Fields and waves in nature and

engineering — the big picture: Copyright ©2021 Reserved — no parts of this

set of lecture notes (Lects. 1-39) may be re-

Fundamental building blocks of matter — electrons and protons at atomic produced without permission from the author.

scales — interact with one another gravitationally and via “electromagnetic”

forces. These interactions are most conveniently described in terms of suit-
ably defined “vector fields” that permeate space and time, or simply the space-
time (x, y, z, t) ≡ (r, t). Interactions attributed to particle masses can be
formulated by gravitational fields g(r, t) specified in reference frames where
spatial coordinates r = (x, y, z) are defined. Far stronger interactions at-
tributed to particle charges, on the other hand, are formulated in terms of
a pair of vector fields, E(r, t) and B(r, t), known as electric and magnetic
fields, respectively.

Electric and magnetic fields:

A particle with charge q and mass m as well as position and velocity vectors
r and v = dr
dt specified at an instant t within a measurement frame (or “lab”
frame) will be accelerated in accordance with
dv
m = q(E(r, t) + v × B(r, t)), (1)
dt

1
which is Newton’s 2nd law of motion 1 for a particle under the influence of
Lorentz force
F = q(E + v × B). (2)
In view of (1), the operational definitions of fields E(r, t) and B(r, t) arise
from particle acceleration a = dv dt
that can be measured in the lab frame:
the electric field E is evidently force per unit stationary charge (i.e., v = 0)
whereas field B describes an additional force per charge in transport (i.e.,
qv) that acts in a direction perpendicular to v.
There are important differences between gravitational and electromag-
netic interactions: Gravitational interactions are always attractive indicating
that particle masses m that generate the gravitational field g(r, t) must all
have the same algebraic sign (taken to be positive by convention). Electro-
magnetic interactions, on the other hand, are attractive or repulsive depend-
ing on particle charges q which can be positive or negative. By convention
a positive charge q = e ≈ 1.6 × 10−19 C is attributed to the fundamental
particle know as proton, while, again by convention, q = −e for an electron,
the sole companion of the proton within a hydrogen atom2 . Protons and
electrons are charged elementary building blocks3 of all atoms (hydrogen as
1
Valid so long as |v| ≪ c where c is the speed of light in vacuum.
2
Hydrogen atom exists as a consequence of mutual attraction between proton and electron counterbal-
anced by quantum mechanical constraints on allowed energy states — the constraints include the influence
of short-lived virtual particle/anti-particle pairs interacting with the proton and electron in a sporadic
manner.
3
Atoms can also contain in their nuclei varying numbers of an uncharged particle known as the neutron
which is responsible for different isotopes of chemical elements (e.g., the hydrogen isotope known as
deuterium contains a neutron in addition to a proton and an electron). While neutrons have no net

2
well as atoms of heavier elements) that constitute the matter around us. In a
collection of fundamental particles the total mass is always a monotonically
increasing function of the number of particles. However, that is not the case
with total charge since individual particle charges can be positive or nega-
tive. In fact, the net charge density ρ(r, t) found in macroscopic amounts of
matter is typically close to zero as a result of having nearly equal numbers of
protons and electrons in ordinary matter composed of charge-neutral atoms
and molecules4 .

charge, they consist of charged sub-nuclear particles known as up ( 23 e) and down (− 31 e) quarks whose
motions within the neutron establish currents and a magnetic moment.
4
The reason why intrinsically weaker gravity becomes dominant in the macro world.

3
Fields are relative:
Physical laws that we use today to describe our surroundings have been
developed to have identical forms in all reference frames in uniform motion
with respect to one another. For instance, Lorentz force law on a charge q is
F = q(E + v × B) and F′ = q(E′ + v′ × B′ ) (3)
in terms of unprimed and primed variables measured in two reference frames.
Moreover, particle charge q and the speed of light c are assigned invariant5
values in reference frames in relative motion (thus q ′ and c′ are unnecessary
to invoke in physical models). The ramifications of these restrictions con-
stituting the special theory of relativity (first described by Einstein in
1905 and covered at UIUC in PHYS 325) are in full accord with experimen-
tal measurements. They are also well matched by Newtonian relations
(approximate but more intuitive laws of dynamics covered in PHYS 211) if
and when the relative speed of primed and unprimed frames is negligible
compared to the speed of light c.
Since in Newtonian descriptions mass m and acceleration dv dt have invari-
ant values in all reference frames, it follows that if and when |v′ − v| ≪ c,
then F′ = F, in which case (3) implies
E′ + v′ × B′ = E + v × B. (4)
5
Other “relativistic invariants”
p between different
p reference frames include particle (rest) masses and the
so-called “spacetime interval” t2 − L2 /c2 = t′2 − L′2 /c2 between two events ocurring at two locations
and two times separated by a distance L and time-delay t, respectively. Relativistic invariants are the
most prized physical quantitites to focus on in relativistic models (simply because they remain fixed in all
reference frames). Note that distances L 6= L′ and time-delays t 6= t′ are not relativistic invariants!

4
Then, for a stationary charge in the primed frame, we have v′ = 0 and

E′ = E + v × B, (5)

which indicates that force per unit stationary charge in the primed frame
— i.e., the electric field in the primed frame — is a linear combination of
electrical and magnetic forces exerted on the same charge as seen from an-
other reference frame (unprimed) where the charge appears to have a vector
velocity v.
Thus, electric and magnetic fields needed in the formulation of charged
particle interactions are not unrelated to one another — rather, they intermix
in a manner that depends on the reference frame6 being used for analysis
purposes. Note that charges q which are stationary in one reference frame
(and therefore carry no electrical current) will appear to be in motion in
another frame and thus carry electrical currents I. It must therefore be
evident that the equations for E and B in any reference frame must be cross-
coupled and depend on both charge and current densities that are measured
in the same frame.
6
Given E and B measured in the lab, E′ and B′ measured by an observer moving through the lab
with a constant velocity v are well approximated by E′ ≈ E + v × B and B′ ≈ B − v×E c2
so long as
8
|v| ≪ c = 3 × 10 m/s, the speed of light in free space (shown by relativistic analysis discussed in PHYS
225 — exact transformation formulae are E′k = Ek , B′k = Bk , E′⊥ = γ(E⊥ + v × B⊥ ), B′⊥ = γ(B⊥ − v×Ec2
⊥
),
1
where γ = q v2 ).
1−
c2

5
Maxwell’s field equations:
The required set of coupled equations governing E and B was “discovered” in
1864 by James Clerk Maxwell to be (first introduced in PHYS 212 in integral
form and discussed throughout this course)

ρ
∇·E= ǫo Divergence eqn’s ∇·B=0

∇ × E = − ∂B
∂t
Curl eqn’s ∇ × B = µoJ + µoǫo ∂E
∂t

where

H 1 1 F
µo ≡ 4π × 10−7 and ǫo = ≈
m µoc2 36π × 109 m
in mksA units and
1 m
c=√ ≈ 3 × 108
µ o ǫo s
is the speed of light in free space. Furthermore ρ = ρ(r, t) refers to the net
charge density and J = J(r, t) to the current density in the measurement
frame, whereas ∇ · E and ∇ × E refer to the divergence and curl of vector
field E generated by partial differentiation of the orthogonal components of
E (concepts introduced in MATH 241 and reviewed in Lecture 4).

6
Solutions of Maxwell’s equations — waves and static fields (AC/DC):
Maxwell’s partial differential equations shown above, describing the coupled
dynamics of electric and magnetic fields E and B in response to space and
time varying source fields ρ and J, require an extended study to appreciate
their full ramifications and predictions. All predictions of these equations
have been experimentally verified and it has been found out that everything
that is known and observed about electricity and magnetism can be explained
in terms of these equations and their quantized forms. t=0 c
One of their predictions, derived specifically in Lecture 18, is that they
support traveling wave solutions of the form (a)
L
z
E(r, t) ∝ B(r, t) ∝ cos(2πf (t − )) (6)
c
in regions where J = ρ = 0. These are co-sinusoidal field perturbations hav- t = L/2c
c
ing oscillation frequencies f , oscillation periods T = f1 , wavelengths λ = fc ,
(b)
and they travel in 3D space with the speed of light c in free space. Since L

Maxwell’s equations are linear, superpositions of co-sinusoidal waves with

different wavelengths provide additional solutions — these can have arbi-
trary spatial variations and still travel at a fixed speed c. Any such field t = L/c
c

perturbation will travel across a region of size L during a time interval L/c
(c)
as illustrated in the margin. L

Another prediction of Maxwell’s equations is that fields established by

static — i.e., non-time-varying — charge and current densities ρ = ρ(r) and
J = J(r) satisfy two separate sets of decoupled equations

7
 Electrostatics Magnetostatics

ρ
∇·E= ǫo
Divergence eqn’s ∇·B=0

∇×E=0 Curl eqn’s ∇ × B = µo J

shown in the left and right columns above — these were obtained by sim-
ply setting the terms ∂E/∂t and ∂B/∂t in the curl equations to zero. In-
dependent “curl-free” static electric fields E(r) and “divergence-free” static
magnetic fields B(r) satisfying these simplified equations are naturally far
easier to determine than the coupled dynamic fields E(r, t) and B(r, t) to be
encountered in response to time-varying sources ρ(r, t) and J(r, t).

Quasi-static fields:
Even though in practical cases of interest (in physics and engineering) time-
varying sources are the “rule” and static sources an “exception”, learning to
solve the simplified set of electrostatics and magnetostatics equations turns
out to be invaluable. The reason is, static solutions often provide accurate
approximations — known as quasi-static approximation — for time-varying
field problems involving slowly-varying sources ρ(r, t) and J(r, t).
More specifically, if the source variation period T is much longer than the
travel time L/c of field perturbations across a region of size L, that is, if
L
T ≫ , (7)
c
8
then field calculations for the entire region can be done statically using the t=0 c
instantaneous (as opposed to retarded or previous) values of field sources
ρ and J. This is true because under the given condition source strengths (a)
L
will remain nearly constant over time intervals needed to communicate the
new fields to the most distant corners of the region of interest. We can also
re-state the same inequality (7) as t = L/2c
c

c
L ≪ cT = =λ (8) (b)
L
f
using the definition of wavelength λ introduced earlier. The indication is
then, any system with a physical size L that is small in terms of wavelength t = L/c
c
λ of the applied field variations can be analyzed quasi-statically by starting
from Maxwell’s static equations. (c)
L

Fields and circuits:

Lumped circuit analysis techniques introduced in ECE 110 and 210 constitute
practical applications of the quasi-static approach suitable for “electrically
small circuits” consisting of capacitors, inductors, and resistors and slowly
varying AC sources. By contrast, the analysis of “electrically large circuits”
with physical dimensions L approaching or exceeding λ requires taking a
proper account of propagation time delays L/c in the system by developing
a distributed circuit approach based on the full set of Maxwell’s equations.
One practical application area where this need is most acute nowadays
is in chip (integrated circuit) design and packaging suitable for high-speed

9
computing7 . While the physical dimensions of electronic chips and micro-
circuits are generally very small, such elements can still be electrically large
in the sense that L ∼ λ because of reduced wavelengths λ at high clock
speeds f = 1/T . Thus, even the computer engineers (CompE’s) amongst us
need to understand and learn how to mitigate (and take advantage of) the
ramifications of Maxwell’s equations.

7
E.g., Taflove, “Why study electromagnetics”, IEEE APM, 44, 132, 2002.

10
Details and study plan:
So much for the big picture about fields and waves encountered in nature
and engineering systems and circuits. Working details of how fields and wave
effects can be computed and characterized will be provided in the remaining
parts of these notes.
Over the course of 39 lectures we will develop and study, in succession,
the equations and applications of electrostatics (Lectures 1-11), magnetism
(Lectures 12-15), and electromagnetics (Lectures 16-39) with a focus on time
varying (quasi-static as well as wave-like) phenomena.

ECE 329:
We start by finding out how the equations of electrostatics arise from the
familiar Coulomb’s law (like charges repel, unlike charges attract) and the
idea of field superpositions. We learn to solve electrostatic problems using the
notion of electrostatic potential (voltage) and develop the notions of polariza-
tion, conduction, charge continuity, and capacitance in quasi-static settings
of practical importance.
Next we learn how magnetic fields arise from charges in motion (a relative
concept depending on the reference frame of the observer) and develop the
governing laws of magnetostatics (also an extension of Coulomb’s law seen
from different reference frames). The vector potential is introduced for mag-
netic field calculations from prescribed current configurations, and notions
of magnetization and inductance are subsequently developed and applied in

11
quasi-static settings.
Just like time-varying electric fields imply time-varying charge densities
(or vice versa) in electro-quasi-statics (EQS), time-varying currents imply
time-varying magnetic flux in magneto-quasi-statics (MQS). We also learn
that time-varying magnetic-flux is accompanied by time varying electric fields
— a key finding of Faraday’s called induced field with paradigm shifting
ramifications and applications — and requires the modification of curl-free
electric field condition into a dynamic equation known as Faraday’s law.
Finally, the full set of Maxwell’s equations is reached after adding a
time-varying electric field term to the curl equation of magnetostatics. This
change, first introduced by Maxwell in order to make sure that the govern-
ing equations of electricity and magnetism are consistent with conservation
of charge, acknowledges the two-way coupling and feedback between electric
and magnetic fields: time-varying magnetic fields induce time-varying elec-
tric fields — Faraday effect — and time-varying electric fields in turn induce
time-varying magnetic fields (call it the “Maxwell effect”) in order to sustain
electromagnetic field variations in regions far away from charges and current
loops — that is the way nature seems to work (and here we are to observe
all that thanks to Maxwell effect allowing us to be here).
A study of wave solutions of Maxwell’s equations follows, including plane
TEM waves in free space, linear and circular polarized waves, waves in con-
ducting media, normal incidence of waves on planar interfaces of homoge-
neous regions, energy and momentum transfer, guided waves in two-wire
transmission-line (TL) systems, transient response on TL circuits, resonant

12
oscillations in TL cavities, sinusoidal steady-state analysis of TL’s and dis-
tributed circuits, Smith Chart applications, and finally losses in TL systems.
That is the full scope of the 39 lectures of ECE 329 — the course ends
with an intensive study of distributed circuit concepts based on transmission
lines, a study that complements the lumped circuit techniques examined and
mastered in earlier courses.
ECE 329 is only the first half of our first-pass study of the fields and waves
topics essential in electrical engineering education. Important topics such as
radiation and antennas (generation details of electromagnetic waves by time-
varying currents) and dispersion (frequency dependence of wave propagation
speeds in material media) are barely mentioned or not at all in ECE 329.
These constitute the main topics of the follow-on course, ECE 350.

ECE 350:
ECE 350 starts with the discussion of electromagnetic radiation theory and
transmission antennas, continues with propagation and wave guidance ef-
fects (including dispersion, phase and group velocities, Doppler shifts, oblique
incidence, evanescence and tunneling effects, guided modes in parallel-plate,
rectangular, and dielectric slab waveguides), treats cavity fluctuations (in-
cluding resonant modes, blackbody radiation in 3D cavities, thermal noise),
and concludes with a discussion of antenna reception (including effective area,
available power, link equations).

13
Beyond ECE 329 and 350:
Students having gone through ECE 329 and 350 will find themselves ready to
encounter higher level courses in our curriculum focusing on different applica-
tion areas and frequency regimes of the implications of Maxwell’s equations.
It is a life-long endeavor to master these relationships which have precipitated
the scientific upheavals of the 20th century (relativity and quantum mechan-
ics) and have remained intact and essential despite the upheavals unlike most
aspects of classical physics. Our high speed electronics and communication
networks and devices are intrinsically and fundamentally based on fields and
wave concepts. Progress and innovation in these areas will require a deep
understanding of fields and waves and how they interact with novel materials
and structures.
Learn the basics and then go and invent the next thing!

14
 ECE 329 Lecture Notes — Summer 09/11/21, Erhan Kudeki Copyright ©2021 Reserved — no parts of this

set of lecture notes (Lects. 1-39) may be re-

1 Vector fields and Lorentz force produced without permission from the author.

• Interactions between charged particles can be described and modeled8 2

y

in terms of electric and magnetic fields just like gravity can be

formulated in terms of gravitational fields of massive bodies. 1

– In general, charge carrier dynamics and electromagnetic field vari-

ations9 account for all electric and magnetic phenomena observed -2 -1 1 2
x

in nature and engineering applications.

-1

• Electric and magnetic fields E and B generated by charge carriers —

electrons and protons at microscopic scales — permeate all space with -2

proper time delays, and combine additively.

– Consequently we associate with each location of space having Carte-

sian coordinates
(x, y, z) ≡ r
a pair of time-dependent vectors
E(r, t) = (Ex(r, t), Ey (r, t), Ez (r, t))
8
Interactions can also be formulated in terms of past locations (i.e., trajectories) of charge carriers.
Unless the charge carriers are stationary — i.e., their past and present locations are the same — this
formulation becomes impractically complicated compared to field based descriptions.
9
Time-varying fields can exist even in the absence of charge carriers as we will find out in this course
— light propagation in vacuum is a familiar example of this.

15
 and
B(r, t) = (Bx(r, t), By (r, t), Bz (r, t))
that we refer to as E and B for brevity (dependence on position
r and time t is implied ). Maxwell’s equations:

ρ
• Field vectors E and B and electric charge and current densities ρ and ∇·E =
ǫo
∇·B = 0
J — describing the distribution and motions of charge carriers — are ∂B
∇×E = −
related by (i.e., satisfy) a coupled set of linear constraints known as ∂t
∂E
∇×B = µo J + µo ǫ o .
Maxwell’s equations, shown in the margin. ∂t

such that
– Maxwell’s equations are expressed in terms of divergence and curl
F = q(E + v × B),
of field vectors — recall MATH 241 — or, equivalently, in terms
of closed surface and line integrals of the fields enclosing arbitrary with

volumes V and surfaces S in 3D space, as you have first seen in µo ≡ 4π × 10−7

H
,
m
PHYS 212.
and
◦ Maxwell’s equations were “discovered” as a consequence of ex- 1 1 F
ǫo = ≈ ,
perimental and theoretical studies led by 19th century scien- µo c 2 36π × 109 m

tists including Gauss, Ampere, Faraday, and Maxwell. in mksA units, where

1 m
They remain intact and essential despite the scientific upheavals c= √
µo ǫ o
≈ 3 × 108
s
(paradigm shifts) of 20th century: relativity and quantum physics10 .
is the speed of light in free space.

(In Gaussian-cgs units Bc is used

1
10
Fields are are utilized in different ways in classical and quantum electrodynamics, but Maxwell’s field in place of B above, while ǫo = 4π
1 4π
and µo = ǫo c2 = c2 .)
equations remain the same under both paradigms. Relativity theory is an updated model of space and
time relations developed to achieve consistency with the implications of Maxwell’s equations.

16
Given the charge and current densities ρ and J, Maxwell’s equations can be
solved for the fields E and B.

• Field solutions E and B in turn determine how a “test charge” q with

mass m, position r, and velocity v ≡ ṙ = dr
dt
accelerates in accordance
with Lorentz force Lorentz
F = q(E + v × B) force
d
and Newton’s 2nd law F = dt
mv (in classical electrodynamics). As
such Units in mksA sys-
tem:
– electric field E at any location r is the vector force per stationary – q[=]C=sA,
(i.e., v = 0) unit charge (i.e., q = 1 C), – E[=]N/C=V/m,
– B[=]V.s/m2 =Wb/m2 =T,
– magnetic field B describes an additional force per unit charge
– ρ[=]C/m3 ,
which is experienced by charges in motion (v 6= 0) in the reference
– J[=]A/m2 ,
frame — typically called the “lab frame” — where E and v are
where
measured. C, N, V, Wb, and T
are abbreviations for
Coulombs, Newtons, Volts, We-
Since Lorentz force equation has the same form in all inertial reference bers, and Teslas,
frames11 (like all laws of physics, including Maxwell’s equations) while the respectively.

charge velocity v is clearly frame-of-reference dependent, it follows that the

Charge q is quantized in units of
values of fields E and B must also be dependent on the reference frame12 . e = 1.602 × 10−19 C, a relativistic
invariant.
11
Coordinate systems in which particles not subjected to any force — or, if general relativistic effects
are to be retained, particles subjected to gravitational forces only — follow linearly varying trajectories.
12
Given E and B measured in the lab, E′ and B′ measured by an observer moving through the lab
with a constant velocity v are well approximated by E′ ≈ E + v × B and B′ ≈ B − v×E c2
so long as
8
|v| ≪ c = 3 × 10 m/s, the speed of light in free space.

17
 F
• Charge carrier positions r, velocities ṙ, and accelerations r̈ = m , as well
as forces F, fields E and B, and current density J are all described, in
general, in terms of 3D vectors.

• In Cartesian coordinates such vectors and vector functions (of posi-

tion r and/or time t) can be expressed in terms of mutually orthog-
onal unit vectors x̂, ŷ, and ẑ as in
r = (x, y, z) = xx̂+y ŷ+z ẑ and E = (Ex, Ey , Ez ) = Exx̂+Ey ŷ+Ez ẑ etc.,

where
p q
– |r| ≡ x + y + z and |E| ≡ Ex2 + Ey2 + Ez2 etc., are vector
2 2 2
z
r = (x, y, z)
magnitudes,
= xx̂ + y ŷ + z ẑ
r E
– r̂ ≡ |r| and Ê ≡ |E| etc., are associated unit vectors, y
– with dot products ẑ
ŷ
◦ r̂ · r̂ = 1, Ê · Ê = 1, x̂ · x̂ = 1, etc., but x
x̂
◦ x̂ · ŷ = x̂ · ẑ = ŷ · ẑ = 0 UNIT VECTORS AND A POSITION
VECTOR IN RIGHT-HANDED
– and cross products CARETESIAN COORDINATES

◦ x̂ × ŷ = ẑ,
◦ ŷ × ẑ = x̂,
◦ ẑ × x̂ = ŷ,
adopting a right-handed convention (see the margin note in
the next page).

18
 • Recall that Right handed con-
vention: cross product vec-
tor points in the direction indi-
– Dot product A · B is defined as |A| times |B| times the cosine of cated by the thumb of your right
angle θ between A and B. hand when you rotate your fin-
gers from vector A toward vector
◦ Thus dot product is zero when angle θ is 90◦, as in the case B through angle θ you decide to
use.
of x̂ and ŷ, etc.
– Cross product A × B is defined as a vector with a magnitude B = |B|b̂
|A| times |B| times the sine of angle θ between A and B and a
|B| sin θ
direction orthogonal to both A and B in a right-handed sense n̂
θ â
(see margin note) . |B| cos θ A = |A|â
◦ Thus cross product is zero when the vectors cross multiplied A · B = |A||B| cos θ
DOT PRODUCT:product of
are collinear (θ = 0◦) or anti-linear (θ = 180◦). projected vector lengths

A × B = |A||B| sin θâ × n̂

CROSS PRODUCT: right-handed
perpendicular area vector of
the parallelogram formed
Example 1: Given the vectors v = (5, 10, 0) and B = (0, 0, 2) compute the cross and by co-planar vectors

dot products v × B and v · B.

z
Solution: Since we can also write v = 5x̂ + 10ŷ and B = 2ẑ, it follows that r = (x, y, z)
v × B = (5x̂ + 10ŷ) × 2ẑ = 10x̂ × ẑ + 20ŷ × ẑ = −10ŷ + 20x̂.
= xx̂ + y ŷ + z ẑ
y
Alternatively, using the well known determinant method for cross products,
ẑ
x̂ ŷ ẑ ŷ
x
v ×B = 5 10 0 = x̂(10·2−0·0)− ŷ(5·2−0·0)+ ẑ(5·0−10·0) = 20x̂−10ŷ. x̂
UNIT VECTORS AND A POSITION
VECTOR IN RIGHT-HANDED
0 0 2 CARETESIAN COORDINATES

19
 Also, v · B = (5, 10, 0) · (0, 0, 2) = 5 · 0 + 10 · 0 + 0 · 2 = 0.

Having three non-colinear

force measurements Fi cor-
responding to three distinct
test particle velocities vi is
Example 2: A particle with charge q = 1 C passing through the origin r = (x, y, z) = sufficient to determine the
fields E and B at any location
0 of the lab frame is observed to accelerate with forces in space produced by distant
F1 = 2x̂, F2 = 2x̂ − 6ẑ, F3 = 2x̂ + 9ŷ N sources as illustrated by this
example.
when the velocity of the particle is
z
m
v1 = 0, v2 = 2ŷ, v3 = 3ẑ ,
s
in turns. Use the Lorentz force equation y

F = q(E + v × B) v1 = 0
x
to determine the fields E and B at the origin. F1 = 2x̂

Solution: Using the Lorentz force formula first with F = F1 and v =v1, we note that z

2x̂ = (1)(E + 0 × B),

y
which implies that
N V
E = 2x̂ = 2x̂ . v2 = 2ŷ
C m x
Next, we use
F F F2 = 2x̂ − 6ẑ
v×B= − E = − 2x̂ z
q q
with F2 = 2x̂ − 6ẑ and v2 = 2ŷ, as well as E = 2x̂ V/m, to obtain
y
2ŷ × B = −6ẑ ⇒ ŷ × B = −3ẑ;
v3 = 3ẑ F3 = 2x̂ + 9ŷ
x
20
 z
likewise, with F3 = 2x̂ + 9ŷ and v3 = 3ẑ,
y
3ẑ × B = 9ŷ ⇒ ẑ × B = 3ŷ.

Substitute B = Bx x̂ + By ŷ + Bz ẑ in above relations to obtain v1 = 0

x
F1 = 2x̂
ŷ × (Bx x̂ + By ŷ + Bz ẑ) = −Bx ẑ + Bz x̂ = −3ẑ
z
and
ẑ × (Bx x̂ + By ŷ + Bz ẑ) = Bx ŷ − By x̂ = 3ŷ.
Matching the coefficients of x̂, ŷ, and ẑ in each of these relations we find that y

Wb v2 = 2ŷ
Bx = 3 , and By = Bz = 0. x
m2
Hence, vector F2 = 2x̂ − 6ẑ
Wb z
B = 3x̂ 2 .
m
y
v3 = 3ẑ F3 = 2x̂ + 9ŷ
x
• In your first homework you will be asked to do a sequence of vector
exercises, including problems on volume, surface, and line integrals of
vector or scalar functions of space (i.e., “fields”). These problems should
be worked out with the help of your PHYS 212 and/or MATH 241 texts
and notes.

– This course assumes a background of PHYS 212 and MATH

241 (on electromagnetic fields and vector calculus) as well

21
 as ECE 210 (lumped circuits and linear systems con-
cepts including time- and frequency-domain approaches
and phasors).

• The main objective of the course is to build up a firm understand-

ing of electromagnetic field concepts introduced in PHYS 212,
and to learn how to use Maxwell’s equations under static and
time-varying conditions associated with unguided (i.e., wireless) and
guided (mainly transmission lines) electromagnetic waves. The study Prerequisites:
of guided waves is the key to extend the familiar lumped-circuit MATH 241
concepts into the realm of distributed circuits. This is the first half PHYS 212
of a sequence of core electromagnetics courses in our curriculum, the ECE 210
second course being the 3-of-5 elective ECE 350.

– Topical outline: Follow-on:

1. Static electric fields, potential, polarization, quasi- ECE 350
static applications (10 lectures)
2. Static currents and magnetic fields (3 lectures)
3. Time-varying fields and Maxwell’s eqns (4 lectures)
4. Plane wave solutions of Maxwell’s eqns (9 lectures)
5. Guided waves in transmission lines and distributed
circuits (13 lectures)

22
2 Static electric fields — Coulomb’s and Gauss’s
laws
Static electric fields E(r) are produced by static (non-time-varying) distri-
bution of charges and obey and the electrostatic laws shown in the margin
where ρ(r) denotes the net charge density in 3D volume. Over the next few Laws of
lectures we will find out how these laws emerge from Coulomb’s law. electrostatics:
At the most elementary level, each stationary point charge (electron or
proton) Q is surrounded by its radially directed electrostatic field E given ∇ · E = ρ/ǫo
by Coulomb’s law, and in the presence of multiple charges the field vectors ∇×E=0
of all the charges are added vectorially (linear superposition holds) to obtain
a superposition field E.
r̂
• Coulomb’s law specifies the electric field of a stationary charge Q at q
the origin as
Q r = |r|r̂
z
E(r) = r̂ Q
4πǫor2 y
x
1
as a function of position vector r = (x, y, z), where ǫo ≈ 36π×10 9 F/m Force exerted by Q on q:
is a scaling constant known as permittivity of free space, F = qE
p with electric field
r = |r| = x2 + y 2 + z 2 Q
E= r̂
r 4πǫo|r|2
is radial distance from the charge, and r̂ = r radial unit vector pointing With multiple Q’s superpose
away from the charge. multiple E’s

– This Coulomb field E(r) will exert a force F = qE(r) on any

1
 stationary “test charge” q brought within distance r of Q (see
figure in the margin). If qQ > 0, force F is
repulsive (directed along
The existence of a Coulomb field accompanying each charge carrier in its rest r̂), if qQ < 0 it is at-
frame1 is taken to be a fundamental property of charge carriers (established tractive — like charges
by measurements). repel, unlike charges at-
tract.
• When multiple static charges Qn are present in a region, the force
on a stationary test charge q can be described as qE in terms of a
z
superposition field r − rn
X Qn q Qn
E= r̂
2 n rn
n
4πǫ o rn
r

written in terms of the magnitudes and directions of vectors rn pointing

y
from each Qn to q.
O
x
– Equivalently, we can write Position vectors of charges
are referenced with respect
to a common origin O
X Qn r − rn
E(r) = ,
n
4πǫo|r − rn |2 |r − rn |

where r and rn now denote the locations of q and Qn with re-

spect to a common origin — this form is more convenient when
static electric field E is to be calculated for an arbitrary location
r (independent of the test charge notion).
1
In non-inertial rest frames charge carriers will also produce an additional field proportional to the
acceleration of free particles observed in such frames (e.g., Boyer, Am. J. Phys., 47, 129, 1979; Gupta
and Padmanabhan, Phys. Rev. D, 57, 7241, 1998).

2
Example 1: Charges Q1 = 4πǫo and Q2 = −2Q1 are located at coordinates r1 =
(1, 0, 0) = x̂ and r2 = (0, 1, 0) = ŷ, respectively. What is the expression for E(r)
and what is the explicit value of vector E(0)?
y

Solution: Field E due to Q1 and Q2 at an arbitrary point r can be obtained as 2

Q1(r − r1) Q2 (r − r2 )
E(r) = +
4πǫo |r − r1 |3 4πǫo |r − r2|3 1

(r − x̂) 2(r − ŷ) (x − 1, y, z) 2(x, y − 1, z)

= − = − V/m.
|r − x̂|3 |r − ŷ|3 |(x − 1, y, z)|3 |(x, y − 1, z)|3 x
-2 -1 1 2

At the origin where r = (0, 0, 0), this result gives

(−1, 0, 0) 2(0, −1, 0) -1

E(0, 0, 0) = − = −x̂ + 2ŷ V/m.

|(−1, 0, 0)|3 |(0, −1, 0)|3
-2

Field map of a dipole plus a

negative charge
• The vector map shown in the margin depicts samples of unit vec- y

E(r) 10

tors Ê(r) ≡ |E(r)| for the field E(r) obtained in Example 1 on a suit-
able grid established on xy-plane — such plots are useful or visualiza-
tion purposes. Note that arrows emanate out of the positive charge at 5

(x, y) = (1, 0) and converge upon the negative charge at (x, y) = (0, 1).
x
-10 -5 5 10

– Electrostatic fields can be alternatively visualized in terms of so-

called field lines or flux lines, continuous curves which are drawn
tangential to unit vectors Ê(r) at every position r. Try tracing
-5

out the flux lines over the vector map shown in the margin!
-10

3
 E, dS I Q
• According to Coulomb’s law, electrostatic field of a charge Q placed at S E · dS =
ǫo
the origin points out in the radial direction r̂ away from the origin and
S
has a magnitude Q
Q (a)
Er =
4πǫor2
E, dS I Q
E · dS =
that depends on radial distance r, but it does not depend on direction S′
ǫo
r̂. The product of Er with ǫo and the surface area of a sphere at radius (b) S′
Q
r, namely, S = 4πr2, yields I Q
E, dS S ′′ E · dS =
ǫo
ǫ o Er S = Q Q
(c) S ′′
independent of the radius of the sphere. Let’s re-write the same result
as I E, dS I

S ′′′ E · dS = 0
ǫo E · dS = Q, (d)
S S ′′′
Q
where Surface integral depends
only on the net amount of
charge contained within
H the surface --- charges
outside the surface don’t
– the “closed surface integral” S E · dS is called the flux of E over matter; surface shape doesn’t
matter; also charge motion
surface S bounding the volume V = 4π 3
r3 , within the surface does not
matter.

– which in turn denotes the limiting value of the sum of dot products
Ej · ∆Sj computed over all surface elements of S having incremen-
tal areas |∆Sj | and unit vectors ∆Sj /|∆Sj | pointing away from
volume V — the limiting value is obtained as all |∆Sj | approach
zero (i.e., with increasingly finer subdivision of S into |∆Sj | ele-
ments).

4
 H
Although we obtained the equality ǫo S E·dS = Q above only for a spherical
surface S centered about charge Q, we can easily convince ourselves — see the
sketches on the right — that the equality should hold even when we distort
the shape of surface S and/or displace Q away from the center so long as
we do not move Q outside of S. All such variations are permitted because
of inverse r-square dependence of the Coulomb’s law and additive nature of
fields, and if Q is moved outside the surface then the surface integral (flux)
simply goes to zero.
• Hence, given an arbitrary shaped volume V enclosed by an arbitrary
shaped surface S and including a net electrical charge QV , and defining
a displacement field Displacement
D ≡ ǫoE, FV = C
D = ǫoE[=] m m m2
we obtain I
D · dS = QV , Gauss’s law
S
a constraint known as Gauss’s law. At this stage, the introduction of Gauss’s law
D is simply a notational convenience.
Gauss’s law offers an alternative to implementing an explicit sum of Coulomb
fields for calculating static field distributions E or D = ǫoE — the alternative
method can be used when charge distributions have simplifying symmetry
properties as will be illustrated in the next set of examples.
Also, later on we will learn that Gauss’s law is valid even when charges
QV within volume V are non-static (i.e., in motion), a condition under which
Coulomb’s law is no longer valid.

5
 z

Q
∆z =
Example 2: Charged particles Q are located uniformly along the z-axis with an λ
average line density of λ C/m extending from z = −∞ to +∞. We will compute
y
the electrostatic field E of this charge distribution at a distance r from z-axis.
r = r(x̂ cos φ + ŷ sin φ)
Having an average charge density of λ C/m implies that individual charges Q φ
x
are spaced from one another by a distance ∆z = Qλ along the z-axis. Assuming
that charge locations are z = n∆z, where n is any integer, and using Coulomb’s Q point charge in C
law, we find that
∞ ∞
X Q r − ẑn∆z X λ∆z(r − ẑn∆z) λ charge density in C/m
E(r) = 2 |r − ẑn∆z|
= 3
,
n=−∞
4πǫo |r − ẑn∆z| n=−∞
4πǫo |r − ẑn∆z|

which, for position r = r(x̂ cos φ + ŷ sin φ) on xy-plane, at a distance r to the

z-axis, reduces to
∞
X λr(x̂ cos φ + ŷ sin φ)
E= 2 + n2 ∆z 2 )3/2
∆z (microscopic field)
n=−∞
4πǫo (r

because the ẑ component of E proportional to n∆z cancels out (as a result of

summation) due to symmetry in n. This field is “purely radial” in the direction
r̂ ≡ x̂ cos φ + ŷ sin φ
perpendicular to z-axis, and it can be evaluated, for r ≫ ∆z, as an integral
(remember that sums of infinitesimals are in effect definite integrals)
Z ∞ Z ∞
λr λr dz λ
r̂ 2 + z 2 )3/2
dz = r̂ 2 + z 2 )3/2
= r̂ . (macroscopic
−∞ 4πǫo (r 4πǫ o (r 2πǫ o r
| −∞ {z }
2
2/r field)

6
• The result
λ
E = r̂
2πǫor
obtained above, valid for r ≫ ∆z, and labelled as macroscopic
field , also represents at any r (and z) the space average of the mi-
croscopic field taken over small volumes having dimensions of many
∆z’s (inter-particle separations).

– In such a spatial average the rapidly varying structure of micro-

scopic field (in particular at small r, caused by the discrete nature
of charge distribution) is smoothed out as if electrical charge were
distributed in space with a continuous density of λ C/m.
– In realistic applications involving colossal numbers of charge car-
riers (of the order of 1023 in macroscopic chunks of solids) it is
practical (and desirable) to focus our attention on macroscopic
rather than microscopic fields.

We next illustrate how to obtain the macroscopic field E = r̂ 2πǫλo r

directly by using Gauss’s law.

7
 z
λ
Solution using Gauss’s law: We first notice that macroscopic electric field of a S r
charge distribution along the z-axis having an average charge density of λ C/m
λ
Er =
should be pointing in radial direction r̂ away from the z-axis (why?). L 2πǫor
y
Also its magnitude Er should be independent of azimuth angle φ by symmetry.
φ
x
As a consequence, we can apply Gauss’s law
I
D · dS = QV
S
as
ǫo Er 2πrL = λL
over the surface S of a cylindrical volume V of some length L and radius r
centered about the z-axis as shown in the margin — notice that our “clever”
choice of surface S in this problem resulted in the evaluation of the flux integral
in Gauss’s law without doing any calculus.

Clearly, this leads to (as obtained before using a line integral)

λ λ
Er = and E = r̂.
2πǫo r 2πǫo r

8
3 Gauss’s law and static charge densities
z
S
We continue with examples illustrating the use of Gauss’s law in macroscopic
y
field calculations: A ρs
Ex(x) =
2ǫo
Example 1: Point charges Q are distributed over x = 0 plane with an average surface ρS x
charge density of ρs C/m2 . Determine the macroscopic electric field E of this
charge distribution using Gauss’s law.
Ex(x)
Solution: First, invoking Coulomb’s law, we convince ourselves that the field produced
ρs
by surface charge density ρs C/m2 on x = 0 plane will be of the form E = x̂Ex (x) sgn(x)
where Ex (x) is an odd function of x because y- and z-components of the field will 2ǫo
cancel out due to the symmetry of the charge distribution. In that case we can x
apply Gauss’s law over a cylindrical integration surface S having circular caps of
area A parallel to x = 0, and obtain
I
D · dS = QV ⇒ ǫo Ex (x)A − ǫo Ex (−x)A = Aρs ,
S
which leads, with Ex (−x) = −Ex (x), to
ρs
Ex (x) = for x > 0.
2ǫo
Hence, in vector form
ρs
E = x̂ sgn(x),
2ǫo
where sgn(x) is the signum function, equal to ±1 for x ≷ 0.

Note that the macroscopic field calculated above is discontinuous at x = 0 plane

containing the surface charge ρs , and points away from the same surface on both
sides.

1
 z
y

Example 2: Point charges Q are distributed throughout an infinite slab of width W A

ρx
Ex(x) =
located over − W2 < x < W2 with an average charge density of ρ C/m3. Determine ǫo
the macroscopic electric field E of the charged slab inside and outside. ρ x

Solution: Symmetry arguments based on Coulomb’s law once again indicates that we W W
expect a solution of the form E = x̂Ex (x) where Ex (x) is an odd function of x. −
2 2
Ex(x) ρW
In that case, applying Gauss’s law with a cylindrical surface S having circular caps 2ǫo
of area A parallel to x = 0 extending between −x and x < W2 , we obtain
I W W x
−
D · dS = QV ⇒ ǫo Ex (x)A − ǫo Ex (−x)A = ρ2xA, 2 2
S

which leads, with Ex (−x) = −Ex (x), to

ρx W
Ex (x) = for 0 < x < .
ǫo 2
W
For x > 2
, I
D · dS = QV ⇒ ǫo Ex (x)A − ǫo Ex (−x)A = AW ρ,
S
leading to
ρW W
Ex (x) = for x > .
2ǫo 2
These results can be combined as
 ρW
W
−x̂ 2ǫo , for x < − 2

E = x̂Ex (x) = x̂ ρx
ǫ o
, for − W2 < x < W
2
 ρW
for x > W2 .

x̂ 2ǫo ,

2
Note that the field solution depicted in the margin in terms of Ex (x) plot is a con-
tinuous function of x as opposed to the discontinuous Ex (x) solution obtained in
Example 1 for the macroscopic field of a surface charge.

• In future calculations of electrostatic fields, we can use our previous

results, namely

– Coulomb field
Q
E = r̂ of a point charge Q,
4πǫor2
– Field
λ
E = r̂ of constant line density λ,
2πǫor
– Field
ρs
E = x̂ sgn(x) of constant surface density ρs ,
2ǫo
– Field
ρx
E = x̂ of constant volume density ρ
ǫo
as building blocks — that is, the above field equations can be super-
posed to determine the field structure of charge distributions ρ(x, y, z)
that can be expressed as superpositions of simpler charge distributions
with known field structures. Some examples...

3
Example 3: Consider a pair of surface charges ρs > 0 and −ρs C/m2 of equal mag-
nitudes placed on x = − W2 and x = W2 surfaces. Determine the electric field of
this charge distribution depicted in the margin.

Solution: The field of charge density ρs C/m2 on x = − W2 plane should be

ρs W
E+ = x̂ sgn(x + ),
2ǫo 2
pointing away from the discontinuity surface at x = − W2 on both sides. Likewise,
the field of charge density −ρs C/m2 on x = W2 plane should be
ρs W
E− = −x̂ sgn(x − ),
2ǫo 2
ρs
pointing toward x = W
surface from both sides. Superposing the two fields, we E= x̂
2 ǫo
find that z
(
x̂ ρǫos , for − W2 < x < W
2 , ρs x + -y

E = E+ + E− = = x̂ rect( )
0, otherwise, ǫo W
x
as depicted in the margin.
ρs > 0 −ρs
Note that the field lines of our solution point from positive charges on one surface to
the negative charges resting on the other surface — this field has the structure Ex(x)
ρs
of fields encountered in parallel plate capacitors that we will be studying soon. ǫo

W W x
−
2 2

4
 z
−ρ1 < 0 ρ2 > 0

Example 4: An infinite charged slab of width W1, located over −W1 < x < 0, has
E
a negative volumetric charge density of −ρ1 C/m3 , ρ1 > 0. A second slab of - +

width W2 and positive charge density ρ2 is located over 0 < x < W2 as shown x
in the margin. Compute the electric field of this static charge configuration if
W1ρ1 = W2ρ2 , implying that the entire system is charge neutral (i.e., a net charge
−W1 W2
of zero).
E1x(x)
Solution: We note that the field of slab W1 can be written as ρ1W1
 2ǫo
ρ1 W1
x̂ 2ǫo , for x < −W1


W1 −W1 x
ρ (x+ 2 )
E1 = −x̂ 1 ǫo , for − W1 < x < 0

−x̂ ρ1 W1 ,

for x > 0
2ǫo
E2x(x)
as depicted in the margin. Likewise, the field of slab W2 is
 ρ2W2
ρ2 W2
−x̂ 2ǫoW, for x < 0

 2ǫo
2
E2 = x̂ ρ2 (x−ǫo
2 )
, for 0 < x < W2 W2 x
 ρ W
x̂ 2 2 ,

for x > W2.
2ǫo

Note that field strengths ρ2ǫ

1 W1
o
and ρ2ǫ
2 W2
o
showing up in the expressions for E1 and E2 Ex(x)
are equal because of the charge neutrality condition W1ρ1 = W2ρ2 . −W1 W2
x
Consequently, when we superpose E1 and E2 , the fields cancel out outside the region
−W1 < x < W2, so that the total field becomes (as depicted in the margin)
 ρ (x+W ) ρ1W1
−x̂ ǫo , for − W1 < x < 0
1 1 −
ǫo

E = E1 + E2 = x̂ ρ2 (x−W
ǫ o
2)
, for 0 < x < W2

0, otherwise.


5
 Gauss’ Law in terms of
• Charge density formalism which we find convenient to use for macro- charge density:
scopic field calculations can also be “adjusted” to describe the distri-
butions of isolated point charges via the use of impulses or delta I Z
functions in space. D · dS = ρdV
S V
– For example
ρ(x, y, z) = Qδ(x − xo)δ(y − yo )δ(z − zo)
can be regarded as a 3D volumetric charge density function rep- ρ(x, y, z) = Qδ(x)δ(y)δ(z)
resenting a point charge Q located at a coordinate z
r = (x, y, z) = (xo, yo, zo) ≡ ro.
y
◦ This is justified because we can regard δ(x − xo) to be zero Q
x
everywhere except at x = xo. By extension, the product
δ(x − xo)δ(y − yo )δ(z − zo)
3D impulse here
is zero everywhere except at r = ro = (xo, yo, zo ) — therefore where point charge
Q is localized over
the density function ρ(x, y, z) defined above behaves correctly a region of zero
volume
to indicate the absence of charges everywhere with the ex-
ception of ro. Furthermore, the area property of the impulse
implies that the volume integral of the charge density yields
Z Z Z Z
ρdV = Qδ(x − xo)δ(y − yo )δ(z − zo)dxdydz = Q

as it should.

6
 ◦ Notice that the shifted impulses δ(x − xo), etc., must have ρ(x, y, z) = ρs (y, z)δ(x − xo)
m−1 units in order to maintain dimensional consistency in the
above expression. ρs > 0
– Another example is z
ρ(x, y, z) = ρs(y, z)δ(x − xo) y
representing a surface charge density of ρs(y, z) C/m2 on x = xo
plane.
xo x

Example 5: Figure in the margin depicts (for the d = 1) the Ê-field of a pair of
charges ±Q located at (0, 0, ± d2 ) derived from
1.0
Q(r − d2 ẑ) −Q(r + d2 ẑ)
E(r) = +
4πǫo |r − d2 ẑ|3 4πǫo |r + d2 ẑ|3
0.5

Q (x, y, z − d2 ) (x, y, z + d2 )
= [ − ] V/m.
4πǫo |(x, y, z − d2 )|3 |(x, y, z + d2 )|3

z
0.0

R
Determine the electric flux xy E · dS across the entire xy-plane using dS =
−ẑdxdy. -0.5

Solution: Because of linearity, the flux we want to calculate equals the sum of the flux -1.0

due to charge Q at (0, 0, d2 ) above xy-plane and the flux due to charge −Q at -1.0 -0.5 0.0
x
0.5 1.0

(0, 0, − d2 ) above xy-plane.

7
 E · dS = ǫQo for any S surrounding Q, we can, by symmetry,
H
Since by Gauss’s law S
infer that
Q
Z
E · (−ẑdxdy) =
xy 2ǫo
when only charge Q is considered — the logic here is, half of flux S E · dS = ǫQo
H
emanating from charge Q should go up and the remaining half H should go−Q
down
crossing the xy-plane in downward direction. Likewise, since S E · dS = ǫo for
any S surrounding −Q, again by symmetry, we can infer
Q
Z
E · (−ẑdxdy) =
xy 2ǫo
due to charge −Q only — the logic in this case is, half of flux ǫQo “entering” charge
−Q is “coming from” above crossing the xy-plane in downward direction.

Thus, by superposition, we find total

Q Q Q
Z
E · (−ẑdxdy) = + = .
xy 2ǫo 2ǫo ǫo

The above result can be confirmed directly by evaluating the integral

Q (x, y, − d2 ) (x, y, d2 )
Z Z
E(x, y, 0) · (−ẑdxdy) = [ d 3
− ] · (−ẑdxdy)
xy xy 4πǫo |(x, y, − 2 )| |(x, y, d2 )|3
Q d Qd ∞ r
Z Z
= dxdy = dr
4πǫo xy |(x, y, − d2 )|3 2ǫo r=0 (r2 + ( d2 )2)3/2
Q
= .
ǫo
p
Just before the last step we have replaced dxdy by rdrdφ, where r ≡ x2 + y 2 , and
carried out the φ integration before completing the r integration as a last step (which
you should verify).

