Вы находитесь на странице: 1из 8

Physics 505 Fall 2007

Practice Midterm — Solutions

This midterm will be a two hour open book, open notes exam. Do all three problems.

1. A rectangular box has sides of lengths a, b and c


c y
x b

a) For the Dirichlet problem in the interior of the box, the Green’s function may be
expanded as
∞ X

X mπx mπx0 nπy nπy 0
G(x, y, z; x0 , y 0 , z 0 ) = gmn (z, z 0 ) sin sin sin sin
m=1 n=1
a a b b

Write down the appropriate differential equation that gmn (z, z 0 ) must satisfy.
Note that sin kx satisfies the completeness relation

X mπx mπx0 a
sin sin = δ(x − x0 )
a a 2

Hence the Green’s function equation

∇2x0 G(~x, ~x 0 ) = −4πδ 3 (~x − ~x 0 )

has an expansion

mπx0 nπy 0
2 0 mπx nπy
∇x0 gmn (z, z ) sin sin sin sin
a a b b
4 X mπx mπx0 nπy nπy 0
= −4πδ(z − z 0 ) sin sin sin sin
ab m,n a a b b

Working out the x0 and y 0 derivatives on the left-hand side yields

X  d2  mπ 2  nπ 2 
0 mπx mπx0 nπy nπy 0
− − gmn (z, z ) sin sin sin sin
dz 02 a b a a b b
16π X mπx mπx0 nπy nπy 0
=− δ(z − z 0 ) sin sin sin sin
ab m,n
a a b b
However, since sin kx forms an orthogonal basis, each term in this sum must
vanish by itself. This results in the differential equation
d 16π
− γmn gmn (z, z 0 ) = −
δ(z − z 0 ) (1)
dz ab
where γmn = π (m/a)2 + (n/b)2 is given in part c). Note that the Fourier sine
expansion automatically satisfies Dirichlet boundary conditions for x and y. The
remaining boundary condition is that gmn (z, z 0 ) vanishes whenever z or z 0 is equal
to 0 or c.
b) Solve the Green’s function equation for gmn (z, z 0 ) subject to Dirichlet boundary
conditions and write down the result for G(x, y, z; x0 , y 0 , z 0 ).
We may solve the Green’s function equation (1) by first noting that the homoge-
neous equation is of the form
g 00 (z 0 ) − γmn
g(z 0 ) = 0
This is a second-order linear equation with constant coefficients admitting the
familiar solution 0 0
g(z 0 ) = Aeγmn z + Be−γmn z
However, we want g(z 0 ) = 0 when z 0 = 0 or z 0 = c. This motivates us to write
out the solutions
u(z 0 ) = sinh γmn z 0 0 < z0 < z
v(z 0 ) = sinh[γmn (c − z 0 )] z < z0 < c
The Green’s function solution is then given by
Au(z 0 ) z 0 < z

gmn (z, z ) =
Bv(z 0 ) z 0 > z
The matching conditions
d d 16π
g< = g> , g< = g> +
dz 0 dz 0 ab
then give the system
A sinh γmn z = B sinh[γmn (c − z)],
A cosh γmn z = −B cosh[γmn (c − z)] +
which may be solved to yield
16π sinh[γmn (c − z)]
abγmn (cosh γmn z sinh[γmn (c − z)] + sinh γmn z cosh[γmn (c − z)])
16π sinh[γmn (c − z)] 16π
= = v(z)
abγmn sinh γmn c abγmn sinh γmn c

16π sinh γmn z

abγmn (cosh γmn z sinh[γmn (c − z)] + sinh γmn z cosh[γmn (c − z)])
16π sinh γmn z 16π
= = u(z)
abγmn sinh γmn c abγmn sinh γmn c

As a result, the full Green’s function solution is then given by

gmn (z, z 0 ) = u(z< )v(z> )
abγmn sinh γmn c
= sinh γmn z< sinh γmn [(c − z> )]
abγmn sinh γmn c


16π X 1 mπx mπx0 nπy nπy 0

G(~x, ~x 0 ) = sin sin sin sin
ab m,n γmn sinh γmn c a a b b (2)
× sinh γmn z< sinh γmn [(c − z> )]

c) Consider the boundary value problem where the potential on top of the box is
Φ(x, y, c) = V (x, y) while the potential on the other five sides vanish. Using the
Greens’ function obtained above, show that the potential may be written as

