. Usually, the complex part can be neglected, but if the frequency of the electromagnetic
waves is such that ' , it will play an important role.
75
~ and E
~ since
3. There could now be an out of phase component between D
~ = E
~
D
3.4.3
If the dialectric constant is complex, as it is the Drude model, the electromagnetic wave in the
medium will be attenuated since the wave vector, ~k will now has an imaginary component as well.
We recall that the wave equation for the electric field is
~ = o
2 E
~
2E
t2
and so the plane wave solution will admit the following dispersion relation
k =
p
o
and so the wave vector will now be complex, which we can define as k = k + i and the electromagnetic wave will be attenuated according to
76
N e2 X
f
r ' (1 +
).
k =
2
c
c
2m0 2 i
and it can be easily shown that, separating the real and imaginary part of k that we obtain the
following relation for n, the real part of n
, and the attenuation coefficient :
ck
N e2 X
f (2 2 )
'1+
2m0 (2 2 )2 + 2 2
f
N e2 2 X
.
= 2 '
2
m0 c ( 2 )2 + 2 2
n=
Lets examine the above relations for a single type of elements, near a mode frequency and
attenuation . We can see that the index of refraction n goes to a finite value with 0. It
dn
will then increase up to = 1 < , frequencies at which d
= 0. In that region, the dispersion
dn
is said to be normal, since d > 0. In that regime, blue light is more refracted than red light and
the attenuation of light is relatively small. The attenuation is always positive and will increase
from zero to a maximum at = , and will drop for > . The index of refraction will be
exactly n = 1 at = and will then drop to values less than 1. In that regime, where n < 1,
the phase velocity of the electromagnetic wave will be greater than the speed of light
c. It will then be at a minimum at = 2 > and then goes towards n = 1. In the range of
dn
frequencies 1 < < 2 , the dispersion is said to be anomalous, for d
< 0. The attenuation in
this regime is very large, and the medium is relatively opaque.
3.4.4
In a conductor, there is a density of free electrons that are not bound to the atoms (or lattice)
and so we can model it using the Drude model and setting the resonant frequency = 0. Now,
N is the density of free electrons (#/unit volume). The complex conductivity in a conductor is
therefore
r = 1
N e2 /M 0
2 + i
77
where now the sum over all elements is gone since we are now considering only the conduction
electrons. Starting from the Drude model, we can write the equation of motion for the free
electrons as
+ m ~x = eE
~ o eit
m~x
which has for solution
~x =
~
+(e/M )E
2 +i
~v ~x =
~
(e/M )E
i.
From the definition of the current density J~ = N e~v , we can therefore write
~
N (e2 /M )E
J~ = N e~v =
i
and we define the complex conductivity
as
N (e2 /M )
i
~
J~ =
E.
We shall refer to
as the ACconductivity of a conductor.
Note:
1. In the DC limit, we put = 0, we therefore we obtain the DC conductivity
dc =
N e2
m
~ since
~ That implies that current and
2. J~ is not necessarily in phase with E
C and J~ =
E.
the electric field are not necessarily in phase.
78
3.4.5
Propagation in a Conductor
In free space, we have derived the propagation of electromagnetic waves assuming that the density
of electrons and current density J~ was zero. In a conductor, however, there is always a finite
density of free electrons which, if an electric field is applied, will generates a current density
J~ = ~v where is the density free electrons. This latter is dictated in most cases by Ohms law
~
which states that current density of free electrons must be proportional to the electric field, J~ = E.
The Maxwells equation in that case must be modified to account for these free electrons in the
following way:
~ E
~ = /
~
~ E
~ = B
t
~ B
~ =0
~
~ B
~ = E + E.
~
~ E
~ = . Solving this 1st order differential equation for gives
=
t
(t) = e( )t (0)
or, in other words, the free density of charges will dissipate in a characteristic time and
thus is very short for a good conductor 1/ and very long for a poor conductor
1/. This defines a transient time for the charges to flow to the edge, and so we are interested
in the solution of Maxwells equations at a time t which will be given by
~ E
~ =0
~
~ E
~ = B
t
~ B
~ =0
~
~ B
~ = E + E.
