Physics 214 Solution Set 2 Winter 2017

1. [Jackson, problem 11.10]

(a) For the Lorentz boost and rotation matrices K and S show that
(ǫ̂ · S)3 = −ǫ̂ · S , (1)
(ǫ̂′ · K)3 = ǫ̂′ · K ,
where ǫ̂ and ǫ̂′ are any real unit 3-vectors.

We are given
     
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 −1 0
S1 = 
0 0
, S2 =  , S3 =  ,
0 −1 0 0 0 0 0 1 0 0
0 0 1 0 0 −1 0 0 0 0 0 0
     
0 1 0 0 0 0 1 0 0 0 0 1
1 0 0 0 0 0 0 0 0 0 0 0
K1 = 
0
, K2 =  , K3 =  .
0 0 0 1 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0
To prove eq. (1), we note that the first row and column of S1 , S2 and S3 are all zeros. Hence
we can simply focus on the remaining 3 × 3 block. That is, we write
 
0 0k
(Si )jk =  ,
T
0j −ǫijk
where 0 is a row vector of three zeros, 0T is a column vector of three zeros, and


+1 , if (ijk) is an even permutation of (123) ,
ǫijk = −1 , if (ijk) is an odd permutation of (123) ,


0, otherwise ,
is the three-dimensional Levi-Civita tensor. After excluding the first row and column, jk labels
the three remaining rows and columns of Si .
Thus, we can compute (ǫ̂ · S)3 by pretending that the first row and column do not exist. In
detail,1
(ǫ̂ · S)3jk = (ǫ̂ · S)jℓ (ǫ̂ · S)ℓm (ǫ̂ · S)mk = ǫi (Si )jℓ ǫp (Sp )ℓm ǫq (Sq )mk
= −ǫi ǫp ǫq ǫijℓ ǫpℓm ǫqmk = ǫi ǫp ǫq ǫijℓ ǫpmℓ ǫqmk
= ǫi ǫp ǫq (δip δjm − δim δjp )ǫqmk = ǫq ǫqjk − ǫm ǫj ǫq ǫqmk , (2)
1
In eq. (2), we employ the Einstein summation convention. In this derivation, we make use of the antisym-
metry properties of the Levi-Civita tensor and employ the identity ǫijℓ ǫpmℓ = δip δjm − δim δjp .

1
after noting that ǫi ǫi = ǫ̂ · ǫ̂ = 1 since ǫ̂ is an arbitrary real unit vector. We now observe
that ǫm ǫj ǫq ǫqmk = 0 since ǫm ǫq is symmetric under the interchange of m and q whereas ǫqmk is
antisymmetric under the same interchange of indices. Thus, eq. (2) yields
(ǫ̂ · S)3jk = ǫq ǫqjk = −ǫq (Sq )jk = −(ǫ̂ · S)jk ,
which establishes eq. (1). Note that we could have achieved the same result by writing out
explicitly  
0 0 0 0
0 0 −ǫ3 ǫ2 
ǫ̂ · S = 
0
,
ǫ3 0 −ǫ1 
0 −ǫ2 ǫ1 0
and then computing (ǫ̂ · S)3 via matrix multiplication. Indeed,
 
0 0 0 0
0 −ǫ22 − ǫ23 ǫ1 ǫ2 ǫ1 ǫ3 
(ǫ̂ · S)2 = 
0
,
ǫ1 ǫ2 2
−ǫ1 − ǫ3 2
ǫ2 ǫ3 
0 ǫ1 ǫ3 ǫ2 ǫ3 −ǫ21 − ǫ22
and  
0 0 0 0
0 0 −ǫ3 ǫ2 
(ǫ̂ · S)3 = (ǫ̂ · S)2 ǫ̂ · S = −(ǫ21 + ǫ22 + ǫ23 ) 0
 = −ǫ̂ · S ,
ǫ3 0 −ǫ1 
0 −ǫ2 ǫ1 0
after using the fact that ǫ̂ is a real unit 3-vector.
To establish eq. (2), we can follow either of the above two methods. Employing the first
method above, we make use of the block-matrix form for ǫ̂ ′ · K
 
′
0 ǫk
(ǫ̂ ′ · K)jk =  , (3)
′
ǫj 0jk
where 0jk stands for the matrix elements of the 3 × 3 zero matrix. In particular, j labels the
row and k labels the column. Then,
      
′ ′ ′ ′ ′ ′
0 ǫℓ 0 ǫi 0 ǫk ǫ̂ · ǫ̂ 0 i 0 ǫk
(ǫ̂ ′ · K)3jk =    =  
′ ′ ′ T ′ ′ ′
ǫj 0jℓ ǫℓ 0ℓi ǫi 0ik 0j ǫj ǫi ǫi 0ik
      
1 0i 0 ǫ′k 0 ǫ′k 0 ǫ′k
=  = =  = (ǫ̂ ′ · K)jk ,
0Tj ǫ′j ǫ′i ǫ′i 0ik ǫ′j ǫ̂ ′ · ǫ̂ ′ 0jk ǫ′j 0jk
after using the fact that ǫ̂′ is is real unit vector. Thus eq. (2) is established. Alternatively, we
can write out the explicit form for the matrix,
 
0 ǫ′1 ǫ′2 ǫ′3
ǫ′1 0 0 0
ǫ̂ ′ · K = 
ǫ′2
,
0 0 0
ǫ′3 0 0 0

2
and then compute (ǫ̂ ′ · K)3 via matrix multiplication. Indeed,
 ′2 
ǫ1 + ǫ′22 + ǫ′32 0 0 0
 0 ǫ ′2
ǫ ′ ′
1 ǫ2 ǫ′1 ǫ′3 
(ǫ̂ ′ · K)2 = 

1 ,
0 ′ ′
ǫ1 ǫ2 ǫ′22 ǫ′2 ǫ′3 
0 ǫ′1 ǫ′3 ǫ′2 ǫ′3 ǫ′32

and
 
0 ǫ′1 ǫ′2 ǫ′3
ǫ′1 0 0 0
(ǫ̂ ′ · K)3 = (ǫ̂ ′ · K)2 ǫ̂ ′ · K = (ǫ′12 + ǫ′22 + ǫ′32 ) 
ǫ′2
 = ǫ̂ ′ · K ,
0 0 0
ǫ′3 0 0 0

after using the fact that ǫ̂ ′ is a real unit 3-vector.

(b) Use the result of part (a) to show that:

 
exp −ζ β̂ · K = I − β̂ · K sinh ζ + (β̂ · K)2 [cosh ζ − 1] ,

where I is the 4 × 4 identity matrix.

We employ the series expansion for the exponential (which defines the matrix exponential),
  X ∞
(−ζ)n
exp −ζ β̂ · K = (β̂ · K)n . (4)
n=0
n!

In part (a), we established the following result: (β̂ · K)3 = β̂ · K. Hence, it follows that

(β̂ · K)2n = (β̂ · K)2 , (β̂ · K)2n+1 = β̂ · K , for n = 1, 2, 3, . . . .

