You are on page 1of 82

Physics 232:

Classical Electromagnetism

Val Anthony V. Balagon

vvbalagon@up.edu.ph

March-June 2021
Contents

4 Multipoles, Electrostatics of Macroscopic Media, Dielectrics 3


4.1 Multipole Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4.2 Multipole Expansion of the Energy of a Charge Distribution in an External
Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Elementary Treatment of Electrostatics with Ponderable Media . . . . . . . 20
4.4 Boundary-Value Problems in Dielectrics . . . . . . . . . . . . . . . . . . . . 26
4.5 A Dielectric in a Uniform Electric Field . . . . . . . . . . . . . . . . . . . . . 27
4.6 Electrostatic Energy in Dielectric Media . . . . . . . . . . . . . . . . . . . . 27

5 Magnetostatics 28
5.1 Biot-Savart Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Differential Equations of Magnetostatics and Ampere’s Law . . . . . . . . . 32
5.3 The Vector Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.4 Magnetic Fields of a Localized Current Distribution, Magnetic Moment . . . 39
5.5 Magnetic Dipole of a Current Loop . . . . . . . . . . . . . . . . . . . . . . . 44
5.6 Force, Torque and Energy of a Localized Current Distribution in a External
Magnetic Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.7 Magnetic Field of a Charged Rotating Sphere . . . . . . . . . . . . . . . . . 47
5.8 Macroscopic Equations, Boundary Conditions on the Magnetic Induction and
Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.8.1 Hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.9 Methods of Solving Boundary-Value Problems in Magnetostatics. . . . . . . 54
5.10 Magnetic Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.11 Faraday’s Law of Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6 Maxwell Equations, Macroscopic Electromagnetism, Conservation Laws 58


6.1 Green Functions for the Wave Equation . . . . . . . . . . . . . . . . . . . . . 58
6.2 Poynting’s Theorem and Conservation of Energy and Momentum for a System
of Charged Particles and Electromagnetic Fields . . . . . . . . . . . . . . . . 61
6.3 Poynting’s Theorem in Linear Dispersive Media with Losses . . . . . . . . . 66

1
7 Plane Electromagnetic Waves and Wave Propagation 69
7.1 Plane Waves in a Nonconducting Medium . . . . . . . . . . . . . . . . . . . 69
7.2 Linear and Circular Polarization; Stokes Parameters . . . . . . . . . . . . . . 73
7.3 Reflection and Refraction of Electromagnetic Waves at a Plane Interface Be-
tween Dielectrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.3.1 E is perpendicular to Plane of Incidence . . . . . . . . . . . . . . . . 76
7.3.2 E is Parallel to Plane of Incidence . . . . . . . . . . . . . . . . . . . . 79

2
Chapter 4

Multipoles, Electrostatics of
Macroscopic Media, Dielectrics

4.1 Multipole Expansion


*
The potential due to a charge distribution at some position x is

ˆ *0
* 1 ρ(x ) *0
Φ(x) = * *0
d3 x (4.1)
4π0 |x − x |

We recall the Green’s function expansion

∞ X
` `
1 X 1 r<
* *0
= 4π Y ∗ (θ0 , φ0 )Y`,m (θ, φ)
`+1 `,m
(4.2)
|x − x | `=0 m=−`
2` + 1 r>

*
where r< is the lesser of r and r0 , and r> is the greater of r and r0 . Suppose that x is outside
the volume of interest so r< = r0 and r> = r. We plug the expansion to the potential. We
get

ˆ ∞ X
`
* 1 *0 X 1 (r0 )` ∗ 0 0 3 *0
Φ(x) = ρ(x )4π `+1
Y `,m (θ , φ )Y `,m (θ, φ) d x
4π0 `=0 m=−`
2` + 1 r
∞ ` ˆ
1 (r0 )` ∗ 0 0 3 *0
 
1 X X *0
= ρ(x ) Y (θ , φ ) d x Y`,m (θ, φ)
0 `=0 m=−`
2` + 1 r`+1 `,m
∞ X ` ˆ 
* 1 X 1 0 ` *0 ∗ 0 0 3 *0 Y`,m (θ, φ)
Φ(x) = (r ) ρ(x )Y`,m (θ , φ ) d x (4.3)
0 `=0 m=−`
2` + 1 r`+1

We let
ˆ
*0 *0
q`,m = ρ(x )(r0 )` Y`,m

(θ0 , φ0 ) d3 x (4.4)

where the integration is done over the volume of the charge distribution. The expansion
coefficients q`,m are called the multipole moments, which is analogous to moments in the

3
context of probability. It is easy to show that it has the property


q`,−m = (−1)m q`,m . (4.5)

So the multipole expansion formula for the electric potential is then

∞ `
* 1 X X q`,m Y`,m (θ, φ)
Φ(x) = (4.6)
0 `=0 m=−` 2` + 1 r`+1

We compute for the first few q`,m starting with q0,0 .

ˆ
*0 *0
q0,0 = ρ(x )(r0 )0 Y0,0

(θ0 , φ0 ) d3 x
ˆ
1 *0 *0
=√ ρ(x ) d3 x

q
q0,0 =√ (4.7)

where q is the total charge of the distribution. The ` = 1 term refer to the dipole correction.
Before we begin with computing the expansion coefficients, we recall the electric dipole
moment
ˆ
* *0 * *0
p≡ x ρ(x) d3 x . (4.8)

For q1,1

r ˆ
3 *0 0 *0
q1,1 = − ρ(x )r0 e−iφ sin θ0 d3 x

r ˆ
3 *0 *0
=− ρ(x )r0 (cos φ0 − i sin φ0 ) sin θ0 d3 x

r ˆ
3 *0 *0
=− ρ(x )r0 (sin θ0 cos φ0 − i sin θ0 sin φ0 ) d3 x

r ˆ ˆ 
3 *0 0 0 0 3 *0 *0 0 0 0 3 *0
=− ρ(x )r sin θ cos φ d x − i ρ(x )r sin θ sin φ d x .

But x0 = r0 sin θ0 cos φ0 and y 0 = r0 sin θ0 sin φ0 so

r ˆ ˆ 
3 *0 0 3 *0 *0 0 3 *0
=− ρ(x )x d x − i ρ(x )y d x .

´ *0 *0 ´ *0 *0
ρ(x )x0 d3 x is the x−component of the electric dipole moment and ρ(x )y 0 d3 x is the
y−component of the dipole moment. Thus,

r
3
q1,1 = − (px − ipy ). (4.9)

4
For q1,0

ˆ
*0 *0
q1,0 = ρ(x )r0 Y1,0

(θ0 , φ0 ) d3 x
r ˆ
3 *0 *0
= ρ(x )r0 cos θ0 d3 x

But z 0 = r0 cos θ0

r ˆ
3 *0 *0
= ρ(x )z 0 d3 x

The integral is the z−component of the electric dipole moment.

r
3
q1,0 = pz (4.10)

The ` = 2 term is called the quadrupole correction. It is useful that we define a 3 × 3


matrix  
ˆ Q1,1 Q1,2 Q1,3 
*0 *0
(3x0i x0j − r02 δi,j )ρ(x ) d3 x ,
 
Qi,j ≡ Q= Q
 2,1 Q 2,2 Q2,3  ,
 (4.11)
 
Q3,1 Q3,2 Q3,3

called the quardupole moment tensor. The quadrupole tensor is symmetric. We switch
indices i  j,

ˆ
 *0
Qj,i = 3x0j x0i − (r0 )2 δj,i ρ(x ) dτ 0

Since δi,j = δj,i ,

ˆ
 *0
= 3x0i x0j − (r0 )2 δi,j ρ(x ) dτ 0

Qj,i = Qi,j , Symmetric

So Q12 = Q21 , Q13 = Q31 , Q23 = Q32 . Hence, the quadrupole moment tensor is just
 
Q1,1 Q1,2 Q1,3 
 
Q= Q
 1,2 Q2,2 Q 2,3  .
 (4.12)
 
Q1,3 Q2,3 Q3,3

Then, another property is that it is a traceless matrix

3
X
trace(Qi,j ) = Qi,i
i=1

5
3 ˆ
X *0  *0
= ρ( r ) 3x0i x0i − (r0 )2 δi,i d3 x
i=1
ˆ 3
*0 X  *0
= ρ( r ) 3(x0i )2 − (r0 )2 d3 x
ˆ i=1
*0  *0
= ρ( r ) 3(x01 )2 − (r0 )2 ) + (3(x02 )2 − (r0 )2 ) + (3(x03 )2 − (r0 )2 ) d3 x
ˆ
*0  *0
= ρ( r ) 3((x01 )2 + (x02 )2 + (x03 )2 ) − (3r0 )2 ) d3 x
ˆ
*0 0 2  0 2
 0 3 *0

:
= ρ( r ) 3(r )− 3(r ) d x


trace(Qi,j ) = 0

Since we have Q1,1 + Q2,2 + Q3,3 = 0, then Q3,3 = −Q1,1 − Q2,2 . So the sum of the diagonal
of Qi,j is zero. This property reduces the number of elements of Q to compute. For q2,2

ˆ
*0 *0
q2,2 = ρ(x )(r0 )2 Y2,2

(θ0 , φ0 ) d3 x
r ˆ
1 15 *0 0 *0
= ρ(x )(r0 )2 e−2iφ sin2 θ0 d3 x
4 2π
r ˆ
1 15 *0 0 *0
= ρ(x )(r0 e−iφ sin θ0 )2 d3 x
4 2π
r ˆ
1 15 *0 *0
= ρ(x )(x0 − iy 0 )2 d3 x
4 2π
r ˆ
1 15 *0 *0
= ρ(x )(x02 − 2ix0 y 0 − y 02 ) d3 x
4 2π
r ˆ ˆ ˆ 
1 15 *0 02 3 *0 *0 0 0 3 *0 *0 02 3 *0
= ρ(x )x d x − 2i ρ(x )x y d x − ρ(x )y d x
4 2π

´ *0 *0 ´ *0 *0
We call ρ(x )x02 d3 x the second moment of x0 , ρ(x )x0 y 0 d3 x is the first moment of x0 y 0 ,
´ *0 *0
and ρ(x )y 02 d3 x the second moment of y 0 .

r
1 15
q2,2 = (Q1,1 − 2iQ1,2 − Q2,2 ) (4.13)
12 2π

For q2,1

ˆ
*0 *0
q2,1 = ρ(x )(r0 )2 Y2,1

(θ0 , φ0 ) d3 x
r ˆ
15 *0 0 *0
=− ρ(x )(r0 )2 sin θ0 cos θ0 e−iφ d3 x

r ˆ
15 *0 0 *0
=− ρ(x )(r0 cos θ0 )(r0 e−iφ sin θ0 ) d3 x

r ˆ
15 *0 *0
=− ρ(x )z(r0 sin θ0 cos φ0 − r0 sin θ0 sin φ0 ) d3 x

r ˆ
15 *0 *0
=− ρ(x )z 0 (x0 − iy 0 ) d3 x

r
1 15
q2,1 =− (Q1,3 − iQ2,3 ) (4.14)
3 8π

6
For q2,0 ,

ˆ
*0 *0
q2,0 = ρ(x )(r0 )2 Y2,0

(θ0 , φ0 ) d3 x
r ˆ
1 5 *0  *0
= ρ(x )(r0 )2 3 cos2 θ0 − 1 d3 x
2 4π
r ˆ
1 5 *0   *0
= ρ(x ) 3(r0 cos θ0 )2 − (r0 )2 d3 x
2 4π
r ˆ
1 5 *0   *0
= ρ(x ) 3z 02 − (r0 )2 d3 x
2 4π
r
1 5
q2,0 = Q3,3 . (4.15)
2 4π

To compute for q1,−1 , q2,−1 , q2,−2 we use Equation 4.5.

r
∗ 3
q1,−1 = (−1)1 q1,1 = (px + ipy ) (4.16)

r
∗ 1 15
q2,−1 = (−1)1 q2,1 = (Q1,3 + iQ2,3 ) (4.17)
3 8π
r
∗ 1 15
q2,−2 = (−1)2 q2,2 = (Q1,1 + 2iQ1,2 − Q2,2 ). (4.18)
12 2π

Then the first few terms of the multipole expansion is

1 2
* q0,0 Y0,0 (θ, φ) X q1,m Y1,m (θ, φ) X q2,m Y2,m (θ, φ)
0 Φ(x) = 0+1
+ 1+1
+ 2+1
+...
2
| · 0 + 1{z r } m=−1 2 · 1 + 1 r m=−2
2 · 2 + 1 r
`=0 term (monopole) | {z } | {z }
`=1 term (dipole) `=2 term (quadrupole)
1 2
q0,0 1 X 1 X
= Y0,0 (θ, φ) + 2 q1,m Y1,m (θ, φ) + q2,m Y2,m (θ, φ) + . . .
r 3r m=−1
5r3 m=−2

Expanding,

q0,0 1
= Y0,0 (θ, φ) + 2 [q1,−1 Y1,−1 (θ, φ) + q1,0 Y1,0 (θ, φ) + q1,1 Y1,1 (θ, φ)] +
r 3r
1
[q2,−2 Y2,−2 (θ, φ) + q2,−1 Y2,−1 (θ, φ) + q2,0 Y2,0 (θ, φ) + q2,1 Y2,1 (θ, φ) + q2,2 Y2,2 (θ, φ)] + . . .
5r3

We then plug in the corresponding expansion coefficients and spherical harmonics. We shall
do this term-by-term. First for the monopole term

q
Φ0 = (4.19)
4π0 r

The monopole term treats the charge distribution ρ as a point charge at the origin. There
will be higher order corrections due to the distribution not being a point charge. Next, for

7
the dipole term

1
Φ1 = [q1,−1 Y1,−1 (θ, φ) + q1,0 Y1,0 (θ, φ) + q1,1 Y1,1 (θ, φ)]
30 r2
"r r r r
1 3 3 3 3
= 2
(px + ipy )e−iφ sin θ + pz cos θ−
30 r 8π 8π 4π 4π
r r #
3 3
(−1) (px − ipy )eiφ sin θ
8π 8π
 
1 3 −iφ 3 3 iφ
= (px + ipy )e sin θ + pz cos θ + (px − ipy )e sin θ
30 r2 8π 4π 8π
 
1 3  iφ −iφ iφ −iφ
 3
= sin θ px (e + e ) − ipy (e − e ) + pz cos θ
30 r2 8π 4π
 
1 3 3
= sin θ (px · 2 cos φ − ipy · 2i sin φ) + pz cos θ
30 r2 8π 4π
1 3 r
= 2
· [px sin θ cos φ + py sin θ sin φ + pz cos θ] ·
30 r 4π r
1
= [px r sin θ cos φ + py r sin θ sin φ + pz r cos θ]
4π0 r3

Recall that x = sin θ cos φ, y = r sin θ sin φ, z = r cos θ

1
= [px x + py y + pz z]
4π0 r3

*
The term inside the square brackets is just a dot product between the dipole moment p and
*
the position vector x

* *
p·x
Φ1 = . (4.20)
4π0 r3

Lastly, we compute for the quadrupole term Φ2 .

