Ph106bc: Electrodynamics: Sunil Golwala
Ph106bc: Electrodynamics: Sunil Golwala
Electrodynamics
Sunil Golwala
Winter/Spring 2022
Contents
Introduction to Course
1.1 Course Material
1.2 Notation; including Deviations from Griffiths
Contents Page 2
Contents
Advanced Electrostatics
3.1 Intuitive Approach to Laplace’s Equation
3.2 Uniqueness Theorem
3.3 Method of Images
3.4 Formal Solution to Poisson’s Equation: Green Functions
3.5 Introduction to Separation of Variables
3.6 Separation of Variables in Cartesian Coordinates
3.7 Separation of Variables in Spherical Coordinates: General Theory
3.8 Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
3.9 Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
3.10 Multipole Expansions
Electrostatics in Matter
4.1 Polarizability and Polarization
4.2 The Electric Displacement Field
4.3 Linear Dielectrics
4.4 Boundary Value Problems with Linear Dielectrics
4.5 Electrostatic Energy in and Forces on Linear Dielectrics
Contents Page 3
Contents
Magnetostatics
5.1 Study Guidelines
5.2 Lorentz Forces and Current Densities
5.3 Conservation of Charge and the Continuity Equation
5.4 Fields of and Magnetic Forces between Currents
5.5 Curl and Divergence of the Magnetic Field; Ampere’s Law
5.6 Magnetic Vector Potential
5.7 Boundary Conditions on Magnetic Field and Vector Potential
5.8 Magnetic Multipoles
Magnetostatics in Matter
6.1 Paramagnetism and Diamagnetism
6.2 The Field of a Magnetized Object
6.3 ~ and Magnetic Permeability
The Auxiliary Field H
6.4 Boundary Value Problems in Magnetostatics
Electrodynamics
7.1 Currents and Ohm’s Law
7.2 Motional Electromotive Force
7.3 Electromagnetic Induction
7.4 Inductance
7.5 Magnetic Energy and Forces
7.6 Maxwell’s Equations
Contents Page 4
Contents
Conservation Laws
8.1 Motivation and Analogy: Conservation of Charge
8.2 Poynting’s Theorem: Conservation of Energy
8.3 The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
8.4 Conservation of Angular Momentum (Skip)
Electromagnetic Waves
9.1 Introduction and Study Guidelines
9.2 Electromagnetic Waves in Vacuum
9.3 Electromagnetic Waves in Perfectly Nonconducting Matter
9.4 Electromagnetic Waves in Conducting Matter
9.5 Electromagnetic Waves in Dispersive Matter
9.6 Introduction and Study Guide: Guided Waves
9.7 Transmission Lines
9.8 Waveguides
Potentials Revisited
10.1 Potential Formulation
10.2 Retarded Potentials and Fields
Contents Page 5
Contents
Radiation
12.1 Potentials, Fields, and Power Radiated by an Accelerated Point Charge
12.2 General Theory of Radiation
Applications of Radiation
13.1 Classical Scattering Theory
13.2 Antennas
Contents Page 6
Ph106b Lectures
Lectures Page 7
Ph106c Lectures
Lectures Page 8
Lecture 1:
Introduction to Course
Basics of Electrostatics I:
Electric Force and Field
Dirac Delta Function
Integral Form of Gauss’s Law
Date Revised: 2022/01/03 05:30
Date Given: 2022/01/03
Page 9
Section 1
Introduction to Course
1.1 Course Material
1.2 Notation; including Deviations from Griffiths
Page 10
Section 1.1 Introduction to Course: Course Material
Course Material
Overview
This is a course on electrodynamics. It will review the basic material you learned in
Ph1bc but will go beyond in both content as well as in mathematical sophistication.
The intended learning outcome of both Ph106b and Ph106c is for students to acquire
the ability to calculate electric and magnetic potentials, fields, energies, and forces in
a variety of basic physical configurations combined with an understanding of the
underlying physical principles and calculation techniques. This outcome requires both
an understanding of principles as well as the ability to apply them to do calculations!
The course will primarily use and follow Introduction to Electrodynamics by Griffiths
(fourth edition). Supplementary material is drawn from Jackson and from Heald &
Marion, both on reserve in the library. The material presented here will be
self-contained, but past students have found it useful to obtain a copy of Jackson. It
is certainly a book you will want if you continue in physics or a related field.
Prerequisites
Physics:
I Electricity and Magnetism: While Ph1bc is a formal prerequisite for the course,
we will develop the material from scratch. However, review material will be
covered quickly and a basic familiarity with the concepts will be assumed.
I Classical mechanics: Generally, mechanics at the level of Ph1a is sufficient for
this course, though some optional material at the end of Ph106c will make use
of Ph106a material.
I Chapter 1 of Griffiths except for Sections 1.1.5 (“How Vectors Transform”) and
1.5 (“The Dirac Delta Function”). We will review some of prerequisite material
as needed.
I Solutions to second-order linear ordinary differential equations with constant
coefficients (i.e., simple harmonic oscillator).
I Orthonormal functions/bases.
I Over the course, we will develop the following more sophisticated concepts:
I Dirac Delta function.
I Separation of variables to reduce second-order linear partial differential
equations to ordinary differential equations.
I Various specific types of orthonormal functions, specifically sinusoids,
Legendre polynomials, and spherical harmonics.
I Tensor formalism for relativity.
I Key point: Mathematics is the language of physics. You must be competent in
the above basic mathematics to understand and use the material in this course.
Intuition is crucial, but it must be formalized mathematically.
However, mathematics is not just symbolic manipulation or brute force
calculation. Make sure you understand the meaning of every mathematical
expression—i.e., carry along the intuition with the symbols! Only do algebra
and explicit differentiation and integration as a last resort! We will demonstrate
this approach regularly.
Topics to be Covered
Page 16
Section 2
Review of Basics of Electrostatics
2.1 Study Guidelines
2.2 The Assumed Conditions for Electrostatics
2.3 Coulomb’s Law and the Electric Field
2.4 Gauss’s Law
2.5 The Electric Field has Vanishing Curl
2.6 The Electric Potential
2.7 Aside on Techniques
2.8 Boundary Conditions on the Electric Field and Potential
2.9 Poisson’s and Laplace’s Equations
2.10 Electrostatic Energy
2.11 Electric Conductors
2.12 Capacitors and Capacitance
Page 17
Section 2.1 Review of Basics of Electrostatics: Study Guidelines
Study Guidelines
You have seen all the material in this section before in Ph1b. However, the derivations
done there were not as rigorous as they could be because you were simultaneously
learning vector calculus. Our goal in this section is to do more rigorous derivations to
give you some practice in using the mathematical tools. We won’t do any examples in
lecture or the notes because they duplicate Ph1b. But you should make sure you are
comfortable with the examples in Griffiths Chapter 2.
Page 18
Section 2.2 Review of Basics of Electrostatics: The Assumed Conditions for Electrostatics
I All electric charges sourcing the electric field are stationary and have been so for
a sufficiently long time that all fields are static and thus the electric field can be
written in terms of the source charges’ current positions.
I The source charges are held fixed and cannot react to the fields from any test
charges that may be stationary or moving relative to the source charges.
We will see later that, when charges are moving, it takes time for the information
about the position to propagate and thus the fields at a given point depend on the
configuration of the charges at earlier times.
Page 19
Section 2.3 Review of Basics of Electrostatics: Coulomb’s Law and the Electric Field
1 q0 q b
F~ = R with ~ ≡~
R r0
r −~ (2.1)
4 π o R 2
where o = 8.85 × 10−12 C2 N−1 m−2 . The force points along the line from q 0
~ The electric charge is in the
to q as indicated by the sign of the definition of R.
units of Coulombs (C), which is a fundamental unit that cannot be written in
terms of other fundamental units.
Recall that: we use ~ rather than boldface to indicate vectors; R where Griffiths
uses a script r ; and a different convention from Griffiths for the symbols for the
two charges and their position vectors.
I Superposition: the empirical fact that Coulomb’s Law obeys the principle of
superposition: the force on a test charge q at ~r due to N charges {qi0 } at
ri 0 } is obtained by summing the individual vector forces:
positions {~
N N
1 qi0 q b
F~ = F~i = ~i ≡ ~
X X
Ri with R ri 0
r −~ (2.2)
i=1 i=1
4 π o Ri2
q0 b
1
~
F
4π R2
R for a single source charge q 0 at ~
r0
~ (~
E r) = = PN o qi0
1
q i=1 4 π o R 2
R
bi for N source charges {qi0 } at positions {~
ri 0 }
i
(2.3)
1 1 b
Z
~ (~
E r) = r 0)
R dq(~ (2.4)
4 π o R2
~ (~
E r) = (2.5)
1 1 b 1 1
Z Z Z
0 0 0
xb R · x dq(~
r ) + y Rb · yb dq(~
r ) + z b · zb dq(~
R r )
R2 R2 R2
b b b
4 π o
Using the Dirac delta function we will define below, one can write the first two as
special cases of the latter by using delta functions in the dimensions in which the
charge distribution has no extent.
Relating Equation 2.6 to Equation 2.3 offers us both the opportunity to rigorously
connect them as well as a chance to introduce the Dirac delta function. The Dirac
delta function at ~
ro , δ(~r −~ ro ), is defined by what it does when it is multiplied against
an arbitrary function f (~ r ) and integrated: For any function f (~r ) and any volume V
containing the point ~ ro , it holds that
Z
f (~
ro ) ro ∈ V
~
r 0 ) δ(~
f (~ r0 −~
ro ) dτ 0 = (2.9)
0 ro 6∈ V
~
In particular, if f (~
r ) is unity, then the right side of the above integral is unity for
ro ∈ V: the integral of a delta function over the volume containing its ~
~ ro is 1, and,
conversely, the integral of a delta function over any volume not containing its ~ ro
vanishes.
~
r = x xb + y yb + z zb
δ 3 (~
r −~
ro ) = δ(x − xo )δ(y − yo )δ(z − zo ) where
~
ro = xo xb + yo yb + zo zb
(2.10)
With the above, if we define the charge distribution for a set of point charges {qi0 } at
ri 0 } to be
positions {~
N
X
r) =
ρ(~ qi0 δ(~ ri 0 )
r −~ (2.11)
i=1
then, when we do the integral in Equation 2.6 over any volume V containing all N
charges, we recover the discrete version of the expression for the electric field,
Equation 2.3.
N
1 dτ 0 qi0 δ(~
r0 −~ ri 0 ) ~r −~r0
Z X
~ (~
E r) =
4 π o |~
r − ~
r 0 |2 |~
r − ~
r 0|
V i=1
N Z
1 X r −~
~ r0
= dτ 0 qi0 δ(~
r0 −~
ri 0 )
4 π o i=1 V |~
r −~r 0 |3
N N
1 X 0 ~ r −~ri 0 1 X qi0 b
= qi = Ri (2.12)
4 π o i=1 |~r −~ri 0 |3 4 π o i=1 Ri2
Gauss’s Law
Statement of Gauss’s Law
The flux of the electric field through a surface is the integral of the component of the
electric field normal to the surface over the surface:
Z
FS = ~ · nb(~
E r ) da (2.13)
S
Gauss’s Law relates the flux of the electric field through any closed surface to the total
charge enclosed by that surface:
1
I Z
FS = ~ · nb(~
E r ) da = dτ ρ(~
r) (2.14)
S o V(S)
H
where V(S) is the surface enclosed by S and indicates the integral over a closed
surface. Our derivation below will take the surface normal direction to be outward
from the closed volume.
1 dτ 0 ρ(~r 0)
d 2 FS (~ r 0) = r 0 · nb(~
r, ~ r −~
~ r ) da (2.15)
4 π o |~ r 0 |3
r −~
The left side is a double differential because the right side is. If one considers the
geometry (see diagram above), one sees that the quantity (~ r −~ r 0 ) · nb(~r ) da/|~r −~r 0 | is
the projected area of the area element da normal to the unit vector (~ r −~ 0 r −~
r ) /|~ r 0|
from ~r 0 to ~r . Since |~ r −~ r 0 |2 is the square of the distance from ~ r 0 to ~ r , then the
quantity (~ r −~ r 0 ) · nb(~
r ) da/|~ r −~ r 0 |3 is the solid angle dΩ(~ r 0 ) subtended by da at ~
r, ~ r
as viewed from ~ r 0.
1
d 2 FS (~ r 0) =
r, ~ dτ 0 ρ(~
r 0 ) dΩ(~ r 0)
r, ~ (2.16)
4 π o
We know that if we integrate the solid angle over the entire closed surface S
r 0 , we recover 4 π, so:
surrounding our source charge point ~
1 1
I
r 0) =
dFS (~ dτ 0 ρ(~
r 0 ) dΩ(~ r 0) =
r, ~ dτ 0 ρ(~
r 0) (2.17)
4 π o S o
We expect that, due to superposition, if the above is true for the flux due to an
infinitesimal volume of charge, then it holds for the whole distribution of charge
enclosed by S. We can prove this by calculating the flux through S due to the entire
charge distribution, using the fact that the distribution is fully contained inside S (one
of our starting assumptions):
1 dτ 0 ρ(~r 0)
I I Z
~S (~ r 0 · nb(~
FS = E r ) · nb(~
r ) da = 0 |3
r −~
~ r ) da
S S V(S) |~
4 π o
r − ~
r
1 1
I Z I Z
= dτ 0 ρ(~
r 0 ) dΩ(~ r 0) =
r, ~ d 2 FS (~ r 0)
r, ~ (2.18)
4 π o S V(S) 4 π o S V(S)
1 1 1
Z I Z Z
FS = d 2 FS (~ r 0) =
r, ~ r 0) =
dFS (~ dτ 0 ρ(~
r 0 ) (2.19)
4 π o V(S) S o V(S) o V(S)
Note how the proof depended on the 1/r 2 dependence of Coulomb’s Law. The proof
could be done in the opposite direction: Gauss’s Law implies Coulomb’s Law. In
general, for any force, there is a simple Gauss’s Law if and only if the force has a 1/r 2
dependence. Another example is gravity, as you learned in Ph1a and Ph106a.
Section 2.4.3 Proof of Gauss’s Law Page 32
Section 2.4 Review of Basics of Electrostatics: Gauss’s Law
But we are not quite done yet, as we assumed at the start that the charge distribution
vanishes outside of S. Does the result generalize to the case where there is some
~S receives contributions from that charge? Yes, it does.
charge outside of S so that E
Page 34
Section 2.4 Review of Basics of Electrostatics: Gauss’s Law
Since the above holds for any volume V, the integrands must be equal, giving us the
differential version of Gauss’s Law:
~ (~
~ ·E 1
∇ r) = r)
ρ(~ (2.23)
o
We will frequently employ this technique of using an equality between two integrals
over an arbitrary volume or surface to conclude their integrands are equal.
1 dτ 0 ρ(~r 0)
Z
~ (~
~ ·E ~ · r0
∇ r) = ∇ r −~
~ (2.24)
V0 4 π o |~ r 0 |3
r −~
1 r −~
~ r0
Z
∇ ~ (~
~ ·E r) = dτ 0 ρ(~ ~ ·
r 0) ∇ (2.25)
4 π o V0 |~
r −~r 0 |3
One could calculate the above divergence explicitly in any particular coordinate
system. But it is both more rigorous and more instructive to do it using the
divergence theorem.
We can calculate the integral of the above divergence over some arbitrary volume V
(with surface S, with neither V nor S necessarily related to V 0 and S 0 ), as we need to
do for Gauss’s Law, by exchanging the order of integration (no prohibition on doing so
because we don’t move ∇ ~ around) and converting the volume integral over ~ r to an
easier-to-do surface integral using the divergence theorem:
1 ~ · ~ r −~ r0
Z Z Z
dτ ∇ ~ (~
~ ·E r) = dτ dτ 0 ρ(~
r 0) ∇
4 π 0 |~
r − ~
r 0 |3
V V o V
1
Z Z
~
r − ~
r 0
= dτ 0 ρ(~ r 0) dτ ∇~ ·
4 π o V 0 V |~
r −~ r |3
0
1 r −~
~ r0
Z I
= dτ 0 ρ(~ r 0) r) ·
da nb(~ (2.26)
4 π o V 0 S(V) |~
r −~ r 0 |3
We can apply to the surface integral the same argument about solid angles that we
used in proving Gauss’s Law. The integrand above is just the solid angle subtended by
the area element da at ~ r as viewed from ~ r 0 . As before, if ~r 0 is inside V, then the
above integral yields the total solid angle, 4 π. If ~ r 0 is not inside of V, then, for every
area element da at ~ r , there is an area element with an equal and opposite
contribution, making the integral vanish. That is,
~ · ~ r −~r0 r 0 is inside V
Z
4 π if ~
dτ ∇ = (2.27)
V |~
r −~ 0
r | 3 0 r 0 is outside V
if ~
The divergence in the integrand looks a lot like a delta function; more on that later.
The above statement says that the integral over V vanishes if ~ r 0 is not inside V and
yields 4 π if it is inside V. This turns the double integral over V and V 0 into a single
integral over V ∩ V 0 :
1 1
Z Z Z
dτ ∇ ~ (~
~ ·E r) = dτ 0 4 π ρ(~
r 0) = dτ 0 ρ(~
r 0) (2.28)
V 4 π o V∩V 0 o V∩V 0
1
Z Z
dτ ∇ ~ (~
~ ·E r) = dτ 0 ρ(~
r 0) (2.29)
V o V
~ (~
~ ·E 1
∇ r) = r)
ρ(~ (2.30)
o
Now, ρ(~ r ) is an arbitrary function, so we see that the divergence in the integrand acts
like the δ function: it picks out ρ(~ r0 =~r ). Thus, we have also proven
~ · ~r −~r0
∇ r0 −~
= 4 π δ(~ r) (2.32)
|~
r −~r 0 |3
Since the delta function picks out the point where its argument vanishes, it doesn’t
matter what the sign of the argument is. One can prove this explicitly using change of
variables: when the sign of the argument changes, the sign of the differential and of
the limits of integration change also. Those two sign flips cancel each other. Thus
r0 −~
δ(~ r ) = δ(~ r 0)
r −~ (2.33)
Section 2.4.6 Aside: Relation of the Dirac Delta Function to a Divergence, Invariance under Inversion of its Argument Page 39
Section 2.4 Review of Basics of Electrostatics: Gauss’s Law
It may seem that this last property is not true given the above relation between the
delta function and a divergence. In particular, let’s flip the sign on the function the
divergence is acting on:
~ · r −~
~ r0 ~ ·− ~r0 −~
r r0 −~
~ · ~ r ?
r0 −~
4 π δ(~ r) = ∇ 3
=∇ 3
= −∇ = −4 π δ(~
r −~r 0)
|~
r −~ 0
r | |~ 0
r −~ r| |~ r |3
r0 −~
(2.34)
Don’t we have a problem? No, because we failed to recognize that ∇ ~ takes derivatives
with respect to ~ r . Since ~ r −~ r 0 just offsets ~r , then the divergence with respect to
r −~
~ r 0 is the same as the divergence with respect to ~ r . But, when we flip the sign on
r −~
~ r 0 , we should do the same for the divergence: the divergence should be taken with
respect to ~ r0 −~ r . That flips the sign of the divergence operator: ∇ ~ ~r −~r 0 = −∇ ~ ~r 0 −~r .
Finally, ~ r acts like an offset for ~ r 0 , and so the divergence with respect to ~ r0 −~ r is the
same as with respect to ~ r 0 . That is:
Errors of the above type are easy to make and not self-evident! Mathematics in
physics is not just symbol manipulation: there is meaning that must be understood in
order to be sure those manipulations are justified.
Section 2.4.6 Aside: Relation of the Dirac Delta Function to a Divergence, Invariance under Inversion of its Argument Page 40
Section 2.5 Review of Basics of Electrostatics: The Electric Field has Vanishing Curl
The curl of E ~ can be shown to vanish simply by calculating it for an arbitrary charge
distribution:
1 ~ r −~
~ r0
Z
∇ ~ (~
~ ×E r) = ∇× dτ 0 ρ(~
r 0)
4 π o V |~
r −~r 0 |3
1 ~ × ~ r −~r0
Z
= dτ 0 ρ(~
r 0) ∇ (2.36)
4 π o V |~
r −~r 0 |3
Let’s take a simpler, more geometric, and more intuitive approach. As we saw above,
r 0 is just an offset to ~
~ r , thus
Note that, in doing this offset, the curl will be expressed in terms of the components
r −~
of ~ r 0 . This does not change the bounds of integration, but it may make the
expression look complicated because the volume differential is still in the ~r 0 system.
Since we will show this expression, the integrand, vanishes, this bookkeeping
complication is not important. If we define ~s = ~ r −~r 0 , then we have
~ ~r × ~r −~r0 ~ ~s × ~s
∇ =∇ (2.38)
|~
r −~r 0 |3 s3
Now, the function on which the curl is acting has symmetry in the coordinate system
in which the curl is acting, and hence the calculation will be simplified. You can
probably see intuitively that the above curl vanishes, but let’s prove it. (Note also that
the change of variables would require a change to the limits of integration, but, again,
because we will prove the integrand will vanish, this bookkeeping complication will not
be important.)
With the above form, we can trivially apply the formula for the curl in spherical
coordinates, which is listed in Griffiths. For the sake of being explicit, that formula is
1 ∂ vθ 1 1 ∂ vr
~ ×~ ∂ ∂
∇ v = vφ sin θ − rb + − r vφ θb
r sin θ ∂θ ∂φ r sin θ ∂φ ∂r
1 ∂ ∂ vr b
+ (r vθ ) − φ (2.39)
r ∂r ∂θ
Don’t get confused between ~s and ~ r ; the r derivatives and subscripts refer to the
radial coordinate of the coordinate system in which the curl is being taken. In our
case, s is the radial variable and the radial component of ~s /s 3 is 1/s 2 . Thus, ~
v has
only a radial component and that radial component depends only on the radial
distance from the origin. All the derivatives involving the θ and φ components of ~ v
vanish because the components themselves vanish, and the derivatives involving the
radial component vanish because those derivatives are with respect to θ and φ. (Don’t
be confused: ~ v itself depends on θ and φ because the direction of ~ v depends on them;
but the curl formula takes care of that dependence.)
∇ ~ (~
~ ×E r) = 0 (2.40)
Note again that we did not brute-force differentiate, but rather we thought about how
to simplify the calculational aspect (via origin offset) and then saw that made the
result both geometrically/intuitively obvious and easier to demonstrate via calculation.
Section 2.5.1 Calculating the Curl of the Electric Field Page 43
Section 2.5 Review of Basics of Electrostatics: The Electric Field has Vanishing Curl
Stokes’ Theorem (a mathematical theorem we will not prove here but that you saw in
Ma1abc) then tells us that, for any surface S with boundary C(S),
I Z h i
d~ ~ (~
`·E r) = r) · ∇
da nb(~ ~ (~
~ ×E r) = 0 (2.41)
C(S) S
We used above the fact that the line integral of the electric field
around any closed loop C vanishes. If we consider two points along
the loop ~ r2 , C defines two paths along the loop from ~
r1 and ~ r1 to
r2 , C1 and C2 . Let’s difference the line integrals along these two
~
paths, using the vanishing of the line integral around the loop to
see that the difference vanishes:
Z ~
r2 Z ~
r2 Z ~
r2 Z ~
r1
d~ ~ (~
`·E r) − d~ ~ (~
`·E r) = d~ ~ (~
`·E r) + d~ ~ (~
`·E r)
r1
C1 ,~ r1
C2 ,~ r1
C1 ,~ r2
C2 ,~
I
= d~ ~ (~
`·E r) = 0 (2.42)
C
(Be careful again about the endpoint ordering and signs of the two terms!) Therefore,
Z ~
r2 Z ~
r2
d~ ~ (~
`·E r) = d~ ~ (~
`·E r) (2.43)
r1
C1 ,~ r1
C2 ,~
The above relation tells us that the value of the above line integral depends only on
the location of its endpoints, not on the path taken. Thus, we can construct a
function, the electric potential, V (~
r ), defining it via its differences between points:
Z ~
r2
r2 ) − V (~
V (~ r1 ) ≡ − d~ ~ (~
`·E r) (2.44)
~
r1
The fundamental theorem of calculus for line integrals in multiple dimensions implies
Z ~
r2
r2 ) − V (~
V (~ r1 ) = d~ ~ (~
` · ∇V r) (2.45)
~
r1
~ (~
where ∇V r ) is the gradient of the electric potential. The above two formulae hold
regardless of choice of endpoints and path, so the integrands are equal and we have
~ (~
E ~ (~
r ) = −∇V r) (2.46)
The electric potential has units of (N m/C), which we call the volt, V. The electric
field is frequently written in units of V/m instead of N/C.
1
Z
r −~
~ r0
Z ~
r2
~ (~
E r) = dτ 0 ρ(~
r 0) and r2 ) − V (~
V (~ r1 ) ≡ − d~ ~ (~
`·E r)
4 π o V |~
r −~r 0 |3 ~
r1
We can use these to obtain an explicit expression for the potential in terms of the
charge distribution. In practice, trying to do the line integral explicitly using the
~ is tedious and not illuminating.
definition of E
Instead, let us use our understanding of the meaning of the mathematical expression
~ (~
E r ) = −∇V~ (~r ) to make an Ansatz. If we have a point charge at the origin, then the
electric field points radially outward and falls off as 1/r 2 . What function’s derivative
gives that dependence? V (~ r ) = 1/r . This suggests to us
1 1
Z
V (~
r) = dτ 0 ρ(~
r 0) (2.47)
4 π o V |~
r −~r 0|
We may then prove explicitly this form is correct by taking the gradient.
Section 2.6.2 Relation of the Electric Potential to the Charge Distribution Page 47
Section 2.6 Review of Basics of Electrostatics: The Electric Potential
1 1
Z
~ ~r V (~
−∇ r) = − dτ 0 ρ(~ ~ ~r
r 0) ∇ (2.48)
4 π o V |~
r −~r 0|
1 1 1 sb r −~
~ r0
~ ~r
∇ ~ ~r −~r 0
=∇ ~ ~s
=∇ =− 2 =− (2.49)
|~
r −~r 0| |~
r −~r 0| s s |~
r −~r 0 |3
where we used the formula for the gradient in spherical coordinates from Griffiths:
~ (~ ∂T 1 ∂T b 1 ∂T b
∇T r) = rb + θ+ φ (2.50)
∂r r ∂θ r sin θ ∂φ
1 r −~
~ r0
Z
~ (~
−∇V r) = dτ 0 ρ(~
r 0) ~ (~
=E r) (2.51)
4 π o V |~
r −~r 0 |3
Section 2.6.2 Relation of the Electric Potential to the Charge Distribution Page 48
Section 2.6 Review of Basics of Electrostatics: The Electric Potential
Aside on Techniques
It is important to recognize how we almost uniformly avoided brute-force calculations
of divergences, curls, and gradients so far. The only times we did those calculations
explicitly wer when we had rendered the calculations trivial. A key part of doing E&M
successfully and with minimal pain is avoiding algebra and calculus whenever possible
and instead making use of clever arguments of the type we used above. Only do
algebra and calculus as a last resort! There are two reasons for this.
First, the kinds of arguments we used are more physical and help you develop
intuition. For example, in proving the differential version of Gauss’s Law, at no point
~ ! Incredible, right? Instead, we proved that the
did we explicitly take derivatives of E
divergence of the 1/r 2 law is the delta function (again, not explicitly, but by referring
to the geometric proof we made for the integral version of Gauss’s Law) and used that
fact. We could have done the brute-force calculation in Cartesian coordinates, and it
would have given the same result. But you would have derived no intuition from it.
Second, brute-force calculations are prone to oversights — like the one about the sign
~ in the delta-function symmetry derivation — as well as bookkeeping
flip on ∇
mistakes — algebraic sign flips, misapplications of the product and chain rules, etc.
Doing brute-force calculations does not help you understand physics, or even
mathematics. Of course, sometimes brute-force calculations are needed, but try to
avoid them, and keep your wits and intuition about you as you do them!
It takes time to learn how to work this way, but we do derivations (rather than just
quote results) so you can learn these techniques.
Page 50
Lecture 3:
Basics of Electrostatics III:
Boundary Conditions on Electric Field and Potential
Poisson’s and Laplace’s Equations
Electric Potential Energy
Date Revised: 2022/01/06 17:40
Date Given: 2022/01/07
Page 51
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
Page 52
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
where E~ is evaluated over the two faces. We neglect the flux through the cylindrical
wall because we will let dz vanish in the end and so its area will vanish and it will
~ to have a delta-function
contribute no flux. Note that it is not possible for E
singularity that might prevent this vanishing: the most singular type of charge density
is a point charge, and even that only yields a 1/r 2 dependence.
Section 2.8.1 Boundary Condition on the Normal Component of the Electric Field Page 53
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
For Gauss’s Law, the volume integral of the charge density enclosed has two
contributions: from the non-delta-function-like volume charge density in the
half-cylinders and from any delta-function-like surface charge density on the surface.
The contribution of the former will vanish as we let dz → 0. The latter converts the
volume integral to a surface integral:
1 1
Z Z
F= r − S) σ(~
dτ δ(~ r) = da σ(~
r) (2.53)
o V o S
where S is the area at the interface intersected by the cylinder. (Note that this is a
case where the delta function’s argument requires some interpretation to understand
the delta function’s units. It is the S in the argument that implies the function has
units of m−1 rather than m−3 : it is picking out a surface rather than a point and thus
changing the units by one power of distance, not three.) Equating the two expressions
for F , letting dz → 0, and seeing that S1 , S2 → S as dz → 0 in the flux integral yields
1
Z h i Z
r) · E
da nb(~ ~2 (~ ~1 (~
r) − E r) = da σ(~
r) (2.54)
S o S
Section 2.8.1 Boundary Condition on the Normal Component of the Electric Field Page 54
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
This holds for any choice of cylinder and thus any S, so the integrands must be equal:
h
~2 (~ ~1 (~
i 1
r) · E
nb(~ r) − E r) = r)
σ(~ (2.55)
o
That is, the change in the normal component of the electric field across the interface
is proportional to the surface charge density at the interface. If there is no surface
charge at the interface, this component of the electric field must be continuous.
Section 2.8.1 Boundary Condition on the Normal Component of the Electric Field Page 55
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
where we neglect the contributions from the infinitesimal legs because they will vanish
as dz → 0. (Remember, as we discussed earlier, fields cannot have delta-function
singularities and thus they cannot make these infinitesimal portions of the integral
nonzero.)
Section 2.8.2 Boundary Condition on the Tangential Component of the Electric Field Page 56
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
Be careful about the signs of the integrals: d ~ ` for an open contour acquires its
orientiation from the the ordering of the endpoints; it has no intrinsic orientation
until this ordering is specified. Therefore, the sign of d ~` and of the endpoint ordering
do not multiply; they are redundant. Specifically, in this case, the endpoints imply
that d ~
` points along +bs for the second term and −b s for the first term and thus that
the integrands have opposite sign. Do not then think that the opposite polarity of the
endpoint ordering of the two terms implies another relative sign between the two
integrals, with the two relative signs canceling!
Section 2.8.2 Boundary Condition on the Tangential Component of the Electric Field Page 57
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
The fact that the curl of the electric field vanishes ensures the left side is zero.
We can combine the two terms on the right side by changing the endpoint ordering on
the first term and recognizing that C1 → C2 as dz → 0 (remember: C1 and C2
themselves have no orientation: the orientation of the line integrals is set by the
ordering of the endpoints). Thus, we have
Z ~
rb −b r ) dz
n(~ 2
Z ~
rb +b r ) dz
n(~ 2
Z ~
rb
dz→0
h i
0=− ~1 (~
E r) · d~
`+ ~2 (~
E r) · d~
` −→ ~2 (~
E ~1 (~
r) − E r) · d~
`
ra −b
C1 ,~ r ) dz
n(~ 2
ra +b
C2 ,~ r ) dz
n(~ 2
ra
C2 ,~
h i
r) · E
sb(~ ~2 (~ ~1 (~
r) − E r) = 0 (2.57)
This expression holds for any tb and thus sb parallel to the surface, so it tells us that the
tangential component of the electric field is continuous across any boundary
(regardless of whether there is surface charge present).
Section 2.8.2 Boundary Condition on the Tangential Component of the Electric Field Page 58
Section 2.8 Review of Basics of Electrostatics: Boundary Conditions on the Electric Field and Potential
1 h
~2 (~ ~1 (~
i h
~ 2 (~ ~ 1 (~
i
r) · E
r ) = nb(~
σ(~ r) − E r ) · −∇V
r ) = nb(~ r ) + ∇V r)
o
h
~ 2 (~ ~ 1 (~
i 1
r ) · ∇V
=⇒ nb(~ r ) − ∇V r ) = − σ(~
r) (2.58)
o
h i
=⇒ sb(~ ~ 2 (~
r ) · ∇V ~ 1 (~
r ) − ∇V r) = 0 (2.59)
1 ~ (~
~ ·E
r) = ∇
ρ(~ r ) = −∇2 V (~
r) (2.60)
o
1
∇2 V (~
r) = − r)
ρ(~ (2.61)
o
Poisson’s Equation is a partial differential equation. You know from basic calculus that
a differential equation alone is not sufficient to obtain a full solution V (~
r ): constants
of integration are required. For partial differential equations in multiple dimensions,
the constants of integration are given by specifying boundary conditions, conditions for
how the solution or its derivatives must behave on the boundary of the volume in
which we are specifying ρ(~ r ) and would like to determine V (~ r ).
Page 61
Section 2.9 Review of Basics of Electrostatics: Poisson’s and Laplace’s Equations
Our expression for the potential in terms of the charge distribution, Equation 2.47, is
r ) → 0 as
the explicit solution to this equation for a particular boundary condition, V (~
r → ∞. Section 3.11 will develop the concept of a Green Function, which is the
generic tool for solving Poisson’s Equation for arbitrary boundary conditions.
When there is no charge and the right side vanishes, Equation 2.61 is known as
Laplace’s Equation. The importance of this equation is that it implies that, in a region
where there is no charge, the second derivative vanishes everywhere, which implies
there can be no local maxima or minima (they would require a positive or negative
second derivative). We will prove this explicitly in Section 3.1.
For completeness, let’s also rewrite the curl-freeness of the electric field in terms of
the electric potential. There is a mathematical theorem that the curl of a gradient
always vanishes:
~ × (−∇V
∇ ~ )=0 (2.62)
Page 62
Section 2.10 Review of Basics of Electrostatics: Electrostatic Energy
Electrostatic Energy
Electric Potential Energy of a Point Charge in an Electric Field
Consider moving a point charge from ~ r2 along a contour C. The work done on
r1 to ~
the charge is given by doing the line integral of the negative of the electric force along
the path because that is the mechanical force that has to be exerted to move the
charge against the electric force F~e :
Z ~
r2
W12 = − d~
` · F~e (~
r) (2.63)
r1
C,~
That is, the work done on the charge by the mechanical force in going from ~ r1 to ~
r2 is
given by the charge times the change in electric potential between the two positions.
Note the sign: if the potential is higher at the end point, then the work done was
positive.
Section 2.10.1 Electric Potential Energy of a Point Charge in an Electric Field Page 63
Section 2.10 Review of Basics of Electrostatics: Electrostatic Energy
r2 ) − U(~
U(~ r2 ) − V (~
r1 ) = q [V (~ r1 )] (2.65)
That is, the electric potential energy of the charge and the electric potential of the
field are simply related. Since it was defined in terms of work done against a force,
electric potential energy obviously has units of Joules (J). That is explicit in the above
form, which is C (N m/C) = (N m) = J.
Note that the electric field can also do work on the charge. In this case, the sign in
the above line integral for the work is flipped and work is done as the charge loses
potential energy. In this case, the work done by the electric field on a charge is what
gives it the kinetic energy it has at the end: the electric potential energy is converted
to mechanical kinetic energy.
Section 2.10.1 Electric Potential Energy of a Point Charge in an Electric Field Page 64
Section 2.10 Review of Basics of Electrostatics: Electrostatic Energy
N i−1 N
1 X X qi qj 1 X qi qj
U= = (2.67)
4 π o i=1 j=1 |~
ri − ~
rj | 8 π o i,j=1,i6=j
|~
ri − ~rj |
where the factor of 1/2 was introduced to allow i and j to both run from 1 to N.
Generalizing this to a continuous charge distribution, we have
1 r 0)
r ) ρ(~
ρ(~
Z Z
U= dτ dτ 0 (2.68)
8 π o V V |~
r −~ r 0|
We can use the relations between potential, field, and charge density (Equations 2.6,
2.47, and 2.61) and the divergence theorem (Equation 2.20) to obtain an alternate
expression for the electric potential energy in terms of the electric field as follows:
1 ρ(~ r 0)
r ) ρ(~ 1 o
Z Z Z Z
dτ 0 dτ ∇2 V (~
U= dτ = dτ r
ρ(~ ) V (~
r ) = − r ) V (~
r)
8 π o V V |~
r −~ r 0| 2 V 2 V
o o
Z Z
ibp ibp
h i
= − ~ · V (~
dτ ∇ ~ (~
r ) ∇V r) + ~ (~
|∇V r )|2 with = ≡ integration by parts
2 V 2 V
divergence
o i Z
Z
theorem h
o
= da nb · V (~ ~ (~
r) E r) + ~ (~
|∇V r )|2 (2.69)
2 S(V) 2 V
In the last line, the first term is an integral of the product of the potential and the
field at the surface of the volume. In order to get the full energy of the charge
distribution, V must include all the charge. If we assume the charge distribution is
restricted to some finite volume, then V is naturally the volume containing the charge
distribution. But we can add volume that does not contain charge because it
contributes nothing to the initial expression for the electric potential energy.
Section 2.10.3 Electric Potential Energy in Terms of the Electric Field Page 66
Section 2.10 Review of Basics of Electrostatics: Electrostatic Energy
o i Z
Z h
~ (~ o ~ (~
U= da nb · V (~
r) E r) + |∇V r )|2 (2.70)
2 r →∞ 2 all space
Because the charge distribution is restricted to the finite volume V and thus looks like
a point charge as r → ∞, the field and potential fall off like 1/r 2 and 1/r . The surface
area of S only grows as r 2 , so the integral goes like 1/r and thus vanishes as r → ∞.
(If the charge distribution is not restricted to a finite volume, the surface term may
not vanish, requiring one to either keep the surface term or use the initial expression.)
It may seem strange that we can make this choice of S, as changing V and S affects
both integrals in the last expression. The explanation is that the choice of S changes
the two integrals but leaves their sum constant, and taking S to infinity simply zeros
out the first integral, leaving only the contribution of the second integral.
We thus find
o
Z
U= ~ (~
|E r )|2 (2.71)
2
where the integral is over all of space. Correspondingly, the quantity u = 2o |E~ |2 is an
energy density. We interpret this form as indicating that the potential energy created
by assembling the charge distribution is stored in the field: less charge implies a
smaller field and therefore less potential energy.
Section 2.10.3 Electric Potential Energy in Terms of the Electric Field Page 67
Section 2.10 Review of Basics of Electrostatics: Electrostatic Energy
Because the electric potential energy is a quadratic function of the electric field,
The energy of a sum of fields is more than just the sum of the energies of the
individual fields because there is a cross term due to potential energy of the presence
of the charges sourcing the second field in the first field (or vice versa, or half of both).
The infinite self-energy of a point charge reflects the fact that we do not know how to
assemble a point charge. In fundamental particle physics, the existence of point
charges such as the electron is an assumption, not a consequence, of the theory. In
fact, there is scheme, called “renormalization,” by which the infinite self-energy one
calculates for such a charge from Equation 2.71 is “subtracted off” in a self-consistent
fashion across all situations. While this practice is accepted and applied carefully, it is
not understood. String theory, which postulates that all particles are actually vibrating
string-like objects with finite extent, may offer a solution, but string theory currently is
not complete — it does not offer a way to calculate the Standard Model — and there
is no explicit proof it is correct.
Section 2.10.5 Self-energy and Point Charges vs. Continuous Charge Distributions Page 69
Lecture 4:
Basics of Electrostatics IV:
Conductors
Capacitors and Capacitance
Date Revised: 2022/01/10 05:00
Date Given: 2022/01/10
Page 70
Section 2.11 Review of Basics of Electrostatics: Electric Conductors
Electric Conductors
Definition and Behavior of a Conductor
We now talk about electric conductors, both because they are interesting and because
they provide a first opportunity to use boundary conditions to determine properties of
the charge distribution, field, and potential. Notice that we derive these properties
without explicit calculations!
Without any calculation, we know what the response of the conductor will be to an
externally applied electric field: If there is any field present in the conductor, positive
and negative charge densities will separate in response to the field. That separation
results in an additional field whose direction is opposite the applied field because of
the direction the two polarities of charge move in response to the applied field. This
movement occurs until the sum field vanishes, at which point there is no further force
on the charges and the system becomes static. Therefore, E ~ = 0 inside any conductor.
~ = 0 inside a
We may derive the following conductor properties from the fact that E
conductor everywhere:
The exact distribution of q on the surface depends on the geometry. For cases
with some symmetry, we may be able to guess the solution easily.
Consider a conductor with a spherical outer surface. Since there are no field
lines inside the conductor, there is no way the charge in the cavity or on the
inner surface of the conductor can influence the distribution of charge on the
outer surface, even if the inner cavity is non-spherical and/or the charge is not
placed at the center of the cavity. Thus, the charge must distribute itself on the
outer surface of the conductor in the same way as it would if charge +q were
added to a spherical conductor with no cavity. By symmetry, that distribution is
uniform with surface charge density σ = q/4 π r 2 .
Note, however, that, in general, the charge on the inner surface of the conductor
will not be distributed uniformly. It will only be uniform if the inner surface is
spherical and the charge in the cavity is at the center of the cavity, as this
situation has symmetry. (Note that the shape of the outer surface, or the inner
cavity’s location with respect to the outer surface, have no impact, for the same
reasons as the inner cavity does not affect the distribution of charge on the
outer surface.) In any other case, the field lines from the charge in the cavity
will exhibit no symmetry as they terminate on the cavity wall and therefore the
surface charge required to cancel those field lines in the conductor will have no
symmetry.
I If there is no net charge inside a cavity in a conductor, the electric field inside
the cavity vanishes, independent of the external field applied to or net charge
added to the conductor
We use proof by contradiction. Assume there is a
nonzero electric field in the cavity. Since there is no
charge in the cavity, the field lines must start and end
on charges on the surface of the cavity. RTherefore,
there is a path through the cavity with d ~ `·E ~ 6=
0. Now close the path with a segment inside the
conductor. This portion of the now-closed loop C
contributes nothing to the line integral C d ~ ~ over
H
`·E
the entire ~
c 2013 Griffiths, Introduc-
H loop because E = 0 inside the conductor.
tion to Electrodynamics Since C d ~`·E ~ = 0, the contribution from inside
the cavity must vanish also. Contradiction. So the
assumption E ~ 6= 0 in the cavity must be false.
Aside 1: Note the technique of proof by contradiction, which we will use again
in E&M.
Aside 2: This fact is used for shielding of experiments from external electric
fields (and also electromagnetic waves) and is called a Faraday cage. Note that
the conductor can have some net charge on it (and correspondingly sit at some
nonzero electric potential with respect to infinity) and this property still holds.
As we will see later, it also holds in the presence of external electromagnetic
waves, which is the more typical and important application.
~ = σ nb
E (2.72)
o
There is a charge density σ at this point, and an electric field above it, so is there a
force on the charge? Yes, but it is subtle to calculate. The thing to recognize is that
the small element of charge σ da in an infinitesimal area da cannot exert a force on
itself. The field the element of charge is subject to is the field of the charge
distribution excluding that charge element. We find this field by finding the field of
this charge element and subtracting it from the total field. This is an example of one
of the indirect approaches we must apply in E&M: a brute-force approach will not be
successful or generic.
Section 2.11.4 Surface Charge and the Force on the Surface of a Conductor Page 78
Section 2.11 Review of Basics of Electrostatics: Electric Conductors
We know (Griffiths Example 2.5) that the electric field of a charge sheet in the xy
plane is E~ = ±(σ/2 o ) zb where the sign applies depending on whether z > 0 or z < 0.
While the small patch we are considering is not an infinite sheet, it looks like one if we
are infinitesimally close to it. We also know E~other must be continuous at the charge
element because, in the absence of that charge element, there is no charge at the
boundary and thus no surface charge density to cause a discontinuity in the normal
component. (Note that we do not claim we know E ~other , only that we know that it
has this continuity property!) Thus, we may write the equations
E ~other + σ nb
~outside = E ~other − σ nb
~inside = E
E (2.73)
2 o 2 o
where E~other is the field due to the rest of the charge distribution excepting da and,
because of its continuity, the same expression for E ~other appears in both equations.
(Note this technique, which you learned doing story problems in middle-school
pre-algebra, of writing down an equation in which the knowns are not segregated on
one side.) Using E ~outside = (σ/o ) nb and/or E
~inside = 0, we find E
~other = (σ/2 o ) nb.
This is the field that acts on the charge σ da in da. Therefore, the force per unit area
is
F~ ~other
σ da E σ σ2
f~ = = =σ nb = nb (2.74)
da da 2 o 2 o
Section 2.11.4 Surface Charge and the Force on the Surface of a Conductor Page 79
Section 2.11 Review of Basics of Electrostatics: Electric Conductors
~ = (σ/o ) nb:
Writing the force per unit area in terms of the field at the surface E
σ2 o 2
f~ = nb = E nb (2.75)
2 o 2
That is, the surface of a conductor always feels an outward force. Consider what
would happen if you put charge on a balloon with a metallized surface.
Note the force per unit area, which has units of energy density, is actually equal to the
energy density just above the conductor. We could have in fact used the energy
density to derive the force: the force per unit area is the gradient of the energy per
unit area, and moving the conductor surface in or out by an infinitesimal distance dz
would have changed the total energy per unit area by u dz.
Note the indirect technique of proof. Again, we did no integral and we did not use
Coulomb’s Law explicitly.
Section 2.11.4 Surface Charge and the Force on the Surface of a Conductor Page 80
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
Capacitance
Consider two conductors (of arbitrary shapes) and suppose we put equal and opposite
charges Q and −Q on them. The potential difference ∆V between the two is of
course given by the line integral of the electric field from any point on the surface of
one to any point on the surface of the other. How does ∆V scale with the charges?
Q
C = (2.76)
∆V
One can talk about the capacitance of a single conductor with charge Q by implicitly
assuming there is another conductor at infinity that has charge −Q and is defined to
be at V = 0.
Now departing from Griffiths and instead following Jackson §1.11, we can generalize
capacitance to include multiple conductors by simply assuming a generalized linear
relationship between potentials, which we also call voltages, and charges as we argued
above must be true:
N
X
Vi = Dij Qj or V =DQ (2.77)
j=1
where V and Q are N-element column matrices for the voltages and charges on the N
conductors and D is a N × N matrix that connects the two. It is explicit that any
voltage depends linearly on all the charges. The capacitance matrix is then C = D −1 ,
with
N
X
Qi = Cij Vj or Q=CV (2.78)
j=1
This form serves to make it clear that the capacitance is not just a single quantity
between two conductors, but is more general. According to Jackson, the diagonal
element Cii is the “capacitance” of electrode i, and the Cij are termed the “coefficients
of induction” to convey that they indicate the charge induced on electrode i when a
voltage is placed on electrode j. As we will see below, neither of these is what one
would consider the capacitance of a pair of conductors as we discussed initially.
For the simple case of two mirror-symmetric electrodes with equal and opposite
charges ±Q and voltages ±V , we can relate the elements of the capacitance matrix to
the pair capacitance, which is what we usually call the capacitance (e.g., in Ph1b).
We can assume the following form for the capacitance matrix:
Cs −Cm
C = (2.79)
−Cm Cs
Why could we assume the above form? The symmetry of the system implies
C11 = C22 . We shall see below that all capacitance matrices are symmetric matrices,
so C12 = C21 . We chose the negative sign on C12 = −Cm with some foreknowledge of
the result, but that’s a choice and doesn’t affect the value of C12 .
The defining condition of the pair capacitance is that equal and opposite charges are
placed on the two conductors. By symmetry, we can conclude that the conductors
carry equal and opposite voltages (not true for a non-mirror-symmetric configuration).
Thus
Q (Cs + Cm ) V Cs + Cm
C = = = (2.82)
∆V 2V 2
After we have discussed energy, we will return to this system for a more detailed
analysis of what one can say about Cs and Cm .
Page 87
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
∆V (q) ∆V (q) q
dU = dq − − = ∆V (q) dq = dq (2.83)
2 2 C
Note that ∆V is a function of q here: the voltage is not held fixed while the charge is
moved; rather, the voltage and charge increase together (linearly).
1
Z Q 1 Q2
U= q dq = (2.84)
C 0 2 C
Alternatively, using Q = C ∆V ,
1 Q2 1
U= = C (∆V )2 (2.85)
2 C 2
We could have modeled the above process differently. Our transferral of dq from one
electrode to the other is the equivalent of taking charge dq from the negative voltage
electrode, carrying it out to infinity (where we set V = 0), and bringing it back and
putting it on the positive voltage electrode. The equivalence is because the voltage
difference between two points is path-independent. This process is, then, equivalent to
bringing charges dq and −dq in from infinity and putting them on the positive and
negative voltage electrodes, respectively. And the last process is equivalent to bringing
the charges in consecutively rather than simultaneously because we proved earlier the
potential energy does not depend on the order of assembly of the charge distribution.
The above picture is what we need for considering a multi-electrode system: we build
up the charge on each conductor by bringing in charge from infinity and calculating the
work done. Consider bringing charge dqi in from infinity and adding it to electrode i.
The change in the electric potential energy of the system due to adding this charge is
N
X
dUi = Vi dqi = Dij qj dqi (2.86)
j=1
There are two possible double-countings we must avoid: 1) This infinitesimal element
of charge dqi is moved from V = 0 at infinity to V = Vi on the ith electrode, so the
voltages of the other electrodes are irrelevant during this infinitesimal charge transfer
and we should not bring them into the equation; 2) Because the charges on all the
other electrodes j 6= i are physically immobile as dqi is brought in, no work is done on
them, and so there are no other contributions to include (as strange as it may seem
given that their voltages change by dVj = Dji dqi ).
Now, let’s integrate over dqi . We will later do a sum over i. The ordering of the two
steps does not matter because we proved earlier that the electric potential energy does
not depend on the order of assembly. But we do need to worry about the order of how
we have brought in the charges because we should not calculate cross-terms for
charges that do not yet exist. Let’s assume that, if we are integrating the ith charge,
then the first i − 1 charges have already been integrated to their full values {Qj },
j = {1, . . . , i − 1}, and the remaining N − i electrodes j = {i + 1, . . . , N} have no
charge on them yet. Thus, the voltage Vi (qi ; {Qj }j<i ) is given by
N
X i−1
X
Vi (qi ; {Qj }j<i ) = Dij qj = Dii qi + Dij Qj (2.87)
j=1 j=1
Next, we need to sum over i to account for the charging up of all the electrodes:
N N Xi−1
1X X
U= Dii Qi2 + Dij Qi Qj (2.89)
2 i=1 i=1 j=1
N N N
1X 1 X 1 X
U= Dii Qi2 + Dij Qi Qj = Dij Qi Qj
2 i=1 2 i,j=1,i6=j
2 i,j=1
1 T −1
U= Q C Q (2.90)
2
1 T
U= V CV (2.91)
2
Let’s check that this gives the correct result for an elementary capacitor with two
mirror-symmetric electrodes having equal and opposite charges ±Q and voltages ±V .
Using the capacitance matrix we derived earlier (recall, C11 = C22 = Cs and
C12 = C21 = −Cm ),
1
C11 (+V )2 + C22 (−V )2 + C12 (+V )(−V ) + C21 (−V )(+V )
U=
2
1 1
= V 2 [Cs + Cs + Cm + Cm ] = 2 C V 2 = C (∆V )2 (2.92)
2 2
as expected.
1
Uij = Dii Qi2 + Djj Qj2 + Dij Qi Qj (2.93)
2
1
Uji = Dii Qi2 + Djj Qj2 + Dji Qi Qj (2.94)
2
In our initial discussion of the electric potential energy, we argued that the
charging order does not matter. So we may equate the two, Uij = Uji .
Recognizing that Qi and Qj are arbitrary then implies
Section 2.12.4 Properties of the Capacitance Matrix and Its Inverse Page 94
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
1
Ui (all others grounded) = Cii Vi2 (2.96)
2
Since the energy should be positive (it takes work to add charge dqi in the
presence of the same-sign charge qi , as is done when charging up the electrode),
Cii must be positive.
1
Ui (all others neutral) = Dii Qi2 (2.97)
2
Section 2.12.4 Properties of the Capacitance Matrix and Its Inverse Page 95
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
1
dUij = Dii Qi2 + Djj qj dqj + Dij Qi dqj (2.98)
2
(The self-terms and cross-terms vanish for all the electrodes k 6= i, j because
they have Qk = 0.) If we consider the case of Qi , qj positive, and if we bring in
more positive charge dqj , it is obvious that both the change in the jth
self-energy and the energy cross-term should be positive: we are bringing
positive charges in proximity to existing positive charges. We already know the
self-energy terms are positive. In order for the energy cross-term to be positive,
Dij must be positive. In the mirror-symmetric electrode case, we would see via
explicit inversion of C that D’s off-diagonal elements are positive.
Another way to see that the cross-terms must be positive is to recall that the
entire expression must be consistent with our original expression for the electric
potential energy, Equation 2.68. That expression could be broken down into
three integrals, one for each self-energy term and one for the cross-term. When
the charge density is positive, all contributions to that expression are manifestly
positive.
Section 2.12.4 Properties of the Capacitance Matrix and Its Inverse Page 96
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
1
dUij = Cii Vi2 + Cjj vj dvj + Cij Vi dvj (2.99)
2
We already know the first and second terms are positive. The third term is more
challenging. If we want to increment a positive voltage vj by a positive amount
dvj , we need to put positive charge on it. This positive charge will draw
negative charge out of the battery holding Vi constant: some of the field lines
of that new charge on electrode j have to terminate on electrode i if Cij is
non-zero. Again, from Equation 2.68, we know that contribution to the electric
potential energy must be negative even if Vi is positive. Thus, the energy
cross-term must be negative, which requires Cij to be negative. (If Vi is
negative, that implies Qi is negative. It takes positive work to add negative
charge to an electrode that already has negative charge on it, so the cross-term
becomes positive as it should.)
Section 2.12.4 Properties of the Capacitance Matrix and Its Inverse Page 97
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
I | i6=j Cij | ≤ |Cjj |: for a given electrode, the sum of the off-diagonal elements
P
of the capacitance matrix is no larger in magnitude than the corresponding
diagonal element.
Just consider the same situation as just considered. The change in the charge
on the jth electrode is dqj = Cjj dvj . The field lines from those added charges
will terminate either on other electrodes or infinity, so the total negative charge
added to all the other electrodes can be no larger in magnitude than |dqj |.
Therefore,
X
X X
dq i = Cij dvj
≤ dqj = Cjj dvj =⇒ Cij ≤ Cjj (2.100)
i6=j i6=j i6=j
Section 2.12.4 Properties of the Capacitance Matrix and Its Inverse Page 98
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
Maybe we can then do this for the one mirror-symmetric case whose full electric field
configuration we can calculate trivially, the infinite parallel-plate capacitor? No! The
infinite parallel-plate capacitor violates the condition V = 0 at infinity because, if
either plate has non-zero potential, that plate’s non-zero equipotential surface extends
off to infinity in the transverse direction. We violate the assumption that allowed us to
ignore the electrode at infinity. Moreover, infinity is no longer even an equipotential
surface in this configuration! On the equipotentials defined by the two electrodes (at,
e.g., z = ±d/2), the potential at infinity is the potential of the corresponding
electrode. If the two plates have equal and opposite potentials, then the field outside
the plates vanishes and the potential on the surface of that volume at infinity is zero.
The potential on the line z = 0 is also zero. And then, for 0 < |z| < d/2 and
x, y → ∞, the potential is the same linear function of z that it would be at x, y = 0.
Clearly, our assumptions are violated!
Section 2.12.5 Capacitance Matrix of a Mirror-Symmetric Configuration Revisited Page 99
Section 2.12 Review of Basics of Electrostatics: Capacitors and Capacitance
So, we are stymied. In order for the V = 0 at infinity condition to be satisfied, our
electrodes must be finite in extent. But, for electrodes finite in extent, we cannot
calculate the potential in a trivial fashion, so we cannot determine Cs and Cm , or even
C , trivially. We need to develop the full machininery for solving Poisson’s and
Laplace‘’s Equations, which we will begin to do soon.
Page 101
Section 3.1 Advanced Electrostatics: Intuitive Approach to Laplace’s Equation
1 r −~
~ r0 1 r 0)
ρ(~
Z Z
~ (~
E r) = dτ 0 ρ(~
r 0) and V (~
r) = dτ 0 (3.1)
4 π o V |~
r −~r 0 |3 4 π o V |~
r −~ r 0|
are always correct but can be difficult to deal with in practice. Most systems will not
have symmetries that make the integrals easily doable (or avoidable via Gauss’s Law).
Moreover, and this is the greater problem, it is rare that one completely specifies ρ(~
r)
in setting up a problem. Experimentally, what we can control are the shapes,
positions, and potentials (voltages) of conductors. We do not control how the charge
arranges itself on the conductors. Thus, we need to seek alternate ways to solve for
the potential and field over all of space. Laplace’s and Poisson’s Equations are the key.
Page 102
Section 3.1 Advanced Electrostatics: Intuitive Approach to Laplace’s Equation
d 2V
=0 (3.2)
dx 2
V (x) = m x + b (3.3)
Note that these are the only choices in one dimension. Specifying dV /dx at two
points either yields a contradiction (if two different values of dV /dx are given) or
insufficient information (if the same value is given). There are no other quantities to
specify: all higher derivatives vanish thanks to Laplace’s Equation.
1 1
[V (x + a) + V (x − a)] = [(m (x + a) + b) + (m (x − a) + b)]
2 2
= m x + b = V (x) (3.4)
x1 d 2V
dV dV
Z
= + dx = 0 + 0 = 0 (3.5)
dx x1 dx
x0 x0 dx 2
Page 105
Section 3.1 Advanced Electrostatics: Intuitive Approach to Laplace’s Equation
I The value V (~r ) of a solution to Laplace’s Equation at any point is equal to the
average of its value on any sphere centered on that point in the region of
interest:
da0 V (~
r 0)
R
Sa (~
r)
r ) = hV (~
V (~ r )ia ≡ R (3.6)
Sa (~
r) da0
where Sa (~
r ) is the sphere of radius a centered on ~ r . This is straightforward to
show (Griffiths Problem 3.37). Let’s integrate Laplace’s Equation over the
volume enclosed by Sa (~ r ), Va (~
r ), and use the divergence theorem:
Z Z
0= dτ 0 ∇~2r 0 V (~
r 0) = da0 nb(~ ~ ~r 0 V (~
r 0) · ∇ r 0)
Va (~
r) Sa (~
r)
Z
= da0 nb(~ ~ ~r 0 −~r V (~
r 0) · ∇ r 0) (3.7)
Sa (~
r)
where Sa (~s = ~0) is the sphere of radius a centered on the origin of the ~s system
(i.e., the same as the sphere of radius a centered on ~ r 0 coordinate
r in the ~
2
system). If we pull the factor a outside of the integral, the integral is now over
the spherical angles in the ~s coordinate system, while the derivative is in the
radial direction in this coordinate system. Thus, we can pull the derivative
outside the integral too, yielding
1 ∂ V 1 ∂
Z Z
0= a2 dΩs = dΩs V (~s ) (3.9)
4 π a2 s =~
Sa (~ 0) ∂s s=a 4 π ∂a Sa (~s =~0)
One could also prove this by a technique similar to the 1D case, calculating ∇V ~
at any point ~ r 0 in the region by doing a line integral of Laplace’s Equation from
the candidate extremum ~ ~ vanishes at the candidate
r to that point. Since ∇V
extremum (because it is an extremum of V ), and the integrand (∇2 V ) of the
line integral vanishes by Laplace’s Equation, ∇V ~ vanishes at ~ r 0.
Section 3.1.2 Laplace’s Equation in Multiple Dimensions Page 108
Section 3.2 Advanced Electrostatics: Uniqueness Theorem
Uniqueness Theorem
Before obtaining a solution of Laplace’s and Poisson’s Equations, we prove some
uniqueness theorems we will need. This section draws from Jackson §1.8 and §1.9.
If the volume under consideration is not bounded by a surface on which we specify the
boundary conditions, then we must also specify a boundary condition at infinity.
The proof of the Uniqueness Theorem will not show why only one of these types of
boundary conditions may be specified. That proof will be provided soon, in §3.4.1.
The left side vanishes because of the boundary condition (any type). The first term on
the right side vanishes by Laplace’s Equation. Thus, we have
Z
~ 3 |2 = 0
dτ |∇V =⇒ ~ 3 (~
∇V r) = 0 =⇒ V3 = constant (3.15)
V(S)
where we take the second step because the integrand is nonnegative. This result
implies that our two candidate solutions V1 (~
r ) and V2 (~
r ) differ by at most a constant.
Hence, uniqueness is proven.
Section 3.2.3 Generic Uniqueness Proof for Poisson’s Equation Page 111
Section 3.2 Advanced Electrostatics: Uniqueness Theorem
Given the above, we may state/prove three special cases of the uniqueness theorem,
the ones given in Griffiths:
Let each conductor i have surface Si and charge Qi . Since we know the surface
charge density on each conductor is related to the normal component of the
electric field at that conductor, we may see
1 1
I I
da nb(~ ~ (b
r) · E r) = da σ(~
r) = Qi (3.16)
Si o Si o
Now, as before, let’s assume that there are two different solutions V1 (~ r ) and
r ) and their difference is V3 = V2 − V1 . Let’s evaluate the left-hand side of
V2 (~
Equation 3.14 for the BC we are specifying here:
I XI I
~ 3=− ~3 = − ~3
X
da V3 nb · ∇V da V3 nb · E V3,i da nb · E (3.17)
S i Si i Si
where we were able to pull V3 out of the integrals because V1 and V2 have
equipotentials on each surface and so therefore does V3 (with values V3,i , which
we do not need to know). The surface integral of the normal component of E ~3
over each Si vanishes because, as we indicated above, specifying Qi specifies
~1 and E
this surface integral to be the same for E ~2 , so the surface integral
vanishes for E~3 = E
~2 − E~1 . Thus, the LHS of Equation 3.14 also vanishes for
this BC, and so the remainder of the proof of uniqueness carries through.
Note how this proof relied on the boundary surfaces being conductors! Knowing
the total charges on nonconducting boundary surfaces would not be sufficient.
Method of Images
Overview: The Basic Idea of Method of Images
The method of images uses the concept of uniqueness of solutions to Poisson’s
Equation. Basically, given a physical setup involving a true charge distribution ρ(~r)
and Dirichlet boundary conditions for some volume V, one tries to replace the region
outside of V with an image charge distribution ρimage (~ r ) such that, when the image
charge’s potential is summed with that of ρ(~ r ), the potential on the boundary is the
same as that specified by the Dirichlet BC.
The technique works because of the uniqueness theorem: since the potential due to
the image and original charges matches the boundary conditions and satisfies
Poisson’s Equation with the same source term inside V, it is the solution to Poisson’s
Equation for that source term and choice of boundary conditions.
The imagined charge distribution is called image charge because, at least in the
example of the boundary condition being imposed by the presence of a conductor, the
image charges appear to be a (possibly distorted) mirror image of the original charges
through the boundary. “Image charge” is also used (somewhat erroneously) to refer to
the surface charge induced on a conducting boundary that sources the potential that
one models as due to the image charge.
Note that the image charge must be placed outside the volume V because we may not
r ) inside V; that would change the problem we are trying to solve.
change ρ(~
We will see later how the potential due to the image charge distribution (the induced
surface charge) is a component of the particular problem’s Green Function.
Section 3.3.1 Overview: The Basic Idea of Method of Images Page 114
Section 3.3 Advanced Electrostatics: Method of Images
The potential clearly satisfies V (z = 0) = 0. Let’s use this solution to do some other
calculations:
" #
∂ V q z −d z +d
σ = − o = −
∂z z=0 4 π (x 2 + y 2 + (z − d)2 )3/2 3/2
(x 2 + y 2 + (z + d)2 )
z=0
q d
=− (3.19)
2 π (x 2 + y 2 + d 2 )3/2
Section 3.3.2 A Point Charge near a Grounded Infinite Conducting Plane Page 115
Section 3.3 Advanced Electrostatics: Method of Images
We can calculate the total induced surface charge:
Z ∞ Z 2π ∞
−q d 1 1
Qind = r dr dφ = q d √ = −q (3.20)
0 0 2 π (r 2 + d 2 )3/2 r 2 + d 2 0
This is equivalent to just calculating the force on the real charge exerted by the
image charge, which is in general a valid approach. Whether to calculate the
image charge potential and take the gradient or calculate the image charge force
is a matter of choice and convenience.
Section 3.3.2 A Point Charge near a Grounded Infinite Conducting Plane Page 116
Section 3.3 Advanced Electrostatics: Method of Images
∞ 1 q2 ∞ dz 1 q2
Z Z
U=− (−F (z)) dz = − =− (3.22)
d 4 π o 4 d z2 4 π o 4 d
Note that this result is half what one would get for the potential energy of two
equal and opposite point charges separated by a distance 2d:
1 q2
Ualt = − (3.23)
4 π o 2d
There are two ways to understand this. The first is to recognize that, unlike in
the case of two point charges, no energy is gained or lost in moving the negative
charge because it is in the conductor, where V = 0 and thus q V = 0
everywhere. The second is to recognize that the above expression is the energy
stored in all of space in the field of two point charges, but, in this case, the field
is only real in the z > 0 half-space and so the integrated energy is reduced by a
factor of 2.
Section 3.3.2 A Point Charge near a Grounded Infinite Conducting Plane Page 117
Lecture 7:
Advanced Electrostatics III:
Method of Images (cont.)
Green Functions
Date Revised: 2022/01/19 10:15
Date Given: 2022/01/19
Page 118
Section 3.3 Advanced Electrostatics: Method of Images
By symmetry, the appropriate image charge must be on the z axis. Let its value be q 0
and its position be b zb, where b may be positive or negative. We can find q 0 and b by
requiring that V = 0 at ~ r = ±R zb:
1 q q0
0 = V (+R zb) = +
4 π o a − R R −b
1 q q0
0 = V (−R zb) = +
4 π o a + R R +b
R R2
=⇒ q 0 = −q 6= −q b= (3.24)
a a
We see that both values are always physically reasonable because R < a. In particular,
b < R so the image charge remains outside V (i.e., inside the sphere), as we expect.
Note that q 0 6= −q!
Section 3.3.3 A Point Charge near a Grounded, Conducting Sphere Page 119
Section 3.3 Advanced Electrostatics: Method of Images
We can use the above expression to see that the boundary condition V (r = R) = 0 is
satisfied:
q 1 R/a
V (r = R, θ, φ) = −q
p 2 2
4 π o R sin θ + (a − R cos θ)2
2
R 2 sin2 θ + ( Ra − R cos θ)2
q 1 1
= −q
q
4 π o
2 2 2 2 2 2
R sin θ + (a − R cos θ) a sin θ + (R − a cos θ)
=0 (3.27)
Section 3.3.3 A Point Charge near a Grounded, Conducting Sphere Page 120
Section 3.3 Advanced Electrostatics: Method of Images
~ = ∂V /∂r :
Let’s calculate the induced surface charge from nb · ∇V
∂V
σ = − o (3.28)
∂r r =R
2
R
q R sin2 θ − (a − R cos θ) cos θ R R sin2 θ − ( a − R cos θ) cos θ
‘= −
4 π (R 2 sin2 θ + (a − R cos θ)2 )3/2
3/2
a
2
R 2 sin2 θ + ( Ra − R cos θ)2
2
q R − a cos θ a2 R − Ra cos θ
= − 2
4 π (R 2 + a2 − 2 a R cos θ)3/2 R (a2 + R 2 − 2 a R cos θ)3/2
2
R2
q R(1 − Ra 2 ) q R 1− a2
= =−
4 π (R 2 + a2 − 2 a R cos θ)3/2 4 πR 2 a
3/2
R2 R
1+ a2
−2 a
cos θ
One can show by integration that the total induced charge is q 0 . In this geometry, this
makes sense because the volume enclosed by a surface integral of electric field flux at
the boundary encloses the volume containing the image charge. This example
illustrates one case of the theorem stated earlier; in this case, the total induced surface
charge is equal to the image charge. We will see other cases illustrated in the next
example.
The force on the point charge and the electric potential energy can be calculated in a
manner similar to that used for the conducting plane.
Section 3.3.3 A Point Charge near a Grounded, Conducting Sphere Page 121
Section 3.3 Advanced Electrostatics: Method of Images
Physically, this is because the calculation of the total induced surface charge via
Gauss’s Law must be done differently. One method is to use a spherical surface just
outside the boundary, so it is in the conducting volume where the field vanishes. This
implies that the sum of the real and induced charge vanishes, so the induced charge is
the negative of the real charge.
The other method is to put the surface just inside the boundary. Now, the charge
enclosed is only the real charge. As the surface approaches the boundary, though, the
flux integral is equal to the negative of the integral of the surface charge density (up
to o ) because the electric field near a conductor is σ/o (with the negative because
the field is pointed inward). So this tells us the total induced surface charge is the
negative of the real charge too.
Thus, we see illustrated another case of the theorem we stated earlier, that the total
induced surface charge is the image charge, the negative of the real charge, or some
combination of the two. Which one depends on the geometry: is the boundary outside
the volume of interest, inside, or some combination of the two?
In the case of the point charge outside the conducting sphere, we noted that the
Gauss’s Law calculation, with the Gaussian sphere just inside the volume V (i.e.,
having radius infinitesimally larger than a), yields q 0 6= −q. The distinction is whether
the volume V of interest is “outside” the boundary (neglecting the boundary at
infinity) as in the previous case or “inside” the boundary as in this case.
(In the previous case, the Gauss’s Law calculation outside V (i.e., using a Gaussian
sphere of radius less than a) yields no useful information because the sphere doesn’t
contain the induced surface charge. The flux through such a sphere vanishes because
the field is zero inside the conductor, which just tells us that all the induced surface
charge resides, well, on the surface.)
What is the solution for the V 6= 0 sphere on its own? Certainly, the sphere is an
equipotential with the desired value V0 . By symmetry (remember, the point charge is
not present for this problem), the charge is uniformly distributed on the surface. Thus,
we can apply Gauss’s Law to the problem, which tells us that the potential of the
sphere is identical to that of a point charge at the origin. To figure out the value of
the point charge, we require that the point charge’s potential match the boundary
condition:
q0 R
= V0 =⇒ q0 = 4 π o V0 R =⇒ V (r ) = V0 (3.29)
4 π o R |~
r|
q0 Q − q0 Q + q Ra
q0 = Q − q 0 =⇒ V0 = = = (3.31)
4 π o R 4 π o R 4 π o R
Q + q Ra
" #
q 1 R/a
V (r ≥ R, θ, φ) = − 2
+ (3.32)
4 π o |~
r − a zb| r − Ra zb|
|~ 4 π o |~
r|
Notice that this reduces to our original point charge near a sphere solution not when
Q = 0 but rather when Q = q 0 = −q R/a, which is the charge that must flow onto
the sphere for it to stay at V = 0 (i.e., grounded).
1 1
Z
r 0 ) ∇~2r 0
dτ 0 V (~ 0
− 0
∇~2r 0 V (~
r 0)
V(S) |~
r −~r | |~
r −~r |
1 1
I
= r 0 ) nb(~
da V (~ ~ ~r 0
r 0) · ∇ − nb(~ ~ ~r 0 V (~
r 0) · ∇ r 0) (3.33)
|~
r −~ 0
r | |~
r −~ 0
r |
S
Section 3.4.1 Integral Equation for the Electric Potential Page 126
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
1
∇~2r 0 = −4 π δ(~ r 0)
r −~ (3.34)
r0 −~
|~ r|
~ ~r 0 · ~r −~r0 ~ ~r 0 1 r −~
~ r0
∇ = 4 π δ(~ r 0)
r −~ and ∇ =−
|~
r −~r 0 |3 r0 −~
|~ r| |~
r −~r 0 |3
Using the above expression for the Laplacian of |~ r −~ r 0 |−1 , doing the integral over the
delta function, applying Poisson’s Equation, moving the second term on the right side
to the left side, and multiplying everything by − 41π yields, now only for ~ r ∈ V(S):
1
Z 0
r )
ρ(~
r ∈ V(S)) =
V (~ dτ 0 (3.35)
4 π o V(S) |~
r −~ r 0|
1 1 1
I
+ da nb(~ ~ ~r 0 V (~
r 0) · ∇ r 0 ) − V (~
r 0 ) nb(~ ~ ~r 0
r 0) · ∇
4π S |~
r −~ 0
r | |~
r −~ 0
r |
r 0 ∈ V(S)).
r 6∈ V(S) because the integral was over ~
(The left side vanishes for ~
Section 3.4.1 Integral Equation for the Electric Potential Page 127
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
This is a formal equation for the electric potential. The boundary conditions are
present on the right side: in the case of Dirichlet, we specify V (~ r 0 ) for ~
r 0 ∈ S, while
in the case of Neumann, we specify nb(~ ~ ~r 0 V (~
r 0) · ∇ r 0 ) for ~
r 0 ∈ S. Our Uniqueness
Theorem says we should only need to specify one or the other at any given point on
the boundary. In fact, since the Uniqueness Theorem says that knowing one specifies
the other (knowing one gives the full solution, which determines the other), we don’t
have the freedom to specify both independently! Knowing both essentially requires
knowing the solution to the problem. For example, if we consider the simplest possible
case of specifying an equipotential on the boundary, then knowing the other boundary
term requires knowing the normal gradient of the potential at the boundary, which is
equivalent to knowing the surface charge density on the boundary. We would not be
able to guess this except in cases with sufficient symmetry.
r 0)
Therefore, this is not a closed-form solution but rather an integral equation for V (~
r 0 ∈ V(S) ∪ S: the boundary condition does not provide everything on the right
for ~
side, but, if we know the solution, it will satisfy the equation.
So far, however, this integral equation is not very useful. Once we have introduced the
concept of Green Functions, we will see its utility.
Section 3.4.1 Integral Equation for the Electric Potential Page 128
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
O~r f (~
r ) = g (~
r) (3.36)
where O~r is a linear partial differential operator taking derivatives with respect to the
coordinate ~r , f is a generalized potential, and g is a generalized source function.
Poisson’s Equation is an example, with O~r = −o ∇2 , f (~ r ) = V (~
r ), and g (~
r ) = ρ(~
r ).
Is there a general approach for finding f given g ?
Yes, there is, it is called the Green Function approach. The basic idea is to find the
“impulse” response function for the differential equation: the generalized potential one
gets if one has a point-like source. Given the impulse response function, and the
linearity of O~r , one can obtain the generalized potential for an arbitrary source
function by convolving the impulse response function with that source function.
r 0 ) = δ(~
r, ~
O~r G (~ r 0)
r −~ (3.37)
We can check that Equation 3.36 is satisfied by this solution by applying the operator:
Z Z
dτ 0 G (~ r 0 ) g (~
r 0) = dτ 0 O~r G (~ r 0 ) g (~
r 0)
O~r f (~
r ) = O~r r, ~ r, ~ (3.39)
Z
= dτ 0 δ(~
r −~ r 0 ) g (~
r 0 ) = g (~
r) (3.40)
Note how this check relied on the linearity of O~r , which allowed us to bring it inside
the integral. Assuming solutions to the generalized linear partial differential equation
are unique (true for Poisson’s Equation), the Green Function is the only solution we
need to find.
We have already seen that, in the absence of a bounding surface and assuming the
potential falls off at infinity, the Green Function for Poisson’s Equation is
1 1
G (~ r 0) =
r, ~ if V = all space, V (r → ∞) → 0 (3.41)
4 π o |~
r −~r 0|
We can see this explicitly by rewriting our usual expression for the potential for this
boundary condition, Equation 2.47, in terms of G :
1 r 0)
ρ(~
Z Z
V (~
r) = dτ 0 = dτ 0 G (~ r 0 ) ρ(~
r, ~ r 0) (3.42)
4 π o V |~
r −~ r 0| V
Section 3.4.3 General Discussion of Green Functions for Poisson’s Equation Page 131
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
More generally — i.e., for a more complex boundary condition — Poisson’s Equation
implies that its Green Function must decompose into the form
1 1
G (~ r 0) =
r, ~ r 0)
r, ~
+ F (~ with ∇~2r F (~ r 0) = 0
r, ~ (3.43)
4 π o |~
r −~r 0|
where the first term provides the right side of Poisson’s Equation but the second term
is not only allowed by Poisson’s Equation but, we will see, is crucial for satisfying the
boundary conditions for any situation except the trivial one noted above, that of the
potential vanishing at infinity. The F term plays multiple roles, depending on the type
of boundary condition, and we will explain those roles later. Finding G thus consists of
finding F .
Section 3.4.3 General Discussion of Green Functions for Poisson’s Equation Page 132
Lecture 8:
Advanced Electrostatics IV:
Green Functions (cont.)
Obtaining Green Functions from the Method of Images
Date Revised: 2022/02/09 06:30
Patched up errors in derivation of
Dirichlet and Neumann Green Functions
Date Given: 2022/01/21
Page 133
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Note that this equation does not match Equation 3.37, which had the Laplacian
acting on ~ r 0 . We will recover Equation 3.37 later. We then apply the same
r , not ~
kinds of manipulations we did to obtain the integral equation for the potential,
Equation 3.35 (these manipulations rely on Equation 3.44), giving
Z
V (~
r) = dτ 0 ρ(~
r 0 ) G (~ r 0)
r, ~ (3.45)
V
I h i
+ o da0 G (~ r 0 ) nb(~
r, ~ ~ ~r 0 V (~
r 0) · ∇ r 0 ) − V (~
r 0 ) nb(~ ~ ~r 0 G (~
r 0) · ∇ r 0)
r, ~
S(V)
We see that, if we can find the appropriate G for a particular boundary condition and
force the term involving the other boundary condition to vanish, our integral equation
for V (~
r ) reduces to an integration over the source distribution with the Green Function
and over the boundary condition with the Green Function or its normal gradient.
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 134
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
We also see that, even though we assumed Equation 3.44 instead of Equation 3.37 for
the equation defining the Green Function, the result we obtain above is consistent
with Equation 3.38, which states that the source function ρ(~ r 0 ) should be convolved
with the Green Function, integrating over its second argument, to obtain the potential
function in its first argument. We will resolve this apparent inconsistency shortly.
Note that the equation we obtain for V (~ r ) is different from the integral equation for
V (~
r ), Equation 3.35, because there we could not impose such a condition on V (~ r ),
since it is set by the situation under consideration, or on |~ r −~r 0 |−1 (obviously).
r 0 ) is, on the other hand, our tool for solving that integral equation, so we may
r, ~
G (~
design the tool to do its job as long as it respects its defining equation.
We can be more specific about what we mean by “forcing the other BC term to
vanish” by picking a type of boundary condition:
r ∈ S. Therefore, nb(~
r ) is specified for ~
In this case, V (~ ~ ~r V (~
r) · ∇ r ) should be left
unspecified—it should be determined by the solution itself—so we need for it to
not appear in the integral equation. We can eliminate the term containing this
normal derivative if we require the Dirichlet Green Function, GD (~ r 0 ), to satisfy
r, ~
GD (~ r 0) = 0
r, ~ for r 0 ∈ S, ~
~ r ∈ V, S (3.46)
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 135
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Note that we want the above condition to hold for not just ~ r ∈ V but also for
r ∈ S so the expression is usable for calculating the potential on the boundary
~
to ensure the boundary condition remains satisfied (i.e., the expression for V (~
r)
is self-consistent).
Using the interpretation implied by the convolution of the charge density with
the Green Function in Equation 3.45 (admittedly, an interpretation not
obviously consistent with the defining equation, Equation 3.44), the above
condition is equivalent to requiring that charge on the boundary (~r 0 ∈ S) yield
no contribution to the potential elsewhere on the boundary (~ r ∈ S) or in the
volume (~r ∈ V). In one sense, this is what we expect, as the Dirichlet boundary
condition specifies V (~
r ) on the boundary, so any charge that appears on the
boundary to enforce that boundary condition had better do so in a way that
does not modify the boundary condition.
However, in another sense, it is the opposite of what we expect: how can the
induced surface charge on the boundary not affect the potential on the surface
or in the volume? Wasn’t that the whole idea behind the method of images,
that one calculates the additional potential of the induced surface charge on the
boundary by replacing it with an image charge? We resolve this confusion below.
With the above condition, the solution for V (~
r ) reduces to
Z I
V (~
r) = dτ 0 ρ(~
r 0 ) GD (~ r 0 ) − o
r, ~ da0 V (~
r 0 ) nb(~ ~ ~r 0 GD (~
r 0) · ∇ r 0)
r, ~ (3.47)
V S(V)
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 136
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 137
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
When this symmetry property is applied to Equation 3.44, and we also use the
symmetry of the delta function, Equation 3.37 is recovered (after relabeling
r ↔~
~ r 0 ). This resolves the apparent inconsistency between wanting the Green
Function to satisfy Equation 3.37 but having to assume Equation 3.44 at the
start to get Equation 3.45.
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 138
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 139
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
" I #−1
nb(~ 0 ~
r ) · ∇~r 0 GN (~
r, ~ 0
r ) = − o da 0
for r0 ∈S
r ∈ V, S, ~
~ (3.48)
S(V)
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 140
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Z I
V (~
r) = dτ 0 ρ(~
r 0 ) GN (~ r 0 ) + o
r, ~ da0 GN (~ r 0 ) nb(~
r, ~ ~ ~r 0 V (~
r 0) · ∇ r 0 ) + hV (~
r )iS(V)
V S(V)
da0 V (~
r 0)
H
S(V)
with hV (~
r )iS(V) ≡ H (3.49)
S(V) da0
While V (~r ) on the boundary has not been completely eliminated, its only
appearance is via its average value on the boundary. This makes sense, as the
Neumann boundary condition does not specify the potential offset since it only
specifies derivatives of the potential. The appearance of this term reflects the
freedom we have to set the potential offset for problems with Neumann
boundary conditions. Recall that the Uniqueness Theorem only showed
uniqueness up to an overall offset.
What is the interpretation of a Neumann Green Function? Since
nb(~ ~ ~r 0 V (~
r 0) · ∇ r 0 ) specifies the surface charge density on the boundary, GN (~ r 0)
r, ~
simply calculates the potential at a point ~ r in the volume due to this boundary
surface charge density at ~ r 0 . Note that GN is convolved with the volume charge
density and the surface charge density in the same way, reinforcing this
interpretation. A Neumann Green Function thus has a simpler interpretation
than a Dirichlet Green Function. There is no interpretation of GN or FN as
calculating contributions from image charge.
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 141
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
To make further progress in obtaining a functional form for the Green Function, we
must specify the boundary conditions in more detail. We will consider examples of this
next.
Section 3.4.4 Green Functions for Poisson’s Equation with Dirichlet or Neumann Boundary Conditions Page 142
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
1 q q
V (~
r) = − (3.50)
4 π o |~
r − d zb| |~
r + d zb|
We can see by inspection that the Dirichlet Green Function is given by taking
r 0 and −d zb in the second
q = 1 and by replacing d zb in the first term with ~
r 0 mirrored through the x 0 y 0 plane:
term with ~
1 1 1
r 0) =
r, ~
GD (~ − (3.51)
4 π o |~
r −~r 0| r − (x 0 xb + y 0 yb − z 0 zb)|
|~
r 0 ) = q δ(~
One can test this by plugging into Equation 3.46 with ρ(~ r 0 − d zb).
The second term accounts for the fact that induced charge appears on the
grounded conducting plane and calculates the contribution to the potential due
r 0 ) term while the first term is the usual Coulomb’s Law term.
r, ~
to it; it is the F (~
The first term solves Poisson’s Equation while the second term solves Laplace’s
Equation. Both terms depend on the position of the point charge at ~ r 0.
