() :

Classical Electrodynamics (I)

Department of Physics, National Taiwan University
Kwo Ray Chu, Rm. 710, 3366-5113, krchu@yahoo.com.tw
Fall Semester, 2017

1.Textbook and Contents of the Course:

Textbook: J. D. Jackson, Classical Electrodynamics, 3rd
edition. Other books will be referenced in the lecture notes
when needed.
In this course [Classical Electrodynamics , we will cover
the first 7 chapters of the textbook, which elegantly lay down
the foundation of electrodynamics with a systematic exposition
of some key topics in applied mathematics.
Next semester, in Classical Electrodynamics we will
cover selected topics of deep physical insight as well as current
interest. Selected topics (including microwaves, radiating
systems, scattering, diffraction, special theory of relativity, and
synchrotron radiation will help students consolidate the
fundamental principles in Classical Electrodynamics .
2
2. Conduct of Class : Lecture notes will be projected sequentially on
the screen during the class. Physical concepts will be emphasized,
while algebraic details (provided in the lecture notes) will often be
skipped. Questions are encouraged. It is assumed that students
have at least gone through the algebra in the lecture notes before
attending classes (important!).
3. Homework: Approximately 3 problems per chapter will be
assigned for homework. Students are encouraged to do as many
unassigned problems as time allows. Find the problems that appeal
to you. Always do them by yourself. If unsure or unsuccessful,
solutions can be found by, for example, a Google search under
Jackson electrodynamics problem solutions.
4. Teaching Assistants:
alex968.tw@gmail.com, 0921-003-365
d05222007@ntu.edu.tw, 0928-450-365
r05222071@ntu.edu.tw, 0978-007-361
3

5. Grading Policy: Midterm (45%); Final (45); Attendance (5%);

Homework (5%). The overall score will be normalized to reflect
an average consistent with other courses.
6. Lecture Notes:
Starting from basic equations, the lecture notes follow Jackson
closely with algebraic details filled in.
Equations numbered in the format of (1.1), (1.2) refer to
Jackson. Supplementary equations derived in lecture notes, which
will later be referenced, are numbered (1), (2)... [restarting from (1)
in each chapter.] Equations in Appendices A, Bof each chapter
are numbered (A.1), (A.2)and (B.1), (B.2)
Page numbers cited in the text (e.g. p. 120) refer to Jackson.
Section numbers (e.g. Sec. 1.1) refer to Jackson. Main topics
within each section are highlighted by boldfaced characters. Some
words are typed in italicized characters for attention. Technical
terms which are introduced for the first time are underlined. 4
 Chapter 1: Introduction to Electrostatics
1.1 Coulombs Law
Coulomb's law, discovered experimentally in 1785, is a fundamental
law governing all electrostatic phenomena. It states :
1. The force F exerted on point charge q by point charge q1 obeys
F q, q1, and 12 . q er
q1 r F
r
on q
qq1 F is along er (central force).
F e
2 r
4 0 r F is attractive if q and q1 have opposite signs.
F is repulsive if q and q1 have the same sign.
2. For multiple charges, the total force on q is the vector sum of the
2-body Coulomb forces between q and each of its surrounding charges.
Note: The exponent "2" in r 2 has later been determined by precision
measuement to be accurate up to at least 16 decimal points.
Questions: (1) What is the principle of linear superposition?
(2) What force holds our body from falling apart? 5

1.2 Electric Field

The electric field* at point x due to one or more charges is defined
as E( x ) F , *The word "field" was first (1.1)
q used by Faraday in 1845.
where q is a "test" charge at x and F is the total Coulomb force on q.
q q x x1
E at x due to q1 at x1 is E(x) 1 12 e r 1 (1.3)
4 0 r 4 0 |x x1|3
re r = x x1 ( x x1 )e x ( y y1 )e y ( z z1 )e z x x1 er
q1 r E(x)
2 2 2
r | x x1 | ( x x1 ) ( y y1 ) ( z z1 ) x
x1
Question: Why write "rer" as "x x1"? 0
Note: In Jackson (1.2) & (1.3), k 1 in SI units (as used in Jackson).
For a continuous charge distribution, let ( x) be the macroscopic
volume density (charge/unit volume).
( x x) x x E(x)
Then, E(x) due to (x) is x x
(x)(x x) 3 principle of linear x
E( x ) 1 d x x (1.5)
4 0 |x x|3 superposition
0 6
 1.2 Electric Field (continued)
Questions:
q1 x x1
(1) E(x) as x x1. Is this physical?
4 0 |x x1|3
Answer: All charged particles have a finite size (even the electron
has a non-zero, but unknown, size). Hence, they do not produce an
infinite electric field. The particle size is so small that we are always
far away (relative to particle size) from the particle in electrodynamics
problems. Thus, the particle can be treated as a point charge.
(2) Does a point charge experience a force in its own electric field?
Answer: Imagine the charge is distributed in a particle of radius R.
The Coulomb forces within the particle are internal forces, with a zero
vector sum by Newton's third law. This is true even as R 0. Thus, a
point charge experiences no net force in its own electrostatic field.
(3) Any theory on what holds the electron together?
This is out of the scope of this course. See Feynman Lectures II,
Sec. 28-4 if you are interested.
7

