Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Introduction
(4.1)
imposes a constraint between the two quantities (charge density in C m 3 ) and J (the current
density in A m 2 ) and various electrodynamic laws must be consistent with this basic law. A time
varying magnetic ux induces an electric eld through Faradays law,
I
E dl =
d
dt
B dS:
(4.2)
Likewise, a time varying electric eld induces a magnetic eld through the displacement current,
I
d
B dl = 0 "0
dt
C
E dS;
(4.3)
even in the absence of the current J = 0. As is well known, the displacement current played a
crucial role in Maxwells prediction that electromagnetic waves can propagate through vacuum.
The generalized set of Maxwells equations,
r E=
r
E=
(4.4)
@B
;
@t
(4.5)
"0
r B = 0;
r
B=
J + "0
(4.6)
@E
@t
(4.7)
is consistent with the charge conservation law, since the divergence of the LHS of the last equation
indeed vanishes (recall that r (r B) =0 identically) which requires that
r J + "0
4.2
@
@
r E=r J+
= 0:
@t
@t
(4.8)
Faradays Law
In 1831, Faraday discovered that an electric current was induced along a conductor loop when a
magnetic ux enclosed by the loop changed with time. This important discovery gave an answer
to the old question prior to Faradays time whether a magnetic eld could induce an electric eld
because it had been known that an electric eld, via an electric current, could induce magnetic
eld. What Faraday found was that an electric eld (or electromotive force, emf) was induced by
a time varying magnetic ux. The integral form of the Faradays law,
emf =
E dl =
d
dt
B dS;
(4.9)
E=
@B
:
@t
(4.10)
The negative sign is due to Lenz and indicates that the emf is so induced as to oppose the change in
the magnetic ux. It should be noted that many of the experiments originally done by Faraday were
actually due to motional emf in which motion of conductors across a magnetic eld was responsible
for generation of emf without apparent time variation of the magnetic eld itself. An object moving
in a magnetic eld experiences an eective electric eld given by
Eef f = v
B:
(4.11)
This may be seen by noting that change in the magnetic ux enclosed by a loop consists of two
parts, one due to time variation of the magnetic eld,
d
@B
dSdt;
@t
and the other due to the change in the shape of the loop,
d
=
=
=
@S
dt
@t
B (v dl)dt
I
(v B) dldt:
Figure 4-1: Induction of emf in time varying magnetic eld (upper gures) and motional emf in
static magnetic eld (lower gure).
@B
dS
@t
(v
B) dl;
E dl =
@B
dS +
@t
(v
B) dl:
(4.12)
An emf can be generated by letting a conductor move across a stationary magnetic eld as done in
most electric generators.
4.3
@E
;
@t
plays a crucial role for the Maxwells equations to be consistent with the charge conservation law,
@
+ r J = 0:
@t
The magnetic eld induced by the displacement current may be best visualized in a capacitor being
slowly charged as shown in Fig.4-2. The current ows on the surface of electrodes. The radially
Figure 4-2: The surface currents on the capacitor electrode are consistent with the magnetic elds
both outside and inside.
Js1 =
(4.13)
I
2
e ;
(4.14)
(4.15)
is the magnetic eld expected from the Amperes law. In the space between the electrodes, no
conduction current exists but there exists an azimuthal magnetic eld,
B ( )=
1
2
0 "0
@Ez
;
@t
(4.16)
@Ez
:
@t
(4.17)
The magnetic eld is required to exist to satisfy the boundary condition at the inner electrode
surface,
n2 H = Js2 ;
(4.18)
where
I
;
2 a2
Js2 =
(4.19)
is the radially inward surface conduction current on the inner surface of the upper plate. Since
I=
and
Ez =
dq
;
dt
(4.20)
q
;
"0 a2
(4.21)
@Ez 1
:
@t 2
(4.22)
we nd
H =
"0
(4.23)
@Ez
:
@t
(4.24)
H =
"0
As is well known, the displacement current was instrumental for Maxwell to predict that electromagnetic elds obey a wave equation. To see what wave equations the electromagnetic elds
should satisfy, let us take a curl of the Faradays law,
r
r
@B
;
@t
@
r
@t
E =
(r
E) =
(4.25)
B:
(r
E) = r(r E)
r2 E:
Since
r E=
"0
; r
B=
J + "0
@E
@t
1 @2
c2 @t2
E=
1
r +
"0
@J
:
@t
(4.26)
1
"0
:
0
(4.27)
r2
B=
0r
(4.28)
J:
If the displacement current were absent (which, incidentally, is equivalent to the assumption that
c ! 1), both elds would merely satisfy vector Poissons equation,
r2 E =
1
r +
"0
r2 B =
0r
@J
;
@t
J;
1
r +
"0
1
1 @2
c2 @t2
@J
@t
(4.29)
where
1
;
1 @2
2
r
c2 @t2
is the propagator integral operator which yields a retarded solution for the electric eld,
E(r; t) =
=
Z
1
r0
4 "0
Z
1
r
4 "0
(r0 ; t
) 0
dV
0
jr r j
(r0 ; t
) 0
dV
0
jr r j
where
=
jr
@
4 @t
0
r0 j
1
jr
Z
@
r0 j @(t
J(r0 ; t
)dV 0
J(r0 ; t
) 0
dV ;
0
jr r j
In Eq. (4.31),
1
4 "0
(r0 ; t
) 0
dV ; A (r; t) = 0
0
jr r j
4
6
(4.31)
(4.32)
(r; t) =
(4.30)
J(r0 ; t
) 0
dV ;
0
jr r j
r0 j between
are the retarded scalar and vector potential, respectively. For a single charged particle moving at
a velocity vp (t) ; the scalar potential becomes
(r; t) =
=
Z
e
4 "0
e
4 "0 (1
rp t 1c jr r0 j
dV 0
jr r0 j
1
;
n ) jr rp j ret
r0
(4.33)
where
jret means every time dependent quantity is to be evaluated at the retarded time t0 to be
determined implicitly from
1
t t0
r rp t0 = 0:
c
Likewise, the vector potential due to a moving charge is
A (r; t) =
e
4
(1
v
) jr
rp j
(4.34)
ret
These potentials were rst formulated by Lienard and Wiechert. We will use the potentials in
Chapter 8 in formulating the radiation electromagnetic elds due to moving charges.
