PHYSICAL REVIEW E VOLUME 58, NUMBER 4 OCTOBER 1998

Explicit expressions for the electric and magnetic fields of a moving magnetic dipole
José A. Heras
Instituto de Fı́sica, Universidad Nacional Autónoma de México, Apartado Postal 20-364, 01000 México Distrito Federal, Mexico
~Received 9 April 1998!
Explicit expressions for the electric and magnetic fields of an arbitrarily moving particle possessing a
constant magnetic moment are derived from retarded integrals representing the solution of Maxwell’s equa-
tions for electric and magnetic fields of a magnetized source. These expressions exhibit explicitly the useful
separation of the fields into their 1/R, 1/R 2 , and 1/R 3 parts. The total power radiated by this magnetic dipole
is then calculated when the velocity, acceleration, and the derivative of acceleration are parallel. The low
velocity limit of this power and the conservation of energy are used to derive a nonlinear damping force acting
on a nonrelativistic magnetic dipole. @S1063-651X~98!09709-8#

PACS number~s!: 03.50.De, 41.20.Bt, 41.60.Ap

I. INTRODUCTION
It is well known that a dipole at rest ~at the source point
x8 5x0 ! with a constant magnetic moment m yields the mag-
netostatic field: B5 @ 3n(n• m)2 m# /R 3 , where R5 u x2x0 u is
the magnitude of R5(x2x0 ), with x being the field point
and n5R/R. The natural question then arises: How is this
magnetostatic field modified when the dipole is set in arbi-
trary motion? This basic question of classical electrodynam-
ics was treated in the 1960s by several authors @1#. However,
it is somewhat surprising to note that none of them derived
explicit expressions for the fields of an arbitrarily moving
particle with a constant magnetic moment in terms of con-
ventional parameters ~n,b,ḃ,...! and exhibiting the useful
source are derived. In Sec. III these formulas are applied to
finding the fields of an oscillating magnetic dipole. In Sec.
IV the electric and magnetic fields of an arbitrarily moving
particle with a constant magnetic moment are derived. These
fields are expressed in terms of their 1/R, 1/R 2 , and 1/R 3
parts in Sec. V. The power radiated by the magnetic dipole is
obtained in Sec. VI and used to find a radiation damping
force.
II. THE FIELDS OF A MAGNETIZED SOURCE
Consider the time-dependent generalizations of the Cou-
lomb and Biot-Savart laws as given by Jefimenko @4–6#. In
Gaussian units these laws can be written as @3#

S D
separation of the fields into their 1/R, 1/R 2 , and 1/R 3 parts.
However, expressions for the 1/R part of these fields, the
so-called radiation fields, have recently been derived @2,3#.
E5 EEd ~u!
r n ṙ n
R 21
Rc Rc
J̇
2 2 d 3 x 8 dt 8 , ~1a!

In this paper general formulas for the electric and mag-

netic fields due to a magnetized source are derived by mak-
ing the replacements r 50 and J→c¹3M into the general-
ized Coulomb and Biot-Savart laws @4–6# and using certain
B52 EEd ~u! S n3J n3J̇ 3
R 2c
1
Rc 2
D
d x 8 dt 8 , ~1b!

integral relations that are proved in the Appendix. The gen-

where the time integration is from 2` to 1` and the spatial
eral formulas are first applied to find the electric and mag-
integration is over all space; the retardation effect is provided
netic fields of an oscillating magnetic dipole by recognizing
by the delta function d (u) with u5t 8 1R/c2t where R
that the magnetic field contains a delta term that has been
5 u x2x8 u and the overdot means differentiation with respect
ignored by the standard literature. It is shown that this delta
to t 8 . Equations ~1! can be used for deriving general expres-
term accounts for the interaction energy of two oscillating
sions for the electric and magnetic fields of a magnetized
magnetic dipoles. In a second application, novel expressions
source M in vacuum—in the context of this paper, the vector
for the electric and magnetic fields of a moving particle pos-
M is assumed to be the source of the fields E and B. This
sessing a constant magnetic moment are derived. These
means that M must be defined independently of the proper
fields are given in terms of their explicit 1/R, 1/R 2 , and 1/R 3
fields E and B ~as a dipole moment density, for example!.
parts. The total power radiated by this magnetic dipole is
The time derivative in the last term of Eq. ~1a! can be
then determined, in the simple case in which the vectors b,
taken out of the integral by performing an integration by
ḃ, and b̈ are parallel. The low velocity limit of this radiated
parts and using the property ] d (u)/ ] t 8 52 ] d (u)/ ] t. Thus,
power is used to derive a nonlinear damping force from the
conservation of energy. It is pointed out that the linearized one has the equivalent form of Eq. ~1a!:
version of this force corresponds ~up to a constant! with the
expression for a radiation damping force of a quantum par-
ticle with a spin magnetic moment recently derived by
E5 E E d Sr ~u!
n ṙ n 3
1
R 2 Rc D
d x 8 dt 8
Smirnov @7#.
The organization of this paper is as follows. In Sec. II
formulas for electric and magnetic fields of a magnetized
2
]
]
EE d
t
~ u !J 3
Rc 2
d x 8 dt 8 . ~2!

1063-651X/98/58~4!/5047~10!/$15.00 PRE 58 5047 © 1998 The American Physical Society

5048 JOSÉ A. HERAS PRE 58

With the formal substitutions r 50, J→c¹3M, Eq. ~2! be- Equations ~5! and ~10! constitute the solution of Maxwell’s
comes equations

E52
]
]t
EE d ~ u ! ¹ 8 3M 3
Rc
d x 8 dt 8 . ~3!
“•E50, ~11a!

“•B50, ~11b!
The spatial derivative in this expression can be removed by
recasting it into time derivatives. In the Appendix it is shown 1 ]B
that “3E1 50, ~11c!
c ]t

EE d ~ u ! ¹ 8 3M
Rc
d 3 x 8 dt 8
“3B2
1 ]E
c ]t
54 p “3M, ~11d!

