Solutions to Problems in Jackson,
Classical Electrodynamics, Third Edition
Homer Reid
December 8, 1999
Chapter 2: Problems 11-20
Problem 2.11
A line charge with linear charge density τ is placed parallel to, and a
distance R away from, the axis of a conducting cylinder of radius b held
at fixed voltage such that the potential vanishes at infinity. Find
(a) the magnitude and position of the image charge(s);
(b) the potential at any point (expressed in polar coordinates with the
origin at the axis of the cylinder and the direction from the origin
to the line charge as the x axis), including the asymptotic form far
from the cylinder;
(c) the induced surface-charge density, and plot it as a function of angle
for R/b=2,4 in units of τ /2πb;
(d) the force on the charge.
(a) Drawing an analogy to the similar problem of the point charge outside
the conducting sphere, we might expect that the potential on the cylinder can
be made constant by placing an image charge within the cylinder on the line
conducting the line charge with the center of the cylinder, i.e. on the x axis.
Suppose we put the image charge a distance R0 < b from the center of the
cylinder and give it a charge density −τ . Using the expression quoted in Problem
2.3 for the potential of a line charge, the potential at a point x due to the line
charge and its image is
Φ(x)
=
τ
R2
R2
τ
ln
ln
−
4π0 |x − Rî|2
4π0 |x − R0 î|2
1
Homer Reid’s Solutions to Jackson Problems: Chapter 2
2
τ
|x − R0 î|2
ln
.
4π0
|x − Rî|2
=
We want to choose R0 such that the potential is constant when x is on the
cylinder surface. This requires that the argument of the logarithm be equal to
some constant γ at those points:
|x − R0 î|2
|x − Rî|2
=γ
or
b2 + R02 − 2R0 b cos φ = γb2 + γR2 − 2γRb cos φ.
For this to be true everywhere on the cylinder, the φ term must drop out, which
requires R0 = γR. We can then rearrange the remaining terms to find
R0 =
b2
.
R
This is also analogous to the point-charge-and-sphere problem, but there are
differences: in this case the image charge has the same magnitude as the original
line charge, and the potential on the cylinder is constant but not zero.
(b) At a point (ρ, φ), we have
Φ=
τ
ρ2 + R02 − 2ρR0 cos φ
ln 2
.
4π0
ρ + R2 − 2ρR cos φ
For large ρ, this becomes
0
Φ→
1 − 2 Rρ cos φ
τ
ln
.
4π0
1 − 2R
ρ cos φ
Using ln(1 − x) = −(x + x2 /2 + · · ·), we have
Φ →
=
τ 2(R − R0 ) cos φ
4π0
ρ
τ R(1 − b2 /R2 ) cos φ
2π0
ρ
(c)
σ
∂Φ = −0
∂ρ r=b
2b − 2R cos φ
τ
2b − 2R0 cos φ
−
= −
4π b2 + R02 − 2bR0 cos φ b2 + R2 − 2bR cos φ
"
#
2
b − bR cos φ
b − R cos φ
τ
= −
−
2π b2 + Rb42 − 2 bR3 cos φ b2 + R2 − 2bR cos φ
Homer Reid’s Solutions to Jackson Problems: Chapter 2
3
Multiplying the first term by R2 /b2 on top and bottom yields
"
#
R2
τ
b −b
σ = −
2π R2 + b2 − 2bR cos φ
R 2 − b2
τ
= −
2πb R2 + b2 − 2bR cos φ
(d) To find the force on the charge, we note that the potential of the image
charge is
τ
C2
.
Φ(x) = −
ln
4π0 |x − R0 î|2
with C some constant. We can differentiate this to find the electric field due to
the image charge:
E(x)
τ
∇ ln |x − R0 î|2
4π0
τ 2(x − R0 î)
= −
.
4π0 |x − R0 î|2
= −∇Φ(x) = −
The original line charge is at x = R, y = 0, and the field there is
E=−
1
τ
R
τ
î = −
î.
2π0 R − R0
2π0 R2 − b2
The force per unit width on the line charge is
F = τE = −
τ2
R
2π0 R2 − b2
tending to pull the original charge in toward the cylinder.