Equation (4.26) can also be written as
E =
r2
=
r2
1
1 @2
c2 @t2
1
1 @2
c2 @t2
1
r +
"0
@J
@t
1
r +
"0
@Jl
+
@t
@Jt
@t
(4.35)
(4.36)
and Jt = J Jl is the transverse component which does not aect the charge density because of
the identity (or denition)
r Jt = 0:
(4.37)
Since r
Jl = 0; it follows
r(r Jl ) = r2 Jl ;
1 @
r :
r2 @t
(4.38)
1
r +
"0 r2
1
1 @2
c2 @t2
@Jt
:
@t
(4.39)
This formulation is consistent with the choice of Coulomb gauge for the scalar and vector potentials
as we will see in more detail in Chapter 6. Solution for the electric eld is
E(r; t) =
1
r
4 "0
(r0 ; t)
dV 0
jr r0 j
@
4 @t
0
Jt (r0 ; t
) 0
dV :
0
jr r j
(4.40)
Note that in this formulation, the Coulomb electric eld (the rst term in the RHS) is non-retarded
and instantaneous. This unphysical result is a consequence of the Coulomb gauge which singles
out the transverse current Jt . In fact, the non-retarded longitudinal electric led is cancelled by
a term contained in the last term and all physically observable electromagnetic elds are retarded
because of the nite propagation speed c:
4.4
By now, it is clear that there exists two kinds of electric eld, one dened in terms of the scalar
potential,
E1 = r ;
(4.41)
and another originating from time varying magnetic eld,
r
@B
=
@t
E2 =
@
r
@t
A:
(4.42)
@
A
@t
@
Al
@t
@
At ;
@t
where Al and At are the longitudinal and transverse component of the vector potential. Substitution in r E = ="0 yields
@
r E = r2
r Al = :
(4.43)
@t
"0
In Coulomb gauge, Al is so chosen as to satisfy
Coulomb gauge: r Al = 0:
(4.44)
(r; t)
;
"0
(r; t) =
(4.45)
where C is the scalar potential in the Coulomb gauge. It should be emphasized that both C
and are time dependent and in Coulomb gauge the scalar potential responds to a change in the
charge density instantaneously. In Lorenz gauge, r Al is assigned as
Lorenz gauge: r Al +
1 @ L
= 0:
c2 @t
(4.46)
r2
"0
(4.47)
1
r
c
r0 ; t
r0 j
jr
r0
dV 0 :
(4.48)
r0 ; t
1
jr
1
r
c
r0 j
r0
r0
1 @2
c2 @t2
;
1
r
c
r0 ; t
r0
B=
J + "0
@E
@t
1 @2A
c2 @t2
r (r A) +
1 @
c2 @t
0 J:
(4.49)
1 @2A
c2 @t2
1 @2A
c2 @t2
0J
0 Jt ;
1 @
r
c2 @t
since the longitudinal component of the current Jl vanishes through the continuity equation
@
+ r Jl = 0:
@t
9
(4.50)
Note that
@
@
= "0 r E =
@t
@t
and thus
"0
Jl
@
r
@t
"0
@ 2
r
@t
C;
= 0:
r2 A
(4.51)
0 J;
where J = Jl + Jt ; A = Al + At : In Lorenz gauge, the potentials are symmetric in the sense that
both and A satisfy the same wave equation,
r2
1 @2
c2 @t2
=c
A
c
J
(4.52)
and appropriately form a covariant four vector ( =c; A) : In contrast, the formulation of electromagnetic elds in terms of Coulomb gauge is not invariant under the Lorentz transformation. All
potentials and elds are retarded in Lorenz gauge while in the Coulomb gauge, the scalar potential
is non-retarded. The appearance of non-retarded scalar potential is due to the choice r A = 0; or
the assumption that the vector potential is purely transverse.
4.5
E=
"0
@B
;
@t
r B = 0;
r
B=
J + "0
@E
@t
It is noted that the charge density contains all kinds of charges, free charges, bound charges, etc.
Likewise, the current density J contains all kinds of current, conduction currents, magnetization
currents, etc. If free charges are singled out, the rst equation can be written in terms of the
displacement vector D;
r D = free ;
(4.53)
where
D = "0 E + P = "E;
10
(4.54)
with P being the polarization vector, or the electric dipole moment density. The current density
is likewise decomposed into the part due to the motion of free charges and the other due to time
variation of the polarization vector,
@
r P + r Jp = 0;
@t
from which
Jp =
@P
:
@t
(4.55)
B=
J + "0
@E @P
+
@t
@t
J+
@D
@t
(4.56)
M:
If the conduction current is singled out, Eq. (4.56) can be rewritten in terms of the vector H;
r
H = Jc +
where
H=
@D
;
@t
(4.57)
M:
In macroscopic applications, free charges and conduction currents are the quantities that can be
controlled by external means and Eqs. (4.53) and (4.57) are often more convenient than the original
forms.
The Maxwells equations are also consistent with energy conservation law. To show this, we
dene the Poynting vector by
E H; (W m 2 ):
(4.58)
As its dimensions imply, the Poynting vector indicates the ow of electromagnetic energy per unit
area per unit time, that is, power density. The divergence of the Poynting vector is
r (E
H) = (r E) H (r H) E
@D
@B
=
H
Jc +
E
@t
@t
1
@ 1
=
H 2 + "E 2
E J;
@t 2
2
11
(4.59)
respectively. We will return to this problem shortly in the following Section. Integrating over a
volume, we nd
Z
r (E
H) dV =
(E
d
(Um + Ue )
dt
H) dS =
where
Ue =
1
2
"E 2 dV;
Um =
1
2
E JdV;
H 2 dV;
(4.60)
(4.61)
are the total electric and magnetic energies stored in the volume and
Z
E JdV; (J s 1 = W)
(4.62)
is the rate of electromagnetic energy conversion into other forms of energy, e.g., creation of heat
through Joule dissipation and acceleration of charged particles. Therefore,
I
(E H) dS;
(4.63)
S
( E+J
B) dV; (N).