~
and it is easy to show that they lead to the following wave equation for the electromagnetic fields:
79
2~
~
~ 2E
~ = E + E
t2
t
2~
~
~ 2B
~ = B + B
2
t
t
~ t) = E~0 ei(kzt)
and assuming a plane wave solution as usual of the form E(z,
, with k complex, we
obtain the following dispersion relation:
i
k2 = 2 +
where as perviously we have defined k k + i. The electromagnetic field will then be attenuated
as
~ t) = E~0 ez ei(kzt)
E(z,
and so the electromagnetic field will penetrate the conductor over a distance, called skin depth,
corresponding to a decrease in amplitude by 1/e, or
1
.
80
This poem recounts a knights seduction by a vampire. Other titles include: To the Chief Musician
Upon Nabla, an ode to one Professor Tait, who first discussed Hamiltons del operator, (then
known as nabla), and also A Problem in Dynamics, which begins:
An inextensible heavy chain
Lies on a smooth horizontal plane,
An impulsive force is applied at A,
Required the initial motion of K...
[see The Life of James Clerk Maxwell, by Lewis Campbell and William Garnett, 1882, for more]
81
4 Guided Waves
4.1 Transverse Electric and Magnetic Modes
In free space we have shown that Maxwells equations obey a wave equation where the electric,
magnetic and kvector are forming an orthogonal tryad. We will now consider the guided transmission of EM waves in the interior of a hollow pipe. Here, we are interested in a hollow conductor
with walls that are perfectly conducting so that both the electric and magnetic fields must vanish at the material surface. The waves that will propagate inside the hollow pipe will be confined,
and in general such waves will not be transverse, so a component of the electric or magnetic field
along the propagation axis may exist.
4.1.1
We assume a structure with a translational symmetry following the z axis and an arbritary crosssection
z
Hollow Conductor
and we are assuming that the wave is a monochromatic plane wave which propagates along the
zdirection, i.e. ~k = k
z.
~ x, t) = E
~ o (x, y)ei(kzt)
E(~
~ x, t) = B
~ o (x, y)ei(kzt)
B(~
~ o, B
~ o are not a function of z per symmetry argument. Their explicit form is thus E
~o =
i.e. E
o
~ = Bx (x, y)
Ex (x, y)
x + Ey (x, y)
y + Ez (x, y)
z and B
x + By (x, y)
y + Bz (x, y)
z.
~ and B
~ must satisfy the following Maxwells equations
As always, E
82
~ E
~ = 0
~
~ E
~ = B
t
~
~
B = 0
~
~ B
~ = 1 E
c2 t
~ o, B
~ o which only depends on transverse
and the problem is now to solve Maxwells equations for E
variables (x,y).
~ o, B
~ o must satisfy the boundary conditions at the inner wall
We also note that E
~  = 0
E
~ = 0
B
~ o = Ex x
E
+ Ey y + Ez z
o
~
B = Bx x
+ By y + Bz z
and so the goal is to show that if Ez and Bz are known, that the transverse components are
determined too. If that is the case, the problem is reduced to solve for the ttransverse component
only.