52 EEd ~u! S n3M n3Ṁ 3

R 2c
1
Rc 2
D
d x 8 dt 8 . ~4! satisfying the conditions that the fields E and B ~and their
derivatives! vanish at infinity and the source M is confined to
a finite region of space. It should be noted that the term
Equations ~3! and ~4! combine to yield ~8p/3!M in Eq. ~10! is a function evaluated at the field point

EEd S D
and the present time. Without the presence of this ‘‘contact’’
n3Ṁ n3M̈ 3
E5 ~u! 1 d x 8 dt 8 . ~5! term, Eqs. ~5! and ~10! do not strictly satisfy Maxwell’s
R 2c Rc 2 equations. Equation ~10! may be interpreted then as follows.
The field B is formed by two terms: The integral term rep-
Consider now Eq. ~1b! rewritten in the equivalent form resents the value of B outside the magnetized source while

EE d ~ u ! n3J 3 the contact term ~8p/3!M represents its value inside the
B52 d x 8 dt 8 source. A remarkable property of Eqs. ~5! and ~10! is that
R 2c they do not involve spatial derivatives of the vector M,

E E d
] ~ u ! n3J 3 which in most cases simplifies considerably the calculation
2 d x 8 dt 8 . ~6! of the fields. It should be also mentioned that Eqs. ~5! and
] t Rc 2
~10! ~without the contact term! were previously derived in
With the substitution J→c¹3M, Eq. ~6! becomes Ref. @2#.

B52 EE d ~ u ! n3 ~ ¹ 8 3M! 3
R2
d x 8 dt 8
III. THE OSCILLATING MAGNETIC DIPOLE

Consider a particle at rest ~at the point x0 ! with a mag-

EE d
] ~ u ! n3 ~ ¹ 8 3M! 3 netic moment oscillating in time m5 m (t)e, where e is its
2 d x 8 dt 8 . ~7! direction. The associated magnetization vector is given by
]t Rc
M(x,t)5em (t) d (x2x0 ). With this source the integration of
In the Appendix it is shown that Eqs. ~5! and ~10! yields

EE d ~ u ! n3 ~ ¹ 8 3M! ṁ ~ t 8 ! n3e m̈ ~ t 8 ! n3e

d 3 x 8 dt 8 E5 1 , ~12a!
R 2 R 2c Rc 2

52 EE d~ u ! S 3n~ n•M! 2M n~ n•Ṁ! 2Ṁ

R3
1
R 2c
D B5
m ~ t 8 ! $ 3n~ e•n! 2e% ṁ ~ t 8 ! $ 3n~ e•n! 2e%
R3
1
R 2c

8pM m̈ ~ t 8 ! n3 ~ n3e! 8 p m ~ t ! e
3d 3 x 8 dt 8 2 , ~8! 1 1 d ~ x2x0 ! , ~12b!
3 Rc 2 3

EE d ~ u ! n3 ~ ¹ 8 3M! where now n5(x2x0 )/ u x2x0 u and the overdot means dif-
d 3 x 8 dt 8 ferentiation with respect to t 8 5t2R/c with R5 u x2x0 u .
Rc Equations ~12! ~without the delta term! are the well-known

S D
fields of an oscillating magnetic dipole.
52 EEd ~u!
2n~ n•M! n3 ~ n3Ṁ! 3
R 2c
1
Rc 2
d x 8 dt 8 ~9!
The novelty in Eq. ~12b! is the presence of the contact
term (8 p /3) m (t)ed (x2x0 ), which represents the magnetic
field within the dipole. It is somewhat surprising to find that
and therefore Eqs. ~7!–~9! combine to give the expression this delta term is not usually mentioned in the standard

S
literature—in the static regime, however, the analogous delta
B5 EEd ~u!
3n~ n•M! 2M 3n~ n•Ṁ! 2Ṁ
R3
1
R 2c
term is well known @8#; it is precisely the term that accounts
for hyperfine splitting in the ground state of hydrogen @9,10#.

D
However, the delta term is necessary here for the consistency
n3 ~ n3M̈! 3 8pM of Eqs. ~12! with the Maxwell equations. Moreover, this
1 d x 8 dt 8 1 . ~10!
Rc 2 3 delta term accounts for the interaction energy of two oscil-
PRE 58 EXPLICIT EXPRESSIONS FOR THE ELECTRIC AND . . . 5049

lating magnetic dipoles. Indeed, the energy of an oscillating

magnetic dipole m5 m (t)e in the presence of a magnetic
field B is given by U52 m (t)e•B. In particular, the energy
E52 m3
1 d
c dt
Ed 8 „t 1R ~ t 8 ! /c2t… S D n~ t 8 !
R~ t8!2
dt 8

of an oscillating magnetic dipole m1 5 m 1 (t)e1 in the pres-

ence of the magnetic field B of another oscillating magnetic
dipole m2 5 m 2 (t)e2 is
1 d2
2m 2 2
c dt
Ed 8 „t 1R ~ t 8 ! /c2t… S D n~ t 8 !
R~ t8!
dt 8 , ~15!

m 1 ~ t 8 ! m 2 ~ t ! $ 3 ~ e1 •n!~ e2 •n! 2e1 •e2 % where now R(t 8 )5 u x2r(t 8 ) u and n(t 8 )5 @ x2r(t 8 ) # /R(t 8 ).
U52 Using the formula
R3

2
ṁ 1 ~ t 8 ! m 2 ~ t ! $ 3 ~ e1 •n!~ e2 •n! 2e1 •e2 %
R 2c Ed „f ~ t 8 ! 2a…g~ t 8 ! dt 8 5
g~ t 8 !
u d f /dt 8 u
U f ~ t 8 ! 5a
, ~16!

m̈ 1 ~ t 8 ! m 2 ~ t ! $ ~ e1 •n!~ e2 •n! 2e1 •e2 %

2
Rc 2 with f (t 8 )5t 8 1R(t 8 )/c, a5t, and d f /dt 8 512n• b, the
time integrals in Eq. ~15! can be performed, yielding
8p
2 m 1 ~ t ! m 2 ~ t ! d ~ x2x0 !~ e1 2e2 ! , ~13!