Problem 2.12
Starting with the series solution (2.71) for the two-dimensional potential
problem with the potential specified on the surface of a cylinder of radius
b, evaluate the coefficients formally, substitute them into the series, and
sum it to obtain the potential inside the cylinder in the form of Poisson’s
integral:
Z 2π
1
b2 − ρ 2
Φ(ρ, φ) =
Φ(b, φ0 ) 2
dφ0
2π 0
b + ρ2 − 2bρ cos(φ0 − φ)
What modification is necessary if the potential is desired in the region of
space bounded by the cylinder and infinity?
Homer Reid’s Solutions to Jackson Problems: Chapter 2
4
Referring to equation (2.71), we know the bn are all zero, because the ln
term and the negative powers of ρ are singular at the origin. We are left with
Φ(ρ, φ) = a0 +
∞
X
ρn {an sin(nφ) + bn cos(nφ)} .
(1)
n=1
Multiplying both sides successively by 1, sin n0 φ, and cos n0 φ and integrating
at ρ = b gives
Z 2π
1
Φ(b, φ)dφ
(2)
a0 =
2π 0
Z 2π
1
an =
Φ(b, φ) sin(nφ)dφ
(3)
πbn 0
Z 2π
1
bn =
Φ(b, φ) cos(nφ)dφ.
(4)
πbn 0
Plugging back into (1), we find
)
(
Z
∞
1 2π
1 X ρ n
0
0
0
Φ(ρ, φ) =
[sin(nφ) sin(nφ ) + cos(nφ) cos(nφ )] dφ0
+
Φ(b, φ )
π 0
2 n=1 b
)
(
Z
∞
1 2π
1 X ρ n
0
0
=
cos n(φ − φ ) .
(5)
+
Φ(b, φ )
π 0
2 n=1 b
The bracketed term can be expressed in closed form. For simplicity define
x = (ρ/b) and α = (φ − φ0 ). Then
∞
1 X n
+
x cos(nα)
2 n=1
=
=
=
=
=
=
∞
1 1 X n inα
+
x e
+ xn e−inα
2 2 n=1
1 1
1
1
+
+
−
2
2 2 1 − xeiα
1 − xe−iα
1 1 1 − xe−iα − xeiα + 1
+
−
2
2 2 1 − xeiα − xe−iα + x2
1
1 − x cos α
+
−
1
2
1 + x2 − 2x cos α
1
x cos α − x2
+
2 1 + x2 − 2x cos α
1
1 − x2
.
2 1 + x2 − 2x cos α
Plugging this back into (5) gives the advertised result.
Homer Reid’s Solutions to Jackson Problems: Chapter 2
5
Problem 2.13
(a) Two halves of a long hollow conducting cylinder of inner radius b are
separated by small lengthwise gaps on each side, and are kept at different
potentials V1 and V2 . Show that the potential inside is given by
V1 − V 2
2bρ
V1 + V 2
−1
+
tan
cos φ
Φ(ρ, φ) =
2
π
b2 − ρ 2
where φ is measured from a plane perpendicular to the plane through the
gap.
(b) Calculate the surface-charge density on each half of the cylinder.
This problem is just like the previous one. Since we are looking for an
expression for the potential within the cylinder, the correct expansion is (1)
with expansion coefficients given by (2), (3) and (4):
a0
=
=
=
an
=
=
=
=
bn
=
=
=
1
2π
Z
2π
Φ(b, φ)dφ
0
Z 2π π
1
V1
dφ + V2
dφ
2π
0
π
V1 + V 2
2
Z π
Z 2π
1
sin(nφ)dφ
sin(nφ)dφ
+
V
V
2
1
πbn
π
0
h
i
1
π
2π
−
V
|cos
nφ|
+
V
|cos
nφ|
1
2
0
π
nπbn
1
−
[V1 (cos nπ − 1) + V2 (1 − cos nπ)]
n
nπb
0
, n even
2(V1 − V2 )/(nπbn ) , n odd
Z π
Z 2π
1
cos(nφ)dφ
cos(nφ)dφ
+
V
V
2
1
πbn
π
0
h
i
1
π
2π
V
|sin
nφ|
+
V
|sin
nφ|
1
2
0
π
nπbn
0.
Z
With these coefficients, the potential expansion becomes
Φ(ρ, φ) =
V1 + V 2
2(V1 − V2 ) X 1 ρ n
sin nφ.