From Maxwells equations we have
~
~ E
~ + B = 0
t
~
~ B
~ 1 E = 0
2
c t
Examining the vector components of equation (i) component by component yields
83
(i)
(ii)
~ E)
~ z =
(
~ E)
~ x =
(
~ E)
~ y =
(
and since the partial derivation
Ey
z
Ey
Ex
= iBz
x
y
Ey
Ez
= iBx
y
z
Ex Ez
= iBy
z
x
= ikEy and
Ex
z
= ikEx
Ey Ex
= iBz
x
y
EZ
ikEy = iBx
x :
y
EZ
y :
+ ikEx = iBy
x
z :
z :
x :
y :
By
x
B
= i
Ez
x
y
c2
BZ
ikBy = i
Ex
y
c2
Z
B
+ ikBx = i
By
x
c2
Our goal is now to solve for the transverse components (Ex , Ey ) and (Bx , By ) as a function of
longitudinal components Ez , Bz . From the relations above, we can easily show that they result into
Ex =
Ey =
Bx =
By =
i
Ez
Bz
+
k
(/c)2 k 2
x
y
i
Ez
Bz
k
(/c)2 k 2
y
x
i
Bz
Ez
k
2
(/c)2 k 2
x
c y
i
Bz
Ez
+ 2
k
(/c)2 k 2
y
c x
84
~ E
~ =
~ B
~ = 0, so
But every components of the fields must satisfy to
~ E
~ = Ex + Ey + Ez = 0
x
y
z
Which gives, using the equations for Ex and Ey , and with
Ez
z
= ikEz
2
Ez
i
2 Ez
2
E
2
E
z
z
k
+ ikEz = 0
+k
+
(/c2 ) k 2
x2
y 2
xy xy
and therefore we obtain the following equation for Ez
2
2
2 2
2
+
+ (/c ) k Ez = 0
x2 y 2
2
x2
2
y 2
+ Ez = 0
2
2 (/c)2 k 2
known as the Helmholtz equation. Similarly, its can easily be shown that
2
2
2
+
+ Bz = 0
x2 y 2
The wave guide problem is now reduced to solving the Helmholtz equation for Ez and Bz with
the appropriate boundary condition coming from the in xy plane, and with the conducting (or
dialectric) surface separating the two media. Once the Ez or Bz components are known, all others
are easily determined with the equations above.
Helmholtz euqation + Boundary Conditions Ez , Bz (Ex , Ey ), (Bx , By )
85
In the case of TEM modes, we have Ez = Bz = 0. The Maxwells equation will impose that
~ E
~ = 0 Ex + Ey = 0
x
y
~ E
~ = Bz = 0 Ex Ey = 0
t
y
x
~ o = Ex x
E
+ Ey y has zero divergence and zero curl. It can therefore
~ o =
~
E
where is a scalar potential that satisfies the Laplace equation; 2 = 0. However, the boundary
conditions at surface requires that it is an equipotential and since the Laplace equation 2
admits no local minima or maxima, must be defined constant throughout the hollow conductor
and therefore no wave propagation occurs (the electric field inside! is zero everwhere!).
However, note that TEM modes can exist outside of a hollow conductor. For example, propagating
modes surrounding a conducting wire which would propagate at speed c = /~k
y
Rectangular Wave Guide
4.3.1
TE Modes (Ez = 0, Bz 6= 0)
We first consider the TE modes (Ez = 0, Bz 6= 0) where we must solve the Helmholtz equation for
the magnetic field along z:
2
x2
Bzo = 0
2
y 2
2 =
2
c2
k2
Since the boundary condition can easily be expressed in cartesian coordinates, we can solce the
Helmholtz equation in these coordinates by separating the variables. We write
87
d2 X
d2 Y
+
X
+ 2 XY = 0
dx2
dy 2
1 d2 X
1 d2 Y
+
+ 2 = 0
2
2
X
dx
Y
dy
 {z }  {z }
=kx2
=ky2
kx2 + ky2 + 2 = 0
where kx and ky are constant. So the general solution is of the form
X(x) = Asin(kx x) + Bcos(kx x)
for which we must apply the boundary conditions : first, Bx must vanish at the surface i.e. at
x = 0, a. Recalling that
0 for TE mode
7
1
Bz
E
z
Bx =
k
(/c)2 k 2
x
c y
it implies that at the boundary
Bz
X
=0
= 0 X 0 = Akx coskx x Bkx sinkx x = 0
x
x
The boundary condition at x = 0, since cos(0) 6= 0, implies that
A0
and the the boundary condition at x = a imposes that
kx a = m
kx =
m
a
where m = 0, 1, 2, 3 . . ..