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 143
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 144
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 145
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
So, we see that the surface term serves to add the potential offset that the
boundary condition V (z = 0) = V0 requires. Therefore, the solution is now
1 q q
V (~
r) = − + V0 (3.54)
4 π o |~
r − d zb| |~
r + d zb|
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 146
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
This example serves to highlight the fact that one has to be careful about the
self-consistency of boundary conditions, especially when they involve a condition
at infinity. Consider two alternate BCs:
I One cannot set V (z = 0) = V0 and V (r → ∞) = 0 because that is not
self-consistent for z = 0, (x, y ) → ∞: should the BC be V0 or 0 for this
part of the boundary?
I One cannot even require V (z = 0) = V0 and V (z → p ∞) = 0 because it
leaves unspecified the boundary condition for V (z, x 2 + y 2 → ∞). If
one then thinks about what type of BC to specify there, one finds that it
should be impossible to specify something that is consistent with
V (z → ∞) = 0. Think about the case of the conductor held at V0 and
no point charge. We know the solution is a uniform sheet of surface
charge on the conductor, and we know that the field is then a constant
E~ (~ r ) = −(σ/o ) z. This potential
r ) = (σ/o ) zb and the potential is V (~
does not vanish as z → ∞. If one knows that a set of boundary
conditions is not self-consistent for the case of no point charge, then
linearity/superposition tells us there is no way to fix the inconsistency by
adding charges to V: one would have to add a potential that is also not
self-consistent to cancel out the self-inconsistency of the q = 0 potential!
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 147
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
R
" #
1 q q a
V (~
r) = − (3.55)
4 π o |~
r − a zb| R2
|~
r− a
zb|
r 0 = a zb and taking q = 1:
Thus, the Dirichlet Green Function is given by letting ~
0 1 1 R/r 0
r, ~
GD (~ r )= − (3.56)
4 π o |~
r −~r 0| r −~
~
2
r 0 (rR0 )2
Again, the second term accounts for the potential due to the charge induced on
the surface of the sphere and is the term that solves Laplace’s Equation in this
r 0 ) term). And again, one can this test form for GD by
r, ~
situation (the F (~
plugging into Equation 3.46 with ρ(~ r 0 ) = q δ(~
r 0 − a zb).
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 148
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
It is perhaps not so obvious that the second term in this Green Function is
symmetric in its arguments. Let’s rewrite it:
R/r 0 R R
2
= = p (3.57)
|~ r 0 (rR0 )2 |
r −~ r r r 0 − R 2 rb0 |
|b (r r 0 )2 + R 4 − 2 r r 0 R 2 rb · rb0
The same point about the surface integral term as for the conducting plane
r 0 ) = 0 for ~
holds here: that term vanishes because V (~ r 0 ∈ S.
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 149
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
In this case, we can see the effect of the surface integral term in Equation 3.47
because V (~r ) on the boundary does not vanish. The integral term is, from
Equation 3.47:
I
−o da0 V (~
r 0 ) nb(~ ~ ~r 0 GD (~
r 0) · ∇ r 0)
r, ~ (3.58)
S(V)
When we encountered this nonvanishing surface term for the prior case of a
point charge near a conducting plane, we recognized that V (~ r 0 ) = V0 could be
pulled outside the integral and that the integral of the normal gradient of the
Green Function gives the total charge induced on the boundary for a unit charge
at ~
r . To calculate that total induced charge, we invoke the theorem (based on
Gauss’s Law) we discussed earlier. In this case, the surface encloses the image
charge, so the total induced charge is equal to the image charge. That is:
R
I
−o da0 V (~
r 0 ) nb(~ ~ ~r 0 GD (~
r 0) · ∇ r 0 ) = −V0 Qind = −V0 qimage = V0
r, ~
S(V) r
(3.59)
This is again just the potential due to a point charge at the origin whose
magnitude is such that the potential at radius R is V0 .
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 150
Section 3.4 Advanced Electrostatics: Formal Solution to Poisson’s Equation: Green Functions
This is what we found earlier when we discussed the same problem using the
method of images.
The prior situation is identical to this one: specifying the charge on a conductor
is the same as specifying its potential. So the result for V (~
r ) is the same, where
we must take V0 = (Q + (R/a)q)/(4 π o R). Note that, even though we are
talking about a boundary condition in which charge is specified, it is not a
Neumann boundary condition because we do not specify σ(~ r 0 ∈ S), we are still
r 0 ∈ S).
effectively specifying V (~
Section 3.4.5 Obtaining Green Functions from the Method of Images Page 151
Lecture 9:
Advanced Electrostatics V:
Separation of Variables: General Theory
Separation of Variables in Cartesian Coordinates
Date Revised: 2022/01/24 04:30
Date Given: 2022/01/24
Page 152
Section 3.5 Advanced Electrostatics: Introduction to Separation of Variables
We will only develop separation of variables for Laplace’s Equation and, in the near
term, we will only apply it to solving problems with specific types of boundary
conditions rather than trying to use it to find the F piece of the Green Function.
(Recall, F satisfies Laplace’s Equation while G satisfies Poisson’s Equation.) We will
see later, at the tail end of our discussion of separation of variables in spherical
coordinates, that this technique will actually be sufficient to obtain the Green
Function for an arbitrary geometry, which then provides us the solution to Poisson’s
Equation. (One will be able to see that it is not feasible to do separation of variables
for Poisson’s Equation in the same way we do it for Laplace’s Equation: the process
very much relies on the vanishing of one side of the equation!)
where integration over the interval of interest [s, t] is the Hilbert space inner product.
Completeness is defined to be
X
fp∗ (w 0 ) fp (w ) = δ(w 0 − w ) (3.61)
p
X
g (w ) = Ap fp (w ) (3.62)
p
Z t
Ap = dw 0 fp∗ (w 0 ) g (w 0 ) (3.63)
s
We could have derived Equation 3.63 also by applying orthornomality to the expansion
Equation 3.62; this is the usual way we think of finding the {Ap } as we will see below.
They are of course equivalent derivations.
V (~
r ) = X (x) Y (y ) Z (z) (3.64)
d 2X d 2Y d 2Z
Y (y ) Z (z) + X (x) Z (z) + X (x) Y (y ) =0
dx 2 dY 2 dz 2
1 d X2 1 d 2Y 1 2
d Z
2
+ + =0 (3.65)
X (x) dx Y (y ) dY 2 Z (z) dz 2
We have three terms, the first a function of x, the second of y , and the third of z.
Given these mismatched dependences, the only way the equation can hold is if each
term is a constant. That is, it must hold that
1 d 2X 1 d 2Y 1 d 2Z
= K1 = K2 = K3 (3.66)
X (x) dx 2 Y (y ) dY 2 Z (z) dz 2
with K1 + K2 + K3 = 0.
Page 157
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
We know that the solution to these ordinary differential equations are exponentials,
p p
X (x) = A exp(x K1 ) + B exp(−x K1 ) (3.67)
p p
Y (y ) = C exp(y K2 ) + C exp(−y K2 ) (3.68)
p p
Z (z) = E exp(z −(K1 + K2 )) + F exp(−z −(K1 + K2 )) (3.69)
We have not specified which of K1 , K2 , and K3 are positive and which are negative
(clearly, they cannot all be the same sign). That will be determined by the boundary
conditions. Note that we are also neglecting linear solutions that also satisfy the
individual ordinary differential equations; we will see they are not necessary in the
examples we consider here (though they may be needed more generally).
At this point, we cannot make further generic progress; we need to apply a set of
boundary conditions. These will place constraints on the allowed values of the
exponents and coefficients and restrict the family of solutions. There are a number of
examples in Griffiths. To avoid duplication, we use a different one here from Jackson
§2.9.
Page 158
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
Example 3.4: Empty box with five walls grounded and one held at a potential
Consider a box with side lengths a, b, and c in the x, y , and z dimensions and with
one corner at the origin. The boundary conditions are
V (x = 0) = 0 V (y = 0) = 0 V (z = 0) = 0 (3.70)
V (x = a) = 0 V (y = b) = 0 V (z = c) = φ(x, y ) (3.71)
where φ(x, y ) is a function that is given. In SoV, we always apply the homogeneous
BC first because we will see they restrict the functional form of the solutions. The
homogeneous BC in the ith dimension (e.g., y ) can only be satisfied if the ith
function (e.g., Y (y )) satisfies it alone because it must be satisfied for all values of the
other coordinates. Let’s do x, y first for convenience (with foreknowledge of solution):
p p
X (0) = A + B = 0 X (a) = A exp(a K1 ) + B exp(−a K1 ) = 0 (3.72)
p p
Y (0) = C + D = 0 Y (b) = C exp(b K2 ) + D exp(−b K2 ) = 0 (3.73)
Reducing,
h p p i
A exp(a K1 ) − exp(−a K1 ) = 0 (3.74)
h p p i
C exp(b K2 ) − exp(−b K2 ) = 0 (3.75)
Page 159
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
There is no solution to these equations for K1 > 0 and K2 > 0: the unit-normalized
decaying and rising exponentials are only equal when their arguments both vanish, and
they do not. Therefore, let’s take K1 = −α2 and K2 = −β 2 so these become
oscillating exponentials. We thus obtain the conditions
nπ mπ
αn = βm = n, m positive integers (3.77)
a b
where n and m may only be positive integers because negative values are redundant
with the positive ones and n = 0 and m = 0 yield vanishing functions. Thus, we have
∞
X ∞
X
X (x) = An sin αn x Y (y ) = Cm sin βm y (3.78)
n=1 m=1
where the {An } and {Cm } are constants to be determined. These solutions clearly
respect the V = 0 boundary conditions at x = 0, a and y = 0, b because they vanish
at those points.
Page 160
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
Z (0) = E + F = 0 =⇒ F = −E (3.79)
q
Vnm (x, y , z) = Anm sin(αn x) sin(βm y ) sinh(γnm z) with γnm = α2n + βm
2 (3.82)
0 into a single
where we have combined all the arbitrary coefficients Am , Cn , and Enm
coefficient Anm . Each Vnm (~
r ) satisfies all five homogeneous BC.
Page 161
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
Now, we want to apply the last boundary condition, V (x, y , z = c) = φ(x, y ). How?
Not the same way as we applied the previous ones. The prior boundary conditions
were homogeneous, meaning that they forced the solution to vanish somewhere. The
remaining one is inhomogeneous because it requires the solution to take on a particular
functional form on a boundary. These must be treated differently, for two reasons.
I The first involves linearity and uniqueness. Because the right-hand side of a
homogeneous BC is zero, the BC is satisfied by any linear combination of
functions that satisfy the BC. The same is not true of inhomogeneous BC. If it
were possible for two different functions to satisfy the inhomogeneous BC, then
only a subset of linear combinations of them would satisfy the same BC: the
linear combinations in which the coefficients sum to unity. This condition
violates linearity. The only resolution is for there to be precisely one solution to
the inhomogeneous BC. This requirement is consistent with uniqueness: the
inhomogeneous BC is applied last, and it completes the application of the BC,
so the solution should be unique once it is applied.
I From the purely calculational point of view, requiring the solution for a given n,
m to satisfy the inhomogeneous boundary condition would imply
There simply is not enough freedom in the functional form on the left to satisfy
the boundary condition for arbitrary φ(x, y ).
Page 162
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
The only way to have enough freedom to satisfy the inhomogeneous boundary
condition is to consider a linear combination of the individual n, m:
∞
X
V (~
r) = Anm sin(αn x) sin(βm y ) sinh(γnm z) (3.85)
n,m=1
where Anm are now constants to find based on requiring the above linear combination
solution satisfies the inhomogeneous boundary condition at z = c, which now becomes
∞
X
φ(x, y ) = V (x, y , z = c) = Anm sin(αn x) sin(βm y ) sinh(γnm c) (3.86)
n,m=1
This condition will let us determine the Anm , but how, and why are we certain they
exist? We make use of the theory of orthonormal functions we cited earlier.
Page 163
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
p
We will use the fact (not proven here) that the functions { 2/a sin(αn x)} for n ≥ 1
form a complete, orthonormal set p on the x ∈ [0, a] interval (with the given boundary
conditions at x = 0, a), as do { 2/b sin(βn y )} for m ≥ 1 on y ∈ [0, b]. Therefore,
we may recover the Anm by multiplying by them and integrating:
Z a Z b r r
2 2
dx dy φ(x, y ) sin(αp x) sin(βq y )
0 0 a b
Z a Z b ∞ r r
X 2 2
= dx dy Anm sinh(γnm c) sin(αp x) sin(αn x) sin(βm y ) sin(βq y )
0 0 n,m=1
a b
∞ r r √
X a b ab
= Anm sinh(γnm c) δpn δqm = Apq sinh(γpq c) (3.87)
n,m=1
2 2 2
Now,p be aware that wep did more work than necessary above. Once we are told that
the { 2/a sin(αn x) 2/b sin(βm y )} form an orthonormal set, we do not need to do
the integrals on the right-hand side! We only need write the right-hand side of the
original equation in terms of the orthonormal functions, then use orthonormality
(Equation 3.63) to obtain the equations for the individual coefficients; i.e.:
r ∞ r r
ab X 2 2
φ(x, y ) = Anm sin(αn x) sin(βm y ) sinh(γnm c) (3.88)
4 n,m=1 a b
a b
r r r
2 2 ab
Z Z
=⇒ dx dy sin(αp x) sin(βq y ) φ(x, y ) = Apq sinh(γpq c) (3.89)
0 0 a b 4
Page 164
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
Next, we move the coefficients to one side to obtain (replacing pq with mn):
1
Z a Z b 2 2
Anm = dx dy sin(αn x) sin(βm y ) φ(x, y ) (3.90)
sinh(γnm c) 0 0 a b
∞
4 X sinh(γnm z) a 0 b 0
Z Z
V (~
r) = sin(αn x) sin(βm y ) dx dy φ(x 0 , y 0 ) sin(αn x 0 ) sin(βm y 0 )
a b n,m=1 sinh(γnm c) 0 0
(3.91)
A good exercise is to write down the solutions for the five other inhomogeneous
boundary condition cases (especially the ones with the inhomogeneous condition on
the x, y , or z = 0 planes) “by inspection” — i.e., by simply changing the solution we
already have by replacing z with x, y , or a − x, b − y , or c − z — rather than by
rederiving. Clearly, these other problems are not different in any conceptual way, they
are only different calculationally, and only barely. There is no reason to redo all that
calculation from scratch!
Page 165
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
Page 166
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
Referring back to our discussion of Green Functions, the above solution is the surface
term in Equation 3.47 for the particular boundary condition we have applied. By
comparison of the two expressions, we infer (not derive!)
− o nb(~ ~ ~r 0 GD (~
r 0) · ∇ r 0 = x 0 xb + y 0 yb + c zb)
r, ~ (3.92)
∞
4 X sinh(γnm z)
= sin(αn x) sin(βm y ) sin(αn x 0 ) sin(βm y 0 )
ab n,m=1
sinh(γnm c)
Note that this expression does not fully specify GD (or FD )! The above information is
sufficient for the particular physical situation we have set up, which consists of no
physical charge in the volume and the above boundary condition, because:
I The term consisting of the integral of the charge density in the volume
convolved with GD is zero in this case because the charge density vanishes in
the volume. Therefore, we do not need to know GD (or FD ) completely.
I The above surface term is the only one needed because V = 0 on the other
boundaries.
Page 167
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
For the more general problem of an arbitrary charge distribution in the volume and
arbitrary Dirichlet boundary conditions on the surfaces, we would need to find the full
GD . It may seem like one could do as suggested earlier, finding the solution for each
option for which face is held at nonzero potential, then using the results analogous to
the above as six Neumann boundary conditions on GD , and applying separation of
variables to find GD . But one would have to require that GD solve Poisson’s Equation
for a unit point charge, not Laplace’s Equation. This, as we noted earlier, is not
feasible with separation of variables because of the nonzero right side of the equation.
There is a way to deal with this, which we will show a bit later when we develop the
spherical harmonic expansion for the Green Function in spherical coordinates.
Another approach that does work would be the method of images with the condition
V = 0 on all the surfaces. It is left as an exercise for the reader to think about what
set of image charges is appropriate; the situation gets complicated for a charge at an
arbitrary position in the box, but it is solvable. Certainly, from the resulting GD , we
could compute the normal gradient of GD on any surface and thus obtain the general
solution for V in the volume for any Dirichlet boundary condition. We should find that
the normal gradient of GD on the z = c surface is what is given above.
It may seem like separation of variables is unsatisfactory for this reason — the
procedure does not give you the full Green Function, while the method of images
does. But, as we have seen, the method of images is not a systematic procedure —
one has to guess the correct image charge distribution. By contrast, separation of
variables is an entirely algorithmic procedure to give you a solution if a separable one
exists for the particular boundary condition you are applying. It is less general but
more reliable. More importantly, we will show later how, by applying separation of
variables in a more sophisticated way, we can in fact find the full Green Function.
Page 168
Section 3.6 Advanced Electrostatics: Separation of Variables in Cartesian Coordinates
There is, nevertheless, no guarantee that there will be a separable solution; this
depends on the geometry of the boundary conditions. The boundary conditions need
to respect the separability assumed. For example, a boundary condition on a spherical
boundary would not likely yield a solution via separation of variables in Cartesian
coordinates!
Note also that the method of images technique is not appropriate for a Neumann
boundary condition because the method of images solution generally solves the V = 0
Dirichlet BC problem. One needs a technique like separation of variables for such
cases.
Page 169
Lecture 10:
Advanced Electrostatics VI:
Separation of Variables in Spherical Coordinates: General Theory
Separation of Variables in Spherical Coordinates
with Azimuthal Symmetry
Date Revised: 2022/01/26 06:30
Date Given: 2022/01/26
Page 170
Section 3.7 Advanced Electrostatics: Separation of Variables in Spherical Coordinates: General Theory
1 ∂ ∂V 1 ∂V 1 ∂2V
∂
r2 + sin θ + 2
=0 (3.93)
r 2 ∂r ∂r r2 sin θ ∂θ ∂θ r2 sin θ ∂φ2
1 d dR 1 1 d dΘ 1 d 2Φ
sin2 θ r2 + sin θ + =0 (3.95)
R(r ) dr dr Θ(θ) sin θ dθ dθ Φ(φ) dφ2
We see that first term depends only on r and θ while the second term depends only on
φ, so we can immediately assume they are each equal to a constant:
1 d 2Φ
= −m2 (3.96)
Φ(φ) dφ2
The choice of the form of the constant is motivated by what will come next, but we
can see why it needs to be of this form. As we saw in Cartesian coordinates, the above
differential equation is solved either by growing/decaying exponentials (right side
positive) or oscillating exponentials (right side negative). Since φ is a coordinate that
repeats on itself (φ = 2 n π are the same physical coordinate) the solutions Φ(φ) must
also be periodic, forcing the choice of the oscillating exponential. (For the same
reason, the linear solutions we ignored in the Cartesian case are disallowed here.) We
saw before that it is convenient to define the constant to incorporate a squaring.
1 d dR 1 1 d dΘ
sin2 θ r2 + sin θ = m2 (3.98)
R(r ) dr dr Θ(θ) sin θ dθ dθ
1 d dR 1 1 d dΘ m2
r2 + sin θ − 2
=0 (3.99)
R(r ) dr dr Θ(θ) sin θ dθ dθ sin θ
Now, we see that the first term depends only on r and the second only on θ, so we can
separate again by setting the two terms equal to constants that sum to zero. Here, we
rely on prior knowledge of the result to choose the constant to be `(` + 1) so that
1 d dR
r2 = `(` + 1) (3.100)
R(r ) dr dr
1 1 d dΘ m2
sin θ − = −`(` + 1) (3.101)
Θ(θ) sin θ dθ dθ sin2 θ
Note that the radial equation does not depend on m. This implies that the R(r )
functions will not depend on the azimuthal properties of the problem, in particular
whether it has azimuthal symmetry. But R(r ) depends on `, so it will depend on the
polar properties of the problem. Θ(θ) depends on ` and m, so its behavior depends on
both the polar and azimuthal properties of the problem. Φ(φ) looks like it may only
depend on the azimuthal properties because it depends only on m, but m is tied to `
through the polar equation, so there will be some relationship.
Here, we add another item to our “bag of tricks” and define U(r ) by R(r ) = U(r )/r
and plug in. (This is motivated by the r 2 that the second d/dr must act on: assuming
this dependence gets rid of the extra terms arising because of that factor.) We find
d 2U `(` + 1)
− U(r ) = 0 (3.102)
dr 2 r2
Since the two derivatives would reduce the exponent of a power-law solution by 2, and
the second term does the same by dividing by r 2 , the above equation suggests U(r ) is
a power law in r . (Or, try making it work with a transcendental function: you can’t.)
If we plug in such a form U(r ) = r a , we find
U(r ) B
=⇒ R(r ) = = A r a1 −1 + B r a2 −1 = A r ` + `+1 (3.104)
r r
1 d dΘ m2
sin θ + `(` + 1) − Θ(θ) = 0 (3.105)
sin θ dθ dθ sin2 θ
Motivated by the fact that sin θ dθ = −d(cos θ), we add another trick to our bag of
tricks by writing
d dP m2
(1 − x 2 ) + `(` + 1) − P(x) = 0 (3.107)
dx dx 1 − x2
Section 3.7.3 The Polar Equation and the Generalized Legendre Equation Page 175
Section 3.7 Advanced Electrostatics: Separation of Variables in Spherical Coordinates: General Theory
As you have seen in ACM95/100, differential equations of this type can be solved by
assuming the solution is a polynomial in x and requiring termination after a finite
number of terms. That is, one assumes
∞
X
P`m (x) = ak x k (3.108)
k=1
and then, plugging the above form into the differential equation, one requires the
series to terminate (ak = 0 for some k). This condition forces ` to be a nonnegative
integer and −` ≤ m ≤ `. (We already know m is an integer to ensure Φ(φ) is
single-valued.) These polynomials are the associated Legendre polynomials.
Mathematically, there should be a second solution for each `, m because the equation
is second order. These are the solutions one finds by not requiring termination but
simply convergence for −1 < x < 1 (corresponding to 0 < θ < π). If one has a
geometry that excludes the z-axis (where these solutions diverge), these solutions
must be considered. If the z-axis is in the space, then these solutions are unphysical
and can be discarded.
Section 3.7.3 The Polar Equation and the Generalized Legendre Equation Page 176
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Consider the special case of azimuthal symmetry, for which m = 0 and Φ(φ) =
constant. The generalized Legendre Equation reduces to the Legendre Equation:
d dP
(1 − x 2 ) + `(` + 1) P(x) = 0 (3.109)
dx dx
The same series solution applies here with m = 0, so ` must still be a nonnegative
integer. These solutions are the Legendre Polynomials. One can show they obey
Rodrigues’ Formula:
`
1 d
`
P` (x) = x2 − 1 (3.110)
2` `! dx
Section 3.8.1 The Polar Equation Solution with Azimuthal Symmetry: the Legendre Equation and Legendre Polynomials Page 177
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
1
r r
2` + 1 2`0 + 1
Z
dx P` (x) P`0 (x) = δ` `0 (3.111)
−1 2 2
∞
X 2` + 1
P` (x)P` (x 0 ) = δ(x − x 0 ) (3.112)
`=0
2
The coefficients {A` } and {B` } are set by the boundary conditions. If the volume
includes the origin and the boundary conditions imply the potential must be finite
there, the {B` } may be eliminated, and, if the volume includes infinity and the
boundary conditions require the potential be finite (usually zero) there, the {A` } may
be eliminated. In other cases, some or all of the {A` } and {B` } can be nonzero.
Usually, application of the boundary conditions on V will require use of the
orthonormality relations for the Legendre polynomials.
We note that, in the process of doing separation of variables, we have proven that the
angular solution satisfies the eigenvalue-eigenfunction equation
`(` + 1)
∇2 P` (cos θ) = − P` (cos θ) (3.114)
r2
For the angular equation, r acts as a constant and so appears in the eigenvalue.
Section 3.8.3 Full Solution to Laplace’s Equation with Azimuthal Symmetry Page 179
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
We will start first with a case in which the boundary condition is quite obviously
Dirichlet and the application is very much like what we did in Cartesian coordinates.
Generally speaking, however, boundary conditions are not always so obvious. One has
to use whatever information one is given and turn it into boundary conditions of the
type that we know provides uniqueness.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 180
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Let’s consider the two cases together. If the space is r < R, then we require the {B` }
to vanish to ensure a finite potential at the origin. (There is no charge in the volume,
so we are assured that the potential cannot be infinite there.) If the space is r > R,
then we require the {A` } to vanish so the potential goes to zero at infinity. That is:
∞ ∞
X X B`
V (r , θ) = A` r ` P` (cos θ) or V (r , θ) = P (cos θ)
`+1 `
(3.115)
`=0 `=0
r
To apply the boundary condition at R, we evaluate the above equations at that value:
∞ ∞
X X B`
V (R, θ) = A` R ` P` (cos θ) or V (R, θ) = P (cos θ)
`+1 `
(3.116)
`=0 `=0
R
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 181
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Then, to find the coefficients, we apply orthornomality to both sides, as we did for
separation of variables in Cartesian coordinates. For the case of r < R, we have:
2` + 1
Z π
sin θ dθ V (R, θ) P` (cos θ) (3.117)
2 0
∞ π 2` + 1
Z
X 0
= A`0 R ` sin θ dθ P` (cos θ) P`0 (cos θ) (3.118)
0 2
`0 =0
∞
X 0
= A`0 R ` δ``0 = A` R ` (3.119)
`0 =0
which we can solve for A` . Or, based on the orthornormality relation Equation 3.111,
we can just state by inspection (yielding the same result as the above calculation):
2` + 1 1
Z π
A` = sin θ dθ V (R, θ) P` (cos θ) (3.120)
2 R` 0
Notice how R ` appears in the formula for A` . This is analogous to the same way that
sinh(γnm c) appeared in the solution for the coefficients Anm in the Cartesian case
(Equation 3.90).
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 182
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
2 ` + 1 `+1
Z π
B` = R sin θ dθ V (R, θ) P` (cos θ) (3.121)
2 0
∞ Z π
X 2 ` + 1 r`
V (r < R, θ) = `
P ` (cos θ) sin θ 0 dθ 0 V (R, θ 0 ) P` (cos θ 0 ) (3.122)
`=0
2 R 0
∞ Z π
X 2 ` + 1 R `+1
V (r > R, θ) = `+1
P ` (cos θ) sin θ 0 dθ 0 V (R, θ 0 ) P` (cos θ 0 ) (3.123)
`=0
2 r 0
Notice how the units of the coefficients have come out to cancel the powers of r in
the solution so our result has the same units of electrostatic potential as the boundary
condition.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 183
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 184
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 185
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Note the use of two different solutions in the two regions: this is a generally useful
technique.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 186
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Griffiths does the example of an uncharged metal sphere in a uniform electric field in
the z direction, E ~ = E0 zb. The boundary condition is a bit mixed again. Because the
sphere is metal, it is an equipotential. But that doesn’t specify the value of V on the
sphere. Since the field is uniform, we cannot set V to vanish at infinity. Instead,
V (z = 0) = 0 is chosen. From that choice and the fact that the equipotential sphere
is in contact with z = 0, we can conclude that the sphere satisfies V = 0. But now V
at infinity is not specified, so we don’t yet have a Dirichlet boundary condition. The
sensible thing to do is to require the potential approach V (~ r ) = −E0 z at infinity:
whatever induced charge the sphere picks up, its contribution to the potential and
field must fall off at infinity, leaving only the uniform field. Now we have a Dirichlet
boundary condition. Because the potential is allowed to diverge at infinity, we cannot
eliminate the {A` } in this case. But it is easy to see that only A1 is nonzero: for
` > 0, the behavior goes like r ` , and since the potential must go like z = r cos θ at
large r , all the ` > 1 terms must vanish. This large r behavior sets A1 = −E0 . A0 = 0
because the potential has no offset. That leaves the {B` } to be determined.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 187
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
∞
X B`
0 = A1 R cos θ + P (cos θ)
`+1 `
(3.124)
`=0
R
∞
X B`
−A1 R cos θ = P (cos θ)
`+1 `
(3.125)
`=0
R
Since the left side has a ` = 1 term, and the Legendre polynomials are orthonormal,
there can also be only a ` = 1 term on the right side, implying B` = 0 for ` 6= 1 and
B1 /R 2 = −A1 R or B1 = E0 R 3 . Thus, the solution is
R3
r ) = −E0
V (~ r− cos θ (3.126)
r2
Note the use of a nontrivial boundary condition at infinity and the need to realize that
the sphere has the same potential as the z = 0 plane; without these boundary
conditions, it would have been impossible to start the problem.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 188
Lecture 11:
Advanced Electrostatics VII:
Separation of Variables in Spherical Coordinates
with Azimuthal Symmetry (cont.)
Date Revised: 2022/01/31 09:30
Revised lecture break
Date Given: 2022/01/28
Page 189
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Let’s reconsider the situation we looked at before via method of images, the point
charge near the conducting sphere. The setup is as before, with the point charge at
a zb and the sphere centered on the origin with radius R and V = 0 on its surface. One
difficulty is that the presence of the point charge implies Laplace’s equation is not
satisfied in the full volume! It is, however, satisfied separately in the regions R < r < a
and a < r < ∞, and we have the charge density at r = a, so we should somehow solve
separately in the two regions and then join the solutions together (as we did before for
the spherical shell of charge, which we recast as a Neumann boundary condition
(Example 3.6)).
Since we have seen how the method of images can provide the Green Function for a
system, the aforementioned equivalence suggests that we may be able to use
separation of variables to find the full Green Function for a system in the “sum over
orthonormal functions” form rather than in the “system of point charges form.” This
is indeed true and we will do this in general fashion for spherical coordinates later in
§3.9.4 using a technique similar to the one we use for this example.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 190
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
We may guess that the appropriate way to write the charge density at r = a is
q
σ(θ, φ) = δ(θ) (3.127)
2 π a2 sin θ
The rationale for this guess is that a2 sin θ cancels the r 2 sin θ portion of the volume
element and 2 π cancels the φ integral. It has the right units, too, surface charge
density, charge/length2 ; remember, δ(θ)/ sin θ is unitless because θ is unitless. One
can see the form is correct because integration returns q:
Z π Z 2π Z π Z 2π q
da σ(θ, φ) = sin θdθ dφ a2 δ(θ) (3.128)
0 0 0 0 2 π a2 sin θ
π Z 2π
1
Z
= dθ dφ q δ(θ) = q (3.129)
2π 0 0
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 191
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
We can largely apply what we did in the case of the Example 3.6 except that we
cannot eliminate the {B` } for r < a because the inner boundary is at r = R, not
r = 0. Let’s apply the (homogeneous) boundary condition V (r = R) = 0:
∞
!
X B`in
0= Ain
` R `
+ P` (cos θ) (3.130)
`=0
R `+1
where we use the in superscript to indicate these are the coefficients for the solution in
the region inside of the charge at r = a; i.e., the R < r < a region. Since this a
homogeneous boundary condition, we know from prior discussion we can apply it
term-by-term. Perhaps easier to remember/justify is to apply orthonormality to the
sum, which forces the coefficent of P` at each ` to vanish independently:
∞
B`in R 2`+1
X
Ain `
`R =− =⇒ V (r < a, θ) = Ain
` r ` − `+1 P` (cos θ) (3.131)
R `+1 `=0
r
For r > a, we start with the same form for the solution, but of course now with
different coefficients {Aout out
` } and {B` }. Do not confuse these coefficients with the
{Ain
` } and {B in } determined above: these are solutions in different regions, so they are
`
different functions and there is no reason to expect the coefficients are the same! The
out
{A` } must all vanish so the potential vanishes at infinity. So we have
∞
X B`out
V (r > a, θ) = P` (cos θ) (3.132)
`=0
r `+1
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 192
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Next, we join the solutions at the boundary between them by applying the Neumann
boundary condition there, which requires that V be continuous at r = a and that
∂V /∂r be continuous there except at θ = 0, where it has a discontinuity specified by
σ(0). We apply the first (homogeneous) condition, term-by-term like any
homogeneous boundary condition or via the orthonormality of the P` :
R 2`+1 B`out
Ain
` a` − `+1 = `+1 =⇒ B`out = Ain
` a
2`+1
− R 2`+1 (3.133)
a a
Notice the length−1 units of the portion in parentheses, implying that Ain
` will have
units of −1
o (length)
−(`+1) . Next, we apply the derivative matching (Neumann)
condition:
∂ V out ∂ V in
σ(θ)
− =− (3.136)
∂r ∂r
r =a o
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 193
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Evaluating at r = a gives
`
R2
∞ R
∂ V in
in `+1 ` a a
X
= A` a 2 + (` + 1) P` (cos θ) (3.139)
∂r r =a `=0
a a`+2
`
R2
∞ R
∂ V out
X 1 a a
= Ain
`a
`+1
(` + 1) − 2 + P` (cos θ) (3.140)
∂r r =a `=0
a a`+2
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 194
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
When we difference the two, the second terms in the expressions cancel, leaving
∞
X q δ(θ)
− (2 ` + 1)Ain
`a
`−1
P` (cos θ) = − (3.141)
`=0
2 π a2 o sin θ
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 195
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
`
R2
∞ R
q X a` a a
V (r > a, θ) = − P` (cos θ) (3.146)
4 π o `=0 r `+1 r `+1
The form is hardly one we would have guessed! Separation of variables is more
algorithmic than method of images, but it is also less intuitive. We will connect the
two next.
Recognize that the integral over the boundary condition that we expect from past
experience with separation of variables has already been done on the prior page, so it
is not visible here. Also, that integral did not include an integral over φ as we might
have expected. We could have integrated over φ on both sides if we wanted, yielding a
closer analogy to Equation 3.90, but it would have just yielded a common factor of 2 π
on the two sides since neither side has φ dependence. We did not need to do this
because the problem is azimuthally symmetric and thus we know the solution must
include only the m = 0 term.
Section 3.8.4 Examples of Separation of Variables with Azimuthal Symmetry Page 196
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
We will show
X r<∞ `
1
0
= `+1
P` (cos γ) (3.147)
|~
r −~r | r
`=0 >
This will let us go back and forth between separation-of-variables solutions and
functions that look like the Coulomb potential (e.g., point charge near the grounded
sphere!). Griffiths sort of derives this, using a far less interesting and powerful
technique. He also does it in §3.4.1, after the discussion of separation of variables, so
he is unable to use this theorem to connect the method of images and separation of
variables solutions for the point charge near the grounded, conducting sphere.
Section 3.8.5 Connecting Method of Images to Separation of Variables via a Useful Expansion in Legendre Polynomials Page 197
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
To prove this, orient the coordinate system so ~ r 0 = r 0 zb. The function on the
left-hand side of Equation 3.147 is the potential at ~ r of a point charge q = 4 π o in
magnitude (not units!) at r 0 along the z-axis. It satisfies azimuthal symmetry and
thus is expandable in terms of the above solutions of Laplace’s Equation in spherical
coordinates with azimuthal symmetry (because these solutions form a complete,
orthonormal set!):
∞
1 B`
X
`
= A ` r + P` (cos θ) (3.148)
|~
r −~r 0| `=0
r `+1
(Recall, P` (1) = 1.) Thus, the A` are just the coefficients of the power series
expansion of the left side, which we know (recall: (1 − x)−1 = 1 + x + x 2 + · · ·
for 0 < x < 1) is
∞
1 1 1 1 X r `
= r = (3.150)
r0−r r0 1− r0 r 0 `=0 r 0
The series converges because x = r /r 0 < 1. Thus, A` = 1/(r 0 )`+1 . This now
sets the {A` } for arbitrary ~
r (i.e., arbitrary cos γ rather than the special case
cos γ = 1 we have considered).
Section 3.8.5 Connecting Method of Images to Separation of Variables via a Useful Expansion in Legendre Polynomials Page 198
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
I r >r0
We must eliminate the A` coefficients to keep the function finite as r → ∞.
Again, consider ~
r = r zb, which implies
∞
1 1 X B`
= = (3.151)
r −r0 |~
r −~r 0| `=0
r `+1
Combining the above two cases, and generalizing back from cos θ to cos γ, yields
Equation 3.147.
Section 3.8.5 Connecting Method of Images to Separation of Variables via a Useful Expansion in Legendre Polynomials Page 199
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
Section 3.8.5 Connecting Method of Images to Separation of Variables via a Useful Expansion in Legendre Polynomials Page 200
Section 3.8 Advanced Electrostatics: Separation of Variables in Spherical Coordinates with Azimuthal Symmetry
With that theorem proven, we can make the advertised connection. If we compare
Equations 3.145 and 3.146 from separation of variables for the point charge near the
conducting sphere to Equation 3.147, we see that all four terms in the former are of
the form used in the latter. The first term of the first equation has r< = r and r> = a
as appropriate for r < a, while the first term of the second equation has r< = a and
r> = r as needed for r > a. The second terms of both equations are of the same form
with r< = R 2 /a, r> = r and the charge multiplied by −R/a. Thus, we recover
q 1 R/a
V (~
r) = −
2
(3.153)
4 π o |~
r − a zb| r − Ra zb
~
which matches Equation 3.25. Remarkable! This is a case where we were able to use
separation of variables to recover the full potential and thus the full method of images
solution, which we know then gives us the Green Function: it is possible!
Could we have done a similar thing if we had a point charge in the five-sides-grounded
box problem? There is no reason to think it would not work.
In fact, we will later show how to use a similar technique to find the Green Function in
spherical coordinates for systems without azimuthal symmetry.
Section 3.8.5 Connecting Method of Images to Separation of Variables via a Useful Expansion in Legendre Polynomials Page 201
Lecture 12:
Advanced Electrostatics VIII:
Separation of Variables in Spherical Coordinates
without Azimuthal Symmetry
Spherical Harmonic Expansion for Green Functions
Date Revised: 2022/02/02 05:00
Adjusted lecture break
Added detail on steps between Equation 3.178 and 3.179
Date Given: 2022/01/31
Page 202
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
dm
P`m (x) = (−1)m (1 − x 2 )m/2 P` (x) (3.154)
dx m
(−1)m d `+m
P`m (x) = `
(1 − x 2 )m/2 `+m (x 2 − 1)` (3.155)
2 `! dx
which is now valid for all m. It should be clear that P`0 = P` . It should also be clear
that parity in x (evenness/oddness) of the associated Legendre functions is given by
(−1)`+m (where −1 implies oddness): the parity of P` is given by (−1)` , and each
derivative changes the parity by a factor of −1 (note that the powers of (1 − x 2 ) have
no effect on the parity because it is an even function). There are a number of other
properties of these functions, but it is more useful to consider them together with the
φ solutions.
Section 3.9.1 The Full Polar Equation Solution: the Associated Legendre Polynomials Page 203
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The Full Solution to the Angular Piece of Laplace’s Equation: the Spherical
Harmonics
When one combines the P`m (cos θ) and the e imφ solutions of the polar and azimuthal
equations, one obtains the Spherical Harmonics
s
2 ` + 1 (` − m)! m
Y`m (θ, φ) = P (cos θ) e imφ (3.156)
4 π (` + m)! `
They are an orthonormal, complete basis for functions on the sphere (θ, φ) (assuming
the z-axis is part of the sphere; recall our comment about a second set of solutions to
the Legendre equation if it is not). They satisfy numerous important and useful
conditions:
I Conjugation:
I Orthonormality:
Z 2π Z π
dφ sin θ dθY`∗0 m0 (θ, φ)Y`m (θ, φ) = δ``0 δmm0 (3.158)
0 0
Section 3.9.2 The Full Solution to the Angular Piece of Laplace’s Equation: the Spherical Harmonics Page 204
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
This should be obvious from Equation 3.154, the relation between the Legendre
and the associated Legendre polynomials.
I The θ = 0 behavior is simple given Equation 3.154 (the (1 − x 2 ) factor):
I The Addition Theorem for Spherical Harmonics: Given rb and rb0 pointing in the
directions (θ, φ) and (θ 0 , φ 0 ), respectively, then
`
4π X
r · rb0 ) =
P` (b Y ∗ (θ 0 , φ 0 ) Y`m (θ, φ) (3.166)
2 ` + 1 m=−` `m
where rb · rb0 = cos γ = cos θ cos θ 0 + sin θ sin θ 0 cos(φ − φ 0 ). The proof of this
can be found in Jackson §3.6.
Section 3.9.2 The Full Solution to the Angular Piece of Laplace’s Equation: the Spherical Harmonics Page 206
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
X r<∞ `
1
0
= `+1
P` (cos γ)
|~
r −~r | r
`=0 >
∞ X ` `
1 X 1 r< ∗
= 4 π `+1
Y`m (θ 0 , φ 0 )Y`m (θ, φ) (3.167)
|~
r −~r 0| `=0 m=−`
2 ` + 1 r>
The utility of this relation is even more obvious than that of Equation 3.147,
especially for doing integrals over charge distributions with the relative distance
function (i.e., calculating the potential due to Coulomb’s Law): decompose the
charge distribution in terms of spherical harmonics and integrate the charge
distribution in a particular spherical harmonic Y`m over r 0 with weighting by
(r 0 )` to obtain the component of the potential at a distance r from the origin
with spatial dependence Y`m (θ, φ)/r `+1 .
Section 3.9.2 The Full Solution to the Angular Piece of Laplace’s Equation: the Spherical Harmonics Page 207
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
∞ X `
B`m
X
V (r , θ, φ) = A`m r ` + `+1 Y`m (θ, φ) (3.168)
`=0 m=−`
r
Again, the coefficients {A`m } and {B`m } are set by the volume under consideration
and one or the other entire set may vanish. As well, application of the boundary
conditions will require the orthonormality relations for the spherical harmonics.
As with the case of azimuthal symmetry, we note that, in the process of doing
separation of variables, we have proven that the angular solution satisfies the
eigenvalue-eigenfunction equation
`(` + 1)
∇2 Y`m (θ, φ) = − Y`m (θ, φ) (3.169)
r2
As before, the appearance of r 2 on the right side is not surprising. Note also that m
does not appear in the angular equation. This is because Laplace’s Equation itself is
spherically (and therefore azimuthally) symmetric. The charge distribution and
boundary conditions are what may break the spherical symmetry.
Section 3.9.3 The Full Solution of Laplace’s Equation in Spherical Coordinates Page 208
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The fact that the spherical harmonics combined with the usual power laws in radius
solve Laplace’s Equation for problems that are separable in spherical coordinates can
be used to show that the Green Function for such problems will have a convenient
expansion in terms of the radial solutions and spherical harmonics, like
Equation 3.168. It is convenient to recall at this point that a Green Function is
specified (is unique) once one specifies the geometry and the type of boundary
condition; the value of the boundary condition does not affect the Green Function.
So, once we have specified a geometry and type of boundary condition, the expansion
can be determined and is unique. Alternatively, one can think of this expansion as a
generalization of the corollary of the Addition Theorem, Equation 3.167. It is shown
by using the completeness property of the spherical harmonics and the
eigenvalue-eigenfunction equation for the angular solution. But let’s see that this is
true explicitly for a couple example geometries first:
I Free space
The corollary of the Addition Theorem above is the desired expansion of the
Green Function for charge in free space with no finite radius boundaries and
with the condition V → 0 as r → ∞.
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 209
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Using the same corollary of the Addition Theorem, we can immediately write
(using the fact r 0 (R/r 0 )2 < r 0 always because the the image charge is always
at radius < R while the true charge is at r 0 > R):
`
2
∞ ` ` r 0 rR0 Y ∗ (θ 0 , φ 0 ) Y (θ, φ)
0 1 X X r< R `m `m
r, ~
G (~ r )= `+1 − 0
o `=0 m=−` r> r r `+1 2` + 1
(3.171)
∞ ` `+1 #
"
` ∗ (θ 0 , φ 0 ) Y (θ, φ)
1 r< 1 R2 Y`m
X X `m
= `+1
−
o `=0 m=−`
r> R rr0 2` + 1
(3.172)
In both cases, we finally have forms for the Green Function that could plausibly come
from separation of variables. Note, however, that we did not use separation of
variables to obtain it; we used the method of images combined with the corollary of
the Addition Theorem.
Earlier, we solved for the potential of the latter configuration using separation of
variables with azimuthal symmetry, Equations 3.145 and 3.146 reproduced here but
rewritten using the r< , r> notation:
`
R2
∞ ` R
q X r< a a
V (r , θ) = `+1 − P` (cos θ)
4 π o `=0 r> r `+1
Why was this not enough to give us the full Green Function? Because this solution for
the potential in terms of Legendre polynomials assumed the point charge was along
the z-axis.
What we can do is generalize this solution by replacing cos θ with cos γ = rb · rb0 and a
with r 0 followed by application of the Addition Theorem. Then the solution would be
in a form where we could read off the Green Function expansion in spherical
coordinate not assuming azimuthal symmetry. But that is not the same as obtaining
the solution directly, and clearly the above approach does not generalize to a system
that does not respect azimuthal symmetry.
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 211
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The general approach to the problem of finding the Green Function for an arbitrary
(spherical) geometry is to go back to the definition of the Green Function:
and decompose both sides in terms of spherical harmonics. We do not know the Green
Function yet, so its expansion is the arbitrary general form, which here we write
∞ X
X `
r 0) =
r, ~
G (~ r 0 ) Y`m (θ, φ)
A`m (r |~ (3.175)
`=0 m=−`
where the coefficients in the expansion A`m depend on r , as usual, and they also
depend parametrically on ~ r 0 because it is a parameter in the differential equation.
(We do not know the solutions for the radial dependence of the A`m yet for the
general case we are trying to solve (which is not Laplace’s Equation!), so we cannot
assume they are the power laws we saw for solutions to Laplace’s Equation.)
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 212
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The right side can be rewritten using the breakdown of the delta function into delta
functions in each spherical coordinate followed by completeness of the spherical
harmonics. The breakdown of the delta function is:
δ(r − r 0 )
r 0) =
r −~
δ(~ δ(φ − φ 0 ) δ(cos θ − cos θ 0 ) (3.176)
r2
The 1/r 2 on the radial component is required to cancel the r 2 in the volume element
in spherical coordinates. The fact that the delta function in θ is a function of cos θ
and cos θ 0 is because the volume element contains sin θ dθ = d(cos θ). One could
have instead written δ(θ − θ 0 )/ sin θ as we did when rewriting the point charge near
the grounded, conducting sphere as a surface charge density σ(θ), Equation 3.127.
Using completeness of the spherical harmonics, we have
∞ `
δ(r − r 0 ) X X ∗
δ(~ r 0) =
r −~ 2
Y`m (θ 0 , φ 0 ) Y`m (θ, φ) (3.177)
r `=0 m=−`
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 213
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
∞ X
` ∞ `
X δ(r − r 0 ) X X ∗
−o ∇~2r r 0 ) Y`m (θ, φ) =
A`m (r |~ Y`m (θ 0 , φ 0 ) Y`m (θ, φ)
`=0 m=−`
r2 `=0 m=−`
(3.178)
Note that the Laplacian acts on the unprimed coordinates only. When we evaluate the
action of the Laplacian, a cross term ∇~ ~r A`m (r |~ ~ ~r Y`m (θ, φ) appears, but it
r 0) · ∇
vanishes because the first term points along rb while the second is along θb and φ, b
leaving only ∇~2r acting on each factor in the product individually. We wrote down
earlier Equation 3.169, the eigenvalue-eigenfunction equation satisfied by the angular
solutions of Laplace’s Equation, which we use here to evaluate ∇~2r Y`m (θ, φ):
∞ X `
`(` + 1)
X
− o ∇~2r − A`m (r |~
r 0
) Y`m (θ, φ) (3.179)
`=0 m=−`
r2
∞ `
δ(r − r 0 ) X X ∗
= Y`m (θ 0 , φ 0 ) Y`m (θ, φ)
r2 `=0 m=−`
Note that the Laplacian on the left side is now acting with its radial derivatives only
on A`m ; its action on the spherical harmonics has yielded the `(` + 1)/r 2 term.
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 214
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The coefficients of the individual Y`m (θ, φ) on the two sides must be equal because of
the orthonormality relation for the spherical harmonics, implying
`(` + 1) δ(r − r 0 ) ∗
−o ∇~2r − r 0) =
A`m (r |~ Y`m (θ 0 , φ 0 ) (3.180)
r2 r2
A`m (r |r 0 , θ 0 , φ 0 ) = g` (r , r 0 ) Y`m
∗
(θ 0 , φ 0 ) (3.181)
Plugging in this form to the above reduced version of Laplace’s Equation and
∗ (θ 0 , φ 0 ), we get:
canceling Y`m
`(` + 1) δ(r − r 0 )
−o ∇~2r − g` (r , r 0 ) = (3.182)
r2 r2
Only the Laplacian’s radial derivatives yield a nonzero contribution here, so we have
(also multiplying both sides by −r 2 /o ):
d d δ(r − r 0 )
r2 g` (r , r 0 ) − `(` + 1) g` (r , r 0 ) = − (3.183)
dr dr o
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 215
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
∞ X
`
" #
X B`in (r 0 )
0
r, ~
r < r : G (~ r )= 0
Ain 0 `
` (r ) r + ∗
Y`m (θ, φ) Y`m (θ 0 , φ 0 ) (3.185)
`=0 m=−`
r `+1
∞ X `
B`out (r 0 )
X
r > r 0 : G (~ r 0) =
r, ~ Aout
` (r 0 `
) r + `+1
∗
Y`m (θ, φ) Y`m (θ 0 , φ 0 )
`=0 m=−`
r
(3.186)
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 216
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The first term is the integral of a total differential, so it is trivially integrated. For the
second term, the form of g` (r , r 0 ), where it is sum of two terms, each of which
includes a power law in r and some function of r 0 not dependent on r , ensures it
cannot diverge at r = r 0 . Therefore, the second term is an integral of a function with
no singularity at = 0 (i.e., at r = r 0 ) and thus, as → 0, that integral vanishes. The
right side gives −1/0 when integrated. Therefore, we have
r =r 0 +
d 1
lim r 2 g` (r , r 0 )
=−
→0 dr r =r 0 − o
d out d in 1
g` (r , r 0 ) g` (r , r 0 )
− =− (3.188)
dr 0r =r dr 0
r =r (r 0 )2
0
where g`out (r , r 0 ) is the r > r 0 solution and g`in (r , r 0 ) is the r < r 0 solution. This is a
Neumann-type boundary condition as we had for the examples of the arbitrary charge
density on a sphere σ(R, θ) and for the point charge near the conducting sphere
σ(a, θ) = δ(cos θ)/2 π a2 .
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 217
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
We note, as an aside, that we derived the above matching condition Equation 3.188 in
a somewhat different way here than when we considered the above example. In the
example, we used the fact that we knew the boundary condition on the normal
derivative of the potential from Gauss’s Law. Here, we effectively rederived that
boundary condition for the special case of a radial boundary because we have only to
determine the radial function g (r , r 0 ). We could have gone back a step and written
down the boundary condition on the normal derivative of the potential and derived the
same condition above, but it would have required going back to the full potential and
applying orthonormality and completeness again. We circumvented that step by
rederiving the boundary condition considering only the radial function.
Evaluating the above condition explicitly using the r < r 0 and r > r 0 pieces of the
solution, and multiplying both sides by (r 0 )2 , we obtain
h i h i 1
` Ain 0 out 0 0 `+1
` (r ) − A` (r ) (r ) + (` + 1) B`out (r 0 ) − B`in (r 0 ) (r 0 )−` = (3.189)
o
Since Ain in out out all depend on r 0 , all the powers of r 0 match up.
` , B` , A` , and B`
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 218
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Explicitly evaluating this condition, again using the two portions of the solution, yields
h i h i
Ain 0 out 0 0 2 `+1
` (r ) − A` (r ) (r ) + B`in (r 0 ) − B`out (r 0 ) = 0 (3.191)
The above two matching conditions, along with application of the boundary conditions
that define Dirichlet or Neumann Green Functions (Equations 3.46 and 3.48), provide
four conditions for the four unknowns Ain 0 in 0 out 0 out 0
` (r ), B` (r ), A` (r ), and B` (r ), which
should fully specify them. We finally have a completely algorithmic way to obtain the
full Green Function! What a powerful technique! This general approach can be
applied for any coordinate system in which Laplace’s Equation and the boundary
conditions are separable.
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 219
Lecture 13:
Advanced Electrostatics IX:
Spherical Harmonic Expansion for Green Functions (cont.)
Examples of Using
the Spherical Harmonic Expansion for Green Functions
Date Revised: 2022/02/04 05:00
Changed last page of Example 3.11, all of Example 3.12 to
“skipped” color text
Date Given: 2022/02/04
Page 220
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Example 3.9: Expansion in spherical harmonics for the Green Function for
R < r < ∞ with Dirichlet boundary conditions at r = R and r → ∞
Since Y`m (θ, φ) is in general nonzero, this condition can only hold for all θ, φ if
g` (r ∈ S, r 0 ∈ S, V) = 0 (3.194)
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 221
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
We will apply the above condition at the boundaries first, then the matching conditions
at r = r 0 , because the Dirichlet BC are simpler algebraically (this is the same order of
steps we used when we solved this problem using separation of variables).
First, consider the boundary at r = R. Since r 0 > r = R for all ~ r 0 ∈ V, this implies
that we should require g` (r = R, r 0 ) = 0 for the r < r 0 solution, yielding:
Ain 0 ` in 0
` (r ) R + B` (r ) R
−(`+1)
=0 =⇒ B`in (r 0 ) = −R 2 `+1 Ain 0
` (r ) (3.195)
The condition on the change in the radial derivative at r = r 0 yielded Equation 3.189,
which we plug into to obtain
h i 1
` Ain 0 0 `+1
` (r ) (r ) + (` + 1)Ain 0
` (r ) (r )
0 2`+1
− R 2`+1 + R 2`+1 (r 0 )−` =
o
1 1 1
=⇒ Ain 0
` (r ) = (3.197)
2 ` + 1 o (r 0 )`+1
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 222
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Putting it all together, we have that the Green Function for this Dirichlet boundary
condition, expanded in terms of spherical harmonics, is
`+1 ∗
r` R2 Y`m (θ,φ) Y`m (θ 0 ,φ 0 )
P∞ P`
1 1
− r <r0
o `=0 m=−` (r 0 )`+1 R rr0 2 `+1
G (~ r 0) =
r, ~
(r 0 )`
`+1 ∗
R2 Y`m (θ,φ) Y`m (θ 0 ,φ 0 )
P∞ P`
1 1
− r >r0
o `=0 m=−` r `+1 R rr0 2 `+1
(3.198)
∞ ` 2 `+1 #
"
`
r< ∗ (θ 0 , φ 0 )
Y`m (θ, φ) Y`m
1 X X 1 R
= `+1
− 0
(3.199)
o `=0 m=−` r> R rr 2` + 1
This solution is of course consistent with Equation 3.172, where we used the Addition
Theorem for Spherical Harmonics to rewrite the Green Function for this geometry and
type of boundary conditions in terms of the spherical harmonics, except now that we
used separation of variables from the start rather than relying on the method of
images and the Addition Theorem.
Interesting exercises would be to see that the above expression approaches |~ r 0 |−1
r −~
r →~
as ~ r 0 (use the Addition Theorem to recover the method of images solution) and
also to recover the defining differential equation, Equation 3.174.
Section 3.9.4 Expansion of the Green Function in Spherical Coordinates in Terms of the Spherical Harmonics Page 223
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Examples of Using the Expansion of the Green Function in Terms of the
Spherical Harmonics
We did a lot of gymnastics to get the expansion of the Green Function in terms of
spherical harmonics. Let’s see how it can be used. For each of the examples we will
consider, it would be possible to solve for the potential by separation of variables and
application of boundary conditions without explicitly using our expansion. The
advantage of using the Green Function is that it obviates re-solving the same kind of
problem many times by simply providing integrals that need to be done.
For our examples, we will consider charge distributions inside a conducting sphere. We
quote the general result from Jackson for the Green Function expansion in spherical
harmonics for a geometry consisting of the volume between two spheres at r = a and
r = b with Dirichlet BC on the two surfaces:
GD (~ r 0)
r, ~ (3.200)
∞ X ` 2 `+1 # "
" #
Y ∗ (θ 0 , φ 0 ) Y (θ, φ)
1 X 1 a 1 1 r>
` `m
= r` − − h `m
o `=0 m=−` < `+1 b b2
i
a r< 2 `+1
r> 1 − ba (2 ` + 1)
where, as usual, r< = min{r , r 0 } and r> = max{r , r 0 }. Obtaining this more general
result is a matter of doing the same thing as we did to obtain the result for a spherical
conducting boundary at r = R except that the Aout` term cannot be assumed to vanish.
Next, taking the limit a → 0, we get the result we will need for our work below where
we want to solve for the potential inside a sphere at r = b with Dirichlet BC:
∞ `
" #
∗ (θ 0 , φ 0 ) Y (θ, φ)
Y`m
0 1 X X ` 1 1 r> ` `m
r, ~
GD (~ r )= r − (3.201)
o `=0 m=−` < `+1
r> b b2 2` + 1
You will also be able to read off this simpler result from a method of images problem
you will do in homework. On to our examples!
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 225
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Therefore, the potential in the volume for the Dirichlet B.C. V (b, θ, φ) is
X∞ X ` r ` Z
r) =
V (~ Y`m (θ, φ) dΩ0 Y`m
∗
(θ 0 , φ 0 ) V (b, θ 0 , φ 0 ) (3.203)
`=0 m=−`
b
Again, one can check that the charge density is correct by integrating it: the a−2
cancels the (r 0 )2 factor in the volume element and the argument of the θ 0 delta
function is cos θ 0 because the volume element contains d(cos θ 0 ).
As usual using Equation 3.47 with no surface term because the boundary has V = 0,
the potential is then
Z
V (~
r) = dτ 0 ρ(~
r 0 ) GD (~ r 0)
r, ~
V
∞ X `
Q X
= 2
Y`m (θ, φ) (3.205)
2 π o a `=0 m=−`
" #
∗ (θ 0 , φ 0 )
Y`m
1 1 r> `
Z
0 0 0 `
× dτ δ(r − a) δ(cos θ ) r< `+1
−
V r> b b2 2` + 1
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 227
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
where now r< = min{r , a} and r> = max{r , a} because the δ function does the r 0
integral for us, effectively replacing r 0 with a. (The next example will show the case of
a more complex charge distribution for which the radial integral is not done so easily.)
Now, recall P` (0) = 0 for odd ` and P` (0) = [(−1)n (2n − 1)!!]/2n n! for even ` = 2 n,
so we may reduce the above further to (replacing ` with 2 n so n runs over all
nonnegative integers rather than ` running over all nonnegative even integers):
∞
" #
Q X (−1)n (2n − 1)!! 2n 1 1 r> 2n
V (~
r) = r< 2n+1
− P2n (cos θ) (3.208)
4 π o n=0 2n n! r> b b2
where r< = min(r , a) and r> = max(r , a) again: i.e., not surprisingly, the solution has
a different form depending on whether one wants to know the potential inside the ring
(r < a) or outside the ring (r > a). This is now the complete solution.
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 228
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
To get some intuition for the solution, let’s calculate the induced surface charge
density at r = b. We obtain it from the normal gradient of V , which, recall, is just E~,
and the change in its normal component at a boundary gives the surface charge
density. Since the normal gradient is just d/dr for this particular geometry, it does not
act at all on P2n . In calculating this gradient, r< = a and r> = r since we will in the
end evaluate at r = b > a. Therefore:
dV
r ) = o
σ(~
dr r> =r =b,r< =a
∞
Q X (4n + 1) (−1)n (2n − 1)!! a 2n
=− 2
P2n (cos θ)
4 π b n=0 2n n! b
∞
" #
Q X (4n + 1) (−1)n (2n − 1)!! a 2n
=− 1+ P2n (cos θ) (3.209)
4 π b2 n=1
2n n! b
The expression is written in the above suggestive form on the last line so that it is
easy to obtain the total induced surface charge. Since P0 (cos θ) = 1, the integral of
the n > 0 terms over cos θ can be viewed as integrating P2n with P0 ; by
orthonormality of the Legendre polynomials, these terms all yield zero. The first term
yields −Q when integrated over the sphere. This is what we would expect from
Gauss’s Law applied just inside the r = b boundary.
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 229
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
We have seen in this example how the integration of the charge density with the Green
Function breaks the charge density down into its spherical harmonic components,
calculates the potential due to each component individually (and fairly trivially, just
multiplying by a function of the radius of the source charge and the radius at which
the potential is desired) and then sums up those components. The same kind of
correspondence clearly holds for the induced surface charge density.
Note that the additional 4n + 1 factor implies the θ dependence of the induced surface
charge density is different from that of the original ring charge; i.e., the induced
surface charge is not just a ring. If one thinks about how the field lines from the ring
of charge terminate on the grounded conducting sphere, this is obvious: all the field
lines will not simply go to the equator, they will be distributed in θ.
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 230
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
However, if one considers the form of the potential and considers how to translate it
back to a method of images solution, it should be clear that the second term
corresponds to the potential of a ring of radius b 2 /a and total charge
Qimage = −(b/a) Q as one would expect from generalizing the method of images
solution for a point charge inside a conducting sphere. This can be seen by using
Equation 3.147 (the expansion of |~ r −~r 0 |−1 in Legendre polynomials) with ` = 2n,
r< = r (the point at which we want to know the potential, which always has radius
smaller than that of the image charge because the image charge is outside the sphere
of radius b while the point is inside), r> = b 2 /a (the expected radius of the image
charge), and a normalizing factor −b/a for the size and sign of the image charge
relative to the true charge:
Thus, we see the second term has the right form for an image charge at radius
r> = b 2 /a and normalization −b/a relative to the true charge. The ring shape comes
from the weighted sum over Legendre polynomials, which is the same as the
corresponding sum for the potential of the true charge, the first term.
Seeing that the image charge is a ring at radius b 2 /a explains the induced surface
charge density distribution via its proportionality to the field lines from the true charge
to the image charge at the r = b surface. Drawing a picture using the image charge
configuration should make this clear.
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 231
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The first twist is figuring out how to write down the charge density in spherical
coordinates. One could probably rigorously derive the form by writing down the charge
density trivially in Cartesian or cylindrical coordinates and then applying Jacobian
transformation to convert it to spherical coordinates, but there is an easier, more
intuitive way.
It is all present at cos θ = 1 and cos θ = −1, so clearly delta functions for these
positions need to be included. It has azimuthal symmetry, so there will be no φ
dependence, only a factor of 1/2 π. The charge is distributed in radius, so there is
some to-be-determined radial dependence f (r ). To figure out f (r ), let’s write down
the requirement that the integral be the total charge Q:
Q
r) =
ρ(~ f (r ) [δ(cos θ − 1) + δ(cos θ + 1)] (3.212)
2
Zπ
Q= dτ ρ(~ r)
V
Q
Z b Z 1 Z 2π
= dr r 2 f (r ) d(cos θ) [δ(cos θ − 1) + δ(cos θ + 1)] dφ
2π 0 −1 0
Z b
= 2Q dr r 2 f (r )
0
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 232
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
1
Q = 2Q c b =⇒ c= (3.213)
2b
Q
=⇒ r) =
ρ(~ [δ(cos θ − 1) + δ(cos θ + 1)] (3.214)
4 π b r2
Now, since the sphere is grounded, we just need to do the integral of the charge
density with the Dirichlet Green Function:
1
Z
V (~
r) = dτ 0 ρ(~
r 0 ) GD (~ r 0)
r, ~ (3.215)
o V
∞ X `
Q δ(cos θ 0 − 1) + δ(cos θ 0 + 1)
X Z
= dτ 0
4 π o b `=0 m=−` V (r 0 )2
" #
∗ (θ 0 , φ 0 ) Y (θ, φ)
` 1 1 r> ` Y`m `m
× r< `+1
− 2
r> b b 2` + 1
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 233
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
∞ Z b " #
Q X 1 1 r> `
V (~
r) = P` (cos θ) dr 0 r<
`
`+1
− [P` (1) + P` (−1)]
8π 0 b `=0 0 r> b b2
We know P` (1) = 1 and P` (−1) = (−1)` , so the term containing these two factors
yields 2 for even ` and 0 for odd `. Thus, the above reduces to
Z b " #
Q X 1 1 r> `
V (~
r) = P` (cos θ) dr 0 r<
`
`+1
−
4 π 0 b ` even 0 r> b b2
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 234
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
The integral over radius must be broken into two pieces, one for r 0 < r and one for
r 0 > r , because the r< and r> variables take on different values for these two regions
(by definition!). Doing so, and doing the integrals (they are straightforward) yields
" #
Z b 1 1 r> ` 2` + 1 1
r `
dr 0 r<
`
`+1
− = 1 − (3.216)
0 r> b b2 `+1 ` b
The second portion of the above quantity is well-defined for ` 6= 0, but not for ` = 0.
We need to use L’Hôpital’s rule to evaluate it for ` = 0:
h i
d r `
1− r ` r d
1 r ` ln ` b
d` b b b d`
lim 1− = lim d
= − lim d
= ln (3.217)
`→0 ` b `→0
d`
` `→0
d`
` r
Therefore, we may write the full solution as, separating out the ` = 0 term and
rewriting in terms of ` = 2 n,
∞
" #
Q b X 4n + 1 r 2n
V (~
r) = ln + 1− P2n (cos θ) (3.218)
4 π o b r n=1
2n (2n + 1) b
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 235
Section 3.9 Advanced Electrostatics: Separation of Variables in Spherical Coordinates without Azimuthal Symmetry
Let’s calculate the induced surface charge density and the total induced charge again:
∞
" #
∂ V Q X 4n + 1
σ(θ) = o =− 1+ P2n (cos θ) (3.219)
∂r r =b 4 π b2 n=1
2n + 1
Note again how the surface charge density has a different n-dependent weighting than
the potential. Finally, integrating over the sphere to get the total induced charge, all
n ≥ 1 terms vanish, yielding
Z
Qind = b 2 dφ d cos θ σ(θ) = −Q (3.220)
r =b
as we expect from Gauss’s Law. It would again be interesting to rewrite the solution
in the form of a method of images solution, which you have the tools to do.
Section 3.9.5 Examples of Using the Expansion of the Green Function in Terms of the Spherical Harmonics Page 236
Lecture 14:
Advanced Electrostatics X:
Multipoles
Date Revised: 2022/02/04 05:00
Date Given: 2022/02/04
Page 237
Section 3.10 Advanced Electrostatics: Multipole Expansions
Multipole Expansions
Dipoles: Quick Review
Recall from Ph1b the idea of an electric dipole: two charges of equal and opposite
magnitude ±q spaced very close together at ~ r+ and ~ r− . The net charge cancels
almost perfectly, so, rather than the potential falling off like 1/r at large radius, it falls
off as 1/r 2 with functional form
1 p ~ · rb r r r
V (~
r) = as , , →∞ (3.221)
4 π o r 2 |~
r+ | |~
r− | |~
r+ − ~
r− |
where p r+ − ~
~ = q(~ r− ) is the dipole moment.
This idea generalizes. When one has a charge distribution with vanishing net charge,
but inside of which there is a variation in the charge density, that variation is still
noticeable at large distance as a set of potentials that fall off more quickly than 1/r .
The first additional term is the dipole, falling as 1/r 2 , the second is the quadrupole,
falling as 1/r 3 , the third is the octupole, falling as 1/r 4 , and so on. The nomenclature
comes from the minimum number of different source charges one must have to obtain
that moment: one for monopole, two for dipole, four for quadrupole, etc.
1 r 0)
ρ(~
Z
V (~
r) = dτ 0 (3.222)
4 π o V |~
r −~ r 0|
We now use Equation 3.147, taking r< = r 0 and r> = r because r is outside the
charge distribution. Thus,
∞
1 (r 0 )`
Z X
V (~
r) = dτ 0 ρ(~
r 0) P` (cos γ) (3.223)
4 π o V `=0
r `+1
where cos γ = rb · rb0 is the angle between the two vectors. There is a common 1/r we
can factor out, leaving
∞
1 1X 1
Z
`
V (~
r) = dτ 0 ρ(~
r 0 ) r 0 P` (cos γ) (3.224)
4 π o r `=0 r ` V
This is the multipole expansion of the potential of the charge distribution. One can
see that the successive terms fall off as successively higher powers of 1/r .
Let’s write out the first three terms more explicitly to get some physical intuition:
I Monopole term
The first term is
1 1 1 Q
Z
V1 (~
r) = dτ 0 ρ(~
r 0) = (3.225)
4 π o r V 4 π o r
This is the standard Coulomb’s Law term due to the total charge. Far enough
away, all charge distributions look pointlike. But, if Q = 0, this term vanishes
identically and the higher-order terms must be considered. Even if Q 6= 0, if one
is close enough to the charge distribution to see its non-pointlike nature, the
higher-order terms will be important corrections to the monopole term.
Section 3.10.3 The Monopole, Dipole, and Quadrupole Terms Page 240
Section 3.10 Advanced Electrostatics: Multipole Expansions
I Dipole term
The second term is
1 1 1 1
Z Z
V2 (~
r) = 2
dτ 0 ρ(~
r 0 ) r 0 cos γ = 2
r 0 ) r 0 rb0 · rb
dτ 0 ρ(~
4 π o r V 4 π o r V
1 1
Z
= r · dτ 0 ρ(~ r 0) ~
r0 (3.226)
r2
b
4 π o V
1 1
Z
or V2 (~
r) = rb · p
~ where ~=
p dτ 0 ρ(~
r 0) ~
r0 (3.227)
4 π o r 2 V
Z
pj = dτ 0 ρ(~
r 0 ) rj 0 = rbj · p
~ (3.228)
V
Section 3.10.3 The Monopole, Dipole, and Quadrupole Terms Page 241
Section 3.10 Advanced Electrostatics: Multipole Expansions
I Quadrupole term
The third term is
1 1 1
Z
r 0 ) (r 0 )2
dτ 0 ρ(~ 3 cos2 γ − 1
V3 (~
r) =
4 π o r3 V 2
1 1 1
Z
r 0 ) (r 0 )2
dτ 0 ρ(~ 3 rb · rb0 rb0 · rb − 1
= 3
4 π o r V 2
1 1 0 21
Z
0 0 0 0
= r · dτ ρ(~r ) (r ) 3 r r − 1 · rb (3.229)
r3
b b b
4 π o V 2
1 1 1
Z h i
or V3 (~
r) = rb · Q · rb where Q= dτ 0 ρ(~ r 0 − (r 0 )2 1
r 0~
r 0) 3 ~
4 π o r 3 2 V
(3.230)
is the quadrupole moment and 1 = diag(1, 1, 1) is the identity tensor with ones
along the diagonal. Because it is composed of ~ r0~r 0 and 1, Q is a tensor,
implying that one can take a dot product with a vector on each side. Written
out in component form (which is again how you would calculate it):
Z h i
Qjk = r 0 ) 3 rj 0 rk0 − (r 0 )2 δjk = rbj · Q · rbk
dτ 0 ρ(~ (3.231)
V
Section 3.10.3 The Monopole, Dipole, and Quadrupole Terms Page 242
Section 3.10 Advanced Electrostatics: Multipole Expansions
where we define the charge distribution in the new coordinate system ρ 0 (~ r 0 ) in terms
of the original charge distribution ρ(~ r ) to be such that ρ 0 (~
r 0 ) = ρ(~ r0−~
r =~ a) when
r0 =~
~ r +~ a. Thus, an origin shift can induce an artificial dipole moment for a charge
distribution that has a monopole moment. This part of the dipole moment is not real:
it is a reflection of the fact that the multipole potentials are written in terms of
distance from the origin under the assumption that the charge distribution is centered
around the origin. When it is not, this is an unnatural coordinate system to use,
requiring corrections to the standard monopole term (∝ Q/r ) to handle the fact that
the charge distribution is displaced. The above tells us the correction term has the
same form as a dipole term but is dependent on the monopole of the charge
distribution. Obviously, one must choose the origin wisely to avoid such complications.
Note also the somewhat counterintuitive implication that, if Q = 0, then the dipole
moment is independent of origin! This happens because of our assumption
r , r 0 distance to the observation point, which implies that a must also be small so
that no corrections are required.
p cos θ
V2 (~
r) = (3.233)
4 π o r 2
∂ V2 2 p cos θ
=⇒ r) = −
Er (~ =− (3.234)
∂r 4 π o r 3
1 ∂ V2 p sin θ
r) = −
Eθ (~ = (3.235)
r ∂θ 4 π o r 3
1 ∂ V2
r) = −
Eφ (~ =0 (3.236)
r sin θ ∂φ
~ (~ p
or E r) = 3
2 rb cos θ + θb sin θ (3.237)
4 π o r
1 1 1 1 X
V2 (~
r) = r ·p
~ ~= ri pi (3.238)
4 π o r 3 4 π o r 3 i
3/2
Where we used r 3 = r 2 and r 2 = k rk2 to more easily calculate the partial
P
derivative. Therefore, with rj and rbj being the jth Cartesian coordinate and unit vector,
~2 (~ ~ 2 (~
X ∂ V2 1 X
rbj pi 3 ri rj − δij r 2
E r ) = −∇V r) = − rbj =
j
∂rj 4 π o r 5 ij
" ! #
1 X X
= rj rbj 3 pi ri − pj rbj r 2
4 π o r 5 j i
~2 (~ 1
=⇒ E r) = p · rb) rb − p
[3 (~ ~] (3.240)
4 π o r 3
Under the assumption that V (~ r 0 ) varies slowly (but need not be constant!) over the
spatial extent of the charge distribution, we can rewrite this in terms of moments of
the charge distribution and derivatives of the potential. To do so, we need to expand
V (~
r ) about some point in the distribution. Without loss of generality, assume the
charge distribution is centered on the origin, around which we will expand. We use the
multidimensional Taylor expansion of V (~ r ):
3
1 X 0 0 ∂ 2 V
0 0
X
0 ∂V
V (~ r = ~0) +
r ) = V (~ rj + rj rk + ··· (3.242)
∂rj 2 j,k=1 ∂rj ∂rk ~r 0 =~0
r 0 =~
j=1 ~ 0
We can already foresee how integrating the above form for V (~r 0 ) with ρ(~
r 0 ) is going
to result in a dipole moment in the first term and quadrupole moment in the second.
Section 3.10.6 Electrostatic Potential Energy of a Multipole Distribution in an External Potential Page 246
Section 3.10 Advanced Electrostatics: Multipole Expansions
3
∂ 2 V
~ (~0) + 1 X
r 0 ) = V (~0) − ~
V (~ r0·E 3 rj 0 rk0 + ··· (3.243)
6 j,k=1 ∂rj ∂rk ~0
3 ∂ 2 V
~ (~0) + 1 X 0 0
r0·E
= V (~0) − ~ 3 rj rk − δjk (r 0 )2 + ··· (3.244)
6 j,k=1 ∂rj ∂rk ~0
∂ 2 V
X
(r 0 )2 δjk = (r 0 )2 ∇2 V (~
r 0 = 0) = 0 (3.245)
j,k
∂rj ∂rk ~0
because the charge distribution sourcing V is not present near the origin. Remember,
r ) is not the distribution sourcing V ; V is provided to us and is due to some charge
ρ(~
distribution far away from the origin.
Section 3.10.6 Electrostatic Potential Energy of a Multipole Distribution in an External Potential Page 247
Section 3.10 Advanced Electrostatics: Multipole Expansions
integral) Z Z
U = V (~0) dτ 0 ρ(~ ~ (~0) ·
r 0) − E dτ 0 ρ(~
r 0) ~
r0 (3.246)
V V
3
1 X ∂ 2 V
Z h i
+ r 0 ) 3 rj 0 rk0 − δjk (r 0 )2 + · · ·
dτ 0 ρ(~
6 j,k=1 ∂rj ∂rk ~0 V
3
∂ 2 V
~ (~0) + 1 X
= Q V (~0) − p
~·E Qjk + ··· (3.247)
6 j,k=1 ∂rj ∂rk ~0
~ (~ 1 ~ ~ r V (~
r) − p
= Q V (~ ~·E r) + ∇ r · Q · ∇~
~ r) + · · · (3.249)
6
where we have written the last term in tensor dot product form. There are now
contributions to the potential energy from the relative alignment of p ~ and E ~ and from
the alignment of Q’s principal axes relative to the principal axes of the potential’s
curvature matrix. Note that ∇~ ~r acts on the spatial dependence of V (~ r 0 has
r ); ~
already been integrated over to obtain Q.
Section 3.10.6 Electrostatic Potential Energy of a Multipole Distribution in an External Potential Page 248
Section 3.10 Advanced Electrostatics: Multipole Expansions
3
1 ∂ 2 Ej
F~ (~ ~ r ) = Q −∇V
~ (~ ~ p ~ (~
X
r ) = −∇U(~ r) + ∇ ~·E r) + rbm Qjk + ···
6 j,k,m=1
∂rm ∂rk
3
~ (~
~ E ~ (~ 1 X ∂ 2 Ej
=QE r) + p~·∇ r) + rbm Qjk + ···
6 j,k,m=1
∂rm ∂rk
~ (~
~ E ~ (~ 1 ~ h~ ~ (~
i
=QE r) + p~·∇ r) + ∇ ∇· Q ·E r) + · · · (3.250)
6
We see that the total force is a sum of contributions from the interaction of the
monopole with the electric field, the dipole with gradients in the electric field and, the
quadrupole with the local curvature (second derivatives) of the electric field.
The monopole term yields no torque because there is no orientation angle involved: Q
and V (~
r ) are scalars.
Considering the dipole term, we understand that there are only two vectors involved, p ~
and E ~ , and the potential energy only depends on the angle between them. So the
torque will be given by the derivative with respect to this angle, which we call θp to
~ . This angle
differentiate it from the θ coordinate of the system in which we consider E
will be measured from E ~ to p
~. Then,
~ elec = − ∂
N −~p·E ~ (~
r) (3.251)
∂θp
∂
~
~
= p E r ) cos θp = −p E
(~ (~
r ) sin θp
∂θp
=p ~ (~
~×E r) (3.252)
This is a result you are familiar with from Ph1b, indicating the torque acts in a
direction to align the dipole moment with the field direction.
Moving on to the quadrupole term, we recognize from Ph106a that any symmetric
tensor can be diagonalized via a rotation. Let’s write
where the Qi are quadrupole moments along the principal axes of the quadrupole
tensor and R(φQ , θQ , ψQ ) is the rotation matrix that rotates from the frame in which
the coordinate axes align with the quadrupole tensor’s principal axes to the arbitrary
frame we started in, with the three Euler angles (φQ , θQ , ψQ ) defining the orientation
of the principal axes of Q relative to the this arbitrary frame. This kind of
diagonalization should be familiar to you from Ph106a, with R rotating from the
“body” frame (the one fixed to the charge distribution’s quadrupole principal axes) to
the “space” frame. The quadrupole potential energy term is then
1 ~ n o
~ (~
U3 = − ∇~r · R(φQ , θQ , ψQ ) Q [R(φQ , θQ , ψQ )]T · E r) (3.254)
6
To calculate the torque, we need to take the gradient of U3 with respect to the
orientation of the quadrupole. This amounts to taking gradients of R and RT with
respect to this orientation. As you know from the case of the symmetric top, the Euler
angles are particularly useful angles with respect to which these derivatives can be
taken. ∂/∂φQ gives the torque about the z-axis of the space frame. ∂/∂θQ gives the
torque that causes motion in the polar angle direction with respect to the space
frame’s zb. And ∂/∂ψQ calculates the torque about one particular principal axis of the
quadrupole, chosen at will. You are familiar with symmetric tops, with I1 = I2 . Here,
we can have symmetric quadrupoles, with Q1 = Q2 . In this case, the ψQ angle is the
angle about the 3 axis of the quadrupole (the principal axis that aligns with the z-axis
in the body frame). We do not take this further because, as you know from the study
of tops in Ph106a, the phenomenology can be quite rich.
Page 253
Lecture 15:
Electrostatics in Matter I:
Polarizability and Polarization
Bound Charges and their Potential
Electric Displacement Field
Linear Dielectrics
Date Revised: 2022/02/07 06:30
Date Given: 2022/02/07
Page 254
Section 4.1 Electrostatics in Matter: Polarizability and Polarization
I Atoms and molecules are polarizable, meaning that they can acquire a dipole
moment when an external electric field is applied because of the separation of
the positive and negative charge in response to the applied field. The charge
distribution that results is such that its field is in the opposite direction as the
applied field at the location of the atom or molecule.
I We assume that this polarizability is a linear process, so that the induced dipole
moment is linear in the applied electric field, though the response may be
anisotropic. The polarizability tensor α relates the induced dipole moment to
the applied field:
p ~
~ =α·E (4.1)
~ =p
N ~
~×E (4.2)
If the electric field is nonuniform, the dipole feels a force (Equation 3.250)
F~ = p ~ E
~·∇ ~ (4.3)
~ = np
P ~ (4.4)
~ is the induced
where n is the density of polarizable atoms or molecules and p
dipole per atom or molecule.
When a medium is polarized and acquires a polarization vector P, ~ then it can generate
its own electric field. This comes from the superposition of the dipole fields of the
individual polarized atoms or molecules. In Ph1b, you saw how the polarization could
be interpreted as yielding bound charge densities: when the medium polarizes, the
positive components of some dipoles are cancelled by the negative components of
nearby dipoles, but there can appear a net effective charge: on the boundaries, where
the cancellation fails, and in the bulk if the dipole density is not uniform, also causing
the cancellation to fail. This argument was made in Purcell in Ph1b to derive the
bound charge densities, and Griffiths makes it in §4.2.2. Here we derive the
relationship between the polarization vector and the bound charge density in rigorous
fashion.
Section 4.1.2 Bound Charges and the Potential of a Polarizable Material Page 257
Section 4.1 Electrostatics in Matter: Polarizability and Polarization
1
Z ~ r 0 ) · (~
P(~ r −~ r 0)
V (~
r) = dτ 0 (4.5)
4 π o V |~
r −~ r 0 |3
1 1
Z
V (~
r) = ~ r 0) · ∇
dτ 0 P(~ ~ ~r 0 (4.6)
4 π o V |~
r −~r 0|
Section 4.1.2 Bound Charges and the Potential of a Polarizable Material Page 258
Section 4.1 Electrostatics in Matter: Polarizability and Polarization
The first term can be converted to a surface integral via the divergence theorem:
"Z #
1 nb(~ ~ r 0) Z
r 0 ) · P(~ 1
V (~
r) = da0 − dτ 0
∇ ~ r 0)
~ ~r 0 · P(~ (4.8)
4 π o S(V) |~
r −~ r 0| V |~
r −~r 0|
r 0)
We thus see that the potential appears to be that of a surface charge density σb (~
on S(V) and a volume charge density ρb (~r 0 ) in V with (b
n is the outward normal from
the polarizable material):
r 0 ) = nb(~
σb (~ ~ r 0)
r 0 ) · P(~ r 0 ) = −∇
ρb (~ ~ r 0)
~ ~r 0 · P(~ (4.9)
"Z #
1 r 0)
0 σb (~ 0 ρb (~r 0)
Z
V (~
r) = da + dτ (4.10)
4 π o S(V) |~
r −~r 0| V |~
r −~ r 0|
These charges are called “bound charges” because they are bound to the polarizable
medium.
Section 4.1.2 Bound Charges and the Potential of a Polarizable Material Page 259
Section 4.1 Electrostatics in Matter: Polarizability and Polarization
σb = nb(~ ~ = rb · P zb = P cos θ
r) · P (4.11)
This is a problem Griffiths solves in Example 3.9 for a generic σ(θ), and we talked
through the solution earlier. The generic solution was
∞ ∞
X X B`
V (r < R, θ) = A` r ` P` (cos θ) V (r > R, θ) = P (cos θ)
`+1 `
(4.12)
`=0 `=0
r
1
Z π
with A` = dθ 0 sin θ 0 σ(θ 0 ) P` (cos θ 0 ) B` = A` R 2 `+1 (4.13)
2 o R `−1 0
Since σ(θ) = P cos θ = P P1 (cos θ), the orthonormal functions do their job and we get
(making sure to include the normalization factor 2/(2 ` + 1) = 2/3):
P r cos θ P R 3 cos θ
V (r < R, θ) = V (r > R, θ) = (4.14)
3 o 3 o r 2
Section 4.1.2 Bound Charges and the Potential of a Polarizable Material Page 260
Section 4.1 Electrostatics in Matter: Polarizability and Polarization
We can write these more simply. We recognize z = r cos θ and that the total dipole
~ = 4 π R 3 P zb/3, yielding
moment of the sphere is p
Pz ~ · rb
p
V (r < R, θ) = V (r > R, θ) = (4.15)
3 o 4 π o r 2
~ = −P/3
Thus, the field inside the sphere is uniform, E ~ o , and the field outside the
~. Note that the field outside the sphere is a perfect dipole
sphere is that of a dipole p
field all the way to r = R; this is not an approximation (until you get so close to the
surface that you can see the discretization of the dipoles).
