1 Topic 3: Special Techniques

~ ×E
Recall that for electrostatics, because ∇ ~ = 0, we can the electric potential V (~x) such that
~ ~
E = −∇V . Then, in terms of V (~x), the field is determined by Gauss’ law, which becomes

~ ·E
∇ ~ =∇ ~ (~x)) = −∇2 V (~x) = ρ(~x) → ∇2 V (~x) = − ρ(~x) .
~ · (−∇V (1.1)
0 0
When ρ 6= 0 this is called the Poisson equation. In this topic, we will explore some of the
consequences of this equation for our broader understanding of electrostatics.

1.1 Uniqueness and the Laplace Equation

The central piece of information we will use about the Poisson equation is that the solutions are
unique. This means that if we have a strategy for finding A solution for the potential, then we
have also found THE solution for our potential. This is a non-obvious statement, as it is not true
of many other differential equations (including those you will encounter in 100B,C). Therefore,
the easiest way to digest the meaning is to show that it is true directly:
We are trying to solve the poisson equation

ρ(~x)
∇2 V (~x) = − (1.2)
0
Inside a region S with a set of boundary conditions defined on the boundary of the region B = ∂S.
We will NOT assume that S is the entire space. It could also be a finite sized region, like the
inside of a box, or a infinite region with a boundary, like the upper half plane (i.e. the region
z > 0).
Suppose now that we have two solutions to this problem V1 (~x) and V2 (~x). If this is true, then
if we define
V̄ (~x) = V1 (~x) − V2 (~x) (1.3)
then V̄ (~x) must obey the Laplace equation:
1
∇2 V̄ (~x) = ∇2 V1 (~x) − ∇2 V2 (~x) = (ρ2 (~x) − ρ(~x1 )) = 0 (1.4)
0

If the solution V1 (~x) is unique, then it should be true that V2 (~x) = V1 (~x) and therefore V̄ (~x) = 0
is the only solution to the Laplace equation on S.
It is not obvious that the Laplace equation should have no interesting solutions. However, we
can show that it is true with the following observation:

~ · (V̄ (~x)∇
∇ ~ V̄ (~x)) = ∇V
~ (~x) · ∇V
~ (~x) + V̄ (~x)∇2 V̄ (~x) = ∇V
~ (~x) · ∇V
~ (~x) = δ E
~ · δE
~ (1.5)

where we used ∇2 V̄ = 0. Integrating this equation we get

Z I Z
~ ~
dV ∇ · (V̄ (~x)∇V̄ (~x)) = ~
dAn̂ · (V̄ (~x)∇V̄ (~x)) = ~ x) · δ E(~
dV δ E(~ ~ x) (1.6)
S B S

1
This may not look like much, but notice that δ E(~ ~ x) · δ E(~
~ x) ≥ 0. To see why this is important,
suppose that we impose fixed boundary conditions so that our solutions must satisfy that V1 (~x ∈
B) = V2 (~x ∈ B). As a result, it must be true that V̄ (~x ∈ B) = 0 and therefore
Z I
~ ~
dV δ E(~x) · δ E(~x) = ~ V̄ (~x)) = 0
dAn̂ · (V̄ (~x)∇ (1.7)
S B
~ x) = 0, or in other words,
The only way this can be true is if δ E(~
~ 1 (~x) = −∇V1 (~x) = E2 (~x) = −∇V2 (~x) .
E (1.8)
The same conclusion would be true if we impose Neumann boundary conditions so that ∇V1 (~x ∈
B) = ∇V2 (~x∇) so that ∇V̄ (~x ∈ B) = 0.
We will call this result the first uniqueness theorem: The solution to Laplace’s equation
in some region S is unique if V (~x) is specified on the boundary of V , B = ∂B.

1.2 Conductors
The power of the previous result may not be entirely obvious given the problems we solved in
Topic 2. If all we have is a bunch of fixed charges, then we did just find solving for E(~~ x) or
V (~x) using the superposition principle and the field of a point charge. However, this situation
changes dramatically when we consider materials that require us to impose new kinds of boundary
conditions.
The first such material we will consider are conductors. The idea of a conductor is that it is
filled with mobile charges (electrons). Any electric field will immediately cause these charges to
move and thus would not be static. However, it is possible that eventually this the charges can
move to cancel the field so that the charges no longer feel a force and everything is again static.
We can summarize this idea as
~ x ∈ conductor) = 0 .
conductor + electrostatics =⇒ E(~ (1.9)
Therefore, if we have a conductor that fills a region S 0 and no conductor in the region S, then
our electrostatics problem becomes
ρ(~x) ~ (~x ∈ S 0 ) = 0
∇2 V (~x) = − for ~x ∈ S and ∇V (1.10)
0
The following statements all follow from the definition of a conductor in electrostatics:
~ = 0)
1. The electric field vanishes inside a conductor (E

2. There is no charge density inside a conductor ( not including the boundaries). I.e. since
~ = 0 for ~x ∈ S 0 then ∇ · E
E ~ = 0 when ~x ∈ S 0 ∩ ∂S.

3. Any net charges must reside on the surface. I.e. since ρ = 0 inside, the only place any
excess charge can go is the boundary ∂S 0

4. The potential is constant in a conductor (i.e. it is an equipotential surface)

~ ∝ n̂. If this were
5. A the surface of the conductor, the field is in the normal direction, E
not the case, there would be a component of the electric pointing along the conductor and
which would then induce a current in the conductor.

2
Example

Let’s consider the problem of describing the electric field outside a neutral conducting sphere,
that contains an asymetric cavity with a charge q inside. The arrangment is shown in figure 1.
Unlike the problem we have considered before, this does not have any nice symmetry to
exploit. The cavity and charge are not spherically symmetric. Furthermore, the cavity is not
spherically symmetric either so even the superposition principle is likely going to fail us as well.
However, something quite remarkable happens when you use the uniqueness of the solution. If
we are only solving the problem in the region r ≥ R, then because the field inside the conductor is
~ = 0, we have the same problem as if we replace the interior of the conductor with just an empty
E
cavity. As a result, the field the same as a shell of charge with a total charge q. Furthermore,
the problem outside the sphere is spherically symmetric: since the problem outside the sphere is
the same if we remove the inside, the problem outside can’t tell that the interior is asymmetric.
Therefore, our find result is that
~ > R) = q r̂
E(r (1.11)
4π0 r2
This is a remarkable fact about conductors what is qualitatively different than a collection of
charges. Our final for this conductor, our final answer is spherically symmetric, even though the
underlying problem was not. For a collection of charges, the symmetry you get out is only the
symmetry you put into the charge distribution.

y
<latexit sha1_base64="l29WxoUb9DEbvmhLG7jHtZ0OU24=">AAAB6HicbVBNS8NAEJ34WetX1aOXxSJ4KokIeix68diC/YA2lM120q7dbMLuRgihv8CLB0W8+pO8+W/ctjlo64OBx3szzMwLEsG1cd1vZ219Y3Nru7RT3t3bPzisHB23dZwqhi0Wi1h1A6pRcIktw43AbqKQRoHATjC5m/mdJ1Sax/LBZAn6ER1JHnJGjZWa2aBSdWvuHGSVeAWpQoHGoPLVH8YsjVAaJqjWPc9NjJ9TZTgTOC33U40JZRM6wp6lkkao/Xx+6JScW2VIwljZkobM1d8TOY20zqLAdkbUjPWyNxP/83qpCW/8nMskNSjZYlGYCmJiMvuaDLlCZkRmCWWK21sJG1NFmbHZlG0I3vLLq6R9WfPcmte8qtZvizhKcApncAEeXEMd7qEBLWCA8Ayv8OY8Oi/Ou/OxaF1zipkT+APn8wfnvYz9</latexit>
y
<latexit sha1_base64="l29WxoUb9DEbvmhLG7jHtZ0OU24=">AAAB6HicbVBNS8NAEJ34WetX1aOXxSJ4KokIeix68diC/YA2lM120q7dbMLuRgihv8CLB0W8+pO8+W/ctjlo64OBx3szzMwLEsG1cd1vZ219Y3Nru7RT3t3bPzisHB23dZwqhi0Wi1h1A6pRcIktw43AbqKQRoHATjC5m/mdJ1Sax/LBZAn6ER1JHnJGjZWa2aBSdWvuHGSVeAWpQoHGoPLVH8YsjVAaJqjWPc9NjJ9TZTgTOC33U40JZRM6wp6lkkao/Xx+6JScW2VIwljZkobM1d8TOY20zqLAdkbUjPWyNxP/83qpCW/8nMskNSjZYlGYCmJiMvuaDLlCZkRmCWWK21sJG1NFmbHZlG0I3vLLq6R9WfPcmte8qtZvizhKcApncAEeXEMd7qEBLWCA8Ayv8OY8Oi/Ou/OxaF1zipkT+APn8wfnvYz9</latexit>

