Sunteți pe pagina 1din 30

Electrodynamics

Yair Mazal
August 25, 2013

Contents
1 Electrostatics
1.1 Poisson Equation and the Potential . . . . . . . . . . . . .
1.2 Energy of an Electric Field . . . . . . . . . . . . . . . . .
1.3 Laplace Equation . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Mean-Value Theorem . . . . . . . . . . . . . . . .
1.3.2 Uniqueness Theorem . . . . . . . . . . . . . . . . .
1.3.3 Separation of Variables - Cartesian Coordinates . .
1.3.4 Separation of Variables - Cylindrical Coordinates .
1.3.5 Separation of Variables - Spherical Coordinates . .
1.4 Boundary Conditions . . . . . . . . . . . . . . . . . . . . .
1.4.1 Discontinuity in the Normal Field . . . . . . . . .
1.4.2 Continuity in the Tangential Field . . . . . . . . .
1.4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . .
1.5 Multipole Expansion . . . . . . . . . . . . . . . . . . . . .
1.5.1 Two First Terms . . . . . . . . . . . . . . . . . . .
1.6 Electric Field in Matter . . . . . . . . . . . . . . . . . . .
1.6.1 Polarization and Bound Charge . . . . . . . . . . .
1.6.2 Electric Displacement - Gauss Law and Dielectrics
1.6.3 Boundary Conditions in Dielectrics . . . . . . . . .
1.6.4 Linear Dielectrics . . . . . . . . . . . . . . . . . . .
1.6.5 Energy in Dielectrics . . . . . . . . . . . . . . . . .
2 Magnetostatics
2.1 Lorentz Force . . . . . . . . . . . . . . . . .
2.1.1 Currents . . . . . . . . . . . . . . . .
2.1.2 Charge Conservation . . . . . . . . .
2.2 The Biot-Savart Law . . . . . . . . . . . . .
2.2.1 The Vector Potential . . . . . . . . .
2.3 Boundary Conditions . . . . . . . . . . . . .
2.3.1 Continuity in the Normal Field . . .
2.3.2 Discontinuity in the Tangential Field
2.3.3 Conclusion . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

2
2
4
5
5
5
7
9
11
14
14
14
15
16
17
18
18
20
20
21
21

.
.
.
.
.
.
.
.
.

23
23
23
24
25
25
28
28
28
29

Chapter 1

Electrostatics
1.1

Poisson Equation and the Potential

We shall start by reminding ourselves of maxwells first two equations for electrostatics:
~ =0
E
~ = 4
E
~ = .
Due to the first we can deduce: E
Now putting that into the second we arrive at Poisson equation:
= 4

(1.1.1)

Moreover we know from previous courses that given a continuous charge distribution,
one can write the potential as:
Z
(~r0) 3
(~r) =
d ~r0
(1.1.2)
|~r ~r0|
The latter is the solution for Poisson equation. We shall now derive this result. In
order to do that lets get back to Poissons and try to solve it for a point charge, meaning
for a Delta function as the distribution.
First of all we need to notice the spherical symmetry in the problem, and therefore:
g(~r) = g(r) = 4(~r)

(1.1.3)

Using the definition of the Laplacian in spherical coordinates, and dropping derivatives with respect to the angles (due to the symmetry of the problem) we have:
1 2 g(r)
r
= 4(~r)
r2 r
r

and as long as: r 6= 0


1 2 g(r)
g(r)
c
r
= 0 r2
=c g = +b
2
r r
r
r
r
But g(r) goes to zero at infinity and therefore:
g(r) =

c
r

This doesnt work though for r=0, and therefore we shall return to the definition of
the Laplacian to deal with the origin. We know that: = div(grad ), and we shall
use it along with the Gauss theorem in order to solve the problem around the origin.
Integrating equation (1.1.3) over a small volume around the origin we arrive at:
I
Z
Z
~
ds =
div(grad ) dV =
4(~r) dV = 4
(1.1.4)
|~
r|=R

|~
r|<R

|~
r|<R

In the integral on the left of (1.1.5) we dont sum over the origin directly, and
therefore it can be calculated:
I
Z
c 2
~
ds =
r d = 4c = 4 c = 1
(1.1.5)
r2
|~
r|=R
Finally we have the potential of a point charge at the origin:
g(r) =

1
r

(1.1.6)

We shall now generalize this


R result. Using the Delta function we know the following
is true by definition: (~r) = (~r0)(~r ~r0) d~r0. Therefore we have now the potential for
a point charge away from the origin as:
g(|~r ~r0|) =

1
|~r ~r0|

(1.1.7)

Now using the super-position principal for an arbitrary distribution of point charges,
which approaches in the limit a continuous distribution we arrive at Poisson integral
(1.1.2).

1.2

Energy of an Electric Field

For two point charges the energy of the system is defined as: U =

q1 q2
|r~1 r~2 | .

This definition
~
is consistent both with dimensions, and with our requirement that: F = U , that is
known due to the fact that the electric field in electrostatics is rotorless. This can be
generalized according to additivity of energy for a discrete distribution:
U=

1 X qi qj
1X
=
qi i
2
|~
ri r~j |
2

(1.2.1)

i6=j

For a continuous distribution this yields:


Z
1
(~r)(~r) d~r
U=
2 all space

(1.2.2)

R
1
~ d3~r,
and substituting maxwells eq. inside we get: U = 8
r) div E
all space (~
R
R
1
2 3 r ].
~ +
~ ds
and by implementing the product rule: U = 8
[ all space (~r) E
all space E d ~
R
r
~ r2
~
~ 12 , and ds
~ ds
Since 1 , E
(~r) E
0,
r

all space

and we arrive at the conclusion:


1
U=
8

E 2 d3~r

(1.2.3)

all space

The last equation is a bit problematic. We can notice that there is some inconsistency
between our expression for energy in the discrete version which can have any sign, and
the last equation which is definitely positive. This inconsistency arises from the fact
that while our discrete sum evades summing self interactions (i 6= j), our continuous
expression does sum them.
In short one would be wise to exercise caution when calculating electric energy, and
basically act upon the following rules:
When you have a discrete distribution use the discrete version.
When encountered with a continuous configuration use the fields version.
Note that the continuous formula isnt additive (E12 + E22 6= (E1 + E2 )2 ), and
therefore unlike the usual cases the energy of two fields is different than the sum
of separate energies.