8
4 Divergence and curl
Expressing the total charge QV contained in a volume V as a 3D volume
integral of charge density ρ(r), we can write Gauss’s law examined during
the last few lectures in the general form
I Z
D · dS = ρdV.
S V

This equation asserts that the flux of displacement D = ǫoE over any closed
surface S equals the net electrical charge contained in the enclosed volume
V — only the charges included within V affect the flux of D over surface
S, with charges
H outside surface S making no net contribution to the surface
integral S D · dS.
z
• Gauss’s law stated above holds true everywhere in space over all sur- (x, y, z + ∆z)
faces S and their enclosed volumes V , large and small. y (x, y + ∆y, z)

(x, y, z) (x + ∆x, y, z)
• Application of Gauss’s law to a small volume ∆V = ∆x∆y∆z sur-
rounded by a cubic surface ∆S of six faces, leads, in the limit of van-
ishing ∆x, ∆y, and ∆z, to the differential form of Gauss’s law expressed
x
in terms of a divergence operation to be reviewed next:

– Given a sufficiently small volume ∆V = ∆x∆y∆z, we can assume

that Z
ρdV ≈ ρ∆x∆y∆z.
∆V

1
 – Again under the same assumption
I
D·dS ≈ (Dx|2−Dx|1)∆y∆z+(Dy|4−Dy|3)∆x∆z+(Dz|6−Dz|5)∆x∆y
S

with reference to displacement vector components like Dx|2 shown

on cubic surfaces depicted in the margin. Gauss’s law demands
the equality of the two expressions above, namely (after dividing
both sides by ∆x∆y∆z)
Dx|2 − Dx|1 Dy|4 − Dy|3 Dz|6 − Dz|5
+ + ≈ ρ,
∆x ∆y ∆z
in the limit of vanishing ∆x, ∆y, and ∆z. In that limit, we obtain
y
∂Dx ∂Dy ∂Dz z 6
+ + = ρ,
∂x ∂y ∂z 4
2
1
which is known as differential form of Gauss’s law.
5

A more compact way of writing this result is (x, y, z) 3

x
∇ · D = ρ,

where the operator

∂ ∂ ∂
∇≡( , , ),
∂x ∂y ∂z
known as del, is applied on the displacement vector

D = (Dx , Dy , Dz )

2
 ∂
following the usual dot product rules, except that the product of ∂x and Dx,
for instance, is treated as a partial derivative ∂D
∂x . In the left side above
x

∂Dx ∂Dy ∂Dz

∇·D= + + (divergence of D)
∂x ∂y ∂z
is known as divergence of D.

Example 1: Find the divergence of D = x̂5x + ŷ12 C/m2

Solution: In this case

Dx = 5x, Dy = 12, and Dz = 0.
Therefore, divergence of D is
∂Dx ∂Dy ∂Dz
∇·D = + + 3

∂x ∂y ∂z
∂ ∂ ∂ 2

= (5x) + (12) + (0)

∂x ∂y ∂z 1

C
= 5 + 0 + 0 = 5 3.
m

z
0

Note that the divergence of vector D is a scalar quantity which is the volumetric
-1

charge density in space as a consequence of Gauss’s law (in differential form). -2

-3
-3 -2 -1 0 1 2 3
x

3
 z
−ρ1 < 0 ρ2 > 0

Example 2: Find the divergence ∇ · E of electric field vector -

E +
 ρ (x+W ) x
−x̂ ǫo , for − W1 < x < 0
1 1

E = x̂ ρ2 (x−W
ǫ o
2)
, for 0 < x < W2
 −W1 W2
0, otherwise,

E1x(x)
from Example 4, last lecture (see margin figures). ρ1W1
2ǫo

Solution: In this case Ey = Ez = 0, and therefore the divergence of E is −W1 x

 ρ (x+W ) 
 − ǫo ,
1 1
 − ρǫo1 , for − W1 < x < 0
∂Ex ∂ 
ρ2 (x−W2 )

∇·E= = ǫ , = ρǫo2 , for 0 < x < W2 , E2x(x)
∂x ∂x  o 
0, 0, otherwise,
 
ρ2W2
2ǫo
which provides us with ρ(r)/ǫo of Example 4 from last lecture (in accordance W2 x
with Gauss’s law).

Ex(x)
W2
• Summarizing the results so far, Gauss’s law can be expressed in integral
−W1
x
as well as differential forms given by
I Z
ρ1W1
D · dS = ρdV ⇔ ∇ · D = ρ. −
ǫo
S V

– The equivalence of integral and differential forms implies that (af-

ter integrating the differential form of the equation on the right

4
 over volume V on both sides)
I Z
D · dS = ∇ · D dV
S V

which you may recall as the divergence theorem from MATH

241. Divergence thm.
– Note that according to divergence theorem, we can interpret di-
vergence as flux per unit volume.
– We can also think of divergence as a special type of a derivative
applied to vector functions which produces non-zero scalar results
(at each point in space) when the vector function has components
which change in the direction they point.
◦ A second type of vector derivative known as curl which we re-
view next complements the divergence in the sense that these
two types of vector derivatives collectively contain maximal
information about vector fields that they operate on:

Given their curl and divergences, vector fields can be uniquely recon-
structed in regions V of 3D space provided they are known at the
bounding surface S of region V , however large (even infinite) S and V
may be — this is known as Helmholtz theorem (proof outlined in
Lecture 7).

• The curl of a vector field E = E(x, y, z) is defined, in terms of the del

5
 operator ∇, like a cross product

x̂ ŷ ẑ
∂ ∂ ∂ ∂ ∂ ∂
∇ × E ≡ ( , , ) × (Ex , Ey , Ez ) = ∂x ∂y ∂z (curl of E)
∂x ∂y ∂z
Ex Ey Ez
∂Ez ∂Ey ∂Ez ∂Ex ∂Ey ∂Ex
= x̂( − ) − ŷ( − ) + ẑ( − ).
∂y ∂z ∂x ∂z ∂x ∂y

Example 3: Find the curl of the vector field

E = x̂ cos y + ŷ1

Solution: The curl is 4

x̂ ŷ ẑ
∂ ∂ ∂
∇×E = ∂x ∂y ∂z
2

cosy 1 0
∂ ∂ ∂ ∂ ∂ ∂ 0

y
= x̂( 0 − 1) − ŷ( 0 − cos y) + ẑ( 1 − cos y)
∂y ∂z ∂x ∂z ∂x ∂y
= x̂0 − ŷ0 + ẑ(0 + sin y) = ẑ sin y -2

which is another vector field.

-4
-4 -2 0 2 4
x

The diagram in the margin depicts E = x̂ cos y + ŷ1 as a vector map

superposed upon a density plot of |∇ × E| = |ẑ sin y| = | sin y| indicating the
strength of the curl vector ∇ × E (light color corresponds large magnitude).

6
It is apparent that curl ∇ × E is stronger in those regions where E is rapidly
varying in directions orthogonal to the direction of E itself.

• As the above example demonstrates the curl of a vector field is in

general another vector field.

– The only exception is if the curl is identically 0 at all positions

r = (x, y, z)!
◦ In that case, i.e., if ∇ × E = 0, vector field E is said to be
curl-free.

IMPORTANT FACT: All static electric fields E, obtained from

Coulomb’s law, and satisfying Gauss’s law ∇ · D = ρ with static
charge densities ρ = ρ(r), are also found to be curl-free without
exception.

• The proof of curl-free nature of static electric fields can be given by

first showing that Coulomb field of a static charge is curl-free, and then
making use of the superposition principle along with the fact that the
curl of a sum must be the sum of curls — like differentiation, “taking
curl” is a linear operation.

– You should try to show that ∇ × E = 0 with the Coulomb field

of a point charge Q located at the origin.

7
 ◦ The calculation is slightly more complicated than the following z
example (although similar in many ways) where we show that λ
the static electric field of an infinite line charge is curl-free. S r
λ
Er =
L 2πǫor
y
Example 4: Recall that the static field of a line charge λ distributed on the z-axis is φ
x
λ
E(x, y, z) = r̂ ,
2πǫo r
where
x y
r2 = x2 + y 2 and r̂ = x̂ cos φ + ŷ sin φ = ( , , 0).
r r
Show that field E satisfies the condition ∇ × E = 0.

Solution: Clearly, we can express vector E as

λ x y
E= ( , , 0).
2πǫo r2 r2
Since the components rx2 and ry2 of the vector are independent of z, the corre-
sponding curl can be expanded as
x̂ ŷ ẑ x̂ ŷ ẑ
∂ ∂ ∂ λ ∂ ∂ ∂ λ ∂ y ∂ x
∇×E= ∂x ∂y ∂z = ∂x ∂y ∂z = ẑ( − ).
2πǫo x y 2πǫo ∂x r2 ∂y r2
Ex Ey Ez r2 r2 0
But,
∂ y ∂ x ∂ 1 ∂ 1 −2x −2y
− = y − x = y − x = 0,
∂x r2 ∂y r2 ∂x r2 ∂y r2 r4 r4
so ∇ × E = 0 as requested.

8
5 Curl-free fields and electrostatic potential
• Mathematically, we can generate a curl-free vector field E(x, y, z) as
∂V ∂V ∂V
E = −( , , ),
∂x ∂y ∂z
by taking the gradient of any scalar function V (r) = V (x, y, z). The
gradient of V (x, y, z) is defined to be the vector
∂V ∂V ∂V
∇V ≡ ( , , ),
∂x ∂y ∂z
pointing in the direction of increasing V ; in abbreviated notation, curl-
free fields E can be indicated as

E = −∇V.

– Verification: Curl of vector ∇V is

x̂ ŷ ẑ
∂ ∂ ∂
∇ × (∇V ) = ∂x ∂y ∂z = x̂0 − ŷ0 − ẑ0 = 0.
∂V ∂V ∂V
∂x ∂y ∂z

– If E = −∇V represents an electrostatic field, then V is called

the electrostatic potential.
◦ Simple dimensional analysis indicates that units of electro-
static potential must be volts (V).

1
 – The prescription E = −∇V , including the minus sign (optional,
but taken by convention in electrostatics), ensures that electro-
static field E points from regions of “high potential” to “low po-
tential” as illustrated in the next example. Electrostatic fields E
point from regions of
“high V ” to “low V ”
Example 1: Given an electrostatic potential

V (x, y, z) = x2 − 6y V
4

in a certain region of space, determine the corresponding electrostatic field E =

−∇V in the same region.
2

Solution: The electrostatic field is

0
∂ ∂ ∂
E = −∇(x2 − 6y) = −( , , )(x2 − 6y) = (−2x, 6, 0) = −x̂ 2x + ŷ6 V/m.
∂x ∂y ∂z
-2

Note that this field is directed from regions of high potential to low potential. Also note
that electric field vectors are perpendicular everywhere to “equipotential” contours. -4

-4 -2 0 2 4

Light colors indicate “high V ”

Given an electrostatic potential V (x, y, z), finding the corresponding elec- dark colors “low V ”
trostatic field E(x, y, z) is a straightforward procedure (taking the negative
gradient) as already illustrated in Example 1.
The reverse operation of finding V (x, y, z) from a given E(x, y, z) can be
accomplished by performing a vector line integral
Z o
E · dl
p

2
in 3D space, since, as shown below, such integrals are “path independent” for
z p = (xp, yp, zp)
curl-free fields E = −∇V .
Ej
• The vector line integral ∆lj
y
Z o C
E · dl C′
p
o = (xo, yo, zo)
over an integration path C extending from a point p = (xp, yp, zp ) in
3D space to some other point o = (xo, yo, zo ) is defined to be x

– the limiting value of the sum of dot products Ej ·∆lj computed over
all sub-elements of path C having incremental lengths |∆lj | and
unit vectors ∆lj /|∆lj | directed from p towards o — the limiting
value is obtained as all |∆lj | approach zero (i.e., with increasingly
finer subdivision of C into |∆lj | elements).

• Computation of the integral (see example below) involves the use of

infinitesimal displacement vectors
dl = x̂dx + ŷdy + ẑdz = (dx, dy, dz)
and vector dot product
E · dl = (Ex, Ey , Ez ) · (dx, dy, dz) = Ex dx + Ey dy + Ez dz.
The integral
Z o Z o
E · dl = (Exdx + Ey dy + Ez dz)
p p

3
 will in general be path dependent except for when E is curl-free. Curl-free: path-independent
line integrals
Cu
1.0

Example 2: The field E = x̂y ± ŷx is curl-free with the + sign, but not with − as
0.8
verified below by computing ∇ × E. Calculate the line integral of E (for both
signs, ±) from a point o = (0, 0, 0) to point p = (1, 1, 0) for two different paths
0.6
C going through points u = (0, 1, 0) and l = (1, 0, 0), respectively (see margin). Cl

z
Solution: First we note that 0.4

x̂ ŷ ẑ 0.2
∂ ∂ ∂
∇ × (x̂y ± ŷx) = ∂x ∂y ∂z = ẑ(±1 − 1)
y ±x 0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
x

which confirms that E = x̂y ± ŷx is curl-free with with + sign, but not with −. “Curly”: path-dependent line
In either case, the integral to be performed is integrals
Z p Z p Z p Cu
1.0
E · dl = (Exdx + Ey dy + Ez dz) = (y dx ± x dy).
o o o
0.8

For the first path Cu going through u = (0, 1, 0), we have

Z p Z 1 Z 1 0.6

(y dx ± x dy) = (±x) dy |x=0 + y dx|y=1 = 0 + 1 = 1. Cl

z
o y=0 x=0 0.4

For the second path Cl going through l = (1, 0, 0), we have

0.2
Z p Z 1 Z 1
(y dx ± x dy) = y dx|y=0 ± x dy |x=1 = 0 ± 1 = ±1.
o x=0 y=0 0.0

Rp 0.0 0.2 0.4 0.6 0.8 1.0

Clearly, the result shows that the line integral o E · dl is path independent for x

E = x̂y + ŷx which is curl-free, and path dependent for E = x̂y − ŷx in which
case ∇ × E 6= 0.

4
• The mathematical reason why curl-free fields have path-independent
line integrals is because in those occasions the integrals can be written
in terms of exact differentials:

– for curl-free E = x̂y + ŷx we have E · dl as an exact Rdifferential

p
ydx + xdy = d(xy) of the function xy, in which case o E · dl =
xy|po = (1 · 1 − 0 · 0) = 1 over all paths.
– for E = x̂y − ŷx with ∇ × E = −2ẑ 6= 0, on the other hand,
E · dl = ydx − xdy does not form an exact differential −dV , and
thus there is no path-independent integral −V |po, nor an underlying
potential function V . Z
z Vp =
o
· dl
p E
E·dl is guaranteed to be an exact differential if E = −∇V = (− ∂V ∂V ∂V
∂x , − ∂y , − ∂z ), E(r) y
since in that case the differential of V (x, y, z), namely dl

∂V ∂V ∂V
dV ≡ dx+ dy+ dz, is precisely −Ex dx−Ey dy−Ez dz = −E·dl. Vo = 0
∂x ∂y ∂z
– In that case x
Z o Z o Z p
E · dl = − dV = dV = Vp − Vo
p p o

is independent of integration path; thus, if we we call o the “ground”,

and set Vo = 0, then Z o
Vp = E · dl
p
denotes the potential drop from (any) point p to ground o.

5
 • The physical reason why this integral formula for potential Vp works z Vp =
Z
o
· dl
p E
with any integration path is the principle of energy conservation: E(r)
y
Ro dl
– integral p E · dl, namely the “voltage drop” from p to o, repre-
sents the work done per unit charge by the field E in moving Vo = 0
charges from location p to location o1, so if the line integral were
x
path-dependent (in reaching from p to o) there would be ways As long as E is curl-free, line
of creating net energy by making a charge q follow special closed integral is path-independent and
produces the voltage drop from
point p to "ground" o.
paths within the electrostatic field E, in violation of the general
principle of energy conservation (that permits energy conversion
but not creation or destruction).

1
Either to increase the kinetic energy of the charge if charge transport from p to o is unimpeded (as
for a test charge accelarating between a pair of capacitor plates) or else in pushing the charge against
frictional forces (as through a resistive wire) both at the expense of the energy stored in the field. On the
other hand, work done (i.e., the voltage drop) by the field would be negative if charges q > 0 were moved
from p to o against the local electric field (as within a battery), in which case there would be a positive
voltage rise from p to o representing energy gain for the field per unit charge transported from p to o.

6
 z Z
p
Vp = − o E · dl
y
Example 3: Given that Vo = V (0, 0, 0) = 0 and Z
V
E = 2xx̂ + 3z ŷ + 3(y + 1)ẑ ,
m Y
o
determine the electrostatic potential Vp = V (X, Y, Z) at point p = (X, Y, Z) in X x
volts.
A voltmeter with its (+/red)
Solution: Assuming that the field is curl-free (it is), so that any integration path can probe contacting point p and
be used, we find that its (-/black) probe contact-
Z o Z p Z p ing point o would display (by
Vp = E · dl = − E · dl = − (2x dx + 3z dy + 3(y + 1) dz) analog or digital means) the
p o o numerical value of
Z X Z Y Z Z Z o
= − 2x dx|y,z=0 − 3z dy|x=X,z=0 − 3(y + 1) dz|x=X,y=Y Vp = E · dl
p
0 0 0
2
= −X − 0 − 3(Y + 1)Z. with the integration path
consisting of the path defined
This implies by the probe wires. The volt-
V (x, y, x) = −x2 − 3(y + 1)z V. meter reading would be inde-
pendent of the path config-
Note that uration when the field E is
electrostatic.
−∇(−x2 − 3(y + 1)z) = ∇(x2 + 3(y + 1)z)
For the voltmeter not to per-
= x̂2x + ŷ3z + ẑ3(y + 1)
turb the field it is probing,
its input impedance need to
yields the original field E, which is an indication that E is indeed curl-free. be much greater than the
impedance between points p
and o.

7
Alternate Solution — Exact Differential Method: Note that

E · dl = (2xx̂ + 3z ŷ + 3(y + 1)ẑ) · (x̂dx + ŷdy + ẑdz)

= 2xdx + 3zdy + 3(y + 1)dz = 2xdx + 3(ydz + zdy) + 3dz
= d(x2 + 3yz + 3z) = −dV.

Therefore
V (x, y, z) = −x2 − 3yz − 3z + C,
where the integration constant C should chosen so that V (0, 0, 0) = 0. The result
is
V (x, y, z) = −x2 − 3(y + 1)z

as before.

8
Example 5: According to Coulomb’s law electrostatic field of a proton with charge
Q = e (where −e is electronic charge) located at the origin is given as z
e
E= r̂,
4πǫo r2
where z=r
p (x, y, z)
r= x2 + y 2 + z 2 and r̂ = .
r
Determine the electrostatic potential field V established by charge Q = e with
the provision that V → 0 as r → ∞ (i.e., ground at infinity).

Solution: Field E and its potential V will exhibit spherical symmetry in this problem. e y
Therefore, with no loss of generality, we can calculate the line integral from a x
point p at a distance r from the origin to a point o at ∞ (the specified ground)
along, say, the z-axis. Approaching the problem that way, the potential drop
Or else (exact differential
from r to ∞ is method):
Z ∞
e e −e
V (r) = 2
ẑ · ẑdz E·dl = ( r̂)·r̂dr = d( ) = −dV
z=r 4πǫo z 4πǫo r 2 4πǫo r
e ∞ e
= − |r = . leading to
4πǫo z 4πǫo r
e
V (r) =
4πǫo r
(using an integration con-
• To convert electrostatic potential Vp (in volts) at any point p to poten- stant of zero).
tial energy of a charge q brought to the same point, it is sufficient to
multiply Vp with q (or just the sign of q, depending on which energy
units we want to use — see the next example).

9
Example 6: In view of Example 5, what are the potential energies of a proton e and
an electron −e placed at distance r = a away from the proton at the origin,
where distance
4πǫo ~2
a≡ 2 = 0.529 × 10−10 m
e me
stands for Bohr radius — it is the mean distance of the ground state electron in
a hydrogen atom from the center of the atom. Recall that e = 1.602 × 10−19 C
and ǫo ≈ 10−9/36π F/m.

Solution: Let’s first evaluate the potential V (r) at r = a:

e (1.6 × 10−19)36π × 109 9 × 1.6
V (a) = ≈ −10
= = 27.2 V.
4πǫo a 4π × 0.53 × 10 0.53

For the proton, potential energy in Joules is calculated by multiplying V (a) = 27.2
V with q = e = 1.602 × 10−19 C. However, by referring to 1.602 × 10−19 J of
energy as 1 eV (electron-volt), it is more convenient to refer to potential energy
eV (a) of the proton at r = a as

eV (a) = 27.2 eV.

Likewise, for a particle with charge q = −e, i.e., an electron, potential energy at the
same location is
−eV (a) = −27.2 eV.

10
6 Circulation and boundary conditions z
y
Since curl-free static electric fields have path-independent line integrals, it dS o=p
follows that over closed paths C (when points p and o coincide)
C
S
I
E · dl = 0, ∆lj
C Ej
x
H
where the C E · dl is called the circulation of field E over closed path C Closed loop integral over path
bounding a surface S (see margin). C enclosing surface S.
Note that the area increment
dS of surface S is taken by
convention to point in the
right-hand-rule direction
with respect to "circulation"
Example 1: Consider the static electric field variation direction C.

ρx
E(x, y, z) = x̂ z y
ǫo
that will be encountered within a uniformly charged slab of an infinite extent in
y and z directions and a finite width in xH direction centered about x = 0. Show (-3,4,0) (3,4,0)

that this field E satisfies the condition C E · dl = 0 for a rectangular closed

path C with vertices at (x, y, z) = (−3, 0, 0), (3, 0, 0), (3, 4, 0), and (−3, 4, 0)
traversed in the order of the vertices given. (-3,0,0)
C (3,0,0) x

Solution: Integration path C is shown in Hthe figure in the margin. With the help of
the figure we expand the circulation C E · dl as
Z 3 Z 4 Z −3 Z 0
ρx ρ3 ρx ρ(−3)
E = x̂ · x̂dx + x̂ · ŷdy + x̂ · x̂dx + x̂ · ŷdy
x=−3 ǫo y=0 ǫo x=3 ǫo y=4 ǫo
Z 3 Z −3
ρx ρx
= dx + 0 + dx + 0 = 0.
x=−3 ǫo x=3 ǫo

1
 H
Note that in expanding C E · dl above for the given path C, we took dl as x̂dx
and ŷdy in turns (along horizontal and vertical edges of C, respectively) and
ordered the integration limits in x and y to traverse C in a counter-clockwise
direction as indicated in the diagram.
z
y

• Vector fields E having zero circulations over all closed paths C are dS
known as conservative fields (for obvious reasons having to do with C
S
their use in modeling static fields compatible with conservation theo- dl
rems). E
x
– The concepts of curl-free and conservative fields overlap, that is STOKE’S THM:
Circulation of E around close
I path C equals the flux over
enclosed surface S of the curl
E · dl = 0 ⇔ ∇ × E = 0 of E taken in direction of dS.

C dS points in right-hand-rule
direction with respect to
over all closed paths C and at each r. "circulation" direction C.

• The above relationship between circulation and curl is also a conse-

quence of Stoke’s theorem (discussed in MATH 241) which asserts
that I Z Stoke’s thm.
E · dl = ∇ × E · dS,
C S
where

– the integration surface S on the right is bounded by the closed

integration contour C of the left side, and

2
 – the incremental area element dS on the right points across area S z
in the direction indicated by a right-hand rule as follows: y
dS
Point your right thumb in chosen circulation direction C; then your
right fingers point through surface S in the direction that should be C
S
adopted for dS. dl
H E
– Given Stoke’s theorem, C E · dl = 0 follows immediately for all x
STOKE’S THM:
C, if ∇ × E = 0 is true over all r. Circulation of E around close
path C equals the flux over
enclosed surface S of the curl
• Stoke’s theorem clearly implies that curl is circulation per unit of E taken in direction of dS.

area, just as the divergence theorem showed that divergence is flux dS points in right-hand-rule
direction with respect to
per unit volume. "circulation" direction C.

– The only difference is, curl also has a direction, which is the normal
C
unit of the plane that contains the maximal value of circulation z 4
2
per unit area found at that location (over all possible orientations (x, y, z + ∆z) dS = ∆y∆z x̂

of dS). y
1 (x, y + ∆y, z)
3
(x, y, z)
• We will verify Stoke’s thm after explaining the circulation per unit
area notion in steps:

– Let us first calculate the circulation of a vector field E taken about

x
an arbitrary point (x, y, z) on a constant x plane around a square
contour with small edge dimensions ∆y and ∆z parallel to y and
z axes as shown in the margin.

3
 C
– For a small rectangular contour “Cx ” on a constant x plane with z 4
2
sufficiently small ∆y and ∆z dimensions parallel to y and z axes (x, y, z + ∆z) dS = ∆y∆z x̂
(see figure in the margin), we have y
1 (x, y + ∆y, z)
I 3
(x, y, z)
E · dl ≈ Ez|2∆z − Ey|4∆y − Ez|1 ∆z + Ey|3∆y
Cx
= (Ez|2 − Ez|1 )∆z − (Ey|4 − Ey|3)∆y.

It follows that x
1 Ez|2 − Ez|1 Ey|4 − Ey|3
I
E · dl ≈ ( − )
∆y∆z Cx ∆y ∆z
and
1 ∂Ez ∂Ey
I
lim E · dl = ( − ) = x̂ · ∇ × E,
∆y,∆z→0 ∆y∆z Cx ∂y ∂z
meaning that the x component of ∇ × E is the circulation of E
per unit area on a constant x surface.
– Likewise, y and z components of ∇ × E are circulations of E per
unit area on constant y and z surfaces, and in general
1 1 1
I I I
∇×E = lim ( E·dl, E·dl, E·dl).
∆x,∆y,∆z→0 ∆y∆z C ∆x∆z Cy ∆x∆y Cz
x

– Furthermore, based on the above result, we can recognize that vec-

tors ∇ × E point everywhere in directions perpendicular to planes
of maximum circulations per unit area in the E field and have

4
 magnitudes corresponding to the maximum values of circulations C
per unit area at every point. S

• Now, to confirm Stoke’s theorem

I Z
E · dl = ∇ × E · dS dl
C S E

pertinent for a closed path C and its enclosed surface S, we will make Sum of circulations over small
squares cancel in the interior
use of the diagram shown in the margin. edges and only survive around the
exterior path C. This way,
circulation around C matches
the sum of the fluxes of curl E
– The circulations over small squares shown in the diagram are ap- calculated over the small squares.

proximately equal to the products of their areas and the normal Laws of
components of ∇ × E calculated at the center points (basd on electrostatics:
what we learned above).
– When allR such circulations covering surface S are added up, the
∇×E = 0
result is S ∇ × E · dS in the limit of vanishing size for the squares,
H ∇ · ǫo E = ρ
– as well as C E · dl because in the grand sum of the circulations
They also apply “quasi-statically”
over all the squares, all the contributions mutually cancel out (like over a region of dimension L
the overlapping edges of red and blue squares) except for those when a time-varying field source
ρ(r, t) has a time-constant τ much
calculated along the periphery C! longer than the propagation time
delay L/c of E(r, t) field varia-
tions across the region (c is the
We can now summarize the general constraints governing static electric fields speed of light).
as
In electro-quasistatics (EQS)
∇ × E(r) = 0, ∇ · D(r) = ρ(r), where D(r) = ǫoE(r). E(r, t) will be accompanied by
a slowly varying magnetic field
B(r, t) (to be studied starting in
Lecture 12).
5
• Vector fields E(r) and D(r) governed by these equations will in general
be continuous functions of position coordinates r = (x, y, z) except at
boundary surfaces where charge density function ρ(r) requires a repre-
sentation in terms of a surface charge density ρs (r).

– For instance, according to our earlier results, static electric field

of a charge density (see sketch at the margin)
2

ρ(r) = ρs δ(z)
1

would be
ρs ρs 0

z
E(r) = ẑ sgn(z) ⇒ D(r) = ẑ sgn(z).
2ǫo 2
-1

◦ Consider a superposition of these fields with fields Eo(r) and

Do(r) = ǫoEo(r) produced by arbitrary continuous sources, -2

namely (macroscopic) fields

-2 -1 0 1 2
x

ρs ρs
E(r) = ẑ sgn(z)+Eo(r) and D(r) = ẑ sgn(z)+ǫoEo(r).
2ǫo 2

Since fields Eo(r) and Do(r) vary continuously, these field expressions
must satisfy

ẑ · (D+ − D− ) = ρs and ẑ × (E+ − E− ) = 0

where
E+ ≡ E(x, y, 0+) and E− ≡ E(x, y, 0− )

6
 D+
refer to limiting values of E at z = 0 plane from above and below,
ẑ D−
respectively, and likewise for z=0

D+ ≡ D(x, y, 0+) and D− ≡ D(x, y, 0− ).

• The above “boundary condition equations” can be written in a more

general form (see margin for justification) as
D+
n̂ · (D+ − D− ) = ρs and n̂ × (E+ − E− ) = 0 n̂
w D−
where n̂ denotes a unit vector normal to any surface of an arbitrary
orientation carrying a surface charge density ρs, while field vectors with
superscripts + and − indicate limiting values of fields measured on Constraint
I
either side of the charged surface (with n̂ pointing from − to +). E · dl = 0
C

– The equations can be further simplified as around the dotted path yields

Et+ = Et−
Dn+ − Dn− = ρs and Et+ = Et−
in w → 0 limit.

where Dn and Et refer to normal component of D and tangential

component of E, respectively. Clearly, these boundary condi- Gauss’s law
tions say that at any surface S, I
D · dS = QV
◦ tangential component of electric field E needs to be continu- S

ous, but applied over the dotted volume (seen in

profile) yields
◦ normal component of D can change by an amount equal to
Dn+ − Dn− = ρs
the charge density ρs carried by the surface.
in w → 0 limit.
7
 D = 0 for x < 0.
ρs = 2C/m2 ρso = ?
z
Example 2: Measurements indicate that D = 0 in the region x < 0.

Also, x = 0 and x = 5 m planes contain surface charge densities of ρs = 2 C/m2 and

ρso , respectively. y
Determine ρso and D for −∞ < x < ∞ if there are no other charge distributions.

Solution: Since the normal component of D must increase by ρs = 2 C/m2 when x

x=5m
we cross the charged surface x = 0, we must have D = x̂2 C/m2 in the region
0 < x < 5 m.

Having D = 0 in the region x < 0 requires that the field due to surface charge ρso
on x = 5 m plane must cancel the field due ρs = 2 C/m2 on x = 0 plane — this
requires that ρso be −2 C/m2 .

In that case D = 0 in the region x > 5 m, because D must increase by ρso = −2

C/m2 when we cross the charged surface at x = 5 m.

8
 D = 3ŷ for x < 0.
ρs = 2C/m2 ρs = −6C/m2
z
Example 3: In the region x < 0 measurements indicate a constant displacement field
D = 3ŷ C/m2. Also, x = 0 and x = 5 m planes contain surface charge densities
of ρs = 2 C/m2 and ρs = −6 C/m2 respectively. Determine D for x > 0 if D is
known to be uniform in the intervals 0 < x < 5 m and x > 5 m. y
Solution: First we note that E = D ǫo
= ŷ ǫ3o V/m is tangential to x = 0 and x = 5 m
surfaces. Since the tangential component of E cannot change at any boundary, x
we will have a uniform Ey = ǫ3o in all regions, −∞ < x < ∞, implying that x=5m
Dy = 3 C/m2 throughout (caused by charges at |y| → ∞).

Second, we note that normal component of D with respect to x = 0 and x = 5 m

surfaces, namely Dx , is zero in z < 0. Since the normal component of D must
increase by an amount ρs when we cross a charged surface, we must have Dx = 2
C/m2 in the region 0 < x < 5 m, and Dx = 2 + (−6) = −4 C/m2 in x > 5 m.

In summary, 
ŷ3,
 for x < 0,
D = x̂2 + ŷ3, C.
for 0 < x < 5 m m2

−x̂4 + ŷ3, for x > 5 m


9
7 Poisson’s and Laplace’s equations
Summarizing the properties of electrostatic fields we have learned so far, they
satisfy the laws of electrostatics shown in the margin and, in addition, Laws of
electrostatics:
E = −∇V as a consequence of ∇ × E = 0.
∇ · E = ρ/ǫo
• Using these relations, we can re-write Gauss’s law as
∇×E=0
ρ
∇ · E = −∇ · (∇V ) = ,
ǫo
from which it follows that
ρ
∇2V = − , (Poisson’s eqn)
ǫo
where
2 ∂ 2V ∂ 2V ∂ 2V
∇V ≡ + +
∂x2 ∂y 2 ∂z 2
is known as Laplacian of V . Poisson’s eqn:
ρ
– A special case of Poisson’s equation corresponding to having ∇2 V = −
ǫo
ρ(x, y, z) = 0

everywhere in the region of interest is Laplace’s eqn:

∇2V = 0. (Laplace’s eqn)
∇2 V = 0

1
Focusing our attention first on Laplace’s equation, we note that the equation z V (d) = −3 V

can be used in charge free-regions to determine the electrostatic potential

V (x, y, z) by matching it to specified potentials at boundaries as illustrated y
V (0) = 0
in the following examples: V (z) =?
z=0
x

Example 1: Consider a pair of parallel conducting metallic plates of infinite extents z

in x and y directions but separated in z direction by a finite distance of d = 2
m (as shown in the margin). The conducting plates have non-zero surface charge z=d=2m

densities (to be determined in Example 2), which are known to be responsible for
an electrostatic field E = ẑEz measured in between the plates. Each plate has V (z) = Az + B
some unique and constant electrostatic potential V since neither E(r) nor V (r)
can dependent the coordinates x or y given the geometry of the problem. V (z)

Using Laplace’s equation determine V (z) and E(z) between the plates if the potential
of the plate at z = 0 is 0 (the ground), while the potential of the plate at z = d
is −3 V.

Solution: Since the potential function V = V (z) between the plates is only dependent
on z, it follows that Laplace’s equation simplifies as

2 ∂ 2V ∂ 2V ∂ 2V ∂ 2V
∇V = + + = = 0.
∂x2 ∂y 2 ∂z 2 ∂z 2
This equation can be satisfied by
V (z) = Az + B
where A and B are constants to be determined. Now applying the given boundary
conditions, we first notice that (at the lower plate)
V (0) = (Az + B)|z=0 = B = 0.

2
 z V (d) = −3 V
Applying the second boundary condition (at the top plate) we find
3V y
V (2) = (Az + 0)|z=2 = 2A = −3 V ⇒ A = − . V (0) = 0
2m V (z) =?
z=0
x
The upshot is, potential function
3
V (z) = − z, for 0 < z < 2 m.
2 z
Finally, we determine the electric field between the plates as z=d=2m
−3 3
3 ∂ 3 3V V (z) = − z
2
E = −∇V = −∇(− z) = ẑ ( z) = ẑ . 3 3
2 ∂z 2 2m E = −∇(− z) = ẑ
2 2
0
V (z)

Example 2: In Example 1 what are the surface charge densities of the metallic plates
located at z = 0 and z = 2 m surfaces?

Solution: Since the electric field

3V
E = ẑ
2m
in between the plates, comparing this field with the field
ρs
E = ẑ
ǫo
of a pair of parallel surfaces carrying surface charge densities ρs and −ρs (at
z = 0 and z = 2 m), we find that
3
ρs = ǫo
2
on the surface at z = 0. The surface at z = 2 m has ρs = − 32 ǫo .

3
 Notice that our solution with equal and opposite charge densities on the parallel
surfaces implies that electrostatic fields are zero within the conducting plates
where the fields due to two charged surfaces are canceling out. This conclusion is
consistent with having constant electrostatic potentials within conducting regions
as will be discussed in the next lecture.

z
Example 3: A pair of copper blocks separated by a distance d = 3 m in x direction
Vo = 0
hold surface charge densities of ρs = ±2 C/m2 on surfaces facing one another as +
+
-
-
shown in the margin. The blocks are assigned constant potentials Vo = 0 and Vp + -
+ -
(see figure). What is the potential difference Vp ? + - Vp =?
− + +
E =0 + E -
+ -
Solution: Let D = x̂ǫo Ex denote the displacement vector in between the blocks, and + -
+
let D− = 0 denote the displacement vector within the block with a surface at +
-
- x
+
x = 0. Then the boundary condition equation used at x = 0 implies that 3 m
-

C 2
x̂ · (D+ − D− ) = ǫo Ex = 2 ⇒ Ex = .
m2 ǫo
In that case, potential difference between the blocks is
2 6
V = Ex d = 3= .
ǫo ǫo
Since the block on the left is at a higher potential (electric field vectors point
from high to low potential) assigned as Vo = 0, we must have
6
Vp = − .
ǫo

4
 z
Poisson’s equation −ρ1 < 0 ρ2 > 0
ρ
∇2 V = −
ǫo
E
is used in regions where the charge density ρ(r) is non-zero. The following - +

x
example illustrates a possible use of Poisson’s equation.
−W1 W2
E1x(x)
Example 4: An infinite charged slab of width W1, located over −W1 < x < 0, has a ρ1W1
negative volumetric charge density of −ρ1 C/m3 , ρ1 > 0. A second slab of width 2ǫo
W2 and positive charge density ρ2 is located over 0 < x < W2 as shown in the
−W1 x
margin. The electric field of this static charge configuration under the constraint
W1ρ1 = W2ρ2 was computed in an earlier section as
(
−x̂ ρ1 (x+Wǫo
1)
, for − W1 < x < 0 E2x(x)
E=
x̂ ρ2 (x−W
ǫo
2)
, for 0 < x < W2 ρ2W2
2ǫo

and is depicted in the margin. Determine the electrostatic potential in the re- W2 x
gion and the potential difference V21 ≡ V (W2) − V (−W1) satisfying Poisson’s
equation.

Solution: This is a one dimensional geometry where E and potential V depend only on Ex(x)
−W1 W2
coordinate x. Therefore, Poisson’s equation ∇2V = −ρ/ǫo takes the simplified
form x
d2 V ρ(x)
= − .
dx2 ǫo
ρ1W1
−
Integral of this equation over x yields in the left dV
dx = −Ex , which implies, given ǫo
the electric field result from above,
(
ρ1 (x+W1 )
dV ǫo , for − W1 < x < 0
= ρ2 (x−W2 )
dx − ǫo , for 0 < x < W2

5
 Integrating dV
dx
once more (i.e., finding suitable anti-derivatives with integration
constants), we find
ρ1 (x+W1 )2
(
2ǫo + V1 , for − W1 < x < 0
V (x) = 2
− ρ2 (x−W2ǫo
2)
+ V2 , for 0 < x < W2 z
−ρ1 < 0 ρ2 > 0
where the integration constants included on each line have been selected so that
V2 = V (W2), V1 = V (−W1). E
- +

Requiring a unique potential value at x = 0 (we can only associate a single potential x
energy level with each position in space) compatible with this expression for V (x),
we obtain −W1
ρ1 (0 + W1)2 ρ2 (0 − W2)2 Ex(x)
W2
+ V1 = − + V2 , −W1 W2
2ǫo 2ǫo
x
from which
ρ2 W22 + ρ1 W12 ρ2 W2 (W1 + W2 ) ρ1 W1(W1 + W2)
V21 = V2 − V1 = = = . ρ1W1
2ǫo 2ǫo 2ǫo −
ǫo

Note that the equation above can be solved for W1, W2, and W2 + W1 in terms of
V12 , ρ2 , and ρ1 , providing useful formulas for diode design (see ECE 440). We V (x)
V2
can also get useful specific formulae for V1 and V2 by imposing V (0) = 0, i.e.,
choosing x = 0 to be the reference point.
−W1
W2 x
V1

• The solution of Poisson’s equation

ρ
∇2 V = −
ǫo

6
with an arbitrary ρ existing over a finite region in space can be obtained z ρ(r′ )
′
as r−r
ρ(r′)
Z
V (r) = ′
d3 r′ r r′
4πǫo|r − r |
where d3r′ ≡ dx′ dy ′ dz ′ and the 3D integral on the right over the primed
coordinates is performed over the entire region where the charge density y
is non-zero. O
x
– Verification: The solution above can be verified by combining a
number of results we have seen earlier on:
1. In Lecture 5 we learned that the electric potential V (r) of a
point charge e at the origin is
e
V (r) = .
4πǫo|r|
Clearly, this singular result is a solution of Poisson’s equa-
tion above (and the stated boundary condition) for a charge
density input of
ρ(r) = eδ(r).
2. Using ECE 210-like terminology and notation, the above re-
sult can be represented as
1
δ(r) → Poisson’s Eqn →
4πǫo|r|

7
 identifying the output on the right as 3D “impulse response”
of the linear and shift-invariant (LSI) system represented
by Poisson’s equation.
3. Because of shift-invariance, we have
1
δ(r − r′) → Poisson’s Eqn → ,
4πǫo|r − r′|
meaning that a shifted impulse causes a shifted impulse re-
sponse.
The shifted impulse response is usually called “Green’s
function” G(r, r′) in EM theory.
4. Because of linearity, we are allowed to use superpositioning
arguments like
1
Z Z
ρ(r′ )δ(r−r′)d3r′ = ρ(r) → Poisson’s Eqn → ρ(r′ ) ′
d3r′ = V (r),
4πǫo|r − r |
which concludes our verification of the electrostatic1 potential
solution. Note how we made use of the sifting property of the
impulse (from ECE 210) in above calculation.

1
Also, in quasi-statics we use ρ(r′ , t) to obtain V (r, t) over regions small compared to λ = c/f , with f
the highest frequency in ρ(r′ , t).

8
 • As an application of the general solution of Poisson’s equation, namely
ρ ρ(r′ )
Z
2 3 ′
∇ V =− ⇒ V (r) = d r,
ǫo 4πǫo|r − r′ |
we next provide an outline of the proof of Helmholtz theorem (see
p states that any vector field F(x, y, z) that vanishes in
Lecture 4) which
the limit r = x2 + y 2 + z 2 → ∞ can be reconstructed uniquely from
its divergence and curl:

– First, with no loss of generality, we write

F = −∇V + ∇ × A

in terms of scalar and vector fields V (x, y, z) and A(x, y, z) to be

identified as follows2 :
– Taking first the divergence of F (and using ∇ · ∇ × A = 0), we
find that
∇′ · F(r′ ) 3 ′
Z
2
∇ · F = −∇ V ⇒ V (r) = dr
4π|r − r′ |
in analogy with Poisson’s equation (with ∇′·F(r′) replacing ρ(r′ )/ǫo
where ∇′ is “del” in (x′, y ′ , z ′ )-space).
2
This is possible because of the vector identity −∇2 G = ∇ × (∇ × G) − ∇(∇ · G) — call −∇2 G ≡ F,
which, according to this identity, is equal to the curl of a vector ∇ × G ≡ A (with ∇ · A = ∇ · ∇ × G = 0),
minus the gradient of a scalar ∇ · G ≡ V , as claimed. The challenge is in figuring out the underlying G
for a given F, which is what Helmholtz theorem is all about.

9
 – Likewise, the curl of F (with ∇ × ∇V = 0) leads us to, with a
divergence-free 3 A, to
∇′ × F(r′ ) 3 ′
Z
2 2
∇×F = ∇×∇×A = ∇(∇·A)−∇ A = −∇ A ⇒ A(r) = dr
4π|r − r′|
once again in analogy with Poisson’s equation4 .

These results validate Helmholtz theorem for fields F vanishing at infin-

ity, since, evidently, V and A needed to reconstruct F can be uniquely
specified in terms of ∇ · F and ∇ × F, respectively.

3
To confirm ∇ · A = 0 directly, use the identity ∇ · (αG) = α∇ · G + G · ∇α to expand A(r) as

∇′ × F(r′ ) 3 ′ ∇′ × F(r′ ) 1 ∇′ × F(r′ ) ′ 1 ∇′ · (∇′ × F(r′ )) 3 ′

Z Z Z Z
3 ′ 3 ′
∇· d r = ·∇ d r = − ·∇ d r = d r = 0,
4π|r − r′ | 4π |r − r′ | 4π |r − r′ | 4π|r − r′ |

after also using integration by parts for an integrand that vanishes as |r| → ∞ and a symmetry relation
∇|r − r′ |−1 = −∇′ |r − r′ |−1 which is easy to confirm.
4
While the vector field A identified above is divergence-free, ∇×A in the F = −∇V +∇×A expansion
can also be replaced with ∇×A′ so long as A′ = A+∇Ψ since ∇×∇Ψ is unconditionally zero independent
of the choice of Ψ. Note it is possible to specify Ψ so that ∇ · A′ = ∇ · ∇Ψ = ∇2 Ψ 6= 0 in which case A′
will be a divergent solution of the ∇ × F equation above! The additive term ∇Ψ in A′ is analogous to
allowing a constant number to be added to V ! The freedom to specify Ψ and thus ∇ · A′ at will is known
as gauge freedom and any choice of Ψ making ∇ · A′ = 0 is known as Coulomb’s gauge.

10
 −ρs
8 Conductors, dielectrics, and polarization (a)
- - - - - - - - - -

ρs
So far in this course we have examined static field configurations of charge Eo = ẑ
ǫo
distributions assumed to be fixed in free space in the absence of nearby
materials (solid, liquid, or gas) composed of neutral atoms and molecules. + + + + + + + + + + ρs
In the presence of material bodies composed of large number of charge-
neutral atoms (in fluid or solid states) static charge distributions giving rise (b)- - - - - - - - - - −ρs
to electrostatic fields can be typically1 found: ρs + + + + + + + + +
Eo
+

σ>0 E=0
1. On exterior surfaces of conductors in “steady-state”, - - - - - - - - - -
−ρs
Eo
2. In crystal lattices occupied by ionized atoms, as in depletion regions of + + + + + + + + + + ρs
semiconductor junctions in diodes and transistors.
A conducting slab inserted into a
region with field E_o (as shown
In this lecture we will examine these configurations and response of materials in b)develops surface charge which
cancels out E_o within the slab.
to applied electric fields. E_o relates to surface charge
as dictated by Gauss’s law and
superposition principle.

Conductivity and static charges on conductor surfaces:

• Conductivity σ is an emergent property of materials bodies con-
taining free charge carriers (e.g., unbound electrons, ionized atoms or
molecules) which relates the applied electric field E (V/m) to the elec-
trical current density J (A/m2 ) conducted in the material via a linear
1
More generally, materials containing charge carriers exhibiting divergence free flows will also exhibit
static charge distributions.

1
 relation2
J = σE. (Ohm’s Law)
(a)
- - - - - - - - - - −ρs
• Simple physics-based models for σ will be discussed later in Lecture 11. ρs
Eo = ẑ
For now it is sufficient to note that: ǫo

– σ → ∞ corresponds to a perfect electrical conductor 3 (PEC) for + + + + + + + + + + ρs

which it is necessary that E = 0 (in analogy with V = 0 across a
short circuit element) independent of J. (b)
- - - - - - - - - - −ρs
Eo
– σ → 0 corresponds to a perfect insulator for which it is necessary ρs + + + + + + + + + +
that J = 0 (in analogy with I = 0 through an open circuit element) σ>0 E=0
- - - - - - - - - -
independent of E. −ρs
Eo
+ + + + + + + + + + ρs
• While (macroscopic) E = 0 in PEC’s unconditionally, a conductor with
a finite σ (e.g., copper or sea water) will also have E = 0 in “steady- A conducting slab inserted into a
region with field E_o (as shown
in b)develops surface charge which
state” after the decay of transient currents J that may be initiated cancels out E_o within the slab.

within the conductor after applying an external electric field Eo (see E_o relates to surface charge
as dictated by Gauss’s law and
margin). superposition principle.