1
0 Φ2 = [q2,−2 Y2,−2 (θ, φ) + q2,−1 Y2,−1 (θ, φ) + q2,0 Y2,0 (θ, φ) + q2,1 Y2,1 (θ, φ) + q2,2 Y2,2 (θ, φ)]
5r3

We evaluate each term in the parenthesis separately.

r r
1 15 1 15 −2iφ 2
q2,−2 Y2,−2 (θ, φ) = (Q1,1 + 2iQ1,2 − Q2,2 ) · e sin θ
12 2π 4 2π
5
= (Q1,1 + 2iQ1,2 − Q2,2 )(e−iφ sin θ)2
32π
5 h r i2
= (Q1,1 + 2iQ1,2 − Q2,2 ) (sin θ cos φ − i sin θ sin φ) ·
32π r
5
= (Q1,1 + 2iQ1,2 − Q2,2 )(r sin θ cos φ − ir sin θ sin φ)2
32πr2

We recall that x = r sin θ cos φ and y = r sin θ sin φ

5
q2,−2 Y2,−2 (θ, φ) = (Q1,1 + 2iQ1,2 − Q2,2 )(x − iy)2
32πr2

8
Next,

r r
1 15 1 15 −iφ
q2,−1 Y2,−1 (θ, φ) = (Q1,3 + iQ2,3 ) · e sin θ cos θ
3 8π 2 2π
5 r2
= (Q1,3 + iQ2,3 )(cos φ − i sin φ) sin θ cos θ · 2
8π r
5
= (Q1,3 + iQ2,3 )(r sin θ cos φ − ir sin θ sin φ)(r cos θ)
8πr2

Recall that z = r cos θ

5
q2,−1 Y2,−1 (θ, φ) = (Q1,3 + iQ2,3 )(x − iy)z
8πr2

Next,

r r
1 5 1 5
q2,0 Y2,0 (θ, φ) = Q3,3 · (3 cos2 θ − 1)
2 4π 4 π
5 r2
= (3 cos2 θ − 1) · 2 Q3,3
16π r
5
= (3r2 cos2 θ − r2 )Q3,3
16πr2
5
= (3z 2 − (x2 + y 2 + z 2 ))Q3,3
16πr2
5
q2,0 Y2,0 (θ, φ) = 2
(2z 2 − x2 − y 2 )Q3,3
16πr

Next,

r r
1 15 1 15 iφ
q2,1 Y2,1 (θ, φ) = − (Q1,3 − iQ2,3 ) · − e sin θ cos θ
3 8π 2 2π
5 r2
= (Q1,3 − iQ2,3 )(cos φ + i sin φ) sin θ cos θ · 2
8π r
5
= (Q1,3 − iQ2,3 )(r sin θ cos φ + ir sin θ sin φ)(r cos θ)
8πr2
5
q2,1 Y2,1 (θ, φ) = (Q1,3 − iQ2,3 )(x + iy)z
8πr2

Lastly,

r r
1 15 1 15 2iφ 2
q2,2 Y2,2 (θ, φ) = (Q1,1 − 2iQ1,2 − Q2,2 ) · e sin θ
12 2π 4 2π
5
= (Q1,1 − 2iQ1,2 − Q2,2 )(eiφ sin θ)2
32π
5 h r i2
= (Q1,1 − 2iQ1,2 − Q2,2 ) (sin θ cos φ + i sin θ sin φ) ·
32π r
5
= (Q1,1 − 2iQ1,2 − Q2,2 )(r sin θ cos φ + ir sin θ sin φ)2
32πr2
5
q2,2 Y2,2 (θ, φ) = 2
(Q1,1 − 2iQ1,2 − Q2,2 )(x + iy)2
32πr

9
Summing them all up

0 5r3 · Φ2 = q2,−2 Y2,−2 (θ, φ) + q2,2 Y2,2 (θ, φ)+

q2,−1 Y2,−1 (θ, φ) + q2,1 Y2,1 (θ, φ)+

q2,0 Y2,0 (θ, φ)


5 5
= 2
(Q1,1 + 2iQ1,2 − Q2,2 )(x − iy)2 + 2
(Q1,1 − 2iQ1,2 − Q2,2 )(x + iy)2 +
32πr 32πr
5 5
(Q1,3 + iQ2,3 )(x − iy)z + (Q1,3 − iQ2,3 )(x + iy)z+
8πr2 8πr2
5
(2z 2 − x2 − y 2 )Q3,3
16πr2
5  2
x (Q1,1 − Q2,2 ) + 4xyQ1,2 + y 2 (Q2,2 − Q1,1 ) +

= 2
16πr
5z 5
2
(xQ 1,3 + yQ 2,3 ) + 2
(2z 2 − x2 − y 2 )Q3,3
4πr 16πr

Simplifying

5 
0 5r3 · Φ2 = 4x(yQ1,2 + zQ1,3 ) + 4yzQ2,3 + x2 (Q1,1 − Q2,2 − Q3,3 ) +
16πr2
2z 2 Q3,3 − y 2 (Q1,1 − Q2,2 + Q3,3 )


16π0 r5 · Φ2 = 4xyQ1,2 + 4yzQ2,3 + 4xzQ1,3 + x2 Q1,1 + y 2 Q2,2 + z 2 Q3,3 +

x2 (−Q2,2 − Q3,3 ) + y 2 (−Q1,1 − Q3,3 ) + z 2 Q3,3

= 4xyQ1,2 + 4yzQ2,3 + 4xzQ1,3 + x2 Q1,1 + y 2 Q2,2 + z 2 Q3,3 +

− x2 Q2,2 − y 2 Q1,1 + (−x2 − y 2 + z 2 )Q3,3

We use Q3,3 = −Q1,1 − Q2,2

= 4xyQ1,2 + 4yzQ2,3 + 4xzQ1,3 + x2 Q1,1 + y 2 Q2,2 + z 2 Q3,3 +

− x2 Q2,2 − y 2 Q1,1 + (−x2 − y 2 + z 2 )(−Q1,1 − Q2,2 )

= 4xyQ1,2 + 4yzQ2,3 + 4xzQ1,3 + x2 Q1,1 + y 2 Q2,2 + z 2 Q3,3 +

x2 Q1,1 + y 2 Q2,2 + z 2 Q3,3

= 2xyQ1,2 + 2yxQ1,2 + 2yzQ2,3 + 2zyQ2,3 + 2xzQ1,3 + 2zxQ1,3 +

2x2 Q1,1 + 2y 2 Q2,2 + 2z 2 Q3,3

But since the quadrupole tensor is symmetric Q1,2 = Q2,1 , Q1,3 = Q3,1 ,

16π0 r3 · Φ2 = 2xyQ1,2 + 2yxQ2,1 + 2yzQ2,3 + 2zyQ3,2 + 2xzQ1,3 + 2zxQ3,1 +

2x2 Q1,1 + 2y 2 Q2,2 + 2z 2 Q3,3

10
Rearranging

16π0 r3 · Φ2 = 2x2 Q1,1 + 2xyQ1,2 + 2xzQ1,3 + 2yxQ2,1 + 2y 2 Q2,2 + 2yzQ2,3 +

2zxQ3,1 + 2zyQ3,2 + 2z 2 Q3,3

We can rewrite the RHS in terms of the following summation:

3 X
X 3
16π0 · Φ2 = 2 xi xj Q1,j
i=1 j=1

where x1 = x, x2 = y, x3 = z. Hence, the quadrupole contribution is

3 3
1 X X xi xj
Φ2 = Qi,j . (4.21)
4π0 r3 i=1 j=1 2

Therefore, the potential can be expressed as the following expansion

* *
" #
* 1 q p · x 1 X x x
i j
Φ(x) = + 3 + Qi,j 5 + . . . (4.22)
4π0 r r 2 i,j r

∞ `
* 1 X X q`,m Y`,m (θ, φ)
Φ(x) = (4.23)
0 `=0 m=−` 2` + 1 r`+1

Normally we wouldn’t go beyond ` = 2 since it would be too cumbersome. To get the electric
field, we take the negative gradient of Equation 4.6

* * *
E(x) = −∇Φ
* * ∂Φ 1 ∂Φ 1 ∂Φ
−E(x) = r̂ + θ̂ + φ̂
∂r r ∂θ r sin θ ∂φ
" ∞ `
#
∂ 1 X X q`,m Y`,m (θ, φ)
= `+1

∂r 0 `=0 m=−`
2` + 1 r
" ∞ `
#
1 ∂ 1 X X q`,m Y`,m (θ, φ)
+ θ̂
r ∂θ 0 `=0 m=−` 2` + 1 r`+1
" ∞ `
#
1 ∂ 1 X X q`,m Y`,m (θ, φ)
+ φ̂
r sin θ ∂φ 0 `=0 m=−` 2` + 1 r`+1

We note that the multipole moments q`,m (r0 , θ0 , φ0 ) are not functions of (r, θ, φ)

∞ X
`    
* * X q`,m ∂ 1 1 ∂Y`,m (θ, φ) 1 ∂Y`,m (θ, φ)
−0 E(x) = Y`,m (θ, φ) r̂ + `+2 θ̂ + `+2 φ̂
`=0 m=−`
2` + 1 ∂r r`+1 r ∂θ r sin θ ∂φ

11
∂Y`,m (θ,φ) ∂Y`,m (θ,φ)
We note that ∂φ
= imY`,m (θ, φ). We will not evaluate ∂θ
because the resulting
equation is too long to write down.

∞ X
`  
* * X q`,m (` + 1) 1 ∂Y`,m (θ, φ) im
E(x) = Y`,m (θ, φ) r̂ − `+2 θ̂ − `+2 Y`,m (θ, φ) φ̂
`=0 m=−`
0 (2` + 1) r`+2 r ∂θ r sin θ

(4.24)

From this, the components of the electric field can be written as

∞ X
`
X (` + 1) Y`,m (θ, φ)
Er = q`,m
`=0 m=−`
0 (2` + 1) r`+2
∞ X
`
X 1 ∂
Eθ = − q`,m `+2
Y`,m (θ, φ) (4.25)
`=0 m=−`
0 (2` + 1)r ∂θ
∞ X
`
X im
Eφ = − q`,m Y`,m (θ, φ)
`=0 m=−`
0 (2` + 1)r`+2 sin θ

*
Electric Field of a Dipole Located at x0
*
We shall use Equation 4.20 to compute for the electric field E due to a dipole. We can
also use Equation 4.25 to compute for the electric field of a dipole but it will be more
time-consuming.
* *
!
* * * * p·x
E dip (x) = −∇Φ = −∇
4π0 r3

We shall use spherical coordinates

* * ∂Φ 1 ∂Φ 1 ∂Φ
−E dip (x) = r̂ + θ̂ + φ̂
∂r r ∂θ r sin θ ∂φ
:0
     
* * ∂ pr cos θ 1 ∂ pr cos θ 1 ∂ pr cos θ

−4π0 E dip (x) = r̂ + θ̂ +  
φ̂
∂r r3 r ∂θ r3 r sin θ 
∂φ r3
 
 
∂ 1 p ∂
= p cos θ 2
r̂ + 3 (cos θ) θ̂
∂r r r ∂θ
 
−2 p
= p cos θ 3
r̂ + 3 (− sin θ) θ̂
r r
* * 2p cos θ p sin θ
E dip (x) = 3
r̂ + θ̂ (4.26)
4π0 r 4π0 r3

Reading off the components


2p cos θ
Edip,r =
4π0 r3
p sin θ
Edip,θ = (4.27)
4π0 r3
Edip,φ = 0

We can convert Equation 4.26 into coordinate-free form.

* *
* * 1 3n̂( p · n̂) − p
E dip (x) = (4.28)
4π0 r3

12
*
We can generalize this by placing the electric dipole at some arbitrary point x0 . So, the
electric field is now
* *
* * 1 3n̂( p · n̂) − p
E dip (x) = (4.29)
4π0 |* *
x − x0 |3

*
Example 4.1. Consider a localized charge distribution ρ(x) that gives rise to an electric
*
field throughout space. We wish to calculate the integral of E over the volume of a sphere
of radius R. We choose the origin to be the center of the sphere. We consider two extreme
cases, where the charge distribution is (a) inside and (b) outside the sphere.

ˆ * *
ˆ *
3* * *
E(x) d x = − ∇Φ(x) d3 x
V V

We use the divergence theorem,

ˆ
* *
=− Φ(x) d a
S

*
where a = n̂R2 dΩ. n̂ is the direction outward from the surface of the sphere. Then

ˆ ˆ *0
1 ρ(x ) *0 *
=− * *0
d3 x d a
4π0 S V |x − x |
ˆ 2 ˆ ˆ
* * * R *0 3 *0 n̂
E(x) d3 x = − ρ(x ) d x * *0
dΩ
V 4π0 V S |x − x |

Invoking the multipole expansion

ˆ ˆ ˆ ∞ X
` `
* *
3* R2 *0 3 *0
X 1 r<
E(x) d x = − ρ(x ) d x 4π Y ∗ (θ0 , φ0 )Y`,m (θ, φ)n̂ dΩ
`+1 `,m
V 4π0 V S `=0 m=−`
2` + 1 r>
(4.30)
We note that the unit vector n̂ in Cartesian coordinates written in terms of the spherical
angles is
n̂ = sin θ cos φ î + sin θ sin φ ĵ + cos θ k̂.

As expected, |n̂| = 1. We need to write n̂ in terms of a linear combination of spherical


harmonics. We list down the ` = 1 spherical harmonics.

r r r
1 3 −iφ 1 3 1 3 iφ
Y1,−1 (θ, φ) = e sin θ, Y1,0 (θ, φ) = cos θ, Y1,1 (θ, φ) = − e sin θ
2 2π 2 π 2 2π

13
We note the orthogonality condition

ˆ π ˆ 2π
Y`,m Y`∗0 ,m0 dΩ = δ`,`0 δm,m0
θ=0 φ=0

The easiest to re-express is the z-component. Looking at the table of spherical harmonics

r r
3 4π
Y1,0 (θ, φ) = cos θ ⇒ cos θ = Y1,0 (θ, φ)
4π 3

Next, we use a linear combination of Y1,−1 (θ, φ) and Y1,1 (θ, φ)

sin θ cos φ = aY1,−1 (θ, φ) + bY1,1 (θ, φ) (4.31)


We find for a and b. Starting with a, we multiply both sides by Y1,−1 (θ, φ)

∗ ∗ ∗
Y1,−1 (θ, φ) sin θ cos φ = aY1,−1 (θ, φ)Y1,−1 (θ, φ) + bY1,1 (θ, φ)Y1,−1 (θ, φ)

We then integrate over the solid angle.

ˆ ˆ :1

 ˆ :0

 
∗ ∗ ∗ 
Y1,−1 (θ, φ) sin θ cos φ dΩ
=a Y1,−1
(θ,φ)Y
1,−1 (θ, φ) dΩ +b Y1,1 (θ,
 φ)Y


1,−1 (θ, φ) dΩ
  
r 

a=
3


Next, we find for b. We multiply Equation 4.31 by Y1,1 (θ, φ)

∗ ∗ ∗
Y1,1 (θ, φ) sin θ cos φ = aY1,−1 (θ, φ)Y1,1 (θ, φ) + bY1,1 (θ, φ)Y1,1 (θ, φ)

Integrating over the solid angle, r



b=− .
3

Thus, we can write sin θ cos φ as

r

sin θ cos φ = [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] (4.32)
3

Next, we do the same for sin θ sin φ. We can write sin θ sin φ as

sin θ sin φ = cY1,−1 (θ, φ) + dY1,1 (θ, φ). (4.33)

We find that c and d are r r


2π 2π
c=i , d=i .
3 3

14
Hence, r

sin θ sin φ = i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] . (4.34)
3

So, n̂ in terms of spherical harmonics is

r r r
2π 2π 4π
n̂ = [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] î + i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + Y1,0 (θ, φ) k̂.
3 3 3
(4.35)

We plug this to Equation 4.30.

ˆ ˆ " ∞ X
` ` ˆ #
* *
3* R2 *0 3 *0
X 1 r<
E(x) d x = − ρ(x ) d x Y ∗ (θ0 , φ0 )
`+1 `,m
Y`,m (θ, φ)n̂ dΩ
V 0 V `=0 m=−`
2` + 1 r> S
(4.36)
We focus on the Ω integral first.

Continued at the next page.

15
ˆ ∞ X
` ` ˆ (r r r )
X X 1 r< 2π 2π 4π
[. . .] dΩ [. . .] = Y ∗ (θ0 , φ0 )
`+1 `,m
Y`,m (θ, φ) [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] î + i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + Y1,0 (θ, φ) k̂ dΩ
`,m S `=0 m=−`
2` + 1 r> S 3 3 3

Due to orthogonality, the ` = 1 terms only survive.

r 1 ˆ √
2π r< X 1 ∗ 0 0
n o
= Y (θ , φ ) Y 1,m (θ, φ) [Y1,−1 (θ, φ) − Y 1,1 (θ, φ)] î + i [Y 1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + 2Y 1,0 (θ, φ) k̂ dΩ
3 r> 2
m=−1
3 1,m S
r  ˆ √
1 2π r< ∗ 0 0
n o
= 2
Y1,−1 (θ , φ ) Y1,−1 (θ, φ) [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] î + i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + 2Y1,0 (θ, φ) k̂ dΩ +
3 3 r>
ˆ S
n √ o
∗ 0 0
Y1,0 (θ , φ ) Y1,0 (θ, φ) [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] î + i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + 2Y1,0 (θ, φ) k̂ dΩ +
ˆS n √ o 
∗ 0 0
Y1,1 (θ , φ ) Y1,1 (θ, φ) [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] î + i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + 2Y1,0 (θ, φ) k̂ dΩ
S

∗ ∗
Recall the property Y`,m (θ, φ) = (−1)m Y`,−m (θ, φ). We apply the complex conjugate to the said property, so Y`,m (θ, φ) = (−1)m Y`,−m (θ, φ).

r  ˆ √
1 2π r< ∗ 0 0
n

  ∗
 o
= 2
Y 1,−1 (θ , φ ) Y 1,−1 (θ, φ) Y 1,−1 (θ, φ) − (−1)Y 1,−1 (θ, φ) î + i Y1,−1 (θ, φ) + (−1)Y 1,−1 (θ, φ) ĵ + 2Y 1,0 (θ, φ) k̂ dΩ +
3 3 r>
ˆ S
n √ o
∗ 0 0 ∗
Y1,0 (θ , φ ) Y1,0 (θ, φ) [Y1,−1 (θ, φ) − Y1,1 (θ, φ)] î + i [Y1,−1 (θ, φ) + Y1,1 (θ, φ)] ĵ + 2(−1)0 Y1,0 (θ, φ) k̂ dΩ +
ˆS n √ o 
∗ 0 0 ∗ ∗
  
Y1,1 (θ , φ ) Y1,1 (θ, φ) −Y1,1 (θ, φ) − Y1,1 (θ, φ) î + i −Y1,1 (θ, φ) + Y1,1 (θ, φ) ĵ + 2Y1,0 (θ, φ) k̂ dΩ
S
16
We use the orthogonality condition of spherical harmonics.