+
2
π
n b
n odd
(6)
6
Homer Reid’s Solutions to Jackson Problems: Chapter 2
Here we need an auxiliary result:
X 1
xn sin nφ =
n
n odd
=
=
1 X 1
(x = iy)
(iy)n [einπ − e−inφ ]
2i
n
n odd
∞
1 X (−1)n iφ 2n+1
(ye )
− (ye−iφ )2n+1
2 n=0 2n + 1
1 −1 iφ
tan (ye ) − tan−1 (ye−iφ )
2
(7)
where in the last line we just identified the Taylor series for the inverse tangent
function. Next we need an identity:
γ1 − γ 2
−1
−1
−1
tan γ1 − tan γ2 = tan
.
1 + γ 1 γ2
(I derived this one by drawing some triangles and doing some algebra.) With
this, (7) becomes
X 1
1
2iy sin φ
xn sin nφ =
tan−1
n
2
1 + y2
n odd
2x sin φ
1
−1
tan
.
=
2
1 − x2
Using this in (6) with x = ρ/b gives
V1 − V 2
V1 + V 2
+
tan−1
Φ(ρ, b) =
π
π
2ρb sin φ
b2 − ρ 2
(Evidently, Jackson and I defined the angle φ differently).
.
Homer Reid’s Solutions to Jackson Problems: Chapter 2
7
Problem 2.15
(a) Show that the Green function G(x, y; x0 , y 0 ) appropriate for Dirichlet
boundary conditions for a square two-dimensional region, 0 ≤ x ≤ 1, 0 ≤
y ≤ 1, has an expansion
G(x, y; x0 , y 0 ) = 2
∞
X
gn (y, y 0 ) sin(nπx) sin(nπx0 )
n=1
where gn (y, y 0 ) satisfies
2
∂
2 2
−
n
π
gn (y, y 0 ) = δ(y 0 − y)
∂y 2
and gn (y, 0) = gn (y, 1) = 0.
(b) Taking for gn (y, y 0 ) appropriate linear combinations of sinh(nπy 0 )
and cosh(nπy 0 ) in the two regions y 0 < y and y 0 > y, in accord with the
boundary conditions and the discontinuity in slope required by the source
delta function, show that the explicit form of G is
G(x, y; x0 , y 0 ) =
∞
X
1
−2
sin(nπx) sin(nπx0 ) sinh(nπy< ) sinh[nπ(1 − y> )]
nπ
sinh(nπ)
n=1
where y< (y> ) is the smaller (larger) of y and y 0 .
(I have taken out a factor −4π from the expressions for gn and G, in accordance
with my convention for Green’s functions; see the Green’s functions review
above.)
(a) To use as a Green’s function in a Dirichlet boundary value problem G must
satisfy two conditions. The first is that G vanish on the boundary of the region
of interest. The suggested expansion of G clearly satisfies this. First, sin(nπx0 )
is 0 when x0 is 0 or 1. Second, g(y, y 0 ) vanishes when y 0 is 0 or 1. So G(x, y; x0 , y 0 )
vanishes for points (x0 , y 0 ) on the boundary.
The second condition on G is
2
∂
∂2
∇2 G =
G = δ(x − x0 ) δ(y − y 0 ).
(8)
+
∂x02
∂y 02
With the suggested expansion, we have
∞
X
∂2
G = 2
gn (y, y 0 ) sin(nπx) −n2 π 2 sin(nπx0 )
02
∂x
n=1
∞
X ∂ 02
∂2
G = 2
g (y, y 0 ) sin(nπx) sin(nπx0 )
02
2 n
∂y
∂y
n=1
8
Homer Reid’s Solutions to Jackson Problems: Chapter 2
We can add these together and use the differential equation satisfied by gn to
find
∇2 G = δ(y − y 0 ) · 2
∞
X
sin(nπx) sin(nπx0 )
n=1
0
= δ(y − y ) · δ(x − x0 )
since the infinite sum is just a well-known representation of the δ function.
(b) The suggestion is to take
An1 sinh(nπy 0 ) + Bn1 cosh(nπy 0 ),
gn (y, y 0 ) =
An2 sinh(nπy 0 ) + Bn2 cosh(nπy 0 ),
y 0 < y;
y 0 > y.
(9)
The idea to use hyperbolic sines and cosines comes from the fact that sinh(nπy)
and cosh(nπy) satisfy a homogeneous version of the differential equation for g n
(i.e. satisfy that differential equation with the δ function replaced by zero).