(4.64)
Substituting
= "0 r E
and
J=
1
0
"0
@E
;
@t
we nd
dPm
dt
=
=
where
Z
Z
"0 (r E)E +
!
r T
!
T = Tij = "0 Ei Ej
1 @
E
c2 @t
1
"0 E 2
2
ij
"0
@E
@t
B dV
H dV;
1
0
12
Bi Bj
(4.65)
1
2
B2
0
ij ;
(N m
(4.66)
is the Maxwells stress tensor and use is made of the following vector identities,
rE 2 = r(E E) = 2E (r
E) + 2 (E r) E;
r (EE) = (r E)E + (E r) E
rB 2 = r(B B) = 2B
(r
B) + 2 (B r) B
r (BB) = (r B)B + (B r) B = (B r) B:
Eq. (4.65) suggests that the vector
1
E
c2
H; (N s m
(4.67)
H; (N m
(4.68)
(E
H); (J s m
(4.69)
(4.70)
Its ux density is
1
R= r
c
(E
H); (J m
):
(4.71)
Both the momentum and angular momentum densities are proportional to the Poynting vector,
namely, the energy ow. If for example a system is losing energy through radiation of electromagnetic energy, the system is necessarily losing momentum and angular momentum as well. Consider
a charged particle undergoing circular motion, e.g., electron in a magnetic eld. Since the electron is continuously accelerated by the centripetal force, it radiates electromagnetic energy. At
the same time it loses its angular momentum to radiation. Therefore, it is natural to expect that
electromagnetic elds (with proper polarization) carry an angular momentum with them.
13
4.6
One special mode of electromagnetic waves in free space is a plane wave in which the amplitude
of electric and magnetic eld remains constant. Without loss of generality, we may assume wave
propagation in the z direction and an electric eld in the x direction,
E(z; t) = E0 ei(kz
!t)
ex :
1 @2
c2 @t2
provided
E(z; t) = 0;
(4.72)
!
= c:
k
From
r
E=
(4.73)
@B
;
@t
we nd
E = !B;
(4.74)
k
E0 ei(kz
!
!t)
ey ;
(4.75)
E0
E0
E0
; or H0 =
;
=
c
c 0
Z
where
Z=
"0
(4.76)
= 376:8; ( )
(4.77)
0H
(4.78)
This equipartition of wave energy is similar to that in mechanical waves in which kinetic and
potential energies are equal. The total wave energy density is therefore given by
u=2
1
"0 E 2 = "0 E 2 ; (J m
2
14
(4.79)
and the Poynting ux may be written in terms of either the electric or magnetic eld as
Sz = Ex Hy = c"0 Ex2 =
Ex2
; (W m
Z
(4.80)
or
Sz = c
2
0 Hy
= ZHy2 :
(4.81)
For a harmonic wave with an amplitude E0 , the average (rms) wave energy density is given by
1
uave = "0 E02 ; (W m
2
(4.82)
):
(4.83)
@B
;
@t
E=
B=
@("E)
;
@t
(4.84)
where " is the permittivity which in general depends on the wave frequency and spatial position
and also the electric eld. The origin of the permittivity is in the current induced by the electric
eld in a material medium. In the magnetic induction equation
r
B=
J + "0
@E
@t
if the current density is proportional to the electric eld through a conductivity ; J = E; we have
r
B=
E + "0
@E
@t
0"
@E
;
@t
(4.85)
!"0
(4.86)
It should be noted that the current density is due to deviation of electron orbit from bound harmonic
15
motion in molecules. The equation of motion for an electron placed in an oscillating electric eld is
m
d2
+ ! 20 x =
dt2
eE0 e
i!t
(4.87)
where m is the electron mass and ! 0 is the frequency of bound harmonic motion. The current
density is
dx
i! eE
J = ne
= 2
;
dt
!
! 20 m
and the conductivity becomes
=
i!
!2
ne2
;
! 20 m
(4.88)
! 2p
! 20
!2
(4.89)
ne2
:
"0 m
(4.90)
!
1
=p
k
"(!)
d!
=
dk
(4.91)
!
1
:
1 ! d" k
1+
2 "(!) d!
(4.92)
(At frequencies remotely separated from the resonance frequency ! 0 ; the group velocity coincides
with the energy propagation velocity. Near the resonance, however, the group velocity exceeds c and
it loses the meaning of energy propagation velocity. The concept of signal velocity was introduced
by Brillouin.) The impedance is accordingly modied as
E
Z(!) =
=
H
"(!)
16
(4.93)
"(!)
E 2 ; (W m
):
1+
1 ! d"
2 "(!) d!
"(!)E 2
1 d
1
[!"(!)]E 2 +
H2
2 d!
2 0
1
1
1 ! 2p (! 2 + ! 20 )
"0 E 2 +
"0 E 2 +
2
2
2
2 (!
!0)
2
0H
; (J m
0H
(4.94)
is used. This result is valid only if the group velocity can be regarded as energy propagation velocity
which may not hold near the resonance ! ' ! 0 if the dielectric is dissipative.
The origin of the additional factor
1 d" 2
! E ;
2 d!