∞ X

X mπx nπy
Φ(x, y, z) = Amn sin sin sinh γmn z
m=1 n=1
a b

where γmn = π (m/a)2 + (n/b)2 and

Z a Z b
4 mπx nπy
Amn = dx dy V (x, y) sin sin
ab sinh γmn c 0 0 a b

Since we only have to worry about the potential on the top of the box (and since
we assume there is no charge inside the box), the Green’s function solution may
be written
∂G(~x, ~x 0 ) 0
Φ(~x ) = − Φ(~x 0 ) da
4π z0 =c ∂n0
x, ~x 0 ) 0 0
1 0 0 ∂G(~
=− V (x , y ) dz dy
4π z0 =c ∂n0
Noting that the outward-pointing normal n̂0 on the top of the box is in the +ẑ 0
direction, we compute the normal derivative of (2)

∂G(~x, ~x 0 ) ∂G(~x, ~x 0 )

∂n0 0 = ∂z 0 0
z =c z =c
16π X 1 mπx mπx0 nπy nπy 0
= sin sin sin sin
ab m,n γmn sinh γmn c a a b b
× sinh γmn z − γmn cosh γmn [(c − z 0 )]
z 0 =c
X 4 mπx nπy
= −4π sin sin sinh γmn z
ab sinh γmn c a b
mπx0 nπy 0
× sin sin
a b
Inserting this into (3) then straightforwardly gives the desired result. Note that
the primes may be dropped from the double integral Amn once it has been isolated
from the rest of the expression.

2. The potential on the surface of a sphere of radius a is specified by

0, 0≤θ<β
V0 V (θ, φ) = V0 , β ≤ θ ≤ π − β
0, π−β <θ ≤π

There are no other charges in this problem.

a) Show that the potential outside the sphere may be expressed as
X  a l+1
Φ(r, θ, φ) = V0 [Pl+1 (cos β) − Pl−1 (cos β)] Pl (cos θ)

where we take P−1 (x) = 0. Note that Legendre polynomials satisfy the relation
0 0
(2l + 1)Pl (x) = Pl+1 (x) − Pl−1 (x).
There are several ways of solving this problem. Perhaps the most straightforward
is to realize from azimuthal symmetry that the potential necessarily admits a
Legendre expansion
X  a l+1
Φ(~x ) = αl Pl (cos θ)
The boundary conditions at r = a gives
V (θ) = αl Pl (cos θ)

This is clearly a Legendre expansion for V (θ). The legendre orthogonality relation
allows us to write the expansion coefficients αl as
Z 1
2l + 1
αl = V (cos θ) Pl (cos θ) d(cos θ)
2 −1

For the specified potential, this becomes

Z cos β Z cos β
2l + 1 V0
αl = V0 Pl (cos θ) d(cos θ) = (2l + 1)Pl (x) dx
2 − cos β 2 − cos β

Using the Legendre relation given in the problem, we obtain

V0 cos β 0
αl = [P (x) − Pl−1 (x)] dx
2 − cos β l+1
V0 h i
= Pl+1 (cos β) − Pl−1 (cos β) − Pl+1 (− cos β) + Pl−1 (− cos β)
= (1 + (−)l )[Pl+1 (cos β) − Pl−1 (cos β)]

This vanishes unless l is even (which should be obvious from the z → −z sym-
metry of the problem). The result is then
X  a l+1
Φ(r, θ) = V0 [Pl+1 (cos β) − Pl−1 (cos β)] Pl (cos θ)
l even

Note that l = 0 is allowed, and gives the monopole contribution. (An earlier
version of this practice midterm excluded l = 0 from the sum, and this was a
This problem could also have been solved by using the Dirichlet Green’s function
outside a sphere

X 1 1 ∗
G(~x, ~x ) = 4π l
r< − l+1 l+1
Ylm (Ω0 )Ylm (Ω)
2l + 1 r< r>

After slight manipulation, we would have ended up with a similar Legendre poly-
nomial integration.
b) For fixed V0 , what angle β maximizes the quadrupole moment?
The quadrupole is given by l = 2. Hence the quadrupole moment is related to
the α2 term in the expansion. Since we do not care about normalization (we only
care to maximize the moment) it is sufficient to write

q2,0 ∼ α2 ∼ P3 (cos β) − P1 (cos β)