Thus gn as defined in (9) satisfies its differential equation (at all points except
y = y 0 ) for any choice of the As and Bs. This leaves us free to choose these
coefficients as required to satisfy the boundary conditions and the differential
equation at y = y 0 .
First let’s consider the boundary conditions. Since y is somewhere between
0 and 1, the condition that gn vanish for y 0 = 0 is only relevant to the top line
of (9), where it requires taking Bn1 = 0 but leaves An1 undetermined for now.
The condition that gn vanish for y 0 = 1 only affects the lower line of (9), where
it requires that
0 = An2 sinh(nπ) + Bn2 cosh(nπ)
= (An2 + Bn2 )enπ + (−An2 + Bn2 )e−nπ
(10)
One way to make this work is to take
An2 + Bn2 = −e−nπ
and
Bn2 = enπ + An2
→
− An2 + Bn2 = enπ .
Then
so An2 = − cosh(nπ)
and
2An2 = −enπ − e−nπ
Bn2 = sinh(nπ).
With this choice of coefficients, the lower line in (9) becomes
gn (y, y 0 ) = − cosh(nπ) sinh(nπy 0 )+sinh(nπ) cosh(nπy 0 ) = sinh[nπ(1−y 0 )] (11)
for (y 0 > y). Actually, we haven’t completely determined An2 and Bn2 ; we could
multiply (11) by an arbitrary constant γn and (10) would still be satisfied.
Next we need to make sure that the two halves of (9) match up at y 0 = y:
An1 sinh(nπy) = γn sinh[nπ(1 − y)].
(12)
9
Homer Reid’s Solutions to Jackson Problems: Chapter 2
70000
60000
g(yprime)
50000
40000
30000
20000
10000
0
0
0.2
0.4
0.6
0.8
1
yprime
Figure 1: gn (y, y 0 ) from Problem 2.15 with n=5, y=.41
This obviously happens when
An1 = βn sinh[nπ(1 − y)] and γn = βn sinh(nπy)
where βn is any constant. In other words, we have
βn sinh[nπ(1 − y)] sinh(nπy 0 ),
gn (y, y 0 ) =
βn sinh[nπ(1 − y 0 )] sinh(nπy),
= βn sinh[nπ(1 − y> )] sinh(nπy< )
y 0 < y;
y 0 > y.
(13)
with y< and y> defined as in the problem. Figure 1 shows a graph of this
function n = 5, y = .41.
The final step is to choose the normalization constant βn such that gn satisfies its differential equation:
2
∂
2 2
gn (y, y 0 ) = δ(y − y 0 ).
(14)
−
n
π
∂ 2 y 02
To say that the left-hand side “equals” the delta function requires two things:
• that the left-hand side vanish at all points y 0 6= y, and
• that its integral over any interval (y1 , y2 ) equal 1 if the interval contains
the point y 0 = y, and vanish otherwise.
The first condition is clearly satisfied regardless of the choice of βn . The second
condition may be satisfied by making gn continuous, which we have already
done, but giving its first derivative a finite jump of unit magnitude at y 0 = y:
Homer Reid’s Solutions to Jackson Problems: Chapter 2
10
y0 =y+
∂
0 g
(y,
y
)
n
0 − = 1.
∂y 0
y =y
Differentiating (13), we find this condition to require
nπβn [− cosh[nπ(1 − y)] sinh(nπy) − sinh[nπ(1 − y)] cosh(nπy)] = −nπβn sinh(nπ) = 1
so (14) is satisfied if
βn = −
1
.
nπ sinh(nπ)
Then (13) is
gn (y, y 0 ) = −
sinh[nπ(1 − y> )] sinh(nπy< )
nπ sinh(nπ)
and the composite Green’s function is
G(x, y; x0 , y 0 )
= 2
∞
X
gn (y, y 0 ) sin(nπx) sin(nπx0 )
n=1
∞
X
= −2
sinh[nπ(1 − y> )] sinh(nπy< ) sin(nπx) sin(nπx0 )
(15)
.
nπ sinh(nπ)
n=1
Problem 2.16
A two-dimensional potential exists on a unit square area (0 ≤ x ≤ 1, 0 ≤
y ≤ 1) bounded by “surfaces” held at zero potential. Over the entire square
there is a uniform charge density of unit strength (per unit length in z).