1.3 Gausss Law

Consider a point charge q inside or outside of a closed surface S .
da : a differential surface area on S S n
d : solid angle from q to da q
E e
n : unit vector normal to da 40r2 r
d da
Let
and pointing out of S
q r rer
e r : unit vector along r
: angle between n and e r 2d cos
r

e nda
cos da = 4 d d r 2
q q q
E n da e r nda
4 0r 2 4 0r 2
0 r
d 0 d 0
n q outside S:
d (leaving S ) 0 q inside S: n
d (entering S ) 0 d 4 q d 0 n d 0
n
S q S
q , q inside S Gauss's law for
s E n da 4 d 0
q
0 a single charge (1.9)
0, q outside S 8
 1.3 Gausss Law (continued)
By the principle of linear superposition, Gauss's law for a discrete
set of charges is
s E n da 10 qi [all qi inside S ] v n (1.10)
i
and Gauss's law for a distribution of charge is (x)
3
s E n da 10 v (x)d x [ inside S ] S (1.11)
Discussion: n
1. (1.11) is the integral form of Gauss's law. As will be shown in
2 examples below, it is a powerful mathematical expression derived
from Coulomb's law: E(x)
( x )
1 ( x)( xx) 3
E(x) 4 3 d x x
x (1.5)
0 |xx|
0
2. (1.5) applies to infinite space. (1.11) applies to any volume v.
A note on notation: (x) is used when x is the only variable [as in
(1.11)]. (x) is used to dintinguish 2 variables, x and x [as in (1.5)].
9

1.3 Gausss Law (continued)

Example 1: Find E on the surface of an infinite, plane conductor
with a uniform surface charge density (charge per unit area).
It will be complicated if we use Coulomb's law (see prob. 2, p. 36
of lecture notes, but trivial if we use Gauss's law.
A
Draw a pillbox of infinitesimal thickness Conductor E
(dashed line), which encloses an area A of the (zero field)
conductor surface. E 0 inside the conductor.
By symmetry, E on the right of must be to . Applying Gauss's
law to the pillbox: s E n da EA A E
q
0 0 0
(a) the conductor surface is curved , or

Question: If (b) is non-uniform, or ,
(c) there are charges away from the conductor
E at any point x on the conductor surface is still given by (x) / 0
pointing normal to (x). Why? (See prob. 2, lecture notes, p. 36.)
Note: This is true only for E on (not away from) the surface.
10
 1.3 Gausss Law (continued)

Example 2: Shell theorems another application of Gausss law

Halliday, Resnick, and Walker, Fundamentals of Physics:
The two shell theorems that we found so useful in our study of
gravitation hold equally well in electrostatics:
Theorem 1: A uniform spherical shell of charge behaves, for
external points, as if all its charge were concentrated at its center.
Theorem 2: A uniform spherical shell of charge exerts no force on a
charge particle placed inside the shell.
a
Proof : r Er er
Symmetry consideration E Er e r
Q , r a a: radius of shell
Gauss' law 4 r Er 0
2
0, r a Q: total charge on shell
Q , r a (as if Q were at r 0)
Er 4 0r 2
0, r a (Q produces no E) 11

1.3 Gausss Law (continued)

( x)(x x) 3
Exercise: Coulomb's law states E(x) 1 d x.
4 0 |x x|3
Assume (x) is a finite function of x. Use the integral form of
Gauss's law and order-of-magnitude considerations to explain why
E(x) as x x [x is any point in ( x)].
Solution:
If E(x) , it must be due to charge(s) closest to x.
Apply Guess's law: s E nda 1 v (x)d 3 x to a sphere of
0
infinistesimal radius around x. Then, S 2 and v 3
E (| x || x | ) 2 1 3
0
E due to charges inside the sphere of radius
E (| x || x | ) 0 as 0

12
 1.4 Differential Form of Gausss Law
Using the divergence theorem:
n is a unit vector normal to the
3 surface element da and pointing
v Ad x s A nda, away from the volume v enclosed
by surface S .
we obtain from s E n da 1 v (x)d 3 x [(1.11)]
0
3 3 S
1
s E n da v Ed x 0 v (x)d x da n
3
v ( E ) d x 0 [for any v] v (1.12)
0

E
0
differential form of Gauss's law (1.13)

If v f (x)d 3 x 0 for any volume v, then f (x) 0 everywhere.

This is the basis for obtaining (1.13) from (1.12).
Question: If s A nda 0 for any closed surface S , does it imply
A = 0 everywhere? Ans.: No (Example: B - field)
13

1.5 Another Equation of Electrostatics

and the Scalar Potential
| x x |n
n
x [( x x) 2 ( y y) 2 ( z z ) 2 ] 2 e x operates on x.
n
y [( x x) 2 ( y y) 2 ( z z ) 2 ] 2 e y operates on x.
n | x x ' |n | x x |n
z [( x x) 2 ( y y) 2 ( z z ) 2 ] 2 e z
n
2 2 2 2 1
2 [( x x) ( y y) ( z z ) ] 2( x x)e x
n x x
n 1
n2 [( x x) 2 ( y y) 2 ( z z ) 2 ] 2 2( y y)e y x
n x
2 2 1
n2 [( x x) ( y y) ( z z ) 2 ] 2 2( z z )e z 0
n | x x |n2 (x x) (1)
Examples : (frequently used later)
(1) | x x | xx ; (2) 1 xx3 ; (3) 1 3 3 xx5
|xx| |xx| |xx| |xx| |xx|
14
 1.5 Another Equation of Electrostatics and the Scalar Potential (continued)

1 (x)(xx) 1 (x) 3
E(x) 4 d 3 x 4 d x
0 |xx|3 0 |xx|
operates on x, hence can be |x 1x| x x3
3
moved out of the d x-integral. |x x|
1 ( x) d 3 x (x) : electrostatic potential
4
0
|xx|
1 (x) 3 x: finite value
where (x) 4 d x (1.17)
0 |xx| x : any value, e.g. x
Question: What and where is the reference value of in (1.17)?
It is simple to show that if E , then E = 0 (1.14)
Note: E = 0 also gives E (Griffith, 3rd ed., Sec. 1.6.2)
Stokes's theorem: c A d s ( A) n da loop C
d
d : a differential length on a closed loop C n
da S (open
S : arbitrary open surface bounded by loop C surface
n: unit vector normal to surface element da in bounded
the direction given by the right -hand rule by loop C )
15

