CHAPTER 2: Boundary-Value Problems

in Electrostatics: I

Applications of Greens theorem

1
 2.6 Green Function for the Sphere; General
Solution for the Potential
The general electrostatic problem (upper figure):
2 (x) 1 (x) with b.c.
s s
0 s

has the formal solution: (see Sec. 1.10)

(x) 1 v (x)GD (x, x)d 3 x x x ((x)
4 0
1 x S
0
4
s ( x ) n
G D ( x , x ) da , (1 44)
(1.44)
where the Green function GD (x, x) is the solution of (lower figure)
2GD (x, x) 4 (x x) with b GD (x, x) 0 for
i h b.c. f x on S
GD (x, x) 0
GD (x, x) can be regarded as the potential
d to a unit
due i source (q 4 0 , p. 64) source
i point unit point
at an arbitrary position x inside the same GD (x, x)
surface
f S , but
b t with
ith the
th homogeneous
h b.c.
b S
x x (same S as in
GD (x, x) = 0 for x on S. 0 upper figure) 2
 2.6 Green Function for the Sphere (continued)

(1 44) to find due to a point charge q at x b

Example 1: Use (1.44)
in infinite space.
q
2 (x) (x b) with b ( ) 0
ith b.c.
0
In order to use ((1.44),
) we first obtain the Green function from
2GD (x, x) 4 (x x) with GD (x, x) 0 for x on S (2)
1
The solution of (2) is GD (x, x)
| x x |
Sub q (x b) for (x) and 1/ | x x | for GD (x, x) into (1
Sub. (1.44)
44)
1
q ( xb ) |x x|
1
1 0
GD (x, x)
3
( x) v
(x ) GD (x, x ) d x s (x ) da
4 0 4 n
1 q

4 0 | x b |
3
 2.6 Green Function for the Sphere (continued)

Example 2: 2 (x) 0 with b.c. (r a ) (a, , )

Find (x) in the region r a (see left figure).
First, find GD (x, x) from x
the equation (see right figure) n x n x
2GD (x, x) 4 (x x) S a S a

with GD ( x, x) 0 on S . (a, , ) GD (x, x) 0

Note : n points outward from the volume of interest
interest.
This equation has the solution (see Sec. 2.2):
1 a 2
GD (x, x) 4 (x x)
GD (x, x)
| x x | x x a 2 x in region of interest (r a )
x2
1 1
(3)

12 12
x x2 2 xx cos
2 x2 x2 a 2 2 xx cos
a2
Note: GD (x, x) GD (x, x) angle between x and x 4
 2.6 Green Function for the Sphere (continued)

GD (x, x) GD (x, x) ( x2 a2 )

n '
x a x
x a a ( x 2
a 2
2 ax cos ) 32

Sub. (3) into (1.44)

0
1 1 GD (x, x)
3
( x) ( x ) G ( x , x ) d x ( x ) da
4 0 v 4 s
D
n
1 a( x 2 a 2 )
s (a, , ) 2 d (2
(2.19)
19)
4 2
( x a 2ax cos ) 3 2

Questions:
1. In (3), we have GD (x, x) |x1x| a as a solution of
x xa 2x/x2
2GD (x, x) 4 (x x)). But GD (x, x) |x1x| apparently also
satisfies the same equation. Does this violate the uniqueness thm.?
2 C
2. Can th
the solu ti off 2GD (x, x) 4 (x x) be
l tion b written
itt ini the
th
form GD (x, x) GD (x x)? why? 5
 2.1 Method of Images

The method of images is not a general method. It works for some

problems with a simple geometry.
geometry Consider a point charge q located
in front of an infinite and grounded plane conductor (see figure). The
region
g of interest is x 0 and is ggoverned byy the Poisson equation:
q
s 0 0 image
s s0 0
2 q s
( x) ( x y ) x charge
x
0 q q q
subject to the boundary y y y
condition x x
0 0 region of
region of
( x 0) 0. interest interest
In order to maintain a zero potential on the conductor, surface
charge
h will
ill be
b induced
i d d (by(b q ) on the
h conductor.
d W may simulate
We i l
the effects of the surface charge with a hypothetical "image charge",
q, located symmetrically behind the conductor.
conductor Then,
hen 6
 2.1 Method of Images (continued)