Using the Green function of Problem 2.15, show that the solution can be
written as
∞
4 X sin[(2m + 1)πx]
cosh[(2m + 1)π(y − (1/2))]
Φ(x, y) = 3
.
1−
π 0 m=0
(2m + 1)3
cosh[(2m + 1)π/2]
Referring to my Green’s functions review above, the potential at a point x0
within the square is given by
Z
I 1
∂G 0
0 ∂Φ Φ(x0 ) = −
−
G(x
;
x
)
G(x0 ; x0 )ρ(x0 )dV 0 +
Φ(x0 )
0
0 dA .
0 V
∂n
∂n
0
S
x
x
(16)
In this case the surface integral vanishes, because we’re given that Φ vanishes
on the boundary, and G vanishes there by construction. We’re also given that
11
Homer Reid’s Solutions to Jackson Problems: Chapter 2
ρ(x0 )dV 0 = dx0 dy 0 throughout the entire volume. Then we can plug in (15) to
find
Z 1Z 1
∞
2 X
1
Φ(x0 ) =
sinh[nπ(1−y> )] sinh(nπy< ) sin(nπx0 ) sin(nπx0 )dx0 dy 0 .
π0 n=1 n sinh(nπ) 0 0
(17)
The integrals can be done separately. The x integral is
sin(nπx0 )
Z
1
sin(nπx0 )dx0
0
sin(nπx0 )
= −
[cos(nπ) − 1]
nπ
(2 sin(nπx0 ))/nπ , n odd
=
0
, n even
(18)
The y integral is
sinh[nπ(1 − y0 )]
=
=
=
Z
y0
0
0
sinh(nπy )dy + sinh(nπy0 )
0
Z
1
sinh[nπ(1 − y 0 )]dy 0
y0
y0
1 o
1 n
sinh[nπ(1 − y0 )] · cosh(nπy 0 )0 − sinh[nπy0 ] · cosh[nπ(1 − y 0 )]y0
nπ
1
{sinh[nπ(1 − y0 )] cosh(nπy0 ) + sinh(nπy0 ) cosh[nπ(1 − y0 )] − sinh(nπy0 ) − sinh[nπ(1 − y0 )]}
nπ
1
{sinh[nπ] − sinh[nπ(1 − y0 )] − sinh(nπy0 )}.
(19)
nπ
Inserting (18) and (19) in (17), we have
X sin(nπx0 ) 4
sinh[nπ(1 − y0 )] + sinh(nπy0 )
Φ(x0 ) = 3
1−
.
π 0
n3
sinh(nπ)
n odd
The thing in brackets is equal to what Jackson has, but this is tedious to show
so I’ll skip the proof.
12
Homer Reid’s Solutions to Jackson Problems: Chapter 2
Problem 2.17
(a) Construct the free-space Green function G(x, y; x0 , y 0 ) for twodimensional electrostatics by integrating 1/R with respect to z 0 − z
between the limits ±Z, where Z is taken to be very large. Show that
apart from an inessential constant, the Green function can be written
alternately as
G(x, y; x0 , y 0 ) = − ln[(x − x0 )2 + (y − y 0 )2 ]
= − ln[ρ2 + ρ02 − 2ρρ0 cos(φ − φ0 )].
(b) Show explicitly by separation of variables in polar coordinates that the
Green function can be expressed as a Fourier series in the azimuthal
coordinate,
∞
1 X im(φ−φ0 )
e
gm (ρ, ρ0 )
G=
2π −∞
where the radial Green functions satisfy
m2
δ(ρ − ρ0 )
1 ∂
0 ∂gm
ρ
− 02 gm =
.
0
0
0
ρ ∂ρ
∂ρ
ρ
ρ
Note that gm (ρ, ρ0 ) for fixed ρ is a different linear combination of the
solutions of the homogeneous radial equation (2.68) for ρ0 < ρ and
for ρ0 > ρ, with a discontinuity of slope at ρ0 = ρ determined by the
source delta function.
(c) Complete the solution and show that the free-space Green function has
the expansion
G(ρ, φ; ρ0 , φ0 ) =
m
∞
1 X 1 ρ<
1
ln(ρ2> ) −
· cos[m(φ − φ0 )]
4π
2π m=1 m ρ>
where ρ< (ρ> ) is the smaller (larger) of ρ and ρ0 .