r  ˆ √ ∗ 0 0 ˆ
1 2π r< ∗ 0 0

∗ ∗


= 2
Y1,−1 (θ , φ ) Y1,−1 (θ, φ) Y1,−1 (θ, φ) î − iY1,−1 (θ, φ) ĵ dΩ + 2Y1,0 (θ , φ ) Y1,0 (θ, φ)Y1,0 (θ, φ) k̂ dΩ +
3 3 r> S S
ˆ   
∗ 0 0 ∗ ∗
Y1,1 (θ , φ ) Y1,1 (θ, φ) −Y1,1 (θ, φ) î − iY1,1 (θ, φ) ĵ dΩ
S
√ ∗ 0 0
r
1 2π r< h ∗ 0 0 ∗ 0 0
i
= 2
Y 1,−1 (θ , φ )( î − i ĵ) + 2Y 1,0 (θ , φ ) k̂ − Y 1,1 (θ , φ )(î + i ĵ)
3 3 r>
√ ∗ 0 0 i
r
1 2π r< h ∗ 0 0 ∗ 0 0
 ∗ 0 0 ∗ 0 0

= 2
Y 1,−1 (θ , φ ) − Y 1,1 (θ , φ ) î − i Y 1,−1 (θ , φ ) + Y 1,1 (θ , φ ) ĵ + 2Y1,0 (θ , φ )k̂
3 3 r>
"r #

r r r
1 2π r< 3 3 1 3
= 2
sin θ0 cos φ0 î − i × i sin θ0 sin φ0 ĵ + 2 × cos θ0 k̂
3 3 r> 2π 2π 2 π
1 r< h 0 0 0 0 0
i
= 2
sin θ cos φ î + sin θ sin φ ĵ + cos θ k̂
3 r>
ˆ
X 1 r< 0
[. . .] dΩ [. . .] = 2
n̂ (4.37)
`,m S 3 r>
17
Then we plug Equation 4.37 to Equation 4.36

ˆ ˆ
* *
3*R2 *0 r< 0 3 *0
E(x) d x = − ρ(x ) 2
n̂ d x (4.38)
V 30 V r>

a) For the first case, where the sphere completely encloses the charge distribution, r< = r0
and r> = R. Then,

ˆ ˆ
* *
3* R2 *0 r0 0 3 *0
E(x) d x = −

ρ(x ) n̂ d x
V 30 V
ˆ R2

1 *0 *0
=− ρ(x )r0 n̂0 d3 x
30 V

*0
We note that r0 n̂0 = r . So

ˆ
1 *0 *0 *0
=− r ρ(x ) d3 x
30 V

*
The integral is the expression for the electric dipole moment p. So if the bounding
*
surface of the charge density ρ(x) is a sphere, then the volume integral of the electric
field is
ˆ *
* *
3* p
E(x) d x = − . (4.39)
r<R 30

We note that the size of the bounding surface does not change the value of the volume
integral. In order to be consistent with Equation 4.39, we must update Equation 4.29
as
* *
" #
* * 1 3n̂( p · n̂) − p 4π * * *0
E dip (x) = * * − pδ(x − x0 ) . (4.40)
4π0 |x − x0 |3 3

So if we place a pure dipole of magnitude p inside the volume V , the electric field must
be the expression above.

b) For the second case, where the charge distribution is exterior to the bounding surface,
then r< = R and r> = r0 .

ˆ ˆ 0
* *
3* R3 *0 n̂ *0
E(x) d x = − ρ(x ) 02 d3 x
V 30 V r
ˆ * *0
R3 *0 0 − n 3 *0
= ρ(x ) * d x
30 V | 0 − x̂0 |3

* * *
The integral is just Coulomb’s law where the position vector x is at the origin x = 0.

ˆ * * * 4πR3 * * *
E(x) d3 x = E(x = 0)
V 30

So the average value of the electric field over a spherical volume containing no chage

18
is the value of the field at the center of the sphere.

4.2 Multipole Expansion of the Energy of a Charge

Distribution in an External Field

We place a charge density ρ at an external potential Φ. This means that the energy is

ˆ
* * *
W = ρ(x)Φ(x)d3 x (4.41)

We choose the origin of the coordinate system to be inside the charge distribution. We
perform a multivariate Taylor expansion1 on the external potential Φ:

* * * * * 1 XX ∂ 2Φ
Φ(x) = Φ( 0) + x · ∇Φ( 0) + xi xj (0) + · · ·
2 i j ∂xi ∂xj

* *
∂Φ
Since E = −∇Φ, or in component form Ei = − ∂x i

* * * * * 1 XX ∂
Φ(x) = Φ( 0) − x · E( 0) − xi xj Ej (0) + · · ·
2 i j ∂xi

* * * *
Since ∇ · E = 0, then we can add a quantity 61 r2 ∇ · E( 0) = 0 without changing the sum.
We rewrite it in component form,

1 2* * * 1 X ∂ 1 XX ∂
r ∇ · E( 0) = r2 Ei (0) = r2 Ej (0)δi,j = 0
6 6 i
∂xi 6 i j
∂xi

So,

* * * 1 XX
* * ∂ 1 XX ∂
Φ(x) = Φ( 0) − x · E( 0) − xi xj Ej (0) + r2 Ej (0)δi,j + · · ·
2 i j ∂xi 6 i j
∂xi
XX 1 r2 ∂

* * * * ∂
= Φ( 0) − x · E( 0) + − xi xj Ej (0) + Ej (0)δi,j + · · ·
i j
2 ∂xi 6 ∂xi

* * * * * 1 XX  ∂
Φ(x) = Φ( 0) − x · E( 0) − 3xi xj + r2 δi,j Ej (0) + · · · (4.42)
6 i j ∂xi

*
We then plug Φ(x) into W , we obtain

ˆ " #
* *1 XX * * * ∂ *
3xi xj + r2 δi,j Ej (0) + · · · d3 x

W = ρ(x) Φ( 0) − x · E( 0) −
6 i j ∂xi
ˆ ˆ  ˆ
* * 3* * * 3* * * 1 XX ∂ *  *
= Φ( 0) ρ(x)d x − ρ(x)xd x · E( 0) − Ej (0) ρ(x) 3xi xj + r2 δi,j d3 x + · · ·
6 i j ∂xi
1
Check https://math.stackexchange.com/questions/2648512/taylor-expansion-for-vector-valued-function

19
The first integral corresponds to the total charge of the system, we denote it as q, while the
*
second and third integrals correspond to the dipole moment p and the quadrupole moment
tensor Qi,j , respectively.

* * * * 1 XX ∂
W = qΦ( 0) − p · E( 0) − Qi,j Ej (0) + · · · (4.43)
6 i j ∂xi

In a way, one reads this as the charge q interacts with the potential, the dipole moment
interacts with the electric field, and the quadrupole tensor interacts with the electric field
gradient.
* *
What is the interaction energy between two dipoles p 1 and p 2 ? We note that the
electric field of a dipole is given by Equation 4.40. We know that dipoles do not have
* * * *
monopole and quadrupole moments. We place p 1 at x1 and p 2 at x2 . It is necessary
* * *
that x1 6= x2 . In the context of this section, the charge distribution is p 1 and the thing that
*
gives the external field is p 2 .

* * * *
W12 = − p 1 (x1 ) · E 2 (x1 )
* *
" #
* 1 3n̂1,2 ( p 2 · n̂1,2 ) − p 2 4π * * * :

 0
= −p1 · − 1 − x2 )

* * p2
δ(
x
4π0 |x1 − x2 | 3 3
* * * *
p 1 · p 2 − 3( p 1 · n̂1,2 )( p 2 · n̂1,2 )
W12 = * * (4.44)
4π0 |x1 − x2 |3

* *
where n̂1,2 is the direction of x1 − x2 . Conversely, we can compute for W21 by switching the
* *
roles of p 1 and p 2

* * * *
W21 = − p 2 (x2 ) · E 1 (x2 )
* *
" #
* * 1 3n̂2,1 ( p 1 · n̂2,1 ) − p 1 4π * * * :0

= − p 2 (x2 ) · − p δ(
x2 − x1 )
4π0 * *
|x2 − x1 |3 3 1
* * * *
1 p 2 · p 1 − 3( p 2 · n̂2,1 )( p 1 · n̂2,1 )
W21 = * *
4π0 |x2 − x1 |3

We see that W12 = W21 , as expected (it is easy to show that the second terms of W1,2 and
W2,1 are equivalent).

4.3 Elementary Treatment of Electrostatics with Pon-

derable Media

In physics, ponderable means to have a detectable amount of matter. In E&M, a ponderable


medium is a substance with detectable amount of matter that does not conduct electricity.
So basically a ponderable medium is an insulator.

20
We note that vaccuum is a perfect insulator but it is not a ponderable medium. Air is also
an insulator but since it isn’t dense enough we do not consider it as a ponderable medium.
In this section, we only consider ponderable media that are dielectrics. All dielectrics are
insulators, but not all insulators are dielectrics.
A dielectric is a ponderable media whose charge distribution can be slightly deformed -
or in technical terms, polarized - by an external electric field. Before we proceed, let us
differentiate macroscopic and microscopic electric fields.

Consider an arbitrary charge distribution ρ. We then choose some volume ∆V inside the
charge distribution, which contains a large number of charges. The individual charges sets
*
up an electric field, which we will call the microscopic electric field E micro,i . Averaging the
*
electric field contributions from each charge determines the macroscopic electric field E macro .
We find that
* * * * *
∇ × E micro = 0 → E micro = −∇Φmicro .

Since the macroscopic electric field is just the average of the microscopic electric fields, thus

* * * * *
∇ × E macro = 0 → E macro = −∇Φmacro .

* *
From now on we will name the macroscopic electric field as E macro = E. Suppose that
we have a dielectric and the external electric field is turned off. The macroscopic electric
field will be setup by free charges ρ or net charges in the medium. When we turn on an
*
external electric field E ext the molecules of the dielectric composed of many atoms will be
*
deflected/polarized. Suppose that E ext is positive in magnitude, the electron cloud (which is
negatively charged) will be attracted to the direction of the external field, while the positively
charged nucleus (due to the proton) will be repelled. In other words, there is a net separation
of charges.
By applying an external electric field the dielectric in response has an in-
duced effective dipole moment due to the redistribution of the charges in the
medium. Each dipole will now setup an electric field. We define the electric polarization

21
Figure 4.1: Each atom in the medium has a positive charge (nucleus)
and a negative charge (electron). By applying an external electric
field the nucleus and electron cloud are deflected.

*
or polarization density at point x as

* * X *
P (x) = Ni h p i i
i

*
where Ni is the average number of molecules per unit volume, and h p i i is the average dipole
*
moment for the ith molecule. The of the electric polarization P is in dipole moment per
volume. Then the free charge density is

* X
ρ(x) = Ni hei i + ρexcess ,
i

where hei i is the average net charge for each molecule, and ρexcess are the charges that are
*
dislodged from the orbital clouds of the atoms. We call ρ(x) the free charge.

*
The goal is to find the electric field (macroscopic) at some point x inside the distribution.
Consider a small volume ∆V in some arbitrary charge distribution. The free charges and
the dipoles in this small volume contribute a potential ∆Φ.

* * *0
" #
* 1 q p · (x − x )
∆Φ(x) = *0
+ * *0
4π0 |*
x−x| |x − x |3

22
* *
The total charge q in ∆V is just ρ∆V , while the net dipole moment p in ∆V is just P ∆V .

" * * *0
#
1 (ρ∆V ) (P ∆V ) · (x − x )
= *0
+ *0
4π0 |*
x−x|
*
|x − x |3
" * * *0
#
* 1 ρ P · (x − x )
∆Φ(x) = *0
+ * *0 ∆V
4π0 |*
x−x| |x − x |3

The term inside the square brackets above is the multipole expansion up to the dipole con-
tribution where the higher order do not have any contributions. To get the full contribution
from all molecules we perform a linear superposition of all ∆Φi . The sum now becomes an
integral.

ˆ " *0 * *0 * *0
#
* 1 ρ(x ) P (x ) · (x − x ) *0
Φ(x) = * *0
+ * *0
d3 x
4π0 |x − x | |x − x |3

* *0
x−x
We can write * *0 as
| x − x |3

* *0
!
x−x n̂ *0 1
* *0
= * *0
=∇ * *0
|x − x |3 |x − x |2 |x − x |

* *0
where n̂ is the separation vector of x and x . Then

ˆ ˆ *0
!
* 1 ρ 1 3 *0
* 1 *0
Φ(x) = * *0
d x + P ·∇ * *0
d3 x
4π0 |x − x | 4π0 |x − x |
| {z }

We use the following on the second term

* ! ! !
*0 P 1 *0 * * *0 1
∇ · * *0
= * *0
∇ ·P +P ·∇ * *0
|x − x | |x − x | |x − x |
| {z }

Then we plug this back to Φ,

ˆ ˆ *0
 * ! *

* 1 ρ *0 1 *0 P ∇ · P  3 *0
Φ(x) = * *0
d3 x + ∇ ·
* *0
− * *0
d x
4π0 |x − x | 4π0 |x − x | |x − x |

We use the divergence theorem on the first term inside the square brackets.

ˆ ˆ * ˆ *0 *
* 1 ρ 1 3 *0 P · n̂ 1 ∇ ·P *0
Φ(x) = * *0
d x + * *0
da0 − * *0
d3 x
4π0 |x − x | 4π0 S |x − x | 4π0 |x − x |

where n̂ is unit vector normal to the surface S. By integrating over all space, the surface

23
integral vanishes. Thus, the second term cancels out. Therefore, the potential is

ˆ *0 *0 * *0
* 1 ρ(x ) − ∇ · P (x ) *0
Φ(x) = * *0
d3 x (4.45)
4π0 all space |x − x |

* *
To find the electric field, we use E = −∇Φ,

ˆ  *0
 !
* * * 1 *0 * *0 * 1 *0
E = −∇Φ(x) = − ρ(x ) − ∇ · P (x ) ∇ * *0
d3 x
4π0 |x − x |

We take the divergence

ˆ  *0
 !
* 1
* *0 * *0 1 *0
∇·E =− ρ(x ) − ∇ · P (x ) ∇2 * *0
d3 x
4π0 |x − x |

 
1 * *0
But ∇2 * *0 = −4πδ(x − x ),
|x−x |

ˆ  * 0 * *0

1 *0 * *0 *0
=+ ρ(x ) − ∇ · P (x ) δ(x − x ) d3 x (4.46)
0
* * * * * *
∇ · E = ρ(x) − ∇ · P (x) (4.47)

* *
This looks like a modification of Gauss’s law. If P (x) is non-uniform there can be a net
increase or decrease of charge within any small volume. We define the electric displacement
*
D as
* * *
D ≡ 0 E + P . (4.48)

* *
Re-arranging ∇ · E,
* * * * * *
0 ∇ · E + ∇ · P (x) = ρ(x)

Therefore,
* * *
∇ · D = ρ(x) . (4.49)

*
This is known as the Gauss’s law for dielectrics. A special case is when the polarization P
*
is parallel to the macroscopic electric field E.

* *
P = 0 χe E. (4.50)

All media that follows the equation above are called linear dielectrics. Then the electric
displacement is

* * *
D = 0 E + P
* *
= 0 E + (0 χe E)

24
*
= 0 (1 + χe )E
* *
D = E , (4.51)

where χe is called the electric susceptibility and  = 0 (1+χe ) is called the electric permittivity.

The quantity 0
is called the dielectric constant. Generally, the electric permittivity can vary
*
based on position  = (x), but for simpler media the properties are uniform all throughout.
This means that the electric permittivity is a constant,  = const. Plugging this to the
Gauss’s law for dielectrics, we get a familiar equation

* * ρ
∇·E = (4.52)


Suppose we have two distinct media identified by electric permittivities 1 (right) and 2
(left).

Figure 4.2: The boundary of two distinct media.

Using the integral form of Equation 4.52, we apply a Gaussian pillbox at the interface to
produce a boundary condition.

ˆ * *
D · da = q

As the width becomes smaller we are only left with the perpendicular components. For the
RHS the total charge at the pillbox is q = σA

(D2⊥ − D1⊥ )A = σA

We cancel the arbitrarily small area A both sides. We define a normal vector to the interface
n̂2,1 , which points towards the second medium. So,

* *
(D 2 − D 1 ) · n̂2,1 = σ (4.53)

* *
Next, we take the integral form of curl of the macroscopic electric field, ∇ × E = 0, where

25
the path is a square loop.
˛ * *
E ·dl = 0

By making the width smaller, the contribution at the sides go to zero leaving us with

k k
E2 − E1 = 0

In terms of n̂2,1
* *
(E 2 − E 1 ) × n̂2,1 = 0 (4.54)

4.4 Boundary-Value Problems in Dielectrics

* *
Consider the system described by the figure above. We want to find E and D at the two
media. We note that there is no surface charge distribution at the interface. This problem
can be solved by the method of images. Listing down the relevant Maxwell equation at the
regions

• Region I:
* *
1 ∇ · E 1 = ρ = qδ(x)δ(y)δ(z − d) (4.55)

• Region II:
* *
2 ∇ · E 2 = 0 (4.56)

We apply Equation 4.53 at the interface

* * 0
(D 2 − D 1 ) · n̂2,1 = 
σ
>

* *
(2 E 2 − 1 E 1 ) · (−k̂) = 0

−2 E2,z + 1 E1,z = 0

26
Thus,

2 E2,z = 1 E1,z (4.57)

at the boundary. Next, we apply Equation 4.54

* *
(E 2 − E 1 ) × (−k̂) = 0
* *
E 2 × k̂ = E 1 × k̂
*0

 *0


E2,x î × k̂ + E2,y ĵ × k̂ + E2,z  × k̂
k̂  = E1,x î × k̂ + E1,y ĵ × k̂ + E1,z  × k̂
k̂ 

We can see that


E2,x = E1,x , E2,y = E1,y (4.58)

at the boundary.
[Continue]

4.5 A Dielectric in a Uniform Electric Field

[Section 4.4]

4.6 Electrostatic Energy in Dielectric Media

[In OneNote]

27
Chapter 5

Magnetostatics

In this chapter we consider charges in motion with respect to a reference frame. The moving

charges exerts a force field that interacts with the moving charge. This force field is embodied
*
by the magnetic-flux density B.
* * *
F = Qv × B (5.1)

In a certain frame of reference we may only see the magnetic field. Generally, the force on
a given charge is
* * * *
F = Q(E + v × B) (5.2)

However, in magnetostatics we only consider the magnetic-flux density.