Thus, we can rewrite the series given in eq. (4) as

  X ζn X ζn
exp −ζ β̂ · K = I − β̂ · K + (β̂ · K)2 , (5)
n! n even n!
n odd
n≥1 n≥2

after using the fact that (β̂ · K)0 = I is the 4 × 4 identity matrix. Using,
∞
X X∞
ζ 2n+1 ζ 2n
= sinh ζ , = cosh ζ ,
n=0
(2n + 1)! n=0
(2n)!

and noting that the last summation in eq. (5) starts at n = 2, we end up with
 
exp −ζ β̂ · K = I − β̂ · K sinh ζ + (β̂ · K)2 [cosh ζ − 1] , (6)

which is the desired result.

3
REMARKS:
To understand the significance of eq. (6), let us write it explicitly in matrix form. It is
convenient to use the block matrix form of eq. (3), where j labels the row and k labels the
column,
     
1 0k 0 β̂ k 1 0k
I = , (β̂ · K)jk =  , (β̂ · K)2jk =  .
T
0j δjk β̂ j 0jk 0Tj β̂ j β̂ k
(7)
Then, eq. (6) yields
 
  cosh ζ −β̂ k sinh ζ
exp −ζ β̂ · K =  .
−β̂ j sinh ζ δjk + β̂ j β̂ k (cosh ζ − 1)

In class, we identified ζ = tanh−1 β as the rapidity, which satisfies

1
γ=p = cosh ζ , βγ = sinh ζ .
1 − β2

~ = β β̂ = (β1 , β2 , β3 ), it follows that

Hence, after writing β
 
   γ −γβk 
exp −ζ β̂ · K = 

,
βj βk 
−γβj δjk + (γ − 1) 2
β

which we recognize as the boost matrix defined in eq. (11.98) of Jackson.

2. [Jackson, problem 11.13] An infinitely long straight wire of negligible cross-sectional area
is at rest and has a uniform linear charge density q0 in the inertial frame K ′ . The frame K ′
(and the wire) move with velocity ~ v parallel to the direction of the wire with respect to the
laboratory frame K.
(a) Write down the electric and magnetic fields in cylindrical coordinates in the rest frame
of the wire. Using the Lorentz transformation properties of the fields, find the components of
the electric and magnetic fields in the laboratory.

In the rest frame of the wire (i.e. frame K ′ ), choose the z-axis to point along the wire. Then,
to compute the electric field, we draw a cylinder of length L and radius r ′ , whose symmetry
axis coincides with the z-axis. Applying Gauss’ law in gaussian units,
I
E~ ′ · n̂ da = 4πQ , (8)
S

4
where Q is the total charge enclosed inside the cylinder. In cylindrical coordinates (r ′ , φ′ , z ′ ),2
the symmetry of the problem implies that E ~ ′ (r ′ , φ′ , z ′ ) = E ′ (r ′ )ρ̂, where E ′ (r ′ ) depends only
on the radial distance from the symmetry axis. Choosing the surface S to be the surface of the
cylinder, we have n̂ = r̂, and so eq. (8) reduces to

2πρ′ LE ′ (r ′ ) = 4πQ .

Defining the linear charge density (i.e. charge per unit length) by q0 = Q/L, we conclude that3

~ ′ (r ′ ) = 2q0 r̂ .
E (9)
r′
~ ′ = 0.
Since there are no moving charges in the rest frame of the wire, it follows that B
The transformation laws for the electric and magnetic field between reference frames K and
K ′ are given by4
h i γ2 ~ ~ ~ ′
~ ~ ′ ~ ~
E =γ E −β×B − ′
β(β · E ) ,
γ+1
h i 2
~ =γ B
B ~′+β ~ ′ − γ β(
~×E ~ β
~·B ~ ′) .
γ+1

For this problem, β ~ = β ẑ. Using the results of part (a), and noting that r = r ′ (since the
radial direction is perpendicular to the direction of the velocity of frame K ′ with respect to K),
it follows that
E~ = 2γq0 r̂ , B~ = 2γβq0 φ̂ , (10)
r r
where we have used ẑ · r̂ = 0 and ẑ × r̂ = φ̂.

(b) What are the charge and current densities associated with the wire in its rest frame? In
the laboratory?

In reference frame K there are no moving charges, so that J~ ′ = 0. The corresponding charge
density is
q0
ρ′ (r ′ ) = ′
δ(r ′ ) . (11)
2πr
To check this, let us integrate over a cylinder of length L and arbitrary nonzero radius, whose
symmetry axis coincides with the z-axis. Then,
Z Z Z
ρ (r ) dV = ρ (r ) r dr dφ dz = q0 dr ′ δ(r ′ )dz ′ = q0 L = Q .
′ ′ ′ ′ ′ ′ ′

2
We denote the radial coordinate of cylindrical coordinates in frame K ′ to be r′ rather than the more
traditional ρ′ , since we reserve the letter ρ for charge density.
3
The direction of the unit vectors r̂, φ̂ and ẑ are the same in frames K and K ′ , so no extra primed-superscript
is required on these quantities.
4
Eq. (11.149) of Jackson provides the equations to transform the fields from reference frame K to reference
frame K ′ . To transform the fields from K ′ to K, simply change the sign of β. ~

5
Since J µ = (cρ ; J~) is a four-vector, the relevant transformation law between frames K and K ′
are:
~ · J~ ′ ) ,
cρ = γ(cρ′ + β (12)
γ − 1 ~ ~′ ~
J~ = J~ ′ + 2
~ ′.
(β · J )β + γ βcρ (13)
β

Plugging in J~ ′ = 0 and the result of eq. (11), and noting that β ~ = β ẑ and r ′ = r, it follows
that5
γq0 γβc q0
ρ(r) = δ(r) , J~ = ẑ δ(r) = ρ(r)v ẑ = ρ(r)~
v, (14)
2πr 2πr
after using v ≡ β c.

(c) From the laboratory charge and current densities, calculate directly the electric and
magnetic fields in the laboratory. Compare with the results of part (a).

This is an electrostatics and magnetostatics problem, so we can use Gauss’ law to compute
~ and Ampère’s law to compute B.
E ~ The computation of E ~ is identical to the one given in
part (a) with q0 replaced by γq0 . Hence, it immediately follows from eq. (9) that

~ 2γq0
E(r) = r̂ ,
r
in agreement with eq. (10). Ampère’s law in gaussian units is
I
~ · d~ 4πI
B ℓ= ,
C c

where I is the current enclosed in the loop C. With J~ given by eq. (14),
Z Z
I= J~ · n̂ da = ρ(r)v r dr dφ = γq0 v ,
A

after noting that n̂ = ẑ points along the direction of the current flow and da = rdrdφ is
the infinitesimal area element perpendicular to the current flow. Using the symmetry of the
problem, B ~ = B(r)φ̂. Thus, evaluating Ampère’s law with a contour C given by a circle
centered at r = 0 that lies in a plane perpendicular to the current flow, d~ ~ and
ℓ = rdφφ
4πI 4πγq0 v
2πrB(r) = = ,
c c
which yields
~ 2γβq0 v
B(r) = φ̂ ,
r
after using v = βc, in agreement with eq. (10).
5
We can interpret q ≡ γq0 as the linear charge density as observed in reference frame K. This is not
unexpected due to the phenomenon of length contraction.

6
3. [Jackson, problem 11.15] In a certain reference frame, a static uniform electric field E0 is
parallel to the x-axis, and a static uniform magnetic field B0 = 2E0 lies in the x–y plane,
making an angle θ with respect to the x-axis. Determine the relative velocity of a reference
frame in which the electric and magnetic fields are parallel. What are the fields in this frame
for θ ≪ 1 and θ → 12 π?