in the electric energy density is due to electron kinetic and potential energies in an oscillating
electric eld. From the equation of motion of a bound electron,
@2
+ ! 20 x =
@t2
e
E;
m
(4.95)
(4.96)
Therefore, the total energy density associated with the electric eld is
1
1 ! 2 ! 2p
1 ! 20 ! 2p
2
"0 E 2 +
"
E
+
"0 E 2 ; (J m
0
2
2 (! 2 ! 0 )2
2 (! 2 ! 0 )2
(4.97)
(4.98)
Since the wave under consideration is strongly dispersive, there is no simple energy equipartition
as in the case of nondispersive waves. Note that the electron kinetic and potential energies are the
17
@D
@ ("E)
=E
;
@t
@t
(4.99)
(4.100)
can be seen if we realize that the permittivity " is in general frequency dependent. Waves become
dispersive and wave amplitude is bound to decrease although slowly. In this case,
@ ("E)
@t
i!" (!) E0 e
i!t
' ( i! 0 + ) " (! 0 ) + i
=
where
i! 0 " (! 0 ) E0 e
i!t
@"
E0 e
@! 0
+ " (! 0 ) + ! 0
i!t
@" dE0
e
@! 0 dt
i!t
(4.101)
(4.102)
@ ("E)
1
@" dE 2
=
" (! 0 ) + ! 0
;
@t
2
@! 0 dt
(4.103)
1
@"
" (! 0 ) + ! 0
E2:
2
@! 0
(4.104)
1
2
(! 0 ) + ! 0
@
H 2:
@! 0
(4.105)
4.7
Reection and transmission of plane electromagnetic waves at a boundary of two dielectrics can be
conveniently analyzed in terms of impedance mismatch. Let a plane wave of amplitude Ei in air be
incident normally on a at dielectric surface of impedance Z2 as shown in Fig.4-3: The impedance
of air is very close to the vacuum impedance, Z1 = 377 : The incident Poynting ux is split into
18
):
(4.106)
At the boundary, continuity of the electric elds, which are all tangential to the surface, yields
Ei + Er = Et :
(4.107)
Z2 Z1
Ei;
Z2 + Z1
(4.108)
Et =
2Z2
Ei :
Z2 + Z1
(4.109)
(4.110)
Hi + Hr = Ht ;
(4.111)
yield
Hr =
Z1 Z2
Hi;
Z1 + Z2
Ht =
2Z1
Hi :
Z1 + Z2
(4.112)
Note that the polarity of either electric or magnetic eld of the reected wave is reversed. This
is understandable because the incident and reected elds must satisfy the following vectorial
relationships,
k Ei = ! 0 Hi ;
19
Er = !
0 Hr :
Either Er or Hr must change sign on reection for the reected Poynting ux to be in opposite
direction relative to the incident ux.
It should be pointed out that the analysis presented above is solely based on energy conservation and no consideration was given to momentum conservation. In fact, if only the momenta
associated with the three waves (incident, reected, and transmitted) are considered, momentum
is not conserved. Since the momentum ux densities associated with each wave are
1 Ei2
= "1 Ei2 ;
c1 Z1
Reected wave : "1 Er2 ;
Incident wave :
(4.113)
"2 Et2 =
2(Z12 Z22 )
2(Z1 Z2 )
"1 Ei2 =
"1 Ei2 :
2
(Z1 + Z2 )
Z1 + Z2
(4.114)
This is actually taken up by the dielectric body as mechanical momentum. Recall that an innitely
massive body can absorb momentum without absorbing energy. If Z2 > Z1 ; the dielectric is pushed
to the left. In fact, at the boundary, the electric energy densities are discontinuous, and the
dierence is
1
"2 Et2
2
1
2Z22
"1 Et2 = ("2 "1 )
E2
2
(Z1 + Z2 )2 i
2(Z12 Z22 )
=
"1 Ei2
(Z1 + Z2 )2
2(Z1 Z2 )
=
"1 Ei2 :
Z1 + Z2
This appears as a force per unit area on the boundary surface acting from the higher energy density
side to the lower because electric force in the direction perpendicular to the eld appears as pressure.
Reection at a conductor surface can be analyzed in a similar manner by modifying the impedance appropriately. For a medium having a conductivity ; the impedance is given by
Z=
i! 0
:
i!"0
This can be seen from the Maxwells equation in the presence of conduction current,
r
@E
@t
i!"0 E;
H = J + "0
=
20
(4.115)
i!
"e
i! 0
:
i!"0
For ordinary conductors, the conduction current far dominates over the displacement current even
in microwave frequency range. Then,
Z'
i!
= (1
r
!
i)
2
(4.116)
p
and electromagnetic waves
is much smaller than the free space impedance Z0 =
0 ="0 ' 377
incident on a conductor surface suers strong reection. However, reection can never be complete.
Bath room mirrors coated with aluminum has power reection coe cient of about 90% at optical
frequency ! ' 1014 rad/sec.
Example 1 Impedance Matching
A dielectric lm a quarter wavelength thick coated on a surface of another dielectric (say, optical
glass) can eliminate reection of electromagnetic waves normally incident if the impedance of the
lm is chosen to be a geometric mean,
Zf =
p
Z0 Zg :
(4.118)
This condition follows from cancellation between two reected waves, one at the air-lm boundary
and another at the lm-glass boundary,
Zf Z0
Zg Zf
+ ei
= 0;
Zf + Z0
Zg + Zf
f
21
for Zf ; we nd
p
Z0 Zg :
Zf =
Likewise, reection from a conductor plate can be avoided by placing a thin conducting plate
of thickness d at a quarter wavelength in front of the conductor surface if the conductance of the
plate is chosen to be
1
=
:
(4.119)
Z0 d
4.8
For an arbitrary incident angle 1 ; reection and transmission at a dielectric boundary can be
analyzed by exploiting the boundary conditions for the electric and magnetic elds as follows. An
incident wave continues to be assumed plane polarized. The refracted angle 2 is related to the
incident angle 1 through the well known Snells law,
sin
sin
n2
;
n1
(4.120)
p
p
where n1 = "1 ="0 and n2 = "2 ="0 are the indices of refraction of respective media. The Snells
law follows from the conservation of the wavenumber parallel to the boundary,
k1 sin
= k2 sin
2;
(4.121)
!
1
=p
k2
"2
;
0
(4.122)
k2
k1
=
:
n1
n2
(4.123)
= k2
n1
n2
sin2
1:
>
n2
; n2 < n1 ;
n1
This is the condition for total reection which occurs when a wave is incident on a medium with a
smaller index of refraction (e.g., from glass to air), n2 < n1 : An imaginary wavenumber indicates
exponential damping in the region n2 from the surface,
E0 e
kz z
22
kz = k2
n1
n2
sin2
1:
n1
sin :
n2
H=
@D
; or k
@t
H=
!D;
E=
@B
; or k
@t
E = !B;
4.8.1
We rst consider the case in which the magnetic eld of the incident wave is in the incident plane,
or the electric eld is parallel to the boundary surface. The reected and refracted electric elds
are also parallel to the surface and the continuity of the tangential component of the electric eld
gives
Ei + Er = Et :
(4.124)
23
Figure 4-4: A plane wave with incident angle 1 when the magnetic eld is in the incidence plane
and the electric eld is parallel to the boundary surface.