– The reason is, mobile free charges (e.g., electrons in metallic con-
ductors) within the conductor will be pulled or pushed by the
applied field Eo to pile up on exterior surfaces of the conductor
2
Linear behavior is possible provided charge carriers suffer occasional collisions within the medium.
3
PEC is an “idealization” that has no real counterpart, even though it is convenient to treat high
conductivity materials such as copper as PEC in certain approximate models and calculations. For “su-
perconducting materials” σ → ∞ only in the low frequency limit.

2
 until a surface charge density ρs that is generated produces a sec- (a)
- - - - - - - - - - −ρs

ondary field −Eo that exactly cancels out the applied Eo within ρs
Eo = ẑ
the interior of the conductor. ǫo

– E = 0 in the interior at steady-state implies that potential V =const., + + + + + + + + + + ρs

as well as ρ = ∇ · D = ∇ · ǫoE = 0.
– Surface charge density ρs and the exterior field on a conductor
(b)
- - - - - - - - - - −ρs
surface will satisfy the boundary condition equations Eo
ρs + + + + + + + + + +

n̂ · D = ρs and n̂ × E = 0, σ>0 E=0

- - - - - - - - - -
−ρs
Eo
with n̂ denoting the outward unit normal. + + + + + + + + + + ρs

• The transient “time-constant” τ for the decay of charge density ρ (and A conducting slab inserted into a
region with field E_o (as shown
hence E, as claimed above) in a homogeneous4 conductor (constant σ) in b)develops surface charge which
cancels out E_o within the slab.

can be obtained using the continuity equation E_o relates to surface charge
as dictated by Gauss’s law and
superposition principle.
∂ρ
+∇·J=0
∂t
representing the mathematical statement of charge conservation (de-
rived in Lecture 16). Using J = σE and ∇ · E = ρ/ǫo, we have
σ
∇ · J = σ∇ · E = ρ
ǫo
4
See Fisher and Varney, Am. J. Phys., 44, 464 (1976), for a discussion of contact potential between
different metals.

3
 above, from which it follows that
∂ρ σ − ǫσo t
+ ρ = 0 with a damped solution ρ(t) = ρ(0)e .
∂t ǫo
The decay time-constant
ǫo
τ=
σ
is typically very short (∼ 10−18 s) in metallic conductors, which is why
such conductors are usually considered to be in steady-state (and have
zero interior fields).

• As a consequence: in electrostatic5 problems conducting volumes

of materials (e.g., chunks of copper) can be treated as equipotentials
having zero internal fields and finite surface charge densities ρs = n̂ · D
expressed in terms of external fields D normal to the surface.

5
Also applicable quasi-statically when externally applied fields Eo (t) change slowly with time-constants
much longer than ǫo /σ. The way conductors are treated in high frequency electromagnetic problems will
be described later on.

4
Dielectric materials and polarization:
• Dielectric materials consist of a large number of charge-neutral atoms - - - - - - - - - -
or molecules and ideally contain no mobile charge carriers (i.e., σ = 0). Eo
Dielectric P
• Electric fields produced by charges located outside or within a dielectric slab E = Eo −
ǫo
material will polarize the dielectric — meaning that its constituent Eo
atoms or molecules will be “stretched out” to expose their internal or + + + + + + + + + +
A dielectric slab inserted into a
“bound” charges, electrons and protons — which will in turn cause the region with an initial field E_o
will become polarized.
electric field inside the dielectric to become weaker than (but not zero,
Inside the polarized dielectric the
as in conductors) what the field would have been in the absence of field will be weaker than E_o, but
not reduced to zero as in a
polarization effect. conductor.

We will next examine this polarization process and see how Gauss’s law can
be re-stated to facilitate field calculations in dielectric materials containing
bound charge carriers, i.e., atomic/molecular electrons and protons which
are not free to drift away from one another indefinitely (neglecting possible
ionization events).

• Consider a static free-charge density ρ(z) that would produce a macro-

scopic field Eo satisfying ρ = ǫo∇ · Eo in free space, producing, instead,
a field E = ẑEz inside a dielectric medium composed of an array of
neutral atoms or molecules.

Our objective is to relate the field E to Eo and ρ, and find a way

of calculating E when ρ is given.

5
 ρs + + + + + + + + + + ρs
−ẑ ≡ E1
• In the presence of an electric field E = ẑEz in the dielectric each neutral −ρs - -
d
- - - - - - - - ǫo
atom of the medium will be in a distorted (but not ripped apart) state ∆z 0 = E2
forming a ẑ oriented electric dipole, which can be visualized as a ρs + + + + + + + + + +
−ẑ
ρs
−ρs - ǫo
proton-electron pair with a small proton displacement d in z direction - - - - - - - - -

0
with respect to the electron. ρs + + + + + + + + + + ρs
−ẑ
−ρs - - - - - - - - - - ǫo
– Consider a regular array of such dipoles 0
ρs + + + + + + + + + + ρs
p ≡ edẑ, −ẑ
ǫo
−ρs - - - - - - - - - -

with ∆x, ∆y, and ∆z spacings between the dipoles (see margin), e
0
ρs =
so that the volumetric dipole density is ∆x∆y

1
Nd ≡ m−3,
∆x∆y∆z
within the array, and, furthermore,
e C
ρs =
∆x∆y m2
is the magnitude of charge density of the adjacent proton and
electron layers (see margin again) formed by arrays of adjacent
dipoles displaced in z by intervals ∆z.
– Assuming that the array is infinite in extent in x and y directions,
the proton and electron layers with surface charge densities ±ρs
will produce interior electric fields
ρs e/ǫo
E1 = −ẑ = −ẑ
ǫo ∆x∆y
6
 ρs + + + + + + + + + + ρs
(pointing in opposite direction to E = ẑEz ), and exterior fields d −ẑ
ǫo
≡ E1
−ρs - - - - - - - - - -

∆z 0 = E2
E2 = 0 ρs + + + + + + + + + + ρs
−ẑ
−ρs - - - - - - - - - - ǫo
in between the dipole layers. Space averaged macroscopic electric
0
field within the array (with a spatial weighting proportional to ρs + + + + + + + + + + ρs
−ẑ
the size of regions with the fields E1 and E2) produced by the −ρs - - - - - - - - - - ǫo

polarized dipoles will then be 0

ρs + + + + + + + + + + ρs
−ẑ
d ∆z − d ed/ǫo Nd edẑ P −ρs - - - - - - - - - - ǫo
Ep = E1 + E2 = −ẑ =− =− ,
∆z ∆z ∆x∆y∆z ǫo ǫo e
0
ρs =
∆x∆y
where
P ≡ Nd edẑ = Nd p
is, by definition, macroscopic polarization field of the dielectric,
measured in units of C/m2 (same units as a surface charge density).
- - - - - - - - - -
– The total macroscopic field E in the dielectric is then the sum of Eo
field Eo produced by the free charge density ρ in the region and P
Dielectric
the polarization field Ep = − ǫPo produced by bound charge carriers slab E = Eo −
ǫo
of the neutral atoms and/or molecules of the dielectric, i.e., Eo
+ + + + + + + + + +
P A dielectric slab inserted into a
E = Eo − , region with an initial field E_o
ǫo will become polarized.

Inside the polarized dielectric the

a result that shows a “reduced field strength” E (compared to Eo) field will be weaker than E_o, but
not reduced to zero as in a
inside the dielectric since P and Eo are colinear. conductor.

7
• Let’s re-arrange the expression for E from above as

ǫ o E + P = ǫ o Eo

after multiplying it with ǫo and moving P to the left. Now, the term on
the right is ǫoEo = Do respresenting the displacement vector outside the
dielectric slab, and if we were to “adopt” the left hand side expression
as the “displacement vector” for the interior, i.e, take

D = ǫo E + P - - - - - - - - - -
Eo
in regions with non-zero P, then we would see that Dielectric P
slab E = Eo −
ǫo
– D = Do, i.e., the displacement is the same inside and outside the
Eo
slab, while electric fields E and Eo inside and outside differ by a + + + + + + + + + +

non-zero −P/ǫo, and furthermore, A dielectric slab inserted into a

region with an initial field E_o
will become polarized.
– this generalized definition of electric displacement is consistent
Inside the polarized dielectric the
with (by now familiar) D = ǫoE for free space as P 6= 0 only field will be weaker than E_o, but
not reduced to zero as in a
within dielectrics. conductor.

• To express Gauss’s law ∇ · (ǫoE) = ρ in a form applicable with our

new revised D = ǫoE + P we first note that Gauss’s law also holds in
material media so long as ρ = ρf + ρb is a total charge density function,
a sum of charge densities ρf and ρb associated with free and bound
charge carriers that could present in the dielectric region. As such:

8
 – outside dielectrics, Gauss’s law is ∇ · (ǫoE) = ρf , since ρb = 0 in
that case, and this can be expressed as ∇ · D = ρf , with D = ǫoE
as usual;
– using ∇ · D = ρf also within dielectrics, with D = ǫoE + P, leads
to ∇ · (ǫoE) = ρf − ∇ · P, which is the expected form for Gauss’s
law within dielectrics with ρb = −∇ · P;
– in conclusion, we can write Gauss’s law in a general form applicable - - - - - - - - - -
everywere as Eo
∇ · D = ρf , Dielectric P
slab E = Eo −
ǫo
with the understanding that D ≡ ǫoE + P and “includes” the
Eo
effect of bound charge density ρb = −∇ · P which may be non- + + + + + + + + + +

zero6 within dielectrical materials. A dielectric slab inserted into a

region with an initial field E_o
will become polarized.
– Typcially subscript f of ρf is dropped in Gauss’s law, with the
Inside the polarized dielectric the
understanding that ρ refers to ρf because any non-zero ρb = −∇·P field will be weaker than E_o, but
not reduced to zero as in a
effects have already been “lumped” into D = ǫoE + P definition. conductor.

• The differential form of Gauss’s law that we will now write (without
the subscript f on ρf ) as

∇ · D = ρ, [Gauss’s law inside material medium]

appears in integral form (after applying Divergence theorem to volume

6
Note that if P is a constant inside a dielectric and zero outside then ρb = −∇ · P will be a surface
charge density confined to the surface of the dielectric.

9
 integral of the differential form) as
I Z
D · dS = ρdV,
S V

where the right side denotes the net free charge inside volume V .

• In a large class of dielectric materials macroscopic polarization P and

electric field E turn out to be linearly related (see Lecture 11) as

P = ǫoχeE,

where χe ≥ 0 is a dimensionless quantity called electric susceptibil-

ity. For such materials

D = ǫoE + P = ǫo(1 + χe )E = ǫE,

where
ǫ = ǫo(1 + χe) ≡ ǫr ǫo
is known as the permittivity of the dielectric, and

ǫ r = 1 + χe

its relative permittivity or dielectric constant.

– Dielectric constant of free space is 1,

◦ for air ǫr ≈ 1.0006,

10
 ◦ for glass 4 − 10,
◦ dry-to-wet earth 5 − 10, silicon 11 − 12, distilled water 81.
In certain materials χe and ǫ are found to be tensors — mean-
ing that P and D are no longer aligned with E. Such materials
are said to be anisotropic, but they will not be studied in this
course. Also, there is an exception to the condition χe ≥ 0 — in
collisionless plasmas χe < 0, as discussed in ECE 450.

• In Gauss’s law applicable in material media ρ denotes the free charge

carrier density (after the revisions we have agreed to make). Further-
more, in perfect dielectrics there are no mobile free carriers and Gauss’s
law typically reduces to ∇ · D = 0, while the corresponding boundary D+
n̂
condition equation for surfaces separating perfect dielectrics becomes D−
n̂ · (D+ − D− ) = 0 ⇒ Dn+ = Dn−,
which says that normal component of displacement D is continuous on
such surfaces. This is accompanied by

n̂ × (E+ − E−) = 0 ⇒ Et+ = Et−

D+
n̂
stating the continuity of tangential components of E, which is univer- D−
sally true as we have seen earlier.

11
9 Static fields in dielectric media
• Summarizing important results from last lecture:

– within a dielectric medium, displacement

D = ǫE = ǫoE + P,

and if the permittivity ǫ = ǫr ǫo is known, D and E can be calcu-

lated from free surface charge ρs or volume charge ρ in the region
without resorting to P.
– on surfaces separating perfect dielectrics, n̂ · (D+ − D−) = 0 typ- D+
ically, while n̂ · D+ = ρs on a conductor-dielectric interface (with n̂
D−
n̂ pointing from the conductor toward the dielectric).
– Gauss’s law ∇ · D = ρ (and its integral counterpart) includes only
the free charge density on its right side, which is typically zero in
many practical problems.
– once D and E have been calculated (typically using the boundary
condition equations), polarization P can be obtained as

P = D − ǫo E

if needed.

These rules will be used in the examples in this section.

1
 z

Example 1: A perfect dielectric slab having a finite thickness W in the x direction E = 18x̂ E = 3x̂ E = 18x̂
is surrounded by free space and has a constant electric field E = 18x̂ V/m in
its exterior. Induced polarization of bound charges inside dielectric reduces the
electric field strength inside the slab from 18x̂ V/m to E = 3x̂ V/m. What are ǫ = ǫo ǫ = ǫr ǫo ǫ = ǫo
the displacement field D and polarization P outside and inside the slab, and
what are the dielectric constant ǫr and electric susceptibility χe of the slab? x

Solution: Displacement field outside the slab, where ǫ = ǫo , must be

C
D = ǫo E = x̂18ǫo .
m2
The outside polarization P is of course zero. Boundary conditions at the interface
of the slab with free space require the continuity of normal component of D and
tangential component of E — both of these conditions would be satisfied if we
were to take D = x̂18ǫo C/m2 also within the dielectric slab. Thus, with E = 3x̂
V/m inside the slab, the condition D = ǫslabE within the slab requires that
ǫslab = 6ǫo .
Consequently, the dielectric constant of the slab is
ǫslab
ǫr = 1 + χe = =6
ǫo
and its electric susceptibility is
χe = ǫr − 1 = 5.
Finally, since D = ǫo E + P in general, polarization P inside the slab is
C
P = D − ǫo E = x̂18ǫo − ǫo 3x̂ = x̂15ǫo .
m2

2
• Our revised definition of displacement D = ǫE, where ǫ = ǫr ǫo, implies,
when combined with E = −∇V and ∇ · D = ρ, a revised form of
Poisson’s equation
ρ
∇2 V = − ,
ǫ
– provided that dielectric constant ǫr is independent of position so
that ∇ · D = ∇ · (ǫE) = ǫ∇ · E is a valid intermediate step in the
derivation of Poisson’s equation.
– Under the same condition Laplace’s equation ∇2V = 0 also re-
mains valid.
– Dielectrics where ǫr is independent of position are said to be ho-
mogeneous.
◦ In inhomogeneous dielectrics where ǫ varies with position
neither equation is valid, and one has to resort to the full
form of Gauss’s law in field and potential calculations.
In other words, don’t use Laplace’s/Poisson’s equations
in inhomogeneous media.
In the next example we have two homogeneous slabs side-by-side
making up an inhomogeneous configuration. In that case we can
use Laplace/Poisson within the slabs one at a time and then match
the results at the boundary using boundary condition equations
as shown.

3
 ρs = 2ǫo V (2) =?
z

Example 2: A pair of infinite conducting plates at z = 0 and z = 2 m carry equal

and opposite surface charge densities of −2ǫo C/m2 and 2ǫo C/m2 , respectively.
2ǫo
Determine V (2) if V (0) = 0 and regions 0 < z < 1 m and 1 < z < 2 m are
occupied by perfect dielectrics with permittivities of ǫo and 2ǫo , respectively. V (0) = 0
ǫo x
Solution: Given that V (0) = 0, we assume V (z) = Az, for some constant A in the
ρs = −2ǫo
homogeneous region 0 < z < 1 m, since V (z) = Az satisfies the Laplace’s
equation as well as the boundary condition at z = 0.
z
This gives V (1) = A at z = 1 m, which then implies that we can take V (z) =
A + B(z − 1) for the second homogeneous region 1 < z < 2 m having a different z=2
permittivity than the region below.
2ǫo V (z) = A + B(z − 1)
To determine the constants A and B, we will make use of boundary conditions at z=1

z = 0 and z = 1 m interfaces: ǫo V (z) = Az

V (z)
• In the region 0 < z < 1 m, the electric field E = −∇(Az) = −Aẑ, and,
therefore displacement D = ǫ1 E = −ǫo Aẑ. Hence, the pertinent boundary
condition ẑ · D(0) = ρs yields

ẑ · D(0) = −ǫo A = −2ǫo ⇒ A = 2.

• Just below z = 1 m the displacement is D(1−) = −ǫoAẑ = −2ǫoẑ as we

found out above. Above z = 1 m, the electric field is E = −∇(A + B(z −
1)) = −B ẑ, and, therefore, D(1+) = −2ǫo B ẑ just above z = 1 m. Hence,
the pertinent boundary condition ẑ · (D(1+) − D(1−) = 0 yields

ẑ · (−2ǫoB ẑ − (−2ǫo ẑ)) = −2ǫo B + 2ǫo = 0 ⇒ B = 1.

4
 Based on above calculations of constants A and B, the potential solution for the
region is (
2z V, 0<z<1
V (z) =
2 + (z − 1) V, 1 < z < 2.
It follows that V (2) = 3 V.

Note that electric fields −2ẑ V/m and −ẑ V/m in the bottom and top layers point
from high to low potential regions. Electric field E is discontinuous at the bound-
ary at z = 1 m while displacement D is continuous — the continuity of normally
directed D is demanded by boundary condition equations in the absence of sur-
face charge. z

E=0 V =0
d
ρsd = ?
d1 ρs
Example 3: A pair of infinite conducting plates at z = 0 and z = d are grounded
and have equal potentials, say, V = 0. The region 0 < z < d is occupied by E=?
free space (i.e., ǫ = ǫo ) except that an infinite charge sheet with a static surface ρs0 = ?
charge density ρs is located at z = d1 < d. Determine (a) the electrostatic field 0
E(z) in regions 0 < z < d1 and d1 < z < d, and (b) the surface charge densities E=0 V =0
ρs0 and ρsd at z = 0 and z = d on conductor surfaces if d1 = d/2.
If ρs in Example 3 is a slowly-
Solution: (a) Laplace’s equation for the given geometry requires a linear (in z) poten- varying function of time, then
tial solution in regions 0 < z < d1 and d1 < z < d. Since electrostatic E = −∇V , slowly varying E, ρs0 , and ρsd cal-
we can therefore represent the electric field in these regions as culated with instantaneous values
( of ρs would constitute quasi-static
solutions which are valid so long
−ẑVo /d1, 0 < z < d1
E= as d ≪ c/f , with f the highest
+ẑVo /d2, d1 < z < d frequency in ρs (t).

5
 where Vo ≡ V (d1) and d2 ≡ d − d1. Hence,
(
−ẑǫo Vo /d1 , 0 < z < d1
D = ǫo E = ,
+ẑǫo Vo /d2, d1 < z < d

and Maxwell’s boundary condition equation applied on z = d1 surface is

 
1 1
ẑ · (D(d+ −
1 ) − D(d1 )) = ρs ⇒ ǫo Vo + = ρs .
d2 d1
Thus  −1
ρs 1 1 ρs d1d2 ρs d1d2
Vo = + = = .
ǫo d2 d1 ǫo d1 + d2 ǫo d
Substituting Vo back into the expression for E, we have
(
−ẑ ρǫos dd2 , 0 < z < d1
E=
+ẑ ρǫos dd1 , d1 < z < d.

(b) The surface charge at z = 0 can be found by evaluating ẑ · D = ẑ · ǫo E at z = 0.

Hence,
d2 −−−−−→ ρs
ρs0 = ẑ · ǫo E(0) = − ρs d1 = d/2 − .
d 2
Likewise,
d1 −−−−−→ ρs
ρsd = −ẑ · ǫo E(d) = − ρs d1 = d/2 − .
d 2

6
 z
V (2) =?
ρs = −2ǫo
2ǫo

Example 4: Between a pair of infinite conducting plates at z = 0 and z = 2 m, the

V (0) = 0
medium is a perfect dielectric with an inhomogeneous permittivity of
ρs = 2ǫo x
4ǫo ǫo
ǫ(z) = .
4−z
Determine the electric potential V (2) on the top plate if V (0) = 0 and the
surface charge density is ρs = 2ǫo C/m2 on the bottom plate at z = 0. Note z
that Laplace’s equation cannot be used in this problem since the medium is 2ǫo
z=2
inhomogeneous. 4ǫo
ǫ(z) =
4−z
Solution: Consider Gauss’s law
∇ · (ǫE) = ρ
Ez (z)
with ρ = 0 in the region 0 < z < 2 m. Assuming that E = ẑEz (z), because the
geometry is invariant in x and y, we have ǫo
∂
∇ · (ǫE) = 0 ⇒ (ǫEz ) = 0 ⇒ ǫEz = constant.
∂z
Thus the product ǫEz is invariant with respect to coordinate z, which implies
that
ǫ(0) z
ǫ(z)Ez (z) = ǫ(0)Ez (0) ⇒ Ez (z) = Ez (0) = Ez (0)(1 − )
ǫ(z) 4
after substituting for ǫ(z). To identify Ez (0), we apply the bottom boundary
condition ẑ · D(0) = ρs , and obtain
2ǫo V
Dz (0) = ǫ(0)Ez (0) = 2ǫo ⇒ Ez (0) = =2m .
ǫ(0)

7
To determine V (2), we integrate E = ẑ2(1 − 4z ) V/m from top to bottom plate
(grounded), obtaining
Z 0 Z 0
z
V (2) = E · dl = 2(1 − )dz
z=2 z=2 4
2
z 4 3
= 2(z − )|02 = −2(2 − ) = −2 · = −3 V.
8 8 2

8
10 Capacitance and conductance
z A A
Parallel-plate capacitor: Consider a pair of conducting plates with surface
areas A separated by some distance d in free space (see margin).
The plates are initially charge neutral, but then some amount of electrons
Q y −Q
are transferred from one plate to the other so that the plates acquire equal
and opposite charges Q and −Q, distributed with surface densities of ± Q A on
plate surfaces facing one another (as shown in the margin). x
√ d
• That way, in steady state and for d ≪ A, a field configuration con-
fined mainly to the region between the plates is acquired, satisfying the
condition that static field inside a conductor
√ should be zero. A weak
“fringing field” can be ignored if d ≪ A and thus the geometry well
approximates the case with infinite plates.

– A constant displacement field

Q
D = x̂
A
satisfies the normal boundary condition at the left plate boundary
as well as Gauss’s law ∇ · D = 0 in the region between the plates.
The corresponding electrostatic field is
D Q
E= = x̂ ,
ǫo ǫo A
and the voltage drop from (positive charged) left plate to (negative

1
 charged) right plate is
Z (d,0,0) z A A
d
Q d
Z
V = E · dl = dx = Q.
(0,0,0) x=0 ǫ o A ǫ o A
Q y−Q
The last result can be expressed as a linear charge-voltage relation

Q = CV x

with d
A
C ≡ ǫo
d
representing the capacitance of the parallel conducting plate ar-
dV
rangement that we call parallel plate capacitor. I = C
dt
• By differentiating Q = CV we obtain the charging rate of the capacitor
as + V (t) -
dQ dV
I= =C
dt dt
which is only possible, for ideal capacitors, if the capacitor plates are
externally connected to a circuit supplying a current as shown on the
right where the direction of I = dQ
dt
is in the direction of voltage drop V
across the capacitor, from the positively charged plate to the negatively
charged plate as shown.

– In lossy capacitors when the medium between the plates is con-

ducting, the charging rate of the capacitor plate will be smaller as

2
 given by Q −Q
dQ dV +
+
-
=C = I − GV, + E
-
-
dt dt +
+
-
-
J = σE
where G stands for the conductance of the capacitor (derived later +
+
-
-
x
in this lecture) and I the external current flowing into the non- +
+
-
-
ideal capacitor. +
+
-
-
– Therefore, for non-ideal capacitors the external current d

dV V (t)
I=C + GV, + -
dt I = GV (t) + C
dV
dt C
meaning that part of I goes into changing stored charge Q = CV
of capacitor plates and the rest to conduct a GV amount of leakage
1
current of the capacitor plates, and the equivalent circuit of the R=
G
non-ideal capacitor there contains a “shunt resistance” R = G1
accompanying C as shown in the margin.

• Returning to the IV -relation

dV
I=C
dt
of the ideal capacitor, this IV -relation was √obtained from the QV -
relation above quasi-statically assuming that A ≪ λ = c/f , where f
is the highest frequency of V (t). The power absorbed by the capacitor
is then calculated as
dV d 1
P =VI =VC = ( CV 2),
dt dt 2
3
 implying a stored energy of
1 1
W = CV 2 = ǫo|Ex|2Ad
2 2
when the capacitor is in a charged state.

• Notice that stored energy is

1 1
ǫoEx2 = ǫoE · E
2 2
times the volume Ad occupied by the field E between the capacitor
plates. That suggests that
1
w = ǫo E · E
2
can be interpreted as stored electrostatic energy per unit volume in
general.

– Also both capacitance C and stored energies W and w would have

ǫ replacing ǫo in dielectric media.

A capacitor with a perfect dielectric between its plates will hold its charge
and stored energy indefinitely. However, if the dielectric is imperfect and
has a finite conductivity σ, charge will be transported from the positive to
negative plate by a volumetric current density

J = σE,

4
which will result in a quasi-static discharge of the capacitor and the loss of Q −Q
+ -
the stored energy W to Ohmic dissipation in the imperfect dielectric. +
+
-
E -
Just as capacitance C characterizes the energy and charge storage “ca- +
+
-
-
+ J = σE
pacity” of the capacitor, we can define a conductance G that relates the +
-
-
x
+ -
quasi-static discharge current I in between the plates of a capacitor to po- +
+
-
-
tential drop V : + -
d

• Discharge current I is the product of current density V (t)

+ -
Jx = σEx I = GV (t) + C
dV
C
dt

in A/m2 units and the plate area A. Since Ex = Vd , we obtain a linear

current-voltage relation 1
R=
I = GV G
System above behaves like a resis-
with conductance tor R = 1/G for
A
G≡σ ω≪
G
=
1
=
σ
d C RC ǫ
for the parallel plate capacitor. and like a capacitor C in the com-
plementary frequency band. To
obtain capacitor behavior at low
– Notice that G = σǫ C, a relation that will hold true for other types frequencies make sure that σ is
of capacitors that we will be examining. sufficiently small.

– Also, Alternatively, with large σ the

1 d system becomes a good electri-
R≡ = cal connector, a resistor R with
G Aσ a small resistance R ∝ 1/σ.
is the corresponding resistance that scales inversely with con-
ductivity σ of the material — large σ materials will have small

5
 resistance, but for a given σ, R increases with length d and de-
r
creases with increasing cross-sectional area A. Simple conductivity b a
models and J will be discussed next lecture.
z
Coaxial Cable: When we study guided wave propagation later in the course ℓ
we will learn about coaxial cables.

• A coax cable consists of two conducting regions — a central cylindrical

conductor with a cross-sectional radius a, enclosed by a conducting
pipe of a radius b > a (see margin), with some dielectric ǫ filling in
the space. We will next calculate the capacitance and conductance of
a coax segment of some length ℓ.

• For ℓ ≫ b, field E can be assumed to point out radially away from

the inner conductor with radius a to the outer conductor with radius b.
In that case Gauss’s law in integral form can be utilized to determine
the radial field Er . Considering a cylindrical integration surface with a
radius r > a centered about the inner conductor, we re-write Gauss’s
law I
ǫ E · dS = QV
s
as
ǫEr 2πrℓ = Q
where Q is the total charge distributed over the inner conductor and ǫ
the permittivity of the dielectric separating the two conductors.

6
 – It follows that
Q
Er = ,
2πǫℓr
and voltage drop from inner to outer conductor is
Z b Z b Z b
Q Q dr Q b
V = Er dr = dr = = ln .
r=a r=a 2πǫℓr 2πℓǫ r=a r 2πℓǫ a
Clearly, once again Q = CV , with
2π
C = b ℓǫ
ln a
representing the capacitance of the coax segment of length l.

• The capacitance of the coax per unit length is

2π
C = b ǫ.
ln a
– Conductance of the coax per unit length can likewise be shown
to be
2π
G = b σ.
ln a
This result is a consequence of the general relation G = σǫ C men-
tioned earlier.
– Per length parameters C and G of the coax will play an important
role when we study guided wave propagation in coaxial transmis-
sion lines with lengths for which quasi-static approximation may
be violated.

7
 z
Diode junctions: In Example 4 in Lecture 7 we derived the expression
−ρ1 < 0 ρ2 > 0
for potential drop V across a charged region of a total width of W1 + W2,
such that in region 1 where −W1 < x < 0 the charge density ρ = −ρ1 is
negative, while in region 2 where 0 < x < W2 the charge density ρ = ρ2 -
E +
is positive, with the additional constraint that the entire region is charge
x
neutral, meaning that ρ1W1 = ρ2W2.
By solving Poisson’s equation for this charge density configuration (see
margin) encountered in junction regions of semiconductor diodes (described −W1
Ex(x)
W2
in detail ECE 440) we had established that the voltage drop from x = W2 −W1 W2
to x = −W1 across the junction is given by x

ρ2W2(W1 + W2) ρ1 W1(W1 + W2)

V = = .
2ǫo 2ǫo ρ1W1
−
ǫo
The above equation implies that
r
2ǫoV 2ǫoV ρ1 + ρ2
W1 = and W2 = ⇒ W1+W2 = 2ǫoV . V (x)
(W1 + W2)ρ1 (W1 + W2)ρ2 ρ1ρ2 V2

Using the expressions above for junction voltage V and width W1 + W2,
−W1
we will next derive an expression for small signal capacitance of the diode
junction:
W2 x
V1
• In region 2 where x > 0, the junction holds a total positive charge of
Q = ρ2 W2A per cross-sectional area A.
Q
• Therefore, substituting A
for ρ2W2 in the expression for V above, and

8
 √ z
also using the W1 + W2 ∝ V expression derived above, we obtain
−ρ1 < 0 ρ2 > 0
q
ρ1 +ρ2
ρ2W2(W1 + W2) Q 2ǫoV ρ1 ρ2
V = = , E
2ǫo 2ǫoA - +

which can be re-arranged as x

2ǫoρ1 ρ2 √
r
Q=A V W2
ρ1 + ρ2 −W1
Ex(x)
−W1 W2
representing a non-linear charge-voltage relation (for a given charge x
profiles satisfying ρ1W1 = ρ2 W2).

– In a linear charge-voltage relation Q = CV , the capacitance pa- ρ1W1

rameter C represents the slope Q V
of a Q vs V curve. −
ǫo

dQ
The slope of any Q vs V curve is given by the derivative dV , whether or
dQ
not the curve is linear. The slope dV of a non-linear charge-voltage curve V (x)
V2
can be interpreted as a small signal capacitance C. For a diode junction,
differentiating the above equation, we find that −W1

dQ
r
ǫo ρ1ρ2 W2 x
C= =A . V1
dV 2V (ρ1 + ρ2 )
Small changes dV in junction voltage will accompany small changes dQ =
CdV in stored charge Q of the junction, but the amount CdV will itself
depend on V because C ∝ V −1/2.

9
11 Lorentz-Drude models for conductivity and (a)

susceptibility and polarization current E=0

In this lecture we will describe simple microscopic models for conductivity σ q>0
and electric susceptibility χe of material media composed of free and bound In the absence of an applied electric
field E, free charge q exhibits a
charge carriers. The models were first developed by Lorentz and Drude prior "random walk" between collisions such
that its average velocity v is zero.

to the establishment of quantum mechanics. In these models free charge By collisions we refer to the inreactions
of q with zero-mean miscroscopic electric
fields within the conductor due to charges
carriers motions are described using Newtonian dynamics and atoms are entrapped in the lattice.

represented as electric dipoles p = −er (r is electron displacement from

(b)
atomic nucleus) behaving like damped 2nd order systems.
Conductivity: Conducting materials such as copper, sea water, ionized
gases (plasmas) contain a finite density N of mobile and free charge carriers E
at the microscopic level (in addition to neutral atoms and molecules sharing q>0
the same macroscopic space) — these elementary mobile carriers can be In the presence of an applied electric
field E, the mean position of free charge
electrons, positive or negative ions, or positive “holes” (in semi-conductor q>0 drifts in the direction of field
vector E with some non-zero mean
velocity v.
materials). Avg. drift velocity v reperesents a
balance between acceleration force due
to E and an opposing friction force
• Each elementary charge carrier with a charge q and mass m and subject produced by collisions of q with the
lattice at random intervals with
some mean value τ .
to a macroscopic electrical force qE will be modelled by a dynamic
equation
dv v
m = qE − m ,
dt τ
which is effectively Newton’s second law — “force equals mass times
acceleration” — in which v denotes the macroscopic velocity1 of charge
1
Think of microscopic velocity of each charge carrier as v + δv, where δv is an independent zero-mean

1
 carriers and −m vτ denotes a macroscopic friction force proportional to 1 m

−v. Friction is a consequence of “collisions” of charge carriers with the

neutral background at a frequency of ν = τ1 collisions per unit time, v
q
and causes the decay of v as
1 m
1m
−t/τ
v(t) = v(0)e
A cube of unity volume contains
in the absence of field E. Therefore, when E = 0 the carriers settle N charge carriers q each moving
with an average velocity of
v=1 m/s towards the shaded surface
down to a steady state with v = 0 (in t ≫ τ limit), meaning that of the cube of a unity surface area.

no macroscopic current density J will be found in the absence of E in In 1 second all N charges in the
volume cross the shaded surface,
regions with homogeneous charge carrier densities. transporting Nq Coulombs of charge
per second per unit area.

By contrast, charge transport rate

• With a constant but non-zero E, steady-state solution of the above across the same surface is

equation is N qv C/s/m2 = N qv A/m2

qτ qτ if the average charge velocity
v = E, where | | is known as mobility. is an arbitrary v.
m m Thus, in vector notation, we can
define the current density in the
region as
– Assuming N charge carriers per unit volume each moving (on the
average) with this steady-state velocity in a given material, we can J = N qv
calculate the average flux density of charge through the region as such that current I (in A) across
any surface S having area elements
dS is given by
N q 2 C/s
J = N qv = E Z

mν m2 I= S J · dS

which is commonly referred to as current density (see margin).

If a given material contains several species of carriers with charge,
random variable for each charge carrier whereas macroscopic velocity v corresponds to the statistical
average of all v + δv.

2
 mass, collision frequency, and number density of qs, ms, νs, and
Ns, respectively, then current density can be expressed as Ohm’s Law
and
J = σE, DC
with conductivity
X Ns qs2
σ= σs and σs =
s
msνs
denoting the medium and species conductivities, respectively, un-
der DC conditions.

• With a time varying field E the corresponding current density will also
be time varying, in which case conductivity σ should be defined in the
frequency domain using phasor techniques (remember ECE 210).

– Briefly, using phasors Ẽ and J̃ such that

E(t) = Re{Ẽejωt} and J(t) = Re{J̃ejωt}, etc.,

we have a phasor transformed Newton’s force balance equation

dv v ṽ
m = qE − m ⇒ mjω ṽ = q Ẽ − m ,
dt τ τ
from which it follows that

J̃ = σ Ẽ,

3
 with
X Nsqs2
σ= σs and σs = .
s
ms(νs + jω)
AC
– Note that the AC conductivity just derived can be approximated conductivity
by the DC conductivity derived earlier for all AC frequencies ω
much smaller than species collision frequencies νs .
◦ In many cases of practical interest, this condition can be easily
met, and we are often well justified to ignore the frequency
dependence and complex character of conductivity σ revealed
in above derivation.
– More advanced quantum mechanical derivations of σs give the
same results except with effective masses specified by quantum
theory replacing the particle masses ms used in classical models.

• Typical DC conductivities:

– For silver, copper, gold, σ ∼ several × 107 S/m

– For seawater σ ≈ 4 S/m
– For intrinsic silicon σ = 1.6 × 10−3 S/m
– For dry earth σ ∼ 10−5 S/m
– For glass σ ∼ 10−10 − 10−14 S/m

Superconductivity occurs in certain materials at low temperatures

4
 when the DC conductivity vanishes as a consequence of correlated
charge carrier motions which are ignored in the Lorentz-Drude
model.

Susceptibility: In perfect dielectrics there are no free charge carriers and so + E -

σ = 0. However, in general such materials are polarizable and therefore they +
+
-
-
have a non-zero susceptibility χe and a dielectric constant ǫr = 1 + χe > 1. +
+ −e r e -
-
+ -
• In Lorentz-Drude model, each polarized atom or molecule is considered +
r e -
to be a dipole p = −er, with r representing the displacement vector + −e -
+ −e r e -
of an atomic electron from atomic nucleus when the atom is polarized +
+
-
because of an applied electric field. -

• If the polarizing force on the atom is removed, observations indicate

that the dipole field of the atom Ep ∝ p ∝ r will decay as a damped
co-sinusoid with a pdecay time constant τd = α1 and a characteristic
damped frequency ωo2 − α2 ≈ ωo satisfying a condition ωo ≫ α = τ1
d
(strongly underdamped).

– Possible values of ωo for a given atom can be obtained from the en-
ergy levels of bound states of the atom (calculated using standard
quantum2 models like in PHYS 214) and time constants τd = α1
(which are finite because energies ~ωo radiated away are also finite)
are related to observed line widths (2α) in the emission spectra of
excited atoms.
2
For a quantum mechanical derivation of susceptibility, see, e.g., Mott, “Elements of Wave Mechanics”,
Chapt 4, Sect 10 (1958); Miller, “Quantum Mechanics for Scientists and Engineers”, Sect 7.3 (2008).

5
 Electron displacement having the inferred damped co-sinusoid form
−t/τd
cos( ωo2 − α2 t) ≈ roe−t/τd cos(ωo t)
p
r(t) = roe
+ E -
+
is “zero-input response” (remember ECE 210) of a linear second-order +
-
-
ODE that can be constructed using Newton’s second law of classical +
+ −e r e -
-
+
mechanics: +
-
+ −e r e -
-
d2 r + −e r e
– If we assume that mass m times acceleration dt2
of a displaced +
-
-
+
electron equals the sum of a -

◦ force −eE exerted by an applied macroscopic electric field E,

◦ a spring-like restoring force −mωo2r responsible for the binding
of the electron to the nucleus, and
◦ a friction-like dissipative force −m2α dr
dt ,

we get
d2 r dr
m 2 = −eE − mωo2r − m2α ,
dt dt
for which r(t) given above is the zero-input solution in the absence
of E.

• To find the DC susceptibility of a dielectric composed of dipoles con-

strained by the above equation, we note that steady-state solution of
the equation with a non-zero constant field E is
e
r=− E.
mωo2

6
 Consequently, dipole moment of a single polarized atom is
e2
p = −er = E,
mωo2 + E -
+ -
+
and polarization field in a dielectric with a dipole density of Nd is + r e
-
-
+ −e -
+
Nd e2 +
-
P = Nd p = E. + −e r e -
mωo2 + −e r e -
-
+ -
+
This result can also be written as -

P = ǫoχeE,

where DC
Nd e2/mǫo susceptibility
χe ≡ 2
ωo
is DC susceptibility. AC susceptibility can be derived using phasor
techniques, but at frequencies ω ≪ ωo , AC susceptibility is well ap-
proximated by the DC susceptibility derived above.

Polarization current: Consider the case of a time varying electric field

E(t) in a dielectric medium at a frequency ω ≪ ωo such that the relations
e
r=− E and P = ǫoχeE
mωo2
from above are accurate.

7
• Time variation of E will imply the time variation of electron displace-
ment r, so that there will be in effect a non-zero electron velocity
dr e dE
v= =−
dt mωo2 dt
capable of producing a current.

– With Nd such electrons per unit volume, each carrying a charge

−e, we will have a net flux density of charge in the region given
by
Nd e2 dE dP
Jp = −eNd v = = .
mωo2 dt dt
This flux is effectively an AC current density carried by bound
charges found in a dielectric medium. Even though, a DC cur-
rent is not possible in a perfect dielectric containing only bound
charges, evidently AC currents Polarization
dP current
Jp = Polarization current density density
dt
are possible — we call this type of AC current as polarization
current density.

• In our studies of time-varying electromagnetic fields we will include the

effects of polarization currents dP
dt
along with the effects of conduction
currents σE.

8
12 Magnetic force and fields and Ampere’s law
Pairs of wires carrying currents I running in the same (opposite) direction
are known to attract (repel) one another. In this lecture we will explain the
mechanism — the phenomenon is a relativistic1 consequence of electrostatic
charge interactions, but it is more commonly described in terms of magnetic
fields. This will be our introduction to magnetic field effects in this course. I

1
F
Brief summary of special relativity: Observations indicate that light (EM) waves can be
“counted” like particles and yet travel at one and the same speed c = 3 × 108 m/s in all reference
frames in relative motion. As first recognized by Albert Einstein, these facts preclude the possibility that
F
a particle velocity u could appear as I
u′ = u − v (Newtonian)

to an observer approaching the particle with a velocity v; instead, u must transform to the observer’s
frame as
u−v
u′ = , (relativistic) “Things should be made as simple
1 − uv
c2 as possible – but no simpler.”
so that if u = c, then u′ = c also. This “relativistic” velocity transformation in turn requires that positions — Albert Einstein
x and times t of physical events transform (between the frames) as
v
x′ = γ(x − vt) and t′ = γ(t − x), (relativistic)
c2
where γ ≡ √ 1
, rather than as
1−v2 /c2

x′ = x − vt and t′ = t, (Newtonian)
′
so that dxdt
= u and dx
dt′
= u′ are related by the relativistic formula for u′ given above.
Relativistic transformations imply a number of “counter-intuitive” effects ordinarily not noticed unless
|v| is very close to c. One of them is Lorentz contraction, implied by dx = dx′ /γ at a fixed t: since γ > 1,
dx < dx′ , and moving objects having velocities v appear shorter then they are when viewed from other
reference frames where v is determined. A second one is time dilation, implied by dt′ = dt/γ at a fixed x′ :
since γ > 1, dt′ < dt, and moving clocks having velocities v and fixed x′ run slower than clocks in other
reference frames where v is determined. Consider taking PHYS 325 to learn more about special relativity.

1
 • Consider a current carrying stationary wire in the lab frame:

– the wire has a stationary lattice of positive ions,

– electrons are moving to the left through the lattice with an average
speed v, and (a) Neutral wire carrying current I
in the "lab frame":
– a current I > 0 is flowing to the right as shown in the figure.
◦ If the wire is electrically uncharged — which will be true if λ+ = −λ− I
+ _ + _ + _ + _ + _ + _ + _ + _
electron and ion charge densities in the wire, λ− < 0 and
λ− v
λ+ > 0, respectively, have equal magnitudes — then the wire
will produce no electrostatic field E, and any stationary charge
q placed near the wire will not be subject to any force2. (b) In the "electron frame" the wire
appears positively charged:
◦ The current carried by the wire is I = v|λ−| = vλ+ in terms
of the magnitudes of electron velocity and charge density.
v I λ′+ = γλ+
• An uncharged wire in the lab frame appears as “charged” in the refer- +
_
+
_
+ + + +
_ _
+
_
+ + +
_ _
ence frame of the electrons carrying the current: r λ′− = λ−/γ
λ′ ′
E = r̂
– this is a relativistic effect due to “Lorentz contraction” of the dis- 2πǫor
tances between the charges in the wire.
2
This is true for zero-resistivity wires. Current carrying wires with finite resistivity will however support ′ v 2 Iv
λ ≈ λ+ 2 = 2 = Ivµoǫo
surface charge densities with axial gradients to produce the static field within the wire needed to drive c c
the current — e.g., in Am. J. Phys.: Jefimenko, 30, 19 (1962); Parker, 38, 720 (1970); Preyer, 68, 1002
(2000).

2
 – In the electron frame the wire is found to have a positive charge
density λ′, and thus it has a radial electrostatic field
λ′ ′
E = r̂
2πǫor
implying an electrostatic force F′ = qE′ on a stationary charge q.

– Relativistic calculations3 show that

′ v2 I v2 (a) Neutral wire carrying current I
λ ≈ λ+ 2 = ( ) 2 = Ivǫoµo in the "lab frame":
c v c
3
(i) Electron spacings dx′ measured in the electron reference frame will appear as λ+ = −λ− I
r + _ + _ + _ + _ + _ + _ + _ + _
v2
dx = 1 − 2 dx′ v
c λ−
in the lab frame because of Lorentz contraction. Charge density of the electrons in the lab frame,
λ′−
λ− = p , (b) In the "electron frame" the wire
1 − v 2 /c2 appears positively charged:
is therefore greater in magnitude than the electron charge density λ′− in the electron frame. Furthermore,
λ− = −λ+ in order to maintain a charge neutral wire in the lab frame. (ii) Once again because of Lorentz
contraction, the charge density of positive ions will appear in the electron frame as
v I λ′+ = γλ+
λ+
λ′+ = p . +
_
+
_
+ + + +
_ _
+
_
+ + +
_ _
1 − v 2 /c2
r λ′− = λ−/γ
(iii) Thus, the total charge density of the wire in the electron frame is
λ′
E′ = r̂
λ+ p λ+ p λ+ v 2 /c2 λ+ v 2 2πǫor
λ′ = λ′+ + λ′− = p + λ− 1 − v 2 /c2 = p − λ+ 1 − v 2 /c2 = p ≈ 2 ,
1 − v 2 /c2 1 − v 2 /c2 1 − v 2 /c2 c

a positive density for non-zero |v| ≪ c. (e.g., articles in Am. J. Phys.: Webster, 29, 262, 1961; Matzek
and Russel, 36, 905, 1968; Arista and Lopez, 43, 525, 1975; Zapolsky, 56, 1137, 1988). ′ v 2 Iv
λ ≈ λ+ 2 = 2 = Ivµoǫo
c c
3
 and force F′ = qE′ can be transformed back to the lab frame, (a) In the "electron frame" the wire
appears positively charged and
repelsa test charge q with
where q appears to be moving with velocity v, as (with no ap- force F’=qE’

proximation4 ) v I λ′+ = γλ+

µo I + + + + + + + + + +
F = qv × φ̂, _ _ _ _ _ _ _
2πr r λ′− = λ−/γ
where φ̂ is the unit vector in the direction given by the right-hand- q
′ λ′ ′ µoI
F = qE = q r̂ ≈ qv r̂
rule (see margin) and µo = 4π × 10−7 H/m is permeability of free 2πǫor 2πr
space. v 2 Iv
′
λ ≈ λ+ 2 = 2 = Ivµoǫo
c c
• We find it convenient to define (b) In the lab frame force F~F’ of
moving charge q is attributed to
µo I magnetic field B produced by
current I and velocity v of the
B≡ φ̂ charge in F=qvXB combination.
2πr
to be the “magnetic flux density” of current filament I at a distance r,
and attribute the force I
F = qv × B q µoI
v B= φ̂
on the moving charge q to the magnetic field B produced by current I 2πr
F = qv × B
(rather than to the electrostatic field of the wire seen by q in its own Magnetic field B curls around
current I in a right handed
reference frame). direction designated by azimuthal
unit vectorφ̂

While we assumed q to be stationary in the reference frame of the electrons Magnetic field lines close upon
themselves unlike electric field
in the above discussion (for the sake of simplicity), the results obtained above lines which start and stop on
point charges.
are found to be valid for all particle velocities v measured in the lab frame. Right hand rule: point your
right thumb in the direction
4
We also get the same result using the approximation F = F′ that can be justified when |v| ≪ c, which of current flow; your fingers
is typically true by a large margin for electron speeds in current carrying conducting metals — see HW. will point in direction φ̂.