In frame K, we have
~ = E0 x̂ ,
E ~ = Bx x̂ + By ŷ ,
B (15)
with
~ ·B
E ~ = |E|
~ |B|
~ cos θ = E0 B0 cos θ = 2E02 cos θ , (16)
~ = E0 and |B|
after writing |E| ~ = B0 = 2E0 . It follows that

Bx = 2E0 cos θ , By = 2E0 sin θ . (17)

The reference frame K ′ , in which the electric and magnetic fields are parallel, is moving at
a velocity ~ ~ with respect to reference frame K. That is,
v ≡ cβ
~′×B
E ~ ′ = 0. (18)
The electric and magnetic fields in frame K ′ are related to the corresponding fields in frame K
by eq. (11.149) of Jackson. These relations can be rewritten in the following form,
~ k′ = E
E ~k , B~ k′ = B
~k , (19)
   
~ ⊥′ = γ E
E ~×B
~⊥ + β ~⊥ , ~ ⊥′ = γ B
B ~×E
~⊥ − β ~⊥ . (20)

~ and fields with a ⊥ subscript

In our notation above, fields with a k subscript are parallel to β
~ For example,
are perpendicular to β.
~×E
β ~·E
~ k = 0 and β ~⊥ = 0 ,

which implies that

~ ~ ~ ~ ~ ~
E~ k = (β · E)β and E ~⊥ = E~ − (β · E)β .
β2 β2
The form of eqs. (19) and (20) suggests that the relative velocity ~ v should lie in the z-
direction. That is,
~ = β ẑ ,
β
in which case E~ k = Ez ẑ and B ~ k = Bz ẑ. Since Ez = Bz = 0, it follows from eq. (19) that
′ ′
Ez = Bz = 0. Using eq. (20), the transverse fields are given by
Ex′ = γ(Ex − βBy ) = γE0 (1 − 2β sin θ) , Ey′ = γ(Ey + βBx ) = 2βγE0 cos θ , (21)
Bx′ = γ(Bx + βEy ) = 2γE0 cos θ , By′ = γ(Ey + βBx ) = γE0 (2 sin θ − β) , (22)
after using eqs. (15)—(17). Note that eq. (18) implies that
~′×B
Ex′ By′ − Ey′ Bx′ = (E ~ ′ )z = 0 .

7
Inserting the above results into eqs. (21) and (22) yields,
γ 2 E02 (1 − 2β sin θ)(2 sin θ − β) − 4βγ 2 E02 cos2 θ = 0 .
Multiplying out the factors above and writing cos2 θ = 1 − sin2 θ, the above equation simplifies
to
2β 2 sin θ − 5β + 2 sin θ = 0 .
This is a quadratic equation in β which is easily solved. The larger of the two roots is greater
than 1, which we reject since 0 ≤ β ≤ 1 (i.e., 0 ≤ v ≤ c). The smaller of the two roots is
non-negative and less than 1. Thus, we conclude that
p
v 5 − 25 − 16 sin2 θ
β= = . (23)
c 4 sin θ
The two limiting cases are easily analyzed. In the case of θ ≪ 1, we can work to first order
in θ. From eq. (23) we find that β ≃ 52 θ. Since θ ≪ 1 it follows that β ≪ 1, in which case

γ = (1 − β 2 )−1/2 ≃ 1 + O(β 2 ) .
Since we are working to first order in θ, we also must work to first order in β. In particular we
can neglect terms such as βθ. Hence, in this limiting case, eqs. (21) and (22) yield
~ ′ = 1B
E ~ ′ = E0 (x̂ + 2β ŷ) , for β ≃ 52 θ ≪ 1 ,
2

where we have neglected terms that are second√ order (or higher) in β. Finally, in the limit of
1 1 6
θ → 2 π, eq. (23) yields β = 2 . Then γ = 2/ 3, and eqs. (21) and (22) yield
√
E~ ′ = 0, B~ ′ = 3E0 ŷ , for θ = 12 π . (24)

REMARK:
In fact, it is easy to show that eq. (23) implies that 0 ≤ β ≤ 12 . If we multiply the numerator
p
and denominator of eq. (23) by 5 + 25 − 16 sin2 θ, we obtain,
4 sin θ
β= p .
5+ 25 − 16 sin2 θ
Since the polar angle lies in the range 0 ≤ θ ≤ π or equivalently 0 ≤ sin θ ≤ 1, it follows
immediately that β ≥ 0 (where β = 0 corresponds to sin θ = 0 as expected). Finally, it is easy
to verify that
4 sin θ 1
p ≤ . (25)
2
5 + 25 − 16 sin θ 2
6
Recall that in class, we showed that the quantity F µν Feµν = 12 ǫµναβ F µν F αβ = −4E·
~ B~ is a Lorentz invariant.
This means that if E ~ and B ~ are perpendicular in one frame, then they must be perpendicular in all frames.
1
Thus, if θ = 2 π in frame K and θ = 0 in frame K ′ , then it must be true that either the electric field or the
magnetic field (or both) vanish in frame K ′ , since the only vector that is both perpendicular and parallel to a
given fixed nonzero vector is the zero vector. This is indeed the case here, as can be seen in eq. (24).

8
Since the denominator on the left hand side above is positive, we can rewrite eq. (25) as
 p 
4 sin θ ≤ 21 5 + 25 − 16 sin2 θ . (26)

This inequality is manifestly true for sin θ = 0. For sin θ > 0, eq. (26) can be rearranged into
the following form p
8 sin θ − 5 ≤ 25 − 16 sin2 θ . (27)
Squaring both sides and simplifying the resulting expression then yields

sin θ (sin θ − 1) ≤ 0 . (28)

Dividing both sides of the equation by sin θ (which is assumed positive) yields sin θ ≤ 1, which
is valid for all polar angles θ. Hence, eq. (25) is established. The inequality becomes an equality
if sin θ = 1, in which case β = 21 .

4. [Jackson, problem 11.18] The electric and magnetic fields of a particle of charge q moving
in a straight line with speed v = βc, given by eq. (11.52) of Jackson, become more and more
concentrated as β → 1, as indicated in Fig. 11.9 on p. 561 of Jackson. Choose axes so that
the charge moves along the z axis in the positive direction, passing the origin at t = 0. Let
the spatial coordinates of the observation point be (x, y, z) and define the transverse vector ~
r⊥,
with components x and y. Consider the fields and the source in the limit of β = 1.
(a) Show that the fields can be written as

~ = 2q ~
E
r⊥
δ(ct − z) , ~ = 2q v̂ × ~
B
r⊥
δ(ct − z) , (29)
2 2
r⊥ r⊥

where v̂ is a unit vector in the direction of the particle’s velocity.

We begin with eq. (11.154) on p. 560 of Jackson,

qR~
~=
E , (30)
R3 γ 2 (1 − β 2 sin2 ψ)3/2

where ψ is the angle between the vectors ~ ~ I have modified Jackson’s notation by
v and R.
~
employing the symbol R for the vector that points from the charge q to the observation point
r = (x, y, z) in reference frame K.7 Eq. (30) was also derived in class along with the corre-
~
sponding result for the magnetic field,

q(~
v × R)~
~ =
B . (31)
cR3 γ 2 (1 − β 2 sin2 ψ)3/2
7
Jackson denotes the vector that points from the charge q to the observation point (x, y, z) by ~r. However,
I prefer to employ ~
r to represent the vector that points from the origin of reference frame K to the observation
point, as shown in Fig. 1.