0 !H;
(4.125)
1 (Ei
Er ) = k2 cos
2 Et :
(4.126)
sin(
sin(
Et =
2 cos
sin(
sin(
sin(
1)
2
1
2)
Ei '
sin
sin
1)
1+
2)
1 sin 2
1
2)
1 ; 2 ),
sin
1 + sin
2
where
Zi =
Ei ;
Ei :
(4.128)
we recover
Ei =
"i
(4.127)
n 1 n2
Z2 Z1
Ei =
Ei ;
n2 + n1
Z2 + Z1
(4.129)
(4.130)
24
E2
H2 cos
r
i
=
2
"2
n1
n2
sin2
(4.131)
is also pure imaginary which indeed ensures total reection of electromagnetic waves,
Z1
i Z2TE
= 1:
TE
Z1 + i Z2
In total reection, the phase dierence between the incident and reected waves is
TE
4.8.2
= 2 tan
Z1
Z2TE
= 2 tan
sin2
(n2 =n1 )2
cos
(4.132)
Figure 4-5: The electric eld is in the incidence plane and the magnetioc eld is paralllel to the
boundary.
In this case, the magnetic eld is parallel to the boundary plane and the continuity of tangential
component of the magnetic eld is simply
Hi + Hr = Ht :
(4.133)
For each wave, the electric eld is related to the magnetic eld through
E=
1
k
!"
25
H:
(4.134)
1 (Hi
1
cos
n2
Hr ) =
2 Ht
1
cos
n2
2 (Hi
+ Hr ):
(4.135)
Solving for Hr ; we nd
Hr =
=
sin 1 cos
sin 1 cos
tan( 1
tan ( 1 +
sin
+
sin
1
2)
Hi :
2)
2 cos 2
2 cos 2
Hi
(4.136)
(4.137)
1+
tan(
tan (
2)
1+
2)
Hi =
tan(
tan (
1)
2
1
2 sin 1 cos
1 + 2 ) cos(
sin(
2)
1
1
2)
Hi :
Ei ;
(4.138)
2 sin 2 cos
sin( 1 + 2 ) cos(
1
2)
Ei :
(4.139)
If
1
(4.140)
the reected wave vanishes completely. Under this condition, Snells law becomes
n2
sin
=
n1
sin
sin
cos
= tan
1;
(4.141)
= tan
n2
n1
(4.142)
is called the Brewsters angle. If incident wave is plane polarized with the magnetic eld oriented
parallel to, say, a surface of glass, reection can be avoided at the Brewsters angle. This principle
is often exploited in designing reecting mirrors in lasers so that output laser beam has a high
degree of planar polarization.
Because the reection coe cient depends on wave polarization, unpolarized wave with random
polarization becomes partially polarized on reection and transmission at a dielectric boundary.
The magnetic reection coe cient derived in Eq. (4.136) will play an important role in analyzing
the transition radiation discussed in Chapter 8.
When the conditions for total reection are met, the impedance dened by the ratio between
26
E2 cos
=
H2
r
i
"2
n1
n2
sin2
1:
(4.143)
4.9
= 2 tan
Z1
Z2TM
= 2 tan
!
p
sin2 1 (n2 =n1 )2
:
(n2 =n1 )2 cos 1
(4.144)
Planar polarization discussed in the preceding section is a highly idealized mode of propagation of
electromagnetic waves. A plane polarized wave can be decomposed into two circularly polarized
plane waves of opposite helicity, one with positive helicity and another with negative helicity.
Helicity of an electromagnetic wave is closely related with the angular momentum carried by the
wave.
Circularly polarized waves propagating in the z-direction can be described by the electric eld
vectors,
E (z; t) = E0 (ex iey )ei(kz !t) ;
(4.145)
where the positive sign is for positive helicity and minus sign is for negative helicity. The sum of
these two waves of opposite helicity trivially yields a plane polarized wave with the electric eld in
the x-direction. Corresponding magnetic elds are
B
=
=
1
k E
!
k
E0 (ey
!
iex )ei(kz
!t)
(4.146)
and Poynting ux is
S=E
H =2
E02
ez ;
Z
(4.147)
where the factor 2 accounts for the two independent modes of equal amplitude.
A general form of mixed helicity may be written as
E(z; t) = (E1 ex + E2 iey )ei(kz
!t)
(4.148)
k
(E1 ey
!
E2 iex )ei(kz
!t)
(4.149)
27
(4.150)
Figure 4-6: Trace of the head of the rotating electric eld vector associated with a circularly
polarized wave with positive helicity. In the case of negative helicity, the direction of rotation
relative to the wavevector is reversed.
tan(
tan(
1)
2
1
Ek0 ;
2)
E?r =
1)
2
2
1)
E?0 :
2 sin 1 cos
sin( 1 + 2 ) cos(
E?t =
2 cos
sin(
2
1
1 sin 2
1
2)
2)
Ek0 ;
E?0 :
Both reected and transmitted waves are elliptically polarized. In particular, if the incident angle is
at the Brewsters angle, 1 = =2 2 ; Ekr vanishes and the reected wave becomes plane polarized.
28
On total reection of a circularly polarized electromagnetic wave, the reected wave becomes
elliptically polarized because the phases of TE and TM components dier as evident from Eqs.
(4.132) and (4.144).