4
Also, if there are multiple current filaments In , each generating its own field
Bn, force F on q can be calculated using a superposition method as with
electrostatic fields.
Magnetic field B of the infinite current filament I obtained above can
also be obtained by superposing the magnetic field increments
µoIdl × r̂
dB ≡ (Biot-Savart law)
4πr2
of directed current increments Idl, where r = rr̂ is a position vector extend- z
ing from the location of the current increment to the field position where dB
is being specified — this formula, known as Biot-Savart law, is only valid
when used in terms of infinitesimal segments Idl of time-invarying current I
loops.
µoI
r B= φ̂
• Magnetic field B of the infinite line current I “derived” above satisfies 2πr

a circulation relation I
B · dl = µoIC ,
C
with IC = I.
This integral for the circulation of static magnetic field B is found to be
valid (experimentally) for all closed circulation paths C, and is known
as Ampere’s law (for static magnetic fields). In Ampere’s law

– IC stands for the net sum of all filament currents In crossing any
surface S bounded by path C,
◦ flowing in the direction given by the “right-hand-rule”:

5
 when the right thumb is pointed in the direction of dl along path
C, the direction of filament current In is specified as the direction
of the fingers of your right hand through surface S bounded by
contour C.
◦ Filament currents not crossing S — i.e., current filaments not
“linked” to path C — should not be included on the right hand
side of Ampere’s law.

• Ampere’s law can also be expressed as

I Z
H · dl = J · dS, z
C S

where
I

– we have defined µoI

H ≡ µ−1
o B
r B=
2πr
φ̂

for the sake of convenience, and

– J is the volumetric current density measured in A/m2 units (e.g.,
σE in a conducting region as discussed in last lecture) having a
total flux Z
IC = J · dS
S
across any surface S bounded by a path C,
◦ with dS pointing across S in the direction compatible with
right-hand-rule as in Stoke’s theorem (recall Lecture 6).

6
• Stoke’s theorem re-stated for a vector field H as
I Z
H · dl = ∇ × H · dS
C S

implies that the differential form of Ampere’s law should be Laws of

magnetostatics:
∇ × H = J.

This differential relation is accompanied by ∇×H = J

∇ · B = 0, ∇·B = 0
They also apply “quasi-statically”
satisfied by static magnetic field of the line current as well as by any over a region of dimension L
when a time-varying field source
other magnetic field — static as well as non-static, as determined ex- J(r, t) has a time-constant τ
perimentally and described in more detail later on. much longer than the propagation
time delay L/c of field variations
across the region (c is the speed
• Current density vector field J invoked above in Ampere’s law expres- of light).
sions, measured nominally in units of A/m2 , can also be adjusted to
In magneto-quasistatics (MQS)
describe the distributions of surface or line currents in 3D space. B(r,t) = µo H(r, t) will be ac-
companied by a slowly varying
electric field E(r, t) (derived from
– For example, Faraday’s law discussed in Lec-
J(x, y, z) = Js(y, z)δ(x − xo) ture 14).

can be regarded as volumetric current density representation

of a surface current density Js (x, y) measured in A/m units
flowing on x = xo surface.

7
 – Likewise,
J = ẑ y rect(y − 0.5) δ(x)
J(x, y, z) = ẑI(z)δ(x − xo)δ(y − yo )
representats a line current I(z) measured in A units flowing in
z-direction along a filament defined by the intersections of x = xo
z y
and y = yo surfaces.
– As a most extreme case,
J(x, y, z, t) = Qvδ(x − xo)δ(y − yo )δ(z − zo)
represents the time-varying current density of a point charge Q xo x
at coordinates (x, y, z) = (xo(t), yo(t), zo(t)) moving with velocity
v = (ẋo(t), ẏo(t), żo(t)).

Example 1: Consider a surface current density of

Js = ẑy rect(y − 0.5) A/m

flowing on x = 0 plane (as shown in the margin). What is the total current I
flowing on the same plane measured in A units?

Solution: To go from a surface current density Js in A/m to a total current I in A,

we need to perform an appropriate integration operation on the surface were Js
is defined. For the specified Js in this problem we find that
Z ∞ Z 1
y2 1 1
I= Js · ẑdy = y dy = |0 = A.
y=−∞ y=0 2 2

8
13 Current sheet, solenoid, vector potential and
current loops Js = ẑJs

In the following examples we will calculate the magnetic fields B = µoH

established by some simple current configurations by using the integral form y

of static Ampere’s law.

B L
C B = ŷB(x)
W

x
Example 1: Consider a uniform surface current density Js = Js ẑ A/m flowing on µoJs
B(x) =
x = 0 plane (see figure in the margin) — the current sheet extends infinitely in 2
y and z directions. Determine B and H.

Solution: Since the current sheet extends infinitely in y and z directions we expect B As shown in Example 1 mag-
netic field of a current sheet
to depend only on coordinate x. Also, the field should be the superposition of the is independent of distance
fields of an infinite number of current filaments, which suggests, by right-hand- |x| from the current sheet.
rule, B = ŷB(x), where B(x) is an odd function of x. To determine B(x), such Also H changes discontinu-
that B(−x) = −B(x), we apply Ampere’s law by computing the circulation of ously across the current sheet
by an amount Js .
B around the rectangular path C shown in the figure in the margin. We expand
I
B · dl = µo IC
C
as
B(x)L + 0 − B(−x)L + 0 = µo Js L,
from which we obtain
µo Js µo Js Js
B(x) = ⇒ B = ŷ sgn(x) and H = ŷ sgn(x).
2 2 2

1
 Example 2: Consider a slab of thickness W over − W2 < x < W2 which extends in-
finitely in y and z directions and conducts a uniform current density of J = ẑJo
A/m2 . Determine H if the current density is zero outside the slab.

Solution: Given the geometric similarities between this problem and Example 1, we
postulate that B = ŷB(x), where B(x) is an odd function of x, that is B(−x) =
−B(x). To determine B(x) we apply Ampere’s law by computing the circulation
of B around the rectangular path C shown in the figure in the margin. For
x < W2 , we expand I
B · dl = µo IC
C
as
z
B(x)L + 0 − B(−x)L + 0 = µo Jo 2xL ⇒ B(x) = µo Jo x.
y
W
For x ≥ 2 , the expansion gives
Joẑ
W
B(x)L + 0 − B(−x)L + 0 = µo Jo W L ⇒ B(x) = µo Jo . B L B = ŷB(x)
2 C
W x
Hence, we find that
( W W
ŷJo x, |x| < W2 −
2 2
H= By (x)
ŷJo W2 sgn(x), otherwise. µoJoW
2
Note that the solution plotted in the margin shows no discontinuity at x = ± W2
or elsewhere. −
W W x
2 2

The figure in the margin depicts a finite section of an infinite solenoid.

A solenoid can be constructed in practice by winding a long wire into a

2
 B = ẑB
multi loop coil as depicted. A solenoid with its loop carrying a current I
in φ̂ direction (as shown), produces effectively a surface current density of
Js = IN φ̂ A/m, where N is the number density (1/m) of current loops in
C
the solenoid. In Example 3 we compute the magnetic field of the infinite
solenoid using Ampere’s law.
L I

Example 3: An infinite solenoid having N loops per unit length is stacked in z-

direction, each loop carrying a current of I A in counter-clockwise direction when
viewed from the top (see margin). Determine H. Infinite solenoid
with N loops per
unit length carrying
I amps per loop
Solution: Assuming that B = 0 outside the solenoid, and also B is independent of
z within the solenoid, we find that Ampere’s law indicates for the circulation C B = µoIN
shown in the margin
I
B · dl = µo IC ⇒ LB = µo IN L.
C
This leads to
B = µo IN and H = ẑIN
for the field within the solenoid.

The assumption of zero magnetic flux density B = 0 for the exterior region is justified
because:
(a) if the exterior field is non-zero, then it must be independent of x and y (follows
from Ampere’s law applied to any exterior path C with IC = 0), and
(b) the finite interior flux Ψ = µo IN πa2 can only be matched with the flux of
the infinitely extended exterior region when the constant exterior flux density
(because of (a)) is vanishingly small.

3
 • Static electric fields: Curl-free and are governed by
∇ × E = 0, ∇ · D = ρ where D = ǫE
with ǫ = ǫr ǫo.

• Static magnetic fields: Divergence-free and are governed by

∇ · B = 0, ∇ × H = J where B = µH
with µ = µr µo — relative permeabilities µr other than unity (for free
space) will be explained later on.

Mathematically, we can generate a divergence-free vector field B(x, y, z)

as
B=∇×A
by taking the curl of any vector field A = A(x, y, z) (just like we can generate
a curl-free E by taking the gradient of any scalar field −V (x, y, z)).
Verification: Notice that
∂ ∂ ∂
∂ ∂ ∂ ∂x ∂y ∂z
∂ ∂ ∂
∇·∇×A = (∇ × A)x + (∇ × A)y + (∇ × A)z = ∂x ∂y ∂z
∂x ∂y ∂z
Ax Ay Az
∂ ∂Az ∂Ay ∂ ∂Az ∂Ax ∂ ∂Ay ∂Ax
= ( − )− ( − )+ ( − ) = 0.
∂x ∂y ∂z ∂y ∂x ∂z ∂z ∂x ∂y
• If B = ∇ × A represents a magnetostatic field, then A is called mag-
netostatic potential or vector potential.

4
 – Vector potential A can be used in magnetostatics in similar ways
to how electrostatic potential V is used in electrostatics.
◦ In electrostatics we can assign V = 0 to any point in space
that is convenient in a given problem.
◦ In magnetostatics we can assign ∇ · A to any scalar that is
convenient in a given problem.
– For example, if we make the assignment1

∇ · A = 0,

then we find that

∇ × B = ∇ × ∇ × A = ∇(∇ · A) − ∇2A = −∇2A.

This is a nice and convenient outcome, because, when combined with

∇ × H = J ⇒ ∇ × B = µoJ,

it produces
∇2A = −µoJ,
which is the magnetostatic version of Poisson’s equation
ρ
∇2 V = − .
ǫo
1
With this assignment — known as Coulomb gauge — A acquires the physical meaning of “potential
momentum per unit charge”, just as scalar potential V is “potential enegy per unit charge” (see Konopinski,
Am. J. Phys., 46, 499, 1978).

5
 – In analogy with solution
ρ(r′ )
Z
3 ′
V (r) = d r z J(r′)
4πǫo|r − r′ | r−r ′

of Poisson’s equation, it has a solution

r r′
µoJ(r′) 3 ′
Z
A(r) = d r.
4π|r − r′ |
y
2
Given any static current density J(r), the above equation can be used to O
obtain the corresponding vector potential A that simultaneously satisfies x

∇ · A = 0 and ∇ × A = B.
Once A is available, obtaining B = ∇ × A is then just a matter of taking a
curl.
• Magnetic flux density B of a single current loop I can be calculated
after determining its vector potential as follows: z
y
– For a loop of radius a on z = 0 plane, we can express the corresponding current
density as
(−y ′ , x′, 0)
a
p
′ ′ ′2 ′2
J(r ) = Iδ(z )δ( x + y − a) p
x′2 + y ′2
x
I
where the ratio on the right is the unit vector φ̂′ .
r
– Inserting this into the general solution for vector potential, and performing
the integration over z ′ , we obtain Jφ = Iδ(z)δ( x2 + y 2 − a)
2
Also, in quasi-statics we use J(r′ , t) to obtain A(r, t) and B = ∇ × A over regions small compared to
λ = c/f , with f the highest frequency in J(r′, t).

6
 µo I (−y ′ , x′, 0)
Z p
A(r) = ′2 ′2
δ( x + y − a) p p dx′ dy ′
4π (x − x′)2 + (y − y ′ )2 + z 2 x′2 + y ′2
µo I (−y ′, x′, 0)
Z
′
= δ(r − a) p r′ dr′ dφ′
4π (x − x′ )2 + (y − y ′ )2 + z 2 r′
Z π
µo I (−a sin φ′ , a cos φ′ , 0)
= p dφ′ ≡ x̂Ax (r) + ŷAy (r).
4π −π (x − a cos φ ) + (y − a sin φ ) + z
′ 2 ′ 2 2

– Given that Az = 0, it can be shown that B = ∇ × A leads to

∂Ay ∂Ax ∂Ay ∂Ax
Bx = − , By = , Bz = − .
∂z ∂z ∂x ∂y
– From the expected azimuthal symmetry of B about the z-axis, it is sufficient
to evaluate these on, say, y = 0 plane — after some algebra, and dropping the
primes, we find, on y = 0 plane,
µo aI π z cos φ
Z
Bx = dφ,
4π −π (x2 + a2 + z 2 − 2ax cos φ)3/2
µo aI π z sin φ
Z
By = dφ,
4π −π (x2 + a2 + z 2 − 2ax cos φ)3/2
and
µo aI π a − x cos φ
Z
Bz = dφ.
4π −π (x2 + a2 + z 2 − 2ax cos φ)3/2
– We note that By = 0 since the By integrand above is odd in φ and the
integration limits are centered about the origin. Hence, the field on y = 0
plane is given as
B = x̂Bx + ẑBz
with Bx and Bz defined above.
– There are no closed form expressions for the Bx and Bz integrals above for an
arbitrary (x, z).

7
 ◦ However, it can be easily seen that if x = 0 (i.e., along the z-axis), Bx = 0
(as symmetry would dictate) and
µo aI π a µo Ia2
Z
Bz = dφ = .
4π −π (a2 + z 2 )3/2 2(a2 + z 2 )3/2 3

For |z| ≫ a, 2

µo Ia2
Bz ≈ , 1

2|z|3

z
0

which is positive and varies with the inverse third power of distance |z|.
-1 I
– Also, Bx and Bz integrals can be performed numerically. Figure -2

in the margin depicts the pattern of B̂ on y = 0 plane for a loop -3

of radius a = 1 computed using Mathematica. -3 -2 -1 0

x
1 2 3

600
H
• Note that circulation C B · dl around each closed field line (“linking” 400

the current loop) equals a fixed value of µoI — this dictates that the
200

average field strength |B| of a current loop is stronger on shorter field

z
0

lines closer to the current loop than on longer field lines linking the
loop further out. As a result |B| can be shown to vary as r−3 for large
-200

r. -400

-600
-600 -400 -200 0 200 400 600

• It can be shown that the equations for magnetic field lines of a current x

loop on, say, y = 0 plane, can be expressed as

r = L sin2 θ

in terms of radial distance r from the origin and zenith angle θ

measured from the z axis. Clearly, parameter L in this formula is the

8
 radial distance of the field line on θ = 90o plane, and the field line B∝I
formula is accurate only for r ≫ a. The Earth’s magnetic field had
qv × B
such a magnetic dipole topology as shown. qv × B

I
• Lorentz force due to the magnetic fields of a pair of current loops — also
known as magnetic dipoles — turns out to be “attractive” when the cur- I
rent directions agree (see margin). Bar magnets carrying “equivalent”
Loops with parallel
current loops of atomic origins interact with one another in exactly the currents attract one
another

same way — i.e., as governed by the second term of Lorentz force.

9
14 Faraday’s law and induced emf
Michael Faraday discovered (in 1831, less than 200 years ago) that a chang-
ing current in a wire loop induces current flows in nearby wires — today
we describe this phenomenon as electromagnetic induction: the current
change in the first loop causes the magnetic field produced by the current to
change, and magnetic field change, in turn, is said to induce 1 (i.e., produce)
electric fields which drive the currents in nearby wires. Definitions of E and
• While static electric fields produced by static charge distributions are B have not changed:
unconditionally curl-free, time-varying electric fields produced by cur-
rent distributions with time-varying components are found to have, in recall that
accordance with Faraday’s observations, non-zero curls specified by • E is force per unit sta-
tionary charge
∂B
∇×E=− Faraday’s law • B gives an additional
∂t force v × B per unit
at all positions r in all reference frames of measurement. Using Stoke’s charge in motion with
velocity v in the mea-
theorem, the same constraint can also be expressed in integral form as surement frame.
∂B
I Z
E · dl = − · dS Faraday’s law
C S ∂t
for all surfaces S bounded by all closed and directed paths C (with
the direction of C, indicated by an arrow, and direction of vector dS
related by right hand rule).
1
Relativistic derivation of static B given in Lecture 12 can be extended to show that Coulomb interactions of charges
in time-varying motions require a description in terms of time-varying B and E — see, e.g., Am. J. Phys.: Tessman, 34,
1048 (1966); Tessman and Finnel, 35, 523 (1967); Kobe, 54, 631 (1986). Thus, the cause of induced E is not really the
time-varying B, but rather the time-varying current J that is also producing the variation in B.

1
• The right hand side of the integral form equation above includes the
flux of rate of change of magnetic field B over surface S. C(∆t) δS

If contour C bounding S is “fixed” (unchanging) in the measurement v∆t

frame, then the equation can also be expressed as S

C
dl
d
I Z
Z

E · dl = − B · dS, Ψ(0) = S
B(r, 0) · dS, and

C dt S Ψ(∆t) =
Z
B(r, ∆t) · dS +
Z
B(r, ∆t) · dS.
S δS

where the right hand side is now expressed in terms of the rate of change Ψ(∆t) − Ψ(0)
Thus, =
of magnetic flux Z
∆t
Z B(r, ∆t) − B(r, 0) Z dS
· dS + δS B(r, ∆t) · .
S ∆t ∆t
Ψ≡ B · dS
S Hence in limit ∆t → 0
linking contour C over any surface S bounded by C.
dΨ Z ∂B Z
= S · dS − C v × B · dl,
dt ∂t
• This modification (the exchange of the order of integration and time
since
derivative on the right side) would not be permissible if path C were Z dS Z ∆tv × dl
B(r, ∆t) · = C B(r, ∆t) ·
moving within the measurement frame or being deformed in time — δS ∆t
Z
∆t
=− (v × B) · dl
but in such cases we could still express Faraday’s integral form equation C
because
with − dΨ
dt
on the right side, provided that we also modify the left side B · (v × dl) = dl · (B × v),
as in both representing the volume of
d
I Z
a parallelepiped formed by the vectors
(E + v × B) · dl = − B · dS dl, v, and B.
C dt S Note that velocity v does not have to be
where v denotes the velocity of motion or deformation of path C. constant around contour C.

– This is equivalent to the original equation, since, as shown in the margin,

∂B d
Z Z I
· dS = B · dS + v × B · dl
S ∂t dt S C
when C is changing continuously with velocities v.

2
 • A physical interpretation of the final equation

d
I Z
(E + v × B) · dl = − B · dS Integral form Faraday’s Law
C dt
| {z S }
dΨ
E =− Faraday’s eqn.
dt
is as follows: Magnetic field lines con-
tributing to Ψ form links
– the circulation integral on the left is the “voltage drop” once with path C (bounding S)
around the directed closed path C, representing the work done like the links in an ordinary
per unit charge (by the Lorentz force ∝ E + v × B) taken a chain — hence, Ψ is said to
full circle around C, which was denoted by Michael Faraday with a be the flux linking path C.
symbol E and called the emf (short for electro-motive force, which
B
is a bad name since E is work, and not force, per unit charge) for
the closed path, equaling the decay rate − dΨdt
of its linked mag-
v
netic flux Ψ (due to all sources of magnetic flux density B in the
region). dS
S
– if/when path C is occupied by a conducting wire loop of some
total conductance G = R1 , and a resistance R = G1 , a current C

I = GE = RE will flow around the loop in the circulation direction2 ,

2
I = Aσ|E + v × B| for a homogeneous wire loop with a conductivity σ and cross sectional area A. If
the loop length
H is L, then the loop conductance is G = Aσ
L
and therefore we find that I = GE, as claimed,
since E = C (E + v × B) · dl = |E + v × B|L around a homogeneous loop.

3
 driven by the non-zero field E + v × B within the wire accounting
for the non-zero E if/when − dΨ
dt is non-zero.
– in equivalent circuit models of conducting wire loops, Faraday’s
equation, re-written as RI = − dΨdt
, is effectively Kirchfoff’s voltage
law (KVL) applied to the loop, with RI on the left denoting the
(sum of all) voltage drops in the direction of C, while − dΨdt
on the
right denoting a voltage rise also in the direction of C.
◦ note that the emf E describes both the voltage drop RI and
voltage rise − dΨ
dt
appearing in the circuit model for the con-
ducting wire loop since E = RI and E = − dΨ dt are both true.
– in modern parlance (since Maxwell) the term emf and its symbol
E are used to refer to and denote sources of energy, e.g., battery
voltages and magnetic flux rate − dΨdt that drive currents I = R
E
B
around closed circuits3 .
v
• If path C is fixed in the measurement frame, then v = 0, and KVL for
such a stationary loop reads as dS
S
dΨ
I
E · dl = − ; C
C dt

– otherwise, that is if C is in motion, then

3
see Saslow, Am. J. Phys., 58, 22 (2021), for a discussion of Maxwell’s interpretation of emf and
electrical energy production in batteries. Also see Scanlon et al., Am. J. Phys., 37, 689 (1969) for a
discussion of E = C (E + v × B) · dl vs E = − dΨ
H
dt
.

4
 dΨ
I
(E + v × B) · dl = −
C dt
because in that case force per unit charge advected with path C S
will be E + v × B according to Lorentz force (note: any additional
velocity vq of a moving charge along C does not contribute because C I=
E
R
(vq × B) · dl = 0 if dl and vq are parallel).
Think of EMF as the sum of all the "voltage
rises" around the loop traversed in the
– In either case, if C is a physical conducting path with a total direction of loop current I that needs to
match the total "voltage drop" RI around

resistance R, then the emf − dΨ

dt drives a current
the same loop traversed in the same
direction.
That way, KVL which states that

− dΨ Sum of voltage rises = Sum of voltage drops,

I = dt is fulfilled.

R
around C in the circulation direction (determined by dl and dS
directions used in accordance with the right-hand-rule).

• The minus sign present in Faraday’s equation, E = − dΨ

dt
, assures that
induced current I produces an induced magnetic field that opposes the
flux change producing the emf — this fact is known as Lenz’s rule
and is in full accord with observations4 ./newpage

• According to Faraday’s law it appears that magnetic flux variations

− dΨ
dt can produce a non-zero emf independent of how the variations are
produced — the possibilities are:
4
Faraday’s law not having the minus sign (or in conflict with Lenz’s rule) would be non-physical, as it
would lead to unbounded growth of induced currents and fields (by aiding rather than opposing the flux
change producing the emf).

5
 1. Fixed C, but time-varying B,
2. B =const. (in space and time), but time-varying C (rotating or
changing size),
3. An inhomogeneous static B = B(r) in the measurement frame
and C in motion.
B
• Note that even in the absence of any electric field E in the measurement
frame, a non-zero emf
v
dΨ
I
(v × B) · dl = −
C dt S
E
can exist because of the motion of C through an inhomogeneous mag- C I=
R
netic field (if the field is homogeneous then dΨ
dt will be zero, implying
zero E), which will of course appear as an emf
dΨ′
I
′ ′
E · dl = − ′
C dt
for a second observer moving with C who sees a time varying electric
field E′ = v × B in her own frame (in addition to the inhomogeneous
but constant magnetic field B of the first frame appearing as a time-
varying magnetic field B′)5 .
E′
See Scanlon et. al., Am. J. Phys., 37, 698 (1969), for a discussion of I ′ =
5
R
for rigid C with resistance
R observed from different reference frames.

6
– Thus, having non-zero electric field circulations
I
E′ · dl′
C

under time-varying magnetic field conditions appears to be quite

comprehensible after all!
– Magnetic fields B in one frame will morph into electric fields E′ in
other frames because of (near) invariance of Lorentz force between
reference frames.
– Moreover a morphed E′ can even be non-conservative — i.e., non
curl-free — when B is inhomogeneous in space (or time) as we
have just seen.

7
 z
Boe−t/τ ẑ
y

Example 1: If
B = Bo e−t/τ ẑ,
what is the emf E taken over a stationary circular loop C of radius r = 10 m on x
z = 0 plane in counter-clockwise direction (looking down on z = 0 plane)? What
is current I if the loop resistance is R? C

Solution: Since counter-clockwise circulation is requested we take dS pointing in ẑ

direction to be consistent with the right hand rule. We then have
Z
Ψ = B · dS = (Bo e−t/τ ẑ) · (π102ẑ) = π102 Bo e−t/τ
S

over the circular surface S. Thus, the emf

dΨ Bo
E=− = π102 e−t/τ .
dt τ
The loop current will be I = RE in counter-clockwise direction of the computed cir-
culation E, which will be positive and counteract (i.e., strengthen) the weakening
Bz .

8
Example 2: Consider the magnetic flux density
µo I
B= φ̂
2πr
y
produced by current I flowing along the x axis. What is the emf E of a square
loop C of area 4 m2 moving on xy-plane with edges parallel to x- and y-axes, if 2 m/s
its center is located at
H y = 2t m as a function of time? Compute the emf E first
dΨ
as − dt and then as C (v × B) · dl to verify that the same values are obtained. 2m
2t
Solution: Given the described geometry, we have C
Z 1 Z 2t+1
µo I µo I 2t + 1 I x
Ψ(t) = dx dy = ln( ).
−1 2t−1 2πy π 2t − 1
Thus, the emf E is
dΨ µo I 2t − 1 ∂ 2t + 1 µo I 4 µo I
− =− ( ) ( )= = .
dt π 2t + 1 ∂t 2t − 1 π (2t + 1)(2t − 1) π(t2 − 41 )
µo I
Alternatively, since v = 2ŷ m/s, and v × B = 2 2πr x̂, we find, using dl = ±x̂dx
and ±ŷdy in turns,
µo I µo I µo I
I
E = (v × B) · dl = 2 2−2 2=
C 2π(2t − 1) 2π(2t + 1) π(t2 − 41 )
in consistency with the above result.

9
Example 3: A conducting loop of a radius r = 0.1 m (see figure in the margin) is
ω
being rotated about the x axis with frequency of f = 2π = 60 Hz in a region
with a DC magnetic field of B = 10ẑ T. Determine the induced current in the
loop if the loop resistance is 12 Ω.

Solution: The maximum value of the magnetic flux linking the loop should be

Ψo = π(0.1)210 = 0.1π Wb.

The time-varying flux linking the rotating loop is therefore

Ψ(t) = Ψo cos(ωt) = 0.1π cos(120πt).

The corresponding emf is

dΨ
E=− = (120π)0.1π sin(120πt).
dt
Therefore, the induced current around the loop must be
E 12π 2 sin(120πt)
I= = = π 2 sin(120πt) A.
R 12

10
 z
C ~v = 3x̂ m/s

X X X X X X X
Example 4: A conducting bar of resistance R1 = 1 Ω ohms is moved in the x-direction R2 2m
R1 = 1 Ω
with a velocity v = 3x̂ m/s on a pair of perfect conducting (R = 0) stationary X X X X X X X
x
3t
rails 2 m apart terminated with a load resistance R2 at x = 0, all constituting a
rectangular contour C to be taken counterclockwise. A constant magnetic field
of B = 1ŷ T is linked throught contour C such that the flux Ψ = −1 × 2 × 3t
and the emf E = −dΨ/dt = 6 V. Hence, Faraday’s law demands that
I Z t Z b
(E + v × B) · dl = (E + v × B)1 · dl + (E)2 · dl = 6 Moving bar in the presence
C b t of a constant magnetic field
produces an emf and electric
where the two integrals (with b and t referring to bottom and top rail contact fields in the lab frame that
points) correspond to voltage drops across resistors R1 and R2 , respectively. But drive a loop current I.
since Z t
(v × B)1 · dl = 3 × 1 × 2 = 6, Example
H 4 illustrates
H how the
b E · dl part of emf (E + v ×
it follows that B) · dl caused by a motion v =
Z t Z b 3x̂ m/s is zero (with non-zero
(E)1 · dl + (E)2 · dl = 0 ⇒ Ez1 − Ez2 = 0 ⇒ Ez2 = Ez1, static Ez components)!!
b t
i.e., identical static fields within the moving and stationary bars across the perfect
conducting rails. This may be a surprising claim/result — let’s give two examples
to illustrate how this happens:
1. Let R2 = 2 Ω ohms. Then I = 6/3 = 2 A. It follows that voltage drops (E + v × B)1 · 2ẑ = 2
V across R1 and (E)2 · (−2ẑ) = 4 V across R2 , yielding Ez1 = Ez2 = −2 V/m.

2. Let R2 = ∞ — open ckt load to the moving conductor. Then I = 6/∞ = 0 A. It follows that
(E + v × B)1 · 2ẑ = 0 V across R1 and (E)2 · (−2ẑ) = 6 V across R2 , yielding Ez1 = Ez2 = −3
V/m. Note that in this
R t case the entire emf appears across the open termination (gap in the
loop C and the emf b (E + v × B)1 · dl = 0 across resistor R1 ).

11
Example 5: An infinite solenoid producing a constant − dΨ dt = 8 V, passes through
small a loop consisting of a 1 Ω resistor on the right and a 3 Ω resistor on the
left, connected in series — see margin plot. What is the current Ic through this
resistor loop, and what voltages would be measured (by a voltmeter) across the
individual resistors?

Solution: The magnetic flux produced by the solenoid will be confined to its interior
Transformers which operate
as long as dI/dt (and thus dΨ/dt, as specified) is constant and emf E = −dΨ/dt based on an inductive cou-
is non-time varying (see below). In that case, with constant emf E = − dΨ
dt = 8
pling principle, and electric
dynamos (and motors) which
V of the encircling resistor loop in the setup, the loop current Ic is the ratio of produce motion induced emfs
E (and rotating coils) are stud-
E and the total loop resistance 4 Ω, i.e., Ic = R = 2 A. Consequently, 1 and 3
ied in depth in power courses
Ω resistors will develop 2 and 6 V drops, respectively, in the direction of the 2A starting with ECE 330.
current!! Note that:

• the loop has no battery to support this current flow — it has instead been excited
“inductively”.
• with constant dI/dt, there is zero magnetic field at the locations of the loop wire and
resistors (static E in the solenoid exterior is curl-free!) — thus, the emf of the loop is
not being produced by a time varying local magnetic field; it is rather a consequence of
the time-varing current I(t) in the solenoid (which is also responsible for time-varying
Ψ), with the relation E = −dΨ/dt being “incidental”!
• what a voltmeter measures across the resistors — whether 2 or 6 V — depends on
whether its probes contacting points A and B are placed to the right or to the left of
the solenoid!! That’s because the field E produced by the time-varying current I(t) is
no longer conservative across the system and consequently the line integral of E is path
dependent — we have to be more careful about what we mean by voltage in these new
situations!

12
 y
A

Example 6: Consider a square conducting loop of 1 m2 cross sectional area bordered R1 R2

by R1 = 2 Ω and R2 = 1 Ω resistors as shown in the margin. The loop is C

linked with a magnetic flux Ψ due to time varying magnetic field described as B

x
B = (12 − 3t)ẑ T.

• Hence, Ψ = 12 − 3t Wb and the emf E = −dΨ/dt = 3 V. In the presence of time vary-

ing magnetic flux, voltage of
• Loop current I = 3V/3Ω = 1 A in the circulation direction. R
a path P , defined as P (E +
v × B) · dl, will in general be
• Voltage drop V1 = 2 V across R1 from point A to point B. path dependent!

• Voltage drop V2 = −1 V across R2 from point A to point B. A voltmeter reads and dis-
plays the voltage of its
• A voltmeter connected from A (positive lead) to B will read 2 V if and only if its own path constituted by the
leads form a path identical to the path defined by R1 (from A to B). placement of its own probe
wires contacting the mea-
• A voltmeter connected from A (positive lead) to B will read -1 V if and only if surement nodes A and B.
its leads form a path identical to the path defined by R2.

• A voltmeter connected from A (postive lead) to B will read 0.5 V if its leads form
a diagonal path from A to B.

– To see this, notice that Faraday’s law applied for the triangular loop in-
cluding the voltmeter and R2 would have an emf of 1.5 V equaling the sum
of voltmeter reading VR and 1 V drop across resistor R2 .

13
 (a) A one turn coil with current I
generates its own linked magnetic
15 Inductance — coil, solenoid, shorted coax flux LI as shown, where a non-
negative L is the inductance of
the coil.

• Given a circular coil with some resistance R and conducting some cur- 3

rent I, the magnetic flux Ψ produced by I and “linking” the coil itself
2

— see figure on the right — can be expressed as

1

Ψ = LI Ψ

z
0

using a non-negative proportionality constant -1 I

Ψ -2

L=
I -3
-3 -2 -1 0 1 2 3

termed the self -inductance of the coil measured in units of Henries

x
(b) An equivalent circuit model
for the coil expressed in terms of
(H=Wb/A)1 . lumped resistor R and inductor L
forming a loop carrying the loop
current I R
• Given Ψ = LI, and its time derivative
dΨ dI dI
=L , I, E = −L
dt dt dt
it follows that Faraday’s equation applied to the coil is + dI -
V (t) = L
dΨ dI dt
E =− = −L , The emf RI=-LdI/dt of the coil
dt dt appears as a voltage rise across
the inductor in the ckt model,
as well as a voltage drop across
indicating a self -emf −L dI dt
representing a voltage rise around the the resistor, both taken in the
direction of current I. Voltage
coil in the direction of current flow I = E/R — see an equivalent circuit drop V across the inductor in the
current direction is LdI/dt, as
we learned in our circuit courses.
model for the coil derived from these relations shown on the right.
1
As opposed to a mutual inductance M, also measured in Henries, relating the flux linking a coil C to
a current Io flowing in a second coil Co .

1
 (a) A one turn coil with current I
– The current I and self-emf E are then the solutions of differential generates its own linked magnetic
flux LI as shown, where a non-
equations negative L is the inductance of
the coil.
dI dE
RI = −L and RE = −L ,
dt dt 3

respectively, and exhibit an exponential decay with a time constant 2

of τ = L/R (just like in LR circuits seen in ckt courses, and in 1

analogy with time constant τ = RC that governs voltage decays Ψ

z
0

in RC circuits).
-1 I
◦ Note that τ = L/R implies that when the inducance L is
large, so is time constant τ , and current decay in the induc-
-2

tor is slow — inductors with large L will behave like slowly -3

-3 -2 -1 0 1 2 3
x

time-varying current sources (just like capacitors behaving like (b) An equivalent circuit model
for the coil expressed in terms of
time-varying voltage sources) as they relase their stored energy lumped resistor R and inductor L
forming a loop carrying the loop
(while maintaining a voltage rise −L dI dt determined by other
current I R
elements in their connected circuits).
dI
• For an inductor consisting of n-loops, the emf E measured across all I, E = −L
dt
n-loops is naturally (since n emf’s add up)
+ dI -
V (t) = L
d d dI dt
E = n(− Ψ) = − nΨ ≡ −L The emf RI=-LdI/dt of the coil
dt dt dt appears as a voltage rise across
the inductor in the ckt model,
as well as a voltage drop across
implying an inductance the resistor, both taken in the
direction of current I. Voltage
nΨ drop V across the inductor in the
L≡ . current direction is LdI/dt, as
I we learned in our circuit courses.

2
 Example 1: An n-turn coil has a resistance R = 1 Ω and inductance of 1 µH. If it is
conducting 3 A current at t = 0, determine I(t) for t > 0.

Solution: Current flow in the resistive n-turn coil will be driven by self-emf E = −L dI
dt
matching a voltage drop RI. Hence
dI dI R R 6
RI = −L ↔ + I = 0 ⇒ I(t) = I(0)e− L t = 3e−10 t A.
dt dt L

• As illustrated by above example, current I around a resistive loop C

will in general be obtained by solving a differential equation constructed
using the emf of the loop.

– The algebraic I = RE solution used last lecture assumed that self-

emf −L dI
dt
produced by the induced current I(t) is small compared
to an externally produced emf.

We continue with typical inductance calculations.

3
 Inductance of long solenoid: Consider a long solenoid with length ℓ, B = ẑB
cross-sectional area A, and a density of N loops per unit length as examined
in Example 3 of Lecture 12 (see figure in the margin). As determined in
Example 3, the magnetic flux density in the interior of the solenoid is

B = µoIN ẑ

while n = N ℓ is the number of turns of the solenoid. Thus, the inductance

of the solenoid with n = N ℓ turns is ℓ I
nΨ N ℓ(µoIN )A
L= = = N 2 µoAℓ.
I I
• As we know from our circuit courses, an inductor L such as the solenoid
coil considered above can be used to store energy. An inductor con-
nected to an external circuit with a quasi-static current I develops a
voltage drop V = L dI 2 B = µoIN
dt across its terminals and absorbs power at an
instantaneous rate
dI d 1
P =VI =L I = ( LI 2),
dt dt 2
implying a stored energy of
1 2 1 2 2 |Bz |2 1
W = LI = N µoAℓI = Aℓ = µo|Hz |2Aℓ
2 2 2µo 2
in an inductor in a conducting state.
2
Assuming a physical size much smaller than a wavelength λ = c/f for the highest frequency in I(t).

4
• Notice that the stored energy of the solenoid is
1 1
µo|Hz |2 = µoH · H
2 2
times its volume Aℓ occupied by the field H inside the solenoid. That
suggests that
1
w = µo H · H
2
can be interpreted as stored magnetostatic energy per unit volume in
general.

– Also both inductance L and stored energies W and w would have

µ replacing µo in material media with permeabilities

µ = (1 + χm )µo

and magnetic susceptibilities χm , in analogy with the concepts of

permittivity ǫ = (1 + χe )ǫo and electrical susceptibility χe.
◦ Permeability and magnetic susceptibility notions will be ex-
amined in a future lecture.

5
Inductance of shorted coax: Consider a coaxial cable of some length ℓ I
which is “shorted” at one end (with a wire connecting the inner and outer
r Short
conductors), so that a steady current I can flow on the inner conductor of b a I
radius a to return on the interior surface of the outer conductor at radius z
b after having circulated through the short. We will next determine the B
inductance L of such an inductor after first computing the magnetic flux
density Bφ that will be produced by the inner conductor current I. In Bφ
calculation we will assume ℓ ≫ b so that an “infinite coax” approximation ℓ
can be invoked. Shorted coax circulates
a current I linking a
• Expanding the integral form of Ampere’s law magnetic flux Ψ
confined to a region
I bounded by the outer
conductor of the coax.
B · dl = µoIC
C
as
Bφ2πr = µoI
over a circular integration contour C of a radius r > a, we find that
the magnetic flux density in the interior of the coax cable is
µo I
Bφ = .
2πr
• Therefore, the magnetic flux linked by the closed current path I (see
figure in the margin) is
Z b
µo dr µo b
Z
Ψ = B · dS = ℓ I = ℓ ln I.
S 2π a r 2π a
6
Clearly, we have a linear relation Ψ = LI, with

ln ab
L≡ ℓµo,
2π
which is the inductance of a shorted coax of a finite length ℓ.

– The inductance of the coax per unit length is

ln ab
L= µo ,
2π
which should be contrasted with capacitance per unit length
2π
C= ǫo
ln ab
of the same coax configuration.
Notice how L and C are proportional to ǫo and µo, respectively,
having proportionality constants which are inverses of one another.

7
Inductance of shorted parallel plates: If a pair of parallel plates of
areas A = W ℓ and separation d were shorted at one end, we would obtain
effectively an inductor with a per length inductance
d
L= µo
W
that accompanies per length capacitance
W
C= ǫo
d
of the same parallel plate configuration. This follows from a generalization of
our finding above that the proportionality constants of L and C are arithmetic
inverses of one another.

8
16 Charge conservation, continuity eqn, displace- (a) At t=0 volume V contains
a neutral atom but no net charge

ment current, Maxwell’s equations S

V
H
t=0
• Total electric charge is conserved in nature in the following sense: QV = 0
if a process generates (or eliminates) a positive charge, it always does
so as accompanied by a negative charge of equal magnitude.
(b) At t=t1 volume V contains
a proton and a free electron after
the ionization of the hydrogen
– Example: Photoionization of atoms and molecules can generate atom. There is still no net charge
in the volume.
free positive ions and free negative electrons in pairs (see margin S
-e
figure). Photoionization is a process that converts bound charge V e
carriers into free charge carriers. t = t1
QV = 0
– Example: Recombination when a positive ion and an electron
get together to produce a charge neutral atom or molecule.
(c) At t=t2 volume V now contains
– Example: Annihilation of an electron (negative charge) by a only a proton after the exit of
free electron through surface S.
Now V contains a net charge e.
positron (positive charge of equal magnitude) and the reverse pro- -e
S
cess of pair creation.
V
t = t2 e
As a consequence, if the total electric charge QV contained in any finite QV = e
volume V changes as a function of time, this change must be attributed
to a net transport of charge, i.e., electric current, across the bounding
surface S of volume V as detailed below.

1
• Consider two distinct surfaces S1 and S2 bounded by the same closed
loop C (as shown in the margin) such that a volume V is contained
between the two surfaces.

– Let dS2
Z S2
I1 = J · dS1 V dS
S1
and Z dS1
I2 = J · dS2 C
S2
S1
denote currents flowing through surfaces S1 and S2, respectively.
– Note that current I1 through surface S1 enters volume V , while
current I2 through surface S2 exits volume V (with the directions
assigned to dS1 and dS2).
– If I1 6= I2, then current out is not matched by the current in,
and as a result, the net charge QV contained in volume V increases
with time at a rate I1 − I2 provided that charge is conserved in
the sense discussed above. In that case, we have
dQV
= I1 − I2 .
dt
This relationship can be expressed as
d
Z Z Z
ρdV = J · dS1 − J · dS2
dt V S1 S2

2
 since, in terms of charge density ρ, charge in volume V is
Z
QV = ρdV.
V

The expression can also be cast as

∂ρ
Z I
dV = − J · dS
V ∂t S

where S is the union of surfaces S1 and S2 enclosing V , and dS

is an outward area element of S (see margin). This relationship is
known as continuity equation. Its differential form is Continuity
equation
∂ρ
= −∇ · J,
∂t
which follows from the integral form above as a consequence of
divergence theorem (recall Lecture 4).

Continuity equation is a mathematical re-statement of the

principle of conservation of charge.

3
• While Faraday’s law
∂B
∇×E=−
∂t
indicates that time-varying B induces time-varying electric fields E,
Ampere’s law, written as

∇ × H = J,

makes no such claim about a time-varying E inducing a time-varying

B = µoH.

– This “asymmetry” was noted by James Clerk Maxwell who realized

that the form of Ampere’s law given above must be “incomplete”
under time-varying situations. Revised
– Noting the inconsistency of Ampere’s law with the continuity equa- Ampere’s
tion under time varying conditions, he re-wrote the Ampere’s law law (with
as “displacement
∂D current”)
∇×H=J+
∂t
in 1861 by adding the term on the right which is now called the
“displacement current”.
◦ Maxwell postulated that the displacement current term is needed
in Ampere’s law because only then the divergence of Ampere’s
law avoids falling into conflict with charge conservation (under
time varying conditions).

4
 Verification of Maxwell’s claim: Since ∇ × H is divergence-free
(just like the curl of vector potential A, namely B), it follows that
the divergence of Maxwell’s modified Ampere’s law — often called
Ampere-Maxwell equation — is
∂
∇ · (∇ × H) = ∇ · J + ∇ · D = 0.
∂t
– In the absence of the second term due to displacement current,
this results would be inconsistent with the continuity equation
∂ρ
+ ∇ · J = 0,
∂t
∂ρ
unless ∂t = 0 (the static case).
– By, contrast, including the second term, the result above is rec-
ognized as the continuity equation per se, since by Gauss’s law
— assuming that it applies with no change under time varying
situations —
∂ ∂ρ
∇·D= .
∂t ∂t
• The modified Ampere’s law
∂D
∇×H=J+
∂t
postulated by Maxwell under the assumption that Gauss’s law is also
valid under time-varying conditions, leads to some specific predictions
about how time-varying fields should behave.

5
 • These predictions — concerning the propagation of electromagnetic
waves — were validated experimentally by Heinrich Hertz around 1888.

– The experiments confirmed that time-varying electric and mag-

netic fields obey collectively (and at microscopic scales) the differ-
ential relations Maxwell’s
equations
∇·D = ρ Gauss’s law
∇·B = 0
∂B
∇×E = − Faraday’s law
∂t
∂D
∇×H = J+ , Ampere’s law
∂t
where
D = ǫoE and B = µoH
provided that ρ and J describe the distributions of all charges and
currents associated with free and bound charge carriers1 . Alter-
natively, the same differential relations — known collectively as
Maxwell’s equations — are also valid for macroscopic fields, pro-
vided that ρ and J describe only the free charge contributions
and
D = ǫE and B = µH
1
In the classical domain, down to scales of about ℏ/mc, the Compton wavelength — at shorter scales
quantized and generalized versions (known as electroweak theory) are needed.

6
 in terms of suitably defined permittivities and permeabilities ǫ and
µ — see next Lecture.

• The unnamed Maxwell equation

∇·B=0

can be viewed to be a consequence of Faraday’s law

∂B
∇×E=−
∂t
and the fact that magnetic monopoles have never been observed.
Explanation: Since ∇ × E is divergence-free, taking the divergence
of Faraday’s law, we get
∂
∇ · (∇ × E) = − ∇ · B = 0.
∂t
This constraint requires ∇ · B to an invariant scalar at all locations
in space. As a consequence, if ∇ · B = 0 at some instant in time, it
should remain so at all times. Given that ∇ · B = 0 for static fields,
this relationship must also continue to be valid when B starts changing
with time.

The fact that ∇ · B remains fixed at a zero value everywhere, whereas ∇ · D

varies like ρ, is in fact a consequence of the fact that there appears to be no
magnetic charges (monopoles) in nature. Had there been “point charges for

7
magnetic fields” in nature, ∇ · B would have equaled the density of those
charges, and magnetic field lines would have started and stopped on them
(rather than looping into themselves). But no one has observed of any evi-
dence for such magnetic charges anywhere, even going back to the very early D+
n̂
times in the history of the universe (accessible by making observations of w D−
very far astronomical objects). So, ∇ · B = 0.

• Finally, the full set of Maxwell’s boundary condition equations concern-

Constraint
ing any interface with a normal unit vector n̂ are
∂D
I Z
H · dl = (J + ) · dS
∂t
n̂ · (D+ − D−) = ρs C S

around the dotted path yields

n̂ · (B+ − B−) = 0
n̂ × (H+ − H− ) = Js
n̂ × (E+ − E−) = 0
in w → 0 limit.
n̂ × (H+ − H−) = Js

– We had already seen how the first and third boundary condition
Constraint
equations arise. I
B · dS = 0
– The second boundary condition equation concerning the normal S

component of B is another consequence of the absence of magnetic applied over the dotted volume (seen in
profile) yields
charges (see margin).
Bn+ − Bn− = 0
– A detailed justification of the last boundary condition concerning
tangential H will be given explicitly during Lecture 19. This equa- in w → 0 limit.

tion allows a discontinuous change in the tangential component of

H if the interface contains a non-zero surface current Js .

8
17 Magnetization current, Maxwell’s equations
in material media
• Consider the microscopic-form Maxwell’s equations
∇·D = ρ Gauss’s law
∇·B = 0
∂B
∇×E = − Faraday’s law
∂t
∂D
∇×H = J+ , Ampere’s law
∂t
where
D = ǫo E
B = µoH.

• Direct applications of these equations in material media containing a

colossal number of bound charges is impractical.

• Macroscopic-form Maxwell’s equations suitable for material media are

obtained by first expressing ρ and J above as the macroscopic quantities
ρ = ρf − ∇ · P
and
∂P
J = Jf + +∇×M
∂t
where

1
 – subscripts f indicate charge and current density contributions due
to free charge carriers,
– the term −∇ · P denotes the bound charge density,
– the term ∂P∂t denotes the polarization current density due to
oscillating dipoles (already discussed in Lecture 11), and
– ∇×M is a “magnetization current density” also due to bound
charges, an effect that we will discuss and clarify later in this
section.