F G
3
1 d n
E52 m3
c dt R ~ 12n• b!
2
ret

F G
where x2x0 is the separation of the dipoles. When the di- 2
1 d n
poles are separated the delta term in Eq. ~13! can be ignored. 2m 2 2 . ~17!
However, when the dipoles are at the same place the delta- c dt R ~ 12n• b! ret
term contribution accounts for the interaction energy.

IV. A MOVING PARTICLE POSSESSING A CONSTANT This equation can also be written in the convenient form
MAGNETIC MOMENT

F S D S DG
The problem of finding the fields of a point dipole in
arbitrary motion is somewhat different from that of comput- 1 d n 1 d2 n
E5 1 3 m,
ing the fields of a point charge in arbitrary motion @2#. In the c dt R ~ 12n• b!
2
c dt R ~ 12n• b!
2 2
ret
case of a moving charge e one solves the usual Maxwell’s ~18!
equations with the sources r (x,t)5e d $ x2r(t) % and J(x,t)
5ev(t) d $ x2r(t) % . By analogy one might think that for a on the understanding that @ dF/dt # ret means dF(t 8 )/dt and
moving dipole possessing a constant magnetic moment m the not dF(t 8 )/dt 8 , that is, the ‘‘ret’’ outside the square brackets
problem consists in solving Maxwell’s equations ~11! with applies to the arguments of the functions inside and not to
the source M(x,t)5 md $ x2r(t) % . However, there is a subtle the variable of differentiation @11#.
difference between the two cases. The electric charge is con- By a similar procedure, when M(x,t)5 md „x2r(t)… is
served and Lorentz invariant and thereby it is necessarily inserted into Eq. ~10! and the integrations over the resulting
independent of motion. Nevertheless, the magnetic dipole expression are performed one obtains the magnetic field of
moment is not a Lorentz invariant. However, in the model an arbitrarily moving magnetic dipole:
assumed in this paper the point dipole is observed in a frame
where there is only magnetization and it is given by

F S D
M(x,t)5 md „x2r(t)….
3n~ n• m! 2 m 1 d 3n~ n• m! 2 m
Equation ~5! can be written in the equivalent form B5 1
R 3 ~ 12n• b! c dt R 2 ~ 12n• b!

E5
1 ]
c ]t
EE d ~ t 8 1R/c2t ! S D
n3M 3
R2
d x 8 dt 8 1 S
1 d 2 n3 ~ n3 m!
c 2 dt 2 R ~ 12n• b! DG 1
ret
8p
3
md ~ x2r~ t ! ….
~19!
1
1 ]2
c2 ]t2
EE d ~ t 8 1R/c2t !S D n3M 3
R
d x 8 dt 8 .
A similar expression but without the delta term was obtained
by Monaghan in Ref. @1#. In the derivation of Eq. ~19!, how-
~14! ever, the assumption that m is a constant vector has not yet
been used. This means that Eq. ~19! is valid even in the case
With the magnetization M(x,t)5 md $ x2r(t) % , the volume that m is a function of time. However, by performing some
integrals in Eq. ~14! can be done immediately, yielding derivatives in Eq. ~19! and making d m/dt50 one obtains
5050 JOSÉ A. HERAS PRE 58

B5 F 3n~ n• m! 2 m
1 2
3 1 dn
R ~ 12n• b! R ~ 12n• b! c dt
3 ~ n• m! 1n
1 dn
c dt
•m H S DJ 1 $ 3n~ n• m! 2 m%
1 d
S 1
c dt R ~ 12n• b!
2 D
1 $ n3 ~ n3 m! %
1 d2 1
S
c dt R ~ 12n• b!
2 2 1
1
R ~ 12n• b! c dt
2 D
2 d 2n
H
2 3 ~ n3 m ! 1 m3
1 d 2n
c 2 dt 2
3n 1
2 dn
c dt
3
1 dn
c dt
•m S D S DJ
1 H 4 dn
c dt
3 ~ n3 m! 12 m3
1 dn
c dt
3n S DJ S DG
1 d 1
c dt R ~ 12n• b!
1
ret
8p
3
md $ x2r~ t ! % . ~20!

This is an expression for the magnetic field of a moving dipole with constant magnetic moment.

V. THE EXPLICIT 1/R 3 , 1/R 2 , AND 1/R PARTS OF THE FIELDS

Although Eqs. ~18! and ~19! are relatively simple, they do not exhibit explicitly the useful separation of the fields into their
1/R, 1/R 2 , and 1/R 3 parts. Such a separation of the fields, however, can be accomplished by performing all the specified time
derivatives in Eqs. ~18! and ~20!. This task, although straightforward, is extremely laborious. It involves long and complicated
vector manipulations and the full expressions obtained for the fields turn out to be very lengthy.
It is convenient to begin with Eq. ~18! rewritten as

E5 n FH S1 d 1
c dt R ~ 12n• b!
2
1 d2
1 2 2
1
c dt R ~ 12n• b!D S DJ 1
1 dn
H1
1
2 d 1
c dt R 2 ~ 12n• b! c dt R ~ 12n• b! S DJ
1
1 d 2n 1
S
c dt R ~ 12n• b!
2 2 DG ret
3 m. ~21!

By using the result @3# 1 d 1

S
c dt R ~ 12n• b!
2 3 5 2
3n• ḃ
D
R ~ 12n• b! 5 c
1 d
~ R i ~ 12n• b! j ! ~ n3 b! 2 22 b• ~ n2b!
c dt 2 .
R 3 ~ 12n• b! 5

5 jR i ~ 12n• b! j22 2 S n• ḃ ~ n3 b! 2
c
1
R
, D ~23d!

The last two expressions are used to derive the following

result:
2iR i21
~ 12n• b! j21
~ n• b! ~ i, j integers! , ~22!

