Self-torque and angular momentum balance for a spinning charged sphere

Béatrice Bonga, Eric Poisson, and Huan Yang

Citation: American Journal of Physics 86, 839 (2018); doi: 10.1119/1.5054590

View online: https://doi.org/10.1119/1.5054590
View Table of Contents: http://aapt.scitation.org/toc/ajp/86/11
Published by the American Association of Physics Teachers

Articles you may be interested in

Kepler's equation and some of its pearls
American Journal of Physics 86, 849 (2018); 10.1119/1.5055760

The exponential benefits of logarithmic thinking

American Journal of Physics 86, 859 (2018); 10.1119/1.5058771

Relativistic description of the photoelectric effect

American Journal of Physics 86, 825 (2018); 10.1119/1.5052359

A hybrid approach for using programming exercises in introductory physics

American Journal of Physics 86, 831 (2018); 10.1119/1.5058449

Experimental verification of the adiabatic transfer in Wilberforce pendulum normal modes

American Journal of Physics 86, 818 (2018); 10.1119/1.5051179

The Great Silence: The Science and Philosophy of Fermi's Paradox

American Journal of Physics 86, 878 (2018); 10.1119/1.5053112
Self-torque and angular momentum balance for a spinning charged
sphere
atrice Bonga
Be
Perimeter Institute for Theoretical Physics, Waterloo, Ontario N2L 2Y5, Canada
Eric Poisson
Department of Physics, University of Guelph, Guelph, Ontario N1G 2W1, Canada
Huan Yang
Department of Physics, University of Guelph, Guelph, Ontario, N1G 2W1, Canada and Perimeter Institute
for Theoretical Physics, Waterloo, Ontario N2L 2Y5, Canada
(Received 24 April 2018; accepted 30 August 2018)
Angular momentum balance is examined in the context of the electrodynamics of a spinning
charged sphere, which is allowed to possess any variable angular velocity. We calculate the
electric and magnetic fields of the (hollow) sphere, and express them as expansions in powers of
s/tc  1, the ratio of the light-travel time s across the sphere and the characteristic time scale tc
of variation of the angular velocity. From the fields we compute the self-torque exerted by the
fields on the sphere, and argue that only a piece of this self-torque can be associated with
radiation reaction. Then we obtain the rate at which angular momentum is radiated away by the
shell, and the total angular momentum contained in the electromagnetic field. With these results
we demonstrate explicitly that the field angular momentum is lost in part to radiation and in part
to the self-torque; angular momentum balance is thereby established. Finally, we examine the
angular motion of the sphere under the combined action of the self-torque and an additional
torque supplied by an external agent. VC 2018 American Association of Physics Teachers.
https://doi.org/10.1119/1.5054590

I. INTRODUCTION AND SUMMARY
The aim of this paper is to investigate the electrodynamics
of a spinning charged sphere. Our wish is to bring together
two themes of the standard textbook literature, one rather
neglected, the other shrouded in mystery.
The neglected theme is electromagnetic angular momen-
tum, which tends to be mentioned in passing in most texts,
but rarely explored beyond a very small number of exam-
ples. For example, Griffiths1 introduces the field angular
momentum in Sec. 8.2.4 and computes it explicitly for a
solenoid encased between two cylindrical shells.2 Apart
from another appearance in Problem 8.17, this is the sole
mention of electromagnetic angular momentum in this popu-
lar text. As another example, Jackson3 mentions the field
angular momentum on a few occasions, but the topic is not
deemed worthy of a section of its own.
The theme shrouded in mystery is self-force and radiation
reaction. Despite valiant attempts to elucidate this topic in
the standard texts (see, for example, Sec. 11.2.2 in Griffiths,
or Chapter 16 of Jackson), these aspects of electromagnetic
theory remain conceptually challenging to most students
(and instructors), thanks to the lurking infinities, mass
renormalization, runaway solutions, pre-acceleration, and
the like. This topic has been the subject of a vast literature,
and American Journal of Physics has been an active partici-
pant. A simple pedagogical introduction to the subject can
be found in Boyer’s paper,4 and a useful historical survey
was provided by Rohrlich.5 In Ref. 6, Griffiths et al. compare
the predictions of the standard equation of motion for a
charged particle, the Abraham-Lorentz equation, to those of
the less pathological Landau-Lifshitz equation (see Sec. 75
of Ref. 7). The issue of mass renormalization is discussed by
Griffiths and Owen.8 A clever derivation of the self-force
was devised by Griffiths and Szeto9 on the basis of a dumb-
bell model for an electric charge; their construction was
revisited and generalized by Ori and Rosenthal.10,11
In our opinion, the conceptual difficulties associated with
self-force and radiation reaction stem from a misguided
insistence that the Maxwell-Lorentz theory should apply to a
point particle. While the electromagnetic field of a point par-
ticle is perfectly well defined, its action back on the particle
is not, and much contortion is required to obtain self-
consistent equations of motion for the particle. Our point of
view is that in the classical framework of the Maxwell-
Lorentz theory, equations of motion should only be formu-
lated for extended blobs of charge, and a pointlike limit
should be recovered only when the electromagnetic field
varies on a length scale that is long compared with the size
of the blob. In this context the equations of motion are neces-
sarily approximate, and self-consistency can be achieved
only within the limits of the approximation. For a rigorous
implementation of this idea, see Ref. 12.
Our interest in this paper is with the electrodynamics of an
infinitely thin, spherical shell that is uniformly charged and
spinning on a central axis with a variable angular velocity.
The spinning motion of the sphere implies that angular
momentum shall be one of its most important attributes, and
thus is the first theme introduced.13 The fact that the angular
velocity is variable implies that the shell emits electromag-
netic radiation, which takes angular momentum away from
the system; this loss must be balanced by a self-torque acting
on the shell, and we have our second theme. Such issues
were explored previously in a paper by Stump and Pollack,14
who calculated the self-torque acting on a point magnetic
dipole. We feel that the spinning shell, with its finite exten-
sion, provides a much more satisfactory starting point for
this discussion.