Using these expressions in Gauss’s and Ampere’s laws

∇ · ǫo E = ρ Gauss’s law
∂ǫoE
∇ × µ−1
o B = J + , Ampere’s law
∂t
we obtain

∇ · (ǫoE + P) = ρf Gauss’s law

∂
∇ × (µ−1
o B − M) = Jf + (ǫoE + P), Ampere’s law.
∂t
Now, re-define D and H as

D = ǫeE + P = ǫE

and
H = µ−1 −1
o B − M = µ B,

2
 respectively, and drop the subscripts f which will no longer be needed.
Using these new definitions, the full set of Maxwell’s equations now
read as (the same form as before)
∇·D = ρ Gauss’s law
∇·B = 0
∂B
∇×E = − Faraday’s law
∂t
∂D
∇×H = J+ , Ampere’s law
∂t
with
D = ǫE
B = µH,
where ρ and J are understood to be due to free charge carriers only.

• We had already seen many aspects of the above procedure for obtaining
the macroscopic form field equations earlier on (e.g., in Lectures 8 and
11).

– In particular we were already familiar with the revised definition

of D = ǫE along with the concept of medium permittivity ǫ.
– The new feature above that requires further discussions is the rela-
tion B = µH along with the concept of medium permeability
µ. The details of this relation are connected to the concept of
“magnetization current” which we discuss next.

3
• Just like “free charge” density and currents, “bound charge” densities
and currents also have to satisfy the continuity equation
∂ρ
+ ∇ · J = 0.
∂t
– This equation is automatically satisfied if we substitute

ρ = ρb = −∇ · P

and
∂P
J = Jb =
∂t
in it.
Verification:
∂ρb ∂ ∂P
+ ∇ · Jb = (−∇ · P) + ∇ · =0
∂t ∂t ∂t
since the order of time derivative and divergence can be exchanged
on the right.
– But the same equation is also satisfied if we take
∂P
Jb = +∇×M
∂t
for any vector field M simply because vector ∇ × M is divergence
free.

4
 Consequently, it is not sufficient to represent bound currents in mate-
rial media as simply ∂P
∂t , if bound carriers can also conduct divergence-
free currents due to closed-loop orbits.

– In fact, electrons “orbiting” atomic nuclei certainly produce such

divergence-free current loops at microscopic scales — we account
for such currents at macroscopic scales by including a magnetiza-
tion current term ∇ × M in Jb .
– Also, bound charge motions within nucleons1 — proton and neu-
trons — produce magnetization currents ∇ × M.
– Even bare electrons can produce magnetization currents ∇ × M
because of their intrinsic spin 2.

Once ∇ × M is included in Jb , it follows from Ampere’s law that

H = µ−1
o B−M

where M is referred to as magnetization field.

1
Physical models of nucleons involve bound charge carriers known as quarks which cannot be observed
in a free state.
2
All elementary charge carriers carry an intrinsic magnetization proportional to charge-to-mass ratio
q
m
and a “spin angular momentum” having quantized values of ± ~2 N.m.s in any measurement direction.
Using Heisenberg’s uncertainty principle, ∆p∆r ≥ ~2 , we can interpret the spin angular momentum of
an elementary particle as the lower bound of ∆p∆r, the product of quantum uncertainties in particle
momentum and position. There is no classical interpretation of spin angular momentum for point particles.

5
• To get a physical picture about magnetization M and the physical
origin of H = µ−1 o B − M consider a solenoid wound around some
cylindrical shaped material as shown in the margin. We know that
with a solenoid current Io, we would have Ho = N Io ẑ in the interior of
a solenoid with N loops per unit legth aligned with the z-axis, and a
corresponding magnetic flux density Bo = µoN Ioẑ when the solenoid
Io
core is free space. This will be modified to some B = Bo + µoM when
a material core is introduced into the same space, where µoM stands Within the core, stacks of
atomic loop currents are
effective solenoids giving
for the (additional) macroscopic (space averaged) magnetic flux den- rise to an additional
magnetic flux density of
an average magnitude

sity produced by microscopic current loops localized within the atoms Bo = µoIoN
inside the solenoid
µo Al Il Nl = µoM
but outside the cylindrical that adds to Bo
constituting the core. magnetic core

1
– If there are Na = ∆x∆y∆z atoms per unit volume in the core, with
∆x separations in x direction and so forth, loop currents Il of a
stack of atoms with ∆z separations in z would produce an effective
Il
solenoid an internal z-directed magnetic flux density of µo ∆z ẑ and
zero exterior field.
– Since one such atomic stack solenoid with a loop area of Al will be
found for every ∆x∆y cross-sectional area of the core, a macro-
scopic average magnetic flux density produced by these atomic
solenoids would be calculated as (this calculation is similar to
finding the average polarization field in a dielectric as discussed in
Al Il
Lecture 8) ∆x∆y × µo ∆z ẑ = µoNaIl Al ẑ ≡ µoM, with M = Na m,
m ≡ Il Al ẑ.

6
 ◦ Here m is the magnetic dipole moment of each current loop
(analogous to electric dipole p = qr), M is the magnetization
field vector (analogous to P = Nap), which is a simple product
of m per magnetized atom and the atomic number density Na
in the core.
– Superposing the magnetic flux densities of µoM andBo, we obtain
B = Bo +µoM for the core region, or for any region of space having
a non-zero magnetization M, which then leads to the general result
H = µ−1o B − M, which is further discussed below.
– Notice, whether the flux density B = Bo +µoM inside the material
medium is stronger or weaker in magnitude than Bo depends on
the direction of M, which, in turn, depends on the algebraic sign
of microscopic loop currents Il introduced above.
◦ Negative Il is found in diamagnetic materials where |B| <
|Bo|, while positive Il in paramagnetic and ferromagnetic
materials where |B| > |Bo|, as discussed below.
– Also, the expression H = µ−1 −1
o B − M leads to H = µo B = Ho
in the exterior region where M = 0, indicating that while fields B
and Boif the interior and exterior are different, H is the same in
both regions (analogous with D in dielectrics).

7
 • Lab measurements — e.g., inductances L measured for coils wound
around magnetic materials3 — show that for a large class of materials

M ≡ µ−1
o B−H

varies linearly with H (which is of course possible only when B also

varies linearly with H).

– In that case we write

M = χm H,
where χm is a dimensionless parameter called magnetic suscep-
tibility, and obtain a relation Magnetic
susceptibility
B = µo(1 + χm )H = µH, and
where permeability
µ = µo(1 + χm)
is called the permeability of the medium.

3
Recall from Lecture 15 that L ∝ µ when inductors are wound around materials with permeability µ.

8
 • For a large class of materials with M ∝ H, it is observed that |χm| ≪ 1.
In that case, the material is called

– Diamagnetic if χm < 0:
◦ Diamagnetism occurs when an applied magnetic field induces
electron orbital angular momentum in a collection of atoms
having no net permanent magnetization M — in such materi-
als electron clouds around atomic nuclei spin up in accordance
with Lenz’s to generate magnetic fields opposing the applied
magnetic field so as to keep B = µH smaller than µoH. This
happens in materials that we ordinarily think of being non-
magnetic (wood, glass, water, etc.). Diamagnetic materials
are in fact very weakly repulsed by permanent magnets since
µ ≈ µo in all diamagnetic materials.
– Paramagnetic if χm > 0:
◦ Paramagnetism occurs in materials composed of atoms hav-
ing permanent magnetic dipole moments due to electron spin
angular momentum — magnetic dipoles of such atoms co-
align with the applied magnetic field due to v × B related
torques, leading to M pointing in the applied B direction4 .
This happens for atoms with unfilled inner electron shells, be-
cause in filled shells electron spins are opposite (due to Pauli
4
In these materials the described paramagnetism overcomes the diamagnetic tendency of the material
caused by the orbital angular momenta of its electrons around atomic nuclei.

9
 exclusion principle) and cancel one another. Unfilled outer
shells do not usually give rise to paramagnetism because in-
teractions between adjacent atoms in that case give rise to
opposite spins of their outer shell electrons. Paramagnetic
materials are very weakly attracted to permanent magnets
(e.g., aluminum, lithium, tungsten).

• In a small class of materials known as ferromagnets — iron, nickel,

and cobalt, which are metals with atoms having unfilled inner elec-
tron shells, and their various alloys — M can arise spontaneously (be-
cause permanent magnetic dipole moments of nearby atoms produced
by electron spins become co-aligned as a consequence of conduction
electrons moving through the lattice) and turns out to be a non-linear
function of present and past values of H, in which case experimentally
obtained relations, denoted as

B = B(H),

need to be used in Maxwell’s equations. It is even possible to have

non-zero B in such materials with zero H — permanent magnets have
that property.

• First principles modeling of χm or the B = B(H) relation requires

quantum mechanics (classical models turn out to be not accurate enough).
Overall, the models give rise to frequency dependent results, involving
loss as well as resonance features (also exhibited in Lorentz-Drude mod-

10
els of χe examined in Lecture 11) relevant for applications including
various magnetic imaging techniques.

11
18 Wave equation and plane TEM waves in source-
free media
With this lecture we start our study of the full set of Maxwell’s equations
shown in the margin by first restricting our attention to homogeneous and
non-conducting media with constant ǫ and µ and zero σ.

• Our first objective is to show that non-trivial (i.e., non-zero) time- ∇·D = ρ
varying field solutions of these equations can be obtained even in the ∇·B = 0
absence of ρ and J. ∂B
∇×E = −
∂t
– We already know static ρ and J to be the source of static electric ∂D
∇×H = J+ .
and magnetic fields. ∂t
– We will come to understand that time varying ρ and J, which
necessarily obey the continuity equation
∂ρ
+ ∇ · J = 0,
∂t
constitute the source of time-varying electromagnetic fields.

Despite these intimate connections between the sources ρ and J and

the fields
D = ǫE and B = µH,
non-trivial field solutions can exist in source-free media as we will see
shortly.

1
• Such field solutions in fact represent electromagnetic waves, a familiar
example of which is light.

• Another example is radiowaves that we use when we communicate

using wireless devices such as radios, cell-phones, WiFi, etc.

• Different types of electromagnetic waves are distinguished by their os-

cillation frequencies, and include

– radiowaves,
– microwaves,
– infrared,
– light,
– ultraviolet,
– X-rays, and gamma rays,

going across the electromagnetic spectrum from low to high fre-

quencies.
We are well aware that these types of electromagnetic waves can travel
across empty regions of space — e.g., from sun to Earth — transporting
energy and heat as well as momentum.

– Next, we will discover their general properties by examining Maxwell’s

equations under the restriction ρ = J = 0.

2
• In source-free and homogeneous regions where ρ = J = 0 and ǫ and
µ are constant, we can simplify Maxwell’s equations as shown in the
margin.

– If there are non-trivial solutions of these equations, namely E(r, t) 6= ∇·E = 0

0 and H(r, t) 6= 0, they evidently need to be divergence-free. ∇·H = 0
∂H
– They also have to be “curly” according to the last two equations: ∇ × E = −µ
Faraday’s and Ampere’s laws. ∂t
∂E
∇×H = ǫ .
• Next we will make use of vector identity ∂t

∇ × (∇ × E) = ∇(∇ · E) − ∇2E
which should be familiar from an earlier homework problem.

– Since the electric field E is divergence-free in the absence of sources,

this identity simplifies as
∇ × (∇ × E) = −∇2E
where in the right side ∇2E is the Laplacian of E.
– Using this result we can express the curl of Faraday’s law as
∂H ∂
∇ × [∇ × E = −µ ] ⇒ −∇2E = −µ ∇ × H,
∂t ∂t
which combines with the Ampere’s law to produce
2 ∂ 2E
∇ E = µǫ 2 ,
∂t
3
 which can be written explicitly as 3D vector
wave
∂ 2E ∂ 2E ∂ 2E ∂ 2E
+ 2 + 2 = µǫ 2 . equation
∂x2 ∂y ∂z ∂t
Recall that our objective is to see whether a non-trivial time-varying solution
of Maxwell’s equations can exist in source-free media.

Our objective at this stage is not finding a general solution; it is instead

identifying a simple example of a non-trivial time-varying E(r, t), if we can.

For example, can a field solution

E(r, t) = x̂Ex (z, t)

that only depends on z and t and “polarized” in x-direction exist? If it can

exist, what would be the properties of this x-polarized solution?
• To find out, we note that with E = x̂Ex(z, t), the above “wave equation”
is reduced to 1D scalar
2 2
∂ Ex ∂ Ex wave
2
= µǫ 2
,
∂z ∂t equation
an equation that is known as a 1D scalar wave equation, as opposed
to the 3D vector wave equation above.

– Now, by substitution, we can easily show that

√
Ex = cos(ω(t − µǫz)),

4
 satisfies the 1D wave equation and represents an x-polarized time-
periodic field solution with an oscillation frequency ω.
– 1D wave equation can also be satisfied by
√
Ex = cos(ω(t + µǫz)).

Let us jointly refer to these solutions as

z
Ex = cos(ω(t ∓ )),
v
where
1
v≡√
µǫ
has the dimensions of m/s (i.e., velocity) and the algebraic signs ∓
distinguish between the “travel directions” of these possible “wave solu-
tions” as elaborated later on.

• Let us next find out the magnetic field intensity H that accompanies
the x-polarized electric field wave solution
z
E = x̂ cos(ω(t ∓ )).
v
– Since the curl of E is
x̂ ŷ ẑ
∂ ∂ ∂ ∂Ex z ω
∇×E= ∂x ∂y ∂z = ŷ = ±ŷ sin(ω(t ∓ )) ,
∂z v v
Ex 0 0

5
 Faraday’s law z
∂H y
∇ × E = −µ
∂t E×H
requires that H should satisfy
∂H z ω H
−µ = ±ŷ sin(ω(t ∓ )) . z
∂t v v E = x̂f (t − )
v
Finding the time-dependent anti-derivative (and remembering v = x
√
1/ µǫ), we obtain
r
ǫ z z y
H = ±ŷ cos(ω(t ∓ )).
µ v

• The results above, namely our x-polarized non-trivial field solutions of x

Maxwell’s equations in source-free homogeneous space, can be repre-
H z
sented more compactly as E = x̂f (t + )
v
z f (t ∓ zv )
E = x̂f (t ∓ ) and H = ±ŷ , E×H
v η
where
jωt ejωt + e−jωt
f (t) ≡ cos(ωt) = Re{e }=
2
is the field waveform, r
µ
η≡
ǫ
is known as intrinsic impedance (and measured in units of ohms).

6
 • Since Maxwell’s equations with constant µ and ǫ are linear and time-
invariant (LTI), the field solutions above can be further generalized by
using their weighted and time-shifted superpositions such as
X
f (t) = An cos(ωn t + θn )
n

and ∞
1
Z
f (t) = F (ω)ejωtdω
2π −∞
having frequency dependent weighting factors An and F (ω). And since
according to Fourier analysis all practical signals f (t) can be synthe-
sized in these forms, it follows that the field solutions above are valid
with arbitrary waveforms f (t). d’Alembert
wave
Solutions solutions
z
E, H ∝ f (t ∓ )
v
of the 1D scalar wave equation with arbitrary f (t) are known as d’Alembert
wave solutions.

• d’Alembert solution
z
E, H ∝ f (t − )
v
describes electromagnetic waves traveling in +z direction, whereas so-
lution
z
E, H ∝ f (t + )
v

7
 Traveling wave in +z direction with speed v=c:

Time plots at z=0 and z>0:

describes electromagnetic waves traveling in −z direction (see margin). z

Ex (t, 0) = △(t) Ex(t, z) = △(t − )
c
In each case the travel speed is
1 −−−−−−→ 1
≡ c ≈ 3 × 108 m/s.
z
v=√ free space √ τ =1 c
t
µǫ µ o ǫo
Position plots at t=0 and t>0

• H solution can be obtained from E by dividing it with η and rotating

z
it by 90◦ so that vector E × H points in direction the waves travel. z
Ex(0, z) = △(− )
c
Ex(t, z) = △(t − )
c

• E can be obtained from H by multiplying it with η and rotating it by ct z

90◦ so that vector E × H — called Poynting vector — once again ∆z = cτ

Note: ct=300 m in 1 microsec

points in direction the waves travel. ct=300 km in 1 millisec

In each case the intrinsic impedance is

r r
µ −−−−−−→ µo Fundamental signal waveforms: REVIEW
η= free space ≡ ηo ≈ 120π ohms. u(t) u(t − to)
ǫ ǫo 1 1
Unit-step

Transformation rules above also hold for y-polarized wave solutions 0 t 0 to t

t t − to
z f (t ∓ vz ) rect( )
τ
rect(
τ
)

E = ŷf (t ∓ ) and H = ∓x̂ . 1

Rectangle
1

v η pulse

τ 0 τ t 0 τ to τ t
− −
Question: What about z-polarized waves 2 2 2 2
t t − to
z △( )
τ
△(
τ
)
E = ẑf (t ∓ ), 1
Triangle
1

v pulse

τ 0 τ t 0 τ to τ t
can they exist? −
2 2
−
2 2
t[u(t) − u(t − to)]

Answer: No, z-polarized waves ẑf (t ∓ vz ) traveling in ±z direction cannot

1
Ramp
pulse

exist because they would violate the divergence-free condition ∇·E = 0. 0 to t

8
 z
19 d’Alembert wave solutions, radiation from y
current sheets E×H

• d’Alembert wave solutions of Maxwell’s equations for homogeneous and

H
source-free regions obtained in the last lecture having the forms z
E = x̂f (t − )
z v
E, H ∝ f (t ∓ ) x
v
are classified as uniform plane-TEM waves.
z y
– TEM stands for Transverse ElectroMagnetic, and the reason for
this designation is:

viable solutions satisfying ∇ · E = ∇ · H = 0 conditions have their E x

and H vectors transverse to the direction of propagation which always H z
E = x̂f (t + )
coincides with the direction of vector S ≡ E × H known as Poynting v
vector — more on this later on.
E×H
– d’Alembert wave solutions such as
Poynting vector
z f (t − vz )
E = x̂f (t − ) and H = ŷ E×H
v η
are also designated as uniform plane waves because:

these wave-fields are constant (have the same vector value) at planes
of constant phase, e.g., on planes defined by
z
t − = const.,
v
1
 which are planes transverse to the propagation direction (direction of
vector E × H).

Not all waves solutions of Maxwell’s equations are uniform plane — for in-
stance non-uniform TEM waves with spherical surfaces of constant phase are
ubiquitous, but they will be examined later on (in ECE 450, mainly).
After the next set of examples we will examine how uniform plane waves
can be radiated by infinite planes of surface currents. By contrast, spherical
waves are produced by compact antennas having finite dimensions.

Example 1: Let
t − y/c
E = x̂△( )
τ
be a wave solution in free space where △( τt ) is a triangular waveform of duration
τ peaking at t = 0 (defined in ECE 210). We will next provide two different
solutions demonstrating how the wave field B accompanying E can be found.

Solution 1: We recognize the given wave field E as a TEM uniform plane wave travel-
ing in y-direction given the t−y/c dependence of phase. Consequently, we obtain
H by dividing E with η = ηo and rotating it by 90◦ from x̂-direction to co-align
it with E × H vector. As a result,

△( t−y/c
τ ) △( t−y/c )
H = −ẑ = −ẑ p τ .
ηo µo /ǫo
Hence,
√ t − y/c △( t−y/c
τ
)
B = µo H = −ẑ µo ǫo △( ) = −ẑ .
τ c

2
Solution 2: According to Faraday’s law,
x̂ ŷ ẑ
∂B ∂ ∂ ∂ ∂Ex
= −∇ × E = − ∂x ∂y ∂z = ẑ
∂t ∂y
Ex 0 0
t − y/c ∂ t − y/c −1 t − y/c
= ẑ△′ ( ) ( ) = ẑ △′ ( )
τ ∂y τ cτ τ
d
with the help of chain rule of differentiation, where △’(u) ≡ du △(u). Finding
the time-dependent anti derivative, we directly obtain (as before)

△( t−y/c
τ )
B = −ẑ .
c

Example 2: Consider the Lorentz force

F = q(E + v × B)
on a test charge q in the lab where E and B are the plane wave fields considered
in Example 1. Show that electrical force term qE will dominate the magnetic
force term qv × B unless the particle speed v = |v| is close to the speed of light
c (i.e., test charge is relativistic).

Solution: Since
t − y/c △( t−y/c
τ )
E = x̂△( ) and B = −ẑ ,
τ c
it follows that Lorentz force
t − y/c ẑ
F = q(E + v × B) = q△( )(x̂ − v × ).
τ c

3
 Clearly, the first term of F proportional to x̂ is dominant, unless v = |v| is close
to c.

Example 3: Consider an x̂-polarized plane TEM wave field in free space propagating
in +z direction such that
z t
E(z, t) = x̂f (t − ), with f (t) = At rect( ),
c τ
where c = 3 × 108 m/s = 300 m/µs is the speed of light in free space, τ = 1 µs,
V/m
and A = 2 µs . A plot of f (t) vs t (labelled in µs units is shown in the margin.
Determine the corresponding H(z, t) and make the following plots:

• (a) t-plots at fixed z’s: Ex (0, t) and Ex (z = 600 m, t),

• (b) z-plots at fixed t’s: Ex (z, 0) and Ex (z, 2 µs).

Solution: (a) t-plots at fixed z’s: Since z/c = 2 µs for z = 600 m, it follows that
t − 2µs V
Ex (600 m, t) = 2 (t − 2µs) rect( )
1µs m
is a shifted version of
t V
Ex (0, t) = 2 t rect(
)
1µs m
already plotted above. A graph showing both waveforms (black for z = 0 and
red for z = 600 m) is in the margin.

4
(b) z-plots at fixed t’s: In this case we wish to deptict
z 0 − z/c V
Ex (z, 0) = 2 (0 − ) rect( )
c 1µ m
and
z 2µ − z/c V
Ex (z, 2 µs) = 2 (2µ − ) rect( ) .
c 1µ m
The minus sign in front of z in the first term on the right indicates that the slopes
of the curves to be plotted are negative. Hence, we end up with the descending
ramp waveforms (black for t = 0 and red for t = 2µs) shown in the margin.

• Plane electromagnetic waves discussed above propagate in free-space in

regions of zero ρ and J (per our derivation).

– But what generates such waves?

• The answer must be, far away ρ and J variations (linked by continuity
equation) that we have not considered in our equations so far.

• We will next describe how plane TEM waves can be produced — radi-
ated — by time-varying infinite current sheets by starting from familiar
static and quasi-static solutions:

5
 z
• Consider first a static and constant surface current density
Js = x̂Jx A/m y
+
H
flowing on z = 0 surface as shown graphically in the margin. This
infinite surface current will produce a static magnetic field Js = x̂Jx
Jx x
A/m for z ≷ 0
H(z) = ∓ŷ
2
also shown in the margin as we learned in Lecture 13. H−

– Note that the fields point in opposing directions above and below
the surface current in compliance with the right hand rule and
obey the boundary condition equation for tangential H.
– Also, H is not accompanied by an electric field E since static
currents produce only static magnetic fields.

• What if the surface current Jx varies with time, i.e., Jx = Jx(t). In

that case we have quasi-statically
Jx(t)
H(z, t) ≈ ∓ŷ
A/m for z ≷ 0,
2
but only as an approximation for positions very close to z = 0 sur-
face where propagation time-delay zv of d’Alembert solutions can be
neglected1 .
1
This solution surely cannot be an exact solution since if it were, it would imply instantaneous changes
in H in response Jx at arbirarily large distances, implying an infinite speed of propagation — we know
that is not true!

6
 • But the exact field solution of Maxwell’s equtions valid for all z is z
equally easy to obtain: just replace Jx (t) above with Jx(t ∓ vz ) and E+ × H+
E+
replace ≈ with = so that y
z H+
Jx (t ∓ v)
H(z, t) = ∓ŷ A/m for z ≷ 0
2 Js = x̂Jx (t)

complies with plane TEM d’Alembert solutions2 of Maxwell’s equations x

in homogeneous and sourece free regions z ≷ 0.
H−
E−
• As always, there is an accompanying E(z, t) that is obtained by multi-
plying H(z, t) with η and replacing its unit vector so that vector E × H E− × H−

points in the direction of propagation, away from the z = 0 in this case

— hence, as illustrated in the margin,
η z
E(z, t) = −x̂ Jx(t ∓ ) V/m for z ≷ 0.
2 v

Since Maxwell’s eqn’s + boundary conditions have unique solutions in given

settings, we are assured that any solution that complies with both (as in this
case) is the solution for the given setting (surface current on z = 0, in this
case) — it was surprisingly easy to solve this radiation problem by starting
from simple static and quasi-static solutions.

2
We use Jx (t ∓ vz ) rather than Jx (t ± vz ) for z ≷ 0 because we assume that Jx (t) on z = 0 surface is the
only field source — in that case causality principle dictates that we use only the solutions propagating
away from the source (just like when a pebble drops in a pond, ripples propagate away).

7
 Conclusion: Evidently, a time varying surface current

Js = x̂f (t) on z = 0 plane

produces plane electromagnetic waves

± ηf (t ∓ vz ) ± f (t ∓ vz )
E = −x̂ and H = ∓ŷ in regions z ≷ 0
2 2
propagating away from the z = 0 plane.
Note that: z

E+ × H+
1. Ex and Hy waveforms are proportional to delayed versions of surface E+
y
current Jx(t) at each location z above and below the current sheet, with +
H
the reference directions of E and Js opposing one another.
Js = x̂Jx (t)
±
2. fields E are continuous on z = 0 surface in compliance with tangential
x
boundary condition equations.
H−
3. fields H± exhibit a discontinuity on z = 0 surface that matches the E−
current density of the same surface, once again in compliance with
E− × H−
tangential boundary condition equations.

Opposing E and Js vectors on z = 0 plane indicate that the surface is acting

as a source of radiated energy (the energy that feeds the waves radiated away
from the surface) — this interpretation will be discussed in more detail in
the next lecture.

8
Example 4: A current sheet on z = 0 surface is described by
t
Js (t) = x̂f (t), with f (t) = At rect( ),
τ
A/m
where τ = 1 µs and A = 1 µs . A plot of the current waveform f (t) is plotted in
the margin. Assuming that the current sheet is embedded in free space, construct
the following plots:

• (a) Radiated Hy (z, t = 2µs) vs z,

• (b) RadiatedEx(z, t = 2µs) vs z.
Solution: (a) From the theory developed above, we have using delayed copies of half
the surface current density,
1 z 2µ ∓ zc A
Hy (z, 2µs) = ∓ (2µ ∓ ) rect( ) for z ≷ 0,
2 c 1µ m
as plotted in the margin. Notice that the propagated field waveforms — c ×
2µs=600 m has been covered in 2 µs — are re-scaled and shifted replicas of the
source function f (t).

(b) We have, multiplying Hy with ηo = 120π Ω, and adjusting the signs so that E
and Js are pointing in opposite directions,
z 2µ ∓ zc V
Ex (z, 2µs) = −60π(2µ ∓ ) rect( ) for z ≷ 0.
c 1µ m
Plots are shown in the margin.

9
20 Poynting theorem and monochromatic waves
• The magnitude of Poynting vector

S=E×H

represents the amount of power transported — often called energy flux

— by electromagnetic fields E and H over a unit area transverse to the
E × H direction.

This interpretation of the Poynting vector is obtained from a conservation

law extracted from Maxwell’s equations (see margin) as follows:

1. Dot multiply Faraday’s law by H, dot multiply Ampere’s law by E, ∇·D = ρ

∂B ∇·B = 0
(∇ × E = − )·H ∂B
∂t ∇×E = −
∂D ∂t
(∇ × H = J + )·E ∂D
∂t ∇×H = J+ .
∂t
and take their difference:
∂D ∂B
H · ∇ × E − E · ∇ × H = − · E − · H −J · E.
| {z }
| ∂t {z ∂t }
∂ 1 1
∇ · (E × H) − ( ǫE · E + µH · H)
∂t 2 2

2. After re-arrangements shown above, the result can be written as

1
 ∂ 1 1
( ǫE · E + µH · H) + ∇ · (E × H) + J · E = 0.
∂t 2 2

• Poynting theorem derived above is a conservation law just like the

continuity equation ∂ρ
∂t
+ ∇ · J = 0: Poynting theorem

– The first term on the left,

∂ 1 1
( ǫE · E + µH · H),
∂t 2 2
is time rate of change of total electric and magnetic energy den-
sity.

Hence, Poynting theorem is the conservation law for electro-

magnetic energy, just like continuity equation is the conservation law
for electric charge.

– The second term

∇ · (E × H)
accounts for energy transport in Poynting theorem, just like ∇ · J
accounts for charge transport in the continuity equation. There-
fore
S≡E×H

2
 is energy flux per unit area measured in
VA W J/s
= 2= 2
mm m m

units, just like J is charge flux per unit area in C /s A

m2 = m2 units.
– Finally, the last term in Poynting theorem (repeated in the mar-
gin), Poynting thm:
J·E
∂ 1 1
is called Joule heating, and it represents power absorbed per ∂t 2( ǫE · E + 2 µH · H) +

unit volume (which can only be non-zero in the presence of J). ∇ · (E × H) + J · E = 0

If J·E is negative in any region, then J in that region is acting as a
source of electromagnetic energy, just like any circuit component
with negative vi is acting as an energy source in the electrical
circuit.
Note that J · E had a negative value on the current sheet radiator
examined in last lecture. We return to the current sheet radiator
in the next example.

3
 z
y
Example 1: On z = 0 plane we have a time-harmonic surface current specified as S+
+ η z
E = −x̂ f (t − )
2 v Js = x̂f (t)
A
Js = x̂f (t) = x̂2 cos(ωt)
m
where ω is some frequency of oscillation.
1 z
η z H− = ŷ f (t + )
(a) Determine the radiated TEM wave fields E(z, t) and H(z, t) in the regions z ≷ 0, E = −x̂ f (t + )
−
2 v
2 v
S−
(b) The associated Poynting vectors E × H, and

(c) Js · E on the current sheet.

Solution: (a) With reference to the solution of the current sheet radiator depicted
in the margin (from last lecture), we that an x-polarized surface current f (t)
produces the wave fields
η z 1 z
Ex = − f (t ∓ ) and Hy = ∓ f (t ∓ )
2 v 2 v
in the surrounding regions propagating away from the current sheet on both sides.
Given that f (t) = 2 cos(ωt), this implies that
Ex = −η cos(ωt ∓ βz) and Hy = ∓ cos(ωt ∓ βz)
where
ω
β= and η = ηo ≈ 120π Ω
c
since the current sheet is surrounded by vacuum. Hence in vector form we have
V A,
E(z, t) = −η cos(ωt ∓ βz)x̂ and H(z, t) = ∓ cos(ωt ∓ βz)ŷ m
m
where the upper signs are for z > 0, and lower signs for z < 0.

4
(b) The associated Poynting vectors are
W
S = E × H = ±η cos2(ωt ∓ βz)ẑ .
m2
Note that the time-average value of vector S points in the direction of wave
propagation on both sides of the current sheet.

(c) Since on z = 0 surface of the current sheet the electric field vector is
V
E(0, t) = −η cos(ωt)x̂ ,
m
it follows that Js · E on the same surface is
A V W
Js (t) · E(0, t) = (x̂2 cos(ωt) ) · (−η cos(ωt)x̂ ) = −2η cos2 (ωt) 2 .
m m m

• In the above example, a time-harmonic source current oscillating at

some frequency ω produced “monochromatic waves” of radiated fields
propagating away from the current sheet on both sides.

– The calculations showed time-varying Poynting vectors E × H.

The time-averaged values of these time-varying vectors can be eas-
ily determined by making use of the trig identity
1
cos2(ωt + φ) = [1 + cos(2ωt + 2φ)].
2
Since the time-average of the second term on the right is zero, we

5
 can express the time-average of this identity as z
y
1 1 S+
hcos2(ωt + φ)i = h [1 + cos(2ωt + 2φ)]i = , + η z
E = −x̂ f (t − )
2 2 2 v Js = x̂f (t)

where the angular brackets denote the time-averaging procedure.

1 z
• Consequently, the result η z
E = −x̂ f (t + )
− H− = ŷ f (t + )
2 v
2 v
S−
W
E × H = ±η cos2(ωt ∓ βz)ẑ 2
m
from Example 1 implies that
1 W W
hE × Hi = ±η ẑ 2 = ±60π ẑ 2 ,
2 m m
which represent the time-average power per unit area transported by
the waves radiated by the current sheet.

• In Poynting theorem the Joule heating term J · E is power absorbed

per unit volume, and, accordingly, −J · E is power injected per
unit volume.

– Likewise, ±Js·E can be interpreted as power absorbed/injected

per unit area on a surface.

In Example 1 above we calculated an instantaneous injected power

density of

6
 W
−Js · E = 2η cos2(ωt) .
m2
Clearly, its time-aveage works out as
W W
h−Js · Ei = η = 120π .
m2 m2
– Note that h−Js · Ei exactly matches the sum of |hE × Hi| calcu-
lated on both sides of the current sheet, in conformity with energy
conservation principle (Poynting theorem).

7
21 Monochromatic waves and phasor notation
• Recall that we reached the traveling-wave d’Alembert solutions
z
E, H ∝ f (t ∓ )
v cos(ωt)
Period
1 2π
via the superposition of time-shifted and amplitude-scaled versions of T =
ω

f (t) = cos(ωt),
t
namely the monochromatic waves
-1
z
A cos[ω(t ∓ )] = A cos(ωt ∓ βz), cos(βz)
Wavelength
v 1 2π
λ=
with amplitudes A where β

ω √
β≡ = ω µǫ z
v
can be called wave-number in analogy with wave-frequency ω. -1

– As depicted in the margin, monochromatic solutions A cos(ωt∓βz)

are periodic in position and time, with the wave-number β being
essentially a spatial-frequency, the spatial counterpart of ω.

This is an important point that you should try to understand

well — it has implications for signal processing courses related
to images and vision.

1
 Period
– In general, monochromatic solutions of 1D wave-equations ob- 1
cos(ωt)
2π
tained in various branches of science and engineering can all be rep- T =
ω
resented in the same format as above in terms of wave-frequency
/ wave-wavenumber pairs ω and β having a ratio t
ω
v≡ -1
β
Wavelength
cos(βz)
recognized as the wave-speed and specific dispersion relations 1 2π
λ=
such as: β

1. TEM waves in perfect dielectrics: z

√
β = ω µǫ, -1

Dispersion relations
2. Acoustic waves in monoatomic gases with temperature T (K) between
and atomic mass m (kg): wavefrequency ω
r
m and
β=ω 5 , wavenumber β
3 KT
determine the
3. TEM waves in collisionless
q plasmas (ionized gases) with plasma propagation veloc-
frequency ω = Ne :
2 ity
p mǫo
ω
1q 2 v= = λf
β= ω − ωp2 . β
c
for all types of
wave motions.
2
 – For any type of wave solution — TEM, acoustic, plasma wave
— once the dispersion relation is available (meaning that it has
been derived from fundamental physical laws governing the specific
wave type), wave propagation velocity is always obtained as
ω
v=
β
or, equivalently, as
λ
v= = λf
T
where
2π
λ≡ Wavelength
β
and
2π 1
T = ≡ Waveperiod.
ω f

propagatingWaveCos.eps

3
4
 • Monochromatic x-polarized waves
V
E = Eo cos(ωt ∓ βz) x̂
m
can also be expressed in phasor form as
V
Ẽ = Eoe∓jβz x̂
m
such that
Re{Ẽejωt} = Eo cos(ωt ∓ βz) x̂ = E
in view of Euler’s identity.

Example 1: Study the following table to understand monochromatic wave

fields and their phasors.

Field Phasor Comment

E = cos(ωt + βy) ẑ Ẽ = ejβy ẑ z-polarized wave propagating in −y direction
jβy
H̃ = − e η x̂ magnetic phasor that accompanies Ẽ above
H = sin(ωt − βz) ŷ H̃ = −je−jβz ŷ wave propagating in +z direction
Ẽ = −jηe−jβz x̂ electric field phasor of H̃ above
E = η sin(ωt − βz) x̂ which is an x-polarized field (see the right column)

5
Example 2: Given that

H = x̂H + cos(ωt − βz) + ŷH − sin(ωt + βz)

representing the sum of wave fields propagating in opposite directions, the corre-
sponding phasor
H̃ = x̂H +e−jβz − j ŷH − ejβz .
The corresponding E-field phasor is

Ẽ = −ŷηH + e−jβz + j x̂ηH − ejβz ,

from which

E = −ŷηH + cos(ωt − βz) − x̂ηH − sin(ωt + βz).

Make sure to check that all the signs make sense, and if you think you have
caught an error, let us know.

• In general, we transform between plane TEM wave phasors Ẽ and H̃

as follows:

1. To obtain H̃ from Ẽ: divide Ẽ by η and rotate the vector direction

so that vector S̃ ≡ Ẽ × H̃∗ points in the propagation direction of the
wave — more on complex vector S̃ later on.

2. To obtain Ẽ from H̃: multiply H̃ by η and rotate the vector direction

so that vector S̃ ≡ Ẽ × H̃∗ points in the propagation direction of the

6
 wave. z
y
S+
+ η z
E = −x̂ f (t − )
2 v Js = x̂f (t)

Example 3: On z = 0 plane we have a monochromatic surface current specified as

A 1 z
Js = x̂f (t) = x̂2 cos(ωt) = Re{x̂2 ejωt}. η z
E = −x̂ f (t + )
− H− = ŷ f (t + )
2 v
m 2 v
S−
± ±
Determine wave field phasors Ẽ and H̃ for plane TEM waves propagating away
from the z = 0 surface on both sides (assumed vacuum).

Solution: We know that an x-polarized surface current f (t) produces

η z 1 z
Ex = − f (t ∓ ) and Hy = ∓ f (t ∓ )
2 v 2 v
in surrounding regions. Given that f (t) = 2 cos(ωt), this implies

Ex = −η cos(ωt ∓ βz) and Hy = ∓ cos(ωt ∓ βz)

where
ω
β= and η = ηo ≈ 120π Ω
c
since the current sheet is surrounded by vacuum. Converting these into phasors,
we find
Ẽ± = −ηe∓jβz x̂ and H̃± = ∓e∓jβz ŷ.

7
 z
• In the last lecture we calculated the time-average E × H and Js · E of y
the fields examined in Example 3 using a time-domain approach. The + η z
S+
E = −x̂ f (t − )
same calculations can be carried out in terms of phasors Ẽ, H̃, and J̃s 2 v Js = x̂f (t)

as follows:
1 1 1 z
hE × Hi = Re{Ẽ × H̃∗} and hJs · Ei = Re{J̃s · Ẽ∗} η z
E = −x̂ f (t + )
− H− = ŷ f (t + )
2 v
2 2 2 v
S−
where Ẽ × H̃∗ ≡ S̃ is called complex Poynting vector. Instantaneous power

– The proof of these are analogous to the proof of p(t) = v(t)i(t)

1 with time-harmonic signals is

hp(t)i = Re{V I ∗ }
2 V ejωt + cc Iejωt + cc
v(t)i(t) = ( )( )
2 2
for the average power of a circuit component in terms of voltage and current
phasors V and I (see margin). where V and I are phasors of v(t) and
i(t) and cc indicates the conjugate of
the term to the left of + sign.
For, instance, given that This can be expanded as

V I ∗ + cc V Iej2ωt + cc
A V v(t)i(t) = + .
J̃s = 2x̂ and Ẽ±(z) = −ηe∓jβz x̂ 4 4
m m The second term has a zero time aver-
age. It follows that time-average power
in Example 3, it follows that
V I ∗ + cc 1
hv(t)i(t)i = = Re{V I ∗ }
1 W 4 2
h−Js(t) · E(0, t)i = Re{−J̃s · Ẽ∗(0)} = η ≈ 120π 2 , since
2 m
V I ∗ + cc = V I ∗ + V ∗ I = 2Re{V I ∗ }.
in conformity with the result from last lecture.
(Also see ECE 210 text.)

8
22 Phasor form of Maxwell’s equations and
damped waves in conducting media
• When the fields and the sources in Maxwell’s equations are all monochro-
matic functions of time expressed in terms of their phasors, Maxwell’s
equations can be transformed into the phasor domain. ∇·D = ρ
∇·B = 0
– In the phasor domain all ∂B
∇×E = −
∂t
∂ ∂D
→ jω ∇×H = J+ .
∂t ∂t
and all variables D, ρ, etc. are replaced by their phasors D̃, ρ̃, and
so on. With those changes Maxwell’s equations take the form shown
in the margin.
– Also in these equations it is implied that
∇ · D̃ = ρ̃
D̃ = ϵẼ ∇ · B̃ = 0
B̃ = µH̃ ∇ × Ẽ = −jω B̃
J̃ = σ Ẽ ∇ × H̃ = J̃ + jω D̃

where ϵ, µ, and σ could be a function of frequency ω (as, strictly

speaking, they all are as seen in Lecture 11).
– We can derive from the phasor form Maxwell’s equations shown in
the margin the TEM wave properties obtained earlier on using the
time-domain equations by assuming ρ̃ = J̃ = 0.

1
 We will do that, and and after that relax the requirement J̃ = 0 with
J̃ = σ Ẽ to examine how TEM waves propagate in conducting media.

• With ρ̃ = J̃ = 0 the phasor form Maxwell’s equation take their simplified

forms shown in the margin.

– Using ∇ · Ẽ = 0
∇ · H̃ = 0
∇ × [∇ × Ẽ = −jωµH̃] ⇒ −∇2Ẽ = −jωµ∇ × H̃
∇ × Ẽ = −jωµH̃
which combines with the Ampere’s law to produce ∇ × H̃ = jωϵẼ

∇2Ẽ + ω 2µϵẼ = 0.

– For x-polarized waves with phasors

Ẽ = x̂Ẽx(z),

the phasor wave equation above simplifies as

∂2 2
Ẽ x + ω µϵẼx = 0.
∂z 2
– Try solutions of the form

Ẽx(z) = e−γz or eγz

where γ is to be determined.

2
 – Upon substitution into wave equation both of these lead to

(γ 2 + ω 2µϵ)Ẽx = 0,

which yields

γ 2 + ω 2 µϵ = 0 ⇒ γ 2 = −ω 2µϵ

from which one possibility is

√
γ = jβ, with β ≡ ω µϵ.

Thus viable phasor solutions are

Ẽx (z) = e∓jβz

as we already knew.
– Furthermore, using the phasor form Faraday’s law it is easy to show
that
e∓jβz
!
µ
H̃y = ± with η = .
η ϵ

Note that we have recovered above the familiar properties

of plane TEM waves using phasor methods.
Next, the phasor method carries us to a new domain that cannot be easily
examined using time-domain methods.

3
• With ρ̃ = 0 but J̃ = σ Ẽ ̸= 0, implying non-zero conductivity σ, the
pertinent phasor form equations are as shown in the margin.

– This is the same set as before, except that ∇ · Ẽ = 0

∇ · H̃ = 0
jωϵ has been replaced by σ + jωϵ.
∇ × Ẽ = −jωµH̃
Thus, we can make use of phasor wave solutions above after applying ∇ × H̃ = σ Ẽ + jωϵẼ
the following modifications to γ and η: = (σ + jωϵ)Ẽ

1.

2 2 ⇒⇒ "
γ = −ω µϵ = (jωµ)(jωϵ) γ = (jωµ)(σ + jωϵ)
σ ̸= 0

2. ! # #
µ jωµ ⇒⇒ jωµ
η= = η= .
ϵ jωϵ σ ̸= 0 σ + jωϵ
Note that the modified γ and η satisfy
γ γη
γη = jωµ and = σ + jωϵ µ =
η jω
γ
leading to useful relations shown in the margin (assuming real valued σ = Re{ }
η
σ and ϵ). 1 γ
ϵ = Im{ }
ω η

4
 (a) Damped wave snapshot at t=0
• In terms of γ and η above, we can express an x-polarized plane wave together with exponential envelope

propagating in z direction in terms of phasors e−αz

Eo ∓γz
Ẽ = x̂Eoe∓γz and H̃ = ±ŷ e z
η
e−αz cos(ωt − βz)|t=0
where Eo is an arbitrary complex constant (complex wave amplitude).

• In expanded forms γ and η appear as: (b) Snaphot at t>0, with t=0 waveform
for comparison
"
γ = (jωµ)(σ + jωϵ) ≡ α+jβ, so that α = Re{γ} and β = Im{γ},

and z
# # #
e−αz cos(ωt − βz)
jωµ jωµ jωµ
η= ≡ |η|ejτ so that |η| = | | and τ = ∠ .
σ + jωϵ σ + jωϵ σ + jωϵ
– β appears within cosine
argument and deter-
1. In the special case of a perfect dielectric with σ = 0, we find mines the wavelength
! 2π
√ µ λ=
β
γ = jω µϵ ≡ jβ and η = ,
ϵ and propagation speed

and, therefore, ω
vp = .
β

∓jβz ŷEoe∓jβz – α controls wave attenu-

Ẽ = x̂Eoe and H̃ = ± ation by
η
e∓αz
as before. In this case α = τ = 0.
factor in propagation
direction.

5
2. Another case of imperfect dielectric (or “lousy” conductor) occurs
when σ is not zero, but it is so small that are justified in using

(1 ± a)p ≈ 1 ± pa, if |a| ≪ 1,

1 σ
with p = 2
as follows: For ωϵ
≪ 1,
!
" √ σ √ σ σ µ √
γ = (jωµ)(σ + jωϵ) = jω µϵ(1−j )1/2 ≈ jω µϵ(1−j )= +jω µϵ;
ωϵ 2ωϵ 2 ϵ
hence
!
σ µ √
Ẽ ≈ x̂Eoe∓(α+jβ)z with α = and β = ω µϵ;
2 ϵ
also in the same case
ŷEoe∓(α+jβ)z
! !
µ µ σ µ j tan−1 σ
!
H̃ ≈ ± with η = σ ≈ (1+j )≈ e 2ωϵ ,
η ϵ(1 − j ωϵ ) ϵ 2ωϵ ϵ
such that !
µ σ
|η| ≈
and τ = ∠η ≈ .
ϵ 2ωϵ
Note: γ and η both are complex valued, the consequences of which will
be examined later on.
σ
3. A third case of good conductor corresponds to ωϵ ≫ 1. In that case,
! ! ! ! !
σ σ ωµσ µ jωµ ωµ jπ/4
!
γ = jω µϵ(1 − j ) ≈ ω jµ = (1+j) and η ≈ = = e .
ωϵ ω 2 −j ωσ σ σ
6
 (a) Damped wave snapshot at t=0
Hence, together with exponential envelope

! ! e−αz
ωµσ " ωµ
α≈β≈ = πf µσ while |η| = and τ = ∠η = 45o.
2 σ z

4. Finally, perfect conductor case corresponds to σ → ∞, in which e−αz cos(ωt − βz)|t=0

case Ẽx → 0 as we will show later on. Wave fields cannot exist in perfect
conductors. (b) Snaphot at t>0, with t=0 waveform
for comparison

• Summarizing, in a homogeneous medium with arbitrary but con-

stant µ, ϵ, and σ, time-harmonic plane TEM waves are in terms of
z
E = x̂Re{Eoe∓(α+jβ)z ejωt} = x̂|Eo|e∓αz cos(ωt ∓ βz + ∠Eo)
e−αz cos(ωt − βz)

and accompanying magnetic fields

Eo |Eo| ∓αz • β appears within cosine
H = ±ŷRe{ e∓(α+jβ)z ejωt} = ±ŷ e cos(ωt ∓ βz + ∠Eo − ∠η). argument and deter-
η |η| mines the wavelength
2π
λ=
• Propagation velocity β
ω ω and propagation speed
vp = = " ,
β Im{ (jωµ)(σ + jωϵ)} vp =
ω
.
β
now depends on frequency ω and it describes the speed of the nodes • α controls wave attenu-
(zero-crossings, not modified by the attenuation factor) of the field wave- ation by
form. Subscript p is introduced to distinguish vp — also called phase e∓αz
velocity — from group velocity vg discussed in ECE 450 (velocity of factor in propagation
narrowband wave packets). direction.