q
(x) 4 1 1
0 |xy| |xy|
0 0
s s
image s s 0 0
x charge
and, by symmetry, (x) satisfies q q x q
the boundary condition y y y
x x
( x 0)) 0. 0
region of
0
region of
Operate (x) with 2 interest interest
q
2 (x) [ (x y ) (x y )] (1)
0
In the region of interest (x 0), we have ( x y ) 0. Thus,
(x) obeys the original Poisson equation
q
2 (x) ( x y ) This shows that we must put the image
0 charge
g outside the region
g of interest
Since (x) satisfies both the Poisson equation and the boundary
condition in the region of interest, it is a solution. By the uniqueness
theorem, it is the only solution. Note that the Poisson equation (1) and
the solution (x) are irrelevant outside the region of interest. 7
 2.2 Point Charge in the Presence of a Grounded
Conducting Sphere n
Refer to the conducting sphere of radius a x
shown in the figure.
g Assume a point
p charge
g q B qn
a q y
is at r y ( a ). To find for r a, we put an
image charge q at r y ( a ). Then, y
q 4 0 q 4 0
( x)
xy x y y a a2
First set , or y ' = ,
First,
q 4 0 q 4 0 a y' y

xn yn xn yn y
so that n a n = a n n
y
B d condition
Bounday diti requires
i
Note: y ' a; hence, q' lies outside
q 4 0 q 4 0 the region of interest.
(a) 0
a n ay n y a nn q q'
y Next, set so that RHS 0.
q 4 0 aq 4 0 a y'
( x) y' a
xy a 2 Thi gives
This i q ' q q.
y x 2 y a y
y 8
 2.2 Point Charge in the Presence of a Grounded Conducting Sphere (continued)

q 4 0 aq 4 0 This is eq
equivalent
i alent to (2
(2.1)
1)
Rewrite ( x) and (2.4) of Jackson.
xy 2
y x a 2 y
y
2 q
In the region of interest (r a), we have ( x) ( x y ).
0
Thus as in the case of the plane cond
Thus, uctor satisfies the Poisson
conductor,
equation and the b.c. It is hence the only solution. The E-field lines
aaree shown
s ow in tthee figure
gu e below.
be ow.

n
x
B qn
a q y

y

9
 2.2 Point Charge in the Presence of a Grounded Conducting Sphere (continued)

Surface charge density on the sphere: The solution for (x) can be
expressed in terms of scalars as
q 1 a
(x) 4 [ 2 2 1/ 2
4 2 ]
0 ( x y 2 xy cos ) y ( x a 2 2 xay cos )
2 1/ 2
y n
where is the angle between x and y.
x
By Gauss's law, the surface charge B qn
density at point B is a q y
y
0 Er ( x a ) 0 x xa
a 2
q 2a 2 y cos a (2a 2 y cos )
8 [ 2 2 3/ 2
4 3 ]
(a y 2ayy cos ) y (a 2 a 2 2 ay cos )3/ 2
2
y
1 a2
q a y
( ) 2 (2 5)
(2.5)
(1 a 2 2 ay cos )
2 y
4 a 3/ 2
y 10
 2.2 Point Charge in the Presence of a Grounded Conducting Sphere (continued)

Total charge on the sphere:

The total surface charge can be obtained n
x
by integrating over the spherical surface.
B
However, it can be deduced from a simple n
a q y q
argument: In the region r a, the electric
field due to the surface charge is exactly y
the same as that due to the image charge q.
Hence, by
b GGauss's' llaw, the
h totall surface
f charge
h must bbe q( ay q ).
)
Force on q:
Since, at the position of charge q, the field produced by the image
charge q is the same as that produced by the surface charge, the force
on q is the Coulomb force betwen q and q.

3
q ( a q) a
1 qq 1 y 1 q 2 y
F=
n
n 2 n (2.6)
4 0 ( y y) 2 4 0 ( y a ) 2
2
4 0 a (1 a ) 2
2
y y2 11
 2.3 Point Charge in the Presence of a Charged,
Insulated Conducting Sphere (with Total Charge Q)
Insulated,
If the sphere is insulated with total charge Q
on its surface, we may obtain in two steps.
Step 1: Ground the sphere
same problem as in Sec. 2.2
q 4 0 aq 4 0
( x)
xy y xa 2y/y 2 s
00

with total surface charge q aq y. Q y

aq

Stepp 2: Disconnect the g

ground wire.
Add Q aq / y to the sphere so that the
total charge on the sphere is Q. Then,
Q aq y will be distributed uniformly
on the surface because the charges were already in static equilibrium.
Q aq y
due to Q aq y is ( x)
4 0 x 12
 2.3 Point Charge in the Presence of a Charged, Insulated, Conducting Sphere (continued)