We remind the reader of the Ph1b technique, where we obtained this same result by
treating the sphere as two spheres of uniform charge density ρ = q/(4 π R 3 /3) with
their centers displaced by d~ = p
~/q. The field inside a uniform sphere of charge is
proportional to the radial vector outward from its center, so the two vectors ~ ~
r − d/2
and ~ ~ end up differencing (because the two spheres have opposite charge) to
r + d/2
~ yielding the uniform internal field. Outside the spheres, they look like point
yield d,
~.
charges, so the system looks like a point dipole p
One could also use this argument to figure out that the charge density on the surface
is σ = P cos θ and evaluate the potential and field of that charge distribution.
Section 4.1.2 Bound Charges and the Potential of a Polarizable Material Page 261
Section 4.2 Electrostatics in Matter: The Electric Displacement Field
These are analogues of Coulomb’s law for ρb , so the potential and field due to the
polarization density satisfy
1 ~b = 1 ρb = − 1 ∇
~ ·E ~
~ ·P
∇2 Vb = − ρb ∇ (4.17)
o o o
1 ~ = 1 ρf − ∇
~ ·E
~
~ ·P
∇2 V = − (ρf + ρb ) ∇ (4.18)
o o
We will see later that it will be convenient to have a field that depends only on the
free charge density. Thus, we define the electric displacement field by
~ = o E
D ~ +P
~ (4.19)
The Helmholtz Theorem tells us that any vector field can be written as the sum of a
curl-free component (sourced by the divergence of the field) and a divergence-free
~ we also
component (sourced by the curl of the field). Thus, to fully understand D,
need to determine its curl:
∇ ~ = o ∇
~ ×D ~ +∇
~ ×E ~ =∇
~ ×P ~
~ ×P (4.21)
Because the right side may not vanish, the left side may not vanish. This possibly
nonzero curl is an important distinction between D~ and E
~.
~ ×P
Note the extra condition ∇ ~ = 0 that has to be specified; this reflects the fact that
~ has more degrees of freedom than a scalar field ρb , so those extra degrees of
P
freedom need to be specified (via the curl-free condition) for ρb to tell the whole story
~ to be derivable from ρf ).
(and thus for D
The situation will simplify somewhat when we consider linear, uniform dielectrics
where P~ ∝E ~ ; then ∇
~ ×P ~ = 0 is guaranteed, though the requirement that P
~ be
normal to any boundaries may still create complications.
~1 = 1 σ
nb · E~2 − E sb · E~2 − E
~1 = 0 (4.22)
o
where nb is the normal vector pointing from region 1 into region 2 and sb is any
tangential vector (i.e., sb · nb = 0). We derived the equation for the normal component
using the divergence of E ~ . So, here, we can use the fact that ∇
~ ·D~ = ρf , which yields
nb · D~2 − D
~ 1 = σf (4.23)
nb · P~2 − P
~ 1 = −σb (4.24)
We could also have used ρb = −∇~ ·P ~ and followed the same type of derivation as
used for E~ and D.
~ The sign on the right side of the boundary condition enters because
~ = −ρb .
~ ·P
of the sign in ∇
sb · D~2 − D
~ 1 = sb · P~2 − P
~1 (4.25)
Note that, even in the case of linear dielectrics, the right side can be nonzero, as we
will see below.
Linear Dielectrics
Susceptibility, Permittivity, and Dielectric Constant
So far, we have considered situations where P ~ has been specified for us. But, it is
usually caused by an external field, and so what we really want to do is figure out
what observed potential and field arise by summing the externally applied
potential/field and that due to the polarization of the dielectric in response to that
external potential/field. For most substances, at least at low fields, the relation
between the two is linear and there is a simple scalar constant of proportionality:
~ = o χe E
P ~ (4.26)
where χe is the electric susceptibility. Such materials are called linear dielectrics. An
immediate implication of the above is:
D ~ +P
~ = o E ~ = o (1 + χe ) E
~ ≡ E
~ (4.27)
~ = E (r ) rb
E ~ = D(r ) rb
D ~ = P(r ) rb
P (4.28)
This ensures that the curl of all three vanish and that, at the boundaries, we have no
tangential components of D ~ and P.
~ We have now satisfied all the conditions required
~ directly from the free charge by Gauss’s Law, which yields
for us to be able to derive D
~ r) = Q
D(~ rb r >a (4.29)
4 π r2
~ =E
(D ~ =P
~ = 0 for r < a.) Then we just apply the relation between D
~ and E
~:
Q Q/4 π r 2 rb
~ (~ a<r <b
E r) = rb = (4.30)
4 π (r ) r 2 Q/4 π o r 2 rb b<r
The electric field is screened (reduced) inside the dielectric and unchanged outside.
~ r ) = o χe (r ) E
~ (~ ~ (~ (r ) − o Q
P(~ r ) = ((r ) − o ) E r) = rb
(r ) 4 π r2
−o Q
4 π r2
rb a < r < b
= (4.31)
0 b<r
ρb = −∇~ ·P ~ =0 (4.32)
(
−b ~ = a) = − −o Q 2
r · P(r r =a
σb = 4πa (4.33)
~
rb · P(r = b) = − o Q
r =b
4 π b2
Note the in the denominator! We see that P ~ is radially outward and decreasing with
~ does. Note that, even though P
r like 1/r 2 as E ~ is position-dependent, its divergence
vanishes, so there is no bound charge density. There is surface charge density, negative
at r = a and positive at r = b. This is to be expected, as the dielectric polarizes so
the negative ends of the dipoles are attracted to Q on the conducting sphere and the
positive ends are repelled, leaving uncancelled layers of negative charge on the inner
boundary and positive charge on the outer boundary.
The electric field is reduced inside the dielectric because the negative charge on the
inner boundary screens (generates a field that partially cancels) the field of the free
charge on the conducting sphere: the total surface charge density σf + σb at r = a is
less than Q/4 π a2 , and it is the total charge that determines E~.
Note that, because of the neutrality of the dielectric, the total surface charge on the
outer boundary cancels that on the inner boundary, so the net charge enclosed inside a
sphere of radius r > b is just Q: outside the dielectric, no screening effect is present.
It is worth thinking about the above a bit: it occurs both because the dielectric has no
net charge and the problem is spherically symmetric. In contrast, we will see a
dielectric sphere can polarize in an external field and generate a field outside itself in
spite of having no net charge, which is possible because spherical symmetry is broken
in that case. But there is no monopole field, only a dipole field.
Note also that, once you have calculated σb and ρb , you can ignore the presence of
the dielectric: as we stated earlier, the total field is sourced by the sum of the free and
bound charge densities and the dielectric has no further effect, which one can see here
from noticing that E~ in the dielectric is what one would have calculated if one had
been given σf + σb at r = a.
~ (not D!):
Finally, let’s calculate the electric potential from E ~
Z ~
r Z r
r) = −
V (~ d~s 0 · E~ (~
r 0) = − dr 0 E (r 0 )
∞ ∞
Z r
Q 1 Q 1
V (r > b) = − dr 0 2
= (4.34)
4π ∞ o r 4 π o r
Z b Z r
Q 1 0 1
V (a < r < b) = − dr 0 + dr
4π ∞ o r 2 b r2
" b r #
Q 1 1 Q 1 1 1 1
= + = − + (4.35)
4 π o r ∞ r b 4 π b o r
Q 1 1 1 1
V (r < a) = V (r = a) = − + (4.36)
4 π b o a
A final comment: if one takes the → ∞ limit, one can see that one recovers the
behavior one would have if the entire region r < b were filled with conductor. A
~ =0
conductor can be considered to be an infinitely polarizable dielectric, with E
inside, which requires χe → ∞.
Page 272
Section 4.3 Electrostatics in Matter: Linear Dielectrics
Let the capacitor plates lie parallel to the xy -plane at z = 0 (negative plate) and
z = a (positive plate) so zb is the unit vector pointing from the negative plate to the
positive one. In such a geometry, we know from symmetry that E ~ , D,
~ and P~ are all
parallel to zb and independent of xy , assuming we ignore the capacitor edges. Thus, at
the interfaces at z = 0 and z = a, all these vectors are normal to the interface and so
no tangential components are present. These features of the fields imply that we can
apply Gauss’s Law to the free charge density to find D. ~
The free charge density is σf = ±Q/A where Q is the charge on the plates (+Q at
z = a and −Q at z = 0) and A is the plate area. Gauss’s Law for an infinite sheet of
charge (Griffiths Example 2.5) tells us that the field of a single sheet is E = σ/2 o .
Therefore, we have for this case
−Q
~ = zb 0<z<a
D A (4.37)
0 z < 0, z > a
because the fields of the two plates cancel for z < 0 and z > a but add for 0 < z < a,
and there is no o because we are calculating D,~ not E~.
This implies:
− 1 Q
− − Q
o
~= zb 0<z<a ~ = zb 0<z<a
E A P A (4.38)
0 z < 0, z > a 0 z < 0, z > a
ρb = −∇ ~ ·P~ =0 (4.39)
(
~ = a) − − o Q
~ = zb · P(z z=a A
z=a
σb = nb · P ~ = 0) = −o Q (4.40)
−b z · P(z z=0 A
z=0
We have negative bound surface charge near the positive plate and positive bound
surface charge near the negative plate. Finally, the voltage is
Z z Z z
1 Q
1 Q
V (0 < z < a) = − ~ (~
d~s 0 · E r 0) = − dz 0 − = z (4.41)
0 0 A A
From this, we can calculate the capacitance, which comes out as expected:
Q Q A
C = = = = r Cvac (4.42)
∆V (1/) (Q/A) a a
C is increased because ∆V is reduced because the surface charge densities screen the
electric field inside the dielectric. The electric field inside the dielectric is the field one
expects from surface charge densities σf + σb = ±(o /) (Q/A).
The volume bound charge density vanishes again. The surface charge density at the
top and bottom has the same expression, but again with being evaluated for the
particular value of z. The surface bound charge density at the z = a interface is
~ 2 = Q − 1 − o + 2 − o
~ 1 + nb2 · P
σb (z = a) = nb1 · P ~ 2 = zb · P
~ 1 − zb · P (4.43)
A 1 2
Depending on which dielectric constant is greater, this can be positive or negative. Of
course, it vanishes if 1 = 2 . The potential and capacitance are
1 Q 1 Q 1 Q
V (0 < z < a) = z V (a < z < b) = a+ (z − a) (4.44)
1 A 1 A 2 A
−1
Q a b−a A
C = = + A = eff = eff ,r Cvac (4.45)
∆V 1 2 b
where 1/eff = [a/1 + (b − a)/2 ]/b is the thickness-weighted inverse mean of the
dielectric constants and eff ,r = eff /o . This is the same as two capacitors in series,
which is not surprising since that problem has the same equipotential surfaces. The
total field is that of three sheets of surface charge σf + σb , with σf = 0 at the
interface between the two dielectrics.
1
C = C1 + C2 = (1 A1 + 2 A2 ) (4.46)
a
But let’s derive this from scratch, appreciating the subtlety at the interface.
Because the voltage difference between the two plates is independent of (they are
~ is the same in 1 and 2 : this is the
equipotentials), it is reasonable to guess that E
key insight! Because the dielectrics are uniform in z, it is also reasonable to assume it
is independent of z as one would have in the single-dielectric case. So, our guess for
the form of the fields is:
~ = −E0 zb D
~ = −1 E0 zb in V1 ~ = − (1 − o ) E0 zb in V1
E P (4.47)
−2 E0 zb in V2 − (2 − o ) E0 zb in V2
We see this form respects the tangential boundary conditions at the interface between
the two dielectrics, as it has to:
zb · E~2 − E
~1 = 0 zb · D~2 − D
~ 1 = (1 − 2 ) E0 = zb · P~2 − P
~1 (4.48)
Because D ~ is different in the two volumes, we must allow the free (and bound) charge
densities to be different. This provides us a set of equations to solve for E0 :
1 E0 = σf ,1 2 E0 = σf ,2 A1 σf ,1 + A2 σf ,2 = Q (4.49)
1 Q 1 A1 + 2 A2
=⇒ E0 = eff = A = A1 + A2 (4.50)
eff A A1 + A2
Q Q A
C = = = eff = eff ,r Cvac (4.51)
∆V a E0 a
− 1 Q
− 1−oQ
( (
zb in V1 zb in V1
~ =
D eff
2
A ~ =
P eff A
2 −o Q ρb = −∇ ~ =0
~ ·P
Q
− eff A
zb in V2 − A
zb in V2
eff
(4.52)
1 Q 1 −o Q
( (
eff A
in V1 eff A
zb in V1
|σf | = 2 Q |σb | = 2 −o Q (4.53)
eff A
in V2 eff A
zb in V2
σb always has the opposite sign as σf . For Q > 0, the sign of σf is positive at z = a
and negative at z = 0. Note that, because P ~ is different in V1 and V2 , so too does σb
differ between the two dielectrics.
1 − o Q Q
1 o
σt,1 = σf ,1 + σb,1 = − = (4.54)
eff eff A eff A
2 − o Q Q
2 o
σt,2 = σf ,2 + σb,2 = − = (4.55)
eff eff A eff A
This makes sense: since the electric field is the same in V1 and V2 , the total (free +
bound) surface charge density sourcing it should be the same. The total charge
density is a factor o /eff smaller than would be present in the absence of dielectrics
because the bound charge density screens the free charge density. The free charge
density is different in the two regions because the opposite-sign bound charge density
is different because of the different dielectric constants. In contrast to our naive
expectation, the free charge density is not uniform on the conductor; rather, it
redistributes itself so the fundamental condition, that the conductors be
equipotentials, is satisfied when one includes the effect of the dielectric. Instead, the
total charge density is uniform, which yields a field independent of (x, y ), which is
what ensures the equipotential condition is met.
− 0 ~ − 0 − 0
ρb = −∇ ~ = −∇
~ ·P ~ · D =− ∇ ~ =−
~ ·D ρf (4.56)
(Homogeneity is required so the gradient does not act on .) Therefore, if there is no
free charge density in a linear, homogeneous dielectric, there is no bound charge
density either. Thus, the dielectric volume satisfies Laplace’s Equation. All our
machinery for solving Laplace’s Equation applies here.
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 279
Section 4.4 Electrostatics in Matter: Boundary Value Problems with Linear Dielectrics
We always need boundary conditions, though, and we can use the ones we derived
~ and D
earlier (the tangential E ~ conditions will yield the same condition on V , so we
start with the simpler one):
h i h i
nb · D~2 − D
~ 1 = σf sb · E~2 − E
~1 = 0 (4.57)
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 280
Lecture 17:
Electrostatics in Matter III:
Boundary Value Problems with Linear Dielectrics (cont.)
Electrostatic Energy in Linear Dielectrics
Date Revised: 2022/02/11 00:00
Date Given: 2022/02/09
Page 281
Section 4.4 Electrostatics in Matter: Boundary Value Problems with Linear Dielectrics
Example 4.6: Spherical cavity in a dielectric medium with uniform field applied
Let’s apply the above to a spherical cavity of radius R in a medium with permittivity
~ = E0 zb applied. There is no free charge anywhere. Our
with a uniform field E
boundary conditions therefore are
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 282
Section 4.4 Electrostatics in Matter: Boundary Value Problems with Linear Dielectrics
where we have applied the requirement that V be finite at the origin to eliminate the
1/r `+1 terms for Vin . Recall that we cannot eliminate the r ` terms for Vout because
the potential does not vanish at infinity.
Let’s first apply the r → ∞ condition. We did this before in the case of a metal
sphere in a uniform field, and we found
Aout
1 = −E0 Aout
`6=1 = 0 (4.62)
B1out B`6out
=1
Ain
1 R = −E0 R + Ain `
`6=1 R = (4.63)
R2 R `+1
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 283
Section 4.4 Electrostatics in Matter: Boundary Value Problems with Linear Dielectrics
Finally, let’s take the radial derivative and apply the matching condition on it, again
using orthonormality:
2 out B`6out
=1
0 Ain
1 = − E0 + 3
B1 0 Ain
`6=1 ` R
`−1
= − (` + 1) (4.64)
R R `+2
− o 3
Ain out
`6=1 = B`6=1 = 0 B1out = − E0 R 3 Ain
1 =− E0 (4.65)
2 − o 2 + o
3 3
Vin (r ) = V (r < R) = − E0 r cos θ = − E0 z (4.66)
2 + o 2 + o
− o R3
Vout (r ) = V (r > R) = −E0 r cos θ − E0 2 cos θ (4.67)
2 + o r
~ · rb
p 4π 3 3 o
= −E0 z + with p ~=− R E0 ( − o ) zb
4 π o r 2 3 2 + o
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 284
Section 4.4 Electrostatics in Matter: Boundary Value Problems with Linear Dielectrics
The potential inside the cavity is that of a uniform electric field in the same direction
as the applied field but multiplied by the factor 3 /(2 + o ) > 1, while the potential
outside is that of the uniform field plus that of a dipole whose orientation is opposite
the uniform field and whose magnitude is given above. It is as if the cavity acquired a
polarization density in the negative z direction, though of course that cannot happen
because χe (r < R) = 0 there and thus P(r ~ < R) = o χe (r < R)E ~ (r < R) = 0. The
polarization density outside the cavity is just the total (not the applied uniform) field
times − 0 (which is not particularly illuminating).
~ = R) = nb · ( − o ) E
σb = nb · P(r ~out (r = R)
R3
∂ − o − o
= ( − o ) −b
r · E0 zb − E0 2 cos θ = −3 o E0 cos θ
∂r 2 + o r r =R 2 + o
(Notice that nb = −b r because nb is taken to point out of the dielectric medium in the
definition of σb .) We see the boundary of the cavity acquires a surface charge density
with the same magnitude and cosine dependence as the bound charge on the surface
of a uniformly polarized sphere, though with opposite sign (so there is negative charge
at the +z end and positive charge at the −z end). The sign follows naturally from our
arguments about cancellation of dipole charge.
The field is enhanced in the cavity for two reasons: first, there is no polarizable
material to screen the electric field, and, second there is surface charge density on the
cavity’s boundary that creates an additional field in the direction of the uniform field.
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 285
Section 4.4 Electrostatics in Matter: Boundary Value Problems with Linear Dielectrics
For reference, we note that the solution for a dielectric sphere (Griffiths Example 4.7)
in a uniform field looks very similar:
3 o ~ · rb
p
V (r < R) = − E0 z V (r > R) = −E0 z + (4.68)
2 o + 4 π o r 2
4π 3 3 o 4π 3 ~
with ~=
p R E0 ( − o ) zb ≡ R P(r < R) (4.69)
3 2 o + 3
− o
σb = 3 o E0 cos θ (4.70)
2 o +
Basically, exchange o and everywhere to go between the two results. In this case,
the sphere acquires a polarization density 3 o ( − o )/(2 o + ), now in the direction
of the applied field. The surface charge density is also of same form as the cavity case
with the ↔ o exchange. That exchange flips the sign so that the +z end acquires a
positive charge, again as expected from the dipole charge cancellation argument.
From the polarized sphere, one can recover the case of a conducting sphere in an
external uniform field by taking → ∞ as noted earlier.
Section 4.4.1 General Conditions for Linear, Homogeneous Dielectrics Page 286
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Section 4.5.1 Electrostatic Potential Energy due to an Assembly of Free Charge in the Presence of Dielectrics Page 287
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Assuming the potential falls off at infinity, the surface term can be taken out to
infinity to vanish. So, we are then left with
Z ~
D Z
U= ~ (~
dτ 0 E ~ r 0)
r 0 ) · d D(~ (4.73)
0 V
~ from zero
There are two integrals here, one over volume and one over the value of D
~ is of course tied to D
to its final value. E ~ and they vary together.
Section 4.5.1 Electrostatic Potential Energy due to an Assembly of Free Charge in the Presence of Dielectrics Page 288
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
For the case of a linear (but perhaps not homogeneous) dielectric, we may use
~ r ) = (~
D(~ ~ (~
r )E r ) and therefore
Z ~
E Z
U= dτ 0 (~ ~ (~
r 0) E ~ (~
r 0) · d E r 0)
0 V
~Z
E
1
Z h i
= dτ 0 (~r 0) d E ~ (~ ~ (~
r 0) · E r 0)
2 0 V
1 1
Z Z
= r 0 ) E 2 (~
dτ 0 (~ r 0) = dτ 0 E~ (~ ~ r 0)
r 0 ) · D(~ (4.74)
2 V 2 V
If the medium is linear and homogeneous, one can pull outside the integral at any
point, yielding
1
Z Z
U= dτ 0 E 2 (~
r 0) = dτ 0 D 2 (~
r 0) (4.75)
2 V 2 V
Section 4.5.1 Electrostatic Potential Energy due to an Assembly of Free Charge in the Presence of Dielectrics Page 289
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
By contrast, if we wanted to know the total electrostatic potential energy stored in the
assembly of the free and bound charge, we would just do the usual volume integral of
E 2 with o instead of . That energy is smaller because > o . The reason for this
difference is that assembling the medium in the first place, which consists of bringing
positive and negative charges together, creates a system with negative potential
energy, and thus the total potential energy of the system would be lower if we
accounted for the energy of assembling the medium. But we will never pull the
dielectric apart, so it is natural to treat that component of the potential energy as an
offset that is inaccessible and neglect it in the electrostatic potential energy.
Section 4.5.1 Electrostatic Potential Energy due to an Assembly of Free Charge in the Presence of Dielectrics Page 290
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Suppose we start with a system with a free charge distribution ρf that sources a field
~1 in a dielectric medium 1 , yielding a displacement D
E ~ 1 = 1 E
~1 . The initial energy is
1
Z
U1 = dτ E~1 · D
~1 (4.76)
2
Now, with the charges sourcing E ~1 held fixed, let’s introduce a piece of dielectric
occupying the volume V2 and having dielectric constant 2 , replacing the dielectric of
dielectric constant 1 there. The remainder of space outside V2 is occupied by 1 in
both configurations. The electric field and displacement field everywhere change to E ~2
and D ~ 2 , where D
~ 2 (~
r ) = (~ ~2 (~
r) E r ). Note that E ~1 and E ~2 are not identical outside V2 ,
and the same is true for D ~ 1 and D ~ 2 . The dielectric affects the field everywhere, not
just inside V2 . The energy is now
1
Z
U2 = dτ E~2 · D
~2 (4.77)
2
The difference in energy between the two configurations is therefore
1
Z h i
U2 − U1 = dτ E~2 · D
~2 − E~1 · D
~1 (4.78)
2
1 ~2 + 1
Z h i Z h i h i
U2 − U1 = dτ E~2 · D
~1 − E
~1 · D dτ E~2 + E
~1 · D~2 − D
~1 (4.79)
2 2
h i
~ × E
It holds that ∇ ~2 + E
~1 = 0, so it can be derived from a potential V , so the
second integral becomes
1
Z h i
− ~
dτ ∇V · D~2 − D
~1 (4.80)
2
~2 − D
We integrate by parts (the surface term vanishes because it depends on D ~1,
which should vanish as one goes far from the dielectric) to obtain
1
Z h i
~ · D
dτ V ∇ ~2 − D
~1 (4.81)
2
This divergence vanishes because the free charge has not changed between the two
~ = ρf ).
~ ·D
configurations (recall, ∇
1
Z h i
U2 − U1 = dτ E~2 · D
~1 − E
~1 · D
~2 (4.82)
2
~ 2 = 1 E
Now, outside V2 , it holds that D ~2 (remember, only changed inside V2 ), and
~ 1 = 1 E
recall also D ~1 everywhere, so the two terms cancel each other there and the
integrand vanishes outside V2 . Therefore, we can restrict the integral to V2 :
1
Z
U2 − U1 = − ~2 · E
dτ (2 − 1 ) E ~1 (4.83)
2 V2
This is already interesting — even though the field changes in all of space, we need
only look at the before and after fields in the volume V2 rather than the entire system.
1 1 ~ ~
Z
W = U2 − U1 = − ~ ·E
dτ P ~1 ⇐⇒ w =− P · E1 (4.84)
2 V2 2
where we recall that E ~1 is the electric field in the absence of the dielectric and P
~ is the
polarization density of the dielectric, and w refers to an energy density. This is just
like the energy of a dipole in an external electric field, except that the factor of 1/2
accounts for the integration from zero field to actual field, from the fact that the
dielectric polarizes in response to the applied field. We see that the introduction of
the dielectric into an existing electric field in vacuum, holding the source charges fixed,
reduces the overall electrostatic energy.
1 ~2 = − 1 ~2 = 1
~ 2
Z Z Z
− ~ ·E
dτ P dτ ~ 2 − o E
D ~2 · E dτ o E ~ ~
2 − D2 · E2
2 V2 2 V2 2 V2
(4.85)
This is some sort of difference between the total electrostatic potential energy in V2
and the electrostatic potential energy neglecting that associated with the assembly of
the dielectric medium. The expression has two problems: there is no differencing with
the initial configuration, and it neglects the energy stored in V1 . It is part of the
energy difference we are interested in, but not all of. The use of D ~2 · E~2 would suffer
the same problems.
Page 296
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
∂W ∂W ∂C
Fξ Q = − =− (4.86)
∂ξ Q ∂C Q ∂ξ
where we made the second step because, if Q is held fixed, the variation of the system
energy is given entirely by the variation of the capacitance. ξ can be a spatial
displacement coordinate like x, y , or z, or it can be an angular orientation coordinate,
in which case the generalized force is actually a torque.
Any system of conductors can be reduced to a capacitance matrix, so the above can
also be written using Equation 2.77 (recall, D = C −1 )
N N
∂ 1 X 1 X ∂ Dij 1
∂ −1
= − QT
Fξ
Q
=− Qi Qj Dij =− Qi Qj C Q
∂ξ 2 i,j=1 2 i,j=1 ∂ξ 2 ∂ξ
Q
(4.87)
(We have intentionally avoided using the confusing notation Cij−1 , using Dij instead.)
Section 4.5.3 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Free Charge Fixed Page 297
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Let’s consider a parallel-plate capacitor with plate separation d, plate side dimensions
` and w , and with a slab of linear, homogeneous dielectric partially inserted between
the plates, with vacuum from 0 to x and dielectric from x to ` with 0 < x < `.
Let’s do this by calculating the total energy of the slab in the capacitor, with E
dependent on the position of the slab. The energy is (using the calculation of C from
the earlier example)
1 Q2 o w x + w (` − x)
W = with C = (4.88)
2 C d
Therefore,
1 Q 2 dC 1 Q 2 (o − ) w 1 w
F x |Q = − − 2
= = − V 2 ( − o ) (4.89)
2 C dx 2 C2 d 2 d
Section 4.5.3 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Free Charge Fixed Page 298
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Intuitively, the dielectric is pulled in because it lowers the energy of the configuration:
the field energy density is proportional to |E ~ |2 , and |E
~ | ∝ −1 , so the field energy
−1
density is ∝ : larger implies lower energy.
Microscopically, what is happening is that the fringing field of the capacitor polarizes
the dielectric, leading to bound charge on the surface. The bound charge on the
surface is attracted to the free charge on the capacitor plates, causing the dielectric to
be pulled in. It’s a runaway effect, with the movement of the dielectric into the
capacitor leading to greater polarization of the fringing field region, increasing the
bound surface charge density and thereby leading to a greater attractive force. The
system only reaches equilibrium when the dielectric is maximally contained in the
capacitor.
Section 4.5.3 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Free Charge Fixed Page 299
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
In general, we do not encounter the above situation. Rather, we hold the voltages
constant on a set of electrodes while we consider the work done during a virtual
displacement dξ.
Before we get into it, though, let’s ask ourselves what we expect to have happen.
Should the force change depending on whether we hold the voltage or the charge
fixed? No, because the force is due to the arrangement of charges on the conductors
and the dielectric at the current instant in time, not at some point in the future that
is affected by whether the charges or voltages are kept constant.
Let’s model the fixed voltage situation in two steps, first disconnecting the batteries
and holding the charge fixed while we move the dielectric as we did above, then
reconnecting the batteries so that charge flows on to or off of the electrodes and
restores them to their original potentials.
Section 4.5.4 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Voltages Fixed Page 300
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Since we are now focusing on a situation with voltages on electrodes, it makes sense
to think about a set of electrodes i = 1...N with voltages Vi and charges Qi . The
electrodes have a capacitance matrix C . Let’s first consider the change in electrostatic
energy for the first step with the charges held fixed (again, using D = C −1 ):
N N
1 X 1 X
dWfield |Q =d Qi Qj Dij = Qi Qj dDij (4.90)
2 i,j=1 2 i,j=1
Q
The change in the inverse capacitance matrix results in a change in the voltages on
the electrodes given by
N
X
dVi |Q = dDij Qj (4.91)
j=1
Section 4.5.4 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Voltages Fixed Page 301
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
Now, let’s return the voltages to their original values by allowing charge to flow on/off
the electrodes from batteries while holding the dielectrics fixed (i.e., Dij held
constant). The charge transfer at fixed voltage required to undo the above voltage
changes at fixed charge is
N
X N
X
dQk |V = Cki (−dVi )Q = − Cki Qj dDij (4.92)
i=1 i,j=1
The change in the electrostatic energy of the configuration (energy flowing out of the
battery into the field) due to this flow of charge is
N
X N
X N
X
bat
dWfield = Vk dQk |V = − Vk Cki Qj dDij = − Qi Qj dDij
V
k=1 i,j,k=1 i,j=1
= −2 dWfield |Q (4.93)
PN
where we used Cki = Cik and k=1 Vk Cik = Qi .
Section 4.5.4 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Voltages Fixed Page 302
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
As we explained earlier, the force cannot depend on whether the charge is held fixed or
the voltage is held fixed. To ensure we get the same force in the two cases, we
therefore must conclude
∂ Wfield ∂ Wfield
Fξ V = =− = Fξ Q (4.95)
∂ξ V ∂ξ Q
That is, when the battery is involved, we must consider the energy of the entire
system and take the positive gradient of the field energy, rather than considering only
the energy of the field and taking the negative gradient of that energy. The reason
these two gradients are different, with a sign between them, is because the derivative
is calculationally different depending on whether V or Q is held fixed.
Section 4.5.4 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Voltages Fixed Page 303
Section 4.5 Electrostatics in Matter: Electrostatic Energy in and Forces on Linear Dielectrics
∂ Wfield ∂ Wfield
Fξ V = =− = Fξ Q (4.97)
∂ξ V ∂ξ Q
One can check this result using the parallel plate capacitor example by starting with
W = C V 2 /2 instead of W = Q 2 /2 C . Taking the positive derivative at fixed V gives
the same result as taking the negative derivative at fixed Q because C is in the
numerator in the first case while C is in the denominator in the second.
Section 4.5.4 Force and Torque on a Linear, Homogeneous Dielectric in an External Field with Voltages Fixed Page 304
Section 5
Magnetostatics
5.1 Study Guidelines
5.2 Lorentz Forces and Current Densities
5.3 Conservation of Charge and the Continuity Equation
5.4 Fields of and Magnetic Forces between Currents
5.5 Curl and Divergence of the Magnetic Field; Ampere’s Law
5.6 Magnetic Vector Potential
5.7 Boundary Conditions on Magnetic Field and Vector Potential
5.8 Magnetic Multipoles
Page 305
Section 5.1 Magnetostatics: Study Guidelines
Study Guidelines
As with basic electrostatics, you have seen much of the material in this section before
in Ph1c. As with electrostatics, we will use more rigor here. We will also consider
some more advanced topics such as the multipole expansion of the magnetic vector
potential, off-axis fields for azimuthally symmetric configurations, etc. As with basic
electrostatics, we won’t do any examples in lecture or the notes where they would
duplicate Ph1c. But you should be review the examples in Griffiths Chapter 5 and
make sure you are comfortable with them.
Page 306
Section 5.2 Magnetostatics: Lorentz Forces and Current Densities
F~mag = q ~ ~
v ×B (5.1)
Note that the electrostatic force on q is not modified by the fact that it is moving.
See the nice examples in Griffiths of cyclotron and cycloid motion (Examples 5.1 and
5.2). These are at the level of Ph1c, so we do not spend time in lecture on them.
Section 5.2.1 Force on a Moving Point Charge in a Magnetic Field Page 307
Section 5.2 Magnetostatics: Lorentz Forces and Current Densities
The one exception to this is the case of intrinsic magnetic moments of fundamental
particles, which emerge from quantum field theory. In such cases, the magnetic
moment is not identified with a current loop, it is just an intrinsic property of the
particle. Since our proof above requires the Lorentz Force Law, and such moments are
not assocated with a current that experiences the Lorentz Force, the proof does not
apply. In cases concerning such moments, work can be done by the field of the
moment or on the magnetic moment by an external magnetic field because no battery
is required to maintain the magnetic moment.
Line Currents
A current carried by a wire can be modeled as a constant line charge density λ that is
moving at fixed speed v :
I = λv (5.3)
For the sake of the generalizations we will consider below, let us write this as a
position-dependent vector
~I(~ r) ~
r ) = λ(~ v (~
r) (5.4)
where ~v (~
r ) is a function of position and its direction follows the wire. By conservation
of charge, the only position dependence of ~I(~ r ) can be its direction. This implies that
any position dependence in λ(~ r ) must be canceled by the position dependence of the
magnitude of ~ r ). If λ is position-independent, then only the direction of ~
v (~ v may
change with position.
For magnetostatics, we assume that such a line current, and the surface and volume
current densities that follow below, are time-independent, or steady: they were set up
an infinitely long time ago and have been flowing at their current values since then.
We also ignore the discretization of the charge density (in this case λ) and consider it
to be a continuous quantity. This is called the steady-state assumption or
approximation.
Z h i
F~mag = d` ~I(~ ~ r)
r ) × B(~ (5.6)
C
where we have used the fact that d ~ v , and ~I are all in the same direction at any
`, ~
point on the wire because the current flows in the wire. Now, realizing that I is
independent of position along the wire (due to conservation of charge as noted
above), we can pull it out in front of the integral, yielding
Z h i
F~mag = I d~ ~ r)
` × B(~ (5.7)
C
Griffiths Example 5.3 is a nice example of calculating the force on a current loop and
also illustrates the point of the battery supplying the energy to do the work that
appears to be done by the magnetic field. The magnetic field acts like a constraint
force to redirect that work.
Current Densities
Just as we generalized point charges to line, surface, and volume charge densities, we
can generalize single moving point charges to line, surface, and volume current
densities. We have already made the first generalization, which is straightforward to
understand since one intuitively thinks of a current as an ensemble of point charges
moving through a wire.
A surface current density is a current flowing in a sheet; think of water flowing over
the surface of an object. The surface current density K~ is defined by
d~I(~ ~ (~
r) = K r ) d`⊥ = K
b
r) × d~
(~
~
`K (~
r) (5.8)
If one thinks about the surface current density as a moving distribution of a surface
charge density, then
~ (~
K r) ~
r ) = σ(~ v (~
r) (5.9)
A volume current density is a current flowing in a bulk volume; think of water flowing
in a pipe or in a river. The volume current density J~ is defined by
d~I(~ ~ r ) da⊥ = J(~
r ) = J(~ ~ r)
b r ) · nb da J(~ (5.10)
where nb is the normal to the area element da. (If we had defined a normal nb to the
line element d ~
` in the plane of the sheet, we could have used a dot product instead of
a cross product in the definition of the surface current density. But it is conventional
to do it as we have done it.)
If one thinks about the volume current density as a moving distribution of a volume
charge density, then
~ r ) = ρ(~
J(~ r) ~
v (~
r) (5.11)
Z h i
F~mag = da K~ (~ ~ r)
r ) × B(~ (5.13)
S
Z h i Z h i
F~mag = dq ~v (~ ~ r) =
r ) × B(~ r) ~
dτ ρ(~ v (~ ~ r)
r ) × B(~ (5.14)
V
Z h i
F~mag = ~ r ) × B(~
dτ J(~ ~ r) (5.15)
V
It should be clear that we could have considered Equation 5.15 to be the fundamental
statement of the Lorentz Force Law and derived the lower-dimensional versions by
~ Such a reduction
inclusion of appropriate delta functions in the definition of ρ or J.
would be cumbersome because the sheet or line carrying the current may not be easy
to parameterize, but the reduction is conceptually straightforward.
If we take S to be a closed surface, we may apply the divergence theorem to the above:
I Z
~ r) =
r ) · J(~
da nb(~ dτ ∇ ~ r)
~ · J(~ (5.17)
S V(S)
d d r)
∂ ρ(~
Z Z
IS = − QV(S) = − r) = −
dτ ρ(~ dτ (5.18)
dt dt V(S) V(S) ∂t
Page 314
Section 5.3 Magnetostatics: Conservation of Charge and the Continuity Equation
Thus, we have
r)
∂ ρ(~
Z Z
dτ ∇ ~ r) = −
~ · J(~ dτ (5.19)
V(S) V(S) ∂t
Since the surface S is arbitrary, it must hold that the integrands are equal everywhere:
r)
~ r ) = − ∂ ρ(~
~ · J(~
∇ (5.20)
∂t
This is the continuity equation and is effectively the differential version of conservation
of charge.
With this equation, we can define our steady-state assumption more mathematically:
it corresponds to ∂ρ/∂t = 0, which then implies ∇~ · J~ = 0. The interpretation is that
the charge density at any point cannot change with time, which implies that the net
current flow into or out of any point vanishes.
Page 315
Lecture 19:
Magnetostatics II:
Biot-Savart Law
Ampere’s Law, Divergence of B
Vector Potential
Date Revised: 2022/02/16 09:30
Date Given: 2022/02/16
Page 316
Section 5.4 Magnetostatics: Fields of and Magnetic Forces between Currents
~ r ) = µo
~I(~
r 0 ) × (~r −~ r 0) µo d~
` 0 (~
r 0 ) × (~ r 0)
r −~
Z Z
B(~ d`0 0 3
= I 0 3
(5.21)
4π |~
r −~ r | 4π |~
r −~ r |
µo = 4 π × 10−7 N A−2 is the permeability of free space. The magnetic field carries
units of teslas, T = N/(A · m). The Biot-Savart Law is the analogue in magnetostatics
of Coulomb’s Law in electrostatics, and it has the same 1/r 2 dependence.
You are well aware of the result that the field of a straight wire along the z-axis
carrying current I at a transverse distance s from the wire is
~ r ) = µo I φ
B(~ b (5.22)
2π s
where φ
b is the azimuthal unit vector in cylindrical coordinates. The field forms circles
around the wire with orientation set by the right-hand rule. This is derived in Griffiths
Example 5.5, which we will not repeat here since you saw it in Ph1c.
Section 5.4.1 Biot-Savart Law Page 317
Section 5.4 Magnetostatics: Fields of and Magnetic Forces between Currents
Consider the special case of both wires running parallel to the z axis separated by s sb
in the xy -plane, with the first wire on the z-axis itself. Then d ~
` = zb dz, d ~
` 0 = zb dz 0 ,
~ r 0 = s sb + z 0 zb.
r = z zb, ~
Therefore,
h i
d~ r) × d~
` 0 (~
r 0 ) × (~ r 0 ) = dz dz 0 zb × zb × (z − z 0 ) zb − s sb
`(~ r −~ (5.25)
0
= dz dz s sb (5.26)
3/2
r 0 |3 = (z − z 0 )2 + s 2
and |~
r −~ (5.27)
Thus, µo
Z ∞ Z ∞ 1
F~mag = I1 I2 s sb dz dz 0 (5.28)
4π −∞ [(z − z 0 )2 + s 2 ]3/2
−∞
Z ∞ " #∞
µo z0 − z
= I1 I2 s sb dz (5.29)
4π 1/2
−∞ 2 0 2
s [(z − z ) + s ] 2
−∞
Z ∞ Z ∞
µo 2 µo I1 I2
= I1 I2 s sb dz 2 = sb dz (5.30)
4π −∞ s 2π s −∞
where we did the integral using the trigonometric substitution z 0 − z = s tan θ. The
total force is infinite, but we can abstract out of the above expression the force per
unit length on the first wire, which is attractive (pointing towards the second wire) if
the currents flow in the same direction:
µo I1 I2
f~mag = sb (5.31)
2π s
~ (~
K r 0 ) × (~
r −~ r 0) ~ r 0 ) × (~
J(~ r −~ r 0)
~ r ) = µo ~ r ) = µo
Z Z
B(~ da0 B(~ dτ 0
4π |~
r −~ r 0 |3 4π |~r −~r 0 |3
(5.32)
As with the Lorentz Force Law, it should also be clear that one could consider the
volume version to be the fundamental statement of the Biot-Savart Law and one can
derive the lower-dimensional versions by including delta functions in the definition of
~ This does not apply to a reduction to zero dimensionality, as noted above.
J.
There are good examples of the use of the Biot-Savart Law in Griffiths. Again, these
are at the level of Ph1c, so we do not spend time in lecture on them.
Section 5.4.3 General Expressions for Fields due to Current Densities Page 320
Section 5.4 Magnetostatics: Fields of and Magnetic Forces between Currents
~ r 0)
J(~
~ × µo
Z
~ r) = ∇
B(~ dτ 0 (5.34)
4π V |~
r −~ r 0|
We note that, while our derivation of this equation did not appear to require any
assumptions about the way the current behaves at infinity, we will see later that the
steady-state assumption does imply the net current through any sphere must vanish.
Section 5.4.4 Another Form for the Biot-Savart Law Page 321
Section 5.5 Magnetostatics: Curl and Divergence of the Magnetic Field; Ampere’s Law
From the field of a current-carring wire, Equation 5.22, we get the clear impression
that B~ has curl and that the curl is related to the current sourcing the field. Here, we
explicitly calculate this curl from the Biot-Savart Law. Griffiths Section 5.3.2 provides
one technique for this; we use Jackson’s technique instead to avoid duplication.
We take the curl of Equation 5.34 and apply the BAC − CAB rule for the triple vector
product, ∇ ~ × (∇~ ×~ ~ ∇
a) = ∇( ~ ·~ a ) − ∇2 ~ a, writing the coordinate that ∇ ~ acts on
explicitly: " #
~ r 0)
J(~
~ ~r × µo
Z
∇ ~ r) = ∇
~ ~r × B(~ ~ ~r × ∇ dτ 0 (5.35)
4π V |~
r −~ r 0|
" !! !#
µo ~
Z ~ r 0)
J(~
Z ~ r 0)
J(~
= ∇~r ∇ ~ ~r · dτ 0 − ∇~2r dτ 0
4π |~
r −~ 0
r | |~
r −~ r 0|
V V
(5.36)
" Z ! Z !#
µo ~ ~
J(~ 0
r ) ~
J(~ 0
r )
= ∇~r dτ 0 ∇ ~ ~r · − dτ 0 ∇~2r (5.37)
4π V |~r −~ r 0| V |~
r −~ r 0|
We were able to bring ∇ ~ ~r and ∇2 inside the integrals because ∇ ~ ~r is with respect to ~
r
~
r
and the integral is over ~ ~ ~r is with respect to ~
r 0 . Similarly, because ∇ r and J~ is a
r 0 , J~ passes through the divergence in the first term and the Laplacian in
function of ~
the second one, preserving the necessary dot product in the first term and the
vectorial nature of the second term:
1 1
Z Z
∇ ~ r ) = µo ∇
~ ~r × B(~ ~ ~r ~ r 0) · ∇
dτ 0 J(~ ~ ~r − ~ r 0 ) ∇2
dτ 0 J(~
0 ~
r 0
4π V |~
r −~r | V |~
r −~r |
(5.38)
1 1
~ ~r
∇ ~ ~r 0
= −∇
|~
r −~r 0| |~
r −~r 0|
1
∇~2r = −4 π δ(~ r 0)
r −~
|~
r −~r 0|
The first equation may seem surprising if one considers the exchange ~ r ↔~ r 0 , but one
can see it is true by simply evaluating the gradient on both sides or by defining
r −~
~s = ~ r 0 and applying the offset and inversion tricks we used in electrostatics. The
second is Equation 3.34 with the exchange ~ r ↔~ r 0 (where here there is no sign flip
because the Laplacian is quadratic in the derivatives and the delta function is
symmetric in its argument). Applying them, we obtain
1
~ r ) = µo −∇
Z Z
~ × B(~
∇ ~ ~r ~ r 0) · ∇
dτ 0 J(~ ~ ~r 0 + 4 π dτ 0 ~ 0
J(~
r ) r
δ(~ − ~
r 0
)
4π V |~
r −~r 0| V
(5.39)
~ r ), yielding
The second term just becomes 4 π J(~
1
~ r ) = µo −∇
Z
~ × B(~
∇ ~ ~r ~ r 0) · ∇
dτ 0 J(~ ~ ~r 0 ~ r 0)
+ µo J(~ (5.40)
4π V |~
r −~r 0|
~ · (f ~
We can apply the product rule ∇ ~ +f∇
a · ∇f
a) = ~ ~ ·~
a to rewrite the first term:
!
Z
1
Z ~ r 0)
J(~
Z ~ r 0)
~ ~r 0 · J(~
∇
~ r 0) · ∇
dτ 0 J(~ ~ ~r 0 = ~ ~r 0 ·
dτ 0 ∇ − dτ 0
V |~
r −~r 0| V |~
r −~ r 0| V |~r −~ r 0|
(5.41)
!
I ~ r 0)
J(~
0 0
= r )·
da nb(~ =0 (5.42)
S(V) |~
r −~ r 0|
We used the divergence theorem to transform the first term into a surface integral,
and then we take the surface to infinity. Assuming the currents are localized, the
integrand vanishes on that surface, causing the first term to vanish. The second term
~ r 0 ) = 0 under the steady-state assumption by the continuity
~ ~r 0 · J(~
vanishes because ∇
equation with ∂ρ/∂t = 0. Thus, we obtain, under the steady-state assumption,
~ r ) = µo J(~
~ × B(~
∇ ~ r) (5.43)
This equation is the differential version of Ampere’s Law, which we will return to
shortly.
Let’s discuss some subtleties in the above derivation connected to the vanishing of the
~ ∇
∇( ~ ·~a) term. There are two points to make:
~ · (∇
The vector identity ∇ ~ ×~
a) = 0 combined with Equation 5.34 immediately implies
∇ ~ r) = 0
~ · B(~ (5.44)
The magnetic field has no divergence. This immediately implies there are no magnetic
point charges: magnetic fields are sourced by currents only. It should be realized that
this apparent fact is really an assumption inherent in the Biot-Savart Law. If we had
added to the Biot-Savart Law a second term that looks like Coulomb’s Law, due to
magnetic monopoles, then the above divergence would have yielded that density of
magnetic charge on the right side. It is an empirical observation that there are no
magnetic monopoles, and hence we assume that magnetic fields are only sourced by
currents via the Biot-Savart Law. That magnetic fields are sourced by currents at all is
also an empirical observation; the Biot-Savart Law simplify codifies that observation.
General Thoughts on the Curl and Divergence of the Electric and Magnetic
Field
Considering the corresponding expressions for electrostatics, we recognize that the
electric field has divergence equal to the charge density because of the empirical
observation of Coulomb’s Law describing the electric field. It has a vanishing curl
because of the empirical absence of a current that sources electric fields in the way
that electric currents source magnetic fields; if there were a Biot-Savart-like term that
added to Coulomb’s Law, then the electric field would have curl. We can in fact guess
that, if magnetic monopoles existed, moving magnetic monopoles would generate an
electric field in the same way that moving electric monopoles generate a magnetic
field.
The key point in all of the above is that the nature of the divergence and the curl of
the electric and magnetic fields reflect empirical observations about the way these
fields are generated. These are not derivable results: they are inherent in the formulae
we wrote down for the electric and magnetic fields, which themselves are based on
observations.
We will see later that we can replace the assumption of Coulomb’s Law and the
Biot-Savart Law with an assumption about a potential from which the electric and
magnetic fields can be derived. But, again, we can only make that assumption because
it yields the correct empirical relations, Coulomb’s Law and the Biot-Savart Law.
Section 5.5.3 General Thoughts on the Curl and Divergence of the Electric and Magnetic Field Page 328
Section 5.5 Magnetostatics: Curl and Divergence of the Magnetic Field; Ampere’s Law
The left side can be transformed using Stokes’ Theorem into a line integral around the
edge of S, which we denote by the closed contour C(S), while the right side is just
total current passing through C(S), Iencl :
I
d~ ~ r ) = µo Iencl
` · B(~ (5.46)
C(S)
As before, there are a number of examples in Griffiths that are at the level of Ph1c, so
we do not spend time on them here.
But this form, implied by the Biot-Savart Law, is not the only form. We had freedom
with the electrostatic potential to add an offset. Here, we can add any curl-less
~ without affecting B.
function to A ~ The form above corresponds to the additional
condition
∇ ~ r) = 0
~ · A(~ (5.48)
~=A
A ~0 + ∇λ
~ (5.49)
Then
∇ ~=∇
~ ·A ~ 0 + ∇2 λ
~ ·A (5.50)
If we require the left side to vanish, then we have a version of Poisson’s Equation:
∇2 λ = −∇ ~0
~ ·A (5.51)
~
One thus sees one of the motivations for the assumed form ∇λ.
Let’s choose boundary conditions that place the boundary at infinity with the field
falling off at infinity. For these boundary conditions, we know from Coulomb’s Law
that the solution to Poisson’s Equation is
1
Z ~ ·A
∇ ~0 (~
r 0)
r) =
λ(~ dτ 0 (5.52)
4π V |~
r −~r 0|
~ × ∇λ
The vector calculus identity ∇ ~ = 0 implies that ∇ ~ ×A ~=∇ ~ ×A ~0 and thus the
magnetic field is the same for the two vector potentials (our second motivation for the
~
choice to add ∇λ). We thus have an explicit formula for the term that has to be
added to A~0 so that the resulting form A
~ is divergenceless while leaving the magnetic
field unchanged.
The above explicit formula may not be valid if we assume different boundary
conditions, but we know Poisson’s Equation always has a solution, so we are
guaranteed that the desired function λ(~
r ) exists.
Let us make a final point about how the above relates to the connection between
~ ·A
∇ ~ = 0 and the behavior of the currents at infinity. It is not true that starting with
~ ·A
∇ ~0 6= 0 corresponds to a different physical assumption about the currents at
infinity: changing ∇~ ·A~ has no effect on the fields and thus can have no effect on the
currents. Our standard formula for A ~ is only valid under the assumption ∇ ~ ·A~ = 0,
and so the relation between ∇ ~ ·A~ and the assumption about how the currents behave
is only valid for that form. If one assumes a different form for A,~ one that has
~ ·A
∇ ~ 6= 0, then taking its divergence will not necessarily result in the particular
expressions that we encountered before in deriving the differential form of Ampere’s
Law, so the interpretation of ∇ ~ ·A~ = 0 will be different, and the mathematical
manifestation of the currents vanishing at infinity will also change. One benefit of the
choice ∇ ~ ·A~ = 0 is that this mathematical manifestation is simple.
Note that the vector components of A ~ and J~ line up. Thus, the last equation is a
component-by-component Poisson’s Equation. Again, under the assumption that the
currents are localized and for appropriate boundary conditions (as we assumed in
providing the alternate version of the Biot-Savart Law that we previously used to
~ we know the solution:
define A),
~ r 0)
J(~
~ r ) = µo
Z
~ r ) = −µo J(~
~ r) localized currents
∇2 A(~ ⇐⇒ A(~ dτ 0 (5.56)
4π V |~
r −~ r 0|
This is just Equation 5.47 again. Essentially, we can think of the three components of
the current density as sourcing the three components of the vector potential in the
same way that the electric charge density sources the electric potential.
Section 5.6.3 Alternate Proof of the Form for the Magnetic Vector Potential Page 334
Section 5.6 Magnetostatics: Magnetic Vector Potential
We can consider the specific cases of line and surface current densities as volume
current densities that include delta functions specifying the localization to a line or
sheet. When one does the volume integral, the delta function reduces the
three-dimensional integral over the volume to one- or two-dimensional integrals over a
line or sheet, yielding:
~I(~
r 0) K~ (~
r 0)
~ r ) = µo ~ r ) = µo
Z Z
A(~ d` A(~ da0 (5.57)
4π C |~
r −~ r 0| 4π S |~
r −~ r 0|
Note that the units of the vector potential are unchanged: the change in the units of
the current densities are canceled by the change in the units of the measure of
integration.
Section 5.6.4 The Vector Potential for Line and Surface Currents Page 335
Section 5.6 Magnetostatics: Magnetic Vector Potential
The intuition part is to recognize that, because B ~ is along the z-axis inside the
solenoid and vanishing outside and because A ~ “wraps around” B, ~ it is natural to
assume A~ is along φ.
b Then one can do the calculation in the same way as one applies
Ampere’s Law, except that instead of current through a surface (“enclosed current”),
we have enclosed magnetic flux, and, instead of a line integral of magnetic field
around the edge of the surface, we have a line integral of vector potential. Please
study the details in Griffiths, as a variant on this problem will be given in homework.
Section 5.6.4 The Vector Potential for Line and Surface Currents Page 336
Lecture 20:
Magnetostatics III:
Uniqueness Theorem
Boundary Conditions
Date Revised: 2022/02/18 00:00
Date Given: 2022/02/18
Page 337
Section 5.6 Magnetostatics: Magnetic Vector Potential
Just as we did for electric fields, we can show that, given a current distribution and a
well-defined set of boundary conditions, the magnetic field obtained is unique. We
assume that a current distribution J(~~ r ) in a volume V is specified. We will see later
how specific we must be about the boundary conditions.
First, we need something analogous to the Green’s Identities we used in the case of
electrostatics. Using the vector identity ∇ a×~
~ · (~ b) = ~
b·∇~ ×~ a−~ a·∇~ ×~ ~
b, letting u
and ~v be two arbitrary vector fields, and applying the identity with ~
a=u ~ and
~
b=∇ ~ ×~ v , we may write
Z Z h i
~ · (~
dτ ∇ ~ ×~
u × (∇ v )) = ~ ×~
dτ (∇ ~ ×u
v ) · (∇ ~) − u ~ × (∇
~ · (∇ ~ ×~
v )) (5.59)
V V
Since the expression on the left-hand side is a divergence, we may turn it into a
surface integral using the divergence theorem:
I Z h i
da nb · (~ ~ ×~
u × (∇ v )) = ~ ×u
dτ (∇ ~ ×~
~) · (∇ v) − u ~ × (∇
~ · (∇ ~ ×~
v )) (5.60)
S(V) V
From the above equation, we can see what (minimal) boundary condition information
we must have to obtain uniqueness of B: ~ we must have that, at any given point on
~ B,
the surface, A, ~ nb × A,
~ or nb × B~ is specified. If this is true, then A
~3 = A ~2 − A~1 = 0
where A ~ is specified, B
~3 = B~2 − B ~ 1 = 0 where B~ is specified,
~3 = nb × (A
nb × A ~2 − A~1 ) = 0 where nb × A~ is specified, and nb × B~ 3 = nb × (B
~2 − B
~1 ) = 0
where nb × B ~ is specified. Requiring one of these four conditions at every point on
S(V) ensures the integrand on the right side vanishes at every point on S(V) and thus
the right side vanishes. Since the integrand on the left side is nonnegative, it must
therefore vanish everywhere: B ~ 3 = 0. Hence, B ~1 = B ~ 2 and the fields are identical and
the field solution is unique.
We have already discussed how the A ~ that generates a particular B~ is unique up to the
gradient of an additional function if its divergence is left unspecified. The above
theorem for the uniqueness of the magnetic field therefore now tells us that
specification of J~ in the volume and of A,
~ B,
~ nb × A,
~ or nb × B ~ on the boundary gives a
vector potential that is unique up to the gradient of an additional function if its
divergence is unspecified. But what do we need to know to completely determine the
vector potential?
Obtaining a unique vector potential is the equivalent of being able to also know the λ
function (up to an offset). We showed that λ satisfies Poisson’s Equation with ∇~ ·A~
~ we would need to
as the source, Equation 5.51. So, clearly, to obtain a unique A,
~ We also would need appropriate boundary conditions for this Poisson
~ · A.
specify ∇
Equation. We may conclude from our proof of the uniquess of the scalar potential (up
~ on the boundary to obtain a
to an offset) that we must either specify λ or nb · ∇λ
unique λ (again, up to an offset) and thus a unique A. ~
Which of the above conditions provide the necessary boundary condition on λ? Only
~ on the boundary is certain to be sufficient. This gives ∇λ
specification of A ~ and thus
~
nb · ∇λ, a Neumann boundary condition for λ and thus sufficient to render λ unique.
~ r ) = −∇U(~
B(~ ~ r) (5.64)
One must take some care, though: in addition to being current-free, the region under
consideration must be simply connected. Griffiths Problem 5.29 shows a situation
~ ×B
where the current in a region may vanish but ∇ ~ 6= 0 because the region is not
simply connected and the enclosed volume outside the region contains current.
∇2 U(~
r ) = −∇ ~ r) = 0
~ · B(~ (5.65)
Our usual assumption of simple boundary conditions — everything falls off to zero at
infinity — yields a trivial result here, U(~
r ) = 0, so we must assume less trivial
boundary conditions to obtain a nonzero U. We will return to the use of the
magnetostatic scalar potential in connection with magnetically polarizable materials.
Recall that Gauss’s Law, ∇ ~ ·E~ = ρ/o , implied that the normal component of the
electric field satisfied Equation 2.55
h
~2 (~ ~1 (~
i 1
r) · E
nb(~ r) − E r) = r)
σ(~ (5.66)
o
That is, the normal component of the magnetic field is continuous at any boundary.
For the tangential component, we return to the derivation leading to Equation 2.57.
In that case, we considered a contour C that consisted of two legs C1 and C2 parallel
to the interface and to each other and two legs normal to the interface whose length
would be shrunk to zero. We saw
I Z ~
rb −b r ) dz
n(~ 2
Z ~
rb +b r ) dz
n(~ 2
d~ ~ (~
`·E r) = − ~1 (~
E r) · d~
`+ ~2 (~
E r) · d~
` (5.68)
C ra −b
C1 ,~ r ) dz
n(~ 2
ra +b
C2 ,~ r ) dz
n(~ 2
Z ~
rb h i
dz→0 ~2 (~ ~1 (~
−→ E r) − E r) · d~
` (5.69)
ra
C2 ,~
h i h i
~ 2 (~
B ~ 1 (~
r) − B r ) · sb(~
r ) = µo K~ (~
r ) × nb(~
r ) · sb(~
r) (5.73)
Note that this condition holds for any sb tangential to the interface. To give some
~ has the magnitude of K
intuition, nb × K ~ (because nb ⊥ K~ always) but points in a
~ while still tangent to the interface. The sign is set by the
direction perpendicular to K
cross-product right-hand rule.
~ 2 (~
B ~ 1 (~
r) − B ~ (~
r ) = µo K r ) × nb(~
r) (5.74)
captures both boundary conditions: the projection of B ~ normal to the interface (along
nb) is continuous because the projection of the right side along that direction vanishes,
and the projection of B~ along any sb parallel to the interface can be discontinuous by
the projection of µo K~ × nb along that direction. This is a very nice relation: given K
~,
it provides a way to calculate the change in the entire magnetic field across the
interface, not just the change of a component.
We can rewrite the above in another way. Take the cross product of both sides with
nb(~
r ) from the left. The right side becomes a triple vector product, which we can
rewrite using the BAC − CAB rule, ~ a × (~ b×~ c) = ~
b(~a·~ c) − ~ a·~
c (~ b). The second term
has nb · K~ , which vanishes, while the first term has nb · nb = 1. Thus, we have
h i
r) × B
nb(~ ~ 2 (~ ~ 1 (~
r) − B ~ (~
r ) = µo K r) (5.75)
The earlier form is more useful when K ~ is specified, and the second form would more
~ if the fields are specified. Note, however, that this form does not
easily yield K
preserve the information about the normal component of B ~ because the contribution
of that component to the left side vanishes.
~2 (~
A ~1 (~
r) − A r) = 0 (5.76)
While the vector potential itself is continuous, its derivatives are not necessarily
~ which is not necessarily continuous.
continuous because its derivatives are related to B,
Evaluating these discontinuities is a bit harder than in the case of the electric potential
because the derivatives are not related in a trivial component-by-component way to
the field. We need an expression involving second derivatives of A ~ if we want to obtain
boundary conditions on the first derivatives of A.~ Let’s use Equation 5.56:
~ r ) = −µo J(~
∇2 A(~ ~ r) (5.77)
We have used Cartesian coordinates rather than a coordinate system using nb, tb, and sb
because the latter vary in direction depending on where one is on the surface; their
derivatives do not vanish, so we would not have been able to pull them inside the
Laplacian as we did with xb.
Given the above, we now apply the same kind of geometry we used to derive the
boundary condition on the normal component of E~ . That yields
h i
~ xb · A
nb · ∇ ~2 (~
r) − ∇ ~ xb · A ~1 (~ ~ (~
r ) = −µo xb · K r) (5.79)
h i
~ xb · A
nb · ∇ ~2 (~ ~1 (~
r ) − xb · A r) = (5.80)
~ and K
The above argument holds for the yb and zb projections of A ~ also, so we may
combine them to obtain
h i
~ A
nb · ∇ ~2 (~ ~1 (~
r) − A ~ (~
r ) = −µo K r) (5.81)
Thus, we see that the normal derivative of each component of the vector potential has
a discontinuity set by the surface current density in the direction of that component of
the vector potential. This is a lot like the discontinuity in the normal component of
the electric potential being determined by the surface charge density at the boundary.
We may derive, from the above, conditions in the normal and tangential directions by
recognizing that
~ nb = 0
nb · ∇ ~ sb = 0
nb · ∇ (5.82)
These relations should be intuitively obvious: the direction of nb, sb, and tb change as
one moves transversely along the surface (along sb or tb), but they simply are not
defined off the surface and thus they can have no derivative in that direction. This
implies that the normal derivative of the normal component of A ~ has no discontinuity
since there can be no surface current in that direction:
n h io
~2 (~
~ nb · A
nb · ∇ ~1 (~
r) − A r) =0 (5.83)
It also implies that the normal gradient of the vector potential in a particular direction
parallel to the interface changes by the surface current density in that direction:
n h io
~2 (~
~ sb · A
nb · ∇ ~1 (~
r) − A r) ~ (~
= −µo sb · K r) (5.84)
Next, let’s consider the tangential derivatives of the vector potential. Here, we use the
vector identity
~ ×∇
∇ ~ r) = 0
~ A(~ (5.85)
where again we consider each component of A ~ as a scalar function and the above
equation holds for all three components. If we again project by Cartesian components;
e.g.
~ ×∇
∇ ~ r) = 0
~ xb · A(~ (5.86)
then we can apply the same type of argument as we applied for calculating the
~ , which in this case yields
boundary condition on the tangential components of E
h i
~ xb · A
sb · ∇ ~2 (~
r) − ∇ ~ xb · A ~1 (~
r) = 0 (5.87)
h i
~ xb · A
sb · ∇ ~2 (~ ~1 (~
r ) − xb · A r) = (5.88)
Since the argument again generalizes to any Cartesian component, we may combine
the three expressions to obtain
h i
~ A
sb · ∇ ~2 (~ ~1 (~
r) − A r) = 0 (5.89)
Page 356
Section 5.8 Magnetostatics: Magnetic Multipoles
Magnetic Multipoles
Derivation of Magnetic Multipole Expansion
Since the vector potential is sourced by the current distribution in a manner similar to
the way the charge distribution sources the electric potential, it is natural to develop
the same multipole expansion. We follow Jackson for the sake of generality and
variety; you can of course read the derivation in Griffiths, too. We continue to make
the steady-state assumption, and now we also make the assumption they are localized.
We start with the equation for the vector potential in terms of the current distribution:
~ r 0)
J(~
~ r ) = µo
Z
A(~ dτ 0 (5.90)
4π V |~
r −~ r 0|
∞
X r< `
1
0
= `+1
P` (cos γ) (5.91)
|~
r −~r | r
`=0 >
where r< and r> and the smaller and larger of r and r 0 .
As with the multipole expansion for the electrostatic potential, we will take r r 0 :
we want to know what the potential looks like far away from the current distribution.
Therefore, r< = r 0 and r> = r :
∞
~ r ) = µo (r 0 )`
Z
~ r 0)
X
A(~ dτ 0 J(~ P` (cos γ) (5.92)
4π V `=0
r `+1
∞
~ r ) = µo 1 1
Z
~ r 0 ) r 0 ` P` (cos γ)
X
dτ 0 J(~
A(~ `
(5.93)
4 π r `=0 r V
Now, consider the first term, which is just the volume integral of the current density.
Under the steady-state assumption, it is intuitively clear this integral must vanish. To
~ · (f ~
prove this explicitly, we first use the vector identity ∇ ~ ·~
a) = f ∇ ~ with
a∇f
a+~
a = J~ and f = ri any of the Cartesian coordinates:
~
3 3
~ = ri ∇
~ · (ri J) ~ · J~ + J~ · ∇r
~ i =0+
X ∂ X
∇ Jj ri = Jj δij = Ji (5.94)
j=1
∂rj j=1
where the first term vanishes because of the steady-state assumption and so continuity
~ · J~ = 0. With this, we can compute the integral using the divergence
implies ∇
theorem:
Z Z h i I h i
r 0) =
dτ 0 Ji (~ dτ 0 ∇ ~ r 0) =
~ 0 · r 0 J(~ da0 nb(~ ~ r 0) = 0
r 0 ) · ri 0 J(~ (5.95)
i
V V S(V)
where the surface integral in the last term vanishes because the current distribution is
localized.
∞
~ r ) = µo 1 1
Z
~ r 0 ) r 0 ` P` (cos γ)
X
dτ 0 J(~
A(~ `
(5.96)
4 π r `=1 r V
This is the multipole expansion of the vector potential of the current distribution. As
with the multipole expansion of the electric potential, one can see that the successive
terms fall off as successively higher powers of 1/r .
It makes sense that there is no monopole term because ∇ ~ ·B~ = 0: if there were a way
to make a current distribution look like a monopole from far away, then one would
have a field configuration with a nonzero Gauss’s law integral of magnetic flux through
a closed surface containing the current distribution, which is not allowed by ∇ ~ = 0.
~ ·B
We must first prove an identity. We start with the same vector identity as before, now
with f = ri rj and ~ ~
a = J:
~ = ri rj ∇
~ · (ri rj J)
∇ ~ · J~ + J~ · ∇(r
~ i rj ) = 0 + rj J~ · ∇r
~ i + ri J~ · ∇r
~ j (5.99)
= rj Ji + ri Jj (5.100)
where we have again used ∇~ · J~ = 0. We apply the same technique of integrating over
volume and turning the left side into a surface term that vanishes, so we are left with
Z h i
dτ 0 ri 0 Jj (~
r 0 ) + rj 0 Ji (~
r 0) = 0 (5.101)
V
~2 term as:
We can use this identity to rewrite the A
3
µo 1 X 1h
Z i
~2 (~
A r) = ri rj dτ 0 r 0 ) rj 0 − Jj (~
Ji (~ r 0 ) ri 0 (5.102)
3
b
4 π r i,j=1 V 2
where we split out half of the Ji rj 0 factor and used the identity to exchange the
indices. You have learned in Ph106a and hopefully elsewhere that the cross-product
can be written
3 1 for cyclic index permutations
a×~
X
(~ b)k = kmn am bn with kmn = −1 for anticyclic index permutations
0 when any two indices are identical
m,n=1
(5.103)
where kmn is the Levi-Civita symbol. Multiplying this definition by ijk and summing
over k gives
3 3 3
a×~
X X X
ijk (~ b)k = ijk kmn am bn = kij kmn am bn (5.104)
k=1 k,m,n=1 k,m,n=1
µo 1 1
Z
0
0 0
=− r×
~ dτ ~ r × J(~ r ) (5.109)
4 π r3 2 V
~ r) = 1 ~
Z
M(~ ~ r)
r × J(~ and ~ =
m ~ r 0)
dτ 0 M(~ (5.110)
2 V
~2 (~ ~ ×~
µo m r
A r) = (5.111)
4 π r3
Interestingly, this form has the same radial dependence as that of the electrostatic
potential of a dipole, but the cross-product in the numerator differs from the dot
product in the numerator of the electric dipole potential. However, because the
magnetic field is obtained from the curl of the vector potential, while the electric field
is obtained from the gradient of the electric potential, we will see that the two forms
result in the same field configuration (up to normalization)!
Now, let us consider a current loop. The only assumption we make is that the current
throughout the loop is the same so that we can extract it from the integral. The
volume integral reduces to a line integral over the loop contour:
µo 1 1 µo 1 r 0 × d~
~ ` 0 (~
r 0)
I I
~2 (~
A r) = − r×
~ r 0 × I d~
~ ` 0 (~
r 0) = − r ×I
~ (5.112)
4 π r3 2 C 4 π r3 C 2
The integral is now just a geometric quantity that has units of area. Separating out
the magnetic moment, we have
~ loop × ~
µo m r r 0 × d~
~ ` 0 (~
r 0)
I
~ loop (~
A r) = ~ loop = I
m (5.113)
2
4π r3 C 2
Section 5.8.3 Specialization of Magnetic Dipole Potential to a Current Loop Page 365
Section 5.8 Magnetostatics: Magnetic Multipoles
For the case of a loop confined to a plane that contains the origin, the quantity
r 0 × d~
~ ` 0 /2 is the differential area element for the loop: it is the area of the triangle
formed by ~ r 0 , the vector from the origin to a point on the loop, and d ~` 0 , the line
element tangent to the loop at ~ 0
r and in the direction of the current, and this cross
product has the standard right-hand-rule orientation. The integral thus calculates the
area of the loop! Thus, for a planar loop, the above reduces to
~2 (~ µo 1
A r) = − r × I nb a
~ (5.114)
4 π r3
where a is the loop area and nb is the normal to the loop with orientation defined by
the current via the right-hand rule. Therefore, for this case, we have
Section 5.8.3 Specialization of Magnetic Dipole Potential to a Current Loop Page 366
Section 5.8 Magnetostatics: Magnetic Multipoles
~2 (~ µo m sin θ b
A r) = φ ≡ A2,φ φ
b (5.116)
4π r2
1 ∂ µo m cos θ
B2,r (~
r) = (sin θ A2,φ ) = 2 (5.117)
r sin θ ∂θ 4π r3
1 ∂ µo m sin θ
r) = −
B2,θ (~ (r A2,φ ) = (5.118)
r ∂r 4π r3
B2,φ (~
r) = 0 (5.119)
~ 2 (~ µo m
or B r) = 2 r cos θ + b sin θ
θ (5.120)
4 π r3
b
~ = m zb:
Let’s derive the more generic result by releasing the condition m
3 3
~ 2 (~ ~=
~ ×A
X ∂ Ak µo X ∂ m r
` m
B r) = ∇ ijk rbi = ijk rbi k`m (5.121)
i,j,k=1
∂rj 4π i,j,k,`,m=1
∂rj r3
3
m` δjm 3 m` rm
µo X
= ijk k`m rbi − 2rj (5.122)
4π i,j,k,`,m=1
r3 2 r5
We
P3 use the cyclicity of the Levi-Civita Psymbol in its indices and the identities
3
k=1 kij k`m = δi` δjm − δim δj` and j,k=1 jki jk` = 2δi` to rewrite the above in a
form identical to that of the electric dipole, Equation 3.240:
3 3 3
~ 2 (~ µ o
X 2m i 3 X X
B r) = rbi 3 − 5 mi rj rj − ri mj rj (5.123)
4 π i=1 r r j=1 j=1
3
µo X 3 ri (m~ ·~
r ) − mi (~
r ·~
r)
= rbi (5.124)
4 π i=1 r5
~ 2 (~ ~ · rb) rb − m
µo 3 (m ~
=⇒ B r) = (5.125)
4π r3
As we did for electric multipoles, let’s consider the problem of the force and torque on
a magnetic dipole. However, because there is no magnetic potential energy function,
we must begin from the Lorentz Force on the current distribution, which is given by
Z
F~mag = ~ r ) × B(~
dτ J(~ ~ r) (5.126)
V
~ r ).
As we did in the case of the force on an electric multipole, we Taylor expand B(~
Again, as we did for electrostatics, we place the multipole at the origin and will
generalize the result later. The expansion is
3
X ∂ Bk
Bk (~ r = ~0) +
r ) = Bk (~ rm + ··· (5.127)
m=1
∂rm ~r =~0
3 Z
F~mag =
X
ijk rbi dτ Jj (~
r ) Bk (~
r) (5.128)
i,j,k=1 V
3 3
" Z #
∂ Bk
X Z X
= ijk rbi Bk (~0) dτ Jj (~
r) + dτ Jj rm + · · · (5.129)
i,j,k=1 V m=1
∂r m ~
0 V
We have done both these integrals before. The first one contains the monopole of the
current distribution, which vanishes as in Equation 5.95. Since we will see that the
second term is in general nonzero and is proportional to the magnetic dipole moment,
let’s call it F~dip and focus on it, dropping the higher-order terms. It is very similar in
structure to what we encountered in calculating the dipole term in Equation 5.98.
Applying the same tricks we used there to obtain Equation 5.107, we may rewrite it as
3 Z
∂ Bk 1
h i
F~dip = ~ r) × ~
X
ijk rbi dτ jmn J(~ r (5.130)
i,j,k,m,n=1
∂rm ~0
V 2 n
3
∂ Bk 1
Z h i
~ r)
X
=− ijk jmn rbi mn with ~ =
m dτ ~r × J(~
i,j,k,m,n=1
∂rm ~0 2 V
(5.131)
We use the vector identity Equation 5.105, 3j=1 jik jmn = δim δkn − δin δkm , and also
P
use ijk = −jik to adjust the indices to match this expression, yielding
3
∂ Bk
F~dip =
X
(δim δkn − δin δkm ) rbi mn (5.132)
i,k,m,n=1
∂rm ~0
3
" ! ! #
X ∂ Bk ∂ Bk
= rbi mk − mi (5.133)
i,k=1
∂ri ~ ∂rk ~
0 0
= ∇~ m ~ ·B ~ − m ~ ∇~ ·B~ (5.134)
~0 ~ 0
The second term vanishes. Generalizing the first term to a dipole at an arbitrary
position, we have
1
h i Z h i
F~dip = ∇
~ m ~ r)
~ · B(~ with m ~ = dτ ~ ~ r 0)
r 0 × J(~ (5.135)
2 V
The force causes the magnetic dipole to move to a local maximum of m ~ Note how
~ · B.
it is identical to the force on an electric dipole in an electric field, Equation 3.250.
We’ll address below the implication that the magnetic field can do work on the dipole.
We may obtain from the Lorentz Force Law on a current distribution the
corresponding torque:
Z h i
~ mag =
N ~ r ) × B(~
r × J(~
dτ ~ ~ r) (5.136)
V
where we have just added up the torque volume element by volume element in the
same way we summed the force. When we Taylor expand the magnetic field, we have
Z h i
~ mag =
N ~ r ) × B(
r × J(~
dτ ~ ~ ~0) + · · · (5.137)
V
Because of the ~r × inside the integrand, the zeroth-order term no longer vanishes and
so we do not need to consider the next order term in the Taylor expansion. We will
write the zeroth-order term as N ~ dip below for reasons that will become clear.
To get the above expression into a familiar form, we need to repeat the same kinds of
vector arithmetic tricks we have used before. First, we apply the BAC − CAB rule,
a × (~
~ b×~ c) = ~ a·~
b(~ c) − ~ a·~
c (~ b), which we can do without having to write things in
terms of indices because there are no derivatives floating around:
Z h i Z h i Z h i
~ dip =
N ~ r ) × B(
r × J(~
dτ ~ ~ ~0) = ~ r) ~
dτ J(~ ~ ~0) −
r · B( ~ ~0) ~
dτ B( ~ r)
r · J(~
V V V
(5.138)
We can make the second term vanish by the same kinds of tricks we used earlier
during the vector potential multipole expansion:
~ r ) = 1 ∇r ~ r) = 1 ∇
h i h i n h i o
~ r ) = r ∇r
r · J(~
~ ~ · J(~ ~ 2 · J(~ ~ r) − r2 ∇
~ · r 2 J(~ ~ r)
~ · J(~ (5.139)
2 2
In this expression, the second term vanishes under the steady-state assumption, and
~ r ). Since we
the first term can be turned into a surface integral with integrand r 2 J(~
are considering a localized current distribution, the surface can be taken far enough
~ r ) vanishes on the surface.
out that J(~
The first term looks again like the expression we have encountered in Equation 5.98,
which becomes apparent when we write it out in component form:
3 Z
~ dip =
X
N rbi Bj (~0) dτ Ji (~
r ) rj (5.140)
i,j=1 V
1
Z h i
~ ~0) × m
= −B( ~ with ~ =
m dτ ~ ~ r)
r × J(~ (5.142)
2 V
1
Z h i
~ dip = m
N ~ r)
~ × B(~ with ~ =
m dτ 0 ~ ~ r 0)
r 0 × J(~ (5.143)
2 V
The magnetic dipole feels a torque that tends to align it with the magnetic field (the
torque vanishes when m ~ again like the situation for an electric
~ is aligned with B),
dipole in an electric field.
r ) = −m
U(~ ~ r)
~ · B(~ (5.144)
This form for the potential energy expresses two features of magnetic dipoles: they like
~
~ · B.
to be aligned with the local magnetic field, and they seek the region of largest m
The thing that should be concerning about this expression is that we argued earlier
that magnetic fields can do no work, yet here we have the possibility of such work.
That is because we are assuming m ~ is held fixed. For a finite current loop, there must
be a battery doing work to keep the current fixed as m~ moves or turns relative to B: ~
such motion yields changing magnetic fields, which, as you know from Ph1c, generate
voltages around the loop in which the current for m~ flows. The battery will be the
thing doing the work to counter these voltages and keep the current flowing. If m ~ is a
property of a fundamental particle, then there is no explicit battery: it is simply an
empirical fact that |m|
~ cannot change, and one that we must incorporate as a
postulate.
Page 376
Lecture 22:
Magnetostatics in Matter I:
Field of a Magnetized Object
Date Revised: 2022/02/24 23:00
Date Given: 2022/02/24
Page 377
Section 6.1 Magnetostatics in Matter: Paramagnetism and Diamagnetism
Page 378
Section 6.2 Magnetostatics in Matter: The Field of a Magnetized Object
dm ~ r ) dτ
~ = M(~ (6.1)
M~ is not to be confused with the magnetization density M(~ r ); the latter can be for
some arbitrary current distribution, while the former is specifically to be considered to
be a density of magnetic dipole moments. M(~ ~ r ) for this special case
r ) should give M(~
of pure dipoles. We will, confusingly, drop “macroscopic” from here on out. Assuming
we are looking at the dipoles from a macroscopic enough scale that the dipole
approximation is valid, we may use our expression for the vector potential of a
magnetic dipole, Equation 5.111, to calculate the contribution to the vector potential
at ~
r due to the above infinitesimal volume at ~r 0:
~ r 0 ) × (~
µo d m(~ r −~r 0) ~ r 0 ) × (~
µo dτ 0 M(~ r 0)
r −~
~ r) =
d A(~ = (6.2)
4π |~
r −~ 0
r | 3 4π |~
r −~ 0
r | 3
~ r 0 ) × (~
M(~ r −~ r 0)
~ r ) = µo
Z
A(~ dτ 0 0 3
(6.3)
4π V |~
r −~ r |
1
~ r ) = µo
Z
A(~ ~ r 0) × ∇
dτ 0 M(~ ~ ~r 0 (6.4)
4π V |~
r −~ r 0|
!
µo
Z ~ ~
∇ 0 × M(~ 0
r ) µo
Z ~ r 0)
M(~
= dτ 0 ~r − dτ 0 ~
∇ ~
r 0 × (6.5)
4π V |~ r 0|
r −~ 4π V |~
r −~ r 0|
µo
Z ~ r 0)
~ 0 × M(~
∇ µo
Z ~ r 0 ) × nb(~
M(~ r 0)
= dτ 0 ~r 0
+ da0 0
(6.6)
4π V |~
r −~
r | 4 π S(V) |~
r −~ r |
where, in the last step, we have used a vector identity that we will prove on the
following slide.
Let’s prove the vector identity we just used, which is a corollary of the divergence
theorem for the curl. Let ~
a(~r ) be an arbitrary vector field and let ~
c be an arbitrary
constant vector. Then the divergence theorem tells us
Z I
~ · [~
dτ ∇ r) × ~
a(~ c] = r ) · [~
da nb(~ r) × ~
a(~ c] (6.7)
V S(V)
Now, apply the cyclicity of triple scalar products (along with the fact that ~
c is
constant and can thus it can be moved past ∇) ~ and bring ~ c outside the integrals
(since it is a constant vector):
Z h i I
c·
~ dτ ∇~ ×~ c·
r) = ~
a(~ da [b r) × ~
n(~ a(~
r )] (6.8)
V S(V)
Since ~
c is arbitrary, the expression must hold for any ~
c and thus:
Z h i I
dτ ∇~ ×~
a(~
r) = da [b r) × ~
n(~ a(~
r )] (6.9)
V S(V)
Making some definitions, we recognize that the vector potential can be considered to
be sourced by a bound volume current density J~b (~
r ) and a bound surface current
~ b (~
density K r ):
J~b (~
r) = ∇ ~ r)
~ × M(~ ~ b (~
K ~ r ) × nb(~
r ) = M(~ r) (6.10)
~ r ) = µo J~b (~
r 0) µo ~ b (~
K r 0)
Z I
A(~ dτ 0 + da0 (6.11)
4π V |~
r −~ r 0| 4 π S(V) |~
r −~ r 0|
The way in which these current densities source A ~ is identical to the way in which free
current densities do. Moreover, we can see the clear analogy to bound volume and
surface charges in the case of polarized materials.
Griffiths Section 6.2.2 gives a nice discussion of the physical inerpretation of bound
currrents that will be presented in class, but there is not much to add here.
J~b (~ ~ × M zb = 0
r) = ∇ ~ b (~
K r ) = M zb × nb = M zb × rb = M sin θ φ
b (6.12)
We need to calculate
Z 2π Z π
~ r ) = µo R 2 M sin θ 0 φb
A(~ dφ 0 dθ 0 sin θ 0 0|
(6.13)
4π 0 0 |~
r − ~
r
µo 2 2 π
Z π
M sin θ 0 (−bx sin φ 0 + yb cos φ 0 )
Z
0
= R dφ dθ 0 sin θ 0 (6.14)
4π 0 0 |~
r −~r 0|
(The R 2 out front is because an area integral, not just a solid angle integral, needs to
be done.) This is done in Griffiths Example 5.11 via explicit integration. For the sake
of variety, let’s use a different technique. We use Equation 3.167, the Spherical
Harmonic Addition Theorem Corollary, which expands |~ r −~r 0 |−1 in terms of spherical
harmonics, recognizing |~ r 0 | = R because the integral is over the sphere of radius R:
∞ X ` `
1 X 1 r< ∗
0
= 4π `+1
Y`m (θ 0 , φ 0 )Y`m (θ, φ) (6.15)
|~
r −~r | `=0 m=−`
2 ` + 1 r>
Let’s consider the xb piece of the above angular integral; the other term will be similar
in spirit. We will write theRnumerator
R in terms of spherical harmonics and use the
expansion. We abbreviate 02 π dφ 0 0π dθ 0 sin θ 0 = dΩ0 and recall
R
Y`,−m = (−1)m Y`,m ∗ . Applying these facts yields
sin θ 0 (− sin φ 0 )
Z
dΩ0 = (6.16)
|~
r −~r 0|
∞ X ` `
8 π Y1,1 (θ 0 , φ 0 ) + Y1,−1 (θ 0 , φ 0 )
r
1 r<
Z X
dΩ0 4π Y ∗ (θ 0 , φ 0 )Y`m (θ, φ)
`+1 `m
3 2i `=0 m=−`
2 ` + 1 r>
The integral over Ω0 gives δ`,1 δm,1 and δ`,1 δm,−1 , eliminating the sum and yielding
`
sin θ 0 (− sin φ 0 )
r
4π 8 π 1 r<
Z
dΩ0 = `+1
[Y1,1 (θ, φ) + Y1,−1 (θ, φ)] (6.17)
|~
r −~r 0| 2i 3 3 r>
4 π r<
=− 2
sin θ sin φ (6.18)
3 r>
where the 1/3 came from 1/(2 ` + 1). We get back the same type of angular
r 0 | turned into the prefactor shown, which has the
r −~
dependence, but with the 1/|~
correct dimensions.