7
 (a) Damped wave snapshot at t=0
• Wavelength together with exponential envelope
2π vp e−αz
λ= =
β f
now depends on frequency f via both the numerator and the denomi- z

nator, and measures twice the distance between successive nodes of the
e−αz cos(ωt − βz)|t=0
waveform.
• Penetration depth (also called skin depth if very small) (b) Snaphot at t>0, with t=0 waveform
for comparison
1 1
δ≡ = "
α Re{ (jωµ)(σ + jωϵ)}
is the distance for the field strength to be reduced by e−1 factor in its z

direction of propagation. e−αz cos(ωt − βz)

– For a fixed σ, and a sufficiently large ω, the penetration depth

– β appears within cosine
2 argument and deter-
δ ≈ " µ Imperfect dielectric formula mines the wavelength
σ ϵ
2π
which can be very small if σ is large — with small δ the wave is λ=
β
severely attenuated as it propagates. and propagation speed
– For a fixed σ, and a sufficiently small ω, vp =
ω
.
! β
2 1
δ≈ =√ Good conductor ”skin depth” formula – α controls wave attenu-
µωσ πf µσ ation by
which, although small with large σ, increases as ω decreases, making e∓αz
low frequencies to be preferable in applications requiring propagat-
factor in propagation
ing through lossy media with large σ, such as in sea-water. direction.

8
23 Imperfect dielectrics, good conductors
2π 1
Condition β α |η| τ λ= β
δ= α
Perfect √ pµ 2π
σ=0 ω ǫµ 0 ǫ 0 √
ω ǫµ ∞
dielectric
Imperfect σ √ pµ pµ q
ωǫ
≪1 ∼ ω ǫµ β 12 ωǫ
σ
= σ
2 ǫ
∼ ǫ
∼ σ
2ωǫ
∼ 2π
√
ω ǫµ
2
σ
ǫ
µ
dielectric
Good σ
√ √ p ωµ
≫1 ∼ πf µσ ∼ πf µσ 45◦ ∼ √ 2π ∼ √ 1
conductor ωǫ σ πf µσ πf µσ

Perfect
σ=∞ ∞ ∞ 0 - 0 0
conductor
• The table above summarizes TEM wave parameters in homogeneous
conducting media where the propagation velocity x-polarized phasor
ω
vp = Ẽ = x̂Eoe∓αz e∓jβz
β
(note that it can be frequency dependent) and field phasors can be accompanied by
expressed in formats similar to that shown in the margin, keeping in
mind that propagation direction coincides with vector H̃ = ±ŷ Eηo e∓αz e∓jβz .
S̃ ≡ Ẽ × H̃∗

such that
1
hSi = hE × Hi = Re{S̃}
2
is the average energy flux per unit area (time-average Poynting vector).

1
Example 1: Consider the plane TEM wave
V
Ẽ = ŷ2e−αz e−jβz
,
m
in an imperfect dielectric. Determine H̃ and time-average Poynting vector
hSi. Compute hSi at z = 0 and z = 10 m, if ǫ = 4ǫo , µ = µo , σ = 10−3 S/m,
and ω = 2π · 109 rad/s

Solution: Using right hand rule, so that E × H points in propagation direction ẑ, we
find that
2 2 A
H̃ = −x̂ e−αz e−jβz ≈ −x̂ p e−αz e−jβz e−jτ
η µ/ǫ m
pµ
using |η| = ǫ from the table above for a perfect dielectric.

The avg. Poynting vector is

1 1 2
hSi = Re{Ẽ × H̃∗ } = Re{ŷ2e−αz e−jβz × (−x̂ p e−αz e−jβz e−jτ )∗}
2 2 µ/ǫ
1 2 2
= − Re{ŷ2e−αz × x̂ p e−αz ejτ } = ẑ p e−2αz cos τ.
2 µ/ǫ µ/ǫ
With the given parameters,
σ 10−3 · 36π × 109 9 −3
= = 10 ≪ 1,
ωǫ 2π · 109 · 4 2
σ 9
τ ≈ ≈ 10−3 rad
2ωǫ r
r 4
µ µo ηo
|η| ≈ = = = 60π Ω
ǫ 4ǫo 2
r
σ µ 1 −3 1
α ≈ = 10 60π = 30π · 10−3 .
2 ǫ 2 m

2
 Hence, at z = 0,
2 2 1 W
hSi = ẑ p cos τ ≈ ẑ = ẑ ,
µ/ǫ 60π 30π m2

whereas, at z = 10 m,
2 2 −6π/10 0.15 W
e−2·30π·10 ·10 cos τ ≈ ẑ
−3
hSi = ẑ p e ≈ ẑ .
µ/ǫ 60π 30π m2

• Note that in above example power transmitted per unit area has dropped
to 15% of its value upon propapagating over a relatively short distance
of 10 m.

– In the physical terms, the lost power of the wave is gained by the
propagation medium in the form of heat — average Joule heating ⇐This is what we
hJ · Ei in the medium will be positive and account for the loss of want to happen in a
the wave power (as seen in a HW problem). microwave oven.

From a communications perspective, this rapid attenuation is problematic

since it is evident that the signal energy is being wasted as heat in the
medium rather than being transmitted efficiently to distant communication
targets.
As the next example shows, we are better off using lower frequencies in
under-water communcations.

3
Example 2: Repeat Example 1 for ω = 2π · 103 rad/s and σ = 4 S/m (sea water) in
which case the propagation medium becomes a good conductor.

Solution: Using right hand rule, so that E × H points in propagation direction ẑ, we
have
2 2 A
H̃ = −x̂ e−αz e−jβz ≈ −x̂ e−αz e−jβz e−jτ
η |η| m
as well as
1 1 2
hSi = Re{Ẽ × H̃∗ } = Re{ŷ2e−αz e−jβz × (−x̂ e−αz e−jβz e−jτ )∗}
2 2 |η|
1 2 2
= − Re{ŷ2e−αz × x̂ e−αz ejτ } = ẑ e−2αz cos τ.
2 |η| |η|

With the given parameters,

σ 4 · 36π × 109
= 3
= 18 · 106 ≫ 1,
ωǫ 2π · 10 · 4
which confirms that the medium behaves as a good conductor at this small ω,
and using the appropriate formulae from the table,
π
τ ≈ rad
4 r √
r
ωµ 2π · 103 · 4π · 10−7 p π 2
|η| ≈ = = π 2 × 10−4 ≈ Ω
σ 4 100
p √ √ π 1
α ≈ πf µσ = π · 103 · 4π · 10−7 · 4 = 42π 2 10−4 = .
25 m
Hence, at z = 0,
2 200 π 100 W
hSi = ẑ cos τ ≈ ẑ √ cos = ẑ ,
|η| π 2 4 π m2

4
 whereas, at z = 10 m,
100 −2· π ·10 100 W
hSi = ẑ e 25 ≈ ẑ 0.081 2 .
π π m

• As Example 2 illustrates, at a frequency of ω = 2π · 103 rad/s or f = 1

kHz, wave power is reduced to about 8% over a 10 m distance in sea
water. Less reduction in power is possible over the same distance if at
√
a smaller frequency f since α ∝ f .

– The disadvantage of being forced to use smaler frequencies is of

course having a smaller available bandwidth at small frequencies.
Thus communication with submarines at great depths will only be
possible at very slow rates.

The next example identifies the penetration depth in sea water at 1 kHz.

5
Example 3: What is the penetration depth δ = α−1 in a medium with σ = 4 S/m,
ǫ = 81ǫo , and µ = µo for ω = 2π · 103 rad/s.

Solution: With the given parameters we have

σ 4 · 36π × 109 72 × 109
= 3
= 3
≈ 106 ≫ 1,
ωǫ 2π · 10 · 81 81 × 10
i.e., good conductor situation. Hence the penetration depth is
1 1 1 100 25
δ≈√ =√ =√ = = ≈ 7.95 m.
πf µσ π103 · 4π · 10−74 42π 2 · 10−4 4π π

6
24 Signal transmission, circular polarization
Since in perfect dielectrics the propagation velocity vp = v and the intrinsic
impedance η are frequency independent (i.e., propagation is non-dispersive),
d’Alembert plane wave solutions of the form
z f (t − zv )
E = x̂f (t − ) and H = ŷ
v η
are valid in such media.
6
m(t)
• Consider a waveform 4

f (t) = m(t) cos(ωt), -4 -2 2 4 6

-2 t
where -4

-6

2π
– ω is some specific frequency having a corresponding period T = ω, 6
m(t) cos(ωt)
– m(t) is some arbitrary signal (e.g., a voice signal, a message) 4

changing slowly compared to period T . 2

-4 -2 2 4 6
-2
In that case, -4
t
-6
– f (t) specified above can be called narrowband AM, and
– ω the carrier frequency of modulating cosine of the message
signal m(t).

1
 The corresponding x-polarized wave fields propagating in z direction
can then be represented as Field 1
z m(t − vz )
E = m(t − ) cos(ωt − βz)x̂ and H = cos(ωt − βz)ŷ
v η
√
where β = ω µǫ as usual1.

• With reference to the expressions above, we could say that the AM

wave field has an x-polarized carrier.

• By contrast, Field 2
z
E = m(t − ) cos(ωt − βz)ŷ
v
represents an AM wave field with a y-polarized carrier, and so does Field 3
z
E = m(t − ) sin(ωt − βz)ŷ
v
but with a carrier that has been time-delayed by a quarter period.

• Suppose Fields 1 and 3 above were transmitted simultaneously and

therefore superpose. In that case we will have a wave field with Circular
z polarized
E = m(t − )[cos(ωt − βz)x̂ + sin(ωt − βz)ŷ] carrier
v
1
In dispersive media where β is a non-linear function of ω, narrowband AM can propagate as
z ∂ω
m(t − ) cos(ωt − βz)x̂ where vg ≡
vg ∂β
is known as group velocity — covered in detail in ECE 450.

2
 CIRCULAR POLARIZATION:
which has a circular polarized carrier. Since this is just a superpo- Field vector rotates instead
sition of two d’Alembert waves, the accompanying H is easily found to of oscillating.

The rotation frequency is also

be the wave frequency.

z
H = m(t − )[cos(ωt − βz)ŷ − sin(ωt − βz)x̂]/η. cos(ωt − βz)x̂ + sin(ωt − βz)ŷ
v z
y

– Circular-polarized AM wave fields just introduced are in some t>0

practical applications better to use than the linear-polarized waves E x

t=0
because of, say, the peculiarities of a propagation medium (e.g,
Earth’s ionosphere or the interplanetary medium).
RIGHT CIRCULAR
– Since a circular-polarized wave field is a linear combination of
linear-polarized waves, it has a phasor that is a linear combination
of phasors of its linear components, as in Right-circular

cos(ωt−βz)x̂+sin(ωt−βz)ŷ ⇔ e−jβz x̂−je−jβz ŷ = (x̂−j ŷ)e−jβz

or Left-circular

cos(ωt−βz)x̂−sin(ωt−βz)ŷ ⇔ e−jβz x̂+je−jβz ŷ = (x̂+j ŷ)e−jβz . cos(ωt − βz)x̂ − sin(ωt − βz)ŷ

y
z

• In the last step above, we have introduced two flavors of circularly

polarized waves, which correspond to fields vectors rotating in opposite t=0
E x
directions at any position in space when viewed toward the direction the
t>0
wave propagates— clockwise for right-circular, counter-clockwise
for left circular. LEFT CIRCULAR

When left-hand thumb is pointed

along propagation direction z
the fingers curl in the rotation
direction of the field vector.

3
 x̂ − j ŷ
• Also, y
z

– for the right-circular wave propagating in z direction, the field t>0

vector simplified at z = 0 as
t=0 E x

cos(ωt)x̂ + sin(ωt)ŷ ⇔ x̂ − j ŷ
RIGHT CIRCULAR
rotates in the direction that your right-hand fingers curl when x-comp leads y-comp because of -j

the thumb is directed in propagation direction z, whereas

x̂ + j ŷ
y
– for the left-circular wave propagating in z direction, likewise, z

vector
cos(ωt)x̂ − sin(ωt)ŷ ⇔ x̂ + j ŷ t=0
E x
rotates in the direction that your left-hand fingers curl when the t>0
thumb is directed in propagation direction z.
LEFT CIRCULAR

x-comp lags y-comp because of +j

In general, the “handednes” or “helicity” of a circular polarized
wave is always obtained by matching your right or left hand to
the specified propagation and rotation directions — see example
below.

Furthermore, the rotation direction is most easily seen if the

wave is expressed in phasor form by seeing which component leads
(or lags) which. Here is an explanation by example:

4
 Example 1: A circular polarized wave field vector is given as

Ẽ = (ẑ + j ŷ)ejβx .

Determine the propagation and rotation directions of the field vector as well as
its helicity.

Solution: The propagation direction is −x since the exponent in ejβx lacks a minus
sign.

At x = 0, the wave field vector rotates as

E = Re{(ẑ + j ŷ)ejωt} = ẑ cos(ωt) − ŷ sin(ωt), ẑ + j ŷ

z y
t=0
of which the y-component leads the z-component by 90◦ of phase, or, equivalently,
E
by a quarter period in time — therefore, the vector points in y-direction before it
t>0
points in z-direction (or in z-direction before it points in −y-direction), rotating
x
from y- toward z-axis.

When I direct my right thumb in −x direction, my fingers curl from z- toward y-axis,
LEFT CIRCULAR
which is curling in the wrong direction. Hence this wave is not right-circular! It
(i) z-comp lags y-comp because of +j.
is left-circular.
(ii)E-vector rotates from y towards z.

(iii) since the wave propagates in -x

direction it is LEFT CIRCULAR

Given any propagation direction, a carrier field of an arbitrary polar-

ization can always be expressed as weighted superpositions of any pair of
orthogonal polarized carrier fields — such orthogonal pairs are considered
to be complete sets of basis functions for expressing waves with arbitrary

5
polarizations.
• EXAMPLE: Right- and left circular waves propagating in z directions
are weighted superpositions of orthogonal x- and y-polarized fields
as in (expressed in terms of phasors): basis functions Circulars
in terms of
x̂e−jβz and ŷe−jβz
linears
superpose to form right- and left-circular waves
(x̂ − j ŷ)e−jβz and (x̂ + j ŷ)e−jβz
using the weights
1, −j and 1, j
respectively.

• EXAMPLE: x- and y-polarized waves propagating in z directions

are weighted superpositions of orthogonal right- and left-circular
fields as in (expressed in terms of phasors): basis functions Linears
in terms of
(x̂ − j ŷ)e−jβz and (x̂ + j ŷ)e−jβz
circulars
superpose to form linear polarized waves
x̂e−jβz and ŷe−jβz
using the weights
1 1 1 1
, and − ,
2 2 2j 2j
respectively.

6
 • It can be argued that right- and left-circular wave pair forms an in-
trinsically more fundamental set of basis functions than, say, x̂- and
ŷ-polarized waves, because while the selection of which direction is x
and which direction is y can be arbitrary, there is no arbitrariness in
how helicity is assigned to circular polarized modes propagating in a
given direction2 .

• Also, oscillating charges will radiate linear-polarized fields, whereas ro-

tating charges will radiate circular-polarized fields (in the direction nor-
mal to the rotation plane) — so, source dynamics selects the radiated
wave polarization.

• Wave polarization is important because

– it depends on physical geometry and dynamics of the wave source,

– it may depend on the physical properties of the region the wave
propagates through,
– it will determine the direction of Lorentz force on any test charge
or electrical load,
– angular momentum carried by the wave depends on polarization,
etc.
2
Furthermore RCP and LCP plane waves consist of photons with spin angular momenta of +~ and
−~, respectively, corresponding to the eigenvalues of the quantum mechanical spin operator, while the
photons constituting LP waves will be in a “superposition state” of the eigenvectors of the spin operator
having the eigenvalues ±~ — upon spin measurements the photons of a LP wave will furnish one of +~
and −~ with equal (50%) probabilities, unlike the RCP and LCP wave photons furnishing +~ and −~,
respectively, with 100% probabilities.

7
 Note that this figure
only shows one linear
component of the sur-
Example 2: On z = 0 plane we have a time-varying surface current density
face current on
A z = 0 plane. One linear
Js (t) = m(t)[cos(ωt)x̂ + sin(ωt)ŷ]
m
component causes a lin-
with a carrier frequency of ω. Determine the radiated wave fields E± and the ear polarized radiation.
polarization (and the helicity if appropriate) of the carrier.
An orthogonal pair of
Solution: We have already learned that a surface current Js (t) on z = 0 plane will linear components will
produce TEM wave fields
η z conspire to radiate a cir-
E± = − Js (t ∓ )
2 v cular polarized wave as
in surrounding regions. With the given Js (t) , this implies in Example 2 when they
η z V are 90◦ out of phase.
E± = − m(t ∓ )[cos(ωt ∓ βz)x̂ + sin(ωt ∓ βz)ŷ] , z
2 v m
y
which has a circular-polarized carrier
S+
+ η z
cos(ωt ∓ βz)x̂ + sin(ωt ∓ βz)ŷ E = −x̂ f (t − )
2 v Js = x̂f (t)

that varies, on z = 0 plane, as

cos(ωt)x̂ + sin(ωt)ŷ. 1 z
η z H− = ŷ f (t + )
E = −x̂ f (t + )
−
2 v
This vector rotates from x- toward y-axis, and therefore the carrier of E+ is 2 v
S−
right-circular and the carrier of E− is left-circular.

8
 Example 3: In Example 2, what is the average power density of the circular polarized
carrier signal
V
Ec = cos(ωt − βz)x̂ + sin(ωt − βz)ŷ
m
in the region z > 0, assumed to be vacuum?

Solution: In phasor notation Ec and is given as

V
Ẽc = (x̂ − j ŷ)e−jβz .
m
The corresponding Hc phasor is
1 V
H̃c = (ŷ + j x̂)e−jβz .
ηo m
Therefore, the average power density is found to be
1 1 1 1
Re{Ẽc × H̃∗c } = Re{(x̂ − j ŷ) × (ŷ + j x̂)∗ } = (ẑ + ẑ) = ẑ.
2 2ηo 2ηo ηo
This is twice the power content of a linearly polarized wave field of an equal
amplitude!

Make sure you check and follow all the sign changes that take place in
Example 3.

9
25 Wave reflection and transmission
In this lecture we will examine the phenomenon of plane-wave reflections at
an interface separating two homogeneous regions where Maxwell’s equations
allow for traveling TEM wave solutions. The solutions will also need to
satisfy the boundary condition equations repeated in the margin. We will
consider a propagation scenario in which (see margin): n̂ · (D+ − D−) = ρs
n̂ · (B+ − B−) = 0
1. Region 1 where z < 0 is occupied by a perfect dielectric with medium n̂ × (E+ − E−) = 0
parameters µ1, ǫ1, and σ1 = 0,
n̂ × (H+ − H− ) = Js
2. Region 2 where z > 0 is homogeneous with medium parameters µ2, ǫ2,
y
and σ2,
Hi Ht

3. Interface z = 0 contains no surface charge or current except possibly Ei Et

in σ2 → ∞ limit which will be considered separately at the end. Region 1

x Region 2

z
• In Region 1 we envision an incident plane-wave with linear-polarized field phasors
Er
Eo −jβ1 z Hr
Ẽi = x̂Eoe−jβ1 z and H̃i = ŷ e ,
η1
where

– Eo is the wave amplitude due to far away source located in z → −∞ region,

q
µ1 √
– η1 = ǫ1 and β1 = ω µ1 ǫ1 .

1
Fields above satisfy Maxwell’s equations in Region 1, but if there were no
other fields in Regions 1 and 2 boundary condition equations requiring
continuous tangential E and H at the z = 0 interface would be violated.
In order to comply with the boundary condition equations we postulate
a set of reflected and transmitted wave fields in Regions 1 and 2 as follows: Incident:
• In Region 1 we postulate a reflected plane-wave with linear-polarized field phasors
jβ1 z ΓEo jβ1 z Ẽi = x̂Eoe−jβ1 z ,
Ẽr = x̂ΓEoe and H̃r = −ŷ e
η1 Eo −jβ1 z
including an unknown Γ that we will refer to as reflection coefficient.
H̃i = ŷ e ,
η1
– Note that the reflected wave propagates in −z direction (direction of H̃r and
the exponential terms have been adjusted accordingly). Reflected:

• In Region 2 we postulate a transmitted plane-wave with linear-polarized field

Ẽr = x̂ΓEoejβ1 z ,
phasors
τ Eo −γ2 z ΓEo jβ1z
Ẽt = x̂τ Eoe−γ2 z and H̃t = ŷ e H̃r = −ŷ e ,
η2 η1
including an unknown τ that we will refer to as transmission coefficient.
Transmitted:
– Note that the transmitted wave propagates in z direction, and
– since Region 2 is conducting we have
Ẽt = x̂τ Eoe−γ2 z ,
s
jωµ2
η2 =
σ2 + jωǫ2
τ Eo −γ2 z
and H̃t = ŷ e .
p η2
γ2 = (jωµ2 )(σ2 + jωǫ2 ) = α2 + jβ2 .

2
 • To determine the unknowns Γ and τ we enforce the following boundary
conditions at z = 0 where the fields simplify as shown in the margin: Incident at z = 0:

1. Tangential Ẽ continuous at z = 0: This requires Ẽix + Ẽrx = Ẽtx ,

Eo
leading to Ẽi = x̂Eo, H̃i = ŷ ,
η1
(1 + Γ)Eo = τ Eo ⇒ 1+Γ=τ

2. Tangential H̃ continuous at z = 0: This requires H̃iy + H̃ry = H̃ty , Reflected at z = 0:

leading to
Eo Eo
(1 − Γ) =τ ⇒ 1-Γ= ηη12 τ ΓEo
η1 η2 Ẽr = x̂ΓEo, H̃r = −ŷ
η1
Replacing τ by 1 + Γ in the second equation, we can solve for the
reflection coefficient as Transmitted at z = 0:
η2 −η1
Γ= η2 +η1
and substituting this in turn in the first equation we can solve for the τ Eo
Ẽt = x̂τ Eo, H̃t = ŷ
transmission coefficient as η2
2η2
τ = η2 +η1

The results are summarized in the margin on the next page.

3
 Special cases: Reflection coeff.:
η2 − η1
1. Region 2 is a perfect conductor with σ2 → ∞: In that case Γ= ,
η2 + η1
η2 → 0, and consequently

Γ = −1 and τ = 0. Transmission coeff.:

2η2
Incident wave cannot penetrate the perfect conductor, and it reflects τ= .
η2 + η1
totally back into Region 1 — we will study this idealized limiting case
more carefully later on. Memorize the Γ for-
Practical application of total reflection: mirrors mula, and memorize
τ as “one plus Γ”.
2. Region 2 is the same as Region 1: In that case η2 = η1, and
consequently Above,
Γ = 0 and τ = 1. r
µ1
η1 =
This is the matched impedance case when no reflection takes place ǫ1
and the incident wave is transmitted in its entirety.
and
3. Region 2 is lossless, i.e., σ2 = 0: Unless η2 = η1 there will be s
reflected as well as transmitted waves. jωµ2
η2 = .
σ2 + jωǫ2
Partial reflections can be reduced by applying a “anti-glare” coat-
ing1 on the surface, a practice known as “impedance matching”.
1 √
This is a λ/4 thick layer of a material having a characteristic impedance given by η1 η2 — the reason
for why this “quarter-wave matching” works will be discussed when we study transmission lines later on.

4
 y

Hi Ht
Ei Et
Example 1: An plane-wave in vacuum,
x Region 2
√ Region 1
Ẽi = x̂ 120πe−jβ1 z m
V
,
z
9
is incident at z = 0 on a dielectric medium with µ = µo and ǫ = ǫ.
4 o
Determine Er
the average Poynting vectors hSi i, hSr i, and hSt i of the incident, reflected, and Hr
transmitted fields.

Solution: The intrinsic impedance of the second medium occupying z > 0 is

µo 2
r
η2 = 9 = ηo .
4 ǫo
3
Therefore, the reflection coefficient is
2 2
η2 − η1 3 ηo − ηo 3 −1 2−3 1
Γ= = 2 = 2 = =−
η2 + η1 3 η o + η o 3 +1 2+3 5
and the transmission coefficient is
1 4
τ =1+Γ=1− = .
5 5

The reflected wave therefore has the field phasors

1 √ 1 √
Ẽr = − x̂ 120πejβ1 z and H̃r = ŷ 120πejβ1 z
5 5ηo
and
1 1 1 120π 1 1 W
hSr i = Re{Ẽr × H̃∗r } = −ẑ ( )2 ≈ −ẑ ( )2 2 .
2 2 5 ηo 2 5 m

5
 The transmitted wave, likewise, has the field phasors
4 √ 4 √
Ẽt = x̂ 120πe−jβ2 z and H̃t = 2 ŷ 120πe−jβ2z
5 5 3 ηo
and
1 1 4 3 120π 1 4 3 W
hSt i = Re{Ẽt × H̃∗t } = ẑ ( )2 ≈ ẑ ( )2 .
2 2 5 2 ηo 2 5 2 m2
As for the incident wave
√ 1 √
Ẽi = x̂ 120πe−jβ1 z and H̃i = ŷ 120πe−jβ1 z
ηo
and
1 1 120π 1 W
hSi i = Re{Ẽi × H̃∗i } = ẑ ≈ ẑ .
2 2 ηo 2 m2
Note: We have
1 1 16 3 1 1 24 1
|hSr i| + |hSt i| = ( + ) = ( + ) = = |hSi i|
2 25 25 2 2 25 25 2
in compliance with energy conservation (as expected) — energy flux per unit
area of the transmitted and reflected waves add up the that of the
incident wave!

6
26 Standing waves, radiation pressure y

Hi Ht = 0
We continue in this lecture with our studies of wave reflection and transmis-
Ei Et = 0
sion at a plane boundary between two homogeneous media.
Region 1
x Region 2
• In case of total reflection from a perfectly conducting mirror placed at
z
z = 0 surface, Γ = −1, and the incident and reflected waves in z < 0
Er
region combine to produce standing waves of electric and magnetic Hr
field:

– Incident wave (a traveling wave going in z-direction):

Eo −jβ1 z
Ẽi = x̂Eoe−jβ1 z and H̃i = ŷ e ,
η1

– Reflected wave (a traveling wave going in −z-direction):

Eo jβ1z
Ẽr = −x̂Eoejβ1 z and H̃r = ŷ e ,
η1
– Standing wave:
Eo −jβ1 z jβ1 z
Ẽ = Ẽi+Ẽr = x̂Eo(e−jβ1 z −ejβ1z ) and H̃ = H̃i+H̃r = ŷ (e +e )
η1
which simplify as
2Eo
Ẽ = −j x̂2Eo sin(β1z) and H̃ = ŷ cos(β1z). Standing
η1
waves
1
 These are called standing wave phasors because when we go Ex(z, t) ∝ sin(βz) sin(ωt)

to the time-domain (by multiplying with ejωt and taking the real
time as usual) we obtain:
2Eo z
E(z, t) = x̂2Eo sin(β1z) sin(ωt) and H(z, t) = ŷ cos(β1z) cos(ωt);
η1

these, unlike d’Alembert solutions of the format f (t ∓ zv ), describe os- λ=

2π
cillations in time t, with different amplitudes at different positions z β
Hy (z, t) ∝ cos(βz) cos(ωt)
(see margin and the animation linked in class calendar).

– Standing waves carry no net energy, that is, with standing wave
fields we have
z
hE × Hi = 0,
because of the cancellation of the power transported by its travel-
ing wave components in opposite directions. λ
Verification: Using the phasors 2
Note: Nulls in Ex and Hy are
2Eo separated by half wavelength.
Ẽ = −j x̂2Eo sin(β1z) and H̃ = ŷ cos(β1 z),
η1 Adjacent nulls of Ex and Hy
are separated by quarter
we have wavelength.

1 1 2Eo It is useful to think of nulls

hE × Hi = Re{Ẽ × H̃∗ } = Re{−j x̂2Eo sin(β1z) × ŷ cos(β1z)} of Ex as "shorts" in analogy
2 2 η1 to shorts in circuits where v=0.
2Eo2 Conductor shorts Ex on its
= ẑ sin(β1 z) cos(β1z)Re{−j} = 0. surface where a current flows.
η1
Also useful to think of nulls
of Hy as "opens" in analogy
to opens in circuits where i=0.

2
 – Note that E(0, t) = 0 on z = 0 surface satisfying the tangential Ex(z, t) ∝ sin(βz) sin(ωt)

electric boundary condition as expected (recall that the fields are

zero within the perfect conducting mirror).
– Also note that z
2Eo
H(0, t) = ŷ cos(ωt)
η1
on z = 0 surface. Since this tangential magnetic field is not zero, 2π
λ=
boundary condition equations imply that there must be an oscil- β
lating surface current Hy (z, t) ∝ cos(βz) cos(ωt)

2Eo A,
Js = x̂ cos(ωt) m
η1
z
satisfying
−ẑ × H(0, t) = Js.
λ
Js on mirror surfaces is really a convenient idealization of volume cur-
2
rents flowing in thin layers — just a few skin depths — near good- Note: Nulls in Ex and Hy are
separated by half wavelength.
conductor surfaces (real-life mirrors are good but not perfect conduc-
tors!). Radiation due to Js in effect causes the reflected wave and also Adjacent nulls of Ex and Hy
are separated by quarter
cancels out the incident wave field in z > 0. wavelength.

It is useful to think of nulls

of Ex as "shorts" in analogy
Next we examine reflections from a good conductor and see of how to shorts in circuits where v=0.

the limiting case of a perfect conductor is naturally reached. Conductor shorts Ex on its
surface where a current flows.

Also useful to think of nulls

of Hy as "opens" in analogy
to opens in circuits where i=0.

3
 • Going back to the partial reflection case, consider the transmitted fields
Ẽt = x̂τ Eoe−γ2 z
Ẽt and H̃t in Region 2 shown in the margin. Also shown in the margin
are the phasors for current density Jt and magnetic flux density Bt. τ Eo −γ2 z
H̃t = ŷ e
η2
In the box below we integrate the volumetric current density Jt of a good
−γ2 z
conductor from z = 0 to ∞ and find out that this “depth integral” matches J̃t = σ2Ẽt = x̂σ2τ Eoe
the surface current density found above for the case of perfect conductor. In B̃ = µ H̃ = ŷ µ2τ Eo e−γ2 z
t 2 t
this calculation we assume that Region 1 is vacuum, and also take µ2 = µo: η2

Effective surface current: Assuming that Region 2 is a good conductor,

s r
jωµ2 jωµ2 p p
η2 = ≈ and γ2 = jωµ2 (σ2 + jωǫ2 ) ≈ jωµ2 σ2
σ2 + jωǫ2 σ2

and therefore
√
2η2 2η2 2σ2η2 2 jωµ2 σ2 2γ2
τ= ≈ and σ2 τ ≈ = = .
ηo + η2 ηo ηo ηo ηo
The depth integral of the volumetric current density in Region 2, that is, the effective
surface current of the region is then
Z ∞ Z ∞
1 2Eo
J̃t dz = x̂ σ2τ Eo e−γ2 z dz = x̂Eo (σ2τ ) = x̂
0 0 γ2 ηo
in phasor form, matching the phasor of the time-domain result from above, namely
2Eo A
Js = x̂ cos(ωt) m
ηo
representing the surface current on an idealized perfect conductor surface.

4
 Surface resistance: Let J̃s stand for the effective surface current of a good con-
ductor with a propagation constant
p
γ ≈ jωµσ = α + jβ = α + jα

and a volumetric current density J̃(z) such that

Z ∞ Z ∞
J̃(0)
J̃s = J̃(z)dz = J̃(0)e−γz dz = .
z=0 z=0 γ
In that case
J̃s γ −γz
J̃(z) = J̃s γe−γz and E(z) = e
σ
inside the good conductor in terms of the effective surface current J̃s , and the average
power dissipated per unit volume (Joule heating) is
1 e−2αz 1 2α2e−2αz
hJ(z) · E(z)i = |J̃s γ|2 = |J̃s |2 .
2 σ 2 σ
The depth integral of the same quantity, that is the power dissipated per unit area,
is then Z ∞
1
hJ(z) · E(z)idz = Rs |J̃s |2 ,
0 2
with √ r
α πf µσ πf µ
Rs ≡ = = (Ω)
σ σ σ
called the surface resistance.
The surface resistance concept is useful to model loss effects in waveguides and
cavity resonators as studied in ECE 450. Also, we can make use of surface resistance
when modeling lossy transmission lines (see Lecture 39).

5
 • Let’s finally calculate the magnetic component of the Lorentz force on
charge carriers of a good conductor due to the penetrating fields: Radiation pressure propor-
tional to

Radiation pressure: If there are N free charge carriers per unit volume inside a hSi i/c
reflecting mirror, then
N F = N qv × Bt = Jt × Bt shows that plane-TEM waves
not only carry and transport
will be the force per unit volume of the mirror, expressed in terms of current density energy, but also momentum.
Jt = N qv and the magnetic flux density Bt .
Its integral over all z can be interpreted as the total force per unit area of the
mirror, Z ∞ TEM waves not only
Prad = Jt × Bt dz, heat, but also push!
0
having a magnitude known as radiation pressure of the reflecting wave. This is a
time-varying quantity, with a time-average
Z ∞ ((It can also be shown that the momen-
1 tum density of the wave is
hPrad i = Re{J̃ × B̃∗}dz
0Z 2
∞ hSi i/c2 N.s/m3
1 µ2τ Eo −2α2z
= ẑ Re{(σ2τ Eo )( )}e dz
0 2 η2 and (spin) angular momentum density

|Eo |2 2γ2 µ2 2η2 1 |Eo |2 Re{γ2} µ2 ±hSi i/ωc N.s/m2

= ẑ Re{( )( )} = ẑ2
2 ηo η2 ηo 2α2 2ηo α2 ηo
2 for right- and left-circular waves. Mo-
|Eo | µo mentum per photon of energy ~ω can be
= ẑ2 = 2hSii/c, obtained by dividing the above expres-
2ηo ηo
sions by |hSi i|/ωc~, the number density
where of photons in the wave field.))
|Eo |2
hSi i ≡ ẑ
2ηo
is the time-average Poynting vector of the incident wave reflected from the mirror (factor
of 2 in hPrad i is due to the recoil of the wave off the mirror; see Rothman and Boughn,
Am. J. Phys., 77 , 122, 1977).

6
27 Guided TEM waves on TL systems
• An x̂ polarized plane TEM wave propagating in z direction is depicted in the mar-
x
gin.
E z
– A pair of conducting plates placed at x = 0 and x = d would not perturb E×H

the fields except that charge and current density variations would be induced H
Unguided uniform plane
wave propagation in a
on plate surfaces at x = 0 and x = d (on both sides) to satisfy Maxwell’s homogeneous medium
y
boundary condition equations.

• If charge and currents were confined only to interior surfaces of the plates facing one
x
another, fields E and H accompanying them would be restricted to the region in d
I
between the plates, constituting what we would call guided waves. E z
E×H Plate 1
– Such a guided wave field confined to the region between the plates will sat- W
H
I
Plate 2
isfy Maxwell’s equations including a minor fringing component that can be
y
neglected when the plate width W is much larger than plate separation d.

In the following discussion of guided waves in parallel-plate transmission lines

(TL) we will assume W ≫ d and neglect the effects of fringing fields.

– Guided waves produce wavelike surface charge and current variations on plate
surfaces.
– Conversely, wavelike charge and current variations on plate surfaces would
produce guided wave fields.

It is sufficient to apply a time-varying current and/or charge density at some location

z on a parallel-plate TL — e.g., by a time-varying voltage or current source — in
order to “excite” the TL with propagating guided fields.

1
 x
How such excitations propagate away from their “source points” on TL systems will d
be our main subject of study for the rest of the semester. I
E z
E×H Plate 1
• In a parallel-plate TL we ignore any fringing fields and assume that H
I
W Plate 2
TEM wave fields y
E = x̂Ex (z, t) and H = ŷHy (z, t)
occupy the region between the plates. For these fields uniform in x and
y, Faraday’s and Ampere’s laws reduce to scalar expressions
∂H ∂Ex ∂Hy
∇ × E = −µ ⇒ = −µ
∂t ∂z ∂t
and
∂E ∂Hy ∂Ex
∇ × H = σE + ǫ ⇒ − = σEx + ǫ .
∂t ∂z ∂t
• Now, multiply both equations by d and let Note that voltage drop
Z 1
V ≡ Ex d voltage drop from plate 2 to plate 1 V = E · dl = Ex d
2

to obtain is uniquely defined — inde-

∂V ∂Hy ∂Hy ∂V pendent of integration path
= −µd and −d =ǫ + σV. — on constant z surfaces be-
∂z ∂t ∂z ∂t cause with TEM fields

• Next, multiply these with W and let Bz = µHz = 0,

I ≡ Hy W current in z-direction on plate 2 and consequently circulation

d
I Z
(because Jsz = Hy on plate 2) to obtain E · dl = − B · dS = 0
C dt S
∂V ∂I ∂I ∂V when C is on constant z plane
W = −µd and −d = ǫW + σW V.
∂z ∂t ∂z ∂t and dS = ±dxdyẑ.

2
 • We can re-write these equations as
∂V ∂I ∂I ∂V
− =L and − =C + GV
∂z ∂t ∂z ∂t
utilizing
d W W
L=µ , C=ǫ , G=σ
W d d
appropriate for the parallel-plate TL — we recognize these parameters
as inductance, capacitance, and conductance of the parallel plate TL. Telegrapher’s
equations:
– In the equations above the GV term accounts for Ohmic losses of
wave fields having to do with currents leaking between the wires
(plates) of the TL.
∂V ∂I
– Another possible loss term that we have not picked up — because − = L
∂z ∂t
we assumed infinite conducting plates — is a missing RI term in ∂I ∂V
− = C
the right-hand-side of the first equation. ∂z ∂t

Rather than correcting for that at this stage, we will drop the GV term
from the second equation, and focus our attention for a while (until
the last day of the semester, in fact) on ideal lossless transmission lines
governed by the equations shown in the margin — they are known as
known as telegrapher’s equations1.
1
Telegrapher’s equations were first compiled by Oliver Heaviside (of close-up method, unit-step, and
countless other contributions) in 1880’s. Telegraphy was being used worldwide by 1850’s as a means of
rapid communications.

3
 ∂
• Except for − ∂z on the left, the telegrapher’s equations look like the
V − I relations of inductors and capacitors (which is the best way of
remembering them). Telegrapher’s
equations:
• The equations can be readily combined to obtain a 1D scalar wave
equation
∂ 2V ∂ 2V
= LC 2 . ∂V ∂I
∂z 2 ∂t − = L
In analogy to ∂z ∂t
∂ 2 Ex ∂ 2 Ex ∂I ∂V
= µǫ 2 , − = C
∂z 2 ∂t ∂z ∂t
the wave equation for V has d’Alembert wave solutions where

z 1 1 µ
V (z, t) = f (t ∓ ) where v ≡ √ =√ . C = ǫGF, L =
GF
,
v LC µǫ
with “geometrical factor”
• In that case the second telegrapher’s equation demands
W
GF = parallel-plate
∂I ∂V z d
− =C = Cf ′ (t ∓ )
∂z ∂t v =
2π
coax
ln ab
implying an anti-derivative π
= twin-lead
cosh−1 2aD
z f (t ∓ vz )
I(z, t) = ±Cv f (t ∓ ) = ±
v Zo
with √ r r
1 LC L 1 µ
Zo ≡ = = = .
Cv C C GF ǫ
4
 • In summary, d’Alembert wave solutions of telegrapher’s equations are
z f (t ∓ vz )
V (z, t) = f (t ∓ ) and I(z, t) = ±
v Zo
with a propagation speed
1 1
v=√ =√ a) t <0
LC µǫ ++ Wire 2

that equals the wave speed of the associated electric and magnetic fields, 3V
Wire 1
and voltage-to-current ratio representing a characteristic impedance --
r 0 z
r
L 1 µ
Zo = = .
C GF ǫ
b) t = 0+
++++I Wire 2
Telegrapher’s equations and their d’Alembert solutions provide us with a
3V V
“handle” on the following physics: Wire 1
----
I
• Suppose that + and - terminals of a 3 V battery makes contact with
the terminals of a charge neutral TL at t = 0 as depicted in the margin.
We will assume that V (z, t) = I(z, t) = 0 on the TL for t < 0. c) t > 0+
+++++++I Wire 2

As soon as contact is made between the terminals of the battery and the
3V V z = vt
TL, the excess + and - charges on battery terminals will “spill onto” the TL Wire 1

terminals as shown in the figure for t = 0+: -------

I

• what really happens is,

5
 a) t <0
– electrons move from the - terminal of the battery onto the bottom ++ Wire 2
wire of the TL,
3V
– replenished by an equal amount of electrons moving from the top Wire 1
--
wire into the battery via its + terminal, 0 z

giving the overall impression of current flows I (in opposite direction

to electron motion) as marked on the two wires in the diagram. b) t = 0+
++++I Wire 2

– currents I and voltage V marked in the diagram are confined only 3V V

to location z = 0+ at t = 0+, while there is still zero current on ----
Wire 1

I
the rest of the TL!!!

Having unequal currents on a length of wire is in conflict with our

notions from earlier circuit courses, but that’s because earlier courses c) t > 0+
+++++++I Wire 2
taught us “lumped-circuit analysis”, an approximate technique jus-
V z = vt
tified when it’s OK to ignore certain time delays of charge movements 3V
Wire 1
in the circuit (when wire lengths are sufficiently short). -------
I

Having unequal currents on the TL wire is really what happens

– because, for instance, electrons at some z > 0 on the top wire will
start moving towards the battery terminal only after the neighbor-
ing electrons at z − deplete the region leaving some excess positive
charge.

Thus, currents I on the wires, and voltage V defined and measured

across the wires, spread out of z = 0 region at a finite speed v, in

6
 analogy with ripples spreading out on a pond surface when perturbed
by a falling pebble.

• Telegrapher’s equations and their d’Alembert solutions will

allow us to calculate how I and V evolve on the TL in quan-
titative terms.
To appreciate the distinction between lumped and distributed circuit anal-
ysis, we next develop a lumped circuit model of a very short length of a TL
over which lumped circuit notions may be applicable:
• Let us re-write the first telegrapher’s equation as ∂V ∂I
− = L
∂I ∂z ∂t
−∆V ≡ V (z, t) − V (z + ∆z, t) = ∆zL ∂I ∂V
∂t − = C
∂z ∂t
after approximating the left side as a ratio of infinitesimals.

– This relation shows that in the current flow direction there is an

infinitesimal inductive voltage drop of ∆zL ∂I
∂t
between points z
I(z) I(z + ∆z)
and z +∆z on the wire carrying current I ≡ I(z, t) ≈ I(z +∆z, t).
Wire 2 + ∆zL
• Likewise, the second equation re-arranged as V (z) ∆zC
∂V Wire 1 -
−∆I ≡ I(z, t) − I(z + ∆z, t) = ∆zC , z z + ∆z
∂t
– this shows that an infinitesimal capacitor current ∆zC ∂V
∂t
flows out
of a node located between z and z + ∆z on the wire with current
I into a node on the second wire at the same location.

7
Evidently, a short section ∆z of the TL has an equivalent T-network with I(z) I(z + ∆z)
1. a series inductance ∆zL carrying a current I(z, t) ≈ I(z + ∆z, t), and Wire 2 + ∆zL
V (z) ∆zC
2. a shunt capacitance ∆zC carrying a voltage V (z + ∆z, t) ≈ V (z, t)
Wire 1 -
as shown in the margin. z z + ∆z
This lumped-circuit equivalent is only accurate for ∆z so small that
I(z, t) ≈ I(z + ∆z, t) and V (z + ∆z, t) ≈ V (z, t)
are both true, which is of course possible only if ∆z ≪ λ, λ being the shortest
wavelength in I(z, t) ∝ H(z, t) and V (z, t) ∝ E(z, t) waveforms. TL’s can also support non-TEM

modes having non-zero compo-

• Going back to parallel-plate TL in TEM mode, observe that the total
nents of Hz or Ez . These modes
power transported in the guide will be the Poynting vector E × H =
are non-propagating (evanescent)
Ex Hy ẑ times the cross-sectional area of the guide, namely, W d.
at low frequencies and remain lo-
Thus, power transported in z direction is calized near their excitation re-

p(z, t) = W dEx(z, t)Hy (z, t), gions (e.g., discontinuity points on

λ
the line) if d < (pp TL) or if
= (Ex(z, t)d)(Hy (z, t)W ) = V (z, t)I(z, t) 2

a+b < λ
π (coax). At high frequen-
the familiar formula from circuit theory. cies when these modes cannot be

Hence, the circuit theory formula avoided with practical dimensions

d, a, and b, it may be practicable

1
P = Re{Ṽ I˜∗} to use them rather than the TEM
2
mode. Use single-wire waveguides
for average power will also hold in sinusoidal-steady state TL problems when
in that case instead of two-wire
˜ to represent the V (z, t) and I(z, t) waveforms.
we use phasors Ṽ (z) and I(z)
TL’s.

8
28 Distributed circuits and bounce diagrams
Last lecture we learned that voltage and current variations on TL’s are gov-
erned by telegrapher’s equations and their d’Alembert solutions — the latter
can be expressed as
z z f (t − vz ) g(t + zv )
V (z, t) = f (t − ) + g(t + ) and I(z, t) = −
v v Zo Zo
in terms of r
1 L Source
ckt
Transmission line
v=√ and Zo = Rg I(z, t) Wire 2
IL
LC C + + Load
+ V (z, t) Zo VL RL
and functions f (t) and g(t) corresponding to signal waveforms propagated in -
fi (t) I(z, t) - Wire 1 -
+z and −z directions, respectively. 0 l z

• In this lecture we will learn how to solve distributed circuit prob-

lems containing TL segments and two terminal elements such as resis-
tors and voltage (or current) sources. In solving the problems, we will
apply the usual rules of lumped circuit analysis at element terminals
and treat the TL’s in terms of d’Alembert solutions above.

• Consider a TL with a characteristic impedance Zo extending from z = 0

to z = l, where a two-terminal source circuit (e.g., a receiving antenna)
modeled by a Thevenin equivalent with voltage fi(t) and resistance
Rg is connected between the TL terminals at z = 0 and a load (e.g.,
a receiver circuit) modeled by a resistance RL terminates the line at
z = l (see margin).

1
 – We want to determine voltage and current signals V (z, t) and
I(z, t) on the TL and the load RL for time t > 0 in terms of
source signal fi(t) assuming that fi (t) = 0 for t < 0.