1.3

Laplace Equation

Solving Poissons integral (1.1.2), could theoretically solve all of our problems with electrostatics. Unfortunately solving this integral can prove extremely hard, and therefore
other ways to deal with these problems were developed. One of which is solving Laplace
equation.
Laplace equation is the homogenous case of Poisson, i.e. = 0. As a result once were
able to solve the homogenous case we only need to find private solutions. This can
prove hard as well. The real advantage of the looking at the homogenous case is the
a differential equation can be solved by separate functions on various regions. Laplace
equation will determine along with sufficient boundary conditions a unique potential, in
regions in which no charge is present.
Now lets take a look at Laplace equation:
= 0

(1.3.1)

First of all we can notice that we can write, and attempt to solve this equation in any
coordinate system. Before we try we shall notice some general qualities of this equation.

1.3.1

Mean-Value Theorem

Functions satisfying Laplace equation are called harmonic functions, and as such they
satisfy the mean value theorem for harmonic functions. For three dimensions it states
that the potential at a point ~r in space is equal to the average of the potential over a
sphere surrounding it, where the point is at the origin.
I
1
(~r) =
da
4R2 |~r|=R
The same applies for an average over a circle in 2D.
An important consequence of this is theorem is that the potential cannot have local
extrema point, for if there were any I could draw a sphere around them and obviously
the theorem would be contradicted.

1.3.2

Uniqueness Theorem

As we know from PDEs, also here sufficient boundary conditions will determine the
uniqueness of the solution. We shall dwell in that for a few moments to determine what
is sufficient?
First Case
The solution for Laplace equation is uniquely defined over a volume V if the potential
is defined all over its surface S. It can be easily proved because if we would assume two
independent solutions 1 , and 2 they would still have to be equal at the surface in order
to solve the equation. Moreover since the equation is linear their difference 3 = 2 1
is also a solution. We also know that 3 = 0 over the surface (the actual value, not the

average). Since all extreme points take place on the boundary it follows that 3 = 0
everywhere and 1 = 2 .
Therefore in order to know the potential uniquely in a region, we need to know it over
the surface of the region, and to know the charge distribution within the volume.
Second Case
This statement is relevant to configurations involving conductors. We say that the
Electric field is uniquely determined over a region V surrounded by conductors, and
containing a known density , as long as the total charge on each conductor (the one at
the surface, and another inside the volume if exists) is known. The region can be as I
said bounded by a conductor or, unbounded which is like saying its bounded at infinity.
Proof/explanation - Lets suppose that there are two fields satisfying the conditions of
the problem. In such case both fields abide Gauss integral and differential laws:

E~1 = 4,
E~2 = 4
I
~ = 4Qi ,
~ = 4Qi
E~1 ds
E~2 ds

ith conductor

ith conductor

The same applies to the outer boundary either if it is finite or infinite:


I
I
~
~ = 4Qtotal
~
E1 ds = 4Qtotal ,
E~2 ds
outer boundary

outer boundary

Again we shall look at the difference between the two and notice that:
I
~ =0
~
E3 = 0,
E~3 ds

(1.3.2)

every boundary

One other thing we need to notice is that all conducting surfaces are equipotential (each
has a different but constant potential). Here comes the trick:
(3 E~3 ) = 3 ( E~3 ) + E~3 (3 ) = E32
The latter exploits our knowledge about the relation of field and potential, as well as
our knowledge of (1.3.2) Now well try to integrate over the entire volume in question:
Z
I
Z
~
~
~
(3 E3 ) dV = 3 E3 dS = E32 dV
V

If we look at the integral at the middle we can see that over each boundary (conducting
surface) the potential is constant and can be taken out of the integral and according to
(1.3.2) the whole thing equals zero. Hence we have:
Z
E32 dV = 0
V

But this integrand is non-negative and therefore is the integral is zero, the integrand
must be zero everywhere. Now we got our result E~1 = E~2 .
6

1.3.3

Separation of Variables - Cartesian Coordinates

First, few sentences about this method in general. When trying to find a solution for a
PDE we often encounter great difficulties. one of the ways to overcome this problem is
to guess that we can find a solution using separation of variables, i.e. that we can find a
solution that is built from several functions. Each of these functions will be a function
of one variable only, and their multiplication will yield the solution for our equation.
This assumption will prove very helpful if such a solution can actually be found. Sadly
it is very hard to find solutions that do not obey this principle, and we are often in the
dark once it fails. On the other hand if we can find such a solution were obviously very
happy, and the to the uniqueness theorem we can be sure that it is the only one.
Now lets talk about Cartesian coordinates. Laplace equation will take the form of
=

2 2 2
+
+ 2 =0
x2
y 2
z

(1.3.3)

Now assuming: = X(x)Y (y)Z(z), and dividing by , we arrive at


1 2X
1 2Y
1 2Z
+
+
=0
X x2
Y y 2
Z z 2
Now using the assumption of separation of variables we get:
1 d2 X
= C1 ,
X dx2