+ +
+ - + + +

R
<latexit sha1_base64="NfFBQKCbADihv37V0PCkDJe5Zns=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8dW7Ae0oWy2k3btZhN2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4dua3n1BpHssHM0nQj+hQ8pAzaqzUuO+XK27VnYOsEi8nFchR75e/eoOYpRFKwwTVuuu5ifEzqgxnAqelXqoxoWxMh9i1VNIItZ/ND52SM6sMSBgrW9KQufp7IqOR1pMosJ0RNSO97M3E/7xuasJrP+MySQ1KtlgUpoKYmMy+JgOukBkxsYQyxe2thI2ooszYbEo2BG/55VXSuqh6btVrXFZqN3kcRTiBUzgHD66gBndQhyYwQHiGV3hzHp0X5935WLQWnHzmGP7A+fwBrKGM1g==</latexit>
<latexit
r>R
<latexit sha1_base64="Qr9t3PXndq0Rdji6V4Ungdq/ELo=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoCcJeMkxPvKAZAmzk9lkyOzsMtMrhCWf4MWDIl79Im/+jZNkD5pY0FBUddPdFSRSGHTdb6ewtr6xuVXcLu3s7u0flA+PWiZONeNNFstYdwJquBSKN1Gg5J1EcxoFkreD8e3Mbz9xbUSsHnGScD+iQyVCwSha6UHf3PfLFbfqzkFWiZeTCuRo9MtfvUHM0ogrZJIa0/XcBP2MahRM8mmplxqeUDamQ961VNGIGz+bnzolZ1YZkDDWthSSufp7IqORMZMosJ0RxZFZ9mbif143xfDaz4RKUuSKLRaFqSQYk9nfZCA0ZygnllCmhb2VsBHVlKFNp2RD8JZfXiWti6rnVr27y0qtnsdRhBM4hXPw4ApqUIcGNIHBEJ7hFd4c6bw4787HorXg5DPH8AfO5w8EOY2g</latexit>

+ + x
<latexit sha1_base64="f2yzimwbR/Dgjzp6tZ360fHRqNI=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cW7Ae0oWy2k3btZhN2N2IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4dua3H1FpHst7M0nQj+hQ8pAzaqzUeOqXK27VnYOsEi8nFchR75e/eoOYpRFKwwTVuuu5ifEzqgxnAqelXqoxoWxMh9i1VNIItZ/ND52SM6sMSBgrW9KQufp7IqOR1pMosJ0RNSO97M3E/7xuasJrP+MySQ1KtlgUpoKYmMy+JgOukBkxsYQyxe2thI2ooszYbEo2BG/55VXSuqh6btVrXFZqN3kcRTiBUzgHD66gBndQhyYwQHiGV3hzHpwX5935WLQWnHzmGP7A+fwB5jmM/A==</latexit>
<latexit

+ ~ =0
E
<latexit sha1_base64="JshK7sxcIikJx9bMJNBj0pYtSYI=">AAAB8HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0ItQFMFjBfshbSib7aRdupuE3U2hhP4KLx4U8erP8ea/cdvmoNUHA4/3ZpiZFySCa+O6X05hZXVtfaO4Wdra3tndK+8fNHWcKoYNFotYtQOqUfAIG4Ybge1EIZWBwFYwupn5rTEqzePowUwS9CUdRDzkjBorPXbHyMjtFXF75Ypbdecgf4mXkwrkqPfKn91+zFKJkWGCat3x3MT4GVWGM4HTUjfVmFA2ogPsWBpRidrP5gdPyYlV+iSMla3IkLn6cyKjUuuJDGynpGaol72Z+J/XSU146Wc8SlKDEVssClNBTExm35M+V8iMmFhCmeL2VsKGVFFmbEYlG4K3/PJf0jyrem7Vuz+v1K7zOIpwBMdwCh5cQA3uoA4NYCDhCV7g1VHOs/PmvC9aC04+cwi/4Hx8A0B8j2A=</latexit>

+ x
<latexit sha1_base64="f2yzimwbR/Dgjzp6tZ360fHRqNI=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cW7Ae0oWy2k3btZhN2N2IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4dua3H1FpHst7M0nQj+hQ8pAzaqzUeOqXK27VnYOsEi8nFchR75e/eoOYpRFKwwTVuuu5ifEzqgxnAqelXqoxoWxMh9i1VNIItZ/ND52SM6sMSBgrW9KQufp7IqOR1pMosJ0RNSO97M3E/7xuasJrP+MySQ1KtlgUpoKYmMy+JgOukBkxsYQyxe2thI2ooszYbEo2BG/55VXSuqh6btVrXFZqN3kcRTiBUzgHD66gBndQhyYwQHiGV3hzHpwX5935WLQWnHzmGP7A+fwB5jmM/A==</latexit>
<latexit

- q
<latexit sha1_base64="6st+2unI49vTKzHpFrX8LltbNi8=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cW7Ae0oWy2k3btZhN3N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4dua3n1BpHst7M0nQj+hQ8pAzaqzUeOyXK27VnYOsEi8nFchR75e/eoOYpRFKwwTVuuu5ifEzqgxnAqelXqoxoWxMh9i1VNIItZ/ND52SM6sMSBgrW9KQufp7IqOR1pMosJ0RNSO97M3E/7xuasJrP+MySQ1KtlgUpoKYmMy+JgOukBkxsYQyxe2thI2ooszYbEo2BG/55VXSuqh6btVrXFZqN3kcRTiBUzgHD66gBndQhyYwQHiGV3hzHpwX5935WLQWnHzmGP7A+fwB252M9Q==</latexit>
- ~ =0
E
+ - +
<latexit sha1_base64="JshK7sxcIikJx9bMJNBj0pYtSYI=">AAAB8HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0ItQFMFjBfshbSib7aRdupuE3U2hhP4KLx4U8erP8ea/cdvmoNUHA4/3ZpiZFySCa+O6X05hZXVtfaO4Wdra3tndK+8fNHWcKoYNFotYtQOqUfAIG4Ybge1EIZWBwFYwupn5rTEqzePowUwS9CUdRDzkjBorPXbHyMjtFXF75Ypbdecgf4mXkwrkqPfKn91+zFKJkWGCat3x3MT4GVWGM4HTUjfVmFA2ogPsWBpRidrP5gdPyYlV+iSMla3IkLn6cyKjUuuJDGynpGaol72Z+J/XSU146Wc8SlKDEVssClNBTExm35M+V8iMmFhCmeL2VsKGVFFmbEYlG4K3/PJf0jyrem7Vuz+v1K7zOIpwBMdwCh5cQA3uoA4NYCDhCV7g1VHOs/PmvC9aC04+cwi/4Hx8A0B8j2A=</latexit>

- + +
+ +

Figure 1: Electric field around a conducting sphere containing a cavity with a change q inside. Since the
~ = 0, the problem of finding the field outside the sphere is the same
field inside the conductor vanishes, E
as a shell of charge q.

Some of you may be unsatisfied with this derivation. The use of the uniqueness theorem
is indeed enough to remove the interior of the sphere. However, I made an assumption in the
process that the distribution of charge on the surface was spherically symmetric with a total
charge q. Fortunately, these are two properties that we can easily check.

3
 • Spherical Symmetry: Suppose that charge distribution was not symmetric so that the
potential is spherical coordinates, V (r, θ, φ), depends on the angles θ, φ at the surface r = R.
However, it is a property of being a conductor that the transverse electric field must vanish,
~ · ϕ̂ = E
so E ~ · θ̂ = 0. This implies

1 ∂ 1 ∂
Eθ = − V (r, θ, ϕ) = 0 Eϕ = − V (r, θ, ϕ) = 0 =⇒ V (r, θ, ϕ) = V (r),
r ∂θ r sin θ ∂ϕ
(1.12)
so that the potential can only be radial, and hence spherically symmetric.