• Using the d’Alembert solutions V (z, t) and I(z, t) from above at z = l,

we have
Source Transmission line
V (ℓ, t) f (t − vℓ ) + g(t + vℓ ) f (t − vℓ ) + g(t + vℓ ) VL ckt
Rg I(z, t) Wire 2
IL
= = Zo = = RL , +
I(ℓ, t) f (t− vℓ )
−
g(t+ vℓ ) f (t − vℓ ) − g(t + vℓ ) IL + V (z, t) Zo
+ Load
VL RL
Zo Zo -
fi (t) I(z, t) - Wire 1 -
from which we obtain 0 l z

l RL − Zo l 2l
g(t + ) = f (t − ) ⇒ g(t) = ΓLf (t − )
v |RL {z
+ Zo} v v
ΓL
where
RL − Zo
ΓL =
RL + Zo
is the load reflection coefficient in the TL circuit. We can re-write
the d’Alembert solution for V (z, t) and I(z, t) in terms of only f (t) as
z z 2l f (t − zv ) ΓLf (t + zv − 2lv )
V (z, t) = f (t− )+ΓLf (t+ − ) and I(z, t) = − .
v v v Zo Zo
• Assuming that fi(t) = 0 = f (t) for t < 0, we can relate f (t) to fi (t) in
t > 0 interval using the KVL equation at z = 0 that states
fi(t) = Rg I(0, t) + V (0, t),

2
 which is, using V (z, t) and I(z, t) at z = 0,

f (t) ΓLf (t − 2lv ) 2l

fi(t) = Rg ( − ) + f (t) + ΓLf (t − ) .
| Zo {z Zo } | {z v}
I(0, t) V (0, t)

Now, since f (t − 2lv ) = 0 for t − 2l

v
< 0, we find out that for the epoch
(or time interval ) 0 < t < 2lv ,

f (t) Zo
fi (t) = Rg + f (t) ⇒ f (t) = fi (t) Source Transmission line
Zo Rg + Zo ckt
Rg I(z, t) IL
| {z } Wire 2
+ + Load
τg + V (z, t) Zo VL RL
-
fi (t) I(z, t) - Wire 1 -
where 0 l z
Zo
τg =
Rg + Zo
is the injection coefficient of the TL circuit1 .
2l
• Thus, for the epoch 0 < t < v, we have the voltage and current
solutions
z z 2l τg fi(t − vz ) ΓLτg fi (t + vz − 2lv )
V (z, t) = τg fi(t− )+ΓL τg fi (t+ − ) and I(z, t) = −
v v v Zo Zo
on the line.
1
Note how f (t) appears to be related to fi (t) according to a voltage division rule with Zo representing
the resistance across which voltage f (t) is measured.

3
 – So far fi(t) function is arbitrary and the above results would also
be valid for fi(t) = δ(t), Dirac’s impulse, in which case

z z 2l τg δ(t − zv ) ΓLτg δ(t + vz − 2lv )

V (z, t) = τg δ(t− )+ΓLτg δ(t+ − ) and I(z, t) = −
v v v Zo Zo
would be the voltage and current impulse response functions
of the TL circuit for the 0 < t < 2lv epoch.
Rg I(z, t) IL

• To extend the impulse response functions above to the “next epoch” + +

2l 4l + V (z, t) Zo VL RL
v < t < v , we note that at z = 0 the KVL equation with fi (t) = δ(t)
-
δ(t) I(z, t) - -
reads as 0 l z

f (t) ΓLf (t − 2lv ) 2l Bounce diagram

δ(t) = Rg ( − ) + f (t) + ΓLf (t − ) . l

| Zo {z Zo v} z
} | {z τg
l
I(0, t) V (0, t) τ g ΓL v
2l
v τ g ΓL Γg
which can be re-arranged as 3l
τg Γ2L Γg v
Rg Rg 2l 4l
δ(t) = (1 + )f (t) + (1 − )ΓL f (t − ), v τg Γ2L Γ2g
Zo Zo | {z v }
τg Γ3L Γ2g
2l
τg δ(t − )
v t

where for f (t − 2lv ) we used a delayed copy of f (t) = τg fi (t) solution

for f (t) from the previous epoch in view of the time delay 2lv contained
within f (t − 2lv ).

4
– Hence, solving this for f (t), we find, for this epoch,
Rg − Zo 2l
f (t) = τg δ(t) + ΓLτg δ(t − ),
R +Z v
| g {z o}
Γg

where
Rg − Zo
Γg =
Rg + Zo
is the source reflection coefficient of the TL circuit.
2l 4l
– Substituting f (t) for the epoch v
< t < v
within voltage and
current formulae
z z 2l f (t − vz ) ΓLf (t + vz − 2lv )
V (z, t) = f (t− )+ΓL f (t+ − ) and I(z, t) = −
v v v Zo Zo
we obtain the “extended” voltage and current impulse response
functions
z z 2l z 2l z 4l
V (z, t) = τg δ(t− )+ΓLτg δ(t+ − )+Γg ΓLτg δ(t− − )+Γg Γ2Lτg δ(t+ − )
v v v v v v v
and
z z 2l z 2l z 4l
I(z, t) = Zo−1 [τg δ(t− )−ΓLτg δ(t+ − )+Γg ΓL τg δ(t− − )−Γg Γ2L τg δ(t+ − )]
v v v v v v v
respectively.

5
 ◦ At this point the algebra is pretty messy, but a straightforward Rg I(z, t) IL

pattern is emerging (to obviate the need for algebraic analysis + +

+ V (z, t) Zo VL RL
for the upcoming epochs) that is best appreciated with the δ(t)
-
I(z, t) - -
help of bounce diagrams explained next: 0 l z

– A bounce diagram is a plot of the “trajectories” of traveling im- Bounce diagram

pulses found on transmission line segments excited by impulse in- τg
l
z

puts. l
τ g ΓL v
– The horizontal axis represents position z of the traveling impulses 2l
v τ g ΓL Γg
while time t is represented by a downward pointing axis. 3l
τg Γ2L Γg v
– The first slanted line on the top of the diagram, representing the 4l
v τg Γ2L Γ2g
traveling impulse
z
τg δ(t − ), τg Γ3L Γ2g
v
(first term of hz (t) = V (z, t)) is “reflected” at time t = vℓ from load t

RL to turn into a backward propagating impulse

z 2ℓ
τg ΓLδ(t + − )
v v
represented by the second line of the diagram.
2ℓ
– The backward propagating impulse reaches z = 0 at t = v and is
reflected once more with a reflection coefficient
Rg − Zo
Γg =
Rg + Zo

6
 to become a forward propagating impulse
z 2ℓ
τg ΓLΓg δ(t −
− )
v v
represented by the third line of the diagram.
◦ Reflection at Rg is in effect the same physical process as re-
flection at RL and therefore its coefficient Γg is identical with Rg IL
I(z, t)
ΓL except for the replacement of RL by Rg . + +
+ V (z, t) Zo VL RL
– The bounce diagram is advanced in time with further reflections δ(t)
-
I(z, t) - -
occurring at both ends. 0 l z

– We show the calculated weights of traveling impulses directly on Bounce diagram

the diagram just above the slanted lines representing the trajec- l
z
τg
tories of each traveling impulse (each having a lifetime of ℓ/v) l
τ g ΓL v
• Using the bounce diagram, the full expressions for the voltage and 2l
v τ g ΓL Γg
current impulse response functions of the circuit can be written as 3l
∞ v
X z 2ℓ τg Γ2L Γg
V (z, t) = τg (ΓL Γg )nδ(t − −n ) 4l
n=0
v v v τg Γ2L Γ2g
∞
X z 2ℓ τg Γ3L Γ2g
+τg ΓL (ΓL Γg )n δ(t + − (n + 1) )
n=0
v v
t
and
∞
τg
X z 2ℓ
I(z, t) = Zo (ΓLΓg )n δ(t − −n )
n=0
v v
∞
τg X z 2ℓ
− ΓL (ΓL Γg )n δ(t + − (n + 1) ).
Zo n=0 v v

7
• Although these series formulae look daunting, only the lower order
terms usually matter — that is true because |ΓL| ≤ 1 and |Γg | ≤ 1 and Rg I(z, t) IL

thus (ΓLΓg )n is typically a rapidly diminishing function of n (unless + +

+ V (z, t) Zo VL RL
the ckt is “dissipation free” and resonant, a concept explored in Lecture δ(t)
-
I(z, t) - -
31). 0 l z

• We typically rely on the bounce diagram technique more so than the Bounce diagram
l
z
series expressions developed above. This will be illustrated by several τg
l
examples in the next lecture. τ g ΓL v
2l
v τ g ΓL Γg
– The main idea is to combine delayed versions of the circuit input 3l
fi(t) with the impulse weights indicated on the bounce diagram, τg Γ2L Γg v
4l
since, in general, the convolution δ(t − Tz ) ∗ fi (t) = fi (t − Tz ) for v τg Γ2L Γ2g

any z-dependent delay such as zv , zv − 2ℓv , etc... τg Γ3L Γ2g

8
29 Bounce diagrams
• Last lecture we obtained the implulse-response functions Source matched to
line:
z z 2l Zo I(z, t)
V (z, t) = τg [δ(t − ) + ΓLδ(t + − )]
v v v +
+ V (z, t) Zo RL
-
and f (t) I(z, t) -
τg z z 2l 0 l z
I(z, t) = [δ(t − ) − ΓLδ(t + − )]
Zo v v v Bounce diagram l
for the voltage and current in the TL circuit shown in the margin where z
τg
l
the source is matched to the line so that τg = 21 — circuit response with τg Γ L v
an arbitrary input f (t) is obtained by convolving these with f (t) (as 2l
v
shown in Example 1 in last lecture).

• The impulse-response for V (z, t) is depicted in the margin in the form

of a bounce diagram, in which t

– the trajectories of the impulses constituting the impulse response

are plotted, with
◦ z axis in the horizontal, and
◦ t axis in the vertical extending from top to bottom
– and coefficients of each impulse noted in the diagram next to the
trajectory lines.
– the blue line sloping down on the top is a depiction of forward
propagating impulse τg δ(t − vz ),

1
 – the next line down is the depiction of backward propagating im-
pulse τg ΓLδ(t + zv − 2lv ).

Bounce diagrams are graphical representations of impulse re-

sponse functions derived in TL circuit problems, and are pri-
marily used to determine the impulse response functions, Rg I(z, t)
+
rather than the other way around as will be illustrated below. +
- V (z, t) Zo RL
f (t) I(z, t) -
• We show in the margin a circuit with an arbitrary 0 l z

Zo Bounce diagram
Rg and τg = , l
Rg + Zo τg
z

l
2l
for which the bounce diagram is not terminated at t = because the
v τ g ΓL v
2l
backward propagating impulse on the line arriving at z = 0 at time v τ g ΓL Γg

t = 2lv is reflected from z = 0 with a reflection coefficient of 3l

v
τg Γ2L Γg
4l
Rg − Zo v τg Γ2L Γ2g
Γg = .
Rg + Zo τg Γ3L Γ2g

– Reflections of negative-going impulses incident on the source cir- t

cuit are justified because these impulses just see the resistor Rg at
the generator end — the source voltage f (t) = δ(t) is by then just
a short is series with Rg — unmatched to Zo , just like the forward
going impulses seeing a load RL unmatched to Zo and reflecting
with a coefficient
RL − Zo
ΓL = .
RL + Zo
2
• Once the bounce diagram for voltage has been constructed as shown Bounce diagram
l
above, then the impulse response can be written by inspection as τg
z

l
∞
X z 2l τ g ΓL v
V (z, t) = τg (ΓLΓg )n δ(t − −n ) 2l
n=0
v v v τ g ΓL Γg
∞ 3l
X
n z 2l v
+τg ΓL (ΓLΓg ) δ(t + − (n + 1) ). τg Γ2L Γg

n=0
v v 4l
v τg Γ2L Γ2g

Also, τg Γ3L Γ2g

∞
τg
X z 2l
I(z, t) = Zo (ΓLΓg )n δ(t − −n ) t
n=0
v v
∞
τg X z 2l
− ΓL (ΓLΓg )n δ(t + − (n + 1) ).
Zo n=0 v v

– It can also be shown that the first term of V (z, t) above is derived
from the formal solution of the equation
2l
V +(t) = τg δ(t) + ΓLΓg V +(t − )
v
which is obtained from
δ(t) − V (0, t)
I(0, t) =
Rg
enforced at z = 0. We have effectively by-passed such a formal
approach to the problem by using the bounce diagram technique.

3
• These awful series formulae above are hardly needed in most appli-
cations when only the first few terms of the series are sufficient for
reasonably accurate results (like in the next example).

4
 z = 1200 m
I(z, t)
+
+
- V (z, t) RL = 100 Ω
u(t) I(z, t) - Zo = 50 Ω
Example 1: Consider a TL circuit where Zo = 50 Ω, v = c, l = 2400 m, Rg = 0, and 0 l z
RL = 100 Ω. Determine and plot V (1200, t) if f (t) = u(t).
Bounce diagram
l
z
Solution: For this circuit Γg = −1 1 ΓL =
1
3 l
Zo Rg − Zo RL − Zo 1 1
v
τg = = 1, Γg = = −1, and ΓL = = . 3
Rg + Zo Rg + Zo RL + Zo 3 2l 1
v −
3
Also, the transit time across the TL is 1
3l
− v
9
l 2400 m 4l 1
= = 8 µs. v
v 300 × 106 m/s 9
1
From the bounce diagram shown in the margin, the impulse response for z = 1200 27

m (the location marked by the vertical dashed line) is found to be −

1
t 27
1 1 1 1
V (1200, t) = δ(t − 4) + δ(t − 12) − δ(t − 20) − δ(t − 28) + δ(t − 36) + · · ·
3 3 9 9 vH4,tL
2.0

Replacing the δ(t) in this expression with the unit-step u(t), the specified source 1.5

function f (t), we get

1.0
1 1 1 1
V (1200, t) = u(t − 4) + u(t − 12) − u(t − 20) − u(t − 28) + u(t − 36) + · · ·
3 3 9 9 0.5

which is plotted in the margin. 0 10 20 30 40 50 60

t us

Animated version of this is

linked in the class calendar.

5
 Bounce diagram
l
1 z
• Note that as t → ∞, V (1200, t) → 1 V in Example 1, as if DC condi- Γg = −1 1 ΓL =
3 l
1
tions prevail and the TL becomes a pair of wires in the lumped circuit 3
v
2l
sense. v −
1
3
3l
1
− v
– DC steady-state corresponds to ω = 0 and signal wavelength λ → 4l
9
1
∞. In that limit l ≪ λ is always valid and TL can be treated like v
9
1
an ordinary lumped circuit. 27
1
– Of course this simplification can only occur with f (t) ∝ u(t), or t
−
27

its delayed versions, which are all asymptotically DC in t → ∞

limit. The simplification does not apply for f (t) = sin(ωt)u(t), for vH4,tL
2.0

example. 1.5

1.0

0.5

t us
0 10 20 30 40 50 60

6
Example 2: In the TL circuit described in Example 1, determine V (z, t) and I(z, t)
for a new source signal f (t) = rect( Tt )+2rect( t−T
T ), T = 1 µs. Plot V (z, t) versus
z at t = 3 µs and t = 11 µs.

Solution: With τg = 1, Γg = −1, ΓL = 31 , and 2lc = 16 µs, we obtain, by convolving fHtL

2.0

with the general impulse response, the voltage response

1.5
∞ ∞
X 1 n z 1X 1 n z
V (z, t) = (− ) f (t − − n16) + (− ) f (t + − (n + 1)16) 1.0

n=0
3 c 3 n=0 3 c
0.5
z
where c is to be entered in µs units. Also,
t us
-2 -1 1 2 3 4
∞ ∞
1 X 1 n z 1 1X 1 n z VHz,3L
I(z, t) = (− ) f (t − − n16) − (− ) f (t + − (n + 1)16). 2.0
50 n=0 3 c 50 3 n=0 3 c
1.5

At t = 3 µs, the voltage variation is

1.0

∞ ∞
X 1 z 1X 1 n z
V (z, 3) = (− )nf (3 − − n16) + (− ) f (3 + − (n + 1)16), 0.5

n=0
3 c 3 n=0 3 c
zm
500 1000 1500 2000

which is plotted in the margin using f (t) = rect(t) + 2rect(t − 1). Likewise, at VHz,11L
2.0
t = 11 µs,
1.5
∞ ∞
X 1 n z 1X 1 n z
V (z, 11) = (− ) f (11 − − n16) + (− ) f (11 + − (n + 1)16). 1.0

n=0
3 c 3 n=0 3 c
0.5

zm
0 500 1000 1500 2000

Animated version of this is

linked in the class calendar.

7
30 Multi-line circuits Rg = 2Z1

Z1 Z2 = 2Z1
Load
+ RL = Z2
-
v1 v2 = 2v1
f (t)
• In this lecture we will extend the bounce diagram technique to solve 0 l/2 l

distributed circuit problems involving multiple transmission lines. 1 4

ΓL = 0
τg = Γg = Γ12 = τ12 =
0 3 3 l z
• One example of such a circuit is shown in the margin where two distinct 1

TL’s of equal lengths have been joined directly at a distance 2l away

3
4
9
1
from the generator. 9

– The impulse response of the system can be found by first con- 27

4
81

structing the bounce diagram for the TL system as shown in the 1

81

margin. t

– In this bounce diagram, z = 2l happens to be the location of ad-

ditional reflections as well as transmissions because of the sudden
change of Zo from Z1 to Z2 = 2Z2.

These reflections and transmissions between line j and k — transmis-

sion from j to k, and reflection from k back to j — can be computed
with reflection coefficient
Zk − Zj
Γjk =
Zk + Zj
and transmission coefficient
τjk = 1 + Γjk
that ensure the voltage and current continuity at the junction

1
 Rg = 2Z1
– Zj is the characteristic impedance of the line of the incident pulse, Load
Z1 Z2 = 2Z1
while +
-
f (t)
v1 v2 = 2v1
RL = Z2

0 l/2 l
– Zk is the impedance of the cascaded line into which the transmitted
pulse is injected. τg = Γg = Γ12 =
1
3
τ12 =
4
3
ΓL = 0
0 l z

Verification:
1
3
4
9
1

– Let 9

Vj+(1 + Γjk ) and Vj+(1 − Γjk )/Zj 1

27
4

denote the total voltage and current on line Zj expressed in terms 1

81

81

of incident voltage wave V + (of d’Alembert type), and t

– let
Vj+τjk and Vj+τjk /Zk
denote the voltage and current on line Zk adjacent to line Zj .

This notation identifies Γjk and τjk as reflection and transmission coef-
ficients at the junction.

– Taking
Vj+(1 + Γjk ) = Vj+τjk
and
Vj+(1 − Γjk )/Zj = Vj+τjk /Zk
in order to enforce voltage and current continuity, we can solve for
Γjk and τjk expressions stated above.

2
 Rg = 2Z1
Load
+ Z1 Z2 = 2Z1
- RL = Z2
v1 v2 = 2v1
f (t)
0 l/2 l
Example 1: In the circuit shown in the margin with two TL segments, line 2 has twice
the characteristic impedance and propagation velocity of line 1, i.e., 1 4
ΓL = 0
τg = Γg = Γ12 = τ12 =
0 3 3 l z
Z2 = 2Z1 and v2 = 2v1. 1
3

Determine L2 and C2 in terms of L1 and C1 . 4

9
1
9

Solution: We have
L2 L1 1
Z2 = 2Z1 ⇒ =4 27
C2 C1 4
81

and 1
81
1 1
v2 = 2v1 ⇒ =4 . t
L2 C2 L1C1
The product of the two equations gives
1 1 1
= 16 ⇒ C2 = C1 ,
C22 C12 4
while their ratio leads to
L2 = L1 .

3
 Rg = 2Z1
Load
+ Z1 Z2 = 2Z1
- RL = Z2
v1 v2 = 2v1
f (t)
Example 2: In the circuit of Example 1, determine V (z, t) and I(z, t) if 0 l/2 l

f (t) = sin(2πt)u(t), t in µs, 1 4

τg = Γg = Γ12 = τ12 = ΓL = 0
3 3
and l = 2400 m, v1 = 150 m/µs, and Z1 = 25 Ω. 0 l z
1

Solution: From the bounce diagram we infer the following impulse-response for the 3
4

voltage variable: 1
9

9
∞
1 X 1 2n z l 1 z l
V (z, t) = ( ) [δ(t − − n ) + δ(t + − (n + 1) )] 1
3 n=0 3 v1 v1 3 v1 v1 27
4
81
l
for z < 2
, and 1
81
∞
1 X 1 2n 4 z l/2 t
V (z, t) = ( ) δ(t − − (4n + 1) )
3 n=0 3 3 v2 v2
l
for 2 < z < l. The impulse response for the current is t

∞
1 X 1 2n z l 1 z l VHz,3 usL

I(z, t) = ( ) [δ(t − − n ) − δ(t + − (n + 1) )] 1.0

3Z1 n=0 3 v1 v1 3 v1 v1 0.5

for z < 2l , and

zm
500 1000 1500 2000

-0.5

∞
1 X 1 4 z l/2
( )2n δ(t − − (4n + 1) )
-1.0

I(z, t) =
3Z2 n=0
3 3 v2 v2
l
for 2
< z < l. Using
l 2400
= = 16 µs
v1 150
and replacing δ(t) with f (t) = sin(2πt)u(t) the plot depicted in the margin was
obtained.

4
 V+−V− V ++
Z1 Z2
+ +
Example 3: Two TL’s with characteristic impedances Z1 and Z2 are joined at a junc- V + + V −- -V ++
Z1 R Z2
tion that also includes a “shunt” resistance R as shown in the diagram in the
margin. Determine the reflection coefficient Γ12 and transmission coefficient τ12 z
at the junction.

Solution: Consider a voltage wave

z
V +(t − )
v1
coming from the left producing reflected and transmitted waves
z z
V −(t + ) and V ++(t − )
v1 v2
on lines 1 and 2 traveling to the left and right, respectively, on two sides of the
junction. Using an abbreviated notation, KVL and KCL applied at the junction
can be expressed as
V+ V− V ++ V ++
V + + V − = V ++ and − = + ,
Z1 Z1 R Z2
where in the KCL equation the first term on the right is the current flowing down
the resistor R, and the second term is the TL current on line 2 (as marked in the
circuit diagrams in the margin). The equations can be rearranged as
V + + V − = V ++
Z1 ++
V+−V− = V ,
Zeq
where
RZ2
Zeq ≡
R + Z2

5
 is the parallel combination of R and Z2 . Solving these equations, we find that
V− Zeq − Z1
Γ12 ≡ =
V+ Zeq + Z1
and
V ++ 2Zeq
τ12 = + = .
V Zeq + Z1

By, symmetry, the coefficients

Zeq − Z2
Γ21 =
Zeq + Z2
and
2Zeq
τ21 =
Zeq + Z2
would describe reflection and transmission when a wave is incident from right
provided that
RZ1 V+−V− V ++
Zeq ≡
R + Z1 Z1 Z2
is used. + R +
V++V− V ++
Z1 - - Z2

z
Hint: in this ckt Γ12 has the
usual form in terms of Zeq ≡
R+Z2 . For τ12 , we need 1+Γ12
multiplied by a voltage divi-
Exercise: Two TL’s with characteristic impedances Z1 and Z2 are joined at a junction sion factor Z2 /(R + Z2 ).
that also includes a series resistance R as shown in the margin. Determine the
reflection coefficient Γ12 and transmission coefficient τ12 at the junction.

6
31 Periodic oscillations in lossless TL ckts +
C I(t) V (t)
• Lossless LC circuits (see margin) can support source-free and co-sinusoidal L -
voltage and current oscillations at a frequency of
1
ω=√ |V (z, t)|
LC |I|

known as LC resonance frequency.

• Lossless TL circuits can also support source-free voltage and current

oscillations, but the number of resonance frequencies is infinite and the Open λ
4
Open

oscillation waveforms are not restricted to co-sinusoidal forms. 0 l z

A TL segment — a “stub”,
so to speak — open circuited
– Resonance frequencies of lossless TL’s are harmonically related, at both ends can support
voltage and current oscilla-
and therefore superpositions of resonant oscillations on TL’s can tions such that the current
add up to arbitrary periodic waveforms as in Fourier series repre- waveform vanishes at both
ends. Absolute values of a
sentation of periodic functions. possible set of voltage and
current waveforms satisfying
this boundary condition are
In this lecture we will examine the periodic oscillations and resonances depicted above.
encountered in lossless and source-free TL circuits.

1
• Consider first a TL segment of some length ℓ having no electrical con-
nection to any elements at either end, as shown in the margin.
|V (z, t)|
– Effectively, both ends of the TL have been “open circuited”, and |I|
thus TL current I(z, t) needs to vanish at z = 0 and z = ℓ. Since
f (t − vz ) g(t + vz )
I(z, t) = −
Zo Zo Open λ Open
4

in general, these boundary conditions 0 l z

A TL “stub” open circuited
f (t) g(t) at both ends can support
I(0, t) = − =0 voltage and current oscilla-
Zo Zo tions such that the current
waveform vanishes at both
and ends. Absolute values of a
f (t − vℓ ) g(t + vℓ ) possible set of voltage and
I(l, t) = − =0 current waveforms satisfying
Zo Zo this boundary condition are
depicted above.
require that
◦ g(t) = f (t)
◦ f (t − vℓ ) = f (t + vℓ ) ⇒ f (t) = f (t + 2ℓv ).
– the first condition says that forward and backward going waves
are described in terms of a single time function f (t),
– while the second condition indicates that function f (t) is neces-
sarily periodic with a
◦ period T = 2ℓv
2π πv
◦ fundamental frequency ωo = T = ℓ

2
 Since no other constraint is imposed, any waveform with the spec-
ified period is admissible, and the most general such expression is
given by the Fourier series
∞
X
f (t) = Fo + Fn cos(nωot + θn )
n=1

having harmonically related frequencies nωo and arbitrary Fourier

coefficients Fn and θn .
– Hence, in general, the line current
f (t − vz ) − f (t + vz )
I(z, t) =
Zo
∞
X Fn
= [cos(nωot + θn − nβoz) − cos(nωot + θn + nβoz)]
n=1 o
Z

where βo ≡ ωo/v = π/ℓ is the fundamental wavenumber.

– The same result written in phasor form is
∞ ∞
˜ =
X Fn jθn −jnβo z jnβo z
X Fn
I(z) e [e −e ]= ejθn (−2j) sin(nβoz),
n=1
Zo n=1
Zo

which also means that (back in the time domain)

∞
X 2Fn
I(z, t) = sin(nωot + θn ) sin(nβoz).
n=1
Zo

3
 Also1 ∞ 1.0
X
V (z, t) = 2Fn cos(nωot + θn ) cos(nβoz) 0.5

n=1
z HmL
jθn −jnβo z jnβo z 50 100 150 200 250 300
P
from the phasor Ṽ (z) = n Fn e [e +e ].
-0.5

In summary: -1.0

A time-snapshot of the cur-

rent standing wave modes
– Periodic variations of arbitrary complexity — or timbre, in analogy n = 1, 2, 3, 4 on a TL segment
with musical instruments — in V (z, t) and I(z, t) are allowed on an 300 m long, open ended on
both sides. Each mode n has
open circuited (on both ends) TL segment of length ℓ and consist n half wavelengths fitted into
of superpositions of resonant modes (see margin) the line length l and high n
modes oscillate with higher
πv π πv π frequencies. See animation
cos(n t + θn ) cos(n z) and sin(n t + θn ) sin(n z), of these modes linked in the
ℓ ℓ ℓ ℓ class calendar.

respectively, in the range n ≥ 1, each one being a standing wave.

– Each resonant mode or standing wave of index n ≥ 1 has a
◦ resonance frequency
πv v
ω= n rad/s or f = n Hz
ℓ 2ℓ
◦ resonance wavelength
v 2ℓ
λ= = ,
f n
1
Note that an arbitrary DC term can also be included in V (z, t).

4
 implying that
λ
ℓ=n ,
2
that is, the line length is an integer multiple of half-
wavelength at each resonance.

• The resonances examined above also apply to TL’s of length ℓ shorted

at both ends, provided that the mode equations above are swapped
between voltage and current — that is, periodic variations of arbitrary
complexity in I(z, t) and V (z, t) consist of superpositions of resonant
modes
πv π πv π
cos(n t + θn ) cos(n z) and sin(n t + θn ) sin(n z),
ℓ ℓ ℓ ℓ
respectively, in the range n ≥ 1.
Note that in this case the voltage modes vanish at z = 0 and z = ℓ as
required by the boundary condition V (0, t) = V (ℓ, t) = 0 imposed by
having shorts at both ends.

5
 |I|
• For TL’s of length ℓ open at one end shorted at the other end, |V |
resonant wavelengths and frequencies can be identified by requiring ℓ
to be an odd multiple of λ4

– the reason for this is, the nulls of waveforms ∝ cos(βz) and sin(βz) Open λ
4
Short

are separated by odd multiples of 0 l z

A TL stub open at one end
λ 2π/β π short at the other can sup-
= = . port voltage and current os-
4 4 2β cillations such that the cur-
rent waveform vanishes at
– Hence, resonance condition is the open end while the volt-
age waveform vanishes at the
λ shorted end. Absolute values
ℓ= (2n + 1), n ≥ 0, of a possible set of voltage
4 and current waveforms sat-
isfying this boundary condi-
and since tion are depicted above. Res-
onant standing waves modes
fλ = v on this line will have voltage
and current nulls separated
it follows that the resonance frequencies are by an odd multiple of a quar-
ter wavelength.
v 1 πv 1
f= ( + n) and ω = ( + n) for n ≥ 0.
2ℓ 2 ℓ 2

6
 1.0

0.5
Example 1: A lossless TL of 600 m length is left open at z = 0 and shorted at
z = l = 600 m. Determine (a) resonant frequencies of the line, (b) resonant 100 200 300 400 500 600
z HmL

current modes, (c) resonant voltage modes obtained from the current modes
using the telegrapher’s equations. The line has a characteristic impedance of -0.5

Zo = 50 Ω and a propagation velocity v = c. -1.0

A time-snapshot of the volt-

Solution: (a) The line must be an odd multiple of quarter wavelengths at the resonant age standing wave modes n =
frequencies. Therefore, 0, 1, 2, 3 on a TL segment 600
m long, open ended at z =
λ c/f 0 and shorted at z = 600
600 m = (2n + 1) ⇒ 600 m = (2n + 1) m. Each mode n has 2n +
4 4
1 quarter wavelengths fitted
leading to into the line length l and the
300 m/µs 1 high n modes oscillate with
f = (2n + 1) = (2n + 1) MHz, n ≥ 0. higher frequencies. See ani-
4 · 600 m 8
mation of these modes linked
(b) Since the current modes need to vanish at z = 0, we can express them in terms in the class calendar.
of a sine function as
sin(βz) sin(ωt)
where
π
ω = 2πf = (2n + 1) Mrad/s,
4
and
2π π
β= = (2n + 1) rad/m.
λ 1200
In explicit terms, current modes are
π π
In (z, t) = sin((2n + 1) z) sin((2n + 1) t)
1200 4
where z is in m and t in µs.

7
(c) Let’s find the voltage modes Vn (z, t) from the current modes above using one of
the telegrapher’s equations,
∂V ∂I
− =L .
∂z ∂t
Substituting In (z, t) into this equation and differentiating we find
∂V π π π
= −(2n + 1) L sin((2n + 1) z) cos((2n + 1) t).
∂z 4 1200 4
Next finding the anti-derivative of the above, we conclude
π π
Vn (z, t) = 300L cos((2n + 1) z) cos((2n + 1) t).
1200 4
A snapshot of the animation of resonant voltage modes is shown in the margin. 1.0

0.5
Note that the amplitudes of current and voltage modes cannot be assigned indepedent
of one another — above we set the current amplitudes to unity and obtained z HmL
100 200 300 400 500 600
300L for voltage amplitudes.
-0.5

-1.0

8
 t = 1 µs
• How can one get source free oscillations in a TL?
t=0
One answer is, the TL might have been connected to a source in the 3V V (z, 1)
past before being disconnected from it. z
0 600 m

– Consider the circuit shown in the margin where a 3V battery is

switched in and out for 1 µs on a line of length l = 600 m. Also, resistors R at tempera-
ture T connected to TL ter-
For v = c, we can write the voltage and the current on the line at minals can transfer thermal
t = 1 µs (by inspection) as noise energy to the TL. If

z − 150 3rect( z−150

300
) the resistors are disconnected
V (z, 1) = 3rect( ) V and I(z, 1) = A. at some point in time, the
300 Zo
energy left on the TL will
After t = 1 µs both ends of the TL will be open, and, therefore, only be shared between its res-
periodic waveforms with a onant modes (up to a fre-
2ℓ 1200 quency limit KT /~ imposed
– fundamental period of T = v = 300 = 4 µs and
by quantum mechanics) at an
2π
– fundamental frequency of ωo = T
= π2 Mrad/s average level of KT joules

will be allowed on the source free line. (per mode) where K is the
Boltzmann constant. Lossy
Therefore, V (z, t) and I(z, t) for t > 1 µs can be expressed as a weighted
lines with finite conductivity
superposition of the resonant modes of the line with resonant frequen-
also produce thermal noise.
cies nωo, subject to the initial conditions V (z, 1) and I(z, 1) given
Thermal noise is easy to
above.
detect and routinely inter-
feres with weak communi-
cation signals that we care
about!
9
32 Input impedance and microwave resonators Series Parallel

• The input impedance and admittance of the series and parallel

L
LC resonators shown in the margin are, respectively,
L C
1 1 C
Zs = j(ωL − ) and Yp = j(ωC − ),
ωC ωL
both of which vanish at the common resonance frequency of these net-
works, namely
1
ω=√ ≡ ωo .
LC
– Recall that LC resonators play an important role in the design of
filter and tuning circuits.

In this lecture we will examine the input impedance of microwave

resonators consisting of open or short circuited TL stubs.

• In the last lecture we learned that when a shorted stub is open circuited
at its input port, it shows resonance if the stub length ℓ is an odd
multiple of λ4 .

– The corresponding resonant frequencies are

v 1
f= ( + n) for n = 0, 1, 2, 3, · · ·
2ℓ 2

1
 and the input port of the stub coincides with a voltage max and
a current null, i.e., I(0, t) = 0 — thus the input impedance Zin of
the stub is infinite at these resonances, just like the impedance of
the parallel LC-circuit depicted above.
– Thus this set of resonant frequencies are referred to as parallel
resonances of the shorted stub.

• We also learned that when the stub length ℓ is an an integer multiple

of λ2 , its voltage at the input terminal is necessarily zero, implying that
the input impedance Zin must also be zero.

– The corresponding resonant frequencies are

v
f= n for n = 1, 2, 3, · · ·
2ℓ
and are termed series resonances of the shorted stub, in analogy
v
with the zero impedance of the series LC-circuit depicted above. 4ℓ
Parallel Resonances

0
v f
Series Resonances
2ℓ
short
The diagram in the margin marks the locations of parallel and series
resonance frequencies of the shorted stub associated with infinite and zero l
input impedance Zin . +
Thus, a shorted stub, included in a circuit such as the one shown in the + RL VL(t)
margin, will exhibit extreme behavior at these special frequencies — namely fi(t) - -

it will appear as a

2
 • short at its series resonances, causing the entire input signal fi (t) to
appear as VL (t) across the load RL , and v
Parallel Resonances
4ℓ

• open at its parallel resonances, causing VL (t) across the load RL to be 0

f
v Series Resonances
notched out. 2ℓ
short

We next focus our attention on how Zin of the stub appears at other l
frequencies not coinciding with any of the resonances discussed
+
above.
+ RL VL(t)
- -
• In the following we will assume that the TL stub, as well as the circuit fi(t)
it is connected to, are all in sinusoidal steady state at a frequency
determined by the frequency of the sinusoidal source fi (t).

– In that case d’Alembert solutions will also be co-sinusoidal at the

source frequency ω and we can express V (z, t) and I(z, t) on the
line as
z z
V (z, t) = Re{V +ejω(t− v )}+Re{V − ejω(t+ v )} ⇔ Ṽ (z) = V +e−jβz +V − ejβz

and
z z
Re{V +ejω(t− v )} − Re{V − ejω(t+ v )} ˜ = V + −jβz
e − V − jβz
e
I(z, t) = ⇔ I(z) ,
Zo Zo
where
ω
√
– β= v = ω LC is the wavenumber at frequency ω, and

3
 – V + and V − are phasors of forward and backward propagating
voltage waves on the line evaluated at z = 0.

We have expressed the phasor counterparts of co-sinusoidal waves V (z, t)

and I(z, t) above on the right, for it will be necessary to use phasors in
defining an input impedance — the impedance concept belongs to the
frequency domain!
GENERATOR, LOAD
• Before applying the boundary condition at the shorted end of the TL Circuit
stub, it will be convenient to shift the origin of our coordinate system Input
Short
to coincide with the shorted termination rather than the input port of port Zo

the TL.

• It will also be convenient to refer to “−z” as “d”, with the variable d −l l 0 z

growing to the left from the short termination toward the input terminal
of the line. d l 0

In that case, the input impedance of the shorted stub can be denoted
as
Ṽ (d = l)
Z(l) = ,
˜ = l)
I(d
where
+ jβd − −jβd ˜ V +ejβd − V − e−jβd
Ṽ (d) ≡ V e +V e and I(d) ≡ .
Zo
• An immediate benefit of our new notation comes when we apply the
voltage boundary condition at the short termination.

4
 GENERATOR, LOAD
– We apply it as Circuit

Input
V (0, t) = 0 ⇔ Ṽ (0) = V + + V − = 0 port Zo Short

from which it follows that

0 z
V − = −V +
−l l

and thus d l 0

Ṽ (d) ≡ V +(ejβd − e−jβd ) = j2V + sin(βd)

and
+ jβd −jβd
˜ ≡ V (e + e )
I(d) = Yo2V + cos(βd),
Zo
where
1
Yo ≡ Characteristic admittance.
Zo tanHΒlL

– Finally the input impedance of the shorted stub is 5

Ṽ (l)
Z(l) = = jZo tan(βl). Π Π 3Π 5Π 3Π 7Π
2Π
Βl

˜
Π
4 2 4 4 2 4

I(l)
-5

Note that: input impedance Z(l) = 0 + jX(l) is (see margin for X(l))
1. purely reactive for all l,

2. has a positive imaginary part and therefore it is inductive for

2π π λ
βl = l < rad = 90o ⇒ 0 < l < = Quarter wavelength.
λ 2 4
5
3. has a negative imaginary part and therefore it is capacitive for
π 2π λ λ
< βl = l < π rad = 180o ⇒ < l < = Half wavelength.
2 λ 4 2
λ
4. is periodic with a period of 2
over l, which means that
all possible reactive impedances of the form jX are realized for
0 < l < λ2 .

• a shorted TL stub of length 0 < l < λ2 spans all possible impedances

that can be provided by all possible inductors and capacitors!
λ
for a length l ≪ 4 shorted stub is a pure inductor with impedance
Open
input
L √
r
Z(l) = jZo tan(βl) ≈ jZoβl = j ω LCl = jωLl.
C Y (l) = 0 Zo Short
termn.

• here we used tan(βl) ≈ βl, which is valid when βl ≪ 1 in radians.

λ
l=
4
λ
5. at l = 4
the input admittance of the shorted stub,
1 1
Y (l) = = = −jYo cot(βl), Short
Z(l) jZo tan(βl) input

vanishes, meaning that Z(l) = 0 Zo Open

termn.

λ
• a shorted stub of length l = 4 appears at its input terminals like λ
l=
an open (see margin). 4

6
 6. at l = λ2 the input impedance of the shorted stub returns back to
zero, which in turn indicates, in view of (5), that
short
λ
• an open ended stub of length l = 4
must appear at its input l
terminals like a short (see margin).
+

RL VL(t)
Next set of examples illustrate the uses of shorted/opened TL stubs as circuit +
- -
elements. f i (t)

Amplitude Resp.
1.0
Example 1: A shorted TL stub of length l = 3 m is connected in series with a resistor
RL = 50 Ω as shown in the diagram in the margin. Plot the magnitude of the 0.8

frequency response H(ω) = ṼF̃L as a function of frequency f = 2πω

if Zo = 50 Ω 0.6

and v = c on the stub. Interpret the amplitude response curve |H(ω)| in terms 0.4

of resonance frequencies of the shorted line. 0.2

f HMHzL
ω
Solution: Using β = c and voltage division, we find that frequency response 50 100 150 200 250 300

ṼL RL 1
H(ω) = = = .
F̃i RL + jZo tan(βl) 1 + j tan( ωc l)
The plot of |H(ω)| with the given parameters is shown in the margin. The peaks
of the amplitude response occur at the series resonance frequencies of the shorted
stub when its input impedance is zero (an effective short). The nulls of the
amplitude response correspond to parallel resonances of the stub when it appears
like an open at its input terminals.

7
 Series network:

short
Example 2: Consider a shorted TL connected at d = l to an inductor L. Determine
the resonances of the combined network. L l
Solution: The input impedance of the shorted line is
√
Z(l) = jZo tan(βl) = jZo tan(ω LCl)
whereas inductor L has an impedance ZL = jωL. If the inductor and shorted stub
are connected in series (see margin), then the series resonances of the network
will be observed when the network input impedance
√ Parallel network:
ZL + Z(l) = jωL + jZo tan(ω LCl)
short
equals zero. The parallel resonances of the network will be observed when the
impedance is infinite. While the series resonance frequencies of the stub will be
shifted because of the inductor, parallel resonances will not shift (infinities due l
to tan function cannot be shifted by the finite additive term due to the inductor).
The shifted series resonance frequencies ωn can be found graphically by plotting
ZL + Z(l) and looking for the zero crossings.
L
If the inductor and shorted stub are connected in parallel, then the parallel resonances
of the network will be observed when the network input admittance
1 1
YL + Y (l) = + √
jωL jZo tan(ω LCl)
equals zero (same as infinite input impedance). The series resonances, on the
other hand, will be observed when the admittance is infinite (same as zero input
impedance). Series resonances of the stub will not be shifted with, unlike its
parallel resonances. The shifted parallel resonance frequencies ωn will equal the
series resonance frequencies of the series connected network described above.

8
 short

l
Example 3: A shorted TL stub of length l = 3 m is connected in series with a a +
C
capacitor C = 10 pf and a resistor RL = 50 Ω as shown in the diagram in the RL VL(t)
+
margin. Plot the magnitude of the frequency response H(ω) = ṼF̃L as a function - -
ω fi(t)
of frequency f = 2π if Zo = 50 Ω and v = c on the stub. Interpret the amplitude
Can change the short
response curve |H(ω)| in terms of resonance frequencies of the shorted line. location on the stub
to tune the filter
ω
Solution: Using β = c and voltage division, we find that frequency response

ṼL RL 1
H(ω) = = = . Amplitude Resp.

F̃i RL + 1
jωC + jZo tan(βl) 1 + 1
jωRL C + j tan( ωc l) 1.0

0.8

The plot of |H(ω)| with the given parameters is shown in the margin. The peaks 0.6
of the amplitude response occur at the shifted series resonance frequencies of the
0.4
shorted T.L. stub. The nulls of the amplitude response correspond to parallel
resonances of the stub when it appears open. 0.2

f HMHzL
50 100 150 200 250 300

9
Example 4: (a) If in the TL circuit shown in the margin IR = 2∠0o A, what is the
line length l in terms of wavelength λ of the given source frequency on the line?

(b) Repeat for IR = 0.

Note: starting in this example we are dropping the tildes on

the phasors.

Solution: (a) If IR = 2∠0o A, then KCL application at the source terminal implies
that I(l) = 0.

In that case the TL has an open at d = l. Since d = 0 is also an open,

we need to have l to be an integer multiple of λ2 .

In other words, the condition of IR = 2∠0o will only be realized in the

above circuit if the source frequency is such that the TL length l happens to be
some integer multiple of λ2 at the given frequency.

(b) If IR = 0, then V (l) = (50 Ω)IR = 0, implying that the T.L. has a short at d = l. I(l) I(0) = 0

IR + +
26 V (l)
Since d = 0 is an open, we need to have l some odd multiple of λ4 .
0A V (0)
50 Ω - -

d
l 0

10
33 TL circuits with half- and quarter-wave trans-
formers GENERATOR, LOAD
Circuit I(d)

• Last lecture we established that phasor solutions of telegrapher’s equa- Input

+
V (d)
port Zo
tions for TL’s in sinusoidal steady-state can be expressed as - ZL = RL + jXL

+ jβd − −jβd V +ejβd − V −e−jβd 0 z

V (d) = V e + V e and I(d) = −l l
Zo
in a new coordinate system shown in the margin. d l 0

By convention the load is located on the right, at z = 0 = d, and the

TL input connected to a generator or some source circuit is shown on
the left at d = l.

We have replaced the short termination of the previous lecture with an

arbitrary load impedance
ZL = RL + jXL.

In this lecture we will discuss sinusoidal steady-state TL cir-

cuit problems having arbitrary reactive loads but with line
lengths l constrained to be integer multiples of λ4 (at the op-
eration frequency).
The constraint will be lifted next lecture when we will de-
velop the general analysis tools for sinusoidal steady-state TL
circuits.

1
• In the TL circuit shown in the margin an arbitrary load ZL is connected
to a TL of length l = λ2 at the source frequency.
Given that
λ 2π λ
e±jβ 2 = e±j λ 2 = e±jπ = −1, Half-wave transformer:

the general phasor relations Iin IL = −Iin

+ +
V +ejβd − V −e−jβd Zo
V (d) = V +ejβd + V −e−jβd and I(d) = Vin ZL VL = −Vin
Zo - -
λ 0
imply 2
λ
Vin ≡ V ( ) = −V + − V − = −V (0) = −VL,
2
λ −V + + V −
Iin ≡ I( ) = = −I(0) = −IL.
2 Zo
We conclude that a λ2 -transformer

– inverts the algebraic sign of its voltage and current inputs at the
load end (and vice versa), and
– has an input impedance identical with the load impedance since
Vin −VL
Zin ≡ = = ZL.
Iin −IL
These very simple results are easy to remember and use.

2
• In the TL circuit shown in the margin an arbitrary load ZL is connected
to a TL of length l = λ4 at the source frequency. Quarter-wave transformer:

Given that Vin

±jβ λ4 ±j 2π λ ±j π2 Iin IL = −j
e =e λ 4 =e = ±j, Zo
+ +
general phasor relations Vin ZL VL = −jIin Zo
- Zo -
+ jβd − −jβd
V e −V e λ 0
V (d) = V +ejβd + V −e−jβd and I(d) = 4
Zo
imply ZinZL = Zo2
λ
Vin ≡ V ( ) = jV + − jV − = jI(0)Zo = jIL Zo,
4
λ jV + + jV − V (0) VL
Iin ≡ I( ) = =j =j .
4 Zo Zo Zo
We conclude that a λ4 -transformer Quarter-wave
current-forcing
– has an input impedance
equation:
Vin jIL Zo Zo2 Zo2
Zin ≡ = = = , Vin
Iin jVL/Zo VL/IL ZL IL = −j .
Zo
– and provides a load current
Vin Load voltage
IL = −j ,
Zo
VL = ZLIL
proportional to input voltage Vin but independent of load impedance
ZL. once IL is available
from above equation.
3
 Iin
+ +
Vin ZL = 50 + j50 Ω
λ
Example 1: Given ZL = 50 + j50 Ω, what is Zin for a 4
transformer with Zo = 50 Ω? Z
- o = 50 Ω -
λ 0
Solution: It is 4 Zo2
Zo2 502 50 50 1 − j1 Zin =
ZL
Zin = = = = = 25 − j25 Ω.
ZL 50 + j50 1 + j1 1 + j1 1 − j1
λ
Notice that an inductive ZL has been turned into a capacitive Zin by 4 trans-
former.

+
+ 50 + j50 Ω
Vg = 1006 0 V -
Example 2: The load and the transformer of Example 1 are connected to a source Zo = 50 Ω -
with voltage phasor Vg = 100∠0o V at the input port. What is the load current λ 0
IL and what is the average power absorbed by the load? 4

Solution: Since Vin = Vg = 100∠0o V, the current-forcing formula for the quarter-wave
transformer implies
Vin 100
IL = −j = −j = −j2 A.
Zo 50
To find the average power absorbed, we first note that load voltage
VL = ZLIL = (50 + j50)(−j2) = 100 − j100 V.
Thus,
1 1
PL = Re{VL IL∗ } = Re{(100 − j100)(j2)} = 100 W.
2 2

4
 I(d)
Zg = 25 Ω
Vg = +
ZL =
V (d)
j10 V +- Zo = 50 Ω
100 Ω
-

Example 3: Load ZL = 100 Ω is connected to a T.L. with length l = 0.75λ. At the

d 0.75λ 0.5λ 0.25λ 0
generator end, d = 0.75λ, a source with open circuit voltage Vg = j10 V and
Thevenin impedance Zg = 25 Ω is connected. Determine VL and IL if Zo = 50 Ω.