1 d2 Y
= C2 ,
Y dy 2

1 d2 Z
= (C1 + C2 )
Z dz 2

Now we only need to solve three simple ODE according to boundary conditions. A few
notes are in order to help figuring these out, as they arent always as simple as they
seem:
In problems which are independent of one of the coordinates we always have the
constants from the separation equal but opposite in sign. In three dimensions we
have more freedom of choice regarding their sign. Keep in mind that even though
some choices are wiser than others, they all lead to the same conclusion, if youre
unsure go with the flow.
If you want to make it easier, choose the coordinates in which the solution must
be periodical to have the negative constants.
Symmetry isnt always on your side, meaning, if you have a problem with two
planes at zero potential chose the coordinate system in such a way that you can
write: sin( na x ) with compliance with the conditions, where a is the distance
between the planes. It will make your life easier and less confusing than other
choices.
Lets say we have a charge distribution of a certain plane, for example (x = 0) =
(y, z). In that case we can relate discontinuity in the electric field normal to the
plane to the distribution like this:

|
| + = (|x=0 |x=0+ ) n
= 4
~n x=0
~n x=0
7

(1.3.4)

Do remember though that if you actually take the derivative of the Fourier series
that you got one element at a time, you a get new series which represents the field
function at a given direction. This function suffers a discontinuity at the surface
charge and therefore the Fourier series converges to the mean value, and not to
the value of the function with accordance to the Dirichlet theorem.
The solution I presented isnt actually the complete solution. The complete solution to the differential equation includes also a case in which at least one of the
constants generated by the separation is zero. In such a case the solution will be
linear at least in one of the variables. This case can usually be ignored on physical
grounds, but for instance this is the solution for a plate capacitor.
The last but most complex situation is a boundary condition in the form of a
function of the potential over the coordinates on a surface. In such a case Fourier
series must be used either in 1D or 2D.
In one variable lets assume for the sake of discussion that we have a boundary
condition in the form of (y = 0) = (y = a) = 0. Such condition will result
in periodical solution in y in the form of n sin( na y ). If we got a periodical
solution in y from separation of variables well get an exponential solution in x (we
assumed a problem in 2D). Therefore we have:
n = (An ekn x + Bn ekn x ) sin(kn y)
In the last equation kn =

n
a ,

(1.3.5)

and then the complete solution for the problem is:


X
(x, y) =
n (x, y)
(1.3.6)
n

The only problem is that we dont know yet the coefficients An , and Bn . Those
two must be found using the orthogonality of Fouriers system. Since finding both
of them can prove hard we can break the problem into two problems. In each
one of the boundaries will have zero potential and the other will have the given
function. The final solution will be the sum of both.
For the sake of simplicity lets assume we found all the coefficients Bn to be zero.
This could happen for example if we solve for an open boundary in the positive x
values, and therefore we require Bn = 0, on the physical grounds of finite potential
everywhere. We still need to find An . Lets take another arbitrary boundary
condition in the form of (x = 0) = 0 (y). Putting that into (1.3.6) we get
X
0 (y) = (0, y) =
An sin(kn y)
(1.3.7)
n

Now multiplying the last expression by sin(km y), and integrating over the domain
[0, a], we get using the orthogonality of Fouriers system
Z
2 a
An =
0 (y) sin(kn y) dy
(1.3.8)
a 0
8

A note must bee made that I implicitly used my ability to choose the origin in such
a manner the even modes drop off. This isnt necessarily possible. Theoretically
two more degrees of freedom are in order (since we have 4 boundary conditions),
but a wise choice of coordinates can often remove some without any effort.
The 3D case is the same as the 2D in most aspects. There are two differences
though. The first is that once we acknowledge that the solution is periodical in
one coordinate we dont no yet about the rest. The other is that using Fourier is
a bit more tricky even though its basically the same. In 3D if we chose wisely we
can get
n,m (x, y, z) = (An,m ekn,m x + Bn,m ekn,m x ) sin(

n y
m z
) sin(
)
a
b

(1.3.9)

For simplicity again we shall drop one of coefficients, and using the same logic as
before we get:
Z a Z b
4
n y
m z
An,m =
dy
dz 0 (y, z) sin(
) sin(
)
(1.3.10)
ab 0
a
b
0

1.3.4

Separation of Variables - Cylindrical Coordinates

In cylindrical coordinates we shall assume the potential doesnt depend on the z coordinate (it is feasible to solve without the assumption, but it simplifies things and in all of
the problems I saw we could make this assumption. Therefore we have = R(r)(),
and the Laplace equation takes the form
=

1 2
R
R 2
1
(r ) + 2 2 =
(r
)+ 2 2 =0
r r r
r
r r r
r

(1.3.11)

Multiplying the last result by r2 , and dividing by


r R
1 2
(r
)+
=0
R r r
2
Assuming separation of variables
r R
(r
) = C1 ,
R r r

1 2
= C1
2

A wiser choice here is to assume C1 = m2 (meaning its positive), because the solution
must be periodical in . We get two ODE
1 2
= m2 m = eim
2
And

r R
(r
) = m2
R r r
9

Rm = Am rm + Bm rm

m 6= 0

R0 = A0 ln r + B0
The complete solution will be
X
(r, ) = A0 ln r + B0 +
(Am rm + Bm rm )(Cn cos m + Dn sin m)

(1.3.12)

A few anecdotes:
As before in order to define the solution uniquely we need two physical boundary conditions. They can be any combination of conditions regarding charge distribution (Neumann condition), or the potential over specific surfaces (Dirichlet
condition).
Regarding charge distribution conditions equation (1.3.4) still holds. In cylindrical
coordinates it can take the form of
(r = r0 ) = ()

|r=r
| + = (|r=r |r=r+ ) r = 4
0
0
0
~n
~n r=r0

The convergence to the mean value is the same as in the cartesian case.
It is advisable to remember that since we didnt define yet our Fourier constants in
any way, we can play a bit with the way things look in order to make things easier.
For example we can assume an arbitrary boundary condition: (r = R) = 0 ().
We shall attempt to find the potential for smaller and bigger radii (unbounded).
For smaller radii r < R, we must assume some of the constants vanish due to
finiteness of the potential, and for larger other vanish:
r<R:

(r, ) = B0 +

X
r
( )n (Cn cos n + Dn sin n)
R

n=1

r>R:

(r, ) = F0 + A0 ln(r/R) +

X
R
( )n (Tn cos n + Sn sin n)
r

n=1

It can be noticed that even though the expression changed its appearance it is
essentially the same, except for the mentioned requirements. We can exploit continuity of the potential in order to gain some more knowledge.
(R , ) = (R+ , ) B0 +

(Cn cos n+Dn sin n) = F0 +

n=1

(Tn cos n+Sn sin n) = 0 ()

n=1

This gives us due to the uniqueness of Fourier series


B0 = F0 , Cn = Tn , Dn = Sn

10

We can now find Fourier coefficients:


Z 2
1
B0 =
0 () d
2 0
Z
1 2
Cn =
0 () cos n d
0
Z
1 2
0 () sin n d
Dn =
0
Note that we havent found A0 , since we provided only one physical boundary
condition. It can be found for example by providing the charge distribution over
a certain radius.
I assumed here a solution over an unbounded region. If the region was to be
bounded (for instance a problem over a ring), we couldnt drop automatically all
coefficients for positive or negative powers of r.

1.3.5

Separation of Variables - Spherical Coordinates

We shall now solve Laplace equation in spherical coordinates. The Laplace equation will
take the form
=

1 2
1

1
2
(r
)
+
(sin

)
+
=0
r2 r
r
r2 sin

r2 sin2 2

(1.3.13)

Assuming separation of variables: (~r) = R(r)Y (, ) we arrive at


2Y
Y 2 R
R

Y
R
(r
)
+
(sin

)
+
=0
2
r2 r
r
r2 sin

r2 sin 2
Dividing by and multiplying by r2
1

Y
1
2Y
1 2 R
(r
)+
(sin
)+
=0
2
R r
r
Y sin

Y sin 2
And assuming separation of variables
1 2 R
(r
) = l(l + 1)
R r
r

(1.3.14)

(It can be shown that the constant generated by the separation of variables must fit the
form of l (l + 1) for the solution not to explode.)
1

Y
1
2Y
(sin
)+
= l(l + 1)
Y sin

Y sin2 2

11

(1.3.15)

We can Solve for r: Rl (r) = Al rl + Bl r(l+1) Now we shall attempt to separate again
assuming: Y (, ) = ()().
Putting that into (1.3.15), we get
sin

1 2
(sin
)+
= l(l + 1) sin2

(1.3.16)

Again we have a function of alone, and a function of alone. Lets solve for :
d2
= m2 () = eim
d2
and since it should be periodical with a 2 period, m is an integer. Now using this result
in (1.3.16) we arrive at
1

m2
] = 0
(sin
) + [l(l + 1)
sin

sin2
For now we shall assume m = 0, and use u = cos . By these with a bit of effort we can
arrive at the Legendre equation:
d
d
[(1 u2 ) ] + l(l + 1) = 0
du
du

(1.3.17)

This equation has a solution in the form of Legendre polynomials. These functions form
an orthogonal basis (which can be normalized too), and form a complete orthonormal
system, meaning functions can be decomposed using these polynomials the same way
we use Fourier system.
A few points about Legendre equation and its solution:
It can be proved that our requirement of the separation constant to be in the form
of = l(l + 1) is needed in order to have finite solutions over all of our domain.
Our domain is [1, 1] which can be justified if we remember that u = cos .
There are several ways to derive Legendre polynomials. One of which is:
Pl (u) =

1 dl 2
(u 1)l
2l l! dul

These functions form an orthogonal system using an inner product of:


Z 1
2
Pl Pm du =
l,m
2l
+1
1
In theory were supposed to have two independent solutions to the equation since
its a second order equation. Our solution using a series has yielded the Legendre
polynomials, but what about the second solution. It happens that the second
solutions blow up at the ends of our domain, and therefore are irrelevant to our
case on physical grounds.
12

The last solution holds only for m = 0. Under these circumstances ( is independent
of ) the complete solution is:
X
(r, ) =
(Al rl + Bl rl )Pl (cos )
l

The specific solution can be derived from boundary conditions. Since the calculation of
the coefficients using spectral decomposition isnt feasible without a computer for any
but the simplest cases, usually exercises have a boundary condition in the form of one
the first polynomials, or you can write the solution only symbolically. For this reason
its a good idea to know to recognize the first few (they appear in our formulae page).
If we want to return to the general case (m 6= 0) the separation of variables still
works, but the solution is a bit more complex, even though its essentially the same.
There are several conventions about normalization but basically:
P`m (x) = (1)m (1 x2 )m/2
s
Y`m (, ) =

dm
(P` (x))
dxm

(2` + 1) (` m)! m
P (cos ) eim
4 (` + m)! `

Which yields:
Z

=0

=0

d = ``0 mm0
Y`m Y`m
0

Therefore the most general solution in spherical coordinates is:


X
(r, , ) =
R(r)Y`m (, )
`,m

13

1.4

Boundary Conditions

When trying to solve the Laplace equation we always solve it over regions. We already
know the general solution in different regions, but the important part is applying the
correct boundary conditions.