* *
* µ0 v × x
B≡ q 3 (5.3)
4π |*x |

*
Charges in motion are described by the current density J . We consider charges in motion
along a surface. Then we consider a small area δA that is perpendicular to the current. Over
some time δt, the amount of charge δQ passes through δA.

δQ
J=
δA δt

28
Then at a length δ`

δQ δ` δQ δ` δQ δ`
J= · = · = ·
δA δt δ` δA δ` δt δV
|{z} δt
|{z}
charge density velocity

Therefore,
* *
J = ρ v. (5.4)

As expected the current density flows in the direction of the velocity of the charges. Charge
must be conserved. So, the charge density must obey a continuity equation:

∂ρ * *
+∇·J =0 (5.5)
∂t

∂ρ
In magnetostatics, we consider charge densities that do not change in time, ∂t
= 0 (steady
current). Implying
* *
∇·J =0 (5.6)

If this is not zero we are dealing with a time-dependent charge distribution.

5.1 Biot-Savart Law

We consider a current density in a wire of a small cross-sectional area. The total current
passing through the wire is defined as

ˆ * *
I≡ J · dS.

29
We consider a small cylindrical volume with surface area A and length d`. We know that
* *
the current passing through this small volume is I = J · A. We multiply both sides by d`

* *
I d` = J · Ad`
*
= ρ v · δV n̂
*
= ρδV v · n̂
*
I d` = δq v · n̂

*
But v, n̂ are parallel
I d`
I d` = δqv → v= (5.7)
δq
*
This means that a charge δq is moving in velocity v. This small charge then sets up a small
*
magnetic field δ B,

* *
* µ0 v × x
δB = δq *3
4π |x |
* *
µ0 δq I d ` × x
= 3
4π δq |*

x |
* *
* µ0 I d ` × x
δB = 3
4π |* x |

We integrate to obtain all contributions

˛ * *
* µ0 I d` × x
B= *3
. (5.8)
4π |x |

The reasoning for the closed integral is because the wire must be closed (no charge is created
or destroyed). Open wires do not sustain a flow of charges since charge is destroyed at the
end. We consider two circuits of current I1 and I2 , respectively. We want to calculate the

force on circuit 2 due to circuit 1. The infinitesimal force on a small line element in circuit

30
1 is due to an external magnetic field produced by circuit 2:

* * *
dF = Id ` 1 × B 2

We integrate over the loop in circuit 1

*
˛ * *
F = I1 d ` 1 × B 2
˛ ˛ * *
* µ0 d ` 2 × x1,2
= I1 d ` 1 × I2 *
4π |x1,2 |3
˛ ˛ * * *
µ0 I1 I2 d ` 1 × d ` 2 × x1,2
= *
4π |x1,2 |3

* * * * * *
We can use the triple product rule for cross products d ` 1 × d ` 2 × x1,2 = d ` 2 (d ` 1 · x1,2 ) −
* * *
(d ` 1 · d ` 2 )x1,2

˛ ˛
µ0 I1 I2 1 h * * * * * * i
= * d ` 2 (d ` 1 · x1,2 ) − (d ` 1 · d ` 2 )x1,2
4π |x1,2 |3
˛ ˛ * * * ˛ ˛
 
µ0 I1 I2  d ` 2 (d ` 1 · x1,2 ) 1 * * *
= * − * (d ` 1 · d ` 2 )x1,2 
4π |x1,2 |3 |x1,2 |3

¸ *
But we know that d ` 2 = 0. Therefore, the force on circuit 1 due to circuit 2 is

˛ ˛ *
*µ0 I1 I2 x1,2 * *
F =− * d`1 · d`2 . (5.9)
4π |x1,2 |3

Example 5.1. Let us obtain the force per unit length on wire 1.

The differential force on a small portion d`1 of wire 1 due to wire 2 is

ˆ ∞ *
* µ0 * * x12
dF 1 = − I1 I2 d`1 · d`2 *
4π −∞ |x12 |3
ˆ ∞ *
µ0 x12
= − I1 I2 d`1 dx *
4π −∞ |x12 |3

31
* * √
But x12 = −x î − d ĵ and so, |x12 | = x2 + d 2

* ˆ ∞
dF 1 µ0 −x î − d ĵ
= − I1 I2 dx 2
d`1 4π (x + d2 )3/2
" ˆ −∞ ˆ ∞
:0
#

µ0 xdx  
 dx
= I1 I2 î + d ĵ

2 2 3/2 2 2 3/2
4π (x
−∞ +d ) −∞ (x + d )
ˆ ∞
µ0 dx
= I1 I2 d 2 2 3/2

4π −∞ (x + d )
 
µ0 2
= I1 I2 d ĵ
4π d2
*
dF 1 µ0 I1 I2
= ĵ
d`1 2πd

We are presented with two cases.



* 
dF 1 Attractive, if I1 I2 > 0 (the current flow in the same direction).

= (5.10)
d`1 
Repulsive,
 if I1 I2 < 0 (the currents flow in the opposite direction).

* * *
If a current density J (x) is in an external magnetic-flux density B(x), the elementary
force law implies that the total force on the distribution is

*
ˆ * * * *
F = J (x) × B(x) d3 x (5.11)

and the total torque is


*
ˆ * * * *
*
N= x × (J (x) × B(x)) d3 x (5.12)

5.2 Differential Equations of Magnetostatics and Am-

pere’s Law

Recall that for a current-carrying wire the magnetic field is given by the Biot-Savart law

* *
*µ0 Id ` × x
dB = .
4π |* x|3

32
We now consider a bulk and find the magnetic field it set up. The current in the bulk can
* * *
be expressed as Id ` = J dA d `

* *
µ0 J dA d` × x
*
dB = *
4π |x|3

* * *0
We let x → x − x

* * *0
µ0 J × ( x − x )
= *0
dA d`
4π |*x − x |3

We integrate over all contributions, we get the general form of the magnetic induction field
* *
for any current density J (x)

ˆ * *0
* * µ0 * *0 x−x
B(x) = J (x ) × * *0
d3 x0 (5.13)
4π |x − x |3

* *0
x−x
But the expression * *0 can be expressed as
|x−x |

* *0
x−x * 1 * 1
* *0
= −∇ *
x * *0
= ∇*
x
0
* *0
|x − x | |x − x | |x − x |

So,

ˆ !
* * µ0 * *0 * 1
B(x) = J (x ) × −∇*
x * *0
d 3 x0
4π |x − x |

* * * * * *
[WHY] We shall use the identity ∇ × (ψ A) = ψ(∇ × A) + (∇ψ) × A

* *0
* J (x ) 1 * * * :0 * 1 * *
 0

∇*
x
× * *0
= * *0

 *J (a )

x
+ ∇ *
x * *0
× J (x)
|x − x | |x − x | |x − x |

So,
ˆ * *0
* µ0 * J (x )
B= ∇* × * *0
d 3 x0 (5.14)
4π x |x − x |
We know that the divergence of a curl is always zero. So,

" ˆ * *0 #
* * µ0 * * J (x ) !
∇·B = ∇ · ∇*
x
× * *0
d3 x0 = 0
4π |x − x |

Therefore,
* *
∇·B =0. (5.15)

33
*
Next, we take the curl of B.

" ˆ * *0 #
* µ0 *
* * J (x ) 3 0
∇×B = ∇* × ∇* × * *0
dx
4π x x
|x − x |

* * * * * * *
We shall use the identity ∇ × (∇ × A) = ∇(∇ · A) − ∇2 A,

* *0 * *0 ! * *0
* * J (x ) * * J (x ) J (x )
∇*
x
× ∇*
x
× * *0
= ∇*
x
∇*
x
· * *0
− ∇2 * *0
|x − x | |x − x | |x − x |

We use the following

* 1 * 1 1 * *0
−∇ *
x * *0
= ∇*
x
0
* *0
and ∇2 * *0 = −4πδ(x − x )
|x − x | |x − x | x−x

Hence,
ˆ
* µ0 *
* * *0 * 1
∇×B =− ∇ J (x )∇*
x * *0
d3 x0
4π |x − x |
[Fill-up please] We do integration by parts we get the second Maxwell’s equation in
magnetostatics.
* * * *
∇ × B = µ0 J (x) (5.16)

A consequence of this is

5.3 The Vector Potential


* * * * *
We solve for B of a bulk where J = 0. Since ∇ × B = 0, it is implied that we can write B
in terms of a scalar potential ΦM .

* * * *
∇×B =0 → B = −∇ΦM (5.17)

* *
And since ∇ · B = 0, we obtain a Laplace’s equation

* * * *
∇ · B = −∇ · ∇ΦM = 0

−∇2 ΦM = 0 (5.18)

We can obtain the magnetic field using the magnetic scalar potential from the Laplace’s
* *
equation in the absence of a current density J . Generally, we cannot do this if J 6= 0. We
* * * *
note that ∇ · B = 0 implies that we can express B in terms of a vector potential A:

* * *
B =∇×A (5.19)

34
We compare this to Equation 5.14 we can see that

ˆ * *0
* µ0 J (x ) *
A= * *0
d3 x0 + ∇ψ (5.20)
4π |x − x |

where ψ is an arbitrary scalar. This term is needed because the curl of a gradient is always
* * *
zero, ∇ × (∇ψ) = 0. This implies that A is not unique. It can be determined uniquely by
* * *
J (x) up to an arbitrary ∇ψ. If we choose to reformulate magnetostatics in terms of the
* *
vector potential A, then the transformation of A given by

* * *
A → A + ∇ψ (5.21)

does not in any way alter the solutions to the field equations. This transformation is known
* *
as a gauge transformation. We have a freedom to choose A(x) to be the most convenient
* * * * *
form of A(x). From ∇ × B = µ0 J

* * * *
∇ × (∇ × A) − µ0 J

* *
If ∇ · A 6= 0, then we can always perform a gauge transformation

* *0 * *
A → A = A + ∇φ (5.22)

* *0
with the requirement that ∇ · A = 0. Then,

* *0
∇·A =0
* * *
∇ · (A + ∇φ) = 0
* *
∇ · A + ∇2 φ = 0

Hence,
* *
∇2 φ = − |∇{z
· A} . (5.23)
not zero

This has an infinite number of solutions. We can make a convenient choice of gauge (known
* *
as the Coulomb gauge) ∇ · A = 0 so that

* *
∇2 A = −µ0 J . (5.24)

We write this in component form

∇2 Ax = µ0 Jx

35
∇2 Ay = µ0 Jy

∇2 Az = µ0 Jz

We can use the Green’s function method to find the general solution of Equation 5.24. We
* *0 * *0 * *0 1
note that ∇2 G(x, x ) = −4πδ(x − x ) → G(x, x ) = * *0 . Each component has the
|x−x |

solution
ˆ *
* µ0 Jk (x)
Ak (x) = * *0
d 3 x0 , k = x, y, z (5.25)
4π |x − x |
Or in vector form,
ˆ * *
* * µ0 J (x)
A(x) = * *0
d3 x0 . (5.26)
4π |x − x |
*
Example 5.2. Consider a circular loop with constant current density J . We must find the
magnetic potential at some point P .

Solution

The current density is just

*
J = Jφ φ̂ = −Jφ sin φ0 î + Jφ cos φ0 ĵ

We must find Jφ . In a current loop, we generally have vdq = Id`. We integrate along the
loop.

ˆ ˛
v dq = Id`

Qv = I(2πa)
I
v= (2πa)
Q

We know that the velocity of the v charges are constant which implies that the current is
also constant throughout the loop. This means that the charge density is

Qδ(r0 − a)δ(cos θ0 )
ρ= .
2πa2

So,
Qvδ(r0 − a)δ(cos θ0 ) I Iδ(r0 − a)δ(cos θ0 )
Jφ ≡ ρv = · (2πa) =
2πa2 Q a

We note that δ(cos θ0 ) is equivalent to

π
δ(cos θ0 ) = sin δ(cos θ0 ) = sin θ0 δ(cos θ0 )
2

36
Plugging this fact to Jφ
I sin θ0 δ(r0 − a)δ(cos θ0 )
Jφ =
a
*
We plug J to the vector potential

ˆ * *0
* µ0 J (x )
A= * *0
d3 x0
4π |x − x |
ˆ
µ0 −Jφ sin φ0 î + Jφ cos φ0 ĵ
= * *0
d 3 x0
4π |x − x |

*
The loop has cylindrical symmetry. To simplify the integration we place x at the xz−plane.
*
So, x = (x, 0, z). Since there will be no contribution in the x−component (because φ0 = 0
at the xz-plane), the vector potential will only have a φ component

ˆ
µ0 Jφ cos φ0
Aφ (r, θ, 0) = * *0
d 3 x0
4π |x − x |
ˆ
µ0 I sin θ0 δ(r0 − a)δ(cos θ0 ) cos φ0 02
= · * *0 r sin θ0 dΩ0 dr0
4π a |x − x |
ˆ 2π ˆ a ˆ π
µ0 I 0 0 0 02 0 sin2 θ0 δ(cos θ0 )
= dφ cos φ dr r δ(r − a) dθ0 2
4πa 0 0 0 [r + r02 − 2rr0 (cos θ cos θ0 + sin θ sin θ0 ) cos φ0 ]1/2
ˆ ˆ a
µ0 I 2π 0 0 0 r02 δ(r0 − a)
= dφ cos φ dr p
4πa 0 0 r2 + r02 − 2r0 r sin θ cos φ0
ˆ 2π
µ0 I 0 0 a2
= dφ cos φ p
4πa 0 r2 + a2 − 2r0 r sin θ cos φ0

We are left with

ˆ 2π
µ0 Ia cos φ0
Aφ (r, θ, 0) = dφ0 (5.27)
4π 0 (a2 + r2 − 2ar sin θ cos φ0 )1/2

This cannot be expressed in elementary functions. Instead we write this in terms of the
complete elliptic integrals K and E.

(2 − k 2 )K(k) − 2E(k)
 
µ0 4Ia
Aφ (r, θ) = √ (5.28)
4π a2 + r2 + 2ar sin θ k2

where
4ar sin θ
k2 = (5.29)
a2 + r2 + 2ar sin θ

We note that the complete elliptic integrals of the first and second kind are given by

ˆ π/2 ˆ π
1 2
q
K(k) = p dϕ, E(k) = 1 − k 2 sin2 ϕ dϕ (5.30)
0 1 − k 2 sin2 ϕ 0

37
*
Now that we have the vector potential, we can now find the magnetic induction B.


r̂ θ̂ r sin θ φ̂

* 1 ∂ ∂ ∂

B= 2
r sin θ ∂r ∂θ ∂φ

0 0 Aφ r sin θ

1 ∂ 1 ∂
= (Aφ sin θ) r̂ + 2 (−1) (rAφ ) θ̂ + 0
r2 sin θ ∂θ r sin θ ∂r

The components are

1 ∂
Br = (Aφ sin θ) (5.31)
r2
sin θ ∂θ
1 ∂
Bθ = − 2 (rAφ ) (5.32)
r sin θ ∂r
Bφ = 0 (5.33)

We can express Aφ as a power series

µ0 Ia2 r sin θ 15a2 r2 sin2 θ


 
Aφ = 1+ + ··· (5.34)
4(a2 + r2 )3/2 8(a2 + r2 )2

We can examine this in the limiting cases where r  a and r  a. The magnetic induction
field components are

µ0 Ia2 cos θ 15a2 r2 sin2 θ


 
Br = 1+ + ···
2(a2 + r2 )3/2 4(a2 + r2 )2
µ0 Ia2 sin θ 15a2 r2 sin2 θ(4a2 − 3r2 )
 
2 2
Bθ = − 2 2a − r + + ...
4(a + r2 )5/2 8(a2 + r2 )2

For the case r  a (far from the loop) we have

µ0 cos θ
Br = (Iπa2 ) (5.35)
2π r3
µ0 sin θ
Bθ = 3
(Iπa2 ). (5.36)
4π r

38
This looks like a dipole. We introduce the magnetic dipole moment

m = πIa2 . (5.37)

This expression is independent of the shape.