839 Am. J. Phys. 86 (11), November 2018 http://aapt.org/ajp C 2018 American Association of Physics Teachers
V 839
 The system’s angular-momentum vector J is subjected to
two conservation equations, which are established in Sec. II.
The first concerns Jem , the electromagnetic piece of the
angular momentum, and it reads
d em
N em  T em ;
J ¼ N (1)
dt
where N em is the rate at which angular momentum is radi-
ated away, and T em is the electromagnetic torque—the self-
torque—acting on the shell. The equation states that the field
angular momentum is communicated partly to the radiation
and partly to the shell. The second conservation statement
concerns Jmech , the shell’s mechanical angular momentum,
and it reads
d mech
J ¼ T em þ T ext ; (2)
dt
where T em is again the self-torque, and T ext is the torque sup-
plied by an external agent acting on the shell. The equation
states that the shell’s angular momentum changes because of
the combined action of both torques. The sum of Jem and
Jmech is the total angular momentum J, and according to Eqs.
(1) and (2), it satisfies the conservation statement dJ=dt
¼ N N em  T ext . The total angular momentum is not con-
served, because of radiative losses and the torque exerted by
the external agent.
The conservation statement of Eq. (1) can be verified by
calculating the field angular momentum, the radiated losses,
and the self-torque. This requires expressions for the shell’s
electric and magnetic fields at any time t and any position x.
Provided that we introduce a mild and sensible assumption,
the fields of a spinning charged sphere can be calculated for
any variable angular velocity. The assumption is that the
angular velocity is taken to vary on a characteristic time
scale tc that is very long compared with s ¼ R/c, the light-
travel time across (half) the sphere; R is the sphere’s radius
and c is the speed of light. The fields are calculated in Sec.
III and presented as expansions in powers of s/tc  1. In
principle, these can be carried out to an arbitrarily high
order; we truncate them after a representative number of
terms.
With the fields in hand, we calculate the self-torque
in Sec. IV, and express it also as an expansion in powers of
s/tc. We find that only pieces of the self-torque can be asso-
ciated with radiation reaction; the remaining pieces (which
include the dominant contribution to the self-torque) have
nothing to do with radiation reaction. As we explain in Sec.
IV, the self-torque can be decomposed into terms that come
with even powers of s/tc (and therefore even powers of c),
and terms that come with odd powers of s/tc (and therefore
odd powers of c). The odd terms are sensitive to the choice
of boundary condition at infinity; they change sign when
the retarded solution describing an outgoing wave is
switched to the advanced solution describing an incoming
wave. These odd terms, therefore, can be associated with
radiation reaction, with the justification that a switch of
boundary condition will produce a change in the self-
torque. The even terms, on the other hand, are insensitive
to the choice of boundary condition, and stay the same after
a switch; these terms have nothing to do with radiation
reaction. The dominant contribution to the self-torque is
even in s/tc.
840 Am. J. Phys., Vol. 86, No. 11, November 2018
The radiated angular momentum is calculated in Sec. V,
and presented as an expansion in powers of s/tc. As can be
expected from the preceding discussion, each term is odd in c,
and therefore sensitive to the choice of boundary condition at
infinity. We note that the radiated angular momentum (like
the self-torque) is proportional to q, the total charge of the
spinning sphere. This provides a vivid illustration of the fact
that the flux of angular momentum depends not only on the
radiative, 1/r part of the electromagnetic field, but also on its
Coulombic, 1/r2 part; this point was made forcibly in a recent
paper by Ashtekar and Bonga.15 The field’s angular momen-
tum is calculated in Sec. VI, and Eq. (1) is verified explicitly.
In Sec. VII, we turn to the shell’s angular motion, gov-
erned by Eq. (2). We imagine that an external agent applies a
torque on the shell, in addition to the self-torque supplied by
the electromagnetic field. The resulting equation of motion is
again presented as an expansion in powers of s/tc, with the
external torque giving the dominant contribution, and the
self-torque splitting up into radiation-reaction terms (odd in
s/tc) and terms (even in s/tc) that have nothing to do with
radiation reaction. A feature of the equation of motion is that
it is written in terms of an effective mass Meff, the sum of the
shell’s mass and the additional inertia contributed by the
electrostatic field; this is a form of mass renormalization, a
ubiquitous procedure of self-force calculations. Another fea-
ture of the equation of motion is that it requires a reduction
of order (see Sec. 75 of Ref. 7 or Ref. 16) to avoid the emer-
gence of runaway solutions. In our case, this procedure is
natural and justified on the grounds of the underlying
assumption that s/tc  1. We examine several examples of
the shell’s angular motion, corresponding to different exter-
nal torques.
The electrodynamics of a spinning charged sphere unlocks
a number of important lessons. First, it presents an excellent
opportunity to reflect on the role of angular momentum in
electromagnetism, a largely neglected subject. Second, it
provides a vivid example of the fact that the radiated angular
momentum may not solely depend on the 1/r, radiative part
of the electromagnetic field, but can also depend on the 1/r2,
Coulombic part. Third, it provides a conceptually clean setup
to understand how the field can act back on its source; the
self-torque is well defined and can be computed explicitly,
and this should help lift some of the aura of mystery that
tends to accompany this topic. Fourth, it reveals that the
notions of self-torque and radiation reaction must be clearly
distinguished. Indeed, the self-torque contains two types of
terms: time-antisymmetric terms that reflect the choice of
outgoing-wave boundary condition and can therefore be
associated with radiation reaction, and time-symmetric terms
that are insensitive to the choice of boundary condition and
have therefore nothing to do with radiation reaction. And
fifth, it provides an instance in which the equation of motion
for the source of an electromagnetic field is well defined and
can be solved explicitly. These lessons can all be learned at a
reasonable cost: All calculations are straightforward and
accessible to a starting graduate student, and they might also
be tractable to a well-educated senior undergraduate (instruc-
tors should be able to cope as well). We believe that these
lessons are well worth learning, and that this paper could
form the basis of an independent study module in a graduate
course in electromagnetism.
This paper is dedicated to the memory of Steve Detweiler,
who understood better than most the power of simple models
to foster understanding of complicated things.
Bonga, Poisson, and Yang 840
II. CONSERVATION STATEMENTS d em
a  Ta ;
J ¼ Nem em
(8)
dt a
In this section, we establish a number of conservation
statements relating to the momentum and angular momen- where
tum of an electromagnetic field. These are formulated in þ
some generality, with the assumption that the field is sourced
by a bounded charge distribution supported by a fluid; for N em
a :¼  em
abc xb Tcd dSd (9)
S
simplicity we take the fluid to be electrically and magneti-
cally unpolarized. The conservation laws will then apply, as is the field angular momentum crossing out of S per unit
a special case, to the field sourced by a spinning charged time, and
sphere. ð
Throughout the paper we denote three-dimensional vec-
T a :¼ abc xb fcem d 3 x
em
(10)
tors with a bold-italic symbol such as E, and denote individ- V
ual components with Ea or ðEÞa . We shall move freely
between the notations, making the most convenient choice is the torque exerted by the fields on the charge distribution
given the context. We adopt the Einstein summation conven- in V. Equation (8) states that the rate at which the field angu-
tion, according to which repeated vector (or tensor) indices lar momentum leaves V is equal to the rate at which it
are summed over. As examples of our notation, we write crosses out of S added to the torque exerted on all charges
jEj2 ¼ E  E ¼ Ea Ea and ðE  BÞa ¼ abc Eb Bc , where abc is within V. The equation is established by taking the time
the completely antisymmetric permutation symbol. We let derivative inside the integral in Eq. (7), substituting Eq. (6),
@ t :¼ @/@t stand for partial differentiation with respect to and integrating by parts.
time t, and @ a :¼ @/@xa for partial differentiation with respect The fluid that supports the charge distribution contributes
to the spatial coordinate xa. mechanical momentum and angular momentum. Let qm be
We begin with a conservation statement for the linear the fluid’s mass density, and let va be its velocity field. Mass
momentum contained in the electromagnetic field. Let conservation is expressed by the continuity equation
@ tqm þ @ a(qmva) ¼ 0, and in the absence of charges and
gem
a :¼ 0 ðE  BÞa (3) fields, the fluid is subjected to the momentum theorem