9
 x

(x, y, z)

~
r ~
R
ψ
vt
~
z
q
Figure 1: A charge q moving at constant velocity ~ v in the z-direction as seen from reference
frame K. The angle ψ is defined so that v̂ · R̂ = cos ψ.

The reference frame K is exhibited in Fig. 1. It is evident from this figure that
~ =~
R r −~
vt . (32)

The velocity vector is taken to lie along the z-direction. That is, ~
v = v ẑ.
It is convenient to introduce the notation where

r ⊥ = xx̂ + y ŷ ,
~ r k = z~
~ z,

so that ~ v = 0 and ~
r⊥ · ~ rk × ~ ~ into components
v = 0. Likewise, we can resolve the vector R
parallel and perpendicular to the velocity vector,
~ =R
R ~k + R
~⊥ ,

where
~ k ≡ Rk ẑ = (z − vt)ẑ ,
R ~⊥ = ~
R r⊥ . (33)
~ ⊥ | ≡ R⊥ = R sin ψ. It follows that
after making use of eq. (32). In particular, note that |R

R3 (1 − β 2 sin2 ψ)3/2 = (R2 − β 2 R2 sin2 ψ)3/2 = (R⊥

2
+ Rk2 − β 2 R⊥
2 3/2
)

= [Rk2 + R⊥
2
(1 − β 2 )]3/2 = (Rk2 + R⊥
2
/γ 2 )3/2 . (34)

Note that in obtaining eq. (34) we used R2 = R⊥ 2

+ Rk2 and γ ≡ (1 − β 2 )−1/2 . Moreover, since
~⊥ = ~
R r⊥ [cf. eq. (33)], we may replace R⊥ with r⊥ ≡ |~ r ⊥ | = (x2 + y 2)1/2 in the above formulae.
Eqs. (30), (33) and (34) then yield
 
~= γq ~
r ⊥ + (z − vt)ẑ
E . (35)
(γ 2 Rk2 + r⊥
2 3/2
)

Likewise, eqs.(31), (33) and (34) yield

~ = γq(~
v×~ r⊥)
B 2 2 2 3/2
. (36)
c(γ Rk + r⊥ )

10
 Consider the limit of β → 1. In this limit, γ → ∞, and we see that
(
γ 0, if Rk 6= 0 ,
lim 2 2 2 3/2
=
γ→∞ (γ R + r ) ∞, if Rk = 0 .
k ⊥

This implies that

γ
lim = kδ(Rk ) , (37)
γ→∞ (γ 2 R2 2 3/2
+ r⊥ )
k

for some constant k. Note that in light of eq. (33),

lim Rk = z − ct ,
γ→∞

since γ → ∞ in the limit of v → c. To determine k, we integrate eq. (37) from −∞ to ∞,

since Rk can be any real number (either positive, negative or zero) depending on the value of
the time t. Thus, employing the substitution u = γRk ,
Z ∞ Z ∞ ∞
γ dRk du u  2
k= = =  = 2 .
2 2 2 3/2 2 2 3/2 2 2 2 1/2 
−∞ (γ Rk + r⊥ ) −∞ (u + r⊥ ) r⊥ (u + r⊥ ) −∞ r⊥

Hence, we conclude that

γ 2
lim = 2 δ(z − ct) .
γ→∞ (γ 2 Rk2 2 3/2
+ r⊥ ) r⊥
Inserting this result back into eqs. (35) and (36) yields

~ = 2q ~
E
r⊥
δ(z − ct) , ~ = 2q v̂ × ~
B
r⊥
δ(z − ct) . (38)
2 2
r⊥ r⊥
~ above, we noted that in the limit of v → c, the z-component of the electric field
In obtaining E
is proportional to (z − ct)δ(z − ct) = 0 due to a property of the delta function. In obtaining
B~ above, we noted that limv→c ~v /c = v̂. Finally, since the delta function is an even function
of its argument, we can write δ(z − ct) = δ(ct − z), and eq. (29) is verified.

(b) Show by substitution into the Maxwell equations that these fields are consistent with
the 4-vector source density
J α = qcv α δ (2) (~
r ⊥ )δ(ct − z) ,
where the 4-vector v α = (1 ; v̂).

The four-vector current is given by J µ = (cρ ; J~). Hence, using the Maxwell equations in
gaussian units,
0
~ E
∇· ~ = 4πρ = 4πJ .
c
Hence, using eq. (38) and noting that Ez = 0, it follows that
   
0 c ~ ~ c ~ ~ ∂Ez qc ~ ~
r⊥
J = ∇· E = ∇⊥ · E + = δ(z − ct)∇⊥ · 2
, (39)
4π 4π ∂z 2π r⊥

11
 ~ ⊥ ≡ x̂ ∂/∂x + ŷ ∂/∂y. For ~
where ∇ r ⊥ ≡ x x̂ + y ŷ 6= 0, an elementary computation yields
     
~ ~
r⊥ ∂ x ∂ y y 2 − x2 x2 − y 2
∇⊥ · 2
= + = + = 0. (40)
r⊥ ∂x x2 + y 2 ∂y x2 + y 2 (x2 + y 2)2 (x2 + y 2)2

To determine the behavior at ~

r ⊥ = 0, we consider the two-dimensional analogue of the diver-
gence theorem,
Z   I Z 2π
~ ~
r⊥ ~
r⊥
dxdy ∇⊥ · 2
= r⊥ dφ 2 · r̂ ⊥ = dφ = 2π , (41)
A r⊥ C r⊥ 0

where A is a circular disk and C is the circular boundary of the disk. Note that r̂ ⊥ = ~
r ⊥ /r⊥
is the outward normal to the circular boundary.
Eqs. (40) and (41) imply that
 
~ ~
r⊥
∇⊥ · 2
= 2πδ (2) (~
r⊥) , (42)
r⊥

where δ (2) (~
r⊥ ) is a two-dimensional delta function. Inserting this result into eq. (39), we end
up with
J 0 = qcδ (2) (~
r ⊥ )δ(z − ct) . (43)
Next, we employ the Maxwell equation,
~
∇ ~ − 1 ∂ E = 4π J~ ,
~ ×B (44)
c ∂t c

to evaluate J~. First, we compute

r ⊥ = ẑ × (xx̂ + y ŷ) = xŷ − y x̂ ,

v̂ × ~ (45)

where we have used the fact that ~

v points in the z direction. It then follows that
 
x̂ ŷ ẑ
   
v̂ × ~
r ⊥  ∂ ∂ ∂
~ ×
∇ 2
δ(z − ct) = det 
 ∂x ∂y

∂z 
r⊥  
−y x
δ(z − ct) δ(z − ct) 0
x2+ y2 x2 + y2
    
xx̂ + y ŷ ′ ∂ x ∂ y
=− 2 δ (z − ct) + + δ(z − ct)
x + y2 ∂x x + y2
2 ∂y x + y2
2

 
~
r⊥ ~ ⊥· ~
r⊥
= − 2 δ ′ (z − ct) + ẑ ∇ 2
δ(z − ct)
r⊥ r⊥

~
r⊥ ′
=− 2
δ (z − ct) + 2π ẑ δ (2) (~
r ⊥ )δ(z − ct) , (46)
r⊥

12
after making use of eq. (42). The prime refers to differentiation with respect to z. Finally, we
compute    
∂ ~ r⊥ ∂ ~ r⊥ c~
r⊥
2
δ(z − ct) = −c 2
δ(z − ct) = − 2 δ ′ (z − ct) . (47)
∂t r⊥ ∂z r⊥ r⊥
Inserting eq. (38) into eq. (44) and using eqs. (46) and (47), we obtain
   