( ḃ5d b/dt 8 ) one obtains

1 d2 1
c dt R ~ 12n–b!
2 2 S5 D
3 ~ n• ḃ! 2
R ~ 12n• b! c 5 21
n• b̈
R ~ 12n• b! 4 c 2

1 d
S 1
c dt R ~ 12n• b!
5
n• ḃ
D b• ~ n2 b!
3 1 2
R ~ 12n• b! c R ~ 12n• b! 3
, 1
2 $ 2 b• ~ n2 b! 2 ~ n3 b! 2 % ~ n• ḃ!
R 2 ~ 12n• b! 5 c
~23a!
n3 ~ n3 b! • ḃ22 b• ḃ1n• ḃ
1
R 2 ~ 12n• b! 4 c
1 d
S 1
c dt R 2 ~ 12n• b!
5 2
n• ḃ
D 1
b• ~ n2 b!
R ~ 12n• b! 3 c R 3 ~ 12n• b! 3 1
b• ~ n2 b ! $ 2 b• ~ n2 b! 2 ~ n3 b! 2 %
R 3 ~ 12n• b! 5
n• b
1 , ~23b!
R ~ 12n• b! 2
3
~ n3 b! 2
2 , ~23e!
R 3 ~ 12n• b! 4
1 d
S1
c dt R ~ 12n• b! 3 5
3n• ḃ
D
R ~ 12n• b! 5 c where b̈5d 2 b/dt 8 2 . The time derivatives of the unit vector
n are given by
2 ~ n3 b! 2 2 b• ~ n2 b!
2 ,
R 2 ~ 12n• b! 5 1 dn n3 ~ n3 b!
5 , ~23f!
~23c! c dt R ~ 12n• b!
PRE 58 EXPLICIT EXPRESSIONS FOR THE ELECTRIC AND . . . 5051

1 d 2 n n3 $ ~ n2 b! 3 ḃ% n3 ~ n3 b!~ 12 b 2 ! n~ n3 b! 2 n3 ~ n3 b!
2 5 1 2 2 2 . ~23g!
2
c dt R ~ 12n• b! c
3
R ~ 12n• b!
2 3
R ~ 12n• b! 2 R 2 ~ 12n• b!

This last expression was derived by Wang @12#. Substitution of Eqs. ~23a!, ~23b!, and ~23e!–~23g! into Eq. ~21! yields an
expression whose terms are arranged in order of decreasing powers of 1/R:

E5 F n@ b• ~ n2 b! $ 2 b• ~ n2 b! 2 ~ n3 b! 2 % # n3 ~ n3 b! $ 2 b• ~ n2 b! 112 b 2 % 2n~ n3 b! 2 2n$ b• ~ n2 b! %

R 3 ~ 12n• b! 5
1
R 3 ~ 12n• b! 4
1 3
R ~ 12n• b! 3

2n$ 2 b• ~ n2 b! 2 ~ n3 b! 2 % ~ n• ḃ! n$ n3 ~ n3 b! • ḃ22 b• ḃ1n• ḃ% 1n3 $ ~ n2 b! 3 ḃ% 12n3 ~ n3 b!~ n• ḃ! %
1 1
R 2 ~ 12n• b! 5 c R 2 ~ 12n• b! 4 c

1
n~ n• ḃ!
1
3n~ n• ḃ! 2
R ~ 12n• b! c R ~ 12n• b! c
2 3 5 21
n~ n• b̈!
R ~ 12n• b! 4 c 2
G 3 m.
ret
~24a!

By a similar procedure, the use of Eqs. ~23a!, ~23b!, ~23e!–~23g! into Eq. ~20! yields

B5 F n3 ~ n3 m! $ 2 b• ~ n2 b! 2 ~ n3 b! 2 %$ b• ~ n2 b! % n3 ~ n3 m!~ n3 b! 2
R 3 ~ 12n• b! 5
2
R 3 ~ 12n• b! 4
m3 ~ n3 b!@ 112 ~ n• b! 23 b 2 # 12 $ n3 ~ n3 b! % 3 ~ n3 m!@ 112 ~ n• b! 23 b 2 #
1
R 3 ~ 12n• b! 4
$ 3n~ n• m! 2 m%$ b• ~ n2 b! % 22n3 ~ n3 m!~ n3 b! 2 12n3 ~ n3 b! $ n3 ~ n3 b! • m% 22 m~ n3 b! 2
1
R 3 ~ 12n• b! 3
3n3 ~ n3 b!~ n• m! 13n$ n3 ~ n3 b! • m% 1 $ 3n~ n• m! 2 m% ~ n• b! 2 2 $ n3 ~ n3 b! % 3 ~ n3 m! 1 m3 ~ n3 b!
1
R 3 ~ 12n• b! 2

3n~ n• m! 2 m 2n3 ~ n3 m! $ 2 b• ~ n2 b! 2 ~ n3 b! 2 % ~ n• ḃ! m3 @ n3 $ ḃ1n3 ~ b3 ḃ! % #

1 1 1
R 3 ~ 12n• b! R 2 ~ 12n• b! 5 c R 2 ~ 12n• b! 4 c

n3 ~ n3 m! $ n3 ~ n3 b! • ḃ22 b• ḃ1n• ḃ% 1 @ 2n3 $ ~ n2 b! 3 ḃ% # 3 ~ n3 m!

1
R 2 ~ 12n• b! 4 c

4 $ n3 ~ n3 b! % 3 ~ n3 m!~ n• ḃ! 12 m3 ~ n3 b!~ n• ḃ! $ 3n~ n• m! 2 m% ~ n• ḃ! 3n3 ~ n3 m!~ n• ḃ! 2

1 1 1
R 2 ~ 12n• b! 4 c R 2 ~ 12n• b! 3 c R ~ 12n• b! 5 c 2

1
n3 ~ n3 m!~ n• b̈!
R ~ 12n• b! 4 c 2
G 1
ret
8p
3
md $ x2r~ t ! % . ~24b!