We can repeat the same kind of manipulation for the yb term, yielding
Therefore,
Recall that |~r 0 | = R because the surface integral was over the sphere of radius R, so
r> (r< ) is replaced by R in the first (second) expression above.
~ r ) is continuous at r = R, as we expect.
Note that A(~
Evaluating the curl of the first term to obtain the magnetic field, we have inside the
sphere
~ ≤ R) = 1 µo M 2 rb cos θ − θb sin θ = 2 µo M
h i
~ ≤ R) = ∇
B(r ~ × A(r ~ (6.22)
3 3
For r ≥ R, we have
~ × rb
~ ≥ R) = µo m 4π 3 ~
A(r ~ =
m R M (6.23)
4 π r2 3
which is the vector potential (thus yielding the field of) a pure dipole with magnetic
moment given by integrating the uniform magnetization density over the sphere. This
form is exact for all r ≥ R.
~ (~ 1 ~ ~ r ) = 2 µo M
~
r ≤R E r) = − P B(~ (6.24)
3 o 3
1 p ~ · rb ~ r ) = µo m~ × rb
r ≥R V (~
r) = A(~ (6.25)
4 π o r 2 4 π r2
4π 3~ 4π 3 ~
~=
p R P m~ = R M (6.26)
3 3
Inside the sphere, the difference is a factor of −2 and the exchange of 1/o for µo .
Outside the sphere, the two potentials result in fields identical up to the replacement
~ by M
of P ~ and again 1/o by µo . The difference in the r ≤ R expressions reflects the
~ is aligned with M
fact that the magnetic field of the bound surface current (i.e., of M) ~
~ is opposite to
while the electric field of the surface bound charge density (i.e., of P)
~ This sign difference is a generic phenomenon, resulting in the very different
P.
behavior of electrostatic and magnetostatic fields in matter.
Page 388
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
We saw that A ~ is sourced by the bound current density J~b = ∇ ~ ×M~ in the same way
it would be sourced by a free current density J~f . Therefore, Ampere’s Law is satisfied
with the sum of the two currents:
1 ~ ~ = J~f + J~b = J~f + ∇ ~
~ ×M
∇×B (6.27)
µo
If we want to write an Ampere’s Law in terms of the free currents only, in the same
way that we wanted to write Gauss’s Law in terms of the free charges only, then we
can define the auxiliary field
~
~ ≡ B −M
H ~ (6.28)
µo
In contrast to electrostatics, where the displacement field was the sum of the electric
field and the polarization density, here the auxiliary field is the difference of the
magnetic field and the magnetization density. The sign flip comes from the differing
signs in the definition of the bound charge and current densities: ρb = −∇ ~ ·P~ while
J~b = ∇ ~ which itself comes from the commutative vs. anticommutative natures
~ × M,
of the dot and cross product.
~ we then have
With this definition of H,
∇ ~ = 1 ∇
~ ×H ~ −∇
~ ×B ~ = J~f + J~b − J~b = J~f
~ ×M (6.29)
µo
Therefore, we have a modified Ampere’s Law
I Z
∇ ~ = J~f
~ ×H ⇐⇒ d~ ~ r) =
` · H(~ r ) · J~f (~
da nb(~ r ) = If ,enc (6.30)
C S(C)
Thus, as intended, we have an Ampere’s Law in terms of the free currents only, which
~ The fact that H
(partially) source H. ~ satisfies Ampere’s Law in the free current leads
some to use the name applied field for it. That may be misleading, though, because
the free current does not tell one everything one must know to determine H ~ (in the
same way that ρf does not completely determine the displacement field D). ~
∇ ~ = −∇
~ ·H ~
~ ·M (6.31)
~
~ ≤ R) = B(r ≤ R) − M
H(r ~ = 2M ~ = −1 M
~ −M ~ (6.32)
µo 3 3
~ ~ =
field of the magnetic dipole m 4π ~
R3 M
~ ≥ R) = B(r ≥ R) =
H(r 3
(6.33)
µo µo
This all makes sense given the Helmholtz theorem: since ∇ ~ ·H~ does not vanish, H ~ is
not just the curl of a vector potential, but must be the sum of the gradient of a scalar
potential and the curl of a vector potential. Ampere’s Law for H ~ tells us that the free
current density sources the vector potential, while −∇ ~ sources the scalar potential.
~ ·M
We will see later that the latter point allows us to use our electrostatic boundary value
problem techniques.
We can make the same point about ρf not being the only source of D; ~ when ∇ ~ ×P~ is
nonzero, then D~ receives an additional sourcing term. It was not convenient to make
this point when we discussed D ~ initially because we had not yet learned about vector
potentials and how to discuss sourcing of D ~ by a vector field, ∇ ~ But now we do,
~ × P.
~ received a contribution that is sourced by ∇
and so it should be clear that D ~ ×P ~ in
the same way that H ~ receives a contribution that is sourced by ∇ ~ ×H~ = J~f .
In particular, in Example 4.5, the capacitor with two side-by-side dielectrics, we saw
such a situation, manifested by the discontinuity in the tangential component of P ~ at
the interface between the two dielectrics.
~
Who Cares About H?
~ any more useful than D
Is H ~ was?
On the other hand, H ~ is sourced by the free currents, which is the thing one explicitly
controls in the lab. For that reason alone, we expect H~ is of greater utility than D.
~
We will see this more clearly when we consider specific types of permeable materials.
The other reason we will find H~ more useful is that, in reality, we frequently come
across ferromagnets, where M ~ is provided and thus we are given the ∇ ~ ·M~ source for
~ but we rarely encounter ferroelectrics, where P
H, ~ and thus the ∇ ~ ×P ~ source for D
~
are provided. We would find D ~ useful as a calculation tool if we were given a system
in which ∇~ ×P ~ were nonzero or, more likely, P
~ were tangent to boundaries between a
ferroelectric and vacuum or between different ferroelectrics. Then ∇ ~ ×P ~ and any
~ would source D
discontinuity in nb × P ~ in the same way that J~ and a boundary K ~
~
source B.
Section 6.3.3 ~
Who Cares About H? Page 393
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
~
Boundary Conditions on H
h i h i
r) · H
nb(~ ~ 2 (~ ~ 1 (~
r) − H r ) = −b r) · M
n(~ ~ 2 (~ ~ 1 (~
r) − M r) (6.34)
Applying the same arguments using Ampere’s Law for H~ as we did using Ampere’s
~ we can also conclude the analogy of Equation 5.73:
Law for B,
h i h i
~ 2 (~
H ~ 1 (~
r) − H r ) · sb(~
r) = K~ f (~
r ) × nb(~
r ) · sb(~
r) (6.35)
Section 6.3.4 ~
Boundary Conditions on H Page 394
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
Recall that we found alternative forms of the corresponding boundary conditions for
~ Equations 5.74 and 5.75:
B,
~ 2 (~
B ~ 1 (~
r) − B ~ (~
r ) = µo K r ) × nb(~
r)
h i
nb(~ ~ 2 (~
r) × B ~ 1 (~
r) − B r ) = µo K ~ (~
r)
There is no trivial analogue of the first one because it relied on the normal component
~ being continuous. However, we can obtain the analogue of the second equation,
of B
though we have to do it in a different way because, for B, ~ we used the first equation
above to obtain the second one.
Section 6.3.4 ~
Boundary Conditions on H Page 395
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
We start by using sb = tb × nb and then applying the cyclicity of the triple scalar product
on both sides:
h i
~2 − H
~ 1 · tb × nb = [b ~f
H n × sb] · K (6.36)
h i
tb · nb × H ~2 − H ~f
~ 1 = tb · K (6.37)
The same equation holds trivially with tb replaced by nb: the left side vanishes because
nb is perpendicular to any cross product involving nb and the right side vanishes because
K~ f is always perpendicular to nb. This, combined with the fact that tb in the above can
be any vector in the plane of the boundary, implies the more general statement
h i
r) × H
nb(~ ~ 2 (~ ~ 1 (~
r) − H ~ f (~
r) = K r) (6.38)
which is the analogue of the second equation on the previous slide. But note that this
equation provides no information about the normal component of H ~ because it is
related to the normal component of M.~
Section 6.3.4 ~
Boundary Conditions on H Page 396
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
~ = χm H
M ~ (6.39)
~ is sometimes called the applied field!) Since B
(One can see why H ~ = µo H~ +M~ ,
we have
~ = µo H
B ~ +M~ = µo (1 + χm ) H ~ ≡ µH ~ (6.40)
Diamagnetic materials exist via the same kind of classical argument, now involving the
response of currents in materials to applied fields.
The analogous paraelectric materials (χe < 0) do not exist for the most part — it is
hard to understand how one can get an electrically polarizable material to have χe < 0.
Metals can have negative permittivity at high frequencies (optical), but not DC.
~ M,
With the linear relationship between H, ~ and B,~ we can rewrite the boundary
conditions we derived earlier in a somewhat simpler form.
~ implies
The continuity of the normal component of B
h i
nb(~ ~ 1 (~
r ) · µ1 H ~ 2 (~
r ) − µ2 H r) = 0 (6.41)
µ1 ~ µ2 ~
r) ·
nb(~ r) −
M1 (~ M2 (~
r) = 0 (6.42)
χm,1 χm,2
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 399
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
" #
~ 2 (~
B r) ~ 1 (~
B r) h i
− · sb(~
r) = K~ f (~
r ) × nb(~
r ) · sb(~
r) (6.43)
µ2 µ1
" #
~ 2 (~
B r) ~ 1 (~
B r)
or r) ×
nb(~ − ~ f (~
=K r) (6.44)
µ2 µ1
and
" #
~ 2 (~
M r) ~ 1 (~
M r) h i
− · sb(~
r) = K~ f (~
r ) × nb(~
r ) · sb(~
r) (6.45)
χm,2 χm,1
" #
~ 2 (~
M r) ~ 1 (~
M r)
or r) ×
nb(~ − ~ f (~
=K r) (6.46)
χm,2 χm,1
Vanishing of Kf will of course simplify these expressions, yielding the continuity of the
~
tangential component of B/µ ~ m.
and M/χ
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 400
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
~ tells us
Let’s first see how far we can get without using χm . Ampere’s Law for H
I Z
~ · d~
H `= da nb · J~f (6.47)
C S(C)
I ~ I s2 b
s≤R: 2 π s H(s) = π s 2 ⇐⇒ H(s) = φ (6.48)
π R2 2πs R2
I ~ I
s≥R: 2 π s H(s) = π R 2 ⇐⇒ H(s) = φ
b (6.49)
π R2 2πs
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 401
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
B(s ~ ≥ R) = µo I φ
~ ≥ R) = µo H(s b (6.50)
2πs
~ ≤ R) = χm H(s
~ ≤ R) = χm I s 2 b µ − µo I s 2 b
M(s φ= φ (6.51)
2 π s R2 µo 2 π s R2
and therefore
2
B(s ~ ≤ R) = µ I s φ
~ ≤ R) = µ H(s b (6.52)
2 π s R2
All three fields are azimuthal inside and outside R. For paramagnetic materials,
χm ≥ 0 (µ ≥ µo ), so M ~ is parallel to H
~ and |B|
~ > µo |H|~ inside R. For diamagnetic
materials, χm < 0 (µ ≤ µo ), so M ~ is antiparallel to H
~ and |B|
~ ≤ µo |H|
~ inside R.
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 402
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
Let’s check the boundary conditions. All the fields are tangential at the boundary, so
the normal conditions — continuity of the normal components of B, ~ µ H,
~ and
~ m — are trivially satisfied. There is no free surface current density, so we
µ M/χ
~ B/µ,
expect the tangential components of H, ~ ~ m to be continuous. We see
and M/χ
this indeed holds, with them taking on the values
~ ~
b · H(s
φ b · B(s = R) = φ
~ = R) = φ b · M(s = R) = I (6.53)
µ χm 2πR
The last one is a bit tricky because both the numerator M ~ and the denominator χm
vanish for s > R, but L’Hopital’s rule allows evaluation of the ratio in the limit
χm → 0. The zb tangential components are trivially continuous since they all vanish.
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 403
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
For the sake of completeness, let’s calculate the bound surface current and check that
the boundary conditions on B ~ are correct. The bound surface current is K ~b = M~ × nb
(Equation 6.10). In this case, nb = sb, the radial unit vector in cylindrical coordinates, so
~ b (s = R) = M(s = R) φ
b × sb = −χm I
K zb (6.54)
2πR
For a paramagnetic materials (χm > 0), the surface current points along −b z while, for
diamagnetic materials (χm < 0), it points along +b z . One can see this physically by
considering the direction of alignment of the dipoles and which direction the
uncancelled current on the boundary flows. From the direction of this surface current,
one can then see that the field of this surface current adds to the field of the free
current for the paramagnetic case and partially cancels it for the diamagnetic case.
~ It has no normal component in
Finally, let’s check the boundary conditions on B.
either region, so continuity of the normal component is trivially satisified. The
discontinuity in the tangential component matches Equation 5.75:
h
~2 − B
i
~ 1 = sb × [µo − µ] I b I ~b
nb × B φ = −µo χm zb = µo K (6.55)
2πR 2πR
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 404
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
Let’s also calculate the bound volume current density, J~b = ∇ ~ from
~ ×M
Equation 6.10. It is
J~b (~ ~ = χm ∇
~ ×M
r) = ∇ ~ = χm J~f = χm I zb
~ ×H (6.56)
πR 2
For paramagnetic materials, J~b is parallel to J~f and thus its field adds to the field of
the free current, while, for diamagnetic materials, it is antiparallel and it partially
cancels the free current’s field.
Note that the integral of J~b over the cross section and the integral of K
~ b over the
circumference are equal in magnitude and opposite in sign, canceling perfectly. This is
why the magnetic field outside the wire is only that due to the free current.
A modest extension to this problem would be to include a free surface current in the zb
direction, which would then cause a discontinuity in the φ ~ B/µ
b component of H, ~ and
~ m . You should try this on your own.
M/χ
Section 6.3.6 Boundary Conditions for Linear Magnetic Materials Page 405
Lecture 24:
Magnetostatics in Matter III:
Nonlinear Materials and Ferromagnetism
Boundary Value Problems in Magnetostatics
Date Revised: 2022/03/02 09:30
Date Given: 2022/03/02
Page 406
~ and Magnetic Permeability
Section 6.3 Magnetostatics in Matter: The Auxiliary Field H
Both phenomena are caused by unpaired electrons as paramagnetism is; one might
have guessed this by the fact that all three phenomena involve the alignment of
magnetic dipoles with the applied field. The additional dipole-dipole interaction that
causes nonlinearity is due to the exchange effect in quantum mechanics. We will
explain this in the following.
As you know, electrons in atoms occupy shells corresponding to different energies for
the electron-nucleus Coulomb interaction. Only a certain number of states are allowed
for each shell (n2 for shell n), and electrons can be put in a shell with spin “up” or
spin “down” (multiplying by 2 the number of allowed states).
When a shell is partially filled, the electrons prefer to be unpaired, meaning that they
have different orbital wavefunctions (probability distributions) and the same spin
direction (i.e., aligned spins) rather than the same orbital wavefunction and different
spin directions. This behavior, where electrons prefer to be in different orbitals but
have the same spin, is frequently termed Hund’s Rules.
The reason for this preference against having the same orbital wavefunction is that the
electrostatic repulsive energy of two electrons in the same orbital state is high: in
quantum mechanics, that energy is determined by the integral of the product of their
wavefunctions weighted by 1/|~ r −~r 0 | where ~ r 0 are their positions, so the less
r and ~
similar their wavefunctions are, the lower the (positive) electrostatic repulsive energy is.
Next, considering the Pauli exclusion principle, the overall state (spatial wavefunction
and spin) must be antisymmetric under exchange of the two electrons. One could
achieve this by taking the spatial wavefunction to be antisymmetric and the spin state
to be symmetric under exchange or vice versa. It turns out that, when one calculates
the Coulomb repulsion energy integral, there is a second term that arises due to the
extra terms created by requiring the overall state to be symmetric or antisymmetric
under exchange. Moreover, because of the symmetry constraints on the overall state,
this exchange term is negative when the spin state is symmetric and positive when it
is antisymmetric. Thus, the exchange term ensures that, not only must the overall
state be antisymmetric with the two electrons in different spatial wavefunctions, but
also that the antisymmetry must be in the spatial wavefunction, not the spin state.
With the above, it would still be possible for the electrons to either have the same spin
projection or for them to have opposite spin projections and the spin state to be the
symmetric combination of the two possible anti-aligned states. It turns out that
spin-orbit coupling causes the latter state to be higher energy, so the case in which the
two electrons are aligned is preferred. Thus, we are able to explain Hund’s Rules.
In addition, though, one needs a mechanism for unpaired electrons in nearby atoms to
align with each other; alignment of the unpaired electrons in a single atom is not
enough. A similar exchange interaction is required, of which there are many types that
depend on the details of the material and how the electrons in nearby atoms interact.
The key requirement for such exchange effects to occur, though, is delocalization —
the electron wavefunctions must be large enough that they spread to nearby atoms —
so that there can be exchange interactions between electrons in adjacent atoms. This
explains why nonlinearity occurs only in atoms with d- and f -shell electrons — the
electrons in these orbitals are more weakly bound than s- and p-shell electrons,
providing the necessary delocalization.
Ferromagnets have domains, which are regions of coaligned magnetic dipoles, caused
by the aforementioned nonlinearity: it is energetically favorable for the magnetic
moments of unpaired electrons in nearby atoms to align. By default, these domains
are macroscopic in size (fractions of a mm, perhaps), but they do not align with each
other because alignment would create a large magnetic field outside the material,
which is not a low-energy state (which we will see when we talk about magnetic
energy). When a large field is applied, though, the energy cost of not aligning with the
magnetic field (m~ ·B ~ applied ) is larger than the energy savings of not having a large field
~
energy (|Bmaterial |), and so the domains align with the applied field.
We then must consider the phenomenon of saturation, whereby, at large fields, one
gets to the point where all the unpaired electrons’ dipole moments are aligned with
the field and there are no more magnetic dipoles left to align. The magnetization
density stops increasing and saturates. The applied field H ~ may continue to be
increased, but B~ = µo (H~ +M ~ sat ), where M~ sat is the saturated magnetization density.
Thus, B ~ increases due to the first term only. (At lower fields, M ~ increases with H,
~
leading to a large amplification of H ~ to yield B.)
~
After bringing a nonlinear paramagnetic material into saturation, what happens when
one turns off the applied field? The large exchange interaction energy makes it
favorable for the moments to remain aligned with the direction of the applied
magnetic field that has been removed. It is not that full alignment is the
lowest-energy state, but that there is an energy barrier between the fully aligned state
and the lower-energy state with randomly aligned domains. In fact, to reduce M ~ and
B~ to zero requires a significant H
~ in the direction opposite to B.
~ After B~ goes
through zero, it can then begin to align with H ~ again and one can reach saturation in
the other direction. And so on.
This phenomenon of the magnetization (and thus B) ~ being dependent on past history
~ ~ but, in addition, B
is called hysteresis: not only is B a nonlinear function of H, ~
depends on the history of H.~ Hysteresis curves are shown in Griffiths Figures 6.28 and
6.29.
There is not much more we can say about ferromagnetism without considering specific
cases.
µ − µo ~ µ − µo µ − µo
J~b = ∇ ~ =∇
~ ×M ~ × H = ~ =
~ ×H
∇ J~f (6.57)
µo µo µo
~ =∇
B ~
~ ×A ~ = H(
H ~ B)
~ ~ = J~f
~ ×H
∇ (6.58)
∇ ~ ∇
~ × H( ~ = J~f
~ × A) (6.59)
~ and B
If the relation between H ~ is not simple, the above equation may be difficult to
solve.
1 ~
~ ×
∇ ~
∇×A = J~f (6.60)
µ
If we further specify that µ is constant over some region, then in that region we have
~ × ∇
∇ ~ =∇
~ ×A ~ ∇ ~ − ∇2 A
~ ·A ~ = µ J~f (6.61)
~ = −µ J~f
∇2 A (6.62)
In principle, one can apply the same techniques as we used for solving Poisson’s
Equation in electrostatics to solve this component by component. Boundary
~
conditions must be specified either directly (recall that we proved that if any one of A,
~ nb × A,
B, ~ or nb × B~ is specified at every point on the boundary, then the resulting field
(though not necessarily the vector potential) is unique) or by matching using the
conditions on the normal and tangential components at boundaries.
Another technical challenge associated with the above equation is that it only
separates cleanly into component-by-component Poisson Equations in Cartesian
coordinates. If the current distribution it not naturally represented in Cartesian
coordinates (e.g., even a simple circular current loop), then separation of variables
may not be feasible. Method of images may work, or one may have to resort to other
techniques or numerical solution. None of this technical complication takes away from
the fact that there will be a unique solution for each component independently. The
technical complication just makes it hard to actually obtain that solution.
~ · µo
∇ ~ +M
H ~ =0 (6.63)
−∇2 VM + ∇ ~ =0
~ ·M (6.64)
∇2 VM = −ρM with ρM = −∇ ~
~ ·M (6.65)
(note the canceling minus signs in the definitions!) where ρM is termed the
magnetostatic charge density. Note the close similarity to the definition of the bound
charge density ρb = −∇~ ·P~ for dielectrics. This equation can be solved by the
standard techniques for solving Poisson’s Equation.
In particular, if the boundary is at infinity and we require the fields to fall off to zero
there, we know the Green Function for the above equation, which yields
1
Z
∇ ~ r 0)
~ ~r 0 · M(~
r) = −
VM (~ dτ 0 (6.66)
4π V |~r −~ r 0|
1 1
Z
VM (~
r) = ~ r 0) · ∇
dτ 0 M(~ ~ ~r 0 (6.67)
4π V |~
r −~r 0|
r 0 to ~
We change variables on the gradient from ~ r in the usual way, picking up a sign:
1 1
Z
r) = −
VM (~ ~ r 0) · ∇
dτ 0 M(~ ~ ~r (6.68)
4π V |~
r −~r 0|
1 ~
Z ~ r 0)
M(~
r) = −
VM (~ ∇~r · dτ 0 (6.69)
4π V |~
r −~r 0|
If we want to know the potential and field far from the region that is magnetized, and
we can assume the magnetization M ~ is confined to a finite region (localized), we can
make the approximation |~
r −~r 0 |−1 ≈ r −1 and pull this factor outside the integral,
which gives
1 ~ 1
Z
r) = −
VM (~ ∇~r · ~ r 0)
dτ 0 M(~ (6.70)
4π r V
1 m~ ·~r
Z
= with m~ = ~ r 0)
dτ 0 M(~ (6.71)
4π r 3
V
~ = o m,
That is, the scalar potential is equal to that of an electric dipole with p ~
implying the field is equal to that of a magnetic dipole m.~ (The factor of µo will
reappear when one calculates B ~ instead of H).
~ Any magnetized object looks like a
dipole from far enough away, which is not surprising.
1
Z ~ r 0)
~ ~r 0 · M(~
∇ 1
I
nb(~ ~ r 0)
r 0 ) · M(~
r) = −
VM (~ dτ 0 + da0 (6.72)
4π V |~r −~ r 0| 4 π S(V) |~
r −~ r 0|
This second term looks like the bound surface charge density term in the corresponding
expression in electrostatics, so we define a magnetostatic surface charge density
σM (~
r ) = nb(~ ~ r)
r ) · M(~ (6.73)
and see that it sources the magnetostatic scalar potential in the same way that ρM
does. Together, both terms look identical to Equation 4.8. One must take some care
about the sign of the surface term. nb is defined to be the outwardly directed normal
from the magnetized region out into vacuum. This is why σM has the sign definition
that it does. This convention is consistent with the definition of σb , which also used
the outwardly directed normal.
Mz m~ · rb 4π ~
VM (r ≤ R) = VM (r ≥ R) = with ~ =
m π R3 M (6.74)
3 4 π r2 3
(
~
~ = −∇V
~ M = −M r ≤R
H 3 (6.75)
~ field of a magnetic dipole m
H ~ r ≥R
1 ~ ~ = 2 µo M
~ = µo H
B ~ +M
~ =⇒ ~ ≤ R) = µo
B(r M +M − ~ (6.76)
3 3
B(r ~ =B
~ ≥ R) = µo H ~ field of a magnetic dipole m
~
(6.77)
This matches our previous solution for the magnetic field of this system that we
obtained by calculating the vector potential of the bound surface current.
J~b (~ ~ r)
~ × M(~
r) = ∇ ~ b (~
K ~ r ) × nb(~
r ) = M(~ r)
~ r ) = µo J~b (~
r 0) µo ~ b (~
K r 0)
Z I
A(~ dτ 0 0
+ da0
4π V |~
r −~ r | 4 π S(V) |~
r −~ r 0|
We can, in fact, directly calculate the field from the bound currents using the
Biot-Savart Law. The approach described above of using the magnetostatic scalar
potential for such cases will in general be calculationally easier if the problem is
amenable to the techniques for solving Poisson’s Equation, but the Biot-Savart Law is
certainly always guaranteed to work.
Page 424
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
~ = −∇V
H ~ M (~
r) (6.78)
~ = B(
Again, if we know the relationship B ~ H),
~ then we can use the divergence
equation:
∇ ~ −∇V
~ ·B ~ M =0 (6.79)
~ and B
Again, if the relation between H ~ is not simple, the above equation may be
difficult to solve.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 425
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
In a region where µ is constant, it can be passed through the divergence and we can
reduce this to
∇2 V M = 0 (6.81)
The importance of boundary conditions should be even more clear in such cases: since
there is no source term in the equation, the boundary conditions entirely determine
the solution.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 426
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
This is now a “soft,” linear material, where we cannot take M~ to be fixed. But it is a
situation with no free currents, so Laplace’s Equation holds (except at the r = R
boundary, but we develop matching conditions there).
Fortunately, we do not need to solve the boundary value problem from scratch because
this problem is directly analogous the case of a dielectrically polarizable sphere in an
external electric field. We have the following correspondence:
~ = −o ∇V
o E ~ ~ = −∇V
H ~ M (6.82)
2 2
o ∇ V = 0 ∇ VM = 0 (6.83)
~ = − o E
P ~ ~ = µ − µo H
M ~ (6.84)
o µo
~ = o E
D ~ +P
~ ~ o =H
B/µ ~ +M ~ (6.85)
~ r−→
o E
→∞ ~0
o E ~ r−→
H
→∞ ~
B0 /µo (6.86)
~ r−→
D
→∞ ~0
o E ~ o r−→
B/µ
→∞ ~
B0 /µo (6.87)
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 427
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
Let’s also compare the matching conditions. We want to use the matching conditions
that incorporate only the free charge densities because we do not know the bound
charge densities ahead of time. For the electrostatic case, we used
h i
nb · D ~ > (R) − D~ < (R) = σf = 0 (6.88)
h i
sb · o E ~> (R) − o E
~< (R) = 0 (6.89)
Thus, not only is there a perfect correspondence between fields, potentials, and
r → ∞ boundary conditions in the two problems, there is also a correspondence
between matching conditions at r = R. Thus, we can just apply the solution to the
~ → H,
electrostatic problem with the substitutions o E ~ o V → VM , P ~ → M,~ and
~ → B/µ
D ~ o.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 428
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
3 µo 3 µo B0
VM (r < R) = − H0 z = − z (6.92)
2 µo + µ 2 µo + µ µo
m~ · rb B0 m~ · rb
VM (r > R) = −H0 z + =− z+ (6.93)
4 π r2 µo 4 π r2
4π 3 ~ 4π 3 3 (µ − µo ) 4 π 3 B0 3 (µ − µo )
m~ ≡ R M(r < R) = R H0 zb = R zb
3 3 2 µo + µ 3 µo 2 µo + µ
(6.94)
From the above, we calculate the fields and the magnetostatic surface charge density
(ρM = 0 because M~ is uniform):
3 µo ~0
B ~0
B ~ < R)
M(r
~ < R) =
H(r = − (6.95)
2 µo + µ µo µo 3
µ − µo B~ 0 µ − µo B0
~ < R) = 3
M(r σM = 3 cos θ (6.96)
2 µo + µ µo 2 µo + µ µo
" #
~ ~ < R)
~ < R) = µo B0 − M(r
h i
~ < R) = µo H(r
B(r ~ < R) + M(r ~ < R)
+ M(r
µo 3
2 3
~ 0 + µo M(r ~ < R) = µ ~0
=B B (6.97)
3 2 µo + µ
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 429
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
I Like E~, H~ is uniform inside the sphere and points in the direction of the uniform
field. For χm > 0, like for χe > 0, it is smaller in magnitude than the uniform
field at infinity.
I The magnetization density is in the direction of the uniform field for χm > 0 as
it was for P~ and χe > 0.
I The magnetostatic surface charge density has a cos θ dependence and is positive
at the north pole for χm > 0, as it was for the electrostatic surface charge
density and χe > 0.
I B~ is enhanced relative to the uniform field for χm > 0. We did not calculate D ~
in the electrostatic case, but we would have found that it, too, was enhanced
relative to the uniform field.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 430
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
There is a shortcut method that is much faster, so good to know from the point of
view of technique. It makes the ansatz that the sphere magnetizes uniformly so then
the total field is the superposition of a uniform field and a uniformly magnetized
sphere (Equation 6.22). This assumption is made initially without relating M ~ and H.
~
It then uses the relation M~ = χm H ~ (equivalently, B~ = µ H)
~ to relate the two and
solve for the fields.
~ < R) = B
B(r ~ uniform + B ~ 0 + 2 µo M
~ sphere = B ~ (6.98)
3
!
~ sphere
B
~ < R) = H
H(r ~ uniform + H
~ sphere = H
~ uniform + ~ sphere
−M
µo
~0
B 1 ~
= − M (6.99)
µo 3
~ < R) = µ H(r
Then we apply B(r ~ < R) to relate the above two equations and solve
~ One finds one gets the same result. One can then calculate the field at r ≥ R
for M.
from superposition. Admittedly, this technique is somewhat backhanded; when trying
to understand things for the first time, reapplying the scalar potential to the full
problem is more straightforward.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 431
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
There are no free currents, so we may use the magnetostatic scalar potential
technique. Furthemore, ∇ ~ ·H~ = 0 in each region since µ is constant in each region.
So the scalar potential VM satisfies Laplace’s Equation, allowing us to apply our
techniques for the solution of Laplace’s Equation from electrostatics.
In particular, given the azimuthal symmetry, we may assume the solution in each of
the three regions is of the form given in Equation 3.113:
∞
X
VM (r < a, θ) ≡ V1 (r , θ) = A` r ` P` (cos θ) (6.100)
`=0
∞
D`
X
VM (a < r < b, θ) ≡ V2 (r , θ) = C` r ` + P` (cos θ) (6.101)
`=0
r `+1
∞
X E`
VM (r > b, θ) ≡ V3 (r , θ) = −H0 r cos θ + P (cos θ)
`+1 `
(6.102)
`=0
r
where we have already applied the requirements that VM be finite as r → 0 and that
it yield the uniform field as r → ∞ with H0 = B0 /µo . We have also assumed that VM
has no constant offset as r → ∞.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 432
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
There are no free currents, so our matching conditions are (as for the magnetically
permeable sphere, Equations 6.90 and 6.91) that the normal component of B ~ and the
~ ~ ~
tangential component of H be continuous. Using H = −∇VM , we thus have the four
conditions
∂ V1 ∂ V2 ∂ V2 ∂ V3
µo = µ µ = µo (6.103)
∂r a ∂r a ∂r b ∂r b
∂ V1 ∂ V2 ∂ V2 ∂ V3
= = (6.104)
∂θ a ∂θ a ∂θ b ∂θ b
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 433
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
Before we dive into a lot of calculation, let’s see what we can figure out without doing
much work. The radial derivative equations only connect terms on the two sides of the
equations with the same ` because they do not modify the orthonormal P` (cos θ).
What about the angular derivative equations? Recall Equation 3.154:
dm
P`m (x) = (−1)m (1 − x 2 )m/2 P` (x) (6.105)
dx m
∂ P` (cos θ)
Let’s write ∂θ
using this:
where we note that, since 0 < θ < π, there is no sign ambiguity and thus
sin θ = (1 − cos2 θ)1/2 . The P`1 (cos θ) are also orthonormal polynomials (the P`m over
all ` at fixed m form an orthonormal set in order for the Y`m to form an orthonormal
set), so the same point we made above about the equations connecting terms at the
same ` holds for these equations also. Note however that, for ` = 0, the ∂/∂θ
matching condition yields zero.
Note also that, for ` ≥ 1, these equations are the same as one would have obtained by
requiring continuity of VM since ∂/∂θ doesn’t modify the radial factor of each term.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 434
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
Taking the necessary derivatives for the radial derivative equations and then equating
the two sides of all six equations (four for ` > 0, only two for ` = 0) term-by-term
gives us:
D`
`>0: µo ` A` a`−1 = µ ` C` a`−1 − µ (` + 1) (6.108)
a`+2
D` E`
µ ` C` b `−1 − µ (` + 1) `+2 = −µo H0 δ`1 − µo (` + 1) `+2 (6.109)
b b
D`
A` a` = C` a` + `+1 (6.110)
a
D` E`
C` b ` + `+1 = −H0 b δ`1 + `+1 (6.111)
b b
D0 D0 E0
`=0: 0 = −µ −µ = −µo 2 (6.112)
a2 b2 b
We explicitly write out the ` = 0 equations because they yield qualitatively different
conditions than the ` > 0 terms.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 435
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
For ` > 1, solving for C` and D` results in both vanishing, so then A` and E` vanish
for ` > 1.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 436
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
For ` = 1, we can do a lot of algebra to find explicit formulae for all the coefficients
(you can find these in Jackson §5.12). These formulae are not particularly
illuminating, but they become more intuitive when we take the limit µ/µo 1.
Inserting those coefficients into the solutions, we obtain
µ
µo
1 9
V1 (r , θ) = A1 r cos θ = − H0 r cos θ (6.113)
µ a3
2 µo
1− b3
µ
1 D1 3 1 a3
µo
V2 (r , θ) = C1 r + 2 cos θ = − H0 r+ cos θ (6.114)
2 r2
3
r µ
1− a
µo b3
a3
µ
µo
1
E1
b 3 3 1 + 12 b3
V3 (r , θ) = −H0 r + 2 cos θ = H0 −r + 2 1 − cos θ
r r µ a3
µ
1− b3
o
(6.115)
Note that we include the term of order µo /µ in the r > b solution so we can see that
the matching condition on the tangential derivative at r = b (equivalent to matching
of VM itself) is explicitly satisfied even in the limit µ/µo 1.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 437
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
µ
µo
1 9 H0 zb
~ 1 (r , θ)
H = − ~ 1 (r , θ) = µo H
B ~ 1 (r , θ) (6.116)
3
2 µ
1 − ba3
µo
µ
µo
1 3 H0 zb ~ a · rb) rb − m
3 (m ~a
~ 2 (r , θ)
H = + ~ 2 (r , θ) = µ H
B ~ 2 (r , θ) (6.117)
4 π r3
3
µ
µo
1 − ba3
µ
µo
1 ~ b · rb) rb − m
3 (m ~b
~ 3 (r , θ)
H = H0 zb + ~ 3 (r , θ) = µo H
B ~ 3 (r , θ) (6.118)
4 π r3
9 H0 4π 3
with ~a = −
m 3
a zb (6.119)
2 µ
1− a 3
µo b3
a3
3 1 + 21 b3
4π 3
~ b = 3 1 −
m 3
H0 b zb (6.120)
µ
1− a 3
µ o b3
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 438
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 440
Lecture 26:
Magnetostatics in Matter V:
Boundary Value Problems in Magnetostatics (cont.)
Electrodynamics I:
Currents and Ohm’s Law
Date Revised: 2022/03/28 05:00
Revised lecture break
Date Given: 2022/03/07
Page 441
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
Let’s now consider the analogy to electrostatic shielding. Electrostatic shielding is
easily provided by conductors, and perfect conductors (/o → ∞) provide perfect
electrostatic shielding. They do this by setting up surface charge that perfectly cancels
the externally applied field.
The magnetostatic shielding effect is very similar, though it occurs for completely
different reasons. In this case, the high magnetic permeability of the materials causes
magnetic dipoles to be set up to almost perfectly cancel the externally applied
magnetic field (a residual field of order µo /µ times the externally applied field reaches
the interior). This occurs because the oriented dipoles yield a large bound surface
current whose magnetic field cancels the externally applied magnetic field. If one had
the equivalent of an electric conductor, with µ/µo → ∞, the shielding would be
perfect, as for an electric conductor. In such a material, H → 0 and M ~ = B/µ
~ o , just
as in a perfect conductor one has E → 0 (and P ~ = D).
~
That said, one could calculate the magnetostatic surface charge density σM from the
discontinuity in M~ and one would see that σM would look very much like σb for the
case of a spherical shell of high dielectric susceptibility (/o 1) and, in the limit
µ/µo → ∞, σM would mimic the surface charge density of the electrical conductor
limit (which is the same as /o → ∞).
Either way one does it, this calculation has important practical implications: such
highly permeable materials are in widespread use for magnetic shielding from,
especially, Earth’s field in magnetically sensitive experiments and equipment such as
SQUIDs (very sensitive magnetometers) and photomultiplier tubes (where the
electrons’ paths can be substantially bent and thus the gain modified by magnetic
fields).
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 442
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
While we have benefited from our boundary-value problem techniques to get directly
to the fields without having to calculate the bound surface currents, it would be nice
to see how the bound surface currents give the observed fields. Recall Equation 5.75,
which gives the bound surface current density from the change in the tangential
component of the magnetic field:
~ (~ 1 h
~ > (~ ~ < (~
i
K r) = r) × B
nb(~ r) − B r) (6.121)
µo
where nb points from the < region to the > region. In our case, K ~ =K~ b because there
~ only has
are no free currents. Since nb = rb for our spherical surfaces and B
components in the rb and θb directions, this reduces to
" #
b = 1 B>,θ − B<,θ φ b= 1 µ> ∂ VM µ< ∂ VM
Kb (r ) φ − + b (6.122)
φ
µo µo r ∂θ r> r ∂θ r<
µ< − µ> 1 ∂ VM b
= φ (6.123)
µo r ∂θ r
where < and > indicate the two sides of the particular boundary at r and we use the
fact that the tangential component of H,~ which is given by −(1/r ) ∂VM /∂θ here, is
continuous and thus has the same value on both sides of the interface at r . So it is
straightforward to calculate the surface currents given VM .
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 443
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
We also know how to calculate B ~ given surface currents derived from a uniform
magnetization: we did it in our first calculation of the magnetic field of the
permanently magnetized sphere (Equation 6.22) and saw (valid only for K ~b ∝ φ
b sin θ!)
~ b
~ ~ (r < R) = 2 µo M
B ~ = 2 µo Kb · φ zb
M 3 3 sin θ
· − ~ b
~ ~ (r > R) =
B
µ o 3 ( ~
m r
b ) r
b ~
m 4 ~ = 4 π R 3 Kb · φ zb
~ = π R3 M
m
M 4π r3 3 3 sin θ
where the relation between M ~ and K ~ b comes from the definition of the bound surface
~b = M
current, K ~ × nb = M zb × rb = φ
b M sin θ. This applies here because the directions
of the magnetizations and surface currents are the same as we have here. (The fact
that we have permeable materials present is irrelevant for the calculation of B:~ once
one has all the bound currents, one can calculate B ~ directly from them.) So, we
expect that, in this case, we can just add the field of the above form due to the bound
currents to the uniform applied field to get the total field in the three regions.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 444
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
~a , K
That is, we expect (again, valid only for K ~b ∝ φ
b sin θ!)
~ b ~
~ 0 + 2 µo Kb (a) · φ + Kb (b) · φ zb
~ 1 (r , θ) = B
b
B (6.124)
3 sin θ
2 ~ b (b) · φ
K ~ a · rb) rb − m
µo 3 (m ~a
~ ~
b
B2 (r , θ) = B0 + µo zb + (6.125)
3 sin θ 4π r3
~ 0 + µo 3 (m
~ 3 (r , θ) = B ~ b · rb) rb − m~b
B (6.126)
4π r3
4 K~ b (a) · φ
b 4 ~ b (b) · φ
K b
m~ a = π a3 zb ~b = m
m ~ a + π b3 zb (6.127)
3 sin θ 3 sin θ
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 445
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
µ − µo 1 ∂ VM
µ
BK ∼ µo K ∼ ±µo ∼ µo − 1 Hθ (6.128)
µo r ∂θ µo
1 0 ! 1
µ µo µo
∼ µo O +O O H0 (6.129)
µo µ µ
" 1 #
µo 0 µo
∼ O +O B0 (6.130)
µ µ
(In the second line, we used H ∼ O(µo /µ)1 H0 , which one can see from the
expressions for H~ 1 and H
~ 2 . It is not so obvious that this is true for H
~ 3 at r ∼ b, but it
must be true because Hθ is continuous. It turns out to be true because the dipole
field cancels the applied field to first order in H0 (i.e., zeroth order in µo /µ) at the
equator, leaving a residual field of order O(µo /µ)1 H0 . The cancellation does not
happen at the poles, but Hθ = 0 at the poles.)
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 446
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
We now see the problem. The term that is O(µo /µ)0 will cancel the B ~ 0 term. So then
the O(µo /µ)1 B0 term is all that is left and is our full field, as expected. But we have
not done the approximation self-consistently. We would have obtained a term of the
same order by including terms O(µo /µ)2 in the expression for H because they would
yield O(µo /µ)1 terms when multiplied by the O(µ/µo )1 term from the (µ/µo − 1)
prefactor. Without including that term, we will get the incorrect coefficient for the
residual field.
We could have included that higher order term, but then we would run into the same
problem at the next order: our calculation of the field using the surface currents would
be correct to O(µo /µ)1 , but our expression for the fields would have terms of order
O(µo /µ)2 that we would not be able to fully reproduce. Given that it would be
algebraically challenging to do this even to O(µo /µ)1 correctly, we punt on trying to
calculate the residual field.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 447
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
O(µo /µ)0
≈ b 9
−φ
B0
sin θ (6.131)
a3 µo
2 1− b3
a3
3 1+ 1
− 1 V B0
~ b (b, θ) = µ < µ> ∂ M µ 2 b3
K φ b=φ b −1 sin θ
µo b ∂θ b µo µ a3 µo
µo
1 − b3
3
1 a
O(µo /µ)0 3 1 + 2 b3 B
0
≈ φ
b
3
sin θ (6.132)
1− a µo
b3
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 448
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
O(µo /µ)0 2 B0 9 1 a3
~ 1 (r , θ)
zb · B ≈ B0 + − +3 1+ =0 (6.133)
3 1 − a33 2 2 b3
b
~ 1 (r , θ)
B O(µo /µ)0
~ 1 (r , θ)
zb · H = ≈ 0 (6.134)
µo
As expected, both the magnetic and auxiliary fields vanish to zeroth order in µo /µ
inside the cavity.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 449
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
The total magnetic field in the permeable material is of order B0 because both terms
shown are of order B0 . In the limit a b, one recovers 3 B0 as we expect from the
case of the permeable sphere (Equation 6.97 with µo /µ → 0). The auxiliary field
vanishes in the permeable material to order (µo /µ)0 because one must divide the
entire expression by µ to get H from B, which combines with the µo in the expression
for B to give a prefactor of µo /µ that vanishes at the level of approximation we are
considering.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 450
Section 6.4 Magnetostatics in Matter: Boundary Value Problems in Magnetostatics
So, in the end, we see that, to the level of approximation for which we can
self-consistently do calculations, the fields we calculate from the surface currents
match the fields that we used to calculate those surface currents.
Section 6.4.5 No Free Currents, Linear Materials via Scalar Potential Page 451
Section 7
Electrodynamics
7.1 Currents and Ohm’s Law
7.2 Motional Electromotive Force
7.3 Electromagnetic Induction
7.4 Inductance
7.5 Magnetic Energy and Forces
7.6 Maxwell’s Equations
Page 452
Section 7.1 Electrodynamics: Currents and Ohm’s Law
J~ ∝ f~ (7.1)
I The first is the random thermal motion of the charge carriers. The forces we
can apply yield velocities that are small perturbations to this random thermal
motion. So the mean speed of the carriers is dominated by the thermal speed
vthermal .
I The second is scattering. This scattering is in fact the cause of the randomness
of the thermal motion. The charge carriers scatter off of impurities and defects
in the material and off of the thermal vibrations present in the material. This
scattering is elastic in general, resulting in no loss of energy but in a redirection
of velocity.
Section 7.1.1 Ohm’s Law and Joule Heating: Differential Version Page 453
Section 7.1 Electrodynamics: Currents and Ohm’s Law
In the presence of such effects, our picture should not be of a charge carrier smoothly
accelerating under the influence of an external force, but rather of a carrier with a
large randomly directed velocity, scattering frequently, and with acceleration by the
force between scatters resulting in a net motion in the direction of the electric force.
The scattering randomly redirects the velocity, so the velocity due to the externally
applied force is, on average, reset to zero after each collision. If the thermal speed is
vthermal and the typical distance traveled between scatters is λ, then the time available
for the externally applied force to accelerate a carrier between scatters is
λ
t= (7.2)
vthermal
The average velocity acquired from the applied force during this time is
1 1 f~ λ
~
vave = ~
at = (7.3)
2 2 m vthermal
This velocity is the average overall velocity because of the zeroing of the
instantaneous velocity after each collision.
Section 7.1.1 Ohm’s Law and Joule Heating: Differential Version Page 454
Section 7.1 Electrodynamics: Currents and Ohm’s Law
If we then use J~ = n q ~
vave where n is the number density of charge carriers and q is
the charge per carrier, and we use f~ = q E~ , we then can write
n q2 λ n q2 λ
J~ = ~
E =⇒ J~ = σ E
~ σ= (7.4)
2 m vthermal 2 m vthermal
There is power dissipated in this process — the work done on the charge carriers by
the electric field is lost to random motion when they scatter. The inifinitesimal
amount of energy lost per unit time dP in an infinitesimal volume dτ is equal to the
work done by the electric field on the charge carriers:
force J~
dP = number density · velocity · vave · f~ dτ = n
dτ = n ~ ~ dτ = J~ · E
· qE ~ dτ
carrier nq
(7.5)
We note that the possibility of E~ 6= 0 does not contradict our earlier discussions of
conductors in electrostatics; here, we have non-stationary charges, where in that case
we considered the final static situation after any currents had flowed.
Section 7.1.1 Ohm’s Law and Joule Heating: Differential Version Page 455
Section 7.1 Electrodynamics: Currents and Ohm’s Law
Let’s assume the cross-sectional area of the conductor is constant and the conductor is
~ . If we then do
uniform. This lets us do the area integral trivially, yielding I = σ A nb · E
a line integral directed along the wire, such that d ~` ∝ nb, we have
Z Z Z
I` = d` I = σ A ~ = σA
d` nb · E d~`·E~ =⇒ I ` = σA V (7.7)
` 1 ` 1
=⇒ V = IR with R= ≡ ρ with ρ= (7.8)
A σ A σ
which is the familiar version of Ohm’s law in terms of current, voltage, and resistance.
This is the integral version of Ohm’s Law while J~ = σ E ~ is the differential (or local)
version. We also define the resistivity ρ as the reciprocal of the conductivity σ. We
can also integrate the Joule heating expression to get the usual integral expression for
Joule heating:
2
~ = I V = I 2R = V
Z Z Z Z
I
P= dP = dτ J~ · E
~= da d` nb · E (7.9)
V V S A R
Section 7.1.2 Integral Version of Ohm’s Law and Joule Heating Page 456
Lecture 27:
Electrodynamics II:
Vanishing Charge Density with Steady-State Currents
Electric Field in a Wire
Electromotive Force
Date Revised: 2022/03/28 05:00
Date Given: 2022/03/28
Page 457
Section 7.1 Electrodynamics: Currents and Ohm’s Law
~= 1∇
~ ·E
∇ ~ · J~ = 0 (7.10)
σ
where the first step was possible by Ohm’s Law and the assumed uniformity of the
conductivity and the second step by the steady-state assumption on (macroscopic)
scales. So the answer is no, as long as the conductivity is uniform and the system is
steady-state, no charge density accumulates. Note that this is not a circular argument:
the steady-state assumption corresponded to ∂ρ/∂t = 0, not ρ = 0. Now, with the
combination of ∂ρ/∂t = 0 and Ohm’s Law, we conclude ρ = 0.
Later, we will see how it is possible for charge to accumulate when we consider
non-steady-state systems (in particular, with sinusoidal currents).
Note that our microscopic picture is not consistent with the steady-state assumption,
but, averaged over time, our macroscopic picture is.
Section 7.1.3 Steady-State Assumption and Uniform Conductivity =⇒ Zero Charge Density Page 458
Section 7.1 Electrodynamics: Currents and Ohm’s Law
We implicitly assumed in proving the integral version of Ohm’s Law above that the
uniformity of the conductor implied that the field and thus the current were uniform
over the cross-sectional area. We can prove this. We did not explicitly require that the
electric field also be uniform with position along the wire, but we can prove that, too.
We proved above that the charge density vanishes in a uniform conductor with steady
currents. Therefore, the conductor satisfies Laplace’s Equation. Dirichlet boundary
conditions are set at the two ends of the conductor by the potential difference ∆V .
We assume these equipotentials are (by connections to a battery) transverse to the
wire axis at z = 0 and z = `. On the outer surface of the wire, J~ · nb = 0 because no
current flows out of the wire, which implies that E~ · nb = 0, which provides a boundary
condition on the normal gradient of the potential (a Neumann boundary condition).
(Equivalently, this implies the charge density vanishes at the surface.)
∆V ~ = − ∆V zb
~ = −∇V
V (~
r) = z =⇒ E (7.11)
` `
Note that we do not need the sinusoidal solutions from separation of variables here —
we only need the linear solution (which we ignored in our discussion of separation of
variables in linear coordinates). This will be of relevance for the homework, too!
Therefore, it is valid to assume that the field is uniform over the cross-sectional area
of the wire and along the length of the wire if the wire is of fixed cross-sectional area,
the conductivity is uniform, and the currents are steady-state. The latter two
conditions told us Laplace’s Equation is satisfied, while the former one provided the
z-translation symmetry needed to guess the solution.
What happens to this argument if the wire changes in some way along its length; e.g.,
the conductivity changes, or the wire diameter changes?
~ N
B
Let’s consider the force on a charge carrier in the portion of the wire that intersects the
field. If ~ ~ = −B zb (into the page), then these charges feel a Lorentz force
v = v xb and B
Since this force is aligned with the vertical portion of the wire, the carriers in that
section can move. Assuming for the moment the charge carriers are positive (the
argument can be reversed if they are negative), they would start to collect at b at the
top end of the vertical portion and a deficit would appear at a at the bottom end. The
local electrostatic repulsion between like charges would cause the charge carriers to
start flowing through the rest of the circuit and would prevent this clumping of
carriers. In this way, a current is generated around the loop without the influence of a
large-scale electric field in the circuit. If the loop is pulled at constant speed, one
satisfies the steady-state assumption, with no charge buildup.
There is work done on a charge carrier during its movement up the vertical portion of
the wire
Z b
Wab = d~
` · F~mag = q v B h (7.13)
a
We will see below that this work is not done by the Lorentz force as suggested above
(recall, the Lorentz force can do no work because F~mag ⊥ ~ v ), but it is nevertheless
done. The energy gained by the charge carriers via this work is dissipated as Joule
heating in the resistor because the carriers quickly reach some steady-state velocity
and a steady-state current flows.
We define the work done per unit charge on the charges as they move from a to b as
the motional electromotive force or motional emf:
Wab
E= =vBh (7.14)
q
Let’s think about how we can interpret E. Let I be the current that flows. I is q times
the number of charges that flow per unit time past a given point. Therefore I E is the
power being supplied to the ensemble of charges, the work done on them per unit
time. By conservation of energy, it is also the power being dissipated in Joule heating
in R. But we know that latter quantity is also I2 R. Equating the two, we see
E = IR (7.15)
That is, E plays the role of voltage in Ohm’s Law for the resistor. E has the right units
for this purpose. In fact, if one attaches a voltmeter across the resistor R, it will
report a voltage V = E: a voltmeter works essentially by measuring the current in a
very large resistor R 0 R placed in parallel with R, and the current that will flow
through R 0 is identical to what would flow if a battery E were placed across R with R 0
in parallel. So, what appeared to just be a work done on a unit charge now can be
interpreted as equivalent to a voltage! But be sure to remember that the current is
generated by movement of the circuit in a magnetic field; it is not due to an electric
field! We will return to the distinction between E and a voltage later when we consider
electromagnetic induction and Faraday’s Law.
Let’s now think about what force is doing the work. As we discussed some time ago,
the Lorentz force does no work because F~ ⊥ ~ v . However, a force must pull the loop.
There is a force counteracting this force that the pulling force must match to keep the
loop at constant speed: the Lorentz force due to the velocity the carriers have
~ = u yb. This force is
acquired in the yb direction, which we will denote by u
−F~pull = F~mag
0
= q u yb × −B zb = −q u B xb (7.16)
~ =~
w ~ = v xb + u yb
v +u (7.17)
force is
dWpull
= F~pull · w
~ = q u B xb · (v xb + u yb) = q u B v (7.18)
dt
Note that the charge carriers move on a diagonal line relative to the lab frame as they
move from point a to point b on the wire, with this line partly in the direction of F~pull .
It takes the charge carriers a time t = h/u to move on this trajectory since their yb
direction speed is u. Therefore, the work done by F~pull during the movement of a
charge from a to b is:
dWpull h
Wpull = = qBv h (7.19)
dt u
Wpull
=⇒ =Bvh=E (7.20)
q
That is, the work done by the pulling force, per unit charge, matches the motional
emf. The pulling force provides the energy that is eventually dissipated as heat as the
carriers flow through the resistor.
Mechanically, how does this work? A magnetic field does no work, so it should only
change the direction of the velocity of the charge carriers. So, initially, when the
pulling force begins to act and the carriers start to move in the x direction and feel a
Lorentz force in the y direction, their x velocity starts to be transformed into y
velocity. But the loop is being pulled at constant speed v xb, so the walls of the wire
exert a force so the carriers’ x velocity remains equal to v xb as the magnetic force
acts. Similarly, as the carriers acquire a velocity in the y direction, they feel F~mag
0 in
the −b x direction, and the walls of the wire must exert a force to keep them moving at
v xb in the x direction. By Newton’s third law, the charge carriers exert a reaction force
on the walls of the wire, which would slow down the loop if there were not a force
pulling it. Thus, we see it is the force pulling the loop that ultimately provides the
work to drive the current.
And note: All this motion is accomplished without a large-scale electric field. Of
course, it relies on the microscopic Coulomb repulsion between like charge carriers and
the Coulomb binding to the wire that keeps the charge carriers from flying out of the
wire.
Returning to the emf itself, we can rewrite it in a useful form. We define the magnetic
~ dotted into the normal to a surface over the surface:
flux to be the integral of B
Z
Φ= ~ r)
da nb · B(~ (7.21)
S
Φ = B hx (7.22)
dΦ dx
=Bh = −B h v (7.23)
dt dt
(x decreases with time for v > 0) which is just the negative of the motional emf. That
is, we have
dΦ
E=− (7.24)
dt
We can see, through some work, that the motion ~ v also describes the change in the
area of and flux through the loop. The flux changes by
Z Z
dΦ = ~−
da nbC · B ~
da nbC · B (7.25)
S(C(t+dt)) S(C(t))
where we have used the fact that nbS = −b nC on the S(C(t + dt)) surface and nbS = nbC
on the S(C(t)) surface. (The direction of nbC is set by the direction of d ~
` in the figure
and the right-hand rule.) The negative sign is present in the former because the
orientation of nbC that maintains its direction as the contour moves has nbC (t + dt) on
this surface pointing into the enclosed volume rather than outward.
The area element on the ribbon (with outwardly directed normal as in the above
integral) is given by
r × d~
nbS da = d~ ` (7.28)
where: d ~` is the line element along C(t) with orientation set by consistency with nbC
for S(C(t)) and the right-hand rule; and d~ r is the change in the vector position of
that line element between t and t + dt. The difference between these two positions is
related to ~ r =~
v , d~ v dt, so:
v dt × d ~
nbS da = ~ ` (7.29)
I
Therefore, dΦ = v dt × d ~
~ ~
` ·B (7.30)
C(t)
We turned an area integral into a line integral, but it still calculates magnetic flux.
~ k d~
Since u ~ to ~
`, we can add u ~ without affecting the integral:
v to obtain w
I
dΦ = ~ dt × d ~
w ~
` ·B (7.31)
C(t)
Using the cyclic property of the triple vector product, reversing the resulting cross
~ ×w
product B ~ , and moving dt to the left side, we obtain
dΦ
I
=− d~
`· w ~
~ ×B (7.32)
dt C(t)
dΦ F~mag
I
=− d~
`· (7.33)
dt C(t) q
The integral of the Lorentz force per unit charge integrated around the loop is the
generalization for arbitrary loops of our earlier expression for the motional emf for the
rectangular loop (earlier, we integrated over only the section of length h from a to b
of the rectangular loop for which F~mag was nonzero), so
F~mag dΦ
I
E= d~
`· =− moving circuit (7.34)
C(t) q dt
The motional emf, as defined by the line integral of the Lorentz force per unit charge
around the loop, is given by the negative of the rate of change of the magnetic flux
through the loop. The signs of the line integral and the flux are set by requiring that
the orientation of the line integral (via d ~
`) be consistent via the right-hand rule with
the orientation of the surface normal nbC used for the flux calculation.
Page 474
Section 7.2 Electrodynamics: Motional Electromotive Force
~ · nb(t) = A B cos ωt
Φ(t) = a2 B (7.35)
dΦ
E(t) = − = A B ω sin ωt (7.36)
dt
It is instructive to think again about the Lorentz force experienced by the charge
carriers in the loop and see how it generates the motional emf. Let the magnetic field
be B ~ = B zb and let the axis of rotation be +b y . Suppose the loop is just moving past
having nb = −b x , as shown in the figure. Then the carriers all have a velocity parallel to
±bx due to the motion of the loop (this is ~ v ). (They also have motion in the zb
direction, but this is parallel to B~ and thus no Lorentz force is generated.) The
carriers in the sections of the loop parallel to zb (perpendicular to the axis of rotation,
parallel to the field) cannot move in response to this force because they feel a force in
the yb direction, transverse to the section of wire they are in. Those in the parts of the
loop parallel to ±b y (parallel to axis of rotation) also feel a force along yb, and they can
move along yb. As the loop turns away from this orientation, the arm at +b z a/2 has
velocity in the +b x direction and vice versa for the arm at −b z a/2. Positive charge
carriers in these arms feel forces in the −b y and +b y directions, respectively. This forces
a current to flow in direction defined by the −b n = +bx orientation by right-hand rule,
generating a field through the loop in the −b n = +b x direction.
As the loop passes through this orientation, the flux is zero and is changing from
negative (bn·B~ < 0) to positive (bn·B ~ > 0). One can see that the driven current is in
the direction needed for its field to counter the change in magnetic flux. This is a
manifestation of Lenz’s Law, which we will return to later.
If one taps the loop as is typical for such a generator, as shown in the figure, the tap
connected to the +b z a/2 arm will have positive voltage and the tap connected to the
−b z a/2 arm will have negative voltage because they need to drive a current in an
external circuit that carries current in the direction consistent with that argued above,
from the +b z a/2 arm to the −b z a/2 arm.
Note the polarity of the above statement: we decide the sign of the voltage at the
taps not by what is needed to drive the current in the loop (which is driven by the
Lorentz force, not by this voltage) but rather by the sign needed to drive the current
in the external load (the resistor) so that current exits the loop, goes through the
load, and returns to the loop where it is needed to conserve charge.
Electromagnetic Induction
Faraday’s Law
I Moving loops: As we considered above, the magnetic field is stationary but the
loop is moving.
I Moving magnetic fields: The loop is held fixed but the magnetic field is
changing because the currents sourcing the field are being translated.
I Changing magnetic fields: Both the loop and the sources of the field are
stationary, but the currents sourcing the field are changing.
We just proved using the Lorentz Force Law that the first situation results in a
motional emf: a force that causes the flow of a current around the loop, given by
Equation 7.34:
F~mag dΦ
I
E= d~
`· =− moving circuit (7.37)
C(t) q dt
Faraday’s Law consists of the empirical observation that the same rule applies for the
second and third situations. The subtlety is this: this law could not have been derived
using the Lorentz Force applied to the situation described above of a fixed loop and a
moving and/or changing magnetic field: there is no magnetic force if the charge
carriers are not moving. A natural and important corollary is that the emf that
appears is not due to a magnetic force. Rather, since the loop is at rest in the second
and third situations, the force that appears arises from a true electric field.
Mathematically, we write Faraday’s Law as
F~elec dΦ
I
E= d~
`· =− moving or changing magnetic field (7.38)
C(t) q dt
We see that it is identical in form to the Lorentz Force law applied to a moving loop
with the replacement of F~mag by F~elec .
Combining the two forms, and defining E ~ind = F~elec /q where ind indicates that the
electric field here is not an electrostatic one due to Coulomb’s Law but rather an
“induced” field due to the changing magnetic flux, we then may write a common law
that applies in any situation:
" #
~
~ind + Fmag = − dΦ = − d
I Z
E= d~
`· E ~ r , t)
r , t) · B(~
da nb(~ (7.39)
C(t) q dt dt S(C(t))
If there is any ambiguity in the sign, one should apply Lenz’s Law: the emf has a sign
such that the polarity of the current it would drive produces a magnetic field that
counters the change in magnetic field. We will prove Lenz’s Law explicitly later.
Quasistatic Assumption
Note that we have implicitly assumed in our derivations that the current everywhere in
the loop responds instantaneously to the total emf on the left side, that there is no
time delay between a buildup of charge at one point in the circuit and the driving of a
current around the loop. We made the same assumption in deriving Ohm’s Law. This
is the “quasistatic assumption,” that all fields and currents everywhere change
instantaneously and that information is propagated infinitely quickly. Formally, this
assumption consists of saying that, given a typical physical scale for a system ` and a
typical timescale for variation t, we have
t `/c (7.40)
We will release this assumption when we discuss electromagnetic waves and radiation.
Consider the problem of the magnetic field moving at fixed velocity. One could go to
the rest frame of the magnetic field and consider the loop to be moving at fixed
velocity as in our moving loop cases. The magnetic force implied by the motional emf
appears. In Galilean relativity, forces are invariant upon change of inertial (fixed
velocity) frame. This would imply that the magnetic force in the field-fixed frame is
still present in the loop-fixed frame, but now we interpret it as an electric force
because the loop is not moving.
In the case of changing magnetic fields, we simply have to invoke the expectation that
the loop has no way of knowing whether it experiences a changing field because the
current sourcing the field is moving or because it is changing: it only knows about the
field that results, not the source of the field.
This Galilean relativity argument was, however, not recognized until after Faraday’s
observation.
Section 7.3.3 Motional EMF, Faraday’s Law, Galilean Relativity, and Galilean Field Transformations Page 482
Section 7.3 Electrodynamics: Electromagnetic Induction
We can make use of this argument to understand how electric and magnetic fields mix
with each other under such Galilean (nonrelativistic) transformations. Let’s assume we
have written down our law, Equation 7.39, in both the rest frame of the loop and in
the lab frame in which the loop is moving. The fields and position vectors in the loop
rest frame are given 0 symbols, the ones in the lab frame have no primes. The total
emf can be determined explicitly using a voltmeter to measure the voltage across the
resistor in the loop, and it is a scalar that is independent of frame (the reading on the
voltmeter doesn’t change if you see the voltmeter moving with the loop!). So we can
equate the lab and rest frame expressions through E:
" #
~
~ind + Fmag
I I
d~ ~0 =
`0 · Eind d~
`· E (7.41)
C0 C(t) q
(C 0 = C(t = 0) can be assumed by appropriate choice of when the lab and loop rest
frame coordinate systems coincide). Now, let’s use our expression for the magnetic
~ contribution that we
force term from our derivation of Equation 7.34, dropping the u
had added in:
I I h i
d~ ~0 =
`0 · Eind d~
`· E ~ind + ~ ~
v ×B (7.42)
C0 C(t)
Section 7.3.3 Motional EMF, Faraday’s Law, Galilean Relativity, and Galilean Field Transformations Page 483
Section 7.3 Electrodynamics: Electromagnetic Induction
~0 =E
E ~ind + ~ ~
v ×B (7.43)
ind
~0 =E
E ~ +~ ~
v ×B ~ =E
E ~Coul + E
~ind (7.44)
Therefore, this is a rule for how electric fields transform from one frame to another
under Galilean relativity, regardless of the source of the field. Electric fields are not the
same in a fixed and a moving frame if magnetic fields are present, even before special
relativity is considered! Special relativity then only adds correction coefficients to the
above equation.
It is important to note that the expectation that the electrostatic fields do not depend
on frame has been an assertion so far, based on the assumption that Coulomb’s Law is
unaffected by whether the charges are moving or not. We will return to this point
later in connection to Maxwell’s Equations, as it will lead to a symmetrization of the
above equation between E ~ and B.
~
Galilean relativity is consistent with the quasistatic assumption. We need only consider
special relativity when the nonzero travel time of light becomes important because
special relativity says the speed of light is the same in all frames.
Section 7.3.3 Motional EMF, Faraday’s Law, Galilean Relativity, and Galilean Field Transformations Page 484
Section 7.3 Electrodynamics: Electromagnetic Induction
Recall the previous example of an AC generator that used a rotating square loop in a
constant magnetic field. Instead, hold the loop fixed but assume that the magnetic
~
field is being varied sinusoidally, B(t) ~ 0 cos ωt. Then the flux is
=B
~
Φ(t) = A B(t) · nb = A B0 cos ωt (7.45)
dΦ
E(t) = − = A B0 ω sin ωt (7.46)
dt
just as before.
Note, again, the polarity of the emf! As before, the emf’s polarity is such that it
causes current to flow in an external resistor attached to the two ends of the circuit in
a direction consistent with the current that flows in the loop.
Section 7.3.3 Motional EMF, Faraday’s Law, Galilean Relativity, and Galilean Field Transformations Page 485
Section 7.3 Electrodynamics: Electromagnetic Induction
Something one has to be careful about is incorrectly believing that, because of the
emf’s sign, it should also drive a current in the zero-resistance loop in the direction
implied by the emf. That erroneous belief arises because one is assuming the electric
~ around a loop vanishes. No: that sign of
field is conservative, that the integral of E
emf would drive current in the wrong direction! For the current flowing in the loop,
the emf measures the work per unit charge done by E ~ind + F~mag /q as they push the
current around the loop, but they are not pushing the charges down an electrostatic
potential! The effect of having this current flow is that the same current flows
through the resistor, creating an apparent potential drop across the resistor that we
can measure with a voltmeter. But the voltmeter is just measuring the current flowing
through a known resistance, which, by Ohm’s Law, is proportional to the line integral
of the electric field through the resistor. The voltmeter’s ability to measure something
that looks like a voltage does not imply that an electrostatic potential can be defined
everywhere in the loop and resistor!
In thinking about what causes the current to flow, it is better to visualize the electric
field: one recognizes that the changing magnetic field generates an electric field that
pushes current in the direction it needs to flow to counter the change in magnetic
field. This electric field has nonzero loop integral around the circuit! Therefore, the
existence of the emf E at the ends of the circuit does not imply the same emf is
experienced by the current flowing in the loop itelf; the nonzero loop integral of the
electric field invalidates the rule that the total voltage drop around a loop must
vanish, which is the source of the misconception that E, appearing at the ends of the
circuit, is also the driver of the current in the loop.
Section 7.3.3 Motional EMF, Faraday’s Law, Galilean Relativity, and Galilean Field Transformations Page 486
Section 7.3 Electrodynamics: Electromagnetic Induction
~ =−d
I Z
d~
`·E ~ r , t)
r ) · B(~
da nb(~ (7.47)
C dt S(C)
Let’s use Stokes’ Theorem on the left side, and, since the contour is time-independent,
we can move the time derivative inside the integral on the right side. We turn it into a
partial derivative to make it clear that we do not need to worry about any possible
time-dependence of ~ r (of which there is none here). This yields
I h i Z ~ r , t)
∂ B(~
r) · ∇
da nb(~ ~ (~
~ ×E r) = − r) ·
da nb(~ (7.48)
S(C) S(C) ∂t
~ r , t)
∂ B(~
∇ ~ (~
~ ×E r) = − (7.49)
∂t
~ = 0 for
~ ×E
This differential version of Faraday’s Law is the generalization of ∇
time-dependent situations. We now explicitly see what was said in the previous
example: a changing B ~ creates a nonconservative electric field!
Biot-Savart and Ampere’s Law for the Induced Electric Field in the Absence of
Charges
If we consider the special case of no charge density, then we have
~
∇ ~ =0
~ ·E ~ = −∂B
~ ×E
∇ (7.50)
∂t
∇ ~ =0
~ ·B ~ = µo J~
~ ×B
∇ (7.51)
In magnetostatics, we saw that the above two equations, combined with the
assumption ∇ ~ = 0, yielded Poisson’s Equation for A
~ ·A ~ with µo J~ as the source
(Equation 5.56). By correspondence, we can thus state
~
∂B
~ =∇
E ~E
~ ×A ~E =
∇2 A ~E = 0
~ ·A
∇ (7.52)
∂t
Section 7.3.5 Biot-Savart and Ampere’s Law for the Induced Electric Field in the Absence of Charges Page 488
Section 7.3 Electrodynamics: Electromagnetic Induction
~ r 0)
∂ B(~
1
Z
~E (~
A r) = − dτ 0 ∂t
(7.53)
4π V |~
r −~ r 0|
Finally, we may take the curl of the above expression to recover the analogue of the
Biot-Savart Law. We did this backwards in the case of magnetostatics: we started
with the empirical Biot-Savart Law and derived that the field could be written as the
curl of the form of the vector potential corresponding to the above. Nevertheless, that
proof could be reversed, so we may conclude that the analogous Biot-Savart Law
holds (compare to Equation 5.32)
~ r 0)
∂ B(~
1 × (~ r 0)
r −~ 1 ∂ ~ r 0 ) × (~
B(~ r −~ r 0)
Z Z
~ (~
E r) = − dτ 0 ∂t
=− dτ 0
4π V |~ r 0 |3
r −~ 4 π ∂t V |~
r −~ 0
r | 3
(7.54)
b where we pulled the time derivative outside the integral under the assumption that
the volume itself is time-independent.
Section 7.3.5 Biot-Savart and Ampere’s Law for the Induced Electric Field in the Absence of Charges Page 489
Section 7.3 Electrodynamics: Electromagnetic Induction
Caution: We have made the quasistatic assumption, that all time derivatives are small
enough that the propagation time for disturbances in the magnetic fields is much less
than the timescales on which the field vary. This is what allows us to use the
magnetostatic formulae in time-varying situations. If the time derivatives become
large, then one needs the full formalism of electromagnetic waves, which we will
develop later.
Section 7.3.5 Biot-Savart and Ampere’s Law for the Induced Electric Field in the Absence of Charges Page 490
Section 7.3 Electrodynamics: Electromagnetic Induction
We may rewrite the left side using the vector identity for the curl of the curl as we did
when deriving Poisson’s Equation for A~ in terms of J~ (Equation 5.56):
~
~ind = −µo ∂ J
~ ∇
∇ ~ind − ∇2 E
~ ·E (7.56)
∂t
If we again assume no charge density (valid since we are considering only the induced
~ind ) and that the currents are localized so the fields fall off appropriately
electric field E
at infinity, we have a Poisson’s Equation for E ~ind , whose solution we know:
~ r 0)
∂ J(~
∂ J~ µo
Z
~ind = µo localized currents ~ind (~
∇2 E ⇐⇒ E r) = − dτ 0 ∂t
(7.57)
∂t 4π V |~
r −~ r 0|
Example 7.3: Induced Electric Field for Coaxial Conductors (Griffiths 7.16)
An alternating current I = I0 cos ωt flows down a long straight wire of negligible radius
and returns along a coaxial conducting tube of radius a and negligible thickness. Both
conductors are assumed to be perfect (infinite conductivity). We want to find the
induced electric field as a function of the transverse radius s in cylindrical coordinates.
For reasons that we will be able to explain later when we discuss EM waves in the
presence of conductors, the currents flow in sheets at the surfaces of the conductors
because they have infinite conductivity.
In the region between the wire and the outer conductor, the field of the wire is the
~ t) = φ
usual B(s, b µo I(t)/2 π s. The magnetic field of the return-current cylinder is
zero inside (consider an Amperian loop in the xy -plane with radius s < a: none of the
return current flows through the surface enclosed by that loop). Outside the
return-current sheet, its magnetic field is that of a wire carrying the total return
current, which has the same magnitude but opposite sign of the field of the inner wire.
Thus, the total magnetic field is the inner conductor’s magnetic field between the
conductors and is zero outside the outer conductor.
The system has azimuthal and z-translation symmetry, so the induced electric field
~ = Es (s) sb + Eφ (s) φ
must have the form E b + Ez (s) zb.
Section 7.3.6 Poisson’s Equation for Induced Electric Field, Proof of Lenz’s Law Page 492
Section 7.3 Electrodynamics: Electromagnetic Induction
If we think about what kind of Amperian loop has a nonzero flux of ∂ B/∂t ~ ~ it
(not J!),
is a loop in the sz plane with normal in the φ b direction. Let’s first consider a loop of
this kind with one z leg at infinity and the other at s > a. The contributions to the
loop integral of the electric field along the two radial legs cancel, and the contribution
from the leg at infinity vanishes assuming the fields fall off as s → ∞, so this loop
only gets a contribution from the z leg at finite radius, which picks out Ez (s > a).
~
The enclosed flux of ∂ B/∂t vanishes, so we can conclude Ez (s > a) = 0.
Now, repeat with one z leg at s between 0 and a and one z leg outside the outer
conductor. The radial legs cancel and the z leg outside the outer conductor
~
contributes nothing. When calculating the enclosed flux of ∂ B/∂t, a similar thing
holds: there is no magnetic field outside a, so the area integral only goes from s to a.
If the loop’s z dimension is `, we have
` ∂ Bφ (s 0 , t)
a µo ∂ I a ds 0
Z Z Z
Ez (s < a) ` = − dz ds 0
=− ` (7.58)
0 s ∂t 2 π ∂t s s0
µo a
= ω I0 ` sin ωt ln (7.59)
2π s
µo a
=⇒ Ez (s < a) = ω I0 sin ωt ln (7.60)
2π s
Note the sign: taking the loop normal to be φ b implies that the z leg with the nonzero
contribution yields a positive contribution. Then the usual minus sign enters, which is
cancelled by the sign of the derivative of cos ωt.
Section 7.3.6 Poisson’s Equation for Induced Electric Field, Proof of Lenz’s Law Page 493
Section 7.3 Electrodynamics: Electromagnetic Induction
We can see Eφ (s) vanishes by using a loop in the sφ plane that has radial legs (φ
constant) and azimuthal legs (s constant). One azimuthal leg can be taken to infinity
so it yields no contribution, and the radial legs’ contributions cancel, leaving only the
contribution from the azimuthal leg at finite radius. But, unlike the Ez case, this loop
~
has no magnetic flux and thus no ∂ B/∂t through it, so Eφ (s) = 0.
Section 7.3.6 Poisson’s Equation for Induced Electric Field, Proof of Lenz’s Law Page 494
Section 7.3 Electrodynamics: Electromagnetic Induction
Thus,
~ (s < a, t) = zb µo ω I0 sin ωt ln a
E ~ (s > a, t) = 0
E (7.61)
2π s
One can easily see the sign makes sense. This E ~ tries to drive a current parallel or
antiparallel to the current already flowing in the wire. When the current is decreasing,
the electric field is increasing to try to drive a current in the same direction in which
current is being lost by the decreasing current. It tries to generates a magnetic field
that would compensate for the magnetic field that is being removed by the decreasing
central conductor current. And vice versa for an increasing current.
Note how, when we can calculate the induced electric field directly, there is no
ambiguity about which direction the driven current would flow, unlike when we talk
about E.
Section 7.3.6 Poisson’s Equation for Induced Electric Field, Proof of Lenz’s Law Page 495
Lecture 29:
Electrodynamics IV:
Inductance
Magnetic Energy of Currents and Fields
Date Revised: 2022/04/04 05:00
Adjusted lecture break
Date Given: 2022/04/01
Page 496
Section 7.4 Electrodynamics: Inductance
Mutual Inductance
We have so far considered magnetic fields and fluxes in the abstract, without any
concern about where they come from. But they are generated by currents, so it is
natural to want to connect the Faraday’s Law emf to changing currents. We do that
through mutual inductance.
where we used the fact that B~ is derived from a vector potential followed by Stokes’
Theorem. Now, let’s use the relation between the current in C1 and A ~1 using the usual
~1 (assuming appropriate boundary conditions):
solution of the Poisson’s Equation for A
µo I1 d ~
`1 µo d~`2 · d ~
`1
I I I I
Φ21 = d~
`2 · = I1 (7.63)
4π C2 C1 |~
r2 − ~ r1 | 4π C2 C1 |~
r2 − ~r1 |
µo d~`2 · d ~
`1
I I
Φ21 = M21 I1 M21 = Neumann Formula (7.64)
4π C2 C1 |~
r2 − ~r1 |
where M21 is the mutual inductance between C1 and C2 and has units of Henries
(volt-second/amp). Two important characteristics:
We may now take the time derivative to calculate the emf at C2 due to a change in I1 :
dΦ21 dI1
E2 = − = −M21 (7.65)
dt dt
Self-Inductance
The above derivation works even when C1 and C2 are identical: a current loop induces
an emf on itself. In practice, calculating the integral can be difficult because of the
singularity at ~
r1 = ~
r2 , but one can be assured that self-inductance exists and is not
infinite. The symbol used is L and the corresponding equations are
µo d~`2 · d ~
`1 dI
I I
Φ = LI L= E = −L (7.66)
4π C C |~
r2 − ~r1 | dt
In both the cases of mutual inductance and self-inductance, one rarely does the
integral directly. Instead, one tries to find the field using Ampere’s Law, then calculate
the flux, and finally get M or L from Φ/I. This eliminates the need to deal directly
with the singularity in the above integral.
~ r 0)
~ r) = 1 J(~
Z
A(~ dτ 0 (7.67)
4π V |~
r −~ r 0|
µo 1
Z Z ~ r ) · J(~
J(~ ~ r 0)
=⇒ Mij = dτ dτ 0 (7.68)
4π Ii Ij Vi Vj |~
r −~ r 0|
µo 1
Z Z ~ r ) · J(~
J(~ ~ r 0)
L= dτ dτ 0 (7.69)
4π I2 V V |~r − r 0|
~
where Vi is the volume of the ith inductor. We notice that the currents do not drop
out as cleanly, but, assuming linear behavior of the current flow (the current does not
flow differently as the overall magnitude of the current is changed), we expect J~ ∝ I
and indeed, once the functional dependence of the current density on position has
been established, the inductances are purely geometrical quantities as for the case of
line currents.
This is not a rigorous proof. We will return to this when we discuss magnetic energy.
~ = µo n I zb
B (7.70)
where n is the number of turns per unit length, the solenoid axis is along zb, and the
current flows along the φb direction. The magnetic flux threading the solenoid and the
self-inductance are therefore
Φ = n ` π a2 B = µo n2 ` π a2 I =⇒ L = µo n2 ` π a2 (7.71)
If we have two interpenetrating solenoids with turn densities n1 and n2 , radii a1 < a2 ,
and lengths `1 < `2 , then the flux into solenoid 1 of the field from solenoid 2 and the
mutual inductance are
It is interesting and useful to note that we may also calculate Φ21 using M given the
symmetry of M. This is very convenient, as calculating the contribution to Φ21 from
the portion of solenoid 1’s field past its ends would be nontrivial. It is also interesting
to see that the mutual inductance is not manifestly symmetric under index exchange
1 ↔ 2. This reflects the asymmetry of the setup between solenoids 1 and 2.
That is, when a varying current is driven through an inductive object, it has to be
driven against a force per unit charge whose line integral (note that we did not say
potential!) along the current’s path is E. (The force is due to an induced electric field
for this case of a stationary loop that is experiencing a dΦ/dt due to its own current
varying.) The force that must be exerted, and the work that must be done, is above
and beyond the force needed to overcome the inertia of the charge carriers (i.e., the
Newton’s Law force F = m a).
The rate at which this work is being done is given by the same expression we derived
before for the work done by the pulling force in the case that the field was fixed but
the loop was moving: it is the work done per unit charge by the battery to push the
current against the back emf, −E, times the charge flowing past a given point per unit
time, I:
dW dI
= Power = −I E = L I (7.73)
dt dt
We can integrate this over time to get the total work done and the magnetic energy
created:
1 2
W = LI (7.74)
2
It is natural to ask why, when we considered the situation with the loop being pulled
through a magnetic field, we did not worry about this magnetic energy: we said that
the work done by the pulling force was completely dissipated in the Joule heating of
the resistor. Or, put another way, why did we not need to include a resistor in the
calculation here? When we include the resistor, some of the work done by the pulling
force as the loop was accelerated from rest to ~v goes into this magnetic energy, the
L I2 /2 energy. Once at fixed velocity, however, the current and thus this energy stay
constant. The pulling force continues to do work, however. Since we specifically made
the steady-state assumption — that the loop had been moving at fixed ~ v for all time
and would stay moving for all time — this transient process of creating L I2 /2 was not
relevant, and our conservation of energy argument was valid; we just neglected noting
the path that the energy took through the magnetic energy in the steady state. Now,
without the resistor, we are focused entirely on the transient portion of the process,
hence the importance of the magnetic energy.
Let’s generalize the above result. Consider a system of N inductive elements with
inductance matrix Mij (Mii ≡ Li , Mij = Mji ). We turn on the currents in the order
i = 1, 2, · · · , N. We first have to maintain the current Ii against the emf on inductor i
felt due to its changing current. Once it has reached its final value, we also have to
maintain it as the currents in the inductors j > i are increased from 0 to their final
values (note: j > i, not j < i as we had in electrostatics). The required power is:
N N N N
dW X dWi X X dIi
X dIj
= = (−Ii Ei ) = Ii Mii + Ii Mij
dt i=1
dt i=1 i=1
dt j>i
dt
N N N
X 1 2
X 1 X
=⇒ W = Mii Ii + Ii Mij Ij = Mij Ii Ij (7.75)
i=1
2 j>i
2 i,j=1
where we have symmetrized the sum over j by including a factor of 1/2, and then we
combined the cross-terms with the self-terms. If we rewrite all our relations using
matrix notation, with I being a column vector of currents, Φ being a column vector of
fluxes, and M being the matrix of mutual inductances, we have
1 T 1
Φ=MI W = Φ I = IT M I (7.76)
2 2
Note: we could have calculated the above somewhat differently, considering the work
done to maintain the loops j < i at their final current values, plus the work done in
loop i itself, while loop i is begin ramped to its final value. The result would be the
same.
Section 7.5.1 Magnetic Energy in Terms of Currents Page 504
Section 7.5 Electrodynamics: Magnetic Energy and Forces
1
Z
W = dτ J~ · A
~ (7.79)
2 V
Aside: The above equation now justifies Equation 7.69: if one uses Equation 5.56 to
~ in terms of J~ and then calculates L = 2 W /I 2 , one recovers Equation 7.69. By
write A
considering two separate volume current distributions, one can recover Equation 7.68
also.