~ qc ~ v̂ × ~r⊥ q ∂ ~ r⊥
J= ∇× 2
δ(z − ct) − 2
δ(z − ct)
2π r⊥ 2π ∂t r⊥
 
qc ~
r⊥ ′ (2) ~
r⊥ ′
= − 2 δ (z − ct) + 2π ẑ δ (~ r ⊥ )δ(z − ct) + 2 δ (z − ct)
2π r⊥ r⊥

= qcv̂ δ (2) (~
r ⊥ )δ(z − ct) , (48)
after using the fact that v̂ = ẑ. Combining eqs. (43) and (48), we can write
J α = qcv α δ (2) (~
r ⊥ )δ(z − ct) ,
where the four-vector v α = (1 ; v̂).

(c) Show that the fields of part (a) are derivable from either of the following 4-vector
potentials:
A0 = Az = −2qδ(ct − z) ln(λr⊥ ) , A~⊥ = 0 , (49)
or
A0 = Az = 0 , ~ ⊥ = −2qΘ(ct − z)∇
A ~ ⊥ ln(λr⊥ ) , (50)
where λ is an irrelevant parameter setting the scale of the logarithm. Show that the two
potentials differ by a gauge transformation and find the corresponding gauge function χ.

~ Given the four-vector potential, the electromagnetic

The four-vector potential is Aµ = (Φ ; A).
fields are determined by
~
E ~ 0 − 1 ∂A ,
~ = −∇A ~ =∇
B ~ ×A
~.
c ∂t
Inserting the scalar and vector potentials given in eq. (49),
h i
~ = 2q ∇
E ~ δ(ct − z) ln(λr⊥ ) + 2q ẑ ln(λr⊥ ) ∂ δ(ct − z)
c ∂t
 h i  
∂ ∂ 1 2 2 ∂ 1∂
= 2qδ(z − ct) x̂ + ŷ 2
ln(x + y ) + ln λ + 2q ẑ ln(λr⊥ ) + δ(ct − z)
∂x ∂y ∂z c ∂t
~
r⊥
= 2q 2
δ(z − ct) ,
r⊥
2
after using ~
r⊥ = xx̂ + y ŷ and r⊥ = x2 + y 2 . In particular, note that
 
∂ 1∂
+ f (ct − z) = 0 ,
∂z c ∂t

13
for any function of ct − z. Using eq. (49) to compute the magnetic field,
 
x̂ ŷ ẑ
 
∂ ∂ ∂ 
~ =∇
B ~ ×A ~ = det  
 ∂x ∂y ∂z 
 
0 0 −2q ln(λr⊥ )δ(ct − z)
 
∂ ∂
= −2qδ(ct − z) x̂ ln(λr⊥ ) − ŷ ln(λr⊥ )
∂y ∂x
 
∂ 1 2 2
 ∂ 1 2 2

= −2qδ(ct − z) x̂ ln(x + y ) + ln λ − ŷ ln(x + y ) + ln λ
∂y 2 ∂x 2

2q v̂ × ~r⊥
=− 2
(y x̂ − xŷ)δ(ct − z) = 2q 2
δ(ct − z) ,
r⊥ r⊥
after employing eq. (45).
Repeating these calculations using eq. (50),
~⊥  
~ 1 ∂A ∂ ∂ 1  ~
r⊥
E=− = 2qδ(ct − z) x̂ + ŷ 2
ln(x2 + y 2 ) + ln λ = 2q 2 δ(ct − z) ,
c ∂t ∂x ∂y r⊥
d
after using the relation between the delta function and the step function, δ(x) = Θ(x). In
dx
the computation of the magnetic field, we require the following result:
 
∂ ∂ 1  xx̂ + y ŷ
∇~ ⊥ ln(λr⊥ ) = x̂ + ŷ ln(x2
+ y 2
) + ln λ = 2 .
2
∂x ∂y x + y2
Hence, it follows that
 
x̂ ŷ ẑ
 
 ∂ ∂ ∂
~ =∇
B ~ ×A ~ = −2q det  


 ∂x ∂y ∂z 
 x y 
Θ(ct − z) 2 Θ(ct − z) 0
x + y2 x2 + y 2
    
y x̂ − xŷ ∂ y ∂ x
= 2 δ(ct − z) + ẑ Θ(ct − z) −
x + y2 ∂x x2 + y 2 ∂y x2 + y 2

v̂ × ~r⊥
= 2q 2
δ(ct − z) ,
r⊥
after employing eq. (45) and noting that
   
∂ y ∂ x 2xy 2xy
2 2
− 2 2
=− 2 2 2
+ 2 = 0.
∂x x + y ∂y x + y (x + y ) (x + y 2 )2

14
 Finally, we demonstrate that eqs. (49) and (50) differ by a gauge transformation. Under a
gauge transformation (using gaussian units),

~→A
~′ = A
~ + ∇χ
~ , 1 ∂χ
A A0 → A′ 0 = A0 − .
c ∂t
Denoting Aµ by eq. (49) and A′ µ by eq. (50), it follows that

∂χ
= −2qcδ(ct − z) ln(λr⊥ ) ,
∂t
~ ⊥ χ = −2qΘ(ct − z)∇
∇ ~ ⊥ ln(λr⊥ ) ,

∂χ
= 2qδ(ct − z) ln(λr⊥ ) .
∂z
The solution to these equations can be determined by inspection,

χ(~
x, t) = −2qΘ(ct − z) ln(λr⊥ ) ,

up to an overall additive constant.

5. [Jackson, problem 11.27]

(a) A charge density ρ′ of zero total charge, but with a dipole moment p~ ′ , exists in reference
′ ′ ′
frame K . There is no current density in K . The frame K moves with velocity ~ v = βc~ in the
frame K. Find the charge and current densities ρ and J~ in the frame K and show that there
is a magnetic dipole moment m ~ = 21 (~ ~ correct to first order in β. What is the electric
p ′ × β),
dipole moment p ~ in K to the same order in β?