5.4 Magnetic Fields of a Localized Current Distribu-

tion, Magnetic Moment


*
Suppose we have a localized current density J on a bulk of volume V . Suppose we enclose
the bulk with a volume V 0 , where V 0 > V . The computation for the vector potential

ˆ * *0
* * µ0 J (x )
A(x) = * *0
d3 x0
4π |x − x |

*
where d3 x0 refers to V 0 . Since J = 0 at V 0 , then the integration along V 0 does not matter.
* *0 * *0
We consider that the field point is far away |x|  |x |. Then |x − x | is

q
* *0 *2 *02 * *0
|x − x | = x + x − 2x · x
v
u *0 * *0
* t
u |x | 2x · x
= |x| 1 + * − *
|x|2 |x|2

*0 * *0
| x |2 |2 x · x | * *0
But *  * . At this limiting condition |x − x | approximates to
| x |2 | x |2

v
u * *0
* *0 * t
u 2x · x
|x − x | ≈ |x| 1 − *
|x|2

We can perform a binomial expansion

* *0 * *0
" # " #
* *0 * 1 2x · x * x·x
|x − x | ≈ |x| 1 + + ··· = |x| 1 + + ···
2 |*x|2
*
|x|2

39
1
Then, for * *0 we do a geometric series:
|x−x |

* *0
!
1 1 1 x·x
* *0
= *
h * *0
i= * 1+ * + ··· (5.38)
|x − x | |x| 1 + x·x
* + ··· |x| |x|2
| x |2

* *
Next, we plug this to A(x)

ˆ " * *0
#
* * µ0 1 x·x * *0
A(x) = * + * + · · · J (x ) d3 x0
4π |x| |x|3

* *
We consider the ith component of A(x):

" ˆ * ˆ #
* µ0 1 *0 x *0 *0
Ai (x) = * Ji (x ) d3 x0 + * · Ji (x )x d3 x0 + · · · (5.39)
4π x |x|3

* *
We wish to show that the first term is zero. We let G = f g J where f and g are continuous
*
functions. We assume that J is localized. We take the volume integral of the divergence of
*
G. We use the Green’s reciprocity theorem

ˆ * *
ˆ * *
∇ · G dV = ∇ · G dV 0
V V0

We use the divergence theorem

ˆ * *
˛ * *0
0
∇ · G dV = G · dS . (5.40)
V0 S0

But the current density at the surface of V 0 is zero. Hence, it is true that

˛ * *0
G · dS = 0. (5.41)
S0

´ * * ´ *0
Next, we need to show that V
∇ · G dV = V
Ji (x ) d3 x0 . We use the triple product rule of
the divergence

ˆ ˆ
 
* * * * *
0 * * *
*


∇ · G dV = ∇· J  d3 x0
f J · ∇g + g J · ∇f + f g 
V

* *
Recall that in magnetostatics ∇ · J = 0. If f = 1 and g = x0i , then

ˆ ˆ 0
* * * *0 * * 0
∇ · G dV = (1)J · ∇ x0i + gJ · ∇ f d3 x0
V
ˆ * *0
= (J · ∇ )g d3 x0

40
ˆ  
∂ ∂ ∂
= Jx 0 + Jy 0 + Jz 0 x0i d3 x0
∂x ∂y ∂z
ˆ * * ˆ
*0
∇ · G dV = Ji (x ) d3 x0
V

´ * *
But since V
∇ · G dV = 0, then

ˆ
*0 !
Ji (x ) d3 x0 = 0 (5.42)

So, the first term of Equation 5.39 is zero. Next, we examine the second term.

* ˆ
* µ0 x *0 *0
Ai (x) = · Ji (x )x d3 x0 (5.43)
4π |*
x|3

This is equivalent to

ˆ 3 ˆ
* *0 *0 3 0
X *0
x· Ji (x )x d x = xj Ji (x )x0j d3 x0 (5.44)
j=1

´ * *
Next, from V
∇ · G dV

ˆ * *
ˆ  * *0 * *0

∇ · G dV = f J · ∇ g + g J · ∇ f d 3 x0 = 0
V

We use f = x0i and g = x0j .

ˆ  * *0 * *0
 ˆ
x0i J x0j x0j J x0i 3 0
x0i Jj + x0j Ji d3 x0 = 0
 
= ·∇ + ·∇ dx =

So,

ˆ ˆ
x0i Jj 3 0
d x =− x0j Ji d3 x0 (5.45)

Then, from Equation 5.44

ˆ 3 ˆ 3 ˆ h
* *0 *0 3 0
X *0 1X *0 *0
i
x· Ji (x )x d x = xj Ji (x )x0j 3 0
dx = xj xj Ji (x ) + xj Ji (x ) d3 x0
0 0

j=1
2 j=1

We plug Equation 5.45 on the second term

3 ˆ h
1X *0 *0
i
= xj x0j Ji (x ) − x0i Jj (x ) d3 x0
2 j=1

This looks like a cross-product. It is equivalent to

ˆ  ˆ 
* *0 *0 1 * 3 0 *0 *
0
x· Ji (x )x d x = x × x × J dx (5.46)
2 i

41
We can define the magnetic moment density or magnetization

* * 1* * *
M(x) ≡ x × J (x) (5.47)
2

*
and its integral as the magnetic moment m

ˆ
* 1 *0 * *0
m= x × J (x ) dx0 (5.48)
2

Hence, the leading contributor to the vector potential is the magnetic dipole.

* *
* * µ0 m × x
A(x) = 3 (5.49)
4π |*
x |

We note that since the first term is zero, there is no magnetic monopole. The corresponding
magnetic field of a magnetic dipole is

* * *
!
* µ0 *
* * m×x µ0 * * x
B =∇×A= ∇× *3
= ∇ × m × *3 (5.50)
4π |x | 4π |x |

* * * * * * * * * * * * * * *
We use the identity ∇ × C × D = C(∇ · D) − D(∇ · C) + (D × ∇)C − (C × ∇)D.

* * * *
" ! ! #
* * µ0 * * x x * * x * * * * x
B(x) = m ∇· * − * (∇ · m) + * · ∇ m − (m · ∇) *
4π |x|3 |x|3 |x|3 |x|3

* *0
But since m is a function of x , then the second and third terms cancel out

* * * *
" ! !  #
µ0 * * x x * *  x  ** * * x
= m ∇· * − *  (∇
 · m) + * · ∇ m − (m · ∇) *
4π |x|3 3
|x| | x|3 |x|3

* *
" ! #
* * µ0 * * x * * x
B(x) = m ∇· * − (m · ∇) * (5.51)
4π |x|3 |x|3

We evaluate the dot products. For the first term we use the product rule

*
!
* x * * 1 * * 1
∇· * = (∇ · x) * +x· ∇ * (5.52)
|x|3 |x|3 |x|3

We evaluate the first term.

  
* * ∂ ∂ ∂ 
∇ · x = î + ĵ + k̂ · îx + ĵy + k̂z = 1 + 1 + 1 = 3
∂x ∂y ∂z

And for the second term.

 
* 1 ∂ ∂ ∂
∇ * = î + ĵ + k̂ (x2 + y 2 + z 2 )−3/2
|x|3 ∂x ∂y ∂z

42
3  
=− * îx + ĵy + k̂z
|x|5
* 1 3 *
∇ * =− * x
|x|3 |x|5

Hence, !
* * 1 * 3 * 3 * * 3 * 3
x· ∇ * = x·− * x=− * (x · x) = − * (|x|2 ) = − *
|x|3 |x|5 |x|5 |x|5 |x|3

Plugging these to Equation 5.52

*
* x 3 3
∇· * = * − * =0 (5.53)
|x|3 |x|3 |x|3

So, Equation 5.51 becomes.

*
" #
* * µ0 * x * µ0 1 * * * 1 * * *
B(x) = − (m · ∇) * = − * (m · ∇) x + x(m · ∇) *
4π |x|3 4π |x|3 |x|3

We need to simplify this further. For the first term,

 
* * * ∂ ∂ ∂
(m · ∇)x = (î mx + ĵ my + k̂ mz ) · î + ĵ + k̂ (î x + ĵ y + k̂ z)
∂x ∂y ∂z
∂ ∂ ∂
= (mx + my + mz )(î x + ĵ y + k̂ z)
∂x ∂y ∂xz
* * * *
(m · ∇)x = î mx + ĵ my + k̂ mz = m

and for the second term,

 
* * 1 ∂ ∂ ∂ 1
(m · ∇) * = (î mx + ĵ my + k̂ mz ) · î + ĵ + k̂
|x|3 ∂x ∂y ∂z (x + y + z 2 )3/2
2 2
!
3 xmx ymy zmz
=− * * + * + *
|x|4 |x| |x| |x|
*
3 * x
=− * m· *
|x|4 |x|
*
* * 1 3m · n̂
(m · ∇) * =− *
|x|3 |x|4

*
Plugging these to B

*
" !#
* * µ0 1 * * 3m · n̂
B(x) = − (m) + x − *
4π |*x|3 |x|4
*
" ! *#
µ0 1 * 3m · n̂ x
=− (m) + − *
4π |*x|3 |x|3
*
|x|
*
" ! #
µ0 1 * 3m · n̂
=− (m) + − * n̂
4π |*x|3 |x|3

43
Which leads us to the magnetic induction of a dipole,

* *
" #
* * µ0 3n̂(n̂ · m) − m
B(x) = * (5.54)
4π |x|3

5.5 Magnetic Dipole of a Current Loop

We compute the magnetic dipole of a loop of wire with a constant current density. The

magnetic dipole moment is

ˆ
* 1 *0 * *0
m= x × J (x ) d3 x0
2
ˆ
1 *0 * *0
= x × J (x ) dS 0 d`0
2
ˆ *0
1 *0 * *0
= x × J (x ) dS 0 d `
2
ˆ *0
ˆ
1 *0
= x × d` dI
2
ˆ *0
* I *0
m= x × d` (5.55)
2

This is the general expression for the magnetic moment of a loop of wire with constant
current density. Suppose we have a plane loop. The magnitude of the magnetic moment is

ˆ *0
* 1 *0
|m| = I | x × d ` | = IA (5.56)
2

where A is the total area enclosed by the loop. This holds regardless of the shape of the
n* o
loop. Suppose we have a collection of point charges {qi } at positions xi with velocities
n* o
v i . The total current density for this system is

* X * * *0
J= qi v i δ(x − xi )
i

The corresponding magnetic moment is

ˆ
* 1 *0 * *0
m= x × J (x ) d3 x0
2

44
ˆ !
1 *0 X * *0 *0
= x × qi v i δ(x − xi ) d3 x0
2
ˆ i
1X *0 * *0 *0
= qi x × v i δ(x − xi ) d3 x0
2 i
1X * *
= qi x i × v i
2 i
1X * * mi
= qi x i × v i ·
2 i
mi
1 X qi * *
= xi × p i
2 i mi

*
where p is called the linear momentum.

* 1 X qi *
m= Li
2 i mi

*
where L is called the angular momentum. If it happens that qi and mi are the same for all
particles (e.g. the particles are electrons), then

* 1 e X* 1 e *
m= Li = L (5.57)
2 me i 2 me

*
where L is the total angular momentum of the system. Suppose we have a bulk with a
* *
localized charge density J (x) and an arbitrary volume V .

Suppose that the current density is inside V . The volume integral of the magnetic
induction is
ˆ * 2µ0 *
B d3 x0 = m (5.58)
V 3

This requires us to add a correction term to the magnetic induction of a magnetic dipole:

* *
" #
* * µ0 3n̂(n̂ · m) − m 8π * *
B(x) = * + mδ(x) (5.59)
4π |x|3 3

The delta function only has a contribution when the source is at the origin.

45
5.6 Force, Torque and Energy of a Localized Current

Distribution in a External Magnetic Induction

According to Ampere’s law, if a localized distribution of current is placed in an external


*
magnetic field B, it experiences a force and a torque. We assume that the magnetic field
varies slowly over the region of the current density. Due to this we can perform a Taylor
*
series expansion at the origin. The kth component of B has the expansion:

* * *
Bk (x) = Bk (0) + x · ∇Bk (0) + · · · (5.60)

Then we can write the force as

*
ˆ * *0 * *0
F = J (x ) × B(x ) d3 x0
ˆ
*0 * 0 *
 
* *0 *
= J (x ) × B(0) + (x · ∇ )B(0) + · · · d3 x0
ˆ ˆ *
* *0 * *0 *0 * 0 *
= J (x ) × B(0) d x + J (x ) × (x · ∇ )B(0) d3 x0 + · · ·
3 0

ˆ ˆ *
* *0 * *0 *0 * 0 *
= J (x ) d x ×B(0) + J (x ) × (x · ∇ )B(0) d3 x0 + · · ·
3 0

| {z }
=0

We cut-off the higher-order terms.

*
ˆ * *0 **
*0
F = J (x ) × x d3 x0 ·∇B(0)
| {z }
*
m
* **
= −m · ∇B(0)

* * * ** * * * *
Using the product rule of gradients, ∇(A · B) = ()A∇)B + (B · ∇)A

* * * *0
*
 * *
= ∇·B )
∇(m · B) − m( 

* * * *
F = ∇(m · B) (5.61)

As for work,

ˆ * *
W = F · d`
ˆ * * *
*
= (∇(m · B)) · d `

46
Using the fundament theorem of calculus

* *
W = −m · B (5.62)

5.7 Magnetic Field of a Charged Rotating Sphere

Suppose we have a sphere of radius R with a total charge q uniformly distributed on the
*
surface and rotating at a constant angular velocity ω. We want to calculate the vector
*
potential at some point x, and the magnetic induction inside and outside the sphere. First,
*
we need to calculate J .

* * *
J = ρ(x) v

* * *
But v = ω × x.

* * *
= ρ(x)(ω × x)

* qδ(r−R)
Since the system has spherical symmetry, then ρ(x) = 4πR2
.

* qδ(r − R) * *
J= (ω × x) (5.63)
4πR2

*
We plug this to A

ˆ * *
* * µ0 J (x)
A(x) = * *0
d 3 x0
4π |x − x |
"ˆ *0
#
µ0 q * δ(r0 − R)x 3 0
= ω× * *0
dx
(4πR)2 |x − x |

We let the integral be

ˆ *0
* δ(r0 − R)x
G= * *0
d 3 x0
|x − x |

*
We can place the z−axis in the direction of x. We note that

*0
x = r0 cos θ0 sin φ0 î + r0 sin θ0 sin φ0 ĵ + r0 cos θ0 k̂

47
*
We plug this to G,

ˆ
* δ(r0 − R)
G= (r0 cos θ0 sin φ0 î + r0 sin θ0 sin φ0 ĵ + r0 cos θ0 k̂) d3 x0
(r2 + r02 − 2rr0 cos θ0 )1/2

But the azimuthal angle integrals in the x− and y−components vanish because of spherical
symmetry. We are left with

ˆ π ˆ ∞
*
0 δ(r0 − R)
G = 2π dθ (r0 cos θ0 sin φ0 î + r0 sin θ0 sin φ0 ĵ + r0 cos θ0 k̂)r02 sin θ0 dr0
(r2 + r02 − 2rr0 cos θ0 )1/2
ˆ0 π ˆ0 ∞
δ(r0 − R)r03 cos θ0 sin θ0
= 2π dθ0 k̂ dr0
0 0 (r2 + r02 − 2rr0 cos θ0 )1/2

But k̂ = r̂.

ˆ π
R2 cos θ0 sin θ0
= 2π dθ0 r̂
0 (r2 + r02 − 2rR cos θ0 )1/2

We can make a change of variable ξ = cos θ0 . Then,

ˆ π
ξ
= 2πR 2
dξ 0 r̂
0 + (r2 r02 − 2rRξ)1/2
2πR
*
(r2 + R2 − rR)(r + R) − (r2 + R2 + rR)|r − R| r̂
 
G=
r

The value of |r − R| depends on if r > R or r < R. Inside the sphere we find that

* 2πR  2
(r + R2 − rR)(r + R) − (r2 + R2 + rR)(R − r) r̂

G(r < R) =
r
4πRr
= r̂
3
*
4πRr x
=
3 |* x|
* 4πR *
G(r < R) = x (5.64)
3

Outside the sphere, |r − R| = r − R

* 2πR  2
(r + R2 − rR)(r + R) − (r2 + R2 + rR)|r − R| r̂

G(r > R) =
r
4πR4
= r̂
3r2
*
4πR4 x
=
3r2 |*x|
* 4πR4 *
G(r > R) = x (5.65)
3r3

48
Hence, the vector potential is

µ0 q * *

*


12πR
ω ×x inside
A= (5.66)
* *
 µ0 qR2 ω× x


12π * outside
| x |3

* * * *
Then we compute for the magnetic induction B using B = ∇ × A for the region inside the
* * * * * * * * * * * * * * *
sphere. We shall use the identity ∇×(C × D) = C(∇· D)− D(∇· C)+(D· ∇)C −(C · ∇)D.
* * * *
We use C = ω and D = x

* µ0 q h * * * * * * * * * * * *
i
B(r < R) = ω(∇ · x) − x(∇ · ω) + (x · ∇)ω − (ω · ∇)x
12πR

*
But since ω = const, then the second and third terms are zero.

* µ0 q h * * * * * *
i
B(r < R) = ω(∇ · x) − (ω · ∇)x
12πR

This simplifies to

µ0 q h * * i
= 3ω − ω
12πR
* µ0 q *
B(r < R) = ω (5.67)
6πR

Next, we look for the magnetic induction outside the sphere,

* *
* µ0 qR2 * ω × x
B(r > R) = ∇× *
12π |x|3
* *
" ! #
µ0 qR2 * * x * * x
= ω ∇· * − (ω · ∇) *
12π |x|3 |x|3

*
x
We let r̂ = * .
|x|

 
µ0 qR2 * µ0 qR2 *
12π
ω · r̂ r̂ − 12π
ω
= *
|x|3

Thus,
 
qR2 * qR2 *
* µ0 3π
ω · r̂ r̂ − 3π
ω
B(r > R) = * (5.68)
4π |x|3

By inspection, we can infer that the magnetic dipole moment is

*
* qR2 ω
m= . (5.69)

49
*
This means that the sphere looks like a magnetic dipole. The magnetic induction B is due
*
to a magnetic dipole with magnetic moment m.