• Surface Charge: Using Gauss’s law and the fact the conductor is neutral, we can enclose
the entire conductor with a sphere of radius r > R to find

~ =q
Z
dAr̂ · E (1.13)
Sr>R 

We can add a second surface just inside with r0 = R −  < R so that we pick up only the
surface charge
~ = qsurface
Z Z
~
dAr̂ · E + dA(−r̂) · E (1.14)
Sr>R Sr0 <R 

Now we notice that since E(r~ 0 < R) = 0 the second term does contribute so that

qsurface ~ = q =⇒ qsurface = q . (1.15)

Z Z Z
~
dAr̂ · E + ~
dA(−r̂) · E = = dAr̂ · E
Sr>R Sr0 <R  Sr>R 

The take away from this example is that conductors screen change. The problem of finding
the electric field in a region that is separated from a charge by a conductor is that same as solving
the problem as if there was no change there. This is what enables us to screen electric fields via
Faraday cages, namely if you surround yourself with a conductor, the field inside the cage will
be zero as if you were inside a solid piece of conductor. This strategy can be effective even for a
mesh of conductor, not even with a solid walls.
This example generalizes to the second uniqueness theorem: The solution to Laplace’s
equation in some region V bounded by conductors is uniquely determined by the charge on each
conductor.
The proof of this claims is just repeating the previous argument: if I know the total charge
on a conductor, then field outside the conductor is the same as a shell with the same total change
and no transverse potential. As a result, we can use the first uniqueness theorem to determine
the field outside the conductors is unique.

Grounded Conductors

In the previous example, we induced a conductor with the requirement that the total charge was
zero. As a result, we had screened the charge q on one boundary, but had to induce another
surface charge on the other boundary to keep the total charged fixed. An alternate situation that
arises in many physical situations is a grounded conductor, where the conductor is connected by
a wire to a larger disconnected system. Since the potential inside a conductor is constant, the

4
entire conductor aquires the same potential as our distant point, which we can simply call V = 0.
However, in the process, this enables our conductor to change its total charge, thus giving us
different results.
To see how this works, imagine that we consider the same problem as we saw in Figure 1 but
now with the requirement that V (~x) = 0 inside the conductor. We are now solving the problem

∇2 V (~x) = 0 V (r = R) = 0 =⇒ V (r > R) = 0 ~ > R) = 0

E(r (1.16)

In other words, we now have no more surface charge on the outside and hence no electric field.
The total charge of the system is therefore q + qconductor = 0 and therefore we have a total charge
of −q on the conductor, localized to the interior boundary.

1.3 Method of Images

When solving boundary value problems with unique solutions, one can use tricks that generate a
valid solution that look like a different physics problem. This is not limited to electrostatics, but
comes up in many different physical situations. The method of images a very common example
of such a strategy.
The best way to understand the method of images is with an example. Suppose we are solving
for the electric field of a charge q located at x = y = 0 and z = z0 > 0, above an infinite grounded
conductor that covers the entire x-y plane at z = 0. This problem is tricky, but you realize your
only job is to find a potential V (x, y, z) that satisfies
q
∇2 V (x, y, z) = − δ(x)δ(y)δ(z − z0 ) for z > 0 and V (z = 0) = 0 . (1.17)
0
Now consider the potential due to TWO charges, on with +q at z = z0 and other with −q at
z = −z0 . The potential for this charge configuration is
!
q 1 1
V (x, y, z) = p −p =0 (1.18)
4π0 x2 + y 2 + (z − z0 )2 x2 + y 2 + (z + z0 )2
We notice that when z = 0
!
q 1 1
V (x, y, z = 0) = p −p =0 (1.19)
4π0 x2 + y 2 + (−z0 )2 x2 + y 2 + (z0 )2
This charge configuration solves our boundary condition. At face value, it doesn’t look like it
solves the right equation
q q
∇2 V (x, y, z) = − δ(x)δ(y)δ(z − z0 ) + δ(x)δ(y)δ(z + z0 ) (1.20)
0 0
But since δ(z + z0 ) = 0 for all z > 0, in the region z > 0, this is actually just
q
∇2 V (x, y, z) = − δ(x)δ(y)δ(z − z0 ) for z > 0 (1.21)
0
Therefore, we see that this charge configuration generates a V (x, y, z > 0) that satisfies our
equations and boundary conditions. Now, for z < 0, it is important that this image charge
doesn’t exist. There we are solve

∇2 V (x, y, z) = 0 for z < 0 and V (z = 0) = 0 . (1.22)

5
and therefore the only solution is V (x, y, z) = 0 for z < 0. Altogether, we have
  
 q

√ 2 2 1
−√ 2 2 1
z>0
V (x, y, z) = 4π0 x +y +(z−z0 )2 x +y +(z+z0 )2 (1.23)
0 z≤0


This is another example of a conductor that screen the charge. On the other side, z < 0, there
is no evidence of the +q charge and the electric field is completely cancelled out.
Importantly, the image charge doesn’t exist. That is responsible for the apparent change is an
accumulation of charge on the edge of the conductor, which shields it from the electric field. We
can calculate what this surface charge is using Gauss’s law. Our electric field is also determined
by the image charge, so we have
  
 q (x,y,z−z0 ) (x,y,z+z0 )
− z>0
~
E(x, y, z) = 4π0 (x +y +(z−z0 ) )
2 2 2 3/2 (x2 +y 2 +(z+z0 )2 )( 3/2)
(1.24)
0 z<0

Notice specifically, at this we talk E(x, y, z = 0+ ) (where 0+ means z =  > 0 and then take the
limit  → 0) that
−2q
 
~ + (0, 0, z0 )
E(x, y, 0 ) = (1.25)
4π0 (x2 + y 2 + (z0 )2 )3/2
On the other hand, when z = 0− , we just have E ~ = 0. Integrating over a tiny pillbox that crosses
z = 0 we have
−2q
 
z0 1 1
Z Z
~ + ~ −
dAẑ E(x, y, 0 )+ dA(−ẑ)E(x, y, 0 ) = A = Qenclosed = Aσ(x, y)
4π0 (x2 + y 2 + (z0 )2 )3/2 0 0
(1.26)
where A is the area of the cross-section of the pillbox and we assumed A is small enough that E ~
is effectively constant in the x and y directions. We find that the charge per unit area is given by
−2q
 
z0
σ(x, y) = (1.27)
4π (x2 + y 2 + (z0 )2 )3/2

Conducting Sphere

Any discussion of the method of images requires one demonstration that is a bit shocking. A
popular such example starts with the following observation: take two charges q and q 0 and place
them on the positive z axis a distance a and b from the origin. The electric potential from these
charges can be written in spherical coordinates as
q 1 q0 1
V (r, θ, ϕ) = √ + √ (1.28)
4π0 r2 + a2 − 2ra cos θ 4π0 r2 + b2 − 2rb cos θ
Now we choose q 0 = −Rq/a and b = R2 /a. Plugging in, we get
!
q 1 1
V (r, θ, ϕ) = √ − a
p (1.29)
4π0 2 2
r + a − 2ra cos θ R r2 + R4 /a2 − 2rR2 cos θ/a
!
q 1 1
= √ −p (1.30)
4π0 r2 + a2 − 2ra cos θ (ra/R)2 + R2 − 2ra cos θ

6
Now we can observe that when r = R, we have
 
q 1 1
V (r = R, θ, ϕ) = √ −√ =0 (1.31)
4π0 R2 + a2 − 2Ra cos θ a2 + R2 − 2Ra cos θ
As a result, if we want to find the electric potential outside (or inside) a conducting sphere in
the presence of a change q (or q 0 ), we have determined where to put the image charge to impose
the condition that a conducting sphere of radius R is grounded.
There are many more such examples. There is a mathematical physics literature where
strategies for finding these solutions are developed using complex analysis. You are not expected
to be able to come up with these clever examples on your own, but given one idea you should
be able to use it in other circumstances where the same basic idea is applicable. Specifically, we
have seen the idea of how to solve for the field in the presence of one infinite conducting plane
and you should be able to use that idea to solve problems involving multiple infinite conducting
planes.