(As in Problem 2.15, I modified the text of the problem to match with my
convention for Green’s functions.)
(a)
R
= [(x − x0 )2 + (y − y 0 )2 + (z − z 0 )2 ]1/2
≡ [a2 + u2 ]1/2 , a = [(x − x0 )2 + (y − y 0 )2 ]1/2
, u = (z − z 0 ).
Integrating,
Z
Z
−Z
du
2
[a + u2 ]1/2
h
i +Z
= ln (a2 + u2 )1/2 + u −Z
Homer Reid’s Solutions to Jackson Problems: Chapter 2
= ln
(Z 2 + a2 )1/2 + Z
(Z 2 + a2 )1/2 − Z
= ln
(1 + (a2 /Z 2 ))1/2 + 1
(1 + (a2 /Z 2 ))1/2 − 1
≈ ln
2+
13
a2
2Z 2
a2
2Z 2
2
4Z + a2
a2
2
= ln[4Z + a2 ] − ln a2 .
= ln
Since Z is much bigger than a, the first term is essentially independent of a and
is the ’nonessential constant’ Jackson is talking about. The remaining term is
the 2D Green’s function:
G = − ln a2
= − ln[(x − x0 )2 + (y − y 0 )2 ] in rectangular coordinates
= − ln[ρ2 + ρ02 − 2ρρ0 cos(φ − φ0 )] in cylindrical coordinates.
(b) The 2d Green’s function is defined by
Z
∇2 G(ρ, φ; ρ0 , φ0 )ρ0 dρ0 dφ0 = 1
but ∇2 G = 0 at points other than (ρ, φ). These conditions are met if
∇2 G(ρ, φ; ρ0 , φ0 ) =
1
δ(ρ − ρ0 )δ(φ − φ0 ).
ρ0
(20)
You need the ρ0 on the bottom there to cancel out the ρ0 in the area element in
the integral. The Laplacian in two-dimensional cylindrical coordinates is
1 ∂
1 ∂
2
0 ∂
∇ = 0 0 ρ
− 02 02 .
ρ ∂ρ
∂ρ0
ρ ∂φ
Applying this to the suggested expansion for G gives
∞ 0
m2
1 X 1 ∂
0 ∂gm
ρ
− 02 gm eim(φ−φ ) .
∇ G(ρ, φ; ρ , φ ) =
0
0
0
2π −∞ ρ ∂ρ
∂ρ
ρ
2
0
0
If gm satisfies its differential equation as specified in the problem, the term in
brackets equals δ(ρ − ρ0 )/ρ0 for all m and may be removed from the sum, leaving
∇2 G(ρ, φ; ρ0 , φ0 )
∞
1 X im(φ−φ0 )
e
2π −∞
=
δ(ρ − ρ0 )
ρ0
·
=
δ(ρ − ρ0 )
ρ0
δ(φ − φ0 ).
Homer Reid’s Solutions to Jackson Problems: Chapter 2
14
(c) As in Problem 2.15, we’ll construct the functions gm by finding solutions of
the homogenous radial differential equation in the two regions and piecing them
together at ρ = ρ0 such that the function is continuous but its derivative has a
finite jump of magnitude 1/ρ.
For m ≥ 1, the solution to the homogenous equation
m2
1 ∂
0 ∂
ρ
− 02 f (ρ0 ) = 0
ρ0 ∂ρ0
∂ρ0
ρ
is
f (ρ0 ) = Am ρ0m + Bm ρ0−m .
Thus we take
gm =
A1m ρ0m + B1m ρ0−m
A2m ρ0m + B2m ρ0−m
, ρ0 < ρ
, ρ0 > ρ.
In order that the first solution be finite at the origin, and the second solution
be finite at infinity, we have to take B1m = A2m = 0. Then the condition that
the two solutions match at ρ = ρ0 is
A1m ρm = B2m ρ−m
which requires
A1m = γm ρ−m
B2m = ρm γm
for some constant γm . Now we have

m
 γ m ρ0
ρ m
gm =
 γm ρ0
ρ
,
ρ0 < ρ
,
ρ0 > ρ
The finite-derivative step condition is
dgm 1
dgm −
=
dρ0 ρ0 =ρ+
dρ0 ρ0 =ρ−
ρ
or
−mγm
so
1 1
+
ρ ρ
γm = −
=
1
ρ
1
.