5.8 Macroscopic Equations, Boundary Conditions on

the Magnetic Induction and Magnetic Field


*
The previous assumption that we had is that J is completely known as a function of position.
*
But in macroscopic cases involving bulk materials J is usually not known. Electrons in
the material may constitute time-varying electric currents. For macroscopic materials we
*
consider the average magnetic values. We call the bulk material’s magnetic induction is B.
Suppose that the microscopic magnetic induction satisfies

* *
∇ · B micro = 0 (5.70)

Then it implies that


* * * * *
∇·B =0 → B =∇×A (5.71)

*
The large number of molecules per unit volume, each with its magnetic moment mi , gives
rise to an average macroscopic magnetization or magnetic moment density given by

* * X *
M (x) = Ni hmi i (5.72)
i

*
where Ni is the average number of molecules per unit volume and hmi i is the average
*
magnetic moment in a small volume at x. The total contribution of these molecules to the

ˆ * *0 ˆ * *0 * *0
* * µ0 J (x ) 3 0 µ0 M (x ) × (x − x )
A(x) = * *0
dx + * *0
d3 x0 (5.73)
4π |x − x | 4π |x − x |
| {z } | {z }
contribution of flow of free charge contribution by the magnetic moments per molecule

*
where J is due to free charges (i.e. dislodged electrons). We can write the second term as

ˆ * * * *0 ˆ *0
!
M (x) × (x − x ) * *0 1
* *0
d3 x0 = M (x ) × ∇ * *0
d 3 x0 (5.74)
|x − x |3 |x − x |

We can use

* *0 ! !
* M (x ) 1 * * *0 * 1 * *0
∇× * *0
= * *0
∇ × M (x ) + ∇ * *0
× M (x ) (5.75)
|x − x | |x − x | |x − x |

50
*
Then we plug this to A

ˆ *0
! ˆ " * *
* *0 !#
* *0 1 3 0 1 0
* *0
0 M (x )
M (x ) × ∇ * *0
dx = * *0
∇ × M (x ) − ∇ × * *0
d3 x0
|x − x | |x − x | |x − x |
(5.76)
We convert the second term to a surface integral (follows from the Gauss’s theorem)

ˆ *
* *0 ! ˆ * *0
0 M (x ) 3 0 M (x ) *
∇ × * *0
dx = * *0
· dS (5.77)
|x − x | |x − x |

*
We assume that M is localized.

Then the surface integral goes to zero. So, we are left with

ˆ *0
! ˆ *
* *0 1 3 0 1 * *0
M (x ) × ∇ * *0
dx = * *0
∇0 × M (x ) d3 x0 (5.78)
|x − x | |x − x |

Hence, the vector potential is

ˆ * *0 *0 * *0
* * µ0 J (x ) + ∇ × M (x )
A(x) = * *0
d3 x0 (5.79)
4π |x − x |

*0 * *0 *
We notice that ∇ × M (x ) has the same units as J . The magnetization induces an effective
* *0 * *0
current density J M ≡ ∇ × M (x ). So the total current density comes from the free charges
and the magnetization. Essentially, the curl of the microscopic magnetic induction comes
from
* * * * * * *
∇ × B micro = µ0 J micro → ∇ × B = µ0 (J + J M ) (5.80)

* *
This is the macroscopic expression of ∇ × B

* * * * *
∇ × B − µ0 (∇ × M ) = µ0 J
* * *
µ0 (∇ × (B/µ0 − M )) =

51
We introduce the quantity

*
* B *
H≡ −M
µ0

We can then write

* * *
∇×H =J (5.81)
* *
∇·B =0 (5.82)

*
where J is the current density arising from free charges in the bulk material. the relationship
* *
between H and B depends on the material. For isotropic diamagnetic and paramagnetic
substances:
* *
B = µH (5.83)

where µ, the paramagnetic permeability, is a characteristic property of the medium. µ > µ0 is


a paramagnetic substance, while µ < µ0 is a diamagnetic substance. Diamagnetic materials
are repelled by magnetic fields. An applied magnetic field creates an induced magnetic field in
the diamagnetic material in the opposite direction causing repulsiveness. Paramagnetic and
ferromagnetic materials are attracted by external magnetic fields. Ferromagnetic substances
have the property
* * *
B = F (H) (5.84)

*
where F is some vector function.

5.8.1 Hysteresis

* * *
Increasing H also increases B. Slowly decreasing H back to the initial value traces a
weird curve. It does not retrace the original path. This phenomenon is called hysteresis.
* * * *
This implies that B is not a single-valued function of H. The function F (H) depends on

52
the history of the preparation of the material.

* * * *
We can then infer the boundary conditions for B and H. We convert ∇ · B = 0 →
¸ * *
B ·d a = 0 using the divergence theorem. Then we place a Gaussian pillbox at the interface

˛ * *
B · da = 0
ˆ *
ˆ *
ˆ *
* * *
B top · d a + B bottom · d a + B sides · d a = 0

We shrink the sides  → 0.

(B2⊥ − B1⊥ )A = 0

We can rewrite this as


* *
(B 2 − B 1 ) · n̂ = 0 (5.85)

* * * ¸ * *
Next, we apply Ampere’s law, ∇ × H = J → H · d ` = I, to a rectangular loop at the
interface

˛ * *
H · d` = I

k k
(H2 − H1 )` = K`
k k
H2 − H1 = µ0 K

We can rewrite this as


* * *
n̂ × (H 2 − H 1 ) = K (5.86)

If there are no free charges and currents, then we can write 5.86 as

* * * µ2 *
B 2 · n̂ = B 1 · n̂, n̂ × H 2 = n̂ × H1 (5.87)
µ1

or
* 1 * * *
H 2 · n̂ = H 1 · n̂, n̂ × H 2 − = n̂ × H 1 (5.88)
2
* *
These boundary conditions completely determine B and H in a system with a boundary.

53
5.9 Methods of Solving Boundary-Value Problems in

Magnetostatics.

The basic equations are


* * * * *
∇ · B = 0, ∇×H =J (5.89)

A. Generally applicable method of vector potential


* *
Since ∇ · B = 0 then we can write the magnetic induction as

* * *
B =∇×A (5.90)

* * * *
We want to write everything in terms of A. Suppose that H = H(B) where µ is the
permeability of the bulk material.. Then

* * * * *
∇ × H(∇ × A) = J

* *
But for linear media B = µH where µ is the permeability of the bulk material. We assume
that µ is isotropic in the bulk material.

* 1 *
∇× (B) =
µ
* 1 * * *
∇× (∇ × A) = J
µ

* * * * * * *
We make use of the identity ∇ × ∇ × A = ∇(∇ · A) − ∇2 A. We assume the Coulomb
* *
gauge ∇ · A = 0 which leads us to

* *
∇2 A = −µJ (5.91)

We can use the method of Green’s functions to obtain the general solution:

ˆ * *0
* * µ J (x )
A(x) = * *0
d3 x0 (5.92)
4π |x − x |

*
B. J = 0; Magnetic Scalar Potential
*
If J = 0, then we can write the magnetic field as

* *
H = −∇ΦM (5.93)

54
where ΦM is called the magnetic scalar potential. For a linear medium with isotropic per-
* *
meability ∇ · B = 0 becomes

* *
∇ · (µH) = 0
* *
−∇ · (µ∇ΦM ) = 0

∇ 2 ΦM = 0 (5.94)

which is Laplace’s equation.

C. Hard Ferromagnets

A hard ferromagnet is called “hard” if it has a magnetization that is independent of applied


fields for moderate field strengths. Such materials can be treated as if they had a fixed,
* *
specified magnetization M (x).

a) Scalar Potential
*
Since J = 0, then we can use the magnetic scalar potential. Recall that in the presence of
*
a magnetization M
* * *
B = µ0 (M + H) (5.95)

We take the divergence

* *0
*
 * * *
∇·B
  = µ0 ∇ · (M + H)
* * * *
∇ · H = −∇ · M

* *
But since H = −∇ΦM ,
* *
∇ 2 ΦM = ∇ · M (5.96)

* *
This is the same as the Poisson equation in electrostatics. We can let ρM = −∇ · M .

∇2 ΦM = −ρM (5.97)

We can use the method of Green’s functions to obtain the general solution:

ˆ *0 * *0
* 1 ∇ · M (x )
ΦM (x) = − * *0
d3 x0 (5.98)
4π |x − x |

55
This is analogous to the electric potential in the electrostatics case. Suppose that the mag-
netic “charge” distribution is localized, then the integrand can be manipulated from:

* *0 ! *0 * *0 ! *0 * *0 !
*0 M (x ) ∇ · M (x ) * *0 * 0 1 ∇ · M (x ) * *0 * 1
∇ · * *0
= * *0
+ M ( x )∇ · * *0
= * *0
− M (x )∇ · * *0
|x − x | |x − x | |x − x | |x − x | |x − x |

*0 *
1
where we used ∇ * *0 = −∇ * 1*0 . So, the term in the integrand is
|x−x | |x−x |

*0 * *0 * *0 ! !
∇ · M (x ) *0 M (x ) * *0 * 1
* *0
=∇ · * *0
+ M (x )∇ · * *0
(5.99)
|x − x | |x − x | |x − x |

We integrate both sides

ˆ *0 * *0 ˆ *0
* *0 ! ˆ !
∇ · M (x ) M (x ) * *0 * 1 *
* *0
d 3 x0 = ∇ · * *0
d3 x0 + M (x )∇ · * *0
d3 x0 = 4πΦM (x)
|x − x | |x − x | |x − x |

Since the body is localized then the integral in the second term would vanish if we consider
all space.

ˆ * *0
* 1 * M (x )
ΦM (x) = − ∇ · * *0
d 3 x0 (5.100)
4π |x − x |

*0 *
1
where we used ∇ * *0 = −∇ * 1*0 again. If we observe from far away
|x−x | |x−x |

 ˆ *
1 *
* 1 * *0
3 0 m · r̂
ΦM (x) = − ∇ · M (x ) d x = (5.101)
4π r 4πr2
| {z }
*
m

*
where m is the total magnetic moment. So

* *
*m·x
ΦM (x) = (5.102)
4πr3

* * * *
What if M is discontinuous (outside the object has M = 0)? So outside the surface ∇ · M
*
becomes zero. If a ferromagnet has volume V and surface S, we specify M inside V and
assume that it falls suddenly to zero at the surface. There has to be an induced magnetization
on the surface. We consider a small volume (pillbox) at the surface. We use the divergence
theorem on the pillbox
ˆ * *
" *
∇ · F dV = F̂ · n̂ dS (5.103)
V

So,

ˆ ˆ * *
3 0
ρM d x = − ∇ · M d 3 x0

56
ˆ *
QM = − M · n̂ dS 0
S
ˆ *
ˆ * :0

0 
QM = − · n̂ dS 0
M · n̂ dS −M 


So for any arbitrary surface area


*
σM = M · n̂ (5.104)

There is an in effective magnetic surface-charge density where n̂ is outward directed normal.


Therefore, ΦM must have a correction term that contains σM

ˆ *0 * *0 ˛ *0 * *0
* 1 ∇ · M (x ) 13 0 ∇ · M (x )
ΦM (x) = − * *0
dx + * *0
da0 (5.105)
4π |x − x | 4π S |x − x |

*
This only applies when M is localized.

5.10 Magnetic Shielding

5.11 Faraday’s Law of Induction

57
Chapter 6

Maxwell Equations, Macroscopic


Electromagnetism, Conservation Laws

6.1 Green Functions for the Wave Equation

So far we have encountered the inhomogeneous wave equations involving the scalar and
vector potentials:

1 ∂ 2Φ ρ
∇2 Φ − 2 2
=− (6.1)
c ∂t 0
*
*1 ∂ 2A *
∇2 A − 2 2 = −µ0 J (6.2)
c ∂t

where we have assumed the Lorenz gauge. They have the general form:

1 ∂ 2Ψ *
∇2 Ψ − 2 2
= −4πf (x, t) (6.3)
c ∂t

*
where f (x, t) is a source term. This may either act as a charge or current distribution. We
* *
take the Fourier transform of ψ(x, t) and f (x, t) in the frequency domain:

ˆ ∞
* *
Ψ(x, ω) = Ψ(x, t)eiωt dt
ˆ−∞

* *
f (x, ω) = f (x, t)eiωt dt
−∞

Their corresponding inverse Fourier transform is


ˆ ∞
* *
Ψ(x, t) = Ψ(x, ω)e−iωt dt
ˆ−∞
∞ (6.4)
* * −iωt
f (x, t) = f (x, ω)e dt
−∞

58
We plug this to Equation 6.3

ˆ ∞ ˆ ∞ ˆ ∞
1 ∂2
  
2 * −iωt * −iωt * −iωt
∇ Ψ(x, ω)e dt − 2 2 Ψ(x, ω)e dt = −4π f (x, ω)e dt
−∞ c ∂t −∞ −∞
ˆ ∞ ˆ ∞
1 ∂2

* −iωt *
2
∇ − 2 2 Ψ(x, ω)e dt = f (x, ω)e−iωt dt
−∞ c ∂t −∞
ˆ ∞ 2
 ˆ ∞
ω * *
∇2 + 2 Ψ(x, ω)e−iωt dt = f (x, ω)e−iωt dt
−∞ c −∞

This is true for any arbitrary time t. So we obtain

* *
(∇2 + k 2 )Ψ(x, ω) = −4πf (x, ω) (6.5)

* *0
where k = ωc . The differential operator has a corresponding Green’s function G(x, x ). The
Green’s function satisfies

* *0 * *0
(∇2 + k 2 )G(x − x ) = −4πδ(x − x ) (6.6)

Assuming that there are no boundary surfaces, the Green’s function is spherically symmetric
* * *0
because the dependence is only along R = x − x . Then the Laplacian in the differential
*
operator is only the derivative with respect to R = |R|.

1 d2 * *0 2 * *0 * *0
(RG( x − x )) + k G( x − x ) = −4πδ( x − x) (6.7)
R dR2

* * *0
Suppose that R 6= 0, means that x − x 6= 0 (the field point is not at the source point), then
the delta function becomes 0.

1 d2 * *0 * *0
2
(RG(x − x )) + k 2 G(x − x ) = 0 (6.8)
R dR

This second-order differential equation has the general solution:

e−ikR eikR
G(R) = A +B (6.9)
R R

We want to know what A and B are. But if k → 0 we just obtain the Poisson equation.

1 1
lim G(R) = (A + B) = (6.10)
k→0 R R

where we required A + B = 1. We define

e±ikR
G± (R) = (6.11)
R

59
so that we can rewrite the general solution as

G(R) = AG+ (R) + BG− (R) (6.12)

We can interpret G+ (R) as an outgoing spherical wave with amplitude A, while G− (R)
represents an incoming spherical wave with amplitude B.
In order to understand the time dependence on each term, we need to introduce time-
dependent Green’s functions that must satisfy

1 ∂2
 
* *0
∇ − 2 2 G± (R, t, t0 ) = −4πδ(x − x )δ(t − t0 ).
2
c ∂t

We apply the Fourier transform

ˆ ∞ ˆ ∞
1 ∂2
 
± 0 iωt * *0
2
∇ − 2 2 G (R, t, t )e dt = −4πδ(x − x ) δ(t − t0 )eiωt dt
c ∂t
−∞
ˆ ∞ −∞

∇ − k 2 G± (R, t, t0 )eiωt dt =
 2 
−∞

Then we inverse Fourier transform:

0 * *0
∇ + k 2 G± (R, ω)eiωt = −4πeiωt δ(x − x )
 2 
(6.13)

But the solutions to this differential equation are

0
G± (R, ω) = G± (R)eiωt

Then the time-dependent Green’s functions are

ˆ ∞
(±) 1 e±ikR −iωτ
G (R, τ ) = ·e dω (6.14)
2π −∞ R

where τ = t − t0 is the relative time appearing in Equation 6.13. The infinite-space Green’s
function is thus a function of only the relative distance R and the relative time τ between
the source and observation point. In a nondispersive medium where k = ω/c, the integral in
Equation 6.14 is a delta function. The Green’s functions are

 
(±) 1 R
G (R, τ ) = δ τ ∓ (6.15)
R c

* *0
In terms of x, x , t, t0 :

* *0
" #!
(±) * *0 0 1 |x − x |
0
G (x, x , t, t ) = * *0 δ t − t ∓ (6.16)
|x − x | c

60
This is called the retarded Green’s function because it exhibits the casual behavior associated
with a wave disturbance.