1.4 Separation of Variables in Cartesian Coordinates

We are now ready to attack the problem of finding the potential head-on. We are going to solve
the electric potential in a region with no changes,
∂2 ∂2 ∂2
∇2 V (x, y, z) = V (x, y, z) + V (x, y, z) + V (x, y, z) = 0 (1.32)
∂x2 ∂y 2 ∂z 2
subject to a set of boundary conditions. As we have seen already, the boundary conditions are
what sets that strategy for solving a problem as much as the equation. To see why this is, let us
first assume that we have a specified the potential on the boundaries of a infinitely long box:

V (x = 0, L, y, z) = V0 V (x, y = 0, L, z) = 0 z ∈ [−∞, ∞] (1.33)

Because the boundary conditions are easily state in Cartesian coordinates. So, we want to solve
the equations in Cartesian coordinates as well.
Now we notice something nice: Let’s guess that we have a solution that can be written in the
form
V (x, y, z) = m(x)n(y)p(z) . (1.34)
Then we have

∇2 V (x, y, z) = m00 (x)n(y)p(z) + m(x)n00 (y)p(z) + m(x)n(y)p00 (z) = 0 (1.35)

If we divide by V (x, y, z) = m(x)n(y)p(z) we have

∇2 V (x, y, z) m00 (x) n00 (y) p00 (z)
= + + =0 (1.36)
V (x, y, z) m(x) n(y) p(z)
Notice that each term is a function of a different viable. Since no term depends on the same
variable, it must be true that none of these terms depend on x,y, or z. In other words, each term
must be a constant:
m00 (x) n00 (y) p00 (z)
= c1 = c2 = c3 = −c1 − c2 (1.37)
m(x) n(y) p(z)

7
We have succeed in reducing a partial differential equation (PDE) we don’t know how to solve
with 3 ordinary differential equations (ODEs) we can imagine solving.The part that we haven’t
addressed is that we know nothing about c1 and c2 other than the fact that they are constant.
They could be real, imaginary or complex.
First, let’s impose the obvious symmetry we have in the problem. The box is infinitely long
and the boundary conditions are independent of z. For this specific problem, we have a z → z + c
symmetry, and therefore p(z) = constant and c3 = p00 /p = 0. As a result, we have c2 = −c1 and
therefore we have one unknown constant. The two remaining equations are

m00 (x) n00 (y)

= c1 = −c1 (1.38)
m(x) n(y)

Let’s start with the equation for y as those boundary conditions are slightly easier to impose.
We need to have
V (x, y = 0, z) = V (x, y = L, z) = 0 ∀x ∈ [0, L] . (1.39)
We can accomplish this easily if n(y = 0) = n(y = L) = 0 (since it multiplies m(x), making the
solution vanish at those boundary). This is really why the separation of variables is effective: we
have a problem where at least some of the boundary conditions conditions separate as well, so
that it really just becomes an ODE in certain directions.
Given the boundary conditions for n(y) we now set out to find viable solutions. We follow
the usual strategy of making an educated guess that n(y) = Aeλy :
√
n00 (y) + c1 n(y) = 0 n(y) = Aeλy → Aeλy (λ2 + c1 ) = 0 → λ = ± −c1 (1.40)

We have two possible options c1 > 0 or c1 < 0. Let’s first assume c1 = −`2 < 0 and see what
happens. Recall that the superposition principles holds, so we can add all possible solutions to
these equations:
n(y) = Ae`y + Be−`y (1.41)
Now we want to find the allowed values of A, B and ` by imposing boundary conditions. We
need to have
V (x, y = 0, z) = V (x, y = L, z) = 0 ∀x ∈ [0, L] . (1.42)
We can accomplish this easily if n(y = 0) = n(y = L) = 0 (since it multiplies m(x), making the
solution vanish at those boundary). Imposing the conditions we get

A+B =0 Ae`L + Be−`L = 0 → B = −A Ae−`L (e2`L − 1) = 0 → A = 0 (1.43)

In short, c1 < 0 doesn’t give us solution that can satisfy our boundary conditions. We can try
again with c1 = 0:
n00 (y) = 0 → n(y) = Ay + B , (1.44)
but again when we impose boundary conditions we find A = B = 0.
Our last hope is to take c1 = k 2 > 0 so that λ = ±ik. We then have

n(y) = Aeiky + Be−iky (1.45)

8
Solving the same boundary conditions gives us
qπ
A+B =0 AeikL + Be−ikL = 0 → B = −A 2iA sin(kL) = 0 → k = q = 1, 2, 3, ...
L
(1.46)
These are interesting solutions. Note that we still haven’t specified A.
Given the solution for n(y) we can go back and solve m(x)
m00 (x) − k 2 m(x) = 0 m(x) = Cekx + De−kx = Ceqπx/L + De−qπx/L (1.47)
Imposing the boundary conditions is now non-trivial for V0 6= 0. Specifically, if we were to simply
impose it is not enough to impose m(0) = V0 , because we still have a function of y that multiplies
it. Our strategy requires that the sum of all the possible y-dependent functions is somehow
y-independent. So we write the most general solution consitent with the y-boundary conditions.
∞
X
V (x, y) = (Cq eqπx/L + Dq e−qπx/L ) sin(qπy/L) (1.48)
q=1

and require that

∞
X
V (x = 0, y) = (Cq + Di ) sin(qπy/L) = V0 (1.49)
q=1
∞
X
V (x = L, y) = (Cq eqπ + Dq e−qπ ) sin(qπy/L) = V0 (1.50)
q=1

In order to make sense of what to do, we need to recall a few integrals: for an integers p, q 6= 0,
Z L Z L
L L L L
dy cos pπy/L = sin pπy/L|0 = 0 dy sin qπy/L = − cos qπy/L|L 0 = (1−(−1)q )
0 πp 0 πq πq
(1.51)
Now using the trig-identity
1
sin(qπy/L) sin(q 0 πy/L) = cos((q − q 0 )πy/L) − cos((q + q 0 )πy/L)


(1.52)
2
we find for positive integers q, q 0 > 0
(
L
q = q0
Z L
dy sin(qπy/L) sin(q 0 πy/L) = 2 (1.53)
0 0 q 6= q 0
This is remarkable useful formula, as we can integrate both side of our previous equation as
follows:
Z L Z L
0
sin(q πy/L)V (x = 0, y) = = sin(q 0 πy/L)V0 (1.54)
0 0
Z a ∞
X Z a
sin(q 0 πy/L) (Cq + Dq ) sin(qπy/L) = sin(q 0 πy/L)V0 (1.55)
0 q=1 0
∞ a
L
X Z
(Cq + Dq )δq,q0 = sin(q 0 πy/L)V0 (1.56)
2 0
q=1
L L 0
(Cq0 + Dq0 ) = (1 − (−1)q )V0 (1.57)
2 πq 0

9
so that when q is odd
4V0
Cq + D q = . (1.58)
πq
Repeating the same integral for x = L then gives
4V0
(Cq eqπ + Dq e−qπ ) = (1.59)
πq
Taking the difference between these equations gives
eqπ/2 − e−qπ/2
Cq (eqπ − 1) = −Dq (e−qπ − 1) → Dq = −Cq eqπ = Cq eqπ (1.60)
e−qπ/2 − eqπ/2
and plugging this into our equation for the boundary condition at x = L, we get
(
4V0
qπ q = odd
Cq = πq(1+e ) (1.61)
0 q = even
And therefore, the full solution is given by
∞
4V0 1 1 1
(eqπ(x− 2 )/L + e−qπ(x− 2 )/L ) sin(qπy/L)
X
V (x, y) = −qπ/2 qπ/2
(1.62)
πq e +e
q=odd

It practice, these kinds of answers are reasonably well approximated by the first few terms in the
sum but rarely have nice closed form expression that doesn’t involve the sum.

Generalizations and Orthogonal Functions

We can consider a related problem, that on the surface looks harder, but is actually easier where
we have an infinite z direction but also x ∈ [0, ∞) with the boundary conditions
V (x = 0, L, y, z) = V0 (y) V (x → ∞, y, z) → 0 V (x, y = 0, L, z) = 0 (1.63)
The boundary conditions for y is unchanged, so we still have the solution
X∞
V (x, y) = (Cq eqπx/L + Dq e−qπx/L ) sin(qπy/L) (1.64)
q=1

Now we need to impose that as x → ∞, V (x, y) → 0 and therefore Cq = 0. Now we are left with
boundary condition
∞
X
V (x = 0, y) = Dq sin(qπy/L) = V0 (y) (1.65)
q=1
But we simply repeat what we did before to find
Z L Z L
sin(q 0 πy/L)V (x = 0, y) = = sin(q 0 πy/L)V0 (y) (1.66)
0 0
Z a ∞
X Z a
sin(q 0 πy/L) Dq sin(qπy/L) = sin(q 0 πy/L)V0 (y) (1.67)
0 q=1 0
∞ a
L
X Z
Dq δq,q0 = sin(q 0 πy/L)V0 (y) (1.68)
2 0
q=1
a
L
Z
Dq 0 = sin(q 0 πy/L)V0 (y) (1.69)
2 0

10
In principle, this gives us a recipe for finding Dq0 but how do we know that this procedure
reproduces V0 (y)?
The answer is that what we are really doing is claiming that we can write a function in terms
of a infinite series of other functions
∞
X
V0 (y) = aq sin(qπy/L) . (1.70)
q=1

This is a remarkably common strategy you will learn more about in Physics 105a. The basic idea
is to think of functions the same way we think of a vector, just living in an infinite number of
dimensions. For a vector, we think of
~ = Ex x̂ + Ey ŷ + Ez ẑ
E x̂ · ŷ = 0 = x̂ · ẑ = ŷ · ẑ (1.71)

This means I can pick out the component Ex by just taking the dot product x̂ · E.~
The idea with functions is that the dot product is replaced with integration: imagine I have
an infinite set of function fn (x) where n = 1, 2, 3, ... defined so that
Z L
dxfn (x)fm (x) = δn,m ≡ fn · fm (1.72)
0

We will call this “a basis of orthogonal functions” the same way that x̂ and ŷ are a orthogonal
basis of vectors. So now, IF I can represent a function G(x) as
∞
X Z L
G(x) = gn fn (x) → gn = dxfn (x)G(x) (1.73)
n=1 0

There are two magical facts that you will encounter in Phys 105a and in the rest of your physics
courses:

1. The list of solutions to most ordinary differential equations automatically furnish a basis
of orthogonal functions.

2. For most functions we care about, these series representations converge so that it is exactly
true that G(x) = ∞
P
n=1 gn fn (x)

The case where fn (x) = sin(nπx/L) is the one of the most common examples of such a decom-
position (known as the fourier series). However, we will encounter more examples in this course
and you will see any more in electromagnetism, quantum mechanics, etc.