be the field’s momentum density, and @t gmech

a
mech
þ @b Tab ¼ faext ; (11)
 
1 where gmech ¼ qm va is the mechanical momentum density,
em
Tab :¼ 0 Ea Eb  dab jEj2 mech
a
2 Tab is the mechanical stress tensor, and faext is the force
  density supplied by external agents. In the case of a perfect
1 1
 Ba Bb  dab jBj2 (4) fluid, for example, the mechanical stress tensor is given by
l0 2 mech
Tab ¼ qm va vb þ p dab , where p is the pressure, and the
em momentum theorem becomes qm dva =dt þ @a p ¼ faext . This is
be the field’s stress tensor, defined such that Tab dSb is the recognized as Euler’s equation.
momentum crossing out of an element of surface dSb per If we now allow the fluid to be charged, and allow for the
unit time.17 Finally, let presence of an electromagnetic field, the momentum theorem
becomes
faem :¼ qEa þ ðj  BÞa (5)
@t ga þ @b Tab ¼ faext ; (12)
be the Lorentz force density on a charge distribution
described by the charge density q and current density j. Then where ga :¼ gmech
a þ gem
a is the total momentum density,
the conservation statement Tab :¼ Tab þ Tab is the total stress tensor, and faext is the
mech em

force density provided by external agents (excluding the

@t gem em em
a þ @b Tab ¼ fa (6) Lorentz force density). In the case of a perfect fluid, the
momentum theorem becomes qm dva =dt þ @a p  faem ¼ faext ,
follows as a consequence of Maxwell’s equations. This is the Euler’s equation for a charged fluid.
momentum theorem for the electromagnetic field, which is The momentum theorem for a charged fluid gives rise to a
proved in the standard texts (see, for example, Sec. 8.2.2 of conservation statement for the total angular momentum con-
Ref. 1 or Sec. 6.7 of Ref. 3). tained in the region V
The momentum theorem gives rise to an angular- ð
momentum theorem, which we formulate in integral form.
Let V be a region of three-dimensional space bounded by a Ja :¼ ðx  gÞa d3 x ¼ Jamech þ Jaem : (13)
V
closed two-surface S. We assume that V and S are fixed in
space. Let We assume that the fluid occupies a bounded region within
ð V, and that this region does not extend all the way to S. In
Jaem :¼ ðx  gem Þa d 3 x (7) this case Tab evaluated on S consists of a field contribution
V only, and the mechanical contribution vanishes. The conser-
vation statement is
be the total angular momentum of the electromagnetic field
in V; x denotes the position vector. Then a consequence of d
a  Ta ;
Ja ¼ Nem ext
(14)
Eq. (6) is the conservation statement dt

841 Am. J. Phys., Vol. 86, No. 11, November 2018 Bonga, Poisson, and Yang 841
where Nem a is the field angular momentum crossing out of S Because ja is time dependent (via m), the computation of
per unit time, as given in Eq. (9), and the vector potential is more involved. It is helpful to work
ð with Ay – iAx, noting that jz ¼ 0 and jy  ijx / sin h ei/ , a
T ext :¼  xb fcext d 3 x (15) spherical harmonic with labels ‘ ¼ 1 and m ¼ 1. We take
a abc
V advantage of a decomposition of the Green’s function in
spherical harmonics, provided by
is the torque supplied by the external agents. X
Equations (8) and (14) imply Gðt; x; t0 ; x0 Þ ¼ g‘ ðt; r; t0 ; r0 ÞY‘m