We shall assume that in frame K ′ , we have J~ ′ = 0 and the charge density, ρ′ (~x ′ ) is time-
independent. Furthermore, we assume that the total charge in frame K ′ is zero, i.e.
Z
x ′ ) d3x′ = 0 ,
ρ′ (~ (51)

whereas the electric dipole moment in frame K ′ , denoted by p ~ ′ , is assumed to be nonzero. By

definition, the electric dipole moment is given by
Z
′
~ = ~
p x ′ ρ′ (~
x ′ ) d3 x′ . (52)

Frame K ′ is assumed to move with velocity ~ ~ with respect to frame K. We shall employ
v = βc
eqs. (12) and (13), which we repeat here:

~ · J~ ′ ) ,
cρ = γ(cρ′ + β (53)
γ − 1 ~ ~′ ~ ~ ′.
J~ = J~ ′ + (β · J )β + γ βcρ (54)
β2

15
Setting J~ ′ = 0, it follows that

x ′) ,
x) = γρ′ (~
ρ(~ J~(~ ~ ′ (~
x) = γ βcρ ~ x) .
x ′ ) = βcρ(~ (55)

Hence, the electric dipole moment in frame K is given by

Z
~= ~
p x) d3 x ,
x ρ(~ (56)

and the magnetic dipole moment is defined (in gaussian units) in frame K as
Z Z Z 
1 ~ 3 1 ~ 3 1 3 ~,
~ =
m x × J (~
~ x) d x = 2 ~ x × β ρ(~x) d x = 2 x ρ(~
~ x) d x × β
2c

after employing eq. (55) in the second step above. In light of eq. (56), it follows that

~ = 21 p
m ~.
~×β (57)

In order to relate p~ to p
~ ′ , we need to compare ρ(~ x ′ ). Eq. (11.19) of Jackson gives
x) and ρ′ (~
the four-vector x′ µ in terms of xµ . The inverse relation is obtained by changing the sign of β, ~
which yields

~·~ γ−1 ~ ~ + γ βx
~ ′.
x0 = γ(x′0 + β x ′) , x′ +
x=~
~ x ′ )β
(β · ~ 0 (58)
β2

Working to first order in β, we can set γ = (1 − β 2 )−1/2 ≃ 1, and

~·~
x0 ≃ x′0 + β x′ , x≃~
~ ~ ′0 .
x ′ + βx (59)

In the same approximation, d3 x′ ≃ d3 x and ρ(~ x ′ ). Hence, using eq. (51),

x) ≃ ρ′ (~
Z Z Z
~= ~
p x ρ(~ 3
x) d x ≃ (~ ′ ~ ′ ′
x + βx0 )ρ (~ ′ 3 ′
x )d x = ~ x ′ ρ′ (~
x ′ ) d3 x′ = p
~′ . (60)

to first order in β. Using eqs. (57) and (60), we conclude that to first order in β,

~ = 21 p
m ~.
~′ × β

(b) Instead of the charge density, but no current density, in K ′ , consider no charge density,
but a current density J~ ′ that has a magnetic dipole moment m ~ ′ . Find the charge and current
densities in K and show that to first order in β there is an electric dipole moment p ~=β ~×m ~ ′
in addition to the magnetic dipole moment.

x ′ ) = 0 in frame K ′ . Then, using the continuity equation

Suppose that ρ′ (~

~ ~ ′ ∂ρ′
∇·J + = 0, (61)
∂t

16
and it follows that J~ ′ is a steady current; that is,
~ ′ · J~ ′ (~
∇ x ′) = 0 . (62)

~ ′ . By definition, the
The magnetic dipole moment in frame K ′ is nonzero and is denoted by m
magnetic dipole moment is given by
Z
1
′
~ =
m x ′ × J~ ′ (~
~ x ′ ) d3 x′ . (63)
2c
Then, using eqs. (53) and (54),
γ ~ ~′ ′
ρ(~
x) = β · J (~ x ), (64)
c
γ − 1 ~ ~′ ′ ~
J~(~
x) = J~ ′ (~
x ′) + [β · J (~
x )] β . (65)
β2
~
Taking the dot product of eq. (65) with β,
~ · J~(~
β ~ · J~ ′ (~
x) = γ β x ′) . (66)

Inserting this result into eq. (64) yields

~
β
ρ(~
x) = · J~(~
x) . (67)
c
Using eq. (67), the electric dipole moment in frame K is given by
Z Z
1

~= ~
p x ρ(~ 3
x) d x = ~ ~ · J~(~
x β x) d3 x . (68)
c
Working to first order in β, we can set γ = (1 − β 2 )−1/2 ≃ 1, in which case d3 x′ ≃ d3 x and
~ · J~(~
β ~ · J~ ′ (~
x) = β x ′ ). Hence,
Z
1

~≃
p x β
~ ~ · J~ ′ (~
x ′ ) d3 x′ .
c
We now apply using eq. (59) which yields,
Z ~ x′0 Z
1 ′ ~

3 ′ β
~≃
p ~ ′
x β · J (~
~ ′
x) dx + ~
β · J~ ′ (~
x ′ ) d3 x′ . (69)
c c
This result can be simplified by using the following trick. One can express the ith component
of the current as follows,
~ ′ · J~ ′ = ∇′k (J ′ k x′ i ) ,
J ′ i = ∇′k (J ′ k x′ i ) − x′ i ∇ (70)

after applying eq. (62) in the final step. Thus,

Z Z
J d x = ∇′k (J ′ k x′ i ) d3x′ = 0 ,
′i 3 ′
(71)

17
after converting to a surface integral at infinity, which vanishes under the assumption of a
localized current. Hence, the last term in eq. (69) vanishes, and we are left with
Z
1

~≃
p x′ β
~ ~ · J~ ′ (~
x ′ ) d3 x′ . (72)
c
We can rewrite the integrand of eq. (72) by using the vector identity



x′ β
~ ~ · J~ ′ (~ ~× ~
x ′) = β x ′ × J~ ′ (~
x ′ ) + J~ ′ (~ ~·~
x ′ )(β x ′) .

Inserting this result into eq. (72) and using eq. (63), it follows that
Z
~ 1 ~·~
~ = 2β × m
p ~ +′
J~ ′ (~
x ′ )(β x ′ ) d3x′ . (73)
c

The remaining integral can evaluated by employing eq. (70) for J~ ′ . In particular, the ith
component of the integral in eq. (73) is
Z Z Z Z
1 ′i j ′j 3 ′ βj ′j ′ ′k ′i 3 ′ βj ′k ′i ′ ′j 3 ′ βj
J β x dx = x ∇k (J x ) d x = − J x ∇k (x ) d x = − J ′ j x′ i d3 x′ ,
c c c c
(74)
after an integration by parts, where the surface integral at infinity can be dropped as it vanishes
for a localized current. In the final step of eq. (74), we used ∇′k (x′ j ) = δ jk and summed over
the repeated index k. Thus, eq. (74) yields
Z Z
1

~ ′
J (~ ′ ~
x )(β · ~ ′ 3 ′
x )d x = − x′ β
~ ~ · J~ ′ (~
x ′ d3 x′ ≃ −~p,
c
where we have used eq. (72) in the final step. Inserting this result back into eq. (73) yields

~×m
~ = 2β
p ~ ′−p
~.

Solving for p
~, we end up with
~×m
~=β
p ~ ′. (75)
For completeness, we shall evaluate m ~ in terms of m ~ ′ , keeping only terms to first order
in β. In this approximation, we have d3 x ≃ d3 x′ and J~(~ x) = J~ ′ (~
x ′ ). Using these results along
with eq. (59), it follows that
Z Z Z
1 ~ 1 ~ ′ 1 ~ ′ ) × J~ ′ (~
m~ = x × J (~
~ 3
x) d x = ′
x × J (~
~ 3 ′
x )d x = (~x ′ + βx 0 x ′ ) d3 x′
2c 2c 2c
Z
x′0 ~
=m ~ +′
β × J~ ′ (~x ′ ) d3 x′ ,
2c
after using eq. (63). However, in light of eq. (71), the remaining integral above vanishes. Hence,
to first order in β,
m~ =m ~ ′. (76)

18
ADDENDUM to Problem 5

Although Jackson only asks for the the results to first order in β, it is not too hard to derive
the exact results. This derivation will be given below.