2m
Then
gm

0 m
 − 1 ρ
, ρ0 < ρ
2m
ρ m
=
 − 1 ρ0
, ρ0 > ρ
2m ρ
m
1
ρ<
.
= −
2m ρ>
Homer Reid’s Solutions to Jackson Problems: Chapter 2
15
Plugging this back into the expansion gives
m
∞
0
1 X 1 ρ<
eim(φ−φ )
G = −
4π −∞ m ρ>
m
∞
1 X 1 ρ<
= −
cos[m(φ − φ0 )].
2π 1 m ρ>
Jackson seems to be adding a ln term to this, which comes from the m = 0
solution of the radial equation, but I have left it out because it doesn’t vanish
as ρ0 → ∞.
Problem 2.18
(a) By finding appropriate solutions of the radial equation in part b of
Problem 2.17, find the Green function for the interior Dirichlet problem of a cylinder of radius b [gm (ρ, ρ0 = b) = 0. See (1.40)]. First
find the series expansion akin to the free-space Green function of
Problem 2.17. Then show that it can be written in closed form as
2 02
ρ ρ + b4 − 2ρρ0 b2 cos(φ − φ0 )
G = ln 2 2
b (ρ + ρ02 − 2ρρ0 cos(φ − φ0 ))
or
(b2 − ρ2 )(b2 − ρ02 ) + b2 |ρ − ρ0 |2
.
G = ln
b2 |ρ − ρ0 |2
(b) Show that the solution of the Laplace equation with the potential
given as Φ(b, φ) on the cylinder can be expressed as Poisson’s integral of Problem 2.12.
(c) What changes are necessary for the Green function for the exterior
problem (b < ρ < ∞), for both the Fourier expansion and the
closed form? [Note that the exterior Green function is not rigorously
correct because it does not vanish for ρ or ρ0 → ∞. For situations
in which the potential falls of fast enough as ρ → ∞, no mistake is
made in its use.]
(a) As before, we write the general solution of the radial equation for gm in the
two distinct regions:
A1m ρ0m + B1m ρ0−m , ρ0 < ρ
0
gm (ρ, ρ ) =
(21)
A2m ρ0m + B2m ρ0−m , ρ0 > ρ.
The first boundary conditions are that gm remain finite at the origin and
vanish on the cylinder boundary. This requires that
B1m = 0
16
Homer Reid’s Solutions to Jackson Problems: Chapter 2
and
A2m bm + B2m b−m = 0
so
B2m = −γm bm
A2m = γm b−m
for some constant γm .
Next, gm must be continuous at ρ = ρ0 :
m ρ m
b
m
A1m ρ
= γm
−
b
ρ
m b
γ m ρ m
.
−
A1m =
ρm
b
ρ
With this we have
m 0 m
ρ
b
gm (ρ, ρ ) = γm
−
b
ρ
ρ
0 m m b
ρ
−
= γm
,
b
ρ0
ρ m
0
,
ρ0 < ρ
ρ0 > ρ.
Finally, dgm /dρ0 must have a finite jump of magnitude 1/ρ at ρ0 = ρ.
dgm dgm 1
=
−
ρ
dρ0 ρ0 =ρ+
dρ0 ρ0 =ρ−
m−1
m bm
ρ
ρ m
1
b
= mγm
+ m+1 − mγm
−
bm
ρ
b
ρ
ρ
m
1
b
= 2mγm
ρ
ρ
so
γm =
and
0
gm (ρ, ρ ) =
=
1 ρ m
2m b
0 m 0 m 1
ρρ
ρ
−
2
2m
b
ρ
0 m m ρ
ρρ
1
−
2
2m
b
ρ0
or
gm (ρ, ρ0 ) =
1
2m
ρρ0
b2
m
−
ρ<
ρ>
,
ρ0 < ρ
,
ρ0 > ρ.
m .
Plugging into the expansion for G gives
G(ρ, φ, ρ0 , φ0 ) =
0 m m ∞
1 X 1
ρρ
ρ<
cos m(φ − φ0 ).