6.2 Poynting’s Theorem and Conservation of Energy

and Momentum for a System of Charged Particles

and Electromagnetic Fields


* *
Suppose we have an external electromagnetic field E and B. The total rate of doing work
* * *
done by the external electromagnetic field on a charged point particle is q v · E, where v is
the velocity of the particle. We note that there is no mention of the magnetic field doing
work because we know for a fact that magnetic fields cannot do work since the velocity and
the magnetic force are always perpendicular to each other.
Suppose we have an arbitrary localized charge and current distribution, then the total
rate of doing work by the external field in a volume V is

ˆ * *
J · E d3 x. (6.17)
V

This equation is interpreted as power. This might represent as a conversion of electromagnetic


energy into mechanical or thermal energy. Using the macroscopic form of Ampere’s law,
* * * *
∂D
∇×H =J + ∂t
,

ˆ ˆ "* *#
* * * ∂ D *
J · E d3 x = ∇×H − · E d3 x
V V ∂t
ˆ "* * * * ∂D
*#

= E · (∇ × H) − E · d3 x (6.18)
V ∂t

* * *
We make use of the cross product rule for the quantity ∇ · (E × H):

* * * * * * * * *
∇ · (E × H) = (∇ × E) · H − (∇ × H) · E

* * *
Then the quantity E · (∇ × H) in Equation 6.18 is just

* * * * * * * * *
(∇ × H) · E = (∇ × E) · H − ∇ · (E × H).

* * *
But according to Faraday’s law in Maxwell’s equations ∇ × E = − ∂∂tB :

*
* * ∂B ** * * *
(∇ × H) · E = − · H − ∇ · (E × H) (6.19)
∂t

61
We plug this to Equation 6.18

ˆ "
* ! *#
∂B * * * * * ∂D
= − · H − ∇ · (E × H) − E · d3 x
V ∂t ∂t
ˆ * * ˆ " * * *#
∂ B * * * * ∂D
J · E d3 x = − · H + ∇ · (E × H) + E · d3 x. (6.20)
V V ∂t ∂t

First we assume that the material is a linear medium with susceptibility  and permeability µ,
* * * *
meaning that E = 1 D and B = µH. There will be no dispersion because the susceptibility
and permeability are isotropic and frequency independent. We define a new quantity

1 * * * * 
u≡ E·D+B·H (6.21)
2
* * * *
∂B ∂D ∂u
We can express ∂t
·H +E· ∂t
as ∂t
because

∂u 1 ∂ * * * * 
= E·D+B·H
∂t 2 ∂t
 
1 ∂ * * ∂ * *
= (E · D) + (B · H)
2 ∂t ∂t
" * * * * #
1 * ∂E * ∂D * ∂B * ∂H
= D· +E· +H · +B·
2 ∂t ∂t ∂t ∂t
" * * * *#
1 1 * ∂E 1 * ∂E * ∂B * ∂B
= E· + E· + µB · + µB ·
2  ∂t  ∂t ∂t ∂t
" * *#
1 2 * ∂E * ∂B
= E· + 2µB ·
2  ∂t ∂t
* *
∂u * ∂ D * ∂B
=E· +H · X
∂t ∂t ∂t

So
ˆ ˆ  
* *
3 ∂u * * *
− J ·E d x= + ∇ · (E × H) d3 x (6.22)
V V ∂t

This is true for an arbitrary differential volume in V . So, we must have

* * ∂u * * *
−J · E = + ∇ · (E × H) (6.23)
∂t

* * *
We define the quantity S ≡ E × H, which is the Poynting vector.

* * ∂u * *
−J · E = +∇·S (6.24)
∂t

The Poynting vector has dimensions of energy/area × time, which is an energy flux vector
for the electromagnetic field. We can interpret Equation 6.22 as:
The time rate of change of electromagnetic energy within a certain volume, plus the energy

62
flowing out through the boundary surfaces of the volume per unit time, is equal to the negative
of the total work done by the fields on the sources within the volume.
This is Poynting’s theorem. This can be viewed as a conservation of energy. This may
not apply to all materials because we assumed in our computations that the material we are
considering is linear, and we ignored dispersion or losses.
* *
So far we have examined J · E. We denote the total energy of the particles within a
volume V as Emech and assume that no particles can move out of the volume. So the power
exerted by the charged particles is

ˆ
dEmech * *
= J · E d3 x
dt V

Suppose we turn on an external electromagnetic field. Then the total power of the system is

˛
dE d * *
= (Emech + Efield ) = − n · S da (6.25)
dt dt S

We can find the total energy by the electromagnetic field by integrating Equation 6.21. We
consider that we are in vacuum so that u becomes

  
1 *  * * 1 *
u= E · 0 E + B · B
2 µ0
 
0 *
2 1 *2
= |E| + |B|
2 0 µ0
0 * * 
u= |E|2 + c2 |B|2 (6.26)
2

Then we compute for the total energy due to the electromagnetic field.

ˆ ˆ *
3 0 * 
Efield = ud x = |E|2 + c2 |B|2 d3 x (6.27)
V 2 V

Next, we should also consider conservation of linear momentum. We recall the Lorentz force
law:
* * * *
F = q(E + v × B) (6.28)

*
where F is the force acting on the charged particle due to the electromagnetic field.

* * * *
F = qE + q v × B

´ * * ´ *
We can express q as q = V
ρ d3 x, and q v as q v = V
J d3 x:

ˆ *
ˆ * *
3
= ρE d x + J × B d3 x
V V

63
* *
But F is just the time derivative of the linear momentum of the system P mech .

* ˆ 
dP mech * * *
= ρE + J × B d3 x. (6.29)
dt V

* * *
We want to re-express ρE + J × B. We recall Gauss’s law and Ampere’s law with Maxwell’s
correction:

*!
*ρ * * * * ∂E
∇·E = , ∇ × B = µ0 J + 0
0 ∂t

So we can write the charge and current densities as

*
* * * 1 * * ∂E
ρ = 0 ∇ · E, J = ∇ × B − 0 . (6.30)
µ0 ∂t

So,

*!
* * *  * * * 1 * * ∂E *
ρE + J × B = 0 ∇ · E E + ∇ × B − 0 ×B
µ0 ∂t
" * #
*
 * * 1 * * * ∂E *
= 0 E ∇·E + (∇ × B) × B − ×B
0 µ0 ∂t

We switch the cross-products:

" * #
* * *  * *  * ∂E
* * * *
ρE + J × B = 0 E ∇·E +B× − c2 B × (∇ × B) (6.31)
∂t

* *
We take the partial time derivative of E × B

* *
∂ * * ∂E * * ∂B
(E × B) = ×B+E×
∂t ∂t ∂t
* *
∂E * ∂B
*
= −B × +E×
∂t ∂t

* *

so that B × ∂t
(E) is
* *
* ∂E ∂ * * * ∂B
B× = − (E × B) + E × . (6.32)
∂t ∂t ∂t

We plug this back:

" * ! #
* * * * * * ∂ * * * ∂B * * *
ρE + J × B = 0 E ∇·E + − (E × B) + E × − c2 B × (∇ × B)
∂t ∂t

64
* * *
We add a zero, c2 B(∇ · B) = 0 (due to the No name law), inside the parenthesis.

" * #
* * * * * * ∂ * * * ∂ B * * *
= 0 E ∇ · E + c2 B(∇ · B) − (E × B) + E × − c2 B × (∇ × B)
∂t ∂t

* * *
But ∇ × E = − ∂∂tB

* * * h*  * * * * * * * * * * * i ∂ * *
ρE + J × B = 0 E ∇ · E + c2 B(∇ · B) − E × (∇ × E) − c2 B × (∇ × B) − 0 (E × B)
∂t
(6.33)

Then we plug this back to Equation 6.29.

* ˆ  h*  * * 
dP mech * * * * * * * * * i ∂ * *
= 0 E ∇ · E + c B(∇ · B) − E × (∇ × E) − c B × (∇ × B) − 0 (E × B) d3 x
2 2
dt ∂t
V ˆ 
d * * *
P mech + 0 (E × B) d3 x =
dt
ˆ hV *  * *  * * * * * * * * * i
0 E ∇ · E + c B(∇ · B) − E × (∇ × E) − c B × (∇ × B) d3 x
2 2
V

(6.34)

But the second term in the LHS is

ˆ * *
ˆ * * *
3
0 E × B d x = µ0 0 E × H d3 x ≡ P field , (6.35)
V V

* *
where P field is the total linear momentum of the electromagnetic field. We can treat µ0 0 E ×
*
H as a momentum density. Thus, Equation 6.34 is a conservation law for momentum.
We introduce the Maxwell stress tensor Tαβ :

 
2 1 * * 2
* *
Tαβ = 0 Eα Eβ + c Bα Bβ − (E · E + c B · B)δαβ (6.36)
2

This allows us to express Equation 6.34 more compactly as

* * ˆ X
dP mech dP field ∂
+ = Tαβ d3 x.
dt dt V β ∂xβ

Then we use the divergence theorem to recast the RHS in terms of a surface integral

˛ X
d * *  ∂
P mech + P field = Tαβ nβ da (6.37)
dt S β ∂xβ

*
where n is the outward normal to the closed surface S. Again, this expression is a statement
of conservation of linear momentum. The integrand β ∂x∂ β Tαβ nβ represents the α-th com-
P

ponent of the flow per unit area of momentum across the surface S into the volume V . It is

65
the force per unit area transmitted across the surface S and acting on the combined system
of particles and fields inside V .

6.3 Poynting’s Theorem in Linear Dispersive Media

with Losses

In the prior section we restricted ourselves to studying Poynting’s theorem in linear media
with no dispersion or losses. Actual materials can have dispersion and losses. We need
to Fourier decompose the electric field (and the electric displacement), and the magnetic
induction (and the magnetic field as well).

* *
ˆ ∞ *
*
E(x, t) = E(x, ω)e−iωt dω
* *
ˆ−∞
∞ * *
D(x, t) = D(x, ω)e−iωt dω
−∞

* *
ˆ ∞ *
*
B(x, t) = B(x, ω)e−iωt dω
* *
ˆ−∞
∞ * *
H(x, t) = H(x, ω)e−iωt dω
−∞

* * * * * *
where we assumed that the material is linear D(x, ω) = (ω)E(x, ω) and B(x, ω) =
* *
µ(ω)H(x, ω). We note that the permeability and the susceptibility are now frequency-
dependent. The reality constraints are:

*∗ * * * *∗ * * *
E (x, ω) = E(x, −ω), D (x, ω) = D(x, −ω), ∗ (ω) = (−ω)

* *
∂D
Then we plug the Fourier integrals to Equation 6.20. We consider the E · ∂t
term first.

* ˆ ∞ * ˆ ∞ ∗ 
* ∂D * 0 ∂
−iω 0 t
* *
0−iωt
E· = E(x, ω )e dω · D(x, ω)e dω
∂t −∞ ∂t −∞
ˆ ∞ ∗  ˆ ∞ 
* * * *
0 iω 0 t 0 −iωt
= E (x, ω )e dω · (ω)E(x, ω)(−iω)e dω
−∞ −∞
* ˆ ˆ *∗
* ∂D * 0
E· = dω dω 0 E (ω 0 )[−iω(ω)] · E(ω)e−i(ω−ω )t (6.38)
∂t

Next we split the integral into two equal parts

* ˆ ˆ *∗
* ∂D 1 * 0
E· = dω dω 0 E (ω 0 )[−iω(ω)] · E(ω)e−i(ω−ω )t +
∂t 2
ˆ ˆ *∗
1 * 0
dω dω 0 E (ω 0 )[−iω(ω)] · E(ω)e−i(ω−ω )t
2

66
For the second integral, we make a change of variables ω → −ω 0 and ω 0 → −ω:

* ˆ ˆ *∗
* ∂D 1 * 0
E· = dω dω 0 E (ω 0 )[−iω(ω)] · E(ω)e−i(ω−ω )t +
∂t 2
ˆ ˆ *∗
1 0
* 0
dω dω E (−ω)[−i(−ω 0 )(−ω 0 )] · E(−ω 0 )e−i(−ω −(−ω))t
2
ˆ ˆ *∗
1 * 0
= dω dω 0 E (ω 0 )[−iω(ω)] · E(ω)e−i(ω−ω )t +
2
ˆ ˆ *∗
1 0
* 0
dω dω E (−ω)[iω 0 (−ω 0 )] · E(−ω 0 )e−i(ω−ω )t
2

We take the complex conjugate of the reality conditions and we plug them to the equation
above:

* ˆ ˆ *∗
* ∂D 1 * 0
E· = dω dω 0 E (ω 0 )[−iω(ω)] · E(ω)e−i(ω−ω )t +
∂t 2
ˆ ˆ *∗
1 0
* 0
dω dω E(ω)[iω 0 ∗ (ω 0 )] · E (ω 0 )e−i(ω−ω )t
2
* ∂D
* ˆ ˆ *∗
1 * 0
E· = dω dω 0 E (ω 0 )[−iω(ω) + iω 0 ∗ (ω 0 )] · E(ω)e−i(ω−ω )t (6.39)
∂t 2

Next we expand the term inside the square brackets at ω = ω 0 :

d
[−iω(ω) + iω 0 ∗ (ω 0 )] = 2ω Im(ω) − i(ω − ω 0 ) (ω∗ (ω)) + · · · (6.40)

* *
∂D
We plug this to E · ∂t
:

* ˆ ˆ *∗
* ∂D 1 d * 0
E· = dω dω 0 E (ω 0 )[2ω Im(ω) − i(ω − ω 0 ) (ω∗ (ω)) + · · · ] · E(ω)e−i(ω−ω )t
∂t 2 dω
ˆ ˆ *∗
1 * 0
= dω dω 0 E (ω 0 )[2ω Im(ω)] · E(ω)e−i(ω−ω )t −
2
ˆ ˆ *∗
1 d * 0
dω dω 0 E (ω 0 )[i(ω − ω 0 ) (ω∗ (ω))] · E(ω)e−i(ω−ω )t + · · ·
2 dω

∂ −i(ω−ω 0 )t 0
But ∂t
e = −i(ω − ω 0 )e−i(ω−ω )t . So

* ˆ ˆ *∗
* ∂D 1 * 0
E· = dω dω 0 E (ω 0 )[2ω Im(ω)] · E(ω)e−i(ω−ω )t +
∂t 2
ˆ ˆ *∗
1∂ d * 0
dω dω 0 E (ω 0 ) (ω∗ (ω)) · E(ω)e−i(ω−ω )t + · · · (6.41)
2 ∂t dω

* * * *
∂B
For H · ∂t
we just replace E → H and  → µ:

* ˆ ˆ *∗
* ∂B 1 * 0
H· = dω dω 0 H (ω 0 )[2ω Imµ(ω)] · H(ω)e−i(ω−ω )t +
∂t 2
ˆ ˆ *∗
1∂ d * 0
dω dω 0 H (ω 0 ) (ωµ∗ (ω)) · H(ω)e−i(ω−ω )t + · · · (6.42)
2 ∂t dω

67
If the permeability and susceptibility are real and frequency-independent, we would retrieve
Equation 6.20. Suppose we are considering a narrow range of frequencies and that

* * * *
E = Ẽ(t) cos(ω0 t + α), H = H̃(t) cos(ω0 t + β) (6.43)

* *
1
where Ẽ(t) and H̃(t) are slowly varying relative to ω0
and the inverse of the frequency range
over which (ω) changes appreciably. We add up Equations 6.41 and 6.42 (disregard the
2nd-order term and up) and take the expectation value over a period of frequency ω0 :

* *
* ∂D * ∂B * * * * * * * * ∂ueff
hE · +H · i = ω0 Im (ω0 )hE(x, t) · E(x, t)i + ω0 Im µ(ω0 )hH(x, t) · H(x, t)i +
∂t ∂t ∂t
(6.44)

where the effective electromagnetic energy density is

   
1 d(ω) * * * * 1 d(ωµ) * * * *
ueff = Re (ω0 ) hE(x, t) · E(x, t)i + Re (ω0 ) hH(x, t) · H(x, t)i (6.45)
2 dω 2 dω

So from Equation 6.44 Poynting’s theorem for dispersive media is

∂ueff * * * * * * * * * * * *
+ ∇· S = −ω0 Im (ω0 )hE(x, t)· E(x, t)i−ω0 Im µ(ω0 )hH(x, t)· H(x, t)i− J · E (6.46)
∂t

68
Chapter 7

Plane Electromagnetic Waves and


Wave Propagation

7.1 Plane Waves in a Nonconducting Medium

In the absence of sources (charges and currents) Maxwell’s equations become

*
* * * ∂B *
∇ · B = 0, ∇×E+ =0 (7.1)
∂t
*
* * * ∂D*
∇ · D = 0, ∇×H − =0 (7.2)
∂t

* * * * * * * *
We assume that we can express E(x, t) = E(x, ω)eiωt and B(x, t) = B(x, ω)eiωt (applies
* *
also for D and H)so that

* * * * * * * *
∇ · B(x, ω) = 0, ∇ × E(x, ω) − iω B(x, ω) = 0 (7.3)
* * * * * * * *
∇ · D(x, ω) = 0, ∇ × H(x, ω) + iω D(x, ω) = 0 (7.4)

* * * *
Consider uniform isotropic linear media D = E and B = µH. We take the cross product
of

* * * * * * *
∇ × (∇ × E(x, ω)) = iω(∇ × B(x, ω))

* *  *0
*
 * * * *
∇· E ) − ∇2 E) = iω ∇ × B(x, ω)
(∇(

* * *
But ∇ × B = iωµE

* *
∇2 E = iω(iµωE)
* *
∇2 E + ω 2 µE = 0
*
(∇2 + ω 2 µ)E = 0 (7.5)

69
Next,

* * * * * * *
∇ × (∇ × B(x, ω)) = −iωµ∇ × E(x, ω)

* **
*0
 * *
∇ · B ) − ∇2 B)
(∇( = −iωµ(iω B)
*
(∇2 + ω 2 µ)B = 0 (7.6)

Equations 7.5 and 7.6 are Helmholtz wave equations. We let


k = ω µ, (7.7)

where k is the wave number. Thus, we can write

*
(∇2 + k 2 )E = 0
*
. (7.8)
2 2
(∇ + k )B = 0

The phase velocity of the wave is

ω 1 c
v≡ =√ = . (7.9)
k µ n

q
µ
where n = µ0 0
is called the index of refraction, which is usually a function of frequency.
In one-dimension, the general solution to the wave equation is

u(x, t) = aeikx−iωt + be−ikx−iωt (7.10)

Since ω = kv, we can rewrite

uk (x, t) = aeik(x−vt) + be−ik(x+vt) (7.11)

We can write this in a more general form as

u(x, t) = f (x − vt) + g(x + vt) (7.12)

where f and g are arbitrary functions which represent a wave traveling in the positive and
negative x−directions, respectively, with phase velocity v. This might not be applicable if
the medium is dispersive, since the wave changes shape as it propagates.
*
Suppose we have an electromagnetic wave of frequency ω and wave vector k = k n̂. An

70
obvious solution to Equation 7.8 are plane waves:

* * * **
E(x, t) = E 0 eik n· x −iωt
* * * **
(7.13)
ik n· x −iωt
B(x, t) = B 0 e

The physical electric and magnetic fields can be obtained by taking only the real part of the
* *
equations above. The amplitudes E 0 , B 0 are constant vectors, and n̂ (unit vector) is the
direction of the wave propagation.
* * * *
Since ∇ · E = 0 and ∇ · B = 0, then it must be that

* *
n̂ · E0 = 0, n̂ · B0 = 0. (7.14)

There is no component along along the direction of propagation. This means that the electric
and magnetic fields of an electromagnetic wave are always perpendicular to the direction of
propagation. This is called a transverse wave. The curl equations on the other hand:

* * *
(∇ × E − iω B)α = 0

We consider the α−component only

* * *
(∇ × E − iω B)α = αβγ ∂β Eγ − iωBα = 0
* ** * **
= αβγ ∂β (E 0 eik n· x −iωt )γ − iω(B 0 eik n· x −iωt )α

= αβγ ∂β (E0,γ eiknζ xζ −iωt ) − iω(B0,α eiknζ xζ −iωt )

= αβγ (δβζ iknζ )(E0,γ eiknζ xζ −iωt ) − iω(B0,α eiknζ xζ −iωt )

We must have ζ = β.