1.5 Separation of Variables in Spherical Coordinates

While separation of variables in the case of Cartesian coordinates is a bit more obvious, the same
procedure in spherical coordinates will be far more useful (except on the occasion where you want
to study electric fields confined to rectangular regions). We again return to study the Laplace
equation, ∇2 V (r, θ, ϕ) = 0 in r, θ and ϕ:

1 ∂2 2 1 ∂ ∂ 1 ∂2
(r V (r, θ, ϕ)) + (sin θ V (r, θ, ϕ)) + V (r, θ, ϕ) = 0 (1.74)
r2 ∂r2 r2 sin θ ∂θ ∂θ r2 sin2 θ ∂ϕ2

11
We will again make a guess that V (r, θ, ϕ) = g(r)p(θ)q(ϕ) so that

g 00 + 2g 0 /r p00 + p0 / tan θ q 00
+ + =0 (1.75)
g r2 p sin θ2 r2 q

First we multiply by r2 to get

r2 g 00 + 2rg 0 p00 + p0 / tan θ q 00

 
=− + (1.76)
g p sin θ2 q
So the LHS and RHS must be equal to some constant L.

Angular Equations

Now focusing on the angular part, we have

p00 + p0 / tan θ q 00 00 0
2 p − p / tan θ 2 q 00
+ = −L → sin θ + L sin θ = − (1.77)
p sin θ2 q p q

We again equate this to a constant, m2 so that

q 00
− = m2 → q 00 + m2 q = 0 → q = eimϕ (1.78)
q

The choice of this constant −m2 makes sense because our function q(ϕ) has to be periodic in ϕ:

q(ϕ + n2π) = q(ϕ) → m = 0, ±1, ±2, ... (1.79)

We see that m is a integer as one would expect.

Now the tricky part, our equation for p(θ) is

p00 + p0 / tan θ
sin2 θ + L sin2 θ = m2 (1.80)
p
It is not obvious what to do with this. But it is useful to remember that the volume element in
three dimensions is r2 sin θdrdθdϕ → r2 drd cos θdϕ, which suggests maybe cos θ is a more useful
variable than θ itself. Using ∂cos θ = − sin1 θ ∂θ we have
1
∂θ (sin θ∂θ f ) = ∂cos θ (1 − cos2 θ)∂cos θ (1.81)
sin θ
If we define c = cos θ ∈ [−1, 1] then our equation becomes

m2
∂c (1 − c2 )∂c p(c) + Lp(c) − p(c) = 0 (1.82)
1 − c2
That this point, we have reached the point where, for now, we are going to accept that a miracle
occurs and we can write the solutions to this equation in terms of functions that look familiar.
You will learn more about where they come from in Physics 105a.
Demanding that the solutions to these equations make sense requires that L = `(` + 1) where
` = 0, 1, 2, 3.... Specifically, we are assuming that there is no specific point on the surface of the
sphere (with r > 0) where V (r, θ, φ) will diverge. Without this assumption, we find answer that

12
are infinite at cos θ = ±1, but we have no reason to expect anything special to happen at those
points (since we can always rotate the sphere to make those point a different value of the angles).
With this constraint (much like our periodicity constraint that told us m was an integer), our
solutions are
p(c) = P`m (cos θ) (1.83)
where P`m are called the associated Legendre polynomials. We will mostly be interested in the
case where the m = 0 (V (r, θ, ϕ) → V (r, θ)) and

P`m (cos θ) = P` (cos θ) (1.84)

are called the Legrendre polynomials. These functions can be computed using the Rodrigues’
formula:
1 ` 2
P` = P`m=0 = ∂ (c − 1)` (1.85)
2 `! c
`
(−1)m
P`m (c) = (1 − c2 )m/2 ∂c`+m (c2 − 1)` (1.86)
2` `!
Clearly P` is just a polynomial of order `. Here are the first few polynomials:
1
P0 (cos θ) = 1 P1 (cos θ) = cos θ P2 (c) = (3 cos2 θ − 1) (1.87)
2
Using the definition, one finds these form a basis of orthogonal function, now on cos θ ∈ [−1, 1]
1
2
Z
d cos θP` (cos θ)P`0 (cos θ) = δ`,`0 . (1.88)
−1 2` + 1

You could imagine arriving at the Legrendre polynomials a much less clever way, just by building
a set of polynomials interatively. This is what you would get if you thought P` (c) = `n=0 dn cn
P

and then, starting from ` = 0 and increasing ` one at a time, fix the dn by requiring the integral
with the lower order polynomials vanish.
The second part of this formula also makes sense if you remember that sin θ = (1 − c2 )1/2
(remember that θ ∈ [0, π] so sin θ is always positive. The associate Legendre polynomials are just
the Legrendre polynomials (P` (cos θ)) where we are switching out cosines and replacing them
with sines. If you think of this a basis for possible angular dependence, this makes a lot of sense.
If and when you talk about this in quantum mechanics, this will make even more sense as you
can show that m 6= 0 comes from m = 0 by some kind of rotation.
In terms of the associated Legendre polynomials, P`m (cos θ), we have the following relation

(` − m)! m
P`−m (c) = (−1)m P (c) (1.89)
(` + m)! `

and the orthogonality property

Z 1
2(` + m)!
dcP`m (c)P`m
0 (c) = δ`,`0 (1.90)
−1 (2` + 1)(` − m)!

13
With this information, we have enough to define the spherical harmonics:
s
2` + 1 (` − m)! m
Y`,m = P (cos θ)eimϕ (1.91)
4π (` + m)! `
The normalization was chosen to so that
Z
d cos θdϕY`,m Y`∗0 ,m0 = δ`,`0 δm,m0 (1.92)

This basically just combines known properties of the individual function eimϕ and P`m (cos θ).
Notice that
Y`∗0 ,m0 = (−1)m Y`0 ,−m0 (1.93)
This relation is somewhat trivial because we used
s
(` − m)! m (` − (−m))! −m
P` (c) = P (c) (1.94)
(` + m)! (` + (−m))! `
so complex conjugation is just changing the phase and we are just relabeling everything else in
terms of −m, but they were real functions to start.
Now recall that the spherical harmonic were defined so that
~ 2 Y m (cos θ, ϕ) = −`(` + 1)Y m (cos θ, ϕ)
∇ (1.95)
` `

This is important because (a) it doesn’t depend on m and (b) `(` + 1) is the eigenvalue that will
appear in the radial equation.

Radial Equation

Now we are left with the radial profile, which solves the equation:
r2 g 00 + 2rg 0
= L = `(` + 1) (1.96)
g
Remember that m does not appear in this equation because it only appeared as a constant in
separating θ and ϕ. Our equations is therefore:

r2 g 00 + 2rg 0 − `(` + 1)g = 0 (1.97)

where we know that ` = 0, 1, 2....

This is the kind of equation that is easy to solve if we guess it is a power law g(r) = ra :

rρ (a(a − 1) + 2a − `(` + 1)) = 0 → a = `, −` − 1 (1.98)

So our full solution is given by

∞
X X
V (r, θ, ϕ) = (c`,m r` + d`,m r−`−1 )Y`m (cos θ, ϕ) (1.99)
`=0 m=0,±1,..