−
2π n=1 m
b2
ρ>
(22)
Homer Reid’s Solutions to Jackson Problems: Chapter 2
17
Here we need to work out an auxiliary result:
∞
∞ Z x
X
X
1 n
n−1
0
u
du cos m(φ − φ0 )
x cos n(φ − φ ) =
n
0
n=1
n=1
)
Z x( X
∞
1
n
0
=
u cos n(φ − φ ) du
u n=1
0
Z x
cos(φ − φ0 ) − u
du
=
1 + u2 − 2u cos(φ − φ0 )
0
x
1
= − ln(1 − 2u cos(φ − φ0 ) + u2 )0
2
1
= − ln[1 − 2x cos(φ − φ0 ) + x2 ].
2
(I summed the infinite series here back in Problem 2.12. The integral in the
second-to-last step can be done by partial fraction decomposition, although I
cheated and looked it up on www.integrals.com). We can apply this result
individually to the two terms in (22):
1
1 + (ρρ0 /b2 )2 − 2(ρρ0 /b2 ) cos(φ − φ0 )
0
0
G(ρ, φ; ρ , φ ) = −
ln
4π
1 + (ρ< /ρ> )2 − 2(ρ< /ρ> ) cos(φ − φ0 )
2 4
ρ> b + ρ2 ρ02 − 2ρρ0 b2 cos(φ − φ0 )
1
= −
ln
4π
b4
ρ2> + ρ2< − 2ρ< ρ> cos(φ − φ0 )
2 4
ρ> b + ρ2 ρ02 − 2ρρ0 b2 cos(φ − φ0 )
1
= −
ln
4π
b4
ρ2> + ρ2< − 2ρ< ρ> cos(φ − φ0 )
2
ρ>
1
= −
ln
4π
b2
4
b + ρ2 ρ02 − 2ρρ0 b2 cos(φ − φ0 )
1
(23)
ln 2 2
−
4π
b (ρ + ρ02 − 2ρρ0 cos(φ − φ0 ))
This is Jackson’s result, with an additional ln term thrown in for good measure.
I’m not sure why Jackson didn’t quote this term as part of his answer; he did
include it in his answer to problem 2.17 (c). Did I do something wrong?
(b) Now we want to plug the expression for G above into (16) to compute
the potential within the cylinder. If there is no charge inside the cylinder, the
volume integral vanishes, and we are left with the surface integral:
Z
∂G dA0 .
(24)
Φ(ρ, φ) = Φ(b, φ0 )
∂ρ0 ρ0 =b
where the integral is over the surface of the cylinder.
For this we need the normal derivative of (23) on the cylinder:
∂G
1
=−
0
∂ρ
4π
2ρ0 − 2ρ cos(φ − φ0 )
2ρ2 ρ0 − 2ρb2 cos(φ − φ0 )
− 2
4
2
02
0
2
0
b + ρ ρ − 2ρρ b cos(φ − φ ) ρ + ρ02 − 2ρρ0 cos(φ − φ0 )
.
Homer Reid’s Solutions to Jackson Problems: Chapter 2
18
Evaluated at ρ0 = b this is
ρ2 − b 2
∂G 1
.
=−
∂ρ0 ρ0 =b
2π b(ρ2 + b2 − 2ρb cos(φ − φ0 ))
In the surface integral, the extra factor of b on the bottom is cancelled by the
factor of b in the area element dA0 , and (24) becomes just the result of Problem
2.12.
(c) For the exterior problem we again start with the solution (21). Now the
boundary conditions are different; the condition at ∞ gives A2m = 0, while the
condition at b gives
B1m = −γm bm .
A1m = γm b−m
From the continuity condition at ρ0 = ρ we find
m b
ρ m
−
A2m = γm ρm
.
b
ρ
The finite derivative jump condition gives
m m 1
b
b
1
ρ m
1
ρ m
− mγm
=
−
+
−mγm
b
ρ
ρ
b
ρ
ρ
ρ
or
γm = −
1
2m
m
b
.
ρ
Putting it all together we have for the exterior problem
2 m m 1
b
ρ<
gm =
.
−
2m
ρρ0
ρ>
This is the same gm we came up with before, but with b2 and ρρ0 terms flipped
in first term. But the closed-form expression was symmetrical in those two
expressions (except for the mysterious ln term) so the closed-form expression
for the exterior Green’s function should be the same as the interior Green’s
function.