ðh0 ; /0 ÞY‘m ðh; /Þ;
‘m
d mech
J ¼ T em
a þ Ta ;
ext
(16) (21)
dt a
where (r, h, /) are the spherical coordinates attached to x,
the statement that the mechanical angular momentum while ðr0 ; h0 ; /0 Þ are those attached to x0 , and where
changes because of the combined action of the electromag-
netic and external torques. 2pc  
g‘ ðt; r; t0 ; r 0 Þ ¼ H D  jr  r0 j Hðr þ r 0  DÞP‘ ðnÞ;
rr0
(22)
III. FIELDS OF A SPINNING CHARGED SPHERE
We consider an infinitely thin, spherical shell of radius R with H(u) denoting the step function, P‘(n) the Legendre
polynomials, D :¼ cðt  t0 Þ, and
spinning on a central axis with a time-changing angular
velocity X(t). The shell possesses a tangential pressure that
r2 þ r02  D2
supports it against gravity, and we assume that its material is n :¼ : (23)
not polarized (either electrically or magnetically) by the 2rr 0
electromagnetic field. We align the z-direction with the rota- Because this decomposition does not seem to be widely
tion axis, and employ the usual spherical coordinates (r, h, known, but can be immensely useful, we provide a deriva-
/) in addition to the Cartesian coordinates (x, y, z). Each tion in the Appendix.
coordinate system comes with a basis of unit vectors. For the We insert Eq. (21) within the integral for Ay – iAx, write
Cartesian coordinates we have ð^ x ; ^y ; ^z Þ, and for the spherical d3 x0 ¼ r 02 dr 0 dX0 where dX0 :¼ sin h0 dh0 d/0 , perform the
^ /Þ.
coordinates we have ð^r ; h; ^
angular integration by exploiting the orthonormality of
The shell carries a charge q, and its charge density is given spherical harmonics, and evaluate the radial integral with the
by help of the delta function contained in j. We arrive at
q l0
q¼ dðr  RÞ: (17) A¼ ^
Cðt; rÞsin h /; (24)
4pR2 4p
The velocity of an element of charge on the shell is where
^ The current density is j ¼ qv, and the
v ¼ RXðtÞ sin h /. ð
shell’s magnetic moment is 3
Cðt; rÞ :¼ mðt0 Þg1 ðt; r; t0 ; RÞ dt0 : (25)
ð 4pR
1 1
m :¼ x  j d 3 x ¼ qR2 XðtÞ ^z : (18)
2 3 Inserting Eq. (22) with ‘ ¼ 1 and accounting for the step
functions, this function becomes
In the following, we shall express X(t) in terms of m(t), with
m :¼ jmj ¼ 13 qR2 X. ð tjrRj=c h i
3c 2
In the usual Lorenz gauge, the potentials created by the Cðt; r Þ ¼ mðt0 Þ r2 þ R2  c2 ðt  t0 Þ dt0 :
spinning shell satisfy wave equations, with solutions 4R3 r 2 tðrþRÞ=c
ð (26)
1
Uðt; xÞ ¼ qðt0 ; x0 ÞGðt; x; t0 ; x0 Þ dt0 d 3 x0 ; (19a)
4p0 In principle, the integral can be evaluated only once mðt0 Þ is
ð specified. As we shall see, however, we can avoid making an
l
Aa ðt; xÞ ¼ 0 ja ðt0 ; x0 ÞGðt; x; t0 ; x0 Þ dt0 d3 x0 ; (19b) explicit identification, and obtain C(t, r) for any mðt0 Þ, at the
4p
small price of a sensible assumption. We shall assume that
where Gðt; x; t0 ; x0 Þ is the retarded Green’s function for the tc, the characteristic time scale of variation of mðt0 Þ, is very
wave equation. Because q is static and spherically symmet- long compared with s: ¼ R/c, the light-travel time across
ric, the scalar potential is given simply by (half) the sphere. The assumption implies, for example, that
sm_  m, that s2 m €  sm,_ and so on, with overdots indicat-
1 q ing differentiation with respect to time.
Uin ¼ ; (20a) When r < R, the bounds of integration in Eq. (26) become
4p0 R
t – R/c – r/c and t – R/c þ r/c, respectively, and the integra-
1 q tion takes place over a short time interval 2r/c < 2s around
Uout ¼ ; (20b)
4p0 r t ¼ R/c. Over this interval mðt0 Þ can be expressed as a Taylor
expansion about t0 ¼ t  R=c, and the integral can be evalu-
with Uin applying when r < R, and Uout when r > R. ated. We obtain

842 Am. J. Phys., Vol. 86, No. 11, November 2018 Bonga, Poisson, and Yang 842
 
r 1 r 2  ð2Þ ð 3Þ
 The magnetic field B ¼ $  A ¼ Br ^r þ Bh h^ has the non-
Cin ðt; r Þ ¼ 3 ðm þ sm_ Þ þ m þ sm vanishing components
R 10 c2
4  
þ
1 r ð 4Þ
þ ð5Þ l0 2 1 r 2  ð 2Þ 
280 c4
m sm Bin
r ¼ 3
ðm þ sm_ Þ þ 2
m þ smð3Þ
 4p R 10 c
1 r 6  ð 6Þ ð7Þ

1 r 4 
þ m þ sm þ    ; (27) þ mð4Þ þ smð5Þ
15120 c6 tR=c 280 c 4

where a number within brackets in a superscript gives the 1 r6  ð6Þ ð 7Þ

þ m þ sm þ  cos h;
number of differentiations, and the subscript after the end 15120 c6 tR=c
square bracket indicates that m and its derivatives are evalu-
ated at the time t – R/c. We remark that m and its derivatives (31a)

could be further Taylor expanded about the time t, but this l0 2 1 r2  
form for Cin happens to be convenient to work with. Bin
h ¼ 3
ðm þ sm_ Þ þ 2 mð2Þ þ smð3Þ
4p R 5c
When r > R, the bounds of integration in Eq. (26) become
4 
t – r/c – R/c and t – r/c þ R/c, respectively, and the integra- 3 r
þ mð4Þ þ smð5Þ
tion again takes place over a short time interval 2R/c ¼ 2s. 280 c 4
Over this interval mðt0 Þ can again be expressed as a Taylor 
1 r6  ð6Þ ð 7Þ

expansion, this time about t0 ¼ t  r=c, and the integral can þ m þ sm þ  sin h
3780 c6 tR=c
again be evaluated. We obtain
    (31b)
1 r 1 2 ð2Þ r ð3Þ
Cout ðt;rÞ ¼ 2 m þ m_ þ s m þ m inside the sphere, and
r c 10 c
     
1 4 ð4Þ r ð5Þ l0 2 r 1 2 ð2Þ r ð3Þ
þ s m þ m out
Br ¼ m þ m_ þ s m þ m
280 c 4p r3 c 10 c
    
1 6 ð 6Þ r ð7Þ 1 4 ð4Þ r ð5Þ
þ s m þ m þ  ; (28) þ s m þ m
15120 c tr=c 280 c
  
1 6 ð6Þ r ð 7Þ
where m and its derivatives are now evaluated at the time þ s m þ m þ  cos h;
t – r/c. It is easy to verify that C is continuous at r ¼ R. 15120 c tr=c
Inserting Eqs. (27) and (28) within Eq. (24) gives us the vec- (32a)
tor potential of a rotating charged sphere.  
With the potentials in hand it is a straightforward exercise l0 1 r r2
to obtain the fields. The electric field can be decomposed Bout
h ¼ m þ m_ þ 2 m
€
4p r3 c c
into a Coulomb piece defined by EC ¼ $U and an induc-  
tion piece defined by EI ¼ @t A. These are given by 1 2 r r2
þ s mð2Þ þ mð3Þ þ 2 mð4Þ
10 c c
 
ECin ¼ 0; (29a) 1 4 ð4Þ r ð5Þ r2 ð6Þ
 þ s m þ m þ 2m
l0 r 1 r2  ð3Þ  280 c c
ð 4Þ   
EIin ¼  ð _
m þ sm€ Þ þ m þ sm 1 r ð7Þ r 2 ð8Þ
4p R3 10 c2 þ 6 ð6Þ
s m þ m þ 2m þ  sinh
4  15120 c c tr=c
1 r
þ 4
mð5Þ þ smð6Þ (32b)
280 c