1
Z
W = ~· ∇
dτ A ~
~ ×B (7.80)
2 µo V
1
Z h i
W = dτ B ~· ∇~ ×A ~ −∇~ · A ~×B ~ (7.81)
2 µo V
1 ~ 2− 1
Z I
= dτ |B| da nb · A~×B ~ (7.82)
2 µo V 2 µo S(V)
Now, the original volume integral was over only the region containing the current, but
the volume integral could be extended to a larger region since there would be no
additional contribution. So we do the usual thing and expand the volume to include
all of space and take the bounding surface to infinity.
We assume that A ~×B ~ falls off more quickly than 1/r 2 (true for finite current
distributions) so that the surface term goes to zero, or that the particulars of the
configuration ensure the integral vanishes even if the current distribution is not finite
and we expect a finite energy. Therefore,
1
Z
W = ~ 2
dτ |B| (7.83)
2 µo
Thus we see that the magnetic energy is just given by the integral of the square of the
field. We now see that the magnetic energy created as the currents are ramped from
zero to their final values is stored in the field. (We specifically avoided saying earlier
where it was stored!)
On the point about the surface term: For an infinite solenoid, the surface term only
includes the endcaps of the solenoid, since B ~ vanishes outside the solenoid. The
contributions of the two endcaps vanish because A ~×B ~ points along sb in cylindrical
coordinates, but the endcap’s normal is along zb. For an infinite wire, even when
calculated per unit length, all the terms are logarithmically infinite. This is because the
current and the fields do not die off quickly enough at infinity. The calculation fails
even if one does the calculation using J~ · A,
~ and even for a finite diameter wire (if one
accounts for the fact that J~ becomes a surface current in the perfect conductor case).
1 1
W = π a2 ` B 2 = µo n2 I2 π a2 ` (7.84)
2 µo 2
Note that we can extract from this the self-inductance using W = L I2 /2, yielding
L = µo n2 π a2 ` as we obtained by calculating the flux. To put some numbers on this,
the LHC CMS experiment (http://home.web.cern.ch/about/experiments/cms) has
a solenoid with a field of 4 T with radius a = 3 m and length 13 m. The stored energy
is therefore about 2.5 gigajoules, an enormous number.
µo 2 b µo b
W = I ` ln L= ` ln (7.85)
4π a 2π a
We followed Griffiths in first developing concepts for inductance and energy using
current loops and then generalizing both to volume current distributions in a fairly
obvious way. That said, there is a more rigorous way to do all of this by first
considering the work done in maintaining current densities in the presence of the
electric field generated by changes in those current densities. This then leads to the
idea of J~ · A
~ being the energy density in the magnetic field. Then one can define
inductances by writing the field energy in terms of the total currents that normalize
the current distributions. It can then be shown that these inductances relate currents
to fluxes and thus rates of change of currents to emfs. This alternative logical path is
followed in Jackson §5.16–5.17.
Page 510
Section 7.5 Electrodynamics: Magnetic Energy and Forces
This is done similarly to the electrostatic case and follows Jackson §5.16. Recall that
we discussed the distinction between the total energy needed to assemble the final
configuration, including the construction of the bound dipoles, and the energy needed
to bring the free charges in assuming the bound dipoles already exist and neglecting
the potential energy of creating them. In this case, we assume the bound magnetic
dipoles are created and maintained by someone else — someone else has built them
and raised their currents to their full values for us and also maintains those currents in
the presence of back emf generated when the free currents change — and we need
only consider the work that has to be done to turn on some free currents in the
presence of these bound magnetic dipoles.
This separation is not academic: all naturally occurring magnetic materials rely on the
magnetic dipole moments of fundamental particles. Those magnetic dipoles are
unchangeable, and thus the energy stored in them is effectively a constant offset that
we have no experimental access to. It therefore makes sense to want to ignore it in
calculations of magnetic energy.
Section 7.5.4 Magnetic Energy of an Assembly of Free Currents in the Presence of Magnetizable Materials Page 511
Section 7.5 Electrodynamics: Magnetic Energy and Forces
Consider the differential of work the battery must do to maintain the free currents J~f
during a change in the magnetic field δ B.~ (We need not specify whether this change is
due to a change in J~f and/or the location of the magnetizable materials — all that
~ Equation 7.78 implies that the change in energy would be (holding the
matters is δ B.)
geometry fixed)
"Z #
dΦ
I I
δW = −I E δt = I δt = I δΦ = I δ ~
da nb · B = I δ d~ ~ =I
`·A d~ ~
` · δA
dt S(C) C C
(7.86)
for which the volume generalization would be
Z Z
δW = dτ J~f · δ A
~= dτ ∇ ~ · δA
~ ×H ~ (7.87)
V V
Apply the same algebra and the same discarding of the surface term as in free space:
Z Z
δW = ~· ∇
dτ H ~ × δA~ = ~ · δB
dτ H ~ (7.88)
V V
~ H)
For nonlinear materials, we would need to apply the specific B( ~ function go further.
~ = µδ H
For linear materials, we use δ B ~ to do the integral and obtain the expected
analogue to the free-space result:
1 ~ 2= µ ~ 2= 1
Z Z Z
W = dτ |B| dτ |H| ~ ·B
dτ H ~ (7.89)
2µ V 2 V 2 V
Section 7.5.4 Magnetic Energy of an Assembly of Free Currents in the Presence of Magnetizable Materials Page 512
Section 7.5 Electrodynamics: Magnetic Energy and Forces
We can calculate this along the lines of the derivation we did for polarizable materials,
following Jackson §5.16. Let’s assume that we have a configuration of currents that
~ 1 and H
generates fields B ~ 1 in a volume containing a permeable material µ1 . Now,
bring in a material of permeability µ2 such that it occupies a volume V2 contained in
V while holding the free source currents fixed. The fields (everywhere) change to B ~2
and H~2 .
1
Z h i
U2 − U1 = dτ B~2 · H
~2 − B
~1 · H
~1 (7.90)
2
Section 7.5.5 Magnetic Energy of a Magnetizable Material in an External Field Page 513
Section 7.5 Electrodynamics: Magnetic Energy and Forces
We can apply similar manipulations as we did for the electrostatic case. First, we
rewrite the above as
1 ~2 + 1
Z h i Z h i h i
U2 − U1 = dτ B~2 · H
~1 − B
~1 · H dτ B ~1 + B
~2 · H ~2 − H
~1 (7.91)
2 2
h i
~ · B
Since ∇ ~1 + B ~ allowing us to
~ 2 = 0, it can be derived from a vector potential A,
rewrite the second term as
1
Z h i
dτ H ~2 − H
~1 · ∇ ~ ×A~ (7.92)
2
1
Z
dτ A~·∇
~ × H ~2 − H~1 (7.93)
2
~ 2 and H
The curl in the integrand vanishes because H ~ 1 are sourced by the same free
currents.
Section 7.5.5 Magnetic Energy of a Magnetizable Material in an External Field Page 514
Section 7.5 Electrodynamics: Magnetic Energy and Forces
We are thus left with the first term from the equation we started with
1
Z h i
U2 − U1 = dτ B ~2 · H
~1 − B
~1 · H
~2 (7.94)
2
~ = µ H,
Applying linearity, B ~ we then obtain
1
Z
U2 − U1 = ~2 · H
dτ (µ2 − µ1 ) H ~1 (7.95)
2
1 1 1 1
Z Z
U2 − U1 = ~2 · H
dτ (µ2 − µ1 ) H ~1 = dτ − ~2 · B
B ~1 (7.96)
2 V2 2 V2 µ1 µ2
This is the analogue of Equation 4.83 aside from a sign flip, which mechanically is due
~ = µH
to the fact that B ~ (rather than H~ = µ B).
~ If we take µ1 = µ0 and µ2 = µ, we
can use M~ 2 = (µ2 /µ0 − 1)H~ 2 = (µ/µ0 − 1)H
~ 2 to rewrite this as
1 1 ~ ~
Z
W = U2 − U1 = ~ ·B
dτ M ~1 ⇐⇒ w = M · B1 (7.97)
2 V2 2
How do we understand this sign flip conceptually? Trying to track the sign through
the derivation is not illuminating. But it can be understood by comparing to our
calculation of the energy of magnetic dipole in an external field, Equation 5.144,
where we assumed that the magnetic dipole moment and field were given and held
fixed without our having to account for how this was done. In that case, the potential
energy of the configuration was U = −m ~ (The factor of 1/2 here comes from the
~ · B.
linear relationship between m ~ and B ~ and the integration from zero field to B, ~ which is
not important for this discussion). We see that we have a sign flip relative to that
situation. It is sensible, then, to attribute the sign flip to the fact that, in deriving the
expression w = |B| ~ 2 /2 µ that was the starting point for this derivation, we accounted
for the work done by the batteries to maintain the free currents as the permeable
material was brought in. No such work was required in the previously considered case
of a fixed dipole moment m ~
~ and fixed field B.
Note that, importantly, we do not account for how the magnetization density M ~ is
maintained. This is to be distinguished from not considering how M ~ is created, which
we argued was just an unchangeable offset. We may ignore this additional
consideration here because, again, the magnetization density is, in naturally occurring
systems, due to fundamental magnetic dipoles that require no batteries to maintain
their magnetic moments.
When we compare to the electrostatic analogy, Equation 4.84, we recognize a sign flip,
too. The rationale is the same: in the electrostatic case, we do not have to do any
work to maintain the free charges sourcing the applied field E~ at their nominal
positions, while here we do have to do work with a battery to maintain currents at the
nominal values and positions due to the back emf from the changing M. ~
Section 7.5.5 Magnetic Energy of a Magnetizable Material in an External Field Page 516
Section 7.5 Electrodynamics: Magnetic Energy and Forces
N ∂ Mij−1
∂ Wfield 1 X 1 T ∂
−1
Fξ = − =− Φi Φj =− Φ M Φ (7.98)
∂ξ Φ 2 i,j=1 ∂ξ 2 ∂ξ
Φ Φ Φ
It’s not entirely clear at a microscopic level (i.e., what has to happen to the currents)
how one maintains fixed fluxes as inductors are moved around. But, certainly, one is
assured that, if one sets up a system of inductors with currents and then disconnects
them from their batteries, any movement of the loops must keep the fluxes fixed and
change the currents accordingly since there are no batteries to work against the emfs
and maintain the currents. This issue will be revisited in homework and is discussed in
Griffiths Section 8.3 (4th edition).
Section 7.5.6 Magnetic Forces from Magnetic Energy with Fluxes Fixed Page 517
Section 7.5 Electrodynamics: Magnetic Energy and Forces
N
1 X h i
dWfield = Φi Φj d M −1 (7.99)
Φ 2 i,j=1 ij
N
X h i
dIi = d M −1 Φj (7.100)
Φ ij
j=1
N
X N
X h i
dΦk = Mki (−dIi )Φ = − Mki d M −1 Φj (7.101)
I ij
i=1 i,j=1
Section 7.5.7 Magnetic Forces from Magnetic Energy with Currents Fixed Page 518
Section 7.5 Electrodynamics: Magnetic Energy and Forces
The above infinitesimal flux changes cause emfs. The currents Ii must be maintained
by batteries in the presence of these emfs. The work done by the batteries over the
infinitesimal time dt needed to make the flux changes is
n n
bat
dWfield
bat
X X dΦk
dWfield = dt = dt (−Ik Ek ) = dt Ik (7.102)
dt dt
I
I k=1 k=1 I
n
X n
X h i
= Ik dΦk = − Ik Mki d M −1 Φj (7.103)
I ij
k=1 i,j,k=1
n
X h i
=− Φi Φj d M −1 = −2 dWfield (7.104)
ij Φ
i,j=1
Note that we did not need to worry about the work done by the battery to make the
canceling change in current dIk because this current change would be multiplied
against dt Ek = −dΦk , which is already infinitesimal. We need only consider the
above term consisting of the nominal currents Ik multiplied against dt Ek . We had the
same situation in electrostatics, where we did not consider the dVk dQk terms, only
the Vk dQk terms.
Section 7.5.7 Magnetic Forces from Magnetic Energy with Currents Fixed Page 519
Section 7.5 Electrodynamics: Magnetic Energy and Forces
The total change in the field energy is then obtained by adding the two contributions
to the field energy, the field energy change at fixed flux followed by the energy added
to the field by the batteries as they return the currents to their initial values:
bat
dWfield = dWfield + dWfield = dWfield − 2 dWfield = −dWfield (7.105)
I Φ I Φ Φ Φ
∂ Wfield ∂ Wfield
Fξ I = =− = Fξ Φ (7.106)
∂ξ I ∂ξ Φ
That is, just like in the electrostatic case, when the battery is involved and we
consider the energy of the entire system, we see we must take the positive gradient of
the field energy at fixed current, rather than considering only the energy of the field
and taking the negative gradient of the field energy at fixed current. The reason these
two gradients are different, with a sign between them, is because the derivative is
calculationally different depending on whether I or Φ is held fixed.
Section 7.5.7 Magnetic Forces from Magnetic Energy with Currents Fixed Page 520
Section 7.5 Electrodynamics: Magnetic Energy and Forces
Since ∂M −1 /∂ξ = −M −1 [∂M/∂ξ]M −1 (one can see this in the same way we proved
the analogous relationship for C ), this form yields Equation 7.98 for Fξ Φ . Thus,
∂ Wfield ∂ Wfield
Fξ I = =− = Fξ Φ (7.108)
∂ξ I ∂ξ Φ
Section 7.5.7 Magnetic Forces from Magnetic Energy with Currents Fixed Page 521
Lecture 31:
Electrodynamics VI:
Displacement Current
Maxwell’s Equations
Date Revised: 2022/04/06 06:45
Date Given: 2022/04/06
Page 522
Section 7.6 Electrodynamics: Maxwell’s Equations
Maxwell’s Equations
The Inconsistency in our Equations
Let’s write the full set of equations we have come to:
∇ ~= ρ
~ ·E ~ =0
~ ·B
∇ (7.109)
o
~
∇ ~ = −∂B
~ ×E ~ = µo J~
~ ×B
∇ (7.110)
∂t
Now, we know that the divergence of a curl vanishes: it’s a vector identity. We should
check that it holds! For the electric field, we have
~
~ ·∇
∇ ~ ×E ~ · −∂B = − ∂ ∇
~ =∇ ~ =0
~ ·B (7.111)
∂t ∂t
~ we obtain
If we repeat with B,
~ ·∇
∇ ~ ×B ~ · J~ = −µo ∂ ρ 6= 0
~ = µo ∇ in general (7.112)
∂t
There is a more physical way to see this by applying Ampere’s Law to a circuit
containing a parallel-plate capacitor. Construct an Ampere’s Law loop around the wire
carrying the current. Ampere’s Law is satisfied because there is magnetic field in an
~ and there is
azimuthal direction around the wire (giving a nonzero line integral of B)
current passing through the disk-like surface whose boundary is the contour.
Now pick another surface that passes between the capacitor plates. This is an equally
valid surface; nothing about our proof of Ampere’s Law from the Biot-Savart Law
assumed a particular choice of surface for the Ampere’s Law surface integral. But this
surface has no current intersecting it because it passes through the capacitor!
The reason this problem happens and we never noticed it before is because we have a
non-steady-state case here: charge piles up on the capacitor plates giving ∂ρ/∂t 6= 0;
we had assumed all along during magnetostatics and during our discussion of
induction that ∂ρ/∂t = 0.
While charge cannot jump across the capacitor plates so that there can be a current to
keep Ampere’s Law satisfied, we do recognize that, as charge enters one plate of the
capacitor, an equal amount of charge leaves the other plate, ensuring that dQ/dt = 0
for the capacitor as a whole. This is suggestive of the idea that perhaps there is some
sort of current flowing across the capacitor gap, just not the physical movement of
charges we are used to. This new current will be called the displacement current.
Was it ok to do this? Does it violate any of our previous conclusions? The only
equation we have modified is the ∇ ~ ×B ~ equation, so we only need to consider our
study of magnetostatics, where we applied this equation. The addition preserves the
~ ×B
usual behavior of ∇ ~ for magnetostatics because ∂ E ~ /∂t = 0 in magnetostatics.
Why? Two things can result in time dependence of E ~ . The first is time dependence in
ρ. But in magnetostatics, we make the steady-state assumption, explicitly requiring no
buildup of charge and hence ∂ρ/∂t = 0. The second is time dependence of B, ~ which
can yield time dependence of E ~ via Faraday’s Law. But magnetostatics assumes B ~ is
constant in time, so there is no worry there.
Section 7.6.2 The Displacement Current Page 525
Section 7.6 Electrodynamics: Maxwell’s Equations
~
∂E
J~d ≡ o (7.115)
∂t
More importantly, we also now see for the first time that a changing electric field
sources a magnetic field. Unlike with Faraday’s Law, however, there is no negative
sign and the induced magnetic field does not act in such a way as to try to cancel the
changing electric field.
~ = σ nb = 1 Q nb
E (7.116)
o o A
where nb is the normal from the positive plate to the negative plate. Therefore, the
displacement current is
~
∂E 1 dQ I
J~d = o = nb = nb (7.117)
∂t A dt A
The integral form of Ampere’s Law with the displacement current is therefore
I Z
d~ ~ = µo Iencl + µo
`·B da nb · J~d (7.118)
C S(C)
If we choose the first surface we discussed earlier, the flat surface in the plane of the
contour C, we get the first term but the second term vanishes, yielding µo I. If we
choose the second surface, the one between the capacitor plates, the first term
vanishes but the second term gives µo I. Thus, the inconsistency seen earlier has been
eliminated.
∂ µo a ω 2 I0 a
J~d (s < a) = o zb ω I0 sin ωt ln = zb µo o cos ωt ln (7.119)
∂t 2 π s 2π s
Let’s integrate over the (s, φ) plane to get the total displacement current:
Z Z 2π ω 2 I0
Z a
a
Id = nb · J~D = dφ µo o cos ωt s ds ln
S(C) 0 2π 0 s
2 0 !
x 1
= µo o ω 2 I0 cos ωt a2
ln x −
2 2 x=1
a2
= µo o ω 2 I0 cos ωt (7.120)
4
We did not include the displacement current in the calculation of the magnetic field in
the system. Is that a problem?
Well, the problem is, in principle, even worse: we ought to include the displacement
current in the calculation of B, ~ but then our calculation of E ~ via Faraday’s Law needs
to also be corrected for the B ~ due to the displacement current, yielding a correction to
E~ , which itself will yield a correction to J~d , and so on. The proper way to handle this
is to develop the formalism for electromagnetic waves, where we self-consistently solve
all of Maxwell’s Equations.
For now, it is instructive to look at the relative size of J~d and J~ so we can understand
why these corrections are small and thus why our previous results, while not precisely
correct, are an excellent approximation.
The ratio of the amplitudes of the displacement current and the true current, up to
factors of order unity, is
The numerator of the final expression is the square of the light travel time over the
length scale of the problem, a. The denominator is, up to a factor (2 π)2 , the square
of the oscillation period. Thus, this quantity is a measure of how quasistatic the
system is. We have mentioned before that, if a/c 1/ω is not satisfied, then our
quasistatic approximation is invalid. This corroborates that: if the oscillation period
becomes comparable to the light travel time so the system is no longer quasistatic,
then the displacement current will approach the real current in magnitude and our
~ ignoring the displacement current will be a bad approximation.
prior calculation of B
The ratio of the displacement current to the true current scales as ω 2 , so one must go
to high frequency to notice it. Quantitatively, if we ask how high in frequency one
must go to obtain Id /I = 0.01 if we take a = 2 mm as the dimension of the coaxial
conductor, we obtain
1 3 × 1011 mm/s √
r
ω 1 c Id
ν= = = 0.01 ≈ 2 GHz (7.122)
2π 2π a I 2π 2 mm
GHz oscillators were not available in Faraday’s time, so the fact that he did not
observe the effects of the displacement current is not surprising.
Section 7.6.2 The Displacement Current Page 530
Section 7.6 Electrodynamics: Maxwell’s Equations
~ ~
∇ ~= ρ
~ ·E ∇ ~ =0
~ ·B ~ = −∂B
~ ×E
∇ ~ = µo J~ + o µo ∂ E
~ ×B
∇
o ∂t ∂t
(7.123)
~ · J~ = − ∂ ρ
F~ = q E~ +~ ~
v ×B ∇ (7.124)
∂t
~ ~
∇ ~= ρ
~ ·E ∇ ~ =0
~ ·B ~ + ∂B = 0
~ ×E
∇ ~ − o µo ∂ E = µo J~
~ ×B
∇
o ∂t ∂t
(7.125)
ρb = −∇ ~
~ ·P (7.126)
~
∂P ~
J~p = ⇐⇒ ∇ ~ · ∂P = − ∂ ∇
~ · J~p = −∇ ~ = − ∂ ρb
~ ·P (7.127)
∂t ∂t ∂t ∂t
That is, the definition on the left naturally gives the continuity relation between J~p
and ρb one would like.
Intuitively, think of J~p as follows. Suppose one has a cylinder of polarizable material of
length dz and cross-sectional area da and with polarization vector P ~ = P zb. The
definition ρb = −∇ ~ ·P ~ implies that there is a bound surface charge
Q = σ da = nb · P ~ da = ±P da at each end. If, for example, we allow P ~ to vary
sinusoidally, P~ =P ~ 0 sin ωt, which corresponds to the surface charge obeying
Q(t) = P0 da sin ωt, then the current is
dQ
Ip = J~p · nb da = P0 da ω cos ωt = (7.128)
dt
as would be necessary to transfer charge back and forth between the two ends of the
cylinder to yield the corresponding time-dependent surface charge. This current is,
literally, the motion of the charges that make up the dipoles as they flip back and
forth sinuisodally.
~ Recall that M
Do we have to worry about time dependence of M? ~ yields a bound
current density
J~b = ∇ ~
~ ×M (7.129)
Let’s use all this to rewrite Maxwell’s Equations in terms of free charges and currents.
The charge and current densities have the following parts:
~
~
~ ·P
ρ = ρf + ρb = ρf − ∇ J~ = J~f + J~b + J~p = J~f + ∇ ~ + ∂P
~ ×M (7.130)
∂t
~ = ρf − ∇
~ ·E
Using Gauss’s law, o ∇ ~ and the definition of the displacement field,
~ · P,
~ = o E
D ~ + P,
~ we obtain
∇ ~ = ρf
~ ·D (7.131)
~
∇ ~ = J~f + ∂ D
~ ×H (7.133)
∂t
Now it is clear why the last term is called the displacement current — it is the
~
apparent current due to the time variation of the displacement vector D!
Faraday’s Law and ∇ ~ ·B~ = 0 are not affected since they do not depend on the free
and bound currents. Thus, Maxwell’s Equations in matter are (again, putting all the
fields on the left sides and the sources on the right):
∇ ~ = ρf
~ ·D ~ =0
~ ·B
∇ (7.134)
~ ~
∇ ~ + ∂B = 0
~ ×E ~ − ∂ D = J~f
~ ×H
∇ (7.135)
∂t ∂t
~
These equations must be supplemented by specific constitutive relations between E
~ and between B
and D ~ and H
~ to completely specify the behavior (and, of course,
boundary conditions must be provided). For linear media, these relations are:
linear media: ~ = χe o E
P ~ ~ = χm H
M ~ ~ = E
D ~ ~ = µH
B ~ (7.136)
Page 537
Lecture 32:
Conservation Laws:
Energy
Electromagnetic Waves I:
Electromagnetic Waves in Vacuum:
Wave Equation
Date Revised: 2022/04/07 23:30
Date Given: 2022/04/08
Page 538
Section 8.1 Conservation Laws: Motivation and Analogy: Conservation of Charge
r)
~ r ) = − ∂ ρ(~
~ · J(~
∇ (8.1)
∂t
In electrodynamics, we want to ask the same question for energy and momentum
because we want to understand whether the fields we have constructed have true
physical meaning or are just mathematical constructs. Determining whether they carry
energy and momentum is one way to answer that question, and such a consideration
leads to the question of conservation of these quantities.
Page 539
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
o 1
Z Z
We = ~ |2
dτ |E Wm = ~ 2
dτ |B| (8.2)
2 2 µo
Recall that this is the work needed to move new charge in from infinity due to the
repulsion from the charge already there, or the work that needs to be done to raise a
current from zero to its final value against a back emf. It is thus natural to expect
that the total energy density in the electric and magnetic fields is
1 1 ~ 2
ufield = ~ |2 +
o |E |B| (8.3)
2 µo
Will show this is valid by considering the exchange of energy between the fields and
charges/currents. We use the term electromagnetic field to reflect the fact that the
fields influence each other and their energies are on the same footing.
Page 540
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Given a single particle of charge q acted on by the electromagnetic field, the work
done on it as it moves by d ~` is
dW = F~ · d ~
`=q E~ +~ ~ dt = q E
v ×B ~ ·~
v dt (8.4)
dW
Z
= ~ · J~
dτ E (8.5)
dt
~
~ · J~ = 1 E ~ · ∂E
E ~· ∇ ~ − o E
~ ×B (8.6)
µo ∂t
One subtlety here: we started off talking about J~ being acted upon by an
electromagnetic field, and now it seems like we are treating J~ as the source of that
field. It is not the sole source of the field because, now with the displacement current
term combined with Faraday’s Law, there can be electric and magnetic fields that are
sourced by each other’s time variation rather than by physical currents. The above
substitution is nevertheless valid because the second term subtracts off the
displacement current term that is due to changing fields rather than physical current:
one should not be able to do work on the displacement current!
Page 541
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Another subtlety is the issue of whether J~ can create fields that do work on itself.
This is entirely possible, as we saw in the example of the time-varying current in the
coaxial conductor: a time-varying current generated a time-varying magnetic field that
generated a time-varying electric field aligned with the original current. If there were
no battery driving the current, then the work being done by the field on the current
should reduce the energy in the current in exactly the way that would be needed to
conserve energy. Of course, if a battery is involved, then it can supply energy and we
do not expect the energy of the currents and fields alone to be conserved.
Page 542
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Incorporating the above and applying the divergence theorem to the last term, we thus
obtain Poynting’s Theorem:
" Z #
dW d 1 ~ |2 + 1 |B|
I
=− dτ o |E ~ 2 + ~
da nb · S (8.10)
dt dt V 2 µo S(V)
~= 1 E
with the Poynting vector defined to be S ~ ×B
~ (8.11)
µo
Poynting’s Theorem says that the work per unit time done on the charges and
currents in a volume V by electromagnetic forces is equal to negative of the sum of
the change per unit time of the energy in the fields and the energy flowing outward
through the surface of V. S~ has units of energy per unit time per unit area and is
considered the energy flux density.
Page 543
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Another useful form is given by putting the field energy density term on the left side:
dW d 1 1 ~ 2 1
Z I
+ dτ ~ |2 +
o |E |B| =− ~
da nb · S (8.12)
dt dt V 2 µo µo S(V)
d
I
(Emech + Efield ) = − ~
da nb · S (8.13)
dt S(V)
The rate of change of the total energy in the volume is given by the flux of the
Poynting vector through the boundary of the volume — this much more explicitly puts
mechanical and field energy on the same footing and shows that both can be
transported by the Poynting flux.
Note that this allays our fears about a current doing work on itself: while it may do
so, energy remains conserved as long as one takes into account the field energy.
Page 544
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
We may write both versions in a local form by recognizing that the volume being
integrated overR is arbitrary. If we define umech (~
r ) to be the density of mechanical
energy (W = V dτ umech ) as a function of position and
1 1 ~
ufield = ~ (~
o |E r )|2 + r )|2
|B(~ (8.14)
2 µo
to be the energy density of the electromagnetic field, then our two versions of
~
Poynting’s theorem yield the local relations (after converting the surface integral of S
back to a volume integral using the divergence theorem):
∂ umech (~
r) ∂ ufield (~
r) ~ r)
~ · S(~ ∂ ~ r)
~ · S(~
=− −∇ ⇐⇒ [umech (~ r )] = −∇
r ) + ufield (~
∂t ∂t ∂t
(8.15)
This is the kind of local conservation theorem we wanted, relating the rate of change
of a density (here the energy density) to the divergence of a current density (here the
Poynting vector).
When there is no change in mechanical energy — e.g., in empty space — then we can
specialize the above to obtain the continuity equation for the energy of the
electromagnetic field:
dEfield ~ r ) = − ∂ ufield (~
r)
I
umech = 0 : =− ~ ⇐⇒ ∇
da nb · S ~ · S(~ (8.16)
dt S(V) ∂t
Page 545
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Example 8.1: Power Transported Down a Coaxial Cable (Griffiths Problem 8.1)
Consider a coaxial cable with central conductor of diameter a and outer conductor of
radius b and zero thickness. A static current flows along +b z on the central conductor
and back along −b z on the outer shell. We used a similar geometry in a previous
example, but with a time-varying current in that case. The inner conductor is held at
voltage V and the outer conductor at V = 0 (ground) at one end of the cylinder, and
there is a resistive sheet of sheet conductivity σ (definition to be provided) capping
the other end.
Because the inner conductor is assumed to have infinite conductivity, there can be no
electric field inside and thus all the current must flow on the surface (consequence of
Ohm’s Law: J~ = 0 because E ~ = 0). The calculation of the magnetic field is thus the
same as the prior example in the same geometry with time-varying current. The
magnetic field between the conductors is
~ µo I b
B(s) = φ (8.17)
2πs
where I is the current due to V (value to be determined). In the prior example, we did
not explicitly have a voltage on the inner conductor (effectively, the conductivity of
the sheet at the end was infinite).
Page 546
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Here, since we have such a voltage, there is a line charge density on the inner
conductor and a radial electric field. You are no doubt familiar with the Gauss’ Law
calculation of this configuration, which yields
~ (s) = λ
E sb (8.18)
2 π o s
Let’s find λ by matching to the applied voltage. The potential and field are
λ b ~ (s) = V
V (s) ∝ ln s V = V (a) − V (b) = ln =⇒ E b
sb (8.19)
2 π o a s ln a
~= 1 E
S ~ ×B
~ = IV
zb (8.20)
b
µo 2 π s 2 ln a
The energy and energy current are between the conductors, not in them! The power
flowing down the cable is found by integrating the Poynting vector over the
cross-sectional area where the fields are:
Z b Z 2π Z b
IV IV ds
Z
P= ~
da nb · S(s) = s ds dφ = = IV (8.21)
2 b b
S a 0 2 π s ln a ln a a s
Page 547
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
Let’s calculate the power dissipated in the resistive sheet at the end. The sheet does
not disturb the potential because the sheet continues to satisfy Laplace’s equation
with the same boundary conditions in s: V (a) = V , V (b) = 0. Therefore, our electric
field above is valid in the conducting sheet, and the surface current density and total
current are
K ~ (s) = σ V sb
~ (s) = σ E (8.22)
s ln ba
where σ is the conductivity per square, which can be thought of as σ = limt→0 σ/t
where σ is the usual conductivity and t is the thickness of the sheet. The total current
is
Z 2π 2 π σ V
I= s dφ K (s) = 2 π s K (s) = b
(8.23)
0 ln a
Page 548
Section 8.2 Conservation Laws: Poynting’s Theorem: Conservation of Energy
b
V ln a
R= = (8.24)
I 2 π σ
2 π σ V V2
= b
= I V = I2 R = (8.26)
ln a
R
as expected since this is the power coming down the central conductor and it cannot
go beyond the resistive sheet since the fields go to zero out there (no current or
charge density beyond the sheet).
Page 549
Lecture 54:
Conservation Laws:
Stress Tensor/Linear Momentum
Torque Tensor/Angular Momentum
Date Revised: TBD
Date Given: TBD
These two sections,
consisting of material colored green and orange,
is optional!
Page 550
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
The Maxwell Stress Tensor: Conservation of Linear Momentum
(Skip)
We showed in the previous section that fields carry energy and that one must account
for that energy in order for conservation of energy to hold. The natural next question
to ask is whether the electromagnetic fields carry momentum. The matter and fields
are related by the fields exerting forces on the matter, so let’s use these forces to
connect the momentum of the matter and fields. The Lorentz Force Law is
~ mech
dP
= F~ = q E~ +~ ~
v ×B (8.27)
dt
~ mech
dP
Z
= dτ ~ + J~ × B
ρE ~ (8.28)
dt V
Using Maxwell’s Equations, we can write this purely in terms of the fields:
" # !
~ mech
dP 1 ~ ~
~ − o ∂ E
Z h i
= dτ o ∇ ~ E
~ ·E ~+ ∇×B ~
×B (8.29)
dt V µo ∂t
We recall the same subtleties as for energy: ρ and J~ now being taken as source of
fields, and the last term subtracts off the displacement current since the magnetic field
exerts no force on it.
Page 551
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
" #
~ mech
dP ~
~ · T − o µo ∂ S
Z
= dτ ∇ (8.30)
dt V ∂t
3
1 1 1
X
T (~
r) = Tij (~
r ) rbi rbj r ) = o Ei Ej − δij E 2 +
Tij (~ Bi Bj − δij B 2
i,j=1
2 µo 2
(8.31)
(we do not show the fields’ dependence on position for brevity) and where the vector
dot products and divergence of T are given by
3 3 3
X X
~ ·T =
X ∂
a·T =
~ ai Tij rbj T ·~
a= rbi Tij aj ∇ rbj Tij (8.32)
i=1 j=1 i=1
∂ri
Note that Tij is symmetric in its indices. We are not terribly concerned in this course
with the transformation properties of scalars, vectors, and tensors under coordinate
system rotations, so we will not comment further on what a tensor is. Recall we
encountered the quadrupole moment tensor earlier.
Page 552
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
Using the divergence theorem on the first term and moving the time derivative in the
second term outside the integral, we obtain
~ mech
dP
I
d
Z
= da nb(~ r ) − o µo
r ) · T (~ ~ r)
dτ S(~ (8.33)
dt S(V) dt V
This equation states that the rate of change of the mechanical momentum in a
volume V is equal to the integral over the surface of the volume of the stress tensor’s
flux through that surface minus the rate of change of the volume integral of the
Poynting vector.
Let us consider a situation in which the second term vanishes and we are left with the
flux of T over the surface. This justifies the naming of T : it gives the force per unit
area due to the electromagnetic fields, or the stress. Tij is the force per unit area
acting in the ith direction on an area element who normal is in the jth direction. The
diagonal elements are pressures and the off-diagonal forces are shears. More generally,
the force per unit area in the nb1 direction on an area element whose normal is in the
nb2 direction (not necessarily parallel or perpendicular to nb1 ), or vice versa, is
F (~
r , nb1 , nb2 )
= nb1 · T (~
r ) · nb2 (8.34)
A
Page 553
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
We may abstract out mechanical momentum and force densities p r ) and f~(~
~mech (~ r );
i.e., per unit volume expressions:
~ r)
∂ S(~
r ) ≡ ρm (~
~mech (~
p r )~
v (~
r) f~(~ ~ · T (~
r) ≡ ∇ r ) − o µo (8.35)
∂t
We may conclude that these quantities are related locally because of the arbitrariness
of the volume over which we are integrating:
Z
~mech
∂p d
Z ~ mech
dP
Z
dτ = ~mech =
dτ p = dτ f~ (8.36)
V ∂t dt V dt V
~mech (~
∂p r) ~ r)
∂ S(~
=⇒ = f~(~ ~ · T (~
r) = ∇ r ) − o µo (8.37)
∂t ∂t
This is the kind of conservation law we wanted to get to, a local one that relates the
rate of change of the local momentum density to the divergence of the local stress
tensor and the rate of change of the Poynting vector. It can also be viewed as a local
force law, the generalization of Newton’s Second Law.
Page 554
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
As we did with the energy, and motivated by the appearance of a time derivative on
the right side, we may rewrite the above as
d
Z I
~ mech + o µo
P ~
dτ S = da nb · T (8.38)
dt V S(V)
We are thus motivated to define the linear momentum density and linear momentum
of the electromagnetic field as
Z
r) ≡ ~
~field (~
p g (~ ~ r ) = o E
r ) ≡ o µo S(~ ~ (~ ~ r)
r ) × B(~ ~ field =
P dτ ~
g (8.39)
V
Page 555
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
d ~ I
~ field = ∂ ~ · T (~
Pmech + P da nb · T ⇐⇒ [~ r) + ~
pmech (~ r )] = ∇
g (~ r)
dt S(V) ∂t
(8.40)
We thus see that the rate of change of the total (mechanical + field) linear
momentum in a volume is given by the integral of the stress tensor over the surface, or
that the rate of change of the total momentum density at a point is given by the
divergence of the stress tensor at that point. The stress tensor is thus seen to be the
momentum current density in the same way that J~ is the electric current density and
~ is the energy current density (up to a sign): all satisfy local continuity equations.
S
The second equation can also be considered a generalized force law, where now we
consider the rate of change of the momentum of both the particles and the fields, with
~ · T being the “force” that acts on both.
∇
~ field
dP
I
∂~
g (~
r)
~mech = 0 :
p = da nb · T ⇐⇒ ~ · T (~
∇ r) = (8.41)
dt S(V) ∂t
Page 556
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
Page 557
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
We have calculated the magnetic field for a similar configuration when we calculated
the field of the uniformly magnetized sphere, which was
~ ≤ R) = 2 µo M zb
B(r (8.42)
3
~ · rb)b
~ ≥ R) = µo 3 (m r −m
~ 4
B(r with ~ =
m π R 3 M zb (8.43)
4π r3 3
The surface current was K ~ =φ b M sin θ. In the new problem, the surface current is
~ =φ
K b σ ω R sin θ, so we just need to replace M with σ ω R, giving
~ ≤ R) = 2 µo σ ω R zb
B(r (8.44)
3
~ ~ · rb)b
µo 3 (m r −m
~ 4
B(r ≥ R) = with ~ =
m π R 4 σ ω zb (8.45)
4π r3 3
Page 558
Section 8.3 Conservation Laws: The Maxwell Stress Tensor: Conservation of Linear Momentum (Skip)
To calculate the force, we would nominally expect to calculate the flux of the stress
tensor over the hemisphere (the plane at z = 0 for r < R and the hemispherical shell
r = R at z > 0). However, the derivation implies that any volume containing the
matter on which we would like to calculate the force suffices for the calculation. So
let’s do the calculation more easily by setting the surface to be the z = 0 plane. The
force will only be in the z direction by symmetry, so we need only the T3i components.
Moreover, because the plane we want to do the calculation for has a surface normal
only in the z direction, we can restrict to the T33 component:
1 2
T33 = B2 =⇒ T33 (r < R, z = 0) = µo σ 2 ω 2 R 2 (8.46)
2 µo z 9
µo σ 2 ω 2 R 8
T33 (r > R, z = 0) = (8.47)
18 r 6
n = −b
We can do the area integral easily (b z because we want the force on the upper
half space and −b
z is the outward surface normal):
Z 2π Z R Z ∞
Fz = − dφ r dr T33 (r < R, z = 0) + r dr T33 (r > R, z = 0) (8.48)
0 0 R
R2 2 µo σ 2 ω 2 R 8
π
= −2 π µo σ 2 ω 2 R 2 + =− µo σ 2 R 4 ω 2 (8.49)
2 9 72 R 4 4
Page 559
Section 8.4 Conservation Laws: Conservation of Angular Momentum (Skip)
d~
Lmech ~ =~
=N r × F~ = ~
r ×q E~ +~ ~
v ×B (8.50)
dt
d~
Lmech
Z
= dτ ~ ~ + J~ × B
r × ρE ~ (8.51)
dt V
d~
Lmech
~ · T − o µo ∂ ~
Z
= dτ ~r× ∇ ~
r ×S (8.52)
dt V ∂t
Page 560
Section 8.4 Conservation Laws: Conservation of Angular Momentum (Skip)
r× ∇
Let’s manipulate the expression ~ ~ · T : we would obviously like to turn it into a
pure divergence. Using Equation 8.32,
3 3
~ · T (~
X ∂ Tij X ∂ Tij
r× ∇
~ r) = ~r× rbj = rbk × rbj rk (8.53)
i,j=1
∂ri i,j,k=1
∂ri
3 3
!
X ∂ X ∂ rk
= rbk × rbj rk Tij − rbk × rbj Tij (8.54)
i,j,k=1
∂r i i,j,k=1
∂ri
3 3
X ∂ X
= rk rbk × Tij rbj − rbk × rbj Tij δik (8.55)
i,j,k=1
∂ri i,j,k=1
3 3
X ∂ X
=− Tij rbj × rk rbk − rbk × rbj Tkj (8.56)
i,j,k=1
∂ri i,j,k=1
~ · −T (~
=∇ r) × ~
r =∇ ~ · M(~ r) with M(~ r ) = −T (~
r) × ~
r
(8.57)
(we did not show the explicit ~
r dependence for the intermediate steps for brevity)
where the second term in the penultimate line vanishes because it is the product of
rk × rbj ) and symmetric in j and k (Tkj ).
quantities that are antisymmetric in j and k (b
M is the analogue of the stress tensor, but now for torque, which we will call the
torque tensor. We reordered ~r and T to obtain −T × ~ r × T so the
r rather than ~
coordinate index of the divergence matches up with the first coordinate index of M.
Page 561
Section 8.4 Conservation Laws: Conservation of Angular Momentum (Skip)
Explicitly (undoing the reordering of T and ~
r so we get ~r ×, not ×~
r , below, and not
showing explicitly the position dependence for brevity):
3
X 3
X
M = −T × ~
r =− Tij rbi rbj × rk rbk = − rbi rbm mjk Tij rk (8.58)
i,j,k=1 i,j,k,m=1
3
1 1 1
X
= rbm rbi rk mkj o Ej Ei − δji E 2 + Bj Bi − δji B 2 (8.59)
i,j,k,m=1
2 µo 2
3
~+ 1 ~ ~−1 1 2
h i h i
~ E ~ B
X
= o ~r ×E r ×B rbm rbi rk mki o E 2 + B (8.60)
µo 2 i,k,m=1
µo
0 z −y
1 1 1
h i h i
= o ~ ~ E
r ×E ~+ ~ ~ B
r ×B ~+ 2
o E + 2
B −z 0 x
µo 2 µo y −x 0
(8.61)
Aside: In Ph106a, one shows that angular momentum is more rigorously written as an
antisymmetric second-rank (pseudo)tensor, but, because such an object has only 3
independent quantities, it can be reduced to a (pseudo)vector (first-rank tensor) using
cross-product notation. That applies here to both ~ ~ By
r × S.
Lmech and to ~
extrapolation, M may be written as an a completely antisymmetric third-rank
(pseudo)tensor. Since we do not use any of the transformation properties of these
objects under rotations in this course, there is no need to use these higher-rank
objects and so we stick with the less sophisticated vector notation for cross products.
But this concept will return when we consider the relativistic generalization of M
because the reduction to a second-rank tensor is only possible in three dimensions.
Page 562
Section 8.4 Conservation Laws: Conservation of Angular Momentum (Skip)
With our expression in terms of the divergence of M, we may write the analogue of
Equation 8.30 for torque:
d~
Lmech
~ · M − o µo ∂ ~
Z
= dτ ∇ ~
r ×S (8.62)
dt V ∂t
Using the divergence theorem, we may rewrite as we did the force equation
d~
Lmech d
I Z
= r ) · M(~
da nb(~ r ) − o µo dτ ~ ~ r)
r × S(~ (8.63)
dt S(V) dt V
We thus have a relation between the rate of change of mechanical angular momentum
and the flux of the torque tensor M into/out of the volume and the rate of the
change of integral of the funny quantity containing the Poynting vector.
Page 563
Section 8.4 Conservation Laws: Conservation of Angular Momentum (Skip)
Let’s turn this into a differential version. We need to define the mechanical
momentum density and the torque density:
~ ~ · M(~ ∂ ~ r)
r) ≡ ~
`mech (~ r ×p
~mech (~ r × ρm (~
r) = ~ r )~
v (~
r) r) ≡ ∇
~torque (~
n r ) − o µo r × S(~
~
∂t
(8.64)
Then we have
∂~
`mech d d~
Lmech
Z Z Z
dτ = dτ ~
`mech = = ~torque
dτ n (8.65)
V ∂t dt V dt V
∂~
`mech (~
r) ~ · M(~ ∂ ~ r)
=⇒ =n r) = ∇
~torque (~ r ) − o µo r × S(~
~ (8.66)
∂t ∂t
We thus obtain a local conservation law that relates the rate of change of the local
angular momentum density to the divergence of the local torque tensor and the rate of
change of the rate of the change of the funny quantity containing the Poynting vector.
It can also be viewed as a local torque version of Newton’s Second Law.
Page 564
Section 8.4 Conservation Laws: Conservation of Angular Momentum (Skip)
As before, it is natural to define a field angular momentum density and move it to the
left side of the above equations:
Z
~ r ×~
r) = ~
`field (~ r ) = o µo ~
g (~ ~ r ) = o ~
r × S(~ ~ (~
r × (E ~ r ))
r ) × B(~ ~
Lfield = dτ ~
`field
V
(8.67)
With that definition, we obtain
d ~ I ∂ ~
Lmech + ~
Lfield = da nb · M ⇐⇒ r) + ~
`mech (~ ~ · M(~
r) = ∇
`field (~ r)
dt S(V) ∂t
(8.68)
Again, we obtain an integral conservation equation relating the rate of change of the
total angular momentum in a volume to the integral of the torque tensor over the
surface and a local conservation equation relating the rate of change of the total
angular momentum density to the divergence of a current density, here now the
angular momentum current density (which has units of angular momentum per unit
area per unit time). The second equation is a generalized local “torque” equation.
Note the choice of sign for M follows the same convention as for the stress tensor: it
gives a continuity equation with a sign flip but is the correct sign for torque. Be aware
that this sign convention is the opposite of Jackson’s (his Problem 6.9).
Note that field angular momentum is not the same as photon spin or circular
polarization; we will come back to this later when we discuss polarization of EM waves.
Page 565
Section 9
Electromagnetic Waves
9.1 Introduction and Study Guidelines
9.2 Electromagnetic Waves in Vacuum
9.3 Electromagnetic Waves in Perfectly Nonconducting Matter
9.4 Electromagnetic Waves in Conducting Matter
9.5 Electromagnetic Waves in Dispersive Matter
9.6 Introduction and Study Guide: Guided Waves
9.7 Transmission Lines
9.8 Waveguides
Page 566
Section 9.1 Electromagnetic Waves: Introduction and Study Guidelines
We deviate from Griffiths’ ordering of topics because you have seen the wave equation
three times before, in Ph1c, Ph2/12a, and Ph106a, so we do not need to reintroduce
it from scratch. Let’s just launch into it and bring the formalism of waves in as we go.
Page 567
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
~ ~
∇ ~ =0
~ ·E ∇ ~ =0
~ ·B ~ + ∂B = 0
~ ×E
∇ ~ − o µo ∂ E = 0
~ ×B
∇ (9.1)
∂t ∂t
~ = −o µo ∂ ∇ ~ = −o µo ∂ ∇
~ ∇
∇ ~ − ∇2 E
~ ·E ~
~ ×B ~ ∇
∇( ~ − ∇2 B
~ · B) ~
~ ×E
∂t ∂t
2~ 2~
~ = o µo ∂ E
∇2 E ~ = o µo ∂ B
∇2 B
∂t 2 ∂t 2
where ∇ ~ = 0 because there is no charge density. These are copies of the wave
~ ·E
equation.
Section 9.2.1 From Maxwell’s Equations to the Wave Equation Page 568
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
1 ∂2
∇2 f (~
r , t) = f (~
r , t) (9.2)
v 2 ∂t 2
f (~
r , t) = g (w ) with w =~
k ·~
r −ωt (9.3)
3 3 3
X ∂2f X ∂ dg ∂ w X ∂ dg
∇2 f (~
r , t) = 2
= = ki (9.4)
i=1
∂ri i=1
∂ri dw ∂ri i=1
∂ri dw
3 3
d 2g ∂ w X d 2g d 2g
= |~
X
= ki 2
= ki2 2
k|2 (9.5)
i=1
dw ∂ri i=1
dw dw 2
∂2 ∂ dg ∂ w ∂ dg d 2g ∂ w d 2g
2
f (~
r , t) = = (−ω) = −ω 2
= ω2 (9.6)
∂t ∂t dw ∂t ∂t dw dw ∂t dw 2
Section 9.2.1 From Maxwell’s Equations to the Wave Equation Page 569
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
The above condition says we can rewrite the argument, eliminating either ω or |~
k|, as
ω b
w = ±~ r − |~
k ·~ k| v t = ±k · ~
r −vt (9.8)
v
r
δ~
δw = 0 =⇒ ω ±kb · δ~
r − v δt = 0 =⇒ ±kb · =v (9.9)
δt
That is, the surfaces of constant w propagate in space along the direction ±kb at
speed v . This implies that the “shape function” g (w ) propagates at this speed.
Returning to our electromagnetic wave equations, we thus see that these waves in the
√
electric and magnetic fields propagate at speed v = 1/ o µo which is now, by
definition, the speed of light, denoted by c.
The interpretation of ω and k = |~ k| are not clear yet, and in fact they are no longer
strictly necessary (the factor ω/v = k could be absorbed into g (w ) now that we know
ω and k are not independent), but they will become so below when we consider
sinusoidal waves.
Section 9.2.1 From Maxwell’s Equations to the Wave Equation Page 570
Lecture 33:
Electromagnetic Waves II:
Electromagnetic Waves in Vacuum (cont.):
Generic Properties,
Time-Averaging, Complex Notation,
Types of Polarization
Date Revised: 2022/04/11 10:00
Date Given: 2022/04/11
Page 571
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
Section 9.2.2 General Properties of Solutions to the EM Wave Equations Page 572
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
I Transversality
We can rewrite the divergence equations:
~
~ = d E · ∇w
~ = − ωE
~ ·E
0=∇ ~0 sin ~
kbE · E kE · ~
r − ωE t + δ E (9.12)
dw c
d ~
B ω
~ =
~ ·B
0=∇ ~ =− B
· ∇w ~ 0 sin ~
kbB · B kB · ~
r − ωB t + δB (9.13)
dw c
~0 and
For the above equations to hold at all points in space, it is necessary for E
~ 0 to be perpendicular to their respective propagation directions. EM waves are
B
thus transverse waves: the field disturbance is in the direction perpendicular to
propagation.
Section 9.2.2 General Properties of Solutions to the EM Wave Equations Page 573
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
3 3
~=
~ ×E
X ∂ Ek X dEk ∂ w
∇ ijk rbi = ijk rbi
i,j,k=1
∂rj i,j,k=1
dw ∂rj
3 ~ ~
dEk dE ωE b dE
kE ,j = ~
X
= ijk rbi kE × = kE ×
i,j,k=1
dw dw c dw
ωE b
~0 sin ~
=− kE × E kE · ~
r − ωE t + δE (9.14)
c
~
∂E dE~ ∂w dE~
= = −ωE ~0 sin ~
ωE E kE · ~
r − ωE t + δ E (9.15)
∂t dw ∂t dw
~ has similar derivatives. Plugging the above into Faraday’s Law and Ampere’s
B
Law:
ωE b
~0 sin ~
~ 0 sin ~
− kE × E kE · ~
r − ωE t + δE = −ωB B kB · ~
r − ωB t + δ B
c
(9.16)
ωB b
~ 0 sin ~
ω E
~0 sin ~
kB × B kB · ~
r − ωB t + δB = − 2 E kE · ~
r − ωE t + δE
c c
(9.17)
Section 9.2.2 General Properties of Solutions to the EM Wave Equations Page 574
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
~ and B:
In the end, we therefore have the following relation between E ~
~ (~
E ~0 cos ~
r , t) = E k ·~
r −ωt +δ (9.18)
~ r , t) = B
B(~ ~ 0 cos ~k ·~
r −ωt +δ (9.19)
~ 0 = 1 kb × E
B ~0 ⇐⇒ ~0 = −c kb × B
E ~0 (9.20)
c
Section 9.2.2 General Properties of Solutions to the EM Wave Equations Page 575
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
~ 0 · nb2 = 1 E
B2 = B ~0 · nb1 = E1 ~ 0 · nb1 = − 1 E
B1 = B ~0 · nb2 = − E2 (9.21)
c c c c
We thus see that, aside from picking consistent k, b nb1 , and nb2 , there is no connection
between the (E1 , B2 ) pair and the (E2 , B1 ) pair. The waves in the two complementary
polarizations can have different ω and thus different k. They are two completely
independent waves. There is no fixed relationship between the waves in the two
polarizations, and they can get out of phase with each other as they propagate if they
have different ω.
If we consider two waves that have the same ω, then the two waves propagate
together — their relative phase does not change with time or position. But there
remains no condition connecting E1 and E2 , or B1 and B2 , so there is no requirement
that the complementary polarizations have matching amplitude or phase (δ). We will
see later that this independence of the two polarization amplitudes and phases can be
used to generate a diverse set of possible polarizations: linear, circular, and elliptical.
Section 9.2.2 General Properties of Solutions to the EM Wave Equations Page 576
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
~ 0 = 1 kb × E µo
r
H ~0 Z0 = (9.22)
Z0 o
The quantity Z0 ≈ 377 Ω is known as the impedance of free space and has units of
resistance (impedance). We see that H ~ is now only a factor of 377 smaller than E
~.
~ ~
We also recall that H has units of surface current density. This foreshadows the way H
will be related to the surface currents that the electric field drives in
polarizable/magnetizable media and in conductors.
Section 9.2.3 Rewriting EM Waves using the Auxiliary Field Page 577
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
1 1 2 1
o µo 2
u= o E 2 + B = o E 2 + E = o E 2 (9.23)
2 µo 2 µo
2
~= 1 E
S ~ = E kb = c o E 2 kb = c u kb
~ ×B (9.24)
µo c µo
Thus, we see that the energy transported by the electromagnetic wave travels at the
speed of light, just as the wave does. The momentum density vector is
~= u b o E 2 b
~
g = o µo S k= k (9.25)
c c
Note that the energy flux and momentum density differ by a factor of c 2 , not just c
(as we would expect for a relation between energy and momentum) because one is a
flux (energy/area/time) and the other is a density (momentum/volume); the difference
in spatial and temporal units introduces another factor of velocity between the two.
We can also write down the stress tensor. First, consider the special case kb = k zb so
nb1 = xb and nb2 = yb. Consider a wave polarized along xb. Then the fields are
~ = E xb
E ~ = B yb = E yb = õo o E yb
B (9.26)
c
1 2 1 1
T11 = o E2 − E + − B2 = 0 (9.27)
2 µo 2
1 1 1
2
T22 = o − E + B − B2 = 0
2
(9.28)
2 µo 2
1 1 1 2
2
T33 = o − E + − B = −u (9.29)
2 µo 2
T12 = T13 = T23 = 0 (9.30)
=⇒ T = −u zb zb (9.31)
It is reasonable to extrapolate from the above that the generic stress tensor (now
making the time dependence explicit) is
T = − kb kb o E02 cos2 ~k ·~
r −ωt +δ (9.32)
One explanation of the reason for the negative sign is that T is the negative of the
momentum current density.
From the stress tensor, we can calculate the radiation pressure, the force per unit area
that would be applied to an object that absorbs the electromagnetic wave. Recall that
nb1 · T · nb2 gives the force acting in the nb1 direction on a surface element whose normal
is in the nb2 direction. Since T ∝ −kb k, b the force is only nonzero (and positive) in the
kb direction on an area element whose outward normal is in the −kb direction. (Recall
how, in our example of using the stress tensor to calculate the force between the two
spinning charged hemispheres, the surface normal was in the −b z direction for
calculating the force on the hemisphere in the upper half-space.) The radiation
pressure in the kb direction is then
Pressure = kb · T · −kb = E02 cos2 ~
k ·~
r −ωt +δ = u (9.33)
We will see later that, if the wave is not absorbed but reflected, the wave maintains its
amplitude |E~ | but its kb reverses sign, implying that the momentum transfer and thus
the pressure are increased by a factor of 2.
It is not particularly useful to write down the angular momentum density and the
angular momentum tensor for a plane wave. They are position dependent, reflecting
the fact that, if a charged particle absorbs energy and momentum from a plane wave
at some nonzero distance from the origin, it acquires a linear momentum p ~ and thus
an angular momentum ~ r ×p~. The latter carries no information beyond that of the
former. Only if the wave has a nontrivial dependence of E ~ and B~ on position — for
example, E ~ ×B ~ ∝φ b — is the angular momentum of the wave interesting. Such waves
are beyond the scope of our current discussion of plane waves.
The average power per unit area transported by the wave is the intensity
D E 1
~ = c o E 2
I = |S| (9.36)
0
2
Note that the magnitude refers to the vector magnitude. Starting with the next slide,
~ in such a way as
when we consider complex notation for fields, we will always define S
to be real, so the magnitude will continue to refer to vector magnitude even for
complex fields.
For the sake of convenience in manipulation, we will from now on use complex
notation for plane waves,
~ ~ r , t) = k × nb R Ee0 e i(~k·~r −ω t)
h i h i
~ (~
b
E r , t) = nb R Ee0 e i(k·~r −ω t) B(~ (9.37)
c
Ee0 = E0 e iδ (9.38)
where Ee0 is now a complex number into which we have absorbed the phase factor e iδ
~ and
and R means “take the real part.” We will not carry along tildes on the vectors E
~0 because it would be too cumbersome for the notation. It will be clear from context
E
whether we mean the complex or real fields.
~ ~
To calculate quadratic quantities like u, S, g , and T with full space- and
time-dependence requires that one first take the real part and then apply the
previously provided formulae for these quadratic quantities.
where the first expression for each quantity is always valid (even later for conductors
when E ~ and B~ can be out of phase) while the final evaluation is only valid for the
pw vac
plane waves in vacuum we have been considering (indicated by the = notation).
The factors of 1/2 in front come from time-averaging. Note that there may be spatial
dependence remaining in the result if the wave amplitude has a dependence on
position outside of the sinusoidal wave-propagation factor (not possible for plane EM
waves, but it will happen for radiation). It is not necessary to take the real part for u
and certain pieces of T because they are manifestly real.
Types of Polarization
~ at a
So far, we have only discussed linear polarization, wherein the direction of E
particular point in space stays constant over time (aside from sign flips). However, by
combining two orthogonal linear polarizations with appropriate complex coefficients,
one can obtain more “complex” behavior.
The simplest extension is to consider what happens when you add two orthogonal
polarizations of the same amplitude but with a possible phase shift:
Ee0 ~
E r , t) = √ nb1 + nb2 e iδ e i (k·~r −ω t )
~ (~ (9.43)
2
But if δ = ±π/2, then the wave polarized along nb2 is π/2 out of phase with the wave
polarized along nb1 : when the nb1 mode has zero amplitude, the nb2 mode has maximum
amplitude and vice versa. If we take the real part, it is clear what is going on:
h
~ (~
i E0 h π i
R E r , t) = √ nb1 cos ~ r − ω t + δ0 + nb2 cos ~
k ·~ k ·~r − ω t + δ0 ±
2 2
(9.44)
E0 h i
= √ nb1 cos ~ r − ω t + δ0 ∓ nb2 sin ~
k ·~ k ·~
r − ω t + δ0 (9.45)
2
√
The polarization vector maintains an amplitude E0 / 2 but it sweeps around in a circle
with period T = 2 π/ω: this is circular polarization. To understand which direction
the polarization vector rotates, let’s look into the wave (toward −k)b while sitting at a
fixed point in space (fixed ~r ). The time-varying component of the arguments of the
sinusoids is −ω t, and thus, as time evolves positively, the arguments of the sinusoids
evolve negatively. For δ = +π/2, the sign on the second term is negative, the rotation
is counterclockwise, and the wave is called left circularly polarized. Conversely,
δ = −π/2 yields clockwise rotation and is called right circularly polarized. One also
speaks in terms of helicity, in which case one considers the rotation of the polarization
relative to the direction of motion using the right-hand rule. The left circularly
polarized wave has positive helicity because the polarization vector rotates around +kb
according to the right-hand rule (thumb along +k). b The right circularly polarized
wave has negative helicity because it obeys the left-hand rule.
It turns out that elliptically polarized waves are easier to analyze if they are rewritten
in terms of the two helicities (or circular polarizations). That is, if we take as our
polarization basis and field decomposition
1 ~
nb± = √ nb1 ± e i π/2 nb2 E r , t) = Ee+ nb+ + Ee− nb− e i (k·~r −ω t )
~ (~ (9.47)
2
then the parameters of the ellipse traced out by the polarization vector are:
Ee− semiminor axis 1 − r θ
r ei θ = = angle wrt nb1 = (9.48)
Ee+ semimajor axis 1+r 2
where the angle is measured looking into the wave (i.e., looking in the −kb direction).
Section 9.2.7 Types of Polarization Page 587
Section 9.2 Electromagnetic Waves: Electromagnetic Waves in Vacuum
It is interesting to note that a circularly polarized plane wave is, in terms of angular
momentum, no different from a linearly polarized plane wave according to the
definition of the angular momentum density, Equation 8.67:
2
~ = o E0 ~
~ r× E
`field = o ~ ~ ×B r × kb (9.49)
c
One can see that the angular momentum has to do with the relative orientation of the
propagation direction and the position vector, not with the nature of the polarization.
This reflects the fact that, in quantum mechanics, the helicity of the wave becomes
the intrinsic spin angular momentum of the photon, while the quantity calculated
above is the orbital angular momentum of the photon and has to do with the spatial
distribution of the EM wave, in much the same way that orbital angular momentum in
quantum mechanics is determined by the spatial distribution of the wavefunction and
is unassociated with the particle’s spin.
Page 589
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
~ ~
∇ ~ =0
~ ·D ∇ ~ =0
~ ·B ~ + ∂B = 0
~ ×E
∇ ~ − ∂D = 0
~ ×H
∇ (9.50)
∂t ∂t
As noted earlier, we need relations between D ~ and E~ and between H ~ and B~ to make
use of these. If we assume linear, isotropic media ( and µ scalars, not tensors)
~ = E
D ~ ~ = µH
B ~ (9.51)
~ ~
∇ ~ =0
~ ·E ∇ ~ =0
~ ·B ~ + ∂B = 0
~ ×E
∇ ~ − µ ∂E = 0
~ ×B
∇ (9.52)
∂t ∂t
These are the same as our equations in vacuum, leading to the same kinds of waves,
but with the modification
1 c µ
r
v = √ = with n= = index of refraction (9.53)
µ n o µo
Most materials in which waves can propagate (as we will see, p this means materials
√
that do not have high conductivities) have µ ≈ µo , so n ≈ /o = r . Since r > 1
in general (there are very few paraelectric materials that enhance the field rather than
act to decrease it), light generally goes more slowly in dielectrics. Though, creation of
metamaterials in which the effective index of refraction is less than unity (over a
limited frequency range via use of resonant structures) is an area of active research!
~ and E
The relation between B ~ , Equation 9.20, is modified in the obvious manner:
~ r , t) = 1 kb × E
B(~ ~ (~
r , t) (9.54)
v
~ and E
As wepdid for free space, we also have that H ~ are related by an impedance,
Z = µ/, which we now call the wave impedance. With it, we have
~ r , t)
~ r , t) = B(~
H(~ =
1 b ~
k × E (~
r , t) (9.55)
µ Z
Energy density, the Poynting vector, the momentum vector, and the stress tensor take
on unsurprising forms given the above modification:
1 1 ~= 1E
u= E 2 + B2 = E 2 S ~∗ × B
~ = v E 2 kb = v u kb (9.56)
2 µ µ
~= E2 b u
~
g = µS k = kb T = − E 2 kb kb = −u kb kb (9.57)
v v
The time averages for a sinusoidal wave of (real) amplitude E0 are (now including
intensity):
1 ~ = 1 E
D E D
~ = 1 v E 2 kb = v hui kb
~∗ × B
E
hui = E02 S 0 (9.58)
2 2µ 2
D E 1
~ = v E 2 = v hui
I = |S| (9.59)
0
2
D E 1 E2 hui D E 1
h~ ~ =
gi= µ S 0 b
k= kb T = − E02 kb kb = −hui kb kb (9.60)
2 v v 2
Boundary Conditions
Recalling our boundary conditions for linear media (Equations 2.57, 4.23, 6.41, 6.43)
and applying our assumption of no free currents, we have (bn = normal from 1 to 2, sb
= any tangential vector at interface):
~1
B ~2
B
~1 = nb · 2 E
nb · 1 E ~2 ~ 1 = nb · B
nb · B ~2 ~1 = sb · E
sb · E ~2 sb · = sb · (9.61)
µ1 µ2
We will apply these to calculate the reflection and refraction of EM waves at the
interface between different linear media. We will write the magnetic field boundary
condition in terms of H~ because it makes them look like the electric field boundary
conditions, which will be convenient during our discussion of reflection and refraction:
~ 1 = nb · µ2 H
nb · µ1 H ~2 ~ 1 = sb · H
sb · H ~2 (9.62)
We use H ~ instead of B~ because the boundary conditions are more easily written in
~ We have already applied the condition that the frequencies of the three
terms of H.
waves are identical. This is necessary for any boundary conditions connecting them to
be applicable at all time. Then
v1 n2
ki v1 = kr v1 = kt v2 = ω =⇒ kr = ki kt = ki = ki (9.66)
v2 n1
Since k = 2 π/λ, this implies the wavelength differs by n2 /n1 in the two media!
Section 9.3.3 Reflection and Refraction: General Considerations Page 594
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
Each propagation vector makes an angle with the interface normal, nb. We label them
θi , θr , and θt . The projection of a given wavevector parallel to the normal is therefore
|~
k| cos θ while the projection perpendicular to the normal (in the plane of the
interface) is |~ k| sin θ. Since we have argued that these projections into the plane of the
interface are equal, we thus have
ki sin θi = kr sin θr = kt sin θt (9.70)
n2
Now, using our relations kr = ki and kt = n1
ki , we may conclude:
Snell’s Law tells us that, if n2 > n1 , the light ray bends toward the normal, and it
bends away from the normal if n2 < n1 .
Total internal reflection occurs when n1 > n2 and sin θi > n2 /n1 : in this case,
sin θt > 1 and there is no solution for kt . This happens because, when n2 < n1 , then
the magnitude kt < ki , so then the projection ki sin θi can be (but doesn’t have to be,
depending on the value of sin θi ) too large for kt to match. We will study this case in
more detail in a homework problem.
Note that none of these results depending on knowing anything about Maxwell’s
Equations or boundary conditions: these are generic properties of any waves, which is
why they hold for sound waves, phonons, etc. They result only from matching time
and space dependences at the boundary.
Section 9.3.3 Reflection and Refraction: General Considerations Page 596
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
Electromagnetic Waves
Electromagnetic in inPerfect
Waves Matter
Perfect Matter (cont.)
(cont.)
Reflected and Transmitted Field Relations
Now, let’s apply our electromagnetic boundary conditions to figure out how the
Now, let’s apply our electromagnetic
amplitudes
Electromagnetic
Each and boundary
energies
Waves
propagation arein
vector conditions
related.
Perfect
makes an angle mayto
We with
Matter
the figure
drop all the
(cont.)
interface outexponential
normal.how the
We label factors
them θi , because
we have
Electromagnetic
amplitudes and energies established
θr , and θt . The
Waves
are related. We that
in theydrop
projection
Perfect
may ofare identical
a given
Matter
all the atexponential
wavevector the
(cont.) interface.
parallel We isneed
to the normal
factors to define two
therefore
because
�
|k| cos θi while the projection perpendicular to the normal — i.e., in the plane of the
new vectors:
we have established that
Now,they are
applyidentical
let’sinterface — is |�k|
our sin θiat theweinterface.
. Since
electromagnetic have We
argued that
boundary need
these to
projections
conditions define two
into theout
to figure plane of the
how
new unit vectors:
Now, let’s apply the
amplitudes
interface
ourand are equal, we thus
electromagnetic
energies �k × areb
have
n Electromagnetic
boundary
related.
�k ×We n conditions
b may�kdrop×b nWaves
to
all figure in
outPerfect
how the
the exponential Matter
factors (con
because
i r t
amplitudes we haveand energies
established b
uare=that
related. =
We
are�may drop = θall=thethe exponential wbWe b
=factors
n ×b utobecause (1.58)
�i ×they n| ki sin
b
identical
|krθat =b k|r sin at|r�kt ×
interface.
kt sin
n| θWe
b
need define two
(1.55)
vectors: that they|kare
i× nthe t
we havenew established identical interface. need to define two
Now, let’sn2apply our electromagnetic boundary condition
new vectors: b Now, usingb our relations kr = ki and kt = n ki , we may conclude:
ub ∝ nb × kbi = nIt b× kr =matter
doesn’t
nb × kwhich t b×we
�ki k
amplitudes1 and energies are related. We may drop all th
�k u �because
b
n usewe
b tor ×define
have b
n established kt × b n thatwe argued earlier that all of them
they
b = ub × nb
w have the same kprojection � b
u =n � in � the= plane �kof =b
×the n �interface,θiwhich = θr w b is=areb
n×
the
identical
b
u (1.56)
piece
at the
that matters
interfa
(1.58)
i ×b
nr (9.73)
klaw ×ofbnreflection:
b
u = cross |k=i × b | |�k=
new r × nt |
b
vectors: | kt × b n |wb =k, b
n ×sin b
u (1.58) to the
for the above product. b
u is also normal to n1 sin θi �
all the = nso2 itθtis perpendicular
|�klaw
i × of refraction
b
n | |�k r
(Snell’s Law):
× b
n | | �k t × bn | (1.57)
plane in which the propagation vectors lie. w b is then the �ki × obvious � third direction � and it
It doesn’t matter which b we
k �k results
usein. to Thus,
defineonour b
u knowing
because b
n Maxwell’s ×b
kr can nthat b
allktof×them
nas
= we conditions
argued earlier
defines Note
the that none
plane that ofthe
these live depending boundaryanything about be=written
Equations bor boundary conditions: these are generic
b
uproperties of any =waves, which is
It doesn’t havematter
the samewhichprojection
k we use in to the define plane b
u because we argued
of the interface, earlier
|�kwhich n| is that
the all
|�kr piecenof them
| that |�ktmatters
why they hold for sound waves, phonons,»etc. – i ×b ×b ×b n|
have theforsame the projection
“
above crossin”product.
the planeb u of the interface,
is also normal b
uargued “ theis�k,
towhich
all ”it is »perpendicular
thesopiece –
that matters
b
u of � to the
It doesn’t matter which b k we
n · �1in � use
E0,i product.
+ theto
� define
E0,r propagation
= b
n ·also u because
�
�2 Enormal we �
the ·k, Eso � earlier
it+is � that all · Ethe
for the above cross b
u is vectorstolie. allw bEobvious = third direction
perpendicular to 0,t (1.59)
b
plane which 0,r
bb is then 0,ithe 0,r and it
b
It doesn’t w
matter which k wecontaining
use to wb
define b
u because we
them lie in the plane
sameinplane which with
defines nb, so thethe
propagation direction
�k livelie.
vectors in. w normal
b is then–ourto the the plane
obvious third direction and it
“the plane that �” have
Thus,
» theb
u same 1
boundary
“
projection
conditions
in ” plane
the » canof
b
u
–bethe
written
1
as
interface, w
defines the b
them and nb is independent plane
n · ofB that
� 0,i + the
which B k live
� 0,rone = is n in.
b ·used. Thus,(The
� 0,r
B our boundary equality
· conditions
�of
B the
+ �
B can=be written
three · as B �
» –cross 0,i 0,r b
“ ” for thew babove µ1 “then, u
product. ”u is »also wb –
normal
µ2
to0,tall the
cross-products follows“from b Equations
n · Section
�1 E �1.20,i”+ E � 0,r = b9.68 and
· �2 E
nElectromagnetic
9.69.)
»
� 0,rplane
Waves: in
– Clearly,
b
u
which
“ the �propagation
·Waves
E in” +E » is
� Matter perpendicular
– b
u
vectors lie. ·E �Page
b is
w then th
u = �w (1.59)
b 26 (1.60)
u Electromagnetic
b b 0,i Perfect 0,r b 0,t
to this plane in which
n · �1 n
b � 0,iand
E +E �the
0,r propagation
=b n · �2 E � 0,r vectors
defineslie. · –w
the �b
w
Eplane �that
0,iis+then
E 0,r the = �kobvious
the live in. · Eb third
Thus, our boundary
(1.59)
b»
w b » 0,t –
w
b
“ ” 1 � “ ” 1 �
direction, and it and “ nb define the
� 0,i ”+ plane
� 0,r of = that �the
» k –livebuin. �»0,r those – b u
1b � »· µb – 0,t“
b
where n ·the
b B stacking B b
ub nand·B wb is
0,r u just meant
b 1b““ �· to B” 0,i”+ that
indicate B u = equations
b apply
B
b
n · with
B� 0,i either
+B � 0,rb u on=b ·B
nboth� 0,rsides or w bw · �1w �sides.
B0,i+ µ1+� B� 0,r == � 0,r · w B0,t u2 · E � 0,i +
b on n ·both
b
µ1
E 0,i E0,r b
n · �2w bE
µ2 b
w (1.60)
“ ” » –
(1.60) “
Section 1.2 Electromagnetic Waves: Electromagnetic � 0,i + B � 0,rWaves � 0,r Matter b u 1 � Pa
where the stacking of b u and w b is just b
n ·meant B to indicate =b nin ·Perfect
that B those equations · applyB 0,i +
wb µ1
where the with stacking
either b uofon b
u both
and w bsides
is just or w meant
b on both to indicate
sides. that those equations apply
Section 9.3.4 Reflected and Transmitted Field Relations Page 597
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
where the stacking of ub and wb is just meant to indicate that those equations apply
with either ub on both sides or w
b on both sides.
Now, we must consider two cases for the polarization of the incoming wave: in
(parallel to) the plane of incidence or perpendicular to the plane of incidence, also
termed transverse magnetic (TM) and transverse electric (TE) for obvious reasons:
perpendicular to the
~0,i · nb ~ 0,i · ub = 0
plane of incidence E =0 H (9.77)
w
or transverse electric (TE)
b
w
b: Ee0,i cos θi + Ee0,r cos θr = Ee0,t cos θt (9.85)
1 e 1 e
ub : E0,i − Ee0,r = E0,t (9.86)
Z1 Z2
Since the incident wave amplitude must be allowed to be arbitrary, we expect to only
be able to determine the ratios Eer ,0 /Eei,0 and Eet,0 /Eei,0 . Only two of the above
equations can therefore be independent. (One can easily see that the nb and ub
equations are equivalent via Snell’s Law.) Picking the last two because they are
easiest to work with, we obtain Fresnel’s Equations in (parallel to) the plane of
incidence (for TM waves):
cos θt Z1
α= β= (9.87)
cos θi Z2
Fresnel’s
Equations
Ee0,r α−β Ee0,t 2
= = in (parallel to) (9.88)
Ee0,i α+β Ee0,i α+β the plane of
incidence (TM)
µ1 e µ2 e
nb : E0,i sin θi + Ee0,r sin θr = E0,t sin θt (9.95)
Z1 Z2
1 e 1 e
w
b: E0,i cos θi − Ee0,r cos θr = E0,t cos θt (9.96)
Z1 Z2
ub : Ee0,i + Ee0,r = Ee0,t (9.97)
Again, we only need two of the equations, so we use the latter two because the nb
equation can be reduced to the ub equation via Snell’s Law. We solve to obtain
Fresnel’s Equations perpendicular to the plane of incidence (TE):
Fresnel’s
Equations
Ee0,r 1 − αβ Ee0,t 2
= = perpendicular (9.98)
Ee0,i 1 + αβ Ee0,i 1 + αβ to the plane of
incidence (TE)
Page 603
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 604
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
We know that, for a given pair of media, either sin θt / sin θi < 1 or
sin θt / sin θi > 1 is true for all angles because this ratio is set by Snell’s Law,
sin θt / sin θi = n1 /n2 . Since sin and cos are monotonic over the first quadrant,
we can conclude that α = cos θt / cos θi is also either smaller than or greater
than 1 for all angles, with the case being determined by n1 /n2 :
n1 sin θt cos θt
>1 ⇐⇒ >1 ⇐⇒ <1 ⇐⇒ α<1 (9.99)
n2 sin θi cos θi
n1 sin θt cos θt
<1 ⇐⇒ <1 ⇐⇒ >1 ⇐⇒ α>1 (9.100)
n2 sin θi cos θi
However, β = Z1 /Z2 = (µ1 /µ2 )(n2 /n1 ), so the size of n2 /n1 relative to unity
does not completely determine the size of β relative to unity. No generic
statement can be made about the relative size of α, β, and 1/β.
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 605
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
That said, we can make conditional statements that depend on whether we are
considering the parallel (TM) or perpendicular (TE) incidence cases and
whether β is smaller or larger than unity. We break this up into two steps: what
is the sign of the reflection at normal incidence, and whether that sign changes
as the angle of incidence changes:
I Sign of reflected wave at normal incidence, θi = θt = 1 (parallel (TM)
and perpendicular (TE) incidence degenerate at this angle)
The sign of the denominator of the expression for the reflected amplitude
is always positive, and α = 1 because θi = θt = 1 at normal incidence, so
the sign of the reflection at normal incidence is set by the sign of the
numerator 1 − β:
!
µ1 n2 Ee0,r
if =β>1 =⇒ sign <0 (9.101)
µ2 n1 Ee0,i θi =0
!
µ1 n2 Ee0,r
if =β<1 =⇒ sign >0 (9.102)
µ2 n1 Ee0,i θi =0
I Does the sign of the reflected wave change with angle of incidence?
We can answer this question by determining whether there is an angle at
which zero reflection occurs. If so, the sign of the reflected wave will
change from it sign at normal incidence to the opposite sign. Whether
there is such an angle depends on the relative sizes of α and β. We will
call this angle Brewster’s Angle, though that nomenclature generally only
applies for the case µ1 = µ2 = µo , which we will consider shortly.
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 606
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 607
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
n12
1− n22
sin2 θB 1
1
−1
β2
= ⇐⇒ sin2 θB = (9.106)
1 − sin2 θB β2 1
−
n12
β2 n22
This is basically just the converse of the TM case, so the same set
of requirements for a solution for θB implies
µ1 n2 n1 n1 µ1
if =β>1 =⇒ need > 1, < (9.107)
µ2 n1 n2 n2 µ2
µ1 n2 n1 n1 µ1
if =β<1 =⇒ need < 1, > (9.108)
µ2 n1 n2 n2 µ2
Note how the pairing of the two n1 /n2 conditions changes between the
TM and TE cases! The above behavior will be seen in the plots at the
end of this section.
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 608
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
n1 n2 cos θt
>1 ⇐⇒ β= <1 and α= <1 (9.109)
n2 n1 cos θi
n1 n2 cos θt
<1 ⇐⇒ β= >1 and α= >1 (9.110)
n2 n1 cos θi
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 609
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
n12
1− n22
sin2 θB n22 n22
= ⇐⇒ sin2 θB = (9.111)
1 − sin2 θB n12 n12 + n22
Brewster’s
n2 Z1
⇐⇒ tan θB = =β= Angle for (9.112)
n1 Z2 µ1 = µ2
Note that 0 < θB < π/4 for n2 < n1 and π/4 < θB < π/2 for n2 > n1 .
Therefore, we may summarize this case as (using 1 − β = (n1 − n2 )/n1 ):
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 610
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 611
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 612
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
I Relation between sign of reflected wave and possibility of total internal reflection
Both sets of equations imply that total internal reflection occurs only if n2 < n1 ,
which also implies that the sign of the reflected wave at normal incidence is
equal to the sign of the incident wave. Thus, the two conditions are equivalent:
no sign flip at normal incidence ⇐⇒ total internal reflection possible
Section 9.3.5 Implications for Signs and Magnitudes of Fields Page 613
Section 9.3 Electromagnetic Waves: Electromagnetic Waves in Perfectly Nonconducting Matter
~j | = 1 j vj E 2 cos θj = 1 c E 2 cos θj
D E
I j = |S j (9.118)
2 2 Zj j
We can calculate from this the reflected and transmitted energy or power ratios:
!2 2
Ee0,r α−β
Ir
R= = = parallel (9.119)
Ii Ee0,i α+β
1 − αβ 2
= perpendicular (9.120)
1 + αβ
!2 2
Z1 Ee0,t cos θt 2
It
T = = = αβ parallel (9.121)
Ii Z2 Ee0,i cos θi α+β
2
2
= αβ perpendicular (9.122)
1 + αβ
Normal Incidence
Let’s summarize the results for normal incidence, θi = θr = θt = 0, where things
simplify substantially. This implies α = 1, for which the parallel and perpendicular
cases are equivalent, yielding:
2
Z1 Z1
Ee0,r 1−β 1− Z2
1−β
2 1− Z2
= = Z1
R= = Z1
(9.123)
Ee0,i 1+β 1+ 1+β 1+
Z2 Z2
2
2
Ee0,t 2 2 2 Z1 2
= = T =β = (9.124)
Ee0,i 1+β 1 + ZZ1 1+β Z2 1 + Z 1
Z2
2
We will see that we get similar equations for transmission lines and waveguides. For
the case µ1 = µ2 = µ0 , β = Z1 /Z2 reduces to n2 /n1 = v1 /v2 :
Ee0,r n1 − n2 v2 − v1 n2 − n1 2 v2 − v1 2
= = R= = (9.125)
Ee0,i n1 + n2 v2 + v1 n2 + n1 v2 + v1
2 2
Ee0,t 2 2 n2 2 v1 2
= = T = = (9.126)
Ee0,i n1 + n2 v2 + v1 n1 n1 + n2 v2 v2 + v1
One recovers the results for a wave on a string you learned about in Ph2/12.
Typical Behavior
We first show the simpler case µ1 /µ2 = 1 and then consider the more general case µ1 /µ2 6= 1.
Notice the
nonmonotonic
n2/n1 = 1.4, µ2/µ1 = 1.0, β = 1.4 n2/n1 = 0.7, µ2/µ1 = 1.0, β = 0.7
behavior in R and 2 2
T near θB for the par. (TM), refl.
par. (TM), transm.
par. (TM), refl.
par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
parallel incidence perp. (TE), transm. perp. (TE), transm.
case: this is
necessary to yield 1 1
the zero in the
Er/Ei, Et/Ei
Er/Ei, Et/Ei
reflected electric
field and in R at
θB . The same 0 0
behavior is not
present in the θTIR
transmitted field θBTM θBTM
-1 -1
amplitude; T = 1 1.0 1.0
comes about
because of the α 0.8 0.8
factor in T .
Notice also how
0.6 0.6
the transmitted
R, T
R, T
field amplitude
appears to diverge 0.4 0.4
but the
transmitted energy 0.2 0.2
vanishes at θTIR
n
θTIR = sin−1 n2 ; θTM
B θTM
B
1 0.0 0.0
again, the α factor 0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
in T explains this.
n2/n1 = 1.4, µ2/µ1 = 1.4, β = 1.0 n2/n1 = 1.4, µ2/µ1 = 0.7, β = 2.0
2 2
par. (TM), refl. par. (TM), refl.
par. (TM), transm. par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
perp. (TE), transm. perp. (TE), transm.
1 1
Er/Ei, Et/Ei
Er/Ei, Et/Ei
0 0
θTE
B θTM
B
-1 -1
1.0 1.0
0.8 0.8
0.6 0.6
R, T
R, T
0.4 0.4
0.2 0.2
θBTE θBTM
0.0 0.0
0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
n2/n1 = 1.4, µ2/µ1 = 1.7, β = 0.8 n2/n1 = 1.4, µ2/µ1 = 0.6, β = 2.4
2 2
par. (TM), refl. par. (TM), refl.
par. (TM), transm. par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
perp. (TE), transm. perp. (TE), transm.
1 1
Er/Ei, Et/Ei
Er/Ei, Et/Ei
0 0
θTE
B θTM
B
-1 -1
1.0 1.0
0.8 0.8
0.6 0.6
R, T
R, T
0.4 0.4
0.2 0.2
θBTE θBTM
0.0 0.0
0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
n2/n1 = 1.4, µ2/µ1 = 1.2, β = 1.2 n2/n1 = 1.4, µ2/µ1 = 0.8, β = 1.7
2 2
par. (TM), refl. par. (TM), refl.
par. (TM), transm. par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
perp. (TE), transm. perp. (TE), transm.