1.5 Another Equation of Electrostatics and the Scalar Potential (continued)

Work done to bring charge q from position A to B

position B along any path: E
B
W A F d Question: Is W the work done d
d
by E or by an external force? A
q AB E d

x e x y e y z e z ; d dxe x dye y dze z
q AB d

d x dx y dy z dz d
q AB d
d is an infinitesimal change of due to an
q(B A ) infinitesimal displacement d .
Thus, W depends only on the values of at A and B, and it is
independent of the charge's path from A to B, which justifies the
concept of potential energy. Thus, the total work done on q along
any closed path C is 0, i.e. C E d 0 (1.21)
Stokes's thm. C E d S ( E) n da S ( E) n da 0.
S is an arbitrary open surface We again obtain E 0 [(1.14)].
16
 1.7 Poisson and Laplace Equations
(Sec. 1.6 will be covered later)
E (1.13)
Rewrite 0
E (1.16)
Sub. (1.13) into (1.16), we obtain the Poisson equation

2 (1.28)
0
In a charge-free region, (1.28) reduces to the Laplace euqation
2 0 (1.29)
Questions:
1. What do we need to know to solve a differential equation such as

2 ? (Ans.: "region of interest" and "boundary conditions")
0
2
2. In , does include charges on the boundary? (Ans.: No).
0
3. If 0 and 0, where are the charges that produce ?
(Ans.: Charges on and/or outside the boundary). 17

Summary of Secs. 1-5 and 7:

q1
F
qq1 E e
er 4 0r 2 r re r x x1
4 0 r 2
q (xx1) r x x1
E lim qF 1
q0 4 0|xx1|3 principle of
er 1 (x)(xx) d 3 x linear superposition
E 4
r |xx|3
q1
x1 x v
S 0
| x x |n
41 (x) d 3 x n | x x |n2 (x x)
x |xx|
0 x 0
1 (x) 3
4 d x
0 |x x|
0
derived in Sec. using E qe r
f 0 4 0r 2
s E n da 1 v ( x ) d x
3
0
E 0 divergence thm.
E 0
2 0 18
 Questions on Secs. 1-5 and 7:
1. Can one calculate E by using E = 0 alone? Ans.: No
Helmholtzs Theorem (Griffiths, Intro. to Electrodynamics, Sec.
1.6.1): If a vector field goes to 0 at infinity, it is uniquely determined
by its divergence and curl.
2. To solve 2 = 0, we need to know the region of interest and
boundary conditions. Do we need the same information to solve
1 (x) 3
(x) 4 d x?
0 |xx|
1 (x) 3
3. Why break one equation [ (x) 4 d x ] into 2 equations:
0 |xx|
E = 0 and E = 0 ?
4. Coulombs law gives E = 0 and E = /0 . Can it give any
other independent relation for E?
Ans.: No (by Helmholtzs Theorem)
19

5. Is the integral form of Gausss law mathematically equivalent

to the differential form of Gausss law?
Answer: Yes. To prove the equivalence, we need to show that
the integral form of Gausss law is both a sufficient and necessary
condition for the differential form of Gausss law. This can be
demonstrated as follows:
s E nda 10 v (x)d x
3
(1.11)
divergence thm.
3
v ( E 0 ) d x 0 (for arbitrary volume v)

E (1.13)
0
Downward manipulation () shows that (1.11) is a sufficient
condition for (1.13). Upward manipulation () shows that (1.11) is a
necessary condition for (1.13). Thus, the two forms of Gausss law
are mathematically equivalent (hence always physically equivalent).
20
 6. Is Gausss law mathematically equivalent to Coulombs law?
Answer: No, because Coulombs law is a sufficient but not a
necessary condition for Gausss law. That is, we may derive Gausss
law from Coulombs law, but not the reverse.
Coulombs law completely determines E, but Gausss law cannot
completely determine E. This becomes clear if we write Gausss law
in its differential form, E = 0 . By Helmholtzs Theorem, we
also need E to completely determine E. In electrostatics, this is
given by E = 0. In general, it is given by Faradays law,
E B
t [Ch. 5]
Thus, E 0 gives different E in static and dynamic cases.
1 (x)(xx) 3
Note: E(x) 4 d x is valid only for static E. In
0 |xx|3
later chapters we show s E nda 1 v (x)d 3 x (or E 0 )
0
also applies to time-dependent E. 21

7. Is Gausss law physically equivalent to Coulombs law?

Answer: Yes, but only in static cases, in which E is spherically
symmetric about point charge q, so that E = Erer.
Note: E is asymmetric about a moving charge (See Fig. 11.9).
On a spherical surface of radius r centered at
a static charge q, we have E Er er by symmetry.
q q r Er er
2 q
Thus, E nda
s 0 gives E 4 r
r 0
q q
Er E Er er e
4 0r 2 4 0r 2 r
In the above, we have derived Coulombs law from Gausss law.
In Sec. 1.3, we have also derived Gausss law from Coulombs law.
Hence, the two laws are equivalent in electrostatics. The fact they are
conditionally equivalent (in electrostatics only) gives another reason
that they are not mathematically equivalent.
Note : Electrostatics requires not only v 0 for all q, but also
d v 0 ( d v 0 can lead to radiation, as will be studied in Ch. 14).
dt dt 22
 1.6 Surface Distributions of Charges and Dipoles and
Discontinuities in the Electric Field and Potential
Single Layer :
Consider a layer of surface charges. Its density at x defined as
( x ) lim a
q The layer has
a 0 zero thickness.
a, q
Note: and have different units.
q n2
Apply Gau ss's law, E n da , to an
0
infinitesimally thin pillbox, we obtain n1 E2
q
( E1 n1 E 2 n 2 ) a E1 pillbox
0
n1 n 2 (thickness 0)
q
(E E ) n 1
2 1 2 0 a 0 (1.22)
The tangential component of E can be shown to be continuous
across the layer by applying C E d 0 to the loop drawn in
dashed lines in the figure. 23