1 r 6  ð 7Þ ð8Þ

^
þ m þ sm þ  sin h / outside the sphere.
15120 c6 tR=c
(29b)
IV. SELF-TORQUE
inside the sphere; we recall that s ¼ R/c. Outside the sphere
we have The torque exerted by the electromagnetic field on the
1 q spinning sphere—the self-torque—is given in Eq. (10), in
ECout ¼ ^r ; (30a) which we insert the Lorentz force density of Eq. (5). The cal-
4p0 r 2 culation requires some care because the fields are discontinu-
    ous at r ¼ R. It is easy to see, however, that the Coulomb
I l0 1 r 1 2 ð3Þ r ð4Þ
Eout ¼  m_ þ m € þ s m þ m piece of the electric field (which is discontinuous) does not
4p r2 c 10 c
  contribute to the torque. The induction piece does contribute,
1 4 r and this contribution is unambiguous because EI happens to
þ s mð5Þ þ mð6Þ
280 c be continuous at r ¼ R. The magnetic field is discontinuous,
  
1 ð 7Þ r ð 8Þ but calculation shows that the magnetic field makes no con-
þ 6
s m þ m þ  ^
sin h /:
15120 c tribution to the torque.
tr=c
We find after a straightforward calculation that the self-
(30b) torque’s only nonvanishing component is

843 Am. J. Phys., Vol. 86, No. 11, November 2018 Bonga, Poisson, and Yang 843
 
l0 q 1   Inspection of Eq. (34) reveals that the leading terms
T em
z ¼ € Þ þ s2 mð3Þ þ smð4Þ
ðm_ þ sm involving m_ and m(3) come with even powers of c and are
6pR 10
1 4  ð 5Þ  therefore time-symmetric; these contributions to the self-
þ s m þ smð6Þ torque have nothing to do with radiation reaction. The lead-
280  ing radiation-reaction term occurs at order s3 and is propor-
1  
þ s6 mð7Þ þ smð8Þ þ    : (33) tional to m(4).
15120 tR=c

An alternative expression, with m and its derivatives further V. RADIATED ANGULAR MOMENTUM
expanded about the time t, is
The radiated flux of angular momentum is calculated on

l0 q 2 1 the basis of Eq. (9). We choose S to be a sphere of radius r0,
Tz ¼ 
em
m_  s2 mð3Þ þ s3 mð4Þ and we take r0 to be very large compared with kc, a charac-
6pR 5 3
teristic wavelength of the electromagnetic radiation. With tc
6 4 ð5Þ 1 5 ð6Þ
 s m þ s m denoting the characteristic time scale of variation of the
35 15  magnetic moment (this quantity was introduced in Sec. III),
4 6 ð7Þ 1 7 ð8Þ we have that kc ¼ ctc, and we therefore assume that r0  ctc
 s m þ s m þ  : (34)
189 175 t  R.
After inserting Eq. (4) and writing dSa ¼ r^a r02 dX for the
The expression of Eq. (33) is useful for the purpose of estab- surface element (where dX :¼ sin h dhd/), we find that the
lishing angular momentum balance, as we do in Sec. VI. angular-momentum flux can be expressed as
Equation (34) is a better starting point for the calculation of ð
the angular motion of the spinning shell, which we present in N a ¼ 0 r03 ð^r  EÞð^r  EÞa dX
em
Sec. VII. It is also more convenient to single out the contri-
butions associated with radiation reaction, as we do in the ð
1
next paragraphs.  r 3 ð^r  BÞð^r  BÞa dX: (35)
l0 0
One would normally expect the self-force or self-torque
acting on a distribution of charge to be associated with the
We make the substitutions from Eqs. (30) and (32), neglect
electromagnetic radiation emitted by the distribution, and
terms that are suppressed by powers of kc/r0, perform the
one would attach the words “radiation reaction” to the phe-
angular integration, and obtain that the only nonvanishing
nomenon. Our result for the self-torque reveals, however,
that this expectation is not entirely fulfilled: the self-torque is component of the radiated flux is
not the sole result of radiation reaction. In fact, while radia- 
l0 q 1 1 4 ð 6Þ
tion reaction does contribute, it is not the dominant Nem
z ¼ € þ s2 mð4Þ þ
m s m
contribution. 6pc 10 280

The radiation-reaction terms in Eq. (34) can be 1
identified as those that come with an odd power of c (recall þ s 6 m ð 8Þ þ    ; (36)
15120 tr0 =c
that s ¼ R/c); the terms that come with even powers of c
have nothing to do with radiation reaction. To understand where m and its derivatives are evaluated at the time t – r0/c;
this, recall that the fields of Sec. III were calculated on the we recall that s ¼ R/c. The expression is proportional to q
basis of the retarded Green’s function. These fields describe and linear in the magnetic moment; this indicates that the
an outgoing electromagnetic wave that carries angular radiated angular momentum results from an interplay
momentum outward, with the radiated loss of angular between the Coulomb and induction pieces of the electric
momentum balanced by a contribution to the self-torque. field.
Suppose, however, that we had adopted the advanced solu- The radiated flux of Eq. (36) is associated with the
tion instead of the retarded solution. In this case, we would
retarded solution to Maxwell’s equation. As discussed in
have had an incoming wave that carries angular momentum
Sec. IV, the advanced solution can be obtained with the for-
inward, and the radiated gain of angular momentum would
mal switch c ! –c, and the advanced version of the radiated
have been balanced by an equal and opposite contribution to
flux would come with an overall minus sign, and would
the self-torque. The advanced solution can be obtained from
the retarded solution by the formal switch c ! –c, and the involve m and its derivatives evaluated at the time t þ r0/c.
contributions to the self-torque that are sensitive to the
choice of solution are therefore those that are odd in c; these VI. FIELD ANGULAR MOMENTUM
are the radiation-reaction terms. The terms that are even in c
are the same regardless of the choice of solution, and these The self-torque of Eq. (33) and the radiated angular
contributions to the self-torque—though they still have to do momentum of Eq. (36) come at the expense of the field’s
with the electromagnetic field created by the spinning own angular momentum, as implied by Eq. (8). The angular
sphere—have nothing to do with radiation reaction. momentum in the electromagnetic field can be calculated
Another way of stating all this is to say that the complete from Eq. (7), which we write in the form
self-torque is a sum of time-antisymmetric terms (odd in c) ð
that make up the radiation-reaction piece, and time- Ja ¼ 0 ½ðx  BÞEa  ðx  EÞBa  d3 x;
em
(37)
symmetric terms (even in c) that have nothing to do with V
radiation reaction. The notions of self-torque and radiation
reaction must therefore be distinguished. or