4.10
StokesParameters
Consider an electromagnetic wave propagating in the z direction with electric eld components
E(r; t) = (Ex ex + Ey ey: )ei(kz
!t)
(4.151)
The amplitudes Ex and Ey may be complex allowing for nite phase dierence,
Ex = jEx j ei x ; Ey = jEy j ei y :
(4.152)
is nite,
If = x
y = 0; the led is a simple superposition of two linearly polarized waves. If
the wave is in general elliptically polarized. In optics, direct measurement of the phase dierence
is not easy. What is normally measured is the intensity or the quadratic quantities of the electric
eld,
jEx j2 ; jEy j2 ; jEx j jEy j cos( x
(4.153)
y ); and jEx j jEy j sin( x
y ):
Stokesparameters are dened, with I = jEx j2 + jEy j2 the total intensity, by
s0 =
s1 =
s2 =
s3 =
1
I
1
I
2
I
2
I
jEx j2 + jEy j2 = 1;
jEx j2
jEy j2 ;
x
x
(4.154)
y );
y );
and satisfy
s21 + s22 + s23 = 1;
(4.155)
2
2
Re Ex Ey ; s3 = Im Ex Ey ;
I
I
(4.156)
(4.157)
Natural light is characterized by a collection of many waves with random phases and complete
depolarization, s1 = s2 = s3 = 0 even though each wave may be highly monochromatic. For a
plane wave polarized in the x direction,
s1 = 1; s2 = s3 = 0:
29
s1 = s3 = 0; s2 = 1:
For a circularly polarized wave with positive (negative) helicity,
s1 = s2 = 0; s3 =
1:
sin ) jEx je
cos + jEy je
sin
y ):
1
jEx j2 + jEy j2 + 2jEx jjEy j cos(
2
y)
(4.158)
Finally, exploiting a uniaxial (or biaxial) crystal, a so-called quarter wavelength plate can be fabricated to induce =2 relative phase delay between Ex and Ey due to the dierent propagation
velocities of ordinary and extraordinary modes. The intensity in this case is
I 0 ( ) = (jEx jei
sin ) jEx je
cos
ijEy je
sin
y)
1
jEx j2 + jEy j2
2
y)
(4.159)
I ( = 0) ; I
; I
and I 0
(4.160)
is called polarization tensor. Unless the eld is purely coherent, the time averaged intensity Ei Ej
may still vary slowly with time. Iik is Hermitian and thus can be diagonalized through the eigen-
30
values
and
ij )
= 0:
(4.161)
Corresponding two polarization eigenvectors n(1) and n(2) can be determined from
(1)
Iij nj
(1)
1 ni ;
(2)
Iij nj
(2)
1 ni :
(4.162)
In terms of the eigenvectors, the polarization tensor can be written in the form
!
I =
(1) (1)
1n n
(2) (2)
2n n :
(4.163)
As an example, let us consider superposition of two plane polarized waves, one in the x direction
and another in the direction (cos ; sin ) in the x y plane. The relative phase between the two
waves is assumed to be random and the intensities of the waves are I1 and I2 : The electric eld is
E = (E1 + E2 ei cos ; E2 ei sin );
where I1 = E12 ; I2 = E22 ; and
Iij =
Eigenvalues are
1;2
1
=
2
min =
I1 + I2
max
I2 cos sin
I2 sin2
q
(I1 + I2 )2
The ratio
=
(4.164)
I1 + I2
I1 + I2 +
p
p
(4.165)
4I1 I2 sin2
(I1 + I2 )2
4I1 I2 sin2
(I1 + I2 )2
4I1 I2 sin2
(4.166)
(4.167)
=1
cos ;
(4.169)
1 cos
:
1 + cos
(4.170)
cos ; sin
2
2
; n(2) =
31
sin ; cos
2
2
(4.171)
ij
3
X
m=1
where
(1)
ij
(m)
sm ij
1
0
0
1
(2)
ij
1 + s1 s2
s2 + is3 1
1
= I
2
0 1
1 0
(3)
ij
0
i
is3
s1
i
0
(4.172)
(4.173)
1 0
0 0
1
Iij = I
2
1
i
i
1
= =4: When s3 = 1; s1 = s2 = 0;
1; s1 = s2 = 0 and
4.11
In this section, we analyze propagation of electromagnetic waves along a conductor rod of radius
a and conductivity : A simple transverse Magnetic (TM) mode is considered with the following
eld components, E ; Ez ; and H : The axial electric eld Ez outside the conductor rod satises the
scalar wave equation,
@2
1 @
@2
+
+
@ 2
@
@z 2
Ez ( ; z; t) =
32
1 @2
Ez ( ; z; t):
c2 @t2
(4.174)
Figure 4-7: Field proles of electromagentic wave propagating along a conductor rod.
i!t ;
this reduces to
@2
1 @
@2
!2
+
+
+
@ 2
@
@z 2
c2
Ez ( ; z) = 0:
(4.175)
Furthermore, since the wave is propagating along the rod, we may single out the z dependence in
the form eikz and reduce the wave equation to
d2
1 d
+
+
2
d
d
where
2
!
c
Ez ( ) = 0;
k2 :
(4.177)
H0 (
) = J0 (
(4.176)
iN0 (
);
); N0 (
known as Hankel functions of the rst and second kind, respectively. Their asymptotic behavior at
large argument is
r
h
i
2
(1)
H0 (x) !
exp i x
;
x
4
r
h
i
2
(2)
H0 (x) !
exp i x
;
x
4
indicating radially outward and inward propagation, respectively. In the present case, both k
and are complex because of dissipation in the rod. It is reasonable to assume radially outward
33
Ez ( ) = E0 H0 (
):
(4.180)
E ( )=
and from r
E = i!
0 H;
);
(4.181)
(1)
E0 H1 (
i!"0
E0
(1)
H1 (
);
(4.182)
d
(1)
(1)
xH1 (x) = xH0 (x);
dx
d (1)
(1)
H (x) = H1 (x):
dx 0
Z=
i!
(surface impedance);
(4.183)
(1)
H0 ( a)
(1)
H1 ( a)
= i!"0 Z:
(4.184)
For a given wave frequency !; this equation determines and thus k; the axial wavenumber and
in this respect, it is a dispersion relation.
For a small impedance Z; approaches 0, and the axial wavenumber k becomes
k=
!
;
c
as expected. Noting
2
(1)
lim H0 (x) ! 1 + i (ln(x=2) +
x!0
E) ;
(1)
lim H1 (x) !
x!0
21
;
x
(4.185)
we see that the axial electric eld becomes negligible and the transverse elds approach those of
TEM mode,
E ( )
a
E ( ) = E0 ; H ( ) =
;
(4.186)
Z0
p
where Z0 =
0 ="0 is the impedance of free space. The elds in coaxial cables used for transmission
of electromagnetic waves can be approximated by those given above,
E =
V
1
;
ln(b=a)
H =
34
E
I
=
;
Z
2
(4.187)
where V is the potential dierence between the inner and outer conductor with radii a and
p
b;respectively, I is the current and Z =
0 =" with " the permittivity of the insulating material lling the cable. The characteristic impedance of the cable is
Zcable
V
=
=
I
0 ="
b
a
ln
(4.188)
Similarly, the impedance of a parallel wire transmission line with conductor radius a and separation
distance D is approximately given by
Z'
4.12
0 ="0
D
a
ln
a:
(4.189)
In a conductor, the conduction current dominates over the displacement current and a simple Ohms
law,
J = E;
(4.190)
may be assumed where
Maxwells equations
@B
; r
@t
E=
B'
E;
@B
:
@t
If the eld is oscillating at a frequency !; we obtain the following ordinary dierential equation,
r2 B + i!