1.6 Surface Distributions of Charges and Dipoles (continued)

Example : (see figure)

a (uniform distribution of
Apply the shell theorems Q on a layer of radius a )

Q , ra
4 a
Q0 At r a,
is continuous.
Er is discontinuous.
4 r , r a
0 a r
Questions :
1. E and of a point charge diverge as one moves infinistesimally
close to the charge. Explain why E and due to surface charges do
not diverge as one moves infinistesimally close to the surface.
Answer: A point charge is a microscopic quantity (finite amount of
charge at a point ). is macroscopic quantity, which varies smoothly
on the surface (no point charges). One must integrate over a finite
area to obtain a finite amount of charge. Hence, there is only an
infinistesimal amount of charge at any surface point ( no divergence).
2. Why is continuous across the layer? 24
 1.6 Surface Distributions of Charges and Dipoles (continued)
Dipole Layer :
A dipole layer is formed of 2 closely n
(x) |xx xx|
spaced single layers with and . Its
strength is D(x) lim (x)d (x). (x)
d ( x)0
Its direction (n) is from to . x

0 r
x Observation point
1 (x) d 3 x 1 [ (x) da (xnd ) da]
(x) 4
0 |xx| 4 0 s |xx| s ' |x(xnd )|
Since d 0, and of a differential segment of the dipole
layer can be assumed to be at the same point x. For the same reason,
S and S become one surface and da da. Thus,
1
(x) 4 s (x)[|x1x| |x(x1nd )|]da
0 25

1.6 Surface Distributions of Charges and Dipoles (continued)

Using the binomial expansion: (This expansion will be used

throughout the course) n can be negative
and/or a non-integer.
n(n1) n2 2
( x y )n x n nx n1 y x y , (2)
2!
we obtain y n
x
1 1 1 2 12
1 b (1 2 2 2 )
a a b a a+b
|b+a| b b
(b2 a2 2ab) 2 b
2
1 (1 a 2 a2b ) 1 a3b [valid for a b]
b 2b b b b
a
b 0
Let
b x x
a nd
1 1 dn xx3 [valid for d | x x |]
|x(xnd )| |xx| |xx|
26
 1.6 Surface Distributions of Charges and Dipoles (continued)

Sub. 1 1 dn xx3
|x( xnd )| |xx| |xx|
into (x) 1 s (x)[ 1 1 ]da, we obtain
4 0 |xx| |x(xnd )|
|x1x| |x1x| [See (1)]

1
(x) 4 s ( x)d (x) n xx3 da 1 s D (x)n 1 da
0 |xx| 4 0 |xx|
D ( x) (1.24)
and d appear as a product here, so its meaningful
to define the product as the dipole layer strength.
or (x) 1 s D (x) n xx 1 2 da 1 s D (x)d (1.26)
4 0 |xx| |xx| 4 0


cos 1 r 2
d 0, if cos 0

d |xx xx|
d 0, if cos 0

Observation point r 27

1.6 Surface Distributions of Charges and Dipoles (continued)

1 D (x)d (1.26)
4 0 s
Rewrite : (x)
1 1
4 s D (x)n |xx| da (1.24)
0
Note: (1) The direction of n and sign of d are shown below with
respect to the polarity of the dipole layer:
direction of n : sign of d : See derivation
n of (1.26).
x

+ d 0
+ x
d 0
+
(2) The RHS of (1.24) is an explicit function of x (point of
observation). The RHS of (1.26) is an implicit function
of x, because the total solid angle depends on x.
Questions: (1) Under what condition are (1.24) and (1.26) invalid?
(2) What is the reference point for in these 2 eqs.? 28
 1.6 Surface Distributions of Charges and Dipoles (continued)

Special case 1: A double layer with D const.

1
4 D(x)d (1.26) d
s 2D
0
0 2
D
n
2D D D
0 ( 2 0 ) 0
0
(1.27)
electric field between layers: E D . Question: Is E continuous
0d across the dipole layer?
is discontinuous across the dipole layer. p
a
Special case 2: Point dipole d q q
n
p lim a n
D da
point
a 0
dipole a 0 d
a =q
dipole layer dipole layer point dipole
lim a n( d )da n d a nqd
a 0

(x) 41 lim 1 da 1 p(xx)

0 a 0 a
D ( x )n
|xx| 4 0 |xx|3
(1.25)


(1.24) xx3
|xx| 29

Delta Functions (pp. 26 - 27)

( x a)
Definition of delta function:
( x a ) 0 , if x a
a2
a1 ( x a )dx 1 , if a1 a a2 x
a1 a
a2
Note: Since the delta function is defined in terms of an integral,
it takes an integration to bring out its full meaning.
Question: What is the value of a 2 ( x a )dx?
a

Properties of delta function:

(i) a 2 f ( x) ( x a) dx f (a )
a
(3)
1
0
(ii) a f ( x) ( x a )dx f ( x) ( x a ) |aa2 a 2 f ( x) ( x a )dx
a2 a
1 1 1

f (a ) (4)
30
 Delta Functions (continued)

(iii) Let x a be the root of f ( x) 0, then f ( x)

f ( a2 ) f ( a1 )
f (a )
a12 [ f ( x)]dx f ( a12) [ f ( x)] d f1( x) df ( x)
a
a1 a a2
dx
x
f ( a2 ) 1 ( f )df 1 1 , f (a) 0 f ( x ) f (a ) f (a )
f ( a1 ) f f (a ) f ( a ) 1 2

ff ((aa1 )) 1 ( f )df 1 1 , f (a ) 0 a1 a a 2 x
2 f f ( a ) f ( a )
Note: In both expressions above, the integration is from a samller
value to a larger value, as in the definition of the delta function.
Compare with (3) [ f ( x)] 1 ( x a ) [ 1 ( x a )] (5a)
f ( a ) f ( x )
If f ( x) has multiple roots xi [f ( xi ) 0, i 1, 2,], then
[ f ( x)] 1 ( x x ) [ 1 ( x x )] (5b)
i f ( xi ) i f ( x )
i i