We will be particularly interested in the case where only m = 0 contributes, such that
∞
X
V (r, θ) = (c` r` + d` r−`−1 )P` (cos θ) (1.100)
`=0

14
Examples

Example 1: Imagine we are given a sphere of radius R where the potential is held at

V (r = R, θ) = V0 cos 2θ = V0 (2 cos2 θ − 1) (1.101)

First let’s find the potential outside, r > R. So we have

∞
X
V (r, θ) = (c` r` + d` r−`−1 )P` (cos θ) V (R, θ) = V0 (2 cos2 θ − 1) (1.102)
`=0

First, we require that V (r → ∞) → 0 so c` = 0. Next we have to impose the boundary condition

at r = R so
X∞
d` R−`−1 P` (cos θ) = V0 (2 cos2 θ − 1) (1.103)
`=0

Integrating both sides by P`0 (cos θ)

and using
Z 1 1
2
Z
d cos θP` (cos θ)P`0 (cos θ) ≡ dcP` (c)P`0 (c) = δ`,`0 . (1.104)
−1 −1 2` + 1
we have
1
2 1
Z
−1
2d0 R = dcV0 (2c2 − 1) = − V0 → d0 = − RV0 (1.105)
−1 3 3
Z 1
2
d1 R−2 = dccV0 (2c2 − 1) = 0 → d1 = 0 (1.106)
3 −1
Z 1
2 1 8 4
d2 R−3 = dc (3c2 − 1)V0 (2c2 − 1) = V0 → d2 = R3 (1.107)
5 −1 2 15 3
(1.108)

with d`>2 = 0, where we used

1
P0 (cos θ) = 1 P1 (cos θ) = cos θ P2 (c) = (3 cos2 θ − 1) (1.109)
2
Using these results, our solution for r > R is
R3
 
1R 2 2
V (r ≥ R, θ) = V0 − + (3 cos θ − 1) 3 (1.110)
3r 3 r
Where we notice that when r = R we have
 
1 2
V (r = R, θ) = V0 − + (3 cos θ − 1) = V0 (2 cos2 θ − 1) .
2
(1.111)
3 3
In this way, we could have determined the coefficients d0 and d2 just by matching the terms in
the polynomial.
Now we can repeat this for the interior, r < R. We again have
∞
X
V (r, θ) = (c` r` + d` r−`−1 )P` (cos θ) V (R, θ) = V0 (2 cos2 θ − 1) (1.112)
`=0

15
We impose that the solution inside is finite, so we must have d` = 0 to prevent the solution from
diverging at r = 0, so now we are solving
∞
X
V (r, θ) = c` r` P` (cos θ) V (R, θ) = V0 (2 cos2 θ − 1) (1.113)
`=0

Now we notice that the integrals to determine c` will be exactly the same so we can just write
down the answer
1 4
c0 = − V0 c2 = R−2 (1.114)
3 3
so that
r2
 
1 2
V (r ≤ R, θ) = V0 − + (3 cos2 θ − 1) 2 (1.115)
3 3 R
Which again agrees with our boundary conditions.
Example 2: A neutral conducting sphere of radius R is placed in a uniform electric field pointing
~ = E0 ẑ. We would like to find the field for r > R after we introduce the
in the z-direction, E
conducing sphere.
The key idea here is that the electric field introduces a charge distribution on the conducting
sphere which, in turn, modifies the electric field. The key idea in this problem is that because of
the uniform electric field, the potential at infinity is no longer zero. Instead, when we are very
far from the sphere, it should look like our undisturbed constant field

~  R) → E0 ẑ
E(r V (r  R) → −E0 z = −Er cos θ (1.116)

so that −∇V~ = E0 ẑ. In the last line, we used the formula z = r cos θ from the definition of
spherical coordinates. We know the field inside the conductor is

~ < R) →→ V (r < R) = constant

E(r (1.117)

Since the overall constant value of V (~x) is no longer important (normally we use our freedom to
choose this constant to set V (r  R) → 0), we can choose this constant so thatb

V (r ≤ R) = 0 . (1.118)

Therefore, the problem we are solving is

∇2 V (r, θ, ϕ) = 0 V (r = R, θ, ϕ) = 0 V (r → ∞, θ, ϕ) = −E0 r cos θ (1.119)

Since our function is independent of ϕ, we can use our m = 0 solution to the equation
∞
X
V (r, θ) = (c` r` + d` r−`−1 )P` (cos θ) (1.120)
`=0

Now all we have to do is impose boundary conditions: first we impose V (r = R, θ) = 0 so

∞
X
(c` R` + d` R−`−1 )P` (cos θ) = 0 (1.121)
`=0

16
We can integrate both sides to see
∞
2
Z
0 0
X
0
d cos θP`0 (cos θ) (c` R` + d` R−`−1 )P` (cos θ) = (c`0 R` + d` R−` −1 ) = 0 (1.122)
2` + 1
`=0

or
0 0
c`0 R` = −d`0 R−` −1 (1.123)
Now we impose that
∞
X
V (r → ∞, θ) = c` r` P` (cos θ) = −E0 r cos θ (1.124)
`=0

Integrating both sides again gives

∞
2
Z X
d cos θP`0 (cos θ) c` r` P` (cos θ) = c`0 r` = −δ`0 ,1 E0 r (1.125)
2`0 + 1
`=0

where we used cos θ ≡ P1 (cos θ). Now we have

c`0 6=1 = 0 c1 = −E0 d1 = E0 R3 . (1.126)

Putting this altogether, we get

 3 !
R
V (r ≥ R, θ) = −E0 r cos θ 1 − . (1.127)
r

1.6 Multipole Expansion

Electric Potential of a Dipole

We will start by considering the electric potential far away from a configuration of two equal and
opposite charges, a dipole, shown on the left side of Figure 2. The two charges are separated by
a distance d. The potential at a point ~r at a distance r from the midpoint of the two charges is
given by  
q 1 1
V (r) = − (1.128)
4π0 r+ r−
where r± are the distances to the ±q charges. We can think of the line pointing from +q to the
point r as a vector ~r+ (and similarly for −q). As we have done many times, we notice that ~r is
the sum of the vector ~r+ and the vector from the midpoint to +q. Therefore:

d 2 d d d2
~r = ẑ + ~r+ → r+ = ~r+ · ~r+ = (~r − ẑ) · (~r − ẑ) = r2 + − rd cos θ (1.129)
2 2 2 4
Similarly
2 d2
r− = r2 + + rd cos θ (1.130)
4
Now if we take r  d, then we have
2
r± ≈ r2 ∓ rd cos θ (1.131)

17
and
1 1 1 ∓rd cos θ
≈ − . √ (1.132)
∓ rd cos θ r 2 r3 r2
Applying this to the potential, we find
 
q 1 1 q d cos θ
V (r  d) ≈ √ −√ ≈ (1.133)
4π0 2
r − rd cos θ 2
r + rd cos θ 4π0 r2

Notice that this is the ` = 1 term in the expansion of our radial solution to the Laplace equation
with the boundary conditions V (r → ∞) → 0:
∞
X
V (r, θ) = d` r−`−1 P` (cos θ) (1.134)
`=0

At long distances, the leading contribution to the potential from this dipole is just a ` = 1 term
with no contribution at ` = 0, which is what we get from just a single point change or a monopole.
The pattern continues to more complicated charge configurations illustrated on the right side of
Figure 2

+q
+q
<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

r+
+q
<latexit sha1_base64="36dJZP7DvlT2hWRo9ey+BCTdDvo=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBZBEEoiBT0WBPFY0X5AG8pmu2mXbjZhdyKU0J/gxYMiXv1F3vw3btsctPXBwOO9GWbmBYkUBl332ymsrW9sbhW3Szu7e/sH5cOjlolTzXiTxTLWnYAaLoXiTRQoeSfRnEaB5O1gfDPz209cGxGrR5wk3I/oUIlQMIpWetD9i3654lbdOcgq8XJSgRyNfvmrN4hZGnGFTFJjup6boJ9RjYJJPi31UsMTysZ0yLuWKhpx42fzU6fkzCoDEsbalkIyV39PZDQyZhIFtjOiODLL3kz8z+umGF77mVBJilyxxaIwlQRjMvubDITmDOXEEsq0sLcSNqKaMrTplGwI3vLLq6R1WfXcqndfq9Rv8ziKcAKncA4eXEEd7qABTWAwhGd4hTdHOi/Ou/OxaC04+cwx/IHz+QP6mY2Y</latexit>