1.4.1

Discontinuity in the Normal Field

The electric field is always discontinuous in the normal component when encountering a
boundary with a charge distribution.
Lets assume our configuration has a boundary in the form of a surface charge . We
do not require to be constant in general. Now we shall construct a closed surface of
height , and face of area A. We shall take A to be small enough so that doesnt vary
along the surface, and  small enough so that the surface is practically a plane in our
analysis,
and the contribution to the field is normal
to this2. plane.
CHAPTER
ELECTROSTATICS
88

Figure 2.36

Now using Gauss law over this small box


I
~ = 4Qin = 4A
~ ds
E

You may have noticed, in studying


Exs. 2.4 and 2.5, or working problems such as 2.7,
S
2.11, and 2.16, that the electric field always undergoes a discontinUity when you cross a
Since
the charge
sides of
boxitcontribute
nothing
duethe
toamount
the dotbyproduct
were left
a. the
In fact,
is a simple matter
to find
which E changes
at with
surface
such a boundary. Suppose we draw a wafer-thin Gaussian pillbox, extending just barely

Ebelow
=states
4 that
above
over the edge in each direction (Fig.E2.36).
Gauss's
law

=
=
I
Continuity in the Tangential Field
E. da

1.4.2

EO

EO

A is thefield
area is
of continuous
the pillbox lid.
(If a boundaries.
varies from point
to pointtoorsee
thethat
surface
Thewhere
tangential
across
In order
weisshall look
A
to
be
extremely
small.)
Now,
the
sides
of
the
pillbox
contribute
curved,
we
must
pick
at an infinitesimal loop, and attempt to calculate to circulation of the electric field along
it. nothing to the flux, in the limit as the thickness E goes to zero, so we are left with
1..
Eabove -

1..
Ebelow

1
-a,

(2.31)

EO

where Ei1bove denotes the component of E that is perpendicular to the surface immediately
above, and Etelow is the same, only just below the surface. For consistency, we let "upward"
14 normal component of E is discontinuous
be the positive direction for both. Conclusion: The
by an amount a / EO at any boundary. In particular, where there is no surface charge, E1.. is
continuous, as for instance at the surface of a uniformly charged solid sphere.
The tangential component of E, by contrast, is always continuous. For if we apply
Eq.2.19,

E dl

= 0,

to the thin rectangular loop of Fig. 2.37, the ends give nothing (as
.

( I

give Eabovel - Ebelowl), so

---+ 0), and the sides

2.3. ELECTRIC POTENTIAL

89

2.3. ELECTRIC POTENTIAL

89

Figure 2.37

Since we know that

~ =0
~ d`
E

where Ell stands for the components of E parallel to the surface. The boundaryk conditionsk

and if on
weEtake
 0, then the sides contribute nothing we arrive at: Eabove = Ebelow
(Eqs. 2.31 and 2.32) can be combined into a single formula:
Figure 2.37

Eabove - Ebelow

1.4.3

=-

EO

Conclusion

n,

(2.33)

Ellisstands
the perpendicular
components oftoEthe
parallel
to the
surface.
The
boundary
conditions
a unit for
vector
surface,
pointing
from
"below"
to "above.,,6
where n

Our conclusion
can
be written
both
for the field,
and formula:
for the potential:
2.31 and
2.32) can is
becontinuous
combined into
a single
on EThe
(Eqs.
potential,
meanwhile,
across
any boundary
(Fig. 2.38), since

~ below E
~ above = above
a below = 4
E
n
b
Eabove - Ebelow = - n,
Vabove -

Vbelow

=-

Or in scalar notation:

EOE . dl;

(2.33)

n is length
a unit vector
to does
the surface,
pointing from "below" to "above.,,6
where
as
the path
shrinksperpendicular
to zero, so too
the integral:
below
above
The potential,
meanwhile,
is continuous across any boundary (Fig. 2.38), since

= (above below ) n
= 4
n

(2.34)
Vabove = Vbelow.b

Vabove - Vbelow
= - if Ewe. dl;
The potential itself though is continuous
because
take
Z
as the path length shrinks to zero, so too does the integral:
~
~ d`
above below = E
Vabove

Vbelow.

(2.34)

Figure 2.38

and

6Notice that it doesn't matter which side you call "above" and which "below," since reversal would switch the
direction of n. Incidentally, if you're only interested in the field due to the (essentially flat) local patch of surface
above the
surface, and -(a /2Eo)n immediately below. This
charge itself, the answer is (a /2Eo)n immediately
Figure
2.38
if you
are close
enough
patch =
it "looks"
follows from
2.4, forof
taking
the Ex.
length

0, we
gettotheabove
belowlike
. an infinite plane. Evidently the entire
discontinuity in E is attributable to this local patch of charge.
6Notice that it doesn't matter which side you call "above" and which "below," since reversal would switch the
direction of n. Incidentally, if you're only interested in the field due to the (essentially flat) local patch of surface
charge itself, the answer is (a /2Eo)n immediately above the surface, and -(a /2Eo)n immediately below. This
follows from Ex. 2.4, for if you are close enough to the patch it "looks" like an infinite plane. Evidently the entire
discontinuity in E is attributable to this local patch of charge.