88
By
y
=0
ky =
n
n = 0, 1, 2 . . .
b
We now notice that the k vectors along x and y are now discrete, and are labeled according to
the m, n indices. The complete solution for Bz (x, y) is therefore
ny
mx
cos
m, n N
Bz = B0 cos
a
b
where we define B0 the amplitude of the mode. Finally, to obey the Helmholtz equation, the kx
and ky must obey
m2 n2
+ 2
a2
b
= 2 /c2 k 2
s
2
2
m
n2
2
k=
+ 2 .
c
a2
b
m2 n2
+ 2
a2
b
so that
k=
1p 2
2 .
mn
c
c
c
We immediately see that there is a frequency cutoff mn
for which if < mn
kCk=
Re{k} + i. That means that the wave will not propagate below that frequency cutoff, and instead
c
c , the EM wave and does not
be exponentially damped i.e if mn
the cutoff, then when < mn
propagate.
89
If a > b, the cutoff is for the T E modes is the T E10 mode associated wth the 10 frequency
10 =
c
a
NOTE: Given Bz , we can easily obtain Ex , Ey , Bx , By using the equations given above. We call
this solution the T Emn solution.
4.3.2
TM Modes (Ez 6= 0, Bz = 0)
The approach to finding the TM solutions is the same as for TE modes, but here we set Bz = 0
and we are solving for Ez with the conditions that Ez must vanish at the surfaces. By solving the
Helmholtz equation in a similar way by separation of variables, the solution is
Ez = E0 sin
mn
mx
a
ny
b
sin
q
2
2
= c m
+ nb2
a2
but we must note that here both indices m AND n must be nonzero for waves to propagate.
The lowest frequency of the TM mode will therefore be the T M11 mode.
4.3.3
m2 n2
+ 2
a2
b
Assuming that a > b, then the lowest cutoff frequency is 10 = c/b. The T E10 is therefore the
dominant mode. We typically design waveguides such that only the dominant mode propagates,
and we set its frequencies at the propagating frequencies of the waves that we want to propagate.
A hollow waveguide cane be considered as propagating electromagnetic waves dispersively. Lets
calculate the phase and group velocity for the rectangular wave guide:
Phase velocity:
v
c
=p
>c
2
k
1 ( mn
)
90
c . The phase velocity inside a wave guide is there fore greater than the speed
assuming that > mn
of light.
Group velocity:
r
1
mn 2
vg =
=c 1(
)
dk/d
velocity
The group velocity can be expressed as a simple ratio of the speed of light and the phase velocity,
and we for the rectangular wave guide with perfectly conducting walls, the following relation holds
vg v2 = c2 .
v
c
vg
mn
Group Velocity
Consider the TM modes and set = = 1. We must solve the Helmholtz equation
~ 2 + 2 ]E o = 0
[
z
(*)
xmn
a
where J(xmn ) = 0. ie., xmn are the Bessel function zeros, or the nth roots of the order m Bessel
function. The dispersion relation becomes
2
+ c2 k 2
2 = mn
mn = cxamn
where m 0, n 1.
The general solution for the TM modes can therefore be written as
Ez (r, ) =
o
Emn
Jm
xmn r
cos(m + mn )
a
For a mn mode, m indicates the order of the Bessel function, n are the roots of mth order Bessel
function.
TE modes: Similarly, the TE modes are given with boundary conditions that
and so
0
Jm
(a) = 0 =
ymn
a
where J 0 (ymn ) = 0 and ymn are the nth roots of the derivative of Jm .
92
Bz
x,y
= 0 at r = a,
Bz (r, ) =
o
Bmn
Jm
ymn r
a
cos(m + mn )
2
2 = Nmn
+ c2 k 2
mn = cyamn
Once the Ez (or Bz ) components are known, we can calculate the transverse components of the
field by using ( = = 1)
Er =
E =
Br =
B =
Ez
1 i 1 Bz
+ ik
2 c r
r
1
i Bz
1 Ez
ik
2
c
r
1
i 1 Ez
Bz
+ ik
2
c r
r
1 i Ez
1 Bz
+ ik
2 c
r
What is an EM cavity?