(a) By assumption, ρ′ (~x ′ ) is a time-independent charge density in frame K ′ which satisfies

eq. (51) [i.e. the total charge vanishes]. We are asked to compute p ~ and m~ in frame K. By
assumption, this means that we should evaluate these quantities at a fixed time x0 in frame K.
Using eq. (58), it follows that
x0 ~
x′0 = −β·~ x′ .
γ
Inserting this result back into eq. (58) then yields
 
′ γ−1 ~·~ ~ + βx
~ 0.
x=~
~ x + − γ (β x ′ )β
β2

We can simplify this expression by noting that γ = (1 − β 2 )−1/2 implies that β 2 γ 2 = γ 2 − 1.

Substituting for β 2 above,
γ−1 γ
2
−γ = − . (77)
β γ+1
Hence,
γ ~·~ ~ + βx
~ 0.
x=~
~ x′ − (β x ′ )β (78)
γ+1
We can evaluate the Jacobian of the transformation, eq. (78),8

∂xi γ
′
= δij − βi βj , (79)
∂xj γ+1

which we evaluate at fixed x0 . To compute the determinant of the Jacobian, we use the following
general result which is easily proved (see the Appendix following this addendum):

a| 2 .
det(δij − ai aj ) = 1 − |~

Hence,    
∂xi γβ 2 1 γ2 − 1 1
det =1− = γ+1− = .
∂x′j γ+1 γ+1 γ γ
It follows that  
3 ∂xi d3 x′
d x = det d3 x′ = , (80)
∂x′j γ
which is just the well-known length contraction in special relativity. Using eqs. (55) and (80),
it follows that the charge dq located inside an infinitesimal volume element is

dq = ρ(~ x ′ ) d3x′ ,
x) d3 x = ρ′ (~ (81)
8
In eq. (79), we do not distinguish between lowered and raised indices, as all involved quantities are three-
dimensional.

19
since the factors of γ cancel out. Eq. (81) is expected since the electric charge is a Lorentz
scalar, which must be independent of the reference frame used to evaluate it.
We now can compute p ~ in frame K. Using eqs. (78) and (81),
Z Z Z  
′ ′ γ ~·~ ′ ~ + βx
~ 0 ρ′ (~
~= ~
p x ρ(~ 3
x) d x = ~ ′
x ρ (~ 3 ′
x )d x = x −
~ (β x )β x ′ ) d3x′ . (82)
γ+1
Using eq. (52), it follows that
γ ~ ~
p ~′ −
~=p ~ ′ )β
(β · p (83)
γ+1
where we have used eq. (51) [i.e., the total charge in frame K ′ vanishes] to eliminate the last
term in eq. (82). Indeed, if we work to first order in β, then the second term on the right hand
side of eq. (83) can be dropped and we recover the result of eq. (60).
To obtain m,~ we start with the exact result obtained in eq. (57). Inserting eq. (83) for p ~
~ ~
and using β × β = 0, it follows that 9

~ = 12 p
m ~,
~′ × β (84)
which is exact to all orders in β.

REMARK:

One can derive eq. (83) more directly as follows. Using eqs. (55) and (58),
Z Z  
′ γ−1 ~ ′ ~ ~
~= ~
p x ρ(~ 3
x) d x = γ x +
~ (β · ~
x )β + γ βx ′ ′
0 ρ (~x ′ ) d3 x
β2
Z  
′ γ−1 ~ ′ ~ ~
=γ x +
~ (β · ~x )β + γ βx0 ρ′ (~
′
x ′ ) δ(x0 − ct)d4 x .
β2
R
In the last step, I inserted the integral δ(x0 − ct)dx0 = 1 which does not change the result.
This is useful, since d4 x = d4 x′ = d3 x′ dx′0 . Using eq. (58) to express x0 in terms of x′0 ,
Z    
′ γ−1 ~ ′ ~ ~ ′ ′ ′ ′ ~ ′ ct
~=
p x +
~ (β · ~
x )β + γ βx0 ρ (~ x ) δ x0 + β·~ x − d3 x′ dx′0 ,
β2 γ
~ x ′ − ct/γ). Integrating over x′ yields
after writing γδ(x0 − ct) = δ([x0 − ct]/γ) = δ(x′0 + β·~ 0
Z     Z
γ −1 ~·~ ~ ρ′ (~ ~ ρ′ (~
~=
p x′ +
~ − γ (β x ′ )β x ′ ) d3x′ + ctβ x ′ ) d3x′ .
β2
R ′ ′ ′ 3 ′
Using eqs. (51) and (77) along with p~′ = ~ x ρ (~ x ) d x , we end up with
γ ~ ~,
~′ −
~=p
p ~ ′ )β
(β · p
γ+1
which reproduces eq. (83). Of course, the two derivations are equivalent.
9
In the literature, one often finds the result of eq. (84) quoted without the factor of 12 . For a discussion
of this discrepancy, see V. Hnizdo, Magnetic dipole moment of a moving electric dipole, American Journal of
Physics 80, 645–647 (2012).

20
 (b) By assumption, J~ ′ (~
x ′ ) is a time-independent, steady current in frame K ′ which satisfies
eq. (62). We are asked to compute p ~ and m~ in frame K. In this computation, it will be quite
useful to derive a formula for the following integral,
Z
x′ i J ′j d3 x′ .

This can be accomplished using a identity similar to that of eq. (70). First we write

x′ i J ′ j = 21 (x′ i J ′j + x′ j J ′ i ) + 21 (x′ i J ′ j − x′ j J ′ i ) .

We then observe that

′
~ · J~ ′ = ∇′k (x′ i x′ j J ′ k ) ,
x′ i J ′ j + x′ j J ′ i = ∇′k (x′ i x′ j J ′ k ) + x′ i x′ j ∇

after applying eq. (62), and

x ′ × J~ ′ )k .
x′ i J ′j − x′ j J ′ i = ǫijk (~

Consequently,
Z Z Z
′i ′j 3 ′ ′i ′j ′j ′i 3 ′
x J d x = 1
2
(x J + x J ) d x + 1
2
(x′ i J ′ j − x′ j J ′ i ) d3 x′
Z Z
= 1
2
∇′k (x′ i x′ j J ′ k ) d3x′ + 1 ijk
2
ǫ x ′ × J~ ′ )k d3 x′ .
(~ (85)