( x)
Exercise: Use (5a) to show (a x) ( x a ) and (cx) c
31

Delta Functions (continued)

Extension to 3 dimensions :
1. Cartesian coordinates: x = (x1 , x2 , x3 )
(x x) ( x1 x1 ) ( x2 x2 ) ( x3 x3 ) (6)
V (x x)d 3 x ( x1 x1 )dx1 ( x2 x2 )dx2 ( x3 x3 )dx3

0, if x lies outside V
1, if x lies inside V
x3 x

2. Cylindrical coordinates: x = ( , , z )
x2
(x x) 1 ( ) ( ) ( z z ) (7)
x1
V (x x)d 3 x V (x x) d d dz z x
( ) d ( ) d ( z z )dz

0, if x lies outside V
1, if x lies inside V
Question: If x and x both have the dimension of cm, what are the
dimensions of (x) and (x)? [See Appendix (A), Eq. (A.9).] 32
 Delta Functions (continued)
r
3. Spherical coordinates: r = (r , , )
r
2 1 (r r ) ( ) ( ), or

(r r) r sin (8)
1
r 2 (r r ) (cos cos ) ( )

By (4), (cos cos ) 1 ( ) 1 ( ), 0
sin sin
3 (r r) 2
V (r r)d x V r 2 (cos cos ) ( ) r
drd (cos ) d

0, if r lies outside V d 3x
[see (9) below]
1, if r lies inside V
Note: There are 2 ways to write d 3 x in spherical coordinates.
2 2 2
0 dr 0 rd 0 r sin d 0 r dr sin d 0 d
0
1 2
0 r 2 dr 1 d (cos ) 0 d 11 d (cos )
2
d 3 x r 2 sin drd d , or d (cos ) is to be integrated from
(9)
r drd (cos )d a smaller cos to a larger cos . 33

Delta Functions (continued)

Approximate representations of the delta function :

The delta function, ( x), can be represented analytically by the
following functions because they satisfy the definition of the delta
function in the limit 0 ( > 0).
1
( x) lim
0 x 2 2
2
x2
1
( x) lim e 2
0 2
1 , for x
( x) lim 2 2
0 0, otherwise

34
 Delta Functions (continued)
Representation of volume charge density (x) by delta functions:
of a point charge q at the origin

(x) of a uniform line charge on the z -axis (from to )
of a uniform surface charge on the entire x -y plane
q ( x) ( y ) ( z ), triple infinity

( x) ( y ), double infinity
( z ), single infinity
1
due to q at the origin r 2 (most divergent)
1
E due to on the z -axis r (less divergent)
is
due to on the x -y plane a constant value ( )
2 0
as one approaches the charge.
of point, line, and surface charge densities all have infinite values
because the volume is 0, although the "degree of infinity" is different.
The conceptual difficulty of infinite can be resolved by giving the
point, line, or surface a finite radius a or thickness t and let a or t 0.
35

Delta Functions (continued)

Problem 1: A total charge Q is uniformly distributed around

a circular ring of radius a and infinitesimal thickness. Write the
voulme charge density (x) in cylindrical coordinates.
Solution:
Let (x) = K (r a ) (z ) and find K as follows.
3
(x)d x K (r a) ( z )rdrd dz z
2 Ka Q circular ring
Q
K
2 a
a
Q r
( x) (r a) ( z )
2 a
Note: has the dimension of "charge/volume" as expected.

36
 Delta Functions (continued)
Problem 2: A circular disk of radius a has a uniform . Calculate
and E on its axis (i.e. z -axis). z

circular disk
( z ), if r a
Solution: (x)
0, if r a a
Total charge: Q (x)d x 3 r
2
0 rdr 0 d ( z )dz 2 0 rdr a 2 , as expected
a a

( z , r 0) 41 (x) d 3 x a r dr 2 d ( z)
dz

0 |xx| 4 0 0 0 ( z r2 )1/ 2
2

2 0 2 1 2 1/ 2 r dr 2 [( z 2 a 2 )1/ 2 z ], for z 0 a 2
a
0 ( z r ) 0 (2)

( z a, r 0) 2 [ z (1 2 ) z ] 2 [ z (1 a 2 ) z ) 4 1z
2 1/ 2 2 Q
a
0 z 0 2z 0

E( z 0, r 0) ( z , r 0) 2 [1 2 2 1/ 2 ]e z
z
0 ( z a ) Coulomb's
E( z 0, r 0, a ) 2 e z , as expected law
0
E( z 0, r 0) e z E on the surface is independent of a (disk
2 0

radius) E on the surface is due to an infinistesimal area of . 37

Delta Functions (continued)

Problem 3: Prove 2 1r 4 (r )
(r ) 0, if r 0 r
Solution: Definition of (r ) : 3
(r )d x 1
r
Hence, we need to prove
(i) 2 1 0, if r 0
r
(ii) 2 r1 d 3 x 4 (r )d 3 x 4
It is convenient to use the spherical coordinates. To prove (i), we
we write 2 as (see back cover of Jackson)
2 12 (r 2 ) 2 1 (sin ) 1 2
r r r r sin r 2 sin 2 2
2
2 r1 12 d (r 2 d 1r ) 12 d ( r 2 ) 0 if r 0
r dr dr r dr r
2
Note: r 2 is undetermined at r 0. However, here we are only
r
concerned with the region r 0. 38
 Delta Functions (continued)