An electromagnetic cavity is simply a conductor closed on all sides, and for which the EM field can
oscillate at some frequencies. No monochromatic wave can travel in it, but, we can have standing
waves at discrete frequencies mnr , forming inside it.
~ B
~
The problem is to find the natural frequencies of the cavity, as well as the configuration of the E,
fields for each oscillating mode.
We consider a rectangular waveguide to which we add two surfaces, one at each end at distance
z = 0, d. Since, we know the solution for an unbounded conductor waveguide, we can consider the
solution as the sum (along z) of an incident plus reflected wave in z direction which interfere with
each other and that of the waveguide along the x and y directions. The interference along z will
create standing waves.
Consider TE modes, ie Ez = 0. In the zdirection, at the new interfaces, the Bz component must
obey to the new boundary conditions along z
93
y
Finite Rectangular Wave Guide
Bz = 0 at z = 0, d
1
(Bz eikz Bz eikz ) = Bz sinkz
2i
Applying the boundary conditions: z = 0 gives simply Bz sinkz = 0 (i.e. this boundary conditions
gives no new information) but at z = d, it gives Bz sinkd = 0, which will be true only when
kd = r, r N
The dispersion relation is, for the wave guide
2
2 = mn
+ c2 k 2
k =
r
d
2
s
mnr =
2
mn
94
rc
d
2
mx
ny
rz
Bz = Bmnr cos
cos
sin
a
b
d
4.5.2
= c
r2
m2 n2
+
+
a2
b2
d2
A priori it seems impossible to excite an EM cavity unless the excitation frequency is exactly that
of the cavity resonance, given by the frequencies mnr . In practice, however, there will be some
energy losses which will make the spectrum of the cavity going from a theoretical delta function
( mnr ) to that of a spectrum of peaks with finite widths owing to the finite conductivity of
the walls.
We define the quality factor, Q of a cavity as
o Energy
Power Loss
d<U >
dt
>
= o <U
Q
so that
~ B
~ 2
< U > E,
E(t) = Eo eo t/2Q
i.e. electric and magnetic fields damped exponentially in time as well. We are interested in whats
happening in the spectrum of frequencies mnr , and to gain that information, we shall Fourier
transform the timedependent field
Z
E() =
E()2
( o
)2
1
+ (/2Q)2
where w0 is the frequency of a given mode under cosideration. This functional form is known as a
Lorentzian curve which is characterized by its width, , at half height, given here by
o
Q
and therefore the larger the quality factor Q is, the more closely will the frequency spectrum around
o resemble a function. To calculate the Qfactor of the cavity, one typically consider the energy
loss for a given mode by the ohmic loss from
1
U=
2
1 ~ 2
2
~
d~x o E + B
o
~ B
~ fields are being damped by the conductors having a finite conductivity at the
where the E,
walls.
96
5 Electromagnetic Radiation
5.1 Fields of Moving Charges
5.1.1
We recall that scalar and vector potentials for electrodynamics are given by
~
~ =
~ A
E
t
~
~
~
B = A
And we have shown previously that they obey a wave equation with a source (In the Lorentz gauge):
o
2~
A = o J~
2 =
with
~ 2 o o
2
Our goal is now to solve these equations when
5.1.2
2
t2
6= 0.