The complicated expressions in Eqs. ~24! represent the elec-
tric and magnetic fields of an arbitrarily moving particle with
a constant magnetic moment. It is now possible to answer the
question in the Introduction: How is the magnetostatic field
of a magnetic dipole modified when it is set in arbitrary
motion? Answer: An arbitrarily moving magnetic dipole
modifies strongly its magnetostatic field in accordance with
Eqs. ~24!. However, a detailed interpretation of these equa-
tions is a very complicated task. But there are some points
that are relevant for an interpretation of these equations:
~1! Near, intermediate, and far fields. As may be seen in
Eqs. ~24!, the fields separate naturally into three parts: The
near fields Enear and Bnear , which vary as 1/R 3 and depend
on the velocity and magnetic moment; the intermediate fields
Eint and Bint , which vary as 1/R 2 and depend on the velocity,
acceleration, and the magnetic moment; and the far fields
Efar and Bfar , which vary as 1/R and depend on the velocity,
acceleration, time derivative of acceleration, and the mag-
netic moment. Thus, the complete fields read E5Enear1Eint
1Efar , and B5Bnear1Bint1Bfar1Bdel , where Bdel denotes
the delta-function term (8 p /3) md $ x2r(t) % , which is evalu-
ated at the field point and the present time—this term repre-
sents the magnetic field within the moving dipole; it is es-
sential for achieving the consistency of Eqs. ~24! with the
Maxwell equations ~11!.
~2! Static limit. When the velocity, acceleration, and de-
rivative of acceleration of the magnetic dipole are zero, that
is, when the dipole is at absolute rest, Eq. ~24a! yields E
50 and Eq. ~24b! reduces to the well known static form:
B5 $ 3n(n• m)2 m% /R 3 1(8 p /3) md $ x2x0 % , where x0 is the
point where the dipole is at rest @8#.
~3! Uniform motion. It follows from Eqs. ~24! that, in
contrast to the simple form of the fields of a charge moving
with constant velocity, the fields of a magnetic dipole in
5052 JOSÉ A. HERAS PRE 58

uniform motion ( ḃ5 b̈50) exhibit an exceedingly compli-

cated form. The electric field is given by the 1/R 3 part of Eq.
~24a! and the magnetic field by the 1/R 3 part of Eq. ~24b!
plus the delta term, that is, E5Enear and B5Bnear1Bdel .
~4! Coulombian acceleration fields. The presence of the
acceleration vector in the intermediate fields is remarkable—
for a moving charge the acceleration appears only in the far
fields. This result is better recognized when one assumes low
velocities. Indeed, if the velocity of the dipole is small com-
pared with that of light ( b !1) then the intermediate fields in
Eqs. ~24! reduce to

Eint5 F 3n~ n• ḃ! 2 ḃ

R 2c
G ret
3 m, ~25a!
FIG. 1. Geometry of the radiation process.

d P ~ t 8 ! dN ~ t ! c

F G
5 @ 12n• b# ret5 u REu 2 @ 12n• b# ret .
$ 6n~ n• m! 22 m% ~ n• ḃ! 2n~ m• ḃ! 2 ḃ~ n• m! dV dV 4p
Bint5 . ~28!
R 2c ret
~25b!
By using Eq. ~26a! one obtains

F H
These fields may be called ‘‘Coulombian acceleration
fields.’’ The point here is that these fields, though depending d P~ t8! 1 9 ~ n• ḃ! 4 6 ~ n• b! 2 ~ n• b̈!
linearly on the acceleration, are not radiation fields since 5 ~ n3 m ! 2
1
dV 4pc3 ~ 12n• b! 9 ~ 12n• b! 8
they vary as 1/R 2 .
~5! Radiation fields. Evidently, the far fields appearing in
Eqs. ~24!, 1
~ n• b̈! 2
~ 12n• b! 7
JG . ~29!

F G
ret
3n3 m~ n• ḃ! 2 n3 m~ n• b̈!
Erad5 5 21 , ~26a! In order to find the total radiated power P(t 8 ) at a fixed time
R ~ 12n• b! c R ~ 12n• b! 4 c 2 ret
t 8 , it is necessary to specify the vectors m, b, ḃ, and b̈. The
simplest example of Eq. ~29! is one in which the vectors b,
Brad5 @ n# ret3Erad , ~26b!
ḃ, and b̈ are parallel. For the sake of simplicity consider a
are radiation fields. These fields depend on the velocity and magnetic dipole that is moving along the Z axis. The geom-
linearly on the derivative of the acceleration as well as on the etry is illustrated in Fig. 1. Therefore, b5ẑb , ḃ5ẑ ḃ , b̈
square of the acceleration. Now, in general, b, ḃ, and b̈ are 5ẑb̈ , and m5ẑm . With these specific values and with dV
independent. This allows one to assume b50 and ḃ50 at 5sin ududf and n5x̂(sin u cos f)1ŷ(sin u sin f)1ẑ cos u,
least instantaneously. In this case Eqs. ~26! reduce to Eq. ~29! is first integrated over f:

Erad5 F n3 m~ n• b̈!
Rc 2
G ret
, ~27a!
P~ t8!5
9m2
F E
ḃ 4
p sin3 u cos4 u
du G
2c 3 0 ~ 12 b cos u ! 9
ret

F E G
Brad5 @ n# ret3E. ~27b!
3m2 p sin3 u cos3 u
1 ḃ 2 b̈ du
Therefore, even when both the velocity and the acceleration c3 0 ~ 12 b cos u ! 8
ret
of a magnetic dipole are instantaneously equal to zero ~at the
retarded time!, the dipole can still produce a radiation field
on account of the derivative of its acceleration. 1
m
2c 3
2

F E
b̈ 2
0
p sin u cos u
3 2

~ 12 b cos u ! 7
du G ret
. ~30!

VI. TOTAL RADIATED POWER BY A MOVING

The integrals over u are computed directly and the resulting
MAGNETIC DIPOLE
expressions can be written in terms of powers of g 5(1
Consider now the energy flux associated with the radia- 2 b 2 ) 21/2:
tion fields. It is given by the Poynting vector S
5(c/4p )Erad3Brad5(c/4p ) u Eradu 2 n. With this vector one
defines the radiated power dN(t)/dV5(S•n)R 2
5(c/4p ) u REradu . This is the energy per unit time and unit
2
E0
p sin3 u cos4 u
~ 12 b cos u ! 9
d u 52
4 21b 4 118b 2 11
S
35 ~ b 21 ! 7 ~ b 11 ! 7 D
solid angle that is radiated in the direction n at time t. The 4
radiated power N(t) is connected with the radiated power 5 ~ 40g 14260g 12121g 10! ,
35
P(t 8 ) expressed in terms of the dipole’s own time by means
of the relationship @13# ~31a!
PRE 58 EXPLICIT EXPRESSIONS FOR THE ELECTRIC AND . . . 5053