To prove (ii), we integrate 2 1r over a spherical volume V

21 3 1 3 1 da r 2 1 d 4
v r d x v r d x s er
r s r 2
divergence thm. 12 er r 2 d
r
Note: r > 0 on S No need to evaluate r 2 / r 2 at r = 0.
Change to a coordinate system in which r x x
r
and r x x . We obatin from 2 1r 4 (r ) x
x
2 1 4 (x x) (1.31)
|x x| 0
Note: Jackson outlined the same method to prove (1.31) on p. 35.
( x) (x) 3
Problem 4: Derive 2 (x) from (x) 1 d x
0 4 0 |x x|
Solution: 2 (x) 1 (x) 2 1 d 3 x
4 0 |x x|
( x)
1 (x) 4 (x x) d 3 x [same as (1.28)]
4 0 0
Note: Jackson did this problem with a different method (p. 35). 39

1.8 Greens Theorem

Green's theorem - a powerful theorem for treating electrostatic
boundary value problems
3
v A d x s A n da [divergence theorem]
Let A = , where and are arbitrary functions of position.
A 2

S
A n n n da n
Sub. these 2 expressions for A and A n in the v
divergence theorem, we obtain Green's first identity,
2 3
v ( )d x s n da (1.34)
Interchange and in (1.34).

v ( 2 )d 3 x s da
n
Subtract these two equations, we obtain Green's second identity,
2 2 3
v ( )d x s ( n n )da (1.35)
40
 1.8 Greens Theorem (continued)

Green's theorem relates a volume integral to a surface integral and

the volume integral contains the operator 2 . These features are useful
for the manipulateion of the Poisson equation in bounded space.
For example, applying Green's second identity:
2 2 3
v ( )d x s ( n n )da (1.35)
we may convert the Poisson equation into an integral equation.
In (1.35), letting be R 1 ( 1 ), be the electrostatic potential
|xx|
2
(thus, ), and x be the integration variable, we obtain
0

v [4 (x x) 01R (x)]d x s [ n ( R1 ) R1 n ]da
3

x inside v
( x) 4 1 ( x) 3 1
[ 1 ( 1 )]da

0 v x x
d x

4 s R n n R
(1.36)
Note: (1.36) is an integral equation (not a solution) for . However,
in infinite space, ( R ) R 1 . The the surface integral (2nd term)
1 (x) 3
vanishes and (1.36) reduces to the solution: ( x) 4 d x.
0 |xx| 41

1.9 Uniqueness of Solution with Dirichlet or

Neumann Boundary Conditions
Dirichlet boundary condition: s specified

Neumann boundary condition: n s specified
Cauchy boundary condition: and both specified
s n s
As another application of Green's theorem, we use it to prove the
uniqueness theorem for the solution of the Poisson equation.
Let there be two solutions, 1 and 2 , both satisfying
2 n on S (Dirichlet b.c.), or
0 with

n n on S (Neumann b.c.) V
n

21 0 1,2 n on S , or
i.e. 2 with
2 0
n 1,2 n on S S

2 U n n 0 on S , or
Define U 1 2 , then U 0 with
n U n n 0 on S 42

 1.9 Uniqueness of Solution (continued)

Rewrite Green's 1st identity: v ( 2 )d 3 x s n da
Let U ( 1 2 ) b.c. U 0 or U / n 0 on S
U U U )d 3 x s U U
2 3 2
v (U
n da 0 v U d x0
0
U 0 everywhere within V V n
0, if U 0 on S
U 1 2 U
const , if n 0 on S
S

1 and 2 differ by at most a constant, hence are the same solution.

Note: Since the solution is uniquely determined by specifying either
or / n on S , the Cauchy b.c. ( and / n both specified
on S ) is an over-specification, which may lead to inconsistency.
Question: What is a "solution"?
Exercise : Prove that there cannot be any static E inside a closed,
hollow conductor if there is no charge in the hollow region. 43

1.10 Formal Solution of Electrostatic Boundary-Value

Problem with Green Function
The Green function
is the solution of a boundary-value problem
with a point source. In Jackson (Sec. 1.10), 3 notations appear:
G (x, x) : Green function under either Dirichlet or Neumann b.c.

GD (x, x) : Green function under Dirichlet b.c.
GN (x, x) : Green function under Neumann b.c.
These all refer to Green functions in electrostatics. GN (x, x) is
rarely used. We will consider only GD (x, x).
GD 0
Green Function under Dirichlet B.C. point
source S
GD (x, x) is the solution of (see figure)
2GD (x, x) 4 (x x) with GD (x, x) 0 for x on S , (10)
where x is treated as a variable and x (the arbitrary position of a point
source within S ) is treated as a constant.
Physically, if 4 q / 0 , GD (x, x) is the potential at x due to a
point charge q at x subject to the b. c. that the potential vanishes on S .
44
 1.10 Formal Solution of Electrostatic Boundary-Value Problem(continued)

Symmetry Property of GD (x, x) : GD 0 for

Consider two equations: one with a point y on S x
x
source at x, the other with a point source at x: 0
S
2y GD (y , x) 4 (y x), b.c. GD (y , x) 0 for y on S
2
y GD (y , x) 4 (y x), b.c. GD (y , x) 0 for y on S

Rewrite: v ( 2y 2y )d 3 y s [ n n ]da (1.35)
Let GD (y , x) and GD (y , x), where y is the variable.
4 ( y x) 4 ( y x)

v [GD (y , x) 2y GD (y , x) GD (y , x') 2y GD (y , x)]d 3 y
s [ GD ( y, x) GD (y , x) GD (y , x) GD ( y, x)]da 0
n n
GD (x, x) GD ( x, x) [symmetry property of GD (x, x)]
Application: GD ( x, x) 0 for x on S GD (x, x) 0 for x on S.
Exercise: Give 2 physical examples to show GD ( x, x) GD (x, x).
45