1 1
Er/Ei, Et/Ei
Er/Ei, Et/Ei
0 0
θTM
B θTM
B
-1 -1
1.0 1.0
0.8 0.8
0.6 0.6
R, T
R, T
0.4 0.4
0.2 0.2
θBTM θBTM
0.0 0.0
0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
n2/n1 = 0.7, µ2/µ1 = 1.4, β = 0.5 n2/n1 = 0.7, µ2/µ1 = 0.7, β = 1.0
2 2
par. (TM), refl. par. (TM), refl.
par. (TM), transm. par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
perp. (TE), transm. perp. (TE), transm.
1 1
Er/Ei, Et/Ei
Er/Ei, Et/Ei
0 0
θTIR θTIR
θTM
B θTM
B
-1 -1
1.0 1.0
0.8 0.8
0.6 0.6
R, T
R, T
0.4 0.4
0.2 0.2
θTIR θTIR
θBTM θBTM
0.0 0.0
0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
n2/n1 = 0.7, µ2/µ1 = 1.7, β = 0.4 n2/n1 = 0.7, µ2/µ1 = 0.6, β = 1.2
2 2
par. (TM), refl. par. (TM), refl.
par. (TM), transm. par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
perp. (TE), transm. perp. (TE), transm.
1 1
Er/Ei, Et/Ei
Er/Ei, Et/Ei
0 0
θTIR θTIR
θTM
B θTE
B
-1 -1
1.0 1.0
0.8 0.8
0.6 0.6
R, T
R, T
0.4 0.4
0.2 0.2
θTIR θTIR
θBTM θBTE
0.0 0.0
0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
n2/n1 = 0.7, µ2/µ1 = 1.2, β = 0.6 n2/n1 = 0.7, µ2/µ1 = 0.8, β = 0.8
2 2
par. (TM), refl. par. (TM), refl.
par. (TM), transm. par. (TM), transm.
perp. (TE), refl. perp. (TE), refl.
perp. (TE), transm. perp. (TE), transm.
1 1
Er/Ei, Et/Ei
Er/Ei, Et/Ei
0 0
θTIR θTIR
θTM
B θTM
B
-1 -1
1.0 1.0
0.8 0.8
0.6 0.6
R, T
R, T
0.4 0.4
0.2 0.2
θTIR θTIR
θBTM θBTM
0.0 0.0
0 20 40 60 80 0 20 40 60 80
angle of incidence [degrees] angle of incidence [degrees]
Page 623
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
Do we need to worry about free charge? No. If we combine the continuity equation,
~ · J~f = −∂ρf /∂t with Gauss’s Law, we obtain
∇
∂ ρf
~ ·E~ = − σ ρf
= −σ ∇ (9.128)
∂t
The solution to this equation (with σ/ > 0) is a decaying exponential in time,
ρf (t) = ρf (t = 0) exp(−t/τ ) with τ = /σ. Thus, any free charge decays away. The
time constant for the decay as compared to the relevant field variation timescale (the
wave oscillation period T = 2 π/ω) is a measure of how good the conductor is at the
frequencies of interest: for a good conductor, one must have τ T (or, equivalently,
ω τ 1, neglecting the 2 π). Regardless, the appearance of any free charge is
transient, while we will see the wave, though decaying in space, is stable in time.
Section 9.4.1 Maxwell’s Equations and the Wave Equation for Conductors; Plane-Wave Solutions Page 624
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
~ ~
∇ ~ =0
~ ·E ~ =0
~ ·B
∇ ~ + ∂B = 0
~ ×E
∇ ~ ×B
∇ ~ − µ ∂E = 0
~ − µσE
∂t ∂t
(9.129)
∂2E~ ~
∂E ∂2B~ ~
∂B
~ = µ
∇2 E +σµ ~ = µ
∇2 B +σµ (9.130)
∂t 2 ∂t ∂t 2 ∂t
As we did for nonconducting matter, we assume plane wave solutions because any
solution can be built from them by linearity:
~ (~
E ~0 e i (~k·~r −ω t )
r , t) = E B(~ ~ 0 e i (~k·~r −ω t )
~ r , t) = B (9.131)
We dispense with repeating the proof that ~ k and ω are the same for E~ and B~ (the
proof is similar in form) and simply assume it and check it below. We determine later
the relation between E ~0 and B
~ 0 (including any possible δ, which they absorb when
using complex notation).
Section 9.4.1 Maxwell’s Equations and the Wave Equation for Conductors; Plane-Wave Solutions Page 625
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
~
k ·~
k = µ ω2 + i σ µ ω (9.132)
~ (~
E r , t) = E
b r /δ ) i (k k·~
~0 e −(k·~ e
b r −ω t ) ~ 0 e −(k·~
~ r , t) = B
B(~
b r /δ ) i (k k·~
e
b r −ω t )
(9.135)
Section 9.4.1 Maxwell’s Equations and the Wave Equation for Conductors; Plane-Wave Solutions Page 626
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
Let’s consider two limits, the poor conductor limit and the good conductor limit:
I Poor conductor limit
This is the limit ω τ 1. (Note that we have to wait many wave periods after
the appearance of the wave in order for the free charge to decay away because
ω τ 1 (τ T /2 π), but we do reach ρf = 0 if we wait this long.) The second
term under the inner square root is small and one can Taylor expand to obtain
1/2
ωτ 1 kµ 1 1
k −→ √ 1+ + 1 ≈ kµ (9.136)
2 2 ω2 τ 2
1/2 r
1 1 kµ Zµ σ
ωτ 1 ω σ µ
κ −→ √ 1+ − 1 ≈ = = (9.137)
vµ 2 2 ω2 τ 2 2ωτ 2 2
ωτ 1 2 2 2ρ
r
δ −→ = = (9.138)
σ µ Zµ σ Zµ
In this case, the wave propagation speed and wavelength are the same as in a
nonconducting medium. The form for the skin depth is particularly interesting
because it involves the ratio of the resistivity and the wave impedance, which
connect to the Ohmic and displacement currents, respectively. We can write the
skin depth another way, now using the wavelength in the medium:
ωτ 1 2ωτ 2τ
δ −→ = λµ λµ (9.139)
kµ T
2 τ /T 1 counts the number of wavelengths over which the amplitude decays:
in a poor conductor, the decay of the wave happens over many wavelengths.
Section 9.4.1 Maxwell’s Equations and the Wave Equation for Conductors; Plane-Wave Solutions Page 627
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
" r #1/2 s r
ωτ 1 ω 1 1 ω µσω
k, κ −→ = = (9.140)
vµ 2 ω2 τ 2 2 τ
2 vµ 2
s
ωτ 1 2
δ −→ (9.141)
µσω
There is no propagating wave because the decay constant and the propagation
constant are the same; put another way, δ = 1/κ = 1/k = λ/2 π, indicating the
wave decays in about 1/6 of a wavelength!
Notice how becomes irrelevant for a good conductor. We will see that Zµ is
effectively replaced by Zσ ∼ 1/σ δ so δ ∼ 1/Zσ σ, a form similar to the poor
conductor case. However, Zσ Zµ for a good conductor, so H ∼ E /Z is
much larger relative to E than it is for a poor conductor: in a good conductor,
the Ohmic currents and magnetic field dominate over the displacement current.
We will return to this aspect later.
Section 9.4.1 Maxwell’s Equations and the Wave Equation for Conductors; Plane-Wave Solutions Page 628
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
General Results
Coming back to our general plane-wave solution, we write (with nb defining the
~)
polarization of E
~ equation to find
~ ×E
and we use the ∇
~ r , t) = k + i κ Ee0 kb × nb e −(k·~
B(~
b r /δ ) i (k k·~
e
b r −ω t )
(9.143)
ω
We may write
"r #1/2
p σ 2
k + i κ = K ei φ K = k 2 + κ2 = kµ 1+ (9.144)
ω
q 1/2
σ 2
κ 1+ ω
− 1
tan φ = = q (9.145)
k
σ 2
1+ ω
+1
Notice that we are writing 1/ω τ as σ/ ω so the physical origin of the “good
conductor/poor conductor” distinction is clear.
With the above definition, we can more simply write the relation between the phases
~ and B:
and amplitudes of E ~
Be0 k +iκ K ei φ ei φ
σ 2 1/4
= = = 1+ (9.146)
Ee0 ω kµ vµ vµ ω
~ and E
It now also makes sense to introduce the relation between H ~ using H
~ = B/µ:
~
~ r , t)
~ r , t) = B(~
H(~ =
k +iκ e b
E0 k × nb e −(k·~r /δ) e i (k k·~r −ω t )
b b
(9.147)
µ ωµ
He0 1 Be0 ei φ
σ 2 1/4 ei φ
σ 2 1/4
= = 1+ = 1+ (9.148)
Ee0 µ Ee0 vµ µ ω Zµ ω
Thus, we see that there is now a phase shift between electric and magnetic fields, with
the magnetic field lagging behind the electric field. The amplitudes are related in a
somewhat complicated manner that depends on the wave velocity as before but now
with a modification factor that depends on the conductivity. The auxiliary field
behaves in the same way except that it is now the wave impedance with the same
modification factor that relates the electric and auxiliary fields.
1 σ Be0 1 He0 1
poor conductor : = 1 =⇒ → , → (9.149)
ωτ ω Ee0 vµ Ee0 Zµ
e iπ/4
r
1 σ B
e0 1 1
good conductor : = 1 =⇒ → (9.150)
ωτ ω E0
e vµ ωτ vµ
e iπ/4
r
He0 1 1
→ (9.151)
Ee0 Zµ ωτ Zµ
In a poor conductor, the relation approaches the nonconducting case (as we expect).
In contrast, in a good conductor, the magnetic and auxiliary fields are enhanced
(because there are free currents to enhance the magnetic field but no free charges to
enhance the electric field) and they lag the electric field by exactly π/4. Why π/4 and
not π/2? The two current densities are Jd = i ωE and Jf = σ E . So
|Jd | = ω/σ|Jf | |Jf | in the good conductor limit, and thus the imaginary piece
i σ ω dominates the right side of the wave equations, Equations 9.130, and therefore
~
k ·~k ∝ i = e i π/2 . But it is (~
k ·~
k)1/2 /ω = (k + i κ)/ω that appears as the coefficient
relating E to B (from Faraday’s Law), so then (k + i κ)/ω ∝ e i π/4 hence the π/4
phase shift.
In a poor conductor, the ohmic term is negligible and so we only get the piece due to
Jd = ∂E /∂t. It is in phase in spite of the ∂/∂t because ∂E /∂t is related by Ampere’s
Law to spatial derivatives of B, which introduce a canceling factor of i = e i π/2 . This
causes ~
k ·~
k to be real and positive, so no additional phase is introduced.
Section 9.4.2 General Results Page 631
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
For the good conductor case, we can rewrite the relation between the electric and
auxiliary fields in a more suggestive form if we recognize the skin depth is present:
e iπ/4
r r
H 1 σ
e0 r
= = e iπ/4 (9.152)
Ee0 Zµ ωτ µ ω
s
σ2 2 1+i 1+i 1
= e iπ/4 = σδ ≡ (9.153)
2 σµω 2 2 Zσ
where Zσ = 1/σ δ. Why does the ratio have this form? In a good conductor, the
fields, and thus the ohmic current density, decay to zero over the distance δ — the
bulk of the current flows in a layer of thickness δ. If we imagine integrating this
current over the skin depth to get an effective surface current density, we would
multiply J ≈ σ E by δ to get K ≈ σ δE and thus K /E ≈ σδ = 1/Zσ . Recall that H
has the same units as K , so it makes perfect sense that H ∼ K and the ratio
H/E ∼ K /E ≈ 1/Zσ . The prefactor is only present to take care of the phase issue,
and it has (almost) unity magnitude. In fact, we can define the surface impedance of
the conductor by
2 H
e0 1
Zs = Zσ = (1 − i) Zσ = (9.154)
1+i E0
e Z s
We will come back to this in the context of waveguides with good but not perfectly
conducting walls.
Section 9.4.3 Energy Density, Poynting Vector, and Intensity Page 633
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
The poor conductor expressions match our expressions for a wave in a nonconducting
medium except that they decay with depth as one would expect.
Section 9.4.3 Energy Density, Poynting Vector, and Intensity Page 634
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
From the above, we can calculate the total power dissipated in the conductive medium
per unit area. We just need to know how much power flows in per unit area at the
kb · ~
r = 0 plane: the wave is propagating along k,
b and there is no power flowing
backward along k. b This power flow at kb · ~
r = 0 is I(kb · ~
r = 0) because the plane we
are considering has surface normal k.
b Therefore, the power dissipated per unit area is
dPdiss 1 1 δ
= I(kb · ~
r = 0) = Je0 Ee0 δ = Je0 Ee0
dA 4 2 2
which is as one would expect: Je0 Ee0 /2 is the Joule power dissipation per unit
volume (including time-averaging), and δ/2 is the energy decay length and so provides
the thickness over which that power is dissipated to give a power dissipation per unit
area.
Section 9.4.3 Energy Density, Poynting Vector, and Intensity Page 635
Lecture 37:
Electromagnetic Waves VI:
Electromagnetic Waves in Conducting Matter (cont.):
Reflection at a Conductor at Normal Incidence
Electromagnetic Waves in Dispersive Matter
Date Revised: 2022/04/20 05:00
Date Given: 2022/04/20
Page 636
Section 9.4 Electromagnetic Waves: Electromagnetic Waves in Conducting Matter
Because free charges and currents are now possible, our boundary conditions differ
from those for an interface between two nonconducting media:
h i h i
~1 − 2 E
nb · 1 E ~2 = σf nb · B~1 − B
~2 = 0 (9.168)
" #
h i ~1
B ~2
B
sb · E~1 − E
~2 = 0 sb · − = K~ f × nb · sb (9.169)
µ1 µ2
The general case is very complicated to analyze; a measure of the difficulty is that not
even Jackson tries to do it! Let’s consider only the case of normal incidence. Since
~ = 0, we have σf = 0 and the situation is similar to analyzing the case of
nb · E
nonconducting media.
~i (~ ~ i (~ 1 e
E b e i(ki nb·~r −ω t)
r , t) = Ee0,i w H r , t) = E0,i ub e i(ki nb·~r −ω t) (9.170)
Zµ,1
~r (~ ~ r (~ 1 e
E b e i(−ki nb·~r −ω t)
r , t) = Ee0,r w H r , t) = − E0,r ub e i(−ki nb·~r −ω t) (9.171)
Zµ,1
~t (~ ~ t (~ kt + i κt e
E b e i(kt nb·~r −ω t)
r , t) = Ee0,t w H r , t) = E0,t ub e i(kt nb·~r −ω t) (9.172)
µ2 ω
× e −(κt nb·~r ) × e −(κt nb·~r )
p
where Zµ,1 = µ1 /1 . Inserting these expressions into the boundary condition
equations, evaluated with nb = zb and at z = 0, we obtain
Solving, we obtain
! !
Ee0,r 1 − βe Ee0,t 2
= = (9.175)
Ee0,i 1 + βe Ee0,i 1 + βe
Zµ,1 Zµ,1 Kt e i φ Kt e i φt
with βe ≡ (kt + i κt ) = =β (9.176)
µ2 ω Zµ,2 ω/vµ,2 kµ,2
p p √
where Zµ,1 = µ1 /1 , Zµ,2 = µ2 /2 , vµ,2 = 1/ 2 µ2 , kµ,2 = ω/vµ,2 ,
i φ
Kt e t = kt + i κt , and β = µ1 n2 /µ2 n1 = Zµ,1 /Zµ,2 was defined in Equation 9.87.
The reflected power is
Ee 2 1 − βe 2
0,r
R= = (9.177)
Ee0,i 1 + βe
The expression for the transmitted power, if calculated directly from the Poynting
vector in the conductor, is rather complicated. Instead, one can just use T = 1 − R.
These expressions are very similar in spirit to the normal incidence equations for
nonconducting media, Equations 9.123 and 9.124, except that β is replaced by β. e Of
course, this causes the reflected and transmitted waves to acquire nontrivial phase
shifts relative to the incident wave.
The poor conductor limit is the normal incidence case for two nonconducting media,
Equations 9.123 and 9.124, and the good conductor limit gives perfect reflection with
a sign flip for the reflected wave.
The case of a wave not at normal incidence becomes more algebraically challenging
because three of the boundary condition equations are required and two of them
involve the angles of incidence and transmission, as we saw in the nonconducting case.
The complex nature of k + i κ further complicates the situation. You will calculate an
approximate solution for this case in homework.
Section 9.4.4 Reflection at a Conducting Surface Page 640
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
We will build a simple model for dispersion. The key elements of the model are:
where x is the displacement from the equilibrium position and ω0 is the natural
frequency of the oscillation. The assumed quadratic potential is
U(x) = m ω02 x 2 /2.
Section 9.5.1 Classical Model for Frequency Dependence of Permittivity — Dispersive Matter Page 641
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
I There is some damping force by which the electron can lose energy as it
oscillates about the minimum of the potential. In most materials, this damping
force is from emission and absorption of phonons (quantized crystal acoustic
vibrations). This can also be thought of as “scattering from” phonons, which
will become relevant when we consider the “conductor” limit later. The
damping force is assumed to follow a standard damping force law
dx
Fdamping = −m γ (9.185)
dt
The wave angular frequency ω and the natural frequency of oscillation ω0 are
not assumed to be the same.
Section 9.5.1 Classical Model for Frequency Dependence of Permittivity — Dispersive Matter Page 642
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
d 2x
m = F = Fbinding + Fdamping + Fdriving (9.187)
dt 2
d 2x dx q
=⇒ +γ + ω02 x = E0 cos ωt (9.188)
dt 2 dt m
You’ve seen this equation, the driven, damped simple harmonic oscillator, in Ph1a and
Ph106a. You know that we switch to complex notation so that the right side becomes
Ee0 e −i ω t and we assume a harmonic solution xe(t) = xe0 e −i ω t . Plugging in this
solution gives an algebraic equation that can be solved for xe0 , which gives the
amplitude and phase of the solution:
q/m
xe0 = Ee0 (9.189)
ω02 − ω 2 − i γ ω
From this, we can calculate the dipole moment of the electron, which will lead us to
the polarization density of the medium:
q 2 /m
pe(t) = q xe(t) = Ee0 e −i ω t (9.190)
ω02 − ω2 − i γ ω
Section 9.5.1 Classical Model for Frequency Dependence of Permittivity — Dispersive Matter Page 643
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
We can obtain the polarization density by the usual formula of summing over all the
electrons. Suppose there are multiple electrons per atom or molecule (which we will
term a “site”) with different binding and damping forces. If there are fj electrons per
site with natural frequency ωj and damping γj , and there are N sites per unit volume,
then the polarization density is
~ N q 2 X fj ~
P= E (9.191)
m j
ωj2 − ω 2 − i γj ω
~ and P
where E ~ are complex vectors. Clearly, there is a proportionality between E ~ and
~ and we may define a complex susceptibility, χ
P, ee , complex permittivity, e, and
complex dielectric constant, e
r ,
~ =χ ~ N q2 X fj
P ee o E =⇒ r = =1+χ
ee = 1 + (9.192)
e
m o j ωj2 − ω 2 − i γj ω
e
o
We will use this complex permittivity in the wave equation for EM waves.
Section 9.5.1 Classical Model for Frequency Dependence of Permittivity — Dispersive Matter Page 644
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
∂2E~
~ =e
∇2 E µo (9.193)
∂t 2
with e
being a complex number. If we assume a plane wave solution, this becomes
very much like the wave equation in conducting matter:
complex permittivity: ~
k ·~
kE~0 e i (~k·~r −ω t ) = e ~0 e i (~k·~r −ω t )
µo ω 2 E (9.194)
~ (~ ) (~ )
k ·~ ~0 e i k·~r −ω t 2 ~0 e i k·~
r −ω t
conducting matter: kE = µω + i σµω E (9.195)
~ (~
E ~0 e −(κ k·~
r , t) = E
b r ) i (k k·~
e
b r −ω t )
(9.197)
The wave speed is v = ω/k, the index of refraction is n = c k/ω, and the intensity
(energy, power) attenuation constant is α = 2 κ.
Section 9.5.2 Wave Solutions for Dispersive Matter Page 645
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
Now, let’s figure out what k and κ are. Let’s specialize to gases where the density and
thus the N-proportional term in e
r is small so we can Taylor expand the square root:
p
2
ωp ω N q2 X fj
µo ω =
k +iκ= e r ≈ 1+ (9.198)
2 m o j ωj2 − ω 2 − i γj ω
e
c c
We separate the real and imaginary parts by making the denominator real to obtain
ck N q2 X fj ωj2 − ω 2
n= ≈1+ 2 (9.199)
ω 2 m o j
ωj2 − ω 2 + γj2 ω 2
α ω2 N q2 X γj fj
=κ= 2 (9.200)
2 c 2 m o j
ωj − ω 2 + γj2 ω 2
2
Our assumption of α being small is very similar to the poor conductor case: the wave
propagates into the medium, with B e0 = Ee0 /v = Ee0 n/c and with the magnetic and
electric fields in phase, and with the wave decaying over a large number of
wavelengths given by the ratio δ/λ = k/(2 π κ). (There are no free charges in the
medium because the oscillating electrons are a bound charge density whose effect is
incorporated into e , so the caveat about the validity of this limit is lifted.)
We see how the damping terms γj are the equivalent of Ohmic losses in a conductor:
they lead to loss of energy from the wave. Because the current is a bound current,
Ohmic loss is absorbed into e rather than provided by an explicit σ.
Section 9.5.2 Wave Solutions for Dispersive Matter Page 646
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
Low-Frequency Behavior, ω ωj
If the frequency is much lower than any of the resonant frequencies, ω ωj , and the
damping term is also small, γj ω (weak damping), then we may drop the imaginary
piece in the denominator of Equation 9.198 and Taylor expand the remainder, yielding:
ωωj N q2 X fj N q2 X fj
n −→ 1 + =1+ (9.201)
2 m o j ωj2 − ω 2
2 m o j ω2
ωj2 1− ωj2
!
N q 2 X fj ω2 N q 2 X fj N q 2 X fj
≈1+ 1+ =1+ + ω2 (9.202)
2 m o j ωj2 ωj2 2 m o j ωj2 2 m o j ωj4
B
n =1+A 1+ 2 (9.203)
λ
This is Cauchy’s Formula, with A being the coefficient of refraction (the offset from
unity) and B being the coefficient of dispersion (the normalization of the
frequency-dependent, dispersive term). The rising index of refraction with frequency
reflects the increasing ability of the individual electrons to be displaced by the field and
thus the sites to be polarized as the frequency approaches the resonant frequencies.
The reason that the expression asymptotes to a constant at low frequency is that, in
this limit, the index of refraction is just the square root of the static dielectric
constant, which depends on the effective spring constant for static distortions of the
electron distribution about the sites. We can see this by going back to the dielectric
constant (so no factor of 1/2), dropping the quadratic term, and rewriting in terms of
the spring constants kj = m ωj2 :
N q 2 X fj
r = =1+ (9.204)
e
e
o o j
kj
Conductor Behavior, ω0 = 0, ω ωj
If we set ω0 = 0 but assume ω ωj for j > 0, we obtain (going back to
Equation 9.198 and looking at e
/o not n, so no factor of 1/2):
N q 2 X fj Ne q 2 1 1
r = =1+ +i + O(ω 2 ) (9.205)
e
e
o o j
k j o m γ0 − i ω ω
(ω/γ0 1 implies the damping time 1/γ0 is much less than the wave period T .)
That is, a conductor can be treated like a dispersive medium with one “resonant
frequency,” ω0 = 0, which corresponds to totally unbound electrons with no restoring
force! Nonconductors have their first resonance at ω1 > 0. Thus, we can view
conductors and nonconductors in a unified way depending on whether there is a
“resonant frequency” at ω = 0.
Section 9.5.4 Conductor Behavior, ω0 = 0, ω ωj Page 649
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
The reason our model can accommodate conductors is because it contains the
resistive damping that results in a terminal velocity for the electrons in the same way
that the Drude model results in a linear relationship between field and electron speed.
Recall our definition of the damping force, |Fdamping | = m γ v , which we can rewrite as
1 v
= (9.209)
γ |Fdamping /m|
which has units of time: 1/γ is a characteristic damping time. Recall also our Drude
model for DC conductivity, with
Ne q 2 λ
σ= (9.210)
2 m vthermal
v 1 m λ τ
= = σ= = (9.211)
|Fdamping |/m γ0 Ne q 2 2 vthermal 2
where τ is the time between scatters in the Drude model. Recall (Equation 7.3) that
vave = (τ /2) (q E /m) is average drift velocity in the Drude model and thus τ /2 is the
damping timescale for the motion driven by the DC electric field. That is exactly the
definition of γ0−1 , the damping time for the DC electric field.
Page 651
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
Plasma Behavior, ω ωj
Again, we can ignore the damping behavior because now ω ωj (the ω 2 − ωj2 term
dominates and we can neglect its ωj2 piece), providing
ωωj N q 2 X fj 1 ωp2 N q2 X N Z q2
n −→ 1 − ≡1− ωp2 ≡ fj = (9.212)
2 m o j ω 2 2 ω2 m o j m o
P
(Z = j fj is the number of electrons per site.) This is the same relationship one
would have obtained if one had ignored Fbinding and Fdamping : this behavior arises only
from the force being given by the driving electric field. At these high frequencies, the
oscillation occurs much faster than any binding (ω ωj ) or damping (ω γj ) forces
can have effect and the electrons can be considered to be free. The asymptotic return
of the index of refraction back toward unity reflects the fact that, due to the electrons’
inertia, the distance they can move during one cycle of the EM wave, and hence the
polarizability of the medium, decreases as 1/m ω 2 at high frequency. ωp is called the
plasma frequency because it defines the behavior of a plasma of free electrons.
Notice that n < 1 is possible in the plasma limit, yielding wave speeds that are larger
than c. Griffiths discusses how the group velocity of waves prevents a violation of
special relativity.
Strangely enough, the plasma behavior can hold at much lower frequencies in tenuous
plasmas where the electrons are free and binding forces are negligible. In these cases,
one has to return to the full permittivity (Equation 9.192) and set ωj2 and γj to zero
because the approximation that the dispersive term in the refractive index is small
compared to unity fails. In such cases, one has a permittivity
ωp2
=1− 2 (9.213)
e
o ω
It is possible for this relation to hold even for ω < ωp , in which case, e becomes
negative. Note that this is not equivalent to the perfect conductor limit: in that case,
acquires a significant imaginary component, as we just saw in the previous section.
e
The implications of the purely negative e are that there is reflection at an interface
into such a medium (total internal reflection for all angles) and the field falls off
exponentially into the medium with skin depth and decay constant
1 ωp
=κ= (9.214)
δ c
It turns out that this plasma phenomenon, not the perfect conductor limit, explains
the optical reflectivity of metals. There are resonances in the ultraviolet in metals that
then bring e
> 0 and allow metals to be ultraviolet transparent. This phenomenon also
explains how the ionosphere reflects AM (500–1600 kHz) and shortwave
(2.3–26 MHz) radio waves. AM radio reflection off the ionosphere is
conditions-dependent and is better at night than during day.
Section 9.5.5 Plasma Behavior, ω ωj Page 653
Section 9.5 Electromagnetic Waves: Electromagnetic Waves in Dispersive Matter
Near-Resonance Behavior
The behavior of the refractive index and skin depth near a resonance are shown below.
The decay constant displays a maximum (the skin depth displays a minimum) on
resonance because the system is best able to absorb (and dissipate) energy on
resonance.
One interesting feature is that, above ωj , n can drop below unity, yielding v > c.
Another interesting feature arises from the anomalous dispersion. Let’s relate the
negative slope in n(ω) to the group velocity to see its implications:
k dn c dk k dω v
n=c =⇒ = −c 2 =⇒ vg = = (9.215)
ω dω ω dω ω dk 1 + dd ln
ln n
ω
χrot
r ,rot = 1 +
e (9.216)
1 − i ω τrot
where χrot is the zero-frequency contribution to the susceptibility and τrot is the time
constant for relaxation of an imposed polarization due to thermal fluctuations in the
−1
absence of an electric field. The real part falls off for ω > τrot while the imaginary
−1
part has a peak at ω = τrot . The real part yields standard non-conducting matter
behavior while the imaginary part yields conductor-like behavior in terms of the
relative phase of B and E and of dissipation (into the thermal modes that cause
relaxation of polarization). One never arrives, however, at the good conductor limit
because the second term has an upper limit of χrot while the corresponding term for a
conductor (Equation 9.207) is σ/( ω), which has no upper limit, and because χrot
itself has an upper limit set by full alignment of the intrinsic dipole moments and E ~.
Another way of saying this is that the electrons that are responding are still bound and
thus can never display the almost-free propagation characteristic of a good conductor.
The generic behavior of the real and imaginary parts of e /o = er (usually denoted by
0 and 00 ), in the rotational/conduction limits and through the anomalous dispersion
features, is shown here.
We will first consider transmission lines from the circuit theory perspective, then
consider Maxwell’s Equations in confined regions and the propagating solutions. The
material on transmission lines can be found in Heald and Marion §7.1. The material
on waveguides is found in Heald and Marion §7.3–7.5 and Jackson §8.1–8.5.
Page 657
Section 9.7 Electromagnetic Waves: Transmission Lines
Transmission Lines
Circuit Transmission Line Theory
Here, we consider a specialized arrangement consisting of two electrodes,
translation-invariant in the z direction, with a voltage between them and currents
flowing on them. We allow for the voltage and current to be position-dependent,
which is a necessity for a propagating wave. We shall see that the critical ingredients
necessary for propagating solutions are inductance and capacitance.
We start with the figure below, which consists of two wires on which there is a
position-dependent voltage difference and a position-dependent current. We treat the
two wires completely (anti)symmetrically: if there is a voltage V (z) on one wire, there
is a voltage −V (z) on the other wire at the same point. Thus, the voltage difference
between the wires is ∆V (z) = 2 V (z), which is the quantity we will be primarily
concerned with. There is a current flowing on each wire, again antisymmetric, with
values I(z) and −I(z). There may be a net charge Q(z) and −Q(z) on the wires. We
assume the system has an inductance per unit length for the pair of wires of L and a
capacitance per unit length between the wires of C.
dQ = C dz d(∆V ) (9.217)
∂I ∂ ∆V
= −C (9.219)
∂z ∂t
The voltage drop due to the inductance (we assume zero resistance) is:
∂I L ∂I
d(∆V ) = −L dz =⇒ dV = − dz (9.220)
∂t 2 ∂t
To understand the sign, think about the loop formed by the four points in the
diagram. For I > 0 and ∂I/∂t > 0, the magnetic flux through the loop is into the
page and increasing in magnitude. So an emf is induced, with the polarity such that it
tries to drive the current the opposite way. But where does the emf appear?
Remember our discussion about closing loops with resistors in connection with
Faraday’s law. Here, we should think of there being a resistor at z + dz between the
two electrodes, and, for the above polarity, the upper wire’s voltage must decrease and
the lower wire’s voltage must increase in order to drive a current across the resistor
and around the loop in the desired direction (counterclockwise). So, a positive ∂I/∂t
should make ∆V less positive. This explains the sign.
To understand the factor of 1/2. we just need to recognize that the full emf affects
∆V , so, when we want to calculate the dV of a single wire, it is ∆V /2.
What about the fact that now we have a voltage difference dV (< 0 for ∂I/∂t > 0)
along the wire when we said we couldn’t have electric fields in wires? That voltage
difference is not because of an electric field, it is because of finite propagation speed.
∂ ∆V ∂I
= −L (9.221)
∂z ∂t
We can take derivatives of our two equations (Equations 9.219 and 9.221) and
combine them to obtain
∂2I ∂2I ∂ 2 ∆V ∂ 2 ∆V
= LC 2 = LC (9.222)
∂z 2 ∂t ∂z 2 ∂t 2
We thus
√ have wave equations describing waves in one dimension moving with speed
v = 1/ L C.
In one dimension and for a fixed frequency ω, there are two solutions with velocity +v
and −v and they have the general form (as we derived earlier for EM waves)
ω ω
I± (z, t) = I± F ± z − ω t ∆V (z, t) = V± (z, t) = Z± I± F ± z − ω t
v v
(9.223)
We see that the wave propagating to +z has current flow to the right in the top wire
when the top wire has a positive voltage while the −z mode has current flow to the
left when the top wire has a positive voltage. Rather than allowing the negative
impedance −ZLC , we will consider these two modes to have the same characteristic
impedance, ZLC , remembering that the currents flow in opposite directions. While
ZLC has units of ohms because it is the ratio of a voltage to a current, realize that it
has nothing to do with resistivity in the wires! It is set by the geometry of the
electrodes and the materials ( and µ) surrounding them, which set L and C.
b
2π µ ln a
C= b
L= (9.225)
ln a
2π
ln ba
r r
1 1 c L µ
v = √ = √ = ZLC = = (9.226)
LC µ n C 2π
Recall that , µ, n, v , and Zµ characterize the dielectric between the conductors, not
√
the perfect conductors themselves! For µ ≈ µo , this gives ZLC ≈ (60Ω) ln ba / r .
Teflon and polyethylene are frequently used as the dielectric, having r = 2.0 for teflon
and r = 2.3 for high-density polyethylene. Standard coaxial cable impedances are
50 Ω and 75 Ω. 75 Ω is a typical cable impedance for television antennas because it is
a good match to the antenna impedance (which we will discuss later). Not that any of
you are old enough to have seen a television antenna...
This structure consists of two thin metal ribbons of width w and separation h w .
The capacitance is just that of a parallel plate capacitor. The inductance is easily
obtained by calculating the field between two infinite sheets of current flowing in
opposite directions (B = µ K ) and then calculating from that the flux per unit length
(Φ/` = µ K h) and dividing by the current (I = K w ). Thus, we have
w µh
C= L= (9.227)
h w
r r
1 1 c L h µ
v = √ = √ = ZLC = = (9.228)
LC µ n C w
Page 665
Section 9.7 Electromagnetic Waves: Transmission Lines
Section 9.7.2 Generalization to Arbitrary Electrode Shapes via Wave Formulation Page 666
Section 9.7 Electromagnetic Waves: Transmission Lines
To do this, let’s think about the fields and the charge and current densities rather
than just voltages and currents. We aim to find a solution to Maxwell’s Equations
that takes the form of a wave propagating in the z-direction and that has transverse
fields (consistent with our transmission line picture involving capacitance and
inductance per unit length). We use H ~ instead of B~ because the former is more
directly related to the surface current density. Recall k vµ = ω. We assume:
~ (~
E x Ex (x, y ) + yb Ey (x, y )] e i(k z−ω t)
r , t) = [b r , t) = ρ0 (x, y ) e i(k z−ω t)
ρ(~ (9.229)
~ r , t) = [b
H(~ x Hx (x, y ) + yb Hy (x, y )] e i(k z−ω t) ~ r , t) = zb J0 (x, y ) e i(k z−ω t)
J(~ (9.230)
ρ0 and J0 are assumed to be surface charge and current densities, so they have δ
functions in the transverse coordinates in them, but we don’t write them out explicitly
because it is unnecessary for our discussion here.
We will show that, under the assumptions made above, in which the charge and
current densities are the same in form and have a wave dependence on z, then the
propagating wave solution is just a static solution in (x, y ) multiplied by the e i(k z−ω t)
dependence, where the static solution is that sourced by ρ0 and J0 = vµ ρ0 .
When we get to waveguides, we will see that this special case yields the TEM
(transverse-electric-magnetic) modes, while more general current distributions yield
other, more complex modes.
Section 9.7.2 Generalization to Arbitrary Electrode Shapes via Wave Formulation Page 667
Section 9.7 Electromagnetic Waves: Transmission Lines
To do this, let’s write out Maxwell’s Equations assuming the above form. The four
equations involving x and y derivatives are
h i ~
∂B ∂ Ey ∂ Ex
∇ ~
~ ×E =− · zb =⇒ − =0 (9.231)
z ∂t ∂x ∂y
h
~ = ρ
~ ·E
i ∂ Ex ∂ Ey ρ0 (x, y )
∇ =⇒ + = (9.232)
∂x ∂y
~
~ = J~f · zb + ∂ E · zb ∂ Hy ∂ Hx
h i
~ ×H
∇ =⇒ − = J0 (x, y ) (9.233)
z ∂t ∂x ∂y
h
~ ~
i ∂ Hx ∂ Hy
∇·H =0 =⇒ + =0 (9.234)
∂x ∂y
These four equations are satisfied by the form we have chosen; the exponential factor
multiplies both sides of every equation and thus cancels out. They will yield a static
solution because they are sourced by the charge and current densities ρ0 and J0 , which
are not time-dependent.
Section 9.7.2 Generalization to Arbitrary Electrode Shapes via Wave Formulation Page 668
Section 9.7 Electromagnetic Waves: Transmission Lines
h i∂B ~
∇ ~
~ ×E =−
· xb =⇒ −i k Ey = i ω µ Hx (9.235)
∂tx
~
~ = − ∂ B · yb
h i
~ ×E
∇ =⇒ i k Ex = i ω µ Hy (9.236)
y ∂t
~
~ = J~f · xb + ∂ E · xb
h i
~ ×H
∇ =⇒ −i k Hy = −i ω Ex (9.237)
x ∂t
~
~ = J~f · yb + E · yb
∂
h i
~ ×H
∇ =⇒ i k Hx = −i ω Ey (9.238)
y ∂t
These equations are satisfied by the exponential factor along with the relation
1 ~ = 1 b ~
xb Hx + yb Hy = (−b
x Ey + yb Ex ) =⇒ H k ×E (9.239)
Zµ Zµ
Section 9.7.2 Generalization to Arbitrary Electrode Shapes via Wave Formulation Page 669
Section 9.7 Electromagnetic Waves: Transmission Lines
~ · J~ + ∂ρ/∂t = 0:
Let’s relate the charge and current densities by applying continuity, ∇
In addition to relating the two densities, this equation tells us that current in a perfect
conductor flows only on the surfaces: since we know there is no charge density inside a
perfect conductor (the decay constant for the charge density was τ = /σ → 0, in the
language of our solutions for EM waves in conductors), the above equation tells us
there is no current density inside the conductor either.
p
Another way of seeing this would be to recognize that the skin depth δ = 2/(σ µ ω)
vanishes as σ → ∞: the fields do not penetrate into the conductors, so neither does
the current density. This happens because a surface current density is driven that
cancels the EM field inside the conductor.
This, along with the fact that xb Ex + yb Ey and xb Hx + yb Hy are sourced by the
constant charge density ρ0 and current density J0 , shows what we set out to prove,
that a transmission line has propagating wave solutions in z whose transverse field
configuration is identical to that of the two-dimensional static problem sourced by
charge density ρ0 and current density J0 = vµ ρ0 . Note that these are not the only
solutions, as we will show later.
Section 9.7.2 Generalization to Arbitrary Electrode Shapes via Wave Formulation Page 670
Section 9.7 Electromagnetic Waves: Transmission Lines
Now let’s recover the circuit formulation from the wave formulation. Given the wave
formulation solution, we can calculate the voltage and current via
Z ~
r−
∆V (z, t) = d~ ~
`·E (9.241)
~
r+
1
I I I
I(z, t) = ± d` K0 = ±vµ d` σ0 = ± ~
d` nb · E (9.242)
C± C± Zµ C±
where we used the relation between the surface charge density and the discontinuity in
the normal ~
p component of D in obtaining the last expression, noting
vµ = /µ = 1/Zµ . K0 is the surface current density, with J0 and K0 related by a
δ function that defines the conductors’ surfaces, and σ0 is the surface charge density,
with σ0 and ρ0 also related by this δ function. The voltage integral is from any point
on the + electrode to any point on the − electrode (since each is an equipotential in
(x, y )) and the current (charge) integral is over the entire + or − electrode.
Section 9.7.3 Connection between Circuit and Wave Formulations of Transmission Lines Page 671
Section 9.7 Electromagnetic Waves: Transmission Lines
where C0 = C o / is the capacitance per unit length if the medium is replaced with
o . We see that the characteristic impedance is proportional to Zµ with the constant
of proportionality being related to the capacitance per unit length scaled by . This is
a purely geometric quantity (not even dependent on or µ). Clearly, the above
quantity is determined only by the physical shape of the solutions, implying it is set
purely by the geometry. This is what we found earlier in our coaxial cable and stripline
examples, that the transmission line impedance ZLC was Zµ times a factor
determined by the geometry alone.
Section 9.7.3 Connection between Circuit and Wave Formulations of Transmission Lines Page 672
Section 9.7 Electromagnetic Waves: Transmission Lines
1 1
LC = µ =⇒ vµ = √ = √ (9.245)
LC µ
µ L L L0
ZLC = Zµ = Zµ = Zµ = Zµ (9.246)
C CL µ µ µ0
Finally, we note that the combination of the above relations tell us that the
geometrical factor in C and L is the same (up to reciprocation):
L L0 o
LC = µ ⇐⇒ = ⇐⇒ = (9.247)
µ C µo C0
Section 9.7.3 Connection between Circuit and Wave Formulations of Transmission Lines Page 673
Section 9.7 Electromagnetic Waves: Transmission Lines
The Poynting vector has units of power/area. Therefore, if we integrate it over the
xy -plane, we get the total time-averaged power flowing past a given point:
Z D E 1 Z 2
hPi = ~ = vµ
da zb · S da Ee0
(9.249)
S 2 S
The integral (with the 1/2) is the time-averaged energy per unit length in the electric
field, which can be rewritten using the capacitance per unit length and the voltages
(recall, this is how we derived ue = 21 E 2 ). Thus, we have
1 1 C 1 |V0 |2 1 1
hPi = vµ C |V0 |2 = √ |V0 |2 = = |I0 |2 ZLC = R (I∗0 V0 ) (9.250)
2 2 LC 2 ZLC 2 2
Now, let’s consider a wave incident on a junction between two transmission lines (at
z = 0) of characteristic impedances Z1 and Z2 . We assume a solution consisting of an
incident and reflected wave at z < 0 and a transmitted wave at z > 0. The voltage
must be continuous as always, and the current must be continuous to avoid a buildup
~ and B
of charge at the boundary. (If we choose to match fields instead: E ~ are
transverse and there are no free charges or currents at the boundary, so their normal
components vanish and their transverse components are continuous, yielding these
conditions on V and I .) We write the waves as (including a reflected wave for z < 0):
Section 9.7.5 Reflection and Transmission at a Transmission Line Junction Page 675
Section 9.7 Electromagnetic Waves: Transmission Lines
1 − re te Z2 − Z1 2 Z2
1 + re = te = =⇒ re = te = (9.255)
Z1 Z2 Z2 + Z1 Z1 + Z2
We see that these equations are identical to Equations 9.123 and 9.124 for the fields
in the case of normal incidence of an EM wave on the interface between two different
perfectly non-conducting materials, except that now the impedances are the
characteristic impedances of the transmission lines ZLC instead of the wave
impedances Zµ . Of course, the former incorporates the latter, but it also depends on
the transmission line geometry, which does not exist for the free space case.
1 1 e2
r |2 /Z1
2 2
|e Z2 − Z1 |t | /Z2 Z1 2 Z2
2 2
R= 1
= T = 1
= (9.256)
2
|1|2 /Z1 Z2 + Z1 2
|1| 2 /Z
1 Z2 Z1 + Z2
One can easily see R + T = 1 and that these expressions also match what we derived
for an EM wave at normal incidence.
Section 9.7.5 Reflection and Transmission at a Transmission Line Junction Page 676
Section 9.7 Electromagnetic Waves: Transmission Lines
te −i ω t
z>0: V (t) = te e −i ω t I(t) = e (9.257)
ZL
This doesn’t change the matching conditions, so we have the same relations as
Equations 9.255 and 9.256 with Z2 replaced by ZL . In particular, there is full
transmission of the power from the transmission line into the load if ZL = Z1 : this is
another version of impedance matching. Note that ZL = 0 and ZL = ∞ both result in
unity reflection, though with opposite signs for re.
Page 678
Section 9.7 Electromagnetic Waves: Transmission Lines
Let’s assume that there are left- and right-going waves on this length ` of transmission
line with amplitudes V+ and V− to be determined. (We won’t write the e −i ω t time
dependence for now; think of it as considering everything at t = 0.) Define the
“complex-wave ratio” as the ratio of the left-going and right-going amplitudes:
V− (z)
r (z) =
V+ (z)
V (z) 1 + r (z)
Z (z) = = ZLC (9.258)
I(z) 1 − r (z)
We know or can now calculate the functions V+ (z), V− (z), r (z), and Z (z):
Now, to find Zi = Z (0), the effective input impedance of the transmission line, we
simply set Z (`) = ZL (this is the same as requiring that V (`) = ZL I(`)), find r (0)
from the relation between Z (z) and r (0), and then find Zi = Z (0) from r (0):
ZL
ZLC
−1
Z (`) = ZL =⇒ r (0) = e 2 i k ` ZL
(9.261)
ZLC
+1
1 + r (0) ZL − i ZLC tan k`
=⇒ Zi = Z (0) = ZLC = ZLC (9.262)
1 − r (0) ZLC − i ZL tan k`
Based on this formula, some examples of the behavior are given below. It is crucial to
recognize that the behavior depends on k so is frequency-dependent!
I Short-circuit termination
λ λ
(2 n) < ` < (2 n + 1) =⇒ I(Zi ) < 0 =⇒ inductive (9.264)
4 4
λ λ
(2 n − 1) < ` < (2 n) =⇒ I(Zi ) > 0 =⇒ capacitive (9.265)
4 4
Note that the sign conventions for inductive and capacitive reactances are the
opposite of what is used in Ph1c and in engineering because we use here a
e −i ω t time dependence instead of the usual e i ω t engineering time dependence.
I Open-circuit termination
I Quarter-wavelength
2
ZLC
λ
`= Zi = (9.267)
4 ZL
I Half-wavelength
λ
`= Zi = ZL (9.268)
2
No change in impedance.
Waveguides
What is a Waveguide?
~ (~
E ~0 (~
r , t) = E r⊥ ) e i(k z−ω t) ~ r , t) = H
H(~ ~ 0 (~
r⊥ ) e i(k z−ω t) (9.269)
∂ Ey ∂ Ex ∂ Hy ∂ Hx
− = i µ ω Hz − = −i ω Ez (9.270)
∂x ∂y ∂x ∂y
∂ Ez ∂ Hz
− i k Ey = i µ ω Hx − i k Hy = −i ω Ex (9.271)
∂y ∂y
∂ Ez ∂ Hz
i k Ex − = i µ ω Hy i k Hx − = −i ω Ey (9.272)
∂x ∂x
We allow for the medium in the waveguide to have uniform , µ in the above
equations. We see the symmetry between E and H in the equations.
We may solve these reduced equations for the transverse components in terms of the
z, or longitudinal components. (e.g., plug the equation with Hy on the RHS into the
equation for Ex on the RHS and solve for Ex and one gets an equation for Ex in terms
of Ez and Hz .) Defining any vector with a ⊥ subscript p
to be the vector’s component
in the xy -plane, and recalling kµ = ω/vµ and Zµ = µ/, we obtain
1 ~ ⊥ H0,z + i kµ zb × ∇
~ 0,⊥ =
H i k ∇ ~ ⊥ E0,z (9.273)
2 − k2
kµ Zµ
~0,⊥ = 1 h
~ ⊥ E0,z − i kµ Zµ zb × ∇
~ ⊥ H0,z
i
E 2 − k2
ik∇ (9.274)
kµ
Note that the transverse components end up out of phase with the longitudinal
component by π/2 due to the factor i. Note that one cannot find a generic, simple
~0,⊥ and H
relationship between E ~ 0,⊥ like we had before for transverse waves. We will
see below that we can obtain such a relationship, but differently for different modes.
For propagation in the −bz direction, one can work through the algebra to see that
only the sign on k in the first terms changes; the second terms are unchanged. This is
because the prefactors in the second terms, as well as the sign of zb, do not depend on
k.
In order for the units on the two sides to match, the factor of 1/k cancels the units of
~ ⊥ operator, and there is the usual factor of Zµ connecting E and H.
the ∇
Section 9.8.2 Eigenvector-Eigenvalue Equation for Waveguide Solutions Page 685
Section 9.8 Electromagnetic Waves: Waveguides
Next, let’s use the Maxwell divergence equations by taking the divergence of the
above. We note that
~ ⊥ = xb ∂ + yb ∂ + zb ∂ ∂ ∂
~ · zb × ∇
∇ · yb − xb =0 (9.275)
∂x ∂y ∂z ∂x ∂y
Therefore,
0=∇ ~ = e i(k z−ω t) ∇
~ ·E ~0,⊥ + i k E0,z
~⊥·E (9.276)
ik
= 2 − k2
∇2⊥ E0,z + i k E0,z (9.277)
kµ
~ we obtain
~ · B,
Rewriting, and doing the same with ∇
ω
∇2⊥ E0,z + kµ
2
− k 2 E0,z = 0 ∇2⊥ H0,z + kµ
2
− k 2 H0,z = 0
kµ = (9.278)
vµ
If we choose E0,z = 0, then the waves are called transverse electric (TE) modes, and if
we choose H0,z = 0, they are called transverse magnetic (TM) modes. Note that,
because E~0,⊥ depends on E0,z and H0,z , setting E0,z = 0 still allows for nonzero E ~0,⊥
for TE modes, and similarly for H ~ 0,⊥ for TM modes. By linearity, any arbitrary sum of
TE and TM modes is still a valid field configuration. Note that the mode wavelength,
set by k, is not necessarily the same as it is in unconfined space or transmission lines,
which would be given by kµ . Once one has specified TE or TM, a simple relationship
between the transverse fields is obtained (from Equations 9.273 amd 9.274):
Note that these relations are k-dependent (⇔ ν-dependent) because the ratio k/kµ
or kµ /k appears. So, while they are generic relations, there will be no such simple
relation if one sums modes with different k (⇔ ν) or sums TE and TM modes.
There are also transverse electric and magnetic (TEM) modes. These are not obtained
by setting E0,z = H0,z = 0 in the above wave equations. Rather, one must go back to
the curl equations and set the z components to be zero there. This results in
~0,⊥ and H
simplified equations for E ~ 0,⊥ as we derived in our transmission line theory
discussion (Equations 9.229 and 9.230): static solutions in two dimensions give the
transverse field behavior for the propagating TEM mode (with H ~ 0,⊥ = zb × E
~0,⊥ /Zµ
and k = kµ ). In fact, “transmission line” and “TEM mode” should be considered
synonymous, especially given what we will prove soon, which is that hollow
waveguides cannot have TEM modes.
Section 9.8.2 Eigenvector-Eigenvalue Equation for Waveguide Solutions Page 687
Section 9.8 Electromagnetic Waves: Waveguides
We have the partial differential equations for the fields, but we need boundary
conditions to have a fully defined mathematical problem to solve, so let’s consider
those now.
There will be free charges and currents in the walls to generate the fields, so we may
~ and tb · H
not assume anything about nb · E ~ at the walls.
For a perfect conductor, the skin depth vanishes. Therefore, there is no magnetic or
~ = 0, implies nb · H
~ ·B
electric field in the conductor. This, along with ∇ ~ = 0 at the
walls.
~ =0
The vanishing of the fields in the conductor along with Faraday’s Law tells us tb · E
at the walls. (Remember, fields cannot have a δ-function spatial dependence, so the
right side of Faraday’s Law yields no contribution to the surface integral that is related
to the usual loop integral we do to determine tangential boundary conditions like this.)
~ =0
~ ·E ∂ E0,x ∂ E0,y h
~
~ ×E
i ∂ E0,y ∂ E0,x
∇ =⇒ + =0 ∇ =0 =⇒ − =0
∂x ∂y z ∂x ∂y
The other curl equation terms vanish because E ~0 (x, y ) has no z dependence and E0,z
vanishes. Thus, E ~0 has no curl, so it is the gradient of a scalar potential V0 (x, y ). The
vanishing of ∇~ ·E ~ tells us that this potential V0 (x, y ) satisfies Laplace’s (instead of
Poisson’s) Equation in two dimensions. The boundary condition tb · E ~ = 0 implies any
connected boundary of the waveguide is an equipotential for V0 (at fixed z). If there
is only one boundary, then we may use the fact that solutions to Laplace’s Equation
have no local minima or maxima to conclude that the potential must be constant
across the (x, y )-plane and thus that E ~0,⊥ = 0. Given the generic relation for TEM
modes, H ~ 0,⊥ = zb × E~0,⊥ /Zµ , this also implies H
~ 0,⊥ = 0. (The lack of fields also
implies ρ0 (x, y ) and J0 (x, y ) vanish too.)
This argument does not apply for a non-hollow waveguide because there are at least
two boundaries that may not be at the same potential (e.g., coaxial cable).
This result makes sense: transmission lines need two (and exactly two) non-grounded
electrodes with complementary currents and voltages to define L and C.
Section 9.8.4 No TEM Mode in Hollow Waveguide Page 689
Section 9.8 Electromagnetic Waves: Waveguides
We now return to the generic case in which either E0,z 6= 0 or H0,z 6= 0. Since we
have reduced the problem to one of finding these z components of the fields, we need
boundary conditions on E0,z and H0,z on the walls, which we will define by a contour
C (consisting of multiple disconnected pieces when there is more than one conductor).
We may now apply the previously developed generic boundary conditions to these
fields and use Equations 9.273 and 9.274.
E0,z |C = 0 (9.280)
Note that this holds even for (especially for!) TM modes, which have nonzero E0,z .
Even though E0,z is nonzero in general for TM modes, it does vanish at the walls, and
the vanishing at the walls does not imply it vanishes everywhere!
We can also apply Faraday’s law with the orthogonal tangent vector, tb⊥ = zb × nb. This
~ = 0 and
vector is tangent to the walls and lies in the xy plane. Using tb⊥ · E
Equation 9.274:
~ = tb⊥ · E
~0,⊥ = tb⊥ · 1 h
~ ⊥ E0,z − i Zµ kµ zb × ∇
~ ⊥ H0,z
i
0 = tb⊥ · E 2 2
ik∇
kµ −k
The first term on the right side vanishes because it calculates the gradient of E0,z
along tb⊥ , which vanishes because E0,z vanishes everywhere on the walls. The second
term can be cleaned up using the vector identity ~a · (~ ~ = (~
b × ∇) a×~ ~ to yield
b) · ∇
another boundary condition on the z components of the fields.:
~ ⊥ H0,z = tb⊥ × zb · ∇
~ ⊥ H0,z ~ ⊥ H0,z = 0
0 = tb⊥ · zb × ∇ =⇒ nb · ∇ (9.281)
C
1 ~ ⊥ H0,z + i kµ zb × ∇
~ = nb · H
0 = nb · H ~ 0,⊥ = nb · ik∇ ~ ⊥ E0,z
2
kµ −k 2 Zµ
∇2⊥ ψ + γ 2 ψ = 0 γ 2 = kµ
2
− k2 ψ|C = 0 (9.282)
Again, the specific solutions depend on the shape of the boundary chosen.
It should be clear that the TM and TE solutions are complementary in that they solve
the same eigenvalue-eigenvector equation but with complementary boundary
conditions, Dirichlet and Neumann. These two types of boundary conditions are the
only types of boundary conditions implied by Maxwell’s Equations, so the solutions,
along with the TEM mode solutions if they exist, form a complete set to describe any
propagating field in the waveguide. As noted above, there is no reason for the γn ’s of
the TE and TM modes to be the same, and certainly the {ψnTM } and {ψnTE } will be
different due to the different boundary conditions.
Page 695
Section 9.8 Electromagnetic Waves: Waveguides
Section 9.8.9 Propagation Properties: Propagation Constant, Wave Speed, Dispersion Relations, and Wave Impedance Page 696
Section 9.8 Electromagnetic Waves: Waveguides
On the other hand, the TE and TM modes satisfy the condition (from γ 2 = kµ
2 − k2
ω 2 = vµ
2
k 2 + γn2
(9.286)
Thus, there is a cutoff frequency ωc,n = vµ γn such that, for ω < ωc,n , we find
k 2 < 0, yielding a decaying mode with decay constant
s s
ωc,n = vµ γn ω2 ωc,n ω2
: κn (ω) = −i kn (ω) = γn2 − 2 = 1− (9.287)
ω < ωc,n vµ vµ 2
ωc,n
Note that the decay constant depends on which type (TM or TE) and number n of
mode one considers! We see that the decay length dn (ω) = 1/κn (ω) → ∞ as
ω → ωc,n from below because energy-conserving propagation must become possible
for ω > ωc,n .
Section 9.8.9 Propagation Properties: Propagation Constant, Wave Speed, Dispersion Relations, and Wave Impedance Page 697
Section 9.8 Electromagnetic Waves: Waveguides
For the propagating modes with ω > ωc,n , the propagation constant, wavelength, and
wave speed are
s s s
2
ωc,n 2
ωc,n
ωc,n = vµ γn ω2 ω
: kn (ω) = − γn2 = 1− = kµ 1− (9.288)
ω > ωc,n 2
vµ vµ ω2 ω2
ω vµ 2π λµ
vn (ω) = = r λn (ω) = = r (9.289)
kn (ω) ω2 kn (ω) ω2
1 − ωc,n
2 1 − ωc,n
2
We see that the propagation constant, wavelength, and the wave speed go to the
unguided values kµ = ω/vµ , λµ = 2 π/kµ = vµ /ν, and vµ , respectively, as
ω → ∞: the system approaches the transmission-line/free-space limit. If one goes
back to Equations 9.273 and 9.274, one sees that ratio of the transverse fields to the
2 − k 2 ) as ω → ∞ (i.e., k → k ), making the
longitudinal field increases as 1/(kµ µ
longitudinal component small in this limit, as one would expect. Essentially,
propagation looks like that in free space as the wavelength becomes small compared
to the transverse dimensions: the wave no longer notices the boundaries.
As ω → ωc,n , the propagation constant vanishes, the wavelength becomes infinite, and
the wave speed becomes infinite. (The latter two must become infinite together
because their ratio is ν, which remains finite. Similarly, because kn (ω) vn (ω) = ω,
kn (ω) → 0 as vn (ω) → ∞ so ω can remain constant.)
Section 9.8.9 Propagation Properties: Propagation Constant, Wave Speed, Dispersion Relations, and Wave Impedance Page 698
Section 9.8 Electromagnetic Waves: Waveguides
We recall from our discussion of EM waves in matter that the wave impedance is
~ and E
implicitly defined via the relation between H ~ , Equation 9.55. The generalization
in the case of waveguides uses the transverse components of H ~ and E~ (following the
relation between the transverse components of H ~ and E ~ , Equations 9.279):
~ 0,⊥ = 1 kb × E
H ~0,⊥ (9.290)
Z
Using our relations between H ~ 0,⊥ and E
~0,⊥ for the TM and TE modes and the
expression for kn (ω) we just derived, we obtain
s
2
k TM (ω) ωc,n kµ Zµ
ZnTM (ω) = n Zµ = Zµ 1− ZnTE (ω) = Zµ = r (9.291)
kµ ω2 knTE (ω) ω2
1 − ωc,n
2
The wave impedance behaves differently as ω → ωc,n for the TM and TE modes! This
is explained by the current flow. In the TE mode, H ~ ⊥ ∝ k TE (ω) → 0 as ω → ωc,n
n
(while E ~⊥ ∝ kµ Zµ ). Thus, no current flows longitudinally while the voltage remains
finite, so we need ZnTE (ω) → ∞ to obtain this. On the other hand, in the TM mode,
E~⊥ ∝ k TM (ω) → 0 as ω → ωc,n (while H ~ ⊥ ∝ kµ /Zµ ). In this case, current flows
n
longitudinally but no voltage drop appears, so we need ZnTM (ω) → 0 to obtain this.
In the ω → ∞ limit, we recover Zµ for both impedances, just as we obtained the
free-space limits for propagation constant, wave speed, and wavelength in this limit.
Section 9.8.9 Propagation Properties: Propagation Constant, Wave Speed, Dispersion Relations, and Wave Impedance Page 699
Section 9.8 Electromagnetic Waves: Waveguides
Let’s now specialize to a particular geometry, which will let us find the modes that
solve the previously mentioned eigenvector-eigenvalue problem. Our first example will
be rectangular waveguide of height a in x and width b in y . One corner of the
waveguide is at (x, y ) = 0 and the diagonal corner is at (x, y ) = (a, b).
1 d 2X 1 d 2Y 2
− k2 = 0
+ + kµ (9.292)
X (x) dx 2 Y (y ) dy 2
1 d 2X 1 d 2Y
= −kx2 = −ky2 k 2 = kµ
2
− kx2 − ky2 (9.293)
X (x) dx 2 Y (y ) dy 2
The solutions to the differential equations are clearly exponentials, and they have
imaginary arguments for kx2 ≥ 0, which we will find it is most convenient to write as
sinusoids:
X (x) = Ax sin kx x + Bx cos kx x (9.294)
~ 0,z = 0. Explicitly,
Our boundary coundition is nb · ∇H
C
dX mπ mπx
=0 =⇒ Ax = 0 kx = X (x) ∝ cos (9.295)
dx x=0,a a a
dY nπ nπy
=0 =⇒ Ay = 0 ky = Y (y ) ∝ cos (9.296)
dy y =0,b b b
rectangular
mπx nπy 2 m2 π 2 n2 π 2
=⇒ waveguide H0,z ∝ cos cos γmn = kx2 + ky2 = + (9.297)
TEmn mode a b a2 b2
~ 0,⊥ and E
This is called the TEmn mode. From H0,z , we can obviously obtain H ~0,⊥ via
Equations 9.285.
On the next few slides, we show some visualizations of the fields and surface charge
and current densities for the TE10 mode.
0.8
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0
0.8
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0
rectangular
s
m2 n2
waveguide ωc,mn = vµ π 2
+ 2 (9.299)
TEmn mode a b
Griffiths has a nice discussion of interpreting these modes as the result of a free EM
wave propagating through the waveguide with an infinite number of reflections off the
walls. It can help to build some intuition for how these modes arise. The actual
calculation clearly can only be applied in the simple case of rectangular waveguide.
It is straightforward to obtain the TM modes. The same math applies except the
boundary condition is X (x)|x=0,a = 0 and Y (y )|y =0,b = 0, so one obtains sines
instead of cosines (like particle in a box in quantum mechanics):
rectangular
s
mπx nπy m2 n2
waveguide E0,z ∝ sin sin ωc,mn = vµ π 2
+ 2 (9.300)
TMmn mode a b a b
These sine modes are of course necessary to ensure E0,z vanishes at the walls as
~ 0,⊥ and
required by our boundary conditions. From E0,z , we can obviously obtain H
~0,⊥ via Equations 9.283.
E
In general, the TM modes for any system require less work to determine because the
boundary condition is easier to calculate and enforce. Whenever you do a problem
that asks for both the TE and TM modes, do the TM modes first (in contrast to what
we have done here!).
~ = 1 b ~
H k ×E (9.301)
Zµ
We thus know the electrostatic and magnetostatic fields for this configuration have
the form
coaxial
waveguide ~0,⊥ = A sb
E ~ 0,⊥ =
H
A b
φ (9.302)
TEM mode s Zµ s
where the relative normalization has been determined from the aforementioned
relation between the fields.
Note that such cables do also support TE and TM modes; see, for example,
http://www.microwaves101.com/encyclopedias/coax#TE11.
~∗ × H
Let’s evaluate E ~ for a particular mode TM mode n (remember, H0,z = 0!):
~∗ × H
E ~ = E ∗ zb × H
~ 0,⊥ + E ~ ∗ × H0,z zb + E~∗ × H ~ 0,⊥
0,z 0,⊥ 0,⊥
1 i kµ
TM,n ∗
= E0,z,n zb × zb × ∇~ ⊥ E0,z,n
Zµ γn2
TM ∗
i kn (ω) ~ i kµ
+ ∇⊥ E0,z,n × ~ ⊥ E0,z,n
zb × ∇
2
γn Zµ γn 2
The expression is complicated because the fields are not purely transverse and there
are phase shifts among the components.
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 709
Section 9.8 Electromagnetic Waves: Waveguides
~ ⊥ = 0, we
Using the BAC − CAB rule for the triple-vector products, and also zb · ∇
obtain (now showing the TE analogue)
The first term points transverse to the direction of propagation. Since the walls are
perfectly conducting, there can be no time-averaged, net energy flow in that direction.
So, when we time-average, we neglect it. That leaves us with
TM TE
~ TM = zb ω kn (ω) ∇ ~ TE = zb ω kn (ω) µ ∇
D E 2 D E 2
S ~ ⊥ E0,z,n S ~ ⊥ H0,z,n (9.306)
n n
2 γn4 2 γn4
~ instead of B
Notice how the use of H ~ makes the expressions for the two modes very
similar in form.
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 710
Section 9.8 Electromagnetic Waves: Waveguides
D E
Let’s integrate S~ · zb over the waveguide cross-sectional area to get the total power.
Using ψn to represent E0,z,n or H0,z,n as appropriate, we can manipulate the area
integral using the two-dimensional analogue of Green’s First Identity (Equation 3.12):
Z 2 I Z
~ ~ ⊥ ψn −
da ∇ ⊥ ψn = d` ψn∗ nb · ∇ da ψn∗ ∇2⊥ ψn (9.307)
S C(S) S
(We are considering power in a single mode for now. We’ll generalize shortly.)
~ ⊥ ψn always
The first term vanishes due to our boundary conditions: either ψn or nb · ∇
vanishes on the boundary. The second term can be transformed using the
eigenvector-eigenvalue equation that yields the solutions ψn . Thus,
Z 2 Z
da |ψn |2
~ 2
da ∇ ⊥ ψn = γn (9.308)
S
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 711
Section 9.8 Electromagnetic Waves: Waveguides
s
2
1 ω knTM (ω) 1 ω2 ωc,n
D E Z Z
2
PnTM = da |E0,z,n | = vµ 2 1− da |E0,z,n |2 (9.309)
2 γn2 S 2 ωc,n ω2 S
s
2
1 ω knTE (ω) 1 ω2 ωc,n
D E Z Z
2
PnTE = µ da |H0,z,n | = vµ 2 1− µ da |H0,z,n |2 (9.310)
2 γn2 S 2 ωc,n ω2 S
(Note the symmetry between E and H in the above; the expressions would look more
different if B were used, the µ would be in the denominator.) If we do similar integrals
to calculate the time-averaged energy per unit length in the waveguide, and we obtain
D E 1 ω2 Z D E 1 ω2 Z
UnTM = 2
da |E0,z,n |2 UnTE = 2
µ da |H0,z,n |2 (9.311)
2 ωc,n S 2 ωc,n S
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 712
Section 9.8 Electromagnetic Waves: Waveguides
What about when one has a wave with power in multiple modes? Do the above
formulae hold? Yes. Suppose one has power in modes a and b. Then the
time-averaged Poynting vector will be
D E 1 D ∗ E
~ = R
S ~a + cb E
ca E ~b × ca H
~ a + cb H
~b (9.312)
2
where ca and cb are the coefficients of the two modes. Consider one of the cross
terms, under the assumption that both a and b are TM modes, and apply the same
kind of algebra we used before to obtain:
Again, the first term does not yield propagating power, so we neglect it. When we
integrate the second term over the waveguide cross section, we can see it vanishes via
the same kind of application of the eigenvector-eigenvalue equation:
Z Z
~ ⊥ ψ∗ · ∇
∇ ~ ⊥ ψb = γ 2 da ψa∗ ψb = 0 (9.314)
a b
S S
where the final vanishing occurs because the solutions of any eigenvector-eigenvalue
equation form an orthonormal basis and the integration over the cross section is the
appropriate inner product (like hψa |ψb i in quantum mechanics).
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 713
Section 9.8 Electromagnetic Waves: Waveguides
Via a similar calculation, one can show the other cross term vanishes when integrated
over the cross section S. This leaves only the non-cross-terms, which yield the power
in the individual modes. That is:
D E D E D E
P TM = |ca |2 PaTM + |cb |2 PbTM (9.315)
One can show the same thing via a similar technique when a and b are both TE
modes. When one is TE and one is TM, we apply the same technique yet again.
Orthogonality still applies: the TE and TM modes solve the same
eigenvector-eigenvalue equation but with different boundary conditions, so any pair of
TE and TM modes must be orthogonal. Thus, we have the general result
X D E X D E
TM 2 TE
hPi = |cm | Pm + |cnTE |2 PnTE (9.316)
m n
where cmTM and c TE are the coefficients of the TM,m and TE,n modes. The same of
n
course holds for the energy density:
X D E X D E
TM 2 TE
hUi = |cm | Um + |cnTE |2 UnTE (9.317)
m n
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 714
Section 9.8 Electromagnetic Waves: Waveguides
It is interesting here that we do not have hPi = vµ hUi as we had in free space, or even
hPi = vn (ω)hUi as we might expect given the guided wave speed. This makes sense,
as vn (ω) > vµ , so vn (ω) can exceed the speed of light. The quantity relating hPi and
hUi is
s
2
ωc,n dω
vµ 1 − 2 = (9.318)
ω dkn (ω)
s
2
ωc,n
2 dω 2 kn (ω)
2 ω dω = vµ (2 kn dkn + 0) =⇒ = vµ = vµ 1 − 2 (9.319)
dkn ω ω
This speed is the group velocity, vg ,n (ω). You are aware from Ph2a/12a how the
group velocity is the speed at which a wave packet propagates. The group velocity
here can never exceed vµ , thus preventing the power flow from going faster than the
speed of light. So, we have
s
2
ωc,n 2
hPn i = vg ,n (ω)hUn i vg ,n (ω) = vµ 1 − 2 vg ,n (ω) vn (ω) = vµ (9.320)
ω
Section 9.8.12 Energy, Poynting Vector, and Group Velocity Page 715
Lecture 42:
Electromagnetic Waves XI:
Waveguides with Finite Conductivity
Date Revised: 2022/05/02 07:00
Date Given: 2022/05/02
Page 716
Section 9.8 Electromagnetic Waves: Waveguides
I The conductivity is infinite, so, even when E ~ is varying, the surface charge
~c , the field inside the conductor, vanishes. Our
redistributes itself so that E
boundary conditions from Gauss’s Law and Faraday’s Law imply that there is a
~ due to this surface charge density while tb · E
discontinuity in nb · E ~ is continuous,
so the electric field outside the conductor is normal to its surface. Our TE and
TM mode boundary conditions are consistent with these conditions.
I Let’s revisit the continuity of, and thus vanishing of, the tangential component
~ in light of the surface current density. This continuity continues to hold:
of E
~ if the
Faraday’s Law tells us that there will only be a discontinuity in tb · E
s = tb × nb) has δ-function behavior at the
magnetic field in the sb direction (b
surface, and we have argued many times before that fields cannot have such
singularities (though charge and current densities can).
I This surface current density’s magnetic field cancels that of the external field to
ensure no magnetic field propagates into the interior of the conductor (much the
same way as the surface charge density does this for the electric field).
The above fact that magnetic fields vanish in perfect conductors when there is no free
current source in the conductor has been hinted at before in homework and in our
discussion of TEM modes for transmission lines, but we have never discussed it in as
general a setting as this before. For our waveguide mode derivation, we only required
~ ⊥ H0,z vanish at the boundary.
nb · ∇
Because of the importance of the surface current density, we will continue to work
~ rather than B,
with H ~ as we have been doing for some time.
Now, let’s consider a waveguide with walls having good conductivity. There are two
adjustments to be made:
I Moreover, because σ is finite, Ohm’s Law is applicable and thus we also know
that a δ-function-like surface current density is not physically allowed because it
would require a δ-function-like electric field for finite σ. Therefore, there is no
surface current to shield the conductor from the magnetic field in the waveguide
medium, while there is nonzero J driven by Ec . Nonzero J and finite σ therefore
imply there is power dissipation in the walls due to the currents flowing in a
resistive material, and thus there must be attenuation of the wave.
It is quite difficult to include these effects exactly, but we can derive some approximate
results in the case of a good conductor.
Specifically, we assume a successive approximation scheme where the fields outside the
conductor may now have tb · E ~ and nb · B
~ components, but they are small compared to
~ and tb · B.
nb · E ~ We obtain the dominant components of the field outside the walls
from the boundary value problem solutions for perfectly conductive walls (our
waveguide solutions), we use the good-conductor boundary conditions to obtain from
them the dominant components of the fields inside the walls, and then we again use
the good-conductor boundary conditions to obtain from these the corrections to the
fields outside the walls, and finally we use the boundary conditions again to find the
correction fields inside the walls. We find these correction fields in the waveguide are
indeed small compared to the perfect-conductor fields in the waveguide, justifying the
approximation scheme. More importantly, finding the fields inside the walls allows us
to calculate the Joule dissipation and thus the attenuation of the wave in the
waveguide. For the latter, after zooming in to find the fields in the conducting walls
and the correction fields, we will zoom back out, treating the quickly decaying current
density as a perfect sheet current and the fields as going to zero perfectly.
Let’s also think about the propagation directions inside and outside the conductor.
Outside the conductor, the E ~ field is close to normal to the surface because tb · E
~ ≈0
~ 6= 0. That is, θi ≈ π/2: we have glancing incidence! You will show in
while nb · E
Ph106c Problem Set 4 #1 that the general expression for the propagation vector
inside a good conductor is
~ 1+i ω
kt = −b
n +w
b sin θi (9.321)
δ vµ
where the first term has −b n rather than nb because we use here a direction convention
for nb opposite
p to that in Problem Set 4. Since ω/vµ = kµ and
1/δ → kµ σ/(2 ω) kµ in the good conductor limit (Equation 9.140), we see
that kbt ≈ −bn even if θi = π/2! Thus, we are assured that the propagation direction in
the conductor is normal to the interface. (Note, for our waveguide solutions, k < kµ ,
so the above dominance of the normal component holds even accounting for the full
frequency dependence of k from our discussion of dispersion relations.)
Section 9.8.14 Fields Near the Walls for Waveguides with Finite Conductivity Page 722
Section 9.8 Electromagnetic Waves: Waveguides
With this information about the propagation direction in the conductor, we define ξ to
be the coordinate along kbt = −b
n (with the minus sign because we have defined nb to
point out of the conductor for waveguides) and η to be the coordinate perpendicular
to kbt (along the waveguide boundary). Then, we have (with 0 subscripts used so we
can avoid writing everywhere the e i(k z−ω t) factor)
~ 0,c (ξ, η) = H
H ~ 0 (ξ = 0, η) e −ξ/δ e i ξ/δ (9.322)
~0,c (ξ, η) = 2 Zσ nb × H
E ~ 0,c (ξ, η) = 2 Zσ nb × H
~ 0 (ξ = 0, η) e −ξ/δ e i ξ/δ (9.323)
1+i 1+i
where again nb points outward from the conductor into the waveguide medium. We
thus have the dominant components of the fields inside the conductor.
Section 9.8.14 Fields Near the Walls for Waveguides with Finite Conductivity Page 723
Section 9.8 Electromagnetic Waves: Waveguides
Rewriting these fields with the full spatial and time dependence:
~ c (ξ, η, z, t) = H
H ~ 0 (ξ = 0, η) e −ξ/δ e i ξ/δ e i(k z−ω t) (9.324)
~c (ξ, η, z, t) = 2 Zσ nb × H
E ~ 0 (ξ = 0, η) e −ξ/δ e i ξ/δ e i(k z−ω t) (9.325)
1+i
Using boundary conditions to connect the fields inside the conductor to those outside,
we may now calculate the correction terms to the fields outside the conductor. The
~c (ξ = 0, η, z, t) by
tangential electric field outside the conductor is given by E
continuity of this component. The normal component of H ~ outside the conductor is
found using Faraday’s Law (now referring to the full fields, without 0 subscripts,
because we need the full spatial and time dependence for Faraday’s Law):
~
∂H
−µ nb · = nb · ∇ ~
~ ×E (9.326)
∂t
ξ=0
ξ=0
For evaluating the curl, consider a temporary primed coordinate system, distinct from
the one we have been using from the waveguide, with zb 0 = nb, xb 0 ∝ H ~ 0,c (ξ = 0, η),
~0,c (ξ = 0), η) ∝ nb × H
and yb 0 ∝ E ~ 0,c (ξ = 0, η). In this coordinate system,
∂Ey 0 ∂Ex 0
nb · ∇ ~ = zb 0 · ∇
~ ×E ~ =
~ ×E − = x 0 ~
· ∇ y 0 ~
· E − y 0 ~
· ∇ x 0 ~
· E
∂x 0 ∂y 0
b b b b
(9.327)
Section 9.8.14 Fields Near the Walls for Waveguides with Finite Conductivity Page 724
Section 9.8 Electromagnetic Waves: Waveguides
~ = 0, t) = − µc δ
h i
nb · H(ξ h ~
b·∇ ~ = 0)
b · H(ξ
h (9.329)
µ 1−i
where we have defined h b = xb 0 to be the tangential unit vector at the boundary that is
~ at the boundary (which will be along tb⊥ = zb × nb for a TM mode but will
parallel to H
be some linear combination of tb⊥ and zb for a TE mode).
What is the typical magnitude of the normal magnetic field? h b·∇~ has two
components: one along tb⊥ and one along zb. The zb piece will yield something
proportional to k, while the tb⊥ piece will yield something proportional to γn , where
γn2 = kx2 + ky2 . Since k 2 + γn2 = kµ
2 , the h ~ derivative will always bring in factor
b·∇
scaling like kµ . (Note that it is kµ that appears, not k, so we need not worry about
k → 0 as ω → ωc,n .) Thus, the magnitude of the above gradient is given by kµ H0,z ,
so the ratio of the normal component of the magnetic field to the tangential
component is δ/λµ , which is small, as we expected.
Section 9.8.14 Fields Near the Walls for Waveguides with Finite Conductivity Page 725
Section 9.8 Electromagnetic Waves: Waveguides
Aside from seeing they are small, it is interesting to look at the magnitudes of the
correction fields relative to the original waveguide fields more carefully. We have, with
all fields evaluated right at the boundary (ξ = 0):
Hc = H0
Zσ Zσ Zµ,c Zσ Zµ,c
Ec ≈ Zσ H0 = Zµ,c H0 = Zµ H0 ≈ E0
Zµ,c Zµ,c Zµ Zµ,c Zµ
Zσ
∼ E0 E0
Zµ,c
~ = Ec E0
Et ≡ tb · E
C µc
~
Hn ≡ nb · H ≈
δ kµ H0 ∼ δ kµ H0 H0
C µ
r r r
Zσ 1 c 1 µc σ ω c c ω ωτ kµ,c δ
r r
Next, using = = = = =
Zµ,c σ δ µc σ 2 µc 2σ 2 2
Et Zσ Zµ,c kµ,c δ Zµ,c kµ δ kµ,c Zµ,c
we see = = =
E0 Zµ,c Zµ 2 Zµ 2 kµ Zµ
kµ δ c µc µc µc kµ δ Hn
r
= = ≈ as we might have expected!
2 µ c µ µ 2 H0
Section 9.8.14 Fields Near the Walls for Waveguides with Finite Conductivity Page 726
Section 9.8 Electromagnetic Waves: Waveguides
dhPJ i D ~∗ ~ E σ ~ 2 σ ω 2 δ 2 2 ~ 2
= J ·E = E0,c (ξ, η) = µc H0 (ξ = 0, η) e −2ξ/δ (9.330)
dτ 2 2 2
where the z and t dependence have been eliminated when we take the magnitude,
leaving only the normalizing 0 fields that depend on the transverse coordinates (ξ, η).
(Note that we have used the definitions of Zσ and δ to manipulate the prefactor.)
Section 9.8.15 Power Flow for Waveguides with Finite Conductivity Page 727
Section 9.8 Electromagnetic Waves: Waveguides
When we integrate this over depth to get the Joule dissipation per unit area,
dξ e −2ξ/δ yields δ/2. After some algebra (recall, δ = 2/µc σ ω and
R p
dhPJ i 1 ~ 2 Zσ ~ 2
= H0 (ξ = 0, η) = K (η) (9.331)
da 2σδ 2
where we have used the fact that, for the perfect conductor waveguide solution, the
tangential component of the auxiliary field gives the surface current density. (There is
no inconsistency in reintroducing K ~ at this point: recall that K
~ is the σ → ∞, δ → 0
~ Or, looked at from a length scale large compared to δ, the quickly
limit of δ J.
decaying J~ looks like an infinitesimally thin sheet of current. From that same
perspective, we recover the perfect conductor limit because the exponential decay of
the fields is compressed to zero thickness, yielding an apparent discontinuity in the
fields at the conductor boundary.) Looks like I2 R/2!
Integrating over the boundary yields the rate of power loss per unit length along the
waveguide
Section 9.8.15 Power Flow for Waveguides with Finite Conductivity Page 728
Section 9.8 Electromagnetic Waves: Waveguides
i kµ
K0,z (η) = H(ξ = 0, η) = b⊥ · zb × ∇
t ~ ⊥ E0,z (9.333)
Zµ γn2
ξ=0,η
i kµ
~ ⊥ E0,z
i kµ ~ ⊥ E0,z
= tb⊥ × zb · ∇ = n · ∇ (9.334)
Zµ γn2 2
b
ξ=0,η Z µ γn
ξ=0,η
where we used the cyclicity of theptriple scalar product and tb⊥ = zb × nb. Plugging into
Equation 9.332 and using Zµ = µ/ = µ vµ , kµ = ω/vµ , and ωc,n = vµ γn
(putting the n subscripts in now and not showing the (ξ, η) dependence for brevity):
d PnTM Zσ ω 2
1 2
I
~
− = dη nb · ∇ ⊥ E0,z,n (9.335)
dz 2
2 ωc,n µ2 ωc,n
2
C(S)
One can obtain the corresponding expression for TE modes (which has two terms, one
~ components):
each for the currents associated with the transverse and longitudinal H
" ! #
2 2 ω2
d PnTE Zσ ω 2
vµ ωc,n 2
I
− = dη 1− ~ ⊥ H0,z,n + c,n |H0,z,n |2
nb × ∇
dz 2
2 ωc,n 2
ωc,n ω2 ω2
C(S)
(9.336)
Section 9.8.15 Power Flow for Waveguides with Finite Conductivity Page 729
Section 9.8 Electromagnetic Waves: Waveguides
To calculate the power loss per unit length explicitly, we must evaluate the above
integrals using the known waveguide solutions, so the exact result will depend on the
waveguide geometry. However, we can determine the approximate size of the above
expressions generically. Recall that, for perfectly conducting walls, the
eigenvalue-eigenvector equation implies (see Equations 9.307 and 9.308)
Z 2 Z
da |ψn |2
~ 2
da ∇ ⊥ ψn = γn (9.337)
S
where the averages are over the boundary or the surface, not time. Thus, we can
relate the line integral over the boundary to an area integral over the cross section:
dη ~ 2 C
I Z I Z
nb · ∇⊥ E0,z,n = ξn da |E0,z,n |2 with C = dη A= da (9.338)
2
γn A S
C(S) C(S) S
where ξn is a fudge factor that has to be calculated for the specific geometry and is
defined by the above equations. There is a similar expression for the TE modes. The
area integral on the right is related to the power flow down the waveguide,
Equation 9.309.
Section 9.8.15 Power Flow for Waveguides with Finite Conductivity Page 730
Section 9.8 Electromagnetic Waves: Waveguides
In fact, by combining Equations 9.309 and Equation 9.310 and the analogue of the
above equation for TE modes, one can show that the field decay constant is
q
ω " #
2
ωc,n
1 dhPn i Zσ (ωc,n ) C ωc,n
κn = − = r ξn + ηn 2 (9.339)
2hPn i dz Zµ 2A ω2 ω
1 − ωc,n
2
where Zσ is evaluated at ωc,n , ξn is the order unity quantity defined above for TM
modes (and can be defined analogously for TE modes), and ηn is an analogous
quantity that is nonzero only for TE modes (corresponding to the second term in the
2 /ω 2 , which comes from the current associated with
TE expression, proportional to ωc,n
the longitudinal field). For any given waveguide geometry, one can find ξn , ηn , and κn
by evaluating the explicit dhPn i/dz and hPn i formulae. Note that ξn and ηn are
unrelated to the earlier (ξ, η) coordinates.
The useful and interesting thing about the above expression is that it exhibits all the
frequency dependence. ξn and ηn are unitless, O(1) factors that depend on the
solution to the eigenvector-eigenvalue equation, which is geometry- but not
frequency-dependent.