B = 0:
By = 0:
(4.191)
i!
1 + ip
= p
!
2
(4.192)
The magnetic eld decays exponentially from the conductor surface and the quantity
=
2
!
35
; (m)
(4.193)
is called the skin depth. Damping of the electromagnetic elds is evidently due to Ohmic dissipation
in the conductor.
For a cylindrical conductor rod with radius a( c=!) in which an axial current Jz ( )e i!t is
excited, Jz obeys
1 d
d2
+
+ i! 0
Jz ( ) = 0;
(4.194)
2
d
d
This has a bounded solution
Jz ( ) = E0
J0 (k )
;
J0 (ka)
(4.195)
where E0 is the electric eld at the rod surface. The impedance per unit length of the rod can thus
be dened by
Z
E0
k J0 (ka)
=
=
; (Ohms/m):
(4.196)
l
I
2 a J1 (ka)
where use is made of
I=2
a
J1 (ka)
E0
:
k
J0 (ka)
Jz ( ) d = 2
(4.197)
J0 (x) ' 1
x2
x
; J1 (x) '
4
2
x2
8
i!
(4.198)
Li
= 0 (H/m)
l
8
(4.199)
is the well known internal inductance of a cylindrical rod carrying a uniform current (no skin eect).
In the high frequency limit (strong skin eect), the ratio J0 (ka)=J1 (ka) approaches i; and the
impedance is
Z
1
=
(1 i):
(4.200)
l
2 a
This is also reasonable since with strong skin eect, the current ow is limited in a thin layer with
an area 2 a : Note that the inductive reactance is identical to the resistance in this limit.
If a conductor cylinder is placed in an oscillating axial magnetic eld Hz0 e i!t as in inductive
rf heating, the current ows in the azimuthal direction. J ( ) satises
1 d
d2
+
2
d
d
1
2
+ i!
J ( ) = 0;
< a:
An appropriate solution is
J ( )=k
J1 (k )
Hz0 ;
J0 (ka)
36
< a;
(4.201)
4.13
i!
A charge neutral plasma contains equal amount of positive and negative charge density. When
placed in an electric eld oscillating at a high frequency, a plasma current due to electron motion
is induced. The conductivity can be found by letting ! 0 = 0 in Eq. (4.88) (because in a plasma,
electrons are free),
i ne2
=
;
(4.202)
! m
and corresponding permittivity
" = "0
! 2p
!2
:
c;
! 2p
! (! + i c )
(4.203)
this is modied as
!
(4.204)
as can be easily worked out by introducing a nite collision frequency in the equation of motion.
For a frequency much smaller than the plasma frequency, the permittivity becomes negative,
"'
"0
! 2p
;
!2
(4.205)
!p
;
c
(4.206)
!t)
= E0 exp
!p
z e
c
i!t
(4.207)
where z is the distance in the plasma from the vacuum-plasma boundary. This means that an
electromagnetic wave incident on a plasma cannot penetrate into plasma except for a distance of
the order of the skin depth dened by
c
=
:
(4.208)
!p
The wave is completely reected if the plasma is collisionless. Reection of low frequency radio
waves by the ionoshperic plasma and is a well known example.
37
!p;
c;
!
:
2 c
(4.209)
e
(E + v
m
B)
@f
= 0:
@v
(4.210)
As it is, it is a nonlinear equation because the electromagnetic elds E and B associated with a
wave aect the distribution function f: Let us assume a wave propagating along the z direction
with electric eld polarized in the x direction Ex (z; t) = E0 ei(kz !t) : The distribution function may
be linearized as f = f1 + FM ; where f1 is the perturbation and FM is unperturbed Maxwellian
distribution. Noting
mv
@FM (v 2 )
=
FM (v 2 );
(4.211)
@v
Te
and thus
(v
B)
we nd
f1 =
@FM (v 2 )
= 0;
@v
e
Ex
Te
vx FM :
i (kvz !)
(4.212)
The current density can be found from the rst order moment,
Z
Je = ne vf1 dv:
(4.213)
= !=kvT e with vT e =
ne2
Ex
i
Te
vx2
kvz
FM dv
Z 1
ne2
1
e
i
Ex p
mkvT e
1 t
2
ne
i
Ex Z ( ) ;
mkvT e
t2
dt
(4.214)
38
!
! 2p
Z( ) :
1+
!kvT e
(4.215)
1;
! 2p
!2
p
Z( )'i ;
and
! 2p
p
" ' i "0
:
!kvT e
1=6
'
!p
c
2=3
!
vT e
1=3
(4.216)
This is often called anomalous skin eect. (This is probably misnomer because there is nothing
anomalous in the derivation.)
4.14
Some crystals exhibit anisotropy in polarizability. The permittivity in such media becomes a tensor,
and the displacement vector D and electric eld E are related through a dielectric tensor ";
D = " E; or Di ="ij Ej ;
(4.217)
where " is a diagonal tensor consisting of three permittivities in each axial direction, x; y; and z;
1
"x 0 0
B
C
" = @ 0 "y 0 A :
0 0 "z
0
(4.218)
In uniaxial crystals, polarization occurs preferentially along one axis, say, "x = "y 6= "z : In
anisotropic media, the phase and group velocities are in general oriented in dierent directions
and the well known double refraction phenomenon (already known in the 17th century) can be
39
@B
; r
@t
E=
B=
@D
=
@t
Then
r
E=
@2E
;
@t2
0"
r (r E) r2 E+
0"
0"
@2E
= 0:
@t2
@E
:
@t
(4.219)
(4.220)
(4.221)
Note that in an anisotropic medium, the divergence of the electric eld, r E; does not necessarily
vanish, although
r D = free = 0;
(4.222)
must hold. After Fourier decomposition, Eq. (4.221) reduces to
k(k E)
k2 E + !2
0"
E = 0;
(4.223)
or
(k 2
!2
ki kj
ij
0 "ij )Ej
= 0:
(4.224)
The dispersion relation of electromagnetic waves in a uniaxial crystal is therefore given by the
determinant,
det(k 2 ij ki kj ! 2 0 "ij ) = 0:
(4.225)
Introducing ^"ij = "ij ="0 and the index of refraction,
n=
c
k;
!