844 Am. J. Phys., Vol. 86, No. 11, November 2018 Bonga, Poisson, and Yang 844
 ð r0
VII. SHELL MOTION
Jaem ¼ ja dr (38)
0 The self-torque of Eq. (34) affects the angular motion of
with the shell. The motion is governed by Eq. (16), dJamech =dt
ð a þ T a , where Ja
¼ T ext em mech
is the shell’s mechanical angular
momentum, T a is the torque supplied by an external agent,
ext
ja :¼ 0 r 3 ½ð^r  BÞEa  ð^r  EÞBa  dX; (39)
and T em
a is the self-torque. In this section, we explore the
consequences of this equation.
we recall that r0 is the radius of the sphere S bounding V.
Substitution from Eqs. (29) and (31) and evaluation of the A. Equation of angular motion
angular integral reveals that ja ¼ 0 when r < R. For r > R we
get from Eqs. (30) and (32) that the only nonvanishing com- We take the shell to have a uniform mass density, and its
ponent of ja is mechanical angular momentum is related to the magnetic
"  moment by
l0 q m m_ m€
jz ¼ þ þ 2M
6p r2 cr c2 Jmech ¼ m; (43)
  q
1 2 mð2Þ mð3Þ mð4Þ
þ s þ þ 2
10  r 2 cr c  where M is the shell’s total mass. The shell’s equation of
1 4 m ð4Þ
m ð5Þ
mð6Þ motion is therefore
þ s 2
þ þ 2 
280 r cr c # 2M l q 2 1
  m_ ¼ T ext  0 m_  s2 mð3Þ þ s3 mð4Þ
ð 6Þ
1 6 m m ð7Þ
mð8Þ q 6pR 5 3
þ s þ þ 2 þ  :
15120 r2 cr c 6 4 ð 5Þ 1 5 ð 6Þ
tr=c  s m þ s m
(40) 35 15 
4 6 ð7Þ 1 7 ð 8Þ
 s m þ s m þ  ; (44)
This must be integrated with respect to r, and it is a fortunate 189 175
circumstance that each term in jz is a derivative with respect
to r. For example, where T ext is the z-component of the external torque (we
  assume that the remaining components vanish). We transfer
m m_ m€ @ m m_ the m_ term on the right-hand side of the equation to the other
þ þ 2¼ þ ; (41)
r2 cr c @r r c side, and write the combination as ð2Meff =qÞm, _ where
Meff :¼ M þ dM is an effective mass, with an electrostatic
where the partial derivative holds t fixed. All remaining correction given by
terms can be written in a similar way. The integration is per-
formed from r ¼ R to r ¼ r0, and after neglecting terms that q2
dMc2 ¼ : (45)
are suppressed by powers of kc/r0, we arrive at 12p0 R

l0 q 1  
This contribution to the effective mass can be thought of as
em
Jz ¼ ðm þ sm_ Þ þ s2 mð2Þ þ smð3Þ
6pR 10 additional inertia provided by the electromagnetic field.
1 4  ð4Þ  Curiously, dMc2 is a factor 2/3 smaller than the electrostatic
þ s m þ smð5Þ energy of a hollow sphere, given by U ¼ q2/(8p0R).
280 
  Introducing the notation
1 6 ð6Þ ð7Þ
þ s m þ sm þ 
15120 tR=c q 1 q2 2 dM
 T :¼ T ext ; e :¼ 2
¼ ; (46)
l q 1 1 4 ð 5Þ 2Meff Meff c 30p0 R 5 Meff
 0 m_ þ s2 mð3Þ þ sm
6pc 10 280
 we put the shell’s equation of motion in the form
1
þ s6 mð7Þ þ    : (42) 
15120 5 3 1
tr0 =c m_ ¼ T þ e s2 mð3Þ  s3 mð4Þ þ s4 mð5Þ  s5 mð6Þ
6 7 6
As indicated, the first set of terms is evaluated at the time 
t – R/c, while the second set is evaluated at t – r0/c. The first 10 6 ð7Þ 1 7 ð8Þ
þ s m  s m þ  ; (47)
set can be thought of as the contribution to the angular 189 70
momentum coming from the near-zone fields in the vicinity
of the spinning shell. The second set can be thought of as the where, we recall, s :¼ R/c. We note that up to a factor of
contribution from the wave-zone fields. order unity, e is the ratio of the electrostatic contribution to
Differentiation of Eq. (42) with respect to t and compari- the mass to the total effective mass. This ratio is small for a
son with Eqs. (33) and (36) reveals that the field angular macroscopic body. To take the measure of this, we consider
momentum satisfies dJaem =dt ¼ N em a  T a . This is Eq. (8),
em
a sphere of mass M ¼ 10 g and radius R ¼ 1 cm, on which we
the statement of angular momentum balance that was first deposit a charge q ¼ 0.1 C; for this object e ’ 1  106 . The
established in Sec. II. It is deeply satisfying that the state- assumption placed on the variation of m, s/tc  1, ensures
ment can be verified explicitly. that s2 mð3Þ  m;
_ s3 mð4Þ  s2 mð3Þ , and so on. The upshot is

845 Am. J. Phys., Vol. 86, No. 11, November 2018 Bonga, Poisson, and Yang 845
that the external term T is strongly dominant in Eq. (47), and
the self-torque terms provide small corrections. We recall
that in Eq. (47), the terms with odd powers of s are
radiation-reaction terms; the terms with even powers of s
(including the leading-order term proportional to m(3)) have
nothing to do with radiation reaction.
Equation (47) contains high-order derivatives, and it
would include terms of even higher order if the expansion in
powers of s/tc were pursued beyond what is displayed there.
Such equations are typically pathological and produce run-
away solutions that strongly violate the underlying assump-
tion that s/tc  1. A coping strategy presents itself on the
grounds that e is small, and that each successive term con-
tributes a smaller correction. The strategy is to perform a
reduction of order, by substituting m_ ¼ T þ OðeÞ on the
right-hand side of the equation. This yields
 cubic function of t. The leading term, growing as t3, is con-
5 3
m_ ¼ T þ e s2 Tð2Þ  s3 Tð3Þ þ s4 Tð4Þ
6 7