1.10 Formal Solution of Electrostatic Boundary-Value Problem(continued)

Formal Solution of Electrostatic Boundary -Value Problem :
Rewrite 2GD ( x, x) 4 ( x x), GD ( x, x) 0 on S (10)
and Green's 2nd identity: GD 0
2 2 3 point
v [ (x) (x) (x) (x)]d x source S
s [ (x) n (x) (x) n (x)]da (1.35)

Consider a general electrostatic problem with Dirichlet b.c.:
2 (x) (x) / 0 with (x) s (x) for x on S (11)
In (1.35), let (x) be the solution of (11) and let (x) [ G D (x,x)]
be the solution of (10) in the same region enclosed by S . Then,
4 ( xx) ( x)
0 s
2 2 3
v [ (x) GD (x, x) GD (x, x) (x)]d x S
s [ (x) n GD (x, x) GD (x, x) n (x)]da

0 on S
GD ( x,x)
v (x)GD (x, x)d x 41 s (x)
1 3
(x) 4 n
da (1.44)
0 46
 1.10 Formal Solution of Electrostatic Boundary-Value Problem(continued)
Rewrite s
2 (x) ( x) / 0 with ( x) s( x) on S S (11)
2
GD (x, x) 4 (x x), with GD ( x, x) 0 on S (10)
(x) 1 ( ) ( , ) 3
1 (x) GD ( x,x)da (1.44)

4 0 v
x G D x x d x 4 s n
1 (x) 3
Note: 1. (x) 4 d x is applicable only to infinite space.
0 |xx|
2. (1.44) is a solution applicable to any region with Dirichlet b.c.
3. See (1.46) for the counterpart of (1.44) with Neumann b.c.
Steps to evaluate (1.44): GD 0
point S

1. Solve (10) for GD (x, x ). source
2. Substitute ( x), GD ( x, x), s ( x), and GD ( x, x) / n into (1.44).
It is often much simpler to solve GD (x, x) from (10) than solving
directly from (11), since (10) has the simple b.c. of GD (x, x) 0 on S .
Applications of (1.44) can be found in Chs. 2 and 3. The problem
below gives an application without the need to solve (10) for G (x, x). 47

1.10 Formal Solution of Electrostatic Boundary-Value Problem(continued)

Problem: A hollow cube (see figure) has six
square sides. There is no charge inside. Five sides
are grounded. The sixth side, insulated from the
others, is held at a constant potential 0. Find at
the center of the cube.
Solution: Let the center of the cube be at x 0 and rewrite (1.44):
(x) 41
v (x) GD (x, x)d x 41 s (x) n GD (x, x)da (1.44)
3
0
If all 6 sides had the potential 0 , then ( x 0) 0 and by (1.44)
0 41 s 0 n GD (0, x)da (12)
For the present problem, we have 0 on side 1, 0 on the
other 5 sides, and the same GD ( x, x). By (1.44), at the center is
(x 0) 41 side 10 n GD (0, x)da 16 41 s 0 n GD (0, x)da

0 by (12)
16 0 n GD (0, x) is symmetric
with respect to all six sides. 48
 1.11 Electrostatic Potential Energy and Energy
Density; Capacitance
Electric Field Energy: Let be due to . The work done to add is
W (x) (x)d 3 x Using ( a) a a
v
we obtain
Change to E by 0 E
E ( E) E
0 v (x) E( x)d 3 x ( E) E E
0 v ( E) d 3 x 0 v E Ed 3 x 0 v E Ed 3 x

By conservation of s
E da 0 Note: For this integral to
energy, we postulate 1 1 r2 vanish, the integration
r r2 r must be over all space.
W to be the total
E-field energy. 1 dE 2 ( x ) E 2 (x)

2
E ( x) 0 2
W 0 v d x 0 E(x) dE(x) 2 v E(x) d 3 x [v ]
3
(1.54)
2
E E E E ( E) E ( E)
0
0
as r

0
W 2 v (x) (x)d x 2 s E da 12 v (x) (x)d 3 x
1 3
(1.53)
49

1.11 Electrostatic Potential Energy (continued)

An alternative derivation of (1.53) and (1.54):

Consider a state in which a charge density (x) has produced
an electrostatic potential (x), i.e.
(x) (x).

Then, by the principle of superposition,
(x) (x), where 0 1.
To find the electric field energy, we consider the energy needed to
build up (x) from 0 (no charge and no potential) to 1 (the
present state). At any stage in the build-up process, the relative
charge density (hence the relative potential) remains the same;
namely, the intermediate state is characterized by the charge density
(x) and potential (x).
In such a build-up process, when the potential is (x), the work
done by adding an incremental charge (x)d is
dW v d 3 x (x) (x)d 50
 1.11 Electrostatic Potential Energy (continued)

Hence, the total work done from 0 to 1 is 0

2

1 1

W v d 3
x (x) (x) 0 d
1 d 3 x ( x) ( x)
0 dW 2 v (1.53)
Green's 1st identity This integral 0 only if r


20 v 2 d 3 x 20 [ v d 3 x s ( n )da]
2
20 v E d 3 x [v ] ~ 1r ~ r12 r 2 (1.54)
Questions: 1. If we bring q and q toward each other, the work done
2
is negative. Why is then W 20 v E d 3 x always positive?
2. Give one example to show that the E-field carries energy.
3. If we do work to bring q from 1 to 2 , where does the work end in?
2
Electric Field Energy Density : We postulate wE 20 E (1.55)
Question : Can the field energy density of multiple charges be
separately calculated, then linearly summed?
2 2
Answer: No, because wE 20 E 20 ( E j ) ( E j ) 20 E j
51