Retarded Potentials
2 = o
and
~
2 A
= o J~
97
(~x) =
~ x) =
A(~
1
4o
o
4
x)
d~x0 ~x(~
~
x0 
~
x
d~x0 ~xJ~
~
x0 
Imagine now that the charges arent static anymore, i.e. so there will be a time lag between
propagation of information which travels at speed of light to the observe. The delay for the
information to travel from the source location (variable x0 ) to the observer (variable x) location
~ and B
~ is known as the retarded time, and is defined as
where we wish to calculate the fields E
tr t
r
c
where
r ~x ~x0 
So for moving charges, the potentials should be generalized as
Z
1
(x~0 , tr ) ~0
(~x, t) =
dx
4o
r
Z ~ ~0
J(x , tr ) ~0
o
~
dx
A(~x, t) =
4
r
where (~x, tr ) is the charge density that prevailed at x~0 and retarded time, tr . These are known as
the retarted potentials, and are reminiscent from the fact that fields are not instantaneous.
Similarly, we can define an advanced time, ta as
ta t +
r
c
(~x, t) =
~ x, t) =
A(~
5.1.3
Z
1
(x~0 , tr ) ~0
dx
4o
~x x~0 
Z ~ ~0
o
J(x , tr ) ~0
dx
4
~x x~0 
LienartWiechert Potentials
We shall now consider the potentials of a point charge moving on a specific trajectory. By point
charge we mean that the charge has no distribution, no size. We define its path or trajectory at a
time, t, by the vector
(x~0 , tr )dx~0 =
q
~v /c
1R
(~x, t) =
~ x, t) =
A(~
qc
1
~ v)
4o (RcR~
o
qc~v
~ v)
4 (RcR~
99
~v
(~x, t).
c2
5.1.4
~v =~(tr ) where as usual ~x is the posiiton of the field observer, (tr ) is the position at time tr . The
r ) must be taken at retarded time t r determined by ~x
~ and~(t
R
~ (tr ) = c(t tr ) = f (~x, t) and
the calculations of the gradient operator is rather cumbersome, so we shall only give the results
here for the fields, and not the derivation. Making use of the notation
~v
~u cR
the fields due to a moving charge are given by
q
R
~ (~u ~a))]
[(c2 v 2 )~u + (R
3
~
4o (R ~u)
~ mpc (~x, t) = 1 R
~ E
~ mpc (~x, t)
B
c
~ mpc (~x, t) =
E
~
where ~a dv/dt
is the acceleration of the point charge at retarded time, tr . We note that
~
~ and to the R
~ (vector to retarded position).
B is perpendicular to E
Note that in the special case where if ~v = 0 = ~a, i.e. when charges are not moving, then ~u cR
and so the electric field is
~ mpc (~x, t)
E
=
=
q Rc3 R
~ ~u)3
4o (R
~ 3
q Rc
4o R3 c3
q R
4o R2
and so
lim
~
u0,~a0
~ mpc (~x, t) =
E
100
qR
.
4o R2
The electric field hat is recovered in the limit of no velocity and no acceleration is that of static
charges, as it should.
~ mpc , given by
The electric field of moving charge E
~ mpc (~x, t) =
E
q
R
~ (~u ~a))]
[(c2 v 2 )~u + (R
~ ~u)3
4o (R
~v , where
effectively contains two terms, one that depends on the velocity of the charges ~u cR
2
2
~v d~
(tr )/dt, and which gives the field in (c v )~u. This term is the generalized Coulomb
field, and is also termed the velocity field Ev 1/R2 . The second term depending on the
~ (~u ~a) is termed the acceleration field
acceleration of the moving charges ~a, and given by R
, Ea 1/R. The fact that the velocity field falls much faster than the acceleration field will have
important consequences on the radiation of moving charges, as we shall see in the next section.
~ mpc since it is given by
Finally, we note that similar results are obtained for the magnetic field B
~ mpc = R E(~
~ x, t).
B
c
Radiation in general
E
t
t 2
o
o
so electromagnetic waves in vacuum propagates out to infinity carrying energy with them.