<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

q
r
<latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>

<latexit sha1_base64="mn7csi4iWmxMmPCldZXDs+IFgTA=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lEqMeCIB5bsB/QhrLZTtq1m03Y3Qgl9Bd48aCIV3+SN/+N2zYHbX0w8Hhvhpl5QSK4Nq777RQ2Nre2d4q7pb39g8Oj8vFJW8epYthisYhVN6AaBZfYMtwI7CYKaRQI7AST27nfeUKleSwfzDRBP6IjyUPOqLFSUw3KFbfqLkDWiZeTCuRoDMpf/WHM0gilYYJq3fPcxPgZVYYzgbNSP9WYUDahI+xZKmmE2s8Wh87IhVWGJIyVLWnIQv09kdFI62kU2M6ImrFe9ebif14vNeGNn3GZpAYlWy4KU0FMTOZfkyFXyIyYWkKZ4vZWwsZUUWZsNiUbgrf68jppX1U9t+o1ryv1uzyOIpzBOVyCBzWowz00oAUMEJ7hFd6cR+fFeXc+lq0FJ585hT9wPn8A3lWM+g==</latexit>
<latexit

Monopole Dipole
✓
<latexit sha1_base64="XUikHvXBDyRHGULLBn0pna3R+C4=">AAAB7XicbVDLSgNBEJyNrxhfUY9eBoPgKeyKoMeAIB4jmAckS5iddJIxszPLTK8QlvyDFw+KePV/vPk3TpI9aGJBQ1HVTXdXlEhh0fe/vcLa+sbmVnG7tLO7t39QPjxqWp0aDg2upTbtiFmQQkEDBUpoJwZYHEloReObmd96AmOFVg84SSCM2VCJgeAMndTs4giQ9coVv+rPQVdJkJMKyVHvlb+6fc3TGBRyyaztBH6CYcYMCi5hWuqmFhLGx2wIHUcVi8GG2fzaKT1zSp8OtHGlkM7V3xMZi62dxJHrjBmO7LI3E//zOikOrsNMqCRFUHyxaJBKiprOXqd9YYCjnDjCuBHuVspHzDCOLqCSCyFYfnmVNC+qgV8N7i8rtds8jiI5IafknATkitTIHamTBuHkkTyTV/Lmae/Fe/c+Fq0FL585Jn/gff4ApSuPLA==</latexit>
<latexit

q +q
d
<latexit sha1_base64="VFuwgu+V5p0BTRBgWq/UWrlORlI=">AAAB6HicbVBNS8NAEJ34WetX1aOXxSJ4KokIeix68diC/YA2lM1m0q7dbMLuRiihv8CLB0W8+pO8+W/ctjlo64OBx3szzMwLUsG1cd1vZ219Y3Nru7RT3t3bPzisHB23dZIphi2WiER1A6pRcIktw43AbqqQxoHATjC+m/mdJ1SaJ/LBTFL0YzqUPOKMGis1w0Gl6tbcOcgq8QpShQKNQeWrHyYsi1EaJqjWPc9NjZ9TZTgTOC33M40pZWM6xJ6lksao/Xx+6JScWyUkUaJsSUPm6u+JnMZaT+LAdsbUjPSyNxP/83qZiW78nMs0MyjZYlGUCWISMvuahFwhM2JiCWWK21sJG1FFmbHZlG0I3vLLq6R9WfPcmte8qtZvizhKcApncAEeXEMd7qEBLWCA8Ayv8OY8Oi/Ou/OxaF1zipkT+APn8wfH6Yzo</latexit>
<latexit
+q
<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>
q +q <latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>

<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

r
<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

<latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>

<latexit sha1_base64="1YQM63vxxOYH66oarH15wTHNiNg=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBbBiyWRgh4LgnisaD+gDWWz3bRLN5uwOxFK6E/w4kERr/4ib/4bt20O2vpg4PHeDDPzgkQKg6777RTW1jc2t4rbpZ3dvf2D8uFRy8SpZrzJYhnrTkANl0LxJgqUvJNoTqNA8nYwvpn57SeujYjVI04S7kd0qEQoGEUrPej+Rb9ccavuHGSVeDmpQI5Gv/zVG8QsjbhCJqkxXc9N0M+oRsEkn5Z6qeEJZWM65F1LFY248bP5qVNyZpUBCWNtSyGZq78nMhoZM4kC2xlRHJllbyb+53VTDK/9TKgkRa7YYlGYSoIxmf1NBkJzhnJiCWVa2FsJG1FNGdp0SjYEb/nlVdK6rHpu1buvVeq3eRxFOIFTOAcPrqAOd9CAJjAYwjO8wpsjnRfn3flYtBacfOYY/sD5/AH9oY2a</latexit>

<latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>
q
<latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>
q +q
<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

<latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>
q +q
<latexit sha1_base64="NCBSIKU4a0QdB8A9m1ZOTtia1Dw=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBZBEEoigh4LgnisYj+gDWWz3bRLN5u4OxFK6D/w4kERr/4jb/4bt20O2vpg4PHeDDPzgkQKg6777RRWVtfWN4qbpa3tnd298v5B08SpZrzBYhnrdkANl0LxBgqUvJ1oTqNA8lYwup76rSeujYjVA44T7kd0oEQoGEUr3Z899soVt+rOQJaJl5MK5Kj3yl/dfszSiCtkkhrT8dwE/YxqFEzySambGp5QNqID3rFU0YgbP5tdOiEnVumTMNa2FJKZ+nsio5Ex4yiwnRHFoVn0puJ/XifF8MrPhEpS5IrNF4WpJBiT6dukLzRnKMeWUKaFvZWwIdWUoQ2nZEPwFl9eJs3zqudWvbuLSu0mj6MIR3AMp+DBJdTgFurQAAYhPMMrvDkj58V5dz7mrQUnnzmEP3A+fwBDG40u</latexit>

<latexit sha1_base64="ItscO4txWYBj02eJ7v4dQltvgsM=">AAAB6XicbVBNS8NAEJ3Ur1q/qh69LBbBiyURQY8FQTxWsR/QhrLZbtqlm03cnQgl9B948aCIV/+RN/+N2zYHbX0w8Hhvhpl5QSKFQdf9dgorq2vrG8XN0tb2zu5eef+gaeJUM95gsYx1O6CGS6F4AwVK3k40p1EgeSsYXU/91hPXRsTqAccJ9yM6UCIUjKKV7s8ee+WKW3VnIMvEy0kFctR75a9uP2ZpxBUySY3peG6CfkY1Cib5pNRNDU8oG9EB71iqaMSNn80unZATq/RJGGtbCslM/T2R0ciYcRTYzoji0Cx6U/E/r5NieOVnQiUpcsXmi8JUEozJ9G3SF5ozlGNLKNPC3krYkGrK0IZTsiF4iy8vk+Z51XOr3t1FpXaTx1GEIziGU/DgEmpwC3VoAIMQnuEV3pyR8+K8Ox/z1oKTzxzCHzifP0YljTA=</latexit>
q Quadrupole Octopole

Figure 2: Left: Vectors determining the field of a dipole configuration. Right: Arrangement of charges
giving monopoles, dipoles, quadrupoles and octopoles.

General Charge Distribution

Now we consider the problem of finding the potential for a general charge distribution ρ(~x0 ) were
we assume that ρ(~x0 ) = 0 when |~x0 | > R some charateristic size of the distribution. We want to
find the change potential at some point ~x at a distance r  R. In general, we can always write
1 1
Z
V (~x) = d3 x0 ρ(~x0 ) (1.135)
4π0 |~x − ~x0 |

18
We again expand |~x| = r  |~x0 | = r0 using

|~x0 − ~x|2 = r2 + r02 − 2rr0 cos θ (1.136)

where θ is the angle between ~x and ~x0 (said differently, we are placing ~x = rẑ for simplicity). We
now expand in r  r0 to get
 0
r02

1 1 1 1 11 r
= q = + 2 cos θ − 2 + . . . (1.137)
|~x − ~x0 | r 1 − (2 r0 cos θ − r02 ) r 2r r r
r r2

The result, which is not obvious ahead of time, is that

∞  0 `
1 1X r
= P` (cos θ) (1.138)
|~x − ~x0 | r r
`=0

Two aspects of this formula are easy to understand. When ~x 6= ~x0 , we recall that
1 1
∇~2x 0
= ∇~2x0 =0 (1.139)
|~x − ~x | |~x − ~x0 |

But we know that any solution to this equation in radial coordinate takes the form
∞
1 1 X
∇~2x 0
→ 0
= (c` (~x0 )r` + d` (~x0 )r−`−1 )P` (cos θ) (1.140)
|~x − ~x | |~x − ~x |
`=0

and
∞
1 1 X
∇~2x0 → = (c0` (~x0 )r0` + d0` (~x0 )r0−`−1 )P` (cos θ) (1.141)
|~x − ~x0 | |~x − ~x0 |
`=0

If we take r  r0 , then we expect that this solution should be well behaved when r → ∞ or
r0 → 0 and therefore the only way to satisfy both equations is if
∞  0 `
1 1X r
= d` P` (cos θ) (1.142)
|~x − ~x0 | r r
`=0