Solution: First we determine input impedance Zin by noting that ZL = 100 Ω trans-
forms to itself, namely 100 Ω at d = 0.5λ, but then it transforms from d = 0.5λ
to 0.75λ as
Zo2 502
Zin = = = 25 Ω.
Z(0.5λ) 100
Hence, using voltage division, we find,
Zin 25
Vin = Vg = j10 = j5 V.
Zg + Zin 25 + 25
Next, using half-wave transformer rule, we notice that

V (0.25λ) = −Vin = −j5 V,

and finally applying the quarter-wave current forcing equation with V (0.25λ) we
get
V (0.25λ) −j5
IL = −j = −j = −0.1 A.
Zo 50
Clearly, then, the load voltage is

VL = ZL IL = (100 Ω)(−0.1 A) = −10 V.

5
 Vin
IL2 = −j
Zo

Example 4: In the circuit shown in the margin, ZL1 = 50 Ω, ZL2 = 100 Ω, and Zo1 = ZL2
Zo2 Vin
Zo2 = 50 Ω. Determine IL1 and IL2 if Vin = 5 V. Both T.L. sections are quarter- IL1 = −j
Zo
wave transformers.
+
Vin Zo1 ZL1
Solution: Using the current-forcing equation, we have -
Vin 5 λ 0
IL1 = IL2 = −j = −j = −j0.1 A. 4
Zo 50
Consequently,
VL1 = IL1ZL1 = −j0.1 A × 50 Ω = −j5 V
and
VL2 = IL2ZL2 = −j0.1 A × 100 Ω = −j10 V.
Thus, total avg power absorbed is
1 ∗ 1
P = Re{VL1IL1 } + Re{VL2I ∗L2}
2 2
1 1
= = Re{−j5 × j0.1} + Re{−j10 × j0.1} = 0.75 W.
2 2

6
34 Line impedance, generalized reflection coef- LOAD
GENERATOR,
I(d)
ficient, Smith Chart Circuit
+
Input
port Zo V (d)
• Consider a TL of an arbitrary length l terminated by an arbitrary load - ZL = RL + jXL

ZL = RL + jXL. −l l 0 z

as depicted in the margin. d l 0

Voltage and current phasors are known to vary on the line as

+ jβd − −jβd V +ejβd − V − e−jβd

V (d) = V e +V e and I(d) = .
Zo
In this lecture we will develop the general analysis tools needed
to determine the unknowns of these phasors, namely V + and
V − , in terms of source circuit specifications.

• Our analysis starts at the load end of the TL where V (0) and I(0)
stand for the load voltage and current, obeying Ohm’s law

V (0) = ZL I(0).

Hence, using V (0) and I(0) from above, we have

+ − V+−V− ZL − Zo +
V +V = ZL ⇒ V− = V .
Zo ZL + Zo
1
– Define a load reflection coefficient “Load reflection coefficient”
ZL − Zo is a well justified name for
ΓL ≡
ZL + Zo ΓL since the forward travel-
ing wave with phasor V + ejβd
and re-write the voltage and current phasors as
gets reflected from the load.
+ jβd −j2βd V +ejβd [1 − ΓLe−j2βd]
V (d) = V e [1 + ΓLe ] and I(d) = .
Zo
– Define a generalized reflection coefficient The term “generalized reflec-
tion coefficient” is also well
Γ(d) ≡ ΓLe−j2βd
justified even if there is no
and re-write the voltage and current phasors as reflection taking place at ar-
bitrary d — the reason is, if
+ jβd V +ejβd[1 − Γ(d)]
V (d) = V e [1 + Γ(d)] and I(d) = . the line were cut at location
Zo
d and the stub with the load
– Line impedance is then defined as were replaced by a lumped
V (d) 1 + Γ(d) load having a reflection co-
Z(d) = = Zo
I(d) 1 − Γ(d) efficient equal to Γ(d), then
there would be no modifica-
for all values of d on the line extending from the load point d = 0
tion of the voltage and cur-
all the way to the input port at d = l.
rent variations on the line to-
wards the generator.
With the dependence on d of Z(d) as well as Γ(d) tacitly implied,
we can re-write this important relation and its inverse as Each location d on the line

Z 1+Γ Z − Zo has an impedance Z and a re-

= ⇔ Γ= . flection coefficient Γ linked by
Zo 1 − Γ Z + Zo
these equations.
2
Properties of Z(d) = R(d) + jX(d) and Γ(d) = ΓLe−j2βd linked by the
relations
Z 1+Γ Z − Zo
= ⇔ Γ= :
Zo 1 − Γ Z + Zo
1. For real valued Zo and R(d) ≥ 0 , |Γ(d)| ≤ 1:
Verification:
p
|Z − Zo| |(R − Zo) + jX| (R − Zo)2 + X 2
|Γ| = = =p .
|Z + Zo | |(R + Zo ) + jX| (R + Zo)2 + X 2
Since with R ≥ 0
p p
(R − Zo) + X ≤ (R + Zo )2 + X 2 ⇒ |Γ| ≤ 1.
2 2

2. Since
|Γ| = |ΓL| and ∠Γ(d) = ∠ΓL − 2βd
property (1) implies that Γ(d) is a complex number which is constrained
to be on or within the unit-circle on the complex plane.

3. Relationships
Z 1+Γ Z − Zo
= ⇔ Γ=
Zo 1 − Γ Z + Zo
between Γ and Z are known as bilinear transformations — here the
term bilinear refers to the numerator as well as the denominator of
these transformations being linear in the variable being transformed
(from right to left).

3
 Bilinear (or Möbius) transformations are known to have the general
property of mapping straight lines into circles on the complex num-
ber plane.

• Bilinear transformations between Im

r=const.

Γ ≡ Γr + jΓi ≡ (Γr , Γi) z=r+jx

and
Z Unit

≡ z ≡ r + jx, Radius
Circle
Zo
known as normalized impedance, lead to an ingenious graphical aid Re

known as the Smith Chart. x=const.

– On a Smith Chart (SC), straight lines on the right hand side of

the complex number plane (see margin), represented by
r = const. and x = const.,
are mapped onto circular loci of
Z − Zo z − 1
(Γr , Γi) = Γ = =
Z + Zo z + 1
occupying the region of the plane bordered by the unit circle.

Circles corresponding to z = const. + jx and z = r + jconst.constitute

a griding of the unit circle and its interior. By means of this grid,
the normalized impedance z corresponding to every possible Γ can be
directly read off the SC.

4
• SC can be constructed by first noting that
z − 1 r + jx − 1 [(r − 1) + jx][(r + 1) − jx] (r2 + x2 − 1) + j2x
Γ= = = = ≡ Γr +jΓi ;
z + 1 r + jx + 1 (r + 1)2 + x2 (r + 1)2 + x2
thus Γi
2 2
(r + x − 1) 2x SmithChart
Γr = and Γ i = , 1

(r + 1)2 + x2 (r + 1)2 + x2 .5 2

and by direct substitution we can verify the following equations

.2 x=5

r 2 1 2 1 1 2. + 1. ä

(Γr − ) + Γ2i = ( ) and (Γr − 1)2 + (Γi − )2 = ( )2 80.447214, 0.0368959<

r+1 r+1 x x 0
-1
.2 .5

0.4 - 0.2 ä
1 2 2.61803 r=5
1
Γr
describing r and x dependent circles, respectively, on complex plane -.2 x=-5

constituting the grid lines of the SC.

-.5 -2

• Typical SC usage: -1

Smith Chart is the unit cir-

cle and its interior on the
1. Locate and mark z(0) — normalized load impedance — on the complex number plane corre-
sponding to the generalized
SC, which places you at a distance |Γ(0)| = |ΓL| from the origin reflection coefficient Γ. The
of the complex plane (and the SC), at an angle of θ = ∠Γ(0). gridding allows direct identi-
fication of the bilinear trans-
2. Draw a constant |Γ| = |ΓL| circle with a compass going through form of Γ, namely the nor-
malized line impedance z.
point z(0) on the SC (the read circle in the margin). Rotate clock-
wise on the circle by an angle of
4π d
2βd = d rad = 360◦
λ λ/2
to land on z(d) that can be read off using the SC gridding.

5
 SmithChart
– Rotation by an angle of 2βd amounts to 1

rotation by full circle for d = λ2 , .5 2

rotation by half circle for d = λ4 ,

rotation by quarter circle for d = λ8 , etc. .2
2. + 1. ä
x=5

3. Also, 0 .2 .5 1
80.447214, 0.0368959<
2 2.61803 r=5

1 0.4 - 0.2 ä

y(d) ≡
z(d) -.2 x=-5

which is the normalized line admittance is located on the SC

on the constant |Γ| = |ΓL| circle across the point corresponding to -.5 -2

z(d). -1

Verification: Since
1+Γ 1 1 − Γ 1 + (−Γ)
z= ⇒ y= = = ;
1−Γ z 1 + Γ 1 − (−Γ)
hence whereas z is the transform of Γ, y is the transform of −Γ,
having the same magnitude as Γ but an angle off by ±180◦.
– Therefore, “reflect” on the SC across the origin to jump
from z(d) to y(d) if you need the value of the normalized
admittance.

Our first SC example is given next.

6
 I(d)

+
Zo = 50 Ω V (d) ZL = 50 + j100 Ω
-
Example 1: A transmission line is terminated by an inductive load of
d 0.125λ 0
ZL = 50 + j100 Ω.

Determine the input impedance Zin = Z(l) of the line at a distance

(a) At load point:
λ SmithChart
1
d=l=
8 .5 2

if the characteristic impedance of the line is Zo = 50 Ω. Also determine the 1. + 2. ä

normalized input admittance y(l). .2 80.707107, 0.0625< x=5

Solution: The normalized load impedance is 0 .2 .5 1 2 r=5

5.82843

ZL 50 + j100
z(0) = = = 1 + j2.
Zo 50 -.2 0.2 - 0.4 ä x=-5

λ
Enter z(0) on the SC and then rotate clockwise by 8 ⇔(quarter circle) to obtain
the normalized input impedance
-.5 -2

-1

z(l) = 1 − j2, (b) at input point:

SmithChart
1

and the normalized input admittance .5 2

y(l) = 0.2 + j0.4 0.2 + 0.4 ä

.2 x=5

right across z(l). The input impedance is

.2 .5 1 2 r=5
Zin = Zo z(l) = 50(1 − j2) = 50 − j100 Ω. 0 5.82843

-.2 1. - 2. ä x=-5

80.707107, -0.0625<

-.5 -2

7 -1
Blow up of the SC’s used in Example 1:
SmithChart SmithChart
1 1

.5 2 .5 2

1. + 2. ä 0.2 + 0.4 ä

.2 80.707107, 0.0625< x=5 .2 x=5

0 .2 .5 1 2 r=5
5.82843 0 .2 .5 1 2 r=5
5.82843

-.2 0.2 - 0.4 ä x=-5 -.2 1. - 2. ä x=-5

80.707107, -0.0625<

-.5 -2 -.5 -2

(a) At load point -1 (b) at input point -1

• A SmithChartTool linked from the class calendar (a javascript util-

ity that requires a Safari or Firefox browser to work properly) marks
and prints z(d) in red and y(d) in magenta across from z(d) on the
constant-|ΓL | circle (shown in red) as in the above examples. Also
– printed in black is the real valued normalized impedance z(dmax ) discussed in
the upcoming lectures (also known as VSWR).
– also printed in red is |ΓL |∠Γ(d) where the second entry is expressed in terms
of an equivalent λd such that λd = 0.5 corresponds to an angle of 360◦ . This
way of referring to ∠Γ(d) will be convenient in many SC applications that we
will see.

8
35 Smith Chart examples .5
SmithChart
1

.2 x!5
2. # 1. $

!0.447214, 0.0368959"
Example 1: A load ZL = 100 + j50 Ω is connected across a TL with Zo = 50 Ω and 0 .2 .5 1 2 2.61803 r!5

l = 0.4λ. At the generator end, d = l, the line is shunted by an impedance 0.4 " 0.2 $

Zs = 100 Ω. What are the input impedance Zin and admittance Yin of the line, ".2 x!"5

including the shunt connected element.

".5 "2
Solution: Normalized load impedance "1
SmithChart
ZL 100 + j50 1

z(0) = = = 2 + j1
Zo 50 .5 2

is entered in the SC shown in the margin on the top. Clockwise rotation (from 0.600056 # 0.663401 $

!0.447214, 0.136896"
load toward generator) at fixed |Γ| (red circle) by
.2 x!5

0.4λ ⇔ 0.8 × 360◦ = 288◦ 0 .2 .5 1 2 2.61803 r!5

takes us to ".2
0.749913 " 0.829077 $
x!"5

z(l) ≈ 0.6 + j0.66 and y(l) ≈ 0.75 − j0.83

as shown on the SC in the middle. Hence, including the shunt element with ".5 "2

normalized input impedance zsi = 2 and admittance ysi = 12 , we obtain "1

SmithChart
1

yin = y(l) + ysi ≈ 1.25 − j0.83 .5 2

for the overall normalized input admittance of the shunted line as shown on the
SC in the bottom — the corresponding normalized input impedance is .2 0.555603 # 0.368536 $ x!5

!0.361132, 0.176393"
1
zin = ≈ 0.56 + j0.37. 0 .2 .5 1 2
2.13054 r!5

yi 1.24991 " 0.829077 $

Hence, the unnormalized input impedance and admittance are ".2 x!"5

Zin = Zo zin ≈ 27.8 + j18.4 Ω and Yin = Yo yin ≈ 0.025 − j0.017 S.

".5 "2

"1

1
 I(d)
Zg = 25 Ω
Vg = +
Zs = 100 Ω ZL =
100! 0 V +- V (d)
100 + j50 Ω
Example 2: The TL network described in Example 1 is connected to a generator with
Zo = 50 Ω
-

open circuit voltage phasor Vg = 100∠0 V and internal impedance Zg = 25 Ω.

What is the average power (a) input of the shunted line, (b) delivered to the d 0.4λ 0

shunt element, delivered to the load.

Solution:

(a) Using the input impedance

Zin ≈ 27.8 + j18.4 Ω,
from Example 1, we can write
Zin Vg
Vin = Vg and Iin = .
Zg + Zin Zg + Zin
Therefore, the average power input of the shunted line is
1 ∗ 1 Vg Zin Vg
P = Re{Vin Iin } = Re{ ( )∗ }
2 2 Zg + Zin Zg + Zin
|Vg |2 1002
= Re{Zin } = 27.8 ≈ 44.44 W.
2|Zg + Zin |2 2|25 + 27.8 + j18.4|2

(b) The shunt element Zs = 100 Ω sees the same voltage Vin and conducts a current
Vin /Zs . Therefore it absorbs an average power of
1 Vin ∗ |Vin |2 |Vg Zin|2
P = Re{Vin ( ) } = =
2 Zs 2Zs 2Zs|Zg + Zin |2
|100 · (27.8 + j18.4)|2
≈ ≈ 17.78 W.
2 · 100 · |25 + 27.8 + j18.4|2
The remainder of 44.44 W will be absorbed in ZL .

2
 I(d) SmithChart
1

+ .5 2
Zin = 50 + j50 Ω V (d)
Zo = 50 Ω
ZL =?
- 1. # 1. $
.2 x!5
!0.447214, 0.0881041"

d 0.3λ 0 .2 .5 1 2 2.61803 r!5

0

0.5 " 0.5 $

".2 x!"5

Example 3: A TL of length l = 0.3λ has an input impedance Zin = 50 + j50 Ω.

Determine the load impedance ZL = Z(0) and YL = Y (0) given that Zo = 50 Ω
for the line. ".5 "2

"1

Solution: First enter the normalized inpur impedance SmithChart

1

Zin 50 + j50 .5 2

zin = = =1+j
Zo 50 0.594079 # 0.655216 $

in the SC as shown in the margin on the top. Counter-clockwise rotation (from .2 x!5

generator toward load) at fixed |Γ| (red circle) by

0 .2 .5 1 2 2.61803 r!5

0.3λ ⇔ 0.6 × 360◦ = 216◦

takes us to ".2
0.759461 " 0.837617 $
x!"5

!0.447214, "0.111896"
z(0) ≈ 0.76 − j0.84 and y(0) ≈ 0.59 + j0.66
as shown on the next SC at the load point. Hence, we find ".5 "2

ZL = Zo z(0) ≈ 50 · (0.76 − j0.84) = 37.97 − j41.88 Ω "1

and
1
YL = Yo y(0) ≈ (0.59 + j0.66) = 0.012 + j0.013 S.
50

3
 SmithChart
ΓLs = −0.5 1

.5 2

Zo = 50 Ω
.2 x!5
I(d)

0.333333 3.
+ .2 .5 1 2 3. r!5
!0.5, 0"
0

V (d)
Zin =? Zo = 50 Ω
ΓL = 0.5
- ".2 x!"5

d 0.5λ 0.2λ 0 ".5 "2

"1
SmithChart
1

.5 2

Example 4: A TL of length l = 0.5λ and Zo = 50 Ω has a load reflection coefficient

ΓL = 0.5 and and a shunt connected TL at d = 0.2λ. The shunt connected TL .2 1.70075 # 1.3329 $ x!5

has l = 0.3λ, Zo = 50 Ω, and a load reflection coefficient ΓL = −0.5. Determine

the input impedance of the line. 0 .2 .5 1 2 3. r!5

0.364251 " 0.285469 $

Solution: Recall that the SC covers the unit circle of the complex plane and therefore ".2 !0.5, "0.2" x!"5

the complex number

ΓL = 0.5 + j0 = 0.5 ".5 "2

can be entered directly in the SC as shown on the top SC in the margin. Clockwise "1
SmithChart
rotation (from load toward generator) at fixed |Γ| (red circle) by 1

.5 2

◦ ◦
0.2λ ⇔ 0.4 × 360 = 144
takes us to
.2 x!5

0.333333 3.

z(0.2λ) ≈ 0.36 − j0.29 and y(0.2λ) ≈ 1.7 + j1.33 0 .2

!0.5, 0.25"
.5 1 2 3. r!5

as shown on the SC in the middle. Likewise, entering ".2 x!"5

ΓLs = −0.5 + j0 = −0.5

".5 "2

"1

4
 SmithChart
1

.5 2

.2 0.364251 # 0.285469 $ x!5

for the shunt connected stub in the third SC and rotating clockwise by
.2 .5 1 2 3. r!5
0.3λ ⇔ 0.6 × 360◦ = 216◦
0

1.70075 " 1.3329 $

!0.5, "0.05"
we obtain ".2 x!"5

zs (0.3λ) ≈ 1.7 − j1.33 and ys (0.3λ) ≈ 0.36 + j0.29. ".5 "2

"1

We proceed by combining the normalized admittances as SmithChart

1

.5 2

yc = y(0.2λ) + ys (0.3λ) ≈ (1.7 + j1.33) + (0.36 + j0.29) = 2.065 + j1.61837,

and entering it in the next SC. Finally rotating clockwise once again by .2
2.065 # 1.61837 $
x!5

0.3λ ⇔ 0.6 × 360◦ = 216◦ 0 .2 .5 1 2 r!5

3.53468

0.3 " 0.235114 $

we obtain, from the last SC !0.558955, "0.209976"

".2 x!"5

zin ≈ 3.38 − j0.69 ⇒ Zin = zin Zo ≈ 169 − j34.4 Ω.

".5 "2

"1
SmithChart
1

.5 2

.2 x!5

0.283936 # 0.0577194 $

0 .2 .5 1 2 3.38216 " 0.687537 $

r!5
3.53468

!0.558955, "0.00997561"

".2 x!"5

".5 "2

"1

5
Example 5: What is the load impedance ZLs terminating the shunt connected stub
in Example 4?

Solution: Given that the corresponding reflection coefficient is

ΓLs = −0.5,

it follows from the bilinear transformation linking zLs and ΓLs that
1 + ΓLs 1 − 0.5 1
zLs = = = .
1 − ΓLs 1 + 0.5 3
Hence, the impedance is
50
ZLs = Zo zLs = Ω.
3

Example 6: What is the load impedance ZL in Example 4?

Solution: This is similar to Example 5. Given that the load reflection coefficient is
ΓL = 0.5,
it follows from the bilinear transformation linking zL and ΓL that
1 + ΓL 1 + 0.5
zL = = = 3.
1 − ΓL 1 − 0.5
Hence, the impedance is
ZL = Zo zL = 150 Ω.

6
36 Smith Chart and VSWR Generator
I(d)
Transmission line
Wire 2
+ Load
Zg
• Consider the general phasor expressions
+ V (d) Zo
- ZL
F = Vg - Wire 1
0
V +ejβd(1 − ΓLe−j2βd )
l d
+ jβd
V (d) = V e (1 + ΓLe −j2βd
) and I(d) = |V (d)|max
|V (d)|
Zo
describing the voltage and current variations on TL’s in sinusoidal |V (d)|min dmax

steady-state. dmin

– Unless ΓL = 0, these phasors contain reflected components, which Complex addition displayed
graphically superposed on a
means that voltage and current variations on the line “contain” Smith Chart

standing waves. SmithChart

1

.5 2

In that case the phasors go through cycles of magnitude variations as a

function of d, and in the voltage magnitude in particular (see margin)
Γ(d) z(dmax )
.2 x!5
)
varying as 1 + Γ (d =VSWR
0 .2 .5 1 2 VSWR r!5
+ −j2βd +
|V (d)| = |V ||1 + ΓLe | = |V ||1 + Γ(d)| 1 Γ(dmax ) = |ΓL|

takes maximum and minimum values of ".2 x!"5

|V (d)|max = |V +|(1 + |ΓL|) and |V (d)|min = |V +|(1 − |ΓL|) ".5 "2

"1

at locations d = dmax and dmin such that |1 + Γ(d)| maximizes for d = dmax
Γ(dmax) = ΓLe−j2βdmax = |ΓL| and Γ(dmin) = ΓLe−j2βdmin = −|ΓL|,
|1 + Γ(d)| minimizes for d = dmin
and such that Γ(dmin) = −Γ(dmax)
λ
dmax − dmin is an odd multiple of .
4

1
 – These results can be most easily understood and verified graphi- Generator Transmission line
I(d) Wire 2

cally on a SC as shown in the margin. + Zg

+
V (d) Zo
Load

- ZL
F = Vg - Wire 1
0
• We define a parameter known as voltage standing wave ratio, or
l d
|V (d)|max

VSWR for short, by |V (d)|

VSWR − 1
|V (d)|min
|V (dmax )| 1 + |ΓL| dmax
VSWR ≡ = ⇔ |ΓL| = .
VSWR + 1
dmin
|V (dmin)| 1 − |ΓL|
Notice that the VSWR and |ΓL| form a bilinear transform pair just Complex addition displayed
graphically superposed on a
like Smith Chart
1+Γ z−1
z= ⇔ Γ= . SmithChart
1
1−Γ z+1 .5 2

Since
1 + Γ(dmax)
Γ(dmax ) = |ΓL| ⇒ VSWR = , .2
) Γ(d) z(dmax )
x!5

1 − Γ(dmax ) 1 + Γ (d =VSWR

this analogy between the transform pairs also implies that 0 .2 .5 1 2 VSWR r!5

1 Γ(dmax ) = |ΓL|
z(dmax ) = VSWR, ".2 x!"5

as explicitly marked on the the SC shown in the margin . Consequently, ".5 "2

"1

– the VSWR of any TL can be directly read off from its SC plot |1 + Γ(d)| maximizes for d = dmax

as the normalized impedance value z(dmax ) on constant-|ΓL | circle

|1 + Γ(d)| minimizes for d = dmin
crossing the positive real axis of the complex plane. such that Γ(dmin) = −Γ(dmax)

2
• The extreme values the VSWR can take are:

1. VSWR=1 if |ΓL| = 0 and the TL carries no reflected wave.

2. VSWR=∞ if |ΓL| = 1 corresponding to having a short, open, or
a purely reactive load that causes a total reflection.

SmithChart
1

.5 2

.2
z(d x!5)
|1 + Γ(d)| maximizes for d = dmax
Γ ( d) Γ(d) max
1+ =VSWR
0 .2 .5 1 2 VSWR r!5

1 Γ(dmax ) = |ΓL|
x!"5
".2
|1 + Γ(d)| minimizes for d = dmin
such that Γ(dmin) = −Γ(dmax)
".5 "2

"1

3
 • In the lab it is easy and useful to determine the VSWR and dmax or Generator Transmission line
I(d) Wire 2

dmin of a TL circuit with an unknown load, since + Zg

+
V (d) Zo
Load

- ZL
F = Vg - Wire 1

1. given the VSWR, l d 0

VSWR − 1 |V (d)|max
|V (d)|
|ΓL| =
VSWR + 1
|V (d)|min
is easily determined, and
dmax

dmin

2. given dmax or dmin the complex ΓL or its transform zL can be easily

Complex addition displayed
obtained. graphically superposed on a
Smith Chart

SmithChart
1

Say dmax is known: then, .5 2

• since (as we have seen above)

Γ(d) z(dmax )
.2 x!5
)
1 + Γ (d =VSWR
Γ(dmax ) = ΓLe−j2βdmax = |ΓL| .2 .5 1 2 VSWR r!5
0

1 Γ(dmax ) = |ΓL|
it follows that
".2 x!"5

1 + ΓL
ΓL = |ΓL|ej2βdmax ⇒ zL = .
1 − ΓL ".5 "2

"1

• alternatively, zL can be obtained directly on the SC by rotating counter- |1 + Γ(d)| maximizes for d = dmax
clockwise by dmax from the location of
|1 + Γ(d)| minimizes for d = dmin
z(dmax ) = VSWR. such that Γ(dmin) = −Γ(dmax)

These techniques are illustrated in the next example.

4
 SmithChart
1

.5 2

Example 1: An unknown load ZL on a Zo = 50 Ω TL has

.2 x!5

V (dmin) = 20 V, , dmin = 0.125λ and VSWR=4.

0.25 4.
0 .2 .5 1 2 r!5
4.
Determine (a) the load impedance ZL , and (b) the average power PL absorbed !0.6, 0.25"

by the load.
".2 x!"5

Solution:

(a) As shown in the top SC in the margin, VSWR=4 is entered in the SC as ".5 "2

z(dmax ) = 4 + j0, and constant |ΓL | circle is then drawn (red circle) passing "1
SmithChart
through z(dmax ) = 4. 1

.5 2

Right across z(dmax) = 4 on the circle is z(dmin) = 0.25. 0.470588 $ 0.882353 #

A counter-clockwise rotation from z(dmin) = 0.25 by one fourth of a full circle corre- .2 x!5

sponding to a displacement of dmin = 0.125λ (a full circle corresponds to a λ/2

displacement) takes us to 0 .2 .5 1 2 r!5
4.

zL ≈ 0.4706 − j0.8823
".2 x!"5

as shown in the second SC. Hence, this gives 0.470588 " 0.882353 #

!0.6, "0.125"

ZL = Zo zL = 50(0.4706 − j0.8823) = 23.53 − j44.12 Ω. ".5 "2

"1

(b) We will calculate PL by using V (dmin) and I(dmin). Since

1
z(dmin) = 0.25 it follows that Z(dmin) = 50 Ω = 12.5 Ω.
4

5
 Therefore the voltage and current phasors at the voltage minimum location are
20 V
V (dmin) = 20 V and I(dmin) = .
12.5 Ω
Generator Transmission line
Average power transported toward the load at d − dmin is, therefore, I(d) Wire 2
+ Load
1 1 20 400 Zg
P (dmin ) = Re{V (dmin)I(dmin)∗} = Re{20 }= W = 16 W. +
- V (d) Zo ZL
2 2 12.5 25 F = Vg - Wire 1
l d 0
Since the TL is assumed to be lossless we should have |V (d)|max
|V (d)|
PL = P (dmin ) = 16 W.
|V (d)|min dmax

dmin

Example 2: If the TL circuit in Example 1 has l = 0.625λ, and a generator with an

internal impedance Zg = 50 Ω, determine the generator voltage Vg .

Solution: Given that l = 0.625λ and dmin = 0.125λ, we note that there is just one
half-wave transformer between l = 0.625λ and dmin = 0.125λ. Therefore
Vin = −V (dmin) = −20 V and Zin = Z(dmin) = 12.5 Ω.
But also
Zin
Vin = Vg .
Zg + Zin
Consequently,
Zg + Zin 50 + 12.5 62.5
Vg = Vin = −20 = −20 = −100 V.
Zin 12.5 12.5

6
 SmithChart
1

.5 2

Example 3: Determine V + and V − in the circuit of Examples 1 and 2 above such that .2 x!5
the voltage phasor on the line is given by
0.25 4.
V (d) = V + ejβd + V − e−jβd . 0 .2 .5 1 2 r!5
4.
!0.6, 0.25"

Solution: Looking back to Example 1 (also see the SC’s in the margin), we first note ".2 x!"5

that
VSWR − 1 4 − 1
|ΓL | = = = 0.6 = Γ(dmax) = −Γ(dmin).
VSWR + 1 4+1 ".5 "2

Hence, evaluating V (d) at d = dmin , we have "1

SmithChart
1
+ jβdmin
V (dmin) = V e (1 + Γ(dmin)) .5 2
λ
+ j 2π j π4 +
= V (e λ 8 )(1 + (−0.6)) = 0.4e V = 20 V, 0.470588 $ 0.882353 #

from which .2 x!5

+ −j π4
V = 50e V.
.2 .5 1 2 r!5
Since 0 4.

j π2
ΓL = Γ(0) = Γ(dmin)ej2βdmin = −0.6e ,
it follows that ".2
0.470588 " 0.882353 #
x!"5

π π π
V − = ΓL V + = −0.6ej 2 × 50e−j 4 = −30ej 4 V. !0.6, "0.125"

".5 "2

"1

7
 SmithChart
1

.5 2

Example 4: Determine the load voltage and current VL = V (0) and IL = I(0) in the .2 x!5
circuit of Examples 1-3 above.
0.25 4.
Solution: In general, 0 .2 .5 1 2 r!5
4.
!0.6, 0.25"

V + ejβd − V − e−jβd
V (d) = V + ejβd − V − e−jβd and I(d) = . ".2 x!"5
Zo
Therefore,
".5 "2

+ −
V −V "1
VL = V (0) = V + + V − and IL = I(0) = . SmithChart
Zo 1

.5 2
Using Zo = 50 Ω and 0.470588 $ 0.882353 #

−j π4 j π4
V + = 50e V and V − = −30e V .2 x!5

from Example 3, we find that

0 .2 .5 1 2 r!5
4.
π π
−j π4 j π4 50e−j 4 + 30ej 4 π π
VL = 50e − 30e V and IL = = e−j 4 + 0.6ej 4 A.
50
".2 x!"5
0.470588 " 0.882353 #

!0.6, "0.125"

".5 "2

"1

8
37 Smith Chart and impedance matching
• In lossless TL circuits the average power input Pin at the generator end I(d)
+ Load
precisely matches the average power delivered to the load, PL. +
-
Zg
V (d) Zo ZL
F = Vg -
In fact, Pin and PL also match the average power P (d) transported on
the line at an arbitrary d. d l 0

• We have in general
1
P (d) = Re{V (d)I ∗(d)}
2
+ jβd
1 + jβd − −jβd V e − V − e−jβd ∗ Power tx’ed toward the load:
= Re{(V e + V e )( )}
2 Zo |V +|2
1 |V + |2 |V −|2 V − V +∗e−j2βd − (V −V +∗e−j2βd )∗ .
= Re{ − + } 2Zo
2 Zo Zo Zo
|V +|2 |V − |2
= − . Power tx’ed toward the gen-
2Zo 2Zo
erator:
– Note that P (d) is the difference of power transported |V − |2
|V + |2 .
2Zo
toward the load by the “forward-going” wave, and 2Zo
|V − |2
2Zo
toward the generator by the reflected wave.
– Also note that Power reflection coefficient:

|V +|2 |V − |2 |V +|2 |ΓL|2.

P (d) = − = (1 − |ΓL|2)
2Zo 2Zo 2Zo
Power transmission coeff.:
so that |ΓL|2 is an effective power reflection coefficient.
1 − |ΓL|2.
1
 • In TL circuits with load impedances ZL unmatched to the character-
istic impedance Zo, the reflected power I(d)
+ Load
+ 2 Zg
|V | + V (d) Zo
|ΓL|2 -
F = Vg -
ZL
2Zo
will be non-zero and the VSWR>1. d l 0

This a condition not favored by practical signal generators used in

TL circuits.
• Most generators are designed (in their biasing arrangements) to oper-
ate in circuits with low VSWR (close to unity), requiring Zin closely
matched to Rg , most frequently 50 Ω, an optimal characteristic impedance
value for coax-lines (when line losses are taken into account).
• Thus a standard procedure is to use TL’s with Zo = Rg , and uti-
lize a lossless impedance matching network on the TL if the load
impedance ZL 6= Zo.

– This practice is called impedance matching.

Impedance matching achieves VSWR=1 between the generator and the match-
ing network inserted at a location between the load and the generator.
• The inserted network should be designed to yield an input impedance
equal Zo at its input terminals.
The following examples illustrate different ways of achieving an
impedance match.

2
 Quarter-wave matching
λ
4

q
Zq = Zo Z(d1 )
Zo ZL
Example 1: Quarter-wave matching of resistive loads:

Consider a TL with ZL = 25 Ω and Rg = Zo = 50 Ω. Since ZL 6= Zo the load is d l d1 0

unmatched and the VSWR>1.

To reduce the VSWR on the line connected to the generator to unity, we can insert
a quarter-wave transformer right after ZL — i.e., at d1 = 0 in the circuit
shown in the margin — with a characteristic impedance
√ √
Zq = 25 × 50 = 1250 = 35.35 Ω.

The impedance at the input terminals of the quarter-wave transformer (on the left)
is then Zo , i.e., 50 Ω, implying a perfect impedance match.

• Quarter-wave matching illustrated above is a very commonly used match-

ing technique.

• It is a straightforward application of the quarter-wave transformer impedance

formula
Zq2
Zin =
ZL
for a transformer with characteristic impedance Zq .

3
 Quarter-wave matching
λ
4

q
Zq = Zo Z(d1 )
Example 2: Quarter-wave matching of reactive loads: Zo ZL

Consider a TL with ZL = 50 + j50 Ω and Rg = Zo = 50 Ω. Since ZL 6= Zo the load d l d1 0

is unmatched and the VSWR>1.
SmithChart
We cannot insert the quarter-wave transformer right after the load because then we 1

would need a complex valued Zq implying a lossy matching network. .5 2

Instead, we insert a quarter wave transformer a distance d1 to the left of ZL , 1. + 1. ä

where d1 is selected, using a SC, to have a purely resistive Z(d1). In that case, .2 x=5
80.447214, 0.0881041<
the quarter-wave transformer impedance formula
p .2 .5 1 2 2.61803 r=5
Zq = Z(d1) × 50 0

yields a real valued Zq as needed. This procedure leads to having d1 = dmax or 0.5 - 0.5 ä
x=-5
d1 = dmin corresponding to the positions of voltage maxima and minima on the -.2

line.
-.5 -2
As shown in the margin,
-1

Z(d1) = 50(2.62 + j0) = 131 Ω. Note that:

for
d1 ≈ 0.250λ − 0.162λ = 0.088λ z(d1 ) = z(dmax ) = VSWR ≈ 2.62

is a suitable choice for quarter-wave matching. In that case we need as marked on the SC.
√ √ Also
Zq = 131 × 50 = 50 × 2.62 Ω
dmax ≈ 0.088λ
for the quarter wave transformer in order match to load to a line with Zo = 50 Ω. since, as marked on the SC,
the angle of ΓL is 0.088λ.

4
 Single-stub
tuning Shorted
stub
ls

Example 3: Single-stub tuning:

Consider a TL with ZL = 100 − j50 Ω and Rg = Zo = 50 Ω. Since ZL 6= Zo the load Zo ZL

is unmatched and the VSWR>1.
d l d1 0
We will insert a shorted-stub a distance d1 to the left of ZL in parallel with the line
to achieve an impedance match. Want y(d1 ) + ystub =1
SmithChart
1
Distance d1 will be selected, using a SC, to have a normalized admittance of
.5 2

y(d1) = 1 + jb
so that a stub, with a normalized input admittance .2 x=5
0.4 + 0.2 ä
ystub = −jb,
.2 .5 1 2 2.61803 r=5
can be added in parallel to have a combined admittance of 0
2. - 1. ä

y(d1) + ystub = 1 + j0 80.447214, -0.0368959<

-.2 x=-5
and achieve a perfect impedance match (i.e., VSWR=1).

In specific -.5 -2
ZL 1
zL = = 2 − j1 and yL = = 0.4 + j0.2
Zo zL
-1
SmithChart
1
as shown on the SC on the top in the margin. We rotate clockwise on the SC by
an amount corresponding to d1 to obtain .5 2

y(d1 ) = 1 + j1 1. + 1. ä
.2 x=5
on the “g = 1” or “y = 1 + jb” circle as shown in the bottom SC. From the
amount of rotation we determine
0 .2 .5 1 2 2.61803 r=5
d1 ≈ 0.162λ − 0.037λ = 0.125λ.
0.5 - 0.5 ä
x=-5
5
-.2
80.447214, -0.161896<
The required input impedance of the shorted stub to achieve

y(d1) + ystub = 1 + j0

is
ystub = −1j. SmithChart
1

.5 2
To achieve this input admittance the required stub length is
λ
ls = = 0.125λ .2 x=5
8 0.4 + 0.2 ä

as determined from the SC — start at y = ∞ point on the SC on the far right

.2 .5 1 2 2.61803 r=5
(corresponding to the short termination), and then rotate clockwise (toward the 0
2. - 1. ä
generator) until the normalized admittance reads −j1; the amount of rotation
80.447214, -0.0368959<
indicates the required ls. -.2 x=-5

-.5 -2

• Another matching technique called double-stub tuning uses two -1

SmithChart
shorted stubs of lengths l1 and l2 located at fixed values of d1 and d2. 1

.5 2

– Typically d1 is zero or λ4 , and

1. + 1. ä

3 λ8 .
.2 x=5
– d2 = d1 +
Vary l1 and l2 until VSWR is reduced to 1 near the generator end. 0 .2 .5 1 2 2.61803 r=5

The advantage of double-stub tuning is avoiding changes of stub loca- 0.5 - 0.5 ä
-.2 x=-5
tions when ZL is changed. It’s implementation on a SC is considerably 80.447214, -0.161896<

more complicated than single-stub tuning.

-.5 -2

-1

6
38 Distribution networks
• A corporate ladder network that combines 4
CORPORATE LADDER NETWORK
identical loads ZL into a single equivalent input ZL ZL ZL ZL
Loads
impedance ZL is shown in the margin.

• In this network 6 different quarter-wave trans-

Quarter-
formers with arbitrary but identical characteris- wave tx’s
tic impedances Zo are utilized.
You should be able to compute the load volt-
ages VL in the network in terms of input volt-
Quarter-
age Vin applied across the input port by using wave tx’s
All quarter-wave
the current-forcing formula for the quarter-wave tx’s have the
same Z_o
transformer introduced earlier on. Input
port Zin = ZL
Input impedance is the
same as load impedance

By symmetry, the loads absorb equal avg power,

a quarter of the input power each.

• If, in a corporate ladder network, ZL = Zo , then TL segment lengths

connected to each ZL can be varied at will without affecting the input
impedance ZL = Zo (why?).

• Allowing variable length TL’s connected to each ZL makes it possible to

adjust and vary the phase of the voltage and current of each ZL — this
is useful, for instance, in feeding phased antenna arrays (ZL represents
an antenna load) to achieve steerable radiation patterns.

1
• A hybrid combiner network shown in the mar- HYBRID COMBINER --- SUM and DIFFERENCE OF ISOLATED
gin can be used to excite two identical TL loads GENERATOR SIGNALS V_A AND V_B ARE APPLIED TO LOADS R

R (e.g., antenna arrays impedance matched to VA − VB VA VA + VB

−j √ −j √
have input impedances R) with independent sig- 2 2 2 2 2

√ R √
nal generators VA and VB having equal internal R
Zo = 2R
+
Zo = 2R
R
LOAD - VA LOAD
resistances R matched to the load resistance. INPUT A Zo =
√
2R
√
Zo = 2R

• The hybrid “rat-race” combiner is built with 6 VB

2
quarter-wave transformers of identical Zo =
√
2R Zo =
√
2R
R
λ +
√ 4 - VB

Zo = 2R, INPUT B

in which case the generators VA and VB see

impedance-matched loads (at the hybrid inputs
where they are connected) and produce load
voltages proportional to VA ± VB as shown in
the diagram.
• Generators A and B with open ckt voltages VA and VB are isolated from one another’s influence
because of “destructive interference” between the two paths from each generator to the other one
(two paths have a λ2 length difference).

• This very special situation allows one to calculate the various terminal voltages on the hybrid
due to VA and VB one-at-a-time as if loads R were isolated from generator-B and -A (by “virtual
shorts” existing across generator terminals when VB and VA are suppressed) in turns, and then
superpose the results.

2
• Terminal voltages obtained with that procedure (those shown on the diagram) turn
out to be valid when both generators are active as can easily be checked for self-
consistency by using the current-forcing equations introduced earlier. For instance,
the total current into generator-A terminal (flowing from both sides) is
j VA − VB j VA + VB VA
IA = − √ (−j √ ) − √ (−j √ ) = − ,
R 2 2 2 R 2 2 2 2R
and hence the voltage drop from the same terminal to the ground is
VA VA
IA R + V A = − R + VA =
2R 2
as marked explicitly on the diagram. All self-consistency tests that can be applied
with the given expressions are passed, and so the results given are valid.

• The input and output ports of a HYBRID COMBINER --- SUM and DIFFERENCE OF ISOLATED
hybrid combiner can be swapped GENERATOR SIGNALS V_A AND V_B ARE APPLIED TO LOADS R

while still maintaining the proper- VA − VB VA VA + VB

−j √ −j √
ties of the hybrid — namely, input 2 2 2 2 2

√ R √
impedance R, and output signals the R
Zo = 2R
+
Zo = 2R
R
LOAD - VA LOAD
sum and difference of generator volt- INPUT A Zo =
√
2R
√
ages. Zo = 2R

VB
2
√ √ R
Zo = 2R Zo = 2R
λ +
- VB
4

INPUT B

3
39 Lossy lines IDEAL LOSSLESS T.L.:

I(z) I(z + ∆z)

+ ∆zL
V (z) ∆zC
-
• Lossless TL’s we have been studying so far are idealizations of real TL’s z z + ∆z
which are invariably lossy.
REALISTIC LOSSY T.L.:
– Here, we are making reference to Ohmic energy losses in the
I(z) ∆zR I(z + ∆z)
conducting wires of the TL, as well as to losses in the imperfect + ∆zL
dielectric separating the two conductors. V (z) ∆zC ∆zG
-
z z + ∆z
• The effect of wire losses in TL’s is modelled by adding a ∆zR resis-
tance in series with ∆zL inductor in the equivalent circuit model of an
infinitesimal (∆z ≪ λ) TL section as shown in the margin.

• In addition, a shunt conductance ∆zG in parallel with capacitance ∆zC

accounts in the lossy model for dielectric losses. Using perturbation theory, it can
be shown that for a coax of inner
and outer radii a and b,
• While the phasor form of telegrapher’s equations for a lossless TL is r
fµ 1 1
R= ( + ),
∂V ∂I πσ a b
− = jωLI and − = jωCV,
∂z ∂z while for a parallel-plate transmis-
sion line of width W ,
for lossy lines — where impedance per unit length jωL must be replaced r
4π f µ
R= ,
by jωL + R and conductance per unit length jωC by jωC + G — the W πσ
in terms of conductivity σ and per-
equations take the form meability µ of the T.L. conductors.

∂V ∂I
− = (jωL + R)I and − = (jωC + G)V.
∂z ∂z
1
• We will next show that

1. lossless line solutions can be readily modified to account for loss

effects introduced by Ohmic energy losses in R and G,
2. lossless line results we have learned up till now are by and large
valid even on lossy lines provided that
(a) frequency ω is sufficiently large, and
V ± ±jβd
(b) voltage and current solutions V ±e±jβd and ±Z o
e are mod-
ified by multiplying an attenuation term e±αd which only mat-
ters in practice when d ≫ λ.

• Note that lossless line solutions of telegrapher’s equations can be re-

stated as
± ±γd V ± ±γd
V =V e and I = e ,
±Zo
where
s
√
r
p L jωL
γ = jβ = jω LC = (jωL)(jωC) and Zo = = .
C jωC

– Replacing jωL by jωL + R, and jωC by jωC + G, we obtain

s
p jωL + R
γ = (jωL + R)(jωC + G) and Zo =
jωC + G
in the lossy case.

2
While waves governed by lossy γ and Zo (see margin) can exhibit substan-
tially different beavior than the lossless waves (examined in the previous
sections), at high frequencies the wave properties are reasonably similar as
alluded in item (2) above. We next examine this simplified high-frequency
limit.
p
• At high frequencies ω, such that ωL ≫ R and ωC ≫ G, we have γ= (jωL + R)(jωC + G)

– characteristic impedance s
jωL + R
Zo =
s r
jωL + R L jωC + G
Zo = ≈
jωC + G C

just as in the lossless case1, and

– complex propagation constant
s s
p √ R G
γ = (jωL + R)(jωC + G) = jω LC 1+ 1+
jωL jωC
√ R G √ 1 R
≈ jω LC(1 + )(1 + ) ≈ jω LC + ( + GZo)
j2ωL j2ωC 2 Zo
= jβ + α

with
√ R G 1 R
β ≈ ω LC and α ≈ β( + ) = ( + GZo).
q
2ωL 2ωC 2 Zo
L L
1
In fact Zo = C
is exact even for a lossy line if R
= GC .

3
 2π
• Note that β = λ
is the same as in the lossless case, and since
R G
α ≈ β( + ) ≪ β,
2ωL 2ωC
the “penetration depth” δ ≡ α1 of voltage and current waves on the TL
is much longer than a wavelength λ = 2π β
in this regime.
In summary, in the high-frequency regime, characteristic impedance Zo
and wavenumber β are (practically) the same as they are on lossless lines,
but signals do attenuate by a factor e±αd which should not be (and cannot
be) neglected over long distances d exceeding many wavelengths λ.
• At lower frequencies where the above approximations cannot be justi-
fied, a more careful analysis of lossy line equations is warranted.

• Finally, for an air-filled coax with inner and outer radii a and b, it can
be shown that the attenuation constant
q
f µo 1 b
1R 1 πσ b (1 + a )
α= = ηo ,
2 Zo 2 2π ln( ab )
which minimizes, at a fixed outer radius b, for ab ≈ 3.6, which in turn
results in an “optimal” characteristic impedance of
ηo b b
Zo = ln( ) = 60 ln( ) Ω ≈ 75 Ω
2π a a
for the same coax. Note that this result is independent of σ, the con-
ductivity of inner and outer conductors of the coax.

4
– For a dielectric filled coax having ǫ = 94 ǫo — implying vp = 23 c =
2×108 m/s — the same ratio ab ≈ 3.6 of outer and inner conductor
radii leads to Zo ≈ 50 Ω, the most common Zo encountered in
practical applications.
– The above result should also explain why having a thicker coax —
larger b — is better when losses are a concern.