E 0
p sin3 u cos3 u
~ 12 b cos u ! 8
d u 5
32
S
7 b 3 13 b
105 ~ b 21 ! 6 ~ b 11 ! 6 D the integrals on the right-hand side of Eq. ~34! can be per-
formed by parts

5
32b
105
~ 10g 1227 g 10! , ~31b! E t1
t2
Frad•vdt52
18m 2 2
35c 7
@ a a•v #
t2
t1
2
2m2
15c 5
t
@ ȧ•a2ä•v# t 2
1

E p sin3 u cos2 u
~ 12 b cos u ! 7 d u 52
4
S
7 b 2 11
15 ~ b 21 ! 6 ~ b 11 ! 6 D 1
54m 2
35c 7
E t2

t1
a 2 ȧ•vdt2
2m2
15c 5
E
t1
t2
&
a•vdt,
0
~36!
4
5 ~ 8 g 1027 g 8 ! . ~31c! where & a5ẑa^ . The question now is under what conditions the
15
first two terms in Eq. ~36! vanish. A first case would be when
Substituting these values into Eq. ~30! and performing some the motion is such that a 2 a•v50 and ȧ•a2ä•v50 at t
manipulation one ends up with the total power radiated by 5t 1 and t5t 2 . A second case would be when the motion is
the magnetic dipole when the vectors m, b, ḃ, and b̈ are periodic since then both quantities a 2 a•v and ȧ•a2ä•v have
parallel: the same value at t5t 1 and t5t 2 . In a third case one might
assume that the time interval t 2 2t 1 is sufficiently short in
18m 2 4 such a way that the state of the system is approximately the
P~ t8!5 @ ḃ ~ 40g 14260g 12121g 10!# ret same at t5t 1 and t5t 2 @5#. In any case one ends up with the
35c 3
expression
32m 2

ES D
1 @ bḃ 2 b̈ ~ 10g 1227 g 10!# ret t2 54m 2 a 2 ȧ 2 m 2&
a
35c 3
Frad2 1 •vdt50. ~37!
t1 35c 7 15c 5
2m2
1 @ b̈ 2 ~ 8 g 1027 g 8 !# ret . ~32!
15c 3 From this equation one can identify the following radiation
reaction force:
Evidently, the dynamics of the dipole is disturbed by this
radiation loss since it provokes a reaction force back on the 54m 2 2 2 m 2
dipole. To find an expression for the radiation reaction force Frad5 ȧa 2 &
a. ~38!
35c 7 15c 5
Frad , consider the low velocity limit of Eq. ~32!. The ap-
proximation b !1 implies g '1 and thereby bg 5 Ag 2 21 This force is really unusual; the second of its terms is pro-
'0. This approximation also implies that the effect of retar- portional to the third derivative of the acceleration, that is, to
dation becomes unimportant @10#. Therefore, by writing ḃ the fifth derivative of the position r(t). Hence, the equation
5a/c and b̈ 5ȧ/c Eq. ~32! reduces to of motion is a fifth order differential equation:

P5
18m 2 a 4 2 m 2 ȧ 2
35c 7
1
15c 5
. ~33! m
d 2r
dt 2
5F ext1
54m 2 d 2 r
35c 7 dt 2 S D 2
d 3r 2 m 2 d 5r
2
dt 3 15c 5 dt 5
, ~39!

It should be noted that an analogous formula for the electric where m is the mass of the dipole and Fext is an external
dipole has been recently derived @3#. As expected, the for- force. Therefore, it is necessary to specify five initial condi-
mulas for the total power radiated by electric and magnetic tions in order to solve Eq. ~39!. When a 2 ȧ and &
a are the same
dipoles exhibit exactly the same structure. Therefore, the as- order over a brief interval, the first term in Eq. ~39! is neg-
sociated radiation reaction forces will exhibit necessarily the ligible when compared with the second one and thus Eq. ~39!
same form. can be approximated by its linear term
To determine Frad from the conservation of energy it is
necessary that the work done by this force on the magnetic 2m2
Frad52 &
a. ~40!
dipole in the interval t 1 ,t,t 2 must be equal to the negative 15c 5
of the energy radiated, that is, * t2
t1 Frad•vdt52 * t1 Pdt. The
t2

use of Eq. ~33! yields Using Newton’s second law this force takes the form

E t2

t1
Frad•vdt52
18m 2
35c 7
Et1
t2
4
a dt2
2m2
15c 5
E
t1
t2
ȧ dt. ~34!
2
a52 t 3&
a,

where t is a characteristic time defined by

~41!

With the aid of the results

d
t5 S 2m2
15mc 5 D 1/3
. ~42!
a 4 5 ~ a 2 a•v! 23a 2 ȧ•v, ~35a!
dt For example, the characteristic time for an electron ~m
59.28310221 erg G21 and m59.11310228 G! is t 58.05
d
&•v, 310223 sec. It is interesting to note that this characteristic
ȧ2 5 ~ a•ȧ2ä•v! 1a ~35b!
dt time is one order of magnitude greater than the characteristic
5054 JOSÉ A. HERAS PRE 58

time involved in the usual Abraham-Lorentz formula associ- where « i pq is the three-dimensional Levi-Civita symbol with
ated to the charge of the electron, a5 t ȧ, which has the value « 12351; M i 5(M) i ; (¹ 8 ) i 5 ] / ] x 8 i , and R5 u x i 2x 8 i u . The
t 56.26310224 sec. summation convention on repeated indices is adopted. The
The three linearly independent solutions of Eq. ~41! are right-hand side of Eq. ~A1! can be expressed as

a1 ~ t ! 5k1 e 2t/ t ,

a2 ~ t ! 5k2 e t/2t cos~ )t/2t ! ,

~43a!