The part that we probably couldn’t have just guessed ahead of time is that d` = 1 for all `.
So far this has just been a bunch of Taylor expanding an algebra. But when we plug this
back into a formula for V (~x) we get
∞
1 1 1 1X 1
Z Z
3 0 0
V (~x) = d x ρ(~
x ) = d3 x0 r0` P` (cos θ)ρ(~x0 ) (1.143)
4π0 |~x − ~x0 | 4π0 r r`
`=0

Notice is that notice is that the final integral no longer depends on r. What this tells us about
the behavior very far away r  R. Since the charge distribution has a characteristic size R, we
might expect that Z
d3 x0 r0` P` (cos θ)ρ(~x0 ) = c` R` (1.144)

19
where c` = O(Q) are the dimensionless number characterizing the shape of the charge distribution
and Q = d3 x0 |ρ(~x0 )| is the total charge. We can write the potential as
R

∞  ` !
1 1 X R
V (~x) = Q+ c` (1.145)
4π0 r r
`=1

To understand the implications, imagine that for any given experiment, to predict the outcome I
need to be able to model V (r) to 1% accurate. If R = 1 cm and r = 10 cm, then I should expect
to need to keep up to ` = 2, since the ` = 2 contribution is c2 (1/10)2 = 10−2 c2 . Now if we go to
r = 1 m, now we only need ` = 0 and ` = 1 and I expect the mistake I am making by dropping
c2 (1/100)2 = 10−4 c2 is about 0.01%. For r > 1 m, I can probably treat the potential as if it was
a point charge with negligible error.
This phenomenon, is known as “decoupling”, and is one of the most profound and important
concepts in physics. In essence, it tells us that we don’t need to know exactly what is happening at
short distances to understand the world at longer distances. Instead, a few number characterizing
the short distance behavior is good enough to have a very accurate picture of the work. This is
why Newton was able to use gravity to explain the solar system, the tides, etc. even though he
had no idea how the sun worked, or the composition of the planets. It is the same reason we can
understand the working of atoms, i.e. chemistry, without understanding the nucleus, quarks, or
modern particle physics.
However, this idea also underscores why it is so difficult to make new discoveries: imagine
you want to know the inner workings of a charge distribution. If you are far away, it is really
hard. You either have to make very very precise measurements, or you have to get a lot closer.
Most experiments cannot tell apart two distributions that have the same total charge, dipole and
quadrupole moment. This is the same reasons it is hard to find violations of the laws of physics
we know, whether that is Einstein’s theory of gravity, or our full model of particle physics.

1.7 Electric Field of a Dipole

In the next topic, we will refer to the dipole a lot, so here was want to expand a bit on what a
dipole looks like. The dipole is important because electric forces really don’t like objects to have
a total net charge Q. Instead, matter tends to be made in almost exactly numbers of protons
and electrons so that the total charge cancels out (i.e. is N is the number of atoms in a system
Q  eN ). On the other hand, the quantum mechanics forbids that electrons and protons can
be in the same place, so there is good reason for atoms and molecules to have a non-zero dipole
moment. Given that the size of an atom is tiny compared to the macroscopic scales we are
discussing, we can typically ignore all the higher multipoles and just focus on the dipole.
The general definition of the dipole is of a group of N charges is
N
X
p~ = qi ~xi (1.146)
i=1

(NOTE: here i is not a vector index, it is just labelling all N particles). We can see this matches

20
our previous definition as follows: for this configuration
N
X
ρ(~x0 ) = qi δ(~x0 − ~xi ) (1.147)
i=1

Z N
X N
X
d3 x0 r0 P1 (cos θ)ρ(~x0 ) = qi |~xi | cos θi = r̂ · qi ~xi = r̂ · p~ (1.148)
i=1 i=1
so that the potential form the dipole is
1 p~ · ~x 1 p~ · ~x
V (~x) = 3
= (1.149)
4π0 r 4π0 |~x · ~x|3/2

where we used r̂ = ~x/r and r2 = ~x · ~x. In this form, it is easy to find the electric field using
cartesian coordinates:
p~ · ~x 3 p~ · ~x
   
~ ~ 1 ~ 1 p~
E = −∇V (~x) = − ∇ =− − 2~x (1.150)
4π0 |~x · ~x|3/2 4π0 |~x · ~x|3/2 2 |~x · ~x|5/2
1 1
= (3(~p · r̂)r̂ − p~) (1.151)
4π0 r3
where we again used r̂ = ~x/x to relate cartesian and spherical coordinates.

1.8 Multipoles, Translations, and Rotations

The multipole expansion has some somewhat non-intuitive properties when it comes to thinking
about how we move around the space of coordinates. There are two important features that are
useful for getting one’s bearings

Origin of Coordinates: As written, the expressions for the multipole moments (i.e. c` ) have
some not ideal properties. For one, it depends on exactly where you define r = 0. This should
be familiar, as the quantities we are calculating are equivalent to the moment of inertia and
generalizations therefore. In the context of mechanics, you may recall the right moment of
inertia depended on the axis around which it was rotating.
The good new is that the first coefficient (the biggest term) in the series does not dependent
on where we place r = 0. For example, the ` = 0 term (the monopole) is given by
Z
d3 x0 ρ(~x0 ) = Q . (1.152)

which is the total charge. We would definitely hope the total charge doesn’t depend on how we
define our coordinates. But now suppose that Q = 0 and therefore there is no monopole. So we
go to the dipole (` = 1):
1 1
Z
V (~x) ≈ d3 cos θr0 ρ(~x0 ) = c1 R (1.153)
4π0 r2
But now we notice that r cos θr0 = ~x · ~x0 so we can write
1 ~x 1 ~x · p~
Z Z
V (~x) ≈ 3
· d3 x0 ~x0 ρ(~x0 ) = p~ ≡ d3 x0 ~x0 ρ(~x0 ) (1.154)
4π0 r 4π0 r3

21
Now we can see what happens if we shift the origin of our coordinate system ~x0 , ~x → (~x0 −~a), (~x−~a)
for some constant vector ~a, then in the coordinate centered around ~a we have
Z
p~ → d3 x0 (~x0 − ~a)ρ(~x0 ) = p~ − ~aQ = p~ (1.155)

So in the presence of a non-zero total charge, the dipole changes by −~aQ but when the leading
term vanishes, Q = 0, the second term is independent of the coordinates.
One can use more or less the same idea (and some fancier ways or writing this expression) to
show that it c` = 0 for ` < L then cL 6= 0 doesn’t depend on the coordinates (i.e. the expression
of the influence of ~a on cL depends on c`<L and therefore vanishes when it is the leading term.

Rotations and Multipole Moments: In writing the multiple expansion, we made the choice
of placing ~x = rẑ so that the cosine between ~x and ~x0 was the same as the polar angle θ in
spherical coordinates. In other words, we can find the electric potential at any point ~x is we
rotate the object holding ~x = rẑ fixed (which is equivalently to rotating ~x and holding the object
fixed.
Since this is just a rotation, it is natural to think we can simply use the rotation matrix to
figure out the electric potential. This is actually true, but more complicated than you might
think. Let’s go back to our formula for ~x = rẑ:
∞  ` !
1 1 X R
V (~x) = Q+ c` (1.156)
4π0 r r
`=1

where Z
c` R =`
d3 x0 r0` P` (cos θ)ρ(~x0 ) . (1.157)

So the question is, how do Q and c` change if we were to rotate the object (or the coordinates).
The way to see how this works, is to notice that

rr0 cos θ ≡ ~x · ~x0 = xi x0i r0 cos θ = x̂i x0i (1.158)

so that
1
Z Z Z
3 0 0 3 0 0i 0
Q= d x ρ(~x ) c1 = x̂i d x x ρ(~x ) c2 = x̂i x̂j d3 x0 (3x0i x0j − δ ij )ρ(~x0 ) (1.159)
2
If we want to rotate the objects, holding the ~x coordinate fixed that means we want to rotate ~x0
and not ~x. Now we see a patter: Q is a scalar under rotations (there are no indices to rotate), c1
is a vector under rotations of ~x0 , c2 is a two-index tensor under rotations of ~x0 , etc. In this sense,
it would have made more sense to write
∞  ` !
1 1 X
i1 ..i` R
V (~x) = Q+ (x̂i1 ...x̂i` )c (1.160)
4π0 r r
`=1

where now we defined Z

d3 x0 r0` P` (cos θ)ρ(~x0 ) = R` (x̂i1 ...x̂i` )ci1 ..i` (1.161)

where the multipole moments ci1 ..i` transforms under rotations as if each index in is like a vector
(hence the whole thing is a `-index tensor).

22