(4.226)
ij
ni nj
^"ij ) = 0:
(4.227)
In uniaxial crystals, polarizability along one axis (say, z-axis) diers from those along other
axes. We assume
"x = "y "? ; " z "k :
(4.228)
Since the crystal is symmetric about the z-axis, the wave vector k can be assumed to be in the
x z plane without loss of generality,
kx = k? ;
40
kz = kk ;
(4.229)
or
n? =
ck?
;
!
nk =
ckk
:
!
(4.230)
^"? 0
n? nk
n? nk
n2
0
^"? 0
n2?
= 0:
(4.231)
^"k
(4.232)
and
^"k n2k + ^"? n2?
^"k^"? = 0:
(4.233)
The rst mode of propagation is independent of the propagation angle as if in an isotropic medium.
It is called the ordinary mode because it maintains the properties of electromagnetic waves in
isotropic media. The electric eld is in the y-direction and the dispersion relation in Eq. (4.232) is
equivalent to the following dierential equation,
@2
@2
+
@x2 @z 2
0 "? !
Ey (x; y) = 0:
(4.234)
In this mode, the phase and group velocities are in the same direction (in the direction of k)
although in magnitude they dier because of the frequency dependence of the permittivity "? (!):
The second mode has a peculiar property and for this reason is called extraordinary mode. Let
the angle between k and the z-axis be : The dispersion relation in Eq. (4.233) becomes
ck
!
s
!
cos2
=c
k
^"?
^"k^"? = 0:
sin2
ek ;
^"k
(4.235)
(4.236)
=
=
@!
1 @!
ek +
e
@k
k@
!
c2
1
ek +
k
!=k ^"k
1
^"?
sin cos e :
(4.237)
This is evidently not parallel to the phase velocity unless ^"k = ^"? (isotropic dielectric). The
magnitude of the group velocity is always larger than that of the phase velocity.
Electromagnetic waves in a plasma conned by a magnetic eld is another example of anisotropic
medium which accommodates variety of waves. A simple case of cold plasma will be considered in
41
Problem 4.11.
Figure 4-8: Double refraction of randomly polarized incident wave by a uniaxial crystal. Propagation of the O-mode (ordinary mode) is unaected but that of the X-mode (extraordinary mode) is.
The group velocity of the X-mode in the crystal deviates from the incident direction.
An important consequence of the presence of the ordinary and extraordinary modes is the well
known double refraction caused by some crystals. Consider a light beam with random polarizations
incident normal to a surface of a uniaxial crystal as shown in Fig. 4-8. Unless the optical axis (zaxis in the geometry assumed) coincides with the direction of the incident beam, the beam is split
into two beams at the surface. In the crystal, one beam propagates along the direction of the
incident beam and another at an angle. The phase velocities of the ordinary and extraordinary
modes are in the same direction as the incident beam but the group velocity of the extraordinary
mode deviates from the incident direction. Wave energy propagates at the group velocity and thus
beam splitting occurs.
Some isotropic dielectrics can become uniaxial media if placed in an electric eld. This is
because the permittivity is in general nonlinear and the component in the direction of the eld
becomes eld-dependent,
" = "(0) + E n ;
(4.238)
where is a constant, n = 1 is for Pockels eect and n = 2 is for Kerr eect. Pockels eect in
some liquids is widely used for laser switching and optical modulation.
42
Problems
4.1 A sphere of radius a carries a charge q which is decreasing due to emission of charge in
every radial direction. Show that there is no magnetic eld even though there exists a radial
conduction current density.
4.2 A cylindrical permanent magnet with radius a and a uniform axial magnetization Mz is
rotating at an angular frequency ! about its axis. What is the emf induced in the closed
circuit shown?
! 2pe
!2
! pe =
ne2
:
" 0 me
43
0H
= "0 E 2 ;
! 2p
!2
! 20
Show that the group velocity does not exceed c: What if a nite dissipation is allowed,
! 2p
! 2 + 2i !
"(!) = "0
! 20
1
3"0 0
0
B
C
!
" = @ 0 3"0 0 A :
0
0 2"0
A light beam in air is incident on a at surface(x y plane)of the crystal at an angle i
from the normal which coincides with the optical axis z: Show that the refraction law for the
44
extraordinary mode is
tan
p
^"? sin
=q
^"k ^"k
sin2
;
i
where ^"? = 3; ^"k = 2: What is the refraction law for the ordinary mode?
4.11 Find the impedance per unit length of a copper wire of radius 3 mm at f = 60, 103 and 106
Hz. Copper conductivity is = 5:8 107 S/m at room temperature.
4.12 Using the equation of motion for electrons in a cold magnetized plasma,
me
@v
=
@t
e (E + v
B0 ) ;
where B0 (external magnetic eld) is in z direction, show that the dielectric tensor is given
by
3
2
! 2p
! 2p
i
0
7
6 1 !2
2
2)
!(! 2
7
6
2
2
7
6
!
!
!
p
ps
7
" = "0 6
1
0
7
6 i
2)
2
!2
7
6 !(! 2
4
! 2p 5
0
0
1
!2
p
where ! p = ne2 =me "0 is the plasma frequency and e = eB0 =me is the cyclotron frequency.
The wavevector k may be assumed to be k = k? ex +kk ez without loss of generality because of
axial symmetry. Note that the tensor is Hermitian. (In this analysis, the electron temperature
is ignored.)
4.13 A laser beam passes through a glass window of refractive index 1.50 into water with refractive
index 1.33. The beam is E-polarized (electric eld in the incident plane). Design the glass
window to avoid reection at both surfaces.
45