15

1.5

Multipole Expansion

If were very far away from the a charge configuration, we can generally say that its
geometry becomes less relevant, and the potential/field assume a form closer the form
of a point charge. We shall try to implement this logic in this section.
Important note: The expression developed here assumes azimuthal symmetry, and
therefore holds only in such cases. For the general case refer to Gedalins lecture notes
in the third lecture.
In order to take this forward we need to remember poissons integral (1.1.2). We
1
1
shall try to develop |~r~
r0| in powers of r .
We can write
 r0 
 r0 2
2
|~r ~r0|2 = r2 [1 +
cos 0]
r
r
or defining
 r0   r0


2 cos 0
r
r
we can write

|~r ~r0| = r 1 + 
Now using Taylor expansion
1
1
1
1
= (1 + ) 2 =
|~r ~r0|
r
r



1
3 2
5 3
1  +   +
2
8
16

The remarkable thing about this result is that it can be written also as:

1
1 X  r0 n
=
Pn (cos 0)
|~r ~r0|
r
r
n=0

Incorporating this result into (1.1.2) we get


(~r) =

X
n=0

1
rn+1

(r0)n Pn (cos 0)(~r0) dV 0

(1.5.1)

This expression is a solution for Poissons equation in powers of 1r . The first term is the
monopole contribution, the second is the dipole, the third is the quadrupole, and so on.
When using this approximation it can easily be seen, that far away form the configuration
the field and potential acts mainly according to the first non-zero term of this series.
This can easily tell us how fast does the potential decay, even without solving explicitly
just by knowing which terms vanish.

16

1.5.1

Two First Terms

The first term can be easily


derived from the general formula (n = 0). In this case we
R
can see that since Q = dV , we have
=

Q
r

The dipole potential (second term) will be:


Z
Z
1
1
dipole (~r) = 2
r0 cos 0 (~r0) dV 0 = 2 r ~r0 (~r0) dV 0
r V
r
V
We can notice that the last integral doesnt depend on ~r, and if we define
Z
p~ ~r0 (~r0) dV 0
V

we can write:
dipole (~r) =

p~ ~r
r3

(1.5.2)

For a discrete configuration we can write:


p~

qi~r0i

Another thing we can notice is that the first term of the series is independent of our
choice of origin. Moreover in case that the first term vanishes (meaning Q = 0) then the
second term is independent of choice of origin since if we move it by ~a we get:
Z
Z
p~1 = ~r01 (~r0) dV 0 =
(~r0 ~a) (~r0) dV 0 = p~ Q~a = p~
V

~ =
Using the result for we find the electric field of a dipole by: E

~ = r3 (~
E
p ~r) p~ ~r r3
(~
p ~r) = ~r ( p~) + p~ ( ~r) + (~r ) p~ + (~
p ) ~r = px

~r
r3 = 3 5
r
p ~r) ~r r2 p~
~ = 3 (~
E
r5

17

~r
~r
~r
+ py
+ pz
= p~
x
y
z

1.6
1.6.1

Electric Field in Matter


Polarization and Bound Charge

Inside a material there are many dipoles. Inside the dielectric the dipoles cannot move
since they are bounded (its a dielectric after all), so all they can do is to rotate. Therefore
the sole effect that an electric field has over the dipoles in a dielectric is that it aligns
them.
This phenomenon can be simply explained if we imagine an electric field acting on
an ideal dipole (two opposite charges). In such a case it is obvious that half of the
dipole is attracted to the field, while the other is repelled by it. Since all it can do is
rotate, the outcome is that the dipole is aligned parallel to the filed. This argument
is an oversimplification of reality since many dipoles arent ideal meaning they do not
consist of two opposite charges. for instance the water molecule which is very polarizable
or carbon dioxide, or any other neutral molecule built of three atoms for that matter.
In general the dipole moment is a second order tensor (it reduces into a vector for two
molecules due to symmetry around the axis of the dipole). We can say that the torque
acting upon it under the assumption of uniform field (uniform on the molecules scale)
is given by:
~
N = p~ E
In case it isnt uniform we can get an expression for the force in the form of:
~
F~ = (~
p ) E
and then the torque would be :
~ = p~ E
~ + ~r F~
N
~ is uniform the expression for the force can be written as:
Note that in case E


~
F~ = p~ E
This notation teaches us that in such a case we can write the energy contained in such
a dipole in an electric field:
~
U = ~
pE
I will now do something confusing, and I apologize in advance. As youve noticed
so far in this course we use the letter p to denote many things. This multiple use of
the same letter is extremely frustrating. As Im sure that the reader is used to being
frustrated since he studies physics which can prove quite hard, I will not change this
convention, and youll have to forgive me.
I shall now (and only during this section) use the P (not p) to denote the following:
P dipole moment density
18

In order to make things more unclear we shall invent a confusing name for this dipole
density, we shall call it Polarization. This name is the same as polarization of waves,
and it resembles it in the meaning that both talk about the direction of fields.
Now we shall attempt to understand what is the potential and field that such an
induced dipole moment produces. If we will remember what is the potential which is
produced by a usual ideal dipole (1.5.2), and if we remember the fact that now our
potential isnt point like, and hence we shall use something similar to (1.1.2) we can
convince ourselves that this would work:
Z
(~r ~r0) P~ (~r0) 3
(~r) =
d ~r0
|~r ~r0|3

I shall now denote ~r ~r0 = ~ .


Here comes the first nasty trick, we can notice that the following is true:
 

1
= 2
0
|~ |

r
r

Therefore the last integral can be written as :


 
Z
1
(~r) = P~ 0
d3~r0
|~ |

Yet a second nasty trick would be to use integration by parts over the last expression to
obtain:
!
Z
Z 
I ~
Z


P ~
P~
1
1
3
3
~
(~r) = 0
d ~r0
0 P d ~r0 =
ds0
0 P~ d3~r0

The sensible question after this is why? The answer is that we can get a physical
interpretation of this expression which makes sense. We can define two sizes which
could interest us in the following manner:
b P~ n

where n
is a unity vector perpendicular to the surface of the dielectric, and we can define:
b P~
These two give some meaning to this polarization vector.
With these notations we get:
I
Z
b
b 3
(~r) =
ds0 +
d ~r0
S

It means that the potential of a dielectric can be expressed or represented by a corresponding surface charge distribution on its surface, plus a a volumetric bound charge
density.
19

1.6.2

Electric Displacement - Gauss Law and Dielectrics

We have shown that within dielectrics we have dipoles that can create charge accumulations , and due to it electric potential as a reaction to external fields. These reactions
obviously create fields too (since they create potentials). These interactions meddle
with the fields within the dielectric. The compound field is usually of more interest (on
a practical point of view) to us the the original since this is what we can measure.
We can divide the charge density within the dielectric to the density due to polarization
- bound charge, and the density due to everything else - the free charge:
= b + f
using this devision in Maxwells first equation reads:


~ = 4 = 4 (b + f ) = 4 P~ + f
E
Switching sides we can write:


 
~ + 4 P~ = D
~ = 4f
E
where we define:
~ E
~ + 4 P~
D
as the Electric Displacement.