~43b!
« i pq d ~ u ! ] M q
Rc
5
]
]x8p ]x8p S
« i pq d ~ u ! M q
Rc
2
Rc D
« i pq M q ] d ~ u !
]x8p

a3 ~ t ! 5k3 e t/2t
sin~ )t/2t ! , ~43c! 2
« i pq d ~ u ! M q ]
c ]x8 R
p
1
. SD ~A2!

where k1 , k2 , and k3 are vectorial constants. Evidently, a2

The derivatives in the last two terms are
and a3 are runaway solutions. However, a1 is a reasonable
solution from a physical point of view since it predicts that
acceleration spontaneously decreases exponentially with
time, which clearly agrees with the law of inertia. This pre-
]d~ u !
]x8p
52
n p ]d~ u !
c ]t8
,
]
SD
1
]x8p R
np
5 2,
R
~A3!

diction is contrary to that of the usual Abraham-Lorentz

where n p 5(R) p /R. With these derivatives, Eq. ~A2! takes
equation of the point charge which has no natural solution
the form
consistent with the law of inertia.
Finally, it is interesting to note that Eq. ~40! corresponds
~up to a constant! with the following expression derived by
Smirnov @7#:
« i pq d ~ u ! ] M q
Rc ]x8 p5
]
]x8 p S
« i pq d ~ u ! M q
Rc
2 D
d ~ u ! « ipq n p M q
R 2c
« i pq n p M q ] d ~ u !
2m2 1 . ~A4!
Frad52 5 &a, ~44! Rc 2 ]t8
3c

for a quantum nonrelativistic particle with zero electric The last term may be rewritten as

S D
charge, mass m, and spin magnetic moment m5g m 0 s @here
« i pq n p M q ] d ~ u ! ] d ~ u ! « i pq n p M q
g is the g factor of the particle, m 0 is the corresponding 5
magneton, and s5 $ s i % (i51,2,3) is the set of Pauli matri- Rc 2
]t8 ]t8 Rc 2
ces#. d ~ u ! « i pq n p ] M q
2 , ~A5!
ACKNOWLEDGMENTS
Rc 2 ]t8

The author thanks Lorea Chaos, Jaime Albarrán, and and therefore Eq. ~A4! takes the final form

S D
Javier Benitéz who collaborated with him in some aspects of
this work. Valuable comments by Dr. Octavio Novaro, Dr. « i pq d ~ u ! ] M q « i pq n p M q « i pq n p ] M q
p 52 d ~ u ! 1
Eugenio Ley Koo, and Dr. Karo Michaelian are gratefully Rc ]x8 R 2c Rc 2 ] t 8

S D
acknowledged.
] d ~ u ! « i pq n p M q
1
APPENDIX: DERIVATION OF EQS. „4…, „8…, AND „9… ]t8 Rc 2

Consider the ith Cartesian component of the integrand of

the left-hand side of Eq. ~4!:
1
]
]x8p Rc S
« i pq d ~ u ! M q
. D ~A6!

S d ~ u ! ¹ 8 3M
Rc D i
5
« ipq d ~ u ! ] M q
Rc ]x8p
, ~A1!
When this expression is integrated over space and time one
obtains

EE « i pq d ~ u ! ] M q 3
Rc ]x8p
d x 8 dt 8 52 EE d S ~u!
« i pq n p M q « i pq n p ] M q 3
R 2c
1
Rc 2 ] t 8 D
d x 8 dt 8

1 E F E ]]8 S d
t Rc 2 D G
~ u ! « i pq n p M q
dt 8 d 3 x 8 1 E F E ] ]8 S x p
« i pq d ~ u ! M q 3
Rc D G
d x 8 dt 8 .

~A7!

The time integration in the second term of the right-hand side of Eq. ~A7! gives zero because the delta function vanishes for
t 8 56`. The volume integral of the last term of Eq. ~A7! becomes a surface integral, and hence vanishes at infinity assuming
the magnetization is contained in a finite volume. Thus, Eq. ~A7! reduces to
PRE 58 EXPLICIT EXPRESSIONS FOR THE ELECTRIC AND . . . 5055

EE « ipq d ~ u ! ] M q 3
Rc ]x8p
d x 8 dt 8 52 EE d~ u ! S « i pq n p M q « i pq n p ] M q 3
R 2c
1
Rc 2 ] t 8 D
d x 8 dt 8 . ~A8!

The vector version of this expression is precisely Eq. ~4!. With these derivatives, Eq. ~A10! becomes
Equation ~8! will be now demonstrated. The ith Cartesian
component of the integrand of the left-hand side of Eq. ~8!, d ~ u ! « i pq « qst n p ] M t « i pq « qst n p n s M t ] d ~ u !

S D
5
d ~ u ! ¹ 8 3 ~ ¹ 8 3M! i
d ~ u ! « ipq « qst n p ] M t R2 ] x s8 R 2c ]t8
5 . ~A9!
Rc

The right-hand side of this equation can be expressed as

R2 ] x s8
2 d ~ u ! « i pq « qst M t S 3n p n s 2 d sp
R3
D
d ~ u ! « i pq « qst n p ] M t « ipq « qst n p M t ] d ~ u ! 4p
52 1 d ~ u ! « i pq « qst M t d sp d ~ x2x8 !
R 2
] x s8 R2 ] x s8 3

2d~ u !« ipq
« qst M
] np
t
S D
] x s8 R 2
1
] x s8
S
] d ~ u ! « i pq « qst n p M t
R2
. D
S D
~A12!
] d ~ u ! « ipq « qst n p M t
1 .
] x s8 R2
The first term of the right-hand side can be written as
~A10!
The derivatives in the first two terms of the right-hand side
are
« i pq « qst n p n s M t ] d ~ u !
2
R c ]t8
5
]t8 S
] « i pq « qst n p n s M t
R 2c D
]d~ u ! ns ]d~ u ! d ~ u ! « i pq « qst n p n s ] M t
52 , 2
] x s8 c ]t8 R 2c ]t8
~A13!
] np
S D
] x s8 R 2
5
3n p n s 2 d sp 4 p s
R3
5 d d ~ x2x8 ! .
3 p
~A11!
and therefore Eq. ~A12! takes the form

d ~ u ! « i pq « qst n p ] M t
R 2
] x s8
52
d ~ u ! « ipq « qst n p n s ] M t
2
R c ]t8
2 d ~ u ! « i pq
« qst M t
R 3S
3n p n s 2 d sp
1
4p
3
D
d ~ u ! « i pq « qst M t d sp d ~ x2x8 !