INSERT FIGURE RadiationWaves
The radiation of charges will be described
ower radiated, or energy loss per unit time,
H in term of p
~ d~s, but the charges will carry some energy with them.
and is given by the Poynting vector P = S
~ ,
To obtain the total power radiated, one must integrate over a sphere at infinity, so that R
and so
I
Prad = lim
~ R)
~ d~s
S(
We have seen that the electric field for moving point charges can be separated into a velocity
~ v and an acceleration field E
~ a . Lets consider first the dependence with respect to the field
field E
observer distance. We note that
101
~ = E
~ +E
~
E
v a
1
1
+
2
r v
r ~a
so that the power radiated at infinity which is the Poynting vector times the surface of a sphere at
infinity with surface S r2 ), which is given by the square of the field will go as
~
S
=
Prad
~v + S
~~a
S
*0 1
1
r2 +
r2
4
2
r
r
~v
~v
So only the acceleration field will radiate at infinity and hence we need only to consider the term
~a =
E
q
R
~ (~u ~a)]
[R
~ ~u)3
4o (R
~
~ a ] = Ea  R
~
R Ea )E
[Ea  R (
Sa =
o c
o c.
~v = c(R
~v /c). For a nonrelativistic moving charge with ~v /c << 1,
Recall that ~u = cR
we can make the approximation
~u cR
or, in other words at a time tr , the charge is almost at rest so that
~ ~u)3 c3 R3
(R
which gives for the acceleration field
~a =
E
=
q
R ~
~a)]
[R (cR
4o c3 R3
q
(R
~a)]
[R
4o c2 R
102
~a)R
~a].
~ a = o q [(R
E
4R
5.2.2
To calculate the total power radiated by a particale at time tr , we need to consider a sphere of
radius R centered at position of the particle at time tr , keeping in mind that the radiation will take
a time (t tr ) = R/c to reach the sphere.
So,
2
~a = Ea  R
S
o c
Ea 2 =
=
=
=
=
2o q 2
~a)R
~a)((R
~a)R
~a)
((R
(4R)2
2o q 2
2 + (R
a)2 )
(~a2 2(~a R)
(4R)2
2o q 2
a)2 )
(~a2 (R
(4R)2
2o q 2 2
a (1 cos2 )
(4R)2
2o q 2 2 2
a sin
(4R)2
And therefore
2
~a = o q a2 sin2 R
S
c(4R)2
103
sin
R
2
I
P
=
=
=
=
=
~sa d~s
Z
2
o q 2 a2
sin2 2
R sindd
(4)2 c
R
Z
Z 2
o q 2 a2
3
sin d
d
(4)2 c 0
0
o q 2 a2 4
2
(4)2 c 3
o q 2 a2
6c
P =
o q 2 a2
6c
which is the Larmor Formula for a moving point charge (v/c << 1)
The Larmour formula is only valid if v/c << 1. As the velocity of the moving charge increases,
~v = c(R
~v /c).
one needs to take into account the v/c that we neglected in ~u = cR
We can show that in this case, the Larmour formula will be generalized by
2
o q 2 6 2 ~v ~a
P =
a
6c
c
1
= p
1 v 2 /c2
P =
0 q 2 6 2
a
6c
104
P =
0 q 2 6 a2
0 q 2 6 a2 1
[1 v 2 /c2 ] =
6c
6c 2
P =
0 q 2 4 a2
6c
5.2.3
In Ritherfords picture of the atom, the electron orbit is in a classical orbit around the nucelas.
Lets consider the simplest possible case, the hydrodgen atom:
INSERT FIGURE: hydrogen
Assuming that the e is orbiting classically, i.e. according to Newtonian mechanics and classical
electromagnetic theory, how long would the e remain on a stable orbit? We have shown that since
~a =
6 = 0 there will be power radiated equal to:
P =
0 q 2 4 a2
6c
P =
1
v 2 /c2
for ~v ~a
' 1, we get:
0 e2 a2
dE
=
6c
dt
Remembering that radiated power is energy lost by the system. Further, we assume that the
energy loss per revolutio nis small compared with the total energy of the atom. In the field, which
is generated by the Coulomb potential of the hydrogen nucleus, the total energy and acceleration
of the electron are:
E=
mv 2
e2
2
40 r
105
a=
v2
r
Recall the virial theorem: 2hT i = nhV i where n = 1 is the power of the potentials rdependence.