Section 9.8.15 Power Flow for Waveguides with Finite Conductivity Page 731
Section 9.8 Electromagnetic Waves: Waveguides
Summing everything up, the fields in the waveguide medium including the decay along
the waveguide are (kn and κn functions of ω):
~n (~
E ~0,n (~
r , t) = E r⊥ ) e i(kn (ω) z−ω t) e −κn (ω) z (9.340)
~ n (~
H ~ 0,n (~
r , t) = H r⊥ ) e i(kn (ω) z−ω t)
e −κn (ω) z
(9.341)
r ∈ C)
rC is a vector ~
The fields in the walls are (~
We do not work out the full spatial dependences of the correction fields in the
waveguide medium, i.e., the correction fields that yield nonzero (bn × h) ~ and nb · H
b ·E ~
at the waveguide walls, because that would require an additional step of re-solving the
eigenvalue-eigenvector equations with a correction term to the boundary conditions, so
~0,n and H
the transverse spatial dependences E ~ 0,n in the waveguide medium remain the
perfect waveguide solutions.
Section 9.8.15 Power Flow for Waveguides with Finite Conductivity Page 732
Section 10
Potentials Revisited
10.1 Potential Formulation
10.2 Retarded Potentials and Fields
Page 733
Lecture 43:
Potentials Revisited:
Potential Formulation
Retarded Potentials
Date Revised: 2022/05/04 07:00
Date Given: 2022/05/04
Page 734
Section 10.1 Potentials Revisited: Potential Formulation
Potential Formulation
Revisiting Scalar and Vector Potentials
However, we have not revisited the static potentials that we developed in the static
field cases. Let’s do that.
~ =∇
B ~
~ ×A (10.1)
~ 6= 0 when B
~ ×E
However, Faraday’s Law implies that ∇ ~ has time dependence.
Therefore, we cannot assume E~ = −∇V ~ =∇
~ . However, using B ~ we see that
~ × A,
!
~ ~
~ = −∂B = − ∂ ∇ ~ + ∂A
~ ×E
∇ ~
~ ×A =⇒ ~ ×
∇ E =0 (10.2)
∂t ∂t ∂t
~ ~
~ + ∂ A = −∇V
E ~ =⇒ E ~ − ∂A
~ = −∇V (10.3)
∂t ∂t
~
Aside: The interpretation is that, by subtracting off ∂ A/∂t, we remove the
nonconservative part of the electric field, leaving a conservative part that can be
described by a scalar potential. Moreover, we now have a very straightforward way to
calculate the Faraday’s Law E~ind : it is just −∂ A/∂t!
~
~
∇ ~ =0
~ ·B ~ + ∂B = 0
~ ×E
∇ (10.4)
∂t
Let’s check whether Gauss’s Law and Ampere’s Law can be satisfied. In terms of the
potentials, Gauss’s Law becomes
ρ ~ = −∇2 V − ∂ ∇
~ ·E
~
~ ·A
=∇ (10.5)
o ∂t
~ 2~
~ ∂ V − o µo ∂ A
~ = µo J~ + o µo ∂ E = µo J~ − o µo ∇
~ × ∇
∇ ~ ×A (10.6)
∂t ∂t ∂t 2
~ × (∇
which we may rewrite using the usual ∇ ~ ×~ ~ ∇
a) = ∇( ~ ·~
a ) − ∇2 ~
a identity:
!
∂2A~
~ + o µo ∂ V = −µo J~
2~ ~ ∇ ~ ·A
∇ A − o µo −∇ (10.7)
∂t 2 ∂t
Equations 10.5 and 10.7 are second-order coupled partial differential equations for V
~ and, so far, we see no reason they cannot be solved.
and A
We can write Gauss’ Law (Equation 10.5) and Ampere’s Law (Equation 10.7) in terms
of the potentials in the symmetrical form (defining the d’Alembertian 2 with the
opposite sign convention as Griffiths to be consistent with our later discussion of
special relativity):
∂L ρ ~ + ∇L
~ = µo J~
2 V − = 2 A (10.8)
∂t o
∂2 ~ + o µo
~ ·A ∂V
with 2 ≡ o µo − ∇2 L≡∇ (10.9)
∂t 2 ∂t
Gauge freedom answers the question: how much can we change A ~ and V while still
~ and
obtaining the same fields? That is, let us assume two sets of potentials (V , A)
~ 0 ) that differ by functions α
(V 0 , A ~ and β:
~ 0 (~
A ~ r , t) + α
r , t) = A(~ ~ (~
r , t) V 0 (~
r , t) = V (~
r , t) + β(~
r , t) (10.10)
The requirement that these two sets of potentials yield the same fields gives the
equations:
h i
0=∇ ~ × A ~ 0 (~ ~ r , t) = ∇
r , t) − A(~ ~ ×α ~ (~r , t) (10.11)
~ V (~
0 ∂ h
~ 0 (~ ~ r , t)
i
0 = −∇ r , t) − V (~r , t) − A r , t) − A(~
∂t
~ ~ (~
∂α r , t)
= −∇β(~ r , t) − (10.12)
∂t
~ (~
α ~ r , t)
r , t) = ∇λ(~ (10.13)
r , t)
~ β(~ ∂ λ(~
∇ r , t) + =0 (10.14)
∂t
r , t)
∂ λ(~
r , t) +
β(~ = k(t) (10.15)
∂t
Rt
We can absorb k(t) into λ: add to λ the position-independent term 0 dt 0 k(t 0 ),
whose gradient vanishes and thus does not affect α(~
r , t). Thus
r , t) = −∂λ(~
β(~ r , t)/∂t and we have the relations
~ 0 (~ ~ r , t) + ∇λ(~
~ r , t) r , t)
∂ λ(~
A r , t) = A(~ V 0 (~ r , t) −
r , t) = V (~ (10.16)
∂t
Coulomb Gauge
This is the gauge we chose earlier for magnetostatics,
ρ
Coulomb Gauge ∇2 V = − (10.18)
o
That is, the charge density sets the potential in the same way as in electrostatics, so
changes in charge density propagate into the potential instantaneously. Of course, you
know from special relativity that this is not possible. We will see that there are
~
corrections from ∂ A/∂t that prevent E ~ from responding instantaneously to such
changes. The differential equation for A ~ becomes
∂2A~
∂V
Coulomb Gauge ~ − o µo
∇2 A = −µo J~ + o µo ∇
~ (10.19)
∂t 2 ∂t
Lorenz Gauge
The reason this choice is made is that it sets L = 0 in Equation 10.9 so that we obtain
symmetric, inhomogeneous (have source terms) wave equations for V and A: ~
ρ ~ = µo J~ ∂2
Lorenz Gauge 2 V = 2 A 2 ≡ o µo − ∇2 (10.21)
o ∂t 2
with, again, the opposite sign convention as Griffiths for 2 . This will be the natural
gauge for calculating radiation of EM waves from moving charges and currents,
especially in a manner that is manifestly consistent with relativity.
The Lorentz Force Law in Potential Form and Implications for Classical and
Quantum Mechanics (Skip)
Let’s write the Lorentz Force Law in terms of the potentials:
" #
~
dp ~
~ − ∂A + ~
F~ = = q −∇V v× ∇ ~
~ ×A (10.22)
dt ∂t
Using the BAC − CAB rule for the triple vector product and recognizing ~
v is not a
function of position, we may rewrite as
" #
d~
v ~
∂A
m = −q ~ ~ ~
v ·∇ A+∇ V −~
+ ~ ~
v ·A (10.23)
dt ∂t
~
∂A ~
~ = dA
We may write + ~ ~ A
v ·∇ (10.24)
∂t dt
Let’s be explicit about why the convective derivative correctly gives the total
~ Consider a time interval dt in which the particle moves d~
derivative of A. r =~v dt:
~ = A(~
dA ~ r +~ ~ r , t)
v dt, t + dt) − A(~ (10.25)
" #
~
∂A ~
∂A ~
∂A ~
∂A ~
~ + ∂A
= vx dt + vy dt + vz dt + dt = dt ~ ~ A
v ·∇ (10.26)
∂x ∂y ∂z ∂t ∂t
d
~ = −∇
h
~ q V −~ ~
i
m~
v +qA v ·A (10.27)
dt
d~
π ~ v ~
h
~ r , t)
i
= −∇U π = m~
~ v +qA r, ~
Uv (~ r , t) − ~
v , t) = q V (~ v · A(~ (10.28)
dt
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 744
Section 10.1 Potentials Revisited: Potential Formulation
You may have seen this idea of a velocity-dependent potential in Ph106a. Recall that
the Lagrangian formulation of mechanics is derivable as a special case, for
conservative forces, of the generalized equation of motion:
!
(nc) ri
∂~ d ∂T ∂T
F~i
X
· = Fk = − (10.29)
i
∂qk dt ∂ q̇k ∂qk
(nc)
where the F~i are the set of non-constraint forces, the qk are generalized
coordinates, and T is the kinetic energy in terms of the generalized coordinates and
velocities. For a conservative force, the force term on the left side can be written as a
gradient of a potential energy U that is a function only of the coordinates. This form
permits one to define the Lagrangian function L = T − U and incorporate the gradient
of U into the second term on the right side. (Such a U does not contribute to the first
term because it does not depend on the generalized velocities.) The result is the
Euler-Lagrange Equation
!
d ∂L ∂L
− =0 (10.30)
dt ∂ q̇k ∂qk
For nonconservative, velocity-dependent forces like the Lorentz Force, it is clear that,
if one can construct the appropriate function Uv such that its partial derivatives with
respect to qk and q̇k generate the appropriate force terms on the left side, then the
same construction applies, now with L = T − Uv and Uv being a function generalized
coordinates and velocities, a velocity-dependent potential (energy).
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 745
Section 10.1 Potentials Revisited: Potential Formulation
Let’s see how it works out in detail in this case, which will also lead us to the
appropriate Hamiltonian operator for quantum mechanics in the presence of
electromagnetic fields. Using the velocity-dependent potential energy we defined in
Equations 10.28, we write down the corresponding Lagrangian:
1 h
~ r , t)
i
L = T − Uv = m v 2 − q V (~
r , t) − ~
v · A(~ (10.31)
2
~
where we used the expression we derived earlier for the total derivative of A.
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 746
Section 10.1 Potentials Revisited: Potential Formulation
Moving terms around and being careful not to reorder derivatives now that ~
r and
v = ~r˙ are no longer independent, we have
~
3 3
" # " #
∂V ∂ Ak X ∂ An X ∂ Ak
m r¨k = q − − +q r˙n − r˙n (10.34)
∂rk ∂t n=1
∂rk n=1
∂rn
The term in the second set of brackets on the left can be rewritten using the
component version of the BAC − CAB rule:
" #
∂V ∂ Ak h i
m r¨k = q − − + q ~r˙ × ∇ ~
~ ×A (10.35)
∂rk ∂t k
Therefore,
" #
~
∂A
m ~¨
r =q ~ −
−∇V + ~r˙ × ∇ ~
~ ×A (10.36)
∂t
as we expect: both the Lorentz Force Law and a second-order ODE for the particle
trajectory ~
r (t).
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 747
Section 10.1 Potentials Revisited: Potential Formulation
Let’s also calculate the canonical momentum and the Hamiltonian. Normally, one
writes the Hamiltonian in terms of the canonical momentum and the generalized
coordinates, not the generalized coordinates and velocities, but we will do the latter
first so we can see that it yields the correct total energy:
∂L
πk = = m r˙k + q Ak (10.37)
∂ q̇k
πk q̇k − L = m ~r˙ + q A
~ · ~r˙ − L
X
H= (10.38)
k
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 748
Section 10.1 Potentials Revisited: Potential Formulation
Let’s rewrite the Hamiltonian as it should be done, in terms of the canonical
momentum, so that we can see how one again recovers the Lorentz Force Law as well
as obtains the quantum
mechanical
Hamiltonian operator. Start from the second line
above, using ~v = ~π−qA ~ /m:
1 1
~ − q V − 1 ~
2
π·
H =~ π−qA
~ ~ −
π−qA ~ ·A
π−qA ~ (10.40)
m 2m m
~
1 2
~ + q V
= π−qA (10.41)
2m
~
3
dπk ∂H 1 X ∂ ∂V
π̇k = =− =− (πn − q An ) (πn − q An ) − q
dt ∂rk m n=1 ∂rk ∂rk
3
q X ∂ An ∂V
= (πn − q An ) −q
m n=1 ∂rk ∂rk
drk ∂H 1
r˙k = = = (πk − q Ak )
dt ∂πk m
In calculating the above, one must remember that πk and rj are independent variables,
∂πk /∂rj = 0, during the process of taking the derivatives in Hamilton’s Equations
(similar to the way rk and r˙k are independent in the process of taking the derivatives
to obtain the Euler-Lagrange Equation).
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 749
Section 10.1 Potentials Revisited: Potential Formulation
After we take the derivatives, though, we drop this independence and substitute the
second equation in the first one to eliminate πk and obtain a second-order ODE in rk :
3
d X ∂ An ∂V
(m r˙k + q Ak ) = q r˙n −q
dt n=1
∂rk ∂rk
d ∂ h ~ = − ∂ Uv
i
(m r˙k + q Ak ) = − q V − ~r˙ · A
dt ∂rk ∂rk
which recovers Equation 10.28 with the understanding that derivatives with respect to
rk should not be interpreted as acting on r˙k in order to get the last equation.
The greater point of interest about writing out the Hamiltonian is to see the
motivation for the quantum mechanical Hamiltonian. To go from classical mechanics
to quantum mechanics, we promote the canonical momentum to the momentum
~ (or, more correctly, we promote ~
operator −i ~ ∇ π to the quantum mechanical
momentum operator Π, ~ whose representation in the position basis is −i ~ ∇).
~ The
above derivation shows why the Hamiltonian in the presence of an arbitrary EM field
~ in it, which otherwise goes unexplained in quantum mechanics.
has A
The presence of A ~ in the Hamiltonian is also why the vector potential plays a more
important role in quantum mechanics than it does in classical mechanics. Its presence
in H gives it the opportunity to affect the phase of the quantum mechanical state.
Section 10.1.5 The Lorentz Force Law in Potential Form and Implications for Classical and Quantum Mechanics (Skip) Page 750
Section 10.1 Potentials Revisited: Potential Formulation
We can also find an equation relating the time derivative of the particle energy to the
potentials. This will become useful when we consider the relativistic generalization of
the Lorentz Force Law, which will include an energy component. We obtain this by
calculating the work done per unit time by the Lorentz Force:
" #
dT ~
∂A n o
=~ ~
v ·F =q ~ ~
v · −∇V − ~
v· v· ~
+~ ~
v × ∇×A~ (10.42)
dt ∂t
The last term vanishes because the triple cross product is perpendicular to ~
v,
reflecting the fact that magnetic fields can do not work. Now, let’s add the total
derivative of q V ,
d
~ (q V ) + ∂ (q V )
(q V ) = ~v ·∇ (10.43)
dt ∂t
d ∂ h ~ = ∂ Uv
i
(T + q V ) = q V −~
v ·A (10.44)
dt ∂t ∂t
We begin by showing that one can derive the Schrödinger Equation by considering the
wavefunction ψ(~
r , t) to be a classical field with Lagrangian density (Lagrangian per
unit volume):
1 ∂ ∗ ∗ 1
∂
~
2
L= ψ∗ i ~ ψ+ψ i~ ψ − ψ ∗ −i ~ ∇ ψ (10.45)
2 ∂t ∂t 2m
With a Lagrangian density, we consider the field at every point in space ψ(~ r , t) to be
an independent trajectory in time, analogous to ~ x (t) for a single particle. Therefore,
r , t) with ~
our generalized coordinate is ψ(~ r acting as a label or index, with each point
in space corresponding to an independent variable. When one has a complex field, one
can show that one should actually use ψ ∗ as the independent variable for the
Euler-Lagrange Equation; that is, the Euler-Lagrange Equation is
d
∂L ∂L
− =0 (10.46)
dt ∂ ψ̇ ∗ ∂ψ ∗
d i~ i ~ ∂ψ ~2
− ψ − + ∇2 ψ = 0 (10.47)
dt 2 2 ∂t 2m
1 ~
2 ∂ψ
−i ~ ∇ ψ = i ~ (10.48)
2m ∂t
r , t) → ψ 0 (~
ψ(~ r , t) e i χ
r , t) = ψ(~ with χ a constant (10.49)
∂ ψ0 ∂L ∂ ψ 0 i~ ∗ ~2
Z Z Z Z
I= dτ pψ = dτ = dτ ψ (i ψ) = − r , t)|2
dτ |ψ(~
∂χ ∂ ψ̇ ∂χ 2 2
(10.50)
That is, the fact that the phase transformation is a symmetry transformation of the
Lagrangian density implies that the total probability of the wavefunction is conserved
(the negative sign is unimportant, as are the other prefactors).
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 753
Section 10.1 Potentials Revisited: Potential Formulation
It is a general principle of modern field theory that we should expect global symmetry
transformations like the one above to also be symmetry transformations of the
Lagrangian when the transformation is made local. One reason we make this
requirement is that a symmetry transformation that is the same everywhere in space is
inconsistent with special relativity: how does the information propagate
instantaneously to all points in space? We also know simultaneity is frame-dependent
in special relativity, which also calls into question the idea of applying a transformation
that is the same at all points in space. In this case, that would imply
r , t) → ψ 0 (~
ψ(~ r , t) e i χ(~r ,t)
r , t) = ψ(~ (10.51)
However, we quickly see that, when we try to apply such a local transformation to the
Lagrangian density, the spatial and time derivatives act on χ(~
r , t) and create all kinds
of additional terms such that L is not invariant under the local version of the
transformation. Specifically:
∂ ψ0 ∂ ψ iχ ∂χ
= e + i ψ e iχ (10.52)
∂t ∂t ∂t
h i h i
~ ·∇
∇2 ψ 0 = ∇ ~ ψ e iχ = ∇ ~ · e iχ ∇ψ
~ + i ψ e iχ ∇χ
~ (10.53)
iχ 2
= e ∇ ψ + 2i e iχ ~ · ∇χ
∇ψ ~ −e iχ ~ · ∇χ
ψ ∇χ ~ +iψe ∇ χ iχ 2
(10.54)
In each expression, the first term is what we had for the global version of the
transformation, but the additional terms break the symmetry.
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 754
Section 10.1 Potentials Revisited: Potential Formulation
There is a standard prescription for fixing this kind of problem. We need counterterms
to cancel the additional terms that appeared above. We make the replacement
p ~
~~r = π~r = −i ~ ∇ −→ ~ r , t) = −i ~ ∇
~~r = π~r − q A(~
p ~ r , t)
~ − q A(~ (10.55)
∂ ∂
pt = πt = i ~ −→ pt = πt − q V (~
r , t) = i ~ − q V (~
r , t) (10.56)
∂t ∂t
~ r , t) and V (~
where the new functions A(~ r , t) also participate in the local phase
transformation, now called, for obvious reasons, a gauge transformation:
r , t)
∂ λ(~
V 0 (~ r , t) −
r , t) = V (~ (10.57)
∂t
~ 0 (~
A ~ r , t) + ∇λ(~
r , t) = A(~ ~ r , t) (10.58)
with χ(~
r , t) = q λ(~
r , t)/~ the same function used in the transformation of the
wavefunction up to a constant factor, so the coupled gauge transformation of the
wavefunction is
r , t) → ψ 0 (~
ψ(~ r , t) e i q λ(~r ,t)/~
r , t) = ψ(~ (10.59)
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 755
Section 10.1 Potentials Revisited: Potential Formulation
With this replacement, the Lagrangian is
∗
1 1
∂ ∂ 2
~ ψ
~ −qA
L= ψ∗ i ~ −qV ψ+ψ i~ −qV ψ∗ − ψ ∗ −i ~ ∇
2 ∂t ∂t 2m
(10.60)
We can check this Lagrangian density is invariant under the gauge transformation.
The space piece is:
~ −qA
−i ~ ∇ ~ ψ −→
~ −qA
−i ~ ∇ ~ 0 ψ0 (10.61)
h q i h i
~ + e i q λ/~ ψ i
= −i ~ e i q λ/~ ∇ψ ~
∇λ −q A ~ + ∇λ
~ e i q λ/~ ψ
~
= e i q λ/~ −i ~ ∇~ −qA~ ψ (10.62)
2
=⇒ −i ~ ∇~ −qA ~ ψ −→
2 h i
~ −qA
−i ~ ∇ ~ 0 ψ 0 = −i ~ ∇ ~ −qA ~ 0 · e i q λ/~ −i ~ ∇ ~ −qA ~ ψ
h i
~ − i ~∇
= e i q λ/~ q ∇λ ~ −qA ~ − q ∇λ
~ · −i ~ ∇ ~ −qA ~ ψ
2
= e i q λ/~ −i ~ ∇~ −qA~ ψ (10.63)
The transformation yields just the simple phase factor in front that we need for the
cancellation to work.
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 756
Section 10.1 Potentials Revisited: Potential Formulation
Again, the transformation yields just the simple phase factor in front that we need for
the cancellation to work.
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 757
Section 10.1 Potentials Revisited: Potential Formulation
d i~ i ~ ∂ψ 1 1 1
2
− ψ − − qV ψ− qV ψ− −i ~ ∇ ~ ψ =0
~ −qA
dt 2 2 ∂t 2 2 2m
(10.68)
1 2
~ ψ+qV ψ = i~
~ −qA ∂ ψ
−i ~ ∇ (10.69)
2m ∂t
which is exactly the Schrödinger Equation with coupling of the particle of charge q to
an electrostatic potential V (~ ~ r , t)!
r , t) and a vector potential A(~
This is the same Schrödinger Equation that get if we promote the classical
Hamiltonian from Equation 10.41 to a quantum mechanical Hamiltonian operator by
~ (in the position basis), as indicated earlier .
π with −i ~ ∇
replacing ~
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 758
Section 10.1 Potentials Revisited: Potential Formulation
Lastly, with the coupling between charge and the potentials now present, we can
reinterpret our Noether’s Theorem result: not only does it imply conservation of
probability but, because the particle now has a charge associated with it, it implies
conservation of charge, too! This is conservation of integrated charge over all of space
for each particle, not local conservation of charge.
~j(~ i ~ h ∗~ ~ ∗
i
r , t) = − ψ ∇ψ − ψ ∇ψ
2 m
~ · ~j + ∂ |ψ|2 = 0
∇
∂t
which is the continuity equation for probability. If one multiplies all of the above by q,
one obtains the continuity equation for charge density too!
Section 10.1.6 Gauge Transformations and Coupling of Matter to Electromagnetic Fields (Skip) Page 759
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
|~
r −~r 0|
retarded time tr = t − (10.70)
c
1 ρ(~r 0 , tr ) ~ r 0 , tr )
J(~
~ r , t) = µo
Z Z
retarded
V (~
r , t) = dτ 0 A(~ dτ 0
potentials 4 π o V |~
r −~ r 0| 4π V |~
r −~ r 0|
(10.71)
We check below that these forms in fact solve the inhomogeneous wave equations.
r 0 , tr )
f (~
Z
r , t) =
φ(~ dτ 0 (10.72)
V |~r −~ r 0|
∂2 r 0 , tr )
f (~ 1 ∂2 1 ∂2
= r 0 , tr ) =
f (~ r 0 , tr )
f (~ (10.73)
∂t 2 |~r −~ r 0| |~
r −~r 0 | ∂t 2 |~
r −~r 0 | ∂tr2
because ∂∂ttr = 1. For the space term in 2 , we will need ∇2 f (~r 0 , tr ) . Calculating this
~
by brute force is difficult because ∇ is with respect to ~r but tr depends on |~ r −~r 0 |.
We’ll employ a similar trick as we did when doing such manipulations in electrostatics,
which is to define ~s = ~r −~r 0 (so that tr = t − s/c) and recognize that
r 0 , tr )~r = ∇~2s f (~
∇~2r f (~ r 0 , tr )~s =~r −~r 0
(10.74)
because the expression is a scalar and thus cannot depend on the coordinate system
origin aside from evaluating it at the correct physical location (the business with the
arguments in the subscripts). Note that ~s and ~ r 0 are independent variables, just as ~ r
and ~r 0 were independent, so ∇~2s does not act on the ~ r 0 dependence of f (~
r 0 , tr )!
Let’s evaluate the easier expression (recognizing tr only depends on s = |~s |):
1 ∂ 1 ∂ ∂ f ∂ tr
∂
∇~2s f (~
r 0 , tr ) = s2 r 0 , tr ) =
f (~ s2 (10.75)
s 2 ∂s
∂s s 2 ∂s ∂tr ∂s
1 ∂ ∂f 1 1 ∂2f 2 ∂f
= 2 s2 − = − (10.76)
s ∂s ∂tr c c 2 ∂tr2 s c ∂tr
Again, ∂/∂s did not act on the ~ r 0 dependence of f (~ r 0 , tr ) because the transformation
r, ~
of variables was from (~ 0
r , t) to (~s , ~ 0
r , t) and so the independence of ~ r and ~r0
translates to independence of ~s and ~ r 0 . From the above, we obtain the expression we
will need below:
∇~2r f (~
r 0 , tr ) 1 ∂2f 1 2 ∂f 1
= − (10.77)
|~ r 0|
r −~ c 2 ∂tr2 |~
r −~r 0| c ∂tr |~
r −~r 0 |2
1 ~ ~s 1
h i h i
2 ∇~ ~r f (~ ~ ~r
r 0 , tr ) · ∇ =2 ∇ ~ ~s f (~
r 0 , tr ) · ∇ (10.78)
|~
r −~r 0| s
2 ∂f 2 ∂f 1
= = (10.79)
c s 2 ∂tr c ∂tr |~ r −~ r 0 |2
Now, to proceed with the second term in 2 involving ∇2 . Let’s work on it using the
product rule (this is a bit cleaner than the way Griffiths does it):
" #
r 0 , tr )
f (~
∇~ ~r f (~
r 0 , tr )
1
∇~2r ~
= ∇~r · ~
+ ∇~r · f (~ 0 ~
r , tr ) ∇~r (10.80)
|~r −~ r 0| |~
r −~ r 0| |~r −~ r 0|
"
2
∇~r f (~ 0
r , tr ) 1
= + f (~ r 0 , tr ) ∇~2r
|~
r −~ r 0| |~
r −~ r 0|
1
h i
+2 ∇ ~ ~r f (~
r 0 , tr ) · ∇ ~ ~r (10.81)
|~
r −~ r 0|
1 1 ∂ f 2
0 0
= + f (~r , tr ) 4 πδ(~ r − ~
r ) (10.82)
|~
r −~ r 0 | c 2 ∂tr2
where we plugged in the expressions we derived above for the first and third terms on
the second line and discarded the cancelling terms, and we used the usual relation
between the delta function and the Laplacian of the inverse distance. In the end result
we have, we see that the first term will cancel the (1/c 2 ) ∂ 2 /∂t 2 term from 2 . The
second term with the delta function can be easily evaluated (note that doing so gives
tr = t, also).
1 ∂2 r 0 , tr )
f (~
Z
2 φ(~
r , t) = dτ 0 − ∇ ~
2
r (10.83)
V c 2 ∂t 2 |~r −~ r 0|
Z
= 4π dτ 0 f (~
r 0 , tr ) δ(~ r 0 ) = 4 π f (~
r −~ r , t) (10.84)
V
r , t)
ρ(~ ~ r , t) = µo J(~
~ r , t)
2 V (~
r , t) = 2 A(~ (10.85)
o
as desired.
We need to confirm the Lorenz gauge condition is satisfied since we assumed it. We
~ Recall:
will need to calculate the divergence of the integrand in the expression for A.
~ ~r 1 r −~
~ r0 r0 −~
~ r ~ ~r 0 1
∇ =− = = −∇ (10.86)
|~
r −~r 0| |~
r −~r 0 |3 |~
r −~r 0 |3 |~
r −~r 0|
Therefore,
!
~ r 0 , tr )
J(~ ~ r 0 , tr )
~ ~r · J(~
∇
1
~ ~r ·
∇ = + J(~~ r 0 , tr ) · ∇
~ ~r (10.87)
|~
r −~ r 0| |~
r −~ 0
r | |~r −~ 0
r |
~
∇~ · J(~~ 0
r , tr )
1
= r 0
− J(~~ r 0 , tr ) · ∇
~ ~r 0
0
(10.88)
|~
r −~ r | |~
r −~ r |
!
∇ ~ r 0 , tr )
~ ~r · J(~ ~ r 0 , tr )
J(~ ∇~ 0 · J(~~ r 0 , tr )
= −∇ ~ ~r 0 · + ~r (10.89)
|~
r −~ r 0| |~
r −~ r 0| |~
r −~ r 0|
!
~ r 0 , tr )
= −∇~ ~r 0 · J(~ (10.90)
|~
r −~ r 0|
" #
1 ∂ J~ ~ ∂ρ ∂ J~ ~
+ · ∇~r tr − + · ∇~r 0 tr
|~
r −~ r 0 | ∂tr ∂tr ∂tr
where the continuity equation was used when evaluating ∇ ~ r 0 , tr ) but not for
~ ~r 0 · J(~
~ ~
∇~r · J(~ 0 ~
r , tr ) because the argument of J must match the divergence’s variable for
∂
continuity to apply. Note that the ∂t r 0 , tr ).
are evaluated at (~
r
Using the same kind of argument as we used on the previous page for showing
~ ~r (1/|~
∇ r −~ ~ ~r 0 (1/|~
r 0 |) = −∇ r 0 |), we have
r −~
~ ~r tr = − 1 ∇
∇ ~ ~r |~
r −~
1 ~
r 0| = ∇ r 0 |~
~ r −~ ~ ~r 0 tr
r 0 | = −∇ (10.91)
c c
~
We may now calculate the divergence of A:
!
~ r 0 , tr )
J(~ ~ r 0 , tr )
J(~
~ ~r · µo µo
Z Z
∇ ~ r , t) = ∇
~ ~r · A(~ dτ 0 = dτ 0 ~
∇ ~
r · (10.93)
4π V |~r −~ r 0| 4π V |~
r −~ r 0|
" ! #
~ r 0 , tr )
µo ~ ~r 0 · J(~ 1 ∂ρ
Z
= dτ 0 −∇ − (10.94)
4π V |~
r −~ r 0| |~
r −~ r 0 | ∂tr
∂ 1 ρ(~r 0 , tr ) ∂
Z
= −o µo dτ 0 = −o µo V (~r , t) (10.95)
∂t 4 π o V |~
r −~ r 0| ∂t
where the first term in the second line has been transformed into a surface integral at
infinity that vanishes because the current distribution is assumed to be finite in extent.
We convert ∂/∂tr to ∂/∂t by using ∂t/∂tr = 1. The Lorenz condition is thus satisfied.
Page 767
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
r 0 , tr )
f (~ 1 ∂ f ∂ tr 1
∂ 0 ∂
= + f (~r , tr ) (10.96)
∂ri |~r −~ r 0||~
r −~ r 0 | ∂tr ∂ri ∂ri |~ r −~ r 0|
1 1 ∂f r 0 , tr ) (~
f (~ r −~ r 0 ) · rbi
= − 0
− 0 2
(10.97)
c |~ r −~ r | ∂tr |~r −~ r | |~
r −~ r 0|
∂ tr 1 ∂ 1 (~ r −~ r 0 ) · rbi
r 0 = −
where =− r −~
~ (10.98)
∂ri c ∂ri c |~ r −~ r 0|
~ (~ ~ ~r V (~ ∂ ~
Then, E r , t) = −∇ r , t) − A(~ r , t) (10.99)
∂t
" !#
1
Z
1 ~
ρ(~ r 0 , tr )
∂ ~ r 0 , tr )
J(~
=− dτ 0 ∇~r + µo (10.100)
4π V o |~
r −~ r 0| ∂t |~
r −~ r 0|
" #
1 r −~
~ r0 1 ∂ρ ~ r −~ r0 1 ∂ J~ 1
Z
0 0
= dτ ρ(~ r , tr ) + − 2
4 π o V |~
r −~ r 0 |3 c ∂tr |~ r −~ r 0 |2 c ∂tr |~ r −~r 0|
(10.101)
where we obtained a sign flip in the last step by exchanging the j and k indices.
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 769
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
" #
1 r −~
~ r0 1 ∂ρ ~ r −~r0 1 ∂ J~ 1
Z
~ (~
E r , t) = dτ 0 ρ(~
r 0 , tr ) + − 2
4 π o |~
r −~ 0
r | 3 c ∂tr |~
r −~ 0
r | 2 c ∂tr |~r −~r 0|
V
(10.106)
" #
~ r 0 , tr )
J(~ 1 1 ∂ J~ ~ r0
r −~
~ r , t) = µo
Z
B(~ dτ 0 + × (10.107)
4π V |~
r −~ r 0 |2 c |~
r −~r 0 | ∂tr |~ r 0|
r −~
These equations are not incredibly useful in practice because it is usually easier to
calculate the retarded potential and differentiate to find the field, but they give a clear
physical understanding of how disturbances in the source distributions determine the
fields at a distance.
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 770
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
We note that the ∂/∂tr terms in each equation fall off like 1/|~ r −~r 0 | while the other
r 0 |2 , so it is these ∂/∂tr terms that dominate at large
r −~
terms fall off like 1/|~
distances. We won’t do the calculations using the above expressions (we will use the
retarded potentials), but this is an interesting fact because it tells us that it is the time
variation of the sources that generates that the strongest fields at large distances (the
radiated fields). A heuristic (i.e., handwaving) way of understanding this is to attribute
it to the fact that time variation generates EM waves, which are self-propagating,
while static sources and charges must generate the distant field directly.
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 771
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
Heald and Marion §8.2 has some nice discussion of the interpretation of these
equations as well as references to some papers that discuss the topic further. We note
two of their points here:
I It is frequently said (as we have said) that changing electric and magnetic fields
induce magnetic and electric fields, respectively, and this is how EM waves can
propagate. While this is true mathematically in Maxwell’s Equations, this is
something of an artifact. Maxwell’s Equations are local equations, meaning they
relate the fields, their derivatives, and the sources at a particular point in space
and time. But if we look at Jefimemko’s Equations, which are exactly true, we
see that fields at a distance are sourced by charges and currents. In effect,
Faraday’s Law and the displacement current term in Ampere’s Law are
intermediaries necessary to make Maxwell’s Equations local.
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 772
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
can be seen by inference by the following argument. Let’s Taylor expand the
source charge and current distributions around t = tr :
r 0| 1 ∂ 2 ρ r 0| 2
|~
r −~ |~
r −~
∂ ρ
r 0 , t) = ρ(~
ρ(~ r 0 , tr ) + + 2
+ ···
∂t ~r 0 ,tr
c 2 ∂t ~r 0 ,tr
c
~
~ r 0 , tr ) + ∂ J |~
r −~r 0| 1 ∂ 2 J~ |~
r −~r 0| 2
~ r 0 , t) = J(~
J(~ + + ···
∂t 0 c 2 ∂t 2 0 c
~
r ,tr ~
r ,tr
That is, we can obtain the first two terms in both of Jefimemko’s Equations
simply by accounting for retardation. The third term in the electric field
equation is the manifestation of Faraday’s Law, as we said above. But there is
no corresponding term in the magnetic field expression for the displacement
current, so we are forced to conclude that the retardation calculation accounts
for it. Amazing!
(We also see from the above Taylor expansion how, again, the first terms in
Jefimemko’s Equations are Coulomb’s Law and the Biot-Savart Law evaluated
at tr .)
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 773
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
1 1 ∂f
= (10.108)
τs f ∂tr
|~
r −~r 0| 1 ∂ f |~
r −~r 0 |/c
α= = (10.109)
c f ∂tr τs
which is the ratio of the light travel time to the source term variation timescale. The
quasistatic approximation consists of assuming that α 1 , which justifies neglecting
these terms and setting tr = t.
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 774
Section 10.2 Potentials Revisited: Retarded Potentials and Fields
The third term in the E ~ equation requires a bit more analysis. It involves the current,
but we cannot compare it to the current terms in the B ~ equation because E
~ and B~
have different units. So we must compare it to the other terms in the E ~ equation,
which involve ρ. Let’s write J = ρ v (drop the vector nature, it’s not important for
this analysis), so then the ratio of the third to the first term is
|~
r −~r 0| 1 ∂ J |~
r −~r 0 | 1 ∂ (ρv ) |~
r −~r 0| 1 ∂ρ 1 ∂v
= = β + (10.110)
c ρ c ∂t c ρ c ∂t c ρ ∂t c ∂t
|~r −~ r 0 |/c
= αβ +
τa
Section 10.2.2 Fields from Retarded Potentials: Jefimemko’s Equations Page 775
Section 11
Relativity and Electrodynamics
11.1 Study Guide
11.2 Fundamentals of Special Relativity
11.3 Relativistically Covariant Formulation of Electrodynamics
11.4 Relativistic Dynamics with Electromagnetic Fields
11.5 Lagrangian Formulation of Relativistic Electrodynamics (Skip)
11.6 Relativistic Conservation Laws (Skip)
Page 776
Section 11.1 Relativity and Electrodynamics: Study Guide
Study Guide
Since you already studied special relativity in Ph1b, we will not repeat the discussion
of all the thought experiments that lead one to special relativity (Griffiths §12.1). As
with much of the course, we will take a more rigorous approach instead, deriving the
Lorentz Transformation via consideration of one physical situation combined with
principles-based requirements on the transformation’s properties. We will quickly
review the basic results of special relativity you learned in Ph1b (Griffiths §12.1.3) .
We will then introduce four-vectors, the metric, and tensors (Griffith §12.1.4, with
added rigor), and proceed to discuss relativistic kinematics, energy-momentum, and
collisions (Griffiths §12.2.1–12.2.3). We will then switch to a discussion of how to
rewrite electrodynamics in relativistic notation, including writing the total
electromagnetic force (electric + magnetic) in a form that respects special relativity
and uses the relativistic notation (Griffiths §12.3.3–12.3.5, §12.2.4), including
derivation of the separate transformation properties of the electric and magnetic fields
(yielding the results in Griffiths §12.3.2 via a different approach). In the latter, you will
see that we derive no new physics but rather that electrodynamics is written far more
naturally and concisely in this notation.
Page 777
Section 11.1 Relativity and Electrodynamics: Study Guide
Page 778
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
So far in this course, we have not concerned ourselves much with how the physical
objects we have defined — e.g., charge, currents, fields, forces, etc. — depend on the
frame in which they are written. That is generally a topic more relevant for
mechanics. But special relativity is mechanics and, more importantly, transformation
properties of these various quantities will play a central role in determining how we
write down electrodynamics in a way that is manifestly consistent with special
relativity. So let’s define a few terms that we will use in that discussion.
I Reference Frame
As you know, special relativity is concerned with how physical quantities behave
when we look at them in different inertial reference frames — frames that are
moving at fixed velocity. A frame that moves at fixed velocity is simply one in
which Newton’s Laws are obeyed. We will denote such frames by F , Fe, etc.
Section 11.2.1 Reference Frames, Coordinate Representations, Transformation Properties, Invariance, and Covariance Page 779
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
I Coordinate Representation
Any physical quantity has an existence independent of any specific coordinate
system or reference frame. (We use both terms because one can imagine
different coordinate systems that describe the same reference frame, and one
might not consider the coordinate systems of two reference frames to be
different if they are aligned and they coincide at one instant in time.) When we
write down physical laws, such as Newton’s Law or Coulomb’s Law or the
Biot-Savart Law or Maxwell’s Equations, those are statements about
relationships between physical quantities that are independent of coordinate
system or reference frame. Frequently, though, to actually execute those
relationships, we do need to pick a coordinate system or reference frame. The
coordinate representation of a physical quantity, such as an electric current, is
the set of numbers that define it in a given coordinate system or reference
frame. This representation can vary between coordinate systems or reference
frames. But that variation does not change the physical meaning of the quantity
to which one is referring. For the example of a current, which is a vector, its
coordinate representation will depend on coordinate system orientation. But it is
still the same current independent of that choice.
Section 11.2.1 Reference Frames, Coordinate Representations, Transformation Properties, Invariance, and Covariance Page 780
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
Section 11.2.1 Reference Frames, Coordinate Representations, Transformation Properties, Invariance, and Covariance Page 781
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
There is clearly great value in writing down the laws of physics in terms of invariant
and covariant quantities: this ensures our laws can be written down in a manner that
is manifestly independent of coordinate system or reference frame. After reviewing
special relativity, out goal is to formulate electrodynamics in such a way that the
transformation properties of the various quantities we deal with — source
distributions, potentials, fields, etc. — are made clear. We will see that
electrodynamics is deeply related to relativity, a realization that in fact initiated the
development of special relativity by Poincaré, Lorentz, Einstein, and others. In fact,
Einstein’s first paper on special relativity is titled “On the Electrodynamics of Moving
Bodies” (Annalen der Physik, 17: 891 (1905)).
Section 11.2.1 Reference Frames, Coordinate Representations, Transformation Properties, Invariance, and Covariance Page 782
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
Sign Convention
As with our definition of 2 , we deviate from Griffiths and follow the more standard
sign convention for relativity where the inner product of a four-vector is of the form
(time component)2 −(space components)2 .
It is the second postulate that was Einstein’s brilliant leap and that leads to all the
nonintuitive implications of special relativity. To some extent, the second postulate is
a corollary of the first once one realizes that electromagnetic waves do not travel in a
medium. If the laws of electromagnetism, which give rise to the speed of light, are to
be the same in all frames, then the speed of light must of necessity be the same in all
frames. But it took the Michelson-Morley experiment to kill the concept of the ether,
a medium in which light propagates.
Consider two inertial reference frames F and Fe. Let them have coordinate axes
(x, y , z, t) and (e
x , ye, ze, te). We include t as a coordinate and allow it to be different in
the Fe frame because it will be necessary to avoid a contradiction with the second
postulate. Let the two systems’ axes and origins coincide at t = te = 0. Let the Fe
frame be moving at speed v = c β along the +x axis with respect to F , which means
that the position of the Fe origin obeys x = β c t in the F frame.
This information is summarized in our space-time
diagram in the F system, a plot of t vs. x with the
origin of the Fe system represented by the solid line
of slope β −1 and the path of a light ray emitted
from the origin shown by the dashed line of slope 1.
Points in the space-time diagram are referred to as
events because they are not just points in space, but
in time also. Obviously, one can generalize space-
time to more than one spatial dimension, it just
becomes hard to visualize.
Let us consider the path of a light ray in the two frames, requiring that the second
postulate hold. Suppose the light ray is emitted from the Fe origin at time c tee = −L e
in the +x direction, it hits a mirror at xer = L
e at c ter = 0 and is reflected, returning to
the Fe origin at c tea = L.e (We use L
e instead of L to avoid confusion when we later
discuss length contraction.) The light ray has ye = ze = 0 for all time. In a space-time
diagram of the Fe frame, the reflection event (e xr , c ter ) is obtained by the intersection
of light rays propagating forward in time from the emission event (e xe , c tee ) = (0, −L)
e
and backward in time from the absorption event (e xa , c ta ) = (0, L). The intersection is
e e
at (exr , c ter ) = (L,
e 0).
Let’s examine the path in the F system, assuming we do not yet know the
transformation law for coordinates between reference frames. The light ray travels
only along the x direction, so it satifies y = z = 0 for all time also. Let (xe , c te ),
(xr , c tr ) and (xa , c ta ) indicate the coordinates of the three events in the F frame.
The emission and absorption events must occur on the solid line representing the
position of the Fe origin, which means their space and time coordinates are related by
the slope β −1 . The line of this slope through the origin thus gives the space-time
location of the c te axis in F . Time reversal invariance says that these events’
symmetric occurence in Fe implies they occur at symmetric times in F . So the two
events must be at (−c ta β, −c ta ) and (c ta β, c ta ). Let’s calculate where light rays
from these two events would intersect, which will give us the position of the reflection
event in F , which yields the space-time location of the xe axis in F . In the following, r
and s begin as undetermined parameters indicating how much time passes in the F
frame between the emission or absorption event and the reflection event, respectively:
(−c ta β, −c ta ) + r c (1, 1) = (c ta β, c ta ) + s c (1, −1) (11.1)
r − s = 2 β ta r + s = 2 ta (11.2)
r = (1 + β) ta s = (1 − β) ta (11.3)
(xr , c tr ) = (−c ta β, −c ta ) + r c (1, 1) = (c ta , β c ta ) (11.4)
The last line implies that the reflection event (xr , c tr ) sits on the line through the
origin with slope β. Its position on that line depends on the value of ta , the time of
the absorption event — that is, on L e and β. Since the reflection event (e xr , c ter ) is
always at c ter = 0, the line thus tells us where the xe axis sits in the F space-time
diagram. So, in sum, we have that the te axis is a line of slope β −1 and the xe axis is a
line of slope β, both going through the F origin.
Section 11.2.4 Derivation of the Lorentz Transformation Page 787
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
We may construct the general form for the transformation of coordinates from Fe to F
by using the above information along with expected properties of the transformation:
1. Linearity: We have seen that events on the xe axis lie on a line of slope β and
events on the c te axis lie on a line of slope β −1 in F . If we assume the
transformation is linear in the space-time coordinates (ex , c te) then the
transformation for an arbitrary space time event with Fe frame coordinates
(e
x , c te) to the F frame can be written
3. Symmetry of the two frames: By symmetry, the slope of the x and t axes in
the Fe space-time diagram must be the same as the slope of the xe and te axes in
the F space-time diagram, so γ(β) = γ(−β) and γ̃(β) = γ̃(−β) is required (the
sign flip has already been accounted for). So we have
xe = γ (x − β c t) c te = −γ (β x − c t) Lorentz
(11.11)
x = γ xe + β c te c t = γ βe x + c te Transformation
Consider again two frames F and Fe, with Fe moving at speed β in the +x direction
relative to F , and assume the origins of the two frames coincide at t = te = 0. In this
case, we will emit a light ray from the origin in the +y direction and reflect it back to
the origin. The emission, reflection, and absorption events are given by (neglecting the
z coordinate, which is not involved at all):
Let’s now repeat our argument regarding the coordinates of the three events in F ,
with the complication that our space-time now has three dimensions, though the
motion is only in two of them.
Symmetry about the origin relates the time coordinates and the known velocity
vector of Fe in F relates the space coordinates of the two events.
3. Though the reflection event occurs at ye 6= 0 and y 6= 0, the emission and
absorption events occur at the origin and thus are unaffected by the existence of
the y dimension. They must thus obey the Lorentz transformation rule, which
implies
(xe , ye , c te ) = −γ L
e (β, 0, 1) (11.17)
(xa , ya , c ta ) = γ L
e (β, 0, 1) (11.18)
4. Let us now calculate the position of the intersection of light rays sent out from
the emission and absorption events. We use the postulate that the speed of
light
√ is c in F . The space-time displacement of a light ray in one unit of c t is
( 1 − δ 2 , δ, 1), where δ allows for freedom in the direction (in the xy plane). δ
carries the same sign as yr . We have:
p p
1 − δ 2 , δ, 1) = (xa , ya , c ta ) + s (− 1 − δ 2 , δ, −1) (11.19)
(xe , ye , c te ) + r (
p p
e (β, 0, 1) + r ( 1 − δ 2 , δ, 1) = γ L
−γ L e (β, 0, 1) + s (− 1 − δ 2 , δ, −1) (11.20)
Let’s explain this a bit more. We may assume the velocity vector has the same
components on the two sides, with just a sign flip in the y component, because
the light path must be symmetric about t = te = 0 because the motion is along
x, not y . The signs are obtained as follows:
I The signs on the left-side velocity term are obtained by simple arguments:
I If ye = 0 and yr > 0, then the y light velocity must be positive
between the two events.
I In the limit β 1, we must recover the nonrelativistic limit, and in
that limit we know that if the light ray always has xe = 0 and the xe
origin is moving in the +x direction, then x ≥ 0 is required for the
position of the light ray.
Simplifying, we obtain
p p
r =s =γL
e 1 − δ2 = β δ= 1 − β 2 = γ −1 (11.22)
Page 794
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
I Time dilation
Consider two events occuring at a fixed point in space in the frame Fe; for
example, two ticks of a clock. They are separated by the vector
(e
x , c te) = (0, c τ ). What is the separation of the two events in the frame F ,
relative to which Fe is moving at speed β? The Lorentz transformation tells us
The time between the events in the F frame is larger than in Fe. Hence, the
term “time dilation”: time “slows down” in the moving frame, the two events
have a smaller time separation in their rest frame than in any other frame. One
is not obtaining something for nothing, though, because the spatial separation
of the two events has become nonzero. That is, we are no longer measuring just
the time separation of two events that occur at the same point in space; we are
measuring a separation with both time and space components.
I Length contraction
The length L e of an object at rest in Fe can be viewed as two events with
separation (∆e x , c ∆te) = (L,
e 0), representing the left and right ends of the object
at some common time c te. The time separation between these two events will
become nonzero in the F frame because of their nonzero spatial separation, so
these two events are not valid as a length measurement in F .
Explicitly, and without loss of generality, we let the two events corresponding to
the Fe length measurement be
(e
x1 , c te1 ) = (0, 0) (e
x2 , c te2 ) = (L,
e 0) (11.26)
confirming the expectation that the length measurement events in Fe do not give
a length measurement in F . To make a length measurement in F , we must pick
points on the two trajectories that are separated by ∆t = 0, which the above
expression implies is not possible if te1 = te2 . Let us allow te1 6= te2 and apply the
requirement ∆t = 0, using the Lorentz transformation to write this condition in
terms of Fe coordinates:
0 = c ∆t = c t2 − c t1 = γ β (e
x2 − xe1 ) + c te2 − c te1 (11.29)
Section 11.2.5 Implications of the Lorentz Transformation Page 796
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
Without loss of generality, we again assume event 1 occurs when the two origins
intersect, so xe1 = x1 = c te1 = c t1 = 0. We choose c t2 = 0 also in order to
obtain a length measurement in F . The assumption about the origins implies
that xe2 = L
e for all time because the object is at rest in Fe. With these
assumptions, the c ∆t = 0 requirement reduces to
c te2 = −β xe2 = −β L
e (11.30)
(x1 , c t1 ) = (0, 0) (e
x1 , c te1 ) = (0, 0) (11.32)
−1 e e −β L)
(x2 , c t2 ) = (γ L, 0) (e
x2 , c te2 ) = (L, e (11.33)
We see that in order to be simultaneous in the F frame, the two events must
occur with negative time separation, i.e., with c ∆te = c te2 − c te1 = −β L,
e in the
rest frame of the object. The object is moving to the right with the Fe frame, so
the right end of the object is not at x = L
e yet in the F frame when the
measurement event occurs in F , hence the apparent length contraction.
Section 11.2.5 Implications of the Lorentz Transformation Page 797
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
I Relativity of Simultaneity
The length contraction example shows that two events that are temporally
simultaneous but spatially separated in one frame may not be temporally
simultaneous in another; i.e., c ∆te = 0 does not imply c ∆t = 0 unless ∆e x =0
also. Generically, then, simultaneity of physically separated events is no longer
well defined.
I Transformations of Areas
The matrix of partial derivatives of our transformation is
γ −β γ
J(β) = (11.34)
−β γ γ
So, space-time areas are preserved:
d xe c d te = |J| dx c dt = dx c dt (11.35)
because γ 2 − β 2 γ 2 = 1. This is a necessity, as there must be symmetry between
the two directions for the Lorentz transformation, which would not hold if the
Jacobian determinant were not unity.
I Invariant Interval
From the Lorentz transformation and again using γ 2 − β 2 γ 2 = 1, one can show
(c t)2 − x 2 = (c te)2 − (e
x )2 ≡ s 2 (11.36)
The quantity s 2 is the invariant interval associated with the space-time vector
(x, c t) and (e
x , c te). It can be thought of like the magnitude of a vector in
space, which is invariant under spatial rotations. The invariant interval is
invariant under Lorentz transformations, also known as boosts.
Section 11.2.5 Implications of the Lorentz Transformation Page 798
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
So, the Lorentz transformation looks something like a coordinate rotation, except by
an imaginary angle i η. More importantly, though, we see the contours of constant xe
or c te as β (i.e., η) varies form hyperbolic curves in the x t plane. That is, the event
(e
x , 0) in the Fe frame appears on the hyperbola (x, c t) = xe (cosh η, sinh η) in F , with
η increasing as β increases. Similarly, the event (0, c te) appears on the hyperbola
(x, c t) = c te (sinh η, cosh η) in F . This is illustrated in the figure on the following slide.
NOTE: Even though we have made the above geometrical intepretation, one has to be
a bit careful about overinterpreting it. The difficulty is that the quantity left constant
by a Lorentz transformation, the invariant interval, does not correspond to a curve of
fixed distance from the origin on the above plots. Rather, it corresponds to the
hyperbolic curves that the fiducial points follow as β is changed. Another way of
seeing this is that the entire first and third quadrants of the Fe frame occupy the area
between the two corresponding slanted space-time axes displayed in the right plot.
Therefore, spatial distances and areas are not preserved by the mapping. Invariant
interval is preserved but does not correspond to a fixed distance on the plot!
Section 11.2.6 Geometric Interpretation of Lorentz Transformation Page 799
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
x , c te) = (1, 0), (0, 1), (−1, 0), (0, −1) in F for η = 10−1 , 10−3/4 ,
Left: position of (e
10−1/2 , 10−1/4 , 1, 101/4 (moving sequentially outward). This shows how a particular
event is seen in another frame as the relative speed β is increased. Right: Fe
space-time axes in F for same values of η (larger η =⇒ increasingly oblique). This
plot shows how the Fe space-time axes appear “squeezed together” when seen in the F
frame. Both plots are for positive β and η. Negative β and η would occupy the other
two quadrants.
The geometrical interpretation and invariant interval will make it possible to prove
that the relativity of simultaneity can never produce causality problems.
First, let us define the light cone as the region in space-time that can be reached from
a given event; refer to the figure on the slide at the end of this section. The light cone
of the event at the origin of a space-time diagram consists of all events with |c t| > |x|.
The region c t < 0, the past light cone, consists of all events that could have causal
influence on the event (x, c t) = (0, 0). The region c t > 0, the future light cone,
consists of all events that the event (x, c t) = (0, 0) can have causal influence on.
Since we know that, under Lorentz transformations, an event slides along a hyperbolic
curve in space-time as the speed β is varied, we are assured that events that are in the
future light cone of the event at the origin in one reference frame are also in that
event’s future light cone in any other reference frame. Similarly for past light cones.
What about simultaneous events? Two events are simultaneous in a particular frame
Fe if they occur at the same value of te, the time in the frame Fe. Let one event be at
the origin (0, 0) and the other event at (e x , 0), so the space-time vector separating
them is (ex , 0). This event is outside the light cone of the origin, so, in this frame, the
two events are out of causal contact. When we transform to a different frame F , but
one whose origin coincides with that of Fe at t = te = 0, the space-time vector between
the two events will slide on the hyperbolic curve (cosh η, sinh η). The event at (e x , 0) in
Fe may move to t < 0 or t > 0 in other frames, raising causality questions. The
causality worries are put to rest by the fact that the hyperbola is entirely outside the
light cone of the first event, so the second event is always outside of causal contact
with the first event.
Whether two events are causally connected is determined by the sign of the invariant
interval of the space-time vector separating them. If s 2 > 0, then |c t| > |x| and the
two events connected by the vector are in causal contact. The vector is called
time-like. Our argument about hyperbolic curves ensures that the sign of the time
component of the vector does not change, preserving causal relationships. If s 2 < 0,
then |c t| < |x| and the two events are never in causal contact, regardless of frame.
The vector is called space-like. The sign of the time component of the vector may
depend on the reference frame. If s 2 = 0, the vector is called null or light-like since
only light (or, as we shall see, any other massless particle) can travel on such a path in
space-time. The concepts are illustrated on the following slide.
Space-time diagrams illustrating light cones, invariant intervals, and the preservation
of causality. The dashed lines indicate the light cone of the event at (x, c t) = (0, 0).
(Left) Surface of constant s 2 > 0, corresponding to a time-like interval. Lorentz
transformations move space-time events along each half of this surface but cannot
cause events to move from one half to the other half and thus cannot change the sign
of t or affect causality. (Right) Surface of constant s 2 < 0, corresponding to a
space-like interval. Lorentz transformations also move space-time events along this
surface, and the sign of t may change. However, none of these events are in the light
cone of (0, 0) and thus sign changes in t cannot affect causality.
r µ = (r 0 , r 1 , r 2 , r 3 ) = (ct, x, y , z) (11.40)
Such four-vectors are termed “Lorentz-covariant” to indicate that, while they do vary
between reference frames, there are rules for how they vary and those rules are given
by the Lorentz transformation law.
A quick way to check the sign of the matrix is to check the transformation properties
of the origin of the F frame, with r µ = (ct, 0, 0, 0). Using the above matrix, we get
reµ = (γct, γβct, 0, 0): the x = re1 coordinate is positive, as expected if F moves at
c β xb with respect to Fe.
βi βj
Λ00 = γ Λ0i = Λi 0 = γ βi Λi j = δij + (γ − 1) i, j = 1, 2, 3 (11.41)
β2
g is the metric. One can check that this particular definition of g and the resulting
⇒
definition of | r |2 is indeed independent of reference frame by applying the Lorentz
transformation to show explicitly
The chain of logic is important: g cannot be derived from some more fundamental
principle; it has this form because this form provides the invariant quantity we wanted.
Note that the · notation for the Lorentz scalar product will quickly become useless for
tensors, which may have more than two indices. We introduce it for analogy to the
three-dimensional dot product. The contraction notation is Lorentz-covariant and is
thus entirely sufficient. The Invariant Norm of a Four-Vector and the Metric
Section 11.2.9 Page 806
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
Tensors
More generally, a n-th rank tensor T is an object that has 4n components organized
with n indices, T µ1 ···µn , and that transforms under a change of reference frame via
the rule
i.e., contraction with Lorentz transformation matrices in all the indices. A four-vector
is a first-rank tensor. We will see second-rank tensors soon enough. A four-scalar (or
just scalar) is a zeroth-rank tensor and thus is invariant under Lorentz transformations;
conversely, any quantity that is invariant under Lorentz transformations is a scalar.
From now on, when we talk about tensors, we also include four-vectors in the
category. Given its transformation properties, the metric is a second-rank tensor.
Page 808
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
If we are going to lower indices, we need a way to raise them, so we define in any frame
where the −1 implies the matrix inversion operation and where both versions of δ are
the identity matrix with ones along the diagonal in any reference frame. We can check
that this definition of g µν raises indices in a manner consistent with our definition of
how indices are lowered:
g µν rν = g µν gνσ r σ = δ µσ r σ = r µ (11.47)
That is, we recover the raised-index four-vector we started with if we use the above
definitions.
Now, because g is so simple, it turns out that g µν = gµν (just try it out in any given
reference frame). Note also that
g µν = g µσ gσν = δ µν (11.48)
Moreover, it doesn’t make much sense to write δ µν or δµν because they are the metric,
and thus writing δµσ or δ µσ would be confusing: some of their elements would be -1!
It is ok to write g µν or gµν , but it is preferred to use δ µν or δµν to avoid obfuscation.
We can derive Lorentz transformation properties for “lowered” indices by using the
known properties of the raised indices and the metric:
where we used geµν = gµν (definition of metric) in the first step, the definition of the
raising operation in the second step, and then we applied a self-consistent use of
raising and lower operations on Λνσ in the third step. If one works it out in detail, one
can see that the components Λµν are given by Λµν with a sign flip on the velocity.
For the sake of developing intuition, consider a coordinate system rotation instead of
Lorentz transformation. The rotation changes the coordinate axes. But a point in
space remains fixed — the rotation is just a relabeling of points in space. So its
coordinates in the new coordinate system are different than those in the old coordinate
system. The amount by which the coordinates must change is exactly the opposite of
the amount that the axes changed. For example, suppose one rotates about the z-axis
by +30◦ so that the new x-axis is in the first quadrant of the old coordinate system.
Then a point that was on the x-axis of the old coordinate system is now in the fourth
quadrant of the new coordinate system. When the unit vectors are rotated in the
positive direction, the coordinates of points appear to rotate in the negative direction.
Covariant Gradient
The next obvious tool we need is a four-vector version of the gradient operator. The
natural thing to want is for this gradient
⇒operator
to do the right thing for Taylor
expansions. Given a scalar function S r , we would like
⇒ ⇒
⇒ ⇒ ⇒
⇒
S r + d r − S r = ∇S · d r = ∇S dr µ (11.52)
µ
where we have used the Lorentz scalar product to ensure both sides are Lorentz scalars
even though the right side is made of coordinate-system-dependent — contravariant
and covariant under Lorentz transformation — quantities. We know that, since r µ is a
contravariant four-vector, its differential version dr µ is also a contravariant
four-vector. This, along with the known form of Taylor expansions for the coordinate
representation in a particular reference frame, implies that
⇒ ∂S ∂ ∂ ∂ ∂
∇S = ≡ ∂µ S =⇒ ∂µ ≡ , , , (11.53)
µ ∂r µ ∂r 0 ∂r 1 ∂r 2 ∂r 3
would yield the desired result. Thus, we use the above as the definition of the
four-gradient.
⇒ ⇒ ∂ ∂ ∂ ∂
2 = ∇ · ∇ = ∂µ ∂ µ = − − − (11.55)
∂(r 0 )2 ∂(r 1 )2 ∂(r 2 )2 ∂(r 3 )2
⇒ ⇒ 1 ∂ ~2
=⇒ 2 = ∇ · ∇ = ∂µ ∂ µ = −∇ (11.56)
c 2 ∂t 2
It is natural to ask whether it is possible to create a space-time vector for the velocity,
an entity that transforms like the space-time position four-vector. This would provide
a “unified” treatment of velocity and position. The trick, of course, is to find a set of
four numbers that are covariant, that transform in the appropriate way under Lorentz
transformations.
First, we must define the proper time. Consider a frame that is moving with the
particle whose velocity we want to specify. In that frame, the particle’s space-time
position is always r µ = (c τ, 0, 0, 0) because the particle is at the origin. c τ specifies
the time in the frame moving with the particle. That may seem frame-specific; but
remember that c 2 τ 2 is the invariant interval of the particle’s position. This will be the
same in all reference frames that share the same origin at their time origin, so it is
reasonable to think of it as a quantity available in all frames.
Section 11.2.13 Proper Time, the Covariant Four-Velocity, and Velocity Addition Page 814
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
Section 11.2.13 Proper Time, the Covariant Four-Velocity, and Velocity Addition Page 815
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
So, we have a good formal definition. What does it look like in terms of quantities we
have easy access to, three-velocities? Consider a particle moving at velocity c β ~p
(possibly a function of time) in some frame F . (We use the subscript p to distinguish
the particle velocity from that of the frame F relative to another frame Fe.) The
particle’s trajectory is r µ (t) in this frame. We have the obvious differential relations
where the second relation is simply time dilation. These relations are instantaneously
true even if the particle is accelerating. Thus, the four-velocity in F is
2 −1/2
~ ~p .
where γp = 1 − β p is the γ factor associated with the particle velocity β
Thus, given a particle trajectory, we can easily calculate the four-velocity in any frame.
We note that we could have also obtained the four-velocity by Lorentz transformation
of the four-vector (1, ~0) from one frame F to a frame Fe in which F is moving a
velocity c β~p , making the argument that (1, ~0) is the four-velocity of the F frame origin
in F . It is hardly obvious that that would have been the right thing to do, though!
Section 11.2.13 Proper Time, the Covariant Four-Velocity, and Velocity Addition Page 816
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
We may now use the Lorentz transformation of the four-velocity to derive the rule for
⇒
velocity addition. Start with a particle with four-velocity v with coordinate
µ
representation v1 in F and suppose F is moving at velocity c β2 xb with respect to the
⇒
frame Fe. This will give us the addition of v and the four-velocity
⇒
u = γ2 c (1, β2 , 0, 0). It is:
γ2 γ2 β2 0 0 1 γ2 (1 + β2 β1,x )
γ2 β2 γ2 0 0
γ1 c β1,x = γ1 c γ2 (β2 + β1,x ) (11.63)
veµ =
0 0 1 0 β1,y β1,y
0 0 0 1 β1,z β1,z
⇒
~12 where β
Because v is a four-vector, we expect veµ = γ12 c 1, β ~12 is the velocity
The expression for γ12 could be derived more algebraically, but the above is clearly the
simplest way to obtain it.
Section 11.2.13 Proper Time, the Covariant Four-Velocity, and Velocity Addition Page 817
Section 11.2 Relativity and Electrodynamics: Fundamentals of Special Relativity
One could of course consider a more general velocity addition by allowing F to move
in an arbitrary direction with respect to Fe. The algebra is worse but there is no
conceptual difference.
It is possible to derive the above, or understand it, by considering all the length
contractions and time dilations involved. Clearly, using the Lorentz transformation is
much faster.
Section 11.2.13 Proper Time, the Covariant Four-Velocity, and Velocity Addition Page 818
Section 11.3 Relativity and Electrodynamics: Relativistically Covariant Formulation of Electrodynamics
ρ = γ ρo J~ = ρ ~
v = ρo γ ~
v (11.65)
⇒
This is easily summarized using the four-velocity v we just defined:
⇒ ⇒
J = ρo v J µ = ρo v µ = γ (ρo c, ρo ~
v ) = (ρ c, ρ ~
v) (11.66)
⇒
Since we are multiplying a four-vector by the invariant quantity ρo , we are assured J
is also a four-vector, which we call the covariant source density. Note: ρ itself is not
an invariant quantity, but the rest-frame charge density ρo is invariant because the
total amount of charge cannot depend on the reference frame (it just consists of
counting up charges!). The continuity equation is then written in manifestly
Lorentz-invariant form (using the Lorentz scalar product):
∂ρ ⇒ ⇒
0= ~ · J~ = ∂µ J µ = ∇ · J
+∇ (11.67)
∂t
V ~
Aµ ≡ ,A (11.69)
c
then the above equations can be rewritten as the covariant wave equation
⇒ ⇒
2 A = µo J (11.70)
⇒ ⇒
Since J is a four-vector and 2 and µo are scalars, we are assured that A is a
four-vector as desired. Of course, the above equations only apply in Lorenz gauge, the
condition for which can be written in a manifestly Lorentz-invariant form:
∂V ⇒ ⇒
~ = ∂µ Aµ = ∇ · A
~ ·A
0 = o µo +∇ (11.71)
∂t
The Lorenz gauge condition is invariant under change of reference frame because it is
~ = 0 (which is
~ ·A
a four-scalar by definition. Note how it is the natural extension of ∇
a scalar under rotations and thus independent of coordinate axis orientation.)
In the rest frame F of a point charge q, the covariant potential has components
1 q ~ r , t) = 0
V (~
r , t) = A(~ (11.72)
4 π o [x 2 + y 2 + z 2 ]1/2
Now, to have the charge move with velocity ~v =cβ ~ = c β xb, we simply need to obtain
⇒
A in a lab frame Fe relative to which the charge’s rest frame F moves at ~v . That is,
⇒
we use the Lorentz transformation, Equation 11.39, on A :
1 e 1 ex = γ Ax + β 1 V = γ β V
γ
V =γ V + β Ax = V A (11.73)
c c c c c
and A
ey = Ay = 0 and A
ez = Az = 0.
This is half the work. V is of course written in terms of the rest frame coordinates
r µ = (c t, x, y , z), so we need to rewrite it in terms of the lab frame coordinates
reµ = (c te, xe, ye, ze). These coordinates are also related by Lorentz transformation, but
we need the one going in the opposite direction (because we are now writing F
coordinates in terms of Fe coordinates)
c t = γ c te − β xe x = γ −β c te + xe (11.74)
and y = ye and z = ze. (You can check that this is the correct direction for the
transformation by considering the position of the F frame origin in Fe frame
coordinates: (x = 0, t) should obey xe = β c te.) Combining the Lorentz transformation
of the potential with the above transformation of the coordinates, we obtain:
1 γq
r µ) =
Ve (e i1/2 (11.75)
4 π o
h 2
γ2 β c te − xe + ye2 + ze2
1 q
= i1/2 (11.76)
4 π o
hn 2 o
xe − v te + ye2 + ze2 − β 2 (e
y 2 + ze2 )
The coordinates in the above are the current position of the charge in Fe.
Let’s generalize the above expression for an arbitrary direction of motion by defining
Then we have
1 q 1
r µ) =
Ve (e (11.78)
r µ ) 1 − β 2 sin2 θ 1/2
4 π o R(e
1 1 q vi 1
A r µ) = 2
ei (e (11.79)
r µ ) 1 − β 2 sin2 θ 1/2
c 4 π o R(e
where θ is now the angle between the direction of motion ~ ~ from the
v and the vector R
particle’s current position to the position at which we want to know the potentials.
We see that the potentials are unmodified along the particle’s trajectory (ahead of or
behind) but that the potential strengthens as one moves away from that trajectory,
becoming strongest transverse to the direction of motion.
We will see later it will be useful to rewrite these expressions in terms of the retarded
position of the particle — the position at the retarded time — but we will avoid that
digression for now.
Page 824
Section 11.3 Relativity and Electrodynamics: Relativistically Covariant Formulation of Electrodynamics
⇒ ⇒ 1
F=∇∧ A ⇐⇒ F µν = ∂ µ Aν − ∂ ν Aµ ⇐⇒ −F 0j = F j0 = Ej F ij = −ijk Bk
c
(11.80)
where we have introduced the wedge product, ∧, which is the four-dimensional
generalization of the cross-product, and where i, j, k run over 1, 2, 3 and ijk is the
completely antisymmetric Levi-Civita symbol. (We define Ej ≡ E j and Bj ≡ B j ; we
are free to do this because they are not components of a four-vector, for which the
relation between contravariant and covariant components is fixed). From the above,
0 −Ex /c −Ey /c −Ez /c
Ex /c 0 −Bz By
F µν =
(11.81)
Ey /c Bz 0 −Bx
Ez /c −By Bx 0
Thanks to the definition of F in terms of the covariant gradient and the covariant
potential, we are assured it is a second-rank tensor. F is clearly the Lorentz-covariant
way to represent the electric and magnetic fields: its second-rank tensor
transformation properties provide all the information we need:
Let’s write these transformation laws in terms of the fields explicitly so we can see the
transformation rules more clearly. As noted above, we define lowered roman indices for
three-tensors by Ej ≡ E j , Bj ≡ B j , βi ≡ β i , vi ≡ v i = c β i , and δjk ≡ δ j k (we will not
use v µ here so there is no confusion); for the space components of four-tensors,
ri = −r i , g j k = −g jk continue to hold. We have:
where we have expanded out the sums into space and time components. We use
F 00 = 0 and the generic forms of the components of Λ, Equation 11.41,
βj βk βj β`
Eej = −c γ δjk + (γ − 1) 2 F 0k + γ βk γ βj F k0 + γ βk δj` + (γ − 1) 2 F k`
β β
(11.85)
βj βk 2
Eej = −c γ F 0j + 2 2
0k kj
γ − γ − β γ F + γ β k F (11.86)
β2
βj βk
Eej = γ Ej + (1 − γ) Ek − γ vk jk` B` (11.87)
β2
We will reduce this a simpler form soon, but let’s obtain the analogous result for the
~ from F , we need the identity ijk ijm = 2 δkm . With
magnetic field first. To extract B
this,
1 1 1
− jk` F k` = jk` k`m Bm = k`j k`m Bm = δjm Bm = Bj (11.88)
2 2 2
Therefore,
ej = − 1 jk` γ βk F 0` − γ β` F 0k + F k` − γ − 1 βk βm F `m + β` βn F nk
B
2 β2
(11.91)
then, using jk` = −j`k , F 0a = −Ea /c, −jk` F k` /2 = Bj and F ab = −abc Bc :
which we see is a form very similar to the one we obtained for the transformation of
the electric field, Equation 11.87.
~˜ = γ E
E ~ + (1 − γ) βb βb · E
~ −γ~ v ×B ~ (11.97)
~˜ = γ B ~ + γ ~
B ~ + (1 − γ) βb βb · B ~
v ×E (11.98)
c2
Let us resolve the fields into pieces parallel and perpendicular to the velocity:
~˜|| = βb βb · E
E ~ ~⊥ = E
E ~ − βb βb · E
~ ~˜ || = βb βb · B
B ~ ~⊥ = B
B ~ − βb βb · B
~
(11.99)
Thus, we have
~˜⊥ = γ E
h i
Ee|| = E|| E ~⊥ − ~ ~⊥
v ×B (11.100)
~⊥ + 1 ~
B
e|| = B|| ~˜ ⊥ = γ B
B v × ~⊥
E (11.101)
c2
recovering the results in Griffiths §12.3.2, where the transformation properties of the
fields are derived. (These are the equivalent of Equations 12.109 in Griffiths,
accounting for the fact that those equations give the field in F in terms of the field in
Fe and thus have a sign flip on the velocity.) We see that these properties have fallen
out of the Lorentz transformation of the F tensor.
We first saw the mixing of electric and magnetic fields via transformation between
inertial frames in the context of Faraday’s Law. The field tensor codifies this mixing
(with γ’s added) and gives it a satisfying rationale by tying it to the general
requirement of Lorentz covariance (i.e., special relativity). One must of course
remember that both Faraday’s Law and the invariance of the speed of light that lead
to special relativity, and more generally the principle that physical laws are
independent of inertial reference frame, are empirical observations that we codify via
an underlying principle. The principles are profound, which may seem sufficient
justification to assume them, but, ultimately, emprical verification of predictions is
what justifies those principles. This is the difference between modern empiricism and
the ancient Greek approach to science, where the principles were deemed sufficient
and formulation and testing of predictions deemed unnecessary!
~ (~ 1 q ~ r , t) = 0
E r , t) = ~
r B(~ (11.102)
4 π o r 3
We take ~
v = v xb. Applying the Lorentz transformation to the fields, we obtain
1 q
Eex (r µ ) = Ee|| = E|| = Ex = x (11.103)
4 π o r 3
Bex (r µ ) = Be|| = B|| = Bx = 0 (11.104)
~˜yz (r µ ) = E
~˜⊥ = γ E ~ ⊥ = 1 γ q (y yb + z zb)
h i
E ~⊥ − ~
v ×B (11.105)
4 π o r 3
1 ~⊥ = 1 ~
~˜ yz (r µ ) = B
B ~˜ ⊥ = γ B ~⊥ + v ×E
~ v ×E~˜ (11.106)
c2 c2
These fields are still in terms of the F -frame coordinates, so they are not very useful
as is.
1 γq
Eex = i3/2 xe − v te (11.108)
4 π o
h 2
γ2 2
xe − v te + ye + ze2
~˜yz = 1 γq
E i3/2 (e
y yb + ze zb) (11.109)
4 π o
h 2
γ2 xe − v te + ye2 + ze2
~˜ (e 1 γq ~ r µ)
E r µ) = i3/2 R(e (11.110)
4 π o
h 2
γ2 2
xe − v t + ye + ze
e 2
1 q ~ r µ)
= i3/2 R(e (11.111)
4 π o
hn 2 o
γ2 xe − v te + ye2 + ze2 − β 2 (e
y 2 + ze2 )
~ =e
where R r −~
~ v te.
Just as we did for the Lienard-Wiechert potentials, we may generalize this result in
terms of current position:
1 ~ r µ)
q R(e 1 − β2
~˜ =
E (11.112)
3
4 π o [R(er )] 1 − β 2 sin2 θ 3/2
µ
~˜ r µ ) = 1 β
B(e ~˜ (e
~×E r µ) (11.113)
c
We see the phenomenology is different from that for potentials (Equations 11.78 and
11.79): along and behind the direction of motion, the field is weakened by a factor
γ −2 while, transverse to the direction of motion, the field is enhanced by a factor γ.
Then, Fµν F µν consists of multiplying the two matrices not with standard matrix
multiplication but rather element by element and then summing over all elements.
That yields the Lorentz invariant:
2 ~ ~ ~ ·B
~ c2
Fµν F µν = − E · E + 2B =⇒ − Fµν F µν = E 2 − c 2 B 2
c2 2
(11.116)
This implies that the field strengths scale together from frame to frame: if one
increases or decreases, so does the other. Finding this invariant without first having
the Faraday tensor would have been quite difficult!
Section 11.3.6 Electromagnetic Field Tensor Invariant Page 834
Section 11.3 Relativity and Electrodynamics: Relativistically Covariant Formulation of Electrodynamics
The Dual Electromagnetic Field Tensor and a Second Field Tensor Invariant
We may define the dual tensor G using the completely antisymmetric four-index
Levi-Civita symbol µνλσ :
1 µνλσ
G µν = Fλσ (11.117)
2
which yields
G 00 = 0 G jj = 0 (11.118)
1 1
−G 0j
=G j0
= 0j λσ F λσ G = ijλσ F λσ
ij
(11.119)
2 2
1 1
= 0j ik F ik = ij0k F 0k + ijk0 F k0
2 2
1
= − jik ik` B` = ijk0 F k0
2
1 1
= jik `ik B` = ijk Ek
2 c
= δj` B` = Bj (11.120)
0j 0j
where λσ = ik on the left side because µνλσ
vanishes if any of its indices repeat,
so the j, λ, and σ indices must all take on space values 1, 2, 3 and thus all three must
be latin indices. Similarly, on the right side, because its first two indices are latin
indices, ij λσ is only nonzero if one of λ or σ is zero, which then requires the other to
become a latin index so it takes on space values.
Section 11.3.7 The Dual Electromagnetic Field Tensor and a Second Field Tensor Invariant Page 835
Section 11.3 Relativity and Electrodynamics: Relativistically Covariant Formulation of Electrodynamics
~ /c → B
which is obtained from F µν by the replacement E ~ and B
~ → −E
~ /c.
4 ~ ~ c ~ ·B
~
Fµν G µν = − E ·B =⇒ − Fµν G µν = E (11.122)
c 4
This invariant implies two things: 1) if the fields are perpendicular in one frame (or
one vanishes), then they remain perpendicular (or one vanishes) in any other frame;
2) if the above invariant is nonzero in some frame, then, since the field strengths
increase or decrease together, the magnitude of the cosine of the angle between them
changes in the opposite fashion, so the angle between them changes in the same
fashion as the fields; e.g., if the field strengths increase, they must become more
perpendicular to each other.
Section 11.3.7 The Dual Electromagnetic Field Tensor and a Second Field Tensor Invariant Page 836
Section 11.3 Relativity and Electrodynamics: Relativistically Covariant Formulation of Electrodynamics
Maxwell’s Equations
Finally, we can rewrite Maxwell’s Equations in a very simple form using the
electromagnetic field tensor and its dual. The inhomogeneous equations (the ones
with source terms on the right side, Gauss’s Law and Ampere’s Law) are:
∂µ F µν = ∂µ ∂ µ Aν − ∂µ ∂ ν Aµ = 2 Aν − ∂ ν ∂µ Aµ = µo J ν − 0 (11.123)
µν ν
∂µ F = µo J (11.124)
where the second term vanishes due to the Lorenz gauge condition (Equation 11.71).
The homogenous equations are
1
∂µ G µν = ∂µ µνλσ (∂λ Aσ − ∂σ Aλ ) = ∂µ µνλσ ∂λ Aσ = 0 (11.125)
2
∂µ G µν = 0 (11.126)
where the second step is possible because µνλσ is antisymmetric under exchange of λ
and σ (doing this exchange in the second term yields a copy of the first term) and the
third step holds because ∂µ ∂λ is symmetric under exchange of these indices. It is
interesting that physics statements — Faraday’s law and the divergencelessness of B~
— reduce to a mathematical identity thanks to the definitions of F and G. This is
something we have seen before: ∇ ~ = 0 can be viewed as a consequence of
~ ·B
B~ =∇ ~ It is a sign that we have defined G well!
~ × A.
It is straightforward to see that these equations yield the standard Maxwell Equations:
∂µ F µ0 = µo J 0 ∂µ F µi = µo J i (11.127)
00 i0 1
∂0 F + ∂i F = µo ρ c ∂t F 0i + ∂j F ji = µo J i (11.128)
c
∂ Ei 1 1 ∂Ei ∂Bk
0+ = ρ − 2 − jik = µo J i (11.129)
∂r i c c o c ∂t ∂r j
~
∇ ~= ρ
~ ·E ~ = µo J~ + o µo ∂ E
~ ×B
∇ (11.130)
o ∂t
∂µ G µ0 = 0 ∂µ G µi = 0 (11.131)
1 ~ ~
∂B
~ =0
~ ·B
∇ − ~
∇×E = c o µo (11.132)
c ∂t
∂B~
∇ ~
~ ×E =− (11.133)
∂t
d dx i
m = q E i = q c F i0 (11.134)
dt dt
dp µ
= q F µν vν (11.135)
dτ
This expression is now relativistically covariant because both sides of the equation are
rank 1 tensors (four-vectors).
Section 11.4.1 Lorentz Force and Four-Momentum Page 839
Section 11.4 Relativity and Electrodynamics: Relativistic Dynamics with Electromagnetic Fields
Of course, we need to check that this generalization yields the correct Lorentz Force in
an arbitrary frame, so let’s evaluate it in a frame in which the charged particle is not
at rest. We have for the space components (recall vµ = (γ c, −γ ~ v ), and writing F~ · rbi ,
~ · rbi to avoid sign ambiguities for space components (note that F~ is force
v · rbi , and p
~
while F µν is the field tensor!):
dp i
= q F i0 γ c − F ij γ ~
v · rbj (11.136)
dτ
dt dp i Ei
=q γ c + ijk Bk γ ~ v · rbj (11.137)
dτ dt c
γ F~ · ri = γ q Ei + ~
b v ×B ~ (11.138)
i
=⇒ F~ = q E~ +~ v ×B ~ (11.139)
where dt/dτ = γ (time dilation) and we note the subtlety that, in relativity, the
natural generalization of Newton’s Second Law is to replace F~ · rbi = (d/dt)(~
p · rbi ) by
dp i d
F~ · rbi = = v · rbi )
(γ m ~ (11.140)
dt dt
This generalization ensures the force and momentum can get as large as one wants
without |~v | exceeding c. We note that Griffiths defines K µ = dp µ /dτ as the
Minkowski force because it is a four-vector (transforms by the Lorentz transformation
rule); dp µ /dt is not.)
Section 11.4.1 Lorentz Force and Four-Momentum Page 840
Section 11.4 Relativity and Electrodynamics: Relativistic Dynamics with Electromagnetic Fields
Let’s look at the time component that comes along for free in the above. It is
d
γ m c = q F 00 γ c − F 0i γ ~ v · rbi (11.141)
dτ
dt d Ei
γ m c = q (0)(c) − − γ~ v · rbi (11.142)
dτ dt c
d Ei
γ γmc = γq v · rbi
~ (11.143)
dt c
dU ~
= q~ v ·E (11.144)
dt
where we have been motivated to define U = γ m c 2 as the relativistic generalization
of the energy of the particle. This is just conservation of energy: the rate of change of
the particle energy is given by the work being done by the electric field (the magnetic
field does no work).
We may now see the full meaning of the four-momentum we defined earlier:
U
⇒ ⇒
p =m v ⇐⇒ p µ = m v µ = γ m (c, ~
v) = ~
,γ p (11.145)
c
which is entirely consistent with what we have derived above. It may seem strange
⇒
that the interpretation of the time component of p as the particle energy comes as
an afterthought rather than being something we design in, but this makes sense given
our interest in covariant quantities: we could not necessarily know ahead of time what
quantity should be combined with p ~ to make a four-vector. Formulating a
relativistically covariant generalization of the Lorentz Force made the meaning of the
⇒
time component of p self-evident.
Section 11.4.1 Lorentz Force and Four-Momentum Page 841
Section 11.5 Relativity and Electrodynamics: Lagrangian Formulation of Relativistic Electrodynamics (Skip)
where τ is the proper time, not the volume element! We can check that this
Lagrangian recovers the nonrelativistic free particle limit:
!
β1 v |2
1 |~ 1
L −→ −m c 2
1− v |2
= −m c 2 + m |~ (11.148)
2 c2 2
which, aside from an irrelevant constant, is what we expect. Note the importance of
the negative sign in the definition of the full relativistic Lagrangian.