Hence, the total is

1 Qaq y
( x) [ xqy aq
] (2.8)
4 0 2
y xa y/y 2 x
q (Qaq / y ) y force due to
The force on q is the force in (2.6) plus
4 0 y3 added charge
g
1 q qa3 (2 y 2 a 2 ) y Q y q
F Q a (2.9)
4 0 y
2 2 2 2
y( y a ) y
qQ
As y , F 4 y 2 (Coulomb force between point charges)
0
As y a, F is always attractive even if q and Q have the same
sign.
Question: If there is an excess of electrons on the surface, why
don'tt they leave the surface due to mutual repulsion?
don
(See p. 61 for a discussion on the work function of a metal.)
13
 2.7 Conducting Spheres with Hemisphere
(to be discussed
d scussed in Sec. 3.3)
2.8 Orthogonal Functions and Expansions
Definition of Orthogonal Functions :
Consider a set of real or complex functions U n ( ) (n 1,2, )
which are square integrable on the interval a b.
inner product

b 0,
0 mn
orthogonal, if a U n ( )U m ( )d
U n ( )'s are 0, m n
b 0 mn
0,
orthonormal, if U n ( )U m ( )d mn
a 1, m n
G
Geometrical
i l analogue:
l ex, ey, andd ez are an orthonormal
th l sett off
unit vectors, i.e. emen = mn. By comparison, the dot product emen
is similar to the inner product . But the algebraic set Un() can be
infinite in number. 14
 2.8 Orthogonal Functions and Expansions (continued)

Linearly Independent Functions :

The set of U n ( ) 's are said to be linearly independent if the
only solution of anU n ( ) 0 (for every in the range of
n
a b) is an 0 for any n.
If a set off ffunctions
i are orthogonal,
h l they
h are also
l linearly
li l
independent.
Prooff :
P
anU n ( ) 0
n
0, unless
mn
b b
a anU n ( )U m ( )d an a U n ( )U m ( )d
n n
b 2
an a U n ( ) d 0
an 0 for
f any n
15
 2.8 Orthogonal Functions and Expansions (continued)
Gram-Schmidt Orthogonalization Procedure:
Orthogonality is a sufficient, but not necessary, condition for
linear independence, i.e. linearly independent functions do not
have to be orthogonal. However, they can be reconstructed into an
orthogonal set by the Gram-Schmidt orthogonalization procedure.
Consider two vectors, e x and (e x e y ), as a simple example.
These two vectors are not orthogonal, because e x (e x e y ) 0, but
are linearly independent because ae x b(e x e y ) 0 a b 0.
We may form two new vectors as linear combinations of the old
vectors, e1 e x and e 2 e x +e y + e x , and demand e1 e 2 0.
e1 e 2 0 1+ 0 1 e 2 e y
The new set, e1 ( e x ) and e 2 ( e y ), are thus orthogonal (as well
as linearly independent).
The same procedure can be applied to algebraic functions.
16
 2.8 Orthogonal Functions and Expansions (continued)

Completeness of a Set of Functions :

Expand an arbitrary, square integrable function f ( ) in terms of
a finite number (N ) of functions in the orthonormal set U n ( ),
N
f ( ) anU n ( ) (2.30)
n 1
and
d ddefine
fi the
th mean square error (M N ) as
2
b N
M N a f ( ) anU n ( ) d .
n 1
If there exists a finite number N0 such that for N > N0 the mean
square error MN can be b made d smaller
ll than
th any arbitrarily
bit il smallll
positive quantity by proper choice of ans, then the set Un( ) is
said to be complete and the series representation

anU n ( ) f ( ) (2.33)
n 1
is said to converge in the mean to f ( ). 17
 2.8 Orthogonal Functions and Expansions (continued)

R
Rewrite (2 33) f ( ) anU n ( )
it (2.33): (2 33)
(2.33)
n 1
Using the orthonormal property of U n ( ) 's, we get
b
an a U n* ( ) f ( ) d (2.32)
Ch
Change (2 32) tto and
iin (2.32) d substitute
b tit t (2.32)
(2 32) iinto
t (2.33)
(2 33)
b *
f ( ) a U n ( )U n ( ) f ( ) d ((2.34))
n 1

f ( ) is arbitrary U n* ( )U n ( ) ( ) (2.35)
n1

(completeness or closure relation)
Jackson, p. 68: "All orthonormal sets of functions normally
occurring in mathematical physics have been proved to be
complete."
l " (Thi
(This statement will
ill be
b illustrated
ill d in
i Sec.
S 2.9.)
29)
18
 2.8 Orthogonal Functions and Expansions (continued)