Therefore:
hV i
2
hV i
hEi =
2
hT i =
hV i
e2
=
2
80 r
hEi =
e2
40 rm
v2 =
mv 2
e2
e2
=
2
40 r
80 r
and
a=
So that
e2
40 r2 m
e2 dr
0 e2 a2
0 e2
dhEi dr
=
=
dt dt
80 r2 dt
6c
6c
dr
dt
e2
40 r2 m
2
2
o e2
e2
4o r2
=
6c 4o r2 m
e2
o e2
e2
o e4
=
=
2
3c 4o r m
3c(4o )r2 m2
or
3c(4o )m2 2
r dr
o e4
So that the total time the electron will spend in orbit will be:
dt =
Z
=
0
3c(4o )m2
dt =
o e4
r2 dr
a0
Where a0 is the Bohr radius. Were integrating over the radial distance as the electron falls from
its initial radius into the nucleus (r = 0). This gives:
c(4o )m2 3
a0
o e4
106
' 1010 s
4 107 5 1076
1020
,
a very short lifespan indeed!
p0 cos
sin[(t r/c)]
(r, , t) =
4o c
r
~ , t) = o p0 sin[(t r/c)]
A(r,
z
4r
~ =
~ A,
~ the electric and magnetic fields for
~ =
~ A~ and B
Using the field equations E
t
an oscillating electric dipole become:
~ = o p0 cos[(t r/c)]
E
4
2
~ = o p0 cos[(t r/c)]
B
4c
~=
As usual, the power radiated is obtained by integrating the Poynting vector S
Hence:
107
1 ~
~
o E B
at infinity.
2
o p0 2 sin
~
S=
cos[(t r/c)] r
c
4
r
and if we average it in time (remember hcos2 ti = 1/2i),
~ =
hSi
o p20 4 sin2
r
32 2 c r2
Such that the differential radiated power for an electric dipole is:
Note:
i) dP
d
dP
d
=
e.dip.
o p20 4
sin2
32 2 c
5.3.2
We want to study the scattering of EM waves (light!) by free (Thomson) or bound (Rayleigh)
electrons. Picture this process:
INSERT FIGURE scattering
Lets assume the incident wave is linearly polarized with an electric field of the form:
~ =E
~ 0 ei(~k~xt)
E
Recall the Drude model in section 3.3.2: we had developed a theory of the electron gas and found
that the electrons form an oscillating dipole in response to the driving electric field:
p~0 =
~0
E
2
m 02 2 i
p~ = p~0 eit
dP
d
e.dip.
o p20 4
=
sin2 =
32 2 c
dP
d
=
e.dip.
~2
E
o 4
0
sin2
32 2 c (02 2 )2 + 2 2
A fraction of the incident EM wave energy is absorbed by the motion of the electrons and given
back in the form of radiation. We define the differential cross section as the power emitted in a
given direction per unit solid angle per unit of incident energy flux:
d
d
=
e.dip.
(dP/d)
~ i
hSi
with:
~ =
hSi
1 ~
~ = 1 E
~ 0 2 = co E
~ 0 2
hE Bi
o
o c
so,
d
d
~ 2 sin2
E
~ 2 (02 2 )2 + 2 2
32 2 m2 c2 o E
o c2 r02 4 sin2
o 32 2 [(02 2 )2 + 2 2 ]
o e4 4
=
e.dip.
Where we have used e2 /mc2 r0 , the classical radius of the electron ' 3 1013 cm.
d
d
=
e.dip.
o c 2 2 2
4
r
sin
o 32 2 0
(02 2 )2 + 2 2
Z
tot =
d
d
d
o c2 2
4
r0 2
o 12 (0 2 )2 + 2 2
109
tot =
o c2 2
r 6= f ()
o 12 0
tot
o c2 2 4
r
=
o 12 0 0
110