1.11 Electrostatic Potential Energy (continued)

Force on the Surface of a Conductor: see Griffiths, Sec. 2.5.3)
Consider a conductor with on it. At any point conductor
(observation point) on the surface, the total field
(Etotal) must be to the surface and Etotal (inside)0. Etotal
By Gausss law, we have Etotal (outside), where
is the local value at the observation point (see upper figure).
The local will produce equal and opposite fields (call it self see
lower figure) on both sides of and, by Gausss law, Eself Eself
Eself 2 , which is half of Etotal (outside). E E
0 ext ext
Note: Eself is due to on an infinitesimal area (see notes p. 36).
Since Etotal (inside)0, all the external charges away from the local
must have produced an external field (Eext and Eext=Eself so
that it cancels self inside and hence doubles self outside. The local
can only experience a force due to Eext (not Eself ). Thus,
Force on the surface E 2 (see pp. 42-43)
unit area ext 2 0 52
 Appendix A: Unit Systems and Dimensions
Unit Systems:
Two systems of electromagnetic units are in common use today:
the SI and Gaussian systems. Regardless of ones personal
preference, it is important to be familiar with both systems and, in
particular, the conversion from one system to the other. Conversion
formulae can be divided into two categories: symbol/equation
conversion such as E and E = q/(40r2)] and unit conversion
(such as coulomb).
Conversion formulae for symbols and equations are listed in
Table 3 on p. 782 of Jackson and conversion formulae for units in
Table 4 on p. 783 (both tables attached on next page). These two
tables are all we need to convert between SI and Gaussian systems.
Correct use of the tables requires practices.
53

Appendix A: Unit Systems and Dimensions (continued)

Jackson, p. 782, Table 3 Jackson, p. 783, Table 4

54
 Appendix A: Unit Systems and Dimensions (continued)

Conversion of symbols and equations:

Consider, for example, the conversion of the SI equation
q
E (A.1)
4 0r 2
into the Gaussian system.
This involves the conversion of symbols and equations. So we use
Table 3. First, we note from Table 3 (top) that mechanical symbols
(e.g. time, length, mass, force, energy, and frequency) are unchanged
in the conversion. Thus, we only need to deal with electromagnetic
symbols on both sides of (A.1).
G
From Table 3, we find E SI E and q SI 4 0 q G (A.2)
4 0
Sub. E G / 4 0 and 4 0 q G , respectively, for E and q in (A.1),
we obtain the corresponding equation in the Gaussian system:
4 0 qG qG
EG E G
2 (A.3)
4 0 4 0r 2 r 55

Appendix A: Unit Systems and Dimensions (continued)

Conversion of units and evaluation of physical quantities:

q
Consider again the SI equation : E (A.1)
4 0r 2
Given r 0.01 m, q 1 statcoulomb, we may evaluate E in 3 steps:
Step 1: Express r , q, and 0 in SI units. From Table 3 (bottom)
and Table 4, we find
0 8.854 1012 Farad/m 1 Farad/m
36 109
r 0.01 m (same as given) (A.4)
q ( 1 statcoulomb) 1 9 coulomb
310
Step 2: Sub. the numbers (but not the units ) from (A.4) into (A.1).
1
This gives E
q
3109 3 104
4 0 r2 4 1
9
(0.01)2
36 10
Step 3: Look up Table 4 for the SI unit of E. As shown in Table 4,
the SI unit of E is V/m. Thus, E 3 104 V/m (A.5)
56
 Appendix A: Unit Systems and Dimensions (continued)

As another exercise, we write (A.1) in the Gaussian system :

q
E 2 (A.3)
r
and evaluate E for the same r ( 0.01 m) and q ( 1 statcoulomb).
Step 1: Express r and q in Gaussian units. From Table 4, we find

r ( 0.01 m) 1 cm
q 1 statcoulomb (same as given)
(A.6)

Step 2: Sub. the numbers (but not the units) from (A.6) into (A.3).
q
This gives E 2 11 1
r
Step 3: Look up Table 4 for the Gaussian unit of E. We find the
unit to be statvolt/cm. Thus, E 1 statvolt/cm (A.7)
Table 4 shows 1 statvolt/cm 3 10 4 V/m. Hence, the 2 results
E 3 104 V/m
in (A.5) and (A.7): are identical as expected.
E 1 statvolt/cm
57

Appendix A: Unit Systems and Dimensions (continued)

Units and Dimensions :

In the Gaussian system, the basic units are length (), mass (m), and
time (t ). In the SI system, they are the above plus the current ( I ). [See
Table 1 (top) on p. 779 of Jackson.] All other units are derived units.
If a physical quantity is expressed in term of the basic units, we
have the dimension of this quantity.
A mechanical quantity has the same dimension in both systems.
For example, the acceleration a ( d 2 x / dt 2 ) has the dimension of
t 2 . From f ma, we obtain the dimension of force : mt 2 , which
in turn gives the dimension of work (f ) or energy: m 2t 2 .
An electromagnetic quantity has different dimensions in different
systems. For example, the charge q has the SI dimension of It. From
the Gaussian equation f q1q2 / r 2 and the dimensions of force and
length, we find the Gaussian dimension of q to be m1/23/2t 1. Since
q has the dimension of energy (m 2t 2 ), the potential has the SI
dimension of m 2t 3 I 1 and the Gaussian dimension of m1/ 21/ 2t 1. 58
 Appendix A: Unit Systems and Dimensions (continued)

All physical quantities in an equation must be expressed in the

same unit system and all terms must have the same dimension. For
example, by Stokes's theorem, we have
C E d S ( E) n da (A.8)
where both terms have the dimension of (the dimension of E ).
In the definition of the delta function:
a12 ( x a)dx 1,
a
(A.9)
the RHS is dimensionless. Thus, if x has the dimension of , ( x a )
must have the dimension of 1. However, "0" is not to be regarded
as a dimensionless quantity. This is clear if we write (A.8) as
C E d S ( E) n da 0.
Well-known equations need not be checked for dimensional
consistency. However, for newly derived equations, a dimensional
check can be a convenient way to find mistakes. 59