Fourier Series : example p of completep set of orthogonal

g functions
Exponential representation of f ( x) on the interval a2 x a2 :
ikn x f ( x)
f ( x) an e

n
a
x (4)
2a a
kn 2 n ; an 1 2a f ( x)e ikn x dx 2
a a
2
In (4), f ( x) is in general a complex function and, even when f ( x)
is real
real, an is in general a complex constant.
constant
In the case f ( x) is real , we have the realty condition: an a*n
P f : f ( x) reall f ( x) f * ( x)
Proof n n
ik n x ik x
an e an* e n a*n eikn x
n n n
ikn x
an a*n (since e is linearly independent)
Questions: 1. Why "n to " instead of "n 0 to " ?
2. Why kn 2 n a instead of kn n a ? 19
 2.8 Orthogonal Functions and Expansions (continued)
Sinusoidal representation of f ( x) on the interval a2 x a2 :

f ( x) an e ikn x
a0 an eikn x a n e ikn x
n n 1

a0 an cos kn x a n cos kn x i an sin kn x a n sin kn x
n 1

a0 an a n cos kn x i an a n sin kn x
n 1 n 1
A0
f ( x) An cos kn x Bn sin kn x , kn 2a n (5)
2 n1
where Same as ((2.36)) and (2.37)
( )
a a

An an a n a a
2



ik

ik


1 2 f ( x) e n e n dx 2 2 f ( x) cos k xdx
x x
a a
2
n
(n 0 ) 2cos kn x
a a

(n 1 ) a
2

ik

x

ik x


Bn i an a n i 2a f ( x) e n e n dx 2 2a f ( x) sin kn xdx

a
2
2i sin kn x 20
 2.8 Orthogonal Functions and Expansions (continued)

Discussion:
Di i : It is often more convenient
con enient to express
e press a physical
ph sical
quantity (a real number) in the exponential representation than in
th sinusoidal
the i id l representation,
t ti b
because th complex
the l coefficient
ffi i t (a
( n)
of an exponential term carries twice the information of the real
coefficient
ffi i t (An or Bn) off a sinusoidal
i id l term.
t F example,
For l if
x(t) = Re[aeit]
i the
is h displacement
di l off a simple
i l harmonic
h i oscillator,
ill the
h complexl a
contains both the magnitude and phase of the displacement. In the
sinusoidal representation, the same quantity will be written
x(t) = Acos(t) + Bsin(t).
Exponential terms are also easier to manipulate (such as
multiplication and differentiation). This point will be further
discussed in Ch. 7.
21
 2.8 Orthogonal Functions and Expansions (continued)
Fourier Transform :
If the interval becomes infinite (a ), we obtain the Fourier
transform (see Jackson p.68).
1 ikx
f ( x ) A ( k ) e dk (2.44)
2

A(k ) 1 f ( x)eikx dx
2

(2.45)

Change x to x in (2.45) and sub. (2.45) into (2.44)

1 ik ( x x)
f ( x) dxf ( x) e dk
2

( x x)
1 ikk ( x x)

e dk ( x x) [completeness relation] (2.47)
2
Questionn 1: Does A(k ) contain any more or any less information
Questio
than f ( x) ?
22
 2.8 Orthogonal Functions and Expansions (continued)

Questi on 2 : Does an in f ( x) an eikn x have the same dimension

stion
n

as A(k ) in f ( x) 1 A(k )eikx dk ?
2
[assuming x is a dimensional quantity.]
1 ik ( x x)
Rewrite (2.47):
(2 47):
e dk ( x x)
2
Interchange x and k
1 i ( k k ) x

e dx (k k ), [orthogonality condition] (2.46)
2
Let y k k and sub. it into (2.46)
1 ixy

e dx ( y )
2
Since (y ) ( y ), we may write more generally,
1 ixy

e dx ( y ) (6)
2 23
 2.8 Orthogonal Functions and Expansions (continued)

There are two useful theorems concerning the Fourier integral:

(1) Parseval's theorem :
2 2
Parseval'ss theorem states f ( x) dx A( k ) dk
The Parseval (7)
Proof :
f ( x) 1 ikx
2
A( k ) e dk (2.44)
(2 44)
Rewrite the Fourier transform:
A(k ) 1
f ( x)eikx dx (2.45)
(2 45)
2
2 *

f ( x ) dx
f ( x ) f ( x)dx

1 ikx 1 *
ik x
dx A(k )e dk A (k )e dk
2 2
* 1 i ( k k ) x 2

dkA ( k )
dk A ( k )
dxe
A ( k ) dk
2

( k k ) 24
 2.8 Orthogonal Functions and Expansions (continued)

(2) Convolution theorem :

The convolution theorem states
1 [ f ( x ) f ( )d ]eikx dx A (k ) A (k )
2
1 2 1 2 (8)

This is called the convolution of f1 ( x) and f 2 ( x)
where the factor 1 follows the convention in (2.44) and (2.45).
2
1 1 ikx
k
Proof : LHS of (8) f
2
( ) d
f1 ( x )e dx
2 2
Let x (dx d )
1 1 ik ( )
2
f ( ) d f
1
( )e d
2 2
1 ik 1 ik
2
f ( ) d 1
f ( )e d
2 2
A1 (k ) A2 (k ) 25
 2.9 Separation of Variables, Laplace Equation in
Rectangular Coordinates
2 2 2
2 Laplace equation in
2 2 2 0 Cartesian ((2.48))
x y z di t
C t i coordinates
Let ( x, y, z )= X ( x)Y ( y ) Z ( z ) (2.49)
1 d 2 X 1 d 2Y 1 d 2 Z
2
2
2
0 (2.50)
X dx Y dy Z dz
Since this equation holds for arbitrary values of x, y , and z , each
of the three terms must be separately constant.
d2X 2 d 2Y 2 d 2Z 2 2 2 2
2 X ; Y ; Z subject to
dx dy 2 d 2
dz
ei x ei y e z
X ( x) i x ; Y ( y ) i y ; Z ( z ) z with 2 2
e e e
26
 2.9 Separation of Variables, Laplace Equation in Rectangular Coordinates (continued)

Example : Find inside a charge-free

charge free rectangular box (see figure)
with the b.c. (x, y, z c) = V (x, y ) and 0 on other sides.
X ( x) Aei x Bei x
X (0) 0 B A X A(ei x ei x ) A 'sin x
n , n 1
X ( a ) 0 n a 1, 22, z

X ( x) An sin n x
n 1
Similarly, Y ( y ) Aei y Bei y .
Y (0) 0 and Y (b) 0 give y

Y ( y ) Am sin m y, m bm
m1
Solution for Z : Z ( z ) Ae z Be z
x
Z (0) 0 B A Z ( z ) A(e z e z ) A ''sinh z

Anm sin n x sin m y sinh nm z , nm 2n 2m (2.56)
n,m 1 27
 2.9 Separation of Variables, Laplace Equation in Rectangular Coordinates (continued)

b c on the z c plane:
To find Anm , we apply the b.c.
( x, y , z c ) V ( x, y )

V ( x, y ) Anm sin n x sin m y sinh nm c (2.57)
n,m1
4 a b
Anm dx dyV ( x, y )sin n x sin m y (2.58)
ab sinh nm c 0 0

Questions:
Q ti
1. The method of images is not a general method, but the method of
expansion
i in
i orthogonal
th l functions
f ti i Why?
is. Wh ?
2. In electrostatics, only charges can produce . In this problem,
= 0,
0 how
h can there
th be b ?
3. Can we find the surface charge distribution ( ) on the walls from
the
h knowledge
k l d off inside
i id the
h box?
b ? If so, under
d what
h condition?
di i ?
28
 2.9 Separation of Variables, Laplace Equation in Rectangular Coordinates (continued)

Discussion:

Rewrite (2.57): V ( x, y ) Anm sin n x sin m y sinh nm c ,
n,m 1
where n an and m bm .
z

This is a good example to substantiate

the following statement on physics ground:
"All orthonormal sets of functions normally y
occurring in mathematical physics have been
proved to be complete." (p. 68) x
In (2.57),
( i ( n x) andd sin
) sin( i ( m y ) are orthogonal
h l functions
f i
generated in a physics problem. Physically, we expect the problem
to have a solution for any boundary condition on the surface z c,
i.e. for any function V ( x, y ) specified in (2.57). Thus, sin( n x) and
and sin ( m y ) must each form a complete p set in order to represent
p
an arbitrary V ( x, y ). 29
 Homework of Chap. 2

Problems: 1,, 2,, 3,, 4,, 5,,

9, 23, 26

30