Section 11.5.1 Lagrangian Formulation for Particle-Field Interaction Page 842
Section 11.5 Relativity and Electrodynamics: Lagrangian Formulation of Relativistic Electrodynamics (Skip)
V
Lint = −q v µ Aµ = −q c −~ ~ = −q V − ~
v ·A ~
v ·A (11.150)
c
We can confirm that it yields a sensible canonical momentum (again using ~v · rbi for
the components of the three-velocity to avoid confusion with the space components of
the four-velocity):
1 v · rbi )
2 (~
∂L
πi = = −m c 2 γ − 2
− q −Ai = γ m (~
v · rbi ) + q Ai (11.151)
v · rbi )
∂ (~ 2 c
π = m~
which is what we expect from the nonrelativistic version ~ ~ along with the
v +qA
v → γm~
relativistic replacement m ~ v.
d ∂L ∂L
0= − (11.152)
v · rbi )
dt ∂ (~ ∂x i
d ∂ ~
= γ m (~ v · rbi ) + q Ai + q i V − ~ v ·A (11.153)
dt ∂x
" #
d ∂Ai ~ i ∂V ∂ ~
= (γ m (~ v · rbi )) + q + ~ v ·∇ A +q i −q i ~ v ·A (11.154)
dt ∂t ∂x ∂x
" #
i
d ∂V ∂A h i
~ · rbi
~ ×A
= (γ m (~ v · rbi )) + q + −q ~ v× ∇ (11.155)
dt ∂x i ∂t
where we used the convective derivative (Equation 10.24) to expand dAi /dt and then
we used the BAC − CAB rule to reconstruct the triple cross-product in the last
equation. Separating the time derivative of the momentum from the other terms, we
find
" #
d ∂V ∂Ai h i
~ · rbi
~ ×A
v · rbi )) = q − i −
(γ m (~ +q ~v× ∇ (11.156)
dt ∂x ∂t
h i
=q E ~ +~ ~ · rbi
v ×B (11.157)
which is exactly the relativistic version of the Lorentz Force Law we derived earlier
(Equations 11.139 and 11.140).
Section 11.5.1 Lagrangian Formulation for Particle-Field Interaction Page 844
Section 11.5 Relativity and Electrodynamics: Lagrangian Formulation of Relativistic Electrodynamics (Skip)
Did we have other options for Lint ? The only quantities we have for the EM field from
which to construct a Lagrangian are the covariant potential Aµ , the field strength
tensor F µν , and its dual G µν , and the only quantity from the particle is the four
velocity v µ . Had we tried to construct Lint using either of the field strength tensors
instead, we would have obtained a Lagrangian quadratic in v µ since these tensors
have two indices to contract over. If one starts to work out the resulting equations of
motion, one finds that they just aren’t sensible and cannot yield the nonrelativistic
limit. So we are left to use only Aµ and v µ , yielding the form we already checked
above.
The obvious question is: what do we write down for the Lagrangian density for the
EM field? A good starting principle is that we want the Lagrangian density to be a
Lorentz scalar. We know of two Lorentz scalars we can build with the EM field tensor,
Fµν F µν and Fµν G µν . We did note it before, but the latter version is undesirable
because G µν is a pseudotensor, meaning that it does not change sign under coordinate
inversion. The Lagrangian density itself should be invariant under coordinate
inversion, so the first form is but the latter form is not. So we choose the first form.
1
L=− Fµν F µν − Jµ Aµ
4 µo
Section 11.5.2 Lagrangian Density for the EM Field and Recovery of Maxwell’s Equations Page 846
Section 11.5 Relativity and Electrodynamics: Lagrangian Formulation of Relativistic Electrodynamics (Skip)
Next, we must obtain the Euler-Lagrange equations for this Lagrangian. We have to
choose the degrees of freedom. While it might seem like they should be the
components of F µν , we know these components are highly redundant. We have the
mathematical constraint that arises via construction that ∂µ G µν = 0
(Equation 11.124). This, along with the presence of Aµ in the interaction Lagrangian
density, suggests that we should treat Aµ as our degrees of freedom.
The second question is: what are the relativistically covariant Euler-Lagrange
Equations? It turns out the correspondence is (see, e.g., Goldstein, Poole, and Safko)
!
d ∂L ∂L
∂L ∂L
− =0 −→ ∂µ − =0
dt ∂ q̇k ∂qk ∂ (∂ µ Aν ) ∂Aν
Given the above two points, it makes sense to rewrite the Lagragian density in a
manner that makes the dependence on Aν explicit and also makes all appearances of
∂ µ and Aν consistent with the version in the Euler-Lagrange Equations:
1
L=− gαβ gγδ ∂ β Aδ − ∂ δ Aβ (∂ α Aγ − ∂ γ Aα ) − Jµ Aµ
4 µo
Section 11.5.2 Lagrangian Density for the EM Field and Recovery of Maxwell’s Equations Page 847
Section 11.5 Relativity and Electrodynamics: Lagrangian Formulation of Relativistic Electrodynamics (Skip)
∂L 1 h i
µ ν
=− gαβ gγσ δµβ δνσ F αγ − δµσ δνβ F αγ + δµα δνγ F βσ − δµα δνγ F βσ
∂ (∂ A ) 4 µo
1
= − Fµν by symmetry of gµν and antisymmetry of Fµν
µo
∂L
= −Jµ
∂Aν
Thus, we have
1
∂µ − Fµν + Jµ = 0 =⇒ ∂µ F µν = µo J µ
µo
Section 11.5.2 Lagrangian Density for the EM Field and Recovery of Maxwell’s Equations Page 848
Section 11.6 Relativity and Electrodynamics: Relativistic Conservation Laws (Skip)
1 1
T µν = F µλ F λν + g µν Fλσ F λσ (11.158)
µo 4
The components of T are (recall, we found the relativistic invariant Fµν F µν before):
1 1 2
T 00 = F 00 F 00 + gij F 0j F i0 + g 00 − 2 E 2 − c2 B2
(11.159)
µo 4 c
1
E 2 2
1 E
o
− B2 E 2 + c 2 B 2 = ufield
= (−1) − 2 − = (11.160)
µo c 2 c2 2
Section 11.6.1 Maxwell Energy-Momentum Tensor and Conservation of Energy-Momentum Page 849
Section 11.6 Relativity and Electrodynamics: Relativistic Conservation Laws (Skip)
1 g ij E 2
T ij = F iλ gλσ F σj − − B 2
(11.161)
µo 2 c2
1
δ ij E 2
= g00 F i0 F 0j + gk` F ik F `j + − B 2
(11.162)
µo 2 c2
1
δ ij E 2
= F i0 F 0j − F ik F kj + − B 2
(11.163)
µo 2 c2
1 1 δ ij E 2
2
= − 2 Ei Ej − ik` B` kjm Bm + − B (11.164)
µo c 2 c2
δ ij
= o −Ei Ej + c 2 i`k B` jmk Bm + E 2 − c 2B 2
(11.165)
2
δ ij
= o −Ei Ej + c 2 δij δ`m − δim δ`j B` Bm + E 2 − c 2B 2
(11.166)
2
δ ij
= o −Ei Ej − c 2 Bi Bj + E 2 + c 2B 2 = − T
(11.167)
2 ij
i.e., the negative of the Maxwell Stress Tensor we defined in Equation 8.31. (To be
clear, T is the four-tensor while T is the three-tensor.) Note that we allow space-space
(roman) indices to be contracted without requiring one raised and one lowered index.
Section 11.6.1 Maxwell Energy-Momentum Tensor and Conservation of Energy-Momentum Page 850
Section 11.6 Relativity and Electrodynamics: Relativistic Conservation Laws (Skip)
1 0λ 1 0j 1 1
T 0i = F gλσ F σi = F gjk F ki = − F 0j δjk F ki = − F 0j F ji (11.168)
µo µo µo µo
1 Ej 1 1 1
T 0i = −
− −jik Bk = ijk Ej Bk = Si = c pfield,i (11.169)
µo c µo c c
~=E
where S ~ × B/µ
~ o is the Poynting vector. Summarizing,
ufield c~
g
T µν = (11.170)
c~g −T
Section 11.6.1 Maxwell Energy-Momentum Tensor and Conservation of Energy-Momentum Page 851
Section 11.6 Relativity and Electrodynamics: Relativistic Conservation Laws (Skip)
With the energy-momentum tensor in hand, it is natural to rewrite our energy and
linear momentum conservation laws using it. If we take the four-divergence of the
energy-momentum tensor, we find
1 1
∂µ T µν = ∂µ F µλ F λν + F µλ ∂µ F λν + ∂ ν Fλσ F λσ
(11.171)
µo 4
1 h i
∂µ T µν − Jλ F λν = Fµλ ∂ µ F λν + ∂ µ F λν + ∂ ν F µλ (11.172)
2 µo
We may rewrite the last two terms using the homogeneous Maxwell Equation, which
we rewrite as:
1
0 = ∂σ G σλ = ∂σ σλνµ Fνµ = ∂ λ F νµ + ∂ µ F λν + ∂ ν F µλ (11.173)
2
−∂ λ F νµ = ∂ µ F λν + ∂ ν F µλ (11.174)
The last step in the first line is obtained by just writing out all the terms and
combining the ones that differ by a single flip of two indices; it may seem strange that
four equations (λ = 0, 1, 2, 3) became 64 equations (λ, µ, ν = 0, 1, 2, 3), but many
of them vanish and the others are redundant (see how convenient the dual tensor is!).
Section 11.6.1 Maxwell Energy-Momentum Tensor and Conservation of Energy-Momentum Page 852
Section 11.6 Relativity and Electrodynamics: Relativistic Conservation Laws (Skip)
Using this in the equation for the divergence of the energy-momentum tensor, and
then reordering the indices on the second term on each side (and picking up minus
signs), we have
1 h i
∂µ T µν − Jλ F λν = Fµλ ∂ µ F λν − ∂ λ F νµ (11.175)
2 µo
1 h i
∂µ T µν + F νλ Jλ = Fµλ ∂ µ F λν + ∂ λ F µν (11.176)
2 µo
The quantity in brackets on the right side is now symmetric in µ and λ, while Fµλ is
antisymmetric, so the right side vanishes. Moving the field-current term to the right
side, we obtain
∂µ T µν = −F νλ Jλ (11.177)
If we write out the time and space components of this four-vector equation, we obtain
∂ ufield ~ ·S ~ = − ∂ umech
~ = −J~ · E
+∇ (11.178)
∂t ∂t
∂~ g h
~ · T = − ρE ~ + J~ × B
i ~mech
~ = −∂p
−∇ (11.179)
∂t ∂t
which are Equations 8.15 and 8.40 from our discussion of conservation laws.
Section 11.6.1 Maxwell Energy-Momentum Tensor and Conservation of Energy-Momentum Page 853
Section 11.6 Relativity and Electrodynamics: Relativistic Conservation Laws (Skip)
M µνσ = − [T µν r σ − T µσ r ν ] = T µσ r ν − T µν r σ (11.180)
where the ∧ indicates the antisymmetric difference over the index of the space-time
position four-vector and the last index of the stress tensor.
Let’s write down a conservation law for this tensor using ∂µ T µν = −F νλ Jλ and
∂µ r ν = δµν :
where, in the penultimate step, we used the symmetry of the stress tensor and, in the
last step, we used the asymmetry of the wedge product (the νσ indices are the
wedge-product indices). Let’s break this down piece-by-piece.
where we have written down the differential version of Equation 8.51, replacing the
cross products in that expression with wedge products, and we have also replaced the
mechanical angular momentum density cross product ~ `mech with the wedge product
version ` . Thus, we have
mech
∂ ~ ·M=− ∂ `
` −∇ (11.196)
∂t field ∂t mech
which is Equation 8.68 showing that the rate of change of the angular momentum
density is given by the divergence of the torque tensor.
Page 858
Lecture 48:
Radiation I:
Potentials and Fields of a Fixed-Velocity Point Charge
Date Revised: 2022/05/16 07:15
Date Given: 2022/05/16
Page 859
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
We proceed in a very different manner than Griffiths. In §10.3.1, Griffiths uses the
retarded potential formulae (Equation 10.71) to do a direct calculation of the
Lienard-Wiechert potential, which is challenging. Then, in §10.3.2, Griffiths brute-force
differentiates the potentials to get the fields. Instead, we have used relativity to obtain
the potentials of a particle moving at fixed velocity — the Lienard-Wiechert potentials
— and its fields in terms of the current position of the particle. We can rewrite the
results in terms of the retarded time. Then, we return to the Lienard-Wiechert
potentials and incorporate the effect of acceleration through possible time-dependence
of the velocity by direct differentiation allowing the velocity to vary, which is much
easier now that we have accounted for all other time dependences via the
relativity-based derivation of the Lienard-Wiechert potentials and corresponding fields.
Finally, from the fields we calculate the radiated power pattern for the general case,
specializing to slowly moving particles at the end to obtain the Larmor Formula.
Our treatment of the calculation of the fields follows, partially, that of M. Cross’s
Ph106c lecture notes, which do not appear to follow any specific textbook. The
treatment of radiation follows that of Heald and Marion Sections 8.7 and 8.8.
We will not cover the topic of radiation reaction (Griffiths 11.2.2 and 11.2.3) — it’s
interesting, but there is not much to add to what Griffiths says.
Let’s first reiterate the potentials and fields of a moving point charge that we
calculated using Lorentz Transformations in a form that will be useful going forward.
From Equations 11.78 and 11.79:
Lienard-Wiechert
1 q 1
scalar potential V (~
r , t) = p (12.1)
using current position 4 π o R(~
r , t) 1 − β 2 sin2 θ
Lienard-Wiechert ~
vector potential ~ r , t) = µo q c β p
A(~
1
(12.2)
using current position 4 π R(~
r , t) 1 − β 2 sin2 θ
It is very important for us to note that, even though the above potentials were derived
using Lorentz Transformations — i.e., fixed velocity — they turn out to be the same
as what one would get by working from the general forms for the retarded potentials
(Equation 10.71). That derivation is done in Griffiths §10.3.1.
Section 12.1.2 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Current Position Page 862
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
There is an intuitive reason for this matchup, which does not hold for the fields. The
retarded potential expressions involve the position of the point charge. The process of
doing the integral can involve Jacobians, which involves first derivatives and thus
might involve the velocity. But there is no way for second derivatives to appear, and
thus acceleration cannot be relevant. If acceleration does not matter, then the
fixed-velocity derivation using Lorentz Transformations must be valid even in the
accelerating case.
Electric field ~
~ (~ q R 1 − β2
of moving point charge E r , t) = 3
(12.3)
4 π o R 1 − β 2 sin2 θ 3/2
using current position
Magnetic field ~ ~
~ r , t) = µo q c β × R 1 − β2
of moving point charge B(~ 3 3/2 (12.4)
using current position 4π R 1 − β 2 sin2 θ
Section 12.1.2 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Current Position Page 863
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Comments:
I The dependences on q, v = c β, and R ~ are unchanged from the β = 0 limit.
I The forms suggest that the potential and field information at the field point
acts as if the particle had kept moving at the velocity it had when the potential
information left the particle. Effectively, this retarded potential carries
information not just about the position of the particle at the retarded time but
also about its velocity! And, of course, it is not affected yet by later time events!
I The potentials and fields are forward-backward symmetric in magnitude. We
shall see that, when we rewrite using retarded time, they are not!
I The potentials
p are enhanced in the transverse direction by a factor
γ = 1/ 1 − β 2 and take on the static values in the forward and backward
directions along the direction of motion.
I The fieldsphave enhanced angular and β dependence. The field is enhanced by a
factor 1/ 1 − β 2 (the Lorentz factor of the particle, γ) in the plane transverse
to the direction of motion (at θ = π/2) and reduced by a factor 1 − β 2 along
the axis of motion (θ = 0 or θ = π).
I The relation B~ ∝β ~×E ~ along with E
~ ∝R
~ implies that B
~ wraps around β
~
following the right-hand rule.
Section 12.1.2 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Current Position Page 864
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
AO : ~ r (t) = ~
R r −w
~ (tr ) BO : ~
R(t) r −w
=~ ~ (t) (12.6)
~ r and a boldface R for our R.
Griffiths uses a boldface script r for our R ~ These are the
retarded relative position and the current relative position.
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 865
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
~ r and R
because the right triangles with R ~ as hypotenuses share the same vertical side,
PO. We also have
We use the above two relations to substitute for cos α and sin α in the expression for
cos γ:
β Rr + R cos θ R sin θ
cos γ = sin θ sin θ − cos θ = β sin θ (12.12)
Rr Rr
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 866
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
q
R ~·R
1 − β 2 sin2 θ = Rr 1 − β br (12.16)
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 867
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Inserting this into Equations 12.1 and 12.2, we obtain
Lienard-Wiechert
1 q 1
scalar potential V (~
r , t) =
4 π o Rr (~ ~ r) · R
r , t) 1 − β(t br (~
r , t)
using retarded position
(12.17)
Lienard-Wiechert ~
vector potential ~ r , t) = µo q c β(tr )
A(~
1
(12.18)
4 π Rr (~ ~ r) · R
r , t) 1 − β(t br (~
r , t)
using retarded position
These are versions of the Lienard-Wiechert potentials that use the retarded position of
the particle. It has been assumed that the particle has constant velocity between tr
and t. The β → 0 limit is also clear here: the correction term becomes unity.
Comments:
I These potentials are what one might expect based on the retarded time, but
with a correction factor related to the angle between the direction of motion
and the position vector between the moving particle and the observer.
I While β~·R br < 0 is possible, the correction factor is always nonnegative because
β < 1.
I The potentials are stronger than the steady-state case in the half-space ahead of
the particle and weaker in the half-space behind it, with the steady-state value
obtained on the boundary. This effect can be understood in terms of a “piling
up” effect that is like that of the non-relativistic Doppler shift. This
phenomenon is discussed in more detail in Griffiths §10.3.1.
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 868
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 869
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
We can also use the above form to obtain the fields in terms of the retarded position.
Recall the relation that we used to convert the Lienard-Wiechert potentials from
retarded position to current position:
q
R ~·R
1 − β 2 sin2 θ = Rr 1 − β br (12.19)
~ ∝R
In addition, we note that, because E ~ = Rr (R ~ then (R
br − β), ~ ×E
br − β) ~ = 0 and
therefore
~ = 1β ~= 1 β ~= 1R
h i
B ~×E ~+ R ~ ×E
br − β ~
br × E (12.21)
c c c
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 870
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Inserting the above relations into the equations for the fields in terms of the current
position, we have
Electric field R~r − β
~ Rr
1 − β2
~ (~ q
of moving point charge E r , t) = i3 (12.22)
4 π o Rr3
h
using retarded position ~·R
1−β br
Magnetic field ~ ~ 2
of moving point charge ~ r , t) = µo q c β × Rr h 1 − β i
B(~ (12.23)
4π Rr3 3
using retarded position 1−β~·Rbr
These match Griffiths Equations 10.72 and 10.73 with zero acceleration (~ ~
a = 0). B
~ as for the current position fields.
wraps around β
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 871
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Comments:
I These fields have reduction factor 1 − β 2 in the numerator, but the overall
reduction/enhancement relative to the static case is more complex.
I Unlike the current position fields, there is a forward-backward asymmetry with
enhancement (1 − β 2 )/(1 − β)3 in the forward direction and reduction
(1 − β 2 )/(1 + β)3 in the backward direction.
I These factors are different from those for the potentials, which had only one
power of 1 − β ~·Rbr in the denominator.
I The electric field does not point along R~ r : the field points along the current
relative position vector, not the retarded relative position vector
I Again, information about both the position and velocity seems to be
transmitted in the fields.
Section 12.1.3 Potentials and Fields of a Fixed-Velocity Point Charge in Terms of Retarded Position Page 872
Lecture 49:
Radiation II:
Potentials and Fields of an Accelerated Point Charge
Bremsstrahlung
Date Revised: 2022/05/18 14:30
Revised lecture break
Date Given: 2022/05/18
Page 873
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
We will find that it is more convenient (or, rather, less inconvenient) to use the
retarded position expressions for the Lienard-Wiechert potentials.
We assume the particle has a trajectory w~ (t) and that its instantaneous velocity is
~
β(t) =w~ 0 (t)/c. We will end up evaluating these and derivatives thereof at tr .
c 2 (t − tr )2 = (~
r −w
~ (tr )) · (~
r −w
~ (tr )) (12.25)
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 874
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Let’s take the partial derivatives of both sides. First, with respect to t holding ~
r fixed,
and recalling that c (t − tr ) = Rr and β~=w ~ 0 /c,
∂ tr ∂ tr
2 c 2 (t − tr ) 1 − = 2 (~
r − ~
w (tr )) · −w~ 0
(tr ) (12.26)
∂t ~r ∂t ~r
~ r ) ∂ tr = −c Rr
h i
~ r · c β(t
− c Rr − R (12.27)
∂t ~r
∂ tr 1
= (12.28)
∂t ~r ~
1 − β(tr ) · Rbr
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 875
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
~ defining β
From these, we can calculate derivatives of β, ~0 = w ~ 00 /c as the first
~ or second derivative of w
derivative of β ~ /c with respect to its argument:
∂ βi
∂ tr ~
∂β
∂ tr
= βi0 (tr ) =⇒ ~ 0 (tr )
=β (12.32)
∂t ~r ∂t ~r ∂t ∂t ~r
~
r
∂ βi ∂ tr ~
∂β ∂ tr
= βi0 (tr ) =⇒ ~ 0 (tr )
=β (12.33)
∂rj ∂rj ∂rj ∂rj
t t t t
∂ X Rr ,i X Rr ,i ∂ tr
∂ ~ b br ∂ tr
βi0 ~0 ·R
β · Rr = βi = = β (12.34)
∂t ∂t R R ∂t ∂t
r r
b
~
r ,R
br
i
i ~
r ,R
br ~r ,Rr
~
r ,R
br
∂ ~ b
∂ X Rr ,i X Rr ,i ∂ tr
br ∂ tr
~0 ·R
β · Rr = βi = βi0 = β (12.35)
∂rj ∂rj i Rr b R ∂r ∂ri b
b
i r i
t,Rr t,Rr b t,Rr t,Rr
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 876
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Next, to calculate E~ , we start from Equations 12.17 and 12.18 but now allow β~ to be
differentiated also. We only show the new terms, understanding that β ~ and β
~ 0 are
evaluated at tr :
∂V ∂V 1 q −1 ∂ h
~·R
i
− = − − i 2 − β (12.36)
br
∂ri ∂ri 4 π o Rr 1 − β ∂ri
h
~
β ~·Rbr b
t,Rr
∂ V q 1 Rr ,i 1 ~0 ·R
= − + i3 β br (12.37)
∂ri ~ 4 π o c Rr 1 − β
h
β ~·R b r
~ b V = −∇
~ ~r V + q 1 R br 1 ~0
=⇒ −∇~r ,R
br · β
i3 R (12.38)
r ~
4 π o c Rr 1 − β
h
β ~·R br
!
~
∂A ~
∂A µo q c −β ~ ∂ ~ b
h i
1 ~
∂β
− = − − i2 − β · Rr +
∂t ∂t ~ 4 π Rr ∂t ~·Rbr ∂t b
h
β ~
1 − β · Rr
b ~
r ,Rbr 1 − β ~
r ,Rr
(12.39)
" #
~
∂A µo q c 1
= − − i3 β~ β~0 ·R ~·R
br + 1 − β br β~0
∂t ~ 4 π Rr 1 − β
h
β ~·Rbr
(12.40)
~ + q 1 1
h i
~= E
E i3 R ~ R
br − β br · β ~·R
~0 + 1−β ~0
br β
~ 4 π o c Rr 1 − β
h
β ~·R r
b
(12.41)
Finally, using the BAC − CAB rule and adding powers of Rr to turn R ~r :
br into R
h i Electric field
~ ~ ~ ~0 of accelerated
~ (~ ~ (~
q 1 1 Rr × Rr − β Rr × β
E r , t) = E r , t) + point charge
i3
~ 4 π o Rr3 c
h
β ~·R
1−β br using retarded
position
(12.42)
(β~ and β~ 0 are evaluated at tr .) We see that the expression is very similar to the
~ ~r − β~ Rr 1 − β 2 is replaced with the
fixed-β term, Equation 12.22, except that R
triple cross product. We also see that this term is smaller than the fixed-β ~ term by a
factor (Rr /c) β 0 , which is sensible: this factor is the ratio of the light travel time to
the characteristic relativistic acceleration timescale (i.e., where the latter is normalized
to c, not to c β, as we discussed in the context of Jefimemko’s Equations,
Equations 10.106 and 10.107.)
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 878
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Similarly,
" #
X ∂ µo q c βk (tr ) 1
Bi = Bi |β~ + ijk (12.43)
∂rj 4 π Rr (~ ~ r) · R
r , t) 1 − β(t
br (~
r , t) t,Rb
j,k r
∂ βk
P
j,k ijk ∂r
P ∂ ~
β · Rr
j,k βj ∂r
b
µo q c j t,R br k t,R
br
= Bi |β~ + − (12.44)
~·R i2
4 π Rr
h
1−β br
1−β ~·R b r
P R
R
0 − c rR,j
P ~0 ·R − c rR,k
j,k ijk βk βj β br
µo q c r j,k r
= Bi |β~ + − (12.45)
i2 i3
4 π Rr
h h
~
1 − β · Rr
b ~
1 − β · Rr
b
µ q R ~0
br × β ~×R
β
~ + o
br
=⇒ ~ = B
B − h
i2 + h i3 R ~ 0
br · β (12.46)
~ 4 π Rr
β ~·R ~·R
1−β br 1−β br
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 879
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Therefore,
~ ~ ~0 ~ ~ ~0 ~ ~
~ ~
µo q Rr − β · Rr β × Rr + Rr · β β × Rr
B(~
r , t) = B(~
r , t) + (12.47)
i3
~ 4 π Rr3
h
β ~·R
1−β b r
~ = 1R
B ~ 6= 1 β
br × E ~×E
~ Magnetic field of accelerated point
(12.48)
c c charge using retarded position
~ and β
(β ~ 0 are evaluated at tr .) As with E ~ , we see a form similar to the fixed-β
~ term
with replacement of β ~×R ~ r (1 − β 2 ) by the complicated expression in the numerator.
This term is also smaller than the fixed-β ~ term by a factor (Rr /c) β 0 . Only one of the
~ ~
relations between E and B is preserved.
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 880
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
The direction of these acceleration fields is dramatically different from that of the
fixed-velocity fields. The latter were similar in direction to the static field, with E ~ ∝ R,~
the current position vector and thus being radial, and B ~ wrapping around β ~ in the
usual azimuthal manner. By contrast, the acceleration electric field is perpendicular to
~ r ; thus it is transverse to the vector from the retarded position. We will see below
R
that the remaining direction choice for E ~ is set by the acceleration vector β~ 0 : the
electric field is in the plane formed by R ~ r and β
~ 0 . The magnetic field direction is
completely determined by R ~ r and E
~ : it is perpendicular to both, making it also
transverse relative to the retarded position (and thus also determined by R ~ r and β~ 0 ).
It is striking how the fixed-velocity fields seem to emanate from the current position of
the particle, and thus they carry information about the particle’s position and velocity
at the retarded time but transfer it forward to the current position, while the
acceleration fields clearly emanate from the particle’s retarded position and are
determined by its acceleration at the retarded time. The acceleration fields “make no
assumption” about the acceleration being fixed at times in the future, while the
fixed-velocity fields assume fixed velocity into the future. (The latter is presumably a
consequence of Galilean relativity.)
Section 12.1.4 Potentials and Fields of an Accelerated Point Charge Page 881
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
With the above detailed formulae for the fields due to an accelerated charge, we can
make our first study of radiation, which consists of the propagating fields created by
accelerated charges. We will of course calculate the power radiated using the Poynting
vector, S ~=E ~ × B/µ
~ 0 (no need for complex conjugations or taking real parts because
all fields are real here; also, we don’t time average because we are not yet considering
sinusoidal behavior). We can quickly see that only the terms involving the acceleration
~
a=cβ ~ 0 are important at large distances by extracting the dependences of various
terms from the full expressions:
~ : q 1 1
~ : q 1 β 1
E 2 2
B (12.49)
β~ o γ Rr β~ o c γ 2 Rr2
~−E
~ : q 1 ~0 1 ~−B
~ : q 1 ~ 0 1 (12.50)
E β B (1 + β) β
β~ o c Rr β~ o c 2 Rr
The |β~ terms are called the velocity terms and the others are called the acceleration
terms. Only the latter are important at large distances, the so-called far field.
Section 12.1.5 Poynting Vector Radiated by an Accelerating Point Charge Page 882
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
I We again recognize that the acceleration terms differ from the velocity terms by
a factor (Rr /c) β 0 , which is the ratio of the light travel time to the relativistic
acceleration timescale τa . This factor was introduced by the derivatives of β ~
that were taken to obtain the fields from the potentials.
I Note, too, that this factor causes the replacement of a factor of Rr in the
denominator with the length quantity c/β 0 , which is the distance light travels in
an acceleration timescale. This is a key replacement, as it is what make the
acceleration fields dominant at large distances, yielding the 1/R 2 law for the
radiated power!
I The acceleration electric field is perpendicular to R br (because it is the result of a
cross-product including R br ), which points along the line-of-sight from the
retarded position of the point charge to the point at which we want to know the
fields. (Let’s use the term “field point” for this point, ~ r .) This may be easier to
visualize if you imagine looking at the retarded position from the perspective of
the field point: Rbr is the vector pointing at you along that line-of-sight.
I The acceleration magnetic field is perpendicular to both this line-of-sight and
the acceleration electric field (because the magnetic field is proportional to the
cross product of R br and the electric field).
Section 12.1.5 Poynting Vector Radiated by an Accelerating Point Charge Page 883
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
~≈ 1 1
1
S ~−E
E ~ × B~−B ~ = ~−E
E ~ × br × E
R ~−E
~
µo β~ β~ 2 µo β~ c β~
2
~≈ 1 R ~−E ~
S br E (12.51)
c µo ~
β
where the last step used the BAC − CAB rule and R ~ −E
br · (E ~ | ~ ) = 0. We see that the
β
Poynting vector is along the line-of-sight from the retarded position to the field point.
We’ll consider the cases of the acceleration parallel and perpendicular to the velocity
separately and then add the two to get the fully relativistic result. We can then either
take its non-relativistic limit to get the Larmor Formula, or we can start from a
non-relativistic version of the electric field to get the same result.
Section 12.1.5 Poynting Vector Radiated by an Accelerating Point Charge Page 884
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Acceleration Parallel to Velocity: Bremsstrahlung
Let’s first do the parallel acceleration case. A standard application of this case is to
calculate bremsstrahlung (“braking radiation”), the radiation an electron gives off as it
is decelerated by interaction with matter (primarily the Coulomb force from the
positive charge of nuclei.) We set β ~×β ~ 0 = 0. The relevant piece of Equation 12.42 is
br × R
R br × 1 β~0 1 ~0
q 1 c q 1 β
E~~ 0 ~ = i3 =− c ⊥
i3 (12.52)
β ||β 4 π o Rr
h
4 π o Rr 1 − β
h
1−β ~·Rb r
~·R b r
projection of acceleration
with ~0 = β
β ~0 −R
br R ~0
br · β perpendicular to line-of-sight from (12.53)
⊥
the retarded position (Rbr )
or ~ a−R
a⊥ = ~ br Rbr · ~
a (12.54)
where we used the BAC − CAB rule to evaluate the triple cross product. The result
tells us that only the projection of the acceleration perpendicular to the line-of-sight is
responsible for the far-field radiation. (Be sure not to confuse this with the case of
acceleration perpendicular to the velocity! In fact, in this case, the acceleration is
parallel to the velocity.) A bit strange: velocity and acceleration in one direction only
yields radiation at angles away from that direction, and the radiation is strongest in
the direction perpendicular to the acceleration and velocity. If one thinks about the
relationship between the acceleration and the electric field direction, though, it makes
perfect sense: the electric field in a given direction is only affected by acceleration
parallel to that direction.
Section 12.1.6 Acceleration Parallel to Velocity: Bremsstrahlung Page 885
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
Next, if we define ~a to be the z-axis of a spherical coordinate system and θ the polar
angle of the field point, then we have ~a = a zb, |~ ~·R
a⊥ | = a sin θ, and β br = β cos θ, so
~ = µo q
2 1 a2 sin2 θ
S R
br (12.55)
16 π c Rr [1 − β cos θ]6
2 2
(We rewrote o in terms of c and µo .) The energy radiated into a unit area per unit
field point time interval is d 2 U/dt dA = R ~
br · S.
However, we will usually be interested in knowing the energy lost by the particle per
unit time, so we need to convert the dt in the above to a dtr by dividing by ∂tr /∂t,
which we calculated in Equation 12.28. The distinction between the two is the same
(nonrelativistic!) “piling up” effect that we discussed earlier in the context of
calculating the Lienard-Wiechert potential. Therefore (multiplying by Rr2 to convert
from power per unit area to power per unit solid angle):
!−1 !−1
dP d 2U d 2U ∂ tr ~ ∂ tr
= = = Rr2 R
br · S
dΩ dtr dΩ dt dΩ ∂t β~ ∂t β~
µo q 2 1 a2 sin2 θ
~·R
= 2 6
1−β br (12.56)
16 π c [1 − β cos θ]
The following figure is a polar plot that illustrates the shape of this function for a
particle moving from left to right. When β 1, one obtains a distribution symmetric
between the forward and the reverse direction. As β increases, the denominator begins
to take effect and the radiation is strongly directed into the forward hemisphere
(though the radiation along the direction of motion continues to vanish), with the
peak intensity found on a cone.
One can integrate the above over all angles to obtain the total radiated power
where γ = (1 − β 2 )−1/2 is the usual relativistic Lorentz factor. Note the strong
dependence on γ!
Page 889
Section 12.1 Radiation: Potentials, Fields, and Power Radiated by an Accelerated Point Charge
θ is the polar angle relative to the particle’s direction of motion (i.e., angle away from
zb in this case), while φ is the azimuthal angle around the particle’s direction of
motion, with φ = 0 being along the acceleration direction (b x in this case).
The dashed lines indicate the angle at which the radiated power vanishes, and the ×
factors indicate the enhancement of the “backward” lobe for visualization (which, in
fact, moves to the forward hemisphere as β → 1).
Combining the two total power formulae, which only differ by a factor of γ 2 , yields the
total power radiated for an arbitrary angle between velocity and acceleration:
~
2 total power radiated
µo q 2 a2 1 − β × ab
µo q 2 a2 6
2
by an accelerated
~
P= = γ 1 − β × ab (12.62)
6 π c (1 − β 2 )3 6π c point charge
Lienard’s Formula
Non-relativistic Limit
The nonrelativistic or “slowly moving charge” limit is obtained by letting β → 0 but
allowing β 0 6= 0 in Equation 12.42. The electric field simplifies to
~
h
~ ~0
i Electric field
~ (~ ~ (~
q Rr × Rr × β of accelerated
E r , t) = E r , t) + (12.63)
~
β 4 π o c Rr3 point charge in
non-relativistic limit
where the first, fixed-velocity term still falls off more quickly than the second,
acceleration term in this limit.
We will calculate the radiated power from Equation 12.51. If we again define ~ a to be
the z-axis of a spherical coordinate system and θ the polar angle of the field point so
that ~ ~ r = rb (= R
a = a zb, R ~ because β → 0), then we have
2 a r2 2
~ 0 = a r rb × [b b sin θ = a r θb sin θ
h i
~r × R
R ~r × β r × zb] = − rb × φ (12.64)
c c c
Therefore, canceling out factors of r 2 = Rr2 ,
2 q 2 a2 sin2 θ
~−E
~ =
E (12.65)
~
β 4 π o c4 r 2
Therefore,
1 b ~ 2 2 2 2
~=
S ~ = rb µo q a sin θ
Rr E − E (12.66)
cµo ~
β 16 π 2 c r 2
We could have obtained this limit from the fully relativistic Equation 12.61 by letting
β → 0 and β 0 6= 0 as we did here. We would also see that the angular radiation
pattern of both the parallel and perpendicular acceleration cases (bremsstrahlung and
synchrotron) both converge to the above simple, symmetric sin2 θ dependence because
the direction of motion does not break the symmetry.
We note three length scales that will be important in the discussion and on whose
relative sizes we will base various approximations:
Fourier Transforms
We are going to use Fourier techniques to simplify the derivations in the case of
radiation from an arbitrary source distribution, so we need to define the formalism.
The function ge(f ) is the Fourier Transform of the function g (t). It has the units of g
divided by frequency. For g (t) to which no boundary conditions have been applied
(typical BC would be to assume periodicity with a particular frequency f0 or to assume
Dirichlet or Neumann BC at two times ta and tb ), there is no restriction on the values
of f — any value must be allowed. (In real life, the rate at which you can sample g (t)
sets an upper limit on the values of f for which information is available — the
so-called Nyquist criterion — but we won’t worry about real life here...).
Note that our Fourier Transform sign convention for the argument of the exponential
is the opposite of that used in electrical engineering (which gives the impedance of
inductors and capacitors as i ω L and 1/i ω C ). As long as one is self-consistent, one
can choose whichever convention one likes. But be careful when comparing to other
texts.
r ) e −i ω t
r , t) = ρ0 (~
ρ(~ ~ r , t) = J~0 (~
J(~ r ) e −i ω t (12.70)
~ (~
E ~ 0 , J~0 ](~
r , t) = E[ρ r , f ) e −i ω t ~ r , t) = B[ρ
B(~ ~ 0 , J~0 ](~
r , f ) e −i ω t (12.71)
~ ](~
where E[ r , f ) and B[ ~ ](~ r , f ) indicate functional dependence: put in the spatial
r ), J~0 (~
functions ρ0 (~ r ) and the frequency f and what you get out are the spatial
~ r , f ), B(~
functions E(~ ~ r , f ). The linearity of Maxwell’s Equations assures us that the
harmonic time dependence is carried through from the sources to the fields. More
importantly, linearity assures us that, if we have a source distribution with arbitrary
time dependence that can be broken down in terms of components with harmonic
time dependence like that shown above, then we can calculate the fields for each
component using E[ ~ ](~ r , f ) and B[ ~ ](~
r , f ) and then sum them up to get the total field.
The Fourier Transform is the tool for doing all this.
The motivation for these expressions is that multiplication by the e i ω t factor followed
r ) and J~0 (~
by the integration would pick out the ρ0 (~ r ) with the e −i ω t time
dependence shown on the previous page. Then, from the above solution that
determines fields from sources, we know
~˜ (~
E r , f ) = E[e
˜
r , f ), ~J(~
~ ρ(~ r , f )] ~˜ r , f ) = B[e
B(~ ~ ρ(~ ˜
r , f ), ~J(~
r , f )] (12.73)
~ ](~
Therefore, we only need determine the functions E[ ~ ](~
r , f ) and B[ r , f ) by
determining the fields for a harmonic time dependence e i ω t of the sources and then
we can use Fourier Transforms to calculate the fields for arbitrary source time
dependence.
Linearity also holds, surprisingly, for quadratic quantities like energy and radiation if
one time averages. Specifically, let’s calculate the time-averaged Poynting vector:
D E
~ r , t) = 1 D ~∗ ~ r , t)
E
S(~ R E (~r , t) × B(~ (12.75)
2 µo
Z ∞ Z ∞
1
= R df1 ~˜ ∗ (~
df2 e i(ω1 −ω2 )t E ~˜ r , f2 )
r , f1 ) × B(~ (12.76)
2 µo −∞ −∞
Z ∞ Z ∞
1
~˜ ∗ (~ ~˜ r , f2 )
D E
= R df1 df2 e i(ω1 −ω2 )t E r , f1 ) × B(~ (12.77)
2 µo −∞ −∞
Now,
R T /2
D E 1
Z T /2
−T /2
dt e i ω t δ(f )
iωt iωt
e = lim dt e = lim R T /2 = (12.78)
T →∞ T −T /2 T →∞ dt δ(f = 0)
−T /2
There is a funny normalizing factor that corrects for the fact that the Fourier
Transforms of the fields have the units of field divided by frequency; since there is only
one integral over frequency left, the δ(f = 0) provides another unit of frequency in the
numerator as needed. Recall, δ(f = 0) = limT →∞ T is present because of the
time-averaging. It will always be canceled by a similar factor in the numerator,
eliminating what appears to be division by ∞. (We will not actually be calculating
power from the Fourier Transforms directly, so this will turn out not to be an issue
below.) More importantly, the above tells us that the time-averaged Poynting vector is
obtained by summing up the contributions from each frequency in a linear way: the
power at different frequencies just adds up, so we can calculate the power for a given
frequency and then do sums to get total power. The cross-terms drop away in the
time average.
A final point to make is that time derivatives become powers of ω for Fourier
transforms. This is seen by differentiating the expression for g (t):
Z ∞ Z ∞ Z ∞
dg d d −i ω t
= df ge(f ) e −i ω t = df ge(f ) e = df (−i ω) ge(f ) e −i ω t
dt dt −∞ −∞ dt −∞
(12.81)
r ) e −i ω t
r , t) = ρ0 (~
ρ(~ ~ r , t) = J~0 (~
J(~ r ) e −i ω t (12.82)
We assume that ρ0 and J~0 are zero outside some volume V near the origin. The
retarded scalar and vector potentials are (Equations 10.71; recall, tr = t − |~ r 0 |/c)
r −~
1 ρ (~
r , tr ) ~ r , t) = µo J~ (~
r , tr )
Z Z
V (~
r , t) = dτ 0 A(~ dτ 0 (12.83)
4 π o V |~
r −~ r 0| 4π V |~
r −~ r 0|
0 ω
e −i ω tr = e −i ω t e i k |~r −~r |
with k= (12.84)
c
Let’s rewrite the retarded potentials using this information. With the assumption of
harmonic time dependence and no sources outside V (ρ = 0, J~ = 0 outside V),
Ampere’s Law tells us that, outside V,
~
∂E
o µo =∇ ~
~ ×B =⇒ ~ = c2 i ∇
E ~
~ ×B (12.85)
∂t ω
~ It becomes
Therefore, we only need to calculate A.
0
~ r , t) = µo e −i ω t e i k |~r −~r |
Z
A(~ dτ 0 J~0 (~
r 0) (12.86)
4π V |~r −~ r 0|
Page 904
Section 12.2 Radiation: General Theory of Radiation
Now that the dependences are clear, let’s figure out what we need to keep. The first
factor is independent of r 0 and multiplies the whole expression, so no issue there. In
the second factor, there are power law dependences on r 0 and r , and there are no
direct power-law dependences elsewhere, so we may look at this factor alone. In the
limit d r , we may keep the first term and discard the second and remaining terms
because they falls off as higher powers of r .
Making these approximations and inserting into our expression for the vector potential:
It may seem strange that we dropped the rb · rb0 r 0 term from the expansion of
|~
r −~r 0 |−1 , the one that yielded the electric dipole and magnetic dipole potentials.
The reason we don’t need it here is because of the exponential: its variation over the
scale of the source distribution (note that we did not make the approximation
k d = 2 π d/λ 1!) prevents the cancellation that occurred for the monopole terms
in the static case and necessitated keeping the rb · rb0 r 0 term.
This expression makes very explicit the physical picture that the harmonic dependence
of the source current drives a spherical outgoing wave: the e i(k r −ω t) /r factor.
From the above, we can calculate the fields, time-averaged Poynting vector, radiation
pattern, and total power radiated (for harmonic time dependence!):
~ =∇
B ~
~ ×A E~ = c2 i ∇ ~ ×B~
ω
1 dP dP
D E D E D E Z
~ =
S R E ~∗ × B
~ ~
= r 2 rb · S hPi = dΩ
2 µo dΩ dΩ
radiation from
arbitrary ∞
~ r , t) = µo e
i (k r −ω t) X (−i k)m
Z
m
source distribution A(~ dτ 0 J~0 (~
r 0 ) r 0 rb · rb0
vector potential 4π r m=0
m! V
d r, d λ
(12.92)
We have dropped the d 2 /λ
r requirement because it is implied by the other two
requirements. This is now the multipole expansion for radiation: successive terms
probe finer and finer structure of the source distribution. We see next how these
various terms give electric dipole, magnetic dipole, and electric quadrupole radiation.
The different multipole terms are not an analogue of the multipole terms we
calculated for the electric scalar and magnetic vector potentials (Equations 3.224 and
5.96): in those cases, we had [(r 0 )m /r m+1 ] Pm (b r · rb0 ) in the integrand, here we have
[(k r 0 )m /r ] (b
r · rb0 )m in the integrand: same units, different formula. In the radiation
case, all the multipole terms fall off as 1/r , but they depend on different moments of
the current distribution. In contrast, for the static potential multipoles, the higher
terms fell off with increasing powers of r (and depended on higher moments of the
charge and current distributions, but calculated in a different manner).
~ · ri J~0 e −i ω t + ri ∂ ρ0 e −i ω t
~ · ri J~0 e −i ω t − ri ∇
J0,i e −i ω t = ∇ ~ · J~0 e −i ω t = ∇
∂t
=∇~ · ri J~0 e −i ω t − i ω ri ρ0 e −i ω t (12.94)
When we do the integral, the first term vanishes: it can be turned into a surface
integral of ri J~ at the boundary of V and, since the sources are contained in V, there
can be no current flowing through the boundary of V. That leaves the second term, so
the vector potential becomes
i (k r −ω t) Z ~0 i (k r −ω t)
~ r , t) = µo e
A(~ dτ 0 (−i ω) ρ0 (~
r 0) ~
r 0 = −i
µo ω p
e (12.95)
4π r V 4π r
where we have used the definition of the dipole moment, Equation 3.227. We see
that, in the static limit (ω → 0), the expression vanishes as we expect.
To calculate the magnetic field, we need to take the curl. The curl will act on both
the 1/r and the phase factor in the exponent. But the action of the curl on 1/r will
yield 1/r 2 , which we can drop if we also assume k r 1 or λ r . So we calculate
~ × p
∇ ~0 e i k r = −~ ~ i k r = −~
p0 × ∇e ~ k r ) = −~
p0 × e i k r ∇(i p0 × e i k r i k rb (12.96)
Therefore
2
~ r , t) = µo ω rb × p
B(~
~0 i (k r −ω t)
e d λr (12.97)
4π cr
where the factor of ω/c came from k. We obtain the electric field from
~ = c 2 (i/ω) ∇
E ~ ×B~ = c 2 (−k/ω) rb × B ~ = −c rb × B,
~ where we used a similar
technique for the calculation of the curl (involving more terms, but again one drops all
terms of order 1/r 2 ). This yields (using µo = 1/(c 2 o ))
~ (~ 1 ω 2 rb × (br ×p
~0 ) i (k r −ω t)
E r , t) = − e d λr (12.98)
4 π o c 2r
The geometry of the fields is as follows. The magnetic field is normal to both the
line-of-sight to the dipole at the origin (b ~0 . So, for a
r ) and to the dipole’s direction p
dipole along zb, B~ oscillates in the φ
b direction. The electric field is normal to the
~ so it oscillates in the θb direction. Both θb and φ
line-of-sight and to B, b are in the plane
normal to rb.
Let’s now generalize these using our discussion of Fourier Transforms. Let’s assume an
arbitrary time dependence of the charge density ρ(~ ~(t).
r , t) and the dipole moment p
The former has a Fourier decomposition (Equation 12.72), which we can use to
Fourier decompose the latter:
Z ∞ Z ∞
ρe(~
r, f ) = r , t) e i ω t
dt ρ(~ ⇐⇒ r , t) =
ρ(~ r , f ) e −i ω t
df ρe(~
−∞ −∞
Z Z Z ∞
=⇒ ~(~
p r , t) = dτ 0 ρ(~
r 0 , t) ~
r0 = dτ 0 ~
r0 r 0 , f ) e −i ω t
df ρe(~ (12.99)
V V −∞
Z ∞ Z
= ˜(f ) e −i ω t
df ~p with ˜(f ) =
~p dτ 0 ρe(~
r 0, f ) ~
r0 (12.100)
−∞ V
˜
~˜ r , f ) e −i ω t = −i µo ω ~p (f ) e i (k r −ω t)
A(~ (12.101)
4π r
where we have factored the e −i ω t dependence out of the vector potential and added
~˜ r , f ). Now, sum up
˜(f ) generates its own A(~
the f argument to indicate that each ~p
over frequency components using the Fourier Transform:
" #
Z ∞ Z ∞ ˜(f )
µo ω ~p
~ r , t) =
A(~ ~˜ r , f ) e −i ω t =
df A(~ df −i e ikr
e −i ω t (12.102)
−∞ −∞ 4π r
Next, recall how time derivatives are related to factors of −i ω in Fourier Transforms,
Equation 12.81, and also recall Equation 12.84, which tells us that
e i (k r −ω t) = e −i ω tr (recall, r r 0 is assumed!). These let us rewrite the above as
∞ ˙
~ r , t) = µo 1 ˜(f ) e −i ω tr = µo ~p (tr )
Z h i
A(~ df −i ω ~p (12.103)
4π r −∞ 4π r
One can do similar Fourier analyses for B~ and E~ . One could instead take the curl and
time derivative of the above expression, using again the r r 0 approximation to
make these derivatives easy to take. The Fourier analysis path seems more physical
because it makes a bit clearer how the rb× factors below arise.
The result is the full generalization of our harmonic time dependence expressions:
electric dipole radiation ¨
~ (~ 1 rb × rb × ~p (tr )
electric field E r , t) = (12.106)
d λr 4 π o c 2r
Note that the dipole moment derivatives are evaluated at tr ! These equations match
Equations 11.54, 11.56, and 11.57 of Griffiths.
Let’s define a coordinate system with ~p¨ ∝ zb and θ the spherical coordinate polar
angle, so then
rb × zb = −φ
b sin θ rb × (b
r × zb) = rb × −φb sin θ = θb sin θ (12.107)
~ no
Thus, for an arbitrary time dependence, we obtain (using dP/dΩ = r 2 rb · S,
complex conjugations or real parts necessary because these formulae do not assume
harmonic time dependence, and S ~∝E ~ ×B ~ ∝ θb × −φ
b = rb):
electric dipole
dP µo p̈ 2 sin2 θ µo p̈ 2
radiation pattern = P= (12.108)
d λr dΩ 16 π 2 c 6π c
Notice the similarity to the Larmor formula! Had we calculated the Larmor formula for
two point charges forming a dipole, going through the fields to see how the factors of
q d add coherently, we would have obtained the above result. (The dipole
approximation automatically incorporates β → 0.) If we assume harmonic time
dependence and time-average (also now taking the necessary complex conjugations
and real parts if we use complex notation), we obtain:
electric dipole
radiation pattern
dP µo p02 ω 4 sin2 θ µo p02 ω 4
for harmonic = hPi = (12.109)
time dependence dΩ 32 π 2 c 12 π c
d λr
i(k r −ω t)
~ r , t) = µo e
Z
A(~ (−i k) dτ 0 J~0 (~
r 0 ) rb · rb0 r 0 (12.110)
4π r V
This term is a factor of order k d = 2 π d/λ smaller than the electric dipole term. We
also calculated this term in the process of deriving the vector potential due to a
magnetic dipole, starting at Equation 5.97. Again, we now have to allow
~ · J~0 = −∂ρ0 /∂t 6= 0. This yields (many steps left out!)
∇
1 h 0 i 1 h i
J~0 (~
r 0 ) rb · rb0 r 0 = r × J~0 × rb + r 0 rb · rb0 J~0 + rb · J~0 rb0
~ (12.111)
2 2
The second two terms involve quantities similar to those we dealt with for the electric
dipole term. We can integrate these terms by parts to obtain expressions involving
~ · J~0 (and therefore bringing in one more power of r 0 to cancel the dimensions of ∇),
∇ ~
which can be evaluated using continuity, ∇ ~ · J~0 = −∂ρ0 /∂t = i ω ρ0 . They therefore
result in terms containing i ω ρ0 and two powers of r 0 . These are electric quadrupole
terms. They are of the same order of magnitude as the terms we will keep, but they
are complicated and so we will drop them for this study.
Section 12.2.5 Magnetic Dipole and Electric Quadrupole Radiation Page 914
Section 12.2 Radiation: General Theory of Radiation
The first term recalls the definition of the magnetic dipole moment (Equation 5.110)
1 0
Z
~0 =
m dτ 0 r × J~0 (~
~ r 0) (12.112)
V 2
~ r , t) = i µo ω rb × m
A(~
~ 0 i(k r −ω t)
e d λr (12.113)
4π cr
This expression is smaller than the analogous expression for electric dipole radiation
(Equation 12.95) by a factor (m0 /c)/p, which we can see is k d as expected:
1 m0 1 d 3J d 1 ωd ρd
= 3 = =kd (12.114)
p c d ρd c ρd c
where J ∝ ω d ρ follows from our evaluation of J~ in the electric dipole radiation case.
The fields are easily derived using the same procedure as for the electric dipole term to
evaluate the curls, yielding
2
~ r , t) = − µo ω rb × (b
B(~
r ×m
~ 0 ) i (k r −ω t)
e d λr (12.115)
4π c 2r
Section 12.2.5 Magnetic Dipole and Electric Quadrupole Radiation Page 915
Section 12.2 Radiation: General Theory of Radiation
~ = −c rb × B
For the electric field, we can use the harmonic relation E ~ (from
~ 2 ~ ~
E = c (i/ω) ∇ × B) to find
~ (~ 1 ω 2 rb × (b
r × (b r ×m
~ 0 )) i (k r −ω t)
E r , t) = e (12.116)
4 π o c 3r
~ ∝ (1/c) rb × E
In the electric dipole case, we had B ~ ∝ rb × (b
r × (b
r ×p~)), but we also
~
derived B ∝ rb × p~. We thus suspect that the quadruple vector product in E ~ for
magnetic dipole radiation reduces to rb × m
~ 0 . Let’s prove that explicitly using the
BAC − CAB rule for an arbitrary vector ~ a:
rb × [b
r × (b
r ×~ r · (b
a)] = rb [b r ×~
a)] − (b
r ×~ r · rb) = −b
a) (b r ×~
a (12.117)
~ (~ 1 ω 2 rb × m~ 0 i (k r −ω t)
E r , t) = − e d λr (12.118)
4 π o c 3r
Griffiths derives the magnetic dipole fields for the special case m ~ 0 ∝ zb. We saw for the
ideal electric dipole that rb × (b
r × zb) = θb sin θ and rb × (b
r × (b
r × zb)) = φb sin θ.
Applying that here yields B ~ ∝ −θb sin θ and E ~ ∝φb sin θ, thus matching Griffiths
Equations 11.36 and 11.37.
Section 12.2.5 Magnetic Dipole and Electric Quadrupole Radiation Page 916
Section 12.2 Radiation: General Theory of Radiation
As we did with the electric dipole radiation field, we can generalize these expressions
using Fourier Transforms, yielding
magnetic dipole radiation
µo rb × rb × m~¨ (tr )
magnetic field ~
B(~
r , t) = (12.120)
d λr 4π c 2r
where again all the dipole moments and derivatives thereof are evaluated at the
retarded time. The vector potential is perpendicular to both the vector rate of change
of the dipole and the position vector, the electric field is perpendicular to the second
derivative of the dipole and the position vector, and the magnetic field is
perpendicular to the second derivative of the dipole and the electric field. There is no
analogue of these generic results in Griffiths, who only considers the special case of
~ ∝ zb and harmonic time dependence, as we derived above. For the special case of a
m
dipole m ~˙ ∝ zb and m
~ ∝ zb that also has m ~¨ ∝ zb, the vector potential and electric field
are along φb and the magnetic field is along θ.
b Note how A ~ and E
~ have the same
direction as the infinitesimal current.
Section 12.2.5 Magnetic Dipole and Electric Quadrupole Radiation Page 917
Section 12.2 Radiation: General Theory of Radiation
To determine the power pattern, we can use the calculations we did for the electric
dipole case because the dependences are similar: up to signs and normalization, B~ for
~ for the magnetic case and E
the electric case matches E ~ for the electric case matches
~ for the magnetic case. In both cases, E
B ~ ×B ~ ∝ +b
r as is necessary for outgoing
radiation, so we can dispense with the signs. The normalization is set by the
replacement p → m/c. So, we have (again, dP/dΩ = r 2 rb · S,~ all fields real)
magnetic dipole
dP µo m̈2 sin2 θ µo m̈2
radiation pattern = P= (12.122)
d λr dΩ 16 π 2 c3 6 π c3
magnetic dipole
radiation pattern
dP µo m02 ω 4 sin2 θ µo m02 ω 4
for harmonic = hPi = (12.123)
time dependence dΩ 32 π 2 c3 12 π c 3
d λr
Given the replacement p → m/c, the magnetic dipole radiation power is down by a
factor of [(m/c)/p]2 = (k d)2 = (2 π d/λ)2 relative to electric dipole radiation.
Section 12.2.5 Magnetic Dipole and Electric Quadrupole Radiation Page 918
Section 12.2 Radiation: General Theory of Radiation
Postscript
Griffiths is unfortunately lacking in examples on the topic of radiation. The reason for
this is that, with the above formalism in hand, the basic examples consist largely of
calculating electric and magnetic dipole moments and then plugging into the above
formulae. Rather than do a lot of this plug-and-chug, it makes sense to move on to
classical scattering theory and antennas, which provide more meaty and meaningful
applications of the above. We will do this in Ph106c.
If it seems like the above discussion was a bit tortured — the term-by-term
manipulation of Equation 12.92, the by-hand splitting of the m = 1 term into
magnetic dipole and electric quadrupole radiation, the reappearance of the same
electric and magnetic field dependences — that impression is correct. There is a more
unified way of treating the multipole expansion, which is given in Jackson Chapter 16,
but which we do not have time to present.
Page 920
Lecture 52:
Applications of Radiation I:
Classical Scattering
Antennas
Date Revised: 2022/05/25 10:00
Date Given: 2022/05/25
Page 921
Section 13.1 Applications of Radiation: Classical Scattering Theory
~ (~
E ~0 e i (~k·~r −ω t )
r , t) = E (13.1)
~ = α(ω) E
~ = α(ω) E
p ~0 e −i ω t (13.2)
(Recall we derived models for α(ω) in Section 5.) This is an oscillating, accelerating
~, ~p˙ , and ~p
electric dipole with p ¨ in the same direction as the incoming polarization
vector. We can calculate its radiated power pattern using Equation 12.109, which
adds the assumption d λ r (scatterer much smaller than wavelength, which is
smaller than the distance to field point):
dP
~0 |2 ω 4 sin2 θ
µo α2 |E ~0 |2 ω 4
µo α2 |E
= 2
hPi = (13.3)
dΩ 32 π c 12 π c
where θ is the polar angle relative to the polarization vector of the incoming wave.
Note that the emitted power pattern has no dependence on φ, the azimuthal angle
around the polarization vector, implying the scattered light does not care about the
incoming wave direction, only its polarization.
Section 13.1.2 Scattering to Large Distances of Polarized Light by a Small Object, d λ r Page 923
Section 13.1 Applications of Radiation: Classical Scattering Theory
linearly polarized 2 2
dσpol ω4 8π ω4
α α
cross section = sin2 θ σpol = (13.5)
d λr dΩ 4 π o c4 3 4 π o c4
Because α/4 π o carries units of volume, σ carries units of area and dσ/dΩ units of
area/solid angle. The reason for this is that we assume an incoming plane wave of
infinite transverse extent, so we cannot calculate the fraction of its total power
scattered: that would vanish because the power is infinite. Rather, we calculate the
power scattered given an incoming power per unit area, so we have to multiply by an
area to get the right output units.
Section 13.1.2 Scattering to Large Distances of Polarized Light by a Small Object, d λ r Page 924
Section 13.1 Applications of Radiation: Classical Scattering Theory
Section 13.1.3 Scattering to Large Distances of Unpolarized Light by a Small Object, d λ r Page 925
Section 13.1 Applications of Radiation: Classical Scattering Theory
For the polarization component perpendicular to the scattering plane, the scattering
plane makes an angle π/2 with the polarization vector. The angle θsc is the φ angle in
a coordinate system whose z-axis is this polarization vector. Thus, for any θsc , the
angle θ in the differential cross section formula is θ⊥ = π/2 in the diagram. Assuming
the incoming light is unpolarized implies half of the incoming power is in this
polarization component, yielding
2 4
dσ⊥ 1 dσ 1
α ω
= = (13.6)
dΩ 2 dΩ θ=π/2 2 4 π o c4
For the polarization component parallel to the scattering plane, the scattering plane
lies in the xz-plane of the coordinate system whose z-axis is this polarization vector
and whose ±x-axis is the incident direction k.
b Thus, the polar angle in the scattering
coordinate system and the polar angle in the incident polarization coordinate system
are complementary, so sin θ|| = cos θsc and therefore
2 4
dσ||
1 dσ 1
α ω
= = cos2 θsc (13.7)
dΩ 2 dΩ θ=π/2−θsc
2 4 π o c4
We want to combine these cross sections to get the total cross section. Do we need to
worry about constructive or destructive addition of the scattered waves in the two
polarizations? No, a fact we can see both conceptually and mechanically.
Section 13.1.3 Scattering to Large Distances of Unpolarized Light by a Small Object, d λ r Page 926
Section 13.1 Applications of Radiation: Classical Scattering Theory
We can see this conceptually as follows. We first recognize that the two incoming
polarizations excite dipoles parallel to the incoming field directions and therefore
perpendicular to each other:
¨ ∝E
~p ~in,|| ¨⊥ ∝ E
~p ~in,⊥ ~in,|| ⊥ E
E ~in,⊥ =⇒ ~p ¨⊥
¨ ⊥ ~p (13.8)
|| ||
Then, we recall that the electric field of the radiated wave is in the direction of the
projection of ~p¨ transverse to the line-of-sight direction (b
rsc ). Therefore, the radiated
electric fields for the two incoming polarizations must be perpendicular to each other
(and to the line of sight). We know that fields in orthogonal polarizations do not
interfere with each other and their powers just add, hence we just add the cross
sections, accounting for the fact that half the power is in each polarization.
Section 13.1.3 Scattering to Large Distances of Unpolarized Light by a Small Object, d λ r Page 927
Section 13.1 Applications of Radiation: Classical Scattering Theory
We can see this mechanically by calculating the Poynting vector of the summed wave:
∗ ∗
~tot = E
µo S ~rad,|| + E
~rad,⊥ × B ~ rad,|| + B
~ rad,⊥ (13.9)
The fields of the scattered (radiated) wave satisfy (Equations 12.105 and 12.106)
E ¨ ∝ rb rb · E
~rad ∝ rb × rb × ~p ~in − E
~in ~ rad ∝ −b
B ¨ ∝ rb × E
r × ~p ~in (13.10)
Therefore, the cross terms in the Poynting vector are (applying the BAC − CAB rule
many times):
h i h i
~∗ ~ ~ ~ ~
E rad,⊥ × Brad,|| ∝ rb rb · Ein,|| − Ein,|| × rb × Ein,⊥ (13.11)
h i
~in,|| rb × rb × E
= rb · E ~in,⊥ − E ~in,|| × rb × E ~in,⊥ (13.12)
h i
~in,|| rb rb · E
= rb · E ~in,⊥ − E ~in,⊥ (13.13)
h i
− rb E ~in,|| · E
~in,⊥ − E ~in,⊥ rb · E ~in,|| (13.14)
The third term vanishes because the polarizations are perpendicular to one another.
The second term cancels the fourth term. The first term vanishes because, given the
~in,⊥ . So the whole cross term vanishes.
geometry, rb is in the plane perpendicular to E
The other cross term is obtained by exchanging || and ⊥ and so it also vanishes.
Section 13.1.3 Scattering to Large Distances of Unpolarized Light by a Small Object, d λ r Page 928
Section 13.1 Applications of Radiation: Classical Scattering Theory
Since the power radiated in the two polarizations just sum, we may add the two cross
sections together incoherently. We also integrate over all angles to get the total
scattering cross section:
unpolarized 2 2
dσunpol ω 4 1 + cos2 θsc 8π ω4
cross α α
= σunpol = (13.15)
section dΩ 4 π o c4 2 3 4 π o c4
d λr
The total scattering cross section is the same as the polarized case because the
polarized cross section has no dependence on polarization angle.
Section 13.1.3 Scattering to Large Distances of Unpolarized Light by a Small Object, d λ r Page 929
Section 13.1 Applications of Radiation: Classical Scattering Theory
Recall our discussion of the permittivity and index of refraction of dispersive materials
and how it depends on frequency. In particular, it is fairly independent of frequency in
regions well away from any known bound states. For the atmosphere, the relevant
states are either vibrational and rotational states of molecules or electronic states of
the constituent atoms. The atmosphere is dominated by N2 and O2 , which are
symmetric molecules with no net dipole moment; hence, in our approximation, there is
no coupling of light to their rotational and vibrational states. The electronic states of
N, O, and the third major component, Ar, are above optical frequencies. Therefore,
the permittivity becomes constant over the optical part of the spectrum. We may infer
from this that the polarizability α becomes constant. (Given the above, it actually
vanishes for N2 and O2 and the scattering is entirely by the individual atoms, not
molecules.)
Thus, the frequency dependence of the scattering cross section comes entirely from
the ω 4 term. This strong dependence on frequency implies that the blue portion of the
solar spectrum scatters much more than the red portion: seven times comparing
400 nm to 650 nm.
Section 13.1.4 Application: Rayleigh Scattering, or Why the Sky is Blue Page 930
Section 13.1 Applications of Radiation: Classical Scattering Theory
I When we look away from the sun, the light we are seeing is primarily scattered
light, which is dominantly blue.
I When we look at the sun at sunrise or sunset, when the path length through the
atmosphere is large, the blue light has been scattered out of the line-of-sight
and we see a red sunrise or sunset.
Section 13.1.4 Application: Rayleigh Scattering, or Why the Sky is Blue Page 931
Section 13.1 Applications of Radiation: Classical Scattering Theory
A simple situation in which to apply our theory of scattering quantitatively is the case
of Thomson scattering, which is scattering of an electromagnetic wave off of a free
electron. Recall that we calculated the relation between a bound, damped electron’s
dipole moment and the incident electric field to be (Equation 9.190)
q 2 /m
pe(t) = e xe(t) = Ee0 e −i ω t (13.16)
ω02 − ω2 − i γ ω
√
If we take the limit ω ω0 and ω γ ω (the plasma limit we defined in Section 5),
we obtain
pefree (t) e2
αfree = =− (13.17)
Ee0 e −i ω t m ω2
free-electron 2 2
dσThomson e2 1 + cos2 θsc 8π e2
(Thomson) = σThomson =
cross section dΩ 4 π o me c 2 2 3 4 π o me c 2
(13.18)
The total cross section is called the Thomson cross section and is ubiquitous because
it sets the scale of scattering of EM waves off of electrons, even when quantum
mechanical effects are taken into account. Notice that it is independent of frequency
because the ω −2 dependence of α cancels the ω 4 dependence of dipole radiation.
Though, recall we have assumed (p/e) ∼ d λ r . This assumptions fails at high
enough frequency.
e2
rclassical = (13.19)
4 π o me c 2
because it is the radius one obtains by equating the rest mass energy of the electron,
me c 2 , to the Coulomb energy required to assemble of sphere of uniform charge
density of total charge e and radius rclassical . One the one hand, it is remarkable that
this radius gives σThomson up to a factor of 8/3; on the other hand, it is the only
length scale available, so perhaps we should not be surprised.
Section 13.1.5 Application: Thomson Scattering Page 933
Section 13.1 Applications of Radiation: Classical Scattering Theory
− o ~
~ = 4 π o d 3
p E0 (13.20)
+ 2 o
where (k d)4 comes from (ω/c)4 d 4 . This is also termed “Rayleigh scattering”
because of the similar frequency dependence.
where n is the density of atoms or molecules and σ is the scattering cross section per
atom or molecule. To find σ, we need the polarizability, α. Recalling the definition of
the dielectric constant r , we have
~ = np ~ ~ = ( − o ) E
~ = χe o E ~ − o
P ~ = nαE and P =⇒ α= (13.24)
n
Inserting this into our expression for σunpol , Equation 13.15, we obtain
2 2
8π − o ω4 8 π3 − o 1
γdilute = n σ = n 4
= (13.25)
3 4 π o n c 3 o n λ4
γdense = (# of cells per unit volume) × (cross section per cell) (13.26)
1 8 π α δNd 2 ω 4 8 π 3 − o 2 1 δNd2
= 3 = (13.27)
d 3 4 π o c4 3 o n λ4 n d 3
In an ideal gas (not what we are dealing with), we should make the cell size match the
volume per atom because the correlation length vanishes for an ideal gas. Thus, the
cell volume d 3 is 1/n and the rms fluctuation per cell is 1 because that is the rms for
a Poisson distribution with mean value of 1. With these values, we recover the prior
dilute expression.
where βT is the isothermal compressibility (basically, how hard it is stuff more atoms
or molecules into a given volume). With this, we have
2
8 π3 − o 1
γ= kB T βT (13.29)
3 o n λ4
This kind of Rayleigh scattering is quite important for light propagation in liquids;
scattering, rather than absorption, is the primary reason for light attenuation in liquids.
One perhaps surprising overlap of particle physics and thermodynamics is the fact
that, in almost any detector that involves the propagation of Cerenkov or scintillation
light in a liquid — e.g., water Cerenkov and liquid scintillator detectors of neutrinos
such as in the SuperKamiokande, Sudbury Neutrino Observatory, LSND, MiniBooNE,
and Daya Bay experiments (among many others), and liquid noble detectors based on
scintillation light used for these purposes as well as dark matter searches and coherent
neutrino-nucleus scattering measurements — Rayleigh scattering is the dominant
mechanism for attenuation and must be modeled well to accurately simulate the
detector behavior.
Also, interestingly, βT can diverge at the liquid-gas critical point, giving rise to
extremely large scattering called critical opalescence, which provides a useful way of
measuring the critical point of a particular material.
Section 13.1.8 Application: Scattering in a Dense Medium Page 937
Section 13.2 Applications of Radiation: Antennas
Antennas
Introduction and Study Guide
Antennas are our second major application of the theory of radiation we have
developed. This material is not covered in Griffiths, it is largely from Heald and
Marion §9.4, §9.5, and §9.7.
We see the first hint of a concept of an antenna’s radiation resistance, which gives the
relation between mean-square current and the radiated power. We will see that, for an
antenna radiating into
free space, the maximal efficiency for radiated power is
obtained when hPi/ I2 = Z0 . This short line antenna does not do a good job of
radiating into free space because of the prefactor (kd)2 /6π 1. We immediately see
that, to match free-space well, an antenna must have a size of order the wavelength it
is radiating. The approximations made for the multipole expansion and electric and
magnetic dipole radiation fail and we must go back to the more general expressions.
Page 940
Section 13.2 Applications of Radiation: Antennas
i (k r −ω t)
~ r , t) = µo e
Z
0
r0
A(~ dτ 0 J~0 (~
r 0 ) e −i k r r ·b
b
4π r V
We did not derive expressions for the fields in this general case, so we do so here. We
recall that, with the above assumed harmonic time dependence, the application of the
curl to get the magnetic and electric fields yields i k rb × factors from the argument of
the exponential (neglecting the terms of higher order in r that come from
differentiating the 1/r dependence). Therefore,
antenna i (k r −ω t)
~ r , t) = i µo ω e
Z
0
r0
magnetic B(~ rb × dτ 0 J~0 (~
r 0 ) e −i k r r ·b
b
(13.32)
field 4π c r V
antenna
~ (~ i c2 ~ ~ = −c rb × B
~
electric E r , t) = ∇×B (13.33)
field ω
Antenna Gain
An antenna is fundamentally a device for either radiating power or receiving power
with a desired pattern. They are frequently intended to radiate to or receive from a
particular direction. This can be quantified using the idea of antenna gain:
D E
dP
dΩ
(b
r)
antenna gain G (b
r) = (13.34)
hP/4 πi
which is just the ratio of the power per unit solid angle into a particular direction
divided by the total power per unit solid angle into all directions. The antenna
radiates more power in directions for which this number is large. If an antenna is fully
omnidirectional, then G (br ) = 1 in all directions. We shall see that this same number
also quantifies how directional an antenna is in reception mode.
G (b
r ) [dB] = 10 log10 G (b
r) (13.35)
Thus, an omnidirectional antenna has 0 dB gain in all directions while other antennas
have positive dB gain in some directions (the directions they are good at transmitting
to) and negative dB gain in other directions.
Antenna Impedance
Since the antenna may not be small compared to λ, the mean-square current I2 may
not be the same everywhere in the antenna. For specificity, we choose to evaluate I2
at the point where the antenna is fed by wires or a waveguide, the so-called “gap”
drive point. We then define the antenna impedance and radiation resistance:
For an ideal antenna, meaning one that radiates all the power incident it receives from
a transmission line, Zant = Rrad (we use R instead of Z for the radiation component
because it is always real by definition). For non-ideal antennas, there may be
additional resistive or reactive components.
Of particular importance is that Zant determines the relation between power flowing in
on a transmission line and power received by the antenna (which can then be
radiated). If Zant = Zline , then the antenna receives all power from the transmission
line and reflects none. If this is not true, as may be the case because transmission
lines have a set of standard impedances, then a transformer can be used to impedance
match such a transmission line to an antenna. We discussed earlier how appropriate
lengths of transmission line could be used as transformers. These can work quite well
for antennas intended for single-frequency use. Broadband impedance matching —
matching over a large fractional bandwidth ∆ν/ν — is more difficult.
d
I(z) = I0 sin k − |z| (13.37)
2
This form is sinusoidal as desired and also satisfies the boundary condition that the
current vanish at the end. The derivative is not continuous at z = 0 except for certain
values of d. The current at the gap drive point is Ig ,0 = I0 sin k2d . Given the current
distribution, we may calculate the magnetic field, noting that rb × d` zb = −φ b sin θ:
i(k r −ω t) d/2
b µo ω I0 e d
Z
0
~ r , t) = −i φ
B(~ sin θ dz 0 e −i k z cos θ
sin k − |z 0 |
4π c r −d/2 2
(13.38)
Let us rewrite this as the product of a function of λ and a function of d/λ and θ:
center-driven i(k r −ω t)
b µo ω I0 λ e d
line antenna ~ r , t) = −i φ
B(~ f ,θ (13.39)
magnetic field 4π c 2 r λ
center-driven
d
sin θ
Z π d/λ
πd
line antenna
f ,θ = dζ e −i ζ cos θ sin − |ζ| (13.40)
magnetic field λ π −π d/λ λ
pattern
Comparing with our calculation of the radiation field due to a line current electric
dipole, we see that if we make the correspondence,
d
λ
I0 d ↔ I0 sin θ ↔ f ,θ (13.41)
2 λ
and note that the relation between E ~ and B~ is the same as for the line current electric
dipole, then the expressions calculated for the line current electric dipole can be
translated to this case, giving
λ2
µo ω 2 I20 4
2 2
dP d = Z0 I2 f d , θ
= 2
f , θ 0 (13.42)
dΩ 32 π c λ 32 λ
2
antenna d
,θ 2
1 d
f λ D E Z
|f |2 =
gain G (θ) = D E with dΩ f , θ (13.43)
pattern |f |2 4π λ
D E
Z0
antenna
hPi 32
I20 4 π |f |2 π 1 D E
radiation Rrad =
2 = 1 2
= |f |2 Z0 (13.44)
resistance Ig I
2 g ,0
4 sin2 k2d
For the sinusoidal current we have assumed, we can evaluate the field pattern for
arbitrary d. This is done by using the trigonometric product identities
1
cos A sin B = [sin (A + B) − sin (A − B)] (13.45)
2
1
sin A sin B = [cos (A − B) − cos (A + B)] (13.46)
2
0
and also using the Euler formula for e −i k z cos θ . This yields
center-driven
kd
kd
line antenna
d
2 cos 2
cos θ − cos 2
f ,θ = (13.47)
radiation λ π sin θ
field pattern
We can calculate these quantities for some specific antenna lengths. A plot comparing
the radiation patterns of our idealized electric dipole and half-wave and full-wave
antennas is given at the end.
Ig ,0 = I0 Rrad ≈ 73 Ω (13.51)
I full-wave antenna, d = λ
The current is
The current at the drive point formally vanishes and the antenna’s radiation
resistance is formally infinite: basically, no matter what voltage the transmission
line has on it, it cannot drive a current at the antenna drive point. Clearly, not a
very good antenna to drive with a transmission line! The corrections we ignored
in making the assumption of perfectly sinusoidal current will enable Ig 6= 0.
However, we can still calculate the radiation pattern using our formula for f
with k d/2 = π, giving
π
4 cos2 2 cos θ
d 2 cos (π cos θ) − cos π
f = 1, θ = = (13.53)
λ π sin θ π sin θ
D E
|f |2 = 0.672 (13.54)
The full-wave antenna has a more peaked radiation pattern than the half-wave
antenna, though it is harder to drive.
This higher radiation resistance is a better match to free space, which has
Z0 = 377 Ω as noted before. Flat-pair or twisted-pair transmission line, which
have impedances of about 300 Ω, are best for feeding this antenna.
I Arbitrary Length
Given on the following slide is a plot of the antenna impedance
(Equation 13.36) as a function of length in units of λ. The length is marked
with tickmarks on the trajectory. The impedance is a complex number whose
resistive (real) part is the horizontal axis and whose reactive (imaginary) part is
the vertical axis. The real part is dominated by the radiation resistance
(Equation 13.36), which is always real by definition. The reactive component
simply indicates that the current and voltage at the drive point (which are
supplied by the feeding transmission line) are out of phase, with the sign of the
phase defining whether the antenna has a capacitive or inductive reactance.
The right side of the trajectory, at d/λ = 1, would go off to infinity for an ideal
full-wave antenna; nonidealities keep it finite. At d/λ = 3/2, the antenna looks
similar to a half-wave antenna but with a somewhat higher impedance.
One can use a plot like this to pick the transmission line that best impedance
matches the antenna one wants to use, the choice of which is usually driven by
the radiation pattern or the available space. If there is an impedance mismatch,
one might use a transmission line transformer to transform the impedance or
lumped element capacitors or inductors to “tune out” (cancel) the reactive part
of the antenna impedance.
0.6
0.4
0.2
2 1 1 2
0.2
0.4
0.6
Polar plot of gain patterns (power units) of ideal electric dipole (black), half-wave
antenna (red), and full-wave antenna (blue). The greater gain of the full-wave
antenna is due to its cos4 dependence (vs. cos2 for the half-wave antenna). Figure
courtesy of M. Cross.
λ2 b × incident flux
Pr (k)
b = G (−k) (13.57)
4π
The quantity (λ2 /4 π) G (−k) b is the effective area of the antenna in the direction kb
(sort
R of like the cross section for scattering we discussed earlier.) By its definition,
dΩ G (br ) = 4 π, so the above also tells us that the total effective area of the antenna
integrated over all possible incoming angles is λ2 . This is an important theorem in
antenna theory because it provides a normalizing factor for measuring antenna
efficiency: if you surround an antenna with a blackbody radiator, you know the
incident flux. The antenna should receive a power λ2 times the flux incident from the
blackbody. You can then measure the power exiting the antenna onto a transmission
line and take the ratio of observed to expected power to determine the antenna’s
overall efficiency. This could also be used, for example, to determine the antenna’s
radiation resistance.
The proof of this theorem is interesting because it involves the Lorentz Reciprocity
Theorem and can be found in Drabowich et al., Modern Antennas, §4.1.
Suppose one has an array of N antennas, each displaced from the origin by ∆ ~ j, j = 1
~0 be the radiation field pattern of
to N, all fed identically with current, in phase. Let E
an individual antenna situated at the origin. Then the radiation pattern of the array is
easily calculated by recognizing that each antenna’s contribution to the total electric
0 0
field is the same up to the phase offsets due to the e −i k r br ·br factor arising from the
displacement in ~ 0
r between antennas:
N
~=
X i c2 ~ h
~ ×A~j (~
i
E ∇× ∇ r , t) (13.58)
j=1
ω
N
" #
i c2 ~ µo e i(k r −ω t)
Z
~ ~ r0
X
0 0 −i k br ·~
= ∇× ∇× d ` I(~ r )e (13.59)
j=1
ω 4π r Cj
N
" #
i c2 ~ ~ × µo e
i(k r −ω t) Z
~ 0
d~ r 0 ) e −i k br ·(∆j +~r )
X
= ∇× ∇ ` 0 I(~ (13.60)
j=1
ω 4π r C0
N
~0
X ~
=E e −i k br ·∆j (13.61)
j=1
~ ~0 in what remains.
where we factored e −i k br ·∆j out of the integral and recognized E
You hopefully recognize the sum over phase factors of this type as the same kind of
sum present in the calculation of the interference pattern due to an array of slits in a
screen. Let us consider an analogue of that, an array of N center-fed antennas oriented
with their currents flowing along the z-axis and displaced from the origin with uniform
spacing ∆. Let the first antenna be at ~r0 with r0 r . The radiation field pattern is
N N
~ ~
~ =E
~0 ~0 e −i k br ·(~r0 +(j−1) ∆)
X X
E e −i k br ·∆j = E (13.62)
j=1 j=1
N ~
~0 e −i k br ·~r0
X ~ ~0 e −i k br ·~r0 1 − e −i k N br ·∆
=E e −i k br ·∆ (j−1) = E ~
(13.63)
j=1 1− e −i k br ·∆
PN
where we have used the standard result r j−1 = (1 − r N )/(1 − r ) for geometric
j=1
series. Thus, if f0 (d/λ, θ) is the radiation pattern corresponding to E ~0 , we have
2 ~ 2
2
−i k N b r ·∆
f d , θ = f0 d , θ 1 − e
(13.64)
λ λ ~
1 − e −i k br ·∆
driven-array ~
sin2 N k2br ·∆
2 2
f d , θ = f0 d , θ
antenna ~
(13.65)
radiation pattern
λ λ
sin2 k br2·∆
The factor at the end is the same factor one finds in the calculation of the interference
pattern from an array of slits. It has the following properties:
These results for the peak height and width are obtained by Taylor expanding the sines
~ = 2 n π to first and third order, while the secondary maxima are found by
near k rb · ∆
~ = m π/N 6= 2 n π.
requiring k rb · ∆
Forward
direction
I There is one driven antenna, which is a folded half-wave antenna (for reasons
discussed earlier).
I There is one antenna placed behind the the drive antenna that is intended as a
back-reflector: it is spaced λ/4 behind the driven antenna so that the wave that
reflects from it is in phase in the forward direction (remember the extra π phase
shift due to a reflection from a conductor). This is the equivalent of saying that
the wave that is radiated by the back-reflector, which is excited by the driven
antenna, is in-phase with the driven antenna’s wave in the forward direction and
out of phase in the backward direction.
I There are multiple “director” antennas placed ahead of the driven antenna.
They are intended to constructively interfere with the driven antenna in the
forward direction and destructively in the backward direction. If there were just
one, it would be placed λ/2 in front of the driven antenna (so the reflected
wave in the reverse direction is out of phase). With multiple directors, the
optimal spacing turns out to be about λ/3.
I A more sophisticated discussion implies that the lengths of the passive antennas
are also important. The spacing argument is an oversimplification because it
assumes the back-reflector acts like an infinite conducting plane. Really, we
should think about the antenna impedance and ask what needs to be done to
~ phase factor that arises from the spacing between the
cancel the −i k rb · ∆
driven antenna and the back-reflector or directors. This issue did not arise for
the phased array because, using identical antennas, there are no path length
differences for the direction of maximum constructive interference, perpendicular
to the spacing ∆ ~ between elements. But here we want constructive interference
in the direction in which there are path length differences, and so we need to
cancel out the corresponding phase factors using the antenna impedances.
Recall the figure showing the antenna impedance as a function of size: antennas
that have d < λ/2 are capacitive and those with λ/2 < d < λ are inductive. In
~ is negative imaginary. With our
the forward direction, the phase factor −i k rb · ∆
−i ω t time dependence, a capacitive element has a positive imaginary
component to its impedance (positive phase factor), so the directors should be
capacitive to cancel this negative imaginary phase and yield constructive
interference in the forward direction (and, conversely, destructive interference in
the reverse direction). So, the directors should have dd < λ/2. Conversely, the
back-reflector should be inductive and thus have λ/2 < dr < d. Usually, dd and
dr are not much different from d.