1
]t8 S
] d ~ u ! « ipq « qst n p n s M t
R c2 1 D
] x s8
S
] d ~ u ! « i pq « qst n p M t
R2
. D ~A14!

Using the identity « ipq « qst 5« ipq « stq 5 d is d tp 2 d it d sp in the first three terms of the right-hand side of Eq. ~A14! it becomes

d ~ u ! « i pq « qst n p ] M t
R 2
] x s8
52 d ~ u ! S
3n i n t M t 2M i n i n t Ṁ t 2Ṁ i
R 3 1 2
R c
2
8p
3
D
d ~ u ! M i d ~ x2x8 ! 1
] d ~ u ! « i pq « qst n p n s M t
]t8 R 2c S D
1
] x s8
S
] d ~ u ! « ipq « qst n p M t
R2
. D ~A15!

Integration over space and time of this expression yields

EE d ~ u ! « i pq « qst n p ] M t 3
R2 ] x s8
d x 8 dt 8 52 EE d S ~u!
3n i n t M t 2M i n i n t Ṁ t 2Ṁ i 3
R3
1
R 2c
D
d x 8 dt 8

2
8p
3
EE d ~ u ! M i d ~ x2x8 ! d 3 x 8 dt 8 1 E F E ]]8 S d t
~ u ! « i pq « qst n p n s M t
R 2c D G
dt 8 d 3 x 8

1 E F E ]] 8 S d
xs R2 D G
~ u ! « i pq « qst n p M t 3
d x 8 dt 8 . ~A16!
5056 JOSÉ A. HERAS PRE 58

The time integration in the second term of the right-hand side of Eq. ~A16! gives zero because the delta function vanishes for
t 8 56`. The volume integral of the last term of Eq. ~A16! becomes a surface integral, and hence vanishes at infinity assuming
the magnetization is confined to a finite region of space. Thus, Eq. ~A16! reduces to

EE d ~ u ! « ipq « qst n p ] M t 3
R2 ] x s8
d x 8 dt 8 52 EE d ~u! S 3n i n t M t 2M i n i n t Ṁ t 2Ṁ i 3
R 3 1 2
R c
D
d x 8 dt 8 2
8p i
3
M. ~A17!

When this expression is written in vector notation Eq. ~8! is obtained.

The derivation of Eq. ~9! is similar to that of Eq. ~8!. Accordingly, after a manipulation similar to that for obtaining Eq.
~A14!, the following identity is derived:

d ~ u ! « i pq « qst n p ] M t
Rc ] x s8
S
52 d ~ u !
2n i n t M t n i n t Ṁ t 2Ṁ i
R3
1
Rc 2
1
]t8
D S
] d ~ u ! « i pq « qst n p n s M t
Rc 2
1
] x s8
D Rc S
] d ~ u ! « i pq « qst n p M t
. D
~A18!
If this identity is integrated over space and time the result is

EE d ~ u ! « i pq « qst n p ] M t 3
Rc ] x s8
d x 8 dt 8 52 EE d~ u ! S 2n i n t M t n i n t Ṁ t 2Ṁ i 3
R3
1
Rc 2
D
d x 8 dt 8

1 E S ]]8 S d
t
~ u ! « i pq « qst n p n s M t
Rc 2 D D
dt 8 d 3 x 8 1 E S ]] 8 S d
xs Rc D D
~ u ! « i pq « qst n p M t 3
d x 8 dt 8 .

~A19!
The time integration in the second term of the right-hand side of Eq. ~A19! gives zero because the delta function vanishes for
t 8 56`. The volume integral of the last term of Eq. ~A19! becomes a surface integral, and hence vanishes at infinity assuming
the magnetization is confined to a finite region of space. Thus, Eq. ~A19! reduces to

EE d ~ u ! « ipq « qst n p ] M t 3
Rc ] x s8
d x 8 dt 8 52 EE d ~u! S R3
1
Rc 2
D
2n i n t M t n i n t Ṁ t 2Ṁ i 3
d x 8 dt 8 . ~A20!

The vector version of this equation is precisely Eq. ~9!.

@1# J. R. Ellis, Proc. Cambridge Philos. Soc. 59, 759 ~1963!; G. N. J. A. Heras, ibid. 62, 525 ~1994!; 63, 928 ~1995!; M. Bornatici
Ward, Proc. R. Soc. London, Ser. A 279, 562 ~1964!; Proc. and U. Bellotti, ibid. 64, 568 ~1996!.
Cambridge Philos. Soc. 61, 547 ~1965!; J. R. Ellis, J. Math. @7# A. Y. Smirnov, J. Phys. A 30, 1135 ~1997!.
Phys. 7, 1185 ~1966!; M. Kolsrud and E. Leer, Physica Nor- @8# D. J. Griffiths, Am. J. Phys. 60, 979 ~1992!.
vegica 2, 181 ~1967!; J. J. Monaghan, J. Phys. A 1, 112 ~1968!. @9# D. J. Griffiths, Am. J. Phys. 50, 698 ~1982!.
@2# J. A. Heras, Am. J. Phys. 62, 1109 ~1994!. @10# J. D. Jackson, Classical Electrodynamics ~Wiley, New York,
@3# J. A. Heras, Phys. Lett. A 237, 343 ~1998!. 1975!, 2nd ed., Sec. 14.5.
@4# O. D. Jefimenko, Electricity and Magnetism ~Electrect Scien- @11# L. Eiges, The Classical Electromagnetic Field ~Dover, New
tific Co., Star City, WV, 1989!, 2nd ed., p. 516. York, 1972!, Sec. 14.3 and Sec. 14.5.
@5# J. B. Marion and M. A. Heald, Classical Electromagnetic Ra- @12# C. Wang, Phys. Rev. E 47, 4358 ~1993!.
diation ~Saunders, Forth Worth, TX, 1995!, Sec. 8.2. @13# G. Scharf, From Electrostatics to Optics ~Springer-Verlag,
@6# D. J. Griffiths and M. A. Heald, Am. J. Phys. 59, 111 ~1991!; Berlin, 1994!, Sec. 3.5.