Be Careful!
There is a tricky point regarding the electric displacement. One could think that in
dielectrics we solve problems just the same as we would solve usually, and only switch
~ to D.
~ This is wrong. We have defined the displacement using divergence,
to f and E
but the divergence alone doesnt define a vector quantity. it defines only its monopole
sources but not the curl sources. If we try to evaluate its curl we arrive at:
~ =E
~ + 4 P~ = 4 P~ 6= 0
D
~ isnt 0 in general, and therefore it cannot be evaluated by the divergence
The curl of D
alone.

1.6.3

Boundary Conditions in Dielectrics

If we look back at the derivation of the boundary conditions a few sections back we
notice that the same logic can be implemented here, but since things are slightly different
(mainly P~ 6= 0) so the conditions too change:
~k
~k
~k
~k
D
above Dbelow = Pabove Pbelow
20

For the perpendicular component we can write:

Dabove
Dbelow
= 4f

1.6.4

Linear Dielectrics

Many substances can be regarded as linear dielectrics. Such a behavior means that there
is a linear relationship in the form of:
~
P~ = e E
If such a relation exists we can write:
~ =E
~ + 4 P~ = (1 + 4e ) E
~ = E
~
D
We call e the electric susceptibility, and  is the permittivity. In such cases we also
look at this from another angle:
 
4e
e ~
b = P~ =
D =
f

1 + 4e
and therefore when there is no free charge density there is no bound charge density
either. This doesnt imply anything about the bound charge though over the surface b ,
only about the volume density.
We can also rewrite the boundary conditions as:

= f
below Ebelow
above Eabove

Or:

above
below
below
= f
n

Of course the potential is still continuous and above = below .


above

1.6.5

Energy in Dielectrics

In dielectrics the energy density of space can be written as:


Z
Z
1
1
2
~ E
~ dV
W =
E dV =
D
8
8
In order to explain it lets suppose that we have a dielectric and we add an incremental
amount of free charge f . We try to analyze what is the work that will be done on
this free charge.
Z
W =

(f ) dV
21

We can use the previous relations to achieve:


Z


1
~ ] dV
W =
[ D
4
Using the product rule we obtain:
Z

Z 



1
~ ] dV +
~ E
~ dV
W =
[ D
D
4
As we have done before with the energy, turning the first integral into a surface integral,
and integrating over the whole space makes the first integral vanish
Now if the dielectric is linear we can write:
 
 

1 ~ ~ 1
~ E
~
~ E
~ = D
D E = E 2 = E
2
2
Using this:

W =
And at last

1
8

1
W =
8

~ E
~ dV
D

~ E
~ dV
D

all space

22

Chapter 2

Magnetostatics
2.1

Lorentz Force

Lorentz force gives us an expression for the force which acts upon a given charge, and
by super-position on configurations of charges. We shall not prove that this expression
is really the right expression. We assume its validity, and this axiom is one of the
foundations of electrodynamics.
For a point charge the force take the form of:


~v
~
~
~
FL = q E + B
c
One important point which we can notice already, and which is sometimes forgotten
is that the magnetic force never does any work, as it is perpendicular to the velocity,
and therefore its power is zero.

2.1.1

Currents

Current is a physical size which describes the flow of charges. It relates both to the
amount of charge and to its velocity. We look at it either as:
I~ = ~v
where is the linear charge distribution and ~v is the charge velocity, or as:
dq
dt
The difference between definitions can be confusing, especially since in one the current
is a vector, and in the other its a scalar. Both can be used, and we always need to
remember the basic meaning and adjust the math accordingly.
If we try to find now an expression for the Lorentz force which acts upon a current
we may write:
Z
Z
Z
1
1 ~ ~
1
~
~
~
Fmag =
~v B dq =
~v B dl =
I B dl
c
c
c
I=

23

Usually the magnitude of the current is constant along the wire and we get:
Z
I ~
~
~
Fmag =
dl B
c
Sometimes we have a surface current, and then our definition using the linear charge
density doesnt hold. This charge density is special because as we shall see later on it
defines boundary conditions just like surface charge. we shall define the surface current
density as:
~
~ dI
K
dl
or in a more friendly manner:
~ = ~v
K
We could now write the Lorentz force for such a current as:
Z
Z
Z
1
1 ~
1
~
~
~ da
~
~v B dq =
~v B da =
K B
Fmag =
c
c
c

Caution!
As Ive written, and as well see soon, the magnetic field suffers a discontinuity at at
surface current. Therefore a question can be asked which field to use. The correct answer
is to use the average field in the integral.
Last we shall define a volume current density. We denote this one as we always have
as:

dI~
J~
da

or in a more friendly manner:


J~ = ~v
The resulting Loretz force will be:
Z
Z
Z
1
1
1 ~ ~
~
~
~
Fmag =
~v B dq =
~v B dV =
J B dV
c
c
c

2.1.2

Charge Conservation

If we use the definition of J~ we could get the following:


Z
Z
I = J da = J~ d~a
S

24

Using Gauss theorem we arrive at:


I

J~ d~a =

J~ dV

Since charge is conserved we can always say that what flows from the source, must
leave the source (or using the time derivative definition) we can write:
Z
Z
Z

J~ dV =
dV =
dV
t
t
V
and we arrive at the differential charge conservation:

J~ =
t

2.2

The Biot-Savart Law

Until now we have discussed how charges are affected by a magnetic field, but how do
you create such a field?
A magnetic field is created by moving charges (current). In this chapter since we limit
ourselves to the case of magnetostatics nothing varies with time. This statement includes
the charge density and therefore we can state:

= 0 J~ = 0
t

2.2.1

The Vector Potential

We can either start building the theory from Biot Savart and state that:
~ (~r) = 1
B
c

Z ~
J (~r0) (~r ~r0) 3
d ~r0
|~r ~r0|3

Another option is to begin from Maxwells equations and arrive at Biot Savart, we shall
try this.
Beginning with Maxwell eq. we start with:
~ =0
B
~ =
B

4 ~
J
c

~ doesnt have a divergence we may claim that we can write it as a curl of another
Since B
~
vector. We shall denote this vector potential by A.
~ =A
~
B
25

~ uniquely we can choose how to calibrate it. Putting A


~ into
Since we didnt define A
Maxwells fourth equation we can see that we get:


~ A
~ = 4 J~
grad div A
c
~ = 0, will make our
Looking at the last expression it is easy to see why the choice div A
life easier. This choice is called the Coulomb Gauge. In order to justify this choice we
should show that finding such a vector potential is always feasible.
Coulomb Gauge Proof
~ 0 . Now lets
Lets assume that we found a vector potential which isnt divergenceless - A
~
~ 0 + .
add to it the gradient of a scalar function . So we get a new potential: A = A
The new divergence will be:
~ =A
~ 0 + 2
A
~ 0 we can find such so
We can prove coulombs gauge then by proving that for every A
that:
~0
2 = A
~0
Mathematically speaking the last row is just the regular poisson equation where A
is the source. We have already proven that solutions for this equation can always be
found at least in theory.
This concludes our proof that Coulombs gauge doesnt violate any assumption , and
doesnt put any restrictions on our soultion.
Using this choice we can get a simple version of the equation stated before:
~=
A

4 ~
J
c

If we write this equation in cartesian coordinates we get three simple poisson equations, one for each coordinate. Therefore we can understand that the solution for this
differential equation is poisson integral:
~ (~r) = 1
A
c

J~
d3~r0
|~r ~r0|

For line or surface currents we could write then:


Z
1
I
~
A (~r) =
d~`
c
|~r ~r0|
~ (~r) = 1
A
c

26

~
K
ds
|~r ~r0|

~ according to this we will get:


If we take the curl of A,
1
c

~ =A
~ =
B

!
J~
d3~r0
|~r ~r0|

Since we integrate with respect to the primed coordinates, and take the derivative with
respect to the unprimed we can take the derivative into the integral and write:
!
Z
Z
1
J~
1 J~ (~r0) (~r ~r0) 3
3

d ~r0 =
d ~r0
c
|~r ~r0|
c
|~r ~r0|3
(We dont need to take the derivative of J, as its a function of the primed coordinates.)
So now we got Biot Savart law from the vector potential

27

(5.72)

As for the tangential components, an amperian loop running perpendicular to the current
(Fig. 5.50) yields

B dl

= fL01enc = fLo KI ,

2.3 or Boundary Conditions


II
B above

I
Bbe10w

fLo K .

(5.73)

The magnetic field works in an opposite way to the electric. We shall now see why, and
Thus the component of B that is parallel to the surface but perpendicular to the current is
whatdiscontinuous
are the boundary
conditions for the magnetic field.
in the amount fLo K. A similar amperian loop running parallel to the current
reveals that the parallel component is continuous. These results can be summarized in a
single formula:

2.3.1

Continuity in theBabove
Normal
Field
- Bbelow
= fLoCK x n),

(5.74)

The normal
is vector
continuous
acrosstoboundaries.
To see"upward."
why lets look at a small, and
where n isfield
a unit
perpendicular
the surface, pointing
~
thin box, which is halved by a surface which holds a surface current K.
B1above

Applying Maxwells third eq.


I

2.3.2


5.49
~ Figure
B
= 0 we get:

~ = 0 B
~ ds
B
above = Bbelow

Discontinuity in the Tangential Field

In order to see why the tangential field isnt continuous we shall look at an infinitesimal
CHAPTER 5. MAGNETOSTATICS
current242
loop, and calculate the circulation around it.

Applying Maxwells fourth eq. we get: Figure 5.50


I

k
~ = Bk
~ d`
B
B
above

below

`=

4
K`
c

Like the scalar potential in electrostatics, the vector potential is continuous across
boundary:
(5.751
Aabove = Abelow,

28

for V . A = 0 guarantees 15 that the normal component is continuous, and V x A = B, in


the form

fA, dl

B . da

<1>,

means that the tangential components are continuous (the flux through an amperian loop of
vanishing thickness is zero). But the derivative of A inherits the discontinuity of B:

aAabove
an

aAbe!ow
an

- - - = -ftoK.

(5.761

2.3.3

Conclusion

Adding that up we can write:


~ above B
~ below = 4 K
~ n
B

c
where n
is a unit vector perpendicular to the surface, and pointing above.
Regarding the potential we can say that the potential is continuous across boundaries
~ = 0 guarantees the normal component, and if we
(using Coulomb gauge) because A
calculate the circulation:
I
Z
~
~ = m
~
~ ds
A d` = B

Now if we take the length of

29

S-ar putea să vă placă și