The first term in the second line of eq. (85) can be converted into a surface integral at infinity,
which vanishes under the assumption that the current is localized. The second term is related
to the magnetic moment via eq. (63). Hence,
Z
1
x′ i J ′j d3 x′ = ǫijk m′ k . (86)
c
We are now ready to compute m. ~ We again perform the computation at fixed x0 , in which
case we can use eqs. (78) and (80). Employing eq. (65),
Z
1
~ =
m x × J~(~
~ x) d3 x
2c
Z    
1 γ ~ ′ ~ ~ γ − 1 ~ ~
= ′
x −
~ x )β + βx0 × J~ (~
(β · ~ ′ ′
x )+ [β · J~ (~
′ ′
x )] β d3 x′
2γc γ+1 β2
Z  
~ ′
m 1 γ − 1 ~ ~′ ′ ′ ~ γ ~ ′ ~ ~ ′ ′ ~ ~ ′ ′
= + [β · J (~x )]~
x ×β− (β · ~
x )β × J (~ x ) + x0 β × J (~ x ) d3 x′ ,
γ 2γc β2 γ+1
~×β
where some of the cross-terms vanish due to β ~ = 0. It is convenient to rewrite the above
expression in component form,
Z  
k m′ k 1 ijk γ − 1 ′i ′ℓ j ℓ γ
m = + ǫ x J β β − x J β β + x0 J β d3 x′ .
′ℓ ′j i ℓ ′j i
γ 2γc β2 γ+1

21
Using eqs. (71) and (86), the integrals above are easily evaluated,
 
k m′ k 1 ijk γ − 1 iℓn ′ n j ℓ γ ℓjn ′ n i ℓ
m = + ǫ ǫ m β β − ǫ m ββ .
γ 2γ β2 γ+1
Summing over the product of ǫ-tensors,
 
k m′ k 1 ′ n γ − 1 jℓ kn jn kℓ j ℓ γ iℓ kn in ℓk i ℓ
m = + m (δ δ − δ δ )β β − (δ δ − δ δ )β β
γ 2γ β2 γ+1
 
m′ k 1 ′ n 2 kn k n γ−1 γ
= + m (β δ − β β ) −
γ 2γ β2 γ+1
m′ k γ − 1 h 2 ′k ~ ′ k
i
= + β m − (β · m ~ )β ,
γ 2(γ + 1)
after using β 2 = (γ 2 − 1)/γ 2 . We can also use the latter to combine the first two terms above,
 
1 β 2 (γ − 1) 1 1
+ = 1+ 2 .
γ 2(γ + 1) 2 γ
Thus, we end up with
 
1 1 γ−1 ~ ~.
~ =
m 1+ 2 ~ ′−
m ~ ′ )β
(β · m
2 γ 2(γ + 1)
To first order in β, we can approximate γ ≃ 1, and it follows that m ~ ≃m ~ ′ , which confirms the
result of eq. (76).
Next, we evaluate p ~ at fixed x0 starting from eq. (68),
Z Z
1
3 1

~=
p ~ ~
x β · J (~
~ x) d x = x β
~ ~ · J~ ′ (~
x ′ ) d3 x′ ,
c c
after using eqs. (66) and (80). We now employ eq. (78) and obtain
Z  
1 γ ~ ~

~=
p ~ ′
x − β (β · ~ ′ ~
x ) + x0 β ~ · J~ ′ (~
β x ′ ) d3 x′ ,
c γ+1
which we rewrite in component form,
Z  
i 1 ′i γ
p = x − β β x + x0 β β j J ′ j d3 x′ .
i k ′k i
c γ+1
Using eqs. (71) and (86), the integrals above are easily evaluated,
 
γ ~×m
i j
p =β ǫ m − ijℓ ′ ℓ i k kjℓ ′ ℓ
ββ ǫ m = (β ~ ′ )k ,
γ+1
after noting that ǫkjℓ β j β k = 0. Hence,
~×m
~=β
p ~ ′,
which is exact to all orders in β.

22
REMARK:
The results of parts (a) and (b) can be combined by linear superposition into the following
result. Suppose a time-independent localized charge density ρ′ (~ x ′ ) of zero total charge and a
time-independent, steady localized current density J~ ′ (~
x ′ ) exist in a reference frame K ′ . In this
reference frame, the system of charges and currents possesses an electric dipole moment p ~ ′ and
a magnetic dipole moment m ′
~ . The frame K ′ moves with velocity ~ v = βc~ in the frame K.
Then, in frame K there is a charge density and current density given by:

γ − 1 ~ ~′ ′
~
x) = γ cρ′ (~
cρ(~ ~ · J~ ′ (~
x ′) + β x ′) , J~(~
x) = J~ ′ (~
x ′) + β · J (~ ~ ′ (~
x ) β + γ βcρ x ′) ,
β2

where γ = (1 − β 2 )−1/2 . Moreover, the electric and magnetic dipole moments in frame K are:
γ ~·p ~+β ~×m
~′ −
~=p
p (β ~ ′) β ~ ′,
γ+1
 
1 1 γ −1 ~ ~+1p ~.
~ =
m 1+ 2 m ~ ′− ~ ′) β
(β · m 2
~′ × β
2 γ 2(γ + 1)

The above results are exact to all orders in β. If one sets m ~ ′ = 0 [or p
~ ′ = 0], then one recovers
the results of part (a) [part (b)] above, respectively.
An equivalent result can be found by defining the unit vector β̂ ≡ β/β. ~ Then, using
2 2 2 10
β = (γ − 1)/γ , we end up with
 
1 ~×m
p~=p ′
~ − 1− ~ ′ ) β̂ + β
(β̂ · p ~ ′, (87)
γ
(   2 )
1 1 ′ 1 ′ ~ .
~ =
m 1+ 2 m ~ − 1− (β̂ · m ~′×β
~ ) β̂ + p (88)
2 γ γ

10
Eqs. (87) and (88) were obtained in George P. Fisher, The electric dipole moment of a moving magnetic
dipole, American Journal of Physics 39, 1528–1533 (1971). See also Marijan Ribarič and Luka Šušteršič, Moving
pointlike charges and electric and magnetic dipoles, American Journal of Physics 60, 513–519 (1992).

23
 APPENDIX: Proof of a determinantal formula

Theorem: Let A be an n × n matrix, whose matrix elements are given by:

Aij = δij − ai aj . (89)

Then,
n
X
2 2
det A = 1 − |~
a| , where |~
a| = a2i . (90)
i=1

Proof: The determinant of A is defined as

det A = ǫi1 i2 ···in A1i1 A2i2 · · · Anin ,

where there is an implicit sum over repeated indices. Plugging in eq. (89),

det A = ǫi1 i2 ···iN (δ1i1 − a1 ai1 )(δ2i2 − a2 ai2 ) · · · (δnin − an ain ) .

Expanding the product above, and using the Kronecker deltas to perform the sums, we obtain

det A = ǫ123···n − a1 ai1 ǫi1 23···n − a2 ai2 ǫ1i2 3···n − · · · − an ain ǫ123···in
= ǫ123···n (1 − a21 − a22 − · · · − a2n ) , (91)

after performing the final set of summations and using the fact that ǫi1 i2 ···in vanishes unless all
of its indices are distinct. Note the absence of any terms in eq. (91) that are quartic (or higher
order) in the ai , since all such terms will be symmetric under the interchange of two indices
that are summed against two corresponding indices of the Levi-Civita tensor. For example,

ǫi1 i2 ···iN a1 ai1 a2 ai2 = 0 ,

since ai1 ai2 is symmetric under the interchange of i1 and i2 whereas ǫi1 i2 ···iN is antisymmetric
under this interchange of indices.
Since ǫ123···n = 1, eq. (91) yields
n
X
det A = 1 − a2i ,
i=1

which completes the proof.

24