= αβγ (knβ )E0,γ − ωB0,α = 0

giving us

* √ *
B0 = µn̂ × E 0 . (7.15)

* *
which implies that E and B are always perpendicular to each other. The other curl equation
* * * *
also yields Equation 7.15. Instead of using E and B, we can use D and H.

*
* D0
H 0 = n̂ × (7.16)
Z

71
p
where Z = µ/ is an impedance. Since the electromagnetic wave is transverse, then the
electric and magnetic field share the same plane. It is useful to introduce mutually orthogonal
unit vectors ˆ1 and ˆ2 . The unit vector n̂ is real and perpendicular to ˆ1 and ˆ2 . We let the

amplitude vectors be
* * √
E 0 = ˆ1 E0 , B 0 = ˆ2 µE0 (7.17)

or alternatively
* * √
E 0 = ˆ2 E00 , 1 µE00 .
B 0 = −ˆ (7.18)

where E0 and E00 are complex constants. Next we want to find the Poynting vector:

* 1* *∗
S = E×H
2
1 1 * *∗
= E×B

1 1  * ik* *
 * **
∗
= E 0 e n· x −iωt × B 0 eik n· x −iωt

1 1 −ik* n·
*
x +iωt (ik
*
(*

((( *
x −iωt
*
= e (( e E0 × B0
( ( (

(

11 √
= 1 |E0 |) × (ˆ
(ˆ 2 µ|E0 |)

r
* 1 
S= |E0 |2 n̂ (7.19)
2 µ

Next, for the time-averaged energy density

1 * *∗ 1 * * ∗
 
u= E · E + B · B
4 µ
* * ∗ * * ∗
  
1 * * *
ik x −iωt
n·
 *
ik x −iωt
n·
 1  * ik * *
x −iωt
n·
 *
ik x −iωt
n·
=  E 0 e · E 0
e + B0 e · B0
e
4 µ
 
1 * * 1* *
= E 0 · E 0 + B 0 · B 0
4 µ
 
1 1 √ √
= 1 E0 ) · (ˆ
(ˆ 1 E0 ) + (ˆ2 µE0 ) · (ˆ2 µE0 )
4 µ
1
|E0 |2 + |E0 |2

=
4

u= |E0 |2 (7.20)
2

72
7.2 Linear and Circular Polarization; Stokes Parame-

ters
*
Suppose we have two waves propagating in the same direction k = k n̂ with electric fields

* **

E 1 = ˆ1 E1 ei k · x −iωt (7.21)


* **

E 2 = ˆ2 E2 ei k · x −iωt (7.22)

and magnetic fields

* *
* √ k × E1
B1 = µ (7.23)
k
* *
* √ k × E2
B2 = µ (7.24)
k

They can be combined to a more general wave propagating in the direction

* * **

E(x, t) = (1 E1 + 2 E2 )ei k · x −iωt (7.25)

The polarization of the electromagnetic wave is determined by the phases of wave 1 and
wave 2. If the phases of E1 and E2 are equal (in-phase), then the resulting wave is a linearly
polarized wave. The angle
If they are out of phase, then we have an elliptically polarized wave (out of phase). For
a circular polarized wave we have E1 = E2 but the phase difference is 90◦ .

Figure 7.1: Polarization of an electromagnetic wave. Taken from


https://www.youtube.com/watch?v=Q0qrU4nprB0

For a circular polarized EM-wave the electric field is

* * **

2 )ei k · x −iωt
1 ± iˆ
E(x, t) = E0 (ˆ

73
Suppose we choose that the wave is propagating along the z-direction. We can choose ˆ1
and ˆ2 be the x and y basis vectors. So,

* * * * * *
E(x, t) = (E0 x̂ ± iE0 ŷ)(cos( k · x − ωt) + i sin( k · x − ωt))
h * * * * i h * * * * i
= E0 (cos( k · x − ωt) + i sin( k · x − ωt)) x̂ ± iE0 (cos( k · x − ωt) + i sin( k · x − ωt)) ŷ
h * * * * i h * * * * i
= E0 (cos( k · x − ωt) + i sin( k · x − ωt)) x̂ + E0 (i cos( k · x − ωt) ∓ sin( k · x − ωt)) ŷ

We take the real components of the electric field:

*
Ex (x, t) = E0 cos(kz − ωt)
*
Ey (x, t) = ∓E0 sin(kz − ωt) (7.26)
*
Ez (x, t) = 0

*
The case ˆ1 +iˆ
2 represents a rotation of E that is counterclockwise (left circularly polarized ),
*
while ˆ1 − iˆ
2 represents a clockwise rotation of E

7.3 Reflection and Refraction of Electromagnetic Waves

at a Plane Interface Between Dielectrics

Next, we want to know how a plane wave reflects and refracts as it meets a medium of
different permeability and susceptibility.

Suppose we have an incident wave

* * **

E = E 0 ei k · x −iωt
* * (7.27)
* √ k×E
B= µ
k

74
Then we must have refracted and

*0 *0 **

E = E 0 ei k · x −iωt
* *0 (7.28)
*0 √ k×E
B = µ
k

reflected wave
* 00 * 00 **

E = E 0 ei k · x −iωt
* * 00 (7.29)
* 00 √ k×E
B = µ .
k
The wave vectors are

* *00 √
| k| = k = kω µ (7.30)
*0 *0 p
| k | = k = kω µ0 0 (7.31)

Figure 7.2: View at the plane of incidence.

We want to be able to determine the changes amplitudes and phase for the reflected and
refracted waves. At z = 0 the phase factors are all equal:

* * *0 * *00 *
( k · x)z=0 = ( k · x)z=0 = ( k · x)z=0 (7.32)

* *0 *00
This implies that k, k , and k lie on a plane. By definition, these dot products produce

k sin i = k 0 sin r = k 00 sin r0 (7.33)

But since k = k 00 (the reflected wave same magnitude but different direction), then we must
have
k sin i
 = k 00 sin r0 → i = r0 (7.34)

which means that the angle of incidence is equal to the angle of reflection. Thus, we have

75
derived Snell’s law from geometrical optics:

√ √1
sin i k0 ω µ0 0 µ0 0 n0
·

= = √ = (7.35)
sin r k ω µ √1 n
 µ0 0

We note the boundary conditions at the interface (z = 0):

(i) E ⊥ = 0 E 0⊥

(ii) B ⊥ = B 0⊥

(iii) E k = E 0k

1 k
(iv) µ
B = 1
µ0
B 0k

This leads to

* 00 *0
 
* *
0
(E 0 + E 0 ) −  E 0 · n = 0 (7.36)
*00 * 00 *0 *0
 
* * *
k × E0 + k × E0 − k × E0 · n = 0 (7.37)
* * 00 *0 *
(E 0 + E 0 − E 0 ) × n = 0 (7.38)
*00 * 00 1 *0 * 0
 
1 * * *
(k × E 0 + k × E 0 ) − 0 (k × E 0) × n = 0 (7.39)
µ µ

7.3.1 E is perpendicular to Plane of Incidence


*
Suppose that the incident electric field E is perpendicular to the interface. We recall that

*
× means that E is pointing into the page (the +y−axis). The first boundary condition
(Equation 7.36) simply tells us that

* 00 *0
 
* *
0
(E 0 + E 0 ) −  E 0 · z = 0

0=0

76
which is meaningless. For the third boundary condition, we know that the electric fields are
all along the y−axis. So, the cross product

ŷ × ẑ = x̂

So, we produce
(E0 + E000 − E00 )x̂ = 0

Then the x−component of this equation tells us that

E0 + E000 − E00 = 0 (7.40)

Next, for the third boundary condition. We can decompose the wave vectors as

*
k = −k sin ix̂ − k cos iẑ (7.41)
*0
k = k 0 sin rx̂ + k 0 cos rẑ (7.42)
*00
k = k 00 sin r0 x̂ + k 00 cos r0 ẑ (7.43)

* *00 *00
But we know that | k| = | k |, so k = k sin i x̂ + k cos i ẑ. So that

* *
k × E 0 = −(k sin ix̂ + k cos iẑ) × (E0 ŷ)

= −kE0 sin i(x̂ × ŷ) − kE0 cos i(ẑ × ŷ)


* *
k × E 0 = kE0 cos i x̂ − kE0 sin i ẑ

This applies to the other cross-products as well. So,

1 *
((kE0 cos i x̂ − kE0 sin i ẑ) + (kE000 sin i ẑ − kE000 cos i x̂)) × z−
µ
1 0 0 *
0
(k E0 sin r ẑ − k 0 E00 cos r x̂) × z = 0
µ
k0 0
 
k 00 *
(E0 − E0 ) cos i + 0 E0 cos r x̂ × z = 0
µ µ
k0 0
 
k 00
− (E0 − E0 ) cos i − 0 E0 cos r ŷ = 0 (sign error)
µ µ

Hence

s
0 0
r

(E0 − E000 ) cos i − E cos r = 0 (7.44)
µ µ0 0

77
Next, from Equation 7.40 we have E000 = E00 − E0 . We plug it to Equation 7.44.

s
0 0
r

(E0 − (E00 − E0 )) cos i = E cos r
µ µ0 0
s
0 0
r

(2E0 − E00 ) cos i = E cos r
µ µ0 0
"s #
0
r r
 
2E0 cos i = cos r + cos i E00 (7.45)
µ µ0 µ

So
q
E00 2 µ cos i
=q q
E0 
cos i + 0
cos r
µ µ0

√ √
But n = c µ and n0 = c µ0 0


2  cos i
=√ q0
 cos i + c µµ0 cos r
2n cos i
= q0
µ0
n cos i + c µµ0 · µ0
cos r
2n cos i
=
n cos i + µµ0 n0 cos r

p
But cos r = 1 − sin2 r (from Pythagorean trig. identity).

2n cos i
= p
n cos i + µµ0 n0 1 − sin2 r
2n cos i
= p
µ
n cos i + n02 − n02 sin2 r
µ0

n0
But according to Snell’s law: sin i
sin r
= n
. So, n0 sin r = n sin i

E00 2n cos i
= √ (7.46)
E0 n cos i + µµ0 n02 − n2 sin i

Next Equation 7.40 says that E00 = E0 + E000 . We plug this to Equation 7.44.

s
0
r

(E0 − E000 ) cos i = (E0 + E000 ) cos r
µ µ0
"r s # "r s #
 0  0
cos i − cos r E0 = cos i + cos r E000
µ µ0 µ µ0

78
Then
q q
 0
E000 µ
cos i − µ0
cos r
=q q
E0  0
cos i + µ 0 cos r
µ
q q p
√  0 2
c µ µ cos i − µ0 1 − sin r
= √ q q p
c µ  cos i + 0 1 − sin2 r
µ µ0
p
µ
n cos i − µ0
n02 − n02 sin2 r
= p
µ
n cos i + n02 − n02 sin2 r
µ0
p
E000 n cos i − µµ0 n02 − n2 sin2 i
= p (7.47)
E0 n cos i + µµ0 n02 − n2 sin2 i

Equations 7.46 and 7.47 are called the Fresnel equations.

7.3.2 E is Parallel to Plane of Incidence


*
Next we consider the case where E points along the plane of incidence.

We note that the symbol · means that the vector is pointing outwards the page (pointing
at us). We write again the boundary conditions.

* 00 *0
 
* *
0
(E 0 + E 0 ) −  E 0 · n = 0
*00 * 00 *0 *0
 
* * *
k × E0 + k × E0 − k × E0 · n = 0
* * 00 *0 *
(E 0 + E 0 − E 0 ) × n = 0
*00 * 00 1 *0 * 0
 
1 * * *
(k × E 0 + k × E 0 ) − 0 (k × E 0) × n = 0
µ µ

*
Since B has no component along the plane of incidence, then the second boundary

79
condition is just 0 = 0, which doesn’t mean anything. Next, the third boundary condition

* * 00 *0
(E 0 + E 0 − E 0 ) × ẑ = 0

which produces
cos i(E0 − E000 ) − cos rE00 = 0 (7.48)

The fourth boundary condition says that

*00 * 00 1 *0 * 0
 
1 * *
( k × E 0 + k × E 0 ) − 0 ( k × E 0 ) × ẑ = 0
µ µ

* *
The vector k × E 0 is pointing along the y−axis. So, if we cross this with ẑ we get a vector
along the x−axis, and hence,

s
0 0
r

(E0 + E000 ) − E =0 (7.49)
µ µ0 0

Equation 7.48 says that

cos r 0
E000 = E0 − E
cos i 0

We plug this to Equation 7.49

s
0 0
r 
 cos r 0 
E0 + E0 − E0 = E
µ cos i µ0 0
s
0 0
r r
  cos r 0
2 E0 = E0 + E
µ µ 0 µ cos i 0

So
q
E00 2 µ
=q
E0
q
0
µ0
+ µ cos r
cos i
q
2 µ cos i
=q q
0
µ0
cos i + µ cos r
q
2 µ cos i √
c µ
=q · √
cos i + µ cos r c µ
q
0
µ0

2n cos i n0
=  q  · 0
µ0
cos i + n cos r n
0
µ c µ 0 · µ0

2nn0 cos i
= µ 02
µ0
n cos i + nn0 cos r

80
2nn0 cos i
= p
µ 02
µ0 n cos i + n n02 − n02 sin2 r
E00 2nn0 cos i
= p (7.50)
E0 µ 02
0 n cos i + n n02 − n2 sin2 i
µ

Next, Equation 7.48 says that


cos i
E00 = (E0 − E000 ) (7.51)
cos r

We plug this to Equation 7.49:

s 
0 cos i
r 
 00 00
(E0 + E0 ) − (E0 − E0 ) = 0
µ µ0 cos r
"r s # " s r #
 0 cos i 0 cos i 
− E0 = − − E000
µ µ0 cos r µ0 cos r µ

So
q q
0 cos i  √
E000 µ0 cos r
− cos r c µ
µ
=q q · · √
E0 0 cos i
+ µ cos r c µ
µ0 cos r
q √
c cos i − c  cos r √µ
0 µ
µ0
= q0 √ ·√
c µµ0 cos i + c  cos r µ
q
0 0 √
cµ µ 0 · µµ0 cos i − c µ cos r
= q 0 √
0
cµ µ 0 · µµ0 cos i + c µ cos r
µ √
µ0
(c 0 µ0 ) cos i − n cos r n0
= µ √0 0 ·
µ0
(c  µ ) cos i + n cos r n0
µ 02
µ0
n cos i − nn0 cos r
= µ 02
µ0
cos i + nn0 cos r
n
p
µ 02 0
µ0 n cos i − nn 1 − sin2 r
= p
µ 02
µ0
n cos i + nn0 1 − sin2 r
p
µ 02
µ 0 n cos i − n n02 − n02 sin2 r
= p
µ 02
µ0
n cos i + n0 n02 − n02 sin2 r
p
µ 02
E000 µ 0 n cos i − n n02 − n2 sin2 i
= p (7.52)
E0 µ 02
0 n cos i + n0 n02 − n2 sin2 i
µ

81

You might also like