We extremize this quantity by taking a derivative with respect to β and setting

the result to zero. Noting that taking a derivative simply ‘undoes’ the integration,
we end up solving
∼ P2 (cos β) sin β = 12 (3 cos2 β − 1) sin β = 0

This is solved by β = 0 and β = cos−1 (±1/ 3). Clearly the quadrupole moment
vanishes when β = 0 (as the entire sphere √ is at a constant zero potential). Hence
the maximum is when β = cos (1/ 3), provided we restrict 0 ≤ β ≤ π/2.
Note that β > π/2 does not really make sense, except in a formal manner where
V0 → −V0 (corresponding to interchanging the integration limits). Technically,
the problem should have asked to maximize the magnitude of the quadrupole
moment instead of to make q2,0 as positive as possible (which would depend on
the sign of V0 ).

3. A line charge on the z axis extends from z = −a to z = +a and has linear charge
density varying as
λ0 z α ,

y λ(z) = α
−λ0 |z| , −a ≤ z < 0
−λ 0 |z |
where α is a positive constant. The total charge on the 0 < z ≤ a segment is Q (and
the charge on the −a ≤ z < 0 segment is −Q).
a) Calculate all of the multipole moments of the charge distribution. Make sure to
indicate which moments are non-vanishing.
Noting that a uniformly charged line charge on the +z axis has charge density
ρ= δ(cosθ − 1)
we see that a varying line charge yields
ρ= δ(cos θ − 1)
This may be checked by observing

3 2 dφ
dq = ρ d x = ρ r dr dφ d(cos θ) = λ(r)dr
2π cos θ=1

(Note that there is no distinction between r and z for cos θ = 1.) Hence, for the
positive and negative line charge, we have

λ0 r α
ρ= [δ(cos θ − 1) − δ(cos θ + 1)]

We may normalize λ0 by integrating from 0 to a to obtain the total charge Q

a a
λ0 aα+1
Q= λ(z) dz = λ0 z α dz =
0 0 α+1

α + 1  r α
ρ=Q [δ(cos θ − 1) − δ(cos θ + 1)]
2πar2 a
The multipole moments are then

qlm = rl Ylm (Ω)ρ r2 dr dΩ
α + 1 a l  r α
∗ dΩ
=Q r dr Ylm (Ω)[δ(cos θ − 1) − δ(cos θ + 1)]
a 0 a 2π

By azimuthal symmetry, only the m = 0 moments are non-vanishing

Z 1r
α + 1 al+1 2l + 1
ql,0 =Q Pl (cos θ)[δ(cos θ − 1) − δ(cos θ + 1)]d(cos θ)
a α + l + 1 −1 4π
α + 1 2l + 1
= Qal [Pl (1) − Pl (−1)]
α+l+1 4π
l α+1 2l + 1
= Qa [1 − (−)l ]
α+l+1 4π

Hence all moments vanish unless l is odd and m is zero. Then

α+1 2l + 1
ql,0 = Qal l odd
α+l+1 π

b) Write down the multipole expansion for the potential in explicit form up to the
first two non-vanishing terms.
The multipole expansion gives

1 X 4π Ylm (θ, φ)
Φ= qlm
4π0 2l + 1 rl+1
1 X 4π Pl (cos θ)
= ql,0
4π0 2l + 1 rl+1
l odd
2Q X α + 1  a l (4)
= Pl (cos θ)
4π0 r α+l+1 r
l odd
2Q (α + 1)a Pl (cos θ) (α + 1)a3 P3 (cos θ)
= + + ···
4π0 α+2 r2 α+4 r4
2Q (α + 1)a cos θ (α + 1)a3 5 cos3 θ − 3 cos θ
= + + ···
4π0 α + 2 r2 2(α + 4) r4

c) What is the dipole moment p~ in terms of Q, a and α?

Note that the dipole term in (4) has the form

1 2Q(α + 1)a z
4π0 α+2 r3

Comparing this with the dipole expression

1 p~ · ~x
4π0 r3

2Q(α + 1)a
p~ = ẑ
Alternatively, one could compute directly

p~ = ~xρ d x = ~xQ(α + 1) α+1 sgn(z) dz
−a a x=y=0

The z component is the only non-vanishing component

2Q(α + 1)a
pz = 2Q(α + 1) α+1
dz =
0 a α+2