1 10 6 ð6Þ 1 torque. It is interesting to note that the self-torque term is
 s5 Tð5Þ þ s T  s7 Tð7Þ þ    ; (48)
6 189 70
where we neglect all terms of order e2 and higher.18
Equation (48) is free of pathologies, and its solution is
immediate
ðt  as shown in Eq. (36). The apparent paradox is resolved by
5 3
mðtÞ ¼ Tðt Þ dt þ e s2 Tð1Þ  s3 Tð2Þ þ s4 Tð3Þ
0 0
6 7
1 5 ð 4Þ 10 6 ð5Þ
 s T þ s T
6 189 
1 T ¼ T4 :¼ mk4 t3 ;
 s7 Tð6Þ þ    þ constant; (49)
70
where the constant is determined by the initial condition.
With T assumed to vary on a time scale tc that is long com-
pared with s ¼ R/c, Eq. (49) is guaranteed to describe a m(t)
that also varies on a long time scale.
B. Examples
For the purpose of illustration we examine a number of
concrete examples.
To begin we consider the simplest case, that of a vanishing
external torque. Setting T ¼ T1 :¼ 0 in Eq. (49) returns
m1(t) ¼ constant, and we see that as expected, the sphere
rotates with a constant angular velocity when the external
torque is zero. Equation (34) implies that the self-torque van-
ishes, Eq. (36) shows that there is no radiated flux of angular
momentum, and Eq. (42) reveals that the field’s angular
momentum is J em ¼ l0 qm1 =ð6pRÞ, a constant. In this rather
trivial case, angular momentum is strictly conserved.
Next on our list of examples is that of a constant external
torque, T ¼ T2 :¼ constant. In this case Eq. (49) returns
m2 ðtÞ ¼ m2 ð0Þ þ T2 t, and we find that the angular velocity
increases linearly with time. Equation (34) implies that the
self-torque is given by
l0 q dM ext 5 10 6
T em ¼  T2 ¼  T ¼  e T ext ; (50)
6pR Meff 2 189
where we used Eqs. (45) and (46). This self-torque was
already moved to the left-hand side of the equation of motion
in Eq. (47), and shown to give rise to the shift M
846 Am. J. Phys., Vol. 86, No. 11, November 2018
! M þ dM ¼ Meff in the sphere’s mass. The previous equa-
tion reveals another interpretation for the self-torque (valid
in this specific case only): it produces a slight offset in the
external torque. Equation (36) shows that there is no radiated
flux of angular momentum.
Next we consider an external torque given by
T ¼ T3 :¼ mk3 t2 ; (51)
where m is a constant with unit of magnetic moment, and
k is a constant with unit of inverse time. The solution to Eq.
(49) is
1 3
m3 ðtÞ ¼ m3 ð0Þ þ mðktÞ þ 2emf2 ðktÞ; (52)
3
where f: ¼ ks  1. In this case, the magnetic moment is a
tributed by the external torque, and the subleading term, pro-
portional to e and growing as t, is contributed by the self-
proportional to f2 / s2 / c–2 and therefore even in c.
Recalling the discussion of Sec. IV, we conclude that the
self-torque correction has nothing to do with radiation reac-
tion. This is in spite of the fact that a magnetic moment that
grows approximately as t3 does radiate angular momentum,
noting that the radiated losses come at the expense of the
field’s angular momentum, and need not be balanced by a
radiation-reaction contribution to the self-torque.
As a slight variation of the preceding example, we take
(53)
where m and k are again constants. The corresponding solu-
tion to Eq. (49) is
h i
1 4 2
m4 ðtÞ ¼ m4 ð0Þ þ mðktÞ þ em 3f2 ðktÞ  5f3 ðktÞ ;
4
(54)
where f: ¼ ks  1. The leading term that grows as t4 is con-
tributed by the external torque, and the remaining terms are
contributed by the self-torque. In this case, we see that the
self-torque correction contains a term proportional to f2,
which is again even in c; this term has nothing to do with
radiation reaction. But there is also a term proportional to f3,
which is odd in c; this is a radiation-reaction contribution.
Next we consider an oscillating external torque given by
T ¼ T5 :¼ mx cos xt; (55)
where m is a constant with unit of magnetic moment, and x
is a constant angular frequency. The corresponding solution
to the equation of motion is
 
3
m5 ðtÞ ¼ m5 ð0Þ þ m 1  e f2  f4
7

þ f þ    sin xt
 
5 3 1 5 1
þ em f  f þ f7 þ    ðcos xt  1Þ;
6 6 70
(56)
Bonga, Poisson, and Yang 846
 h i
where f :¼ xs  1. In the first set of terms, proportional to
Gðt; x; t0 ; x0 Þ ¼ 2cHðt  t0 Þ d c2 ðt  t0 Þ2  jx  x0 j2 ;
sin xt, the self-torque produces an amplitude correction that
is even in f; this correction is not to be associated with radia- (A2)
tion reaction. The second set of terms is also contributed by
the self-torque, and it represents a correction to the phase of with H(u) denoting the step function.
the magnetic moment; these terms are odd in f, and they are We wish to express the Green’s function as a spherical-
directly associated with radiation reaction. harmonic decomposition. We write
For a final example, we consider an external torque that
X
gradually switches on and then gradually switches off. We Gðt; x; t0 ; x0 Þ ¼ g‘ ðt; r; t0 ; r0 ÞY‘m

ðh0 ; /0 ÞY‘m ðh; /Þ;
describe it in terms of a Gaussian function ‘m

T ¼ T6 :¼ mkek t ;
2 2
(57) (A3)

so that the external torque is active during a time interval where g‘ ðt; r; t0 ; r0 Þ is a reduced Green’s function for each
proportional to k–1 around t ¼ 0. The corresponding solu- multipole order ‘; the label m does not appear because the
tion is reduced wave equation satisfied by g‘ is independent of m.
The strategy to obtain g‘ is to subject Eq. (A2) to a projection
1 pffiffiffi to each one of its multipole components.
m6 ðtÞ ¼ m6 ð1Þ þ pm½erf ðktÞ þ 1 We introduce the notation D :¼ cðt  t0 Þ; s :¼ jx  x0 j,
2
h i and express Eq. (A2) in the condensed form
2 5 3 2
ð Þ
 em 2f kt þ f 2 kt  1 ð Þ G¼ 2cHðDÞdðD2  s2 Þ. We substitute this on the left-hand
3
h i side of Eq. (A3), multiply each side by Y‘0 m0 ðh0 ; /0 Þ, and inte-
12 3 grate over dX0 ¼ sin h0 dh0 d/0 . The result is
þ f4 2ðktÞ  3ðktÞ
7 h i
2 5 4 2
ð
ð Þ
þ f 4 kt  12 kt þ 3 ð Þ 2cHðDÞ dðD2  s2 ÞY‘m ðh0 ; /0 Þ dX0 ¼ g‘ Y‘m ðh; /Þ:
3 h i
80 6 5 3
þ ð Þ ð
f 4 kt  20 kt þ 15 kt Þ ð Þ (A4)
189 h i
4 6 4 2 We next set m ¼ 0 and use the fact that Y‘0 ðh; /Þ
þ f7 8ðktÞ  60ðktÞ þ 90ðktÞ  15
35 / P‘ ðcos hÞ. The previous equation reduces to
2 2
þ    ek t ; (58) ð
2cHðDÞ dðD2  s2 ÞP‘ ðcos h0 Þ d cos h0 d/0
v
Ð
1=2 u 2
where erfðuÞ :¼ 2p 0 e dv is the error function. The
leading term in Eq. (58) describes a gradual increase of m6 ¼ g‘ P‘ ðcos hÞ: (A5)
value m6(–1) to its final value m6 ðþ1Þ
from its initialpffiffiffi
¼ m6 ð1Þ þ pm. The corrections contributed by the self- Finally, we set cos h ¼ 1 and use the fact that P‘(1) ¼ 1. This
torque are slight modulations superposed to the leading gives
behavior described by the error function. ð
g‘ ¼ 2cHðDÞ dðD2  s2 Þj cos h¼1 P‘ ðcos h0 Þ d cos h0 d/0 :
ACKNOWLEDGMENTS
(A6)
Conversations with Bernie Nickel were greatly appreciated.
This work was supported by the Natural Sciences and
Engineering Research Council of Canada and by Perimeter Because s2 evaluated at cos h ¼ 1 is r2  2rr0 cos h0 þ r 02 , we
Institute for Theoretical Physics. Research at Perimeter have that
Institute is supported by the Government of Canada through ð
the Department of Innovation, Science and Economic g‘ ¼ 4pcHðDÞ dðD2  r2 þ 2rr 0 cos h0  r 02 Þ
Development Canada and by the Province of Ontario through
the Ministry of Research, Innovation and Science.  P‘ ðcos h0 Þ d cos h0 ; (A7)

or
ð
APPENDIX: SPHERICAL-HARMONIC 2pcHðDÞ
g‘ ¼ dðcos h0  nÞP‘ ðcos h0 Þ d cos h0 ; (A8)
DECOMPOSITION OF THE RETARDED GREEN’S rr0
FUNCTION
The retarded Green’s function for the wave equation is where n :¼ ðr 2 þ r02  D2 Þ=ð2rr0 Þ.
given by The integral is nonzero whenever n lies in the interval
between 1 and þ1; when this condition is satisfied it evalu-
dðt  t0  jx  x0 j=cÞ ates to g‘ ¼ 2pcHðDÞP‘ ðnÞ=ðrr0 Þ. The condition 1 < n
Gðt; x; t0 ; x0 Þ ¼ : (A1) implies 2rr0 < r2 þ r02  D2 , so that D < r þ r0 . The con-
jx  x0 j
dition n < 1 implies 2rr0 > r2 þ r 02  D2 , so that D > jr
Alternatively, it can be expressed as  r 0 j. This last condition supersedes the requirement that

847 Am. J. Phys., Vol. 86, No. 11, November 2018 Bonga, Poisson, and Yang 847
 10
D > 0, which comes from the step function appearing in G.
Altogether, we find that the reduced Green’s function is
11
given by
2pc   12
g‘ ðt; r; t0 ; r0 Þ ¼ H D  jr  r0 j Hðr þ r0  DÞP‘ ðnÞ:
rr0
13
(A9)
14
The temporal support of the reduced Green’s function is the
interval jr  r0 j < D < r þ r0 . 15
1
D. J. Griffiths, Introduction to Electrodynamics, 4th ed. (Pearson, Boston,
2013). 16
2
This variant of Feynman’s disk paradox was first proposed by Romer (Ref.
19) and Boos (Ref. 20); the original paradox is discussed in Sec. 17.4 of 17
Volume II of the Feynman lectures (Ref. 21), and in Ref. 22.
3
J. D. Jackson, Classical Electrodynamics, 3rd ed. (Wiley, New York,
18
1999).
4
T. H. Boyer, “Mass renormalization and radiation damping for a charged
particle in uniform circular motion,” Am. J. Phys. 40, 1843–1846
(1972).
5 19
F. Rohrlich, “The dynamics of a charged sphere and the electron,” Am. J.
Phys. 65, 1051–1056 (1997).
6 20
D. J. Griffiths, T. C. Proctor, and D. F. Schroeter, “Abraham–Lorentz ver-
sus Landau–Lifshitz,” Am. J. Phys. 78, 391–402 (2010).
7 21
L. D. Landau and E. M. Lifshitz, The Classical Theory of Fields, 4th ed.
(Butterworth-Heinemann, Oxford, England, 2000).
8 22
D. J. Griffiths and R. E. Owen, “Mass renormalization in classical electro-
dynamics,” Am. J. Phys. 51, 1120–1126 (1983).
9 23
D. J. Griffiths and E. W. Szeto, “Dumbbell model for the classical radia-
tion reaction,” Am. J. Phys. 46, 244–248 (1978).
848 Am. J. Phys., Vol. 86, No. 11, November 2018
A. Ori and E. Rosenthal, “Universal self force from an extended-object
approach,” Phys. Rev. D 68, 041701 (2003); e-print arXiv:gr-qc/0205003.
A. Ori and E. Rosenthal, “Calculation of the self force using the extended-
object approach,” J. Math. Phys. 45, 2347–2364 (2004); e-print arXiv:gr-qc/
0309102.
S. E. Gralla, A. I. Harte, and R. M. Wald, “Rigorous derivation of electromag-
netic self-force,” Phys. Rev. D 80, 024031 (2009); e-print arXiv:0905.2391.
The electromagnetic angular momentum of a spinning shell with constant
angular velocity was previously calculated by de Castro (Ref. 23).
D. R. Stump and G. L. Pollack, “Magnetic dipole oscillations and radiation
damping,” Am. J. Phys. 65, 81–87 (1997).
A. Ashtekar and B. Bonga, “On the ambiguity in the notion of transverse
traceless modes of gravitational waves,” Gen. Rel. Grav. 49, 122 (2017);
e-print arXiv:1707.09914.
R. J. Cook, “Radiation reaction revisited,” Am. J. Phys. 52, 894–895
(1984).
The stress tensor is sometimes defined with plus signs, with the alternative
em
convention that Tab dSb is the momentum crossing into the element of sur-
face per unit time.
The reduction of order works even when e is taken to be of order unity. In
this case, the procedure relies entirely on the assumption that s/tc  1, and
it produces a very similar form for the reduced equation of motion, with
coefficients that acquire e-dependent corrections.
R. H. Romer, “Angular momentum of static electromagnetic fields,” Am.
J. Phys. 34, 772–778 (1966).
F. L. Boos, Jr., “More on the Feynman’s disk paradox,” Am. J. Phys. 52,
756–757 (1984).
R. P. Feynman, R. B. Leighton, and M. Sands, The Feynman Lectures on
Physics, The New Millennium Edition (Basic Books, USA, 2011).
G. G. Lombardi, “Feynman’s disk paradox,” Am. J. Phys. 51, 213–214
(1983).
A. S. de Castro, “Electromagnetic angular momentum for a rotating
charged shell,” Am. J. Phys. 59, 180–181 (1991).
Bonga, Poisson, and Yang 848